?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
oom: allow a non-CAP_SYS_RESOURCE proces to oom_score_adj down
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / proc / base.c
blob9d096e82b201090b5a1fdac646f811cc6fa8bc57
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 /*
195 * Return zero if current may access user memory in @task, -error if not.
196 */
197 static int check_mem_permission(struct task_struct *task)
198 {
199 /*
200 * A task can always look at itself, in case it chooses
201 * to use system calls instead of load instructions.
202 */
203 if (task == current)
204 return 0;
205
206 /*
207 * If current is actively ptrace'ing, and would also be
208 * permitted to freshly attach with ptrace now, permit it.
209 */
210 if (task_is_stopped_or_traced(task)) {
211 int match;
212 rcu_read_lock();
213 match = (tracehook_tracer_task(task) == current);
214 rcu_read_unlock();
215 if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH))
216 return 0;
217 }
218
219 /*
220 * Noone else is allowed.
221 */
222 return -EPERM;
223 }
224
225 struct mm_struct *mm_for_maps(struct task_struct *task)
226 {
227 struct mm_struct *mm;
228
229 if (mutex_lock_killable(&task->signal->cred_guard_mutex))
230 return NULL;
231
232 mm = get_task_mm(task);
233 if (mm && mm != current->mm &&
234 !ptrace_may_access(task, PTRACE_MODE_READ)) {
235 mmput(mm);
236 mm = NULL;
237 }
238 mutex_unlock(&task->signal->cred_guard_mutex);
239
240 return mm;
241 }
242
243 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
244 {
245 int res = 0;
246 unsigned int len;
247 struct mm_struct *mm = get_task_mm(task);
248 if (!mm)
249 goto out;
250 if (!mm->arg_end)
251 goto out_mm; /* Shh! No looking before we're done */
252
253 len = mm->arg_end - mm->arg_start;
254
255 if (len > PAGE_SIZE)
256 len = PAGE_SIZE;
257
258 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
259
260 // If the nul at the end of args has been overwritten, then
261 // assume application is using setproctitle(3).
262 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
263 len = strnlen(buffer, res);
264 if (len < res) {
265 res = len;
266 } else {
267 len = mm->env_end - mm->env_start;
268 if (len > PAGE_SIZE - res)
269 len = PAGE_SIZE - res;
270 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
271 res = strnlen(buffer, res);
272 }
273 }
274 out_mm:
275 mmput(mm);
276 out:
277 return res;
278 }
279
280 static int proc_pid_auxv(struct task_struct *task, char *buffer)
281 {
282 int res = 0;
283 struct mm_struct *mm = get_task_mm(task);
284 if (mm) {
285 unsigned int nwords = 0;
286 do {
287 nwords += 2;
288 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
289 res = nwords * sizeof(mm->saved_auxv[0]);
290 if (res > PAGE_SIZE)
291 res = PAGE_SIZE;
292 memcpy(buffer, mm->saved_auxv, res);
293 mmput(mm);
294 }
295 return res;
296 }
297
298
299 #ifdef CONFIG_KALLSYMS
300 /*
301 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
302 * Returns the resolved symbol. If that fails, simply return the address.
303 */
304 static int proc_pid_wchan(struct task_struct *task, char *buffer)
305 {
306 unsigned long wchan;
307 char symname[KSYM_NAME_LEN];
308
309 wchan = get_wchan(task);
310
311 if (lookup_symbol_name(wchan, symname) < 0)
312 if (!ptrace_may_access(task, PTRACE_MODE_READ))
313 return 0;
314 else
315 return sprintf(buffer, "%lu", wchan);
316 else
317 return sprintf(buffer, "%s", symname);
318 }
319 #endif /* CONFIG_KALLSYMS */
320
321 #ifdef CONFIG_STACKTRACE
322
323 #define MAX_STACK_TRACE_DEPTH 64
324
325 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
326 struct pid *pid, struct task_struct *task)
327 {
328 struct stack_trace trace;
329 unsigned long *entries;
330 int i;
331
332 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
333 if (!entries)
334 return -ENOMEM;
335
336 trace.nr_entries = 0;
337 trace.max_entries = MAX_STACK_TRACE_DEPTH;
338 trace.entries = entries;
339 trace.skip = 0;
340 save_stack_trace_tsk(task, &trace);
341
342 for (i = 0; i < trace.nr_entries; i++) {
343 seq_printf(m, "[<%p>] %pS\n",
344 (void *)entries[i], (void *)entries[i]);
345 }
346 kfree(entries);
347
348 return 0;
349 }
350 #endif
351
352 #ifdef CONFIG_SCHEDSTATS
353 /*
354 * Provides /proc/PID/schedstat
355 */
356 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
357 {
358 return sprintf(buffer, "%llu %llu %lu\n",
359 (unsigned long long)task->se.sum_exec_runtime,
360 (unsigned long long)task->sched_info.run_delay,
361 task->sched_info.pcount);
362 }
363 #endif
364
365 #ifdef CONFIG_LATENCYTOP
366 static int lstats_show_proc(struct seq_file *m, void *v)
367 {
368 int i;
369 struct inode *inode = m->private;
370 struct task_struct *task = get_proc_task(inode);
371
372 if (!task)
373 return -ESRCH;
374 seq_puts(m, "Latency Top version : v0.1\n");
375 for (i = 0; i < 32; i++) {
376 struct latency_record *lr = &task->latency_record[i];
377 if (lr->backtrace[0]) {
378 int q;
379 seq_printf(m, "%i %li %li",
380 lr->count, lr->time, lr->max);
381 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
382 unsigned long bt = lr->backtrace[q];
383 if (!bt)
384 break;
385 if (bt == ULONG_MAX)
386 break;
387 seq_printf(m, " %ps", (void *)bt);
388 }
389 seq_putc(m, '\n');
390 }
391
392 }
393 put_task_struct(task);
394 return 0;
395 }
396
397 static int lstats_open(struct inode *inode, struct file *file)
398 {
399 return single_open(file, lstats_show_proc, inode);
400 }
401
402 static ssize_t lstats_write(struct file *file, const char __user *buf,
403 size_t count, loff_t *offs)
404 {
405 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
406
407 if (!task)
408 return -ESRCH;
409 clear_all_latency_tracing(task);
410 put_task_struct(task);
411
412 return count;
413 }
414
415 static const struct file_operations proc_lstats_operations = {
416 .open = lstats_open,
417 .read = seq_read,
418 .write = lstats_write,
419 .llseek = seq_lseek,
420 .release = single_release,
421 };
422
423 #endif
424
425 static int proc_oom_score(struct task_struct *task, char *buffer)
426 {
427 unsigned long points = 0;
428
429 read_lock(&tasklist_lock);
430 if (pid_alive(task))
431 points = oom_badness(task, NULL, NULL,
432 totalram_pages + total_swap_pages);
433 read_unlock(&tasklist_lock);
434 return sprintf(buffer, "%lu\n", points);
435 }
436
437 struct limit_names {
438 char *name;
439 char *unit;
440 };
441
442 static const struct limit_names lnames[RLIM_NLIMITS] = {
443 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
444 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
445 [RLIMIT_DATA] = {"Max data size", "bytes"},
446 [RLIMIT_STACK] = {"Max stack size", "bytes"},
447 [RLIMIT_CORE] = {"Max core file size", "bytes"},
448 [RLIMIT_RSS] = {"Max resident set", "bytes"},
449 [RLIMIT_NPROC] = {"Max processes", "processes"},
450 [RLIMIT_NOFILE] = {"Max open files", "files"},
451 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
452 [RLIMIT_AS] = {"Max address space", "bytes"},
453 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
454 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
455 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
456 [RLIMIT_NICE] = {"Max nice priority", NULL},
457 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
458 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
459 };
460
461 /* Display limits for a process */
462 static int proc_pid_limits(struct task_struct *task, char *buffer)
463 {
464 unsigned int i;
465 int count = 0;
466 unsigned long flags;
467 char *bufptr = buffer;
468
469 struct rlimit rlim[RLIM_NLIMITS];
470
471 if (!lock_task_sighand(task, &flags))
472 return 0;
473 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
474 unlock_task_sighand(task, &flags);
475
476 /*
477 * print the file header
478 */
479 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
480 "Limit", "Soft Limit", "Hard Limit", "Units");
481
482 for (i = 0; i < RLIM_NLIMITS; i++) {
483 if (rlim[i].rlim_cur == RLIM_INFINITY)
484 count += sprintf(&bufptr[count], "%-25s %-20s ",
485 lnames[i].name, "unlimited");
486 else
487 count += sprintf(&bufptr[count], "%-25s %-20lu ",
488 lnames[i].name, rlim[i].rlim_cur);
489
490 if (rlim[i].rlim_max == RLIM_INFINITY)
491 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
492 else
493 count += sprintf(&bufptr[count], "%-20lu ",
494 rlim[i].rlim_max);
495
496 if (lnames[i].unit)
497 count += sprintf(&bufptr[count], "%-10s\n",
498 lnames[i].unit);
499 else
500 count += sprintf(&bufptr[count], "\n");
501 }
502
503 return count;
504 }
505
506 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
507 static int proc_pid_syscall(struct task_struct *task, char *buffer)
508 {
509 long nr;
510 unsigned long args[6], sp, pc;
511
512 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
513 return sprintf(buffer, "running\n");
514
515 if (nr < 0)
516 return sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
517
518 return sprintf(buffer,
519 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
520 nr,
521 args[0], args[1], args[2], args[3], args[4], args[5],
522 sp, pc);
523 }
524 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
525
526 /************************************************************************/
527 /* Here the fs part begins */
528 /************************************************************************/
529
530 /* permission checks */
531 static int proc_fd_access_allowed(struct inode *inode)
532 {
533 struct task_struct *task;
534 int allowed = 0;
535 /* Allow access to a task's file descriptors if it is us or we
536 * may use ptrace attach to the process and find out that
537 * information.
538 */
539 task = get_proc_task(inode);
540 if (task) {
541 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
542 put_task_struct(task);
543 }
544 return allowed;
545 }
546
547 static int proc_setattr(struct dentry *dentry, struct iattr *attr)
548 {
549 int error;
550 struct inode *inode = dentry->d_inode;
551
552 if (attr->ia_valid & ATTR_MODE)
553 return -EPERM;
554
555 error = inode_change_ok(inode, attr);
556 if (error)
557 return error;
558
559 if ((attr->ia_valid & ATTR_SIZE) &&
560 attr->ia_size != i_size_read(inode)) {
561 error = vmtruncate(inode, attr->ia_size);
562 if (error)
563 return error;
564 }
565
566 setattr_copy(inode, attr);
567 mark_inode_dirty(inode);
568 return 0;
569 }
570
571 static const struct inode_operations proc_def_inode_operations = {
572 .setattr = proc_setattr,
573 };
574
575 static int mounts_open_common(struct inode *inode, struct file *file,
576 const struct seq_operations *op)
577 {
578 struct task_struct *task = get_proc_task(inode);
579 struct nsproxy *nsp;
580 struct mnt_namespace *ns = NULL;
581 struct path root;
582 struct proc_mounts *p;
583 int ret = -EINVAL;
584
585 if (task) {
586 rcu_read_lock();
587 nsp = task_nsproxy(task);
588 if (nsp) {
589 ns = nsp->mnt_ns;
590 if (ns)
591 get_mnt_ns(ns);
592 }
593 rcu_read_unlock();
594 if (ns && get_task_root(task, &root) == 0)
595 ret = 0;
596 put_task_struct(task);
597 }
598
599 if (!ns)
600 goto err;
601 if (ret)
602 goto err_put_ns;
603
604 ret = -ENOMEM;
605 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
606 if (!p)
607 goto err_put_path;
608
609 file->private_data = &p->m;
610 ret = seq_open(file, op);
611 if (ret)
612 goto err_free;
613
614 p->m.private = p;
615 p->ns = ns;
616 p->root = root;
617 p->event = ns->event;
618
619 return 0;
620
621 err_free:
622 kfree(p);
623 err_put_path:
624 path_put(&root);
625 err_put_ns:
626 put_mnt_ns(ns);
627 err:
628 return ret;
629 }
630
631 static int mounts_release(struct inode *inode, struct file *file)
632 {
633 struct proc_mounts *p = file->private_data;
634 path_put(&p->root);
635 put_mnt_ns(p->ns);
636 return seq_release(inode, file);
637 }
638
639 static unsigned mounts_poll(struct file *file, poll_table *wait)
640 {
641 struct proc_mounts *p = file->private_data;
642 unsigned res = POLLIN | POLLRDNORM;
643
644 poll_wait(file, &p->ns->poll, wait);
645 if (mnt_had_events(p))
646 res |= POLLERR | POLLPRI;
647
648 return res;
649 }
650
651 static int mounts_open(struct inode *inode, struct file *file)
652 {
653 return mounts_open_common(inode, file, &mounts_op);
654 }
655
656 static const struct file_operations proc_mounts_operations = {
657 .open = mounts_open,
658 .read = seq_read,
659 .llseek = seq_lseek,
660 .release = mounts_release,
661 .poll = mounts_poll,
662 };
663
664 static int mountinfo_open(struct inode *inode, struct file *file)
665 {
666 return mounts_open_common(inode, file, &mountinfo_op);
667 }
668
669 static const struct file_operations proc_mountinfo_operations = {
670 .open = mountinfo_open,
671 .read = seq_read,
672 .llseek = seq_lseek,
673 .release = mounts_release,
674 .poll = mounts_poll,
675 };
676
677 static int mountstats_open(struct inode *inode, struct file *file)
678 {
679 return mounts_open_common(inode, file, &mountstats_op);
680 }
681
682 static const struct file_operations proc_mountstats_operations = {
683 .open = mountstats_open,
684 .read = seq_read,
685 .llseek = seq_lseek,
686 .release = mounts_release,
687 };
688
689 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
690
691 static ssize_t proc_info_read(struct file * file, char __user * buf,
692 size_t count, loff_t *ppos)
693 {
694 struct inode * inode = file->f_path.dentry->d_inode;
695 unsigned long page;
696 ssize_t length;
697 struct task_struct *task = get_proc_task(inode);
698
699 length = -ESRCH;
700 if (!task)
701 goto out_no_task;
702
703 if (count > PROC_BLOCK_SIZE)
704 count = PROC_BLOCK_SIZE;
705
706 length = -ENOMEM;
707 if (!(page = __get_free_page(GFP_TEMPORARY)))
708 goto out;
709
710 length = PROC_I(inode)->op.proc_read(task, (char*)page);
711
712 if (length >= 0)
713 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
714 free_page(page);
715 out:
716 put_task_struct(task);
717 out_no_task:
718 return length;
719 }
720
721 static const struct file_operations proc_info_file_operations = {
722 .read = proc_info_read,
723 .llseek = generic_file_llseek,
724 };
725
726 static int proc_single_show(struct seq_file *m, void *v)
727 {
728 struct inode *inode = m->private;
729 struct pid_namespace *ns;
730 struct pid *pid;
731 struct task_struct *task;
732 int ret;
733
734 ns = inode->i_sb->s_fs_info;
735 pid = proc_pid(inode);
736 task = get_pid_task(pid, PIDTYPE_PID);
737 if (!task)
738 return -ESRCH;
739
740 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
741
742 put_task_struct(task);
743 return ret;
744 }
745
746 static int proc_single_open(struct inode *inode, struct file *filp)
747 {
748 return single_open(filp, proc_single_show, inode);
749 }
750
751 static const struct file_operations proc_single_file_operations = {
752 .open = proc_single_open,
753 .read = seq_read,
754 .llseek = seq_lseek,
755 .release = single_release,
756 };
757
758 static int mem_open(struct inode* inode, struct file* file)
759 {
760 file->private_data = (void*)((long)current->self_exec_id);
761 /* OK to pass negative loff_t, we can catch out-of-range */
762 file->f_mode |= FMODE_UNSIGNED_OFFSET;
763 return 0;
764 }
765
766 static ssize_t mem_read(struct file * file, char __user * buf,
767 size_t count, loff_t *ppos)
768 {
769 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
770 char *page;
771 unsigned long src = *ppos;
772 int ret = -ESRCH;
773 struct mm_struct *mm;
774
775 if (!task)
776 goto out_no_task;
777
778 if (check_mem_permission(task))
779 goto out;
780
781 ret = -ENOMEM;
782 page = (char *)__get_free_page(GFP_TEMPORARY);
783 if (!page)
784 goto out;
785
786 ret = 0;
787
788 mm = get_task_mm(task);
789 if (!mm)
790 goto out_free;
791
792 ret = -EIO;
793
794 if (file->private_data != (void*)((long)current->self_exec_id))
795 goto out_put;
796
797 ret = 0;
798
799 while (count > 0) {
800 int this_len, retval;
801
802 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
803 retval = access_process_vm(task, src, page, this_len, 0);
804 if (!retval || check_mem_permission(task)) {
805 if (!ret)
806 ret = -EIO;
807 break;
808 }
809
810 if (copy_to_user(buf, page, retval)) {
811 ret = -EFAULT;
812 break;
813 }
814
815 ret += retval;
816 src += retval;
817 buf += retval;
818 count -= retval;
819 }
820 *ppos = src;
821
822 out_put:
823 mmput(mm);
824 out_free:
825 free_page((unsigned long) page);
826 out:
827 put_task_struct(task);
828 out_no_task:
829 return ret;
830 }
831
832 #define mem_write NULL
833
834 #ifndef mem_write
835 /* This is a security hazard */
836 static ssize_t mem_write(struct file * file, const char __user *buf,
837 size_t count, loff_t *ppos)
838 {
839 int copied;
840 char *page;
841 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
842 unsigned long dst = *ppos;
843
844 copied = -ESRCH;
845 if (!task)
846 goto out_no_task;
847
848 if (check_mem_permission(task))
849 goto out;
850
851 copied = -ENOMEM;
852 page = (char *)__get_free_page(GFP_TEMPORARY);
853 if (!page)
854 goto out;
855
856 copied = 0;
857 while (count > 0) {
858 int this_len, retval;
859
860 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
861 if (copy_from_user(page, buf, this_len)) {
862 copied = -EFAULT;
863 break;
864 }
865 retval = access_process_vm(task, dst, page, this_len, 1);
866 if (!retval) {
867 if (!copied)
868 copied = -EIO;
869 break;
870 }
871 copied += retval;
872 buf += retval;
873 dst += retval;
874 count -= retval;
875 }
876 *ppos = dst;
877 free_page((unsigned long) page);
878 out:
879 put_task_struct(task);
880 out_no_task:
881 return copied;
882 }
883 #endif
884
885 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
886 {
887 switch (orig) {
888 case 0:
889 file->f_pos = offset;
890 break;
891 case 1:
892 file->f_pos += offset;
893 break;
894 default:
895 return -EINVAL;
896 }
897 force_successful_syscall_return();
898 return file->f_pos;
899 }
900
901 static const struct file_operations proc_mem_operations = {
902 .llseek = mem_lseek,
903 .read = mem_read,
904 .write = mem_write,
905 .open = mem_open,
906 };
907
908 static ssize_t environ_read(struct file *file, char __user *buf,
909 size_t count, loff_t *ppos)
910 {
911 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
912 char *page;
913 unsigned long src = *ppos;
914 int ret = -ESRCH;
915 struct mm_struct *mm;
916
917 if (!task)
918 goto out_no_task;
919
920 if (!ptrace_may_access(task, PTRACE_MODE_READ))
921 goto out;
922
923 ret = -ENOMEM;
924 page = (char *)__get_free_page(GFP_TEMPORARY);
925 if (!page)
926 goto out;
927
928 ret = 0;
929
930 mm = get_task_mm(task);
931 if (!mm)
932 goto out_free;
933
934 while (count > 0) {
935 int this_len, retval, max_len;
936
937 this_len = mm->env_end - (mm->env_start + src);
938
939 if (this_len <= 0)
940 break;
941
942 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
943 this_len = (this_len > max_len) ? max_len : this_len;
944
945 retval = access_process_vm(task, (mm->env_start + src),
946 page, this_len, 0);
947
948 if (retval <= 0) {
949 ret = retval;
950 break;
951 }
952
953 if (copy_to_user(buf, page, retval)) {
954 ret = -EFAULT;
955 break;
956 }
957
958 ret += retval;
959 src += retval;
960 buf += retval;
961 count -= retval;
962 }
963 *ppos = src;
964
965 mmput(mm);
966 out_free:
967 free_page((unsigned long) page);
968 out:
969 put_task_struct(task);
970 out_no_task:
971 return ret;
972 }
973
974 static const struct file_operations proc_environ_operations = {
975 .read = environ_read,
976 .llseek = generic_file_llseek,
977 };
978
979 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
980 size_t count, loff_t *ppos)
981 {
982 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
983 char buffer[PROC_NUMBUF];
984 size_t len;
985 int oom_adjust = OOM_DISABLE;
986 unsigned long flags;
987
988 if (!task)
989 return -ESRCH;
990
991 if (lock_task_sighand(task, &flags)) {
992 oom_adjust = task->signal->oom_adj;
993 unlock_task_sighand(task, &flags);
994 }
995
996 put_task_struct(task);
997
998 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
999
1000 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1001 }
1002
1003 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1004 size_t count, loff_t *ppos)
1005 {
1006 struct task_struct *task;
1007 char buffer[PROC_NUMBUF];
1008 long oom_adjust;
1009 unsigned long flags;
1010 int err;
1011
1012 memset(buffer, 0, sizeof(buffer));
1013 if (count > sizeof(buffer) - 1)
1014 count = sizeof(buffer) - 1;
1015 if (copy_from_user(buffer, buf, count)) {
1016 err = -EFAULT;
1017 goto out;
1018 }
1019
1020 err = strict_strtol(strstrip(buffer), 0, &oom_adjust);
1021 if (err)
1022 goto out;
1023 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
1024 oom_adjust != OOM_DISABLE) {
1025 err = -EINVAL;
1026 goto out;
1027 }
1028
1029 task = get_proc_task(file->f_path.dentry->d_inode);
1030 if (!task) {
1031 err = -ESRCH;
1032 goto out;
1033 }
1034
1035 task_lock(task);
1036 if (!task->mm) {
1037 err = -EINVAL;
1038 goto err_task_lock;
1039 }
1040
1041 if (!lock_task_sighand(task, &flags)) {
1042 err = -ESRCH;
1043 goto err_task_lock;
1044 }
1045
1046 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
1047 err = -EACCES;
1048 goto err_sighand;
1049 }
1050
1051 if (oom_adjust != task->signal->oom_adj) {
1052 if (oom_adjust == OOM_DISABLE)
1053 atomic_inc(&task->mm->oom_disable_count);
1054 if (task->signal->oom_adj == OOM_DISABLE)
1055 atomic_dec(&task->mm->oom_disable_count);
1056 }
1057
1058 /*
1059 * Warn that /proc/pid/oom_adj is deprecated, see
1060 * Documentation/feature-removal-schedule.txt.
1061 */
1062 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, "
1063 "please use /proc/%d/oom_score_adj instead.\n",
1064 current->comm, task_pid_nr(current),
1065 task_pid_nr(task), task_pid_nr(task));
1066 task->signal->oom_adj = oom_adjust;
1067 /*
1068 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1069 * value is always attainable.
1070 */
1071 if (task->signal->oom_adj == OOM_ADJUST_MAX)
1072 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
1073 else
1074 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
1075 -OOM_DISABLE;
1076 err_sighand:
1077 unlock_task_sighand(task, &flags);
1078 err_task_lock:
1079 task_unlock(task);
1080 put_task_struct(task);
1081 out:
1082 return err < 0 ? err : count;
1083 }
1084
1085 static const struct file_operations proc_oom_adjust_operations = {
1086 .read = oom_adjust_read,
1087 .write = oom_adjust_write,
1088 .llseek = generic_file_llseek,
1089 };
1090
1091 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1092 size_t count, loff_t *ppos)
1093 {
1094 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1095 char buffer[PROC_NUMBUF];
1096 int oom_score_adj = OOM_SCORE_ADJ_MIN;
1097 unsigned long flags;
1098 size_t len;
1099
1100 if (!task)
1101 return -ESRCH;
1102 if (lock_task_sighand(task, &flags)) {
1103 oom_score_adj = task->signal->oom_score_adj;
1104 unlock_task_sighand(task, &flags);
1105 }
1106 put_task_struct(task);
1107 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1108 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1109 }
1110
1111 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1112 size_t count, loff_t *ppos)
1113 {
1114 struct task_struct *task;
1115 char buffer[PROC_NUMBUF];
1116 unsigned long flags;
1117 long oom_score_adj;
1118 int err;
1119
1120 memset(buffer, 0, sizeof(buffer));
1121 if (count > sizeof(buffer) - 1)
1122 count = sizeof(buffer) - 1;
1123 if (copy_from_user(buffer, buf, count)) {
1124 err = -EFAULT;
1125 goto out;
1126 }
1127
1128 err = strict_strtol(strstrip(buffer), 0, &oom_score_adj);
1129 if (err)
1130 goto out;
1131 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1132 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1133 err = -EINVAL;
1134 goto out;
1135 }
1136
1137 task = get_proc_task(file->f_path.dentry->d_inode);
1138 if (!task) {
1139 err = -ESRCH;
1140 goto out;
1141 }
1142
1143 task_lock(task);
1144 if (!task->mm) {
1145 err = -EINVAL;
1146 goto err_task_lock;
1147 }
1148
1149 if (!lock_task_sighand(task, &flags)) {
1150 err = -ESRCH;
1151 goto err_task_lock;
1152 }
1153
1154 if (oom_score_adj < task->signal->oom_score_adj_min &&
1155 !capable(CAP_SYS_RESOURCE)) {
1156 err = -EACCES;
1157 goto err_sighand;
1158 }
1159
1160 if (oom_score_adj != task->signal->oom_score_adj) {
1161 if (oom_score_adj == OOM_SCORE_ADJ_MIN)
1162 atomic_inc(&task->mm->oom_disable_count);
1163 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1164 atomic_dec(&task->mm->oom_disable_count);
1165 }
1166 task->signal->oom_score_adj = oom_score_adj;
1167 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1168 task->signal->oom_score_adj_min = oom_score_adj;
1169 /*
1170 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1171 * always attainable.
1172 */
1173 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1174 task->signal->oom_adj = OOM_DISABLE;
1175 else
1176 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1177 OOM_SCORE_ADJ_MAX;
1178 err_sighand:
1179 unlock_task_sighand(task, &flags);
1180 err_task_lock:
1181 task_unlock(task);
1182 put_task_struct(task);
1183 out:
1184 return err < 0 ? err : count;
1185 }
1186
1187 static const struct file_operations proc_oom_score_adj_operations = {
1188 .read = oom_score_adj_read,
1189 .write = oom_score_adj_write,
1190 .llseek = default_llseek,
1191 };
1192
1193 #ifdef CONFIG_AUDITSYSCALL
1194 #define TMPBUFLEN 21
1195 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1196 size_t count, loff_t *ppos)
1197 {
1198 struct inode * inode = file->f_path.dentry->d_inode;
1199 struct task_struct *task = get_proc_task(inode);
1200 ssize_t length;
1201 char tmpbuf[TMPBUFLEN];
1202
1203 if (!task)
1204 return -ESRCH;
1205 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1206 audit_get_loginuid(task));
1207 put_task_struct(task);
1208 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1209 }
1210
1211 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1212 size_t count, loff_t *ppos)
1213 {
1214 struct inode * inode = file->f_path.dentry->d_inode;
1215 char *page, *tmp;
1216 ssize_t length;
1217 uid_t loginuid;
1218
1219 if (!capable(CAP_AUDIT_CONTROL))
1220 return -EPERM;
1221
1222 rcu_read_lock();
1223 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1224 rcu_read_unlock();
1225 return -EPERM;
1226 }
1227 rcu_read_unlock();
1228
1229 if (count >= PAGE_SIZE)
1230 count = PAGE_SIZE - 1;
1231
1232 if (*ppos != 0) {
1233 /* No partial writes. */
1234 return -EINVAL;
1235 }
1236 page = (char*)__get_free_page(GFP_TEMPORARY);
1237 if (!page)
1238 return -ENOMEM;
1239 length = -EFAULT;
1240 if (copy_from_user(page, buf, count))
1241 goto out_free_page;
1242
1243 page[count] = '\0';
1244 loginuid = simple_strtoul(page, &tmp, 10);
1245 if (tmp == page) {
1246 length = -EINVAL;
1247 goto out_free_page;
1248
1249 }
1250 length = audit_set_loginuid(current, loginuid);
1251 if (likely(length == 0))
1252 length = count;
1253
1254 out_free_page:
1255 free_page((unsigned long) page);
1256 return length;
1257 }
1258
1259 static const struct file_operations proc_loginuid_operations = {
1260 .read = proc_loginuid_read,
1261 .write = proc_loginuid_write,
1262 .llseek = generic_file_llseek,
1263 };
1264
1265 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1266 size_t count, loff_t *ppos)
1267 {
1268 struct inode * inode = file->f_path.dentry->d_inode;
1269 struct task_struct *task = get_proc_task(inode);
1270 ssize_t length;
1271 char tmpbuf[TMPBUFLEN];
1272
1273 if (!task)
1274 return -ESRCH;
1275 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1276 audit_get_sessionid(task));
1277 put_task_struct(task);
1278 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1279 }
1280
1281 static const struct file_operations proc_sessionid_operations = {
1282 .read = proc_sessionid_read,
1283 .llseek = generic_file_llseek,
1284 };
1285 #endif
1286
1287 #ifdef CONFIG_FAULT_INJECTION
1288 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1289 size_t count, loff_t *ppos)
1290 {
1291 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1292 char buffer[PROC_NUMBUF];
1293 size_t len;
1294 int make_it_fail;
1295
1296 if (!task)
1297 return -ESRCH;
1298 make_it_fail = task->make_it_fail;
1299 put_task_struct(task);
1300
1301 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1302
1303 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1304 }
1305
1306 static ssize_t proc_fault_inject_write(struct file * file,
1307 const char __user * buf, size_t count, loff_t *ppos)
1308 {
1309 struct task_struct *task;
1310 char buffer[PROC_NUMBUF], *end;
1311 int make_it_fail;
1312
1313 if (!capable(CAP_SYS_RESOURCE))
1314 return -EPERM;
1315 memset(buffer, 0, sizeof(buffer));
1316 if (count > sizeof(buffer) - 1)
1317 count = sizeof(buffer) - 1;
1318 if (copy_from_user(buffer, buf, count))
1319 return -EFAULT;
1320 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1321 if (*end)
1322 return -EINVAL;
1323 task = get_proc_task(file->f_dentry->d_inode);
1324 if (!task)
1325 return -ESRCH;
1326 task->make_it_fail = make_it_fail;
1327 put_task_struct(task);
1328
1329 return count;
1330 }
1331
1332 static const struct file_operations proc_fault_inject_operations = {
1333 .read = proc_fault_inject_read,
1334 .write = proc_fault_inject_write,
1335 .llseek = generic_file_llseek,
1336 };
1337 #endif
1338
1339
1340 #ifdef CONFIG_SCHED_DEBUG
1341 /*
1342 * Print out various scheduling related per-task fields:
1343 */
1344 static int sched_show(struct seq_file *m, void *v)
1345 {
1346 struct inode *inode = m->private;
1347 struct task_struct *p;
1348
1349 p = get_proc_task(inode);
1350 if (!p)
1351 return -ESRCH;
1352 proc_sched_show_task(p, m);
1353
1354 put_task_struct(p);
1355
1356 return 0;
1357 }
1358
1359 static ssize_t
1360 sched_write(struct file *file, const char __user *buf,
1361 size_t count, loff_t *offset)
1362 {
1363 struct inode *inode = file->f_path.dentry->d_inode;
1364 struct task_struct *p;
1365
1366 p = get_proc_task(inode);
1367 if (!p)
1368 return -ESRCH;
1369 proc_sched_set_task(p);
1370
1371 put_task_struct(p);
1372
1373 return count;
1374 }
1375
1376 static int sched_open(struct inode *inode, struct file *filp)
1377 {
1378 return single_open(filp, sched_show, inode);
1379 }
1380
1381 static const struct file_operations proc_pid_sched_operations = {
1382 .open = sched_open,
1383 .read = seq_read,
1384 .write = sched_write,
1385 .llseek = seq_lseek,
1386 .release = single_release,
1387 };
1388
1389 #endif
1390
1391 #ifdef CONFIG_SCHED_AUTOGROUP
1392 /*
1393 * Print out autogroup related information:
1394 */
1395 static int sched_autogroup_show(struct seq_file *m, void *v)
1396 {
1397 struct inode *inode = m->private;
1398 struct task_struct *p;
1399
1400 p = get_proc_task(inode);
1401 if (!p)
1402 return -ESRCH;
1403 proc_sched_autogroup_show_task(p, m);
1404
1405 put_task_struct(p);
1406
1407 return 0;
1408 }
1409
1410 static ssize_t
1411 sched_autogroup_write(struct file *file, const char __user *buf,
1412 size_t count, loff_t *offset)
1413 {
1414 struct inode *inode = file->f_path.dentry->d_inode;
1415 struct task_struct *p;
1416 char buffer[PROC_NUMBUF];
1417 long nice;
1418 int err;
1419
1420 memset(buffer, 0, sizeof(buffer));
1421 if (count > sizeof(buffer) - 1)
1422 count = sizeof(buffer) - 1;
1423 if (copy_from_user(buffer, buf, count))
1424 return -EFAULT;
1425
1426 err = strict_strtol(strstrip(buffer), 0, &nice);
1427 if (err)
1428 return -EINVAL;
1429
1430 p = get_proc_task(inode);
1431 if (!p)
1432 return -ESRCH;
1433
1434 err = nice;
1435 err = proc_sched_autogroup_set_nice(p, &err);
1436 if (err)
1437 count = err;
1438
1439 put_task_struct(p);
1440
1441 return count;
1442 }
1443
1444 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1445 {
1446 int ret;
1447
1448 ret = single_open(filp, sched_autogroup_show, NULL);
1449 if (!ret) {
1450 struct seq_file *m = filp->private_data;
1451
1452 m->private = inode;
1453 }
1454 return ret;
1455 }
1456
1457 static const struct file_operations proc_pid_sched_autogroup_operations = {
1458 .open = sched_autogroup_open,
1459 .read = seq_read,
1460 .write = sched_autogroup_write,
1461 .llseek = seq_lseek,
1462 .release = single_release,
1463 };
1464
1465 #endif /* CONFIG_SCHED_AUTOGROUP */
1466
1467 static ssize_t comm_write(struct file *file, const char __user *buf,
1468 size_t count, loff_t *offset)
1469 {
1470 struct inode *inode = file->f_path.dentry->d_inode;
1471 struct task_struct *p;
1472 char buffer[TASK_COMM_LEN];
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 p = get_proc_task(inode);
1481 if (!p)
1482 return -ESRCH;
1483
1484 if (same_thread_group(current, p))
1485 set_task_comm(p, buffer);
1486 else
1487 count = -EINVAL;
1488
1489 put_task_struct(p);
1490
1491 return count;
1492 }
1493
1494 static int comm_show(struct seq_file *m, void *v)
1495 {
1496 struct inode *inode = m->private;
1497 struct task_struct *p;
1498
1499 p = get_proc_task(inode);
1500 if (!p)
1501 return -ESRCH;
1502
1503 task_lock(p);
1504 seq_printf(m, "%s\n", p->comm);
1505 task_unlock(p);
1506
1507 put_task_struct(p);
1508
1509 return 0;
1510 }
1511
1512 static int comm_open(struct inode *inode, struct file *filp)
1513 {
1514 return single_open(filp, comm_show, inode);
1515 }
1516
1517 static const struct file_operations proc_pid_set_comm_operations = {
1518 .open = comm_open,
1519 .read = seq_read,
1520 .write = comm_write,
1521 .llseek = seq_lseek,
1522 .release = single_release,
1523 };
1524
1525 /*
1526 * We added or removed a vma mapping the executable. The vmas are only mapped
1527 * during exec and are not mapped with the mmap system call.
1528 * Callers must hold down_write() on the mm's mmap_sem for these
1529 */
1530 void added_exe_file_vma(struct mm_struct *mm)
1531 {
1532 mm->num_exe_file_vmas++;
1533 }
1534
1535 void removed_exe_file_vma(struct mm_struct *mm)
1536 {
1537 mm->num_exe_file_vmas--;
1538 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
1539 fput(mm->exe_file);
1540 mm->exe_file = NULL;
1541 }
1542
1543 }
1544
1545 void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1546 {
1547 if (new_exe_file)
1548 get_file(new_exe_file);
1549 if (mm->exe_file)
1550 fput(mm->exe_file);
1551 mm->exe_file = new_exe_file;
1552 mm->num_exe_file_vmas = 0;
1553 }
1554
1555 struct file *get_mm_exe_file(struct mm_struct *mm)
1556 {
1557 struct file *exe_file;
1558
1559 /* We need mmap_sem to protect against races with removal of
1560 * VM_EXECUTABLE vmas */
1561 down_read(&mm->mmap_sem);
1562 exe_file = mm->exe_file;
1563 if (exe_file)
1564 get_file(exe_file);
1565 up_read(&mm->mmap_sem);
1566 return exe_file;
1567 }
1568
1569 void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
1570 {
1571 /* It's safe to write the exe_file pointer without exe_file_lock because
1572 * this is called during fork when the task is not yet in /proc */
1573 newmm->exe_file = get_mm_exe_file(oldmm);
1574 }
1575
1576 static int proc_exe_link(struct inode *inode, struct path *exe_path)
1577 {
1578 struct task_struct *task;
1579 struct mm_struct *mm;
1580 struct file *exe_file;
1581
1582 task = get_proc_task(inode);
1583 if (!task)
1584 return -ENOENT;
1585 mm = get_task_mm(task);
1586 put_task_struct(task);
1587 if (!mm)
1588 return -ENOENT;
1589 exe_file = get_mm_exe_file(mm);
1590 mmput(mm);
1591 if (exe_file) {
1592 *exe_path = exe_file->f_path;
1593 path_get(&exe_file->f_path);
1594 fput(exe_file);
1595 return 0;
1596 } else
1597 return -ENOENT;
1598 }
1599
1600 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1601 {
1602 struct inode *inode = dentry->d_inode;
1603 int error = -EACCES;
1604
1605 /* We don't need a base pointer in the /proc filesystem */
1606 path_put(&nd->path);
1607
1608 /* Are we allowed to snoop on the tasks file descriptors? */
1609 if (!proc_fd_access_allowed(inode))
1610 goto out;
1611
1612 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1613 out:
1614 return ERR_PTR(error);
1615 }
1616
1617 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1618 {
1619 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1620 char *pathname;
1621 int len;
1622
1623 if (!tmp)
1624 return -ENOMEM;
1625
1626 pathname = d_path(path, tmp, PAGE_SIZE);
1627 len = PTR_ERR(pathname);
1628 if (IS_ERR(pathname))
1629 goto out;
1630 len = tmp + PAGE_SIZE - 1 - pathname;
1631
1632 if (len > buflen)
1633 len = buflen;
1634 if (copy_to_user(buffer, pathname, len))
1635 len = -EFAULT;
1636 out:
1637 free_page((unsigned long)tmp);
1638 return len;
1639 }
1640
1641 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1642 {
1643 int error = -EACCES;
1644 struct inode *inode = dentry->d_inode;
1645 struct path path;
1646
1647 /* Are we allowed to snoop on the tasks file descriptors? */
1648 if (!proc_fd_access_allowed(inode))
1649 goto out;
1650
1651 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1652 if (error)
1653 goto out;
1654
1655 error = do_proc_readlink(&path, buffer, buflen);
1656 path_put(&path);
1657 out:
1658 return error;
1659 }
1660
1661 static const struct inode_operations proc_pid_link_inode_operations = {
1662 .readlink = proc_pid_readlink,
1663 .follow_link = proc_pid_follow_link,
1664 .setattr = proc_setattr,
1665 };
1666
1667
1668 /* building an inode */
1669
1670 static int task_dumpable(struct task_struct *task)
1671 {
1672 int dumpable = 0;
1673 struct mm_struct *mm;
1674
1675 task_lock(task);
1676 mm = task->mm;
1677 if (mm)
1678 dumpable = get_dumpable(mm);
1679 task_unlock(task);
1680 if(dumpable == 1)
1681 return 1;
1682 return 0;
1683 }
1684
1685
1686 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1687 {
1688 struct inode * inode;
1689 struct proc_inode *ei;
1690 const struct cred *cred;
1691
1692 /* We need a new inode */
1693
1694 inode = new_inode(sb);
1695 if (!inode)
1696 goto out;
1697
1698 /* Common stuff */
1699 ei = PROC_I(inode);
1700 inode->i_ino = get_next_ino();
1701 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1702 inode->i_op = &proc_def_inode_operations;
1703
1704 /*
1705 * grab the reference to task.
1706 */
1707 ei->pid = get_task_pid(task, PIDTYPE_PID);
1708 if (!ei->pid)
1709 goto out_unlock;
1710
1711 if (task_dumpable(task)) {
1712 rcu_read_lock();
1713 cred = __task_cred(task);
1714 inode->i_uid = cred->euid;
1715 inode->i_gid = cred->egid;
1716 rcu_read_unlock();
1717 }
1718 security_task_to_inode(task, inode);
1719
1720 out:
1721 return inode;
1722
1723 out_unlock:
1724 iput(inode);
1725 return NULL;
1726 }
1727
1728 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1729 {
1730 struct inode *inode = dentry->d_inode;
1731 struct task_struct *task;
1732 const struct cred *cred;
1733
1734 generic_fillattr(inode, stat);
1735
1736 rcu_read_lock();
1737 stat->uid = 0;
1738 stat->gid = 0;
1739 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1740 if (task) {
1741 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1742 task_dumpable(task)) {
1743 cred = __task_cred(task);
1744 stat->uid = cred->euid;
1745 stat->gid = cred->egid;
1746 }
1747 }
1748 rcu_read_unlock();
1749 return 0;
1750 }
1751
1752 /* dentry stuff */
1753
1754 /*
1755 * Exceptional case: normally we are not allowed to unhash a busy
1756 * directory. In this case, however, we can do it - no aliasing problems
1757 * due to the way we treat inodes.
1758 *
1759 * Rewrite the inode's ownerships here because the owning task may have
1760 * performed a setuid(), etc.
1761 *
1762 * Before the /proc/pid/status file was created the only way to read
1763 * the effective uid of a /process was to stat /proc/pid. Reading
1764 * /proc/pid/status is slow enough that procps and other packages
1765 * kept stating /proc/pid. To keep the rules in /proc simple I have
1766 * made this apply to all per process world readable and executable
1767 * directories.
1768 */
1769 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1770 {
1771 struct inode *inode;
1772 struct task_struct *task;
1773 const struct cred *cred;
1774
1775 if (nd && nd->flags & LOOKUP_RCU)
1776 return -ECHILD;
1777
1778 inode = dentry->d_inode;
1779 task = get_proc_task(inode);
1780
1781 if (task) {
1782 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1783 task_dumpable(task)) {
1784 rcu_read_lock();
1785 cred = __task_cred(task);
1786 inode->i_uid = cred->euid;
1787 inode->i_gid = cred->egid;
1788 rcu_read_unlock();
1789 } else {
1790 inode->i_uid = 0;
1791 inode->i_gid = 0;
1792 }
1793 inode->i_mode &= ~(S_ISUID | S_ISGID);
1794 security_task_to_inode(task, inode);
1795 put_task_struct(task);
1796 return 1;
1797 }
1798 d_drop(dentry);
1799 return 0;
1800 }
1801
1802 static int pid_delete_dentry(const struct dentry * dentry)
1803 {
1804 /* Is the task we represent dead?
1805 * If so, then don't put the dentry on the lru list,
1806 * kill it immediately.
1807 */
1808 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1809 }
1810
1811 static const struct dentry_operations pid_dentry_operations =
1812 {
1813 .d_revalidate = pid_revalidate,
1814 .d_delete = pid_delete_dentry,
1815 };
1816
1817 /* Lookups */
1818
1819 typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1820 struct task_struct *, const void *);
1821
1822 /*
1823 * Fill a directory entry.
1824 *
1825 * If possible create the dcache entry and derive our inode number and
1826 * file type from dcache entry.
1827 *
1828 * Since all of the proc inode numbers are dynamically generated, the inode
1829 * numbers do not exist until the inode is cache. This means creating the
1830 * the dcache entry in readdir is necessary to keep the inode numbers
1831 * reported by readdir in sync with the inode numbers reported
1832 * by stat.
1833 */
1834 static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1835 char *name, int len,
1836 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1837 {
1838 struct dentry *child, *dir = filp->f_path.dentry;
1839 struct inode *inode;
1840 struct qstr qname;
1841 ino_t ino = 0;
1842 unsigned type = DT_UNKNOWN;
1843
1844 qname.name = name;
1845 qname.len = len;
1846 qname.hash = full_name_hash(name, len);
1847
1848 child = d_lookup(dir, &qname);
1849 if (!child) {
1850 struct dentry *new;
1851 new = d_alloc(dir, &qname);
1852 if (new) {
1853 child = instantiate(dir->d_inode, new, task, ptr);
1854 if (child)
1855 dput(new);
1856 else
1857 child = new;
1858 }
1859 }
1860 if (!child || IS_ERR(child) || !child->d_inode)
1861 goto end_instantiate;
1862 inode = child->d_inode;
1863 if (inode) {
1864 ino = inode->i_ino;
1865 type = inode->i_mode >> 12;
1866 }
1867 dput(child);
1868 end_instantiate:
1869 if (!ino)
1870 ino = find_inode_number(dir, &qname);
1871 if (!ino)
1872 ino = 1;
1873 return filldir(dirent, name, len, filp->f_pos, ino, type);
1874 }
1875
1876 static unsigned name_to_int(struct dentry *dentry)
1877 {
1878 const char *name = dentry->d_name.name;
1879 int len = dentry->d_name.len;
1880 unsigned n = 0;
1881
1882 if (len > 1 && *name == '0')
1883 goto out;
1884 while (len-- > 0) {
1885 unsigned c = *name++ - '0';
1886 if (c > 9)
1887 goto out;
1888 if (n >= (~0U-9)/10)
1889 goto out;
1890 n *= 10;
1891 n += c;
1892 }
1893 return n;
1894 out:
1895 return ~0U;
1896 }
1897
1898 #define PROC_FDINFO_MAX 64
1899
1900 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1901 {
1902 struct task_struct *task = get_proc_task(inode);
1903 struct files_struct *files = NULL;
1904 struct file *file;
1905 int fd = proc_fd(inode);
1906
1907 if (task) {
1908 files = get_files_struct(task);
1909 put_task_struct(task);
1910 }
1911 if (files) {
1912 /*
1913 * We are not taking a ref to the file structure, so we must
1914 * hold ->file_lock.
1915 */
1916 spin_lock(&files->file_lock);
1917 file = fcheck_files(files, fd);
1918 if (file) {
1919 if (path) {
1920 *path = file->f_path;
1921 path_get(&file->f_path);
1922 }
1923 if (info)
1924 snprintf(info, PROC_FDINFO_MAX,
1925 "pos:\t%lli\n"
1926 "flags:\t0%o\n",
1927 (long long) file->f_pos,
1928 file->f_flags);
1929 spin_unlock(&files->file_lock);
1930 put_files_struct(files);
1931 return 0;
1932 }
1933 spin_unlock(&files->file_lock);
1934 put_files_struct(files);
1935 }
1936 return -ENOENT;
1937 }
1938
1939 static int proc_fd_link(struct inode *inode, struct path *path)
1940 {
1941 return proc_fd_info(inode, path, NULL);
1942 }
1943
1944 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1945 {
1946 struct inode *inode;
1947 struct task_struct *task;
1948 int fd;
1949 struct files_struct *files;
1950 const struct cred *cred;
1951
1952 if (nd && nd->flags & LOOKUP_RCU)
1953 return -ECHILD;
1954
1955 inode = dentry->d_inode;
1956 task = get_proc_task(inode);
1957 fd = proc_fd(inode);
1958
1959 if (task) {
1960 files = get_files_struct(task);
1961 if (files) {
1962 rcu_read_lock();
1963 if (fcheck_files(files, fd)) {
1964 rcu_read_unlock();
1965 put_files_struct(files);
1966 if (task_dumpable(task)) {
1967 rcu_read_lock();
1968 cred = __task_cred(task);
1969 inode->i_uid = cred->euid;
1970 inode->i_gid = cred->egid;
1971 rcu_read_unlock();
1972 } else {
1973 inode->i_uid = 0;
1974 inode->i_gid = 0;
1975 }
1976 inode->i_mode &= ~(S_ISUID | S_ISGID);
1977 security_task_to_inode(task, inode);
1978 put_task_struct(task);
1979 return 1;
1980 }
1981 rcu_read_unlock();
1982 put_files_struct(files);
1983 }
1984 put_task_struct(task);
1985 }
1986 d_drop(dentry);
1987 return 0;
1988 }
1989
1990 static const struct dentry_operations tid_fd_dentry_operations =
1991 {
1992 .d_revalidate = tid_fd_revalidate,
1993 .d_delete = pid_delete_dentry,
1994 };
1995
1996 static struct dentry *proc_fd_instantiate(struct inode *dir,
1997 struct dentry *dentry, struct task_struct *task, const void *ptr)
1998 {
1999 unsigned fd = *(const unsigned *)ptr;
2000 struct file *file;
2001 struct files_struct *files;
2002 struct inode *inode;
2003 struct proc_inode *ei;
2004 struct dentry *error = ERR_PTR(-ENOENT);
2005
2006 inode = proc_pid_make_inode(dir->i_sb, task);
2007 if (!inode)
2008 goto out;
2009 ei = PROC_I(inode);
2010 ei->fd = fd;
2011 files = get_files_struct(task);
2012 if (!files)
2013 goto out_iput;
2014 inode->i_mode = S_IFLNK;
2015
2016 /*
2017 * We are not taking a ref to the file structure, so we must
2018 * hold ->file_lock.
2019 */
2020 spin_lock(&files->file_lock);
2021 file = fcheck_files(files, fd);
2022 if (!file)
2023 goto out_unlock;
2024 if (file->f_mode & FMODE_READ)
2025 inode->i_mode |= S_IRUSR | S_IXUSR;
2026 if (file->f_mode & FMODE_WRITE)
2027 inode->i_mode |= S_IWUSR | S_IXUSR;
2028 spin_unlock(&files->file_lock);
2029 put_files_struct(files);
2030
2031 inode->i_op = &proc_pid_link_inode_operations;
2032 inode->i_size = 64;
2033 ei->op.proc_get_link = proc_fd_link;
2034 d_set_d_op(dentry, &tid_fd_dentry_operations);
2035 d_add(dentry, inode);
2036 /* Close the race of the process dying before we return the dentry */
2037 if (tid_fd_revalidate(dentry, NULL))
2038 error = NULL;
2039
2040 out:
2041 return error;
2042 out_unlock:
2043 spin_unlock(&files->file_lock);
2044 put_files_struct(files);
2045 out_iput:
2046 iput(inode);
2047 goto out;
2048 }
2049
2050 static struct dentry *proc_lookupfd_common(struct inode *dir,
2051 struct dentry *dentry,
2052 instantiate_t instantiate)
2053 {
2054 struct task_struct *task = get_proc_task(dir);
2055 unsigned fd = name_to_int(dentry);
2056 struct dentry *result = ERR_PTR(-ENOENT);
2057
2058 if (!task)
2059 goto out_no_task;
2060 if (fd == ~0U)
2061 goto out;
2062
2063 result = instantiate(dir, dentry, task, &fd);
2064 out:
2065 put_task_struct(task);
2066 out_no_task:
2067 return result;
2068 }
2069
2070 static int proc_readfd_common(struct file * filp, void * dirent,
2071 filldir_t filldir, instantiate_t instantiate)
2072 {
2073 struct dentry *dentry = filp->f_path.dentry;
2074 struct inode *inode = dentry->d_inode;
2075 struct task_struct *p = get_proc_task(inode);
2076 unsigned int fd, ino;
2077 int retval;
2078 struct files_struct * files;
2079
2080 retval = -ENOENT;
2081 if (!p)
2082 goto out_no_task;
2083 retval = 0;
2084
2085 fd = filp->f_pos;
2086 switch (fd) {
2087 case 0:
2088 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
2089 goto out;
2090 filp->f_pos++;
2091 case 1:
2092 ino = parent_ino(dentry);
2093 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2094 goto out;
2095 filp->f_pos++;
2096 default:
2097 files = get_files_struct(p);
2098 if (!files)
2099 goto out;
2100 rcu_read_lock();
2101 for (fd = filp->f_pos-2;
2102 fd < files_fdtable(files)->max_fds;
2103 fd++, filp->f_pos++) {
2104 char name[PROC_NUMBUF];
2105 int len;
2106
2107 if (!fcheck_files(files, fd))
2108 continue;
2109 rcu_read_unlock();
2110
2111 len = snprintf(name, sizeof(name), "%d", fd);
2112 if (proc_fill_cache(filp, dirent, filldir,
2113 name, len, instantiate,
2114 p, &fd) < 0) {
2115 rcu_read_lock();
2116 break;
2117 }
2118 rcu_read_lock();
2119 }
2120 rcu_read_unlock();
2121 put_files_struct(files);
2122 }
2123 out:
2124 put_task_struct(p);
2125 out_no_task:
2126 return retval;
2127 }
2128
2129 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
2130 struct nameidata *nd)
2131 {
2132 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
2133 }
2134
2135 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
2136 {
2137 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
2138 }
2139
2140 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
2141 size_t len, loff_t *ppos)
2142 {
2143 char tmp[PROC_FDINFO_MAX];
2144 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
2145 if (!err)
2146 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
2147 return err;
2148 }
2149
2150 static const struct file_operations proc_fdinfo_file_operations = {
2151 .open = nonseekable_open,
2152 .read = proc_fdinfo_read,
2153 .llseek = no_llseek,
2154 };
2155
2156 static const struct file_operations proc_fd_operations = {
2157 .read = generic_read_dir,
2158 .readdir = proc_readfd,
2159 .llseek = default_llseek,
2160 };
2161
2162 /*
2163 * /proc/pid/fd needs a special permission handler so that a process can still
2164 * access /proc/self/fd after it has executed a setuid().
2165 */
2166 static int proc_fd_permission(struct inode *inode, int mask, unsigned int flags)
2167 {
2168 int rv;
2169
2170 if (flags & IPERM_FLAG_RCU)
2171 return -ECHILD;
2172 rv = generic_permission(inode, mask, flags, NULL);
2173 if (rv == 0)
2174 return 0;
2175 if (task_pid(current) == proc_pid(inode))
2176 rv = 0;
2177 return rv;
2178 }
2179
2180 /*
2181 * proc directories can do almost nothing..
2182 */
2183 static const struct inode_operations proc_fd_inode_operations = {
2184 .lookup = proc_lookupfd,
2185 .permission = proc_fd_permission,
2186 .setattr = proc_setattr,
2187 };
2188
2189 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2190 struct dentry *dentry, struct task_struct *task, const void *ptr)
2191 {
2192 unsigned fd = *(unsigned *)ptr;
2193 struct inode *inode;
2194 struct proc_inode *ei;
2195 struct dentry *error = ERR_PTR(-ENOENT);
2196
2197 inode = proc_pid_make_inode(dir->i_sb, task);
2198 if (!inode)
2199 goto out;
2200 ei = PROC_I(inode);
2201 ei->fd = fd;
2202 inode->i_mode = S_IFREG | S_IRUSR;
2203 inode->i_fop = &proc_fdinfo_file_operations;
2204 d_set_d_op(dentry, &tid_fd_dentry_operations);
2205 d_add(dentry, inode);
2206 /* Close the race of the process dying before we return the dentry */
2207 if (tid_fd_revalidate(dentry, NULL))
2208 error = NULL;
2209
2210 out:
2211 return error;
2212 }
2213
2214 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2215 struct dentry *dentry,
2216 struct nameidata *nd)
2217 {
2218 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2219 }
2220
2221 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2222 {
2223 return proc_readfd_common(filp, dirent, filldir,
2224 proc_fdinfo_instantiate);
2225 }
2226
2227 static const struct file_operations proc_fdinfo_operations = {
2228 .read = generic_read_dir,
2229 .readdir = proc_readfdinfo,
2230 .llseek = default_llseek,
2231 };
2232
2233 /*
2234 * proc directories can do almost nothing..
2235 */
2236 static const struct inode_operations proc_fdinfo_inode_operations = {
2237 .lookup = proc_lookupfdinfo,
2238 .setattr = proc_setattr,
2239 };
2240
2241
2242 static struct dentry *proc_pident_instantiate(struct inode *dir,
2243 struct dentry *dentry, struct task_struct *task, const void *ptr)
2244 {
2245 const struct pid_entry *p = ptr;
2246 struct inode *inode;
2247 struct proc_inode *ei;
2248 struct dentry *error = ERR_PTR(-ENOENT);
2249
2250 inode = proc_pid_make_inode(dir->i_sb, task);
2251 if (!inode)
2252 goto out;
2253
2254 ei = PROC_I(inode);
2255 inode->i_mode = p->mode;
2256 if (S_ISDIR(inode->i_mode))
2257 inode->i_nlink = 2; /* Use getattr to fix if necessary */
2258 if (p->iop)
2259 inode->i_op = p->iop;
2260 if (p->fop)
2261 inode->i_fop = p->fop;
2262 ei->op = p->op;
2263 d_set_d_op(dentry, &pid_dentry_operations);
2264 d_add(dentry, inode);
2265 /* Close the race of the process dying before we return the dentry */
2266 if (pid_revalidate(dentry, NULL))
2267 error = NULL;
2268 out:
2269 return error;
2270 }
2271
2272 static struct dentry *proc_pident_lookup(struct inode *dir,
2273 struct dentry *dentry,
2274 const struct pid_entry *ents,
2275 unsigned int nents)
2276 {
2277 struct dentry *error;
2278 struct task_struct *task = get_proc_task(dir);
2279 const struct pid_entry *p, *last;
2280
2281 error = ERR_PTR(-ENOENT);
2282
2283 if (!task)
2284 goto out_no_task;
2285
2286 /*
2287 * Yes, it does not scale. And it should not. Don't add
2288 * new entries into /proc/<tgid>/ without very good reasons.
2289 */
2290 last = &ents[nents - 1];
2291 for (p = ents; p <= last; p++) {
2292 if (p->len != dentry->d_name.len)
2293 continue;
2294 if (!memcmp(dentry->d_name.name, p->name, p->len))
2295 break;
2296 }
2297 if (p > last)
2298 goto out;
2299
2300 error = proc_pident_instantiate(dir, dentry, task, p);
2301 out:
2302 put_task_struct(task);
2303 out_no_task:
2304 return error;
2305 }
2306
2307 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2308 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2309 {
2310 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2311 proc_pident_instantiate, task, p);
2312 }
2313
2314 static int proc_pident_readdir(struct file *filp,
2315 void *dirent, filldir_t filldir,
2316 const struct pid_entry *ents, unsigned int nents)
2317 {
2318 int i;
2319 struct dentry *dentry = filp->f_path.dentry;
2320 struct inode *inode = dentry->d_inode;
2321 struct task_struct *task = get_proc_task(inode);
2322 const struct pid_entry *p, *last;
2323 ino_t ino;
2324 int ret;
2325
2326 ret = -ENOENT;
2327 if (!task)
2328 goto out_no_task;
2329
2330 ret = 0;
2331 i = filp->f_pos;
2332 switch (i) {
2333 case 0:
2334 ino = inode->i_ino;
2335 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2336 goto out;
2337 i++;
2338 filp->f_pos++;
2339 /* fall through */
2340 case 1:
2341 ino = parent_ino(dentry);
2342 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2343 goto out;
2344 i++;
2345 filp->f_pos++;
2346 /* fall through */
2347 default:
2348 i -= 2;
2349 if (i >= nents) {
2350 ret = 1;
2351 goto out;
2352 }
2353 p = ents + i;
2354 last = &ents[nents - 1];
2355 while (p <= last) {
2356 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2357 goto out;
2358 filp->f_pos++;
2359 p++;
2360 }
2361 }
2362
2363 ret = 1;
2364 out:
2365 put_task_struct(task);
2366 out_no_task:
2367 return ret;
2368 }
2369
2370 #ifdef CONFIG_SECURITY
2371 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2372 size_t count, loff_t *ppos)
2373 {
2374 struct inode * inode = file->f_path.dentry->d_inode;
2375 char *p = NULL;
2376 ssize_t length;
2377 struct task_struct *task = get_proc_task(inode);
2378
2379 if (!task)
2380 return -ESRCH;
2381
2382 length = security_getprocattr(task,
2383 (char*)file->f_path.dentry->d_name.name,
2384 &p);
2385 put_task_struct(task);
2386 if (length > 0)
2387 length = simple_read_from_buffer(buf, count, ppos, p, length);
2388 kfree(p);
2389 return length;
2390 }
2391
2392 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2393 size_t count, loff_t *ppos)
2394 {
2395 struct inode * inode = file->f_path.dentry->d_inode;
2396 char *page;
2397 ssize_t length;
2398 struct task_struct *task = get_proc_task(inode);
2399
2400 length = -ESRCH;
2401 if (!task)
2402 goto out_no_task;
2403 if (count > PAGE_SIZE)
2404 count = PAGE_SIZE;
2405
2406 /* No partial writes. */
2407 length = -EINVAL;
2408 if (*ppos != 0)
2409 goto out;
2410
2411 length = -ENOMEM;
2412 page = (char*)__get_free_page(GFP_TEMPORARY);
2413 if (!page)
2414 goto out;
2415
2416 length = -EFAULT;
2417 if (copy_from_user(page, buf, count))
2418 goto out_free;
2419
2420 /* Guard against adverse ptrace interaction */
2421 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2422 if (length < 0)
2423 goto out_free;
2424
2425 length = security_setprocattr(task,
2426 (char*)file->f_path.dentry->d_name.name,
2427 (void*)page, count);
2428 mutex_unlock(&task->signal->cred_guard_mutex);
2429 out_free:
2430 free_page((unsigned long) page);
2431 out:
2432 put_task_struct(task);
2433 out_no_task:
2434 return length;
2435 }
2436
2437 static const struct file_operations proc_pid_attr_operations = {
2438 .read = proc_pid_attr_read,
2439 .write = proc_pid_attr_write,
2440 .llseek = generic_file_llseek,
2441 };
2442
2443 static const struct pid_entry attr_dir_stuff[] = {
2444 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2445 REG("prev", S_IRUGO, proc_pid_attr_operations),
2446 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2447 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2448 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2449 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2450 };
2451
2452 static int proc_attr_dir_readdir(struct file * filp,
2453 void * dirent, filldir_t filldir)
2454 {
2455 return proc_pident_readdir(filp,dirent,filldir,
2456 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2457 }
2458
2459 static const struct file_operations proc_attr_dir_operations = {
2460 .read = generic_read_dir,
2461 .readdir = proc_attr_dir_readdir,
2462 .llseek = default_llseek,
2463 };
2464
2465 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2466 struct dentry *dentry, struct nameidata *nd)
2467 {
2468 return proc_pident_lookup(dir, dentry,
2469 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2470 }
2471
2472 static const struct inode_operations proc_attr_dir_inode_operations = {
2473 .lookup = proc_attr_dir_lookup,
2474 .getattr = pid_getattr,
2475 .setattr = proc_setattr,
2476 };
2477
2478 #endif
2479
2480 #ifdef CONFIG_ELF_CORE
2481 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2482 size_t count, loff_t *ppos)
2483 {
2484 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2485 struct mm_struct *mm;
2486 char buffer[PROC_NUMBUF];
2487 size_t len;
2488 int ret;
2489
2490 if (!task)
2491 return -ESRCH;
2492
2493 ret = 0;
2494 mm = get_task_mm(task);
2495 if (mm) {
2496 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2497 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2498 MMF_DUMP_FILTER_SHIFT));
2499 mmput(mm);
2500 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2501 }
2502
2503 put_task_struct(task);
2504
2505 return ret;
2506 }
2507
2508 static ssize_t proc_coredump_filter_write(struct file *file,
2509 const char __user *buf,
2510 size_t count,
2511 loff_t *ppos)
2512 {
2513 struct task_struct *task;
2514 struct mm_struct *mm;
2515 char buffer[PROC_NUMBUF], *end;
2516 unsigned int val;
2517 int ret;
2518 int i;
2519 unsigned long mask;
2520
2521 ret = -EFAULT;
2522 memset(buffer, 0, sizeof(buffer));
2523 if (count > sizeof(buffer) - 1)
2524 count = sizeof(buffer) - 1;
2525 if (copy_from_user(buffer, buf, count))
2526 goto out_no_task;
2527
2528 ret = -EINVAL;
2529 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2530 if (*end == '\n')
2531 end++;
2532 if (end - buffer == 0)
2533 goto out_no_task;
2534
2535 ret = -ESRCH;
2536 task = get_proc_task(file->f_dentry->d_inode);
2537 if (!task)
2538 goto out_no_task;
2539
2540 ret = end - buffer;
2541 mm = get_task_mm(task);
2542 if (!mm)
2543 goto out_no_mm;
2544
2545 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2546 if (val & mask)
2547 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2548 else
2549 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2550 }
2551
2552 mmput(mm);
2553 out_no_mm:
2554 put_task_struct(task);
2555 out_no_task:
2556 return ret;
2557 }
2558
2559 static const struct file_operations proc_coredump_filter_operations = {
2560 .read = proc_coredump_filter_read,
2561 .write = proc_coredump_filter_write,
2562 .llseek = generic_file_llseek,
2563 };
2564 #endif
2565
2566 /*
2567 * /proc/self:
2568 */
2569 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2570 int buflen)
2571 {
2572 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2573 pid_t tgid = task_tgid_nr_ns(current, ns);
2574 char tmp[PROC_NUMBUF];
2575 if (!tgid)
2576 return -ENOENT;
2577 sprintf(tmp, "%d", tgid);
2578 return vfs_readlink(dentry,buffer,buflen,tmp);
2579 }
2580
2581 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2582 {
2583 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2584 pid_t tgid = task_tgid_nr_ns(current, ns);
2585 char *name = ERR_PTR(-ENOENT);
2586 if (tgid) {
2587 name = __getname();
2588 if (!name)
2589 name = ERR_PTR(-ENOMEM);
2590 else
2591 sprintf(name, "%d", tgid);
2592 }
2593 nd_set_link(nd, name);
2594 return NULL;
2595 }
2596
2597 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2598 void *cookie)
2599 {
2600 char *s = nd_get_link(nd);
2601 if (!IS_ERR(s))
2602 __putname(s);
2603 }
2604
2605 static const struct inode_operations proc_self_inode_operations = {
2606 .readlink = proc_self_readlink,
2607 .follow_link = proc_self_follow_link,
2608 .put_link = proc_self_put_link,
2609 };
2610
2611 /*
2612 * proc base
2613 *
2614 * These are the directory entries in the root directory of /proc
2615 * that properly belong to the /proc filesystem, as they describe
2616 * describe something that is process related.
2617 */
2618 static const struct pid_entry proc_base_stuff[] = {
2619 NOD("self", S_IFLNK|S_IRWXUGO,
2620 &proc_self_inode_operations, NULL, {}),
2621 };
2622
2623 /*
2624 * Exceptional case: normally we are not allowed to unhash a busy
2625 * directory. In this case, however, we can do it - no aliasing problems
2626 * due to the way we treat inodes.
2627 */
2628 static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
2629 {
2630 struct inode *inode;
2631 struct task_struct *task;
2632
2633 if (nd->flags & LOOKUP_RCU)
2634 return -ECHILD;
2635
2636 inode = dentry->d_inode;
2637 task = get_proc_task(inode);
2638 if (task) {
2639 put_task_struct(task);
2640 return 1;
2641 }
2642 d_drop(dentry);
2643 return 0;
2644 }
2645
2646 static const struct dentry_operations proc_base_dentry_operations =
2647 {
2648 .d_revalidate = proc_base_revalidate,
2649 .d_delete = pid_delete_dentry,
2650 };
2651
2652 static struct dentry *proc_base_instantiate(struct inode *dir,
2653 struct dentry *dentry, struct task_struct *task, const void *ptr)
2654 {
2655 const struct pid_entry *p = ptr;
2656 struct inode *inode;
2657 struct proc_inode *ei;
2658 struct dentry *error;
2659
2660 /* Allocate the inode */
2661 error = ERR_PTR(-ENOMEM);
2662 inode = new_inode(dir->i_sb);
2663 if (!inode)
2664 goto out;
2665
2666 /* Initialize the inode */
2667 ei = PROC_I(inode);
2668 inode->i_ino = get_next_ino();
2669 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2670
2671 /*
2672 * grab the reference to the task.
2673 */
2674 ei->pid = get_task_pid(task, PIDTYPE_PID);
2675 if (!ei->pid)
2676 goto out_iput;
2677
2678 inode->i_mode = p->mode;
2679 if (S_ISDIR(inode->i_mode))
2680 inode->i_nlink = 2;
2681 if (S_ISLNK(inode->i_mode))
2682 inode->i_size = 64;
2683 if (p->iop)
2684 inode->i_op = p->iop;
2685 if (p->fop)
2686 inode->i_fop = p->fop;
2687 ei->op = p->op;
2688 d_set_d_op(dentry, &proc_base_dentry_operations);
2689 d_add(dentry, inode);
2690 error = NULL;
2691 out:
2692 return error;
2693 out_iput:
2694 iput(inode);
2695 goto out;
2696 }
2697
2698 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2699 {
2700 struct dentry *error;
2701 struct task_struct *task = get_proc_task(dir);
2702 const struct pid_entry *p, *last;
2703
2704 error = ERR_PTR(-ENOENT);
2705
2706 if (!task)
2707 goto out_no_task;
2708
2709 /* Lookup the directory entry */
2710 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2711 for (p = proc_base_stuff; p <= last; p++) {
2712 if (p->len != dentry->d_name.len)
2713 continue;
2714 if (!memcmp(dentry->d_name.name, p->name, p->len))
2715 break;
2716 }
2717 if (p > last)
2718 goto out;
2719
2720 error = proc_base_instantiate(dir, dentry, task, p);
2721
2722 out:
2723 put_task_struct(task);
2724 out_no_task:
2725 return error;
2726 }
2727
2728 static int proc_base_fill_cache(struct file *filp, void *dirent,
2729 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2730 {
2731 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2732 proc_base_instantiate, task, p);
2733 }
2734
2735 #ifdef CONFIG_TASK_IO_ACCOUNTING
2736 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2737 {
2738 struct task_io_accounting acct = task->ioac;
2739 unsigned long flags;
2740
2741 if (whole && lock_task_sighand(task, &flags)) {
2742 struct task_struct *t = task;
2743
2744 task_io_accounting_add(&acct, &task->signal->ioac);
2745 while_each_thread(task, t)
2746 task_io_accounting_add(&acct, &t->ioac);
2747
2748 unlock_task_sighand(task, &flags);
2749 }
2750 return sprintf(buffer,
2751 "rchar: %llu\n"
2752 "wchar: %llu\n"
2753 "syscr: %llu\n"
2754 "syscw: %llu\n"
2755 "read_bytes: %llu\n"
2756 "write_bytes: %llu\n"
2757 "cancelled_write_bytes: %llu\n",
2758 (unsigned long long)acct.rchar,
2759 (unsigned long long)acct.wchar,
2760 (unsigned long long)acct.syscr,
2761 (unsigned long long)acct.syscw,
2762 (unsigned long long)acct.read_bytes,
2763 (unsigned long long)acct.write_bytes,
2764 (unsigned long long)acct.cancelled_write_bytes);
2765 }
2766
2767 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2768 {
2769 return do_io_accounting(task, buffer, 0);
2770 }
2771
2772 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2773 {
2774 return do_io_accounting(task, buffer, 1);
2775 }
2776 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2777
2778 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2779 struct pid *pid, struct task_struct *task)
2780 {
2781 seq_printf(m, "%08x\n", task->personality);
2782 return 0;
2783 }
2784
2785 /*
2786 * Thread groups
2787 */
2788 static const struct file_operations proc_task_operations;
2789 static const struct inode_operations proc_task_inode_operations;
2790
2791 static const struct pid_entry tgid_base_stuff[] = {
2792 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2793 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2794 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2795 #ifdef CONFIG_NET
2796 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2797 #endif
2798 REG("environ", S_IRUSR, proc_environ_operations),
2799 INF("auxv", S_IRUSR, proc_pid_auxv),
2800 ONE("status", S_IRUGO, proc_pid_status),
2801 ONE("personality", S_IRUSR, proc_pid_personality),
2802 INF("limits", S_IRUGO, proc_pid_limits),
2803 #ifdef CONFIG_SCHED_DEBUG
2804 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2805 #endif
2806 #ifdef CONFIG_SCHED_AUTOGROUP
2807 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2808 #endif
2809 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2810 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2811 INF("syscall", S_IRUSR, proc_pid_syscall),
2812 #endif
2813 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2814 ONE("stat", S_IRUGO, proc_tgid_stat),
2815 ONE("statm", S_IRUGO, proc_pid_statm),
2816 REG("maps", S_IRUGO, proc_maps_operations),
2817 #ifdef CONFIG_NUMA
2818 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2819 #endif
2820 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2821 LNK("cwd", proc_cwd_link),
2822 LNK("root", proc_root_link),
2823 LNK("exe", proc_exe_link),
2824 REG("mounts", S_IRUGO, proc_mounts_operations),
2825 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2826 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2827 #ifdef CONFIG_PROC_PAGE_MONITOR
2828 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2829 REG("smaps", S_IRUGO, proc_smaps_operations),
2830 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2831 #endif
2832 #ifdef CONFIG_SECURITY
2833 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2834 #endif
2835 #ifdef CONFIG_KALLSYMS
2836 INF("wchan", S_IRUGO, proc_pid_wchan),
2837 #endif
2838 #ifdef CONFIG_STACKTRACE
2839 ONE("stack", S_IRUSR, proc_pid_stack),
2840 #endif
2841 #ifdef CONFIG_SCHEDSTATS
2842 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2843 #endif
2844 #ifdef CONFIG_LATENCYTOP
2845 REG("latency", S_IRUGO, proc_lstats_operations),
2846 #endif
2847 #ifdef CONFIG_PROC_PID_CPUSET
2848 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2849 #endif
2850 #ifdef CONFIG_CGROUPS
2851 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2852 #endif
2853 INF("oom_score", S_IRUGO, proc_oom_score),
2854 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2855 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2856 #ifdef CONFIG_AUDITSYSCALL
2857 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2858 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2859 #endif
2860 #ifdef CONFIG_FAULT_INJECTION
2861 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2862 #endif
2863 #ifdef CONFIG_ELF_CORE
2864 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2865 #endif
2866 #ifdef CONFIG_TASK_IO_ACCOUNTING
2867 INF("io", S_IRUGO, proc_tgid_io_accounting),
2868 #endif
2869 };
2870
2871 static int proc_tgid_base_readdir(struct file * filp,
2872 void * dirent, filldir_t filldir)
2873 {
2874 return proc_pident_readdir(filp,dirent,filldir,
2875 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2876 }
2877
2878 static const struct file_operations proc_tgid_base_operations = {
2879 .read = generic_read_dir,
2880 .readdir = proc_tgid_base_readdir,
2881 .llseek = default_llseek,
2882 };
2883
2884 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2885 return proc_pident_lookup(dir, dentry,
2886 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2887 }
2888
2889 static const struct inode_operations proc_tgid_base_inode_operations = {
2890 .lookup = proc_tgid_base_lookup,
2891 .getattr = pid_getattr,
2892 .setattr = proc_setattr,
2893 };
2894
2895 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2896 {
2897 struct dentry *dentry, *leader, *dir;
2898 char buf[PROC_NUMBUF];
2899 struct qstr name;
2900
2901 name.name = buf;
2902 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2903 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2904 if (dentry) {
2905 shrink_dcache_parent(dentry);
2906 d_drop(dentry);
2907 dput(dentry);
2908 }
2909
2910 name.name = buf;
2911 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2912 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2913 if (!leader)
2914 goto out;
2915
2916 name.name = "task";
2917 name.len = strlen(name.name);
2918 dir = d_hash_and_lookup(leader, &name);
2919 if (!dir)
2920 goto out_put_leader;
2921
2922 name.name = buf;
2923 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2924 dentry = d_hash_and_lookup(dir, &name);
2925 if (dentry) {
2926 shrink_dcache_parent(dentry);
2927 d_drop(dentry);
2928 dput(dentry);
2929 }
2930
2931 dput(dir);
2932 out_put_leader:
2933 dput(leader);
2934 out:
2935 return;
2936 }
2937
2938 /**
2939 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2940 * @task: task that should be flushed.
2941 *
2942 * When flushing dentries from proc, one needs to flush them from global
2943 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2944 * in. This call is supposed to do all of this job.
2945 *
2946 * Looks in the dcache for
2947 * /proc/@pid
2948 * /proc/@tgid/task/@pid
2949 * if either directory is present flushes it and all of it'ts children
2950 * from the dcache.
2951 *
2952 * It is safe and reasonable to cache /proc entries for a task until
2953 * that task exits. After that they just clog up the dcache with
2954 * useless entries, possibly causing useful dcache entries to be
2955 * flushed instead. This routine is proved to flush those useless
2956 * dcache entries at process exit time.
2957 *
2958 * NOTE: This routine is just an optimization so it does not guarantee
2959 * that no dcache entries will exist at process exit time it
2960 * just makes it very unlikely that any will persist.
2961 */
2962
2963 void proc_flush_task(struct task_struct *task)
2964 {
2965 int i;
2966 struct pid *pid, *tgid;
2967 struct upid *upid;
2968
2969 pid = task_pid(task);
2970 tgid = task_tgid(task);
2971
2972 for (i = 0; i <= pid->level; i++) {
2973 upid = &pid->numbers[i];
2974 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2975 tgid->numbers[i].nr);
2976 }
2977
2978 upid = &pid->numbers[pid->level];
2979 if (upid->nr == 1)
2980 pid_ns_release_proc(upid->ns);
2981 }
2982
2983 static struct dentry *proc_pid_instantiate(struct inode *dir,
2984 struct dentry * dentry,
2985 struct task_struct *task, const void *ptr)
2986 {
2987 struct dentry *error = ERR_PTR(-ENOENT);
2988 struct inode *inode;
2989
2990 inode = proc_pid_make_inode(dir->i_sb, task);
2991 if (!inode)
2992 goto out;
2993
2994 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2995 inode->i_op = &proc_tgid_base_inode_operations;
2996 inode->i_fop = &proc_tgid_base_operations;
2997 inode->i_flags|=S_IMMUTABLE;
2998
2999 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
3000 ARRAY_SIZE(tgid_base_stuff));
3001
3002 d_set_d_op(dentry, &pid_dentry_operations);
3003
3004 d_add(dentry, inode);
3005 /* Close the race of the process dying before we return the dentry */
3006 if (pid_revalidate(dentry, NULL))
3007 error = NULL;
3008 out:
3009 return error;
3010 }
3011
3012 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3013 {
3014 struct dentry *result;
3015 struct task_struct *task;
3016 unsigned tgid;
3017 struct pid_namespace *ns;
3018
3019 result = proc_base_lookup(dir, dentry);
3020 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3021 goto out;
3022
3023 tgid = name_to_int(dentry);
3024 if (tgid == ~0U)
3025 goto out;
3026
3027 ns = dentry->d_sb->s_fs_info;
3028 rcu_read_lock();
3029 task = find_task_by_pid_ns(tgid, ns);
3030 if (task)
3031 get_task_struct(task);
3032 rcu_read_unlock();
3033 if (!task)
3034 goto out;
3035
3036 result = proc_pid_instantiate(dir, dentry, task, NULL);
3037 put_task_struct(task);
3038 out:
3039 return result;
3040 }
3041
3042 /*
3043 * Find the first task with tgid >= tgid
3044 *
3045 */
3046 struct tgid_iter {
3047 unsigned int tgid;
3048 struct task_struct *task;
3049 };
3050 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3051 {
3052 struct pid *pid;
3053
3054 if (iter.task)
3055 put_task_struct(iter.task);
3056 rcu_read_lock();
3057 retry:
3058 iter.task = NULL;
3059 pid = find_ge_pid(iter.tgid, ns);
3060 if (pid) {
3061 iter.tgid = pid_nr_ns(pid, ns);
3062 iter.task = pid_task(pid, PIDTYPE_PID);
3063 /* What we to know is if the pid we have find is the
3064 * pid of a thread_group_leader. Testing for task
3065 * being a thread_group_leader is the obvious thing
3066 * todo but there is a window when it fails, due to
3067 * the pid transfer logic in de_thread.
3068 *
3069 * So we perform the straight forward test of seeing
3070 * if the pid we have found is the pid of a thread
3071 * group leader, and don't worry if the task we have
3072 * found doesn't happen to be a thread group leader.
3073 * As we don't care in the case of readdir.
3074 */
3075 if (!iter.task || !has_group_leader_pid(iter.task)) {
3076 iter.tgid += 1;
3077 goto retry;
3078 }
3079 get_task_struct(iter.task);
3080 }
3081 rcu_read_unlock();
3082 return iter;
3083 }
3084
3085 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
3086
3087 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3088 struct tgid_iter iter)
3089 {
3090 char name[PROC_NUMBUF];
3091 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
3092 return proc_fill_cache(filp, dirent, filldir, name, len,
3093 proc_pid_instantiate, iter.task, NULL);
3094 }
3095
3096 /* for the /proc/ directory itself, after non-process stuff has been done */
3097 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3098 {
3099 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3100 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
3101 struct tgid_iter iter;
3102 struct pid_namespace *ns;
3103
3104 if (!reaper)
3105 goto out_no_task;
3106
3107 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
3108 const struct pid_entry *p = &proc_base_stuff[nr];
3109 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
3110 goto out;
3111 }
3112
3113 ns = filp->f_dentry->d_sb->s_fs_info;
3114 iter.task = NULL;
3115 iter.tgid = filp->f_pos - TGID_OFFSET;
3116 for (iter = next_tgid(ns, iter);
3117 iter.task;
3118 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3119 filp->f_pos = iter.tgid + TGID_OFFSET;
3120 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
3121 put_task_struct(iter.task);
3122 goto out;
3123 }
3124 }
3125 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3126 out:
3127 put_task_struct(reaper);
3128 out_no_task:
3129 return 0;
3130 }
3131
3132 /*
3133 * Tasks
3134 */
3135 static const struct pid_entry tid_base_stuff[] = {
3136 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3137 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3138 REG("environ", S_IRUSR, proc_environ_operations),
3139 INF("auxv", S_IRUSR, proc_pid_auxv),
3140 ONE("status", S_IRUGO, proc_pid_status),
3141 ONE("personality", S_IRUSR, proc_pid_personality),
3142 INF("limits", S_IRUGO, proc_pid_limits),
3143 #ifdef CONFIG_SCHED_DEBUG
3144 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3145 #endif
3146 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3147 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3148 INF("syscall", S_IRUSR, proc_pid_syscall),
3149 #endif
3150 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3151 ONE("stat", S_IRUGO, proc_tid_stat),
3152 ONE("statm", S_IRUGO, proc_pid_statm),
3153 REG("maps", S_IRUGO, proc_maps_operations),
3154 #ifdef CONFIG_NUMA
3155 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
3156 #endif
3157 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3158 LNK("cwd", proc_cwd_link),
3159 LNK("root", proc_root_link),
3160 LNK("exe", proc_exe_link),
3161 REG("mounts", S_IRUGO, proc_mounts_operations),
3162 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3163 #ifdef CONFIG_PROC_PAGE_MONITOR
3164 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3165 REG("smaps", S_IRUGO, proc_smaps_operations),
3166 REG("pagemap", S_IRUSR, proc_pagemap_operations),
3167 #endif
3168 #ifdef CONFIG_SECURITY
3169 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3170 #endif
3171 #ifdef CONFIG_KALLSYMS
3172 INF("wchan", S_IRUGO, proc_pid_wchan),
3173 #endif
3174 #ifdef CONFIG_STACKTRACE
3175 ONE("stack", S_IRUSR, proc_pid_stack),
3176 #endif
3177 #ifdef CONFIG_SCHEDSTATS
3178 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3179 #endif
3180 #ifdef CONFIG_LATENCYTOP
3181 REG("latency", S_IRUGO, proc_lstats_operations),
3182 #endif
3183 #ifdef CONFIG_PROC_PID_CPUSET
3184 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3185 #endif
3186 #ifdef CONFIG_CGROUPS
3187 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3188 #endif
3189 INF("oom_score", S_IRUGO, proc_oom_score),
3190 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3191 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3192 #ifdef CONFIG_AUDITSYSCALL
3193 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3194 REG("sessionid", S_IRUSR, proc_sessionid_operations),
3195 #endif
3196 #ifdef CONFIG_FAULT_INJECTION
3197 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3198 #endif
3199 #ifdef CONFIG_TASK_IO_ACCOUNTING
3200 INF("io", S_IRUGO, proc_tid_io_accounting),
3201 #endif
3202 };
3203
3204 static int proc_tid_base_readdir(struct file * filp,
3205 void * dirent, filldir_t filldir)
3206 {
3207 return proc_pident_readdir(filp,dirent,filldir,
3208 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3209 }
3210
3211 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3212 return proc_pident_lookup(dir, dentry,
3213 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3214 }
3215
3216 static const struct file_operations proc_tid_base_operations = {
3217 .read = generic_read_dir,
3218 .readdir = proc_tid_base_readdir,
3219 .llseek = default_llseek,
3220 };
3221
3222 static const struct inode_operations proc_tid_base_inode_operations = {
3223 .lookup = proc_tid_base_lookup,
3224 .getattr = pid_getattr,
3225 .setattr = proc_setattr,
3226 };
3227
3228 static struct dentry *proc_task_instantiate(struct inode *dir,
3229 struct dentry *dentry, struct task_struct *task, const void *ptr)
3230 {
3231 struct dentry *error = ERR_PTR(-ENOENT);
3232 struct inode *inode;
3233 inode = proc_pid_make_inode(dir->i_sb, task);
3234
3235 if (!inode)
3236 goto out;
3237 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3238 inode->i_op = &proc_tid_base_inode_operations;
3239 inode->i_fop = &proc_tid_base_operations;
3240 inode->i_flags|=S_IMMUTABLE;
3241
3242 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
3243 ARRAY_SIZE(tid_base_stuff));
3244
3245 d_set_d_op(dentry, &pid_dentry_operations);
3246
3247 d_add(dentry, inode);
3248 /* Close the race of the process dying before we return the dentry */
3249 if (pid_revalidate(dentry, NULL))
3250 error = NULL;
3251 out:
3252 return error;
3253 }
3254
3255 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3256 {
3257 struct dentry *result = ERR_PTR(-ENOENT);
3258 struct task_struct *task;
3259 struct task_struct *leader = get_proc_task(dir);
3260 unsigned tid;
3261 struct pid_namespace *ns;
3262
3263 if (!leader)
3264 goto out_no_task;
3265
3266 tid = name_to_int(dentry);
3267 if (tid == ~0U)
3268 goto out;
3269
3270 ns = dentry->d_sb->s_fs_info;
3271 rcu_read_lock();
3272 task = find_task_by_pid_ns(tid, ns);
3273 if (task)
3274 get_task_struct(task);
3275 rcu_read_unlock();
3276 if (!task)
3277 goto out;
3278 if (!same_thread_group(leader, task))
3279 goto out_drop_task;
3280
3281 result = proc_task_instantiate(dir, dentry, task, NULL);
3282 out_drop_task:
3283 put_task_struct(task);
3284 out:
3285 put_task_struct(leader);
3286 out_no_task:
3287 return result;
3288 }
3289
3290 /*
3291 * Find the first tid of a thread group to return to user space.
3292 *
3293 * Usually this is just the thread group leader, but if the users
3294 * buffer was too small or there was a seek into the middle of the
3295 * directory we have more work todo.
3296 *
3297 * In the case of a short read we start with find_task_by_pid.
3298 *
3299 * In the case of a seek we start with the leader and walk nr
3300 * threads past it.
3301 */
3302 static struct task_struct *first_tid(struct task_struct *leader,
3303 int tid, int nr, struct pid_namespace *ns)
3304 {
3305 struct task_struct *pos;
3306
3307 rcu_read_lock();
3308 /* Attempt to start with the pid of a thread */
3309 if (tid && (nr > 0)) {
3310 pos = find_task_by_pid_ns(tid, ns);
3311 if (pos && (pos->group_leader == leader))
3312 goto found;
3313 }
3314
3315 /* If nr exceeds the number of threads there is nothing todo */
3316 pos = NULL;
3317 if (nr && nr >= get_nr_threads(leader))
3318 goto out;
3319
3320 /* If we haven't found our starting place yet start
3321 * with the leader and walk nr threads forward.
3322 */
3323 for (pos = leader; nr > 0; --nr) {
3324 pos = next_thread(pos);
3325 if (pos == leader) {
3326 pos = NULL;
3327 goto out;
3328 }
3329 }
3330 found:
3331 get_task_struct(pos);
3332 out:
3333 rcu_read_unlock();
3334 return pos;
3335 }
3336
3337 /*
3338 * Find the next thread in the thread list.
3339 * Return NULL if there is an error or no next thread.
3340 *
3341 * The reference to the input task_struct is released.
3342 */
3343 static struct task_struct *next_tid(struct task_struct *start)
3344 {
3345 struct task_struct *pos = NULL;
3346 rcu_read_lock();
3347 if (pid_alive(start)) {
3348 pos = next_thread(start);
3349 if (thread_group_leader(pos))
3350 pos = NULL;
3351 else
3352 get_task_struct(pos);
3353 }
3354 rcu_read_unlock();
3355 put_task_struct(start);
3356 return pos;
3357 }
3358
3359 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3360 struct task_struct *task, int tid)
3361 {
3362 char name[PROC_NUMBUF];
3363 int len = snprintf(name, sizeof(name), "%d", tid);
3364 return proc_fill_cache(filp, dirent, filldir, name, len,
3365 proc_task_instantiate, task, NULL);
3366 }
3367
3368 /* for the /proc/TGID/task/ directories */
3369 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3370 {
3371 struct dentry *dentry = filp->f_path.dentry;
3372 struct inode *inode = dentry->d_inode;
3373 struct task_struct *leader = NULL;
3374 struct task_struct *task;
3375 int retval = -ENOENT;
3376 ino_t ino;
3377 int tid;
3378 struct pid_namespace *ns;
3379
3380 task = get_proc_task(inode);
3381 if (!task)
3382 goto out_no_task;
3383 rcu_read_lock();
3384 if (pid_alive(task)) {
3385 leader = task->group_leader;
3386 get_task_struct(leader);
3387 }
3388 rcu_read_unlock();
3389 put_task_struct(task);
3390 if (!leader)
3391 goto out_no_task;
3392 retval = 0;
3393
3394 switch ((unsigned long)filp->f_pos) {
3395 case 0:
3396 ino = inode->i_ino;
3397 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3398 goto out;
3399 filp->f_pos++;
3400 /* fall through */
3401 case 1:
3402 ino = parent_ino(dentry);
3403 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3404 goto out;
3405 filp->f_pos++;
3406 /* fall through */
3407 }
3408
3409 /* f_version caches the tgid value that the last readdir call couldn't
3410 * return. lseek aka telldir automagically resets f_version to 0.
3411 */
3412 ns = filp->f_dentry->d_sb->s_fs_info;
3413 tid = (int)filp->f_version;
3414 filp->f_version = 0;
3415 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3416 task;
3417 task = next_tid(task), filp->f_pos++) {
3418 tid = task_pid_nr_ns(task, ns);
3419 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3420 /* returning this tgid failed, save it as the first
3421 * pid for the next readir call */
3422 filp->f_version = (u64)tid;
3423 put_task_struct(task);
3424 break;
3425 }
3426 }
3427 out:
3428 put_task_struct(leader);
3429 out_no_task:
3430 return retval;
3431 }
3432
3433 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3434 {
3435 struct inode *inode = dentry->d_inode;
3436 struct task_struct *p = get_proc_task(inode);
3437 generic_fillattr(inode, stat);
3438
3439 if (p) {
3440 stat->nlink += get_nr_threads(p);
3441 put_task_struct(p);
3442 }
3443
3444 return 0;
3445 }
3446
3447 static const struct inode_operations proc_task_inode_operations = {
3448 .lookup = proc_task_lookup,
3449 .getattr = proc_task_getattr,
3450 .setattr = proc_setattr,
3451 };
3452
3453 static const struct file_operations proc_task_operations = {
3454 .read = generic_read_dir,
3455 .readdir = proc_task_readdir,
3456 .llseek = default_llseek,
3457 };
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree