]>
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> |
5a0e3ad6 | 84 | #include <linux/slab.h> |
1da177e4 LT |
85 | #include "internal.h" |
86 | ||
0f2fe20f EB |
87 | /* NOTE: |
88 | * Implementing inode permission operations in /proc is almost | |
89 | * certainly an error. Permission checks need to happen during | |
90 | * each system call not at open time. The reason is that most of | |
91 | * what we wish to check for permissions in /proc varies at runtime. | |
92 | * | |
93 | * The classic example of a problem is opening file descriptors | |
94 | * in /proc for a task before it execs a suid executable. | |
95 | */ | |
96 | ||
1da177e4 | 97 | struct pid_entry { |
1da177e4 | 98 | char *name; |
c5141e6d | 99 | int len; |
1da177e4 | 100 | mode_t mode; |
c5ef1c42 | 101 | const struct inode_operations *iop; |
00977a59 | 102 | const struct file_operations *fop; |
20cdc894 | 103 | union proc_op op; |
1da177e4 LT |
104 | }; |
105 | ||
61a28784 | 106 | #define NOD(NAME, MODE, IOP, FOP, OP) { \ |
20cdc894 | 107 | .name = (NAME), \ |
c5141e6d | 108 | .len = sizeof(NAME) - 1, \ |
20cdc894 EB |
109 | .mode = MODE, \ |
110 | .iop = IOP, \ | |
111 | .fop = FOP, \ | |
112 | .op = OP, \ | |
113 | } | |
114 | ||
631f9c18 AD |
115 | #define DIR(NAME, MODE, iops, fops) \ |
116 | NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} ) | |
117 | #define LNK(NAME, get_link) \ | |
61a28784 | 118 | NOD(NAME, (S_IFLNK|S_IRWXUGO), \ |
20cdc894 | 119 | &proc_pid_link_inode_operations, NULL, \ |
631f9c18 AD |
120 | { .proc_get_link = get_link } ) |
121 | #define REG(NAME, MODE, fops) \ | |
122 | NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {}) | |
123 | #define INF(NAME, MODE, read) \ | |
61a28784 | 124 | NOD(NAME, (S_IFREG|(MODE)), \ |
20cdc894 | 125 | NULL, &proc_info_file_operations, \ |
631f9c18 AD |
126 | { .proc_read = read } ) |
127 | #define ONE(NAME, MODE, show) \ | |
be614086 EB |
128 | NOD(NAME, (S_IFREG|(MODE)), \ |
129 | NULL, &proc_single_file_operations, \ | |
631f9c18 | 130 | { .proc_show = show } ) |
1da177e4 | 131 | |
aed54175 VN |
132 | /* |
133 | * Count the number of hardlinks for the pid_entry table, excluding the . | |
134 | * and .. links. | |
135 | */ | |
136 | static unsigned int pid_entry_count_dirs(const struct pid_entry *entries, | |
137 | unsigned int n) | |
138 | { | |
139 | unsigned int i; | |
140 | unsigned int count; | |
141 | ||
142 | count = 0; | |
143 | for (i = 0; i < n; ++i) { | |
144 | if (S_ISDIR(entries[i].mode)) | |
145 | ++count; | |
146 | } | |
147 | ||
148 | return count; | |
149 | } | |
150 | ||
7c2c7d99 | 151 | static int get_fs_path(struct task_struct *task, struct path *path, bool root) |
1da177e4 LT |
152 | { |
153 | struct fs_struct *fs; | |
7c2c7d99 HD |
154 | int result = -ENOENT; |
155 | ||
0494f6ec MS |
156 | task_lock(task); |
157 | fs = task->fs; | |
7c2c7d99 HD |
158 | if (fs) { |
159 | read_lock(&fs->lock); | |
160 | *path = root ? fs->root : fs->pwd; | |
161 | path_get(path); | |
162 | read_unlock(&fs->lock); | |
163 | result = 0; | |
164 | } | |
0494f6ec | 165 | task_unlock(task); |
7c2c7d99 | 166 | return result; |
0494f6ec MS |
167 | } |
168 | ||
99f89551 EB |
169 | static int get_nr_threads(struct task_struct *tsk) |
170 | { | |
99f89551 EB |
171 | unsigned long flags; |
172 | int count = 0; | |
173 | ||
174 | if (lock_task_sighand(tsk, &flags)) { | |
175 | count = atomic_read(&tsk->signal->count); | |
176 | unlock_task_sighand(tsk, &flags); | |
177 | } | |
178 | return count; | |
179 | } | |
180 | ||
3dcd25f3 | 181 | static int proc_cwd_link(struct inode *inode, struct path *path) |
0494f6ec | 182 | { |
99f89551 | 183 | struct task_struct *task = get_proc_task(inode); |
0494f6ec | 184 | int result = -ENOENT; |
99f89551 EB |
185 | |
186 | if (task) { | |
7c2c7d99 | 187 | result = get_fs_path(task, path, 0); |
99f89551 EB |
188 | put_task_struct(task); |
189 | } | |
1da177e4 LT |
190 | return result; |
191 | } | |
192 | ||
3dcd25f3 | 193 | static int proc_root_link(struct inode *inode, struct path *path) |
1da177e4 | 194 | { |
99f89551 | 195 | struct task_struct *task = get_proc_task(inode); |
1da177e4 | 196 | int result = -ENOENT; |
99f89551 EB |
197 | |
198 | if (task) { | |
7c2c7d99 | 199 | result = get_fs_path(task, path, 1); |
99f89551 EB |
200 | put_task_struct(task); |
201 | } | |
1da177e4 LT |
202 | return result; |
203 | } | |
204 | ||
638fa202 RM |
205 | /* |
206 | * Return zero if current may access user memory in @task, -error if not. | |
207 | */ | |
208 | static int check_mem_permission(struct task_struct *task) | |
209 | { | |
210 | /* | |
211 | * A task can always look at itself, in case it chooses | |
212 | * to use system calls instead of load instructions. | |
213 | */ | |
214 | if (task == current) | |
215 | return 0; | |
216 | ||
217 | /* | |
218 | * If current is actively ptrace'ing, and would also be | |
219 | * permitted to freshly attach with ptrace now, permit it. | |
220 | */ | |
0d094efe RM |
221 | if (task_is_stopped_or_traced(task)) { |
222 | int match; | |
223 | rcu_read_lock(); | |
224 | match = (tracehook_tracer_task(task) == current); | |
225 | rcu_read_unlock(); | |
226 | if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH)) | |
227 | return 0; | |
228 | } | |
638fa202 RM |
229 | |
230 | /* | |
231 | * Noone else is allowed. | |
232 | */ | |
233 | return -EPERM; | |
234 | } | |
1da177e4 | 235 | |
831830b5 AV |
236 | struct mm_struct *mm_for_maps(struct task_struct *task) |
237 | { | |
704b836c ON |
238 | struct mm_struct *mm; |
239 | ||
240 | if (mutex_lock_killable(&task->cred_guard_mutex)) | |
241 | return NULL; | |
00f89d21 | 242 | |
704b836c ON |
243 | mm = get_task_mm(task); |
244 | if (mm && mm != current->mm && | |
245 | !ptrace_may_access(task, PTRACE_MODE_READ)) { | |
246 | mmput(mm); | |
247 | mm = NULL; | |
13f0feaf | 248 | } |
704b836c ON |
249 | mutex_unlock(&task->cred_guard_mutex); |
250 | ||
831830b5 | 251 | return mm; |
831830b5 AV |
252 | } |
253 | ||
1da177e4 LT |
254 | static int proc_pid_cmdline(struct task_struct *task, char * buffer) |
255 | { | |
256 | int res = 0; | |
257 | unsigned int len; | |
258 | struct mm_struct *mm = get_task_mm(task); | |
259 | if (!mm) | |
260 | goto out; | |
261 | if (!mm->arg_end) | |
262 | goto out_mm; /* Shh! No looking before we're done */ | |
263 | ||
264 | len = mm->arg_end - mm->arg_start; | |
265 | ||
266 | if (len > PAGE_SIZE) | |
267 | len = PAGE_SIZE; | |
268 | ||
269 | res = access_process_vm(task, mm->arg_start, buffer, len, 0); | |
270 | ||
271 | // If the nul at the end of args has been overwritten, then | |
272 | // assume application is using setproctitle(3). | |
273 | if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) { | |
274 | len = strnlen(buffer, res); | |
275 | if (len < res) { | |
276 | res = len; | |
277 | } else { | |
278 | len = mm->env_end - mm->env_start; | |
279 | if (len > PAGE_SIZE - res) | |
280 | len = PAGE_SIZE - res; | |
281 | res += access_process_vm(task, mm->env_start, buffer+res, len, 0); | |
282 | res = strnlen(buffer, res); | |
283 | } | |
284 | } | |
285 | out_mm: | |
286 | mmput(mm); | |
287 | out: | |
288 | return res; | |
289 | } | |
290 | ||
291 | static int proc_pid_auxv(struct task_struct *task, char *buffer) | |
292 | { | |
293 | int res = 0; | |
294 | struct mm_struct *mm = get_task_mm(task); | |
295 | if (mm) { | |
296 | unsigned int nwords = 0; | |
dfe6b7d9 | 297 | do { |
1da177e4 | 298 | nwords += 2; |
dfe6b7d9 | 299 | } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */ |
1da177e4 LT |
300 | res = nwords * sizeof(mm->saved_auxv[0]); |
301 | if (res > PAGE_SIZE) | |
302 | res = PAGE_SIZE; | |
303 | memcpy(buffer, mm->saved_auxv, res); | |
304 | mmput(mm); | |
305 | } | |
306 | return res; | |
307 | } | |
308 | ||
309 | ||
310 | #ifdef CONFIG_KALLSYMS | |
311 | /* | |
312 | * Provides a wchan file via kallsyms in a proper one-value-per-file format. | |
313 | * Returns the resolved symbol. If that fails, simply return the address. | |
314 | */ | |
315 | static int proc_pid_wchan(struct task_struct *task, char *buffer) | |
316 | { | |
ffb45122 | 317 | unsigned long wchan; |
9281acea | 318 | char symname[KSYM_NAME_LEN]; |
1da177e4 LT |
319 | |
320 | wchan = get_wchan(task); | |
321 | ||
9d65cb4a | 322 | if (lookup_symbol_name(wchan, symname) < 0) |
f83ce3e6 JE |
323 | if (!ptrace_may_access(task, PTRACE_MODE_READ)) |
324 | return 0; | |
325 | else | |
326 | return sprintf(buffer, "%lu", wchan); | |
9d65cb4a AD |
327 | else |
328 | return sprintf(buffer, "%s", symname); | |
1da177e4 LT |
329 | } |
330 | #endif /* CONFIG_KALLSYMS */ | |
331 | ||
2ec220e2 KC |
332 | #ifdef CONFIG_STACKTRACE |
333 | ||
334 | #define MAX_STACK_TRACE_DEPTH 64 | |
335 | ||
336 | static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns, | |
337 | struct pid *pid, struct task_struct *task) | |
338 | { | |
339 | struct stack_trace trace; | |
340 | unsigned long *entries; | |
341 | int i; | |
342 | ||
343 | entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL); | |
344 | if (!entries) | |
345 | return -ENOMEM; | |
346 | ||
347 | trace.nr_entries = 0; | |
348 | trace.max_entries = MAX_STACK_TRACE_DEPTH; | |
349 | trace.entries = entries; | |
350 | trace.skip = 0; | |
351 | save_stack_trace_tsk(task, &trace); | |
352 | ||
353 | for (i = 0; i < trace.nr_entries; i++) { | |
354 | seq_printf(m, "[<%p>] %pS\n", | |
355 | (void *)entries[i], (void *)entries[i]); | |
356 | } | |
357 | kfree(entries); | |
358 | ||
359 | return 0; | |
360 | } | |
361 | #endif | |
362 | ||
1da177e4 LT |
363 | #ifdef CONFIG_SCHEDSTATS |
364 | /* | |
365 | * Provides /proc/PID/schedstat | |
366 | */ | |
367 | static int proc_pid_schedstat(struct task_struct *task, char *buffer) | |
368 | { | |
172ba844 | 369 | return sprintf(buffer, "%llu %llu %lu\n", |
826e08b0 IM |
370 | (unsigned long long)task->se.sum_exec_runtime, |
371 | (unsigned long long)task->sched_info.run_delay, | |
2d72376b | 372 | task->sched_info.pcount); |
1da177e4 LT |
373 | } |
374 | #endif | |
375 | ||
9745512c AV |
376 | #ifdef CONFIG_LATENCYTOP |
377 | static int lstats_show_proc(struct seq_file *m, void *v) | |
378 | { | |
379 | int i; | |
13d77c37 HS |
380 | struct inode *inode = m->private; |
381 | struct task_struct *task = get_proc_task(inode); | |
9745512c | 382 | |
13d77c37 HS |
383 | if (!task) |
384 | return -ESRCH; | |
385 | seq_puts(m, "Latency Top version : v0.1\n"); | |
9745512c AV |
386 | for (i = 0; i < 32; i++) { |
387 | if (task->latency_record[i].backtrace[0]) { | |
388 | int q; | |
389 | seq_printf(m, "%i %li %li ", | |
390 | task->latency_record[i].count, | |
391 | task->latency_record[i].time, | |
392 | task->latency_record[i].max); | |
393 | for (q = 0; q < LT_BACKTRACEDEPTH; q++) { | |
9c246247 | 394 | char sym[KSYM_SYMBOL_LEN]; |
9745512c AV |
395 | char *c; |
396 | if (!task->latency_record[i].backtrace[q]) | |
397 | break; | |
398 | if (task->latency_record[i].backtrace[q] == ULONG_MAX) | |
399 | break; | |
400 | sprint_symbol(sym, task->latency_record[i].backtrace[q]); | |
401 | c = strchr(sym, '+'); | |
402 | if (c) | |
403 | *c = 0; | |
404 | seq_printf(m, "%s ", sym); | |
405 | } | |
406 | seq_printf(m, "\n"); | |
407 | } | |
408 | ||
409 | } | |
13d77c37 | 410 | put_task_struct(task); |
9745512c AV |
411 | return 0; |
412 | } | |
413 | ||
414 | static int lstats_open(struct inode *inode, struct file *file) | |
415 | { | |
13d77c37 | 416 | return single_open(file, lstats_show_proc, inode); |
d6643d12 HS |
417 | } |
418 | ||
9745512c AV |
419 | static ssize_t lstats_write(struct file *file, const char __user *buf, |
420 | size_t count, loff_t *offs) | |
421 | { | |
13d77c37 | 422 | struct task_struct *task = get_proc_task(file->f_dentry->d_inode); |
9745512c | 423 | |
13d77c37 HS |
424 | if (!task) |
425 | return -ESRCH; | |
9745512c | 426 | clear_all_latency_tracing(task); |
13d77c37 | 427 | put_task_struct(task); |
9745512c AV |
428 | |
429 | return count; | |
430 | } | |
431 | ||
432 | static const struct file_operations proc_lstats_operations = { | |
433 | .open = lstats_open, | |
434 | .read = seq_read, | |
435 | .write = lstats_write, | |
436 | .llseek = seq_lseek, | |
13d77c37 | 437 | .release = single_release, |
9745512c AV |
438 | }; |
439 | ||
440 | #endif | |
441 | ||
1da177e4 LT |
442 | /* The badness from the OOM killer */ |
443 | unsigned long badness(struct task_struct *p, unsigned long uptime); | |
444 | static int proc_oom_score(struct task_struct *task, char *buffer) | |
445 | { | |
446 | unsigned long points; | |
447 | struct timespec uptime; | |
448 | ||
449 | do_posix_clock_monotonic_gettime(&uptime); | |
19c5d45a | 450 | read_lock(&tasklist_lock); |
495789a5 | 451 | points = badness(task->group_leader, uptime.tv_sec); |
19c5d45a | 452 | read_unlock(&tasklist_lock); |
1da177e4 LT |
453 | return sprintf(buffer, "%lu\n", points); |
454 | } | |
455 | ||
d85f50d5 NH |
456 | struct limit_names { |
457 | char *name; | |
458 | char *unit; | |
459 | }; | |
460 | ||
461 | static const struct limit_names lnames[RLIM_NLIMITS] = { | |
cff4edb5 | 462 | [RLIMIT_CPU] = {"Max cpu time", "seconds"}, |
d85f50d5 NH |
463 | [RLIMIT_FSIZE] = {"Max file size", "bytes"}, |
464 | [RLIMIT_DATA] = {"Max data size", "bytes"}, | |
465 | [RLIMIT_STACK] = {"Max stack size", "bytes"}, | |
466 | [RLIMIT_CORE] = {"Max core file size", "bytes"}, | |
467 | [RLIMIT_RSS] = {"Max resident set", "bytes"}, | |
468 | [RLIMIT_NPROC] = {"Max processes", "processes"}, | |
469 | [RLIMIT_NOFILE] = {"Max open files", "files"}, | |
470 | [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"}, | |
471 | [RLIMIT_AS] = {"Max address space", "bytes"}, | |
472 | [RLIMIT_LOCKS] = {"Max file locks", "locks"}, | |
473 | [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"}, | |
474 | [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"}, | |
475 | [RLIMIT_NICE] = {"Max nice priority", NULL}, | |
476 | [RLIMIT_RTPRIO] = {"Max realtime priority", NULL}, | |
8808117c | 477 | [RLIMIT_RTTIME] = {"Max realtime timeout", "us"}, |
d85f50d5 NH |
478 | }; |
479 | ||
480 | /* Display limits for a process */ | |
481 | static int proc_pid_limits(struct task_struct *task, char *buffer) | |
482 | { | |
483 | unsigned int i; | |
484 | int count = 0; | |
485 | unsigned long flags; | |
486 | char *bufptr = buffer; | |
487 | ||
488 | struct rlimit rlim[RLIM_NLIMITS]; | |
489 | ||
a6bebbc8 | 490 | if (!lock_task_sighand(task, &flags)) |
d85f50d5 | 491 | return 0; |
d85f50d5 NH |
492 | memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS); |
493 | unlock_task_sighand(task, &flags); | |
d85f50d5 NH |
494 | |
495 | /* | |
496 | * print the file header | |
497 | */ | |
498 | count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n", | |
499 | "Limit", "Soft Limit", "Hard Limit", "Units"); | |
500 | ||
501 | for (i = 0; i < RLIM_NLIMITS; i++) { | |
502 | if (rlim[i].rlim_cur == RLIM_INFINITY) | |
503 | count += sprintf(&bufptr[count], "%-25s %-20s ", | |
504 | lnames[i].name, "unlimited"); | |
505 | else | |
506 | count += sprintf(&bufptr[count], "%-25s %-20lu ", | |
507 | lnames[i].name, rlim[i].rlim_cur); | |
508 | ||
509 | if (rlim[i].rlim_max == RLIM_INFINITY) | |
510 | count += sprintf(&bufptr[count], "%-20s ", "unlimited"); | |
511 | else | |
512 | count += sprintf(&bufptr[count], "%-20lu ", | |
513 | rlim[i].rlim_max); | |
514 | ||
515 | if (lnames[i].unit) | |
516 | count += sprintf(&bufptr[count], "%-10s\n", | |
517 | lnames[i].unit); | |
518 | else | |
519 | count += sprintf(&bufptr[count], "\n"); | |
520 | } | |
521 | ||
522 | return count; | |
523 | } | |
524 | ||
ebcb6734 RM |
525 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
526 | static int proc_pid_syscall(struct task_struct *task, char *buffer) | |
527 | { | |
528 | long nr; | |
529 | unsigned long args[6], sp, pc; | |
530 | ||
531 | if (task_current_syscall(task, &nr, args, 6, &sp, &pc)) | |
532 | return sprintf(buffer, "running\n"); | |
533 | ||
534 | if (nr < 0) | |
535 | return sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc); | |
536 | ||
537 | return sprintf(buffer, | |
538 | "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n", | |
539 | nr, | |
540 | args[0], args[1], args[2], args[3], args[4], args[5], | |
541 | sp, pc); | |
542 | } | |
543 | #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */ | |
544 | ||
1da177e4 LT |
545 | /************************************************************************/ |
546 | /* Here the fs part begins */ | |
547 | /************************************************************************/ | |
548 | ||
549 | /* permission checks */ | |
778c1144 | 550 | static int proc_fd_access_allowed(struct inode *inode) |
1da177e4 | 551 | { |
778c1144 EB |
552 | struct task_struct *task; |
553 | int allowed = 0; | |
df26c40e EB |
554 | /* Allow access to a task's file descriptors if it is us or we |
555 | * may use ptrace attach to the process and find out that | |
556 | * information. | |
778c1144 EB |
557 | */ |
558 | task = get_proc_task(inode); | |
df26c40e | 559 | if (task) { |
006ebb40 | 560 | allowed = ptrace_may_access(task, PTRACE_MODE_READ); |
778c1144 | 561 | put_task_struct(task); |
df26c40e | 562 | } |
778c1144 | 563 | return allowed; |
1da177e4 LT |
564 | } |
565 | ||
6d76fa58 LT |
566 | static int proc_setattr(struct dentry *dentry, struct iattr *attr) |
567 | { | |
568 | int error; | |
569 | struct inode *inode = dentry->d_inode; | |
570 | ||
571 | if (attr->ia_valid & ATTR_MODE) | |
572 | return -EPERM; | |
573 | ||
574 | error = inode_change_ok(inode, attr); | |
1e8123fd JJ |
575 | if (!error) |
576 | error = inode_setattr(inode, attr); | |
6d76fa58 LT |
577 | return error; |
578 | } | |
579 | ||
c5ef1c42 | 580 | static const struct inode_operations proc_def_inode_operations = { |
6d76fa58 LT |
581 | .setattr = proc_setattr, |
582 | }; | |
583 | ||
a1a2c409 MS |
584 | static int mounts_open_common(struct inode *inode, struct file *file, |
585 | const struct seq_operations *op) | |
1da177e4 | 586 | { |
99f89551 | 587 | struct task_struct *task = get_proc_task(inode); |
cf7b708c | 588 | struct nsproxy *nsp; |
6b3286ed | 589 | struct mnt_namespace *ns = NULL; |
a1a2c409 | 590 | struct path root; |
5addc5dd AV |
591 | struct proc_mounts *p; |
592 | int ret = -EINVAL; | |
1da177e4 | 593 | |
99f89551 | 594 | if (task) { |
cf7b708c PE |
595 | rcu_read_lock(); |
596 | nsp = task_nsproxy(task); | |
597 | if (nsp) { | |
598 | ns = nsp->mnt_ns; | |
863c4702 AD |
599 | if (ns) |
600 | get_mnt_ns(ns); | |
601 | } | |
cf7b708c | 602 | rcu_read_unlock(); |
7c2c7d99 HD |
603 | if (ns && get_fs_path(task, &root, 1) == 0) |
604 | ret = 0; | |
99f89551 EB |
605 | put_task_struct(task); |
606 | } | |
5addc5dd | 607 | |
a1a2c409 MS |
608 | if (!ns) |
609 | goto err; | |
7c2c7d99 | 610 | if (ret) |
a1a2c409 MS |
611 | goto err_put_ns; |
612 | ||
a1a2c409 MS |
613 | ret = -ENOMEM; |
614 | p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL); | |
615 | if (!p) | |
616 | goto err_put_path; | |
617 | ||
618 | file->private_data = &p->m; | |
619 | ret = seq_open(file, op); | |
620 | if (ret) | |
621 | goto err_free; | |
622 | ||
623 | p->m.private = p; | |
624 | p->ns = ns; | |
625 | p->root = root; | |
626 | p->event = ns->event; | |
627 | ||
628 | return 0; | |
629 | ||
630 | err_free: | |
631 | kfree(p); | |
632 | err_put_path: | |
633 | path_put(&root); | |
634 | err_put_ns: | |
635 | put_mnt_ns(ns); | |
636 | err: | |
1da177e4 LT |
637 | return ret; |
638 | } | |
639 | ||
640 | static int mounts_release(struct inode *inode, struct file *file) | |
641 | { | |
a1a2c409 MS |
642 | struct proc_mounts *p = file->private_data; |
643 | path_put(&p->root); | |
644 | put_mnt_ns(p->ns); | |
1da177e4 LT |
645 | return seq_release(inode, file); |
646 | } | |
647 | ||
5addc5dd AV |
648 | static unsigned mounts_poll(struct file *file, poll_table *wait) |
649 | { | |
650 | struct proc_mounts *p = file->private_data; | |
31b07093 | 651 | unsigned res = POLLIN | POLLRDNORM; |
5addc5dd | 652 | |
9f5596af AV |
653 | poll_wait(file, &p->ns->poll, wait); |
654 | if (mnt_had_events(p)) | |
31b07093 | 655 | res |= POLLERR | POLLPRI; |
5addc5dd AV |
656 | |
657 | return res; | |
658 | } | |
659 | ||
a1a2c409 MS |
660 | static int mounts_open(struct inode *inode, struct file *file) |
661 | { | |
662 | return mounts_open_common(inode, file, &mounts_op); | |
663 | } | |
664 | ||
00977a59 | 665 | static const struct file_operations proc_mounts_operations = { |
1da177e4 LT |
666 | .open = mounts_open, |
667 | .read = seq_read, | |
668 | .llseek = seq_lseek, | |
669 | .release = mounts_release, | |
5addc5dd | 670 | .poll = mounts_poll, |
1da177e4 LT |
671 | }; |
672 | ||
2d4d4864 RP |
673 | static int mountinfo_open(struct inode *inode, struct file *file) |
674 | { | |
675 | return mounts_open_common(inode, file, &mountinfo_op); | |
676 | } | |
677 | ||
678 | static const struct file_operations proc_mountinfo_operations = { | |
679 | .open = mountinfo_open, | |
680 | .read = seq_read, | |
681 | .llseek = seq_lseek, | |
682 | .release = mounts_release, | |
683 | .poll = mounts_poll, | |
684 | }; | |
685 | ||
b4629fe2 CL |
686 | static int mountstats_open(struct inode *inode, struct file *file) |
687 | { | |
a1a2c409 | 688 | return mounts_open_common(inode, file, &mountstats_op); |
b4629fe2 CL |
689 | } |
690 | ||
00977a59 | 691 | static const struct file_operations proc_mountstats_operations = { |
b4629fe2 CL |
692 | .open = mountstats_open, |
693 | .read = seq_read, | |
694 | .llseek = seq_lseek, | |
695 | .release = mounts_release, | |
696 | }; | |
697 | ||
1da177e4 LT |
698 | #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */ |
699 | ||
700 | static ssize_t proc_info_read(struct file * file, char __user * buf, | |
701 | size_t count, loff_t *ppos) | |
702 | { | |
2fddfeef | 703 | struct inode * inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
704 | unsigned long page; |
705 | ssize_t length; | |
99f89551 EB |
706 | struct task_struct *task = get_proc_task(inode); |
707 | ||
708 | length = -ESRCH; | |
709 | if (!task) | |
710 | goto out_no_task; | |
1da177e4 LT |
711 | |
712 | if (count > PROC_BLOCK_SIZE) | |
713 | count = PROC_BLOCK_SIZE; | |
99f89551 EB |
714 | |
715 | length = -ENOMEM; | |
e12ba74d | 716 | if (!(page = __get_free_page(GFP_TEMPORARY))) |
99f89551 | 717 | goto out; |
1da177e4 LT |
718 | |
719 | length = PROC_I(inode)->op.proc_read(task, (char*)page); | |
720 | ||
721 | if (length >= 0) | |
722 | length = simple_read_from_buffer(buf, count, ppos, (char *)page, length); | |
723 | free_page(page); | |
99f89551 EB |
724 | out: |
725 | put_task_struct(task); | |
726 | out_no_task: | |
1da177e4 LT |
727 | return length; |
728 | } | |
729 | ||
00977a59 | 730 | static const struct file_operations proc_info_file_operations = { |
1da177e4 LT |
731 | .read = proc_info_read, |
732 | }; | |
733 | ||
be614086 EB |
734 | static int proc_single_show(struct seq_file *m, void *v) |
735 | { | |
736 | struct inode *inode = m->private; | |
737 | struct pid_namespace *ns; | |
738 | struct pid *pid; | |
739 | struct task_struct *task; | |
740 | int ret; | |
741 | ||
742 | ns = inode->i_sb->s_fs_info; | |
743 | pid = proc_pid(inode); | |
744 | task = get_pid_task(pid, PIDTYPE_PID); | |
745 | if (!task) | |
746 | return -ESRCH; | |
747 | ||
748 | ret = PROC_I(inode)->op.proc_show(m, ns, pid, task); | |
749 | ||
750 | put_task_struct(task); | |
751 | return ret; | |
752 | } | |
753 | ||
754 | static int proc_single_open(struct inode *inode, struct file *filp) | |
755 | { | |
756 | int ret; | |
757 | ret = single_open(filp, proc_single_show, NULL); | |
758 | if (!ret) { | |
759 | struct seq_file *m = filp->private_data; | |
760 | ||
761 | m->private = inode; | |
762 | } | |
763 | return ret; | |
764 | } | |
765 | ||
766 | static const struct file_operations proc_single_file_operations = { | |
767 | .open = proc_single_open, | |
768 | .read = seq_read, | |
769 | .llseek = seq_lseek, | |
770 | .release = single_release, | |
771 | }; | |
772 | ||
1da177e4 LT |
773 | static int mem_open(struct inode* inode, struct file* file) |
774 | { | |
775 | file->private_data = (void*)((long)current->self_exec_id); | |
776 | return 0; | |
777 | } | |
778 | ||
779 | static ssize_t mem_read(struct file * file, char __user * buf, | |
780 | size_t count, loff_t *ppos) | |
781 | { | |
2fddfeef | 782 | struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode); |
1da177e4 LT |
783 | char *page; |
784 | unsigned long src = *ppos; | |
785 | int ret = -ESRCH; | |
786 | struct mm_struct *mm; | |
787 | ||
99f89551 EB |
788 | if (!task) |
789 | goto out_no_task; | |
790 | ||
638fa202 | 791 | if (check_mem_permission(task)) |
1da177e4 LT |
792 | goto out; |
793 | ||
794 | ret = -ENOMEM; | |
e12ba74d | 795 | page = (char *)__get_free_page(GFP_TEMPORARY); |
1da177e4 LT |
796 | if (!page) |
797 | goto out; | |
798 | ||
799 | ret = 0; | |
800 | ||
801 | mm = get_task_mm(task); | |
802 | if (!mm) | |
803 | goto out_free; | |
804 | ||
805 | ret = -EIO; | |
806 | ||
807 | if (file->private_data != (void*)((long)current->self_exec_id)) | |
808 | goto out_put; | |
809 | ||
810 | ret = 0; | |
811 | ||
812 | while (count > 0) { | |
813 | int this_len, retval; | |
814 | ||
815 | this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count; | |
816 | retval = access_process_vm(task, src, page, this_len, 0); | |
638fa202 | 817 | if (!retval || check_mem_permission(task)) { |
1da177e4 LT |
818 | if (!ret) |
819 | ret = -EIO; | |
820 | break; | |
821 | } | |
822 | ||
823 | if (copy_to_user(buf, page, retval)) { | |
824 | ret = -EFAULT; | |
825 | break; | |
826 | } | |
827 | ||
828 | ret += retval; | |
829 | src += retval; | |
830 | buf += retval; | |
831 | count -= retval; | |
832 | } | |
833 | *ppos = src; | |
834 | ||
835 | out_put: | |
836 | mmput(mm); | |
837 | out_free: | |
838 | free_page((unsigned long) page); | |
839 | out: | |
99f89551 EB |
840 | put_task_struct(task); |
841 | out_no_task: | |
1da177e4 LT |
842 | return ret; |
843 | } | |
844 | ||
845 | #define mem_write NULL | |
846 | ||
847 | #ifndef mem_write | |
848 | /* This is a security hazard */ | |
63967fa9 | 849 | static ssize_t mem_write(struct file * file, const char __user *buf, |
1da177e4 LT |
850 | size_t count, loff_t *ppos) |
851 | { | |
f7ca54f4 | 852 | int copied; |
1da177e4 | 853 | char *page; |
2fddfeef | 854 | struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode); |
1da177e4 LT |
855 | unsigned long dst = *ppos; |
856 | ||
99f89551 EB |
857 | copied = -ESRCH; |
858 | if (!task) | |
859 | goto out_no_task; | |
860 | ||
638fa202 | 861 | if (check_mem_permission(task)) |
99f89551 | 862 | goto out; |
1da177e4 | 863 | |
99f89551 | 864 | copied = -ENOMEM; |
e12ba74d | 865 | page = (char *)__get_free_page(GFP_TEMPORARY); |
1da177e4 | 866 | if (!page) |
99f89551 | 867 | goto out; |
1da177e4 | 868 | |
f7ca54f4 | 869 | copied = 0; |
1da177e4 LT |
870 | while (count > 0) { |
871 | int this_len, retval; | |
872 | ||
873 | this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count; | |
874 | if (copy_from_user(page, buf, this_len)) { | |
875 | copied = -EFAULT; | |
876 | break; | |
877 | } | |
878 | retval = access_process_vm(task, dst, page, this_len, 1); | |
879 | if (!retval) { | |
880 | if (!copied) | |
881 | copied = -EIO; | |
882 | break; | |
883 | } | |
884 | copied += retval; | |
885 | buf += retval; | |
886 | dst += retval; | |
887 | count -= retval; | |
888 | } | |
889 | *ppos = dst; | |
890 | free_page((unsigned long) page); | |
99f89551 EB |
891 | out: |
892 | put_task_struct(task); | |
893 | out_no_task: | |
1da177e4 LT |
894 | return copied; |
895 | } | |
896 | #endif | |
897 | ||
85863e47 | 898 | loff_t mem_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
899 | { |
900 | switch (orig) { | |
901 | case 0: | |
902 | file->f_pos = offset; | |
903 | break; | |
904 | case 1: | |
905 | file->f_pos += offset; | |
906 | break; | |
907 | default: | |
908 | return -EINVAL; | |
909 | } | |
910 | force_successful_syscall_return(); | |
911 | return file->f_pos; | |
912 | } | |
913 | ||
00977a59 | 914 | static const struct file_operations proc_mem_operations = { |
1da177e4 LT |
915 | .llseek = mem_lseek, |
916 | .read = mem_read, | |
917 | .write = mem_write, | |
918 | .open = mem_open, | |
919 | }; | |
920 | ||
315e28c8 JP |
921 | static ssize_t environ_read(struct file *file, char __user *buf, |
922 | size_t count, loff_t *ppos) | |
923 | { | |
924 | struct task_struct *task = get_proc_task(file->f_dentry->d_inode); | |
925 | char *page; | |
926 | unsigned long src = *ppos; | |
927 | int ret = -ESRCH; | |
928 | struct mm_struct *mm; | |
929 | ||
930 | if (!task) | |
931 | goto out_no_task; | |
932 | ||
006ebb40 | 933 | if (!ptrace_may_access(task, PTRACE_MODE_READ)) |
315e28c8 JP |
934 | goto out; |
935 | ||
936 | ret = -ENOMEM; | |
937 | page = (char *)__get_free_page(GFP_TEMPORARY); | |
938 | if (!page) | |
939 | goto out; | |
940 | ||
941 | ret = 0; | |
942 | ||
943 | mm = get_task_mm(task); | |
944 | if (!mm) | |
945 | goto out_free; | |
946 | ||
947 | while (count > 0) { | |
948 | int this_len, retval, max_len; | |
949 | ||
950 | this_len = mm->env_end - (mm->env_start + src); | |
951 | ||
952 | if (this_len <= 0) | |
953 | break; | |
954 | ||
955 | max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count; | |
956 | this_len = (this_len > max_len) ? max_len : this_len; | |
957 | ||
958 | retval = access_process_vm(task, (mm->env_start + src), | |
959 | page, this_len, 0); | |
960 | ||
961 | if (retval <= 0) { | |
962 | ret = retval; | |
963 | break; | |
964 | } | |
965 | ||
966 | if (copy_to_user(buf, page, retval)) { | |
967 | ret = -EFAULT; | |
968 | break; | |
969 | } | |
970 | ||
971 | ret += retval; | |
972 | src += retval; | |
973 | buf += retval; | |
974 | count -= retval; | |
975 | } | |
976 | *ppos = src; | |
977 | ||
978 | mmput(mm); | |
979 | out_free: | |
980 | free_page((unsigned long) page); | |
981 | out: | |
982 | put_task_struct(task); | |
983 | out_no_task: | |
984 | return ret; | |
985 | } | |
986 | ||
987 | static const struct file_operations proc_environ_operations = { | |
988 | .read = environ_read, | |
989 | }; | |
990 | ||
1da177e4 LT |
991 | static ssize_t oom_adjust_read(struct file *file, char __user *buf, |
992 | size_t count, loff_t *ppos) | |
993 | { | |
2fddfeef | 994 | struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode); |
8578cea7 | 995 | char buffer[PROC_NUMBUF]; |
1da177e4 | 996 | size_t len; |
28b83c51 KM |
997 | int oom_adjust = OOM_DISABLE; |
998 | unsigned long flags; | |
1da177e4 | 999 | |
99f89551 EB |
1000 | if (!task) |
1001 | return -ESRCH; | |
28b83c51 KM |
1002 | |
1003 | if (lock_task_sighand(task, &flags)) { | |
1004 | oom_adjust = task->signal->oom_adj; | |
1005 | unlock_task_sighand(task, &flags); | |
1006 | } | |
1007 | ||
99f89551 EB |
1008 | put_task_struct(task); |
1009 | ||
8578cea7 | 1010 | len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust); |
0c28f287 AM |
1011 | |
1012 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
1da177e4 LT |
1013 | } |
1014 | ||
1015 | static ssize_t oom_adjust_write(struct file *file, const char __user *buf, | |
1016 | size_t count, loff_t *ppos) | |
1017 | { | |
99f89551 | 1018 | struct task_struct *task; |
5d863b89 KM |
1019 | char buffer[PROC_NUMBUF]; |
1020 | long oom_adjust; | |
28b83c51 | 1021 | unsigned long flags; |
5d863b89 | 1022 | int err; |
1da177e4 | 1023 | |
8578cea7 EB |
1024 | memset(buffer, 0, sizeof(buffer)); |
1025 | if (count > sizeof(buffer) - 1) | |
1026 | count = sizeof(buffer) - 1; | |
1da177e4 LT |
1027 | if (copy_from_user(buffer, buf, count)) |
1028 | return -EFAULT; | |
5d863b89 KM |
1029 | |
1030 | err = strict_strtol(strstrip(buffer), 0, &oom_adjust); | |
1031 | if (err) | |
1032 | return -EINVAL; | |
8ac773b4 AD |
1033 | if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) && |
1034 | oom_adjust != OOM_DISABLE) | |
1da177e4 | 1035 | return -EINVAL; |
5d863b89 | 1036 | |
2fddfeef | 1037 | task = get_proc_task(file->f_path.dentry->d_inode); |
99f89551 EB |
1038 | if (!task) |
1039 | return -ESRCH; | |
28b83c51 KM |
1040 | if (!lock_task_sighand(task, &flags)) { |
1041 | put_task_struct(task); | |
1042 | return -ESRCH; | |
1043 | } | |
1044 | ||
1045 | if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) { | |
1046 | unlock_task_sighand(task, &flags); | |
8fb4fc68 GJ |
1047 | put_task_struct(task); |
1048 | return -EACCES; | |
1049 | } | |
28b83c51 KM |
1050 | |
1051 | task->signal->oom_adj = oom_adjust; | |
1052 | ||
1053 | unlock_task_sighand(task, &flags); | |
99f89551 | 1054 | put_task_struct(task); |
5d863b89 KM |
1055 | |
1056 | return count; | |
1da177e4 LT |
1057 | } |
1058 | ||
00977a59 | 1059 | static const struct file_operations proc_oom_adjust_operations = { |
1da177e4 LT |
1060 | .read = oom_adjust_read, |
1061 | .write = oom_adjust_write, | |
1062 | }; | |
1063 | ||
1da177e4 LT |
1064 | #ifdef CONFIG_AUDITSYSCALL |
1065 | #define TMPBUFLEN 21 | |
1066 | static ssize_t proc_loginuid_read(struct file * file, char __user * buf, | |
1067 | size_t count, loff_t *ppos) | |
1068 | { | |
2fddfeef | 1069 | struct inode * inode = file->f_path.dentry->d_inode; |
99f89551 | 1070 | struct task_struct *task = get_proc_task(inode); |
1da177e4 LT |
1071 | ssize_t length; |
1072 | char tmpbuf[TMPBUFLEN]; | |
1073 | ||
99f89551 EB |
1074 | if (!task) |
1075 | return -ESRCH; | |
1da177e4 | 1076 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", |
0c11b942 | 1077 | audit_get_loginuid(task)); |
99f89551 | 1078 | put_task_struct(task); |
1da177e4 LT |
1079 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); |
1080 | } | |
1081 | ||
1082 | static ssize_t proc_loginuid_write(struct file * file, const char __user * buf, | |
1083 | size_t count, loff_t *ppos) | |
1084 | { | |
2fddfeef | 1085 | struct inode * inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
1086 | char *page, *tmp; |
1087 | ssize_t length; | |
1da177e4 LT |
1088 | uid_t loginuid; |
1089 | ||
1090 | if (!capable(CAP_AUDIT_CONTROL)) | |
1091 | return -EPERM; | |
1092 | ||
7dc52157 PM |
1093 | rcu_read_lock(); |
1094 | if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) { | |
1095 | rcu_read_unlock(); | |
1da177e4 | 1096 | return -EPERM; |
7dc52157 PM |
1097 | } |
1098 | rcu_read_unlock(); | |
1da177e4 | 1099 | |
e0182909 AV |
1100 | if (count >= PAGE_SIZE) |
1101 | count = PAGE_SIZE - 1; | |
1da177e4 LT |
1102 | |
1103 | if (*ppos != 0) { | |
1104 | /* No partial writes. */ | |
1105 | return -EINVAL; | |
1106 | } | |
e12ba74d | 1107 | page = (char*)__get_free_page(GFP_TEMPORARY); |
1da177e4 LT |
1108 | if (!page) |
1109 | return -ENOMEM; | |
1110 | length = -EFAULT; | |
1111 | if (copy_from_user(page, buf, count)) | |
1112 | goto out_free_page; | |
1113 | ||
e0182909 | 1114 | page[count] = '\0'; |
1da177e4 LT |
1115 | loginuid = simple_strtoul(page, &tmp, 10); |
1116 | if (tmp == page) { | |
1117 | length = -EINVAL; | |
1118 | goto out_free_page; | |
1119 | ||
1120 | } | |
99f89551 | 1121 | length = audit_set_loginuid(current, loginuid); |
1da177e4 LT |
1122 | if (likely(length == 0)) |
1123 | length = count; | |
1124 | ||
1125 | out_free_page: | |
1126 | free_page((unsigned long) page); | |
1127 | return length; | |
1128 | } | |
1129 | ||
00977a59 | 1130 | static const struct file_operations proc_loginuid_operations = { |
1da177e4 LT |
1131 | .read = proc_loginuid_read, |
1132 | .write = proc_loginuid_write, | |
1133 | }; | |
1e0bd755 EP |
1134 | |
1135 | static ssize_t proc_sessionid_read(struct file * file, char __user * buf, | |
1136 | size_t count, loff_t *ppos) | |
1137 | { | |
1138 | struct inode * inode = file->f_path.dentry->d_inode; | |
1139 | struct task_struct *task = get_proc_task(inode); | |
1140 | ssize_t length; | |
1141 | char tmpbuf[TMPBUFLEN]; | |
1142 | ||
1143 | if (!task) | |
1144 | return -ESRCH; | |
1145 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", | |
1146 | audit_get_sessionid(task)); | |
1147 | put_task_struct(task); | |
1148 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); | |
1149 | } | |
1150 | ||
1151 | static const struct file_operations proc_sessionid_operations = { | |
1152 | .read = proc_sessionid_read, | |
1153 | }; | |
1da177e4 LT |
1154 | #endif |
1155 | ||
f4f154fd AM |
1156 | #ifdef CONFIG_FAULT_INJECTION |
1157 | static ssize_t proc_fault_inject_read(struct file * file, char __user * buf, | |
1158 | size_t count, loff_t *ppos) | |
1159 | { | |
1160 | struct task_struct *task = get_proc_task(file->f_dentry->d_inode); | |
1161 | char buffer[PROC_NUMBUF]; | |
1162 | size_t len; | |
1163 | int make_it_fail; | |
f4f154fd AM |
1164 | |
1165 | if (!task) | |
1166 | return -ESRCH; | |
1167 | make_it_fail = task->make_it_fail; | |
1168 | put_task_struct(task); | |
1169 | ||
1170 | len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail); | |
0c28f287 AM |
1171 | |
1172 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
f4f154fd AM |
1173 | } |
1174 | ||
1175 | static ssize_t proc_fault_inject_write(struct file * file, | |
1176 | const char __user * buf, size_t count, loff_t *ppos) | |
1177 | { | |
1178 | struct task_struct *task; | |
1179 | char buffer[PROC_NUMBUF], *end; | |
1180 | int make_it_fail; | |
1181 | ||
1182 | if (!capable(CAP_SYS_RESOURCE)) | |
1183 | return -EPERM; | |
1184 | memset(buffer, 0, sizeof(buffer)); | |
1185 | if (count > sizeof(buffer) - 1) | |
1186 | count = sizeof(buffer) - 1; | |
1187 | if (copy_from_user(buffer, buf, count)) | |
1188 | return -EFAULT; | |
cba8aafe VL |
1189 | make_it_fail = simple_strtol(strstrip(buffer), &end, 0); |
1190 | if (*end) | |
1191 | return -EINVAL; | |
f4f154fd AM |
1192 | task = get_proc_task(file->f_dentry->d_inode); |
1193 | if (!task) | |
1194 | return -ESRCH; | |
1195 | task->make_it_fail = make_it_fail; | |
1196 | put_task_struct(task); | |
cba8aafe VL |
1197 | |
1198 | return count; | |
f4f154fd AM |
1199 | } |
1200 | ||
00977a59 | 1201 | static const struct file_operations proc_fault_inject_operations = { |
f4f154fd AM |
1202 | .read = proc_fault_inject_read, |
1203 | .write = proc_fault_inject_write, | |
1204 | }; | |
1205 | #endif | |
1206 | ||
9745512c | 1207 | |
43ae34cb IM |
1208 | #ifdef CONFIG_SCHED_DEBUG |
1209 | /* | |
1210 | * Print out various scheduling related per-task fields: | |
1211 | */ | |
1212 | static int sched_show(struct seq_file *m, void *v) | |
1213 | { | |
1214 | struct inode *inode = m->private; | |
1215 | struct task_struct *p; | |
1216 | ||
43ae34cb IM |
1217 | p = get_proc_task(inode); |
1218 | if (!p) | |
1219 | return -ESRCH; | |
1220 | proc_sched_show_task(p, m); | |
1221 | ||
1222 | put_task_struct(p); | |
1223 | ||
1224 | return 0; | |
1225 | } | |
1226 | ||
1227 | static ssize_t | |
1228 | sched_write(struct file *file, const char __user *buf, | |
1229 | size_t count, loff_t *offset) | |
1230 | { | |
1231 | struct inode *inode = file->f_path.dentry->d_inode; | |
1232 | struct task_struct *p; | |
1233 | ||
43ae34cb IM |
1234 | p = get_proc_task(inode); |
1235 | if (!p) | |
1236 | return -ESRCH; | |
1237 | proc_sched_set_task(p); | |
1238 | ||
1239 | put_task_struct(p); | |
1240 | ||
1241 | return count; | |
1242 | } | |
1243 | ||
1244 | static int sched_open(struct inode *inode, struct file *filp) | |
1245 | { | |
1246 | int ret; | |
1247 | ||
1248 | ret = single_open(filp, sched_show, NULL); | |
1249 | if (!ret) { | |
1250 | struct seq_file *m = filp->private_data; | |
1251 | ||
1252 | m->private = inode; | |
1253 | } | |
1254 | return ret; | |
1255 | } | |
1256 | ||
1257 | static const struct file_operations proc_pid_sched_operations = { | |
1258 | .open = sched_open, | |
1259 | .read = seq_read, | |
1260 | .write = sched_write, | |
1261 | .llseek = seq_lseek, | |
5ea473a1 | 1262 | .release = single_release, |
43ae34cb IM |
1263 | }; |
1264 | ||
1265 | #endif | |
1266 | ||
4614a696 JS |
1267 | static ssize_t comm_write(struct file *file, const char __user *buf, |
1268 | size_t count, loff_t *offset) | |
1269 | { | |
1270 | struct inode *inode = file->f_path.dentry->d_inode; | |
1271 | struct task_struct *p; | |
1272 | char buffer[TASK_COMM_LEN]; | |
1273 | ||
1274 | memset(buffer, 0, sizeof(buffer)); | |
1275 | if (count > sizeof(buffer) - 1) | |
1276 | count = sizeof(buffer) - 1; | |
1277 | if (copy_from_user(buffer, buf, count)) | |
1278 | return -EFAULT; | |
1279 | ||
1280 | p = get_proc_task(inode); | |
1281 | if (!p) | |
1282 | return -ESRCH; | |
1283 | ||
1284 | if (same_thread_group(current, p)) | |
1285 | set_task_comm(p, buffer); | |
1286 | else | |
1287 | count = -EINVAL; | |
1288 | ||
1289 | put_task_struct(p); | |
1290 | ||
1291 | return count; | |
1292 | } | |
1293 | ||
1294 | static int comm_show(struct seq_file *m, void *v) | |
1295 | { | |
1296 | struct inode *inode = m->private; | |
1297 | struct task_struct *p; | |
1298 | ||
1299 | p = get_proc_task(inode); | |
1300 | if (!p) | |
1301 | return -ESRCH; | |
1302 | ||
1303 | task_lock(p); | |
1304 | seq_printf(m, "%s\n", p->comm); | |
1305 | task_unlock(p); | |
1306 | ||
1307 | put_task_struct(p); | |
1308 | ||
1309 | return 0; | |
1310 | } | |
1311 | ||
1312 | static int comm_open(struct inode *inode, struct file *filp) | |
1313 | { | |
1314 | int ret; | |
1315 | ||
1316 | ret = single_open(filp, comm_show, NULL); | |
1317 | if (!ret) { | |
1318 | struct seq_file *m = filp->private_data; | |
1319 | ||
1320 | m->private = inode; | |
1321 | } | |
1322 | return ret; | |
1323 | } | |
1324 | ||
1325 | static const struct file_operations proc_pid_set_comm_operations = { | |
1326 | .open = comm_open, | |
1327 | .read = seq_read, | |
1328 | .write = comm_write, | |
1329 | .llseek = seq_lseek, | |
1330 | .release = single_release, | |
1331 | }; | |
1332 | ||
925d1c40 MH |
1333 | /* |
1334 | * We added or removed a vma mapping the executable. The vmas are only mapped | |
1335 | * during exec and are not mapped with the mmap system call. | |
1336 | * Callers must hold down_write() on the mm's mmap_sem for these | |
1337 | */ | |
1338 | void added_exe_file_vma(struct mm_struct *mm) | |
1339 | { | |
1340 | mm->num_exe_file_vmas++; | |
1341 | } | |
1342 | ||
1343 | void removed_exe_file_vma(struct mm_struct *mm) | |
1344 | { | |
1345 | mm->num_exe_file_vmas--; | |
1346 | if ((mm->num_exe_file_vmas == 0) && mm->exe_file){ | |
1347 | fput(mm->exe_file); | |
1348 | mm->exe_file = NULL; | |
1349 | } | |
1350 | ||
1351 | } | |
1352 | ||
1353 | void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) | |
1354 | { | |
1355 | if (new_exe_file) | |
1356 | get_file(new_exe_file); | |
1357 | if (mm->exe_file) | |
1358 | fput(mm->exe_file); | |
1359 | mm->exe_file = new_exe_file; | |
1360 | mm->num_exe_file_vmas = 0; | |
1361 | } | |
1362 | ||
1363 | struct file *get_mm_exe_file(struct mm_struct *mm) | |
1364 | { | |
1365 | struct file *exe_file; | |
1366 | ||
1367 | /* We need mmap_sem to protect against races with removal of | |
1368 | * VM_EXECUTABLE vmas */ | |
1369 | down_read(&mm->mmap_sem); | |
1370 | exe_file = mm->exe_file; | |
1371 | if (exe_file) | |
1372 | get_file(exe_file); | |
1373 | up_read(&mm->mmap_sem); | |
1374 | return exe_file; | |
1375 | } | |
1376 | ||
1377 | void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm) | |
1378 | { | |
1379 | /* It's safe to write the exe_file pointer without exe_file_lock because | |
1380 | * this is called during fork when the task is not yet in /proc */ | |
1381 | newmm->exe_file = get_mm_exe_file(oldmm); | |
1382 | } | |
1383 | ||
1384 | static int proc_exe_link(struct inode *inode, struct path *exe_path) | |
1385 | { | |
1386 | struct task_struct *task; | |
1387 | struct mm_struct *mm; | |
1388 | struct file *exe_file; | |
1389 | ||
1390 | task = get_proc_task(inode); | |
1391 | if (!task) | |
1392 | return -ENOENT; | |
1393 | mm = get_task_mm(task); | |
1394 | put_task_struct(task); | |
1395 | if (!mm) | |
1396 | return -ENOENT; | |
1397 | exe_file = get_mm_exe_file(mm); | |
1398 | mmput(mm); | |
1399 | if (exe_file) { | |
1400 | *exe_path = exe_file->f_path; | |
1401 | path_get(&exe_file->f_path); | |
1402 | fput(exe_file); | |
1403 | return 0; | |
1404 | } else | |
1405 | return -ENOENT; | |
1406 | } | |
1407 | ||
008b150a | 1408 | static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd) |
1da177e4 LT |
1409 | { |
1410 | struct inode *inode = dentry->d_inode; | |
1411 | int error = -EACCES; | |
1412 | ||
1413 | /* We don't need a base pointer in the /proc filesystem */ | |
1d957f9b | 1414 | path_put(&nd->path); |
1da177e4 | 1415 | |
778c1144 EB |
1416 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1417 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1418 | goto out; |
1da177e4 | 1419 | |
3dcd25f3 | 1420 | error = PROC_I(inode)->op.proc_get_link(inode, &nd->path); |
1da177e4 | 1421 | out: |
008b150a | 1422 | return ERR_PTR(error); |
1da177e4 LT |
1423 | } |
1424 | ||
3dcd25f3 | 1425 | static int do_proc_readlink(struct path *path, char __user *buffer, int buflen) |
1da177e4 | 1426 | { |
e12ba74d | 1427 | char *tmp = (char*)__get_free_page(GFP_TEMPORARY); |
3dcd25f3 | 1428 | char *pathname; |
1da177e4 LT |
1429 | int len; |
1430 | ||
1431 | if (!tmp) | |
1432 | return -ENOMEM; | |
0c28f287 | 1433 | |
cf28b486 | 1434 | pathname = d_path(path, tmp, PAGE_SIZE); |
3dcd25f3 JB |
1435 | len = PTR_ERR(pathname); |
1436 | if (IS_ERR(pathname)) | |
1da177e4 | 1437 | goto out; |
3dcd25f3 | 1438 | len = tmp + PAGE_SIZE - 1 - pathname; |
1da177e4 LT |
1439 | |
1440 | if (len > buflen) | |
1441 | len = buflen; | |
3dcd25f3 | 1442 | if (copy_to_user(buffer, pathname, len)) |
1da177e4 LT |
1443 | len = -EFAULT; |
1444 | out: | |
1445 | free_page((unsigned long)tmp); | |
1446 | return len; | |
1447 | } | |
1448 | ||
1449 | static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen) | |
1450 | { | |
1451 | int error = -EACCES; | |
1452 | struct inode *inode = dentry->d_inode; | |
3dcd25f3 | 1453 | struct path path; |
1da177e4 | 1454 | |
778c1144 EB |
1455 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1456 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1457 | goto out; |
1da177e4 | 1458 | |
3dcd25f3 | 1459 | error = PROC_I(inode)->op.proc_get_link(inode, &path); |
1da177e4 LT |
1460 | if (error) |
1461 | goto out; | |
1462 | ||
3dcd25f3 JB |
1463 | error = do_proc_readlink(&path, buffer, buflen); |
1464 | path_put(&path); | |
1da177e4 | 1465 | out: |
1da177e4 LT |
1466 | return error; |
1467 | } | |
1468 | ||
c5ef1c42 | 1469 | static const struct inode_operations proc_pid_link_inode_operations = { |
1da177e4 | 1470 | .readlink = proc_pid_readlink, |
6d76fa58 LT |
1471 | .follow_link = proc_pid_follow_link, |
1472 | .setattr = proc_setattr, | |
1da177e4 LT |
1473 | }; |
1474 | ||
28a6d671 EB |
1475 | |
1476 | /* building an inode */ | |
1477 | ||
1478 | static int task_dumpable(struct task_struct *task) | |
1da177e4 | 1479 | { |
28a6d671 EB |
1480 | int dumpable = 0; |
1481 | struct mm_struct *mm; | |
1da177e4 | 1482 | |
28a6d671 EB |
1483 | task_lock(task); |
1484 | mm = task->mm; | |
1485 | if (mm) | |
6c5d5238 | 1486 | dumpable = get_dumpable(mm); |
28a6d671 EB |
1487 | task_unlock(task); |
1488 | if(dumpable == 1) | |
1489 | return 1; | |
1490 | return 0; | |
1491 | } | |
1da177e4 | 1492 | |
1da177e4 | 1493 | |
61a28784 | 1494 | static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task) |
28a6d671 EB |
1495 | { |
1496 | struct inode * inode; | |
1497 | struct proc_inode *ei; | |
c69e8d9c | 1498 | const struct cred *cred; |
1da177e4 | 1499 | |
28a6d671 | 1500 | /* We need a new inode */ |
1da177e4 | 1501 | |
28a6d671 EB |
1502 | inode = new_inode(sb); |
1503 | if (!inode) | |
1504 | goto out; | |
1505 | ||
1506 | /* Common stuff */ | |
1507 | ei = PROC_I(inode); | |
1508 | inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; | |
28a6d671 EB |
1509 | inode->i_op = &proc_def_inode_operations; |
1510 | ||
1511 | /* | |
1512 | * grab the reference to task. | |
1513 | */ | |
1a657f78 | 1514 | ei->pid = get_task_pid(task, PIDTYPE_PID); |
28a6d671 EB |
1515 | if (!ei->pid) |
1516 | goto out_unlock; | |
1517 | ||
28a6d671 | 1518 | if (task_dumpable(task)) { |
c69e8d9c DH |
1519 | rcu_read_lock(); |
1520 | cred = __task_cred(task); | |
1521 | inode->i_uid = cred->euid; | |
1522 | inode->i_gid = cred->egid; | |
1523 | rcu_read_unlock(); | |
1da177e4 | 1524 | } |
28a6d671 EB |
1525 | security_task_to_inode(task, inode); |
1526 | ||
1da177e4 | 1527 | out: |
28a6d671 EB |
1528 | return inode; |
1529 | ||
1530 | out_unlock: | |
1531 | iput(inode); | |
1532 | return NULL; | |
1da177e4 LT |
1533 | } |
1534 | ||
28a6d671 | 1535 | static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat) |
1da177e4 | 1536 | { |
1da177e4 | 1537 | struct inode *inode = dentry->d_inode; |
28a6d671 | 1538 | struct task_struct *task; |
c69e8d9c DH |
1539 | const struct cred *cred; |
1540 | ||
28a6d671 | 1541 | generic_fillattr(inode, stat); |
1da177e4 | 1542 | |
28a6d671 EB |
1543 | rcu_read_lock(); |
1544 | stat->uid = 0; | |
1545 | stat->gid = 0; | |
1546 | task = pid_task(proc_pid(inode), PIDTYPE_PID); | |
1547 | if (task) { | |
1548 | if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) || | |
1549 | task_dumpable(task)) { | |
c69e8d9c DH |
1550 | cred = __task_cred(task); |
1551 | stat->uid = cred->euid; | |
1552 | stat->gid = cred->egid; | |
1da177e4 LT |
1553 | } |
1554 | } | |
28a6d671 | 1555 | rcu_read_unlock(); |
d6e71144 | 1556 | return 0; |
1da177e4 LT |
1557 | } |
1558 | ||
1da177e4 LT |
1559 | /* dentry stuff */ |
1560 | ||
1561 | /* | |
1562 | * Exceptional case: normally we are not allowed to unhash a busy | |
1563 | * directory. In this case, however, we can do it - no aliasing problems | |
1564 | * due to the way we treat inodes. | |
1565 | * | |
1566 | * Rewrite the inode's ownerships here because the owning task may have | |
1567 | * performed a setuid(), etc. | |
99f89551 EB |
1568 | * |
1569 | * Before the /proc/pid/status file was created the only way to read | |
1570 | * the effective uid of a /process was to stat /proc/pid. Reading | |
1571 | * /proc/pid/status is slow enough that procps and other packages | |
1572 | * kept stating /proc/pid. To keep the rules in /proc simple I have | |
1573 | * made this apply to all per process world readable and executable | |
1574 | * directories. | |
1da177e4 LT |
1575 | */ |
1576 | static int pid_revalidate(struct dentry *dentry, struct nameidata *nd) | |
1577 | { | |
1578 | struct inode *inode = dentry->d_inode; | |
99f89551 | 1579 | struct task_struct *task = get_proc_task(inode); |
c69e8d9c DH |
1580 | const struct cred *cred; |
1581 | ||
99f89551 EB |
1582 | if (task) { |
1583 | if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) || | |
1584 | task_dumpable(task)) { | |
c69e8d9c DH |
1585 | rcu_read_lock(); |
1586 | cred = __task_cred(task); | |
1587 | inode->i_uid = cred->euid; | |
1588 | inode->i_gid = cred->egid; | |
1589 | rcu_read_unlock(); | |
1da177e4 LT |
1590 | } else { |
1591 | inode->i_uid = 0; | |
1592 | inode->i_gid = 0; | |
1593 | } | |
9ee8ab9f | 1594 | inode->i_mode &= ~(S_ISUID | S_ISGID); |
1da177e4 | 1595 | security_task_to_inode(task, inode); |
99f89551 | 1596 | put_task_struct(task); |
1da177e4 LT |
1597 | return 1; |
1598 | } | |
1599 | d_drop(dentry); | |
1600 | return 0; | |
1601 | } | |
1602 | ||
28a6d671 | 1603 | static int pid_delete_dentry(struct dentry * dentry) |
99f89551 | 1604 | { |
28a6d671 EB |
1605 | /* Is the task we represent dead? |
1606 | * If so, then don't put the dentry on the lru list, | |
1607 | * kill it immediately. | |
1608 | */ | |
1609 | return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first; | |
1610 | } | |
1611 | ||
d72f71eb | 1612 | static const struct dentry_operations pid_dentry_operations = |
28a6d671 EB |
1613 | { |
1614 | .d_revalidate = pid_revalidate, | |
1615 | .d_delete = pid_delete_dentry, | |
1616 | }; | |
1617 | ||
1618 | /* Lookups */ | |
1619 | ||
c5141e6d ED |
1620 | typedef struct dentry *instantiate_t(struct inode *, struct dentry *, |
1621 | struct task_struct *, const void *); | |
61a28784 | 1622 | |
1c0d04c9 EB |
1623 | /* |
1624 | * Fill a directory entry. | |
1625 | * | |
1626 | * If possible create the dcache entry and derive our inode number and | |
1627 | * file type from dcache entry. | |
1628 | * | |
1629 | * Since all of the proc inode numbers are dynamically generated, the inode | |
1630 | * numbers do not exist until the inode is cache. This means creating the | |
1631 | * the dcache entry in readdir is necessary to keep the inode numbers | |
1632 | * reported by readdir in sync with the inode numbers reported | |
1633 | * by stat. | |
1634 | */ | |
61a28784 EB |
1635 | static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir, |
1636 | char *name, int len, | |
c5141e6d | 1637 | instantiate_t instantiate, struct task_struct *task, const void *ptr) |
61a28784 | 1638 | { |
2fddfeef | 1639 | struct dentry *child, *dir = filp->f_path.dentry; |
61a28784 EB |
1640 | struct inode *inode; |
1641 | struct qstr qname; | |
1642 | ino_t ino = 0; | |
1643 | unsigned type = DT_UNKNOWN; | |
1644 | ||
1645 | qname.name = name; | |
1646 | qname.len = len; | |
1647 | qname.hash = full_name_hash(name, len); | |
1648 | ||
1649 | child = d_lookup(dir, &qname); | |
1650 | if (!child) { | |
1651 | struct dentry *new; | |
1652 | new = d_alloc(dir, &qname); | |
1653 | if (new) { | |
1654 | child = instantiate(dir->d_inode, new, task, ptr); | |
1655 | if (child) | |
1656 | dput(new); | |
1657 | else | |
1658 | child = new; | |
1659 | } | |
1660 | } | |
1661 | if (!child || IS_ERR(child) || !child->d_inode) | |
1662 | goto end_instantiate; | |
1663 | inode = child->d_inode; | |
1664 | if (inode) { | |
1665 | ino = inode->i_ino; | |
1666 | type = inode->i_mode >> 12; | |
1667 | } | |
1668 | dput(child); | |
1669 | end_instantiate: | |
1670 | if (!ino) | |
1671 | ino = find_inode_number(dir, &qname); | |
1672 | if (!ino) | |
1673 | ino = 1; | |
1674 | return filldir(dirent, name, len, filp->f_pos, ino, type); | |
1675 | } | |
1676 | ||
28a6d671 EB |
1677 | static unsigned name_to_int(struct dentry *dentry) |
1678 | { | |
1679 | const char *name = dentry->d_name.name; | |
1680 | int len = dentry->d_name.len; | |
1681 | unsigned n = 0; | |
1682 | ||
1683 | if (len > 1 && *name == '0') | |
1684 | goto out; | |
1685 | while (len-- > 0) { | |
1686 | unsigned c = *name++ - '0'; | |
1687 | if (c > 9) | |
1688 | goto out; | |
1689 | if (n >= (~0U-9)/10) | |
1690 | goto out; | |
1691 | n *= 10; | |
1692 | n += c; | |
1693 | } | |
1694 | return n; | |
1695 | out: | |
1696 | return ~0U; | |
1697 | } | |
1698 | ||
27932742 MS |
1699 | #define PROC_FDINFO_MAX 64 |
1700 | ||
3dcd25f3 | 1701 | static int proc_fd_info(struct inode *inode, struct path *path, char *info) |
28a6d671 EB |
1702 | { |
1703 | struct task_struct *task = get_proc_task(inode); | |
1704 | struct files_struct *files = NULL; | |
1705 | struct file *file; | |
1706 | int fd = proc_fd(inode); | |
99f89551 | 1707 | |
99f89551 | 1708 | if (task) { |
28a6d671 EB |
1709 | files = get_files_struct(task); |
1710 | put_task_struct(task); | |
1711 | } | |
1712 | if (files) { | |
1713 | /* | |
1714 | * We are not taking a ref to the file structure, so we must | |
1715 | * hold ->file_lock. | |
1716 | */ | |
1717 | spin_lock(&files->file_lock); | |
1718 | file = fcheck_files(files, fd); | |
1719 | if (file) { | |
3dcd25f3 JB |
1720 | if (path) { |
1721 | *path = file->f_path; | |
1722 | path_get(&file->f_path); | |
1723 | } | |
27932742 MS |
1724 | if (info) |
1725 | snprintf(info, PROC_FDINFO_MAX, | |
1726 | "pos:\t%lli\n" | |
1727 | "flags:\t0%o\n", | |
1728 | (long long) file->f_pos, | |
1729 | file->f_flags); | |
28a6d671 EB |
1730 | spin_unlock(&files->file_lock); |
1731 | put_files_struct(files); | |
1732 | return 0; | |
99f89551 | 1733 | } |
28a6d671 EB |
1734 | spin_unlock(&files->file_lock); |
1735 | put_files_struct(files); | |
99f89551 | 1736 | } |
28a6d671 | 1737 | return -ENOENT; |
99f89551 EB |
1738 | } |
1739 | ||
3dcd25f3 | 1740 | static int proc_fd_link(struct inode *inode, struct path *path) |
27932742 | 1741 | { |
3dcd25f3 | 1742 | return proc_fd_info(inode, path, NULL); |
27932742 MS |
1743 | } |
1744 | ||
1da177e4 LT |
1745 | static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd) |
1746 | { | |
1747 | struct inode *inode = dentry->d_inode; | |
99f89551 | 1748 | struct task_struct *task = get_proc_task(inode); |
aed7a6c4 | 1749 | int fd = proc_fd(inode); |
1da177e4 | 1750 | struct files_struct *files; |
c69e8d9c | 1751 | const struct cred *cred; |
1da177e4 | 1752 | |
99f89551 EB |
1753 | if (task) { |
1754 | files = get_files_struct(task); | |
1755 | if (files) { | |
1756 | rcu_read_lock(); | |
1757 | if (fcheck_files(files, fd)) { | |
1758 | rcu_read_unlock(); | |
1759 | put_files_struct(files); | |
1760 | if (task_dumpable(task)) { | |
c69e8d9c DH |
1761 | rcu_read_lock(); |
1762 | cred = __task_cred(task); | |
1763 | inode->i_uid = cred->euid; | |
1764 | inode->i_gid = cred->egid; | |
1765 | rcu_read_unlock(); | |
99f89551 EB |
1766 | } else { |
1767 | inode->i_uid = 0; | |
1768 | inode->i_gid = 0; | |
1769 | } | |
9ee8ab9f | 1770 | inode->i_mode &= ~(S_ISUID | S_ISGID); |
99f89551 EB |
1771 | security_task_to_inode(task, inode); |
1772 | put_task_struct(task); | |
1773 | return 1; | |
1774 | } | |
b835996f | 1775 | rcu_read_unlock(); |
1da177e4 | 1776 | put_files_struct(files); |
1da177e4 | 1777 | } |
99f89551 | 1778 | put_task_struct(task); |
1da177e4 LT |
1779 | } |
1780 | d_drop(dentry); | |
1781 | return 0; | |
1782 | } | |
1783 | ||
d72f71eb | 1784 | static const struct dentry_operations tid_fd_dentry_operations = |
1da177e4 LT |
1785 | { |
1786 | .d_revalidate = tid_fd_revalidate, | |
1787 | .d_delete = pid_delete_dentry, | |
1788 | }; | |
1789 | ||
444ceed8 | 1790 | static struct dentry *proc_fd_instantiate(struct inode *dir, |
c5141e6d | 1791 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
1da177e4 | 1792 | { |
c5141e6d | 1793 | unsigned fd = *(const unsigned *)ptr; |
444ceed8 EB |
1794 | struct file *file; |
1795 | struct files_struct *files; | |
1796 | struct inode *inode; | |
1797 | struct proc_inode *ei; | |
1798 | struct dentry *error = ERR_PTR(-ENOENT); | |
1da177e4 | 1799 | |
61a28784 | 1800 | inode = proc_pid_make_inode(dir->i_sb, task); |
1da177e4 LT |
1801 | if (!inode) |
1802 | goto out; | |
1803 | ei = PROC_I(inode); | |
aed7a6c4 | 1804 | ei->fd = fd; |
1da177e4 LT |
1805 | files = get_files_struct(task); |
1806 | if (!files) | |
444ceed8 | 1807 | goto out_iput; |
1da177e4 | 1808 | inode->i_mode = S_IFLNK; |
ca99c1da DS |
1809 | |
1810 | /* | |
1811 | * We are not taking a ref to the file structure, so we must | |
1812 | * hold ->file_lock. | |
1813 | */ | |
1814 | spin_lock(&files->file_lock); | |
1da177e4 LT |
1815 | file = fcheck_files(files, fd); |
1816 | if (!file) | |
444ceed8 | 1817 | goto out_unlock; |
aeb5d727 | 1818 | if (file->f_mode & FMODE_READ) |
1da177e4 | 1819 | inode->i_mode |= S_IRUSR | S_IXUSR; |
aeb5d727 | 1820 | if (file->f_mode & FMODE_WRITE) |
1da177e4 | 1821 | inode->i_mode |= S_IWUSR | S_IXUSR; |
ca99c1da | 1822 | spin_unlock(&files->file_lock); |
1da177e4 | 1823 | put_files_struct(files); |
444ceed8 | 1824 | |
1da177e4 LT |
1825 | inode->i_op = &proc_pid_link_inode_operations; |
1826 | inode->i_size = 64; | |
1827 | ei->op.proc_get_link = proc_fd_link; | |
1828 | dentry->d_op = &tid_fd_dentry_operations; | |
1829 | d_add(dentry, inode); | |
cd6a3ce9 EB |
1830 | /* Close the race of the process dying before we return the dentry */ |
1831 | if (tid_fd_revalidate(dentry, NULL)) | |
444ceed8 | 1832 | error = NULL; |
1da177e4 | 1833 | |
444ceed8 EB |
1834 | out: |
1835 | return error; | |
1836 | out_unlock: | |
ca99c1da | 1837 | spin_unlock(&files->file_lock); |
1da177e4 | 1838 | put_files_struct(files); |
444ceed8 | 1839 | out_iput: |
1da177e4 | 1840 | iput(inode); |
cd6a3ce9 | 1841 | goto out; |
1da177e4 LT |
1842 | } |
1843 | ||
27932742 MS |
1844 | static struct dentry *proc_lookupfd_common(struct inode *dir, |
1845 | struct dentry *dentry, | |
1846 | instantiate_t instantiate) | |
444ceed8 EB |
1847 | { |
1848 | struct task_struct *task = get_proc_task(dir); | |
1849 | unsigned fd = name_to_int(dentry); | |
1850 | struct dentry *result = ERR_PTR(-ENOENT); | |
1851 | ||
1852 | if (!task) | |
1853 | goto out_no_task; | |
1854 | if (fd == ~0U) | |
1855 | goto out; | |
1856 | ||
27932742 | 1857 | result = instantiate(dir, dentry, task, &fd); |
444ceed8 EB |
1858 | out: |
1859 | put_task_struct(task); | |
1860 | out_no_task: | |
1861 | return result; | |
1862 | } | |
1863 | ||
27932742 MS |
1864 | static int proc_readfd_common(struct file * filp, void * dirent, |
1865 | filldir_t filldir, instantiate_t instantiate) | |
28a6d671 | 1866 | { |
2fddfeef | 1867 | struct dentry *dentry = filp->f_path.dentry; |
28a6d671 EB |
1868 | struct inode *inode = dentry->d_inode; |
1869 | struct task_struct *p = get_proc_task(inode); | |
457c2510 | 1870 | unsigned int fd, ino; |
28a6d671 | 1871 | int retval; |
28a6d671 | 1872 | struct files_struct * files; |
1da177e4 | 1873 | |
28a6d671 EB |
1874 | retval = -ENOENT; |
1875 | if (!p) | |
1876 | goto out_no_task; | |
1877 | retval = 0; | |
28a6d671 EB |
1878 | |
1879 | fd = filp->f_pos; | |
1880 | switch (fd) { | |
1881 | case 0: | |
1882 | if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0) | |
1883 | goto out; | |
1884 | filp->f_pos++; | |
1885 | case 1: | |
1886 | ino = parent_ino(dentry); | |
1887 | if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0) | |
1888 | goto out; | |
1889 | filp->f_pos++; | |
1890 | default: | |
1891 | files = get_files_struct(p); | |
1892 | if (!files) | |
1893 | goto out; | |
1894 | rcu_read_lock(); | |
28a6d671 | 1895 | for (fd = filp->f_pos-2; |
9b4f526c | 1896 | fd < files_fdtable(files)->max_fds; |
28a6d671 | 1897 | fd++, filp->f_pos++) { |
27932742 MS |
1898 | char name[PROC_NUMBUF]; |
1899 | int len; | |
28a6d671 EB |
1900 | |
1901 | if (!fcheck_files(files, fd)) | |
1902 | continue; | |
1903 | rcu_read_unlock(); | |
1904 | ||
27932742 MS |
1905 | len = snprintf(name, sizeof(name), "%d", fd); |
1906 | if (proc_fill_cache(filp, dirent, filldir, | |
1907 | name, len, instantiate, | |
1908 | p, &fd) < 0) { | |
28a6d671 EB |
1909 | rcu_read_lock(); |
1910 | break; | |
1911 | } | |
1912 | rcu_read_lock(); | |
1913 | } | |
1914 | rcu_read_unlock(); | |
1915 | put_files_struct(files); | |
1916 | } | |
1917 | out: | |
1918 | put_task_struct(p); | |
1919 | out_no_task: | |
1920 | return retval; | |
1921 | } | |
1922 | ||
27932742 MS |
1923 | static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry, |
1924 | struct nameidata *nd) | |
1925 | { | |
1926 | return proc_lookupfd_common(dir, dentry, proc_fd_instantiate); | |
1927 | } | |
1928 | ||
1929 | static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir) | |
1930 | { | |
1931 | return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate); | |
1932 | } | |
1933 | ||
1934 | static ssize_t proc_fdinfo_read(struct file *file, char __user *buf, | |
1935 | size_t len, loff_t *ppos) | |
1936 | { | |
1937 | char tmp[PROC_FDINFO_MAX]; | |
3dcd25f3 | 1938 | int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp); |
27932742 MS |
1939 | if (!err) |
1940 | err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp)); | |
1941 | return err; | |
1942 | } | |
1943 | ||
1944 | static const struct file_operations proc_fdinfo_file_operations = { | |
1945 | .open = nonseekable_open, | |
1946 | .read = proc_fdinfo_read, | |
1947 | }; | |
1948 | ||
00977a59 | 1949 | static const struct file_operations proc_fd_operations = { |
28a6d671 EB |
1950 | .read = generic_read_dir, |
1951 | .readdir = proc_readfd, | |
1da177e4 LT |
1952 | }; |
1953 | ||
8948e11f AD |
1954 | /* |
1955 | * /proc/pid/fd needs a special permission handler so that a process can still | |
1956 | * access /proc/self/fd after it has executed a setuid(). | |
1957 | */ | |
e6305c43 | 1958 | static int proc_fd_permission(struct inode *inode, int mask) |
8948e11f AD |
1959 | { |
1960 | int rv; | |
1961 | ||
1962 | rv = generic_permission(inode, mask, NULL); | |
1963 | if (rv == 0) | |
1964 | return 0; | |
1965 | if (task_pid(current) == proc_pid(inode)) | |
1966 | rv = 0; | |
1967 | return rv; | |
1968 | } | |
1969 | ||
1da177e4 LT |
1970 | /* |
1971 | * proc directories can do almost nothing.. | |
1972 | */ | |
c5ef1c42 | 1973 | static const struct inode_operations proc_fd_inode_operations = { |
1da177e4 | 1974 | .lookup = proc_lookupfd, |
8948e11f | 1975 | .permission = proc_fd_permission, |
6d76fa58 | 1976 | .setattr = proc_setattr, |
1da177e4 LT |
1977 | }; |
1978 | ||
27932742 MS |
1979 | static struct dentry *proc_fdinfo_instantiate(struct inode *dir, |
1980 | struct dentry *dentry, struct task_struct *task, const void *ptr) | |
1981 | { | |
1982 | unsigned fd = *(unsigned *)ptr; | |
1983 | struct inode *inode; | |
1984 | struct proc_inode *ei; | |
1985 | struct dentry *error = ERR_PTR(-ENOENT); | |
1986 | ||
1987 | inode = proc_pid_make_inode(dir->i_sb, task); | |
1988 | if (!inode) | |
1989 | goto out; | |
1990 | ei = PROC_I(inode); | |
1991 | ei->fd = fd; | |
1992 | inode->i_mode = S_IFREG | S_IRUSR; | |
1993 | inode->i_fop = &proc_fdinfo_file_operations; | |
1994 | dentry->d_op = &tid_fd_dentry_operations; | |
1995 | d_add(dentry, inode); | |
1996 | /* Close the race of the process dying before we return the dentry */ | |
1997 | if (tid_fd_revalidate(dentry, NULL)) | |
1998 | error = NULL; | |
1999 | ||
2000 | out: | |
2001 | return error; | |
2002 | } | |
2003 | ||
2004 | static struct dentry *proc_lookupfdinfo(struct inode *dir, | |
2005 | struct dentry *dentry, | |
2006 | struct nameidata *nd) | |
2007 | { | |
2008 | return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate); | |
2009 | } | |
2010 | ||
2011 | static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir) | |
2012 | { | |
2013 | return proc_readfd_common(filp, dirent, filldir, | |
2014 | proc_fdinfo_instantiate); | |
2015 | } | |
2016 | ||
2017 | static const struct file_operations proc_fdinfo_operations = { | |
2018 | .read = generic_read_dir, | |
2019 | .readdir = proc_readfdinfo, | |
2020 | }; | |
2021 | ||
2022 | /* | |
2023 | * proc directories can do almost nothing.. | |
2024 | */ | |
2025 | static const struct inode_operations proc_fdinfo_inode_operations = { | |
2026 | .lookup = proc_lookupfdinfo, | |
2027 | .setattr = proc_setattr, | |
2028 | }; | |
2029 | ||
2030 | ||
444ceed8 | 2031 | static struct dentry *proc_pident_instantiate(struct inode *dir, |
c5141e6d | 2032 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
444ceed8 | 2033 | { |
c5141e6d | 2034 | const struct pid_entry *p = ptr; |
444ceed8 EB |
2035 | struct inode *inode; |
2036 | struct proc_inode *ei; | |
bd6daba9 | 2037 | struct dentry *error = ERR_PTR(-ENOENT); |
444ceed8 | 2038 | |
61a28784 | 2039 | inode = proc_pid_make_inode(dir->i_sb, task); |
444ceed8 EB |
2040 | if (!inode) |
2041 | goto out; | |
2042 | ||
2043 | ei = PROC_I(inode); | |
2044 | inode->i_mode = p->mode; | |
2045 | if (S_ISDIR(inode->i_mode)) | |
2046 | inode->i_nlink = 2; /* Use getattr to fix if necessary */ | |
2047 | if (p->iop) | |
2048 | inode->i_op = p->iop; | |
2049 | if (p->fop) | |
2050 | inode->i_fop = p->fop; | |
2051 | ei->op = p->op; | |
2052 | dentry->d_op = &pid_dentry_operations; | |
2053 | d_add(dentry, inode); | |
2054 | /* Close the race of the process dying before we return the dentry */ | |
2055 | if (pid_revalidate(dentry, NULL)) | |
2056 | error = NULL; | |
2057 | out: | |
2058 | return error; | |
2059 | } | |
2060 | ||
1da177e4 LT |
2061 | static struct dentry *proc_pident_lookup(struct inode *dir, |
2062 | struct dentry *dentry, | |
c5141e6d | 2063 | const struct pid_entry *ents, |
7bcd6b0e | 2064 | unsigned int nents) |
1da177e4 | 2065 | { |
cd6a3ce9 | 2066 | struct dentry *error; |
99f89551 | 2067 | struct task_struct *task = get_proc_task(dir); |
c5141e6d | 2068 | const struct pid_entry *p, *last; |
1da177e4 | 2069 | |
cd6a3ce9 | 2070 | error = ERR_PTR(-ENOENT); |
1da177e4 | 2071 | |
99f89551 EB |
2072 | if (!task) |
2073 | goto out_no_task; | |
1da177e4 | 2074 | |
20cdc894 EB |
2075 | /* |
2076 | * Yes, it does not scale. And it should not. Don't add | |
2077 | * new entries into /proc/<tgid>/ without very good reasons. | |
2078 | */ | |
7bcd6b0e EB |
2079 | last = &ents[nents - 1]; |
2080 | for (p = ents; p <= last; p++) { | |
1da177e4 LT |
2081 | if (p->len != dentry->d_name.len) |
2082 | continue; | |
2083 | if (!memcmp(dentry->d_name.name, p->name, p->len)) | |
2084 | break; | |
2085 | } | |
7bcd6b0e | 2086 | if (p > last) |
1da177e4 LT |
2087 | goto out; |
2088 | ||
444ceed8 | 2089 | error = proc_pident_instantiate(dir, dentry, task, p); |
1da177e4 | 2090 | out: |
99f89551 EB |
2091 | put_task_struct(task); |
2092 | out_no_task: | |
cd6a3ce9 | 2093 | return error; |
1da177e4 LT |
2094 | } |
2095 | ||
c5141e6d ED |
2096 | static int proc_pident_fill_cache(struct file *filp, void *dirent, |
2097 | filldir_t filldir, struct task_struct *task, const struct pid_entry *p) | |
61a28784 EB |
2098 | { |
2099 | return proc_fill_cache(filp, dirent, filldir, p->name, p->len, | |
2100 | proc_pident_instantiate, task, p); | |
2101 | } | |
2102 | ||
28a6d671 EB |
2103 | static int proc_pident_readdir(struct file *filp, |
2104 | void *dirent, filldir_t filldir, | |
c5141e6d | 2105 | const struct pid_entry *ents, unsigned int nents) |
28a6d671 EB |
2106 | { |
2107 | int i; | |
2fddfeef | 2108 | struct dentry *dentry = filp->f_path.dentry; |
28a6d671 EB |
2109 | struct inode *inode = dentry->d_inode; |
2110 | struct task_struct *task = get_proc_task(inode); | |
c5141e6d | 2111 | const struct pid_entry *p, *last; |
28a6d671 EB |
2112 | ino_t ino; |
2113 | int ret; | |
2114 | ||
2115 | ret = -ENOENT; | |
2116 | if (!task) | |
61a28784 | 2117 | goto out_no_task; |
28a6d671 EB |
2118 | |
2119 | ret = 0; | |
28a6d671 EB |
2120 | i = filp->f_pos; |
2121 | switch (i) { | |
2122 | case 0: | |
2123 | ino = inode->i_ino; | |
2124 | if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0) | |
2125 | goto out; | |
2126 | i++; | |
2127 | filp->f_pos++; | |
2128 | /* fall through */ | |
2129 | case 1: | |
2130 | ino = parent_ino(dentry); | |
2131 | if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0) | |
2132 | goto out; | |
2133 | i++; | |
2134 | filp->f_pos++; | |
2135 | /* fall through */ | |
2136 | default: | |
2137 | i -= 2; | |
2138 | if (i >= nents) { | |
2139 | ret = 1; | |
2140 | goto out; | |
2141 | } | |
2142 | p = ents + i; | |
7bcd6b0e EB |
2143 | last = &ents[nents - 1]; |
2144 | while (p <= last) { | |
61a28784 | 2145 | if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0) |
28a6d671 EB |
2146 | goto out; |
2147 | filp->f_pos++; | |
2148 | p++; | |
2149 | } | |
2150 | } | |
2151 | ||
2152 | ret = 1; | |
2153 | out: | |
61a28784 EB |
2154 | put_task_struct(task); |
2155 | out_no_task: | |
28a6d671 | 2156 | return ret; |
1da177e4 LT |
2157 | } |
2158 | ||
28a6d671 EB |
2159 | #ifdef CONFIG_SECURITY |
2160 | static ssize_t proc_pid_attr_read(struct file * file, char __user * buf, | |
2161 | size_t count, loff_t *ppos) | |
2162 | { | |
2fddfeef | 2163 | struct inode * inode = file->f_path.dentry->d_inode; |
04ff9708 | 2164 | char *p = NULL; |
28a6d671 EB |
2165 | ssize_t length; |
2166 | struct task_struct *task = get_proc_task(inode); | |
2167 | ||
28a6d671 | 2168 | if (!task) |
04ff9708 | 2169 | return -ESRCH; |
28a6d671 EB |
2170 | |
2171 | length = security_getprocattr(task, | |
2fddfeef | 2172 | (char*)file->f_path.dentry->d_name.name, |
04ff9708 | 2173 | &p); |
28a6d671 | 2174 | put_task_struct(task); |
04ff9708 AV |
2175 | if (length > 0) |
2176 | length = simple_read_from_buffer(buf, count, ppos, p, length); | |
2177 | kfree(p); | |
28a6d671 | 2178 | return length; |
1da177e4 LT |
2179 | } |
2180 | ||
28a6d671 EB |
2181 | static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf, |
2182 | size_t count, loff_t *ppos) | |
2183 | { | |
2fddfeef | 2184 | struct inode * inode = file->f_path.dentry->d_inode; |
28a6d671 EB |
2185 | char *page; |
2186 | ssize_t length; | |
2187 | struct task_struct *task = get_proc_task(inode); | |
2188 | ||
2189 | length = -ESRCH; | |
2190 | if (!task) | |
2191 | goto out_no_task; | |
2192 | if (count > PAGE_SIZE) | |
2193 | count = PAGE_SIZE; | |
2194 | ||
2195 | /* No partial writes. */ | |
2196 | length = -EINVAL; | |
2197 | if (*ppos != 0) | |
2198 | goto out; | |
2199 | ||
2200 | length = -ENOMEM; | |
e12ba74d | 2201 | page = (char*)__get_free_page(GFP_TEMPORARY); |
28a6d671 EB |
2202 | if (!page) |
2203 | goto out; | |
2204 | ||
2205 | length = -EFAULT; | |
2206 | if (copy_from_user(page, buf, count)) | |
2207 | goto out_free; | |
2208 | ||
107db7c7 DH |
2209 | /* Guard against adverse ptrace interaction */ |
2210 | length = mutex_lock_interruptible(&task->cred_guard_mutex); | |
2211 | if (length < 0) | |
2212 | goto out_free; | |
2213 | ||
28a6d671 | 2214 | length = security_setprocattr(task, |
2fddfeef | 2215 | (char*)file->f_path.dentry->d_name.name, |
28a6d671 | 2216 | (void*)page, count); |
107db7c7 | 2217 | mutex_unlock(&task->cred_guard_mutex); |
28a6d671 EB |
2218 | out_free: |
2219 | free_page((unsigned long) page); | |
2220 | out: | |
2221 | put_task_struct(task); | |
2222 | out_no_task: | |
2223 | return length; | |
2224 | } | |
2225 | ||
00977a59 | 2226 | static const struct file_operations proc_pid_attr_operations = { |
28a6d671 EB |
2227 | .read = proc_pid_attr_read, |
2228 | .write = proc_pid_attr_write, | |
2229 | }; | |
2230 | ||
c5141e6d | 2231 | static const struct pid_entry attr_dir_stuff[] = { |
631f9c18 AD |
2232 | REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations), |
2233 | REG("prev", S_IRUGO, proc_pid_attr_operations), | |
2234 | REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2235 | REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2236 | REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2237 | REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
28a6d671 EB |
2238 | }; |
2239 | ||
72d9dcfc | 2240 | static int proc_attr_dir_readdir(struct file * filp, |
28a6d671 EB |
2241 | void * dirent, filldir_t filldir) |
2242 | { | |
2243 | return proc_pident_readdir(filp,dirent,filldir, | |
72d9dcfc | 2244 | attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff)); |
28a6d671 EB |
2245 | } |
2246 | ||
00977a59 | 2247 | static const struct file_operations proc_attr_dir_operations = { |
1da177e4 | 2248 | .read = generic_read_dir, |
72d9dcfc | 2249 | .readdir = proc_attr_dir_readdir, |
1da177e4 LT |
2250 | }; |
2251 | ||
72d9dcfc | 2252 | static struct dentry *proc_attr_dir_lookup(struct inode *dir, |
28a6d671 EB |
2253 | struct dentry *dentry, struct nameidata *nd) |
2254 | { | |
7bcd6b0e EB |
2255 | return proc_pident_lookup(dir, dentry, |
2256 | attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2257 | } |
2258 | ||
c5ef1c42 | 2259 | static const struct inode_operations proc_attr_dir_inode_operations = { |
72d9dcfc | 2260 | .lookup = proc_attr_dir_lookup, |
99f89551 | 2261 | .getattr = pid_getattr, |
6d76fa58 | 2262 | .setattr = proc_setattr, |
1da177e4 LT |
2263 | }; |
2264 | ||
28a6d671 EB |
2265 | #endif |
2266 | ||
698ba7b5 | 2267 | #ifdef CONFIG_ELF_CORE |
3cb4a0bb KH |
2268 | static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf, |
2269 | size_t count, loff_t *ppos) | |
2270 | { | |
2271 | struct task_struct *task = get_proc_task(file->f_dentry->d_inode); | |
2272 | struct mm_struct *mm; | |
2273 | char buffer[PROC_NUMBUF]; | |
2274 | size_t len; | |
2275 | int ret; | |
2276 | ||
2277 | if (!task) | |
2278 | return -ESRCH; | |
2279 | ||
2280 | ret = 0; | |
2281 | mm = get_task_mm(task); | |
2282 | if (mm) { | |
2283 | len = snprintf(buffer, sizeof(buffer), "%08lx\n", | |
2284 | ((mm->flags & MMF_DUMP_FILTER_MASK) >> | |
2285 | MMF_DUMP_FILTER_SHIFT)); | |
2286 | mmput(mm); | |
2287 | ret = simple_read_from_buffer(buf, count, ppos, buffer, len); | |
2288 | } | |
2289 | ||
2290 | put_task_struct(task); | |
2291 | ||
2292 | return ret; | |
2293 | } | |
2294 | ||
2295 | static ssize_t proc_coredump_filter_write(struct file *file, | |
2296 | const char __user *buf, | |
2297 | size_t count, | |
2298 | loff_t *ppos) | |
2299 | { | |
2300 | struct task_struct *task; | |
2301 | struct mm_struct *mm; | |
2302 | char buffer[PROC_NUMBUF], *end; | |
2303 | unsigned int val; | |
2304 | int ret; | |
2305 | int i; | |
2306 | unsigned long mask; | |
2307 | ||
2308 | ret = -EFAULT; | |
2309 | memset(buffer, 0, sizeof(buffer)); | |
2310 | if (count > sizeof(buffer) - 1) | |
2311 | count = sizeof(buffer) - 1; | |
2312 | if (copy_from_user(buffer, buf, count)) | |
2313 | goto out_no_task; | |
2314 | ||
2315 | ret = -EINVAL; | |
2316 | val = (unsigned int)simple_strtoul(buffer, &end, 0); | |
2317 | if (*end == '\n') | |
2318 | end++; | |
2319 | if (end - buffer == 0) | |
2320 | goto out_no_task; | |
2321 | ||
2322 | ret = -ESRCH; | |
2323 | task = get_proc_task(file->f_dentry->d_inode); | |
2324 | if (!task) | |
2325 | goto out_no_task; | |
2326 | ||
2327 | ret = end - buffer; | |
2328 | mm = get_task_mm(task); | |
2329 | if (!mm) | |
2330 | goto out_no_mm; | |
2331 | ||
2332 | for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) { | |
2333 | if (val & mask) | |
2334 | set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2335 | else | |
2336 | clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2337 | } | |
2338 | ||
2339 | mmput(mm); | |
2340 | out_no_mm: | |
2341 | put_task_struct(task); | |
2342 | out_no_task: | |
2343 | return ret; | |
2344 | } | |
2345 | ||
2346 | static const struct file_operations proc_coredump_filter_operations = { | |
2347 | .read = proc_coredump_filter_read, | |
2348 | .write = proc_coredump_filter_write, | |
2349 | }; | |
2350 | #endif | |
2351 | ||
28a6d671 EB |
2352 | /* |
2353 | * /proc/self: | |
2354 | */ | |
2355 | static int proc_self_readlink(struct dentry *dentry, char __user *buffer, | |
2356 | int buflen) | |
2357 | { | |
488e5bc4 | 2358 | struct pid_namespace *ns = dentry->d_sb->s_fs_info; |
b55fcb22 | 2359 | pid_t tgid = task_tgid_nr_ns(current, ns); |
28a6d671 | 2360 | char tmp[PROC_NUMBUF]; |
b55fcb22 | 2361 | if (!tgid) |
488e5bc4 | 2362 | return -ENOENT; |
b55fcb22 | 2363 | sprintf(tmp, "%d", tgid); |
28a6d671 EB |
2364 | return vfs_readlink(dentry,buffer,buflen,tmp); |
2365 | } | |
2366 | ||
2367 | static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd) | |
2368 | { | |
488e5bc4 | 2369 | struct pid_namespace *ns = dentry->d_sb->s_fs_info; |
b55fcb22 | 2370 | pid_t tgid = task_tgid_nr_ns(current, ns); |
7fee4868 AV |
2371 | char *name = ERR_PTR(-ENOENT); |
2372 | if (tgid) { | |
2373 | name = __getname(); | |
2374 | if (!name) | |
2375 | name = ERR_PTR(-ENOMEM); | |
2376 | else | |
2377 | sprintf(name, "%d", tgid); | |
2378 | } | |
2379 | nd_set_link(nd, name); | |
2380 | return NULL; | |
2381 | } | |
2382 | ||
2383 | static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd, | |
2384 | void *cookie) | |
2385 | { | |
2386 | char *s = nd_get_link(nd); | |
2387 | if (!IS_ERR(s)) | |
2388 | __putname(s); | |
28a6d671 EB |
2389 | } |
2390 | ||
c5ef1c42 | 2391 | static const struct inode_operations proc_self_inode_operations = { |
28a6d671 EB |
2392 | .readlink = proc_self_readlink, |
2393 | .follow_link = proc_self_follow_link, | |
7fee4868 | 2394 | .put_link = proc_self_put_link, |
28a6d671 EB |
2395 | }; |
2396 | ||
801199ce EB |
2397 | /* |
2398 | * proc base | |
2399 | * | |
2400 | * These are the directory entries in the root directory of /proc | |
2401 | * that properly belong to the /proc filesystem, as they describe | |
2402 | * describe something that is process related. | |
2403 | */ | |
c5141e6d | 2404 | static const struct pid_entry proc_base_stuff[] = { |
61a28784 | 2405 | NOD("self", S_IFLNK|S_IRWXUGO, |
801199ce | 2406 | &proc_self_inode_operations, NULL, {}), |
801199ce EB |
2407 | }; |
2408 | ||
2409 | /* | |
2410 | * Exceptional case: normally we are not allowed to unhash a busy | |
2411 | * directory. In this case, however, we can do it - no aliasing problems | |
2412 | * due to the way we treat inodes. | |
2413 | */ | |
2414 | static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd) | |
2415 | { | |
2416 | struct inode *inode = dentry->d_inode; | |
2417 | struct task_struct *task = get_proc_task(inode); | |
2418 | if (task) { | |
2419 | put_task_struct(task); | |
2420 | return 1; | |
2421 | } | |
2422 | d_drop(dentry); | |
2423 | return 0; | |
2424 | } | |
2425 | ||
d72f71eb | 2426 | static const struct dentry_operations proc_base_dentry_operations = |
801199ce EB |
2427 | { |
2428 | .d_revalidate = proc_base_revalidate, | |
2429 | .d_delete = pid_delete_dentry, | |
2430 | }; | |
2431 | ||
444ceed8 | 2432 | static struct dentry *proc_base_instantiate(struct inode *dir, |
c5141e6d | 2433 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
801199ce | 2434 | { |
c5141e6d | 2435 | const struct pid_entry *p = ptr; |
801199ce | 2436 | struct inode *inode; |
801199ce | 2437 | struct proc_inode *ei; |
444ceed8 | 2438 | struct dentry *error = ERR_PTR(-EINVAL); |
801199ce EB |
2439 | |
2440 | /* Allocate the inode */ | |
2441 | error = ERR_PTR(-ENOMEM); | |
2442 | inode = new_inode(dir->i_sb); | |
2443 | if (!inode) | |
2444 | goto out; | |
2445 | ||
2446 | /* Initialize the inode */ | |
2447 | ei = PROC_I(inode); | |
2448 | inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; | |
801199ce EB |
2449 | |
2450 | /* | |
2451 | * grab the reference to the task. | |
2452 | */ | |
1a657f78 | 2453 | ei->pid = get_task_pid(task, PIDTYPE_PID); |
801199ce EB |
2454 | if (!ei->pid) |
2455 | goto out_iput; | |
2456 | ||
801199ce EB |
2457 | inode->i_mode = p->mode; |
2458 | if (S_ISDIR(inode->i_mode)) | |
2459 | inode->i_nlink = 2; | |
2460 | if (S_ISLNK(inode->i_mode)) | |
2461 | inode->i_size = 64; | |
2462 | if (p->iop) | |
2463 | inode->i_op = p->iop; | |
2464 | if (p->fop) | |
2465 | inode->i_fop = p->fop; | |
2466 | ei->op = p->op; | |
2467 | dentry->d_op = &proc_base_dentry_operations; | |
2468 | d_add(dentry, inode); | |
2469 | error = NULL; | |
2470 | out: | |
801199ce EB |
2471 | return error; |
2472 | out_iput: | |
2473 | iput(inode); | |
2474 | goto out; | |
2475 | } | |
2476 | ||
444ceed8 EB |
2477 | static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry) |
2478 | { | |
2479 | struct dentry *error; | |
2480 | struct task_struct *task = get_proc_task(dir); | |
c5141e6d | 2481 | const struct pid_entry *p, *last; |
444ceed8 EB |
2482 | |
2483 | error = ERR_PTR(-ENOENT); | |
2484 | ||
2485 | if (!task) | |
2486 | goto out_no_task; | |
2487 | ||
2488 | /* Lookup the directory entry */ | |
7bcd6b0e EB |
2489 | last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1]; |
2490 | for (p = proc_base_stuff; p <= last; p++) { | |
444ceed8 EB |
2491 | if (p->len != dentry->d_name.len) |
2492 | continue; | |
2493 | if (!memcmp(dentry->d_name.name, p->name, p->len)) | |
2494 | break; | |
2495 | } | |
7bcd6b0e | 2496 | if (p > last) |
444ceed8 EB |
2497 | goto out; |
2498 | ||
2499 | error = proc_base_instantiate(dir, dentry, task, p); | |
2500 | ||
2501 | out: | |
2502 | put_task_struct(task); | |
2503 | out_no_task: | |
2504 | return error; | |
2505 | } | |
2506 | ||
c5141e6d ED |
2507 | static int proc_base_fill_cache(struct file *filp, void *dirent, |
2508 | filldir_t filldir, struct task_struct *task, const struct pid_entry *p) | |
61a28784 EB |
2509 | { |
2510 | return proc_fill_cache(filp, dirent, filldir, p->name, p->len, | |
2511 | proc_base_instantiate, task, p); | |
2512 | } | |
2513 | ||
aba76fdb | 2514 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
297c5d92 AR |
2515 | static int do_io_accounting(struct task_struct *task, char *buffer, int whole) |
2516 | { | |
940389b8 | 2517 | struct task_io_accounting acct = task->ioac; |
5995477a AR |
2518 | unsigned long flags; |
2519 | ||
2520 | if (whole && lock_task_sighand(task, &flags)) { | |
2521 | struct task_struct *t = task; | |
2522 | ||
2523 | task_io_accounting_add(&acct, &task->signal->ioac); | |
2524 | while_each_thread(task, t) | |
2525 | task_io_accounting_add(&acct, &t->ioac); | |
2526 | ||
2527 | unlock_task_sighand(task, &flags); | |
297c5d92 | 2528 | } |
aba76fdb AM |
2529 | return sprintf(buffer, |
2530 | "rchar: %llu\n" | |
2531 | "wchar: %llu\n" | |
2532 | "syscr: %llu\n" | |
2533 | "syscw: %llu\n" | |
2534 | "read_bytes: %llu\n" | |
2535 | "write_bytes: %llu\n" | |
2536 | "cancelled_write_bytes: %llu\n", | |
7c44319d AB |
2537 | (unsigned long long)acct.rchar, |
2538 | (unsigned long long)acct.wchar, | |
2539 | (unsigned long long)acct.syscr, | |
2540 | (unsigned long long)acct.syscw, | |
2541 | (unsigned long long)acct.read_bytes, | |
2542 | (unsigned long long)acct.write_bytes, | |
2543 | (unsigned long long)acct.cancelled_write_bytes); | |
297c5d92 AR |
2544 | } |
2545 | ||
2546 | static int proc_tid_io_accounting(struct task_struct *task, char *buffer) | |
2547 | { | |
2548 | return do_io_accounting(task, buffer, 0); | |
aba76fdb | 2549 | } |
297c5d92 AR |
2550 | |
2551 | static int proc_tgid_io_accounting(struct task_struct *task, char *buffer) | |
2552 | { | |
2553 | return do_io_accounting(task, buffer, 1); | |
2554 | } | |
2555 | #endif /* CONFIG_TASK_IO_ACCOUNTING */ | |
aba76fdb | 2556 | |
47830723 KC |
2557 | static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns, |
2558 | struct pid *pid, struct task_struct *task) | |
2559 | { | |
2560 | seq_printf(m, "%08x\n", task->personality); | |
2561 | return 0; | |
2562 | } | |
2563 | ||
28a6d671 EB |
2564 | /* |
2565 | * Thread groups | |
2566 | */ | |
00977a59 | 2567 | static const struct file_operations proc_task_operations; |
c5ef1c42 | 2568 | static const struct inode_operations proc_task_inode_operations; |
20cdc894 | 2569 | |
c5141e6d | 2570 | static const struct pid_entry tgid_base_stuff[] = { |
631f9c18 AD |
2571 | DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations), |
2572 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), | |
2573 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), | |
b2211a36 | 2574 | #ifdef CONFIG_NET |
631f9c18 | 2575 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), |
b2211a36 | 2576 | #endif |
631f9c18 AD |
2577 | REG("environ", S_IRUSR, proc_environ_operations), |
2578 | INF("auxv", S_IRUSR, proc_pid_auxv), | |
2579 | ONE("status", S_IRUGO, proc_pid_status), | |
2580 | ONE("personality", S_IRUSR, proc_pid_personality), | |
2581 | INF("limits", S_IRUSR, proc_pid_limits), | |
43ae34cb | 2582 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 2583 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
ebcb6734 | 2584 | #endif |
4614a696 | 2585 | REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations), |
ebcb6734 | 2586 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
631f9c18 | 2587 | INF("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 2588 | #endif |
631f9c18 AD |
2589 | INF("cmdline", S_IRUGO, proc_pid_cmdline), |
2590 | ONE("stat", S_IRUGO, proc_tgid_stat), | |
2591 | ONE("statm", S_IRUGO, proc_pid_statm), | |
2592 | REG("maps", S_IRUGO, proc_maps_operations), | |
28a6d671 | 2593 | #ifdef CONFIG_NUMA |
631f9c18 | 2594 | REG("numa_maps", S_IRUGO, proc_numa_maps_operations), |
28a6d671 | 2595 | #endif |
631f9c18 AD |
2596 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
2597 | LNK("cwd", proc_cwd_link), | |
2598 | LNK("root", proc_root_link), | |
2599 | LNK("exe", proc_exe_link), | |
2600 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
2601 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
2602 | REG("mountstats", S_IRUSR, proc_mountstats_operations), | |
1e883281 | 2603 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 AD |
2604 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
2605 | REG("smaps", S_IRUGO, proc_smaps_operations), | |
2606 | REG("pagemap", S_IRUSR, proc_pagemap_operations), | |
28a6d671 EB |
2607 | #endif |
2608 | #ifdef CONFIG_SECURITY | |
631f9c18 | 2609 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
2610 | #endif |
2611 | #ifdef CONFIG_KALLSYMS | |
631f9c18 | 2612 | INF("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 2613 | #endif |
2ec220e2 KC |
2614 | #ifdef CONFIG_STACKTRACE |
2615 | ONE("stack", S_IRUSR, proc_pid_stack), | |
28a6d671 EB |
2616 | #endif |
2617 | #ifdef CONFIG_SCHEDSTATS | |
631f9c18 | 2618 | INF("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 2619 | #endif |
9745512c | 2620 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 2621 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 2622 | #endif |
8793d854 | 2623 | #ifdef CONFIG_PROC_PID_CPUSET |
631f9c18 | 2624 | REG("cpuset", S_IRUGO, proc_cpuset_operations), |
a424316c PM |
2625 | #endif |
2626 | #ifdef CONFIG_CGROUPS | |
631f9c18 | 2627 | REG("cgroup", S_IRUGO, proc_cgroup_operations), |
28a6d671 | 2628 | #endif |
631f9c18 AD |
2629 | INF("oom_score", S_IRUGO, proc_oom_score), |
2630 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations), | |
28a6d671 | 2631 | #ifdef CONFIG_AUDITSYSCALL |
631f9c18 AD |
2632 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
2633 | REG("sessionid", S_IRUGO, proc_sessionid_operations), | |
28a6d671 | 2634 | #endif |
f4f154fd | 2635 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 2636 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
f4f154fd | 2637 | #endif |
698ba7b5 | 2638 | #ifdef CONFIG_ELF_CORE |
631f9c18 | 2639 | REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations), |
3cb4a0bb | 2640 | #endif |
aba76fdb | 2641 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
631f9c18 | 2642 | INF("io", S_IRUGO, proc_tgid_io_accounting), |
aba76fdb | 2643 | #endif |
28a6d671 | 2644 | }; |
1da177e4 | 2645 | |
28a6d671 | 2646 | static int proc_tgid_base_readdir(struct file * filp, |
1da177e4 LT |
2647 | void * dirent, filldir_t filldir) |
2648 | { | |
2649 | return proc_pident_readdir(filp,dirent,filldir, | |
28a6d671 | 2650 | tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff)); |
1da177e4 LT |
2651 | } |
2652 | ||
00977a59 | 2653 | static const struct file_operations proc_tgid_base_operations = { |
1da177e4 | 2654 | .read = generic_read_dir, |
28a6d671 | 2655 | .readdir = proc_tgid_base_readdir, |
1da177e4 LT |
2656 | }; |
2657 | ||
28a6d671 | 2658 | static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){ |
7bcd6b0e EB |
2659 | return proc_pident_lookup(dir, dentry, |
2660 | tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
2661 | } |
2662 | ||
c5ef1c42 | 2663 | static const struct inode_operations proc_tgid_base_inode_operations = { |
28a6d671 | 2664 | .lookup = proc_tgid_base_lookup, |
99f89551 | 2665 | .getattr = pid_getattr, |
6d76fa58 | 2666 | .setattr = proc_setattr, |
1da177e4 | 2667 | }; |
1da177e4 | 2668 | |
60347f67 | 2669 | static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid) |
1da177e4 | 2670 | { |
48e6484d | 2671 | struct dentry *dentry, *leader, *dir; |
8578cea7 | 2672 | char buf[PROC_NUMBUF]; |
48e6484d EB |
2673 | struct qstr name; |
2674 | ||
2675 | name.name = buf; | |
60347f67 PE |
2676 | name.len = snprintf(buf, sizeof(buf), "%d", pid); |
2677 | dentry = d_hash_and_lookup(mnt->mnt_root, &name); | |
48e6484d | 2678 | if (dentry) { |
29f12ca3 | 2679 | shrink_dcache_parent(dentry); |
48e6484d EB |
2680 | d_drop(dentry); |
2681 | dput(dentry); | |
2682 | } | |
1da177e4 | 2683 | |
48e6484d | 2684 | name.name = buf; |
60347f67 PE |
2685 | name.len = snprintf(buf, sizeof(buf), "%d", tgid); |
2686 | leader = d_hash_and_lookup(mnt->mnt_root, &name); | |
48e6484d EB |
2687 | if (!leader) |
2688 | goto out; | |
1da177e4 | 2689 | |
48e6484d EB |
2690 | name.name = "task"; |
2691 | name.len = strlen(name.name); | |
2692 | dir = d_hash_and_lookup(leader, &name); | |
2693 | if (!dir) | |
2694 | goto out_put_leader; | |
2695 | ||
2696 | name.name = buf; | |
60347f67 | 2697 | name.len = snprintf(buf, sizeof(buf), "%d", pid); |
48e6484d EB |
2698 | dentry = d_hash_and_lookup(dir, &name); |
2699 | if (dentry) { | |
2700 | shrink_dcache_parent(dentry); | |
2701 | d_drop(dentry); | |
2702 | dput(dentry); | |
1da177e4 | 2703 | } |
48e6484d EB |
2704 | |
2705 | dput(dir); | |
2706 | out_put_leader: | |
2707 | dput(leader); | |
2708 | out: | |
2709 | return; | |
1da177e4 LT |
2710 | } |
2711 | ||
0895e91d RD |
2712 | /** |
2713 | * proc_flush_task - Remove dcache entries for @task from the /proc dcache. | |
2714 | * @task: task that should be flushed. | |
2715 | * | |
2716 | * When flushing dentries from proc, one needs to flush them from global | |
60347f67 | 2717 | * proc (proc_mnt) and from all the namespaces' procs this task was seen |
0895e91d RD |
2718 | * in. This call is supposed to do all of this job. |
2719 | * | |
2720 | * Looks in the dcache for | |
2721 | * /proc/@pid | |
2722 | * /proc/@tgid/task/@pid | |
2723 | * if either directory is present flushes it and all of it'ts children | |
2724 | * from the dcache. | |
2725 | * | |
2726 | * It is safe and reasonable to cache /proc entries for a task until | |
2727 | * that task exits. After that they just clog up the dcache with | |
2728 | * useless entries, possibly causing useful dcache entries to be | |
2729 | * flushed instead. This routine is proved to flush those useless | |
2730 | * dcache entries at process exit time. | |
2731 | * | |
2732 | * NOTE: This routine is just an optimization so it does not guarantee | |
2733 | * that no dcache entries will exist at process exit time it | |
2734 | * just makes it very unlikely that any will persist. | |
60347f67 PE |
2735 | */ |
2736 | ||
2737 | void proc_flush_task(struct task_struct *task) | |
2738 | { | |
9fcc2d15 | 2739 | int i; |
9b4d1cbe | 2740 | struct pid *pid, *tgid; |
130f77ec PE |
2741 | struct upid *upid; |
2742 | ||
130f77ec | 2743 | pid = task_pid(task); |
9b4d1cbe | 2744 | tgid = task_tgid(task); |
130f77ec | 2745 | |
9fcc2d15 | 2746 | for (i = 0; i <= pid->level; i++) { |
130f77ec PE |
2747 | upid = &pid->numbers[i]; |
2748 | proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr, | |
9b4d1cbe | 2749 | tgid->numbers[i].nr); |
130f77ec | 2750 | } |
6f4e6433 PE |
2751 | |
2752 | upid = &pid->numbers[pid->level]; | |
2753 | if (upid->nr == 1) | |
2754 | pid_ns_release_proc(upid->ns); | |
60347f67 PE |
2755 | } |
2756 | ||
9711ef99 AB |
2757 | static struct dentry *proc_pid_instantiate(struct inode *dir, |
2758 | struct dentry * dentry, | |
c5141e6d | 2759 | struct task_struct *task, const void *ptr) |
444ceed8 EB |
2760 | { |
2761 | struct dentry *error = ERR_PTR(-ENOENT); | |
2762 | struct inode *inode; | |
2763 | ||
61a28784 | 2764 | inode = proc_pid_make_inode(dir->i_sb, task); |
444ceed8 EB |
2765 | if (!inode) |
2766 | goto out; | |
2767 | ||
2768 | inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO; | |
2769 | inode->i_op = &proc_tgid_base_inode_operations; | |
2770 | inode->i_fop = &proc_tgid_base_operations; | |
2771 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 VN |
2772 | |
2773 | inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff, | |
2774 | ARRAY_SIZE(tgid_base_stuff)); | |
444ceed8 EB |
2775 | |
2776 | dentry->d_op = &pid_dentry_operations; | |
2777 | ||
2778 | d_add(dentry, inode); | |
2779 | /* Close the race of the process dying before we return the dentry */ | |
2780 | if (pid_revalidate(dentry, NULL)) | |
2781 | error = NULL; | |
2782 | out: | |
2783 | return error; | |
2784 | } | |
2785 | ||
1da177e4 LT |
2786 | struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd) |
2787 | { | |
cd6a3ce9 | 2788 | struct dentry *result = ERR_PTR(-ENOENT); |
1da177e4 | 2789 | struct task_struct *task; |
1da177e4 | 2790 | unsigned tgid; |
b488893a | 2791 | struct pid_namespace *ns; |
1da177e4 | 2792 | |
801199ce EB |
2793 | result = proc_base_lookup(dir, dentry); |
2794 | if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT) | |
2795 | goto out; | |
2796 | ||
1da177e4 LT |
2797 | tgid = name_to_int(dentry); |
2798 | if (tgid == ~0U) | |
2799 | goto out; | |
2800 | ||
b488893a | 2801 | ns = dentry->d_sb->s_fs_info; |
de758734 | 2802 | rcu_read_lock(); |
b488893a | 2803 | task = find_task_by_pid_ns(tgid, ns); |
1da177e4 LT |
2804 | if (task) |
2805 | get_task_struct(task); | |
de758734 | 2806 | rcu_read_unlock(); |
1da177e4 LT |
2807 | if (!task) |
2808 | goto out; | |
2809 | ||
444ceed8 | 2810 | result = proc_pid_instantiate(dir, dentry, task, NULL); |
1da177e4 | 2811 | put_task_struct(task); |
1da177e4 | 2812 | out: |
cd6a3ce9 | 2813 | return result; |
1da177e4 LT |
2814 | } |
2815 | ||
1da177e4 | 2816 | /* |
0804ef4b | 2817 | * Find the first task with tgid >= tgid |
0bc58a91 | 2818 | * |
1da177e4 | 2819 | */ |
19fd4bb2 EB |
2820 | struct tgid_iter { |
2821 | unsigned int tgid; | |
0804ef4b | 2822 | struct task_struct *task; |
19fd4bb2 EB |
2823 | }; |
2824 | static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter) | |
2825 | { | |
0804ef4b | 2826 | struct pid *pid; |
1da177e4 | 2827 | |
19fd4bb2 EB |
2828 | if (iter.task) |
2829 | put_task_struct(iter.task); | |
454cc105 | 2830 | rcu_read_lock(); |
0804ef4b | 2831 | retry: |
19fd4bb2 EB |
2832 | iter.task = NULL; |
2833 | pid = find_ge_pid(iter.tgid, ns); | |
0804ef4b | 2834 | if (pid) { |
19fd4bb2 EB |
2835 | iter.tgid = pid_nr_ns(pid, ns); |
2836 | iter.task = pid_task(pid, PIDTYPE_PID); | |
0804ef4b EB |
2837 | /* What we to know is if the pid we have find is the |
2838 | * pid of a thread_group_leader. Testing for task | |
2839 | * being a thread_group_leader is the obvious thing | |
2840 | * todo but there is a window when it fails, due to | |
2841 | * the pid transfer logic in de_thread. | |
2842 | * | |
2843 | * So we perform the straight forward test of seeing | |
2844 | * if the pid we have found is the pid of a thread | |
2845 | * group leader, and don't worry if the task we have | |
2846 | * found doesn't happen to be a thread group leader. | |
2847 | * As we don't care in the case of readdir. | |
2848 | */ | |
19fd4bb2 EB |
2849 | if (!iter.task || !has_group_leader_pid(iter.task)) { |
2850 | iter.tgid += 1; | |
0804ef4b | 2851 | goto retry; |
19fd4bb2 EB |
2852 | } |
2853 | get_task_struct(iter.task); | |
0bc58a91 | 2854 | } |
454cc105 | 2855 | rcu_read_unlock(); |
19fd4bb2 | 2856 | return iter; |
1da177e4 LT |
2857 | } |
2858 | ||
7bcd6b0e | 2859 | #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff)) |
0804ef4b | 2860 | |
61a28784 | 2861 | static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir, |
19fd4bb2 | 2862 | struct tgid_iter iter) |
61a28784 EB |
2863 | { |
2864 | char name[PROC_NUMBUF]; | |
19fd4bb2 | 2865 | int len = snprintf(name, sizeof(name), "%d", iter.tgid); |
61a28784 | 2866 | return proc_fill_cache(filp, dirent, filldir, name, len, |
19fd4bb2 | 2867 | proc_pid_instantiate, iter.task, NULL); |
61a28784 EB |
2868 | } |
2869 | ||
1da177e4 LT |
2870 | /* for the /proc/ directory itself, after non-process stuff has been done */ |
2871 | int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir) | |
2872 | { | |
1da177e4 | 2873 | unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY; |
2fddfeef | 2874 | struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode); |
19fd4bb2 | 2875 | struct tgid_iter iter; |
b488893a | 2876 | struct pid_namespace *ns; |
1da177e4 | 2877 | |
61a28784 EB |
2878 | if (!reaper) |
2879 | goto out_no_task; | |
2880 | ||
7bcd6b0e | 2881 | for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) { |
c5141e6d | 2882 | const struct pid_entry *p = &proc_base_stuff[nr]; |
61a28784 | 2883 | if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0) |
801199ce | 2884 | goto out; |
1da177e4 LT |
2885 | } |
2886 | ||
b488893a | 2887 | ns = filp->f_dentry->d_sb->s_fs_info; |
19fd4bb2 EB |
2888 | iter.task = NULL; |
2889 | iter.tgid = filp->f_pos - TGID_OFFSET; | |
2890 | for (iter = next_tgid(ns, iter); | |
2891 | iter.task; | |
2892 | iter.tgid += 1, iter = next_tgid(ns, iter)) { | |
2893 | filp->f_pos = iter.tgid + TGID_OFFSET; | |
2894 | if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) { | |
2895 | put_task_struct(iter.task); | |
0804ef4b | 2896 | goto out; |
1da177e4 | 2897 | } |
0bc58a91 | 2898 | } |
0804ef4b EB |
2899 | filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET; |
2900 | out: | |
61a28784 EB |
2901 | put_task_struct(reaper); |
2902 | out_no_task: | |
0bc58a91 EB |
2903 | return 0; |
2904 | } | |
1da177e4 | 2905 | |
28a6d671 EB |
2906 | /* |
2907 | * Tasks | |
2908 | */ | |
c5141e6d | 2909 | static const struct pid_entry tid_base_stuff[] = { |
631f9c18 AD |
2910 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), |
2911 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fd_operations), | |
2912 | REG("environ", S_IRUSR, proc_environ_operations), | |
2913 | INF("auxv", S_IRUSR, proc_pid_auxv), | |
2914 | ONE("status", S_IRUGO, proc_pid_status), | |
2915 | ONE("personality", S_IRUSR, proc_pid_personality), | |
2916 | INF("limits", S_IRUSR, proc_pid_limits), | |
43ae34cb | 2917 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 2918 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
ebcb6734 | 2919 | #endif |
4614a696 | 2920 | REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations), |
ebcb6734 | 2921 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
631f9c18 | 2922 | INF("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 2923 | #endif |
631f9c18 AD |
2924 | INF("cmdline", S_IRUGO, proc_pid_cmdline), |
2925 | ONE("stat", S_IRUGO, proc_tid_stat), | |
2926 | ONE("statm", S_IRUGO, proc_pid_statm), | |
2927 | REG("maps", S_IRUGO, proc_maps_operations), | |
28a6d671 | 2928 | #ifdef CONFIG_NUMA |
631f9c18 | 2929 | REG("numa_maps", S_IRUGO, proc_numa_maps_operations), |
28a6d671 | 2930 | #endif |
631f9c18 AD |
2931 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
2932 | LNK("cwd", proc_cwd_link), | |
2933 | LNK("root", proc_root_link), | |
2934 | LNK("exe", proc_exe_link), | |
2935 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
2936 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
1e883281 | 2937 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 AD |
2938 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
2939 | REG("smaps", S_IRUGO, proc_smaps_operations), | |
2940 | REG("pagemap", S_IRUSR, proc_pagemap_operations), | |
28a6d671 EB |
2941 | #endif |
2942 | #ifdef CONFIG_SECURITY | |
631f9c18 | 2943 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
2944 | #endif |
2945 | #ifdef CONFIG_KALLSYMS | |
631f9c18 | 2946 | INF("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 2947 | #endif |
2ec220e2 KC |
2948 | #ifdef CONFIG_STACKTRACE |
2949 | ONE("stack", S_IRUSR, proc_pid_stack), | |
28a6d671 EB |
2950 | #endif |
2951 | #ifdef CONFIG_SCHEDSTATS | |
631f9c18 | 2952 | INF("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 2953 | #endif |
9745512c | 2954 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 2955 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 2956 | #endif |
8793d854 | 2957 | #ifdef CONFIG_PROC_PID_CPUSET |
631f9c18 | 2958 | REG("cpuset", S_IRUGO, proc_cpuset_operations), |
a424316c PM |
2959 | #endif |
2960 | #ifdef CONFIG_CGROUPS | |
631f9c18 | 2961 | REG("cgroup", S_IRUGO, proc_cgroup_operations), |
28a6d671 | 2962 | #endif |
631f9c18 AD |
2963 | INF("oom_score", S_IRUGO, proc_oom_score), |
2964 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations), | |
28a6d671 | 2965 | #ifdef CONFIG_AUDITSYSCALL |
631f9c18 AD |
2966 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
2967 | REG("sessionid", S_IRUSR, proc_sessionid_operations), | |
28a6d671 | 2968 | #endif |
f4f154fd | 2969 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 2970 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
f4f154fd | 2971 | #endif |
297c5d92 | 2972 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
631f9c18 | 2973 | INF("io", S_IRUGO, proc_tid_io_accounting), |
297c5d92 | 2974 | #endif |
28a6d671 EB |
2975 | }; |
2976 | ||
2977 | static int proc_tid_base_readdir(struct file * filp, | |
2978 | void * dirent, filldir_t filldir) | |
2979 | { | |
2980 | return proc_pident_readdir(filp,dirent,filldir, | |
2981 | tid_base_stuff,ARRAY_SIZE(tid_base_stuff)); | |
2982 | } | |
2983 | ||
2984 | static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){ | |
7bcd6b0e EB |
2985 | return proc_pident_lookup(dir, dentry, |
2986 | tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
2987 | } |
2988 | ||
00977a59 | 2989 | static const struct file_operations proc_tid_base_operations = { |
28a6d671 EB |
2990 | .read = generic_read_dir, |
2991 | .readdir = proc_tid_base_readdir, | |
2992 | }; | |
2993 | ||
c5ef1c42 | 2994 | static const struct inode_operations proc_tid_base_inode_operations = { |
28a6d671 EB |
2995 | .lookup = proc_tid_base_lookup, |
2996 | .getattr = pid_getattr, | |
2997 | .setattr = proc_setattr, | |
2998 | }; | |
2999 | ||
444ceed8 | 3000 | static struct dentry *proc_task_instantiate(struct inode *dir, |
c5141e6d | 3001 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
444ceed8 EB |
3002 | { |
3003 | struct dentry *error = ERR_PTR(-ENOENT); | |
3004 | struct inode *inode; | |
61a28784 | 3005 | inode = proc_pid_make_inode(dir->i_sb, task); |
444ceed8 EB |
3006 | |
3007 | if (!inode) | |
3008 | goto out; | |
3009 | inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO; | |
3010 | inode->i_op = &proc_tid_base_inode_operations; | |
3011 | inode->i_fop = &proc_tid_base_operations; | |
3012 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 VN |
3013 | |
3014 | inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff, | |
3015 | ARRAY_SIZE(tid_base_stuff)); | |
444ceed8 EB |
3016 | |
3017 | dentry->d_op = &pid_dentry_operations; | |
3018 | ||
3019 | d_add(dentry, inode); | |
3020 | /* Close the race of the process dying before we return the dentry */ | |
3021 | if (pid_revalidate(dentry, NULL)) | |
3022 | error = NULL; | |
3023 | out: | |
3024 | return error; | |
3025 | } | |
3026 | ||
28a6d671 EB |
3027 | static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd) |
3028 | { | |
3029 | struct dentry *result = ERR_PTR(-ENOENT); | |
3030 | struct task_struct *task; | |
3031 | struct task_struct *leader = get_proc_task(dir); | |
28a6d671 | 3032 | unsigned tid; |
b488893a | 3033 | struct pid_namespace *ns; |
28a6d671 EB |
3034 | |
3035 | if (!leader) | |
3036 | goto out_no_task; | |
3037 | ||
3038 | tid = name_to_int(dentry); | |
3039 | if (tid == ~0U) | |
3040 | goto out; | |
3041 | ||
b488893a | 3042 | ns = dentry->d_sb->s_fs_info; |
28a6d671 | 3043 | rcu_read_lock(); |
b488893a | 3044 | task = find_task_by_pid_ns(tid, ns); |
28a6d671 EB |
3045 | if (task) |
3046 | get_task_struct(task); | |
3047 | rcu_read_unlock(); | |
3048 | if (!task) | |
3049 | goto out; | |
bac0abd6 | 3050 | if (!same_thread_group(leader, task)) |
28a6d671 EB |
3051 | goto out_drop_task; |
3052 | ||
444ceed8 | 3053 | result = proc_task_instantiate(dir, dentry, task, NULL); |
28a6d671 EB |
3054 | out_drop_task: |
3055 | put_task_struct(task); | |
3056 | out: | |
3057 | put_task_struct(leader); | |
3058 | out_no_task: | |
3059 | return result; | |
3060 | } | |
3061 | ||
0bc58a91 EB |
3062 | /* |
3063 | * Find the first tid of a thread group to return to user space. | |
3064 | * | |
3065 | * Usually this is just the thread group leader, but if the users | |
3066 | * buffer was too small or there was a seek into the middle of the | |
3067 | * directory we have more work todo. | |
3068 | * | |
3069 | * In the case of a short read we start with find_task_by_pid. | |
3070 | * | |
3071 | * In the case of a seek we start with the leader and walk nr | |
3072 | * threads past it. | |
3073 | */ | |
cc288738 | 3074 | static struct task_struct *first_tid(struct task_struct *leader, |
b488893a | 3075 | int tid, int nr, struct pid_namespace *ns) |
0bc58a91 | 3076 | { |
a872ff0c | 3077 | struct task_struct *pos; |
1da177e4 | 3078 | |
cc288738 | 3079 | rcu_read_lock(); |
0bc58a91 EB |
3080 | /* Attempt to start with the pid of a thread */ |
3081 | if (tid && (nr > 0)) { | |
b488893a | 3082 | pos = find_task_by_pid_ns(tid, ns); |
a872ff0c ON |
3083 | if (pos && (pos->group_leader == leader)) |
3084 | goto found; | |
0bc58a91 | 3085 | } |
1da177e4 | 3086 | |
0bc58a91 | 3087 | /* If nr exceeds the number of threads there is nothing todo */ |
a872ff0c ON |
3088 | pos = NULL; |
3089 | if (nr && nr >= get_nr_threads(leader)) | |
3090 | goto out; | |
1da177e4 | 3091 | |
a872ff0c ON |
3092 | /* If we haven't found our starting place yet start |
3093 | * with the leader and walk nr threads forward. | |
0bc58a91 | 3094 | */ |
a872ff0c ON |
3095 | for (pos = leader; nr > 0; --nr) { |
3096 | pos = next_thread(pos); | |
3097 | if (pos == leader) { | |
3098 | pos = NULL; | |
3099 | goto out; | |
3100 | } | |
1da177e4 | 3101 | } |
a872ff0c ON |
3102 | found: |
3103 | get_task_struct(pos); | |
3104 | out: | |
cc288738 | 3105 | rcu_read_unlock(); |
0bc58a91 EB |
3106 | return pos; |
3107 | } | |
3108 | ||
3109 | /* | |
3110 | * Find the next thread in the thread list. | |
3111 | * Return NULL if there is an error or no next thread. | |
3112 | * | |
3113 | * The reference to the input task_struct is released. | |
3114 | */ | |
3115 | static struct task_struct *next_tid(struct task_struct *start) | |
3116 | { | |
c1df7fb8 | 3117 | struct task_struct *pos = NULL; |
cc288738 | 3118 | rcu_read_lock(); |
c1df7fb8 | 3119 | if (pid_alive(start)) { |
0bc58a91 | 3120 | pos = next_thread(start); |
c1df7fb8 ON |
3121 | if (thread_group_leader(pos)) |
3122 | pos = NULL; | |
3123 | else | |
3124 | get_task_struct(pos); | |
3125 | } | |
cc288738 | 3126 | rcu_read_unlock(); |
0bc58a91 EB |
3127 | put_task_struct(start); |
3128 | return pos; | |
1da177e4 LT |
3129 | } |
3130 | ||
61a28784 EB |
3131 | static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir, |
3132 | struct task_struct *task, int tid) | |
3133 | { | |
3134 | char name[PROC_NUMBUF]; | |
3135 | int len = snprintf(name, sizeof(name), "%d", tid); | |
3136 | return proc_fill_cache(filp, dirent, filldir, name, len, | |
3137 | proc_task_instantiate, task, NULL); | |
3138 | } | |
3139 | ||
1da177e4 LT |
3140 | /* for the /proc/TGID/task/ directories */ |
3141 | static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir) | |
3142 | { | |
2fddfeef | 3143 | struct dentry *dentry = filp->f_path.dentry; |
1da177e4 | 3144 | struct inode *inode = dentry->d_inode; |
7d895244 | 3145 | struct task_struct *leader = NULL; |
0bc58a91 | 3146 | struct task_struct *task; |
1da177e4 LT |
3147 | int retval = -ENOENT; |
3148 | ino_t ino; | |
0bc58a91 | 3149 | int tid; |
b488893a | 3150 | struct pid_namespace *ns; |
1da177e4 | 3151 | |
7d895244 GC |
3152 | task = get_proc_task(inode); |
3153 | if (!task) | |
3154 | goto out_no_task; | |
3155 | rcu_read_lock(); | |
3156 | if (pid_alive(task)) { | |
3157 | leader = task->group_leader; | |
3158 | get_task_struct(leader); | |
3159 | } | |
3160 | rcu_read_unlock(); | |
3161 | put_task_struct(task); | |
99f89551 EB |
3162 | if (!leader) |
3163 | goto out_no_task; | |
1da177e4 LT |
3164 | retval = 0; |
3165 | ||
ee568b25 | 3166 | switch ((unsigned long)filp->f_pos) { |
1da177e4 LT |
3167 | case 0: |
3168 | ino = inode->i_ino; | |
ee6f779b | 3169 | if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0) |
1da177e4 | 3170 | goto out; |
ee6f779b | 3171 | filp->f_pos++; |
1da177e4 LT |
3172 | /* fall through */ |
3173 | case 1: | |
3174 | ino = parent_ino(dentry); | |
ee6f779b | 3175 | if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0) |
1da177e4 | 3176 | goto out; |
ee6f779b | 3177 | filp->f_pos++; |
1da177e4 LT |
3178 | /* fall through */ |
3179 | } | |
3180 | ||
0bc58a91 EB |
3181 | /* f_version caches the tgid value that the last readdir call couldn't |
3182 | * return. lseek aka telldir automagically resets f_version to 0. | |
3183 | */ | |
b488893a | 3184 | ns = filp->f_dentry->d_sb->s_fs_info; |
2b47c361 | 3185 | tid = (int)filp->f_version; |
0bc58a91 | 3186 | filp->f_version = 0; |
ee6f779b | 3187 | for (task = first_tid(leader, tid, filp->f_pos - 2, ns); |
0bc58a91 | 3188 | task; |
ee6f779b | 3189 | task = next_tid(task), filp->f_pos++) { |
b488893a | 3190 | tid = task_pid_nr_ns(task, ns); |
61a28784 | 3191 | if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) { |
0bc58a91 EB |
3192 | /* returning this tgid failed, save it as the first |
3193 | * pid for the next readir call */ | |
2b47c361 | 3194 | filp->f_version = (u64)tid; |
0bc58a91 | 3195 | put_task_struct(task); |
1da177e4 | 3196 | break; |
0bc58a91 | 3197 | } |
1da177e4 LT |
3198 | } |
3199 | out: | |
99f89551 EB |
3200 | put_task_struct(leader); |
3201 | out_no_task: | |
1da177e4 LT |
3202 | return retval; |
3203 | } | |
6e66b52b EB |
3204 | |
3205 | static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat) | |
3206 | { | |
3207 | struct inode *inode = dentry->d_inode; | |
99f89551 | 3208 | struct task_struct *p = get_proc_task(inode); |
6e66b52b EB |
3209 | generic_fillattr(inode, stat); |
3210 | ||
99f89551 | 3211 | if (p) { |
99f89551 | 3212 | stat->nlink += get_nr_threads(p); |
99f89551 | 3213 | put_task_struct(p); |
6e66b52b EB |
3214 | } |
3215 | ||
3216 | return 0; | |
3217 | } | |
28a6d671 | 3218 | |
c5ef1c42 | 3219 | static const struct inode_operations proc_task_inode_operations = { |
28a6d671 EB |
3220 | .lookup = proc_task_lookup, |
3221 | .getattr = proc_task_getattr, | |
3222 | .setattr = proc_setattr, | |
3223 | }; | |
3224 | ||
00977a59 | 3225 | static const struct file_operations proc_task_operations = { |
28a6d671 EB |
3226 | .read = generic_read_dir, |
3227 | .readdir = proc_task_readdir, | |
3228 | }; |