]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/proc/base.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | * | |
6 | * proc base directory handling functions | |
7 | * | |
8 | * 1999, Al Viro. Rewritten. Now it covers the whole per-process part. | |
9 | * Instead of using magical inumbers to determine the kind of object | |
10 | * we allocate and fill in-core inodes upon lookup. They don't even | |
11 | * go into icache. We cache the reference to task_struct upon lookup too. | |
12 | * Eventually it should become a filesystem in its own. We don't use the | |
13 | * rest of procfs anymore. | |
e070ad49 ML |
14 | * |
15 | * | |
16 | * Changelog: | |
17 | * 17-Jan-2005 | |
18 | * Allan Bezerra | |
19 | * Bruna Moreira <bruna.moreira@indt.org.br> | |
20 | * Edjard Mota <edjard.mota@indt.org.br> | |
21 | * Ilias Biris <ilias.biris@indt.org.br> | |
22 | * Mauricio Lin <mauricio.lin@indt.org.br> | |
23 | * | |
24 | * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT | |
25 | * | |
26 | * A new process specific entry (smaps) included in /proc. It shows the | |
27 | * size of rss for each memory area. The maps entry lacks information | |
28 | * about physical memory size (rss) for each mapped file, i.e., | |
29 | * rss information for executables and library files. | |
30 | * This additional information is useful for any tools that need to know | |
31 | * about physical memory consumption for a process specific library. | |
32 | * | |
33 | * Changelog: | |
34 | * 21-Feb-2005 | |
35 | * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT | |
36 | * Pud inclusion in the page table walking. | |
37 | * | |
38 | * ChangeLog: | |
39 | * 10-Mar-2005 | |
40 | * 10LE Instituto Nokia de Tecnologia - INdT: | |
41 | * A better way to walks through the page table as suggested by Hugh Dickins. | |
42 | * | |
43 | * Simo Piiroinen <simo.piiroinen@nokia.com>: | |
44 | * Smaps information related to shared, private, clean and dirty pages. | |
45 | * | |
46 | * Paul Mundt <paul.mundt@nokia.com>: | |
47 | * Overall revision about smaps. | |
1da177e4 LT |
48 | */ |
49 | ||
50 | #include <asm/uaccess.h> | |
51 | ||
1da177e4 LT |
52 | #include <linux/errno.h> |
53 | #include <linux/time.h> | |
54 | #include <linux/proc_fs.h> | |
55 | #include <linux/stat.h> | |
5995477a | 56 | #include <linux/task_io_accounting_ops.h> |
1da177e4 | 57 | #include <linux/init.h> |
16f7e0fe | 58 | #include <linux/capability.h> |
1da177e4 | 59 | #include <linux/file.h> |
9f3acc31 | 60 | #include <linux/fdtable.h> |
1da177e4 LT |
61 | #include <linux/string.h> |
62 | #include <linux/seq_file.h> | |
63 | #include <linux/namei.h> | |
6b3286ed | 64 | #include <linux/mnt_namespace.h> |
1da177e4 | 65 | #include <linux/mm.h> |
a63d83f4 | 66 | #include <linux/swap.h> |
b835996f | 67 | #include <linux/rcupdate.h> |
1da177e4 | 68 | #include <linux/kallsyms.h> |
2ec220e2 | 69 | #include <linux/stacktrace.h> |
d85f50d5 | 70 | #include <linux/resource.h> |
5096add8 | 71 | #include <linux/module.h> |
1da177e4 LT |
72 | #include <linux/mount.h> |
73 | #include <linux/security.h> | |
74 | #include <linux/ptrace.h> | |
0d094efe | 75 | #include <linux/tracehook.h> |
87ebdc00 | 76 | #include <linux/printk.h> |
a424316c | 77 | #include <linux/cgroup.h> |
1da177e4 LT |
78 | #include <linux/cpuset.h> |
79 | #include <linux/audit.h> | |
5addc5dd | 80 | #include <linux/poll.h> |
1651e14e | 81 | #include <linux/nsproxy.h> |
8ac773b4 | 82 | #include <linux/oom.h> |
3cb4a0bb | 83 | #include <linux/elf.h> |
60347f67 | 84 | #include <linux/pid_namespace.h> |
22d917d8 | 85 | #include <linux/user_namespace.h> |
5ad4e53b | 86 | #include <linux/fs_struct.h> |
5a0e3ad6 | 87 | #include <linux/slab.h> |
640708a2 | 88 | #include <linux/flex_array.h> |
48f6a7a5 | 89 | #include <linux/posix-timers.h> |
f133ecca CM |
90 | #ifdef CONFIG_HARDWALL |
91 | #include <asm/hardwall.h> | |
92 | #endif | |
43d2b113 | 93 | #include <trace/events/oom.h> |
1da177e4 | 94 | #include "internal.h" |
faf60af1 | 95 | #include "fd.h" |
1da177e4 | 96 | |
0f2fe20f EB |
97 | /* NOTE: |
98 | * Implementing inode permission operations in /proc is almost | |
99 | * certainly an error. Permission checks need to happen during | |
100 | * each system call not at open time. The reason is that most of | |
101 | * what we wish to check for permissions in /proc varies at runtime. | |
102 | * | |
103 | * The classic example of a problem is opening file descriptors | |
104 | * in /proc for a task before it execs a suid executable. | |
105 | */ | |
106 | ||
1da177e4 | 107 | struct pid_entry { |
cedbccab | 108 | const char *name; |
c5141e6d | 109 | int len; |
d161a13f | 110 | umode_t mode; |
c5ef1c42 | 111 | const struct inode_operations *iop; |
00977a59 | 112 | const struct file_operations *fop; |
20cdc894 | 113 | union proc_op op; |
1da177e4 LT |
114 | }; |
115 | ||
61a28784 | 116 | #define NOD(NAME, MODE, IOP, FOP, OP) { \ |
20cdc894 | 117 | .name = (NAME), \ |
c5141e6d | 118 | .len = sizeof(NAME) - 1, \ |
20cdc894 EB |
119 | .mode = MODE, \ |
120 | .iop = IOP, \ | |
121 | .fop = FOP, \ | |
122 | .op = OP, \ | |
123 | } | |
124 | ||
631f9c18 AD |
125 | #define DIR(NAME, MODE, iops, fops) \ |
126 | NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} ) | |
127 | #define LNK(NAME, get_link) \ | |
61a28784 | 128 | NOD(NAME, (S_IFLNK|S_IRWXUGO), \ |
20cdc894 | 129 | &proc_pid_link_inode_operations, NULL, \ |
631f9c18 AD |
130 | { .proc_get_link = get_link } ) |
131 | #define REG(NAME, MODE, fops) \ | |
132 | NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {}) | |
133 | #define INF(NAME, MODE, read) \ | |
61a28784 | 134 | NOD(NAME, (S_IFREG|(MODE)), \ |
20cdc894 | 135 | NULL, &proc_info_file_operations, \ |
631f9c18 AD |
136 | { .proc_read = read } ) |
137 | #define ONE(NAME, MODE, show) \ | |
be614086 EB |
138 | NOD(NAME, (S_IFREG|(MODE)), \ |
139 | NULL, &proc_single_file_operations, \ | |
631f9c18 | 140 | { .proc_show = show } ) |
1da177e4 | 141 | |
aed54175 VN |
142 | /* |
143 | * Count the number of hardlinks for the pid_entry table, excluding the . | |
144 | * and .. links. | |
145 | */ | |
146 | static unsigned int pid_entry_count_dirs(const struct pid_entry *entries, | |
147 | unsigned int n) | |
148 | { | |
149 | unsigned int i; | |
150 | unsigned int count; | |
151 | ||
152 | count = 0; | |
153 | for (i = 0; i < n; ++i) { | |
154 | if (S_ISDIR(entries[i].mode)) | |
155 | ++count; | |
156 | } | |
157 | ||
158 | return count; | |
159 | } | |
160 | ||
f7ad3c6b | 161 | static int get_task_root(struct task_struct *task, struct path *root) |
1da177e4 | 162 | { |
7c2c7d99 HD |
163 | int result = -ENOENT; |
164 | ||
0494f6ec | 165 | task_lock(task); |
f7ad3c6b MS |
166 | if (task->fs) { |
167 | get_fs_root(task->fs, root); | |
7c2c7d99 HD |
168 | result = 0; |
169 | } | |
0494f6ec | 170 | task_unlock(task); |
7c2c7d99 | 171 | return result; |
0494f6ec MS |
172 | } |
173 | ||
7773fbc5 | 174 | static int proc_cwd_link(struct dentry *dentry, struct path *path) |
0494f6ec | 175 | { |
7773fbc5 | 176 | struct task_struct *task = get_proc_task(dentry->d_inode); |
0494f6ec | 177 | int result = -ENOENT; |
99f89551 EB |
178 | |
179 | if (task) { | |
f7ad3c6b MS |
180 | task_lock(task); |
181 | if (task->fs) { | |
182 | get_fs_pwd(task->fs, path); | |
183 | result = 0; | |
184 | } | |
185 | task_unlock(task); | |
99f89551 EB |
186 | put_task_struct(task); |
187 | } | |
1da177e4 LT |
188 | return result; |
189 | } | |
190 | ||
7773fbc5 | 191 | static int proc_root_link(struct dentry *dentry, struct path *path) |
1da177e4 | 192 | { |
7773fbc5 | 193 | struct task_struct *task = get_proc_task(dentry->d_inode); |
1da177e4 | 194 | int result = -ENOENT; |
99f89551 EB |
195 | |
196 | if (task) { | |
f7ad3c6b | 197 | result = get_task_root(task, path); |
99f89551 EB |
198 | put_task_struct(task); |
199 | } | |
1da177e4 LT |
200 | return result; |
201 | } | |
202 | ||
2ca66ff7 AD |
203 | static int proc_pid_cmdline(struct seq_file *m, struct pid_namespace *ns, |
204 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 205 | { |
2ca66ff7 AD |
206 | /* |
207 | * Rely on struct seq_operations::show() being called once | |
208 | * per internal buffer allocation. See single_open(), traverse(). | |
209 | */ | |
210 | BUG_ON(m->size < PAGE_SIZE); | |
211 | m->count += get_cmdline(task, m->buf, PAGE_SIZE); | |
212 | return 0; | |
1da177e4 LT |
213 | } |
214 | ||
f9ea536e AD |
215 | static int proc_pid_auxv(struct seq_file *m, struct pid_namespace *ns, |
216 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 217 | { |
e7dcd999 | 218 | struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ); |
2fadaef4 | 219 | if (mm && !IS_ERR(mm)) { |
1da177e4 | 220 | unsigned int nwords = 0; |
dfe6b7d9 | 221 | do { |
1da177e4 | 222 | nwords += 2; |
dfe6b7d9 | 223 | } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */ |
f9ea536e | 224 | seq_write(m, mm->saved_auxv, nwords * sizeof(mm->saved_auxv[0])); |
1da177e4 | 225 | mmput(mm); |
f9ea536e AD |
226 | return 0; |
227 | } else | |
228 | return PTR_ERR(mm); | |
1da177e4 LT |
229 | } |
230 | ||
231 | ||
232 | #ifdef CONFIG_KALLSYMS | |
233 | /* | |
234 | * Provides a wchan file via kallsyms in a proper one-value-per-file format. | |
235 | * Returns the resolved symbol. If that fails, simply return the address. | |
236 | */ | |
edfcd606 AD |
237 | static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns, |
238 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 239 | { |
ffb45122 | 240 | unsigned long wchan; |
9281acea | 241 | char symname[KSYM_NAME_LEN]; |
1da177e4 LT |
242 | |
243 | wchan = get_wchan(task); | |
244 | ||
9d65cb4a | 245 | if (lookup_symbol_name(wchan, symname) < 0) |
f83ce3e6 JE |
246 | if (!ptrace_may_access(task, PTRACE_MODE_READ)) |
247 | return 0; | |
248 | else | |
edfcd606 | 249 | return seq_printf(m, "%lu", wchan); |
9d65cb4a | 250 | else |
edfcd606 | 251 | return seq_printf(m, "%s", symname); |
1da177e4 LT |
252 | } |
253 | #endif /* CONFIG_KALLSYMS */ | |
254 | ||
a9712bc1 AV |
255 | static int lock_trace(struct task_struct *task) |
256 | { | |
257 | int err = mutex_lock_killable(&task->signal->cred_guard_mutex); | |
258 | if (err) | |
259 | return err; | |
260 | if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) { | |
261 | mutex_unlock(&task->signal->cred_guard_mutex); | |
262 | return -EPERM; | |
263 | } | |
264 | return 0; | |
265 | } | |
266 | ||
267 | static void unlock_trace(struct task_struct *task) | |
268 | { | |
269 | mutex_unlock(&task->signal->cred_guard_mutex); | |
270 | } | |
271 | ||
2ec220e2 KC |
272 | #ifdef CONFIG_STACKTRACE |
273 | ||
274 | #define MAX_STACK_TRACE_DEPTH 64 | |
275 | ||
276 | static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns, | |
277 | struct pid *pid, struct task_struct *task) | |
278 | { | |
279 | struct stack_trace trace; | |
280 | unsigned long *entries; | |
a9712bc1 | 281 | int err; |
2ec220e2 KC |
282 | int i; |
283 | ||
284 | entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL); | |
285 | if (!entries) | |
286 | return -ENOMEM; | |
287 | ||
288 | trace.nr_entries = 0; | |
289 | trace.max_entries = MAX_STACK_TRACE_DEPTH; | |
290 | trace.entries = entries; | |
291 | trace.skip = 0; | |
2ec220e2 | 292 | |
a9712bc1 AV |
293 | err = lock_trace(task); |
294 | if (!err) { | |
295 | save_stack_trace_tsk(task, &trace); | |
296 | ||
297 | for (i = 0; i < trace.nr_entries; i++) { | |
b81a618d | 298 | seq_printf(m, "[<%pK>] %pS\n", |
a9712bc1 AV |
299 | (void *)entries[i], (void *)entries[i]); |
300 | } | |
301 | unlock_trace(task); | |
2ec220e2 KC |
302 | } |
303 | kfree(entries); | |
304 | ||
a9712bc1 | 305 | return err; |
2ec220e2 KC |
306 | } |
307 | #endif | |
308 | ||
1da177e4 LT |
309 | #ifdef CONFIG_SCHEDSTATS |
310 | /* | |
311 | * Provides /proc/PID/schedstat | |
312 | */ | |
f6e826ca AD |
313 | static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns, |
314 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 315 | { |
f6e826ca | 316 | return seq_printf(m, "%llu %llu %lu\n", |
826e08b0 IM |
317 | (unsigned long long)task->se.sum_exec_runtime, |
318 | (unsigned long long)task->sched_info.run_delay, | |
2d72376b | 319 | task->sched_info.pcount); |
1da177e4 LT |
320 | } |
321 | #endif | |
322 | ||
9745512c AV |
323 | #ifdef CONFIG_LATENCYTOP |
324 | static int lstats_show_proc(struct seq_file *m, void *v) | |
325 | { | |
326 | int i; | |
13d77c37 HS |
327 | struct inode *inode = m->private; |
328 | struct task_struct *task = get_proc_task(inode); | |
9745512c | 329 | |
13d77c37 HS |
330 | if (!task) |
331 | return -ESRCH; | |
332 | seq_puts(m, "Latency Top version : v0.1\n"); | |
9745512c | 333 | for (i = 0; i < 32; i++) { |
34e49d4f JP |
334 | struct latency_record *lr = &task->latency_record[i]; |
335 | if (lr->backtrace[0]) { | |
9745512c | 336 | int q; |
34e49d4f JP |
337 | seq_printf(m, "%i %li %li", |
338 | lr->count, lr->time, lr->max); | |
9745512c | 339 | for (q = 0; q < LT_BACKTRACEDEPTH; q++) { |
34e49d4f JP |
340 | unsigned long bt = lr->backtrace[q]; |
341 | if (!bt) | |
9745512c | 342 | break; |
34e49d4f | 343 | if (bt == ULONG_MAX) |
9745512c | 344 | break; |
34e49d4f | 345 | seq_printf(m, " %ps", (void *)bt); |
9745512c | 346 | } |
9d6de12f | 347 | seq_putc(m, '\n'); |
9745512c AV |
348 | } |
349 | ||
350 | } | |
13d77c37 | 351 | put_task_struct(task); |
9745512c AV |
352 | return 0; |
353 | } | |
354 | ||
355 | static int lstats_open(struct inode *inode, struct file *file) | |
356 | { | |
13d77c37 | 357 | return single_open(file, lstats_show_proc, inode); |
d6643d12 HS |
358 | } |
359 | ||
9745512c AV |
360 | static ssize_t lstats_write(struct file *file, const char __user *buf, |
361 | size_t count, loff_t *offs) | |
362 | { | |
496ad9aa | 363 | struct task_struct *task = get_proc_task(file_inode(file)); |
9745512c | 364 | |
13d77c37 HS |
365 | if (!task) |
366 | return -ESRCH; | |
9745512c | 367 | clear_all_latency_tracing(task); |
13d77c37 | 368 | put_task_struct(task); |
9745512c AV |
369 | |
370 | return count; | |
371 | } | |
372 | ||
373 | static const struct file_operations proc_lstats_operations = { | |
374 | .open = lstats_open, | |
375 | .read = seq_read, | |
376 | .write = lstats_write, | |
377 | .llseek = seq_lseek, | |
13d77c37 | 378 | .release = single_release, |
9745512c AV |
379 | }; |
380 | ||
381 | #endif | |
382 | ||
8d8b97ba AV |
383 | #ifdef CONFIG_CGROUPS |
384 | static int cgroup_open(struct inode *inode, struct file *file) | |
385 | { | |
386 | struct pid *pid = PROC_I(inode)->pid; | |
387 | return single_open(file, proc_cgroup_show, pid); | |
388 | } | |
389 | ||
390 | static const struct file_operations proc_cgroup_operations = { | |
391 | .open = cgroup_open, | |
392 | .read = seq_read, | |
393 | .llseek = seq_lseek, | |
394 | .release = single_release, | |
395 | }; | |
396 | #endif | |
397 | ||
398 | #ifdef CONFIG_PROC_PID_CPUSET | |
399 | ||
400 | static int cpuset_open(struct inode *inode, struct file *file) | |
401 | { | |
402 | struct pid *pid = PROC_I(inode)->pid; | |
403 | return single_open(file, proc_cpuset_show, pid); | |
404 | } | |
405 | ||
406 | static const struct file_operations proc_cpuset_operations = { | |
407 | .open = cpuset_open, | |
408 | .read = seq_read, | |
409 | .llseek = seq_lseek, | |
410 | .release = single_release, | |
411 | }; | |
412 | #endif | |
413 | ||
1da177e4 LT |
414 | static int proc_oom_score(struct task_struct *task, char *buffer) |
415 | { | |
a7f638f9 | 416 | unsigned long totalpages = totalram_pages + total_swap_pages; |
b95c35e7 | 417 | unsigned long points = 0; |
1da177e4 | 418 | |
19c5d45a | 419 | read_lock(&tasklist_lock); |
b95c35e7 | 420 | if (pid_alive(task)) |
a7f638f9 DR |
421 | points = oom_badness(task, NULL, NULL, totalpages) * |
422 | 1000 / totalpages; | |
19c5d45a | 423 | read_unlock(&tasklist_lock); |
1da177e4 LT |
424 | return sprintf(buffer, "%lu\n", points); |
425 | } | |
426 | ||
d85f50d5 | 427 | struct limit_names { |
cedbccab AD |
428 | const char *name; |
429 | const char *unit; | |
d85f50d5 NH |
430 | }; |
431 | ||
432 | static const struct limit_names lnames[RLIM_NLIMITS] = { | |
cff4edb5 | 433 | [RLIMIT_CPU] = {"Max cpu time", "seconds"}, |
d85f50d5 NH |
434 | [RLIMIT_FSIZE] = {"Max file size", "bytes"}, |
435 | [RLIMIT_DATA] = {"Max data size", "bytes"}, | |
436 | [RLIMIT_STACK] = {"Max stack size", "bytes"}, | |
437 | [RLIMIT_CORE] = {"Max core file size", "bytes"}, | |
438 | [RLIMIT_RSS] = {"Max resident set", "bytes"}, | |
439 | [RLIMIT_NPROC] = {"Max processes", "processes"}, | |
440 | [RLIMIT_NOFILE] = {"Max open files", "files"}, | |
441 | [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"}, | |
442 | [RLIMIT_AS] = {"Max address space", "bytes"}, | |
443 | [RLIMIT_LOCKS] = {"Max file locks", "locks"}, | |
444 | [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"}, | |
445 | [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"}, | |
446 | [RLIMIT_NICE] = {"Max nice priority", NULL}, | |
447 | [RLIMIT_RTPRIO] = {"Max realtime priority", NULL}, | |
8808117c | 448 | [RLIMIT_RTTIME] = {"Max realtime timeout", "us"}, |
d85f50d5 NH |
449 | }; |
450 | ||
451 | /* Display limits for a process */ | |
1c963eb1 AD |
452 | static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns, |
453 | struct pid *pid, struct task_struct *task) | |
d85f50d5 NH |
454 | { |
455 | unsigned int i; | |
d85f50d5 | 456 | unsigned long flags; |
d85f50d5 NH |
457 | |
458 | struct rlimit rlim[RLIM_NLIMITS]; | |
459 | ||
a6bebbc8 | 460 | if (!lock_task_sighand(task, &flags)) |
d85f50d5 | 461 | return 0; |
d85f50d5 NH |
462 | memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS); |
463 | unlock_task_sighand(task, &flags); | |
d85f50d5 NH |
464 | |
465 | /* | |
466 | * print the file header | |
467 | */ | |
1c963eb1 | 468 | seq_printf(m, "%-25s %-20s %-20s %-10s\n", |
d85f50d5 NH |
469 | "Limit", "Soft Limit", "Hard Limit", "Units"); |
470 | ||
471 | for (i = 0; i < RLIM_NLIMITS; i++) { | |
472 | if (rlim[i].rlim_cur == RLIM_INFINITY) | |
1c963eb1 | 473 | seq_printf(m, "%-25s %-20s ", |
d85f50d5 NH |
474 | lnames[i].name, "unlimited"); |
475 | else | |
1c963eb1 | 476 | seq_printf(m, "%-25s %-20lu ", |
d85f50d5 NH |
477 | lnames[i].name, rlim[i].rlim_cur); |
478 | ||
479 | if (rlim[i].rlim_max == RLIM_INFINITY) | |
1c963eb1 | 480 | seq_printf(m, "%-20s ", "unlimited"); |
d85f50d5 | 481 | else |
1c963eb1 | 482 | seq_printf(m, "%-20lu ", rlim[i].rlim_max); |
d85f50d5 NH |
483 | |
484 | if (lnames[i].unit) | |
1c963eb1 | 485 | seq_printf(m, "%-10s\n", lnames[i].unit); |
d85f50d5 | 486 | else |
1c963eb1 | 487 | seq_putc(m, '\n'); |
d85f50d5 NH |
488 | } |
489 | ||
1c963eb1 | 490 | return 0; |
d85f50d5 NH |
491 | } |
492 | ||
ebcb6734 | 493 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 AD |
494 | static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns, |
495 | struct pid *pid, struct task_struct *task) | |
ebcb6734 RM |
496 | { |
497 | long nr; | |
498 | unsigned long args[6], sp, pc; | |
a9712bc1 AV |
499 | int res = lock_trace(task); |
500 | if (res) | |
501 | return res; | |
ebcb6734 RM |
502 | |
503 | if (task_current_syscall(task, &nr, args, 6, &sp, &pc)) | |
09d93bd6 | 504 | seq_puts(m, "running\n"); |
a9712bc1 | 505 | else if (nr < 0) |
09d93bd6 | 506 | seq_printf(m, "%ld 0x%lx 0x%lx\n", nr, sp, pc); |
a9712bc1 | 507 | else |
09d93bd6 | 508 | seq_printf(m, |
ebcb6734 RM |
509 | "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n", |
510 | nr, | |
511 | args[0], args[1], args[2], args[3], args[4], args[5], | |
512 | sp, pc); | |
a9712bc1 AV |
513 | unlock_trace(task); |
514 | return res; | |
ebcb6734 RM |
515 | } |
516 | #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */ | |
517 | ||
1da177e4 LT |
518 | /************************************************************************/ |
519 | /* Here the fs part begins */ | |
520 | /************************************************************************/ | |
521 | ||
522 | /* permission checks */ | |
778c1144 | 523 | static int proc_fd_access_allowed(struct inode *inode) |
1da177e4 | 524 | { |
778c1144 EB |
525 | struct task_struct *task; |
526 | int allowed = 0; | |
df26c40e EB |
527 | /* Allow access to a task's file descriptors if it is us or we |
528 | * may use ptrace attach to the process and find out that | |
529 | * information. | |
778c1144 EB |
530 | */ |
531 | task = get_proc_task(inode); | |
df26c40e | 532 | if (task) { |
006ebb40 | 533 | allowed = ptrace_may_access(task, PTRACE_MODE_READ); |
778c1144 | 534 | put_task_struct(task); |
df26c40e | 535 | } |
778c1144 | 536 | return allowed; |
1da177e4 LT |
537 | } |
538 | ||
6b4e306a | 539 | int proc_setattr(struct dentry *dentry, struct iattr *attr) |
6d76fa58 LT |
540 | { |
541 | int error; | |
542 | struct inode *inode = dentry->d_inode; | |
543 | ||
544 | if (attr->ia_valid & ATTR_MODE) | |
545 | return -EPERM; | |
546 | ||
547 | error = inode_change_ok(inode, attr); | |
1025774c CH |
548 | if (error) |
549 | return error; | |
550 | ||
1025774c CH |
551 | setattr_copy(inode, attr); |
552 | mark_inode_dirty(inode); | |
553 | return 0; | |
6d76fa58 LT |
554 | } |
555 | ||
0499680a VK |
556 | /* |
557 | * May current process learn task's sched/cmdline info (for hide_pid_min=1) | |
558 | * or euid/egid (for hide_pid_min=2)? | |
559 | */ | |
560 | static bool has_pid_permissions(struct pid_namespace *pid, | |
561 | struct task_struct *task, | |
562 | int hide_pid_min) | |
563 | { | |
564 | if (pid->hide_pid < hide_pid_min) | |
565 | return true; | |
566 | if (in_group_p(pid->pid_gid)) | |
567 | return true; | |
568 | return ptrace_may_access(task, PTRACE_MODE_READ); | |
569 | } | |
570 | ||
571 | ||
572 | static int proc_pid_permission(struct inode *inode, int mask) | |
573 | { | |
574 | struct pid_namespace *pid = inode->i_sb->s_fs_info; | |
575 | struct task_struct *task; | |
576 | bool has_perms; | |
577 | ||
578 | task = get_proc_task(inode); | |
a2ef990a XF |
579 | if (!task) |
580 | return -ESRCH; | |
0499680a VK |
581 | has_perms = has_pid_permissions(pid, task, 1); |
582 | put_task_struct(task); | |
583 | ||
584 | if (!has_perms) { | |
585 | if (pid->hide_pid == 2) { | |
586 | /* | |
587 | * Let's make getdents(), stat(), and open() | |
588 | * consistent with each other. If a process | |
589 | * may not stat() a file, it shouldn't be seen | |
590 | * in procfs at all. | |
591 | */ | |
592 | return -ENOENT; | |
593 | } | |
594 | ||
595 | return -EPERM; | |
596 | } | |
597 | return generic_permission(inode, mask); | |
598 | } | |
599 | ||
600 | ||
601 | ||
c5ef1c42 | 602 | static const struct inode_operations proc_def_inode_operations = { |
6d76fa58 LT |
603 | .setattr = proc_setattr, |
604 | }; | |
605 | ||
1da177e4 LT |
606 | #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */ |
607 | ||
608 | static ssize_t proc_info_read(struct file * file, char __user * buf, | |
609 | size_t count, loff_t *ppos) | |
610 | { | |
496ad9aa | 611 | struct inode * inode = file_inode(file); |
1da177e4 LT |
612 | unsigned long page; |
613 | ssize_t length; | |
99f89551 EB |
614 | struct task_struct *task = get_proc_task(inode); |
615 | ||
616 | length = -ESRCH; | |
617 | if (!task) | |
618 | goto out_no_task; | |
1da177e4 LT |
619 | |
620 | if (count > PROC_BLOCK_SIZE) | |
621 | count = PROC_BLOCK_SIZE; | |
99f89551 EB |
622 | |
623 | length = -ENOMEM; | |
e12ba74d | 624 | if (!(page = __get_free_page(GFP_TEMPORARY))) |
99f89551 | 625 | goto out; |
1da177e4 LT |
626 | |
627 | length = PROC_I(inode)->op.proc_read(task, (char*)page); | |
628 | ||
629 | if (length >= 0) | |
630 | length = simple_read_from_buffer(buf, count, ppos, (char *)page, length); | |
631 | free_page(page); | |
99f89551 EB |
632 | out: |
633 | put_task_struct(task); | |
634 | out_no_task: | |
1da177e4 LT |
635 | return length; |
636 | } | |
637 | ||
00977a59 | 638 | static const struct file_operations proc_info_file_operations = { |
1da177e4 | 639 | .read = proc_info_read, |
87df8424 | 640 | .llseek = generic_file_llseek, |
1da177e4 LT |
641 | }; |
642 | ||
be614086 EB |
643 | static int proc_single_show(struct seq_file *m, void *v) |
644 | { | |
645 | struct inode *inode = m->private; | |
646 | struct pid_namespace *ns; | |
647 | struct pid *pid; | |
648 | struct task_struct *task; | |
649 | int ret; | |
650 | ||
651 | ns = inode->i_sb->s_fs_info; | |
652 | pid = proc_pid(inode); | |
653 | task = get_pid_task(pid, PIDTYPE_PID); | |
654 | if (!task) | |
655 | return -ESRCH; | |
656 | ||
657 | ret = PROC_I(inode)->op.proc_show(m, ns, pid, task); | |
658 | ||
659 | put_task_struct(task); | |
660 | return ret; | |
661 | } | |
662 | ||
663 | static int proc_single_open(struct inode *inode, struct file *filp) | |
664 | { | |
c6a34058 | 665 | return single_open(filp, proc_single_show, inode); |
be614086 EB |
666 | } |
667 | ||
668 | static const struct file_operations proc_single_file_operations = { | |
669 | .open = proc_single_open, | |
670 | .read = seq_read, | |
671 | .llseek = seq_lseek, | |
672 | .release = single_release, | |
673 | }; | |
674 | ||
b409e578 | 675 | static int __mem_open(struct inode *inode, struct file *file, unsigned int mode) |
1da177e4 | 676 | { |
496ad9aa | 677 | struct task_struct *task = get_proc_task(file_inode(file)); |
e268337d LT |
678 | struct mm_struct *mm; |
679 | ||
680 | if (!task) | |
681 | return -ESRCH; | |
682 | ||
b409e578 | 683 | mm = mm_access(task, mode); |
e268337d LT |
684 | put_task_struct(task); |
685 | ||
686 | if (IS_ERR(mm)) | |
687 | return PTR_ERR(mm); | |
688 | ||
6d08f2c7 ON |
689 | if (mm) { |
690 | /* ensure this mm_struct can't be freed */ | |
691 | atomic_inc(&mm->mm_count); | |
692 | /* but do not pin its memory */ | |
693 | mmput(mm); | |
694 | } | |
695 | ||
e268337d LT |
696 | file->private_data = mm; |
697 | ||
1da177e4 LT |
698 | return 0; |
699 | } | |
700 | ||
b409e578 CW |
701 | static int mem_open(struct inode *inode, struct file *file) |
702 | { | |
bc452b4b DH |
703 | int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH); |
704 | ||
705 | /* OK to pass negative loff_t, we can catch out-of-range */ | |
706 | file->f_mode |= FMODE_UNSIGNED_OFFSET; | |
707 | ||
708 | return ret; | |
b409e578 CW |
709 | } |
710 | ||
572d34b9 ON |
711 | static ssize_t mem_rw(struct file *file, char __user *buf, |
712 | size_t count, loff_t *ppos, int write) | |
1da177e4 | 713 | { |
e268337d | 714 | struct mm_struct *mm = file->private_data; |
572d34b9 ON |
715 | unsigned long addr = *ppos; |
716 | ssize_t copied; | |
1da177e4 | 717 | char *page; |
1da177e4 | 718 | |
e268337d LT |
719 | if (!mm) |
720 | return 0; | |
99f89551 | 721 | |
30cd8903 KM |
722 | page = (char *)__get_free_page(GFP_TEMPORARY); |
723 | if (!page) | |
e268337d | 724 | return -ENOMEM; |
1da177e4 | 725 | |
f7ca54f4 | 726 | copied = 0; |
6d08f2c7 ON |
727 | if (!atomic_inc_not_zero(&mm->mm_users)) |
728 | goto free; | |
729 | ||
1da177e4 | 730 | while (count > 0) { |
572d34b9 | 731 | int this_len = min_t(int, count, PAGE_SIZE); |
1da177e4 | 732 | |
572d34b9 | 733 | if (write && copy_from_user(page, buf, this_len)) { |
1da177e4 LT |
734 | copied = -EFAULT; |
735 | break; | |
736 | } | |
572d34b9 ON |
737 | |
738 | this_len = access_remote_vm(mm, addr, page, this_len, write); | |
739 | if (!this_len) { | |
1da177e4 LT |
740 | if (!copied) |
741 | copied = -EIO; | |
742 | break; | |
743 | } | |
572d34b9 ON |
744 | |
745 | if (!write && copy_to_user(buf, page, this_len)) { | |
746 | copied = -EFAULT; | |
747 | break; | |
748 | } | |
749 | ||
750 | buf += this_len; | |
751 | addr += this_len; | |
752 | copied += this_len; | |
753 | count -= this_len; | |
1da177e4 | 754 | } |
572d34b9 | 755 | *ppos = addr; |
30cd8903 | 756 | |
6d08f2c7 ON |
757 | mmput(mm); |
758 | free: | |
30cd8903 | 759 | free_page((unsigned long) page); |
1da177e4 LT |
760 | return copied; |
761 | } | |
1da177e4 | 762 | |
572d34b9 ON |
763 | static ssize_t mem_read(struct file *file, char __user *buf, |
764 | size_t count, loff_t *ppos) | |
765 | { | |
766 | return mem_rw(file, buf, count, ppos, 0); | |
767 | } | |
768 | ||
769 | static ssize_t mem_write(struct file *file, const char __user *buf, | |
770 | size_t count, loff_t *ppos) | |
771 | { | |
772 | return mem_rw(file, (char __user*)buf, count, ppos, 1); | |
773 | } | |
774 | ||
85863e47 | 775 | loff_t mem_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
776 | { |
777 | switch (orig) { | |
778 | case 0: | |
779 | file->f_pos = offset; | |
780 | break; | |
781 | case 1: | |
782 | file->f_pos += offset; | |
783 | break; | |
784 | default: | |
785 | return -EINVAL; | |
786 | } | |
787 | force_successful_syscall_return(); | |
788 | return file->f_pos; | |
789 | } | |
790 | ||
e268337d LT |
791 | static int mem_release(struct inode *inode, struct file *file) |
792 | { | |
793 | struct mm_struct *mm = file->private_data; | |
71879d3c | 794 | if (mm) |
6d08f2c7 | 795 | mmdrop(mm); |
e268337d LT |
796 | return 0; |
797 | } | |
798 | ||
00977a59 | 799 | static const struct file_operations proc_mem_operations = { |
1da177e4 LT |
800 | .llseek = mem_lseek, |
801 | .read = mem_read, | |
802 | .write = mem_write, | |
803 | .open = mem_open, | |
e268337d | 804 | .release = mem_release, |
1da177e4 LT |
805 | }; |
806 | ||
b409e578 CW |
807 | static int environ_open(struct inode *inode, struct file *file) |
808 | { | |
809 | return __mem_open(inode, file, PTRACE_MODE_READ); | |
810 | } | |
811 | ||
315e28c8 JP |
812 | static ssize_t environ_read(struct file *file, char __user *buf, |
813 | size_t count, loff_t *ppos) | |
814 | { | |
315e28c8 JP |
815 | char *page; |
816 | unsigned long src = *ppos; | |
b409e578 CW |
817 | int ret = 0; |
818 | struct mm_struct *mm = file->private_data; | |
315e28c8 | 819 | |
b409e578 CW |
820 | if (!mm) |
821 | return 0; | |
315e28c8 | 822 | |
315e28c8 JP |
823 | page = (char *)__get_free_page(GFP_TEMPORARY); |
824 | if (!page) | |
b409e578 | 825 | return -ENOMEM; |
315e28c8 | 826 | |
d6f64b89 | 827 | ret = 0; |
b409e578 CW |
828 | if (!atomic_inc_not_zero(&mm->mm_users)) |
829 | goto free; | |
315e28c8 | 830 | while (count > 0) { |
e8905ec2 DH |
831 | size_t this_len, max_len; |
832 | int retval; | |
315e28c8 | 833 | |
e8905ec2 | 834 | if (src >= (mm->env_end - mm->env_start)) |
315e28c8 JP |
835 | break; |
836 | ||
e8905ec2 DH |
837 | this_len = mm->env_end - (mm->env_start + src); |
838 | ||
839 | max_len = min_t(size_t, PAGE_SIZE, count); | |
840 | this_len = min(max_len, this_len); | |
315e28c8 | 841 | |
b409e578 | 842 | retval = access_remote_vm(mm, (mm->env_start + src), |
315e28c8 JP |
843 | page, this_len, 0); |
844 | ||
845 | if (retval <= 0) { | |
846 | ret = retval; | |
847 | break; | |
848 | } | |
849 | ||
850 | if (copy_to_user(buf, page, retval)) { | |
851 | ret = -EFAULT; | |
852 | break; | |
853 | } | |
854 | ||
855 | ret += retval; | |
856 | src += retval; | |
857 | buf += retval; | |
858 | count -= retval; | |
859 | } | |
860 | *ppos = src; | |
315e28c8 | 861 | mmput(mm); |
b409e578 CW |
862 | |
863 | free: | |
315e28c8 | 864 | free_page((unsigned long) page); |
315e28c8 JP |
865 | return ret; |
866 | } | |
867 | ||
868 | static const struct file_operations proc_environ_operations = { | |
b409e578 | 869 | .open = environ_open, |
315e28c8 | 870 | .read = environ_read, |
87df8424 | 871 | .llseek = generic_file_llseek, |
b409e578 | 872 | .release = mem_release, |
315e28c8 JP |
873 | }; |
874 | ||
fa0cbbf1 DR |
875 | static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count, |
876 | loff_t *ppos) | |
877 | { | |
496ad9aa | 878 | struct task_struct *task = get_proc_task(file_inode(file)); |
fa0cbbf1 DR |
879 | char buffer[PROC_NUMBUF]; |
880 | int oom_adj = OOM_ADJUST_MIN; | |
881 | size_t len; | |
882 | unsigned long flags; | |
883 | ||
884 | if (!task) | |
885 | return -ESRCH; | |
886 | if (lock_task_sighand(task, &flags)) { | |
887 | if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX) | |
888 | oom_adj = OOM_ADJUST_MAX; | |
889 | else | |
890 | oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) / | |
891 | OOM_SCORE_ADJ_MAX; | |
892 | unlock_task_sighand(task, &flags); | |
893 | } | |
894 | put_task_struct(task); | |
895 | len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj); | |
896 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
897 | } | |
898 | ||
899 | static ssize_t oom_adj_write(struct file *file, const char __user *buf, | |
900 | size_t count, loff_t *ppos) | |
901 | { | |
902 | struct task_struct *task; | |
903 | char buffer[PROC_NUMBUF]; | |
904 | int oom_adj; | |
905 | unsigned long flags; | |
906 | int err; | |
907 | ||
908 | memset(buffer, 0, sizeof(buffer)); | |
909 | if (count > sizeof(buffer) - 1) | |
910 | count = sizeof(buffer) - 1; | |
911 | if (copy_from_user(buffer, buf, count)) { | |
912 | err = -EFAULT; | |
913 | goto out; | |
914 | } | |
915 | ||
916 | err = kstrtoint(strstrip(buffer), 0, &oom_adj); | |
917 | if (err) | |
918 | goto out; | |
919 | if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) && | |
920 | oom_adj != OOM_DISABLE) { | |
921 | err = -EINVAL; | |
922 | goto out; | |
923 | } | |
924 | ||
496ad9aa | 925 | task = get_proc_task(file_inode(file)); |
fa0cbbf1 DR |
926 | if (!task) { |
927 | err = -ESRCH; | |
928 | goto out; | |
929 | } | |
930 | ||
931 | task_lock(task); | |
932 | if (!task->mm) { | |
933 | err = -EINVAL; | |
934 | goto err_task_lock; | |
935 | } | |
936 | ||
937 | if (!lock_task_sighand(task, &flags)) { | |
938 | err = -ESRCH; | |
939 | goto err_task_lock; | |
940 | } | |
941 | ||
942 | /* | |
943 | * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum | |
944 | * value is always attainable. | |
945 | */ | |
946 | if (oom_adj == OOM_ADJUST_MAX) | |
947 | oom_adj = OOM_SCORE_ADJ_MAX; | |
948 | else | |
949 | oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE; | |
950 | ||
951 | if (oom_adj < task->signal->oom_score_adj && | |
952 | !capable(CAP_SYS_RESOURCE)) { | |
953 | err = -EACCES; | |
954 | goto err_sighand; | |
955 | } | |
956 | ||
957 | /* | |
958 | * /proc/pid/oom_adj is provided for legacy purposes, ask users to use | |
959 | * /proc/pid/oom_score_adj instead. | |
960 | */ | |
87ebdc00 | 961 | pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n", |
fa0cbbf1 DR |
962 | current->comm, task_pid_nr(current), task_pid_nr(task), |
963 | task_pid_nr(task)); | |
964 | ||
965 | task->signal->oom_score_adj = oom_adj; | |
966 | trace_oom_score_adj_update(task); | |
967 | err_sighand: | |
968 | unlock_task_sighand(task, &flags); | |
969 | err_task_lock: | |
970 | task_unlock(task); | |
971 | put_task_struct(task); | |
972 | out: | |
973 | return err < 0 ? err : count; | |
974 | } | |
975 | ||
976 | static const struct file_operations proc_oom_adj_operations = { | |
977 | .read = oom_adj_read, | |
978 | .write = oom_adj_write, | |
979 | .llseek = generic_file_llseek, | |
980 | }; | |
981 | ||
a63d83f4 DR |
982 | static ssize_t oom_score_adj_read(struct file *file, char __user *buf, |
983 | size_t count, loff_t *ppos) | |
984 | { | |
496ad9aa | 985 | struct task_struct *task = get_proc_task(file_inode(file)); |
a63d83f4 | 986 | char buffer[PROC_NUMBUF]; |
a9c58b90 | 987 | short oom_score_adj = OOM_SCORE_ADJ_MIN; |
a63d83f4 DR |
988 | unsigned long flags; |
989 | size_t len; | |
990 | ||
991 | if (!task) | |
992 | return -ESRCH; | |
993 | if (lock_task_sighand(task, &flags)) { | |
994 | oom_score_adj = task->signal->oom_score_adj; | |
995 | unlock_task_sighand(task, &flags); | |
996 | } | |
997 | put_task_struct(task); | |
a9c58b90 | 998 | len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj); |
a63d83f4 DR |
999 | return simple_read_from_buffer(buf, count, ppos, buffer, len); |
1000 | } | |
1001 | ||
1002 | static ssize_t oom_score_adj_write(struct file *file, const char __user *buf, | |
1003 | size_t count, loff_t *ppos) | |
1004 | { | |
1005 | struct task_struct *task; | |
1006 | char buffer[PROC_NUMBUF]; | |
1007 | unsigned long flags; | |
0a8cb8e3 | 1008 | int oom_score_adj; |
a63d83f4 DR |
1009 | int err; |
1010 | ||
1011 | memset(buffer, 0, sizeof(buffer)); | |
1012 | if (count > sizeof(buffer) - 1) | |
1013 | count = sizeof(buffer) - 1; | |
723548bf DR |
1014 | if (copy_from_user(buffer, buf, count)) { |
1015 | err = -EFAULT; | |
1016 | goto out; | |
1017 | } | |
a63d83f4 | 1018 | |
0a8cb8e3 | 1019 | err = kstrtoint(strstrip(buffer), 0, &oom_score_adj); |
a63d83f4 | 1020 | if (err) |
723548bf | 1021 | goto out; |
a63d83f4 | 1022 | if (oom_score_adj < OOM_SCORE_ADJ_MIN || |
723548bf DR |
1023 | oom_score_adj > OOM_SCORE_ADJ_MAX) { |
1024 | err = -EINVAL; | |
1025 | goto out; | |
1026 | } | |
a63d83f4 | 1027 | |
496ad9aa | 1028 | task = get_proc_task(file_inode(file)); |
723548bf DR |
1029 | if (!task) { |
1030 | err = -ESRCH; | |
1031 | goto out; | |
1032 | } | |
d19d5476 DR |
1033 | |
1034 | task_lock(task); | |
1035 | if (!task->mm) { | |
1036 | err = -EINVAL; | |
1037 | goto err_task_lock; | |
1038 | } | |
1039 | ||
a63d83f4 | 1040 | if (!lock_task_sighand(task, &flags)) { |
723548bf | 1041 | err = -ESRCH; |
d19d5476 | 1042 | goto err_task_lock; |
a63d83f4 | 1043 | } |
d19d5476 | 1044 | |
a9c58b90 | 1045 | if ((short)oom_score_adj < task->signal->oom_score_adj_min && |
a63d83f4 | 1046 | !capable(CAP_SYS_RESOURCE)) { |
723548bf DR |
1047 | err = -EACCES; |
1048 | goto err_sighand; | |
a63d83f4 DR |
1049 | } |
1050 | ||
a9c58b90 | 1051 | task->signal->oom_score_adj = (short)oom_score_adj; |
dabb16f6 | 1052 | if (has_capability_noaudit(current, CAP_SYS_RESOURCE)) |
a9c58b90 | 1053 | task->signal->oom_score_adj_min = (short)oom_score_adj; |
43d2b113 | 1054 | trace_oom_score_adj_update(task); |
01dc52eb | 1055 | |
723548bf | 1056 | err_sighand: |
a63d83f4 | 1057 | unlock_task_sighand(task, &flags); |
d19d5476 DR |
1058 | err_task_lock: |
1059 | task_unlock(task); | |
a63d83f4 | 1060 | put_task_struct(task); |
723548bf DR |
1061 | out: |
1062 | return err < 0 ? err : count; | |
a63d83f4 DR |
1063 | } |
1064 | ||
1065 | static const struct file_operations proc_oom_score_adj_operations = { | |
1066 | .read = oom_score_adj_read, | |
1067 | .write = oom_score_adj_write, | |
6038f373 | 1068 | .llseek = default_llseek, |
a63d83f4 DR |
1069 | }; |
1070 | ||
1da177e4 LT |
1071 | #ifdef CONFIG_AUDITSYSCALL |
1072 | #define TMPBUFLEN 21 | |
1073 | static ssize_t proc_loginuid_read(struct file * file, char __user * buf, | |
1074 | size_t count, loff_t *ppos) | |
1075 | { | |
496ad9aa | 1076 | struct inode * inode = file_inode(file); |
99f89551 | 1077 | struct task_struct *task = get_proc_task(inode); |
1da177e4 LT |
1078 | ssize_t length; |
1079 | char tmpbuf[TMPBUFLEN]; | |
1080 | ||
99f89551 EB |
1081 | if (!task) |
1082 | return -ESRCH; | |
1da177e4 | 1083 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", |
e1760bd5 EB |
1084 | from_kuid(file->f_cred->user_ns, |
1085 | audit_get_loginuid(task))); | |
99f89551 | 1086 | put_task_struct(task); |
1da177e4 LT |
1087 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); |
1088 | } | |
1089 | ||
1090 | static ssize_t proc_loginuid_write(struct file * file, const char __user * buf, | |
1091 | size_t count, loff_t *ppos) | |
1092 | { | |
496ad9aa | 1093 | struct inode * inode = file_inode(file); |
1da177e4 LT |
1094 | char *page, *tmp; |
1095 | ssize_t length; | |
1da177e4 | 1096 | uid_t loginuid; |
e1760bd5 | 1097 | kuid_t kloginuid; |
1da177e4 | 1098 | |
7dc52157 PM |
1099 | rcu_read_lock(); |
1100 | if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) { | |
1101 | rcu_read_unlock(); | |
1da177e4 | 1102 | return -EPERM; |
7dc52157 PM |
1103 | } |
1104 | rcu_read_unlock(); | |
1da177e4 | 1105 | |
e0182909 AV |
1106 | if (count >= PAGE_SIZE) |
1107 | count = PAGE_SIZE - 1; | |
1da177e4 LT |
1108 | |
1109 | if (*ppos != 0) { | |
1110 | /* No partial writes. */ | |
1111 | return -EINVAL; | |
1112 | } | |
e12ba74d | 1113 | page = (char*)__get_free_page(GFP_TEMPORARY); |
1da177e4 LT |
1114 | if (!page) |
1115 | return -ENOMEM; | |
1116 | length = -EFAULT; | |
1117 | if (copy_from_user(page, buf, count)) | |
1118 | goto out_free_page; | |
1119 | ||
e0182909 | 1120 | page[count] = '\0'; |
1da177e4 LT |
1121 | loginuid = simple_strtoul(page, &tmp, 10); |
1122 | if (tmp == page) { | |
1123 | length = -EINVAL; | |
1124 | goto out_free_page; | |
1125 | ||
1126 | } | |
81407c84 EP |
1127 | |
1128 | /* is userspace tring to explicitly UNSET the loginuid? */ | |
1129 | if (loginuid == AUDIT_UID_UNSET) { | |
1130 | kloginuid = INVALID_UID; | |
1131 | } else { | |
1132 | kloginuid = make_kuid(file->f_cred->user_ns, loginuid); | |
1133 | if (!uid_valid(kloginuid)) { | |
1134 | length = -EINVAL; | |
1135 | goto out_free_page; | |
1136 | } | |
e1760bd5 EB |
1137 | } |
1138 | ||
1139 | length = audit_set_loginuid(kloginuid); | |
1da177e4 LT |
1140 | if (likely(length == 0)) |
1141 | length = count; | |
1142 | ||
1143 | out_free_page: | |
1144 | free_page((unsigned long) page); | |
1145 | return length; | |
1146 | } | |
1147 | ||
00977a59 | 1148 | static const struct file_operations proc_loginuid_operations = { |
1da177e4 LT |
1149 | .read = proc_loginuid_read, |
1150 | .write = proc_loginuid_write, | |
87df8424 | 1151 | .llseek = generic_file_llseek, |
1da177e4 | 1152 | }; |
1e0bd755 EP |
1153 | |
1154 | static ssize_t proc_sessionid_read(struct file * file, char __user * buf, | |
1155 | size_t count, loff_t *ppos) | |
1156 | { | |
496ad9aa | 1157 | struct inode * inode = file_inode(file); |
1e0bd755 EP |
1158 | struct task_struct *task = get_proc_task(inode); |
1159 | ssize_t length; | |
1160 | char tmpbuf[TMPBUFLEN]; | |
1161 | ||
1162 | if (!task) | |
1163 | return -ESRCH; | |
1164 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", | |
1165 | audit_get_sessionid(task)); | |
1166 | put_task_struct(task); | |
1167 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); | |
1168 | } | |
1169 | ||
1170 | static const struct file_operations proc_sessionid_operations = { | |
1171 | .read = proc_sessionid_read, | |
87df8424 | 1172 | .llseek = generic_file_llseek, |
1e0bd755 | 1173 | }; |
1da177e4 LT |
1174 | #endif |
1175 | ||
f4f154fd AM |
1176 | #ifdef CONFIG_FAULT_INJECTION |
1177 | static ssize_t proc_fault_inject_read(struct file * file, char __user * buf, | |
1178 | size_t count, loff_t *ppos) | |
1179 | { | |
496ad9aa | 1180 | struct task_struct *task = get_proc_task(file_inode(file)); |
f4f154fd AM |
1181 | char buffer[PROC_NUMBUF]; |
1182 | size_t len; | |
1183 | int make_it_fail; | |
f4f154fd AM |
1184 | |
1185 | if (!task) | |
1186 | return -ESRCH; | |
1187 | make_it_fail = task->make_it_fail; | |
1188 | put_task_struct(task); | |
1189 | ||
1190 | len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail); | |
0c28f287 AM |
1191 | |
1192 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
f4f154fd AM |
1193 | } |
1194 | ||
1195 | static ssize_t proc_fault_inject_write(struct file * file, | |
1196 | const char __user * buf, size_t count, loff_t *ppos) | |
1197 | { | |
1198 | struct task_struct *task; | |
1199 | char buffer[PROC_NUMBUF], *end; | |
1200 | int make_it_fail; | |
1201 | ||
1202 | if (!capable(CAP_SYS_RESOURCE)) | |
1203 | return -EPERM; | |
1204 | memset(buffer, 0, sizeof(buffer)); | |
1205 | if (count > sizeof(buffer) - 1) | |
1206 | count = sizeof(buffer) - 1; | |
1207 | if (copy_from_user(buffer, buf, count)) | |
1208 | return -EFAULT; | |
cba8aafe VL |
1209 | make_it_fail = simple_strtol(strstrip(buffer), &end, 0); |
1210 | if (*end) | |
1211 | return -EINVAL; | |
16caed31 DJ |
1212 | if (make_it_fail < 0 || make_it_fail > 1) |
1213 | return -EINVAL; | |
1214 | ||
496ad9aa | 1215 | task = get_proc_task(file_inode(file)); |
f4f154fd AM |
1216 | if (!task) |
1217 | return -ESRCH; | |
1218 | task->make_it_fail = make_it_fail; | |
1219 | put_task_struct(task); | |
cba8aafe VL |
1220 | |
1221 | return count; | |
f4f154fd AM |
1222 | } |
1223 | ||
00977a59 | 1224 | static const struct file_operations proc_fault_inject_operations = { |
f4f154fd AM |
1225 | .read = proc_fault_inject_read, |
1226 | .write = proc_fault_inject_write, | |
87df8424 | 1227 | .llseek = generic_file_llseek, |
f4f154fd AM |
1228 | }; |
1229 | #endif | |
1230 | ||
9745512c | 1231 | |
43ae34cb IM |
1232 | #ifdef CONFIG_SCHED_DEBUG |
1233 | /* | |
1234 | * Print out various scheduling related per-task fields: | |
1235 | */ | |
1236 | static int sched_show(struct seq_file *m, void *v) | |
1237 | { | |
1238 | struct inode *inode = m->private; | |
1239 | struct task_struct *p; | |
1240 | ||
43ae34cb IM |
1241 | p = get_proc_task(inode); |
1242 | if (!p) | |
1243 | return -ESRCH; | |
1244 | proc_sched_show_task(p, m); | |
1245 | ||
1246 | put_task_struct(p); | |
1247 | ||
1248 | return 0; | |
1249 | } | |
1250 | ||
1251 | static ssize_t | |
1252 | sched_write(struct file *file, const char __user *buf, | |
1253 | size_t count, loff_t *offset) | |
1254 | { | |
496ad9aa | 1255 | struct inode *inode = file_inode(file); |
43ae34cb IM |
1256 | struct task_struct *p; |
1257 | ||
43ae34cb IM |
1258 | p = get_proc_task(inode); |
1259 | if (!p) | |
1260 | return -ESRCH; | |
1261 | proc_sched_set_task(p); | |
1262 | ||
1263 | put_task_struct(p); | |
1264 | ||
1265 | return count; | |
1266 | } | |
1267 | ||
1268 | static int sched_open(struct inode *inode, struct file *filp) | |
1269 | { | |
c6a34058 | 1270 | return single_open(filp, sched_show, inode); |
43ae34cb IM |
1271 | } |
1272 | ||
1273 | static const struct file_operations proc_pid_sched_operations = { | |
1274 | .open = sched_open, | |
1275 | .read = seq_read, | |
1276 | .write = sched_write, | |
1277 | .llseek = seq_lseek, | |
5ea473a1 | 1278 | .release = single_release, |
43ae34cb IM |
1279 | }; |
1280 | ||
1281 | #endif | |
1282 | ||
5091faa4 MG |
1283 | #ifdef CONFIG_SCHED_AUTOGROUP |
1284 | /* | |
1285 | * Print out autogroup related information: | |
1286 | */ | |
1287 | static int sched_autogroup_show(struct seq_file *m, void *v) | |
1288 | { | |
1289 | struct inode *inode = m->private; | |
1290 | struct task_struct *p; | |
1291 | ||
1292 | p = get_proc_task(inode); | |
1293 | if (!p) | |
1294 | return -ESRCH; | |
1295 | proc_sched_autogroup_show_task(p, m); | |
1296 | ||
1297 | put_task_struct(p); | |
1298 | ||
1299 | return 0; | |
1300 | } | |
1301 | ||
1302 | static ssize_t | |
1303 | sched_autogroup_write(struct file *file, const char __user *buf, | |
1304 | size_t count, loff_t *offset) | |
1305 | { | |
496ad9aa | 1306 | struct inode *inode = file_inode(file); |
5091faa4 MG |
1307 | struct task_struct *p; |
1308 | char buffer[PROC_NUMBUF]; | |
0a8cb8e3 | 1309 | int nice; |
5091faa4 MG |
1310 | int err; |
1311 | ||
1312 | memset(buffer, 0, sizeof(buffer)); | |
1313 | if (count > sizeof(buffer) - 1) | |
1314 | count = sizeof(buffer) - 1; | |
1315 | if (copy_from_user(buffer, buf, count)) | |
1316 | return -EFAULT; | |
1317 | ||
0a8cb8e3 AD |
1318 | err = kstrtoint(strstrip(buffer), 0, &nice); |
1319 | if (err < 0) | |
1320 | return err; | |
5091faa4 MG |
1321 | |
1322 | p = get_proc_task(inode); | |
1323 | if (!p) | |
1324 | return -ESRCH; | |
1325 | ||
2e5b5b3a | 1326 | err = proc_sched_autogroup_set_nice(p, nice); |
5091faa4 MG |
1327 | if (err) |
1328 | count = err; | |
1329 | ||
1330 | put_task_struct(p); | |
1331 | ||
1332 | return count; | |
1333 | } | |
1334 | ||
1335 | static int sched_autogroup_open(struct inode *inode, struct file *filp) | |
1336 | { | |
1337 | int ret; | |
1338 | ||
1339 | ret = single_open(filp, sched_autogroup_show, NULL); | |
1340 | if (!ret) { | |
1341 | struct seq_file *m = filp->private_data; | |
1342 | ||
1343 | m->private = inode; | |
1344 | } | |
1345 | return ret; | |
1346 | } | |
1347 | ||
1348 | static const struct file_operations proc_pid_sched_autogroup_operations = { | |
1349 | .open = sched_autogroup_open, | |
1350 | .read = seq_read, | |
1351 | .write = sched_autogroup_write, | |
1352 | .llseek = seq_lseek, | |
1353 | .release = single_release, | |
1354 | }; | |
1355 | ||
1356 | #endif /* CONFIG_SCHED_AUTOGROUP */ | |
1357 | ||
4614a696 JS |
1358 | static ssize_t comm_write(struct file *file, const char __user *buf, |
1359 | size_t count, loff_t *offset) | |
1360 | { | |
496ad9aa | 1361 | struct inode *inode = file_inode(file); |
4614a696 JS |
1362 | struct task_struct *p; |
1363 | char buffer[TASK_COMM_LEN]; | |
830e0fc9 | 1364 | const size_t maxlen = sizeof(buffer) - 1; |
4614a696 JS |
1365 | |
1366 | memset(buffer, 0, sizeof(buffer)); | |
830e0fc9 | 1367 | if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count)) |
4614a696 JS |
1368 | return -EFAULT; |
1369 | ||
1370 | p = get_proc_task(inode); | |
1371 | if (!p) | |
1372 | return -ESRCH; | |
1373 | ||
1374 | if (same_thread_group(current, p)) | |
1375 | set_task_comm(p, buffer); | |
1376 | else | |
1377 | count = -EINVAL; | |
1378 | ||
1379 | put_task_struct(p); | |
1380 | ||
1381 | return count; | |
1382 | } | |
1383 | ||
1384 | static int comm_show(struct seq_file *m, void *v) | |
1385 | { | |
1386 | struct inode *inode = m->private; | |
1387 | struct task_struct *p; | |
1388 | ||
1389 | p = get_proc_task(inode); | |
1390 | if (!p) | |
1391 | return -ESRCH; | |
1392 | ||
1393 | task_lock(p); | |
1394 | seq_printf(m, "%s\n", p->comm); | |
1395 | task_unlock(p); | |
1396 | ||
1397 | put_task_struct(p); | |
1398 | ||
1399 | return 0; | |
1400 | } | |
1401 | ||
1402 | static int comm_open(struct inode *inode, struct file *filp) | |
1403 | { | |
c6a34058 | 1404 | return single_open(filp, comm_show, inode); |
4614a696 JS |
1405 | } |
1406 | ||
1407 | static const struct file_operations proc_pid_set_comm_operations = { | |
1408 | .open = comm_open, | |
1409 | .read = seq_read, | |
1410 | .write = comm_write, | |
1411 | .llseek = seq_lseek, | |
1412 | .release = single_release, | |
1413 | }; | |
1414 | ||
7773fbc5 | 1415 | static int proc_exe_link(struct dentry *dentry, struct path *exe_path) |
925d1c40 MH |
1416 | { |
1417 | struct task_struct *task; | |
1418 | struct mm_struct *mm; | |
1419 | struct file *exe_file; | |
1420 | ||
7773fbc5 | 1421 | task = get_proc_task(dentry->d_inode); |
925d1c40 MH |
1422 | if (!task) |
1423 | return -ENOENT; | |
1424 | mm = get_task_mm(task); | |
1425 | put_task_struct(task); | |
1426 | if (!mm) | |
1427 | return -ENOENT; | |
1428 | exe_file = get_mm_exe_file(mm); | |
1429 | mmput(mm); | |
1430 | if (exe_file) { | |
1431 | *exe_path = exe_file->f_path; | |
1432 | path_get(&exe_file->f_path); | |
1433 | fput(exe_file); | |
1434 | return 0; | |
1435 | } else | |
1436 | return -ENOENT; | |
1437 | } | |
1438 | ||
008b150a | 1439 | static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd) |
1da177e4 LT |
1440 | { |
1441 | struct inode *inode = dentry->d_inode; | |
408ef013 | 1442 | struct path path; |
1da177e4 LT |
1443 | int error = -EACCES; |
1444 | ||
778c1144 EB |
1445 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1446 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1447 | goto out; |
1da177e4 | 1448 | |
408ef013 CH |
1449 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1450 | if (error) | |
1451 | goto out; | |
1452 | ||
b5fb63c1 | 1453 | nd_jump_link(nd, &path); |
408ef013 | 1454 | return NULL; |
1da177e4 | 1455 | out: |
008b150a | 1456 | return ERR_PTR(error); |
1da177e4 LT |
1457 | } |
1458 | ||
3dcd25f3 | 1459 | static int do_proc_readlink(struct path *path, char __user *buffer, int buflen) |
1da177e4 | 1460 | { |
e12ba74d | 1461 | char *tmp = (char*)__get_free_page(GFP_TEMPORARY); |
3dcd25f3 | 1462 | char *pathname; |
1da177e4 LT |
1463 | int len; |
1464 | ||
1465 | if (!tmp) | |
1466 | return -ENOMEM; | |
0c28f287 | 1467 | |
7b2a69ba | 1468 | pathname = d_path(path, tmp, PAGE_SIZE); |
3dcd25f3 JB |
1469 | len = PTR_ERR(pathname); |
1470 | if (IS_ERR(pathname)) | |
1da177e4 | 1471 | goto out; |
3dcd25f3 | 1472 | len = tmp + PAGE_SIZE - 1 - pathname; |
1da177e4 LT |
1473 | |
1474 | if (len > buflen) | |
1475 | len = buflen; | |
3dcd25f3 | 1476 | if (copy_to_user(buffer, pathname, len)) |
1da177e4 LT |
1477 | len = -EFAULT; |
1478 | out: | |
1479 | free_page((unsigned long)tmp); | |
1480 | return len; | |
1481 | } | |
1482 | ||
1483 | static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen) | |
1484 | { | |
1485 | int error = -EACCES; | |
1486 | struct inode *inode = dentry->d_inode; | |
3dcd25f3 | 1487 | struct path path; |
1da177e4 | 1488 | |
778c1144 EB |
1489 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1490 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1491 | goto out; |
1da177e4 | 1492 | |
7773fbc5 | 1493 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1da177e4 LT |
1494 | if (error) |
1495 | goto out; | |
1496 | ||
3dcd25f3 JB |
1497 | error = do_proc_readlink(&path, buffer, buflen); |
1498 | path_put(&path); | |
1da177e4 | 1499 | out: |
1da177e4 LT |
1500 | return error; |
1501 | } | |
1502 | ||
faf60af1 | 1503 | const struct inode_operations proc_pid_link_inode_operations = { |
1da177e4 | 1504 | .readlink = proc_pid_readlink, |
6d76fa58 LT |
1505 | .follow_link = proc_pid_follow_link, |
1506 | .setattr = proc_setattr, | |
1da177e4 LT |
1507 | }; |
1508 | ||
28a6d671 EB |
1509 | |
1510 | /* building an inode */ | |
1511 | ||
6b4e306a | 1512 | struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task) |
28a6d671 EB |
1513 | { |
1514 | struct inode * inode; | |
1515 | struct proc_inode *ei; | |
c69e8d9c | 1516 | const struct cred *cred; |
1da177e4 | 1517 | |
28a6d671 | 1518 | /* We need a new inode */ |
1da177e4 | 1519 | |
28a6d671 EB |
1520 | inode = new_inode(sb); |
1521 | if (!inode) | |
1522 | goto out; | |
1523 | ||
1524 | /* Common stuff */ | |
1525 | ei = PROC_I(inode); | |
85fe4025 | 1526 | inode->i_ino = get_next_ino(); |
28a6d671 | 1527 | inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; |
28a6d671 EB |
1528 | inode->i_op = &proc_def_inode_operations; |
1529 | ||
1530 | /* | |
1531 | * grab the reference to task. | |
1532 | */ | |
1a657f78 | 1533 | ei->pid = get_task_pid(task, PIDTYPE_PID); |
28a6d671 EB |
1534 | if (!ei->pid) |
1535 | goto out_unlock; | |
1536 | ||
28a6d671 | 1537 | if (task_dumpable(task)) { |
c69e8d9c DH |
1538 | rcu_read_lock(); |
1539 | cred = __task_cred(task); | |
1540 | inode->i_uid = cred->euid; | |
1541 | inode->i_gid = cred->egid; | |
1542 | rcu_read_unlock(); | |
1da177e4 | 1543 | } |
28a6d671 EB |
1544 | security_task_to_inode(task, inode); |
1545 | ||
1da177e4 | 1546 | out: |
28a6d671 EB |
1547 | return inode; |
1548 | ||
1549 | out_unlock: | |
1550 | iput(inode); | |
1551 | return NULL; | |
1da177e4 LT |
1552 | } |
1553 | ||
6b4e306a | 1554 | int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat) |
1da177e4 | 1555 | { |
1da177e4 | 1556 | struct inode *inode = dentry->d_inode; |
28a6d671 | 1557 | struct task_struct *task; |
c69e8d9c | 1558 | const struct cred *cred; |
0499680a | 1559 | struct pid_namespace *pid = dentry->d_sb->s_fs_info; |
c69e8d9c | 1560 | |
28a6d671 | 1561 | generic_fillattr(inode, stat); |
1da177e4 | 1562 | |
28a6d671 | 1563 | rcu_read_lock(); |
dcb0f222 EB |
1564 | stat->uid = GLOBAL_ROOT_UID; |
1565 | stat->gid = GLOBAL_ROOT_GID; | |
28a6d671 EB |
1566 | task = pid_task(proc_pid(inode), PIDTYPE_PID); |
1567 | if (task) { | |
0499680a VK |
1568 | if (!has_pid_permissions(pid, task, 2)) { |
1569 | rcu_read_unlock(); | |
1570 | /* | |
1571 | * This doesn't prevent learning whether PID exists, | |
1572 | * it only makes getattr() consistent with readdir(). | |
1573 | */ | |
1574 | return -ENOENT; | |
1575 | } | |
28a6d671 EB |
1576 | if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) || |
1577 | task_dumpable(task)) { | |
c69e8d9c DH |
1578 | cred = __task_cred(task); |
1579 | stat->uid = cred->euid; | |
1580 | stat->gid = cred->egid; | |
1da177e4 LT |
1581 | } |
1582 | } | |
28a6d671 | 1583 | rcu_read_unlock(); |
d6e71144 | 1584 | return 0; |
1da177e4 LT |
1585 | } |
1586 | ||
1da177e4 LT |
1587 | /* dentry stuff */ |
1588 | ||
1589 | /* | |
1590 | * Exceptional case: normally we are not allowed to unhash a busy | |
1591 | * directory. In this case, however, we can do it - no aliasing problems | |
1592 | * due to the way we treat inodes. | |
1593 | * | |
1594 | * Rewrite the inode's ownerships here because the owning task may have | |
1595 | * performed a setuid(), etc. | |
99f89551 EB |
1596 | * |
1597 | * Before the /proc/pid/status file was created the only way to read | |
1598 | * the effective uid of a /process was to stat /proc/pid. Reading | |
1599 | * /proc/pid/status is slow enough that procps and other packages | |
1600 | * kept stating /proc/pid. To keep the rules in /proc simple I have | |
1601 | * made this apply to all per process world readable and executable | |
1602 | * directories. | |
1da177e4 | 1603 | */ |
0b728e19 | 1604 | int pid_revalidate(struct dentry *dentry, unsigned int flags) |
1da177e4 | 1605 | { |
34286d66 NP |
1606 | struct inode *inode; |
1607 | struct task_struct *task; | |
c69e8d9c DH |
1608 | const struct cred *cred; |
1609 | ||
0b728e19 | 1610 | if (flags & LOOKUP_RCU) |
34286d66 NP |
1611 | return -ECHILD; |
1612 | ||
1613 | inode = dentry->d_inode; | |
1614 | task = get_proc_task(inode); | |
1615 | ||
99f89551 EB |
1616 | if (task) { |
1617 | if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) || | |
1618 | task_dumpable(task)) { | |
c69e8d9c DH |
1619 | rcu_read_lock(); |
1620 | cred = __task_cred(task); | |
1621 | inode->i_uid = cred->euid; | |
1622 | inode->i_gid = cred->egid; | |
1623 | rcu_read_unlock(); | |
1da177e4 | 1624 | } else { |
dcb0f222 EB |
1625 | inode->i_uid = GLOBAL_ROOT_UID; |
1626 | inode->i_gid = GLOBAL_ROOT_GID; | |
1da177e4 | 1627 | } |
9ee8ab9f | 1628 | inode->i_mode &= ~(S_ISUID | S_ISGID); |
1da177e4 | 1629 | security_task_to_inode(task, inode); |
99f89551 | 1630 | put_task_struct(task); |
1da177e4 LT |
1631 | return 1; |
1632 | } | |
1633 | d_drop(dentry); | |
1634 | return 0; | |
1635 | } | |
1636 | ||
d855a4b7 ON |
1637 | static inline bool proc_inode_is_dead(struct inode *inode) |
1638 | { | |
1639 | return !proc_pid(inode)->tasks[PIDTYPE_PID].first; | |
1640 | } | |
1641 | ||
1dd704b6 DH |
1642 | int pid_delete_dentry(const struct dentry *dentry) |
1643 | { | |
1644 | /* Is the task we represent dead? | |
1645 | * If so, then don't put the dentry on the lru list, | |
1646 | * kill it immediately. | |
1647 | */ | |
d855a4b7 | 1648 | return proc_inode_is_dead(dentry->d_inode); |
1dd704b6 DH |
1649 | } |
1650 | ||
6b4e306a | 1651 | const struct dentry_operations pid_dentry_operations = |
28a6d671 EB |
1652 | { |
1653 | .d_revalidate = pid_revalidate, | |
1654 | .d_delete = pid_delete_dentry, | |
1655 | }; | |
1656 | ||
1657 | /* Lookups */ | |
1658 | ||
1c0d04c9 EB |
1659 | /* |
1660 | * Fill a directory entry. | |
1661 | * | |
1662 | * If possible create the dcache entry and derive our inode number and | |
1663 | * file type from dcache entry. | |
1664 | * | |
1665 | * Since all of the proc inode numbers are dynamically generated, the inode | |
1666 | * numbers do not exist until the inode is cache. This means creating the | |
1667 | * the dcache entry in readdir is necessary to keep the inode numbers | |
1668 | * reported by readdir in sync with the inode numbers reported | |
1669 | * by stat. | |
1670 | */ | |
f0c3b509 | 1671 | bool proc_fill_cache(struct file *file, struct dir_context *ctx, |
6b4e306a | 1672 | const char *name, int len, |
c5141e6d | 1673 | instantiate_t instantiate, struct task_struct *task, const void *ptr) |
61a28784 | 1674 | { |
f0c3b509 | 1675 | struct dentry *child, *dir = file->f_path.dentry; |
1df98b8b | 1676 | struct qstr qname = QSTR_INIT(name, len); |
61a28784 | 1677 | struct inode *inode; |
1df98b8b AV |
1678 | unsigned type; |
1679 | ino_t ino; | |
61a28784 | 1680 | |
1df98b8b | 1681 | child = d_hash_and_lookup(dir, &qname); |
61a28784 | 1682 | if (!child) { |
1df98b8b AV |
1683 | child = d_alloc(dir, &qname); |
1684 | if (!child) | |
1685 | goto end_instantiate; | |
1686 | if (instantiate(dir->d_inode, child, task, ptr) < 0) { | |
1687 | dput(child); | |
1688 | goto end_instantiate; | |
61a28784 EB |
1689 | } |
1690 | } | |
61a28784 | 1691 | inode = child->d_inode; |
147ce699 AV |
1692 | ino = inode->i_ino; |
1693 | type = inode->i_mode >> 12; | |
61a28784 | 1694 | dput(child); |
f0c3b509 | 1695 | return dir_emit(ctx, name, len, ino, type); |
1df98b8b AV |
1696 | |
1697 | end_instantiate: | |
1698 | return dir_emit(ctx, name, len, 1, DT_UNKNOWN); | |
61a28784 EB |
1699 | } |
1700 | ||
640708a2 PE |
1701 | #ifdef CONFIG_CHECKPOINT_RESTORE |
1702 | ||
1703 | /* | |
1704 | * dname_to_vma_addr - maps a dentry name into two unsigned longs | |
1705 | * which represent vma start and end addresses. | |
1706 | */ | |
1707 | static int dname_to_vma_addr(struct dentry *dentry, | |
1708 | unsigned long *start, unsigned long *end) | |
1709 | { | |
1710 | if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2) | |
1711 | return -EINVAL; | |
1712 | ||
1713 | return 0; | |
1714 | } | |
1715 | ||
0b728e19 | 1716 | static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags) |
640708a2 PE |
1717 | { |
1718 | unsigned long vm_start, vm_end; | |
1719 | bool exact_vma_exists = false; | |
1720 | struct mm_struct *mm = NULL; | |
1721 | struct task_struct *task; | |
1722 | const struct cred *cred; | |
1723 | struct inode *inode; | |
1724 | int status = 0; | |
1725 | ||
0b728e19 | 1726 | if (flags & LOOKUP_RCU) |
640708a2 PE |
1727 | return -ECHILD; |
1728 | ||
1729 | if (!capable(CAP_SYS_ADMIN)) { | |
41735818 | 1730 | status = -EPERM; |
640708a2 PE |
1731 | goto out_notask; |
1732 | } | |
1733 | ||
1734 | inode = dentry->d_inode; | |
1735 | task = get_proc_task(inode); | |
1736 | if (!task) | |
1737 | goto out_notask; | |
1738 | ||
2344bec7 CW |
1739 | mm = mm_access(task, PTRACE_MODE_READ); |
1740 | if (IS_ERR_OR_NULL(mm)) | |
640708a2 PE |
1741 | goto out; |
1742 | ||
1743 | if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) { | |
1744 | down_read(&mm->mmap_sem); | |
1745 | exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end); | |
1746 | up_read(&mm->mmap_sem); | |
1747 | } | |
1748 | ||
1749 | mmput(mm); | |
1750 | ||
1751 | if (exact_vma_exists) { | |
1752 | if (task_dumpable(task)) { | |
1753 | rcu_read_lock(); | |
1754 | cred = __task_cred(task); | |
1755 | inode->i_uid = cred->euid; | |
1756 | inode->i_gid = cred->egid; | |
1757 | rcu_read_unlock(); | |
1758 | } else { | |
dcb0f222 EB |
1759 | inode->i_uid = GLOBAL_ROOT_UID; |
1760 | inode->i_gid = GLOBAL_ROOT_GID; | |
640708a2 PE |
1761 | } |
1762 | security_task_to_inode(task, inode); | |
1763 | status = 1; | |
1764 | } | |
1765 | ||
1766 | out: | |
1767 | put_task_struct(task); | |
1768 | ||
1769 | out_notask: | |
1770 | if (status <= 0) | |
1771 | d_drop(dentry); | |
1772 | ||
1773 | return status; | |
1774 | } | |
1775 | ||
1776 | static const struct dentry_operations tid_map_files_dentry_operations = { | |
1777 | .d_revalidate = map_files_d_revalidate, | |
1778 | .d_delete = pid_delete_dentry, | |
1779 | }; | |
1780 | ||
1781 | static int proc_map_files_get_link(struct dentry *dentry, struct path *path) | |
1782 | { | |
1783 | unsigned long vm_start, vm_end; | |
1784 | struct vm_area_struct *vma; | |
1785 | struct task_struct *task; | |
1786 | struct mm_struct *mm; | |
1787 | int rc; | |
1788 | ||
1789 | rc = -ENOENT; | |
1790 | task = get_proc_task(dentry->d_inode); | |
1791 | if (!task) | |
1792 | goto out; | |
1793 | ||
1794 | mm = get_task_mm(task); | |
1795 | put_task_struct(task); | |
1796 | if (!mm) | |
1797 | goto out; | |
1798 | ||
1799 | rc = dname_to_vma_addr(dentry, &vm_start, &vm_end); | |
1800 | if (rc) | |
1801 | goto out_mmput; | |
1802 | ||
70335abb | 1803 | rc = -ENOENT; |
640708a2 PE |
1804 | down_read(&mm->mmap_sem); |
1805 | vma = find_exact_vma(mm, vm_start, vm_end); | |
1806 | if (vma && vma->vm_file) { | |
1807 | *path = vma->vm_file->f_path; | |
1808 | path_get(path); | |
1809 | rc = 0; | |
1810 | } | |
1811 | up_read(&mm->mmap_sem); | |
1812 | ||
1813 | out_mmput: | |
1814 | mmput(mm); | |
1815 | out: | |
1816 | return rc; | |
1817 | } | |
1818 | ||
1819 | struct map_files_info { | |
7b540d06 | 1820 | fmode_t mode; |
640708a2 PE |
1821 | unsigned long len; |
1822 | unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */ | |
1823 | }; | |
1824 | ||
c52a47ac | 1825 | static int |
640708a2 PE |
1826 | proc_map_files_instantiate(struct inode *dir, struct dentry *dentry, |
1827 | struct task_struct *task, const void *ptr) | |
1828 | { | |
7b540d06 | 1829 | fmode_t mode = (fmode_t)(unsigned long)ptr; |
640708a2 PE |
1830 | struct proc_inode *ei; |
1831 | struct inode *inode; | |
1832 | ||
640708a2 PE |
1833 | inode = proc_pid_make_inode(dir->i_sb, task); |
1834 | if (!inode) | |
c52a47ac | 1835 | return -ENOENT; |
640708a2 PE |
1836 | |
1837 | ei = PROC_I(inode); | |
1838 | ei->op.proc_get_link = proc_map_files_get_link; | |
1839 | ||
1840 | inode->i_op = &proc_pid_link_inode_operations; | |
1841 | inode->i_size = 64; | |
1842 | inode->i_mode = S_IFLNK; | |
1843 | ||
7b540d06 | 1844 | if (mode & FMODE_READ) |
640708a2 | 1845 | inode->i_mode |= S_IRUSR; |
7b540d06 | 1846 | if (mode & FMODE_WRITE) |
640708a2 PE |
1847 | inode->i_mode |= S_IWUSR; |
1848 | ||
1849 | d_set_d_op(dentry, &tid_map_files_dentry_operations); | |
1850 | d_add(dentry, inode); | |
1851 | ||
c52a47ac | 1852 | return 0; |
640708a2 PE |
1853 | } |
1854 | ||
1855 | static struct dentry *proc_map_files_lookup(struct inode *dir, | |
00cd8dd3 | 1856 | struct dentry *dentry, unsigned int flags) |
640708a2 PE |
1857 | { |
1858 | unsigned long vm_start, vm_end; | |
1859 | struct vm_area_struct *vma; | |
1860 | struct task_struct *task; | |
c52a47ac | 1861 | int result; |
640708a2 PE |
1862 | struct mm_struct *mm; |
1863 | ||
c52a47ac | 1864 | result = -EPERM; |
640708a2 PE |
1865 | if (!capable(CAP_SYS_ADMIN)) |
1866 | goto out; | |
1867 | ||
c52a47ac | 1868 | result = -ENOENT; |
640708a2 PE |
1869 | task = get_proc_task(dir); |
1870 | if (!task) | |
1871 | goto out; | |
1872 | ||
c52a47ac | 1873 | result = -EACCES; |
eb94cd96 | 1874 | if (!ptrace_may_access(task, PTRACE_MODE_READ)) |
640708a2 PE |
1875 | goto out_put_task; |
1876 | ||
c52a47ac | 1877 | result = -ENOENT; |
640708a2 | 1878 | if (dname_to_vma_addr(dentry, &vm_start, &vm_end)) |
eb94cd96 | 1879 | goto out_put_task; |
640708a2 PE |
1880 | |
1881 | mm = get_task_mm(task); | |
1882 | if (!mm) | |
eb94cd96 | 1883 | goto out_put_task; |
640708a2 PE |
1884 | |
1885 | down_read(&mm->mmap_sem); | |
1886 | vma = find_exact_vma(mm, vm_start, vm_end); | |
1887 | if (!vma) | |
1888 | goto out_no_vma; | |
1889 | ||
05f56484 SK |
1890 | if (vma->vm_file) |
1891 | result = proc_map_files_instantiate(dir, dentry, task, | |
1892 | (void *)(unsigned long)vma->vm_file->f_mode); | |
640708a2 PE |
1893 | |
1894 | out_no_vma: | |
1895 | up_read(&mm->mmap_sem); | |
1896 | mmput(mm); | |
640708a2 PE |
1897 | out_put_task: |
1898 | put_task_struct(task); | |
1899 | out: | |
c52a47ac | 1900 | return ERR_PTR(result); |
640708a2 PE |
1901 | } |
1902 | ||
1903 | static const struct inode_operations proc_map_files_inode_operations = { | |
1904 | .lookup = proc_map_files_lookup, | |
1905 | .permission = proc_fd_permission, | |
1906 | .setattr = proc_setattr, | |
1907 | }; | |
1908 | ||
1909 | static int | |
f0c3b509 | 1910 | proc_map_files_readdir(struct file *file, struct dir_context *ctx) |
640708a2 | 1911 | { |
640708a2 PE |
1912 | struct vm_area_struct *vma; |
1913 | struct task_struct *task; | |
1914 | struct mm_struct *mm; | |
f0c3b509 AV |
1915 | unsigned long nr_files, pos, i; |
1916 | struct flex_array *fa = NULL; | |
1917 | struct map_files_info info; | |
1918 | struct map_files_info *p; | |
640708a2 PE |
1919 | int ret; |
1920 | ||
41735818 | 1921 | ret = -EPERM; |
640708a2 PE |
1922 | if (!capable(CAP_SYS_ADMIN)) |
1923 | goto out; | |
1924 | ||
1925 | ret = -ENOENT; | |
f0c3b509 | 1926 | task = get_proc_task(file_inode(file)); |
640708a2 PE |
1927 | if (!task) |
1928 | goto out; | |
1929 | ||
1930 | ret = -EACCES; | |
eb94cd96 | 1931 | if (!ptrace_may_access(task, PTRACE_MODE_READ)) |
640708a2 PE |
1932 | goto out_put_task; |
1933 | ||
1934 | ret = 0; | |
f0c3b509 AV |
1935 | if (!dir_emit_dots(file, ctx)) |
1936 | goto out_put_task; | |
640708a2 | 1937 | |
f0c3b509 AV |
1938 | mm = get_task_mm(task); |
1939 | if (!mm) | |
1940 | goto out_put_task; | |
1941 | down_read(&mm->mmap_sem); | |
640708a2 | 1942 | |
f0c3b509 | 1943 | nr_files = 0; |
640708a2 | 1944 | |
f0c3b509 AV |
1945 | /* |
1946 | * We need two passes here: | |
1947 | * | |
1948 | * 1) Collect vmas of mapped files with mmap_sem taken | |
1949 | * 2) Release mmap_sem and instantiate entries | |
1950 | * | |
1951 | * otherwise we get lockdep complained, since filldir() | |
1952 | * routine might require mmap_sem taken in might_fault(). | |
1953 | */ | |
640708a2 | 1954 | |
f0c3b509 AV |
1955 | for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) { |
1956 | if (vma->vm_file && ++pos > ctx->pos) | |
1957 | nr_files++; | |
1958 | } | |
1959 | ||
1960 | if (nr_files) { | |
1961 | fa = flex_array_alloc(sizeof(info), nr_files, | |
1962 | GFP_KERNEL); | |
1963 | if (!fa || flex_array_prealloc(fa, 0, nr_files, | |
1964 | GFP_KERNEL)) { | |
1965 | ret = -ENOMEM; | |
1966 | if (fa) | |
1967 | flex_array_free(fa); | |
1968 | up_read(&mm->mmap_sem); | |
1969 | mmput(mm); | |
1970 | goto out_put_task; | |
640708a2 | 1971 | } |
f0c3b509 AV |
1972 | for (i = 0, vma = mm->mmap, pos = 2; vma; |
1973 | vma = vma->vm_next) { | |
1974 | if (!vma->vm_file) | |
1975 | continue; | |
1976 | if (++pos <= ctx->pos) | |
1977 | continue; | |
1978 | ||
1979 | info.mode = vma->vm_file->f_mode; | |
1980 | info.len = snprintf(info.name, | |
1981 | sizeof(info.name), "%lx-%lx", | |
1982 | vma->vm_start, vma->vm_end); | |
1983 | if (flex_array_put(fa, i++, &info, GFP_KERNEL)) | |
1984 | BUG(); | |
640708a2 | 1985 | } |
640708a2 | 1986 | } |
f0c3b509 AV |
1987 | up_read(&mm->mmap_sem); |
1988 | ||
1989 | for (i = 0; i < nr_files; i++) { | |
1990 | p = flex_array_get(fa, i); | |
1991 | if (!proc_fill_cache(file, ctx, | |
1992 | p->name, p->len, | |
1993 | proc_map_files_instantiate, | |
1994 | task, | |
1995 | (void *)(unsigned long)p->mode)) | |
1996 | break; | |
1997 | ctx->pos++; | |
640708a2 | 1998 | } |
f0c3b509 AV |
1999 | if (fa) |
2000 | flex_array_free(fa); | |
2001 | mmput(mm); | |
640708a2 | 2002 | |
640708a2 PE |
2003 | out_put_task: |
2004 | put_task_struct(task); | |
2005 | out: | |
2006 | return ret; | |
2007 | } | |
2008 | ||
2009 | static const struct file_operations proc_map_files_operations = { | |
2010 | .read = generic_read_dir, | |
f0c3b509 | 2011 | .iterate = proc_map_files_readdir, |
640708a2 PE |
2012 | .llseek = default_llseek, |
2013 | }; | |
2014 | ||
48f6a7a5 PE |
2015 | struct timers_private { |
2016 | struct pid *pid; | |
2017 | struct task_struct *task; | |
2018 | struct sighand_struct *sighand; | |
57b8015e | 2019 | struct pid_namespace *ns; |
48f6a7a5 PE |
2020 | unsigned long flags; |
2021 | }; | |
2022 | ||
2023 | static void *timers_start(struct seq_file *m, loff_t *pos) | |
2024 | { | |
2025 | struct timers_private *tp = m->private; | |
2026 | ||
2027 | tp->task = get_pid_task(tp->pid, PIDTYPE_PID); | |
2028 | if (!tp->task) | |
2029 | return ERR_PTR(-ESRCH); | |
2030 | ||
2031 | tp->sighand = lock_task_sighand(tp->task, &tp->flags); | |
2032 | if (!tp->sighand) | |
2033 | return ERR_PTR(-ESRCH); | |
2034 | ||
2035 | return seq_list_start(&tp->task->signal->posix_timers, *pos); | |
2036 | } | |
2037 | ||
2038 | static void *timers_next(struct seq_file *m, void *v, loff_t *pos) | |
2039 | { | |
2040 | struct timers_private *tp = m->private; | |
2041 | return seq_list_next(v, &tp->task->signal->posix_timers, pos); | |
2042 | } | |
2043 | ||
2044 | static void timers_stop(struct seq_file *m, void *v) | |
2045 | { | |
2046 | struct timers_private *tp = m->private; | |
2047 | ||
2048 | if (tp->sighand) { | |
2049 | unlock_task_sighand(tp->task, &tp->flags); | |
2050 | tp->sighand = NULL; | |
2051 | } | |
2052 | ||
2053 | if (tp->task) { | |
2054 | put_task_struct(tp->task); | |
2055 | tp->task = NULL; | |
2056 | } | |
2057 | } | |
2058 | ||
2059 | static int show_timer(struct seq_file *m, void *v) | |
2060 | { | |
2061 | struct k_itimer *timer; | |
57b8015e PE |
2062 | struct timers_private *tp = m->private; |
2063 | int notify; | |
cedbccab | 2064 | static const char * const nstr[] = { |
57b8015e PE |
2065 | [SIGEV_SIGNAL] = "signal", |
2066 | [SIGEV_NONE] = "none", | |
2067 | [SIGEV_THREAD] = "thread", | |
2068 | }; | |
48f6a7a5 PE |
2069 | |
2070 | timer = list_entry((struct list_head *)v, struct k_itimer, list); | |
57b8015e PE |
2071 | notify = timer->it_sigev_notify; |
2072 | ||
48f6a7a5 | 2073 | seq_printf(m, "ID: %d\n", timer->it_id); |
57b8015e PE |
2074 | seq_printf(m, "signal: %d/%p\n", timer->sigq->info.si_signo, |
2075 | timer->sigq->info.si_value.sival_ptr); | |
2076 | seq_printf(m, "notify: %s/%s.%d\n", | |
2077 | nstr[notify & ~SIGEV_THREAD_ID], | |
2078 | (notify & SIGEV_THREAD_ID) ? "tid" : "pid", | |
2079 | pid_nr_ns(timer->it_pid, tp->ns)); | |
15ef0298 | 2080 | seq_printf(m, "ClockID: %d\n", timer->it_clock); |
48f6a7a5 PE |
2081 | |
2082 | return 0; | |
2083 | } | |
2084 | ||
2085 | static const struct seq_operations proc_timers_seq_ops = { | |
2086 | .start = timers_start, | |
2087 | .next = timers_next, | |
2088 | .stop = timers_stop, | |
2089 | .show = show_timer, | |
2090 | }; | |
2091 | ||
2092 | static int proc_timers_open(struct inode *inode, struct file *file) | |
2093 | { | |
2094 | struct timers_private *tp; | |
2095 | ||
2096 | tp = __seq_open_private(file, &proc_timers_seq_ops, | |
2097 | sizeof(struct timers_private)); | |
2098 | if (!tp) | |
2099 | return -ENOMEM; | |
2100 | ||
2101 | tp->pid = proc_pid(inode); | |
57b8015e | 2102 | tp->ns = inode->i_sb->s_fs_info; |
48f6a7a5 PE |
2103 | return 0; |
2104 | } | |
2105 | ||
2106 | static const struct file_operations proc_timers_operations = { | |
2107 | .open = proc_timers_open, | |
2108 | .read = seq_read, | |
2109 | .llseek = seq_lseek, | |
2110 | .release = seq_release_private, | |
2111 | }; | |
640708a2 PE |
2112 | #endif /* CONFIG_CHECKPOINT_RESTORE */ |
2113 | ||
c52a47ac | 2114 | static int proc_pident_instantiate(struct inode *dir, |
c5141e6d | 2115 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
444ceed8 | 2116 | { |
c5141e6d | 2117 | const struct pid_entry *p = ptr; |
444ceed8 EB |
2118 | struct inode *inode; |
2119 | struct proc_inode *ei; | |
444ceed8 | 2120 | |
61a28784 | 2121 | inode = proc_pid_make_inode(dir->i_sb, task); |
444ceed8 EB |
2122 | if (!inode) |
2123 | goto out; | |
2124 | ||
2125 | ei = PROC_I(inode); | |
2126 | inode->i_mode = p->mode; | |
2127 | if (S_ISDIR(inode->i_mode)) | |
bfe86848 | 2128 | set_nlink(inode, 2); /* Use getattr to fix if necessary */ |
444ceed8 EB |
2129 | if (p->iop) |
2130 | inode->i_op = p->iop; | |
2131 | if (p->fop) | |
2132 | inode->i_fop = p->fop; | |
2133 | ei->op = p->op; | |
fb045adb | 2134 | d_set_d_op(dentry, &pid_dentry_operations); |
444ceed8 EB |
2135 | d_add(dentry, inode); |
2136 | /* Close the race of the process dying before we return the dentry */ | |
0b728e19 | 2137 | if (pid_revalidate(dentry, 0)) |
c52a47ac | 2138 | return 0; |
444ceed8 | 2139 | out: |
c52a47ac | 2140 | return -ENOENT; |
444ceed8 EB |
2141 | } |
2142 | ||
1da177e4 LT |
2143 | static struct dentry *proc_pident_lookup(struct inode *dir, |
2144 | struct dentry *dentry, | |
c5141e6d | 2145 | const struct pid_entry *ents, |
7bcd6b0e | 2146 | unsigned int nents) |
1da177e4 | 2147 | { |
c52a47ac | 2148 | int error; |
99f89551 | 2149 | struct task_struct *task = get_proc_task(dir); |
c5141e6d | 2150 | const struct pid_entry *p, *last; |
1da177e4 | 2151 | |
c52a47ac | 2152 | error = -ENOENT; |
1da177e4 | 2153 | |
99f89551 EB |
2154 | if (!task) |
2155 | goto out_no_task; | |
1da177e4 | 2156 | |
20cdc894 EB |
2157 | /* |
2158 | * Yes, it does not scale. And it should not. Don't add | |
2159 | * new entries into /proc/<tgid>/ without very good reasons. | |
2160 | */ | |
7bcd6b0e EB |
2161 | last = &ents[nents - 1]; |
2162 | for (p = ents; p <= last; p++) { | |
1da177e4 LT |
2163 | if (p->len != dentry->d_name.len) |
2164 | continue; | |
2165 | if (!memcmp(dentry->d_name.name, p->name, p->len)) | |
2166 | break; | |
2167 | } | |
7bcd6b0e | 2168 | if (p > last) |
1da177e4 LT |
2169 | goto out; |
2170 | ||
444ceed8 | 2171 | error = proc_pident_instantiate(dir, dentry, task, p); |
1da177e4 | 2172 | out: |
99f89551 EB |
2173 | put_task_struct(task); |
2174 | out_no_task: | |
c52a47ac | 2175 | return ERR_PTR(error); |
1da177e4 LT |
2176 | } |
2177 | ||
f0c3b509 | 2178 | static int proc_pident_readdir(struct file *file, struct dir_context *ctx, |
c5141e6d | 2179 | const struct pid_entry *ents, unsigned int nents) |
28a6d671 | 2180 | { |
f0c3b509 AV |
2181 | struct task_struct *task = get_proc_task(file_inode(file)); |
2182 | const struct pid_entry *p; | |
28a6d671 | 2183 | |
28a6d671 | 2184 | if (!task) |
f0c3b509 | 2185 | return -ENOENT; |
28a6d671 | 2186 | |
f0c3b509 AV |
2187 | if (!dir_emit_dots(file, ctx)) |
2188 | goto out; | |
2189 | ||
2190 | if (ctx->pos >= nents + 2) | |
2191 | goto out; | |
28a6d671 | 2192 | |
f0c3b509 AV |
2193 | for (p = ents + (ctx->pos - 2); p <= ents + nents - 1; p++) { |
2194 | if (!proc_fill_cache(file, ctx, p->name, p->len, | |
2195 | proc_pident_instantiate, task, p)) | |
2196 | break; | |
2197 | ctx->pos++; | |
2198 | } | |
28a6d671 | 2199 | out: |
61a28784 | 2200 | put_task_struct(task); |
f0c3b509 | 2201 | return 0; |
1da177e4 LT |
2202 | } |
2203 | ||
28a6d671 EB |
2204 | #ifdef CONFIG_SECURITY |
2205 | static ssize_t proc_pid_attr_read(struct file * file, char __user * buf, | |
2206 | size_t count, loff_t *ppos) | |
2207 | { | |
496ad9aa | 2208 | struct inode * inode = file_inode(file); |
04ff9708 | 2209 | char *p = NULL; |
28a6d671 EB |
2210 | ssize_t length; |
2211 | struct task_struct *task = get_proc_task(inode); | |
2212 | ||
28a6d671 | 2213 | if (!task) |
04ff9708 | 2214 | return -ESRCH; |
28a6d671 EB |
2215 | |
2216 | length = security_getprocattr(task, | |
2fddfeef | 2217 | (char*)file->f_path.dentry->d_name.name, |
04ff9708 | 2218 | &p); |
28a6d671 | 2219 | put_task_struct(task); |
04ff9708 AV |
2220 | if (length > 0) |
2221 | length = simple_read_from_buffer(buf, count, ppos, p, length); | |
2222 | kfree(p); | |
28a6d671 | 2223 | return length; |
1da177e4 LT |
2224 | } |
2225 | ||
28a6d671 EB |
2226 | static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf, |
2227 | size_t count, loff_t *ppos) | |
2228 | { | |
496ad9aa | 2229 | struct inode * inode = file_inode(file); |
28a6d671 EB |
2230 | char *page; |
2231 | ssize_t length; | |
2232 | struct task_struct *task = get_proc_task(inode); | |
2233 | ||
2234 | length = -ESRCH; | |
2235 | if (!task) | |
2236 | goto out_no_task; | |
2237 | if (count > PAGE_SIZE) | |
2238 | count = PAGE_SIZE; | |
2239 | ||
2240 | /* No partial writes. */ | |
2241 | length = -EINVAL; | |
2242 | if (*ppos != 0) | |
2243 | goto out; | |
2244 | ||
2245 | length = -ENOMEM; | |
e12ba74d | 2246 | page = (char*)__get_free_page(GFP_TEMPORARY); |
28a6d671 EB |
2247 | if (!page) |
2248 | goto out; | |
2249 | ||
2250 | length = -EFAULT; | |
2251 | if (copy_from_user(page, buf, count)) | |
2252 | goto out_free; | |
2253 | ||
107db7c7 | 2254 | /* Guard against adverse ptrace interaction */ |
9b1bf12d | 2255 | length = mutex_lock_interruptible(&task->signal->cred_guard_mutex); |
107db7c7 DH |
2256 | if (length < 0) |
2257 | goto out_free; | |
2258 | ||
28a6d671 | 2259 | length = security_setprocattr(task, |
2fddfeef | 2260 | (char*)file->f_path.dentry->d_name.name, |
28a6d671 | 2261 | (void*)page, count); |
9b1bf12d | 2262 | mutex_unlock(&task->signal->cred_guard_mutex); |
28a6d671 EB |
2263 | out_free: |
2264 | free_page((unsigned long) page); | |
2265 | out: | |
2266 | put_task_struct(task); | |
2267 | out_no_task: | |
2268 | return length; | |
2269 | } | |
2270 | ||
00977a59 | 2271 | static const struct file_operations proc_pid_attr_operations = { |
28a6d671 EB |
2272 | .read = proc_pid_attr_read, |
2273 | .write = proc_pid_attr_write, | |
87df8424 | 2274 | .llseek = generic_file_llseek, |
28a6d671 EB |
2275 | }; |
2276 | ||
c5141e6d | 2277 | static const struct pid_entry attr_dir_stuff[] = { |
631f9c18 AD |
2278 | REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations), |
2279 | REG("prev", S_IRUGO, proc_pid_attr_operations), | |
2280 | REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2281 | REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2282 | REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
2283 | REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations), | |
28a6d671 EB |
2284 | }; |
2285 | ||
f0c3b509 | 2286 | static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx) |
28a6d671 | 2287 | { |
f0c3b509 AV |
2288 | return proc_pident_readdir(file, ctx, |
2289 | attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2290 | } |
2291 | ||
00977a59 | 2292 | static const struct file_operations proc_attr_dir_operations = { |
1da177e4 | 2293 | .read = generic_read_dir, |
f0c3b509 | 2294 | .iterate = proc_attr_dir_readdir, |
6038f373 | 2295 | .llseek = default_llseek, |
1da177e4 LT |
2296 | }; |
2297 | ||
72d9dcfc | 2298 | static struct dentry *proc_attr_dir_lookup(struct inode *dir, |
00cd8dd3 | 2299 | struct dentry *dentry, unsigned int flags) |
28a6d671 | 2300 | { |
7bcd6b0e EB |
2301 | return proc_pident_lookup(dir, dentry, |
2302 | attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2303 | } |
2304 | ||
c5ef1c42 | 2305 | static const struct inode_operations proc_attr_dir_inode_operations = { |
72d9dcfc | 2306 | .lookup = proc_attr_dir_lookup, |
99f89551 | 2307 | .getattr = pid_getattr, |
6d76fa58 | 2308 | .setattr = proc_setattr, |
1da177e4 LT |
2309 | }; |
2310 | ||
28a6d671 EB |
2311 | #endif |
2312 | ||
698ba7b5 | 2313 | #ifdef CONFIG_ELF_CORE |
3cb4a0bb KH |
2314 | static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf, |
2315 | size_t count, loff_t *ppos) | |
2316 | { | |
496ad9aa | 2317 | struct task_struct *task = get_proc_task(file_inode(file)); |
3cb4a0bb KH |
2318 | struct mm_struct *mm; |
2319 | char buffer[PROC_NUMBUF]; | |
2320 | size_t len; | |
2321 | int ret; | |
2322 | ||
2323 | if (!task) | |
2324 | return -ESRCH; | |
2325 | ||
2326 | ret = 0; | |
2327 | mm = get_task_mm(task); | |
2328 | if (mm) { | |
2329 | len = snprintf(buffer, sizeof(buffer), "%08lx\n", | |
2330 | ((mm->flags & MMF_DUMP_FILTER_MASK) >> | |
2331 | MMF_DUMP_FILTER_SHIFT)); | |
2332 | mmput(mm); | |
2333 | ret = simple_read_from_buffer(buf, count, ppos, buffer, len); | |
2334 | } | |
2335 | ||
2336 | put_task_struct(task); | |
2337 | ||
2338 | return ret; | |
2339 | } | |
2340 | ||
2341 | static ssize_t proc_coredump_filter_write(struct file *file, | |
2342 | const char __user *buf, | |
2343 | size_t count, | |
2344 | loff_t *ppos) | |
2345 | { | |
2346 | struct task_struct *task; | |
2347 | struct mm_struct *mm; | |
2348 | char buffer[PROC_NUMBUF], *end; | |
2349 | unsigned int val; | |
2350 | int ret; | |
2351 | int i; | |
2352 | unsigned long mask; | |
2353 | ||
2354 | ret = -EFAULT; | |
2355 | memset(buffer, 0, sizeof(buffer)); | |
2356 | if (count > sizeof(buffer) - 1) | |
2357 | count = sizeof(buffer) - 1; | |
2358 | if (copy_from_user(buffer, buf, count)) | |
2359 | goto out_no_task; | |
2360 | ||
2361 | ret = -EINVAL; | |
2362 | val = (unsigned int)simple_strtoul(buffer, &end, 0); | |
2363 | if (*end == '\n') | |
2364 | end++; | |
2365 | if (end - buffer == 0) | |
2366 | goto out_no_task; | |
2367 | ||
2368 | ret = -ESRCH; | |
496ad9aa | 2369 | task = get_proc_task(file_inode(file)); |
3cb4a0bb KH |
2370 | if (!task) |
2371 | goto out_no_task; | |
2372 | ||
2373 | ret = end - buffer; | |
2374 | mm = get_task_mm(task); | |
2375 | if (!mm) | |
2376 | goto out_no_mm; | |
2377 | ||
2378 | for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) { | |
2379 | if (val & mask) | |
2380 | set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2381 | else | |
2382 | clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2383 | } | |
2384 | ||
2385 | mmput(mm); | |
2386 | out_no_mm: | |
2387 | put_task_struct(task); | |
2388 | out_no_task: | |
2389 | return ret; | |
2390 | } | |
2391 | ||
2392 | static const struct file_operations proc_coredump_filter_operations = { | |
2393 | .read = proc_coredump_filter_read, | |
2394 | .write = proc_coredump_filter_write, | |
87df8424 | 2395 | .llseek = generic_file_llseek, |
3cb4a0bb KH |
2396 | }; |
2397 | #endif | |
2398 | ||
aba76fdb | 2399 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
297c5d92 AR |
2400 | static int do_io_accounting(struct task_struct *task, char *buffer, int whole) |
2401 | { | |
940389b8 | 2402 | struct task_io_accounting acct = task->ioac; |
5995477a | 2403 | unsigned long flags; |
293eb1e7 | 2404 | int result; |
5995477a | 2405 | |
293eb1e7 VK |
2406 | result = mutex_lock_killable(&task->signal->cred_guard_mutex); |
2407 | if (result) | |
2408 | return result; | |
2409 | ||
2410 | if (!ptrace_may_access(task, PTRACE_MODE_READ)) { | |
2411 | result = -EACCES; | |
2412 | goto out_unlock; | |
2413 | } | |
1d1221f3 | 2414 | |
5995477a AR |
2415 | if (whole && lock_task_sighand(task, &flags)) { |
2416 | struct task_struct *t = task; | |
2417 | ||
2418 | task_io_accounting_add(&acct, &task->signal->ioac); | |
2419 | while_each_thread(task, t) | |
2420 | task_io_accounting_add(&acct, &t->ioac); | |
2421 | ||
2422 | unlock_task_sighand(task, &flags); | |
297c5d92 | 2423 | } |
293eb1e7 | 2424 | result = sprintf(buffer, |
aba76fdb AM |
2425 | "rchar: %llu\n" |
2426 | "wchar: %llu\n" | |
2427 | "syscr: %llu\n" | |
2428 | "syscw: %llu\n" | |
2429 | "read_bytes: %llu\n" | |
2430 | "write_bytes: %llu\n" | |
2431 | "cancelled_write_bytes: %llu\n", | |
7c44319d AB |
2432 | (unsigned long long)acct.rchar, |
2433 | (unsigned long long)acct.wchar, | |
2434 | (unsigned long long)acct.syscr, | |
2435 | (unsigned long long)acct.syscw, | |
2436 | (unsigned long long)acct.read_bytes, | |
2437 | (unsigned long long)acct.write_bytes, | |
2438 | (unsigned long long)acct.cancelled_write_bytes); | |
293eb1e7 VK |
2439 | out_unlock: |
2440 | mutex_unlock(&task->signal->cred_guard_mutex); | |
2441 | return result; | |
297c5d92 AR |
2442 | } |
2443 | ||
2444 | static int proc_tid_io_accounting(struct task_struct *task, char *buffer) | |
2445 | { | |
2446 | return do_io_accounting(task, buffer, 0); | |
aba76fdb | 2447 | } |
297c5d92 AR |
2448 | |
2449 | static int proc_tgid_io_accounting(struct task_struct *task, char *buffer) | |
2450 | { | |
2451 | return do_io_accounting(task, buffer, 1); | |
2452 | } | |
2453 | #endif /* CONFIG_TASK_IO_ACCOUNTING */ | |
aba76fdb | 2454 | |
22d917d8 EB |
2455 | #ifdef CONFIG_USER_NS |
2456 | static int proc_id_map_open(struct inode *inode, struct file *file, | |
ccf94f1b | 2457 | const struct seq_operations *seq_ops) |
22d917d8 EB |
2458 | { |
2459 | struct user_namespace *ns = NULL; | |
2460 | struct task_struct *task; | |
2461 | struct seq_file *seq; | |
2462 | int ret = -EINVAL; | |
2463 | ||
2464 | task = get_proc_task(inode); | |
2465 | if (task) { | |
2466 | rcu_read_lock(); | |
2467 | ns = get_user_ns(task_cred_xxx(task, user_ns)); | |
2468 | rcu_read_unlock(); | |
2469 | put_task_struct(task); | |
2470 | } | |
2471 | if (!ns) | |
2472 | goto err; | |
2473 | ||
2474 | ret = seq_open(file, seq_ops); | |
2475 | if (ret) | |
2476 | goto err_put_ns; | |
2477 | ||
2478 | seq = file->private_data; | |
2479 | seq->private = ns; | |
2480 | ||
2481 | return 0; | |
2482 | err_put_ns: | |
2483 | put_user_ns(ns); | |
2484 | err: | |
2485 | return ret; | |
2486 | } | |
2487 | ||
2488 | static int proc_id_map_release(struct inode *inode, struct file *file) | |
2489 | { | |
2490 | struct seq_file *seq = file->private_data; | |
2491 | struct user_namespace *ns = seq->private; | |
2492 | put_user_ns(ns); | |
2493 | return seq_release(inode, file); | |
2494 | } | |
2495 | ||
2496 | static int proc_uid_map_open(struct inode *inode, struct file *file) | |
2497 | { | |
2498 | return proc_id_map_open(inode, file, &proc_uid_seq_operations); | |
2499 | } | |
2500 | ||
2501 | static int proc_gid_map_open(struct inode *inode, struct file *file) | |
2502 | { | |
2503 | return proc_id_map_open(inode, file, &proc_gid_seq_operations); | |
2504 | } | |
2505 | ||
f76d207a EB |
2506 | static int proc_projid_map_open(struct inode *inode, struct file *file) |
2507 | { | |
2508 | return proc_id_map_open(inode, file, &proc_projid_seq_operations); | |
2509 | } | |
2510 | ||
22d917d8 EB |
2511 | static const struct file_operations proc_uid_map_operations = { |
2512 | .open = proc_uid_map_open, | |
2513 | .write = proc_uid_map_write, | |
2514 | .read = seq_read, | |
2515 | .llseek = seq_lseek, | |
2516 | .release = proc_id_map_release, | |
2517 | }; | |
2518 | ||
2519 | static const struct file_operations proc_gid_map_operations = { | |
2520 | .open = proc_gid_map_open, | |
2521 | .write = proc_gid_map_write, | |
2522 | .read = seq_read, | |
2523 | .llseek = seq_lseek, | |
2524 | .release = proc_id_map_release, | |
2525 | }; | |
f76d207a EB |
2526 | |
2527 | static const struct file_operations proc_projid_map_operations = { | |
2528 | .open = proc_projid_map_open, | |
2529 | .write = proc_projid_map_write, | |
2530 | .read = seq_read, | |
2531 | .llseek = seq_lseek, | |
2532 | .release = proc_id_map_release, | |
2533 | }; | |
22d917d8 EB |
2534 | #endif /* CONFIG_USER_NS */ |
2535 | ||
47830723 KC |
2536 | static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns, |
2537 | struct pid *pid, struct task_struct *task) | |
2538 | { | |
a9712bc1 AV |
2539 | int err = lock_trace(task); |
2540 | if (!err) { | |
2541 | seq_printf(m, "%08x\n", task->personality); | |
2542 | unlock_trace(task); | |
2543 | } | |
2544 | return err; | |
47830723 KC |
2545 | } |
2546 | ||
28a6d671 EB |
2547 | /* |
2548 | * Thread groups | |
2549 | */ | |
00977a59 | 2550 | static const struct file_operations proc_task_operations; |
c5ef1c42 | 2551 | static const struct inode_operations proc_task_inode_operations; |
20cdc894 | 2552 | |
c5141e6d | 2553 | static const struct pid_entry tgid_base_stuff[] = { |
631f9c18 AD |
2554 | DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations), |
2555 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), | |
640708a2 PE |
2556 | #ifdef CONFIG_CHECKPOINT_RESTORE |
2557 | DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations), | |
2558 | #endif | |
631f9c18 | 2559 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 2560 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
b2211a36 | 2561 | #ifdef CONFIG_NET |
631f9c18 | 2562 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), |
b2211a36 | 2563 | #endif |
631f9c18 | 2564 | REG("environ", S_IRUSR, proc_environ_operations), |
f9ea536e | 2565 | ONE("auxv", S_IRUSR, proc_pid_auxv), |
631f9c18 | 2566 | ONE("status", S_IRUGO, proc_pid_status), |
35a35046 | 2567 | ONE("personality", S_IRUSR, proc_pid_personality), |
1c963eb1 | 2568 | ONE("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 2569 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 2570 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
5091faa4 MG |
2571 | #endif |
2572 | #ifdef CONFIG_SCHED_AUTOGROUP | |
2573 | REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations), | |
ebcb6734 | 2574 | #endif |
4614a696 | 2575 | REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations), |
ebcb6734 | 2576 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 | 2577 | ONE("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 2578 | #endif |
2ca66ff7 | 2579 | ONE("cmdline", S_IRUGO, proc_pid_cmdline), |
631f9c18 AD |
2580 | ONE("stat", S_IRUGO, proc_tgid_stat), |
2581 | ONE("statm", S_IRUGO, proc_pid_statm), | |
b7643757 | 2582 | REG("maps", S_IRUGO, proc_pid_maps_operations), |
28a6d671 | 2583 | #ifdef CONFIG_NUMA |
b7643757 | 2584 | REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations), |
28a6d671 | 2585 | #endif |
631f9c18 AD |
2586 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
2587 | LNK("cwd", proc_cwd_link), | |
2588 | LNK("root", proc_root_link), | |
2589 | LNK("exe", proc_exe_link), | |
2590 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
2591 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
2592 | REG("mountstats", S_IRUSR, proc_mountstats_operations), | |
1e883281 | 2593 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 2594 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
b7643757 | 2595 | REG("smaps", S_IRUGO, proc_pid_smaps_operations), |
32ed74a4 | 2596 | REG("pagemap", S_IRUSR, proc_pagemap_operations), |
28a6d671 EB |
2597 | #endif |
2598 | #ifdef CONFIG_SECURITY | |
631f9c18 | 2599 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
2600 | #endif |
2601 | #ifdef CONFIG_KALLSYMS | |
edfcd606 | 2602 | ONE("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 2603 | #endif |
2ec220e2 | 2604 | #ifdef CONFIG_STACKTRACE |
35a35046 | 2605 | ONE("stack", S_IRUSR, proc_pid_stack), |
28a6d671 EB |
2606 | #endif |
2607 | #ifdef CONFIG_SCHEDSTATS | |
f6e826ca | 2608 | ONE("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 2609 | #endif |
9745512c | 2610 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 2611 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 2612 | #endif |
8793d854 | 2613 | #ifdef CONFIG_PROC_PID_CPUSET |
631f9c18 | 2614 | REG("cpuset", S_IRUGO, proc_cpuset_operations), |
a424316c PM |
2615 | #endif |
2616 | #ifdef CONFIG_CGROUPS | |
631f9c18 | 2617 | REG("cgroup", S_IRUGO, proc_cgroup_operations), |
28a6d671 | 2618 | #endif |
631f9c18 | 2619 | INF("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 2620 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 2621 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
28a6d671 | 2622 | #ifdef CONFIG_AUDITSYSCALL |
631f9c18 AD |
2623 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
2624 | REG("sessionid", S_IRUGO, proc_sessionid_operations), | |
28a6d671 | 2625 | #endif |
f4f154fd | 2626 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 2627 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
f4f154fd | 2628 | #endif |
698ba7b5 | 2629 | #ifdef CONFIG_ELF_CORE |
631f9c18 | 2630 | REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations), |
3cb4a0bb | 2631 | #endif |
aba76fdb | 2632 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
1d1221f3 | 2633 | INF("io", S_IRUSR, proc_tgid_io_accounting), |
aba76fdb | 2634 | #endif |
f133ecca CM |
2635 | #ifdef CONFIG_HARDWALL |
2636 | INF("hardwall", S_IRUGO, proc_pid_hardwall), | |
2637 | #endif | |
22d917d8 EB |
2638 | #ifdef CONFIG_USER_NS |
2639 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
2640 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 2641 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
22d917d8 | 2642 | #endif |
48f6a7a5 PE |
2643 | #ifdef CONFIG_CHECKPOINT_RESTORE |
2644 | REG("timers", S_IRUGO, proc_timers_operations), | |
2645 | #endif | |
28a6d671 | 2646 | }; |
1da177e4 | 2647 | |
f0c3b509 | 2648 | static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 2649 | { |
f0c3b509 AV |
2650 | return proc_pident_readdir(file, ctx, |
2651 | tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
2652 | } |
2653 | ||
00977a59 | 2654 | static const struct file_operations proc_tgid_base_operations = { |
1da177e4 | 2655 | .read = generic_read_dir, |
f0c3b509 | 2656 | .iterate = proc_tgid_base_readdir, |
6038f373 | 2657 | .llseek = default_llseek, |
1da177e4 LT |
2658 | }; |
2659 | ||
00cd8dd3 AV |
2660 | static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
2661 | { | |
7bcd6b0e EB |
2662 | return proc_pident_lookup(dir, dentry, |
2663 | tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
2664 | } |
2665 | ||
c5ef1c42 | 2666 | static const struct inode_operations proc_tgid_base_inode_operations = { |
28a6d671 | 2667 | .lookup = proc_tgid_base_lookup, |
99f89551 | 2668 | .getattr = pid_getattr, |
6d76fa58 | 2669 | .setattr = proc_setattr, |
0499680a | 2670 | .permission = proc_pid_permission, |
1da177e4 | 2671 | }; |
1da177e4 | 2672 | |
60347f67 | 2673 | static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid) |
1da177e4 | 2674 | { |
48e6484d | 2675 | struct dentry *dentry, *leader, *dir; |
8578cea7 | 2676 | char buf[PROC_NUMBUF]; |
48e6484d EB |
2677 | struct qstr name; |
2678 | ||
2679 | name.name = buf; | |
60347f67 | 2680 | name.len = snprintf(buf, sizeof(buf), "%d", pid); |
4f522a24 | 2681 | /* no ->d_hash() rejects on procfs */ |
60347f67 | 2682 | dentry = d_hash_and_lookup(mnt->mnt_root, &name); |
48e6484d | 2683 | if (dentry) { |
29f12ca3 | 2684 | shrink_dcache_parent(dentry); |
48e6484d EB |
2685 | d_drop(dentry); |
2686 | dput(dentry); | |
2687 | } | |
1da177e4 | 2688 | |
48e6484d | 2689 | name.name = buf; |
60347f67 PE |
2690 | name.len = snprintf(buf, sizeof(buf), "%d", tgid); |
2691 | leader = d_hash_and_lookup(mnt->mnt_root, &name); | |
48e6484d EB |
2692 | if (!leader) |
2693 | goto out; | |
1da177e4 | 2694 | |
48e6484d EB |
2695 | name.name = "task"; |
2696 | name.len = strlen(name.name); | |
2697 | dir = d_hash_and_lookup(leader, &name); | |
2698 | if (!dir) | |
2699 | goto out_put_leader; | |
2700 | ||
2701 | name.name = buf; | |
60347f67 | 2702 | name.len = snprintf(buf, sizeof(buf), "%d", pid); |
48e6484d EB |
2703 | dentry = d_hash_and_lookup(dir, &name); |
2704 | if (dentry) { | |
2705 | shrink_dcache_parent(dentry); | |
2706 | d_drop(dentry); | |
2707 | dput(dentry); | |
1da177e4 | 2708 | } |
48e6484d EB |
2709 | |
2710 | dput(dir); | |
2711 | out_put_leader: | |
2712 | dput(leader); | |
2713 | out: | |
2714 | return; | |
1da177e4 LT |
2715 | } |
2716 | ||
0895e91d RD |
2717 | /** |
2718 | * proc_flush_task - Remove dcache entries for @task from the /proc dcache. | |
2719 | * @task: task that should be flushed. | |
2720 | * | |
2721 | * When flushing dentries from proc, one needs to flush them from global | |
60347f67 | 2722 | * proc (proc_mnt) and from all the namespaces' procs this task was seen |
0895e91d RD |
2723 | * in. This call is supposed to do all of this job. |
2724 | * | |
2725 | * Looks in the dcache for | |
2726 | * /proc/@pid | |
2727 | * /proc/@tgid/task/@pid | |
2728 | * if either directory is present flushes it and all of it'ts children | |
2729 | * from the dcache. | |
2730 | * | |
2731 | * It is safe and reasonable to cache /proc entries for a task until | |
2732 | * that task exits. After that they just clog up the dcache with | |
2733 | * useless entries, possibly causing useful dcache entries to be | |
2734 | * flushed instead. This routine is proved to flush those useless | |
2735 | * dcache entries at process exit time. | |
2736 | * | |
2737 | * NOTE: This routine is just an optimization so it does not guarantee | |
2738 | * that no dcache entries will exist at process exit time it | |
2739 | * just makes it very unlikely that any will persist. | |
60347f67 PE |
2740 | */ |
2741 | ||
2742 | void proc_flush_task(struct task_struct *task) | |
2743 | { | |
9fcc2d15 | 2744 | int i; |
9b4d1cbe | 2745 | struct pid *pid, *tgid; |
130f77ec PE |
2746 | struct upid *upid; |
2747 | ||
130f77ec | 2748 | pid = task_pid(task); |
9b4d1cbe | 2749 | tgid = task_tgid(task); |
130f77ec | 2750 | |
9fcc2d15 | 2751 | for (i = 0; i <= pid->level; i++) { |
130f77ec PE |
2752 | upid = &pid->numbers[i]; |
2753 | proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr, | |
9b4d1cbe | 2754 | tgid->numbers[i].nr); |
130f77ec | 2755 | } |
60347f67 PE |
2756 | } |
2757 | ||
c52a47ac AV |
2758 | static int proc_pid_instantiate(struct inode *dir, |
2759 | struct dentry * dentry, | |
2760 | struct task_struct *task, const void *ptr) | |
444ceed8 | 2761 | { |
444ceed8 EB |
2762 | struct inode *inode; |
2763 | ||
61a28784 | 2764 | inode = proc_pid_make_inode(dir->i_sb, task); |
444ceed8 EB |
2765 | if (!inode) |
2766 | goto out; | |
2767 | ||
2768 | inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO; | |
2769 | inode->i_op = &proc_tgid_base_inode_operations; | |
2770 | inode->i_fop = &proc_tgid_base_operations; | |
2771 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 | 2772 | |
bfe86848 MS |
2773 | set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff, |
2774 | ARRAY_SIZE(tgid_base_stuff))); | |
444ceed8 | 2775 | |
fb045adb | 2776 | d_set_d_op(dentry, &pid_dentry_operations); |
444ceed8 EB |
2777 | |
2778 | d_add(dentry, inode); | |
2779 | /* Close the race of the process dying before we return the dentry */ | |
0b728e19 | 2780 | if (pid_revalidate(dentry, 0)) |
c52a47ac | 2781 | return 0; |
444ceed8 | 2782 | out: |
c52a47ac | 2783 | return -ENOENT; |
444ceed8 EB |
2784 | } |
2785 | ||
00cd8dd3 | 2786 | struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags) |
1da177e4 | 2787 | { |
335eb531 | 2788 | int result = -ENOENT; |
1da177e4 | 2789 | struct task_struct *task; |
1da177e4 | 2790 | unsigned tgid; |
b488893a | 2791 | struct pid_namespace *ns; |
1da177e4 | 2792 | |
dbcdb504 | 2793 | tgid = name_to_int(&dentry->d_name); |
1da177e4 LT |
2794 | if (tgid == ~0U) |
2795 | goto out; | |
2796 | ||
b488893a | 2797 | ns = dentry->d_sb->s_fs_info; |
de758734 | 2798 | rcu_read_lock(); |
b488893a | 2799 | task = find_task_by_pid_ns(tgid, ns); |
1da177e4 LT |
2800 | if (task) |
2801 | get_task_struct(task); | |
de758734 | 2802 | rcu_read_unlock(); |
1da177e4 LT |
2803 | if (!task) |
2804 | goto out; | |
2805 | ||
444ceed8 | 2806 | result = proc_pid_instantiate(dir, dentry, task, NULL); |
1da177e4 | 2807 | put_task_struct(task); |
1da177e4 | 2808 | out: |
c52a47ac | 2809 | return ERR_PTR(result); |
1da177e4 LT |
2810 | } |
2811 | ||
1da177e4 | 2812 | /* |
0804ef4b | 2813 | * Find the first task with tgid >= tgid |
0bc58a91 | 2814 | * |
1da177e4 | 2815 | */ |
19fd4bb2 EB |
2816 | struct tgid_iter { |
2817 | unsigned int tgid; | |
0804ef4b | 2818 | struct task_struct *task; |
19fd4bb2 EB |
2819 | }; |
2820 | static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter) | |
2821 | { | |
0804ef4b | 2822 | struct pid *pid; |
1da177e4 | 2823 | |
19fd4bb2 EB |
2824 | if (iter.task) |
2825 | put_task_struct(iter.task); | |
454cc105 | 2826 | rcu_read_lock(); |
0804ef4b | 2827 | retry: |
19fd4bb2 EB |
2828 | iter.task = NULL; |
2829 | pid = find_ge_pid(iter.tgid, ns); | |
0804ef4b | 2830 | if (pid) { |
19fd4bb2 EB |
2831 | iter.tgid = pid_nr_ns(pid, ns); |
2832 | iter.task = pid_task(pid, PIDTYPE_PID); | |
0804ef4b EB |
2833 | /* What we to know is if the pid we have find is the |
2834 | * pid of a thread_group_leader. Testing for task | |
2835 | * being a thread_group_leader is the obvious thing | |
2836 | * todo but there is a window when it fails, due to | |
2837 | * the pid transfer logic in de_thread. | |
2838 | * | |
2839 | * So we perform the straight forward test of seeing | |
2840 | * if the pid we have found is the pid of a thread | |
2841 | * group leader, and don't worry if the task we have | |
2842 | * found doesn't happen to be a thread group leader. | |
2843 | * As we don't care in the case of readdir. | |
2844 | */ | |
19fd4bb2 EB |
2845 | if (!iter.task || !has_group_leader_pid(iter.task)) { |
2846 | iter.tgid += 1; | |
0804ef4b | 2847 | goto retry; |
19fd4bb2 EB |
2848 | } |
2849 | get_task_struct(iter.task); | |
0bc58a91 | 2850 | } |
454cc105 | 2851 | rcu_read_unlock(); |
19fd4bb2 | 2852 | return iter; |
1da177e4 LT |
2853 | } |
2854 | ||
021ada7d | 2855 | #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 1) |
0804ef4b | 2856 | |
1da177e4 | 2857 | /* for the /proc/ directory itself, after non-process stuff has been done */ |
f0c3b509 | 2858 | int proc_pid_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 2859 | { |
19fd4bb2 | 2860 | struct tgid_iter iter; |
db963164 | 2861 | struct pid_namespace *ns = file->f_dentry->d_sb->s_fs_info; |
f0c3b509 | 2862 | loff_t pos = ctx->pos; |
1da177e4 | 2863 | |
021ada7d | 2864 | if (pos >= PID_MAX_LIMIT + TGID_OFFSET) |
f0c3b509 | 2865 | return 0; |
1da177e4 | 2866 | |
021ada7d | 2867 | if (pos == TGID_OFFSET - 1) { |
db963164 AV |
2868 | struct inode *inode = ns->proc_self->d_inode; |
2869 | if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK)) | |
f0c3b509 | 2870 | return 0; |
021ada7d AV |
2871 | iter.tgid = 0; |
2872 | } else { | |
2873 | iter.tgid = pos - TGID_OFFSET; | |
2874 | } | |
19fd4bb2 | 2875 | iter.task = NULL; |
19fd4bb2 EB |
2876 | for (iter = next_tgid(ns, iter); |
2877 | iter.task; | |
2878 | iter.tgid += 1, iter = next_tgid(ns, iter)) { | |
f0c3b509 AV |
2879 | char name[PROC_NUMBUF]; |
2880 | int len; | |
2881 | if (!has_pid_permissions(ns, iter.task, 2)) | |
2882 | continue; | |
0499680a | 2883 | |
f0c3b509 AV |
2884 | len = snprintf(name, sizeof(name), "%d", iter.tgid); |
2885 | ctx->pos = iter.tgid + TGID_OFFSET; | |
2886 | if (!proc_fill_cache(file, ctx, name, len, | |
2887 | proc_pid_instantiate, iter.task, NULL)) { | |
19fd4bb2 | 2888 | put_task_struct(iter.task); |
f0c3b509 | 2889 | return 0; |
1da177e4 | 2890 | } |
0bc58a91 | 2891 | } |
f0c3b509 | 2892 | ctx->pos = PID_MAX_LIMIT + TGID_OFFSET; |
0bc58a91 EB |
2893 | return 0; |
2894 | } | |
1da177e4 | 2895 | |
28a6d671 EB |
2896 | /* |
2897 | * Tasks | |
2898 | */ | |
c5141e6d | 2899 | static const struct pid_entry tid_base_stuff[] = { |
631f9c18 | 2900 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), |
3835541d | 2901 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 2902 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
631f9c18 | 2903 | REG("environ", S_IRUSR, proc_environ_operations), |
f9ea536e | 2904 | ONE("auxv", S_IRUSR, proc_pid_auxv), |
631f9c18 | 2905 | ONE("status", S_IRUGO, proc_pid_status), |
35a35046 | 2906 | ONE("personality", S_IRUSR, proc_pid_personality), |
1c963eb1 | 2907 | ONE("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 2908 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 2909 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
ebcb6734 | 2910 | #endif |
4614a696 | 2911 | REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations), |
ebcb6734 | 2912 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 | 2913 | ONE("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 2914 | #endif |
2ca66ff7 | 2915 | ONE("cmdline", S_IRUGO, proc_pid_cmdline), |
631f9c18 AD |
2916 | ONE("stat", S_IRUGO, proc_tid_stat), |
2917 | ONE("statm", S_IRUGO, proc_pid_statm), | |
b7643757 | 2918 | REG("maps", S_IRUGO, proc_tid_maps_operations), |
81841161 CG |
2919 | #ifdef CONFIG_CHECKPOINT_RESTORE |
2920 | REG("children", S_IRUGO, proc_tid_children_operations), | |
2921 | #endif | |
28a6d671 | 2922 | #ifdef CONFIG_NUMA |
b7643757 | 2923 | REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations), |
28a6d671 | 2924 | #endif |
631f9c18 AD |
2925 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
2926 | LNK("cwd", proc_cwd_link), | |
2927 | LNK("root", proc_root_link), | |
2928 | LNK("exe", proc_exe_link), | |
2929 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
2930 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
1e883281 | 2931 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 2932 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
b7643757 | 2933 | REG("smaps", S_IRUGO, proc_tid_smaps_operations), |
32ed74a4 | 2934 | REG("pagemap", S_IRUSR, proc_pagemap_operations), |
28a6d671 EB |
2935 | #endif |
2936 | #ifdef CONFIG_SECURITY | |
631f9c18 | 2937 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
2938 | #endif |
2939 | #ifdef CONFIG_KALLSYMS | |
edfcd606 | 2940 | ONE("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 2941 | #endif |
2ec220e2 | 2942 | #ifdef CONFIG_STACKTRACE |
35a35046 | 2943 | ONE("stack", S_IRUSR, proc_pid_stack), |
28a6d671 EB |
2944 | #endif |
2945 | #ifdef CONFIG_SCHEDSTATS | |
f6e826ca | 2946 | ONE("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 2947 | #endif |
9745512c | 2948 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 2949 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 2950 | #endif |
8793d854 | 2951 | #ifdef CONFIG_PROC_PID_CPUSET |
631f9c18 | 2952 | REG("cpuset", S_IRUGO, proc_cpuset_operations), |
a424316c PM |
2953 | #endif |
2954 | #ifdef CONFIG_CGROUPS | |
631f9c18 | 2955 | REG("cgroup", S_IRUGO, proc_cgroup_operations), |
28a6d671 | 2956 | #endif |
631f9c18 | 2957 | INF("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 2958 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 2959 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
28a6d671 | 2960 | #ifdef CONFIG_AUDITSYSCALL |
631f9c18 | 2961 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
26ec3c64 | 2962 | REG("sessionid", S_IRUGO, proc_sessionid_operations), |
28a6d671 | 2963 | #endif |
f4f154fd | 2964 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 2965 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
f4f154fd | 2966 | #endif |
297c5d92 | 2967 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
1d1221f3 | 2968 | INF("io", S_IRUSR, proc_tid_io_accounting), |
297c5d92 | 2969 | #endif |
f133ecca CM |
2970 | #ifdef CONFIG_HARDWALL |
2971 | INF("hardwall", S_IRUGO, proc_pid_hardwall), | |
2972 | #endif | |
22d917d8 EB |
2973 | #ifdef CONFIG_USER_NS |
2974 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
2975 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 2976 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
22d917d8 | 2977 | #endif |
28a6d671 EB |
2978 | }; |
2979 | ||
f0c3b509 | 2980 | static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx) |
28a6d671 | 2981 | { |
f0c3b509 AV |
2982 | return proc_pident_readdir(file, ctx, |
2983 | tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
2984 | } |
2985 | ||
00cd8dd3 AV |
2986 | static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
2987 | { | |
7bcd6b0e EB |
2988 | return proc_pident_lookup(dir, dentry, |
2989 | tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
2990 | } |
2991 | ||
00977a59 | 2992 | static const struct file_operations proc_tid_base_operations = { |
28a6d671 | 2993 | .read = generic_read_dir, |
f0c3b509 | 2994 | .iterate = proc_tid_base_readdir, |
6038f373 | 2995 | .llseek = default_llseek, |
28a6d671 EB |
2996 | }; |
2997 | ||
c5ef1c42 | 2998 | static const struct inode_operations proc_tid_base_inode_operations = { |
28a6d671 EB |
2999 | .lookup = proc_tid_base_lookup, |
3000 | .getattr = pid_getattr, | |
3001 | .setattr = proc_setattr, | |
3002 | }; | |
3003 | ||
c52a47ac | 3004 | static int proc_task_instantiate(struct inode *dir, |
c5141e6d | 3005 | struct dentry *dentry, struct task_struct *task, const void *ptr) |
444ceed8 | 3006 | { |
444ceed8 | 3007 | struct inode *inode; |
61a28784 | 3008 | inode = proc_pid_make_inode(dir->i_sb, task); |
444ceed8 EB |
3009 | |
3010 | if (!inode) | |
3011 | goto out; | |
3012 | inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO; | |
3013 | inode->i_op = &proc_tid_base_inode_operations; | |
3014 | inode->i_fop = &proc_tid_base_operations; | |
3015 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 | 3016 | |
bfe86848 MS |
3017 | set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff, |
3018 | ARRAY_SIZE(tid_base_stuff))); | |
444ceed8 | 3019 | |
fb045adb | 3020 | d_set_d_op(dentry, &pid_dentry_operations); |
444ceed8 EB |
3021 | |
3022 | d_add(dentry, inode); | |
3023 | /* Close the race of the process dying before we return the dentry */ | |
0b728e19 | 3024 | if (pid_revalidate(dentry, 0)) |
c52a47ac | 3025 | return 0; |
444ceed8 | 3026 | out: |
c52a47ac | 3027 | return -ENOENT; |
444ceed8 EB |
3028 | } |
3029 | ||
00cd8dd3 | 3030 | static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags) |
28a6d671 | 3031 | { |
c52a47ac | 3032 | int result = -ENOENT; |
28a6d671 EB |
3033 | struct task_struct *task; |
3034 | struct task_struct *leader = get_proc_task(dir); | |
28a6d671 | 3035 | unsigned tid; |
b488893a | 3036 | struct pid_namespace *ns; |
28a6d671 EB |
3037 | |
3038 | if (!leader) | |
3039 | goto out_no_task; | |
3040 | ||
dbcdb504 | 3041 | tid = name_to_int(&dentry->d_name); |
28a6d671 EB |
3042 | if (tid == ~0U) |
3043 | goto out; | |
3044 | ||
b488893a | 3045 | ns = dentry->d_sb->s_fs_info; |
28a6d671 | 3046 | rcu_read_lock(); |
b488893a | 3047 | task = find_task_by_pid_ns(tid, ns); |
28a6d671 EB |
3048 | if (task) |
3049 | get_task_struct(task); | |
3050 | rcu_read_unlock(); | |
3051 | if (!task) | |
3052 | goto out; | |
bac0abd6 | 3053 | if (!same_thread_group(leader, task)) |
28a6d671 EB |
3054 | goto out_drop_task; |
3055 | ||
444ceed8 | 3056 | result = proc_task_instantiate(dir, dentry, task, NULL); |
28a6d671 EB |
3057 | out_drop_task: |
3058 | put_task_struct(task); | |
3059 | out: | |
3060 | put_task_struct(leader); | |
3061 | out_no_task: | |
c52a47ac | 3062 | return ERR_PTR(result); |
28a6d671 EB |
3063 | } |
3064 | ||
0bc58a91 EB |
3065 | /* |
3066 | * Find the first tid of a thread group to return to user space. | |
3067 | * | |
3068 | * Usually this is just the thread group leader, but if the users | |
3069 | * buffer was too small or there was a seek into the middle of the | |
3070 | * directory we have more work todo. | |
3071 | * | |
3072 | * In the case of a short read we start with find_task_by_pid. | |
3073 | * | |
3074 | * In the case of a seek we start with the leader and walk nr | |
3075 | * threads past it. | |
3076 | */ | |
9f6e963f ON |
3077 | static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos, |
3078 | struct pid_namespace *ns) | |
0bc58a91 | 3079 | { |
d855a4b7 | 3080 | struct task_struct *pos, *task; |
9f6e963f ON |
3081 | unsigned long nr = f_pos; |
3082 | ||
3083 | if (nr != f_pos) /* 32bit overflow? */ | |
3084 | return NULL; | |
1da177e4 | 3085 | |
cc288738 | 3086 | rcu_read_lock(); |
d855a4b7 ON |
3087 | task = pid_task(pid, PIDTYPE_PID); |
3088 | if (!task) | |
3089 | goto fail; | |
3090 | ||
3091 | /* Attempt to start with the tid of a thread */ | |
9f6e963f | 3092 | if (tid && nr) { |
b488893a | 3093 | pos = find_task_by_pid_ns(tid, ns); |
d855a4b7 | 3094 | if (pos && same_thread_group(pos, task)) |
a872ff0c | 3095 | goto found; |
0bc58a91 | 3096 | } |
1da177e4 | 3097 | |
0bc58a91 | 3098 | /* If nr exceeds the number of threads there is nothing todo */ |
9f6e963f | 3099 | if (nr >= get_nr_threads(task)) |
c986c14a | 3100 | goto fail; |
1da177e4 | 3101 | |
a872ff0c ON |
3102 | /* If we haven't found our starting place yet start |
3103 | * with the leader and walk nr threads forward. | |
0bc58a91 | 3104 | */ |
d855a4b7 | 3105 | pos = task = task->group_leader; |
c986c14a | 3106 | do { |
9f6e963f | 3107 | if (!nr--) |
c986c14a | 3108 | goto found; |
d855a4b7 | 3109 | } while_each_thread(task, pos); |
c986c14a ON |
3110 | fail: |
3111 | pos = NULL; | |
3112 | goto out; | |
a872ff0c ON |
3113 | found: |
3114 | get_task_struct(pos); | |
3115 | out: | |
cc288738 | 3116 | rcu_read_unlock(); |
0bc58a91 EB |
3117 | return pos; |
3118 | } | |
3119 | ||
3120 | /* | |
3121 | * Find the next thread in the thread list. | |
3122 | * Return NULL if there is an error or no next thread. | |
3123 | * | |
3124 | * The reference to the input task_struct is released. | |
3125 | */ | |
3126 | static struct task_struct *next_tid(struct task_struct *start) | |
3127 | { | |
c1df7fb8 | 3128 | struct task_struct *pos = NULL; |
cc288738 | 3129 | rcu_read_lock(); |
c1df7fb8 | 3130 | if (pid_alive(start)) { |
0bc58a91 | 3131 | pos = next_thread(start); |
c1df7fb8 ON |
3132 | if (thread_group_leader(pos)) |
3133 | pos = NULL; | |
3134 | else | |
3135 | get_task_struct(pos); | |
3136 | } | |
cc288738 | 3137 | rcu_read_unlock(); |
0bc58a91 EB |
3138 | put_task_struct(start); |
3139 | return pos; | |
1da177e4 LT |
3140 | } |
3141 | ||
3142 | /* for the /proc/TGID/task/ directories */ | |
f0c3b509 | 3143 | static int proc_task_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3144 | { |
d855a4b7 ON |
3145 | struct inode *inode = file_inode(file); |
3146 | struct task_struct *task; | |
b488893a | 3147 | struct pid_namespace *ns; |
f0c3b509 | 3148 | int tid; |
1da177e4 | 3149 | |
d855a4b7 | 3150 | if (proc_inode_is_dead(inode)) |
f0c3b509 | 3151 | return -ENOENT; |
1da177e4 | 3152 | |
f0c3b509 | 3153 | if (!dir_emit_dots(file, ctx)) |
d855a4b7 | 3154 | return 0; |
1da177e4 | 3155 | |
0bc58a91 EB |
3156 | /* f_version caches the tgid value that the last readdir call couldn't |
3157 | * return. lseek aka telldir automagically resets f_version to 0. | |
3158 | */ | |
f0c3b509 AV |
3159 | ns = file->f_dentry->d_sb->s_fs_info; |
3160 | tid = (int)file->f_version; | |
3161 | file->f_version = 0; | |
d855a4b7 | 3162 | for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns); |
0bc58a91 | 3163 | task; |
f0c3b509 AV |
3164 | task = next_tid(task), ctx->pos++) { |
3165 | char name[PROC_NUMBUF]; | |
3166 | int len; | |
b488893a | 3167 | tid = task_pid_nr_ns(task, ns); |
f0c3b509 AV |
3168 | len = snprintf(name, sizeof(name), "%d", tid); |
3169 | if (!proc_fill_cache(file, ctx, name, len, | |
3170 | proc_task_instantiate, task, NULL)) { | |
0bc58a91 EB |
3171 | /* returning this tgid failed, save it as the first |
3172 | * pid for the next readir call */ | |
f0c3b509 | 3173 | file->f_version = (u64)tid; |
0bc58a91 | 3174 | put_task_struct(task); |
1da177e4 | 3175 | break; |
0bc58a91 | 3176 | } |
1da177e4 | 3177 | } |
d855a4b7 | 3178 | |
f0c3b509 | 3179 | return 0; |
1da177e4 | 3180 | } |
6e66b52b EB |
3181 | |
3182 | static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat) | |
3183 | { | |
3184 | struct inode *inode = dentry->d_inode; | |
99f89551 | 3185 | struct task_struct *p = get_proc_task(inode); |
6e66b52b EB |
3186 | generic_fillattr(inode, stat); |
3187 | ||
99f89551 | 3188 | if (p) { |
99f89551 | 3189 | stat->nlink += get_nr_threads(p); |
99f89551 | 3190 | put_task_struct(p); |
6e66b52b EB |
3191 | } |
3192 | ||
3193 | return 0; | |
3194 | } | |
28a6d671 | 3195 | |
c5ef1c42 | 3196 | static const struct inode_operations proc_task_inode_operations = { |
28a6d671 EB |
3197 | .lookup = proc_task_lookup, |
3198 | .getattr = proc_task_getattr, | |
3199 | .setattr = proc_setattr, | |
0499680a | 3200 | .permission = proc_pid_permission, |
28a6d671 EB |
3201 | }; |
3202 | ||
00977a59 | 3203 | static const struct file_operations proc_task_operations = { |
28a6d671 | 3204 | .read = generic_read_dir, |
f0c3b509 | 3205 | .iterate = proc_task_readdir, |
6038f373 | 3206 | .llseek = default_llseek, |
28a6d671 | 3207 | }; |