?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge branches 'core-fixes-for-linus' and 'irq-fixes-for-linus' of git://git.kernel...
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / proc / base.c
blobdc8bca72b002d9b01036bb7e659495eb972af322
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.
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.
48 */
49
50 #include <asm/uaccess.h>
51
52 #include <linux/errno.h>
53 #include <linux/time.h>
54 #include <linux/proc_fs.h>
55 #include <linux/stat.h>
56 #include <linux/task_io_accounting_ops.h>
57 #include <linux/init.h>
58 #include <linux/capability.h>
59 #include <linux/file.h>
60 #include <linux/fdtable.h>
61 #include <linux/string.h>
62 #include <linux/seq_file.h>
63 #include <linux/namei.h>
64 #include <linux/mnt_namespace.h>
65 #include <linux/mm.h>
66 #include <linux/swap.h>
67 #include <linux/rcupdate.h>
68 #include <linux/kallsyms.h>
69 #include <linux/stacktrace.h>
70 #include <linux/resource.h>
71 #include <linux/module.h>
72 #include <linux/mount.h>
73 #include <linux/security.h>
74 #include <linux/ptrace.h>
75 #include <linux/tracehook.h>
76 #include <linux/cgroup.h>
77 #include <linux/cpuset.h>
78 #include <linux/audit.h>
79 #include <linux/poll.h>
80 #include <linux/nsproxy.h>
81 #include <linux/oom.h>
82 #include <linux/elf.h>
83 #include <linux/pid_namespace.h>
84 #include <linux/fs_struct.h>
85 #include <linux/slab.h>
86 #include "internal.h"
87
88 /* NOTE:
89 * Implementing inode permission operations in /proc is almost
90 * certainly an error. Permission checks need to happen during
91 * each system call not at open time. The reason is that most of
92 * what we wish to check for permissions in /proc varies at runtime.
93 *
94 * The classic example of a problem is opening file descriptors
95 * in /proc for a task before it execs a suid executable.
96 */
97
98 struct pid_entry {
99 char *name;
100 int len;
101 mode_t mode;
102 const struct inode_operations *iop;
103 const struct file_operations *fop;
104 union proc_op op;
105 };
106
107 #define NOD(NAME, MODE, IOP, FOP, OP) { \
108 .name = (NAME), \
109 .len = sizeof(NAME) - 1, \
110 .mode = MODE, \
111 .iop = IOP, \
112 .fop = FOP, \
113 .op = OP, \
114 }
115
116 #define DIR(NAME, MODE, iops, fops) \
117 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
118 #define LNK(NAME, get_link) \
119 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
120 &proc_pid_link_inode_operations, NULL, \
121 { .proc_get_link = get_link } )
122 #define REG(NAME, MODE, fops) \
123 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
124 #define INF(NAME, MODE, read) \
125 NOD(NAME, (S_IFREG|(MODE)), \
126 NULL, &proc_info_file_operations, \
127 { .proc_read = read } )
128 #define ONE(NAME, MODE, show) \
129 NOD(NAME, (S_IFREG|(MODE)), \
130 NULL, &proc_single_file_operations, \
131 { .proc_show = show } )
132
133 /*
134 * Count the number of hardlinks for the pid_entry table, excluding the .
135 * and .. links.
136 */
137 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
138 unsigned int n)
139 {
140 unsigned int i;
141 unsigned int count;
142
143 count = 0;
144 for (i = 0; i < n; ++i) {
145 if (S_ISDIR(entries[i].mode))
146 ++count;
147 }
148
149 return count;
150 }
151
152 static int get_task_root(struct task_struct *task, struct path *root)
153 {
154 int result = -ENOENT;
155
156 task_lock(task);
157 if (task->fs) {
158 get_fs_root(task->fs, root);
159 result = 0;
160 }
161 task_unlock(task);
162 return result;
163 }
164
165 static int proc_cwd_link(struct inode *inode, struct path *path)
166 {
167 struct task_struct *task = get_proc_task(inode);
168 int result = -ENOENT;
169
170 if (task) {
171 task_lock(task);
172 if (task->fs) {
173 get_fs_pwd(task->fs, path);
174 result = 0;
175 }
176 task_unlock(task);
177 put_task_struct(task);
178 }
179 return result;
180 }
181
182 static int proc_root_link(struct inode *inode, struct path *path)
183 {
184 struct task_struct *task = get_proc_task(inode);
185 int result = -ENOENT;
186
187 if (task) {
188 result = get_task_root(task, path);
189 put_task_struct(task);
190 }
191 return result;
192 }
193
194 static struct mm_struct *__check_mem_permission(struct task_struct *task)
195 {
196 struct mm_struct *mm;
197
198 mm = get_task_mm(task);
199 if (!mm)
200 return ERR_PTR(-EINVAL);
201
202 /*
203 * A task can always look at itself, in case it chooses
204 * to use system calls instead of load instructions.
205 */
206 if (task == current)
207 return mm;
208
209 /*
210 * If current is actively ptrace'ing, and would also be
211 * permitted to freshly attach with ptrace now, permit it.
212 */
213 if (task_is_stopped_or_traced(task)) {
214 int match;
215 rcu_read_lock();
216 match = (tracehook_tracer_task(task) == current);
217 rcu_read_unlock();
218 if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH))
219 return mm;
220 }
221
222 /*
223 * No one else is allowed.
224 */
225 mmput(mm);
226 return ERR_PTR(-EPERM);
227 }
228
229 /*
230 * If current may access user memory in @task return a reference to the
231 * corresponding mm, otherwise ERR_PTR.
232 */
233 static struct mm_struct *check_mem_permission(struct task_struct *task)
234 {
235 struct mm_struct *mm;
236 int err;
237
238 /*
239 * Avoid racing if task exec's as we might get a new mm but validate
240 * against old credentials.
241 */
242 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
243 if (err)
244 return ERR_PTR(err);
245
246 mm = __check_mem_permission(task);
247 mutex_unlock(&task->signal->cred_guard_mutex);
248
249 return mm;
250 }
251
252 struct mm_struct *mm_for_maps(struct task_struct *task)
253 {
254 struct mm_struct *mm;
255 int err;
256
257 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
258 if (err)
259 return ERR_PTR(err);
260
261 mm = get_task_mm(task);
262 if (mm && mm != current->mm &&
263 !ptrace_may_access(task, PTRACE_MODE_READ)) {
264 mmput(mm);
265 mm = ERR_PTR(-EACCES);
266 }
267 mutex_unlock(&task->signal->cred_guard_mutex);
268
269 return mm;
270 }
271
272 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
273 {
274 int res = 0;
275 unsigned int len;
276 struct mm_struct *mm = get_task_mm(task);
277 if (!mm)
278 goto out;
279 if (!mm->arg_end)
280 goto out_mm; /* Shh! No looking before we're done */
281
282 len = mm->arg_end - mm->arg_start;
283
284 if (len > PAGE_SIZE)
285 len = PAGE_SIZE;
286
287 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
288
289 // If the nul at the end of args has been overwritten, then
290 // assume application is using setproctitle(3).
291 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
292 len = strnlen(buffer, res);
293 if (len < res) {
294 res = len;
295 } else {
296 len = mm->env_end - mm->env_start;
297 if (len > PAGE_SIZE - res)
298 len = PAGE_SIZE - res;
299 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
300 res = strnlen(buffer, res);
301 }
302 }
303 out_mm:
304 mmput(mm);
305 out:
306 return res;
307 }
308
309 static int proc_pid_auxv(struct task_struct *task, char *buffer)
310 {
311 struct mm_struct *mm = mm_for_maps(task);
312 int res = PTR_ERR(mm);
313 if (mm && !IS_ERR(mm)) {
314 unsigned int nwords = 0;
315 do {
316 nwords += 2;
317 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
318 res = nwords * sizeof(mm->saved_auxv[0]);
319 if (res > PAGE_SIZE)
320 res = PAGE_SIZE;
321 memcpy(buffer, mm->saved_auxv, res);
322 mmput(mm);
323 }
324 return res;
325 }
326
327
328 #ifdef CONFIG_KALLSYMS
329 /*
330 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
331 * Returns the resolved symbol. If that fails, simply return the address.
332 */
333 static int proc_pid_wchan(struct task_struct *task, char *buffer)
334 {
335 unsigned long wchan;
336 char symname[KSYM_NAME_LEN];
337
338 wchan = get_wchan(task);
339
340 if (lookup_symbol_name(wchan, symname) < 0)
341 if (!ptrace_may_access(task, PTRACE_MODE_READ))
342 return 0;
343 else
344 return sprintf(buffer, "%lu", wchan);
345 else
346 return sprintf(buffer, "%s", symname);
347 }
348 #endif /* CONFIG_KALLSYMS */
349
350 static int lock_trace(struct task_struct *task)
351 {
352 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
353 if (err)
354 return err;
355 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
356 mutex_unlock(&task->signal->cred_guard_mutex);
357 return -EPERM;
358 }
359 return 0;
360 }
361
362 static void unlock_trace(struct task_struct *task)
363 {
364 mutex_unlock(&task->signal->cred_guard_mutex);
365 }
366
367 #ifdef CONFIG_STACKTRACE
368
369 #define MAX_STACK_TRACE_DEPTH 64
370
371 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
372 struct pid *pid, struct task_struct *task)
373 {
374 struct stack_trace trace;
375 unsigned long *entries;
376 int err;
377 int i;
378
379 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
380 if (!entries)
381 return -ENOMEM;
382
383 trace.nr_entries = 0;
384 trace.max_entries = MAX_STACK_TRACE_DEPTH;
385 trace.entries = entries;
386 trace.skip = 0;
387
388 err = lock_trace(task);
389 if (!err) {
390 save_stack_trace_tsk(task, &trace);
391
392 for (i = 0; i < trace.nr_entries; i++) {
393 seq_printf(m, "[<%pK>] %pS\n",
394 (void *)entries[i], (void *)entries[i]);
395 }
396 unlock_trace(task);
397 }
398 kfree(entries);
399
400 return err;
401 }
402 #endif
403
404 #ifdef CONFIG_SCHEDSTATS
405 /*
406 * Provides /proc/PID/schedstat
407 */
408 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
409 {
410 return sprintf(buffer, "%llu %llu %lu\n",
411 (unsigned long long)task->se.sum_exec_runtime,
412 (unsigned long long)task->sched_info.run_delay,
413 task->sched_info.pcount);
414 }
415 #endif
416
417 #ifdef CONFIG_LATENCYTOP
418 static int lstats_show_proc(struct seq_file *m, void *v)
419 {
420 int i;
421 struct inode *inode = m->private;
422 struct task_struct *task = get_proc_task(inode);
423
424 if (!task)
425 return -ESRCH;
426 seq_puts(m, "Latency Top version : v0.1\n");
427 for (i = 0; i < 32; i++) {
428 struct latency_record *lr = &task->latency_record[i];
429 if (lr->backtrace[0]) {
430 int q;
431 seq_printf(m, "%i %li %li",
432 lr->count, lr->time, lr->max);
433 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
434 unsigned long bt = lr->backtrace[q];
435 if (!bt)
436 break;
437 if (bt == ULONG_MAX)
438 break;
439 seq_printf(m, " %ps", (void *)bt);
440 }
441 seq_putc(m, '\n');
442 }
443
444 }
445 put_task_struct(task);
446 return 0;
447 }
448
449 static int lstats_open(struct inode *inode, struct file *file)
450 {
451 return single_open(file, lstats_show_proc, inode);
452 }
453
454 static ssize_t lstats_write(struct file *file, const char __user *buf,
455 size_t count, loff_t *offs)
456 {
457 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
458
459 if (!task)
460 return -ESRCH;
461 clear_all_latency_tracing(task);
462 put_task_struct(task);
463
464 return count;
465 }
466
467 static const struct file_operations proc_lstats_operations = {
468 .open = lstats_open,
469 .read = seq_read,
470 .write = lstats_write,
471 .llseek = seq_lseek,
472 .release = single_release,
473 };
474
475 #endif
476
477 static int proc_oom_score(struct task_struct *task, char *buffer)
478 {
479 unsigned long points = 0;
480
481 read_lock(&tasklist_lock);
482 if (pid_alive(task))
483 points = oom_badness(task, NULL, NULL,
484 totalram_pages + total_swap_pages);
485 read_unlock(&tasklist_lock);
486 return sprintf(buffer, "%lu\n", points);
487 }
488
489 struct limit_names {
490 char *name;
491 char *unit;
492 };
493
494 static const struct limit_names lnames[RLIM_NLIMITS] = {
495 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
496 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
497 [RLIMIT_DATA] = {"Max data size", "bytes"},
498 [RLIMIT_STACK] = {"Max stack size", "bytes"},
499 [RLIMIT_CORE] = {"Max core file size", "bytes"},
500 [RLIMIT_RSS] = {"Max resident set", "bytes"},
501 [RLIMIT_NPROC] = {"Max processes", "processes"},
502 [RLIMIT_NOFILE] = {"Max open files", "files"},
503 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
504 [RLIMIT_AS] = {"Max address space", "bytes"},
505 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
506 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
507 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
508 [RLIMIT_NICE] = {"Max nice priority", NULL},
509 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
510 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
511 };
512
513 /* Display limits for a process */
514 static int proc_pid_limits(struct task_struct *task, char *buffer)
515 {
516 unsigned int i;
517 int count = 0;
518 unsigned long flags;
519 char *bufptr = buffer;
520
521 struct rlimit rlim[RLIM_NLIMITS];
522
523 if (!lock_task_sighand(task, &flags))
524 return 0;
525 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
526 unlock_task_sighand(task, &flags);
527
528 /*
529 * print the file header
530 */
531 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
532 "Limit", "Soft Limit", "Hard Limit", "Units");
533
534 for (i = 0; i < RLIM_NLIMITS; i++) {
535 if (rlim[i].rlim_cur == RLIM_INFINITY)
536 count += sprintf(&bufptr[count], "%-25s %-20s ",
537 lnames[i].name, "unlimited");
538 else
539 count += sprintf(&bufptr[count], "%-25s %-20lu ",
540 lnames[i].name, rlim[i].rlim_cur);
541
542 if (rlim[i].rlim_max == RLIM_INFINITY)
543 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
544 else
545 count += sprintf(&bufptr[count], "%-20lu ",
546 rlim[i].rlim_max);
547
548 if (lnames[i].unit)
549 count += sprintf(&bufptr[count], "%-10s\n",
550 lnames[i].unit);
551 else
552 count += sprintf(&bufptr[count], "\n");
553 }
554
555 return count;
556 }
557
558 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
559 static int proc_pid_syscall(struct task_struct *task, char *buffer)
560 {
561 long nr;
562 unsigned long args[6], sp, pc;
563 int res = lock_trace(task);
564 if (res)
565 return res;
566
567 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
568 res = sprintf(buffer, "running\n");
569 else if (nr < 0)
570 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
571 else
572 res = sprintf(buffer,
573 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
574 nr,
575 args[0], args[1], args[2], args[3], args[4], args[5],
576 sp, pc);
577 unlock_trace(task);
578 return res;
579 }
580 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
581
582 /************************************************************************/
583 /* Here the fs part begins */
584 /************************************************************************/
585
586 /* permission checks */
587 static int proc_fd_access_allowed(struct inode *inode)
588 {
589 struct task_struct *task;
590 int allowed = 0;
591 /* Allow access to a task's file descriptors if it is us or we
592 * may use ptrace attach to the process and find out that
593 * information.
594 */
595 task = get_proc_task(inode);
596 if (task) {
597 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
598 put_task_struct(task);
599 }
600 return allowed;
601 }
602
603 int proc_setattr(struct dentry *dentry, struct iattr *attr)
604 {
605 int error;
606 struct inode *inode = dentry->d_inode;
607
608 if (attr->ia_valid & ATTR_MODE)
609 return -EPERM;
610
611 error = inode_change_ok(inode, attr);
612 if (error)
613 return error;
614
615 if ((attr->ia_valid & ATTR_SIZE) &&
616 attr->ia_size != i_size_read(inode)) {
617 error = vmtruncate(inode, attr->ia_size);
618 if (error)
619 return error;
620 }
621
622 setattr_copy(inode, attr);
623 mark_inode_dirty(inode);
624 return 0;
625 }
626
627 static const struct inode_operations proc_def_inode_operations = {
628 .setattr = proc_setattr,
629 };
630
631 static int mounts_open_common(struct inode *inode, struct file *file,
632 const struct seq_operations *op)
633 {
634 struct task_struct *task = get_proc_task(inode);
635 struct nsproxy *nsp;
636 struct mnt_namespace *ns = NULL;
637 struct path root;
638 struct proc_mounts *p;
639 int ret = -EINVAL;
640
641 if (task) {
642 rcu_read_lock();
643 nsp = task_nsproxy(task);
644 if (nsp) {
645 ns = nsp->mnt_ns;
646 if (ns)
647 get_mnt_ns(ns);
648 }
649 rcu_read_unlock();
650 if (ns && get_task_root(task, &root) == 0)
651 ret = 0;
652 put_task_struct(task);
653 }
654
655 if (!ns)
656 goto err;
657 if (ret)
658 goto err_put_ns;
659
660 ret = -ENOMEM;
661 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
662 if (!p)
663 goto err_put_path;
664
665 file->private_data = &p->m;
666 ret = seq_open(file, op);
667 if (ret)
668 goto err_free;
669
670 p->m.private = p;
671 p->ns = ns;
672 p->root = root;
673 p->event = ns->event;
674
675 return 0;
676
677 err_free:
678 kfree(p);
679 err_put_path:
680 path_put(&root);
681 err_put_ns:
682 put_mnt_ns(ns);
683 err:
684 return ret;
685 }
686
687 static int mounts_release(struct inode *inode, struct file *file)
688 {
689 struct proc_mounts *p = file->private_data;
690 path_put(&p->root);
691 put_mnt_ns(p->ns);
692 return seq_release(inode, file);
693 }
694
695 static unsigned mounts_poll(struct file *file, poll_table *wait)
696 {
697 struct proc_mounts *p = file->private_data;
698 unsigned res = POLLIN | POLLRDNORM;
699
700 poll_wait(file, &p->ns->poll, wait);
701 if (mnt_had_events(p))
702 res |= POLLERR | POLLPRI;
703
704 return res;
705 }
706
707 static int mounts_open(struct inode *inode, struct file *file)
708 {
709 return mounts_open_common(inode, file, &mounts_op);
710 }
711
712 static const struct file_operations proc_mounts_operations = {
713 .open = mounts_open,
714 .read = seq_read,
715 .llseek = seq_lseek,
716 .release = mounts_release,
717 .poll = mounts_poll,
718 };
719
720 static int mountinfo_open(struct inode *inode, struct file *file)
721 {
722 return mounts_open_common(inode, file, &mountinfo_op);
723 }
724
725 static const struct file_operations proc_mountinfo_operations = {
726 .open = mountinfo_open,
727 .read = seq_read,
728 .llseek = seq_lseek,
729 .release = mounts_release,
730 .poll = mounts_poll,
731 };
732
733 static int mountstats_open(struct inode *inode, struct file *file)
734 {
735 return mounts_open_common(inode, file, &mountstats_op);
736 }
737
738 static const struct file_operations proc_mountstats_operations = {
739 .open = mountstats_open,
740 .read = seq_read,
741 .llseek = seq_lseek,
742 .release = mounts_release,
743 };
744
745 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
746
747 static ssize_t proc_info_read(struct file * file, char __user * buf,
748 size_t count, loff_t *ppos)
749 {
750 struct inode * inode = file->f_path.dentry->d_inode;
751 unsigned long page;
752 ssize_t length;
753 struct task_struct *task = get_proc_task(inode);
754
755 length = -ESRCH;
756 if (!task)
757 goto out_no_task;
758
759 if (count > PROC_BLOCK_SIZE)
760 count = PROC_BLOCK_SIZE;
761
762 length = -ENOMEM;
763 if (!(page = __get_free_page(GFP_TEMPORARY)))
764 goto out;
765
766 length = PROC_I(inode)->op.proc_read(task, (char*)page);
767
768 if (length >= 0)
769 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
770 free_page(page);
771 out:
772 put_task_struct(task);
773 out_no_task:
774 return length;
775 }
776
777 static const struct file_operations proc_info_file_operations = {
778 .read = proc_info_read,
779 .llseek = generic_file_llseek,
780 };
781
782 static int proc_single_show(struct seq_file *m, void *v)
783 {
784 struct inode *inode = m->private;
785 struct pid_namespace *ns;
786 struct pid *pid;
787 struct task_struct *task;
788 int ret;
789
790 ns = inode->i_sb->s_fs_info;
791 pid = proc_pid(inode);
792 task = get_pid_task(pid, PIDTYPE_PID);
793 if (!task)
794 return -ESRCH;
795
796 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
797
798 put_task_struct(task);
799 return ret;
800 }
801
802 static int proc_single_open(struct inode *inode, struct file *filp)
803 {
804 return single_open(filp, proc_single_show, inode);
805 }
806
807 static const struct file_operations proc_single_file_operations = {
808 .open = proc_single_open,
809 .read = seq_read,
810 .llseek = seq_lseek,
811 .release = single_release,
812 };
813
814 static int mem_open(struct inode* inode, struct file* file)
815 {
816 file->private_data = (void*)((long)current->self_exec_id);
817 /* OK to pass negative loff_t, we can catch out-of-range */
818 file->f_mode |= FMODE_UNSIGNED_OFFSET;
819 return 0;
820 }
821
822 static ssize_t mem_read(struct file * file, char __user * buf,
823 size_t count, loff_t *ppos)
824 {
825 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
826 char *page;
827 unsigned long src = *ppos;
828 int ret = -ESRCH;
829 struct mm_struct *mm;
830
831 if (!task)
832 goto out_no_task;
833
834 ret = -ENOMEM;
835 page = (char *)__get_free_page(GFP_TEMPORARY);
836 if (!page)
837 goto out;
838
839 mm = check_mem_permission(task);
840 ret = PTR_ERR(mm);
841 if (IS_ERR(mm))
842 goto out_free;
843
844 ret = -EIO;
845
846 if (file->private_data != (void*)((long)current->self_exec_id))
847 goto out_put;
848
849 ret = 0;
850
851 while (count > 0) {
852 int this_len, retval;
853
854 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
855 retval = access_remote_vm(mm, src, page, this_len, 0);
856 if (!retval) {
857 if (!ret)
858 ret = -EIO;
859 break;
860 }
861
862 if (copy_to_user(buf, page, retval)) {
863 ret = -EFAULT;
864 break;
865 }
866
867 ret += retval;
868 src += retval;
869 buf += retval;
870 count -= retval;
871 }
872 *ppos = src;
873
874 out_put:
875 mmput(mm);
876 out_free:
877 free_page((unsigned long) page);
878 out:
879 put_task_struct(task);
880 out_no_task:
881 return ret;
882 }
883
884 static ssize_t mem_write(struct file * file, const char __user *buf,
885 size_t count, loff_t *ppos)
886 {
887 int copied;
888 char *page;
889 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
890 unsigned long dst = *ppos;
891 struct mm_struct *mm;
892
893 copied = -ESRCH;
894 if (!task)
895 goto out_no_task;
896
897 mm = check_mem_permission(task);
898 copied = PTR_ERR(mm);
899 if (IS_ERR(mm))
900 goto out_task;
901
902 copied = -EIO;
903 if (file->private_data != (void *)((long)current->self_exec_id))
904 goto out_mm;
905
906 copied = -ENOMEM;
907 page = (char *)__get_free_page(GFP_TEMPORARY);
908 if (!page)
909 goto out_mm;
910
911 copied = 0;
912 while (count > 0) {
913 int this_len, retval;
914
915 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
916 if (copy_from_user(page, buf, this_len)) {
917 copied = -EFAULT;
918 break;
919 }
920 retval = access_remote_vm(mm, dst, page, this_len, 1);
921 if (!retval) {
922 if (!copied)
923 copied = -EIO;
924 break;
925 }
926 copied += retval;
927 buf += retval;
928 dst += retval;
929 count -= retval;
930 }
931 *ppos = dst;
932 free_page((unsigned long) page);
933 out_mm:
934 mmput(mm);
935 out_task:
936 put_task_struct(task);
937 out_no_task:
938 return copied;
939 }
940
941 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
942 {
943 switch (orig) {
944 case 0:
945 file->f_pos = offset;
946 break;
947 case 1:
948 file->f_pos += offset;
949 break;
950 default:
951 return -EINVAL;
952 }
953 force_successful_syscall_return();
954 return file->f_pos;
955 }
956
957 static const struct file_operations proc_mem_operations = {
958 .llseek = mem_lseek,
959 .read = mem_read,
960 .write = mem_write,
961 .open = mem_open,
962 };
963
964 static ssize_t environ_read(struct file *file, char __user *buf,
965 size_t count, loff_t *ppos)
966 {
967 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
968 char *page;
969 unsigned long src = *ppos;
970 int ret = -ESRCH;
971 struct mm_struct *mm;
972
973 if (!task)
974 goto out_no_task;
975
976 ret = -ENOMEM;
977 page = (char *)__get_free_page(GFP_TEMPORARY);
978 if (!page)
979 goto out;
980
981
982 mm = mm_for_maps(task);
983 ret = PTR_ERR(mm);
984 if (!mm || IS_ERR(mm))
985 goto out_free;
986
987 ret = 0;
988 while (count > 0) {
989 int this_len, retval, max_len;
990
991 this_len = mm->env_end - (mm->env_start + src);
992
993 if (this_len <= 0)
994 break;
995
996 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
997 this_len = (this_len > max_len) ? max_len : this_len;
998
999 retval = access_process_vm(task, (mm->env_start + src),
1000 page, this_len, 0);
1001
1002 if (retval <= 0) {
1003 ret = retval;
1004 break;
1005 }
1006
1007 if (copy_to_user(buf, page, retval)) {
1008 ret = -EFAULT;
1009 break;
1010 }
1011
1012 ret += retval;
1013 src += retval;
1014 buf += retval;
1015 count -= retval;
1016 }
1017 *ppos = src;
1018
1019 mmput(mm);
1020 out_free:
1021 free_page((unsigned long) page);
1022 out:
1023 put_task_struct(task);
1024 out_no_task:
1025 return ret;
1026 }
1027
1028 static const struct file_operations proc_environ_operations = {
1029 .read = environ_read,
1030 .llseek = generic_file_llseek,
1031 };
1032
1033 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
1034 size_t count, loff_t *ppos)
1035 {
1036 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1037 char buffer[PROC_NUMBUF];
1038 size_t len;
1039 int oom_adjust = OOM_DISABLE;
1040 unsigned long flags;
1041
1042 if (!task)
1043 return -ESRCH;
1044
1045 if (lock_task_sighand(task, &flags)) {
1046 oom_adjust = task->signal->oom_adj;
1047 unlock_task_sighand(task, &flags);
1048 }
1049
1050 put_task_struct(task);
1051
1052 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
1053
1054 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1055 }
1056
1057 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1058 size_t count, loff_t *ppos)
1059 {
1060 struct task_struct *task;
1061 char buffer[PROC_NUMBUF];
1062 long oom_adjust;
1063 unsigned long flags;
1064 int err;
1065
1066 memset(buffer, 0, sizeof(buffer));
1067 if (count > sizeof(buffer) - 1)
1068 count = sizeof(buffer) - 1;
1069 if (copy_from_user(buffer, buf, count)) {
1070 err = -EFAULT;
1071 goto out;
1072 }
1073
1074 err = strict_strtol(strstrip(buffer), 0, &oom_adjust);
1075 if (err)
1076 goto out;
1077 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
1078 oom_adjust != OOM_DISABLE) {
1079 err = -EINVAL;
1080 goto out;
1081 }
1082
1083 task = get_proc_task(file->f_path.dentry->d_inode);
1084 if (!task) {
1085 err = -ESRCH;
1086 goto out;
1087 }
1088
1089 task_lock(task);
1090 if (!task->mm) {
1091 err = -EINVAL;
1092 goto err_task_lock;
1093 }
1094
1095 if (!lock_task_sighand(task, &flags)) {
1096 err = -ESRCH;
1097 goto err_task_lock;
1098 }
1099
1100 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
1101 err = -EACCES;
1102 goto err_sighand;
1103 }
1104
1105 if (oom_adjust != task->signal->oom_adj) {
1106 if (oom_adjust == OOM_DISABLE)
1107 atomic_inc(&task->mm->oom_disable_count);
1108 if (task->signal->oom_adj == OOM_DISABLE)
1109 atomic_dec(&task->mm->oom_disable_count);
1110 }
1111
1112 /*
1113 * Warn that /proc/pid/oom_adj is deprecated, see
1114 * Documentation/feature-removal-schedule.txt.
1115 */
1116 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, "
1117 "please use /proc/%d/oom_score_adj instead.\n",
1118 current->comm, task_pid_nr(current),
1119 task_pid_nr(task), task_pid_nr(task));
1120 task->signal->oom_adj = oom_adjust;
1121 /*
1122 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1123 * value is always attainable.
1124 */
1125 if (task->signal->oom_adj == OOM_ADJUST_MAX)
1126 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
1127 else
1128 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
1129 -OOM_DISABLE;
1130 err_sighand:
1131 unlock_task_sighand(task, &flags);
1132 err_task_lock:
1133 task_unlock(task);
1134 put_task_struct(task);
1135 out:
1136 return err < 0 ? err : count;
1137 }
1138
1139 static const struct file_operations proc_oom_adjust_operations = {
1140 .read = oom_adjust_read,
1141 .write = oom_adjust_write,
1142 .llseek = generic_file_llseek,
1143 };
1144
1145 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1146 size_t count, loff_t *ppos)
1147 {
1148 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1149 char buffer[PROC_NUMBUF];
1150 int oom_score_adj = OOM_SCORE_ADJ_MIN;
1151 unsigned long flags;
1152 size_t len;
1153
1154 if (!task)
1155 return -ESRCH;
1156 if (lock_task_sighand(task, &flags)) {
1157 oom_score_adj = task->signal->oom_score_adj;
1158 unlock_task_sighand(task, &flags);
1159 }
1160 put_task_struct(task);
1161 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1162 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1163 }
1164
1165 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1166 size_t count, loff_t *ppos)
1167 {
1168 struct task_struct *task;
1169 char buffer[PROC_NUMBUF];
1170 unsigned long flags;
1171 long oom_score_adj;
1172 int err;
1173
1174 memset(buffer, 0, sizeof(buffer));
1175 if (count > sizeof(buffer) - 1)
1176 count = sizeof(buffer) - 1;
1177 if (copy_from_user(buffer, buf, count)) {
1178 err = -EFAULT;
1179 goto out;
1180 }
1181
1182 err = strict_strtol(strstrip(buffer), 0, &oom_score_adj);
1183 if (err)
1184 goto out;
1185 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1186 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1187 err = -EINVAL;
1188 goto out;
1189 }
1190
1191 task = get_proc_task(file->f_path.dentry->d_inode);
1192 if (!task) {
1193 err = -ESRCH;
1194 goto out;
1195 }
1196
1197 task_lock(task);
1198 if (!task->mm) {
1199 err = -EINVAL;
1200 goto err_task_lock;
1201 }
1202
1203 if (!lock_task_sighand(task, &flags)) {
1204 err = -ESRCH;
1205 goto err_task_lock;
1206 }
1207
1208 if (oom_score_adj < task->signal->oom_score_adj_min &&
1209 !capable(CAP_SYS_RESOURCE)) {
1210 err = -EACCES;
1211 goto err_sighand;
1212 }
1213
1214 if (oom_score_adj != task->signal->oom_score_adj) {
1215 if (oom_score_adj == OOM_SCORE_ADJ_MIN)
1216 atomic_inc(&task->mm->oom_disable_count);
1217 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1218 atomic_dec(&task->mm->oom_disable_count);
1219 }
1220 task->signal->oom_score_adj = oom_score_adj;
1221 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1222 task->signal->oom_score_adj_min = oom_score_adj;
1223 /*
1224 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1225 * always attainable.
1226 */
1227 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1228 task->signal->oom_adj = OOM_DISABLE;
1229 else
1230 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1231 OOM_SCORE_ADJ_MAX;
1232 err_sighand:
1233 unlock_task_sighand(task, &flags);
1234 err_task_lock:
1235 task_unlock(task);
1236 put_task_struct(task);
1237 out:
1238 return err < 0 ? err : count;
1239 }
1240
1241 static const struct file_operations proc_oom_score_adj_operations = {
1242 .read = oom_score_adj_read,
1243 .write = oom_score_adj_write,
1244 .llseek = default_llseek,
1245 };
1246
1247 #ifdef CONFIG_AUDITSYSCALL
1248 #define TMPBUFLEN 21
1249 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1250 size_t count, loff_t *ppos)
1251 {
1252 struct inode * inode = file->f_path.dentry->d_inode;
1253 struct task_struct *task = get_proc_task(inode);
1254 ssize_t length;
1255 char tmpbuf[TMPBUFLEN];
1256
1257 if (!task)
1258 return -ESRCH;
1259 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1260 audit_get_loginuid(task));
1261 put_task_struct(task);
1262 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1263 }
1264
1265 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1266 size_t count, loff_t *ppos)
1267 {
1268 struct inode * inode = file->f_path.dentry->d_inode;
1269 char *page, *tmp;
1270 ssize_t length;
1271 uid_t loginuid;
1272
1273 if (!capable(CAP_AUDIT_CONTROL))
1274 return -EPERM;
1275
1276 rcu_read_lock();
1277 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1278 rcu_read_unlock();
1279 return -EPERM;
1280 }
1281 rcu_read_unlock();
1282
1283 if (count >= PAGE_SIZE)
1284 count = PAGE_SIZE - 1;
1285
1286 if (*ppos != 0) {
1287 /* No partial writes. */
1288 return -EINVAL;
1289 }
1290 page = (char*)__get_free_page(GFP_TEMPORARY);
1291 if (!page)
1292 return -ENOMEM;
1293 length = -EFAULT;
1294 if (copy_from_user(page, buf, count))
1295 goto out_free_page;
1296
1297 page[count] = '\0';
1298 loginuid = simple_strtoul(page, &tmp, 10);
1299 if (tmp == page) {
1300 length = -EINVAL;
1301 goto out_free_page;
1302
1303 }
1304 length = audit_set_loginuid(current, loginuid);
1305 if (likely(length == 0))
1306 length = count;
1307
1308 out_free_page:
1309 free_page((unsigned long) page);
1310 return length;
1311 }
1312
1313 static const struct file_operations proc_loginuid_operations = {
1314 .read = proc_loginuid_read,
1315 .write = proc_loginuid_write,
1316 .llseek = generic_file_llseek,
1317 };
1318
1319 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1320 size_t count, loff_t *ppos)
1321 {
1322 struct inode * inode = file->f_path.dentry->d_inode;
1323 struct task_struct *task = get_proc_task(inode);
1324 ssize_t length;
1325 char tmpbuf[TMPBUFLEN];
1326
1327 if (!task)
1328 return -ESRCH;
1329 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1330 audit_get_sessionid(task));
1331 put_task_struct(task);
1332 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1333 }
1334
1335 static const struct file_operations proc_sessionid_operations = {
1336 .read = proc_sessionid_read,
1337 .llseek = generic_file_llseek,
1338 };
1339 #endif
1340
1341 #ifdef CONFIG_FAULT_INJECTION
1342 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1343 size_t count, loff_t *ppos)
1344 {
1345 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1346 char buffer[PROC_NUMBUF];
1347 size_t len;
1348 int make_it_fail;
1349
1350 if (!task)
1351 return -ESRCH;
1352 make_it_fail = task->make_it_fail;
1353 put_task_struct(task);
1354
1355 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1356
1357 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1358 }
1359
1360 static ssize_t proc_fault_inject_write(struct file * file,
1361 const char __user * buf, size_t count, loff_t *ppos)
1362 {
1363 struct task_struct *task;
1364 char buffer[PROC_NUMBUF], *end;
1365 int make_it_fail;
1366
1367 if (!capable(CAP_SYS_RESOURCE))
1368 return -EPERM;
1369 memset(buffer, 0, sizeof(buffer));
1370 if (count > sizeof(buffer) - 1)
1371 count = sizeof(buffer) - 1;
1372 if (copy_from_user(buffer, buf, count))
1373 return -EFAULT;
1374 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1375 if (*end)
1376 return -EINVAL;
1377 task = get_proc_task(file->f_dentry->d_inode);
1378 if (!task)
1379 return -ESRCH;
1380 task->make_it_fail = make_it_fail;
1381 put_task_struct(task);
1382
1383 return count;
1384 }
1385
1386 static const struct file_operations proc_fault_inject_operations = {
1387 .read = proc_fault_inject_read,
1388 .write = proc_fault_inject_write,
1389 .llseek = generic_file_llseek,
1390 };
1391 #endif
1392
1393
1394 #ifdef CONFIG_SCHED_DEBUG
1395 /*
1396 * Print out various scheduling related per-task fields:
1397 */
1398 static int sched_show(struct seq_file *m, void *v)
1399 {
1400 struct inode *inode = m->private;
1401 struct task_struct *p;
1402
1403 p = get_proc_task(inode);
1404 if (!p)
1405 return -ESRCH;
1406 proc_sched_show_task(p, m);
1407
1408 put_task_struct(p);
1409
1410 return 0;
1411 }
1412
1413 static ssize_t
1414 sched_write(struct file *file, const char __user *buf,
1415 size_t count, loff_t *offset)
1416 {
1417 struct inode *inode = file->f_path.dentry->d_inode;
1418 struct task_struct *p;
1419
1420 p = get_proc_task(inode);
1421 if (!p)
1422 return -ESRCH;
1423 proc_sched_set_task(p);
1424
1425 put_task_struct(p);
1426
1427 return count;
1428 }
1429
1430 static int sched_open(struct inode *inode, struct file *filp)
1431 {
1432 return single_open(filp, sched_show, inode);
1433 }
1434
1435 static const struct file_operations proc_pid_sched_operations = {
1436 .open = sched_open,
1437 .read = seq_read,
1438 .write = sched_write,
1439 .llseek = seq_lseek,
1440 .release = single_release,
1441 };
1442
1443 #endif
1444
1445 #ifdef CONFIG_SCHED_AUTOGROUP
1446 /*
1447 * Print out autogroup related information:
1448 */
1449 static int sched_autogroup_show(struct seq_file *m, void *v)
1450 {
1451 struct inode *inode = m->private;
1452 struct task_struct *p;
1453
1454 p = get_proc_task(inode);
1455 if (!p)
1456 return -ESRCH;
1457 proc_sched_autogroup_show_task(p, m);
1458
1459 put_task_struct(p);
1460
1461 return 0;
1462 }
1463
1464 static ssize_t
1465 sched_autogroup_write(struct file *file, const char __user *buf,
1466 size_t count, loff_t *offset)
1467 {
1468 struct inode *inode = file->f_path.dentry->d_inode;
1469 struct task_struct *p;
1470 char buffer[PROC_NUMBUF];
1471 long nice;
1472 int err;
1473
1474 memset(buffer, 0, sizeof(buffer));
1475 if (count > sizeof(buffer) - 1)
1476 count = sizeof(buffer) - 1;
1477 if (copy_from_user(buffer, buf, count))
1478 return -EFAULT;
1479
1480 err = strict_strtol(strstrip(buffer), 0, &nice);
1481 if (err)
1482 return -EINVAL;
1483
1484 p = get_proc_task(inode);
1485 if (!p)
1486 return -ESRCH;
1487
1488 err = nice;
1489 err = proc_sched_autogroup_set_nice(p, &err);
1490 if (err)
1491 count = err;
1492
1493 put_task_struct(p);
1494
1495 return count;
1496 }
1497
1498 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1499 {
1500 int ret;
1501
1502 ret = single_open(filp, sched_autogroup_show, NULL);
1503 if (!ret) {
1504 struct seq_file *m = filp->private_data;
1505
1506 m->private = inode;
1507 }
1508 return ret;
1509 }
1510
1511 static const struct file_operations proc_pid_sched_autogroup_operations = {
1512 .open = sched_autogroup_open,
1513 .read = seq_read,
1514 .write = sched_autogroup_write,
1515 .llseek = seq_lseek,
1516 .release = single_release,
1517 };
1518
1519 #endif /* CONFIG_SCHED_AUTOGROUP */
1520
1521 static ssize_t comm_write(struct file *file, const char __user *buf,
1522 size_t count, loff_t *offset)
1523 {
1524 struct inode *inode = file->f_path.dentry->d_inode;
1525 struct task_struct *p;
1526 char buffer[TASK_COMM_LEN];
1527
1528 memset(buffer, 0, sizeof(buffer));
1529 if (count > sizeof(buffer) - 1)
1530 count = sizeof(buffer) - 1;
1531 if (copy_from_user(buffer, buf, count))
1532 return -EFAULT;
1533
1534 p = get_proc_task(inode);
1535 if (!p)
1536 return -ESRCH;
1537
1538 if (same_thread_group(current, p))
1539 set_task_comm(p, buffer);
1540 else
1541 count = -EINVAL;
1542
1543 put_task_struct(p);
1544
1545 return count;
1546 }
1547
1548 static int comm_show(struct seq_file *m, void *v)
1549 {
1550 struct inode *inode = m->private;
1551 struct task_struct *p;
1552
1553 p = get_proc_task(inode);
1554 if (!p)
1555 return -ESRCH;
1556
1557 task_lock(p);
1558 seq_printf(m, "%s\n", p->comm);
1559 task_unlock(p);
1560
1561 put_task_struct(p);
1562
1563 return 0;
1564 }
1565
1566 static int comm_open(struct inode *inode, struct file *filp)
1567 {
1568 return single_open(filp, comm_show, inode);
1569 }
1570
1571 static const struct file_operations proc_pid_set_comm_operations = {
1572 .open = comm_open,
1573 .read = seq_read,
1574 .write = comm_write,
1575 .llseek = seq_lseek,
1576 .release = single_release,
1577 };
1578
1579 /*
1580 * We added or removed a vma mapping the executable. The vmas are only mapped
1581 * during exec and are not mapped with the mmap system call.
1582 * Callers must hold down_write() on the mm's mmap_sem for these
1583 */
1584 void added_exe_file_vma(struct mm_struct *mm)
1585 {
1586 mm->num_exe_file_vmas++;
1587 }
1588
1589 void removed_exe_file_vma(struct mm_struct *mm)
1590 {
1591 mm->num_exe_file_vmas--;
1592 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
1593 fput(mm->exe_file);
1594 mm->exe_file = NULL;
1595 }
1596
1597 }
1598
1599 void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1600 {
1601 if (new_exe_file)
1602 get_file(new_exe_file);
1603 if (mm->exe_file)
1604 fput(mm->exe_file);
1605 mm->exe_file = new_exe_file;
1606 mm->num_exe_file_vmas = 0;
1607 }
1608
1609 struct file *get_mm_exe_file(struct mm_struct *mm)
1610 {
1611 struct file *exe_file;
1612
1613 /* We need mmap_sem to protect against races with removal of
1614 * VM_EXECUTABLE vmas */
1615 down_read(&mm->mmap_sem);
1616 exe_file = mm->exe_file;
1617 if (exe_file)
1618 get_file(exe_file);
1619 up_read(&mm->mmap_sem);
1620 return exe_file;
1621 }
1622
1623 void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
1624 {
1625 /* It's safe to write the exe_file pointer without exe_file_lock because
1626 * this is called during fork when the task is not yet in /proc */
1627 newmm->exe_file = get_mm_exe_file(oldmm);
1628 }
1629
1630 static int proc_exe_link(struct inode *inode, struct path *exe_path)
1631 {
1632 struct task_struct *task;
1633 struct mm_struct *mm;
1634 struct file *exe_file;
1635
1636 task = get_proc_task(inode);
1637 if (!task)
1638 return -ENOENT;
1639 mm = get_task_mm(task);
1640 put_task_struct(task);
1641 if (!mm)
1642 return -ENOENT;
1643 exe_file = get_mm_exe_file(mm);
1644 mmput(mm);
1645 if (exe_file) {
1646 *exe_path = exe_file->f_path;
1647 path_get(&exe_file->f_path);
1648 fput(exe_file);
1649 return 0;
1650 } else
1651 return -ENOENT;
1652 }
1653
1654 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1655 {
1656 struct inode *inode = dentry->d_inode;
1657 int error = -EACCES;
1658
1659 /* We don't need a base pointer in the /proc filesystem */
1660 path_put(&nd->path);
1661
1662 /* Are we allowed to snoop on the tasks file descriptors? */
1663 if (!proc_fd_access_allowed(inode))
1664 goto out;
1665
1666 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1667 out:
1668 return ERR_PTR(error);
1669 }
1670
1671 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1672 {
1673 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1674 char *pathname;
1675 int len;
1676
1677 if (!tmp)
1678 return -ENOMEM;
1679
1680 pathname = d_path(path, tmp, PAGE_SIZE);
1681 len = PTR_ERR(pathname);
1682 if (IS_ERR(pathname))
1683 goto out;
1684 len = tmp + PAGE_SIZE - 1 - pathname;
1685
1686 if (len > buflen)
1687 len = buflen;
1688 if (copy_to_user(buffer, pathname, len))
1689 len = -EFAULT;
1690 out:
1691 free_page((unsigned long)tmp);
1692 return len;
1693 }
1694
1695 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1696 {
1697 int error = -EACCES;
1698 struct inode *inode = dentry->d_inode;
1699 struct path path;
1700
1701 /* Are we allowed to snoop on the tasks file descriptors? */
1702 if (!proc_fd_access_allowed(inode))
1703 goto out;
1704
1705 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1706 if (error)
1707 goto out;
1708
1709 error = do_proc_readlink(&path, buffer, buflen);
1710 path_put(&path);
1711 out:
1712 return error;
1713 }
1714
1715 static const struct inode_operations proc_pid_link_inode_operations = {
1716 .readlink = proc_pid_readlink,
1717 .follow_link = proc_pid_follow_link,
1718 .setattr = proc_setattr,
1719 };
1720
1721
1722 /* building an inode */
1723
1724 static int task_dumpable(struct task_struct *task)
1725 {
1726 int dumpable = 0;
1727 struct mm_struct *mm;
1728
1729 task_lock(task);
1730 mm = task->mm;
1731 if (mm)
1732 dumpable = get_dumpable(mm);
1733 task_unlock(task);
1734 if(dumpable == 1)
1735 return 1;
1736 return 0;
1737 }
1738
1739 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1740 {
1741 struct inode * inode;
1742 struct proc_inode *ei;
1743 const struct cred *cred;
1744
1745 /* We need a new inode */
1746
1747 inode = new_inode(sb);
1748 if (!inode)
1749 goto out;
1750
1751 /* Common stuff */
1752 ei = PROC_I(inode);
1753 inode->i_ino = get_next_ino();
1754 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1755 inode->i_op = &proc_def_inode_operations;
1756
1757 /*
1758 * grab the reference to task.
1759 */
1760 ei->pid = get_task_pid(task, PIDTYPE_PID);
1761 if (!ei->pid)
1762 goto out_unlock;
1763
1764 if (task_dumpable(task)) {
1765 rcu_read_lock();
1766 cred = __task_cred(task);
1767 inode->i_uid = cred->euid;
1768 inode->i_gid = cred->egid;
1769 rcu_read_unlock();
1770 }
1771 security_task_to_inode(task, inode);
1772
1773 out:
1774 return inode;
1775
1776 out_unlock:
1777 iput(inode);
1778 return NULL;
1779 }
1780
1781 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1782 {
1783 struct inode *inode = dentry->d_inode;
1784 struct task_struct *task;
1785 const struct cred *cred;
1786
1787 generic_fillattr(inode, stat);
1788
1789 rcu_read_lock();
1790 stat->uid = 0;
1791 stat->gid = 0;
1792 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1793 if (task) {
1794 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1795 task_dumpable(task)) {
1796 cred = __task_cred(task);
1797 stat->uid = cred->euid;
1798 stat->gid = cred->egid;
1799 }
1800 }
1801 rcu_read_unlock();
1802 return 0;
1803 }
1804
1805 /* dentry stuff */
1806
1807 /*
1808 * Exceptional case: normally we are not allowed to unhash a busy
1809 * directory. In this case, however, we can do it - no aliasing problems
1810 * due to the way we treat inodes.
1811 *
1812 * Rewrite the inode's ownerships here because the owning task may have
1813 * performed a setuid(), etc.
1814 *
1815 * Before the /proc/pid/status file was created the only way to read
1816 * the effective uid of a /process was to stat /proc/pid. Reading
1817 * /proc/pid/status is slow enough that procps and other packages
1818 * kept stating /proc/pid. To keep the rules in /proc simple I have
1819 * made this apply to all per process world readable and executable
1820 * directories.
1821 */
1822 int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1823 {
1824 struct inode *inode;
1825 struct task_struct *task;
1826 const struct cred *cred;
1827
1828 if (nd && nd->flags & LOOKUP_RCU)
1829 return -ECHILD;
1830
1831 inode = dentry->d_inode;
1832 task = get_proc_task(inode);
1833
1834 if (task) {
1835 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1836 task_dumpable(task)) {
1837 rcu_read_lock();
1838 cred = __task_cred(task);
1839 inode->i_uid = cred->euid;
1840 inode->i_gid = cred->egid;
1841 rcu_read_unlock();
1842 } else {
1843 inode->i_uid = 0;
1844 inode->i_gid = 0;
1845 }
1846 inode->i_mode &= ~(S_ISUID | S_ISGID);
1847 security_task_to_inode(task, inode);
1848 put_task_struct(task);
1849 return 1;
1850 }
1851 d_drop(dentry);
1852 return 0;
1853 }
1854
1855 static int pid_delete_dentry(const struct dentry * dentry)
1856 {
1857 /* Is the task we represent dead?
1858 * If so, then don't put the dentry on the lru list,
1859 * kill it immediately.
1860 */
1861 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1862 }
1863
1864 const struct dentry_operations pid_dentry_operations =
1865 {
1866 .d_revalidate = pid_revalidate,
1867 .d_delete = pid_delete_dentry,
1868 };
1869
1870 /* Lookups */
1871
1872 /*
1873 * Fill a directory entry.
1874 *
1875 * If possible create the dcache entry and derive our inode number and
1876 * file type from dcache entry.
1877 *
1878 * Since all of the proc inode numbers are dynamically generated, the inode
1879 * numbers do not exist until the inode is cache. This means creating the
1880 * the dcache entry in readdir is necessary to keep the inode numbers
1881 * reported by readdir in sync with the inode numbers reported
1882 * by stat.
1883 */
1884 int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1885 const char *name, int len,
1886 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1887 {
1888 struct dentry *child, *dir = filp->f_path.dentry;
1889 struct inode *inode;
1890 struct qstr qname;
1891 ino_t ino = 0;
1892 unsigned type = DT_UNKNOWN;
1893
1894 qname.name = name;
1895 qname.len = len;
1896 qname.hash = full_name_hash(name, len);
1897
1898 child = d_lookup(dir, &qname);
1899 if (!child) {
1900 struct dentry *new;
1901 new = d_alloc(dir, &qname);
1902 if (new) {
1903 child = instantiate(dir->d_inode, new, task, ptr);
1904 if (child)
1905 dput(new);
1906 else
1907 child = new;
1908 }
1909 }
1910 if (!child || IS_ERR(child) || !child->d_inode)
1911 goto end_instantiate;
1912 inode = child->d_inode;
1913 if (inode) {
1914 ino = inode->i_ino;
1915 type = inode->i_mode >> 12;
1916 }
1917 dput(child);
1918 end_instantiate:
1919 if (!ino)
1920 ino = find_inode_number(dir, &qname);
1921 if (!ino)
1922 ino = 1;
1923 return filldir(dirent, name, len, filp->f_pos, ino, type);
1924 }
1925
1926 static unsigned name_to_int(struct dentry *dentry)
1927 {
1928 const char *name = dentry->d_name.name;
1929 int len = dentry->d_name.len;
1930 unsigned n = 0;
1931
1932 if (len > 1 && *name == '0')
1933 goto out;
1934 while (len-- > 0) {
1935 unsigned c = *name++ - '0';
1936 if (c > 9)
1937 goto out;
1938 if (n >= (~0U-9)/10)
1939 goto out;
1940 n *= 10;
1941 n += c;
1942 }
1943 return n;
1944 out:
1945 return ~0U;
1946 }
1947
1948 #define PROC_FDINFO_MAX 64
1949
1950 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1951 {
1952 struct task_struct *task = get_proc_task(inode);
1953 struct files_struct *files = NULL;
1954 struct file *file;
1955 int fd = proc_fd(inode);
1956
1957 if (task) {
1958 files = get_files_struct(task);
1959 put_task_struct(task);
1960 }
1961 if (files) {
1962 /*
1963 * We are not taking a ref to the file structure, so we must
1964 * hold ->file_lock.
1965 */
1966 spin_lock(&files->file_lock);
1967 file = fcheck_files(files, fd);
1968 if (file) {
1969 if (path) {
1970 *path = file->f_path;
1971 path_get(&file->f_path);
1972 }
1973 if (info)
1974 snprintf(info, PROC_FDINFO_MAX,
1975 "pos:\t%lli\n"
1976 "flags:\t0%o\n",
1977 (long long) file->f_pos,
1978 file->f_flags);
1979 spin_unlock(&files->file_lock);
1980 put_files_struct(files);
1981 return 0;
1982 }
1983 spin_unlock(&files->file_lock);
1984 put_files_struct(files);
1985 }
1986 return -ENOENT;
1987 }
1988
1989 static int proc_fd_link(struct inode *inode, struct path *path)
1990 {
1991 return proc_fd_info(inode, path, NULL);
1992 }
1993
1994 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1995 {
1996 struct inode *inode;
1997 struct task_struct *task;
1998 int fd;
1999 struct files_struct *files;
2000 const struct cred *cred;
2001
2002 if (nd && nd->flags & LOOKUP_RCU)
2003 return -ECHILD;
2004
2005 inode = dentry->d_inode;
2006 task = get_proc_task(inode);
2007 fd = proc_fd(inode);
2008
2009 if (task) {
2010 files = get_files_struct(task);
2011 if (files) {
2012 rcu_read_lock();
2013 if (fcheck_files(files, fd)) {
2014 rcu_read_unlock();
2015 put_files_struct(files);
2016 if (task_dumpable(task)) {
2017 rcu_read_lock();
2018 cred = __task_cred(task);
2019 inode->i_uid = cred->euid;
2020 inode->i_gid = cred->egid;
2021 rcu_read_unlock();
2022 } else {
2023 inode->i_uid = 0;
2024 inode->i_gid = 0;
2025 }
2026 inode->i_mode &= ~(S_ISUID | S_ISGID);
2027 security_task_to_inode(task, inode);
2028 put_task_struct(task);
2029 return 1;
2030 }
2031 rcu_read_unlock();
2032 put_files_struct(files);
2033 }
2034 put_task_struct(task);
2035 }
2036 d_drop(dentry);
2037 return 0;
2038 }
2039
2040 static const struct dentry_operations tid_fd_dentry_operations =
2041 {
2042 .d_revalidate = tid_fd_revalidate,
2043 .d_delete = pid_delete_dentry,
2044 };
2045
2046 static struct dentry *proc_fd_instantiate(struct inode *dir,
2047 struct dentry *dentry, struct task_struct *task, const void *ptr)
2048 {
2049 unsigned fd = *(const unsigned *)ptr;
2050 struct file *file;
2051 struct files_struct *files;
2052 struct inode *inode;
2053 struct proc_inode *ei;
2054 struct dentry *error = ERR_PTR(-ENOENT);
2055
2056 inode = proc_pid_make_inode(dir->i_sb, task);
2057 if (!inode)
2058 goto out;
2059 ei = PROC_I(inode);
2060 ei->fd = fd;
2061 files = get_files_struct(task);
2062 if (!files)
2063 goto out_iput;
2064 inode->i_mode = S_IFLNK;
2065
2066 /*
2067 * We are not taking a ref to the file structure, so we must
2068 * hold ->file_lock.
2069 */
2070 spin_lock(&files->file_lock);
2071 file = fcheck_files(files, fd);
2072 if (!file)
2073 goto out_unlock;
2074 if (file->f_mode & FMODE_READ)
2075 inode->i_mode |= S_IRUSR | S_IXUSR;
2076 if (file->f_mode & FMODE_WRITE)
2077 inode->i_mode |= S_IWUSR | S_IXUSR;
2078 spin_unlock(&files->file_lock);
2079 put_files_struct(files);
2080
2081 inode->i_op = &proc_pid_link_inode_operations;
2082 inode->i_size = 64;
2083 ei->op.proc_get_link = proc_fd_link;
2084 d_set_d_op(dentry, &tid_fd_dentry_operations);
2085 d_add(dentry, inode);
2086 /* Close the race of the process dying before we return the dentry */
2087 if (tid_fd_revalidate(dentry, NULL))
2088 error = NULL;
2089
2090 out:
2091 return error;
2092 out_unlock:
2093 spin_unlock(&files->file_lock);
2094 put_files_struct(files);
2095 out_iput:
2096 iput(inode);
2097 goto out;
2098 }
2099
2100 static struct dentry *proc_lookupfd_common(struct inode *dir,
2101 struct dentry *dentry,
2102 instantiate_t instantiate)
2103 {
2104 struct task_struct *task = get_proc_task(dir);
2105 unsigned fd = name_to_int(dentry);
2106 struct dentry *result = ERR_PTR(-ENOENT);
2107
2108 if (!task)
2109 goto out_no_task;
2110 if (fd == ~0U)
2111 goto out;
2112
2113 result = instantiate(dir, dentry, task, &fd);
2114 out:
2115 put_task_struct(task);
2116 out_no_task:
2117 return result;
2118 }
2119
2120 static int proc_readfd_common(struct file * filp, void * dirent,
2121 filldir_t filldir, instantiate_t instantiate)
2122 {
2123 struct dentry *dentry = filp->f_path.dentry;
2124 struct inode *inode = dentry->d_inode;
2125 struct task_struct *p = get_proc_task(inode);
2126 unsigned int fd, ino;
2127 int retval;
2128 struct files_struct * files;
2129
2130 retval = -ENOENT;
2131 if (!p)
2132 goto out_no_task;
2133 retval = 0;
2134
2135 fd = filp->f_pos;
2136 switch (fd) {
2137 case 0:
2138 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
2139 goto out;
2140 filp->f_pos++;
2141 case 1:
2142 ino = parent_ino(dentry);
2143 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2144 goto out;
2145 filp->f_pos++;
2146 default:
2147 files = get_files_struct(p);
2148 if (!files)
2149 goto out;
2150 rcu_read_lock();
2151 for (fd = filp->f_pos-2;
2152 fd < files_fdtable(files)->max_fds;
2153 fd++, filp->f_pos++) {
2154 char name[PROC_NUMBUF];
2155 int len;
2156
2157 if (!fcheck_files(files, fd))
2158 continue;
2159 rcu_read_unlock();
2160
2161 len = snprintf(name, sizeof(name), "%d", fd);
2162 if (proc_fill_cache(filp, dirent, filldir,
2163 name, len, instantiate,
2164 p, &fd) < 0) {
2165 rcu_read_lock();
2166 break;
2167 }
2168 rcu_read_lock();
2169 }
2170 rcu_read_unlock();
2171 put_files_struct(files);
2172 }
2173 out:
2174 put_task_struct(p);
2175 out_no_task:
2176 return retval;
2177 }
2178
2179 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
2180 struct nameidata *nd)
2181 {
2182 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
2183 }
2184
2185 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
2186 {
2187 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
2188 }
2189
2190 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
2191 size_t len, loff_t *ppos)
2192 {
2193 char tmp[PROC_FDINFO_MAX];
2194 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
2195 if (!err)
2196 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
2197 return err;
2198 }
2199
2200 static const struct file_operations proc_fdinfo_file_operations = {
2201 .open = nonseekable_open,
2202 .read = proc_fdinfo_read,
2203 .llseek = no_llseek,
2204 };
2205
2206 static const struct file_operations proc_fd_operations = {
2207 .read = generic_read_dir,
2208 .readdir = proc_readfd,
2209 .llseek = default_llseek,
2210 };
2211
2212 /*
2213 * /proc/pid/fd needs a special permission handler so that a process can still
2214 * access /proc/self/fd after it has executed a setuid().
2215 */
2216 static int proc_fd_permission(struct inode *inode, int mask, unsigned int flags)
2217 {
2218 int rv;
2219
2220 if (flags & IPERM_FLAG_RCU)
2221 return -ECHILD;
2222 rv = generic_permission(inode, mask, flags, NULL);
2223 if (rv == 0)
2224 return 0;
2225 if (task_pid(current) == proc_pid(inode))
2226 rv = 0;
2227 return rv;
2228 }
2229
2230 /*
2231 * proc directories can do almost nothing..
2232 */
2233 static const struct inode_operations proc_fd_inode_operations = {
2234 .lookup = proc_lookupfd,
2235 .permission = proc_fd_permission,
2236 .setattr = proc_setattr,
2237 };
2238
2239 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2240 struct dentry *dentry, struct task_struct *task, const void *ptr)
2241 {
2242 unsigned fd = *(unsigned *)ptr;
2243 struct inode *inode;
2244 struct proc_inode *ei;
2245 struct dentry *error = ERR_PTR(-ENOENT);
2246
2247 inode = proc_pid_make_inode(dir->i_sb, task);
2248 if (!inode)
2249 goto out;
2250 ei = PROC_I(inode);
2251 ei->fd = fd;
2252 inode->i_mode = S_IFREG | S_IRUSR;
2253 inode->i_fop = &proc_fdinfo_file_operations;
2254 d_set_d_op(dentry, &tid_fd_dentry_operations);
2255 d_add(dentry, inode);
2256 /* Close the race of the process dying before we return the dentry */
2257 if (tid_fd_revalidate(dentry, NULL))
2258 error = NULL;
2259
2260 out:
2261 return error;
2262 }
2263
2264 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2265 struct dentry *dentry,
2266 struct nameidata *nd)
2267 {
2268 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2269 }
2270
2271 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2272 {
2273 return proc_readfd_common(filp, dirent, filldir,
2274 proc_fdinfo_instantiate);
2275 }
2276
2277 static const struct file_operations proc_fdinfo_operations = {
2278 .read = generic_read_dir,
2279 .readdir = proc_readfdinfo,
2280 .llseek = default_llseek,
2281 };
2282
2283 /*
2284 * proc directories can do almost nothing..
2285 */
2286 static const struct inode_operations proc_fdinfo_inode_operations = {
2287 .lookup = proc_lookupfdinfo,
2288 .setattr = proc_setattr,
2289 };
2290
2291
2292 static struct dentry *proc_pident_instantiate(struct inode *dir,
2293 struct dentry *dentry, struct task_struct *task, const void *ptr)
2294 {
2295 const struct pid_entry *p = ptr;
2296 struct inode *inode;
2297 struct proc_inode *ei;
2298 struct dentry *error = ERR_PTR(-ENOENT);
2299
2300 inode = proc_pid_make_inode(dir->i_sb, task);
2301 if (!inode)
2302 goto out;
2303
2304 ei = PROC_I(inode);
2305 inode->i_mode = p->mode;
2306 if (S_ISDIR(inode->i_mode))
2307 inode->i_nlink = 2; /* Use getattr to fix if necessary */
2308 if (p->iop)
2309 inode->i_op = p->iop;
2310 if (p->fop)
2311 inode->i_fop = p->fop;
2312 ei->op = p->op;
2313 d_set_d_op(dentry, &pid_dentry_operations);
2314 d_add(dentry, inode);
2315 /* Close the race of the process dying before we return the dentry */
2316 if (pid_revalidate(dentry, NULL))
2317 error = NULL;
2318 out:
2319 return error;
2320 }
2321
2322 static struct dentry *proc_pident_lookup(struct inode *dir,
2323 struct dentry *dentry,
2324 const struct pid_entry *ents,
2325 unsigned int nents)
2326 {
2327 struct dentry *error;
2328 struct task_struct *task = get_proc_task(dir);
2329 const struct pid_entry *p, *last;
2330
2331 error = ERR_PTR(-ENOENT);
2332
2333 if (!task)
2334 goto out_no_task;
2335
2336 /*
2337 * Yes, it does not scale. And it should not. Don't add
2338 * new entries into /proc/<tgid>/ without very good reasons.
2339 */
2340 last = &ents[nents - 1];
2341 for (p = ents; p <= last; p++) {
2342 if (p->len != dentry->d_name.len)
2343 continue;
2344 if (!memcmp(dentry->d_name.name, p->name, p->len))
2345 break;
2346 }
2347 if (p > last)
2348 goto out;
2349
2350 error = proc_pident_instantiate(dir, dentry, task, p);
2351 out:
2352 put_task_struct(task);
2353 out_no_task:
2354 return error;
2355 }
2356
2357 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2358 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2359 {
2360 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2361 proc_pident_instantiate, task, p);
2362 }
2363
2364 static int proc_pident_readdir(struct file *filp,
2365 void *dirent, filldir_t filldir,
2366 const struct pid_entry *ents, unsigned int nents)
2367 {
2368 int i;
2369 struct dentry *dentry = filp->f_path.dentry;
2370 struct inode *inode = dentry->d_inode;
2371 struct task_struct *task = get_proc_task(inode);
2372 const struct pid_entry *p, *last;
2373 ino_t ino;
2374 int ret;
2375
2376 ret = -ENOENT;
2377 if (!task)
2378 goto out_no_task;
2379
2380 ret = 0;
2381 i = filp->f_pos;
2382 switch (i) {
2383 case 0:
2384 ino = inode->i_ino;
2385 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2386 goto out;
2387 i++;
2388 filp->f_pos++;
2389 /* fall through */
2390 case 1:
2391 ino = parent_ino(dentry);
2392 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2393 goto out;
2394 i++;
2395 filp->f_pos++;
2396 /* fall through */
2397 default:
2398 i -= 2;
2399 if (i >= nents) {
2400 ret = 1;
2401 goto out;
2402 }
2403 p = ents + i;
2404 last = &ents[nents - 1];
2405 while (p <= last) {
2406 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2407 goto out;
2408 filp->f_pos++;
2409 p++;
2410 }
2411 }
2412
2413 ret = 1;
2414 out:
2415 put_task_struct(task);
2416 out_no_task:
2417 return ret;
2418 }
2419
2420 #ifdef CONFIG_SECURITY
2421 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2422 size_t count, loff_t *ppos)
2423 {
2424 struct inode * inode = file->f_path.dentry->d_inode;
2425 char *p = NULL;
2426 ssize_t length;
2427 struct task_struct *task = get_proc_task(inode);
2428
2429 if (!task)
2430 return -ESRCH;
2431
2432 length = security_getprocattr(task,
2433 (char*)file->f_path.dentry->d_name.name,
2434 &p);
2435 put_task_struct(task);
2436 if (length > 0)
2437 length = simple_read_from_buffer(buf, count, ppos, p, length);
2438 kfree(p);
2439 return length;
2440 }
2441
2442 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2443 size_t count, loff_t *ppos)
2444 {
2445 struct inode * inode = file->f_path.dentry->d_inode;
2446 char *page;
2447 ssize_t length;
2448 struct task_struct *task = get_proc_task(inode);
2449
2450 length = -ESRCH;
2451 if (!task)
2452 goto out_no_task;
2453 if (count > PAGE_SIZE)
2454 count = PAGE_SIZE;
2455
2456 /* No partial writes. */
2457 length = -EINVAL;
2458 if (*ppos != 0)
2459 goto out;
2460
2461 length = -ENOMEM;
2462 page = (char*)__get_free_page(GFP_TEMPORARY);
2463 if (!page)
2464 goto out;
2465
2466 length = -EFAULT;
2467 if (copy_from_user(page, buf, count))
2468 goto out_free;
2469
2470 /* Guard against adverse ptrace interaction */
2471 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2472 if (length < 0)
2473 goto out_free;
2474
2475 length = security_setprocattr(task,
2476 (char*)file->f_path.dentry->d_name.name,
2477 (void*)page, count);
2478 mutex_unlock(&task->signal->cred_guard_mutex);
2479 out_free:
2480 free_page((unsigned long) page);
2481 out:
2482 put_task_struct(task);
2483 out_no_task:
2484 return length;
2485 }
2486
2487 static const struct file_operations proc_pid_attr_operations = {
2488 .read = proc_pid_attr_read,
2489 .write = proc_pid_attr_write,
2490 .llseek = generic_file_llseek,
2491 };
2492
2493 static const struct pid_entry attr_dir_stuff[] = {
2494 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2495 REG("prev", S_IRUGO, proc_pid_attr_operations),
2496 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2497 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2498 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2499 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2500 };
2501
2502 static int proc_attr_dir_readdir(struct file * filp,
2503 void * dirent, filldir_t filldir)
2504 {
2505 return proc_pident_readdir(filp,dirent,filldir,
2506 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2507 }
2508
2509 static const struct file_operations proc_attr_dir_operations = {
2510 .read = generic_read_dir,
2511 .readdir = proc_attr_dir_readdir,
2512 .llseek = default_llseek,
2513 };
2514
2515 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2516 struct dentry *dentry, struct nameidata *nd)
2517 {
2518 return proc_pident_lookup(dir, dentry,
2519 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2520 }
2521
2522 static const struct inode_operations proc_attr_dir_inode_operations = {
2523 .lookup = proc_attr_dir_lookup,
2524 .getattr = pid_getattr,
2525 .setattr = proc_setattr,
2526 };
2527
2528 #endif
2529
2530 #ifdef CONFIG_ELF_CORE
2531 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2532 size_t count, loff_t *ppos)
2533 {
2534 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2535 struct mm_struct *mm;
2536 char buffer[PROC_NUMBUF];
2537 size_t len;
2538 int ret;
2539
2540 if (!task)
2541 return -ESRCH;
2542
2543 ret = 0;
2544 mm = get_task_mm(task);
2545 if (mm) {
2546 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2547 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2548 MMF_DUMP_FILTER_SHIFT));
2549 mmput(mm);
2550 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2551 }
2552
2553 put_task_struct(task);
2554
2555 return ret;
2556 }
2557
2558 static ssize_t proc_coredump_filter_write(struct file *file,
2559 const char __user *buf,
2560 size_t count,
2561 loff_t *ppos)
2562 {
2563 struct task_struct *task;
2564 struct mm_struct *mm;
2565 char buffer[PROC_NUMBUF], *end;
2566 unsigned int val;
2567 int ret;
2568 int i;
2569 unsigned long mask;
2570
2571 ret = -EFAULT;
2572 memset(buffer, 0, sizeof(buffer));
2573 if (count > sizeof(buffer) - 1)
2574 count = sizeof(buffer) - 1;
2575 if (copy_from_user(buffer, buf, count))
2576 goto out_no_task;
2577
2578 ret = -EINVAL;
2579 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2580 if (*end == '\n')
2581 end++;
2582 if (end - buffer == 0)
2583 goto out_no_task;
2584
2585 ret = -ESRCH;
2586 task = get_proc_task(file->f_dentry->d_inode);
2587 if (!task)
2588 goto out_no_task;
2589
2590 ret = end - buffer;
2591 mm = get_task_mm(task);
2592 if (!mm)
2593 goto out_no_mm;
2594
2595 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2596 if (val & mask)
2597 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2598 else
2599 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2600 }
2601
2602 mmput(mm);
2603 out_no_mm:
2604 put_task_struct(task);
2605 out_no_task:
2606 return ret;
2607 }
2608
2609 static const struct file_operations proc_coredump_filter_operations = {
2610 .read = proc_coredump_filter_read,
2611 .write = proc_coredump_filter_write,
2612 .llseek = generic_file_llseek,
2613 };
2614 #endif
2615
2616 /*
2617 * /proc/self:
2618 */
2619 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2620 int buflen)
2621 {
2622 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2623 pid_t tgid = task_tgid_nr_ns(current, ns);
2624 char tmp[PROC_NUMBUF];
2625 if (!tgid)
2626 return -ENOENT;
2627 sprintf(tmp, "%d", tgid);
2628 return vfs_readlink(dentry,buffer,buflen,tmp);
2629 }
2630
2631 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2632 {
2633 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2634 pid_t tgid = task_tgid_nr_ns(current, ns);
2635 char *name = ERR_PTR(-ENOENT);
2636 if (tgid) {
2637 name = __getname();
2638 if (!name)
2639 name = ERR_PTR(-ENOMEM);
2640 else
2641 sprintf(name, "%d", tgid);
2642 }
2643 nd_set_link(nd, name);
2644 return NULL;
2645 }
2646
2647 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2648 void *cookie)
2649 {
2650 char *s = nd_get_link(nd);
2651 if (!IS_ERR(s))
2652 __putname(s);
2653 }
2654
2655 static const struct inode_operations proc_self_inode_operations = {
2656 .readlink = proc_self_readlink,
2657 .follow_link = proc_self_follow_link,
2658 .put_link = proc_self_put_link,
2659 };
2660
2661 /*
2662 * proc base
2663 *
2664 * These are the directory entries in the root directory of /proc
2665 * that properly belong to the /proc filesystem, as they describe
2666 * describe something that is process related.
2667 */
2668 static const struct pid_entry proc_base_stuff[] = {
2669 NOD("self", S_IFLNK|S_IRWXUGO,
2670 &proc_self_inode_operations, NULL, {}),
2671 };
2672
2673 static struct dentry *proc_base_instantiate(struct inode *dir,
2674 struct dentry *dentry, struct task_struct *task, const void *ptr)
2675 {
2676 const struct pid_entry *p = ptr;
2677 struct inode *inode;
2678 struct proc_inode *ei;
2679 struct dentry *error;
2680
2681 /* Allocate the inode */
2682 error = ERR_PTR(-ENOMEM);
2683 inode = new_inode(dir->i_sb);
2684 if (!inode)
2685 goto out;
2686
2687 /* Initialize the inode */
2688 ei = PROC_I(inode);
2689 inode->i_ino = get_next_ino();
2690 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2691
2692 /*
2693 * grab the reference to the task.
2694 */
2695 ei->pid = get_task_pid(task, PIDTYPE_PID);
2696 if (!ei->pid)
2697 goto out_iput;
2698
2699 inode->i_mode = p->mode;
2700 if (S_ISDIR(inode->i_mode))
2701 inode->i_nlink = 2;
2702 if (S_ISLNK(inode->i_mode))
2703 inode->i_size = 64;
2704 if (p->iop)
2705 inode->i_op = p->iop;
2706 if (p->fop)
2707 inode->i_fop = p->fop;
2708 ei->op = p->op;
2709 d_add(dentry, inode);
2710 error = NULL;
2711 out:
2712 return error;
2713 out_iput:
2714 iput(inode);
2715 goto out;
2716 }
2717
2718 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2719 {
2720 struct dentry *error;
2721 struct task_struct *task = get_proc_task(dir);
2722 const struct pid_entry *p, *last;
2723
2724 error = ERR_PTR(-ENOENT);
2725
2726 if (!task)
2727 goto out_no_task;
2728
2729 /* Lookup the directory entry */
2730 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2731 for (p = proc_base_stuff; p <= last; p++) {
2732 if (p->len != dentry->d_name.len)
2733 continue;
2734 if (!memcmp(dentry->d_name.name, p->name, p->len))
2735 break;
2736 }
2737 if (p > last)
2738 goto out;
2739
2740 error = proc_base_instantiate(dir, dentry, task, p);
2741
2742 out:
2743 put_task_struct(task);
2744 out_no_task:
2745 return error;
2746 }
2747
2748 static int proc_base_fill_cache(struct file *filp, void *dirent,
2749 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2750 {
2751 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2752 proc_base_instantiate, task, p);
2753 }
2754
2755 #ifdef CONFIG_TASK_IO_ACCOUNTING
2756 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2757 {
2758 struct task_io_accounting acct = task->ioac;
2759 unsigned long flags;
2760
2761 if (whole && lock_task_sighand(task, &flags)) {
2762 struct task_struct *t = task;
2763
2764 task_io_accounting_add(&acct, &task->signal->ioac);
2765 while_each_thread(task, t)
2766 task_io_accounting_add(&acct, &t->ioac);
2767
2768 unlock_task_sighand(task, &flags);
2769 }
2770 return sprintf(buffer,
2771 "rchar: %llu\n"
2772 "wchar: %llu\n"
2773 "syscr: %llu\n"
2774 "syscw: %llu\n"
2775 "read_bytes: %llu\n"
2776 "write_bytes: %llu\n"
2777 "cancelled_write_bytes: %llu\n",
2778 (unsigned long long)acct.rchar,
2779 (unsigned long long)acct.wchar,
2780 (unsigned long long)acct.syscr,
2781 (unsigned long long)acct.syscw,
2782 (unsigned long long)acct.read_bytes,
2783 (unsigned long long)acct.write_bytes,
2784 (unsigned long long)acct.cancelled_write_bytes);
2785 }
2786
2787 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2788 {
2789 return do_io_accounting(task, buffer, 0);
2790 }
2791
2792 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2793 {
2794 return do_io_accounting(task, buffer, 1);
2795 }
2796 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2797
2798 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2799 struct pid *pid, struct task_struct *task)
2800 {
2801 int err = lock_trace(task);
2802 if (!err) {
2803 seq_printf(m, "%08x\n", task->personality);
2804 unlock_trace(task);
2805 }
2806 return err;
2807 }
2808
2809 /*
2810 * Thread groups
2811 */
2812 static const struct file_operations proc_task_operations;
2813 static const struct inode_operations proc_task_inode_operations;
2814
2815 static const struct pid_entry tgid_base_stuff[] = {
2816 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2817 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2818 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2819 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2820 #ifdef CONFIG_NET
2821 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2822 #endif
2823 REG("environ", S_IRUSR, proc_environ_operations),
2824 INF("auxv", S_IRUSR, proc_pid_auxv),
2825 ONE("status", S_IRUGO, proc_pid_status),
2826 ONE("personality", S_IRUGO, proc_pid_personality),
2827 INF("limits", S_IRUGO, proc_pid_limits),
2828 #ifdef CONFIG_SCHED_DEBUG
2829 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2830 #endif
2831 #ifdef CONFIG_SCHED_AUTOGROUP
2832 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2833 #endif
2834 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2835 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2836 INF("syscall", S_IRUGO, proc_pid_syscall),
2837 #endif
2838 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2839 ONE("stat", S_IRUGO, proc_tgid_stat),
2840 ONE("statm", S_IRUGO, proc_pid_statm),
2841 REG("maps", S_IRUGO, proc_maps_operations),
2842 #ifdef CONFIG_NUMA
2843 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2844 #endif
2845 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2846 LNK("cwd", proc_cwd_link),
2847 LNK("root", proc_root_link),
2848 LNK("exe", proc_exe_link),
2849 REG("mounts", S_IRUGO, proc_mounts_operations),
2850 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2851 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2852 #ifdef CONFIG_PROC_PAGE_MONITOR
2853 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2854 REG("smaps", S_IRUGO, proc_smaps_operations),
2855 REG("pagemap", S_IRUGO, proc_pagemap_operations),
2856 #endif
2857 #ifdef CONFIG_SECURITY
2858 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2859 #endif
2860 #ifdef CONFIG_KALLSYMS
2861 INF("wchan", S_IRUGO, proc_pid_wchan),
2862 #endif
2863 #ifdef CONFIG_STACKTRACE
2864 ONE("stack", S_IRUGO, proc_pid_stack),
2865 #endif
2866 #ifdef CONFIG_SCHEDSTATS
2867 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2868 #endif
2869 #ifdef CONFIG_LATENCYTOP
2870 REG("latency", S_IRUGO, proc_lstats_operations),
2871 #endif
2872 #ifdef CONFIG_PROC_PID_CPUSET
2873 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2874 #endif
2875 #ifdef CONFIG_CGROUPS
2876 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2877 #endif
2878 INF("oom_score", S_IRUGO, proc_oom_score),
2879 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2880 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2881 #ifdef CONFIG_AUDITSYSCALL
2882 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2883 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2884 #endif
2885 #ifdef CONFIG_FAULT_INJECTION
2886 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2887 #endif
2888 #ifdef CONFIG_ELF_CORE
2889 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2890 #endif
2891 #ifdef CONFIG_TASK_IO_ACCOUNTING
2892 INF("io", S_IRUGO, proc_tgid_io_accounting),
2893 #endif
2894 };
2895
2896 static int proc_tgid_base_readdir(struct file * filp,
2897 void * dirent, filldir_t filldir)
2898 {
2899 return proc_pident_readdir(filp,dirent,filldir,
2900 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2901 }
2902
2903 static const struct file_operations proc_tgid_base_operations = {
2904 .read = generic_read_dir,
2905 .readdir = proc_tgid_base_readdir,
2906 .llseek = default_llseek,
2907 };
2908
2909 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2910 return proc_pident_lookup(dir, dentry,
2911 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2912 }
2913
2914 static const struct inode_operations proc_tgid_base_inode_operations = {
2915 .lookup = proc_tgid_base_lookup,
2916 .getattr = pid_getattr,
2917 .setattr = proc_setattr,
2918 };
2919
2920 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2921 {
2922 struct dentry *dentry, *leader, *dir;
2923 char buf[PROC_NUMBUF];
2924 struct qstr name;
2925
2926 name.name = buf;
2927 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2928 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2929 if (dentry) {
2930 shrink_dcache_parent(dentry);
2931 d_drop(dentry);
2932 dput(dentry);
2933 }
2934
2935 name.name = buf;
2936 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2937 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2938 if (!leader)
2939 goto out;
2940
2941 name.name = "task";
2942 name.len = strlen(name.name);
2943 dir = d_hash_and_lookup(leader, &name);
2944 if (!dir)
2945 goto out_put_leader;
2946
2947 name.name = buf;
2948 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2949 dentry = d_hash_and_lookup(dir, &name);
2950 if (dentry) {
2951 shrink_dcache_parent(dentry);
2952 d_drop(dentry);
2953 dput(dentry);
2954 }
2955
2956 dput(dir);
2957 out_put_leader:
2958 dput(leader);
2959 out:
2960 return;
2961 }
2962
2963 /**
2964 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2965 * @task: task that should be flushed.
2966 *
2967 * When flushing dentries from proc, one needs to flush them from global
2968 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2969 * in. This call is supposed to do all of this job.
2970 *
2971 * Looks in the dcache for
2972 * /proc/@pid
2973 * /proc/@tgid/task/@pid
2974 * if either directory is present flushes it and all of it'ts children
2975 * from the dcache.
2976 *
2977 * It is safe and reasonable to cache /proc entries for a task until
2978 * that task exits. After that they just clog up the dcache with
2979 * useless entries, possibly causing useful dcache entries to be
2980 * flushed instead. This routine is proved to flush those useless
2981 * dcache entries at process exit time.
2982 *
2983 * NOTE: This routine is just an optimization so it does not guarantee
2984 * that no dcache entries will exist at process exit time it
2985 * just makes it very unlikely that any will persist.
2986 */
2987
2988 void proc_flush_task(struct task_struct *task)
2989 {
2990 int i;
2991 struct pid *pid, *tgid;
2992 struct upid *upid;
2993
2994 pid = task_pid(task);
2995 tgid = task_tgid(task);
2996
2997 for (i = 0; i <= pid->level; i++) {
2998 upid = &pid->numbers[i];
2999 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
3000 tgid->numbers[i].nr);
3001 }
3002
3003 upid = &pid->numbers[pid->level];
3004 if (upid->nr == 1)
3005 pid_ns_release_proc(upid->ns);
3006 }
3007
3008 static struct dentry *proc_pid_instantiate(struct inode *dir,
3009 struct dentry * dentry,
3010 struct task_struct *task, const void *ptr)
3011 {
3012 struct dentry *error = ERR_PTR(-ENOENT);
3013 struct inode *inode;
3014
3015 inode = proc_pid_make_inode(dir->i_sb, task);
3016 if (!inode)
3017 goto out;
3018
3019 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3020 inode->i_op = &proc_tgid_base_inode_operations;
3021 inode->i_fop = &proc_tgid_base_operations;
3022 inode->i_flags|=S_IMMUTABLE;
3023
3024 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
3025 ARRAY_SIZE(tgid_base_stuff));
3026
3027 d_set_d_op(dentry, &pid_dentry_operations);
3028
3029 d_add(dentry, inode);
3030 /* Close the race of the process dying before we return the dentry */
3031 if (pid_revalidate(dentry, NULL))
3032 error = NULL;
3033 out:
3034 return error;
3035 }
3036
3037 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3038 {
3039 struct dentry *result;
3040 struct task_struct *task;
3041 unsigned tgid;
3042 struct pid_namespace *ns;
3043
3044 result = proc_base_lookup(dir, dentry);
3045 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3046 goto out;
3047
3048 tgid = name_to_int(dentry);
3049 if (tgid == ~0U)
3050 goto out;
3051
3052 ns = dentry->d_sb->s_fs_info;
3053 rcu_read_lock();
3054 task = find_task_by_pid_ns(tgid, ns);
3055 if (task)
3056 get_task_struct(task);
3057 rcu_read_unlock();
3058 if (!task)
3059 goto out;
3060
3061 result = proc_pid_instantiate(dir, dentry, task, NULL);
3062 put_task_struct(task);
3063 out:
3064 return result;
3065 }
3066
3067 /*
3068 * Find the first task with tgid >= tgid
3069 *
3070 */
3071 struct tgid_iter {
3072 unsigned int tgid;
3073 struct task_struct *task;
3074 };
3075 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3076 {
3077 struct pid *pid;
3078
3079 if (iter.task)
3080 put_task_struct(iter.task);
3081 rcu_read_lock();
3082 retry:
3083 iter.task = NULL;
3084 pid = find_ge_pid(iter.tgid, ns);
3085 if (pid) {
3086 iter.tgid = pid_nr_ns(pid, ns);
3087 iter.task = pid_task(pid, PIDTYPE_PID);
3088 /* What we to know is if the pid we have find is the
3089 * pid of a thread_group_leader. Testing for task
3090 * being a thread_group_leader is the obvious thing
3091 * todo but there is a window when it fails, due to
3092 * the pid transfer logic in de_thread.
3093 *
3094 * So we perform the straight forward test of seeing
3095 * if the pid we have found is the pid of a thread
3096 * group leader, and don't worry if the task we have
3097 * found doesn't happen to be a thread group leader.
3098 * As we don't care in the case of readdir.
3099 */
3100 if (!iter.task || !has_group_leader_pid(iter.task)) {
3101 iter.tgid += 1;
3102 goto retry;
3103 }
3104 get_task_struct(iter.task);
3105 }
3106 rcu_read_unlock();
3107 return iter;
3108 }
3109
3110 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
3111
3112 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3113 struct tgid_iter iter)
3114 {
3115 char name[PROC_NUMBUF];
3116 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
3117 return proc_fill_cache(filp, dirent, filldir, name, len,
3118 proc_pid_instantiate, iter.task, NULL);
3119 }
3120
3121 /* for the /proc/ directory itself, after non-process stuff has been done */
3122 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3123 {
3124 unsigned int nr;
3125 struct task_struct *reaper;
3126 struct tgid_iter iter;
3127 struct pid_namespace *ns;
3128
3129 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3130 goto out_no_task;
3131 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3132
3133 reaper = get_proc_task(filp->f_path.dentry->d_inode);
3134 if (!reaper)
3135 goto out_no_task;
3136
3137 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
3138 const struct pid_entry *p = &proc_base_stuff[nr];
3139 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
3140 goto out;
3141 }
3142
3143 ns = filp->f_dentry->d_sb->s_fs_info;
3144 iter.task = NULL;
3145 iter.tgid = filp->f_pos - TGID_OFFSET;
3146 for (iter = next_tgid(ns, iter);
3147 iter.task;
3148 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3149 filp->f_pos = iter.tgid + TGID_OFFSET;
3150 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
3151 put_task_struct(iter.task);
3152 goto out;
3153 }
3154 }
3155 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3156 out:
3157 put_task_struct(reaper);
3158 out_no_task:
3159 return 0;
3160 }
3161
3162 /*
3163 * Tasks
3164 */
3165 static const struct pid_entry tid_base_stuff[] = {
3166 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3167 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3168 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3169 REG("environ", S_IRUSR, proc_environ_operations),
3170 INF("auxv", S_IRUSR, proc_pid_auxv),
3171 ONE("status", S_IRUGO, proc_pid_status),
3172 ONE("personality", S_IRUGO, proc_pid_personality),
3173 INF("limits", S_IRUGO, proc_pid_limits),
3174 #ifdef CONFIG_SCHED_DEBUG
3175 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3176 #endif
3177 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3178 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3179 INF("syscall", S_IRUGO, proc_pid_syscall),
3180 #endif
3181 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3182 ONE("stat", S_IRUGO, proc_tid_stat),
3183 ONE("statm", S_IRUGO, proc_pid_statm),
3184 REG("maps", S_IRUGO, proc_maps_operations),
3185 #ifdef CONFIG_NUMA
3186 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
3187 #endif
3188 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3189 LNK("cwd", proc_cwd_link),
3190 LNK("root", proc_root_link),
3191 LNK("exe", proc_exe_link),
3192 REG("mounts", S_IRUGO, proc_mounts_operations),
3193 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3194 #ifdef CONFIG_PROC_PAGE_MONITOR
3195 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3196 REG("smaps", S_IRUGO, proc_smaps_operations),
3197 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3198 #endif
3199 #ifdef CONFIG_SECURITY
3200 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3201 #endif
3202 #ifdef CONFIG_KALLSYMS
3203 INF("wchan", S_IRUGO, proc_pid_wchan),
3204 #endif
3205 #ifdef CONFIG_STACKTRACE
3206 ONE("stack", S_IRUGO, proc_pid_stack),
3207 #endif
3208 #ifdef CONFIG_SCHEDSTATS
3209 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3210 #endif
3211 #ifdef CONFIG_LATENCYTOP
3212 REG("latency", S_IRUGO, proc_lstats_operations),
3213 #endif
3214 #ifdef CONFIG_PROC_PID_CPUSET
3215 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3216 #endif
3217 #ifdef CONFIG_CGROUPS
3218 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3219 #endif
3220 INF("oom_score", S_IRUGO, proc_oom_score),
3221 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3222 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3223 #ifdef CONFIG_AUDITSYSCALL
3224 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3225 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3226 #endif
3227 #ifdef CONFIG_FAULT_INJECTION
3228 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3229 #endif
3230 #ifdef CONFIG_TASK_IO_ACCOUNTING
3231 INF("io", S_IRUGO, proc_tid_io_accounting),
3232 #endif
3233 };
3234
3235 static int proc_tid_base_readdir(struct file * filp,
3236 void * dirent, filldir_t filldir)
3237 {
3238 return proc_pident_readdir(filp,dirent,filldir,
3239 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3240 }
3241
3242 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3243 return proc_pident_lookup(dir, dentry,
3244 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3245 }
3246
3247 static const struct file_operations proc_tid_base_operations = {
3248 .read = generic_read_dir,
3249 .readdir = proc_tid_base_readdir,
3250 .llseek = default_llseek,
3251 };
3252
3253 static const struct inode_operations proc_tid_base_inode_operations = {
3254 .lookup = proc_tid_base_lookup,
3255 .getattr = pid_getattr,
3256 .setattr = proc_setattr,
3257 };
3258
3259 static struct dentry *proc_task_instantiate(struct inode *dir,
3260 struct dentry *dentry, struct task_struct *task, const void *ptr)
3261 {
3262 struct dentry *error = ERR_PTR(-ENOENT);
3263 struct inode *inode;
3264 inode = proc_pid_make_inode(dir->i_sb, task);
3265
3266 if (!inode)
3267 goto out;
3268 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3269 inode->i_op = &proc_tid_base_inode_operations;
3270 inode->i_fop = &proc_tid_base_operations;
3271 inode->i_flags|=S_IMMUTABLE;
3272
3273 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
3274 ARRAY_SIZE(tid_base_stuff));
3275
3276 d_set_d_op(dentry, &pid_dentry_operations);
3277
3278 d_add(dentry, inode);
3279 /* Close the race of the process dying before we return the dentry */
3280 if (pid_revalidate(dentry, NULL))
3281 error = NULL;
3282 out:
3283 return error;
3284 }
3285
3286 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3287 {
3288 struct dentry *result = ERR_PTR(-ENOENT);
3289 struct task_struct *task;
3290 struct task_struct *leader = get_proc_task(dir);
3291 unsigned tid;
3292 struct pid_namespace *ns;
3293
3294 if (!leader)
3295 goto out_no_task;
3296
3297 tid = name_to_int(dentry);
3298 if (tid == ~0U)
3299 goto out;
3300
3301 ns = dentry->d_sb->s_fs_info;
3302 rcu_read_lock();
3303 task = find_task_by_pid_ns(tid, ns);
3304 if (task)
3305 get_task_struct(task);
3306 rcu_read_unlock();
3307 if (!task)
3308 goto out;
3309 if (!same_thread_group(leader, task))
3310 goto out_drop_task;
3311
3312 result = proc_task_instantiate(dir, dentry, task, NULL);
3313 out_drop_task:
3314 put_task_struct(task);
3315 out:
3316 put_task_struct(leader);
3317 out_no_task:
3318 return result;
3319 }
3320
3321 /*
3322 * Find the first tid of a thread group to return to user space.
3323 *
3324 * Usually this is just the thread group leader, but if the users
3325 * buffer was too small or there was a seek into the middle of the
3326 * directory we have more work todo.
3327 *
3328 * In the case of a short read we start with find_task_by_pid.
3329 *
3330 * In the case of a seek we start with the leader and walk nr
3331 * threads past it.
3332 */
3333 static struct task_struct *first_tid(struct task_struct *leader,
3334 int tid, int nr, struct pid_namespace *ns)
3335 {
3336 struct task_struct *pos;
3337
3338 rcu_read_lock();
3339 /* Attempt to start with the pid of a thread */
3340 if (tid && (nr > 0)) {
3341 pos = find_task_by_pid_ns(tid, ns);
3342 if (pos && (pos->group_leader == leader))
3343 goto found;
3344 }
3345
3346 /* If nr exceeds the number of threads there is nothing todo */
3347 pos = NULL;
3348 if (nr && nr >= get_nr_threads(leader))
3349 goto out;
3350
3351 /* If we haven't found our starting place yet start
3352 * with the leader and walk nr threads forward.
3353 */
3354 for (pos = leader; nr > 0; --nr) {
3355 pos = next_thread(pos);
3356 if (pos == leader) {
3357 pos = NULL;
3358 goto out;
3359 }
3360 }
3361 found:
3362 get_task_struct(pos);
3363 out:
3364 rcu_read_unlock();
3365 return pos;
3366 }
3367
3368 /*
3369 * Find the next thread in the thread list.
3370 * Return NULL if there is an error or no next thread.
3371 *
3372 * The reference to the input task_struct is released.
3373 */
3374 static struct task_struct *next_tid(struct task_struct *start)
3375 {
3376 struct task_struct *pos = NULL;
3377 rcu_read_lock();
3378 if (pid_alive(start)) {
3379 pos = next_thread(start);
3380 if (thread_group_leader(pos))
3381 pos = NULL;
3382 else
3383 get_task_struct(pos);
3384 }
3385 rcu_read_unlock();
3386 put_task_struct(start);
3387 return pos;
3388 }
3389
3390 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3391 struct task_struct *task, int tid)
3392 {
3393 char name[PROC_NUMBUF];
3394 int len = snprintf(name, sizeof(name), "%d", tid);
3395 return proc_fill_cache(filp, dirent, filldir, name, len,
3396 proc_task_instantiate, task, NULL);
3397 }
3398
3399 /* for the /proc/TGID/task/ directories */
3400 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3401 {
3402 struct dentry *dentry = filp->f_path.dentry;
3403 struct inode *inode = dentry->d_inode;
3404 struct task_struct *leader = NULL;
3405 struct task_struct *task;
3406 int retval = -ENOENT;
3407 ino_t ino;
3408 int tid;
3409 struct pid_namespace *ns;
3410
3411 task = get_proc_task(inode);
3412 if (!task)
3413 goto out_no_task;
3414 rcu_read_lock();
3415 if (pid_alive(task)) {
3416 leader = task->group_leader;
3417 get_task_struct(leader);
3418 }
3419 rcu_read_unlock();
3420 put_task_struct(task);
3421 if (!leader)
3422 goto out_no_task;
3423 retval = 0;
3424
3425 switch ((unsigned long)filp->f_pos) {
3426 case 0:
3427 ino = inode->i_ino;
3428 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3429 goto out;
3430 filp->f_pos++;
3431 /* fall through */
3432 case 1:
3433 ino = parent_ino(dentry);
3434 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3435 goto out;
3436 filp->f_pos++;
3437 /* fall through */
3438 }
3439
3440 /* f_version caches the tgid value that the last readdir call couldn't
3441 * return. lseek aka telldir automagically resets f_version to 0.
3442 */
3443 ns = filp->f_dentry->d_sb->s_fs_info;
3444 tid = (int)filp->f_version;
3445 filp->f_version = 0;
3446 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3447 task;
3448 task = next_tid(task), filp->f_pos++) {
3449 tid = task_pid_nr_ns(task, ns);
3450 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3451 /* returning this tgid failed, save it as the first
3452 * pid for the next readir call */
3453 filp->f_version = (u64)tid;
3454 put_task_struct(task);
3455 break;
3456 }
3457 }
3458 out:
3459 put_task_struct(leader);
3460 out_no_task:
3461 return retval;
3462 }
3463
3464 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3465 {
3466 struct inode *inode = dentry->d_inode;
3467 struct task_struct *p = get_proc_task(inode);
3468 generic_fillattr(inode, stat);
3469
3470 if (p) {
3471 stat->nlink += get_nr_threads(p);
3472 put_task_struct(p);
3473 }
3474
3475 return 0;
3476 }
3477
3478 static const struct inode_operations proc_task_inode_operations = {
3479 .lookup = proc_task_lookup,
3480 .getattr = proc_task_getattr,
3481 .setattr = proc_setattr,
3482 };
3483
3484 static const struct file_operations proc_task_operations = {
3485 .read = generic_read_dir,
3486 .readdir = proc_task_readdir,
3487 .llseek = default_llseek,
3488 };
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree