]>
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++) { | |
8b927d73 | 446 | seq_printf(m, "[<%pK>] %pB\n", |
a9712bc1 AV |
447 | (void *)entries[i], (void *)entries[i]); |
448 | } | |
449 | unlock_trace(task); | |
2ec220e2 KC |
450 | } |
451 | kfree(entries); | |
452 | ||
a9712bc1 | 453 | return err; |
2ec220e2 KC |
454 | } |
455 | #endif | |
456 | ||
5968cece | 457 | #ifdef CONFIG_SCHED_INFO |
1da177e4 LT |
458 | /* |
459 | * Provides /proc/PID/schedstat | |
460 | */ | |
f6e826ca AD |
461 | static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns, |
462 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 463 | { |
5968cece NR |
464 | if (unlikely(!sched_info_on())) |
465 | seq_printf(m, "0 0 0\n"); | |
466 | else | |
467 | seq_printf(m, "%llu %llu %lu\n", | |
25ce3191 JP |
468 | (unsigned long long)task->se.sum_exec_runtime, |
469 | (unsigned long long)task->sched_info.run_delay, | |
470 | task->sched_info.pcount); | |
471 | ||
472 | return 0; | |
1da177e4 LT |
473 | } |
474 | #endif | |
475 | ||
9745512c AV |
476 | #ifdef CONFIG_LATENCYTOP |
477 | static int lstats_show_proc(struct seq_file *m, void *v) | |
478 | { | |
479 | int i; | |
13d77c37 HS |
480 | struct inode *inode = m->private; |
481 | struct task_struct *task = get_proc_task(inode); | |
9745512c | 482 | |
13d77c37 HS |
483 | if (!task) |
484 | return -ESRCH; | |
485 | seq_puts(m, "Latency Top version : v0.1\n"); | |
9745512c | 486 | for (i = 0; i < 32; i++) { |
34e49d4f JP |
487 | struct latency_record *lr = &task->latency_record[i]; |
488 | if (lr->backtrace[0]) { | |
9745512c | 489 | int q; |
34e49d4f JP |
490 | seq_printf(m, "%i %li %li", |
491 | lr->count, lr->time, lr->max); | |
9745512c | 492 | for (q = 0; q < LT_BACKTRACEDEPTH; q++) { |
34e49d4f JP |
493 | unsigned long bt = lr->backtrace[q]; |
494 | if (!bt) | |
9745512c | 495 | break; |
34e49d4f | 496 | if (bt == ULONG_MAX) |
9745512c | 497 | break; |
34e49d4f | 498 | seq_printf(m, " %ps", (void *)bt); |
9745512c | 499 | } |
9d6de12f | 500 | seq_putc(m, '\n'); |
9745512c AV |
501 | } |
502 | ||
503 | } | |
13d77c37 | 504 | put_task_struct(task); |
9745512c AV |
505 | return 0; |
506 | } | |
507 | ||
508 | static int lstats_open(struct inode *inode, struct file *file) | |
509 | { | |
13d77c37 | 510 | return single_open(file, lstats_show_proc, inode); |
d6643d12 HS |
511 | } |
512 | ||
9745512c AV |
513 | static ssize_t lstats_write(struct file *file, const char __user *buf, |
514 | size_t count, loff_t *offs) | |
515 | { | |
496ad9aa | 516 | struct task_struct *task = get_proc_task(file_inode(file)); |
9745512c | 517 | |
13d77c37 HS |
518 | if (!task) |
519 | return -ESRCH; | |
9745512c | 520 | clear_all_latency_tracing(task); |
13d77c37 | 521 | put_task_struct(task); |
9745512c AV |
522 | |
523 | return count; | |
524 | } | |
525 | ||
526 | static const struct file_operations proc_lstats_operations = { | |
527 | .open = lstats_open, | |
528 | .read = seq_read, | |
529 | .write = lstats_write, | |
530 | .llseek = seq_lseek, | |
13d77c37 | 531 | .release = single_release, |
9745512c AV |
532 | }; |
533 | ||
534 | #endif | |
535 | ||
6ba51e37 AD |
536 | static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns, |
537 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 538 | { |
a7f638f9 | 539 | unsigned long totalpages = totalram_pages + total_swap_pages; |
b95c35e7 | 540 | unsigned long points = 0; |
1da177e4 | 541 | |
ef419398 ON |
542 | points = oom_badness(task, NULL, NULL, totalpages) * |
543 | 1000 / totalpages; | |
25ce3191 JP |
544 | seq_printf(m, "%lu\n", points); |
545 | ||
546 | return 0; | |
1da177e4 LT |
547 | } |
548 | ||
d85f50d5 | 549 | struct limit_names { |
cedbccab AD |
550 | const char *name; |
551 | const char *unit; | |
d85f50d5 NH |
552 | }; |
553 | ||
554 | static const struct limit_names lnames[RLIM_NLIMITS] = { | |
cff4edb5 | 555 | [RLIMIT_CPU] = {"Max cpu time", "seconds"}, |
d85f50d5 NH |
556 | [RLIMIT_FSIZE] = {"Max file size", "bytes"}, |
557 | [RLIMIT_DATA] = {"Max data size", "bytes"}, | |
558 | [RLIMIT_STACK] = {"Max stack size", "bytes"}, | |
559 | [RLIMIT_CORE] = {"Max core file size", "bytes"}, | |
560 | [RLIMIT_RSS] = {"Max resident set", "bytes"}, | |
561 | [RLIMIT_NPROC] = {"Max processes", "processes"}, | |
562 | [RLIMIT_NOFILE] = {"Max open files", "files"}, | |
563 | [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"}, | |
564 | [RLIMIT_AS] = {"Max address space", "bytes"}, | |
565 | [RLIMIT_LOCKS] = {"Max file locks", "locks"}, | |
566 | [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"}, | |
567 | [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"}, | |
568 | [RLIMIT_NICE] = {"Max nice priority", NULL}, | |
569 | [RLIMIT_RTPRIO] = {"Max realtime priority", NULL}, | |
8808117c | 570 | [RLIMIT_RTTIME] = {"Max realtime timeout", "us"}, |
d85f50d5 NH |
571 | }; |
572 | ||
573 | /* Display limits for a process */ | |
1c963eb1 AD |
574 | static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns, |
575 | struct pid *pid, struct task_struct *task) | |
d85f50d5 NH |
576 | { |
577 | unsigned int i; | |
d85f50d5 | 578 | unsigned long flags; |
d85f50d5 NH |
579 | |
580 | struct rlimit rlim[RLIM_NLIMITS]; | |
581 | ||
a6bebbc8 | 582 | if (!lock_task_sighand(task, &flags)) |
d85f50d5 | 583 | return 0; |
d85f50d5 NH |
584 | memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS); |
585 | unlock_task_sighand(task, &flags); | |
d85f50d5 NH |
586 | |
587 | /* | |
588 | * print the file header | |
589 | */ | |
1c963eb1 | 590 | seq_printf(m, "%-25s %-20s %-20s %-10s\n", |
25ce3191 | 591 | "Limit", "Soft Limit", "Hard Limit", "Units"); |
d85f50d5 NH |
592 | |
593 | for (i = 0; i < RLIM_NLIMITS; i++) { | |
594 | if (rlim[i].rlim_cur == RLIM_INFINITY) | |
1c963eb1 | 595 | seq_printf(m, "%-25s %-20s ", |
25ce3191 | 596 | lnames[i].name, "unlimited"); |
d85f50d5 | 597 | else |
1c963eb1 | 598 | seq_printf(m, "%-25s %-20lu ", |
25ce3191 | 599 | lnames[i].name, rlim[i].rlim_cur); |
d85f50d5 NH |
600 | |
601 | if (rlim[i].rlim_max == RLIM_INFINITY) | |
1c963eb1 | 602 | seq_printf(m, "%-20s ", "unlimited"); |
d85f50d5 | 603 | else |
1c963eb1 | 604 | seq_printf(m, "%-20lu ", rlim[i].rlim_max); |
d85f50d5 NH |
605 | |
606 | if (lnames[i].unit) | |
1c963eb1 | 607 | seq_printf(m, "%-10s\n", lnames[i].unit); |
d85f50d5 | 608 | else |
1c963eb1 | 609 | seq_putc(m, '\n'); |
d85f50d5 NH |
610 | } |
611 | ||
1c963eb1 | 612 | return 0; |
d85f50d5 NH |
613 | } |
614 | ||
ebcb6734 | 615 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 AD |
616 | static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns, |
617 | struct pid *pid, struct task_struct *task) | |
ebcb6734 RM |
618 | { |
619 | long nr; | |
620 | unsigned long args[6], sp, pc; | |
25ce3191 JP |
621 | int res; |
622 | ||
623 | res = lock_trace(task); | |
a9712bc1 AV |
624 | if (res) |
625 | return res; | |
ebcb6734 RM |
626 | |
627 | if (task_current_syscall(task, &nr, args, 6, &sp, &pc)) | |
09d93bd6 | 628 | seq_puts(m, "running\n"); |
a9712bc1 | 629 | else if (nr < 0) |
09d93bd6 | 630 | seq_printf(m, "%ld 0x%lx 0x%lx\n", nr, sp, pc); |
a9712bc1 | 631 | else |
09d93bd6 | 632 | seq_printf(m, |
ebcb6734 RM |
633 | "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n", |
634 | nr, | |
635 | args[0], args[1], args[2], args[3], args[4], args[5], | |
636 | sp, pc); | |
a9712bc1 | 637 | unlock_trace(task); |
25ce3191 JP |
638 | |
639 | return 0; | |
ebcb6734 RM |
640 | } |
641 | #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */ | |
642 | ||
1da177e4 LT |
643 | /************************************************************************/ |
644 | /* Here the fs part begins */ | |
645 | /************************************************************************/ | |
646 | ||
647 | /* permission checks */ | |
778c1144 | 648 | static int proc_fd_access_allowed(struct inode *inode) |
1da177e4 | 649 | { |
778c1144 EB |
650 | struct task_struct *task; |
651 | int allowed = 0; | |
df26c40e EB |
652 | /* Allow access to a task's file descriptors if it is us or we |
653 | * may use ptrace attach to the process and find out that | |
654 | * information. | |
778c1144 EB |
655 | */ |
656 | task = get_proc_task(inode); | |
df26c40e | 657 | if (task) { |
caaee623 | 658 | allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS); |
778c1144 | 659 | put_task_struct(task); |
df26c40e | 660 | } |
778c1144 | 661 | return allowed; |
1da177e4 LT |
662 | } |
663 | ||
6b4e306a | 664 | int proc_setattr(struct dentry *dentry, struct iattr *attr) |
6d76fa58 LT |
665 | { |
666 | int error; | |
2b0143b5 | 667 | struct inode *inode = d_inode(dentry); |
6d76fa58 LT |
668 | |
669 | if (attr->ia_valid & ATTR_MODE) | |
670 | return -EPERM; | |
671 | ||
31051c85 | 672 | error = setattr_prepare(dentry, attr); |
1025774c CH |
673 | if (error) |
674 | return error; | |
675 | ||
1025774c CH |
676 | setattr_copy(inode, attr); |
677 | mark_inode_dirty(inode); | |
678 | return 0; | |
6d76fa58 LT |
679 | } |
680 | ||
0499680a VK |
681 | /* |
682 | * May current process learn task's sched/cmdline info (for hide_pid_min=1) | |
683 | * or euid/egid (for hide_pid_min=2)? | |
684 | */ | |
685 | static bool has_pid_permissions(struct pid_namespace *pid, | |
686 | struct task_struct *task, | |
687 | int hide_pid_min) | |
688 | { | |
689 | if (pid->hide_pid < hide_pid_min) | |
690 | return true; | |
691 | if (in_group_p(pid->pid_gid)) | |
692 | return true; | |
caaee623 | 693 | return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS); |
0499680a VK |
694 | } |
695 | ||
696 | ||
697 | static int proc_pid_permission(struct inode *inode, int mask) | |
698 | { | |
699 | struct pid_namespace *pid = inode->i_sb->s_fs_info; | |
700 | struct task_struct *task; | |
701 | bool has_perms; | |
702 | ||
703 | task = get_proc_task(inode); | |
a2ef990a XF |
704 | if (!task) |
705 | return -ESRCH; | |
796f571b | 706 | has_perms = has_pid_permissions(pid, task, HIDEPID_NO_ACCESS); |
0499680a VK |
707 | put_task_struct(task); |
708 | ||
709 | if (!has_perms) { | |
796f571b | 710 | if (pid->hide_pid == HIDEPID_INVISIBLE) { |
0499680a VK |
711 | /* |
712 | * Let's make getdents(), stat(), and open() | |
713 | * consistent with each other. If a process | |
714 | * may not stat() a file, it shouldn't be seen | |
715 | * in procfs at all. | |
716 | */ | |
717 | return -ENOENT; | |
718 | } | |
719 | ||
720 | return -EPERM; | |
721 | } | |
722 | return generic_permission(inode, mask); | |
723 | } | |
724 | ||
725 | ||
726 | ||
c5ef1c42 | 727 | static const struct inode_operations proc_def_inode_operations = { |
6d76fa58 LT |
728 | .setattr = proc_setattr, |
729 | }; | |
730 | ||
be614086 EB |
731 | static int proc_single_show(struct seq_file *m, void *v) |
732 | { | |
733 | struct inode *inode = m->private; | |
734 | struct pid_namespace *ns; | |
735 | struct pid *pid; | |
736 | struct task_struct *task; | |
737 | int ret; | |
738 | ||
739 | ns = inode->i_sb->s_fs_info; | |
740 | pid = proc_pid(inode); | |
741 | task = get_pid_task(pid, PIDTYPE_PID); | |
742 | if (!task) | |
743 | return -ESRCH; | |
744 | ||
745 | ret = PROC_I(inode)->op.proc_show(m, ns, pid, task); | |
746 | ||
747 | put_task_struct(task); | |
748 | return ret; | |
749 | } | |
750 | ||
751 | static int proc_single_open(struct inode *inode, struct file *filp) | |
752 | { | |
c6a34058 | 753 | return single_open(filp, proc_single_show, inode); |
be614086 EB |
754 | } |
755 | ||
756 | static const struct file_operations proc_single_file_operations = { | |
757 | .open = proc_single_open, | |
758 | .read = seq_read, | |
759 | .llseek = seq_lseek, | |
760 | .release = single_release, | |
761 | }; | |
762 | ||
5381e169 ON |
763 | |
764 | struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode) | |
1da177e4 | 765 | { |
5381e169 ON |
766 | struct task_struct *task = get_proc_task(inode); |
767 | struct mm_struct *mm = ERR_PTR(-ESRCH); | |
e268337d | 768 | |
5381e169 | 769 | if (task) { |
caaee623 | 770 | mm = mm_access(task, mode | PTRACE_MODE_FSCREDS); |
5381e169 | 771 | put_task_struct(task); |
e268337d | 772 | |
5381e169 ON |
773 | if (!IS_ERR_OR_NULL(mm)) { |
774 | /* ensure this mm_struct can't be freed */ | |
f1f10076 | 775 | mmgrab(mm); |
5381e169 ON |
776 | /* but do not pin its memory */ |
777 | mmput(mm); | |
778 | } | |
779 | } | |
780 | ||
781 | return mm; | |
782 | } | |
783 | ||
784 | static int __mem_open(struct inode *inode, struct file *file, unsigned int mode) | |
785 | { | |
786 | struct mm_struct *mm = proc_mem_open(inode, mode); | |
e268337d LT |
787 | |
788 | if (IS_ERR(mm)) | |
789 | return PTR_ERR(mm); | |
790 | ||
e268337d | 791 | file->private_data = mm; |
1da177e4 LT |
792 | return 0; |
793 | } | |
794 | ||
b409e578 CW |
795 | static int mem_open(struct inode *inode, struct file *file) |
796 | { | |
bc452b4b DH |
797 | int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH); |
798 | ||
799 | /* OK to pass negative loff_t, we can catch out-of-range */ | |
800 | file->f_mode |= FMODE_UNSIGNED_OFFSET; | |
801 | ||
802 | return ret; | |
b409e578 CW |
803 | } |
804 | ||
572d34b9 ON |
805 | static ssize_t mem_rw(struct file *file, char __user *buf, |
806 | size_t count, loff_t *ppos, int write) | |
1da177e4 | 807 | { |
e268337d | 808 | struct mm_struct *mm = file->private_data; |
572d34b9 ON |
809 | unsigned long addr = *ppos; |
810 | ssize_t copied; | |
1da177e4 | 811 | char *page; |
272ddc8b | 812 | unsigned int flags; |
1da177e4 | 813 | |
e268337d LT |
814 | if (!mm) |
815 | return 0; | |
99f89551 | 816 | |
0ee931c4 | 817 | page = (char *)__get_free_page(GFP_KERNEL); |
30cd8903 | 818 | if (!page) |
e268337d | 819 | return -ENOMEM; |
1da177e4 | 820 | |
f7ca54f4 | 821 | copied = 0; |
388f7934 | 822 | if (!mmget_not_zero(mm)) |
6d08f2c7 ON |
823 | goto free; |
824 | ||
f511c0b1 | 825 | flags = FOLL_FORCE | (write ? FOLL_WRITE : 0); |
6347e8d5 | 826 | |
1da177e4 | 827 | while (count > 0) { |
572d34b9 | 828 | int this_len = min_t(int, count, PAGE_SIZE); |
1da177e4 | 829 | |
572d34b9 | 830 | if (write && copy_from_user(page, buf, this_len)) { |
1da177e4 LT |
831 | copied = -EFAULT; |
832 | break; | |
833 | } | |
572d34b9 | 834 | |
6347e8d5 | 835 | this_len = access_remote_vm(mm, addr, page, this_len, flags); |
572d34b9 | 836 | if (!this_len) { |
1da177e4 LT |
837 | if (!copied) |
838 | copied = -EIO; | |
839 | break; | |
840 | } | |
572d34b9 ON |
841 | |
842 | if (!write && copy_to_user(buf, page, this_len)) { | |
843 | copied = -EFAULT; | |
844 | break; | |
845 | } | |
846 | ||
847 | buf += this_len; | |
848 | addr += this_len; | |
849 | copied += this_len; | |
850 | count -= this_len; | |
1da177e4 | 851 | } |
572d34b9 | 852 | *ppos = addr; |
30cd8903 | 853 | |
6d08f2c7 ON |
854 | mmput(mm); |
855 | free: | |
30cd8903 | 856 | free_page((unsigned long) page); |
1da177e4 LT |
857 | return copied; |
858 | } | |
1da177e4 | 859 | |
572d34b9 ON |
860 | static ssize_t mem_read(struct file *file, char __user *buf, |
861 | size_t count, loff_t *ppos) | |
862 | { | |
863 | return mem_rw(file, buf, count, ppos, 0); | |
864 | } | |
865 | ||
866 | static ssize_t mem_write(struct file *file, const char __user *buf, | |
867 | size_t count, loff_t *ppos) | |
868 | { | |
869 | return mem_rw(file, (char __user*)buf, count, ppos, 1); | |
870 | } | |
871 | ||
85863e47 | 872 | loff_t mem_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
873 | { |
874 | switch (orig) { | |
875 | case 0: | |
876 | file->f_pos = offset; | |
877 | break; | |
878 | case 1: | |
879 | file->f_pos += offset; | |
880 | break; | |
881 | default: | |
882 | return -EINVAL; | |
883 | } | |
884 | force_successful_syscall_return(); | |
885 | return file->f_pos; | |
886 | } | |
887 | ||
e268337d LT |
888 | static int mem_release(struct inode *inode, struct file *file) |
889 | { | |
890 | struct mm_struct *mm = file->private_data; | |
71879d3c | 891 | if (mm) |
6d08f2c7 | 892 | mmdrop(mm); |
e268337d LT |
893 | return 0; |
894 | } | |
895 | ||
00977a59 | 896 | static const struct file_operations proc_mem_operations = { |
1da177e4 LT |
897 | .llseek = mem_lseek, |
898 | .read = mem_read, | |
899 | .write = mem_write, | |
900 | .open = mem_open, | |
e268337d | 901 | .release = mem_release, |
1da177e4 LT |
902 | }; |
903 | ||
b409e578 CW |
904 | static int environ_open(struct inode *inode, struct file *file) |
905 | { | |
906 | return __mem_open(inode, file, PTRACE_MODE_READ); | |
907 | } | |
908 | ||
315e28c8 JP |
909 | static ssize_t environ_read(struct file *file, char __user *buf, |
910 | size_t count, loff_t *ppos) | |
911 | { | |
315e28c8 JP |
912 | char *page; |
913 | unsigned long src = *ppos; | |
b409e578 CW |
914 | int ret = 0; |
915 | struct mm_struct *mm = file->private_data; | |
a3b609ef | 916 | unsigned long env_start, env_end; |
315e28c8 | 917 | |
8148a73c MK |
918 | /* Ensure the process spawned far enough to have an environment. */ |
919 | if (!mm || !mm->env_end) | |
b409e578 | 920 | return 0; |
315e28c8 | 921 | |
0ee931c4 | 922 | page = (char *)__get_free_page(GFP_KERNEL); |
315e28c8 | 923 | if (!page) |
b409e578 | 924 | return -ENOMEM; |
315e28c8 | 925 | |
d6f64b89 | 926 | ret = 0; |
388f7934 | 927 | if (!mmget_not_zero(mm)) |
b409e578 | 928 | goto free; |
a3b609ef MG |
929 | |
930 | down_read(&mm->mmap_sem); | |
931 | env_start = mm->env_start; | |
932 | env_end = mm->env_end; | |
933 | up_read(&mm->mmap_sem); | |
934 | ||
315e28c8 | 935 | while (count > 0) { |
e8905ec2 DH |
936 | size_t this_len, max_len; |
937 | int retval; | |
315e28c8 | 938 | |
a3b609ef | 939 | if (src >= (env_end - env_start)) |
315e28c8 JP |
940 | break; |
941 | ||
a3b609ef | 942 | this_len = env_end - (env_start + src); |
e8905ec2 DH |
943 | |
944 | max_len = min_t(size_t, PAGE_SIZE, count); | |
945 | this_len = min(max_len, this_len); | |
315e28c8 | 946 | |
272ddc8b | 947 | retval = access_remote_vm(mm, (env_start + src), page, this_len, 0); |
315e28c8 JP |
948 | |
949 | if (retval <= 0) { | |
950 | ret = retval; | |
951 | break; | |
952 | } | |
953 | ||
954 | if (copy_to_user(buf, page, retval)) { | |
955 | ret = -EFAULT; | |
956 | break; | |
957 | } | |
958 | ||
959 | ret += retval; | |
960 | src += retval; | |
961 | buf += retval; | |
962 | count -= retval; | |
963 | } | |
964 | *ppos = src; | |
315e28c8 | 965 | mmput(mm); |
b409e578 CW |
966 | |
967 | free: | |
315e28c8 | 968 | free_page((unsigned long) page); |
315e28c8 JP |
969 | return ret; |
970 | } | |
971 | ||
972 | static const struct file_operations proc_environ_operations = { | |
b409e578 | 973 | .open = environ_open, |
315e28c8 | 974 | .read = environ_read, |
87df8424 | 975 | .llseek = generic_file_llseek, |
b409e578 | 976 | .release = mem_release, |
315e28c8 JP |
977 | }; |
978 | ||
c5317167 AV |
979 | static int auxv_open(struct inode *inode, struct file *file) |
980 | { | |
981 | return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS); | |
982 | } | |
983 | ||
984 | static ssize_t auxv_read(struct file *file, char __user *buf, | |
985 | size_t count, loff_t *ppos) | |
986 | { | |
987 | struct mm_struct *mm = file->private_data; | |
988 | unsigned int nwords = 0; | |
06b2849d LY |
989 | |
990 | if (!mm) | |
991 | return 0; | |
c5317167 AV |
992 | do { |
993 | nwords += 2; | |
994 | } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */ | |
995 | return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv, | |
996 | nwords * sizeof(mm->saved_auxv[0])); | |
997 | } | |
998 | ||
999 | static const struct file_operations proc_auxv_operations = { | |
1000 | .open = auxv_open, | |
1001 | .read = auxv_read, | |
1002 | .llseek = generic_file_llseek, | |
1003 | .release = mem_release, | |
1004 | }; | |
1005 | ||
fa0cbbf1 DR |
1006 | static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count, |
1007 | loff_t *ppos) | |
1008 | { | |
496ad9aa | 1009 | struct task_struct *task = get_proc_task(file_inode(file)); |
fa0cbbf1 DR |
1010 | char buffer[PROC_NUMBUF]; |
1011 | int oom_adj = OOM_ADJUST_MIN; | |
1012 | size_t len; | |
fa0cbbf1 DR |
1013 | |
1014 | if (!task) | |
1015 | return -ESRCH; | |
f913da59 MH |
1016 | if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX) |
1017 | oom_adj = OOM_ADJUST_MAX; | |
1018 | else | |
1019 | oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) / | |
1020 | OOM_SCORE_ADJ_MAX; | |
fa0cbbf1 DR |
1021 | put_task_struct(task); |
1022 | len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj); | |
1023 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
1024 | } | |
1025 | ||
1d5f0acb MH |
1026 | static int __set_oom_adj(struct file *file, int oom_adj, bool legacy) |
1027 | { | |
1028 | static DEFINE_MUTEX(oom_adj_mutex); | |
44a70ade | 1029 | struct mm_struct *mm = NULL; |
1d5f0acb MH |
1030 | struct task_struct *task; |
1031 | int err = 0; | |
1032 | ||
1033 | task = get_proc_task(file_inode(file)); | |
1034 | if (!task) | |
1035 | return -ESRCH; | |
1036 | ||
1037 | mutex_lock(&oom_adj_mutex); | |
1038 | if (legacy) { | |
1039 | if (oom_adj < task->signal->oom_score_adj && | |
1040 | !capable(CAP_SYS_RESOURCE)) { | |
1041 | err = -EACCES; | |
1042 | goto err_unlock; | |
1043 | } | |
1044 | /* | |
1045 | * /proc/pid/oom_adj is provided for legacy purposes, ask users to use | |
1046 | * /proc/pid/oom_score_adj instead. | |
1047 | */ | |
1048 | pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n", | |
1049 | current->comm, task_pid_nr(current), task_pid_nr(task), | |
1050 | task_pid_nr(task)); | |
1051 | } else { | |
1052 | if ((short)oom_adj < task->signal->oom_score_adj_min && | |
1053 | !capable(CAP_SYS_RESOURCE)) { | |
1054 | err = -EACCES; | |
1055 | goto err_unlock; | |
1056 | } | |
1057 | } | |
1058 | ||
44a70ade MH |
1059 | /* |
1060 | * Make sure we will check other processes sharing the mm if this is | |
1061 | * not vfrok which wants its own oom_score_adj. | |
1062 | * pin the mm so it doesn't go away and get reused after task_unlock | |
1063 | */ | |
1064 | if (!task->vfork_done) { | |
1065 | struct task_struct *p = find_lock_task_mm(task); | |
1066 | ||
1067 | if (p) { | |
1068 | if (atomic_read(&p->mm->mm_users) > 1) { | |
1069 | mm = p->mm; | |
f1f10076 | 1070 | mmgrab(mm); |
44a70ade MH |
1071 | } |
1072 | task_unlock(p); | |
1073 | } | |
1074 | } | |
1075 | ||
1d5f0acb MH |
1076 | task->signal->oom_score_adj = oom_adj; |
1077 | if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE)) | |
1078 | task->signal->oom_score_adj_min = (short)oom_adj; | |
1079 | trace_oom_score_adj_update(task); | |
44a70ade MH |
1080 | |
1081 | if (mm) { | |
1082 | struct task_struct *p; | |
1083 | ||
1084 | rcu_read_lock(); | |
1085 | for_each_process(p) { | |
1086 | if (same_thread_group(task, p)) | |
1087 | continue; | |
1088 | ||
1089 | /* do not touch kernel threads or the global init */ | |
1090 | if (p->flags & PF_KTHREAD || is_global_init(p)) | |
1091 | continue; | |
1092 | ||
1093 | task_lock(p); | |
1094 | if (!p->vfork_done && process_shares_mm(p, mm)) { | |
1095 | 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", | |
1096 | task_pid_nr(p), p->comm, | |
1097 | p->signal->oom_score_adj, oom_adj, | |
1098 | task_pid_nr(task), task->comm); | |
1099 | p->signal->oom_score_adj = oom_adj; | |
1100 | if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE)) | |
1101 | p->signal->oom_score_adj_min = (short)oom_adj; | |
1102 | } | |
1103 | task_unlock(p); | |
1104 | } | |
1105 | rcu_read_unlock(); | |
1106 | mmdrop(mm); | |
1107 | } | |
1d5f0acb MH |
1108 | err_unlock: |
1109 | mutex_unlock(&oom_adj_mutex); | |
1110 | put_task_struct(task); | |
1111 | return err; | |
1112 | } | |
f913da59 | 1113 | |
b72bdfa7 DR |
1114 | /* |
1115 | * /proc/pid/oom_adj exists solely for backwards compatibility with previous | |
1116 | * kernels. The effective policy is defined by oom_score_adj, which has a | |
1117 | * different scale: oom_adj grew exponentially and oom_score_adj grows linearly. | |
1118 | * Values written to oom_adj are simply mapped linearly to oom_score_adj. | |
1119 | * Processes that become oom disabled via oom_adj will still be oom disabled | |
1120 | * with this implementation. | |
1121 | * | |
1122 | * oom_adj cannot be removed since existing userspace binaries use it. | |
1123 | */ | |
fa0cbbf1 DR |
1124 | static ssize_t oom_adj_write(struct file *file, const char __user *buf, |
1125 | size_t count, loff_t *ppos) | |
1126 | { | |
fa0cbbf1 DR |
1127 | char buffer[PROC_NUMBUF]; |
1128 | int oom_adj; | |
fa0cbbf1 DR |
1129 | int err; |
1130 | ||
1131 | memset(buffer, 0, sizeof(buffer)); | |
1132 | if (count > sizeof(buffer) - 1) | |
1133 | count = sizeof(buffer) - 1; | |
1134 | if (copy_from_user(buffer, buf, count)) { | |
1135 | err = -EFAULT; | |
1136 | goto out; | |
1137 | } | |
1138 | ||
1139 | err = kstrtoint(strstrip(buffer), 0, &oom_adj); | |
1140 | if (err) | |
1141 | goto out; | |
1142 | if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) && | |
1143 | oom_adj != OOM_DISABLE) { | |
1144 | err = -EINVAL; | |
1145 | goto out; | |
1146 | } | |
1147 | ||
fa0cbbf1 DR |
1148 | /* |
1149 | * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum | |
1150 | * value is always attainable. | |
1151 | */ | |
1152 | if (oom_adj == OOM_ADJUST_MAX) | |
1153 | oom_adj = OOM_SCORE_ADJ_MAX; | |
1154 | else | |
1155 | oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE; | |
1156 | ||
1d5f0acb | 1157 | err = __set_oom_adj(file, oom_adj, true); |
fa0cbbf1 DR |
1158 | out: |
1159 | return err < 0 ? err : count; | |
1160 | } | |
1161 | ||
1162 | static const struct file_operations proc_oom_adj_operations = { | |
1163 | .read = oom_adj_read, | |
1164 | .write = oom_adj_write, | |
1165 | .llseek = generic_file_llseek, | |
1166 | }; | |
1167 | ||
a63d83f4 DR |
1168 | static ssize_t oom_score_adj_read(struct file *file, char __user *buf, |
1169 | size_t count, loff_t *ppos) | |
1170 | { | |
496ad9aa | 1171 | struct task_struct *task = get_proc_task(file_inode(file)); |
a63d83f4 | 1172 | char buffer[PROC_NUMBUF]; |
a9c58b90 | 1173 | short oom_score_adj = OOM_SCORE_ADJ_MIN; |
a63d83f4 DR |
1174 | size_t len; |
1175 | ||
1176 | if (!task) | |
1177 | return -ESRCH; | |
f913da59 | 1178 | oom_score_adj = task->signal->oom_score_adj; |
a63d83f4 | 1179 | put_task_struct(task); |
a9c58b90 | 1180 | len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj); |
a63d83f4 DR |
1181 | return simple_read_from_buffer(buf, count, ppos, buffer, len); |
1182 | } | |
1183 | ||
1184 | static ssize_t oom_score_adj_write(struct file *file, const char __user *buf, | |
1185 | size_t count, loff_t *ppos) | |
1186 | { | |
a63d83f4 | 1187 | char buffer[PROC_NUMBUF]; |
0a8cb8e3 | 1188 | int oom_score_adj; |
a63d83f4 DR |
1189 | int err; |
1190 | ||
1191 | memset(buffer, 0, sizeof(buffer)); | |
1192 | if (count > sizeof(buffer) - 1) | |
1193 | count = sizeof(buffer) - 1; | |
723548bf DR |
1194 | if (copy_from_user(buffer, buf, count)) { |
1195 | err = -EFAULT; | |
1196 | goto out; | |
1197 | } | |
a63d83f4 | 1198 | |
0a8cb8e3 | 1199 | err = kstrtoint(strstrip(buffer), 0, &oom_score_adj); |
a63d83f4 | 1200 | if (err) |
723548bf | 1201 | goto out; |
a63d83f4 | 1202 | if (oom_score_adj < OOM_SCORE_ADJ_MIN || |
723548bf DR |
1203 | oom_score_adj > OOM_SCORE_ADJ_MAX) { |
1204 | err = -EINVAL; | |
1205 | goto out; | |
1206 | } | |
a63d83f4 | 1207 | |
1d5f0acb | 1208 | err = __set_oom_adj(file, oom_score_adj, false); |
723548bf DR |
1209 | out: |
1210 | return err < 0 ? err : count; | |
a63d83f4 DR |
1211 | } |
1212 | ||
1213 | static const struct file_operations proc_oom_score_adj_operations = { | |
1214 | .read = oom_score_adj_read, | |
1215 | .write = oom_score_adj_write, | |
6038f373 | 1216 | .llseek = default_llseek, |
a63d83f4 DR |
1217 | }; |
1218 | ||
1da177e4 | 1219 | #ifdef CONFIG_AUDITSYSCALL |
b4eb4f7f | 1220 | #define TMPBUFLEN 11 |
1da177e4 LT |
1221 | static ssize_t proc_loginuid_read(struct file * file, char __user * buf, |
1222 | size_t count, loff_t *ppos) | |
1223 | { | |
496ad9aa | 1224 | struct inode * inode = file_inode(file); |
99f89551 | 1225 | struct task_struct *task = get_proc_task(inode); |
1da177e4 LT |
1226 | ssize_t length; |
1227 | char tmpbuf[TMPBUFLEN]; | |
1228 | ||
99f89551 EB |
1229 | if (!task) |
1230 | return -ESRCH; | |
1da177e4 | 1231 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", |
e1760bd5 EB |
1232 | from_kuid(file->f_cred->user_ns, |
1233 | audit_get_loginuid(task))); | |
99f89551 | 1234 | put_task_struct(task); |
1da177e4 LT |
1235 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); |
1236 | } | |
1237 | ||
1238 | static ssize_t proc_loginuid_write(struct file * file, const char __user * buf, | |
1239 | size_t count, loff_t *ppos) | |
1240 | { | |
496ad9aa | 1241 | struct inode * inode = file_inode(file); |
1da177e4 | 1242 | uid_t loginuid; |
e1760bd5 | 1243 | kuid_t kloginuid; |
774636e1 | 1244 | int rv; |
1da177e4 | 1245 | |
7dc52157 PM |
1246 | rcu_read_lock(); |
1247 | if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) { | |
1248 | rcu_read_unlock(); | |
1da177e4 | 1249 | return -EPERM; |
7dc52157 PM |
1250 | } |
1251 | rcu_read_unlock(); | |
1da177e4 | 1252 | |
1da177e4 LT |
1253 | if (*ppos != 0) { |
1254 | /* No partial writes. */ | |
1255 | return -EINVAL; | |
1256 | } | |
1da177e4 | 1257 | |
774636e1 AD |
1258 | rv = kstrtou32_from_user(buf, count, 10, &loginuid); |
1259 | if (rv < 0) | |
1260 | return rv; | |
81407c84 EP |
1261 | |
1262 | /* is userspace tring to explicitly UNSET the loginuid? */ | |
1263 | if (loginuid == AUDIT_UID_UNSET) { | |
1264 | kloginuid = INVALID_UID; | |
1265 | } else { | |
1266 | kloginuid = make_kuid(file->f_cred->user_ns, loginuid); | |
774636e1 AD |
1267 | if (!uid_valid(kloginuid)) |
1268 | return -EINVAL; | |
e1760bd5 EB |
1269 | } |
1270 | ||
774636e1 AD |
1271 | rv = audit_set_loginuid(kloginuid); |
1272 | if (rv < 0) | |
1273 | return rv; | |
1274 | return count; | |
1da177e4 LT |
1275 | } |
1276 | ||
00977a59 | 1277 | static const struct file_operations proc_loginuid_operations = { |
1da177e4 LT |
1278 | .read = proc_loginuid_read, |
1279 | .write = proc_loginuid_write, | |
87df8424 | 1280 | .llseek = generic_file_llseek, |
1da177e4 | 1281 | }; |
1e0bd755 EP |
1282 | |
1283 | static ssize_t proc_sessionid_read(struct file * file, char __user * buf, | |
1284 | size_t count, loff_t *ppos) | |
1285 | { | |
496ad9aa | 1286 | struct inode * inode = file_inode(file); |
1e0bd755 EP |
1287 | struct task_struct *task = get_proc_task(inode); |
1288 | ssize_t length; | |
1289 | char tmpbuf[TMPBUFLEN]; | |
1290 | ||
1291 | if (!task) | |
1292 | return -ESRCH; | |
1293 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", | |
1294 | audit_get_sessionid(task)); | |
1295 | put_task_struct(task); | |
1296 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); | |
1297 | } | |
1298 | ||
1299 | static const struct file_operations proc_sessionid_operations = { | |
1300 | .read = proc_sessionid_read, | |
87df8424 | 1301 | .llseek = generic_file_llseek, |
1e0bd755 | 1302 | }; |
1da177e4 LT |
1303 | #endif |
1304 | ||
f4f154fd AM |
1305 | #ifdef CONFIG_FAULT_INJECTION |
1306 | static ssize_t proc_fault_inject_read(struct file * file, char __user * buf, | |
1307 | size_t count, loff_t *ppos) | |
1308 | { | |
496ad9aa | 1309 | struct task_struct *task = get_proc_task(file_inode(file)); |
f4f154fd AM |
1310 | char buffer[PROC_NUMBUF]; |
1311 | size_t len; | |
1312 | int make_it_fail; | |
f4f154fd AM |
1313 | |
1314 | if (!task) | |
1315 | return -ESRCH; | |
1316 | make_it_fail = task->make_it_fail; | |
1317 | put_task_struct(task); | |
1318 | ||
1319 | len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail); | |
0c28f287 AM |
1320 | |
1321 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
f4f154fd AM |
1322 | } |
1323 | ||
1324 | static ssize_t proc_fault_inject_write(struct file * file, | |
1325 | const char __user * buf, size_t count, loff_t *ppos) | |
1326 | { | |
1327 | struct task_struct *task; | |
774636e1 | 1328 | char buffer[PROC_NUMBUF]; |
f4f154fd | 1329 | int make_it_fail; |
774636e1 | 1330 | int rv; |
f4f154fd AM |
1331 | |
1332 | if (!capable(CAP_SYS_RESOURCE)) | |
1333 | return -EPERM; | |
1334 | memset(buffer, 0, sizeof(buffer)); | |
1335 | if (count > sizeof(buffer) - 1) | |
1336 | count = sizeof(buffer) - 1; | |
1337 | if (copy_from_user(buffer, buf, count)) | |
1338 | return -EFAULT; | |
774636e1 AD |
1339 | rv = kstrtoint(strstrip(buffer), 0, &make_it_fail); |
1340 | if (rv < 0) | |
1341 | return rv; | |
16caed31 DJ |
1342 | if (make_it_fail < 0 || make_it_fail > 1) |
1343 | return -EINVAL; | |
1344 | ||
496ad9aa | 1345 | task = get_proc_task(file_inode(file)); |
f4f154fd AM |
1346 | if (!task) |
1347 | return -ESRCH; | |
1348 | task->make_it_fail = make_it_fail; | |
1349 | put_task_struct(task); | |
cba8aafe VL |
1350 | |
1351 | return count; | |
f4f154fd AM |
1352 | } |
1353 | ||
00977a59 | 1354 | static const struct file_operations proc_fault_inject_operations = { |
f4f154fd AM |
1355 | .read = proc_fault_inject_read, |
1356 | .write = proc_fault_inject_write, | |
87df8424 | 1357 | .llseek = generic_file_llseek, |
f4f154fd | 1358 | }; |
e41d5818 DV |
1359 | |
1360 | static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf, | |
1361 | size_t count, loff_t *ppos) | |
1362 | { | |
1363 | struct task_struct *task; | |
9049f2f6 AM |
1364 | int err; |
1365 | unsigned int n; | |
e41d5818 | 1366 | |
1203c8e6 AM |
1367 | err = kstrtouint_from_user(buf, count, 0, &n); |
1368 | if (err) | |
1369 | return err; | |
1370 | ||
e41d5818 DV |
1371 | task = get_proc_task(file_inode(file)); |
1372 | if (!task) | |
1373 | return -ESRCH; | |
1203c8e6 | 1374 | WRITE_ONCE(task->fail_nth, n); |
e41d5818 | 1375 | put_task_struct(task); |
1203c8e6 | 1376 | |
e41d5818 DV |
1377 | return count; |
1378 | } | |
1379 | ||
1380 | static ssize_t proc_fail_nth_read(struct file *file, char __user *buf, | |
1381 | size_t count, loff_t *ppos) | |
1382 | { | |
1383 | struct task_struct *task; | |
bfc74093 AM |
1384 | char numbuf[PROC_NUMBUF]; |
1385 | ssize_t len; | |
e41d5818 DV |
1386 | |
1387 | task = get_proc_task(file_inode(file)); | |
1388 | if (!task) | |
1389 | return -ESRCH; | |
1203c8e6 AM |
1390 | len = snprintf(numbuf, sizeof(numbuf), "%u\n", |
1391 | READ_ONCE(task->fail_nth)); | |
bfc74093 | 1392 | len = simple_read_from_buffer(buf, count, ppos, numbuf, len); |
1203c8e6 | 1393 | put_task_struct(task); |
bfc74093 AM |
1394 | |
1395 | return len; | |
e41d5818 DV |
1396 | } |
1397 | ||
1398 | static const struct file_operations proc_fail_nth_operations = { | |
1399 | .read = proc_fail_nth_read, | |
1400 | .write = proc_fail_nth_write, | |
1401 | }; | |
f4f154fd AM |
1402 | #endif |
1403 | ||
9745512c | 1404 | |
43ae34cb IM |
1405 | #ifdef CONFIG_SCHED_DEBUG |
1406 | /* | |
1407 | * Print out various scheduling related per-task fields: | |
1408 | */ | |
1409 | static int sched_show(struct seq_file *m, void *v) | |
1410 | { | |
1411 | struct inode *inode = m->private; | |
74dc3384 | 1412 | struct pid_namespace *ns = inode->i_sb->s_fs_info; |
43ae34cb IM |
1413 | struct task_struct *p; |
1414 | ||
43ae34cb IM |
1415 | p = get_proc_task(inode); |
1416 | if (!p) | |
1417 | return -ESRCH; | |
74dc3384 | 1418 | proc_sched_show_task(p, ns, m); |
43ae34cb IM |
1419 | |
1420 | put_task_struct(p); | |
1421 | ||
1422 | return 0; | |
1423 | } | |
1424 | ||
1425 | static ssize_t | |
1426 | sched_write(struct file *file, const char __user *buf, | |
1427 | size_t count, loff_t *offset) | |
1428 | { | |
496ad9aa | 1429 | struct inode *inode = file_inode(file); |
43ae34cb IM |
1430 | struct task_struct *p; |
1431 | ||
43ae34cb IM |
1432 | p = get_proc_task(inode); |
1433 | if (!p) | |
1434 | return -ESRCH; | |
1435 | proc_sched_set_task(p); | |
1436 | ||
1437 | put_task_struct(p); | |
1438 | ||
1439 | return count; | |
1440 | } | |
1441 | ||
1442 | static int sched_open(struct inode *inode, struct file *filp) | |
1443 | { | |
c6a34058 | 1444 | return single_open(filp, sched_show, inode); |
43ae34cb IM |
1445 | } |
1446 | ||
1447 | static const struct file_operations proc_pid_sched_operations = { | |
1448 | .open = sched_open, | |
1449 | .read = seq_read, | |
1450 | .write = sched_write, | |
1451 | .llseek = seq_lseek, | |
5ea473a1 | 1452 | .release = single_release, |
43ae34cb IM |
1453 | }; |
1454 | ||
1455 | #endif | |
1456 | ||
5091faa4 MG |
1457 | #ifdef CONFIG_SCHED_AUTOGROUP |
1458 | /* | |
1459 | * Print out autogroup related information: | |
1460 | */ | |
1461 | static int sched_autogroup_show(struct seq_file *m, void *v) | |
1462 | { | |
1463 | struct inode *inode = m->private; | |
1464 | struct task_struct *p; | |
1465 | ||
1466 | p = get_proc_task(inode); | |
1467 | if (!p) | |
1468 | return -ESRCH; | |
1469 | proc_sched_autogroup_show_task(p, m); | |
1470 | ||
1471 | put_task_struct(p); | |
1472 | ||
1473 | return 0; | |
1474 | } | |
1475 | ||
1476 | static ssize_t | |
1477 | sched_autogroup_write(struct file *file, const char __user *buf, | |
1478 | size_t count, loff_t *offset) | |
1479 | { | |
496ad9aa | 1480 | struct inode *inode = file_inode(file); |
5091faa4 MG |
1481 | struct task_struct *p; |
1482 | char buffer[PROC_NUMBUF]; | |
0a8cb8e3 | 1483 | int nice; |
5091faa4 MG |
1484 | int err; |
1485 | ||
1486 | memset(buffer, 0, sizeof(buffer)); | |
1487 | if (count > sizeof(buffer) - 1) | |
1488 | count = sizeof(buffer) - 1; | |
1489 | if (copy_from_user(buffer, buf, count)) | |
1490 | return -EFAULT; | |
1491 | ||
0a8cb8e3 AD |
1492 | err = kstrtoint(strstrip(buffer), 0, &nice); |
1493 | if (err < 0) | |
1494 | return err; | |
5091faa4 MG |
1495 | |
1496 | p = get_proc_task(inode); | |
1497 | if (!p) | |
1498 | return -ESRCH; | |
1499 | ||
2e5b5b3a | 1500 | err = proc_sched_autogroup_set_nice(p, nice); |
5091faa4 MG |
1501 | if (err) |
1502 | count = err; | |
1503 | ||
1504 | put_task_struct(p); | |
1505 | ||
1506 | return count; | |
1507 | } | |
1508 | ||
1509 | static int sched_autogroup_open(struct inode *inode, struct file *filp) | |
1510 | { | |
1511 | int ret; | |
1512 | ||
1513 | ret = single_open(filp, sched_autogroup_show, NULL); | |
1514 | if (!ret) { | |
1515 | struct seq_file *m = filp->private_data; | |
1516 | ||
1517 | m->private = inode; | |
1518 | } | |
1519 | return ret; | |
1520 | } | |
1521 | ||
1522 | static const struct file_operations proc_pid_sched_autogroup_operations = { | |
1523 | .open = sched_autogroup_open, | |
1524 | .read = seq_read, | |
1525 | .write = sched_autogroup_write, | |
1526 | .llseek = seq_lseek, | |
1527 | .release = single_release, | |
1528 | }; | |
1529 | ||
1530 | #endif /* CONFIG_SCHED_AUTOGROUP */ | |
1531 | ||
4614a696 JS |
1532 | static ssize_t comm_write(struct file *file, const char __user *buf, |
1533 | size_t count, loff_t *offset) | |
1534 | { | |
496ad9aa | 1535 | struct inode *inode = file_inode(file); |
4614a696 JS |
1536 | struct task_struct *p; |
1537 | char buffer[TASK_COMM_LEN]; | |
830e0fc9 | 1538 | const size_t maxlen = sizeof(buffer) - 1; |
4614a696 JS |
1539 | |
1540 | memset(buffer, 0, sizeof(buffer)); | |
830e0fc9 | 1541 | if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count)) |
4614a696 JS |
1542 | return -EFAULT; |
1543 | ||
1544 | p = get_proc_task(inode); | |
1545 | if (!p) | |
1546 | return -ESRCH; | |
1547 | ||
1548 | if (same_thread_group(current, p)) | |
1549 | set_task_comm(p, buffer); | |
1550 | else | |
1551 | count = -EINVAL; | |
1552 | ||
1553 | put_task_struct(p); | |
1554 | ||
1555 | return count; | |
1556 | } | |
1557 | ||
1558 | static int comm_show(struct seq_file *m, void *v) | |
1559 | { | |
1560 | struct inode *inode = m->private; | |
1561 | struct task_struct *p; | |
1562 | ||
1563 | p = get_proc_task(inode); | |
1564 | if (!p) | |
1565 | return -ESRCH; | |
1566 | ||
1567 | task_lock(p); | |
1568 | seq_printf(m, "%s\n", p->comm); | |
1569 | task_unlock(p); | |
1570 | ||
1571 | put_task_struct(p); | |
1572 | ||
1573 | return 0; | |
1574 | } | |
1575 | ||
1576 | static int comm_open(struct inode *inode, struct file *filp) | |
1577 | { | |
c6a34058 | 1578 | return single_open(filp, comm_show, inode); |
4614a696 JS |
1579 | } |
1580 | ||
1581 | static const struct file_operations proc_pid_set_comm_operations = { | |
1582 | .open = comm_open, | |
1583 | .read = seq_read, | |
1584 | .write = comm_write, | |
1585 | .llseek = seq_lseek, | |
1586 | .release = single_release, | |
1587 | }; | |
1588 | ||
7773fbc5 | 1589 | static int proc_exe_link(struct dentry *dentry, struct path *exe_path) |
925d1c40 MH |
1590 | { |
1591 | struct task_struct *task; | |
925d1c40 MH |
1592 | struct file *exe_file; |
1593 | ||
2b0143b5 | 1594 | task = get_proc_task(d_inode(dentry)); |
925d1c40 MH |
1595 | if (!task) |
1596 | return -ENOENT; | |
cd81a917 | 1597 | exe_file = get_task_exe_file(task); |
925d1c40 | 1598 | put_task_struct(task); |
925d1c40 MH |
1599 | if (exe_file) { |
1600 | *exe_path = exe_file->f_path; | |
1601 | path_get(&exe_file->f_path); | |
1602 | fput(exe_file); | |
1603 | return 0; | |
1604 | } else | |
1605 | return -ENOENT; | |
1606 | } | |
1607 | ||
6b255391 | 1608 | static const char *proc_pid_get_link(struct dentry *dentry, |
fceef393 AV |
1609 | struct inode *inode, |
1610 | struct delayed_call *done) | |
1da177e4 | 1611 | { |
408ef013 | 1612 | struct path path; |
1da177e4 LT |
1613 | int error = -EACCES; |
1614 | ||
6b255391 AV |
1615 | if (!dentry) |
1616 | return ERR_PTR(-ECHILD); | |
1617 | ||
778c1144 EB |
1618 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1619 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1620 | goto out; |
1da177e4 | 1621 | |
408ef013 CH |
1622 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1623 | if (error) | |
1624 | goto out; | |
1625 | ||
6e77137b | 1626 | nd_jump_link(&path); |
408ef013 | 1627 | return NULL; |
1da177e4 | 1628 | out: |
008b150a | 1629 | return ERR_PTR(error); |
1da177e4 LT |
1630 | } |
1631 | ||
3dcd25f3 | 1632 | static int do_proc_readlink(struct path *path, char __user *buffer, int buflen) |
1da177e4 | 1633 | { |
0ee931c4 | 1634 | char *tmp = (char *)__get_free_page(GFP_KERNEL); |
3dcd25f3 | 1635 | char *pathname; |
1da177e4 LT |
1636 | int len; |
1637 | ||
1638 | if (!tmp) | |
1639 | return -ENOMEM; | |
0c28f287 | 1640 | |
7b2a69ba | 1641 | pathname = d_path(path, tmp, PAGE_SIZE); |
3dcd25f3 JB |
1642 | len = PTR_ERR(pathname); |
1643 | if (IS_ERR(pathname)) | |
1da177e4 | 1644 | goto out; |
3dcd25f3 | 1645 | len = tmp + PAGE_SIZE - 1 - pathname; |
1da177e4 LT |
1646 | |
1647 | if (len > buflen) | |
1648 | len = buflen; | |
3dcd25f3 | 1649 | if (copy_to_user(buffer, pathname, len)) |
1da177e4 LT |
1650 | len = -EFAULT; |
1651 | out: | |
1652 | free_page((unsigned long)tmp); | |
1653 | return len; | |
1654 | } | |
1655 | ||
1656 | static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen) | |
1657 | { | |
1658 | int error = -EACCES; | |
2b0143b5 | 1659 | struct inode *inode = d_inode(dentry); |
3dcd25f3 | 1660 | struct path path; |
1da177e4 | 1661 | |
778c1144 EB |
1662 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1663 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1664 | goto out; |
1da177e4 | 1665 | |
7773fbc5 | 1666 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1da177e4 LT |
1667 | if (error) |
1668 | goto out; | |
1669 | ||
3dcd25f3 JB |
1670 | error = do_proc_readlink(&path, buffer, buflen); |
1671 | path_put(&path); | |
1da177e4 | 1672 | out: |
1da177e4 LT |
1673 | return error; |
1674 | } | |
1675 | ||
faf60af1 | 1676 | const struct inode_operations proc_pid_link_inode_operations = { |
1da177e4 | 1677 | .readlink = proc_pid_readlink, |
6b255391 | 1678 | .get_link = proc_pid_get_link, |
6d76fa58 | 1679 | .setattr = proc_setattr, |
1da177e4 LT |
1680 | }; |
1681 | ||
28a6d671 EB |
1682 | |
1683 | /* building an inode */ | |
1684 | ||
68eb94f1 EB |
1685 | void task_dump_owner(struct task_struct *task, mode_t mode, |
1686 | kuid_t *ruid, kgid_t *rgid) | |
1687 | { | |
1688 | /* Depending on the state of dumpable compute who should own a | |
1689 | * proc file for a task. | |
1690 | */ | |
1691 | const struct cred *cred; | |
1692 | kuid_t uid; | |
1693 | kgid_t gid; | |
1694 | ||
1695 | /* Default to the tasks effective ownership */ | |
1696 | rcu_read_lock(); | |
1697 | cred = __task_cred(task); | |
1698 | uid = cred->euid; | |
1699 | gid = cred->egid; | |
1700 | rcu_read_unlock(); | |
1701 | ||
1702 | /* | |
1703 | * Before the /proc/pid/status file was created the only way to read | |
1704 | * the effective uid of a /process was to stat /proc/pid. Reading | |
1705 | * /proc/pid/status is slow enough that procps and other packages | |
1706 | * kept stating /proc/pid. To keep the rules in /proc simple I have | |
1707 | * made this apply to all per process world readable and executable | |
1708 | * directories. | |
1709 | */ | |
1710 | if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) { | |
1711 | struct mm_struct *mm; | |
1712 | task_lock(task); | |
1713 | mm = task->mm; | |
1714 | /* Make non-dumpable tasks owned by some root */ | |
1715 | if (mm) { | |
1716 | if (get_dumpable(mm) != SUID_DUMP_USER) { | |
1717 | struct user_namespace *user_ns = mm->user_ns; | |
1718 | ||
1719 | uid = make_kuid(user_ns, 0); | |
1720 | if (!uid_valid(uid)) | |
1721 | uid = GLOBAL_ROOT_UID; | |
1722 | ||
1723 | gid = make_kgid(user_ns, 0); | |
1724 | if (!gid_valid(gid)) | |
1725 | gid = GLOBAL_ROOT_GID; | |
1726 | } | |
1727 | } else { | |
1728 | uid = GLOBAL_ROOT_UID; | |
1729 | gid = GLOBAL_ROOT_GID; | |
1730 | } | |
1731 | task_unlock(task); | |
1732 | } | |
1733 | *ruid = uid; | |
1734 | *rgid = gid; | |
1735 | } | |
1736 | ||
db978da8 AG |
1737 | struct inode *proc_pid_make_inode(struct super_block * sb, |
1738 | struct task_struct *task, umode_t mode) | |
28a6d671 EB |
1739 | { |
1740 | struct inode * inode; | |
1741 | struct proc_inode *ei; | |
1da177e4 | 1742 | |
28a6d671 | 1743 | /* We need a new inode */ |
1da177e4 | 1744 | |
28a6d671 EB |
1745 | inode = new_inode(sb); |
1746 | if (!inode) | |
1747 | goto out; | |
1748 | ||
1749 | /* Common stuff */ | |
1750 | ei = PROC_I(inode); | |
db978da8 | 1751 | inode->i_mode = mode; |
85fe4025 | 1752 | inode->i_ino = get_next_ino(); |
078cd827 | 1753 | inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode); |
28a6d671 EB |
1754 | inode->i_op = &proc_def_inode_operations; |
1755 | ||
1756 | /* | |
1757 | * grab the reference to task. | |
1758 | */ | |
1a657f78 | 1759 | ei->pid = get_task_pid(task, PIDTYPE_PID); |
28a6d671 EB |
1760 | if (!ei->pid) |
1761 | goto out_unlock; | |
1762 | ||
68eb94f1 | 1763 | task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid); |
28a6d671 EB |
1764 | security_task_to_inode(task, inode); |
1765 | ||
1da177e4 | 1766 | out: |
28a6d671 EB |
1767 | return inode; |
1768 | ||
1769 | out_unlock: | |
1770 | iput(inode); | |
1771 | return NULL; | |
1da177e4 LT |
1772 | } |
1773 | ||
a528d35e DH |
1774 | int pid_getattr(const struct path *path, struct kstat *stat, |
1775 | u32 request_mask, unsigned int query_flags) | |
1da177e4 | 1776 | { |
a528d35e | 1777 | struct inode *inode = d_inode(path->dentry); |
28a6d671 | 1778 | struct task_struct *task; |
a528d35e | 1779 | struct pid_namespace *pid = path->dentry->d_sb->s_fs_info; |
c69e8d9c | 1780 | |
28a6d671 | 1781 | generic_fillattr(inode, stat); |
1da177e4 | 1782 | |
28a6d671 | 1783 | rcu_read_lock(); |
dcb0f222 EB |
1784 | stat->uid = GLOBAL_ROOT_UID; |
1785 | stat->gid = GLOBAL_ROOT_GID; | |
28a6d671 EB |
1786 | task = pid_task(proc_pid(inode), PIDTYPE_PID); |
1787 | if (task) { | |
796f571b | 1788 | if (!has_pid_permissions(pid, task, HIDEPID_INVISIBLE)) { |
0499680a VK |
1789 | rcu_read_unlock(); |
1790 | /* | |
1791 | * This doesn't prevent learning whether PID exists, | |
1792 | * it only makes getattr() consistent with readdir(). | |
1793 | */ | |
1794 | return -ENOENT; | |
1795 | } | |
68eb94f1 | 1796 | task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid); |
1da177e4 | 1797 | } |
28a6d671 | 1798 | rcu_read_unlock(); |
d6e71144 | 1799 | return 0; |
1da177e4 LT |
1800 | } |
1801 | ||
1da177e4 LT |
1802 | /* dentry stuff */ |
1803 | ||
1804 | /* | |
1805 | * Exceptional case: normally we are not allowed to unhash a busy | |
1806 | * directory. In this case, however, we can do it - no aliasing problems | |
1807 | * due to the way we treat inodes. | |
1808 | * | |
1809 | * Rewrite the inode's ownerships here because the owning task may have | |
1810 | * performed a setuid(), etc. | |
99f89551 | 1811 | * |
1da177e4 | 1812 | */ |
0b728e19 | 1813 | int pid_revalidate(struct dentry *dentry, unsigned int flags) |
1da177e4 | 1814 | { |
34286d66 NP |
1815 | struct inode *inode; |
1816 | struct task_struct *task; | |
c69e8d9c | 1817 | |
0b728e19 | 1818 | if (flags & LOOKUP_RCU) |
34286d66 NP |
1819 | return -ECHILD; |
1820 | ||
2b0143b5 | 1821 | inode = d_inode(dentry); |
34286d66 NP |
1822 | task = get_proc_task(inode); |
1823 | ||
99f89551 | 1824 | if (task) { |
68eb94f1 EB |
1825 | task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid); |
1826 | ||
9ee8ab9f | 1827 | inode->i_mode &= ~(S_ISUID | S_ISGID); |
1da177e4 | 1828 | security_task_to_inode(task, inode); |
99f89551 | 1829 | put_task_struct(task); |
1da177e4 LT |
1830 | return 1; |
1831 | } | |
1da177e4 LT |
1832 | return 0; |
1833 | } | |
1834 | ||
d855a4b7 ON |
1835 | static inline bool proc_inode_is_dead(struct inode *inode) |
1836 | { | |
1837 | return !proc_pid(inode)->tasks[PIDTYPE_PID].first; | |
1838 | } | |
1839 | ||
1dd704b6 DH |
1840 | int pid_delete_dentry(const struct dentry *dentry) |
1841 | { | |
1842 | /* Is the task we represent dead? | |
1843 | * If so, then don't put the dentry on the lru list, | |
1844 | * kill it immediately. | |
1845 | */ | |
2b0143b5 | 1846 | return proc_inode_is_dead(d_inode(dentry)); |
1dd704b6 DH |
1847 | } |
1848 | ||
6b4e306a | 1849 | const struct dentry_operations pid_dentry_operations = |
28a6d671 EB |
1850 | { |
1851 | .d_revalidate = pid_revalidate, | |
1852 | .d_delete = pid_delete_dentry, | |
1853 | }; | |
1854 | ||
1855 | /* Lookups */ | |
1856 | ||
1c0d04c9 EB |
1857 | /* |
1858 | * Fill a directory entry. | |
1859 | * | |
1860 | * If possible create the dcache entry and derive our inode number and | |
1861 | * file type from dcache entry. | |
1862 | * | |
1863 | * Since all of the proc inode numbers are dynamically generated, the inode | |
1864 | * numbers do not exist until the inode is cache. This means creating the | |
1865 | * the dcache entry in readdir is necessary to keep the inode numbers | |
1866 | * reported by readdir in sync with the inode numbers reported | |
1867 | * by stat. | |
1868 | */ | |
f0c3b509 | 1869 | bool proc_fill_cache(struct file *file, struct dir_context *ctx, |
6b4e306a | 1870 | const char *name, int len, |
c5141e6d | 1871 | instantiate_t instantiate, struct task_struct *task, const void *ptr) |
61a28784 | 1872 | { |
f0c3b509 | 1873 | struct dentry *child, *dir = file->f_path.dentry; |
1df98b8b | 1874 | struct qstr qname = QSTR_INIT(name, len); |
61a28784 | 1875 | struct inode *inode; |
1df98b8b AV |
1876 | unsigned type; |
1877 | ino_t ino; | |
61a28784 | 1878 | |
1df98b8b | 1879 | child = d_hash_and_lookup(dir, &qname); |
61a28784 | 1880 | if (!child) { |
3781764b AV |
1881 | DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq); |
1882 | child = d_alloc_parallel(dir, &qname, &wq); | |
1883 | if (IS_ERR(child)) | |
1df98b8b | 1884 | goto end_instantiate; |
3781764b AV |
1885 | if (d_in_lookup(child)) { |
1886 | int err = instantiate(d_inode(dir), child, task, ptr); | |
1887 | d_lookup_done(child); | |
1888 | if (err < 0) { | |
1889 | dput(child); | |
1890 | goto end_instantiate; | |
1891 | } | |
61a28784 EB |
1892 | } |
1893 | } | |
2b0143b5 | 1894 | inode = d_inode(child); |
147ce699 AV |
1895 | ino = inode->i_ino; |
1896 | type = inode->i_mode >> 12; | |
61a28784 | 1897 | dput(child); |
f0c3b509 | 1898 | return dir_emit(ctx, name, len, ino, type); |
1df98b8b AV |
1899 | |
1900 | end_instantiate: | |
1901 | return dir_emit(ctx, name, len, 1, DT_UNKNOWN); | |
61a28784 EB |
1902 | } |
1903 | ||
640708a2 PE |
1904 | /* |
1905 | * dname_to_vma_addr - maps a dentry name into two unsigned longs | |
1906 | * which represent vma start and end addresses. | |
1907 | */ | |
1908 | static int dname_to_vma_addr(struct dentry *dentry, | |
1909 | unsigned long *start, unsigned long *end) | |
1910 | { | |
1911 | if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2) | |
1912 | return -EINVAL; | |
1913 | ||
1914 | return 0; | |
1915 | } | |
1916 | ||
0b728e19 | 1917 | static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags) |
640708a2 PE |
1918 | { |
1919 | unsigned long vm_start, vm_end; | |
1920 | bool exact_vma_exists = false; | |
1921 | struct mm_struct *mm = NULL; | |
1922 | struct task_struct *task; | |
640708a2 PE |
1923 | struct inode *inode; |
1924 | int status = 0; | |
1925 | ||
0b728e19 | 1926 | if (flags & LOOKUP_RCU) |
640708a2 PE |
1927 | return -ECHILD; |
1928 | ||
2b0143b5 | 1929 | inode = d_inode(dentry); |
640708a2 PE |
1930 | task = get_proc_task(inode); |
1931 | if (!task) | |
1932 | goto out_notask; | |
1933 | ||
caaee623 | 1934 | mm = mm_access(task, PTRACE_MODE_READ_FSCREDS); |
2344bec7 | 1935 | if (IS_ERR_OR_NULL(mm)) |
640708a2 PE |
1936 | goto out; |
1937 | ||
1938 | if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) { | |
1939 | down_read(&mm->mmap_sem); | |
1940 | exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end); | |
1941 | up_read(&mm->mmap_sem); | |
1942 | } | |
1943 | ||
1944 | mmput(mm); | |
1945 | ||
1946 | if (exact_vma_exists) { | |
68eb94f1 EB |
1947 | task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid); |
1948 | ||
640708a2 PE |
1949 | security_task_to_inode(task, inode); |
1950 | status = 1; | |
1951 | } | |
1952 | ||
1953 | out: | |
1954 | put_task_struct(task); | |
1955 | ||
1956 | out_notask: | |
640708a2 PE |
1957 | return status; |
1958 | } | |
1959 | ||
1960 | static const struct dentry_operations tid_map_files_dentry_operations = { | |
1961 | .d_revalidate = map_files_d_revalidate, | |
1962 | .d_delete = pid_delete_dentry, | |
1963 | }; | |
1964 | ||
6b255391 | 1965 | static int map_files_get_link(struct dentry *dentry, struct path *path) |
640708a2 PE |
1966 | { |
1967 | unsigned long vm_start, vm_end; | |
1968 | struct vm_area_struct *vma; | |
1969 | struct task_struct *task; | |
1970 | struct mm_struct *mm; | |
1971 | int rc; | |
1972 | ||
1973 | rc = -ENOENT; | |
2b0143b5 | 1974 | task = get_proc_task(d_inode(dentry)); |
640708a2 PE |
1975 | if (!task) |
1976 | goto out; | |
1977 | ||
1978 | mm = get_task_mm(task); | |
1979 | put_task_struct(task); | |
1980 | if (!mm) | |
1981 | goto out; | |
1982 | ||
1983 | rc = dname_to_vma_addr(dentry, &vm_start, &vm_end); | |
1984 | if (rc) | |
1985 | goto out_mmput; | |
1986 | ||
70335abb | 1987 | rc = -ENOENT; |
640708a2 PE |
1988 | down_read(&mm->mmap_sem); |
1989 | vma = find_exact_vma(mm, vm_start, vm_end); | |
1990 | if (vma && vma->vm_file) { | |
1991 | *path = vma->vm_file->f_path; | |
1992 | path_get(path); | |
1993 | rc = 0; | |
1994 | } | |
1995 | up_read(&mm->mmap_sem); | |
1996 | ||
1997 | out_mmput: | |
1998 | mmput(mm); | |
1999 | out: | |
2000 | return rc; | |
2001 | } | |
2002 | ||
2003 | struct map_files_info { | |
7b540d06 | 2004 | fmode_t mode; |
9a87fe0d | 2005 | unsigned int len; |
640708a2 PE |
2006 | unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */ |
2007 | }; | |
2008 | ||
bdb4d100 CO |
2009 | /* |
2010 | * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the | |
2011 | * symlinks may be used to bypass permissions on ancestor directories in the | |
2012 | * path to the file in question. | |
2013 | */ | |
2014 | static const char * | |
6b255391 | 2015 | proc_map_files_get_link(struct dentry *dentry, |
fceef393 AV |
2016 | struct inode *inode, |
2017 | struct delayed_call *done) | |
bdb4d100 CO |
2018 | { |
2019 | if (!capable(CAP_SYS_ADMIN)) | |
2020 | return ERR_PTR(-EPERM); | |
2021 | ||
fceef393 | 2022 | return proc_pid_get_link(dentry, inode, done); |
bdb4d100 CO |
2023 | } |
2024 | ||
2025 | /* | |
6b255391 | 2026 | * Identical to proc_pid_link_inode_operations except for get_link() |
bdb4d100 CO |
2027 | */ |
2028 | static const struct inode_operations proc_map_files_link_inode_operations = { | |
2029 | .readlink = proc_pid_readlink, | |
6b255391 | 2030 | .get_link = proc_map_files_get_link, |
bdb4d100 CO |
2031 | .setattr = proc_setattr, |
2032 | }; | |
2033 | ||
c52a47ac | 2034 | static int |
640708a2 PE |
2035 | proc_map_files_instantiate(struct inode *dir, struct dentry *dentry, |
2036 | struct task_struct *task, const void *ptr) | |
2037 | { | |
7b540d06 | 2038 | fmode_t mode = (fmode_t)(unsigned long)ptr; |
640708a2 PE |
2039 | struct proc_inode *ei; |
2040 | struct inode *inode; | |
2041 | ||
db978da8 AG |
2042 | inode = proc_pid_make_inode(dir->i_sb, task, S_IFLNK | |
2043 | ((mode & FMODE_READ ) ? S_IRUSR : 0) | | |
2044 | ((mode & FMODE_WRITE) ? S_IWUSR : 0)); | |
640708a2 | 2045 | if (!inode) |
c52a47ac | 2046 | return -ENOENT; |
640708a2 PE |
2047 | |
2048 | ei = PROC_I(inode); | |
6b255391 | 2049 | ei->op.proc_get_link = map_files_get_link; |
640708a2 | 2050 | |
bdb4d100 | 2051 | inode->i_op = &proc_map_files_link_inode_operations; |
640708a2 | 2052 | inode->i_size = 64; |
640708a2 PE |
2053 | |
2054 | d_set_d_op(dentry, &tid_map_files_dentry_operations); | |
2055 | d_add(dentry, inode); | |
2056 | ||
c52a47ac | 2057 | return 0; |
640708a2 PE |
2058 | } |
2059 | ||
2060 | static struct dentry *proc_map_files_lookup(struct inode *dir, | |
00cd8dd3 | 2061 | struct dentry *dentry, unsigned int flags) |
640708a2 PE |
2062 | { |
2063 | unsigned long vm_start, vm_end; | |
2064 | struct vm_area_struct *vma; | |
2065 | struct task_struct *task; | |
c52a47ac | 2066 | int result; |
640708a2 PE |
2067 | struct mm_struct *mm; |
2068 | ||
c52a47ac | 2069 | result = -ENOENT; |
640708a2 PE |
2070 | task = get_proc_task(dir); |
2071 | if (!task) | |
2072 | goto out; | |
2073 | ||
c52a47ac | 2074 | result = -EACCES; |
caaee623 | 2075 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
640708a2 PE |
2076 | goto out_put_task; |
2077 | ||
c52a47ac | 2078 | result = -ENOENT; |
640708a2 | 2079 | if (dname_to_vma_addr(dentry, &vm_start, &vm_end)) |
eb94cd96 | 2080 | goto out_put_task; |
640708a2 PE |
2081 | |
2082 | mm = get_task_mm(task); | |
2083 | if (!mm) | |
eb94cd96 | 2084 | goto out_put_task; |
640708a2 PE |
2085 | |
2086 | down_read(&mm->mmap_sem); | |
2087 | vma = find_exact_vma(mm, vm_start, vm_end); | |
2088 | if (!vma) | |
2089 | goto out_no_vma; | |
2090 | ||
05f56484 SK |
2091 | if (vma->vm_file) |
2092 | result = proc_map_files_instantiate(dir, dentry, task, | |
2093 | (void *)(unsigned long)vma->vm_file->f_mode); | |
640708a2 PE |
2094 | |
2095 | out_no_vma: | |
2096 | up_read(&mm->mmap_sem); | |
2097 | mmput(mm); | |
640708a2 PE |
2098 | out_put_task: |
2099 | put_task_struct(task); | |
2100 | out: | |
c52a47ac | 2101 | return ERR_PTR(result); |
640708a2 PE |
2102 | } |
2103 | ||
2104 | static const struct inode_operations proc_map_files_inode_operations = { | |
2105 | .lookup = proc_map_files_lookup, | |
2106 | .permission = proc_fd_permission, | |
2107 | .setattr = proc_setattr, | |
2108 | }; | |
2109 | ||
2110 | static int | |
f0c3b509 | 2111 | proc_map_files_readdir(struct file *file, struct dir_context *ctx) |
640708a2 | 2112 | { |
640708a2 PE |
2113 | struct vm_area_struct *vma; |
2114 | struct task_struct *task; | |
2115 | struct mm_struct *mm; | |
f0c3b509 AV |
2116 | unsigned long nr_files, pos, i; |
2117 | struct flex_array *fa = NULL; | |
2118 | struct map_files_info info; | |
2119 | struct map_files_info *p; | |
640708a2 PE |
2120 | int ret; |
2121 | ||
640708a2 | 2122 | ret = -ENOENT; |
f0c3b509 | 2123 | task = get_proc_task(file_inode(file)); |
640708a2 PE |
2124 | if (!task) |
2125 | goto out; | |
2126 | ||
2127 | ret = -EACCES; | |
caaee623 | 2128 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
640708a2 PE |
2129 | goto out_put_task; |
2130 | ||
2131 | ret = 0; | |
f0c3b509 AV |
2132 | if (!dir_emit_dots(file, ctx)) |
2133 | goto out_put_task; | |
640708a2 | 2134 | |
f0c3b509 AV |
2135 | mm = get_task_mm(task); |
2136 | if (!mm) | |
2137 | goto out_put_task; | |
2138 | down_read(&mm->mmap_sem); | |
640708a2 | 2139 | |
f0c3b509 | 2140 | nr_files = 0; |
640708a2 | 2141 | |
f0c3b509 AV |
2142 | /* |
2143 | * We need two passes here: | |
2144 | * | |
2145 | * 1) Collect vmas of mapped files with mmap_sem taken | |
2146 | * 2) Release mmap_sem and instantiate entries | |
2147 | * | |
2148 | * otherwise we get lockdep complained, since filldir() | |
2149 | * routine might require mmap_sem taken in might_fault(). | |
2150 | */ | |
640708a2 | 2151 | |
f0c3b509 AV |
2152 | for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) { |
2153 | if (vma->vm_file && ++pos > ctx->pos) | |
2154 | nr_files++; | |
2155 | } | |
2156 | ||
2157 | if (nr_files) { | |
2158 | fa = flex_array_alloc(sizeof(info), nr_files, | |
2159 | GFP_KERNEL); | |
2160 | if (!fa || flex_array_prealloc(fa, 0, nr_files, | |
2161 | GFP_KERNEL)) { | |
2162 | ret = -ENOMEM; | |
2163 | if (fa) | |
2164 | flex_array_free(fa); | |
2165 | up_read(&mm->mmap_sem); | |
2166 | mmput(mm); | |
2167 | goto out_put_task; | |
640708a2 | 2168 | } |
f0c3b509 AV |
2169 | for (i = 0, vma = mm->mmap, pos = 2; vma; |
2170 | vma = vma->vm_next) { | |
2171 | if (!vma->vm_file) | |
2172 | continue; | |
2173 | if (++pos <= ctx->pos) | |
2174 | continue; | |
2175 | ||
2176 | info.mode = vma->vm_file->f_mode; | |
2177 | info.len = snprintf(info.name, | |
2178 | sizeof(info.name), "%lx-%lx", | |
2179 | vma->vm_start, vma->vm_end); | |
2180 | if (flex_array_put(fa, i++, &info, GFP_KERNEL)) | |
2181 | BUG(); | |
640708a2 | 2182 | } |
640708a2 | 2183 | } |
f0c3b509 AV |
2184 | up_read(&mm->mmap_sem); |
2185 | ||
2186 | for (i = 0; i < nr_files; i++) { | |
2187 | p = flex_array_get(fa, i); | |
2188 | if (!proc_fill_cache(file, ctx, | |
2189 | p->name, p->len, | |
2190 | proc_map_files_instantiate, | |
2191 | task, | |
2192 | (void *)(unsigned long)p->mode)) | |
2193 | break; | |
2194 | ctx->pos++; | |
640708a2 | 2195 | } |
f0c3b509 AV |
2196 | if (fa) |
2197 | flex_array_free(fa); | |
2198 | mmput(mm); | |
640708a2 | 2199 | |
640708a2 PE |
2200 | out_put_task: |
2201 | put_task_struct(task); | |
2202 | out: | |
2203 | return ret; | |
2204 | } | |
2205 | ||
2206 | static const struct file_operations proc_map_files_operations = { | |
2207 | .read = generic_read_dir, | |
f50752ea AV |
2208 | .iterate_shared = proc_map_files_readdir, |
2209 | .llseek = generic_file_llseek, | |
640708a2 PE |
2210 | }; |
2211 | ||
b18b6a9c | 2212 | #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS) |
48f6a7a5 PE |
2213 | struct timers_private { |
2214 | struct pid *pid; | |
2215 | struct task_struct *task; | |
2216 | struct sighand_struct *sighand; | |
57b8015e | 2217 | struct pid_namespace *ns; |
48f6a7a5 PE |
2218 | unsigned long flags; |
2219 | }; | |
2220 | ||
2221 | static void *timers_start(struct seq_file *m, loff_t *pos) | |
2222 | { | |
2223 | struct timers_private *tp = m->private; | |
2224 | ||
2225 | tp->task = get_pid_task(tp->pid, PIDTYPE_PID); | |
2226 | if (!tp->task) | |
2227 | return ERR_PTR(-ESRCH); | |
2228 | ||
2229 | tp->sighand = lock_task_sighand(tp->task, &tp->flags); | |
2230 | if (!tp->sighand) | |
2231 | return ERR_PTR(-ESRCH); | |
2232 | ||
2233 | return seq_list_start(&tp->task->signal->posix_timers, *pos); | |
2234 | } | |
2235 | ||
2236 | static void *timers_next(struct seq_file *m, void *v, loff_t *pos) | |
2237 | { | |
2238 | struct timers_private *tp = m->private; | |
2239 | return seq_list_next(v, &tp->task->signal->posix_timers, pos); | |
2240 | } | |
2241 | ||
2242 | static void timers_stop(struct seq_file *m, void *v) | |
2243 | { | |
2244 | struct timers_private *tp = m->private; | |
2245 | ||
2246 | if (tp->sighand) { | |
2247 | unlock_task_sighand(tp->task, &tp->flags); | |
2248 | tp->sighand = NULL; | |
2249 | } | |
2250 | ||
2251 | if (tp->task) { | |
2252 | put_task_struct(tp->task); | |
2253 | tp->task = NULL; | |
2254 | } | |
2255 | } | |
2256 | ||
2257 | static int show_timer(struct seq_file *m, void *v) | |
2258 | { | |
2259 | struct k_itimer *timer; | |
57b8015e PE |
2260 | struct timers_private *tp = m->private; |
2261 | int notify; | |
cedbccab | 2262 | static const char * const nstr[] = { |
57b8015e PE |
2263 | [SIGEV_SIGNAL] = "signal", |
2264 | [SIGEV_NONE] = "none", | |
2265 | [SIGEV_THREAD] = "thread", | |
2266 | }; | |
48f6a7a5 PE |
2267 | |
2268 | timer = list_entry((struct list_head *)v, struct k_itimer, list); | |
57b8015e PE |
2269 | notify = timer->it_sigev_notify; |
2270 | ||
48f6a7a5 | 2271 | seq_printf(m, "ID: %d\n", timer->it_id); |
25ce3191 JP |
2272 | seq_printf(m, "signal: %d/%p\n", |
2273 | timer->sigq->info.si_signo, | |
2274 | timer->sigq->info.si_value.sival_ptr); | |
57b8015e | 2275 | seq_printf(m, "notify: %s/%s.%d\n", |
25ce3191 JP |
2276 | nstr[notify & ~SIGEV_THREAD_ID], |
2277 | (notify & SIGEV_THREAD_ID) ? "tid" : "pid", | |
2278 | pid_nr_ns(timer->it_pid, tp->ns)); | |
15ef0298 | 2279 | seq_printf(m, "ClockID: %d\n", timer->it_clock); |
48f6a7a5 PE |
2280 | |
2281 | return 0; | |
2282 | } | |
2283 | ||
2284 | static const struct seq_operations proc_timers_seq_ops = { | |
2285 | .start = timers_start, | |
2286 | .next = timers_next, | |
2287 | .stop = timers_stop, | |
2288 | .show = show_timer, | |
2289 | }; | |
2290 | ||
2291 | static int proc_timers_open(struct inode *inode, struct file *file) | |
2292 | { | |
2293 | struct timers_private *tp; | |
2294 | ||
2295 | tp = __seq_open_private(file, &proc_timers_seq_ops, | |
2296 | sizeof(struct timers_private)); | |
2297 | if (!tp) | |
2298 | return -ENOMEM; | |
2299 | ||
2300 | tp->pid = proc_pid(inode); | |
57b8015e | 2301 | tp->ns = inode->i_sb->s_fs_info; |
48f6a7a5 PE |
2302 | return 0; |
2303 | } | |
2304 | ||
2305 | static const struct file_operations proc_timers_operations = { | |
2306 | .open = proc_timers_open, | |
2307 | .read = seq_read, | |
2308 | .llseek = seq_lseek, | |
2309 | .release = seq_release_private, | |
2310 | }; | |
b5946bea | 2311 | #endif |
640708a2 | 2312 | |
5de23d43 JS |
2313 | static ssize_t timerslack_ns_write(struct file *file, const char __user *buf, |
2314 | size_t count, loff_t *offset) | |
2315 | { | |
2316 | struct inode *inode = file_inode(file); | |
2317 | struct task_struct *p; | |
2318 | u64 slack_ns; | |
2319 | int err; | |
2320 | ||
2321 | err = kstrtoull_from_user(buf, count, 10, &slack_ns); | |
2322 | if (err < 0) | |
2323 | return err; | |
2324 | ||
2325 | p = get_proc_task(inode); | |
2326 | if (!p) | |
2327 | return -ESRCH; | |
2328 | ||
4b2bd5fe JS |
2329 | if (p != current) { |
2330 | if (!capable(CAP_SYS_NICE)) { | |
2331 | count = -EPERM; | |
2332 | goto out; | |
2333 | } | |
5de23d43 | 2334 | |
4b2bd5fe JS |
2335 | err = security_task_setscheduler(p); |
2336 | if (err) { | |
2337 | count = err; | |
2338 | goto out; | |
2339 | } | |
904763e1 JS |
2340 | } |
2341 | ||
7abbaf94 JS |
2342 | task_lock(p); |
2343 | if (slack_ns == 0) | |
2344 | p->timer_slack_ns = p->default_timer_slack_ns; | |
2345 | else | |
2346 | p->timer_slack_ns = slack_ns; | |
2347 | task_unlock(p); | |
2348 | ||
2349 | out: | |
5de23d43 JS |
2350 | put_task_struct(p); |
2351 | ||
2352 | return count; | |
2353 | } | |
2354 | ||
2355 | static int timerslack_ns_show(struct seq_file *m, void *v) | |
2356 | { | |
2357 | struct inode *inode = m->private; | |
2358 | struct task_struct *p; | |
7abbaf94 | 2359 | int err = 0; |
5de23d43 JS |
2360 | |
2361 | p = get_proc_task(inode); | |
2362 | if (!p) | |
2363 | return -ESRCH; | |
2364 | ||
4b2bd5fe | 2365 | if (p != current) { |
5de23d43 | 2366 | |
4b2bd5fe JS |
2367 | if (!capable(CAP_SYS_NICE)) { |
2368 | err = -EPERM; | |
2369 | goto out; | |
2370 | } | |
2371 | err = security_task_getscheduler(p); | |
2372 | if (err) | |
2373 | goto out; | |
2374 | } | |
904763e1 | 2375 | |
7abbaf94 JS |
2376 | task_lock(p); |
2377 | seq_printf(m, "%llu\n", p->timer_slack_ns); | |
2378 | task_unlock(p); | |
2379 | ||
2380 | out: | |
5de23d43 JS |
2381 | put_task_struct(p); |
2382 | ||
2383 | return err; | |
2384 | } | |
2385 | ||
2386 | static int timerslack_ns_open(struct inode *inode, struct file *filp) | |
2387 | { | |
2388 | return single_open(filp, timerslack_ns_show, inode); | |
2389 | } | |
2390 | ||
2391 | static const struct file_operations proc_pid_set_timerslack_ns_operations = { | |
2392 | .open = timerslack_ns_open, | |
2393 | .read = seq_read, | |
2394 | .write = timerslack_ns_write, | |
2395 | .llseek = seq_lseek, | |
2396 | .release = single_release, | |
2397 | }; | |
2398 | ||
c52a47ac | 2399 | static int proc_pident_instantiate(struct inode *dir, |
c5141e6d | 2400 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
444ceed8 | 2401 | { |
c5141e6d | 2402 | const struct pid_entry *p = ptr; |
444ceed8 EB |
2403 | struct inode *inode; |
2404 | struct proc_inode *ei; | |
444ceed8 | 2405 | |
db978da8 | 2406 | inode = proc_pid_make_inode(dir->i_sb, task, p->mode); |
444ceed8 EB |
2407 | if (!inode) |
2408 | goto out; | |
2409 | ||
2410 | ei = PROC_I(inode); | |
444ceed8 | 2411 | if (S_ISDIR(inode->i_mode)) |
bfe86848 | 2412 | set_nlink(inode, 2); /* Use getattr to fix if necessary */ |
444ceed8 EB |
2413 | if (p->iop) |
2414 | inode->i_op = p->iop; | |
2415 | if (p->fop) | |
2416 | inode->i_fop = p->fop; | |
2417 | ei->op = p->op; | |
fb045adb | 2418 | d_set_d_op(dentry, &pid_dentry_operations); |
444ceed8 EB |
2419 | d_add(dentry, inode); |
2420 | /* Close the race of the process dying before we return the dentry */ | |
0b728e19 | 2421 | if (pid_revalidate(dentry, 0)) |
c52a47ac | 2422 | return 0; |
444ceed8 | 2423 | out: |
c52a47ac | 2424 | return -ENOENT; |
444ceed8 EB |
2425 | } |
2426 | ||
1da177e4 LT |
2427 | static struct dentry *proc_pident_lookup(struct inode *dir, |
2428 | struct dentry *dentry, | |
c5141e6d | 2429 | const struct pid_entry *ents, |
7bcd6b0e | 2430 | unsigned int nents) |
1da177e4 | 2431 | { |
c52a47ac | 2432 | int error; |
99f89551 | 2433 | struct task_struct *task = get_proc_task(dir); |
c5141e6d | 2434 | const struct pid_entry *p, *last; |
1da177e4 | 2435 | |
c52a47ac | 2436 | error = -ENOENT; |
1da177e4 | 2437 | |
99f89551 EB |
2438 | if (!task) |
2439 | goto out_no_task; | |
1da177e4 | 2440 | |
20cdc894 EB |
2441 | /* |
2442 | * Yes, it does not scale. And it should not. Don't add | |
2443 | * new entries into /proc/<tgid>/ without very good reasons. | |
2444 | */ | |
bac5f5d5 AD |
2445 | last = &ents[nents]; |
2446 | for (p = ents; p < last; p++) { | |
1da177e4 LT |
2447 | if (p->len != dentry->d_name.len) |
2448 | continue; | |
2449 | if (!memcmp(dentry->d_name.name, p->name, p->len)) | |
2450 | break; | |
2451 | } | |
bac5f5d5 | 2452 | if (p >= last) |
1da177e4 LT |
2453 | goto out; |
2454 | ||
444ceed8 | 2455 | error = proc_pident_instantiate(dir, dentry, task, p); |
1da177e4 | 2456 | out: |
99f89551 EB |
2457 | put_task_struct(task); |
2458 | out_no_task: | |
c52a47ac | 2459 | return ERR_PTR(error); |
1da177e4 LT |
2460 | } |
2461 | ||
f0c3b509 | 2462 | static int proc_pident_readdir(struct file *file, struct dir_context *ctx, |
c5141e6d | 2463 | const struct pid_entry *ents, unsigned int nents) |
28a6d671 | 2464 | { |
f0c3b509 AV |
2465 | struct task_struct *task = get_proc_task(file_inode(file)); |
2466 | const struct pid_entry *p; | |
28a6d671 | 2467 | |
28a6d671 | 2468 | if (!task) |
f0c3b509 | 2469 | return -ENOENT; |
28a6d671 | 2470 | |
f0c3b509 AV |
2471 | if (!dir_emit_dots(file, ctx)) |
2472 | goto out; | |
2473 | ||
2474 | if (ctx->pos >= nents + 2) | |
2475 | goto out; | |
28a6d671 | 2476 | |
bac5f5d5 | 2477 | for (p = ents + (ctx->pos - 2); p < ents + nents; p++) { |
f0c3b509 AV |
2478 | if (!proc_fill_cache(file, ctx, p->name, p->len, |
2479 | proc_pident_instantiate, task, p)) | |
2480 | break; | |
2481 | ctx->pos++; | |
2482 | } | |
28a6d671 | 2483 | out: |
61a28784 | 2484 | put_task_struct(task); |
f0c3b509 | 2485 | return 0; |
1da177e4 LT |
2486 | } |
2487 | ||
28a6d671 EB |
2488 | #ifdef CONFIG_SECURITY |
2489 | static ssize_t proc_pid_attr_read(struct file * file, char __user * buf, | |
2490 | size_t count, loff_t *ppos) | |
2491 | { | |
496ad9aa | 2492 | struct inode * inode = file_inode(file); |
04ff9708 | 2493 | char *p = NULL; |
28a6d671 EB |
2494 | ssize_t length; |
2495 | struct task_struct *task = get_proc_task(inode); | |
2496 | ||
28a6d671 | 2497 | if (!task) |
04ff9708 | 2498 | return -ESRCH; |
28a6d671 EB |
2499 | |
2500 | length = security_getprocattr(task, | |
2fddfeef | 2501 | (char*)file->f_path.dentry->d_name.name, |
04ff9708 | 2502 | &p); |
28a6d671 | 2503 | put_task_struct(task); |
04ff9708 AV |
2504 | if (length > 0) |
2505 | length = simple_read_from_buffer(buf, count, ppos, p, length); | |
2506 | kfree(p); | |
28a6d671 | 2507 | return length; |
1da177e4 LT |
2508 | } |
2509 | ||
28a6d671 EB |
2510 | static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf, |
2511 | size_t count, loff_t *ppos) | |
2512 | { | |
496ad9aa | 2513 | struct inode * inode = file_inode(file); |
bb646cdb | 2514 | void *page; |
28a6d671 EB |
2515 | ssize_t length; |
2516 | struct task_struct *task = get_proc_task(inode); | |
2517 | ||
2518 | length = -ESRCH; | |
2519 | if (!task) | |
2520 | goto out_no_task; | |
b21507e2 SS |
2521 | |
2522 | /* A task may only write its own attributes. */ | |
2523 | length = -EACCES; | |
2524 | if (current != task) | |
2525 | goto out; | |
2526 | ||
28a6d671 EB |
2527 | if (count > PAGE_SIZE) |
2528 | count = PAGE_SIZE; | |
2529 | ||
2530 | /* No partial writes. */ | |
2531 | length = -EINVAL; | |
2532 | if (*ppos != 0) | |
2533 | goto out; | |
2534 | ||
bb646cdb AV |
2535 | page = memdup_user(buf, count); |
2536 | if (IS_ERR(page)) { | |
2537 | length = PTR_ERR(page); | |
28a6d671 | 2538 | goto out; |
bb646cdb | 2539 | } |
28a6d671 | 2540 | |
107db7c7 | 2541 | /* Guard against adverse ptrace interaction */ |
b21507e2 | 2542 | length = mutex_lock_interruptible(¤t->signal->cred_guard_mutex); |
107db7c7 DH |
2543 | if (length < 0) |
2544 | goto out_free; | |
2545 | ||
b21507e2 | 2546 | length = security_setprocattr(file->f_path.dentry->d_name.name, |
bb646cdb | 2547 | page, count); |
b21507e2 | 2548 | mutex_unlock(¤t->signal->cred_guard_mutex); |
28a6d671 | 2549 | out_free: |
bb646cdb | 2550 | kfree(page); |
28a6d671 EB |
2551 | out: |
2552 | put_task_struct(task); | |
2553 | out_no_task: | |
2554 | return length; | |
2555 | } | |
2556 | ||
00977a59 | 2557 | static const struct file_operations proc_pid_attr_operations = { |
28a6d671 EB |
2558 | .read = proc_pid_attr_read, |
2559 | .write = proc_pid_attr_write, | |
87df8424 | 2560 | .llseek = generic_file_llseek, |
28a6d671 EB |
2561 | }; |
2562 | ||
c5141e6d | 2563 | static const struct pid_entry attr_dir_stuff[] = { |
631f9c18 AD |
2564 | REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations), |
2565 | REG("prev", S_IRUGO, proc_pid_attr_operations), | |
2566 | REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2567 | REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2568 | REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2569 | REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
28a6d671 EB |
2570 | }; |
2571 | ||
f0c3b509 | 2572 | static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx) |
28a6d671 | 2573 | { |
f0c3b509 AV |
2574 | return proc_pident_readdir(file, ctx, |
2575 | attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2576 | } |
2577 | ||
00977a59 | 2578 | static const struct file_operations proc_attr_dir_operations = { |
1da177e4 | 2579 | .read = generic_read_dir, |
f50752ea AV |
2580 | .iterate_shared = proc_attr_dir_readdir, |
2581 | .llseek = generic_file_llseek, | |
1da177e4 LT |
2582 | }; |
2583 | ||
72d9dcfc | 2584 | static struct dentry *proc_attr_dir_lookup(struct inode *dir, |
00cd8dd3 | 2585 | struct dentry *dentry, unsigned int flags) |
28a6d671 | 2586 | { |
7bcd6b0e EB |
2587 | return proc_pident_lookup(dir, dentry, |
2588 | attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2589 | } |
2590 | ||
c5ef1c42 | 2591 | static const struct inode_operations proc_attr_dir_inode_operations = { |
72d9dcfc | 2592 | .lookup = proc_attr_dir_lookup, |
99f89551 | 2593 | .getattr = pid_getattr, |
6d76fa58 | 2594 | .setattr = proc_setattr, |
1da177e4 LT |
2595 | }; |
2596 | ||
28a6d671 EB |
2597 | #endif |
2598 | ||
698ba7b5 | 2599 | #ifdef CONFIG_ELF_CORE |
3cb4a0bb KH |
2600 | static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf, |
2601 | size_t count, loff_t *ppos) | |
2602 | { | |
496ad9aa | 2603 | struct task_struct *task = get_proc_task(file_inode(file)); |
3cb4a0bb KH |
2604 | struct mm_struct *mm; |
2605 | char buffer[PROC_NUMBUF]; | |
2606 | size_t len; | |
2607 | int ret; | |
2608 | ||
2609 | if (!task) | |
2610 | return -ESRCH; | |
2611 | ||
2612 | ret = 0; | |
2613 | mm = get_task_mm(task); | |
2614 | if (mm) { | |
2615 | len = snprintf(buffer, sizeof(buffer), "%08lx\n", | |
2616 | ((mm->flags & MMF_DUMP_FILTER_MASK) >> | |
2617 | MMF_DUMP_FILTER_SHIFT)); | |
2618 | mmput(mm); | |
2619 | ret = simple_read_from_buffer(buf, count, ppos, buffer, len); | |
2620 | } | |
2621 | ||
2622 | put_task_struct(task); | |
2623 | ||
2624 | return ret; | |
2625 | } | |
2626 | ||
2627 | static ssize_t proc_coredump_filter_write(struct file *file, | |
2628 | const char __user *buf, | |
2629 | size_t count, | |
2630 | loff_t *ppos) | |
2631 | { | |
2632 | struct task_struct *task; | |
2633 | struct mm_struct *mm; | |
3cb4a0bb KH |
2634 | unsigned int val; |
2635 | int ret; | |
2636 | int i; | |
2637 | unsigned long mask; | |
2638 | ||
774636e1 AD |
2639 | ret = kstrtouint_from_user(buf, count, 0, &val); |
2640 | if (ret < 0) | |
2641 | return ret; | |
3cb4a0bb KH |
2642 | |
2643 | ret = -ESRCH; | |
496ad9aa | 2644 | task = get_proc_task(file_inode(file)); |
3cb4a0bb KH |
2645 | if (!task) |
2646 | goto out_no_task; | |
2647 | ||
3cb4a0bb KH |
2648 | mm = get_task_mm(task); |
2649 | if (!mm) | |
2650 | goto out_no_mm; | |
41a0c249 | 2651 | ret = 0; |
3cb4a0bb KH |
2652 | |
2653 | for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) { | |
2654 | if (val & mask) | |
2655 | set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2656 | else | |
2657 | clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2658 | } | |
2659 | ||
2660 | mmput(mm); | |
2661 | out_no_mm: | |
2662 | put_task_struct(task); | |
2663 | out_no_task: | |
774636e1 AD |
2664 | if (ret < 0) |
2665 | return ret; | |
2666 | return count; | |
3cb4a0bb KH |
2667 | } |
2668 | ||
2669 | static const struct file_operations proc_coredump_filter_operations = { | |
2670 | .read = proc_coredump_filter_read, | |
2671 | .write = proc_coredump_filter_write, | |
87df8424 | 2672 | .llseek = generic_file_llseek, |
3cb4a0bb KH |
2673 | }; |
2674 | #endif | |
2675 | ||
aba76fdb | 2676 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 2677 | static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole) |
297c5d92 | 2678 | { |
940389b8 | 2679 | struct task_io_accounting acct = task->ioac; |
5995477a | 2680 | unsigned long flags; |
293eb1e7 | 2681 | int result; |
5995477a | 2682 | |
293eb1e7 VK |
2683 | result = mutex_lock_killable(&task->signal->cred_guard_mutex); |
2684 | if (result) | |
2685 | return result; | |
2686 | ||
caaee623 | 2687 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) { |
293eb1e7 VK |
2688 | result = -EACCES; |
2689 | goto out_unlock; | |
2690 | } | |
1d1221f3 | 2691 | |
5995477a AR |
2692 | if (whole && lock_task_sighand(task, &flags)) { |
2693 | struct task_struct *t = task; | |
2694 | ||
2695 | task_io_accounting_add(&acct, &task->signal->ioac); | |
2696 | while_each_thread(task, t) | |
2697 | task_io_accounting_add(&acct, &t->ioac); | |
2698 | ||
2699 | unlock_task_sighand(task, &flags); | |
297c5d92 | 2700 | } |
25ce3191 JP |
2701 | seq_printf(m, |
2702 | "rchar: %llu\n" | |
2703 | "wchar: %llu\n" | |
2704 | "syscr: %llu\n" | |
2705 | "syscw: %llu\n" | |
2706 | "read_bytes: %llu\n" | |
2707 | "write_bytes: %llu\n" | |
2708 | "cancelled_write_bytes: %llu\n", | |
2709 | (unsigned long long)acct.rchar, | |
2710 | (unsigned long long)acct.wchar, | |
2711 | (unsigned long long)acct.syscr, | |
2712 | (unsigned long long)acct.syscw, | |
2713 | (unsigned long long)acct.read_bytes, | |
2714 | (unsigned long long)acct.write_bytes, | |
2715 | (unsigned long long)acct.cancelled_write_bytes); | |
2716 | result = 0; | |
2717 | ||
293eb1e7 VK |
2718 | out_unlock: |
2719 | mutex_unlock(&task->signal->cred_guard_mutex); | |
2720 | return result; | |
297c5d92 AR |
2721 | } |
2722 | ||
19aadc98 AD |
2723 | static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns, |
2724 | struct pid *pid, struct task_struct *task) | |
297c5d92 | 2725 | { |
19aadc98 | 2726 | return do_io_accounting(task, m, 0); |
aba76fdb | 2727 | } |
297c5d92 | 2728 | |
19aadc98 AD |
2729 | static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns, |
2730 | struct pid *pid, struct task_struct *task) | |
297c5d92 | 2731 | { |
19aadc98 | 2732 | return do_io_accounting(task, m, 1); |
297c5d92 AR |
2733 | } |
2734 | #endif /* CONFIG_TASK_IO_ACCOUNTING */ | |
aba76fdb | 2735 | |
22d917d8 EB |
2736 | #ifdef CONFIG_USER_NS |
2737 | static int proc_id_map_open(struct inode *inode, struct file *file, | |
ccf94f1b | 2738 | const struct seq_operations *seq_ops) |
22d917d8 EB |
2739 | { |
2740 | struct user_namespace *ns = NULL; | |
2741 | struct task_struct *task; | |
2742 | struct seq_file *seq; | |
2743 | int ret = -EINVAL; | |
2744 | ||
2745 | task = get_proc_task(inode); | |
2746 | if (task) { | |
2747 | rcu_read_lock(); | |
2748 | ns = get_user_ns(task_cred_xxx(task, user_ns)); | |
2749 | rcu_read_unlock(); | |
2750 | put_task_struct(task); | |
2751 | } | |
2752 | if (!ns) | |
2753 | goto err; | |
2754 | ||
2755 | ret = seq_open(file, seq_ops); | |
2756 | if (ret) | |
2757 | goto err_put_ns; | |
2758 | ||
2759 | seq = file->private_data; | |
2760 | seq->private = ns; | |
2761 | ||
2762 | return 0; | |
2763 | err_put_ns: | |
2764 | put_user_ns(ns); | |
2765 | err: | |
2766 | return ret; | |
2767 | } | |
2768 | ||
2769 | static int proc_id_map_release(struct inode *inode, struct file *file) | |
2770 | { | |
2771 | struct seq_file *seq = file->private_data; | |
2772 | struct user_namespace *ns = seq->private; | |
2773 | put_user_ns(ns); | |
2774 | return seq_release(inode, file); | |
2775 | } | |
2776 | ||
2777 | static int proc_uid_map_open(struct inode *inode, struct file *file) | |
2778 | { | |
2779 | return proc_id_map_open(inode, file, &proc_uid_seq_operations); | |
2780 | } | |
2781 | ||
2782 | static int proc_gid_map_open(struct inode *inode, struct file *file) | |
2783 | { | |
2784 | return proc_id_map_open(inode, file, &proc_gid_seq_operations); | |
2785 | } | |
2786 | ||
f76d207a EB |
2787 | static int proc_projid_map_open(struct inode *inode, struct file *file) |
2788 | { | |
2789 | return proc_id_map_open(inode, file, &proc_projid_seq_operations); | |
2790 | } | |
2791 | ||
22d917d8 EB |
2792 | static const struct file_operations proc_uid_map_operations = { |
2793 | .open = proc_uid_map_open, | |
2794 | .write = proc_uid_map_write, | |
2795 | .read = seq_read, | |
2796 | .llseek = seq_lseek, | |
2797 | .release = proc_id_map_release, | |
2798 | }; | |
2799 | ||
2800 | static const struct file_operations proc_gid_map_operations = { | |
2801 | .open = proc_gid_map_open, | |
2802 | .write = proc_gid_map_write, | |
2803 | .read = seq_read, | |
2804 | .llseek = seq_lseek, | |
2805 | .release = proc_id_map_release, | |
2806 | }; | |
f76d207a EB |
2807 | |
2808 | static const struct file_operations proc_projid_map_operations = { | |
2809 | .open = proc_projid_map_open, | |
2810 | .write = proc_projid_map_write, | |
2811 | .read = seq_read, | |
2812 | .llseek = seq_lseek, | |
2813 | .release = proc_id_map_release, | |
2814 | }; | |
9cc46516 EB |
2815 | |
2816 | static int proc_setgroups_open(struct inode *inode, struct file *file) | |
2817 | { | |
2818 | struct user_namespace *ns = NULL; | |
2819 | struct task_struct *task; | |
2820 | int ret; | |
2821 | ||
2822 | ret = -ESRCH; | |
2823 | task = get_proc_task(inode); | |
2824 | if (task) { | |
2825 | rcu_read_lock(); | |
2826 | ns = get_user_ns(task_cred_xxx(task, user_ns)); | |
2827 | rcu_read_unlock(); | |
2828 | put_task_struct(task); | |
2829 | } | |
2830 | if (!ns) | |
2831 | goto err; | |
2832 | ||
2833 | if (file->f_mode & FMODE_WRITE) { | |
2834 | ret = -EACCES; | |
2835 | if (!ns_capable(ns, CAP_SYS_ADMIN)) | |
2836 | goto err_put_ns; | |
2837 | } | |
2838 | ||
2839 | ret = single_open(file, &proc_setgroups_show, ns); | |
2840 | if (ret) | |
2841 | goto err_put_ns; | |
2842 | ||
2843 | return 0; | |
2844 | err_put_ns: | |
2845 | put_user_ns(ns); | |
2846 | err: | |
2847 | return ret; | |
2848 | } | |
2849 | ||
2850 | static int proc_setgroups_release(struct inode *inode, struct file *file) | |
2851 | { | |
2852 | struct seq_file *seq = file->private_data; | |
2853 | struct user_namespace *ns = seq->private; | |
2854 | int ret = single_release(inode, file); | |
2855 | put_user_ns(ns); | |
2856 | return ret; | |
2857 | } | |
2858 | ||
2859 | static const struct file_operations proc_setgroups_operations = { | |
2860 | .open = proc_setgroups_open, | |
2861 | .write = proc_setgroups_write, | |
2862 | .read = seq_read, | |
2863 | .llseek = seq_lseek, | |
2864 | .release = proc_setgroups_release, | |
2865 | }; | |
22d917d8 EB |
2866 | #endif /* CONFIG_USER_NS */ |
2867 | ||
47830723 KC |
2868 | static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns, |
2869 | struct pid *pid, struct task_struct *task) | |
2870 | { | |
a9712bc1 AV |
2871 | int err = lock_trace(task); |
2872 | if (!err) { | |
2873 | seq_printf(m, "%08x\n", task->personality); | |
2874 | unlock_trace(task); | |
2875 | } | |
2876 | return err; | |
47830723 KC |
2877 | } |
2878 | ||
7c23b330 JP |
2879 | #ifdef CONFIG_LIVEPATCH |
2880 | static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns, | |
2881 | struct pid *pid, struct task_struct *task) | |
2882 | { | |
2883 | seq_printf(m, "%d\n", task->patch_state); | |
2884 | return 0; | |
2885 | } | |
2886 | #endif /* CONFIG_LIVEPATCH */ | |
2887 | ||
28a6d671 EB |
2888 | /* |
2889 | * Thread groups | |
2890 | */ | |
00977a59 | 2891 | static const struct file_operations proc_task_operations; |
c5ef1c42 | 2892 | static const struct inode_operations proc_task_inode_operations; |
20cdc894 | 2893 | |
c5141e6d | 2894 | static const struct pid_entry tgid_base_stuff[] = { |
631f9c18 AD |
2895 | DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations), |
2896 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), | |
640708a2 | 2897 | DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations), |
631f9c18 | 2898 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 2899 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
b2211a36 | 2900 | #ifdef CONFIG_NET |
631f9c18 | 2901 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), |
b2211a36 | 2902 | #endif |
631f9c18 | 2903 | REG("environ", S_IRUSR, proc_environ_operations), |
c5317167 | 2904 | REG("auxv", S_IRUSR, proc_auxv_operations), |
631f9c18 | 2905 | ONE("status", S_IRUGO, proc_pid_status), |
35a35046 | 2906 | ONE("personality", S_IRUSR, proc_pid_personality), |
1c963eb1 | 2907 | ONE("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 2908 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 2909 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
5091faa4 MG |
2910 | #endif |
2911 | #ifdef CONFIG_SCHED_AUTOGROUP | |
2912 | REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations), | |
ebcb6734 | 2913 | #endif |
4614a696 | 2914 | REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations), |
ebcb6734 | 2915 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 | 2916 | ONE("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 2917 | #endif |
c2c0bb44 | 2918 | REG("cmdline", S_IRUGO, proc_pid_cmdline_ops), |
631f9c18 AD |
2919 | ONE("stat", S_IRUGO, proc_tgid_stat), |
2920 | ONE("statm", S_IRUGO, proc_pid_statm), | |
b7643757 | 2921 | REG("maps", S_IRUGO, proc_pid_maps_operations), |
28a6d671 | 2922 | #ifdef CONFIG_NUMA |
b7643757 | 2923 | REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations), |
28a6d671 | 2924 | #endif |
631f9c18 AD |
2925 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
2926 | LNK("cwd", proc_cwd_link), | |
2927 | LNK("root", proc_root_link), | |
2928 | LNK("exe", proc_exe_link), | |
2929 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
2930 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
2931 | REG("mountstats", S_IRUSR, proc_mountstats_operations), | |
1e883281 | 2932 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 2933 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
b7643757 | 2934 | REG("smaps", S_IRUGO, proc_pid_smaps_operations), |
493b0e9d | 2935 | REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations), |
32ed74a4 | 2936 | REG("pagemap", S_IRUSR, proc_pagemap_operations), |
28a6d671 EB |
2937 | #endif |
2938 | #ifdef CONFIG_SECURITY | |
631f9c18 | 2939 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
2940 | #endif |
2941 | #ifdef CONFIG_KALLSYMS | |
edfcd606 | 2942 | ONE("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 2943 | #endif |
2ec220e2 | 2944 | #ifdef CONFIG_STACKTRACE |
35a35046 | 2945 | ONE("stack", S_IRUSR, proc_pid_stack), |
28a6d671 | 2946 | #endif |
5968cece | 2947 | #ifdef CONFIG_SCHED_INFO |
f6e826ca | 2948 | ONE("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 2949 | #endif |
9745512c | 2950 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 2951 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 2952 | #endif |
8793d854 | 2953 | #ifdef CONFIG_PROC_PID_CPUSET |
52de4779 | 2954 | ONE("cpuset", S_IRUGO, proc_cpuset_show), |
a424316c PM |
2955 | #endif |
2956 | #ifdef CONFIG_CGROUPS | |
006f4ac4 | 2957 | ONE("cgroup", S_IRUGO, proc_cgroup_show), |
28a6d671 | 2958 | #endif |
6ba51e37 | 2959 | ONE("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 2960 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 2961 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
28a6d671 | 2962 | #ifdef CONFIG_AUDITSYSCALL |
631f9c18 AD |
2963 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
2964 | REG("sessionid", S_IRUGO, proc_sessionid_operations), | |
28a6d671 | 2965 | #endif |
f4f154fd | 2966 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 2967 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
168c42bc | 2968 | REG("fail-nth", 0644, proc_fail_nth_operations), |
f4f154fd | 2969 | #endif |
698ba7b5 | 2970 | #ifdef CONFIG_ELF_CORE |
631f9c18 | 2971 | REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations), |
3cb4a0bb | 2972 | #endif |
aba76fdb | 2973 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 2974 | ONE("io", S_IRUSR, proc_tgid_io_accounting), |
aba76fdb | 2975 | #endif |
f133ecca | 2976 | #ifdef CONFIG_HARDWALL |
d962c144 | 2977 | ONE("hardwall", S_IRUGO, proc_pid_hardwall), |
f133ecca | 2978 | #endif |
22d917d8 EB |
2979 | #ifdef CONFIG_USER_NS |
2980 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
2981 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 2982 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
9cc46516 | 2983 | REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations), |
22d917d8 | 2984 | #endif |
b18b6a9c | 2985 | #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS) |
48f6a7a5 PE |
2986 | REG("timers", S_IRUGO, proc_timers_operations), |
2987 | #endif | |
5de23d43 | 2988 | REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations), |
7c23b330 JP |
2989 | #ifdef CONFIG_LIVEPATCH |
2990 | ONE("patch_state", S_IRUSR, proc_pid_patch_state), | |
2991 | #endif | |
28a6d671 | 2992 | }; |
1da177e4 | 2993 | |
f0c3b509 | 2994 | static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 2995 | { |
f0c3b509 AV |
2996 | return proc_pident_readdir(file, ctx, |
2997 | tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
2998 | } |
2999 | ||
00977a59 | 3000 | static const struct file_operations proc_tgid_base_operations = { |
1da177e4 | 3001 | .read = generic_read_dir, |
f50752ea AV |
3002 | .iterate_shared = proc_tgid_base_readdir, |
3003 | .llseek = generic_file_llseek, | |
1da177e4 LT |
3004 | }; |
3005 | ||
00cd8dd3 AV |
3006 | static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
3007 | { | |
7bcd6b0e EB |
3008 | return proc_pident_lookup(dir, dentry, |
3009 | tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
3010 | } |
3011 | ||
c5ef1c42 | 3012 | static const struct inode_operations proc_tgid_base_inode_operations = { |
28a6d671 | 3013 | .lookup = proc_tgid_base_lookup, |
99f89551 | 3014 | .getattr = pid_getattr, |
6d76fa58 | 3015 | .setattr = proc_setattr, |
0499680a | 3016 | .permission = proc_pid_permission, |
1da177e4 | 3017 | }; |
1da177e4 | 3018 | |
60347f67 | 3019 | static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid) |
1da177e4 | 3020 | { |
48e6484d | 3021 | struct dentry *dentry, *leader, *dir; |
8578cea7 | 3022 | char buf[PROC_NUMBUF]; |
48e6484d EB |
3023 | struct qstr name; |
3024 | ||
3025 | name.name = buf; | |
60347f67 | 3026 | name.len = snprintf(buf, sizeof(buf), "%d", pid); |
4f522a24 | 3027 | /* no ->d_hash() rejects on procfs */ |
60347f67 | 3028 | dentry = d_hash_and_lookup(mnt->mnt_root, &name); |
48e6484d | 3029 | if (dentry) { |
bbd51924 | 3030 | d_invalidate(dentry); |
48e6484d EB |
3031 | dput(dentry); |
3032 | } | |
1da177e4 | 3033 | |
c35a7f18 ON |
3034 | if (pid == tgid) |
3035 | return; | |
3036 | ||
48e6484d | 3037 | name.name = buf; |
60347f67 PE |
3038 | name.len = snprintf(buf, sizeof(buf), "%d", tgid); |
3039 | leader = d_hash_and_lookup(mnt->mnt_root, &name); | |
48e6484d EB |
3040 | if (!leader) |
3041 | goto out; | |
1da177e4 | 3042 | |
48e6484d EB |
3043 | name.name = "task"; |
3044 | name.len = strlen(name.name); | |
3045 | dir = d_hash_and_lookup(leader, &name); | |
3046 | if (!dir) | |
3047 | goto out_put_leader; | |
3048 | ||
3049 | name.name = buf; | |
60347f67 | 3050 | name.len = snprintf(buf, sizeof(buf), "%d", pid); |
48e6484d EB |
3051 | dentry = d_hash_and_lookup(dir, &name); |
3052 | if (dentry) { | |
bbd51924 | 3053 | d_invalidate(dentry); |
48e6484d | 3054 | dput(dentry); |
1da177e4 | 3055 | } |
48e6484d EB |
3056 | |
3057 | dput(dir); | |
3058 | out_put_leader: | |
3059 | dput(leader); | |
3060 | out: | |
3061 | return; | |
1da177e4 LT |
3062 | } |
3063 | ||
0895e91d RD |
3064 | /** |
3065 | * proc_flush_task - Remove dcache entries for @task from the /proc dcache. | |
3066 | * @task: task that should be flushed. | |
3067 | * | |
3068 | * When flushing dentries from proc, one needs to flush them from global | |
60347f67 | 3069 | * proc (proc_mnt) and from all the namespaces' procs this task was seen |
0895e91d RD |
3070 | * in. This call is supposed to do all of this job. |
3071 | * | |
3072 | * Looks in the dcache for | |
3073 | * /proc/@pid | |
3074 | * /proc/@tgid/task/@pid | |
3075 | * if either directory is present flushes it and all of it'ts children | |
3076 | * from the dcache. | |
3077 | * | |
3078 | * It is safe and reasonable to cache /proc entries for a task until | |
3079 | * that task exits. After that they just clog up the dcache with | |
3080 | * useless entries, possibly causing useful dcache entries to be | |
3081 | * flushed instead. This routine is proved to flush those useless | |
3082 | * dcache entries at process exit time. | |
3083 | * | |
3084 | * NOTE: This routine is just an optimization so it does not guarantee | |
3085 | * that no dcache entries will exist at process exit time it | |
3086 | * just makes it very unlikely that any will persist. | |
60347f67 PE |
3087 | */ |
3088 | ||
3089 | void proc_flush_task(struct task_struct *task) | |
3090 | { | |
9fcc2d15 | 3091 | int i; |
9b4d1cbe | 3092 | struct pid *pid, *tgid; |
130f77ec PE |
3093 | struct upid *upid; |
3094 | ||
130f77ec | 3095 | pid = task_pid(task); |
9b4d1cbe | 3096 | tgid = task_tgid(task); |
130f77ec | 3097 | |
9fcc2d15 | 3098 | for (i = 0; i <= pid->level; i++) { |
130f77ec PE |
3099 | upid = &pid->numbers[i]; |
3100 | proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr, | |
9b4d1cbe | 3101 | tgid->numbers[i].nr); |
130f77ec | 3102 | } |
60347f67 PE |
3103 | } |
3104 | ||
c52a47ac AV |
3105 | static int proc_pid_instantiate(struct inode *dir, |
3106 | struct dentry * dentry, | |
3107 | struct task_struct *task, const void *ptr) | |
444ceed8 | 3108 | { |
444ceed8 EB |
3109 | struct inode *inode; |
3110 | ||
db978da8 | 3111 | inode = proc_pid_make_inode(dir->i_sb, task, S_IFDIR | S_IRUGO | S_IXUGO); |
444ceed8 EB |
3112 | if (!inode) |
3113 | goto out; | |
3114 | ||
444ceed8 EB |
3115 | inode->i_op = &proc_tgid_base_inode_operations; |
3116 | inode->i_fop = &proc_tgid_base_operations; | |
3117 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 | 3118 | |
1270dd8d | 3119 | set_nlink(inode, nlink_tgid); |
444ceed8 | 3120 | |
fb045adb | 3121 | d_set_d_op(dentry, &pid_dentry_operations); |
444ceed8 EB |
3122 | |
3123 | d_add(dentry, inode); | |
3124 | /* Close the race of the process dying before we return the dentry */ | |
0b728e19 | 3125 | if (pid_revalidate(dentry, 0)) |
c52a47ac | 3126 | return 0; |
444ceed8 | 3127 | out: |
c52a47ac | 3128 | return -ENOENT; |
444ceed8 EB |
3129 | } |
3130 | ||
00cd8dd3 | 3131 | struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags) |
1da177e4 | 3132 | { |
335eb531 | 3133 | int result = -ENOENT; |
1da177e4 | 3134 | struct task_struct *task; |
1da177e4 | 3135 | unsigned tgid; |
b488893a | 3136 | struct pid_namespace *ns; |
1da177e4 | 3137 | |
dbcdb504 | 3138 | tgid = name_to_int(&dentry->d_name); |
1da177e4 LT |
3139 | if (tgid == ~0U) |
3140 | goto out; | |
3141 | ||
b488893a | 3142 | ns = dentry->d_sb->s_fs_info; |
de758734 | 3143 | rcu_read_lock(); |
b488893a | 3144 | task = find_task_by_pid_ns(tgid, ns); |
1da177e4 LT |
3145 | if (task) |
3146 | get_task_struct(task); | |
de758734 | 3147 | rcu_read_unlock(); |
1da177e4 LT |
3148 | if (!task) |
3149 | goto out; | |
3150 | ||
444ceed8 | 3151 | result = proc_pid_instantiate(dir, dentry, task, NULL); |
1da177e4 | 3152 | put_task_struct(task); |
1da177e4 | 3153 | out: |
c52a47ac | 3154 | return ERR_PTR(result); |
1da177e4 LT |
3155 | } |
3156 | ||
1da177e4 | 3157 | /* |
0804ef4b | 3158 | * Find the first task with tgid >= tgid |
0bc58a91 | 3159 | * |
1da177e4 | 3160 | */ |
19fd4bb2 EB |
3161 | struct tgid_iter { |
3162 | unsigned int tgid; | |
0804ef4b | 3163 | struct task_struct *task; |
19fd4bb2 EB |
3164 | }; |
3165 | static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter) | |
3166 | { | |
0804ef4b | 3167 | struct pid *pid; |
1da177e4 | 3168 | |
19fd4bb2 EB |
3169 | if (iter.task) |
3170 | put_task_struct(iter.task); | |
454cc105 | 3171 | rcu_read_lock(); |
0804ef4b | 3172 | retry: |
19fd4bb2 EB |
3173 | iter.task = NULL; |
3174 | pid = find_ge_pid(iter.tgid, ns); | |
0804ef4b | 3175 | if (pid) { |
19fd4bb2 EB |
3176 | iter.tgid = pid_nr_ns(pid, ns); |
3177 | iter.task = pid_task(pid, PIDTYPE_PID); | |
0804ef4b EB |
3178 | /* What we to know is if the pid we have find is the |
3179 | * pid of a thread_group_leader. Testing for task | |
3180 | * being a thread_group_leader is the obvious thing | |
3181 | * todo but there is a window when it fails, due to | |
3182 | * the pid transfer logic in de_thread. | |
3183 | * | |
3184 | * So we perform the straight forward test of seeing | |
3185 | * if the pid we have found is the pid of a thread | |
3186 | * group leader, and don't worry if the task we have | |
3187 | * found doesn't happen to be a thread group leader. | |
3188 | * As we don't care in the case of readdir. | |
3189 | */ | |
19fd4bb2 EB |
3190 | if (!iter.task || !has_group_leader_pid(iter.task)) { |
3191 | iter.tgid += 1; | |
0804ef4b | 3192 | goto retry; |
19fd4bb2 EB |
3193 | } |
3194 | get_task_struct(iter.task); | |
0bc58a91 | 3195 | } |
454cc105 | 3196 | rcu_read_unlock(); |
19fd4bb2 | 3197 | return iter; |
1da177e4 LT |
3198 | } |
3199 | ||
0097875b | 3200 | #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2) |
0804ef4b | 3201 | |
1da177e4 | 3202 | /* for the /proc/ directory itself, after non-process stuff has been done */ |
f0c3b509 | 3203 | int proc_pid_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3204 | { |
19fd4bb2 | 3205 | struct tgid_iter iter; |
3aa3377f | 3206 | struct pid_namespace *ns = file_inode(file)->i_sb->s_fs_info; |
f0c3b509 | 3207 | loff_t pos = ctx->pos; |
1da177e4 | 3208 | |
021ada7d | 3209 | if (pos >= PID_MAX_LIMIT + TGID_OFFSET) |
f0c3b509 | 3210 | return 0; |
1da177e4 | 3211 | |
0097875b | 3212 | if (pos == TGID_OFFSET - 2) { |
2b0143b5 | 3213 | struct inode *inode = d_inode(ns->proc_self); |
db963164 | 3214 | if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK)) |
f0c3b509 | 3215 | return 0; |
0097875b EB |
3216 | ctx->pos = pos = pos + 1; |
3217 | } | |
3218 | if (pos == TGID_OFFSET - 1) { | |
2b0143b5 | 3219 | struct inode *inode = d_inode(ns->proc_thread_self); |
0097875b EB |
3220 | if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK)) |
3221 | return 0; | |
3222 | ctx->pos = pos = pos + 1; | |
021ada7d | 3223 | } |
0097875b | 3224 | iter.tgid = pos - TGID_OFFSET; |
19fd4bb2 | 3225 | iter.task = NULL; |
19fd4bb2 EB |
3226 | for (iter = next_tgid(ns, iter); |
3227 | iter.task; | |
3228 | iter.tgid += 1, iter = next_tgid(ns, iter)) { | |
f0c3b509 AV |
3229 | char name[PROC_NUMBUF]; |
3230 | int len; | |
3ba4bcee ED |
3231 | |
3232 | cond_resched(); | |
796f571b | 3233 | if (!has_pid_permissions(ns, iter.task, HIDEPID_INVISIBLE)) |
f0c3b509 | 3234 | continue; |
0499680a | 3235 | |
f0c3b509 AV |
3236 | len = snprintf(name, sizeof(name), "%d", iter.tgid); |
3237 | ctx->pos = iter.tgid + TGID_OFFSET; | |
3238 | if (!proc_fill_cache(file, ctx, name, len, | |
3239 | proc_pid_instantiate, iter.task, NULL)) { | |
19fd4bb2 | 3240 | put_task_struct(iter.task); |
f0c3b509 | 3241 | return 0; |
1da177e4 | 3242 | } |
0bc58a91 | 3243 | } |
f0c3b509 | 3244 | ctx->pos = PID_MAX_LIMIT + TGID_OFFSET; |
0bc58a91 EB |
3245 | return 0; |
3246 | } | |
1da177e4 | 3247 | |
1b3044e3 JD |
3248 | /* |
3249 | * proc_tid_comm_permission is a special permission function exclusively | |
3250 | * used for the node /proc/<pid>/task/<tid>/comm. | |
3251 | * It bypasses generic permission checks in the case where a task of the same | |
3252 | * task group attempts to access the node. | |
3253 | * The rationale behind this is that glibc and bionic access this node for | |
3254 | * cross thread naming (pthread_set/getname_np(!self)). However, if | |
3255 | * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0, | |
3256 | * which locks out the cross thread naming implementation. | |
3257 | * This function makes sure that the node is always accessible for members of | |
3258 | * same thread group. | |
3259 | */ | |
3260 | static int proc_tid_comm_permission(struct inode *inode, int mask) | |
3261 | { | |
3262 | bool is_same_tgroup; | |
3263 | struct task_struct *task; | |
3264 | ||
3265 | task = get_proc_task(inode); | |
3266 | if (!task) | |
3267 | return -ESRCH; | |
3268 | is_same_tgroup = same_thread_group(current, task); | |
3269 | put_task_struct(task); | |
3270 | ||
3271 | if (likely(is_same_tgroup && !(mask & MAY_EXEC))) { | |
3272 | /* This file (/proc/<pid>/task/<tid>/comm) can always be | |
3273 | * read or written by the members of the corresponding | |
3274 | * thread group. | |
3275 | */ | |
3276 | return 0; | |
3277 | } | |
3278 | ||
3279 | return generic_permission(inode, mask); | |
3280 | } | |
3281 | ||
3282 | static const struct inode_operations proc_tid_comm_inode_operations = { | |
3283 | .permission = proc_tid_comm_permission, | |
3284 | }; | |
3285 | ||
28a6d671 EB |
3286 | /* |
3287 | * Tasks | |
3288 | */ | |
c5141e6d | 3289 | static const struct pid_entry tid_base_stuff[] = { |
631f9c18 | 3290 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), |
3835541d | 3291 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 3292 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
6ba8ed79 EB |
3293 | #ifdef CONFIG_NET |
3294 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), | |
3295 | #endif | |
631f9c18 | 3296 | REG("environ", S_IRUSR, proc_environ_operations), |
c5317167 | 3297 | REG("auxv", S_IRUSR, proc_auxv_operations), |
631f9c18 | 3298 | ONE("status", S_IRUGO, proc_pid_status), |
35a35046 | 3299 | ONE("personality", S_IRUSR, proc_pid_personality), |
1c963eb1 | 3300 | ONE("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 3301 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 3302 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
ebcb6734 | 3303 | #endif |
1b3044e3 JD |
3304 | NOD("comm", S_IFREG|S_IRUGO|S_IWUSR, |
3305 | &proc_tid_comm_inode_operations, | |
3306 | &proc_pid_set_comm_operations, {}), | |
ebcb6734 | 3307 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 | 3308 | ONE("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 3309 | #endif |
c2c0bb44 | 3310 | REG("cmdline", S_IRUGO, proc_pid_cmdline_ops), |
631f9c18 AD |
3311 | ONE("stat", S_IRUGO, proc_tid_stat), |
3312 | ONE("statm", S_IRUGO, proc_pid_statm), | |
b7643757 | 3313 | REG("maps", S_IRUGO, proc_tid_maps_operations), |
2e13ba54 | 3314 | #ifdef CONFIG_PROC_CHILDREN |
81841161 CG |
3315 | REG("children", S_IRUGO, proc_tid_children_operations), |
3316 | #endif | |
28a6d671 | 3317 | #ifdef CONFIG_NUMA |
b7643757 | 3318 | REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations), |
28a6d671 | 3319 | #endif |
631f9c18 AD |
3320 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
3321 | LNK("cwd", proc_cwd_link), | |
3322 | LNK("root", proc_root_link), | |
3323 | LNK("exe", proc_exe_link), | |
3324 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
3325 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
1e883281 | 3326 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 3327 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
b7643757 | 3328 | REG("smaps", S_IRUGO, proc_tid_smaps_operations), |
493b0e9d | 3329 | REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations), |
32ed74a4 | 3330 | REG("pagemap", S_IRUSR, proc_pagemap_operations), |
28a6d671 EB |
3331 | #endif |
3332 | #ifdef CONFIG_SECURITY | |
631f9c18 | 3333 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
3334 | #endif |
3335 | #ifdef CONFIG_KALLSYMS | |
edfcd606 | 3336 | ONE("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 3337 | #endif |
2ec220e2 | 3338 | #ifdef CONFIG_STACKTRACE |
35a35046 | 3339 | ONE("stack", S_IRUSR, proc_pid_stack), |
28a6d671 | 3340 | #endif |
5968cece | 3341 | #ifdef CONFIG_SCHED_INFO |
f6e826ca | 3342 | ONE("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 3343 | #endif |
9745512c | 3344 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 3345 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 3346 | #endif |
8793d854 | 3347 | #ifdef CONFIG_PROC_PID_CPUSET |
52de4779 | 3348 | ONE("cpuset", S_IRUGO, proc_cpuset_show), |
a424316c PM |
3349 | #endif |
3350 | #ifdef CONFIG_CGROUPS | |
006f4ac4 | 3351 | ONE("cgroup", S_IRUGO, proc_cgroup_show), |
28a6d671 | 3352 | #endif |
6ba51e37 | 3353 | ONE("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 3354 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 3355 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
28a6d671 | 3356 | #ifdef CONFIG_AUDITSYSCALL |
631f9c18 | 3357 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
26ec3c64 | 3358 | REG("sessionid", S_IRUGO, proc_sessionid_operations), |
28a6d671 | 3359 | #endif |
f4f154fd | 3360 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 3361 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
1203c8e6 | 3362 | REG("fail-nth", 0644, proc_fail_nth_operations), |
f4f154fd | 3363 | #endif |
297c5d92 | 3364 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 3365 | ONE("io", S_IRUSR, proc_tid_io_accounting), |
297c5d92 | 3366 | #endif |
f133ecca | 3367 | #ifdef CONFIG_HARDWALL |
d962c144 | 3368 | ONE("hardwall", S_IRUGO, proc_pid_hardwall), |
f133ecca | 3369 | #endif |
22d917d8 EB |
3370 | #ifdef CONFIG_USER_NS |
3371 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
3372 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 3373 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
9cc46516 | 3374 | REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations), |
22d917d8 | 3375 | #endif |
7c23b330 JP |
3376 | #ifdef CONFIG_LIVEPATCH |
3377 | ONE("patch_state", S_IRUSR, proc_pid_patch_state), | |
3378 | #endif | |
28a6d671 EB |
3379 | }; |
3380 | ||
f0c3b509 | 3381 | static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx) |
28a6d671 | 3382 | { |
f0c3b509 AV |
3383 | return proc_pident_readdir(file, ctx, |
3384 | tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
3385 | } |
3386 | ||
00cd8dd3 AV |
3387 | static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
3388 | { | |
7bcd6b0e EB |
3389 | return proc_pident_lookup(dir, dentry, |
3390 | tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
3391 | } |
3392 | ||
00977a59 | 3393 | static const struct file_operations proc_tid_base_operations = { |
28a6d671 | 3394 | .read = generic_read_dir, |
f50752ea AV |
3395 | .iterate_shared = proc_tid_base_readdir, |
3396 | .llseek = generic_file_llseek, | |
28a6d671 EB |
3397 | }; |
3398 | ||
c5ef1c42 | 3399 | static const struct inode_operations proc_tid_base_inode_operations = { |
28a6d671 EB |
3400 | .lookup = proc_tid_base_lookup, |
3401 | .getattr = pid_getattr, | |
3402 | .setattr = proc_setattr, | |
3403 | }; | |
3404 | ||
c52a47ac | 3405 | static int proc_task_instantiate(struct inode *dir, |
c5141e6d | 3406 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
444ceed8 | 3407 | { |
444ceed8 | 3408 | struct inode *inode; |
db978da8 | 3409 | inode = proc_pid_make_inode(dir->i_sb, task, S_IFDIR | S_IRUGO | S_IXUGO); |
444ceed8 EB |
3410 | |
3411 | if (!inode) | |
3412 | goto out; | |
444ceed8 EB |
3413 | inode->i_op = &proc_tid_base_inode_operations; |
3414 | inode->i_fop = &proc_tid_base_operations; | |
3415 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 | 3416 | |
1270dd8d | 3417 | set_nlink(inode, nlink_tid); |
444ceed8 | 3418 | |
fb045adb | 3419 | d_set_d_op(dentry, &pid_dentry_operations); |
444ceed8 EB |
3420 | |
3421 | d_add(dentry, inode); | |
3422 | /* Close the race of the process dying before we return the dentry */ | |
0b728e19 | 3423 | if (pid_revalidate(dentry, 0)) |
c52a47ac | 3424 | return 0; |
444ceed8 | 3425 | out: |
c52a47ac | 3426 | return -ENOENT; |
444ceed8 EB |
3427 | } |
3428 | ||
00cd8dd3 | 3429 | static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags) |
28a6d671 | 3430 | { |
c52a47ac | 3431 | int result = -ENOENT; |
28a6d671 EB |
3432 | struct task_struct *task; |
3433 | struct task_struct *leader = get_proc_task(dir); | |
28a6d671 | 3434 | unsigned tid; |
b488893a | 3435 | struct pid_namespace *ns; |
28a6d671 EB |
3436 | |
3437 | if (!leader) | |
3438 | goto out_no_task; | |
3439 | ||
dbcdb504 | 3440 | tid = name_to_int(&dentry->d_name); |
28a6d671 EB |
3441 | if (tid == ~0U) |
3442 | goto out; | |
3443 | ||
b488893a | 3444 | ns = dentry->d_sb->s_fs_info; |
28a6d671 | 3445 | rcu_read_lock(); |
b488893a | 3446 | task = find_task_by_pid_ns(tid, ns); |
28a6d671 EB |
3447 | if (task) |
3448 | get_task_struct(task); | |
3449 | rcu_read_unlock(); | |
3450 | if (!task) | |
3451 | goto out; | |
bac0abd6 | 3452 | if (!same_thread_group(leader, task)) |
28a6d671 EB |
3453 | goto out_drop_task; |
3454 | ||
444ceed8 | 3455 | result = proc_task_instantiate(dir, dentry, task, NULL); |
28a6d671 EB |
3456 | out_drop_task: |
3457 | put_task_struct(task); | |
3458 | out: | |
3459 | put_task_struct(leader); | |
3460 | out_no_task: | |
c52a47ac | 3461 | return ERR_PTR(result); |
28a6d671 EB |
3462 | } |
3463 | ||
0bc58a91 EB |
3464 | /* |
3465 | * Find the first tid of a thread group to return to user space. | |
3466 | * | |
3467 | * Usually this is just the thread group leader, but if the users | |
3468 | * buffer was too small or there was a seek into the middle of the | |
3469 | * directory we have more work todo. | |
3470 | * | |
3471 | * In the case of a short read we start with find_task_by_pid. | |
3472 | * | |
3473 | * In the case of a seek we start with the leader and walk nr | |
3474 | * threads past it. | |
3475 | */ | |
9f6e963f ON |
3476 | static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos, |
3477 | struct pid_namespace *ns) | |
0bc58a91 | 3478 | { |
d855a4b7 | 3479 | struct task_struct *pos, *task; |
9f6e963f ON |
3480 | unsigned long nr = f_pos; |
3481 | ||
3482 | if (nr != f_pos) /* 32bit overflow? */ | |
3483 | return NULL; | |
1da177e4 | 3484 | |
cc288738 | 3485 | rcu_read_lock(); |
d855a4b7 ON |
3486 | task = pid_task(pid, PIDTYPE_PID); |
3487 | if (!task) | |
3488 | goto fail; | |
3489 | ||
3490 | /* Attempt to start with the tid of a thread */ | |
9f6e963f | 3491 | if (tid && nr) { |
b488893a | 3492 | pos = find_task_by_pid_ns(tid, ns); |
d855a4b7 | 3493 | if (pos && same_thread_group(pos, task)) |
a872ff0c | 3494 | goto found; |
0bc58a91 | 3495 | } |
1da177e4 | 3496 | |
0bc58a91 | 3497 | /* If nr exceeds the number of threads there is nothing todo */ |
9f6e963f | 3498 | if (nr >= get_nr_threads(task)) |
c986c14a | 3499 | goto fail; |
1da177e4 | 3500 | |
a872ff0c ON |
3501 | /* If we haven't found our starting place yet start |
3502 | * with the leader and walk nr threads forward. | |
0bc58a91 | 3503 | */ |
d855a4b7 | 3504 | pos = task = task->group_leader; |
c986c14a | 3505 | do { |
9f6e963f | 3506 | if (!nr--) |
c986c14a | 3507 | goto found; |
d855a4b7 | 3508 | } while_each_thread(task, pos); |
c986c14a ON |
3509 | fail: |
3510 | pos = NULL; | |
3511 | goto out; | |
a872ff0c ON |
3512 | found: |
3513 | get_task_struct(pos); | |
3514 | out: | |
cc288738 | 3515 | rcu_read_unlock(); |
0bc58a91 EB |
3516 | return pos; |
3517 | } | |
3518 | ||
3519 | /* | |
3520 | * Find the next thread in the thread list. | |
3521 | * Return NULL if there is an error or no next thread. | |
3522 | * | |
3523 | * The reference to the input task_struct is released. | |
3524 | */ | |
3525 | static struct task_struct *next_tid(struct task_struct *start) | |
3526 | { | |
c1df7fb8 | 3527 | struct task_struct *pos = NULL; |
cc288738 | 3528 | rcu_read_lock(); |
c1df7fb8 | 3529 | if (pid_alive(start)) { |
0bc58a91 | 3530 | pos = next_thread(start); |
c1df7fb8 ON |
3531 | if (thread_group_leader(pos)) |
3532 | pos = NULL; | |
3533 | else | |
3534 | get_task_struct(pos); | |
3535 | } | |
cc288738 | 3536 | rcu_read_unlock(); |
0bc58a91 EB |
3537 | put_task_struct(start); |
3538 | return pos; | |
1da177e4 LT |
3539 | } |
3540 | ||
3541 | /* for the /proc/TGID/task/ directories */ | |
f0c3b509 | 3542 | static int proc_task_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3543 | { |
d855a4b7 ON |
3544 | struct inode *inode = file_inode(file); |
3545 | struct task_struct *task; | |
b488893a | 3546 | struct pid_namespace *ns; |
f0c3b509 | 3547 | int tid; |
1da177e4 | 3548 | |
d855a4b7 | 3549 | if (proc_inode_is_dead(inode)) |
f0c3b509 | 3550 | return -ENOENT; |
1da177e4 | 3551 | |
f0c3b509 | 3552 | if (!dir_emit_dots(file, ctx)) |
d855a4b7 | 3553 | return 0; |
1da177e4 | 3554 | |
0bc58a91 EB |
3555 | /* f_version caches the tgid value that the last readdir call couldn't |
3556 | * return. lseek aka telldir automagically resets f_version to 0. | |
3557 | */ | |
3aa3377f | 3558 | ns = inode->i_sb->s_fs_info; |
f0c3b509 AV |
3559 | tid = (int)file->f_version; |
3560 | file->f_version = 0; | |
d855a4b7 | 3561 | for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns); |
0bc58a91 | 3562 | task; |
f0c3b509 AV |
3563 | task = next_tid(task), ctx->pos++) { |
3564 | char name[PROC_NUMBUF]; | |
3565 | int len; | |
b488893a | 3566 | tid = task_pid_nr_ns(task, ns); |
f0c3b509 AV |
3567 | len = snprintf(name, sizeof(name), "%d", tid); |
3568 | if (!proc_fill_cache(file, ctx, name, len, | |
3569 | proc_task_instantiate, task, NULL)) { | |
0bc58a91 EB |
3570 | /* returning this tgid failed, save it as the first |
3571 | * pid for the next readir call */ | |
f0c3b509 | 3572 | file->f_version = (u64)tid; |
0bc58a91 | 3573 | put_task_struct(task); |
1da177e4 | 3574 | break; |
0bc58a91 | 3575 | } |
1da177e4 | 3576 | } |
d855a4b7 | 3577 | |
f0c3b509 | 3578 | return 0; |
1da177e4 | 3579 | } |
6e66b52b | 3580 | |
a528d35e DH |
3581 | static int proc_task_getattr(const struct path *path, struct kstat *stat, |
3582 | u32 request_mask, unsigned int query_flags) | |
6e66b52b | 3583 | { |
a528d35e | 3584 | struct inode *inode = d_inode(path->dentry); |
99f89551 | 3585 | struct task_struct *p = get_proc_task(inode); |
6e66b52b EB |
3586 | generic_fillattr(inode, stat); |
3587 | ||
99f89551 | 3588 | if (p) { |
99f89551 | 3589 | stat->nlink += get_nr_threads(p); |
99f89551 | 3590 | put_task_struct(p); |
6e66b52b EB |
3591 | } |
3592 | ||
3593 | return 0; | |
3594 | } | |
28a6d671 | 3595 | |
c5ef1c42 | 3596 | static const struct inode_operations proc_task_inode_operations = { |
28a6d671 EB |
3597 | .lookup = proc_task_lookup, |
3598 | .getattr = proc_task_getattr, | |
3599 | .setattr = proc_setattr, | |
0499680a | 3600 | .permission = proc_pid_permission, |
28a6d671 EB |
3601 | }; |
3602 | ||
00977a59 | 3603 | static const struct file_operations proc_task_operations = { |
28a6d671 | 3604 | .read = generic_read_dir, |
f50752ea AV |
3605 | .iterate_shared = proc_task_readdir, |
3606 | .llseek = generic_file_llseek, | |
28a6d671 | 3607 | }; |
1270dd8d AD |
3608 | |
3609 | void __init set_proc_pid_nlink(void) | |
3610 | { | |
3611 | nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
3612 | nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
3613 | } |