]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/proc/base.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | * | |
6 | * proc base directory handling functions | |
7 | * | |
8 | * 1999, Al Viro. Rewritten. Now it covers the whole per-process part. | |
9 | * Instead of using magical inumbers to determine the kind of object | |
10 | * we allocate and fill in-core inodes upon lookup. They don't even | |
11 | * go into icache. We cache the reference to task_struct upon lookup too. | |
12 | * Eventually it should become a filesystem in its own. We don't use the | |
13 | * rest of procfs anymore. | |
e070ad49 ML |
14 | * |
15 | * | |
16 | * Changelog: | |
17 | * 17-Jan-2005 | |
18 | * Allan Bezerra | |
19 | * Bruna Moreira <bruna.moreira@indt.org.br> | |
20 | * Edjard Mota <edjard.mota@indt.org.br> | |
21 | * Ilias Biris <ilias.biris@indt.org.br> | |
22 | * Mauricio Lin <mauricio.lin@indt.org.br> | |
23 | * | |
24 | * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT | |
25 | * | |
26 | * A new process specific entry (smaps) included in /proc. It shows the | |
27 | * size of rss for each memory area. The maps entry lacks information | |
28 | * about physical memory size (rss) for each mapped file, i.e., | |
29 | * rss information for executables and library files. | |
30 | * This additional information is useful for any tools that need to know | |
31 | * about physical memory consumption for a process specific library. | |
32 | * | |
33 | * Changelog: | |
34 | * 21-Feb-2005 | |
35 | * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT | |
36 | * Pud inclusion in the page table walking. | |
37 | * | |
38 | * ChangeLog: | |
39 | * 10-Mar-2005 | |
40 | * 10LE Instituto Nokia de Tecnologia - INdT: | |
41 | * A better way to walks through the page table as suggested by Hugh Dickins. | |
42 | * | |
43 | * Simo Piiroinen <simo.piiroinen@nokia.com>: | |
44 | * Smaps information related to shared, private, clean and dirty pages. | |
45 | * | |
46 | * Paul Mundt <paul.mundt@nokia.com>: | |
47 | * Overall revision about smaps. | |
1da177e4 LT |
48 | */ |
49 | ||
50 | #include <asm/uaccess.h> | |
51 | ||
1da177e4 LT |
52 | #include <linux/errno.h> |
53 | #include <linux/time.h> | |
54 | #include <linux/proc_fs.h> | |
55 | #include <linux/stat.h> | |
5995477a | 56 | #include <linux/task_io_accounting_ops.h> |
1da177e4 | 57 | #include <linux/init.h> |
16f7e0fe | 58 | #include <linux/capability.h> |
1da177e4 | 59 | #include <linux/file.h> |
9f3acc31 | 60 | #include <linux/fdtable.h> |
1da177e4 LT |
61 | #include <linux/string.h> |
62 | #include <linux/seq_file.h> | |
63 | #include <linux/namei.h> | |
6b3286ed | 64 | #include <linux/mnt_namespace.h> |
1da177e4 | 65 | #include <linux/mm.h> |
a63d83f4 | 66 | #include <linux/swap.h> |
b835996f | 67 | #include <linux/rcupdate.h> |
1da177e4 | 68 | #include <linux/kallsyms.h> |
2ec220e2 | 69 | #include <linux/stacktrace.h> |
d85f50d5 | 70 | #include <linux/resource.h> |
5096add8 | 71 | #include <linux/module.h> |
1da177e4 LT |
72 | #include <linux/mount.h> |
73 | #include <linux/security.h> | |
74 | #include <linux/ptrace.h> | |
0d094efe | 75 | #include <linux/tracehook.h> |
a424316c | 76 | #include <linux/cgroup.h> |
1da177e4 LT |
77 | #include <linux/cpuset.h> |
78 | #include <linux/audit.h> | |
5addc5dd | 79 | #include <linux/poll.h> |
1651e14e | 80 | #include <linux/nsproxy.h> |
8ac773b4 | 81 | #include <linux/oom.h> |
3cb4a0bb | 82 | #include <linux/elf.h> |
60347f67 | 83 | #include <linux/pid_namespace.h> |
5ad4e53b | 84 | #include <linux/fs_struct.h> |
5a0e3ad6 | 85 | #include <linux/slab.h> |
f133ecca CM |
86 | #ifdef CONFIG_HARDWALL |
87 | #include <asm/hardwall.h> | |
88 | #endif | |
43d2b113 | 89 | #include <trace/events/oom.h> |
1da177e4 LT |
90 | #include "internal.h" |
91 | ||
0f2fe20f EB |
92 | /* NOTE: |
93 | * Implementing inode permission operations in /proc is almost | |
94 | * certainly an error. Permission checks need to happen during | |
95 | * each system call not at open time. The reason is that most of | |
96 | * what we wish to check for permissions in /proc varies at runtime. | |
97 | * | |
98 | * The classic example of a problem is opening file descriptors | |
99 | * in /proc for a task before it execs a suid executable. | |
100 | */ | |
101 | ||
1da177e4 | 102 | struct pid_entry { |
1da177e4 | 103 | char *name; |
c5141e6d | 104 | int len; |
d161a13f | 105 | umode_t mode; |
c5ef1c42 | 106 | const struct inode_operations *iop; |
00977a59 | 107 | const struct file_operations *fop; |
20cdc894 | 108 | union proc_op op; |
1da177e4 LT |
109 | }; |
110 | ||
61a28784 | 111 | #define NOD(NAME, MODE, IOP, FOP, OP) { \ |
20cdc894 | 112 | .name = (NAME), \ |
c5141e6d | 113 | .len = sizeof(NAME) - 1, \ |
20cdc894 EB |
114 | .mode = MODE, \ |
115 | .iop = IOP, \ | |
116 | .fop = FOP, \ | |
117 | .op = OP, \ | |
118 | } | |
119 | ||
631f9c18 AD |
120 | #define DIR(NAME, MODE, iops, fops) \ |
121 | NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} ) | |
122 | #define LNK(NAME, get_link) \ | |
61a28784 | 123 | NOD(NAME, (S_IFLNK|S_IRWXUGO), \ |
20cdc894 | 124 | &proc_pid_link_inode_operations, NULL, \ |
631f9c18 AD |
125 | { .proc_get_link = get_link } ) |
126 | #define REG(NAME, MODE, fops) \ | |
127 | NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {}) | |
128 | #define INF(NAME, MODE, read) \ | |
61a28784 | 129 | NOD(NAME, (S_IFREG|(MODE)), \ |
20cdc894 | 130 | NULL, &proc_info_file_operations, \ |
631f9c18 AD |
131 | { .proc_read = read } ) |
132 | #define ONE(NAME, MODE, show) \ | |
be614086 EB |
133 | NOD(NAME, (S_IFREG|(MODE)), \ |
134 | NULL, &proc_single_file_operations, \ | |
631f9c18 | 135 | { .proc_show = show } ) |
1da177e4 | 136 | |
aed54175 VN |
137 | /* |
138 | * Count the number of hardlinks for the pid_entry table, excluding the . | |
139 | * and .. links. | |
140 | */ | |
141 | static unsigned int pid_entry_count_dirs(const struct pid_entry *entries, | |
142 | unsigned int n) | |
143 | { | |
144 | unsigned int i; | |
145 | unsigned int count; | |
146 | ||
147 | count = 0; | |
148 | for (i = 0; i < n; ++i) { | |
149 | if (S_ISDIR(entries[i].mode)) | |
150 | ++count; | |
151 | } | |
152 | ||
153 | return count; | |
154 | } | |
155 | ||
f7ad3c6b | 156 | static int get_task_root(struct task_struct *task, struct path *root) |
1da177e4 | 157 | { |
7c2c7d99 HD |
158 | int result = -ENOENT; |
159 | ||
0494f6ec | 160 | task_lock(task); |
f7ad3c6b MS |
161 | if (task->fs) { |
162 | get_fs_root(task->fs, root); | |
7c2c7d99 HD |
163 | result = 0; |
164 | } | |
0494f6ec | 165 | task_unlock(task); |
7c2c7d99 | 166 | return result; |
0494f6ec MS |
167 | } |
168 | ||
3dcd25f3 | 169 | static int proc_cwd_link(struct inode *inode, struct path *path) |
0494f6ec | 170 | { |
99f89551 | 171 | struct task_struct *task = get_proc_task(inode); |
0494f6ec | 172 | int result = -ENOENT; |
99f89551 EB |
173 | |
174 | if (task) { | |
f7ad3c6b MS |
175 | task_lock(task); |
176 | if (task->fs) { | |
177 | get_fs_pwd(task->fs, path); | |
178 | result = 0; | |
179 | } | |
180 | task_unlock(task); | |
99f89551 EB |
181 | put_task_struct(task); |
182 | } | |
1da177e4 LT |
183 | return result; |
184 | } | |
185 | ||
3dcd25f3 | 186 | static int proc_root_link(struct inode *inode, struct path *path) |
1da177e4 | 187 | { |
99f89551 | 188 | struct task_struct *task = get_proc_task(inode); |
1da177e4 | 189 | int result = -ENOENT; |
99f89551 EB |
190 | |
191 | if (task) { | |
f7ad3c6b | 192 | result = get_task_root(task, path); |
99f89551 EB |
193 | put_task_struct(task); |
194 | } | |
1da177e4 LT |
195 | return result; |
196 | } | |
197 | ||
8b0db9db | 198 | static struct mm_struct *__check_mem_permission(struct task_struct *task) |
638fa202 | 199 | { |
8b0db9db SW |
200 | struct mm_struct *mm; |
201 | ||
202 | mm = get_task_mm(task); | |
203 | if (!mm) | |
204 | return ERR_PTR(-EINVAL); | |
205 | ||
638fa202 RM |
206 | /* |
207 | * A task can always look at itself, in case it chooses | |
208 | * to use system calls instead of load instructions. | |
209 | */ | |
210 | if (task == current) | |
8b0db9db | 211 | return mm; |
638fa202 RM |
212 | |
213 | /* | |
214 | * If current is actively ptrace'ing, and would also be | |
215 | * permitted to freshly attach with ptrace now, permit it. | |
216 | */ | |
0d094efe RM |
217 | if (task_is_stopped_or_traced(task)) { |
218 | int match; | |
219 | rcu_read_lock(); | |
06d98473 | 220 | match = (ptrace_parent(task) == current); |
0d094efe RM |
221 | rcu_read_unlock(); |
222 | if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH)) | |
8b0db9db | 223 | return mm; |
0d094efe | 224 | } |
638fa202 RM |
225 | |
226 | /* | |
25985edc | 227 | * No one else is allowed. |
638fa202 | 228 | */ |
8b0db9db SW |
229 | mmput(mm); |
230 | return ERR_PTR(-EPERM); | |
638fa202 | 231 | } |
1da177e4 | 232 | |
18f661bc | 233 | /* |
8b0db9db SW |
234 | * If current may access user memory in @task return a reference to the |
235 | * corresponding mm, otherwise ERR_PTR. | |
18f661bc | 236 | */ |
8b0db9db | 237 | static struct mm_struct *check_mem_permission(struct task_struct *task) |
18f661bc | 238 | { |
8b0db9db | 239 | struct mm_struct *mm; |
18f661bc SW |
240 | int err; |
241 | ||
242 | /* | |
243 | * Avoid racing if task exec's as we might get a new mm but validate | |
244 | * against old credentials. | |
245 | */ | |
246 | err = mutex_lock_killable(&task->signal->cred_guard_mutex); | |
247 | if (err) | |
8b0db9db | 248 | return ERR_PTR(err); |
18f661bc | 249 | |
8b0db9db | 250 | mm = __check_mem_permission(task); |
18f661bc SW |
251 | mutex_unlock(&task->signal->cred_guard_mutex); |
252 | ||
8b0db9db | 253 | return mm; |
638fa202 | 254 | } |
1da177e4 | 255 | |
831830b5 AV |
256 | struct mm_struct *mm_for_maps(struct task_struct *task) |
257 | { | |
704b836c | 258 | struct mm_struct *mm; |
ec6fd8a4 | 259 | int err; |
704b836c | 260 | |
ec6fd8a4 AV |
261 | err = mutex_lock_killable(&task->signal->cred_guard_mutex); |
262 | if (err) | |
263 | return ERR_PTR(err); | |
00f89d21 | 264 | |
704b836c ON |
265 | mm = get_task_mm(task); |
266 | if (mm && mm != current->mm && | |
267 | !ptrace_may_access(task, PTRACE_MODE_READ)) { | |
268 | mmput(mm); | |
ec6fd8a4 | 269 | mm = ERR_PTR(-EACCES); |
13f0feaf | 270 | } |
9b1bf12d | 271 | mutex_unlock(&task->signal->cred_guard_mutex); |
704b836c | 272 | |
831830b5 | 273 | return mm; |
831830b5 AV |
274 | } |
275 | ||
1da177e4 LT |
276 | static int proc_pid_cmdline(struct task_struct *task, char * buffer) |
277 | { | |
278 | int res = 0; | |
279 | unsigned int len; | |
280 | struct mm_struct *mm = get_task_mm(task); | |
281 | if (!mm) | |
282 | goto out; | |
283 | if (!mm->arg_end) | |
284 | goto out_mm; /* Shh! No looking before we're done */ | |
285 | ||
286 | len = mm->arg_end - mm->arg_start; | |
287 | ||
288 | if (len > PAGE_SIZE) | |
289 | len = PAGE_SIZE; | |
290 | ||
291 | res = access_process_vm(task, mm->arg_start, buffer, len, 0); | |
292 | ||
293 | // If the nul at the end of args has been overwritten, then | |
294 | // assume application is using setproctitle(3). | |
295 | if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) { | |
296 | len = strnlen(buffer, res); | |
297 | if (len < res) { | |
298 | res = len; | |
299 | } else { | |
300 | len = mm->env_end - mm->env_start; | |
301 | if (len > PAGE_SIZE - res) | |
302 | len = PAGE_SIZE - res; | |
303 | res += access_process_vm(task, mm->env_start, buffer+res, len, 0); | |
304 | res = strnlen(buffer, res); | |
305 | } | |
306 | } | |
307 | out_mm: | |
308 | mmput(mm); | |
309 | out: | |
310 | return res; | |
311 | } | |
312 | ||
313 | static int proc_pid_auxv(struct task_struct *task, char *buffer) | |
314 | { | |
2fadaef4 AV |
315 | struct mm_struct *mm = mm_for_maps(task); |
316 | int res = PTR_ERR(mm); | |
317 | if (mm && !IS_ERR(mm)) { | |
1da177e4 | 318 | unsigned int nwords = 0; |
dfe6b7d9 | 319 | do { |
1da177e4 | 320 | nwords += 2; |
dfe6b7d9 | 321 | } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */ |
1da177e4 LT |
322 | res = nwords * sizeof(mm->saved_auxv[0]); |
323 | if (res > PAGE_SIZE) | |
324 | res = PAGE_SIZE; | |
325 | memcpy(buffer, mm->saved_auxv, res); | |
326 | mmput(mm); | |
327 | } | |
328 | return res; | |
329 | } | |
330 | ||
331 | ||
332 | #ifdef CONFIG_KALLSYMS | |
333 | /* | |
334 | * Provides a wchan file via kallsyms in a proper one-value-per-file format. | |
335 | * Returns the resolved symbol. If that fails, simply return the address. | |
336 | */ | |
337 | static int proc_pid_wchan(struct task_struct *task, char *buffer) | |
338 | { | |
ffb45122 | 339 | unsigned long wchan; |
9281acea | 340 | char symname[KSYM_NAME_LEN]; |
1da177e4 LT |
341 | |
342 | wchan = get_wchan(task); | |
343 | ||
9d65cb4a | 344 | if (lookup_symbol_name(wchan, symname) < 0) |
f83ce3e6 JE |
345 | if (!ptrace_may_access(task, PTRACE_MODE_READ)) |
346 | return 0; | |
347 | else | |
348 | return sprintf(buffer, "%lu", wchan); | |
9d65cb4a AD |
349 | else |
350 | return sprintf(buffer, "%s", symname); | |
1da177e4 LT |
351 | } |
352 | #endif /* CONFIG_KALLSYMS */ | |
353 | ||
a9712bc1 AV |
354 | static int lock_trace(struct task_struct *task) |
355 | { | |
356 | int err = mutex_lock_killable(&task->signal->cred_guard_mutex); | |
357 | if (err) | |
358 | return err; | |
359 | if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) { | |
360 | mutex_unlock(&task->signal->cred_guard_mutex); | |
361 | return -EPERM; | |
362 | } | |
363 | return 0; | |
364 | } | |
365 | ||
366 | static void unlock_trace(struct task_struct *task) | |
367 | { | |
368 | mutex_unlock(&task->signal->cred_guard_mutex); | |
369 | } | |
370 | ||
2ec220e2 KC |
371 | #ifdef CONFIG_STACKTRACE |
372 | ||
373 | #define MAX_STACK_TRACE_DEPTH 64 | |
374 | ||
375 | static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns, | |
376 | struct pid *pid, struct task_struct *task) | |
377 | { | |
378 | struct stack_trace trace; | |
379 | unsigned long *entries; | |
a9712bc1 | 380 | int err; |
2ec220e2 KC |
381 | int i; |
382 | ||
383 | entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL); | |
384 | if (!entries) | |
385 | return -ENOMEM; | |
386 | ||
387 | trace.nr_entries = 0; | |
388 | trace.max_entries = MAX_STACK_TRACE_DEPTH; | |
389 | trace.entries = entries; | |
390 | trace.skip = 0; | |
2ec220e2 | 391 | |
a9712bc1 AV |
392 | err = lock_trace(task); |
393 | if (!err) { | |
394 | save_stack_trace_tsk(task, &trace); | |
395 | ||
396 | for (i = 0; i < trace.nr_entries; i++) { | |
b81a618d | 397 | seq_printf(m, "[<%pK>] %pS\n", |
a9712bc1 AV |
398 | (void *)entries[i], (void *)entries[i]); |
399 | } | |
400 | unlock_trace(task); | |
2ec220e2 KC |
401 | } |
402 | kfree(entries); | |
403 | ||
a9712bc1 | 404 | return err; |
2ec220e2 KC |
405 | } |
406 | #endif | |
407 | ||
1da177e4 LT |
408 | #ifdef CONFIG_SCHEDSTATS |
409 | /* | |
410 | * Provides /proc/PID/schedstat | |
411 | */ | |
412 | static int proc_pid_schedstat(struct task_struct *task, char *buffer) | |
413 | { | |
172ba844 | 414 | return sprintf(buffer, "%llu %llu %lu\n", |
826e08b0 IM |
415 | (unsigned long long)task->se.sum_exec_runtime, |
416 | (unsigned long long)task->sched_info.run_delay, | |
2d72376b | 417 | task->sched_info.pcount); |
1da177e4 LT |
418 | } |
419 | #endif | |
420 | ||
9745512c AV |
421 | #ifdef CONFIG_LATENCYTOP |
422 | static int lstats_show_proc(struct seq_file *m, void *v) | |
423 | { | |
424 | int i; | |
13d77c37 HS |
425 | struct inode *inode = m->private; |
426 | struct task_struct *task = get_proc_task(inode); | |
9745512c | 427 | |
13d77c37 HS |
428 | if (!task) |
429 | return -ESRCH; | |
430 | seq_puts(m, "Latency Top version : v0.1\n"); | |
9745512c | 431 | for (i = 0; i < 32; i++) { |
34e49d4f JP |
432 | struct latency_record *lr = &task->latency_record[i]; |
433 | if (lr->backtrace[0]) { | |
9745512c | 434 | int q; |
34e49d4f JP |
435 | seq_printf(m, "%i %li %li", |
436 | lr->count, lr->time, lr->max); | |
9745512c | 437 | for (q = 0; q < LT_BACKTRACEDEPTH; q++) { |
34e49d4f JP |
438 | unsigned long bt = lr->backtrace[q]; |
439 | if (!bt) | |
9745512c | 440 | break; |
34e49d4f | 441 | if (bt == ULONG_MAX) |
9745512c | 442 | break; |
34e49d4f | 443 | seq_printf(m, " %ps", (void *)bt); |
9745512c | 444 | } |
9d6de12f | 445 | seq_putc(m, '\n'); |
9745512c AV |
446 | } |
447 | ||
448 | } | |
13d77c37 | 449 | put_task_struct(task); |
9745512c AV |
450 | return 0; |
451 | } | |
452 | ||
453 | static int lstats_open(struct inode *inode, struct file *file) | |
454 | { | |
13d77c37 | 455 | return single_open(file, lstats_show_proc, inode); |
d6643d12 HS |
456 | } |
457 | ||
9745512c AV |
458 | static ssize_t lstats_write(struct file *file, const char __user *buf, |
459 | size_t count, loff_t *offs) | |
460 | { | |
13d77c37 | 461 | struct task_struct *task = get_proc_task(file->f_dentry->d_inode); |
9745512c | 462 | |
13d77c37 HS |
463 | if (!task) |
464 | return -ESRCH; | |
9745512c | 465 | clear_all_latency_tracing(task); |
13d77c37 | 466 | put_task_struct(task); |
9745512c AV |
467 | |
468 | return count; | |
469 | } | |
470 | ||
471 | static const struct file_operations proc_lstats_operations = { | |
472 | .open = lstats_open, | |
473 | .read = seq_read, | |
474 | .write = lstats_write, | |
475 | .llseek = seq_lseek, | |
13d77c37 | 476 | .release = single_release, |
9745512c AV |
477 | }; |
478 | ||
479 | #endif | |
480 | ||
1da177e4 LT |
481 | static int proc_oom_score(struct task_struct *task, char *buffer) |
482 | { | |
b95c35e7 | 483 | unsigned long points = 0; |
1da177e4 | 484 | |
19c5d45a | 485 | read_lock(&tasklist_lock); |
b95c35e7 | 486 | if (pid_alive(task)) |
a63d83f4 DR |
487 | points = oom_badness(task, NULL, NULL, |
488 | totalram_pages + total_swap_pages); | |
19c5d45a | 489 | read_unlock(&tasklist_lock); |
1da177e4 LT |
490 | return sprintf(buffer, "%lu\n", points); |
491 | } | |
492 | ||
d85f50d5 NH |
493 | struct limit_names { |
494 | char *name; | |
495 | char *unit; | |
496 | }; | |
497 | ||
498 | static const struct limit_names lnames[RLIM_NLIMITS] = { | |
cff4edb5 | 499 | [RLIMIT_CPU] = {"Max cpu time", "seconds"}, |
d85f50d5 NH |
500 | [RLIMIT_FSIZE] = {"Max file size", "bytes"}, |
501 | [RLIMIT_DATA] = {"Max data size", "bytes"}, | |
502 | [RLIMIT_STACK] = {"Max stack size", "bytes"}, | |
503 | [RLIMIT_CORE] = {"Max core file size", "bytes"}, | |
504 | [RLIMIT_RSS] = {"Max resident set", "bytes"}, | |
505 | [RLIMIT_NPROC] = {"Max processes", "processes"}, | |
506 | [RLIMIT_NOFILE] = {"Max open files", "files"}, | |
507 | [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"}, | |
508 | [RLIMIT_AS] = {"Max address space", "bytes"}, | |
509 | [RLIMIT_LOCKS] = {"Max file locks", "locks"}, | |
510 | [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"}, | |
511 | [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"}, | |
512 | [RLIMIT_NICE] = {"Max nice priority", NULL}, | |
513 | [RLIMIT_RTPRIO] = {"Max realtime priority", NULL}, | |
8808117c | 514 | [RLIMIT_RTTIME] = {"Max realtime timeout", "us"}, |
d85f50d5 NH |
515 | }; |
516 | ||
517 | /* Display limits for a process */ | |
518 | static int proc_pid_limits(struct task_struct *task, char *buffer) | |
519 | { | |
520 | unsigned int i; | |
521 | int count = 0; | |
522 | unsigned long flags; | |
523 | char *bufptr = buffer; | |
524 | ||
525 | struct rlimit rlim[RLIM_NLIMITS]; | |
526 | ||
a6bebbc8 | 527 | if (!lock_task_sighand(task, &flags)) |
d85f50d5 | 528 | return 0; |
d85f50d5 NH |
529 | memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS); |
530 | unlock_task_sighand(task, &flags); | |
d85f50d5 NH |
531 | |
532 | /* | |
533 | * print the file header | |
534 | */ | |
535 | count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n", | |
536 | "Limit", "Soft Limit", "Hard Limit", "Units"); | |
537 | ||
538 | for (i = 0; i < RLIM_NLIMITS; i++) { | |
539 | if (rlim[i].rlim_cur == RLIM_INFINITY) | |
540 | count += sprintf(&bufptr[count], "%-25s %-20s ", | |
541 | lnames[i].name, "unlimited"); | |
542 | else | |
543 | count += sprintf(&bufptr[count], "%-25s %-20lu ", | |
544 | lnames[i].name, rlim[i].rlim_cur); | |
545 | ||
546 | if (rlim[i].rlim_max == RLIM_INFINITY) | |
547 | count += sprintf(&bufptr[count], "%-20s ", "unlimited"); | |
548 | else | |
549 | count += sprintf(&bufptr[count], "%-20lu ", | |
550 | rlim[i].rlim_max); | |
551 | ||
552 | if (lnames[i].unit) | |
553 | count += sprintf(&bufptr[count], "%-10s\n", | |
554 | lnames[i].unit); | |
555 | else | |
556 | count += sprintf(&bufptr[count], "\n"); | |
557 | } | |
558 | ||
559 | return count; | |
560 | } | |
561 | ||
ebcb6734 RM |
562 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
563 | static int proc_pid_syscall(struct task_struct *task, char *buffer) | |
564 | { | |
565 | long nr; | |
566 | unsigned long args[6], sp, pc; | |
a9712bc1 AV |
567 | int res = lock_trace(task); |
568 | if (res) | |
569 | return res; | |
ebcb6734 RM |
570 | |
571 | if (task_current_syscall(task, &nr, args, 6, &sp, &pc)) | |
a9712bc1 AV |
572 | res = sprintf(buffer, "running\n"); |
573 | else if (nr < 0) | |
574 | res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc); | |
575 | else | |
576 | res = sprintf(buffer, | |
ebcb6734 RM |
577 | "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n", |
578 | nr, | |
579 | args[0], args[1], args[2], args[3], args[4], args[5], | |
580 | sp, pc); | |
a9712bc1 AV |
581 | unlock_trace(task); |
582 | return res; | |
ebcb6734 RM |
583 | } |
584 | #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */ | |
585 | ||
1da177e4 LT |
586 | /************************************************************************/ |
587 | /* Here the fs part begins */ | |
588 | /************************************************************************/ | |
589 | ||
590 | /* permission checks */ | |
778c1144 | 591 | static int proc_fd_access_allowed(struct inode *inode) |
1da177e4 | 592 | { |
778c1144 EB |
593 | struct task_struct *task; |
594 | int allowed = 0; | |
df26c40e EB |
595 | /* Allow access to a task's file descriptors if it is us or we |
596 | * may use ptrace attach to the process and find out that | |
597 | * information. | |
778c1144 EB |
598 | */ |
599 | task = get_proc_task(inode); | |
df26c40e | 600 | if (task) { |
006ebb40 | 601 | allowed = ptrace_may_access(task, PTRACE_MODE_READ); |
778c1144 | 602 | put_task_struct(task); |
df26c40e | 603 | } |
778c1144 | 604 | return allowed; |
1da177e4 LT |
605 | } |
606 | ||
6b4e306a | 607 | int proc_setattr(struct dentry *dentry, struct iattr *attr) |
6d76fa58 LT |
608 | { |
609 | int error; | |
610 | struct inode *inode = dentry->d_inode; | |
611 | ||
612 | if (attr->ia_valid & ATTR_MODE) | |
613 | return -EPERM; | |
614 | ||
615 | error = inode_change_ok(inode, attr); | |
1025774c CH |
616 | if (error) |
617 | return error; | |
618 | ||
619 | if ((attr->ia_valid & ATTR_SIZE) && | |
620 | attr->ia_size != i_size_read(inode)) { | |
621 | error = vmtruncate(inode, attr->ia_size); | |
622 | if (error) | |
623 | return error; | |
624 | } | |
625 | ||
626 | setattr_copy(inode, attr); | |
627 | mark_inode_dirty(inode); | |
628 | return 0; | |
6d76fa58 LT |
629 | } |
630 | ||
c5ef1c42 | 631 | static const struct inode_operations proc_def_inode_operations = { |
6d76fa58 LT |
632 | .setattr = proc_setattr, |
633 | }; | |
634 | ||
1da177e4 LT |
635 | #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */ |
636 | ||
637 | static ssize_t proc_info_read(struct file * file, char __user * buf, | |
638 | size_t count, loff_t *ppos) | |
639 | { | |
2fddfeef | 640 | struct inode * inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
641 | unsigned long page; |
642 | ssize_t length; | |
99f89551 EB |
643 | struct task_struct *task = get_proc_task(inode); |
644 | ||
645 | length = -ESRCH; | |
646 | if (!task) | |
647 | goto out_no_task; | |
1da177e4 LT |
648 | |
649 | if (count > PROC_BLOCK_SIZE) | |
650 | count = PROC_BLOCK_SIZE; | |
99f89551 EB |
651 | |
652 | length = -ENOMEM; | |
e12ba74d | 653 | if (!(page = __get_free_page(GFP_TEMPORARY))) |
99f89551 | 654 | goto out; |
1da177e4 LT |
655 | |
656 | length = PROC_I(inode)->op.proc_read(task, (char*)page); | |
657 | ||
658 | if (length >= 0) | |
659 | length = simple_read_from_buffer(buf, count, ppos, (char *)page, length); | |
660 | free_page(page); | |
99f89551 EB |
661 | out: |
662 | put_task_struct(task); | |
663 | out_no_task: | |
1da177e4 LT |
664 | return length; |
665 | } | |
666 | ||
00977a59 | 667 | static const struct file_operations proc_info_file_operations = { |
1da177e4 | 668 | .read = proc_info_read, |
87df8424 | 669 | .llseek = generic_file_llseek, |
1da177e4 LT |
670 | }; |
671 | ||
be614086 EB |
672 | static int proc_single_show(struct seq_file *m, void *v) |
673 | { | |
674 | struct inode *inode = m->private; | |
675 | struct pid_namespace *ns; | |
676 | struct pid *pid; | |
677 | struct task_struct *task; | |
678 | int ret; | |
679 | ||
680 | ns = inode->i_sb->s_fs_info; | |
681 | pid = proc_pid(inode); | |
682 | task = get_pid_task(pid, PIDTYPE_PID); | |
683 | if (!task) | |
684 | return -ESRCH; | |
685 | ||
686 | ret = PROC_I(inode)->op.proc_show(m, ns, pid, task); | |
687 | ||
688 | put_task_struct(task); | |
689 | return ret; | |
690 | } | |
691 | ||
692 | static int proc_single_open(struct inode *inode, struct file *filp) | |
693 | { | |
c6a34058 | 694 | return single_open(filp, proc_single_show, inode); |
be614086 EB |
695 | } |
696 | ||
697 | static const struct file_operations proc_single_file_operations = { | |
698 | .open = proc_single_open, | |
699 | .read = seq_read, | |
700 | .llseek = seq_lseek, | |
701 | .release = single_release, | |
702 | }; | |
703 | ||
1da177e4 LT |
704 | static int mem_open(struct inode* inode, struct file* file) |
705 | { | |
706 | file->private_data = (void*)((long)current->self_exec_id); | |
4a3956c7 KH |
707 | /* OK to pass negative loff_t, we can catch out-of-range */ |
708 | file->f_mode |= FMODE_UNSIGNED_OFFSET; | |
1da177e4 LT |
709 | return 0; |
710 | } | |
711 | ||
712 | static ssize_t mem_read(struct file * file, char __user * buf, | |
713 | size_t count, loff_t *ppos) | |
714 | { | |
2fddfeef | 715 | struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode); |
1da177e4 LT |
716 | char *page; |
717 | unsigned long src = *ppos; | |
718 | int ret = -ESRCH; | |
719 | struct mm_struct *mm; | |
720 | ||
99f89551 EB |
721 | if (!task) |
722 | goto out_no_task; | |
723 | ||
1da177e4 | 724 | ret = -ENOMEM; |
e12ba74d | 725 | page = (char *)__get_free_page(GFP_TEMPORARY); |
1da177e4 LT |
726 | if (!page) |
727 | goto out; | |
728 | ||
8b0db9db SW |
729 | mm = check_mem_permission(task); |
730 | ret = PTR_ERR(mm); | |
731 | if (IS_ERR(mm)) | |
1da177e4 LT |
732 | goto out_free; |
733 | ||
734 | ret = -EIO; | |
735 | ||
736 | if (file->private_data != (void*)((long)current->self_exec_id)) | |
737 | goto out_put; | |
738 | ||
739 | ret = 0; | |
740 | ||
741 | while (count > 0) { | |
742 | int this_len, retval; | |
743 | ||
744 | this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count; | |
8b0db9db SW |
745 | retval = access_remote_vm(mm, src, page, this_len, 0); |
746 | if (!retval) { | |
1da177e4 LT |
747 | if (!ret) |
748 | ret = -EIO; | |
749 | break; | |
750 | } | |
751 | ||
752 | if (copy_to_user(buf, page, retval)) { | |
753 | ret = -EFAULT; | |
754 | break; | |
755 | } | |
756 | ||
757 | ret += retval; | |
758 | src += retval; | |
759 | buf += retval; | |
760 | count -= retval; | |
761 | } | |
762 | *ppos = src; | |
763 | ||
764 | out_put: | |
765 | mmput(mm); | |
766 | out_free: | |
767 | free_page((unsigned long) page); | |
768 | out: | |
99f89551 EB |
769 | put_task_struct(task); |
770 | out_no_task: | |
1da177e4 LT |
771 | return ret; |
772 | } | |
773 | ||
63967fa9 | 774 | static ssize_t mem_write(struct file * file, const char __user *buf, |
1da177e4 LT |
775 | size_t count, loff_t *ppos) |
776 | { | |
f7ca54f4 | 777 | int copied; |
1da177e4 | 778 | char *page; |
2fddfeef | 779 | struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode); |
1da177e4 | 780 | unsigned long dst = *ppos; |
8b0db9db | 781 | struct mm_struct *mm; |
1da177e4 | 782 | |
99f89551 EB |
783 | copied = -ESRCH; |
784 | if (!task) | |
785 | goto out_no_task; | |
786 | ||
30cd8903 KM |
787 | copied = -ENOMEM; |
788 | page = (char *)__get_free_page(GFP_TEMPORARY); | |
789 | if (!page) | |
790 | goto out_task; | |
791 | ||
8b0db9db SW |
792 | mm = check_mem_permission(task); |
793 | copied = PTR_ERR(mm); | |
794 | if (IS_ERR(mm)) | |
30cd8903 | 795 | goto out_free; |
1da177e4 | 796 | |
26947f8c SW |
797 | copied = -EIO; |
798 | if (file->private_data != (void *)((long)current->self_exec_id)) | |
8b0db9db | 799 | goto out_mm; |
1da177e4 | 800 | |
f7ca54f4 | 801 | copied = 0; |
1da177e4 LT |
802 | while (count > 0) { |
803 | int this_len, retval; | |
804 | ||
805 | this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count; | |
806 | if (copy_from_user(page, buf, this_len)) { | |
807 | copied = -EFAULT; | |
808 | break; | |
809 | } | |
8b0db9db | 810 | retval = access_remote_vm(mm, dst, page, this_len, 1); |
1da177e4 LT |
811 | if (!retval) { |
812 | if (!copied) | |
813 | copied = -EIO; | |
814 | break; | |
815 | } | |
816 | copied += retval; | |
817 | buf += retval; | |
818 | dst += retval; | |
819 | count -= retval; | |
820 | } | |
821 | *ppos = dst; | |
30cd8903 | 822 | |
8b0db9db SW |
823 | out_mm: |
824 | mmput(mm); | |
30cd8903 KM |
825 | out_free: |
826 | free_page((unsigned long) page); | |
8b0db9db | 827 | out_task: |
99f89551 EB |
828 | put_task_struct(task); |
829 | out_no_task: | |
1da177e4 LT |
830 | return copied; |
831 | } | |
1da177e4 | 832 | |
85863e47 | 833 | loff_t mem_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
834 | { |
835 | switch (orig) { | |
836 | case 0: | |
837 | file->f_pos = offset; | |
838 | break; | |
839 | case 1: | |
840 | file->f_pos += offset; | |
841 | break; | |
842 | default: | |
843 | return -EINVAL; | |
844 | } | |
845 | force_successful_syscall_return(); | |
846 | return file->f_pos; | |
847 | } | |
848 | ||
00977a59 | 849 | static const struct file_operations proc_mem_operations = { |
1da177e4 LT |
850 | .llseek = mem_lseek, |
851 | .read = mem_read, | |
852 | .write = mem_write, | |
853 | .open = mem_open, | |
854 | }; | |
855 | ||
315e28c8 JP |
856 | static ssize_t environ_read(struct file *file, char __user *buf, |
857 | size_t count, loff_t *ppos) | |
858 | { | |
859 | struct task_struct *task = get_proc_task(file->f_dentry->d_inode); | |
860 | char *page; | |
861 | unsigned long src = *ppos; | |
862 | int ret = -ESRCH; | |
863 | struct mm_struct *mm; | |
864 | ||
865 | if (!task) | |
866 | goto out_no_task; | |
867 | ||
315e28c8 JP |
868 | ret = -ENOMEM; |
869 | page = (char *)__get_free_page(GFP_TEMPORARY); | |
870 | if (!page) | |
871 | goto out; | |
872 | ||
315e28c8 | 873 | |
d6f64b89 AV |
874 | mm = mm_for_maps(task); |
875 | ret = PTR_ERR(mm); | |
876 | if (!mm || IS_ERR(mm)) | |
315e28c8 JP |
877 | goto out_free; |
878 | ||
d6f64b89 | 879 | ret = 0; |
315e28c8 JP |
880 | while (count > 0) { |
881 | int this_len, retval, max_len; | |
882 | ||
883 | this_len = mm->env_end - (mm->env_start + src); | |
884 | ||
885 | if (this_len <= 0) | |
886 | break; | |
887 | ||
888 | max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count; | |
889 | this_len = (this_len > max_len) ? max_len : this_len; | |
890 | ||
891 | retval = access_process_vm(task, (mm->env_start + src), | |
892 | page, this_len, 0); | |
893 | ||
894 | if (retval <= 0) { | |
895 | ret = retval; | |
896 | break; | |
897 | } | |
898 | ||
899 | if (copy_to_user(buf, page, retval)) { | |
900 | ret = -EFAULT; | |
901 | break; | |
902 | } | |
903 | ||
904 | ret += retval; | |
905 | src += retval; | |
906 | buf += retval; | |
907 | count -= retval; | |
908 | } | |
909 | *ppos = src; | |
910 | ||
911 | mmput(mm); | |
912 | out_free: | |
913 | free_page((unsigned long) page); | |
914 | out: | |
915 | put_task_struct(task); | |
916 | out_no_task: | |
917 | return ret; | |
918 | } | |
919 | ||
920 | static const struct file_operations proc_environ_operations = { | |
921 | .read = environ_read, | |
87df8424 | 922 | .llseek = generic_file_llseek, |
315e28c8 JP |
923 | }; |
924 | ||
1da177e4 LT |
925 | static ssize_t oom_adjust_read(struct file *file, char __user *buf, |
926 | size_t count, loff_t *ppos) | |
927 | { | |
2fddfeef | 928 | struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode); |
8578cea7 | 929 | char buffer[PROC_NUMBUF]; |
1da177e4 | 930 | size_t len; |
28b83c51 KM |
931 | int oom_adjust = OOM_DISABLE; |
932 | unsigned long flags; | |
1da177e4 | 933 | |
99f89551 EB |
934 | if (!task) |
935 | return -ESRCH; | |
28b83c51 KM |
936 | |
937 | if (lock_task_sighand(task, &flags)) { | |
938 | oom_adjust = task->signal->oom_adj; | |
939 | unlock_task_sighand(task, &flags); | |
940 | } | |
941 | ||
99f89551 EB |
942 | put_task_struct(task); |
943 | ||
8578cea7 | 944 | len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust); |
0c28f287 AM |
945 | |
946 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
1da177e4 LT |
947 | } |
948 | ||
949 | static ssize_t oom_adjust_write(struct file *file, const char __user *buf, | |
950 | size_t count, loff_t *ppos) | |
951 | { | |
99f89551 | 952 | struct task_struct *task; |
5d863b89 | 953 | char buffer[PROC_NUMBUF]; |
0a8cb8e3 | 954 | int oom_adjust; |
28b83c51 | 955 | unsigned long flags; |
5d863b89 | 956 | int err; |
1da177e4 | 957 | |
8578cea7 EB |
958 | memset(buffer, 0, sizeof(buffer)); |
959 | if (count > sizeof(buffer) - 1) | |
960 | count = sizeof(buffer) - 1; | |
723548bf DR |
961 | if (copy_from_user(buffer, buf, count)) { |
962 | err = -EFAULT; | |
963 | goto out; | |
964 | } | |
5d863b89 | 965 | |
0a8cb8e3 | 966 | err = kstrtoint(strstrip(buffer), 0, &oom_adjust); |
5d863b89 | 967 | if (err) |
723548bf | 968 | goto out; |
8ac773b4 | 969 | if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) && |
723548bf DR |
970 | oom_adjust != OOM_DISABLE) { |
971 | err = -EINVAL; | |
972 | goto out; | |
973 | } | |
5d863b89 | 974 | |
2fddfeef | 975 | task = get_proc_task(file->f_path.dentry->d_inode); |
723548bf DR |
976 | if (!task) { |
977 | err = -ESRCH; | |
978 | goto out; | |
979 | } | |
d19d5476 DR |
980 | |
981 | task_lock(task); | |
982 | if (!task->mm) { | |
983 | err = -EINVAL; | |
984 | goto err_task_lock; | |
985 | } | |
986 | ||
28b83c51 | 987 | if (!lock_task_sighand(task, &flags)) { |
723548bf | 988 | err = -ESRCH; |
d19d5476 | 989 | goto err_task_lock; |
28b83c51 KM |
990 | } |
991 | ||
992 | if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) { | |
723548bf DR |
993 | err = -EACCES; |
994 | goto err_sighand; | |
8fb4fc68 | 995 | } |
28b83c51 | 996 | |
51b1bd2a DR |
997 | /* |
998 | * Warn that /proc/pid/oom_adj is deprecated, see | |
999 | * Documentation/feature-removal-schedule.txt. | |
1000 | */ | |
c2142704 | 1001 | printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n", |
be8f684d DR |
1002 | current->comm, task_pid_nr(current), task_pid_nr(task), |
1003 | task_pid_nr(task)); | |
28b83c51 | 1004 | task->signal->oom_adj = oom_adjust; |
a63d83f4 DR |
1005 | /* |
1006 | * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum | |
1007 | * value is always attainable. | |
1008 | */ | |
1009 | if (task->signal->oom_adj == OOM_ADJUST_MAX) | |
1010 | task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX; | |
1011 | else | |
1012 | task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) / | |
1013 | -OOM_DISABLE; | |
43d2b113 | 1014 | trace_oom_score_adj_update(task); |
723548bf | 1015 | err_sighand: |
28b83c51 | 1016 | unlock_task_sighand(task, &flags); |
d19d5476 DR |
1017 | err_task_lock: |
1018 | task_unlock(task); | |
99f89551 | 1019 | put_task_struct(task); |
723548bf DR |
1020 | out: |
1021 | return err < 0 ? err : count; | |
1da177e4 LT |
1022 | } |
1023 | ||
00977a59 | 1024 | static const struct file_operations proc_oom_adjust_operations = { |
1da177e4 LT |
1025 | .read = oom_adjust_read, |
1026 | .write = oom_adjust_write, | |
87df8424 | 1027 | .llseek = generic_file_llseek, |
1da177e4 LT |
1028 | }; |
1029 | ||
a63d83f4 DR |
1030 | static ssize_t oom_score_adj_read(struct file *file, char __user *buf, |
1031 | size_t count, loff_t *ppos) | |
1032 | { | |
1033 | struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode); | |
1034 | char buffer[PROC_NUMBUF]; | |
1035 | int oom_score_adj = OOM_SCORE_ADJ_MIN; | |
1036 | unsigned long flags; | |
1037 | size_t len; | |
1038 | ||
1039 | if (!task) | |
1040 | return -ESRCH; | |
1041 | if (lock_task_sighand(task, &flags)) { | |
1042 | oom_score_adj = task->signal->oom_score_adj; | |
1043 | unlock_task_sighand(task, &flags); | |
1044 | } | |
1045 | put_task_struct(task); | |
1046 | len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj); | |
1047 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
1048 | } | |
1049 | ||
1050 | static ssize_t oom_score_adj_write(struct file *file, const char __user *buf, | |
1051 | size_t count, loff_t *ppos) | |
1052 | { | |
1053 | struct task_struct *task; | |
1054 | char buffer[PROC_NUMBUF]; | |
1055 | unsigned long flags; | |
0a8cb8e3 | 1056 | int oom_score_adj; |
a63d83f4 DR |
1057 | int err; |
1058 | ||
1059 | memset(buffer, 0, sizeof(buffer)); | |
1060 | if (count > sizeof(buffer) - 1) | |
1061 | count = sizeof(buffer) - 1; | |
723548bf DR |
1062 | if (copy_from_user(buffer, buf, count)) { |
1063 | err = -EFAULT; | |
1064 | goto out; | |
1065 | } | |
a63d83f4 | 1066 | |
0a8cb8e3 | 1067 | err = kstrtoint(strstrip(buffer), 0, &oom_score_adj); |
a63d83f4 | 1068 | if (err) |
723548bf | 1069 | goto out; |
a63d83f4 | 1070 | if (oom_score_adj < OOM_SCORE_ADJ_MIN || |
723548bf DR |
1071 | oom_score_adj > OOM_SCORE_ADJ_MAX) { |
1072 | err = -EINVAL; | |
1073 | goto out; | |
1074 | } | |
a63d83f4 DR |
1075 | |
1076 | task = get_proc_task(file->f_path.dentry->d_inode); | |
723548bf DR |
1077 | if (!task) { |
1078 | err = -ESRCH; | |
1079 | goto out; | |
1080 | } | |
d19d5476 DR |
1081 | |
1082 | task_lock(task); | |
1083 | if (!task->mm) { | |
1084 | err = -EINVAL; | |
1085 | goto err_task_lock; | |
1086 | } | |
1087 | ||
a63d83f4 | 1088 | if (!lock_task_sighand(task, &flags)) { |
723548bf | 1089 | err = -ESRCH; |
d19d5476 | 1090 | goto err_task_lock; |
a63d83f4 | 1091 | } |
d19d5476 | 1092 | |
dabb16f6 | 1093 | if (oom_score_adj < task->signal->oom_score_adj_min && |
a63d83f4 | 1094 | !capable(CAP_SYS_RESOURCE)) { |
723548bf DR |
1095 | err = -EACCES; |
1096 | goto err_sighand; | |
a63d83f4 DR |
1097 | } |
1098 | ||
1099 | task->signal->oom_score_adj = oom_score_adj; | |
dabb16f6 MSB |
1100 | if (has_capability_noaudit(current, CAP_SYS_RESOURCE)) |
1101 | task->signal->oom_score_adj_min = oom_score_adj; | |
43d2b113 | 1102 | trace_oom_score_adj_update(task); |
a63d83f4 DR |
1103 | /* |
1104 | * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is | |
1105 | * always attainable. | |
1106 | */ | |
1107 | if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN) | |
1108 | task->signal->oom_adj = OOM_DISABLE; | |
1109 | else | |
1110 | task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) / | |
1111 | OOM_SCORE_ADJ_MAX; | |
723548bf | 1112 | err_sighand: |
a63d83f4 | 1113 | unlock_task_sighand(task, &flags); |
d19d5476 DR |
1114 | err_task_lock: |
1115 | task_unlock(task); | |
a63d83f4 | 1116 | put_task_struct(task); |
723548bf DR |
1117 | out: |
1118 | return err < 0 ? err : count; | |
a63d83f4 DR |
1119 | } |
1120 | ||
1121 | static const struct file_operations proc_oom_score_adj_operations = { | |
1122 | .read = oom_score_adj_read, | |
1123 | .write = oom_score_adj_write, | |
6038f373 | 1124 | .llseek = default_llseek, |
a63d83f4 DR |
1125 | }; |
1126 | ||
1da177e4 LT |
1127 | #ifdef CONFIG_AUDITSYSCALL |
1128 | #define TMPBUFLEN 21 | |
1129 | static ssize_t proc_loginuid_read(struct file * file, char __user * buf, | |
1130 | size_t count, loff_t *ppos) | |
1131 | { | |
2fddfeef | 1132 | struct inode * inode = file->f_path.dentry->d_inode; |
99f89551 | 1133 | struct task_struct *task = get_proc_task(inode); |
1da177e4 LT |
1134 | ssize_t length; |
1135 | char tmpbuf[TMPBUFLEN]; | |
1136 | ||
99f89551 EB |
1137 | if (!task) |
1138 | return -ESRCH; | |
1da177e4 | 1139 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", |
0c11b942 | 1140 | audit_get_loginuid(task)); |
99f89551 | 1141 | put_task_struct(task); |
1da177e4 LT |
1142 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); |
1143 | } | |
1144 | ||
1145 | static ssize_t proc_loginuid_write(struct file * file, const char __user * buf, | |
1146 | size_t count, loff_t *ppos) | |
1147 | { | |
2fddfeef | 1148 | struct inode * inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
1149 | char *page, *tmp; |
1150 | ssize_t length; | |
1da177e4 LT |
1151 | uid_t loginuid; |
1152 | ||
1153 | if (!capable(CAP_AUDIT_CONTROL)) | |
1154 | return -EPERM; | |
1155 | ||
7dc52157 PM |
1156 | rcu_read_lock(); |
1157 | if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) { | |
1158 | rcu_read_unlock(); | |
1da177e4 | 1159 | return -EPERM; |
7dc52157 PM |
1160 | } |
1161 | rcu_read_unlock(); | |
1da177e4 | 1162 | |
e0182909 AV |
1163 | if (count >= PAGE_SIZE) |
1164 | count = PAGE_SIZE - 1; | |
1da177e4 LT |
1165 | |
1166 | if (*ppos != 0) { | |
1167 | /* No partial writes. */ | |
1168 | return -EINVAL; | |
1169 | } | |
e12ba74d | 1170 | page = (char*)__get_free_page(GFP_TEMPORARY); |
1da177e4 LT |
1171 | if (!page) |
1172 | return -ENOMEM; | |
1173 | length = -EFAULT; | |
1174 | if (copy_from_user(page, buf, count)) | |
1175 | goto out_free_page; | |
1176 | ||
e0182909 | 1177 | page[count] = '\0'; |
1da177e4 LT |
1178 | loginuid = simple_strtoul(page, &tmp, 10); |
1179 | if (tmp == page) { | |
1180 | length = -EINVAL; | |
1181 | goto out_free_page; | |
1182 | ||
1183 | } | |
99f89551 | 1184 | length = audit_set_loginuid(current, loginuid); |
1da177e4 LT |
1185 | if (likely(length == 0)) |
1186 | length = count; | |
1187 | ||
1188 | out_free_page: | |
1189 | free_page((unsigned long) page); | |
1190 | return length; | |
1191 | } | |
1192 | ||
00977a59 | 1193 | static const struct file_operations proc_loginuid_operations = { |
1da177e4 LT |
1194 | .read = proc_loginuid_read, |
1195 | .write = proc_loginuid_write, | |
87df8424 | 1196 | .llseek = generic_file_llseek, |
1da177e4 | 1197 | }; |
1e0bd755 EP |
1198 | |
1199 | static ssize_t proc_sessionid_read(struct file * file, char __user * buf, | |
1200 | size_t count, loff_t *ppos) | |
1201 | { | |
1202 | struct inode * inode = file->f_path.dentry->d_inode; | |
1203 | struct task_struct *task = get_proc_task(inode); | |
1204 | ssize_t length; | |
1205 | char tmpbuf[TMPBUFLEN]; | |
1206 | ||
1207 | if (!task) | |
1208 | return -ESRCH; | |
1209 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", | |
1210 | audit_get_sessionid(task)); | |
1211 | put_task_struct(task); | |
1212 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); | |
1213 | } | |
1214 | ||
1215 | static const struct file_operations proc_sessionid_operations = { | |
1216 | .read = proc_sessionid_read, | |
87df8424 | 1217 | .llseek = generic_file_llseek, |
1e0bd755 | 1218 | }; |
1da177e4 LT |
1219 | #endif |
1220 | ||
f4f154fd AM |
1221 | #ifdef CONFIG_FAULT_INJECTION |
1222 | static ssize_t proc_fault_inject_read(struct file * file, char __user * buf, | |
1223 | size_t count, loff_t *ppos) | |
1224 | { | |
1225 | struct task_struct *task = get_proc_task(file->f_dentry->d_inode); | |
1226 | char buffer[PROC_NUMBUF]; | |
1227 | size_t len; | |
1228 | int make_it_fail; | |
f4f154fd AM |
1229 | |
1230 | if (!task) | |
1231 | return -ESRCH; | |
1232 | make_it_fail = task->make_it_fail; | |
1233 | put_task_struct(task); | |
1234 | ||
1235 | len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail); | |
0c28f287 AM |
1236 | |
1237 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
f4f154fd AM |
1238 | } |
1239 | ||
1240 | static ssize_t proc_fault_inject_write(struct file * file, | |
1241 | const char __user * buf, size_t count, loff_t *ppos) | |
1242 | { | |
1243 | struct task_struct *task; | |
1244 | char buffer[PROC_NUMBUF], *end; | |
1245 | int make_it_fail; | |
1246 | ||
1247 | if (!capable(CAP_SYS_RESOURCE)) | |
1248 | return -EPERM; | |
1249 | memset(buffer, 0, sizeof(buffer)); | |
1250 | if (count > sizeof(buffer) - 1) | |
1251 | count = sizeof(buffer) - 1; | |
1252 | if (copy_from_user(buffer, buf, count)) | |
1253 | return -EFAULT; | |
cba8aafe VL |
1254 | make_it_fail = simple_strtol(strstrip(buffer), &end, 0); |
1255 | if (*end) | |
1256 | return -EINVAL; | |
f4f154fd AM |
1257 | task = get_proc_task(file->f_dentry->d_inode); |
1258 | if (!task) | |
1259 | return -ESRCH; | |
1260 | task->make_it_fail = make_it_fail; | |
1261 | put_task_struct(task); | |
cba8aafe VL |
1262 | |
1263 | return count; | |
f4f154fd AM |
1264 | } |
1265 | ||
00977a59 | 1266 | static const struct file_operations proc_fault_inject_operations = { |
f4f154fd AM |
1267 | .read = proc_fault_inject_read, |
1268 | .write = proc_fault_inject_write, | |
87df8424 | 1269 | .llseek = generic_file_llseek, |
f4f154fd AM |
1270 | }; |
1271 | #endif | |
1272 | ||
9745512c | 1273 | |
43ae34cb IM |
1274 | #ifdef CONFIG_SCHED_DEBUG |
1275 | /* | |
1276 | * Print out various scheduling related per-task fields: | |
1277 | */ | |
1278 | static int sched_show(struct seq_file *m, void *v) | |
1279 | { | |
1280 | struct inode *inode = m->private; | |
1281 | struct task_struct *p; | |
1282 | ||
43ae34cb IM |
1283 | p = get_proc_task(inode); |
1284 | if (!p) | |
1285 | return -ESRCH; | |
1286 | proc_sched_show_task(p, m); | |
1287 | ||
1288 | put_task_struct(p); | |
1289 | ||
1290 | return 0; | |
1291 | } | |
1292 | ||
1293 | static ssize_t | |
1294 | sched_write(struct file *file, const char __user *buf, | |
1295 | size_t count, loff_t *offset) | |
1296 | { | |
1297 | struct inode *inode = file->f_path.dentry->d_inode; | |
1298 | struct task_struct *p; | |
1299 | ||
43ae34cb IM |
1300 | p = get_proc_task(inode); |
1301 | if (!p) | |
1302 | return -ESRCH; | |
1303 | proc_sched_set_task(p); | |
1304 | ||
1305 | put_task_struct(p); | |
1306 | ||
1307 | return count; | |
1308 | } | |
1309 | ||
1310 | static int sched_open(struct inode *inode, struct file *filp) | |
1311 | { | |
c6a34058 | 1312 | return single_open(filp, sched_show, inode); |
43ae34cb IM |
1313 | } |
1314 | ||
1315 | static const struct file_operations proc_pid_sched_operations = { | |
1316 | .open = sched_open, | |
1317 | .read = seq_read, | |
1318 | .write = sched_write, | |
1319 | .llseek = seq_lseek, | |
5ea473a1 | 1320 | .release = single_release, |
43ae34cb IM |
1321 | }; |
1322 | ||
1323 | #endif | |
1324 | ||
5091faa4 MG |
1325 | #ifdef CONFIG_SCHED_AUTOGROUP |
1326 | /* | |
1327 | * Print out autogroup related information: | |
1328 | */ | |
1329 | static int sched_autogroup_show(struct seq_file *m, void *v) | |
1330 | { | |
1331 | struct inode *inode = m->private; | |
1332 | struct task_struct *p; | |
1333 | ||
1334 | p = get_proc_task(inode); | |
1335 | if (!p) | |
1336 | return -ESRCH; | |
1337 | proc_sched_autogroup_show_task(p, m); | |
1338 | ||
1339 | put_task_struct(p); | |
1340 | ||
1341 | return 0; | |
1342 | } | |
1343 | ||
1344 | static ssize_t | |
1345 | sched_autogroup_write(struct file *file, const char __user *buf, | |
1346 | size_t count, loff_t *offset) | |
1347 | { | |
1348 | struct inode *inode = file->f_path.dentry->d_inode; | |
1349 | struct task_struct *p; | |
1350 | char buffer[PROC_NUMBUF]; | |
0a8cb8e3 | 1351 | int nice; |
5091faa4 MG |
1352 | int err; |
1353 | ||
1354 | memset(buffer, 0, sizeof(buffer)); | |
1355 | if (count > sizeof(buffer) - 1) | |
1356 | count = sizeof(buffer) - 1; | |
1357 | if (copy_from_user(buffer, buf, count)) | |
1358 | return -EFAULT; | |
1359 | ||
0a8cb8e3 AD |
1360 | err = kstrtoint(strstrip(buffer), 0, &nice); |
1361 | if (err < 0) | |
1362 | return err; | |
5091faa4 MG |
1363 | |
1364 | p = get_proc_task(inode); | |
1365 | if (!p) | |
1366 | return -ESRCH; | |
1367 | ||
1368 | err = nice; | |
1369 | err = proc_sched_autogroup_set_nice(p, &err); | |
1370 | if (err) | |
1371 | count = err; | |
1372 | ||
1373 | put_task_struct(p); | |
1374 | ||
1375 | return count; | |
1376 | } | |
1377 | ||
1378 | static int sched_autogroup_open(struct inode *inode, struct file *filp) | |
1379 | { | |
1380 | int ret; | |
1381 | ||
1382 | ret = single_open(filp, sched_autogroup_show, NULL); | |
1383 | if (!ret) { | |
1384 | struct seq_file *m = filp->private_data; | |
1385 | ||
1386 | m->private = inode; | |
1387 | } | |
1388 | return ret; | |
1389 | } | |
1390 | ||
1391 | static const struct file_operations proc_pid_sched_autogroup_operations = { | |
1392 | .open = sched_autogroup_open, | |
1393 | .read = seq_read, | |
1394 | .write = sched_autogroup_write, | |
1395 | .llseek = seq_lseek, | |
1396 | .release = single_release, | |
1397 | }; | |
1398 | ||
1399 | #endif /* CONFIG_SCHED_AUTOGROUP */ | |
1400 | ||
4614a696 JS |
1401 | static ssize_t comm_write(struct file *file, const char __user *buf, |
1402 | size_t count, loff_t *offset) | |
1403 | { | |
1404 | struct inode *inode = file->f_path.dentry->d_inode; | |
1405 | struct task_struct *p; | |
1406 | char buffer[TASK_COMM_LEN]; | |
1407 | ||
1408 | memset(buffer, 0, sizeof(buffer)); | |
1409 | if (count > sizeof(buffer) - 1) | |
1410 | count = sizeof(buffer) - 1; | |
1411 | if (copy_from_user(buffer, buf, count)) | |
1412 | return -EFAULT; | |
1413 | ||
1414 | p = get_proc_task(inode); | |
1415 | if (!p) | |
1416 | return -ESRCH; | |
1417 | ||
1418 | if (same_thread_group(current, p)) | |
1419 | set_task_comm(p, buffer); | |
1420 | else | |
1421 | count = -EINVAL; | |
1422 | ||
1423 | put_task_struct(p); | |
1424 | ||
1425 | return count; | |
1426 | } | |
1427 | ||
1428 | static int comm_show(struct seq_file *m, void *v) | |
1429 | { | |
1430 | struct inode *inode = m->private; | |
1431 | struct task_struct *p; | |
1432 | ||
1433 | p = get_proc_task(inode); | |
1434 | if (!p) | |
1435 | return -ESRCH; | |
1436 | ||
1437 | task_lock(p); | |
1438 | seq_printf(m, "%s\n", p->comm); | |
1439 | task_unlock(p); | |
1440 | ||
1441 | put_task_struct(p); | |
1442 | ||
1443 | return 0; | |
1444 | } | |
1445 | ||
1446 | static int comm_open(struct inode *inode, struct file *filp) | |
1447 | { | |
c6a34058 | 1448 | return single_open(filp, comm_show, inode); |
4614a696 JS |
1449 | } |
1450 | ||
1451 | static const struct file_operations proc_pid_set_comm_operations = { | |
1452 | .open = comm_open, | |
1453 | .read = seq_read, | |
1454 | .write = comm_write, | |
1455 | .llseek = seq_lseek, | |
1456 | .release = single_release, | |
1457 | }; | |
1458 | ||
925d1c40 MH |
1459 | static int proc_exe_link(struct inode *inode, struct path *exe_path) |
1460 | { | |
1461 | struct task_struct *task; | |
1462 | struct mm_struct *mm; | |
1463 | struct file *exe_file; | |
1464 | ||
1465 | task = get_proc_task(inode); | |
1466 | if (!task) | |
1467 | return -ENOENT; | |
1468 | mm = get_task_mm(task); | |
1469 | put_task_struct(task); | |
1470 | if (!mm) | |
1471 | return -ENOENT; | |
1472 | exe_file = get_mm_exe_file(mm); | |
1473 | mmput(mm); | |
1474 | if (exe_file) { | |
1475 | *exe_path = exe_file->f_path; | |
1476 | path_get(&exe_file->f_path); | |
1477 | fput(exe_file); | |
1478 | return 0; | |
1479 | } else | |
1480 | return -ENOENT; | |
1481 | } | |
1482 | ||
008b150a | 1483 | static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd) |
1da177e4 LT |
1484 | { |
1485 | struct inode *inode = dentry->d_inode; | |
1486 | int error = -EACCES; | |
1487 | ||
1488 | /* We don't need a base pointer in the /proc filesystem */ | |
1d957f9b | 1489 | path_put(&nd->path); |
1da177e4 | 1490 | |
778c1144 EB |
1491 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1492 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1493 | goto out; |
1da177e4 | 1494 | |
3dcd25f3 | 1495 | error = PROC_I(inode)->op.proc_get_link(inode, &nd->path); |
1da177e4 | 1496 | out: |
008b150a | 1497 | return ERR_PTR(error); |
1da177e4 LT |
1498 | } |
1499 | ||
3dcd25f3 | 1500 | static int do_proc_readlink(struct path *path, char __user *buffer, int buflen) |
1da177e4 | 1501 | { |
e12ba74d | 1502 | char *tmp = (char*)__get_free_page(GFP_TEMPORARY); |
3dcd25f3 | 1503 | char *pathname; |
1da177e4 LT |
1504 | int len; |
1505 | ||
1506 | if (!tmp) | |
1507 | return -ENOMEM; | |
0c28f287 | 1508 | |
7b2a69ba | 1509 | pathname = d_path(path, tmp, PAGE_SIZE); |
3dcd25f3 JB |
1510 | len = PTR_ERR(pathname); |
1511 | if (IS_ERR(pathname)) | |
1da177e4 | 1512 | goto out; |
3dcd25f3 | 1513 | len = tmp + PAGE_SIZE - 1 - pathname; |
1da177e4 LT |
1514 | |
1515 | if (len > buflen) | |
1516 | len = buflen; | |
3dcd25f3 | 1517 | if (copy_to_user(buffer, pathname, len)) |
1da177e4 LT |
1518 | len = -EFAULT; |
1519 | out: | |
1520 | free_page((unsigned long)tmp); | |
1521 | return len; | |
1522 | } | |
1523 | ||
1524 | static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen) | |
1525 | { | |
1526 | int error = -EACCES; | |
1527 | struct inode *inode = dentry->d_inode; | |
3dcd25f3 | 1528 | struct path path; |
1da177e4 | 1529 | |
778c1144 EB |
1530 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1531 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1532 | goto out; |
1da177e4 | 1533 | |
3dcd25f3 | 1534 | error = PROC_I(inode)->op.proc_get_link(inode, &path); |
1da177e4 LT |
1535 | if (error) |
1536 | goto out; | |
1537 | ||
3dcd25f3 JB |
1538 | error = do_proc_readlink(&path, buffer, buflen); |
1539 | path_put(&path); | |
1da177e4 | 1540 | out: |
1da177e4 LT |
1541 | return error; |
1542 | } | |
1543 | ||
c5ef1c42 | 1544 | static const struct inode_operations proc_pid_link_inode_operations = { |
1da177e4 | 1545 | .readlink = proc_pid_readlink, |
6d76fa58 LT |
1546 | .follow_link = proc_pid_follow_link, |
1547 | .setattr = proc_setattr, | |
1da177e4 LT |
1548 | }; |
1549 | ||
28a6d671 EB |
1550 | |
1551 | /* building an inode */ | |
1552 | ||
1553 | static int task_dumpable(struct task_struct *task) | |
1da177e4 | 1554 | { |
28a6d671 EB |
1555 | int dumpable = 0; |
1556 | struct mm_struct *mm; | |
1da177e4 | 1557 | |
28a6d671 EB |
1558 | task_lock(task); |
1559 | mm = task->mm; | |
1560 | if (mm) | |
6c5d5238 | 1561 | dumpable = get_dumpable(mm); |
28a6d671 EB |
1562 | task_unlock(task); |
1563 | if(dumpable == 1) | |
1564 | return 1; | |
1565 | return 0; | |
1566 | } | |
1da177e4 | 1567 | |
6b4e306a | 1568 | struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task) |
28a6d671 EB |
1569 | { |
1570 | struct inode * inode; | |
1571 | struct proc_inode *ei; | |
c69e8d9c | 1572 | const struct cred *cred; |
1da177e4 | 1573 | |
28a6d671 | 1574 | /* We need a new inode */ |
1da177e4 | 1575 | |
28a6d671 EB |
1576 | inode = new_inode(sb); |
1577 | if (!inode) | |
1578 | goto out; | |
1579 | ||
1580 | /* Common stuff */ | |
1581 | ei = PROC_I(inode); | |
85fe4025 | 1582 | inode->i_ino = get_next_ino(); |
28a6d671 | 1583 | inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; |
28a6d671 EB |
1584 | inode->i_op = &proc_def_inode_operations; |
1585 | ||
1586 | /* | |
1587 | * grab the reference to task. | |
1588 | */ | |
1a657f78 | 1589 | ei->pid = get_task_pid(task, PIDTYPE_PID); |
28a6d671 EB |
1590 | if (!ei->pid) |
1591 | goto out_unlock; | |
1592 | ||
28a6d671 | 1593 | if (task_dumpable(task)) { |
c69e8d9c DH |
1594 | rcu_read_lock(); |
1595 | cred = __task_cred(task); | |
1596 | inode->i_uid = cred->euid; | |
1597 | inode->i_gid = cred->egid; | |
1598 | rcu_read_unlock(); | |
1da177e4 | 1599 | } |
28a6d671 EB |
1600 | security_task_to_inode(task, inode); |
1601 | ||
1da177e4 | 1602 | out: |
28a6d671 EB |
1603 | return inode; |
1604 | ||
1605 | out_unlock: | |
1606 | iput(inode); | |
1607 | return NULL; | |
1da177e4 LT |
1608 | } |
1609 | ||
6b4e306a | 1610 | int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat) |
1da177e4 | 1611 | { |
1da177e4 | 1612 | struct inode *inode = dentry->d_inode; |
28a6d671 | 1613 | struct task_struct *task; |
c69e8d9c DH |
1614 | const struct cred *cred; |
1615 | ||
28a6d671 | 1616 | generic_fillattr(inode, stat); |
1da177e4 | 1617 | |
28a6d671 EB |
1618 | rcu_read_lock(); |
1619 | stat->uid = 0; | |
1620 | stat->gid = 0; | |
1621 | task = pid_task(proc_pid(inode), PIDTYPE_PID); | |
1622 | if (task) { | |
1623 | if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) || | |
1624 | task_dumpable(task)) { | |
c69e8d9c DH |
1625 | cred = __task_cred(task); |
1626 | stat->uid = cred->euid; | |
1627 | stat->gid = cred->egid; | |
1da177e4 LT |
1628 | } |
1629 | } | |
28a6d671 | 1630 | rcu_read_unlock(); |
d6e71144 | 1631 | return 0; |
1da177e4 LT |
1632 | } |
1633 | ||
1da177e4 LT |
1634 | /* dentry stuff */ |
1635 | ||
1636 | /* | |
1637 | * Exceptional case: normally we are not allowed to unhash a busy | |
1638 | * directory. In this case, however, we can do it - no aliasing problems | |
1639 | * due to the way we treat inodes. | |
1640 | * | |
1641 | * Rewrite the inode's ownerships here because the owning task may have | |
1642 | * performed a setuid(), etc. | |
99f89551 EB |
1643 | * |
1644 | * Before the /proc/pid/status file was created the only way to read | |
1645 | * the effective uid of a /process was to stat /proc/pid. Reading | |
1646 | * /proc/pid/status is slow enough that procps and other packages | |
1647 | * kept stating /proc/pid. To keep the rules in /proc simple I have | |
1648 | * made this apply to all per process world readable and executable | |
1649 | * directories. | |
1da177e4 | 1650 | */ |
6b4e306a | 1651 | int pid_revalidate(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 1652 | { |
34286d66 NP |
1653 | struct inode *inode; |
1654 | struct task_struct *task; | |
c69e8d9c DH |
1655 | const struct cred *cred; |
1656 | ||
34286d66 NP |
1657 | if (nd && nd->flags & LOOKUP_RCU) |
1658 | return -ECHILD; | |
1659 | ||
1660 | inode = dentry->d_inode; | |
1661 | task = get_proc_task(inode); | |
1662 | ||
99f89551 EB |
1663 | if (task) { |
1664 | if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) || | |
1665 | task_dumpable(task)) { | |
c69e8d9c DH |
1666 | rcu_read_lock(); |
1667 | cred = __task_cred(task); | |
1668 | inode->i_uid = cred->euid; | |
1669 | inode->i_gid = cred->egid; | |
1670 | rcu_read_unlock(); | |
1da177e4 LT |
1671 | } else { |
1672 | inode->i_uid = 0; | |
1673 | inode->i_gid = 0; | |
1674 | } | |
9ee8ab9f | 1675 | inode->i_mode &= ~(S_ISUID | S_ISGID); |
1da177e4 | 1676 | security_task_to_inode(task, inode); |
99f89551 | 1677 | put_task_struct(task); |
1da177e4 LT |
1678 | return 1; |
1679 | } | |
1680 | d_drop(dentry); | |
1681 | return 0; | |
1682 | } | |
1683 | ||
fe15ce44 | 1684 | static int pid_delete_dentry(const struct dentry * dentry) |
99f89551 | 1685 | { |
28a6d671 EB |
1686 | /* Is the task we represent dead? |
1687 | * If so, then don't put the dentry on the lru list, | |
1688 | * kill it immediately. | |
1689 | */ | |
1690 | return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first; | |
1691 | } | |
1692 | ||
6b4e306a | 1693 | const struct dentry_operations pid_dentry_operations = |
28a6d671 EB |
1694 | { |
1695 | .d_revalidate = pid_revalidate, | |
1696 | .d_delete = pid_delete_dentry, | |
1697 | }; | |
1698 | ||
1699 | /* Lookups */ | |
1700 | ||
1c0d04c9 EB |
1701 | /* |
1702 | * Fill a directory entry. | |
1703 | * | |
1704 | * If possible create the dcache entry and derive our inode number and | |
1705 | * file type from dcache entry. | |
1706 | * | |
1707 | * Since all of the proc inode numbers are dynamically generated, the inode | |
1708 | * numbers do not exist until the inode is cache. This means creating the | |
1709 | * the dcache entry in readdir is necessary to keep the inode numbers | |
1710 | * reported by readdir in sync with the inode numbers reported | |
1711 | * by stat. | |
1712 | */ | |
6b4e306a EB |
1713 | int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir, |
1714 | const char *name, int len, | |
c5141e6d | 1715 | instantiate_t instantiate, struct task_struct *task, const void *ptr) |
61a28784 | 1716 | { |
2fddfeef | 1717 | struct dentry *child, *dir = filp->f_path.dentry; |
61a28784 EB |
1718 | struct inode *inode; |
1719 | struct qstr qname; | |
1720 | ino_t ino = 0; | |
1721 | unsigned type = DT_UNKNOWN; | |
1722 | ||
1723 | qname.name = name; | |
1724 | qname.len = len; | |
1725 | qname.hash = full_name_hash(name, len); | |
1726 | ||
1727 | child = d_lookup(dir, &qname); | |
1728 | if (!child) { | |
1729 | struct dentry *new; | |
1730 | new = d_alloc(dir, &qname); | |
1731 | if (new) { | |
1732 | child = instantiate(dir->d_inode, new, task, ptr); | |
1733 | if (child) | |
1734 | dput(new); | |
1735 | else | |
1736 | child = new; | |
1737 | } | |
1738 | } | |
1739 | if (!child || IS_ERR(child) || !child->d_inode) | |
1740 | goto end_instantiate; | |
1741 | inode = child->d_inode; | |
1742 | if (inode) { | |
1743 | ino = inode->i_ino; | |
1744 | type = inode->i_mode >> 12; | |
1745 | } | |
1746 | dput(child); | |
1747 | end_instantiate: | |
1748 | if (!ino) | |
1749 | ino = find_inode_number(dir, &qname); | |
1750 | if (!ino) | |
1751 | ino = 1; | |
1752 | return filldir(dirent, name, len, filp->f_pos, ino, type); | |
1753 | } | |
1754 | ||
28a6d671 EB |
1755 | static unsigned name_to_int(struct dentry *dentry) |
1756 | { | |
1757 | const char *name = dentry->d_name.name; | |
1758 | int len = dentry->d_name.len; | |
1759 | unsigned n = 0; | |
1760 | ||
1761 | if (len > 1 && *name == '0') | |
1762 | goto out; | |
1763 | while (len-- > 0) { | |
1764 | unsigned c = *name++ - '0'; | |
1765 | if (c > 9) | |
1766 | goto out; | |
1767 | if (n >= (~0U-9)/10) | |
1768 | goto out; | |
1769 | n *= 10; | |
1770 | n += c; | |
1771 | } | |
1772 | return n; | |
1773 | out: | |
1774 | return ~0U; | |
1775 | } | |
1776 | ||
27932742 MS |
1777 | #define PROC_FDINFO_MAX 64 |
1778 | ||
3dcd25f3 | 1779 | static int proc_fd_info(struct inode *inode, struct path *path, char *info) |
28a6d671 | 1780 | { |
5e442a49 LT |
1781 | struct task_struct *task = get_proc_task(inode); |
1782 | struct files_struct *files = NULL; | |
28a6d671 EB |
1783 | struct file *file; |
1784 | int fd = proc_fd(inode); | |
aa6afca5 | 1785 | |
5e442a49 LT |
1786 | if (task) { |
1787 | files = get_files_struct(task); | |
1788 | put_task_struct(task); | |
1789 | } | |
1790 | if (files) { | |
1791 | /* | |
1792 | * We are not taking a ref to the file structure, so we must | |
1793 | * hold ->file_lock. | |
1794 | */ | |
1795 | spin_lock(&files->file_lock); | |
1796 | file = fcheck_files(files, fd); | |
1797 | if (file) { | |
1798 | unsigned int f_flags; | |
1799 | struct fdtable *fdt; | |
1800 | ||
1801 | fdt = files_fdtable(files); | |
1802 | f_flags = file->f_flags & ~O_CLOEXEC; | |
1803 | if (FD_ISSET(fd, fdt->close_on_exec)) | |
1804 | f_flags |= O_CLOEXEC; | |
1805 | ||
1806 | if (path) { | |
1807 | *path = file->f_path; | |
1808 | path_get(&file->f_path); | |
1809 | } | |
1810 | if (info) | |
1811 | snprintf(info, PROC_FDINFO_MAX, | |
1812 | "pos:\t%lli\n" | |
1813 | "flags:\t0%o\n", | |
1814 | (long long) file->f_pos, | |
1815 | f_flags); | |
1816 | spin_unlock(&files->file_lock); | |
1817 | put_files_struct(files); | |
1818 | return 0; | |
99f89551 | 1819 | } |
5e442a49 LT |
1820 | spin_unlock(&files->file_lock); |
1821 | put_files_struct(files); | |
1822 | } | |
1823 | return -ENOENT; | |
99f89551 EB |
1824 | } |
1825 | ||
3dcd25f3 | 1826 | static int proc_fd_link(struct inode *inode, struct path *path) |
27932742 | 1827 | { |
3dcd25f3 | 1828 | return proc_fd_info(inode, path, NULL); |
27932742 MS |
1829 | } |
1830 | ||
1da177e4 LT |
1831 | static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd) |
1832 | { | |
34286d66 NP |
1833 | struct inode *inode; |
1834 | struct task_struct *task; | |
1835 | int fd; | |
1da177e4 | 1836 | struct files_struct *files; |
c69e8d9c | 1837 | const struct cred *cred; |
1da177e4 | 1838 | |
34286d66 NP |
1839 | if (nd && nd->flags & LOOKUP_RCU) |
1840 | return -ECHILD; | |
1841 | ||
1842 | inode = dentry->d_inode; | |
1843 | task = get_proc_task(inode); | |
1844 | fd = proc_fd(inode); | |
1845 | ||
99f89551 EB |
1846 | if (task) { |
1847 | files = get_files_struct(task); | |
1848 | if (files) { | |
1849 | rcu_read_lock(); | |
1850 | if (fcheck_files(files, fd)) { | |
1851 | rcu_read_unlock(); | |
1852 | put_files_struct(files); | |
1853 | if (task_dumpable(task)) { | |
c69e8d9c DH |
1854 | rcu_read_lock(); |
1855 | cred = __task_cred(task); | |
1856 | inode->i_uid = cred->euid; | |
1857 | inode->i_gid = cred->egid; | |
1858 | rcu_read_unlock(); | |
99f89551 EB |
1859 | } else { |
1860 | inode->i_uid = 0; | |
1861 | inode->i_gid = 0; | |
1862 | } | |
9ee8ab9f | 1863 | inode->i_mode &= ~(S_ISUID | S_ISGID); |
99f89551 EB |
1864 | security_task_to_inode(task, inode); |
1865 | put_task_struct(task); | |
1866 | return 1; | |
1867 | } | |
b835996f | 1868 | rcu_read_unlock(); |
1da177e4 | 1869 | put_files_struct(files); |
1da177e4 | 1870 | } |
99f89551 | 1871 | put_task_struct(task); |
1da177e4 LT |
1872 | } |
1873 | d_drop(dentry); | |
1874 | return 0; | |
1875 | } | |
1876 | ||
d72f71eb | 1877 | static const struct dentry_operations tid_fd_dentry_operations = |
1da177e4 LT |
1878 | { |
1879 | .d_revalidate = tid_fd_revalidate, | |
1880 | .d_delete = pid_delete_dentry, | |
1881 | }; | |
1882 | ||
444ceed8 | 1883 | static struct dentry *proc_fd_instantiate(struct inode *dir, |
c5141e6d | 1884 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
1da177e4 | 1885 | { |
c5141e6d | 1886 | unsigned fd = *(const unsigned *)ptr; |
444ceed8 EB |
1887 | struct file *file; |
1888 | struct files_struct *files; | |
1889 | struct inode *inode; | |
1890 | struct proc_inode *ei; | |
1891 | struct dentry *error = ERR_PTR(-ENOENT); | |
1da177e4 | 1892 | |
61a28784 | 1893 | inode = proc_pid_make_inode(dir->i_sb, task); |
1da177e4 LT |
1894 | if (!inode) |
1895 | goto out; | |
1896 | ei = PROC_I(inode); | |
aed7a6c4 | 1897 | ei->fd = fd; |
1da177e4 LT |
1898 | files = get_files_struct(task); |
1899 | if (!files) | |
444ceed8 | 1900 | goto out_iput; |
1da177e4 | 1901 | inode->i_mode = S_IFLNK; |
ca99c1da DS |
1902 | |
1903 | /* | |
1904 | * We are not taking a ref to the file structure, so we must | |
1905 | * hold ->file_lock. | |
1906 | */ | |
1907 | spin_lock(&files->file_lock); | |
1da177e4 LT |
1908 | file = fcheck_files(files, fd); |
1909 | if (!file) | |
444ceed8 | 1910 | goto out_unlock; |
aeb5d727 | 1911 | if (file->f_mode & FMODE_READ) |
1da177e4 | 1912 | inode->i_mode |= S_IRUSR | S_IXUSR; |
aeb5d727 | 1913 | if (file->f_mode & FMODE_WRITE) |
1da177e4 | 1914 | inode->i_mode |= S_IWUSR | S_IXUSR; |
ca99c1da | 1915 | spin_unlock(&files->file_lock); |
1da177e4 | 1916 | put_files_struct(files); |
444ceed8 | 1917 | |
5e442a49 | 1918 | inode->i_op = &proc_pid_link_inode_operations; |
1da177e4 LT |
1919 | inode->i_size = 64; |
1920 | ei->op.proc_get_link = proc_fd_link; | |
fb045adb | 1921 | d_set_d_op(dentry, &tid_fd_dentry_operations); |
1da177e4 | 1922 | d_add(dentry, inode); |
cd6a3ce9 EB |
1923 | /* Close the race of the process dying before we return the dentry */ |
1924 | if (tid_fd_revalidate(dentry, NULL)) | |
444ceed8 | 1925 | error = NULL; |
1da177e4 | 1926 | |
444ceed8 EB |
1927 | out: |
1928 | return error; | |
1929 | out_unlock: | |
ca99c1da | 1930 | spin_unlock(&files->file_lock); |
1da177e4 | 1931 | put_files_struct(files); |
444ceed8 | 1932 | out_iput: |
1da177e4 | 1933 | iput(inode); |
cd6a3ce9 | 1934 | goto out; |
1da177e4 LT |
1935 | } |
1936 | ||
27932742 MS |
1937 | static struct dentry *proc_lookupfd_common(struct inode *dir, |
1938 | struct dentry *dentry, | |
1939 | instantiate_t instantiate) | |
444ceed8 EB |
1940 | { |
1941 | struct task_struct *task = get_proc_task(dir); | |
1942 | unsigned fd = name_to_int(dentry); | |
1943 | struct dentry *result = ERR_PTR(-ENOENT); | |
1944 | ||
1945 | if (!task) | |
1946 | goto out_no_task; | |
1947 | if (fd == ~0U) | |
1948 | goto out; | |
1949 | ||
27932742 | 1950 | result = instantiate(dir, dentry, task, &fd); |
444ceed8 EB |
1951 | out: |
1952 | put_task_struct(task); | |
1953 | out_no_task: | |
1954 | return result; | |
1955 | } | |
1956 | ||
27932742 MS |
1957 | static int proc_readfd_common(struct file * filp, void * dirent, |
1958 | filldir_t filldir, instantiate_t instantiate) | |
28a6d671 | 1959 | { |
2fddfeef | 1960 | struct dentry *dentry = filp->f_path.dentry; |
28a6d671 EB |
1961 | struct inode *inode = dentry->d_inode; |
1962 | struct task_struct *p = get_proc_task(inode); | |
457c2510 | 1963 | unsigned int fd, ino; |
28a6d671 | 1964 | int retval; |
28a6d671 | 1965 | struct files_struct * files; |
1da177e4 | 1966 | |
28a6d671 EB |
1967 | retval = -ENOENT; |
1968 | if (!p) | |
1969 | goto out_no_task; | |
1970 | retval = 0; | |
28a6d671 EB |
1971 | |
1972 | fd = filp->f_pos; | |
1973 | switch (fd) { | |
1974 | case 0: | |
1975 | if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0) | |
5e442a49 | 1976 | goto out; |
28a6d671 EB |
1977 | filp->f_pos++; |
1978 | case 1: | |
1979 | ino = parent_ino(dentry); | |
1980 | if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0) | |
5e442a49 | 1981 | goto out; |
28a6d671 EB |
1982 | filp->f_pos++; |
1983 | default: | |
1984 | files = get_files_struct(p); | |
1985 | if (!files) | |
5e442a49 | 1986 | goto out; |
28a6d671 | 1987 | rcu_read_lock(); |
28a6d671 | 1988 | for (fd = filp->f_pos-2; |
9b4f526c | 1989 | fd < files_fdtable(files)->max_fds; |
28a6d671 | 1990 | fd++, filp->f_pos++) { |
27932742 MS |
1991 | char name[PROC_NUMBUF]; |
1992 | int len; | |
28a6d671 EB |
1993 | |
1994 | if (!fcheck_files(files, fd)) | |
1995 | continue; | |
1996 | rcu_read_unlock(); | |
1997 | ||
27932742 MS |
1998 | len = snprintf(name, sizeof(name), "%d", fd); |
1999 | if (proc_fill_cache(filp, dirent, filldir, | |
2000 | name, len, instantiate, | |
2001 | p, &fd) < 0) { | |
28a6d671 EB |
2002 | rcu_read_lock(); |
2003 | break; | |
2004 | } | |
2005 | rcu_read_lock(); | |
2006 | } | |
2007 | rcu_read_unlock(); | |
2008 | put_files_struct(files); | |
2009 | } | |
2010 | out: | |
2011 | put_task_struct(p); | |
2012 | out_no_task: | |
2013 | return retval; | |
2014 | } | |
2015 | ||
27932742 MS |
2016 | static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry, |
2017 | struct nameidata *nd) | |
2018 | { | |
2019 | return proc_lookupfd_common(dir, dentry, proc_fd_instantiate); | |
2020 | } | |
2021 | ||
2022 | static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir) | |
2023 | { | |
2024 | return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate); | |
2025 | } | |
2026 | ||
2027 | static ssize_t proc_fdinfo_read(struct file *file, char __user *buf, | |
2028 | size_t len, loff_t *ppos) | |
2029 | { | |
2030 | char tmp[PROC_FDINFO_MAX]; | |
3dcd25f3 | 2031 | int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp); |
27932742 MS |
2032 | if (!err) |
2033 | err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp)); | |
2034 | return err; | |
2035 | } | |
2036 | ||
2037 | static const struct file_operations proc_fdinfo_file_operations = { | |
87df8424 | 2038 | .open = nonseekable_open, |
27932742 | 2039 | .read = proc_fdinfo_read, |
6038f373 | 2040 | .llseek = no_llseek, |
27932742 MS |
2041 | }; |
2042 | ||
00977a59 | 2043 | static const struct file_operations proc_fd_operations = { |
28a6d671 EB |
2044 | .read = generic_read_dir, |
2045 | .readdir = proc_readfd, | |
6038f373 | 2046 | .llseek = default_llseek, |
1da177e4 LT |
2047 | }; |
2048 | ||
8948e11f AD |
2049 | /* |
2050 | * /proc/pid/fd needs a special permission handler so that a process can still | |
2051 | * access /proc/self/fd after it has executed a setuid(). | |
2052 | */ | |
10556cb2 | 2053 | static int proc_fd_permission(struct inode *inode, int mask) |
8948e11f | 2054 | { |
2830ba7f | 2055 | int rv = generic_permission(inode, mask); |
8948e11f AD |
2056 | if (rv == 0) |
2057 | return 0; | |
2058 | if (task_pid(current) == proc_pid(inode)) | |
2059 | rv = 0; | |
2060 | return rv; | |
2061 | } | |
2062 | ||
1da177e4 LT |
2063 | /* |
2064 | * proc directories can do almost nothing.. | |
2065 | */ | |
c5ef1c42 | 2066 | static const struct inode_operations proc_fd_inode_operations = { |
1da177e4 | 2067 | .lookup = proc_lookupfd, |
8948e11f | 2068 | .permission = proc_fd_permission, |
6d76fa58 | 2069 | .setattr = proc_setattr, |
1da177e4 LT |
2070 | }; |
2071 | ||
27932742 MS |
2072 | static struct dentry *proc_fdinfo_instantiate(struct inode *dir, |
2073 | struct dentry *dentry, struct task_struct *task, const void *ptr) | |
2074 | { | |
2075 | unsigned fd = *(unsigned *)ptr; | |
2076 | struct inode *inode; | |
2077 | struct proc_inode *ei; | |
2078 | struct dentry *error = ERR_PTR(-ENOENT); | |
2079 | ||
2080 | inode = proc_pid_make_inode(dir->i_sb, task); | |
2081 | if (!inode) | |
2082 | goto out; | |
2083 | ei = PROC_I(inode); | |
2084 | ei->fd = fd; | |
2085 | inode->i_mode = S_IFREG | S_IRUSR; | |
2086 | inode->i_fop = &proc_fdinfo_file_operations; | |
fb045adb | 2087 | d_set_d_op(dentry, &tid_fd_dentry_operations); |
27932742 MS |
2088 | d_add(dentry, inode); |
2089 | /* Close the race of the process dying before we return the dentry */ | |
2090 | if (tid_fd_revalidate(dentry, NULL)) | |
2091 | error = NULL; | |
2092 | ||
2093 | out: | |
2094 | return error; | |
2095 | } | |
2096 | ||
2097 | static struct dentry *proc_lookupfdinfo(struct inode *dir, | |
2098 | struct dentry *dentry, | |
2099 | struct nameidata *nd) | |
2100 | { | |
2101 | return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate); | |
2102 | } | |
2103 | ||
2104 | static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir) | |
2105 | { | |
2106 | return proc_readfd_common(filp, dirent, filldir, | |
2107 | proc_fdinfo_instantiate); | |
2108 | } | |
2109 | ||
2110 | static const struct file_operations proc_fdinfo_operations = { | |
2111 | .read = generic_read_dir, | |
2112 | .readdir = proc_readfdinfo, | |
6038f373 | 2113 | .llseek = default_llseek, |
27932742 MS |
2114 | }; |
2115 | ||
2116 | /* | |
2117 | * proc directories can do almost nothing.. | |
2118 | */ | |
2119 | static const struct inode_operations proc_fdinfo_inode_operations = { | |
2120 | .lookup = proc_lookupfdinfo, | |
2121 | .setattr = proc_setattr, | |
2122 | }; | |
2123 | ||
2124 | ||
444ceed8 | 2125 | static struct dentry *proc_pident_instantiate(struct inode *dir, |
c5141e6d | 2126 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
444ceed8 | 2127 | { |
c5141e6d | 2128 | const struct pid_entry *p = ptr; |
444ceed8 EB |
2129 | struct inode *inode; |
2130 | struct proc_inode *ei; | |
bd6daba9 | 2131 | struct dentry *error = ERR_PTR(-ENOENT); |
444ceed8 | 2132 | |
61a28784 | 2133 | inode = proc_pid_make_inode(dir->i_sb, task); |
444ceed8 EB |
2134 | if (!inode) |
2135 | goto out; | |
2136 | ||
2137 | ei = PROC_I(inode); | |
2138 | inode->i_mode = p->mode; | |
2139 | if (S_ISDIR(inode->i_mode)) | |
bfe86848 | 2140 | set_nlink(inode, 2); /* Use getattr to fix if necessary */ |
444ceed8 EB |
2141 | if (p->iop) |
2142 | inode->i_op = p->iop; | |
2143 | if (p->fop) | |
2144 | inode->i_fop = p->fop; | |
2145 | ei->op = p->op; | |
fb045adb | 2146 | d_set_d_op(dentry, &pid_dentry_operations); |
444ceed8 EB |
2147 | d_add(dentry, inode); |
2148 | /* Close the race of the process dying before we return the dentry */ | |
2149 | if (pid_revalidate(dentry, NULL)) | |
2150 | error = NULL; | |
2151 | out: | |
2152 | return error; | |
2153 | } | |
2154 | ||
1da177e4 LT |
2155 | static struct dentry *proc_pident_lookup(struct inode *dir, |
2156 | struct dentry *dentry, | |
c5141e6d | 2157 | const struct pid_entry *ents, |
7bcd6b0e | 2158 | unsigned int nents) |
1da177e4 | 2159 | { |
cd6a3ce9 | 2160 | struct dentry *error; |
99f89551 | 2161 | struct task_struct *task = get_proc_task(dir); |
c5141e6d | 2162 | const struct pid_entry *p, *last; |
1da177e4 | 2163 | |
cd6a3ce9 | 2164 | error = ERR_PTR(-ENOENT); |
1da177e4 | 2165 | |
99f89551 EB |
2166 | if (!task) |
2167 | goto out_no_task; | |
1da177e4 | 2168 | |
20cdc894 EB |
2169 | /* |
2170 | * Yes, it does not scale. And it should not. Don't add | |
2171 | * new entries into /proc/<tgid>/ without very good reasons. | |
2172 | */ | |
7bcd6b0e EB |
2173 | last = &ents[nents - 1]; |
2174 | for (p = ents; p <= last; p++) { | |
1da177e4 LT |
2175 | if (p->len != dentry->d_name.len) |
2176 | continue; | |
2177 | if (!memcmp(dentry->d_name.name, p->name, p->len)) | |
2178 | break; | |
2179 | } | |
7bcd6b0e | 2180 | if (p > last) |
1da177e4 LT |
2181 | goto out; |
2182 | ||
444ceed8 | 2183 | error = proc_pident_instantiate(dir, dentry, task, p); |
1da177e4 | 2184 | out: |
99f89551 EB |
2185 | put_task_struct(task); |
2186 | out_no_task: | |
cd6a3ce9 | 2187 | return error; |
1da177e4 LT |
2188 | } |
2189 | ||
c5141e6d ED |
2190 | static int proc_pident_fill_cache(struct file *filp, void *dirent, |
2191 | filldir_t filldir, struct task_struct *task, const struct pid_entry *p) | |
61a28784 EB |
2192 | { |
2193 | return proc_fill_cache(filp, dirent, filldir, p->name, p->len, | |
2194 | proc_pident_instantiate, task, p); | |
2195 | } | |
2196 | ||
28a6d671 EB |
2197 | static int proc_pident_readdir(struct file *filp, |
2198 | void *dirent, filldir_t filldir, | |
c5141e6d | 2199 | const struct pid_entry *ents, unsigned int nents) |
28a6d671 EB |
2200 | { |
2201 | int i; | |
2fddfeef | 2202 | struct dentry *dentry = filp->f_path.dentry; |
28a6d671 EB |
2203 | struct inode *inode = dentry->d_inode; |
2204 | struct task_struct *task = get_proc_task(inode); | |
c5141e6d | 2205 | const struct pid_entry *p, *last; |
28a6d671 EB |
2206 | ino_t ino; |
2207 | int ret; | |
2208 | ||
2209 | ret = -ENOENT; | |
2210 | if (!task) | |
61a28784 | 2211 | goto out_no_task; |
28a6d671 EB |
2212 | |
2213 | ret = 0; | |
28a6d671 EB |
2214 | i = filp->f_pos; |
2215 | switch (i) { | |
2216 | case 0: | |
2217 | ino = inode->i_ino; | |
2218 | if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0) | |
2219 | goto out; | |
2220 | i++; | |
2221 | filp->f_pos++; | |
2222 | /* fall through */ | |
2223 | case 1: | |
2224 | ino = parent_ino(dentry); | |
2225 | if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0) | |
2226 | goto out; | |
2227 | i++; | |
2228 | filp->f_pos++; | |
2229 | /* fall through */ | |
2230 | default: | |
2231 | i -= 2; | |
2232 | if (i >= nents) { | |
2233 | ret = 1; | |
2234 | goto out; | |
2235 | } | |
2236 | p = ents + i; | |
7bcd6b0e EB |
2237 | last = &ents[nents - 1]; |
2238 | while (p <= last) { | |
61a28784 | 2239 | if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0) |
28a6d671 EB |
2240 | goto out; |
2241 | filp->f_pos++; | |
2242 | p++; | |
2243 | } | |
2244 | } | |
2245 | ||
2246 | ret = 1; | |
2247 | out: | |
61a28784 EB |
2248 | put_task_struct(task); |
2249 | out_no_task: | |
28a6d671 | 2250 | return ret; |
1da177e4 LT |
2251 | } |
2252 | ||
28a6d671 EB |
2253 | #ifdef CONFIG_SECURITY |
2254 | static ssize_t proc_pid_attr_read(struct file * file, char __user * buf, | |
2255 | size_t count, loff_t *ppos) | |
2256 | { | |
2fddfeef | 2257 | struct inode * inode = file->f_path.dentry->d_inode; |
04ff9708 | 2258 | char *p = NULL; |
28a6d671 EB |
2259 | ssize_t length; |
2260 | struct task_struct *task = get_proc_task(inode); | |
2261 | ||
28a6d671 | 2262 | if (!task) |
04ff9708 | 2263 | return -ESRCH; |
28a6d671 EB |
2264 | |
2265 | length = security_getprocattr(task, | |
2fddfeef | 2266 | (char*)file->f_path.dentry->d_name.name, |
04ff9708 | 2267 | &p); |
28a6d671 | 2268 | put_task_struct(task); |
04ff9708 AV |
2269 | if (length > 0) |
2270 | length = simple_read_from_buffer(buf, count, ppos, p, length); | |
2271 | kfree(p); | |
28a6d671 | 2272 | return length; |
1da177e4 LT |
2273 | } |
2274 | ||
28a6d671 EB |
2275 | static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf, |
2276 | size_t count, loff_t *ppos) | |
2277 | { | |
2fddfeef | 2278 | struct inode * inode = file->f_path.dentry->d_inode; |
28a6d671 EB |
2279 | char *page; |
2280 | ssize_t length; | |
2281 | struct task_struct *task = get_proc_task(inode); | |
2282 | ||
2283 | length = -ESRCH; | |
2284 | if (!task) | |
2285 | goto out_no_task; | |
2286 | if (count > PAGE_SIZE) | |
2287 | count = PAGE_SIZE; | |
2288 | ||
2289 | /* No partial writes. */ | |
2290 | length = -EINVAL; | |
2291 | if (*ppos != 0) | |
2292 | goto out; | |
2293 | ||
2294 | length = -ENOMEM; | |
e12ba74d | 2295 | page = (char*)__get_free_page(GFP_TEMPORARY); |
28a6d671 EB |
2296 | if (!page) |
2297 | goto out; | |
2298 | ||
2299 | length = -EFAULT; | |
2300 | if (copy_from_user(page, buf, count)) | |
2301 | goto out_free; | |
2302 | ||
107db7c7 | 2303 | /* Guard against adverse ptrace interaction */ |
9b1bf12d | 2304 | length = mutex_lock_interruptible(&task->signal->cred_guard_mutex); |
107db7c7 DH |
2305 | if (length < 0) |
2306 | goto out_free; | |
2307 | ||
28a6d671 | 2308 | length = security_setprocattr(task, |
2fddfeef | 2309 | (char*)file->f_path.dentry->d_name.name, |
28a6d671 | 2310 | (void*)page, count); |
9b1bf12d | 2311 | mutex_unlock(&task->signal->cred_guard_mutex); |
28a6d671 EB |
2312 | out_free: |
2313 | free_page((unsigned long) page); | |
2314 | out: | |
2315 | put_task_struct(task); | |
2316 | out_no_task: | |
2317 | return length; | |
2318 | } | |
2319 | ||
00977a59 | 2320 | static const struct file_operations proc_pid_attr_operations = { |
28a6d671 EB |
2321 | .read = proc_pid_attr_read, |
2322 | .write = proc_pid_attr_write, | |
87df8424 | 2323 | .llseek = generic_file_llseek, |
28a6d671 EB |
2324 | }; |
2325 | ||
c5141e6d | 2326 | static const struct pid_entry attr_dir_stuff[] = { |
631f9c18 AD |
2327 | REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations), |
2328 | REG("prev", S_IRUGO, proc_pid_attr_operations), | |
2329 | REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2330 | REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2331 | REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2332 | REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
28a6d671 EB |
2333 | }; |
2334 | ||
72d9dcfc | 2335 | static int proc_attr_dir_readdir(struct file * filp, |
28a6d671 EB |
2336 | void * dirent, filldir_t filldir) |
2337 | { | |
2338 | return proc_pident_readdir(filp,dirent,filldir, | |
72d9dcfc | 2339 | attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff)); |
28a6d671 EB |
2340 | } |
2341 | ||
00977a59 | 2342 | static const struct file_operations proc_attr_dir_operations = { |
1da177e4 | 2343 | .read = generic_read_dir, |
72d9dcfc | 2344 | .readdir = proc_attr_dir_readdir, |
6038f373 | 2345 | .llseek = default_llseek, |
1da177e4 LT |
2346 | }; |
2347 | ||
72d9dcfc | 2348 | static struct dentry *proc_attr_dir_lookup(struct inode *dir, |
28a6d671 EB |
2349 | struct dentry *dentry, struct nameidata *nd) |
2350 | { | |
7bcd6b0e EB |
2351 | return proc_pident_lookup(dir, dentry, |
2352 | attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2353 | } |
2354 | ||
c5ef1c42 | 2355 | static const struct inode_operations proc_attr_dir_inode_operations = { |
72d9dcfc | 2356 | .lookup = proc_attr_dir_lookup, |
99f89551 | 2357 | .getattr = pid_getattr, |
6d76fa58 | 2358 | .setattr = proc_setattr, |
1da177e4 LT |
2359 | }; |
2360 | ||
28a6d671 EB |
2361 | #endif |
2362 | ||
698ba7b5 | 2363 | #ifdef CONFIG_ELF_CORE |
3cb4a0bb KH |
2364 | static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf, |
2365 | size_t count, loff_t *ppos) | |
2366 | { | |
2367 | struct task_struct *task = get_proc_task(file->f_dentry->d_inode); | |
2368 | struct mm_struct *mm; | |
2369 | char buffer[PROC_NUMBUF]; | |
2370 | size_t len; | |
2371 | int ret; | |
2372 | ||
2373 | if (!task) | |
2374 | return -ESRCH; | |
2375 | ||
2376 | ret = 0; | |
2377 | mm = get_task_mm(task); | |
2378 | if (mm) { | |
2379 | len = snprintf(buffer, sizeof(buffer), "%08lx\n", | |
2380 | ((mm->flags & MMF_DUMP_FILTER_MASK) >> | |
2381 | MMF_DUMP_FILTER_SHIFT)); | |
2382 | mmput(mm); | |
2383 | ret = simple_read_from_buffer(buf, count, ppos, buffer, len); | |
2384 | } | |
2385 | ||
2386 | put_task_struct(task); | |
2387 | ||
2388 | return ret; | |
2389 | } | |
2390 | ||
2391 | static ssize_t proc_coredump_filter_write(struct file *file, | |
2392 | const char __user *buf, | |
2393 | size_t count, | |
2394 | loff_t *ppos) | |
2395 | { | |
2396 | struct task_struct *task; | |
2397 | struct mm_struct *mm; | |
2398 | char buffer[PROC_NUMBUF], *end; | |
2399 | unsigned int val; | |
2400 | int ret; | |
2401 | int i; | |
2402 | unsigned long mask; | |
2403 | ||
2404 | ret = -EFAULT; | |
2405 | memset(buffer, 0, sizeof(buffer)); | |
2406 | if (count > sizeof(buffer) - 1) | |
2407 | count = sizeof(buffer) - 1; | |
2408 | if (copy_from_user(buffer, buf, count)) | |
2409 | goto out_no_task; | |
2410 | ||
2411 | ret = -EINVAL; | |
2412 | val = (unsigned int)simple_strtoul(buffer, &end, 0); | |
2413 | if (*end == '\n') | |
2414 | end++; | |
2415 | if (end - buffer == 0) | |
2416 | goto out_no_task; | |
2417 | ||
2418 | ret = -ESRCH; | |
2419 | task = get_proc_task(file->f_dentry->d_inode); | |
2420 | if (!task) | |
2421 | goto out_no_task; | |
2422 | ||
2423 | ret = end - buffer; | |
2424 | mm = get_task_mm(task); | |
2425 | if (!mm) | |
2426 | goto out_no_mm; | |
2427 | ||
2428 | for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) { | |
2429 | if (val & mask) | |
2430 | set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2431 | else | |
2432 | clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2433 | } | |
2434 | ||
2435 | mmput(mm); | |
2436 | out_no_mm: | |
2437 | put_task_struct(task); | |
2438 | out_no_task: | |
2439 | return ret; | |
2440 | } | |
2441 | ||
2442 | static const struct file_operations proc_coredump_filter_operations = { | |
2443 | .read = proc_coredump_filter_read, | |
2444 | .write = proc_coredump_filter_write, | |
87df8424 | 2445 | .llseek = generic_file_llseek, |
3cb4a0bb KH |
2446 | }; |
2447 | #endif | |
2448 | ||
28a6d671 EB |
2449 | /* |
2450 | * /proc/self: | |
2451 | */ | |
2452 | static int proc_self_readlink(struct dentry *dentry, char __user *buffer, | |
2453 | int buflen) | |
2454 | { | |
488e5bc4 | 2455 | struct pid_namespace *ns = dentry->d_sb->s_fs_info; |
b55fcb22 | 2456 | pid_t tgid = task_tgid_nr_ns(current, ns); |
28a6d671 | 2457 | char tmp[PROC_NUMBUF]; |
b55fcb22 | 2458 | if (!tgid) |
488e5bc4 | 2459 | return -ENOENT; |
b55fcb22 | 2460 | sprintf(tmp, "%d", tgid); |
28a6d671 EB |
2461 | return vfs_readlink(dentry,buffer,buflen,tmp); |
2462 | } | |
2463 | ||
2464 | static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd) | |
2465 | { | |
488e5bc4 | 2466 | struct pid_namespace *ns = dentry->d_sb->s_fs_info; |
b55fcb22 | 2467 | pid_t tgid = task_tgid_nr_ns(current, ns); |
7fee4868 AV |
2468 | char *name = ERR_PTR(-ENOENT); |
2469 | if (tgid) { | |
2470 | name = __getname(); | |
2471 | if (!name) | |
2472 | name = ERR_PTR(-ENOMEM); | |
2473 | else | |
2474 | sprintf(name, "%d", tgid); | |
2475 | } | |
2476 | nd_set_link(nd, name); | |
2477 | return NULL; | |
2478 | } | |
2479 | ||
2480 | static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd, | |
2481 | void *cookie) | |
2482 | { | |
2483 | char *s = nd_get_link(nd); | |
2484 | if (!IS_ERR(s)) | |
2485 | __putname(s); | |
28a6d671 EB |
2486 | } |
2487 | ||
c5ef1c42 | 2488 | static const struct inode_operations proc_self_inode_operations = { |
28a6d671 EB |
2489 | .readlink = proc_self_readlink, |
2490 | .follow_link = proc_self_follow_link, | |
7fee4868 | 2491 | .put_link = proc_self_put_link, |
28a6d671 EB |
2492 | }; |
2493 | ||
801199ce EB |
2494 | /* |
2495 | * proc base | |
2496 | * | |
2497 | * These are the directory entries in the root directory of /proc | |
2498 | * that properly belong to the /proc filesystem, as they describe | |
2499 | * describe something that is process related. | |
2500 | */ | |
c5141e6d | 2501 | static const struct pid_entry proc_base_stuff[] = { |
61a28784 | 2502 | NOD("self", S_IFLNK|S_IRWXUGO, |
801199ce | 2503 | &proc_self_inode_operations, NULL, {}), |
801199ce EB |
2504 | }; |
2505 | ||
444ceed8 | 2506 | static struct dentry *proc_base_instantiate(struct inode *dir, |
c5141e6d | 2507 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
801199ce | 2508 | { |
c5141e6d | 2509 | const struct pid_entry *p = ptr; |
801199ce | 2510 | struct inode *inode; |
801199ce | 2511 | struct proc_inode *ei; |
73d36460 | 2512 | struct dentry *error; |
801199ce EB |
2513 | |
2514 | /* Allocate the inode */ | |
2515 | error = ERR_PTR(-ENOMEM); | |
2516 | inode = new_inode(dir->i_sb); | |
2517 | if (!inode) | |
2518 | goto out; | |
2519 | ||
2520 | /* Initialize the inode */ | |
2521 | ei = PROC_I(inode); | |
85fe4025 | 2522 | inode->i_ino = get_next_ino(); |
801199ce | 2523 | inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; |
801199ce EB |
2524 | |
2525 | /* | |
2526 | * grab the reference to the task. | |
2527 | */ | |
1a657f78 | 2528 | ei->pid = get_task_pid(task, PIDTYPE_PID); |
801199ce EB |
2529 | if (!ei->pid) |
2530 | goto out_iput; | |
2531 | ||
801199ce EB |
2532 | inode->i_mode = p->mode; |
2533 | if (S_ISDIR(inode->i_mode)) | |
bfe86848 | 2534 | set_nlink(inode, 2); |
801199ce EB |
2535 | if (S_ISLNK(inode->i_mode)) |
2536 | inode->i_size = 64; | |
2537 | if (p->iop) | |
2538 | inode->i_op = p->iop; | |
2539 | if (p->fop) | |
2540 | inode->i_fop = p->fop; | |
2541 | ei->op = p->op; | |
801199ce EB |
2542 | d_add(dentry, inode); |
2543 | error = NULL; | |
2544 | out: | |
801199ce EB |
2545 | return error; |
2546 | out_iput: | |
2547 | iput(inode); | |
2548 | goto out; | |
2549 | } | |
2550 | ||
444ceed8 EB |
2551 | static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry) |
2552 | { | |
2553 | struct dentry *error; | |
2554 | struct task_struct *task = get_proc_task(dir); | |
c5141e6d | 2555 | const struct pid_entry *p, *last; |
444ceed8 EB |
2556 | |
2557 | error = ERR_PTR(-ENOENT); | |
2558 | ||
2559 | if (!task) | |
2560 | goto out_no_task; | |
2561 | ||
2562 | /* Lookup the directory entry */ | |
7bcd6b0e EB |
2563 | last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1]; |
2564 | for (p = proc_base_stuff; p <= last; p++) { | |
444ceed8 EB |
2565 | if (p->len != dentry->d_name.len) |
2566 | continue; | |
2567 | if (!memcmp(dentry->d_name.name, p->name, p->len)) | |
2568 | break; | |
2569 | } | |
7bcd6b0e | 2570 | if (p > last) |
444ceed8 EB |
2571 | goto out; |
2572 | ||
2573 | error = proc_base_instantiate(dir, dentry, task, p); | |
2574 | ||
2575 | out: | |
2576 | put_task_struct(task); | |
2577 | out_no_task: | |
2578 | return error; | |
2579 | } | |
2580 | ||
c5141e6d ED |
2581 | static int proc_base_fill_cache(struct file *filp, void *dirent, |
2582 | filldir_t filldir, struct task_struct *task, const struct pid_entry *p) | |
61a28784 EB |
2583 | { |
2584 | return proc_fill_cache(filp, dirent, filldir, p->name, p->len, | |
2585 | proc_base_instantiate, task, p); | |
2586 | } | |
2587 | ||
aba76fdb | 2588 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
297c5d92 AR |
2589 | static int do_io_accounting(struct task_struct *task, char *buffer, int whole) |
2590 | { | |
940389b8 | 2591 | struct task_io_accounting acct = task->ioac; |
5995477a | 2592 | unsigned long flags; |
293eb1e7 | 2593 | int result; |
5995477a | 2594 | |
293eb1e7 VK |
2595 | result = mutex_lock_killable(&task->signal->cred_guard_mutex); |
2596 | if (result) | |
2597 | return result; | |
2598 | ||
2599 | if (!ptrace_may_access(task, PTRACE_MODE_READ)) { | |
2600 | result = -EACCES; | |
2601 | goto out_unlock; | |
2602 | } | |
1d1221f3 | 2603 | |
5995477a AR |
2604 | if (whole && lock_task_sighand(task, &flags)) { |
2605 | struct task_struct *t = task; | |
2606 | ||
2607 | task_io_accounting_add(&acct, &task->signal->ioac); | |
2608 | while_each_thread(task, t) | |
2609 | task_io_accounting_add(&acct, &t->ioac); | |
2610 | ||
2611 | unlock_task_sighand(task, &flags); | |
297c5d92 | 2612 | } |
293eb1e7 | 2613 | result = sprintf(buffer, |
aba76fdb AM |
2614 | "rchar: %llu\n" |
2615 | "wchar: %llu\n" | |
2616 | "syscr: %llu\n" | |
2617 | "syscw: %llu\n" | |
2618 | "read_bytes: %llu\n" | |
2619 | "write_bytes: %llu\n" | |
2620 | "cancelled_write_bytes: %llu\n", | |
7c44319d AB |
2621 | (unsigned long long)acct.rchar, |
2622 | (unsigned long long)acct.wchar, | |
2623 | (unsigned long long)acct.syscr, | |
2624 | (unsigned long long)acct.syscw, | |
2625 | (unsigned long long)acct.read_bytes, | |
2626 | (unsigned long long)acct.write_bytes, | |
2627 | (unsigned long long)acct.cancelled_write_bytes); | |
293eb1e7 VK |
2628 | out_unlock: |
2629 | mutex_unlock(&task->signal->cred_guard_mutex); | |
2630 | return result; | |
297c5d92 AR |
2631 | } |
2632 | ||
2633 | static int proc_tid_io_accounting(struct task_struct *task, char *buffer) | |
2634 | { | |
2635 | return do_io_accounting(task, buffer, 0); | |
aba76fdb | 2636 | } |
297c5d92 AR |
2637 | |
2638 | static int proc_tgid_io_accounting(struct task_struct *task, char *buffer) | |
2639 | { | |
2640 | return do_io_accounting(task, buffer, 1); | |
2641 | } | |
2642 | #endif /* CONFIG_TASK_IO_ACCOUNTING */ | |
aba76fdb | 2643 | |
47830723 KC |
2644 | static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns, |
2645 | struct pid *pid, struct task_struct *task) | |
2646 | { | |
a9712bc1 AV |
2647 | int err = lock_trace(task); |
2648 | if (!err) { | |
2649 | seq_printf(m, "%08x\n", task->personality); | |
2650 | unlock_trace(task); | |
2651 | } | |
2652 | return err; | |
47830723 KC |
2653 | } |
2654 | ||
28a6d671 EB |
2655 | /* |
2656 | * Thread groups | |
2657 | */ | |
00977a59 | 2658 | static const struct file_operations proc_task_operations; |
c5ef1c42 | 2659 | static const struct inode_operations proc_task_inode_operations; |
20cdc894 | 2660 | |
c5141e6d | 2661 | static const struct pid_entry tgid_base_stuff[] = { |
631f9c18 AD |
2662 | DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations), |
2663 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), | |
2664 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), | |
6b4e306a | 2665 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
b2211a36 | 2666 | #ifdef CONFIG_NET |
631f9c18 | 2667 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), |
b2211a36 | 2668 | #endif |
631f9c18 AD |
2669 | REG("environ", S_IRUSR, proc_environ_operations), |
2670 | INF("auxv", S_IRUSR, proc_pid_auxv), | |
2671 | ONE("status", S_IRUGO, proc_pid_status), | |
a9712bc1 | 2672 | ONE("personality", S_IRUGO, proc_pid_personality), |
3036e7b4 | 2673 | INF("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 2674 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 2675 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
5091faa4 MG |
2676 | #endif |
2677 | #ifdef CONFIG_SCHED_AUTOGROUP | |
2678 | REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations), | |
ebcb6734 | 2679 | #endif |
4614a696 | 2680 | REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations), |
ebcb6734 | 2681 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
a9712bc1 | 2682 | INF("syscall", S_IRUGO, proc_pid_syscall), |
43ae34cb | 2683 | #endif |
631f9c18 AD |
2684 | INF("cmdline", S_IRUGO, proc_pid_cmdline), |
2685 | ONE("stat", S_IRUGO, proc_tgid_stat), | |
2686 | ONE("statm", S_IRUGO, proc_pid_statm), | |
2687 | REG("maps", S_IRUGO, proc_maps_operations), | |
28a6d671 | 2688 | #ifdef CONFIG_NUMA |
631f9c18 | 2689 | REG("numa_maps", S_IRUGO, proc_numa_maps_operations), |
28a6d671 | 2690 | #endif |
631f9c18 AD |
2691 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
2692 | LNK("cwd", proc_cwd_link), | |
2693 | LNK("root", proc_root_link), | |
2694 | LNK("exe", proc_exe_link), | |
2695 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
2696 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
2697 | REG("mountstats", S_IRUSR, proc_mountstats_operations), | |
1e883281 | 2698 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 AD |
2699 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
2700 | REG("smaps", S_IRUGO, proc_smaps_operations), | |
ca6b0bf0 | 2701 | REG("pagemap", S_IRUGO, proc_pagemap_operations), |
28a6d671 EB |
2702 | #endif |
2703 | #ifdef CONFIG_SECURITY | |
631f9c18 | 2704 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
2705 | #endif |
2706 | #ifdef CONFIG_KALLSYMS | |
631f9c18 | 2707 | INF("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 2708 | #endif |
2ec220e2 | 2709 | #ifdef CONFIG_STACKTRACE |
a9712bc1 | 2710 | ONE("stack", S_IRUGO, proc_pid_stack), |
28a6d671 EB |
2711 | #endif |
2712 | #ifdef CONFIG_SCHEDSTATS | |
631f9c18 | 2713 | INF("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 2714 | #endif |
9745512c | 2715 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 2716 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 2717 | #endif |
8793d854 | 2718 | #ifdef CONFIG_PROC_PID_CPUSET |
631f9c18 | 2719 | REG("cpuset", S_IRUGO, proc_cpuset_operations), |
a424316c PM |
2720 | #endif |
2721 | #ifdef CONFIG_CGROUPS | |
631f9c18 | 2722 | REG("cgroup", S_IRUGO, proc_cgroup_operations), |
28a6d671 | 2723 | #endif |
631f9c18 AD |
2724 | INF("oom_score", S_IRUGO, proc_oom_score), |
2725 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations), | |
a63d83f4 | 2726 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
28a6d671 | 2727 | #ifdef CONFIG_AUDITSYSCALL |
631f9c18 AD |
2728 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
2729 | REG("sessionid", S_IRUGO, proc_sessionid_operations), | |
28a6d671 | 2730 | #endif |
f4f154fd | 2731 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 2732 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
f4f154fd | 2733 | #endif |
698ba7b5 | 2734 | #ifdef CONFIG_ELF_CORE |
631f9c18 | 2735 | REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations), |
3cb4a0bb | 2736 | #endif |
aba76fdb | 2737 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
1d1221f3 | 2738 | INF("io", S_IRUSR, proc_tgid_io_accounting), |
aba76fdb | 2739 | #endif |
f133ecca CM |
2740 | #ifdef CONFIG_HARDWALL |
2741 | INF("hardwall", S_IRUGO, proc_pid_hardwall), | |
2742 | #endif | |
28a6d671 | 2743 | }; |
1da177e4 | 2744 | |
28a6d671 | 2745 | static int proc_tgid_base_readdir(struct file * filp, |
1da177e4 LT |
2746 | void * dirent, filldir_t filldir) |
2747 | { | |
2748 | return proc_pident_readdir(filp,dirent,filldir, | |
28a6d671 | 2749 | tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff)); |
1da177e4 LT |
2750 | } |
2751 | ||
00977a59 | 2752 | static const struct file_operations proc_tgid_base_operations = { |
1da177e4 | 2753 | .read = generic_read_dir, |
28a6d671 | 2754 | .readdir = proc_tgid_base_readdir, |
6038f373 | 2755 | .llseek = default_llseek, |
1da177e4 LT |
2756 | }; |
2757 | ||
28a6d671 | 2758 | static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){ |
7bcd6b0e EB |
2759 | return proc_pident_lookup(dir, dentry, |
2760 | tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
2761 | } |
2762 | ||
c5ef1c42 | 2763 | static const struct inode_operations proc_tgid_base_inode_operations = { |
28a6d671 | 2764 | .lookup = proc_tgid_base_lookup, |
99f89551 | 2765 | .getattr = pid_getattr, |
6d76fa58 | 2766 | .setattr = proc_setattr, |
1da177e4 | 2767 | }; |
1da177e4 | 2768 | |
60347f67 | 2769 | static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid) |
1da177e4 | 2770 | { |
48e6484d | 2771 | struct dentry *dentry, *leader, *dir; |
8578cea7 | 2772 | char buf[PROC_NUMBUF]; |
48e6484d EB |
2773 | struct qstr name; |
2774 | ||
2775 | name.name = buf; | |
60347f67 PE |
2776 | name.len = snprintf(buf, sizeof(buf), "%d", pid); |
2777 | dentry = d_hash_and_lookup(mnt->mnt_root, &name); | |
48e6484d | 2778 | if (dentry) { |
29f12ca3 | 2779 | shrink_dcache_parent(dentry); |
48e6484d EB |
2780 | d_drop(dentry); |
2781 | dput(dentry); | |
2782 | } | |
1da177e4 | 2783 | |
48e6484d | 2784 | name.name = buf; |
60347f67 PE |
2785 | name.len = snprintf(buf, sizeof(buf), "%d", tgid); |
2786 | leader = d_hash_and_lookup(mnt->mnt_root, &name); | |
48e6484d EB |
2787 | if (!leader) |
2788 | goto out; | |
1da177e4 | 2789 | |
48e6484d EB |
2790 | name.name = "task"; |
2791 | name.len = strlen(name.name); | |
2792 | dir = d_hash_and_lookup(leader, &name); | |
2793 | if (!dir) | |
2794 | goto out_put_leader; | |
2795 | ||
2796 | name.name = buf; | |
60347f67 | 2797 | name.len = snprintf(buf, sizeof(buf), "%d", pid); |
48e6484d EB |
2798 | dentry = d_hash_and_lookup(dir, &name); |
2799 | if (dentry) { | |
2800 | shrink_dcache_parent(dentry); | |
2801 | d_drop(dentry); | |
2802 | dput(dentry); | |
1da177e4 | 2803 | } |
48e6484d EB |
2804 | |
2805 | dput(dir); | |
2806 | out_put_leader: | |
2807 | dput(leader); | |
2808 | out: | |
2809 | return; | |
1da177e4 LT |
2810 | } |
2811 | ||
0895e91d RD |
2812 | /** |
2813 | * proc_flush_task - Remove dcache entries for @task from the /proc dcache. | |
2814 | * @task: task that should be flushed. | |
2815 | * | |
2816 | * When flushing dentries from proc, one needs to flush them from global | |
60347f67 | 2817 | * proc (proc_mnt) and from all the namespaces' procs this task was seen |
0895e91d RD |
2818 | * in. This call is supposed to do all of this job. |
2819 | * | |
2820 | * Looks in the dcache for | |
2821 | * /proc/@pid | |
2822 | * /proc/@tgid/task/@pid | |
2823 | * if either directory is present flushes it and all of it'ts children | |
2824 | * from the dcache. | |
2825 | * | |
2826 | * It is safe and reasonable to cache /proc entries for a task until | |
2827 | * that task exits. After that they just clog up the dcache with | |
2828 | * useless entries, possibly causing useful dcache entries to be | |
2829 | * flushed instead. This routine is proved to flush those useless | |
2830 | * dcache entries at process exit time. | |
2831 | * | |
2832 | * NOTE: This routine is just an optimization so it does not guarantee | |
2833 | * that no dcache entries will exist at process exit time it | |
2834 | * just makes it very unlikely that any will persist. | |
60347f67 PE |
2835 | */ |
2836 | ||
2837 | void proc_flush_task(struct task_struct *task) | |
2838 | { | |
9fcc2d15 | 2839 | int i; |
9b4d1cbe | 2840 | struct pid *pid, *tgid; |
130f77ec PE |
2841 | struct upid *upid; |
2842 | ||
130f77ec | 2843 | pid = task_pid(task); |
9b4d1cbe | 2844 | tgid = task_tgid(task); |
130f77ec | 2845 | |
9fcc2d15 | 2846 | for (i = 0; i <= pid->level; i++) { |
130f77ec PE |
2847 | upid = &pid->numbers[i]; |
2848 | proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr, | |
9b4d1cbe | 2849 | tgid->numbers[i].nr); |
130f77ec | 2850 | } |
6f4e6433 PE |
2851 | |
2852 | upid = &pid->numbers[pid->level]; | |
2853 | if (upid->nr == 1) | |
2854 | pid_ns_release_proc(upid->ns); | |
60347f67 PE |
2855 | } |
2856 | ||
9711ef99 AB |
2857 | static struct dentry *proc_pid_instantiate(struct inode *dir, |
2858 | struct dentry * dentry, | |
c5141e6d | 2859 | struct task_struct *task, const void *ptr) |
444ceed8 EB |
2860 | { |
2861 | struct dentry *error = ERR_PTR(-ENOENT); | |
2862 | struct inode *inode; | |
2863 | ||
61a28784 | 2864 | inode = proc_pid_make_inode(dir->i_sb, task); |
444ceed8 EB |
2865 | if (!inode) |
2866 | goto out; | |
2867 | ||
2868 | inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO; | |
2869 | inode->i_op = &proc_tgid_base_inode_operations; | |
2870 | inode->i_fop = &proc_tgid_base_operations; | |
2871 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 | 2872 | |
bfe86848 MS |
2873 | set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff, |
2874 | ARRAY_SIZE(tgid_base_stuff))); | |
444ceed8 | 2875 | |
fb045adb | 2876 | d_set_d_op(dentry, &pid_dentry_operations); |
444ceed8 EB |
2877 | |
2878 | d_add(dentry, inode); | |
2879 | /* Close the race of the process dying before we return the dentry */ | |
2880 | if (pid_revalidate(dentry, NULL)) | |
2881 | error = NULL; | |
2882 | out: | |
2883 | return error; | |
2884 | } | |
2885 | ||
1da177e4 LT |
2886 | struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd) |
2887 | { | |
73d36460 | 2888 | struct dentry *result; |
1da177e4 | 2889 | struct task_struct *task; |
1da177e4 | 2890 | unsigned tgid; |
b488893a | 2891 | struct pid_namespace *ns; |
1da177e4 | 2892 | |
801199ce EB |
2893 | result = proc_base_lookup(dir, dentry); |
2894 | if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT) | |
2895 | goto out; | |
2896 | ||
1da177e4 LT |
2897 | tgid = name_to_int(dentry); |
2898 | if (tgid == ~0U) | |
2899 | goto out; | |
2900 | ||
b488893a | 2901 | ns = dentry->d_sb->s_fs_info; |
de758734 | 2902 | rcu_read_lock(); |
b488893a | 2903 | task = find_task_by_pid_ns(tgid, ns); |
1da177e4 LT |
2904 | if (task) |
2905 | get_task_struct(task); | |
de758734 | 2906 | rcu_read_unlock(); |
1da177e4 LT |
2907 | if (!task) |
2908 | goto out; | |
2909 | ||
444ceed8 | 2910 | result = proc_pid_instantiate(dir, dentry, task, NULL); |
1da177e4 | 2911 | put_task_struct(task); |
1da177e4 | 2912 | out: |
cd6a3ce9 | 2913 | return result; |
1da177e4 LT |
2914 | } |
2915 | ||
1da177e4 | 2916 | /* |
0804ef4b | 2917 | * Find the first task with tgid >= tgid |
0bc58a91 | 2918 | * |
1da177e4 | 2919 | */ |
19fd4bb2 EB |
2920 | struct tgid_iter { |
2921 | unsigned int tgid; | |
0804ef4b | 2922 | struct task_struct *task; |
19fd4bb2 EB |
2923 | }; |
2924 | static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter) | |
2925 | { | |
0804ef4b | 2926 | struct pid *pid; |
1da177e4 | 2927 | |
19fd4bb2 EB |
2928 | if (iter.task) |
2929 | put_task_struct(iter.task); | |
454cc105 | 2930 | rcu_read_lock(); |
0804ef4b | 2931 | retry: |
19fd4bb2 EB |
2932 | iter.task = NULL; |
2933 | pid = find_ge_pid(iter.tgid, ns); | |
0804ef4b | 2934 | if (pid) { |
19fd4bb2 EB |
2935 | iter.tgid = pid_nr_ns(pid, ns); |
2936 | iter.task = pid_task(pid, PIDTYPE_PID); | |
0804ef4b EB |
2937 | /* What we to know is if the pid we have find is the |
2938 | * pid of a thread_group_leader. Testing for task | |
2939 | * being a thread_group_leader is the obvious thing | |
2940 | * todo but there is a window when it fails, due to | |
2941 | * the pid transfer logic in de_thread. | |
2942 | * | |
2943 | * So we perform the straight forward test of seeing | |
2944 | * if the pid we have found is the pid of a thread | |
2945 | * group leader, and don't worry if the task we have | |
2946 | * found doesn't happen to be a thread group leader. | |
2947 | * As we don't care in the case of readdir. | |
2948 | */ | |
19fd4bb2 EB |
2949 | if (!iter.task || !has_group_leader_pid(iter.task)) { |
2950 | iter.tgid += 1; | |
0804ef4b | 2951 | goto retry; |
19fd4bb2 EB |
2952 | } |
2953 | get_task_struct(iter.task); | |
0bc58a91 | 2954 | } |
454cc105 | 2955 | rcu_read_unlock(); |
19fd4bb2 | 2956 | return iter; |
1da177e4 LT |
2957 | } |
2958 | ||
7bcd6b0e | 2959 | #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff)) |
0804ef4b | 2960 | |
61a28784 | 2961 | static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir, |
19fd4bb2 | 2962 | struct tgid_iter iter) |
61a28784 EB |
2963 | { |
2964 | char name[PROC_NUMBUF]; | |
19fd4bb2 | 2965 | int len = snprintf(name, sizeof(name), "%d", iter.tgid); |
61a28784 | 2966 | return proc_fill_cache(filp, dirent, filldir, name, len, |
19fd4bb2 | 2967 | proc_pid_instantiate, iter.task, NULL); |
61a28784 EB |
2968 | } |
2969 | ||
1da177e4 LT |
2970 | /* for the /proc/ directory itself, after non-process stuff has been done */ |
2971 | int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir) | |
2972 | { | |
d8bdc59f LT |
2973 | unsigned int nr; |
2974 | struct task_struct *reaper; | |
19fd4bb2 | 2975 | struct tgid_iter iter; |
b488893a | 2976 | struct pid_namespace *ns; |
1da177e4 | 2977 | |
d8bdc59f LT |
2978 | if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET) |
2979 | goto out_no_task; | |
2980 | nr = filp->f_pos - FIRST_PROCESS_ENTRY; | |
2981 | ||
2982 | reaper = get_proc_task(filp->f_path.dentry->d_inode); | |
61a28784 EB |
2983 | if (!reaper) |
2984 | goto out_no_task; | |
2985 | ||
7bcd6b0e | 2986 | for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) { |
c5141e6d | 2987 | const struct pid_entry *p = &proc_base_stuff[nr]; |
61a28784 | 2988 | if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0) |
801199ce | 2989 | goto out; |
1da177e4 LT |
2990 | } |
2991 | ||
b488893a | 2992 | ns = filp->f_dentry->d_sb->s_fs_info; |
19fd4bb2 EB |
2993 | iter.task = NULL; |
2994 | iter.tgid = filp->f_pos - TGID_OFFSET; | |
2995 | for (iter = next_tgid(ns, iter); | |
2996 | iter.task; | |
2997 | iter.tgid += 1, iter = next_tgid(ns, iter)) { | |
2998 | filp->f_pos = iter.tgid + TGID_OFFSET; | |
2999 | if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) { | |
3000 | put_task_struct(iter.task); | |
0804ef4b | 3001 | goto out; |
1da177e4 | 3002 | } |
0bc58a91 | 3003 | } |
0804ef4b EB |
3004 | filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET; |
3005 | out: | |
61a28784 EB |
3006 | put_task_struct(reaper); |
3007 | out_no_task: | |
0bc58a91 EB |
3008 | return 0; |
3009 | } | |
1da177e4 | 3010 | |
28a6d671 EB |
3011 | /* |
3012 | * Tasks | |
3013 | */ | |
c5141e6d | 3014 | static const struct pid_entry tid_base_stuff[] = { |
631f9c18 | 3015 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), |
3835541d | 3016 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 3017 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
631f9c18 AD |
3018 | REG("environ", S_IRUSR, proc_environ_operations), |
3019 | INF("auxv", S_IRUSR, proc_pid_auxv), | |
3020 | ONE("status", S_IRUGO, proc_pid_status), | |
a9712bc1 | 3021 | ONE("personality", S_IRUGO, proc_pid_personality), |
3036e7b4 | 3022 | INF("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 3023 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 3024 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
ebcb6734 | 3025 | #endif |
4614a696 | 3026 | REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations), |
ebcb6734 | 3027 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
a9712bc1 | 3028 | INF("syscall", S_IRUGO, proc_pid_syscall), |
43ae34cb | 3029 | #endif |
631f9c18 AD |
3030 | INF("cmdline", S_IRUGO, proc_pid_cmdline), |
3031 | ONE("stat", S_IRUGO, proc_tid_stat), | |
3032 | ONE("statm", S_IRUGO, proc_pid_statm), | |
3033 | REG("maps", S_IRUGO, proc_maps_operations), | |
28a6d671 | 3034 | #ifdef CONFIG_NUMA |
631f9c18 | 3035 | REG("numa_maps", S_IRUGO, proc_numa_maps_operations), |
28a6d671 | 3036 | #endif |
631f9c18 AD |
3037 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
3038 | LNK("cwd", proc_cwd_link), | |
3039 | LNK("root", proc_root_link), | |
3040 | LNK("exe", proc_exe_link), | |
3041 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
3042 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
1e883281 | 3043 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 AD |
3044 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
3045 | REG("smaps", S_IRUGO, proc_smaps_operations), | |
ca6b0bf0 | 3046 | REG("pagemap", S_IRUGO, proc_pagemap_operations), |
28a6d671 EB |
3047 | #endif |
3048 | #ifdef CONFIG_SECURITY | |
631f9c18 | 3049 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
3050 | #endif |
3051 | #ifdef CONFIG_KALLSYMS | |
631f9c18 | 3052 | INF("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 3053 | #endif |
2ec220e2 | 3054 | #ifdef CONFIG_STACKTRACE |
a9712bc1 | 3055 | ONE("stack", S_IRUGO, proc_pid_stack), |
28a6d671 EB |
3056 | #endif |
3057 | #ifdef CONFIG_SCHEDSTATS | |
631f9c18 | 3058 | INF("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 3059 | #endif |
9745512c | 3060 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 3061 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 3062 | #endif |
8793d854 | 3063 | #ifdef CONFIG_PROC_PID_CPUSET |
631f9c18 | 3064 | REG("cpuset", S_IRUGO, proc_cpuset_operations), |
a424316c PM |
3065 | #endif |
3066 | #ifdef CONFIG_CGROUPS | |
631f9c18 | 3067 | REG("cgroup", S_IRUGO, proc_cgroup_operations), |
28a6d671 | 3068 | #endif |
631f9c18 AD |
3069 | INF("oom_score", S_IRUGO, proc_oom_score), |
3070 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations), | |
a63d83f4 | 3071 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
28a6d671 | 3072 | #ifdef CONFIG_AUDITSYSCALL |
631f9c18 | 3073 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
26ec3c64 | 3074 | REG("sessionid", S_IRUGO, proc_sessionid_operations), |
28a6d671 | 3075 | #endif |
f4f154fd | 3076 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 3077 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
f4f154fd | 3078 | #endif |
297c5d92 | 3079 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
1d1221f3 | 3080 | INF("io", S_IRUSR, proc_tid_io_accounting), |
297c5d92 | 3081 | #endif |
f133ecca CM |
3082 | #ifdef CONFIG_HARDWALL |
3083 | INF("hardwall", S_IRUGO, proc_pid_hardwall), | |
3084 | #endif | |
28a6d671 EB |
3085 | }; |
3086 | ||
3087 | static int proc_tid_base_readdir(struct file * filp, | |
3088 | void * dirent, filldir_t filldir) | |
3089 | { | |
3090 | return proc_pident_readdir(filp,dirent,filldir, | |
3091 | tid_base_stuff,ARRAY_SIZE(tid_base_stuff)); | |
3092 | } | |
3093 | ||
3094 | static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){ | |
7bcd6b0e EB |
3095 | return proc_pident_lookup(dir, dentry, |
3096 | tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
3097 | } |
3098 | ||
00977a59 | 3099 | static const struct file_operations proc_tid_base_operations = { |
28a6d671 EB |
3100 | .read = generic_read_dir, |
3101 | .readdir = proc_tid_base_readdir, | |
6038f373 | 3102 | .llseek = default_llseek, |
28a6d671 EB |
3103 | }; |
3104 | ||
c5ef1c42 | 3105 | static const struct inode_operations proc_tid_base_inode_operations = { |
28a6d671 EB |
3106 | .lookup = proc_tid_base_lookup, |
3107 | .getattr = pid_getattr, | |
3108 | .setattr = proc_setattr, | |
3109 | }; | |
3110 | ||
444ceed8 | 3111 | static struct dentry *proc_task_instantiate(struct inode *dir, |
c5141e6d | 3112 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
444ceed8 EB |
3113 | { |
3114 | struct dentry *error = ERR_PTR(-ENOENT); | |
3115 | struct inode *inode; | |
61a28784 | 3116 | inode = proc_pid_make_inode(dir->i_sb, task); |
444ceed8 EB |
3117 | |
3118 | if (!inode) | |
3119 | goto out; | |
3120 | inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO; | |
3121 | inode->i_op = &proc_tid_base_inode_operations; | |
3122 | inode->i_fop = &proc_tid_base_operations; | |
3123 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 | 3124 | |
bfe86848 MS |
3125 | set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff, |
3126 | ARRAY_SIZE(tid_base_stuff))); | |
444ceed8 | 3127 | |
fb045adb | 3128 | d_set_d_op(dentry, &pid_dentry_operations); |
444ceed8 EB |
3129 | |
3130 | d_add(dentry, inode); | |
3131 | /* Close the race of the process dying before we return the dentry */ | |
3132 | if (pid_revalidate(dentry, NULL)) | |
3133 | error = NULL; | |
3134 | out: | |
3135 | return error; | |
3136 | } | |
3137 | ||
28a6d671 EB |
3138 | static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd) |
3139 | { | |
3140 | struct dentry *result = ERR_PTR(-ENOENT); | |
3141 | struct task_struct *task; | |
3142 | struct task_struct *leader = get_proc_task(dir); | |
28a6d671 | 3143 | unsigned tid; |
b488893a | 3144 | struct pid_namespace *ns; |
28a6d671 EB |
3145 | |
3146 | if (!leader) | |
3147 | goto out_no_task; | |
3148 | ||
3149 | tid = name_to_int(dentry); | |
3150 | if (tid == ~0U) | |
3151 | goto out; | |
3152 | ||
b488893a | 3153 | ns = dentry->d_sb->s_fs_info; |
28a6d671 | 3154 | rcu_read_lock(); |
b488893a | 3155 | task = find_task_by_pid_ns(tid, ns); |
28a6d671 EB |
3156 | if (task) |
3157 | get_task_struct(task); | |
3158 | rcu_read_unlock(); | |
3159 | if (!task) | |
3160 | goto out; | |
bac0abd6 | 3161 | if (!same_thread_group(leader, task)) |
28a6d671 EB |
3162 | goto out_drop_task; |
3163 | ||
444ceed8 | 3164 | result = proc_task_instantiate(dir, dentry, task, NULL); |
28a6d671 EB |
3165 | out_drop_task: |
3166 | put_task_struct(task); | |
3167 | out: | |
3168 | put_task_struct(leader); | |
3169 | out_no_task: | |
3170 | return result; | |
3171 | } | |
3172 | ||
0bc58a91 EB |
3173 | /* |
3174 | * Find the first tid of a thread group to return to user space. | |
3175 | * | |
3176 | * Usually this is just the thread group leader, but if the users | |
3177 | * buffer was too small or there was a seek into the middle of the | |
3178 | * directory we have more work todo. | |
3179 | * | |
3180 | * In the case of a short read we start with find_task_by_pid. | |
3181 | * | |
3182 | * In the case of a seek we start with the leader and walk nr | |
3183 | * threads past it. | |
3184 | */ | |
cc288738 | 3185 | static struct task_struct *first_tid(struct task_struct *leader, |
b488893a | 3186 | int tid, int nr, struct pid_namespace *ns) |
0bc58a91 | 3187 | { |
a872ff0c | 3188 | struct task_struct *pos; |
1da177e4 | 3189 | |
cc288738 | 3190 | rcu_read_lock(); |
0bc58a91 EB |
3191 | /* Attempt to start with the pid of a thread */ |
3192 | if (tid && (nr > 0)) { | |
b488893a | 3193 | pos = find_task_by_pid_ns(tid, ns); |
a872ff0c ON |
3194 | if (pos && (pos->group_leader == leader)) |
3195 | goto found; | |
0bc58a91 | 3196 | } |
1da177e4 | 3197 | |
0bc58a91 | 3198 | /* If nr exceeds the number of threads there is nothing todo */ |
a872ff0c ON |
3199 | pos = NULL; |
3200 | if (nr && nr >= get_nr_threads(leader)) | |
3201 | goto out; | |
1da177e4 | 3202 | |
a872ff0c ON |
3203 | /* If we haven't found our starting place yet start |
3204 | * with the leader and walk nr threads forward. | |
0bc58a91 | 3205 | */ |
a872ff0c ON |
3206 | for (pos = leader; nr > 0; --nr) { |
3207 | pos = next_thread(pos); | |
3208 | if (pos == leader) { | |
3209 | pos = NULL; | |
3210 | goto out; | |
3211 | } | |
1da177e4 | 3212 | } |
a872ff0c ON |
3213 | found: |
3214 | get_task_struct(pos); | |
3215 | out: | |
cc288738 | 3216 | rcu_read_unlock(); |
0bc58a91 EB |
3217 | return pos; |
3218 | } | |
3219 | ||
3220 | /* | |
3221 | * Find the next thread in the thread list. | |
3222 | * Return NULL if there is an error or no next thread. | |
3223 | * | |
3224 | * The reference to the input task_struct is released. | |
3225 | */ | |
3226 | static struct task_struct *next_tid(struct task_struct *start) | |
3227 | { | |
c1df7fb8 | 3228 | struct task_struct *pos = NULL; |
cc288738 | 3229 | rcu_read_lock(); |
c1df7fb8 | 3230 | if (pid_alive(start)) { |
0bc58a91 | 3231 | pos = next_thread(start); |
c1df7fb8 ON |
3232 | if (thread_group_leader(pos)) |
3233 | pos = NULL; | |
3234 | else | |
3235 | get_task_struct(pos); | |
3236 | } | |
cc288738 | 3237 | rcu_read_unlock(); |
0bc58a91 EB |
3238 | put_task_struct(start); |
3239 | return pos; | |
1da177e4 LT |
3240 | } |
3241 | ||
61a28784 EB |
3242 | static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir, |
3243 | struct task_struct *task, int tid) | |
3244 | { | |
3245 | char name[PROC_NUMBUF]; | |
3246 | int len = snprintf(name, sizeof(name), "%d", tid); | |
3247 | return proc_fill_cache(filp, dirent, filldir, name, len, | |
3248 | proc_task_instantiate, task, NULL); | |
3249 | } | |
3250 | ||
1da177e4 LT |
3251 | /* for the /proc/TGID/task/ directories */ |
3252 | static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir) | |
3253 | { | |
2fddfeef | 3254 | struct dentry *dentry = filp->f_path.dentry; |
1da177e4 | 3255 | struct inode *inode = dentry->d_inode; |
7d895244 | 3256 | struct task_struct *leader = NULL; |
0bc58a91 | 3257 | struct task_struct *task; |
1da177e4 LT |
3258 | int retval = -ENOENT; |
3259 | ino_t ino; | |
0bc58a91 | 3260 | int tid; |
b488893a | 3261 | struct pid_namespace *ns; |
1da177e4 | 3262 | |
7d895244 GC |
3263 | task = get_proc_task(inode); |
3264 | if (!task) | |
3265 | goto out_no_task; | |
3266 | rcu_read_lock(); | |
3267 | if (pid_alive(task)) { | |
3268 | leader = task->group_leader; | |
3269 | get_task_struct(leader); | |
3270 | } | |
3271 | rcu_read_unlock(); | |
3272 | put_task_struct(task); | |
99f89551 EB |
3273 | if (!leader) |
3274 | goto out_no_task; | |
1da177e4 LT |
3275 | retval = 0; |
3276 | ||
ee568b25 | 3277 | switch ((unsigned long)filp->f_pos) { |
1da177e4 LT |
3278 | case 0: |
3279 | ino = inode->i_ino; | |
ee6f779b | 3280 | if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0) |
1da177e4 | 3281 | goto out; |
ee6f779b | 3282 | filp->f_pos++; |
1da177e4 LT |
3283 | /* fall through */ |
3284 | case 1: | |
3285 | ino = parent_ino(dentry); | |
ee6f779b | 3286 | if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0) |
1da177e4 | 3287 | goto out; |
ee6f779b | 3288 | filp->f_pos++; |
1da177e4 LT |
3289 | /* fall through */ |
3290 | } | |
3291 | ||
0bc58a91 EB |
3292 | /* f_version caches the tgid value that the last readdir call couldn't |
3293 | * return. lseek aka telldir automagically resets f_version to 0. | |
3294 | */ | |
b488893a | 3295 | ns = filp->f_dentry->d_sb->s_fs_info; |
2b47c361 | 3296 | tid = (int)filp->f_version; |
0bc58a91 | 3297 | filp->f_version = 0; |
ee6f779b | 3298 | for (task = first_tid(leader, tid, filp->f_pos - 2, ns); |
0bc58a91 | 3299 | task; |
ee6f779b | 3300 | task = next_tid(task), filp->f_pos++) { |
b488893a | 3301 | tid = task_pid_nr_ns(task, ns); |
61a28784 | 3302 | if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) { |
0bc58a91 EB |
3303 | /* returning this tgid failed, save it as the first |
3304 | * pid for the next readir call */ | |
2b47c361 | 3305 | filp->f_version = (u64)tid; |
0bc58a91 | 3306 | put_task_struct(task); |
1da177e4 | 3307 | break; |
0bc58a91 | 3308 | } |
1da177e4 LT |
3309 | } |
3310 | out: | |
99f89551 EB |
3311 | put_task_struct(leader); |
3312 | out_no_task: | |
1da177e4 LT |
3313 | return retval; |
3314 | } | |
6e66b52b EB |
3315 | |
3316 | static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat) | |
3317 | { | |
3318 | struct inode *inode = dentry->d_inode; | |
99f89551 | 3319 | struct task_struct *p = get_proc_task(inode); |
6e66b52b EB |
3320 | generic_fillattr(inode, stat); |
3321 | ||
99f89551 | 3322 | if (p) { |
99f89551 | 3323 | stat->nlink += get_nr_threads(p); |
99f89551 | 3324 | put_task_struct(p); |
6e66b52b EB |
3325 | } |
3326 | ||
3327 | return 0; | |
3328 | } | |
28a6d671 | 3329 | |
c5ef1c42 | 3330 | static const struct inode_operations proc_task_inode_operations = { |
28a6d671 EB |
3331 | .lookup = proc_task_lookup, |
3332 | .getattr = proc_task_getattr, | |
3333 | .setattr = proc_setattr, | |
3334 | }; | |
3335 | ||
00977a59 | 3336 | static const struct file_operations proc_task_operations = { |
28a6d671 EB |
3337 | .read = generic_read_dir, |
3338 | .readdir = proc_task_readdir, | |
6038f373 | 3339 | .llseek = default_llseek, |
28a6d671 | 3340 | }; |