Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 LT |
2 | /* |
3 | * linux/fs/proc/base.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | * | |
7 | * proc base directory handling functions | |
8 | * | |
9 | * 1999, Al Viro. Rewritten. Now it covers the whole per-process part. | |
10 | * Instead of using magical inumbers to determine the kind of object | |
11 | * we allocate and fill in-core inodes upon lookup. They don't even | |
12 | * go into icache. We cache the reference to task_struct upon lookup too. | |
13 | * Eventually it should become a filesystem in its own. We don't use the | |
14 | * rest of procfs anymore. | |
e070ad49 ML |
15 | * |
16 | * | |
17 | * Changelog: | |
18 | * 17-Jan-2005 | |
19 | * Allan Bezerra | |
20 | * Bruna Moreira <bruna.moreira@indt.org.br> | |
21 | * Edjard Mota <edjard.mota@indt.org.br> | |
22 | * Ilias Biris <ilias.biris@indt.org.br> | |
23 | * Mauricio Lin <mauricio.lin@indt.org.br> | |
24 | * | |
25 | * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT | |
26 | * | |
27 | * A new process specific entry (smaps) included in /proc. It shows the | |
28 | * size of rss for each memory area. The maps entry lacks information | |
29 | * about physical memory size (rss) for each mapped file, i.e., | |
30 | * rss information for executables and library files. | |
31 | * This additional information is useful for any tools that need to know | |
32 | * about physical memory consumption for a process specific library. | |
33 | * | |
34 | * Changelog: | |
35 | * 21-Feb-2005 | |
36 | * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT | |
37 | * Pud inclusion in the page table walking. | |
38 | * | |
39 | * ChangeLog: | |
40 | * 10-Mar-2005 | |
41 | * 10LE Instituto Nokia de Tecnologia - INdT: | |
42 | * A better way to walks through the page table as suggested by Hugh Dickins. | |
43 | * | |
44 | * Simo Piiroinen <simo.piiroinen@nokia.com>: | |
45 | * Smaps information related to shared, private, clean and dirty pages. | |
46 | * | |
47 | * Paul Mundt <paul.mundt@nokia.com>: | |
48 | * Overall revision about smaps. | |
1da177e4 LT |
49 | */ |
50 | ||
7c0f6ba6 | 51 | #include <linux/uaccess.h> |
1da177e4 | 52 | |
1da177e4 LT |
53 | #include <linux/errno.h> |
54 | #include <linux/time.h> | |
55 | #include <linux/proc_fs.h> | |
56 | #include <linux/stat.h> | |
5995477a | 57 | #include <linux/task_io_accounting_ops.h> |
1da177e4 | 58 | #include <linux/init.h> |
16f7e0fe | 59 | #include <linux/capability.h> |
1da177e4 | 60 | #include <linux/file.h> |
9f3acc31 | 61 | #include <linux/fdtable.h> |
94f8f3b0 | 62 | #include <linux/generic-radix-tree.h> |
1da177e4 LT |
63 | #include <linux/string.h> |
64 | #include <linux/seq_file.h> | |
65 | #include <linux/namei.h> | |
6b3286ed | 66 | #include <linux/mnt_namespace.h> |
1da177e4 | 67 | #include <linux/mm.h> |
a63d83f4 | 68 | #include <linux/swap.h> |
b835996f | 69 | #include <linux/rcupdate.h> |
54354c6a | 70 | #include <linux/kallsyms.h> |
2ec220e2 | 71 | #include <linux/stacktrace.h> |
d85f50d5 | 72 | #include <linux/resource.h> |
5096add8 | 73 | #include <linux/module.h> |
1da177e4 LT |
74 | #include <linux/mount.h> |
75 | #include <linux/security.h> | |
76 | #include <linux/ptrace.h> | |
87ebdc00 | 77 | #include <linux/printk.h> |
efb1a57d | 78 | #include <linux/cache.h> |
a424316c | 79 | #include <linux/cgroup.h> |
1da177e4 LT |
80 | #include <linux/cpuset.h> |
81 | #include <linux/audit.h> | |
5addc5dd | 82 | #include <linux/poll.h> |
1651e14e | 83 | #include <linux/nsproxy.h> |
8ac773b4 | 84 | #include <linux/oom.h> |
3cb4a0bb | 85 | #include <linux/elf.h> |
60347f67 | 86 | #include <linux/pid_namespace.h> |
22d917d8 | 87 | #include <linux/user_namespace.h> |
5ad4e53b | 88 | #include <linux/fs_struct.h> |
5a0e3ad6 | 89 | #include <linux/slab.h> |
4eb5aaa3 | 90 | #include <linux/sched/autogroup.h> |
6e84f315 | 91 | #include <linux/sched/mm.h> |
f7ccbae4 | 92 | #include <linux/sched/coredump.h> |
b17b0153 | 93 | #include <linux/sched/debug.h> |
3905f9ad | 94 | #include <linux/sched/stat.h> |
48f6a7a5 | 95 | #include <linux/posix-timers.h> |
04a8682a | 96 | #include <linux/time_namespace.h> |
e79f15a4 | 97 | #include <linux/resctrl.h> |
c2f273eb | 98 | #include <linux/cn_proc.h> |
d21077fb | 99 | #include <linux/ksm.h> |
43d2b113 | 100 | #include <trace/events/oom.h> |
1da177e4 | 101 | #include "internal.h" |
faf60af1 | 102 | #include "fd.h" |
1da177e4 | 103 | |
ac7f1061 AD |
104 | #include "../../lib/kstrtox.h" |
105 | ||
0f2fe20f EB |
106 | /* NOTE: |
107 | * Implementing inode permission operations in /proc is almost | |
108 | * certainly an error. Permission checks need to happen during | |
109 | * each system call not at open time. The reason is that most of | |
110 | * what we wish to check for permissions in /proc varies at runtime. | |
111 | * | |
112 | * The classic example of a problem is opening file descriptors | |
113 | * in /proc for a task before it execs a suid executable. | |
114 | */ | |
115 | ||
efb1a57d AD |
116 | static u8 nlink_tid __ro_after_init; |
117 | static u8 nlink_tgid __ro_after_init; | |
1270dd8d | 118 | |
1da177e4 | 119 | struct pid_entry { |
cedbccab | 120 | const char *name; |
623f594e | 121 | unsigned int len; |
d161a13f | 122 | umode_t mode; |
c5ef1c42 | 123 | const struct inode_operations *iop; |
00977a59 | 124 | const struct file_operations *fop; |
20cdc894 | 125 | union proc_op op; |
1da177e4 LT |
126 | }; |
127 | ||
61a28784 | 128 | #define NOD(NAME, MODE, IOP, FOP, OP) { \ |
20cdc894 | 129 | .name = (NAME), \ |
c5141e6d | 130 | .len = sizeof(NAME) - 1, \ |
20cdc894 EB |
131 | .mode = MODE, \ |
132 | .iop = IOP, \ | |
133 | .fop = FOP, \ | |
134 | .op = OP, \ | |
135 | } | |
136 | ||
631f9c18 AD |
137 | #define DIR(NAME, MODE, iops, fops) \ |
138 | NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} ) | |
139 | #define LNK(NAME, get_link) \ | |
61a28784 | 140 | NOD(NAME, (S_IFLNK|S_IRWXUGO), \ |
20cdc894 | 141 | &proc_pid_link_inode_operations, NULL, \ |
631f9c18 AD |
142 | { .proc_get_link = get_link } ) |
143 | #define REG(NAME, MODE, fops) \ | |
144 | NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {}) | |
631f9c18 | 145 | #define ONE(NAME, MODE, show) \ |
6d9c939d | 146 | NOD(NAME, (S_IFREG|(MODE)), \ |
be614086 | 147 | NULL, &proc_single_file_operations, \ |
631f9c18 | 148 | { .proc_show = show } ) |
6d9c939d CS |
149 | #define ATTR(LSM, NAME, MODE) \ |
150 | NOD(NAME, (S_IFREG|(MODE)), \ | |
151 | NULL, &proc_pid_attr_operations, \ | |
152 | { .lsm = LSM }) | |
1da177e4 | 153 | |
aed54175 VN |
154 | /* |
155 | * Count the number of hardlinks for the pid_entry table, excluding the . | |
156 | * and .. links. | |
157 | */ | |
1270dd8d | 158 | static unsigned int __init pid_entry_nlink(const struct pid_entry *entries, |
aed54175 VN |
159 | unsigned int n) |
160 | { | |
161 | unsigned int i; | |
162 | unsigned int count; | |
163 | ||
1270dd8d | 164 | count = 2; |
aed54175 VN |
165 | for (i = 0; i < n; ++i) { |
166 | if (S_ISDIR(entries[i].mode)) | |
167 | ++count; | |
168 | } | |
169 | ||
170 | return count; | |
171 | } | |
172 | ||
f7ad3c6b | 173 | static int get_task_root(struct task_struct *task, struct path *root) |
1da177e4 | 174 | { |
7c2c7d99 HD |
175 | int result = -ENOENT; |
176 | ||
0494f6ec | 177 | task_lock(task); |
f7ad3c6b MS |
178 | if (task->fs) { |
179 | get_fs_root(task->fs, root); | |
7c2c7d99 HD |
180 | result = 0; |
181 | } | |
0494f6ec | 182 | task_unlock(task); |
7c2c7d99 | 183 | return result; |
0494f6ec MS |
184 | } |
185 | ||
7773fbc5 | 186 | static int proc_cwd_link(struct dentry *dentry, struct path *path) |
0494f6ec | 187 | { |
2b0143b5 | 188 | struct task_struct *task = get_proc_task(d_inode(dentry)); |
0494f6ec | 189 | int result = -ENOENT; |
99f89551 EB |
190 | |
191 | if (task) { | |
f7ad3c6b MS |
192 | task_lock(task); |
193 | if (task->fs) { | |
194 | get_fs_pwd(task->fs, path); | |
195 | result = 0; | |
196 | } | |
197 | task_unlock(task); | |
99f89551 EB |
198 | put_task_struct(task); |
199 | } | |
1da177e4 LT |
200 | return result; |
201 | } | |
202 | ||
7773fbc5 | 203 | static int proc_root_link(struct dentry *dentry, struct path *path) |
1da177e4 | 204 | { |
2b0143b5 | 205 | struct task_struct *task = get_proc_task(d_inode(dentry)); |
1da177e4 | 206 | int result = -ENOENT; |
99f89551 EB |
207 | |
208 | if (task) { | |
f7ad3c6b | 209 | result = get_task_root(task, path); |
99f89551 EB |
210 | put_task_struct(task); |
211 | } | |
1da177e4 LT |
212 | return result; |
213 | } | |
214 | ||
d26d0cd9 LT |
215 | /* |
216 | * If the user used setproctitle(), we just get the string from | |
217 | * user space at arg_start, and limit it to a maximum of one page. | |
218 | */ | |
219 | static ssize_t get_mm_proctitle(struct mm_struct *mm, char __user *buf, | |
220 | size_t count, unsigned long pos, | |
221 | unsigned long arg_start) | |
222 | { | |
223 | char *page; | |
224 | int ret, got; | |
225 | ||
226 | if (pos >= PAGE_SIZE) | |
227 | return 0; | |
228 | ||
229 | page = (char *)__get_free_page(GFP_KERNEL); | |
230 | if (!page) | |
231 | return -ENOMEM; | |
232 | ||
233 | ret = 0; | |
234 | got = access_remote_vm(mm, arg_start, page, PAGE_SIZE, FOLL_ANON); | |
235 | if (got > 0) { | |
236 | int len = strnlen(page, got); | |
237 | ||
238 | /* Include the NUL character if it was found */ | |
239 | if (len < got) | |
240 | len++; | |
241 | ||
242 | if (len > pos) { | |
243 | len -= pos; | |
244 | if (len > count) | |
245 | len = count; | |
246 | len -= copy_to_user(buf, page+pos, len); | |
247 | if (!len) | |
248 | len = -EFAULT; | |
249 | ret = len; | |
250 | } | |
251 | } | |
252 | free_page((unsigned long)page); | |
253 | return ret; | |
254 | } | |
255 | ||
e4b4e441 | 256 | static ssize_t get_mm_cmdline(struct mm_struct *mm, char __user *buf, |
5ab82718 | 257 | size_t count, loff_t *ppos) |
1da177e4 | 258 | { |
c2c0bb44 | 259 | unsigned long arg_start, arg_end, env_start, env_end; |
5ab82718 | 260 | unsigned long pos, len; |
d26d0cd9 | 261 | char *page, c; |
c2c0bb44 | 262 | |
c2c0bb44 | 263 | /* Check if process spawned far enough to have cmdline. */ |
e4b4e441 LT |
264 | if (!mm->env_end) |
265 | return 0; | |
c2c0bb44 | 266 | |
88aa7cc6 | 267 | spin_lock(&mm->arg_lock); |
c2c0bb44 AD |
268 | arg_start = mm->arg_start; |
269 | arg_end = mm->arg_end; | |
270 | env_start = mm->env_start; | |
271 | env_end = mm->env_end; | |
88aa7cc6 | 272 | spin_unlock(&mm->arg_lock); |
c2c0bb44 | 273 | |
5ab82718 LT |
274 | if (arg_start >= arg_end) |
275 | return 0; | |
6a6cbe75 | 276 | |
2ca66ff7 | 277 | /* |
d26d0cd9 LT |
278 | * We allow setproctitle() to overwrite the argument |
279 | * strings, and overflow past the original end. But | |
280 | * only when it overflows into the environment area. | |
2ca66ff7 | 281 | */ |
d26d0cd9 | 282 | if (env_start != arg_end || env_end < env_start) |
5ab82718 | 283 | env_start = env_end = arg_end; |
d26d0cd9 | 284 | len = env_end - arg_start; |
f5b65348 | 285 | |
5ab82718 | 286 | /* We're not going to care if "*ppos" has high bits set */ |
d26d0cd9 LT |
287 | pos = *ppos; |
288 | if (pos >= len) | |
5ab82718 | 289 | return 0; |
d26d0cd9 LT |
290 | if (count > len - pos) |
291 | count = len - pos; | |
292 | if (!count) | |
293 | return 0; | |
294 | ||
295 | /* | |
296 | * Magical special case: if the argv[] end byte is not | |
297 | * zero, the user has overwritten it with setproctitle(3). | |
298 | * | |
299 | * Possible future enhancement: do this only once when | |
300 | * pos is 0, and set a flag in the 'struct file'. | |
301 | */ | |
302 | if (access_remote_vm(mm, arg_end-1, &c, 1, FOLL_ANON) == 1 && c) | |
303 | return get_mm_proctitle(mm, buf, count, pos, arg_start); | |
a0a07b87 | 304 | |
d26d0cd9 LT |
305 | /* |
306 | * For the non-setproctitle() case we limit things strictly | |
307 | * to the [arg_start, arg_end[ range. | |
308 | */ | |
309 | pos += arg_start; | |
3d712546 | 310 | if (pos < arg_start || pos >= arg_end) |
5ab82718 | 311 | return 0; |
3d712546 LT |
312 | if (count > arg_end - pos) |
313 | count = arg_end - pos; | |
5ab82718 LT |
314 | |
315 | page = (char *)__get_free_page(GFP_KERNEL); | |
316 | if (!page) | |
317 | return -ENOMEM; | |
318 | ||
319 | len = 0; | |
320 | while (count) { | |
321 | int got; | |
322 | size_t size = min_t(size_t, PAGE_SIZE, count); | |
323 | ||
3d712546 LT |
324 | got = access_remote_vm(mm, pos, page, size, FOLL_ANON); |
325 | if (got <= 0) | |
5ab82718 | 326 | break; |
3d712546 | 327 | got -= copy_to_user(buf, page, got); |
5ab82718 LT |
328 | if (unlikely(!got)) { |
329 | if (!len) | |
330 | len = -EFAULT; | |
331 | break; | |
c2c0bb44 | 332 | } |
5ab82718 LT |
333 | pos += got; |
334 | buf += got; | |
335 | len += got; | |
336 | count -= got; | |
c2c0bb44 AD |
337 | } |
338 | ||
c2c0bb44 | 339 | free_page((unsigned long)page); |
5ab82718 | 340 | return len; |
1da177e4 LT |
341 | } |
342 | ||
e4b4e441 LT |
343 | static ssize_t get_task_cmdline(struct task_struct *tsk, char __user *buf, |
344 | size_t count, loff_t *pos) | |
345 | { | |
346 | struct mm_struct *mm; | |
347 | ssize_t ret; | |
348 | ||
349 | mm = get_task_mm(tsk); | |
350 | if (!mm) | |
351 | return 0; | |
352 | ||
353 | ret = get_mm_cmdline(mm, buf, count, pos); | |
c2c0bb44 | 354 | mmput(mm); |
e4b4e441 LT |
355 | return ret; |
356 | } | |
357 | ||
358 | static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, | |
359 | size_t count, loff_t *pos) | |
360 | { | |
361 | struct task_struct *tsk; | |
362 | ssize_t ret; | |
363 | ||
364 | BUG_ON(*pos < 0); | |
365 | ||
366 | tsk = get_proc_task(file_inode(file)); | |
367 | if (!tsk) | |
368 | return -ESRCH; | |
369 | ret = get_task_cmdline(tsk, buf, count, pos); | |
370 | put_task_struct(tsk); | |
371 | if (ret > 0) | |
372 | *pos += ret; | |
373 | return ret; | |
1da177e4 LT |
374 | } |
375 | ||
c2c0bb44 AD |
376 | static const struct file_operations proc_pid_cmdline_ops = { |
377 | .read = proc_pid_cmdline_read, | |
378 | .llseek = generic_file_llseek, | |
379 | }; | |
380 | ||
1da177e4 LT |
381 | #ifdef CONFIG_KALLSYMS |
382 | /* | |
383 | * Provides a wchan file via kallsyms in a proper one-value-per-file format. | |
384 | * Returns the resolved symbol. If that fails, simply return the address. | |
385 | */ | |
edfcd606 AD |
386 | static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns, |
387 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 388 | { |
ffb45122 | 389 | unsigned long wchan; |
54354c6a | 390 | char symname[KSYM_NAME_LEN]; |
1da177e4 | 391 | |
54354c6a KC |
392 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
393 | goto print0; | |
1da177e4 | 394 | |
54354c6a KC |
395 | wchan = get_wchan(task); |
396 | if (wchan && !lookup_symbol_name(wchan, symname)) { | |
397 | seq_puts(m, symname); | |
398 | return 0; | |
399 | } | |
25ce3191 | 400 | |
54354c6a KC |
401 | print0: |
402 | seq_putc(m, '0'); | |
25ce3191 | 403 | return 0; |
1da177e4 LT |
404 | } |
405 | #endif /* CONFIG_KALLSYMS */ | |
406 | ||
a9712bc1 AV |
407 | static int lock_trace(struct task_struct *task) |
408 | { | |
f7cfd871 | 409 | int err = down_read_killable(&task->signal->exec_update_lock); |
a9712bc1 AV |
410 | if (err) |
411 | return err; | |
caaee623 | 412 | if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) { |
f7cfd871 | 413 | up_read(&task->signal->exec_update_lock); |
a9712bc1 AV |
414 | return -EPERM; |
415 | } | |
416 | return 0; | |
417 | } | |
418 | ||
419 | static void unlock_trace(struct task_struct *task) | |
420 | { | |
f7cfd871 | 421 | up_read(&task->signal->exec_update_lock); |
a9712bc1 AV |
422 | } |
423 | ||
2ec220e2 KC |
424 | #ifdef CONFIG_STACKTRACE |
425 | ||
426 | #define MAX_STACK_TRACE_DEPTH 64 | |
427 | ||
428 | static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns, | |
429 | struct pid *pid, struct task_struct *task) | |
430 | { | |
2ec220e2 | 431 | unsigned long *entries; |
a9712bc1 | 432 | int err; |
2ec220e2 | 433 | |
f8a00cef JH |
434 | /* |
435 | * The ability to racily run the kernel stack unwinder on a running task | |
436 | * and then observe the unwinder output is scary; while it is useful for | |
437 | * debugging kernel issues, it can also allow an attacker to leak kernel | |
438 | * stack contents. | |
439 | * Doing this in a manner that is at least safe from races would require | |
440 | * some work to ensure that the remote task can not be scheduled; and | |
441 | * even then, this would still expose the unwinder as local attack | |
442 | * surface. | |
443 | * Therefore, this interface is restricted to root. | |
444 | */ | |
445 | if (!file_ns_capable(m->file, &init_user_ns, CAP_SYS_ADMIN)) | |
446 | return -EACCES; | |
447 | ||
6da2ec56 KC |
448 | entries = kmalloc_array(MAX_STACK_TRACE_DEPTH, sizeof(*entries), |
449 | GFP_KERNEL); | |
2ec220e2 KC |
450 | if (!entries) |
451 | return -ENOMEM; | |
452 | ||
a9712bc1 AV |
453 | err = lock_trace(task); |
454 | if (!err) { | |
e988e5ec | 455 | unsigned int i, nr_entries; |
5d008fb4 | 456 | |
e988e5ec TG |
457 | nr_entries = stack_trace_save_tsk(task, entries, |
458 | MAX_STACK_TRACE_DEPTH, 0); | |
a9712bc1 | 459 | |
e988e5ec | 460 | for (i = 0; i < nr_entries; i++) { |
8f5abe84 | 461 | seq_printf(m, "[<0>] %pB\n", (void *)entries[i]); |
a9712bc1 | 462 | } |
e988e5ec | 463 | |
a9712bc1 | 464 | unlock_trace(task); |
2ec220e2 KC |
465 | } |
466 | kfree(entries); | |
467 | ||
a9712bc1 | 468 | return err; |
2ec220e2 KC |
469 | } |
470 | #endif | |
471 | ||
5968cece | 472 | #ifdef CONFIG_SCHED_INFO |
1da177e4 LT |
473 | /* |
474 | * Provides /proc/PID/schedstat | |
475 | */ | |
f6e826ca AD |
476 | static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns, |
477 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 478 | { |
5968cece | 479 | if (unlikely(!sched_info_on())) |
08b55775 | 480 | seq_puts(m, "0 0 0\n"); |
5968cece NR |
481 | else |
482 | seq_printf(m, "%llu %llu %lu\n", | |
25ce3191 JP |
483 | (unsigned long long)task->se.sum_exec_runtime, |
484 | (unsigned long long)task->sched_info.run_delay, | |
485 | task->sched_info.pcount); | |
486 | ||
487 | return 0; | |
1da177e4 LT |
488 | } |
489 | #endif | |
490 | ||
9745512c AV |
491 | #ifdef CONFIG_LATENCYTOP |
492 | static int lstats_show_proc(struct seq_file *m, void *v) | |
493 | { | |
494 | int i; | |
13d77c37 HS |
495 | struct inode *inode = m->private; |
496 | struct task_struct *task = get_proc_task(inode); | |
9745512c | 497 | |
13d77c37 HS |
498 | if (!task) |
499 | return -ESRCH; | |
500 | seq_puts(m, "Latency Top version : v0.1\n"); | |
f6d2f584 | 501 | for (i = 0; i < LT_SAVECOUNT; i++) { |
34e49d4f JP |
502 | struct latency_record *lr = &task->latency_record[i]; |
503 | if (lr->backtrace[0]) { | |
9745512c | 504 | int q; |
34e49d4f JP |
505 | seq_printf(m, "%i %li %li", |
506 | lr->count, lr->time, lr->max); | |
9745512c | 507 | for (q = 0; q < LT_BACKTRACEDEPTH; q++) { |
34e49d4f | 508 | unsigned long bt = lr->backtrace[q]; |
accddc41 | 509 | |
34e49d4f | 510 | if (!bt) |
9745512c | 511 | break; |
34e49d4f | 512 | seq_printf(m, " %ps", (void *)bt); |
9745512c | 513 | } |
9d6de12f | 514 | seq_putc(m, '\n'); |
9745512c AV |
515 | } |
516 | ||
517 | } | |
13d77c37 | 518 | put_task_struct(task); |
9745512c AV |
519 | return 0; |
520 | } | |
521 | ||
522 | static int lstats_open(struct inode *inode, struct file *file) | |
523 | { | |
13d77c37 | 524 | return single_open(file, lstats_show_proc, inode); |
d6643d12 HS |
525 | } |
526 | ||
9745512c AV |
527 | static ssize_t lstats_write(struct file *file, const char __user *buf, |
528 | size_t count, loff_t *offs) | |
529 | { | |
496ad9aa | 530 | struct task_struct *task = get_proc_task(file_inode(file)); |
9745512c | 531 | |
13d77c37 HS |
532 | if (!task) |
533 | return -ESRCH; | |
e02c9b0d | 534 | clear_tsk_latency_tracing(task); |
13d77c37 | 535 | put_task_struct(task); |
9745512c AV |
536 | |
537 | return count; | |
538 | } | |
539 | ||
540 | static const struct file_operations proc_lstats_operations = { | |
541 | .open = lstats_open, | |
542 | .read = seq_read, | |
543 | .write = lstats_write, | |
544 | .llseek = seq_lseek, | |
13d77c37 | 545 | .release = single_release, |
9745512c AV |
546 | }; |
547 | ||
548 | #endif | |
549 | ||
6ba51e37 AD |
550 | static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns, |
551 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 552 | { |
ca79b0c2 | 553 | unsigned long totalpages = totalram_pages() + total_swap_pages; |
b95c35e7 | 554 | unsigned long points = 0; |
9066e5cf YS |
555 | long badness; |
556 | ||
557 | badness = oom_badness(task, totalpages); | |
558 | /* | |
559 | * Special case OOM_SCORE_ADJ_MIN for all others scale the | |
560 | * badness value into [0, 2000] range which we have been | |
561 | * exporting for a long time so userspace might depend on it. | |
562 | */ | |
563 | if (badness != LONG_MIN) | |
564 | points = (1000 + badness * 1000 / (long)totalpages) * 2 / 3; | |
1da177e4 | 565 | |
25ce3191 JP |
566 | seq_printf(m, "%lu\n", points); |
567 | ||
568 | return 0; | |
1da177e4 LT |
569 | } |
570 | ||
d85f50d5 | 571 | struct limit_names { |
cedbccab AD |
572 | const char *name; |
573 | const char *unit; | |
d85f50d5 NH |
574 | }; |
575 | ||
576 | static const struct limit_names lnames[RLIM_NLIMITS] = { | |
cff4edb5 | 577 | [RLIMIT_CPU] = {"Max cpu time", "seconds"}, |
d85f50d5 NH |
578 | [RLIMIT_FSIZE] = {"Max file size", "bytes"}, |
579 | [RLIMIT_DATA] = {"Max data size", "bytes"}, | |
580 | [RLIMIT_STACK] = {"Max stack size", "bytes"}, | |
581 | [RLIMIT_CORE] = {"Max core file size", "bytes"}, | |
582 | [RLIMIT_RSS] = {"Max resident set", "bytes"}, | |
583 | [RLIMIT_NPROC] = {"Max processes", "processes"}, | |
584 | [RLIMIT_NOFILE] = {"Max open files", "files"}, | |
585 | [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"}, | |
586 | [RLIMIT_AS] = {"Max address space", "bytes"}, | |
587 | [RLIMIT_LOCKS] = {"Max file locks", "locks"}, | |
588 | [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"}, | |
589 | [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"}, | |
590 | [RLIMIT_NICE] = {"Max nice priority", NULL}, | |
591 | [RLIMIT_RTPRIO] = {"Max realtime priority", NULL}, | |
8808117c | 592 | [RLIMIT_RTTIME] = {"Max realtime timeout", "us"}, |
d85f50d5 NH |
593 | }; |
594 | ||
595 | /* Display limits for a process */ | |
1c963eb1 AD |
596 | static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns, |
597 | struct pid *pid, struct task_struct *task) | |
d85f50d5 NH |
598 | { |
599 | unsigned int i; | |
d85f50d5 | 600 | unsigned long flags; |
d85f50d5 NH |
601 | |
602 | struct rlimit rlim[RLIM_NLIMITS]; | |
603 | ||
a6bebbc8 | 604 | if (!lock_task_sighand(task, &flags)) |
d85f50d5 | 605 | return 0; |
d85f50d5 NH |
606 | memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS); |
607 | unlock_task_sighand(task, &flags); | |
d85f50d5 NH |
608 | |
609 | /* | |
610 | * print the file header | |
611 | */ | |
afe922c2 AD |
612 | seq_puts(m, "Limit " |
613 | "Soft Limit " | |
614 | "Hard Limit " | |
615 | "Units \n"); | |
d85f50d5 NH |
616 | |
617 | for (i = 0; i < RLIM_NLIMITS; i++) { | |
618 | if (rlim[i].rlim_cur == RLIM_INFINITY) | |
1c963eb1 | 619 | seq_printf(m, "%-25s %-20s ", |
25ce3191 | 620 | lnames[i].name, "unlimited"); |
d85f50d5 | 621 | else |
1c963eb1 | 622 | seq_printf(m, "%-25s %-20lu ", |
25ce3191 | 623 | lnames[i].name, rlim[i].rlim_cur); |
d85f50d5 NH |
624 | |
625 | if (rlim[i].rlim_max == RLIM_INFINITY) | |
1c963eb1 | 626 | seq_printf(m, "%-20s ", "unlimited"); |
d85f50d5 | 627 | else |
1c963eb1 | 628 | seq_printf(m, "%-20lu ", rlim[i].rlim_max); |
d85f50d5 NH |
629 | |
630 | if (lnames[i].unit) | |
1c963eb1 | 631 | seq_printf(m, "%-10s\n", lnames[i].unit); |
d85f50d5 | 632 | else |
1c963eb1 | 633 | seq_putc(m, '\n'); |
d85f50d5 NH |
634 | } |
635 | ||
1c963eb1 | 636 | return 0; |
d85f50d5 NH |
637 | } |
638 | ||
ebcb6734 | 639 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 AD |
640 | static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns, |
641 | struct pid *pid, struct task_struct *task) | |
ebcb6734 | 642 | { |
631b7aba SRRH |
643 | struct syscall_info info; |
644 | u64 *args = &info.data.args[0]; | |
25ce3191 JP |
645 | int res; |
646 | ||
647 | res = lock_trace(task); | |
a9712bc1 AV |
648 | if (res) |
649 | return res; | |
ebcb6734 | 650 | |
631b7aba | 651 | if (task_current_syscall(task, &info)) |
09d93bd6 | 652 | seq_puts(m, "running\n"); |
631b7aba SRRH |
653 | else if (info.data.nr < 0) |
654 | seq_printf(m, "%d 0x%llx 0x%llx\n", | |
655 | info.data.nr, info.sp, info.data.instruction_pointer); | |
a9712bc1 | 656 | else |
09d93bd6 | 657 | seq_printf(m, |
631b7aba SRRH |
658 | "%d 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx\n", |
659 | info.data.nr, | |
ebcb6734 | 660 | args[0], args[1], args[2], args[3], args[4], args[5], |
631b7aba | 661 | info.sp, info.data.instruction_pointer); |
a9712bc1 | 662 | unlock_trace(task); |
25ce3191 JP |
663 | |
664 | return 0; | |
ebcb6734 RM |
665 | } |
666 | #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */ | |
667 | ||
1da177e4 LT |
668 | /************************************************************************/ |
669 | /* Here the fs part begins */ | |
670 | /************************************************************************/ | |
671 | ||
672 | /* permission checks */ | |
51a18734 | 673 | static bool proc_fd_access_allowed(struct inode *inode) |
1da177e4 | 674 | { |
778c1144 | 675 | struct task_struct *task; |
51a18734 | 676 | bool allowed = false; |
df26c40e EB |
677 | /* Allow access to a task's file descriptors if it is us or we |
678 | * may use ptrace attach to the process and find out that | |
679 | * information. | |
778c1144 EB |
680 | */ |
681 | task = get_proc_task(inode); | |
df26c40e | 682 | if (task) { |
caaee623 | 683 | allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS); |
778c1144 | 684 | put_task_struct(task); |
df26c40e | 685 | } |
778c1144 | 686 | return allowed; |
1da177e4 LT |
687 | } |
688 | ||
c1632a0f | 689 | int proc_setattr(struct mnt_idmap *idmap, struct dentry *dentry, |
549c7297 | 690 | struct iattr *attr) |
6d76fa58 LT |
691 | { |
692 | int error; | |
2b0143b5 | 693 | struct inode *inode = d_inode(dentry); |
6d76fa58 LT |
694 | |
695 | if (attr->ia_valid & ATTR_MODE) | |
696 | return -EPERM; | |
697 | ||
c1632a0f | 698 | error = setattr_prepare(&nop_mnt_idmap, dentry, attr); |
1025774c CH |
699 | if (error) |
700 | return error; | |
701 | ||
c1632a0f | 702 | setattr_copy(&nop_mnt_idmap, inode, attr); |
1025774c | 703 | return 0; |
6d76fa58 LT |
704 | } |
705 | ||
0499680a VK |
706 | /* |
707 | * May current process learn task's sched/cmdline info (for hide_pid_min=1) | |
708 | * or euid/egid (for hide_pid_min=2)? | |
709 | */ | |
fa10fed3 | 710 | static bool has_pid_permissions(struct proc_fs_info *fs_info, |
0499680a | 711 | struct task_struct *task, |
e61bb8b3 | 712 | enum proc_hidepid hide_pid_min) |
0499680a | 713 | { |
24a71ce5 AG |
714 | /* |
715 | * If 'hidpid' mount option is set force a ptrace check, | |
716 | * we indicate that we are using a filesystem syscall | |
717 | * by passing PTRACE_MODE_READ_FSCREDS | |
718 | */ | |
719 | if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE) | |
720 | return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS); | |
721 | ||
fa10fed3 | 722 | if (fs_info->hide_pid < hide_pid_min) |
0499680a | 723 | return true; |
fa10fed3 | 724 | if (in_group_p(fs_info->pid_gid)) |
0499680a | 725 | return true; |
caaee623 | 726 | return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS); |
0499680a VK |
727 | } |
728 | ||
729 | ||
4609e1f1 | 730 | static int proc_pid_permission(struct mnt_idmap *idmap, |
549c7297 | 731 | struct inode *inode, int mask) |
0499680a | 732 | { |
fa10fed3 | 733 | struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb); |
0499680a VK |
734 | struct task_struct *task; |
735 | bool has_perms; | |
736 | ||
737 | task = get_proc_task(inode); | |
a2ef990a XF |
738 | if (!task) |
739 | return -ESRCH; | |
fa10fed3 | 740 | has_perms = has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS); |
0499680a VK |
741 | put_task_struct(task); |
742 | ||
743 | if (!has_perms) { | |
fa10fed3 | 744 | if (fs_info->hide_pid == HIDEPID_INVISIBLE) { |
0499680a VK |
745 | /* |
746 | * Let's make getdents(), stat(), and open() | |
747 | * consistent with each other. If a process | |
748 | * may not stat() a file, it shouldn't be seen | |
749 | * in procfs at all. | |
750 | */ | |
751 | return -ENOENT; | |
752 | } | |
753 | ||
754 | return -EPERM; | |
755 | } | |
4609e1f1 | 756 | return generic_permission(&nop_mnt_idmap, inode, mask); |
0499680a VK |
757 | } |
758 | ||
759 | ||
760 | ||
c5ef1c42 | 761 | static const struct inode_operations proc_def_inode_operations = { |
6d76fa58 LT |
762 | .setattr = proc_setattr, |
763 | }; | |
764 | ||
be614086 EB |
765 | static int proc_single_show(struct seq_file *m, void *v) |
766 | { | |
767 | struct inode *inode = m->private; | |
9d78edea | 768 | struct pid_namespace *ns = proc_pid_ns(inode->i_sb); |
76f668be | 769 | struct pid *pid = proc_pid(inode); |
be614086 EB |
770 | struct task_struct *task; |
771 | int ret; | |
772 | ||
be614086 EB |
773 | task = get_pid_task(pid, PIDTYPE_PID); |
774 | if (!task) | |
775 | return -ESRCH; | |
776 | ||
777 | ret = PROC_I(inode)->op.proc_show(m, ns, pid, task); | |
778 | ||
779 | put_task_struct(task); | |
780 | return ret; | |
781 | } | |
782 | ||
783 | static int proc_single_open(struct inode *inode, struct file *filp) | |
784 | { | |
c6a34058 | 785 | return single_open(filp, proc_single_show, inode); |
be614086 EB |
786 | } |
787 | ||
788 | static const struct file_operations proc_single_file_operations = { | |
789 | .open = proc_single_open, | |
790 | .read = seq_read, | |
791 | .llseek = seq_lseek, | |
792 | .release = single_release, | |
793 | }; | |
794 | ||
5381e169 ON |
795 | |
796 | struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode) | |
1da177e4 | 797 | { |
5381e169 ON |
798 | struct task_struct *task = get_proc_task(inode); |
799 | struct mm_struct *mm = ERR_PTR(-ESRCH); | |
e268337d | 800 | |
5381e169 | 801 | if (task) { |
caaee623 | 802 | mm = mm_access(task, mode | PTRACE_MODE_FSCREDS); |
5381e169 | 803 | put_task_struct(task); |
e268337d | 804 | |
5381e169 ON |
805 | if (!IS_ERR_OR_NULL(mm)) { |
806 | /* ensure this mm_struct can't be freed */ | |
f1f10076 | 807 | mmgrab(mm); |
5381e169 ON |
808 | /* but do not pin its memory */ |
809 | mmput(mm); | |
810 | } | |
811 | } | |
812 | ||
813 | return mm; | |
814 | } | |
815 | ||
816 | static int __mem_open(struct inode *inode, struct file *file, unsigned int mode) | |
817 | { | |
818 | struct mm_struct *mm = proc_mem_open(inode, mode); | |
e268337d LT |
819 | |
820 | if (IS_ERR(mm)) | |
821 | return PTR_ERR(mm); | |
822 | ||
e268337d | 823 | file->private_data = mm; |
1da177e4 LT |
824 | return 0; |
825 | } | |
826 | ||
b409e578 CW |
827 | static int mem_open(struct inode *inode, struct file *file) |
828 | { | |
bc452b4b DH |
829 | int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH); |
830 | ||
831 | /* OK to pass negative loff_t, we can catch out-of-range */ | |
832 | file->f_mode |= FMODE_UNSIGNED_OFFSET; | |
833 | ||
834 | return ret; | |
b409e578 CW |
835 | } |
836 | ||
572d34b9 ON |
837 | static ssize_t mem_rw(struct file *file, char __user *buf, |
838 | size_t count, loff_t *ppos, int write) | |
1da177e4 | 839 | { |
e268337d | 840 | struct mm_struct *mm = file->private_data; |
572d34b9 ON |
841 | unsigned long addr = *ppos; |
842 | ssize_t copied; | |
1da177e4 | 843 | char *page; |
272ddc8b | 844 | unsigned int flags; |
1da177e4 | 845 | |
e268337d LT |
846 | if (!mm) |
847 | return 0; | |
99f89551 | 848 | |
0ee931c4 | 849 | page = (char *)__get_free_page(GFP_KERNEL); |
30cd8903 | 850 | if (!page) |
e268337d | 851 | return -ENOMEM; |
1da177e4 | 852 | |
f7ca54f4 | 853 | copied = 0; |
388f7934 | 854 | if (!mmget_not_zero(mm)) |
6d08f2c7 ON |
855 | goto free; |
856 | ||
f511c0b1 | 857 | flags = FOLL_FORCE | (write ? FOLL_WRITE : 0); |
6347e8d5 | 858 | |
1da177e4 | 859 | while (count > 0) { |
d238692b | 860 | size_t this_len = min_t(size_t, count, PAGE_SIZE); |
1da177e4 | 861 | |
572d34b9 | 862 | if (write && copy_from_user(page, buf, this_len)) { |
1da177e4 LT |
863 | copied = -EFAULT; |
864 | break; | |
865 | } | |
572d34b9 | 866 | |
6347e8d5 | 867 | this_len = access_remote_vm(mm, addr, page, this_len, flags); |
572d34b9 | 868 | if (!this_len) { |
1da177e4 LT |
869 | if (!copied) |
870 | copied = -EIO; | |
871 | break; | |
872 | } | |
572d34b9 ON |
873 | |
874 | if (!write && copy_to_user(buf, page, this_len)) { | |
875 | copied = -EFAULT; | |
876 | break; | |
877 | } | |
878 | ||
879 | buf += this_len; | |
880 | addr += this_len; | |
881 | copied += this_len; | |
882 | count -= this_len; | |
1da177e4 | 883 | } |
572d34b9 | 884 | *ppos = addr; |
30cd8903 | 885 | |
6d08f2c7 ON |
886 | mmput(mm); |
887 | free: | |
30cd8903 | 888 | free_page((unsigned long) page); |
1da177e4 LT |
889 | return copied; |
890 | } | |
1da177e4 | 891 | |
572d34b9 ON |
892 | static ssize_t mem_read(struct file *file, char __user *buf, |
893 | size_t count, loff_t *ppos) | |
894 | { | |
895 | return mem_rw(file, buf, count, ppos, 0); | |
896 | } | |
897 | ||
898 | static ssize_t mem_write(struct file *file, const char __user *buf, | |
899 | size_t count, loff_t *ppos) | |
900 | { | |
901 | return mem_rw(file, (char __user*)buf, count, ppos, 1); | |
902 | } | |
903 | ||
85863e47 | 904 | loff_t mem_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
905 | { |
906 | switch (orig) { | |
907 | case 0: | |
908 | file->f_pos = offset; | |
909 | break; | |
910 | case 1: | |
911 | file->f_pos += offset; | |
912 | break; | |
913 | default: | |
914 | return -EINVAL; | |
915 | } | |
916 | force_successful_syscall_return(); | |
917 | return file->f_pos; | |
918 | } | |
919 | ||
e268337d LT |
920 | static int mem_release(struct inode *inode, struct file *file) |
921 | { | |
922 | struct mm_struct *mm = file->private_data; | |
71879d3c | 923 | if (mm) |
6d08f2c7 | 924 | mmdrop(mm); |
e268337d LT |
925 | return 0; |
926 | } | |
927 | ||
00977a59 | 928 | static const struct file_operations proc_mem_operations = { |
1da177e4 LT |
929 | .llseek = mem_lseek, |
930 | .read = mem_read, | |
931 | .write = mem_write, | |
932 | .open = mem_open, | |
e268337d | 933 | .release = mem_release, |
1da177e4 LT |
934 | }; |
935 | ||
b409e578 CW |
936 | static int environ_open(struct inode *inode, struct file *file) |
937 | { | |
938 | return __mem_open(inode, file, PTRACE_MODE_READ); | |
939 | } | |
940 | ||
315e28c8 JP |
941 | static ssize_t environ_read(struct file *file, char __user *buf, |
942 | size_t count, loff_t *ppos) | |
943 | { | |
315e28c8 JP |
944 | char *page; |
945 | unsigned long src = *ppos; | |
b409e578 CW |
946 | int ret = 0; |
947 | struct mm_struct *mm = file->private_data; | |
a3b609ef | 948 | unsigned long env_start, env_end; |
315e28c8 | 949 | |
8148a73c MK |
950 | /* Ensure the process spawned far enough to have an environment. */ |
951 | if (!mm || !mm->env_end) | |
b409e578 | 952 | return 0; |
315e28c8 | 953 | |
0ee931c4 | 954 | page = (char *)__get_free_page(GFP_KERNEL); |
315e28c8 | 955 | if (!page) |
b409e578 | 956 | return -ENOMEM; |
315e28c8 | 957 | |
d6f64b89 | 958 | ret = 0; |
388f7934 | 959 | if (!mmget_not_zero(mm)) |
b409e578 | 960 | goto free; |
a3b609ef | 961 | |
88aa7cc6 | 962 | spin_lock(&mm->arg_lock); |
a3b609ef MG |
963 | env_start = mm->env_start; |
964 | env_end = mm->env_end; | |
88aa7cc6 | 965 | spin_unlock(&mm->arg_lock); |
a3b609ef | 966 | |
315e28c8 | 967 | while (count > 0) { |
e8905ec2 DH |
968 | size_t this_len, max_len; |
969 | int retval; | |
315e28c8 | 970 | |
a3b609ef | 971 | if (src >= (env_end - env_start)) |
315e28c8 JP |
972 | break; |
973 | ||
a3b609ef | 974 | this_len = env_end - (env_start + src); |
e8905ec2 DH |
975 | |
976 | max_len = min_t(size_t, PAGE_SIZE, count); | |
977 | this_len = min(max_len, this_len); | |
315e28c8 | 978 | |
7f7ccc2c | 979 | retval = access_remote_vm(mm, (env_start + src), page, this_len, FOLL_ANON); |
315e28c8 JP |
980 | |
981 | if (retval <= 0) { | |
982 | ret = retval; | |
983 | break; | |
984 | } | |
985 | ||
986 | if (copy_to_user(buf, page, retval)) { | |
987 | ret = -EFAULT; | |
988 | break; | |
989 | } | |
990 | ||
991 | ret += retval; | |
992 | src += retval; | |
993 | buf += retval; | |
994 | count -= retval; | |
995 | } | |
996 | *ppos = src; | |
315e28c8 | 997 | mmput(mm); |
b409e578 CW |
998 | |
999 | free: | |
315e28c8 | 1000 | free_page((unsigned long) page); |
315e28c8 JP |
1001 | return ret; |
1002 | } | |
1003 | ||
1004 | static const struct file_operations proc_environ_operations = { | |
b409e578 | 1005 | .open = environ_open, |
315e28c8 | 1006 | .read = environ_read, |
87df8424 | 1007 | .llseek = generic_file_llseek, |
b409e578 | 1008 | .release = mem_release, |
315e28c8 JP |
1009 | }; |
1010 | ||
c5317167 AV |
1011 | static int auxv_open(struct inode *inode, struct file *file) |
1012 | { | |
1013 | return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS); | |
1014 | } | |
1015 | ||
1016 | static ssize_t auxv_read(struct file *file, char __user *buf, | |
1017 | size_t count, loff_t *ppos) | |
1018 | { | |
1019 | struct mm_struct *mm = file->private_data; | |
1020 | unsigned int nwords = 0; | |
06b2849d LY |
1021 | |
1022 | if (!mm) | |
1023 | return 0; | |
c5317167 AV |
1024 | do { |
1025 | nwords += 2; | |
1026 | } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */ | |
1027 | return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv, | |
1028 | nwords * sizeof(mm->saved_auxv[0])); | |
1029 | } | |
1030 | ||
1031 | static const struct file_operations proc_auxv_operations = { | |
1032 | .open = auxv_open, | |
1033 | .read = auxv_read, | |
1034 | .llseek = generic_file_llseek, | |
1035 | .release = mem_release, | |
1036 | }; | |
1037 | ||
fa0cbbf1 DR |
1038 | static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count, |
1039 | loff_t *ppos) | |
1040 | { | |
496ad9aa | 1041 | struct task_struct *task = get_proc_task(file_inode(file)); |
fa0cbbf1 DR |
1042 | char buffer[PROC_NUMBUF]; |
1043 | int oom_adj = OOM_ADJUST_MIN; | |
1044 | size_t len; | |
fa0cbbf1 DR |
1045 | |
1046 | if (!task) | |
1047 | return -ESRCH; | |
f913da59 MH |
1048 | if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX) |
1049 | oom_adj = OOM_ADJUST_MAX; | |
1050 | else | |
1051 | oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) / | |
1052 | OOM_SCORE_ADJ_MAX; | |
fa0cbbf1 | 1053 | put_task_struct(task); |
66606567 CH |
1054 | if (oom_adj > OOM_ADJUST_MAX) |
1055 | oom_adj = OOM_ADJUST_MAX; | |
fa0cbbf1 DR |
1056 | len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj); |
1057 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
1058 | } | |
1059 | ||
1d5f0acb MH |
1060 | static int __set_oom_adj(struct file *file, int oom_adj, bool legacy) |
1061 | { | |
44a70ade | 1062 | struct mm_struct *mm = NULL; |
1d5f0acb MH |
1063 | struct task_struct *task; |
1064 | int err = 0; | |
1065 | ||
1066 | task = get_proc_task(file_inode(file)); | |
1067 | if (!task) | |
1068 | return -ESRCH; | |
1069 | ||
1070 | mutex_lock(&oom_adj_mutex); | |
1071 | if (legacy) { | |
1072 | if (oom_adj < task->signal->oom_score_adj && | |
1073 | !capable(CAP_SYS_RESOURCE)) { | |
1074 | err = -EACCES; | |
1075 | goto err_unlock; | |
1076 | } | |
1077 | /* | |
1078 | * /proc/pid/oom_adj is provided for legacy purposes, ask users to use | |
1079 | * /proc/pid/oom_score_adj instead. | |
1080 | */ | |
1081 | pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n", | |
1082 | current->comm, task_pid_nr(current), task_pid_nr(task), | |
1083 | task_pid_nr(task)); | |
1084 | } else { | |
1085 | if ((short)oom_adj < task->signal->oom_score_adj_min && | |
1086 | !capable(CAP_SYS_RESOURCE)) { | |
1087 | err = -EACCES; | |
1088 | goto err_unlock; | |
1089 | } | |
1090 | } | |
1091 | ||
44a70ade MH |
1092 | /* |
1093 | * Make sure we will check other processes sharing the mm if this is | |
1094 | * not vfrok which wants its own oom_score_adj. | |
1095 | * pin the mm so it doesn't go away and get reused after task_unlock | |
1096 | */ | |
1097 | if (!task->vfork_done) { | |
1098 | struct task_struct *p = find_lock_task_mm(task); | |
1099 | ||
1100 | if (p) { | |
67197a4f | 1101 | if (test_bit(MMF_MULTIPROCESS, &p->mm->flags)) { |
44a70ade | 1102 | mm = p->mm; |
f1f10076 | 1103 | mmgrab(mm); |
44a70ade MH |
1104 | } |
1105 | task_unlock(p); | |
1106 | } | |
1107 | } | |
1108 | ||
1d5f0acb MH |
1109 | task->signal->oom_score_adj = oom_adj; |
1110 | if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE)) | |
1111 | task->signal->oom_score_adj_min = (short)oom_adj; | |
1112 | trace_oom_score_adj_update(task); | |
44a70ade MH |
1113 | |
1114 | if (mm) { | |
1115 | struct task_struct *p; | |
1116 | ||
1117 | rcu_read_lock(); | |
1118 | for_each_process(p) { | |
1119 | if (same_thread_group(task, p)) | |
1120 | continue; | |
1121 | ||
1122 | /* do not touch kernel threads or the global init */ | |
1123 | if (p->flags & PF_KTHREAD || is_global_init(p)) | |
1124 | continue; | |
1125 | ||
1126 | task_lock(p); | |
1127 | if (!p->vfork_done && process_shares_mm(p, mm)) { | |
44a70ade MH |
1128 | p->signal->oom_score_adj = oom_adj; |
1129 | if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE)) | |
1130 | p->signal->oom_score_adj_min = (short)oom_adj; | |
1131 | } | |
1132 | task_unlock(p); | |
1133 | } | |
1134 | rcu_read_unlock(); | |
1135 | mmdrop(mm); | |
1136 | } | |
1d5f0acb MH |
1137 | err_unlock: |
1138 | mutex_unlock(&oom_adj_mutex); | |
1139 | put_task_struct(task); | |
1140 | return err; | |
1141 | } | |
f913da59 | 1142 | |
b72bdfa7 DR |
1143 | /* |
1144 | * /proc/pid/oom_adj exists solely for backwards compatibility with previous | |
1145 | * kernels. The effective policy is defined by oom_score_adj, which has a | |
1146 | * different scale: oom_adj grew exponentially and oom_score_adj grows linearly. | |
1147 | * Values written to oom_adj are simply mapped linearly to oom_score_adj. | |
1148 | * Processes that become oom disabled via oom_adj will still be oom disabled | |
1149 | * with this implementation. | |
1150 | * | |
1151 | * oom_adj cannot be removed since existing userspace binaries use it. | |
1152 | */ | |
fa0cbbf1 DR |
1153 | static ssize_t oom_adj_write(struct file *file, const char __user *buf, |
1154 | size_t count, loff_t *ppos) | |
1155 | { | |
860a2e7f | 1156 | char buffer[PROC_NUMBUF] = {}; |
fa0cbbf1 | 1157 | int oom_adj; |
fa0cbbf1 DR |
1158 | int err; |
1159 | ||
fa0cbbf1 DR |
1160 | if (count > sizeof(buffer) - 1) |
1161 | count = sizeof(buffer) - 1; | |
1162 | if (copy_from_user(buffer, buf, count)) { | |
1163 | err = -EFAULT; | |
1164 | goto out; | |
1165 | } | |
1166 | ||
1167 | err = kstrtoint(strstrip(buffer), 0, &oom_adj); | |
1168 | if (err) | |
1169 | goto out; | |
1170 | if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) && | |
1171 | oom_adj != OOM_DISABLE) { | |
1172 | err = -EINVAL; | |
1173 | goto out; | |
1174 | } | |
1175 | ||
fa0cbbf1 DR |
1176 | /* |
1177 | * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum | |
1178 | * value is always attainable. | |
1179 | */ | |
1180 | if (oom_adj == OOM_ADJUST_MAX) | |
1181 | oom_adj = OOM_SCORE_ADJ_MAX; | |
1182 | else | |
1183 | oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE; | |
1184 | ||
1d5f0acb | 1185 | err = __set_oom_adj(file, oom_adj, true); |
fa0cbbf1 DR |
1186 | out: |
1187 | return err < 0 ? err : count; | |
1188 | } | |
1189 | ||
1190 | static const struct file_operations proc_oom_adj_operations = { | |
1191 | .read = oom_adj_read, | |
1192 | .write = oom_adj_write, | |
1193 | .llseek = generic_file_llseek, | |
1194 | }; | |
1195 | ||
a63d83f4 DR |
1196 | static ssize_t oom_score_adj_read(struct file *file, char __user *buf, |
1197 | size_t count, loff_t *ppos) | |
1198 | { | |
496ad9aa | 1199 | struct task_struct *task = get_proc_task(file_inode(file)); |
a63d83f4 | 1200 | char buffer[PROC_NUMBUF]; |
a9c58b90 | 1201 | short oom_score_adj = OOM_SCORE_ADJ_MIN; |
a63d83f4 DR |
1202 | size_t len; |
1203 | ||
1204 | if (!task) | |
1205 | return -ESRCH; | |
f913da59 | 1206 | oom_score_adj = task->signal->oom_score_adj; |
a63d83f4 | 1207 | put_task_struct(task); |
a9c58b90 | 1208 | len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj); |
a63d83f4 DR |
1209 | return simple_read_from_buffer(buf, count, ppos, buffer, len); |
1210 | } | |
1211 | ||
1212 | static ssize_t oom_score_adj_write(struct file *file, const char __user *buf, | |
1213 | size_t count, loff_t *ppos) | |
1214 | { | |
860a2e7f | 1215 | char buffer[PROC_NUMBUF] = {}; |
0a8cb8e3 | 1216 | int oom_score_adj; |
a63d83f4 DR |
1217 | int err; |
1218 | ||
a63d83f4 DR |
1219 | if (count > sizeof(buffer) - 1) |
1220 | count = sizeof(buffer) - 1; | |
723548bf DR |
1221 | if (copy_from_user(buffer, buf, count)) { |
1222 | err = -EFAULT; | |
1223 | goto out; | |
1224 | } | |
a63d83f4 | 1225 | |
0a8cb8e3 | 1226 | err = kstrtoint(strstrip(buffer), 0, &oom_score_adj); |
a63d83f4 | 1227 | if (err) |
723548bf | 1228 | goto out; |
a63d83f4 | 1229 | if (oom_score_adj < OOM_SCORE_ADJ_MIN || |
723548bf DR |
1230 | oom_score_adj > OOM_SCORE_ADJ_MAX) { |
1231 | err = -EINVAL; | |
1232 | goto out; | |
1233 | } | |
a63d83f4 | 1234 | |
1d5f0acb | 1235 | err = __set_oom_adj(file, oom_score_adj, false); |
723548bf DR |
1236 | out: |
1237 | return err < 0 ? err : count; | |
a63d83f4 DR |
1238 | } |
1239 | ||
1240 | static const struct file_operations proc_oom_score_adj_operations = { | |
1241 | .read = oom_score_adj_read, | |
1242 | .write = oom_score_adj_write, | |
6038f373 | 1243 | .llseek = default_llseek, |
a63d83f4 DR |
1244 | }; |
1245 | ||
4b7d248b | 1246 | #ifdef CONFIG_AUDIT |
b4eb4f7f | 1247 | #define TMPBUFLEN 11 |
1da177e4 LT |
1248 | static ssize_t proc_loginuid_read(struct file * file, char __user * buf, |
1249 | size_t count, loff_t *ppos) | |
1250 | { | |
496ad9aa | 1251 | struct inode * inode = file_inode(file); |
99f89551 | 1252 | struct task_struct *task = get_proc_task(inode); |
1da177e4 LT |
1253 | ssize_t length; |
1254 | char tmpbuf[TMPBUFLEN]; | |
1255 | ||
99f89551 EB |
1256 | if (!task) |
1257 | return -ESRCH; | |
1da177e4 | 1258 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", |
e1760bd5 EB |
1259 | from_kuid(file->f_cred->user_ns, |
1260 | audit_get_loginuid(task))); | |
99f89551 | 1261 | put_task_struct(task); |
1da177e4 LT |
1262 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); |
1263 | } | |
1264 | ||
1265 | static ssize_t proc_loginuid_write(struct file * file, const char __user * buf, | |
1266 | size_t count, loff_t *ppos) | |
1267 | { | |
496ad9aa | 1268 | struct inode * inode = file_inode(file); |
1da177e4 | 1269 | uid_t loginuid; |
e1760bd5 | 1270 | kuid_t kloginuid; |
774636e1 | 1271 | int rv; |
1da177e4 | 1272 | |
4ea33a97 JA |
1273 | /* Don't let kthreads write their own loginuid */ |
1274 | if (current->flags & PF_KTHREAD) | |
1275 | return -EPERM; | |
1276 | ||
7dc52157 PM |
1277 | rcu_read_lock(); |
1278 | if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) { | |
1279 | rcu_read_unlock(); | |
1da177e4 | 1280 | return -EPERM; |
7dc52157 PM |
1281 | } |
1282 | rcu_read_unlock(); | |
1da177e4 | 1283 | |
1da177e4 LT |
1284 | if (*ppos != 0) { |
1285 | /* No partial writes. */ | |
1286 | return -EINVAL; | |
1287 | } | |
1da177e4 | 1288 | |
774636e1 AD |
1289 | rv = kstrtou32_from_user(buf, count, 10, &loginuid); |
1290 | if (rv < 0) | |
1291 | return rv; | |
81407c84 EP |
1292 | |
1293 | /* is userspace tring to explicitly UNSET the loginuid? */ | |
1294 | if (loginuid == AUDIT_UID_UNSET) { | |
1295 | kloginuid = INVALID_UID; | |
1296 | } else { | |
1297 | kloginuid = make_kuid(file->f_cred->user_ns, loginuid); | |
774636e1 AD |
1298 | if (!uid_valid(kloginuid)) |
1299 | return -EINVAL; | |
e1760bd5 EB |
1300 | } |
1301 | ||
774636e1 AD |
1302 | rv = audit_set_loginuid(kloginuid); |
1303 | if (rv < 0) | |
1304 | return rv; | |
1305 | return count; | |
1da177e4 LT |
1306 | } |
1307 | ||
00977a59 | 1308 | static const struct file_operations proc_loginuid_operations = { |
1da177e4 LT |
1309 | .read = proc_loginuid_read, |
1310 | .write = proc_loginuid_write, | |
87df8424 | 1311 | .llseek = generic_file_llseek, |
1da177e4 | 1312 | }; |
1e0bd755 EP |
1313 | |
1314 | static ssize_t proc_sessionid_read(struct file * file, char __user * buf, | |
1315 | size_t count, loff_t *ppos) | |
1316 | { | |
496ad9aa | 1317 | struct inode * inode = file_inode(file); |
1e0bd755 EP |
1318 | struct task_struct *task = get_proc_task(inode); |
1319 | ssize_t length; | |
1320 | char tmpbuf[TMPBUFLEN]; | |
1321 | ||
1322 | if (!task) | |
1323 | return -ESRCH; | |
1324 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", | |
1325 | audit_get_sessionid(task)); | |
1326 | put_task_struct(task); | |
1327 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); | |
1328 | } | |
1329 | ||
1330 | static const struct file_operations proc_sessionid_operations = { | |
1331 | .read = proc_sessionid_read, | |
87df8424 | 1332 | .llseek = generic_file_llseek, |
1e0bd755 | 1333 | }; |
1da177e4 LT |
1334 | #endif |
1335 | ||
f4f154fd AM |
1336 | #ifdef CONFIG_FAULT_INJECTION |
1337 | static ssize_t proc_fault_inject_read(struct file * file, char __user * buf, | |
1338 | size_t count, loff_t *ppos) | |
1339 | { | |
496ad9aa | 1340 | struct task_struct *task = get_proc_task(file_inode(file)); |
f4f154fd AM |
1341 | char buffer[PROC_NUMBUF]; |
1342 | size_t len; | |
1343 | int make_it_fail; | |
f4f154fd AM |
1344 | |
1345 | if (!task) | |
1346 | return -ESRCH; | |
1347 | make_it_fail = task->make_it_fail; | |
1348 | put_task_struct(task); | |
1349 | ||
1350 | len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail); | |
0c28f287 AM |
1351 | |
1352 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
f4f154fd AM |
1353 | } |
1354 | ||
1355 | static ssize_t proc_fault_inject_write(struct file * file, | |
1356 | const char __user * buf, size_t count, loff_t *ppos) | |
1357 | { | |
1358 | struct task_struct *task; | |
860a2e7f | 1359 | char buffer[PROC_NUMBUF] = {}; |
f4f154fd | 1360 | int make_it_fail; |
774636e1 | 1361 | int rv; |
f4f154fd AM |
1362 | |
1363 | if (!capable(CAP_SYS_RESOURCE)) | |
1364 | return -EPERM; | |
860a2e7f | 1365 | |
f4f154fd AM |
1366 | if (count > sizeof(buffer) - 1) |
1367 | count = sizeof(buffer) - 1; | |
1368 | if (copy_from_user(buffer, buf, count)) | |
1369 | return -EFAULT; | |
774636e1 AD |
1370 | rv = kstrtoint(strstrip(buffer), 0, &make_it_fail); |
1371 | if (rv < 0) | |
1372 | return rv; | |
16caed31 DJ |
1373 | if (make_it_fail < 0 || make_it_fail > 1) |
1374 | return -EINVAL; | |
1375 | ||
496ad9aa | 1376 | task = get_proc_task(file_inode(file)); |
f4f154fd AM |
1377 | if (!task) |
1378 | return -ESRCH; | |
1379 | task->make_it_fail = make_it_fail; | |
1380 | put_task_struct(task); | |
cba8aafe VL |
1381 | |
1382 | return count; | |
f4f154fd AM |
1383 | } |
1384 | ||
00977a59 | 1385 | static const struct file_operations proc_fault_inject_operations = { |
f4f154fd AM |
1386 | .read = proc_fault_inject_read, |
1387 | .write = proc_fault_inject_write, | |
87df8424 | 1388 | .llseek = generic_file_llseek, |
f4f154fd | 1389 | }; |
e41d5818 DV |
1390 | |
1391 | static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf, | |
1392 | size_t count, loff_t *ppos) | |
1393 | { | |
1394 | struct task_struct *task; | |
9049f2f6 AM |
1395 | int err; |
1396 | unsigned int n; | |
e41d5818 | 1397 | |
1203c8e6 AM |
1398 | err = kstrtouint_from_user(buf, count, 0, &n); |
1399 | if (err) | |
1400 | return err; | |
1401 | ||
e41d5818 DV |
1402 | task = get_proc_task(file_inode(file)); |
1403 | if (!task) | |
1404 | return -ESRCH; | |
9f7118b2 | 1405 | task->fail_nth = n; |
e41d5818 | 1406 | put_task_struct(task); |
1203c8e6 | 1407 | |
e41d5818 DV |
1408 | return count; |
1409 | } | |
1410 | ||
1411 | static ssize_t proc_fail_nth_read(struct file *file, char __user *buf, | |
1412 | size_t count, loff_t *ppos) | |
1413 | { | |
1414 | struct task_struct *task; | |
bfc74093 AM |
1415 | char numbuf[PROC_NUMBUF]; |
1416 | ssize_t len; | |
e41d5818 DV |
1417 | |
1418 | task = get_proc_task(file_inode(file)); | |
1419 | if (!task) | |
1420 | return -ESRCH; | |
9f7118b2 | 1421 | len = snprintf(numbuf, sizeof(numbuf), "%u\n", task->fail_nth); |
1203c8e6 | 1422 | put_task_struct(task); |
a44937fe | 1423 | return simple_read_from_buffer(buf, count, ppos, numbuf, len); |
e41d5818 DV |
1424 | } |
1425 | ||
1426 | static const struct file_operations proc_fail_nth_operations = { | |
1427 | .read = proc_fail_nth_read, | |
1428 | .write = proc_fail_nth_write, | |
1429 | }; | |
f4f154fd AM |
1430 | #endif |
1431 | ||
9745512c | 1432 | |
43ae34cb IM |
1433 | #ifdef CONFIG_SCHED_DEBUG |
1434 | /* | |
1435 | * Print out various scheduling related per-task fields: | |
1436 | */ | |
1437 | static int sched_show(struct seq_file *m, void *v) | |
1438 | { | |
1439 | struct inode *inode = m->private; | |
9d78edea | 1440 | struct pid_namespace *ns = proc_pid_ns(inode->i_sb); |
43ae34cb IM |
1441 | struct task_struct *p; |
1442 | ||
43ae34cb IM |
1443 | p = get_proc_task(inode); |
1444 | if (!p) | |
1445 | return -ESRCH; | |
74dc3384 | 1446 | proc_sched_show_task(p, ns, m); |
43ae34cb IM |
1447 | |
1448 | put_task_struct(p); | |
1449 | ||
1450 | return 0; | |
1451 | } | |
1452 | ||
1453 | static ssize_t | |
1454 | sched_write(struct file *file, const char __user *buf, | |
1455 | size_t count, loff_t *offset) | |
1456 | { | |
496ad9aa | 1457 | struct inode *inode = file_inode(file); |
43ae34cb IM |
1458 | struct task_struct *p; |
1459 | ||
43ae34cb IM |
1460 | p = get_proc_task(inode); |
1461 | if (!p) | |
1462 | return -ESRCH; | |
1463 | proc_sched_set_task(p); | |
1464 | ||
1465 | put_task_struct(p); | |
1466 | ||
1467 | return count; | |
1468 | } | |
1469 | ||
1470 | static int sched_open(struct inode *inode, struct file *filp) | |
1471 | { | |
c6a34058 | 1472 | return single_open(filp, sched_show, inode); |
43ae34cb IM |
1473 | } |
1474 | ||
1475 | static const struct file_operations proc_pid_sched_operations = { | |
1476 | .open = sched_open, | |
1477 | .read = seq_read, | |
1478 | .write = sched_write, | |
1479 | .llseek = seq_lseek, | |
5ea473a1 | 1480 | .release = single_release, |
43ae34cb IM |
1481 | }; |
1482 | ||
1483 | #endif | |
1484 | ||
5091faa4 MG |
1485 | #ifdef CONFIG_SCHED_AUTOGROUP |
1486 | /* | |
1487 | * Print out autogroup related information: | |
1488 | */ | |
1489 | static int sched_autogroup_show(struct seq_file *m, void *v) | |
1490 | { | |
1491 | struct inode *inode = m->private; | |
1492 | struct task_struct *p; | |
1493 | ||
1494 | p = get_proc_task(inode); | |
1495 | if (!p) | |
1496 | return -ESRCH; | |
1497 | proc_sched_autogroup_show_task(p, m); | |
1498 | ||
1499 | put_task_struct(p); | |
1500 | ||
1501 | return 0; | |
1502 | } | |
1503 | ||
1504 | static ssize_t | |
1505 | sched_autogroup_write(struct file *file, const char __user *buf, | |
1506 | size_t count, loff_t *offset) | |
1507 | { | |
496ad9aa | 1508 | struct inode *inode = file_inode(file); |
5091faa4 | 1509 | struct task_struct *p; |
860a2e7f | 1510 | char buffer[PROC_NUMBUF] = {}; |
0a8cb8e3 | 1511 | int nice; |
5091faa4 MG |
1512 | int err; |
1513 | ||
5091faa4 MG |
1514 | if (count > sizeof(buffer) - 1) |
1515 | count = sizeof(buffer) - 1; | |
1516 | if (copy_from_user(buffer, buf, count)) | |
1517 | return -EFAULT; | |
1518 | ||
0a8cb8e3 AD |
1519 | err = kstrtoint(strstrip(buffer), 0, &nice); |
1520 | if (err < 0) | |
1521 | return err; | |
5091faa4 MG |
1522 | |
1523 | p = get_proc_task(inode); | |
1524 | if (!p) | |
1525 | return -ESRCH; | |
1526 | ||
2e5b5b3a | 1527 | err = proc_sched_autogroup_set_nice(p, nice); |
5091faa4 MG |
1528 | if (err) |
1529 | count = err; | |
1530 | ||
1531 | put_task_struct(p); | |
1532 | ||
1533 | return count; | |
1534 | } | |
1535 | ||
1536 | static int sched_autogroup_open(struct inode *inode, struct file *filp) | |
1537 | { | |
1538 | int ret; | |
1539 | ||
1540 | ret = single_open(filp, sched_autogroup_show, NULL); | |
1541 | if (!ret) { | |
1542 | struct seq_file *m = filp->private_data; | |
1543 | ||
1544 | m->private = inode; | |
1545 | } | |
1546 | return ret; | |
1547 | } | |
1548 | ||
1549 | static const struct file_operations proc_pid_sched_autogroup_operations = { | |
1550 | .open = sched_autogroup_open, | |
1551 | .read = seq_read, | |
1552 | .write = sched_autogroup_write, | |
1553 | .llseek = seq_lseek, | |
1554 | .release = single_release, | |
1555 | }; | |
1556 | ||
1557 | #endif /* CONFIG_SCHED_AUTOGROUP */ | |
1558 | ||
04a8682a AV |
1559 | #ifdef CONFIG_TIME_NS |
1560 | static int timens_offsets_show(struct seq_file *m, void *v) | |
1561 | { | |
1562 | struct task_struct *p; | |
1563 | ||
1564 | p = get_proc_task(file_inode(m->file)); | |
1565 | if (!p) | |
1566 | return -ESRCH; | |
1567 | proc_timens_show_offsets(p, m); | |
1568 | ||
1569 | put_task_struct(p); | |
1570 | ||
1571 | return 0; | |
1572 | } | |
1573 | ||
1574 | static ssize_t timens_offsets_write(struct file *file, const char __user *buf, | |
1575 | size_t count, loff_t *ppos) | |
1576 | { | |
1577 | struct inode *inode = file_inode(file); | |
1578 | struct proc_timens_offset offsets[2]; | |
1579 | char *kbuf = NULL, *pos, *next_line; | |
1580 | struct task_struct *p; | |
1581 | int ret, noffsets; | |
1582 | ||
1583 | /* Only allow < page size writes at the beginning of the file */ | |
1584 | if ((*ppos != 0) || (count >= PAGE_SIZE)) | |
1585 | return -EINVAL; | |
1586 | ||
1587 | /* Slurp in the user data */ | |
1588 | kbuf = memdup_user_nul(buf, count); | |
1589 | if (IS_ERR(kbuf)) | |
1590 | return PTR_ERR(kbuf); | |
1591 | ||
1592 | /* Parse the user data */ | |
1593 | ret = -EINVAL; | |
1594 | noffsets = 0; | |
1595 | for (pos = kbuf; pos; pos = next_line) { | |
1596 | struct proc_timens_offset *off = &offsets[noffsets]; | |
94d440d6 | 1597 | char clock[10]; |
04a8682a AV |
1598 | int err; |
1599 | ||
1600 | /* Find the end of line and ensure we don't look past it */ | |
1601 | next_line = strchr(pos, '\n'); | |
1602 | if (next_line) { | |
1603 | *next_line = '\0'; | |
1604 | next_line++; | |
1605 | if (*next_line == '\0') | |
1606 | next_line = NULL; | |
1607 | } | |
1608 | ||
94d440d6 | 1609 | err = sscanf(pos, "%9s %lld %lu", clock, |
04a8682a AV |
1610 | &off->val.tv_sec, &off->val.tv_nsec); |
1611 | if (err != 3 || off->val.tv_nsec >= NSEC_PER_SEC) | |
1612 | goto out; | |
94d440d6 AV |
1613 | |
1614 | clock[sizeof(clock) - 1] = 0; | |
1615 | if (strcmp(clock, "monotonic") == 0 || | |
1616 | strcmp(clock, __stringify(CLOCK_MONOTONIC)) == 0) | |
1617 | off->clockid = CLOCK_MONOTONIC; | |
1618 | else if (strcmp(clock, "boottime") == 0 || | |
1619 | strcmp(clock, __stringify(CLOCK_BOOTTIME)) == 0) | |
1620 | off->clockid = CLOCK_BOOTTIME; | |
1621 | else | |
1622 | goto out; | |
1623 | ||
04a8682a AV |
1624 | noffsets++; |
1625 | if (noffsets == ARRAY_SIZE(offsets)) { | |
1626 | if (next_line) | |
1627 | count = next_line - kbuf; | |
1628 | break; | |
1629 | } | |
1630 | } | |
1631 | ||
1632 | ret = -ESRCH; | |
1633 | p = get_proc_task(inode); | |
1634 | if (!p) | |
1635 | goto out; | |
1636 | ret = proc_timens_set_offset(file, p, offsets, noffsets); | |
1637 | put_task_struct(p); | |
1638 | if (ret) | |
1639 | goto out; | |
1640 | ||
1641 | ret = count; | |
1642 | out: | |
1643 | kfree(kbuf); | |
1644 | return ret; | |
1645 | } | |
1646 | ||
1647 | static int timens_offsets_open(struct inode *inode, struct file *filp) | |
1648 | { | |
1649 | return single_open(filp, timens_offsets_show, inode); | |
1650 | } | |
1651 | ||
1652 | static const struct file_operations proc_timens_offsets_operations = { | |
1653 | .open = timens_offsets_open, | |
1654 | .read = seq_read, | |
1655 | .write = timens_offsets_write, | |
1656 | .llseek = seq_lseek, | |
1657 | .release = single_release, | |
1658 | }; | |
1659 | #endif /* CONFIG_TIME_NS */ | |
1660 | ||
4614a696 | 1661 | static ssize_t comm_write(struct file *file, const char __user *buf, |
1662 | size_t count, loff_t *offset) | |
1663 | { | |
496ad9aa | 1664 | struct inode *inode = file_inode(file); |
4614a696 | 1665 | struct task_struct *p; |
860a2e7f | 1666 | char buffer[TASK_COMM_LEN] = {}; |
830e0fc9 | 1667 | const size_t maxlen = sizeof(buffer) - 1; |
4614a696 | 1668 | |
830e0fc9 | 1669 | if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count)) |
4614a696 | 1670 | return -EFAULT; |
1671 | ||
1672 | p = get_proc_task(inode); | |
1673 | if (!p) | |
1674 | return -ESRCH; | |
1675 | ||
c2f273eb | 1676 | if (same_thread_group(current, p)) { |
4614a696 | 1677 | set_task_comm(p, buffer); |
c2f273eb OK |
1678 | proc_comm_connector(p); |
1679 | } | |
4614a696 | 1680 | else |
1681 | count = -EINVAL; | |
1682 | ||
1683 | put_task_struct(p); | |
1684 | ||
1685 | return count; | |
1686 | } | |
1687 | ||
1688 | static int comm_show(struct seq_file *m, void *v) | |
1689 | { | |
1690 | struct inode *inode = m->private; | |
1691 | struct task_struct *p; | |
1692 | ||
1693 | p = get_proc_task(inode); | |
1694 | if (!p) | |
1695 | return -ESRCH; | |
1696 | ||
88b72b31 TH |
1697 | proc_task_name(m, p, false); |
1698 | seq_putc(m, '\n'); | |
4614a696 | 1699 | |
1700 | put_task_struct(p); | |
1701 | ||
1702 | return 0; | |
1703 | } | |
1704 | ||
1705 | static int comm_open(struct inode *inode, struct file *filp) | |
1706 | { | |
c6a34058 | 1707 | return single_open(filp, comm_show, inode); |
4614a696 | 1708 | } |
1709 | ||
1710 | static const struct file_operations proc_pid_set_comm_operations = { | |
1711 | .open = comm_open, | |
1712 | .read = seq_read, | |
1713 | .write = comm_write, | |
1714 | .llseek = seq_lseek, | |
1715 | .release = single_release, | |
1716 | }; | |
1717 | ||
7773fbc5 | 1718 | static int proc_exe_link(struct dentry *dentry, struct path *exe_path) |
925d1c40 MH |
1719 | { |
1720 | struct task_struct *task; | |
925d1c40 MH |
1721 | struct file *exe_file; |
1722 | ||
2b0143b5 | 1723 | task = get_proc_task(d_inode(dentry)); |
925d1c40 MH |
1724 | if (!task) |
1725 | return -ENOENT; | |
cd81a917 | 1726 | exe_file = get_task_exe_file(task); |
925d1c40 | 1727 | put_task_struct(task); |
925d1c40 MH |
1728 | if (exe_file) { |
1729 | *exe_path = exe_file->f_path; | |
1730 | path_get(&exe_file->f_path); | |
1731 | fput(exe_file); | |
1732 | return 0; | |
1733 | } else | |
1734 | return -ENOENT; | |
1735 | } | |
1736 | ||
6b255391 | 1737 | static const char *proc_pid_get_link(struct dentry *dentry, |
fceef393 AV |
1738 | struct inode *inode, |
1739 | struct delayed_call *done) | |
1da177e4 | 1740 | { |
408ef013 | 1741 | struct path path; |
1da177e4 LT |
1742 | int error = -EACCES; |
1743 | ||
6b255391 AV |
1744 | if (!dentry) |
1745 | return ERR_PTR(-ECHILD); | |
1746 | ||
778c1144 EB |
1747 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1748 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1749 | goto out; |
1da177e4 | 1750 | |
408ef013 CH |
1751 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1752 | if (error) | |
1753 | goto out; | |
1754 | ||
1bc82070 | 1755 | error = nd_jump_link(&path); |
1da177e4 | 1756 | out: |
008b150a | 1757 | return ERR_PTR(error); |
1da177e4 LT |
1758 | } |
1759 | ||
5b09c9fe | 1760 | static int do_proc_readlink(const struct path *path, char __user *buffer, int buflen) |
1da177e4 | 1761 | { |
3a72917c | 1762 | char *tmp = kmalloc(PATH_MAX, GFP_KERNEL); |
3dcd25f3 | 1763 | char *pathname; |
1da177e4 LT |
1764 | int len; |
1765 | ||
1766 | if (!tmp) | |
1767 | return -ENOMEM; | |
0c28f287 | 1768 | |
3a72917c | 1769 | pathname = d_path(path, tmp, PATH_MAX); |
3dcd25f3 JB |
1770 | len = PTR_ERR(pathname); |
1771 | if (IS_ERR(pathname)) | |
1da177e4 | 1772 | goto out; |
3a72917c | 1773 | len = tmp + PATH_MAX - 1 - pathname; |
1da177e4 LT |
1774 | |
1775 | if (len > buflen) | |
1776 | len = buflen; | |
3dcd25f3 | 1777 | if (copy_to_user(buffer, pathname, len)) |
1da177e4 LT |
1778 | len = -EFAULT; |
1779 | out: | |
3a72917c | 1780 | kfree(tmp); |
1da177e4 LT |
1781 | return len; |
1782 | } | |
1783 | ||
1784 | static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen) | |
1785 | { | |
1786 | int error = -EACCES; | |
2b0143b5 | 1787 | struct inode *inode = d_inode(dentry); |
3dcd25f3 | 1788 | struct path path; |
1da177e4 | 1789 | |
778c1144 EB |
1790 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1791 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1792 | goto out; |
1da177e4 | 1793 | |
7773fbc5 | 1794 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1da177e4 LT |
1795 | if (error) |
1796 | goto out; | |
1797 | ||
3dcd25f3 JB |
1798 | error = do_proc_readlink(&path, buffer, buflen); |
1799 | path_put(&path); | |
1da177e4 | 1800 | out: |
1da177e4 LT |
1801 | return error; |
1802 | } | |
1803 | ||
faf60af1 | 1804 | const struct inode_operations proc_pid_link_inode_operations = { |
1da177e4 | 1805 | .readlink = proc_pid_readlink, |
6b255391 | 1806 | .get_link = proc_pid_get_link, |
6d76fa58 | 1807 | .setattr = proc_setattr, |
1da177e4 LT |
1808 | }; |
1809 | ||
28a6d671 EB |
1810 | |
1811 | /* building an inode */ | |
1812 | ||
c6eb50d2 | 1813 | void task_dump_owner(struct task_struct *task, umode_t mode, |
68eb94f1 EB |
1814 | kuid_t *ruid, kgid_t *rgid) |
1815 | { | |
1816 | /* Depending on the state of dumpable compute who should own a | |
1817 | * proc file for a task. | |
1818 | */ | |
1819 | const struct cred *cred; | |
1820 | kuid_t uid; | |
1821 | kgid_t gid; | |
1822 | ||
2e0ad552 AD |
1823 | if (unlikely(task->flags & PF_KTHREAD)) { |
1824 | *ruid = GLOBAL_ROOT_UID; | |
1825 | *rgid = GLOBAL_ROOT_GID; | |
1826 | return; | |
1827 | } | |
1828 | ||
68eb94f1 EB |
1829 | /* Default to the tasks effective ownership */ |
1830 | rcu_read_lock(); | |
1831 | cred = __task_cred(task); | |
1832 | uid = cred->euid; | |
1833 | gid = cred->egid; | |
1834 | rcu_read_unlock(); | |
1835 | ||
1836 | /* | |
1837 | * Before the /proc/pid/status file was created the only way to read | |
1838 | * the effective uid of a /process was to stat /proc/pid. Reading | |
1839 | * /proc/pid/status is slow enough that procps and other packages | |
1840 | * kept stating /proc/pid. To keep the rules in /proc simple I have | |
1841 | * made this apply to all per process world readable and executable | |
1842 | * directories. | |
1843 | */ | |
1844 | if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) { | |
1845 | struct mm_struct *mm; | |
1846 | task_lock(task); | |
1847 | mm = task->mm; | |
1848 | /* Make non-dumpable tasks owned by some root */ | |
1849 | if (mm) { | |
1850 | if (get_dumpable(mm) != SUID_DUMP_USER) { | |
1851 | struct user_namespace *user_ns = mm->user_ns; | |
1852 | ||
1853 | uid = make_kuid(user_ns, 0); | |
1854 | if (!uid_valid(uid)) | |
1855 | uid = GLOBAL_ROOT_UID; | |
1856 | ||
1857 | gid = make_kgid(user_ns, 0); | |
1858 | if (!gid_valid(gid)) | |
1859 | gid = GLOBAL_ROOT_GID; | |
1860 | } | |
1861 | } else { | |
1862 | uid = GLOBAL_ROOT_UID; | |
1863 | gid = GLOBAL_ROOT_GID; | |
1864 | } | |
1865 | task_unlock(task); | |
1866 | } | |
1867 | *ruid = uid; | |
1868 | *rgid = gid; | |
1869 | } | |
1870 | ||
7bc3e6e5 EB |
1871 | void proc_pid_evict_inode(struct proc_inode *ei) |
1872 | { | |
1873 | struct pid *pid = ei->pid; | |
1874 | ||
1875 | if (S_ISDIR(ei->vfs_inode.i_mode)) { | |
63f818f4 | 1876 | spin_lock(&pid->lock); |
7bc3e6e5 | 1877 | hlist_del_init_rcu(&ei->sibling_inodes); |
63f818f4 | 1878 | spin_unlock(&pid->lock); |
7bc3e6e5 EB |
1879 | } |
1880 | ||
1881 | put_pid(pid); | |
1882 | } | |
1883 | ||
d919a1e7 | 1884 | struct inode *proc_pid_make_inode(struct super_block *sb, |
db978da8 | 1885 | struct task_struct *task, umode_t mode) |
28a6d671 EB |
1886 | { |
1887 | struct inode * inode; | |
1888 | struct proc_inode *ei; | |
7bc3e6e5 | 1889 | struct pid *pid; |
1da177e4 | 1890 | |
28a6d671 | 1891 | /* We need a new inode */ |
1da177e4 | 1892 | |
28a6d671 EB |
1893 | inode = new_inode(sb); |
1894 | if (!inode) | |
1895 | goto out; | |
1896 | ||
1897 | /* Common stuff */ | |
1898 | ei = PROC_I(inode); | |
db978da8 | 1899 | inode->i_mode = mode; |
85fe4025 | 1900 | inode->i_ino = get_next_ino(); |
200d9421 | 1901 | simple_inode_init_ts(inode); |
28a6d671 EB |
1902 | inode->i_op = &proc_def_inode_operations; |
1903 | ||
1904 | /* | |
1905 | * grab the reference to task. | |
1906 | */ | |
7bc3e6e5 EB |
1907 | pid = get_task_pid(task, PIDTYPE_PID); |
1908 | if (!pid) | |
28a6d671 EB |
1909 | goto out_unlock; |
1910 | ||
7bc3e6e5 EB |
1911 | /* Let the pid remember us for quick removal */ |
1912 | ei->pid = pid; | |
7bc3e6e5 | 1913 | |
68eb94f1 | 1914 | task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid); |
28a6d671 EB |
1915 | security_task_to_inode(task, inode); |
1916 | ||
1da177e4 | 1917 | out: |
28a6d671 EB |
1918 | return inode; |
1919 | ||
1920 | out_unlock: | |
1921 | iput(inode); | |
1922 | return NULL; | |
1da177e4 LT |
1923 | } |
1924 | ||
d919a1e7 ZC |
1925 | /* |
1926 | * Generating an inode and adding it into @pid->inodes, so that task will | |
1927 | * invalidate inode's dentry before being released. | |
1928 | * | |
1929 | * This helper is used for creating dir-type entries under '/proc' and | |
1930 | * '/proc/<tgid>/task'. Other entries(eg. fd, stat) under '/proc/<tgid>' | |
1931 | * can be released by invalidating '/proc/<tgid>' dentry. | |
1932 | * In theory, dentries under '/proc/<tgid>/task' can also be released by | |
1933 | * invalidating '/proc/<tgid>' dentry, we reserve it to handle single | |
1934 | * thread exiting situation: Any one of threads should invalidate its | |
1935 | * '/proc/<tgid>/task/<pid>' dentry before released. | |
1936 | */ | |
1937 | static struct inode *proc_pid_make_base_inode(struct super_block *sb, | |
1938 | struct task_struct *task, umode_t mode) | |
1939 | { | |
1940 | struct inode *inode; | |
1941 | struct proc_inode *ei; | |
1942 | struct pid *pid; | |
1943 | ||
1944 | inode = proc_pid_make_inode(sb, task, mode); | |
1945 | if (!inode) | |
1946 | return NULL; | |
1947 | ||
1948 | /* Let proc_flush_pid find this directory inode */ | |
1949 | ei = PROC_I(inode); | |
1950 | pid = ei->pid; | |
1951 | spin_lock(&pid->lock); | |
1952 | hlist_add_head_rcu(&ei->sibling_inodes, &pid->inodes); | |
1953 | spin_unlock(&pid->lock); | |
1954 | ||
1955 | return inode; | |
1956 | } | |
1957 | ||
b74d24f7 | 1958 | int pid_getattr(struct mnt_idmap *idmap, const struct path *path, |
549c7297 | 1959 | struct kstat *stat, u32 request_mask, unsigned int query_flags) |
1da177e4 | 1960 | { |
a528d35e | 1961 | struct inode *inode = d_inode(path->dentry); |
fa10fed3 | 1962 | struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb); |
28a6d671 | 1963 | struct task_struct *task; |
c69e8d9c | 1964 | |
0d72b928 | 1965 | generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat); |
1da177e4 | 1966 | |
dcb0f222 EB |
1967 | stat->uid = GLOBAL_ROOT_UID; |
1968 | stat->gid = GLOBAL_ROOT_GID; | |
94116929 | 1969 | rcu_read_lock(); |
28a6d671 EB |
1970 | task = pid_task(proc_pid(inode), PIDTYPE_PID); |
1971 | if (task) { | |
fa10fed3 | 1972 | if (!has_pid_permissions(fs_info, task, HIDEPID_INVISIBLE)) { |
0499680a VK |
1973 | rcu_read_unlock(); |
1974 | /* | |
1975 | * This doesn't prevent learning whether PID exists, | |
1976 | * it only makes getattr() consistent with readdir(). | |
1977 | */ | |
1978 | return -ENOENT; | |
1979 | } | |
68eb94f1 | 1980 | task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid); |
1da177e4 | 1981 | } |
28a6d671 | 1982 | rcu_read_unlock(); |
d6e71144 | 1983 | return 0; |
1da177e4 LT |
1984 | } |
1985 | ||
1da177e4 LT |
1986 | /* dentry stuff */ |
1987 | ||
1988 | /* | |
1bbc5513 AV |
1989 | * Set <pid>/... inode ownership (can change due to setuid(), etc.) |
1990 | */ | |
1991 | void pid_update_inode(struct task_struct *task, struct inode *inode) | |
1992 | { | |
1993 | task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid); | |
1994 | ||
1995 | inode->i_mode &= ~(S_ISUID | S_ISGID); | |
1996 | security_task_to_inode(task, inode); | |
1997 | } | |
1998 | ||
1999 | /* | |
1da177e4 LT |
2000 | * Rewrite the inode's ownerships here because the owning task may have |
2001 | * performed a setuid(), etc. | |
99f89551 | 2002 | * |
1da177e4 | 2003 | */ |
1bbc5513 | 2004 | static int pid_revalidate(struct dentry *dentry, unsigned int flags) |
1da177e4 | 2005 | { |
34286d66 NP |
2006 | struct inode *inode; |
2007 | struct task_struct *task; | |
da4d6b9c | 2008 | int ret = 0; |
c69e8d9c | 2009 | |
da4d6b9c SB |
2010 | rcu_read_lock(); |
2011 | inode = d_inode_rcu(dentry); | |
2012 | if (!inode) | |
2013 | goto out; | |
2014 | task = pid_task(proc_pid(inode), PIDTYPE_PID); | |
34286d66 | 2015 | |
99f89551 | 2016 | if (task) { |
1bbc5513 | 2017 | pid_update_inode(task, inode); |
da4d6b9c | 2018 | ret = 1; |
1da177e4 | 2019 | } |
da4d6b9c SB |
2020 | out: |
2021 | rcu_read_unlock(); | |
2022 | return ret; | |
1da177e4 LT |
2023 | } |
2024 | ||
d855a4b7 ON |
2025 | static inline bool proc_inode_is_dead(struct inode *inode) |
2026 | { | |
2027 | return !proc_pid(inode)->tasks[PIDTYPE_PID].first; | |
2028 | } | |
2029 | ||
1dd704b6 DH |
2030 | int pid_delete_dentry(const struct dentry *dentry) |
2031 | { | |
2032 | /* Is the task we represent dead? | |
2033 | * If so, then don't put the dentry on the lru list, | |
2034 | * kill it immediately. | |
2035 | */ | |
2b0143b5 | 2036 | return proc_inode_is_dead(d_inode(dentry)); |
1dd704b6 DH |
2037 | } |
2038 | ||
6b4e306a | 2039 | const struct dentry_operations pid_dentry_operations = |
28a6d671 EB |
2040 | { |
2041 | .d_revalidate = pid_revalidate, | |
2042 | .d_delete = pid_delete_dentry, | |
2043 | }; | |
2044 | ||
2045 | /* Lookups */ | |
2046 | ||
1c0d04c9 EB |
2047 | /* |
2048 | * Fill a directory entry. | |
2049 | * | |
2050 | * If possible create the dcache entry and derive our inode number and | |
2051 | * file type from dcache entry. | |
2052 | * | |
2053 | * Since all of the proc inode numbers are dynamically generated, the inode | |
d2928e85 | 2054 | * numbers do not exist until the inode is cache. This means creating |
1c0d04c9 EB |
2055 | * the dcache entry in readdir is necessary to keep the inode numbers |
2056 | * reported by readdir in sync with the inode numbers reported | |
2057 | * by stat. | |
2058 | */ | |
f0c3b509 | 2059 | bool proc_fill_cache(struct file *file, struct dir_context *ctx, |
a4ef3895 | 2060 | const char *name, unsigned int len, |
c5141e6d | 2061 | instantiate_t instantiate, struct task_struct *task, const void *ptr) |
61a28784 | 2062 | { |
f0c3b509 | 2063 | struct dentry *child, *dir = file->f_path.dentry; |
1df98b8b | 2064 | struct qstr qname = QSTR_INIT(name, len); |
61a28784 | 2065 | struct inode *inode; |
0168b9e3 AV |
2066 | unsigned type = DT_UNKNOWN; |
2067 | ino_t ino = 1; | |
61a28784 | 2068 | |
1df98b8b | 2069 | child = d_hash_and_lookup(dir, &qname); |
61a28784 | 2070 | if (!child) { |
3781764b AV |
2071 | DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq); |
2072 | child = d_alloc_parallel(dir, &qname, &wq); | |
2073 | if (IS_ERR(child)) | |
1df98b8b | 2074 | goto end_instantiate; |
3781764b | 2075 | if (d_in_lookup(child)) { |
0168b9e3 AV |
2076 | struct dentry *res; |
2077 | res = instantiate(child, task, ptr); | |
3781764b | 2078 | d_lookup_done(child); |
0168b9e3 AV |
2079 | if (unlikely(res)) { |
2080 | dput(child); | |
2081 | child = res; | |
d85b399b AV |
2082 | if (IS_ERR(child)) |
2083 | goto end_instantiate; | |
3781764b | 2084 | } |
61a28784 EB |
2085 | } |
2086 | } | |
2b0143b5 | 2087 | inode = d_inode(child); |
147ce699 AV |
2088 | ino = inode->i_ino; |
2089 | type = inode->i_mode >> 12; | |
61a28784 | 2090 | dput(child); |
d85b399b | 2091 | end_instantiate: |
f0c3b509 | 2092 | return dir_emit(ctx, name, len, ino, type); |
61a28784 EB |
2093 | } |
2094 | ||
640708a2 PE |
2095 | /* |
2096 | * dname_to_vma_addr - maps a dentry name into two unsigned longs | |
2097 | * which represent vma start and end addresses. | |
2098 | */ | |
2099 | static int dname_to_vma_addr(struct dentry *dentry, | |
2100 | unsigned long *start, unsigned long *end) | |
2101 | { | |
ac7f1061 AD |
2102 | const char *str = dentry->d_name.name; |
2103 | unsigned long long sval, eval; | |
2104 | unsigned int len; | |
2105 | ||
35318db5 AD |
2106 | if (str[0] == '0' && str[1] != '-') |
2107 | return -EINVAL; | |
ac7f1061 AD |
2108 | len = _parse_integer(str, 16, &sval); |
2109 | if (len & KSTRTOX_OVERFLOW) | |
2110 | return -EINVAL; | |
2111 | if (sval != (unsigned long)sval) | |
2112 | return -EINVAL; | |
2113 | str += len; | |
2114 | ||
2115 | if (*str != '-') | |
640708a2 | 2116 | return -EINVAL; |
ac7f1061 AD |
2117 | str++; |
2118 | ||
35318db5 AD |
2119 | if (str[0] == '0' && str[1]) |
2120 | return -EINVAL; | |
ac7f1061 AD |
2121 | len = _parse_integer(str, 16, &eval); |
2122 | if (len & KSTRTOX_OVERFLOW) | |
2123 | return -EINVAL; | |
2124 | if (eval != (unsigned long)eval) | |
2125 | return -EINVAL; | |
2126 | str += len; | |
2127 | ||
2128 | if (*str != '\0') | |
2129 | return -EINVAL; | |
2130 | ||
2131 | *start = sval; | |
2132 | *end = eval; | |
640708a2 PE |
2133 | |
2134 | return 0; | |
2135 | } | |
2136 | ||
0b728e19 | 2137 | static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags) |
640708a2 PE |
2138 | { |
2139 | unsigned long vm_start, vm_end; | |
2140 | bool exact_vma_exists = false; | |
2141 | struct mm_struct *mm = NULL; | |
2142 | struct task_struct *task; | |
640708a2 PE |
2143 | struct inode *inode; |
2144 | int status = 0; | |
2145 | ||
0b728e19 | 2146 | if (flags & LOOKUP_RCU) |
640708a2 PE |
2147 | return -ECHILD; |
2148 | ||
2b0143b5 | 2149 | inode = d_inode(dentry); |
640708a2 PE |
2150 | task = get_proc_task(inode); |
2151 | if (!task) | |
2152 | goto out_notask; | |
2153 | ||
caaee623 | 2154 | mm = mm_access(task, PTRACE_MODE_READ_FSCREDS); |
2344bec7 | 2155 | if (IS_ERR_OR_NULL(mm)) |
640708a2 PE |
2156 | goto out; |
2157 | ||
2158 | if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) { | |
d8ed45c5 | 2159 | status = mmap_read_lock_killable(mm); |
cd9e2bb8 KK |
2160 | if (!status) { |
2161 | exact_vma_exists = !!find_exact_vma(mm, vm_start, | |
2162 | vm_end); | |
d8ed45c5 | 2163 | mmap_read_unlock(mm); |
cd9e2bb8 | 2164 | } |
640708a2 PE |
2165 | } |
2166 | ||
2167 | mmput(mm); | |
2168 | ||
2169 | if (exact_vma_exists) { | |
68eb94f1 EB |
2170 | task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid); |
2171 | ||
640708a2 PE |
2172 | security_task_to_inode(task, inode); |
2173 | status = 1; | |
2174 | } | |
2175 | ||
2176 | out: | |
2177 | put_task_struct(task); | |
2178 | ||
2179 | out_notask: | |
640708a2 PE |
2180 | return status; |
2181 | } | |
2182 | ||
2183 | static const struct dentry_operations tid_map_files_dentry_operations = { | |
2184 | .d_revalidate = map_files_d_revalidate, | |
2185 | .d_delete = pid_delete_dentry, | |
2186 | }; | |
2187 | ||
6b255391 | 2188 | static int map_files_get_link(struct dentry *dentry, struct path *path) |
640708a2 PE |
2189 | { |
2190 | unsigned long vm_start, vm_end; | |
2191 | struct vm_area_struct *vma; | |
2192 | struct task_struct *task; | |
2193 | struct mm_struct *mm; | |
2194 | int rc; | |
2195 | ||
2196 | rc = -ENOENT; | |
2b0143b5 | 2197 | task = get_proc_task(d_inode(dentry)); |
640708a2 PE |
2198 | if (!task) |
2199 | goto out; | |
2200 | ||
2201 | mm = get_task_mm(task); | |
2202 | put_task_struct(task); | |
2203 | if (!mm) | |
2204 | goto out; | |
2205 | ||
2206 | rc = dname_to_vma_addr(dentry, &vm_start, &vm_end); | |
2207 | if (rc) | |
2208 | goto out_mmput; | |
2209 | ||
d8ed45c5 | 2210 | rc = mmap_read_lock_killable(mm); |
cd9e2bb8 KK |
2211 | if (rc) |
2212 | goto out_mmput; | |
2213 | ||
70335abb | 2214 | rc = -ENOENT; |
640708a2 PE |
2215 | vma = find_exact_vma(mm, vm_start, vm_end); |
2216 | if (vma && vma->vm_file) { | |
08582d67 | 2217 | *path = *file_user_path(vma->vm_file); |
640708a2 PE |
2218 | path_get(path); |
2219 | rc = 0; | |
2220 | } | |
d8ed45c5 | 2221 | mmap_read_unlock(mm); |
640708a2 PE |
2222 | |
2223 | out_mmput: | |
2224 | mmput(mm); | |
2225 | out: | |
2226 | return rc; | |
2227 | } | |
2228 | ||
2229 | struct map_files_info { | |
20d28cde AD |
2230 | unsigned long start; |
2231 | unsigned long end; | |
7b540d06 | 2232 | fmode_t mode; |
640708a2 PE |
2233 | }; |
2234 | ||
bdb4d100 | 2235 | /* |
12886f8a AR |
2236 | * Only allow CAP_SYS_ADMIN and CAP_CHECKPOINT_RESTORE to follow the links, due |
2237 | * to concerns about how the symlinks may be used to bypass permissions on | |
2238 | * ancestor directories in the path to the file in question. | |
bdb4d100 CO |
2239 | */ |
2240 | static const char * | |
6b255391 | 2241 | proc_map_files_get_link(struct dentry *dentry, |
fceef393 AV |
2242 | struct inode *inode, |
2243 | struct delayed_call *done) | |
bdb4d100 | 2244 | { |
12886f8a | 2245 | if (!checkpoint_restore_ns_capable(&init_user_ns)) |
bdb4d100 CO |
2246 | return ERR_PTR(-EPERM); |
2247 | ||
fceef393 | 2248 | return proc_pid_get_link(dentry, inode, done); |
bdb4d100 CO |
2249 | } |
2250 | ||
2251 | /* | |
6b255391 | 2252 | * Identical to proc_pid_link_inode_operations except for get_link() |
bdb4d100 CO |
2253 | */ |
2254 | static const struct inode_operations proc_map_files_link_inode_operations = { | |
2255 | .readlink = proc_pid_readlink, | |
6b255391 | 2256 | .get_link = proc_map_files_get_link, |
bdb4d100 CO |
2257 | .setattr = proc_setattr, |
2258 | }; | |
2259 | ||
0168b9e3 AV |
2260 | static struct dentry * |
2261 | proc_map_files_instantiate(struct dentry *dentry, | |
640708a2 PE |
2262 | struct task_struct *task, const void *ptr) |
2263 | { | |
7b540d06 | 2264 | fmode_t mode = (fmode_t)(unsigned long)ptr; |
640708a2 PE |
2265 | struct proc_inode *ei; |
2266 | struct inode *inode; | |
2267 | ||
0168b9e3 | 2268 | inode = proc_pid_make_inode(dentry->d_sb, task, S_IFLNK | |
db978da8 AG |
2269 | ((mode & FMODE_READ ) ? S_IRUSR : 0) | |
2270 | ((mode & FMODE_WRITE) ? S_IWUSR : 0)); | |
640708a2 | 2271 | if (!inode) |
0168b9e3 | 2272 | return ERR_PTR(-ENOENT); |
640708a2 PE |
2273 | |
2274 | ei = PROC_I(inode); | |
6b255391 | 2275 | ei->op.proc_get_link = map_files_get_link; |
640708a2 | 2276 | |
bdb4d100 | 2277 | inode->i_op = &proc_map_files_link_inode_operations; |
640708a2 | 2278 | inode->i_size = 64; |
640708a2 PE |
2279 | |
2280 | d_set_d_op(dentry, &tid_map_files_dentry_operations); | |
0168b9e3 | 2281 | return d_splice_alias(inode, dentry); |
640708a2 PE |
2282 | } |
2283 | ||
2284 | static struct dentry *proc_map_files_lookup(struct inode *dir, | |
00cd8dd3 | 2285 | struct dentry *dentry, unsigned int flags) |
640708a2 PE |
2286 | { |
2287 | unsigned long vm_start, vm_end; | |
2288 | struct vm_area_struct *vma; | |
2289 | struct task_struct *task; | |
0168b9e3 | 2290 | struct dentry *result; |
640708a2 PE |
2291 | struct mm_struct *mm; |
2292 | ||
0168b9e3 | 2293 | result = ERR_PTR(-ENOENT); |
640708a2 PE |
2294 | task = get_proc_task(dir); |
2295 | if (!task) | |
2296 | goto out; | |
2297 | ||
0168b9e3 | 2298 | result = ERR_PTR(-EACCES); |
caaee623 | 2299 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
640708a2 PE |
2300 | goto out_put_task; |
2301 | ||
0168b9e3 | 2302 | result = ERR_PTR(-ENOENT); |
640708a2 | 2303 | if (dname_to_vma_addr(dentry, &vm_start, &vm_end)) |
eb94cd96 | 2304 | goto out_put_task; |
640708a2 PE |
2305 | |
2306 | mm = get_task_mm(task); | |
2307 | if (!mm) | |
eb94cd96 | 2308 | goto out_put_task; |
640708a2 | 2309 | |
cd9e2bb8 | 2310 | result = ERR_PTR(-EINTR); |
d8ed45c5 | 2311 | if (mmap_read_lock_killable(mm)) |
cd9e2bb8 KK |
2312 | goto out_put_mm; |
2313 | ||
2314 | result = ERR_PTR(-ENOENT); | |
640708a2 PE |
2315 | vma = find_exact_vma(mm, vm_start, vm_end); |
2316 | if (!vma) | |
2317 | goto out_no_vma; | |
2318 | ||
05f56484 | 2319 | if (vma->vm_file) |
0168b9e3 | 2320 | result = proc_map_files_instantiate(dentry, task, |
05f56484 | 2321 | (void *)(unsigned long)vma->vm_file->f_mode); |
640708a2 PE |
2322 | |
2323 | out_no_vma: | |
d8ed45c5 | 2324 | mmap_read_unlock(mm); |
cd9e2bb8 | 2325 | out_put_mm: |
640708a2 | 2326 | mmput(mm); |
640708a2 PE |
2327 | out_put_task: |
2328 | put_task_struct(task); | |
2329 | out: | |
0168b9e3 | 2330 | return result; |
640708a2 PE |
2331 | } |
2332 | ||
2333 | static const struct inode_operations proc_map_files_inode_operations = { | |
2334 | .lookup = proc_map_files_lookup, | |
2335 | .permission = proc_fd_permission, | |
2336 | .setattr = proc_setattr, | |
2337 | }; | |
2338 | ||
2339 | static int | |
f0c3b509 | 2340 | proc_map_files_readdir(struct file *file, struct dir_context *ctx) |
640708a2 | 2341 | { |
640708a2 PE |
2342 | struct vm_area_struct *vma; |
2343 | struct task_struct *task; | |
2344 | struct mm_struct *mm; | |
f0c3b509 | 2345 | unsigned long nr_files, pos, i; |
94f8f3b0 | 2346 | GENRADIX(struct map_files_info) fa; |
f0c3b509 | 2347 | struct map_files_info *p; |
640708a2 | 2348 | int ret; |
5f14b924 | 2349 | struct vma_iterator vmi; |
640708a2 | 2350 | |
94f8f3b0 KO |
2351 | genradix_init(&fa); |
2352 | ||
640708a2 | 2353 | ret = -ENOENT; |
f0c3b509 | 2354 | task = get_proc_task(file_inode(file)); |
640708a2 PE |
2355 | if (!task) |
2356 | goto out; | |
2357 | ||
2358 | ret = -EACCES; | |
caaee623 | 2359 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
640708a2 PE |
2360 | goto out_put_task; |
2361 | ||
2362 | ret = 0; | |
f0c3b509 AV |
2363 | if (!dir_emit_dots(file, ctx)) |
2364 | goto out_put_task; | |
640708a2 | 2365 | |
f0c3b509 AV |
2366 | mm = get_task_mm(task); |
2367 | if (!mm) | |
2368 | goto out_put_task; | |
cd9e2bb8 | 2369 | |
89154dd5 | 2370 | ret = mmap_read_lock_killable(mm); |
cd9e2bb8 KK |
2371 | if (ret) { |
2372 | mmput(mm); | |
2373 | goto out_put_task; | |
2374 | } | |
640708a2 | 2375 | |
f0c3b509 | 2376 | nr_files = 0; |
640708a2 | 2377 | |
f0c3b509 AV |
2378 | /* |
2379 | * We need two passes here: | |
2380 | * | |
c1e8d7c6 ML |
2381 | * 1) Collect vmas of mapped files with mmap_lock taken |
2382 | * 2) Release mmap_lock and instantiate entries | |
f0c3b509 AV |
2383 | * |
2384 | * otherwise we get lockdep complained, since filldir() | |
c1e8d7c6 | 2385 | * routine might require mmap_lock taken in might_fault(). |
f0c3b509 | 2386 | */ |
640708a2 | 2387 | |
5f14b924 LH |
2388 | pos = 2; |
2389 | vma_iter_init(&vmi, mm, 0); | |
2390 | for_each_vma(vmi, vma) { | |
94f8f3b0 KO |
2391 | if (!vma->vm_file) |
2392 | continue; | |
2393 | if (++pos <= ctx->pos) | |
2394 | continue; | |
f0c3b509 | 2395 | |
94f8f3b0 KO |
2396 | p = genradix_ptr_alloc(&fa, nr_files++, GFP_KERNEL); |
2397 | if (!p) { | |
f0c3b509 | 2398 | ret = -ENOMEM; |
89154dd5 | 2399 | mmap_read_unlock(mm); |
f0c3b509 AV |
2400 | mmput(mm); |
2401 | goto out_put_task; | |
640708a2 | 2402 | } |
f0c3b509 | 2403 | |
94f8f3b0 KO |
2404 | p->start = vma->vm_start; |
2405 | p->end = vma->vm_end; | |
2406 | p->mode = vma->vm_file->f_mode; | |
640708a2 | 2407 | } |
89154dd5 | 2408 | mmap_read_unlock(mm); |
fe079a5e | 2409 | mmput(mm); |
f0c3b509 AV |
2410 | |
2411 | for (i = 0; i < nr_files; i++) { | |
20d28cde AD |
2412 | char buf[4 * sizeof(long) + 2]; /* max: %lx-%lx\0 */ |
2413 | unsigned int len; | |
2414 | ||
94f8f3b0 | 2415 | p = genradix_ptr(&fa, i); |
20d28cde | 2416 | len = snprintf(buf, sizeof(buf), "%lx-%lx", p->start, p->end); |
f0c3b509 | 2417 | if (!proc_fill_cache(file, ctx, |
20d28cde | 2418 | buf, len, |
f0c3b509 AV |
2419 | proc_map_files_instantiate, |
2420 | task, | |
2421 | (void *)(unsigned long)p->mode)) | |
2422 | break; | |
2423 | ctx->pos++; | |
640708a2 PE |
2424 | } |
2425 | ||
640708a2 PE |
2426 | out_put_task: |
2427 | put_task_struct(task); | |
2428 | out: | |
94f8f3b0 | 2429 | genradix_free(&fa); |
640708a2 PE |
2430 | return ret; |
2431 | } | |
2432 | ||
2433 | static const struct file_operations proc_map_files_operations = { | |
2434 | .read = generic_read_dir, | |
f50752ea AV |
2435 | .iterate_shared = proc_map_files_readdir, |
2436 | .llseek = generic_file_llseek, | |
640708a2 PE |
2437 | }; |
2438 | ||
b18b6a9c | 2439 | #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS) |
48f6a7a5 PE |
2440 | struct timers_private { |
2441 | struct pid *pid; | |
2442 | struct task_struct *task; | |
2443 | struct sighand_struct *sighand; | |
57b8015e | 2444 | struct pid_namespace *ns; |
48f6a7a5 PE |
2445 | unsigned long flags; |
2446 | }; | |
2447 | ||
2448 | static void *timers_start(struct seq_file *m, loff_t *pos) | |
2449 | { | |
2450 | struct timers_private *tp = m->private; | |
2451 | ||
2452 | tp->task = get_pid_task(tp->pid, PIDTYPE_PID); | |
2453 | if (!tp->task) | |
2454 | return ERR_PTR(-ESRCH); | |
2455 | ||
2456 | tp->sighand = lock_task_sighand(tp->task, &tp->flags); | |
2457 | if (!tp->sighand) | |
2458 | return ERR_PTR(-ESRCH); | |
2459 | ||
2460 | return seq_list_start(&tp->task->signal->posix_timers, *pos); | |
2461 | } | |
2462 | ||
2463 | static void *timers_next(struct seq_file *m, void *v, loff_t *pos) | |
2464 | { | |
2465 | struct timers_private *tp = m->private; | |
2466 | return seq_list_next(v, &tp->task->signal->posix_timers, pos); | |
2467 | } | |
2468 | ||
2469 | static void timers_stop(struct seq_file *m, void *v) | |
2470 | { | |
2471 | struct timers_private *tp = m->private; | |
2472 | ||
2473 | if (tp->sighand) { | |
2474 | unlock_task_sighand(tp->task, &tp->flags); | |
2475 | tp->sighand = NULL; | |
2476 | } | |
2477 | ||
2478 | if (tp->task) { | |
2479 | put_task_struct(tp->task); | |
2480 | tp->task = NULL; | |
2481 | } | |
2482 | } | |
2483 | ||
2484 | static int show_timer(struct seq_file *m, void *v) | |
2485 | { | |
2486 | struct k_itimer *timer; | |
57b8015e PE |
2487 | struct timers_private *tp = m->private; |
2488 | int notify; | |
cedbccab | 2489 | static const char * const nstr[] = { |
57b8015e PE |
2490 | [SIGEV_SIGNAL] = "signal", |
2491 | [SIGEV_NONE] = "none", | |
2492 | [SIGEV_THREAD] = "thread", | |
2493 | }; | |
48f6a7a5 PE |
2494 | |
2495 | timer = list_entry((struct list_head *)v, struct k_itimer, list); | |
57b8015e PE |
2496 | notify = timer->it_sigev_notify; |
2497 | ||
48f6a7a5 | 2498 | seq_printf(m, "ID: %d\n", timer->it_id); |
ba3edf1f | 2499 | seq_printf(m, "signal: %d/%px\n", |
25ce3191 JP |
2500 | timer->sigq->info.si_signo, |
2501 | timer->sigq->info.si_value.sival_ptr); | |
57b8015e | 2502 | seq_printf(m, "notify: %s/%s.%d\n", |
25ce3191 JP |
2503 | nstr[notify & ~SIGEV_THREAD_ID], |
2504 | (notify & SIGEV_THREAD_ID) ? "tid" : "pid", | |
2505 | pid_nr_ns(timer->it_pid, tp->ns)); | |
15ef0298 | 2506 | seq_printf(m, "ClockID: %d\n", timer->it_clock); |
48f6a7a5 PE |
2507 | |
2508 | return 0; | |
2509 | } | |
2510 | ||
2511 | static const struct seq_operations proc_timers_seq_ops = { | |
2512 | .start = timers_start, | |
2513 | .next = timers_next, | |
2514 | .stop = timers_stop, | |
2515 | .show = show_timer, | |
2516 | }; | |
2517 | ||
2518 | static int proc_timers_open(struct inode *inode, struct file *file) | |
2519 | { | |
2520 | struct timers_private *tp; | |
2521 | ||
2522 | tp = __seq_open_private(file, &proc_timers_seq_ops, | |
2523 | sizeof(struct timers_private)); | |
2524 | if (!tp) | |
2525 | return -ENOMEM; | |
2526 | ||
2527 | tp->pid = proc_pid(inode); | |
9d78edea | 2528 | tp->ns = proc_pid_ns(inode->i_sb); |
48f6a7a5 PE |
2529 | return 0; |
2530 | } | |
2531 | ||
2532 | static const struct file_operations proc_timers_operations = { | |
2533 | .open = proc_timers_open, | |
2534 | .read = seq_read, | |
2535 | .llseek = seq_lseek, | |
2536 | .release = seq_release_private, | |
2537 | }; | |
b5946bea | 2538 | #endif |
640708a2 | 2539 | |
5de23d43 JS |
2540 | static ssize_t timerslack_ns_write(struct file *file, const char __user *buf, |
2541 | size_t count, loff_t *offset) | |
2542 | { | |
2543 | struct inode *inode = file_inode(file); | |
2544 | struct task_struct *p; | |
2545 | u64 slack_ns; | |
2546 | int err; | |
2547 | ||
2548 | err = kstrtoull_from_user(buf, count, 10, &slack_ns); | |
2549 | if (err < 0) | |
2550 | return err; | |
2551 | ||
2552 | p = get_proc_task(inode); | |
2553 | if (!p) | |
2554 | return -ESRCH; | |
2555 | ||
4b2bd5fe | 2556 | if (p != current) { |
8da0b4f6 BG |
2557 | rcu_read_lock(); |
2558 | if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) { | |
2559 | rcu_read_unlock(); | |
4b2bd5fe JS |
2560 | count = -EPERM; |
2561 | goto out; | |
2562 | } | |
8da0b4f6 | 2563 | rcu_read_unlock(); |
5de23d43 | 2564 | |
4b2bd5fe JS |
2565 | err = security_task_setscheduler(p); |
2566 | if (err) { | |
2567 | count = err; | |
2568 | goto out; | |
2569 | } | |
904763e1 JS |
2570 | } |
2571 | ||
7abbaf94 JS |
2572 | task_lock(p); |
2573 | if (slack_ns == 0) | |
2574 | p->timer_slack_ns = p->default_timer_slack_ns; | |
2575 | else | |
2576 | p->timer_slack_ns = slack_ns; | |
2577 | task_unlock(p); | |
2578 | ||
2579 | out: | |
5de23d43 JS |
2580 | put_task_struct(p); |
2581 | ||
2582 | return count; | |
2583 | } | |
2584 | ||
2585 | static int timerslack_ns_show(struct seq_file *m, void *v) | |
2586 | { | |
2587 | struct inode *inode = m->private; | |
2588 | struct task_struct *p; | |
7abbaf94 | 2589 | int err = 0; |
5de23d43 JS |
2590 | |
2591 | p = get_proc_task(inode); | |
2592 | if (!p) | |
2593 | return -ESRCH; | |
2594 | ||
4b2bd5fe | 2595 | if (p != current) { |
8da0b4f6 BG |
2596 | rcu_read_lock(); |
2597 | if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) { | |
2598 | rcu_read_unlock(); | |
4b2bd5fe JS |
2599 | err = -EPERM; |
2600 | goto out; | |
2601 | } | |
8da0b4f6 BG |
2602 | rcu_read_unlock(); |
2603 | ||
4b2bd5fe JS |
2604 | err = security_task_getscheduler(p); |
2605 | if (err) | |
2606 | goto out; | |
2607 | } | |
904763e1 | 2608 | |
7abbaf94 JS |
2609 | task_lock(p); |
2610 | seq_printf(m, "%llu\n", p->timer_slack_ns); | |
2611 | task_unlock(p); | |
2612 | ||
2613 | out: | |
5de23d43 JS |
2614 | put_task_struct(p); |
2615 | ||
2616 | return err; | |
2617 | } | |
2618 | ||
2619 | static int timerslack_ns_open(struct inode *inode, struct file *filp) | |
2620 | { | |
2621 | return single_open(filp, timerslack_ns_show, inode); | |
2622 | } | |
2623 | ||
2624 | static const struct file_operations proc_pid_set_timerslack_ns_operations = { | |
2625 | .open = timerslack_ns_open, | |
2626 | .read = seq_read, | |
2627 | .write = timerslack_ns_write, | |
2628 | .llseek = seq_lseek, | |
2629 | .release = single_release, | |
2630 | }; | |
2631 | ||
0168b9e3 AV |
2632 | static struct dentry *proc_pident_instantiate(struct dentry *dentry, |
2633 | struct task_struct *task, const void *ptr) | |
444ceed8 | 2634 | { |
c5141e6d | 2635 | const struct pid_entry *p = ptr; |
444ceed8 EB |
2636 | struct inode *inode; |
2637 | struct proc_inode *ei; | |
444ceed8 | 2638 | |
0168b9e3 | 2639 | inode = proc_pid_make_inode(dentry->d_sb, task, p->mode); |
444ceed8 | 2640 | if (!inode) |
0168b9e3 | 2641 | return ERR_PTR(-ENOENT); |
444ceed8 EB |
2642 | |
2643 | ei = PROC_I(inode); | |
444ceed8 | 2644 | if (S_ISDIR(inode->i_mode)) |
bfe86848 | 2645 | set_nlink(inode, 2); /* Use getattr to fix if necessary */ |
444ceed8 EB |
2646 | if (p->iop) |
2647 | inode->i_op = p->iop; | |
2648 | if (p->fop) | |
2649 | inode->i_fop = p->fop; | |
2650 | ei->op = p->op; | |
1bbc5513 | 2651 | pid_update_inode(task, inode); |
fb045adb | 2652 | d_set_d_op(dentry, &pid_dentry_operations); |
0168b9e3 | 2653 | return d_splice_alias(inode, dentry); |
444ceed8 EB |
2654 | } |
2655 | ||
1da177e4 LT |
2656 | static struct dentry *proc_pident_lookup(struct inode *dir, |
2657 | struct dentry *dentry, | |
d5a572a4 AD |
2658 | const struct pid_entry *p, |
2659 | const struct pid_entry *end) | |
1da177e4 | 2660 | { |
99f89551 | 2661 | struct task_struct *task = get_proc_task(dir); |
0168b9e3 | 2662 | struct dentry *res = ERR_PTR(-ENOENT); |
1da177e4 | 2663 | |
99f89551 EB |
2664 | if (!task) |
2665 | goto out_no_task; | |
1da177e4 | 2666 | |
20cdc894 EB |
2667 | /* |
2668 | * Yes, it does not scale. And it should not. Don't add | |
2669 | * new entries into /proc/<tgid>/ without very good reasons. | |
2670 | */ | |
d5a572a4 | 2671 | for (; p < end; p++) { |
1da177e4 LT |
2672 | if (p->len != dentry->d_name.len) |
2673 | continue; | |
26b95137 AD |
2674 | if (!memcmp(dentry->d_name.name, p->name, p->len)) { |
2675 | res = proc_pident_instantiate(dentry, task, p); | |
1da177e4 | 2676 | break; |
26b95137 | 2677 | } |
1da177e4 | 2678 | } |
99f89551 EB |
2679 | put_task_struct(task); |
2680 | out_no_task: | |
0168b9e3 | 2681 | return res; |
1da177e4 LT |
2682 | } |
2683 | ||
f0c3b509 | 2684 | static int proc_pident_readdir(struct file *file, struct dir_context *ctx, |
c5141e6d | 2685 | const struct pid_entry *ents, unsigned int nents) |
28a6d671 | 2686 | { |
f0c3b509 AV |
2687 | struct task_struct *task = get_proc_task(file_inode(file)); |
2688 | const struct pid_entry *p; | |
28a6d671 | 2689 | |
28a6d671 | 2690 | if (!task) |
f0c3b509 | 2691 | return -ENOENT; |
28a6d671 | 2692 | |
f0c3b509 AV |
2693 | if (!dir_emit_dots(file, ctx)) |
2694 | goto out; | |
2695 | ||
2696 | if (ctx->pos >= nents + 2) | |
2697 | goto out; | |
28a6d671 | 2698 | |
bac5f5d5 | 2699 | for (p = ents + (ctx->pos - 2); p < ents + nents; p++) { |
f0c3b509 AV |
2700 | if (!proc_fill_cache(file, ctx, p->name, p->len, |
2701 | proc_pident_instantiate, task, p)) | |
2702 | break; | |
2703 | ctx->pos++; | |
2704 | } | |
28a6d671 | 2705 | out: |
61a28784 | 2706 | put_task_struct(task); |
f0c3b509 | 2707 | return 0; |
1da177e4 LT |
2708 | } |
2709 | ||
28a6d671 | 2710 | #ifdef CONFIG_SECURITY |
591a22c1 KC |
2711 | static int proc_pid_attr_open(struct inode *inode, struct file *file) |
2712 | { | |
94f0b2d4 LT |
2713 | file->private_data = NULL; |
2714 | __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS); | |
2715 | return 0; | |
591a22c1 KC |
2716 | } |
2717 | ||
28a6d671 EB |
2718 | static ssize_t proc_pid_attr_read(struct file * file, char __user * buf, |
2719 | size_t count, loff_t *ppos) | |
2720 | { | |
496ad9aa | 2721 | struct inode * inode = file_inode(file); |
04ff9708 | 2722 | char *p = NULL; |
28a6d671 EB |
2723 | ssize_t length; |
2724 | struct task_struct *task = get_proc_task(inode); | |
2725 | ||
28a6d671 | 2726 | if (!task) |
04ff9708 | 2727 | return -ESRCH; |
28a6d671 | 2728 | |
6d9c939d | 2729 | length = security_getprocattr(task, PROC_I(inode)->op.lsm, |
c8e477c6 | 2730 | file->f_path.dentry->d_name.name, |
04ff9708 | 2731 | &p); |
28a6d671 | 2732 | put_task_struct(task); |
04ff9708 AV |
2733 | if (length > 0) |
2734 | length = simple_read_from_buffer(buf, count, ppos, p, length); | |
2735 | kfree(p); | |
28a6d671 | 2736 | return length; |
1da177e4 LT |
2737 | } |
2738 | ||
28a6d671 EB |
2739 | static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf, |
2740 | size_t count, loff_t *ppos) | |
2741 | { | |
496ad9aa | 2742 | struct inode * inode = file_inode(file); |
41089b6d | 2743 | struct task_struct *task; |
bb646cdb | 2744 | void *page; |
41089b6d | 2745 | int rv; |
b21507e2 | 2746 | |
bfb819ea | 2747 | /* A task may only write when it was the opener. */ |
591a22c1 | 2748 | if (file->private_data != current->mm) |
bfb819ea KC |
2749 | return -EPERM; |
2750 | ||
41089b6d AD |
2751 | rcu_read_lock(); |
2752 | task = pid_task(proc_pid(inode), PIDTYPE_PID); | |
2753 | if (!task) { | |
2754 | rcu_read_unlock(); | |
2755 | return -ESRCH; | |
2756 | } | |
b21507e2 | 2757 | /* A task may only write its own attributes. */ |
41089b6d AD |
2758 | if (current != task) { |
2759 | rcu_read_unlock(); | |
2760 | return -EACCES; | |
2761 | } | |
35a196be PM |
2762 | /* Prevent changes to overridden credentials. */ |
2763 | if (current_cred() != current_real_cred()) { | |
2764 | rcu_read_unlock(); | |
2765 | return -EBUSY; | |
2766 | } | |
41089b6d | 2767 | rcu_read_unlock(); |
b21507e2 | 2768 | |
28a6d671 EB |
2769 | if (count > PAGE_SIZE) |
2770 | count = PAGE_SIZE; | |
2771 | ||
2772 | /* No partial writes. */ | |
28a6d671 | 2773 | if (*ppos != 0) |
41089b6d | 2774 | return -EINVAL; |
28a6d671 | 2775 | |
bb646cdb AV |
2776 | page = memdup_user(buf, count); |
2777 | if (IS_ERR(page)) { | |
41089b6d | 2778 | rv = PTR_ERR(page); |
28a6d671 | 2779 | goto out; |
bb646cdb | 2780 | } |
28a6d671 | 2781 | |
107db7c7 | 2782 | /* Guard against adverse ptrace interaction */ |
41089b6d AD |
2783 | rv = mutex_lock_interruptible(¤t->signal->cred_guard_mutex); |
2784 | if (rv < 0) | |
107db7c7 DH |
2785 | goto out_free; |
2786 | ||
6d9c939d CS |
2787 | rv = security_setprocattr(PROC_I(inode)->op.lsm, |
2788 | file->f_path.dentry->d_name.name, page, | |
2789 | count); | |
b21507e2 | 2790 | mutex_unlock(¤t->signal->cred_guard_mutex); |
28a6d671 | 2791 | out_free: |
bb646cdb | 2792 | kfree(page); |
28a6d671 | 2793 | out: |
41089b6d | 2794 | return rv; |
28a6d671 EB |
2795 | } |
2796 | ||
00977a59 | 2797 | static const struct file_operations proc_pid_attr_operations = { |
591a22c1 | 2798 | .open = proc_pid_attr_open, |
28a6d671 EB |
2799 | .read = proc_pid_attr_read, |
2800 | .write = proc_pid_attr_write, | |
87df8424 | 2801 | .llseek = generic_file_llseek, |
591a22c1 | 2802 | .release = mem_release, |
28a6d671 EB |
2803 | }; |
2804 | ||
6d9c939d CS |
2805 | #define LSM_DIR_OPS(LSM) \ |
2806 | static int proc_##LSM##_attr_dir_iterate(struct file *filp, \ | |
2807 | struct dir_context *ctx) \ | |
2808 | { \ | |
2809 | return proc_pident_readdir(filp, ctx, \ | |
2810 | LSM##_attr_dir_stuff, \ | |
2811 | ARRAY_SIZE(LSM##_attr_dir_stuff)); \ | |
2812 | } \ | |
2813 | \ | |
2814 | static const struct file_operations proc_##LSM##_attr_dir_ops = { \ | |
2815 | .read = generic_read_dir, \ | |
0a2c2baa | 2816 | .iterate_shared = proc_##LSM##_attr_dir_iterate, \ |
6d9c939d CS |
2817 | .llseek = default_llseek, \ |
2818 | }; \ | |
2819 | \ | |
2820 | static struct dentry *proc_##LSM##_attr_dir_lookup(struct inode *dir, \ | |
2821 | struct dentry *dentry, unsigned int flags) \ | |
2822 | { \ | |
2823 | return proc_pident_lookup(dir, dentry, \ | |
2824 | LSM##_attr_dir_stuff, \ | |
d5a572a4 | 2825 | LSM##_attr_dir_stuff + ARRAY_SIZE(LSM##_attr_dir_stuff)); \ |
6d9c939d CS |
2826 | } \ |
2827 | \ | |
2828 | static const struct inode_operations proc_##LSM##_attr_dir_inode_ops = { \ | |
2829 | .lookup = proc_##LSM##_attr_dir_lookup, \ | |
2830 | .getattr = pid_getattr, \ | |
2831 | .setattr = proc_setattr, \ | |
2832 | } | |
2833 | ||
2834 | #ifdef CONFIG_SECURITY_SMACK | |
2835 | static const struct pid_entry smack_attr_dir_stuff[] = { | |
2836 | ATTR("smack", "current", 0666), | |
2837 | }; | |
2838 | LSM_DIR_OPS(smack); | |
2839 | #endif | |
2840 | ||
6413f852 JJ |
2841 | #ifdef CONFIG_SECURITY_APPARMOR |
2842 | static const struct pid_entry apparmor_attr_dir_stuff[] = { | |
2843 | ATTR("apparmor", "current", 0666), | |
2844 | ATTR("apparmor", "prev", 0444), | |
2845 | ATTR("apparmor", "exec", 0666), | |
2846 | }; | |
2847 | LSM_DIR_OPS(apparmor); | |
2848 | #endif | |
2849 | ||
c5141e6d | 2850 | static const struct pid_entry attr_dir_stuff[] = { |
6d9c939d CS |
2851 | ATTR(NULL, "current", 0666), |
2852 | ATTR(NULL, "prev", 0444), | |
2853 | ATTR(NULL, "exec", 0666), | |
2854 | ATTR(NULL, "fscreate", 0666), | |
2855 | ATTR(NULL, "keycreate", 0666), | |
2856 | ATTR(NULL, "sockcreate", 0666), | |
2857 | #ifdef CONFIG_SECURITY_SMACK | |
2858 | DIR("smack", 0555, | |
2859 | proc_smack_attr_dir_inode_ops, proc_smack_attr_dir_ops), | |
2860 | #endif | |
6413f852 JJ |
2861 | #ifdef CONFIG_SECURITY_APPARMOR |
2862 | DIR("apparmor", 0555, | |
2863 | proc_apparmor_attr_dir_inode_ops, proc_apparmor_attr_dir_ops), | |
2864 | #endif | |
28a6d671 EB |
2865 | }; |
2866 | ||
f0c3b509 | 2867 | static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx) |
28a6d671 | 2868 | { |
f0c3b509 AV |
2869 | return proc_pident_readdir(file, ctx, |
2870 | attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2871 | } |
2872 | ||
00977a59 | 2873 | static const struct file_operations proc_attr_dir_operations = { |
1da177e4 | 2874 | .read = generic_read_dir, |
f50752ea AV |
2875 | .iterate_shared = proc_attr_dir_readdir, |
2876 | .llseek = generic_file_llseek, | |
1da177e4 LT |
2877 | }; |
2878 | ||
72d9dcfc | 2879 | static struct dentry *proc_attr_dir_lookup(struct inode *dir, |
00cd8dd3 | 2880 | struct dentry *dentry, unsigned int flags) |
28a6d671 | 2881 | { |
7bcd6b0e | 2882 | return proc_pident_lookup(dir, dentry, |
d5a572a4 AD |
2883 | attr_dir_stuff, |
2884 | attr_dir_stuff + ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2885 | } |
2886 | ||
c5ef1c42 | 2887 | static const struct inode_operations proc_attr_dir_inode_operations = { |
72d9dcfc | 2888 | .lookup = proc_attr_dir_lookup, |
99f89551 | 2889 | .getattr = pid_getattr, |
6d76fa58 | 2890 | .setattr = proc_setattr, |
1da177e4 LT |
2891 | }; |
2892 | ||
28a6d671 EB |
2893 | #endif |
2894 | ||
698ba7b5 | 2895 | #ifdef CONFIG_ELF_CORE |
3cb4a0bb KH |
2896 | static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf, |
2897 | size_t count, loff_t *ppos) | |
2898 | { | |
496ad9aa | 2899 | struct task_struct *task = get_proc_task(file_inode(file)); |
3cb4a0bb KH |
2900 | struct mm_struct *mm; |
2901 | char buffer[PROC_NUMBUF]; | |
2902 | size_t len; | |
2903 | int ret; | |
2904 | ||
2905 | if (!task) | |
2906 | return -ESRCH; | |
2907 | ||
2908 | ret = 0; | |
2909 | mm = get_task_mm(task); | |
2910 | if (mm) { | |
2911 | len = snprintf(buffer, sizeof(buffer), "%08lx\n", | |
2912 | ((mm->flags & MMF_DUMP_FILTER_MASK) >> | |
2913 | MMF_DUMP_FILTER_SHIFT)); | |
2914 | mmput(mm); | |
2915 | ret = simple_read_from_buffer(buf, count, ppos, buffer, len); | |
2916 | } | |
2917 | ||
2918 | put_task_struct(task); | |
2919 | ||
2920 | return ret; | |
2921 | } | |
2922 | ||
2923 | static ssize_t proc_coredump_filter_write(struct file *file, | |
2924 | const char __user *buf, | |
2925 | size_t count, | |
2926 | loff_t *ppos) | |
2927 | { | |
2928 | struct task_struct *task; | |
2929 | struct mm_struct *mm; | |
3cb4a0bb KH |
2930 | unsigned int val; |
2931 | int ret; | |
2932 | int i; | |
2933 | unsigned long mask; | |
2934 | ||
774636e1 AD |
2935 | ret = kstrtouint_from_user(buf, count, 0, &val); |
2936 | if (ret < 0) | |
2937 | return ret; | |
3cb4a0bb KH |
2938 | |
2939 | ret = -ESRCH; | |
496ad9aa | 2940 | task = get_proc_task(file_inode(file)); |
3cb4a0bb KH |
2941 | if (!task) |
2942 | goto out_no_task; | |
2943 | ||
3cb4a0bb KH |
2944 | mm = get_task_mm(task); |
2945 | if (!mm) | |
2946 | goto out_no_mm; | |
41a0c249 | 2947 | ret = 0; |
3cb4a0bb KH |
2948 | |
2949 | for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) { | |
2950 | if (val & mask) | |
2951 | set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2952 | else | |
2953 | clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2954 | } | |
2955 | ||
2956 | mmput(mm); | |
2957 | out_no_mm: | |
2958 | put_task_struct(task); | |
2959 | out_no_task: | |
774636e1 AD |
2960 | if (ret < 0) |
2961 | return ret; | |
2962 | return count; | |
3cb4a0bb KH |
2963 | } |
2964 | ||
2965 | static const struct file_operations proc_coredump_filter_operations = { | |
2966 | .read = proc_coredump_filter_read, | |
2967 | .write = proc_coredump_filter_write, | |
87df8424 | 2968 | .llseek = generic_file_llseek, |
3cb4a0bb KH |
2969 | }; |
2970 | #endif | |
2971 | ||
aba76fdb | 2972 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 2973 | static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole) |
297c5d92 | 2974 | { |
23202220 | 2975 | struct task_io_accounting acct; |
293eb1e7 | 2976 | int result; |
5995477a | 2977 | |
f7cfd871 | 2978 | result = down_read_killable(&task->signal->exec_update_lock); |
293eb1e7 VK |
2979 | if (result) |
2980 | return result; | |
2981 | ||
caaee623 | 2982 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) { |
293eb1e7 VK |
2983 | result = -EACCES; |
2984 | goto out_unlock; | |
2985 | } | |
1d1221f3 | 2986 | |
1df4bd83 | 2987 | if (whole) { |
23202220 ON |
2988 | struct signal_struct *sig = task->signal; |
2989 | struct task_struct *t; | |
1df4bd83 ON |
2990 | unsigned int seq = 1; |
2991 | unsigned long flags; | |
2992 | ||
2993 | rcu_read_lock(); | |
2994 | do { | |
2995 | seq++; /* 2 on the 1st/lockless path, otherwise odd */ | |
2996 | flags = read_seqbegin_or_lock_irqsave(&sig->stats_lock, &seq); | |
5995477a | 2997 | |
1df4bd83 ON |
2998 | acct = sig->ioac; |
2999 | __for_each_thread(sig, t) | |
3000 | task_io_accounting_add(&acct, &t->ioac); | |
5995477a | 3001 | |
1df4bd83 ON |
3002 | } while (need_seqretry(&sig->stats_lock, seq)); |
3003 | done_seqretry_irqrestore(&sig->stats_lock, seq, flags); | |
3004 | rcu_read_unlock(); | |
23202220 ON |
3005 | } else { |
3006 | acct = task->ioac; | |
297c5d92 | 3007 | } |
23202220 | 3008 | |
25ce3191 JP |
3009 | seq_printf(m, |
3010 | "rchar: %llu\n" | |
3011 | "wchar: %llu\n" | |
3012 | "syscr: %llu\n" | |
3013 | "syscw: %llu\n" | |
3014 | "read_bytes: %llu\n" | |
3015 | "write_bytes: %llu\n" | |
3016 | "cancelled_write_bytes: %llu\n", | |
3017 | (unsigned long long)acct.rchar, | |
3018 | (unsigned long long)acct.wchar, | |
3019 | (unsigned long long)acct.syscr, | |
3020 | (unsigned long long)acct.syscw, | |
3021 | (unsigned long long)acct.read_bytes, | |
3022 | (unsigned long long)acct.write_bytes, | |
3023 | (unsigned long long)acct.cancelled_write_bytes); | |
3024 | result = 0; | |
3025 | ||
293eb1e7 | 3026 | out_unlock: |
f7cfd871 | 3027 | up_read(&task->signal->exec_update_lock); |
293eb1e7 | 3028 | return result; |
297c5d92 AR |
3029 | } |
3030 | ||
19aadc98 AD |
3031 | static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns, |
3032 | struct pid *pid, struct task_struct *task) | |
297c5d92 | 3033 | { |
19aadc98 | 3034 | return do_io_accounting(task, m, 0); |
aba76fdb | 3035 | } |
297c5d92 | 3036 | |
19aadc98 AD |
3037 | static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns, |
3038 | struct pid *pid, struct task_struct *task) | |
297c5d92 | 3039 | { |
19aadc98 | 3040 | return do_io_accounting(task, m, 1); |
297c5d92 AR |
3041 | } |
3042 | #endif /* CONFIG_TASK_IO_ACCOUNTING */ | |
aba76fdb | 3043 | |
22d917d8 EB |
3044 | #ifdef CONFIG_USER_NS |
3045 | static int proc_id_map_open(struct inode *inode, struct file *file, | |
ccf94f1b | 3046 | const struct seq_operations *seq_ops) |
22d917d8 EB |
3047 | { |
3048 | struct user_namespace *ns = NULL; | |
3049 | struct task_struct *task; | |
3050 | struct seq_file *seq; | |
3051 | int ret = -EINVAL; | |
3052 | ||
3053 | task = get_proc_task(inode); | |
3054 | if (task) { | |
3055 | rcu_read_lock(); | |
3056 | ns = get_user_ns(task_cred_xxx(task, user_ns)); | |
3057 | rcu_read_unlock(); | |
3058 | put_task_struct(task); | |
3059 | } | |
3060 | if (!ns) | |
3061 | goto err; | |
3062 | ||
3063 | ret = seq_open(file, seq_ops); | |
3064 | if (ret) | |
3065 | goto err_put_ns; | |
3066 | ||
3067 | seq = file->private_data; | |
3068 | seq->private = ns; | |
3069 | ||
3070 | return 0; | |
3071 | err_put_ns: | |
3072 | put_user_ns(ns); | |
3073 | err: | |
3074 | return ret; | |
3075 | } | |
3076 | ||
3077 | static int proc_id_map_release(struct inode *inode, struct file *file) | |
3078 | { | |
3079 | struct seq_file *seq = file->private_data; | |
3080 | struct user_namespace *ns = seq->private; | |
3081 | put_user_ns(ns); | |
3082 | return seq_release(inode, file); | |
3083 | } | |
3084 | ||
3085 | static int proc_uid_map_open(struct inode *inode, struct file *file) | |
3086 | { | |
3087 | return proc_id_map_open(inode, file, &proc_uid_seq_operations); | |
3088 | } | |
3089 | ||
3090 | static int proc_gid_map_open(struct inode *inode, struct file *file) | |
3091 | { | |
3092 | return proc_id_map_open(inode, file, &proc_gid_seq_operations); | |
3093 | } | |
3094 | ||
f76d207a EB |
3095 | static int proc_projid_map_open(struct inode *inode, struct file *file) |
3096 | { | |
3097 | return proc_id_map_open(inode, file, &proc_projid_seq_operations); | |
3098 | } | |
3099 | ||
22d917d8 EB |
3100 | static const struct file_operations proc_uid_map_operations = { |
3101 | .open = proc_uid_map_open, | |
3102 | .write = proc_uid_map_write, | |
3103 | .read = seq_read, | |
3104 | .llseek = seq_lseek, | |
3105 | .release = proc_id_map_release, | |
3106 | }; | |
3107 | ||
3108 | static const struct file_operations proc_gid_map_operations = { | |
3109 | .open = proc_gid_map_open, | |
3110 | .write = proc_gid_map_write, | |
3111 | .read = seq_read, | |
3112 | .llseek = seq_lseek, | |
3113 | .release = proc_id_map_release, | |
3114 | }; | |
f76d207a EB |
3115 | |
3116 | static const struct file_operations proc_projid_map_operations = { | |
3117 | .open = proc_projid_map_open, | |
3118 | .write = proc_projid_map_write, | |
3119 | .read = seq_read, | |
3120 | .llseek = seq_lseek, | |
3121 | .release = proc_id_map_release, | |
3122 | }; | |
9cc46516 EB |
3123 | |
3124 | static int proc_setgroups_open(struct inode *inode, struct file *file) | |
3125 | { | |
3126 | struct user_namespace *ns = NULL; | |
3127 | struct task_struct *task; | |
3128 | int ret; | |
3129 | ||
3130 | ret = -ESRCH; | |
3131 | task = get_proc_task(inode); | |
3132 | if (task) { | |
3133 | rcu_read_lock(); | |
3134 | ns = get_user_ns(task_cred_xxx(task, user_ns)); | |
3135 | rcu_read_unlock(); | |
3136 | put_task_struct(task); | |
3137 | } | |
3138 | if (!ns) | |
3139 | goto err; | |
3140 | ||
3141 | if (file->f_mode & FMODE_WRITE) { | |
3142 | ret = -EACCES; | |
3143 | if (!ns_capable(ns, CAP_SYS_ADMIN)) | |
3144 | goto err_put_ns; | |
3145 | } | |
3146 | ||
3147 | ret = single_open(file, &proc_setgroups_show, ns); | |
3148 | if (ret) | |
3149 | goto err_put_ns; | |
3150 | ||
3151 | return 0; | |
3152 | err_put_ns: | |
3153 | put_user_ns(ns); | |
3154 | err: | |
3155 | return ret; | |
3156 | } | |
3157 | ||
3158 | static int proc_setgroups_release(struct inode *inode, struct file *file) | |
3159 | { | |
3160 | struct seq_file *seq = file->private_data; | |
3161 | struct user_namespace *ns = seq->private; | |
3162 | int ret = single_release(inode, file); | |
3163 | put_user_ns(ns); | |
3164 | return ret; | |
3165 | } | |
3166 | ||
3167 | static const struct file_operations proc_setgroups_operations = { | |
3168 | .open = proc_setgroups_open, | |
3169 | .write = proc_setgroups_write, | |
3170 | .read = seq_read, | |
3171 | .llseek = seq_lseek, | |
3172 | .release = proc_setgroups_release, | |
3173 | }; | |
22d917d8 EB |
3174 | #endif /* CONFIG_USER_NS */ |
3175 | ||
47830723 KC |
3176 | static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns, |
3177 | struct pid *pid, struct task_struct *task) | |
3178 | { | |
a9712bc1 AV |
3179 | int err = lock_trace(task); |
3180 | if (!err) { | |
3181 | seq_printf(m, "%08x\n", task->personality); | |
3182 | unlock_trace(task); | |
3183 | } | |
3184 | return err; | |
47830723 KC |
3185 | } |
3186 | ||
7c23b330 JP |
3187 | #ifdef CONFIG_LIVEPATCH |
3188 | static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns, | |
3189 | struct pid *pid, struct task_struct *task) | |
3190 | { | |
3191 | seq_printf(m, "%d\n", task->patch_state); | |
3192 | return 0; | |
3193 | } | |
3194 | #endif /* CONFIG_LIVEPATCH */ | |
3195 | ||
76093853 | 3196 | #ifdef CONFIG_KSM |
3197 | static int proc_pid_ksm_merging_pages(struct seq_file *m, struct pid_namespace *ns, | |
3198 | struct pid *pid, struct task_struct *task) | |
3199 | { | |
3200 | struct mm_struct *mm; | |
3201 | ||
3202 | mm = get_task_mm(task); | |
3203 | if (mm) { | |
3204 | seq_printf(m, "%lu\n", mm->ksm_merging_pages); | |
3205 | mmput(mm); | |
3206 | } | |
3207 | ||
3208 | return 0; | |
3209 | } | |
cb4df4ca | 3210 | static int proc_pid_ksm_stat(struct seq_file *m, struct pid_namespace *ns, |
3211 | struct pid *pid, struct task_struct *task) | |
3212 | { | |
3213 | struct mm_struct *mm; | |
3214 | ||
3215 | mm = get_task_mm(task); | |
3216 | if (mm) { | |
3217 | seq_printf(m, "ksm_rmap_items %lu\n", mm->ksm_rmap_items); | |
6080d19f | 3218 | seq_printf(m, "ksm_zero_pages %lu\n", mm->ksm_zero_pages); |
d21077fb SR |
3219 | seq_printf(m, "ksm_merging_pages %lu\n", mm->ksm_merging_pages); |
3220 | seq_printf(m, "ksm_process_profit %ld\n", ksm_process_profit(mm)); | |
cb4df4ca | 3221 | mmput(mm); |
3222 | } | |
3223 | ||
3224 | return 0; | |
3225 | } | |
76093853 | 3226 | #endif /* CONFIG_KSM */ |
3227 | ||
c8d12627 AP |
3228 | #ifdef CONFIG_STACKLEAK_METRICS |
3229 | static int proc_stack_depth(struct seq_file *m, struct pid_namespace *ns, | |
3230 | struct pid *pid, struct task_struct *task) | |
3231 | { | |
3232 | unsigned long prev_depth = THREAD_SIZE - | |
3233 | (task->prev_lowest_stack & (THREAD_SIZE - 1)); | |
3234 | unsigned long depth = THREAD_SIZE - | |
3235 | (task->lowest_stack & (THREAD_SIZE - 1)); | |
3236 | ||
3237 | seq_printf(m, "previous stack depth: %lu\nstack depth: %lu\n", | |
3238 | prev_depth, depth); | |
3239 | return 0; | |
3240 | } | |
3241 | #endif /* CONFIG_STACKLEAK_METRICS */ | |
3242 | ||
28a6d671 EB |
3243 | /* |
3244 | * Thread groups | |
3245 | */ | |
00977a59 | 3246 | static const struct file_operations proc_task_operations; |
c5ef1c42 | 3247 | static const struct inode_operations proc_task_inode_operations; |
20cdc894 | 3248 | |
c5141e6d | 3249 | static const struct pid_entry tgid_base_stuff[] = { |
631f9c18 AD |
3250 | DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations), |
3251 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), | |
640708a2 | 3252 | DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations), |
7bc3fa01 | 3253 | DIR("fdinfo", S_IRUGO|S_IXUGO, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 3254 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
b2211a36 | 3255 | #ifdef CONFIG_NET |
631f9c18 | 3256 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), |
b2211a36 | 3257 | #endif |
631f9c18 | 3258 | REG("environ", S_IRUSR, proc_environ_operations), |
c5317167 | 3259 | REG("auxv", S_IRUSR, proc_auxv_operations), |
631f9c18 | 3260 | ONE("status", S_IRUGO, proc_pid_status), |
35a35046 | 3261 | ONE("personality", S_IRUSR, proc_pid_personality), |
1c963eb1 | 3262 | ONE("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 3263 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 3264 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
5091faa4 MG |
3265 | #endif |
3266 | #ifdef CONFIG_SCHED_AUTOGROUP | |
3267 | REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations), | |
04a8682a AV |
3268 | #endif |
3269 | #ifdef CONFIG_TIME_NS | |
3270 | REG("timens_offsets", S_IRUGO|S_IWUSR, proc_timens_offsets_operations), | |
ebcb6734 | 3271 | #endif |
4614a696 | 3272 | REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations), |
ebcb6734 | 3273 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 | 3274 | ONE("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 3275 | #endif |
c2c0bb44 | 3276 | REG("cmdline", S_IRUGO, proc_pid_cmdline_ops), |
631f9c18 AD |
3277 | ONE("stat", S_IRUGO, proc_tgid_stat), |
3278 | ONE("statm", S_IRUGO, proc_pid_statm), | |
b7643757 | 3279 | REG("maps", S_IRUGO, proc_pid_maps_operations), |
28a6d671 | 3280 | #ifdef CONFIG_NUMA |
b7643757 | 3281 | REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations), |
28a6d671 | 3282 | #endif |
631f9c18 AD |
3283 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
3284 | LNK("cwd", proc_cwd_link), | |
3285 | LNK("root", proc_root_link), | |
3286 | LNK("exe", proc_exe_link), | |
3287 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
3288 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
3289 | REG("mountstats", S_IRUSR, proc_mountstats_operations), | |
1e883281 | 3290 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 3291 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
b7643757 | 3292 | REG("smaps", S_IRUGO, proc_pid_smaps_operations), |
493b0e9d | 3293 | REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations), |
32ed74a4 | 3294 | REG("pagemap", S_IRUSR, proc_pagemap_operations), |
28a6d671 EB |
3295 | #endif |
3296 | #ifdef CONFIG_SECURITY | |
631f9c18 | 3297 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
3298 | #endif |
3299 | #ifdef CONFIG_KALLSYMS | |
edfcd606 | 3300 | ONE("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 3301 | #endif |
2ec220e2 | 3302 | #ifdef CONFIG_STACKTRACE |
35a35046 | 3303 | ONE("stack", S_IRUSR, proc_pid_stack), |
28a6d671 | 3304 | #endif |
5968cece | 3305 | #ifdef CONFIG_SCHED_INFO |
f6e826ca | 3306 | ONE("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 3307 | #endif |
9745512c | 3308 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 3309 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 3310 | #endif |
8793d854 | 3311 | #ifdef CONFIG_PROC_PID_CPUSET |
52de4779 | 3312 | ONE("cpuset", S_IRUGO, proc_cpuset_show), |
a424316c PM |
3313 | #endif |
3314 | #ifdef CONFIG_CGROUPS | |
006f4ac4 | 3315 | ONE("cgroup", S_IRUGO, proc_cgroup_show), |
e79f15a4 CY |
3316 | #endif |
3317 | #ifdef CONFIG_PROC_CPU_RESCTRL | |
3318 | ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show), | |
28a6d671 | 3319 | #endif |
6ba51e37 | 3320 | ONE("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 3321 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 3322 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
4b7d248b | 3323 | #ifdef CONFIG_AUDIT |
631f9c18 AD |
3324 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
3325 | REG("sessionid", S_IRUGO, proc_sessionid_operations), | |
28a6d671 | 3326 | #endif |
f4f154fd | 3327 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 3328 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
168c42bc | 3329 | REG("fail-nth", 0644, proc_fail_nth_operations), |
f4f154fd | 3330 | #endif |
698ba7b5 | 3331 | #ifdef CONFIG_ELF_CORE |
631f9c18 | 3332 | REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations), |
3cb4a0bb | 3333 | #endif |
aba76fdb | 3334 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 3335 | ONE("io", S_IRUSR, proc_tgid_io_accounting), |
aba76fdb | 3336 | #endif |
22d917d8 EB |
3337 | #ifdef CONFIG_USER_NS |
3338 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
3339 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 3340 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
9cc46516 | 3341 | REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations), |
22d917d8 | 3342 | #endif |
b18b6a9c | 3343 | #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS) |
48f6a7a5 PE |
3344 | REG("timers", S_IRUGO, proc_timers_operations), |
3345 | #endif | |
5de23d43 | 3346 | REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations), |
7c23b330 JP |
3347 | #ifdef CONFIG_LIVEPATCH |
3348 | ONE("patch_state", S_IRUSR, proc_pid_patch_state), | |
3349 | #endif | |
c8d12627 AP |
3350 | #ifdef CONFIG_STACKLEAK_METRICS |
3351 | ONE("stack_depth", S_IRUGO, proc_stack_depth), | |
3352 | #endif | |
68bc30bb AL |
3353 | #ifdef CONFIG_PROC_PID_ARCH_STATUS |
3354 | ONE("arch_status", S_IRUGO, proc_pid_arch_status), | |
3355 | #endif | |
0d8315dd YZ |
3356 | #ifdef CONFIG_SECCOMP_CACHE_DEBUG |
3357 | ONE("seccomp_cache", S_IRUSR, proc_pid_seccomp_cache), | |
3358 | #endif | |
76093853 | 3359 | #ifdef CONFIG_KSM |
3360 | ONE("ksm_merging_pages", S_IRUSR, proc_pid_ksm_merging_pages), | |
cb4df4ca | 3361 | ONE("ksm_stat", S_IRUSR, proc_pid_ksm_stat), |
76093853 | 3362 | #endif |
28a6d671 | 3363 | }; |
1da177e4 | 3364 | |
f0c3b509 | 3365 | static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3366 | { |
f0c3b509 AV |
3367 | return proc_pident_readdir(file, ctx, |
3368 | tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
3369 | } |
3370 | ||
00977a59 | 3371 | static const struct file_operations proc_tgid_base_operations = { |
1da177e4 | 3372 | .read = generic_read_dir, |
f50752ea AV |
3373 | .iterate_shared = proc_tgid_base_readdir, |
3374 | .llseek = generic_file_llseek, | |
1da177e4 LT |
3375 | }; |
3376 | ||
3eb39f47 CB |
3377 | struct pid *tgid_pidfd_to_pid(const struct file *file) |
3378 | { | |
30d158b1 | 3379 | if (file->f_op != &proc_tgid_base_operations) |
3eb39f47 CB |
3380 | return ERR_PTR(-EBADF); |
3381 | ||
3382 | return proc_pid(file_inode(file)); | |
3383 | } | |
3384 | ||
00cd8dd3 AV |
3385 | static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
3386 | { | |
7bcd6b0e | 3387 | return proc_pident_lookup(dir, dentry, |
d5a572a4 AD |
3388 | tgid_base_stuff, |
3389 | tgid_base_stuff + ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
3390 | } |
3391 | ||
c5ef1c42 | 3392 | static const struct inode_operations proc_tgid_base_inode_operations = { |
28a6d671 | 3393 | .lookup = proc_tgid_base_lookup, |
99f89551 | 3394 | .getattr = pid_getattr, |
6d76fa58 | 3395 | .setattr = proc_setattr, |
0499680a | 3396 | .permission = proc_pid_permission, |
1da177e4 | 3397 | }; |
1da177e4 | 3398 | |
0895e91d | 3399 | /** |
7bc3e6e5 EB |
3400 | * proc_flush_pid - Remove dcache entries for @pid from the /proc dcache. |
3401 | * @pid: pid that should be flushed. | |
0895e91d | 3402 | * |
7bc3e6e5 EB |
3403 | * This function walks a list of inodes (that belong to any proc |
3404 | * filesystem) that are attached to the pid and flushes them from | |
3405 | * the dentry cache. | |
0895e91d RD |
3406 | * |
3407 | * It is safe and reasonable to cache /proc entries for a task until | |
3408 | * that task exits. After that they just clog up the dcache with | |
3409 | * useless entries, possibly causing useful dcache entries to be | |
7bc3e6e5 EB |
3410 | * flushed instead. This routine is provided to flush those useless |
3411 | * dcache entries when a process is reaped. | |
0895e91d RD |
3412 | * |
3413 | * NOTE: This routine is just an optimization so it does not guarantee | |
7bc3e6e5 EB |
3414 | * that no dcache entries will exist after a process is reaped |
3415 | * it just makes it very unlikely that any will persist. | |
60347f67 PE |
3416 | */ |
3417 | ||
7bc3e6e5 | 3418 | void proc_flush_pid(struct pid *pid) |
60347f67 | 3419 | { |
63f818f4 | 3420 | proc_invalidate_siblings_dcache(&pid->inodes, &pid->lock); |
60347f67 PE |
3421 | } |
3422 | ||
0168b9e3 | 3423 | static struct dentry *proc_pid_instantiate(struct dentry * dentry, |
c52a47ac | 3424 | struct task_struct *task, const void *ptr) |
444ceed8 | 3425 | { |
444ceed8 EB |
3426 | struct inode *inode; |
3427 | ||
d919a1e7 ZC |
3428 | inode = proc_pid_make_base_inode(dentry->d_sb, task, |
3429 | S_IFDIR | S_IRUGO | S_IXUGO); | |
444ceed8 | 3430 | if (!inode) |
0168b9e3 | 3431 | return ERR_PTR(-ENOENT); |
444ceed8 | 3432 | |
444ceed8 EB |
3433 | inode->i_op = &proc_tgid_base_inode_operations; |
3434 | inode->i_fop = &proc_tgid_base_operations; | |
3435 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 | 3436 | |
1270dd8d | 3437 | set_nlink(inode, nlink_tgid); |
1bbc5513 | 3438 | pid_update_inode(task, inode); |
444ceed8 | 3439 | |
fb045adb | 3440 | d_set_d_op(dentry, &pid_dentry_operations); |
0168b9e3 | 3441 | return d_splice_alias(inode, dentry); |
444ceed8 EB |
3442 | } |
3443 | ||
867aaccf | 3444 | struct dentry *proc_pid_lookup(struct dentry *dentry, unsigned int flags) |
1da177e4 LT |
3445 | { |
3446 | struct task_struct *task; | |
1da177e4 | 3447 | unsigned tgid; |
fa10fed3 | 3448 | struct proc_fs_info *fs_info; |
b488893a | 3449 | struct pid_namespace *ns; |
0168b9e3 | 3450 | struct dentry *result = ERR_PTR(-ENOENT); |
1da177e4 | 3451 | |
dbcdb504 | 3452 | tgid = name_to_int(&dentry->d_name); |
1da177e4 LT |
3453 | if (tgid == ~0U) |
3454 | goto out; | |
3455 | ||
fa10fed3 AG |
3456 | fs_info = proc_sb_info(dentry->d_sb); |
3457 | ns = fs_info->pid_ns; | |
de758734 | 3458 | rcu_read_lock(); |
b488893a | 3459 | task = find_task_by_pid_ns(tgid, ns); |
1da177e4 LT |
3460 | if (task) |
3461 | get_task_struct(task); | |
de758734 | 3462 | rcu_read_unlock(); |
1da177e4 LT |
3463 | if (!task) |
3464 | goto out; | |
3465 | ||
24a71ce5 AG |
3466 | /* Limit procfs to only ptraceable tasks */ |
3467 | if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE) { | |
3468 | if (!has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS)) | |
3469 | goto out_put_task; | |
3470 | } | |
3471 | ||
0168b9e3 | 3472 | result = proc_pid_instantiate(dentry, task, NULL); |
24a71ce5 | 3473 | out_put_task: |
1da177e4 | 3474 | put_task_struct(task); |
1da177e4 | 3475 | out: |
0168b9e3 | 3476 | return result; |
1da177e4 LT |
3477 | } |
3478 | ||
1da177e4 | 3479 | /* |
0804ef4b | 3480 | * Find the first task with tgid >= tgid |
0bc58a91 | 3481 | * |
1da177e4 | 3482 | */ |
19fd4bb2 EB |
3483 | struct tgid_iter { |
3484 | unsigned int tgid; | |
0804ef4b | 3485 | struct task_struct *task; |
19fd4bb2 EB |
3486 | }; |
3487 | static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter) | |
3488 | { | |
0804ef4b | 3489 | struct pid *pid; |
1da177e4 | 3490 | |
19fd4bb2 EB |
3491 | if (iter.task) |
3492 | put_task_struct(iter.task); | |
454cc105 | 3493 | rcu_read_lock(); |
0804ef4b | 3494 | retry: |
19fd4bb2 EB |
3495 | iter.task = NULL; |
3496 | pid = find_ge_pid(iter.tgid, ns); | |
0804ef4b | 3497 | if (pid) { |
19fd4bb2 | 3498 | iter.tgid = pid_nr_ns(pid, ns); |
3147d8aa EB |
3499 | iter.task = pid_task(pid, PIDTYPE_TGID); |
3500 | if (!iter.task) { | |
19fd4bb2 | 3501 | iter.tgid += 1; |
0804ef4b | 3502 | goto retry; |
19fd4bb2 EB |
3503 | } |
3504 | get_task_struct(iter.task); | |
0bc58a91 | 3505 | } |
454cc105 | 3506 | rcu_read_unlock(); |
19fd4bb2 | 3507 | return iter; |
1da177e4 LT |
3508 | } |
3509 | ||
0097875b | 3510 | #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2) |
0804ef4b | 3511 | |
1da177e4 | 3512 | /* for the /proc/ directory itself, after non-process stuff has been done */ |
f0c3b509 | 3513 | int proc_pid_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3514 | { |
19fd4bb2 | 3515 | struct tgid_iter iter; |
fa10fed3 | 3516 | struct proc_fs_info *fs_info = proc_sb_info(file_inode(file)->i_sb); |
9d78edea | 3517 | struct pid_namespace *ns = proc_pid_ns(file_inode(file)->i_sb); |
f0c3b509 | 3518 | loff_t pos = ctx->pos; |
1da177e4 | 3519 | |
021ada7d | 3520 | if (pos >= PID_MAX_LIMIT + TGID_OFFSET) |
f0c3b509 | 3521 | return 0; |
1da177e4 | 3522 | |
0097875b | 3523 | if (pos == TGID_OFFSET - 2) { |
fa10fed3 | 3524 | struct inode *inode = d_inode(fs_info->proc_self); |
db963164 | 3525 | if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK)) |
f0c3b509 | 3526 | return 0; |
0097875b EB |
3527 | ctx->pos = pos = pos + 1; |
3528 | } | |
3529 | if (pos == TGID_OFFSET - 1) { | |
fa10fed3 | 3530 | struct inode *inode = d_inode(fs_info->proc_thread_self); |
0097875b EB |
3531 | if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK)) |
3532 | return 0; | |
3533 | ctx->pos = pos = pos + 1; | |
021ada7d | 3534 | } |
0097875b | 3535 | iter.tgid = pos - TGID_OFFSET; |
19fd4bb2 | 3536 | iter.task = NULL; |
19fd4bb2 EB |
3537 | for (iter = next_tgid(ns, iter); |
3538 | iter.task; | |
3539 | iter.tgid += 1, iter = next_tgid(ns, iter)) { | |
e3912ac3 | 3540 | char name[10 + 1]; |
a4ef3895 | 3541 | unsigned int len; |
3ba4bcee ED |
3542 | |
3543 | cond_resched(); | |
fa10fed3 | 3544 | if (!has_pid_permissions(fs_info, iter.task, HIDEPID_INVISIBLE)) |
f0c3b509 | 3545 | continue; |
0499680a | 3546 | |
e3912ac3 | 3547 | len = snprintf(name, sizeof(name), "%u", iter.tgid); |
f0c3b509 AV |
3548 | ctx->pos = iter.tgid + TGID_OFFSET; |
3549 | if (!proc_fill_cache(file, ctx, name, len, | |
3550 | proc_pid_instantiate, iter.task, NULL)) { | |
19fd4bb2 | 3551 | put_task_struct(iter.task); |
f0c3b509 | 3552 | return 0; |
1da177e4 | 3553 | } |
0bc58a91 | 3554 | } |
f0c3b509 | 3555 | ctx->pos = PID_MAX_LIMIT + TGID_OFFSET; |
0bc58a91 EB |
3556 | return 0; |
3557 | } | |
1da177e4 | 3558 | |
1b3044e3 JD |
3559 | /* |
3560 | * proc_tid_comm_permission is a special permission function exclusively | |
3561 | * used for the node /proc/<pid>/task/<tid>/comm. | |
3562 | * It bypasses generic permission checks in the case where a task of the same | |
3563 | * task group attempts to access the node. | |
3564 | * The rationale behind this is that glibc and bionic access this node for | |
3565 | * cross thread naming (pthread_set/getname_np(!self)). However, if | |
3566 | * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0, | |
3567 | * which locks out the cross thread naming implementation. | |
3568 | * This function makes sure that the node is always accessible for members of | |
3569 | * same thread group. | |
3570 | */ | |
4609e1f1 | 3571 | static int proc_tid_comm_permission(struct mnt_idmap *idmap, |
549c7297 | 3572 | struct inode *inode, int mask) |
1b3044e3 JD |
3573 | { |
3574 | bool is_same_tgroup; | |
3575 | struct task_struct *task; | |
3576 | ||
3577 | task = get_proc_task(inode); | |
3578 | if (!task) | |
3579 | return -ESRCH; | |
3580 | is_same_tgroup = same_thread_group(current, task); | |
3581 | put_task_struct(task); | |
3582 | ||
3583 | if (likely(is_same_tgroup && !(mask & MAY_EXEC))) { | |
3584 | /* This file (/proc/<pid>/task/<tid>/comm) can always be | |
3585 | * read or written by the members of the corresponding | |
3586 | * thread group. | |
3587 | */ | |
3588 | return 0; | |
3589 | } | |
3590 | ||
4609e1f1 | 3591 | return generic_permission(&nop_mnt_idmap, inode, mask); |
1b3044e3 JD |
3592 | } |
3593 | ||
3594 | static const struct inode_operations proc_tid_comm_inode_operations = { | |
ccf61486 AS |
3595 | .setattr = proc_setattr, |
3596 | .permission = proc_tid_comm_permission, | |
1b3044e3 JD |
3597 | }; |
3598 | ||
28a6d671 EB |
3599 | /* |
3600 | * Tasks | |
3601 | */ | |
c5141e6d | 3602 | static const struct pid_entry tid_base_stuff[] = { |
631f9c18 | 3603 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), |
7bc3fa01 | 3604 | DIR("fdinfo", S_IRUGO|S_IXUGO, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 3605 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
6ba8ed79 EB |
3606 | #ifdef CONFIG_NET |
3607 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), | |
3608 | #endif | |
631f9c18 | 3609 | REG("environ", S_IRUSR, proc_environ_operations), |
c5317167 | 3610 | REG("auxv", S_IRUSR, proc_auxv_operations), |
631f9c18 | 3611 | ONE("status", S_IRUGO, proc_pid_status), |
35a35046 | 3612 | ONE("personality", S_IRUSR, proc_pid_personality), |
1c963eb1 | 3613 | ONE("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 3614 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 3615 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
ebcb6734 | 3616 | #endif |
1b3044e3 JD |
3617 | NOD("comm", S_IFREG|S_IRUGO|S_IWUSR, |
3618 | &proc_tid_comm_inode_operations, | |
3619 | &proc_pid_set_comm_operations, {}), | |
ebcb6734 | 3620 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 | 3621 | ONE("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 3622 | #endif |
c2c0bb44 | 3623 | REG("cmdline", S_IRUGO, proc_pid_cmdline_ops), |
631f9c18 AD |
3624 | ONE("stat", S_IRUGO, proc_tid_stat), |
3625 | ONE("statm", S_IRUGO, proc_pid_statm), | |
871305bb | 3626 | REG("maps", S_IRUGO, proc_pid_maps_operations), |
2e13ba54 | 3627 | #ifdef CONFIG_PROC_CHILDREN |
81841161 CG |
3628 | REG("children", S_IRUGO, proc_tid_children_operations), |
3629 | #endif | |
28a6d671 | 3630 | #ifdef CONFIG_NUMA |
871305bb | 3631 | REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations), |
28a6d671 | 3632 | #endif |
631f9c18 AD |
3633 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
3634 | LNK("cwd", proc_cwd_link), | |
3635 | LNK("root", proc_root_link), | |
3636 | LNK("exe", proc_exe_link), | |
3637 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
3638 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
1e883281 | 3639 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 3640 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
871305bb | 3641 | REG("smaps", S_IRUGO, proc_pid_smaps_operations), |
493b0e9d | 3642 | REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations), |
32ed74a4 | 3643 | REG("pagemap", S_IRUSR, proc_pagemap_operations), |
28a6d671 EB |
3644 | #endif |
3645 | #ifdef CONFIG_SECURITY | |
631f9c18 | 3646 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
3647 | #endif |
3648 | #ifdef CONFIG_KALLSYMS | |
edfcd606 | 3649 | ONE("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 3650 | #endif |
2ec220e2 | 3651 | #ifdef CONFIG_STACKTRACE |
35a35046 | 3652 | ONE("stack", S_IRUSR, proc_pid_stack), |
28a6d671 | 3653 | #endif |
5968cece | 3654 | #ifdef CONFIG_SCHED_INFO |
f6e826ca | 3655 | ONE("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 3656 | #endif |
9745512c | 3657 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 3658 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 3659 | #endif |
8793d854 | 3660 | #ifdef CONFIG_PROC_PID_CPUSET |
52de4779 | 3661 | ONE("cpuset", S_IRUGO, proc_cpuset_show), |
a424316c PM |
3662 | #endif |
3663 | #ifdef CONFIG_CGROUPS | |
006f4ac4 | 3664 | ONE("cgroup", S_IRUGO, proc_cgroup_show), |
e79f15a4 CY |
3665 | #endif |
3666 | #ifdef CONFIG_PROC_CPU_RESCTRL | |
3667 | ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show), | |
28a6d671 | 3668 | #endif |
6ba51e37 | 3669 | ONE("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 3670 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 3671 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
4b7d248b | 3672 | #ifdef CONFIG_AUDIT |
631f9c18 | 3673 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
26ec3c64 | 3674 | REG("sessionid", S_IRUGO, proc_sessionid_operations), |
28a6d671 | 3675 | #endif |
f4f154fd | 3676 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 3677 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
1203c8e6 | 3678 | REG("fail-nth", 0644, proc_fail_nth_operations), |
f4f154fd | 3679 | #endif |
297c5d92 | 3680 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 3681 | ONE("io", S_IRUSR, proc_tid_io_accounting), |
297c5d92 | 3682 | #endif |
22d917d8 EB |
3683 | #ifdef CONFIG_USER_NS |
3684 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
3685 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 3686 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
9cc46516 | 3687 | REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations), |
22d917d8 | 3688 | #endif |
7c23b330 JP |
3689 | #ifdef CONFIG_LIVEPATCH |
3690 | ONE("patch_state", S_IRUSR, proc_pid_patch_state), | |
3691 | #endif | |
68bc30bb AL |
3692 | #ifdef CONFIG_PROC_PID_ARCH_STATUS |
3693 | ONE("arch_status", S_IRUGO, proc_pid_arch_status), | |
3694 | #endif | |
0d8315dd YZ |
3695 | #ifdef CONFIG_SECCOMP_CACHE_DEBUG |
3696 | ONE("seccomp_cache", S_IRUSR, proc_pid_seccomp_cache), | |
3697 | #endif | |
76093853 | 3698 | #ifdef CONFIG_KSM |
3699 | ONE("ksm_merging_pages", S_IRUSR, proc_pid_ksm_merging_pages), | |
cb4df4ca | 3700 | ONE("ksm_stat", S_IRUSR, proc_pid_ksm_stat), |
76093853 | 3701 | #endif |
28a6d671 EB |
3702 | }; |
3703 | ||
f0c3b509 | 3704 | static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx) |
28a6d671 | 3705 | { |
f0c3b509 AV |
3706 | return proc_pident_readdir(file, ctx, |
3707 | tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
3708 | } |
3709 | ||
00cd8dd3 AV |
3710 | static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
3711 | { | |
7bcd6b0e | 3712 | return proc_pident_lookup(dir, dentry, |
d5a572a4 AD |
3713 | tid_base_stuff, |
3714 | tid_base_stuff + ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
3715 | } |
3716 | ||
00977a59 | 3717 | static const struct file_operations proc_tid_base_operations = { |
28a6d671 | 3718 | .read = generic_read_dir, |
f50752ea AV |
3719 | .iterate_shared = proc_tid_base_readdir, |
3720 | .llseek = generic_file_llseek, | |
28a6d671 EB |
3721 | }; |
3722 | ||
c5ef1c42 | 3723 | static const struct inode_operations proc_tid_base_inode_operations = { |
28a6d671 EB |
3724 | .lookup = proc_tid_base_lookup, |
3725 | .getattr = pid_getattr, | |
3726 | .setattr = proc_setattr, | |
3727 | }; | |
3728 | ||
0168b9e3 AV |
3729 | static struct dentry *proc_task_instantiate(struct dentry *dentry, |
3730 | struct task_struct *task, const void *ptr) | |
444ceed8 | 3731 | { |
444ceed8 | 3732 | struct inode *inode; |
d919a1e7 ZC |
3733 | inode = proc_pid_make_base_inode(dentry->d_sb, task, |
3734 | S_IFDIR | S_IRUGO | S_IXUGO); | |
444ceed8 | 3735 | if (!inode) |
0168b9e3 | 3736 | return ERR_PTR(-ENOENT); |
1bbc5513 | 3737 | |
444ceed8 EB |
3738 | inode->i_op = &proc_tid_base_inode_operations; |
3739 | inode->i_fop = &proc_tid_base_operations; | |
1bbc5513 | 3740 | inode->i_flags |= S_IMMUTABLE; |
aed54175 | 3741 | |
1270dd8d | 3742 | set_nlink(inode, nlink_tid); |
1bbc5513 | 3743 | pid_update_inode(task, inode); |
444ceed8 | 3744 | |
fb045adb | 3745 | d_set_d_op(dentry, &pid_dentry_operations); |
0168b9e3 | 3746 | return d_splice_alias(inode, dentry); |
444ceed8 EB |
3747 | } |
3748 | ||
00cd8dd3 | 3749 | static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags) |
28a6d671 | 3750 | { |
28a6d671 EB |
3751 | struct task_struct *task; |
3752 | struct task_struct *leader = get_proc_task(dir); | |
28a6d671 | 3753 | unsigned tid; |
fa10fed3 | 3754 | struct proc_fs_info *fs_info; |
b488893a | 3755 | struct pid_namespace *ns; |
0168b9e3 | 3756 | struct dentry *result = ERR_PTR(-ENOENT); |
28a6d671 EB |
3757 | |
3758 | if (!leader) | |
3759 | goto out_no_task; | |
3760 | ||
dbcdb504 | 3761 | tid = name_to_int(&dentry->d_name); |
28a6d671 EB |
3762 | if (tid == ~0U) |
3763 | goto out; | |
3764 | ||
fa10fed3 AG |
3765 | fs_info = proc_sb_info(dentry->d_sb); |
3766 | ns = fs_info->pid_ns; | |
28a6d671 | 3767 | rcu_read_lock(); |
b488893a | 3768 | task = find_task_by_pid_ns(tid, ns); |
28a6d671 EB |
3769 | if (task) |
3770 | get_task_struct(task); | |
3771 | rcu_read_unlock(); | |
3772 | if (!task) | |
3773 | goto out; | |
bac0abd6 | 3774 | if (!same_thread_group(leader, task)) |
28a6d671 EB |
3775 | goto out_drop_task; |
3776 | ||
0168b9e3 | 3777 | result = proc_task_instantiate(dentry, task, NULL); |
28a6d671 EB |
3778 | out_drop_task: |
3779 | put_task_struct(task); | |
3780 | out: | |
3781 | put_task_struct(leader); | |
3782 | out_no_task: | |
0168b9e3 | 3783 | return result; |
28a6d671 EB |
3784 | } |
3785 | ||
0bc58a91 EB |
3786 | /* |
3787 | * Find the first tid of a thread group to return to user space. | |
3788 | * | |
3789 | * Usually this is just the thread group leader, but if the users | |
3790 | * buffer was too small or there was a seek into the middle of the | |
3791 | * directory we have more work todo. | |
3792 | * | |
3793 | * In the case of a short read we start with find_task_by_pid. | |
3794 | * | |
3795 | * In the case of a seek we start with the leader and walk nr | |
3796 | * threads past it. | |
3797 | */ | |
9f6e963f ON |
3798 | static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos, |
3799 | struct pid_namespace *ns) | |
0bc58a91 | 3800 | { |
d855a4b7 | 3801 | struct task_struct *pos, *task; |
9f6e963f ON |
3802 | unsigned long nr = f_pos; |
3803 | ||
3804 | if (nr != f_pos) /* 32bit overflow? */ | |
3805 | return NULL; | |
1da177e4 | 3806 | |
cc288738 | 3807 | rcu_read_lock(); |
d855a4b7 ON |
3808 | task = pid_task(pid, PIDTYPE_PID); |
3809 | if (!task) | |
3810 | goto fail; | |
3811 | ||
3812 | /* Attempt to start with the tid of a thread */ | |
9f6e963f | 3813 | if (tid && nr) { |
b488893a | 3814 | pos = find_task_by_pid_ns(tid, ns); |
d855a4b7 | 3815 | if (pos && same_thread_group(pos, task)) |
a872ff0c | 3816 | goto found; |
0bc58a91 | 3817 | } |
1da177e4 | 3818 | |
0bc58a91 | 3819 | /* If nr exceeds the number of threads there is nothing todo */ |
9f6e963f | 3820 | if (nr >= get_nr_threads(task)) |
c986c14a | 3821 | goto fail; |
1da177e4 | 3822 | |
a872ff0c ON |
3823 | /* If we haven't found our starting place yet start |
3824 | * with the leader and walk nr threads forward. | |
0bc58a91 | 3825 | */ |
dce8f8ed | 3826 | for_each_thread(task, pos) { |
9f6e963f | 3827 | if (!nr--) |
c986c14a | 3828 | goto found; |
63993102 | 3829 | } |
c986c14a ON |
3830 | fail: |
3831 | pos = NULL; | |
3832 | goto out; | |
a872ff0c ON |
3833 | found: |
3834 | get_task_struct(pos); | |
3835 | out: | |
cc288738 | 3836 | rcu_read_unlock(); |
0bc58a91 EB |
3837 | return pos; |
3838 | } | |
3839 | ||
3840 | /* | |
3841 | * Find the next thread in the thread list. | |
3842 | * Return NULL if there is an error or no next thread. | |
3843 | * | |
3844 | * The reference to the input task_struct is released. | |
3845 | */ | |
3846 | static struct task_struct *next_tid(struct task_struct *start) | |
3847 | { | |
c1df7fb8 | 3848 | struct task_struct *pos = NULL; |
cc288738 | 3849 | rcu_read_lock(); |
c1df7fb8 | 3850 | if (pid_alive(start)) { |
33a98138 ON |
3851 | pos = __next_thread(start); |
3852 | if (pos) | |
c1df7fb8 ON |
3853 | get_task_struct(pos); |
3854 | } | |
cc288738 | 3855 | rcu_read_unlock(); |
0bc58a91 EB |
3856 | put_task_struct(start); |
3857 | return pos; | |
1da177e4 LT |
3858 | } |
3859 | ||
3860 | /* for the /proc/TGID/task/ directories */ | |
f0c3b509 | 3861 | static int proc_task_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3862 | { |
d855a4b7 ON |
3863 | struct inode *inode = file_inode(file); |
3864 | struct task_struct *task; | |
b488893a | 3865 | struct pid_namespace *ns; |
f0c3b509 | 3866 | int tid; |
1da177e4 | 3867 | |
d855a4b7 | 3868 | if (proc_inode_is_dead(inode)) |
f0c3b509 | 3869 | return -ENOENT; |
1da177e4 | 3870 | |
f0c3b509 | 3871 | if (!dir_emit_dots(file, ctx)) |
d855a4b7 | 3872 | return 0; |
1da177e4 | 3873 | |
0bc58a91 EB |
3874 | /* f_version caches the tgid value that the last readdir call couldn't |
3875 | * return. lseek aka telldir automagically resets f_version to 0. | |
3876 | */ | |
9d78edea | 3877 | ns = proc_pid_ns(inode->i_sb); |
f0c3b509 AV |
3878 | tid = (int)file->f_version; |
3879 | file->f_version = 0; | |
d855a4b7 | 3880 | for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns); |
0bc58a91 | 3881 | task; |
f0c3b509 | 3882 | task = next_tid(task), ctx->pos++) { |
e3912ac3 | 3883 | char name[10 + 1]; |
a4ef3895 | 3884 | unsigned int len; |
0658a096 | 3885 | |
b488893a | 3886 | tid = task_pid_nr_ns(task, ns); |
0658a096 FW |
3887 | if (!tid) |
3888 | continue; /* The task has just exited. */ | |
e3912ac3 | 3889 | len = snprintf(name, sizeof(name), "%u", tid); |
f0c3b509 AV |
3890 | if (!proc_fill_cache(file, ctx, name, len, |
3891 | proc_task_instantiate, task, NULL)) { | |
0bc58a91 EB |
3892 | /* returning this tgid failed, save it as the first |
3893 | * pid for the next readir call */ | |
f0c3b509 | 3894 | file->f_version = (u64)tid; |
0bc58a91 | 3895 | put_task_struct(task); |
1da177e4 | 3896 | break; |
0bc58a91 | 3897 | } |
1da177e4 | 3898 | } |
d855a4b7 | 3899 | |
f0c3b509 | 3900 | return 0; |
1da177e4 | 3901 | } |
6e66b52b | 3902 | |
b74d24f7 | 3903 | static int proc_task_getattr(struct mnt_idmap *idmap, |
549c7297 | 3904 | const struct path *path, struct kstat *stat, |
a528d35e | 3905 | u32 request_mask, unsigned int query_flags) |
6e66b52b | 3906 | { |
a528d35e | 3907 | struct inode *inode = d_inode(path->dentry); |
99f89551 | 3908 | struct task_struct *p = get_proc_task(inode); |
0d72b928 | 3909 | generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat); |
6e66b52b | 3910 | |
99f89551 | 3911 | if (p) { |
99f89551 | 3912 | stat->nlink += get_nr_threads(p); |
99f89551 | 3913 | put_task_struct(p); |
6e66b52b EB |
3914 | } |
3915 | ||
3916 | return 0; | |
3917 | } | |
28a6d671 | 3918 | |
c5ef1c42 | 3919 | static const struct inode_operations proc_task_inode_operations = { |
28a6d671 EB |
3920 | .lookup = proc_task_lookup, |
3921 | .getattr = proc_task_getattr, | |
3922 | .setattr = proc_setattr, | |
0499680a | 3923 | .permission = proc_pid_permission, |
28a6d671 EB |
3924 | }; |
3925 | ||
00977a59 | 3926 | static const struct file_operations proc_task_operations = { |
28a6d671 | 3927 | .read = generic_read_dir, |
f50752ea AV |
3928 | .iterate_shared = proc_task_readdir, |
3929 | .llseek = generic_file_llseek, | |
28a6d671 | 3930 | }; |
1270dd8d AD |
3931 | |
3932 | void __init set_proc_pid_nlink(void) | |
3933 | { | |
3934 | nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
3935 | nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
3936 | } |