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> |
22d917d8 | 84 | #include <linux/user_namespace.h> |
5ad4e53b | 85 | #include <linux/fs_struct.h> |
5a0e3ad6 | 86 | #include <linux/slab.h> |
640708a2 | 87 | #include <linux/flex_array.h> |
f133ecca CM |
88 | #ifdef CONFIG_HARDWALL |
89 | #include <asm/hardwall.h> | |
90 | #endif | |
43d2b113 | 91 | #include <trace/events/oom.h> |
1da177e4 | 92 | #include "internal.h" |
faf60af1 | 93 | #include "fd.h" |
1da177e4 | 94 | |
0f2fe20f EB |
95 | /* NOTE: |
96 | * Implementing inode permission operations in /proc is almost | |
97 | * certainly an error. Permission checks need to happen during | |
98 | * each system call not at open time. The reason is that most of | |
99 | * what we wish to check for permissions in /proc varies at runtime. | |
100 | * | |
101 | * The classic example of a problem is opening file descriptors | |
102 | * in /proc for a task before it execs a suid executable. | |
103 | */ | |
104 | ||
1da177e4 | 105 | struct pid_entry { |
1da177e4 | 106 | char *name; |
c5141e6d | 107 | int len; |
d161a13f | 108 | umode_t mode; |
c5ef1c42 | 109 | const struct inode_operations *iop; |
00977a59 | 110 | const struct file_operations *fop; |
20cdc894 | 111 | union proc_op op; |
1da177e4 LT |
112 | }; |
113 | ||
61a28784 | 114 | #define NOD(NAME, MODE, IOP, FOP, OP) { \ |
20cdc894 | 115 | .name = (NAME), \ |
c5141e6d | 116 | .len = sizeof(NAME) - 1, \ |
20cdc894 EB |
117 | .mode = MODE, \ |
118 | .iop = IOP, \ | |
119 | .fop = FOP, \ | |
120 | .op = OP, \ | |
121 | } | |
122 | ||
631f9c18 AD |
123 | #define DIR(NAME, MODE, iops, fops) \ |
124 | NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} ) | |
125 | #define LNK(NAME, get_link) \ | |
61a28784 | 126 | NOD(NAME, (S_IFLNK|S_IRWXUGO), \ |
20cdc894 | 127 | &proc_pid_link_inode_operations, NULL, \ |
631f9c18 AD |
128 | { .proc_get_link = get_link } ) |
129 | #define REG(NAME, MODE, fops) \ | |
130 | NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {}) | |
131 | #define INF(NAME, MODE, read) \ | |
61a28784 | 132 | NOD(NAME, (S_IFREG|(MODE)), \ |
20cdc894 | 133 | NULL, &proc_info_file_operations, \ |
631f9c18 AD |
134 | { .proc_read = read } ) |
135 | #define ONE(NAME, MODE, show) \ | |
be614086 EB |
136 | NOD(NAME, (S_IFREG|(MODE)), \ |
137 | NULL, &proc_single_file_operations, \ | |
631f9c18 | 138 | { .proc_show = show } ) |
1da177e4 | 139 | |
aed54175 VN |
140 | /* |
141 | * Count the number of hardlinks for the pid_entry table, excluding the . | |
142 | * and .. links. | |
143 | */ | |
144 | static unsigned int pid_entry_count_dirs(const struct pid_entry *entries, | |
145 | unsigned int n) | |
146 | { | |
147 | unsigned int i; | |
148 | unsigned int count; | |
149 | ||
150 | count = 0; | |
151 | for (i = 0; i < n; ++i) { | |
152 | if (S_ISDIR(entries[i].mode)) | |
153 | ++count; | |
154 | } | |
155 | ||
156 | return count; | |
157 | } | |
158 | ||
f7ad3c6b | 159 | static int get_task_root(struct task_struct *task, struct path *root) |
1da177e4 | 160 | { |
7c2c7d99 HD |
161 | int result = -ENOENT; |
162 | ||
0494f6ec | 163 | task_lock(task); |
f7ad3c6b MS |
164 | if (task->fs) { |
165 | get_fs_root(task->fs, root); | |
7c2c7d99 HD |
166 | result = 0; |
167 | } | |
0494f6ec | 168 | task_unlock(task); |
7c2c7d99 | 169 | return result; |
0494f6ec MS |
170 | } |
171 | ||
7773fbc5 | 172 | static int proc_cwd_link(struct dentry *dentry, struct path *path) |
0494f6ec | 173 | { |
7773fbc5 | 174 | struct task_struct *task = get_proc_task(dentry->d_inode); |
0494f6ec | 175 | int result = -ENOENT; |
99f89551 EB |
176 | |
177 | if (task) { | |
f7ad3c6b MS |
178 | task_lock(task); |
179 | if (task->fs) { | |
180 | get_fs_pwd(task->fs, path); | |
181 | result = 0; | |
182 | } | |
183 | task_unlock(task); | |
99f89551 EB |
184 | put_task_struct(task); |
185 | } | |
1da177e4 LT |
186 | return result; |
187 | } | |
188 | ||
7773fbc5 | 189 | static int proc_root_link(struct dentry *dentry, struct path *path) |
1da177e4 | 190 | { |
7773fbc5 | 191 | struct task_struct *task = get_proc_task(dentry->d_inode); |
1da177e4 | 192 | int result = -ENOENT; |
99f89551 EB |
193 | |
194 | if (task) { | |
f7ad3c6b | 195 | result = get_task_root(task, path); |
99f89551 EB |
196 | put_task_struct(task); |
197 | } | |
1da177e4 LT |
198 | return result; |
199 | } | |
200 | ||
1da177e4 LT |
201 | static int proc_pid_cmdline(struct task_struct *task, char * buffer) |
202 | { | |
203 | int res = 0; | |
204 | unsigned int len; | |
205 | struct mm_struct *mm = get_task_mm(task); | |
206 | if (!mm) | |
207 | goto out; | |
208 | if (!mm->arg_end) | |
209 | goto out_mm; /* Shh! No looking before we're done */ | |
210 | ||
211 | len = mm->arg_end - mm->arg_start; | |
212 | ||
213 | if (len > PAGE_SIZE) | |
214 | len = PAGE_SIZE; | |
215 | ||
216 | res = access_process_vm(task, mm->arg_start, buffer, len, 0); | |
217 | ||
218 | // If the nul at the end of args has been overwritten, then | |
219 | // assume application is using setproctitle(3). | |
220 | if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) { | |
221 | len = strnlen(buffer, res); | |
222 | if (len < res) { | |
223 | res = len; | |
224 | } else { | |
225 | len = mm->env_end - mm->env_start; | |
226 | if (len > PAGE_SIZE - res) | |
227 | len = PAGE_SIZE - res; | |
228 | res += access_process_vm(task, mm->env_start, buffer+res, len, 0); | |
229 | res = strnlen(buffer, res); | |
230 | } | |
231 | } | |
232 | out_mm: | |
233 | mmput(mm); | |
234 | out: | |
235 | return res; | |
236 | } | |
237 | ||
238 | static int proc_pid_auxv(struct task_struct *task, char *buffer) | |
239 | { | |
e7dcd999 | 240 | struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ); |
2fadaef4 AV |
241 | int res = PTR_ERR(mm); |
242 | if (mm && !IS_ERR(mm)) { | |
1da177e4 | 243 | unsigned int nwords = 0; |
dfe6b7d9 | 244 | do { |
1da177e4 | 245 | nwords += 2; |
dfe6b7d9 | 246 | } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */ |
1da177e4 LT |
247 | res = nwords * sizeof(mm->saved_auxv[0]); |
248 | if (res > PAGE_SIZE) | |
249 | res = PAGE_SIZE; | |
250 | memcpy(buffer, mm->saved_auxv, res); | |
251 | mmput(mm); | |
252 | } | |
253 | return res; | |
254 | } | |
255 | ||
256 | ||
257 | #ifdef CONFIG_KALLSYMS | |
258 | /* | |
259 | * Provides a wchan file via kallsyms in a proper one-value-per-file format. | |
260 | * Returns the resolved symbol. If that fails, simply return the address. | |
261 | */ | |
262 | static int proc_pid_wchan(struct task_struct *task, char *buffer) | |
263 | { | |
ffb45122 | 264 | unsigned long wchan; |
9281acea | 265 | char symname[KSYM_NAME_LEN]; |
1da177e4 LT |
266 | |
267 | wchan = get_wchan(task); | |
268 | ||
9d65cb4a | 269 | if (lookup_symbol_name(wchan, symname) < 0) |
f83ce3e6 JE |
270 | if (!ptrace_may_access(task, PTRACE_MODE_READ)) |
271 | return 0; | |
272 | else | |
273 | return sprintf(buffer, "%lu", wchan); | |
9d65cb4a AD |
274 | else |
275 | return sprintf(buffer, "%s", symname); | |
1da177e4 LT |
276 | } |
277 | #endif /* CONFIG_KALLSYMS */ | |
278 | ||
a9712bc1 AV |
279 | static int lock_trace(struct task_struct *task) |
280 | { | |
281 | int err = mutex_lock_killable(&task->signal->cred_guard_mutex); | |
282 | if (err) | |
283 | return err; | |
284 | if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) { | |
285 | mutex_unlock(&task->signal->cred_guard_mutex); | |
286 | return -EPERM; | |
287 | } | |
288 | return 0; | |
289 | } | |
290 | ||
291 | static void unlock_trace(struct task_struct *task) | |
292 | { | |
293 | mutex_unlock(&task->signal->cred_guard_mutex); | |
294 | } | |
295 | ||
2ec220e2 KC |
296 | #ifdef CONFIG_STACKTRACE |
297 | ||
298 | #define MAX_STACK_TRACE_DEPTH 64 | |
299 | ||
300 | static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns, | |
301 | struct pid *pid, struct task_struct *task) | |
302 | { | |
303 | struct stack_trace trace; | |
304 | unsigned long *entries; | |
a9712bc1 | 305 | int err; |
2ec220e2 KC |
306 | int i; |
307 | ||
308 | entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL); | |
309 | if (!entries) | |
310 | return -ENOMEM; | |
311 | ||
312 | trace.nr_entries = 0; | |
313 | trace.max_entries = MAX_STACK_TRACE_DEPTH; | |
314 | trace.entries = entries; | |
315 | trace.skip = 0; | |
2ec220e2 | 316 | |
a9712bc1 AV |
317 | err = lock_trace(task); |
318 | if (!err) { | |
319 | save_stack_trace_tsk(task, &trace); | |
320 | ||
321 | for (i = 0; i < trace.nr_entries; i++) { | |
b81a618d | 322 | seq_printf(m, "[<%pK>] %pS\n", |
a9712bc1 AV |
323 | (void *)entries[i], (void *)entries[i]); |
324 | } | |
325 | unlock_trace(task); | |
2ec220e2 KC |
326 | } |
327 | kfree(entries); | |
328 | ||
a9712bc1 | 329 | return err; |
2ec220e2 KC |
330 | } |
331 | #endif | |
332 | ||
1da177e4 LT |
333 | #ifdef CONFIG_SCHEDSTATS |
334 | /* | |
335 | * Provides /proc/PID/schedstat | |
336 | */ | |
337 | static int proc_pid_schedstat(struct task_struct *task, char *buffer) | |
338 | { | |
172ba844 | 339 | return sprintf(buffer, "%llu %llu %lu\n", |
826e08b0 IM |
340 | (unsigned long long)task->se.sum_exec_runtime, |
341 | (unsigned long long)task->sched_info.run_delay, | |
2d72376b | 342 | task->sched_info.pcount); |
1da177e4 LT |
343 | } |
344 | #endif | |
345 | ||
9745512c AV |
346 | #ifdef CONFIG_LATENCYTOP |
347 | static int lstats_show_proc(struct seq_file *m, void *v) | |
348 | { | |
349 | int i; | |
13d77c37 HS |
350 | struct inode *inode = m->private; |
351 | struct task_struct *task = get_proc_task(inode); | |
9745512c | 352 | |
13d77c37 HS |
353 | if (!task) |
354 | return -ESRCH; | |
355 | seq_puts(m, "Latency Top version : v0.1\n"); | |
9745512c | 356 | for (i = 0; i < 32; i++) { |
34e49d4f JP |
357 | struct latency_record *lr = &task->latency_record[i]; |
358 | if (lr->backtrace[0]) { | |
9745512c | 359 | int q; |
34e49d4f JP |
360 | seq_printf(m, "%i %li %li", |
361 | lr->count, lr->time, lr->max); | |
9745512c | 362 | for (q = 0; q < LT_BACKTRACEDEPTH; q++) { |
34e49d4f JP |
363 | unsigned long bt = lr->backtrace[q]; |
364 | if (!bt) | |
9745512c | 365 | break; |
34e49d4f | 366 | if (bt == ULONG_MAX) |
9745512c | 367 | break; |
34e49d4f | 368 | seq_printf(m, " %ps", (void *)bt); |
9745512c | 369 | } |
9d6de12f | 370 | seq_putc(m, '\n'); |
9745512c AV |
371 | } |
372 | ||
373 | } | |
13d77c37 | 374 | put_task_struct(task); |
9745512c AV |
375 | return 0; |
376 | } | |
377 | ||
378 | static int lstats_open(struct inode *inode, struct file *file) | |
379 | { | |
13d77c37 | 380 | return single_open(file, lstats_show_proc, inode); |
d6643d12 HS |
381 | } |
382 | ||
9745512c AV |
383 | static ssize_t lstats_write(struct file *file, const char __user *buf, |
384 | size_t count, loff_t *offs) | |
385 | { | |
13d77c37 | 386 | struct task_struct *task = get_proc_task(file->f_dentry->d_inode); |
9745512c | 387 | |
13d77c37 HS |
388 | if (!task) |
389 | return -ESRCH; | |
9745512c | 390 | clear_all_latency_tracing(task); |
13d77c37 | 391 | put_task_struct(task); |
9745512c AV |
392 | |
393 | return count; | |
394 | } | |
395 | ||
396 | static const struct file_operations proc_lstats_operations = { | |
397 | .open = lstats_open, | |
398 | .read = seq_read, | |
399 | .write = lstats_write, | |
400 | .llseek = seq_lseek, | |
13d77c37 | 401 | .release = single_release, |
9745512c AV |
402 | }; |
403 | ||
404 | #endif | |
405 | ||
1da177e4 LT |
406 | static int proc_oom_score(struct task_struct *task, char *buffer) |
407 | { | |
a7f638f9 | 408 | unsigned long totalpages = totalram_pages + total_swap_pages; |
b95c35e7 | 409 | unsigned long points = 0; |
1da177e4 | 410 | |
19c5d45a | 411 | read_lock(&tasklist_lock); |
b95c35e7 | 412 | if (pid_alive(task)) |
a7f638f9 DR |
413 | points = oom_badness(task, NULL, NULL, totalpages) * |
414 | 1000 / totalpages; | |
19c5d45a | 415 | read_unlock(&tasklist_lock); |
1da177e4 LT |
416 | return sprintf(buffer, "%lu\n", points); |
417 | } | |
418 | ||
d85f50d5 NH |
419 | struct limit_names { |
420 | char *name; | |
421 | char *unit; | |
422 | }; | |
423 | ||
424 | static const struct limit_names lnames[RLIM_NLIMITS] = { | |
cff4edb5 | 425 | [RLIMIT_CPU] = {"Max cpu time", "seconds"}, |
d85f50d5 NH |
426 | [RLIMIT_FSIZE] = {"Max file size", "bytes"}, |
427 | [RLIMIT_DATA] = {"Max data size", "bytes"}, | |
428 | [RLIMIT_STACK] = {"Max stack size", "bytes"}, | |
429 | [RLIMIT_CORE] = {"Max core file size", "bytes"}, | |
430 | [RLIMIT_RSS] = {"Max resident set", "bytes"}, | |
431 | [RLIMIT_NPROC] = {"Max processes", "processes"}, | |
432 | [RLIMIT_NOFILE] = {"Max open files", "files"}, | |
433 | [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"}, | |
434 | [RLIMIT_AS] = {"Max address space", "bytes"}, | |
435 | [RLIMIT_LOCKS] = {"Max file locks", "locks"}, | |
436 | [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"}, | |
437 | [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"}, | |
438 | [RLIMIT_NICE] = {"Max nice priority", NULL}, | |
439 | [RLIMIT_RTPRIO] = {"Max realtime priority", NULL}, | |
8808117c | 440 | [RLIMIT_RTTIME] = {"Max realtime timeout", "us"}, |
d85f50d5 NH |
441 | }; |
442 | ||
443 | /* Display limits for a process */ | |
444 | static int proc_pid_limits(struct task_struct *task, char *buffer) | |
445 | { | |
446 | unsigned int i; | |
447 | int count = 0; | |
448 | unsigned long flags; | |
449 | char *bufptr = buffer; | |
450 | ||
451 | struct rlimit rlim[RLIM_NLIMITS]; | |
452 | ||
a6bebbc8 | 453 | if (!lock_task_sighand(task, &flags)) |
d85f50d5 | 454 | return 0; |
d85f50d5 NH |
455 | memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS); |
456 | unlock_task_sighand(task, &flags); | |
d85f50d5 NH |
457 | |
458 | /* | |
459 | * print the file header | |
460 | */ | |
461 | count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n", | |
462 | "Limit", "Soft Limit", "Hard Limit", "Units"); | |
463 | ||
464 | for (i = 0; i < RLIM_NLIMITS; i++) { | |
465 | if (rlim[i].rlim_cur == RLIM_INFINITY) | |
466 | count += sprintf(&bufptr[count], "%-25s %-20s ", | |
467 | lnames[i].name, "unlimited"); | |
468 | else | |
469 | count += sprintf(&bufptr[count], "%-25s %-20lu ", | |
470 | lnames[i].name, rlim[i].rlim_cur); | |
471 | ||
472 | if (rlim[i].rlim_max == RLIM_INFINITY) | |
473 | count += sprintf(&bufptr[count], "%-20s ", "unlimited"); | |
474 | else | |
475 | count += sprintf(&bufptr[count], "%-20lu ", | |
476 | rlim[i].rlim_max); | |
477 | ||
478 | if (lnames[i].unit) | |
479 | count += sprintf(&bufptr[count], "%-10s\n", | |
480 | lnames[i].unit); | |
481 | else | |
482 | count += sprintf(&bufptr[count], "\n"); | |
483 | } | |
484 | ||
485 | return count; | |
486 | } | |
487 | ||
ebcb6734 RM |
488 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
489 | static int proc_pid_syscall(struct task_struct *task, char *buffer) | |
490 | { | |
491 | long nr; | |
492 | unsigned long args[6], sp, pc; | |
a9712bc1 AV |
493 | int res = lock_trace(task); |
494 | if (res) | |
495 | return res; | |
ebcb6734 RM |
496 | |
497 | if (task_current_syscall(task, &nr, args, 6, &sp, &pc)) | |
a9712bc1 AV |
498 | res = sprintf(buffer, "running\n"); |
499 | else if (nr < 0) | |
500 | res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc); | |
501 | else | |
502 | res = sprintf(buffer, | |
ebcb6734 RM |
503 | "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n", |
504 | nr, | |
505 | args[0], args[1], args[2], args[3], args[4], args[5], | |
506 | sp, pc); | |
a9712bc1 AV |
507 | unlock_trace(task); |
508 | return res; | |
ebcb6734 RM |
509 | } |
510 | #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */ | |
511 | ||
1da177e4 LT |
512 | /************************************************************************/ |
513 | /* Here the fs part begins */ | |
514 | /************************************************************************/ | |
515 | ||
516 | /* permission checks */ | |
778c1144 | 517 | static int proc_fd_access_allowed(struct inode *inode) |
1da177e4 | 518 | { |
778c1144 EB |
519 | struct task_struct *task; |
520 | int allowed = 0; | |
df26c40e EB |
521 | /* Allow access to a task's file descriptors if it is us or we |
522 | * may use ptrace attach to the process and find out that | |
523 | * information. | |
778c1144 EB |
524 | */ |
525 | task = get_proc_task(inode); | |
df26c40e | 526 | if (task) { |
006ebb40 | 527 | allowed = ptrace_may_access(task, PTRACE_MODE_READ); |
778c1144 | 528 | put_task_struct(task); |
df26c40e | 529 | } |
778c1144 | 530 | return allowed; |
1da177e4 LT |
531 | } |
532 | ||
6b4e306a | 533 | int proc_setattr(struct dentry *dentry, struct iattr *attr) |
6d76fa58 LT |
534 | { |
535 | int error; | |
536 | struct inode *inode = dentry->d_inode; | |
537 | ||
538 | if (attr->ia_valid & ATTR_MODE) | |
539 | return -EPERM; | |
540 | ||
541 | error = inode_change_ok(inode, attr); | |
1025774c CH |
542 | if (error) |
543 | return error; | |
544 | ||
1025774c CH |
545 | setattr_copy(inode, attr); |
546 | mark_inode_dirty(inode); | |
547 | return 0; | |
6d76fa58 LT |
548 | } |
549 | ||
0499680a VK |
550 | /* |
551 | * May current process learn task's sched/cmdline info (for hide_pid_min=1) | |
552 | * or euid/egid (for hide_pid_min=2)? | |
553 | */ | |
554 | static bool has_pid_permissions(struct pid_namespace *pid, | |
555 | struct task_struct *task, | |
556 | int hide_pid_min) | |
557 | { | |
558 | if (pid->hide_pid < hide_pid_min) | |
559 | return true; | |
560 | if (in_group_p(pid->pid_gid)) | |
561 | return true; | |
562 | return ptrace_may_access(task, PTRACE_MODE_READ); | |
563 | } | |
564 | ||
565 | ||
566 | static int proc_pid_permission(struct inode *inode, int mask) | |
567 | { | |
568 | struct pid_namespace *pid = inode->i_sb->s_fs_info; | |
569 | struct task_struct *task; | |
570 | bool has_perms; | |
571 | ||
572 | task = get_proc_task(inode); | |
a2ef990a XF |
573 | if (!task) |
574 | return -ESRCH; | |
0499680a VK |
575 | has_perms = has_pid_permissions(pid, task, 1); |
576 | put_task_struct(task); | |
577 | ||
578 | if (!has_perms) { | |
579 | if (pid->hide_pid == 2) { | |
580 | /* | |
581 | * Let's make getdents(), stat(), and open() | |
582 | * consistent with each other. If a process | |
583 | * may not stat() a file, it shouldn't be seen | |
584 | * in procfs at all. | |
585 | */ | |
586 | return -ENOENT; | |
587 | } | |
588 | ||
589 | return -EPERM; | |
590 | } | |
591 | return generic_permission(inode, mask); | |
592 | } | |
593 | ||
594 | ||
595 | ||
c5ef1c42 | 596 | static const struct inode_operations proc_def_inode_operations = { |
6d76fa58 LT |
597 | .setattr = proc_setattr, |
598 | }; | |
599 | ||
1da177e4 LT |
600 | #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */ |
601 | ||
602 | static ssize_t proc_info_read(struct file * file, char __user * buf, | |
603 | size_t count, loff_t *ppos) | |
604 | { | |
2fddfeef | 605 | struct inode * inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
606 | unsigned long page; |
607 | ssize_t length; | |
99f89551 EB |
608 | struct task_struct *task = get_proc_task(inode); |
609 | ||
610 | length = -ESRCH; | |
611 | if (!task) | |
612 | goto out_no_task; | |
1da177e4 LT |
613 | |
614 | if (count > PROC_BLOCK_SIZE) | |
615 | count = PROC_BLOCK_SIZE; | |
99f89551 EB |
616 | |
617 | length = -ENOMEM; | |
e12ba74d | 618 | if (!(page = __get_free_page(GFP_TEMPORARY))) |
99f89551 | 619 | goto out; |
1da177e4 LT |
620 | |
621 | length = PROC_I(inode)->op.proc_read(task, (char*)page); | |
622 | ||
623 | if (length >= 0) | |
624 | length = simple_read_from_buffer(buf, count, ppos, (char *)page, length); | |
625 | free_page(page); | |
99f89551 EB |
626 | out: |
627 | put_task_struct(task); | |
628 | out_no_task: | |
1da177e4 LT |
629 | return length; |
630 | } | |
631 | ||
00977a59 | 632 | static const struct file_operations proc_info_file_operations = { |
1da177e4 | 633 | .read = proc_info_read, |
87df8424 | 634 | .llseek = generic_file_llseek, |
1da177e4 LT |
635 | }; |
636 | ||
be614086 EB |
637 | static int proc_single_show(struct seq_file *m, void *v) |
638 | { | |
639 | struct inode *inode = m->private; | |
640 | struct pid_namespace *ns; | |
641 | struct pid *pid; | |
642 | struct task_struct *task; | |
643 | int ret; | |
644 | ||
645 | ns = inode->i_sb->s_fs_info; | |
646 | pid = proc_pid(inode); | |
647 | task = get_pid_task(pid, PIDTYPE_PID); | |
648 | if (!task) | |
649 | return -ESRCH; | |
650 | ||
651 | ret = PROC_I(inode)->op.proc_show(m, ns, pid, task); | |
652 | ||
653 | put_task_struct(task); | |
654 | return ret; | |
655 | } | |
656 | ||
657 | static int proc_single_open(struct inode *inode, struct file *filp) | |
658 | { | |
c6a34058 | 659 | return single_open(filp, proc_single_show, inode); |
be614086 EB |
660 | } |
661 | ||
662 | static const struct file_operations proc_single_file_operations = { | |
663 | .open = proc_single_open, | |
664 | .read = seq_read, | |
665 | .llseek = seq_lseek, | |
666 | .release = single_release, | |
667 | }; | |
668 | ||
b409e578 | 669 | static int __mem_open(struct inode *inode, struct file *file, unsigned int mode) |
1da177e4 | 670 | { |
e268337d LT |
671 | struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode); |
672 | struct mm_struct *mm; | |
673 | ||
674 | if (!task) | |
675 | return -ESRCH; | |
676 | ||
b409e578 | 677 | mm = mm_access(task, mode); |
e268337d LT |
678 | put_task_struct(task); |
679 | ||
680 | if (IS_ERR(mm)) | |
681 | return PTR_ERR(mm); | |
682 | ||
6d08f2c7 ON |
683 | if (mm) { |
684 | /* ensure this mm_struct can't be freed */ | |
685 | atomic_inc(&mm->mm_count); | |
686 | /* but do not pin its memory */ | |
687 | mmput(mm); | |
688 | } | |
689 | ||
e268337d LT |
690 | file->private_data = mm; |
691 | ||
1da177e4 LT |
692 | return 0; |
693 | } | |
694 | ||
b409e578 CW |
695 | static int mem_open(struct inode *inode, struct file *file) |
696 | { | |
bc452b4b DH |
697 | int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH); |
698 | ||
699 | /* OK to pass negative loff_t, we can catch out-of-range */ | |
700 | file->f_mode |= FMODE_UNSIGNED_OFFSET; | |
701 | ||
702 | return ret; | |
b409e578 CW |
703 | } |
704 | ||
572d34b9 ON |
705 | static ssize_t mem_rw(struct file *file, char __user *buf, |
706 | size_t count, loff_t *ppos, int write) | |
1da177e4 | 707 | { |
e268337d | 708 | struct mm_struct *mm = file->private_data; |
572d34b9 ON |
709 | unsigned long addr = *ppos; |
710 | ssize_t copied; | |
1da177e4 | 711 | char *page; |
1da177e4 | 712 | |
e268337d LT |
713 | if (!mm) |
714 | return 0; | |
99f89551 | 715 | |
30cd8903 KM |
716 | page = (char *)__get_free_page(GFP_TEMPORARY); |
717 | if (!page) | |
e268337d | 718 | return -ENOMEM; |
1da177e4 | 719 | |
f7ca54f4 | 720 | copied = 0; |
6d08f2c7 ON |
721 | if (!atomic_inc_not_zero(&mm->mm_users)) |
722 | goto free; | |
723 | ||
1da177e4 | 724 | while (count > 0) { |
572d34b9 | 725 | int this_len = min_t(int, count, PAGE_SIZE); |
1da177e4 | 726 | |
572d34b9 | 727 | if (write && copy_from_user(page, buf, this_len)) { |
1da177e4 LT |
728 | copied = -EFAULT; |
729 | break; | |
730 | } | |
572d34b9 ON |
731 | |
732 | this_len = access_remote_vm(mm, addr, page, this_len, write); | |
733 | if (!this_len) { | |
1da177e4 LT |
734 | if (!copied) |
735 | copied = -EIO; | |
736 | break; | |
737 | } | |
572d34b9 ON |
738 | |
739 | if (!write && copy_to_user(buf, page, this_len)) { | |
740 | copied = -EFAULT; | |
741 | break; | |
742 | } | |
743 | ||
744 | buf += this_len; | |
745 | addr += this_len; | |
746 | copied += this_len; | |
747 | count -= this_len; | |
1da177e4 | 748 | } |
572d34b9 | 749 | *ppos = addr; |
30cd8903 | 750 | |
6d08f2c7 ON |
751 | mmput(mm); |
752 | free: | |
30cd8903 | 753 | free_page((unsigned long) page); |
1da177e4 LT |
754 | return copied; |
755 | } | |
1da177e4 | 756 | |
572d34b9 ON |
757 | static ssize_t mem_read(struct file *file, char __user *buf, |
758 | size_t count, loff_t *ppos) | |
759 | { | |
760 | return mem_rw(file, buf, count, ppos, 0); | |
761 | } | |
762 | ||
763 | static ssize_t mem_write(struct file *file, const char __user *buf, | |
764 | size_t count, loff_t *ppos) | |
765 | { | |
766 | return mem_rw(file, (char __user*)buf, count, ppos, 1); | |
767 | } | |
768 | ||
85863e47 | 769 | loff_t mem_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
770 | { |
771 | switch (orig) { | |
772 | case 0: | |
773 | file->f_pos = offset; | |
774 | break; | |
775 | case 1: | |
776 | file->f_pos += offset; | |
777 | break; | |
778 | default: | |
779 | return -EINVAL; | |
780 | } | |
781 | force_successful_syscall_return(); | |
782 | return file->f_pos; | |
783 | } | |
784 | ||
e268337d LT |
785 | static int mem_release(struct inode *inode, struct file *file) |
786 | { | |
787 | struct mm_struct *mm = file->private_data; | |
71879d3c | 788 | if (mm) |
6d08f2c7 | 789 | mmdrop(mm); |
e268337d LT |
790 | return 0; |
791 | } | |
792 | ||
00977a59 | 793 | static const struct file_operations proc_mem_operations = { |
1da177e4 LT |
794 | .llseek = mem_lseek, |
795 | .read = mem_read, | |
796 | .write = mem_write, | |
797 | .open = mem_open, | |
e268337d | 798 | .release = mem_release, |
1da177e4 LT |
799 | }; |
800 | ||
b409e578 CW |
801 | static int environ_open(struct inode *inode, struct file *file) |
802 | { | |
803 | return __mem_open(inode, file, PTRACE_MODE_READ); | |
804 | } | |
805 | ||
315e28c8 JP |
806 | static ssize_t environ_read(struct file *file, char __user *buf, |
807 | size_t count, loff_t *ppos) | |
808 | { | |
315e28c8 JP |
809 | char *page; |
810 | unsigned long src = *ppos; | |
b409e578 CW |
811 | int ret = 0; |
812 | struct mm_struct *mm = file->private_data; | |
315e28c8 | 813 | |
b409e578 CW |
814 | if (!mm) |
815 | return 0; | |
315e28c8 | 816 | |
315e28c8 JP |
817 | page = (char *)__get_free_page(GFP_TEMPORARY); |
818 | if (!page) | |
b409e578 | 819 | return -ENOMEM; |
315e28c8 | 820 | |
d6f64b89 | 821 | ret = 0; |
b409e578 CW |
822 | if (!atomic_inc_not_zero(&mm->mm_users)) |
823 | goto free; | |
315e28c8 | 824 | while (count > 0) { |
e8905ec2 DH |
825 | size_t this_len, max_len; |
826 | int retval; | |
315e28c8 | 827 | |
e8905ec2 | 828 | if (src >= (mm->env_end - mm->env_start)) |
315e28c8 JP |
829 | break; |
830 | ||
e8905ec2 DH |
831 | this_len = mm->env_end - (mm->env_start + src); |
832 | ||
833 | max_len = min_t(size_t, PAGE_SIZE, count); | |
834 | this_len = min(max_len, this_len); | |
315e28c8 | 835 | |
b409e578 | 836 | retval = access_remote_vm(mm, (mm->env_start + src), |
315e28c8 JP |
837 | page, this_len, 0); |
838 | ||
839 | if (retval <= 0) { | |
840 | ret = retval; | |
841 | break; | |
842 | } | |
843 | ||
844 | if (copy_to_user(buf, page, retval)) { | |
845 | ret = -EFAULT; | |
846 | break; | |
847 | } | |
848 | ||
849 | ret += retval; | |
850 | src += retval; | |
851 | buf += retval; | |
852 | count -= retval; | |
853 | } | |
854 | *ppos = src; | |
315e28c8 | 855 | mmput(mm); |
b409e578 CW |
856 | |
857 | free: | |
315e28c8 | 858 | free_page((unsigned long) page); |
315e28c8 JP |
859 | return ret; |
860 | } | |
861 | ||
862 | static const struct file_operations proc_environ_operations = { | |
b409e578 | 863 | .open = environ_open, |
315e28c8 | 864 | .read = environ_read, |
87df8424 | 865 | .llseek = generic_file_llseek, |
b409e578 | 866 | .release = mem_release, |
315e28c8 JP |
867 | }; |
868 | ||
fa0cbbf1 DR |
869 | static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count, |
870 | loff_t *ppos) | |
871 | { | |
872 | struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode); | |
873 | char buffer[PROC_NUMBUF]; | |
874 | int oom_adj = OOM_ADJUST_MIN; | |
875 | size_t len; | |
876 | unsigned long flags; | |
877 | ||
878 | if (!task) | |
879 | return -ESRCH; | |
880 | if (lock_task_sighand(task, &flags)) { | |
881 | if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX) | |
882 | oom_adj = OOM_ADJUST_MAX; | |
883 | else | |
884 | oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) / | |
885 | OOM_SCORE_ADJ_MAX; | |
886 | unlock_task_sighand(task, &flags); | |
887 | } | |
888 | put_task_struct(task); | |
889 | len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj); | |
890 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
891 | } | |
892 | ||
893 | static ssize_t oom_adj_write(struct file *file, const char __user *buf, | |
894 | size_t count, loff_t *ppos) | |
895 | { | |
896 | struct task_struct *task; | |
897 | char buffer[PROC_NUMBUF]; | |
898 | int oom_adj; | |
899 | unsigned long flags; | |
900 | int err; | |
901 | ||
902 | memset(buffer, 0, sizeof(buffer)); | |
903 | if (count > sizeof(buffer) - 1) | |
904 | count = sizeof(buffer) - 1; | |
905 | if (copy_from_user(buffer, buf, count)) { | |
906 | err = -EFAULT; | |
907 | goto out; | |
908 | } | |
909 | ||
910 | err = kstrtoint(strstrip(buffer), 0, &oom_adj); | |
911 | if (err) | |
912 | goto out; | |
913 | if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) && | |
914 | oom_adj != OOM_DISABLE) { | |
915 | err = -EINVAL; | |
916 | goto out; | |
917 | } | |
918 | ||
919 | task = get_proc_task(file->f_path.dentry->d_inode); | |
920 | if (!task) { | |
921 | err = -ESRCH; | |
922 | goto out; | |
923 | } | |
924 | ||
925 | task_lock(task); | |
926 | if (!task->mm) { | |
927 | err = -EINVAL; | |
928 | goto err_task_lock; | |
929 | } | |
930 | ||
931 | if (!lock_task_sighand(task, &flags)) { | |
932 | err = -ESRCH; | |
933 | goto err_task_lock; | |
934 | } | |
935 | ||
936 | /* | |
937 | * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum | |
938 | * value is always attainable. | |
939 | */ | |
940 | if (oom_adj == OOM_ADJUST_MAX) | |
941 | oom_adj = OOM_SCORE_ADJ_MAX; | |
942 | else | |
943 | oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE; | |
944 | ||
945 | if (oom_adj < task->signal->oom_score_adj && | |
946 | !capable(CAP_SYS_RESOURCE)) { | |
947 | err = -EACCES; | |
948 | goto err_sighand; | |
949 | } | |
950 | ||
951 | /* | |
952 | * /proc/pid/oom_adj is provided for legacy purposes, ask users to use | |
953 | * /proc/pid/oom_score_adj instead. | |
954 | */ | |
955 | printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n", | |
956 | current->comm, task_pid_nr(current), task_pid_nr(task), | |
957 | task_pid_nr(task)); | |
958 | ||
959 | task->signal->oom_score_adj = oom_adj; | |
960 | trace_oom_score_adj_update(task); | |
961 | err_sighand: | |
962 | unlock_task_sighand(task, &flags); | |
963 | err_task_lock: | |
964 | task_unlock(task); | |
965 | put_task_struct(task); | |
966 | out: | |
967 | return err < 0 ? err : count; | |
968 | } | |
969 | ||
970 | static const struct file_operations proc_oom_adj_operations = { | |
971 | .read = oom_adj_read, | |
972 | .write = oom_adj_write, | |
973 | .llseek = generic_file_llseek, | |
974 | }; | |
975 | ||
a63d83f4 DR |
976 | static ssize_t oom_score_adj_read(struct file *file, char __user *buf, |
977 | size_t count, loff_t *ppos) | |
978 | { | |
979 | struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode); | |
980 | char buffer[PROC_NUMBUF]; | |
a9c58b90 | 981 | short oom_score_adj = OOM_SCORE_ADJ_MIN; |
a63d83f4 DR |
982 | unsigned long flags; |
983 | size_t len; | |
984 | ||
985 | if (!task) | |
986 | return -ESRCH; | |
987 | if (lock_task_sighand(task, &flags)) { | |
988 | oom_score_adj = task->signal->oom_score_adj; | |
989 | unlock_task_sighand(task, &flags); | |
990 | } | |
991 | put_task_struct(task); | |
a9c58b90 | 992 | len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj); |
a63d83f4 DR |
993 | return simple_read_from_buffer(buf, count, ppos, buffer, len); |
994 | } | |
995 | ||
996 | static ssize_t oom_score_adj_write(struct file *file, const char __user *buf, | |
997 | size_t count, loff_t *ppos) | |
998 | { | |
999 | struct task_struct *task; | |
1000 | char buffer[PROC_NUMBUF]; | |
1001 | unsigned long flags; | |
0a8cb8e3 | 1002 | int oom_score_adj; |
a63d83f4 DR |
1003 | int err; |
1004 | ||
1005 | memset(buffer, 0, sizeof(buffer)); | |
1006 | if (count > sizeof(buffer) - 1) | |
1007 | count = sizeof(buffer) - 1; | |
723548bf DR |
1008 | if (copy_from_user(buffer, buf, count)) { |
1009 | err = -EFAULT; | |
1010 | goto out; | |
1011 | } | |
a63d83f4 | 1012 | |
0a8cb8e3 | 1013 | err = kstrtoint(strstrip(buffer), 0, &oom_score_adj); |
a63d83f4 | 1014 | if (err) |
723548bf | 1015 | goto out; |
a63d83f4 | 1016 | if (oom_score_adj < OOM_SCORE_ADJ_MIN || |
723548bf DR |
1017 | oom_score_adj > OOM_SCORE_ADJ_MAX) { |
1018 | err = -EINVAL; | |
1019 | goto out; | |
1020 | } | |
a63d83f4 DR |
1021 | |
1022 | task = get_proc_task(file->f_path.dentry->d_inode); | |
723548bf DR |
1023 | if (!task) { |
1024 | err = -ESRCH; | |
1025 | goto out; | |
1026 | } | |
d19d5476 DR |
1027 | |
1028 | task_lock(task); | |
1029 | if (!task->mm) { | |
1030 | err = -EINVAL; | |
1031 | goto err_task_lock; | |
1032 | } | |
1033 | ||
a63d83f4 | 1034 | if (!lock_task_sighand(task, &flags)) { |
723548bf | 1035 | err = -ESRCH; |
d19d5476 | 1036 | goto err_task_lock; |
a63d83f4 | 1037 | } |
d19d5476 | 1038 | |
a9c58b90 | 1039 | if ((short)oom_score_adj < task->signal->oom_score_adj_min && |
a63d83f4 | 1040 | !capable(CAP_SYS_RESOURCE)) { |
723548bf DR |
1041 | err = -EACCES; |
1042 | goto err_sighand; | |
a63d83f4 DR |
1043 | } |
1044 | ||
a9c58b90 | 1045 | task->signal->oom_score_adj = (short)oom_score_adj; |
dabb16f6 | 1046 | if (has_capability_noaudit(current, CAP_SYS_RESOURCE)) |
a9c58b90 | 1047 | task->signal->oom_score_adj_min = (short)oom_score_adj; |
43d2b113 | 1048 | trace_oom_score_adj_update(task); |
01dc52eb | 1049 | |
723548bf | 1050 | err_sighand: |
a63d83f4 | 1051 | unlock_task_sighand(task, &flags); |
d19d5476 DR |
1052 | err_task_lock: |
1053 | task_unlock(task); | |
a63d83f4 | 1054 | put_task_struct(task); |
723548bf DR |
1055 | out: |
1056 | return err < 0 ? err : count; | |
a63d83f4 DR |
1057 | } |
1058 | ||
1059 | static const struct file_operations proc_oom_score_adj_operations = { | |
1060 | .read = oom_score_adj_read, | |
1061 | .write = oom_score_adj_write, | |
6038f373 | 1062 | .llseek = default_llseek, |
a63d83f4 DR |
1063 | }; |
1064 | ||
1da177e4 LT |
1065 | #ifdef CONFIG_AUDITSYSCALL |
1066 | #define TMPBUFLEN 21 | |
1067 | static ssize_t proc_loginuid_read(struct file * file, char __user * buf, | |
1068 | size_t count, loff_t *ppos) | |
1069 | { | |
2fddfeef | 1070 | struct inode * inode = file->f_path.dentry->d_inode; |
99f89551 | 1071 | struct task_struct *task = get_proc_task(inode); |
1da177e4 LT |
1072 | ssize_t length; |
1073 | char tmpbuf[TMPBUFLEN]; | |
1074 | ||
99f89551 EB |
1075 | if (!task) |
1076 | return -ESRCH; | |
1da177e4 | 1077 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", |
e1760bd5 EB |
1078 | from_kuid(file->f_cred->user_ns, |
1079 | audit_get_loginuid(task))); | |
99f89551 | 1080 | put_task_struct(task); |
1da177e4 LT |
1081 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); |
1082 | } | |
1083 | ||
1084 | static ssize_t proc_loginuid_write(struct file * file, const char __user * buf, | |
1085 | size_t count, loff_t *ppos) | |
1086 | { | |
2fddfeef | 1087 | struct inode * inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
1088 | char *page, *tmp; |
1089 | ssize_t length; | |
1da177e4 | 1090 | uid_t loginuid; |
e1760bd5 | 1091 | kuid_t kloginuid; |
1da177e4 | 1092 | |
7dc52157 PM |
1093 | rcu_read_lock(); |
1094 | if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) { | |
1095 | rcu_read_unlock(); | |
1da177e4 | 1096 | return -EPERM; |
7dc52157 PM |
1097 | } |
1098 | rcu_read_unlock(); | |
1da177e4 | 1099 | |
e0182909 AV |
1100 | if (count >= PAGE_SIZE) |
1101 | count = PAGE_SIZE - 1; | |
1da177e4 LT |
1102 | |
1103 | if (*ppos != 0) { | |
1104 | /* No partial writes. */ | |
1105 | return -EINVAL; | |
1106 | } | |
e12ba74d | 1107 | page = (char*)__get_free_page(GFP_TEMPORARY); |
1da177e4 LT |
1108 | if (!page) |
1109 | return -ENOMEM; | |
1110 | length = -EFAULT; | |
1111 | if (copy_from_user(page, buf, count)) | |
1112 | goto out_free_page; | |
1113 | ||
e0182909 | 1114 | page[count] = '\0'; |
1da177e4 LT |
1115 | loginuid = simple_strtoul(page, &tmp, 10); |
1116 | if (tmp == page) { | |
1117 | length = -EINVAL; | |
1118 | goto out_free_page; | |
1119 | ||
1120 | } | |
e1760bd5 EB |
1121 | kloginuid = make_kuid(file->f_cred->user_ns, loginuid); |
1122 | if (!uid_valid(kloginuid)) { | |
1123 | length = -EINVAL; | |
1124 | goto out_free_page; | |
1125 | } | |
1126 | ||
1127 | length = audit_set_loginuid(kloginuid); | |
1da177e4 LT |
1128 | if (likely(length == 0)) |
1129 | length = count; | |
1130 | ||
1131 | out_free_page: | |
1132 | free_page((unsigned long) page); | |
1133 | return length; | |
1134 | } | |
1135 | ||
00977a59 | 1136 | static const struct file_operations proc_loginuid_operations = { |
1da177e4 LT |
1137 | .read = proc_loginuid_read, |
1138 | .write = proc_loginuid_write, | |
87df8424 | 1139 | .llseek = generic_file_llseek, |
1da177e4 | 1140 | }; |
1e0bd755 EP |
1141 | |
1142 | static ssize_t proc_sessionid_read(struct file * file, char __user * buf, | |
1143 | size_t count, loff_t *ppos) | |
1144 | { | |
1145 | struct inode * inode = file->f_path.dentry->d_inode; | |
1146 | struct task_struct *task = get_proc_task(inode); | |
1147 | ssize_t length; | |
1148 | char tmpbuf[TMPBUFLEN]; | |
1149 | ||
1150 | if (!task) | |
1151 | return -ESRCH; | |
1152 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", | |
1153 | audit_get_sessionid(task)); | |
1154 | put_task_struct(task); | |
1155 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); | |
1156 | } | |
1157 | ||
1158 | static const struct file_operations proc_sessionid_operations = { | |
1159 | .read = proc_sessionid_read, | |
87df8424 | 1160 | .llseek = generic_file_llseek, |
1e0bd755 | 1161 | }; |
1da177e4 LT |
1162 | #endif |
1163 | ||
f4f154fd AM |
1164 | #ifdef CONFIG_FAULT_INJECTION |
1165 | static ssize_t proc_fault_inject_read(struct file * file, char __user * buf, | |
1166 | size_t count, loff_t *ppos) | |
1167 | { | |
1168 | struct task_struct *task = get_proc_task(file->f_dentry->d_inode); | |
1169 | char buffer[PROC_NUMBUF]; | |
1170 | size_t len; | |
1171 | int make_it_fail; | |
f4f154fd AM |
1172 | |
1173 | if (!task) | |
1174 | return -ESRCH; | |
1175 | make_it_fail = task->make_it_fail; | |
1176 | put_task_struct(task); | |
1177 | ||
1178 | len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail); | |
0c28f287 AM |
1179 | |
1180 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
f4f154fd AM |
1181 | } |
1182 | ||
1183 | static ssize_t proc_fault_inject_write(struct file * file, | |
1184 | const char __user * buf, size_t count, loff_t *ppos) | |
1185 | { | |
1186 | struct task_struct *task; | |
1187 | char buffer[PROC_NUMBUF], *end; | |
1188 | int make_it_fail; | |
1189 | ||
1190 | if (!capable(CAP_SYS_RESOURCE)) | |
1191 | return -EPERM; | |
1192 | memset(buffer, 0, sizeof(buffer)); | |
1193 | if (count > sizeof(buffer) - 1) | |
1194 | count = sizeof(buffer) - 1; | |
1195 | if (copy_from_user(buffer, buf, count)) | |
1196 | return -EFAULT; | |
cba8aafe VL |
1197 | make_it_fail = simple_strtol(strstrip(buffer), &end, 0); |
1198 | if (*end) | |
1199 | return -EINVAL; | |
f4f154fd AM |
1200 | task = get_proc_task(file->f_dentry->d_inode); |
1201 | if (!task) | |
1202 | return -ESRCH; | |
1203 | task->make_it_fail = make_it_fail; | |
1204 | put_task_struct(task); | |
cba8aafe VL |
1205 | |
1206 | return count; | |
f4f154fd AM |
1207 | } |
1208 | ||
00977a59 | 1209 | static const struct file_operations proc_fault_inject_operations = { |
f4f154fd AM |
1210 | .read = proc_fault_inject_read, |
1211 | .write = proc_fault_inject_write, | |
87df8424 | 1212 | .llseek = generic_file_llseek, |
f4f154fd AM |
1213 | }; |
1214 | #endif | |
1215 | ||
9745512c | 1216 | |
43ae34cb IM |
1217 | #ifdef CONFIG_SCHED_DEBUG |
1218 | /* | |
1219 | * Print out various scheduling related per-task fields: | |
1220 | */ | |
1221 | static int sched_show(struct seq_file *m, void *v) | |
1222 | { | |
1223 | struct inode *inode = m->private; | |
1224 | struct task_struct *p; | |
1225 | ||
43ae34cb IM |
1226 | p = get_proc_task(inode); |
1227 | if (!p) | |
1228 | return -ESRCH; | |
1229 | proc_sched_show_task(p, m); | |
1230 | ||
1231 | put_task_struct(p); | |
1232 | ||
1233 | return 0; | |
1234 | } | |
1235 | ||
1236 | static ssize_t | |
1237 | sched_write(struct file *file, const char __user *buf, | |
1238 | size_t count, loff_t *offset) | |
1239 | { | |
1240 | struct inode *inode = file->f_path.dentry->d_inode; | |
1241 | struct task_struct *p; | |
1242 | ||
43ae34cb IM |
1243 | p = get_proc_task(inode); |
1244 | if (!p) | |
1245 | return -ESRCH; | |
1246 | proc_sched_set_task(p); | |
1247 | ||
1248 | put_task_struct(p); | |
1249 | ||
1250 | return count; | |
1251 | } | |
1252 | ||
1253 | static int sched_open(struct inode *inode, struct file *filp) | |
1254 | { | |
c6a34058 | 1255 | return single_open(filp, sched_show, inode); |
43ae34cb IM |
1256 | } |
1257 | ||
1258 | static const struct file_operations proc_pid_sched_operations = { | |
1259 | .open = sched_open, | |
1260 | .read = seq_read, | |
1261 | .write = sched_write, | |
1262 | .llseek = seq_lseek, | |
5ea473a1 | 1263 | .release = single_release, |
43ae34cb IM |
1264 | }; |
1265 | ||
1266 | #endif | |
1267 | ||
5091faa4 MG |
1268 | #ifdef CONFIG_SCHED_AUTOGROUP |
1269 | /* | |
1270 | * Print out autogroup related information: | |
1271 | */ | |
1272 | static int sched_autogroup_show(struct seq_file *m, void *v) | |
1273 | { | |
1274 | struct inode *inode = m->private; | |
1275 | struct task_struct *p; | |
1276 | ||
1277 | p = get_proc_task(inode); | |
1278 | if (!p) | |
1279 | return -ESRCH; | |
1280 | proc_sched_autogroup_show_task(p, m); | |
1281 | ||
1282 | put_task_struct(p); | |
1283 | ||
1284 | return 0; | |
1285 | } | |
1286 | ||
1287 | static ssize_t | |
1288 | sched_autogroup_write(struct file *file, const char __user *buf, | |
1289 | size_t count, loff_t *offset) | |
1290 | { | |
1291 | struct inode *inode = file->f_path.dentry->d_inode; | |
1292 | struct task_struct *p; | |
1293 | char buffer[PROC_NUMBUF]; | |
0a8cb8e3 | 1294 | int nice; |
5091faa4 MG |
1295 | int err; |
1296 | ||
1297 | memset(buffer, 0, sizeof(buffer)); | |
1298 | if (count > sizeof(buffer) - 1) | |
1299 | count = sizeof(buffer) - 1; | |
1300 | if (copy_from_user(buffer, buf, count)) | |
1301 | return -EFAULT; | |
1302 | ||
0a8cb8e3 AD |
1303 | err = kstrtoint(strstrip(buffer), 0, &nice); |
1304 | if (err < 0) | |
1305 | return err; | |
5091faa4 MG |
1306 | |
1307 | p = get_proc_task(inode); | |
1308 | if (!p) | |
1309 | return -ESRCH; | |
1310 | ||
2e5b5b3a | 1311 | err = proc_sched_autogroup_set_nice(p, nice); |
5091faa4 MG |
1312 | if (err) |
1313 | count = err; | |
1314 | ||
1315 | put_task_struct(p); | |
1316 | ||
1317 | return count; | |
1318 | } | |
1319 | ||
1320 | static int sched_autogroup_open(struct inode *inode, struct file *filp) | |
1321 | { | |
1322 | int ret; | |
1323 | ||
1324 | ret = single_open(filp, sched_autogroup_show, NULL); | |
1325 | if (!ret) { | |
1326 | struct seq_file *m = filp->private_data; | |
1327 | ||
1328 | m->private = inode; | |
1329 | } | |
1330 | return ret; | |
1331 | } | |
1332 | ||
1333 | static const struct file_operations proc_pid_sched_autogroup_operations = { | |
1334 | .open = sched_autogroup_open, | |
1335 | .read = seq_read, | |
1336 | .write = sched_autogroup_write, | |
1337 | .llseek = seq_lseek, | |
1338 | .release = single_release, | |
1339 | }; | |
1340 | ||
1341 | #endif /* CONFIG_SCHED_AUTOGROUP */ | |
1342 | ||
4614a696 | 1343 | static ssize_t comm_write(struct file *file, const char __user *buf, |
1344 | size_t count, loff_t *offset) | |
1345 | { | |
1346 | struct inode *inode = file->f_path.dentry->d_inode; | |
1347 | struct task_struct *p; | |
1348 | char buffer[TASK_COMM_LEN]; | |
1349 | ||
1350 | memset(buffer, 0, sizeof(buffer)); | |
1351 | if (count > sizeof(buffer) - 1) | |
1352 | count = sizeof(buffer) - 1; | |
1353 | if (copy_from_user(buffer, buf, count)) | |
1354 | return -EFAULT; | |
1355 | ||
1356 | p = get_proc_task(inode); | |
1357 | if (!p) | |
1358 | return -ESRCH; | |
1359 | ||
1360 | if (same_thread_group(current, p)) | |
1361 | set_task_comm(p, buffer); | |
1362 | else | |
1363 | count = -EINVAL; | |
1364 | ||
1365 | put_task_struct(p); | |
1366 | ||
1367 | return count; | |
1368 | } | |
1369 | ||
1370 | static int comm_show(struct seq_file *m, void *v) | |
1371 | { | |
1372 | struct inode *inode = m->private; | |
1373 | struct task_struct *p; | |
1374 | ||
1375 | p = get_proc_task(inode); | |
1376 | if (!p) | |
1377 | return -ESRCH; | |
1378 | ||
1379 | task_lock(p); | |
1380 | seq_printf(m, "%s\n", p->comm); | |
1381 | task_unlock(p); | |
1382 | ||
1383 | put_task_struct(p); | |
1384 | ||
1385 | return 0; | |
1386 | } | |
1387 | ||
1388 | static int comm_open(struct inode *inode, struct file *filp) | |
1389 | { | |
c6a34058 | 1390 | return single_open(filp, comm_show, inode); |
4614a696 | 1391 | } |
1392 | ||
1393 | static const struct file_operations proc_pid_set_comm_operations = { | |
1394 | .open = comm_open, | |
1395 | .read = seq_read, | |
1396 | .write = comm_write, | |
1397 | .llseek = seq_lseek, | |
1398 | .release = single_release, | |
1399 | }; | |
1400 | ||
7773fbc5 | 1401 | static int proc_exe_link(struct dentry *dentry, struct path *exe_path) |
925d1c40 MH |
1402 | { |
1403 | struct task_struct *task; | |
1404 | struct mm_struct *mm; | |
1405 | struct file *exe_file; | |
1406 | ||
7773fbc5 | 1407 | task = get_proc_task(dentry->d_inode); |
925d1c40 MH |
1408 | if (!task) |
1409 | return -ENOENT; | |
1410 | mm = get_task_mm(task); | |
1411 | put_task_struct(task); | |
1412 | if (!mm) | |
1413 | return -ENOENT; | |
1414 | exe_file = get_mm_exe_file(mm); | |
1415 | mmput(mm); | |
1416 | if (exe_file) { | |
1417 | *exe_path = exe_file->f_path; | |
1418 | path_get(&exe_file->f_path); | |
1419 | fput(exe_file); | |
1420 | return 0; | |
1421 | } else | |
1422 | return -ENOENT; | |
1423 | } | |
1424 | ||
008b150a | 1425 | static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd) |
1da177e4 LT |
1426 | { |
1427 | struct inode *inode = dentry->d_inode; | |
408ef013 | 1428 | struct path path; |
1da177e4 LT |
1429 | int error = -EACCES; |
1430 | ||
778c1144 EB |
1431 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1432 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1433 | goto out; |
1da177e4 | 1434 | |
408ef013 CH |
1435 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1436 | if (error) | |
1437 | goto out; | |
1438 | ||
b5fb63c1 | 1439 | nd_jump_link(nd, &path); |
408ef013 | 1440 | return NULL; |
1da177e4 | 1441 | out: |
008b150a | 1442 | return ERR_PTR(error); |
1da177e4 LT |
1443 | } |
1444 | ||
3dcd25f3 | 1445 | static int do_proc_readlink(struct path *path, char __user *buffer, int buflen) |
1da177e4 | 1446 | { |
e12ba74d | 1447 | char *tmp = (char*)__get_free_page(GFP_TEMPORARY); |
3dcd25f3 | 1448 | char *pathname; |
1da177e4 LT |
1449 | int len; |
1450 | ||
1451 | if (!tmp) | |
1452 | return -ENOMEM; | |
0c28f287 | 1453 | |
7b2a69ba | 1454 | pathname = d_path(path, tmp, PAGE_SIZE); |
3dcd25f3 JB |
1455 | len = PTR_ERR(pathname); |
1456 | if (IS_ERR(pathname)) | |
1da177e4 | 1457 | goto out; |
3dcd25f3 | 1458 | len = tmp + PAGE_SIZE - 1 - pathname; |
1da177e4 LT |
1459 | |
1460 | if (len > buflen) | |
1461 | len = buflen; | |
3dcd25f3 | 1462 | if (copy_to_user(buffer, pathname, len)) |
1da177e4 LT |
1463 | len = -EFAULT; |
1464 | out: | |
1465 | free_page((unsigned long)tmp); | |
1466 | return len; | |
1467 | } | |
1468 | ||
1469 | static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen) | |
1470 | { | |
1471 | int error = -EACCES; | |
1472 | struct inode *inode = dentry->d_inode; | |
3dcd25f3 | 1473 | struct path path; |
1da177e4 | 1474 | |
778c1144 EB |
1475 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1476 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1477 | goto out; |
1da177e4 | 1478 | |
7773fbc5 | 1479 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1da177e4 LT |
1480 | if (error) |
1481 | goto out; | |
1482 | ||
3dcd25f3 JB |
1483 | error = do_proc_readlink(&path, buffer, buflen); |
1484 | path_put(&path); | |
1da177e4 | 1485 | out: |
1da177e4 LT |
1486 | return error; |
1487 | } | |
1488 | ||
faf60af1 | 1489 | const struct inode_operations proc_pid_link_inode_operations = { |
1da177e4 | 1490 | .readlink = proc_pid_readlink, |
6d76fa58 LT |
1491 | .follow_link = proc_pid_follow_link, |
1492 | .setattr = proc_setattr, | |
1da177e4 LT |
1493 | }; |
1494 | ||
28a6d671 EB |
1495 | |
1496 | /* building an inode */ | |
1497 | ||
6b4e306a | 1498 | struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task) |
28a6d671 EB |
1499 | { |
1500 | struct inode * inode; | |
1501 | struct proc_inode *ei; | |
c69e8d9c | 1502 | const struct cred *cred; |
1da177e4 | 1503 | |
28a6d671 | 1504 | /* We need a new inode */ |
1da177e4 | 1505 | |
28a6d671 EB |
1506 | inode = new_inode(sb); |
1507 | if (!inode) | |
1508 | goto out; | |
1509 | ||
1510 | /* Common stuff */ | |
1511 | ei = PROC_I(inode); | |
85fe4025 | 1512 | inode->i_ino = get_next_ino(); |
28a6d671 | 1513 | inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; |
28a6d671 EB |
1514 | inode->i_op = &proc_def_inode_operations; |
1515 | ||
1516 | /* | |
1517 | * grab the reference to task. | |
1518 | */ | |
1a657f78 | 1519 | ei->pid = get_task_pid(task, PIDTYPE_PID); |
28a6d671 EB |
1520 | if (!ei->pid) |
1521 | goto out_unlock; | |
1522 | ||
28a6d671 | 1523 | if (task_dumpable(task)) { |
c69e8d9c DH |
1524 | rcu_read_lock(); |
1525 | cred = __task_cred(task); | |
1526 | inode->i_uid = cred->euid; | |
1527 | inode->i_gid = cred->egid; | |
1528 | rcu_read_unlock(); | |
1da177e4 | 1529 | } |
28a6d671 EB |
1530 | security_task_to_inode(task, inode); |
1531 | ||
1da177e4 | 1532 | out: |
28a6d671 EB |
1533 | return inode; |
1534 | ||
1535 | out_unlock: | |
1536 | iput(inode); | |
1537 | return NULL; | |
1da177e4 LT |
1538 | } |
1539 | ||
6b4e306a | 1540 | int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat) |
1da177e4 | 1541 | { |
1da177e4 | 1542 | struct inode *inode = dentry->d_inode; |
28a6d671 | 1543 | struct task_struct *task; |
c69e8d9c | 1544 | const struct cred *cred; |
0499680a | 1545 | struct pid_namespace *pid = dentry->d_sb->s_fs_info; |
c69e8d9c | 1546 | |
28a6d671 | 1547 | generic_fillattr(inode, stat); |
1da177e4 | 1548 | |
28a6d671 | 1549 | rcu_read_lock(); |
dcb0f222 EB |
1550 | stat->uid = GLOBAL_ROOT_UID; |
1551 | stat->gid = GLOBAL_ROOT_GID; | |
28a6d671 EB |
1552 | task = pid_task(proc_pid(inode), PIDTYPE_PID); |
1553 | if (task) { | |
0499680a VK |
1554 | if (!has_pid_permissions(pid, task, 2)) { |
1555 | rcu_read_unlock(); | |
1556 | /* | |
1557 | * This doesn't prevent learning whether PID exists, | |
1558 | * it only makes getattr() consistent with readdir(). | |
1559 | */ | |
1560 | return -ENOENT; | |
1561 | } | |
28a6d671 EB |
1562 | if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) || |
1563 | task_dumpable(task)) { | |
c69e8d9c DH |
1564 | cred = __task_cred(task); |
1565 | stat->uid = cred->euid; | |
1566 | stat->gid = cred->egid; | |
1da177e4 LT |
1567 | } |
1568 | } | |
28a6d671 | 1569 | rcu_read_unlock(); |
d6e71144 | 1570 | return 0; |
1da177e4 LT |
1571 | } |
1572 | ||
1da177e4 LT |
1573 | /* dentry stuff */ |
1574 | ||
1575 | /* | |
1576 | * Exceptional case: normally we are not allowed to unhash a busy | |
1577 | * directory. In this case, however, we can do it - no aliasing problems | |
1578 | * due to the way we treat inodes. | |
1579 | * | |
1580 | * Rewrite the inode's ownerships here because the owning task may have | |
1581 | * performed a setuid(), etc. | |
99f89551 EB |
1582 | * |
1583 | * Before the /proc/pid/status file was created the only way to read | |
1584 | * the effective uid of a /process was to stat /proc/pid. Reading | |
1585 | * /proc/pid/status is slow enough that procps and other packages | |
1586 | * kept stating /proc/pid. To keep the rules in /proc simple I have | |
1587 | * made this apply to all per process world readable and executable | |
1588 | * directories. | |
1da177e4 | 1589 | */ |
0b728e19 | 1590 | int pid_revalidate(struct dentry *dentry, unsigned int flags) |
1da177e4 | 1591 | { |
34286d66 NP |
1592 | struct inode *inode; |
1593 | struct task_struct *task; | |
c69e8d9c DH |
1594 | const struct cred *cred; |
1595 | ||
0b728e19 | 1596 | if (flags & LOOKUP_RCU) |
34286d66 NP |
1597 | return -ECHILD; |
1598 | ||
1599 | inode = dentry->d_inode; | |
1600 | task = get_proc_task(inode); | |
1601 | ||
99f89551 EB |
1602 | if (task) { |
1603 | if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) || | |
1604 | task_dumpable(task)) { | |
c69e8d9c DH |
1605 | rcu_read_lock(); |
1606 | cred = __task_cred(task); | |
1607 | inode->i_uid = cred->euid; | |
1608 | inode->i_gid = cred->egid; | |
1609 | rcu_read_unlock(); | |
1da177e4 | 1610 | } else { |
dcb0f222 EB |
1611 | inode->i_uid = GLOBAL_ROOT_UID; |
1612 | inode->i_gid = GLOBAL_ROOT_GID; | |
1da177e4 | 1613 | } |
9ee8ab9f | 1614 | inode->i_mode &= ~(S_ISUID | S_ISGID); |
1da177e4 | 1615 | security_task_to_inode(task, inode); |
99f89551 | 1616 | put_task_struct(task); |
1da177e4 LT |
1617 | return 1; |
1618 | } | |
1619 | d_drop(dentry); | |
1620 | return 0; | |
1621 | } | |
1622 | ||
6b4e306a | 1623 | const struct dentry_operations pid_dentry_operations = |
28a6d671 EB |
1624 | { |
1625 | .d_revalidate = pid_revalidate, | |
1626 | .d_delete = pid_delete_dentry, | |
1627 | }; | |
1628 | ||
1629 | /* Lookups */ | |
1630 | ||
1c0d04c9 EB |
1631 | /* |
1632 | * Fill a directory entry. | |
1633 | * | |
1634 | * If possible create the dcache entry and derive our inode number and | |
1635 | * file type from dcache entry. | |
1636 | * | |
1637 | * Since all of the proc inode numbers are dynamically generated, the inode | |
1638 | * numbers do not exist until the inode is cache. This means creating the | |
1639 | * the dcache entry in readdir is necessary to keep the inode numbers | |
1640 | * reported by readdir in sync with the inode numbers reported | |
1641 | * by stat. | |
1642 | */ | |
6b4e306a EB |
1643 | int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir, |
1644 | const char *name, int len, | |
c5141e6d | 1645 | instantiate_t instantiate, struct task_struct *task, const void *ptr) |
61a28784 | 1646 | { |
2fddfeef | 1647 | struct dentry *child, *dir = filp->f_path.dentry; |
61a28784 EB |
1648 | struct inode *inode; |
1649 | struct qstr qname; | |
1650 | ino_t ino = 0; | |
1651 | unsigned type = DT_UNKNOWN; | |
1652 | ||
1653 | qname.name = name; | |
1654 | qname.len = len; | |
1655 | qname.hash = full_name_hash(name, len); | |
1656 | ||
1657 | child = d_lookup(dir, &qname); | |
1658 | if (!child) { | |
1659 | struct dentry *new; | |
1660 | new = d_alloc(dir, &qname); | |
1661 | if (new) { | |
1662 | child = instantiate(dir->d_inode, new, task, ptr); | |
1663 | if (child) | |
1664 | dput(new); | |
1665 | else | |
1666 | child = new; | |
1667 | } | |
1668 | } | |
1669 | if (!child || IS_ERR(child) || !child->d_inode) | |
1670 | goto end_instantiate; | |
1671 | inode = child->d_inode; | |
1672 | if (inode) { | |
1673 | ino = inode->i_ino; | |
1674 | type = inode->i_mode >> 12; | |
1675 | } | |
1676 | dput(child); | |
1677 | end_instantiate: | |
1678 | if (!ino) | |
1679 | ino = find_inode_number(dir, &qname); | |
1680 | if (!ino) | |
1681 | ino = 1; | |
1682 | return filldir(dirent, name, len, filp->f_pos, ino, type); | |
1683 | } | |
1684 | ||
640708a2 PE |
1685 | #ifdef CONFIG_CHECKPOINT_RESTORE |
1686 | ||
1687 | /* | |
1688 | * dname_to_vma_addr - maps a dentry name into two unsigned longs | |
1689 | * which represent vma start and end addresses. | |
1690 | */ | |
1691 | static int dname_to_vma_addr(struct dentry *dentry, | |
1692 | unsigned long *start, unsigned long *end) | |
1693 | { | |
1694 | if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2) | |
1695 | return -EINVAL; | |
1696 | ||
1697 | return 0; | |
1698 | } | |
1699 | ||
0b728e19 | 1700 | static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags) |
640708a2 PE |
1701 | { |
1702 | unsigned long vm_start, vm_end; | |
1703 | bool exact_vma_exists = false; | |
1704 | struct mm_struct *mm = NULL; | |
1705 | struct task_struct *task; | |
1706 | const struct cred *cred; | |
1707 | struct inode *inode; | |
1708 | int status = 0; | |
1709 | ||
0b728e19 | 1710 | if (flags & LOOKUP_RCU) |
640708a2 PE |
1711 | return -ECHILD; |
1712 | ||
1713 | if (!capable(CAP_SYS_ADMIN)) { | |
1714 | status = -EACCES; | |
1715 | goto out_notask; | |
1716 | } | |
1717 | ||
1718 | inode = dentry->d_inode; | |
1719 | task = get_proc_task(inode); | |
1720 | if (!task) | |
1721 | goto out_notask; | |
1722 | ||
2344bec7 CW |
1723 | mm = mm_access(task, PTRACE_MODE_READ); |
1724 | if (IS_ERR_OR_NULL(mm)) | |
640708a2 PE |
1725 | goto out; |
1726 | ||
1727 | if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) { | |
1728 | down_read(&mm->mmap_sem); | |
1729 | exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end); | |
1730 | up_read(&mm->mmap_sem); | |
1731 | } | |
1732 | ||
1733 | mmput(mm); | |
1734 | ||
1735 | if (exact_vma_exists) { | |
1736 | if (task_dumpable(task)) { | |
1737 | rcu_read_lock(); | |
1738 | cred = __task_cred(task); | |
1739 | inode->i_uid = cred->euid; | |
1740 | inode->i_gid = cred->egid; | |
1741 | rcu_read_unlock(); | |
1742 | } else { | |
dcb0f222 EB |
1743 | inode->i_uid = GLOBAL_ROOT_UID; |
1744 | inode->i_gid = GLOBAL_ROOT_GID; | |
640708a2 PE |
1745 | } |
1746 | security_task_to_inode(task, inode); | |
1747 | status = 1; | |
1748 | } | |
1749 | ||
1750 | out: | |
1751 | put_task_struct(task); | |
1752 | ||
1753 | out_notask: | |
1754 | if (status <= 0) | |
1755 | d_drop(dentry); | |
1756 | ||
1757 | return status; | |
1758 | } | |
1759 | ||
1760 | static const struct dentry_operations tid_map_files_dentry_operations = { | |
1761 | .d_revalidate = map_files_d_revalidate, | |
1762 | .d_delete = pid_delete_dentry, | |
1763 | }; | |
1764 | ||
1765 | static int proc_map_files_get_link(struct dentry *dentry, struct path *path) | |
1766 | { | |
1767 | unsigned long vm_start, vm_end; | |
1768 | struct vm_area_struct *vma; | |
1769 | struct task_struct *task; | |
1770 | struct mm_struct *mm; | |
1771 | int rc; | |
1772 | ||
1773 | rc = -ENOENT; | |
1774 | task = get_proc_task(dentry->d_inode); | |
1775 | if (!task) | |
1776 | goto out; | |
1777 | ||
1778 | mm = get_task_mm(task); | |
1779 | put_task_struct(task); | |
1780 | if (!mm) | |
1781 | goto out; | |
1782 | ||
1783 | rc = dname_to_vma_addr(dentry, &vm_start, &vm_end); | |
1784 | if (rc) | |
1785 | goto out_mmput; | |
1786 | ||
1787 | down_read(&mm->mmap_sem); | |
1788 | vma = find_exact_vma(mm, vm_start, vm_end); | |
1789 | if (vma && vma->vm_file) { | |
1790 | *path = vma->vm_file->f_path; | |
1791 | path_get(path); | |
1792 | rc = 0; | |
1793 | } | |
1794 | up_read(&mm->mmap_sem); | |
1795 | ||
1796 | out_mmput: | |
1797 | mmput(mm); | |
1798 | out: | |
1799 | return rc; | |
1800 | } | |
1801 | ||
1802 | struct map_files_info { | |
7b540d06 | 1803 | fmode_t mode; |
640708a2 PE |
1804 | unsigned long len; |
1805 | unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */ | |
1806 | }; | |
1807 | ||
1808 | static struct dentry * | |
1809 | proc_map_files_instantiate(struct inode *dir, struct dentry *dentry, | |
1810 | struct task_struct *task, const void *ptr) | |
1811 | { | |
7b540d06 | 1812 | fmode_t mode = (fmode_t)(unsigned long)ptr; |
640708a2 PE |
1813 | struct proc_inode *ei; |
1814 | struct inode *inode; | |
1815 | ||
640708a2 PE |
1816 | inode = proc_pid_make_inode(dir->i_sb, task); |
1817 | if (!inode) | |
1818 | return ERR_PTR(-ENOENT); | |
1819 | ||
1820 | ei = PROC_I(inode); | |
1821 | ei->op.proc_get_link = proc_map_files_get_link; | |
1822 | ||
1823 | inode->i_op = &proc_pid_link_inode_operations; | |
1824 | inode->i_size = 64; | |
1825 | inode->i_mode = S_IFLNK; | |
1826 | ||
7b540d06 | 1827 | if (mode & FMODE_READ) |
640708a2 | 1828 | inode->i_mode |= S_IRUSR; |
7b540d06 | 1829 | if (mode & FMODE_WRITE) |
640708a2 PE |
1830 | inode->i_mode |= S_IWUSR; |
1831 | ||
1832 | d_set_d_op(dentry, &tid_map_files_dentry_operations); | |
1833 | d_add(dentry, inode); | |
1834 | ||
1835 | return NULL; | |
1836 | } | |
1837 | ||
1838 | static struct dentry *proc_map_files_lookup(struct inode *dir, | |
00cd8dd3 | 1839 | struct dentry *dentry, unsigned int flags) |
640708a2 PE |
1840 | { |
1841 | unsigned long vm_start, vm_end; | |
1842 | struct vm_area_struct *vma; | |
1843 | struct task_struct *task; | |
1844 | struct dentry *result; | |
1845 | struct mm_struct *mm; | |
1846 | ||
1847 | result = ERR_PTR(-EACCES); | |
1848 | if (!capable(CAP_SYS_ADMIN)) | |
1849 | goto out; | |
1850 | ||
1851 | result = ERR_PTR(-ENOENT); | |
1852 | task = get_proc_task(dir); | |
1853 | if (!task) | |
1854 | goto out; | |
1855 | ||
1856 | result = ERR_PTR(-EACCES); | |
eb94cd96 | 1857 | if (!ptrace_may_access(task, PTRACE_MODE_READ)) |
640708a2 PE |
1858 | goto out_put_task; |
1859 | ||
1860 | result = ERR_PTR(-ENOENT); | |
1861 | if (dname_to_vma_addr(dentry, &vm_start, &vm_end)) | |
eb94cd96 | 1862 | goto out_put_task; |
640708a2 PE |
1863 | |
1864 | mm = get_task_mm(task); | |
1865 | if (!mm) | |
eb94cd96 | 1866 | goto out_put_task; |
640708a2 PE |
1867 | |
1868 | down_read(&mm->mmap_sem); | |
1869 | vma = find_exact_vma(mm, vm_start, vm_end); | |
1870 | if (!vma) | |
1871 | goto out_no_vma; | |
1872 | ||
05f56484 SK |
1873 | if (vma->vm_file) |
1874 | result = proc_map_files_instantiate(dir, dentry, task, | |
1875 | (void *)(unsigned long)vma->vm_file->f_mode); | |
640708a2 PE |
1876 | |
1877 | out_no_vma: | |
1878 | up_read(&mm->mmap_sem); | |
1879 | mmput(mm); | |
640708a2 PE |
1880 | out_put_task: |
1881 | put_task_struct(task); | |
1882 | out: | |
1883 | return result; | |
1884 | } | |
1885 | ||
1886 | static const struct inode_operations proc_map_files_inode_operations = { | |
1887 | .lookup = proc_map_files_lookup, | |
1888 | .permission = proc_fd_permission, | |
1889 | .setattr = proc_setattr, | |
1890 | }; | |
1891 | ||
1892 | static int | |
1893 | proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir) | |
1894 | { | |
1895 | struct dentry *dentry = filp->f_path.dentry; | |
1896 | struct inode *inode = dentry->d_inode; | |
1897 | struct vm_area_struct *vma; | |
1898 | struct task_struct *task; | |
1899 | struct mm_struct *mm; | |
1900 | ino_t ino; | |
1901 | int ret; | |
1902 | ||
1903 | ret = -EACCES; | |
1904 | if (!capable(CAP_SYS_ADMIN)) | |
1905 | goto out; | |
1906 | ||
1907 | ret = -ENOENT; | |
1908 | task = get_proc_task(inode); | |
1909 | if (!task) | |
1910 | goto out; | |
1911 | ||
1912 | ret = -EACCES; | |
eb94cd96 | 1913 | if (!ptrace_may_access(task, PTRACE_MODE_READ)) |
640708a2 PE |
1914 | goto out_put_task; |
1915 | ||
1916 | ret = 0; | |
1917 | switch (filp->f_pos) { | |
1918 | case 0: | |
1919 | ino = inode->i_ino; | |
1920 | if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0) | |
eb94cd96 | 1921 | goto out_put_task; |
640708a2 PE |
1922 | filp->f_pos++; |
1923 | case 1: | |
1924 | ino = parent_ino(dentry); | |
1925 | if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0) | |
eb94cd96 | 1926 | goto out_put_task; |
640708a2 PE |
1927 | filp->f_pos++; |
1928 | default: | |
1929 | { | |
1930 | unsigned long nr_files, pos, i; | |
1931 | struct flex_array *fa = NULL; | |
1932 | struct map_files_info info; | |
1933 | struct map_files_info *p; | |
1934 | ||
1935 | mm = get_task_mm(task); | |
1936 | if (!mm) | |
eb94cd96 | 1937 | goto out_put_task; |
640708a2 PE |
1938 | down_read(&mm->mmap_sem); |
1939 | ||
1940 | nr_files = 0; | |
1941 | ||
1942 | /* | |
1943 | * We need two passes here: | |
1944 | * | |
1945 | * 1) Collect vmas of mapped files with mmap_sem taken | |
1946 | * 2) Release mmap_sem and instantiate entries | |
1947 | * | |
1948 | * otherwise we get lockdep complained, since filldir() | |
1949 | * routine might require mmap_sem taken in might_fault(). | |
1950 | */ | |
1951 | ||
1952 | for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) { | |
1953 | if (vma->vm_file && ++pos > filp->f_pos) | |
1954 | nr_files++; | |
1955 | } | |
1956 | ||
1957 | if (nr_files) { | |
1958 | fa = flex_array_alloc(sizeof(info), nr_files, | |
1959 | GFP_KERNEL); | |
1960 | if (!fa || flex_array_prealloc(fa, 0, nr_files, | |
1961 | GFP_KERNEL)) { | |
1962 | ret = -ENOMEM; | |
1963 | if (fa) | |
1964 | flex_array_free(fa); | |
1965 | up_read(&mm->mmap_sem); | |
1966 | mmput(mm); | |
eb94cd96 | 1967 | goto out_put_task; |
640708a2 PE |
1968 | } |
1969 | for (i = 0, vma = mm->mmap, pos = 2; vma; | |
1970 | vma = vma->vm_next) { | |
1971 | if (!vma->vm_file) | |
1972 | continue; | |
1973 | if (++pos <= filp->f_pos) | |
1974 | continue; | |
1975 | ||
7b540d06 | 1976 | info.mode = vma->vm_file->f_mode; |
640708a2 PE |
1977 | info.len = snprintf(info.name, |
1978 | sizeof(info.name), "%lx-%lx", | |
1979 | vma->vm_start, vma->vm_end); | |
1980 | if (flex_array_put(fa, i++, &info, GFP_KERNEL)) | |
1981 | BUG(); | |
1982 | } | |
1983 | } | |
1984 | up_read(&mm->mmap_sem); | |
1985 | ||
1986 | for (i = 0; i < nr_files; i++) { | |
1987 | p = flex_array_get(fa, i); | |
1988 | ret = proc_fill_cache(filp, dirent, filldir, | |
1989 | p->name, p->len, | |
1990 | proc_map_files_instantiate, | |
7b540d06 AV |
1991 | task, |
1992 | (void *)(unsigned long)p->mode); | |
640708a2 PE |
1993 | if (ret) |
1994 | break; | |
1995 | filp->f_pos++; | |
640708a2 PE |
1996 | } |
1997 | if (fa) | |
1998 | flex_array_free(fa); | |
1999 | mmput(mm); | |
2000 | } | |
2001 | } | |
2002 | ||
640708a2 PE |
2003 | out_put_task: |
2004 | put_task_struct(task); | |
2005 | out: | |
2006 | return ret; | |
2007 | } | |
2008 | ||
2009 | static const struct file_operations proc_map_files_operations = { | |
2010 | .read = generic_read_dir, | |
2011 | .readdir = proc_map_files_readdir, | |
2012 | .llseek = default_llseek, | |
2013 | }; | |
2014 | ||
2015 | #endif /* CONFIG_CHECKPOINT_RESTORE */ | |
2016 | ||
444ceed8 | 2017 | static struct dentry *proc_pident_instantiate(struct inode *dir, |
c5141e6d | 2018 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
444ceed8 | 2019 | { |
c5141e6d | 2020 | const struct pid_entry *p = ptr; |
444ceed8 EB |
2021 | struct inode *inode; |
2022 | struct proc_inode *ei; | |
bd6daba9 | 2023 | struct dentry *error = ERR_PTR(-ENOENT); |
444ceed8 | 2024 | |
61a28784 | 2025 | inode = proc_pid_make_inode(dir->i_sb, task); |
444ceed8 EB |
2026 | if (!inode) |
2027 | goto out; | |
2028 | ||
2029 | ei = PROC_I(inode); | |
2030 | inode->i_mode = p->mode; | |
2031 | if (S_ISDIR(inode->i_mode)) | |
bfe86848 | 2032 | set_nlink(inode, 2); /* Use getattr to fix if necessary */ |
444ceed8 EB |
2033 | if (p->iop) |
2034 | inode->i_op = p->iop; | |
2035 | if (p->fop) | |
2036 | inode->i_fop = p->fop; | |
2037 | ei->op = p->op; | |
fb045adb | 2038 | d_set_d_op(dentry, &pid_dentry_operations); |
444ceed8 EB |
2039 | d_add(dentry, inode); |
2040 | /* Close the race of the process dying before we return the dentry */ | |
0b728e19 | 2041 | if (pid_revalidate(dentry, 0)) |
444ceed8 EB |
2042 | error = NULL; |
2043 | out: | |
2044 | return error; | |
2045 | } | |
2046 | ||
1da177e4 LT |
2047 | static struct dentry *proc_pident_lookup(struct inode *dir, |
2048 | struct dentry *dentry, | |
c5141e6d | 2049 | const struct pid_entry *ents, |
7bcd6b0e | 2050 | unsigned int nents) |
1da177e4 | 2051 | { |
cd6a3ce9 | 2052 | struct dentry *error; |
99f89551 | 2053 | struct task_struct *task = get_proc_task(dir); |
c5141e6d | 2054 | const struct pid_entry *p, *last; |
1da177e4 | 2055 | |
cd6a3ce9 | 2056 | error = ERR_PTR(-ENOENT); |
1da177e4 | 2057 | |
99f89551 EB |
2058 | if (!task) |
2059 | goto out_no_task; | |
1da177e4 | 2060 | |
20cdc894 EB |
2061 | /* |
2062 | * Yes, it does not scale. And it should not. Don't add | |
2063 | * new entries into /proc/<tgid>/ without very good reasons. | |
2064 | */ | |
7bcd6b0e EB |
2065 | last = &ents[nents - 1]; |
2066 | for (p = ents; p <= last; p++) { | |
1da177e4 LT |
2067 | if (p->len != dentry->d_name.len) |
2068 | continue; | |
2069 | if (!memcmp(dentry->d_name.name, p->name, p->len)) | |
2070 | break; | |
2071 | } | |
7bcd6b0e | 2072 | if (p > last) |
1da177e4 LT |
2073 | goto out; |
2074 | ||
444ceed8 | 2075 | error = proc_pident_instantiate(dir, dentry, task, p); |
1da177e4 | 2076 | out: |
99f89551 EB |
2077 | put_task_struct(task); |
2078 | out_no_task: | |
cd6a3ce9 | 2079 | return error; |
1da177e4 LT |
2080 | } |
2081 | ||
c5141e6d ED |
2082 | static int proc_pident_fill_cache(struct file *filp, void *dirent, |
2083 | filldir_t filldir, struct task_struct *task, const struct pid_entry *p) | |
61a28784 EB |
2084 | { |
2085 | return proc_fill_cache(filp, dirent, filldir, p->name, p->len, | |
2086 | proc_pident_instantiate, task, p); | |
2087 | } | |
2088 | ||
28a6d671 EB |
2089 | static int proc_pident_readdir(struct file *filp, |
2090 | void *dirent, filldir_t filldir, | |
c5141e6d | 2091 | const struct pid_entry *ents, unsigned int nents) |
28a6d671 EB |
2092 | { |
2093 | int i; | |
2fddfeef | 2094 | struct dentry *dentry = filp->f_path.dentry; |
28a6d671 EB |
2095 | struct inode *inode = dentry->d_inode; |
2096 | struct task_struct *task = get_proc_task(inode); | |
c5141e6d | 2097 | const struct pid_entry *p, *last; |
28a6d671 EB |
2098 | ino_t ino; |
2099 | int ret; | |
2100 | ||
2101 | ret = -ENOENT; | |
2102 | if (!task) | |
61a28784 | 2103 | goto out_no_task; |
28a6d671 EB |
2104 | |
2105 | ret = 0; | |
28a6d671 EB |
2106 | i = filp->f_pos; |
2107 | switch (i) { | |
2108 | case 0: | |
2109 | ino = inode->i_ino; | |
2110 | if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0) | |
2111 | goto out; | |
2112 | i++; | |
2113 | filp->f_pos++; | |
2114 | /* fall through */ | |
2115 | case 1: | |
2116 | ino = parent_ino(dentry); | |
2117 | if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0) | |
2118 | goto out; | |
2119 | i++; | |
2120 | filp->f_pos++; | |
2121 | /* fall through */ | |
2122 | default: | |
2123 | i -= 2; | |
2124 | if (i >= nents) { | |
2125 | ret = 1; | |
2126 | goto out; | |
2127 | } | |
2128 | p = ents + i; | |
7bcd6b0e EB |
2129 | last = &ents[nents - 1]; |
2130 | while (p <= last) { | |
61a28784 | 2131 | if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0) |
28a6d671 EB |
2132 | goto out; |
2133 | filp->f_pos++; | |
2134 | p++; | |
2135 | } | |
2136 | } | |
2137 | ||
2138 | ret = 1; | |
2139 | out: | |
61a28784 EB |
2140 | put_task_struct(task); |
2141 | out_no_task: | |
28a6d671 | 2142 | return ret; |
1da177e4 LT |
2143 | } |
2144 | ||
28a6d671 EB |
2145 | #ifdef CONFIG_SECURITY |
2146 | static ssize_t proc_pid_attr_read(struct file * file, char __user * buf, | |
2147 | size_t count, loff_t *ppos) | |
2148 | { | |
2fddfeef | 2149 | struct inode * inode = file->f_path.dentry->d_inode; |
04ff9708 | 2150 | char *p = NULL; |
28a6d671 EB |
2151 | ssize_t length; |
2152 | struct task_struct *task = get_proc_task(inode); | |
2153 | ||
28a6d671 | 2154 | if (!task) |
04ff9708 | 2155 | return -ESRCH; |
28a6d671 EB |
2156 | |
2157 | length = security_getprocattr(task, | |
2fddfeef | 2158 | (char*)file->f_path.dentry->d_name.name, |
04ff9708 | 2159 | &p); |
28a6d671 | 2160 | put_task_struct(task); |
04ff9708 AV |
2161 | if (length > 0) |
2162 | length = simple_read_from_buffer(buf, count, ppos, p, length); | |
2163 | kfree(p); | |
28a6d671 | 2164 | return length; |
1da177e4 LT |
2165 | } |
2166 | ||
28a6d671 EB |
2167 | static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf, |
2168 | size_t count, loff_t *ppos) | |
2169 | { | |
2fddfeef | 2170 | struct inode * inode = file->f_path.dentry->d_inode; |
28a6d671 EB |
2171 | char *page; |
2172 | ssize_t length; | |
2173 | struct task_struct *task = get_proc_task(inode); | |
2174 | ||
2175 | length = -ESRCH; | |
2176 | if (!task) | |
2177 | goto out_no_task; | |
2178 | if (count > PAGE_SIZE) | |
2179 | count = PAGE_SIZE; | |
2180 | ||
2181 | /* No partial writes. */ | |
2182 | length = -EINVAL; | |
2183 | if (*ppos != 0) | |
2184 | goto out; | |
2185 | ||
2186 | length = -ENOMEM; | |
e12ba74d | 2187 | page = (char*)__get_free_page(GFP_TEMPORARY); |
28a6d671 EB |
2188 | if (!page) |
2189 | goto out; | |
2190 | ||
2191 | length = -EFAULT; | |
2192 | if (copy_from_user(page, buf, count)) | |
2193 | goto out_free; | |
2194 | ||
107db7c7 | 2195 | /* Guard against adverse ptrace interaction */ |
9b1bf12d | 2196 | length = mutex_lock_interruptible(&task->signal->cred_guard_mutex); |
107db7c7 DH |
2197 | if (length < 0) |
2198 | goto out_free; | |
2199 | ||
28a6d671 | 2200 | length = security_setprocattr(task, |
2fddfeef | 2201 | (char*)file->f_path.dentry->d_name.name, |
28a6d671 | 2202 | (void*)page, count); |
9b1bf12d | 2203 | mutex_unlock(&task->signal->cred_guard_mutex); |
28a6d671 EB |
2204 | out_free: |
2205 | free_page((unsigned long) page); | |
2206 | out: | |
2207 | put_task_struct(task); | |
2208 | out_no_task: | |
2209 | return length; | |
2210 | } | |
2211 | ||
00977a59 | 2212 | static const struct file_operations proc_pid_attr_operations = { |
28a6d671 EB |
2213 | .read = proc_pid_attr_read, |
2214 | .write = proc_pid_attr_write, | |
87df8424 | 2215 | .llseek = generic_file_llseek, |
28a6d671 EB |
2216 | }; |
2217 | ||
c5141e6d | 2218 | static const struct pid_entry attr_dir_stuff[] = { |
631f9c18 AD |
2219 | REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations), |
2220 | REG("prev", S_IRUGO, proc_pid_attr_operations), | |
2221 | REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2222 | REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2223 | REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2224 | REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
28a6d671 EB |
2225 | }; |
2226 | ||
72d9dcfc | 2227 | static int proc_attr_dir_readdir(struct file * filp, |
28a6d671 EB |
2228 | void * dirent, filldir_t filldir) |
2229 | { | |
2230 | return proc_pident_readdir(filp,dirent,filldir, | |
72d9dcfc | 2231 | attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff)); |
28a6d671 EB |
2232 | } |
2233 | ||
00977a59 | 2234 | static const struct file_operations proc_attr_dir_operations = { |
1da177e4 | 2235 | .read = generic_read_dir, |
72d9dcfc | 2236 | .readdir = proc_attr_dir_readdir, |
6038f373 | 2237 | .llseek = default_llseek, |
1da177e4 LT |
2238 | }; |
2239 | ||
72d9dcfc | 2240 | static struct dentry *proc_attr_dir_lookup(struct inode *dir, |
00cd8dd3 | 2241 | struct dentry *dentry, unsigned int flags) |
28a6d671 | 2242 | { |
7bcd6b0e EB |
2243 | return proc_pident_lookup(dir, dentry, |
2244 | attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2245 | } |
2246 | ||
c5ef1c42 | 2247 | static const struct inode_operations proc_attr_dir_inode_operations = { |
72d9dcfc | 2248 | .lookup = proc_attr_dir_lookup, |
99f89551 | 2249 | .getattr = pid_getattr, |
6d76fa58 | 2250 | .setattr = proc_setattr, |
1da177e4 LT |
2251 | }; |
2252 | ||
28a6d671 EB |
2253 | #endif |
2254 | ||
698ba7b5 | 2255 | #ifdef CONFIG_ELF_CORE |
3cb4a0bb KH |
2256 | static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf, |
2257 | size_t count, loff_t *ppos) | |
2258 | { | |
2259 | struct task_struct *task = get_proc_task(file->f_dentry->d_inode); | |
2260 | struct mm_struct *mm; | |
2261 | char buffer[PROC_NUMBUF]; | |
2262 | size_t len; | |
2263 | int ret; | |
2264 | ||
2265 | if (!task) | |
2266 | return -ESRCH; | |
2267 | ||
2268 | ret = 0; | |
2269 | mm = get_task_mm(task); | |
2270 | if (mm) { | |
2271 | len = snprintf(buffer, sizeof(buffer), "%08lx\n", | |
2272 | ((mm->flags & MMF_DUMP_FILTER_MASK) >> | |
2273 | MMF_DUMP_FILTER_SHIFT)); | |
2274 | mmput(mm); | |
2275 | ret = simple_read_from_buffer(buf, count, ppos, buffer, len); | |
2276 | } | |
2277 | ||
2278 | put_task_struct(task); | |
2279 | ||
2280 | return ret; | |
2281 | } | |
2282 | ||
2283 | static ssize_t proc_coredump_filter_write(struct file *file, | |
2284 | const char __user *buf, | |
2285 | size_t count, | |
2286 | loff_t *ppos) | |
2287 | { | |
2288 | struct task_struct *task; | |
2289 | struct mm_struct *mm; | |
2290 | char buffer[PROC_NUMBUF], *end; | |
2291 | unsigned int val; | |
2292 | int ret; | |
2293 | int i; | |
2294 | unsigned long mask; | |
2295 | ||
2296 | ret = -EFAULT; | |
2297 | memset(buffer, 0, sizeof(buffer)); | |
2298 | if (count > sizeof(buffer) - 1) | |
2299 | count = sizeof(buffer) - 1; | |
2300 | if (copy_from_user(buffer, buf, count)) | |
2301 | goto out_no_task; | |
2302 | ||
2303 | ret = -EINVAL; | |
2304 | val = (unsigned int)simple_strtoul(buffer, &end, 0); | |
2305 | if (*end == '\n') | |
2306 | end++; | |
2307 | if (end - buffer == 0) | |
2308 | goto out_no_task; | |
2309 | ||
2310 | ret = -ESRCH; | |
2311 | task = get_proc_task(file->f_dentry->d_inode); | |
2312 | if (!task) | |
2313 | goto out_no_task; | |
2314 | ||
2315 | ret = end - buffer; | |
2316 | mm = get_task_mm(task); | |
2317 | if (!mm) | |
2318 | goto out_no_mm; | |
2319 | ||
2320 | for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) { | |
2321 | if (val & mask) | |
2322 | set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2323 | else | |
2324 | clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2325 | } | |
2326 | ||
2327 | mmput(mm); | |
2328 | out_no_mm: | |
2329 | put_task_struct(task); | |
2330 | out_no_task: | |
2331 | return ret; | |
2332 | } | |
2333 | ||
2334 | static const struct file_operations proc_coredump_filter_operations = { | |
2335 | .read = proc_coredump_filter_read, | |
2336 | .write = proc_coredump_filter_write, | |
87df8424 | 2337 | .llseek = generic_file_llseek, |
3cb4a0bb KH |
2338 | }; |
2339 | #endif | |
2340 | ||
aba76fdb | 2341 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
297c5d92 AR |
2342 | static int do_io_accounting(struct task_struct *task, char *buffer, int whole) |
2343 | { | |
940389b8 | 2344 | struct task_io_accounting acct = task->ioac; |
5995477a | 2345 | unsigned long flags; |
293eb1e7 | 2346 | int result; |
5995477a | 2347 | |
293eb1e7 VK |
2348 | result = mutex_lock_killable(&task->signal->cred_guard_mutex); |
2349 | if (result) | |
2350 | return result; | |
2351 | ||
2352 | if (!ptrace_may_access(task, PTRACE_MODE_READ)) { | |
2353 | result = -EACCES; | |
2354 | goto out_unlock; | |
2355 | } | |
1d1221f3 | 2356 | |
5995477a AR |
2357 | if (whole && lock_task_sighand(task, &flags)) { |
2358 | struct task_struct *t = task; | |
2359 | ||
2360 | task_io_accounting_add(&acct, &task->signal->ioac); | |
2361 | while_each_thread(task, t) | |
2362 | task_io_accounting_add(&acct, &t->ioac); | |
2363 | ||
2364 | unlock_task_sighand(task, &flags); | |
297c5d92 | 2365 | } |
293eb1e7 | 2366 | result = sprintf(buffer, |
aba76fdb AM |
2367 | "rchar: %llu\n" |
2368 | "wchar: %llu\n" | |
2369 | "syscr: %llu\n" | |
2370 | "syscw: %llu\n" | |
2371 | "read_bytes: %llu\n" | |
2372 | "write_bytes: %llu\n" | |
2373 | "cancelled_write_bytes: %llu\n", | |
7c44319d AB |
2374 | (unsigned long long)acct.rchar, |
2375 | (unsigned long long)acct.wchar, | |
2376 | (unsigned long long)acct.syscr, | |
2377 | (unsigned long long)acct.syscw, | |
2378 | (unsigned long long)acct.read_bytes, | |
2379 | (unsigned long long)acct.write_bytes, | |
2380 | (unsigned long long)acct.cancelled_write_bytes); | |
293eb1e7 VK |
2381 | out_unlock: |
2382 | mutex_unlock(&task->signal->cred_guard_mutex); | |
2383 | return result; | |
297c5d92 AR |
2384 | } |
2385 | ||
2386 | static int proc_tid_io_accounting(struct task_struct *task, char *buffer) | |
2387 | { | |
2388 | return do_io_accounting(task, buffer, 0); | |
aba76fdb | 2389 | } |
297c5d92 AR |
2390 | |
2391 | static int proc_tgid_io_accounting(struct task_struct *task, char *buffer) | |
2392 | { | |
2393 | return do_io_accounting(task, buffer, 1); | |
2394 | } | |
2395 | #endif /* CONFIG_TASK_IO_ACCOUNTING */ | |
aba76fdb | 2396 | |
22d917d8 EB |
2397 | #ifdef CONFIG_USER_NS |
2398 | static int proc_id_map_open(struct inode *inode, struct file *file, | |
2399 | struct seq_operations *seq_ops) | |
2400 | { | |
2401 | struct user_namespace *ns = NULL; | |
2402 | struct task_struct *task; | |
2403 | struct seq_file *seq; | |
2404 | int ret = -EINVAL; | |
2405 | ||
2406 | task = get_proc_task(inode); | |
2407 | if (task) { | |
2408 | rcu_read_lock(); | |
2409 | ns = get_user_ns(task_cred_xxx(task, user_ns)); | |
2410 | rcu_read_unlock(); | |
2411 | put_task_struct(task); | |
2412 | } | |
2413 | if (!ns) | |
2414 | goto err; | |
2415 | ||
2416 | ret = seq_open(file, seq_ops); | |
2417 | if (ret) | |
2418 | goto err_put_ns; | |
2419 | ||
2420 | seq = file->private_data; | |
2421 | seq->private = ns; | |
2422 | ||
2423 | return 0; | |
2424 | err_put_ns: | |
2425 | put_user_ns(ns); | |
2426 | err: | |
2427 | return ret; | |
2428 | } | |
2429 | ||
2430 | static int proc_id_map_release(struct inode *inode, struct file *file) | |
2431 | { | |
2432 | struct seq_file *seq = file->private_data; | |
2433 | struct user_namespace *ns = seq->private; | |
2434 | put_user_ns(ns); | |
2435 | return seq_release(inode, file); | |
2436 | } | |
2437 | ||
2438 | static int proc_uid_map_open(struct inode *inode, struct file *file) | |
2439 | { | |
2440 | return proc_id_map_open(inode, file, &proc_uid_seq_operations); | |
2441 | } | |
2442 | ||
2443 | static int proc_gid_map_open(struct inode *inode, struct file *file) | |
2444 | { | |
2445 | return proc_id_map_open(inode, file, &proc_gid_seq_operations); | |
2446 | } | |
2447 | ||
f76d207a EB |
2448 | static int proc_projid_map_open(struct inode *inode, struct file *file) |
2449 | { | |
2450 | return proc_id_map_open(inode, file, &proc_projid_seq_operations); | |
2451 | } | |
2452 | ||
22d917d8 EB |
2453 | static const struct file_operations proc_uid_map_operations = { |
2454 | .open = proc_uid_map_open, | |
2455 | .write = proc_uid_map_write, | |
2456 | .read = seq_read, | |
2457 | .llseek = seq_lseek, | |
2458 | .release = proc_id_map_release, | |
2459 | }; | |
2460 | ||
2461 | static const struct file_operations proc_gid_map_operations = { | |
2462 | .open = proc_gid_map_open, | |
2463 | .write = proc_gid_map_write, | |
2464 | .read = seq_read, | |
2465 | .llseek = seq_lseek, | |
2466 | .release = proc_id_map_release, | |
2467 | }; | |
f76d207a EB |
2468 | |
2469 | static const struct file_operations proc_projid_map_operations = { | |
2470 | .open = proc_projid_map_open, | |
2471 | .write = proc_projid_map_write, | |
2472 | .read = seq_read, | |
2473 | .llseek = seq_lseek, | |
2474 | .release = proc_id_map_release, | |
2475 | }; | |
22d917d8 EB |
2476 | #endif /* CONFIG_USER_NS */ |
2477 | ||
47830723 KC |
2478 | static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns, |
2479 | struct pid *pid, struct task_struct *task) | |
2480 | { | |
a9712bc1 AV |
2481 | int err = lock_trace(task); |
2482 | if (!err) { | |
2483 | seq_printf(m, "%08x\n", task->personality); | |
2484 | unlock_trace(task); | |
2485 | } | |
2486 | return err; | |
47830723 KC |
2487 | } |
2488 | ||
28a6d671 EB |
2489 | /* |
2490 | * Thread groups | |
2491 | */ | |
00977a59 | 2492 | static const struct file_operations proc_task_operations; |
c5ef1c42 | 2493 | static const struct inode_operations proc_task_inode_operations; |
20cdc894 | 2494 | |
c5141e6d | 2495 | static const struct pid_entry tgid_base_stuff[] = { |
631f9c18 AD |
2496 | DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations), |
2497 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), | |
640708a2 PE |
2498 | #ifdef CONFIG_CHECKPOINT_RESTORE |
2499 | DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations), | |
2500 | #endif | |
631f9c18 | 2501 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 2502 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
b2211a36 | 2503 | #ifdef CONFIG_NET |
631f9c18 | 2504 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), |
b2211a36 | 2505 | #endif |
631f9c18 AD |
2506 | REG("environ", S_IRUSR, proc_environ_operations), |
2507 | INF("auxv", S_IRUSR, proc_pid_auxv), | |
2508 | ONE("status", S_IRUGO, proc_pid_status), | |
a9712bc1 | 2509 | ONE("personality", S_IRUGO, proc_pid_personality), |
3036e7b4 | 2510 | INF("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 2511 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 2512 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
5091faa4 MG |
2513 | #endif |
2514 | #ifdef CONFIG_SCHED_AUTOGROUP | |
2515 | REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations), | |
ebcb6734 | 2516 | #endif |
4614a696 | 2517 | REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations), |
ebcb6734 | 2518 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
a9712bc1 | 2519 | INF("syscall", S_IRUGO, proc_pid_syscall), |
43ae34cb | 2520 | #endif |
631f9c18 AD |
2521 | INF("cmdline", S_IRUGO, proc_pid_cmdline), |
2522 | ONE("stat", S_IRUGO, proc_tgid_stat), | |
2523 | ONE("statm", S_IRUGO, proc_pid_statm), | |
b7643757 | 2524 | REG("maps", S_IRUGO, proc_pid_maps_operations), |
28a6d671 | 2525 | #ifdef CONFIG_NUMA |
b7643757 | 2526 | REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations), |
28a6d671 | 2527 | #endif |
631f9c18 AD |
2528 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
2529 | LNK("cwd", proc_cwd_link), | |
2530 | LNK("root", proc_root_link), | |
2531 | LNK("exe", proc_exe_link), | |
2532 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
2533 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
2534 | REG("mountstats", S_IRUSR, proc_mountstats_operations), | |
1e883281 | 2535 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 2536 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
b7643757 | 2537 | REG("smaps", S_IRUGO, proc_pid_smaps_operations), |
ca6b0bf0 | 2538 | REG("pagemap", S_IRUGO, proc_pagemap_operations), |
28a6d671 EB |
2539 | #endif |
2540 | #ifdef CONFIG_SECURITY | |
631f9c18 | 2541 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
2542 | #endif |
2543 | #ifdef CONFIG_KALLSYMS | |
631f9c18 | 2544 | INF("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 2545 | #endif |
2ec220e2 | 2546 | #ifdef CONFIG_STACKTRACE |
a9712bc1 | 2547 | ONE("stack", S_IRUGO, proc_pid_stack), |
28a6d671 EB |
2548 | #endif |
2549 | #ifdef CONFIG_SCHEDSTATS | |
631f9c18 | 2550 | INF("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 2551 | #endif |
9745512c | 2552 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 2553 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 2554 | #endif |
8793d854 | 2555 | #ifdef CONFIG_PROC_PID_CPUSET |
631f9c18 | 2556 | REG("cpuset", S_IRUGO, proc_cpuset_operations), |
a424316c PM |
2557 | #endif |
2558 | #ifdef CONFIG_CGROUPS | |
631f9c18 | 2559 | REG("cgroup", S_IRUGO, proc_cgroup_operations), |
28a6d671 | 2560 | #endif |
631f9c18 | 2561 | INF("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 2562 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 2563 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
28a6d671 | 2564 | #ifdef CONFIG_AUDITSYSCALL |
631f9c18 AD |
2565 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
2566 | REG("sessionid", S_IRUGO, proc_sessionid_operations), | |
28a6d671 | 2567 | #endif |
f4f154fd | 2568 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 2569 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
f4f154fd | 2570 | #endif |
698ba7b5 | 2571 | #ifdef CONFIG_ELF_CORE |
631f9c18 | 2572 | REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations), |
3cb4a0bb | 2573 | #endif |
aba76fdb | 2574 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
1d1221f3 | 2575 | INF("io", S_IRUSR, proc_tgid_io_accounting), |
aba76fdb | 2576 | #endif |
f133ecca CM |
2577 | #ifdef CONFIG_HARDWALL |
2578 | INF("hardwall", S_IRUGO, proc_pid_hardwall), | |
2579 | #endif | |
22d917d8 EB |
2580 | #ifdef CONFIG_USER_NS |
2581 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
2582 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 2583 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
22d917d8 | 2584 | #endif |
28a6d671 | 2585 | }; |
1da177e4 | 2586 | |
28a6d671 | 2587 | static int proc_tgid_base_readdir(struct file * filp, |
1da177e4 LT |
2588 | void * dirent, filldir_t filldir) |
2589 | { | |
2590 | return proc_pident_readdir(filp,dirent,filldir, | |
28a6d671 | 2591 | tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff)); |
1da177e4 LT |
2592 | } |
2593 | ||
00977a59 | 2594 | static const struct file_operations proc_tgid_base_operations = { |
1da177e4 | 2595 | .read = generic_read_dir, |
28a6d671 | 2596 | .readdir = proc_tgid_base_readdir, |
6038f373 | 2597 | .llseek = default_llseek, |
1da177e4 LT |
2598 | }; |
2599 | ||
00cd8dd3 AV |
2600 | static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
2601 | { | |
7bcd6b0e EB |
2602 | return proc_pident_lookup(dir, dentry, |
2603 | tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
2604 | } |
2605 | ||
c5ef1c42 | 2606 | static const struct inode_operations proc_tgid_base_inode_operations = { |
28a6d671 | 2607 | .lookup = proc_tgid_base_lookup, |
99f89551 | 2608 | .getattr = pid_getattr, |
6d76fa58 | 2609 | .setattr = proc_setattr, |
0499680a | 2610 | .permission = proc_pid_permission, |
1da177e4 | 2611 | }; |
1da177e4 | 2612 | |
60347f67 | 2613 | static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid) |
1da177e4 | 2614 | { |
48e6484d | 2615 | struct dentry *dentry, *leader, *dir; |
8578cea7 | 2616 | char buf[PROC_NUMBUF]; |
48e6484d EB |
2617 | struct qstr name; |
2618 | ||
2619 | name.name = buf; | |
60347f67 PE |
2620 | name.len = snprintf(buf, sizeof(buf), "%d", pid); |
2621 | dentry = d_hash_and_lookup(mnt->mnt_root, &name); | |
48e6484d | 2622 | if (dentry) { |
29f12ca3 | 2623 | shrink_dcache_parent(dentry); |
48e6484d EB |
2624 | d_drop(dentry); |
2625 | dput(dentry); | |
2626 | } | |
1da177e4 | 2627 | |
48e6484d | 2628 | name.name = buf; |
60347f67 PE |
2629 | name.len = snprintf(buf, sizeof(buf), "%d", tgid); |
2630 | leader = d_hash_and_lookup(mnt->mnt_root, &name); | |
48e6484d EB |
2631 | if (!leader) |
2632 | goto out; | |
1da177e4 | 2633 | |
48e6484d EB |
2634 | name.name = "task"; |
2635 | name.len = strlen(name.name); | |
2636 | dir = d_hash_and_lookup(leader, &name); | |
2637 | if (!dir) | |
2638 | goto out_put_leader; | |
2639 | ||
2640 | name.name = buf; | |
60347f67 | 2641 | name.len = snprintf(buf, sizeof(buf), "%d", pid); |
48e6484d EB |
2642 | dentry = d_hash_and_lookup(dir, &name); |
2643 | if (dentry) { | |
2644 | shrink_dcache_parent(dentry); | |
2645 | d_drop(dentry); | |
2646 | dput(dentry); | |
1da177e4 | 2647 | } |
48e6484d EB |
2648 | |
2649 | dput(dir); | |
2650 | out_put_leader: | |
2651 | dput(leader); | |
2652 | out: | |
2653 | return; | |
1da177e4 LT |
2654 | } |
2655 | ||
0895e91d RD |
2656 | /** |
2657 | * proc_flush_task - Remove dcache entries for @task from the /proc dcache. | |
2658 | * @task: task that should be flushed. | |
2659 | * | |
2660 | * When flushing dentries from proc, one needs to flush them from global | |
60347f67 | 2661 | * proc (proc_mnt) and from all the namespaces' procs this task was seen |
0895e91d RD |
2662 | * in. This call is supposed to do all of this job. |
2663 | * | |
2664 | * Looks in the dcache for | |
2665 | * /proc/@pid | |
2666 | * /proc/@tgid/task/@pid | |
2667 | * if either directory is present flushes it and all of it'ts children | |
2668 | * from the dcache. | |
2669 | * | |
2670 | * It is safe and reasonable to cache /proc entries for a task until | |
2671 | * that task exits. After that they just clog up the dcache with | |
2672 | * useless entries, possibly causing useful dcache entries to be | |
2673 | * flushed instead. This routine is proved to flush those useless | |
2674 | * dcache entries at process exit time. | |
2675 | * | |
2676 | * NOTE: This routine is just an optimization so it does not guarantee | |
2677 | * that no dcache entries will exist at process exit time it | |
2678 | * just makes it very unlikely that any will persist. | |
60347f67 PE |
2679 | */ |
2680 | ||
2681 | void proc_flush_task(struct task_struct *task) | |
2682 | { | |
9fcc2d15 | 2683 | int i; |
9b4d1cbe | 2684 | struct pid *pid, *tgid; |
130f77ec PE |
2685 | struct upid *upid; |
2686 | ||
130f77ec | 2687 | pid = task_pid(task); |
9b4d1cbe | 2688 | tgid = task_tgid(task); |
130f77ec | 2689 | |
9fcc2d15 | 2690 | for (i = 0; i <= pid->level; i++) { |
130f77ec PE |
2691 | upid = &pid->numbers[i]; |
2692 | proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr, | |
9b4d1cbe | 2693 | tgid->numbers[i].nr); |
130f77ec | 2694 | } |
60347f67 PE |
2695 | } |
2696 | ||
9711ef99 AB |
2697 | static struct dentry *proc_pid_instantiate(struct inode *dir, |
2698 | struct dentry * dentry, | |
c5141e6d | 2699 | struct task_struct *task, const void *ptr) |
444ceed8 EB |
2700 | { |
2701 | struct dentry *error = ERR_PTR(-ENOENT); | |
2702 | struct inode *inode; | |
2703 | ||
61a28784 | 2704 | inode = proc_pid_make_inode(dir->i_sb, task); |
444ceed8 EB |
2705 | if (!inode) |
2706 | goto out; | |
2707 | ||
2708 | inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO; | |
2709 | inode->i_op = &proc_tgid_base_inode_operations; | |
2710 | inode->i_fop = &proc_tgid_base_operations; | |
2711 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 | 2712 | |
bfe86848 MS |
2713 | set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff, |
2714 | ARRAY_SIZE(tgid_base_stuff))); | |
444ceed8 | 2715 | |
fb045adb | 2716 | d_set_d_op(dentry, &pid_dentry_operations); |
444ceed8 EB |
2717 | |
2718 | d_add(dentry, inode); | |
2719 | /* Close the race of the process dying before we return the dentry */ | |
0b728e19 | 2720 | if (pid_revalidate(dentry, 0)) |
444ceed8 EB |
2721 | error = NULL; |
2722 | out: | |
2723 | return error; | |
2724 | } | |
2725 | ||
00cd8dd3 | 2726 | struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags) |
1da177e4 | 2727 | { |
e656d8a6 | 2728 | struct dentry *result = NULL; |
1da177e4 | 2729 | struct task_struct *task; |
1da177e4 | 2730 | unsigned tgid; |
b488893a | 2731 | struct pid_namespace *ns; |
1da177e4 | 2732 | |
1da177e4 LT |
2733 | tgid = name_to_int(dentry); |
2734 | if (tgid == ~0U) | |
2735 | goto out; | |
2736 | ||
b488893a | 2737 | ns = dentry->d_sb->s_fs_info; |
de758734 | 2738 | rcu_read_lock(); |
b488893a | 2739 | task = find_task_by_pid_ns(tgid, ns); |
1da177e4 LT |
2740 | if (task) |
2741 | get_task_struct(task); | |
de758734 | 2742 | rcu_read_unlock(); |
1da177e4 LT |
2743 | if (!task) |
2744 | goto out; | |
2745 | ||
444ceed8 | 2746 | result = proc_pid_instantiate(dir, dentry, task, NULL); |
1da177e4 | 2747 | put_task_struct(task); |
1da177e4 | 2748 | out: |
cd6a3ce9 | 2749 | return result; |
1da177e4 LT |
2750 | } |
2751 | ||
1da177e4 | 2752 | /* |
0804ef4b | 2753 | * Find the first task with tgid >= tgid |
0bc58a91 | 2754 | * |
1da177e4 | 2755 | */ |
19fd4bb2 EB |
2756 | struct tgid_iter { |
2757 | unsigned int tgid; | |
0804ef4b | 2758 | struct task_struct *task; |
19fd4bb2 EB |
2759 | }; |
2760 | static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter) | |
2761 | { | |
0804ef4b | 2762 | struct pid *pid; |
1da177e4 | 2763 | |
19fd4bb2 EB |
2764 | if (iter.task) |
2765 | put_task_struct(iter.task); | |
454cc105 | 2766 | rcu_read_lock(); |
0804ef4b | 2767 | retry: |
19fd4bb2 EB |
2768 | iter.task = NULL; |
2769 | pid = find_ge_pid(iter.tgid, ns); | |
0804ef4b | 2770 | if (pid) { |
19fd4bb2 EB |
2771 | iter.tgid = pid_nr_ns(pid, ns); |
2772 | iter.task = pid_task(pid, PIDTYPE_PID); | |
0804ef4b EB |
2773 | /* What we to know is if the pid we have find is the |
2774 | * pid of a thread_group_leader. Testing for task | |
2775 | * being a thread_group_leader is the obvious thing | |
2776 | * todo but there is a window when it fails, due to | |
2777 | * the pid transfer logic in de_thread. | |
2778 | * | |
2779 | * So we perform the straight forward test of seeing | |
2780 | * if the pid we have found is the pid of a thread | |
2781 | * group leader, and don't worry if the task we have | |
2782 | * found doesn't happen to be a thread group leader. | |
2783 | * As we don't care in the case of readdir. | |
2784 | */ | |
19fd4bb2 EB |
2785 | if (!iter.task || !has_group_leader_pid(iter.task)) { |
2786 | iter.tgid += 1; | |
0804ef4b | 2787 | goto retry; |
19fd4bb2 EB |
2788 | } |
2789 | get_task_struct(iter.task); | |
0bc58a91 | 2790 | } |
454cc105 | 2791 | rcu_read_unlock(); |
19fd4bb2 | 2792 | return iter; |
1da177e4 LT |
2793 | } |
2794 | ||
e656d8a6 | 2795 | #define TGID_OFFSET (FIRST_PROCESS_ENTRY) |
0804ef4b | 2796 | |
61a28784 | 2797 | static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir, |
19fd4bb2 | 2798 | struct tgid_iter iter) |
61a28784 EB |
2799 | { |
2800 | char name[PROC_NUMBUF]; | |
19fd4bb2 | 2801 | int len = snprintf(name, sizeof(name), "%d", iter.tgid); |
61a28784 | 2802 | return proc_fill_cache(filp, dirent, filldir, name, len, |
19fd4bb2 | 2803 | proc_pid_instantiate, iter.task, NULL); |
61a28784 EB |
2804 | } |
2805 | ||
0499680a VK |
2806 | static int fake_filldir(void *buf, const char *name, int namelen, |
2807 | loff_t offset, u64 ino, unsigned d_type) | |
2808 | { | |
2809 | return 0; | |
2810 | } | |
2811 | ||
1da177e4 LT |
2812 | /* for the /proc/ directory itself, after non-process stuff has been done */ |
2813 | int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir) | |
2814 | { | |
19fd4bb2 | 2815 | struct tgid_iter iter; |
b488893a | 2816 | struct pid_namespace *ns; |
0499680a | 2817 | filldir_t __filldir; |
1da177e4 | 2818 | |
d8bdc59f | 2819 | if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET) |
ae06c7c8 | 2820 | goto out; |
1da177e4 | 2821 | |
b488893a | 2822 | ns = filp->f_dentry->d_sb->s_fs_info; |
19fd4bb2 EB |
2823 | iter.task = NULL; |
2824 | iter.tgid = filp->f_pos - TGID_OFFSET; | |
2825 | for (iter = next_tgid(ns, iter); | |
2826 | iter.task; | |
2827 | iter.tgid += 1, iter = next_tgid(ns, iter)) { | |
0499680a VK |
2828 | if (has_pid_permissions(ns, iter.task, 2)) |
2829 | __filldir = filldir; | |
2830 | else | |
2831 | __filldir = fake_filldir; | |
2832 | ||
19fd4bb2 | 2833 | filp->f_pos = iter.tgid + TGID_OFFSET; |
0499680a | 2834 | if (proc_pid_fill_cache(filp, dirent, __filldir, iter) < 0) { |
19fd4bb2 | 2835 | put_task_struct(iter.task); |
0804ef4b | 2836 | goto out; |
1da177e4 | 2837 | } |
0bc58a91 | 2838 | } |
0804ef4b EB |
2839 | filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET; |
2840 | out: | |
0bc58a91 EB |
2841 | return 0; |
2842 | } | |
1da177e4 | 2843 | |
28a6d671 EB |
2844 | /* |
2845 | * Tasks | |
2846 | */ | |
c5141e6d | 2847 | static const struct pid_entry tid_base_stuff[] = { |
631f9c18 | 2848 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), |
3835541d | 2849 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 2850 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
631f9c18 AD |
2851 | REG("environ", S_IRUSR, proc_environ_operations), |
2852 | INF("auxv", S_IRUSR, proc_pid_auxv), | |
2853 | ONE("status", S_IRUGO, proc_pid_status), | |
a9712bc1 | 2854 | ONE("personality", S_IRUGO, proc_pid_personality), |
3036e7b4 | 2855 | INF("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 2856 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 2857 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
ebcb6734 | 2858 | #endif |
4614a696 | 2859 | REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations), |
ebcb6734 | 2860 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
a9712bc1 | 2861 | INF("syscall", S_IRUGO, proc_pid_syscall), |
43ae34cb | 2862 | #endif |
631f9c18 AD |
2863 | INF("cmdline", S_IRUGO, proc_pid_cmdline), |
2864 | ONE("stat", S_IRUGO, proc_tid_stat), | |
2865 | ONE("statm", S_IRUGO, proc_pid_statm), | |
b7643757 | 2866 | REG("maps", S_IRUGO, proc_tid_maps_operations), |
81841161 CG |
2867 | #ifdef CONFIG_CHECKPOINT_RESTORE |
2868 | REG("children", S_IRUGO, proc_tid_children_operations), | |
2869 | #endif | |
28a6d671 | 2870 | #ifdef CONFIG_NUMA |
b7643757 | 2871 | REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations), |
28a6d671 | 2872 | #endif |
631f9c18 AD |
2873 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
2874 | LNK("cwd", proc_cwd_link), | |
2875 | LNK("root", proc_root_link), | |
2876 | LNK("exe", proc_exe_link), | |
2877 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
2878 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
1e883281 | 2879 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 2880 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
b7643757 | 2881 | REG("smaps", S_IRUGO, proc_tid_smaps_operations), |
ca6b0bf0 | 2882 | REG("pagemap", S_IRUGO, proc_pagemap_operations), |
28a6d671 EB |
2883 | #endif |
2884 | #ifdef CONFIG_SECURITY | |
631f9c18 | 2885 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
2886 | #endif |
2887 | #ifdef CONFIG_KALLSYMS | |
631f9c18 | 2888 | INF("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 2889 | #endif |
2ec220e2 | 2890 | #ifdef CONFIG_STACKTRACE |
a9712bc1 | 2891 | ONE("stack", S_IRUGO, proc_pid_stack), |
28a6d671 EB |
2892 | #endif |
2893 | #ifdef CONFIG_SCHEDSTATS | |
631f9c18 | 2894 | INF("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 2895 | #endif |
9745512c | 2896 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 2897 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 2898 | #endif |
8793d854 | 2899 | #ifdef CONFIG_PROC_PID_CPUSET |
631f9c18 | 2900 | REG("cpuset", S_IRUGO, proc_cpuset_operations), |
a424316c PM |
2901 | #endif |
2902 | #ifdef CONFIG_CGROUPS | |
631f9c18 | 2903 | REG("cgroup", S_IRUGO, proc_cgroup_operations), |
28a6d671 | 2904 | #endif |
631f9c18 | 2905 | INF("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 2906 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 2907 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
28a6d671 | 2908 | #ifdef CONFIG_AUDITSYSCALL |
631f9c18 | 2909 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
26ec3c64 | 2910 | REG("sessionid", S_IRUGO, proc_sessionid_operations), |
28a6d671 | 2911 | #endif |
f4f154fd | 2912 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 2913 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
f4f154fd | 2914 | #endif |
297c5d92 | 2915 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
1d1221f3 | 2916 | INF("io", S_IRUSR, proc_tid_io_accounting), |
297c5d92 | 2917 | #endif |
f133ecca CM |
2918 | #ifdef CONFIG_HARDWALL |
2919 | INF("hardwall", S_IRUGO, proc_pid_hardwall), | |
2920 | #endif | |
22d917d8 EB |
2921 | #ifdef CONFIG_USER_NS |
2922 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
2923 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 2924 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
22d917d8 | 2925 | #endif |
28a6d671 EB |
2926 | }; |
2927 | ||
2928 | static int proc_tid_base_readdir(struct file * filp, | |
2929 | void * dirent, filldir_t filldir) | |
2930 | { | |
2931 | return proc_pident_readdir(filp,dirent,filldir, | |
2932 | tid_base_stuff,ARRAY_SIZE(tid_base_stuff)); | |
2933 | } | |
2934 | ||
00cd8dd3 AV |
2935 | static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
2936 | { | |
7bcd6b0e EB |
2937 | return proc_pident_lookup(dir, dentry, |
2938 | tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
2939 | } |
2940 | ||
00977a59 | 2941 | static const struct file_operations proc_tid_base_operations = { |
28a6d671 EB |
2942 | .read = generic_read_dir, |
2943 | .readdir = proc_tid_base_readdir, | |
6038f373 | 2944 | .llseek = default_llseek, |
28a6d671 EB |
2945 | }; |
2946 | ||
c5ef1c42 | 2947 | static const struct inode_operations proc_tid_base_inode_operations = { |
28a6d671 EB |
2948 | .lookup = proc_tid_base_lookup, |
2949 | .getattr = pid_getattr, | |
2950 | .setattr = proc_setattr, | |
2951 | }; | |
2952 | ||
444ceed8 | 2953 | static struct dentry *proc_task_instantiate(struct inode *dir, |
c5141e6d | 2954 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
444ceed8 EB |
2955 | { |
2956 | struct dentry *error = ERR_PTR(-ENOENT); | |
2957 | struct inode *inode; | |
61a28784 | 2958 | inode = proc_pid_make_inode(dir->i_sb, task); |
444ceed8 EB |
2959 | |
2960 | if (!inode) | |
2961 | goto out; | |
2962 | inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO; | |
2963 | inode->i_op = &proc_tid_base_inode_operations; | |
2964 | inode->i_fop = &proc_tid_base_operations; | |
2965 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 | 2966 | |
bfe86848 MS |
2967 | set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff, |
2968 | ARRAY_SIZE(tid_base_stuff))); | |
444ceed8 | 2969 | |
fb045adb | 2970 | d_set_d_op(dentry, &pid_dentry_operations); |
444ceed8 EB |
2971 | |
2972 | d_add(dentry, inode); | |
2973 | /* Close the race of the process dying before we return the dentry */ | |
0b728e19 | 2974 | if (pid_revalidate(dentry, 0)) |
444ceed8 EB |
2975 | error = NULL; |
2976 | out: | |
2977 | return error; | |
2978 | } | |
2979 | ||
00cd8dd3 | 2980 | static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags) |
28a6d671 EB |
2981 | { |
2982 | struct dentry *result = ERR_PTR(-ENOENT); | |
2983 | struct task_struct *task; | |
2984 | struct task_struct *leader = get_proc_task(dir); | |
28a6d671 | 2985 | unsigned tid; |
b488893a | 2986 | struct pid_namespace *ns; |
28a6d671 EB |
2987 | |
2988 | if (!leader) | |
2989 | goto out_no_task; | |
2990 | ||
2991 | tid = name_to_int(dentry); | |
2992 | if (tid == ~0U) | |
2993 | goto out; | |
2994 | ||
b488893a | 2995 | ns = dentry->d_sb->s_fs_info; |
28a6d671 | 2996 | rcu_read_lock(); |
b488893a | 2997 | task = find_task_by_pid_ns(tid, ns); |
28a6d671 EB |
2998 | if (task) |
2999 | get_task_struct(task); | |
3000 | rcu_read_unlock(); | |
3001 | if (!task) | |
3002 | goto out; | |
bac0abd6 | 3003 | if (!same_thread_group(leader, task)) |
28a6d671 EB |
3004 | goto out_drop_task; |
3005 | ||
444ceed8 | 3006 | result = proc_task_instantiate(dir, dentry, task, NULL); |
28a6d671 EB |
3007 | out_drop_task: |
3008 | put_task_struct(task); | |
3009 | out: | |
3010 | put_task_struct(leader); | |
3011 | out_no_task: | |
3012 | return result; | |
3013 | } | |
3014 | ||
0bc58a91 EB |
3015 | /* |
3016 | * Find the first tid of a thread group to return to user space. | |
3017 | * | |
3018 | * Usually this is just the thread group leader, but if the users | |
3019 | * buffer was too small or there was a seek into the middle of the | |
3020 | * directory we have more work todo. | |
3021 | * | |
3022 | * In the case of a short read we start with find_task_by_pid. | |
3023 | * | |
3024 | * In the case of a seek we start with the leader and walk nr | |
3025 | * threads past it. | |
3026 | */ | |
cc288738 | 3027 | static struct task_struct *first_tid(struct task_struct *leader, |
b488893a | 3028 | int tid, int nr, struct pid_namespace *ns) |
0bc58a91 | 3029 | { |
a872ff0c | 3030 | struct task_struct *pos; |
1da177e4 | 3031 | |
cc288738 | 3032 | rcu_read_lock(); |
0bc58a91 EB |
3033 | /* Attempt to start with the pid of a thread */ |
3034 | if (tid && (nr > 0)) { | |
b488893a | 3035 | pos = find_task_by_pid_ns(tid, ns); |
a872ff0c ON |
3036 | if (pos && (pos->group_leader == leader)) |
3037 | goto found; | |
0bc58a91 | 3038 | } |
1da177e4 | 3039 | |
0bc58a91 | 3040 | /* If nr exceeds the number of threads there is nothing todo */ |
a872ff0c ON |
3041 | pos = NULL; |
3042 | if (nr && nr >= get_nr_threads(leader)) | |
3043 | goto out; | |
1da177e4 | 3044 | |
a872ff0c ON |
3045 | /* If we haven't found our starting place yet start |
3046 | * with the leader and walk nr threads forward. | |
0bc58a91 | 3047 | */ |
a872ff0c ON |
3048 | for (pos = leader; nr > 0; --nr) { |
3049 | pos = next_thread(pos); | |
3050 | if (pos == leader) { | |
3051 | pos = NULL; | |
3052 | goto out; | |
3053 | } | |
1da177e4 | 3054 | } |
a872ff0c ON |
3055 | found: |
3056 | get_task_struct(pos); | |
3057 | out: | |
cc288738 | 3058 | rcu_read_unlock(); |
0bc58a91 EB |
3059 | return pos; |
3060 | } | |
3061 | ||
3062 | /* | |
3063 | * Find the next thread in the thread list. | |
3064 | * Return NULL if there is an error or no next thread. | |
3065 | * | |
3066 | * The reference to the input task_struct is released. | |
3067 | */ | |
3068 | static struct task_struct *next_tid(struct task_struct *start) | |
3069 | { | |
c1df7fb8 | 3070 | struct task_struct *pos = NULL; |
cc288738 | 3071 | rcu_read_lock(); |
c1df7fb8 | 3072 | if (pid_alive(start)) { |
0bc58a91 | 3073 | pos = next_thread(start); |
c1df7fb8 ON |
3074 | if (thread_group_leader(pos)) |
3075 | pos = NULL; | |
3076 | else | |
3077 | get_task_struct(pos); | |
3078 | } | |
cc288738 | 3079 | rcu_read_unlock(); |
0bc58a91 EB |
3080 | put_task_struct(start); |
3081 | return pos; | |
1da177e4 LT |
3082 | } |
3083 | ||
61a28784 EB |
3084 | static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir, |
3085 | struct task_struct *task, int tid) | |
3086 | { | |
3087 | char name[PROC_NUMBUF]; | |
3088 | int len = snprintf(name, sizeof(name), "%d", tid); | |
3089 | return proc_fill_cache(filp, dirent, filldir, name, len, | |
3090 | proc_task_instantiate, task, NULL); | |
3091 | } | |
3092 | ||
1da177e4 LT |
3093 | /* for the /proc/TGID/task/ directories */ |
3094 | static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir) | |
3095 | { | |
2fddfeef | 3096 | struct dentry *dentry = filp->f_path.dentry; |
1da177e4 | 3097 | struct inode *inode = dentry->d_inode; |
7d895244 | 3098 | struct task_struct *leader = NULL; |
0bc58a91 | 3099 | struct task_struct *task; |
1da177e4 LT |
3100 | int retval = -ENOENT; |
3101 | ino_t ino; | |
0bc58a91 | 3102 | int tid; |
b488893a | 3103 | struct pid_namespace *ns; |
1da177e4 | 3104 | |
7d895244 GC |
3105 | task = get_proc_task(inode); |
3106 | if (!task) | |
3107 | goto out_no_task; | |
3108 | rcu_read_lock(); | |
3109 | if (pid_alive(task)) { | |
3110 | leader = task->group_leader; | |
3111 | get_task_struct(leader); | |
3112 | } | |
3113 | rcu_read_unlock(); | |
3114 | put_task_struct(task); | |
99f89551 EB |
3115 | if (!leader) |
3116 | goto out_no_task; | |
1da177e4 LT |
3117 | retval = 0; |
3118 | ||
ee568b25 | 3119 | switch ((unsigned long)filp->f_pos) { |
1da177e4 LT |
3120 | case 0: |
3121 | ino = inode->i_ino; | |
ee6f779b | 3122 | if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0) |
1da177e4 | 3123 | goto out; |
ee6f779b | 3124 | filp->f_pos++; |
1da177e4 LT |
3125 | /* fall through */ |
3126 | case 1: | |
3127 | ino = parent_ino(dentry); | |
ee6f779b | 3128 | if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0) |
1da177e4 | 3129 | goto out; |
ee6f779b | 3130 | filp->f_pos++; |
1da177e4 LT |
3131 | /* fall through */ |
3132 | } | |
3133 | ||
0bc58a91 EB |
3134 | /* f_version caches the tgid value that the last readdir call couldn't |
3135 | * return. lseek aka telldir automagically resets f_version to 0. | |
3136 | */ | |
b488893a | 3137 | ns = filp->f_dentry->d_sb->s_fs_info; |
2b47c361 | 3138 | tid = (int)filp->f_version; |
0bc58a91 | 3139 | filp->f_version = 0; |
ee6f779b | 3140 | for (task = first_tid(leader, tid, filp->f_pos - 2, ns); |
0bc58a91 | 3141 | task; |
ee6f779b | 3142 | task = next_tid(task), filp->f_pos++) { |
b488893a | 3143 | tid = task_pid_nr_ns(task, ns); |
61a28784 | 3144 | if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) { |
0bc58a91 EB |
3145 | /* returning this tgid failed, save it as the first |
3146 | * pid for the next readir call */ | |
2b47c361 | 3147 | filp->f_version = (u64)tid; |
0bc58a91 | 3148 | put_task_struct(task); |
1da177e4 | 3149 | break; |
0bc58a91 | 3150 | } |
1da177e4 LT |
3151 | } |
3152 | out: | |
99f89551 EB |
3153 | put_task_struct(leader); |
3154 | out_no_task: | |
1da177e4 LT |
3155 | return retval; |
3156 | } | |
6e66b52b EB |
3157 | |
3158 | static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat) | |
3159 | { | |
3160 | struct inode *inode = dentry->d_inode; | |
99f89551 | 3161 | struct task_struct *p = get_proc_task(inode); |
6e66b52b EB |
3162 | generic_fillattr(inode, stat); |
3163 | ||
99f89551 | 3164 | if (p) { |
99f89551 | 3165 | stat->nlink += get_nr_threads(p); |
99f89551 | 3166 | put_task_struct(p); |
6e66b52b EB |
3167 | } |
3168 | ||
3169 | return 0; | |
3170 | } | |
28a6d671 | 3171 | |
c5ef1c42 | 3172 | static const struct inode_operations proc_task_inode_operations = { |
28a6d671 EB |
3173 | .lookup = proc_task_lookup, |
3174 | .getattr = proc_task_getattr, | |
3175 | .setattr = proc_setattr, | |
0499680a | 3176 | .permission = proc_pid_permission, |
28a6d671 EB |
3177 | }; |
3178 | ||
00977a59 | 3179 | static const struct file_operations proc_task_operations = { |
28a6d671 EB |
3180 | .read = generic_read_dir, |
3181 | .readdir = proc_task_readdir, | |
6038f373 | 3182 | .llseek = default_llseek, |
28a6d671 | 3183 | }; |