]>
Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 LT |
2 | /* |
3 | * linux/fs/proc/base.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | * | |
7 | * proc base directory handling functions | |
8 | * | |
9 | * 1999, Al Viro. Rewritten. Now it covers the whole per-process part. | |
10 | * Instead of using magical inumbers to determine the kind of object | |
11 | * we allocate and fill in-core inodes upon lookup. They don't even | |
12 | * go into icache. We cache the reference to task_struct upon lookup too. | |
13 | * Eventually it should become a filesystem in its own. We don't use the | |
14 | * rest of procfs anymore. | |
e070ad49 ML |
15 | * |
16 | * | |
17 | * Changelog: | |
18 | * 17-Jan-2005 | |
19 | * Allan Bezerra | |
20 | * Bruna Moreira <bruna.moreira@indt.org.br> | |
21 | * Edjard Mota <edjard.mota@indt.org.br> | |
22 | * Ilias Biris <ilias.biris@indt.org.br> | |
23 | * Mauricio Lin <mauricio.lin@indt.org.br> | |
24 | * | |
25 | * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT | |
26 | * | |
27 | * A new process specific entry (smaps) included in /proc. It shows the | |
28 | * size of rss for each memory area. The maps entry lacks information | |
29 | * about physical memory size (rss) for each mapped file, i.e., | |
30 | * rss information for executables and library files. | |
31 | * This additional information is useful for any tools that need to know | |
32 | * about physical memory consumption for a process specific library. | |
33 | * | |
34 | * Changelog: | |
35 | * 21-Feb-2005 | |
36 | * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT | |
37 | * Pud inclusion in the page table walking. | |
38 | * | |
39 | * ChangeLog: | |
40 | * 10-Mar-2005 | |
41 | * 10LE Instituto Nokia de Tecnologia - INdT: | |
42 | * A better way to walks through the page table as suggested by Hugh Dickins. | |
43 | * | |
44 | * Simo Piiroinen <simo.piiroinen@nokia.com>: | |
45 | * Smaps information related to shared, private, clean and dirty pages. | |
46 | * | |
47 | * Paul Mundt <paul.mundt@nokia.com>: | |
48 | * Overall revision about smaps. | |
1da177e4 LT |
49 | */ |
50 | ||
7c0f6ba6 | 51 | #include <linux/uaccess.h> |
1da177e4 | 52 | |
1da177e4 LT |
53 | #include <linux/errno.h> |
54 | #include <linux/time.h> | |
55 | #include <linux/proc_fs.h> | |
56 | #include <linux/stat.h> | |
5995477a | 57 | #include <linux/task_io_accounting_ops.h> |
1da177e4 | 58 | #include <linux/init.h> |
16f7e0fe | 59 | #include <linux/capability.h> |
1da177e4 | 60 | #include <linux/file.h> |
9f3acc31 | 61 | #include <linux/fdtable.h> |
1da177e4 LT |
62 | #include <linux/string.h> |
63 | #include <linux/seq_file.h> | |
64 | #include <linux/namei.h> | |
6b3286ed | 65 | #include <linux/mnt_namespace.h> |
1da177e4 | 66 | #include <linux/mm.h> |
a63d83f4 | 67 | #include <linux/swap.h> |
b835996f | 68 | #include <linux/rcupdate.h> |
1da177e4 | 69 | #include <linux/kallsyms.h> |
2ec220e2 | 70 | #include <linux/stacktrace.h> |
d85f50d5 | 71 | #include <linux/resource.h> |
5096add8 | 72 | #include <linux/module.h> |
1da177e4 LT |
73 | #include <linux/mount.h> |
74 | #include <linux/security.h> | |
75 | #include <linux/ptrace.h> | |
0d094efe | 76 | #include <linux/tracehook.h> |
87ebdc00 | 77 | #include <linux/printk.h> |
efb1a57d | 78 | #include <linux/cache.h> |
a424316c | 79 | #include <linux/cgroup.h> |
1da177e4 LT |
80 | #include <linux/cpuset.h> |
81 | #include <linux/audit.h> | |
5addc5dd | 82 | #include <linux/poll.h> |
1651e14e | 83 | #include <linux/nsproxy.h> |
8ac773b4 | 84 | #include <linux/oom.h> |
3cb4a0bb | 85 | #include <linux/elf.h> |
60347f67 | 86 | #include <linux/pid_namespace.h> |
22d917d8 | 87 | #include <linux/user_namespace.h> |
5ad4e53b | 88 | #include <linux/fs_struct.h> |
5a0e3ad6 | 89 | #include <linux/slab.h> |
4eb5aaa3 | 90 | #include <linux/sched/autogroup.h> |
6e84f315 | 91 | #include <linux/sched/mm.h> |
f7ccbae4 | 92 | #include <linux/sched/coredump.h> |
b17b0153 | 93 | #include <linux/sched/debug.h> |
3905f9ad | 94 | #include <linux/sched/stat.h> |
640708a2 | 95 | #include <linux/flex_array.h> |
48f6a7a5 | 96 | #include <linux/posix-timers.h> |
43d2b113 | 97 | #include <trace/events/oom.h> |
1da177e4 | 98 | #include "internal.h" |
faf60af1 | 99 | #include "fd.h" |
1da177e4 | 100 | |
ac7f1061 AD |
101 | #include "../../lib/kstrtox.h" |
102 | ||
0f2fe20f EB |
103 | /* NOTE: |
104 | * Implementing inode permission operations in /proc is almost | |
105 | * certainly an error. Permission checks need to happen during | |
106 | * each system call not at open time. The reason is that most of | |
107 | * what we wish to check for permissions in /proc varies at runtime. | |
108 | * | |
109 | * The classic example of a problem is opening file descriptors | |
110 | * in /proc for a task before it execs a suid executable. | |
111 | */ | |
112 | ||
efb1a57d AD |
113 | static u8 nlink_tid __ro_after_init; |
114 | static u8 nlink_tgid __ro_after_init; | |
1270dd8d | 115 | |
1da177e4 | 116 | struct pid_entry { |
cedbccab | 117 | const char *name; |
623f594e | 118 | unsigned int len; |
d161a13f | 119 | umode_t mode; |
c5ef1c42 | 120 | const struct inode_operations *iop; |
00977a59 | 121 | const struct file_operations *fop; |
20cdc894 | 122 | union proc_op op; |
1da177e4 LT |
123 | }; |
124 | ||
61a28784 | 125 | #define NOD(NAME, MODE, IOP, FOP, OP) { \ |
20cdc894 | 126 | .name = (NAME), \ |
c5141e6d | 127 | .len = sizeof(NAME) - 1, \ |
20cdc894 EB |
128 | .mode = MODE, \ |
129 | .iop = IOP, \ | |
130 | .fop = FOP, \ | |
131 | .op = OP, \ | |
132 | } | |
133 | ||
631f9c18 AD |
134 | #define DIR(NAME, MODE, iops, fops) \ |
135 | NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} ) | |
136 | #define LNK(NAME, get_link) \ | |
61a28784 | 137 | NOD(NAME, (S_IFLNK|S_IRWXUGO), \ |
20cdc894 | 138 | &proc_pid_link_inode_operations, NULL, \ |
631f9c18 AD |
139 | { .proc_get_link = get_link } ) |
140 | #define REG(NAME, MODE, fops) \ | |
141 | NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {}) | |
631f9c18 | 142 | #define ONE(NAME, MODE, show) \ |
be614086 EB |
143 | NOD(NAME, (S_IFREG|(MODE)), \ |
144 | NULL, &proc_single_file_operations, \ | |
631f9c18 | 145 | { .proc_show = show } ) |
1da177e4 | 146 | |
aed54175 VN |
147 | /* |
148 | * Count the number of hardlinks for the pid_entry table, excluding the . | |
149 | * and .. links. | |
150 | */ | |
1270dd8d | 151 | static unsigned int __init pid_entry_nlink(const struct pid_entry *entries, |
aed54175 VN |
152 | unsigned int n) |
153 | { | |
154 | unsigned int i; | |
155 | unsigned int count; | |
156 | ||
1270dd8d | 157 | count = 2; |
aed54175 VN |
158 | for (i = 0; i < n; ++i) { |
159 | if (S_ISDIR(entries[i].mode)) | |
160 | ++count; | |
161 | } | |
162 | ||
163 | return count; | |
164 | } | |
165 | ||
f7ad3c6b | 166 | static int get_task_root(struct task_struct *task, struct path *root) |
1da177e4 | 167 | { |
7c2c7d99 HD |
168 | int result = -ENOENT; |
169 | ||
0494f6ec | 170 | task_lock(task); |
f7ad3c6b MS |
171 | if (task->fs) { |
172 | get_fs_root(task->fs, root); | |
7c2c7d99 HD |
173 | result = 0; |
174 | } | |
0494f6ec | 175 | task_unlock(task); |
7c2c7d99 | 176 | return result; |
0494f6ec MS |
177 | } |
178 | ||
7773fbc5 | 179 | static int proc_cwd_link(struct dentry *dentry, struct path *path) |
0494f6ec | 180 | { |
2b0143b5 | 181 | struct task_struct *task = get_proc_task(d_inode(dentry)); |
0494f6ec | 182 | int result = -ENOENT; |
99f89551 EB |
183 | |
184 | if (task) { | |
f7ad3c6b MS |
185 | task_lock(task); |
186 | if (task->fs) { | |
187 | get_fs_pwd(task->fs, path); | |
188 | result = 0; | |
189 | } | |
190 | task_unlock(task); | |
99f89551 EB |
191 | put_task_struct(task); |
192 | } | |
1da177e4 LT |
193 | return result; |
194 | } | |
195 | ||
7773fbc5 | 196 | static int proc_root_link(struct dentry *dentry, struct path *path) |
1da177e4 | 197 | { |
2b0143b5 | 198 | struct task_struct *task = get_proc_task(d_inode(dentry)); |
1da177e4 | 199 | int result = -ENOENT; |
99f89551 EB |
200 | |
201 | if (task) { | |
f7ad3c6b | 202 | result = get_task_root(task, path); |
99f89551 EB |
203 | put_task_struct(task); |
204 | } | |
1da177e4 LT |
205 | return result; |
206 | } | |
207 | ||
c2c0bb44 AD |
208 | static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, |
209 | size_t _count, loff_t *pos) | |
1da177e4 | 210 | { |
c2c0bb44 AD |
211 | struct task_struct *tsk; |
212 | struct mm_struct *mm; | |
213 | char *page; | |
214 | unsigned long count = _count; | |
215 | unsigned long arg_start, arg_end, env_start, env_end; | |
216 | unsigned long len1, len2, len; | |
217 | unsigned long p; | |
6a6cbe75 | 218 | char __user *buf0 = buf; |
c2c0bb44 | 219 | char c; |
6a6cbe75 | 220 | int rv; |
c2c0bb44 AD |
221 | |
222 | BUG_ON(*pos < 0); | |
223 | ||
224 | tsk = get_proc_task(file_inode(file)); | |
225 | if (!tsk) | |
226 | return -ESRCH; | |
227 | mm = get_task_mm(tsk); | |
228 | put_task_struct(tsk); | |
229 | if (!mm) | |
230 | return 0; | |
231 | /* Check if process spawned far enough to have cmdline. */ | |
232 | if (!mm->env_end) { | |
233 | rv = 0; | |
234 | goto out_mmput; | |
235 | } | |
236 | ||
0ee931c4 | 237 | page = (char *)__get_free_page(GFP_KERNEL); |
c2c0bb44 AD |
238 | if (!page) { |
239 | rv = -ENOMEM; | |
240 | goto out_mmput; | |
241 | } | |
242 | ||
88aa7cc6 | 243 | spin_lock(&mm->arg_lock); |
c2c0bb44 AD |
244 | arg_start = mm->arg_start; |
245 | arg_end = mm->arg_end; | |
246 | env_start = mm->env_start; | |
247 | env_end = mm->env_end; | |
88aa7cc6 | 248 | spin_unlock(&mm->arg_lock); |
c2c0bb44 AD |
249 | |
250 | BUG_ON(arg_start > arg_end); | |
251 | BUG_ON(env_start > env_end); | |
252 | ||
253 | len1 = arg_end - arg_start; | |
254 | len2 = env_end - env_start; | |
255 | ||
3581d458 | 256 | /* Empty ARGV. */ |
6a6cbe75 AD |
257 | if (len1 == 0) |
258 | goto end; | |
259 | ||
2ca66ff7 | 260 | /* |
c2c0bb44 AD |
261 | * Inherently racy -- command line shares address space |
262 | * with code and data. | |
2ca66ff7 | 263 | */ |
6a6cbe75 AD |
264 | if (access_remote_vm(mm, arg_end - 1, &c, 1, FOLL_ANON) != 1) |
265 | goto end; | |
c2c0bb44 AD |
266 | |
267 | if (c == '\0') { | |
268 | /* Command line (set of strings) occupies whole ARGV. */ | |
269 | if (len1 <= *pos) | |
6a6cbe75 | 270 | goto end; |
c2c0bb44 AD |
271 | |
272 | p = arg_start + *pos; | |
273 | len = len1 - *pos; | |
274 | while (count > 0 && len > 0) { | |
b42262af AD |
275 | unsigned int nr_read; |
276 | ||
277 | nr_read = min3(count, len, PAGE_SIZE); | |
278 | nr_read = access_remote_vm(mm, p, page, nr_read, FOLL_ANON); | |
279 | if (nr_read == 0) | |
6a6cbe75 | 280 | goto end; |
c2c0bb44 AD |
281 | |
282 | if (copy_to_user(buf, page, nr_read)) { | |
283 | rv = -EFAULT; | |
284 | goto out_free_page; | |
285 | } | |
286 | ||
287 | p += nr_read; | |
288 | len -= nr_read; | |
289 | buf += nr_read; | |
290 | count -= nr_read; | |
c2c0bb44 AD |
291 | } |
292 | } else { | |
c2c0bb44 AD |
293 | /* |
294 | * Command line (1 string) occupies ARGV and | |
295 | * extends into ENVP. | |
296 | */ | |
a0a07b87 AD |
297 | struct { |
298 | unsigned long p; | |
299 | unsigned long len; | |
300 | } cmdline[2] = { | |
301 | { .p = arg_start, .len = len1 }, | |
302 | { .p = env_start, .len = len2 }, | |
303 | }; | |
304 | loff_t pos1 = *pos; | |
305 | unsigned int i; | |
306 | ||
307 | i = 0; | |
308 | while (i < 2 && pos1 >= cmdline[i].len) { | |
309 | pos1 -= cmdline[i].len; | |
310 | i++; | |
c2c0bb44 | 311 | } |
a0a07b87 AD |
312 | while (i < 2) { |
313 | p = cmdline[i].p + pos1; | |
314 | len = cmdline[i].len - pos1; | |
315 | while (count > 0 && len > 0) { | |
6a6b9c4c | 316 | unsigned int nr_read, nr_write; |
a0a07b87 | 317 | |
b42262af AD |
318 | nr_read = min3(count, len, PAGE_SIZE); |
319 | nr_read = access_remote_vm(mm, p, page, nr_read, FOLL_ANON); | |
320 | if (nr_read == 0) | |
6a6cbe75 | 321 | goto end; |
a0a07b87 AD |
322 | |
323 | /* | |
324 | * Command line can be shorter than whole ARGV | |
325 | * even if last "marker" byte says it is not. | |
326 | */ | |
6a6b9c4c | 327 | nr_write = strnlen(page, nr_read); |
a0a07b87 | 328 | |
6a6b9c4c | 329 | if (copy_to_user(buf, page, nr_write)) { |
a0a07b87 AD |
330 | rv = -EFAULT; |
331 | goto out_free_page; | |
332 | } | |
333 | ||
6a6b9c4c AD |
334 | p += nr_write; |
335 | len -= nr_write; | |
336 | buf += nr_write; | |
337 | count -= nr_write; | |
a0a07b87 | 338 | |
6a6b9c4c | 339 | if (nr_write < nr_read) |
6a6cbe75 | 340 | goto end; |
c2c0bb44 AD |
341 | } |
342 | ||
a0a07b87 AD |
343 | /* Only first chunk can be read partially. */ |
344 | pos1 = 0; | |
345 | i++; | |
c2c0bb44 | 346 | } |
c2c0bb44 AD |
347 | } |
348 | ||
6a6cbe75 AD |
349 | end: |
350 | *pos += buf - buf0; | |
351 | rv = buf - buf0; | |
c2c0bb44 AD |
352 | out_free_page: |
353 | free_page((unsigned long)page); | |
354 | out_mmput: | |
355 | mmput(mm); | |
c2c0bb44 | 356 | return rv; |
1da177e4 LT |
357 | } |
358 | ||
c2c0bb44 AD |
359 | static const struct file_operations proc_pid_cmdline_ops = { |
360 | .read = proc_pid_cmdline_read, | |
361 | .llseek = generic_file_llseek, | |
362 | }; | |
363 | ||
1da177e4 LT |
364 | #ifdef CONFIG_KALLSYMS |
365 | /* | |
366 | * Provides a wchan file via kallsyms in a proper one-value-per-file format. | |
367 | * Returns the resolved symbol. If that fails, simply return the address. | |
368 | */ | |
edfcd606 AD |
369 | static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns, |
370 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 371 | { |
ffb45122 | 372 | unsigned long wchan; |
9281acea | 373 | char symname[KSYM_NAME_LEN]; |
1da177e4 | 374 | |
24b2ec21 AD |
375 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
376 | goto print0; | |
1da177e4 | 377 | |
24b2ec21 AD |
378 | wchan = get_wchan(task); |
379 | if (wchan && !lookup_symbol_name(wchan, symname)) { | |
21dae0ad | 380 | seq_puts(m, symname); |
24b2ec21 AD |
381 | return 0; |
382 | } | |
25ce3191 | 383 | |
24b2ec21 AD |
384 | print0: |
385 | seq_putc(m, '0'); | |
25ce3191 | 386 | return 0; |
1da177e4 LT |
387 | } |
388 | #endif /* CONFIG_KALLSYMS */ | |
389 | ||
a9712bc1 AV |
390 | static int lock_trace(struct task_struct *task) |
391 | { | |
392 | int err = mutex_lock_killable(&task->signal->cred_guard_mutex); | |
393 | if (err) | |
394 | return err; | |
caaee623 | 395 | if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) { |
a9712bc1 AV |
396 | mutex_unlock(&task->signal->cred_guard_mutex); |
397 | return -EPERM; | |
398 | } | |
399 | return 0; | |
400 | } | |
401 | ||
402 | static void unlock_trace(struct task_struct *task) | |
403 | { | |
404 | mutex_unlock(&task->signal->cred_guard_mutex); | |
405 | } | |
406 | ||
2ec220e2 KC |
407 | #ifdef CONFIG_STACKTRACE |
408 | ||
409 | #define MAX_STACK_TRACE_DEPTH 64 | |
410 | ||
411 | static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns, | |
412 | struct pid *pid, struct task_struct *task) | |
413 | { | |
414 | struct stack_trace trace; | |
415 | unsigned long *entries; | |
a9712bc1 | 416 | int err; |
2ec220e2 KC |
417 | int i; |
418 | ||
419 | entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL); | |
420 | if (!entries) | |
421 | return -ENOMEM; | |
422 | ||
423 | trace.nr_entries = 0; | |
424 | trace.max_entries = MAX_STACK_TRACE_DEPTH; | |
425 | trace.entries = entries; | |
426 | trace.skip = 0; | |
2ec220e2 | 427 | |
a9712bc1 AV |
428 | err = lock_trace(task); |
429 | if (!err) { | |
430 | save_stack_trace_tsk(task, &trace); | |
431 | ||
432 | for (i = 0; i < trace.nr_entries; i++) { | |
8f5abe84 | 433 | seq_printf(m, "[<0>] %pB\n", (void *)entries[i]); |
a9712bc1 AV |
434 | } |
435 | unlock_trace(task); | |
2ec220e2 KC |
436 | } |
437 | kfree(entries); | |
438 | ||
a9712bc1 | 439 | return err; |
2ec220e2 KC |
440 | } |
441 | #endif | |
442 | ||
5968cece | 443 | #ifdef CONFIG_SCHED_INFO |
1da177e4 LT |
444 | /* |
445 | * Provides /proc/PID/schedstat | |
446 | */ | |
f6e826ca AD |
447 | static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns, |
448 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 449 | { |
5968cece NR |
450 | if (unlikely(!sched_info_on())) |
451 | seq_printf(m, "0 0 0\n"); | |
452 | else | |
453 | seq_printf(m, "%llu %llu %lu\n", | |
25ce3191 JP |
454 | (unsigned long long)task->se.sum_exec_runtime, |
455 | (unsigned long long)task->sched_info.run_delay, | |
456 | task->sched_info.pcount); | |
457 | ||
458 | return 0; | |
1da177e4 LT |
459 | } |
460 | #endif | |
461 | ||
9745512c AV |
462 | #ifdef CONFIG_LATENCYTOP |
463 | static int lstats_show_proc(struct seq_file *m, void *v) | |
464 | { | |
465 | int i; | |
13d77c37 HS |
466 | struct inode *inode = m->private; |
467 | struct task_struct *task = get_proc_task(inode); | |
9745512c | 468 | |
13d77c37 HS |
469 | if (!task) |
470 | return -ESRCH; | |
471 | seq_puts(m, "Latency Top version : v0.1\n"); | |
9745512c | 472 | for (i = 0; i < 32; i++) { |
34e49d4f JP |
473 | struct latency_record *lr = &task->latency_record[i]; |
474 | if (lr->backtrace[0]) { | |
9745512c | 475 | int q; |
34e49d4f JP |
476 | seq_printf(m, "%i %li %li", |
477 | lr->count, lr->time, lr->max); | |
9745512c | 478 | for (q = 0; q < LT_BACKTRACEDEPTH; q++) { |
34e49d4f JP |
479 | unsigned long bt = lr->backtrace[q]; |
480 | if (!bt) | |
9745512c | 481 | break; |
34e49d4f | 482 | if (bt == ULONG_MAX) |
9745512c | 483 | break; |
34e49d4f | 484 | seq_printf(m, " %ps", (void *)bt); |
9745512c | 485 | } |
9d6de12f | 486 | seq_putc(m, '\n'); |
9745512c AV |
487 | } |
488 | ||
489 | } | |
13d77c37 | 490 | put_task_struct(task); |
9745512c AV |
491 | return 0; |
492 | } | |
493 | ||
494 | static int lstats_open(struct inode *inode, struct file *file) | |
495 | { | |
13d77c37 | 496 | return single_open(file, lstats_show_proc, inode); |
d6643d12 HS |
497 | } |
498 | ||
9745512c AV |
499 | static ssize_t lstats_write(struct file *file, const char __user *buf, |
500 | size_t count, loff_t *offs) | |
501 | { | |
496ad9aa | 502 | struct task_struct *task = get_proc_task(file_inode(file)); |
9745512c | 503 | |
13d77c37 HS |
504 | if (!task) |
505 | return -ESRCH; | |
9745512c | 506 | clear_all_latency_tracing(task); |
13d77c37 | 507 | put_task_struct(task); |
9745512c AV |
508 | |
509 | return count; | |
510 | } | |
511 | ||
512 | static const struct file_operations proc_lstats_operations = { | |
513 | .open = lstats_open, | |
514 | .read = seq_read, | |
515 | .write = lstats_write, | |
516 | .llseek = seq_lseek, | |
13d77c37 | 517 | .release = single_release, |
9745512c AV |
518 | }; |
519 | ||
520 | #endif | |
521 | ||
6ba51e37 AD |
522 | static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns, |
523 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 524 | { |
a7f638f9 | 525 | unsigned long totalpages = totalram_pages + total_swap_pages; |
b95c35e7 | 526 | unsigned long points = 0; |
1da177e4 | 527 | |
ef419398 ON |
528 | points = oom_badness(task, NULL, NULL, totalpages) * |
529 | 1000 / totalpages; | |
25ce3191 JP |
530 | seq_printf(m, "%lu\n", points); |
531 | ||
532 | return 0; | |
1da177e4 LT |
533 | } |
534 | ||
d85f50d5 | 535 | struct limit_names { |
cedbccab AD |
536 | const char *name; |
537 | const char *unit; | |
d85f50d5 NH |
538 | }; |
539 | ||
540 | static const struct limit_names lnames[RLIM_NLIMITS] = { | |
cff4edb5 | 541 | [RLIMIT_CPU] = {"Max cpu time", "seconds"}, |
d85f50d5 NH |
542 | [RLIMIT_FSIZE] = {"Max file size", "bytes"}, |
543 | [RLIMIT_DATA] = {"Max data size", "bytes"}, | |
544 | [RLIMIT_STACK] = {"Max stack size", "bytes"}, | |
545 | [RLIMIT_CORE] = {"Max core file size", "bytes"}, | |
546 | [RLIMIT_RSS] = {"Max resident set", "bytes"}, | |
547 | [RLIMIT_NPROC] = {"Max processes", "processes"}, | |
548 | [RLIMIT_NOFILE] = {"Max open files", "files"}, | |
549 | [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"}, | |
550 | [RLIMIT_AS] = {"Max address space", "bytes"}, | |
551 | [RLIMIT_LOCKS] = {"Max file locks", "locks"}, | |
552 | [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"}, | |
553 | [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"}, | |
554 | [RLIMIT_NICE] = {"Max nice priority", NULL}, | |
555 | [RLIMIT_RTPRIO] = {"Max realtime priority", NULL}, | |
8808117c | 556 | [RLIMIT_RTTIME] = {"Max realtime timeout", "us"}, |
d85f50d5 NH |
557 | }; |
558 | ||
559 | /* Display limits for a process */ | |
1c963eb1 AD |
560 | static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns, |
561 | struct pid *pid, struct task_struct *task) | |
d85f50d5 NH |
562 | { |
563 | unsigned int i; | |
d85f50d5 | 564 | unsigned long flags; |
d85f50d5 NH |
565 | |
566 | struct rlimit rlim[RLIM_NLIMITS]; | |
567 | ||
a6bebbc8 | 568 | if (!lock_task_sighand(task, &flags)) |
d85f50d5 | 569 | return 0; |
d85f50d5 NH |
570 | memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS); |
571 | unlock_task_sighand(task, &flags); | |
d85f50d5 NH |
572 | |
573 | /* | |
574 | * print the file header | |
575 | */ | |
1c963eb1 | 576 | seq_printf(m, "%-25s %-20s %-20s %-10s\n", |
25ce3191 | 577 | "Limit", "Soft Limit", "Hard Limit", "Units"); |
d85f50d5 NH |
578 | |
579 | for (i = 0; i < RLIM_NLIMITS; i++) { | |
580 | if (rlim[i].rlim_cur == RLIM_INFINITY) | |
1c963eb1 | 581 | seq_printf(m, "%-25s %-20s ", |
25ce3191 | 582 | lnames[i].name, "unlimited"); |
d85f50d5 | 583 | else |
1c963eb1 | 584 | seq_printf(m, "%-25s %-20lu ", |
25ce3191 | 585 | lnames[i].name, rlim[i].rlim_cur); |
d85f50d5 NH |
586 | |
587 | if (rlim[i].rlim_max == RLIM_INFINITY) | |
1c963eb1 | 588 | seq_printf(m, "%-20s ", "unlimited"); |
d85f50d5 | 589 | else |
1c963eb1 | 590 | seq_printf(m, "%-20lu ", rlim[i].rlim_max); |
d85f50d5 NH |
591 | |
592 | if (lnames[i].unit) | |
1c963eb1 | 593 | seq_printf(m, "%-10s\n", lnames[i].unit); |
d85f50d5 | 594 | else |
1c963eb1 | 595 | seq_putc(m, '\n'); |
d85f50d5 NH |
596 | } |
597 | ||
1c963eb1 | 598 | return 0; |
d85f50d5 NH |
599 | } |
600 | ||
ebcb6734 | 601 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 AD |
602 | static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns, |
603 | struct pid *pid, struct task_struct *task) | |
ebcb6734 RM |
604 | { |
605 | long nr; | |
606 | unsigned long args[6], sp, pc; | |
25ce3191 JP |
607 | int res; |
608 | ||
609 | res = lock_trace(task); | |
a9712bc1 AV |
610 | if (res) |
611 | return res; | |
ebcb6734 RM |
612 | |
613 | if (task_current_syscall(task, &nr, args, 6, &sp, &pc)) | |
09d93bd6 | 614 | seq_puts(m, "running\n"); |
a9712bc1 | 615 | else if (nr < 0) |
09d93bd6 | 616 | seq_printf(m, "%ld 0x%lx 0x%lx\n", nr, sp, pc); |
a9712bc1 | 617 | else |
09d93bd6 | 618 | seq_printf(m, |
ebcb6734 RM |
619 | "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n", |
620 | nr, | |
621 | args[0], args[1], args[2], args[3], args[4], args[5], | |
622 | sp, pc); | |
a9712bc1 | 623 | unlock_trace(task); |
25ce3191 JP |
624 | |
625 | return 0; | |
ebcb6734 RM |
626 | } |
627 | #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */ | |
628 | ||
1da177e4 LT |
629 | /************************************************************************/ |
630 | /* Here the fs part begins */ | |
631 | /************************************************************************/ | |
632 | ||
633 | /* permission checks */ | |
778c1144 | 634 | static int proc_fd_access_allowed(struct inode *inode) |
1da177e4 | 635 | { |
778c1144 EB |
636 | struct task_struct *task; |
637 | int allowed = 0; | |
df26c40e EB |
638 | /* Allow access to a task's file descriptors if it is us or we |
639 | * may use ptrace attach to the process and find out that | |
640 | * information. | |
778c1144 EB |
641 | */ |
642 | task = get_proc_task(inode); | |
df26c40e | 643 | if (task) { |
caaee623 | 644 | allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS); |
778c1144 | 645 | put_task_struct(task); |
df26c40e | 646 | } |
778c1144 | 647 | return allowed; |
1da177e4 LT |
648 | } |
649 | ||
6b4e306a | 650 | int proc_setattr(struct dentry *dentry, struct iattr *attr) |
6d76fa58 LT |
651 | { |
652 | int error; | |
2b0143b5 | 653 | struct inode *inode = d_inode(dentry); |
6d76fa58 LT |
654 | |
655 | if (attr->ia_valid & ATTR_MODE) | |
656 | return -EPERM; | |
657 | ||
31051c85 | 658 | error = setattr_prepare(dentry, attr); |
1025774c CH |
659 | if (error) |
660 | return error; | |
661 | ||
1025774c CH |
662 | setattr_copy(inode, attr); |
663 | mark_inode_dirty(inode); | |
664 | return 0; | |
6d76fa58 LT |
665 | } |
666 | ||
0499680a VK |
667 | /* |
668 | * May current process learn task's sched/cmdline info (for hide_pid_min=1) | |
669 | * or euid/egid (for hide_pid_min=2)? | |
670 | */ | |
671 | static bool has_pid_permissions(struct pid_namespace *pid, | |
672 | struct task_struct *task, | |
673 | int hide_pid_min) | |
674 | { | |
675 | if (pid->hide_pid < hide_pid_min) | |
676 | return true; | |
677 | if (in_group_p(pid->pid_gid)) | |
678 | return true; | |
caaee623 | 679 | return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS); |
0499680a VK |
680 | } |
681 | ||
682 | ||
683 | static int proc_pid_permission(struct inode *inode, int mask) | |
684 | { | |
76f668be | 685 | struct pid_namespace *pid = proc_pid_ns(inode); |
0499680a VK |
686 | struct task_struct *task; |
687 | bool has_perms; | |
688 | ||
689 | task = get_proc_task(inode); | |
a2ef990a XF |
690 | if (!task) |
691 | return -ESRCH; | |
796f571b | 692 | has_perms = has_pid_permissions(pid, task, HIDEPID_NO_ACCESS); |
0499680a VK |
693 | put_task_struct(task); |
694 | ||
695 | if (!has_perms) { | |
796f571b | 696 | if (pid->hide_pid == HIDEPID_INVISIBLE) { |
0499680a VK |
697 | /* |
698 | * Let's make getdents(), stat(), and open() | |
699 | * consistent with each other. If a process | |
700 | * may not stat() a file, it shouldn't be seen | |
701 | * in procfs at all. | |
702 | */ | |
703 | return -ENOENT; | |
704 | } | |
705 | ||
706 | return -EPERM; | |
707 | } | |
708 | return generic_permission(inode, mask); | |
709 | } | |
710 | ||
711 | ||
712 | ||
c5ef1c42 | 713 | static const struct inode_operations proc_def_inode_operations = { |
6d76fa58 LT |
714 | .setattr = proc_setattr, |
715 | }; | |
716 | ||
be614086 EB |
717 | static int proc_single_show(struct seq_file *m, void *v) |
718 | { | |
719 | struct inode *inode = m->private; | |
76f668be CH |
720 | struct pid_namespace *ns = proc_pid_ns(inode); |
721 | struct pid *pid = proc_pid(inode); | |
be614086 EB |
722 | struct task_struct *task; |
723 | int ret; | |
724 | ||
be614086 EB |
725 | task = get_pid_task(pid, PIDTYPE_PID); |
726 | if (!task) | |
727 | return -ESRCH; | |
728 | ||
729 | ret = PROC_I(inode)->op.proc_show(m, ns, pid, task); | |
730 | ||
731 | put_task_struct(task); | |
732 | return ret; | |
733 | } | |
734 | ||
735 | static int proc_single_open(struct inode *inode, struct file *filp) | |
736 | { | |
c6a34058 | 737 | return single_open(filp, proc_single_show, inode); |
be614086 EB |
738 | } |
739 | ||
740 | static const struct file_operations proc_single_file_operations = { | |
741 | .open = proc_single_open, | |
742 | .read = seq_read, | |
743 | .llseek = seq_lseek, | |
744 | .release = single_release, | |
745 | }; | |
746 | ||
5381e169 ON |
747 | |
748 | struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode) | |
1da177e4 | 749 | { |
5381e169 ON |
750 | struct task_struct *task = get_proc_task(inode); |
751 | struct mm_struct *mm = ERR_PTR(-ESRCH); | |
e268337d | 752 | |
5381e169 | 753 | if (task) { |
caaee623 | 754 | mm = mm_access(task, mode | PTRACE_MODE_FSCREDS); |
5381e169 | 755 | put_task_struct(task); |
e268337d | 756 | |
5381e169 ON |
757 | if (!IS_ERR_OR_NULL(mm)) { |
758 | /* ensure this mm_struct can't be freed */ | |
f1f10076 | 759 | mmgrab(mm); |
5381e169 ON |
760 | /* but do not pin its memory */ |
761 | mmput(mm); | |
762 | } | |
763 | } | |
764 | ||
765 | return mm; | |
766 | } | |
767 | ||
768 | static int __mem_open(struct inode *inode, struct file *file, unsigned int mode) | |
769 | { | |
770 | struct mm_struct *mm = proc_mem_open(inode, mode); | |
e268337d LT |
771 | |
772 | if (IS_ERR(mm)) | |
773 | return PTR_ERR(mm); | |
774 | ||
e268337d | 775 | file->private_data = mm; |
1da177e4 LT |
776 | return 0; |
777 | } | |
778 | ||
b409e578 CW |
779 | static int mem_open(struct inode *inode, struct file *file) |
780 | { | |
bc452b4b DH |
781 | int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH); |
782 | ||
783 | /* OK to pass negative loff_t, we can catch out-of-range */ | |
784 | file->f_mode |= FMODE_UNSIGNED_OFFSET; | |
785 | ||
786 | return ret; | |
b409e578 CW |
787 | } |
788 | ||
572d34b9 ON |
789 | static ssize_t mem_rw(struct file *file, char __user *buf, |
790 | size_t count, loff_t *ppos, int write) | |
1da177e4 | 791 | { |
e268337d | 792 | struct mm_struct *mm = file->private_data; |
572d34b9 ON |
793 | unsigned long addr = *ppos; |
794 | ssize_t copied; | |
1da177e4 | 795 | char *page; |
272ddc8b | 796 | unsigned int flags; |
1da177e4 | 797 | |
e268337d LT |
798 | if (!mm) |
799 | return 0; | |
99f89551 | 800 | |
0ee931c4 | 801 | page = (char *)__get_free_page(GFP_KERNEL); |
30cd8903 | 802 | if (!page) |
e268337d | 803 | return -ENOMEM; |
1da177e4 | 804 | |
f7ca54f4 | 805 | copied = 0; |
388f7934 | 806 | if (!mmget_not_zero(mm)) |
6d08f2c7 ON |
807 | goto free; |
808 | ||
f511c0b1 | 809 | flags = FOLL_FORCE | (write ? FOLL_WRITE : 0); |
6347e8d5 | 810 | |
1da177e4 | 811 | while (count > 0) { |
572d34b9 | 812 | int this_len = min_t(int, count, PAGE_SIZE); |
1da177e4 | 813 | |
572d34b9 | 814 | if (write && copy_from_user(page, buf, this_len)) { |
1da177e4 LT |
815 | copied = -EFAULT; |
816 | break; | |
817 | } | |
572d34b9 | 818 | |
6347e8d5 | 819 | this_len = access_remote_vm(mm, addr, page, this_len, flags); |
572d34b9 | 820 | if (!this_len) { |
1da177e4 LT |
821 | if (!copied) |
822 | copied = -EIO; | |
823 | break; | |
824 | } | |
572d34b9 ON |
825 | |
826 | if (!write && copy_to_user(buf, page, this_len)) { | |
827 | copied = -EFAULT; | |
828 | break; | |
829 | } | |
830 | ||
831 | buf += this_len; | |
832 | addr += this_len; | |
833 | copied += this_len; | |
834 | count -= this_len; | |
1da177e4 | 835 | } |
572d34b9 | 836 | *ppos = addr; |
30cd8903 | 837 | |
6d08f2c7 ON |
838 | mmput(mm); |
839 | free: | |
30cd8903 | 840 | free_page((unsigned long) page); |
1da177e4 LT |
841 | return copied; |
842 | } | |
1da177e4 | 843 | |
572d34b9 ON |
844 | static ssize_t mem_read(struct file *file, char __user *buf, |
845 | size_t count, loff_t *ppos) | |
846 | { | |
847 | return mem_rw(file, buf, count, ppos, 0); | |
848 | } | |
849 | ||
850 | static ssize_t mem_write(struct file *file, const char __user *buf, | |
851 | size_t count, loff_t *ppos) | |
852 | { | |
853 | return mem_rw(file, (char __user*)buf, count, ppos, 1); | |
854 | } | |
855 | ||
85863e47 | 856 | loff_t mem_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
857 | { |
858 | switch (orig) { | |
859 | case 0: | |
860 | file->f_pos = offset; | |
861 | break; | |
862 | case 1: | |
863 | file->f_pos += offset; | |
864 | break; | |
865 | default: | |
866 | return -EINVAL; | |
867 | } | |
868 | force_successful_syscall_return(); | |
869 | return file->f_pos; | |
870 | } | |
871 | ||
e268337d LT |
872 | static int mem_release(struct inode *inode, struct file *file) |
873 | { | |
874 | struct mm_struct *mm = file->private_data; | |
71879d3c | 875 | if (mm) |
6d08f2c7 | 876 | mmdrop(mm); |
e268337d LT |
877 | return 0; |
878 | } | |
879 | ||
00977a59 | 880 | static const struct file_operations proc_mem_operations = { |
1da177e4 LT |
881 | .llseek = mem_lseek, |
882 | .read = mem_read, | |
883 | .write = mem_write, | |
884 | .open = mem_open, | |
e268337d | 885 | .release = mem_release, |
1da177e4 LT |
886 | }; |
887 | ||
b409e578 CW |
888 | static int environ_open(struct inode *inode, struct file *file) |
889 | { | |
890 | return __mem_open(inode, file, PTRACE_MODE_READ); | |
891 | } | |
892 | ||
315e28c8 JP |
893 | static ssize_t environ_read(struct file *file, char __user *buf, |
894 | size_t count, loff_t *ppos) | |
895 | { | |
315e28c8 JP |
896 | char *page; |
897 | unsigned long src = *ppos; | |
b409e578 CW |
898 | int ret = 0; |
899 | struct mm_struct *mm = file->private_data; | |
a3b609ef | 900 | unsigned long env_start, env_end; |
315e28c8 | 901 | |
8148a73c MK |
902 | /* Ensure the process spawned far enough to have an environment. */ |
903 | if (!mm || !mm->env_end) | |
b409e578 | 904 | return 0; |
315e28c8 | 905 | |
0ee931c4 | 906 | page = (char *)__get_free_page(GFP_KERNEL); |
315e28c8 | 907 | if (!page) |
b409e578 | 908 | return -ENOMEM; |
315e28c8 | 909 | |
d6f64b89 | 910 | ret = 0; |
388f7934 | 911 | if (!mmget_not_zero(mm)) |
b409e578 | 912 | goto free; |
a3b609ef | 913 | |
88aa7cc6 | 914 | spin_lock(&mm->arg_lock); |
a3b609ef MG |
915 | env_start = mm->env_start; |
916 | env_end = mm->env_end; | |
88aa7cc6 | 917 | spin_unlock(&mm->arg_lock); |
a3b609ef | 918 | |
315e28c8 | 919 | while (count > 0) { |
e8905ec2 DH |
920 | size_t this_len, max_len; |
921 | int retval; | |
315e28c8 | 922 | |
a3b609ef | 923 | if (src >= (env_end - env_start)) |
315e28c8 JP |
924 | break; |
925 | ||
a3b609ef | 926 | this_len = env_end - (env_start + src); |
e8905ec2 DH |
927 | |
928 | max_len = min_t(size_t, PAGE_SIZE, count); | |
929 | this_len = min(max_len, this_len); | |
315e28c8 | 930 | |
7f7ccc2c | 931 | retval = access_remote_vm(mm, (env_start + src), page, this_len, FOLL_ANON); |
315e28c8 JP |
932 | |
933 | if (retval <= 0) { | |
934 | ret = retval; | |
935 | break; | |
936 | } | |
937 | ||
938 | if (copy_to_user(buf, page, retval)) { | |
939 | ret = -EFAULT; | |
940 | break; | |
941 | } | |
942 | ||
943 | ret += retval; | |
944 | src += retval; | |
945 | buf += retval; | |
946 | count -= retval; | |
947 | } | |
948 | *ppos = src; | |
315e28c8 | 949 | mmput(mm); |
b409e578 CW |
950 | |
951 | free: | |
315e28c8 | 952 | free_page((unsigned long) page); |
315e28c8 JP |
953 | return ret; |
954 | } | |
955 | ||
956 | static const struct file_operations proc_environ_operations = { | |
b409e578 | 957 | .open = environ_open, |
315e28c8 | 958 | .read = environ_read, |
87df8424 | 959 | .llseek = generic_file_llseek, |
b409e578 | 960 | .release = mem_release, |
315e28c8 JP |
961 | }; |
962 | ||
c5317167 AV |
963 | static int auxv_open(struct inode *inode, struct file *file) |
964 | { | |
965 | return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS); | |
966 | } | |
967 | ||
968 | static ssize_t auxv_read(struct file *file, char __user *buf, | |
969 | size_t count, loff_t *ppos) | |
970 | { | |
971 | struct mm_struct *mm = file->private_data; | |
972 | unsigned int nwords = 0; | |
06b2849d LY |
973 | |
974 | if (!mm) | |
975 | return 0; | |
c5317167 AV |
976 | do { |
977 | nwords += 2; | |
978 | } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */ | |
979 | return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv, | |
980 | nwords * sizeof(mm->saved_auxv[0])); | |
981 | } | |
982 | ||
983 | static const struct file_operations proc_auxv_operations = { | |
984 | .open = auxv_open, | |
985 | .read = auxv_read, | |
986 | .llseek = generic_file_llseek, | |
987 | .release = mem_release, | |
988 | }; | |
989 | ||
fa0cbbf1 DR |
990 | static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count, |
991 | loff_t *ppos) | |
992 | { | |
496ad9aa | 993 | struct task_struct *task = get_proc_task(file_inode(file)); |
fa0cbbf1 DR |
994 | char buffer[PROC_NUMBUF]; |
995 | int oom_adj = OOM_ADJUST_MIN; | |
996 | size_t len; | |
fa0cbbf1 DR |
997 | |
998 | if (!task) | |
999 | return -ESRCH; | |
f913da59 MH |
1000 | if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX) |
1001 | oom_adj = OOM_ADJUST_MAX; | |
1002 | else | |
1003 | oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) / | |
1004 | OOM_SCORE_ADJ_MAX; | |
fa0cbbf1 DR |
1005 | put_task_struct(task); |
1006 | len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj); | |
1007 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
1008 | } | |
1009 | ||
1d5f0acb MH |
1010 | static int __set_oom_adj(struct file *file, int oom_adj, bool legacy) |
1011 | { | |
1012 | static DEFINE_MUTEX(oom_adj_mutex); | |
44a70ade | 1013 | struct mm_struct *mm = NULL; |
1d5f0acb MH |
1014 | struct task_struct *task; |
1015 | int err = 0; | |
1016 | ||
1017 | task = get_proc_task(file_inode(file)); | |
1018 | if (!task) | |
1019 | return -ESRCH; | |
1020 | ||
1021 | mutex_lock(&oom_adj_mutex); | |
1022 | if (legacy) { | |
1023 | if (oom_adj < task->signal->oom_score_adj && | |
1024 | !capable(CAP_SYS_RESOURCE)) { | |
1025 | err = -EACCES; | |
1026 | goto err_unlock; | |
1027 | } | |
1028 | /* | |
1029 | * /proc/pid/oom_adj is provided for legacy purposes, ask users to use | |
1030 | * /proc/pid/oom_score_adj instead. | |
1031 | */ | |
1032 | pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n", | |
1033 | current->comm, task_pid_nr(current), task_pid_nr(task), | |
1034 | task_pid_nr(task)); | |
1035 | } else { | |
1036 | if ((short)oom_adj < task->signal->oom_score_adj_min && | |
1037 | !capable(CAP_SYS_RESOURCE)) { | |
1038 | err = -EACCES; | |
1039 | goto err_unlock; | |
1040 | } | |
1041 | } | |
1042 | ||
44a70ade MH |
1043 | /* |
1044 | * Make sure we will check other processes sharing the mm if this is | |
1045 | * not vfrok which wants its own oom_score_adj. | |
1046 | * pin the mm so it doesn't go away and get reused after task_unlock | |
1047 | */ | |
1048 | if (!task->vfork_done) { | |
1049 | struct task_struct *p = find_lock_task_mm(task); | |
1050 | ||
1051 | if (p) { | |
1052 | if (atomic_read(&p->mm->mm_users) > 1) { | |
1053 | mm = p->mm; | |
f1f10076 | 1054 | mmgrab(mm); |
44a70ade MH |
1055 | } |
1056 | task_unlock(p); | |
1057 | } | |
1058 | } | |
1059 | ||
1d5f0acb MH |
1060 | task->signal->oom_score_adj = oom_adj; |
1061 | if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE)) | |
1062 | task->signal->oom_score_adj_min = (short)oom_adj; | |
1063 | trace_oom_score_adj_update(task); | |
44a70ade MH |
1064 | |
1065 | if (mm) { | |
1066 | struct task_struct *p; | |
1067 | ||
1068 | rcu_read_lock(); | |
1069 | for_each_process(p) { | |
1070 | if (same_thread_group(task, p)) | |
1071 | continue; | |
1072 | ||
1073 | /* do not touch kernel threads or the global init */ | |
1074 | if (p->flags & PF_KTHREAD || is_global_init(p)) | |
1075 | continue; | |
1076 | ||
1077 | task_lock(p); | |
1078 | if (!p->vfork_done && process_shares_mm(p, mm)) { | |
1079 | pr_info("updating oom_score_adj for %d (%s) from %d to %d because it shares mm with %d (%s). Report if this is unexpected.\n", | |
1080 | task_pid_nr(p), p->comm, | |
1081 | p->signal->oom_score_adj, oom_adj, | |
1082 | task_pid_nr(task), task->comm); | |
1083 | p->signal->oom_score_adj = oom_adj; | |
1084 | if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE)) | |
1085 | p->signal->oom_score_adj_min = (short)oom_adj; | |
1086 | } | |
1087 | task_unlock(p); | |
1088 | } | |
1089 | rcu_read_unlock(); | |
1090 | mmdrop(mm); | |
1091 | } | |
1d5f0acb MH |
1092 | err_unlock: |
1093 | mutex_unlock(&oom_adj_mutex); | |
1094 | put_task_struct(task); | |
1095 | return err; | |
1096 | } | |
f913da59 | 1097 | |
b72bdfa7 DR |
1098 | /* |
1099 | * /proc/pid/oom_adj exists solely for backwards compatibility with previous | |
1100 | * kernels. The effective policy is defined by oom_score_adj, which has a | |
1101 | * different scale: oom_adj grew exponentially and oom_score_adj grows linearly. | |
1102 | * Values written to oom_adj are simply mapped linearly to oom_score_adj. | |
1103 | * Processes that become oom disabled via oom_adj will still be oom disabled | |
1104 | * with this implementation. | |
1105 | * | |
1106 | * oom_adj cannot be removed since existing userspace binaries use it. | |
1107 | */ | |
fa0cbbf1 DR |
1108 | static ssize_t oom_adj_write(struct file *file, const char __user *buf, |
1109 | size_t count, loff_t *ppos) | |
1110 | { | |
fa0cbbf1 DR |
1111 | char buffer[PROC_NUMBUF]; |
1112 | int oom_adj; | |
fa0cbbf1 DR |
1113 | int err; |
1114 | ||
1115 | memset(buffer, 0, sizeof(buffer)); | |
1116 | if (count > sizeof(buffer) - 1) | |
1117 | count = sizeof(buffer) - 1; | |
1118 | if (copy_from_user(buffer, buf, count)) { | |
1119 | err = -EFAULT; | |
1120 | goto out; | |
1121 | } | |
1122 | ||
1123 | err = kstrtoint(strstrip(buffer), 0, &oom_adj); | |
1124 | if (err) | |
1125 | goto out; | |
1126 | if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) && | |
1127 | oom_adj != OOM_DISABLE) { | |
1128 | err = -EINVAL; | |
1129 | goto out; | |
1130 | } | |
1131 | ||
fa0cbbf1 DR |
1132 | /* |
1133 | * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum | |
1134 | * value is always attainable. | |
1135 | */ | |
1136 | if (oom_adj == OOM_ADJUST_MAX) | |
1137 | oom_adj = OOM_SCORE_ADJ_MAX; | |
1138 | else | |
1139 | oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE; | |
1140 | ||
1d5f0acb | 1141 | err = __set_oom_adj(file, oom_adj, true); |
fa0cbbf1 DR |
1142 | out: |
1143 | return err < 0 ? err : count; | |
1144 | } | |
1145 | ||
1146 | static const struct file_operations proc_oom_adj_operations = { | |
1147 | .read = oom_adj_read, | |
1148 | .write = oom_adj_write, | |
1149 | .llseek = generic_file_llseek, | |
1150 | }; | |
1151 | ||
a63d83f4 DR |
1152 | static ssize_t oom_score_adj_read(struct file *file, char __user *buf, |
1153 | size_t count, loff_t *ppos) | |
1154 | { | |
496ad9aa | 1155 | struct task_struct *task = get_proc_task(file_inode(file)); |
a63d83f4 | 1156 | char buffer[PROC_NUMBUF]; |
a9c58b90 | 1157 | short oom_score_adj = OOM_SCORE_ADJ_MIN; |
a63d83f4 DR |
1158 | size_t len; |
1159 | ||
1160 | if (!task) | |
1161 | return -ESRCH; | |
f913da59 | 1162 | oom_score_adj = task->signal->oom_score_adj; |
a63d83f4 | 1163 | put_task_struct(task); |
a9c58b90 | 1164 | len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj); |
a63d83f4 DR |
1165 | return simple_read_from_buffer(buf, count, ppos, buffer, len); |
1166 | } | |
1167 | ||
1168 | static ssize_t oom_score_adj_write(struct file *file, const char __user *buf, | |
1169 | size_t count, loff_t *ppos) | |
1170 | { | |
a63d83f4 | 1171 | char buffer[PROC_NUMBUF]; |
0a8cb8e3 | 1172 | int oom_score_adj; |
a63d83f4 DR |
1173 | int err; |
1174 | ||
1175 | memset(buffer, 0, sizeof(buffer)); | |
1176 | if (count > sizeof(buffer) - 1) | |
1177 | count = sizeof(buffer) - 1; | |
723548bf DR |
1178 | if (copy_from_user(buffer, buf, count)) { |
1179 | err = -EFAULT; | |
1180 | goto out; | |
1181 | } | |
a63d83f4 | 1182 | |
0a8cb8e3 | 1183 | err = kstrtoint(strstrip(buffer), 0, &oom_score_adj); |
a63d83f4 | 1184 | if (err) |
723548bf | 1185 | goto out; |
a63d83f4 | 1186 | if (oom_score_adj < OOM_SCORE_ADJ_MIN || |
723548bf DR |
1187 | oom_score_adj > OOM_SCORE_ADJ_MAX) { |
1188 | err = -EINVAL; | |
1189 | goto out; | |
1190 | } | |
a63d83f4 | 1191 | |
1d5f0acb | 1192 | err = __set_oom_adj(file, oom_score_adj, false); |
723548bf DR |
1193 | out: |
1194 | return err < 0 ? err : count; | |
a63d83f4 DR |
1195 | } |
1196 | ||
1197 | static const struct file_operations proc_oom_score_adj_operations = { | |
1198 | .read = oom_score_adj_read, | |
1199 | .write = oom_score_adj_write, | |
6038f373 | 1200 | .llseek = default_llseek, |
a63d83f4 DR |
1201 | }; |
1202 | ||
1da177e4 | 1203 | #ifdef CONFIG_AUDITSYSCALL |
b4eb4f7f | 1204 | #define TMPBUFLEN 11 |
1da177e4 LT |
1205 | static ssize_t proc_loginuid_read(struct file * file, char __user * buf, |
1206 | size_t count, loff_t *ppos) | |
1207 | { | |
496ad9aa | 1208 | struct inode * inode = file_inode(file); |
99f89551 | 1209 | struct task_struct *task = get_proc_task(inode); |
1da177e4 LT |
1210 | ssize_t length; |
1211 | char tmpbuf[TMPBUFLEN]; | |
1212 | ||
99f89551 EB |
1213 | if (!task) |
1214 | return -ESRCH; | |
1da177e4 | 1215 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", |
e1760bd5 EB |
1216 | from_kuid(file->f_cred->user_ns, |
1217 | audit_get_loginuid(task))); | |
99f89551 | 1218 | put_task_struct(task); |
1da177e4 LT |
1219 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); |
1220 | } | |
1221 | ||
1222 | static ssize_t proc_loginuid_write(struct file * file, const char __user * buf, | |
1223 | size_t count, loff_t *ppos) | |
1224 | { | |
496ad9aa | 1225 | struct inode * inode = file_inode(file); |
1da177e4 | 1226 | uid_t loginuid; |
e1760bd5 | 1227 | kuid_t kloginuid; |
774636e1 | 1228 | int rv; |
1da177e4 | 1229 | |
7dc52157 PM |
1230 | rcu_read_lock(); |
1231 | if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) { | |
1232 | rcu_read_unlock(); | |
1da177e4 | 1233 | return -EPERM; |
7dc52157 PM |
1234 | } |
1235 | rcu_read_unlock(); | |
1da177e4 | 1236 | |
1da177e4 LT |
1237 | if (*ppos != 0) { |
1238 | /* No partial writes. */ | |
1239 | return -EINVAL; | |
1240 | } | |
1da177e4 | 1241 | |
774636e1 AD |
1242 | rv = kstrtou32_from_user(buf, count, 10, &loginuid); |
1243 | if (rv < 0) | |
1244 | return rv; | |
81407c84 EP |
1245 | |
1246 | /* is userspace tring to explicitly UNSET the loginuid? */ | |
1247 | if (loginuid == AUDIT_UID_UNSET) { | |
1248 | kloginuid = INVALID_UID; | |
1249 | } else { | |
1250 | kloginuid = make_kuid(file->f_cred->user_ns, loginuid); | |
774636e1 AD |
1251 | if (!uid_valid(kloginuid)) |
1252 | return -EINVAL; | |
e1760bd5 EB |
1253 | } |
1254 | ||
774636e1 AD |
1255 | rv = audit_set_loginuid(kloginuid); |
1256 | if (rv < 0) | |
1257 | return rv; | |
1258 | return count; | |
1da177e4 LT |
1259 | } |
1260 | ||
00977a59 | 1261 | static const struct file_operations proc_loginuid_operations = { |
1da177e4 LT |
1262 | .read = proc_loginuid_read, |
1263 | .write = proc_loginuid_write, | |
87df8424 | 1264 | .llseek = generic_file_llseek, |
1da177e4 | 1265 | }; |
1e0bd755 EP |
1266 | |
1267 | static ssize_t proc_sessionid_read(struct file * file, char __user * buf, | |
1268 | size_t count, loff_t *ppos) | |
1269 | { | |
496ad9aa | 1270 | struct inode * inode = file_inode(file); |
1e0bd755 EP |
1271 | struct task_struct *task = get_proc_task(inode); |
1272 | ssize_t length; | |
1273 | char tmpbuf[TMPBUFLEN]; | |
1274 | ||
1275 | if (!task) | |
1276 | return -ESRCH; | |
1277 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", | |
1278 | audit_get_sessionid(task)); | |
1279 | put_task_struct(task); | |
1280 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); | |
1281 | } | |
1282 | ||
1283 | static const struct file_operations proc_sessionid_operations = { | |
1284 | .read = proc_sessionid_read, | |
87df8424 | 1285 | .llseek = generic_file_llseek, |
1e0bd755 | 1286 | }; |
1da177e4 LT |
1287 | #endif |
1288 | ||
f4f154fd AM |
1289 | #ifdef CONFIG_FAULT_INJECTION |
1290 | static ssize_t proc_fault_inject_read(struct file * file, char __user * buf, | |
1291 | size_t count, loff_t *ppos) | |
1292 | { | |
496ad9aa | 1293 | struct task_struct *task = get_proc_task(file_inode(file)); |
f4f154fd AM |
1294 | char buffer[PROC_NUMBUF]; |
1295 | size_t len; | |
1296 | int make_it_fail; | |
f4f154fd AM |
1297 | |
1298 | if (!task) | |
1299 | return -ESRCH; | |
1300 | make_it_fail = task->make_it_fail; | |
1301 | put_task_struct(task); | |
1302 | ||
1303 | len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail); | |
0c28f287 AM |
1304 | |
1305 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
f4f154fd AM |
1306 | } |
1307 | ||
1308 | static ssize_t proc_fault_inject_write(struct file * file, | |
1309 | const char __user * buf, size_t count, loff_t *ppos) | |
1310 | { | |
1311 | struct task_struct *task; | |
774636e1 | 1312 | char buffer[PROC_NUMBUF]; |
f4f154fd | 1313 | int make_it_fail; |
774636e1 | 1314 | int rv; |
f4f154fd AM |
1315 | |
1316 | if (!capable(CAP_SYS_RESOURCE)) | |
1317 | return -EPERM; | |
1318 | memset(buffer, 0, sizeof(buffer)); | |
1319 | if (count > sizeof(buffer) - 1) | |
1320 | count = sizeof(buffer) - 1; | |
1321 | if (copy_from_user(buffer, buf, count)) | |
1322 | return -EFAULT; | |
774636e1 AD |
1323 | rv = kstrtoint(strstrip(buffer), 0, &make_it_fail); |
1324 | if (rv < 0) | |
1325 | return rv; | |
16caed31 DJ |
1326 | if (make_it_fail < 0 || make_it_fail > 1) |
1327 | return -EINVAL; | |
1328 | ||
496ad9aa | 1329 | task = get_proc_task(file_inode(file)); |
f4f154fd AM |
1330 | if (!task) |
1331 | return -ESRCH; | |
1332 | task->make_it_fail = make_it_fail; | |
1333 | put_task_struct(task); | |
cba8aafe VL |
1334 | |
1335 | return count; | |
f4f154fd AM |
1336 | } |
1337 | ||
00977a59 | 1338 | static const struct file_operations proc_fault_inject_operations = { |
f4f154fd AM |
1339 | .read = proc_fault_inject_read, |
1340 | .write = proc_fault_inject_write, | |
87df8424 | 1341 | .llseek = generic_file_llseek, |
f4f154fd | 1342 | }; |
e41d5818 DV |
1343 | |
1344 | static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf, | |
1345 | size_t count, loff_t *ppos) | |
1346 | { | |
1347 | struct task_struct *task; | |
9049f2f6 AM |
1348 | int err; |
1349 | unsigned int n; | |
e41d5818 | 1350 | |
1203c8e6 AM |
1351 | err = kstrtouint_from_user(buf, count, 0, &n); |
1352 | if (err) | |
1353 | return err; | |
1354 | ||
e41d5818 DV |
1355 | task = get_proc_task(file_inode(file)); |
1356 | if (!task) | |
1357 | return -ESRCH; | |
9f7118b2 | 1358 | task->fail_nth = n; |
e41d5818 | 1359 | put_task_struct(task); |
1203c8e6 | 1360 | |
e41d5818 DV |
1361 | return count; |
1362 | } | |
1363 | ||
1364 | static ssize_t proc_fail_nth_read(struct file *file, char __user *buf, | |
1365 | size_t count, loff_t *ppos) | |
1366 | { | |
1367 | struct task_struct *task; | |
bfc74093 AM |
1368 | char numbuf[PROC_NUMBUF]; |
1369 | ssize_t len; | |
e41d5818 DV |
1370 | |
1371 | task = get_proc_task(file_inode(file)); | |
1372 | if (!task) | |
1373 | return -ESRCH; | |
9f7118b2 | 1374 | len = snprintf(numbuf, sizeof(numbuf), "%u\n", task->fail_nth); |
bfc74093 | 1375 | len = simple_read_from_buffer(buf, count, ppos, numbuf, len); |
1203c8e6 | 1376 | put_task_struct(task); |
bfc74093 AM |
1377 | |
1378 | return len; | |
e41d5818 DV |
1379 | } |
1380 | ||
1381 | static const struct file_operations proc_fail_nth_operations = { | |
1382 | .read = proc_fail_nth_read, | |
1383 | .write = proc_fail_nth_write, | |
1384 | }; | |
f4f154fd AM |
1385 | #endif |
1386 | ||
9745512c | 1387 | |
43ae34cb IM |
1388 | #ifdef CONFIG_SCHED_DEBUG |
1389 | /* | |
1390 | * Print out various scheduling related per-task fields: | |
1391 | */ | |
1392 | static int sched_show(struct seq_file *m, void *v) | |
1393 | { | |
1394 | struct inode *inode = m->private; | |
76f668be | 1395 | struct pid_namespace *ns = proc_pid_ns(inode); |
43ae34cb IM |
1396 | struct task_struct *p; |
1397 | ||
43ae34cb IM |
1398 | p = get_proc_task(inode); |
1399 | if (!p) | |
1400 | return -ESRCH; | |
74dc3384 | 1401 | proc_sched_show_task(p, ns, m); |
43ae34cb IM |
1402 | |
1403 | put_task_struct(p); | |
1404 | ||
1405 | return 0; | |
1406 | } | |
1407 | ||
1408 | static ssize_t | |
1409 | sched_write(struct file *file, const char __user *buf, | |
1410 | size_t count, loff_t *offset) | |
1411 | { | |
496ad9aa | 1412 | struct inode *inode = file_inode(file); |
43ae34cb IM |
1413 | struct task_struct *p; |
1414 | ||
43ae34cb IM |
1415 | p = get_proc_task(inode); |
1416 | if (!p) | |
1417 | return -ESRCH; | |
1418 | proc_sched_set_task(p); | |
1419 | ||
1420 | put_task_struct(p); | |
1421 | ||
1422 | return count; | |
1423 | } | |
1424 | ||
1425 | static int sched_open(struct inode *inode, struct file *filp) | |
1426 | { | |
c6a34058 | 1427 | return single_open(filp, sched_show, inode); |
43ae34cb IM |
1428 | } |
1429 | ||
1430 | static const struct file_operations proc_pid_sched_operations = { | |
1431 | .open = sched_open, | |
1432 | .read = seq_read, | |
1433 | .write = sched_write, | |
1434 | .llseek = seq_lseek, | |
5ea473a1 | 1435 | .release = single_release, |
43ae34cb IM |
1436 | }; |
1437 | ||
1438 | #endif | |
1439 | ||
5091faa4 MG |
1440 | #ifdef CONFIG_SCHED_AUTOGROUP |
1441 | /* | |
1442 | * Print out autogroup related information: | |
1443 | */ | |
1444 | static int sched_autogroup_show(struct seq_file *m, void *v) | |
1445 | { | |
1446 | struct inode *inode = m->private; | |
1447 | struct task_struct *p; | |
1448 | ||
1449 | p = get_proc_task(inode); | |
1450 | if (!p) | |
1451 | return -ESRCH; | |
1452 | proc_sched_autogroup_show_task(p, m); | |
1453 | ||
1454 | put_task_struct(p); | |
1455 | ||
1456 | return 0; | |
1457 | } | |
1458 | ||
1459 | static ssize_t | |
1460 | sched_autogroup_write(struct file *file, const char __user *buf, | |
1461 | size_t count, loff_t *offset) | |
1462 | { | |
496ad9aa | 1463 | struct inode *inode = file_inode(file); |
5091faa4 MG |
1464 | struct task_struct *p; |
1465 | char buffer[PROC_NUMBUF]; | |
0a8cb8e3 | 1466 | int nice; |
5091faa4 MG |
1467 | int err; |
1468 | ||
1469 | memset(buffer, 0, sizeof(buffer)); | |
1470 | if (count > sizeof(buffer) - 1) | |
1471 | count = sizeof(buffer) - 1; | |
1472 | if (copy_from_user(buffer, buf, count)) | |
1473 | return -EFAULT; | |
1474 | ||
0a8cb8e3 AD |
1475 | err = kstrtoint(strstrip(buffer), 0, &nice); |
1476 | if (err < 0) | |
1477 | return err; | |
5091faa4 MG |
1478 | |
1479 | p = get_proc_task(inode); | |
1480 | if (!p) | |
1481 | return -ESRCH; | |
1482 | ||
2e5b5b3a | 1483 | err = proc_sched_autogroup_set_nice(p, nice); |
5091faa4 MG |
1484 | if (err) |
1485 | count = err; | |
1486 | ||
1487 | put_task_struct(p); | |
1488 | ||
1489 | return count; | |
1490 | } | |
1491 | ||
1492 | static int sched_autogroup_open(struct inode *inode, struct file *filp) | |
1493 | { | |
1494 | int ret; | |
1495 | ||
1496 | ret = single_open(filp, sched_autogroup_show, NULL); | |
1497 | if (!ret) { | |
1498 | struct seq_file *m = filp->private_data; | |
1499 | ||
1500 | m->private = inode; | |
1501 | } | |
1502 | return ret; | |
1503 | } | |
1504 | ||
1505 | static const struct file_operations proc_pid_sched_autogroup_operations = { | |
1506 | .open = sched_autogroup_open, | |
1507 | .read = seq_read, | |
1508 | .write = sched_autogroup_write, | |
1509 | .llseek = seq_lseek, | |
1510 | .release = single_release, | |
1511 | }; | |
1512 | ||
1513 | #endif /* CONFIG_SCHED_AUTOGROUP */ | |
1514 | ||
4614a696 JS |
1515 | static ssize_t comm_write(struct file *file, const char __user *buf, |
1516 | size_t count, loff_t *offset) | |
1517 | { | |
496ad9aa | 1518 | struct inode *inode = file_inode(file); |
4614a696 JS |
1519 | struct task_struct *p; |
1520 | char buffer[TASK_COMM_LEN]; | |
830e0fc9 | 1521 | const size_t maxlen = sizeof(buffer) - 1; |
4614a696 JS |
1522 | |
1523 | memset(buffer, 0, sizeof(buffer)); | |
830e0fc9 | 1524 | if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count)) |
4614a696 JS |
1525 | return -EFAULT; |
1526 | ||
1527 | p = get_proc_task(inode); | |
1528 | if (!p) | |
1529 | return -ESRCH; | |
1530 | ||
1531 | if (same_thread_group(current, p)) | |
1532 | set_task_comm(p, buffer); | |
1533 | else | |
1534 | count = -EINVAL; | |
1535 | ||
1536 | put_task_struct(p); | |
1537 | ||
1538 | return count; | |
1539 | } | |
1540 | ||
1541 | static int comm_show(struct seq_file *m, void *v) | |
1542 | { | |
1543 | struct inode *inode = m->private; | |
1544 | struct task_struct *p; | |
1545 | ||
1546 | p = get_proc_task(inode); | |
1547 | if (!p) | |
1548 | return -ESRCH; | |
1549 | ||
88b72b31 TH |
1550 | proc_task_name(m, p, false); |
1551 | seq_putc(m, '\n'); | |
4614a696 JS |
1552 | |
1553 | put_task_struct(p); | |
1554 | ||
1555 | return 0; | |
1556 | } | |
1557 | ||
1558 | static int comm_open(struct inode *inode, struct file *filp) | |
1559 | { | |
c6a34058 | 1560 | return single_open(filp, comm_show, inode); |
4614a696 JS |
1561 | } |
1562 | ||
1563 | static const struct file_operations proc_pid_set_comm_operations = { | |
1564 | .open = comm_open, | |
1565 | .read = seq_read, | |
1566 | .write = comm_write, | |
1567 | .llseek = seq_lseek, | |
1568 | .release = single_release, | |
1569 | }; | |
1570 | ||
7773fbc5 | 1571 | static int proc_exe_link(struct dentry *dentry, struct path *exe_path) |
925d1c40 MH |
1572 | { |
1573 | struct task_struct *task; | |
925d1c40 MH |
1574 | struct file *exe_file; |
1575 | ||
2b0143b5 | 1576 | task = get_proc_task(d_inode(dentry)); |
925d1c40 MH |
1577 | if (!task) |
1578 | return -ENOENT; | |
cd81a917 | 1579 | exe_file = get_task_exe_file(task); |
925d1c40 | 1580 | put_task_struct(task); |
925d1c40 MH |
1581 | if (exe_file) { |
1582 | *exe_path = exe_file->f_path; | |
1583 | path_get(&exe_file->f_path); | |
1584 | fput(exe_file); | |
1585 | return 0; | |
1586 | } else | |
1587 | return -ENOENT; | |
1588 | } | |
1589 | ||
6b255391 | 1590 | static const char *proc_pid_get_link(struct dentry *dentry, |
fceef393 AV |
1591 | struct inode *inode, |
1592 | struct delayed_call *done) | |
1da177e4 | 1593 | { |
408ef013 | 1594 | struct path path; |
1da177e4 LT |
1595 | int error = -EACCES; |
1596 | ||
6b255391 AV |
1597 | if (!dentry) |
1598 | return ERR_PTR(-ECHILD); | |
1599 | ||
778c1144 EB |
1600 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1601 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1602 | goto out; |
1da177e4 | 1603 | |
408ef013 CH |
1604 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1605 | if (error) | |
1606 | goto out; | |
1607 | ||
6e77137b | 1608 | nd_jump_link(&path); |
408ef013 | 1609 | return NULL; |
1da177e4 | 1610 | out: |
008b150a | 1611 | return ERR_PTR(error); |
1da177e4 LT |
1612 | } |
1613 | ||
3dcd25f3 | 1614 | static int do_proc_readlink(struct path *path, char __user *buffer, int buflen) |
1da177e4 | 1615 | { |
0ee931c4 | 1616 | char *tmp = (char *)__get_free_page(GFP_KERNEL); |
3dcd25f3 | 1617 | char *pathname; |
1da177e4 LT |
1618 | int len; |
1619 | ||
1620 | if (!tmp) | |
1621 | return -ENOMEM; | |
0c28f287 | 1622 | |
7b2a69ba | 1623 | pathname = d_path(path, tmp, PAGE_SIZE); |
3dcd25f3 JB |
1624 | len = PTR_ERR(pathname); |
1625 | if (IS_ERR(pathname)) | |
1da177e4 | 1626 | goto out; |
3dcd25f3 | 1627 | len = tmp + PAGE_SIZE - 1 - pathname; |
1da177e4 LT |
1628 | |
1629 | if (len > buflen) | |
1630 | len = buflen; | |
3dcd25f3 | 1631 | if (copy_to_user(buffer, pathname, len)) |
1da177e4 LT |
1632 | len = -EFAULT; |
1633 | out: | |
1634 | free_page((unsigned long)tmp); | |
1635 | return len; | |
1636 | } | |
1637 | ||
1638 | static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen) | |
1639 | { | |
1640 | int error = -EACCES; | |
2b0143b5 | 1641 | struct inode *inode = d_inode(dentry); |
3dcd25f3 | 1642 | struct path path; |
1da177e4 | 1643 | |
778c1144 EB |
1644 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1645 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1646 | goto out; |
1da177e4 | 1647 | |
7773fbc5 | 1648 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1da177e4 LT |
1649 | if (error) |
1650 | goto out; | |
1651 | ||
3dcd25f3 JB |
1652 | error = do_proc_readlink(&path, buffer, buflen); |
1653 | path_put(&path); | |
1da177e4 | 1654 | out: |
1da177e4 LT |
1655 | return error; |
1656 | } | |
1657 | ||
faf60af1 | 1658 | const struct inode_operations proc_pid_link_inode_operations = { |
1da177e4 | 1659 | .readlink = proc_pid_readlink, |
6b255391 | 1660 | .get_link = proc_pid_get_link, |
6d76fa58 | 1661 | .setattr = proc_setattr, |
1da177e4 LT |
1662 | }; |
1663 | ||
28a6d671 EB |
1664 | |
1665 | /* building an inode */ | |
1666 | ||
c6eb50d2 | 1667 | void task_dump_owner(struct task_struct *task, umode_t mode, |
68eb94f1 EB |
1668 | kuid_t *ruid, kgid_t *rgid) |
1669 | { | |
1670 | /* Depending on the state of dumpable compute who should own a | |
1671 | * proc file for a task. | |
1672 | */ | |
1673 | const struct cred *cred; | |
1674 | kuid_t uid; | |
1675 | kgid_t gid; | |
1676 | ||
2e0ad552 AD |
1677 | if (unlikely(task->flags & PF_KTHREAD)) { |
1678 | *ruid = GLOBAL_ROOT_UID; | |
1679 | *rgid = GLOBAL_ROOT_GID; | |
1680 | return; | |
1681 | } | |
1682 | ||
68eb94f1 EB |
1683 | /* Default to the tasks effective ownership */ |
1684 | rcu_read_lock(); | |
1685 | cred = __task_cred(task); | |
1686 | uid = cred->euid; | |
1687 | gid = cred->egid; | |
1688 | rcu_read_unlock(); | |
1689 | ||
1690 | /* | |
1691 | * Before the /proc/pid/status file was created the only way to read | |
1692 | * the effective uid of a /process was to stat /proc/pid. Reading | |
1693 | * /proc/pid/status is slow enough that procps and other packages | |
1694 | * kept stating /proc/pid. To keep the rules in /proc simple I have | |
1695 | * made this apply to all per process world readable and executable | |
1696 | * directories. | |
1697 | */ | |
1698 | if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) { | |
1699 | struct mm_struct *mm; | |
1700 | task_lock(task); | |
1701 | mm = task->mm; | |
1702 | /* Make non-dumpable tasks owned by some root */ | |
1703 | if (mm) { | |
1704 | if (get_dumpable(mm) != SUID_DUMP_USER) { | |
1705 | struct user_namespace *user_ns = mm->user_ns; | |
1706 | ||
1707 | uid = make_kuid(user_ns, 0); | |
1708 | if (!uid_valid(uid)) | |
1709 | uid = GLOBAL_ROOT_UID; | |
1710 | ||
1711 | gid = make_kgid(user_ns, 0); | |
1712 | if (!gid_valid(gid)) | |
1713 | gid = GLOBAL_ROOT_GID; | |
1714 | } | |
1715 | } else { | |
1716 | uid = GLOBAL_ROOT_UID; | |
1717 | gid = GLOBAL_ROOT_GID; | |
1718 | } | |
1719 | task_unlock(task); | |
1720 | } | |
1721 | *ruid = uid; | |
1722 | *rgid = gid; | |
1723 | } | |
1724 | ||
db978da8 AG |
1725 | struct inode *proc_pid_make_inode(struct super_block * sb, |
1726 | struct task_struct *task, umode_t mode) | |
28a6d671 EB |
1727 | { |
1728 | struct inode * inode; | |
1729 | struct proc_inode *ei; | |
1da177e4 | 1730 | |
28a6d671 | 1731 | /* We need a new inode */ |
1da177e4 | 1732 | |
28a6d671 EB |
1733 | inode = new_inode(sb); |
1734 | if (!inode) | |
1735 | goto out; | |
1736 | ||
1737 | /* Common stuff */ | |
1738 | ei = PROC_I(inode); | |
db978da8 | 1739 | inode->i_mode = mode; |
85fe4025 | 1740 | inode->i_ino = get_next_ino(); |
078cd827 | 1741 | inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode); |
28a6d671 EB |
1742 | inode->i_op = &proc_def_inode_operations; |
1743 | ||
1744 | /* | |
1745 | * grab the reference to task. | |
1746 | */ | |
1a657f78 | 1747 | ei->pid = get_task_pid(task, PIDTYPE_PID); |
28a6d671 EB |
1748 | if (!ei->pid) |
1749 | goto out_unlock; | |
1750 | ||
68eb94f1 | 1751 | task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid); |
28a6d671 EB |
1752 | security_task_to_inode(task, inode); |
1753 | ||
1da177e4 | 1754 | out: |
28a6d671 EB |
1755 | return inode; |
1756 | ||
1757 | out_unlock: | |
1758 | iput(inode); | |
1759 | return NULL; | |
1da177e4 LT |
1760 | } |
1761 | ||
a528d35e DH |
1762 | int pid_getattr(const struct path *path, struct kstat *stat, |
1763 | u32 request_mask, unsigned int query_flags) | |
1da177e4 | 1764 | { |
a528d35e | 1765 | struct inode *inode = d_inode(path->dentry); |
76f668be | 1766 | struct pid_namespace *pid = proc_pid_ns(inode); |
28a6d671 | 1767 | struct task_struct *task; |
c69e8d9c | 1768 | |
28a6d671 | 1769 | generic_fillattr(inode, stat); |
1da177e4 | 1770 | |
28a6d671 | 1771 | rcu_read_lock(); |
dcb0f222 EB |
1772 | stat->uid = GLOBAL_ROOT_UID; |
1773 | stat->gid = GLOBAL_ROOT_GID; | |
28a6d671 EB |
1774 | task = pid_task(proc_pid(inode), PIDTYPE_PID); |
1775 | if (task) { | |
796f571b | 1776 | if (!has_pid_permissions(pid, task, HIDEPID_INVISIBLE)) { |
0499680a VK |
1777 | rcu_read_unlock(); |
1778 | /* | |
1779 | * This doesn't prevent learning whether PID exists, | |
1780 | * it only makes getattr() consistent with readdir(). | |
1781 | */ | |
1782 | return -ENOENT; | |
1783 | } | |
68eb94f1 | 1784 | task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid); |
1da177e4 | 1785 | } |
28a6d671 | 1786 | rcu_read_unlock(); |
d6e71144 | 1787 | return 0; |
1da177e4 LT |
1788 | } |
1789 | ||
1da177e4 LT |
1790 | /* dentry stuff */ |
1791 | ||
1792 | /* | |
1bbc5513 AV |
1793 | * Set <pid>/... inode ownership (can change due to setuid(), etc.) |
1794 | */ | |
1795 | void pid_update_inode(struct task_struct *task, struct inode *inode) | |
1796 | { | |
1797 | task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid); | |
1798 | ||
1799 | inode->i_mode &= ~(S_ISUID | S_ISGID); | |
1800 | security_task_to_inode(task, inode); | |
1801 | } | |
1802 | ||
1803 | /* | |
1da177e4 LT |
1804 | * Rewrite the inode's ownerships here because the owning task may have |
1805 | * performed a setuid(), etc. | |
99f89551 | 1806 | * |
1da177e4 | 1807 | */ |
1bbc5513 | 1808 | static int pid_revalidate(struct dentry *dentry, unsigned int flags) |
1da177e4 | 1809 | { |
34286d66 NP |
1810 | struct inode *inode; |
1811 | struct task_struct *task; | |
c69e8d9c | 1812 | |
0b728e19 | 1813 | if (flags & LOOKUP_RCU) |
34286d66 NP |
1814 | return -ECHILD; |
1815 | ||
2b0143b5 | 1816 | inode = d_inode(dentry); |
34286d66 NP |
1817 | task = get_proc_task(inode); |
1818 | ||
99f89551 | 1819 | if (task) { |
1bbc5513 | 1820 | pid_update_inode(task, inode); |
99f89551 | 1821 | put_task_struct(task); |
1da177e4 LT |
1822 | return 1; |
1823 | } | |
1da177e4 LT |
1824 | return 0; |
1825 | } | |
1826 | ||
d855a4b7 ON |
1827 | static inline bool proc_inode_is_dead(struct inode *inode) |
1828 | { | |
1829 | return !proc_pid(inode)->tasks[PIDTYPE_PID].first; | |
1830 | } | |
1831 | ||
1dd704b6 DH |
1832 | int pid_delete_dentry(const struct dentry *dentry) |
1833 | { | |
1834 | /* Is the task we represent dead? | |
1835 | * If so, then don't put the dentry on the lru list, | |
1836 | * kill it immediately. | |
1837 | */ | |
2b0143b5 | 1838 | return proc_inode_is_dead(d_inode(dentry)); |
1dd704b6 DH |
1839 | } |
1840 | ||
6b4e306a | 1841 | const struct dentry_operations pid_dentry_operations = |
28a6d671 EB |
1842 | { |
1843 | .d_revalidate = pid_revalidate, | |
1844 | .d_delete = pid_delete_dentry, | |
1845 | }; | |
1846 | ||
1847 | /* Lookups */ | |
1848 | ||
1c0d04c9 EB |
1849 | /* |
1850 | * Fill a directory entry. | |
1851 | * | |
1852 | * If possible create the dcache entry and derive our inode number and | |
1853 | * file type from dcache entry. | |
1854 | * | |
1855 | * Since all of the proc inode numbers are dynamically generated, the inode | |
1856 | * numbers do not exist until the inode is cache. This means creating the | |
1857 | * the dcache entry in readdir is necessary to keep the inode numbers | |
1858 | * reported by readdir in sync with the inode numbers reported | |
1859 | * by stat. | |
1860 | */ | |
f0c3b509 | 1861 | bool proc_fill_cache(struct file *file, struct dir_context *ctx, |
6b4e306a | 1862 | const char *name, int len, |
c5141e6d | 1863 | instantiate_t instantiate, struct task_struct *task, const void *ptr) |
61a28784 | 1864 | { |
f0c3b509 | 1865 | struct dentry *child, *dir = file->f_path.dentry; |
1df98b8b | 1866 | struct qstr qname = QSTR_INIT(name, len); |
61a28784 | 1867 | struct inode *inode; |
0168b9e3 AV |
1868 | unsigned type = DT_UNKNOWN; |
1869 | ino_t ino = 1; | |
61a28784 | 1870 | |
1df98b8b | 1871 | child = d_hash_and_lookup(dir, &qname); |
61a28784 | 1872 | if (!child) { |
3781764b AV |
1873 | DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq); |
1874 | child = d_alloc_parallel(dir, &qname, &wq); | |
1875 | if (IS_ERR(child)) | |
1df98b8b | 1876 | goto end_instantiate; |
3781764b | 1877 | if (d_in_lookup(child)) { |
0168b9e3 AV |
1878 | struct dentry *res; |
1879 | res = instantiate(child, task, ptr); | |
3781764b | 1880 | d_lookup_done(child); |
0168b9e3 | 1881 | if (IS_ERR(res)) |
3781764b | 1882 | goto end_instantiate; |
0168b9e3 AV |
1883 | if (unlikely(res)) { |
1884 | dput(child); | |
1885 | child = res; | |
3781764b | 1886 | } |
61a28784 EB |
1887 | } |
1888 | } | |
2b0143b5 | 1889 | inode = d_inode(child); |
147ce699 AV |
1890 | ino = inode->i_ino; |
1891 | type = inode->i_mode >> 12; | |
0168b9e3 | 1892 | end_instantiate: |
61a28784 | 1893 | dput(child); |
f0c3b509 | 1894 | return dir_emit(ctx, name, len, ino, type); |
61a28784 EB |
1895 | } |
1896 | ||
640708a2 PE |
1897 | /* |
1898 | * dname_to_vma_addr - maps a dentry name into two unsigned longs | |
1899 | * which represent vma start and end addresses. | |
1900 | */ | |
1901 | static int dname_to_vma_addr(struct dentry *dentry, | |
1902 | unsigned long *start, unsigned long *end) | |
1903 | { | |
ac7f1061 AD |
1904 | const char *str = dentry->d_name.name; |
1905 | unsigned long long sval, eval; | |
1906 | unsigned int len; | |
1907 | ||
35318db5 AD |
1908 | if (str[0] == '0' && str[1] != '-') |
1909 | return -EINVAL; | |
ac7f1061 AD |
1910 | len = _parse_integer(str, 16, &sval); |
1911 | if (len & KSTRTOX_OVERFLOW) | |
1912 | return -EINVAL; | |
1913 | if (sval != (unsigned long)sval) | |
1914 | return -EINVAL; | |
1915 | str += len; | |
1916 | ||
1917 | if (*str != '-') | |
640708a2 | 1918 | return -EINVAL; |
ac7f1061 AD |
1919 | str++; |
1920 | ||
35318db5 AD |
1921 | if (str[0] == '0' && str[1]) |
1922 | return -EINVAL; | |
ac7f1061 AD |
1923 | len = _parse_integer(str, 16, &eval); |
1924 | if (len & KSTRTOX_OVERFLOW) | |
1925 | return -EINVAL; | |
1926 | if (eval != (unsigned long)eval) | |
1927 | return -EINVAL; | |
1928 | str += len; | |
1929 | ||
1930 | if (*str != '\0') | |
1931 | return -EINVAL; | |
1932 | ||
1933 | *start = sval; | |
1934 | *end = eval; | |
640708a2 PE |
1935 | |
1936 | return 0; | |
1937 | } | |
1938 | ||
0b728e19 | 1939 | static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags) |
640708a2 PE |
1940 | { |
1941 | unsigned long vm_start, vm_end; | |
1942 | bool exact_vma_exists = false; | |
1943 | struct mm_struct *mm = NULL; | |
1944 | struct task_struct *task; | |
640708a2 PE |
1945 | struct inode *inode; |
1946 | int status = 0; | |
1947 | ||
0b728e19 | 1948 | if (flags & LOOKUP_RCU) |
640708a2 PE |
1949 | return -ECHILD; |
1950 | ||
2b0143b5 | 1951 | inode = d_inode(dentry); |
640708a2 PE |
1952 | task = get_proc_task(inode); |
1953 | if (!task) | |
1954 | goto out_notask; | |
1955 | ||
caaee623 | 1956 | mm = mm_access(task, PTRACE_MODE_READ_FSCREDS); |
2344bec7 | 1957 | if (IS_ERR_OR_NULL(mm)) |
640708a2 PE |
1958 | goto out; |
1959 | ||
1960 | if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) { | |
1961 | down_read(&mm->mmap_sem); | |
1962 | exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end); | |
1963 | up_read(&mm->mmap_sem); | |
1964 | } | |
1965 | ||
1966 | mmput(mm); | |
1967 | ||
1968 | if (exact_vma_exists) { | |
68eb94f1 EB |
1969 | task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid); |
1970 | ||
640708a2 PE |
1971 | security_task_to_inode(task, inode); |
1972 | status = 1; | |
1973 | } | |
1974 | ||
1975 | out: | |
1976 | put_task_struct(task); | |
1977 | ||
1978 | out_notask: | |
640708a2 PE |
1979 | return status; |
1980 | } | |
1981 | ||
1982 | static const struct dentry_operations tid_map_files_dentry_operations = { | |
1983 | .d_revalidate = map_files_d_revalidate, | |
1984 | .d_delete = pid_delete_dentry, | |
1985 | }; | |
1986 | ||
6b255391 | 1987 | static int map_files_get_link(struct dentry *dentry, struct path *path) |
640708a2 PE |
1988 | { |
1989 | unsigned long vm_start, vm_end; | |
1990 | struct vm_area_struct *vma; | |
1991 | struct task_struct *task; | |
1992 | struct mm_struct *mm; | |
1993 | int rc; | |
1994 | ||
1995 | rc = -ENOENT; | |
2b0143b5 | 1996 | task = get_proc_task(d_inode(dentry)); |
640708a2 PE |
1997 | if (!task) |
1998 | goto out; | |
1999 | ||
2000 | mm = get_task_mm(task); | |
2001 | put_task_struct(task); | |
2002 | if (!mm) | |
2003 | goto out; | |
2004 | ||
2005 | rc = dname_to_vma_addr(dentry, &vm_start, &vm_end); | |
2006 | if (rc) | |
2007 | goto out_mmput; | |
2008 | ||
70335abb | 2009 | rc = -ENOENT; |
640708a2 PE |
2010 | down_read(&mm->mmap_sem); |
2011 | vma = find_exact_vma(mm, vm_start, vm_end); | |
2012 | if (vma && vma->vm_file) { | |
2013 | *path = vma->vm_file->f_path; | |
2014 | path_get(path); | |
2015 | rc = 0; | |
2016 | } | |
2017 | up_read(&mm->mmap_sem); | |
2018 | ||
2019 | out_mmput: | |
2020 | mmput(mm); | |
2021 | out: | |
2022 | return rc; | |
2023 | } | |
2024 | ||
2025 | struct map_files_info { | |
20d28cde AD |
2026 | unsigned long start; |
2027 | unsigned long end; | |
7b540d06 | 2028 | fmode_t mode; |
640708a2 PE |
2029 | }; |
2030 | ||
bdb4d100 CO |
2031 | /* |
2032 | * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the | |
2033 | * symlinks may be used to bypass permissions on ancestor directories in the | |
2034 | * path to the file in question. | |
2035 | */ | |
2036 | static const char * | |
6b255391 | 2037 | proc_map_files_get_link(struct dentry *dentry, |
fceef393 AV |
2038 | struct inode *inode, |
2039 | struct delayed_call *done) | |
bdb4d100 CO |
2040 | { |
2041 | if (!capable(CAP_SYS_ADMIN)) | |
2042 | return ERR_PTR(-EPERM); | |
2043 | ||
fceef393 | 2044 | return proc_pid_get_link(dentry, inode, done); |
bdb4d100 CO |
2045 | } |
2046 | ||
2047 | /* | |
6b255391 | 2048 | * Identical to proc_pid_link_inode_operations except for get_link() |
bdb4d100 CO |
2049 | */ |
2050 | static const struct inode_operations proc_map_files_link_inode_operations = { | |
2051 | .readlink = proc_pid_readlink, | |
6b255391 | 2052 | .get_link = proc_map_files_get_link, |
bdb4d100 CO |
2053 | .setattr = proc_setattr, |
2054 | }; | |
2055 | ||
0168b9e3 AV |
2056 | static struct dentry * |
2057 | proc_map_files_instantiate(struct dentry *dentry, | |
640708a2 PE |
2058 | struct task_struct *task, const void *ptr) |
2059 | { | |
7b540d06 | 2060 | fmode_t mode = (fmode_t)(unsigned long)ptr; |
640708a2 PE |
2061 | struct proc_inode *ei; |
2062 | struct inode *inode; | |
2063 | ||
0168b9e3 | 2064 | inode = proc_pid_make_inode(dentry->d_sb, task, S_IFLNK | |
db978da8 AG |
2065 | ((mode & FMODE_READ ) ? S_IRUSR : 0) | |
2066 | ((mode & FMODE_WRITE) ? S_IWUSR : 0)); | |
640708a2 | 2067 | if (!inode) |
0168b9e3 | 2068 | return ERR_PTR(-ENOENT); |
640708a2 PE |
2069 | |
2070 | ei = PROC_I(inode); | |
6b255391 | 2071 | ei->op.proc_get_link = map_files_get_link; |
640708a2 | 2072 | |
bdb4d100 | 2073 | inode->i_op = &proc_map_files_link_inode_operations; |
640708a2 | 2074 | inode->i_size = 64; |
640708a2 PE |
2075 | |
2076 | d_set_d_op(dentry, &tid_map_files_dentry_operations); | |
0168b9e3 | 2077 | return d_splice_alias(inode, dentry); |
640708a2 PE |
2078 | } |
2079 | ||
2080 | static struct dentry *proc_map_files_lookup(struct inode *dir, | |
00cd8dd3 | 2081 | struct dentry *dentry, unsigned int flags) |
640708a2 PE |
2082 | { |
2083 | unsigned long vm_start, vm_end; | |
2084 | struct vm_area_struct *vma; | |
2085 | struct task_struct *task; | |
0168b9e3 | 2086 | struct dentry *result; |
640708a2 PE |
2087 | struct mm_struct *mm; |
2088 | ||
0168b9e3 | 2089 | result = ERR_PTR(-ENOENT); |
640708a2 PE |
2090 | task = get_proc_task(dir); |
2091 | if (!task) | |
2092 | goto out; | |
2093 | ||
0168b9e3 | 2094 | result = ERR_PTR(-EACCES); |
caaee623 | 2095 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
640708a2 PE |
2096 | goto out_put_task; |
2097 | ||
0168b9e3 | 2098 | result = ERR_PTR(-ENOENT); |
640708a2 | 2099 | if (dname_to_vma_addr(dentry, &vm_start, &vm_end)) |
eb94cd96 | 2100 | goto out_put_task; |
640708a2 PE |
2101 | |
2102 | mm = get_task_mm(task); | |
2103 | if (!mm) | |
eb94cd96 | 2104 | goto out_put_task; |
640708a2 PE |
2105 | |
2106 | down_read(&mm->mmap_sem); | |
2107 | vma = find_exact_vma(mm, vm_start, vm_end); | |
2108 | if (!vma) | |
2109 | goto out_no_vma; | |
2110 | ||
05f56484 | 2111 | if (vma->vm_file) |
0168b9e3 | 2112 | result = proc_map_files_instantiate(dentry, task, |
05f56484 | 2113 | (void *)(unsigned long)vma->vm_file->f_mode); |
640708a2 PE |
2114 | |
2115 | out_no_vma: | |
2116 | up_read(&mm->mmap_sem); | |
2117 | mmput(mm); | |
640708a2 PE |
2118 | out_put_task: |
2119 | put_task_struct(task); | |
2120 | out: | |
0168b9e3 | 2121 | return result; |
640708a2 PE |
2122 | } |
2123 | ||
2124 | static const struct inode_operations proc_map_files_inode_operations = { | |
2125 | .lookup = proc_map_files_lookup, | |
2126 | .permission = proc_fd_permission, | |
2127 | .setattr = proc_setattr, | |
2128 | }; | |
2129 | ||
2130 | static int | |
f0c3b509 | 2131 | proc_map_files_readdir(struct file *file, struct dir_context *ctx) |
640708a2 | 2132 | { |
640708a2 PE |
2133 | struct vm_area_struct *vma; |
2134 | struct task_struct *task; | |
2135 | struct mm_struct *mm; | |
f0c3b509 AV |
2136 | unsigned long nr_files, pos, i; |
2137 | struct flex_array *fa = NULL; | |
2138 | struct map_files_info info; | |
2139 | struct map_files_info *p; | |
640708a2 PE |
2140 | int ret; |
2141 | ||
640708a2 | 2142 | ret = -ENOENT; |
f0c3b509 | 2143 | task = get_proc_task(file_inode(file)); |
640708a2 PE |
2144 | if (!task) |
2145 | goto out; | |
2146 | ||
2147 | ret = -EACCES; | |
caaee623 | 2148 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
640708a2 PE |
2149 | goto out_put_task; |
2150 | ||
2151 | ret = 0; | |
f0c3b509 AV |
2152 | if (!dir_emit_dots(file, ctx)) |
2153 | goto out_put_task; | |
640708a2 | 2154 | |
f0c3b509 AV |
2155 | mm = get_task_mm(task); |
2156 | if (!mm) | |
2157 | goto out_put_task; | |
2158 | down_read(&mm->mmap_sem); | |
640708a2 | 2159 | |
f0c3b509 | 2160 | nr_files = 0; |
640708a2 | 2161 | |
f0c3b509 AV |
2162 | /* |
2163 | * We need two passes here: | |
2164 | * | |
2165 | * 1) Collect vmas of mapped files with mmap_sem taken | |
2166 | * 2) Release mmap_sem and instantiate entries | |
2167 | * | |
2168 | * otherwise we get lockdep complained, since filldir() | |
2169 | * routine might require mmap_sem taken in might_fault(). | |
2170 | */ | |
640708a2 | 2171 | |
f0c3b509 AV |
2172 | for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) { |
2173 | if (vma->vm_file && ++pos > ctx->pos) | |
2174 | nr_files++; | |
2175 | } | |
2176 | ||
2177 | if (nr_files) { | |
2178 | fa = flex_array_alloc(sizeof(info), nr_files, | |
2179 | GFP_KERNEL); | |
2180 | if (!fa || flex_array_prealloc(fa, 0, nr_files, | |
2181 | GFP_KERNEL)) { | |
2182 | ret = -ENOMEM; | |
2183 | if (fa) | |
2184 | flex_array_free(fa); | |
2185 | up_read(&mm->mmap_sem); | |
2186 | mmput(mm); | |
2187 | goto out_put_task; | |
640708a2 | 2188 | } |
f0c3b509 AV |
2189 | for (i = 0, vma = mm->mmap, pos = 2; vma; |
2190 | vma = vma->vm_next) { | |
2191 | if (!vma->vm_file) | |
2192 | continue; | |
2193 | if (++pos <= ctx->pos) | |
2194 | continue; | |
2195 | ||
20d28cde AD |
2196 | info.start = vma->vm_start; |
2197 | info.end = vma->vm_end; | |
f0c3b509 | 2198 | info.mode = vma->vm_file->f_mode; |
f0c3b509 AV |
2199 | if (flex_array_put(fa, i++, &info, GFP_KERNEL)) |
2200 | BUG(); | |
640708a2 | 2201 | } |
640708a2 | 2202 | } |
f0c3b509 | 2203 | up_read(&mm->mmap_sem); |
fe079a5e | 2204 | mmput(mm); |
f0c3b509 AV |
2205 | |
2206 | for (i = 0; i < nr_files; i++) { | |
20d28cde AD |
2207 | char buf[4 * sizeof(long) + 2]; /* max: %lx-%lx\0 */ |
2208 | unsigned int len; | |
2209 | ||
f0c3b509 | 2210 | p = flex_array_get(fa, i); |
20d28cde | 2211 | len = snprintf(buf, sizeof(buf), "%lx-%lx", p->start, p->end); |
f0c3b509 | 2212 | if (!proc_fill_cache(file, ctx, |
20d28cde | 2213 | buf, len, |
f0c3b509 AV |
2214 | proc_map_files_instantiate, |
2215 | task, | |
2216 | (void *)(unsigned long)p->mode)) | |
2217 | break; | |
2218 | ctx->pos++; | |
640708a2 | 2219 | } |
f0c3b509 AV |
2220 | if (fa) |
2221 | flex_array_free(fa); | |
640708a2 | 2222 | |
640708a2 PE |
2223 | out_put_task: |
2224 | put_task_struct(task); | |
2225 | out: | |
2226 | return ret; | |
2227 | } | |
2228 | ||
2229 | static const struct file_operations proc_map_files_operations = { | |
2230 | .read = generic_read_dir, | |
f50752ea AV |
2231 | .iterate_shared = proc_map_files_readdir, |
2232 | .llseek = generic_file_llseek, | |
640708a2 PE |
2233 | }; |
2234 | ||
b18b6a9c | 2235 | #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS) |
48f6a7a5 PE |
2236 | struct timers_private { |
2237 | struct pid *pid; | |
2238 | struct task_struct *task; | |
2239 | struct sighand_struct *sighand; | |
57b8015e | 2240 | struct pid_namespace *ns; |
48f6a7a5 PE |
2241 | unsigned long flags; |
2242 | }; | |
2243 | ||
2244 | static void *timers_start(struct seq_file *m, loff_t *pos) | |
2245 | { | |
2246 | struct timers_private *tp = m->private; | |
2247 | ||
2248 | tp->task = get_pid_task(tp->pid, PIDTYPE_PID); | |
2249 | if (!tp->task) | |
2250 | return ERR_PTR(-ESRCH); | |
2251 | ||
2252 | tp->sighand = lock_task_sighand(tp->task, &tp->flags); | |
2253 | if (!tp->sighand) | |
2254 | return ERR_PTR(-ESRCH); | |
2255 | ||
2256 | return seq_list_start(&tp->task->signal->posix_timers, *pos); | |
2257 | } | |
2258 | ||
2259 | static void *timers_next(struct seq_file *m, void *v, loff_t *pos) | |
2260 | { | |
2261 | struct timers_private *tp = m->private; | |
2262 | return seq_list_next(v, &tp->task->signal->posix_timers, pos); | |
2263 | } | |
2264 | ||
2265 | static void timers_stop(struct seq_file *m, void *v) | |
2266 | { | |
2267 | struct timers_private *tp = m->private; | |
2268 | ||
2269 | if (tp->sighand) { | |
2270 | unlock_task_sighand(tp->task, &tp->flags); | |
2271 | tp->sighand = NULL; | |
2272 | } | |
2273 | ||
2274 | if (tp->task) { | |
2275 | put_task_struct(tp->task); | |
2276 | tp->task = NULL; | |
2277 | } | |
2278 | } | |
2279 | ||
2280 | static int show_timer(struct seq_file *m, void *v) | |
2281 | { | |
2282 | struct k_itimer *timer; | |
57b8015e PE |
2283 | struct timers_private *tp = m->private; |
2284 | int notify; | |
cedbccab | 2285 | static const char * const nstr[] = { |
57b8015e PE |
2286 | [SIGEV_SIGNAL] = "signal", |
2287 | [SIGEV_NONE] = "none", | |
2288 | [SIGEV_THREAD] = "thread", | |
2289 | }; | |
48f6a7a5 PE |
2290 | |
2291 | timer = list_entry((struct list_head *)v, struct k_itimer, list); | |
57b8015e PE |
2292 | notify = timer->it_sigev_notify; |
2293 | ||
48f6a7a5 | 2294 | seq_printf(m, "ID: %d\n", timer->it_id); |
ba3edf1f | 2295 | seq_printf(m, "signal: %d/%px\n", |
25ce3191 JP |
2296 | timer->sigq->info.si_signo, |
2297 | timer->sigq->info.si_value.sival_ptr); | |
57b8015e | 2298 | seq_printf(m, "notify: %s/%s.%d\n", |
25ce3191 JP |
2299 | nstr[notify & ~SIGEV_THREAD_ID], |
2300 | (notify & SIGEV_THREAD_ID) ? "tid" : "pid", | |
2301 | pid_nr_ns(timer->it_pid, tp->ns)); | |
15ef0298 | 2302 | seq_printf(m, "ClockID: %d\n", timer->it_clock); |
48f6a7a5 PE |
2303 | |
2304 | return 0; | |
2305 | } | |
2306 | ||
2307 | static const struct seq_operations proc_timers_seq_ops = { | |
2308 | .start = timers_start, | |
2309 | .next = timers_next, | |
2310 | .stop = timers_stop, | |
2311 | .show = show_timer, | |
2312 | }; | |
2313 | ||
2314 | static int proc_timers_open(struct inode *inode, struct file *file) | |
2315 | { | |
2316 | struct timers_private *tp; | |
2317 | ||
2318 | tp = __seq_open_private(file, &proc_timers_seq_ops, | |
2319 | sizeof(struct timers_private)); | |
2320 | if (!tp) | |
2321 | return -ENOMEM; | |
2322 | ||
2323 | tp->pid = proc_pid(inode); | |
76f668be | 2324 | tp->ns = proc_pid_ns(inode); |
48f6a7a5 PE |
2325 | return 0; |
2326 | } | |
2327 | ||
2328 | static const struct file_operations proc_timers_operations = { | |
2329 | .open = proc_timers_open, | |
2330 | .read = seq_read, | |
2331 | .llseek = seq_lseek, | |
2332 | .release = seq_release_private, | |
2333 | }; | |
b5946bea | 2334 | #endif |
640708a2 | 2335 | |
5de23d43 JS |
2336 | static ssize_t timerslack_ns_write(struct file *file, const char __user *buf, |
2337 | size_t count, loff_t *offset) | |
2338 | { | |
2339 | struct inode *inode = file_inode(file); | |
2340 | struct task_struct *p; | |
2341 | u64 slack_ns; | |
2342 | int err; | |
2343 | ||
2344 | err = kstrtoull_from_user(buf, count, 10, &slack_ns); | |
2345 | if (err < 0) | |
2346 | return err; | |
2347 | ||
2348 | p = get_proc_task(inode); | |
2349 | if (!p) | |
2350 | return -ESRCH; | |
2351 | ||
4b2bd5fe JS |
2352 | if (p != current) { |
2353 | if (!capable(CAP_SYS_NICE)) { | |
2354 | count = -EPERM; | |
2355 | goto out; | |
2356 | } | |
5de23d43 | 2357 | |
4b2bd5fe JS |
2358 | err = security_task_setscheduler(p); |
2359 | if (err) { | |
2360 | count = err; | |
2361 | goto out; | |
2362 | } | |
904763e1 JS |
2363 | } |
2364 | ||
7abbaf94 JS |
2365 | task_lock(p); |
2366 | if (slack_ns == 0) | |
2367 | p->timer_slack_ns = p->default_timer_slack_ns; | |
2368 | else | |
2369 | p->timer_slack_ns = slack_ns; | |
2370 | task_unlock(p); | |
2371 | ||
2372 | out: | |
5de23d43 JS |
2373 | put_task_struct(p); |
2374 | ||
2375 | return count; | |
2376 | } | |
2377 | ||
2378 | static int timerslack_ns_show(struct seq_file *m, void *v) | |
2379 | { | |
2380 | struct inode *inode = m->private; | |
2381 | struct task_struct *p; | |
7abbaf94 | 2382 | int err = 0; |
5de23d43 JS |
2383 | |
2384 | p = get_proc_task(inode); | |
2385 | if (!p) | |
2386 | return -ESRCH; | |
2387 | ||
4b2bd5fe | 2388 | if (p != current) { |
5de23d43 | 2389 | |
4b2bd5fe JS |
2390 | if (!capable(CAP_SYS_NICE)) { |
2391 | err = -EPERM; | |
2392 | goto out; | |
2393 | } | |
2394 | err = security_task_getscheduler(p); | |
2395 | if (err) | |
2396 | goto out; | |
2397 | } | |
904763e1 | 2398 | |
7abbaf94 JS |
2399 | task_lock(p); |
2400 | seq_printf(m, "%llu\n", p->timer_slack_ns); | |
2401 | task_unlock(p); | |
2402 | ||
2403 | out: | |
5de23d43 JS |
2404 | put_task_struct(p); |
2405 | ||
2406 | return err; | |
2407 | } | |
2408 | ||
2409 | static int timerslack_ns_open(struct inode *inode, struct file *filp) | |
2410 | { | |
2411 | return single_open(filp, timerslack_ns_show, inode); | |
2412 | } | |
2413 | ||
2414 | static const struct file_operations proc_pid_set_timerslack_ns_operations = { | |
2415 | .open = timerslack_ns_open, | |
2416 | .read = seq_read, | |
2417 | .write = timerslack_ns_write, | |
2418 | .llseek = seq_lseek, | |
2419 | .release = single_release, | |
2420 | }; | |
2421 | ||
0168b9e3 AV |
2422 | static struct dentry *proc_pident_instantiate(struct dentry *dentry, |
2423 | struct task_struct *task, const void *ptr) | |
444ceed8 | 2424 | { |
c5141e6d | 2425 | const struct pid_entry *p = ptr; |
444ceed8 EB |
2426 | struct inode *inode; |
2427 | struct proc_inode *ei; | |
444ceed8 | 2428 | |
0168b9e3 | 2429 | inode = proc_pid_make_inode(dentry->d_sb, task, p->mode); |
444ceed8 | 2430 | if (!inode) |
0168b9e3 | 2431 | return ERR_PTR(-ENOENT); |
444ceed8 EB |
2432 | |
2433 | ei = PROC_I(inode); | |
444ceed8 | 2434 | if (S_ISDIR(inode->i_mode)) |
bfe86848 | 2435 | set_nlink(inode, 2); /* Use getattr to fix if necessary */ |
444ceed8 EB |
2436 | if (p->iop) |
2437 | inode->i_op = p->iop; | |
2438 | if (p->fop) | |
2439 | inode->i_fop = p->fop; | |
2440 | ei->op = p->op; | |
1bbc5513 | 2441 | pid_update_inode(task, inode); |
fb045adb | 2442 | d_set_d_op(dentry, &pid_dentry_operations); |
0168b9e3 | 2443 | return d_splice_alias(inode, dentry); |
444ceed8 EB |
2444 | } |
2445 | ||
1da177e4 LT |
2446 | static struct dentry *proc_pident_lookup(struct inode *dir, |
2447 | struct dentry *dentry, | |
c5141e6d | 2448 | const struct pid_entry *ents, |
7bcd6b0e | 2449 | unsigned int nents) |
1da177e4 | 2450 | { |
99f89551 | 2451 | struct task_struct *task = get_proc_task(dir); |
c5141e6d | 2452 | const struct pid_entry *p, *last; |
0168b9e3 | 2453 | struct dentry *res = ERR_PTR(-ENOENT); |
1da177e4 | 2454 | |
99f89551 EB |
2455 | if (!task) |
2456 | goto out_no_task; | |
1da177e4 | 2457 | |
20cdc894 EB |
2458 | /* |
2459 | * Yes, it does not scale. And it should not. Don't add | |
2460 | * new entries into /proc/<tgid>/ without very good reasons. | |
2461 | */ | |
bac5f5d5 AD |
2462 | last = &ents[nents]; |
2463 | for (p = ents; p < last; p++) { | |
1da177e4 LT |
2464 | if (p->len != dentry->d_name.len) |
2465 | continue; | |
2466 | if (!memcmp(dentry->d_name.name, p->name, p->len)) | |
2467 | break; | |
2468 | } | |
bac5f5d5 | 2469 | if (p >= last) |
1da177e4 LT |
2470 | goto out; |
2471 | ||
0168b9e3 | 2472 | res = proc_pident_instantiate(dentry, task, p); |
1da177e4 | 2473 | out: |
99f89551 EB |
2474 | put_task_struct(task); |
2475 | out_no_task: | |
0168b9e3 | 2476 | return res; |
1da177e4 LT |
2477 | } |
2478 | ||
f0c3b509 | 2479 | static int proc_pident_readdir(struct file *file, struct dir_context *ctx, |
c5141e6d | 2480 | const struct pid_entry *ents, unsigned int nents) |
28a6d671 | 2481 | { |
f0c3b509 AV |
2482 | struct task_struct *task = get_proc_task(file_inode(file)); |
2483 | const struct pid_entry *p; | |
28a6d671 | 2484 | |
28a6d671 | 2485 | if (!task) |
f0c3b509 | 2486 | return -ENOENT; |
28a6d671 | 2487 | |
f0c3b509 AV |
2488 | if (!dir_emit_dots(file, ctx)) |
2489 | goto out; | |
2490 | ||
2491 | if (ctx->pos >= nents + 2) | |
2492 | goto out; | |
28a6d671 | 2493 | |
bac5f5d5 | 2494 | for (p = ents + (ctx->pos - 2); p < ents + nents; p++) { |
f0c3b509 AV |
2495 | if (!proc_fill_cache(file, ctx, p->name, p->len, |
2496 | proc_pident_instantiate, task, p)) | |
2497 | break; | |
2498 | ctx->pos++; | |
2499 | } | |
28a6d671 | 2500 | out: |
61a28784 | 2501 | put_task_struct(task); |
f0c3b509 | 2502 | return 0; |
1da177e4 LT |
2503 | } |
2504 | ||
28a6d671 EB |
2505 | #ifdef CONFIG_SECURITY |
2506 | static ssize_t proc_pid_attr_read(struct file * file, char __user * buf, | |
2507 | size_t count, loff_t *ppos) | |
2508 | { | |
496ad9aa | 2509 | struct inode * inode = file_inode(file); |
04ff9708 | 2510 | char *p = NULL; |
28a6d671 EB |
2511 | ssize_t length; |
2512 | struct task_struct *task = get_proc_task(inode); | |
2513 | ||
28a6d671 | 2514 | if (!task) |
04ff9708 | 2515 | return -ESRCH; |
28a6d671 EB |
2516 | |
2517 | length = security_getprocattr(task, | |
2fddfeef | 2518 | (char*)file->f_path.dentry->d_name.name, |
04ff9708 | 2519 | &p); |
28a6d671 | 2520 | put_task_struct(task); |
04ff9708 AV |
2521 | if (length > 0) |
2522 | length = simple_read_from_buffer(buf, count, ppos, p, length); | |
2523 | kfree(p); | |
28a6d671 | 2524 | return length; |
1da177e4 LT |
2525 | } |
2526 | ||
28a6d671 EB |
2527 | static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf, |
2528 | size_t count, loff_t *ppos) | |
2529 | { | |
496ad9aa | 2530 | struct inode * inode = file_inode(file); |
bb646cdb | 2531 | void *page; |
28a6d671 EB |
2532 | ssize_t length; |
2533 | struct task_struct *task = get_proc_task(inode); | |
2534 | ||
2535 | length = -ESRCH; | |
2536 | if (!task) | |
2537 | goto out_no_task; | |
b21507e2 SS |
2538 | |
2539 | /* A task may only write its own attributes. */ | |
2540 | length = -EACCES; | |
2541 | if (current != task) | |
2542 | goto out; | |
2543 | ||
28a6d671 EB |
2544 | if (count > PAGE_SIZE) |
2545 | count = PAGE_SIZE; | |
2546 | ||
2547 | /* No partial writes. */ | |
2548 | length = -EINVAL; | |
2549 | if (*ppos != 0) | |
2550 | goto out; | |
2551 | ||
bb646cdb AV |
2552 | page = memdup_user(buf, count); |
2553 | if (IS_ERR(page)) { | |
2554 | length = PTR_ERR(page); | |
28a6d671 | 2555 | goto out; |
bb646cdb | 2556 | } |
28a6d671 | 2557 | |
107db7c7 | 2558 | /* Guard against adverse ptrace interaction */ |
b21507e2 | 2559 | length = mutex_lock_interruptible(¤t->signal->cred_guard_mutex); |
107db7c7 DH |
2560 | if (length < 0) |
2561 | goto out_free; | |
2562 | ||
b21507e2 | 2563 | length = security_setprocattr(file->f_path.dentry->d_name.name, |
bb646cdb | 2564 | page, count); |
b21507e2 | 2565 | mutex_unlock(¤t->signal->cred_guard_mutex); |
28a6d671 | 2566 | out_free: |
bb646cdb | 2567 | kfree(page); |
28a6d671 EB |
2568 | out: |
2569 | put_task_struct(task); | |
2570 | out_no_task: | |
2571 | return length; | |
2572 | } | |
2573 | ||
00977a59 | 2574 | static const struct file_operations proc_pid_attr_operations = { |
28a6d671 EB |
2575 | .read = proc_pid_attr_read, |
2576 | .write = proc_pid_attr_write, | |
87df8424 | 2577 | .llseek = generic_file_llseek, |
28a6d671 EB |
2578 | }; |
2579 | ||
c5141e6d | 2580 | static const struct pid_entry attr_dir_stuff[] = { |
631f9c18 AD |
2581 | REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations), |
2582 | REG("prev", S_IRUGO, proc_pid_attr_operations), | |
2583 | REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2584 | REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2585 | REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2586 | REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
28a6d671 EB |
2587 | }; |
2588 | ||
f0c3b509 | 2589 | static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx) |
28a6d671 | 2590 | { |
f0c3b509 AV |
2591 | return proc_pident_readdir(file, ctx, |
2592 | attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2593 | } |
2594 | ||
00977a59 | 2595 | static const struct file_operations proc_attr_dir_operations = { |
1da177e4 | 2596 | .read = generic_read_dir, |
f50752ea AV |
2597 | .iterate_shared = proc_attr_dir_readdir, |
2598 | .llseek = generic_file_llseek, | |
1da177e4 LT |
2599 | }; |
2600 | ||
72d9dcfc | 2601 | static struct dentry *proc_attr_dir_lookup(struct inode *dir, |
00cd8dd3 | 2602 | struct dentry *dentry, unsigned int flags) |
28a6d671 | 2603 | { |
7bcd6b0e EB |
2604 | return proc_pident_lookup(dir, dentry, |
2605 | attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2606 | } |
2607 | ||
c5ef1c42 | 2608 | static const struct inode_operations proc_attr_dir_inode_operations = { |
72d9dcfc | 2609 | .lookup = proc_attr_dir_lookup, |
99f89551 | 2610 | .getattr = pid_getattr, |
6d76fa58 | 2611 | .setattr = proc_setattr, |
1da177e4 LT |
2612 | }; |
2613 | ||
28a6d671 EB |
2614 | #endif |
2615 | ||
698ba7b5 | 2616 | #ifdef CONFIG_ELF_CORE |
3cb4a0bb KH |
2617 | static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf, |
2618 | size_t count, loff_t *ppos) | |
2619 | { | |
496ad9aa | 2620 | struct task_struct *task = get_proc_task(file_inode(file)); |
3cb4a0bb KH |
2621 | struct mm_struct *mm; |
2622 | char buffer[PROC_NUMBUF]; | |
2623 | size_t len; | |
2624 | int ret; | |
2625 | ||
2626 | if (!task) | |
2627 | return -ESRCH; | |
2628 | ||
2629 | ret = 0; | |
2630 | mm = get_task_mm(task); | |
2631 | if (mm) { | |
2632 | len = snprintf(buffer, sizeof(buffer), "%08lx\n", | |
2633 | ((mm->flags & MMF_DUMP_FILTER_MASK) >> | |
2634 | MMF_DUMP_FILTER_SHIFT)); | |
2635 | mmput(mm); | |
2636 | ret = simple_read_from_buffer(buf, count, ppos, buffer, len); | |
2637 | } | |
2638 | ||
2639 | put_task_struct(task); | |
2640 | ||
2641 | return ret; | |
2642 | } | |
2643 | ||
2644 | static ssize_t proc_coredump_filter_write(struct file *file, | |
2645 | const char __user *buf, | |
2646 | size_t count, | |
2647 | loff_t *ppos) | |
2648 | { | |
2649 | struct task_struct *task; | |
2650 | struct mm_struct *mm; | |
3cb4a0bb KH |
2651 | unsigned int val; |
2652 | int ret; | |
2653 | int i; | |
2654 | unsigned long mask; | |
2655 | ||
774636e1 AD |
2656 | ret = kstrtouint_from_user(buf, count, 0, &val); |
2657 | if (ret < 0) | |
2658 | return ret; | |
3cb4a0bb KH |
2659 | |
2660 | ret = -ESRCH; | |
496ad9aa | 2661 | task = get_proc_task(file_inode(file)); |
3cb4a0bb KH |
2662 | if (!task) |
2663 | goto out_no_task; | |
2664 | ||
3cb4a0bb KH |
2665 | mm = get_task_mm(task); |
2666 | if (!mm) | |
2667 | goto out_no_mm; | |
41a0c249 | 2668 | ret = 0; |
3cb4a0bb KH |
2669 | |
2670 | for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) { | |
2671 | if (val & mask) | |
2672 | set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2673 | else | |
2674 | clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2675 | } | |
2676 | ||
2677 | mmput(mm); | |
2678 | out_no_mm: | |
2679 | put_task_struct(task); | |
2680 | out_no_task: | |
774636e1 AD |
2681 | if (ret < 0) |
2682 | return ret; | |
2683 | return count; | |
3cb4a0bb KH |
2684 | } |
2685 | ||
2686 | static const struct file_operations proc_coredump_filter_operations = { | |
2687 | .read = proc_coredump_filter_read, | |
2688 | .write = proc_coredump_filter_write, | |
87df8424 | 2689 | .llseek = generic_file_llseek, |
3cb4a0bb KH |
2690 | }; |
2691 | #endif | |
2692 | ||
aba76fdb | 2693 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 2694 | static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole) |
297c5d92 | 2695 | { |
940389b8 | 2696 | struct task_io_accounting acct = task->ioac; |
5995477a | 2697 | unsigned long flags; |
293eb1e7 | 2698 | int result; |
5995477a | 2699 | |
293eb1e7 VK |
2700 | result = mutex_lock_killable(&task->signal->cred_guard_mutex); |
2701 | if (result) | |
2702 | return result; | |
2703 | ||
caaee623 | 2704 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) { |
293eb1e7 VK |
2705 | result = -EACCES; |
2706 | goto out_unlock; | |
2707 | } | |
1d1221f3 | 2708 | |
5995477a AR |
2709 | if (whole && lock_task_sighand(task, &flags)) { |
2710 | struct task_struct *t = task; | |
2711 | ||
2712 | task_io_accounting_add(&acct, &task->signal->ioac); | |
2713 | while_each_thread(task, t) | |
2714 | task_io_accounting_add(&acct, &t->ioac); | |
2715 | ||
2716 | unlock_task_sighand(task, &flags); | |
297c5d92 | 2717 | } |
25ce3191 JP |
2718 | seq_printf(m, |
2719 | "rchar: %llu\n" | |
2720 | "wchar: %llu\n" | |
2721 | "syscr: %llu\n" | |
2722 | "syscw: %llu\n" | |
2723 | "read_bytes: %llu\n" | |
2724 | "write_bytes: %llu\n" | |
2725 | "cancelled_write_bytes: %llu\n", | |
2726 | (unsigned long long)acct.rchar, | |
2727 | (unsigned long long)acct.wchar, | |
2728 | (unsigned long long)acct.syscr, | |
2729 | (unsigned long long)acct.syscw, | |
2730 | (unsigned long long)acct.read_bytes, | |
2731 | (unsigned long long)acct.write_bytes, | |
2732 | (unsigned long long)acct.cancelled_write_bytes); | |
2733 | result = 0; | |
2734 | ||
293eb1e7 VK |
2735 | out_unlock: |
2736 | mutex_unlock(&task->signal->cred_guard_mutex); | |
2737 | return result; | |
297c5d92 AR |
2738 | } |
2739 | ||
19aadc98 AD |
2740 | static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns, |
2741 | struct pid *pid, struct task_struct *task) | |
297c5d92 | 2742 | { |
19aadc98 | 2743 | return do_io_accounting(task, m, 0); |
aba76fdb | 2744 | } |
297c5d92 | 2745 | |
19aadc98 AD |
2746 | static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns, |
2747 | struct pid *pid, struct task_struct *task) | |
297c5d92 | 2748 | { |
19aadc98 | 2749 | return do_io_accounting(task, m, 1); |
297c5d92 AR |
2750 | } |
2751 | #endif /* CONFIG_TASK_IO_ACCOUNTING */ | |
aba76fdb | 2752 | |
22d917d8 EB |
2753 | #ifdef CONFIG_USER_NS |
2754 | static int proc_id_map_open(struct inode *inode, struct file *file, | |
ccf94f1b | 2755 | const struct seq_operations *seq_ops) |
22d917d8 EB |
2756 | { |
2757 | struct user_namespace *ns = NULL; | |
2758 | struct task_struct *task; | |
2759 | struct seq_file *seq; | |
2760 | int ret = -EINVAL; | |
2761 | ||
2762 | task = get_proc_task(inode); | |
2763 | if (task) { | |
2764 | rcu_read_lock(); | |
2765 | ns = get_user_ns(task_cred_xxx(task, user_ns)); | |
2766 | rcu_read_unlock(); | |
2767 | put_task_struct(task); | |
2768 | } | |
2769 | if (!ns) | |
2770 | goto err; | |
2771 | ||
2772 | ret = seq_open(file, seq_ops); | |
2773 | if (ret) | |
2774 | goto err_put_ns; | |
2775 | ||
2776 | seq = file->private_data; | |
2777 | seq->private = ns; | |
2778 | ||
2779 | return 0; | |
2780 | err_put_ns: | |
2781 | put_user_ns(ns); | |
2782 | err: | |
2783 | return ret; | |
2784 | } | |
2785 | ||
2786 | static int proc_id_map_release(struct inode *inode, struct file *file) | |
2787 | { | |
2788 | struct seq_file *seq = file->private_data; | |
2789 | struct user_namespace *ns = seq->private; | |
2790 | put_user_ns(ns); | |
2791 | return seq_release(inode, file); | |
2792 | } | |
2793 | ||
2794 | static int proc_uid_map_open(struct inode *inode, struct file *file) | |
2795 | { | |
2796 | return proc_id_map_open(inode, file, &proc_uid_seq_operations); | |
2797 | } | |
2798 | ||
2799 | static int proc_gid_map_open(struct inode *inode, struct file *file) | |
2800 | { | |
2801 | return proc_id_map_open(inode, file, &proc_gid_seq_operations); | |
2802 | } | |
2803 | ||
f76d207a EB |
2804 | static int proc_projid_map_open(struct inode *inode, struct file *file) |
2805 | { | |
2806 | return proc_id_map_open(inode, file, &proc_projid_seq_operations); | |
2807 | } | |
2808 | ||
22d917d8 EB |
2809 | static const struct file_operations proc_uid_map_operations = { |
2810 | .open = proc_uid_map_open, | |
2811 | .write = proc_uid_map_write, | |
2812 | .read = seq_read, | |
2813 | .llseek = seq_lseek, | |
2814 | .release = proc_id_map_release, | |
2815 | }; | |
2816 | ||
2817 | static const struct file_operations proc_gid_map_operations = { | |
2818 | .open = proc_gid_map_open, | |
2819 | .write = proc_gid_map_write, | |
2820 | .read = seq_read, | |
2821 | .llseek = seq_lseek, | |
2822 | .release = proc_id_map_release, | |
2823 | }; | |
f76d207a EB |
2824 | |
2825 | static const struct file_operations proc_projid_map_operations = { | |
2826 | .open = proc_projid_map_open, | |
2827 | .write = proc_projid_map_write, | |
2828 | .read = seq_read, | |
2829 | .llseek = seq_lseek, | |
2830 | .release = proc_id_map_release, | |
2831 | }; | |
9cc46516 EB |
2832 | |
2833 | static int proc_setgroups_open(struct inode *inode, struct file *file) | |
2834 | { | |
2835 | struct user_namespace *ns = NULL; | |
2836 | struct task_struct *task; | |
2837 | int ret; | |
2838 | ||
2839 | ret = -ESRCH; | |
2840 | task = get_proc_task(inode); | |
2841 | if (task) { | |
2842 | rcu_read_lock(); | |
2843 | ns = get_user_ns(task_cred_xxx(task, user_ns)); | |
2844 | rcu_read_unlock(); | |
2845 | put_task_struct(task); | |
2846 | } | |
2847 | if (!ns) | |
2848 | goto err; | |
2849 | ||
2850 | if (file->f_mode & FMODE_WRITE) { | |
2851 | ret = -EACCES; | |
2852 | if (!ns_capable(ns, CAP_SYS_ADMIN)) | |
2853 | goto err_put_ns; | |
2854 | } | |
2855 | ||
2856 | ret = single_open(file, &proc_setgroups_show, ns); | |
2857 | if (ret) | |
2858 | goto err_put_ns; | |
2859 | ||
2860 | return 0; | |
2861 | err_put_ns: | |
2862 | put_user_ns(ns); | |
2863 | err: | |
2864 | return ret; | |
2865 | } | |
2866 | ||
2867 | static int proc_setgroups_release(struct inode *inode, struct file *file) | |
2868 | { | |
2869 | struct seq_file *seq = file->private_data; | |
2870 | struct user_namespace *ns = seq->private; | |
2871 | int ret = single_release(inode, file); | |
2872 | put_user_ns(ns); | |
2873 | return ret; | |
2874 | } | |
2875 | ||
2876 | static const struct file_operations proc_setgroups_operations = { | |
2877 | .open = proc_setgroups_open, | |
2878 | .write = proc_setgroups_write, | |
2879 | .read = seq_read, | |
2880 | .llseek = seq_lseek, | |
2881 | .release = proc_setgroups_release, | |
2882 | }; | |
22d917d8 EB |
2883 | #endif /* CONFIG_USER_NS */ |
2884 | ||
47830723 KC |
2885 | static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns, |
2886 | struct pid *pid, struct task_struct *task) | |
2887 | { | |
a9712bc1 AV |
2888 | int err = lock_trace(task); |
2889 | if (!err) { | |
2890 | seq_printf(m, "%08x\n", task->personality); | |
2891 | unlock_trace(task); | |
2892 | } | |
2893 | return err; | |
47830723 KC |
2894 | } |
2895 | ||
7c23b330 JP |
2896 | #ifdef CONFIG_LIVEPATCH |
2897 | static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns, | |
2898 | struct pid *pid, struct task_struct *task) | |
2899 | { | |
2900 | seq_printf(m, "%d\n", task->patch_state); | |
2901 | return 0; | |
2902 | } | |
2903 | #endif /* CONFIG_LIVEPATCH */ | |
2904 | ||
28a6d671 EB |
2905 | /* |
2906 | * Thread groups | |
2907 | */ | |
00977a59 | 2908 | static const struct file_operations proc_task_operations; |
c5ef1c42 | 2909 | static const struct inode_operations proc_task_inode_operations; |
20cdc894 | 2910 | |
c5141e6d | 2911 | static const struct pid_entry tgid_base_stuff[] = { |
631f9c18 AD |
2912 | DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations), |
2913 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), | |
640708a2 | 2914 | DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations), |
631f9c18 | 2915 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 2916 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
b2211a36 | 2917 | #ifdef CONFIG_NET |
631f9c18 | 2918 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), |
b2211a36 | 2919 | #endif |
631f9c18 | 2920 | REG("environ", S_IRUSR, proc_environ_operations), |
c5317167 | 2921 | REG("auxv", S_IRUSR, proc_auxv_operations), |
631f9c18 | 2922 | ONE("status", S_IRUGO, proc_pid_status), |
35a35046 | 2923 | ONE("personality", S_IRUSR, proc_pid_personality), |
1c963eb1 | 2924 | ONE("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 2925 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 2926 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
5091faa4 MG |
2927 | #endif |
2928 | #ifdef CONFIG_SCHED_AUTOGROUP | |
2929 | REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations), | |
ebcb6734 | 2930 | #endif |
4614a696 | 2931 | REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations), |
ebcb6734 | 2932 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 | 2933 | ONE("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 2934 | #endif |
c2c0bb44 | 2935 | REG("cmdline", S_IRUGO, proc_pid_cmdline_ops), |
631f9c18 AD |
2936 | ONE("stat", S_IRUGO, proc_tgid_stat), |
2937 | ONE("statm", S_IRUGO, proc_pid_statm), | |
b7643757 | 2938 | REG("maps", S_IRUGO, proc_pid_maps_operations), |
28a6d671 | 2939 | #ifdef CONFIG_NUMA |
b7643757 | 2940 | REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations), |
28a6d671 | 2941 | #endif |
631f9c18 AD |
2942 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
2943 | LNK("cwd", proc_cwd_link), | |
2944 | LNK("root", proc_root_link), | |
2945 | LNK("exe", proc_exe_link), | |
2946 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
2947 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
2948 | REG("mountstats", S_IRUSR, proc_mountstats_operations), | |
1e883281 | 2949 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 2950 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
b7643757 | 2951 | REG("smaps", S_IRUGO, proc_pid_smaps_operations), |
493b0e9d | 2952 | REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations), |
32ed74a4 | 2953 | REG("pagemap", S_IRUSR, proc_pagemap_operations), |
28a6d671 EB |
2954 | #endif |
2955 | #ifdef CONFIG_SECURITY | |
631f9c18 | 2956 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
2957 | #endif |
2958 | #ifdef CONFIG_KALLSYMS | |
edfcd606 | 2959 | ONE("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 2960 | #endif |
2ec220e2 | 2961 | #ifdef CONFIG_STACKTRACE |
35a35046 | 2962 | ONE("stack", S_IRUSR, proc_pid_stack), |
28a6d671 | 2963 | #endif |
5968cece | 2964 | #ifdef CONFIG_SCHED_INFO |
f6e826ca | 2965 | ONE("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 2966 | #endif |
9745512c | 2967 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 2968 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 2969 | #endif |
8793d854 | 2970 | #ifdef CONFIG_PROC_PID_CPUSET |
52de4779 | 2971 | ONE("cpuset", S_IRUGO, proc_cpuset_show), |
a424316c PM |
2972 | #endif |
2973 | #ifdef CONFIG_CGROUPS | |
006f4ac4 | 2974 | ONE("cgroup", S_IRUGO, proc_cgroup_show), |
28a6d671 | 2975 | #endif |
6ba51e37 | 2976 | ONE("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 2977 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 2978 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
28a6d671 | 2979 | #ifdef CONFIG_AUDITSYSCALL |
631f9c18 AD |
2980 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
2981 | REG("sessionid", S_IRUGO, proc_sessionid_operations), | |
28a6d671 | 2982 | #endif |
f4f154fd | 2983 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 2984 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
168c42bc | 2985 | REG("fail-nth", 0644, proc_fail_nth_operations), |
f4f154fd | 2986 | #endif |
698ba7b5 | 2987 | #ifdef CONFIG_ELF_CORE |
631f9c18 | 2988 | REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations), |
3cb4a0bb | 2989 | #endif |
aba76fdb | 2990 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 2991 | ONE("io", S_IRUSR, proc_tgid_io_accounting), |
aba76fdb | 2992 | #endif |
22d917d8 EB |
2993 | #ifdef CONFIG_USER_NS |
2994 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
2995 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 2996 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
9cc46516 | 2997 | REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations), |
22d917d8 | 2998 | #endif |
b18b6a9c | 2999 | #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS) |
48f6a7a5 PE |
3000 | REG("timers", S_IRUGO, proc_timers_operations), |
3001 | #endif | |
5de23d43 | 3002 | REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations), |
7c23b330 JP |
3003 | #ifdef CONFIG_LIVEPATCH |
3004 | ONE("patch_state", S_IRUSR, proc_pid_patch_state), | |
3005 | #endif | |
28a6d671 | 3006 | }; |
1da177e4 | 3007 | |
f0c3b509 | 3008 | static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3009 | { |
f0c3b509 AV |
3010 | return proc_pident_readdir(file, ctx, |
3011 | tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
3012 | } |
3013 | ||
00977a59 | 3014 | static const struct file_operations proc_tgid_base_operations = { |
1da177e4 | 3015 | .read = generic_read_dir, |
f50752ea AV |
3016 | .iterate_shared = proc_tgid_base_readdir, |
3017 | .llseek = generic_file_llseek, | |
1da177e4 LT |
3018 | }; |
3019 | ||
00cd8dd3 AV |
3020 | static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
3021 | { | |
7bcd6b0e EB |
3022 | return proc_pident_lookup(dir, dentry, |
3023 | tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
3024 | } |
3025 | ||
c5ef1c42 | 3026 | static const struct inode_operations proc_tgid_base_inode_operations = { |
28a6d671 | 3027 | .lookup = proc_tgid_base_lookup, |
99f89551 | 3028 | .getattr = pid_getattr, |
6d76fa58 | 3029 | .setattr = proc_setattr, |
0499680a | 3030 | .permission = proc_pid_permission, |
1da177e4 | 3031 | }; |
1da177e4 | 3032 | |
60347f67 | 3033 | static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid) |
1da177e4 | 3034 | { |
48e6484d | 3035 | struct dentry *dentry, *leader, *dir; |
e3912ac3 | 3036 | char buf[10 + 1]; |
48e6484d EB |
3037 | struct qstr name; |
3038 | ||
3039 | name.name = buf; | |
e3912ac3 | 3040 | name.len = snprintf(buf, sizeof(buf), "%u", pid); |
4f522a24 | 3041 | /* no ->d_hash() rejects on procfs */ |
60347f67 | 3042 | dentry = d_hash_and_lookup(mnt->mnt_root, &name); |
48e6484d | 3043 | if (dentry) { |
bbd51924 | 3044 | d_invalidate(dentry); |
48e6484d EB |
3045 | dput(dentry); |
3046 | } | |
1da177e4 | 3047 | |
c35a7f18 ON |
3048 | if (pid == tgid) |
3049 | return; | |
3050 | ||
48e6484d | 3051 | name.name = buf; |
e3912ac3 | 3052 | name.len = snprintf(buf, sizeof(buf), "%u", tgid); |
60347f67 | 3053 | leader = d_hash_and_lookup(mnt->mnt_root, &name); |
48e6484d EB |
3054 | if (!leader) |
3055 | goto out; | |
1da177e4 | 3056 | |
48e6484d EB |
3057 | name.name = "task"; |
3058 | name.len = strlen(name.name); | |
3059 | dir = d_hash_and_lookup(leader, &name); | |
3060 | if (!dir) | |
3061 | goto out_put_leader; | |
3062 | ||
3063 | name.name = buf; | |
e3912ac3 | 3064 | name.len = snprintf(buf, sizeof(buf), "%u", pid); |
48e6484d EB |
3065 | dentry = d_hash_and_lookup(dir, &name); |
3066 | if (dentry) { | |
bbd51924 | 3067 | d_invalidate(dentry); |
48e6484d | 3068 | dput(dentry); |
1da177e4 | 3069 | } |
48e6484d EB |
3070 | |
3071 | dput(dir); | |
3072 | out_put_leader: | |
3073 | dput(leader); | |
3074 | out: | |
3075 | return; | |
1da177e4 LT |
3076 | } |
3077 | ||
0895e91d RD |
3078 | /** |
3079 | * proc_flush_task - Remove dcache entries for @task from the /proc dcache. | |
3080 | * @task: task that should be flushed. | |
3081 | * | |
3082 | * When flushing dentries from proc, one needs to flush them from global | |
60347f67 | 3083 | * proc (proc_mnt) and from all the namespaces' procs this task was seen |
0895e91d RD |
3084 | * in. This call is supposed to do all of this job. |
3085 | * | |
3086 | * Looks in the dcache for | |
3087 | * /proc/@pid | |
3088 | * /proc/@tgid/task/@pid | |
3089 | * if either directory is present flushes it and all of it'ts children | |
3090 | * from the dcache. | |
3091 | * | |
3092 | * It is safe and reasonable to cache /proc entries for a task until | |
3093 | * that task exits. After that they just clog up the dcache with | |
3094 | * useless entries, possibly causing useful dcache entries to be | |
3095 | * flushed instead. This routine is proved to flush those useless | |
3096 | * dcache entries at process exit time. | |
3097 | * | |
3098 | * NOTE: This routine is just an optimization so it does not guarantee | |
3099 | * that no dcache entries will exist at process exit time it | |
3100 | * just makes it very unlikely that any will persist. | |
60347f67 PE |
3101 | */ |
3102 | ||
3103 | void proc_flush_task(struct task_struct *task) | |
3104 | { | |
9fcc2d15 | 3105 | int i; |
9b4d1cbe | 3106 | struct pid *pid, *tgid; |
130f77ec PE |
3107 | struct upid *upid; |
3108 | ||
130f77ec | 3109 | pid = task_pid(task); |
9b4d1cbe | 3110 | tgid = task_tgid(task); |
130f77ec | 3111 | |
9fcc2d15 | 3112 | for (i = 0; i <= pid->level; i++) { |
130f77ec PE |
3113 | upid = &pid->numbers[i]; |
3114 | proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr, | |
9b4d1cbe | 3115 | tgid->numbers[i].nr); |
130f77ec | 3116 | } |
60347f67 PE |
3117 | } |
3118 | ||
0168b9e3 | 3119 | static struct dentry *proc_pid_instantiate(struct dentry * dentry, |
c52a47ac | 3120 | struct task_struct *task, const void *ptr) |
444ceed8 | 3121 | { |
444ceed8 EB |
3122 | struct inode *inode; |
3123 | ||
0168b9e3 | 3124 | inode = proc_pid_make_inode(dentry->d_sb, task, S_IFDIR | S_IRUGO | S_IXUGO); |
444ceed8 | 3125 | if (!inode) |
0168b9e3 | 3126 | return ERR_PTR(-ENOENT); |
444ceed8 | 3127 | |
444ceed8 EB |
3128 | inode->i_op = &proc_tgid_base_inode_operations; |
3129 | inode->i_fop = &proc_tgid_base_operations; | |
3130 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 | 3131 | |
1270dd8d | 3132 | set_nlink(inode, nlink_tgid); |
1bbc5513 | 3133 | pid_update_inode(task, inode); |
444ceed8 | 3134 | |
fb045adb | 3135 | d_set_d_op(dentry, &pid_dentry_operations); |
0168b9e3 | 3136 | return d_splice_alias(inode, dentry); |
444ceed8 EB |
3137 | } |
3138 | ||
00cd8dd3 | 3139 | struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags) |
1da177e4 LT |
3140 | { |
3141 | struct task_struct *task; | |
1da177e4 | 3142 | unsigned tgid; |
b488893a | 3143 | struct pid_namespace *ns; |
0168b9e3 | 3144 | struct dentry *result = ERR_PTR(-ENOENT); |
1da177e4 | 3145 | |
dbcdb504 | 3146 | tgid = name_to_int(&dentry->d_name); |
1da177e4 LT |
3147 | if (tgid == ~0U) |
3148 | goto out; | |
3149 | ||
b488893a | 3150 | ns = dentry->d_sb->s_fs_info; |
de758734 | 3151 | rcu_read_lock(); |
b488893a | 3152 | task = find_task_by_pid_ns(tgid, ns); |
1da177e4 LT |
3153 | if (task) |
3154 | get_task_struct(task); | |
de758734 | 3155 | rcu_read_unlock(); |
1da177e4 LT |
3156 | if (!task) |
3157 | goto out; | |
3158 | ||
0168b9e3 | 3159 | result = proc_pid_instantiate(dentry, task, NULL); |
1da177e4 | 3160 | put_task_struct(task); |
1da177e4 | 3161 | out: |
0168b9e3 | 3162 | return result; |
1da177e4 LT |
3163 | } |
3164 | ||
1da177e4 | 3165 | /* |
0804ef4b | 3166 | * Find the first task with tgid >= tgid |
0bc58a91 | 3167 | * |
1da177e4 | 3168 | */ |
19fd4bb2 EB |
3169 | struct tgid_iter { |
3170 | unsigned int tgid; | |
0804ef4b | 3171 | struct task_struct *task; |
19fd4bb2 EB |
3172 | }; |
3173 | static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter) | |
3174 | { | |
0804ef4b | 3175 | struct pid *pid; |
1da177e4 | 3176 | |
19fd4bb2 EB |
3177 | if (iter.task) |
3178 | put_task_struct(iter.task); | |
454cc105 | 3179 | rcu_read_lock(); |
0804ef4b | 3180 | retry: |
19fd4bb2 EB |
3181 | iter.task = NULL; |
3182 | pid = find_ge_pid(iter.tgid, ns); | |
0804ef4b | 3183 | if (pid) { |
19fd4bb2 EB |
3184 | iter.tgid = pid_nr_ns(pid, ns); |
3185 | iter.task = pid_task(pid, PIDTYPE_PID); | |
0804ef4b EB |
3186 | /* What we to know is if the pid we have find is the |
3187 | * pid of a thread_group_leader. Testing for task | |
3188 | * being a thread_group_leader is the obvious thing | |
3189 | * todo but there is a window when it fails, due to | |
3190 | * the pid transfer logic in de_thread. | |
3191 | * | |
3192 | * So we perform the straight forward test of seeing | |
3193 | * if the pid we have found is the pid of a thread | |
3194 | * group leader, and don't worry if the task we have | |
3195 | * found doesn't happen to be a thread group leader. | |
3196 | * As we don't care in the case of readdir. | |
3197 | */ | |
19fd4bb2 EB |
3198 | if (!iter.task || !has_group_leader_pid(iter.task)) { |
3199 | iter.tgid += 1; | |
0804ef4b | 3200 | goto retry; |
19fd4bb2 EB |
3201 | } |
3202 | get_task_struct(iter.task); | |
0bc58a91 | 3203 | } |
454cc105 | 3204 | rcu_read_unlock(); |
19fd4bb2 | 3205 | return iter; |
1da177e4 LT |
3206 | } |
3207 | ||
0097875b | 3208 | #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2) |
0804ef4b | 3209 | |
1da177e4 | 3210 | /* for the /proc/ directory itself, after non-process stuff has been done */ |
f0c3b509 | 3211 | int proc_pid_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3212 | { |
19fd4bb2 | 3213 | struct tgid_iter iter; |
76f668be | 3214 | struct pid_namespace *ns = proc_pid_ns(file_inode(file)); |
f0c3b509 | 3215 | loff_t pos = ctx->pos; |
1da177e4 | 3216 | |
021ada7d | 3217 | if (pos >= PID_MAX_LIMIT + TGID_OFFSET) |
f0c3b509 | 3218 | return 0; |
1da177e4 | 3219 | |
0097875b | 3220 | if (pos == TGID_OFFSET - 2) { |
2b0143b5 | 3221 | struct inode *inode = d_inode(ns->proc_self); |
db963164 | 3222 | if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK)) |
f0c3b509 | 3223 | return 0; |
0097875b EB |
3224 | ctx->pos = pos = pos + 1; |
3225 | } | |
3226 | if (pos == TGID_OFFSET - 1) { | |
2b0143b5 | 3227 | struct inode *inode = d_inode(ns->proc_thread_self); |
0097875b EB |
3228 | if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK)) |
3229 | return 0; | |
3230 | ctx->pos = pos = pos + 1; | |
021ada7d | 3231 | } |
0097875b | 3232 | iter.tgid = pos - TGID_OFFSET; |
19fd4bb2 | 3233 | iter.task = NULL; |
19fd4bb2 EB |
3234 | for (iter = next_tgid(ns, iter); |
3235 | iter.task; | |
3236 | iter.tgid += 1, iter = next_tgid(ns, iter)) { | |
e3912ac3 | 3237 | char name[10 + 1]; |
f0c3b509 | 3238 | int len; |
3ba4bcee ED |
3239 | |
3240 | cond_resched(); | |
796f571b | 3241 | if (!has_pid_permissions(ns, iter.task, HIDEPID_INVISIBLE)) |
f0c3b509 | 3242 | continue; |
0499680a | 3243 | |
e3912ac3 | 3244 | len = snprintf(name, sizeof(name), "%u", iter.tgid); |
f0c3b509 AV |
3245 | ctx->pos = iter.tgid + TGID_OFFSET; |
3246 | if (!proc_fill_cache(file, ctx, name, len, | |
3247 | proc_pid_instantiate, iter.task, NULL)) { | |
19fd4bb2 | 3248 | put_task_struct(iter.task); |
f0c3b509 | 3249 | return 0; |
1da177e4 | 3250 | } |
0bc58a91 | 3251 | } |
f0c3b509 | 3252 | ctx->pos = PID_MAX_LIMIT + TGID_OFFSET; |
0bc58a91 EB |
3253 | return 0; |
3254 | } | |
1da177e4 | 3255 | |
1b3044e3 JD |
3256 | /* |
3257 | * proc_tid_comm_permission is a special permission function exclusively | |
3258 | * used for the node /proc/<pid>/task/<tid>/comm. | |
3259 | * It bypasses generic permission checks in the case where a task of the same | |
3260 | * task group attempts to access the node. | |
3261 | * The rationale behind this is that glibc and bionic access this node for | |
3262 | * cross thread naming (pthread_set/getname_np(!self)). However, if | |
3263 | * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0, | |
3264 | * which locks out the cross thread naming implementation. | |
3265 | * This function makes sure that the node is always accessible for members of | |
3266 | * same thread group. | |
3267 | */ | |
3268 | static int proc_tid_comm_permission(struct inode *inode, int mask) | |
3269 | { | |
3270 | bool is_same_tgroup; | |
3271 | struct task_struct *task; | |
3272 | ||
3273 | task = get_proc_task(inode); | |
3274 | if (!task) | |
3275 | return -ESRCH; | |
3276 | is_same_tgroup = same_thread_group(current, task); | |
3277 | put_task_struct(task); | |
3278 | ||
3279 | if (likely(is_same_tgroup && !(mask & MAY_EXEC))) { | |
3280 | /* This file (/proc/<pid>/task/<tid>/comm) can always be | |
3281 | * read or written by the members of the corresponding | |
3282 | * thread group. | |
3283 | */ | |
3284 | return 0; | |
3285 | } | |
3286 | ||
3287 | return generic_permission(inode, mask); | |
3288 | } | |
3289 | ||
3290 | static const struct inode_operations proc_tid_comm_inode_operations = { | |
3291 | .permission = proc_tid_comm_permission, | |
3292 | }; | |
3293 | ||
28a6d671 EB |
3294 | /* |
3295 | * Tasks | |
3296 | */ | |
c5141e6d | 3297 | static const struct pid_entry tid_base_stuff[] = { |
631f9c18 | 3298 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), |
3835541d | 3299 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 3300 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
6ba8ed79 EB |
3301 | #ifdef CONFIG_NET |
3302 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), | |
3303 | #endif | |
631f9c18 | 3304 | REG("environ", S_IRUSR, proc_environ_operations), |
c5317167 | 3305 | REG("auxv", S_IRUSR, proc_auxv_operations), |
631f9c18 | 3306 | ONE("status", S_IRUGO, proc_pid_status), |
35a35046 | 3307 | ONE("personality", S_IRUSR, proc_pid_personality), |
1c963eb1 | 3308 | ONE("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 3309 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 3310 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
ebcb6734 | 3311 | #endif |
1b3044e3 JD |
3312 | NOD("comm", S_IFREG|S_IRUGO|S_IWUSR, |
3313 | &proc_tid_comm_inode_operations, | |
3314 | &proc_pid_set_comm_operations, {}), | |
ebcb6734 | 3315 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 | 3316 | ONE("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 3317 | #endif |
c2c0bb44 | 3318 | REG("cmdline", S_IRUGO, proc_pid_cmdline_ops), |
631f9c18 AD |
3319 | ONE("stat", S_IRUGO, proc_tid_stat), |
3320 | ONE("statm", S_IRUGO, proc_pid_statm), | |
b7643757 | 3321 | REG("maps", S_IRUGO, proc_tid_maps_operations), |
2e13ba54 | 3322 | #ifdef CONFIG_PROC_CHILDREN |
81841161 CG |
3323 | REG("children", S_IRUGO, proc_tid_children_operations), |
3324 | #endif | |
28a6d671 | 3325 | #ifdef CONFIG_NUMA |
b7643757 | 3326 | REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations), |
28a6d671 | 3327 | #endif |
631f9c18 AD |
3328 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
3329 | LNK("cwd", proc_cwd_link), | |
3330 | LNK("root", proc_root_link), | |
3331 | LNK("exe", proc_exe_link), | |
3332 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
3333 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
1e883281 | 3334 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 3335 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
b7643757 | 3336 | REG("smaps", S_IRUGO, proc_tid_smaps_operations), |
493b0e9d | 3337 | REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations), |
32ed74a4 | 3338 | REG("pagemap", S_IRUSR, proc_pagemap_operations), |
28a6d671 EB |
3339 | #endif |
3340 | #ifdef CONFIG_SECURITY | |
631f9c18 | 3341 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
3342 | #endif |
3343 | #ifdef CONFIG_KALLSYMS | |
edfcd606 | 3344 | ONE("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 3345 | #endif |
2ec220e2 | 3346 | #ifdef CONFIG_STACKTRACE |
35a35046 | 3347 | ONE("stack", S_IRUSR, proc_pid_stack), |
28a6d671 | 3348 | #endif |
5968cece | 3349 | #ifdef CONFIG_SCHED_INFO |
f6e826ca | 3350 | ONE("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 3351 | #endif |
9745512c | 3352 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 3353 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 3354 | #endif |
8793d854 | 3355 | #ifdef CONFIG_PROC_PID_CPUSET |
52de4779 | 3356 | ONE("cpuset", S_IRUGO, proc_cpuset_show), |
a424316c PM |
3357 | #endif |
3358 | #ifdef CONFIG_CGROUPS | |
006f4ac4 | 3359 | ONE("cgroup", S_IRUGO, proc_cgroup_show), |
28a6d671 | 3360 | #endif |
6ba51e37 | 3361 | ONE("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 3362 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 3363 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
28a6d671 | 3364 | #ifdef CONFIG_AUDITSYSCALL |
631f9c18 | 3365 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
26ec3c64 | 3366 | REG("sessionid", S_IRUGO, proc_sessionid_operations), |
28a6d671 | 3367 | #endif |
f4f154fd | 3368 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 3369 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
1203c8e6 | 3370 | REG("fail-nth", 0644, proc_fail_nth_operations), |
f4f154fd | 3371 | #endif |
297c5d92 | 3372 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 3373 | ONE("io", S_IRUSR, proc_tid_io_accounting), |
297c5d92 | 3374 | #endif |
22d917d8 EB |
3375 | #ifdef CONFIG_USER_NS |
3376 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
3377 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 3378 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
9cc46516 | 3379 | REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations), |
22d917d8 | 3380 | #endif |
7c23b330 JP |
3381 | #ifdef CONFIG_LIVEPATCH |
3382 | ONE("patch_state", S_IRUSR, proc_pid_patch_state), | |
3383 | #endif | |
28a6d671 EB |
3384 | }; |
3385 | ||
f0c3b509 | 3386 | static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx) |
28a6d671 | 3387 | { |
f0c3b509 AV |
3388 | return proc_pident_readdir(file, ctx, |
3389 | tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
3390 | } |
3391 | ||
00cd8dd3 AV |
3392 | static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
3393 | { | |
7bcd6b0e EB |
3394 | return proc_pident_lookup(dir, dentry, |
3395 | tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
3396 | } |
3397 | ||
00977a59 | 3398 | static const struct file_operations proc_tid_base_operations = { |
28a6d671 | 3399 | .read = generic_read_dir, |
f50752ea AV |
3400 | .iterate_shared = proc_tid_base_readdir, |
3401 | .llseek = generic_file_llseek, | |
28a6d671 EB |
3402 | }; |
3403 | ||
c5ef1c42 | 3404 | static const struct inode_operations proc_tid_base_inode_operations = { |
28a6d671 EB |
3405 | .lookup = proc_tid_base_lookup, |
3406 | .getattr = pid_getattr, | |
3407 | .setattr = proc_setattr, | |
3408 | }; | |
3409 | ||
0168b9e3 AV |
3410 | static struct dentry *proc_task_instantiate(struct dentry *dentry, |
3411 | struct task_struct *task, const void *ptr) | |
444ceed8 | 3412 | { |
444ceed8 | 3413 | struct inode *inode; |
0168b9e3 | 3414 | inode = proc_pid_make_inode(dentry->d_sb, task, S_IFDIR | S_IRUGO | S_IXUGO); |
444ceed8 | 3415 | if (!inode) |
0168b9e3 | 3416 | return ERR_PTR(-ENOENT); |
1bbc5513 | 3417 | |
444ceed8 EB |
3418 | inode->i_op = &proc_tid_base_inode_operations; |
3419 | inode->i_fop = &proc_tid_base_operations; | |
1bbc5513 | 3420 | inode->i_flags |= S_IMMUTABLE; |
aed54175 | 3421 | |
1270dd8d | 3422 | set_nlink(inode, nlink_tid); |
1bbc5513 | 3423 | pid_update_inode(task, inode); |
444ceed8 | 3424 | |
fb045adb | 3425 | d_set_d_op(dentry, &pid_dentry_operations); |
0168b9e3 | 3426 | return d_splice_alias(inode, dentry); |
444ceed8 EB |
3427 | } |
3428 | ||
00cd8dd3 | 3429 | static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags) |
28a6d671 | 3430 | { |
28a6d671 EB |
3431 | struct task_struct *task; |
3432 | struct task_struct *leader = get_proc_task(dir); | |
28a6d671 | 3433 | unsigned tid; |
b488893a | 3434 | struct pid_namespace *ns; |
0168b9e3 | 3435 | struct dentry *result = ERR_PTR(-ENOENT); |
28a6d671 EB |
3436 | |
3437 | if (!leader) | |
3438 | goto out_no_task; | |
3439 | ||
dbcdb504 | 3440 | tid = name_to_int(&dentry->d_name); |
28a6d671 EB |
3441 | if (tid == ~0U) |
3442 | goto out; | |
3443 | ||
b488893a | 3444 | ns = dentry->d_sb->s_fs_info; |
28a6d671 | 3445 | rcu_read_lock(); |
b488893a | 3446 | task = find_task_by_pid_ns(tid, ns); |
28a6d671 EB |
3447 | if (task) |
3448 | get_task_struct(task); | |
3449 | rcu_read_unlock(); | |
3450 | if (!task) | |
3451 | goto out; | |
bac0abd6 | 3452 | if (!same_thread_group(leader, task)) |
28a6d671 EB |
3453 | goto out_drop_task; |
3454 | ||
0168b9e3 | 3455 | result = proc_task_instantiate(dentry, task, NULL); |
28a6d671 EB |
3456 | out_drop_task: |
3457 | put_task_struct(task); | |
3458 | out: | |
3459 | put_task_struct(leader); | |
3460 | out_no_task: | |
0168b9e3 | 3461 | return result; |
28a6d671 EB |
3462 | } |
3463 | ||
0bc58a91 EB |
3464 | /* |
3465 | * Find the first tid of a thread group to return to user space. | |
3466 | * | |
3467 | * Usually this is just the thread group leader, but if the users | |
3468 | * buffer was too small or there was a seek into the middle of the | |
3469 | * directory we have more work todo. | |
3470 | * | |
3471 | * In the case of a short read we start with find_task_by_pid. | |
3472 | * | |
3473 | * In the case of a seek we start with the leader and walk nr | |
3474 | * threads past it. | |
3475 | */ | |
9f6e963f ON |
3476 | static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos, |
3477 | struct pid_namespace *ns) | |
0bc58a91 | 3478 | { |
d855a4b7 | 3479 | struct task_struct *pos, *task; |
9f6e963f ON |
3480 | unsigned long nr = f_pos; |
3481 | ||
3482 | if (nr != f_pos) /* 32bit overflow? */ | |
3483 | return NULL; | |
1da177e4 | 3484 | |
cc288738 | 3485 | rcu_read_lock(); |
d855a4b7 ON |
3486 | task = pid_task(pid, PIDTYPE_PID); |
3487 | if (!task) | |
3488 | goto fail; | |
3489 | ||
3490 | /* Attempt to start with the tid of a thread */ | |
9f6e963f | 3491 | if (tid && nr) { |
b488893a | 3492 | pos = find_task_by_pid_ns(tid, ns); |
d855a4b7 | 3493 | if (pos && same_thread_group(pos, task)) |
a872ff0c | 3494 | goto found; |
0bc58a91 | 3495 | } |
1da177e4 | 3496 | |
0bc58a91 | 3497 | /* If nr exceeds the number of threads there is nothing todo */ |
9f6e963f | 3498 | if (nr >= get_nr_threads(task)) |
c986c14a | 3499 | goto fail; |
1da177e4 | 3500 | |
a872ff0c ON |
3501 | /* If we haven't found our starting place yet start |
3502 | * with the leader and walk nr threads forward. | |
0bc58a91 | 3503 | */ |
d855a4b7 | 3504 | pos = task = task->group_leader; |
c986c14a | 3505 | do { |
9f6e963f | 3506 | if (!nr--) |
c986c14a | 3507 | goto found; |
d855a4b7 | 3508 | } while_each_thread(task, pos); |
c986c14a ON |
3509 | fail: |
3510 | pos = NULL; | |
3511 | goto out; | |
a872ff0c ON |
3512 | found: |
3513 | get_task_struct(pos); | |
3514 | out: | |
cc288738 | 3515 | rcu_read_unlock(); |
0bc58a91 EB |
3516 | return pos; |
3517 | } | |
3518 | ||
3519 | /* | |
3520 | * Find the next thread in the thread list. | |
3521 | * Return NULL if there is an error or no next thread. | |
3522 | * | |
3523 | * The reference to the input task_struct is released. | |
3524 | */ | |
3525 | static struct task_struct *next_tid(struct task_struct *start) | |
3526 | { | |
c1df7fb8 | 3527 | struct task_struct *pos = NULL; |
cc288738 | 3528 | rcu_read_lock(); |
c1df7fb8 | 3529 | if (pid_alive(start)) { |
0bc58a91 | 3530 | pos = next_thread(start); |
c1df7fb8 ON |
3531 | if (thread_group_leader(pos)) |
3532 | pos = NULL; | |
3533 | else | |
3534 | get_task_struct(pos); | |
3535 | } | |
cc288738 | 3536 | rcu_read_unlock(); |
0bc58a91 EB |
3537 | put_task_struct(start); |
3538 | return pos; | |
1da177e4 LT |
3539 | } |
3540 | ||
3541 | /* for the /proc/TGID/task/ directories */ | |
f0c3b509 | 3542 | static int proc_task_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3543 | { |
d855a4b7 ON |
3544 | struct inode *inode = file_inode(file); |
3545 | struct task_struct *task; | |
b488893a | 3546 | struct pid_namespace *ns; |
f0c3b509 | 3547 | int tid; |
1da177e4 | 3548 | |
d855a4b7 | 3549 | if (proc_inode_is_dead(inode)) |
f0c3b509 | 3550 | return -ENOENT; |
1da177e4 | 3551 | |
f0c3b509 | 3552 | if (!dir_emit_dots(file, ctx)) |
d855a4b7 | 3553 | return 0; |
1da177e4 | 3554 | |
0bc58a91 EB |
3555 | /* f_version caches the tgid value that the last readdir call couldn't |
3556 | * return. lseek aka telldir automagically resets f_version to 0. | |
3557 | */ | |
76f668be | 3558 | ns = proc_pid_ns(inode); |
f0c3b509 AV |
3559 | tid = (int)file->f_version; |
3560 | file->f_version = 0; | |
d855a4b7 | 3561 | for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns); |
0bc58a91 | 3562 | task; |
f0c3b509 | 3563 | task = next_tid(task), ctx->pos++) { |
e3912ac3 | 3564 | char name[10 + 1]; |
f0c3b509 | 3565 | int len; |
b488893a | 3566 | tid = task_pid_nr_ns(task, ns); |
e3912ac3 | 3567 | len = snprintf(name, sizeof(name), "%u", tid); |
f0c3b509 AV |
3568 | if (!proc_fill_cache(file, ctx, name, len, |
3569 | proc_task_instantiate, task, NULL)) { | |
0bc58a91 EB |
3570 | /* returning this tgid failed, save it as the first |
3571 | * pid for the next readir call */ | |
f0c3b509 | 3572 | file->f_version = (u64)tid; |
0bc58a91 | 3573 | put_task_struct(task); |
1da177e4 | 3574 | break; |
0bc58a91 | 3575 | } |
1da177e4 | 3576 | } |
d855a4b7 | 3577 | |
f0c3b509 | 3578 | return 0; |
1da177e4 | 3579 | } |
6e66b52b | 3580 | |
a528d35e DH |
3581 | static int proc_task_getattr(const struct path *path, struct kstat *stat, |
3582 | u32 request_mask, unsigned int query_flags) | |
6e66b52b | 3583 | { |
a528d35e | 3584 | struct inode *inode = d_inode(path->dentry); |
99f89551 | 3585 | struct task_struct *p = get_proc_task(inode); |
6e66b52b EB |
3586 | generic_fillattr(inode, stat); |
3587 | ||
99f89551 | 3588 | if (p) { |
99f89551 | 3589 | stat->nlink += get_nr_threads(p); |
99f89551 | 3590 | put_task_struct(p); |
6e66b52b EB |
3591 | } |
3592 | ||
3593 | return 0; | |
3594 | } | |
28a6d671 | 3595 | |
c5ef1c42 | 3596 | static const struct inode_operations proc_task_inode_operations = { |
28a6d671 EB |
3597 | .lookup = proc_task_lookup, |
3598 | .getattr = proc_task_getattr, | |
3599 | .setattr = proc_setattr, | |
0499680a | 3600 | .permission = proc_pid_permission, |
28a6d671 EB |
3601 | }; |
3602 | ||
00977a59 | 3603 | static const struct file_operations proc_task_operations = { |
28a6d671 | 3604 | .read = generic_read_dir, |
f50752ea AV |
3605 | .iterate_shared = proc_task_readdir, |
3606 | .llseek = generic_file_llseek, | |
28a6d671 | 3607 | }; |
1270dd8d AD |
3608 | |
3609 | void __init set_proc_pid_nlink(void) | |
3610 | { | |
3611 | nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
3612 | nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
3613 | } |