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> |
1da177e4 LT |
62 | #include <linux/string.h> |
63 | #include <linux/seq_file.h> | |
64 | #include <linux/namei.h> | |
6b3286ed | 65 | #include <linux/mnt_namespace.h> |
1da177e4 | 66 | #include <linux/mm.h> |
a63d83f4 | 67 | #include <linux/swap.h> |
b835996f | 68 | #include <linux/rcupdate.h> |
1da177e4 | 69 | #include <linux/kallsyms.h> |
2ec220e2 | 70 | #include <linux/stacktrace.h> |
d85f50d5 | 71 | #include <linux/resource.h> |
5096add8 | 72 | #include <linux/module.h> |
1da177e4 LT |
73 | #include <linux/mount.h> |
74 | #include <linux/security.h> | |
75 | #include <linux/ptrace.h> | |
0d094efe | 76 | #include <linux/tracehook.h> |
87ebdc00 | 77 | #include <linux/printk.h> |
a424316c | 78 | #include <linux/cgroup.h> |
1da177e4 LT |
79 | #include <linux/cpuset.h> |
80 | #include <linux/audit.h> | |
5addc5dd | 81 | #include <linux/poll.h> |
1651e14e | 82 | #include <linux/nsproxy.h> |
8ac773b4 | 83 | #include <linux/oom.h> |
3cb4a0bb | 84 | #include <linux/elf.h> |
60347f67 | 85 | #include <linux/pid_namespace.h> |
22d917d8 | 86 | #include <linux/user_namespace.h> |
5ad4e53b | 87 | #include <linux/fs_struct.h> |
5a0e3ad6 | 88 | #include <linux/slab.h> |
4eb5aaa3 | 89 | #include <linux/sched/autogroup.h> |
6e84f315 | 90 | #include <linux/sched/mm.h> |
f7ccbae4 | 91 | #include <linux/sched/coredump.h> |
b17b0153 | 92 | #include <linux/sched/debug.h> |
3905f9ad | 93 | #include <linux/sched/stat.h> |
640708a2 | 94 | #include <linux/flex_array.h> |
48f6a7a5 | 95 | #include <linux/posix-timers.h> |
f133ecca CM |
96 | #ifdef CONFIG_HARDWALL |
97 | #include <asm/hardwall.h> | |
98 | #endif | |
43d2b113 | 99 | #include <trace/events/oom.h> |
1da177e4 | 100 | #include "internal.h" |
faf60af1 | 101 | #include "fd.h" |
1da177e4 | 102 | |
0f2fe20f EB |
103 | /* NOTE: |
104 | * Implementing inode permission operations in /proc is almost | |
105 | * certainly an error. Permission checks need to happen during | |
106 | * each system call not at open time. The reason is that most of | |
107 | * what we wish to check for permissions in /proc varies at runtime. | |
108 | * | |
109 | * The classic example of a problem is opening file descriptors | |
110 | * in /proc for a task before it execs a suid executable. | |
111 | */ | |
112 | ||
1270dd8d AD |
113 | static u8 nlink_tid; |
114 | static u8 nlink_tgid; | |
115 | ||
1da177e4 | 116 | struct pid_entry { |
cedbccab | 117 | const char *name; |
623f594e | 118 | unsigned int len; |
d161a13f | 119 | umode_t mode; |
c5ef1c42 | 120 | const struct inode_operations *iop; |
00977a59 | 121 | const struct file_operations *fop; |
20cdc894 | 122 | union proc_op op; |
1da177e4 LT |
123 | }; |
124 | ||
61a28784 | 125 | #define NOD(NAME, MODE, IOP, FOP, OP) { \ |
20cdc894 | 126 | .name = (NAME), \ |
c5141e6d | 127 | .len = sizeof(NAME) - 1, \ |
20cdc894 EB |
128 | .mode = MODE, \ |
129 | .iop = IOP, \ | |
130 | .fop = FOP, \ | |
131 | .op = OP, \ | |
132 | } | |
133 | ||
631f9c18 AD |
134 | #define DIR(NAME, MODE, iops, fops) \ |
135 | NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} ) | |
136 | #define LNK(NAME, get_link) \ | |
61a28784 | 137 | NOD(NAME, (S_IFLNK|S_IRWXUGO), \ |
20cdc894 | 138 | &proc_pid_link_inode_operations, NULL, \ |
631f9c18 AD |
139 | { .proc_get_link = get_link } ) |
140 | #define REG(NAME, MODE, fops) \ | |
141 | NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {}) | |
631f9c18 | 142 | #define ONE(NAME, MODE, show) \ |
be614086 EB |
143 | NOD(NAME, (S_IFREG|(MODE)), \ |
144 | NULL, &proc_single_file_operations, \ | |
631f9c18 | 145 | { .proc_show = show } ) |
1da177e4 | 146 | |
aed54175 VN |
147 | /* |
148 | * Count the number of hardlinks for the pid_entry table, excluding the . | |
149 | * and .. links. | |
150 | */ | |
1270dd8d | 151 | static unsigned int __init pid_entry_nlink(const struct pid_entry *entries, |
aed54175 VN |
152 | unsigned int n) |
153 | { | |
154 | unsigned int i; | |
155 | unsigned int count; | |
156 | ||
1270dd8d | 157 | count = 2; |
aed54175 VN |
158 | for (i = 0; i < n; ++i) { |
159 | if (S_ISDIR(entries[i].mode)) | |
160 | ++count; | |
161 | } | |
162 | ||
163 | return count; | |
164 | } | |
165 | ||
f7ad3c6b | 166 | static int get_task_root(struct task_struct *task, struct path *root) |
1da177e4 | 167 | { |
7c2c7d99 HD |
168 | int result = -ENOENT; |
169 | ||
0494f6ec | 170 | task_lock(task); |
f7ad3c6b MS |
171 | if (task->fs) { |
172 | get_fs_root(task->fs, root); | |
7c2c7d99 HD |
173 | result = 0; |
174 | } | |
0494f6ec | 175 | task_unlock(task); |
7c2c7d99 | 176 | return result; |
0494f6ec MS |
177 | } |
178 | ||
7773fbc5 | 179 | static int proc_cwd_link(struct dentry *dentry, struct path *path) |
0494f6ec | 180 | { |
2b0143b5 | 181 | struct task_struct *task = get_proc_task(d_inode(dentry)); |
0494f6ec | 182 | int result = -ENOENT; |
99f89551 EB |
183 | |
184 | if (task) { | |
f7ad3c6b MS |
185 | task_lock(task); |
186 | if (task->fs) { | |
187 | get_fs_pwd(task->fs, path); | |
188 | result = 0; | |
189 | } | |
190 | task_unlock(task); | |
99f89551 EB |
191 | put_task_struct(task); |
192 | } | |
1da177e4 LT |
193 | return result; |
194 | } | |
195 | ||
7773fbc5 | 196 | static int proc_root_link(struct dentry *dentry, struct path *path) |
1da177e4 | 197 | { |
2b0143b5 | 198 | struct task_struct *task = get_proc_task(d_inode(dentry)); |
1da177e4 | 199 | int result = -ENOENT; |
99f89551 EB |
200 | |
201 | if (task) { | |
f7ad3c6b | 202 | result = get_task_root(task, path); |
99f89551 EB |
203 | put_task_struct(task); |
204 | } | |
1da177e4 LT |
205 | return result; |
206 | } | |
207 | ||
c2c0bb44 AD |
208 | static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, |
209 | size_t _count, loff_t *pos) | |
1da177e4 | 210 | { |
c2c0bb44 AD |
211 | struct task_struct *tsk; |
212 | struct mm_struct *mm; | |
213 | char *page; | |
214 | unsigned long count = _count; | |
215 | unsigned long arg_start, arg_end, env_start, env_end; | |
216 | unsigned long len1, len2, len; | |
217 | unsigned long p; | |
218 | char c; | |
219 | ssize_t rv; | |
220 | ||
221 | BUG_ON(*pos < 0); | |
222 | ||
223 | tsk = get_proc_task(file_inode(file)); | |
224 | if (!tsk) | |
225 | return -ESRCH; | |
226 | mm = get_task_mm(tsk); | |
227 | put_task_struct(tsk); | |
228 | if (!mm) | |
229 | return 0; | |
230 | /* Check if process spawned far enough to have cmdline. */ | |
231 | if (!mm->env_end) { | |
232 | rv = 0; | |
233 | goto out_mmput; | |
234 | } | |
235 | ||
0ee931c4 | 236 | page = (char *)__get_free_page(GFP_KERNEL); |
c2c0bb44 AD |
237 | if (!page) { |
238 | rv = -ENOMEM; | |
239 | goto out_mmput; | |
240 | } | |
241 | ||
242 | down_read(&mm->mmap_sem); | |
243 | arg_start = mm->arg_start; | |
244 | arg_end = mm->arg_end; | |
245 | env_start = mm->env_start; | |
246 | env_end = mm->env_end; | |
247 | up_read(&mm->mmap_sem); | |
248 | ||
249 | BUG_ON(arg_start > arg_end); | |
250 | BUG_ON(env_start > env_end); | |
251 | ||
252 | len1 = arg_end - arg_start; | |
253 | len2 = env_end - env_start; | |
254 | ||
3581d458 AD |
255 | /* Empty ARGV. */ |
256 | if (len1 == 0) { | |
257 | rv = 0; | |
258 | goto out_free_page; | |
259 | } | |
2ca66ff7 | 260 | /* |
c2c0bb44 AD |
261 | * Inherently racy -- command line shares address space |
262 | * with code and data. | |
2ca66ff7 | 263 | */ |
272ddc8b | 264 | rv = access_remote_vm(mm, arg_end - 1, &c, 1, 0); |
c2c0bb44 AD |
265 | if (rv <= 0) |
266 | goto out_free_page; | |
267 | ||
268 | rv = 0; | |
269 | ||
270 | if (c == '\0') { | |
271 | /* Command line (set of strings) occupies whole ARGV. */ | |
272 | if (len1 <= *pos) | |
273 | goto out_free_page; | |
274 | ||
275 | p = arg_start + *pos; | |
276 | len = len1 - *pos; | |
277 | while (count > 0 && len > 0) { | |
278 | unsigned int _count; | |
279 | int nr_read; | |
280 | ||
281 | _count = min3(count, len, PAGE_SIZE); | |
272ddc8b | 282 | nr_read = access_remote_vm(mm, p, page, _count, 0); |
c2c0bb44 AD |
283 | if (nr_read < 0) |
284 | rv = nr_read; | |
285 | if (nr_read <= 0) | |
286 | goto out_free_page; | |
287 | ||
288 | if (copy_to_user(buf, page, nr_read)) { | |
289 | rv = -EFAULT; | |
290 | goto out_free_page; | |
291 | } | |
292 | ||
293 | p += nr_read; | |
294 | len -= nr_read; | |
295 | buf += nr_read; | |
296 | count -= nr_read; | |
297 | rv += nr_read; | |
298 | } | |
299 | } else { | |
c2c0bb44 AD |
300 | /* |
301 | * Command line (1 string) occupies ARGV and | |
302 | * extends into ENVP. | |
303 | */ | |
a0a07b87 AD |
304 | struct { |
305 | unsigned long p; | |
306 | unsigned long len; | |
307 | } cmdline[2] = { | |
308 | { .p = arg_start, .len = len1 }, | |
309 | { .p = env_start, .len = len2 }, | |
310 | }; | |
311 | loff_t pos1 = *pos; | |
312 | unsigned int i; | |
313 | ||
314 | i = 0; | |
315 | while (i < 2 && pos1 >= cmdline[i].len) { | |
316 | pos1 -= cmdline[i].len; | |
317 | i++; | |
c2c0bb44 | 318 | } |
a0a07b87 AD |
319 | while (i < 2) { |
320 | p = cmdline[i].p + pos1; | |
321 | len = cmdline[i].len - pos1; | |
322 | while (count > 0 && len > 0) { | |
323 | unsigned int _count, l; | |
324 | int nr_read; | |
325 | bool final; | |
326 | ||
327 | _count = min3(count, len, PAGE_SIZE); | |
328 | nr_read = access_remote_vm(mm, p, page, _count, 0); | |
329 | if (nr_read < 0) | |
330 | rv = nr_read; | |
331 | if (nr_read <= 0) | |
332 | goto out_free_page; | |
333 | ||
334 | /* | |
335 | * Command line can be shorter than whole ARGV | |
336 | * even if last "marker" byte says it is not. | |
337 | */ | |
338 | final = false; | |
339 | l = strnlen(page, nr_read); | |
340 | if (l < nr_read) { | |
341 | nr_read = l; | |
342 | final = true; | |
343 | } | |
344 | ||
345 | if (copy_to_user(buf, page, nr_read)) { | |
346 | rv = -EFAULT; | |
347 | goto out_free_page; | |
348 | } | |
349 | ||
350 | p += nr_read; | |
351 | len -= nr_read; | |
352 | buf += nr_read; | |
353 | count -= nr_read; | |
354 | rv += nr_read; | |
355 | ||
356 | if (final) | |
357 | goto out_free_page; | |
c2c0bb44 AD |
358 | } |
359 | ||
a0a07b87 AD |
360 | /* Only first chunk can be read partially. */ |
361 | pos1 = 0; | |
362 | i++; | |
c2c0bb44 | 363 | } |
c2c0bb44 AD |
364 | } |
365 | ||
366 | out_free_page: | |
367 | free_page((unsigned long)page); | |
368 | out_mmput: | |
369 | mmput(mm); | |
370 | if (rv > 0) | |
371 | *pos += rv; | |
372 | return rv; | |
1da177e4 LT |
373 | } |
374 | ||
c2c0bb44 AD |
375 | static const struct file_operations proc_pid_cmdline_ops = { |
376 | .read = proc_pid_cmdline_read, | |
377 | .llseek = generic_file_llseek, | |
378 | }; | |
379 | ||
1da177e4 LT |
380 | #ifdef CONFIG_KALLSYMS |
381 | /* | |
382 | * Provides a wchan file via kallsyms in a proper one-value-per-file format. | |
383 | * Returns the resolved symbol. If that fails, simply return the address. | |
384 | */ | |
edfcd606 AD |
385 | static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns, |
386 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 387 | { |
ffb45122 | 388 | unsigned long wchan; |
9281acea | 389 | char symname[KSYM_NAME_LEN]; |
1da177e4 LT |
390 | |
391 | wchan = get_wchan(task); | |
392 | ||
caaee623 JH |
393 | if (wchan && ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS) |
394 | && !lookup_symbol_name(wchan, symname)) | |
25ce3191 | 395 | seq_printf(m, "%s", symname); |
b2f73922 | 396 | else |
1e92a61c | 397 | seq_putc(m, '0'); |
25ce3191 JP |
398 | |
399 | return 0; | |
1da177e4 LT |
400 | } |
401 | #endif /* CONFIG_KALLSYMS */ | |
402 | ||
a9712bc1 AV |
403 | static int lock_trace(struct task_struct *task) |
404 | { | |
405 | int err = mutex_lock_killable(&task->signal->cred_guard_mutex); | |
406 | if (err) | |
407 | return err; | |
caaee623 | 408 | if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) { |
a9712bc1 AV |
409 | mutex_unlock(&task->signal->cred_guard_mutex); |
410 | return -EPERM; | |
411 | } | |
412 | return 0; | |
413 | } | |
414 | ||
415 | static void unlock_trace(struct task_struct *task) | |
416 | { | |
417 | mutex_unlock(&task->signal->cred_guard_mutex); | |
418 | } | |
419 | ||
2ec220e2 KC |
420 | #ifdef CONFIG_STACKTRACE |
421 | ||
422 | #define MAX_STACK_TRACE_DEPTH 64 | |
423 | ||
424 | static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns, | |
425 | struct pid *pid, struct task_struct *task) | |
426 | { | |
427 | struct stack_trace trace; | |
428 | unsigned long *entries; | |
a9712bc1 | 429 | int err; |
2ec220e2 KC |
430 | int i; |
431 | ||
432 | entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL); | |
433 | if (!entries) | |
434 | return -ENOMEM; | |
435 | ||
436 | trace.nr_entries = 0; | |
437 | trace.max_entries = MAX_STACK_TRACE_DEPTH; | |
438 | trace.entries = entries; | |
439 | trace.skip = 0; | |
2ec220e2 | 440 | |
a9712bc1 AV |
441 | err = lock_trace(task); |
442 | if (!err) { | |
443 | save_stack_trace_tsk(task, &trace); | |
444 | ||
445 | for (i = 0; i < trace.nr_entries; i++) { | |
8f5abe84 | 446 | seq_printf(m, "[<0>] %pB\n", (void *)entries[i]); |
a9712bc1 AV |
447 | } |
448 | unlock_trace(task); | |
2ec220e2 KC |
449 | } |
450 | kfree(entries); | |
451 | ||
a9712bc1 | 452 | return err; |
2ec220e2 KC |
453 | } |
454 | #endif | |
455 | ||
5968cece | 456 | #ifdef CONFIG_SCHED_INFO |
1da177e4 LT |
457 | /* |
458 | * Provides /proc/PID/schedstat | |
459 | */ | |
f6e826ca AD |
460 | static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns, |
461 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 462 | { |
5968cece NR |
463 | if (unlikely(!sched_info_on())) |
464 | seq_printf(m, "0 0 0\n"); | |
465 | else | |
466 | seq_printf(m, "%llu %llu %lu\n", | |
25ce3191 JP |
467 | (unsigned long long)task->se.sum_exec_runtime, |
468 | (unsigned long long)task->sched_info.run_delay, | |
469 | task->sched_info.pcount); | |
470 | ||
471 | return 0; | |
1da177e4 LT |
472 | } |
473 | #endif | |
474 | ||
9745512c AV |
475 | #ifdef CONFIG_LATENCYTOP |
476 | static int lstats_show_proc(struct seq_file *m, void *v) | |
477 | { | |
478 | int i; | |
13d77c37 HS |
479 | struct inode *inode = m->private; |
480 | struct task_struct *task = get_proc_task(inode); | |
9745512c | 481 | |
13d77c37 HS |
482 | if (!task) |
483 | return -ESRCH; | |
484 | seq_puts(m, "Latency Top version : v0.1\n"); | |
9745512c | 485 | for (i = 0; i < 32; i++) { |
34e49d4f JP |
486 | struct latency_record *lr = &task->latency_record[i]; |
487 | if (lr->backtrace[0]) { | |
9745512c | 488 | int q; |
34e49d4f JP |
489 | seq_printf(m, "%i %li %li", |
490 | lr->count, lr->time, lr->max); | |
9745512c | 491 | for (q = 0; q < LT_BACKTRACEDEPTH; q++) { |
34e49d4f JP |
492 | unsigned long bt = lr->backtrace[q]; |
493 | if (!bt) | |
9745512c | 494 | break; |
34e49d4f | 495 | if (bt == ULONG_MAX) |
9745512c | 496 | break; |
34e49d4f | 497 | seq_printf(m, " %ps", (void *)bt); |
9745512c | 498 | } |
9d6de12f | 499 | seq_putc(m, '\n'); |
9745512c AV |
500 | } |
501 | ||
502 | } | |
13d77c37 | 503 | put_task_struct(task); |
9745512c AV |
504 | return 0; |
505 | } | |
506 | ||
507 | static int lstats_open(struct inode *inode, struct file *file) | |
508 | { | |
13d77c37 | 509 | return single_open(file, lstats_show_proc, inode); |
d6643d12 HS |
510 | } |
511 | ||
9745512c AV |
512 | static ssize_t lstats_write(struct file *file, const char __user *buf, |
513 | size_t count, loff_t *offs) | |
514 | { | |
496ad9aa | 515 | struct task_struct *task = get_proc_task(file_inode(file)); |
9745512c | 516 | |
13d77c37 HS |
517 | if (!task) |
518 | return -ESRCH; | |
9745512c | 519 | clear_all_latency_tracing(task); |
13d77c37 | 520 | put_task_struct(task); |
9745512c AV |
521 | |
522 | return count; | |
523 | } | |
524 | ||
525 | static const struct file_operations proc_lstats_operations = { | |
526 | .open = lstats_open, | |
527 | .read = seq_read, | |
528 | .write = lstats_write, | |
529 | .llseek = seq_lseek, | |
13d77c37 | 530 | .release = single_release, |
9745512c AV |
531 | }; |
532 | ||
533 | #endif | |
534 | ||
6ba51e37 AD |
535 | static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns, |
536 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 537 | { |
a7f638f9 | 538 | unsigned long totalpages = totalram_pages + total_swap_pages; |
b95c35e7 | 539 | unsigned long points = 0; |
1da177e4 | 540 | |
ef419398 ON |
541 | points = oom_badness(task, NULL, NULL, totalpages) * |
542 | 1000 / totalpages; | |
25ce3191 JP |
543 | seq_printf(m, "%lu\n", points); |
544 | ||
545 | return 0; | |
1da177e4 LT |
546 | } |
547 | ||
d85f50d5 | 548 | struct limit_names { |
cedbccab AD |
549 | const char *name; |
550 | const char *unit; | |
d85f50d5 NH |
551 | }; |
552 | ||
553 | static const struct limit_names lnames[RLIM_NLIMITS] = { | |
cff4edb5 | 554 | [RLIMIT_CPU] = {"Max cpu time", "seconds"}, |
d85f50d5 NH |
555 | [RLIMIT_FSIZE] = {"Max file size", "bytes"}, |
556 | [RLIMIT_DATA] = {"Max data size", "bytes"}, | |
557 | [RLIMIT_STACK] = {"Max stack size", "bytes"}, | |
558 | [RLIMIT_CORE] = {"Max core file size", "bytes"}, | |
559 | [RLIMIT_RSS] = {"Max resident set", "bytes"}, | |
560 | [RLIMIT_NPROC] = {"Max processes", "processes"}, | |
561 | [RLIMIT_NOFILE] = {"Max open files", "files"}, | |
562 | [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"}, | |
563 | [RLIMIT_AS] = {"Max address space", "bytes"}, | |
564 | [RLIMIT_LOCKS] = {"Max file locks", "locks"}, | |
565 | [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"}, | |
566 | [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"}, | |
567 | [RLIMIT_NICE] = {"Max nice priority", NULL}, | |
568 | [RLIMIT_RTPRIO] = {"Max realtime priority", NULL}, | |
8808117c | 569 | [RLIMIT_RTTIME] = {"Max realtime timeout", "us"}, |
d85f50d5 NH |
570 | }; |
571 | ||
572 | /* Display limits for a process */ | |
1c963eb1 AD |
573 | static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns, |
574 | struct pid *pid, struct task_struct *task) | |
d85f50d5 NH |
575 | { |
576 | unsigned int i; | |
d85f50d5 | 577 | unsigned long flags; |
d85f50d5 NH |
578 | |
579 | struct rlimit rlim[RLIM_NLIMITS]; | |
580 | ||
a6bebbc8 | 581 | if (!lock_task_sighand(task, &flags)) |
d85f50d5 | 582 | return 0; |
d85f50d5 NH |
583 | memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS); |
584 | unlock_task_sighand(task, &flags); | |
d85f50d5 NH |
585 | |
586 | /* | |
587 | * print the file header | |
588 | */ | |
1c963eb1 | 589 | seq_printf(m, "%-25s %-20s %-20s %-10s\n", |
25ce3191 | 590 | "Limit", "Soft Limit", "Hard Limit", "Units"); |
d85f50d5 NH |
591 | |
592 | for (i = 0; i < RLIM_NLIMITS; i++) { | |
593 | if (rlim[i].rlim_cur == RLIM_INFINITY) | |
1c963eb1 | 594 | seq_printf(m, "%-25s %-20s ", |
25ce3191 | 595 | lnames[i].name, "unlimited"); |
d85f50d5 | 596 | else |
1c963eb1 | 597 | seq_printf(m, "%-25s %-20lu ", |
25ce3191 | 598 | lnames[i].name, rlim[i].rlim_cur); |
d85f50d5 NH |
599 | |
600 | if (rlim[i].rlim_max == RLIM_INFINITY) | |
1c963eb1 | 601 | seq_printf(m, "%-20s ", "unlimited"); |
d85f50d5 | 602 | else |
1c963eb1 | 603 | seq_printf(m, "%-20lu ", rlim[i].rlim_max); |
d85f50d5 NH |
604 | |
605 | if (lnames[i].unit) | |
1c963eb1 | 606 | seq_printf(m, "%-10s\n", lnames[i].unit); |
d85f50d5 | 607 | else |
1c963eb1 | 608 | seq_putc(m, '\n'); |
d85f50d5 NH |
609 | } |
610 | ||
1c963eb1 | 611 | return 0; |
d85f50d5 NH |
612 | } |
613 | ||
ebcb6734 | 614 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 AD |
615 | static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns, |
616 | struct pid *pid, struct task_struct *task) | |
ebcb6734 RM |
617 | { |
618 | long nr; | |
619 | unsigned long args[6], sp, pc; | |
25ce3191 JP |
620 | int res; |
621 | ||
622 | res = lock_trace(task); | |
a9712bc1 AV |
623 | if (res) |
624 | return res; | |
ebcb6734 RM |
625 | |
626 | if (task_current_syscall(task, &nr, args, 6, &sp, &pc)) | |
09d93bd6 | 627 | seq_puts(m, "running\n"); |
a9712bc1 | 628 | else if (nr < 0) |
09d93bd6 | 629 | seq_printf(m, "%ld 0x%lx 0x%lx\n", nr, sp, pc); |
a9712bc1 | 630 | else |
09d93bd6 | 631 | seq_printf(m, |
ebcb6734 RM |
632 | "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n", |
633 | nr, | |
634 | args[0], args[1], args[2], args[3], args[4], args[5], | |
635 | sp, pc); | |
a9712bc1 | 636 | unlock_trace(task); |
25ce3191 JP |
637 | |
638 | return 0; | |
ebcb6734 RM |
639 | } |
640 | #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */ | |
641 | ||
1da177e4 LT |
642 | /************************************************************************/ |
643 | /* Here the fs part begins */ | |
644 | /************************************************************************/ | |
645 | ||
646 | /* permission checks */ | |
778c1144 | 647 | static int proc_fd_access_allowed(struct inode *inode) |
1da177e4 | 648 | { |
778c1144 EB |
649 | struct task_struct *task; |
650 | int allowed = 0; | |
df26c40e EB |
651 | /* Allow access to a task's file descriptors if it is us or we |
652 | * may use ptrace attach to the process and find out that | |
653 | * information. | |
778c1144 EB |
654 | */ |
655 | task = get_proc_task(inode); | |
df26c40e | 656 | if (task) { |
caaee623 | 657 | allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS); |
778c1144 | 658 | put_task_struct(task); |
df26c40e | 659 | } |
778c1144 | 660 | return allowed; |
1da177e4 LT |
661 | } |
662 | ||
6b4e306a | 663 | int proc_setattr(struct dentry *dentry, struct iattr *attr) |
6d76fa58 LT |
664 | { |
665 | int error; | |
2b0143b5 | 666 | struct inode *inode = d_inode(dentry); |
6d76fa58 LT |
667 | |
668 | if (attr->ia_valid & ATTR_MODE) | |
669 | return -EPERM; | |
670 | ||
31051c85 | 671 | error = setattr_prepare(dentry, attr); |
1025774c CH |
672 | if (error) |
673 | return error; | |
674 | ||
1025774c CH |
675 | setattr_copy(inode, attr); |
676 | mark_inode_dirty(inode); | |
677 | return 0; | |
6d76fa58 LT |
678 | } |
679 | ||
0499680a VK |
680 | /* |
681 | * May current process learn task's sched/cmdline info (for hide_pid_min=1) | |
682 | * or euid/egid (for hide_pid_min=2)? | |
683 | */ | |
684 | static bool has_pid_permissions(struct pid_namespace *pid, | |
685 | struct task_struct *task, | |
686 | int hide_pid_min) | |
687 | { | |
688 | if (pid->hide_pid < hide_pid_min) | |
689 | return true; | |
690 | if (in_group_p(pid->pid_gid)) | |
691 | return true; | |
caaee623 | 692 | return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS); |
0499680a VK |
693 | } |
694 | ||
695 | ||
696 | static int proc_pid_permission(struct inode *inode, int mask) | |
697 | { | |
698 | struct pid_namespace *pid = inode->i_sb->s_fs_info; | |
699 | struct task_struct *task; | |
700 | bool has_perms; | |
701 | ||
702 | task = get_proc_task(inode); | |
a2ef990a XF |
703 | if (!task) |
704 | return -ESRCH; | |
796f571b | 705 | has_perms = has_pid_permissions(pid, task, HIDEPID_NO_ACCESS); |
0499680a VK |
706 | put_task_struct(task); |
707 | ||
708 | if (!has_perms) { | |
796f571b | 709 | if (pid->hide_pid == HIDEPID_INVISIBLE) { |
0499680a VK |
710 | /* |
711 | * Let's make getdents(), stat(), and open() | |
712 | * consistent with each other. If a process | |
713 | * may not stat() a file, it shouldn't be seen | |
714 | * in procfs at all. | |
715 | */ | |
716 | return -ENOENT; | |
717 | } | |
718 | ||
719 | return -EPERM; | |
720 | } | |
721 | return generic_permission(inode, mask); | |
722 | } | |
723 | ||
724 | ||
725 | ||
c5ef1c42 | 726 | static const struct inode_operations proc_def_inode_operations = { |
6d76fa58 LT |
727 | .setattr = proc_setattr, |
728 | }; | |
729 | ||
be614086 EB |
730 | static int proc_single_show(struct seq_file *m, void *v) |
731 | { | |
732 | struct inode *inode = m->private; | |
733 | struct pid_namespace *ns; | |
734 | struct pid *pid; | |
735 | struct task_struct *task; | |
736 | int ret; | |
737 | ||
738 | ns = inode->i_sb->s_fs_info; | |
739 | pid = proc_pid(inode); | |
740 | task = get_pid_task(pid, PIDTYPE_PID); | |
741 | if (!task) | |
742 | return -ESRCH; | |
743 | ||
744 | ret = PROC_I(inode)->op.proc_show(m, ns, pid, task); | |
745 | ||
746 | put_task_struct(task); | |
747 | return ret; | |
748 | } | |
749 | ||
750 | static int proc_single_open(struct inode *inode, struct file *filp) | |
751 | { | |
c6a34058 | 752 | return single_open(filp, proc_single_show, inode); |
be614086 EB |
753 | } |
754 | ||
755 | static const struct file_operations proc_single_file_operations = { | |
756 | .open = proc_single_open, | |
757 | .read = seq_read, | |
758 | .llseek = seq_lseek, | |
759 | .release = single_release, | |
760 | }; | |
761 | ||
5381e169 ON |
762 | |
763 | struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode) | |
1da177e4 | 764 | { |
5381e169 ON |
765 | struct task_struct *task = get_proc_task(inode); |
766 | struct mm_struct *mm = ERR_PTR(-ESRCH); | |
e268337d | 767 | |
5381e169 | 768 | if (task) { |
caaee623 | 769 | mm = mm_access(task, mode | PTRACE_MODE_FSCREDS); |
5381e169 | 770 | put_task_struct(task); |
e268337d | 771 | |
5381e169 ON |
772 | if (!IS_ERR_OR_NULL(mm)) { |
773 | /* ensure this mm_struct can't be freed */ | |
f1f10076 | 774 | mmgrab(mm); |
5381e169 ON |
775 | /* but do not pin its memory */ |
776 | mmput(mm); | |
777 | } | |
778 | } | |
779 | ||
780 | return mm; | |
781 | } | |
782 | ||
783 | static int __mem_open(struct inode *inode, struct file *file, unsigned int mode) | |
784 | { | |
785 | struct mm_struct *mm = proc_mem_open(inode, mode); | |
e268337d LT |
786 | |
787 | if (IS_ERR(mm)) | |
788 | return PTR_ERR(mm); | |
789 | ||
e268337d | 790 | file->private_data = mm; |
1da177e4 LT |
791 | return 0; |
792 | } | |
793 | ||
b409e578 CW |
794 | static int mem_open(struct inode *inode, struct file *file) |
795 | { | |
bc452b4b DH |
796 | int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH); |
797 | ||
798 | /* OK to pass negative loff_t, we can catch out-of-range */ | |
799 | file->f_mode |= FMODE_UNSIGNED_OFFSET; | |
800 | ||
801 | return ret; | |
b409e578 CW |
802 | } |
803 | ||
572d34b9 ON |
804 | static ssize_t mem_rw(struct file *file, char __user *buf, |
805 | size_t count, loff_t *ppos, int write) | |
1da177e4 | 806 | { |
e268337d | 807 | struct mm_struct *mm = file->private_data; |
572d34b9 ON |
808 | unsigned long addr = *ppos; |
809 | ssize_t copied; | |
1da177e4 | 810 | char *page; |
272ddc8b | 811 | unsigned int flags; |
1da177e4 | 812 | |
e268337d LT |
813 | if (!mm) |
814 | return 0; | |
99f89551 | 815 | |
0ee931c4 | 816 | page = (char *)__get_free_page(GFP_KERNEL); |
30cd8903 | 817 | if (!page) |
e268337d | 818 | return -ENOMEM; |
1da177e4 | 819 | |
f7ca54f4 | 820 | copied = 0; |
388f7934 | 821 | if (!mmget_not_zero(mm)) |
6d08f2c7 ON |
822 | goto free; |
823 | ||
f511c0b1 | 824 | flags = FOLL_FORCE | (write ? FOLL_WRITE : 0); |
6347e8d5 | 825 | |
1da177e4 | 826 | while (count > 0) { |
572d34b9 | 827 | int this_len = min_t(int, count, PAGE_SIZE); |
1da177e4 | 828 | |
572d34b9 | 829 | if (write && copy_from_user(page, buf, this_len)) { |
1da177e4 LT |
830 | copied = -EFAULT; |
831 | break; | |
832 | } | |
572d34b9 | 833 | |
6347e8d5 | 834 | this_len = access_remote_vm(mm, addr, page, this_len, flags); |
572d34b9 | 835 | if (!this_len) { |
1da177e4 LT |
836 | if (!copied) |
837 | copied = -EIO; | |
838 | break; | |
839 | } | |
572d34b9 ON |
840 | |
841 | if (!write && copy_to_user(buf, page, this_len)) { | |
842 | copied = -EFAULT; | |
843 | break; | |
844 | } | |
845 | ||
846 | buf += this_len; | |
847 | addr += this_len; | |
848 | copied += this_len; | |
849 | count -= this_len; | |
1da177e4 | 850 | } |
572d34b9 | 851 | *ppos = addr; |
30cd8903 | 852 | |
6d08f2c7 ON |
853 | mmput(mm); |
854 | free: | |
30cd8903 | 855 | free_page((unsigned long) page); |
1da177e4 LT |
856 | return copied; |
857 | } | |
1da177e4 | 858 | |
572d34b9 ON |
859 | static ssize_t mem_read(struct file *file, char __user *buf, |
860 | size_t count, loff_t *ppos) | |
861 | { | |
862 | return mem_rw(file, buf, count, ppos, 0); | |
863 | } | |
864 | ||
865 | static ssize_t mem_write(struct file *file, const char __user *buf, | |
866 | size_t count, loff_t *ppos) | |
867 | { | |
868 | return mem_rw(file, (char __user*)buf, count, ppos, 1); | |
869 | } | |
870 | ||
85863e47 | 871 | loff_t mem_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
872 | { |
873 | switch (orig) { | |
874 | case 0: | |
875 | file->f_pos = offset; | |
876 | break; | |
877 | case 1: | |
878 | file->f_pos += offset; | |
879 | break; | |
880 | default: | |
881 | return -EINVAL; | |
882 | } | |
883 | force_successful_syscall_return(); | |
884 | return file->f_pos; | |
885 | } | |
886 | ||
e268337d LT |
887 | static int mem_release(struct inode *inode, struct file *file) |
888 | { | |
889 | struct mm_struct *mm = file->private_data; | |
71879d3c | 890 | if (mm) |
6d08f2c7 | 891 | mmdrop(mm); |
e268337d LT |
892 | return 0; |
893 | } | |
894 | ||
00977a59 | 895 | static const struct file_operations proc_mem_operations = { |
1da177e4 LT |
896 | .llseek = mem_lseek, |
897 | .read = mem_read, | |
898 | .write = mem_write, | |
899 | .open = mem_open, | |
e268337d | 900 | .release = mem_release, |
1da177e4 LT |
901 | }; |
902 | ||
b409e578 CW |
903 | static int environ_open(struct inode *inode, struct file *file) |
904 | { | |
905 | return __mem_open(inode, file, PTRACE_MODE_READ); | |
906 | } | |
907 | ||
315e28c8 JP |
908 | static ssize_t environ_read(struct file *file, char __user *buf, |
909 | size_t count, loff_t *ppos) | |
910 | { | |
315e28c8 JP |
911 | char *page; |
912 | unsigned long src = *ppos; | |
b409e578 CW |
913 | int ret = 0; |
914 | struct mm_struct *mm = file->private_data; | |
a3b609ef | 915 | unsigned long env_start, env_end; |
315e28c8 | 916 | |
8148a73c MK |
917 | /* Ensure the process spawned far enough to have an environment. */ |
918 | if (!mm || !mm->env_end) | |
b409e578 | 919 | return 0; |
315e28c8 | 920 | |
0ee931c4 | 921 | page = (char *)__get_free_page(GFP_KERNEL); |
315e28c8 | 922 | if (!page) |
b409e578 | 923 | return -ENOMEM; |
315e28c8 | 924 | |
d6f64b89 | 925 | ret = 0; |
388f7934 | 926 | if (!mmget_not_zero(mm)) |
b409e578 | 927 | goto free; |
a3b609ef MG |
928 | |
929 | down_read(&mm->mmap_sem); | |
930 | env_start = mm->env_start; | |
931 | env_end = mm->env_end; | |
932 | up_read(&mm->mmap_sem); | |
933 | ||
315e28c8 | 934 | while (count > 0) { |
e8905ec2 DH |
935 | size_t this_len, max_len; |
936 | int retval; | |
315e28c8 | 937 | |
a3b609ef | 938 | if (src >= (env_end - env_start)) |
315e28c8 JP |
939 | break; |
940 | ||
a3b609ef | 941 | this_len = env_end - (env_start + src); |
e8905ec2 DH |
942 | |
943 | max_len = min_t(size_t, PAGE_SIZE, count); | |
944 | this_len = min(max_len, this_len); | |
315e28c8 | 945 | |
272ddc8b | 946 | retval = access_remote_vm(mm, (env_start + src), page, this_len, 0); |
315e28c8 JP |
947 | |
948 | if (retval <= 0) { | |
949 | ret = retval; | |
950 | break; | |
951 | } | |
952 | ||
953 | if (copy_to_user(buf, page, retval)) { | |
954 | ret = -EFAULT; | |
955 | break; | |
956 | } | |
957 | ||
958 | ret += retval; | |
959 | src += retval; | |
960 | buf += retval; | |
961 | count -= retval; | |
962 | } | |
963 | *ppos = src; | |
315e28c8 | 964 | mmput(mm); |
b409e578 CW |
965 | |
966 | free: | |
315e28c8 | 967 | free_page((unsigned long) page); |
315e28c8 JP |
968 | return ret; |
969 | } | |
970 | ||
971 | static const struct file_operations proc_environ_operations = { | |
b409e578 | 972 | .open = environ_open, |
315e28c8 | 973 | .read = environ_read, |
87df8424 | 974 | .llseek = generic_file_llseek, |
b409e578 | 975 | .release = mem_release, |
315e28c8 JP |
976 | }; |
977 | ||
c5317167 AV |
978 | static int auxv_open(struct inode *inode, struct file *file) |
979 | { | |
980 | return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS); | |
981 | } | |
982 | ||
983 | static ssize_t auxv_read(struct file *file, char __user *buf, | |
984 | size_t count, loff_t *ppos) | |
985 | { | |
986 | struct mm_struct *mm = file->private_data; | |
987 | unsigned int nwords = 0; | |
06b2849d LY |
988 | |
989 | if (!mm) | |
990 | return 0; | |
c5317167 AV |
991 | do { |
992 | nwords += 2; | |
993 | } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */ | |
994 | return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv, | |
995 | nwords * sizeof(mm->saved_auxv[0])); | |
996 | } | |
997 | ||
998 | static const struct file_operations proc_auxv_operations = { | |
999 | .open = auxv_open, | |
1000 | .read = auxv_read, | |
1001 | .llseek = generic_file_llseek, | |
1002 | .release = mem_release, | |
1003 | }; | |
1004 | ||
fa0cbbf1 DR |
1005 | static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count, |
1006 | loff_t *ppos) | |
1007 | { | |
496ad9aa | 1008 | struct task_struct *task = get_proc_task(file_inode(file)); |
fa0cbbf1 DR |
1009 | char buffer[PROC_NUMBUF]; |
1010 | int oom_adj = OOM_ADJUST_MIN; | |
1011 | size_t len; | |
fa0cbbf1 DR |
1012 | |
1013 | if (!task) | |
1014 | return -ESRCH; | |
f913da59 MH |
1015 | if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX) |
1016 | oom_adj = OOM_ADJUST_MAX; | |
1017 | else | |
1018 | oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) / | |
1019 | OOM_SCORE_ADJ_MAX; | |
fa0cbbf1 DR |
1020 | put_task_struct(task); |
1021 | len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj); | |
1022 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
1023 | } | |
1024 | ||
1d5f0acb MH |
1025 | static int __set_oom_adj(struct file *file, int oom_adj, bool legacy) |
1026 | { | |
1027 | static DEFINE_MUTEX(oom_adj_mutex); | |
44a70ade | 1028 | struct mm_struct *mm = NULL; |
1d5f0acb MH |
1029 | struct task_struct *task; |
1030 | int err = 0; | |
1031 | ||
1032 | task = get_proc_task(file_inode(file)); | |
1033 | if (!task) | |
1034 | return -ESRCH; | |
1035 | ||
1036 | mutex_lock(&oom_adj_mutex); | |
1037 | if (legacy) { | |
1038 | if (oom_adj < task->signal->oom_score_adj && | |
1039 | !capable(CAP_SYS_RESOURCE)) { | |
1040 | err = -EACCES; | |
1041 | goto err_unlock; | |
1042 | } | |
1043 | /* | |
1044 | * /proc/pid/oom_adj is provided for legacy purposes, ask users to use | |
1045 | * /proc/pid/oom_score_adj instead. | |
1046 | */ | |
1047 | pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n", | |
1048 | current->comm, task_pid_nr(current), task_pid_nr(task), | |
1049 | task_pid_nr(task)); | |
1050 | } else { | |
1051 | if ((short)oom_adj < task->signal->oom_score_adj_min && | |
1052 | !capable(CAP_SYS_RESOURCE)) { | |
1053 | err = -EACCES; | |
1054 | goto err_unlock; | |
1055 | } | |
1056 | } | |
1057 | ||
44a70ade MH |
1058 | /* |
1059 | * Make sure we will check other processes sharing the mm if this is | |
1060 | * not vfrok which wants its own oom_score_adj. | |
1061 | * pin the mm so it doesn't go away and get reused after task_unlock | |
1062 | */ | |
1063 | if (!task->vfork_done) { | |
1064 | struct task_struct *p = find_lock_task_mm(task); | |
1065 | ||
1066 | if (p) { | |
1067 | if (atomic_read(&p->mm->mm_users) > 1) { | |
1068 | mm = p->mm; | |
f1f10076 | 1069 | mmgrab(mm); |
44a70ade MH |
1070 | } |
1071 | task_unlock(p); | |
1072 | } | |
1073 | } | |
1074 | ||
1d5f0acb MH |
1075 | task->signal->oom_score_adj = oom_adj; |
1076 | if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE)) | |
1077 | task->signal->oom_score_adj_min = (short)oom_adj; | |
1078 | trace_oom_score_adj_update(task); | |
44a70ade MH |
1079 | |
1080 | if (mm) { | |
1081 | struct task_struct *p; | |
1082 | ||
1083 | rcu_read_lock(); | |
1084 | for_each_process(p) { | |
1085 | if (same_thread_group(task, p)) | |
1086 | continue; | |
1087 | ||
1088 | /* do not touch kernel threads or the global init */ | |
1089 | if (p->flags & PF_KTHREAD || is_global_init(p)) | |
1090 | continue; | |
1091 | ||
1092 | task_lock(p); | |
1093 | if (!p->vfork_done && process_shares_mm(p, mm)) { | |
1094 | pr_info("updating oom_score_adj for %d (%s) from %d to %d because it shares mm with %d (%s). Report if this is unexpected.\n", | |
1095 | task_pid_nr(p), p->comm, | |
1096 | p->signal->oom_score_adj, oom_adj, | |
1097 | task_pid_nr(task), task->comm); | |
1098 | p->signal->oom_score_adj = oom_adj; | |
1099 | if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE)) | |
1100 | p->signal->oom_score_adj_min = (short)oom_adj; | |
1101 | } | |
1102 | task_unlock(p); | |
1103 | } | |
1104 | rcu_read_unlock(); | |
1105 | mmdrop(mm); | |
1106 | } | |
1d5f0acb MH |
1107 | err_unlock: |
1108 | mutex_unlock(&oom_adj_mutex); | |
1109 | put_task_struct(task); | |
1110 | return err; | |
1111 | } | |
f913da59 | 1112 | |
b72bdfa7 DR |
1113 | /* |
1114 | * /proc/pid/oom_adj exists solely for backwards compatibility with previous | |
1115 | * kernels. The effective policy is defined by oom_score_adj, which has a | |
1116 | * different scale: oom_adj grew exponentially and oom_score_adj grows linearly. | |
1117 | * Values written to oom_adj are simply mapped linearly to oom_score_adj. | |
1118 | * Processes that become oom disabled via oom_adj will still be oom disabled | |
1119 | * with this implementation. | |
1120 | * | |
1121 | * oom_adj cannot be removed since existing userspace binaries use it. | |
1122 | */ | |
fa0cbbf1 DR |
1123 | static ssize_t oom_adj_write(struct file *file, const char __user *buf, |
1124 | size_t count, loff_t *ppos) | |
1125 | { | |
fa0cbbf1 DR |
1126 | char buffer[PROC_NUMBUF]; |
1127 | int oom_adj; | |
fa0cbbf1 DR |
1128 | int err; |
1129 | ||
1130 | memset(buffer, 0, sizeof(buffer)); | |
1131 | if (count > sizeof(buffer) - 1) | |
1132 | count = sizeof(buffer) - 1; | |
1133 | if (copy_from_user(buffer, buf, count)) { | |
1134 | err = -EFAULT; | |
1135 | goto out; | |
1136 | } | |
1137 | ||
1138 | err = kstrtoint(strstrip(buffer), 0, &oom_adj); | |
1139 | if (err) | |
1140 | goto out; | |
1141 | if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) && | |
1142 | oom_adj != OOM_DISABLE) { | |
1143 | err = -EINVAL; | |
1144 | goto out; | |
1145 | } | |
1146 | ||
fa0cbbf1 DR |
1147 | /* |
1148 | * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum | |
1149 | * value is always attainable. | |
1150 | */ | |
1151 | if (oom_adj == OOM_ADJUST_MAX) | |
1152 | oom_adj = OOM_SCORE_ADJ_MAX; | |
1153 | else | |
1154 | oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE; | |
1155 | ||
1d5f0acb | 1156 | err = __set_oom_adj(file, oom_adj, true); |
fa0cbbf1 DR |
1157 | out: |
1158 | return err < 0 ? err : count; | |
1159 | } | |
1160 | ||
1161 | static const struct file_operations proc_oom_adj_operations = { | |
1162 | .read = oom_adj_read, | |
1163 | .write = oom_adj_write, | |
1164 | .llseek = generic_file_llseek, | |
1165 | }; | |
1166 | ||
a63d83f4 DR |
1167 | static ssize_t oom_score_adj_read(struct file *file, char __user *buf, |
1168 | size_t count, loff_t *ppos) | |
1169 | { | |
496ad9aa | 1170 | struct task_struct *task = get_proc_task(file_inode(file)); |
a63d83f4 | 1171 | char buffer[PROC_NUMBUF]; |
a9c58b90 | 1172 | short oom_score_adj = OOM_SCORE_ADJ_MIN; |
a63d83f4 DR |
1173 | size_t len; |
1174 | ||
1175 | if (!task) | |
1176 | return -ESRCH; | |
f913da59 | 1177 | oom_score_adj = task->signal->oom_score_adj; |
a63d83f4 | 1178 | put_task_struct(task); |
a9c58b90 | 1179 | len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj); |
a63d83f4 DR |
1180 | return simple_read_from_buffer(buf, count, ppos, buffer, len); |
1181 | } | |
1182 | ||
1183 | static ssize_t oom_score_adj_write(struct file *file, const char __user *buf, | |
1184 | size_t count, loff_t *ppos) | |
1185 | { | |
a63d83f4 | 1186 | char buffer[PROC_NUMBUF]; |
0a8cb8e3 | 1187 | int oom_score_adj; |
a63d83f4 DR |
1188 | int err; |
1189 | ||
1190 | memset(buffer, 0, sizeof(buffer)); | |
1191 | if (count > sizeof(buffer) - 1) | |
1192 | count = sizeof(buffer) - 1; | |
723548bf DR |
1193 | if (copy_from_user(buffer, buf, count)) { |
1194 | err = -EFAULT; | |
1195 | goto out; | |
1196 | } | |
a63d83f4 | 1197 | |
0a8cb8e3 | 1198 | err = kstrtoint(strstrip(buffer), 0, &oom_score_adj); |
a63d83f4 | 1199 | if (err) |
723548bf | 1200 | goto out; |
a63d83f4 | 1201 | if (oom_score_adj < OOM_SCORE_ADJ_MIN || |
723548bf DR |
1202 | oom_score_adj > OOM_SCORE_ADJ_MAX) { |
1203 | err = -EINVAL; | |
1204 | goto out; | |
1205 | } | |
a63d83f4 | 1206 | |
1d5f0acb | 1207 | err = __set_oom_adj(file, oom_score_adj, false); |
723548bf DR |
1208 | out: |
1209 | return err < 0 ? err : count; | |
a63d83f4 DR |
1210 | } |
1211 | ||
1212 | static const struct file_operations proc_oom_score_adj_operations = { | |
1213 | .read = oom_score_adj_read, | |
1214 | .write = oom_score_adj_write, | |
6038f373 | 1215 | .llseek = default_llseek, |
a63d83f4 DR |
1216 | }; |
1217 | ||
1da177e4 | 1218 | #ifdef CONFIG_AUDITSYSCALL |
b4eb4f7f | 1219 | #define TMPBUFLEN 11 |
1da177e4 LT |
1220 | static ssize_t proc_loginuid_read(struct file * file, char __user * buf, |
1221 | size_t count, loff_t *ppos) | |
1222 | { | |
496ad9aa | 1223 | struct inode * inode = file_inode(file); |
99f89551 | 1224 | struct task_struct *task = get_proc_task(inode); |
1da177e4 LT |
1225 | ssize_t length; |
1226 | char tmpbuf[TMPBUFLEN]; | |
1227 | ||
99f89551 EB |
1228 | if (!task) |
1229 | return -ESRCH; | |
1da177e4 | 1230 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", |
e1760bd5 EB |
1231 | from_kuid(file->f_cred->user_ns, |
1232 | audit_get_loginuid(task))); | |
99f89551 | 1233 | put_task_struct(task); |
1da177e4 LT |
1234 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); |
1235 | } | |
1236 | ||
1237 | static ssize_t proc_loginuid_write(struct file * file, const char __user * buf, | |
1238 | size_t count, loff_t *ppos) | |
1239 | { | |
496ad9aa | 1240 | struct inode * inode = file_inode(file); |
1da177e4 | 1241 | uid_t loginuid; |
e1760bd5 | 1242 | kuid_t kloginuid; |
774636e1 | 1243 | int rv; |
1da177e4 | 1244 | |
7dc52157 PM |
1245 | rcu_read_lock(); |
1246 | if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) { | |
1247 | rcu_read_unlock(); | |
1da177e4 | 1248 | return -EPERM; |
7dc52157 PM |
1249 | } |
1250 | rcu_read_unlock(); | |
1da177e4 | 1251 | |
1da177e4 LT |
1252 | if (*ppos != 0) { |
1253 | /* No partial writes. */ | |
1254 | return -EINVAL; | |
1255 | } | |
1da177e4 | 1256 | |
774636e1 AD |
1257 | rv = kstrtou32_from_user(buf, count, 10, &loginuid); |
1258 | if (rv < 0) | |
1259 | return rv; | |
81407c84 EP |
1260 | |
1261 | /* is userspace tring to explicitly UNSET the loginuid? */ | |
1262 | if (loginuid == AUDIT_UID_UNSET) { | |
1263 | kloginuid = INVALID_UID; | |
1264 | } else { | |
1265 | kloginuid = make_kuid(file->f_cred->user_ns, loginuid); | |
774636e1 AD |
1266 | if (!uid_valid(kloginuid)) |
1267 | return -EINVAL; | |
e1760bd5 EB |
1268 | } |
1269 | ||
774636e1 AD |
1270 | rv = audit_set_loginuid(kloginuid); |
1271 | if (rv < 0) | |
1272 | return rv; | |
1273 | return count; | |
1da177e4 LT |
1274 | } |
1275 | ||
00977a59 | 1276 | static const struct file_operations proc_loginuid_operations = { |
1da177e4 LT |
1277 | .read = proc_loginuid_read, |
1278 | .write = proc_loginuid_write, | |
87df8424 | 1279 | .llseek = generic_file_llseek, |
1da177e4 | 1280 | }; |
1e0bd755 EP |
1281 | |
1282 | static ssize_t proc_sessionid_read(struct file * file, char __user * buf, | |
1283 | size_t count, loff_t *ppos) | |
1284 | { | |
496ad9aa | 1285 | struct inode * inode = file_inode(file); |
1e0bd755 EP |
1286 | struct task_struct *task = get_proc_task(inode); |
1287 | ssize_t length; | |
1288 | char tmpbuf[TMPBUFLEN]; | |
1289 | ||
1290 | if (!task) | |
1291 | return -ESRCH; | |
1292 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", | |
1293 | audit_get_sessionid(task)); | |
1294 | put_task_struct(task); | |
1295 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); | |
1296 | } | |
1297 | ||
1298 | static const struct file_operations proc_sessionid_operations = { | |
1299 | .read = proc_sessionid_read, | |
87df8424 | 1300 | .llseek = generic_file_llseek, |
1e0bd755 | 1301 | }; |
1da177e4 LT |
1302 | #endif |
1303 | ||
f4f154fd AM |
1304 | #ifdef CONFIG_FAULT_INJECTION |
1305 | static ssize_t proc_fault_inject_read(struct file * file, char __user * buf, | |
1306 | size_t count, loff_t *ppos) | |
1307 | { | |
496ad9aa | 1308 | struct task_struct *task = get_proc_task(file_inode(file)); |
f4f154fd AM |
1309 | char buffer[PROC_NUMBUF]; |
1310 | size_t len; | |
1311 | int make_it_fail; | |
f4f154fd AM |
1312 | |
1313 | if (!task) | |
1314 | return -ESRCH; | |
1315 | make_it_fail = task->make_it_fail; | |
1316 | put_task_struct(task); | |
1317 | ||
1318 | len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail); | |
0c28f287 AM |
1319 | |
1320 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
f4f154fd AM |
1321 | } |
1322 | ||
1323 | static ssize_t proc_fault_inject_write(struct file * file, | |
1324 | const char __user * buf, size_t count, loff_t *ppos) | |
1325 | { | |
1326 | struct task_struct *task; | |
774636e1 | 1327 | char buffer[PROC_NUMBUF]; |
f4f154fd | 1328 | int make_it_fail; |
774636e1 | 1329 | int rv; |
f4f154fd AM |
1330 | |
1331 | if (!capable(CAP_SYS_RESOURCE)) | |
1332 | return -EPERM; | |
1333 | memset(buffer, 0, sizeof(buffer)); | |
1334 | if (count > sizeof(buffer) - 1) | |
1335 | count = sizeof(buffer) - 1; | |
1336 | if (copy_from_user(buffer, buf, count)) | |
1337 | return -EFAULT; | |
774636e1 AD |
1338 | rv = kstrtoint(strstrip(buffer), 0, &make_it_fail); |
1339 | if (rv < 0) | |
1340 | return rv; | |
16caed31 DJ |
1341 | if (make_it_fail < 0 || make_it_fail > 1) |
1342 | return -EINVAL; | |
1343 | ||
496ad9aa | 1344 | task = get_proc_task(file_inode(file)); |
f4f154fd AM |
1345 | if (!task) |
1346 | return -ESRCH; | |
1347 | task->make_it_fail = make_it_fail; | |
1348 | put_task_struct(task); | |
cba8aafe VL |
1349 | |
1350 | return count; | |
f4f154fd AM |
1351 | } |
1352 | ||
00977a59 | 1353 | static const struct file_operations proc_fault_inject_operations = { |
f4f154fd AM |
1354 | .read = proc_fault_inject_read, |
1355 | .write = proc_fault_inject_write, | |
87df8424 | 1356 | .llseek = generic_file_llseek, |
f4f154fd | 1357 | }; |
e41d5818 DV |
1358 | |
1359 | static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf, | |
1360 | size_t count, loff_t *ppos) | |
1361 | { | |
1362 | struct task_struct *task; | |
9049f2f6 AM |
1363 | int err; |
1364 | unsigned int n; | |
e41d5818 | 1365 | |
1203c8e6 AM |
1366 | err = kstrtouint_from_user(buf, count, 0, &n); |
1367 | if (err) | |
1368 | return err; | |
1369 | ||
e41d5818 DV |
1370 | task = get_proc_task(file_inode(file)); |
1371 | if (!task) | |
1372 | return -ESRCH; | |
9f7118b2 | 1373 | task->fail_nth = n; |
e41d5818 | 1374 | put_task_struct(task); |
1203c8e6 | 1375 | |
e41d5818 DV |
1376 | return count; |
1377 | } | |
1378 | ||
1379 | static ssize_t proc_fail_nth_read(struct file *file, char __user *buf, | |
1380 | size_t count, loff_t *ppos) | |
1381 | { | |
1382 | struct task_struct *task; | |
bfc74093 AM |
1383 | char numbuf[PROC_NUMBUF]; |
1384 | ssize_t len; | |
e41d5818 DV |
1385 | |
1386 | task = get_proc_task(file_inode(file)); | |
1387 | if (!task) | |
1388 | return -ESRCH; | |
9f7118b2 | 1389 | len = snprintf(numbuf, sizeof(numbuf), "%u\n", task->fail_nth); |
bfc74093 | 1390 | len = simple_read_from_buffer(buf, count, ppos, numbuf, len); |
1203c8e6 | 1391 | put_task_struct(task); |
bfc74093 AM |
1392 | |
1393 | return len; | |
e41d5818 DV |
1394 | } |
1395 | ||
1396 | static const struct file_operations proc_fail_nth_operations = { | |
1397 | .read = proc_fail_nth_read, | |
1398 | .write = proc_fail_nth_write, | |
1399 | }; | |
f4f154fd AM |
1400 | #endif |
1401 | ||
9745512c | 1402 | |
43ae34cb IM |
1403 | #ifdef CONFIG_SCHED_DEBUG |
1404 | /* | |
1405 | * Print out various scheduling related per-task fields: | |
1406 | */ | |
1407 | static int sched_show(struct seq_file *m, void *v) | |
1408 | { | |
1409 | struct inode *inode = m->private; | |
74dc3384 | 1410 | struct pid_namespace *ns = inode->i_sb->s_fs_info; |
43ae34cb IM |
1411 | struct task_struct *p; |
1412 | ||
43ae34cb IM |
1413 | p = get_proc_task(inode); |
1414 | if (!p) | |
1415 | return -ESRCH; | |
74dc3384 | 1416 | proc_sched_show_task(p, ns, m); |
43ae34cb IM |
1417 | |
1418 | put_task_struct(p); | |
1419 | ||
1420 | return 0; | |
1421 | } | |
1422 | ||
1423 | static ssize_t | |
1424 | sched_write(struct file *file, const char __user *buf, | |
1425 | size_t count, loff_t *offset) | |
1426 | { | |
496ad9aa | 1427 | struct inode *inode = file_inode(file); |
43ae34cb IM |
1428 | struct task_struct *p; |
1429 | ||
43ae34cb IM |
1430 | p = get_proc_task(inode); |
1431 | if (!p) | |
1432 | return -ESRCH; | |
1433 | proc_sched_set_task(p); | |
1434 | ||
1435 | put_task_struct(p); | |
1436 | ||
1437 | return count; | |
1438 | } | |
1439 | ||
1440 | static int sched_open(struct inode *inode, struct file *filp) | |
1441 | { | |
c6a34058 | 1442 | return single_open(filp, sched_show, inode); |
43ae34cb IM |
1443 | } |
1444 | ||
1445 | static const struct file_operations proc_pid_sched_operations = { | |
1446 | .open = sched_open, | |
1447 | .read = seq_read, | |
1448 | .write = sched_write, | |
1449 | .llseek = seq_lseek, | |
5ea473a1 | 1450 | .release = single_release, |
43ae34cb IM |
1451 | }; |
1452 | ||
1453 | #endif | |
1454 | ||
5091faa4 MG |
1455 | #ifdef CONFIG_SCHED_AUTOGROUP |
1456 | /* | |
1457 | * Print out autogroup related information: | |
1458 | */ | |
1459 | static int sched_autogroup_show(struct seq_file *m, void *v) | |
1460 | { | |
1461 | struct inode *inode = m->private; | |
1462 | struct task_struct *p; | |
1463 | ||
1464 | p = get_proc_task(inode); | |
1465 | if (!p) | |
1466 | return -ESRCH; | |
1467 | proc_sched_autogroup_show_task(p, m); | |
1468 | ||
1469 | put_task_struct(p); | |
1470 | ||
1471 | return 0; | |
1472 | } | |
1473 | ||
1474 | static ssize_t | |
1475 | sched_autogroup_write(struct file *file, const char __user *buf, | |
1476 | size_t count, loff_t *offset) | |
1477 | { | |
496ad9aa | 1478 | struct inode *inode = file_inode(file); |
5091faa4 MG |
1479 | struct task_struct *p; |
1480 | char buffer[PROC_NUMBUF]; | |
0a8cb8e3 | 1481 | int nice; |
5091faa4 MG |
1482 | int err; |
1483 | ||
1484 | memset(buffer, 0, sizeof(buffer)); | |
1485 | if (count > sizeof(buffer) - 1) | |
1486 | count = sizeof(buffer) - 1; | |
1487 | if (copy_from_user(buffer, buf, count)) | |
1488 | return -EFAULT; | |
1489 | ||
0a8cb8e3 AD |
1490 | err = kstrtoint(strstrip(buffer), 0, &nice); |
1491 | if (err < 0) | |
1492 | return err; | |
5091faa4 MG |
1493 | |
1494 | p = get_proc_task(inode); | |
1495 | if (!p) | |
1496 | return -ESRCH; | |
1497 | ||
2e5b5b3a | 1498 | err = proc_sched_autogroup_set_nice(p, nice); |
5091faa4 MG |
1499 | if (err) |
1500 | count = err; | |
1501 | ||
1502 | put_task_struct(p); | |
1503 | ||
1504 | return count; | |
1505 | } | |
1506 | ||
1507 | static int sched_autogroup_open(struct inode *inode, struct file *filp) | |
1508 | { | |
1509 | int ret; | |
1510 | ||
1511 | ret = single_open(filp, sched_autogroup_show, NULL); | |
1512 | if (!ret) { | |
1513 | struct seq_file *m = filp->private_data; | |
1514 | ||
1515 | m->private = inode; | |
1516 | } | |
1517 | return ret; | |
1518 | } | |
1519 | ||
1520 | static const struct file_operations proc_pid_sched_autogroup_operations = { | |
1521 | .open = sched_autogroup_open, | |
1522 | .read = seq_read, | |
1523 | .write = sched_autogroup_write, | |
1524 | .llseek = seq_lseek, | |
1525 | .release = single_release, | |
1526 | }; | |
1527 | ||
1528 | #endif /* CONFIG_SCHED_AUTOGROUP */ | |
1529 | ||
4614a696 | 1530 | static ssize_t comm_write(struct file *file, const char __user *buf, |
1531 | size_t count, loff_t *offset) | |
1532 | { | |
496ad9aa | 1533 | struct inode *inode = file_inode(file); |
4614a696 | 1534 | struct task_struct *p; |
1535 | char buffer[TASK_COMM_LEN]; | |
830e0fc9 | 1536 | const size_t maxlen = sizeof(buffer) - 1; |
4614a696 | 1537 | |
1538 | memset(buffer, 0, sizeof(buffer)); | |
830e0fc9 | 1539 | if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count)) |
4614a696 | 1540 | return -EFAULT; |
1541 | ||
1542 | p = get_proc_task(inode); | |
1543 | if (!p) | |
1544 | return -ESRCH; | |
1545 | ||
1546 | if (same_thread_group(current, p)) | |
1547 | set_task_comm(p, buffer); | |
1548 | else | |
1549 | count = -EINVAL; | |
1550 | ||
1551 | put_task_struct(p); | |
1552 | ||
1553 | return count; | |
1554 | } | |
1555 | ||
1556 | static int comm_show(struct seq_file *m, void *v) | |
1557 | { | |
1558 | struct inode *inode = m->private; | |
1559 | struct task_struct *p; | |
1560 | ||
1561 | p = get_proc_task(inode); | |
1562 | if (!p) | |
1563 | return -ESRCH; | |
1564 | ||
1565 | task_lock(p); | |
1566 | seq_printf(m, "%s\n", p->comm); | |
1567 | task_unlock(p); | |
1568 | ||
1569 | put_task_struct(p); | |
1570 | ||
1571 | return 0; | |
1572 | } | |
1573 | ||
1574 | static int comm_open(struct inode *inode, struct file *filp) | |
1575 | { | |
c6a34058 | 1576 | return single_open(filp, comm_show, inode); |
4614a696 | 1577 | } |
1578 | ||
1579 | static const struct file_operations proc_pid_set_comm_operations = { | |
1580 | .open = comm_open, | |
1581 | .read = seq_read, | |
1582 | .write = comm_write, | |
1583 | .llseek = seq_lseek, | |
1584 | .release = single_release, | |
1585 | }; | |
1586 | ||
7773fbc5 | 1587 | static int proc_exe_link(struct dentry *dentry, struct path *exe_path) |
925d1c40 MH |
1588 | { |
1589 | struct task_struct *task; | |
925d1c40 MH |
1590 | struct file *exe_file; |
1591 | ||
2b0143b5 | 1592 | task = get_proc_task(d_inode(dentry)); |
925d1c40 MH |
1593 | if (!task) |
1594 | return -ENOENT; | |
cd81a917 | 1595 | exe_file = get_task_exe_file(task); |
925d1c40 | 1596 | put_task_struct(task); |
925d1c40 MH |
1597 | if (exe_file) { |
1598 | *exe_path = exe_file->f_path; | |
1599 | path_get(&exe_file->f_path); | |
1600 | fput(exe_file); | |
1601 | return 0; | |
1602 | } else | |
1603 | return -ENOENT; | |
1604 | } | |
1605 | ||
6b255391 | 1606 | static const char *proc_pid_get_link(struct dentry *dentry, |
fceef393 AV |
1607 | struct inode *inode, |
1608 | struct delayed_call *done) | |
1da177e4 | 1609 | { |
408ef013 | 1610 | struct path path; |
1da177e4 LT |
1611 | int error = -EACCES; |
1612 | ||
6b255391 AV |
1613 | if (!dentry) |
1614 | return ERR_PTR(-ECHILD); | |
1615 | ||
778c1144 EB |
1616 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1617 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1618 | goto out; |
1da177e4 | 1619 | |
408ef013 CH |
1620 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1621 | if (error) | |
1622 | goto out; | |
1623 | ||
6e77137b | 1624 | nd_jump_link(&path); |
408ef013 | 1625 | return NULL; |
1da177e4 | 1626 | out: |
008b150a | 1627 | return ERR_PTR(error); |
1da177e4 LT |
1628 | } |
1629 | ||
3dcd25f3 | 1630 | static int do_proc_readlink(struct path *path, char __user *buffer, int buflen) |
1da177e4 | 1631 | { |
0ee931c4 | 1632 | char *tmp = (char *)__get_free_page(GFP_KERNEL); |
3dcd25f3 | 1633 | char *pathname; |
1da177e4 LT |
1634 | int len; |
1635 | ||
1636 | if (!tmp) | |
1637 | return -ENOMEM; | |
0c28f287 | 1638 | |
7b2a69ba | 1639 | pathname = d_path(path, tmp, PAGE_SIZE); |
3dcd25f3 JB |
1640 | len = PTR_ERR(pathname); |
1641 | if (IS_ERR(pathname)) | |
1da177e4 | 1642 | goto out; |
3dcd25f3 | 1643 | len = tmp + PAGE_SIZE - 1 - pathname; |
1da177e4 LT |
1644 | |
1645 | if (len > buflen) | |
1646 | len = buflen; | |
3dcd25f3 | 1647 | if (copy_to_user(buffer, pathname, len)) |
1da177e4 LT |
1648 | len = -EFAULT; |
1649 | out: | |
1650 | free_page((unsigned long)tmp); | |
1651 | return len; | |
1652 | } | |
1653 | ||
1654 | static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen) | |
1655 | { | |
1656 | int error = -EACCES; | |
2b0143b5 | 1657 | struct inode *inode = d_inode(dentry); |
3dcd25f3 | 1658 | struct path path; |
1da177e4 | 1659 | |
778c1144 EB |
1660 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1661 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1662 | goto out; |
1da177e4 | 1663 | |
7773fbc5 | 1664 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1da177e4 LT |
1665 | if (error) |
1666 | goto out; | |
1667 | ||
3dcd25f3 JB |
1668 | error = do_proc_readlink(&path, buffer, buflen); |
1669 | path_put(&path); | |
1da177e4 | 1670 | out: |
1da177e4 LT |
1671 | return error; |
1672 | } | |
1673 | ||
faf60af1 | 1674 | const struct inode_operations proc_pid_link_inode_operations = { |
1da177e4 | 1675 | .readlink = proc_pid_readlink, |
6b255391 | 1676 | .get_link = proc_pid_get_link, |
6d76fa58 | 1677 | .setattr = proc_setattr, |
1da177e4 LT |
1678 | }; |
1679 | ||
28a6d671 EB |
1680 | |
1681 | /* building an inode */ | |
1682 | ||
c6eb50d2 | 1683 | void task_dump_owner(struct task_struct *task, umode_t mode, |
68eb94f1 EB |
1684 | kuid_t *ruid, kgid_t *rgid) |
1685 | { | |
1686 | /* Depending on the state of dumpable compute who should own a | |
1687 | * proc file for a task. | |
1688 | */ | |
1689 | const struct cred *cred; | |
1690 | kuid_t uid; | |
1691 | kgid_t gid; | |
1692 | ||
1693 | /* Default to the tasks effective ownership */ | |
1694 | rcu_read_lock(); | |
1695 | cred = __task_cred(task); | |
1696 | uid = cred->euid; | |
1697 | gid = cred->egid; | |
1698 | rcu_read_unlock(); | |
1699 | ||
1700 | /* | |
1701 | * Before the /proc/pid/status file was created the only way to read | |
1702 | * the effective uid of a /process was to stat /proc/pid. Reading | |
1703 | * /proc/pid/status is slow enough that procps and other packages | |
1704 | * kept stating /proc/pid. To keep the rules in /proc simple I have | |
1705 | * made this apply to all per process world readable and executable | |
1706 | * directories. | |
1707 | */ | |
1708 | if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) { | |
1709 | struct mm_struct *mm; | |
1710 | task_lock(task); | |
1711 | mm = task->mm; | |
1712 | /* Make non-dumpable tasks owned by some root */ | |
1713 | if (mm) { | |
1714 | if (get_dumpable(mm) != SUID_DUMP_USER) { | |
1715 | struct user_namespace *user_ns = mm->user_ns; | |
1716 | ||
1717 | uid = make_kuid(user_ns, 0); | |
1718 | if (!uid_valid(uid)) | |
1719 | uid = GLOBAL_ROOT_UID; | |
1720 | ||
1721 | gid = make_kgid(user_ns, 0); | |
1722 | if (!gid_valid(gid)) | |
1723 | gid = GLOBAL_ROOT_GID; | |
1724 | } | |
1725 | } else { | |
1726 | uid = GLOBAL_ROOT_UID; | |
1727 | gid = GLOBAL_ROOT_GID; | |
1728 | } | |
1729 | task_unlock(task); | |
1730 | } | |
1731 | *ruid = uid; | |
1732 | *rgid = gid; | |
1733 | } | |
1734 | ||
db978da8 AG |
1735 | struct inode *proc_pid_make_inode(struct super_block * sb, |
1736 | struct task_struct *task, umode_t mode) | |
28a6d671 EB |
1737 | { |
1738 | struct inode * inode; | |
1739 | struct proc_inode *ei; | |
1da177e4 | 1740 | |
28a6d671 | 1741 | /* We need a new inode */ |
1da177e4 | 1742 | |
28a6d671 EB |
1743 | inode = new_inode(sb); |
1744 | if (!inode) | |
1745 | goto out; | |
1746 | ||
1747 | /* Common stuff */ | |
1748 | ei = PROC_I(inode); | |
db978da8 | 1749 | inode->i_mode = mode; |
85fe4025 | 1750 | inode->i_ino = get_next_ino(); |
078cd827 | 1751 | inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode); |
28a6d671 EB |
1752 | inode->i_op = &proc_def_inode_operations; |
1753 | ||
1754 | /* | |
1755 | * grab the reference to task. | |
1756 | */ | |
1a657f78 | 1757 | ei->pid = get_task_pid(task, PIDTYPE_PID); |
28a6d671 EB |
1758 | if (!ei->pid) |
1759 | goto out_unlock; | |
1760 | ||
68eb94f1 | 1761 | task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid); |
28a6d671 EB |
1762 | security_task_to_inode(task, inode); |
1763 | ||
1da177e4 | 1764 | out: |
28a6d671 EB |
1765 | return inode; |
1766 | ||
1767 | out_unlock: | |
1768 | iput(inode); | |
1769 | return NULL; | |
1da177e4 LT |
1770 | } |
1771 | ||
a528d35e DH |
1772 | int pid_getattr(const struct path *path, struct kstat *stat, |
1773 | u32 request_mask, unsigned int query_flags) | |
1da177e4 | 1774 | { |
a528d35e | 1775 | struct inode *inode = d_inode(path->dentry); |
28a6d671 | 1776 | struct task_struct *task; |
a528d35e | 1777 | struct pid_namespace *pid = path->dentry->d_sb->s_fs_info; |
c69e8d9c | 1778 | |
28a6d671 | 1779 | generic_fillattr(inode, stat); |
1da177e4 | 1780 | |
28a6d671 | 1781 | rcu_read_lock(); |
dcb0f222 EB |
1782 | stat->uid = GLOBAL_ROOT_UID; |
1783 | stat->gid = GLOBAL_ROOT_GID; | |
28a6d671 EB |
1784 | task = pid_task(proc_pid(inode), PIDTYPE_PID); |
1785 | if (task) { | |
796f571b | 1786 | if (!has_pid_permissions(pid, task, HIDEPID_INVISIBLE)) { |
0499680a VK |
1787 | rcu_read_unlock(); |
1788 | /* | |
1789 | * This doesn't prevent learning whether PID exists, | |
1790 | * it only makes getattr() consistent with readdir(). | |
1791 | */ | |
1792 | return -ENOENT; | |
1793 | } | |
68eb94f1 | 1794 | task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid); |
1da177e4 | 1795 | } |
28a6d671 | 1796 | rcu_read_unlock(); |
d6e71144 | 1797 | return 0; |
1da177e4 LT |
1798 | } |
1799 | ||
1da177e4 LT |
1800 | /* dentry stuff */ |
1801 | ||
1802 | /* | |
1803 | * Exceptional case: normally we are not allowed to unhash a busy | |
1804 | * directory. In this case, however, we can do it - no aliasing problems | |
1805 | * due to the way we treat inodes. | |
1806 | * | |
1807 | * Rewrite the inode's ownerships here because the owning task may have | |
1808 | * performed a setuid(), etc. | |
99f89551 | 1809 | * |
1da177e4 | 1810 | */ |
0b728e19 | 1811 | int pid_revalidate(struct dentry *dentry, unsigned int flags) |
1da177e4 | 1812 | { |
34286d66 NP |
1813 | struct inode *inode; |
1814 | struct task_struct *task; | |
c69e8d9c | 1815 | |
0b728e19 | 1816 | if (flags & LOOKUP_RCU) |
34286d66 NP |
1817 | return -ECHILD; |
1818 | ||
2b0143b5 | 1819 | inode = d_inode(dentry); |
34286d66 NP |
1820 | task = get_proc_task(inode); |
1821 | ||
99f89551 | 1822 | if (task) { |
68eb94f1 EB |
1823 | task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid); |
1824 | ||
9ee8ab9f | 1825 | inode->i_mode &= ~(S_ISUID | S_ISGID); |
1da177e4 | 1826 | security_task_to_inode(task, inode); |
99f89551 | 1827 | put_task_struct(task); |
1da177e4 LT |
1828 | return 1; |
1829 | } | |
1da177e4 LT |
1830 | return 0; |
1831 | } | |
1832 | ||
d855a4b7 ON |
1833 | static inline bool proc_inode_is_dead(struct inode *inode) |
1834 | { | |
1835 | return !proc_pid(inode)->tasks[PIDTYPE_PID].first; | |
1836 | } | |
1837 | ||
1dd704b6 DH |
1838 | int pid_delete_dentry(const struct dentry *dentry) |
1839 | { | |
1840 | /* Is the task we represent dead? | |
1841 | * If so, then don't put the dentry on the lru list, | |
1842 | * kill it immediately. | |
1843 | */ | |
2b0143b5 | 1844 | return proc_inode_is_dead(d_inode(dentry)); |
1dd704b6 DH |
1845 | } |
1846 | ||
6b4e306a | 1847 | const struct dentry_operations pid_dentry_operations = |
28a6d671 EB |
1848 | { |
1849 | .d_revalidate = pid_revalidate, | |
1850 | .d_delete = pid_delete_dentry, | |
1851 | }; | |
1852 | ||
1853 | /* Lookups */ | |
1854 | ||
1c0d04c9 EB |
1855 | /* |
1856 | * Fill a directory entry. | |
1857 | * | |
1858 | * If possible create the dcache entry and derive our inode number and | |
1859 | * file type from dcache entry. | |
1860 | * | |
1861 | * Since all of the proc inode numbers are dynamically generated, the inode | |
1862 | * numbers do not exist until the inode is cache. This means creating the | |
1863 | * the dcache entry in readdir is necessary to keep the inode numbers | |
1864 | * reported by readdir in sync with the inode numbers reported | |
1865 | * by stat. | |
1866 | */ | |
f0c3b509 | 1867 | bool proc_fill_cache(struct file *file, struct dir_context *ctx, |
6b4e306a | 1868 | const char *name, int len, |
c5141e6d | 1869 | instantiate_t instantiate, struct task_struct *task, const void *ptr) |
61a28784 | 1870 | { |
f0c3b509 | 1871 | struct dentry *child, *dir = file->f_path.dentry; |
1df98b8b | 1872 | struct qstr qname = QSTR_INIT(name, len); |
61a28784 | 1873 | struct inode *inode; |
1df98b8b AV |
1874 | unsigned type; |
1875 | ino_t ino; | |
61a28784 | 1876 | |
1df98b8b | 1877 | child = d_hash_and_lookup(dir, &qname); |
61a28784 | 1878 | if (!child) { |
3781764b AV |
1879 | DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq); |
1880 | child = d_alloc_parallel(dir, &qname, &wq); | |
1881 | if (IS_ERR(child)) | |
1df98b8b | 1882 | goto end_instantiate; |
3781764b AV |
1883 | if (d_in_lookup(child)) { |
1884 | int err = instantiate(d_inode(dir), child, task, ptr); | |
1885 | d_lookup_done(child); | |
1886 | if (err < 0) { | |
1887 | dput(child); | |
1888 | goto end_instantiate; | |
1889 | } | |
61a28784 EB |
1890 | } |
1891 | } | |
2b0143b5 | 1892 | inode = d_inode(child); |
147ce699 AV |
1893 | ino = inode->i_ino; |
1894 | type = inode->i_mode >> 12; | |
61a28784 | 1895 | dput(child); |
f0c3b509 | 1896 | return dir_emit(ctx, name, len, ino, type); |
1df98b8b AV |
1897 | |
1898 | end_instantiate: | |
1899 | return dir_emit(ctx, name, len, 1, DT_UNKNOWN); | |
61a28784 EB |
1900 | } |
1901 | ||
640708a2 PE |
1902 | /* |
1903 | * dname_to_vma_addr - maps a dentry name into two unsigned longs | |
1904 | * which represent vma start and end addresses. | |
1905 | */ | |
1906 | static int dname_to_vma_addr(struct dentry *dentry, | |
1907 | unsigned long *start, unsigned long *end) | |
1908 | { | |
1909 | if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2) | |
1910 | return -EINVAL; | |
1911 | ||
1912 | return 0; | |
1913 | } | |
1914 | ||
0b728e19 | 1915 | static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags) |
640708a2 PE |
1916 | { |
1917 | unsigned long vm_start, vm_end; | |
1918 | bool exact_vma_exists = false; | |
1919 | struct mm_struct *mm = NULL; | |
1920 | struct task_struct *task; | |
640708a2 PE |
1921 | struct inode *inode; |
1922 | int status = 0; | |
1923 | ||
0b728e19 | 1924 | if (flags & LOOKUP_RCU) |
640708a2 PE |
1925 | return -ECHILD; |
1926 | ||
2b0143b5 | 1927 | inode = d_inode(dentry); |
640708a2 PE |
1928 | task = get_proc_task(inode); |
1929 | if (!task) | |
1930 | goto out_notask; | |
1931 | ||
caaee623 | 1932 | mm = mm_access(task, PTRACE_MODE_READ_FSCREDS); |
2344bec7 | 1933 | if (IS_ERR_OR_NULL(mm)) |
640708a2 PE |
1934 | goto out; |
1935 | ||
1936 | if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) { | |
1937 | down_read(&mm->mmap_sem); | |
1938 | exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end); | |
1939 | up_read(&mm->mmap_sem); | |
1940 | } | |
1941 | ||
1942 | mmput(mm); | |
1943 | ||
1944 | if (exact_vma_exists) { | |
68eb94f1 EB |
1945 | task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid); |
1946 | ||
640708a2 PE |
1947 | security_task_to_inode(task, inode); |
1948 | status = 1; | |
1949 | } | |
1950 | ||
1951 | out: | |
1952 | put_task_struct(task); | |
1953 | ||
1954 | out_notask: | |
640708a2 PE |
1955 | return status; |
1956 | } | |
1957 | ||
1958 | static const struct dentry_operations tid_map_files_dentry_operations = { | |
1959 | .d_revalidate = map_files_d_revalidate, | |
1960 | .d_delete = pid_delete_dentry, | |
1961 | }; | |
1962 | ||
6b255391 | 1963 | static int map_files_get_link(struct dentry *dentry, struct path *path) |
640708a2 PE |
1964 | { |
1965 | unsigned long vm_start, vm_end; | |
1966 | struct vm_area_struct *vma; | |
1967 | struct task_struct *task; | |
1968 | struct mm_struct *mm; | |
1969 | int rc; | |
1970 | ||
1971 | rc = -ENOENT; | |
2b0143b5 | 1972 | task = get_proc_task(d_inode(dentry)); |
640708a2 PE |
1973 | if (!task) |
1974 | goto out; | |
1975 | ||
1976 | mm = get_task_mm(task); | |
1977 | put_task_struct(task); | |
1978 | if (!mm) | |
1979 | goto out; | |
1980 | ||
1981 | rc = dname_to_vma_addr(dentry, &vm_start, &vm_end); | |
1982 | if (rc) | |
1983 | goto out_mmput; | |
1984 | ||
70335abb | 1985 | rc = -ENOENT; |
640708a2 PE |
1986 | down_read(&mm->mmap_sem); |
1987 | vma = find_exact_vma(mm, vm_start, vm_end); | |
1988 | if (vma && vma->vm_file) { | |
1989 | *path = vma->vm_file->f_path; | |
1990 | path_get(path); | |
1991 | rc = 0; | |
1992 | } | |
1993 | up_read(&mm->mmap_sem); | |
1994 | ||
1995 | out_mmput: | |
1996 | mmput(mm); | |
1997 | out: | |
1998 | return rc; | |
1999 | } | |
2000 | ||
2001 | struct map_files_info { | |
7b540d06 | 2002 | fmode_t mode; |
9a87fe0d | 2003 | unsigned int len; |
640708a2 PE |
2004 | unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */ |
2005 | }; | |
2006 | ||
bdb4d100 CO |
2007 | /* |
2008 | * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the | |
2009 | * symlinks may be used to bypass permissions on ancestor directories in the | |
2010 | * path to the file in question. | |
2011 | */ | |
2012 | static const char * | |
6b255391 | 2013 | proc_map_files_get_link(struct dentry *dentry, |
fceef393 AV |
2014 | struct inode *inode, |
2015 | struct delayed_call *done) | |
bdb4d100 CO |
2016 | { |
2017 | if (!capable(CAP_SYS_ADMIN)) | |
2018 | return ERR_PTR(-EPERM); | |
2019 | ||
fceef393 | 2020 | return proc_pid_get_link(dentry, inode, done); |
bdb4d100 CO |
2021 | } |
2022 | ||
2023 | /* | |
6b255391 | 2024 | * Identical to proc_pid_link_inode_operations except for get_link() |
bdb4d100 CO |
2025 | */ |
2026 | static const struct inode_operations proc_map_files_link_inode_operations = { | |
2027 | .readlink = proc_pid_readlink, | |
6b255391 | 2028 | .get_link = proc_map_files_get_link, |
bdb4d100 CO |
2029 | .setattr = proc_setattr, |
2030 | }; | |
2031 | ||
c52a47ac | 2032 | static int |
640708a2 PE |
2033 | proc_map_files_instantiate(struct inode *dir, struct dentry *dentry, |
2034 | struct task_struct *task, const void *ptr) | |
2035 | { | |
7b540d06 | 2036 | fmode_t mode = (fmode_t)(unsigned long)ptr; |
640708a2 PE |
2037 | struct proc_inode *ei; |
2038 | struct inode *inode; | |
2039 | ||
db978da8 AG |
2040 | inode = proc_pid_make_inode(dir->i_sb, task, S_IFLNK | |
2041 | ((mode & FMODE_READ ) ? S_IRUSR : 0) | | |
2042 | ((mode & FMODE_WRITE) ? S_IWUSR : 0)); | |
640708a2 | 2043 | if (!inode) |
c52a47ac | 2044 | return -ENOENT; |
640708a2 PE |
2045 | |
2046 | ei = PROC_I(inode); | |
6b255391 | 2047 | ei->op.proc_get_link = map_files_get_link; |
640708a2 | 2048 | |
bdb4d100 | 2049 | inode->i_op = &proc_map_files_link_inode_operations; |
640708a2 | 2050 | inode->i_size = 64; |
640708a2 PE |
2051 | |
2052 | d_set_d_op(dentry, &tid_map_files_dentry_operations); | |
2053 | d_add(dentry, inode); | |
2054 | ||
c52a47ac | 2055 | return 0; |
640708a2 PE |
2056 | } |
2057 | ||
2058 | static struct dentry *proc_map_files_lookup(struct inode *dir, | |
00cd8dd3 | 2059 | struct dentry *dentry, unsigned int flags) |
640708a2 PE |
2060 | { |
2061 | unsigned long vm_start, vm_end; | |
2062 | struct vm_area_struct *vma; | |
2063 | struct task_struct *task; | |
c52a47ac | 2064 | int result; |
640708a2 PE |
2065 | struct mm_struct *mm; |
2066 | ||
c52a47ac | 2067 | result = -ENOENT; |
640708a2 PE |
2068 | task = get_proc_task(dir); |
2069 | if (!task) | |
2070 | goto out; | |
2071 | ||
c52a47ac | 2072 | result = -EACCES; |
caaee623 | 2073 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
640708a2 PE |
2074 | goto out_put_task; |
2075 | ||
c52a47ac | 2076 | result = -ENOENT; |
640708a2 | 2077 | if (dname_to_vma_addr(dentry, &vm_start, &vm_end)) |
eb94cd96 | 2078 | goto out_put_task; |
640708a2 PE |
2079 | |
2080 | mm = get_task_mm(task); | |
2081 | if (!mm) | |
eb94cd96 | 2082 | goto out_put_task; |
640708a2 PE |
2083 | |
2084 | down_read(&mm->mmap_sem); | |
2085 | vma = find_exact_vma(mm, vm_start, vm_end); | |
2086 | if (!vma) | |
2087 | goto out_no_vma; | |
2088 | ||
05f56484 SK |
2089 | if (vma->vm_file) |
2090 | result = proc_map_files_instantiate(dir, dentry, task, | |
2091 | (void *)(unsigned long)vma->vm_file->f_mode); | |
640708a2 PE |
2092 | |
2093 | out_no_vma: | |
2094 | up_read(&mm->mmap_sem); | |
2095 | mmput(mm); | |
640708a2 PE |
2096 | out_put_task: |
2097 | put_task_struct(task); | |
2098 | out: | |
c52a47ac | 2099 | return ERR_PTR(result); |
640708a2 PE |
2100 | } |
2101 | ||
2102 | static const struct inode_operations proc_map_files_inode_operations = { | |
2103 | .lookup = proc_map_files_lookup, | |
2104 | .permission = proc_fd_permission, | |
2105 | .setattr = proc_setattr, | |
2106 | }; | |
2107 | ||
2108 | static int | |
f0c3b509 | 2109 | proc_map_files_readdir(struct file *file, struct dir_context *ctx) |
640708a2 | 2110 | { |
640708a2 PE |
2111 | struct vm_area_struct *vma; |
2112 | struct task_struct *task; | |
2113 | struct mm_struct *mm; | |
f0c3b509 AV |
2114 | unsigned long nr_files, pos, i; |
2115 | struct flex_array *fa = NULL; | |
2116 | struct map_files_info info; | |
2117 | struct map_files_info *p; | |
640708a2 PE |
2118 | int ret; |
2119 | ||
640708a2 | 2120 | ret = -ENOENT; |
f0c3b509 | 2121 | task = get_proc_task(file_inode(file)); |
640708a2 PE |
2122 | if (!task) |
2123 | goto out; | |
2124 | ||
2125 | ret = -EACCES; | |
caaee623 | 2126 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
640708a2 PE |
2127 | goto out_put_task; |
2128 | ||
2129 | ret = 0; | |
f0c3b509 AV |
2130 | if (!dir_emit_dots(file, ctx)) |
2131 | goto out_put_task; | |
640708a2 | 2132 | |
f0c3b509 AV |
2133 | mm = get_task_mm(task); |
2134 | if (!mm) | |
2135 | goto out_put_task; | |
2136 | down_read(&mm->mmap_sem); | |
640708a2 | 2137 | |
f0c3b509 | 2138 | nr_files = 0; |
640708a2 | 2139 | |
f0c3b509 AV |
2140 | /* |
2141 | * We need two passes here: | |
2142 | * | |
2143 | * 1) Collect vmas of mapped files with mmap_sem taken | |
2144 | * 2) Release mmap_sem and instantiate entries | |
2145 | * | |
2146 | * otherwise we get lockdep complained, since filldir() | |
2147 | * routine might require mmap_sem taken in might_fault(). | |
2148 | */ | |
640708a2 | 2149 | |
f0c3b509 AV |
2150 | for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) { |
2151 | if (vma->vm_file && ++pos > ctx->pos) | |
2152 | nr_files++; | |
2153 | } | |
2154 | ||
2155 | if (nr_files) { | |
2156 | fa = flex_array_alloc(sizeof(info), nr_files, | |
2157 | GFP_KERNEL); | |
2158 | if (!fa || flex_array_prealloc(fa, 0, nr_files, | |
2159 | GFP_KERNEL)) { | |
2160 | ret = -ENOMEM; | |
2161 | if (fa) | |
2162 | flex_array_free(fa); | |
2163 | up_read(&mm->mmap_sem); | |
2164 | mmput(mm); | |
2165 | goto out_put_task; | |
640708a2 | 2166 | } |
f0c3b509 AV |
2167 | for (i = 0, vma = mm->mmap, pos = 2; vma; |
2168 | vma = vma->vm_next) { | |
2169 | if (!vma->vm_file) | |
2170 | continue; | |
2171 | if (++pos <= ctx->pos) | |
2172 | continue; | |
2173 | ||
2174 | info.mode = vma->vm_file->f_mode; | |
2175 | info.len = snprintf(info.name, | |
2176 | sizeof(info.name), "%lx-%lx", | |
2177 | vma->vm_start, vma->vm_end); | |
2178 | if (flex_array_put(fa, i++, &info, GFP_KERNEL)) | |
2179 | BUG(); | |
640708a2 | 2180 | } |
640708a2 | 2181 | } |
f0c3b509 AV |
2182 | up_read(&mm->mmap_sem); |
2183 | ||
2184 | for (i = 0; i < nr_files; i++) { | |
2185 | p = flex_array_get(fa, i); | |
2186 | if (!proc_fill_cache(file, ctx, | |
2187 | p->name, p->len, | |
2188 | proc_map_files_instantiate, | |
2189 | task, | |
2190 | (void *)(unsigned long)p->mode)) | |
2191 | break; | |
2192 | ctx->pos++; | |
640708a2 | 2193 | } |
f0c3b509 AV |
2194 | if (fa) |
2195 | flex_array_free(fa); | |
2196 | mmput(mm); | |
640708a2 | 2197 | |
640708a2 PE |
2198 | out_put_task: |
2199 | put_task_struct(task); | |
2200 | out: | |
2201 | return ret; | |
2202 | } | |
2203 | ||
2204 | static const struct file_operations proc_map_files_operations = { | |
2205 | .read = generic_read_dir, | |
f50752ea AV |
2206 | .iterate_shared = proc_map_files_readdir, |
2207 | .llseek = generic_file_llseek, | |
640708a2 PE |
2208 | }; |
2209 | ||
b18b6a9c | 2210 | #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS) |
48f6a7a5 PE |
2211 | struct timers_private { |
2212 | struct pid *pid; | |
2213 | struct task_struct *task; | |
2214 | struct sighand_struct *sighand; | |
57b8015e | 2215 | struct pid_namespace *ns; |
48f6a7a5 PE |
2216 | unsigned long flags; |
2217 | }; | |
2218 | ||
2219 | static void *timers_start(struct seq_file *m, loff_t *pos) | |
2220 | { | |
2221 | struct timers_private *tp = m->private; | |
2222 | ||
2223 | tp->task = get_pid_task(tp->pid, PIDTYPE_PID); | |
2224 | if (!tp->task) | |
2225 | return ERR_PTR(-ESRCH); | |
2226 | ||
2227 | tp->sighand = lock_task_sighand(tp->task, &tp->flags); | |
2228 | if (!tp->sighand) | |
2229 | return ERR_PTR(-ESRCH); | |
2230 | ||
2231 | return seq_list_start(&tp->task->signal->posix_timers, *pos); | |
2232 | } | |
2233 | ||
2234 | static void *timers_next(struct seq_file *m, void *v, loff_t *pos) | |
2235 | { | |
2236 | struct timers_private *tp = m->private; | |
2237 | return seq_list_next(v, &tp->task->signal->posix_timers, pos); | |
2238 | } | |
2239 | ||
2240 | static void timers_stop(struct seq_file *m, void *v) | |
2241 | { | |
2242 | struct timers_private *tp = m->private; | |
2243 | ||
2244 | if (tp->sighand) { | |
2245 | unlock_task_sighand(tp->task, &tp->flags); | |
2246 | tp->sighand = NULL; | |
2247 | } | |
2248 | ||
2249 | if (tp->task) { | |
2250 | put_task_struct(tp->task); | |
2251 | tp->task = NULL; | |
2252 | } | |
2253 | } | |
2254 | ||
2255 | static int show_timer(struct seq_file *m, void *v) | |
2256 | { | |
2257 | struct k_itimer *timer; | |
57b8015e PE |
2258 | struct timers_private *tp = m->private; |
2259 | int notify; | |
cedbccab | 2260 | static const char * const nstr[] = { |
57b8015e PE |
2261 | [SIGEV_SIGNAL] = "signal", |
2262 | [SIGEV_NONE] = "none", | |
2263 | [SIGEV_THREAD] = "thread", | |
2264 | }; | |
48f6a7a5 PE |
2265 | |
2266 | timer = list_entry((struct list_head *)v, struct k_itimer, list); | |
57b8015e PE |
2267 | notify = timer->it_sigev_notify; |
2268 | ||
48f6a7a5 | 2269 | seq_printf(m, "ID: %d\n", timer->it_id); |
ba3edf1f | 2270 | seq_printf(m, "signal: %d/%px\n", |
25ce3191 JP |
2271 | timer->sigq->info.si_signo, |
2272 | timer->sigq->info.si_value.sival_ptr); | |
57b8015e | 2273 | seq_printf(m, "notify: %s/%s.%d\n", |
25ce3191 JP |
2274 | nstr[notify & ~SIGEV_THREAD_ID], |
2275 | (notify & SIGEV_THREAD_ID) ? "tid" : "pid", | |
2276 | pid_nr_ns(timer->it_pid, tp->ns)); | |
15ef0298 | 2277 | seq_printf(m, "ClockID: %d\n", timer->it_clock); |
48f6a7a5 PE |
2278 | |
2279 | return 0; | |
2280 | } | |
2281 | ||
2282 | static const struct seq_operations proc_timers_seq_ops = { | |
2283 | .start = timers_start, | |
2284 | .next = timers_next, | |
2285 | .stop = timers_stop, | |
2286 | .show = show_timer, | |
2287 | }; | |
2288 | ||
2289 | static int proc_timers_open(struct inode *inode, struct file *file) | |
2290 | { | |
2291 | struct timers_private *tp; | |
2292 | ||
2293 | tp = __seq_open_private(file, &proc_timers_seq_ops, | |
2294 | sizeof(struct timers_private)); | |
2295 | if (!tp) | |
2296 | return -ENOMEM; | |
2297 | ||
2298 | tp->pid = proc_pid(inode); | |
57b8015e | 2299 | tp->ns = inode->i_sb->s_fs_info; |
48f6a7a5 PE |
2300 | return 0; |
2301 | } | |
2302 | ||
2303 | static const struct file_operations proc_timers_operations = { | |
2304 | .open = proc_timers_open, | |
2305 | .read = seq_read, | |
2306 | .llseek = seq_lseek, | |
2307 | .release = seq_release_private, | |
2308 | }; | |
b5946bea | 2309 | #endif |
640708a2 | 2310 | |
5de23d43 JS |
2311 | static ssize_t timerslack_ns_write(struct file *file, const char __user *buf, |
2312 | size_t count, loff_t *offset) | |
2313 | { | |
2314 | struct inode *inode = file_inode(file); | |
2315 | struct task_struct *p; | |
2316 | u64 slack_ns; | |
2317 | int err; | |
2318 | ||
2319 | err = kstrtoull_from_user(buf, count, 10, &slack_ns); | |
2320 | if (err < 0) | |
2321 | return err; | |
2322 | ||
2323 | p = get_proc_task(inode); | |
2324 | if (!p) | |
2325 | return -ESRCH; | |
2326 | ||
4b2bd5fe JS |
2327 | if (p != current) { |
2328 | if (!capable(CAP_SYS_NICE)) { | |
2329 | count = -EPERM; | |
2330 | goto out; | |
2331 | } | |
5de23d43 | 2332 | |
4b2bd5fe JS |
2333 | err = security_task_setscheduler(p); |
2334 | if (err) { | |
2335 | count = err; | |
2336 | goto out; | |
2337 | } | |
904763e1 JS |
2338 | } |
2339 | ||
7abbaf94 JS |
2340 | task_lock(p); |
2341 | if (slack_ns == 0) | |
2342 | p->timer_slack_ns = p->default_timer_slack_ns; | |
2343 | else | |
2344 | p->timer_slack_ns = slack_ns; | |
2345 | task_unlock(p); | |
2346 | ||
2347 | out: | |
5de23d43 JS |
2348 | put_task_struct(p); |
2349 | ||
2350 | return count; | |
2351 | } | |
2352 | ||
2353 | static int timerslack_ns_show(struct seq_file *m, void *v) | |
2354 | { | |
2355 | struct inode *inode = m->private; | |
2356 | struct task_struct *p; | |
7abbaf94 | 2357 | int err = 0; |
5de23d43 JS |
2358 | |
2359 | p = get_proc_task(inode); | |
2360 | if (!p) | |
2361 | return -ESRCH; | |
2362 | ||
4b2bd5fe | 2363 | if (p != current) { |
5de23d43 | 2364 | |
4b2bd5fe JS |
2365 | if (!capable(CAP_SYS_NICE)) { |
2366 | err = -EPERM; | |
2367 | goto out; | |
2368 | } | |
2369 | err = security_task_getscheduler(p); | |
2370 | if (err) | |
2371 | goto out; | |
2372 | } | |
904763e1 | 2373 | |
7abbaf94 JS |
2374 | task_lock(p); |
2375 | seq_printf(m, "%llu\n", p->timer_slack_ns); | |
2376 | task_unlock(p); | |
2377 | ||
2378 | out: | |
5de23d43 JS |
2379 | put_task_struct(p); |
2380 | ||
2381 | return err; | |
2382 | } | |
2383 | ||
2384 | static int timerslack_ns_open(struct inode *inode, struct file *filp) | |
2385 | { | |
2386 | return single_open(filp, timerslack_ns_show, inode); | |
2387 | } | |
2388 | ||
2389 | static const struct file_operations proc_pid_set_timerslack_ns_operations = { | |
2390 | .open = timerslack_ns_open, | |
2391 | .read = seq_read, | |
2392 | .write = timerslack_ns_write, | |
2393 | .llseek = seq_lseek, | |
2394 | .release = single_release, | |
2395 | }; | |
2396 | ||
c52a47ac | 2397 | static int proc_pident_instantiate(struct inode *dir, |
c5141e6d | 2398 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
444ceed8 | 2399 | { |
c5141e6d | 2400 | const struct pid_entry *p = ptr; |
444ceed8 EB |
2401 | struct inode *inode; |
2402 | struct proc_inode *ei; | |
444ceed8 | 2403 | |
db978da8 | 2404 | inode = proc_pid_make_inode(dir->i_sb, task, p->mode); |
444ceed8 EB |
2405 | if (!inode) |
2406 | goto out; | |
2407 | ||
2408 | ei = PROC_I(inode); | |
444ceed8 | 2409 | if (S_ISDIR(inode->i_mode)) |
bfe86848 | 2410 | set_nlink(inode, 2); /* Use getattr to fix if necessary */ |
444ceed8 EB |
2411 | if (p->iop) |
2412 | inode->i_op = p->iop; | |
2413 | if (p->fop) | |
2414 | inode->i_fop = p->fop; | |
2415 | ei->op = p->op; | |
fb045adb | 2416 | d_set_d_op(dentry, &pid_dentry_operations); |
444ceed8 EB |
2417 | d_add(dentry, inode); |
2418 | /* Close the race of the process dying before we return the dentry */ | |
0b728e19 | 2419 | if (pid_revalidate(dentry, 0)) |
c52a47ac | 2420 | return 0; |
444ceed8 | 2421 | out: |
c52a47ac | 2422 | return -ENOENT; |
444ceed8 EB |
2423 | } |
2424 | ||
1da177e4 LT |
2425 | static struct dentry *proc_pident_lookup(struct inode *dir, |
2426 | struct dentry *dentry, | |
c5141e6d | 2427 | const struct pid_entry *ents, |
7bcd6b0e | 2428 | unsigned int nents) |
1da177e4 | 2429 | { |
c52a47ac | 2430 | int error; |
99f89551 | 2431 | struct task_struct *task = get_proc_task(dir); |
c5141e6d | 2432 | const struct pid_entry *p, *last; |
1da177e4 | 2433 | |
c52a47ac | 2434 | error = -ENOENT; |
1da177e4 | 2435 | |
99f89551 EB |
2436 | if (!task) |
2437 | goto out_no_task; | |
1da177e4 | 2438 | |
20cdc894 EB |
2439 | /* |
2440 | * Yes, it does not scale. And it should not. Don't add | |
2441 | * new entries into /proc/<tgid>/ without very good reasons. | |
2442 | */ | |
bac5f5d5 AD |
2443 | last = &ents[nents]; |
2444 | for (p = ents; p < last; p++) { | |
1da177e4 LT |
2445 | if (p->len != dentry->d_name.len) |
2446 | continue; | |
2447 | if (!memcmp(dentry->d_name.name, p->name, p->len)) | |
2448 | break; | |
2449 | } | |
bac5f5d5 | 2450 | if (p >= last) |
1da177e4 LT |
2451 | goto out; |
2452 | ||
444ceed8 | 2453 | error = proc_pident_instantiate(dir, dentry, task, p); |
1da177e4 | 2454 | out: |
99f89551 EB |
2455 | put_task_struct(task); |
2456 | out_no_task: | |
c52a47ac | 2457 | return ERR_PTR(error); |
1da177e4 LT |
2458 | } |
2459 | ||
f0c3b509 | 2460 | static int proc_pident_readdir(struct file *file, struct dir_context *ctx, |
c5141e6d | 2461 | const struct pid_entry *ents, unsigned int nents) |
28a6d671 | 2462 | { |
f0c3b509 AV |
2463 | struct task_struct *task = get_proc_task(file_inode(file)); |
2464 | const struct pid_entry *p; | |
28a6d671 | 2465 | |
28a6d671 | 2466 | if (!task) |
f0c3b509 | 2467 | return -ENOENT; |
28a6d671 | 2468 | |
f0c3b509 AV |
2469 | if (!dir_emit_dots(file, ctx)) |
2470 | goto out; | |
2471 | ||
2472 | if (ctx->pos >= nents + 2) | |
2473 | goto out; | |
28a6d671 | 2474 | |
bac5f5d5 | 2475 | for (p = ents + (ctx->pos - 2); p < ents + nents; p++) { |
f0c3b509 AV |
2476 | if (!proc_fill_cache(file, ctx, p->name, p->len, |
2477 | proc_pident_instantiate, task, p)) | |
2478 | break; | |
2479 | ctx->pos++; | |
2480 | } | |
28a6d671 | 2481 | out: |
61a28784 | 2482 | put_task_struct(task); |
f0c3b509 | 2483 | return 0; |
1da177e4 LT |
2484 | } |
2485 | ||
28a6d671 EB |
2486 | #ifdef CONFIG_SECURITY |
2487 | static ssize_t proc_pid_attr_read(struct file * file, char __user * buf, | |
2488 | size_t count, loff_t *ppos) | |
2489 | { | |
496ad9aa | 2490 | struct inode * inode = file_inode(file); |
04ff9708 | 2491 | char *p = NULL; |
28a6d671 EB |
2492 | ssize_t length; |
2493 | struct task_struct *task = get_proc_task(inode); | |
2494 | ||
28a6d671 | 2495 | if (!task) |
04ff9708 | 2496 | return -ESRCH; |
28a6d671 EB |
2497 | |
2498 | length = security_getprocattr(task, | |
2fddfeef | 2499 | (char*)file->f_path.dentry->d_name.name, |
04ff9708 | 2500 | &p); |
28a6d671 | 2501 | put_task_struct(task); |
04ff9708 AV |
2502 | if (length > 0) |
2503 | length = simple_read_from_buffer(buf, count, ppos, p, length); | |
2504 | kfree(p); | |
28a6d671 | 2505 | return length; |
1da177e4 LT |
2506 | } |
2507 | ||
28a6d671 EB |
2508 | static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf, |
2509 | size_t count, loff_t *ppos) | |
2510 | { | |
496ad9aa | 2511 | struct inode * inode = file_inode(file); |
bb646cdb | 2512 | void *page; |
28a6d671 EB |
2513 | ssize_t length; |
2514 | struct task_struct *task = get_proc_task(inode); | |
2515 | ||
2516 | length = -ESRCH; | |
2517 | if (!task) | |
2518 | goto out_no_task; | |
b21507e2 SS |
2519 | |
2520 | /* A task may only write its own attributes. */ | |
2521 | length = -EACCES; | |
2522 | if (current != task) | |
2523 | goto out; | |
2524 | ||
28a6d671 EB |
2525 | if (count > PAGE_SIZE) |
2526 | count = PAGE_SIZE; | |
2527 | ||
2528 | /* No partial writes. */ | |
2529 | length = -EINVAL; | |
2530 | if (*ppos != 0) | |
2531 | goto out; | |
2532 | ||
bb646cdb AV |
2533 | page = memdup_user(buf, count); |
2534 | if (IS_ERR(page)) { | |
2535 | length = PTR_ERR(page); | |
28a6d671 | 2536 | goto out; |
bb646cdb | 2537 | } |
28a6d671 | 2538 | |
107db7c7 | 2539 | /* Guard against adverse ptrace interaction */ |
b21507e2 | 2540 | length = mutex_lock_interruptible(¤t->signal->cred_guard_mutex); |
107db7c7 DH |
2541 | if (length < 0) |
2542 | goto out_free; | |
2543 | ||
b21507e2 | 2544 | length = security_setprocattr(file->f_path.dentry->d_name.name, |
bb646cdb | 2545 | page, count); |
b21507e2 | 2546 | mutex_unlock(¤t->signal->cred_guard_mutex); |
28a6d671 | 2547 | out_free: |
bb646cdb | 2548 | kfree(page); |
28a6d671 EB |
2549 | out: |
2550 | put_task_struct(task); | |
2551 | out_no_task: | |
2552 | return length; | |
2553 | } | |
2554 | ||
00977a59 | 2555 | static const struct file_operations proc_pid_attr_operations = { |
28a6d671 EB |
2556 | .read = proc_pid_attr_read, |
2557 | .write = proc_pid_attr_write, | |
87df8424 | 2558 | .llseek = generic_file_llseek, |
28a6d671 EB |
2559 | }; |
2560 | ||
c5141e6d | 2561 | static const struct pid_entry attr_dir_stuff[] = { |
631f9c18 AD |
2562 | REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations), |
2563 | REG("prev", S_IRUGO, proc_pid_attr_operations), | |
2564 | REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2565 | REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2566 | REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2567 | REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
28a6d671 EB |
2568 | }; |
2569 | ||
f0c3b509 | 2570 | static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx) |
28a6d671 | 2571 | { |
f0c3b509 AV |
2572 | return proc_pident_readdir(file, ctx, |
2573 | attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2574 | } |
2575 | ||
00977a59 | 2576 | static const struct file_operations proc_attr_dir_operations = { |
1da177e4 | 2577 | .read = generic_read_dir, |
f50752ea AV |
2578 | .iterate_shared = proc_attr_dir_readdir, |
2579 | .llseek = generic_file_llseek, | |
1da177e4 LT |
2580 | }; |
2581 | ||
72d9dcfc | 2582 | static struct dentry *proc_attr_dir_lookup(struct inode *dir, |
00cd8dd3 | 2583 | struct dentry *dentry, unsigned int flags) |
28a6d671 | 2584 | { |
7bcd6b0e EB |
2585 | return proc_pident_lookup(dir, dentry, |
2586 | attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2587 | } |
2588 | ||
c5ef1c42 | 2589 | static const struct inode_operations proc_attr_dir_inode_operations = { |
72d9dcfc | 2590 | .lookup = proc_attr_dir_lookup, |
99f89551 | 2591 | .getattr = pid_getattr, |
6d76fa58 | 2592 | .setattr = proc_setattr, |
1da177e4 LT |
2593 | }; |
2594 | ||
28a6d671 EB |
2595 | #endif |
2596 | ||
698ba7b5 | 2597 | #ifdef CONFIG_ELF_CORE |
3cb4a0bb KH |
2598 | static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf, |
2599 | size_t count, loff_t *ppos) | |
2600 | { | |
496ad9aa | 2601 | struct task_struct *task = get_proc_task(file_inode(file)); |
3cb4a0bb KH |
2602 | struct mm_struct *mm; |
2603 | char buffer[PROC_NUMBUF]; | |
2604 | size_t len; | |
2605 | int ret; | |
2606 | ||
2607 | if (!task) | |
2608 | return -ESRCH; | |
2609 | ||
2610 | ret = 0; | |
2611 | mm = get_task_mm(task); | |
2612 | if (mm) { | |
2613 | len = snprintf(buffer, sizeof(buffer), "%08lx\n", | |
2614 | ((mm->flags & MMF_DUMP_FILTER_MASK) >> | |
2615 | MMF_DUMP_FILTER_SHIFT)); | |
2616 | mmput(mm); | |
2617 | ret = simple_read_from_buffer(buf, count, ppos, buffer, len); | |
2618 | } | |
2619 | ||
2620 | put_task_struct(task); | |
2621 | ||
2622 | return ret; | |
2623 | } | |
2624 | ||
2625 | static ssize_t proc_coredump_filter_write(struct file *file, | |
2626 | const char __user *buf, | |
2627 | size_t count, | |
2628 | loff_t *ppos) | |
2629 | { | |
2630 | struct task_struct *task; | |
2631 | struct mm_struct *mm; | |
3cb4a0bb KH |
2632 | unsigned int val; |
2633 | int ret; | |
2634 | int i; | |
2635 | unsigned long mask; | |
2636 | ||
774636e1 AD |
2637 | ret = kstrtouint_from_user(buf, count, 0, &val); |
2638 | if (ret < 0) | |
2639 | return ret; | |
3cb4a0bb KH |
2640 | |
2641 | ret = -ESRCH; | |
496ad9aa | 2642 | task = get_proc_task(file_inode(file)); |
3cb4a0bb KH |
2643 | if (!task) |
2644 | goto out_no_task; | |
2645 | ||
3cb4a0bb KH |
2646 | mm = get_task_mm(task); |
2647 | if (!mm) | |
2648 | goto out_no_mm; | |
41a0c249 | 2649 | ret = 0; |
3cb4a0bb KH |
2650 | |
2651 | for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) { | |
2652 | if (val & mask) | |
2653 | set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2654 | else | |
2655 | clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2656 | } | |
2657 | ||
2658 | mmput(mm); | |
2659 | out_no_mm: | |
2660 | put_task_struct(task); | |
2661 | out_no_task: | |
774636e1 AD |
2662 | if (ret < 0) |
2663 | return ret; | |
2664 | return count; | |
3cb4a0bb KH |
2665 | } |
2666 | ||
2667 | static const struct file_operations proc_coredump_filter_operations = { | |
2668 | .read = proc_coredump_filter_read, | |
2669 | .write = proc_coredump_filter_write, | |
87df8424 | 2670 | .llseek = generic_file_llseek, |
3cb4a0bb KH |
2671 | }; |
2672 | #endif | |
2673 | ||
aba76fdb | 2674 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 2675 | static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole) |
297c5d92 | 2676 | { |
940389b8 | 2677 | struct task_io_accounting acct = task->ioac; |
5995477a | 2678 | unsigned long flags; |
293eb1e7 | 2679 | int result; |
5995477a | 2680 | |
293eb1e7 VK |
2681 | result = mutex_lock_killable(&task->signal->cred_guard_mutex); |
2682 | if (result) | |
2683 | return result; | |
2684 | ||
caaee623 | 2685 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) { |
293eb1e7 VK |
2686 | result = -EACCES; |
2687 | goto out_unlock; | |
2688 | } | |
1d1221f3 | 2689 | |
5995477a AR |
2690 | if (whole && lock_task_sighand(task, &flags)) { |
2691 | struct task_struct *t = task; | |
2692 | ||
2693 | task_io_accounting_add(&acct, &task->signal->ioac); | |
2694 | while_each_thread(task, t) | |
2695 | task_io_accounting_add(&acct, &t->ioac); | |
2696 | ||
2697 | unlock_task_sighand(task, &flags); | |
297c5d92 | 2698 | } |
25ce3191 JP |
2699 | seq_printf(m, |
2700 | "rchar: %llu\n" | |
2701 | "wchar: %llu\n" | |
2702 | "syscr: %llu\n" | |
2703 | "syscw: %llu\n" | |
2704 | "read_bytes: %llu\n" | |
2705 | "write_bytes: %llu\n" | |
2706 | "cancelled_write_bytes: %llu\n", | |
2707 | (unsigned long long)acct.rchar, | |
2708 | (unsigned long long)acct.wchar, | |
2709 | (unsigned long long)acct.syscr, | |
2710 | (unsigned long long)acct.syscw, | |
2711 | (unsigned long long)acct.read_bytes, | |
2712 | (unsigned long long)acct.write_bytes, | |
2713 | (unsigned long long)acct.cancelled_write_bytes); | |
2714 | result = 0; | |
2715 | ||
293eb1e7 VK |
2716 | out_unlock: |
2717 | mutex_unlock(&task->signal->cred_guard_mutex); | |
2718 | return result; | |
297c5d92 AR |
2719 | } |
2720 | ||
19aadc98 AD |
2721 | static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns, |
2722 | struct pid *pid, struct task_struct *task) | |
297c5d92 | 2723 | { |
19aadc98 | 2724 | return do_io_accounting(task, m, 0); |
aba76fdb | 2725 | } |
297c5d92 | 2726 | |
19aadc98 AD |
2727 | static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns, |
2728 | struct pid *pid, struct task_struct *task) | |
297c5d92 | 2729 | { |
19aadc98 | 2730 | return do_io_accounting(task, m, 1); |
297c5d92 AR |
2731 | } |
2732 | #endif /* CONFIG_TASK_IO_ACCOUNTING */ | |
aba76fdb | 2733 | |
22d917d8 EB |
2734 | #ifdef CONFIG_USER_NS |
2735 | static int proc_id_map_open(struct inode *inode, struct file *file, | |
ccf94f1b | 2736 | const struct seq_operations *seq_ops) |
22d917d8 EB |
2737 | { |
2738 | struct user_namespace *ns = NULL; | |
2739 | struct task_struct *task; | |
2740 | struct seq_file *seq; | |
2741 | int ret = -EINVAL; | |
2742 | ||
2743 | task = get_proc_task(inode); | |
2744 | if (task) { | |
2745 | rcu_read_lock(); | |
2746 | ns = get_user_ns(task_cred_xxx(task, user_ns)); | |
2747 | rcu_read_unlock(); | |
2748 | put_task_struct(task); | |
2749 | } | |
2750 | if (!ns) | |
2751 | goto err; | |
2752 | ||
2753 | ret = seq_open(file, seq_ops); | |
2754 | if (ret) | |
2755 | goto err_put_ns; | |
2756 | ||
2757 | seq = file->private_data; | |
2758 | seq->private = ns; | |
2759 | ||
2760 | return 0; | |
2761 | err_put_ns: | |
2762 | put_user_ns(ns); | |
2763 | err: | |
2764 | return ret; | |
2765 | } | |
2766 | ||
2767 | static int proc_id_map_release(struct inode *inode, struct file *file) | |
2768 | { | |
2769 | struct seq_file *seq = file->private_data; | |
2770 | struct user_namespace *ns = seq->private; | |
2771 | put_user_ns(ns); | |
2772 | return seq_release(inode, file); | |
2773 | } | |
2774 | ||
2775 | static int proc_uid_map_open(struct inode *inode, struct file *file) | |
2776 | { | |
2777 | return proc_id_map_open(inode, file, &proc_uid_seq_operations); | |
2778 | } | |
2779 | ||
2780 | static int proc_gid_map_open(struct inode *inode, struct file *file) | |
2781 | { | |
2782 | return proc_id_map_open(inode, file, &proc_gid_seq_operations); | |
2783 | } | |
2784 | ||
f76d207a EB |
2785 | static int proc_projid_map_open(struct inode *inode, struct file *file) |
2786 | { | |
2787 | return proc_id_map_open(inode, file, &proc_projid_seq_operations); | |
2788 | } | |
2789 | ||
22d917d8 EB |
2790 | static const struct file_operations proc_uid_map_operations = { |
2791 | .open = proc_uid_map_open, | |
2792 | .write = proc_uid_map_write, | |
2793 | .read = seq_read, | |
2794 | .llseek = seq_lseek, | |
2795 | .release = proc_id_map_release, | |
2796 | }; | |
2797 | ||
2798 | static const struct file_operations proc_gid_map_operations = { | |
2799 | .open = proc_gid_map_open, | |
2800 | .write = proc_gid_map_write, | |
2801 | .read = seq_read, | |
2802 | .llseek = seq_lseek, | |
2803 | .release = proc_id_map_release, | |
2804 | }; | |
f76d207a EB |
2805 | |
2806 | static const struct file_operations proc_projid_map_operations = { | |
2807 | .open = proc_projid_map_open, | |
2808 | .write = proc_projid_map_write, | |
2809 | .read = seq_read, | |
2810 | .llseek = seq_lseek, | |
2811 | .release = proc_id_map_release, | |
2812 | }; | |
9cc46516 EB |
2813 | |
2814 | static int proc_setgroups_open(struct inode *inode, struct file *file) | |
2815 | { | |
2816 | struct user_namespace *ns = NULL; | |
2817 | struct task_struct *task; | |
2818 | int ret; | |
2819 | ||
2820 | ret = -ESRCH; | |
2821 | task = get_proc_task(inode); | |
2822 | if (task) { | |
2823 | rcu_read_lock(); | |
2824 | ns = get_user_ns(task_cred_xxx(task, user_ns)); | |
2825 | rcu_read_unlock(); | |
2826 | put_task_struct(task); | |
2827 | } | |
2828 | if (!ns) | |
2829 | goto err; | |
2830 | ||
2831 | if (file->f_mode & FMODE_WRITE) { | |
2832 | ret = -EACCES; | |
2833 | if (!ns_capable(ns, CAP_SYS_ADMIN)) | |
2834 | goto err_put_ns; | |
2835 | } | |
2836 | ||
2837 | ret = single_open(file, &proc_setgroups_show, ns); | |
2838 | if (ret) | |
2839 | goto err_put_ns; | |
2840 | ||
2841 | return 0; | |
2842 | err_put_ns: | |
2843 | put_user_ns(ns); | |
2844 | err: | |
2845 | return ret; | |
2846 | } | |
2847 | ||
2848 | static int proc_setgroups_release(struct inode *inode, struct file *file) | |
2849 | { | |
2850 | struct seq_file *seq = file->private_data; | |
2851 | struct user_namespace *ns = seq->private; | |
2852 | int ret = single_release(inode, file); | |
2853 | put_user_ns(ns); | |
2854 | return ret; | |
2855 | } | |
2856 | ||
2857 | static const struct file_operations proc_setgroups_operations = { | |
2858 | .open = proc_setgroups_open, | |
2859 | .write = proc_setgroups_write, | |
2860 | .read = seq_read, | |
2861 | .llseek = seq_lseek, | |
2862 | .release = proc_setgroups_release, | |
2863 | }; | |
22d917d8 EB |
2864 | #endif /* CONFIG_USER_NS */ |
2865 | ||
47830723 KC |
2866 | static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns, |
2867 | struct pid *pid, struct task_struct *task) | |
2868 | { | |
a9712bc1 AV |
2869 | int err = lock_trace(task); |
2870 | if (!err) { | |
2871 | seq_printf(m, "%08x\n", task->personality); | |
2872 | unlock_trace(task); | |
2873 | } | |
2874 | return err; | |
47830723 KC |
2875 | } |
2876 | ||
7c23b330 JP |
2877 | #ifdef CONFIG_LIVEPATCH |
2878 | static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns, | |
2879 | struct pid *pid, struct task_struct *task) | |
2880 | { | |
2881 | seq_printf(m, "%d\n", task->patch_state); | |
2882 | return 0; | |
2883 | } | |
2884 | #endif /* CONFIG_LIVEPATCH */ | |
2885 | ||
28a6d671 EB |
2886 | /* |
2887 | * Thread groups | |
2888 | */ | |
00977a59 | 2889 | static const struct file_operations proc_task_operations; |
c5ef1c42 | 2890 | static const struct inode_operations proc_task_inode_operations; |
20cdc894 | 2891 | |
c5141e6d | 2892 | static const struct pid_entry tgid_base_stuff[] = { |
631f9c18 AD |
2893 | DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations), |
2894 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), | |
640708a2 | 2895 | DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations), |
631f9c18 | 2896 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 2897 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
b2211a36 | 2898 | #ifdef CONFIG_NET |
631f9c18 | 2899 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), |
b2211a36 | 2900 | #endif |
631f9c18 | 2901 | REG("environ", S_IRUSR, proc_environ_operations), |
c5317167 | 2902 | REG("auxv", S_IRUSR, proc_auxv_operations), |
631f9c18 | 2903 | ONE("status", S_IRUGO, proc_pid_status), |
35a35046 | 2904 | ONE("personality", S_IRUSR, proc_pid_personality), |
1c963eb1 | 2905 | ONE("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 2906 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 2907 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
5091faa4 MG |
2908 | #endif |
2909 | #ifdef CONFIG_SCHED_AUTOGROUP | |
2910 | REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations), | |
ebcb6734 | 2911 | #endif |
4614a696 | 2912 | REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations), |
ebcb6734 | 2913 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 | 2914 | ONE("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 2915 | #endif |
c2c0bb44 | 2916 | REG("cmdline", S_IRUGO, proc_pid_cmdline_ops), |
631f9c18 AD |
2917 | ONE("stat", S_IRUGO, proc_tgid_stat), |
2918 | ONE("statm", S_IRUGO, proc_pid_statm), | |
b7643757 | 2919 | REG("maps", S_IRUGO, proc_pid_maps_operations), |
28a6d671 | 2920 | #ifdef CONFIG_NUMA |
b7643757 | 2921 | REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations), |
28a6d671 | 2922 | #endif |
631f9c18 AD |
2923 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
2924 | LNK("cwd", proc_cwd_link), | |
2925 | LNK("root", proc_root_link), | |
2926 | LNK("exe", proc_exe_link), | |
2927 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
2928 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
2929 | REG("mountstats", S_IRUSR, proc_mountstats_operations), | |
1e883281 | 2930 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 2931 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
b7643757 | 2932 | REG("smaps", S_IRUGO, proc_pid_smaps_operations), |
493b0e9d | 2933 | REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations), |
32ed74a4 | 2934 | REG("pagemap", S_IRUSR, proc_pagemap_operations), |
28a6d671 EB |
2935 | #endif |
2936 | #ifdef CONFIG_SECURITY | |
631f9c18 | 2937 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
2938 | #endif |
2939 | #ifdef CONFIG_KALLSYMS | |
edfcd606 | 2940 | ONE("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 2941 | #endif |
2ec220e2 | 2942 | #ifdef CONFIG_STACKTRACE |
35a35046 | 2943 | ONE("stack", S_IRUSR, proc_pid_stack), |
28a6d671 | 2944 | #endif |
5968cece | 2945 | #ifdef CONFIG_SCHED_INFO |
f6e826ca | 2946 | ONE("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 2947 | #endif |
9745512c | 2948 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 2949 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 2950 | #endif |
8793d854 | 2951 | #ifdef CONFIG_PROC_PID_CPUSET |
52de4779 | 2952 | ONE("cpuset", S_IRUGO, proc_cpuset_show), |
a424316c PM |
2953 | #endif |
2954 | #ifdef CONFIG_CGROUPS | |
006f4ac4 | 2955 | ONE("cgroup", S_IRUGO, proc_cgroup_show), |
28a6d671 | 2956 | #endif |
6ba51e37 | 2957 | ONE("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 2958 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 2959 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
28a6d671 | 2960 | #ifdef CONFIG_AUDITSYSCALL |
631f9c18 AD |
2961 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
2962 | REG("sessionid", S_IRUGO, proc_sessionid_operations), | |
28a6d671 | 2963 | #endif |
f4f154fd | 2964 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 2965 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
168c42bc | 2966 | REG("fail-nth", 0644, proc_fail_nth_operations), |
f4f154fd | 2967 | #endif |
698ba7b5 | 2968 | #ifdef CONFIG_ELF_CORE |
631f9c18 | 2969 | REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations), |
3cb4a0bb | 2970 | #endif |
aba76fdb | 2971 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 2972 | ONE("io", S_IRUSR, proc_tgid_io_accounting), |
aba76fdb | 2973 | #endif |
f133ecca | 2974 | #ifdef CONFIG_HARDWALL |
d962c144 | 2975 | ONE("hardwall", S_IRUGO, proc_pid_hardwall), |
f133ecca | 2976 | #endif |
22d917d8 EB |
2977 | #ifdef CONFIG_USER_NS |
2978 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
2979 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 2980 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
9cc46516 | 2981 | REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations), |
22d917d8 | 2982 | #endif |
b18b6a9c | 2983 | #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS) |
48f6a7a5 PE |
2984 | REG("timers", S_IRUGO, proc_timers_operations), |
2985 | #endif | |
5de23d43 | 2986 | REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations), |
7c23b330 JP |
2987 | #ifdef CONFIG_LIVEPATCH |
2988 | ONE("patch_state", S_IRUSR, proc_pid_patch_state), | |
2989 | #endif | |
28a6d671 | 2990 | }; |
1da177e4 | 2991 | |
f0c3b509 | 2992 | static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 2993 | { |
f0c3b509 AV |
2994 | return proc_pident_readdir(file, ctx, |
2995 | tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
2996 | } |
2997 | ||
00977a59 | 2998 | static const struct file_operations proc_tgid_base_operations = { |
1da177e4 | 2999 | .read = generic_read_dir, |
f50752ea AV |
3000 | .iterate_shared = proc_tgid_base_readdir, |
3001 | .llseek = generic_file_llseek, | |
1da177e4 LT |
3002 | }; |
3003 | ||
00cd8dd3 AV |
3004 | static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
3005 | { | |
7bcd6b0e EB |
3006 | return proc_pident_lookup(dir, dentry, |
3007 | tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
3008 | } |
3009 | ||
c5ef1c42 | 3010 | static const struct inode_operations proc_tgid_base_inode_operations = { |
28a6d671 | 3011 | .lookup = proc_tgid_base_lookup, |
99f89551 | 3012 | .getattr = pid_getattr, |
6d76fa58 | 3013 | .setattr = proc_setattr, |
0499680a | 3014 | .permission = proc_pid_permission, |
1da177e4 | 3015 | }; |
1da177e4 | 3016 | |
60347f67 | 3017 | static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid) |
1da177e4 | 3018 | { |
48e6484d | 3019 | struct dentry *dentry, *leader, *dir; |
e3912ac3 | 3020 | char buf[10 + 1]; |
48e6484d EB |
3021 | struct qstr name; |
3022 | ||
3023 | name.name = buf; | |
e3912ac3 | 3024 | name.len = snprintf(buf, sizeof(buf), "%u", pid); |
4f522a24 | 3025 | /* no ->d_hash() rejects on procfs */ |
60347f67 | 3026 | dentry = d_hash_and_lookup(mnt->mnt_root, &name); |
48e6484d | 3027 | if (dentry) { |
bbd51924 | 3028 | d_invalidate(dentry); |
48e6484d EB |
3029 | dput(dentry); |
3030 | } | |
1da177e4 | 3031 | |
c35a7f18 ON |
3032 | if (pid == tgid) |
3033 | return; | |
3034 | ||
48e6484d | 3035 | name.name = buf; |
e3912ac3 | 3036 | name.len = snprintf(buf, sizeof(buf), "%u", tgid); |
60347f67 | 3037 | leader = d_hash_and_lookup(mnt->mnt_root, &name); |
48e6484d EB |
3038 | if (!leader) |
3039 | goto out; | |
1da177e4 | 3040 | |
48e6484d EB |
3041 | name.name = "task"; |
3042 | name.len = strlen(name.name); | |
3043 | dir = d_hash_and_lookup(leader, &name); | |
3044 | if (!dir) | |
3045 | goto out_put_leader; | |
3046 | ||
3047 | name.name = buf; | |
e3912ac3 | 3048 | name.len = snprintf(buf, sizeof(buf), "%u", pid); |
48e6484d EB |
3049 | dentry = d_hash_and_lookup(dir, &name); |
3050 | if (dentry) { | |
bbd51924 | 3051 | d_invalidate(dentry); |
48e6484d | 3052 | dput(dentry); |
1da177e4 | 3053 | } |
48e6484d EB |
3054 | |
3055 | dput(dir); | |
3056 | out_put_leader: | |
3057 | dput(leader); | |
3058 | out: | |
3059 | return; | |
1da177e4 LT |
3060 | } |
3061 | ||
0895e91d RD |
3062 | /** |
3063 | * proc_flush_task - Remove dcache entries for @task from the /proc dcache. | |
3064 | * @task: task that should be flushed. | |
3065 | * | |
3066 | * When flushing dentries from proc, one needs to flush them from global | |
60347f67 | 3067 | * proc (proc_mnt) and from all the namespaces' procs this task was seen |
0895e91d RD |
3068 | * in. This call is supposed to do all of this job. |
3069 | * | |
3070 | * Looks in the dcache for | |
3071 | * /proc/@pid | |
3072 | * /proc/@tgid/task/@pid | |
3073 | * if either directory is present flushes it and all of it'ts children | |
3074 | * from the dcache. | |
3075 | * | |
3076 | * It is safe and reasonable to cache /proc entries for a task until | |
3077 | * that task exits. After that they just clog up the dcache with | |
3078 | * useless entries, possibly causing useful dcache entries to be | |
3079 | * flushed instead. This routine is proved to flush those useless | |
3080 | * dcache entries at process exit time. | |
3081 | * | |
3082 | * NOTE: This routine is just an optimization so it does not guarantee | |
3083 | * that no dcache entries will exist at process exit time it | |
3084 | * just makes it very unlikely that any will persist. | |
60347f67 PE |
3085 | */ |
3086 | ||
3087 | void proc_flush_task(struct task_struct *task) | |
3088 | { | |
9fcc2d15 | 3089 | int i; |
9b4d1cbe | 3090 | struct pid *pid, *tgid; |
130f77ec PE |
3091 | struct upid *upid; |
3092 | ||
130f77ec | 3093 | pid = task_pid(task); |
9b4d1cbe | 3094 | tgid = task_tgid(task); |
130f77ec | 3095 | |
9fcc2d15 | 3096 | for (i = 0; i <= pid->level; i++) { |
130f77ec PE |
3097 | upid = &pid->numbers[i]; |
3098 | proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr, | |
9b4d1cbe | 3099 | tgid->numbers[i].nr); |
130f77ec | 3100 | } |
60347f67 PE |
3101 | } |
3102 | ||
c52a47ac AV |
3103 | static int proc_pid_instantiate(struct inode *dir, |
3104 | struct dentry * dentry, | |
3105 | struct task_struct *task, const void *ptr) | |
444ceed8 | 3106 | { |
444ceed8 EB |
3107 | struct inode *inode; |
3108 | ||
db978da8 | 3109 | inode = proc_pid_make_inode(dir->i_sb, task, S_IFDIR | S_IRUGO | S_IXUGO); |
444ceed8 EB |
3110 | if (!inode) |
3111 | goto out; | |
3112 | ||
444ceed8 EB |
3113 | inode->i_op = &proc_tgid_base_inode_operations; |
3114 | inode->i_fop = &proc_tgid_base_operations; | |
3115 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 | 3116 | |
1270dd8d | 3117 | set_nlink(inode, nlink_tgid); |
444ceed8 | 3118 | |
fb045adb | 3119 | d_set_d_op(dentry, &pid_dentry_operations); |
444ceed8 EB |
3120 | |
3121 | d_add(dentry, inode); | |
3122 | /* Close the race of the process dying before we return the dentry */ | |
0b728e19 | 3123 | if (pid_revalidate(dentry, 0)) |
c52a47ac | 3124 | return 0; |
444ceed8 | 3125 | out: |
c52a47ac | 3126 | return -ENOENT; |
444ceed8 EB |
3127 | } |
3128 | ||
00cd8dd3 | 3129 | struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags) |
1da177e4 | 3130 | { |
335eb531 | 3131 | int result = -ENOENT; |
1da177e4 | 3132 | struct task_struct *task; |
1da177e4 | 3133 | unsigned tgid; |
b488893a | 3134 | struct pid_namespace *ns; |
1da177e4 | 3135 | |
dbcdb504 | 3136 | tgid = name_to_int(&dentry->d_name); |
1da177e4 LT |
3137 | if (tgid == ~0U) |
3138 | goto out; | |
3139 | ||
b488893a | 3140 | ns = dentry->d_sb->s_fs_info; |
de758734 | 3141 | rcu_read_lock(); |
b488893a | 3142 | task = find_task_by_pid_ns(tgid, ns); |
1da177e4 LT |
3143 | if (task) |
3144 | get_task_struct(task); | |
de758734 | 3145 | rcu_read_unlock(); |
1da177e4 LT |
3146 | if (!task) |
3147 | goto out; | |
3148 | ||
444ceed8 | 3149 | result = proc_pid_instantiate(dir, dentry, task, NULL); |
1da177e4 | 3150 | put_task_struct(task); |
1da177e4 | 3151 | out: |
c52a47ac | 3152 | return ERR_PTR(result); |
1da177e4 LT |
3153 | } |
3154 | ||
1da177e4 | 3155 | /* |
0804ef4b | 3156 | * Find the first task with tgid >= tgid |
0bc58a91 | 3157 | * |
1da177e4 | 3158 | */ |
19fd4bb2 EB |
3159 | struct tgid_iter { |
3160 | unsigned int tgid; | |
0804ef4b | 3161 | struct task_struct *task; |
19fd4bb2 EB |
3162 | }; |
3163 | static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter) | |
3164 | { | |
0804ef4b | 3165 | struct pid *pid; |
1da177e4 | 3166 | |
19fd4bb2 EB |
3167 | if (iter.task) |
3168 | put_task_struct(iter.task); | |
454cc105 | 3169 | rcu_read_lock(); |
0804ef4b | 3170 | retry: |
19fd4bb2 EB |
3171 | iter.task = NULL; |
3172 | pid = find_ge_pid(iter.tgid, ns); | |
0804ef4b | 3173 | if (pid) { |
19fd4bb2 EB |
3174 | iter.tgid = pid_nr_ns(pid, ns); |
3175 | iter.task = pid_task(pid, PIDTYPE_PID); | |
0804ef4b EB |
3176 | /* What we to know is if the pid we have find is the |
3177 | * pid of a thread_group_leader. Testing for task | |
3178 | * being a thread_group_leader is the obvious thing | |
3179 | * todo but there is a window when it fails, due to | |
3180 | * the pid transfer logic in de_thread. | |
3181 | * | |
3182 | * So we perform the straight forward test of seeing | |
3183 | * if the pid we have found is the pid of a thread | |
3184 | * group leader, and don't worry if the task we have | |
3185 | * found doesn't happen to be a thread group leader. | |
3186 | * As we don't care in the case of readdir. | |
3187 | */ | |
19fd4bb2 EB |
3188 | if (!iter.task || !has_group_leader_pid(iter.task)) { |
3189 | iter.tgid += 1; | |
0804ef4b | 3190 | goto retry; |
19fd4bb2 EB |
3191 | } |
3192 | get_task_struct(iter.task); | |
0bc58a91 | 3193 | } |
454cc105 | 3194 | rcu_read_unlock(); |
19fd4bb2 | 3195 | return iter; |
1da177e4 LT |
3196 | } |
3197 | ||
0097875b | 3198 | #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2) |
0804ef4b | 3199 | |
1da177e4 | 3200 | /* for the /proc/ directory itself, after non-process stuff has been done */ |
f0c3b509 | 3201 | int proc_pid_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3202 | { |
19fd4bb2 | 3203 | struct tgid_iter iter; |
3aa3377f | 3204 | struct pid_namespace *ns = file_inode(file)->i_sb->s_fs_info; |
f0c3b509 | 3205 | loff_t pos = ctx->pos; |
1da177e4 | 3206 | |
021ada7d | 3207 | if (pos >= PID_MAX_LIMIT + TGID_OFFSET) |
f0c3b509 | 3208 | return 0; |
1da177e4 | 3209 | |
0097875b | 3210 | if (pos == TGID_OFFSET - 2) { |
2b0143b5 | 3211 | struct inode *inode = d_inode(ns->proc_self); |
db963164 | 3212 | if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK)) |
f0c3b509 | 3213 | return 0; |
0097875b EB |
3214 | ctx->pos = pos = pos + 1; |
3215 | } | |
3216 | if (pos == TGID_OFFSET - 1) { | |
2b0143b5 | 3217 | struct inode *inode = d_inode(ns->proc_thread_self); |
0097875b EB |
3218 | if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK)) |
3219 | return 0; | |
3220 | ctx->pos = pos = pos + 1; | |
021ada7d | 3221 | } |
0097875b | 3222 | iter.tgid = pos - TGID_OFFSET; |
19fd4bb2 | 3223 | iter.task = NULL; |
19fd4bb2 EB |
3224 | for (iter = next_tgid(ns, iter); |
3225 | iter.task; | |
3226 | iter.tgid += 1, iter = next_tgid(ns, iter)) { | |
e3912ac3 | 3227 | char name[10 + 1]; |
f0c3b509 | 3228 | int len; |
3ba4bcee ED |
3229 | |
3230 | cond_resched(); | |
796f571b | 3231 | if (!has_pid_permissions(ns, iter.task, HIDEPID_INVISIBLE)) |
f0c3b509 | 3232 | continue; |
0499680a | 3233 | |
e3912ac3 | 3234 | len = snprintf(name, sizeof(name), "%u", iter.tgid); |
f0c3b509 AV |
3235 | ctx->pos = iter.tgid + TGID_OFFSET; |
3236 | if (!proc_fill_cache(file, ctx, name, len, | |
3237 | proc_pid_instantiate, iter.task, NULL)) { | |
19fd4bb2 | 3238 | put_task_struct(iter.task); |
f0c3b509 | 3239 | return 0; |
1da177e4 | 3240 | } |
0bc58a91 | 3241 | } |
f0c3b509 | 3242 | ctx->pos = PID_MAX_LIMIT + TGID_OFFSET; |
0bc58a91 EB |
3243 | return 0; |
3244 | } | |
1da177e4 | 3245 | |
1b3044e3 JD |
3246 | /* |
3247 | * proc_tid_comm_permission is a special permission function exclusively | |
3248 | * used for the node /proc/<pid>/task/<tid>/comm. | |
3249 | * It bypasses generic permission checks in the case where a task of the same | |
3250 | * task group attempts to access the node. | |
3251 | * The rationale behind this is that glibc and bionic access this node for | |
3252 | * cross thread naming (pthread_set/getname_np(!self)). However, if | |
3253 | * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0, | |
3254 | * which locks out the cross thread naming implementation. | |
3255 | * This function makes sure that the node is always accessible for members of | |
3256 | * same thread group. | |
3257 | */ | |
3258 | static int proc_tid_comm_permission(struct inode *inode, int mask) | |
3259 | { | |
3260 | bool is_same_tgroup; | |
3261 | struct task_struct *task; | |
3262 | ||
3263 | task = get_proc_task(inode); | |
3264 | if (!task) | |
3265 | return -ESRCH; | |
3266 | is_same_tgroup = same_thread_group(current, task); | |
3267 | put_task_struct(task); | |
3268 | ||
3269 | if (likely(is_same_tgroup && !(mask & MAY_EXEC))) { | |
3270 | /* This file (/proc/<pid>/task/<tid>/comm) can always be | |
3271 | * read or written by the members of the corresponding | |
3272 | * thread group. | |
3273 | */ | |
3274 | return 0; | |
3275 | } | |
3276 | ||
3277 | return generic_permission(inode, mask); | |
3278 | } | |
3279 | ||
3280 | static const struct inode_operations proc_tid_comm_inode_operations = { | |
3281 | .permission = proc_tid_comm_permission, | |
3282 | }; | |
3283 | ||
28a6d671 EB |
3284 | /* |
3285 | * Tasks | |
3286 | */ | |
c5141e6d | 3287 | static const struct pid_entry tid_base_stuff[] = { |
631f9c18 | 3288 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), |
3835541d | 3289 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 3290 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
6ba8ed79 EB |
3291 | #ifdef CONFIG_NET |
3292 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), | |
3293 | #endif | |
631f9c18 | 3294 | REG("environ", S_IRUSR, proc_environ_operations), |
c5317167 | 3295 | REG("auxv", S_IRUSR, proc_auxv_operations), |
631f9c18 | 3296 | ONE("status", S_IRUGO, proc_pid_status), |
35a35046 | 3297 | ONE("personality", S_IRUSR, proc_pid_personality), |
1c963eb1 | 3298 | ONE("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 3299 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 3300 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
ebcb6734 | 3301 | #endif |
1b3044e3 JD |
3302 | NOD("comm", S_IFREG|S_IRUGO|S_IWUSR, |
3303 | &proc_tid_comm_inode_operations, | |
3304 | &proc_pid_set_comm_operations, {}), | |
ebcb6734 | 3305 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 | 3306 | ONE("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 3307 | #endif |
c2c0bb44 | 3308 | REG("cmdline", S_IRUGO, proc_pid_cmdline_ops), |
631f9c18 AD |
3309 | ONE("stat", S_IRUGO, proc_tid_stat), |
3310 | ONE("statm", S_IRUGO, proc_pid_statm), | |
b7643757 | 3311 | REG("maps", S_IRUGO, proc_tid_maps_operations), |
2e13ba54 | 3312 | #ifdef CONFIG_PROC_CHILDREN |
81841161 CG |
3313 | REG("children", S_IRUGO, proc_tid_children_operations), |
3314 | #endif | |
28a6d671 | 3315 | #ifdef CONFIG_NUMA |
b7643757 | 3316 | REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations), |
28a6d671 | 3317 | #endif |
631f9c18 AD |
3318 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
3319 | LNK("cwd", proc_cwd_link), | |
3320 | LNK("root", proc_root_link), | |
3321 | LNK("exe", proc_exe_link), | |
3322 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
3323 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
1e883281 | 3324 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 3325 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
b7643757 | 3326 | REG("smaps", S_IRUGO, proc_tid_smaps_operations), |
493b0e9d | 3327 | REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations), |
32ed74a4 | 3328 | REG("pagemap", S_IRUSR, proc_pagemap_operations), |
28a6d671 EB |
3329 | #endif |
3330 | #ifdef CONFIG_SECURITY | |
631f9c18 | 3331 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
3332 | #endif |
3333 | #ifdef CONFIG_KALLSYMS | |
edfcd606 | 3334 | ONE("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 3335 | #endif |
2ec220e2 | 3336 | #ifdef CONFIG_STACKTRACE |
35a35046 | 3337 | ONE("stack", S_IRUSR, proc_pid_stack), |
28a6d671 | 3338 | #endif |
5968cece | 3339 | #ifdef CONFIG_SCHED_INFO |
f6e826ca | 3340 | ONE("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 3341 | #endif |
9745512c | 3342 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 3343 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 3344 | #endif |
8793d854 | 3345 | #ifdef CONFIG_PROC_PID_CPUSET |
52de4779 | 3346 | ONE("cpuset", S_IRUGO, proc_cpuset_show), |
a424316c PM |
3347 | #endif |
3348 | #ifdef CONFIG_CGROUPS | |
006f4ac4 | 3349 | ONE("cgroup", S_IRUGO, proc_cgroup_show), |
28a6d671 | 3350 | #endif |
6ba51e37 | 3351 | ONE("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 3352 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 3353 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
28a6d671 | 3354 | #ifdef CONFIG_AUDITSYSCALL |
631f9c18 | 3355 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
26ec3c64 | 3356 | REG("sessionid", S_IRUGO, proc_sessionid_operations), |
28a6d671 | 3357 | #endif |
f4f154fd | 3358 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 3359 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
1203c8e6 | 3360 | REG("fail-nth", 0644, proc_fail_nth_operations), |
f4f154fd | 3361 | #endif |
297c5d92 | 3362 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 3363 | ONE("io", S_IRUSR, proc_tid_io_accounting), |
297c5d92 | 3364 | #endif |
f133ecca | 3365 | #ifdef CONFIG_HARDWALL |
d962c144 | 3366 | ONE("hardwall", S_IRUGO, proc_pid_hardwall), |
f133ecca | 3367 | #endif |
22d917d8 EB |
3368 | #ifdef CONFIG_USER_NS |
3369 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
3370 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 3371 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
9cc46516 | 3372 | REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations), |
22d917d8 | 3373 | #endif |
7c23b330 JP |
3374 | #ifdef CONFIG_LIVEPATCH |
3375 | ONE("patch_state", S_IRUSR, proc_pid_patch_state), | |
3376 | #endif | |
28a6d671 EB |
3377 | }; |
3378 | ||
f0c3b509 | 3379 | static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx) |
28a6d671 | 3380 | { |
f0c3b509 AV |
3381 | return proc_pident_readdir(file, ctx, |
3382 | tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
3383 | } |
3384 | ||
00cd8dd3 AV |
3385 | static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
3386 | { | |
7bcd6b0e EB |
3387 | return proc_pident_lookup(dir, dentry, |
3388 | tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
3389 | } |
3390 | ||
00977a59 | 3391 | static const struct file_operations proc_tid_base_operations = { |
28a6d671 | 3392 | .read = generic_read_dir, |
f50752ea AV |
3393 | .iterate_shared = proc_tid_base_readdir, |
3394 | .llseek = generic_file_llseek, | |
28a6d671 EB |
3395 | }; |
3396 | ||
c5ef1c42 | 3397 | static const struct inode_operations proc_tid_base_inode_operations = { |
28a6d671 EB |
3398 | .lookup = proc_tid_base_lookup, |
3399 | .getattr = pid_getattr, | |
3400 | .setattr = proc_setattr, | |
3401 | }; | |
3402 | ||
c52a47ac | 3403 | static int proc_task_instantiate(struct inode *dir, |
c5141e6d | 3404 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
444ceed8 | 3405 | { |
444ceed8 | 3406 | struct inode *inode; |
db978da8 | 3407 | inode = proc_pid_make_inode(dir->i_sb, task, S_IFDIR | S_IRUGO | S_IXUGO); |
444ceed8 EB |
3408 | |
3409 | if (!inode) | |
3410 | goto out; | |
444ceed8 EB |
3411 | inode->i_op = &proc_tid_base_inode_operations; |
3412 | inode->i_fop = &proc_tid_base_operations; | |
3413 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 | 3414 | |
1270dd8d | 3415 | set_nlink(inode, nlink_tid); |
444ceed8 | 3416 | |
fb045adb | 3417 | d_set_d_op(dentry, &pid_dentry_operations); |
444ceed8 EB |
3418 | |
3419 | d_add(dentry, inode); | |
3420 | /* Close the race of the process dying before we return the dentry */ | |
0b728e19 | 3421 | if (pid_revalidate(dentry, 0)) |
c52a47ac | 3422 | return 0; |
444ceed8 | 3423 | out: |
c52a47ac | 3424 | return -ENOENT; |
444ceed8 EB |
3425 | } |
3426 | ||
00cd8dd3 | 3427 | static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags) |
28a6d671 | 3428 | { |
c52a47ac | 3429 | int result = -ENOENT; |
28a6d671 EB |
3430 | struct task_struct *task; |
3431 | struct task_struct *leader = get_proc_task(dir); | |
28a6d671 | 3432 | unsigned tid; |
b488893a | 3433 | struct pid_namespace *ns; |
28a6d671 EB |
3434 | |
3435 | if (!leader) | |
3436 | goto out_no_task; | |
3437 | ||
dbcdb504 | 3438 | tid = name_to_int(&dentry->d_name); |
28a6d671 EB |
3439 | if (tid == ~0U) |
3440 | goto out; | |
3441 | ||
b488893a | 3442 | ns = dentry->d_sb->s_fs_info; |
28a6d671 | 3443 | rcu_read_lock(); |
b488893a | 3444 | task = find_task_by_pid_ns(tid, ns); |
28a6d671 EB |
3445 | if (task) |
3446 | get_task_struct(task); | |
3447 | rcu_read_unlock(); | |
3448 | if (!task) | |
3449 | goto out; | |
bac0abd6 | 3450 | if (!same_thread_group(leader, task)) |
28a6d671 EB |
3451 | goto out_drop_task; |
3452 | ||
444ceed8 | 3453 | result = proc_task_instantiate(dir, dentry, task, NULL); |
28a6d671 EB |
3454 | out_drop_task: |
3455 | put_task_struct(task); | |
3456 | out: | |
3457 | put_task_struct(leader); | |
3458 | out_no_task: | |
c52a47ac | 3459 | return ERR_PTR(result); |
28a6d671 EB |
3460 | } |
3461 | ||
0bc58a91 EB |
3462 | /* |
3463 | * Find the first tid of a thread group to return to user space. | |
3464 | * | |
3465 | * Usually this is just the thread group leader, but if the users | |
3466 | * buffer was too small or there was a seek into the middle of the | |
3467 | * directory we have more work todo. | |
3468 | * | |
3469 | * In the case of a short read we start with find_task_by_pid. | |
3470 | * | |
3471 | * In the case of a seek we start with the leader and walk nr | |
3472 | * threads past it. | |
3473 | */ | |
9f6e963f ON |
3474 | static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos, |
3475 | struct pid_namespace *ns) | |
0bc58a91 | 3476 | { |
d855a4b7 | 3477 | struct task_struct *pos, *task; |
9f6e963f ON |
3478 | unsigned long nr = f_pos; |
3479 | ||
3480 | if (nr != f_pos) /* 32bit overflow? */ | |
3481 | return NULL; | |
1da177e4 | 3482 | |
cc288738 | 3483 | rcu_read_lock(); |
d855a4b7 ON |
3484 | task = pid_task(pid, PIDTYPE_PID); |
3485 | if (!task) | |
3486 | goto fail; | |
3487 | ||
3488 | /* Attempt to start with the tid of a thread */ | |
9f6e963f | 3489 | if (tid && nr) { |
b488893a | 3490 | pos = find_task_by_pid_ns(tid, ns); |
d855a4b7 | 3491 | if (pos && same_thread_group(pos, task)) |
a872ff0c | 3492 | goto found; |
0bc58a91 | 3493 | } |
1da177e4 | 3494 | |
0bc58a91 | 3495 | /* If nr exceeds the number of threads there is nothing todo */ |
9f6e963f | 3496 | if (nr >= get_nr_threads(task)) |
c986c14a | 3497 | goto fail; |
1da177e4 | 3498 | |
a872ff0c ON |
3499 | /* If we haven't found our starting place yet start |
3500 | * with the leader and walk nr threads forward. | |
0bc58a91 | 3501 | */ |
d855a4b7 | 3502 | pos = task = task->group_leader; |
c986c14a | 3503 | do { |
9f6e963f | 3504 | if (!nr--) |
c986c14a | 3505 | goto found; |
d855a4b7 | 3506 | } while_each_thread(task, pos); |
c986c14a ON |
3507 | fail: |
3508 | pos = NULL; | |
3509 | goto out; | |
a872ff0c ON |
3510 | found: |
3511 | get_task_struct(pos); | |
3512 | out: | |
cc288738 | 3513 | rcu_read_unlock(); |
0bc58a91 EB |
3514 | return pos; |
3515 | } | |
3516 | ||
3517 | /* | |
3518 | * Find the next thread in the thread list. | |
3519 | * Return NULL if there is an error or no next thread. | |
3520 | * | |
3521 | * The reference to the input task_struct is released. | |
3522 | */ | |
3523 | static struct task_struct *next_tid(struct task_struct *start) | |
3524 | { | |
c1df7fb8 | 3525 | struct task_struct *pos = NULL; |
cc288738 | 3526 | rcu_read_lock(); |
c1df7fb8 | 3527 | if (pid_alive(start)) { |
0bc58a91 | 3528 | pos = next_thread(start); |
c1df7fb8 ON |
3529 | if (thread_group_leader(pos)) |
3530 | pos = NULL; | |
3531 | else | |
3532 | get_task_struct(pos); | |
3533 | } | |
cc288738 | 3534 | rcu_read_unlock(); |
0bc58a91 EB |
3535 | put_task_struct(start); |
3536 | return pos; | |
1da177e4 LT |
3537 | } |
3538 | ||
3539 | /* for the /proc/TGID/task/ directories */ | |
f0c3b509 | 3540 | static int proc_task_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3541 | { |
d855a4b7 ON |
3542 | struct inode *inode = file_inode(file); |
3543 | struct task_struct *task; | |
b488893a | 3544 | struct pid_namespace *ns; |
f0c3b509 | 3545 | int tid; |
1da177e4 | 3546 | |
d855a4b7 | 3547 | if (proc_inode_is_dead(inode)) |
f0c3b509 | 3548 | return -ENOENT; |
1da177e4 | 3549 | |
f0c3b509 | 3550 | if (!dir_emit_dots(file, ctx)) |
d855a4b7 | 3551 | return 0; |
1da177e4 | 3552 | |
0bc58a91 EB |
3553 | /* f_version caches the tgid value that the last readdir call couldn't |
3554 | * return. lseek aka telldir automagically resets f_version to 0. | |
3555 | */ | |
3aa3377f | 3556 | ns = inode->i_sb->s_fs_info; |
f0c3b509 AV |
3557 | tid = (int)file->f_version; |
3558 | file->f_version = 0; | |
d855a4b7 | 3559 | for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns); |
0bc58a91 | 3560 | task; |
f0c3b509 | 3561 | task = next_tid(task), ctx->pos++) { |
e3912ac3 | 3562 | char name[10 + 1]; |
f0c3b509 | 3563 | int len; |
b488893a | 3564 | tid = task_pid_nr_ns(task, ns); |
e3912ac3 | 3565 | len = snprintf(name, sizeof(name), "%u", tid); |
f0c3b509 AV |
3566 | if (!proc_fill_cache(file, ctx, name, len, |
3567 | proc_task_instantiate, task, NULL)) { | |
0bc58a91 EB |
3568 | /* returning this tgid failed, save it as the first |
3569 | * pid for the next readir call */ | |
f0c3b509 | 3570 | file->f_version = (u64)tid; |
0bc58a91 | 3571 | put_task_struct(task); |
1da177e4 | 3572 | break; |
0bc58a91 | 3573 | } |
1da177e4 | 3574 | } |
d855a4b7 | 3575 | |
f0c3b509 | 3576 | return 0; |
1da177e4 | 3577 | } |
6e66b52b | 3578 | |
a528d35e DH |
3579 | static int proc_task_getattr(const struct path *path, struct kstat *stat, |
3580 | u32 request_mask, unsigned int query_flags) | |
6e66b52b | 3581 | { |
a528d35e | 3582 | struct inode *inode = d_inode(path->dentry); |
99f89551 | 3583 | struct task_struct *p = get_proc_task(inode); |
6e66b52b EB |
3584 | generic_fillattr(inode, stat); |
3585 | ||
99f89551 | 3586 | if (p) { |
99f89551 | 3587 | stat->nlink += get_nr_threads(p); |
99f89551 | 3588 | put_task_struct(p); |
6e66b52b EB |
3589 | } |
3590 | ||
3591 | return 0; | |
3592 | } | |
28a6d671 | 3593 | |
c5ef1c42 | 3594 | static const struct inode_operations proc_task_inode_operations = { |
28a6d671 EB |
3595 | .lookup = proc_task_lookup, |
3596 | .getattr = proc_task_getattr, | |
3597 | .setattr = proc_setattr, | |
0499680a | 3598 | .permission = proc_pid_permission, |
28a6d671 EB |
3599 | }; |
3600 | ||
00977a59 | 3601 | static const struct file_operations proc_task_operations = { |
28a6d671 | 3602 | .read = generic_read_dir, |
f50752ea AV |
3603 | .iterate_shared = proc_task_readdir, |
3604 | .llseek = generic_file_llseek, | |
28a6d671 | 3605 | }; |
1270dd8d AD |
3606 | |
3607 | void __init set_proc_pid_nlink(void) | |
3608 | { | |
3609 | nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
3610 | nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
3611 | } |