]>
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 | ||
3dcd25f3 | 169 | static int proc_cwd_link(struct inode *inode, struct path *path) |
0494f6ec | 170 | { |
99f89551 | 171 | struct task_struct *task = get_proc_task(inode); |
0494f6ec | 172 | int result = -ENOENT; |
99f89551 EB |
173 | |
174 | if (task) { | |
7c2c7d99 | 175 | result = get_fs_path(task, path, 0); |
99f89551 EB |
176 | put_task_struct(task); |
177 | } | |
1da177e4 LT |
178 | return result; |
179 | } | |
180 | ||
3dcd25f3 | 181 | static int proc_root_link(struct inode *inode, struct path *path) |
1da177e4 | 182 | { |
99f89551 | 183 | struct task_struct *task = get_proc_task(inode); |
1da177e4 | 184 | int result = -ENOENT; |
99f89551 EB |
185 | |
186 | if (task) { | |
7c2c7d99 | 187 | result = get_fs_path(task, path, 1); |
99f89551 EB |
188 | put_task_struct(task); |
189 | } | |
1da177e4 LT |
190 | return result; |
191 | } | |
192 | ||
638fa202 RM |
193 | /* |
194 | * Return zero if current may access user memory in @task, -error if not. | |
195 | */ | |
196 | static int check_mem_permission(struct task_struct *task) | |
197 | { | |
198 | /* | |
199 | * A task can always look at itself, in case it chooses | |
200 | * to use system calls instead of load instructions. | |
201 | */ | |
202 | if (task == current) | |
203 | return 0; | |
204 | ||
205 | /* | |
206 | * If current is actively ptrace'ing, and would also be | |
207 | * permitted to freshly attach with ptrace now, permit it. | |
208 | */ | |
0d094efe RM |
209 | if (task_is_stopped_or_traced(task)) { |
210 | int match; | |
211 | rcu_read_lock(); | |
212 | match = (tracehook_tracer_task(task) == current); | |
213 | rcu_read_unlock(); | |
214 | if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH)) | |
215 | return 0; | |
216 | } | |
638fa202 RM |
217 | |
218 | /* | |
219 | * Noone else is allowed. | |
220 | */ | |
221 | return -EPERM; | |
222 | } | |
1da177e4 | 223 | |
831830b5 AV |
224 | struct mm_struct *mm_for_maps(struct task_struct *task) |
225 | { | |
704b836c ON |
226 | struct mm_struct *mm; |
227 | ||
228 | if (mutex_lock_killable(&task->cred_guard_mutex)) | |
229 | return NULL; | |
00f89d21 | 230 | |
704b836c ON |
231 | mm = get_task_mm(task); |
232 | if (mm && mm != current->mm && | |
233 | !ptrace_may_access(task, PTRACE_MODE_READ)) { | |
234 | mmput(mm); | |
235 | mm = NULL; | |
13f0feaf | 236 | } |
704b836c ON |
237 | mutex_unlock(&task->cred_guard_mutex); |
238 | ||
831830b5 | 239 | return mm; |
831830b5 AV |
240 | } |
241 | ||
1da177e4 LT |
242 | static int proc_pid_cmdline(struct task_struct *task, char * buffer) |
243 | { | |
244 | int res = 0; | |
245 | unsigned int len; | |
246 | struct mm_struct *mm = get_task_mm(task); | |
247 | if (!mm) | |
248 | goto out; | |
249 | if (!mm->arg_end) | |
250 | goto out_mm; /* Shh! No looking before we're done */ | |
251 | ||
252 | len = mm->arg_end - mm->arg_start; | |
253 | ||
254 | if (len > PAGE_SIZE) | |
255 | len = PAGE_SIZE; | |
256 | ||
257 | res = access_process_vm(task, mm->arg_start, buffer, len, 0); | |
258 | ||
259 | // If the nul at the end of args has been overwritten, then | |
260 | // assume application is using setproctitle(3). | |
261 | if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) { | |
262 | len = strnlen(buffer, res); | |
263 | if (len < res) { | |
264 | res = len; | |
265 | } else { | |
266 | len = mm->env_end - mm->env_start; | |
267 | if (len > PAGE_SIZE - res) | |
268 | len = PAGE_SIZE - res; | |
269 | res += access_process_vm(task, mm->env_start, buffer+res, len, 0); | |
270 | res = strnlen(buffer, res); | |
271 | } | |
272 | } | |
273 | out_mm: | |
274 | mmput(mm); | |
275 | out: | |
276 | return res; | |
277 | } | |
278 | ||
279 | static int proc_pid_auxv(struct task_struct *task, char *buffer) | |
280 | { | |
281 | int res = 0; | |
282 | struct mm_struct *mm = get_task_mm(task); | |
283 | if (mm) { | |
284 | unsigned int nwords = 0; | |
dfe6b7d9 | 285 | do { |
1da177e4 | 286 | nwords += 2; |
dfe6b7d9 | 287 | } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */ |
1da177e4 LT |
288 | res = nwords * sizeof(mm->saved_auxv[0]); |
289 | if (res > PAGE_SIZE) | |
290 | res = PAGE_SIZE; | |
291 | memcpy(buffer, mm->saved_auxv, res); | |
292 | mmput(mm); | |
293 | } | |
294 | return res; | |
295 | } | |
296 | ||
297 | ||
298 | #ifdef CONFIG_KALLSYMS | |
299 | /* | |
300 | * Provides a wchan file via kallsyms in a proper one-value-per-file format. | |
301 | * Returns the resolved symbol. If that fails, simply return the address. | |
302 | */ | |
303 | static int proc_pid_wchan(struct task_struct *task, char *buffer) | |
304 | { | |
ffb45122 | 305 | unsigned long wchan; |
9281acea | 306 | char symname[KSYM_NAME_LEN]; |
1da177e4 LT |
307 | |
308 | wchan = get_wchan(task); | |
309 | ||
9d65cb4a | 310 | if (lookup_symbol_name(wchan, symname) < 0) |
f83ce3e6 JE |
311 | if (!ptrace_may_access(task, PTRACE_MODE_READ)) |
312 | return 0; | |
313 | else | |
314 | return sprintf(buffer, "%lu", wchan); | |
9d65cb4a AD |
315 | else |
316 | return sprintf(buffer, "%s", symname); | |
1da177e4 LT |
317 | } |
318 | #endif /* CONFIG_KALLSYMS */ | |
319 | ||
2ec220e2 KC |
320 | #ifdef CONFIG_STACKTRACE |
321 | ||
322 | #define MAX_STACK_TRACE_DEPTH 64 | |
323 | ||
324 | static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns, | |
325 | struct pid *pid, struct task_struct *task) | |
326 | { | |
327 | struct stack_trace trace; | |
328 | unsigned long *entries; | |
329 | int i; | |
330 | ||
331 | entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL); | |
332 | if (!entries) | |
333 | return -ENOMEM; | |
334 | ||
335 | trace.nr_entries = 0; | |
336 | trace.max_entries = MAX_STACK_TRACE_DEPTH; | |
337 | trace.entries = entries; | |
338 | trace.skip = 0; | |
339 | save_stack_trace_tsk(task, &trace); | |
340 | ||
341 | for (i = 0; i < trace.nr_entries; i++) { | |
342 | seq_printf(m, "[<%p>] %pS\n", | |
343 | (void *)entries[i], (void *)entries[i]); | |
344 | } | |
345 | kfree(entries); | |
346 | ||
347 | return 0; | |
348 | } | |
349 | #endif | |
350 | ||
1da177e4 LT |
351 | #ifdef CONFIG_SCHEDSTATS |
352 | /* | |
353 | * Provides /proc/PID/schedstat | |
354 | */ | |
355 | static int proc_pid_schedstat(struct task_struct *task, char *buffer) | |
356 | { | |
172ba844 | 357 | return sprintf(buffer, "%llu %llu %lu\n", |
826e08b0 IM |
358 | (unsigned long long)task->se.sum_exec_runtime, |
359 | (unsigned long long)task->sched_info.run_delay, | |
2d72376b | 360 | task->sched_info.pcount); |
1da177e4 LT |
361 | } |
362 | #endif | |
363 | ||
9745512c AV |
364 | #ifdef CONFIG_LATENCYTOP |
365 | static int lstats_show_proc(struct seq_file *m, void *v) | |
366 | { | |
367 | int i; | |
13d77c37 HS |
368 | struct inode *inode = m->private; |
369 | struct task_struct *task = get_proc_task(inode); | |
9745512c | 370 | |
13d77c37 HS |
371 | if (!task) |
372 | return -ESRCH; | |
373 | seq_puts(m, "Latency Top version : v0.1\n"); | |
9745512c AV |
374 | for (i = 0; i < 32; i++) { |
375 | if (task->latency_record[i].backtrace[0]) { | |
376 | int q; | |
377 | seq_printf(m, "%i %li %li ", | |
378 | task->latency_record[i].count, | |
379 | task->latency_record[i].time, | |
380 | task->latency_record[i].max); | |
381 | for (q = 0; q < LT_BACKTRACEDEPTH; q++) { | |
9c246247 | 382 | char sym[KSYM_SYMBOL_LEN]; |
9745512c AV |
383 | char *c; |
384 | if (!task->latency_record[i].backtrace[q]) | |
385 | break; | |
386 | if (task->latency_record[i].backtrace[q] == ULONG_MAX) | |
387 | break; | |
388 | sprint_symbol(sym, task->latency_record[i].backtrace[q]); | |
389 | c = strchr(sym, '+'); | |
390 | if (c) | |
391 | *c = 0; | |
392 | seq_printf(m, "%s ", sym); | |
393 | } | |
394 | seq_printf(m, "\n"); | |
395 | } | |
396 | ||
397 | } | |
13d77c37 | 398 | put_task_struct(task); |
9745512c AV |
399 | return 0; |
400 | } | |
401 | ||
402 | static int lstats_open(struct inode *inode, struct file *file) | |
403 | { | |
13d77c37 | 404 | return single_open(file, lstats_show_proc, inode); |
d6643d12 HS |
405 | } |
406 | ||
9745512c AV |
407 | static ssize_t lstats_write(struct file *file, const char __user *buf, |
408 | size_t count, loff_t *offs) | |
409 | { | |
13d77c37 | 410 | struct task_struct *task = get_proc_task(file->f_dentry->d_inode); |
9745512c | 411 | |
13d77c37 HS |
412 | if (!task) |
413 | return -ESRCH; | |
9745512c | 414 | clear_all_latency_tracing(task); |
13d77c37 | 415 | put_task_struct(task); |
9745512c AV |
416 | |
417 | return count; | |
418 | } | |
419 | ||
420 | static const struct file_operations proc_lstats_operations = { | |
421 | .open = lstats_open, | |
422 | .read = seq_read, | |
423 | .write = lstats_write, | |
424 | .llseek = seq_lseek, | |
13d77c37 | 425 | .release = single_release, |
9745512c AV |
426 | }; |
427 | ||
428 | #endif | |
429 | ||
1da177e4 LT |
430 | /* The badness from the OOM killer */ |
431 | unsigned long badness(struct task_struct *p, unsigned long uptime); | |
432 | static int proc_oom_score(struct task_struct *task, char *buffer) | |
433 | { | |
b95c35e7 | 434 | unsigned long points = 0; |
1da177e4 LT |
435 | struct timespec uptime; |
436 | ||
437 | do_posix_clock_monotonic_gettime(&uptime); | |
19c5d45a | 438 | read_lock(&tasklist_lock); |
b95c35e7 ON |
439 | if (pid_alive(task)) |
440 | points = badness(task, uptime.tv_sec); | |
19c5d45a | 441 | read_unlock(&tasklist_lock); |
1da177e4 LT |
442 | return sprintf(buffer, "%lu\n", points); |
443 | } | |
444 | ||
d85f50d5 NH |
445 | struct limit_names { |
446 | char *name; | |
447 | char *unit; | |
448 | }; | |
449 | ||
450 | static const struct limit_names lnames[RLIM_NLIMITS] = { | |
cff4edb5 | 451 | [RLIMIT_CPU] = {"Max cpu time", "seconds"}, |
d85f50d5 NH |
452 | [RLIMIT_FSIZE] = {"Max file size", "bytes"}, |
453 | [RLIMIT_DATA] = {"Max data size", "bytes"}, | |
454 | [RLIMIT_STACK] = {"Max stack size", "bytes"}, | |
455 | [RLIMIT_CORE] = {"Max core file size", "bytes"}, | |
456 | [RLIMIT_RSS] = {"Max resident set", "bytes"}, | |
457 | [RLIMIT_NPROC] = {"Max processes", "processes"}, | |
458 | [RLIMIT_NOFILE] = {"Max open files", "files"}, | |
459 | [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"}, | |
460 | [RLIMIT_AS] = {"Max address space", "bytes"}, | |
461 | [RLIMIT_LOCKS] = {"Max file locks", "locks"}, | |
462 | [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"}, | |
463 | [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"}, | |
464 | [RLIMIT_NICE] = {"Max nice priority", NULL}, | |
465 | [RLIMIT_RTPRIO] = {"Max realtime priority", NULL}, | |
8808117c | 466 | [RLIMIT_RTTIME] = {"Max realtime timeout", "us"}, |
d85f50d5 NH |
467 | }; |
468 | ||
469 | /* Display limits for a process */ | |
470 | static int proc_pid_limits(struct task_struct *task, char *buffer) | |
471 | { | |
472 | unsigned int i; | |
473 | int count = 0; | |
474 | unsigned long flags; | |
475 | char *bufptr = buffer; | |
476 | ||
477 | struct rlimit rlim[RLIM_NLIMITS]; | |
478 | ||
a6bebbc8 | 479 | if (!lock_task_sighand(task, &flags)) |
d85f50d5 | 480 | return 0; |
d85f50d5 NH |
481 | memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS); |
482 | unlock_task_sighand(task, &flags); | |
d85f50d5 NH |
483 | |
484 | /* | |
485 | * print the file header | |
486 | */ | |
487 | count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n", | |
488 | "Limit", "Soft Limit", "Hard Limit", "Units"); | |
489 | ||
490 | for (i = 0; i < RLIM_NLIMITS; i++) { | |
491 | if (rlim[i].rlim_cur == RLIM_INFINITY) | |
492 | count += sprintf(&bufptr[count], "%-25s %-20s ", | |
493 | lnames[i].name, "unlimited"); | |
494 | else | |
495 | count += sprintf(&bufptr[count], "%-25s %-20lu ", | |
496 | lnames[i].name, rlim[i].rlim_cur); | |
497 | ||
498 | if (rlim[i].rlim_max == RLIM_INFINITY) | |
499 | count += sprintf(&bufptr[count], "%-20s ", "unlimited"); | |
500 | else | |
501 | count += sprintf(&bufptr[count], "%-20lu ", | |
502 | rlim[i].rlim_max); | |
503 | ||
504 | if (lnames[i].unit) | |
505 | count += sprintf(&bufptr[count], "%-10s\n", | |
506 | lnames[i].unit); | |
507 | else | |
508 | count += sprintf(&bufptr[count], "\n"); | |
509 | } | |
510 | ||
511 | return count; | |
512 | } | |
513 | ||
ebcb6734 RM |
514 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
515 | static int proc_pid_syscall(struct task_struct *task, char *buffer) | |
516 | { | |
517 | long nr; | |
518 | unsigned long args[6], sp, pc; | |
519 | ||
520 | if (task_current_syscall(task, &nr, args, 6, &sp, &pc)) | |
521 | return sprintf(buffer, "running\n"); | |
522 | ||
523 | if (nr < 0) | |
524 | return sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc); | |
525 | ||
526 | return sprintf(buffer, | |
527 | "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n", | |
528 | nr, | |
529 | args[0], args[1], args[2], args[3], args[4], args[5], | |
530 | sp, pc); | |
531 | } | |
532 | #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */ | |
533 | ||
1da177e4 LT |
534 | /************************************************************************/ |
535 | /* Here the fs part begins */ | |
536 | /************************************************************************/ | |
537 | ||
538 | /* permission checks */ | |
778c1144 | 539 | static int proc_fd_access_allowed(struct inode *inode) |
1da177e4 | 540 | { |
778c1144 EB |
541 | struct task_struct *task; |
542 | int allowed = 0; | |
df26c40e EB |
543 | /* Allow access to a task's file descriptors if it is us or we |
544 | * may use ptrace attach to the process and find out that | |
545 | * information. | |
778c1144 EB |
546 | */ |
547 | task = get_proc_task(inode); | |
df26c40e | 548 | if (task) { |
006ebb40 | 549 | allowed = ptrace_may_access(task, PTRACE_MODE_READ); |
778c1144 | 550 | put_task_struct(task); |
df26c40e | 551 | } |
778c1144 | 552 | return allowed; |
1da177e4 LT |
553 | } |
554 | ||
6d76fa58 LT |
555 | static int proc_setattr(struct dentry *dentry, struct iattr *attr) |
556 | { | |
557 | int error; | |
558 | struct inode *inode = dentry->d_inode; | |
559 | ||
560 | if (attr->ia_valid & ATTR_MODE) | |
561 | return -EPERM; | |
562 | ||
563 | error = inode_change_ok(inode, attr); | |
1e8123fd JJ |
564 | if (!error) |
565 | error = inode_setattr(inode, attr); | |
6d76fa58 LT |
566 | return error; |
567 | } | |
568 | ||
c5ef1c42 | 569 | static const struct inode_operations proc_def_inode_operations = { |
6d76fa58 LT |
570 | .setattr = proc_setattr, |
571 | }; | |
572 | ||
a1a2c409 MS |
573 | static int mounts_open_common(struct inode *inode, struct file *file, |
574 | const struct seq_operations *op) | |
1da177e4 | 575 | { |
99f89551 | 576 | struct task_struct *task = get_proc_task(inode); |
cf7b708c | 577 | struct nsproxy *nsp; |
6b3286ed | 578 | struct mnt_namespace *ns = NULL; |
a1a2c409 | 579 | struct path root; |
5addc5dd AV |
580 | struct proc_mounts *p; |
581 | int ret = -EINVAL; | |
1da177e4 | 582 | |
99f89551 | 583 | if (task) { |
cf7b708c PE |
584 | rcu_read_lock(); |
585 | nsp = task_nsproxy(task); | |
586 | if (nsp) { | |
587 | ns = nsp->mnt_ns; | |
863c4702 AD |
588 | if (ns) |
589 | get_mnt_ns(ns); | |
590 | } | |
cf7b708c | 591 | rcu_read_unlock(); |
7c2c7d99 HD |
592 | if (ns && get_fs_path(task, &root, 1) == 0) |
593 | ret = 0; | |
99f89551 EB |
594 | put_task_struct(task); |
595 | } | |
5addc5dd | 596 | |
a1a2c409 MS |
597 | if (!ns) |
598 | goto err; | |
7c2c7d99 | 599 | if (ret) |
a1a2c409 MS |
600 | goto err_put_ns; |
601 | ||
a1a2c409 MS |
602 | ret = -ENOMEM; |
603 | p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL); | |
604 | if (!p) | |
605 | goto err_put_path; | |
606 | ||
607 | file->private_data = &p->m; | |
608 | ret = seq_open(file, op); | |
609 | if (ret) | |
610 | goto err_free; | |
611 | ||
612 | p->m.private = p; | |
613 | p->ns = ns; | |
614 | p->root = root; | |
615 | p->event = ns->event; | |
616 | ||
617 | return 0; | |
618 | ||
619 | err_free: | |
620 | kfree(p); | |
621 | err_put_path: | |
622 | path_put(&root); | |
623 | err_put_ns: | |
624 | put_mnt_ns(ns); | |
625 | err: | |
1da177e4 LT |
626 | return ret; |
627 | } | |
628 | ||
629 | static int mounts_release(struct inode *inode, struct file *file) | |
630 | { | |
a1a2c409 MS |
631 | struct proc_mounts *p = file->private_data; |
632 | path_put(&p->root); | |
633 | put_mnt_ns(p->ns); | |
1da177e4 LT |
634 | return seq_release(inode, file); |
635 | } | |
636 | ||
5addc5dd AV |
637 | static unsigned mounts_poll(struct file *file, poll_table *wait) |
638 | { | |
639 | struct proc_mounts *p = file->private_data; | |
31b07093 | 640 | unsigned res = POLLIN | POLLRDNORM; |
5addc5dd | 641 | |
9f5596af AV |
642 | poll_wait(file, &p->ns->poll, wait); |
643 | if (mnt_had_events(p)) | |
31b07093 | 644 | res |= POLLERR | POLLPRI; |
5addc5dd AV |
645 | |
646 | return res; | |
647 | } | |
648 | ||
a1a2c409 MS |
649 | static int mounts_open(struct inode *inode, struct file *file) |
650 | { | |
651 | return mounts_open_common(inode, file, &mounts_op); | |
652 | } | |
653 | ||
00977a59 | 654 | static const struct file_operations proc_mounts_operations = { |
1da177e4 LT |
655 | .open = mounts_open, |
656 | .read = seq_read, | |
657 | .llseek = seq_lseek, | |
658 | .release = mounts_release, | |
5addc5dd | 659 | .poll = mounts_poll, |
1da177e4 LT |
660 | }; |
661 | ||
2d4d4864 RP |
662 | static int mountinfo_open(struct inode *inode, struct file *file) |
663 | { | |
664 | return mounts_open_common(inode, file, &mountinfo_op); | |
665 | } | |
666 | ||
667 | static const struct file_operations proc_mountinfo_operations = { | |
668 | .open = mountinfo_open, | |
669 | .read = seq_read, | |
670 | .llseek = seq_lseek, | |
671 | .release = mounts_release, | |
672 | .poll = mounts_poll, | |
673 | }; | |
674 | ||
b4629fe2 CL |
675 | static int mountstats_open(struct inode *inode, struct file *file) |
676 | { | |
a1a2c409 | 677 | return mounts_open_common(inode, file, &mountstats_op); |
b4629fe2 CL |
678 | } |
679 | ||
00977a59 | 680 | static const struct file_operations proc_mountstats_operations = { |
b4629fe2 CL |
681 | .open = mountstats_open, |
682 | .read = seq_read, | |
683 | .llseek = seq_lseek, | |
684 | .release = mounts_release, | |
685 | }; | |
686 | ||
1da177e4 LT |
687 | #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */ |
688 | ||
689 | static ssize_t proc_info_read(struct file * file, char __user * buf, | |
690 | size_t count, loff_t *ppos) | |
691 | { | |
2fddfeef | 692 | struct inode * inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
693 | unsigned long page; |
694 | ssize_t length; | |
99f89551 EB |
695 | struct task_struct *task = get_proc_task(inode); |
696 | ||
697 | length = -ESRCH; | |
698 | if (!task) | |
699 | goto out_no_task; | |
1da177e4 LT |
700 | |
701 | if (count > PROC_BLOCK_SIZE) | |
702 | count = PROC_BLOCK_SIZE; | |
99f89551 EB |
703 | |
704 | length = -ENOMEM; | |
e12ba74d | 705 | if (!(page = __get_free_page(GFP_TEMPORARY))) |
99f89551 | 706 | goto out; |
1da177e4 LT |
707 | |
708 | length = PROC_I(inode)->op.proc_read(task, (char*)page); | |
709 | ||
710 | if (length >= 0) | |
711 | length = simple_read_from_buffer(buf, count, ppos, (char *)page, length); | |
712 | free_page(page); | |
99f89551 EB |
713 | out: |
714 | put_task_struct(task); | |
715 | out_no_task: | |
1da177e4 LT |
716 | return length; |
717 | } | |
718 | ||
00977a59 | 719 | static const struct file_operations proc_info_file_operations = { |
1da177e4 | 720 | .read = proc_info_read, |
87df8424 | 721 | .llseek = generic_file_llseek, |
1da177e4 LT |
722 | }; |
723 | ||
be614086 EB |
724 | static int proc_single_show(struct seq_file *m, void *v) |
725 | { | |
726 | struct inode *inode = m->private; | |
727 | struct pid_namespace *ns; | |
728 | struct pid *pid; | |
729 | struct task_struct *task; | |
730 | int ret; | |
731 | ||
732 | ns = inode->i_sb->s_fs_info; | |
733 | pid = proc_pid(inode); | |
734 | task = get_pid_task(pid, PIDTYPE_PID); | |
735 | if (!task) | |
736 | return -ESRCH; | |
737 | ||
738 | ret = PROC_I(inode)->op.proc_show(m, ns, pid, task); | |
739 | ||
740 | put_task_struct(task); | |
741 | return ret; | |
742 | } | |
743 | ||
744 | static int proc_single_open(struct inode *inode, struct file *filp) | |
745 | { | |
746 | int ret; | |
747 | ret = single_open(filp, proc_single_show, NULL); | |
748 | if (!ret) { | |
749 | struct seq_file *m = filp->private_data; | |
750 | ||
751 | m->private = inode; | |
752 | } | |
753 | return ret; | |
754 | } | |
755 | ||
756 | static const struct file_operations proc_single_file_operations = { | |
757 | .open = proc_single_open, | |
758 | .read = seq_read, | |
759 | .llseek = seq_lseek, | |
760 | .release = single_release, | |
761 | }; | |
762 | ||
1da177e4 LT |
763 | static int mem_open(struct inode* inode, struct file* file) |
764 | { | |
765 | file->private_data = (void*)((long)current->self_exec_id); | |
766 | return 0; | |
767 | } | |
768 | ||
769 | static ssize_t mem_read(struct file * file, char __user * buf, | |
770 | size_t count, loff_t *ppos) | |
771 | { | |
2fddfeef | 772 | struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode); |
1da177e4 LT |
773 | char *page; |
774 | unsigned long src = *ppos; | |
775 | int ret = -ESRCH; | |
776 | struct mm_struct *mm; | |
777 | ||
99f89551 EB |
778 | if (!task) |
779 | goto out_no_task; | |
780 | ||
638fa202 | 781 | if (check_mem_permission(task)) |
1da177e4 LT |
782 | goto out; |
783 | ||
784 | ret = -ENOMEM; | |
e12ba74d | 785 | page = (char *)__get_free_page(GFP_TEMPORARY); |
1da177e4 LT |
786 | if (!page) |
787 | goto out; | |
788 | ||
789 | ret = 0; | |
790 | ||
791 | mm = get_task_mm(task); | |
792 | if (!mm) | |
793 | goto out_free; | |
794 | ||
795 | ret = -EIO; | |
796 | ||
797 | if (file->private_data != (void*)((long)current->self_exec_id)) | |
798 | goto out_put; | |
799 | ||
800 | ret = 0; | |
801 | ||
802 | while (count > 0) { | |
803 | int this_len, retval; | |
804 | ||
805 | this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count; | |
806 | retval = access_process_vm(task, src, page, this_len, 0); | |
638fa202 | 807 | if (!retval || check_mem_permission(task)) { |
1da177e4 LT |
808 | if (!ret) |
809 | ret = -EIO; | |
810 | break; | |
811 | } | |
812 | ||
813 | if (copy_to_user(buf, page, retval)) { | |
814 | ret = -EFAULT; | |
815 | break; | |
816 | } | |
817 | ||
818 | ret += retval; | |
819 | src += retval; | |
820 | buf += retval; | |
821 | count -= retval; | |
822 | } | |
823 | *ppos = src; | |
824 | ||
825 | out_put: | |
826 | mmput(mm); | |
827 | out_free: | |
828 | free_page((unsigned long) page); | |
829 | out: | |
99f89551 EB |
830 | put_task_struct(task); |
831 | out_no_task: | |
1da177e4 LT |
832 | return ret; |
833 | } | |
834 | ||
835 | #define mem_write NULL | |
836 | ||
837 | #ifndef mem_write | |
838 | /* This is a security hazard */ | |
63967fa9 | 839 | static ssize_t mem_write(struct file * file, const char __user *buf, |
1da177e4 LT |
840 | size_t count, loff_t *ppos) |
841 | { | |
f7ca54f4 | 842 | int copied; |
1da177e4 | 843 | char *page; |
2fddfeef | 844 | struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode); |
1da177e4 LT |
845 | unsigned long dst = *ppos; |
846 | ||
99f89551 EB |
847 | copied = -ESRCH; |
848 | if (!task) | |
849 | goto out_no_task; | |
850 | ||
638fa202 | 851 | if (check_mem_permission(task)) |
99f89551 | 852 | goto out; |
1da177e4 | 853 | |
99f89551 | 854 | copied = -ENOMEM; |
e12ba74d | 855 | page = (char *)__get_free_page(GFP_TEMPORARY); |
1da177e4 | 856 | if (!page) |
99f89551 | 857 | goto out; |
1da177e4 | 858 | |
f7ca54f4 | 859 | copied = 0; |
1da177e4 LT |
860 | while (count > 0) { |
861 | int this_len, retval; | |
862 | ||
863 | this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count; | |
864 | if (copy_from_user(page, buf, this_len)) { | |
865 | copied = -EFAULT; | |
866 | break; | |
867 | } | |
868 | retval = access_process_vm(task, dst, page, this_len, 1); | |
869 | if (!retval) { | |
870 | if (!copied) | |
871 | copied = -EIO; | |
872 | break; | |
873 | } | |
874 | copied += retval; | |
875 | buf += retval; | |
876 | dst += retval; | |
877 | count -= retval; | |
878 | } | |
879 | *ppos = dst; | |
880 | free_page((unsigned long) page); | |
99f89551 EB |
881 | out: |
882 | put_task_struct(task); | |
883 | out_no_task: | |
1da177e4 LT |
884 | return copied; |
885 | } | |
886 | #endif | |
887 | ||
85863e47 | 888 | loff_t mem_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
889 | { |
890 | switch (orig) { | |
891 | case 0: | |
892 | file->f_pos = offset; | |
893 | break; | |
894 | case 1: | |
895 | file->f_pos += offset; | |
896 | break; | |
897 | default: | |
898 | return -EINVAL; | |
899 | } | |
900 | force_successful_syscall_return(); | |
901 | return file->f_pos; | |
902 | } | |
903 | ||
00977a59 | 904 | static const struct file_operations proc_mem_operations = { |
1da177e4 LT |
905 | .llseek = mem_lseek, |
906 | .read = mem_read, | |
907 | .write = mem_write, | |
908 | .open = mem_open, | |
909 | }; | |
910 | ||
315e28c8 JP |
911 | static ssize_t environ_read(struct file *file, char __user *buf, |
912 | size_t count, loff_t *ppos) | |
913 | { | |
914 | struct task_struct *task = get_proc_task(file->f_dentry->d_inode); | |
915 | char *page; | |
916 | unsigned long src = *ppos; | |
917 | int ret = -ESRCH; | |
918 | struct mm_struct *mm; | |
919 | ||
920 | if (!task) | |
921 | goto out_no_task; | |
922 | ||
006ebb40 | 923 | if (!ptrace_may_access(task, PTRACE_MODE_READ)) |
315e28c8 JP |
924 | goto out; |
925 | ||
926 | ret = -ENOMEM; | |
927 | page = (char *)__get_free_page(GFP_TEMPORARY); | |
928 | if (!page) | |
929 | goto out; | |
930 | ||
931 | ret = 0; | |
932 | ||
933 | mm = get_task_mm(task); | |
934 | if (!mm) | |
935 | goto out_free; | |
936 | ||
937 | while (count > 0) { | |
938 | int this_len, retval, max_len; | |
939 | ||
940 | this_len = mm->env_end - (mm->env_start + src); | |
941 | ||
942 | if (this_len <= 0) | |
943 | break; | |
944 | ||
945 | max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count; | |
946 | this_len = (this_len > max_len) ? max_len : this_len; | |
947 | ||
948 | retval = access_process_vm(task, (mm->env_start + src), | |
949 | page, this_len, 0); | |
950 | ||
951 | if (retval <= 0) { | |
952 | ret = retval; | |
953 | break; | |
954 | } | |
955 | ||
956 | if (copy_to_user(buf, page, retval)) { | |
957 | ret = -EFAULT; | |
958 | break; | |
959 | } | |
960 | ||
961 | ret += retval; | |
962 | src += retval; | |
963 | buf += retval; | |
964 | count -= retval; | |
965 | } | |
966 | *ppos = src; | |
967 | ||
968 | mmput(mm); | |
969 | out_free: | |
970 | free_page((unsigned long) page); | |
971 | out: | |
972 | put_task_struct(task); | |
973 | out_no_task: | |
974 | return ret; | |
975 | } | |
976 | ||
977 | static const struct file_operations proc_environ_operations = { | |
978 | .read = environ_read, | |
87df8424 | 979 | .llseek = generic_file_llseek, |
315e28c8 JP |
980 | }; |
981 | ||
1da177e4 LT |
982 | static ssize_t oom_adjust_read(struct file *file, char __user *buf, |
983 | size_t count, loff_t *ppos) | |
984 | { | |
2fddfeef | 985 | struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode); |
8578cea7 | 986 | char buffer[PROC_NUMBUF]; |
1da177e4 | 987 | size_t len; |
28b83c51 KM |
988 | int oom_adjust = OOM_DISABLE; |
989 | unsigned long flags; | |
1da177e4 | 990 | |
99f89551 EB |
991 | if (!task) |
992 | return -ESRCH; | |
28b83c51 KM |
993 | |
994 | if (lock_task_sighand(task, &flags)) { | |
995 | oom_adjust = task->signal->oom_adj; | |
996 | unlock_task_sighand(task, &flags); | |
997 | } | |
998 | ||
99f89551 EB |
999 | put_task_struct(task); |
1000 | ||
8578cea7 | 1001 | len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust); |
0c28f287 AM |
1002 | |
1003 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
1da177e4 LT |
1004 | } |
1005 | ||
1006 | static ssize_t oom_adjust_write(struct file *file, const char __user *buf, | |
1007 | size_t count, loff_t *ppos) | |
1008 | { | |
99f89551 | 1009 | struct task_struct *task; |
5d863b89 KM |
1010 | char buffer[PROC_NUMBUF]; |
1011 | long oom_adjust; | |
28b83c51 | 1012 | unsigned long flags; |
5d863b89 | 1013 | int err; |
1da177e4 | 1014 | |
8578cea7 EB |
1015 | memset(buffer, 0, sizeof(buffer)); |
1016 | if (count > sizeof(buffer) - 1) | |
1017 | count = sizeof(buffer) - 1; | |
1da177e4 LT |
1018 | if (copy_from_user(buffer, buf, count)) |
1019 | return -EFAULT; | |
5d863b89 KM |
1020 | |
1021 | err = strict_strtol(strstrip(buffer), 0, &oom_adjust); | |
1022 | if (err) | |
1023 | return -EINVAL; | |
8ac773b4 AD |
1024 | if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) && |
1025 | oom_adjust != OOM_DISABLE) | |
1da177e4 | 1026 | return -EINVAL; |
5d863b89 | 1027 | |
2fddfeef | 1028 | task = get_proc_task(file->f_path.dentry->d_inode); |
99f89551 EB |
1029 | if (!task) |
1030 | return -ESRCH; | |
28b83c51 KM |
1031 | if (!lock_task_sighand(task, &flags)) { |
1032 | put_task_struct(task); | |
1033 | return -ESRCH; | |
1034 | } | |
1035 | ||
1036 | if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) { | |
1037 | unlock_task_sighand(task, &flags); | |
8fb4fc68 GJ |
1038 | put_task_struct(task); |
1039 | return -EACCES; | |
1040 | } | |
28b83c51 KM |
1041 | |
1042 | task->signal->oom_adj = oom_adjust; | |
1043 | ||
1044 | unlock_task_sighand(task, &flags); | |
99f89551 | 1045 | put_task_struct(task); |
5d863b89 KM |
1046 | |
1047 | return count; | |
1da177e4 LT |
1048 | } |
1049 | ||
00977a59 | 1050 | static const struct file_operations proc_oom_adjust_operations = { |
1da177e4 LT |
1051 | .read = oom_adjust_read, |
1052 | .write = oom_adjust_write, | |
87df8424 | 1053 | .llseek = generic_file_llseek, |
1da177e4 LT |
1054 | }; |
1055 | ||
1da177e4 LT |
1056 | #ifdef CONFIG_AUDITSYSCALL |
1057 | #define TMPBUFLEN 21 | |
1058 | static ssize_t proc_loginuid_read(struct file * file, char __user * buf, | |
1059 | size_t count, loff_t *ppos) | |
1060 | { | |
2fddfeef | 1061 | struct inode * inode = file->f_path.dentry->d_inode; |
99f89551 | 1062 | struct task_struct *task = get_proc_task(inode); |
1da177e4 LT |
1063 | ssize_t length; |
1064 | char tmpbuf[TMPBUFLEN]; | |
1065 | ||
99f89551 EB |
1066 | if (!task) |
1067 | return -ESRCH; | |
1da177e4 | 1068 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", |
0c11b942 | 1069 | audit_get_loginuid(task)); |
99f89551 | 1070 | put_task_struct(task); |
1da177e4 LT |
1071 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); |
1072 | } | |
1073 | ||
1074 | static ssize_t proc_loginuid_write(struct file * file, const char __user * buf, | |
1075 | size_t count, loff_t *ppos) | |
1076 | { | |
2fddfeef | 1077 | struct inode * inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
1078 | char *page, *tmp; |
1079 | ssize_t length; | |
1da177e4 LT |
1080 | uid_t loginuid; |
1081 | ||
1082 | if (!capable(CAP_AUDIT_CONTROL)) | |
1083 | return -EPERM; | |
1084 | ||
7dc52157 PM |
1085 | rcu_read_lock(); |
1086 | if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) { | |
1087 | rcu_read_unlock(); | |
1da177e4 | 1088 | return -EPERM; |
7dc52157 PM |
1089 | } |
1090 | rcu_read_unlock(); | |
1da177e4 | 1091 | |
e0182909 AV |
1092 | if (count >= PAGE_SIZE) |
1093 | count = PAGE_SIZE - 1; | |
1da177e4 LT |
1094 | |
1095 | if (*ppos != 0) { | |
1096 | /* No partial writes. */ | |
1097 | return -EINVAL; | |
1098 | } | |
e12ba74d | 1099 | page = (char*)__get_free_page(GFP_TEMPORARY); |
1da177e4 LT |
1100 | if (!page) |
1101 | return -ENOMEM; | |
1102 | length = -EFAULT; | |
1103 | if (copy_from_user(page, buf, count)) | |
1104 | goto out_free_page; | |
1105 | ||
e0182909 | 1106 | page[count] = '\0'; |
1da177e4 LT |
1107 | loginuid = simple_strtoul(page, &tmp, 10); |
1108 | if (tmp == page) { | |
1109 | length = -EINVAL; | |
1110 | goto out_free_page; | |
1111 | ||
1112 | } | |
99f89551 | 1113 | length = audit_set_loginuid(current, loginuid); |
1da177e4 LT |
1114 | if (likely(length == 0)) |
1115 | length = count; | |
1116 | ||
1117 | out_free_page: | |
1118 | free_page((unsigned long) page); | |
1119 | return length; | |
1120 | } | |
1121 | ||
00977a59 | 1122 | static const struct file_operations proc_loginuid_operations = { |
1da177e4 LT |
1123 | .read = proc_loginuid_read, |
1124 | .write = proc_loginuid_write, | |
87df8424 | 1125 | .llseek = generic_file_llseek, |
1da177e4 | 1126 | }; |
1e0bd755 EP |
1127 | |
1128 | static ssize_t proc_sessionid_read(struct file * file, char __user * buf, | |
1129 | size_t count, loff_t *ppos) | |
1130 | { | |
1131 | struct inode * inode = file->f_path.dentry->d_inode; | |
1132 | struct task_struct *task = get_proc_task(inode); | |
1133 | ssize_t length; | |
1134 | char tmpbuf[TMPBUFLEN]; | |
1135 | ||
1136 | if (!task) | |
1137 | return -ESRCH; | |
1138 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", | |
1139 | audit_get_sessionid(task)); | |
1140 | put_task_struct(task); | |
1141 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); | |
1142 | } | |
1143 | ||
1144 | static const struct file_operations proc_sessionid_operations = { | |
1145 | .read = proc_sessionid_read, | |
87df8424 | 1146 | .llseek = generic_file_llseek, |
1e0bd755 | 1147 | }; |
1da177e4 LT |
1148 | #endif |
1149 | ||
f4f154fd AM |
1150 | #ifdef CONFIG_FAULT_INJECTION |
1151 | static ssize_t proc_fault_inject_read(struct file * file, char __user * buf, | |
1152 | size_t count, loff_t *ppos) | |
1153 | { | |
1154 | struct task_struct *task = get_proc_task(file->f_dentry->d_inode); | |
1155 | char buffer[PROC_NUMBUF]; | |
1156 | size_t len; | |
1157 | int make_it_fail; | |
f4f154fd AM |
1158 | |
1159 | if (!task) | |
1160 | return -ESRCH; | |
1161 | make_it_fail = task->make_it_fail; | |
1162 | put_task_struct(task); | |
1163 | ||
1164 | len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail); | |
0c28f287 AM |
1165 | |
1166 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
f4f154fd AM |
1167 | } |
1168 | ||
1169 | static ssize_t proc_fault_inject_write(struct file * file, | |
1170 | const char __user * buf, size_t count, loff_t *ppos) | |
1171 | { | |
1172 | struct task_struct *task; | |
1173 | char buffer[PROC_NUMBUF], *end; | |
1174 | int make_it_fail; | |
1175 | ||
1176 | if (!capable(CAP_SYS_RESOURCE)) | |
1177 | return -EPERM; | |
1178 | memset(buffer, 0, sizeof(buffer)); | |
1179 | if (count > sizeof(buffer) - 1) | |
1180 | count = sizeof(buffer) - 1; | |
1181 | if (copy_from_user(buffer, buf, count)) | |
1182 | return -EFAULT; | |
cba8aafe VL |
1183 | make_it_fail = simple_strtol(strstrip(buffer), &end, 0); |
1184 | if (*end) | |
1185 | return -EINVAL; | |
f4f154fd AM |
1186 | task = get_proc_task(file->f_dentry->d_inode); |
1187 | if (!task) | |
1188 | return -ESRCH; | |
1189 | task->make_it_fail = make_it_fail; | |
1190 | put_task_struct(task); | |
cba8aafe VL |
1191 | |
1192 | return count; | |
f4f154fd AM |
1193 | } |
1194 | ||
00977a59 | 1195 | static const struct file_operations proc_fault_inject_operations = { |
f4f154fd AM |
1196 | .read = proc_fault_inject_read, |
1197 | .write = proc_fault_inject_write, | |
87df8424 | 1198 | .llseek = generic_file_llseek, |
f4f154fd AM |
1199 | }; |
1200 | #endif | |
1201 | ||
9745512c | 1202 | |
43ae34cb IM |
1203 | #ifdef CONFIG_SCHED_DEBUG |
1204 | /* | |
1205 | * Print out various scheduling related per-task fields: | |
1206 | */ | |
1207 | static int sched_show(struct seq_file *m, void *v) | |
1208 | { | |
1209 | struct inode *inode = m->private; | |
1210 | struct task_struct *p; | |
1211 | ||
43ae34cb IM |
1212 | p = get_proc_task(inode); |
1213 | if (!p) | |
1214 | return -ESRCH; | |
1215 | proc_sched_show_task(p, m); | |
1216 | ||
1217 | put_task_struct(p); | |
1218 | ||
1219 | return 0; | |
1220 | } | |
1221 | ||
1222 | static ssize_t | |
1223 | sched_write(struct file *file, const char __user *buf, | |
1224 | size_t count, loff_t *offset) | |
1225 | { | |
1226 | struct inode *inode = file->f_path.dentry->d_inode; | |
1227 | struct task_struct *p; | |
1228 | ||
43ae34cb IM |
1229 | p = get_proc_task(inode); |
1230 | if (!p) | |
1231 | return -ESRCH; | |
1232 | proc_sched_set_task(p); | |
1233 | ||
1234 | put_task_struct(p); | |
1235 | ||
1236 | return count; | |
1237 | } | |
1238 | ||
1239 | static int sched_open(struct inode *inode, struct file *filp) | |
1240 | { | |
1241 | int ret; | |
1242 | ||
1243 | ret = single_open(filp, sched_show, NULL); | |
1244 | if (!ret) { | |
1245 | struct seq_file *m = filp->private_data; | |
1246 | ||
1247 | m->private = inode; | |
1248 | } | |
1249 | return ret; | |
1250 | } | |
1251 | ||
1252 | static const struct file_operations proc_pid_sched_operations = { | |
1253 | .open = sched_open, | |
1254 | .read = seq_read, | |
1255 | .write = sched_write, | |
1256 | .llseek = seq_lseek, | |
5ea473a1 | 1257 | .release = single_release, |
43ae34cb IM |
1258 | }; |
1259 | ||
1260 | #endif | |
1261 | ||
4614a696 JS |
1262 | static ssize_t comm_write(struct file *file, const char __user *buf, |
1263 | size_t count, loff_t *offset) | |
1264 | { | |
1265 | struct inode *inode = file->f_path.dentry->d_inode; | |
1266 | struct task_struct *p; | |
1267 | char buffer[TASK_COMM_LEN]; | |
1268 | ||
1269 | memset(buffer, 0, sizeof(buffer)); | |
1270 | if (count > sizeof(buffer) - 1) | |
1271 | count = sizeof(buffer) - 1; | |
1272 | if (copy_from_user(buffer, buf, count)) | |
1273 | return -EFAULT; | |
1274 | ||
1275 | p = get_proc_task(inode); | |
1276 | if (!p) | |
1277 | return -ESRCH; | |
1278 | ||
1279 | if (same_thread_group(current, p)) | |
1280 | set_task_comm(p, buffer); | |
1281 | else | |
1282 | count = -EINVAL; | |
1283 | ||
1284 | put_task_struct(p); | |
1285 | ||
1286 | return count; | |
1287 | } | |
1288 | ||
1289 | static int comm_show(struct seq_file *m, void *v) | |
1290 | { | |
1291 | struct inode *inode = m->private; | |
1292 | struct task_struct *p; | |
1293 | ||
1294 | p = get_proc_task(inode); | |
1295 | if (!p) | |
1296 | return -ESRCH; | |
1297 | ||
1298 | task_lock(p); | |
1299 | seq_printf(m, "%s\n", p->comm); | |
1300 | task_unlock(p); | |
1301 | ||
1302 | put_task_struct(p); | |
1303 | ||
1304 | return 0; | |
1305 | } | |
1306 | ||
1307 | static int comm_open(struct inode *inode, struct file *filp) | |
1308 | { | |
1309 | int ret; | |
1310 | ||
1311 | ret = single_open(filp, comm_show, NULL); | |
1312 | if (!ret) { | |
1313 | struct seq_file *m = filp->private_data; | |
1314 | ||
1315 | m->private = inode; | |
1316 | } | |
1317 | return ret; | |
1318 | } | |
1319 | ||
1320 | static const struct file_operations proc_pid_set_comm_operations = { | |
1321 | .open = comm_open, | |
1322 | .read = seq_read, | |
1323 | .write = comm_write, | |
1324 | .llseek = seq_lseek, | |
1325 | .release = single_release, | |
1326 | }; | |
1327 | ||
925d1c40 MH |
1328 | /* |
1329 | * We added or removed a vma mapping the executable. The vmas are only mapped | |
1330 | * during exec and are not mapped with the mmap system call. | |
1331 | * Callers must hold down_write() on the mm's mmap_sem for these | |
1332 | */ | |
1333 | void added_exe_file_vma(struct mm_struct *mm) | |
1334 | { | |
1335 | mm->num_exe_file_vmas++; | |
1336 | } | |
1337 | ||
1338 | void removed_exe_file_vma(struct mm_struct *mm) | |
1339 | { | |
1340 | mm->num_exe_file_vmas--; | |
1341 | if ((mm->num_exe_file_vmas == 0) && mm->exe_file){ | |
1342 | fput(mm->exe_file); | |
1343 | mm->exe_file = NULL; | |
1344 | } | |
1345 | ||
1346 | } | |
1347 | ||
1348 | void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) | |
1349 | { | |
1350 | if (new_exe_file) | |
1351 | get_file(new_exe_file); | |
1352 | if (mm->exe_file) | |
1353 | fput(mm->exe_file); | |
1354 | mm->exe_file = new_exe_file; | |
1355 | mm->num_exe_file_vmas = 0; | |
1356 | } | |
1357 | ||
1358 | struct file *get_mm_exe_file(struct mm_struct *mm) | |
1359 | { | |
1360 | struct file *exe_file; | |
1361 | ||
1362 | /* We need mmap_sem to protect against races with removal of | |
1363 | * VM_EXECUTABLE vmas */ | |
1364 | down_read(&mm->mmap_sem); | |
1365 | exe_file = mm->exe_file; | |
1366 | if (exe_file) | |
1367 | get_file(exe_file); | |
1368 | up_read(&mm->mmap_sem); | |
1369 | return exe_file; | |
1370 | } | |
1371 | ||
1372 | void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm) | |
1373 | { | |
1374 | /* It's safe to write the exe_file pointer without exe_file_lock because | |
1375 | * this is called during fork when the task is not yet in /proc */ | |
1376 | newmm->exe_file = get_mm_exe_file(oldmm); | |
1377 | } | |
1378 | ||
1379 | static int proc_exe_link(struct inode *inode, struct path *exe_path) | |
1380 | { | |
1381 | struct task_struct *task; | |
1382 | struct mm_struct *mm; | |
1383 | struct file *exe_file; | |
1384 | ||
1385 | task = get_proc_task(inode); | |
1386 | if (!task) | |
1387 | return -ENOENT; | |
1388 | mm = get_task_mm(task); | |
1389 | put_task_struct(task); | |
1390 | if (!mm) | |
1391 | return -ENOENT; | |
1392 | exe_file = get_mm_exe_file(mm); | |
1393 | mmput(mm); | |
1394 | if (exe_file) { | |
1395 | *exe_path = exe_file->f_path; | |
1396 | path_get(&exe_file->f_path); | |
1397 | fput(exe_file); | |
1398 | return 0; | |
1399 | } else | |
1400 | return -ENOENT; | |
1401 | } | |
1402 | ||
008b150a | 1403 | static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd) |
1da177e4 LT |
1404 | { |
1405 | struct inode *inode = dentry->d_inode; | |
1406 | int error = -EACCES; | |
1407 | ||
1408 | /* We don't need a base pointer in the /proc filesystem */ | |
1d957f9b | 1409 | path_put(&nd->path); |
1da177e4 | 1410 | |
778c1144 EB |
1411 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1412 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1413 | goto out; |
1da177e4 | 1414 | |
3dcd25f3 | 1415 | error = PROC_I(inode)->op.proc_get_link(inode, &nd->path); |
1da177e4 | 1416 | out: |
008b150a | 1417 | return ERR_PTR(error); |
1da177e4 LT |
1418 | } |
1419 | ||
3dcd25f3 | 1420 | static int do_proc_readlink(struct path *path, char __user *buffer, int buflen) |
1da177e4 | 1421 | { |
e12ba74d | 1422 | char *tmp = (char*)__get_free_page(GFP_TEMPORARY); |
3dcd25f3 | 1423 | char *pathname; |
1da177e4 LT |
1424 | int len; |
1425 | ||
1426 | if (!tmp) | |
1427 | return -ENOMEM; | |
0c28f287 | 1428 | |
cf28b486 | 1429 | pathname = d_path(path, tmp, PAGE_SIZE); |
3dcd25f3 JB |
1430 | len = PTR_ERR(pathname); |
1431 | if (IS_ERR(pathname)) | |
1da177e4 | 1432 | goto out; |
3dcd25f3 | 1433 | len = tmp + PAGE_SIZE - 1 - pathname; |
1da177e4 LT |
1434 | |
1435 | if (len > buflen) | |
1436 | len = buflen; | |
3dcd25f3 | 1437 | if (copy_to_user(buffer, pathname, len)) |
1da177e4 LT |
1438 | len = -EFAULT; |
1439 | out: | |
1440 | free_page((unsigned long)tmp); | |
1441 | return len; | |
1442 | } | |
1443 | ||
1444 | static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen) | |
1445 | { | |
1446 | int error = -EACCES; | |
1447 | struct inode *inode = dentry->d_inode; | |
3dcd25f3 | 1448 | struct path path; |
1da177e4 | 1449 | |
778c1144 EB |
1450 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1451 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1452 | goto out; |
1da177e4 | 1453 | |
3dcd25f3 | 1454 | error = PROC_I(inode)->op.proc_get_link(inode, &path); |
1da177e4 LT |
1455 | if (error) |
1456 | goto out; | |
1457 | ||
3dcd25f3 JB |
1458 | error = do_proc_readlink(&path, buffer, buflen); |
1459 | path_put(&path); | |
1da177e4 | 1460 | out: |
1da177e4 LT |
1461 | return error; |
1462 | } | |
1463 | ||
c5ef1c42 | 1464 | static const struct inode_operations proc_pid_link_inode_operations = { |
1da177e4 | 1465 | .readlink = proc_pid_readlink, |
6d76fa58 LT |
1466 | .follow_link = proc_pid_follow_link, |
1467 | .setattr = proc_setattr, | |
1da177e4 LT |
1468 | }; |
1469 | ||
28a6d671 EB |
1470 | |
1471 | /* building an inode */ | |
1472 | ||
1473 | static int task_dumpable(struct task_struct *task) | |
1da177e4 | 1474 | { |
28a6d671 EB |
1475 | int dumpable = 0; |
1476 | struct mm_struct *mm; | |
1da177e4 | 1477 | |
28a6d671 EB |
1478 | task_lock(task); |
1479 | mm = task->mm; | |
1480 | if (mm) | |
6c5d5238 | 1481 | dumpable = get_dumpable(mm); |
28a6d671 EB |
1482 | task_unlock(task); |
1483 | if(dumpable == 1) | |
1484 | return 1; | |
1485 | return 0; | |
1486 | } | |
1da177e4 | 1487 | |
1da177e4 | 1488 | |
61a28784 | 1489 | static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task) |
28a6d671 EB |
1490 | { |
1491 | struct inode * inode; | |
1492 | struct proc_inode *ei; | |
c69e8d9c | 1493 | const struct cred *cred; |
1da177e4 | 1494 | |
28a6d671 | 1495 | /* We need a new inode */ |
1da177e4 | 1496 | |
28a6d671 EB |
1497 | inode = new_inode(sb); |
1498 | if (!inode) | |
1499 | goto out; | |
1500 | ||
1501 | /* Common stuff */ | |
1502 | ei = PROC_I(inode); | |
1503 | inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; | |
28a6d671 EB |
1504 | inode->i_op = &proc_def_inode_operations; |
1505 | ||
1506 | /* | |
1507 | * grab the reference to task. | |
1508 | */ | |
1a657f78 | 1509 | ei->pid = get_task_pid(task, PIDTYPE_PID); |
28a6d671 EB |
1510 | if (!ei->pid) |
1511 | goto out_unlock; | |
1512 | ||
28a6d671 | 1513 | if (task_dumpable(task)) { |
c69e8d9c DH |
1514 | rcu_read_lock(); |
1515 | cred = __task_cred(task); | |
1516 | inode->i_uid = cred->euid; | |
1517 | inode->i_gid = cred->egid; | |
1518 | rcu_read_unlock(); | |
1da177e4 | 1519 | } |
28a6d671 EB |
1520 | security_task_to_inode(task, inode); |
1521 | ||
1da177e4 | 1522 | out: |
28a6d671 EB |
1523 | return inode; |
1524 | ||
1525 | out_unlock: | |
1526 | iput(inode); | |
1527 | return NULL; | |
1da177e4 LT |
1528 | } |
1529 | ||
28a6d671 | 1530 | static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat) |
1da177e4 | 1531 | { |
1da177e4 | 1532 | struct inode *inode = dentry->d_inode; |
28a6d671 | 1533 | struct task_struct *task; |
c69e8d9c DH |
1534 | const struct cred *cred; |
1535 | ||
28a6d671 | 1536 | generic_fillattr(inode, stat); |
1da177e4 | 1537 | |
28a6d671 EB |
1538 | rcu_read_lock(); |
1539 | stat->uid = 0; | |
1540 | stat->gid = 0; | |
1541 | task = pid_task(proc_pid(inode), PIDTYPE_PID); | |
1542 | if (task) { | |
1543 | if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) || | |
1544 | task_dumpable(task)) { | |
c69e8d9c DH |
1545 | cred = __task_cred(task); |
1546 | stat->uid = cred->euid; | |
1547 | stat->gid = cred->egid; | |
1da177e4 LT |
1548 | } |
1549 | } | |
28a6d671 | 1550 | rcu_read_unlock(); |
d6e71144 | 1551 | return 0; |
1da177e4 LT |
1552 | } |
1553 | ||
1da177e4 LT |
1554 | /* dentry stuff */ |
1555 | ||
1556 | /* | |
1557 | * Exceptional case: normally we are not allowed to unhash a busy | |
1558 | * directory. In this case, however, we can do it - no aliasing problems | |
1559 | * due to the way we treat inodes. | |
1560 | * | |
1561 | * Rewrite the inode's ownerships here because the owning task may have | |
1562 | * performed a setuid(), etc. | |
99f89551 EB |
1563 | * |
1564 | * Before the /proc/pid/status file was created the only way to read | |
1565 | * the effective uid of a /process was to stat /proc/pid. Reading | |
1566 | * /proc/pid/status is slow enough that procps and other packages | |
1567 | * kept stating /proc/pid. To keep the rules in /proc simple I have | |
1568 | * made this apply to all per process world readable and executable | |
1569 | * directories. | |
1da177e4 LT |
1570 | */ |
1571 | static int pid_revalidate(struct dentry *dentry, struct nameidata *nd) | |
1572 | { | |
1573 | struct inode *inode = dentry->d_inode; | |
99f89551 | 1574 | struct task_struct *task = get_proc_task(inode); |
c69e8d9c DH |
1575 | const struct cred *cred; |
1576 | ||
99f89551 EB |
1577 | if (task) { |
1578 | if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) || | |
1579 | task_dumpable(task)) { | |
c69e8d9c DH |
1580 | rcu_read_lock(); |
1581 | cred = __task_cred(task); | |
1582 | inode->i_uid = cred->euid; | |
1583 | inode->i_gid = cred->egid; | |
1584 | rcu_read_unlock(); | |
1da177e4 LT |
1585 | } else { |
1586 | inode->i_uid = 0; | |
1587 | inode->i_gid = 0; | |
1588 | } | |
9ee8ab9f | 1589 | inode->i_mode &= ~(S_ISUID | S_ISGID); |
1da177e4 | 1590 | security_task_to_inode(task, inode); |
99f89551 | 1591 | put_task_struct(task); |
1da177e4 LT |
1592 | return 1; |
1593 | } | |
1594 | d_drop(dentry); | |
1595 | return 0; | |
1596 | } | |
1597 | ||
28a6d671 | 1598 | static int pid_delete_dentry(struct dentry * dentry) |
99f89551 | 1599 | { |
28a6d671 EB |
1600 | /* Is the task we represent dead? |
1601 | * If so, then don't put the dentry on the lru list, | |
1602 | * kill it immediately. | |
1603 | */ | |
1604 | return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first; | |
1605 | } | |
1606 | ||
d72f71eb | 1607 | static const struct dentry_operations pid_dentry_operations = |
28a6d671 EB |
1608 | { |
1609 | .d_revalidate = pid_revalidate, | |
1610 | .d_delete = pid_delete_dentry, | |
1611 | }; | |
1612 | ||
1613 | /* Lookups */ | |
1614 | ||
c5141e6d ED |
1615 | typedef struct dentry *instantiate_t(struct inode *, struct dentry *, |
1616 | struct task_struct *, const void *); | |
61a28784 | 1617 | |
1c0d04c9 EB |
1618 | /* |
1619 | * Fill a directory entry. | |
1620 | * | |
1621 | * If possible create the dcache entry and derive our inode number and | |
1622 | * file type from dcache entry. | |
1623 | * | |
1624 | * Since all of the proc inode numbers are dynamically generated, the inode | |
1625 | * numbers do not exist until the inode is cache. This means creating the | |
1626 | * the dcache entry in readdir is necessary to keep the inode numbers | |
1627 | * reported by readdir in sync with the inode numbers reported | |
1628 | * by stat. | |
1629 | */ | |
61a28784 EB |
1630 | static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir, |
1631 | char *name, int len, | |
c5141e6d | 1632 | instantiate_t instantiate, struct task_struct *task, const void *ptr) |
61a28784 | 1633 | { |
2fddfeef | 1634 | struct dentry *child, *dir = filp->f_path.dentry; |
61a28784 EB |
1635 | struct inode *inode; |
1636 | struct qstr qname; | |
1637 | ino_t ino = 0; | |
1638 | unsigned type = DT_UNKNOWN; | |
1639 | ||
1640 | qname.name = name; | |
1641 | qname.len = len; | |
1642 | qname.hash = full_name_hash(name, len); | |
1643 | ||
1644 | child = d_lookup(dir, &qname); | |
1645 | if (!child) { | |
1646 | struct dentry *new; | |
1647 | new = d_alloc(dir, &qname); | |
1648 | if (new) { | |
1649 | child = instantiate(dir->d_inode, new, task, ptr); | |
1650 | if (child) | |
1651 | dput(new); | |
1652 | else | |
1653 | child = new; | |
1654 | } | |
1655 | } | |
1656 | if (!child || IS_ERR(child) || !child->d_inode) | |
1657 | goto end_instantiate; | |
1658 | inode = child->d_inode; | |
1659 | if (inode) { | |
1660 | ino = inode->i_ino; | |
1661 | type = inode->i_mode >> 12; | |
1662 | } | |
1663 | dput(child); | |
1664 | end_instantiate: | |
1665 | if (!ino) | |
1666 | ino = find_inode_number(dir, &qname); | |
1667 | if (!ino) | |
1668 | ino = 1; | |
1669 | return filldir(dirent, name, len, filp->f_pos, ino, type); | |
1670 | } | |
1671 | ||
28a6d671 EB |
1672 | static unsigned name_to_int(struct dentry *dentry) |
1673 | { | |
1674 | const char *name = dentry->d_name.name; | |
1675 | int len = dentry->d_name.len; | |
1676 | unsigned n = 0; | |
1677 | ||
1678 | if (len > 1 && *name == '0') | |
1679 | goto out; | |
1680 | while (len-- > 0) { | |
1681 | unsigned c = *name++ - '0'; | |
1682 | if (c > 9) | |
1683 | goto out; | |
1684 | if (n >= (~0U-9)/10) | |
1685 | goto out; | |
1686 | n *= 10; | |
1687 | n += c; | |
1688 | } | |
1689 | return n; | |
1690 | out: | |
1691 | return ~0U; | |
1692 | } | |
1693 | ||
27932742 MS |
1694 | #define PROC_FDINFO_MAX 64 |
1695 | ||
3dcd25f3 | 1696 | static int proc_fd_info(struct inode *inode, struct path *path, char *info) |
28a6d671 EB |
1697 | { |
1698 | struct task_struct *task = get_proc_task(inode); | |
1699 | struct files_struct *files = NULL; | |
1700 | struct file *file; | |
1701 | int fd = proc_fd(inode); | |
99f89551 | 1702 | |
99f89551 | 1703 | if (task) { |
28a6d671 EB |
1704 | files = get_files_struct(task); |
1705 | put_task_struct(task); | |
1706 | } | |
1707 | if (files) { | |
1708 | /* | |
1709 | * We are not taking a ref to the file structure, so we must | |
1710 | * hold ->file_lock. | |
1711 | */ | |
1712 | spin_lock(&files->file_lock); | |
1713 | file = fcheck_files(files, fd); | |
1714 | if (file) { | |
3dcd25f3 JB |
1715 | if (path) { |
1716 | *path = file->f_path; | |
1717 | path_get(&file->f_path); | |
1718 | } | |
27932742 MS |
1719 | if (info) |
1720 | snprintf(info, PROC_FDINFO_MAX, | |
1721 | "pos:\t%lli\n" | |
1722 | "flags:\t0%o\n", | |
1723 | (long long) file->f_pos, | |
1724 | file->f_flags); | |
28a6d671 EB |
1725 | spin_unlock(&files->file_lock); |
1726 | put_files_struct(files); | |
1727 | return 0; | |
99f89551 | 1728 | } |
28a6d671 EB |
1729 | spin_unlock(&files->file_lock); |
1730 | put_files_struct(files); | |
99f89551 | 1731 | } |
28a6d671 | 1732 | return -ENOENT; |
99f89551 EB |
1733 | } |
1734 | ||
3dcd25f3 | 1735 | static int proc_fd_link(struct inode *inode, struct path *path) |
27932742 | 1736 | { |
3dcd25f3 | 1737 | return proc_fd_info(inode, path, NULL); |
27932742 MS |
1738 | } |
1739 | ||
1da177e4 LT |
1740 | static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd) |
1741 | { | |
1742 | struct inode *inode = dentry->d_inode; | |
99f89551 | 1743 | struct task_struct *task = get_proc_task(inode); |
aed7a6c4 | 1744 | int fd = proc_fd(inode); |
1da177e4 | 1745 | struct files_struct *files; |
c69e8d9c | 1746 | const struct cred *cred; |
1da177e4 | 1747 | |
99f89551 EB |
1748 | if (task) { |
1749 | files = get_files_struct(task); | |
1750 | if (files) { | |
1751 | rcu_read_lock(); | |
1752 | if (fcheck_files(files, fd)) { | |
1753 | rcu_read_unlock(); | |
1754 | put_files_struct(files); | |
1755 | if (task_dumpable(task)) { | |
c69e8d9c DH |
1756 | rcu_read_lock(); |
1757 | cred = __task_cred(task); | |
1758 | inode->i_uid = cred->euid; | |
1759 | inode->i_gid = cred->egid; | |
1760 | rcu_read_unlock(); | |
99f89551 EB |
1761 | } else { |
1762 | inode->i_uid = 0; | |
1763 | inode->i_gid = 0; | |
1764 | } | |
9ee8ab9f | 1765 | inode->i_mode &= ~(S_ISUID | S_ISGID); |
99f89551 EB |
1766 | security_task_to_inode(task, inode); |
1767 | put_task_struct(task); | |
1768 | return 1; | |
1769 | } | |
b835996f | 1770 | rcu_read_unlock(); |
1da177e4 | 1771 | put_files_struct(files); |
1da177e4 | 1772 | } |
99f89551 | 1773 | put_task_struct(task); |
1da177e4 LT |
1774 | } |
1775 | d_drop(dentry); | |
1776 | return 0; | |
1777 | } | |
1778 | ||
d72f71eb | 1779 | static const struct dentry_operations tid_fd_dentry_operations = |
1da177e4 LT |
1780 | { |
1781 | .d_revalidate = tid_fd_revalidate, | |
1782 | .d_delete = pid_delete_dentry, | |
1783 | }; | |
1784 | ||
444ceed8 | 1785 | static struct dentry *proc_fd_instantiate(struct inode *dir, |
c5141e6d | 1786 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
1da177e4 | 1787 | { |
c5141e6d | 1788 | unsigned fd = *(const unsigned *)ptr; |
444ceed8 EB |
1789 | struct file *file; |
1790 | struct files_struct *files; | |
1791 | struct inode *inode; | |
1792 | struct proc_inode *ei; | |
1793 | struct dentry *error = ERR_PTR(-ENOENT); | |
1da177e4 | 1794 | |
61a28784 | 1795 | inode = proc_pid_make_inode(dir->i_sb, task); |
1da177e4 LT |
1796 | if (!inode) |
1797 | goto out; | |
1798 | ei = PROC_I(inode); | |
aed7a6c4 | 1799 | ei->fd = fd; |
1da177e4 LT |
1800 | files = get_files_struct(task); |
1801 | if (!files) | |
444ceed8 | 1802 | goto out_iput; |
1da177e4 | 1803 | inode->i_mode = S_IFLNK; |
ca99c1da DS |
1804 | |
1805 | /* | |
1806 | * We are not taking a ref to the file structure, so we must | |
1807 | * hold ->file_lock. | |
1808 | */ | |
1809 | spin_lock(&files->file_lock); | |
1da177e4 LT |
1810 | file = fcheck_files(files, fd); |
1811 | if (!file) | |
444ceed8 | 1812 | goto out_unlock; |
aeb5d727 | 1813 | if (file->f_mode & FMODE_READ) |
1da177e4 | 1814 | inode->i_mode |= S_IRUSR | S_IXUSR; |
aeb5d727 | 1815 | if (file->f_mode & FMODE_WRITE) |
1da177e4 | 1816 | inode->i_mode |= S_IWUSR | S_IXUSR; |
ca99c1da | 1817 | spin_unlock(&files->file_lock); |
1da177e4 | 1818 | put_files_struct(files); |
444ceed8 | 1819 | |
1da177e4 LT |
1820 | inode->i_op = &proc_pid_link_inode_operations; |
1821 | inode->i_size = 64; | |
1822 | ei->op.proc_get_link = proc_fd_link; | |
1823 | dentry->d_op = &tid_fd_dentry_operations; | |
1824 | d_add(dentry, inode); | |
cd6a3ce9 EB |
1825 | /* Close the race of the process dying before we return the dentry */ |
1826 | if (tid_fd_revalidate(dentry, NULL)) | |
444ceed8 | 1827 | error = NULL; |
1da177e4 | 1828 | |
444ceed8 EB |
1829 | out: |
1830 | return error; | |
1831 | out_unlock: | |
ca99c1da | 1832 | spin_unlock(&files->file_lock); |
1da177e4 | 1833 | put_files_struct(files); |
444ceed8 | 1834 | out_iput: |
1da177e4 | 1835 | iput(inode); |
cd6a3ce9 | 1836 | goto out; |
1da177e4 LT |
1837 | } |
1838 | ||
27932742 MS |
1839 | static struct dentry *proc_lookupfd_common(struct inode *dir, |
1840 | struct dentry *dentry, | |
1841 | instantiate_t instantiate) | |
444ceed8 EB |
1842 | { |
1843 | struct task_struct *task = get_proc_task(dir); | |
1844 | unsigned fd = name_to_int(dentry); | |
1845 | struct dentry *result = ERR_PTR(-ENOENT); | |
1846 | ||
1847 | if (!task) | |
1848 | goto out_no_task; | |
1849 | if (fd == ~0U) | |
1850 | goto out; | |
1851 | ||
27932742 | 1852 | result = instantiate(dir, dentry, task, &fd); |
444ceed8 EB |
1853 | out: |
1854 | put_task_struct(task); | |
1855 | out_no_task: | |
1856 | return result; | |
1857 | } | |
1858 | ||
27932742 MS |
1859 | static int proc_readfd_common(struct file * filp, void * dirent, |
1860 | filldir_t filldir, instantiate_t instantiate) | |
28a6d671 | 1861 | { |
2fddfeef | 1862 | struct dentry *dentry = filp->f_path.dentry; |
28a6d671 EB |
1863 | struct inode *inode = dentry->d_inode; |
1864 | struct task_struct *p = get_proc_task(inode); | |
457c2510 | 1865 | unsigned int fd, ino; |
28a6d671 | 1866 | int retval; |
28a6d671 | 1867 | struct files_struct * files; |
1da177e4 | 1868 | |
28a6d671 EB |
1869 | retval = -ENOENT; |
1870 | if (!p) | |
1871 | goto out_no_task; | |
1872 | retval = 0; | |
28a6d671 EB |
1873 | |
1874 | fd = filp->f_pos; | |
1875 | switch (fd) { | |
1876 | case 0: | |
1877 | if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0) | |
1878 | goto out; | |
1879 | filp->f_pos++; | |
1880 | case 1: | |
1881 | ino = parent_ino(dentry); | |
1882 | if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0) | |
1883 | goto out; | |
1884 | filp->f_pos++; | |
1885 | default: | |
1886 | files = get_files_struct(p); | |
1887 | if (!files) | |
1888 | goto out; | |
1889 | rcu_read_lock(); | |
28a6d671 | 1890 | for (fd = filp->f_pos-2; |
9b4f526c | 1891 | fd < files_fdtable(files)->max_fds; |
28a6d671 | 1892 | fd++, filp->f_pos++) { |
27932742 MS |
1893 | char name[PROC_NUMBUF]; |
1894 | int len; | |
28a6d671 EB |
1895 | |
1896 | if (!fcheck_files(files, fd)) | |
1897 | continue; | |
1898 | rcu_read_unlock(); | |
1899 | ||
27932742 MS |
1900 | len = snprintf(name, sizeof(name), "%d", fd); |
1901 | if (proc_fill_cache(filp, dirent, filldir, | |
1902 | name, len, instantiate, | |
1903 | p, &fd) < 0) { | |
28a6d671 EB |
1904 | rcu_read_lock(); |
1905 | break; | |
1906 | } | |
1907 | rcu_read_lock(); | |
1908 | } | |
1909 | rcu_read_unlock(); | |
1910 | put_files_struct(files); | |
1911 | } | |
1912 | out: | |
1913 | put_task_struct(p); | |
1914 | out_no_task: | |
1915 | return retval; | |
1916 | } | |
1917 | ||
27932742 MS |
1918 | static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry, |
1919 | struct nameidata *nd) | |
1920 | { | |
1921 | return proc_lookupfd_common(dir, dentry, proc_fd_instantiate); | |
1922 | } | |
1923 | ||
1924 | static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir) | |
1925 | { | |
1926 | return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate); | |
1927 | } | |
1928 | ||
1929 | static ssize_t proc_fdinfo_read(struct file *file, char __user *buf, | |
1930 | size_t len, loff_t *ppos) | |
1931 | { | |
1932 | char tmp[PROC_FDINFO_MAX]; | |
3dcd25f3 | 1933 | int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp); |
27932742 MS |
1934 | if (!err) |
1935 | err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp)); | |
1936 | return err; | |
1937 | } | |
1938 | ||
1939 | static const struct file_operations proc_fdinfo_file_operations = { | |
87df8424 | 1940 | .open = nonseekable_open, |
27932742 MS |
1941 | .read = proc_fdinfo_read, |
1942 | }; | |
1943 | ||
00977a59 | 1944 | static const struct file_operations proc_fd_operations = { |
28a6d671 EB |
1945 | .read = generic_read_dir, |
1946 | .readdir = proc_readfd, | |
1da177e4 LT |
1947 | }; |
1948 | ||
8948e11f AD |
1949 | /* |
1950 | * /proc/pid/fd needs a special permission handler so that a process can still | |
1951 | * access /proc/self/fd after it has executed a setuid(). | |
1952 | */ | |
e6305c43 | 1953 | static int proc_fd_permission(struct inode *inode, int mask) |
8948e11f AD |
1954 | { |
1955 | int rv; | |
1956 | ||
1957 | rv = generic_permission(inode, mask, NULL); | |
1958 | if (rv == 0) | |
1959 | return 0; | |
1960 | if (task_pid(current) == proc_pid(inode)) | |
1961 | rv = 0; | |
1962 | return rv; | |
1963 | } | |
1964 | ||
1da177e4 LT |
1965 | /* |
1966 | * proc directories can do almost nothing.. | |
1967 | */ | |
c5ef1c42 | 1968 | static const struct inode_operations proc_fd_inode_operations = { |
1da177e4 | 1969 | .lookup = proc_lookupfd, |
8948e11f | 1970 | .permission = proc_fd_permission, |
6d76fa58 | 1971 | .setattr = proc_setattr, |
1da177e4 LT |
1972 | }; |
1973 | ||
27932742 MS |
1974 | static struct dentry *proc_fdinfo_instantiate(struct inode *dir, |
1975 | struct dentry *dentry, struct task_struct *task, const void *ptr) | |
1976 | { | |
1977 | unsigned fd = *(unsigned *)ptr; | |
1978 | struct inode *inode; | |
1979 | struct proc_inode *ei; | |
1980 | struct dentry *error = ERR_PTR(-ENOENT); | |
1981 | ||
1982 | inode = proc_pid_make_inode(dir->i_sb, task); | |
1983 | if (!inode) | |
1984 | goto out; | |
1985 | ei = PROC_I(inode); | |
1986 | ei->fd = fd; | |
1987 | inode->i_mode = S_IFREG | S_IRUSR; | |
1988 | inode->i_fop = &proc_fdinfo_file_operations; | |
1989 | dentry->d_op = &tid_fd_dentry_operations; | |
1990 | d_add(dentry, inode); | |
1991 | /* Close the race of the process dying before we return the dentry */ | |
1992 | if (tid_fd_revalidate(dentry, NULL)) | |
1993 | error = NULL; | |
1994 | ||
1995 | out: | |
1996 | return error; | |
1997 | } | |
1998 | ||
1999 | static struct dentry *proc_lookupfdinfo(struct inode *dir, | |
2000 | struct dentry *dentry, | |
2001 | struct nameidata *nd) | |
2002 | { | |
2003 | return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate); | |
2004 | } | |
2005 | ||
2006 | static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir) | |
2007 | { | |
2008 | return proc_readfd_common(filp, dirent, filldir, | |
2009 | proc_fdinfo_instantiate); | |
2010 | } | |
2011 | ||
2012 | static const struct file_operations proc_fdinfo_operations = { | |
2013 | .read = generic_read_dir, | |
2014 | .readdir = proc_readfdinfo, | |
2015 | }; | |
2016 | ||
2017 | /* | |
2018 | * proc directories can do almost nothing.. | |
2019 | */ | |
2020 | static const struct inode_operations proc_fdinfo_inode_operations = { | |
2021 | .lookup = proc_lookupfdinfo, | |
2022 | .setattr = proc_setattr, | |
2023 | }; | |
2024 | ||
2025 | ||
444ceed8 | 2026 | static struct dentry *proc_pident_instantiate(struct inode *dir, |
c5141e6d | 2027 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
444ceed8 | 2028 | { |
c5141e6d | 2029 | const struct pid_entry *p = ptr; |
444ceed8 EB |
2030 | struct inode *inode; |
2031 | struct proc_inode *ei; | |
bd6daba9 | 2032 | struct dentry *error = ERR_PTR(-ENOENT); |
444ceed8 | 2033 | |
61a28784 | 2034 | inode = proc_pid_make_inode(dir->i_sb, task); |
444ceed8 EB |
2035 | if (!inode) |
2036 | goto out; | |
2037 | ||
2038 | ei = PROC_I(inode); | |
2039 | inode->i_mode = p->mode; | |
2040 | if (S_ISDIR(inode->i_mode)) | |
2041 | inode->i_nlink = 2; /* Use getattr to fix if necessary */ | |
2042 | if (p->iop) | |
2043 | inode->i_op = p->iop; | |
2044 | if (p->fop) | |
2045 | inode->i_fop = p->fop; | |
2046 | ei->op = p->op; | |
2047 | dentry->d_op = &pid_dentry_operations; | |
2048 | d_add(dentry, inode); | |
2049 | /* Close the race of the process dying before we return the dentry */ | |
2050 | if (pid_revalidate(dentry, NULL)) | |
2051 | error = NULL; | |
2052 | out: | |
2053 | return error; | |
2054 | } | |
2055 | ||
1da177e4 LT |
2056 | static struct dentry *proc_pident_lookup(struct inode *dir, |
2057 | struct dentry *dentry, | |
c5141e6d | 2058 | const struct pid_entry *ents, |
7bcd6b0e | 2059 | unsigned int nents) |
1da177e4 | 2060 | { |
cd6a3ce9 | 2061 | struct dentry *error; |
99f89551 | 2062 | struct task_struct *task = get_proc_task(dir); |
c5141e6d | 2063 | const struct pid_entry *p, *last; |
1da177e4 | 2064 | |
cd6a3ce9 | 2065 | error = ERR_PTR(-ENOENT); |
1da177e4 | 2066 | |
99f89551 EB |
2067 | if (!task) |
2068 | goto out_no_task; | |
1da177e4 | 2069 | |
20cdc894 EB |
2070 | /* |
2071 | * Yes, it does not scale. And it should not. Don't add | |
2072 | * new entries into /proc/<tgid>/ without very good reasons. | |
2073 | */ | |
7bcd6b0e EB |
2074 | last = &ents[nents - 1]; |
2075 | for (p = ents; p <= last; p++) { | |
1da177e4 LT |
2076 | if (p->len != dentry->d_name.len) |
2077 | continue; | |
2078 | if (!memcmp(dentry->d_name.name, p->name, p->len)) | |
2079 | break; | |
2080 | } | |
7bcd6b0e | 2081 | if (p > last) |
1da177e4 LT |
2082 | goto out; |
2083 | ||
444ceed8 | 2084 | error = proc_pident_instantiate(dir, dentry, task, p); |
1da177e4 | 2085 | out: |
99f89551 EB |
2086 | put_task_struct(task); |
2087 | out_no_task: | |
cd6a3ce9 | 2088 | return error; |
1da177e4 LT |
2089 | } |
2090 | ||
c5141e6d ED |
2091 | static int proc_pident_fill_cache(struct file *filp, void *dirent, |
2092 | filldir_t filldir, struct task_struct *task, const struct pid_entry *p) | |
61a28784 EB |
2093 | { |
2094 | return proc_fill_cache(filp, dirent, filldir, p->name, p->len, | |
2095 | proc_pident_instantiate, task, p); | |
2096 | } | |
2097 | ||
28a6d671 EB |
2098 | static int proc_pident_readdir(struct file *filp, |
2099 | void *dirent, filldir_t filldir, | |
c5141e6d | 2100 | const struct pid_entry *ents, unsigned int nents) |
28a6d671 EB |
2101 | { |
2102 | int i; | |
2fddfeef | 2103 | struct dentry *dentry = filp->f_path.dentry; |
28a6d671 EB |
2104 | struct inode *inode = dentry->d_inode; |
2105 | struct task_struct *task = get_proc_task(inode); | |
c5141e6d | 2106 | const struct pid_entry *p, *last; |
28a6d671 EB |
2107 | ino_t ino; |
2108 | int ret; | |
2109 | ||
2110 | ret = -ENOENT; | |
2111 | if (!task) | |
61a28784 | 2112 | goto out_no_task; |
28a6d671 EB |
2113 | |
2114 | ret = 0; | |
28a6d671 EB |
2115 | i = filp->f_pos; |
2116 | switch (i) { | |
2117 | case 0: | |
2118 | ino = inode->i_ino; | |
2119 | if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0) | |
2120 | goto out; | |
2121 | i++; | |
2122 | filp->f_pos++; | |
2123 | /* fall through */ | |
2124 | case 1: | |
2125 | ino = parent_ino(dentry); | |
2126 | if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0) | |
2127 | goto out; | |
2128 | i++; | |
2129 | filp->f_pos++; | |
2130 | /* fall through */ | |
2131 | default: | |
2132 | i -= 2; | |
2133 | if (i >= nents) { | |
2134 | ret = 1; | |
2135 | goto out; | |
2136 | } | |
2137 | p = ents + i; | |
7bcd6b0e EB |
2138 | last = &ents[nents - 1]; |
2139 | while (p <= last) { | |
61a28784 | 2140 | if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0) |
28a6d671 EB |
2141 | goto out; |
2142 | filp->f_pos++; | |
2143 | p++; | |
2144 | } | |
2145 | } | |
2146 | ||
2147 | ret = 1; | |
2148 | out: | |
61a28784 EB |
2149 | put_task_struct(task); |
2150 | out_no_task: | |
28a6d671 | 2151 | return ret; |
1da177e4 LT |
2152 | } |
2153 | ||
28a6d671 EB |
2154 | #ifdef CONFIG_SECURITY |
2155 | static ssize_t proc_pid_attr_read(struct file * file, char __user * buf, | |
2156 | size_t count, loff_t *ppos) | |
2157 | { | |
2fddfeef | 2158 | struct inode * inode = file->f_path.dentry->d_inode; |
04ff9708 | 2159 | char *p = NULL; |
28a6d671 EB |
2160 | ssize_t length; |
2161 | struct task_struct *task = get_proc_task(inode); | |
2162 | ||
28a6d671 | 2163 | if (!task) |
04ff9708 | 2164 | return -ESRCH; |
28a6d671 EB |
2165 | |
2166 | length = security_getprocattr(task, | |
2fddfeef | 2167 | (char*)file->f_path.dentry->d_name.name, |
04ff9708 | 2168 | &p); |
28a6d671 | 2169 | put_task_struct(task); |
04ff9708 AV |
2170 | if (length > 0) |
2171 | length = simple_read_from_buffer(buf, count, ppos, p, length); | |
2172 | kfree(p); | |
28a6d671 | 2173 | return length; |
1da177e4 LT |
2174 | } |
2175 | ||
28a6d671 EB |
2176 | static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf, |
2177 | size_t count, loff_t *ppos) | |
2178 | { | |
2fddfeef | 2179 | struct inode * inode = file->f_path.dentry->d_inode; |
28a6d671 EB |
2180 | char *page; |
2181 | ssize_t length; | |
2182 | struct task_struct *task = get_proc_task(inode); | |
2183 | ||
2184 | length = -ESRCH; | |
2185 | if (!task) | |
2186 | goto out_no_task; | |
2187 | if (count > PAGE_SIZE) | |
2188 | count = PAGE_SIZE; | |
2189 | ||
2190 | /* No partial writes. */ | |
2191 | length = -EINVAL; | |
2192 | if (*ppos != 0) | |
2193 | goto out; | |
2194 | ||
2195 | length = -ENOMEM; | |
e12ba74d | 2196 | page = (char*)__get_free_page(GFP_TEMPORARY); |
28a6d671 EB |
2197 | if (!page) |
2198 | goto out; | |
2199 | ||
2200 | length = -EFAULT; | |
2201 | if (copy_from_user(page, buf, count)) | |
2202 | goto out_free; | |
2203 | ||
107db7c7 DH |
2204 | /* Guard against adverse ptrace interaction */ |
2205 | length = mutex_lock_interruptible(&task->cred_guard_mutex); | |
2206 | if (length < 0) | |
2207 | goto out_free; | |
2208 | ||
28a6d671 | 2209 | length = security_setprocattr(task, |
2fddfeef | 2210 | (char*)file->f_path.dentry->d_name.name, |
28a6d671 | 2211 | (void*)page, count); |
107db7c7 | 2212 | mutex_unlock(&task->cred_guard_mutex); |
28a6d671 EB |
2213 | out_free: |
2214 | free_page((unsigned long) page); | |
2215 | out: | |
2216 | put_task_struct(task); | |
2217 | out_no_task: | |
2218 | return length; | |
2219 | } | |
2220 | ||
00977a59 | 2221 | static const struct file_operations proc_pid_attr_operations = { |
28a6d671 EB |
2222 | .read = proc_pid_attr_read, |
2223 | .write = proc_pid_attr_write, | |
87df8424 | 2224 | .llseek = generic_file_llseek, |
28a6d671 EB |
2225 | }; |
2226 | ||
c5141e6d | 2227 | static const struct pid_entry attr_dir_stuff[] = { |
631f9c18 AD |
2228 | REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations), |
2229 | REG("prev", S_IRUGO, proc_pid_attr_operations), | |
2230 | REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2231 | REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2232 | REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2233 | REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
28a6d671 EB |
2234 | }; |
2235 | ||
72d9dcfc | 2236 | static int proc_attr_dir_readdir(struct file * filp, |
28a6d671 EB |
2237 | void * dirent, filldir_t filldir) |
2238 | { | |
2239 | return proc_pident_readdir(filp,dirent,filldir, | |
72d9dcfc | 2240 | attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff)); |
28a6d671 EB |
2241 | } |
2242 | ||
00977a59 | 2243 | static const struct file_operations proc_attr_dir_operations = { |
1da177e4 | 2244 | .read = generic_read_dir, |
72d9dcfc | 2245 | .readdir = proc_attr_dir_readdir, |
1da177e4 LT |
2246 | }; |
2247 | ||
72d9dcfc | 2248 | static struct dentry *proc_attr_dir_lookup(struct inode *dir, |
28a6d671 EB |
2249 | struct dentry *dentry, struct nameidata *nd) |
2250 | { | |
7bcd6b0e EB |
2251 | return proc_pident_lookup(dir, dentry, |
2252 | attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2253 | } |
2254 | ||
c5ef1c42 | 2255 | static const struct inode_operations proc_attr_dir_inode_operations = { |
72d9dcfc | 2256 | .lookup = proc_attr_dir_lookup, |
99f89551 | 2257 | .getattr = pid_getattr, |
6d76fa58 | 2258 | .setattr = proc_setattr, |
1da177e4 LT |
2259 | }; |
2260 | ||
28a6d671 EB |
2261 | #endif |
2262 | ||
698ba7b5 | 2263 | #ifdef CONFIG_ELF_CORE |
3cb4a0bb KH |
2264 | static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf, |
2265 | size_t count, loff_t *ppos) | |
2266 | { | |
2267 | struct task_struct *task = get_proc_task(file->f_dentry->d_inode); | |
2268 | struct mm_struct *mm; | |
2269 | char buffer[PROC_NUMBUF]; | |
2270 | size_t len; | |
2271 | int ret; | |
2272 | ||
2273 | if (!task) | |
2274 | return -ESRCH; | |
2275 | ||
2276 | ret = 0; | |
2277 | mm = get_task_mm(task); | |
2278 | if (mm) { | |
2279 | len = snprintf(buffer, sizeof(buffer), "%08lx\n", | |
2280 | ((mm->flags & MMF_DUMP_FILTER_MASK) >> | |
2281 | MMF_DUMP_FILTER_SHIFT)); | |
2282 | mmput(mm); | |
2283 | ret = simple_read_from_buffer(buf, count, ppos, buffer, len); | |
2284 | } | |
2285 | ||
2286 | put_task_struct(task); | |
2287 | ||
2288 | return ret; | |
2289 | } | |
2290 | ||
2291 | static ssize_t proc_coredump_filter_write(struct file *file, | |
2292 | const char __user *buf, | |
2293 | size_t count, | |
2294 | loff_t *ppos) | |
2295 | { | |
2296 | struct task_struct *task; | |
2297 | struct mm_struct *mm; | |
2298 | char buffer[PROC_NUMBUF], *end; | |
2299 | unsigned int val; | |
2300 | int ret; | |
2301 | int i; | |
2302 | unsigned long mask; | |
2303 | ||
2304 | ret = -EFAULT; | |
2305 | memset(buffer, 0, sizeof(buffer)); | |
2306 | if (count > sizeof(buffer) - 1) | |
2307 | count = sizeof(buffer) - 1; | |
2308 | if (copy_from_user(buffer, buf, count)) | |
2309 | goto out_no_task; | |
2310 | ||
2311 | ret = -EINVAL; | |
2312 | val = (unsigned int)simple_strtoul(buffer, &end, 0); | |
2313 | if (*end == '\n') | |
2314 | end++; | |
2315 | if (end - buffer == 0) | |
2316 | goto out_no_task; | |
2317 | ||
2318 | ret = -ESRCH; | |
2319 | task = get_proc_task(file->f_dentry->d_inode); | |
2320 | if (!task) | |
2321 | goto out_no_task; | |
2322 | ||
2323 | ret = end - buffer; | |
2324 | mm = get_task_mm(task); | |
2325 | if (!mm) | |
2326 | goto out_no_mm; | |
2327 | ||
2328 | for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) { | |
2329 | if (val & mask) | |
2330 | set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2331 | else | |
2332 | clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2333 | } | |
2334 | ||
2335 | mmput(mm); | |
2336 | out_no_mm: | |
2337 | put_task_struct(task); | |
2338 | out_no_task: | |
2339 | return ret; | |
2340 | } | |
2341 | ||
2342 | static const struct file_operations proc_coredump_filter_operations = { | |
2343 | .read = proc_coredump_filter_read, | |
2344 | .write = proc_coredump_filter_write, | |
87df8424 | 2345 | .llseek = generic_file_llseek, |
3cb4a0bb KH |
2346 | }; |
2347 | #endif | |
2348 | ||
28a6d671 EB |
2349 | /* |
2350 | * /proc/self: | |
2351 | */ | |
2352 | static int proc_self_readlink(struct dentry *dentry, char __user *buffer, | |
2353 | int buflen) | |
2354 | { | |
488e5bc4 | 2355 | struct pid_namespace *ns = dentry->d_sb->s_fs_info; |
b55fcb22 | 2356 | pid_t tgid = task_tgid_nr_ns(current, ns); |
28a6d671 | 2357 | char tmp[PROC_NUMBUF]; |
b55fcb22 | 2358 | if (!tgid) |
488e5bc4 | 2359 | return -ENOENT; |
b55fcb22 | 2360 | sprintf(tmp, "%d", tgid); |
28a6d671 EB |
2361 | return vfs_readlink(dentry,buffer,buflen,tmp); |
2362 | } | |
2363 | ||
2364 | static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd) | |
2365 | { | |
488e5bc4 | 2366 | struct pid_namespace *ns = dentry->d_sb->s_fs_info; |
b55fcb22 | 2367 | pid_t tgid = task_tgid_nr_ns(current, ns); |
7fee4868 AV |
2368 | char *name = ERR_PTR(-ENOENT); |
2369 | if (tgid) { | |
2370 | name = __getname(); | |
2371 | if (!name) | |
2372 | name = ERR_PTR(-ENOMEM); | |
2373 | else | |
2374 | sprintf(name, "%d", tgid); | |
2375 | } | |
2376 | nd_set_link(nd, name); | |
2377 | return NULL; | |
2378 | } | |
2379 | ||
2380 | static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd, | |
2381 | void *cookie) | |
2382 | { | |
2383 | char *s = nd_get_link(nd); | |
2384 | if (!IS_ERR(s)) | |
2385 | __putname(s); | |
28a6d671 EB |
2386 | } |
2387 | ||
c5ef1c42 | 2388 | static const struct inode_operations proc_self_inode_operations = { |
28a6d671 EB |
2389 | .readlink = proc_self_readlink, |
2390 | .follow_link = proc_self_follow_link, | |
7fee4868 | 2391 | .put_link = proc_self_put_link, |
28a6d671 EB |
2392 | }; |
2393 | ||
801199ce EB |
2394 | /* |
2395 | * proc base | |
2396 | * | |
2397 | * These are the directory entries in the root directory of /proc | |
2398 | * that properly belong to the /proc filesystem, as they describe | |
2399 | * describe something that is process related. | |
2400 | */ | |
c5141e6d | 2401 | static const struct pid_entry proc_base_stuff[] = { |
61a28784 | 2402 | NOD("self", S_IFLNK|S_IRWXUGO, |
801199ce | 2403 | &proc_self_inode_operations, NULL, {}), |
801199ce EB |
2404 | }; |
2405 | ||
2406 | /* | |
2407 | * Exceptional case: normally we are not allowed to unhash a busy | |
2408 | * directory. In this case, however, we can do it - no aliasing problems | |
2409 | * due to the way we treat inodes. | |
2410 | */ | |
2411 | static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd) | |
2412 | { | |
2413 | struct inode *inode = dentry->d_inode; | |
2414 | struct task_struct *task = get_proc_task(inode); | |
2415 | if (task) { | |
2416 | put_task_struct(task); | |
2417 | return 1; | |
2418 | } | |
2419 | d_drop(dentry); | |
2420 | return 0; | |
2421 | } | |
2422 | ||
d72f71eb | 2423 | static const struct dentry_operations proc_base_dentry_operations = |
801199ce EB |
2424 | { |
2425 | .d_revalidate = proc_base_revalidate, | |
2426 | .d_delete = pid_delete_dentry, | |
2427 | }; | |
2428 | ||
444ceed8 | 2429 | static struct dentry *proc_base_instantiate(struct inode *dir, |
c5141e6d | 2430 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
801199ce | 2431 | { |
c5141e6d | 2432 | const struct pid_entry *p = ptr; |
801199ce | 2433 | struct inode *inode; |
801199ce | 2434 | struct proc_inode *ei; |
73d36460 | 2435 | struct dentry *error; |
801199ce EB |
2436 | |
2437 | /* Allocate the inode */ | |
2438 | error = ERR_PTR(-ENOMEM); | |
2439 | inode = new_inode(dir->i_sb); | |
2440 | if (!inode) | |
2441 | goto out; | |
2442 | ||
2443 | /* Initialize the inode */ | |
2444 | ei = PROC_I(inode); | |
2445 | inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; | |
801199ce EB |
2446 | |
2447 | /* | |
2448 | * grab the reference to the task. | |
2449 | */ | |
1a657f78 | 2450 | ei->pid = get_task_pid(task, PIDTYPE_PID); |
801199ce EB |
2451 | if (!ei->pid) |
2452 | goto out_iput; | |
2453 | ||
801199ce EB |
2454 | inode->i_mode = p->mode; |
2455 | if (S_ISDIR(inode->i_mode)) | |
2456 | inode->i_nlink = 2; | |
2457 | if (S_ISLNK(inode->i_mode)) | |
2458 | inode->i_size = 64; | |
2459 | if (p->iop) | |
2460 | inode->i_op = p->iop; | |
2461 | if (p->fop) | |
2462 | inode->i_fop = p->fop; | |
2463 | ei->op = p->op; | |
2464 | dentry->d_op = &proc_base_dentry_operations; | |
2465 | d_add(dentry, inode); | |
2466 | error = NULL; | |
2467 | out: | |
801199ce EB |
2468 | return error; |
2469 | out_iput: | |
2470 | iput(inode); | |
2471 | goto out; | |
2472 | } | |
2473 | ||
444ceed8 EB |
2474 | static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry) |
2475 | { | |
2476 | struct dentry *error; | |
2477 | struct task_struct *task = get_proc_task(dir); | |
c5141e6d | 2478 | const struct pid_entry *p, *last; |
444ceed8 EB |
2479 | |
2480 | error = ERR_PTR(-ENOENT); | |
2481 | ||
2482 | if (!task) | |
2483 | goto out_no_task; | |
2484 | ||
2485 | /* Lookup the directory entry */ | |
7bcd6b0e EB |
2486 | last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1]; |
2487 | for (p = proc_base_stuff; p <= last; p++) { | |
444ceed8 EB |
2488 | if (p->len != dentry->d_name.len) |
2489 | continue; | |
2490 | if (!memcmp(dentry->d_name.name, p->name, p->len)) | |
2491 | break; | |
2492 | } | |
7bcd6b0e | 2493 | if (p > last) |
444ceed8 EB |
2494 | goto out; |
2495 | ||
2496 | error = proc_base_instantiate(dir, dentry, task, p); | |
2497 | ||
2498 | out: | |
2499 | put_task_struct(task); | |
2500 | out_no_task: | |
2501 | return error; | |
2502 | } | |
2503 | ||
c5141e6d ED |
2504 | static int proc_base_fill_cache(struct file *filp, void *dirent, |
2505 | filldir_t filldir, struct task_struct *task, const struct pid_entry *p) | |
61a28784 EB |
2506 | { |
2507 | return proc_fill_cache(filp, dirent, filldir, p->name, p->len, | |
2508 | proc_base_instantiate, task, p); | |
2509 | } | |
2510 | ||
aba76fdb | 2511 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
297c5d92 AR |
2512 | static int do_io_accounting(struct task_struct *task, char *buffer, int whole) |
2513 | { | |
940389b8 | 2514 | struct task_io_accounting acct = task->ioac; |
5995477a AR |
2515 | unsigned long flags; |
2516 | ||
2517 | if (whole && lock_task_sighand(task, &flags)) { | |
2518 | struct task_struct *t = task; | |
2519 | ||
2520 | task_io_accounting_add(&acct, &task->signal->ioac); | |
2521 | while_each_thread(task, t) | |
2522 | task_io_accounting_add(&acct, &t->ioac); | |
2523 | ||
2524 | unlock_task_sighand(task, &flags); | |
297c5d92 | 2525 | } |
aba76fdb AM |
2526 | return sprintf(buffer, |
2527 | "rchar: %llu\n" | |
2528 | "wchar: %llu\n" | |
2529 | "syscr: %llu\n" | |
2530 | "syscw: %llu\n" | |
2531 | "read_bytes: %llu\n" | |
2532 | "write_bytes: %llu\n" | |
2533 | "cancelled_write_bytes: %llu\n", | |
7c44319d AB |
2534 | (unsigned long long)acct.rchar, |
2535 | (unsigned long long)acct.wchar, | |
2536 | (unsigned long long)acct.syscr, | |
2537 | (unsigned long long)acct.syscw, | |
2538 | (unsigned long long)acct.read_bytes, | |
2539 | (unsigned long long)acct.write_bytes, | |
2540 | (unsigned long long)acct.cancelled_write_bytes); | |
297c5d92 AR |
2541 | } |
2542 | ||
2543 | static int proc_tid_io_accounting(struct task_struct *task, char *buffer) | |
2544 | { | |
2545 | return do_io_accounting(task, buffer, 0); | |
aba76fdb | 2546 | } |
297c5d92 AR |
2547 | |
2548 | static int proc_tgid_io_accounting(struct task_struct *task, char *buffer) | |
2549 | { | |
2550 | return do_io_accounting(task, buffer, 1); | |
2551 | } | |
2552 | #endif /* CONFIG_TASK_IO_ACCOUNTING */ | |
aba76fdb | 2553 | |
47830723 KC |
2554 | static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns, |
2555 | struct pid *pid, struct task_struct *task) | |
2556 | { | |
2557 | seq_printf(m, "%08x\n", task->personality); | |
2558 | return 0; | |
2559 | } | |
2560 | ||
28a6d671 EB |
2561 | /* |
2562 | * Thread groups | |
2563 | */ | |
00977a59 | 2564 | static const struct file_operations proc_task_operations; |
c5ef1c42 | 2565 | static const struct inode_operations proc_task_inode_operations; |
20cdc894 | 2566 | |
c5141e6d | 2567 | static const struct pid_entry tgid_base_stuff[] = { |
631f9c18 AD |
2568 | DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations), |
2569 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), | |
2570 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), | |
b2211a36 | 2571 | #ifdef CONFIG_NET |
631f9c18 | 2572 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), |
b2211a36 | 2573 | #endif |
631f9c18 AD |
2574 | REG("environ", S_IRUSR, proc_environ_operations), |
2575 | INF("auxv", S_IRUSR, proc_pid_auxv), | |
2576 | ONE("status", S_IRUGO, proc_pid_status), | |
2577 | ONE("personality", S_IRUSR, proc_pid_personality), | |
2578 | INF("limits", S_IRUSR, proc_pid_limits), | |
43ae34cb | 2579 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 2580 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
ebcb6734 | 2581 | #endif |
4614a696 | 2582 | REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations), |
ebcb6734 | 2583 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
631f9c18 | 2584 | INF("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 2585 | #endif |
631f9c18 AD |
2586 | INF("cmdline", S_IRUGO, proc_pid_cmdline), |
2587 | ONE("stat", S_IRUGO, proc_tgid_stat), | |
2588 | ONE("statm", S_IRUGO, proc_pid_statm), | |
2589 | REG("maps", S_IRUGO, proc_maps_operations), | |
28a6d671 | 2590 | #ifdef CONFIG_NUMA |
631f9c18 | 2591 | REG("numa_maps", S_IRUGO, proc_numa_maps_operations), |
28a6d671 | 2592 | #endif |
631f9c18 AD |
2593 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
2594 | LNK("cwd", proc_cwd_link), | |
2595 | LNK("root", proc_root_link), | |
2596 | LNK("exe", proc_exe_link), | |
2597 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
2598 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
2599 | REG("mountstats", S_IRUSR, proc_mountstats_operations), | |
1e883281 | 2600 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 AD |
2601 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
2602 | REG("smaps", S_IRUGO, proc_smaps_operations), | |
2603 | REG("pagemap", S_IRUSR, proc_pagemap_operations), | |
28a6d671 EB |
2604 | #endif |
2605 | #ifdef CONFIG_SECURITY | |
631f9c18 | 2606 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
2607 | #endif |
2608 | #ifdef CONFIG_KALLSYMS | |
631f9c18 | 2609 | INF("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 2610 | #endif |
2ec220e2 KC |
2611 | #ifdef CONFIG_STACKTRACE |
2612 | ONE("stack", S_IRUSR, proc_pid_stack), | |
28a6d671 EB |
2613 | #endif |
2614 | #ifdef CONFIG_SCHEDSTATS | |
631f9c18 | 2615 | INF("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 2616 | #endif |
9745512c | 2617 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 2618 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 2619 | #endif |
8793d854 | 2620 | #ifdef CONFIG_PROC_PID_CPUSET |
631f9c18 | 2621 | REG("cpuset", S_IRUGO, proc_cpuset_operations), |
a424316c PM |
2622 | #endif |
2623 | #ifdef CONFIG_CGROUPS | |
631f9c18 | 2624 | REG("cgroup", S_IRUGO, proc_cgroup_operations), |
28a6d671 | 2625 | #endif |
631f9c18 AD |
2626 | INF("oom_score", S_IRUGO, proc_oom_score), |
2627 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations), | |
28a6d671 | 2628 | #ifdef CONFIG_AUDITSYSCALL |
631f9c18 AD |
2629 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
2630 | REG("sessionid", S_IRUGO, proc_sessionid_operations), | |
28a6d671 | 2631 | #endif |
f4f154fd | 2632 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 2633 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
f4f154fd | 2634 | #endif |
698ba7b5 | 2635 | #ifdef CONFIG_ELF_CORE |
631f9c18 | 2636 | REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations), |
3cb4a0bb | 2637 | #endif |
aba76fdb | 2638 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
631f9c18 | 2639 | INF("io", S_IRUGO, proc_tgid_io_accounting), |
aba76fdb | 2640 | #endif |
28a6d671 | 2641 | }; |
1da177e4 | 2642 | |
28a6d671 | 2643 | static int proc_tgid_base_readdir(struct file * filp, |
1da177e4 LT |
2644 | void * dirent, filldir_t filldir) |
2645 | { | |
2646 | return proc_pident_readdir(filp,dirent,filldir, | |
28a6d671 | 2647 | tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff)); |
1da177e4 LT |
2648 | } |
2649 | ||
00977a59 | 2650 | static const struct file_operations proc_tgid_base_operations = { |
1da177e4 | 2651 | .read = generic_read_dir, |
28a6d671 | 2652 | .readdir = proc_tgid_base_readdir, |
1da177e4 LT |
2653 | }; |
2654 | ||
28a6d671 | 2655 | static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){ |
7bcd6b0e EB |
2656 | return proc_pident_lookup(dir, dentry, |
2657 | tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
2658 | } |
2659 | ||
c5ef1c42 | 2660 | static const struct inode_operations proc_tgid_base_inode_operations = { |
28a6d671 | 2661 | .lookup = proc_tgid_base_lookup, |
99f89551 | 2662 | .getattr = pid_getattr, |
6d76fa58 | 2663 | .setattr = proc_setattr, |
1da177e4 | 2664 | }; |
1da177e4 | 2665 | |
60347f67 | 2666 | static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid) |
1da177e4 | 2667 | { |
48e6484d | 2668 | struct dentry *dentry, *leader, *dir; |
8578cea7 | 2669 | char buf[PROC_NUMBUF]; |
48e6484d EB |
2670 | struct qstr name; |
2671 | ||
2672 | name.name = buf; | |
60347f67 PE |
2673 | name.len = snprintf(buf, sizeof(buf), "%d", pid); |
2674 | dentry = d_hash_and_lookup(mnt->mnt_root, &name); | |
48e6484d | 2675 | if (dentry) { |
29f12ca3 | 2676 | shrink_dcache_parent(dentry); |
48e6484d EB |
2677 | d_drop(dentry); |
2678 | dput(dentry); | |
2679 | } | |
1da177e4 | 2680 | |
48e6484d | 2681 | name.name = buf; |
60347f67 PE |
2682 | name.len = snprintf(buf, sizeof(buf), "%d", tgid); |
2683 | leader = d_hash_and_lookup(mnt->mnt_root, &name); | |
48e6484d EB |
2684 | if (!leader) |
2685 | goto out; | |
1da177e4 | 2686 | |
48e6484d EB |
2687 | name.name = "task"; |
2688 | name.len = strlen(name.name); | |
2689 | dir = d_hash_and_lookup(leader, &name); | |
2690 | if (!dir) | |
2691 | goto out_put_leader; | |
2692 | ||
2693 | name.name = buf; | |
60347f67 | 2694 | name.len = snprintf(buf, sizeof(buf), "%d", pid); |
48e6484d EB |
2695 | dentry = d_hash_and_lookup(dir, &name); |
2696 | if (dentry) { | |
2697 | shrink_dcache_parent(dentry); | |
2698 | d_drop(dentry); | |
2699 | dput(dentry); | |
1da177e4 | 2700 | } |
48e6484d EB |
2701 | |
2702 | dput(dir); | |
2703 | out_put_leader: | |
2704 | dput(leader); | |
2705 | out: | |
2706 | return; | |
1da177e4 LT |
2707 | } |
2708 | ||
0895e91d RD |
2709 | /** |
2710 | * proc_flush_task - Remove dcache entries for @task from the /proc dcache. | |
2711 | * @task: task that should be flushed. | |
2712 | * | |
2713 | * When flushing dentries from proc, one needs to flush them from global | |
60347f67 | 2714 | * proc (proc_mnt) and from all the namespaces' procs this task was seen |
0895e91d RD |
2715 | * in. This call is supposed to do all of this job. |
2716 | * | |
2717 | * Looks in the dcache for | |
2718 | * /proc/@pid | |
2719 | * /proc/@tgid/task/@pid | |
2720 | * if either directory is present flushes it and all of it'ts children | |
2721 | * from the dcache. | |
2722 | * | |
2723 | * It is safe and reasonable to cache /proc entries for a task until | |
2724 | * that task exits. After that they just clog up the dcache with | |
2725 | * useless entries, possibly causing useful dcache entries to be | |
2726 | * flushed instead. This routine is proved to flush those useless | |
2727 | * dcache entries at process exit time. | |
2728 | * | |
2729 | * NOTE: This routine is just an optimization so it does not guarantee | |
2730 | * that no dcache entries will exist at process exit time it | |
2731 | * just makes it very unlikely that any will persist. | |
60347f67 PE |
2732 | */ |
2733 | ||
2734 | void proc_flush_task(struct task_struct *task) | |
2735 | { | |
9fcc2d15 | 2736 | int i; |
9b4d1cbe | 2737 | struct pid *pid, *tgid; |
130f77ec PE |
2738 | struct upid *upid; |
2739 | ||
130f77ec | 2740 | pid = task_pid(task); |
9b4d1cbe | 2741 | tgid = task_tgid(task); |
130f77ec | 2742 | |
9fcc2d15 | 2743 | for (i = 0; i <= pid->level; i++) { |
130f77ec PE |
2744 | upid = &pid->numbers[i]; |
2745 | proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr, | |
9b4d1cbe | 2746 | tgid->numbers[i].nr); |
130f77ec | 2747 | } |
6f4e6433 PE |
2748 | |
2749 | upid = &pid->numbers[pid->level]; | |
2750 | if (upid->nr == 1) | |
2751 | pid_ns_release_proc(upid->ns); | |
60347f67 PE |
2752 | } |
2753 | ||
9711ef99 AB |
2754 | static struct dentry *proc_pid_instantiate(struct inode *dir, |
2755 | struct dentry * dentry, | |
c5141e6d | 2756 | struct task_struct *task, const void *ptr) |
444ceed8 EB |
2757 | { |
2758 | struct dentry *error = ERR_PTR(-ENOENT); | |
2759 | struct inode *inode; | |
2760 | ||
61a28784 | 2761 | inode = proc_pid_make_inode(dir->i_sb, task); |
444ceed8 EB |
2762 | if (!inode) |
2763 | goto out; | |
2764 | ||
2765 | inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO; | |
2766 | inode->i_op = &proc_tgid_base_inode_operations; | |
2767 | inode->i_fop = &proc_tgid_base_operations; | |
2768 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 VN |
2769 | |
2770 | inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff, | |
2771 | ARRAY_SIZE(tgid_base_stuff)); | |
444ceed8 EB |
2772 | |
2773 | dentry->d_op = &pid_dentry_operations; | |
2774 | ||
2775 | d_add(dentry, inode); | |
2776 | /* Close the race of the process dying before we return the dentry */ | |
2777 | if (pid_revalidate(dentry, NULL)) | |
2778 | error = NULL; | |
2779 | out: | |
2780 | return error; | |
2781 | } | |
2782 | ||
1da177e4 LT |
2783 | struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd) |
2784 | { | |
73d36460 | 2785 | struct dentry *result; |
1da177e4 | 2786 | struct task_struct *task; |
1da177e4 | 2787 | unsigned tgid; |
b488893a | 2788 | struct pid_namespace *ns; |
1da177e4 | 2789 | |
801199ce EB |
2790 | result = proc_base_lookup(dir, dentry); |
2791 | if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT) | |
2792 | goto out; | |
2793 | ||
1da177e4 LT |
2794 | tgid = name_to_int(dentry); |
2795 | if (tgid == ~0U) | |
2796 | goto out; | |
2797 | ||
b488893a | 2798 | ns = dentry->d_sb->s_fs_info; |
de758734 | 2799 | rcu_read_lock(); |
b488893a | 2800 | task = find_task_by_pid_ns(tgid, ns); |
1da177e4 LT |
2801 | if (task) |
2802 | get_task_struct(task); | |
de758734 | 2803 | rcu_read_unlock(); |
1da177e4 LT |
2804 | if (!task) |
2805 | goto out; | |
2806 | ||
444ceed8 | 2807 | result = proc_pid_instantiate(dir, dentry, task, NULL); |
1da177e4 | 2808 | put_task_struct(task); |
1da177e4 | 2809 | out: |
cd6a3ce9 | 2810 | return result; |
1da177e4 LT |
2811 | } |
2812 | ||
1da177e4 | 2813 | /* |
0804ef4b | 2814 | * Find the first task with tgid >= tgid |
0bc58a91 | 2815 | * |
1da177e4 | 2816 | */ |
19fd4bb2 EB |
2817 | struct tgid_iter { |
2818 | unsigned int tgid; | |
0804ef4b | 2819 | struct task_struct *task; |
19fd4bb2 EB |
2820 | }; |
2821 | static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter) | |
2822 | { | |
0804ef4b | 2823 | struct pid *pid; |
1da177e4 | 2824 | |
19fd4bb2 EB |
2825 | if (iter.task) |
2826 | put_task_struct(iter.task); | |
454cc105 | 2827 | rcu_read_lock(); |
0804ef4b | 2828 | retry: |
19fd4bb2 EB |
2829 | iter.task = NULL; |
2830 | pid = find_ge_pid(iter.tgid, ns); | |
0804ef4b | 2831 | if (pid) { |
19fd4bb2 EB |
2832 | iter.tgid = pid_nr_ns(pid, ns); |
2833 | iter.task = pid_task(pid, PIDTYPE_PID); | |
0804ef4b EB |
2834 | /* What we to know is if the pid we have find is the |
2835 | * pid of a thread_group_leader. Testing for task | |
2836 | * being a thread_group_leader is the obvious thing | |
2837 | * todo but there is a window when it fails, due to | |
2838 | * the pid transfer logic in de_thread. | |
2839 | * | |
2840 | * So we perform the straight forward test of seeing | |
2841 | * if the pid we have found is the pid of a thread | |
2842 | * group leader, and don't worry if the task we have | |
2843 | * found doesn't happen to be a thread group leader. | |
2844 | * As we don't care in the case of readdir. | |
2845 | */ | |
19fd4bb2 EB |
2846 | if (!iter.task || !has_group_leader_pid(iter.task)) { |
2847 | iter.tgid += 1; | |
0804ef4b | 2848 | goto retry; |
19fd4bb2 EB |
2849 | } |
2850 | get_task_struct(iter.task); | |
0bc58a91 | 2851 | } |
454cc105 | 2852 | rcu_read_unlock(); |
19fd4bb2 | 2853 | return iter; |
1da177e4 LT |
2854 | } |
2855 | ||
7bcd6b0e | 2856 | #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff)) |
0804ef4b | 2857 | |
61a28784 | 2858 | static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir, |
19fd4bb2 | 2859 | struct tgid_iter iter) |
61a28784 EB |
2860 | { |
2861 | char name[PROC_NUMBUF]; | |
19fd4bb2 | 2862 | int len = snprintf(name, sizeof(name), "%d", iter.tgid); |
61a28784 | 2863 | return proc_fill_cache(filp, dirent, filldir, name, len, |
19fd4bb2 | 2864 | proc_pid_instantiate, iter.task, NULL); |
61a28784 EB |
2865 | } |
2866 | ||
1da177e4 LT |
2867 | /* for the /proc/ directory itself, after non-process stuff has been done */ |
2868 | int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir) | |
2869 | { | |
1da177e4 | 2870 | unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY; |
2fddfeef | 2871 | struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode); |
19fd4bb2 | 2872 | struct tgid_iter iter; |
b488893a | 2873 | struct pid_namespace *ns; |
1da177e4 | 2874 | |
61a28784 EB |
2875 | if (!reaper) |
2876 | goto out_no_task; | |
2877 | ||
7bcd6b0e | 2878 | for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) { |
c5141e6d | 2879 | const struct pid_entry *p = &proc_base_stuff[nr]; |
61a28784 | 2880 | if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0) |
801199ce | 2881 | goto out; |
1da177e4 LT |
2882 | } |
2883 | ||
b488893a | 2884 | ns = filp->f_dentry->d_sb->s_fs_info; |
19fd4bb2 EB |
2885 | iter.task = NULL; |
2886 | iter.tgid = filp->f_pos - TGID_OFFSET; | |
2887 | for (iter = next_tgid(ns, iter); | |
2888 | iter.task; | |
2889 | iter.tgid += 1, iter = next_tgid(ns, iter)) { | |
2890 | filp->f_pos = iter.tgid + TGID_OFFSET; | |
2891 | if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) { | |
2892 | put_task_struct(iter.task); | |
0804ef4b | 2893 | goto out; |
1da177e4 | 2894 | } |
0bc58a91 | 2895 | } |
0804ef4b EB |
2896 | filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET; |
2897 | out: | |
61a28784 EB |
2898 | put_task_struct(reaper); |
2899 | out_no_task: | |
0bc58a91 EB |
2900 | return 0; |
2901 | } | |
1da177e4 | 2902 | |
28a6d671 EB |
2903 | /* |
2904 | * Tasks | |
2905 | */ | |
c5141e6d | 2906 | static const struct pid_entry tid_base_stuff[] = { |
631f9c18 | 2907 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), |
3835541d | 2908 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
631f9c18 AD |
2909 | REG("environ", S_IRUSR, proc_environ_operations), |
2910 | INF("auxv", S_IRUSR, proc_pid_auxv), | |
2911 | ONE("status", S_IRUGO, proc_pid_status), | |
2912 | ONE("personality", S_IRUSR, proc_pid_personality), | |
2913 | INF("limits", S_IRUSR, proc_pid_limits), | |
43ae34cb | 2914 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 2915 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
ebcb6734 | 2916 | #endif |
4614a696 | 2917 | REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations), |
ebcb6734 | 2918 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
631f9c18 | 2919 | INF("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 2920 | #endif |
631f9c18 AD |
2921 | INF("cmdline", S_IRUGO, proc_pid_cmdline), |
2922 | ONE("stat", S_IRUGO, proc_tid_stat), | |
2923 | ONE("statm", S_IRUGO, proc_pid_statm), | |
2924 | REG("maps", S_IRUGO, proc_maps_operations), | |
28a6d671 | 2925 | #ifdef CONFIG_NUMA |
631f9c18 | 2926 | REG("numa_maps", S_IRUGO, proc_numa_maps_operations), |
28a6d671 | 2927 | #endif |
631f9c18 AD |
2928 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
2929 | LNK("cwd", proc_cwd_link), | |
2930 | LNK("root", proc_root_link), | |
2931 | LNK("exe", proc_exe_link), | |
2932 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
2933 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
1e883281 | 2934 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 AD |
2935 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
2936 | REG("smaps", S_IRUGO, proc_smaps_operations), | |
2937 | REG("pagemap", S_IRUSR, proc_pagemap_operations), | |
28a6d671 EB |
2938 | #endif |
2939 | #ifdef CONFIG_SECURITY | |
631f9c18 | 2940 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
2941 | #endif |
2942 | #ifdef CONFIG_KALLSYMS | |
631f9c18 | 2943 | INF("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 2944 | #endif |
2ec220e2 KC |
2945 | #ifdef CONFIG_STACKTRACE |
2946 | ONE("stack", S_IRUSR, proc_pid_stack), | |
28a6d671 EB |
2947 | #endif |
2948 | #ifdef CONFIG_SCHEDSTATS | |
631f9c18 | 2949 | INF("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 2950 | #endif |
9745512c | 2951 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 2952 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 2953 | #endif |
8793d854 | 2954 | #ifdef CONFIG_PROC_PID_CPUSET |
631f9c18 | 2955 | REG("cpuset", S_IRUGO, proc_cpuset_operations), |
a424316c PM |
2956 | #endif |
2957 | #ifdef CONFIG_CGROUPS | |
631f9c18 | 2958 | REG("cgroup", S_IRUGO, proc_cgroup_operations), |
28a6d671 | 2959 | #endif |
631f9c18 AD |
2960 | INF("oom_score", S_IRUGO, proc_oom_score), |
2961 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations), | |
28a6d671 | 2962 | #ifdef CONFIG_AUDITSYSCALL |
631f9c18 AD |
2963 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
2964 | REG("sessionid", S_IRUSR, proc_sessionid_operations), | |
28a6d671 | 2965 | #endif |
f4f154fd | 2966 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 2967 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
f4f154fd | 2968 | #endif |
297c5d92 | 2969 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
631f9c18 | 2970 | INF("io", S_IRUGO, proc_tid_io_accounting), |
297c5d92 | 2971 | #endif |
28a6d671 EB |
2972 | }; |
2973 | ||
2974 | static int proc_tid_base_readdir(struct file * filp, | |
2975 | void * dirent, filldir_t filldir) | |
2976 | { | |
2977 | return proc_pident_readdir(filp,dirent,filldir, | |
2978 | tid_base_stuff,ARRAY_SIZE(tid_base_stuff)); | |
2979 | } | |
2980 | ||
2981 | static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){ | |
7bcd6b0e EB |
2982 | return proc_pident_lookup(dir, dentry, |
2983 | tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
2984 | } |
2985 | ||
00977a59 | 2986 | static const struct file_operations proc_tid_base_operations = { |
28a6d671 EB |
2987 | .read = generic_read_dir, |
2988 | .readdir = proc_tid_base_readdir, | |
2989 | }; | |
2990 | ||
c5ef1c42 | 2991 | static const struct inode_operations proc_tid_base_inode_operations = { |
28a6d671 EB |
2992 | .lookup = proc_tid_base_lookup, |
2993 | .getattr = pid_getattr, | |
2994 | .setattr = proc_setattr, | |
2995 | }; | |
2996 | ||
444ceed8 | 2997 | static struct dentry *proc_task_instantiate(struct inode *dir, |
c5141e6d | 2998 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
444ceed8 EB |
2999 | { |
3000 | struct dentry *error = ERR_PTR(-ENOENT); | |
3001 | struct inode *inode; | |
61a28784 | 3002 | inode = proc_pid_make_inode(dir->i_sb, task); |
444ceed8 EB |
3003 | |
3004 | if (!inode) | |
3005 | goto out; | |
3006 | inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO; | |
3007 | inode->i_op = &proc_tid_base_inode_operations; | |
3008 | inode->i_fop = &proc_tid_base_operations; | |
3009 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 VN |
3010 | |
3011 | inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff, | |
3012 | ARRAY_SIZE(tid_base_stuff)); | |
444ceed8 EB |
3013 | |
3014 | dentry->d_op = &pid_dentry_operations; | |
3015 | ||
3016 | d_add(dentry, inode); | |
3017 | /* Close the race of the process dying before we return the dentry */ | |
3018 | if (pid_revalidate(dentry, NULL)) | |
3019 | error = NULL; | |
3020 | out: | |
3021 | return error; | |
3022 | } | |
3023 | ||
28a6d671 EB |
3024 | static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd) |
3025 | { | |
3026 | struct dentry *result = ERR_PTR(-ENOENT); | |
3027 | struct task_struct *task; | |
3028 | struct task_struct *leader = get_proc_task(dir); | |
28a6d671 | 3029 | unsigned tid; |
b488893a | 3030 | struct pid_namespace *ns; |
28a6d671 EB |
3031 | |
3032 | if (!leader) | |
3033 | goto out_no_task; | |
3034 | ||
3035 | tid = name_to_int(dentry); | |
3036 | if (tid == ~0U) | |
3037 | goto out; | |
3038 | ||
b488893a | 3039 | ns = dentry->d_sb->s_fs_info; |
28a6d671 | 3040 | rcu_read_lock(); |
b488893a | 3041 | task = find_task_by_pid_ns(tid, ns); |
28a6d671 EB |
3042 | if (task) |
3043 | get_task_struct(task); | |
3044 | rcu_read_unlock(); | |
3045 | if (!task) | |
3046 | goto out; | |
bac0abd6 | 3047 | if (!same_thread_group(leader, task)) |
28a6d671 EB |
3048 | goto out_drop_task; |
3049 | ||
444ceed8 | 3050 | result = proc_task_instantiate(dir, dentry, task, NULL); |
28a6d671 EB |
3051 | out_drop_task: |
3052 | put_task_struct(task); | |
3053 | out: | |
3054 | put_task_struct(leader); | |
3055 | out_no_task: | |
3056 | return result; | |
3057 | } | |
3058 | ||
0bc58a91 EB |
3059 | /* |
3060 | * Find the first tid of a thread group to return to user space. | |
3061 | * | |
3062 | * Usually this is just the thread group leader, but if the users | |
3063 | * buffer was too small or there was a seek into the middle of the | |
3064 | * directory we have more work todo. | |
3065 | * | |
3066 | * In the case of a short read we start with find_task_by_pid. | |
3067 | * | |
3068 | * In the case of a seek we start with the leader and walk nr | |
3069 | * threads past it. | |
3070 | */ | |
cc288738 | 3071 | static struct task_struct *first_tid(struct task_struct *leader, |
b488893a | 3072 | int tid, int nr, struct pid_namespace *ns) |
0bc58a91 | 3073 | { |
a872ff0c | 3074 | struct task_struct *pos; |
1da177e4 | 3075 | |
cc288738 | 3076 | rcu_read_lock(); |
0bc58a91 EB |
3077 | /* Attempt to start with the pid of a thread */ |
3078 | if (tid && (nr > 0)) { | |
b488893a | 3079 | pos = find_task_by_pid_ns(tid, ns); |
a872ff0c ON |
3080 | if (pos && (pos->group_leader == leader)) |
3081 | goto found; | |
0bc58a91 | 3082 | } |
1da177e4 | 3083 | |
0bc58a91 | 3084 | /* If nr exceeds the number of threads there is nothing todo */ |
a872ff0c ON |
3085 | pos = NULL; |
3086 | if (nr && nr >= get_nr_threads(leader)) | |
3087 | goto out; | |
1da177e4 | 3088 | |
a872ff0c ON |
3089 | /* If we haven't found our starting place yet start |
3090 | * with the leader and walk nr threads forward. | |
0bc58a91 | 3091 | */ |
a872ff0c ON |
3092 | for (pos = leader; nr > 0; --nr) { |
3093 | pos = next_thread(pos); | |
3094 | if (pos == leader) { | |
3095 | pos = NULL; | |
3096 | goto out; | |
3097 | } | |
1da177e4 | 3098 | } |
a872ff0c ON |
3099 | found: |
3100 | get_task_struct(pos); | |
3101 | out: | |
cc288738 | 3102 | rcu_read_unlock(); |
0bc58a91 EB |
3103 | return pos; |
3104 | } | |
3105 | ||
3106 | /* | |
3107 | * Find the next thread in the thread list. | |
3108 | * Return NULL if there is an error or no next thread. | |
3109 | * | |
3110 | * The reference to the input task_struct is released. | |
3111 | */ | |
3112 | static struct task_struct *next_tid(struct task_struct *start) | |
3113 | { | |
c1df7fb8 | 3114 | struct task_struct *pos = NULL; |
cc288738 | 3115 | rcu_read_lock(); |
c1df7fb8 | 3116 | if (pid_alive(start)) { |
0bc58a91 | 3117 | pos = next_thread(start); |
c1df7fb8 ON |
3118 | if (thread_group_leader(pos)) |
3119 | pos = NULL; | |
3120 | else | |
3121 | get_task_struct(pos); | |
3122 | } | |
cc288738 | 3123 | rcu_read_unlock(); |
0bc58a91 EB |
3124 | put_task_struct(start); |
3125 | return pos; | |
1da177e4 LT |
3126 | } |
3127 | ||
61a28784 EB |
3128 | static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir, |
3129 | struct task_struct *task, int tid) | |
3130 | { | |
3131 | char name[PROC_NUMBUF]; | |
3132 | int len = snprintf(name, sizeof(name), "%d", tid); | |
3133 | return proc_fill_cache(filp, dirent, filldir, name, len, | |
3134 | proc_task_instantiate, task, NULL); | |
3135 | } | |
3136 | ||
1da177e4 LT |
3137 | /* for the /proc/TGID/task/ directories */ |
3138 | static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir) | |
3139 | { | |
2fddfeef | 3140 | struct dentry *dentry = filp->f_path.dentry; |
1da177e4 | 3141 | struct inode *inode = dentry->d_inode; |
7d895244 | 3142 | struct task_struct *leader = NULL; |
0bc58a91 | 3143 | struct task_struct *task; |
1da177e4 LT |
3144 | int retval = -ENOENT; |
3145 | ino_t ino; | |
0bc58a91 | 3146 | int tid; |
b488893a | 3147 | struct pid_namespace *ns; |
1da177e4 | 3148 | |
7d895244 GC |
3149 | task = get_proc_task(inode); |
3150 | if (!task) | |
3151 | goto out_no_task; | |
3152 | rcu_read_lock(); | |
3153 | if (pid_alive(task)) { | |
3154 | leader = task->group_leader; | |
3155 | get_task_struct(leader); | |
3156 | } | |
3157 | rcu_read_unlock(); | |
3158 | put_task_struct(task); | |
99f89551 EB |
3159 | if (!leader) |
3160 | goto out_no_task; | |
1da177e4 LT |
3161 | retval = 0; |
3162 | ||
ee568b25 | 3163 | switch ((unsigned long)filp->f_pos) { |
1da177e4 LT |
3164 | case 0: |
3165 | ino = inode->i_ino; | |
ee6f779b | 3166 | if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0) |
1da177e4 | 3167 | goto out; |
ee6f779b | 3168 | filp->f_pos++; |
1da177e4 LT |
3169 | /* fall through */ |
3170 | case 1: | |
3171 | ino = parent_ino(dentry); | |
ee6f779b | 3172 | if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0) |
1da177e4 | 3173 | goto out; |
ee6f779b | 3174 | filp->f_pos++; |
1da177e4 LT |
3175 | /* fall through */ |
3176 | } | |
3177 | ||
0bc58a91 EB |
3178 | /* f_version caches the tgid value that the last readdir call couldn't |
3179 | * return. lseek aka telldir automagically resets f_version to 0. | |
3180 | */ | |
b488893a | 3181 | ns = filp->f_dentry->d_sb->s_fs_info; |
2b47c361 | 3182 | tid = (int)filp->f_version; |
0bc58a91 | 3183 | filp->f_version = 0; |
ee6f779b | 3184 | for (task = first_tid(leader, tid, filp->f_pos - 2, ns); |
0bc58a91 | 3185 | task; |
ee6f779b | 3186 | task = next_tid(task), filp->f_pos++) { |
b488893a | 3187 | tid = task_pid_nr_ns(task, ns); |
61a28784 | 3188 | if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) { |
0bc58a91 EB |
3189 | /* returning this tgid failed, save it as the first |
3190 | * pid for the next readir call */ | |
2b47c361 | 3191 | filp->f_version = (u64)tid; |
0bc58a91 | 3192 | put_task_struct(task); |
1da177e4 | 3193 | break; |
0bc58a91 | 3194 | } |
1da177e4 LT |
3195 | } |
3196 | out: | |
99f89551 EB |
3197 | put_task_struct(leader); |
3198 | out_no_task: | |
1da177e4 LT |
3199 | return retval; |
3200 | } | |
6e66b52b EB |
3201 | |
3202 | static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat) | |
3203 | { | |
3204 | struct inode *inode = dentry->d_inode; | |
99f89551 | 3205 | struct task_struct *p = get_proc_task(inode); |
6e66b52b EB |
3206 | generic_fillattr(inode, stat); |
3207 | ||
99f89551 | 3208 | if (p) { |
99f89551 | 3209 | stat->nlink += get_nr_threads(p); |
99f89551 | 3210 | put_task_struct(p); |
6e66b52b EB |
3211 | } |
3212 | ||
3213 | return 0; | |
3214 | } | |
28a6d671 | 3215 | |
c5ef1c42 | 3216 | static const struct inode_operations proc_task_inode_operations = { |
28a6d671 EB |
3217 | .lookup = proc_task_lookup, |
3218 | .getattr = proc_task_getattr, | |
3219 | .setattr = proc_setattr, | |
3220 | }; | |
3221 | ||
00977a59 | 3222 | static const struct file_operations proc_task_operations = { |
28a6d671 EB |
3223 | .read = generic_read_dir, |
3224 | .readdir = proc_task_readdir, | |
3225 | }; |