mm-only debug patch...
[mmotm.git] / fs / proc / base.c
blobdd5bed0f461afd3d164e99b4d97a66ee3e56efd6
1 /*
2 * linux/fs/proc/base.c
4 * Copyright (C) 1991, 1992 Linus Torvalds
6 * proc base directory handling functions
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.
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>
24 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
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.
33 * Changelog:
34 * 21-Feb-2005
35 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36 * Pud inclusion in the page table walking.
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.
43 * Simo Piiroinen <simo.piiroinen@nokia.com>:
44 * Smaps information related to shared, private, clean and dirty pages.
46 * Paul Mundt <paul.mundt@nokia.com>:
47 * Overall revision about smaps.
50 #include <asm/uaccess.h>
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/utrace.h>
63 #include <linux/seq_file.h>
64 #include <linux/namei.h>
65 #include <linux/mnt_namespace.h>
66 #include <linux/mm.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 "internal.h"
87 /* NOTE:
88 * Implementing inode permission operations in /proc is almost
89 * certainly an error. Permission checks need to happen during
90 * each system call not at open time. The reason is that most of
91 * what we wish to check for permissions in /proc varies at runtime.
93 * The classic example of a problem is opening file descriptors
94 * in /proc for a task before it execs a suid executable.
97 struct pid_entry {
98 char *name;
99 int len;
100 mode_t mode;
101 const struct inode_operations *iop;
102 const struct file_operations *fop;
103 union proc_op op;
106 #define NOD(NAME, MODE, IOP, FOP, OP) { \
107 .name = (NAME), \
108 .len = sizeof(NAME) - 1, \
109 .mode = MODE, \
110 .iop = IOP, \
111 .fop = FOP, \
112 .op = OP, \
115 #define DIR(NAME, MODE, iops, fops) \
116 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
117 #define LNK(NAME, get_link) \
118 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
119 &proc_pid_link_inode_operations, NULL, \
120 { .proc_get_link = get_link } )
121 #define REG(NAME, MODE, fops) \
122 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
123 #define INF(NAME, MODE, read) \
124 NOD(NAME, (S_IFREG|(MODE)), \
125 NULL, &proc_info_file_operations, \
126 { .proc_read = read } )
127 #define ONE(NAME, MODE, show) \
128 NOD(NAME, (S_IFREG|(MODE)), \
129 NULL, &proc_single_file_operations, \
130 { .proc_show = show } )
133 * Count the number of hardlinks for the pid_entry table, excluding the .
134 * and .. links.
136 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
137 unsigned int n)
139 unsigned int i;
140 unsigned int count;
142 count = 0;
143 for (i = 0; i < n; ++i) {
144 if (S_ISDIR(entries[i].mode))
145 ++count;
148 return count;
151 static int get_fs_path(struct task_struct *task, struct path *path, bool root)
153 struct fs_struct *fs;
154 int result = -ENOENT;
156 task_lock(task);
157 fs = task->fs;
158 if (fs) {
159 read_lock(&fs->lock);
160 *path = root ? fs->root : fs->pwd;
161 path_get(path);
162 read_unlock(&fs->lock);
163 result = 0;
165 task_unlock(task);
166 return result;
169 static int get_nr_threads(struct task_struct *tsk)
171 unsigned long flags;
172 int count = 0;
174 if (lock_task_sighand(tsk, &flags)) {
175 count = atomic_read(&tsk->signal->count);
176 unlock_task_sighand(tsk, &flags);
178 return count;
181 static int proc_cwd_link(struct inode *inode, struct path *path)
183 struct task_struct *task = get_proc_task(inode);
184 int result = -ENOENT;
186 if (task) {
187 result = get_fs_path(task, path, 0);
188 put_task_struct(task);
190 return result;
193 static int proc_root_link(struct inode *inode, struct path *path)
195 struct task_struct *task = get_proc_task(inode);
196 int result = -ENOENT;
198 if (task) {
199 result = get_fs_path(task, path, 1);
200 put_task_struct(task);
202 return result;
206 * Return zero if current may access user memory in @task, -error if not.
208 static int check_mem_permission(struct task_struct *task)
211 * A task can always look at itself, in case it chooses
212 * to use system calls instead of load instructions.
214 if (task == current)
215 return 0;
218 * If current is actively ptrace'ing, and would also be
219 * permitted to freshly attach with ptrace now, permit it.
221 if (task_is_stopped_or_traced(task)) {
222 int match;
223 rcu_read_lock();
224 match = (tracehook_tracer_task(task) == current);
225 rcu_read_unlock();
226 if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH))
227 return 0;
231 * Noone else is allowed.
233 return -EPERM;
236 struct mm_struct *mm_for_maps(struct task_struct *task)
238 struct mm_struct *mm;
240 if (mutex_lock_killable(&task->cred_guard_mutex))
241 return NULL;
243 mm = get_task_mm(task);
244 if (mm && mm != current->mm &&
245 !ptrace_may_access(task, PTRACE_MODE_READ)) {
246 mmput(mm);
247 mm = NULL;
249 mutex_unlock(&task->cred_guard_mutex);
251 return mm;
254 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
256 int res = 0;
257 unsigned int len;
258 struct mm_struct *mm = get_task_mm(task);
259 if (!mm)
260 goto out;
261 if (!mm->arg_end)
262 goto out_mm; /* Shh! No looking before we're done */
264 len = mm->arg_end - mm->arg_start;
266 if (len > PAGE_SIZE)
267 len = PAGE_SIZE;
269 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
271 // If the nul at the end of args has been overwritten, then
272 // assume application is using setproctitle(3).
273 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
274 len = strnlen(buffer, res);
275 if (len < res) {
276 res = len;
277 } else {
278 len = mm->env_end - mm->env_start;
279 if (len > PAGE_SIZE - res)
280 len = PAGE_SIZE - res;
281 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
282 res = strnlen(buffer, res);
285 out_mm:
286 mmput(mm);
287 out:
288 return res;
291 static int proc_pid_auxv(struct task_struct *task, char *buffer)
293 int res = 0;
294 struct mm_struct *mm = get_task_mm(task);
295 if (mm) {
296 unsigned int nwords = 0;
297 do {
298 nwords += 2;
299 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
300 res = nwords * sizeof(mm->saved_auxv[0]);
301 if (res > PAGE_SIZE)
302 res = PAGE_SIZE;
303 memcpy(buffer, mm->saved_auxv, res);
304 mmput(mm);
306 return res;
310 #ifdef CONFIG_KALLSYMS
312 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
313 * Returns the resolved symbol. If that fails, simply return the address.
315 static int proc_pid_wchan(struct task_struct *task, char *buffer)
317 unsigned long wchan;
318 char symname[KSYM_NAME_LEN];
320 wchan = get_wchan(task);
322 if (lookup_symbol_name(wchan, symname) < 0)
323 if (!ptrace_may_access(task, PTRACE_MODE_READ))
324 return 0;
325 else
326 return sprintf(buffer, "%lu", wchan);
327 else
328 return sprintf(buffer, "%s", symname);
330 #endif /* CONFIG_KALLSYMS */
332 #ifdef CONFIG_STACKTRACE
334 #define MAX_STACK_TRACE_DEPTH 64
336 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
337 struct pid *pid, struct task_struct *task)
339 struct stack_trace trace;
340 unsigned long *entries;
341 int i;
343 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
344 if (!entries)
345 return -ENOMEM;
347 trace.nr_entries = 0;
348 trace.max_entries = MAX_STACK_TRACE_DEPTH;
349 trace.entries = entries;
350 trace.skip = 0;
351 save_stack_trace_tsk(task, &trace);
353 for (i = 0; i < trace.nr_entries; i++) {
354 seq_printf(m, "[<%p>] %pS\n",
355 (void *)entries[i], (void *)entries[i]);
357 kfree(entries);
359 return 0;
361 #endif
363 #ifdef CONFIG_SCHEDSTATS
365 * Provides /proc/PID/schedstat
367 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
369 return sprintf(buffer, "%llu %llu %lu\n",
370 (unsigned long long)task->se.sum_exec_runtime,
371 (unsigned long long)task->sched_info.run_delay,
372 task->sched_info.pcount);
374 #endif
376 #ifdef CONFIG_LATENCYTOP
377 static int lstats_show_proc(struct seq_file *m, void *v)
379 int i;
380 struct inode *inode = m->private;
381 struct task_struct *task = get_proc_task(inode);
383 if (!task)
384 return -ESRCH;
385 seq_puts(m, "Latency Top version : v0.1\n");
386 for (i = 0; i < 32; i++) {
387 if (task->latency_record[i].backtrace[0]) {
388 int q;
389 seq_printf(m, "%i %li %li ",
390 task->latency_record[i].count,
391 task->latency_record[i].time,
392 task->latency_record[i].max);
393 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
394 char sym[KSYM_SYMBOL_LEN];
395 char *c;
396 if (!task->latency_record[i].backtrace[q])
397 break;
398 if (task->latency_record[i].backtrace[q] == ULONG_MAX)
399 break;
400 sprint_symbol(sym, task->latency_record[i].backtrace[q]);
401 c = strchr(sym, '+');
402 if (c)
403 *c = 0;
404 seq_printf(m, "%s ", sym);
406 seq_printf(m, "\n");
410 put_task_struct(task);
411 return 0;
414 static int lstats_open(struct inode *inode, struct file *file)
416 return single_open(file, lstats_show_proc, inode);
419 static ssize_t lstats_write(struct file *file, const char __user *buf,
420 size_t count, loff_t *offs)
422 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
424 if (!task)
425 return -ESRCH;
426 clear_all_latency_tracing(task);
427 put_task_struct(task);
429 return count;
432 static const struct file_operations proc_lstats_operations = {
433 .open = lstats_open,
434 .read = seq_read,
435 .write = lstats_write,
436 .llseek = seq_lseek,
437 .release = single_release,
440 #endif
442 /* The badness from the OOM killer */
443 unsigned long badness(struct task_struct *p, unsigned long uptime);
444 static int proc_oom_score(struct task_struct *task, char *buffer)
446 unsigned long points;
447 struct timespec uptime;
449 do_posix_clock_monotonic_gettime(&uptime);
450 read_lock(&tasklist_lock);
451 points = badness(task->group_leader, uptime.tv_sec);
452 read_unlock(&tasklist_lock);
453 return sprintf(buffer, "%lu\n", points);
456 struct limit_names {
457 char *name;
458 char *unit;
461 static const struct limit_names lnames[RLIM_NLIMITS] = {
462 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
463 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
464 [RLIMIT_DATA] = {"Max data size", "bytes"},
465 [RLIMIT_STACK] = {"Max stack size", "bytes"},
466 [RLIMIT_CORE] = {"Max core file size", "bytes"},
467 [RLIMIT_RSS] = {"Max resident set", "bytes"},
468 [RLIMIT_NPROC] = {"Max processes", "processes"},
469 [RLIMIT_NOFILE] = {"Max open files", "files"},
470 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
471 [RLIMIT_AS] = {"Max address space", "bytes"},
472 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
473 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
474 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
475 [RLIMIT_NICE] = {"Max nice priority", NULL},
476 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
477 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
480 /* Display limits for a process */
481 static int proc_pid_limits(struct task_struct *task, char *buffer)
483 unsigned int i;
484 int count = 0;
485 unsigned long flags;
486 char *bufptr = buffer;
488 struct rlimit rlim[RLIM_NLIMITS];
490 if (!lock_task_sighand(task, &flags))
491 return 0;
492 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
493 unlock_task_sighand(task, &flags);
496 * print the file header
498 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
499 "Limit", "Soft Limit", "Hard Limit", "Units");
501 for (i = 0; i < RLIM_NLIMITS; i++) {
502 if (rlim[i].rlim_cur == RLIM_INFINITY)
503 count += sprintf(&bufptr[count], "%-25s %-20s ",
504 lnames[i].name, "unlimited");
505 else
506 count += sprintf(&bufptr[count], "%-25s %-20lu ",
507 lnames[i].name, rlim[i].rlim_cur);
509 if (rlim[i].rlim_max == RLIM_INFINITY)
510 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
511 else
512 count += sprintf(&bufptr[count], "%-20lu ",
513 rlim[i].rlim_max);
515 if (lnames[i].unit)
516 count += sprintf(&bufptr[count], "%-10s\n",
517 lnames[i].unit);
518 else
519 count += sprintf(&bufptr[count], "\n");
522 return count;
525 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
526 static int proc_pid_syscall(struct task_struct *task, char *buffer)
528 long nr;
529 unsigned long args[6], sp, pc;
531 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
532 return sprintf(buffer, "running\n");
534 if (nr < 0)
535 return sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
537 return sprintf(buffer,
538 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
540 args[0], args[1], args[2], args[3], args[4], args[5],
541 sp, pc);
543 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
545 /************************************************************************/
546 /* Here the fs part begins */
547 /************************************************************************/
549 /* permission checks */
550 static int proc_fd_access_allowed(struct inode *inode)
552 struct task_struct *task;
553 int allowed = 0;
554 /* Allow access to a task's file descriptors if it is us or we
555 * may use ptrace attach to the process and find out that
556 * information.
558 task = get_proc_task(inode);
559 if (task) {
560 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
561 put_task_struct(task);
563 return allowed;
566 static int proc_setattr(struct dentry *dentry, struct iattr *attr)
568 int error;
569 struct inode *inode = dentry->d_inode;
571 if (attr->ia_valid & ATTR_MODE)
572 return -EPERM;
574 error = inode_change_ok(inode, attr);
575 if (!error)
576 error = inode_setattr(inode, attr);
577 return error;
580 static const struct inode_operations proc_def_inode_operations = {
581 .setattr = proc_setattr,
584 static int mounts_open_common(struct inode *inode, struct file *file,
585 const struct seq_operations *op)
587 struct task_struct *task = get_proc_task(inode);
588 struct nsproxy *nsp;
589 struct mnt_namespace *ns = NULL;
590 struct path root;
591 struct proc_mounts *p;
592 int ret = -EINVAL;
594 if (task) {
595 rcu_read_lock();
596 nsp = task_nsproxy(task);
597 if (nsp) {
598 ns = nsp->mnt_ns;
599 if (ns)
600 get_mnt_ns(ns);
602 rcu_read_unlock();
603 if (ns && get_fs_path(task, &root, 1) == 0)
604 ret = 0;
605 put_task_struct(task);
608 if (!ns)
609 goto err;
610 if (ret)
611 goto err_put_ns;
613 ret = -ENOMEM;
614 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
615 if (!p)
616 goto err_put_path;
618 file->private_data = &p->m;
619 ret = seq_open(file, op);
620 if (ret)
621 goto err_free;
623 p->m.private = p;
624 p->ns = ns;
625 p->root = root;
626 p->event = ns->event;
628 return 0;
630 err_free:
631 kfree(p);
632 err_put_path:
633 path_put(&root);
634 err_put_ns:
635 put_mnt_ns(ns);
636 err:
637 return ret;
640 static int mounts_release(struct inode *inode, struct file *file)
642 struct proc_mounts *p = file->private_data;
643 path_put(&p->root);
644 put_mnt_ns(p->ns);
645 return seq_release(inode, file);
648 static unsigned mounts_poll(struct file *file, poll_table *wait)
650 struct proc_mounts *p = file->private_data;
651 struct mnt_namespace *ns = p->ns;
652 unsigned res = POLLIN | POLLRDNORM;
654 poll_wait(file, &ns->poll, wait);
656 spin_lock(&vfsmount_lock);
657 if (p->event != ns->event) {
658 p->event = ns->event;
659 res |= POLLERR | POLLPRI;
661 spin_unlock(&vfsmount_lock);
663 return res;
666 static int mounts_open(struct inode *inode, struct file *file)
668 return mounts_open_common(inode, file, &mounts_op);
671 static const struct file_operations proc_mounts_operations = {
672 .open = mounts_open,
673 .read = seq_read,
674 .llseek = seq_lseek,
675 .release = mounts_release,
676 .poll = mounts_poll,
679 static int mountinfo_open(struct inode *inode, struct file *file)
681 return mounts_open_common(inode, file, &mountinfo_op);
684 static const struct file_operations proc_mountinfo_operations = {
685 .open = mountinfo_open,
686 .read = seq_read,
687 .llseek = seq_lseek,
688 .release = mounts_release,
689 .poll = mounts_poll,
692 static int mountstats_open(struct inode *inode, struct file *file)
694 return mounts_open_common(inode, file, &mountstats_op);
697 static const struct file_operations proc_mountstats_operations = {
698 .open = mountstats_open,
699 .read = seq_read,
700 .llseek = seq_lseek,
701 .release = mounts_release,
704 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
706 static ssize_t proc_info_read(struct file * file, char __user * buf,
707 size_t count, loff_t *ppos)
709 struct inode * inode = file->f_path.dentry->d_inode;
710 unsigned long page;
711 ssize_t length;
712 struct task_struct *task = get_proc_task(inode);
714 length = -ESRCH;
715 if (!task)
716 goto out_no_task;
718 if (count > PROC_BLOCK_SIZE)
719 count = PROC_BLOCK_SIZE;
721 length = -ENOMEM;
722 if (!(page = __get_free_page(GFP_TEMPORARY)))
723 goto out;
725 length = PROC_I(inode)->op.proc_read(task, (char*)page);
727 if (length >= 0)
728 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
729 free_page(page);
730 out:
731 put_task_struct(task);
732 out_no_task:
733 return length;
736 static const struct file_operations proc_info_file_operations = {
737 .read = proc_info_read,
740 static int proc_single_show(struct seq_file *m, void *v)
742 struct inode *inode = m->private;
743 struct pid_namespace *ns;
744 struct pid *pid;
745 struct task_struct *task;
746 int ret;
748 ns = inode->i_sb->s_fs_info;
749 pid = proc_pid(inode);
750 task = get_pid_task(pid, PIDTYPE_PID);
751 if (!task)
752 return -ESRCH;
754 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
756 put_task_struct(task);
757 return ret;
760 static int proc_single_open(struct inode *inode, struct file *filp)
762 int ret;
763 ret = single_open(filp, proc_single_show, NULL);
764 if (!ret) {
765 struct seq_file *m = filp->private_data;
767 m->private = inode;
769 return ret;
772 static const struct file_operations proc_single_file_operations = {
773 .open = proc_single_open,
774 .read = seq_read,
775 .llseek = seq_lseek,
776 .release = single_release,
779 static int mem_open(struct inode* inode, struct file* file)
781 file->private_data = (void*)((long)current->self_exec_id);
782 return 0;
785 static ssize_t mem_read(struct file * file, char __user * buf,
786 size_t count, loff_t *ppos)
788 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
789 char *page;
790 unsigned long src = *ppos;
791 int ret = -ESRCH;
792 struct mm_struct *mm;
794 if (!task)
795 goto out_no_task;
797 if (check_mem_permission(task))
798 goto out;
800 ret = -ENOMEM;
801 page = (char *)__get_free_page(GFP_TEMPORARY);
802 if (!page)
803 goto out;
805 ret = 0;
807 mm = get_task_mm(task);
808 if (!mm)
809 goto out_free;
811 ret = -EIO;
813 if (file->private_data != (void*)((long)current->self_exec_id))
814 goto out_put;
816 ret = 0;
818 while (count > 0) {
819 int this_len, retval;
821 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
822 retval = access_process_vm(task, src, page, this_len, 0);
823 if (!retval || check_mem_permission(task)) {
824 if (!ret)
825 ret = -EIO;
826 break;
829 if (copy_to_user(buf, page, retval)) {
830 ret = -EFAULT;
831 break;
834 ret += retval;
835 src += retval;
836 buf += retval;
837 count -= retval;
839 *ppos = src;
841 out_put:
842 mmput(mm);
843 out_free:
844 free_page((unsigned long) page);
845 out:
846 put_task_struct(task);
847 out_no_task:
848 return ret;
851 #define mem_write NULL
853 #ifndef mem_write
854 /* This is a security hazard */
855 static ssize_t mem_write(struct file * file, const char __user *buf,
856 size_t count, loff_t *ppos)
858 int copied;
859 char *page;
860 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
861 unsigned long dst = *ppos;
863 copied = -ESRCH;
864 if (!task)
865 goto out_no_task;
867 if (check_mem_permission(task))
868 goto out;
870 copied = -ENOMEM;
871 page = (char *)__get_free_page(GFP_TEMPORARY);
872 if (!page)
873 goto out;
875 copied = 0;
876 while (count > 0) {
877 int this_len, retval;
879 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
880 if (copy_from_user(page, buf, this_len)) {
881 copied = -EFAULT;
882 break;
884 retval = access_process_vm(task, dst, page, this_len, 1);
885 if (!retval) {
886 if (!copied)
887 copied = -EIO;
888 break;
890 copied += retval;
891 buf += retval;
892 dst += retval;
893 count -= retval;
895 *ppos = dst;
896 free_page((unsigned long) page);
897 out:
898 put_task_struct(task);
899 out_no_task:
900 return copied;
902 #endif
904 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
906 switch (orig) {
907 case 0:
908 file->f_pos = offset;
909 break;
910 case 1:
911 file->f_pos += offset;
912 break;
913 default:
914 return -EINVAL;
916 force_successful_syscall_return();
917 return file->f_pos;
920 static const struct file_operations proc_mem_operations = {
921 .llseek = mem_lseek,
922 .read = mem_read,
923 .write = mem_write,
924 .open = mem_open,
927 static ssize_t environ_read(struct file *file, char __user *buf,
928 size_t count, loff_t *ppos)
930 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
931 char *page;
932 unsigned long src = *ppos;
933 int ret = -ESRCH;
934 struct mm_struct *mm;
936 if (!task)
937 goto out_no_task;
939 if (!ptrace_may_access(task, PTRACE_MODE_READ))
940 goto out;
942 ret = -ENOMEM;
943 page = (char *)__get_free_page(GFP_TEMPORARY);
944 if (!page)
945 goto out;
947 ret = 0;
949 mm = get_task_mm(task);
950 if (!mm)
951 goto out_free;
953 while (count > 0) {
954 int this_len, retval, max_len;
956 this_len = mm->env_end - (mm->env_start + src);
958 if (this_len <= 0)
959 break;
961 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
962 this_len = (this_len > max_len) ? max_len : this_len;
964 retval = access_process_vm(task, (mm->env_start + src),
965 page, this_len, 0);
967 if (retval <= 0) {
968 ret = retval;
969 break;
972 if (copy_to_user(buf, page, retval)) {
973 ret = -EFAULT;
974 break;
977 ret += retval;
978 src += retval;
979 buf += retval;
980 count -= retval;
982 *ppos = src;
984 mmput(mm);
985 out_free:
986 free_page((unsigned long) page);
987 out:
988 put_task_struct(task);
989 out_no_task:
990 return ret;
993 static const struct file_operations proc_environ_operations = {
994 .read = environ_read,
997 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
998 size_t count, loff_t *ppos)
1000 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1001 char buffer[PROC_NUMBUF];
1002 size_t len;
1003 int oom_adjust = OOM_DISABLE;
1004 unsigned long flags;
1006 if (!task)
1007 return -ESRCH;
1009 if (lock_task_sighand(task, &flags)) {
1010 oom_adjust = task->signal->oom_adj;
1011 unlock_task_sighand(task, &flags);
1014 put_task_struct(task);
1016 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
1018 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1021 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1022 size_t count, loff_t *ppos)
1024 struct task_struct *task;
1025 char buffer[PROC_NUMBUF];
1026 long oom_adjust;
1027 unsigned long flags;
1028 int err;
1030 memset(buffer, 0, sizeof(buffer));
1031 if (count > sizeof(buffer) - 1)
1032 count = sizeof(buffer) - 1;
1033 if (copy_from_user(buffer, buf, count))
1034 return -EFAULT;
1036 err = strict_strtol(strstrip(buffer), 0, &oom_adjust);
1037 if (err)
1038 return -EINVAL;
1039 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
1040 oom_adjust != OOM_DISABLE)
1041 return -EINVAL;
1043 task = get_proc_task(file->f_path.dentry->d_inode);
1044 if (!task)
1045 return -ESRCH;
1046 if (!lock_task_sighand(task, &flags)) {
1047 put_task_struct(task);
1048 return -ESRCH;
1051 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
1052 unlock_task_sighand(task, &flags);
1053 put_task_struct(task);
1054 return -EACCES;
1057 task->signal->oom_adj = oom_adjust;
1059 unlock_task_sighand(task, &flags);
1060 put_task_struct(task);
1062 return count;
1065 static const struct file_operations proc_oom_adjust_operations = {
1066 .read = oom_adjust_read,
1067 .write = oom_adjust_write,
1070 #ifdef CONFIG_AUDITSYSCALL
1071 #define TMPBUFLEN 21
1072 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1073 size_t count, loff_t *ppos)
1075 struct inode * inode = file->f_path.dentry->d_inode;
1076 struct task_struct *task = get_proc_task(inode);
1077 ssize_t length;
1078 char tmpbuf[TMPBUFLEN];
1080 if (!task)
1081 return -ESRCH;
1082 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1083 audit_get_loginuid(task));
1084 put_task_struct(task);
1085 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1088 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1089 size_t count, loff_t *ppos)
1091 struct inode * inode = file->f_path.dentry->d_inode;
1092 char *page, *tmp;
1093 ssize_t length;
1094 uid_t loginuid;
1096 if (!capable(CAP_AUDIT_CONTROL))
1097 return -EPERM;
1099 if (current != pid_task(proc_pid(inode), PIDTYPE_PID))
1100 return -EPERM;
1102 if (count >= PAGE_SIZE)
1103 count = PAGE_SIZE - 1;
1105 if (*ppos != 0) {
1106 /* No partial writes. */
1107 return -EINVAL;
1109 page = (char*)__get_free_page(GFP_TEMPORARY);
1110 if (!page)
1111 return -ENOMEM;
1112 length = -EFAULT;
1113 if (copy_from_user(page, buf, count))
1114 goto out_free_page;
1116 page[count] = '\0';
1117 loginuid = simple_strtoul(page, &tmp, 10);
1118 if (tmp == page) {
1119 length = -EINVAL;
1120 goto out_free_page;
1123 length = audit_set_loginuid(current, loginuid);
1124 if (likely(length == 0))
1125 length = count;
1127 out_free_page:
1128 free_page((unsigned long) page);
1129 return length;
1132 static const struct file_operations proc_loginuid_operations = {
1133 .read = proc_loginuid_read,
1134 .write = proc_loginuid_write,
1137 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1138 size_t count, loff_t *ppos)
1140 struct inode * inode = file->f_path.dentry->d_inode;
1141 struct task_struct *task = get_proc_task(inode);
1142 ssize_t length;
1143 char tmpbuf[TMPBUFLEN];
1145 if (!task)
1146 return -ESRCH;
1147 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1148 audit_get_sessionid(task));
1149 put_task_struct(task);
1150 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1153 static const struct file_operations proc_sessionid_operations = {
1154 .read = proc_sessionid_read,
1156 #endif
1158 #ifdef CONFIG_FAULT_INJECTION
1159 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1160 size_t count, loff_t *ppos)
1162 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1163 char buffer[PROC_NUMBUF];
1164 size_t len;
1165 int make_it_fail;
1167 if (!task)
1168 return -ESRCH;
1169 make_it_fail = task->make_it_fail;
1170 put_task_struct(task);
1172 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1174 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1177 static ssize_t proc_fault_inject_write(struct file * file,
1178 const char __user * buf, size_t count, loff_t *ppos)
1180 struct task_struct *task;
1181 char buffer[PROC_NUMBUF], *end;
1182 int make_it_fail;
1184 if (!capable(CAP_SYS_RESOURCE))
1185 return -EPERM;
1186 memset(buffer, 0, sizeof(buffer));
1187 if (count > sizeof(buffer) - 1)
1188 count = sizeof(buffer) - 1;
1189 if (copy_from_user(buffer, buf, count))
1190 return -EFAULT;
1191 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1192 if (*end)
1193 return -EINVAL;
1194 task = get_proc_task(file->f_dentry->d_inode);
1195 if (!task)
1196 return -ESRCH;
1197 task->make_it_fail = make_it_fail;
1198 put_task_struct(task);
1200 return count;
1203 static const struct file_operations proc_fault_inject_operations = {
1204 .read = proc_fault_inject_read,
1205 .write = proc_fault_inject_write,
1207 #endif
1210 #ifdef CONFIG_SCHED_DEBUG
1212 * Print out various scheduling related per-task fields:
1214 static int sched_show(struct seq_file *m, void *v)
1216 struct inode *inode = m->private;
1217 struct task_struct *p;
1219 p = get_proc_task(inode);
1220 if (!p)
1221 return -ESRCH;
1222 proc_sched_show_task(p, m);
1224 put_task_struct(p);
1226 return 0;
1229 static ssize_t
1230 sched_write(struct file *file, const char __user *buf,
1231 size_t count, loff_t *offset)
1233 struct inode *inode = file->f_path.dentry->d_inode;
1234 struct task_struct *p;
1236 p = get_proc_task(inode);
1237 if (!p)
1238 return -ESRCH;
1239 proc_sched_set_task(p);
1241 put_task_struct(p);
1243 return count;
1246 static int sched_open(struct inode *inode, struct file *filp)
1248 int ret;
1250 ret = single_open(filp, sched_show, NULL);
1251 if (!ret) {
1252 struct seq_file *m = filp->private_data;
1254 m->private = inode;
1256 return ret;
1259 static const struct file_operations proc_pid_sched_operations = {
1260 .open = sched_open,
1261 .read = seq_read,
1262 .write = sched_write,
1263 .llseek = seq_lseek,
1264 .release = single_release,
1267 #endif
1270 * We added or removed a vma mapping the executable. The vmas are only mapped
1271 * during exec and are not mapped with the mmap system call.
1272 * Callers must hold down_write() on the mm's mmap_sem for these
1274 void added_exe_file_vma(struct mm_struct *mm)
1276 mm->num_exe_file_vmas++;
1279 void removed_exe_file_vma(struct mm_struct *mm)
1281 mm->num_exe_file_vmas--;
1282 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
1283 fput(mm->exe_file);
1284 mm->exe_file = NULL;
1289 void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1291 if (new_exe_file)
1292 get_file(new_exe_file);
1293 if (mm->exe_file)
1294 fput(mm->exe_file);
1295 mm->exe_file = new_exe_file;
1296 mm->num_exe_file_vmas = 0;
1299 struct file *get_mm_exe_file(struct mm_struct *mm)
1301 struct file *exe_file;
1303 /* We need mmap_sem to protect against races with removal of
1304 * VM_EXECUTABLE vmas */
1305 down_read(&mm->mmap_sem);
1306 exe_file = mm->exe_file;
1307 if (exe_file)
1308 get_file(exe_file);
1309 up_read(&mm->mmap_sem);
1310 return exe_file;
1313 void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
1315 /* It's safe to write the exe_file pointer without exe_file_lock because
1316 * this is called during fork when the task is not yet in /proc */
1317 newmm->exe_file = get_mm_exe_file(oldmm);
1320 static int proc_exe_link(struct inode *inode, struct path *exe_path)
1322 struct task_struct *task;
1323 struct mm_struct *mm;
1324 struct file *exe_file;
1326 task = get_proc_task(inode);
1327 if (!task)
1328 return -ENOENT;
1329 mm = get_task_mm(task);
1330 put_task_struct(task);
1331 if (!mm)
1332 return -ENOENT;
1333 exe_file = get_mm_exe_file(mm);
1334 mmput(mm);
1335 if (exe_file) {
1336 *exe_path = exe_file->f_path;
1337 path_get(&exe_file->f_path);
1338 fput(exe_file);
1339 return 0;
1340 } else
1341 return -ENOENT;
1344 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1346 struct inode *inode = dentry->d_inode;
1347 int error = -EACCES;
1349 /* We don't need a base pointer in the /proc filesystem */
1350 path_put(&nd->path);
1352 /* Are we allowed to snoop on the tasks file descriptors? */
1353 if (!proc_fd_access_allowed(inode))
1354 goto out;
1356 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1357 nd->last_type = LAST_BIND;
1358 out:
1359 return ERR_PTR(error);
1362 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1364 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1365 char *pathname;
1366 int len;
1368 if (!tmp)
1369 return -ENOMEM;
1371 pathname = d_path(path, tmp, PAGE_SIZE);
1372 len = PTR_ERR(pathname);
1373 if (IS_ERR(pathname))
1374 goto out;
1375 len = tmp + PAGE_SIZE - 1 - pathname;
1377 if (len > buflen)
1378 len = buflen;
1379 if (copy_to_user(buffer, pathname, len))
1380 len = -EFAULT;
1381 out:
1382 free_page((unsigned long)tmp);
1383 return len;
1386 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1388 int error = -EACCES;
1389 struct inode *inode = dentry->d_inode;
1390 struct path path;
1392 /* Are we allowed to snoop on the tasks file descriptors? */
1393 if (!proc_fd_access_allowed(inode))
1394 goto out;
1396 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1397 if (error)
1398 goto out;
1400 error = do_proc_readlink(&path, buffer, buflen);
1401 path_put(&path);
1402 out:
1403 return error;
1406 static const struct inode_operations proc_pid_link_inode_operations = {
1407 .readlink = proc_pid_readlink,
1408 .follow_link = proc_pid_follow_link,
1409 .setattr = proc_setattr,
1413 /* building an inode */
1415 static int task_dumpable(struct task_struct *task)
1417 int dumpable = 0;
1418 struct mm_struct *mm;
1420 task_lock(task);
1421 mm = task->mm;
1422 if (mm)
1423 dumpable = get_dumpable(mm);
1424 task_unlock(task);
1425 if(dumpable == 1)
1426 return 1;
1427 return 0;
1431 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1433 struct inode * inode;
1434 struct proc_inode *ei;
1435 const struct cred *cred;
1437 /* We need a new inode */
1439 inode = new_inode(sb);
1440 if (!inode)
1441 goto out;
1443 /* Common stuff */
1444 ei = PROC_I(inode);
1445 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1446 inode->i_op = &proc_def_inode_operations;
1449 * grab the reference to task.
1451 ei->pid = get_task_pid(task, PIDTYPE_PID);
1452 if (!ei->pid)
1453 goto out_unlock;
1455 if (task_dumpable(task)) {
1456 rcu_read_lock();
1457 cred = __task_cred(task);
1458 inode->i_uid = cred->euid;
1459 inode->i_gid = cred->egid;
1460 rcu_read_unlock();
1462 security_task_to_inode(task, inode);
1464 out:
1465 return inode;
1467 out_unlock:
1468 iput(inode);
1469 return NULL;
1472 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1474 struct inode *inode = dentry->d_inode;
1475 struct task_struct *task;
1476 const struct cred *cred;
1478 generic_fillattr(inode, stat);
1480 rcu_read_lock();
1481 stat->uid = 0;
1482 stat->gid = 0;
1483 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1484 if (task) {
1485 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1486 task_dumpable(task)) {
1487 cred = __task_cred(task);
1488 stat->uid = cred->euid;
1489 stat->gid = cred->egid;
1492 rcu_read_unlock();
1493 return 0;
1496 /* dentry stuff */
1499 * Exceptional case: normally we are not allowed to unhash a busy
1500 * directory. In this case, however, we can do it - no aliasing problems
1501 * due to the way we treat inodes.
1503 * Rewrite the inode's ownerships here because the owning task may have
1504 * performed a setuid(), etc.
1506 * Before the /proc/pid/status file was created the only way to read
1507 * the effective uid of a /process was to stat /proc/pid. Reading
1508 * /proc/pid/status is slow enough that procps and other packages
1509 * kept stating /proc/pid. To keep the rules in /proc simple I have
1510 * made this apply to all per process world readable and executable
1511 * directories.
1513 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1515 struct inode *inode = dentry->d_inode;
1516 struct task_struct *task = get_proc_task(inode);
1517 const struct cred *cred;
1519 if (task) {
1520 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1521 task_dumpable(task)) {
1522 rcu_read_lock();
1523 cred = __task_cred(task);
1524 inode->i_uid = cred->euid;
1525 inode->i_gid = cred->egid;
1526 rcu_read_unlock();
1527 } else {
1528 inode->i_uid = 0;
1529 inode->i_gid = 0;
1531 inode->i_mode &= ~(S_ISUID | S_ISGID);
1532 security_task_to_inode(task, inode);
1533 put_task_struct(task);
1534 return 1;
1536 d_drop(dentry);
1537 return 0;
1540 static int pid_delete_dentry(struct dentry * dentry)
1542 /* Is the task we represent dead?
1543 * If so, then don't put the dentry on the lru list,
1544 * kill it immediately.
1546 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1549 static const struct dentry_operations pid_dentry_operations =
1551 .d_revalidate = pid_revalidate,
1552 .d_delete = pid_delete_dentry,
1555 /* Lookups */
1557 typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1558 struct task_struct *, const void *);
1561 * Fill a directory entry.
1563 * If possible create the dcache entry and derive our inode number and
1564 * file type from dcache entry.
1566 * Since all of the proc inode numbers are dynamically generated, the inode
1567 * numbers do not exist until the inode is cache. This means creating the
1568 * the dcache entry in readdir is necessary to keep the inode numbers
1569 * reported by readdir in sync with the inode numbers reported
1570 * by stat.
1572 static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1573 char *name, int len,
1574 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1576 struct dentry *child, *dir = filp->f_path.dentry;
1577 struct inode *inode;
1578 struct qstr qname;
1579 ino_t ino = 0;
1580 unsigned type = DT_UNKNOWN;
1582 qname.name = name;
1583 qname.len = len;
1584 qname.hash = full_name_hash(name, len);
1586 child = d_lookup(dir, &qname);
1587 if (!child) {
1588 struct dentry *new;
1589 new = d_alloc(dir, &qname);
1590 if (new) {
1591 child = instantiate(dir->d_inode, new, task, ptr);
1592 if (child)
1593 dput(new);
1594 else
1595 child = new;
1598 if (!child || IS_ERR(child) || !child->d_inode)
1599 goto end_instantiate;
1600 inode = child->d_inode;
1601 if (inode) {
1602 ino = inode->i_ino;
1603 type = inode->i_mode >> 12;
1605 dput(child);
1606 end_instantiate:
1607 if (!ino)
1608 ino = find_inode_number(dir, &qname);
1609 if (!ino)
1610 ino = 1;
1611 return filldir(dirent, name, len, filp->f_pos, ino, type);
1614 static unsigned name_to_int(struct dentry *dentry)
1616 const char *name = dentry->d_name.name;
1617 int len = dentry->d_name.len;
1618 unsigned n = 0;
1620 if (len > 1 && *name == '0')
1621 goto out;
1622 while (len-- > 0) {
1623 unsigned c = *name++ - '0';
1624 if (c > 9)
1625 goto out;
1626 if (n >= (~0U-9)/10)
1627 goto out;
1628 n *= 10;
1629 n += c;
1631 return n;
1632 out:
1633 return ~0U;
1636 #define PROC_FDINFO_MAX 64
1638 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1640 struct task_struct *task = get_proc_task(inode);
1641 struct files_struct *files = NULL;
1642 struct file *file;
1643 int fd = proc_fd(inode);
1645 if (task) {
1646 files = get_files_struct(task);
1647 put_task_struct(task);
1649 if (files) {
1651 * We are not taking a ref to the file structure, so we must
1652 * hold ->file_lock.
1654 spin_lock(&files->file_lock);
1655 file = fcheck_files(files, fd);
1656 if (file) {
1657 if (path) {
1658 *path = file->f_path;
1659 path_get(&file->f_path);
1661 if (info)
1662 snprintf(info, PROC_FDINFO_MAX,
1663 "pos:\t%lli\n"
1664 "flags:\t0%o\n",
1665 (long long) file->f_pos,
1666 file->f_flags);
1667 spin_unlock(&files->file_lock);
1668 put_files_struct(files);
1669 return 0;
1671 spin_unlock(&files->file_lock);
1672 put_files_struct(files);
1674 return -ENOENT;
1677 static int proc_fd_link(struct inode *inode, struct path *path)
1679 return proc_fd_info(inode, path, NULL);
1682 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1684 struct inode *inode = dentry->d_inode;
1685 struct task_struct *task = get_proc_task(inode);
1686 int fd = proc_fd(inode);
1687 struct files_struct *files;
1688 const struct cred *cred;
1690 if (task) {
1691 files = get_files_struct(task);
1692 if (files) {
1693 rcu_read_lock();
1694 if (fcheck_files(files, fd)) {
1695 rcu_read_unlock();
1696 put_files_struct(files);
1697 if (task_dumpable(task)) {
1698 rcu_read_lock();
1699 cred = __task_cred(task);
1700 inode->i_uid = cred->euid;
1701 inode->i_gid = cred->egid;
1702 rcu_read_unlock();
1703 } else {
1704 inode->i_uid = 0;
1705 inode->i_gid = 0;
1707 inode->i_mode &= ~(S_ISUID | S_ISGID);
1708 security_task_to_inode(task, inode);
1709 put_task_struct(task);
1710 return 1;
1712 rcu_read_unlock();
1713 put_files_struct(files);
1715 put_task_struct(task);
1717 d_drop(dentry);
1718 return 0;
1721 static const struct dentry_operations tid_fd_dentry_operations =
1723 .d_revalidate = tid_fd_revalidate,
1724 .d_delete = pid_delete_dentry,
1727 static struct dentry *proc_fd_instantiate(struct inode *dir,
1728 struct dentry *dentry, struct task_struct *task, const void *ptr)
1730 unsigned fd = *(const unsigned *)ptr;
1731 struct file *file;
1732 struct files_struct *files;
1733 struct inode *inode;
1734 struct proc_inode *ei;
1735 struct dentry *error = ERR_PTR(-ENOENT);
1737 inode = proc_pid_make_inode(dir->i_sb, task);
1738 if (!inode)
1739 goto out;
1740 ei = PROC_I(inode);
1741 ei->fd = fd;
1742 files = get_files_struct(task);
1743 if (!files)
1744 goto out_iput;
1745 inode->i_mode = S_IFLNK;
1748 * We are not taking a ref to the file structure, so we must
1749 * hold ->file_lock.
1751 spin_lock(&files->file_lock);
1752 file = fcheck_files(files, fd);
1753 if (!file)
1754 goto out_unlock;
1755 if (file->f_mode & FMODE_READ)
1756 inode->i_mode |= S_IRUSR | S_IXUSR;
1757 if (file->f_mode & FMODE_WRITE)
1758 inode->i_mode |= S_IWUSR | S_IXUSR;
1759 spin_unlock(&files->file_lock);
1760 put_files_struct(files);
1762 inode->i_op = &proc_pid_link_inode_operations;
1763 inode->i_size = 64;
1764 ei->op.proc_get_link = proc_fd_link;
1765 dentry->d_op = &tid_fd_dentry_operations;
1766 d_add(dentry, inode);
1767 /* Close the race of the process dying before we return the dentry */
1768 if (tid_fd_revalidate(dentry, NULL))
1769 error = NULL;
1771 out:
1772 return error;
1773 out_unlock:
1774 spin_unlock(&files->file_lock);
1775 put_files_struct(files);
1776 out_iput:
1777 iput(inode);
1778 goto out;
1781 static struct dentry *proc_lookupfd_common(struct inode *dir,
1782 struct dentry *dentry,
1783 instantiate_t instantiate)
1785 struct task_struct *task = get_proc_task(dir);
1786 unsigned fd = name_to_int(dentry);
1787 struct dentry *result = ERR_PTR(-ENOENT);
1789 if (!task)
1790 goto out_no_task;
1791 if (fd == ~0U)
1792 goto out;
1794 result = instantiate(dir, dentry, task, &fd);
1795 out:
1796 put_task_struct(task);
1797 out_no_task:
1798 return result;
1801 static int proc_readfd_common(struct file * filp, void * dirent,
1802 filldir_t filldir, instantiate_t instantiate)
1804 struct dentry *dentry = filp->f_path.dentry;
1805 struct inode *inode = dentry->d_inode;
1806 struct task_struct *p = get_proc_task(inode);
1807 unsigned int fd, ino;
1808 int retval;
1809 struct files_struct * files;
1811 retval = -ENOENT;
1812 if (!p)
1813 goto out_no_task;
1814 retval = 0;
1816 fd = filp->f_pos;
1817 switch (fd) {
1818 case 0:
1819 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1820 goto out;
1821 filp->f_pos++;
1822 case 1:
1823 ino = parent_ino(dentry);
1824 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1825 goto out;
1826 filp->f_pos++;
1827 default:
1828 files = get_files_struct(p);
1829 if (!files)
1830 goto out;
1831 rcu_read_lock();
1832 for (fd = filp->f_pos-2;
1833 fd < files_fdtable(files)->max_fds;
1834 fd++, filp->f_pos++) {
1835 char name[PROC_NUMBUF];
1836 int len;
1838 if (!fcheck_files(files, fd))
1839 continue;
1840 rcu_read_unlock();
1842 len = snprintf(name, sizeof(name), "%d", fd);
1843 if (proc_fill_cache(filp, dirent, filldir,
1844 name, len, instantiate,
1845 p, &fd) < 0) {
1846 rcu_read_lock();
1847 break;
1849 rcu_read_lock();
1851 rcu_read_unlock();
1852 put_files_struct(files);
1854 out:
1855 put_task_struct(p);
1856 out_no_task:
1857 return retval;
1860 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1861 struct nameidata *nd)
1863 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1866 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1868 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1871 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1872 size_t len, loff_t *ppos)
1874 char tmp[PROC_FDINFO_MAX];
1875 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
1876 if (!err)
1877 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1878 return err;
1881 static const struct file_operations proc_fdinfo_file_operations = {
1882 .open = nonseekable_open,
1883 .read = proc_fdinfo_read,
1886 static const struct file_operations proc_fd_operations = {
1887 .read = generic_read_dir,
1888 .readdir = proc_readfd,
1892 * /proc/pid/fd needs a special permission handler so that a process can still
1893 * access /proc/self/fd after it has executed a setuid().
1895 static int proc_fd_permission(struct inode *inode, int mask)
1897 int rv;
1899 rv = generic_permission(inode, mask, NULL);
1900 if (rv == 0)
1901 return 0;
1902 if (task_pid(current) == proc_pid(inode))
1903 rv = 0;
1904 return rv;
1908 * proc directories can do almost nothing..
1910 static const struct inode_operations proc_fd_inode_operations = {
1911 .lookup = proc_lookupfd,
1912 .permission = proc_fd_permission,
1913 .setattr = proc_setattr,
1916 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
1917 struct dentry *dentry, struct task_struct *task, const void *ptr)
1919 unsigned fd = *(unsigned *)ptr;
1920 struct inode *inode;
1921 struct proc_inode *ei;
1922 struct dentry *error = ERR_PTR(-ENOENT);
1924 inode = proc_pid_make_inode(dir->i_sb, task);
1925 if (!inode)
1926 goto out;
1927 ei = PROC_I(inode);
1928 ei->fd = fd;
1929 inode->i_mode = S_IFREG | S_IRUSR;
1930 inode->i_fop = &proc_fdinfo_file_operations;
1931 dentry->d_op = &tid_fd_dentry_operations;
1932 d_add(dentry, inode);
1933 /* Close the race of the process dying before we return the dentry */
1934 if (tid_fd_revalidate(dentry, NULL))
1935 error = NULL;
1937 out:
1938 return error;
1941 static struct dentry *proc_lookupfdinfo(struct inode *dir,
1942 struct dentry *dentry,
1943 struct nameidata *nd)
1945 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
1948 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
1950 return proc_readfd_common(filp, dirent, filldir,
1951 proc_fdinfo_instantiate);
1954 static const struct file_operations proc_fdinfo_operations = {
1955 .read = generic_read_dir,
1956 .readdir = proc_readfdinfo,
1960 * proc directories can do almost nothing..
1962 static const struct inode_operations proc_fdinfo_inode_operations = {
1963 .lookup = proc_lookupfdinfo,
1964 .setattr = proc_setattr,
1968 static struct dentry *proc_pident_instantiate(struct inode *dir,
1969 struct dentry *dentry, struct task_struct *task, const void *ptr)
1971 const struct pid_entry *p = ptr;
1972 struct inode *inode;
1973 struct proc_inode *ei;
1974 struct dentry *error = ERR_PTR(-ENOENT);
1976 inode = proc_pid_make_inode(dir->i_sb, task);
1977 if (!inode)
1978 goto out;
1980 ei = PROC_I(inode);
1981 inode->i_mode = p->mode;
1982 if (S_ISDIR(inode->i_mode))
1983 inode->i_nlink = 2; /* Use getattr to fix if necessary */
1984 if (p->iop)
1985 inode->i_op = p->iop;
1986 if (p->fop)
1987 inode->i_fop = p->fop;
1988 ei->op = p->op;
1989 dentry->d_op = &pid_dentry_operations;
1990 d_add(dentry, inode);
1991 /* Close the race of the process dying before we return the dentry */
1992 if (pid_revalidate(dentry, NULL))
1993 error = NULL;
1994 out:
1995 return error;
1998 static struct dentry *proc_pident_lookup(struct inode *dir,
1999 struct dentry *dentry,
2000 const struct pid_entry *ents,
2001 unsigned int nents)
2003 struct dentry *error;
2004 struct task_struct *task = get_proc_task(dir);
2005 const struct pid_entry *p, *last;
2007 error = ERR_PTR(-ENOENT);
2009 if (!task)
2010 goto out_no_task;
2013 * Yes, it does not scale. And it should not. Don't add
2014 * new entries into /proc/<tgid>/ without very good reasons.
2016 last = &ents[nents - 1];
2017 for (p = ents; p <= last; p++) {
2018 if (p->len != dentry->d_name.len)
2019 continue;
2020 if (!memcmp(dentry->d_name.name, p->name, p->len))
2021 break;
2023 if (p > last)
2024 goto out;
2026 error = proc_pident_instantiate(dir, dentry, task, p);
2027 out:
2028 put_task_struct(task);
2029 out_no_task:
2030 return error;
2033 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2034 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2036 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2037 proc_pident_instantiate, task, p);
2040 static int proc_pident_readdir(struct file *filp,
2041 void *dirent, filldir_t filldir,
2042 const struct pid_entry *ents, unsigned int nents)
2044 int i;
2045 struct dentry *dentry = filp->f_path.dentry;
2046 struct inode *inode = dentry->d_inode;
2047 struct task_struct *task = get_proc_task(inode);
2048 const struct pid_entry *p, *last;
2049 ino_t ino;
2050 int ret;
2052 ret = -ENOENT;
2053 if (!task)
2054 goto out_no_task;
2056 ret = 0;
2057 i = filp->f_pos;
2058 switch (i) {
2059 case 0:
2060 ino = inode->i_ino;
2061 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2062 goto out;
2063 i++;
2064 filp->f_pos++;
2065 /* fall through */
2066 case 1:
2067 ino = parent_ino(dentry);
2068 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2069 goto out;
2070 i++;
2071 filp->f_pos++;
2072 /* fall through */
2073 default:
2074 i -= 2;
2075 if (i >= nents) {
2076 ret = 1;
2077 goto out;
2079 p = ents + i;
2080 last = &ents[nents - 1];
2081 while (p <= last) {
2082 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2083 goto out;
2084 filp->f_pos++;
2085 p++;
2089 ret = 1;
2090 out:
2091 put_task_struct(task);
2092 out_no_task:
2093 return ret;
2096 #ifdef CONFIG_SECURITY
2097 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2098 size_t count, loff_t *ppos)
2100 struct inode * inode = file->f_path.dentry->d_inode;
2101 char *p = NULL;
2102 ssize_t length;
2103 struct task_struct *task = get_proc_task(inode);
2105 if (!task)
2106 return -ESRCH;
2108 length = security_getprocattr(task,
2109 (char*)file->f_path.dentry->d_name.name,
2110 &p);
2111 put_task_struct(task);
2112 if (length > 0)
2113 length = simple_read_from_buffer(buf, count, ppos, p, length);
2114 kfree(p);
2115 return length;
2118 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2119 size_t count, loff_t *ppos)
2121 struct inode * inode = file->f_path.dentry->d_inode;
2122 char *page;
2123 ssize_t length;
2124 struct task_struct *task = get_proc_task(inode);
2126 length = -ESRCH;
2127 if (!task)
2128 goto out_no_task;
2129 if (count > PAGE_SIZE)
2130 count = PAGE_SIZE;
2132 /* No partial writes. */
2133 length = -EINVAL;
2134 if (*ppos != 0)
2135 goto out;
2137 length = -ENOMEM;
2138 page = (char*)__get_free_page(GFP_TEMPORARY);
2139 if (!page)
2140 goto out;
2142 length = -EFAULT;
2143 if (copy_from_user(page, buf, count))
2144 goto out_free;
2146 /* Guard against adverse ptrace interaction */
2147 length = mutex_lock_interruptible(&task->cred_guard_mutex);
2148 if (length < 0)
2149 goto out_free;
2151 length = security_setprocattr(task,
2152 (char*)file->f_path.dentry->d_name.name,
2153 (void*)page, count);
2154 mutex_unlock(&task->cred_guard_mutex);
2155 out_free:
2156 free_page((unsigned long) page);
2157 out:
2158 put_task_struct(task);
2159 out_no_task:
2160 return length;
2163 static const struct file_operations proc_pid_attr_operations = {
2164 .read = proc_pid_attr_read,
2165 .write = proc_pid_attr_write,
2168 static const struct pid_entry attr_dir_stuff[] = {
2169 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2170 REG("prev", S_IRUGO, proc_pid_attr_operations),
2171 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2172 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2173 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2174 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2177 static int proc_attr_dir_readdir(struct file * filp,
2178 void * dirent, filldir_t filldir)
2180 return proc_pident_readdir(filp,dirent,filldir,
2181 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2184 static const struct file_operations proc_attr_dir_operations = {
2185 .read = generic_read_dir,
2186 .readdir = proc_attr_dir_readdir,
2189 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2190 struct dentry *dentry, struct nameidata *nd)
2192 return proc_pident_lookup(dir, dentry,
2193 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2196 static const struct inode_operations proc_attr_dir_inode_operations = {
2197 .lookup = proc_attr_dir_lookup,
2198 .getattr = pid_getattr,
2199 .setattr = proc_setattr,
2202 #endif
2204 #ifdef CONFIG_ELF_CORE
2205 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2206 size_t count, loff_t *ppos)
2208 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2209 struct mm_struct *mm;
2210 char buffer[PROC_NUMBUF];
2211 size_t len;
2212 int ret;
2214 if (!task)
2215 return -ESRCH;
2217 ret = 0;
2218 mm = get_task_mm(task);
2219 if (mm) {
2220 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2221 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2222 MMF_DUMP_FILTER_SHIFT));
2223 mmput(mm);
2224 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2227 put_task_struct(task);
2229 return ret;
2232 static ssize_t proc_coredump_filter_write(struct file *file,
2233 const char __user *buf,
2234 size_t count,
2235 loff_t *ppos)
2237 struct task_struct *task;
2238 struct mm_struct *mm;
2239 char buffer[PROC_NUMBUF], *end;
2240 unsigned int val;
2241 int ret;
2242 int i;
2243 unsigned long mask;
2245 ret = -EFAULT;
2246 memset(buffer, 0, sizeof(buffer));
2247 if (count > sizeof(buffer) - 1)
2248 count = sizeof(buffer) - 1;
2249 if (copy_from_user(buffer, buf, count))
2250 goto out_no_task;
2252 ret = -EINVAL;
2253 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2254 if (*end == '\n')
2255 end++;
2256 if (end - buffer == 0)
2257 goto out_no_task;
2259 ret = -ESRCH;
2260 task = get_proc_task(file->f_dentry->d_inode);
2261 if (!task)
2262 goto out_no_task;
2264 ret = end - buffer;
2265 mm = get_task_mm(task);
2266 if (!mm)
2267 goto out_no_mm;
2269 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2270 if (val & mask)
2271 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2272 else
2273 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2276 mmput(mm);
2277 out_no_mm:
2278 put_task_struct(task);
2279 out_no_task:
2280 return ret;
2283 static const struct file_operations proc_coredump_filter_operations = {
2284 .read = proc_coredump_filter_read,
2285 .write = proc_coredump_filter_write,
2287 #endif
2290 * /proc/self:
2292 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2293 int buflen)
2295 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2296 pid_t tgid = task_tgid_nr_ns(current, ns);
2297 char tmp[PROC_NUMBUF];
2298 if (!tgid)
2299 return -ENOENT;
2300 sprintf(tmp, "%d", tgid);
2301 return vfs_readlink(dentry,buffer,buflen,tmp);
2304 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2306 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2307 pid_t tgid = task_tgid_nr_ns(current, ns);
2308 char tmp[PROC_NUMBUF];
2309 if (!tgid)
2310 return ERR_PTR(-ENOENT);
2311 sprintf(tmp, "%d", task_tgid_nr_ns(current, ns));
2312 return ERR_PTR(vfs_follow_link(nd,tmp));
2315 static const struct inode_operations proc_self_inode_operations = {
2316 .readlink = proc_self_readlink,
2317 .follow_link = proc_self_follow_link,
2321 * proc base
2323 * These are the directory entries in the root directory of /proc
2324 * that properly belong to the /proc filesystem, as they describe
2325 * describe something that is process related.
2327 static const struct pid_entry proc_base_stuff[] = {
2328 NOD("self", S_IFLNK|S_IRWXUGO,
2329 &proc_self_inode_operations, NULL, {}),
2333 * Exceptional case: normally we are not allowed to unhash a busy
2334 * directory. In this case, however, we can do it - no aliasing problems
2335 * due to the way we treat inodes.
2337 static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
2339 struct inode *inode = dentry->d_inode;
2340 struct task_struct *task = get_proc_task(inode);
2341 if (task) {
2342 put_task_struct(task);
2343 return 1;
2345 d_drop(dentry);
2346 return 0;
2349 static const struct dentry_operations proc_base_dentry_operations =
2351 .d_revalidate = proc_base_revalidate,
2352 .d_delete = pid_delete_dentry,
2355 static struct dentry *proc_base_instantiate(struct inode *dir,
2356 struct dentry *dentry, struct task_struct *task, const void *ptr)
2358 const struct pid_entry *p = ptr;
2359 struct inode *inode;
2360 struct proc_inode *ei;
2361 struct dentry *error = ERR_PTR(-EINVAL);
2363 /* Allocate the inode */
2364 error = ERR_PTR(-ENOMEM);
2365 inode = new_inode(dir->i_sb);
2366 if (!inode)
2367 goto out;
2369 /* Initialize the inode */
2370 ei = PROC_I(inode);
2371 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2374 * grab the reference to the task.
2376 ei->pid = get_task_pid(task, PIDTYPE_PID);
2377 if (!ei->pid)
2378 goto out_iput;
2380 inode->i_mode = p->mode;
2381 if (S_ISDIR(inode->i_mode))
2382 inode->i_nlink = 2;
2383 if (S_ISLNK(inode->i_mode))
2384 inode->i_size = 64;
2385 if (p->iop)
2386 inode->i_op = p->iop;
2387 if (p->fop)
2388 inode->i_fop = p->fop;
2389 ei->op = p->op;
2390 dentry->d_op = &proc_base_dentry_operations;
2391 d_add(dentry, inode);
2392 error = NULL;
2393 out:
2394 return error;
2395 out_iput:
2396 iput(inode);
2397 goto out;
2400 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2402 struct dentry *error;
2403 struct task_struct *task = get_proc_task(dir);
2404 const struct pid_entry *p, *last;
2406 error = ERR_PTR(-ENOENT);
2408 if (!task)
2409 goto out_no_task;
2411 /* Lookup the directory entry */
2412 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2413 for (p = proc_base_stuff; p <= last; p++) {
2414 if (p->len != dentry->d_name.len)
2415 continue;
2416 if (!memcmp(dentry->d_name.name, p->name, p->len))
2417 break;
2419 if (p > last)
2420 goto out;
2422 error = proc_base_instantiate(dir, dentry, task, p);
2424 out:
2425 put_task_struct(task);
2426 out_no_task:
2427 return error;
2430 static int proc_base_fill_cache(struct file *filp, void *dirent,
2431 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2433 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2434 proc_base_instantiate, task, p);
2437 #ifdef CONFIG_TASK_IO_ACCOUNTING
2438 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2440 struct task_io_accounting acct = task->ioac;
2441 unsigned long flags;
2443 if (whole && lock_task_sighand(task, &flags)) {
2444 struct task_struct *t = task;
2446 task_io_accounting_add(&acct, &task->signal->ioac);
2447 while_each_thread(task, t)
2448 task_io_accounting_add(&acct, &t->ioac);
2450 unlock_task_sighand(task, &flags);
2452 return sprintf(buffer,
2453 "rchar: %llu\n"
2454 "wchar: %llu\n"
2455 "syscr: %llu\n"
2456 "syscw: %llu\n"
2457 "read_bytes: %llu\n"
2458 "write_bytes: %llu\n"
2459 "cancelled_write_bytes: %llu\n",
2460 (unsigned long long)acct.rchar,
2461 (unsigned long long)acct.wchar,
2462 (unsigned long long)acct.syscr,
2463 (unsigned long long)acct.syscw,
2464 (unsigned long long)acct.read_bytes,
2465 (unsigned long long)acct.write_bytes,
2466 (unsigned long long)acct.cancelled_write_bytes);
2469 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2471 return do_io_accounting(task, buffer, 0);
2474 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2476 return do_io_accounting(task, buffer, 1);
2478 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2480 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2481 struct pid *pid, struct task_struct *task)
2483 seq_printf(m, "%08x\n", task->personality);
2484 return 0;
2488 * Thread groups
2490 static const struct file_operations proc_task_operations;
2491 static const struct inode_operations proc_task_inode_operations;
2493 static const struct pid_entry tgid_base_stuff[] = {
2494 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2495 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2496 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2497 #ifdef CONFIG_NET
2498 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2499 #endif
2500 REG("environ", S_IRUSR, proc_environ_operations),
2501 INF("auxv", S_IRUSR, proc_pid_auxv),
2502 ONE("status", S_IRUGO, proc_pid_status),
2503 ONE("personality", S_IRUSR, proc_pid_personality),
2504 INF("limits", S_IRUSR, proc_pid_limits),
2505 #ifdef CONFIG_SCHED_DEBUG
2506 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2507 #endif
2508 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2509 INF("syscall", S_IRUSR, proc_pid_syscall),
2510 #endif
2511 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2512 ONE("stat", S_IRUGO, proc_tgid_stat),
2513 ONE("statm", S_IRUGO, proc_pid_statm),
2514 REG("maps", S_IRUGO, proc_maps_operations),
2515 #ifdef CONFIG_NUMA
2516 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2517 #endif
2518 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2519 LNK("cwd", proc_cwd_link),
2520 LNK("root", proc_root_link),
2521 LNK("exe", proc_exe_link),
2522 REG("mounts", S_IRUGO, proc_mounts_operations),
2523 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2524 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2525 #ifdef CONFIG_PROC_PAGE_MONITOR
2526 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2527 REG("smaps", S_IRUGO, proc_smaps_operations),
2528 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2529 #endif
2530 #ifdef CONFIG_SECURITY
2531 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2532 #endif
2533 #ifdef CONFIG_KALLSYMS
2534 INF("wchan", S_IRUGO, proc_pid_wchan),
2535 #endif
2536 #ifdef CONFIG_STACKTRACE
2537 ONE("stack", S_IRUSR, proc_pid_stack),
2538 #endif
2539 #ifdef CONFIG_SCHEDSTATS
2540 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2541 #endif
2542 #ifdef CONFIG_LATENCYTOP
2543 REG("latency", S_IRUGO, proc_lstats_operations),
2544 #endif
2545 #ifdef CONFIG_PROC_PID_CPUSET
2546 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2547 #endif
2548 #ifdef CONFIG_CGROUPS
2549 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2550 #endif
2551 INF("oom_score", S_IRUGO, proc_oom_score),
2552 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2553 #ifdef CONFIG_AUDITSYSCALL
2554 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2555 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2556 #endif
2557 #ifdef CONFIG_FAULT_INJECTION
2558 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2559 #endif
2560 #ifdef CONFIG_ELF_CORE
2561 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2562 #endif
2563 #ifdef CONFIG_TASK_IO_ACCOUNTING
2564 INF("io", S_IRUGO, proc_tgid_io_accounting),
2565 #endif
2568 static int proc_tgid_base_readdir(struct file * filp,
2569 void * dirent, filldir_t filldir)
2571 return proc_pident_readdir(filp,dirent,filldir,
2572 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2575 static const struct file_operations proc_tgid_base_operations = {
2576 .read = generic_read_dir,
2577 .readdir = proc_tgid_base_readdir,
2580 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2581 return proc_pident_lookup(dir, dentry,
2582 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2585 static const struct inode_operations proc_tgid_base_inode_operations = {
2586 .lookup = proc_tgid_base_lookup,
2587 .getattr = pid_getattr,
2588 .setattr = proc_setattr,
2591 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2593 struct dentry *dentry, *leader, *dir;
2594 char buf[PROC_NUMBUF];
2595 struct qstr name;
2597 name.name = buf;
2598 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2599 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2600 if (dentry) {
2601 if (!(current->flags & PF_EXITING))
2602 shrink_dcache_parent(dentry);
2603 d_drop(dentry);
2604 dput(dentry);
2607 name.name = buf;
2608 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2609 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2610 if (!leader)
2611 goto out;
2613 name.name = "task";
2614 name.len = strlen(name.name);
2615 dir = d_hash_and_lookup(leader, &name);
2616 if (!dir)
2617 goto out_put_leader;
2619 name.name = buf;
2620 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2621 dentry = d_hash_and_lookup(dir, &name);
2622 if (dentry) {
2623 shrink_dcache_parent(dentry);
2624 d_drop(dentry);
2625 dput(dentry);
2628 dput(dir);
2629 out_put_leader:
2630 dput(leader);
2631 out:
2632 return;
2636 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2637 * @task: task that should be flushed.
2639 * When flushing dentries from proc, one needs to flush them from global
2640 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2641 * in. This call is supposed to do all of this job.
2643 * Looks in the dcache for
2644 * /proc/@pid
2645 * /proc/@tgid/task/@pid
2646 * if either directory is present flushes it and all of it'ts children
2647 * from the dcache.
2649 * It is safe and reasonable to cache /proc entries for a task until
2650 * that task exits. After that they just clog up the dcache with
2651 * useless entries, possibly causing useful dcache entries to be
2652 * flushed instead. This routine is proved to flush those useless
2653 * dcache entries at process exit time.
2655 * NOTE: This routine is just an optimization so it does not guarantee
2656 * that no dcache entries will exist at process exit time it
2657 * just makes it very unlikely that any will persist.
2660 void proc_flush_task(struct task_struct *task)
2662 int i;
2663 struct pid *pid, *tgid;
2664 struct upid *upid;
2666 pid = task_pid(task);
2667 tgid = task_tgid(task);
2669 for (i = 0; i <= pid->level; i++) {
2670 upid = &pid->numbers[i];
2671 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2672 tgid->numbers[i].nr);
2675 upid = &pid->numbers[pid->level];
2676 if (upid->nr == 1)
2677 pid_ns_release_proc(upid->ns);
2680 static struct dentry *proc_pid_instantiate(struct inode *dir,
2681 struct dentry * dentry,
2682 struct task_struct *task, const void *ptr)
2684 struct dentry *error = ERR_PTR(-ENOENT);
2685 struct inode *inode;
2687 inode = proc_pid_make_inode(dir->i_sb, task);
2688 if (!inode)
2689 goto out;
2691 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2692 inode->i_op = &proc_tgid_base_inode_operations;
2693 inode->i_fop = &proc_tgid_base_operations;
2694 inode->i_flags|=S_IMMUTABLE;
2696 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
2697 ARRAY_SIZE(tgid_base_stuff));
2699 dentry->d_op = &pid_dentry_operations;
2701 d_add(dentry, inode);
2702 /* Close the race of the process dying before we return the dentry */
2703 if (pid_revalidate(dentry, NULL))
2704 error = NULL;
2705 out:
2706 return error;
2709 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2711 struct dentry *result = ERR_PTR(-ENOENT);
2712 struct task_struct *task;
2713 unsigned tgid;
2714 struct pid_namespace *ns;
2716 result = proc_base_lookup(dir, dentry);
2717 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2718 goto out;
2720 tgid = name_to_int(dentry);
2721 if (tgid == ~0U)
2722 goto out;
2724 ns = dentry->d_sb->s_fs_info;
2725 rcu_read_lock();
2726 task = find_task_by_pid_ns(tgid, ns);
2727 if (task)
2728 get_task_struct(task);
2729 rcu_read_unlock();
2730 if (!task)
2731 goto out;
2733 result = proc_pid_instantiate(dir, dentry, task, NULL);
2734 put_task_struct(task);
2735 out:
2736 return result;
2740 * Find the first task with tgid >= tgid
2743 struct tgid_iter {
2744 unsigned int tgid;
2745 struct task_struct *task;
2747 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2749 struct pid *pid;
2751 if (iter.task)
2752 put_task_struct(iter.task);
2753 rcu_read_lock();
2754 retry:
2755 iter.task = NULL;
2756 pid = find_ge_pid(iter.tgid, ns);
2757 if (pid) {
2758 iter.tgid = pid_nr_ns(pid, ns);
2759 iter.task = pid_task(pid, PIDTYPE_PID);
2760 /* What we to know is if the pid we have find is the
2761 * pid of a thread_group_leader. Testing for task
2762 * being a thread_group_leader is the obvious thing
2763 * todo but there is a window when it fails, due to
2764 * the pid transfer logic in de_thread.
2766 * So we perform the straight forward test of seeing
2767 * if the pid we have found is the pid of a thread
2768 * group leader, and don't worry if the task we have
2769 * found doesn't happen to be a thread group leader.
2770 * As we don't care in the case of readdir.
2772 if (!iter.task || !has_group_leader_pid(iter.task)) {
2773 iter.tgid += 1;
2774 goto retry;
2776 get_task_struct(iter.task);
2778 rcu_read_unlock();
2779 return iter;
2782 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2784 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2785 struct tgid_iter iter)
2787 char name[PROC_NUMBUF];
2788 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
2789 return proc_fill_cache(filp, dirent, filldir, name, len,
2790 proc_pid_instantiate, iter.task, NULL);
2793 /* for the /proc/ directory itself, after non-process stuff has been done */
2794 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2796 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2797 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
2798 struct tgid_iter iter;
2799 struct pid_namespace *ns;
2801 if (!reaper)
2802 goto out_no_task;
2804 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
2805 const struct pid_entry *p = &proc_base_stuff[nr];
2806 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
2807 goto out;
2810 ns = filp->f_dentry->d_sb->s_fs_info;
2811 iter.task = NULL;
2812 iter.tgid = filp->f_pos - TGID_OFFSET;
2813 for (iter = next_tgid(ns, iter);
2814 iter.task;
2815 iter.tgid += 1, iter = next_tgid(ns, iter)) {
2816 filp->f_pos = iter.tgid + TGID_OFFSET;
2817 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
2818 put_task_struct(iter.task);
2819 goto out;
2822 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
2823 out:
2824 put_task_struct(reaper);
2825 out_no_task:
2826 return 0;
2830 * Tasks
2832 static const struct pid_entry tid_base_stuff[] = {
2833 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2834 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fd_operations),
2835 REG("environ", S_IRUSR, proc_environ_operations),
2836 INF("auxv", S_IRUSR, proc_pid_auxv),
2837 ONE("status", S_IRUGO, proc_pid_status),
2838 ONE("personality", S_IRUSR, proc_pid_personality),
2839 INF("limits", S_IRUSR, proc_pid_limits),
2840 #ifdef CONFIG_SCHED_DEBUG
2841 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2842 #endif
2843 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2844 INF("syscall", S_IRUSR, proc_pid_syscall),
2845 #endif
2846 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2847 ONE("stat", S_IRUGO, proc_tid_stat),
2848 ONE("statm", S_IRUGO, proc_pid_statm),
2849 REG("maps", S_IRUGO, proc_maps_operations),
2850 #ifdef CONFIG_NUMA
2851 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2852 #endif
2853 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2854 LNK("cwd", proc_cwd_link),
2855 LNK("root", proc_root_link),
2856 LNK("exe", proc_exe_link),
2857 REG("mounts", S_IRUGO, proc_mounts_operations),
2858 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2859 #ifdef CONFIG_PROC_PAGE_MONITOR
2860 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2861 REG("smaps", S_IRUGO, proc_smaps_operations),
2862 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2863 #endif
2864 #ifdef CONFIG_SECURITY
2865 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2866 #endif
2867 #ifdef CONFIG_KALLSYMS
2868 INF("wchan", S_IRUGO, proc_pid_wchan),
2869 #endif
2870 #ifdef CONFIG_STACKTRACE
2871 ONE("stack", S_IRUSR, proc_pid_stack),
2872 #endif
2873 #ifdef CONFIG_SCHEDSTATS
2874 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2875 #endif
2876 #ifdef CONFIG_LATENCYTOP
2877 REG("latency", S_IRUGO, proc_lstats_operations),
2878 #endif
2879 #ifdef CONFIG_PROC_PID_CPUSET
2880 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2881 #endif
2882 #ifdef CONFIG_CGROUPS
2883 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2884 #endif
2885 INF("oom_score", S_IRUGO, proc_oom_score),
2886 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2887 #ifdef CONFIG_AUDITSYSCALL
2888 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2889 REG("sessionid", S_IRUSR, proc_sessionid_operations),
2890 #endif
2891 #ifdef CONFIG_FAULT_INJECTION
2892 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2893 #endif
2894 #ifdef CONFIG_TASK_IO_ACCOUNTING
2895 INF("io", S_IRUGO, proc_tid_io_accounting),
2896 #endif
2899 static int proc_tid_base_readdir(struct file * filp,
2900 void * dirent, filldir_t filldir)
2902 return proc_pident_readdir(filp,dirent,filldir,
2903 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
2906 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2907 return proc_pident_lookup(dir, dentry,
2908 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
2911 static const struct file_operations proc_tid_base_operations = {
2912 .read = generic_read_dir,
2913 .readdir = proc_tid_base_readdir,
2916 static const struct inode_operations proc_tid_base_inode_operations = {
2917 .lookup = proc_tid_base_lookup,
2918 .getattr = pid_getattr,
2919 .setattr = proc_setattr,
2922 static struct dentry *proc_task_instantiate(struct inode *dir,
2923 struct dentry *dentry, struct task_struct *task, const void *ptr)
2925 struct dentry *error = ERR_PTR(-ENOENT);
2926 struct inode *inode;
2927 inode = proc_pid_make_inode(dir->i_sb, task);
2929 if (!inode)
2930 goto out;
2931 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2932 inode->i_op = &proc_tid_base_inode_operations;
2933 inode->i_fop = &proc_tid_base_operations;
2934 inode->i_flags|=S_IMMUTABLE;
2936 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
2937 ARRAY_SIZE(tid_base_stuff));
2939 dentry->d_op = &pid_dentry_operations;
2941 d_add(dentry, inode);
2942 /* Close the race of the process dying before we return the dentry */
2943 if (pid_revalidate(dentry, NULL))
2944 error = NULL;
2945 out:
2946 return error;
2949 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2951 struct dentry *result = ERR_PTR(-ENOENT);
2952 struct task_struct *task;
2953 struct task_struct *leader = get_proc_task(dir);
2954 unsigned tid;
2955 struct pid_namespace *ns;
2957 if (!leader)
2958 goto out_no_task;
2960 tid = name_to_int(dentry);
2961 if (tid == ~0U)
2962 goto out;
2964 ns = dentry->d_sb->s_fs_info;
2965 rcu_read_lock();
2966 task = find_task_by_pid_ns(tid, ns);
2967 if (task)
2968 get_task_struct(task);
2969 rcu_read_unlock();
2970 if (!task)
2971 goto out;
2972 if (!same_thread_group(leader, task))
2973 goto out_drop_task;
2975 result = proc_task_instantiate(dir, dentry, task, NULL);
2976 out_drop_task:
2977 put_task_struct(task);
2978 out:
2979 put_task_struct(leader);
2980 out_no_task:
2981 return result;
2985 * Find the first tid of a thread group to return to user space.
2987 * Usually this is just the thread group leader, but if the users
2988 * buffer was too small or there was a seek into the middle of the
2989 * directory we have more work todo.
2991 * In the case of a short read we start with find_task_by_pid.
2993 * In the case of a seek we start with the leader and walk nr
2994 * threads past it.
2996 static struct task_struct *first_tid(struct task_struct *leader,
2997 int tid, int nr, struct pid_namespace *ns)
2999 struct task_struct *pos;
3001 rcu_read_lock();
3002 /* Attempt to start with the pid of a thread */
3003 if (tid && (nr > 0)) {
3004 pos = find_task_by_pid_ns(tid, ns);
3005 if (pos && (pos->group_leader == leader))
3006 goto found;
3009 /* If nr exceeds the number of threads there is nothing todo */
3010 pos = NULL;
3011 if (nr && nr >= get_nr_threads(leader))
3012 goto out;
3014 /* If we haven't found our starting place yet start
3015 * with the leader and walk nr threads forward.
3017 for (pos = leader; nr > 0; --nr) {
3018 pos = next_thread(pos);
3019 if (pos == leader) {
3020 pos = NULL;
3021 goto out;
3024 found:
3025 get_task_struct(pos);
3026 out:
3027 rcu_read_unlock();
3028 return pos;
3032 * Find the next thread in the thread list.
3033 * Return NULL if there is an error or no next thread.
3035 * The reference to the input task_struct is released.
3037 static struct task_struct *next_tid(struct task_struct *start)
3039 struct task_struct *pos = NULL;
3040 rcu_read_lock();
3041 if (pid_alive(start)) {
3042 pos = next_thread(start);
3043 if (thread_group_leader(pos))
3044 pos = NULL;
3045 else
3046 get_task_struct(pos);
3048 rcu_read_unlock();
3049 put_task_struct(start);
3050 return pos;
3053 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3054 struct task_struct *task, int tid)
3056 char name[PROC_NUMBUF];
3057 int len = snprintf(name, sizeof(name), "%d", tid);
3058 return proc_fill_cache(filp, dirent, filldir, name, len,
3059 proc_task_instantiate, task, NULL);
3062 /* for the /proc/TGID/task/ directories */
3063 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3065 struct dentry *dentry = filp->f_path.dentry;
3066 struct inode *inode = dentry->d_inode;
3067 struct task_struct *leader = NULL;
3068 struct task_struct *task;
3069 int retval = -ENOENT;
3070 ino_t ino;
3071 int tid;
3072 struct pid_namespace *ns;
3074 task = get_proc_task(inode);
3075 if (!task)
3076 goto out_no_task;
3077 rcu_read_lock();
3078 if (pid_alive(task)) {
3079 leader = task->group_leader;
3080 get_task_struct(leader);
3082 rcu_read_unlock();
3083 put_task_struct(task);
3084 if (!leader)
3085 goto out_no_task;
3086 retval = 0;
3088 switch ((unsigned long)filp->f_pos) {
3089 case 0:
3090 ino = inode->i_ino;
3091 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3092 goto out;
3093 filp->f_pos++;
3094 /* fall through */
3095 case 1:
3096 ino = parent_ino(dentry);
3097 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3098 goto out;
3099 filp->f_pos++;
3100 /* fall through */
3103 /* f_version caches the tgid value that the last readdir call couldn't
3104 * return. lseek aka telldir automagically resets f_version to 0.
3106 ns = filp->f_dentry->d_sb->s_fs_info;
3107 tid = (int)filp->f_version;
3108 filp->f_version = 0;
3109 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3110 task;
3111 task = next_tid(task), filp->f_pos++) {
3112 tid = task_pid_nr_ns(task, ns);
3113 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3114 /* returning this tgid failed, save it as the first
3115 * pid for the next readir call */
3116 filp->f_version = (u64)tid;
3117 put_task_struct(task);
3118 break;
3121 out:
3122 put_task_struct(leader);
3123 out_no_task:
3124 return retval;
3127 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3129 struct inode *inode = dentry->d_inode;
3130 struct task_struct *p = get_proc_task(inode);
3131 generic_fillattr(inode, stat);
3133 if (p) {
3134 stat->nlink += get_nr_threads(p);
3135 put_task_struct(p);
3138 return 0;
3141 static const struct inode_operations proc_task_inode_operations = {
3142 .lookup = proc_task_lookup,
3143 .getattr = proc_task_getattr,
3144 .setattr = proc_setattr,
3147 static const struct file_operations proc_task_operations = {
3148 .read = generic_read_dir,
3149 .readdir = proc_task_readdir,