]>
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> |
b835996f | 66 | #include <linux/rcupdate.h> |
1da177e4 | 67 | #include <linux/kallsyms.h> |
2ec220e2 | 68 | #include <linux/stacktrace.h> |
d85f50d5 | 69 | #include <linux/resource.h> |
5096add8 | 70 | #include <linux/module.h> |
1da177e4 LT |
71 | #include <linux/mount.h> |
72 | #include <linux/security.h> | |
73 | #include <linux/ptrace.h> | |
0d094efe | 74 | #include <linux/tracehook.h> |
a424316c | 75 | #include <linux/cgroup.h> |
1da177e4 LT |
76 | #include <linux/cpuset.h> |
77 | #include <linux/audit.h> | |
5addc5dd | 78 | #include <linux/poll.h> |
1651e14e | 79 | #include <linux/nsproxy.h> |
8ac773b4 | 80 | #include <linux/oom.h> |
3cb4a0bb | 81 | #include <linux/elf.h> |
60347f67 | 82 | #include <linux/pid_namespace.h> |
5ad4e53b | 83 | #include <linux/fs_struct.h> |
1da177e4 LT |
84 | #include "internal.h" |
85 | ||
0f2fe20f EB |
86 | /* NOTE: |
87 | * Implementing inode permission operations in /proc is almost | |
88 | * certainly an error. Permission checks need to happen during | |
89 | * each system call not at open time. The reason is that most of | |
90 | * what we wish to check for permissions in /proc varies at runtime. | |
91 | * | |
92 | * The classic example of a problem is opening file descriptors | |
93 | * in /proc for a task before it execs a suid executable. | |
94 | */ | |
95 | ||
1da177e4 | 96 | struct pid_entry { |
1da177e4 | 97 | char *name; |
c5141e6d | 98 | int len; |
1da177e4 | 99 | mode_t mode; |
c5ef1c42 | 100 | const struct inode_operations *iop; |
00977a59 | 101 | const struct file_operations *fop; |
20cdc894 | 102 | union proc_op op; |
1da177e4 LT |
103 | }; |
104 | ||
61a28784 | 105 | #define NOD(NAME, MODE, IOP, FOP, OP) { \ |
20cdc894 | 106 | .name = (NAME), \ |
c5141e6d | 107 | .len = sizeof(NAME) - 1, \ |
20cdc894 EB |
108 | .mode = MODE, \ |
109 | .iop = IOP, \ | |
110 | .fop = FOP, \ | |
111 | .op = OP, \ | |
112 | } | |
113 | ||
631f9c18 AD |
114 | #define DIR(NAME, MODE, iops, fops) \ |
115 | NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} ) | |
116 | #define LNK(NAME, get_link) \ | |
61a28784 | 117 | NOD(NAME, (S_IFLNK|S_IRWXUGO), \ |
20cdc894 | 118 | &proc_pid_link_inode_operations, NULL, \ |
631f9c18 AD |
119 | { .proc_get_link = get_link } ) |
120 | #define REG(NAME, MODE, fops) \ | |
121 | NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {}) | |
122 | #define INF(NAME, MODE, read) \ | |
61a28784 | 123 | NOD(NAME, (S_IFREG|(MODE)), \ |
20cdc894 | 124 | NULL, &proc_info_file_operations, \ |
631f9c18 AD |
125 | { .proc_read = read } ) |
126 | #define ONE(NAME, MODE, show) \ | |
be614086 EB |
127 | NOD(NAME, (S_IFREG|(MODE)), \ |
128 | NULL, &proc_single_file_operations, \ | |
631f9c18 | 129 | { .proc_show = show } ) |
1da177e4 | 130 | |
aed54175 VN |
131 | /* |
132 | * Count the number of hardlinks for the pid_entry table, excluding the . | |
133 | * and .. links. | |
134 | */ | |
135 | static unsigned int pid_entry_count_dirs(const struct pid_entry *entries, | |
136 | unsigned int n) | |
137 | { | |
138 | unsigned int i; | |
139 | unsigned int count; | |
140 | ||
141 | count = 0; | |
142 | for (i = 0; i < n; ++i) { | |
143 | if (S_ISDIR(entries[i].mode)) | |
144 | ++count; | |
145 | } | |
146 | ||
147 | return count; | |
148 | } | |
149 | ||
7c2c7d99 | 150 | static int get_fs_path(struct task_struct *task, struct path *path, bool root) |
1da177e4 LT |
151 | { |
152 | struct fs_struct *fs; | |
7c2c7d99 HD |
153 | int result = -ENOENT; |
154 | ||
0494f6ec MS |
155 | task_lock(task); |
156 | fs = task->fs; | |
7c2c7d99 HD |
157 | if (fs) { |
158 | read_lock(&fs->lock); | |
159 | *path = root ? fs->root : fs->pwd; | |
160 | path_get(path); | |
161 | read_unlock(&fs->lock); | |
162 | result = 0; | |
163 | } | |
0494f6ec | 164 | task_unlock(task); |
7c2c7d99 | 165 | return result; |
0494f6ec MS |
166 | } |
167 | ||
99f89551 EB |
168 | static int get_nr_threads(struct task_struct *tsk) |
169 | { | |
99f89551 EB |
170 | unsigned long flags; |
171 | int count = 0; | |
172 | ||
173 | if (lock_task_sighand(tsk, &flags)) { | |
174 | count = atomic_read(&tsk->signal->count); | |
175 | unlock_task_sighand(tsk, &flags); | |
176 | } | |
177 | return count; | |
178 | } | |
179 | ||
3dcd25f3 | 180 | static int proc_cwd_link(struct inode *inode, struct path *path) |
0494f6ec | 181 | { |
99f89551 | 182 | struct task_struct *task = get_proc_task(inode); |
0494f6ec | 183 | int result = -ENOENT; |
99f89551 EB |
184 | |
185 | if (task) { | |
7c2c7d99 | 186 | result = get_fs_path(task, path, 0); |
99f89551 EB |
187 | put_task_struct(task); |
188 | } | |
1da177e4 LT |
189 | return result; |
190 | } | |
191 | ||
3dcd25f3 | 192 | static int proc_root_link(struct inode *inode, struct path *path) |
1da177e4 | 193 | { |
99f89551 | 194 | struct task_struct *task = get_proc_task(inode); |
1da177e4 | 195 | int result = -ENOENT; |
99f89551 EB |
196 | |
197 | if (task) { | |
7c2c7d99 | 198 | result = get_fs_path(task, path, 1); |
99f89551 EB |
199 | put_task_struct(task); |
200 | } | |
1da177e4 LT |
201 | return result; |
202 | } | |
203 | ||
638fa202 RM |
204 | /* |
205 | * Return zero if current may access user memory in @task, -error if not. | |
206 | */ | |
207 | static int check_mem_permission(struct task_struct *task) | |
208 | { | |
209 | /* | |
210 | * A task can always look at itself, in case it chooses | |
211 | * to use system calls instead of load instructions. | |
212 | */ | |
213 | if (task == current) | |
214 | return 0; | |
215 | ||
216 | /* | |
217 | * If current is actively ptrace'ing, and would also be | |
218 | * permitted to freshly attach with ptrace now, permit it. | |
219 | */ | |
0d094efe RM |
220 | if (task_is_stopped_or_traced(task)) { |
221 | int match; | |
222 | rcu_read_lock(); | |
223 | match = (tracehook_tracer_task(task) == current); | |
224 | rcu_read_unlock(); | |
225 | if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH)) | |
226 | return 0; | |
227 | } | |
638fa202 RM |
228 | |
229 | /* | |
230 | * Noone else is allowed. | |
231 | */ | |
232 | return -EPERM; | |
233 | } | |
1da177e4 | 234 | |
831830b5 AV |
235 | struct mm_struct *mm_for_maps(struct task_struct *task) |
236 | { | |
704b836c ON |
237 | struct mm_struct *mm; |
238 | ||
239 | if (mutex_lock_killable(&task->cred_guard_mutex)) | |
240 | return NULL; | |
00f89d21 | 241 | |
704b836c ON |
242 | mm = get_task_mm(task); |
243 | if (mm && mm != current->mm && | |
244 | !ptrace_may_access(task, PTRACE_MODE_READ)) { | |
245 | mmput(mm); | |
246 | mm = NULL; | |
13f0feaf | 247 | } |
704b836c ON |
248 | mutex_unlock(&task->cred_guard_mutex); |
249 | ||
831830b5 | 250 | return mm; |
831830b5 AV |
251 | } |
252 | ||
1da177e4 LT |
253 | static int proc_pid_cmdline(struct task_struct *task, char * buffer) |
254 | { | |
255 | int res = 0; | |
256 | unsigned int len; | |
257 | struct mm_struct *mm = get_task_mm(task); | |
258 | if (!mm) | |
259 | goto out; | |
260 | if (!mm->arg_end) | |
261 | goto out_mm; /* Shh! No looking before we're done */ | |
262 | ||
263 | len = mm->arg_end - mm->arg_start; | |
264 | ||
265 | if (len > PAGE_SIZE) | |
266 | len = PAGE_SIZE; | |
267 | ||
268 | res = access_process_vm(task, mm->arg_start, buffer, len, 0); | |
269 | ||
270 | // If the nul at the end of args has been overwritten, then | |
271 | // assume application is using setproctitle(3). | |
272 | if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) { | |
273 | len = strnlen(buffer, res); | |
274 | if (len < res) { | |
275 | res = len; | |
276 | } else { | |
277 | len = mm->env_end - mm->env_start; | |
278 | if (len > PAGE_SIZE - res) | |
279 | len = PAGE_SIZE - res; | |
280 | res += access_process_vm(task, mm->env_start, buffer+res, len, 0); | |
281 | res = strnlen(buffer, res); | |
282 | } | |
283 | } | |
284 | out_mm: | |
285 | mmput(mm); | |
286 | out: | |
287 | return res; | |
288 | } | |
289 | ||
290 | static int proc_pid_auxv(struct task_struct *task, char *buffer) | |
291 | { | |
292 | int res = 0; | |
293 | struct mm_struct *mm = get_task_mm(task); | |
294 | if (mm) { | |
295 | unsigned int nwords = 0; | |
dfe6b7d9 | 296 | do { |
1da177e4 | 297 | nwords += 2; |
dfe6b7d9 | 298 | } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */ |
1da177e4 LT |
299 | res = nwords * sizeof(mm->saved_auxv[0]); |
300 | if (res > PAGE_SIZE) | |
301 | res = PAGE_SIZE; | |
302 | memcpy(buffer, mm->saved_auxv, res); | |
303 | mmput(mm); | |
304 | } | |
305 | return res; | |
306 | } | |
307 | ||
308 | ||
309 | #ifdef CONFIG_KALLSYMS | |
310 | /* | |
311 | * Provides a wchan file via kallsyms in a proper one-value-per-file format. | |
312 | * Returns the resolved symbol. If that fails, simply return the address. | |
313 | */ | |
314 | static int proc_pid_wchan(struct task_struct *task, char *buffer) | |
315 | { | |
ffb45122 | 316 | unsigned long wchan; |
9281acea | 317 | char symname[KSYM_NAME_LEN]; |
1da177e4 LT |
318 | |
319 | wchan = get_wchan(task); | |
320 | ||
9d65cb4a | 321 | if (lookup_symbol_name(wchan, symname) < 0) |
f83ce3e6 JE |
322 | if (!ptrace_may_access(task, PTRACE_MODE_READ)) |
323 | return 0; | |
324 | else | |
325 | return sprintf(buffer, "%lu", wchan); | |
9d65cb4a AD |
326 | else |
327 | return sprintf(buffer, "%s", symname); | |
1da177e4 LT |
328 | } |
329 | #endif /* CONFIG_KALLSYMS */ | |
330 | ||
2ec220e2 KC |
331 | #ifdef CONFIG_STACKTRACE |
332 | ||
333 | #define MAX_STACK_TRACE_DEPTH 64 | |
334 | ||
335 | static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns, | |
336 | struct pid *pid, struct task_struct *task) | |
337 | { | |
338 | struct stack_trace trace; | |
339 | unsigned long *entries; | |
340 | int i; | |
341 | ||
342 | entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL); | |
343 | if (!entries) | |
344 | return -ENOMEM; | |
345 | ||
346 | trace.nr_entries = 0; | |
347 | trace.max_entries = MAX_STACK_TRACE_DEPTH; | |
348 | trace.entries = entries; | |
349 | trace.skip = 0; | |
350 | save_stack_trace_tsk(task, &trace); | |
351 | ||
352 | for (i = 0; i < trace.nr_entries; i++) { | |
353 | seq_printf(m, "[<%p>] %pS\n", | |
354 | (void *)entries[i], (void *)entries[i]); | |
355 | } | |
356 | kfree(entries); | |
357 | ||
358 | return 0; | |
359 | } | |
360 | #endif | |
361 | ||
1da177e4 LT |
362 | #ifdef CONFIG_SCHEDSTATS |
363 | /* | |
364 | * Provides /proc/PID/schedstat | |
365 | */ | |
366 | static int proc_pid_schedstat(struct task_struct *task, char *buffer) | |
367 | { | |
172ba844 | 368 | return sprintf(buffer, "%llu %llu %lu\n", |
826e08b0 IM |
369 | (unsigned long long)task->se.sum_exec_runtime, |
370 | (unsigned long long)task->sched_info.run_delay, | |
2d72376b | 371 | task->sched_info.pcount); |
1da177e4 LT |
372 | } |
373 | #endif | |
374 | ||
9745512c AV |
375 | #ifdef CONFIG_LATENCYTOP |
376 | static int lstats_show_proc(struct seq_file *m, void *v) | |
377 | { | |
378 | int i; | |
13d77c37 HS |
379 | struct inode *inode = m->private; |
380 | struct task_struct *task = get_proc_task(inode); | |
9745512c | 381 | |
13d77c37 HS |
382 | if (!task) |
383 | return -ESRCH; | |
384 | seq_puts(m, "Latency Top version : v0.1\n"); | |
9745512c AV |
385 | for (i = 0; i < 32; i++) { |
386 | if (task->latency_record[i].backtrace[0]) { | |
387 | int q; | |
388 | seq_printf(m, "%i %li %li ", | |
389 | task->latency_record[i].count, | |
390 | task->latency_record[i].time, | |
391 | task->latency_record[i].max); | |
392 | for (q = 0; q < LT_BACKTRACEDEPTH; q++) { | |
9c246247 | 393 | char sym[KSYM_SYMBOL_LEN]; |
9745512c AV |
394 | char *c; |
395 | if (!task->latency_record[i].backtrace[q]) | |
396 | break; | |
397 | if (task->latency_record[i].backtrace[q] == ULONG_MAX) | |
398 | break; | |
399 | sprint_symbol(sym, task->latency_record[i].backtrace[q]); | |
400 | c = strchr(sym, '+'); | |
401 | if (c) | |
402 | *c = 0; | |
403 | seq_printf(m, "%s ", sym); | |
404 | } | |
405 | seq_printf(m, "\n"); | |
406 | } | |
407 | ||
408 | } | |
13d77c37 | 409 | put_task_struct(task); |
9745512c AV |
410 | return 0; |
411 | } | |
412 | ||
413 | static int lstats_open(struct inode *inode, struct file *file) | |
414 | { | |
13d77c37 | 415 | return single_open(file, lstats_show_proc, inode); |
d6643d12 HS |
416 | } |
417 | ||
9745512c AV |
418 | static ssize_t lstats_write(struct file *file, const char __user *buf, |
419 | size_t count, loff_t *offs) | |
420 | { | |
13d77c37 | 421 | struct task_struct *task = get_proc_task(file->f_dentry->d_inode); |
9745512c | 422 | |
13d77c37 HS |
423 | if (!task) |
424 | return -ESRCH; | |
9745512c | 425 | clear_all_latency_tracing(task); |
13d77c37 | 426 | put_task_struct(task); |
9745512c AV |
427 | |
428 | return count; | |
429 | } | |
430 | ||
431 | static const struct file_operations proc_lstats_operations = { | |
432 | .open = lstats_open, | |
433 | .read = seq_read, | |
434 | .write = lstats_write, | |
435 | .llseek = seq_lseek, | |
13d77c37 | 436 | .release = single_release, |
9745512c AV |
437 | }; |
438 | ||
439 | #endif | |
440 | ||
1da177e4 LT |
441 | /* The badness from the OOM killer */ |
442 | unsigned long badness(struct task_struct *p, unsigned long uptime); | |
443 | static int proc_oom_score(struct task_struct *task, char *buffer) | |
444 | { | |
445 | unsigned long points; | |
446 | struct timespec uptime; | |
447 | ||
448 | do_posix_clock_monotonic_gettime(&uptime); | |
19c5d45a | 449 | read_lock(&tasklist_lock); |
495789a5 | 450 | points = badness(task->group_leader, uptime.tv_sec); |
19c5d45a | 451 | read_unlock(&tasklist_lock); |
1da177e4 LT |
452 | return sprintf(buffer, "%lu\n", points); |
453 | } | |
454 | ||
d85f50d5 NH |
455 | struct limit_names { |
456 | char *name; | |
457 | char *unit; | |
458 | }; | |
459 | ||
460 | static const struct limit_names lnames[RLIM_NLIMITS] = { | |
cff4edb5 | 461 | [RLIMIT_CPU] = {"Max cpu time", "seconds"}, |
d85f50d5 NH |
462 | [RLIMIT_FSIZE] = {"Max file size", "bytes"}, |
463 | [RLIMIT_DATA] = {"Max data size", "bytes"}, | |
464 | [RLIMIT_STACK] = {"Max stack size", "bytes"}, | |
465 | [RLIMIT_CORE] = {"Max core file size", "bytes"}, | |
466 | [RLIMIT_RSS] = {"Max resident set", "bytes"}, | |
467 | [RLIMIT_NPROC] = {"Max processes", "processes"}, | |
468 | [RLIMIT_NOFILE] = {"Max open files", "files"}, | |
469 | [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"}, | |
470 | [RLIMIT_AS] = {"Max address space", "bytes"}, | |
471 | [RLIMIT_LOCKS] = {"Max file locks", "locks"}, | |
472 | [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"}, | |
473 | [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"}, | |
474 | [RLIMIT_NICE] = {"Max nice priority", NULL}, | |
475 | [RLIMIT_RTPRIO] = {"Max realtime priority", NULL}, | |
8808117c | 476 | [RLIMIT_RTTIME] = {"Max realtime timeout", "us"}, |
d85f50d5 NH |
477 | }; |
478 | ||
479 | /* Display limits for a process */ | |
480 | static int proc_pid_limits(struct task_struct *task, char *buffer) | |
481 | { | |
482 | unsigned int i; | |
483 | int count = 0; | |
484 | unsigned long flags; | |
485 | char *bufptr = buffer; | |
486 | ||
487 | struct rlimit rlim[RLIM_NLIMITS]; | |
488 | ||
a6bebbc8 | 489 | if (!lock_task_sighand(task, &flags)) |
d85f50d5 | 490 | return 0; |
d85f50d5 NH |
491 | memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS); |
492 | unlock_task_sighand(task, &flags); | |
d85f50d5 NH |
493 | |
494 | /* | |
495 | * print the file header | |
496 | */ | |
497 | count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n", | |
498 | "Limit", "Soft Limit", "Hard Limit", "Units"); | |
499 | ||
500 | for (i = 0; i < RLIM_NLIMITS; i++) { | |
501 | if (rlim[i].rlim_cur == RLIM_INFINITY) | |
502 | count += sprintf(&bufptr[count], "%-25s %-20s ", | |
503 | lnames[i].name, "unlimited"); | |
504 | else | |
505 | count += sprintf(&bufptr[count], "%-25s %-20lu ", | |
506 | lnames[i].name, rlim[i].rlim_cur); | |
507 | ||
508 | if (rlim[i].rlim_max == RLIM_INFINITY) | |
509 | count += sprintf(&bufptr[count], "%-20s ", "unlimited"); | |
510 | else | |
511 | count += sprintf(&bufptr[count], "%-20lu ", | |
512 | rlim[i].rlim_max); | |
513 | ||
514 | if (lnames[i].unit) | |
515 | count += sprintf(&bufptr[count], "%-10s\n", | |
516 | lnames[i].unit); | |
517 | else | |
518 | count += sprintf(&bufptr[count], "\n"); | |
519 | } | |
520 | ||
521 | return count; | |
522 | } | |
523 | ||
ebcb6734 RM |
524 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
525 | static int proc_pid_syscall(struct task_struct *task, char *buffer) | |
526 | { | |
527 | long nr; | |
528 | unsigned long args[6], sp, pc; | |
529 | ||
530 | if (task_current_syscall(task, &nr, args, 6, &sp, &pc)) | |
531 | return sprintf(buffer, "running\n"); | |
532 | ||
533 | if (nr < 0) | |
534 | return sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc); | |
535 | ||
536 | return sprintf(buffer, | |
537 | "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n", | |
538 | nr, | |
539 | args[0], args[1], args[2], args[3], args[4], args[5], | |
540 | sp, pc); | |
541 | } | |
542 | #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */ | |
543 | ||
1da177e4 LT |
544 | /************************************************************************/ |
545 | /* Here the fs part begins */ | |
546 | /************************************************************************/ | |
547 | ||
548 | /* permission checks */ | |
778c1144 | 549 | static int proc_fd_access_allowed(struct inode *inode) |
1da177e4 | 550 | { |
778c1144 EB |
551 | struct task_struct *task; |
552 | int allowed = 0; | |
df26c40e EB |
553 | /* Allow access to a task's file descriptors if it is us or we |
554 | * may use ptrace attach to the process and find out that | |
555 | * information. | |
778c1144 EB |
556 | */ |
557 | task = get_proc_task(inode); | |
df26c40e | 558 | if (task) { |
006ebb40 | 559 | allowed = ptrace_may_access(task, PTRACE_MODE_READ); |
778c1144 | 560 | put_task_struct(task); |
df26c40e | 561 | } |
778c1144 | 562 | return allowed; |
1da177e4 LT |
563 | } |
564 | ||
6d76fa58 LT |
565 | static int proc_setattr(struct dentry *dentry, struct iattr *attr) |
566 | { | |
567 | int error; | |
568 | struct inode *inode = dentry->d_inode; | |
569 | ||
570 | if (attr->ia_valid & ATTR_MODE) | |
571 | return -EPERM; | |
572 | ||
573 | error = inode_change_ok(inode, attr); | |
1e8123fd JJ |
574 | if (!error) |
575 | error = inode_setattr(inode, attr); | |
6d76fa58 LT |
576 | return error; |
577 | } | |
578 | ||
c5ef1c42 | 579 | static const struct inode_operations proc_def_inode_operations = { |
6d76fa58 LT |
580 | .setattr = proc_setattr, |
581 | }; | |
582 | ||
a1a2c409 MS |
583 | static int mounts_open_common(struct inode *inode, struct file *file, |
584 | const struct seq_operations *op) | |
1da177e4 | 585 | { |
99f89551 | 586 | struct task_struct *task = get_proc_task(inode); |
cf7b708c | 587 | struct nsproxy *nsp; |
6b3286ed | 588 | struct mnt_namespace *ns = NULL; |
a1a2c409 | 589 | struct path root; |
5addc5dd AV |
590 | struct proc_mounts *p; |
591 | int ret = -EINVAL; | |
1da177e4 | 592 | |
99f89551 | 593 | if (task) { |
cf7b708c PE |
594 | rcu_read_lock(); |
595 | nsp = task_nsproxy(task); | |
596 | if (nsp) { | |
597 | ns = nsp->mnt_ns; | |
863c4702 AD |
598 | if (ns) |
599 | get_mnt_ns(ns); | |
600 | } | |
cf7b708c | 601 | rcu_read_unlock(); |
7c2c7d99 HD |
602 | if (ns && get_fs_path(task, &root, 1) == 0) |
603 | ret = 0; | |
99f89551 EB |
604 | put_task_struct(task); |
605 | } | |
5addc5dd | 606 | |
a1a2c409 MS |
607 | if (!ns) |
608 | goto err; | |
7c2c7d99 | 609 | if (ret) |
a1a2c409 MS |
610 | goto err_put_ns; |
611 | ||
a1a2c409 MS |
612 | ret = -ENOMEM; |
613 | p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL); | |
614 | if (!p) | |
615 | goto err_put_path; | |
616 | ||
617 | file->private_data = &p->m; | |
618 | ret = seq_open(file, op); | |
619 | if (ret) | |
620 | goto err_free; | |
621 | ||
622 | p->m.private = p; | |
623 | p->ns = ns; | |
624 | p->root = root; | |
625 | p->event = ns->event; | |
626 | ||
627 | return 0; | |
628 | ||
629 | err_free: | |
630 | kfree(p); | |
631 | err_put_path: | |
632 | path_put(&root); | |
633 | err_put_ns: | |
634 | put_mnt_ns(ns); | |
635 | err: | |
1da177e4 LT |
636 | return ret; |
637 | } | |
638 | ||
639 | static int mounts_release(struct inode *inode, struct file *file) | |
640 | { | |
a1a2c409 MS |
641 | struct proc_mounts *p = file->private_data; |
642 | path_put(&p->root); | |
643 | put_mnt_ns(p->ns); | |
1da177e4 LT |
644 | return seq_release(inode, file); |
645 | } | |
646 | ||
5addc5dd AV |
647 | static unsigned mounts_poll(struct file *file, poll_table *wait) |
648 | { | |
649 | struct proc_mounts *p = file->private_data; | |
a1a2c409 | 650 | struct mnt_namespace *ns = p->ns; |
31b07093 | 651 | unsigned res = POLLIN | POLLRDNORM; |
5addc5dd AV |
652 | |
653 | poll_wait(file, &ns->poll, wait); | |
654 | ||
655 | spin_lock(&vfsmount_lock); | |
656 | if (p->event != ns->event) { | |
657 | p->event = ns->event; | |
31b07093 | 658 | res |= POLLERR | POLLPRI; |
5addc5dd AV |
659 | } |
660 | spin_unlock(&vfsmount_lock); | |
661 | ||
662 | return res; | |
663 | } | |
664 | ||
a1a2c409 MS |
665 | static int mounts_open(struct inode *inode, struct file *file) |
666 | { | |
667 | return mounts_open_common(inode, file, &mounts_op); | |
668 | } | |
669 | ||
00977a59 | 670 | static const struct file_operations proc_mounts_operations = { |
1da177e4 LT |
671 | .open = mounts_open, |
672 | .read = seq_read, | |
673 | .llseek = seq_lseek, | |
674 | .release = mounts_release, | |
5addc5dd | 675 | .poll = mounts_poll, |
1da177e4 LT |
676 | }; |
677 | ||
2d4d4864 RP |
678 | static int mountinfo_open(struct inode *inode, struct file *file) |
679 | { | |
680 | return mounts_open_common(inode, file, &mountinfo_op); | |
681 | } | |
682 | ||
683 | static const struct file_operations proc_mountinfo_operations = { | |
684 | .open = mountinfo_open, | |
685 | .read = seq_read, | |
686 | .llseek = seq_lseek, | |
687 | .release = mounts_release, | |
688 | .poll = mounts_poll, | |
689 | }; | |
690 | ||
b4629fe2 CL |
691 | static int mountstats_open(struct inode *inode, struct file *file) |
692 | { | |
a1a2c409 | 693 | return mounts_open_common(inode, file, &mountstats_op); |
b4629fe2 CL |
694 | } |
695 | ||
00977a59 | 696 | static const struct file_operations proc_mountstats_operations = { |
b4629fe2 CL |
697 | .open = mountstats_open, |
698 | .read = seq_read, | |
699 | .llseek = seq_lseek, | |
700 | .release = mounts_release, | |
701 | }; | |
702 | ||
1da177e4 LT |
703 | #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */ |
704 | ||
705 | static ssize_t proc_info_read(struct file * file, char __user * buf, | |
706 | size_t count, loff_t *ppos) | |
707 | { | |
2fddfeef | 708 | struct inode * inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
709 | unsigned long page; |
710 | ssize_t length; | |
99f89551 EB |
711 | struct task_struct *task = get_proc_task(inode); |
712 | ||
713 | length = -ESRCH; | |
714 | if (!task) | |
715 | goto out_no_task; | |
1da177e4 LT |
716 | |
717 | if (count > PROC_BLOCK_SIZE) | |
718 | count = PROC_BLOCK_SIZE; | |
99f89551 EB |
719 | |
720 | length = -ENOMEM; | |
e12ba74d | 721 | if (!(page = __get_free_page(GFP_TEMPORARY))) |
99f89551 | 722 | goto out; |
1da177e4 LT |
723 | |
724 | length = PROC_I(inode)->op.proc_read(task, (char*)page); | |
725 | ||
726 | if (length >= 0) | |
727 | length = simple_read_from_buffer(buf, count, ppos, (char *)page, length); | |
728 | free_page(page); | |
99f89551 EB |
729 | out: |
730 | put_task_struct(task); | |
731 | out_no_task: | |
1da177e4 LT |
732 | return length; |
733 | } | |
734 | ||
00977a59 | 735 | static const struct file_operations proc_info_file_operations = { |
1da177e4 LT |
736 | .read = proc_info_read, |
737 | }; | |
738 | ||
be614086 EB |
739 | static int proc_single_show(struct seq_file *m, void *v) |
740 | { | |
741 | struct inode *inode = m->private; | |
742 | struct pid_namespace *ns; | |
743 | struct pid *pid; | |
744 | struct task_struct *task; | |
745 | int ret; | |
746 | ||
747 | ns = inode->i_sb->s_fs_info; | |
748 | pid = proc_pid(inode); | |
749 | task = get_pid_task(pid, PIDTYPE_PID); | |
750 | if (!task) | |
751 | return -ESRCH; | |
752 | ||
753 | ret = PROC_I(inode)->op.proc_show(m, ns, pid, task); | |
754 | ||
755 | put_task_struct(task); | |
756 | return ret; | |
757 | } | |
758 | ||
759 | static int proc_single_open(struct inode *inode, struct file *filp) | |
760 | { | |
761 | int ret; | |
762 | ret = single_open(filp, proc_single_show, NULL); | |
763 | if (!ret) { | |
764 | struct seq_file *m = filp->private_data; | |
765 | ||
766 | m->private = inode; | |
767 | } | |
768 | return ret; | |
769 | } | |
770 | ||
771 | static const struct file_operations proc_single_file_operations = { | |
772 | .open = proc_single_open, | |
773 | .read = seq_read, | |
774 | .llseek = seq_lseek, | |
775 | .release = single_release, | |
776 | }; | |
777 | ||
1da177e4 LT |
778 | static int mem_open(struct inode* inode, struct file* file) |
779 | { | |
780 | file->private_data = (void*)((long)current->self_exec_id); | |
781 | return 0; | |
782 | } | |
783 | ||
784 | static ssize_t mem_read(struct file * file, char __user * buf, | |
785 | size_t count, loff_t *ppos) | |
786 | { | |
2fddfeef | 787 | struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode); |
1da177e4 LT |
788 | char *page; |
789 | unsigned long src = *ppos; | |
790 | int ret = -ESRCH; | |
791 | struct mm_struct *mm; | |
792 | ||
99f89551 EB |
793 | if (!task) |
794 | goto out_no_task; | |
795 | ||
638fa202 | 796 | if (check_mem_permission(task)) |
1da177e4 LT |
797 | goto out; |
798 | ||
799 | ret = -ENOMEM; | |
e12ba74d | 800 | page = (char *)__get_free_page(GFP_TEMPORARY); |
1da177e4 LT |
801 | if (!page) |
802 | goto out; | |
803 | ||
804 | ret = 0; | |
805 | ||
806 | mm = get_task_mm(task); | |
807 | if (!mm) | |
808 | goto out_free; | |
809 | ||
810 | ret = -EIO; | |
811 | ||
812 | if (file->private_data != (void*)((long)current->self_exec_id)) | |
813 | goto out_put; | |
814 | ||
815 | ret = 0; | |
816 | ||
817 | while (count > 0) { | |
818 | int this_len, retval; | |
819 | ||
820 | this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count; | |
821 | retval = access_process_vm(task, src, page, this_len, 0); | |
638fa202 | 822 | if (!retval || check_mem_permission(task)) { |
1da177e4 LT |
823 | if (!ret) |
824 | ret = -EIO; | |
825 | break; | |
826 | } | |
827 | ||
828 | if (copy_to_user(buf, page, retval)) { | |
829 | ret = -EFAULT; | |
830 | break; | |
831 | } | |
832 | ||
833 | ret += retval; | |
834 | src += retval; | |
835 | buf += retval; | |
836 | count -= retval; | |
837 | } | |
838 | *ppos = src; | |
839 | ||
840 | out_put: | |
841 | mmput(mm); | |
842 | out_free: | |
843 | free_page((unsigned long) page); | |
844 | out: | |
99f89551 EB |
845 | put_task_struct(task); |
846 | out_no_task: | |
1da177e4 LT |
847 | return ret; |
848 | } | |
849 | ||
850 | #define mem_write NULL | |
851 | ||
852 | #ifndef mem_write | |
853 | /* This is a security hazard */ | |
63967fa9 | 854 | static ssize_t mem_write(struct file * file, const char __user *buf, |
1da177e4 LT |
855 | size_t count, loff_t *ppos) |
856 | { | |
f7ca54f4 | 857 | int copied; |
1da177e4 | 858 | char *page; |
2fddfeef | 859 | struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode); |
1da177e4 LT |
860 | unsigned long dst = *ppos; |
861 | ||
99f89551 EB |
862 | copied = -ESRCH; |
863 | if (!task) | |
864 | goto out_no_task; | |
865 | ||
638fa202 | 866 | if (check_mem_permission(task)) |
99f89551 | 867 | goto out; |
1da177e4 | 868 | |
99f89551 | 869 | copied = -ENOMEM; |
e12ba74d | 870 | page = (char *)__get_free_page(GFP_TEMPORARY); |
1da177e4 | 871 | if (!page) |
99f89551 | 872 | goto out; |
1da177e4 | 873 | |
f7ca54f4 | 874 | copied = 0; |
1da177e4 LT |
875 | while (count > 0) { |
876 | int this_len, retval; | |
877 | ||
878 | this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count; | |
879 | if (copy_from_user(page, buf, this_len)) { | |
880 | copied = -EFAULT; | |
881 | break; | |
882 | } | |
883 | retval = access_process_vm(task, dst, page, this_len, 1); | |
884 | if (!retval) { | |
885 | if (!copied) | |
886 | copied = -EIO; | |
887 | break; | |
888 | } | |
889 | copied += retval; | |
890 | buf += retval; | |
891 | dst += retval; | |
892 | count -= retval; | |
893 | } | |
894 | *ppos = dst; | |
895 | free_page((unsigned long) page); | |
99f89551 EB |
896 | out: |
897 | put_task_struct(task); | |
898 | out_no_task: | |
1da177e4 LT |
899 | return copied; |
900 | } | |
901 | #endif | |
902 | ||
85863e47 | 903 | loff_t mem_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
904 | { |
905 | switch (orig) { | |
906 | case 0: | |
907 | file->f_pos = offset; | |
908 | break; | |
909 | case 1: | |
910 | file->f_pos += offset; | |
911 | break; | |
912 | default: | |
913 | return -EINVAL; | |
914 | } | |
915 | force_successful_syscall_return(); | |
916 | return file->f_pos; | |
917 | } | |
918 | ||
00977a59 | 919 | static const struct file_operations proc_mem_operations = { |
1da177e4 LT |
920 | .llseek = mem_lseek, |
921 | .read = mem_read, | |
922 | .write = mem_write, | |
923 | .open = mem_open, | |
924 | }; | |
925 | ||
315e28c8 JP |
926 | static ssize_t environ_read(struct file *file, char __user *buf, |
927 | size_t count, loff_t *ppos) | |
928 | { | |
929 | struct task_struct *task = get_proc_task(file->f_dentry->d_inode); | |
930 | char *page; | |
931 | unsigned long src = *ppos; | |
932 | int ret = -ESRCH; | |
933 | struct mm_struct *mm; | |
934 | ||
935 | if (!task) | |
936 | goto out_no_task; | |
937 | ||
006ebb40 | 938 | if (!ptrace_may_access(task, PTRACE_MODE_READ)) |
315e28c8 JP |
939 | goto out; |
940 | ||
941 | ret = -ENOMEM; | |
942 | page = (char *)__get_free_page(GFP_TEMPORARY); | |
943 | if (!page) | |
944 | goto out; | |
945 | ||
946 | ret = 0; | |
947 | ||
948 | mm = get_task_mm(task); | |
949 | if (!mm) | |
950 | goto out_free; | |
951 | ||
952 | while (count > 0) { | |
953 | int this_len, retval, max_len; | |
954 | ||
955 | this_len = mm->env_end - (mm->env_start + src); | |
956 | ||
957 | if (this_len <= 0) | |
958 | break; | |
959 | ||
960 | max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count; | |
961 | this_len = (this_len > max_len) ? max_len : this_len; | |
962 | ||
963 | retval = access_process_vm(task, (mm->env_start + src), | |
964 | page, this_len, 0); | |
965 | ||
966 | if (retval <= 0) { | |
967 | ret = retval; | |
968 | break; | |
969 | } | |
970 | ||
971 | if (copy_to_user(buf, page, retval)) { | |
972 | ret = -EFAULT; | |
973 | break; | |
974 | } | |
975 | ||
976 | ret += retval; | |
977 | src += retval; | |
978 | buf += retval; | |
979 | count -= retval; | |
980 | } | |
981 | *ppos = src; | |
982 | ||
983 | mmput(mm); | |
984 | out_free: | |
985 | free_page((unsigned long) page); | |
986 | out: | |
987 | put_task_struct(task); | |
988 | out_no_task: | |
989 | return ret; | |
990 | } | |
991 | ||
992 | static const struct file_operations proc_environ_operations = { | |
993 | .read = environ_read, | |
994 | }; | |
995 | ||
1da177e4 LT |
996 | static ssize_t oom_adjust_read(struct file *file, char __user *buf, |
997 | size_t count, loff_t *ppos) | |
998 | { | |
2fddfeef | 999 | struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode); |
8578cea7 | 1000 | char buffer[PROC_NUMBUF]; |
1da177e4 | 1001 | size_t len; |
28b83c51 KM |
1002 | int oom_adjust = OOM_DISABLE; |
1003 | unsigned long flags; | |
1da177e4 | 1004 | |
99f89551 EB |
1005 | if (!task) |
1006 | return -ESRCH; | |
28b83c51 KM |
1007 | |
1008 | if (lock_task_sighand(task, &flags)) { | |
1009 | oom_adjust = task->signal->oom_adj; | |
1010 | unlock_task_sighand(task, &flags); | |
1011 | } | |
1012 | ||
99f89551 EB |
1013 | put_task_struct(task); |
1014 | ||
8578cea7 | 1015 | len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust); |
0c28f287 AM |
1016 | |
1017 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
1da177e4 LT |
1018 | } |
1019 | ||
1020 | static ssize_t oom_adjust_write(struct file *file, const char __user *buf, | |
1021 | size_t count, loff_t *ppos) | |
1022 | { | |
99f89551 | 1023 | struct task_struct *task; |
5d863b89 KM |
1024 | char buffer[PROC_NUMBUF]; |
1025 | long oom_adjust; | |
28b83c51 | 1026 | unsigned long flags; |
5d863b89 | 1027 | int err; |
1da177e4 | 1028 | |
8578cea7 EB |
1029 | memset(buffer, 0, sizeof(buffer)); |
1030 | if (count > sizeof(buffer) - 1) | |
1031 | count = sizeof(buffer) - 1; | |
1da177e4 LT |
1032 | if (copy_from_user(buffer, buf, count)) |
1033 | return -EFAULT; | |
5d863b89 KM |
1034 | |
1035 | err = strict_strtol(strstrip(buffer), 0, &oom_adjust); | |
1036 | if (err) | |
1037 | return -EINVAL; | |
8ac773b4 AD |
1038 | if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) && |
1039 | oom_adjust != OOM_DISABLE) | |
1da177e4 | 1040 | return -EINVAL; |
5d863b89 | 1041 | |
2fddfeef | 1042 | task = get_proc_task(file->f_path.dentry->d_inode); |
99f89551 EB |
1043 | if (!task) |
1044 | return -ESRCH; | |
28b83c51 KM |
1045 | if (!lock_task_sighand(task, &flags)) { |
1046 | put_task_struct(task); | |
1047 | return -ESRCH; | |
1048 | } | |
1049 | ||
1050 | if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) { | |
1051 | unlock_task_sighand(task, &flags); | |
8fb4fc68 GJ |
1052 | put_task_struct(task); |
1053 | return -EACCES; | |
1054 | } | |
28b83c51 KM |
1055 | |
1056 | task->signal->oom_adj = oom_adjust; | |
1057 | ||
1058 | unlock_task_sighand(task, &flags); | |
99f89551 | 1059 | put_task_struct(task); |
5d863b89 KM |
1060 | |
1061 | return count; | |
1da177e4 LT |
1062 | } |
1063 | ||
00977a59 | 1064 | static const struct file_operations proc_oom_adjust_operations = { |
1da177e4 LT |
1065 | .read = oom_adjust_read, |
1066 | .write = oom_adjust_write, | |
1067 | }; | |
1068 | ||
1da177e4 LT |
1069 | #ifdef CONFIG_AUDITSYSCALL |
1070 | #define TMPBUFLEN 21 | |
1071 | static ssize_t proc_loginuid_read(struct file * file, char __user * buf, | |
1072 | size_t count, loff_t *ppos) | |
1073 | { | |
2fddfeef | 1074 | struct inode * inode = file->f_path.dentry->d_inode; |
99f89551 | 1075 | struct task_struct *task = get_proc_task(inode); |
1da177e4 LT |
1076 | ssize_t length; |
1077 | char tmpbuf[TMPBUFLEN]; | |
1078 | ||
99f89551 EB |
1079 | if (!task) |
1080 | return -ESRCH; | |
1da177e4 | 1081 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", |
0c11b942 | 1082 | audit_get_loginuid(task)); |
99f89551 | 1083 | put_task_struct(task); |
1da177e4 LT |
1084 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); |
1085 | } | |
1086 | ||
1087 | static ssize_t proc_loginuid_write(struct file * file, const char __user * buf, | |
1088 | size_t count, loff_t *ppos) | |
1089 | { | |
2fddfeef | 1090 | struct inode * inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
1091 | char *page, *tmp; |
1092 | ssize_t length; | |
1da177e4 LT |
1093 | uid_t loginuid; |
1094 | ||
1095 | if (!capable(CAP_AUDIT_CONTROL)) | |
1096 | return -EPERM; | |
1097 | ||
7dc52157 PM |
1098 | rcu_read_lock(); |
1099 | if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) { | |
1100 | rcu_read_unlock(); | |
1da177e4 | 1101 | return -EPERM; |
7dc52157 PM |
1102 | } |
1103 | rcu_read_unlock(); | |
1da177e4 | 1104 | |
e0182909 AV |
1105 | if (count >= PAGE_SIZE) |
1106 | count = PAGE_SIZE - 1; | |
1da177e4 LT |
1107 | |
1108 | if (*ppos != 0) { | |
1109 | /* No partial writes. */ | |
1110 | return -EINVAL; | |
1111 | } | |
e12ba74d | 1112 | page = (char*)__get_free_page(GFP_TEMPORARY); |
1da177e4 LT |
1113 | if (!page) |
1114 | return -ENOMEM; | |
1115 | length = -EFAULT; | |
1116 | if (copy_from_user(page, buf, count)) | |
1117 | goto out_free_page; | |
1118 | ||
e0182909 | 1119 | page[count] = '\0'; |
1da177e4 LT |
1120 | loginuid = simple_strtoul(page, &tmp, 10); |
1121 | if (tmp == page) { | |
1122 | length = -EINVAL; | |
1123 | goto out_free_page; | |
1124 | ||
1125 | } | |
99f89551 | 1126 | length = audit_set_loginuid(current, loginuid); |
1da177e4 LT |
1127 | if (likely(length == 0)) |
1128 | length = count; | |
1129 | ||
1130 | out_free_page: | |
1131 | free_page((unsigned long) page); | |
1132 | return length; | |
1133 | } | |
1134 | ||
00977a59 | 1135 | static const struct file_operations proc_loginuid_operations = { |
1da177e4 LT |
1136 | .read = proc_loginuid_read, |
1137 | .write = proc_loginuid_write, | |
1138 | }; | |
1e0bd755 EP |
1139 | |
1140 | static ssize_t proc_sessionid_read(struct file * file, char __user * buf, | |
1141 | size_t count, loff_t *ppos) | |
1142 | { | |
1143 | struct inode * inode = file->f_path.dentry->d_inode; | |
1144 | struct task_struct *task = get_proc_task(inode); | |
1145 | ssize_t length; | |
1146 | char tmpbuf[TMPBUFLEN]; | |
1147 | ||
1148 | if (!task) | |
1149 | return -ESRCH; | |
1150 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", | |
1151 | audit_get_sessionid(task)); | |
1152 | put_task_struct(task); | |
1153 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); | |
1154 | } | |
1155 | ||
1156 | static const struct file_operations proc_sessionid_operations = { | |
1157 | .read = proc_sessionid_read, | |
1158 | }; | |
1da177e4 LT |
1159 | #endif |
1160 | ||
f4f154fd AM |
1161 | #ifdef CONFIG_FAULT_INJECTION |
1162 | static ssize_t proc_fault_inject_read(struct file * file, char __user * buf, | |
1163 | size_t count, loff_t *ppos) | |
1164 | { | |
1165 | struct task_struct *task = get_proc_task(file->f_dentry->d_inode); | |
1166 | char buffer[PROC_NUMBUF]; | |
1167 | size_t len; | |
1168 | int make_it_fail; | |
f4f154fd AM |
1169 | |
1170 | if (!task) | |
1171 | return -ESRCH; | |
1172 | make_it_fail = task->make_it_fail; | |
1173 | put_task_struct(task); | |
1174 | ||
1175 | len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail); | |
0c28f287 AM |
1176 | |
1177 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
f4f154fd AM |
1178 | } |
1179 | ||
1180 | static ssize_t proc_fault_inject_write(struct file * file, | |
1181 | const char __user * buf, size_t count, loff_t *ppos) | |
1182 | { | |
1183 | struct task_struct *task; | |
1184 | char buffer[PROC_NUMBUF], *end; | |
1185 | int make_it_fail; | |
1186 | ||
1187 | if (!capable(CAP_SYS_RESOURCE)) | |
1188 | return -EPERM; | |
1189 | memset(buffer, 0, sizeof(buffer)); | |
1190 | if (count > sizeof(buffer) - 1) | |
1191 | count = sizeof(buffer) - 1; | |
1192 | if (copy_from_user(buffer, buf, count)) | |
1193 | return -EFAULT; | |
cba8aafe VL |
1194 | make_it_fail = simple_strtol(strstrip(buffer), &end, 0); |
1195 | if (*end) | |
1196 | return -EINVAL; | |
f4f154fd AM |
1197 | task = get_proc_task(file->f_dentry->d_inode); |
1198 | if (!task) | |
1199 | return -ESRCH; | |
1200 | task->make_it_fail = make_it_fail; | |
1201 | put_task_struct(task); | |
cba8aafe VL |
1202 | |
1203 | return count; | |
f4f154fd AM |
1204 | } |
1205 | ||
00977a59 | 1206 | static const struct file_operations proc_fault_inject_operations = { |
f4f154fd AM |
1207 | .read = proc_fault_inject_read, |
1208 | .write = proc_fault_inject_write, | |
1209 | }; | |
1210 | #endif | |
1211 | ||
9745512c | 1212 | |
43ae34cb IM |
1213 | #ifdef CONFIG_SCHED_DEBUG |
1214 | /* | |
1215 | * Print out various scheduling related per-task fields: | |
1216 | */ | |
1217 | static int sched_show(struct seq_file *m, void *v) | |
1218 | { | |
1219 | struct inode *inode = m->private; | |
1220 | struct task_struct *p; | |
1221 | ||
43ae34cb IM |
1222 | p = get_proc_task(inode); |
1223 | if (!p) | |
1224 | return -ESRCH; | |
1225 | proc_sched_show_task(p, m); | |
1226 | ||
1227 | put_task_struct(p); | |
1228 | ||
1229 | return 0; | |
1230 | } | |
1231 | ||
1232 | static ssize_t | |
1233 | sched_write(struct file *file, const char __user *buf, | |
1234 | size_t count, loff_t *offset) | |
1235 | { | |
1236 | struct inode *inode = file->f_path.dentry->d_inode; | |
1237 | struct task_struct *p; | |
1238 | ||
43ae34cb IM |
1239 | p = get_proc_task(inode); |
1240 | if (!p) | |
1241 | return -ESRCH; | |
1242 | proc_sched_set_task(p); | |
1243 | ||
1244 | put_task_struct(p); | |
1245 | ||
1246 | return count; | |
1247 | } | |
1248 | ||
1249 | static int sched_open(struct inode *inode, struct file *filp) | |
1250 | { | |
1251 | int ret; | |
1252 | ||
1253 | ret = single_open(filp, sched_show, NULL); | |
1254 | if (!ret) { | |
1255 | struct seq_file *m = filp->private_data; | |
1256 | ||
1257 | m->private = inode; | |
1258 | } | |
1259 | return ret; | |
1260 | } | |
1261 | ||
1262 | static const struct file_operations proc_pid_sched_operations = { | |
1263 | .open = sched_open, | |
1264 | .read = seq_read, | |
1265 | .write = sched_write, | |
1266 | .llseek = seq_lseek, | |
5ea473a1 | 1267 | .release = single_release, |
43ae34cb IM |
1268 | }; |
1269 | ||
1270 | #endif | |
1271 | ||
4614a696 JS |
1272 | static ssize_t comm_write(struct file *file, const char __user *buf, |
1273 | size_t count, loff_t *offset) | |
1274 | { | |
1275 | struct inode *inode = file->f_path.dentry->d_inode; | |
1276 | struct task_struct *p; | |
1277 | char buffer[TASK_COMM_LEN]; | |
1278 | ||
1279 | memset(buffer, 0, sizeof(buffer)); | |
1280 | if (count > sizeof(buffer) - 1) | |
1281 | count = sizeof(buffer) - 1; | |
1282 | if (copy_from_user(buffer, buf, count)) | |
1283 | return -EFAULT; | |
1284 | ||
1285 | p = get_proc_task(inode); | |
1286 | if (!p) | |
1287 | return -ESRCH; | |
1288 | ||
1289 | if (same_thread_group(current, p)) | |
1290 | set_task_comm(p, buffer); | |
1291 | else | |
1292 | count = -EINVAL; | |
1293 | ||
1294 | put_task_struct(p); | |
1295 | ||
1296 | return count; | |
1297 | } | |
1298 | ||
1299 | static int comm_show(struct seq_file *m, void *v) | |
1300 | { | |
1301 | struct inode *inode = m->private; | |
1302 | struct task_struct *p; | |
1303 | ||
1304 | p = get_proc_task(inode); | |
1305 | if (!p) | |
1306 | return -ESRCH; | |
1307 | ||
1308 | task_lock(p); | |
1309 | seq_printf(m, "%s\n", p->comm); | |
1310 | task_unlock(p); | |
1311 | ||
1312 | put_task_struct(p); | |
1313 | ||
1314 | return 0; | |
1315 | } | |
1316 | ||
1317 | static int comm_open(struct inode *inode, struct file *filp) | |
1318 | { | |
1319 | int ret; | |
1320 | ||
1321 | ret = single_open(filp, comm_show, NULL); | |
1322 | if (!ret) { | |
1323 | struct seq_file *m = filp->private_data; | |
1324 | ||
1325 | m->private = inode; | |
1326 | } | |
1327 | return ret; | |
1328 | } | |
1329 | ||
1330 | static const struct file_operations proc_pid_set_comm_operations = { | |
1331 | .open = comm_open, | |
1332 | .read = seq_read, | |
1333 | .write = comm_write, | |
1334 | .llseek = seq_lseek, | |
1335 | .release = single_release, | |
1336 | }; | |
1337 | ||
925d1c40 MH |
1338 | /* |
1339 | * We added or removed a vma mapping the executable. The vmas are only mapped | |
1340 | * during exec and are not mapped with the mmap system call. | |
1341 | * Callers must hold down_write() on the mm's mmap_sem for these | |
1342 | */ | |
1343 | void added_exe_file_vma(struct mm_struct *mm) | |
1344 | { | |
1345 | mm->num_exe_file_vmas++; | |
1346 | } | |
1347 | ||
1348 | void removed_exe_file_vma(struct mm_struct *mm) | |
1349 | { | |
1350 | mm->num_exe_file_vmas--; | |
1351 | if ((mm->num_exe_file_vmas == 0) && mm->exe_file){ | |
1352 | fput(mm->exe_file); | |
1353 | mm->exe_file = NULL; | |
1354 | } | |
1355 | ||
1356 | } | |
1357 | ||
1358 | void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) | |
1359 | { | |
1360 | if (new_exe_file) | |
1361 | get_file(new_exe_file); | |
1362 | if (mm->exe_file) | |
1363 | fput(mm->exe_file); | |
1364 | mm->exe_file = new_exe_file; | |
1365 | mm->num_exe_file_vmas = 0; | |
1366 | } | |
1367 | ||
1368 | struct file *get_mm_exe_file(struct mm_struct *mm) | |
1369 | { | |
1370 | struct file *exe_file; | |
1371 | ||
1372 | /* We need mmap_sem to protect against races with removal of | |
1373 | * VM_EXECUTABLE vmas */ | |
1374 | down_read(&mm->mmap_sem); | |
1375 | exe_file = mm->exe_file; | |
1376 | if (exe_file) | |
1377 | get_file(exe_file); | |
1378 | up_read(&mm->mmap_sem); | |
1379 | return exe_file; | |
1380 | } | |
1381 | ||
1382 | void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm) | |
1383 | { | |
1384 | /* It's safe to write the exe_file pointer without exe_file_lock because | |
1385 | * this is called during fork when the task is not yet in /proc */ | |
1386 | newmm->exe_file = get_mm_exe_file(oldmm); | |
1387 | } | |
1388 | ||
1389 | static int proc_exe_link(struct inode *inode, struct path *exe_path) | |
1390 | { | |
1391 | struct task_struct *task; | |
1392 | struct mm_struct *mm; | |
1393 | struct file *exe_file; | |
1394 | ||
1395 | task = get_proc_task(inode); | |
1396 | if (!task) | |
1397 | return -ENOENT; | |
1398 | mm = get_task_mm(task); | |
1399 | put_task_struct(task); | |
1400 | if (!mm) | |
1401 | return -ENOENT; | |
1402 | exe_file = get_mm_exe_file(mm); | |
1403 | mmput(mm); | |
1404 | if (exe_file) { | |
1405 | *exe_path = exe_file->f_path; | |
1406 | path_get(&exe_file->f_path); | |
1407 | fput(exe_file); | |
1408 | return 0; | |
1409 | } else | |
1410 | return -ENOENT; | |
1411 | } | |
1412 | ||
008b150a | 1413 | static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd) |
1da177e4 LT |
1414 | { |
1415 | struct inode *inode = dentry->d_inode; | |
1416 | int error = -EACCES; | |
1417 | ||
1418 | /* We don't need a base pointer in the /proc filesystem */ | |
1d957f9b | 1419 | path_put(&nd->path); |
1da177e4 | 1420 | |
778c1144 EB |
1421 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1422 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1423 | goto out; |
1da177e4 | 1424 | |
3dcd25f3 | 1425 | error = PROC_I(inode)->op.proc_get_link(inode, &nd->path); |
1da177e4 | 1426 | out: |
008b150a | 1427 | return ERR_PTR(error); |
1da177e4 LT |
1428 | } |
1429 | ||
3dcd25f3 | 1430 | static int do_proc_readlink(struct path *path, char __user *buffer, int buflen) |
1da177e4 | 1431 | { |
e12ba74d | 1432 | char *tmp = (char*)__get_free_page(GFP_TEMPORARY); |
3dcd25f3 | 1433 | char *pathname; |
1da177e4 LT |
1434 | int len; |
1435 | ||
1436 | if (!tmp) | |
1437 | return -ENOMEM; | |
0c28f287 | 1438 | |
cf28b486 | 1439 | pathname = d_path(path, tmp, PAGE_SIZE); |
3dcd25f3 JB |
1440 | len = PTR_ERR(pathname); |
1441 | if (IS_ERR(pathname)) | |
1da177e4 | 1442 | goto out; |
3dcd25f3 | 1443 | len = tmp + PAGE_SIZE - 1 - pathname; |
1da177e4 LT |
1444 | |
1445 | if (len > buflen) | |
1446 | len = buflen; | |
3dcd25f3 | 1447 | if (copy_to_user(buffer, pathname, len)) |
1da177e4 LT |
1448 | len = -EFAULT; |
1449 | out: | |
1450 | free_page((unsigned long)tmp); | |
1451 | return len; | |
1452 | } | |
1453 | ||
1454 | static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen) | |
1455 | { | |
1456 | int error = -EACCES; | |
1457 | struct inode *inode = dentry->d_inode; | |
3dcd25f3 | 1458 | struct path path; |
1da177e4 | 1459 | |
778c1144 EB |
1460 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1461 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1462 | goto out; |
1da177e4 | 1463 | |
3dcd25f3 | 1464 | error = PROC_I(inode)->op.proc_get_link(inode, &path); |
1da177e4 LT |
1465 | if (error) |
1466 | goto out; | |
1467 | ||
3dcd25f3 JB |
1468 | error = do_proc_readlink(&path, buffer, buflen); |
1469 | path_put(&path); | |
1da177e4 | 1470 | out: |
1da177e4 LT |
1471 | return error; |
1472 | } | |
1473 | ||
c5ef1c42 | 1474 | static const struct inode_operations proc_pid_link_inode_operations = { |
1da177e4 | 1475 | .readlink = proc_pid_readlink, |
6d76fa58 LT |
1476 | .follow_link = proc_pid_follow_link, |
1477 | .setattr = proc_setattr, | |
1da177e4 LT |
1478 | }; |
1479 | ||
28a6d671 EB |
1480 | |
1481 | /* building an inode */ | |
1482 | ||
1483 | static int task_dumpable(struct task_struct *task) | |
1da177e4 | 1484 | { |
28a6d671 EB |
1485 | int dumpable = 0; |
1486 | struct mm_struct *mm; | |
1da177e4 | 1487 | |
28a6d671 EB |
1488 | task_lock(task); |
1489 | mm = task->mm; | |
1490 | if (mm) | |
6c5d5238 | 1491 | dumpable = get_dumpable(mm); |
28a6d671 EB |
1492 | task_unlock(task); |
1493 | if(dumpable == 1) | |
1494 | return 1; | |
1495 | return 0; | |
1496 | } | |
1da177e4 | 1497 | |
1da177e4 | 1498 | |
61a28784 | 1499 | static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task) |
28a6d671 EB |
1500 | { |
1501 | struct inode * inode; | |
1502 | struct proc_inode *ei; | |
c69e8d9c | 1503 | const struct cred *cred; |
1da177e4 | 1504 | |
28a6d671 | 1505 | /* We need a new inode */ |
1da177e4 | 1506 | |
28a6d671 EB |
1507 | inode = new_inode(sb); |
1508 | if (!inode) | |
1509 | goto out; | |
1510 | ||
1511 | /* Common stuff */ | |
1512 | ei = PROC_I(inode); | |
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 | ||
28a6d671 | 1540 | static 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 DH |
1544 | const struct cred *cred; |
1545 | ||
28a6d671 | 1546 | generic_fillattr(inode, stat); |
1da177e4 | 1547 | |
28a6d671 EB |
1548 | rcu_read_lock(); |
1549 | stat->uid = 0; | |
1550 | stat->gid = 0; | |
1551 | task = pid_task(proc_pid(inode), PIDTYPE_PID); | |
1552 | if (task) { | |
1553 | if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) || | |
1554 | task_dumpable(task)) { | |
c69e8d9c DH |
1555 | cred = __task_cred(task); |
1556 | stat->uid = cred->euid; | |
1557 | stat->gid = cred->egid; | |
1da177e4 LT |
1558 | } |
1559 | } | |
28a6d671 | 1560 | rcu_read_unlock(); |
d6e71144 | 1561 | return 0; |
1da177e4 LT |
1562 | } |
1563 | ||
1da177e4 LT |
1564 | /* dentry stuff */ |
1565 | ||
1566 | /* | |
1567 | * Exceptional case: normally we are not allowed to unhash a busy | |
1568 | * directory. In this case, however, we can do it - no aliasing problems | |
1569 | * due to the way we treat inodes. | |
1570 | * | |
1571 | * Rewrite the inode's ownerships here because the owning task may have | |
1572 | * performed a setuid(), etc. | |
99f89551 EB |
1573 | * |
1574 | * Before the /proc/pid/status file was created the only way to read | |
1575 | * the effective uid of a /process was to stat /proc/pid. Reading | |
1576 | * /proc/pid/status is slow enough that procps and other packages | |
1577 | * kept stating /proc/pid. To keep the rules in /proc simple I have | |
1578 | * made this apply to all per process world readable and executable | |
1579 | * directories. | |
1da177e4 LT |
1580 | */ |
1581 | static int pid_revalidate(struct dentry *dentry, struct nameidata *nd) | |
1582 | { | |
1583 | struct inode *inode = dentry->d_inode; | |
99f89551 | 1584 | struct task_struct *task = get_proc_task(inode); |
c69e8d9c DH |
1585 | const struct cred *cred; |
1586 | ||
99f89551 EB |
1587 | if (task) { |
1588 | if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) || | |
1589 | task_dumpable(task)) { | |
c69e8d9c DH |
1590 | rcu_read_lock(); |
1591 | cred = __task_cred(task); | |
1592 | inode->i_uid = cred->euid; | |
1593 | inode->i_gid = cred->egid; | |
1594 | rcu_read_unlock(); | |
1da177e4 LT |
1595 | } else { |
1596 | inode->i_uid = 0; | |
1597 | inode->i_gid = 0; | |
1598 | } | |
9ee8ab9f | 1599 | inode->i_mode &= ~(S_ISUID | S_ISGID); |
1da177e4 | 1600 | security_task_to_inode(task, inode); |
99f89551 | 1601 | put_task_struct(task); |
1da177e4 LT |
1602 | return 1; |
1603 | } | |
1604 | d_drop(dentry); | |
1605 | return 0; | |
1606 | } | |
1607 | ||
28a6d671 | 1608 | static int pid_delete_dentry(struct dentry * dentry) |
99f89551 | 1609 | { |
28a6d671 EB |
1610 | /* Is the task we represent dead? |
1611 | * If so, then don't put the dentry on the lru list, | |
1612 | * kill it immediately. | |
1613 | */ | |
1614 | return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first; | |
1615 | } | |
1616 | ||
d72f71eb | 1617 | static const struct dentry_operations pid_dentry_operations = |
28a6d671 EB |
1618 | { |
1619 | .d_revalidate = pid_revalidate, | |
1620 | .d_delete = pid_delete_dentry, | |
1621 | }; | |
1622 | ||
1623 | /* Lookups */ | |
1624 | ||
c5141e6d ED |
1625 | typedef struct dentry *instantiate_t(struct inode *, struct dentry *, |
1626 | struct task_struct *, const void *); | |
61a28784 | 1627 | |
1c0d04c9 EB |
1628 | /* |
1629 | * Fill a directory entry. | |
1630 | * | |
1631 | * If possible create the dcache entry and derive our inode number and | |
1632 | * file type from dcache entry. | |
1633 | * | |
1634 | * Since all of the proc inode numbers are dynamically generated, the inode | |
1635 | * numbers do not exist until the inode is cache. This means creating the | |
1636 | * the dcache entry in readdir is necessary to keep the inode numbers | |
1637 | * reported by readdir in sync with the inode numbers reported | |
1638 | * by stat. | |
1639 | */ | |
61a28784 EB |
1640 | static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir, |
1641 | char *name, int len, | |
c5141e6d | 1642 | instantiate_t instantiate, struct task_struct *task, const void *ptr) |
61a28784 | 1643 | { |
2fddfeef | 1644 | struct dentry *child, *dir = filp->f_path.dentry; |
61a28784 EB |
1645 | struct inode *inode; |
1646 | struct qstr qname; | |
1647 | ino_t ino = 0; | |
1648 | unsigned type = DT_UNKNOWN; | |
1649 | ||
1650 | qname.name = name; | |
1651 | qname.len = len; | |
1652 | qname.hash = full_name_hash(name, len); | |
1653 | ||
1654 | child = d_lookup(dir, &qname); | |
1655 | if (!child) { | |
1656 | struct dentry *new; | |
1657 | new = d_alloc(dir, &qname); | |
1658 | if (new) { | |
1659 | child = instantiate(dir->d_inode, new, task, ptr); | |
1660 | if (child) | |
1661 | dput(new); | |
1662 | else | |
1663 | child = new; | |
1664 | } | |
1665 | } | |
1666 | if (!child || IS_ERR(child) || !child->d_inode) | |
1667 | goto end_instantiate; | |
1668 | inode = child->d_inode; | |
1669 | if (inode) { | |
1670 | ino = inode->i_ino; | |
1671 | type = inode->i_mode >> 12; | |
1672 | } | |
1673 | dput(child); | |
1674 | end_instantiate: | |
1675 | if (!ino) | |
1676 | ino = find_inode_number(dir, &qname); | |
1677 | if (!ino) | |
1678 | ino = 1; | |
1679 | return filldir(dirent, name, len, filp->f_pos, ino, type); | |
1680 | } | |
1681 | ||
28a6d671 EB |
1682 | static unsigned name_to_int(struct dentry *dentry) |
1683 | { | |
1684 | const char *name = dentry->d_name.name; | |
1685 | int len = dentry->d_name.len; | |
1686 | unsigned n = 0; | |
1687 | ||
1688 | if (len > 1 && *name == '0') | |
1689 | goto out; | |
1690 | while (len-- > 0) { | |
1691 | unsigned c = *name++ - '0'; | |
1692 | if (c > 9) | |
1693 | goto out; | |
1694 | if (n >= (~0U-9)/10) | |
1695 | goto out; | |
1696 | n *= 10; | |
1697 | n += c; | |
1698 | } | |
1699 | return n; | |
1700 | out: | |
1701 | return ~0U; | |
1702 | } | |
1703 | ||
27932742 MS |
1704 | #define PROC_FDINFO_MAX 64 |
1705 | ||
3dcd25f3 | 1706 | static int proc_fd_info(struct inode *inode, struct path *path, char *info) |
28a6d671 EB |
1707 | { |
1708 | struct task_struct *task = get_proc_task(inode); | |
1709 | struct files_struct *files = NULL; | |
1710 | struct file *file; | |
1711 | int fd = proc_fd(inode); | |
99f89551 | 1712 | |
99f89551 | 1713 | if (task) { |
28a6d671 EB |
1714 | files = get_files_struct(task); |
1715 | put_task_struct(task); | |
1716 | } | |
1717 | if (files) { | |
1718 | /* | |
1719 | * We are not taking a ref to the file structure, so we must | |
1720 | * hold ->file_lock. | |
1721 | */ | |
1722 | spin_lock(&files->file_lock); | |
1723 | file = fcheck_files(files, fd); | |
1724 | if (file) { | |
3dcd25f3 JB |
1725 | if (path) { |
1726 | *path = file->f_path; | |
1727 | path_get(&file->f_path); | |
1728 | } | |
27932742 MS |
1729 | if (info) |
1730 | snprintf(info, PROC_FDINFO_MAX, | |
1731 | "pos:\t%lli\n" | |
1732 | "flags:\t0%o\n", | |
1733 | (long long) file->f_pos, | |
1734 | file->f_flags); | |
28a6d671 EB |
1735 | spin_unlock(&files->file_lock); |
1736 | put_files_struct(files); | |
1737 | return 0; | |
99f89551 | 1738 | } |
28a6d671 EB |
1739 | spin_unlock(&files->file_lock); |
1740 | put_files_struct(files); | |
99f89551 | 1741 | } |
28a6d671 | 1742 | return -ENOENT; |
99f89551 EB |
1743 | } |
1744 | ||
3dcd25f3 | 1745 | static int proc_fd_link(struct inode *inode, struct path *path) |
27932742 | 1746 | { |
3dcd25f3 | 1747 | return proc_fd_info(inode, path, NULL); |
27932742 MS |
1748 | } |
1749 | ||
1da177e4 LT |
1750 | static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd) |
1751 | { | |
1752 | struct inode *inode = dentry->d_inode; | |
99f89551 | 1753 | struct task_struct *task = get_proc_task(inode); |
aed7a6c4 | 1754 | int fd = proc_fd(inode); |
1da177e4 | 1755 | struct files_struct *files; |
c69e8d9c | 1756 | const struct cred *cred; |
1da177e4 | 1757 | |
99f89551 EB |
1758 | if (task) { |
1759 | files = get_files_struct(task); | |
1760 | if (files) { | |
1761 | rcu_read_lock(); | |
1762 | if (fcheck_files(files, fd)) { | |
1763 | rcu_read_unlock(); | |
1764 | put_files_struct(files); | |
1765 | if (task_dumpable(task)) { | |
c69e8d9c DH |
1766 | rcu_read_lock(); |
1767 | cred = __task_cred(task); | |
1768 | inode->i_uid = cred->euid; | |
1769 | inode->i_gid = cred->egid; | |
1770 | rcu_read_unlock(); | |
99f89551 EB |
1771 | } else { |
1772 | inode->i_uid = 0; | |
1773 | inode->i_gid = 0; | |
1774 | } | |
9ee8ab9f | 1775 | inode->i_mode &= ~(S_ISUID | S_ISGID); |
99f89551 EB |
1776 | security_task_to_inode(task, inode); |
1777 | put_task_struct(task); | |
1778 | return 1; | |
1779 | } | |
b835996f | 1780 | rcu_read_unlock(); |
1da177e4 | 1781 | put_files_struct(files); |
1da177e4 | 1782 | } |
99f89551 | 1783 | put_task_struct(task); |
1da177e4 LT |
1784 | } |
1785 | d_drop(dentry); | |
1786 | return 0; | |
1787 | } | |
1788 | ||
d72f71eb | 1789 | static const struct dentry_operations tid_fd_dentry_operations = |
1da177e4 LT |
1790 | { |
1791 | .d_revalidate = tid_fd_revalidate, | |
1792 | .d_delete = pid_delete_dentry, | |
1793 | }; | |
1794 | ||
444ceed8 | 1795 | static struct dentry *proc_fd_instantiate(struct inode *dir, |
c5141e6d | 1796 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
1da177e4 | 1797 | { |
c5141e6d | 1798 | unsigned fd = *(const unsigned *)ptr; |
444ceed8 EB |
1799 | struct file *file; |
1800 | struct files_struct *files; | |
1801 | struct inode *inode; | |
1802 | struct proc_inode *ei; | |
1803 | struct dentry *error = ERR_PTR(-ENOENT); | |
1da177e4 | 1804 | |
61a28784 | 1805 | inode = proc_pid_make_inode(dir->i_sb, task); |
1da177e4 LT |
1806 | if (!inode) |
1807 | goto out; | |
1808 | ei = PROC_I(inode); | |
aed7a6c4 | 1809 | ei->fd = fd; |
1da177e4 LT |
1810 | files = get_files_struct(task); |
1811 | if (!files) | |
444ceed8 | 1812 | goto out_iput; |
1da177e4 | 1813 | inode->i_mode = S_IFLNK; |
ca99c1da DS |
1814 | |
1815 | /* | |
1816 | * We are not taking a ref to the file structure, so we must | |
1817 | * hold ->file_lock. | |
1818 | */ | |
1819 | spin_lock(&files->file_lock); | |
1da177e4 LT |
1820 | file = fcheck_files(files, fd); |
1821 | if (!file) | |
444ceed8 | 1822 | goto out_unlock; |
aeb5d727 | 1823 | if (file->f_mode & FMODE_READ) |
1da177e4 | 1824 | inode->i_mode |= S_IRUSR | S_IXUSR; |
aeb5d727 | 1825 | if (file->f_mode & FMODE_WRITE) |
1da177e4 | 1826 | inode->i_mode |= S_IWUSR | S_IXUSR; |
ca99c1da | 1827 | spin_unlock(&files->file_lock); |
1da177e4 | 1828 | put_files_struct(files); |
444ceed8 | 1829 | |
1da177e4 LT |
1830 | inode->i_op = &proc_pid_link_inode_operations; |
1831 | inode->i_size = 64; | |
1832 | ei->op.proc_get_link = proc_fd_link; | |
1833 | dentry->d_op = &tid_fd_dentry_operations; | |
1834 | d_add(dentry, inode); | |
cd6a3ce9 EB |
1835 | /* Close the race of the process dying before we return the dentry */ |
1836 | if (tid_fd_revalidate(dentry, NULL)) | |
444ceed8 | 1837 | error = NULL; |
1da177e4 | 1838 | |
444ceed8 EB |
1839 | out: |
1840 | return error; | |
1841 | out_unlock: | |
ca99c1da | 1842 | spin_unlock(&files->file_lock); |
1da177e4 | 1843 | put_files_struct(files); |
444ceed8 | 1844 | out_iput: |
1da177e4 | 1845 | iput(inode); |
cd6a3ce9 | 1846 | goto out; |
1da177e4 LT |
1847 | } |
1848 | ||
27932742 MS |
1849 | static struct dentry *proc_lookupfd_common(struct inode *dir, |
1850 | struct dentry *dentry, | |
1851 | instantiate_t instantiate) | |
444ceed8 EB |
1852 | { |
1853 | struct task_struct *task = get_proc_task(dir); | |
1854 | unsigned fd = name_to_int(dentry); | |
1855 | struct dentry *result = ERR_PTR(-ENOENT); | |
1856 | ||
1857 | if (!task) | |
1858 | goto out_no_task; | |
1859 | if (fd == ~0U) | |
1860 | goto out; | |
1861 | ||
27932742 | 1862 | result = instantiate(dir, dentry, task, &fd); |
444ceed8 EB |
1863 | out: |
1864 | put_task_struct(task); | |
1865 | out_no_task: | |
1866 | return result; | |
1867 | } | |
1868 | ||
27932742 MS |
1869 | static int proc_readfd_common(struct file * filp, void * dirent, |
1870 | filldir_t filldir, instantiate_t instantiate) | |
28a6d671 | 1871 | { |
2fddfeef | 1872 | struct dentry *dentry = filp->f_path.dentry; |
28a6d671 EB |
1873 | struct inode *inode = dentry->d_inode; |
1874 | struct task_struct *p = get_proc_task(inode); | |
457c2510 | 1875 | unsigned int fd, ino; |
28a6d671 | 1876 | int retval; |
28a6d671 | 1877 | struct files_struct * files; |
1da177e4 | 1878 | |
28a6d671 EB |
1879 | retval = -ENOENT; |
1880 | if (!p) | |
1881 | goto out_no_task; | |
1882 | retval = 0; | |
28a6d671 EB |
1883 | |
1884 | fd = filp->f_pos; | |
1885 | switch (fd) { | |
1886 | case 0: | |
1887 | if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0) | |
1888 | goto out; | |
1889 | filp->f_pos++; | |
1890 | case 1: | |
1891 | ino = parent_ino(dentry); | |
1892 | if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0) | |
1893 | goto out; | |
1894 | filp->f_pos++; | |
1895 | default: | |
1896 | files = get_files_struct(p); | |
1897 | if (!files) | |
1898 | goto out; | |
1899 | rcu_read_lock(); | |
28a6d671 | 1900 | for (fd = filp->f_pos-2; |
9b4f526c | 1901 | fd < files_fdtable(files)->max_fds; |
28a6d671 | 1902 | fd++, filp->f_pos++) { |
27932742 MS |
1903 | char name[PROC_NUMBUF]; |
1904 | int len; | |
28a6d671 EB |
1905 | |
1906 | if (!fcheck_files(files, fd)) | |
1907 | continue; | |
1908 | rcu_read_unlock(); | |
1909 | ||
27932742 MS |
1910 | len = snprintf(name, sizeof(name), "%d", fd); |
1911 | if (proc_fill_cache(filp, dirent, filldir, | |
1912 | name, len, instantiate, | |
1913 | p, &fd) < 0) { | |
28a6d671 EB |
1914 | rcu_read_lock(); |
1915 | break; | |
1916 | } | |
1917 | rcu_read_lock(); | |
1918 | } | |
1919 | rcu_read_unlock(); | |
1920 | put_files_struct(files); | |
1921 | } | |
1922 | out: | |
1923 | put_task_struct(p); | |
1924 | out_no_task: | |
1925 | return retval; | |
1926 | } | |
1927 | ||
27932742 MS |
1928 | static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry, |
1929 | struct nameidata *nd) | |
1930 | { | |
1931 | return proc_lookupfd_common(dir, dentry, proc_fd_instantiate); | |
1932 | } | |
1933 | ||
1934 | static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir) | |
1935 | { | |
1936 | return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate); | |
1937 | } | |
1938 | ||
1939 | static ssize_t proc_fdinfo_read(struct file *file, char __user *buf, | |
1940 | size_t len, loff_t *ppos) | |
1941 | { | |
1942 | char tmp[PROC_FDINFO_MAX]; | |
3dcd25f3 | 1943 | int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp); |
27932742 MS |
1944 | if (!err) |
1945 | err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp)); | |
1946 | return err; | |
1947 | } | |
1948 | ||
1949 | static const struct file_operations proc_fdinfo_file_operations = { | |
1950 | .open = nonseekable_open, | |
1951 | .read = proc_fdinfo_read, | |
1952 | }; | |
1953 | ||
00977a59 | 1954 | static const struct file_operations proc_fd_operations = { |
28a6d671 EB |
1955 | .read = generic_read_dir, |
1956 | .readdir = proc_readfd, | |
1da177e4 LT |
1957 | }; |
1958 | ||
8948e11f AD |
1959 | /* |
1960 | * /proc/pid/fd needs a special permission handler so that a process can still | |
1961 | * access /proc/self/fd after it has executed a setuid(). | |
1962 | */ | |
e6305c43 | 1963 | static int proc_fd_permission(struct inode *inode, int mask) |
8948e11f AD |
1964 | { |
1965 | int rv; | |
1966 | ||
1967 | rv = generic_permission(inode, mask, NULL); | |
1968 | if (rv == 0) | |
1969 | return 0; | |
1970 | if (task_pid(current) == proc_pid(inode)) | |
1971 | rv = 0; | |
1972 | return rv; | |
1973 | } | |
1974 | ||
1da177e4 LT |
1975 | /* |
1976 | * proc directories can do almost nothing.. | |
1977 | */ | |
c5ef1c42 | 1978 | static const struct inode_operations proc_fd_inode_operations = { |
1da177e4 | 1979 | .lookup = proc_lookupfd, |
8948e11f | 1980 | .permission = proc_fd_permission, |
6d76fa58 | 1981 | .setattr = proc_setattr, |
1da177e4 LT |
1982 | }; |
1983 | ||
27932742 MS |
1984 | static struct dentry *proc_fdinfo_instantiate(struct inode *dir, |
1985 | struct dentry *dentry, struct task_struct *task, const void *ptr) | |
1986 | { | |
1987 | unsigned fd = *(unsigned *)ptr; | |
1988 | struct inode *inode; | |
1989 | struct proc_inode *ei; | |
1990 | struct dentry *error = ERR_PTR(-ENOENT); | |
1991 | ||
1992 | inode = proc_pid_make_inode(dir->i_sb, task); | |
1993 | if (!inode) | |
1994 | goto out; | |
1995 | ei = PROC_I(inode); | |
1996 | ei->fd = fd; | |
1997 | inode->i_mode = S_IFREG | S_IRUSR; | |
1998 | inode->i_fop = &proc_fdinfo_file_operations; | |
1999 | dentry->d_op = &tid_fd_dentry_operations; | |
2000 | d_add(dentry, inode); | |
2001 | /* Close the race of the process dying before we return the dentry */ | |
2002 | if (tid_fd_revalidate(dentry, NULL)) | |
2003 | error = NULL; | |
2004 | ||
2005 | out: | |
2006 | return error; | |
2007 | } | |
2008 | ||
2009 | static struct dentry *proc_lookupfdinfo(struct inode *dir, | |
2010 | struct dentry *dentry, | |
2011 | struct nameidata *nd) | |
2012 | { | |
2013 | return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate); | |
2014 | } | |
2015 | ||
2016 | static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir) | |
2017 | { | |
2018 | return proc_readfd_common(filp, dirent, filldir, | |
2019 | proc_fdinfo_instantiate); | |
2020 | } | |
2021 | ||
2022 | static const struct file_operations proc_fdinfo_operations = { | |
2023 | .read = generic_read_dir, | |
2024 | .readdir = proc_readfdinfo, | |
2025 | }; | |
2026 | ||
2027 | /* | |
2028 | * proc directories can do almost nothing.. | |
2029 | */ | |
2030 | static const struct inode_operations proc_fdinfo_inode_operations = { | |
2031 | .lookup = proc_lookupfdinfo, | |
2032 | .setattr = proc_setattr, | |
2033 | }; | |
2034 | ||
2035 | ||
444ceed8 | 2036 | static struct dentry *proc_pident_instantiate(struct inode *dir, |
c5141e6d | 2037 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
444ceed8 | 2038 | { |
c5141e6d | 2039 | const struct pid_entry *p = ptr; |
444ceed8 EB |
2040 | struct inode *inode; |
2041 | struct proc_inode *ei; | |
bd6daba9 | 2042 | struct dentry *error = ERR_PTR(-ENOENT); |
444ceed8 | 2043 | |
61a28784 | 2044 | inode = proc_pid_make_inode(dir->i_sb, task); |
444ceed8 EB |
2045 | if (!inode) |
2046 | goto out; | |
2047 | ||
2048 | ei = PROC_I(inode); | |
2049 | inode->i_mode = p->mode; | |
2050 | if (S_ISDIR(inode->i_mode)) | |
2051 | inode->i_nlink = 2; /* Use getattr to fix if necessary */ | |
2052 | if (p->iop) | |
2053 | inode->i_op = p->iop; | |
2054 | if (p->fop) | |
2055 | inode->i_fop = p->fop; | |
2056 | ei->op = p->op; | |
2057 | dentry->d_op = &pid_dentry_operations; | |
2058 | d_add(dentry, inode); | |
2059 | /* Close the race of the process dying before we return the dentry */ | |
2060 | if (pid_revalidate(dentry, NULL)) | |
2061 | error = NULL; | |
2062 | out: | |
2063 | return error; | |
2064 | } | |
2065 | ||
1da177e4 LT |
2066 | static struct dentry *proc_pident_lookup(struct inode *dir, |
2067 | struct dentry *dentry, | |
c5141e6d | 2068 | const struct pid_entry *ents, |
7bcd6b0e | 2069 | unsigned int nents) |
1da177e4 | 2070 | { |
cd6a3ce9 | 2071 | struct dentry *error; |
99f89551 | 2072 | struct task_struct *task = get_proc_task(dir); |
c5141e6d | 2073 | const struct pid_entry *p, *last; |
1da177e4 | 2074 | |
cd6a3ce9 | 2075 | error = ERR_PTR(-ENOENT); |
1da177e4 | 2076 | |
99f89551 EB |
2077 | if (!task) |
2078 | goto out_no_task; | |
1da177e4 | 2079 | |
20cdc894 EB |
2080 | /* |
2081 | * Yes, it does not scale. And it should not. Don't add | |
2082 | * new entries into /proc/<tgid>/ without very good reasons. | |
2083 | */ | |
7bcd6b0e EB |
2084 | last = &ents[nents - 1]; |
2085 | for (p = ents; p <= last; p++) { | |
1da177e4 LT |
2086 | if (p->len != dentry->d_name.len) |
2087 | continue; | |
2088 | if (!memcmp(dentry->d_name.name, p->name, p->len)) | |
2089 | break; | |
2090 | } | |
7bcd6b0e | 2091 | if (p > last) |
1da177e4 LT |
2092 | goto out; |
2093 | ||
444ceed8 | 2094 | error = proc_pident_instantiate(dir, dentry, task, p); |
1da177e4 | 2095 | out: |
99f89551 EB |
2096 | put_task_struct(task); |
2097 | out_no_task: | |
cd6a3ce9 | 2098 | return error; |
1da177e4 LT |
2099 | } |
2100 | ||
c5141e6d ED |
2101 | static int proc_pident_fill_cache(struct file *filp, void *dirent, |
2102 | filldir_t filldir, struct task_struct *task, const struct pid_entry *p) | |
61a28784 EB |
2103 | { |
2104 | return proc_fill_cache(filp, dirent, filldir, p->name, p->len, | |
2105 | proc_pident_instantiate, task, p); | |
2106 | } | |
2107 | ||
28a6d671 EB |
2108 | static int proc_pident_readdir(struct file *filp, |
2109 | void *dirent, filldir_t filldir, | |
c5141e6d | 2110 | const struct pid_entry *ents, unsigned int nents) |
28a6d671 EB |
2111 | { |
2112 | int i; | |
2fddfeef | 2113 | struct dentry *dentry = filp->f_path.dentry; |
28a6d671 EB |
2114 | struct inode *inode = dentry->d_inode; |
2115 | struct task_struct *task = get_proc_task(inode); | |
c5141e6d | 2116 | const struct pid_entry *p, *last; |
28a6d671 EB |
2117 | ino_t ino; |
2118 | int ret; | |
2119 | ||
2120 | ret = -ENOENT; | |
2121 | if (!task) | |
61a28784 | 2122 | goto out_no_task; |
28a6d671 EB |
2123 | |
2124 | ret = 0; | |
28a6d671 EB |
2125 | i = filp->f_pos; |
2126 | switch (i) { | |
2127 | case 0: | |
2128 | ino = inode->i_ino; | |
2129 | if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0) | |
2130 | goto out; | |
2131 | i++; | |
2132 | filp->f_pos++; | |
2133 | /* fall through */ | |
2134 | case 1: | |
2135 | ino = parent_ino(dentry); | |
2136 | if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0) | |
2137 | goto out; | |
2138 | i++; | |
2139 | filp->f_pos++; | |
2140 | /* fall through */ | |
2141 | default: | |
2142 | i -= 2; | |
2143 | if (i >= nents) { | |
2144 | ret = 1; | |
2145 | goto out; | |
2146 | } | |
2147 | p = ents + i; | |
7bcd6b0e EB |
2148 | last = &ents[nents - 1]; |
2149 | while (p <= last) { | |
61a28784 | 2150 | if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0) |
28a6d671 EB |
2151 | goto out; |
2152 | filp->f_pos++; | |
2153 | p++; | |
2154 | } | |
2155 | } | |
2156 | ||
2157 | ret = 1; | |
2158 | out: | |
61a28784 EB |
2159 | put_task_struct(task); |
2160 | out_no_task: | |
28a6d671 | 2161 | return ret; |
1da177e4 LT |
2162 | } |
2163 | ||
28a6d671 EB |
2164 | #ifdef CONFIG_SECURITY |
2165 | static ssize_t proc_pid_attr_read(struct file * file, char __user * buf, | |
2166 | size_t count, loff_t *ppos) | |
2167 | { | |
2fddfeef | 2168 | struct inode * inode = file->f_path.dentry->d_inode; |
04ff9708 | 2169 | char *p = NULL; |
28a6d671 EB |
2170 | ssize_t length; |
2171 | struct task_struct *task = get_proc_task(inode); | |
2172 | ||
28a6d671 | 2173 | if (!task) |
04ff9708 | 2174 | return -ESRCH; |
28a6d671 EB |
2175 | |
2176 | length = security_getprocattr(task, | |
2fddfeef | 2177 | (char*)file->f_path.dentry->d_name.name, |
04ff9708 | 2178 | &p); |
28a6d671 | 2179 | put_task_struct(task); |
04ff9708 AV |
2180 | if (length > 0) |
2181 | length = simple_read_from_buffer(buf, count, ppos, p, length); | |
2182 | kfree(p); | |
28a6d671 | 2183 | return length; |
1da177e4 LT |
2184 | } |
2185 | ||
28a6d671 EB |
2186 | static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf, |
2187 | size_t count, loff_t *ppos) | |
2188 | { | |
2fddfeef | 2189 | struct inode * inode = file->f_path.dentry->d_inode; |
28a6d671 EB |
2190 | char *page; |
2191 | ssize_t length; | |
2192 | struct task_struct *task = get_proc_task(inode); | |
2193 | ||
2194 | length = -ESRCH; | |
2195 | if (!task) | |
2196 | goto out_no_task; | |
2197 | if (count > PAGE_SIZE) | |
2198 | count = PAGE_SIZE; | |
2199 | ||
2200 | /* No partial writes. */ | |
2201 | length = -EINVAL; | |
2202 | if (*ppos != 0) | |
2203 | goto out; | |
2204 | ||
2205 | length = -ENOMEM; | |
e12ba74d | 2206 | page = (char*)__get_free_page(GFP_TEMPORARY); |
28a6d671 EB |
2207 | if (!page) |
2208 | goto out; | |
2209 | ||
2210 | length = -EFAULT; | |
2211 | if (copy_from_user(page, buf, count)) | |
2212 | goto out_free; | |
2213 | ||
107db7c7 DH |
2214 | /* Guard against adverse ptrace interaction */ |
2215 | length = mutex_lock_interruptible(&task->cred_guard_mutex); | |
2216 | if (length < 0) | |
2217 | goto out_free; | |
2218 | ||
28a6d671 | 2219 | length = security_setprocattr(task, |
2fddfeef | 2220 | (char*)file->f_path.dentry->d_name.name, |
28a6d671 | 2221 | (void*)page, count); |
107db7c7 | 2222 | mutex_unlock(&task->cred_guard_mutex); |
28a6d671 EB |
2223 | out_free: |
2224 | free_page((unsigned long) page); | |
2225 | out: | |
2226 | put_task_struct(task); | |
2227 | out_no_task: | |
2228 | return length; | |
2229 | } | |
2230 | ||
00977a59 | 2231 | static const struct file_operations proc_pid_attr_operations = { |
28a6d671 EB |
2232 | .read = proc_pid_attr_read, |
2233 | .write = proc_pid_attr_write, | |
2234 | }; | |
2235 | ||
c5141e6d | 2236 | static const struct pid_entry attr_dir_stuff[] = { |
631f9c18 AD |
2237 | REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations), |
2238 | REG("prev", S_IRUGO, proc_pid_attr_operations), | |
2239 | REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2240 | REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2241 | REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2242 | REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
28a6d671 EB |
2243 | }; |
2244 | ||
72d9dcfc | 2245 | static int proc_attr_dir_readdir(struct file * filp, |
28a6d671 EB |
2246 | void * dirent, filldir_t filldir) |
2247 | { | |
2248 | return proc_pident_readdir(filp,dirent,filldir, | |
72d9dcfc | 2249 | attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff)); |
28a6d671 EB |
2250 | } |
2251 | ||
00977a59 | 2252 | static const struct file_operations proc_attr_dir_operations = { |
1da177e4 | 2253 | .read = generic_read_dir, |
72d9dcfc | 2254 | .readdir = proc_attr_dir_readdir, |
1da177e4 LT |
2255 | }; |
2256 | ||
72d9dcfc | 2257 | static struct dentry *proc_attr_dir_lookup(struct inode *dir, |
28a6d671 EB |
2258 | struct dentry *dentry, struct nameidata *nd) |
2259 | { | |
7bcd6b0e EB |
2260 | return proc_pident_lookup(dir, dentry, |
2261 | attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2262 | } |
2263 | ||
c5ef1c42 | 2264 | static const struct inode_operations proc_attr_dir_inode_operations = { |
72d9dcfc | 2265 | .lookup = proc_attr_dir_lookup, |
99f89551 | 2266 | .getattr = pid_getattr, |
6d76fa58 | 2267 | .setattr = proc_setattr, |
1da177e4 LT |
2268 | }; |
2269 | ||
28a6d671 EB |
2270 | #endif |
2271 | ||
698ba7b5 | 2272 | #ifdef CONFIG_ELF_CORE |
3cb4a0bb KH |
2273 | static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf, |
2274 | size_t count, loff_t *ppos) | |
2275 | { | |
2276 | struct task_struct *task = get_proc_task(file->f_dentry->d_inode); | |
2277 | struct mm_struct *mm; | |
2278 | char buffer[PROC_NUMBUF]; | |
2279 | size_t len; | |
2280 | int ret; | |
2281 | ||
2282 | if (!task) | |
2283 | return -ESRCH; | |
2284 | ||
2285 | ret = 0; | |
2286 | mm = get_task_mm(task); | |
2287 | if (mm) { | |
2288 | len = snprintf(buffer, sizeof(buffer), "%08lx\n", | |
2289 | ((mm->flags & MMF_DUMP_FILTER_MASK) >> | |
2290 | MMF_DUMP_FILTER_SHIFT)); | |
2291 | mmput(mm); | |
2292 | ret = simple_read_from_buffer(buf, count, ppos, buffer, len); | |
2293 | } | |
2294 | ||
2295 | put_task_struct(task); | |
2296 | ||
2297 | return ret; | |
2298 | } | |
2299 | ||
2300 | static ssize_t proc_coredump_filter_write(struct file *file, | |
2301 | const char __user *buf, | |
2302 | size_t count, | |
2303 | loff_t *ppos) | |
2304 | { | |
2305 | struct task_struct *task; | |
2306 | struct mm_struct *mm; | |
2307 | char buffer[PROC_NUMBUF], *end; | |
2308 | unsigned int val; | |
2309 | int ret; | |
2310 | int i; | |
2311 | unsigned long mask; | |
2312 | ||
2313 | ret = -EFAULT; | |
2314 | memset(buffer, 0, sizeof(buffer)); | |
2315 | if (count > sizeof(buffer) - 1) | |
2316 | count = sizeof(buffer) - 1; | |
2317 | if (copy_from_user(buffer, buf, count)) | |
2318 | goto out_no_task; | |
2319 | ||
2320 | ret = -EINVAL; | |
2321 | val = (unsigned int)simple_strtoul(buffer, &end, 0); | |
2322 | if (*end == '\n') | |
2323 | end++; | |
2324 | if (end - buffer == 0) | |
2325 | goto out_no_task; | |
2326 | ||
2327 | ret = -ESRCH; | |
2328 | task = get_proc_task(file->f_dentry->d_inode); | |
2329 | if (!task) | |
2330 | goto out_no_task; | |
2331 | ||
2332 | ret = end - buffer; | |
2333 | mm = get_task_mm(task); | |
2334 | if (!mm) | |
2335 | goto out_no_mm; | |
2336 | ||
2337 | for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) { | |
2338 | if (val & mask) | |
2339 | set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2340 | else | |
2341 | clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2342 | } | |
2343 | ||
2344 | mmput(mm); | |
2345 | out_no_mm: | |
2346 | put_task_struct(task); | |
2347 | out_no_task: | |
2348 | return ret; | |
2349 | } | |
2350 | ||
2351 | static const struct file_operations proc_coredump_filter_operations = { | |
2352 | .read = proc_coredump_filter_read, | |
2353 | .write = proc_coredump_filter_write, | |
2354 | }; | |
2355 | #endif | |
2356 | ||
28a6d671 EB |
2357 | /* |
2358 | * /proc/self: | |
2359 | */ | |
2360 | static int proc_self_readlink(struct dentry *dentry, char __user *buffer, | |
2361 | int buflen) | |
2362 | { | |
488e5bc4 | 2363 | struct pid_namespace *ns = dentry->d_sb->s_fs_info; |
b55fcb22 | 2364 | pid_t tgid = task_tgid_nr_ns(current, ns); |
28a6d671 | 2365 | char tmp[PROC_NUMBUF]; |
b55fcb22 | 2366 | if (!tgid) |
488e5bc4 | 2367 | return -ENOENT; |
b55fcb22 | 2368 | sprintf(tmp, "%d", tgid); |
28a6d671 EB |
2369 | return vfs_readlink(dentry,buffer,buflen,tmp); |
2370 | } | |
2371 | ||
2372 | static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd) | |
2373 | { | |
488e5bc4 | 2374 | struct pid_namespace *ns = dentry->d_sb->s_fs_info; |
b55fcb22 | 2375 | pid_t tgid = task_tgid_nr_ns(current, ns); |
7fee4868 AV |
2376 | char *name = ERR_PTR(-ENOENT); |
2377 | if (tgid) { | |
2378 | name = __getname(); | |
2379 | if (!name) | |
2380 | name = ERR_PTR(-ENOMEM); | |
2381 | else | |
2382 | sprintf(name, "%d", tgid); | |
2383 | } | |
2384 | nd_set_link(nd, name); | |
2385 | return NULL; | |
2386 | } | |
2387 | ||
2388 | static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd, | |
2389 | void *cookie) | |
2390 | { | |
2391 | char *s = nd_get_link(nd); | |
2392 | if (!IS_ERR(s)) | |
2393 | __putname(s); | |
28a6d671 EB |
2394 | } |
2395 | ||
c5ef1c42 | 2396 | static const struct inode_operations proc_self_inode_operations = { |
28a6d671 EB |
2397 | .readlink = proc_self_readlink, |
2398 | .follow_link = proc_self_follow_link, | |
7fee4868 | 2399 | .put_link = proc_self_put_link, |
28a6d671 EB |
2400 | }; |
2401 | ||
801199ce EB |
2402 | /* |
2403 | * proc base | |
2404 | * | |
2405 | * These are the directory entries in the root directory of /proc | |
2406 | * that properly belong to the /proc filesystem, as they describe | |
2407 | * describe something that is process related. | |
2408 | */ | |
c5141e6d | 2409 | static const struct pid_entry proc_base_stuff[] = { |
61a28784 | 2410 | NOD("self", S_IFLNK|S_IRWXUGO, |
801199ce | 2411 | &proc_self_inode_operations, NULL, {}), |
801199ce EB |
2412 | }; |
2413 | ||
2414 | /* | |
2415 | * Exceptional case: normally we are not allowed to unhash a busy | |
2416 | * directory. In this case, however, we can do it - no aliasing problems | |
2417 | * due to the way we treat inodes. | |
2418 | */ | |
2419 | static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd) | |
2420 | { | |
2421 | struct inode *inode = dentry->d_inode; | |
2422 | struct task_struct *task = get_proc_task(inode); | |
2423 | if (task) { | |
2424 | put_task_struct(task); | |
2425 | return 1; | |
2426 | } | |
2427 | d_drop(dentry); | |
2428 | return 0; | |
2429 | } | |
2430 | ||
d72f71eb | 2431 | static const struct dentry_operations proc_base_dentry_operations = |
801199ce EB |
2432 | { |
2433 | .d_revalidate = proc_base_revalidate, | |
2434 | .d_delete = pid_delete_dentry, | |
2435 | }; | |
2436 | ||
444ceed8 | 2437 | static struct dentry *proc_base_instantiate(struct inode *dir, |
c5141e6d | 2438 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
801199ce | 2439 | { |
c5141e6d | 2440 | const struct pid_entry *p = ptr; |
801199ce | 2441 | struct inode *inode; |
801199ce | 2442 | struct proc_inode *ei; |
444ceed8 | 2443 | struct dentry *error = ERR_PTR(-EINVAL); |
801199ce EB |
2444 | |
2445 | /* Allocate the inode */ | |
2446 | error = ERR_PTR(-ENOMEM); | |
2447 | inode = new_inode(dir->i_sb); | |
2448 | if (!inode) | |
2449 | goto out; | |
2450 | ||
2451 | /* Initialize the inode */ | |
2452 | ei = PROC_I(inode); | |
2453 | inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; | |
801199ce EB |
2454 | |
2455 | /* | |
2456 | * grab the reference to the task. | |
2457 | */ | |
1a657f78 | 2458 | ei->pid = get_task_pid(task, PIDTYPE_PID); |
801199ce EB |
2459 | if (!ei->pid) |
2460 | goto out_iput; | |
2461 | ||
801199ce EB |
2462 | inode->i_mode = p->mode; |
2463 | if (S_ISDIR(inode->i_mode)) | |
2464 | inode->i_nlink = 2; | |
2465 | if (S_ISLNK(inode->i_mode)) | |
2466 | inode->i_size = 64; | |
2467 | if (p->iop) | |
2468 | inode->i_op = p->iop; | |
2469 | if (p->fop) | |
2470 | inode->i_fop = p->fop; | |
2471 | ei->op = p->op; | |
2472 | dentry->d_op = &proc_base_dentry_operations; | |
2473 | d_add(dentry, inode); | |
2474 | error = NULL; | |
2475 | out: | |
801199ce EB |
2476 | return error; |
2477 | out_iput: | |
2478 | iput(inode); | |
2479 | goto out; | |
2480 | } | |
2481 | ||
444ceed8 EB |
2482 | static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry) |
2483 | { | |
2484 | struct dentry *error; | |
2485 | struct task_struct *task = get_proc_task(dir); | |
c5141e6d | 2486 | const struct pid_entry *p, *last; |
444ceed8 EB |
2487 | |
2488 | error = ERR_PTR(-ENOENT); | |
2489 | ||
2490 | if (!task) | |
2491 | goto out_no_task; | |
2492 | ||
2493 | /* Lookup the directory entry */ | |
7bcd6b0e EB |
2494 | last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1]; |
2495 | for (p = proc_base_stuff; p <= last; p++) { | |
444ceed8 EB |
2496 | if (p->len != dentry->d_name.len) |
2497 | continue; | |
2498 | if (!memcmp(dentry->d_name.name, p->name, p->len)) | |
2499 | break; | |
2500 | } | |
7bcd6b0e | 2501 | if (p > last) |
444ceed8 EB |
2502 | goto out; |
2503 | ||
2504 | error = proc_base_instantiate(dir, dentry, task, p); | |
2505 | ||
2506 | out: | |
2507 | put_task_struct(task); | |
2508 | out_no_task: | |
2509 | return error; | |
2510 | } | |
2511 | ||
c5141e6d ED |
2512 | static int proc_base_fill_cache(struct file *filp, void *dirent, |
2513 | filldir_t filldir, struct task_struct *task, const struct pid_entry *p) | |
61a28784 EB |
2514 | { |
2515 | return proc_fill_cache(filp, dirent, filldir, p->name, p->len, | |
2516 | proc_base_instantiate, task, p); | |
2517 | } | |
2518 | ||
aba76fdb | 2519 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
297c5d92 AR |
2520 | static int do_io_accounting(struct task_struct *task, char *buffer, int whole) |
2521 | { | |
940389b8 | 2522 | struct task_io_accounting acct = task->ioac; |
5995477a AR |
2523 | unsigned long flags; |
2524 | ||
2525 | if (whole && lock_task_sighand(task, &flags)) { | |
2526 | struct task_struct *t = task; | |
2527 | ||
2528 | task_io_accounting_add(&acct, &task->signal->ioac); | |
2529 | while_each_thread(task, t) | |
2530 | task_io_accounting_add(&acct, &t->ioac); | |
2531 | ||
2532 | unlock_task_sighand(task, &flags); | |
297c5d92 | 2533 | } |
aba76fdb AM |
2534 | return sprintf(buffer, |
2535 | "rchar: %llu\n" | |
2536 | "wchar: %llu\n" | |
2537 | "syscr: %llu\n" | |
2538 | "syscw: %llu\n" | |
2539 | "read_bytes: %llu\n" | |
2540 | "write_bytes: %llu\n" | |
2541 | "cancelled_write_bytes: %llu\n", | |
7c44319d AB |
2542 | (unsigned long long)acct.rchar, |
2543 | (unsigned long long)acct.wchar, | |
2544 | (unsigned long long)acct.syscr, | |
2545 | (unsigned long long)acct.syscw, | |
2546 | (unsigned long long)acct.read_bytes, | |
2547 | (unsigned long long)acct.write_bytes, | |
2548 | (unsigned long long)acct.cancelled_write_bytes); | |
297c5d92 AR |
2549 | } |
2550 | ||
2551 | static int proc_tid_io_accounting(struct task_struct *task, char *buffer) | |
2552 | { | |
2553 | return do_io_accounting(task, buffer, 0); | |
aba76fdb | 2554 | } |
297c5d92 AR |
2555 | |
2556 | static int proc_tgid_io_accounting(struct task_struct *task, char *buffer) | |
2557 | { | |
2558 | return do_io_accounting(task, buffer, 1); | |
2559 | } | |
2560 | #endif /* CONFIG_TASK_IO_ACCOUNTING */ | |
aba76fdb | 2561 | |
47830723 KC |
2562 | static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns, |
2563 | struct pid *pid, struct task_struct *task) | |
2564 | { | |
2565 | seq_printf(m, "%08x\n", task->personality); | |
2566 | return 0; | |
2567 | } | |
2568 | ||
28a6d671 EB |
2569 | /* |
2570 | * Thread groups | |
2571 | */ | |
00977a59 | 2572 | static const struct file_operations proc_task_operations; |
c5ef1c42 | 2573 | static const struct inode_operations proc_task_inode_operations; |
20cdc894 | 2574 | |
c5141e6d | 2575 | static const struct pid_entry tgid_base_stuff[] = { |
631f9c18 AD |
2576 | DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations), |
2577 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), | |
2578 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), | |
b2211a36 | 2579 | #ifdef CONFIG_NET |
631f9c18 | 2580 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), |
b2211a36 | 2581 | #endif |
631f9c18 AD |
2582 | REG("environ", S_IRUSR, proc_environ_operations), |
2583 | INF("auxv", S_IRUSR, proc_pid_auxv), | |
2584 | ONE("status", S_IRUGO, proc_pid_status), | |
2585 | ONE("personality", S_IRUSR, proc_pid_personality), | |
2586 | INF("limits", S_IRUSR, proc_pid_limits), | |
43ae34cb | 2587 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 2588 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
ebcb6734 | 2589 | #endif |
4614a696 | 2590 | REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations), |
ebcb6734 | 2591 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
631f9c18 | 2592 | INF("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 2593 | #endif |
631f9c18 AD |
2594 | INF("cmdline", S_IRUGO, proc_pid_cmdline), |
2595 | ONE("stat", S_IRUGO, proc_tgid_stat), | |
2596 | ONE("statm", S_IRUGO, proc_pid_statm), | |
2597 | REG("maps", S_IRUGO, proc_maps_operations), | |
28a6d671 | 2598 | #ifdef CONFIG_NUMA |
631f9c18 | 2599 | REG("numa_maps", S_IRUGO, proc_numa_maps_operations), |
28a6d671 | 2600 | #endif |
631f9c18 AD |
2601 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
2602 | LNK("cwd", proc_cwd_link), | |
2603 | LNK("root", proc_root_link), | |
2604 | LNK("exe", proc_exe_link), | |
2605 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
2606 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
2607 | REG("mountstats", S_IRUSR, proc_mountstats_operations), | |
1e883281 | 2608 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 AD |
2609 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
2610 | REG("smaps", S_IRUGO, proc_smaps_operations), | |
2611 | REG("pagemap", S_IRUSR, proc_pagemap_operations), | |
28a6d671 EB |
2612 | #endif |
2613 | #ifdef CONFIG_SECURITY | |
631f9c18 | 2614 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
2615 | #endif |
2616 | #ifdef CONFIG_KALLSYMS | |
631f9c18 | 2617 | INF("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 2618 | #endif |
2ec220e2 KC |
2619 | #ifdef CONFIG_STACKTRACE |
2620 | ONE("stack", S_IRUSR, proc_pid_stack), | |
28a6d671 EB |
2621 | #endif |
2622 | #ifdef CONFIG_SCHEDSTATS | |
631f9c18 | 2623 | INF("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 2624 | #endif |
9745512c | 2625 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 2626 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 2627 | #endif |
8793d854 | 2628 | #ifdef CONFIG_PROC_PID_CPUSET |
631f9c18 | 2629 | REG("cpuset", S_IRUGO, proc_cpuset_operations), |
a424316c PM |
2630 | #endif |
2631 | #ifdef CONFIG_CGROUPS | |
631f9c18 | 2632 | REG("cgroup", S_IRUGO, proc_cgroup_operations), |
28a6d671 | 2633 | #endif |
631f9c18 AD |
2634 | INF("oom_score", S_IRUGO, proc_oom_score), |
2635 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations), | |
28a6d671 | 2636 | #ifdef CONFIG_AUDITSYSCALL |
631f9c18 AD |
2637 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
2638 | REG("sessionid", S_IRUGO, proc_sessionid_operations), | |
28a6d671 | 2639 | #endif |
f4f154fd | 2640 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 2641 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
f4f154fd | 2642 | #endif |
698ba7b5 | 2643 | #ifdef CONFIG_ELF_CORE |
631f9c18 | 2644 | REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations), |
3cb4a0bb | 2645 | #endif |
aba76fdb | 2646 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
631f9c18 | 2647 | INF("io", S_IRUGO, proc_tgid_io_accounting), |
aba76fdb | 2648 | #endif |
28a6d671 | 2649 | }; |
1da177e4 | 2650 | |
28a6d671 | 2651 | static int proc_tgid_base_readdir(struct file * filp, |
1da177e4 LT |
2652 | void * dirent, filldir_t filldir) |
2653 | { | |
2654 | return proc_pident_readdir(filp,dirent,filldir, | |
28a6d671 | 2655 | tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff)); |
1da177e4 LT |
2656 | } |
2657 | ||
00977a59 | 2658 | static const struct file_operations proc_tgid_base_operations = { |
1da177e4 | 2659 | .read = generic_read_dir, |
28a6d671 | 2660 | .readdir = proc_tgid_base_readdir, |
1da177e4 LT |
2661 | }; |
2662 | ||
28a6d671 | 2663 | static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){ |
7bcd6b0e EB |
2664 | return proc_pident_lookup(dir, dentry, |
2665 | tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
2666 | } |
2667 | ||
c5ef1c42 | 2668 | static const struct inode_operations proc_tgid_base_inode_operations = { |
28a6d671 | 2669 | .lookup = proc_tgid_base_lookup, |
99f89551 | 2670 | .getattr = pid_getattr, |
6d76fa58 | 2671 | .setattr = proc_setattr, |
1da177e4 | 2672 | }; |
1da177e4 | 2673 | |
60347f67 | 2674 | static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid) |
1da177e4 | 2675 | { |
48e6484d | 2676 | struct dentry *dentry, *leader, *dir; |
8578cea7 | 2677 | char buf[PROC_NUMBUF]; |
48e6484d EB |
2678 | struct qstr name; |
2679 | ||
2680 | name.name = buf; | |
60347f67 PE |
2681 | name.len = snprintf(buf, sizeof(buf), "%d", pid); |
2682 | dentry = d_hash_and_lookup(mnt->mnt_root, &name); | |
48e6484d | 2683 | if (dentry) { |
29f12ca3 | 2684 | shrink_dcache_parent(dentry); |
48e6484d EB |
2685 | d_drop(dentry); |
2686 | dput(dentry); | |
2687 | } | |
1da177e4 | 2688 | |
48e6484d | 2689 | name.name = buf; |
60347f67 PE |
2690 | name.len = snprintf(buf, sizeof(buf), "%d", tgid); |
2691 | leader = d_hash_and_lookup(mnt->mnt_root, &name); | |
48e6484d EB |
2692 | if (!leader) |
2693 | goto out; | |
1da177e4 | 2694 | |
48e6484d EB |
2695 | name.name = "task"; |
2696 | name.len = strlen(name.name); | |
2697 | dir = d_hash_and_lookup(leader, &name); | |
2698 | if (!dir) | |
2699 | goto out_put_leader; | |
2700 | ||
2701 | name.name = buf; | |
60347f67 | 2702 | name.len = snprintf(buf, sizeof(buf), "%d", pid); |
48e6484d EB |
2703 | dentry = d_hash_and_lookup(dir, &name); |
2704 | if (dentry) { | |
2705 | shrink_dcache_parent(dentry); | |
2706 | d_drop(dentry); | |
2707 | dput(dentry); | |
1da177e4 | 2708 | } |
48e6484d EB |
2709 | |
2710 | dput(dir); | |
2711 | out_put_leader: | |
2712 | dput(leader); | |
2713 | out: | |
2714 | return; | |
1da177e4 LT |
2715 | } |
2716 | ||
0895e91d RD |
2717 | /** |
2718 | * proc_flush_task - Remove dcache entries for @task from the /proc dcache. | |
2719 | * @task: task that should be flushed. | |
2720 | * | |
2721 | * When flushing dentries from proc, one needs to flush them from global | |
60347f67 | 2722 | * proc (proc_mnt) and from all the namespaces' procs this task was seen |
0895e91d RD |
2723 | * in. This call is supposed to do all of this job. |
2724 | * | |
2725 | * Looks in the dcache for | |
2726 | * /proc/@pid | |
2727 | * /proc/@tgid/task/@pid | |
2728 | * if either directory is present flushes it and all of it'ts children | |
2729 | * from the dcache. | |
2730 | * | |
2731 | * It is safe and reasonable to cache /proc entries for a task until | |
2732 | * that task exits. After that they just clog up the dcache with | |
2733 | * useless entries, possibly causing useful dcache entries to be | |
2734 | * flushed instead. This routine is proved to flush those useless | |
2735 | * dcache entries at process exit time. | |
2736 | * | |
2737 | * NOTE: This routine is just an optimization so it does not guarantee | |
2738 | * that no dcache entries will exist at process exit time it | |
2739 | * just makes it very unlikely that any will persist. | |
60347f67 PE |
2740 | */ |
2741 | ||
2742 | void proc_flush_task(struct task_struct *task) | |
2743 | { | |
9fcc2d15 | 2744 | int i; |
9b4d1cbe | 2745 | struct pid *pid, *tgid; |
130f77ec PE |
2746 | struct upid *upid; |
2747 | ||
130f77ec | 2748 | pid = task_pid(task); |
9b4d1cbe | 2749 | tgid = task_tgid(task); |
130f77ec | 2750 | |
9fcc2d15 | 2751 | for (i = 0; i <= pid->level; i++) { |
130f77ec PE |
2752 | upid = &pid->numbers[i]; |
2753 | proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr, | |
9b4d1cbe | 2754 | tgid->numbers[i].nr); |
130f77ec | 2755 | } |
6f4e6433 PE |
2756 | |
2757 | upid = &pid->numbers[pid->level]; | |
2758 | if (upid->nr == 1) | |
2759 | pid_ns_release_proc(upid->ns); | |
60347f67 PE |
2760 | } |
2761 | ||
9711ef99 AB |
2762 | static struct dentry *proc_pid_instantiate(struct inode *dir, |
2763 | struct dentry * dentry, | |
c5141e6d | 2764 | struct task_struct *task, const void *ptr) |
444ceed8 EB |
2765 | { |
2766 | struct dentry *error = ERR_PTR(-ENOENT); | |
2767 | struct inode *inode; | |
2768 | ||
61a28784 | 2769 | inode = proc_pid_make_inode(dir->i_sb, task); |
444ceed8 EB |
2770 | if (!inode) |
2771 | goto out; | |
2772 | ||
2773 | inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO; | |
2774 | inode->i_op = &proc_tgid_base_inode_operations; | |
2775 | inode->i_fop = &proc_tgid_base_operations; | |
2776 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 VN |
2777 | |
2778 | inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff, | |
2779 | ARRAY_SIZE(tgid_base_stuff)); | |
444ceed8 EB |
2780 | |
2781 | dentry->d_op = &pid_dentry_operations; | |
2782 | ||
2783 | d_add(dentry, inode); | |
2784 | /* Close the race of the process dying before we return the dentry */ | |
2785 | if (pid_revalidate(dentry, NULL)) | |
2786 | error = NULL; | |
2787 | out: | |
2788 | return error; | |
2789 | } | |
2790 | ||
1da177e4 LT |
2791 | struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd) |
2792 | { | |
cd6a3ce9 | 2793 | struct dentry *result = ERR_PTR(-ENOENT); |
1da177e4 | 2794 | struct task_struct *task; |
1da177e4 | 2795 | unsigned tgid; |
b488893a | 2796 | struct pid_namespace *ns; |
1da177e4 | 2797 | |
801199ce EB |
2798 | result = proc_base_lookup(dir, dentry); |
2799 | if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT) | |
2800 | goto out; | |
2801 | ||
1da177e4 LT |
2802 | tgid = name_to_int(dentry); |
2803 | if (tgid == ~0U) | |
2804 | goto out; | |
2805 | ||
b488893a | 2806 | ns = dentry->d_sb->s_fs_info; |
de758734 | 2807 | rcu_read_lock(); |
b488893a | 2808 | task = find_task_by_pid_ns(tgid, ns); |
1da177e4 LT |
2809 | if (task) |
2810 | get_task_struct(task); | |
de758734 | 2811 | rcu_read_unlock(); |
1da177e4 LT |
2812 | if (!task) |
2813 | goto out; | |
2814 | ||
444ceed8 | 2815 | result = proc_pid_instantiate(dir, dentry, task, NULL); |
1da177e4 | 2816 | put_task_struct(task); |
1da177e4 | 2817 | out: |
cd6a3ce9 | 2818 | return result; |
1da177e4 LT |
2819 | } |
2820 | ||
1da177e4 | 2821 | /* |
0804ef4b | 2822 | * Find the first task with tgid >= tgid |
0bc58a91 | 2823 | * |
1da177e4 | 2824 | */ |
19fd4bb2 EB |
2825 | struct tgid_iter { |
2826 | unsigned int tgid; | |
0804ef4b | 2827 | struct task_struct *task; |
19fd4bb2 EB |
2828 | }; |
2829 | static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter) | |
2830 | { | |
0804ef4b | 2831 | struct pid *pid; |
1da177e4 | 2832 | |
19fd4bb2 EB |
2833 | if (iter.task) |
2834 | put_task_struct(iter.task); | |
454cc105 | 2835 | rcu_read_lock(); |
0804ef4b | 2836 | retry: |
19fd4bb2 EB |
2837 | iter.task = NULL; |
2838 | pid = find_ge_pid(iter.tgid, ns); | |
0804ef4b | 2839 | if (pid) { |
19fd4bb2 EB |
2840 | iter.tgid = pid_nr_ns(pid, ns); |
2841 | iter.task = pid_task(pid, PIDTYPE_PID); | |
0804ef4b EB |
2842 | /* What we to know is if the pid we have find is the |
2843 | * pid of a thread_group_leader. Testing for task | |
2844 | * being a thread_group_leader is the obvious thing | |
2845 | * todo but there is a window when it fails, due to | |
2846 | * the pid transfer logic in de_thread. | |
2847 | * | |
2848 | * So we perform the straight forward test of seeing | |
2849 | * if the pid we have found is the pid of a thread | |
2850 | * group leader, and don't worry if the task we have | |
2851 | * found doesn't happen to be a thread group leader. | |
2852 | * As we don't care in the case of readdir. | |
2853 | */ | |
19fd4bb2 EB |
2854 | if (!iter.task || !has_group_leader_pid(iter.task)) { |
2855 | iter.tgid += 1; | |
0804ef4b | 2856 | goto retry; |
19fd4bb2 EB |
2857 | } |
2858 | get_task_struct(iter.task); | |
0bc58a91 | 2859 | } |
454cc105 | 2860 | rcu_read_unlock(); |
19fd4bb2 | 2861 | return iter; |
1da177e4 LT |
2862 | } |
2863 | ||
7bcd6b0e | 2864 | #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff)) |
0804ef4b | 2865 | |
61a28784 | 2866 | static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir, |
19fd4bb2 | 2867 | struct tgid_iter iter) |
61a28784 EB |
2868 | { |
2869 | char name[PROC_NUMBUF]; | |
19fd4bb2 | 2870 | int len = snprintf(name, sizeof(name), "%d", iter.tgid); |
61a28784 | 2871 | return proc_fill_cache(filp, dirent, filldir, name, len, |
19fd4bb2 | 2872 | proc_pid_instantiate, iter.task, NULL); |
61a28784 EB |
2873 | } |
2874 | ||
1da177e4 LT |
2875 | /* for the /proc/ directory itself, after non-process stuff has been done */ |
2876 | int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir) | |
2877 | { | |
1da177e4 | 2878 | unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY; |
2fddfeef | 2879 | struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode); |
19fd4bb2 | 2880 | struct tgid_iter iter; |
b488893a | 2881 | struct pid_namespace *ns; |
1da177e4 | 2882 | |
61a28784 EB |
2883 | if (!reaper) |
2884 | goto out_no_task; | |
2885 | ||
7bcd6b0e | 2886 | for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) { |
c5141e6d | 2887 | const struct pid_entry *p = &proc_base_stuff[nr]; |
61a28784 | 2888 | if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0) |
801199ce | 2889 | goto out; |
1da177e4 LT |
2890 | } |
2891 | ||
b488893a | 2892 | ns = filp->f_dentry->d_sb->s_fs_info; |
19fd4bb2 EB |
2893 | iter.task = NULL; |
2894 | iter.tgid = filp->f_pos - TGID_OFFSET; | |
2895 | for (iter = next_tgid(ns, iter); | |
2896 | iter.task; | |
2897 | iter.tgid += 1, iter = next_tgid(ns, iter)) { | |
2898 | filp->f_pos = iter.tgid + TGID_OFFSET; | |
2899 | if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) { | |
2900 | put_task_struct(iter.task); | |
0804ef4b | 2901 | goto out; |
1da177e4 | 2902 | } |
0bc58a91 | 2903 | } |
0804ef4b EB |
2904 | filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET; |
2905 | out: | |
61a28784 EB |
2906 | put_task_struct(reaper); |
2907 | out_no_task: | |
0bc58a91 EB |
2908 | return 0; |
2909 | } | |
1da177e4 | 2910 | |
28a6d671 EB |
2911 | /* |
2912 | * Tasks | |
2913 | */ | |
c5141e6d | 2914 | static const struct pid_entry tid_base_stuff[] = { |
631f9c18 AD |
2915 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), |
2916 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fd_operations), | |
2917 | REG("environ", S_IRUSR, proc_environ_operations), | |
2918 | INF("auxv", S_IRUSR, proc_pid_auxv), | |
2919 | ONE("status", S_IRUGO, proc_pid_status), | |
2920 | ONE("personality", S_IRUSR, proc_pid_personality), | |
2921 | INF("limits", S_IRUSR, proc_pid_limits), | |
43ae34cb | 2922 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 2923 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
ebcb6734 | 2924 | #endif |
4614a696 | 2925 | REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations), |
ebcb6734 | 2926 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
631f9c18 | 2927 | INF("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 2928 | #endif |
631f9c18 AD |
2929 | INF("cmdline", S_IRUGO, proc_pid_cmdline), |
2930 | ONE("stat", S_IRUGO, proc_tid_stat), | |
2931 | ONE("statm", S_IRUGO, proc_pid_statm), | |
2932 | REG("maps", S_IRUGO, proc_maps_operations), | |
28a6d671 | 2933 | #ifdef CONFIG_NUMA |
631f9c18 | 2934 | REG("numa_maps", S_IRUGO, proc_numa_maps_operations), |
28a6d671 | 2935 | #endif |
631f9c18 AD |
2936 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
2937 | LNK("cwd", proc_cwd_link), | |
2938 | LNK("root", proc_root_link), | |
2939 | LNK("exe", proc_exe_link), | |
2940 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
2941 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
1e883281 | 2942 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 AD |
2943 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
2944 | REG("smaps", S_IRUGO, proc_smaps_operations), | |
2945 | REG("pagemap", S_IRUSR, proc_pagemap_operations), | |
28a6d671 EB |
2946 | #endif |
2947 | #ifdef CONFIG_SECURITY | |
631f9c18 | 2948 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
2949 | #endif |
2950 | #ifdef CONFIG_KALLSYMS | |
631f9c18 | 2951 | INF("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 2952 | #endif |
2ec220e2 KC |
2953 | #ifdef CONFIG_STACKTRACE |
2954 | ONE("stack", S_IRUSR, proc_pid_stack), | |
28a6d671 EB |
2955 | #endif |
2956 | #ifdef CONFIG_SCHEDSTATS | |
631f9c18 | 2957 | INF("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 2958 | #endif |
9745512c | 2959 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 2960 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 2961 | #endif |
8793d854 | 2962 | #ifdef CONFIG_PROC_PID_CPUSET |
631f9c18 | 2963 | REG("cpuset", S_IRUGO, proc_cpuset_operations), |
a424316c PM |
2964 | #endif |
2965 | #ifdef CONFIG_CGROUPS | |
631f9c18 | 2966 | REG("cgroup", S_IRUGO, proc_cgroup_operations), |
28a6d671 | 2967 | #endif |
631f9c18 AD |
2968 | INF("oom_score", S_IRUGO, proc_oom_score), |
2969 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations), | |
28a6d671 | 2970 | #ifdef CONFIG_AUDITSYSCALL |
631f9c18 AD |
2971 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
2972 | REG("sessionid", S_IRUSR, proc_sessionid_operations), | |
28a6d671 | 2973 | #endif |
f4f154fd | 2974 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 2975 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
f4f154fd | 2976 | #endif |
297c5d92 | 2977 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
631f9c18 | 2978 | INF("io", S_IRUGO, proc_tid_io_accounting), |
297c5d92 | 2979 | #endif |
28a6d671 EB |
2980 | }; |
2981 | ||
2982 | static int proc_tid_base_readdir(struct file * filp, | |
2983 | void * dirent, filldir_t filldir) | |
2984 | { | |
2985 | return proc_pident_readdir(filp,dirent,filldir, | |
2986 | tid_base_stuff,ARRAY_SIZE(tid_base_stuff)); | |
2987 | } | |
2988 | ||
2989 | static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){ | |
7bcd6b0e EB |
2990 | return proc_pident_lookup(dir, dentry, |
2991 | tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
2992 | } |
2993 | ||
00977a59 | 2994 | static const struct file_operations proc_tid_base_operations = { |
28a6d671 EB |
2995 | .read = generic_read_dir, |
2996 | .readdir = proc_tid_base_readdir, | |
2997 | }; | |
2998 | ||
c5ef1c42 | 2999 | static const struct inode_operations proc_tid_base_inode_operations = { |
28a6d671 EB |
3000 | .lookup = proc_tid_base_lookup, |
3001 | .getattr = pid_getattr, | |
3002 | .setattr = proc_setattr, | |
3003 | }; | |
3004 | ||
444ceed8 | 3005 | static struct dentry *proc_task_instantiate(struct inode *dir, |
c5141e6d | 3006 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
444ceed8 EB |
3007 | { |
3008 | struct dentry *error = ERR_PTR(-ENOENT); | |
3009 | struct inode *inode; | |
61a28784 | 3010 | inode = proc_pid_make_inode(dir->i_sb, task); |
444ceed8 EB |
3011 | |
3012 | if (!inode) | |
3013 | goto out; | |
3014 | inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO; | |
3015 | inode->i_op = &proc_tid_base_inode_operations; | |
3016 | inode->i_fop = &proc_tid_base_operations; | |
3017 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 VN |
3018 | |
3019 | inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff, | |
3020 | ARRAY_SIZE(tid_base_stuff)); | |
444ceed8 EB |
3021 | |
3022 | dentry->d_op = &pid_dentry_operations; | |
3023 | ||
3024 | d_add(dentry, inode); | |
3025 | /* Close the race of the process dying before we return the dentry */ | |
3026 | if (pid_revalidate(dentry, NULL)) | |
3027 | error = NULL; | |
3028 | out: | |
3029 | return error; | |
3030 | } | |
3031 | ||
28a6d671 EB |
3032 | static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd) |
3033 | { | |
3034 | struct dentry *result = ERR_PTR(-ENOENT); | |
3035 | struct task_struct *task; | |
3036 | struct task_struct *leader = get_proc_task(dir); | |
28a6d671 | 3037 | unsigned tid; |
b488893a | 3038 | struct pid_namespace *ns; |
28a6d671 EB |
3039 | |
3040 | if (!leader) | |
3041 | goto out_no_task; | |
3042 | ||
3043 | tid = name_to_int(dentry); | |
3044 | if (tid == ~0U) | |
3045 | goto out; | |
3046 | ||
b488893a | 3047 | ns = dentry->d_sb->s_fs_info; |
28a6d671 | 3048 | rcu_read_lock(); |
b488893a | 3049 | task = find_task_by_pid_ns(tid, ns); |
28a6d671 EB |
3050 | if (task) |
3051 | get_task_struct(task); | |
3052 | rcu_read_unlock(); | |
3053 | if (!task) | |
3054 | goto out; | |
bac0abd6 | 3055 | if (!same_thread_group(leader, task)) |
28a6d671 EB |
3056 | goto out_drop_task; |
3057 | ||
444ceed8 | 3058 | result = proc_task_instantiate(dir, dentry, task, NULL); |
28a6d671 EB |
3059 | out_drop_task: |
3060 | put_task_struct(task); | |
3061 | out: | |
3062 | put_task_struct(leader); | |
3063 | out_no_task: | |
3064 | return result; | |
3065 | } | |
3066 | ||
0bc58a91 EB |
3067 | /* |
3068 | * Find the first tid of a thread group to return to user space. | |
3069 | * | |
3070 | * Usually this is just the thread group leader, but if the users | |
3071 | * buffer was too small or there was a seek into the middle of the | |
3072 | * directory we have more work todo. | |
3073 | * | |
3074 | * In the case of a short read we start with find_task_by_pid. | |
3075 | * | |
3076 | * In the case of a seek we start with the leader and walk nr | |
3077 | * threads past it. | |
3078 | */ | |
cc288738 | 3079 | static struct task_struct *first_tid(struct task_struct *leader, |
b488893a | 3080 | int tid, int nr, struct pid_namespace *ns) |
0bc58a91 | 3081 | { |
a872ff0c | 3082 | struct task_struct *pos; |
1da177e4 | 3083 | |
cc288738 | 3084 | rcu_read_lock(); |
0bc58a91 EB |
3085 | /* Attempt to start with the pid of a thread */ |
3086 | if (tid && (nr > 0)) { | |
b488893a | 3087 | pos = find_task_by_pid_ns(tid, ns); |
a872ff0c ON |
3088 | if (pos && (pos->group_leader == leader)) |
3089 | goto found; | |
0bc58a91 | 3090 | } |
1da177e4 | 3091 | |
0bc58a91 | 3092 | /* If nr exceeds the number of threads there is nothing todo */ |
a872ff0c ON |
3093 | pos = NULL; |
3094 | if (nr && nr >= get_nr_threads(leader)) | |
3095 | goto out; | |
1da177e4 | 3096 | |
a872ff0c ON |
3097 | /* If we haven't found our starting place yet start |
3098 | * with the leader and walk nr threads forward. | |
0bc58a91 | 3099 | */ |
a872ff0c ON |
3100 | for (pos = leader; nr > 0; --nr) { |
3101 | pos = next_thread(pos); | |
3102 | if (pos == leader) { | |
3103 | pos = NULL; | |
3104 | goto out; | |
3105 | } | |
1da177e4 | 3106 | } |
a872ff0c ON |
3107 | found: |
3108 | get_task_struct(pos); | |
3109 | out: | |
cc288738 | 3110 | rcu_read_unlock(); |
0bc58a91 EB |
3111 | return pos; |
3112 | } | |
3113 | ||
3114 | /* | |
3115 | * Find the next thread in the thread list. | |
3116 | * Return NULL if there is an error or no next thread. | |
3117 | * | |
3118 | * The reference to the input task_struct is released. | |
3119 | */ | |
3120 | static struct task_struct *next_tid(struct task_struct *start) | |
3121 | { | |
c1df7fb8 | 3122 | struct task_struct *pos = NULL; |
cc288738 | 3123 | rcu_read_lock(); |
c1df7fb8 | 3124 | if (pid_alive(start)) { |
0bc58a91 | 3125 | pos = next_thread(start); |
c1df7fb8 ON |
3126 | if (thread_group_leader(pos)) |
3127 | pos = NULL; | |
3128 | else | |
3129 | get_task_struct(pos); | |
3130 | } | |
cc288738 | 3131 | rcu_read_unlock(); |
0bc58a91 EB |
3132 | put_task_struct(start); |
3133 | return pos; | |
1da177e4 LT |
3134 | } |
3135 | ||
61a28784 EB |
3136 | static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir, |
3137 | struct task_struct *task, int tid) | |
3138 | { | |
3139 | char name[PROC_NUMBUF]; | |
3140 | int len = snprintf(name, sizeof(name), "%d", tid); | |
3141 | return proc_fill_cache(filp, dirent, filldir, name, len, | |
3142 | proc_task_instantiate, task, NULL); | |
3143 | } | |
3144 | ||
1da177e4 LT |
3145 | /* for the /proc/TGID/task/ directories */ |
3146 | static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir) | |
3147 | { | |
2fddfeef | 3148 | struct dentry *dentry = filp->f_path.dentry; |
1da177e4 | 3149 | struct inode *inode = dentry->d_inode; |
7d895244 | 3150 | struct task_struct *leader = NULL; |
0bc58a91 | 3151 | struct task_struct *task; |
1da177e4 LT |
3152 | int retval = -ENOENT; |
3153 | ino_t ino; | |
0bc58a91 | 3154 | int tid; |
b488893a | 3155 | struct pid_namespace *ns; |
1da177e4 | 3156 | |
7d895244 GC |
3157 | task = get_proc_task(inode); |
3158 | if (!task) | |
3159 | goto out_no_task; | |
3160 | rcu_read_lock(); | |
3161 | if (pid_alive(task)) { | |
3162 | leader = task->group_leader; | |
3163 | get_task_struct(leader); | |
3164 | } | |
3165 | rcu_read_unlock(); | |
3166 | put_task_struct(task); | |
99f89551 EB |
3167 | if (!leader) |
3168 | goto out_no_task; | |
1da177e4 LT |
3169 | retval = 0; |
3170 | ||
ee568b25 | 3171 | switch ((unsigned long)filp->f_pos) { |
1da177e4 LT |
3172 | case 0: |
3173 | ino = inode->i_ino; | |
ee6f779b | 3174 | if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0) |
1da177e4 | 3175 | goto out; |
ee6f779b | 3176 | filp->f_pos++; |
1da177e4 LT |
3177 | /* fall through */ |
3178 | case 1: | |
3179 | ino = parent_ino(dentry); | |
ee6f779b | 3180 | if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0) |
1da177e4 | 3181 | goto out; |
ee6f779b | 3182 | filp->f_pos++; |
1da177e4 LT |
3183 | /* fall through */ |
3184 | } | |
3185 | ||
0bc58a91 EB |
3186 | /* f_version caches the tgid value that the last readdir call couldn't |
3187 | * return. lseek aka telldir automagically resets f_version to 0. | |
3188 | */ | |
b488893a | 3189 | ns = filp->f_dentry->d_sb->s_fs_info; |
2b47c361 | 3190 | tid = (int)filp->f_version; |
0bc58a91 | 3191 | filp->f_version = 0; |
ee6f779b | 3192 | for (task = first_tid(leader, tid, filp->f_pos - 2, ns); |
0bc58a91 | 3193 | task; |
ee6f779b | 3194 | task = next_tid(task), filp->f_pos++) { |
b488893a | 3195 | tid = task_pid_nr_ns(task, ns); |
61a28784 | 3196 | if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) { |
0bc58a91 EB |
3197 | /* returning this tgid failed, save it as the first |
3198 | * pid for the next readir call */ | |
2b47c361 | 3199 | filp->f_version = (u64)tid; |
0bc58a91 | 3200 | put_task_struct(task); |
1da177e4 | 3201 | break; |
0bc58a91 | 3202 | } |
1da177e4 LT |
3203 | } |
3204 | out: | |
99f89551 EB |
3205 | put_task_struct(leader); |
3206 | out_no_task: | |
1da177e4 LT |
3207 | return retval; |
3208 | } | |
6e66b52b EB |
3209 | |
3210 | static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat) | |
3211 | { | |
3212 | struct inode *inode = dentry->d_inode; | |
99f89551 | 3213 | struct task_struct *p = get_proc_task(inode); |
6e66b52b EB |
3214 | generic_fillattr(inode, stat); |
3215 | ||
99f89551 | 3216 | if (p) { |
99f89551 | 3217 | stat->nlink += get_nr_threads(p); |
99f89551 | 3218 | put_task_struct(p); |
6e66b52b EB |
3219 | } |
3220 | ||
3221 | return 0; | |
3222 | } | |
28a6d671 | 3223 | |
c5ef1c42 | 3224 | static const struct inode_operations proc_task_inode_operations = { |
28a6d671 EB |
3225 | .lookup = proc_task_lookup, |
3226 | .getattr = proc_task_getattr, | |
3227 | .setattr = proc_setattr, | |
3228 | }; | |
3229 | ||
00977a59 | 3230 | static const struct file_operations proc_task_operations = { |
28a6d671 EB |
3231 | .read = generic_read_dir, |
3232 | .readdir = proc_task_readdir, | |
3233 | }; |