Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/proc/base.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | * | |
6 | * proc base directory handling functions | |
7 | * | |
8 | * 1999, Al Viro. Rewritten. Now it covers the whole per-process part. | |
9 | * Instead of using magical inumbers to determine the kind of object | |
10 | * we allocate and fill in-core inodes upon lookup. They don't even | |
11 | * go into icache. We cache the reference to task_struct upon lookup too. | |
12 | * Eventually it should become a filesystem in its own. We don't use the | |
13 | * rest of procfs anymore. | |
e070ad49 ML |
14 | * |
15 | * | |
16 | * Changelog: | |
17 | * 17-Jan-2005 | |
18 | * Allan Bezerra | |
19 | * Bruna Moreira <bruna.moreira@indt.org.br> | |
20 | * Edjard Mota <edjard.mota@indt.org.br> | |
21 | * Ilias Biris <ilias.biris@indt.org.br> | |
22 | * Mauricio Lin <mauricio.lin@indt.org.br> | |
23 | * | |
24 | * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT | |
25 | * | |
26 | * A new process specific entry (smaps) included in /proc. It shows the | |
27 | * size of rss for each memory area. The maps entry lacks information | |
28 | * about physical memory size (rss) for each mapped file, i.e., | |
29 | * rss information for executables and library files. | |
30 | * This additional information is useful for any tools that need to know | |
31 | * about physical memory consumption for a process specific library. | |
32 | * | |
33 | * Changelog: | |
34 | * 21-Feb-2005 | |
35 | * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT | |
36 | * Pud inclusion in the page table walking. | |
37 | * | |
38 | * ChangeLog: | |
39 | * 10-Mar-2005 | |
40 | * 10LE Instituto Nokia de Tecnologia - INdT: | |
41 | * A better way to walks through the page table as suggested by Hugh Dickins. | |
42 | * | |
43 | * Simo Piiroinen <simo.piiroinen@nokia.com>: | |
44 | * Smaps information related to shared, private, clean and dirty pages. | |
45 | * | |
46 | * Paul Mundt <paul.mundt@nokia.com>: | |
47 | * Overall revision about smaps. | |
1da177e4 LT |
48 | */ |
49 | ||
50 | #include <asm/uaccess.h> | |
51 | ||
1da177e4 LT |
52 | #include <linux/errno.h> |
53 | #include <linux/time.h> | |
54 | #include <linux/proc_fs.h> | |
55 | #include <linux/stat.h> | |
5995477a | 56 | #include <linux/task_io_accounting_ops.h> |
1da177e4 | 57 | #include <linux/init.h> |
16f7e0fe | 58 | #include <linux/capability.h> |
1da177e4 | 59 | #include <linux/file.h> |
9f3acc31 | 60 | #include <linux/fdtable.h> |
1da177e4 LT |
61 | #include <linux/string.h> |
62 | #include <linux/seq_file.h> | |
63 | #include <linux/namei.h> | |
6b3286ed | 64 | #include <linux/mnt_namespace.h> |
1da177e4 | 65 | #include <linux/mm.h> |
a63d83f4 | 66 | #include <linux/swap.h> |
b835996f | 67 | #include <linux/rcupdate.h> |
1da177e4 | 68 | #include <linux/kallsyms.h> |
2ec220e2 | 69 | #include <linux/stacktrace.h> |
d85f50d5 | 70 | #include <linux/resource.h> |
5096add8 | 71 | #include <linux/module.h> |
1da177e4 LT |
72 | #include <linux/mount.h> |
73 | #include <linux/security.h> | |
74 | #include <linux/ptrace.h> | |
0d094efe | 75 | #include <linux/tracehook.h> |
87ebdc00 | 76 | #include <linux/printk.h> |
a424316c | 77 | #include <linux/cgroup.h> |
1da177e4 LT |
78 | #include <linux/cpuset.h> |
79 | #include <linux/audit.h> | |
5addc5dd | 80 | #include <linux/poll.h> |
1651e14e | 81 | #include <linux/nsproxy.h> |
8ac773b4 | 82 | #include <linux/oom.h> |
3cb4a0bb | 83 | #include <linux/elf.h> |
60347f67 | 84 | #include <linux/pid_namespace.h> |
22d917d8 | 85 | #include <linux/user_namespace.h> |
5ad4e53b | 86 | #include <linux/fs_struct.h> |
5a0e3ad6 | 87 | #include <linux/slab.h> |
640708a2 | 88 | #include <linux/flex_array.h> |
48f6a7a5 | 89 | #include <linux/posix-timers.h> |
f133ecca CM |
90 | #ifdef CONFIG_HARDWALL |
91 | #include <asm/hardwall.h> | |
92 | #endif | |
43d2b113 | 93 | #include <trace/events/oom.h> |
1da177e4 | 94 | #include "internal.h" |
faf60af1 | 95 | #include "fd.h" |
1da177e4 | 96 | |
0f2fe20f EB |
97 | /* NOTE: |
98 | * Implementing inode permission operations in /proc is almost | |
99 | * certainly an error. Permission checks need to happen during | |
100 | * each system call not at open time. The reason is that most of | |
101 | * what we wish to check for permissions in /proc varies at runtime. | |
102 | * | |
103 | * The classic example of a problem is opening file descriptors | |
104 | * in /proc for a task before it execs a suid executable. | |
105 | */ | |
106 | ||
1da177e4 | 107 | struct pid_entry { |
cedbccab | 108 | const char *name; |
c5141e6d | 109 | int len; |
d161a13f | 110 | umode_t mode; |
c5ef1c42 | 111 | const struct inode_operations *iop; |
00977a59 | 112 | const struct file_operations *fop; |
20cdc894 | 113 | union proc_op op; |
1da177e4 LT |
114 | }; |
115 | ||
61a28784 | 116 | #define NOD(NAME, MODE, IOP, FOP, OP) { \ |
20cdc894 | 117 | .name = (NAME), \ |
c5141e6d | 118 | .len = sizeof(NAME) - 1, \ |
20cdc894 EB |
119 | .mode = MODE, \ |
120 | .iop = IOP, \ | |
121 | .fop = FOP, \ | |
122 | .op = OP, \ | |
123 | } | |
124 | ||
631f9c18 AD |
125 | #define DIR(NAME, MODE, iops, fops) \ |
126 | NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} ) | |
127 | #define LNK(NAME, get_link) \ | |
61a28784 | 128 | NOD(NAME, (S_IFLNK|S_IRWXUGO), \ |
20cdc894 | 129 | &proc_pid_link_inode_operations, NULL, \ |
631f9c18 AD |
130 | { .proc_get_link = get_link } ) |
131 | #define REG(NAME, MODE, fops) \ | |
132 | NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {}) | |
631f9c18 | 133 | #define ONE(NAME, MODE, show) \ |
be614086 EB |
134 | NOD(NAME, (S_IFREG|(MODE)), \ |
135 | NULL, &proc_single_file_operations, \ | |
631f9c18 | 136 | { .proc_show = show } ) |
1da177e4 | 137 | |
aed54175 VN |
138 | /* |
139 | * Count the number of hardlinks for the pid_entry table, excluding the . | |
140 | * and .. links. | |
141 | */ | |
142 | static unsigned int pid_entry_count_dirs(const struct pid_entry *entries, | |
143 | unsigned int n) | |
144 | { | |
145 | unsigned int i; | |
146 | unsigned int count; | |
147 | ||
148 | count = 0; | |
149 | for (i = 0; i < n; ++i) { | |
150 | if (S_ISDIR(entries[i].mode)) | |
151 | ++count; | |
152 | } | |
153 | ||
154 | return count; | |
155 | } | |
156 | ||
f7ad3c6b | 157 | static int get_task_root(struct task_struct *task, struct path *root) |
1da177e4 | 158 | { |
7c2c7d99 HD |
159 | int result = -ENOENT; |
160 | ||
0494f6ec | 161 | task_lock(task); |
f7ad3c6b MS |
162 | if (task->fs) { |
163 | get_fs_root(task->fs, root); | |
7c2c7d99 HD |
164 | result = 0; |
165 | } | |
0494f6ec | 166 | task_unlock(task); |
7c2c7d99 | 167 | return result; |
0494f6ec MS |
168 | } |
169 | ||
7773fbc5 | 170 | static int proc_cwd_link(struct dentry *dentry, struct path *path) |
0494f6ec | 171 | { |
2b0143b5 | 172 | struct task_struct *task = get_proc_task(d_inode(dentry)); |
0494f6ec | 173 | int result = -ENOENT; |
99f89551 EB |
174 | |
175 | if (task) { | |
f7ad3c6b MS |
176 | task_lock(task); |
177 | if (task->fs) { | |
178 | get_fs_pwd(task->fs, path); | |
179 | result = 0; | |
180 | } | |
181 | task_unlock(task); | |
99f89551 EB |
182 | put_task_struct(task); |
183 | } | |
1da177e4 LT |
184 | return result; |
185 | } | |
186 | ||
7773fbc5 | 187 | static int proc_root_link(struct dentry *dentry, struct path *path) |
1da177e4 | 188 | { |
2b0143b5 | 189 | struct task_struct *task = get_proc_task(d_inode(dentry)); |
1da177e4 | 190 | int result = -ENOENT; |
99f89551 EB |
191 | |
192 | if (task) { | |
f7ad3c6b | 193 | result = get_task_root(task, path); |
99f89551 EB |
194 | put_task_struct(task); |
195 | } | |
1da177e4 LT |
196 | return result; |
197 | } | |
198 | ||
c2c0bb44 AD |
199 | static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, |
200 | size_t _count, loff_t *pos) | |
1da177e4 | 201 | { |
c2c0bb44 AD |
202 | struct task_struct *tsk; |
203 | struct mm_struct *mm; | |
204 | char *page; | |
205 | unsigned long count = _count; | |
206 | unsigned long arg_start, arg_end, env_start, env_end; | |
207 | unsigned long len1, len2, len; | |
208 | unsigned long p; | |
209 | char c; | |
210 | ssize_t rv; | |
211 | ||
212 | BUG_ON(*pos < 0); | |
213 | ||
214 | tsk = get_proc_task(file_inode(file)); | |
215 | if (!tsk) | |
216 | return -ESRCH; | |
217 | mm = get_task_mm(tsk); | |
218 | put_task_struct(tsk); | |
219 | if (!mm) | |
220 | return 0; | |
221 | /* Check if process spawned far enough to have cmdline. */ | |
222 | if (!mm->env_end) { | |
223 | rv = 0; | |
224 | goto out_mmput; | |
225 | } | |
226 | ||
227 | page = (char *)__get_free_page(GFP_TEMPORARY); | |
228 | if (!page) { | |
229 | rv = -ENOMEM; | |
230 | goto out_mmput; | |
231 | } | |
232 | ||
233 | down_read(&mm->mmap_sem); | |
234 | arg_start = mm->arg_start; | |
235 | arg_end = mm->arg_end; | |
236 | env_start = mm->env_start; | |
237 | env_end = mm->env_end; | |
238 | up_read(&mm->mmap_sem); | |
239 | ||
240 | BUG_ON(arg_start > arg_end); | |
241 | BUG_ON(env_start > env_end); | |
242 | ||
243 | len1 = arg_end - arg_start; | |
244 | len2 = env_end - env_start; | |
245 | ||
3581d458 AD |
246 | /* Empty ARGV. */ |
247 | if (len1 == 0) { | |
248 | rv = 0; | |
249 | goto out_free_page; | |
250 | } | |
2ca66ff7 | 251 | /* |
c2c0bb44 AD |
252 | * Inherently racy -- command line shares address space |
253 | * with code and data. | |
2ca66ff7 | 254 | */ |
c2c0bb44 AD |
255 | rv = access_remote_vm(mm, arg_end - 1, &c, 1, 0); |
256 | if (rv <= 0) | |
257 | goto out_free_page; | |
258 | ||
259 | rv = 0; | |
260 | ||
261 | if (c == '\0') { | |
262 | /* Command line (set of strings) occupies whole ARGV. */ | |
263 | if (len1 <= *pos) | |
264 | goto out_free_page; | |
265 | ||
266 | p = arg_start + *pos; | |
267 | len = len1 - *pos; | |
268 | while (count > 0 && len > 0) { | |
269 | unsigned int _count; | |
270 | int nr_read; | |
271 | ||
272 | _count = min3(count, len, PAGE_SIZE); | |
273 | nr_read = access_remote_vm(mm, p, page, _count, 0); | |
274 | if (nr_read < 0) | |
275 | rv = nr_read; | |
276 | if (nr_read <= 0) | |
277 | goto out_free_page; | |
278 | ||
279 | if (copy_to_user(buf, page, nr_read)) { | |
280 | rv = -EFAULT; | |
281 | goto out_free_page; | |
282 | } | |
283 | ||
284 | p += nr_read; | |
285 | len -= nr_read; | |
286 | buf += nr_read; | |
287 | count -= nr_read; | |
288 | rv += nr_read; | |
289 | } | |
290 | } else { | |
291 | /* | |
292 | * Command line (1 string) occupies ARGV and maybe | |
293 | * extends into ENVP. | |
294 | */ | |
295 | if (len1 + len2 <= *pos) | |
296 | goto skip_argv_envp; | |
297 | if (len1 <= *pos) | |
298 | goto skip_argv; | |
299 | ||
300 | p = arg_start + *pos; | |
301 | len = len1 - *pos; | |
302 | while (count > 0 && len > 0) { | |
303 | unsigned int _count, l; | |
304 | int nr_read; | |
305 | bool final; | |
306 | ||
307 | _count = min3(count, len, PAGE_SIZE); | |
308 | nr_read = access_remote_vm(mm, p, page, _count, 0); | |
309 | if (nr_read < 0) | |
310 | rv = nr_read; | |
311 | if (nr_read <= 0) | |
312 | goto out_free_page; | |
313 | ||
314 | /* | |
315 | * Command line can be shorter than whole ARGV | |
316 | * even if last "marker" byte says it is not. | |
317 | */ | |
318 | final = false; | |
319 | l = strnlen(page, nr_read); | |
320 | if (l < nr_read) { | |
321 | nr_read = l; | |
322 | final = true; | |
323 | } | |
324 | ||
325 | if (copy_to_user(buf, page, nr_read)) { | |
326 | rv = -EFAULT; | |
327 | goto out_free_page; | |
328 | } | |
329 | ||
330 | p += nr_read; | |
331 | len -= nr_read; | |
332 | buf += nr_read; | |
333 | count -= nr_read; | |
334 | rv += nr_read; | |
335 | ||
336 | if (final) | |
337 | goto out_free_page; | |
338 | } | |
339 | skip_argv: | |
340 | /* | |
341 | * Command line (1 string) occupies ARGV and | |
342 | * extends into ENVP. | |
343 | */ | |
344 | if (len1 <= *pos) { | |
345 | p = env_start + *pos - len1; | |
346 | len = len1 + len2 - *pos; | |
347 | } else { | |
348 | p = env_start; | |
349 | len = len2; | |
350 | } | |
351 | while (count > 0 && len > 0) { | |
352 | unsigned int _count, l; | |
353 | int nr_read; | |
354 | bool final; | |
355 | ||
356 | _count = min3(count, len, PAGE_SIZE); | |
357 | nr_read = access_remote_vm(mm, p, page, _count, 0); | |
358 | if (nr_read < 0) | |
359 | rv = nr_read; | |
360 | if (nr_read <= 0) | |
361 | goto out_free_page; | |
362 | ||
363 | /* Find EOS. */ | |
364 | final = false; | |
365 | l = strnlen(page, nr_read); | |
366 | if (l < nr_read) { | |
367 | nr_read = l; | |
368 | final = true; | |
369 | } | |
370 | ||
371 | if (copy_to_user(buf, page, nr_read)) { | |
372 | rv = -EFAULT; | |
373 | goto out_free_page; | |
374 | } | |
375 | ||
376 | p += nr_read; | |
377 | len -= nr_read; | |
378 | buf += nr_read; | |
379 | count -= nr_read; | |
380 | rv += nr_read; | |
381 | ||
382 | if (final) | |
383 | goto out_free_page; | |
384 | } | |
385 | skip_argv_envp: | |
386 | ; | |
387 | } | |
388 | ||
389 | out_free_page: | |
390 | free_page((unsigned long)page); | |
391 | out_mmput: | |
392 | mmput(mm); | |
393 | if (rv > 0) | |
394 | *pos += rv; | |
395 | return rv; | |
1da177e4 LT |
396 | } |
397 | ||
c2c0bb44 AD |
398 | static const struct file_operations proc_pid_cmdline_ops = { |
399 | .read = proc_pid_cmdline_read, | |
400 | .llseek = generic_file_llseek, | |
401 | }; | |
402 | ||
f9ea536e AD |
403 | static int proc_pid_auxv(struct seq_file *m, struct pid_namespace *ns, |
404 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 405 | { |
caaee623 | 406 | struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ_FSCREDS); |
2fadaef4 | 407 | if (mm && !IS_ERR(mm)) { |
1da177e4 | 408 | unsigned int nwords = 0; |
dfe6b7d9 | 409 | do { |
1da177e4 | 410 | nwords += 2; |
dfe6b7d9 | 411 | } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */ |
f9ea536e | 412 | seq_write(m, mm->saved_auxv, nwords * sizeof(mm->saved_auxv[0])); |
1da177e4 | 413 | mmput(mm); |
f9ea536e AD |
414 | return 0; |
415 | } else | |
416 | return PTR_ERR(mm); | |
1da177e4 LT |
417 | } |
418 | ||
419 | ||
420 | #ifdef CONFIG_KALLSYMS | |
421 | /* | |
422 | * Provides a wchan file via kallsyms in a proper one-value-per-file format. | |
423 | * Returns the resolved symbol. If that fails, simply return the address. | |
424 | */ | |
edfcd606 AD |
425 | static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns, |
426 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 427 | { |
ffb45122 | 428 | unsigned long wchan; |
9281acea | 429 | char symname[KSYM_NAME_LEN]; |
1da177e4 LT |
430 | |
431 | wchan = get_wchan(task); | |
432 | ||
caaee623 JH |
433 | if (wchan && ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS) |
434 | && !lookup_symbol_name(wchan, symname)) | |
25ce3191 | 435 | seq_printf(m, "%s", symname); |
b2f73922 | 436 | else |
1e92a61c | 437 | seq_putc(m, '0'); |
25ce3191 JP |
438 | |
439 | return 0; | |
1da177e4 LT |
440 | } |
441 | #endif /* CONFIG_KALLSYMS */ | |
442 | ||
a9712bc1 AV |
443 | static int lock_trace(struct task_struct *task) |
444 | { | |
445 | int err = mutex_lock_killable(&task->signal->cred_guard_mutex); | |
446 | if (err) | |
447 | return err; | |
caaee623 | 448 | if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) { |
a9712bc1 AV |
449 | mutex_unlock(&task->signal->cred_guard_mutex); |
450 | return -EPERM; | |
451 | } | |
452 | return 0; | |
453 | } | |
454 | ||
455 | static void unlock_trace(struct task_struct *task) | |
456 | { | |
457 | mutex_unlock(&task->signal->cred_guard_mutex); | |
458 | } | |
459 | ||
2ec220e2 KC |
460 | #ifdef CONFIG_STACKTRACE |
461 | ||
462 | #define MAX_STACK_TRACE_DEPTH 64 | |
463 | ||
464 | static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns, | |
465 | struct pid *pid, struct task_struct *task) | |
466 | { | |
467 | struct stack_trace trace; | |
468 | unsigned long *entries; | |
a9712bc1 | 469 | int err; |
2ec220e2 KC |
470 | int i; |
471 | ||
472 | entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL); | |
473 | if (!entries) | |
474 | return -ENOMEM; | |
475 | ||
476 | trace.nr_entries = 0; | |
477 | trace.max_entries = MAX_STACK_TRACE_DEPTH; | |
478 | trace.entries = entries; | |
479 | trace.skip = 0; | |
2ec220e2 | 480 | |
a9712bc1 AV |
481 | err = lock_trace(task); |
482 | if (!err) { | |
483 | save_stack_trace_tsk(task, &trace); | |
484 | ||
485 | for (i = 0; i < trace.nr_entries; i++) { | |
b81a618d | 486 | seq_printf(m, "[<%pK>] %pS\n", |
a9712bc1 AV |
487 | (void *)entries[i], (void *)entries[i]); |
488 | } | |
489 | unlock_trace(task); | |
2ec220e2 KC |
490 | } |
491 | kfree(entries); | |
492 | ||
a9712bc1 | 493 | return err; |
2ec220e2 KC |
494 | } |
495 | #endif | |
496 | ||
5968cece | 497 | #ifdef CONFIG_SCHED_INFO |
1da177e4 LT |
498 | /* |
499 | * Provides /proc/PID/schedstat | |
500 | */ | |
f6e826ca AD |
501 | static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns, |
502 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 503 | { |
5968cece NR |
504 | if (unlikely(!sched_info_on())) |
505 | seq_printf(m, "0 0 0\n"); | |
506 | else | |
507 | seq_printf(m, "%llu %llu %lu\n", | |
25ce3191 JP |
508 | (unsigned long long)task->se.sum_exec_runtime, |
509 | (unsigned long long)task->sched_info.run_delay, | |
510 | task->sched_info.pcount); | |
511 | ||
512 | return 0; | |
1da177e4 LT |
513 | } |
514 | #endif | |
515 | ||
9745512c AV |
516 | #ifdef CONFIG_LATENCYTOP |
517 | static int lstats_show_proc(struct seq_file *m, void *v) | |
518 | { | |
519 | int i; | |
13d77c37 HS |
520 | struct inode *inode = m->private; |
521 | struct task_struct *task = get_proc_task(inode); | |
9745512c | 522 | |
13d77c37 HS |
523 | if (!task) |
524 | return -ESRCH; | |
525 | seq_puts(m, "Latency Top version : v0.1\n"); | |
9745512c | 526 | for (i = 0; i < 32; i++) { |
34e49d4f JP |
527 | struct latency_record *lr = &task->latency_record[i]; |
528 | if (lr->backtrace[0]) { | |
9745512c | 529 | int q; |
34e49d4f JP |
530 | seq_printf(m, "%i %li %li", |
531 | lr->count, lr->time, lr->max); | |
9745512c | 532 | for (q = 0; q < LT_BACKTRACEDEPTH; q++) { |
34e49d4f JP |
533 | unsigned long bt = lr->backtrace[q]; |
534 | if (!bt) | |
9745512c | 535 | break; |
34e49d4f | 536 | if (bt == ULONG_MAX) |
9745512c | 537 | break; |
34e49d4f | 538 | seq_printf(m, " %ps", (void *)bt); |
9745512c | 539 | } |
9d6de12f | 540 | seq_putc(m, '\n'); |
9745512c AV |
541 | } |
542 | ||
543 | } | |
13d77c37 | 544 | put_task_struct(task); |
9745512c AV |
545 | return 0; |
546 | } | |
547 | ||
548 | static int lstats_open(struct inode *inode, struct file *file) | |
549 | { | |
13d77c37 | 550 | return single_open(file, lstats_show_proc, inode); |
d6643d12 HS |
551 | } |
552 | ||
9745512c AV |
553 | static ssize_t lstats_write(struct file *file, const char __user *buf, |
554 | size_t count, loff_t *offs) | |
555 | { | |
496ad9aa | 556 | struct task_struct *task = get_proc_task(file_inode(file)); |
9745512c | 557 | |
13d77c37 HS |
558 | if (!task) |
559 | return -ESRCH; | |
9745512c | 560 | clear_all_latency_tracing(task); |
13d77c37 | 561 | put_task_struct(task); |
9745512c AV |
562 | |
563 | return count; | |
564 | } | |
565 | ||
566 | static const struct file_operations proc_lstats_operations = { | |
567 | .open = lstats_open, | |
568 | .read = seq_read, | |
569 | .write = lstats_write, | |
570 | .llseek = seq_lseek, | |
13d77c37 | 571 | .release = single_release, |
9745512c AV |
572 | }; |
573 | ||
574 | #endif | |
575 | ||
6ba51e37 AD |
576 | static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns, |
577 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 578 | { |
a7f638f9 | 579 | unsigned long totalpages = totalram_pages + total_swap_pages; |
b95c35e7 | 580 | unsigned long points = 0; |
1da177e4 | 581 | |
19c5d45a | 582 | read_lock(&tasklist_lock); |
b95c35e7 | 583 | if (pid_alive(task)) |
a7f638f9 DR |
584 | points = oom_badness(task, NULL, NULL, totalpages) * |
585 | 1000 / totalpages; | |
19c5d45a | 586 | read_unlock(&tasklist_lock); |
25ce3191 JP |
587 | seq_printf(m, "%lu\n", points); |
588 | ||
589 | return 0; | |
1da177e4 LT |
590 | } |
591 | ||
d85f50d5 | 592 | struct limit_names { |
cedbccab AD |
593 | const char *name; |
594 | const char *unit; | |
d85f50d5 NH |
595 | }; |
596 | ||
597 | static const struct limit_names lnames[RLIM_NLIMITS] = { | |
cff4edb5 | 598 | [RLIMIT_CPU] = {"Max cpu time", "seconds"}, |
d85f50d5 NH |
599 | [RLIMIT_FSIZE] = {"Max file size", "bytes"}, |
600 | [RLIMIT_DATA] = {"Max data size", "bytes"}, | |
601 | [RLIMIT_STACK] = {"Max stack size", "bytes"}, | |
602 | [RLIMIT_CORE] = {"Max core file size", "bytes"}, | |
603 | [RLIMIT_RSS] = {"Max resident set", "bytes"}, | |
604 | [RLIMIT_NPROC] = {"Max processes", "processes"}, | |
605 | [RLIMIT_NOFILE] = {"Max open files", "files"}, | |
606 | [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"}, | |
607 | [RLIMIT_AS] = {"Max address space", "bytes"}, | |
608 | [RLIMIT_LOCKS] = {"Max file locks", "locks"}, | |
609 | [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"}, | |
610 | [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"}, | |
611 | [RLIMIT_NICE] = {"Max nice priority", NULL}, | |
612 | [RLIMIT_RTPRIO] = {"Max realtime priority", NULL}, | |
8808117c | 613 | [RLIMIT_RTTIME] = {"Max realtime timeout", "us"}, |
d85f50d5 NH |
614 | }; |
615 | ||
616 | /* Display limits for a process */ | |
1c963eb1 AD |
617 | static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns, |
618 | struct pid *pid, struct task_struct *task) | |
d85f50d5 NH |
619 | { |
620 | unsigned int i; | |
d85f50d5 | 621 | unsigned long flags; |
d85f50d5 NH |
622 | |
623 | struct rlimit rlim[RLIM_NLIMITS]; | |
624 | ||
a6bebbc8 | 625 | if (!lock_task_sighand(task, &flags)) |
d85f50d5 | 626 | return 0; |
d85f50d5 NH |
627 | memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS); |
628 | unlock_task_sighand(task, &flags); | |
d85f50d5 NH |
629 | |
630 | /* | |
631 | * print the file header | |
632 | */ | |
1c963eb1 | 633 | seq_printf(m, "%-25s %-20s %-20s %-10s\n", |
25ce3191 | 634 | "Limit", "Soft Limit", "Hard Limit", "Units"); |
d85f50d5 NH |
635 | |
636 | for (i = 0; i < RLIM_NLIMITS; i++) { | |
637 | if (rlim[i].rlim_cur == RLIM_INFINITY) | |
1c963eb1 | 638 | seq_printf(m, "%-25s %-20s ", |
25ce3191 | 639 | lnames[i].name, "unlimited"); |
d85f50d5 | 640 | else |
1c963eb1 | 641 | seq_printf(m, "%-25s %-20lu ", |
25ce3191 | 642 | lnames[i].name, rlim[i].rlim_cur); |
d85f50d5 NH |
643 | |
644 | if (rlim[i].rlim_max == RLIM_INFINITY) | |
1c963eb1 | 645 | seq_printf(m, "%-20s ", "unlimited"); |
d85f50d5 | 646 | else |
1c963eb1 | 647 | seq_printf(m, "%-20lu ", rlim[i].rlim_max); |
d85f50d5 NH |
648 | |
649 | if (lnames[i].unit) | |
1c963eb1 | 650 | seq_printf(m, "%-10s\n", lnames[i].unit); |
d85f50d5 | 651 | else |
1c963eb1 | 652 | seq_putc(m, '\n'); |
d85f50d5 NH |
653 | } |
654 | ||
1c963eb1 | 655 | return 0; |
d85f50d5 NH |
656 | } |
657 | ||
ebcb6734 | 658 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 AD |
659 | static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns, |
660 | struct pid *pid, struct task_struct *task) | |
ebcb6734 RM |
661 | { |
662 | long nr; | |
663 | unsigned long args[6], sp, pc; | |
25ce3191 JP |
664 | int res; |
665 | ||
666 | res = lock_trace(task); | |
a9712bc1 AV |
667 | if (res) |
668 | return res; | |
ebcb6734 RM |
669 | |
670 | if (task_current_syscall(task, &nr, args, 6, &sp, &pc)) | |
09d93bd6 | 671 | seq_puts(m, "running\n"); |
a9712bc1 | 672 | else if (nr < 0) |
09d93bd6 | 673 | seq_printf(m, "%ld 0x%lx 0x%lx\n", nr, sp, pc); |
a9712bc1 | 674 | else |
09d93bd6 | 675 | seq_printf(m, |
ebcb6734 RM |
676 | "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n", |
677 | nr, | |
678 | args[0], args[1], args[2], args[3], args[4], args[5], | |
679 | sp, pc); | |
a9712bc1 | 680 | unlock_trace(task); |
25ce3191 JP |
681 | |
682 | return 0; | |
ebcb6734 RM |
683 | } |
684 | #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */ | |
685 | ||
1da177e4 LT |
686 | /************************************************************************/ |
687 | /* Here the fs part begins */ | |
688 | /************************************************************************/ | |
689 | ||
690 | /* permission checks */ | |
778c1144 | 691 | static int proc_fd_access_allowed(struct inode *inode) |
1da177e4 | 692 | { |
778c1144 EB |
693 | struct task_struct *task; |
694 | int allowed = 0; | |
df26c40e EB |
695 | /* Allow access to a task's file descriptors if it is us or we |
696 | * may use ptrace attach to the process and find out that | |
697 | * information. | |
778c1144 EB |
698 | */ |
699 | task = get_proc_task(inode); | |
df26c40e | 700 | if (task) { |
caaee623 | 701 | allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS); |
778c1144 | 702 | put_task_struct(task); |
df26c40e | 703 | } |
778c1144 | 704 | return allowed; |
1da177e4 LT |
705 | } |
706 | ||
6b4e306a | 707 | int proc_setattr(struct dentry *dentry, struct iattr *attr) |
6d76fa58 LT |
708 | { |
709 | int error; | |
2b0143b5 | 710 | struct inode *inode = d_inode(dentry); |
6d76fa58 LT |
711 | |
712 | if (attr->ia_valid & ATTR_MODE) | |
713 | return -EPERM; | |
714 | ||
715 | error = inode_change_ok(inode, attr); | |
1025774c CH |
716 | if (error) |
717 | return error; | |
718 | ||
1025774c CH |
719 | setattr_copy(inode, attr); |
720 | mark_inode_dirty(inode); | |
721 | return 0; | |
6d76fa58 LT |
722 | } |
723 | ||
0499680a VK |
724 | /* |
725 | * May current process learn task's sched/cmdline info (for hide_pid_min=1) | |
726 | * or euid/egid (for hide_pid_min=2)? | |
727 | */ | |
728 | static bool has_pid_permissions(struct pid_namespace *pid, | |
729 | struct task_struct *task, | |
730 | int hide_pid_min) | |
731 | { | |
732 | if (pid->hide_pid < hide_pid_min) | |
733 | return true; | |
734 | if (in_group_p(pid->pid_gid)) | |
735 | return true; | |
caaee623 | 736 | return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS); |
0499680a VK |
737 | } |
738 | ||
739 | ||
740 | static int proc_pid_permission(struct inode *inode, int mask) | |
741 | { | |
742 | struct pid_namespace *pid = inode->i_sb->s_fs_info; | |
743 | struct task_struct *task; | |
744 | bool has_perms; | |
745 | ||
746 | task = get_proc_task(inode); | |
a2ef990a XF |
747 | if (!task) |
748 | return -ESRCH; | |
0499680a VK |
749 | has_perms = has_pid_permissions(pid, task, 1); |
750 | put_task_struct(task); | |
751 | ||
752 | if (!has_perms) { | |
753 | if (pid->hide_pid == 2) { | |
754 | /* | |
755 | * Let's make getdents(), stat(), and open() | |
756 | * consistent with each other. If a process | |
757 | * may not stat() a file, it shouldn't be seen | |
758 | * in procfs at all. | |
759 | */ | |
760 | return -ENOENT; | |
761 | } | |
762 | ||
763 | return -EPERM; | |
764 | } | |
765 | return generic_permission(inode, mask); | |
766 | } | |
767 | ||
768 | ||
769 | ||
c5ef1c42 | 770 | static const struct inode_operations proc_def_inode_operations = { |
6d76fa58 LT |
771 | .setattr = proc_setattr, |
772 | }; | |
773 | ||
be614086 EB |
774 | static int proc_single_show(struct seq_file *m, void *v) |
775 | { | |
776 | struct inode *inode = m->private; | |
777 | struct pid_namespace *ns; | |
778 | struct pid *pid; | |
779 | struct task_struct *task; | |
780 | int ret; | |
781 | ||
782 | ns = inode->i_sb->s_fs_info; | |
783 | pid = proc_pid(inode); | |
784 | task = get_pid_task(pid, PIDTYPE_PID); | |
785 | if (!task) | |
786 | return -ESRCH; | |
787 | ||
788 | ret = PROC_I(inode)->op.proc_show(m, ns, pid, task); | |
789 | ||
790 | put_task_struct(task); | |
791 | return ret; | |
792 | } | |
793 | ||
794 | static int proc_single_open(struct inode *inode, struct file *filp) | |
795 | { | |
c6a34058 | 796 | return single_open(filp, proc_single_show, inode); |
be614086 EB |
797 | } |
798 | ||
799 | static const struct file_operations proc_single_file_operations = { | |
800 | .open = proc_single_open, | |
801 | .read = seq_read, | |
802 | .llseek = seq_lseek, | |
803 | .release = single_release, | |
804 | }; | |
805 | ||
5381e169 ON |
806 | |
807 | struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode) | |
1da177e4 | 808 | { |
5381e169 ON |
809 | struct task_struct *task = get_proc_task(inode); |
810 | struct mm_struct *mm = ERR_PTR(-ESRCH); | |
e268337d | 811 | |
5381e169 | 812 | if (task) { |
caaee623 | 813 | mm = mm_access(task, mode | PTRACE_MODE_FSCREDS); |
5381e169 | 814 | put_task_struct(task); |
e268337d | 815 | |
5381e169 ON |
816 | if (!IS_ERR_OR_NULL(mm)) { |
817 | /* ensure this mm_struct can't be freed */ | |
818 | atomic_inc(&mm->mm_count); | |
819 | /* but do not pin its memory */ | |
820 | mmput(mm); | |
821 | } | |
822 | } | |
823 | ||
824 | return mm; | |
825 | } | |
826 | ||
827 | static int __mem_open(struct inode *inode, struct file *file, unsigned int mode) | |
828 | { | |
829 | struct mm_struct *mm = proc_mem_open(inode, mode); | |
e268337d LT |
830 | |
831 | if (IS_ERR(mm)) | |
832 | return PTR_ERR(mm); | |
833 | ||
e268337d | 834 | file->private_data = mm; |
1da177e4 LT |
835 | return 0; |
836 | } | |
837 | ||
b409e578 CW |
838 | static int mem_open(struct inode *inode, struct file *file) |
839 | { | |
bc452b4b DH |
840 | int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH); |
841 | ||
842 | /* OK to pass negative loff_t, we can catch out-of-range */ | |
843 | file->f_mode |= FMODE_UNSIGNED_OFFSET; | |
844 | ||
845 | return ret; | |
b409e578 CW |
846 | } |
847 | ||
572d34b9 ON |
848 | static ssize_t mem_rw(struct file *file, char __user *buf, |
849 | size_t count, loff_t *ppos, int write) | |
1da177e4 | 850 | { |
e268337d | 851 | struct mm_struct *mm = file->private_data; |
572d34b9 ON |
852 | unsigned long addr = *ppos; |
853 | ssize_t copied; | |
1da177e4 | 854 | char *page; |
1da177e4 | 855 | |
e268337d LT |
856 | if (!mm) |
857 | return 0; | |
99f89551 | 858 | |
30cd8903 KM |
859 | page = (char *)__get_free_page(GFP_TEMPORARY); |
860 | if (!page) | |
e268337d | 861 | return -ENOMEM; |
1da177e4 | 862 | |
f7ca54f4 | 863 | copied = 0; |
6d08f2c7 ON |
864 | if (!atomic_inc_not_zero(&mm->mm_users)) |
865 | goto free; | |
866 | ||
1da177e4 | 867 | while (count > 0) { |
572d34b9 | 868 | int this_len = min_t(int, count, PAGE_SIZE); |
1da177e4 | 869 | |
572d34b9 | 870 | if (write && copy_from_user(page, buf, this_len)) { |
1da177e4 LT |
871 | copied = -EFAULT; |
872 | break; | |
873 | } | |
572d34b9 ON |
874 | |
875 | this_len = access_remote_vm(mm, addr, page, this_len, write); | |
876 | if (!this_len) { | |
1da177e4 LT |
877 | if (!copied) |
878 | copied = -EIO; | |
879 | break; | |
880 | } | |
572d34b9 ON |
881 | |
882 | if (!write && copy_to_user(buf, page, this_len)) { | |
883 | copied = -EFAULT; | |
884 | break; | |
885 | } | |
886 | ||
887 | buf += this_len; | |
888 | addr += this_len; | |
889 | copied += this_len; | |
890 | count -= this_len; | |
1da177e4 | 891 | } |
572d34b9 | 892 | *ppos = addr; |
30cd8903 | 893 | |
6d08f2c7 ON |
894 | mmput(mm); |
895 | free: | |
30cd8903 | 896 | free_page((unsigned long) page); |
1da177e4 LT |
897 | return copied; |
898 | } | |
1da177e4 | 899 | |
572d34b9 ON |
900 | static ssize_t mem_read(struct file *file, char __user *buf, |
901 | size_t count, loff_t *ppos) | |
902 | { | |
903 | return mem_rw(file, buf, count, ppos, 0); | |
904 | } | |
905 | ||
906 | static ssize_t mem_write(struct file *file, const char __user *buf, | |
907 | size_t count, loff_t *ppos) | |
908 | { | |
909 | return mem_rw(file, (char __user*)buf, count, ppos, 1); | |
910 | } | |
911 | ||
85863e47 | 912 | loff_t mem_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
913 | { |
914 | switch (orig) { | |
915 | case 0: | |
916 | file->f_pos = offset; | |
917 | break; | |
918 | case 1: | |
919 | file->f_pos += offset; | |
920 | break; | |
921 | default: | |
922 | return -EINVAL; | |
923 | } | |
924 | force_successful_syscall_return(); | |
925 | return file->f_pos; | |
926 | } | |
927 | ||
e268337d LT |
928 | static int mem_release(struct inode *inode, struct file *file) |
929 | { | |
930 | struct mm_struct *mm = file->private_data; | |
71879d3c | 931 | if (mm) |
6d08f2c7 | 932 | mmdrop(mm); |
e268337d LT |
933 | return 0; |
934 | } | |
935 | ||
00977a59 | 936 | static const struct file_operations proc_mem_operations = { |
1da177e4 LT |
937 | .llseek = mem_lseek, |
938 | .read = mem_read, | |
939 | .write = mem_write, | |
940 | .open = mem_open, | |
e268337d | 941 | .release = mem_release, |
1da177e4 LT |
942 | }; |
943 | ||
b409e578 CW |
944 | static int environ_open(struct inode *inode, struct file *file) |
945 | { | |
946 | return __mem_open(inode, file, PTRACE_MODE_READ); | |
947 | } | |
948 | ||
315e28c8 JP |
949 | static ssize_t environ_read(struct file *file, char __user *buf, |
950 | size_t count, loff_t *ppos) | |
951 | { | |
315e28c8 JP |
952 | char *page; |
953 | unsigned long src = *ppos; | |
b409e578 CW |
954 | int ret = 0; |
955 | struct mm_struct *mm = file->private_data; | |
a3b609ef | 956 | unsigned long env_start, env_end; |
315e28c8 | 957 | |
8148a73c MK |
958 | /* Ensure the process spawned far enough to have an environment. */ |
959 | if (!mm || !mm->env_end) | |
b409e578 | 960 | return 0; |
315e28c8 | 961 | |
315e28c8 JP |
962 | page = (char *)__get_free_page(GFP_TEMPORARY); |
963 | if (!page) | |
b409e578 | 964 | return -ENOMEM; |
315e28c8 | 965 | |
d6f64b89 | 966 | ret = 0; |
b409e578 CW |
967 | if (!atomic_inc_not_zero(&mm->mm_users)) |
968 | goto free; | |
a3b609ef MG |
969 | |
970 | down_read(&mm->mmap_sem); | |
971 | env_start = mm->env_start; | |
972 | env_end = mm->env_end; | |
973 | up_read(&mm->mmap_sem); | |
974 | ||
315e28c8 | 975 | while (count > 0) { |
e8905ec2 DH |
976 | size_t this_len, max_len; |
977 | int retval; | |
315e28c8 | 978 | |
a3b609ef | 979 | if (src >= (env_end - env_start)) |
315e28c8 JP |
980 | break; |
981 | ||
a3b609ef | 982 | this_len = env_end - (env_start + src); |
e8905ec2 DH |
983 | |
984 | max_len = min_t(size_t, PAGE_SIZE, count); | |
985 | this_len = min(max_len, this_len); | |
315e28c8 | 986 | |
a3b609ef | 987 | retval = access_remote_vm(mm, (env_start + src), |
315e28c8 JP |
988 | page, this_len, 0); |
989 | ||
990 | if (retval <= 0) { | |
991 | ret = retval; | |
992 | break; | |
993 | } | |
994 | ||
995 | if (copy_to_user(buf, page, retval)) { | |
996 | ret = -EFAULT; | |
997 | break; | |
998 | } | |
999 | ||
1000 | ret += retval; | |
1001 | src += retval; | |
1002 | buf += retval; | |
1003 | count -= retval; | |
1004 | } | |
1005 | *ppos = src; | |
315e28c8 | 1006 | mmput(mm); |
b409e578 CW |
1007 | |
1008 | free: | |
315e28c8 | 1009 | free_page((unsigned long) page); |
315e28c8 JP |
1010 | return ret; |
1011 | } | |
1012 | ||
1013 | static const struct file_operations proc_environ_operations = { | |
b409e578 | 1014 | .open = environ_open, |
315e28c8 | 1015 | .read = environ_read, |
87df8424 | 1016 | .llseek = generic_file_llseek, |
b409e578 | 1017 | .release = mem_release, |
315e28c8 JP |
1018 | }; |
1019 | ||
fa0cbbf1 DR |
1020 | static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count, |
1021 | loff_t *ppos) | |
1022 | { | |
496ad9aa | 1023 | struct task_struct *task = get_proc_task(file_inode(file)); |
fa0cbbf1 DR |
1024 | char buffer[PROC_NUMBUF]; |
1025 | int oom_adj = OOM_ADJUST_MIN; | |
1026 | size_t len; | |
1027 | unsigned long flags; | |
1028 | ||
1029 | if (!task) | |
1030 | return -ESRCH; | |
1031 | if (lock_task_sighand(task, &flags)) { | |
1032 | if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX) | |
1033 | oom_adj = OOM_ADJUST_MAX; | |
1034 | else | |
1035 | oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) / | |
1036 | OOM_SCORE_ADJ_MAX; | |
1037 | unlock_task_sighand(task, &flags); | |
1038 | } | |
1039 | put_task_struct(task); | |
1040 | len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj); | |
1041 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
1042 | } | |
1043 | ||
b72bdfa7 DR |
1044 | /* |
1045 | * /proc/pid/oom_adj exists solely for backwards compatibility with previous | |
1046 | * kernels. The effective policy is defined by oom_score_adj, which has a | |
1047 | * different scale: oom_adj grew exponentially and oom_score_adj grows linearly. | |
1048 | * Values written to oom_adj are simply mapped linearly to oom_score_adj. | |
1049 | * Processes that become oom disabled via oom_adj will still be oom disabled | |
1050 | * with this implementation. | |
1051 | * | |
1052 | * oom_adj cannot be removed since existing userspace binaries use it. | |
1053 | */ | |
fa0cbbf1 DR |
1054 | static ssize_t oom_adj_write(struct file *file, const char __user *buf, |
1055 | size_t count, loff_t *ppos) | |
1056 | { | |
1057 | struct task_struct *task; | |
1058 | char buffer[PROC_NUMBUF]; | |
1059 | int oom_adj; | |
1060 | unsigned long flags; | |
1061 | int err; | |
1062 | ||
1063 | memset(buffer, 0, sizeof(buffer)); | |
1064 | if (count > sizeof(buffer) - 1) | |
1065 | count = sizeof(buffer) - 1; | |
1066 | if (copy_from_user(buffer, buf, count)) { | |
1067 | err = -EFAULT; | |
1068 | goto out; | |
1069 | } | |
1070 | ||
1071 | err = kstrtoint(strstrip(buffer), 0, &oom_adj); | |
1072 | if (err) | |
1073 | goto out; | |
1074 | if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) && | |
1075 | oom_adj != OOM_DISABLE) { | |
1076 | err = -EINVAL; | |
1077 | goto out; | |
1078 | } | |
1079 | ||
496ad9aa | 1080 | task = get_proc_task(file_inode(file)); |
fa0cbbf1 DR |
1081 | if (!task) { |
1082 | err = -ESRCH; | |
1083 | goto out; | |
1084 | } | |
1085 | ||
1086 | task_lock(task); | |
1087 | if (!task->mm) { | |
1088 | err = -EINVAL; | |
1089 | goto err_task_lock; | |
1090 | } | |
1091 | ||
1092 | if (!lock_task_sighand(task, &flags)) { | |
1093 | err = -ESRCH; | |
1094 | goto err_task_lock; | |
1095 | } | |
1096 | ||
1097 | /* | |
1098 | * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum | |
1099 | * value is always attainable. | |
1100 | */ | |
1101 | if (oom_adj == OOM_ADJUST_MAX) | |
1102 | oom_adj = OOM_SCORE_ADJ_MAX; | |
1103 | else | |
1104 | oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE; | |
1105 | ||
1106 | if (oom_adj < task->signal->oom_score_adj && | |
1107 | !capable(CAP_SYS_RESOURCE)) { | |
1108 | err = -EACCES; | |
1109 | goto err_sighand; | |
1110 | } | |
1111 | ||
1112 | /* | |
1113 | * /proc/pid/oom_adj is provided for legacy purposes, ask users to use | |
1114 | * /proc/pid/oom_score_adj instead. | |
1115 | */ | |
87ebdc00 | 1116 | pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n", |
fa0cbbf1 DR |
1117 | current->comm, task_pid_nr(current), task_pid_nr(task), |
1118 | task_pid_nr(task)); | |
1119 | ||
1120 | task->signal->oom_score_adj = oom_adj; | |
1121 | trace_oom_score_adj_update(task); | |
1122 | err_sighand: | |
1123 | unlock_task_sighand(task, &flags); | |
1124 | err_task_lock: | |
1125 | task_unlock(task); | |
1126 | put_task_struct(task); | |
1127 | out: | |
1128 | return err < 0 ? err : count; | |
1129 | } | |
1130 | ||
1131 | static const struct file_operations proc_oom_adj_operations = { | |
1132 | .read = oom_adj_read, | |
1133 | .write = oom_adj_write, | |
1134 | .llseek = generic_file_llseek, | |
1135 | }; | |
1136 | ||
a63d83f4 DR |
1137 | static ssize_t oom_score_adj_read(struct file *file, char __user *buf, |
1138 | size_t count, loff_t *ppos) | |
1139 | { | |
496ad9aa | 1140 | struct task_struct *task = get_proc_task(file_inode(file)); |
a63d83f4 | 1141 | char buffer[PROC_NUMBUF]; |
a9c58b90 | 1142 | short oom_score_adj = OOM_SCORE_ADJ_MIN; |
a63d83f4 DR |
1143 | unsigned long flags; |
1144 | size_t len; | |
1145 | ||
1146 | if (!task) | |
1147 | return -ESRCH; | |
1148 | if (lock_task_sighand(task, &flags)) { | |
1149 | oom_score_adj = task->signal->oom_score_adj; | |
1150 | unlock_task_sighand(task, &flags); | |
1151 | } | |
1152 | put_task_struct(task); | |
a9c58b90 | 1153 | len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj); |
a63d83f4 DR |
1154 | return simple_read_from_buffer(buf, count, ppos, buffer, len); |
1155 | } | |
1156 | ||
1157 | static ssize_t oom_score_adj_write(struct file *file, const char __user *buf, | |
1158 | size_t count, loff_t *ppos) | |
1159 | { | |
1160 | struct task_struct *task; | |
1161 | char buffer[PROC_NUMBUF]; | |
1162 | unsigned long flags; | |
0a8cb8e3 | 1163 | int oom_score_adj; |
a63d83f4 DR |
1164 | int err; |
1165 | ||
1166 | memset(buffer, 0, sizeof(buffer)); | |
1167 | if (count > sizeof(buffer) - 1) | |
1168 | count = sizeof(buffer) - 1; | |
723548bf DR |
1169 | if (copy_from_user(buffer, buf, count)) { |
1170 | err = -EFAULT; | |
1171 | goto out; | |
1172 | } | |
a63d83f4 | 1173 | |
0a8cb8e3 | 1174 | err = kstrtoint(strstrip(buffer), 0, &oom_score_adj); |
a63d83f4 | 1175 | if (err) |
723548bf | 1176 | goto out; |
a63d83f4 | 1177 | if (oom_score_adj < OOM_SCORE_ADJ_MIN || |
723548bf DR |
1178 | oom_score_adj > OOM_SCORE_ADJ_MAX) { |
1179 | err = -EINVAL; | |
1180 | goto out; | |
1181 | } | |
a63d83f4 | 1182 | |
496ad9aa | 1183 | task = get_proc_task(file_inode(file)); |
723548bf DR |
1184 | if (!task) { |
1185 | err = -ESRCH; | |
1186 | goto out; | |
1187 | } | |
d19d5476 DR |
1188 | |
1189 | task_lock(task); | |
1190 | if (!task->mm) { | |
1191 | err = -EINVAL; | |
1192 | goto err_task_lock; | |
1193 | } | |
1194 | ||
a63d83f4 | 1195 | if (!lock_task_sighand(task, &flags)) { |
723548bf | 1196 | err = -ESRCH; |
d19d5476 | 1197 | goto err_task_lock; |
a63d83f4 | 1198 | } |
d19d5476 | 1199 | |
a9c58b90 | 1200 | if ((short)oom_score_adj < task->signal->oom_score_adj_min && |
a63d83f4 | 1201 | !capable(CAP_SYS_RESOURCE)) { |
723548bf DR |
1202 | err = -EACCES; |
1203 | goto err_sighand; | |
a63d83f4 DR |
1204 | } |
1205 | ||
a9c58b90 | 1206 | task->signal->oom_score_adj = (short)oom_score_adj; |
dabb16f6 | 1207 | if (has_capability_noaudit(current, CAP_SYS_RESOURCE)) |
a9c58b90 | 1208 | task->signal->oom_score_adj_min = (short)oom_score_adj; |
43d2b113 | 1209 | trace_oom_score_adj_update(task); |
01dc52eb | 1210 | |
723548bf | 1211 | err_sighand: |
a63d83f4 | 1212 | unlock_task_sighand(task, &flags); |
d19d5476 DR |
1213 | err_task_lock: |
1214 | task_unlock(task); | |
a63d83f4 | 1215 | put_task_struct(task); |
723548bf DR |
1216 | out: |
1217 | return err < 0 ? err : count; | |
a63d83f4 DR |
1218 | } |
1219 | ||
1220 | static const struct file_operations proc_oom_score_adj_operations = { | |
1221 | .read = oom_score_adj_read, | |
1222 | .write = oom_score_adj_write, | |
6038f373 | 1223 | .llseek = default_llseek, |
a63d83f4 DR |
1224 | }; |
1225 | ||
1da177e4 LT |
1226 | #ifdef CONFIG_AUDITSYSCALL |
1227 | #define TMPBUFLEN 21 | |
1228 | static ssize_t proc_loginuid_read(struct file * file, char __user * buf, | |
1229 | size_t count, loff_t *ppos) | |
1230 | { | |
496ad9aa | 1231 | struct inode * inode = file_inode(file); |
99f89551 | 1232 | struct task_struct *task = get_proc_task(inode); |
1da177e4 LT |
1233 | ssize_t length; |
1234 | char tmpbuf[TMPBUFLEN]; | |
1235 | ||
99f89551 EB |
1236 | if (!task) |
1237 | return -ESRCH; | |
1da177e4 | 1238 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", |
e1760bd5 EB |
1239 | from_kuid(file->f_cred->user_ns, |
1240 | audit_get_loginuid(task))); | |
99f89551 | 1241 | put_task_struct(task); |
1da177e4 LT |
1242 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); |
1243 | } | |
1244 | ||
1245 | static ssize_t proc_loginuid_write(struct file * file, const char __user * buf, | |
1246 | size_t count, loff_t *ppos) | |
1247 | { | |
496ad9aa | 1248 | struct inode * inode = file_inode(file); |
1da177e4 | 1249 | uid_t loginuid; |
e1760bd5 | 1250 | kuid_t kloginuid; |
774636e1 | 1251 | int rv; |
1da177e4 | 1252 | |
7dc52157 PM |
1253 | rcu_read_lock(); |
1254 | if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) { | |
1255 | rcu_read_unlock(); | |
1da177e4 | 1256 | return -EPERM; |
7dc52157 PM |
1257 | } |
1258 | rcu_read_unlock(); | |
1da177e4 | 1259 | |
1da177e4 LT |
1260 | if (*ppos != 0) { |
1261 | /* No partial writes. */ | |
1262 | return -EINVAL; | |
1263 | } | |
1da177e4 | 1264 | |
774636e1 AD |
1265 | rv = kstrtou32_from_user(buf, count, 10, &loginuid); |
1266 | if (rv < 0) | |
1267 | return rv; | |
81407c84 EP |
1268 | |
1269 | /* is userspace tring to explicitly UNSET the loginuid? */ | |
1270 | if (loginuid == AUDIT_UID_UNSET) { | |
1271 | kloginuid = INVALID_UID; | |
1272 | } else { | |
1273 | kloginuid = make_kuid(file->f_cred->user_ns, loginuid); | |
774636e1 AD |
1274 | if (!uid_valid(kloginuid)) |
1275 | return -EINVAL; | |
e1760bd5 EB |
1276 | } |
1277 | ||
774636e1 AD |
1278 | rv = audit_set_loginuid(kloginuid); |
1279 | if (rv < 0) | |
1280 | return rv; | |
1281 | return count; | |
1da177e4 LT |
1282 | } |
1283 | ||
00977a59 | 1284 | static const struct file_operations proc_loginuid_operations = { |
1da177e4 LT |
1285 | .read = proc_loginuid_read, |
1286 | .write = proc_loginuid_write, | |
87df8424 | 1287 | .llseek = generic_file_llseek, |
1da177e4 | 1288 | }; |
1e0bd755 EP |
1289 | |
1290 | static ssize_t proc_sessionid_read(struct file * file, char __user * buf, | |
1291 | size_t count, loff_t *ppos) | |
1292 | { | |
496ad9aa | 1293 | struct inode * inode = file_inode(file); |
1e0bd755 EP |
1294 | struct task_struct *task = get_proc_task(inode); |
1295 | ssize_t length; | |
1296 | char tmpbuf[TMPBUFLEN]; | |
1297 | ||
1298 | if (!task) | |
1299 | return -ESRCH; | |
1300 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", | |
1301 | audit_get_sessionid(task)); | |
1302 | put_task_struct(task); | |
1303 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); | |
1304 | } | |
1305 | ||
1306 | static const struct file_operations proc_sessionid_operations = { | |
1307 | .read = proc_sessionid_read, | |
87df8424 | 1308 | .llseek = generic_file_llseek, |
1e0bd755 | 1309 | }; |
1da177e4 LT |
1310 | #endif |
1311 | ||
f4f154fd AM |
1312 | #ifdef CONFIG_FAULT_INJECTION |
1313 | static ssize_t proc_fault_inject_read(struct file * file, char __user * buf, | |
1314 | size_t count, loff_t *ppos) | |
1315 | { | |
496ad9aa | 1316 | struct task_struct *task = get_proc_task(file_inode(file)); |
f4f154fd AM |
1317 | char buffer[PROC_NUMBUF]; |
1318 | size_t len; | |
1319 | int make_it_fail; | |
f4f154fd AM |
1320 | |
1321 | if (!task) | |
1322 | return -ESRCH; | |
1323 | make_it_fail = task->make_it_fail; | |
1324 | put_task_struct(task); | |
1325 | ||
1326 | len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail); | |
0c28f287 AM |
1327 | |
1328 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
f4f154fd AM |
1329 | } |
1330 | ||
1331 | static ssize_t proc_fault_inject_write(struct file * file, | |
1332 | const char __user * buf, size_t count, loff_t *ppos) | |
1333 | { | |
1334 | struct task_struct *task; | |
774636e1 | 1335 | char buffer[PROC_NUMBUF]; |
f4f154fd | 1336 | int make_it_fail; |
774636e1 | 1337 | int rv; |
f4f154fd AM |
1338 | |
1339 | if (!capable(CAP_SYS_RESOURCE)) | |
1340 | return -EPERM; | |
1341 | memset(buffer, 0, sizeof(buffer)); | |
1342 | if (count > sizeof(buffer) - 1) | |
1343 | count = sizeof(buffer) - 1; | |
1344 | if (copy_from_user(buffer, buf, count)) | |
1345 | return -EFAULT; | |
774636e1 AD |
1346 | rv = kstrtoint(strstrip(buffer), 0, &make_it_fail); |
1347 | if (rv < 0) | |
1348 | return rv; | |
16caed31 DJ |
1349 | if (make_it_fail < 0 || make_it_fail > 1) |
1350 | return -EINVAL; | |
1351 | ||
496ad9aa | 1352 | task = get_proc_task(file_inode(file)); |
f4f154fd AM |
1353 | if (!task) |
1354 | return -ESRCH; | |
1355 | task->make_it_fail = make_it_fail; | |
1356 | put_task_struct(task); | |
cba8aafe VL |
1357 | |
1358 | return count; | |
f4f154fd AM |
1359 | } |
1360 | ||
00977a59 | 1361 | static const struct file_operations proc_fault_inject_operations = { |
f4f154fd AM |
1362 | .read = proc_fault_inject_read, |
1363 | .write = proc_fault_inject_write, | |
87df8424 | 1364 | .llseek = generic_file_llseek, |
f4f154fd AM |
1365 | }; |
1366 | #endif | |
1367 | ||
9745512c | 1368 | |
43ae34cb IM |
1369 | #ifdef CONFIG_SCHED_DEBUG |
1370 | /* | |
1371 | * Print out various scheduling related per-task fields: | |
1372 | */ | |
1373 | static int sched_show(struct seq_file *m, void *v) | |
1374 | { | |
1375 | struct inode *inode = m->private; | |
1376 | struct task_struct *p; | |
1377 | ||
43ae34cb IM |
1378 | p = get_proc_task(inode); |
1379 | if (!p) | |
1380 | return -ESRCH; | |
1381 | proc_sched_show_task(p, m); | |
1382 | ||
1383 | put_task_struct(p); | |
1384 | ||
1385 | return 0; | |
1386 | } | |
1387 | ||
1388 | static ssize_t | |
1389 | sched_write(struct file *file, const char __user *buf, | |
1390 | size_t count, loff_t *offset) | |
1391 | { | |
496ad9aa | 1392 | struct inode *inode = file_inode(file); |
43ae34cb IM |
1393 | struct task_struct *p; |
1394 | ||
43ae34cb IM |
1395 | p = get_proc_task(inode); |
1396 | if (!p) | |
1397 | return -ESRCH; | |
1398 | proc_sched_set_task(p); | |
1399 | ||
1400 | put_task_struct(p); | |
1401 | ||
1402 | return count; | |
1403 | } | |
1404 | ||
1405 | static int sched_open(struct inode *inode, struct file *filp) | |
1406 | { | |
c6a34058 | 1407 | return single_open(filp, sched_show, inode); |
43ae34cb IM |
1408 | } |
1409 | ||
1410 | static const struct file_operations proc_pid_sched_operations = { | |
1411 | .open = sched_open, | |
1412 | .read = seq_read, | |
1413 | .write = sched_write, | |
1414 | .llseek = seq_lseek, | |
5ea473a1 | 1415 | .release = single_release, |
43ae34cb IM |
1416 | }; |
1417 | ||
1418 | #endif | |
1419 | ||
5091faa4 MG |
1420 | #ifdef CONFIG_SCHED_AUTOGROUP |
1421 | /* | |
1422 | * Print out autogroup related information: | |
1423 | */ | |
1424 | static int sched_autogroup_show(struct seq_file *m, void *v) | |
1425 | { | |
1426 | struct inode *inode = m->private; | |
1427 | struct task_struct *p; | |
1428 | ||
1429 | p = get_proc_task(inode); | |
1430 | if (!p) | |
1431 | return -ESRCH; | |
1432 | proc_sched_autogroup_show_task(p, m); | |
1433 | ||
1434 | put_task_struct(p); | |
1435 | ||
1436 | return 0; | |
1437 | } | |
1438 | ||
1439 | static ssize_t | |
1440 | sched_autogroup_write(struct file *file, const char __user *buf, | |
1441 | size_t count, loff_t *offset) | |
1442 | { | |
496ad9aa | 1443 | struct inode *inode = file_inode(file); |
5091faa4 MG |
1444 | struct task_struct *p; |
1445 | char buffer[PROC_NUMBUF]; | |
0a8cb8e3 | 1446 | int nice; |
5091faa4 MG |
1447 | int err; |
1448 | ||
1449 | memset(buffer, 0, sizeof(buffer)); | |
1450 | if (count > sizeof(buffer) - 1) | |
1451 | count = sizeof(buffer) - 1; | |
1452 | if (copy_from_user(buffer, buf, count)) | |
1453 | return -EFAULT; | |
1454 | ||
0a8cb8e3 AD |
1455 | err = kstrtoint(strstrip(buffer), 0, &nice); |
1456 | if (err < 0) | |
1457 | return err; | |
5091faa4 MG |
1458 | |
1459 | p = get_proc_task(inode); | |
1460 | if (!p) | |
1461 | return -ESRCH; | |
1462 | ||
2e5b5b3a | 1463 | err = proc_sched_autogroup_set_nice(p, nice); |
5091faa4 MG |
1464 | if (err) |
1465 | count = err; | |
1466 | ||
1467 | put_task_struct(p); | |
1468 | ||
1469 | return count; | |
1470 | } | |
1471 | ||
1472 | static int sched_autogroup_open(struct inode *inode, struct file *filp) | |
1473 | { | |
1474 | int ret; | |
1475 | ||
1476 | ret = single_open(filp, sched_autogroup_show, NULL); | |
1477 | if (!ret) { | |
1478 | struct seq_file *m = filp->private_data; | |
1479 | ||
1480 | m->private = inode; | |
1481 | } | |
1482 | return ret; | |
1483 | } | |
1484 | ||
1485 | static const struct file_operations proc_pid_sched_autogroup_operations = { | |
1486 | .open = sched_autogroup_open, | |
1487 | .read = seq_read, | |
1488 | .write = sched_autogroup_write, | |
1489 | .llseek = seq_lseek, | |
1490 | .release = single_release, | |
1491 | }; | |
1492 | ||
1493 | #endif /* CONFIG_SCHED_AUTOGROUP */ | |
1494 | ||
4614a696 | 1495 | static ssize_t comm_write(struct file *file, const char __user *buf, |
1496 | size_t count, loff_t *offset) | |
1497 | { | |
496ad9aa | 1498 | struct inode *inode = file_inode(file); |
4614a696 | 1499 | struct task_struct *p; |
1500 | char buffer[TASK_COMM_LEN]; | |
830e0fc9 | 1501 | const size_t maxlen = sizeof(buffer) - 1; |
4614a696 | 1502 | |
1503 | memset(buffer, 0, sizeof(buffer)); | |
830e0fc9 | 1504 | if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count)) |
4614a696 | 1505 | return -EFAULT; |
1506 | ||
1507 | p = get_proc_task(inode); | |
1508 | if (!p) | |
1509 | return -ESRCH; | |
1510 | ||
1511 | if (same_thread_group(current, p)) | |
1512 | set_task_comm(p, buffer); | |
1513 | else | |
1514 | count = -EINVAL; | |
1515 | ||
1516 | put_task_struct(p); | |
1517 | ||
1518 | return count; | |
1519 | } | |
1520 | ||
1521 | static int comm_show(struct seq_file *m, void *v) | |
1522 | { | |
1523 | struct inode *inode = m->private; | |
1524 | struct task_struct *p; | |
1525 | ||
1526 | p = get_proc_task(inode); | |
1527 | if (!p) | |
1528 | return -ESRCH; | |
1529 | ||
1530 | task_lock(p); | |
1531 | seq_printf(m, "%s\n", p->comm); | |
1532 | task_unlock(p); | |
1533 | ||
1534 | put_task_struct(p); | |
1535 | ||
1536 | return 0; | |
1537 | } | |
1538 | ||
1539 | static int comm_open(struct inode *inode, struct file *filp) | |
1540 | { | |
c6a34058 | 1541 | return single_open(filp, comm_show, inode); |
4614a696 | 1542 | } |
1543 | ||
1544 | static const struct file_operations proc_pid_set_comm_operations = { | |
1545 | .open = comm_open, | |
1546 | .read = seq_read, | |
1547 | .write = comm_write, | |
1548 | .llseek = seq_lseek, | |
1549 | .release = single_release, | |
1550 | }; | |
1551 | ||
7773fbc5 | 1552 | static int proc_exe_link(struct dentry *dentry, struct path *exe_path) |
925d1c40 MH |
1553 | { |
1554 | struct task_struct *task; | |
1555 | struct mm_struct *mm; | |
1556 | struct file *exe_file; | |
1557 | ||
2b0143b5 | 1558 | task = get_proc_task(d_inode(dentry)); |
925d1c40 MH |
1559 | if (!task) |
1560 | return -ENOENT; | |
1561 | mm = get_task_mm(task); | |
1562 | put_task_struct(task); | |
1563 | if (!mm) | |
1564 | return -ENOENT; | |
1565 | exe_file = get_mm_exe_file(mm); | |
1566 | mmput(mm); | |
1567 | if (exe_file) { | |
1568 | *exe_path = exe_file->f_path; | |
1569 | path_get(&exe_file->f_path); | |
1570 | fput(exe_file); | |
1571 | return 0; | |
1572 | } else | |
1573 | return -ENOENT; | |
1574 | } | |
1575 | ||
6b255391 | 1576 | static const char *proc_pid_get_link(struct dentry *dentry, |
fceef393 AV |
1577 | struct inode *inode, |
1578 | struct delayed_call *done) | |
1da177e4 | 1579 | { |
408ef013 | 1580 | struct path path; |
1da177e4 LT |
1581 | int error = -EACCES; |
1582 | ||
6b255391 AV |
1583 | if (!dentry) |
1584 | return ERR_PTR(-ECHILD); | |
1585 | ||
778c1144 EB |
1586 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1587 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1588 | goto out; |
1da177e4 | 1589 | |
408ef013 CH |
1590 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1591 | if (error) | |
1592 | goto out; | |
1593 | ||
6e77137b | 1594 | nd_jump_link(&path); |
408ef013 | 1595 | return NULL; |
1da177e4 | 1596 | out: |
008b150a | 1597 | return ERR_PTR(error); |
1da177e4 LT |
1598 | } |
1599 | ||
3dcd25f3 | 1600 | static int do_proc_readlink(struct path *path, char __user *buffer, int buflen) |
1da177e4 | 1601 | { |
e12ba74d | 1602 | char *tmp = (char*)__get_free_page(GFP_TEMPORARY); |
3dcd25f3 | 1603 | char *pathname; |
1da177e4 LT |
1604 | int len; |
1605 | ||
1606 | if (!tmp) | |
1607 | return -ENOMEM; | |
0c28f287 | 1608 | |
7b2a69ba | 1609 | pathname = d_path(path, tmp, PAGE_SIZE); |
3dcd25f3 JB |
1610 | len = PTR_ERR(pathname); |
1611 | if (IS_ERR(pathname)) | |
1da177e4 | 1612 | goto out; |
3dcd25f3 | 1613 | len = tmp + PAGE_SIZE - 1 - pathname; |
1da177e4 LT |
1614 | |
1615 | if (len > buflen) | |
1616 | len = buflen; | |
3dcd25f3 | 1617 | if (copy_to_user(buffer, pathname, len)) |
1da177e4 LT |
1618 | len = -EFAULT; |
1619 | out: | |
1620 | free_page((unsigned long)tmp); | |
1621 | return len; | |
1622 | } | |
1623 | ||
1624 | static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen) | |
1625 | { | |
1626 | int error = -EACCES; | |
2b0143b5 | 1627 | struct inode *inode = d_inode(dentry); |
3dcd25f3 | 1628 | struct path path; |
1da177e4 | 1629 | |
778c1144 EB |
1630 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1631 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1632 | goto out; |
1da177e4 | 1633 | |
7773fbc5 | 1634 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1da177e4 LT |
1635 | if (error) |
1636 | goto out; | |
1637 | ||
3dcd25f3 JB |
1638 | error = do_proc_readlink(&path, buffer, buflen); |
1639 | path_put(&path); | |
1da177e4 | 1640 | out: |
1da177e4 LT |
1641 | return error; |
1642 | } | |
1643 | ||
faf60af1 | 1644 | const struct inode_operations proc_pid_link_inode_operations = { |
1da177e4 | 1645 | .readlink = proc_pid_readlink, |
6b255391 | 1646 | .get_link = proc_pid_get_link, |
6d76fa58 | 1647 | .setattr = proc_setattr, |
1da177e4 LT |
1648 | }; |
1649 | ||
28a6d671 EB |
1650 | |
1651 | /* building an inode */ | |
1652 | ||
6b4e306a | 1653 | struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task) |
28a6d671 EB |
1654 | { |
1655 | struct inode * inode; | |
1656 | struct proc_inode *ei; | |
c69e8d9c | 1657 | const struct cred *cred; |
1da177e4 | 1658 | |
28a6d671 | 1659 | /* We need a new inode */ |
1da177e4 | 1660 | |
28a6d671 EB |
1661 | inode = new_inode(sb); |
1662 | if (!inode) | |
1663 | goto out; | |
1664 | ||
1665 | /* Common stuff */ | |
1666 | ei = PROC_I(inode); | |
85fe4025 | 1667 | inode->i_ino = get_next_ino(); |
28a6d671 | 1668 | inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; |
28a6d671 EB |
1669 | inode->i_op = &proc_def_inode_operations; |
1670 | ||
1671 | /* | |
1672 | * grab the reference to task. | |
1673 | */ | |
1a657f78 | 1674 | ei->pid = get_task_pid(task, PIDTYPE_PID); |
28a6d671 EB |
1675 | if (!ei->pid) |
1676 | goto out_unlock; | |
1677 | ||
28a6d671 | 1678 | if (task_dumpable(task)) { |
c69e8d9c DH |
1679 | rcu_read_lock(); |
1680 | cred = __task_cred(task); | |
1681 | inode->i_uid = cred->euid; | |
1682 | inode->i_gid = cred->egid; | |
1683 | rcu_read_unlock(); | |
1da177e4 | 1684 | } |
28a6d671 EB |
1685 | security_task_to_inode(task, inode); |
1686 | ||
1da177e4 | 1687 | out: |
28a6d671 EB |
1688 | return inode; |
1689 | ||
1690 | out_unlock: | |
1691 | iput(inode); | |
1692 | return NULL; | |
1da177e4 LT |
1693 | } |
1694 | ||
6b4e306a | 1695 | int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat) |
1da177e4 | 1696 | { |
2b0143b5 | 1697 | struct inode *inode = d_inode(dentry); |
28a6d671 | 1698 | struct task_struct *task; |
c69e8d9c | 1699 | const struct cred *cred; |
0499680a | 1700 | struct pid_namespace *pid = dentry->d_sb->s_fs_info; |
c69e8d9c | 1701 | |
28a6d671 | 1702 | generic_fillattr(inode, stat); |
1da177e4 | 1703 | |
28a6d671 | 1704 | rcu_read_lock(); |
dcb0f222 EB |
1705 | stat->uid = GLOBAL_ROOT_UID; |
1706 | stat->gid = GLOBAL_ROOT_GID; | |
28a6d671 EB |
1707 | task = pid_task(proc_pid(inode), PIDTYPE_PID); |
1708 | if (task) { | |
0499680a VK |
1709 | if (!has_pid_permissions(pid, task, 2)) { |
1710 | rcu_read_unlock(); | |
1711 | /* | |
1712 | * This doesn't prevent learning whether PID exists, | |
1713 | * it only makes getattr() consistent with readdir(). | |
1714 | */ | |
1715 | return -ENOENT; | |
1716 | } | |
28a6d671 EB |
1717 | if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) || |
1718 | task_dumpable(task)) { | |
c69e8d9c DH |
1719 | cred = __task_cred(task); |
1720 | stat->uid = cred->euid; | |
1721 | stat->gid = cred->egid; | |
1da177e4 LT |
1722 | } |
1723 | } | |
28a6d671 | 1724 | rcu_read_unlock(); |
d6e71144 | 1725 | return 0; |
1da177e4 LT |
1726 | } |
1727 | ||
1da177e4 LT |
1728 | /* dentry stuff */ |
1729 | ||
1730 | /* | |
1731 | * Exceptional case: normally we are not allowed to unhash a busy | |
1732 | * directory. In this case, however, we can do it - no aliasing problems | |
1733 | * due to the way we treat inodes. | |
1734 | * | |
1735 | * Rewrite the inode's ownerships here because the owning task may have | |
1736 | * performed a setuid(), etc. | |
99f89551 EB |
1737 | * |
1738 | * Before the /proc/pid/status file was created the only way to read | |
1739 | * the effective uid of a /process was to stat /proc/pid. Reading | |
1740 | * /proc/pid/status is slow enough that procps and other packages | |
1741 | * kept stating /proc/pid. To keep the rules in /proc simple I have | |
1742 | * made this apply to all per process world readable and executable | |
1743 | * directories. | |
1da177e4 | 1744 | */ |
0b728e19 | 1745 | int pid_revalidate(struct dentry *dentry, unsigned int flags) |
1da177e4 | 1746 | { |
34286d66 NP |
1747 | struct inode *inode; |
1748 | struct task_struct *task; | |
c69e8d9c DH |
1749 | const struct cred *cred; |
1750 | ||
0b728e19 | 1751 | if (flags & LOOKUP_RCU) |
34286d66 NP |
1752 | return -ECHILD; |
1753 | ||
2b0143b5 | 1754 | inode = d_inode(dentry); |
34286d66 NP |
1755 | task = get_proc_task(inode); |
1756 | ||
99f89551 EB |
1757 | if (task) { |
1758 | if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) || | |
1759 | task_dumpable(task)) { | |
c69e8d9c DH |
1760 | rcu_read_lock(); |
1761 | cred = __task_cred(task); | |
1762 | inode->i_uid = cred->euid; | |
1763 | inode->i_gid = cred->egid; | |
1764 | rcu_read_unlock(); | |
1da177e4 | 1765 | } else { |
dcb0f222 EB |
1766 | inode->i_uid = GLOBAL_ROOT_UID; |
1767 | inode->i_gid = GLOBAL_ROOT_GID; | |
1da177e4 | 1768 | } |
9ee8ab9f | 1769 | inode->i_mode &= ~(S_ISUID | S_ISGID); |
1da177e4 | 1770 | security_task_to_inode(task, inode); |
99f89551 | 1771 | put_task_struct(task); |
1da177e4 LT |
1772 | return 1; |
1773 | } | |
1da177e4 LT |
1774 | return 0; |
1775 | } | |
1776 | ||
d855a4b7 ON |
1777 | static inline bool proc_inode_is_dead(struct inode *inode) |
1778 | { | |
1779 | return !proc_pid(inode)->tasks[PIDTYPE_PID].first; | |
1780 | } | |
1781 | ||
1dd704b6 DH |
1782 | int pid_delete_dentry(const struct dentry *dentry) |
1783 | { | |
1784 | /* Is the task we represent dead? | |
1785 | * If so, then don't put the dentry on the lru list, | |
1786 | * kill it immediately. | |
1787 | */ | |
2b0143b5 | 1788 | return proc_inode_is_dead(d_inode(dentry)); |
1dd704b6 DH |
1789 | } |
1790 | ||
6b4e306a | 1791 | const struct dentry_operations pid_dentry_operations = |
28a6d671 EB |
1792 | { |
1793 | .d_revalidate = pid_revalidate, | |
1794 | .d_delete = pid_delete_dentry, | |
1795 | }; | |
1796 | ||
1797 | /* Lookups */ | |
1798 | ||
1c0d04c9 EB |
1799 | /* |
1800 | * Fill a directory entry. | |
1801 | * | |
1802 | * If possible create the dcache entry and derive our inode number and | |
1803 | * file type from dcache entry. | |
1804 | * | |
1805 | * Since all of the proc inode numbers are dynamically generated, the inode | |
1806 | * numbers do not exist until the inode is cache. This means creating the | |
1807 | * the dcache entry in readdir is necessary to keep the inode numbers | |
1808 | * reported by readdir in sync with the inode numbers reported | |
1809 | * by stat. | |
1810 | */ | |
f0c3b509 | 1811 | bool proc_fill_cache(struct file *file, struct dir_context *ctx, |
6b4e306a | 1812 | const char *name, int len, |
c5141e6d | 1813 | instantiate_t instantiate, struct task_struct *task, const void *ptr) |
61a28784 | 1814 | { |
f0c3b509 | 1815 | struct dentry *child, *dir = file->f_path.dentry; |
1df98b8b | 1816 | struct qstr qname = QSTR_INIT(name, len); |
61a28784 | 1817 | struct inode *inode; |
1df98b8b AV |
1818 | unsigned type; |
1819 | ino_t ino; | |
61a28784 | 1820 | |
1df98b8b | 1821 | child = d_hash_and_lookup(dir, &qname); |
61a28784 | 1822 | if (!child) { |
3781764b AV |
1823 | DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq); |
1824 | child = d_alloc_parallel(dir, &qname, &wq); | |
1825 | if (IS_ERR(child)) | |
1df98b8b | 1826 | goto end_instantiate; |
3781764b AV |
1827 | if (d_in_lookup(child)) { |
1828 | int err = instantiate(d_inode(dir), child, task, ptr); | |
1829 | d_lookup_done(child); | |
1830 | if (err < 0) { | |
1831 | dput(child); | |
1832 | goto end_instantiate; | |
1833 | } | |
61a28784 EB |
1834 | } |
1835 | } | |
2b0143b5 | 1836 | inode = d_inode(child); |
147ce699 AV |
1837 | ino = inode->i_ino; |
1838 | type = inode->i_mode >> 12; | |
61a28784 | 1839 | dput(child); |
f0c3b509 | 1840 | return dir_emit(ctx, name, len, ino, type); |
1df98b8b AV |
1841 | |
1842 | end_instantiate: | |
1843 | return dir_emit(ctx, name, len, 1, DT_UNKNOWN); | |
61a28784 EB |
1844 | } |
1845 | ||
640708a2 PE |
1846 | /* |
1847 | * dname_to_vma_addr - maps a dentry name into two unsigned longs | |
1848 | * which represent vma start and end addresses. | |
1849 | */ | |
1850 | static int dname_to_vma_addr(struct dentry *dentry, | |
1851 | unsigned long *start, unsigned long *end) | |
1852 | { | |
1853 | if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2) | |
1854 | return -EINVAL; | |
1855 | ||
1856 | return 0; | |
1857 | } | |
1858 | ||
0b728e19 | 1859 | static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags) |
640708a2 PE |
1860 | { |
1861 | unsigned long vm_start, vm_end; | |
1862 | bool exact_vma_exists = false; | |
1863 | struct mm_struct *mm = NULL; | |
1864 | struct task_struct *task; | |
1865 | const struct cred *cred; | |
1866 | struct inode *inode; | |
1867 | int status = 0; | |
1868 | ||
0b728e19 | 1869 | if (flags & LOOKUP_RCU) |
640708a2 PE |
1870 | return -ECHILD; |
1871 | ||
2b0143b5 | 1872 | inode = d_inode(dentry); |
640708a2 PE |
1873 | task = get_proc_task(inode); |
1874 | if (!task) | |
1875 | goto out_notask; | |
1876 | ||
caaee623 | 1877 | mm = mm_access(task, PTRACE_MODE_READ_FSCREDS); |
2344bec7 | 1878 | if (IS_ERR_OR_NULL(mm)) |
640708a2 PE |
1879 | goto out; |
1880 | ||
1881 | if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) { | |
1882 | down_read(&mm->mmap_sem); | |
1883 | exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end); | |
1884 | up_read(&mm->mmap_sem); | |
1885 | } | |
1886 | ||
1887 | mmput(mm); | |
1888 | ||
1889 | if (exact_vma_exists) { | |
1890 | if (task_dumpable(task)) { | |
1891 | rcu_read_lock(); | |
1892 | cred = __task_cred(task); | |
1893 | inode->i_uid = cred->euid; | |
1894 | inode->i_gid = cred->egid; | |
1895 | rcu_read_unlock(); | |
1896 | } else { | |
dcb0f222 EB |
1897 | inode->i_uid = GLOBAL_ROOT_UID; |
1898 | inode->i_gid = GLOBAL_ROOT_GID; | |
640708a2 PE |
1899 | } |
1900 | security_task_to_inode(task, inode); | |
1901 | status = 1; | |
1902 | } | |
1903 | ||
1904 | out: | |
1905 | put_task_struct(task); | |
1906 | ||
1907 | out_notask: | |
640708a2 PE |
1908 | return status; |
1909 | } | |
1910 | ||
1911 | static const struct dentry_operations tid_map_files_dentry_operations = { | |
1912 | .d_revalidate = map_files_d_revalidate, | |
1913 | .d_delete = pid_delete_dentry, | |
1914 | }; | |
1915 | ||
6b255391 | 1916 | static int map_files_get_link(struct dentry *dentry, struct path *path) |
640708a2 PE |
1917 | { |
1918 | unsigned long vm_start, vm_end; | |
1919 | struct vm_area_struct *vma; | |
1920 | struct task_struct *task; | |
1921 | struct mm_struct *mm; | |
1922 | int rc; | |
1923 | ||
1924 | rc = -ENOENT; | |
2b0143b5 | 1925 | task = get_proc_task(d_inode(dentry)); |
640708a2 PE |
1926 | if (!task) |
1927 | goto out; | |
1928 | ||
1929 | mm = get_task_mm(task); | |
1930 | put_task_struct(task); | |
1931 | if (!mm) | |
1932 | goto out; | |
1933 | ||
1934 | rc = dname_to_vma_addr(dentry, &vm_start, &vm_end); | |
1935 | if (rc) | |
1936 | goto out_mmput; | |
1937 | ||
70335abb | 1938 | rc = -ENOENT; |
640708a2 PE |
1939 | down_read(&mm->mmap_sem); |
1940 | vma = find_exact_vma(mm, vm_start, vm_end); | |
1941 | if (vma && vma->vm_file) { | |
1942 | *path = vma->vm_file->f_path; | |
1943 | path_get(path); | |
1944 | rc = 0; | |
1945 | } | |
1946 | up_read(&mm->mmap_sem); | |
1947 | ||
1948 | out_mmput: | |
1949 | mmput(mm); | |
1950 | out: | |
1951 | return rc; | |
1952 | } | |
1953 | ||
1954 | struct map_files_info { | |
7b540d06 | 1955 | fmode_t mode; |
640708a2 PE |
1956 | unsigned long len; |
1957 | unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */ | |
1958 | }; | |
1959 | ||
bdb4d100 CO |
1960 | /* |
1961 | * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the | |
1962 | * symlinks may be used to bypass permissions on ancestor directories in the | |
1963 | * path to the file in question. | |
1964 | */ | |
1965 | static const char * | |
6b255391 | 1966 | proc_map_files_get_link(struct dentry *dentry, |
fceef393 AV |
1967 | struct inode *inode, |
1968 | struct delayed_call *done) | |
bdb4d100 CO |
1969 | { |
1970 | if (!capable(CAP_SYS_ADMIN)) | |
1971 | return ERR_PTR(-EPERM); | |
1972 | ||
fceef393 | 1973 | return proc_pid_get_link(dentry, inode, done); |
bdb4d100 CO |
1974 | } |
1975 | ||
1976 | /* | |
6b255391 | 1977 | * Identical to proc_pid_link_inode_operations except for get_link() |
bdb4d100 CO |
1978 | */ |
1979 | static const struct inode_operations proc_map_files_link_inode_operations = { | |
1980 | .readlink = proc_pid_readlink, | |
6b255391 | 1981 | .get_link = proc_map_files_get_link, |
bdb4d100 CO |
1982 | .setattr = proc_setattr, |
1983 | }; | |
1984 | ||
c52a47ac | 1985 | static int |
640708a2 PE |
1986 | proc_map_files_instantiate(struct inode *dir, struct dentry *dentry, |
1987 | struct task_struct *task, const void *ptr) | |
1988 | { | |
7b540d06 | 1989 | fmode_t mode = (fmode_t)(unsigned long)ptr; |
640708a2 PE |
1990 | struct proc_inode *ei; |
1991 | struct inode *inode; | |
1992 | ||
640708a2 PE |
1993 | inode = proc_pid_make_inode(dir->i_sb, task); |
1994 | if (!inode) | |
c52a47ac | 1995 | return -ENOENT; |
640708a2 PE |
1996 | |
1997 | ei = PROC_I(inode); | |
6b255391 | 1998 | ei->op.proc_get_link = map_files_get_link; |
640708a2 | 1999 | |
bdb4d100 | 2000 | inode->i_op = &proc_map_files_link_inode_operations; |
640708a2 PE |
2001 | inode->i_size = 64; |
2002 | inode->i_mode = S_IFLNK; | |
2003 | ||
7b540d06 | 2004 | if (mode & FMODE_READ) |
640708a2 | 2005 | inode->i_mode |= S_IRUSR; |
7b540d06 | 2006 | if (mode & FMODE_WRITE) |
640708a2 PE |
2007 | inode->i_mode |= S_IWUSR; |
2008 | ||
2009 | d_set_d_op(dentry, &tid_map_files_dentry_operations); | |
2010 | d_add(dentry, inode); | |
2011 | ||
c52a47ac | 2012 | return 0; |
640708a2 PE |
2013 | } |
2014 | ||
2015 | static struct dentry *proc_map_files_lookup(struct inode *dir, | |
00cd8dd3 | 2016 | struct dentry *dentry, unsigned int flags) |
640708a2 PE |
2017 | { |
2018 | unsigned long vm_start, vm_end; | |
2019 | struct vm_area_struct *vma; | |
2020 | struct task_struct *task; | |
c52a47ac | 2021 | int result; |
640708a2 PE |
2022 | struct mm_struct *mm; |
2023 | ||
c52a47ac | 2024 | result = -ENOENT; |
640708a2 PE |
2025 | task = get_proc_task(dir); |
2026 | if (!task) | |
2027 | goto out; | |
2028 | ||
c52a47ac | 2029 | result = -EACCES; |
caaee623 | 2030 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
640708a2 PE |
2031 | goto out_put_task; |
2032 | ||
c52a47ac | 2033 | result = -ENOENT; |
640708a2 | 2034 | if (dname_to_vma_addr(dentry, &vm_start, &vm_end)) |
eb94cd96 | 2035 | goto out_put_task; |
640708a2 PE |
2036 | |
2037 | mm = get_task_mm(task); | |
2038 | if (!mm) | |
eb94cd96 | 2039 | goto out_put_task; |
640708a2 PE |
2040 | |
2041 | down_read(&mm->mmap_sem); | |
2042 | vma = find_exact_vma(mm, vm_start, vm_end); | |
2043 | if (!vma) | |
2044 | goto out_no_vma; | |
2045 | ||
05f56484 SK |
2046 | if (vma->vm_file) |
2047 | result = proc_map_files_instantiate(dir, dentry, task, | |
2048 | (void *)(unsigned long)vma->vm_file->f_mode); | |
640708a2 PE |
2049 | |
2050 | out_no_vma: | |
2051 | up_read(&mm->mmap_sem); | |
2052 | mmput(mm); | |
640708a2 PE |
2053 | out_put_task: |
2054 | put_task_struct(task); | |
2055 | out: | |
c52a47ac | 2056 | return ERR_PTR(result); |
640708a2 PE |
2057 | } |
2058 | ||
2059 | static const struct inode_operations proc_map_files_inode_operations = { | |
2060 | .lookup = proc_map_files_lookup, | |
2061 | .permission = proc_fd_permission, | |
2062 | .setattr = proc_setattr, | |
2063 | }; | |
2064 | ||
2065 | static int | |
f0c3b509 | 2066 | proc_map_files_readdir(struct file *file, struct dir_context *ctx) |
640708a2 | 2067 | { |
640708a2 PE |
2068 | struct vm_area_struct *vma; |
2069 | struct task_struct *task; | |
2070 | struct mm_struct *mm; | |
f0c3b509 AV |
2071 | unsigned long nr_files, pos, i; |
2072 | struct flex_array *fa = NULL; | |
2073 | struct map_files_info info; | |
2074 | struct map_files_info *p; | |
640708a2 PE |
2075 | int ret; |
2076 | ||
640708a2 | 2077 | ret = -ENOENT; |
f0c3b509 | 2078 | task = get_proc_task(file_inode(file)); |
640708a2 PE |
2079 | if (!task) |
2080 | goto out; | |
2081 | ||
2082 | ret = -EACCES; | |
caaee623 | 2083 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
640708a2 PE |
2084 | goto out_put_task; |
2085 | ||
2086 | ret = 0; | |
f0c3b509 AV |
2087 | if (!dir_emit_dots(file, ctx)) |
2088 | goto out_put_task; | |
640708a2 | 2089 | |
f0c3b509 AV |
2090 | mm = get_task_mm(task); |
2091 | if (!mm) | |
2092 | goto out_put_task; | |
2093 | down_read(&mm->mmap_sem); | |
640708a2 | 2094 | |
f0c3b509 | 2095 | nr_files = 0; |
640708a2 | 2096 | |
f0c3b509 AV |
2097 | /* |
2098 | * We need two passes here: | |
2099 | * | |
2100 | * 1) Collect vmas of mapped files with mmap_sem taken | |
2101 | * 2) Release mmap_sem and instantiate entries | |
2102 | * | |
2103 | * otherwise we get lockdep complained, since filldir() | |
2104 | * routine might require mmap_sem taken in might_fault(). | |
2105 | */ | |
640708a2 | 2106 | |
f0c3b509 AV |
2107 | for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) { |
2108 | if (vma->vm_file && ++pos > ctx->pos) | |
2109 | nr_files++; | |
2110 | } | |
2111 | ||
2112 | if (nr_files) { | |
2113 | fa = flex_array_alloc(sizeof(info), nr_files, | |
2114 | GFP_KERNEL); | |
2115 | if (!fa || flex_array_prealloc(fa, 0, nr_files, | |
2116 | GFP_KERNEL)) { | |
2117 | ret = -ENOMEM; | |
2118 | if (fa) | |
2119 | flex_array_free(fa); | |
2120 | up_read(&mm->mmap_sem); | |
2121 | mmput(mm); | |
2122 | goto out_put_task; | |
640708a2 | 2123 | } |
f0c3b509 AV |
2124 | for (i = 0, vma = mm->mmap, pos = 2; vma; |
2125 | vma = vma->vm_next) { | |
2126 | if (!vma->vm_file) | |
2127 | continue; | |
2128 | if (++pos <= ctx->pos) | |
2129 | continue; | |
2130 | ||
2131 | info.mode = vma->vm_file->f_mode; | |
2132 | info.len = snprintf(info.name, | |
2133 | sizeof(info.name), "%lx-%lx", | |
2134 | vma->vm_start, vma->vm_end); | |
2135 | if (flex_array_put(fa, i++, &info, GFP_KERNEL)) | |
2136 | BUG(); | |
640708a2 | 2137 | } |
640708a2 | 2138 | } |
f0c3b509 AV |
2139 | up_read(&mm->mmap_sem); |
2140 | ||
2141 | for (i = 0; i < nr_files; i++) { | |
2142 | p = flex_array_get(fa, i); | |
2143 | if (!proc_fill_cache(file, ctx, | |
2144 | p->name, p->len, | |
2145 | proc_map_files_instantiate, | |
2146 | task, | |
2147 | (void *)(unsigned long)p->mode)) | |
2148 | break; | |
2149 | ctx->pos++; | |
640708a2 | 2150 | } |
f0c3b509 AV |
2151 | if (fa) |
2152 | flex_array_free(fa); | |
2153 | mmput(mm); | |
640708a2 | 2154 | |
640708a2 PE |
2155 | out_put_task: |
2156 | put_task_struct(task); | |
2157 | out: | |
2158 | return ret; | |
2159 | } | |
2160 | ||
2161 | static const struct file_operations proc_map_files_operations = { | |
2162 | .read = generic_read_dir, | |
f50752ea AV |
2163 | .iterate_shared = proc_map_files_readdir, |
2164 | .llseek = generic_file_llseek, | |
640708a2 PE |
2165 | }; |
2166 | ||
b5946bea | 2167 | #ifdef CONFIG_CHECKPOINT_RESTORE |
48f6a7a5 PE |
2168 | struct timers_private { |
2169 | struct pid *pid; | |
2170 | struct task_struct *task; | |
2171 | struct sighand_struct *sighand; | |
57b8015e | 2172 | struct pid_namespace *ns; |
48f6a7a5 PE |
2173 | unsigned long flags; |
2174 | }; | |
2175 | ||
2176 | static void *timers_start(struct seq_file *m, loff_t *pos) | |
2177 | { | |
2178 | struct timers_private *tp = m->private; | |
2179 | ||
2180 | tp->task = get_pid_task(tp->pid, PIDTYPE_PID); | |
2181 | if (!tp->task) | |
2182 | return ERR_PTR(-ESRCH); | |
2183 | ||
2184 | tp->sighand = lock_task_sighand(tp->task, &tp->flags); | |
2185 | if (!tp->sighand) | |
2186 | return ERR_PTR(-ESRCH); | |
2187 | ||
2188 | return seq_list_start(&tp->task->signal->posix_timers, *pos); | |
2189 | } | |
2190 | ||
2191 | static void *timers_next(struct seq_file *m, void *v, loff_t *pos) | |
2192 | { | |
2193 | struct timers_private *tp = m->private; | |
2194 | return seq_list_next(v, &tp->task->signal->posix_timers, pos); | |
2195 | } | |
2196 | ||
2197 | static void timers_stop(struct seq_file *m, void *v) | |
2198 | { | |
2199 | struct timers_private *tp = m->private; | |
2200 | ||
2201 | if (tp->sighand) { | |
2202 | unlock_task_sighand(tp->task, &tp->flags); | |
2203 | tp->sighand = NULL; | |
2204 | } | |
2205 | ||
2206 | if (tp->task) { | |
2207 | put_task_struct(tp->task); | |
2208 | tp->task = NULL; | |
2209 | } | |
2210 | } | |
2211 | ||
2212 | static int show_timer(struct seq_file *m, void *v) | |
2213 | { | |
2214 | struct k_itimer *timer; | |
57b8015e PE |
2215 | struct timers_private *tp = m->private; |
2216 | int notify; | |
cedbccab | 2217 | static const char * const nstr[] = { |
57b8015e PE |
2218 | [SIGEV_SIGNAL] = "signal", |
2219 | [SIGEV_NONE] = "none", | |
2220 | [SIGEV_THREAD] = "thread", | |
2221 | }; | |
48f6a7a5 PE |
2222 | |
2223 | timer = list_entry((struct list_head *)v, struct k_itimer, list); | |
57b8015e PE |
2224 | notify = timer->it_sigev_notify; |
2225 | ||
48f6a7a5 | 2226 | seq_printf(m, "ID: %d\n", timer->it_id); |
25ce3191 JP |
2227 | seq_printf(m, "signal: %d/%p\n", |
2228 | timer->sigq->info.si_signo, | |
2229 | timer->sigq->info.si_value.sival_ptr); | |
57b8015e | 2230 | seq_printf(m, "notify: %s/%s.%d\n", |
25ce3191 JP |
2231 | nstr[notify & ~SIGEV_THREAD_ID], |
2232 | (notify & SIGEV_THREAD_ID) ? "tid" : "pid", | |
2233 | pid_nr_ns(timer->it_pid, tp->ns)); | |
15ef0298 | 2234 | seq_printf(m, "ClockID: %d\n", timer->it_clock); |
48f6a7a5 PE |
2235 | |
2236 | return 0; | |
2237 | } | |
2238 | ||
2239 | static const struct seq_operations proc_timers_seq_ops = { | |
2240 | .start = timers_start, | |
2241 | .next = timers_next, | |
2242 | .stop = timers_stop, | |
2243 | .show = show_timer, | |
2244 | }; | |
2245 | ||
2246 | static int proc_timers_open(struct inode *inode, struct file *file) | |
2247 | { | |
2248 | struct timers_private *tp; | |
2249 | ||
2250 | tp = __seq_open_private(file, &proc_timers_seq_ops, | |
2251 | sizeof(struct timers_private)); | |
2252 | if (!tp) | |
2253 | return -ENOMEM; | |
2254 | ||
2255 | tp->pid = proc_pid(inode); | |
57b8015e | 2256 | tp->ns = inode->i_sb->s_fs_info; |
48f6a7a5 PE |
2257 | return 0; |
2258 | } | |
2259 | ||
2260 | static const struct file_operations proc_timers_operations = { | |
2261 | .open = proc_timers_open, | |
2262 | .read = seq_read, | |
2263 | .llseek = seq_lseek, | |
2264 | .release = seq_release_private, | |
2265 | }; | |
b5946bea | 2266 | #endif |
640708a2 | 2267 | |
5de23d43 JS |
2268 | static ssize_t timerslack_ns_write(struct file *file, const char __user *buf, |
2269 | size_t count, loff_t *offset) | |
2270 | { | |
2271 | struct inode *inode = file_inode(file); | |
2272 | struct task_struct *p; | |
2273 | u64 slack_ns; | |
2274 | int err; | |
2275 | ||
2276 | err = kstrtoull_from_user(buf, count, 10, &slack_ns); | |
2277 | if (err < 0) | |
2278 | return err; | |
2279 | ||
2280 | p = get_proc_task(inode); | |
2281 | if (!p) | |
2282 | return -ESRCH; | |
2283 | ||
2284 | if (ptrace_may_access(p, PTRACE_MODE_ATTACH_FSCREDS)) { | |
2285 | task_lock(p); | |
2286 | if (slack_ns == 0) | |
2287 | p->timer_slack_ns = p->default_timer_slack_ns; | |
2288 | else | |
2289 | p->timer_slack_ns = slack_ns; | |
2290 | task_unlock(p); | |
2291 | } else | |
2292 | count = -EPERM; | |
2293 | ||
2294 | put_task_struct(p); | |
2295 | ||
2296 | return count; | |
2297 | } | |
2298 | ||
2299 | static int timerslack_ns_show(struct seq_file *m, void *v) | |
2300 | { | |
2301 | struct inode *inode = m->private; | |
2302 | struct task_struct *p; | |
2303 | int err = 0; | |
2304 | ||
2305 | p = get_proc_task(inode); | |
2306 | if (!p) | |
2307 | return -ESRCH; | |
2308 | ||
2309 | if (ptrace_may_access(p, PTRACE_MODE_ATTACH_FSCREDS)) { | |
2310 | task_lock(p); | |
2311 | seq_printf(m, "%llu\n", p->timer_slack_ns); | |
2312 | task_unlock(p); | |
2313 | } else | |
2314 | err = -EPERM; | |
2315 | ||
2316 | put_task_struct(p); | |
2317 | ||
2318 | return err; | |
2319 | } | |
2320 | ||
2321 | static int timerslack_ns_open(struct inode *inode, struct file *filp) | |
2322 | { | |
2323 | return single_open(filp, timerslack_ns_show, inode); | |
2324 | } | |
2325 | ||
2326 | static const struct file_operations proc_pid_set_timerslack_ns_operations = { | |
2327 | .open = timerslack_ns_open, | |
2328 | .read = seq_read, | |
2329 | .write = timerslack_ns_write, | |
2330 | .llseek = seq_lseek, | |
2331 | .release = single_release, | |
2332 | }; | |
2333 | ||
c52a47ac | 2334 | static int proc_pident_instantiate(struct inode *dir, |
c5141e6d | 2335 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
444ceed8 | 2336 | { |
c5141e6d | 2337 | const struct pid_entry *p = ptr; |
444ceed8 EB |
2338 | struct inode *inode; |
2339 | struct proc_inode *ei; | |
444ceed8 | 2340 | |
61a28784 | 2341 | inode = proc_pid_make_inode(dir->i_sb, task); |
444ceed8 EB |
2342 | if (!inode) |
2343 | goto out; | |
2344 | ||
2345 | ei = PROC_I(inode); | |
2346 | inode->i_mode = p->mode; | |
2347 | if (S_ISDIR(inode->i_mode)) | |
bfe86848 | 2348 | set_nlink(inode, 2); /* Use getattr to fix if necessary */ |
444ceed8 EB |
2349 | if (p->iop) |
2350 | inode->i_op = p->iop; | |
2351 | if (p->fop) | |
2352 | inode->i_fop = p->fop; | |
2353 | ei->op = p->op; | |
fb045adb | 2354 | d_set_d_op(dentry, &pid_dentry_operations); |
444ceed8 EB |
2355 | d_add(dentry, inode); |
2356 | /* Close the race of the process dying before we return the dentry */ | |
0b728e19 | 2357 | if (pid_revalidate(dentry, 0)) |
c52a47ac | 2358 | return 0; |
444ceed8 | 2359 | out: |
c52a47ac | 2360 | return -ENOENT; |
444ceed8 EB |
2361 | } |
2362 | ||
1da177e4 LT |
2363 | static struct dentry *proc_pident_lookup(struct inode *dir, |
2364 | struct dentry *dentry, | |
c5141e6d | 2365 | const struct pid_entry *ents, |
7bcd6b0e | 2366 | unsigned int nents) |
1da177e4 | 2367 | { |
c52a47ac | 2368 | int error; |
99f89551 | 2369 | struct task_struct *task = get_proc_task(dir); |
c5141e6d | 2370 | const struct pid_entry *p, *last; |
1da177e4 | 2371 | |
c52a47ac | 2372 | error = -ENOENT; |
1da177e4 | 2373 | |
99f89551 EB |
2374 | if (!task) |
2375 | goto out_no_task; | |
1da177e4 | 2376 | |
20cdc894 EB |
2377 | /* |
2378 | * Yes, it does not scale. And it should not. Don't add | |
2379 | * new entries into /proc/<tgid>/ without very good reasons. | |
2380 | */ | |
7bcd6b0e EB |
2381 | last = &ents[nents - 1]; |
2382 | for (p = ents; p <= last; p++) { | |
1da177e4 LT |
2383 | if (p->len != dentry->d_name.len) |
2384 | continue; | |
2385 | if (!memcmp(dentry->d_name.name, p->name, p->len)) | |
2386 | break; | |
2387 | } | |
7bcd6b0e | 2388 | if (p > last) |
1da177e4 LT |
2389 | goto out; |
2390 | ||
444ceed8 | 2391 | error = proc_pident_instantiate(dir, dentry, task, p); |
1da177e4 | 2392 | out: |
99f89551 EB |
2393 | put_task_struct(task); |
2394 | out_no_task: | |
c52a47ac | 2395 | return ERR_PTR(error); |
1da177e4 LT |
2396 | } |
2397 | ||
f0c3b509 | 2398 | static int proc_pident_readdir(struct file *file, struct dir_context *ctx, |
c5141e6d | 2399 | const struct pid_entry *ents, unsigned int nents) |
28a6d671 | 2400 | { |
f0c3b509 AV |
2401 | struct task_struct *task = get_proc_task(file_inode(file)); |
2402 | const struct pid_entry *p; | |
28a6d671 | 2403 | |
28a6d671 | 2404 | if (!task) |
f0c3b509 | 2405 | return -ENOENT; |
28a6d671 | 2406 | |
f0c3b509 AV |
2407 | if (!dir_emit_dots(file, ctx)) |
2408 | goto out; | |
2409 | ||
2410 | if (ctx->pos >= nents + 2) | |
2411 | goto out; | |
28a6d671 | 2412 | |
f0c3b509 AV |
2413 | for (p = ents + (ctx->pos - 2); p <= ents + nents - 1; p++) { |
2414 | if (!proc_fill_cache(file, ctx, p->name, p->len, | |
2415 | proc_pident_instantiate, task, p)) | |
2416 | break; | |
2417 | ctx->pos++; | |
2418 | } | |
28a6d671 | 2419 | out: |
61a28784 | 2420 | put_task_struct(task); |
f0c3b509 | 2421 | return 0; |
1da177e4 LT |
2422 | } |
2423 | ||
28a6d671 EB |
2424 | #ifdef CONFIG_SECURITY |
2425 | static ssize_t proc_pid_attr_read(struct file * file, char __user * buf, | |
2426 | size_t count, loff_t *ppos) | |
2427 | { | |
496ad9aa | 2428 | struct inode * inode = file_inode(file); |
04ff9708 | 2429 | char *p = NULL; |
28a6d671 EB |
2430 | ssize_t length; |
2431 | struct task_struct *task = get_proc_task(inode); | |
2432 | ||
28a6d671 | 2433 | if (!task) |
04ff9708 | 2434 | return -ESRCH; |
28a6d671 EB |
2435 | |
2436 | length = security_getprocattr(task, | |
2fddfeef | 2437 | (char*)file->f_path.dentry->d_name.name, |
04ff9708 | 2438 | &p); |
28a6d671 | 2439 | put_task_struct(task); |
04ff9708 AV |
2440 | if (length > 0) |
2441 | length = simple_read_from_buffer(buf, count, ppos, p, length); | |
2442 | kfree(p); | |
28a6d671 | 2443 | return length; |
1da177e4 LT |
2444 | } |
2445 | ||
28a6d671 EB |
2446 | static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf, |
2447 | size_t count, loff_t *ppos) | |
2448 | { | |
496ad9aa | 2449 | struct inode * inode = file_inode(file); |
bb646cdb | 2450 | void *page; |
28a6d671 EB |
2451 | ssize_t length; |
2452 | struct task_struct *task = get_proc_task(inode); | |
2453 | ||
2454 | length = -ESRCH; | |
2455 | if (!task) | |
2456 | goto out_no_task; | |
2457 | if (count > PAGE_SIZE) | |
2458 | count = PAGE_SIZE; | |
2459 | ||
2460 | /* No partial writes. */ | |
2461 | length = -EINVAL; | |
2462 | if (*ppos != 0) | |
2463 | goto out; | |
2464 | ||
bb646cdb AV |
2465 | page = memdup_user(buf, count); |
2466 | if (IS_ERR(page)) { | |
2467 | length = PTR_ERR(page); | |
28a6d671 | 2468 | goto out; |
bb646cdb | 2469 | } |
28a6d671 | 2470 | |
107db7c7 | 2471 | /* Guard against adverse ptrace interaction */ |
9b1bf12d | 2472 | length = mutex_lock_interruptible(&task->signal->cred_guard_mutex); |
107db7c7 DH |
2473 | if (length < 0) |
2474 | goto out_free; | |
2475 | ||
28a6d671 | 2476 | length = security_setprocattr(task, |
2fddfeef | 2477 | (char*)file->f_path.dentry->d_name.name, |
bb646cdb | 2478 | page, count); |
9b1bf12d | 2479 | mutex_unlock(&task->signal->cred_guard_mutex); |
28a6d671 | 2480 | out_free: |
bb646cdb | 2481 | kfree(page); |
28a6d671 EB |
2482 | out: |
2483 | put_task_struct(task); | |
2484 | out_no_task: | |
2485 | return length; | |
2486 | } | |
2487 | ||
00977a59 | 2488 | static const struct file_operations proc_pid_attr_operations = { |
28a6d671 EB |
2489 | .read = proc_pid_attr_read, |
2490 | .write = proc_pid_attr_write, | |
87df8424 | 2491 | .llseek = generic_file_llseek, |
28a6d671 EB |
2492 | }; |
2493 | ||
c5141e6d | 2494 | static const struct pid_entry attr_dir_stuff[] = { |
631f9c18 AD |
2495 | REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations), |
2496 | REG("prev", S_IRUGO, proc_pid_attr_operations), | |
2497 | REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2498 | REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2499 | REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2500 | REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
28a6d671 EB |
2501 | }; |
2502 | ||
f0c3b509 | 2503 | static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx) |
28a6d671 | 2504 | { |
f0c3b509 AV |
2505 | return proc_pident_readdir(file, ctx, |
2506 | attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2507 | } |
2508 | ||
00977a59 | 2509 | static const struct file_operations proc_attr_dir_operations = { |
1da177e4 | 2510 | .read = generic_read_dir, |
f50752ea AV |
2511 | .iterate_shared = proc_attr_dir_readdir, |
2512 | .llseek = generic_file_llseek, | |
1da177e4 LT |
2513 | }; |
2514 | ||
72d9dcfc | 2515 | static struct dentry *proc_attr_dir_lookup(struct inode *dir, |
00cd8dd3 | 2516 | struct dentry *dentry, unsigned int flags) |
28a6d671 | 2517 | { |
7bcd6b0e EB |
2518 | return proc_pident_lookup(dir, dentry, |
2519 | attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2520 | } |
2521 | ||
c5ef1c42 | 2522 | static const struct inode_operations proc_attr_dir_inode_operations = { |
72d9dcfc | 2523 | .lookup = proc_attr_dir_lookup, |
99f89551 | 2524 | .getattr = pid_getattr, |
6d76fa58 | 2525 | .setattr = proc_setattr, |
1da177e4 LT |
2526 | }; |
2527 | ||
28a6d671 EB |
2528 | #endif |
2529 | ||
698ba7b5 | 2530 | #ifdef CONFIG_ELF_CORE |
3cb4a0bb KH |
2531 | static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf, |
2532 | size_t count, loff_t *ppos) | |
2533 | { | |
496ad9aa | 2534 | struct task_struct *task = get_proc_task(file_inode(file)); |
3cb4a0bb KH |
2535 | struct mm_struct *mm; |
2536 | char buffer[PROC_NUMBUF]; | |
2537 | size_t len; | |
2538 | int ret; | |
2539 | ||
2540 | if (!task) | |
2541 | return -ESRCH; | |
2542 | ||
2543 | ret = 0; | |
2544 | mm = get_task_mm(task); | |
2545 | if (mm) { | |
2546 | len = snprintf(buffer, sizeof(buffer), "%08lx\n", | |
2547 | ((mm->flags & MMF_DUMP_FILTER_MASK) >> | |
2548 | MMF_DUMP_FILTER_SHIFT)); | |
2549 | mmput(mm); | |
2550 | ret = simple_read_from_buffer(buf, count, ppos, buffer, len); | |
2551 | } | |
2552 | ||
2553 | put_task_struct(task); | |
2554 | ||
2555 | return ret; | |
2556 | } | |
2557 | ||
2558 | static ssize_t proc_coredump_filter_write(struct file *file, | |
2559 | const char __user *buf, | |
2560 | size_t count, | |
2561 | loff_t *ppos) | |
2562 | { | |
2563 | struct task_struct *task; | |
2564 | struct mm_struct *mm; | |
3cb4a0bb KH |
2565 | unsigned int val; |
2566 | int ret; | |
2567 | int i; | |
2568 | unsigned long mask; | |
2569 | ||
774636e1 AD |
2570 | ret = kstrtouint_from_user(buf, count, 0, &val); |
2571 | if (ret < 0) | |
2572 | return ret; | |
3cb4a0bb KH |
2573 | |
2574 | ret = -ESRCH; | |
496ad9aa | 2575 | task = get_proc_task(file_inode(file)); |
3cb4a0bb KH |
2576 | if (!task) |
2577 | goto out_no_task; | |
2578 | ||
3cb4a0bb KH |
2579 | mm = get_task_mm(task); |
2580 | if (!mm) | |
2581 | goto out_no_mm; | |
41a0c249 | 2582 | ret = 0; |
3cb4a0bb KH |
2583 | |
2584 | for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) { | |
2585 | if (val & mask) | |
2586 | set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2587 | else | |
2588 | clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2589 | } | |
2590 | ||
2591 | mmput(mm); | |
2592 | out_no_mm: | |
2593 | put_task_struct(task); | |
2594 | out_no_task: | |
774636e1 AD |
2595 | if (ret < 0) |
2596 | return ret; | |
2597 | return count; | |
3cb4a0bb KH |
2598 | } |
2599 | ||
2600 | static const struct file_operations proc_coredump_filter_operations = { | |
2601 | .read = proc_coredump_filter_read, | |
2602 | .write = proc_coredump_filter_write, | |
87df8424 | 2603 | .llseek = generic_file_llseek, |
3cb4a0bb KH |
2604 | }; |
2605 | #endif | |
2606 | ||
aba76fdb | 2607 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 2608 | static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole) |
297c5d92 | 2609 | { |
940389b8 | 2610 | struct task_io_accounting acct = task->ioac; |
5995477a | 2611 | unsigned long flags; |
293eb1e7 | 2612 | int result; |
5995477a | 2613 | |
293eb1e7 VK |
2614 | result = mutex_lock_killable(&task->signal->cred_guard_mutex); |
2615 | if (result) | |
2616 | return result; | |
2617 | ||
caaee623 | 2618 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) { |
293eb1e7 VK |
2619 | result = -EACCES; |
2620 | goto out_unlock; | |
2621 | } | |
1d1221f3 | 2622 | |
5995477a AR |
2623 | if (whole && lock_task_sighand(task, &flags)) { |
2624 | struct task_struct *t = task; | |
2625 | ||
2626 | task_io_accounting_add(&acct, &task->signal->ioac); | |
2627 | while_each_thread(task, t) | |
2628 | task_io_accounting_add(&acct, &t->ioac); | |
2629 | ||
2630 | unlock_task_sighand(task, &flags); | |
297c5d92 | 2631 | } |
25ce3191 JP |
2632 | seq_printf(m, |
2633 | "rchar: %llu\n" | |
2634 | "wchar: %llu\n" | |
2635 | "syscr: %llu\n" | |
2636 | "syscw: %llu\n" | |
2637 | "read_bytes: %llu\n" | |
2638 | "write_bytes: %llu\n" | |
2639 | "cancelled_write_bytes: %llu\n", | |
2640 | (unsigned long long)acct.rchar, | |
2641 | (unsigned long long)acct.wchar, | |
2642 | (unsigned long long)acct.syscr, | |
2643 | (unsigned long long)acct.syscw, | |
2644 | (unsigned long long)acct.read_bytes, | |
2645 | (unsigned long long)acct.write_bytes, | |
2646 | (unsigned long long)acct.cancelled_write_bytes); | |
2647 | result = 0; | |
2648 | ||
293eb1e7 VK |
2649 | out_unlock: |
2650 | mutex_unlock(&task->signal->cred_guard_mutex); | |
2651 | return result; | |
297c5d92 AR |
2652 | } |
2653 | ||
19aadc98 AD |
2654 | static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns, |
2655 | struct pid *pid, struct task_struct *task) | |
297c5d92 | 2656 | { |
19aadc98 | 2657 | return do_io_accounting(task, m, 0); |
aba76fdb | 2658 | } |
297c5d92 | 2659 | |
19aadc98 AD |
2660 | static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns, |
2661 | struct pid *pid, struct task_struct *task) | |
297c5d92 | 2662 | { |
19aadc98 | 2663 | return do_io_accounting(task, m, 1); |
297c5d92 AR |
2664 | } |
2665 | #endif /* CONFIG_TASK_IO_ACCOUNTING */ | |
aba76fdb | 2666 | |
22d917d8 EB |
2667 | #ifdef CONFIG_USER_NS |
2668 | static int proc_id_map_open(struct inode *inode, struct file *file, | |
ccf94f1b | 2669 | const struct seq_operations *seq_ops) |
22d917d8 EB |
2670 | { |
2671 | struct user_namespace *ns = NULL; | |
2672 | struct task_struct *task; | |
2673 | struct seq_file *seq; | |
2674 | int ret = -EINVAL; | |
2675 | ||
2676 | task = get_proc_task(inode); | |
2677 | if (task) { | |
2678 | rcu_read_lock(); | |
2679 | ns = get_user_ns(task_cred_xxx(task, user_ns)); | |
2680 | rcu_read_unlock(); | |
2681 | put_task_struct(task); | |
2682 | } | |
2683 | if (!ns) | |
2684 | goto err; | |
2685 | ||
2686 | ret = seq_open(file, seq_ops); | |
2687 | if (ret) | |
2688 | goto err_put_ns; | |
2689 | ||
2690 | seq = file->private_data; | |
2691 | seq->private = ns; | |
2692 | ||
2693 | return 0; | |
2694 | err_put_ns: | |
2695 | put_user_ns(ns); | |
2696 | err: | |
2697 | return ret; | |
2698 | } | |
2699 | ||
2700 | static int proc_id_map_release(struct inode *inode, struct file *file) | |
2701 | { | |
2702 | struct seq_file *seq = file->private_data; | |
2703 | struct user_namespace *ns = seq->private; | |
2704 | put_user_ns(ns); | |
2705 | return seq_release(inode, file); | |
2706 | } | |
2707 | ||
2708 | static int proc_uid_map_open(struct inode *inode, struct file *file) | |
2709 | { | |
2710 | return proc_id_map_open(inode, file, &proc_uid_seq_operations); | |
2711 | } | |
2712 | ||
2713 | static int proc_gid_map_open(struct inode *inode, struct file *file) | |
2714 | { | |
2715 | return proc_id_map_open(inode, file, &proc_gid_seq_operations); | |
2716 | } | |
2717 | ||
f76d207a EB |
2718 | static int proc_projid_map_open(struct inode *inode, struct file *file) |
2719 | { | |
2720 | return proc_id_map_open(inode, file, &proc_projid_seq_operations); | |
2721 | } | |
2722 | ||
22d917d8 EB |
2723 | static const struct file_operations proc_uid_map_operations = { |
2724 | .open = proc_uid_map_open, | |
2725 | .write = proc_uid_map_write, | |
2726 | .read = seq_read, | |
2727 | .llseek = seq_lseek, | |
2728 | .release = proc_id_map_release, | |
2729 | }; | |
2730 | ||
2731 | static const struct file_operations proc_gid_map_operations = { | |
2732 | .open = proc_gid_map_open, | |
2733 | .write = proc_gid_map_write, | |
2734 | .read = seq_read, | |
2735 | .llseek = seq_lseek, | |
2736 | .release = proc_id_map_release, | |
2737 | }; | |
f76d207a EB |
2738 | |
2739 | static const struct file_operations proc_projid_map_operations = { | |
2740 | .open = proc_projid_map_open, | |
2741 | .write = proc_projid_map_write, | |
2742 | .read = seq_read, | |
2743 | .llseek = seq_lseek, | |
2744 | .release = proc_id_map_release, | |
2745 | }; | |
9cc46516 EB |
2746 | |
2747 | static int proc_setgroups_open(struct inode *inode, struct file *file) | |
2748 | { | |
2749 | struct user_namespace *ns = NULL; | |
2750 | struct task_struct *task; | |
2751 | int ret; | |
2752 | ||
2753 | ret = -ESRCH; | |
2754 | task = get_proc_task(inode); | |
2755 | if (task) { | |
2756 | rcu_read_lock(); | |
2757 | ns = get_user_ns(task_cred_xxx(task, user_ns)); | |
2758 | rcu_read_unlock(); | |
2759 | put_task_struct(task); | |
2760 | } | |
2761 | if (!ns) | |
2762 | goto err; | |
2763 | ||
2764 | if (file->f_mode & FMODE_WRITE) { | |
2765 | ret = -EACCES; | |
2766 | if (!ns_capable(ns, CAP_SYS_ADMIN)) | |
2767 | goto err_put_ns; | |
2768 | } | |
2769 | ||
2770 | ret = single_open(file, &proc_setgroups_show, ns); | |
2771 | if (ret) | |
2772 | goto err_put_ns; | |
2773 | ||
2774 | return 0; | |
2775 | err_put_ns: | |
2776 | put_user_ns(ns); | |
2777 | err: | |
2778 | return ret; | |
2779 | } | |
2780 | ||
2781 | static int proc_setgroups_release(struct inode *inode, struct file *file) | |
2782 | { | |
2783 | struct seq_file *seq = file->private_data; | |
2784 | struct user_namespace *ns = seq->private; | |
2785 | int ret = single_release(inode, file); | |
2786 | put_user_ns(ns); | |
2787 | return ret; | |
2788 | } | |
2789 | ||
2790 | static const struct file_operations proc_setgroups_operations = { | |
2791 | .open = proc_setgroups_open, | |
2792 | .write = proc_setgroups_write, | |
2793 | .read = seq_read, | |
2794 | .llseek = seq_lseek, | |
2795 | .release = proc_setgroups_release, | |
2796 | }; | |
22d917d8 EB |
2797 | #endif /* CONFIG_USER_NS */ |
2798 | ||
47830723 KC |
2799 | static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns, |
2800 | struct pid *pid, struct task_struct *task) | |
2801 | { | |
a9712bc1 AV |
2802 | int err = lock_trace(task); |
2803 | if (!err) { | |
2804 | seq_printf(m, "%08x\n", task->personality); | |
2805 | unlock_trace(task); | |
2806 | } | |
2807 | return err; | |
47830723 KC |
2808 | } |
2809 | ||
28a6d671 EB |
2810 | /* |
2811 | * Thread groups | |
2812 | */ | |
00977a59 | 2813 | static const struct file_operations proc_task_operations; |
c5ef1c42 | 2814 | static const struct inode_operations proc_task_inode_operations; |
20cdc894 | 2815 | |
c5141e6d | 2816 | static const struct pid_entry tgid_base_stuff[] = { |
631f9c18 AD |
2817 | DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations), |
2818 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), | |
640708a2 | 2819 | DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations), |
631f9c18 | 2820 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 2821 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
b2211a36 | 2822 | #ifdef CONFIG_NET |
631f9c18 | 2823 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), |
b2211a36 | 2824 | #endif |
631f9c18 | 2825 | REG("environ", S_IRUSR, proc_environ_operations), |
f9ea536e | 2826 | ONE("auxv", S_IRUSR, proc_pid_auxv), |
631f9c18 | 2827 | ONE("status", S_IRUGO, proc_pid_status), |
35a35046 | 2828 | ONE("personality", S_IRUSR, proc_pid_personality), |
1c963eb1 | 2829 | ONE("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 2830 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 2831 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
5091faa4 MG |
2832 | #endif |
2833 | #ifdef CONFIG_SCHED_AUTOGROUP | |
2834 | REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations), | |
ebcb6734 | 2835 | #endif |
4614a696 | 2836 | REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations), |
ebcb6734 | 2837 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 | 2838 | ONE("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 2839 | #endif |
c2c0bb44 | 2840 | REG("cmdline", S_IRUGO, proc_pid_cmdline_ops), |
631f9c18 AD |
2841 | ONE("stat", S_IRUGO, proc_tgid_stat), |
2842 | ONE("statm", S_IRUGO, proc_pid_statm), | |
b7643757 | 2843 | REG("maps", S_IRUGO, proc_pid_maps_operations), |
28a6d671 | 2844 | #ifdef CONFIG_NUMA |
b7643757 | 2845 | REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations), |
28a6d671 | 2846 | #endif |
631f9c18 AD |
2847 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
2848 | LNK("cwd", proc_cwd_link), | |
2849 | LNK("root", proc_root_link), | |
2850 | LNK("exe", proc_exe_link), | |
2851 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
2852 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
2853 | REG("mountstats", S_IRUSR, proc_mountstats_operations), | |
1e883281 | 2854 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 2855 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
b7643757 | 2856 | REG("smaps", S_IRUGO, proc_pid_smaps_operations), |
32ed74a4 | 2857 | REG("pagemap", S_IRUSR, proc_pagemap_operations), |
28a6d671 EB |
2858 | #endif |
2859 | #ifdef CONFIG_SECURITY | |
631f9c18 | 2860 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
2861 | #endif |
2862 | #ifdef CONFIG_KALLSYMS | |
edfcd606 | 2863 | ONE("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 2864 | #endif |
2ec220e2 | 2865 | #ifdef CONFIG_STACKTRACE |
35a35046 | 2866 | ONE("stack", S_IRUSR, proc_pid_stack), |
28a6d671 | 2867 | #endif |
5968cece | 2868 | #ifdef CONFIG_SCHED_INFO |
f6e826ca | 2869 | ONE("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 2870 | #endif |
9745512c | 2871 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 2872 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 2873 | #endif |
8793d854 | 2874 | #ifdef CONFIG_PROC_PID_CPUSET |
52de4779 | 2875 | ONE("cpuset", S_IRUGO, proc_cpuset_show), |
a424316c PM |
2876 | #endif |
2877 | #ifdef CONFIG_CGROUPS | |
006f4ac4 | 2878 | ONE("cgroup", S_IRUGO, proc_cgroup_show), |
28a6d671 | 2879 | #endif |
6ba51e37 | 2880 | ONE("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 2881 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 2882 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
28a6d671 | 2883 | #ifdef CONFIG_AUDITSYSCALL |
631f9c18 AD |
2884 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
2885 | REG("sessionid", S_IRUGO, proc_sessionid_operations), | |
28a6d671 | 2886 | #endif |
f4f154fd | 2887 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 2888 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
f4f154fd | 2889 | #endif |
698ba7b5 | 2890 | #ifdef CONFIG_ELF_CORE |
631f9c18 | 2891 | REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations), |
3cb4a0bb | 2892 | #endif |
aba76fdb | 2893 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 2894 | ONE("io", S_IRUSR, proc_tgid_io_accounting), |
aba76fdb | 2895 | #endif |
f133ecca | 2896 | #ifdef CONFIG_HARDWALL |
d962c144 | 2897 | ONE("hardwall", S_IRUGO, proc_pid_hardwall), |
f133ecca | 2898 | #endif |
22d917d8 EB |
2899 | #ifdef CONFIG_USER_NS |
2900 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
2901 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 2902 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
9cc46516 | 2903 | REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations), |
22d917d8 | 2904 | #endif |
48f6a7a5 PE |
2905 | #ifdef CONFIG_CHECKPOINT_RESTORE |
2906 | REG("timers", S_IRUGO, proc_timers_operations), | |
2907 | #endif | |
5de23d43 | 2908 | REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations), |
28a6d671 | 2909 | }; |
1da177e4 | 2910 | |
f0c3b509 | 2911 | static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 2912 | { |
f0c3b509 AV |
2913 | return proc_pident_readdir(file, ctx, |
2914 | tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
2915 | } |
2916 | ||
00977a59 | 2917 | static const struct file_operations proc_tgid_base_operations = { |
1da177e4 | 2918 | .read = generic_read_dir, |
f50752ea AV |
2919 | .iterate_shared = proc_tgid_base_readdir, |
2920 | .llseek = generic_file_llseek, | |
1da177e4 LT |
2921 | }; |
2922 | ||
00cd8dd3 AV |
2923 | static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
2924 | { | |
7bcd6b0e EB |
2925 | return proc_pident_lookup(dir, dentry, |
2926 | tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
2927 | } |
2928 | ||
c5ef1c42 | 2929 | static const struct inode_operations proc_tgid_base_inode_operations = { |
28a6d671 | 2930 | .lookup = proc_tgid_base_lookup, |
99f89551 | 2931 | .getattr = pid_getattr, |
6d76fa58 | 2932 | .setattr = proc_setattr, |
0499680a | 2933 | .permission = proc_pid_permission, |
1da177e4 | 2934 | }; |
1da177e4 | 2935 | |
60347f67 | 2936 | static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid) |
1da177e4 | 2937 | { |
48e6484d | 2938 | struct dentry *dentry, *leader, *dir; |
8578cea7 | 2939 | char buf[PROC_NUMBUF]; |
48e6484d EB |
2940 | struct qstr name; |
2941 | ||
2942 | name.name = buf; | |
60347f67 | 2943 | name.len = snprintf(buf, sizeof(buf), "%d", pid); |
4f522a24 | 2944 | /* no ->d_hash() rejects on procfs */ |
60347f67 | 2945 | dentry = d_hash_and_lookup(mnt->mnt_root, &name); |
48e6484d | 2946 | if (dentry) { |
bbd51924 | 2947 | d_invalidate(dentry); |
48e6484d EB |
2948 | dput(dentry); |
2949 | } | |
1da177e4 | 2950 | |
c35a7f18 ON |
2951 | if (pid == tgid) |
2952 | return; | |
2953 | ||
48e6484d | 2954 | name.name = buf; |
60347f67 PE |
2955 | name.len = snprintf(buf, sizeof(buf), "%d", tgid); |
2956 | leader = d_hash_and_lookup(mnt->mnt_root, &name); | |
48e6484d EB |
2957 | if (!leader) |
2958 | goto out; | |
1da177e4 | 2959 | |
48e6484d EB |
2960 | name.name = "task"; |
2961 | name.len = strlen(name.name); | |
2962 | dir = d_hash_and_lookup(leader, &name); | |
2963 | if (!dir) | |
2964 | goto out_put_leader; | |
2965 | ||
2966 | name.name = buf; | |
60347f67 | 2967 | name.len = snprintf(buf, sizeof(buf), "%d", pid); |
48e6484d EB |
2968 | dentry = d_hash_and_lookup(dir, &name); |
2969 | if (dentry) { | |
bbd51924 | 2970 | d_invalidate(dentry); |
48e6484d | 2971 | dput(dentry); |
1da177e4 | 2972 | } |
48e6484d EB |
2973 | |
2974 | dput(dir); | |
2975 | out_put_leader: | |
2976 | dput(leader); | |
2977 | out: | |
2978 | return; | |
1da177e4 LT |
2979 | } |
2980 | ||
0895e91d RD |
2981 | /** |
2982 | * proc_flush_task - Remove dcache entries for @task from the /proc dcache. | |
2983 | * @task: task that should be flushed. | |
2984 | * | |
2985 | * When flushing dentries from proc, one needs to flush them from global | |
60347f67 | 2986 | * proc (proc_mnt) and from all the namespaces' procs this task was seen |
0895e91d RD |
2987 | * in. This call is supposed to do all of this job. |
2988 | * | |
2989 | * Looks in the dcache for | |
2990 | * /proc/@pid | |
2991 | * /proc/@tgid/task/@pid | |
2992 | * if either directory is present flushes it and all of it'ts children | |
2993 | * from the dcache. | |
2994 | * | |
2995 | * It is safe and reasonable to cache /proc entries for a task until | |
2996 | * that task exits. After that they just clog up the dcache with | |
2997 | * useless entries, possibly causing useful dcache entries to be | |
2998 | * flushed instead. This routine is proved to flush those useless | |
2999 | * dcache entries at process exit time. | |
3000 | * | |
3001 | * NOTE: This routine is just an optimization so it does not guarantee | |
3002 | * that no dcache entries will exist at process exit time it | |
3003 | * just makes it very unlikely that any will persist. | |
60347f67 PE |
3004 | */ |
3005 | ||
3006 | void proc_flush_task(struct task_struct *task) | |
3007 | { | |
9fcc2d15 | 3008 | int i; |
9b4d1cbe | 3009 | struct pid *pid, *tgid; |
130f77ec PE |
3010 | struct upid *upid; |
3011 | ||
130f77ec | 3012 | pid = task_pid(task); |
9b4d1cbe | 3013 | tgid = task_tgid(task); |
130f77ec | 3014 | |
9fcc2d15 | 3015 | for (i = 0; i <= pid->level; i++) { |
130f77ec PE |
3016 | upid = &pid->numbers[i]; |
3017 | proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr, | |
9b4d1cbe | 3018 | tgid->numbers[i].nr); |
130f77ec | 3019 | } |
60347f67 PE |
3020 | } |
3021 | ||
c52a47ac AV |
3022 | static int proc_pid_instantiate(struct inode *dir, |
3023 | struct dentry * dentry, | |
3024 | struct task_struct *task, const void *ptr) | |
444ceed8 | 3025 | { |
444ceed8 EB |
3026 | struct inode *inode; |
3027 | ||
61a28784 | 3028 | inode = proc_pid_make_inode(dir->i_sb, task); |
444ceed8 EB |
3029 | if (!inode) |
3030 | goto out; | |
3031 | ||
3032 | inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO; | |
3033 | inode->i_op = &proc_tgid_base_inode_operations; | |
3034 | inode->i_fop = &proc_tgid_base_operations; | |
3035 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 | 3036 | |
bfe86848 MS |
3037 | set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff, |
3038 | ARRAY_SIZE(tgid_base_stuff))); | |
444ceed8 | 3039 | |
fb045adb | 3040 | d_set_d_op(dentry, &pid_dentry_operations); |
444ceed8 EB |
3041 | |
3042 | d_add(dentry, inode); | |
3043 | /* Close the race of the process dying before we return the dentry */ | |
0b728e19 | 3044 | if (pid_revalidate(dentry, 0)) |
c52a47ac | 3045 | return 0; |
444ceed8 | 3046 | out: |
c52a47ac | 3047 | return -ENOENT; |
444ceed8 EB |
3048 | } |
3049 | ||
00cd8dd3 | 3050 | struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags) |
1da177e4 | 3051 | { |
335eb531 | 3052 | int result = -ENOENT; |
1da177e4 | 3053 | struct task_struct *task; |
1da177e4 | 3054 | unsigned tgid; |
b488893a | 3055 | struct pid_namespace *ns; |
1da177e4 | 3056 | |
dbcdb504 | 3057 | tgid = name_to_int(&dentry->d_name); |
1da177e4 LT |
3058 | if (tgid == ~0U) |
3059 | goto out; | |
3060 | ||
b488893a | 3061 | ns = dentry->d_sb->s_fs_info; |
de758734 | 3062 | rcu_read_lock(); |
b488893a | 3063 | task = find_task_by_pid_ns(tgid, ns); |
1da177e4 LT |
3064 | if (task) |
3065 | get_task_struct(task); | |
de758734 | 3066 | rcu_read_unlock(); |
1da177e4 LT |
3067 | if (!task) |
3068 | goto out; | |
3069 | ||
444ceed8 | 3070 | result = proc_pid_instantiate(dir, dentry, task, NULL); |
1da177e4 | 3071 | put_task_struct(task); |
1da177e4 | 3072 | out: |
c52a47ac | 3073 | return ERR_PTR(result); |
1da177e4 LT |
3074 | } |
3075 | ||
1da177e4 | 3076 | /* |
0804ef4b | 3077 | * Find the first task with tgid >= tgid |
0bc58a91 | 3078 | * |
1da177e4 | 3079 | */ |
19fd4bb2 EB |
3080 | struct tgid_iter { |
3081 | unsigned int tgid; | |
0804ef4b | 3082 | struct task_struct *task; |
19fd4bb2 EB |
3083 | }; |
3084 | static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter) | |
3085 | { | |
0804ef4b | 3086 | struct pid *pid; |
1da177e4 | 3087 | |
19fd4bb2 EB |
3088 | if (iter.task) |
3089 | put_task_struct(iter.task); | |
454cc105 | 3090 | rcu_read_lock(); |
0804ef4b | 3091 | retry: |
19fd4bb2 EB |
3092 | iter.task = NULL; |
3093 | pid = find_ge_pid(iter.tgid, ns); | |
0804ef4b | 3094 | if (pid) { |
19fd4bb2 EB |
3095 | iter.tgid = pid_nr_ns(pid, ns); |
3096 | iter.task = pid_task(pid, PIDTYPE_PID); | |
0804ef4b EB |
3097 | /* What we to know is if the pid we have find is the |
3098 | * pid of a thread_group_leader. Testing for task | |
3099 | * being a thread_group_leader is the obvious thing | |
3100 | * todo but there is a window when it fails, due to | |
3101 | * the pid transfer logic in de_thread. | |
3102 | * | |
3103 | * So we perform the straight forward test of seeing | |
3104 | * if the pid we have found is the pid of a thread | |
3105 | * group leader, and don't worry if the task we have | |
3106 | * found doesn't happen to be a thread group leader. | |
3107 | * As we don't care in the case of readdir. | |
3108 | */ | |
19fd4bb2 EB |
3109 | if (!iter.task || !has_group_leader_pid(iter.task)) { |
3110 | iter.tgid += 1; | |
0804ef4b | 3111 | goto retry; |
19fd4bb2 EB |
3112 | } |
3113 | get_task_struct(iter.task); | |
0bc58a91 | 3114 | } |
454cc105 | 3115 | rcu_read_unlock(); |
19fd4bb2 | 3116 | return iter; |
1da177e4 LT |
3117 | } |
3118 | ||
0097875b | 3119 | #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2) |
0804ef4b | 3120 | |
1da177e4 | 3121 | /* for the /proc/ directory itself, after non-process stuff has been done */ |
f0c3b509 | 3122 | int proc_pid_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3123 | { |
19fd4bb2 | 3124 | struct tgid_iter iter; |
3aa3377f | 3125 | struct pid_namespace *ns = file_inode(file)->i_sb->s_fs_info; |
f0c3b509 | 3126 | loff_t pos = ctx->pos; |
1da177e4 | 3127 | |
021ada7d | 3128 | if (pos >= PID_MAX_LIMIT + TGID_OFFSET) |
f0c3b509 | 3129 | return 0; |
1da177e4 | 3130 | |
0097875b | 3131 | if (pos == TGID_OFFSET - 2) { |
2b0143b5 | 3132 | struct inode *inode = d_inode(ns->proc_self); |
db963164 | 3133 | if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK)) |
f0c3b509 | 3134 | return 0; |
0097875b EB |
3135 | ctx->pos = pos = pos + 1; |
3136 | } | |
3137 | if (pos == TGID_OFFSET - 1) { | |
2b0143b5 | 3138 | struct inode *inode = d_inode(ns->proc_thread_self); |
0097875b EB |
3139 | if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK)) |
3140 | return 0; | |
3141 | ctx->pos = pos = pos + 1; | |
021ada7d | 3142 | } |
0097875b | 3143 | iter.tgid = pos - TGID_OFFSET; |
19fd4bb2 | 3144 | iter.task = NULL; |
19fd4bb2 EB |
3145 | for (iter = next_tgid(ns, iter); |
3146 | iter.task; | |
3147 | iter.tgid += 1, iter = next_tgid(ns, iter)) { | |
f0c3b509 AV |
3148 | char name[PROC_NUMBUF]; |
3149 | int len; | |
3150 | if (!has_pid_permissions(ns, iter.task, 2)) | |
3151 | continue; | |
0499680a | 3152 | |
f0c3b509 AV |
3153 | len = snprintf(name, sizeof(name), "%d", iter.tgid); |
3154 | ctx->pos = iter.tgid + TGID_OFFSET; | |
3155 | if (!proc_fill_cache(file, ctx, name, len, | |
3156 | proc_pid_instantiate, iter.task, NULL)) { | |
19fd4bb2 | 3157 | put_task_struct(iter.task); |
f0c3b509 | 3158 | return 0; |
1da177e4 | 3159 | } |
0bc58a91 | 3160 | } |
f0c3b509 | 3161 | ctx->pos = PID_MAX_LIMIT + TGID_OFFSET; |
0bc58a91 EB |
3162 | return 0; |
3163 | } | |
1da177e4 | 3164 | |
1b3044e3 JD |
3165 | /* |
3166 | * proc_tid_comm_permission is a special permission function exclusively | |
3167 | * used for the node /proc/<pid>/task/<tid>/comm. | |
3168 | * It bypasses generic permission checks in the case where a task of the same | |
3169 | * task group attempts to access the node. | |
3170 | * The rationale behind this is that glibc and bionic access this node for | |
3171 | * cross thread naming (pthread_set/getname_np(!self)). However, if | |
3172 | * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0, | |
3173 | * which locks out the cross thread naming implementation. | |
3174 | * This function makes sure that the node is always accessible for members of | |
3175 | * same thread group. | |
3176 | */ | |
3177 | static int proc_tid_comm_permission(struct inode *inode, int mask) | |
3178 | { | |
3179 | bool is_same_tgroup; | |
3180 | struct task_struct *task; | |
3181 | ||
3182 | task = get_proc_task(inode); | |
3183 | if (!task) | |
3184 | return -ESRCH; | |
3185 | is_same_tgroup = same_thread_group(current, task); | |
3186 | put_task_struct(task); | |
3187 | ||
3188 | if (likely(is_same_tgroup && !(mask & MAY_EXEC))) { | |
3189 | /* This file (/proc/<pid>/task/<tid>/comm) can always be | |
3190 | * read or written by the members of the corresponding | |
3191 | * thread group. | |
3192 | */ | |
3193 | return 0; | |
3194 | } | |
3195 | ||
3196 | return generic_permission(inode, mask); | |
3197 | } | |
3198 | ||
3199 | static const struct inode_operations proc_tid_comm_inode_operations = { | |
3200 | .permission = proc_tid_comm_permission, | |
3201 | }; | |
3202 | ||
28a6d671 EB |
3203 | /* |
3204 | * Tasks | |
3205 | */ | |
c5141e6d | 3206 | static const struct pid_entry tid_base_stuff[] = { |
631f9c18 | 3207 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), |
3835541d | 3208 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 3209 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
6ba8ed79 EB |
3210 | #ifdef CONFIG_NET |
3211 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), | |
3212 | #endif | |
631f9c18 | 3213 | REG("environ", S_IRUSR, proc_environ_operations), |
f9ea536e | 3214 | ONE("auxv", S_IRUSR, proc_pid_auxv), |
631f9c18 | 3215 | ONE("status", S_IRUGO, proc_pid_status), |
35a35046 | 3216 | ONE("personality", S_IRUSR, proc_pid_personality), |
1c963eb1 | 3217 | ONE("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 3218 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 3219 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
ebcb6734 | 3220 | #endif |
1b3044e3 JD |
3221 | NOD("comm", S_IFREG|S_IRUGO|S_IWUSR, |
3222 | &proc_tid_comm_inode_operations, | |
3223 | &proc_pid_set_comm_operations, {}), | |
ebcb6734 | 3224 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 | 3225 | ONE("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 3226 | #endif |
c2c0bb44 | 3227 | REG("cmdline", S_IRUGO, proc_pid_cmdline_ops), |
631f9c18 AD |
3228 | ONE("stat", S_IRUGO, proc_tid_stat), |
3229 | ONE("statm", S_IRUGO, proc_pid_statm), | |
b7643757 | 3230 | REG("maps", S_IRUGO, proc_tid_maps_operations), |
2e13ba54 | 3231 | #ifdef CONFIG_PROC_CHILDREN |
81841161 CG |
3232 | REG("children", S_IRUGO, proc_tid_children_operations), |
3233 | #endif | |
28a6d671 | 3234 | #ifdef CONFIG_NUMA |
b7643757 | 3235 | REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations), |
28a6d671 | 3236 | #endif |
631f9c18 AD |
3237 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
3238 | LNK("cwd", proc_cwd_link), | |
3239 | LNK("root", proc_root_link), | |
3240 | LNK("exe", proc_exe_link), | |
3241 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
3242 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
1e883281 | 3243 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 3244 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
b7643757 | 3245 | REG("smaps", S_IRUGO, proc_tid_smaps_operations), |
32ed74a4 | 3246 | REG("pagemap", S_IRUSR, proc_pagemap_operations), |
28a6d671 EB |
3247 | #endif |
3248 | #ifdef CONFIG_SECURITY | |
631f9c18 | 3249 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
3250 | #endif |
3251 | #ifdef CONFIG_KALLSYMS | |
edfcd606 | 3252 | ONE("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 3253 | #endif |
2ec220e2 | 3254 | #ifdef CONFIG_STACKTRACE |
35a35046 | 3255 | ONE("stack", S_IRUSR, proc_pid_stack), |
28a6d671 | 3256 | #endif |
5968cece | 3257 | #ifdef CONFIG_SCHED_INFO |
f6e826ca | 3258 | ONE("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 3259 | #endif |
9745512c | 3260 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 3261 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 3262 | #endif |
8793d854 | 3263 | #ifdef CONFIG_PROC_PID_CPUSET |
52de4779 | 3264 | ONE("cpuset", S_IRUGO, proc_cpuset_show), |
a424316c PM |
3265 | #endif |
3266 | #ifdef CONFIG_CGROUPS | |
006f4ac4 | 3267 | ONE("cgroup", S_IRUGO, proc_cgroup_show), |
28a6d671 | 3268 | #endif |
6ba51e37 | 3269 | ONE("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 3270 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 3271 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
28a6d671 | 3272 | #ifdef CONFIG_AUDITSYSCALL |
631f9c18 | 3273 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
26ec3c64 | 3274 | REG("sessionid", S_IRUGO, proc_sessionid_operations), |
28a6d671 | 3275 | #endif |
f4f154fd | 3276 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 3277 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
f4f154fd | 3278 | #endif |
297c5d92 | 3279 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 3280 | ONE("io", S_IRUSR, proc_tid_io_accounting), |
297c5d92 | 3281 | #endif |
f133ecca | 3282 | #ifdef CONFIG_HARDWALL |
d962c144 | 3283 | ONE("hardwall", S_IRUGO, proc_pid_hardwall), |
f133ecca | 3284 | #endif |
22d917d8 EB |
3285 | #ifdef CONFIG_USER_NS |
3286 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
3287 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 3288 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
9cc46516 | 3289 | REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations), |
22d917d8 | 3290 | #endif |
28a6d671 EB |
3291 | }; |
3292 | ||
f0c3b509 | 3293 | static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx) |
28a6d671 | 3294 | { |
f0c3b509 AV |
3295 | return proc_pident_readdir(file, ctx, |
3296 | tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
3297 | } |
3298 | ||
00cd8dd3 AV |
3299 | static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
3300 | { | |
7bcd6b0e EB |
3301 | return proc_pident_lookup(dir, dentry, |
3302 | tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
3303 | } |
3304 | ||
00977a59 | 3305 | static const struct file_operations proc_tid_base_operations = { |
28a6d671 | 3306 | .read = generic_read_dir, |
f50752ea AV |
3307 | .iterate_shared = proc_tid_base_readdir, |
3308 | .llseek = generic_file_llseek, | |
28a6d671 EB |
3309 | }; |
3310 | ||
c5ef1c42 | 3311 | static const struct inode_operations proc_tid_base_inode_operations = { |
28a6d671 EB |
3312 | .lookup = proc_tid_base_lookup, |
3313 | .getattr = pid_getattr, | |
3314 | .setattr = proc_setattr, | |
3315 | }; | |
3316 | ||
c52a47ac | 3317 | static int proc_task_instantiate(struct inode *dir, |
c5141e6d | 3318 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
444ceed8 | 3319 | { |
444ceed8 | 3320 | struct inode *inode; |
61a28784 | 3321 | inode = proc_pid_make_inode(dir->i_sb, task); |
444ceed8 EB |
3322 | |
3323 | if (!inode) | |
3324 | goto out; | |
3325 | inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO; | |
3326 | inode->i_op = &proc_tid_base_inode_operations; | |
3327 | inode->i_fop = &proc_tid_base_operations; | |
3328 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 | 3329 | |
bfe86848 MS |
3330 | set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff, |
3331 | ARRAY_SIZE(tid_base_stuff))); | |
444ceed8 | 3332 | |
fb045adb | 3333 | d_set_d_op(dentry, &pid_dentry_operations); |
444ceed8 EB |
3334 | |
3335 | d_add(dentry, inode); | |
3336 | /* Close the race of the process dying before we return the dentry */ | |
0b728e19 | 3337 | if (pid_revalidate(dentry, 0)) |
c52a47ac | 3338 | return 0; |
444ceed8 | 3339 | out: |
c52a47ac | 3340 | return -ENOENT; |
444ceed8 EB |
3341 | } |
3342 | ||
00cd8dd3 | 3343 | static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags) |
28a6d671 | 3344 | { |
c52a47ac | 3345 | int result = -ENOENT; |
28a6d671 EB |
3346 | struct task_struct *task; |
3347 | struct task_struct *leader = get_proc_task(dir); | |
28a6d671 | 3348 | unsigned tid; |
b488893a | 3349 | struct pid_namespace *ns; |
28a6d671 EB |
3350 | |
3351 | if (!leader) | |
3352 | goto out_no_task; | |
3353 | ||
dbcdb504 | 3354 | tid = name_to_int(&dentry->d_name); |
28a6d671 EB |
3355 | if (tid == ~0U) |
3356 | goto out; | |
3357 | ||
b488893a | 3358 | ns = dentry->d_sb->s_fs_info; |
28a6d671 | 3359 | rcu_read_lock(); |
b488893a | 3360 | task = find_task_by_pid_ns(tid, ns); |
28a6d671 EB |
3361 | if (task) |
3362 | get_task_struct(task); | |
3363 | rcu_read_unlock(); | |
3364 | if (!task) | |
3365 | goto out; | |
bac0abd6 | 3366 | if (!same_thread_group(leader, task)) |
28a6d671 EB |
3367 | goto out_drop_task; |
3368 | ||
444ceed8 | 3369 | result = proc_task_instantiate(dir, dentry, task, NULL); |
28a6d671 EB |
3370 | out_drop_task: |
3371 | put_task_struct(task); | |
3372 | out: | |
3373 | put_task_struct(leader); | |
3374 | out_no_task: | |
c52a47ac | 3375 | return ERR_PTR(result); |
28a6d671 EB |
3376 | } |
3377 | ||
0bc58a91 EB |
3378 | /* |
3379 | * Find the first tid of a thread group to return to user space. | |
3380 | * | |
3381 | * Usually this is just the thread group leader, but if the users | |
3382 | * buffer was too small or there was a seek into the middle of the | |
3383 | * directory we have more work todo. | |
3384 | * | |
3385 | * In the case of a short read we start with find_task_by_pid. | |
3386 | * | |
3387 | * In the case of a seek we start with the leader and walk nr | |
3388 | * threads past it. | |
3389 | */ | |
9f6e963f ON |
3390 | static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos, |
3391 | struct pid_namespace *ns) | |
0bc58a91 | 3392 | { |
d855a4b7 | 3393 | struct task_struct *pos, *task; |
9f6e963f ON |
3394 | unsigned long nr = f_pos; |
3395 | ||
3396 | if (nr != f_pos) /* 32bit overflow? */ | |
3397 | return NULL; | |
1da177e4 | 3398 | |
cc288738 | 3399 | rcu_read_lock(); |
d855a4b7 ON |
3400 | task = pid_task(pid, PIDTYPE_PID); |
3401 | if (!task) | |
3402 | goto fail; | |
3403 | ||
3404 | /* Attempt to start with the tid of a thread */ | |
9f6e963f | 3405 | if (tid && nr) { |
b488893a | 3406 | pos = find_task_by_pid_ns(tid, ns); |
d855a4b7 | 3407 | if (pos && same_thread_group(pos, task)) |
a872ff0c | 3408 | goto found; |
0bc58a91 | 3409 | } |
1da177e4 | 3410 | |
0bc58a91 | 3411 | /* If nr exceeds the number of threads there is nothing todo */ |
9f6e963f | 3412 | if (nr >= get_nr_threads(task)) |
c986c14a | 3413 | goto fail; |
1da177e4 | 3414 | |
a872ff0c ON |
3415 | /* If we haven't found our starting place yet start |
3416 | * with the leader and walk nr threads forward. | |
0bc58a91 | 3417 | */ |
d855a4b7 | 3418 | pos = task = task->group_leader; |
c986c14a | 3419 | do { |
9f6e963f | 3420 | if (!nr--) |
c986c14a | 3421 | goto found; |
d855a4b7 | 3422 | } while_each_thread(task, pos); |
c986c14a ON |
3423 | fail: |
3424 | pos = NULL; | |
3425 | goto out; | |
a872ff0c ON |
3426 | found: |
3427 | get_task_struct(pos); | |
3428 | out: | |
cc288738 | 3429 | rcu_read_unlock(); |
0bc58a91 EB |
3430 | return pos; |
3431 | } | |
3432 | ||
3433 | /* | |
3434 | * Find the next thread in the thread list. | |
3435 | * Return NULL if there is an error or no next thread. | |
3436 | * | |
3437 | * The reference to the input task_struct is released. | |
3438 | */ | |
3439 | static struct task_struct *next_tid(struct task_struct *start) | |
3440 | { | |
c1df7fb8 | 3441 | struct task_struct *pos = NULL; |
cc288738 | 3442 | rcu_read_lock(); |
c1df7fb8 | 3443 | if (pid_alive(start)) { |
0bc58a91 | 3444 | pos = next_thread(start); |
c1df7fb8 ON |
3445 | if (thread_group_leader(pos)) |
3446 | pos = NULL; | |
3447 | else | |
3448 | get_task_struct(pos); | |
3449 | } | |
cc288738 | 3450 | rcu_read_unlock(); |
0bc58a91 EB |
3451 | put_task_struct(start); |
3452 | return pos; | |
1da177e4 LT |
3453 | } |
3454 | ||
3455 | /* for the /proc/TGID/task/ directories */ | |
f0c3b509 | 3456 | static int proc_task_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3457 | { |
d855a4b7 ON |
3458 | struct inode *inode = file_inode(file); |
3459 | struct task_struct *task; | |
b488893a | 3460 | struct pid_namespace *ns; |
f0c3b509 | 3461 | int tid; |
1da177e4 | 3462 | |
d855a4b7 | 3463 | if (proc_inode_is_dead(inode)) |
f0c3b509 | 3464 | return -ENOENT; |
1da177e4 | 3465 | |
f0c3b509 | 3466 | if (!dir_emit_dots(file, ctx)) |
d855a4b7 | 3467 | return 0; |
1da177e4 | 3468 | |
0bc58a91 EB |
3469 | /* f_version caches the tgid value that the last readdir call couldn't |
3470 | * return. lseek aka telldir automagically resets f_version to 0. | |
3471 | */ | |
3aa3377f | 3472 | ns = inode->i_sb->s_fs_info; |
f0c3b509 AV |
3473 | tid = (int)file->f_version; |
3474 | file->f_version = 0; | |
d855a4b7 | 3475 | for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns); |
0bc58a91 | 3476 | task; |
f0c3b509 AV |
3477 | task = next_tid(task), ctx->pos++) { |
3478 | char name[PROC_NUMBUF]; | |
3479 | int len; | |
b488893a | 3480 | tid = task_pid_nr_ns(task, ns); |
f0c3b509 AV |
3481 | len = snprintf(name, sizeof(name), "%d", tid); |
3482 | if (!proc_fill_cache(file, ctx, name, len, | |
3483 | proc_task_instantiate, task, NULL)) { | |
0bc58a91 EB |
3484 | /* returning this tgid failed, save it as the first |
3485 | * pid for the next readir call */ | |
f0c3b509 | 3486 | file->f_version = (u64)tid; |
0bc58a91 | 3487 | put_task_struct(task); |
1da177e4 | 3488 | break; |
0bc58a91 | 3489 | } |
1da177e4 | 3490 | } |
d855a4b7 | 3491 | |
f0c3b509 | 3492 | return 0; |
1da177e4 | 3493 | } |
6e66b52b EB |
3494 | |
3495 | static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat) | |
3496 | { | |
2b0143b5 | 3497 | struct inode *inode = d_inode(dentry); |
99f89551 | 3498 | struct task_struct *p = get_proc_task(inode); |
6e66b52b EB |
3499 | generic_fillattr(inode, stat); |
3500 | ||
99f89551 | 3501 | if (p) { |
99f89551 | 3502 | stat->nlink += get_nr_threads(p); |
99f89551 | 3503 | put_task_struct(p); |
6e66b52b EB |
3504 | } |
3505 | ||
3506 | return 0; | |
3507 | } | |
28a6d671 | 3508 | |
c5ef1c42 | 3509 | static const struct inode_operations proc_task_inode_operations = { |
28a6d671 EB |
3510 | .lookup = proc_task_lookup, |
3511 | .getattr = proc_task_getattr, | |
3512 | .setattr = proc_setattr, | |
0499680a | 3513 | .permission = proc_pid_permission, |
28a6d671 EB |
3514 | }; |
3515 | ||
00977a59 | 3516 | static const struct file_operations proc_task_operations = { |
28a6d671 | 3517 | .read = generic_read_dir, |
f50752ea AV |
3518 | .iterate_shared = proc_task_readdir, |
3519 | .llseek = generic_file_llseek, | |
28a6d671 | 3520 | }; |