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.
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.
35 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36 * Pud inclusion in the page table walking.
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/seq_file.h>
63 #include <linux/namei.h>
64 #include <linux/mnt_namespace.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 #ifdef CONFIG_HARDWALL
87 #include <asm/hardwall.h>
92 * Implementing inode permission operations in /proc is almost
93 * certainly an error. Permission checks need to happen during
94 * each system call not at open time. The reason is that most of
95 * what we wish to check for permissions in /proc varies at runtime.
97 * The classic example of a problem is opening file descriptors
98 * in /proc for a task before it execs a suid executable.
105 const struct inode_operations *iop;
106 const struct file_operations *fop;
110 #define NOD(NAME, MODE, IOP, FOP, OP) { \
112 .len = sizeof(NAME) - 1, \
119 #define DIR(NAME, MODE, iops, fops) \
120 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
121 #define LNK(NAME, get_link) \
122 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
123 &proc_pid_link_inode_operations, NULL, \
124 { .proc_get_link = get_link } )
125 #define REG(NAME, MODE, fops) \
126 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
127 #define INF(NAME, MODE, read) \
128 NOD(NAME, (S_IFREG|(MODE)), \
129 NULL, &proc_info_file_operations, \
130 { .proc_read = read } )
131 #define ONE(NAME, MODE, show) \
132 NOD(NAME, (S_IFREG|(MODE)), \
133 NULL, &proc_single_file_operations, \
134 { .proc_show = show } )
137 * Count the number of hardlinks for the pid_entry table, excluding the .
140 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
147 for (i = 0; i < n; ++i) {
148 if (S_ISDIR(entries[i].mode))
155 static int get_task_root(struct task_struct *task, struct path *root)
157 int result = -ENOENT;
161 get_fs_root(task->fs, root);
168 static int proc_cwd_link(struct inode *inode, struct path *path)
170 struct task_struct *task = get_proc_task(inode);
171 int result = -ENOENT;
176 get_fs_pwd(task->fs, path);
180 put_task_struct(task);
185 static int proc_root_link(struct inode *inode, struct path *path)
187 struct task_struct *task = get_proc_task(inode);
188 int result = -ENOENT;
191 result = get_task_root(task, path);
192 put_task_struct(task);
197 static struct mm_struct *mm_access(struct task_struct *task, unsigned int mode)
199 struct mm_struct *mm;
202 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
206 mm = get_task_mm(task);
207 if (mm && mm != current->mm &&
208 !ptrace_may_access(task, mode)) {
210 mm = ERR_PTR(-EACCES);
212 mutex_unlock(&task->signal->cred_guard_mutex);
217 struct mm_struct *mm_for_maps(struct task_struct *task)
219 return mm_access(task, PTRACE_MODE_READ);
222 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
226 struct mm_struct *mm = get_task_mm(task);
230 goto out_mm; /* Shh! No looking before we're done */
232 len = mm->arg_end - mm->arg_start;
237 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
239 // If the nul at the end of args has been overwritten, then
240 // assume application is using setproctitle(3).
241 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
242 len = strnlen(buffer, res);
246 len = mm->env_end - mm->env_start;
247 if (len > PAGE_SIZE - res)
248 len = PAGE_SIZE - res;
249 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
250 res = strnlen(buffer, res);
259 static int proc_pid_auxv(struct task_struct *task, char *buffer)
261 struct mm_struct *mm = mm_for_maps(task);
262 int res = PTR_ERR(mm);
263 if (mm && !IS_ERR(mm)) {
264 unsigned int nwords = 0;
267 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
268 res = nwords * sizeof(mm->saved_auxv[0]);
271 memcpy(buffer, mm->saved_auxv, res);
278 #ifdef CONFIG_KALLSYMS
280 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
281 * Returns the resolved symbol. If that fails, simply return the address.
283 static int proc_pid_wchan(struct task_struct *task, char *buffer)
286 char symname[KSYM_NAME_LEN];
288 wchan = get_wchan(task);
290 if (lookup_symbol_name(wchan, symname) < 0)
291 if (!ptrace_may_access(task, PTRACE_MODE_READ))
294 return sprintf(buffer, "%lu", wchan);
296 return sprintf(buffer, "%s", symname);
298 #endif /* CONFIG_KALLSYMS */
300 static int lock_trace(struct task_struct *task)
302 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
305 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
306 mutex_unlock(&task->signal->cred_guard_mutex);
312 static void unlock_trace(struct task_struct *task)
314 mutex_unlock(&task->signal->cred_guard_mutex);
317 #ifdef CONFIG_STACKTRACE
319 #define MAX_STACK_TRACE_DEPTH 64
321 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
322 struct pid *pid, struct task_struct *task)
324 struct stack_trace trace;
325 unsigned long *entries;
329 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
333 trace.nr_entries = 0;
334 trace.max_entries = MAX_STACK_TRACE_DEPTH;
335 trace.entries = entries;
338 err = lock_trace(task);
340 save_stack_trace_tsk(task, &trace);
342 for (i = 0; i < trace.nr_entries; i++) {
343 seq_printf(m, "[<%pK>] %pS\n",
344 (void *)entries[i], (void *)entries[i]);
354 #ifdef CONFIG_SCHEDSTATS
356 * Provides /proc/PID/schedstat
358 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
360 return sprintf(buffer, "%llu %llu %lu\n",
361 (unsigned long long)task->se.sum_exec_runtime,
362 (unsigned long long)task->sched_info.run_delay,
363 task->sched_info.pcount);
367 #ifdef CONFIG_LATENCYTOP
368 static int lstats_show_proc(struct seq_file *m, void *v)
371 struct inode *inode = m->private;
372 struct task_struct *task = get_proc_task(inode);
376 seq_puts(m, "Latency Top version : v0.1\n");
377 for (i = 0; i < 32; i++) {
378 struct latency_record *lr = &task->latency_record[i];
379 if (lr->backtrace[0]) {
381 seq_printf(m, "%i %li %li",
382 lr->count, lr->time, lr->max);
383 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
384 unsigned long bt = lr->backtrace[q];
389 seq_printf(m, " %ps", (void *)bt);
395 put_task_struct(task);
399 static int lstats_open(struct inode *inode, struct file *file)
401 return single_open(file, lstats_show_proc, inode);
404 static ssize_t lstats_write(struct file *file, const char __user *buf,
405 size_t count, loff_t *offs)
407 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
411 clear_all_latency_tracing(task);
412 put_task_struct(task);
417 static const struct file_operations proc_lstats_operations = {
420 .write = lstats_write,
422 .release = single_release,
427 static int proc_oom_score(struct task_struct *task, char *buffer)
429 unsigned long points = 0;
431 read_lock(&tasklist_lock);
433 points = oom_badness(task, NULL, NULL,
434 totalram_pages + total_swap_pages);
435 read_unlock(&tasklist_lock);
436 return sprintf(buffer, "%lu\n", points);
444 static const struct limit_names lnames[RLIM_NLIMITS] = {
445 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
446 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
447 [RLIMIT_DATA] = {"Max data size", "bytes"},
448 [RLIMIT_STACK] = {"Max stack size", "bytes"},
449 [RLIMIT_CORE] = {"Max core file size", "bytes"},
450 [RLIMIT_RSS] = {"Max resident set", "bytes"},
451 [RLIMIT_NPROC] = {"Max processes", "processes"},
452 [RLIMIT_NOFILE] = {"Max open files", "files"},
453 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
454 [RLIMIT_AS] = {"Max address space", "bytes"},
455 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
456 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
457 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
458 [RLIMIT_NICE] = {"Max nice priority", NULL},
459 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
460 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
463 /* Display limits for a process */
464 static int proc_pid_limits(struct task_struct *task, char *buffer)
469 char *bufptr = buffer;
471 struct rlimit rlim[RLIM_NLIMITS];
473 if (!lock_task_sighand(task, &flags))
475 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
476 unlock_task_sighand(task, &flags);
479 * print the file header
481 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
482 "Limit", "Soft Limit", "Hard Limit", "Units");
484 for (i = 0; i < RLIM_NLIMITS; i++) {
485 if (rlim[i].rlim_cur == RLIM_INFINITY)
486 count += sprintf(&bufptr[count], "%-25s %-20s ",
487 lnames[i].name, "unlimited");
489 count += sprintf(&bufptr[count], "%-25s %-20lu ",
490 lnames[i].name, rlim[i].rlim_cur);
492 if (rlim[i].rlim_max == RLIM_INFINITY)
493 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
495 count += sprintf(&bufptr[count], "%-20lu ",
499 count += sprintf(&bufptr[count], "%-10s\n",
502 count += sprintf(&bufptr[count], "\n");
508 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
509 static int proc_pid_syscall(struct task_struct *task, char *buffer)
512 unsigned long args[6], sp, pc;
513 int res = lock_trace(task);
517 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
518 res = sprintf(buffer, "running\n");
520 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
522 res = sprintf(buffer,
523 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
525 args[0], args[1], args[2], args[3], args[4], args[5],
530 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
532 /************************************************************************/
533 /* Here the fs part begins */
534 /************************************************************************/
536 /* permission checks */
537 static int proc_fd_access_allowed(struct inode *inode)
539 struct task_struct *task;
541 /* Allow access to a task's file descriptors if it is us or we
542 * may use ptrace attach to the process and find out that
545 task = get_proc_task(inode);
547 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
548 put_task_struct(task);
553 int proc_setattr(struct dentry *dentry, struct iattr *attr)
556 struct inode *inode = dentry->d_inode;
558 if (attr->ia_valid & ATTR_MODE)
561 error = inode_change_ok(inode, attr);
565 if ((attr->ia_valid & ATTR_SIZE) &&
566 attr->ia_size != i_size_read(inode)) {
567 error = vmtruncate(inode, attr->ia_size);
572 setattr_copy(inode, attr);
573 mark_inode_dirty(inode);
577 static const struct inode_operations proc_def_inode_operations = {
578 .setattr = proc_setattr,
581 static int mounts_open_common(struct inode *inode, struct file *file,
582 const struct seq_operations *op)
584 struct task_struct *task = get_proc_task(inode);
586 struct mnt_namespace *ns = NULL;
588 struct proc_mounts *p;
593 nsp = task_nsproxy(task);
600 if (ns && get_task_root(task, &root) == 0)
602 put_task_struct(task);
611 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
615 file->private_data = &p->m;
616 ret = seq_open(file, op);
623 p->m.poll_event = ns->event;
637 static int mounts_release(struct inode *inode, struct file *file)
639 struct proc_mounts *p = file->private_data;
642 return seq_release(inode, file);
645 static unsigned mounts_poll(struct file *file, poll_table *wait)
647 struct proc_mounts *p = file->private_data;
648 unsigned res = POLLIN | POLLRDNORM;
650 poll_wait(file, &p->ns->poll, wait);
651 if (mnt_had_events(p))
652 res |= POLLERR | POLLPRI;
657 static int mounts_open(struct inode *inode, struct file *file)
659 return mounts_open_common(inode, file, &mounts_op);
662 static const struct file_operations proc_mounts_operations = {
666 .release = mounts_release,
670 static int mountinfo_open(struct inode *inode, struct file *file)
672 return mounts_open_common(inode, file, &mountinfo_op);
675 static const struct file_operations proc_mountinfo_operations = {
676 .open = mountinfo_open,
679 .release = mounts_release,
683 static int mountstats_open(struct inode *inode, struct file *file)
685 return mounts_open_common(inode, file, &mountstats_op);
688 static const struct file_operations proc_mountstats_operations = {
689 .open = mountstats_open,
692 .release = mounts_release,
695 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
697 static ssize_t proc_info_read(struct file * file, char __user * buf,
698 size_t count, loff_t *ppos)
700 struct inode * inode = file->f_path.dentry->d_inode;
703 struct task_struct *task = get_proc_task(inode);
709 if (count > PROC_BLOCK_SIZE)
710 count = PROC_BLOCK_SIZE;
713 if (!(page = __get_free_page(GFP_TEMPORARY)))
716 length = PROC_I(inode)->op.proc_read(task, (char*)page);
719 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
722 put_task_struct(task);
727 static const struct file_operations proc_info_file_operations = {
728 .read = proc_info_read,
729 .llseek = generic_file_llseek,
732 static int proc_single_show(struct seq_file *m, void *v)
734 struct inode *inode = m->private;
735 struct pid_namespace *ns;
737 struct task_struct *task;
740 ns = inode->i_sb->s_fs_info;
741 pid = proc_pid(inode);
742 task = get_pid_task(pid, PIDTYPE_PID);
746 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
748 put_task_struct(task);
752 static int proc_single_open(struct inode *inode, struct file *filp)
754 return single_open(filp, proc_single_show, inode);
757 static const struct file_operations proc_single_file_operations = {
758 .open = proc_single_open,
761 .release = single_release,
764 static int mem_open(struct inode* inode, struct file* file)
766 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
767 struct mm_struct *mm;
772 mm = mm_access(task, PTRACE_MODE_ATTACH);
773 put_task_struct(task);
778 /* OK to pass negative loff_t, we can catch out-of-range */
779 file->f_mode |= FMODE_UNSIGNED_OFFSET;
780 file->private_data = mm;
785 static ssize_t mem_read(struct file * file, char __user * buf,
786 size_t count, loff_t *ppos)
790 unsigned long src = *ppos;
791 struct mm_struct *mm = file->private_data;
796 page = (char *)__get_free_page(GFP_TEMPORARY);
803 int this_len, retval;
805 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
806 retval = access_remote_vm(mm, src, page, this_len, 0);
813 if (copy_to_user(buf, page, retval)) {
825 free_page((unsigned long) page);
829 static ssize_t mem_write(struct file * file, const char __user *buf,
830 size_t count, loff_t *ppos)
834 unsigned long dst = *ppos;
835 struct mm_struct *mm = file->private_data;
840 page = (char *)__get_free_page(GFP_TEMPORARY);
846 int this_len, retval;
848 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
849 if (copy_from_user(page, buf, this_len)) {
853 retval = access_remote_vm(mm, dst, page, this_len, 1);
866 free_page((unsigned long) page);
870 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
874 file->f_pos = offset;
877 file->f_pos += offset;
882 force_successful_syscall_return();
886 static int mem_release(struct inode *inode, struct file *file)
888 struct mm_struct *mm = file->private_data;
894 static const struct file_operations proc_mem_operations = {
899 .release = mem_release,
902 static ssize_t environ_read(struct file *file, char __user *buf,
903 size_t count, loff_t *ppos)
905 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
907 unsigned long src = *ppos;
909 struct mm_struct *mm;
915 page = (char *)__get_free_page(GFP_TEMPORARY);
920 mm = mm_for_maps(task);
922 if (!mm || IS_ERR(mm))
927 int this_len, retval, max_len;
929 this_len = mm->env_end - (mm->env_start + src);
934 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
935 this_len = (this_len > max_len) ? max_len : this_len;
937 retval = access_process_vm(task, (mm->env_start + src),
945 if (copy_to_user(buf, page, retval)) {
959 free_page((unsigned long) page);
961 put_task_struct(task);
966 static const struct file_operations proc_environ_operations = {
967 .read = environ_read,
968 .llseek = generic_file_llseek,
971 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
972 size_t count, loff_t *ppos)
974 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
975 char buffer[PROC_NUMBUF];
977 int oom_adjust = OOM_DISABLE;
983 if (lock_task_sighand(task, &flags)) {
984 oom_adjust = task->signal->oom_adj;
985 unlock_task_sighand(task, &flags);
988 put_task_struct(task);
990 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
992 return simple_read_from_buffer(buf, count, ppos, buffer, len);
995 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
996 size_t count, loff_t *ppos)
998 struct task_struct *task;
999 char buffer[PROC_NUMBUF];
1001 unsigned long flags;
1004 memset(buffer, 0, sizeof(buffer));
1005 if (count > sizeof(buffer) - 1)
1006 count = sizeof(buffer) - 1;
1007 if (copy_from_user(buffer, buf, count)) {
1012 err = kstrtoint(strstrip(buffer), 0, &oom_adjust);
1015 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
1016 oom_adjust != OOM_DISABLE) {
1021 task = get_proc_task(file->f_path.dentry->d_inode);
1033 if (!lock_task_sighand(task, &flags)) {
1038 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
1044 * Warn that /proc/pid/oom_adj is deprecated, see
1045 * Documentation/feature-removal-schedule.txt.
1047 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1048 current->comm, task_pid_nr(current), task_pid_nr(task),
1050 task->signal->oom_adj = oom_adjust;
1052 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1053 * value is always attainable.
1055 if (task->signal->oom_adj == OOM_ADJUST_MAX)
1056 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
1058 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
1061 unlock_task_sighand(task, &flags);
1064 put_task_struct(task);
1066 return err < 0 ? err : count;
1069 static const struct file_operations proc_oom_adjust_operations = {
1070 .read = oom_adjust_read,
1071 .write = oom_adjust_write,
1072 .llseek = generic_file_llseek,
1075 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1076 size_t count, loff_t *ppos)
1078 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1079 char buffer[PROC_NUMBUF];
1080 int oom_score_adj = OOM_SCORE_ADJ_MIN;
1081 unsigned long flags;
1086 if (lock_task_sighand(task, &flags)) {
1087 oom_score_adj = task->signal->oom_score_adj;
1088 unlock_task_sighand(task, &flags);
1090 put_task_struct(task);
1091 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1092 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1095 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1096 size_t count, loff_t *ppos)
1098 struct task_struct *task;
1099 char buffer[PROC_NUMBUF];
1100 unsigned long flags;
1104 memset(buffer, 0, sizeof(buffer));
1105 if (count > sizeof(buffer) - 1)
1106 count = sizeof(buffer) - 1;
1107 if (copy_from_user(buffer, buf, count)) {
1112 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1115 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1116 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1121 task = get_proc_task(file->f_path.dentry->d_inode);
1133 if (!lock_task_sighand(task, &flags)) {
1138 if (oom_score_adj < task->signal->oom_score_adj_min &&
1139 !capable(CAP_SYS_RESOURCE)) {
1144 task->signal->oom_score_adj = oom_score_adj;
1145 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1146 task->signal->oom_score_adj_min = oom_score_adj;
1148 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1149 * always attainable.
1151 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1152 task->signal->oom_adj = OOM_DISABLE;
1154 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1157 unlock_task_sighand(task, &flags);
1160 put_task_struct(task);
1162 return err < 0 ? err : count;
1165 static const struct file_operations proc_oom_score_adj_operations = {
1166 .read = oom_score_adj_read,
1167 .write = oom_score_adj_write,
1168 .llseek = default_llseek,
1171 #ifdef CONFIG_AUDITSYSCALL
1172 #define TMPBUFLEN 21
1173 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1174 size_t count, loff_t *ppos)
1176 struct inode * inode = file->f_path.dentry->d_inode;
1177 struct task_struct *task = get_proc_task(inode);
1179 char tmpbuf[TMPBUFLEN];
1183 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1184 audit_get_loginuid(task));
1185 put_task_struct(task);
1186 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1189 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1190 size_t count, loff_t *ppos)
1192 struct inode * inode = file->f_path.dentry->d_inode;
1197 if (!capable(CAP_AUDIT_CONTROL))
1201 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1207 if (count >= PAGE_SIZE)
1208 count = PAGE_SIZE - 1;
1211 /* No partial writes. */
1214 page = (char*)__get_free_page(GFP_TEMPORARY);
1218 if (copy_from_user(page, buf, count))
1222 loginuid = simple_strtoul(page, &tmp, 10);
1228 length = audit_set_loginuid(current, loginuid);
1229 if (likely(length == 0))
1233 free_page((unsigned long) page);
1237 static const struct file_operations proc_loginuid_operations = {
1238 .read = proc_loginuid_read,
1239 .write = proc_loginuid_write,
1240 .llseek = generic_file_llseek,
1243 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1244 size_t count, loff_t *ppos)
1246 struct inode * inode = file->f_path.dentry->d_inode;
1247 struct task_struct *task = get_proc_task(inode);
1249 char tmpbuf[TMPBUFLEN];
1253 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1254 audit_get_sessionid(task));
1255 put_task_struct(task);
1256 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1259 static const struct file_operations proc_sessionid_operations = {
1260 .read = proc_sessionid_read,
1261 .llseek = generic_file_llseek,
1265 #ifdef CONFIG_FAULT_INJECTION
1266 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1267 size_t count, loff_t *ppos)
1269 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1270 char buffer[PROC_NUMBUF];
1276 make_it_fail = task->make_it_fail;
1277 put_task_struct(task);
1279 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1281 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1284 static ssize_t proc_fault_inject_write(struct file * file,
1285 const char __user * buf, size_t count, loff_t *ppos)
1287 struct task_struct *task;
1288 char buffer[PROC_NUMBUF], *end;
1291 if (!capable(CAP_SYS_RESOURCE))
1293 memset(buffer, 0, sizeof(buffer));
1294 if (count > sizeof(buffer) - 1)
1295 count = sizeof(buffer) - 1;
1296 if (copy_from_user(buffer, buf, count))
1298 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1301 task = get_proc_task(file->f_dentry->d_inode);
1304 task->make_it_fail = make_it_fail;
1305 put_task_struct(task);
1310 static const struct file_operations proc_fault_inject_operations = {
1311 .read = proc_fault_inject_read,
1312 .write = proc_fault_inject_write,
1313 .llseek = generic_file_llseek,
1318 #ifdef CONFIG_SCHED_DEBUG
1320 * Print out various scheduling related per-task fields:
1322 static int sched_show(struct seq_file *m, void *v)
1324 struct inode *inode = m->private;
1325 struct task_struct *p;
1327 p = get_proc_task(inode);
1330 proc_sched_show_task(p, m);
1338 sched_write(struct file *file, const char __user *buf,
1339 size_t count, loff_t *offset)
1341 struct inode *inode = file->f_path.dentry->d_inode;
1342 struct task_struct *p;
1344 p = get_proc_task(inode);
1347 proc_sched_set_task(p);
1354 static int sched_open(struct inode *inode, struct file *filp)
1356 return single_open(filp, sched_show, inode);
1359 static const struct file_operations proc_pid_sched_operations = {
1362 .write = sched_write,
1363 .llseek = seq_lseek,
1364 .release = single_release,
1369 #ifdef CONFIG_SCHED_AUTOGROUP
1371 * Print out autogroup related information:
1373 static int sched_autogroup_show(struct seq_file *m, void *v)
1375 struct inode *inode = m->private;
1376 struct task_struct *p;
1378 p = get_proc_task(inode);
1381 proc_sched_autogroup_show_task(p, m);
1389 sched_autogroup_write(struct file *file, const char __user *buf,
1390 size_t count, loff_t *offset)
1392 struct inode *inode = file->f_path.dentry->d_inode;
1393 struct task_struct *p;
1394 char buffer[PROC_NUMBUF];
1398 memset(buffer, 0, sizeof(buffer));
1399 if (count > sizeof(buffer) - 1)
1400 count = sizeof(buffer) - 1;
1401 if (copy_from_user(buffer, buf, count))
1404 err = kstrtoint(strstrip(buffer), 0, &nice);
1408 p = get_proc_task(inode);
1413 err = proc_sched_autogroup_set_nice(p, &err);
1422 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1426 ret = single_open(filp, sched_autogroup_show, NULL);
1428 struct seq_file *m = filp->private_data;
1435 static const struct file_operations proc_pid_sched_autogroup_operations = {
1436 .open = sched_autogroup_open,
1438 .write = sched_autogroup_write,
1439 .llseek = seq_lseek,
1440 .release = single_release,
1443 #endif /* CONFIG_SCHED_AUTOGROUP */
1445 static ssize_t comm_write(struct file *file, const char __user *buf,
1446 size_t count, loff_t *offset)
1448 struct inode *inode = file->f_path.dentry->d_inode;
1449 struct task_struct *p;
1450 char buffer[TASK_COMM_LEN];
1452 memset(buffer, 0, sizeof(buffer));
1453 if (count > sizeof(buffer) - 1)
1454 count = sizeof(buffer) - 1;
1455 if (copy_from_user(buffer, buf, count))
1458 p = get_proc_task(inode);
1462 if (same_thread_group(current, p))
1463 set_task_comm(p, buffer);
1472 static int comm_show(struct seq_file *m, void *v)
1474 struct inode *inode = m->private;
1475 struct task_struct *p;
1477 p = get_proc_task(inode);
1482 seq_printf(m, "%s\n", p->comm);
1490 static int comm_open(struct inode *inode, struct file *filp)
1492 return single_open(filp, comm_show, inode);
1495 static const struct file_operations proc_pid_set_comm_operations = {
1498 .write = comm_write,
1499 .llseek = seq_lseek,
1500 .release = single_release,
1503 static int proc_exe_link(struct inode *inode, struct path *exe_path)
1505 struct task_struct *task;
1506 struct mm_struct *mm;
1507 struct file *exe_file;
1509 task = get_proc_task(inode);
1512 mm = get_task_mm(task);
1513 put_task_struct(task);
1516 exe_file = get_mm_exe_file(mm);
1519 *exe_path = exe_file->f_path;
1520 path_get(&exe_file->f_path);
1527 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1529 struct inode *inode = dentry->d_inode;
1530 int error = -EACCES;
1532 /* We don't need a base pointer in the /proc filesystem */
1533 path_put(&nd->path);
1535 /* Are we allowed to snoop on the tasks file descriptors? */
1536 if (!proc_fd_access_allowed(inode))
1539 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1541 return ERR_PTR(error);
1544 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1546 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1553 pathname = d_path(path, tmp, PAGE_SIZE);
1554 len = PTR_ERR(pathname);
1555 if (IS_ERR(pathname))
1557 len = tmp + PAGE_SIZE - 1 - pathname;
1561 if (copy_to_user(buffer, pathname, len))
1564 free_page((unsigned long)tmp);
1568 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1570 int error = -EACCES;
1571 struct inode *inode = dentry->d_inode;
1574 /* Are we allowed to snoop on the tasks file descriptors? */
1575 if (!proc_fd_access_allowed(inode))
1578 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1582 error = do_proc_readlink(&path, buffer, buflen);
1588 static const struct inode_operations proc_pid_link_inode_operations = {
1589 .readlink = proc_pid_readlink,
1590 .follow_link = proc_pid_follow_link,
1591 .setattr = proc_setattr,
1595 /* building an inode */
1597 static int task_dumpable(struct task_struct *task)
1600 struct mm_struct *mm;
1605 dumpable = get_dumpable(mm);
1612 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1614 struct inode * inode;
1615 struct proc_inode *ei;
1616 const struct cred *cred;
1618 /* We need a new inode */
1620 inode = new_inode(sb);
1626 inode->i_ino = get_next_ino();
1627 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1628 inode->i_op = &proc_def_inode_operations;
1631 * grab the reference to task.
1633 ei->pid = get_task_pid(task, PIDTYPE_PID);
1637 if (task_dumpable(task)) {
1639 cred = __task_cred(task);
1640 inode->i_uid = cred->euid;
1641 inode->i_gid = cred->egid;
1644 security_task_to_inode(task, inode);
1654 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1656 struct inode *inode = dentry->d_inode;
1657 struct task_struct *task;
1658 const struct cred *cred;
1660 generic_fillattr(inode, stat);
1665 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1667 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1668 task_dumpable(task)) {
1669 cred = __task_cred(task);
1670 stat->uid = cred->euid;
1671 stat->gid = cred->egid;
1681 * Exceptional case: normally we are not allowed to unhash a busy
1682 * directory. In this case, however, we can do it - no aliasing problems
1683 * due to the way we treat inodes.
1685 * Rewrite the inode's ownerships here because the owning task may have
1686 * performed a setuid(), etc.
1688 * Before the /proc/pid/status file was created the only way to read
1689 * the effective uid of a /process was to stat /proc/pid. Reading
1690 * /proc/pid/status is slow enough that procps and other packages
1691 * kept stating /proc/pid. To keep the rules in /proc simple I have
1692 * made this apply to all per process world readable and executable
1695 int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1697 struct inode *inode;
1698 struct task_struct *task;
1699 const struct cred *cred;
1701 if (nd && nd->flags & LOOKUP_RCU)
1704 inode = dentry->d_inode;
1705 task = get_proc_task(inode);
1708 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1709 task_dumpable(task)) {
1711 cred = __task_cred(task);
1712 inode->i_uid = cred->euid;
1713 inode->i_gid = cred->egid;
1719 inode->i_mode &= ~(S_ISUID | S_ISGID);
1720 security_task_to_inode(task, inode);
1721 put_task_struct(task);
1728 static int pid_delete_dentry(const struct dentry * dentry)
1730 /* Is the task we represent dead?
1731 * If so, then don't put the dentry on the lru list,
1732 * kill it immediately.
1734 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1737 const struct dentry_operations pid_dentry_operations =
1739 .d_revalidate = pid_revalidate,
1740 .d_delete = pid_delete_dentry,
1746 * Fill a directory entry.
1748 * If possible create the dcache entry and derive our inode number and
1749 * file type from dcache entry.
1751 * Since all of the proc inode numbers are dynamically generated, the inode
1752 * numbers do not exist until the inode is cache. This means creating the
1753 * the dcache entry in readdir is necessary to keep the inode numbers
1754 * reported by readdir in sync with the inode numbers reported
1757 int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1758 const char *name, int len,
1759 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1761 struct dentry *child, *dir = filp->f_path.dentry;
1762 struct inode *inode;
1765 unsigned type = DT_UNKNOWN;
1769 qname.hash = full_name_hash(name, len);
1771 child = d_lookup(dir, &qname);
1774 new = d_alloc(dir, &qname);
1776 child = instantiate(dir->d_inode, new, task, ptr);
1783 if (!child || IS_ERR(child) || !child->d_inode)
1784 goto end_instantiate;
1785 inode = child->d_inode;
1788 type = inode->i_mode >> 12;
1793 ino = find_inode_number(dir, &qname);
1796 return filldir(dirent, name, len, filp->f_pos, ino, type);
1799 static unsigned name_to_int(struct dentry *dentry)
1801 const char *name = dentry->d_name.name;
1802 int len = dentry->d_name.len;
1805 if (len > 1 && *name == '0')
1808 unsigned c = *name++ - '0';
1811 if (n >= (~0U-9)/10)
1821 #define PROC_FDINFO_MAX 64
1823 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1825 struct task_struct *task = get_proc_task(inode);
1826 struct files_struct *files = NULL;
1828 int fd = proc_fd(inode);
1831 files = get_files_struct(task);
1832 put_task_struct(task);
1836 * We are not taking a ref to the file structure, so we must
1839 spin_lock(&files->file_lock);
1840 file = fcheck_files(files, fd);
1842 unsigned int f_flags;
1843 struct fdtable *fdt;
1845 fdt = files_fdtable(files);
1846 f_flags = file->f_flags & ~O_CLOEXEC;
1847 if (FD_ISSET(fd, fdt->close_on_exec))
1848 f_flags |= O_CLOEXEC;
1851 *path = file->f_path;
1852 path_get(&file->f_path);
1855 snprintf(info, PROC_FDINFO_MAX,
1858 (long long) file->f_pos,
1860 spin_unlock(&files->file_lock);
1861 put_files_struct(files);
1864 spin_unlock(&files->file_lock);
1865 put_files_struct(files);
1870 static int proc_fd_link(struct inode *inode, struct path *path)
1872 return proc_fd_info(inode, path, NULL);
1875 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1877 struct inode *inode;
1878 struct task_struct *task;
1880 struct files_struct *files;
1881 const struct cred *cred;
1883 if (nd && nd->flags & LOOKUP_RCU)
1886 inode = dentry->d_inode;
1887 task = get_proc_task(inode);
1888 fd = proc_fd(inode);
1891 files = get_files_struct(task);
1894 if (fcheck_files(files, fd)) {
1896 put_files_struct(files);
1897 if (task_dumpable(task)) {
1899 cred = __task_cred(task);
1900 inode->i_uid = cred->euid;
1901 inode->i_gid = cred->egid;
1907 inode->i_mode &= ~(S_ISUID | S_ISGID);
1908 security_task_to_inode(task, inode);
1909 put_task_struct(task);
1913 put_files_struct(files);
1915 put_task_struct(task);
1921 static const struct dentry_operations tid_fd_dentry_operations =
1923 .d_revalidate = tid_fd_revalidate,
1924 .d_delete = pid_delete_dentry,
1927 static struct dentry *proc_fd_instantiate(struct inode *dir,
1928 struct dentry *dentry, struct task_struct *task, const void *ptr)
1930 unsigned fd = *(const unsigned *)ptr;
1932 struct files_struct *files;
1933 struct inode *inode;
1934 struct proc_inode *ei;
1935 struct dentry *error = ERR_PTR(-ENOENT);
1937 inode = proc_pid_make_inode(dir->i_sb, task);
1942 files = get_files_struct(task);
1945 inode->i_mode = S_IFLNK;
1948 * We are not taking a ref to the file structure, so we must
1951 spin_lock(&files->file_lock);
1952 file = fcheck_files(files, fd);
1955 if (file->f_mode & FMODE_READ)
1956 inode->i_mode |= S_IRUSR | S_IXUSR;
1957 if (file->f_mode & FMODE_WRITE)
1958 inode->i_mode |= S_IWUSR | S_IXUSR;
1959 spin_unlock(&files->file_lock);
1960 put_files_struct(files);
1962 inode->i_op = &proc_pid_link_inode_operations;
1964 ei->op.proc_get_link = proc_fd_link;
1965 d_set_d_op(dentry, &tid_fd_dentry_operations);
1966 d_add(dentry, inode);
1967 /* Close the race of the process dying before we return the dentry */
1968 if (tid_fd_revalidate(dentry, NULL))
1974 spin_unlock(&files->file_lock);
1975 put_files_struct(files);
1981 static struct dentry *proc_lookupfd_common(struct inode *dir,
1982 struct dentry *dentry,
1983 instantiate_t instantiate)
1985 struct task_struct *task = get_proc_task(dir);
1986 unsigned fd = name_to_int(dentry);
1987 struct dentry *result = ERR_PTR(-ENOENT);
1994 result = instantiate(dir, dentry, task, &fd);
1996 put_task_struct(task);
2001 static int proc_readfd_common(struct file * filp, void * dirent,
2002 filldir_t filldir, instantiate_t instantiate)
2004 struct dentry *dentry = filp->f_path.dentry;
2005 struct inode *inode = dentry->d_inode;
2006 struct task_struct *p = get_proc_task(inode);
2007 unsigned int fd, ino;
2009 struct files_struct * files;
2019 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
2023 ino = parent_ino(dentry);
2024 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2028 files = get_files_struct(p);
2032 for (fd = filp->f_pos-2;
2033 fd < files_fdtable(files)->max_fds;
2034 fd++, filp->f_pos++) {
2035 char name[PROC_NUMBUF];
2038 if (!fcheck_files(files, fd))
2042 len = snprintf(name, sizeof(name), "%d", fd);
2043 if (proc_fill_cache(filp, dirent, filldir,
2044 name, len, instantiate,
2052 put_files_struct(files);
2060 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
2061 struct nameidata *nd)
2063 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
2066 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
2068 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
2071 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
2072 size_t len, loff_t *ppos)
2074 char tmp[PROC_FDINFO_MAX];
2075 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
2077 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
2081 static const struct file_operations proc_fdinfo_file_operations = {
2082 .open = nonseekable_open,
2083 .read = proc_fdinfo_read,
2084 .llseek = no_llseek,
2087 static const struct file_operations proc_fd_operations = {
2088 .read = generic_read_dir,
2089 .readdir = proc_readfd,
2090 .llseek = default_llseek,
2094 * /proc/pid/fd needs a special permission handler so that a process can still
2095 * access /proc/self/fd after it has executed a setuid().
2097 static int proc_fd_permission(struct inode *inode, int mask)
2099 int rv = generic_permission(inode, mask);
2102 if (task_pid(current) == proc_pid(inode))
2108 * proc directories can do almost nothing..
2110 static const struct inode_operations proc_fd_inode_operations = {
2111 .lookup = proc_lookupfd,
2112 .permission = proc_fd_permission,
2113 .setattr = proc_setattr,
2116 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2117 struct dentry *dentry, struct task_struct *task, const void *ptr)
2119 unsigned fd = *(unsigned *)ptr;
2120 struct inode *inode;
2121 struct proc_inode *ei;
2122 struct dentry *error = ERR_PTR(-ENOENT);
2124 inode = proc_pid_make_inode(dir->i_sb, task);
2129 inode->i_mode = S_IFREG | S_IRUSR;
2130 inode->i_fop = &proc_fdinfo_file_operations;
2131 d_set_d_op(dentry, &tid_fd_dentry_operations);
2132 d_add(dentry, inode);
2133 /* Close the race of the process dying before we return the dentry */
2134 if (tid_fd_revalidate(dentry, NULL))
2141 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2142 struct dentry *dentry,
2143 struct nameidata *nd)
2145 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2148 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2150 return proc_readfd_common(filp, dirent, filldir,
2151 proc_fdinfo_instantiate);
2154 static const struct file_operations proc_fdinfo_operations = {
2155 .read = generic_read_dir,
2156 .readdir = proc_readfdinfo,
2157 .llseek = default_llseek,
2161 * proc directories can do almost nothing..
2163 static const struct inode_operations proc_fdinfo_inode_operations = {
2164 .lookup = proc_lookupfdinfo,
2165 .setattr = proc_setattr,
2169 static struct dentry *proc_pident_instantiate(struct inode *dir,
2170 struct dentry *dentry, struct task_struct *task, const void *ptr)
2172 const struct pid_entry *p = ptr;
2173 struct inode *inode;
2174 struct proc_inode *ei;
2175 struct dentry *error = ERR_PTR(-ENOENT);
2177 inode = proc_pid_make_inode(dir->i_sb, task);
2182 inode->i_mode = p->mode;
2183 if (S_ISDIR(inode->i_mode))
2184 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2186 inode->i_op = p->iop;
2188 inode->i_fop = p->fop;
2190 d_set_d_op(dentry, &pid_dentry_operations);
2191 d_add(dentry, inode);
2192 /* Close the race of the process dying before we return the dentry */
2193 if (pid_revalidate(dentry, NULL))
2199 static struct dentry *proc_pident_lookup(struct inode *dir,
2200 struct dentry *dentry,
2201 const struct pid_entry *ents,
2204 struct dentry *error;
2205 struct task_struct *task = get_proc_task(dir);
2206 const struct pid_entry *p, *last;
2208 error = ERR_PTR(-ENOENT);
2214 * Yes, it does not scale. And it should not. Don't add
2215 * new entries into /proc/<tgid>/ without very good reasons.
2217 last = &ents[nents - 1];
2218 for (p = ents; p <= last; p++) {
2219 if (p->len != dentry->d_name.len)
2221 if (!memcmp(dentry->d_name.name, p->name, p->len))
2227 error = proc_pident_instantiate(dir, dentry, task, p);
2229 put_task_struct(task);
2234 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2235 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2237 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2238 proc_pident_instantiate, task, p);
2241 static int proc_pident_readdir(struct file *filp,
2242 void *dirent, filldir_t filldir,
2243 const struct pid_entry *ents, unsigned int nents)
2246 struct dentry *dentry = filp->f_path.dentry;
2247 struct inode *inode = dentry->d_inode;
2248 struct task_struct *task = get_proc_task(inode);
2249 const struct pid_entry *p, *last;
2262 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2268 ino = parent_ino(dentry);
2269 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2281 last = &ents[nents - 1];
2283 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2292 put_task_struct(task);
2297 #ifdef CONFIG_SECURITY
2298 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2299 size_t count, loff_t *ppos)
2301 struct inode * inode = file->f_path.dentry->d_inode;
2304 struct task_struct *task = get_proc_task(inode);
2309 length = security_getprocattr(task,
2310 (char*)file->f_path.dentry->d_name.name,
2312 put_task_struct(task);
2314 length = simple_read_from_buffer(buf, count, ppos, p, length);
2319 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2320 size_t count, loff_t *ppos)
2322 struct inode * inode = file->f_path.dentry->d_inode;
2325 struct task_struct *task = get_proc_task(inode);
2330 if (count > PAGE_SIZE)
2333 /* No partial writes. */
2339 page = (char*)__get_free_page(GFP_TEMPORARY);
2344 if (copy_from_user(page, buf, count))
2347 /* Guard against adverse ptrace interaction */
2348 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2352 length = security_setprocattr(task,
2353 (char*)file->f_path.dentry->d_name.name,
2354 (void*)page, count);
2355 mutex_unlock(&task->signal->cred_guard_mutex);
2357 free_page((unsigned long) page);
2359 put_task_struct(task);
2364 static const struct file_operations proc_pid_attr_operations = {
2365 .read = proc_pid_attr_read,
2366 .write = proc_pid_attr_write,
2367 .llseek = generic_file_llseek,
2370 static const struct pid_entry attr_dir_stuff[] = {
2371 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2372 REG("prev", S_IRUGO, proc_pid_attr_operations),
2373 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2374 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2375 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2376 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2379 static int proc_attr_dir_readdir(struct file * filp,
2380 void * dirent, filldir_t filldir)
2382 return proc_pident_readdir(filp,dirent,filldir,
2383 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2386 static const struct file_operations proc_attr_dir_operations = {
2387 .read = generic_read_dir,
2388 .readdir = proc_attr_dir_readdir,
2389 .llseek = default_llseek,
2392 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2393 struct dentry *dentry, struct nameidata *nd)
2395 return proc_pident_lookup(dir, dentry,
2396 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2399 static const struct inode_operations proc_attr_dir_inode_operations = {
2400 .lookup = proc_attr_dir_lookup,
2401 .getattr = pid_getattr,
2402 .setattr = proc_setattr,
2407 #ifdef CONFIG_ELF_CORE
2408 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2409 size_t count, loff_t *ppos)
2411 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2412 struct mm_struct *mm;
2413 char buffer[PROC_NUMBUF];
2421 mm = get_task_mm(task);
2423 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2424 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2425 MMF_DUMP_FILTER_SHIFT));
2427 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2430 put_task_struct(task);
2435 static ssize_t proc_coredump_filter_write(struct file *file,
2436 const char __user *buf,
2440 struct task_struct *task;
2441 struct mm_struct *mm;
2442 char buffer[PROC_NUMBUF], *end;
2449 memset(buffer, 0, sizeof(buffer));
2450 if (count > sizeof(buffer) - 1)
2451 count = sizeof(buffer) - 1;
2452 if (copy_from_user(buffer, buf, count))
2456 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2459 if (end - buffer == 0)
2463 task = get_proc_task(file->f_dentry->d_inode);
2468 mm = get_task_mm(task);
2472 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2474 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2476 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2481 put_task_struct(task);
2486 static const struct file_operations proc_coredump_filter_operations = {
2487 .read = proc_coredump_filter_read,
2488 .write = proc_coredump_filter_write,
2489 .llseek = generic_file_llseek,
2496 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2499 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2500 pid_t tgid = task_tgid_nr_ns(current, ns);
2501 char tmp[PROC_NUMBUF];
2504 sprintf(tmp, "%d", tgid);
2505 return vfs_readlink(dentry,buffer,buflen,tmp);
2508 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2510 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2511 pid_t tgid = task_tgid_nr_ns(current, ns);
2512 char *name = ERR_PTR(-ENOENT);
2516 name = ERR_PTR(-ENOMEM);
2518 sprintf(name, "%d", tgid);
2520 nd_set_link(nd, name);
2524 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2527 char *s = nd_get_link(nd);
2532 static const struct inode_operations proc_self_inode_operations = {
2533 .readlink = proc_self_readlink,
2534 .follow_link = proc_self_follow_link,
2535 .put_link = proc_self_put_link,
2541 * These are the directory entries in the root directory of /proc
2542 * that properly belong to the /proc filesystem, as they describe
2543 * describe something that is process related.
2545 static const struct pid_entry proc_base_stuff[] = {
2546 NOD("self", S_IFLNK|S_IRWXUGO,
2547 &proc_self_inode_operations, NULL, {}),
2550 static struct dentry *proc_base_instantiate(struct inode *dir,
2551 struct dentry *dentry, struct task_struct *task, const void *ptr)
2553 const struct pid_entry *p = ptr;
2554 struct inode *inode;
2555 struct proc_inode *ei;
2556 struct dentry *error;
2558 /* Allocate the inode */
2559 error = ERR_PTR(-ENOMEM);
2560 inode = new_inode(dir->i_sb);
2564 /* Initialize the inode */
2566 inode->i_ino = get_next_ino();
2567 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2570 * grab the reference to the task.
2572 ei->pid = get_task_pid(task, PIDTYPE_PID);
2576 inode->i_mode = p->mode;
2577 if (S_ISDIR(inode->i_mode))
2578 set_nlink(inode, 2);
2579 if (S_ISLNK(inode->i_mode))
2582 inode->i_op = p->iop;
2584 inode->i_fop = p->fop;
2586 d_add(dentry, inode);
2595 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2597 struct dentry *error;
2598 struct task_struct *task = get_proc_task(dir);
2599 const struct pid_entry *p, *last;
2601 error = ERR_PTR(-ENOENT);
2606 /* Lookup the directory entry */
2607 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2608 for (p = proc_base_stuff; p <= last; p++) {
2609 if (p->len != dentry->d_name.len)
2611 if (!memcmp(dentry->d_name.name, p->name, p->len))
2617 error = proc_base_instantiate(dir, dentry, task, p);
2620 put_task_struct(task);
2625 static int proc_base_fill_cache(struct file *filp, void *dirent,
2626 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2628 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2629 proc_base_instantiate, task, p);
2632 #ifdef CONFIG_TASK_IO_ACCOUNTING
2633 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2635 struct task_io_accounting acct = task->ioac;
2636 unsigned long flags;
2639 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2643 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2648 if (whole && lock_task_sighand(task, &flags)) {
2649 struct task_struct *t = task;
2651 task_io_accounting_add(&acct, &task->signal->ioac);
2652 while_each_thread(task, t)
2653 task_io_accounting_add(&acct, &t->ioac);
2655 unlock_task_sighand(task, &flags);
2657 result = sprintf(buffer,
2662 "read_bytes: %llu\n"
2663 "write_bytes: %llu\n"
2664 "cancelled_write_bytes: %llu\n",
2665 (unsigned long long)acct.rchar,
2666 (unsigned long long)acct.wchar,
2667 (unsigned long long)acct.syscr,
2668 (unsigned long long)acct.syscw,
2669 (unsigned long long)acct.read_bytes,
2670 (unsigned long long)acct.write_bytes,
2671 (unsigned long long)acct.cancelled_write_bytes);
2673 mutex_unlock(&task->signal->cred_guard_mutex);
2677 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2679 return do_io_accounting(task, buffer, 0);
2682 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2684 return do_io_accounting(task, buffer, 1);
2686 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2688 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2689 struct pid *pid, struct task_struct *task)
2691 int err = lock_trace(task);
2693 seq_printf(m, "%08x\n", task->personality);
2702 static const struct file_operations proc_task_operations;
2703 static const struct inode_operations proc_task_inode_operations;
2705 static const struct pid_entry tgid_base_stuff[] = {
2706 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2707 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2708 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2709 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2711 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2713 REG("environ", S_IRUSR, proc_environ_operations),
2714 INF("auxv", S_IRUSR, proc_pid_auxv),
2715 ONE("status", S_IRUGO, proc_pid_status),
2716 ONE("personality", S_IRUGO, proc_pid_personality),
2717 INF("limits", S_IRUGO, proc_pid_limits),
2718 #ifdef CONFIG_SCHED_DEBUG
2719 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2721 #ifdef CONFIG_SCHED_AUTOGROUP
2722 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2724 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2725 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2726 INF("syscall", S_IRUGO, proc_pid_syscall),
2728 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2729 ONE("stat", S_IRUGO, proc_tgid_stat),
2730 ONE("statm", S_IRUGO, proc_pid_statm),
2731 REG("maps", S_IRUGO, proc_maps_operations),
2733 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2735 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2736 LNK("cwd", proc_cwd_link),
2737 LNK("root", proc_root_link),
2738 LNK("exe", proc_exe_link),
2739 REG("mounts", S_IRUGO, proc_mounts_operations),
2740 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2741 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2742 #ifdef CONFIG_PROC_PAGE_MONITOR
2743 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2744 REG("smaps", S_IRUGO, proc_smaps_operations),
2745 REG("pagemap", S_IRUGO, proc_pagemap_operations),
2747 #ifdef CONFIG_SECURITY
2748 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2750 #ifdef CONFIG_KALLSYMS
2751 INF("wchan", S_IRUGO, proc_pid_wchan),
2753 #ifdef CONFIG_STACKTRACE
2754 ONE("stack", S_IRUGO, proc_pid_stack),
2756 #ifdef CONFIG_SCHEDSTATS
2757 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2759 #ifdef CONFIG_LATENCYTOP
2760 REG("latency", S_IRUGO, proc_lstats_operations),
2762 #ifdef CONFIG_PROC_PID_CPUSET
2763 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2765 #ifdef CONFIG_CGROUPS
2766 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2768 INF("oom_score", S_IRUGO, proc_oom_score),
2769 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2770 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2771 #ifdef CONFIG_AUDITSYSCALL
2772 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2773 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2775 #ifdef CONFIG_FAULT_INJECTION
2776 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2778 #ifdef CONFIG_ELF_CORE
2779 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2781 #ifdef CONFIG_TASK_IO_ACCOUNTING
2782 INF("io", S_IRUSR, proc_tgid_io_accounting),
2784 #ifdef CONFIG_HARDWALL
2785 INF("hardwall", S_IRUGO, proc_pid_hardwall),
2789 static int proc_tgid_base_readdir(struct file * filp,
2790 void * dirent, filldir_t filldir)
2792 return proc_pident_readdir(filp,dirent,filldir,
2793 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2796 static const struct file_operations proc_tgid_base_operations = {
2797 .read = generic_read_dir,
2798 .readdir = proc_tgid_base_readdir,
2799 .llseek = default_llseek,
2802 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2803 return proc_pident_lookup(dir, dentry,
2804 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2807 static const struct inode_operations proc_tgid_base_inode_operations = {
2808 .lookup = proc_tgid_base_lookup,
2809 .getattr = pid_getattr,
2810 .setattr = proc_setattr,
2813 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2815 struct dentry *dentry, *leader, *dir;
2816 char buf[PROC_NUMBUF];
2820 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2821 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2823 shrink_dcache_parent(dentry);
2829 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2830 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2835 name.len = strlen(name.name);
2836 dir = d_hash_and_lookup(leader, &name);
2838 goto out_put_leader;
2841 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2842 dentry = d_hash_and_lookup(dir, &name);
2844 shrink_dcache_parent(dentry);
2857 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2858 * @task: task that should be flushed.
2860 * When flushing dentries from proc, one needs to flush them from global
2861 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2862 * in. This call is supposed to do all of this job.
2864 * Looks in the dcache for
2866 * /proc/@tgid/task/@pid
2867 * if either directory is present flushes it and all of it'ts children
2870 * It is safe and reasonable to cache /proc entries for a task until
2871 * that task exits. After that they just clog up the dcache with
2872 * useless entries, possibly causing useful dcache entries to be
2873 * flushed instead. This routine is proved to flush those useless
2874 * dcache entries at process exit time.
2876 * NOTE: This routine is just an optimization so it does not guarantee
2877 * that no dcache entries will exist at process exit time it
2878 * just makes it very unlikely that any will persist.
2881 void proc_flush_task(struct task_struct *task)
2884 struct pid *pid, *tgid;
2887 pid = task_pid(task);
2888 tgid = task_tgid(task);
2890 for (i = 0; i <= pid->level; i++) {
2891 upid = &pid->numbers[i];
2892 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2893 tgid->numbers[i].nr);
2896 upid = &pid->numbers[pid->level];
2898 pid_ns_release_proc(upid->ns);
2901 static struct dentry *proc_pid_instantiate(struct inode *dir,
2902 struct dentry * dentry,
2903 struct task_struct *task, const void *ptr)
2905 struct dentry *error = ERR_PTR(-ENOENT);
2906 struct inode *inode;
2908 inode = proc_pid_make_inode(dir->i_sb, task);
2912 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2913 inode->i_op = &proc_tgid_base_inode_operations;
2914 inode->i_fop = &proc_tgid_base_operations;
2915 inode->i_flags|=S_IMMUTABLE;
2917 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
2918 ARRAY_SIZE(tgid_base_stuff)));
2920 d_set_d_op(dentry, &pid_dentry_operations);
2922 d_add(dentry, inode);
2923 /* Close the race of the process dying before we return the dentry */
2924 if (pid_revalidate(dentry, NULL))
2930 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2932 struct dentry *result;
2933 struct task_struct *task;
2935 struct pid_namespace *ns;
2937 result = proc_base_lookup(dir, dentry);
2938 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2941 tgid = name_to_int(dentry);
2945 ns = dentry->d_sb->s_fs_info;
2947 task = find_task_by_pid_ns(tgid, ns);
2949 get_task_struct(task);
2954 result = proc_pid_instantiate(dir, dentry, task, NULL);
2955 put_task_struct(task);
2961 * Find the first task with tgid >= tgid
2966 struct task_struct *task;
2968 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2973 put_task_struct(iter.task);
2977 pid = find_ge_pid(iter.tgid, ns);
2979 iter.tgid = pid_nr_ns(pid, ns);
2980 iter.task = pid_task(pid, PIDTYPE_PID);
2981 /* What we to know is if the pid we have find is the
2982 * pid of a thread_group_leader. Testing for task
2983 * being a thread_group_leader is the obvious thing
2984 * todo but there is a window when it fails, due to
2985 * the pid transfer logic in de_thread.
2987 * So we perform the straight forward test of seeing
2988 * if the pid we have found is the pid of a thread
2989 * group leader, and don't worry if the task we have
2990 * found doesn't happen to be a thread group leader.
2991 * As we don't care in the case of readdir.
2993 if (!iter.task || !has_group_leader_pid(iter.task)) {
2997 get_task_struct(iter.task);
3003 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
3005 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3006 struct tgid_iter iter)
3008 char name[PROC_NUMBUF];
3009 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
3010 return proc_fill_cache(filp, dirent, filldir, name, len,
3011 proc_pid_instantiate, iter.task, NULL);
3014 /* for the /proc/ directory itself, after non-process stuff has been done */
3015 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3018 struct task_struct *reaper;
3019 struct tgid_iter iter;
3020 struct pid_namespace *ns;
3022 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3024 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3026 reaper = get_proc_task(filp->f_path.dentry->d_inode);
3030 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
3031 const struct pid_entry *p = &proc_base_stuff[nr];
3032 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
3036 ns = filp->f_dentry->d_sb->s_fs_info;
3038 iter.tgid = filp->f_pos - TGID_OFFSET;
3039 for (iter = next_tgid(ns, iter);
3041 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3042 filp->f_pos = iter.tgid + TGID_OFFSET;
3043 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
3044 put_task_struct(iter.task);
3048 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3050 put_task_struct(reaper);
3058 static const struct pid_entry tid_base_stuff[] = {
3059 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3060 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3061 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3062 REG("environ", S_IRUSR, proc_environ_operations),
3063 INF("auxv", S_IRUSR, proc_pid_auxv),
3064 ONE("status", S_IRUGO, proc_pid_status),
3065 ONE("personality", S_IRUGO, proc_pid_personality),
3066 INF("limits", S_IRUGO, proc_pid_limits),
3067 #ifdef CONFIG_SCHED_DEBUG
3068 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3070 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3071 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3072 INF("syscall", S_IRUGO, proc_pid_syscall),
3074 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3075 ONE("stat", S_IRUGO, proc_tid_stat),
3076 ONE("statm", S_IRUGO, proc_pid_statm),
3077 REG("maps", S_IRUGO, proc_maps_operations),
3079 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
3081 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3082 LNK("cwd", proc_cwd_link),
3083 LNK("root", proc_root_link),
3084 LNK("exe", proc_exe_link),
3085 REG("mounts", S_IRUGO, proc_mounts_operations),
3086 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3087 #ifdef CONFIG_PROC_PAGE_MONITOR
3088 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3089 REG("smaps", S_IRUGO, proc_smaps_operations),
3090 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3092 #ifdef CONFIG_SECURITY
3093 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3095 #ifdef CONFIG_KALLSYMS
3096 INF("wchan", S_IRUGO, proc_pid_wchan),
3098 #ifdef CONFIG_STACKTRACE
3099 ONE("stack", S_IRUGO, proc_pid_stack),
3101 #ifdef CONFIG_SCHEDSTATS
3102 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3104 #ifdef CONFIG_LATENCYTOP
3105 REG("latency", S_IRUGO, proc_lstats_operations),
3107 #ifdef CONFIG_PROC_PID_CPUSET
3108 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3110 #ifdef CONFIG_CGROUPS
3111 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3113 INF("oom_score", S_IRUGO, proc_oom_score),
3114 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3115 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3116 #ifdef CONFIG_AUDITSYSCALL
3117 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3118 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3120 #ifdef CONFIG_FAULT_INJECTION
3121 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3123 #ifdef CONFIG_TASK_IO_ACCOUNTING
3124 INF("io", S_IRUSR, proc_tid_io_accounting),
3126 #ifdef CONFIG_HARDWALL
3127 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3131 static int proc_tid_base_readdir(struct file * filp,
3132 void * dirent, filldir_t filldir)
3134 return proc_pident_readdir(filp,dirent,filldir,
3135 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3138 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3139 return proc_pident_lookup(dir, dentry,
3140 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3143 static const struct file_operations proc_tid_base_operations = {
3144 .read = generic_read_dir,
3145 .readdir = proc_tid_base_readdir,
3146 .llseek = default_llseek,
3149 static const struct inode_operations proc_tid_base_inode_operations = {
3150 .lookup = proc_tid_base_lookup,
3151 .getattr = pid_getattr,
3152 .setattr = proc_setattr,
3155 static struct dentry *proc_task_instantiate(struct inode *dir,
3156 struct dentry *dentry, struct task_struct *task, const void *ptr)
3158 struct dentry *error = ERR_PTR(-ENOENT);
3159 struct inode *inode;
3160 inode = proc_pid_make_inode(dir->i_sb, task);
3164 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3165 inode->i_op = &proc_tid_base_inode_operations;
3166 inode->i_fop = &proc_tid_base_operations;
3167 inode->i_flags|=S_IMMUTABLE;
3169 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3170 ARRAY_SIZE(tid_base_stuff)));
3172 d_set_d_op(dentry, &pid_dentry_operations);
3174 d_add(dentry, inode);
3175 /* Close the race of the process dying before we return the dentry */
3176 if (pid_revalidate(dentry, NULL))
3182 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3184 struct dentry *result = ERR_PTR(-ENOENT);
3185 struct task_struct *task;
3186 struct task_struct *leader = get_proc_task(dir);
3188 struct pid_namespace *ns;
3193 tid = name_to_int(dentry);
3197 ns = dentry->d_sb->s_fs_info;
3199 task = find_task_by_pid_ns(tid, ns);
3201 get_task_struct(task);
3205 if (!same_thread_group(leader, task))
3208 result = proc_task_instantiate(dir, dentry, task, NULL);
3210 put_task_struct(task);
3212 put_task_struct(leader);
3218 * Find the first tid of a thread group to return to user space.
3220 * Usually this is just the thread group leader, but if the users
3221 * buffer was too small or there was a seek into the middle of the
3222 * directory we have more work todo.
3224 * In the case of a short read we start with find_task_by_pid.
3226 * In the case of a seek we start with the leader and walk nr
3229 static struct task_struct *first_tid(struct task_struct *leader,
3230 int tid, int nr, struct pid_namespace *ns)
3232 struct task_struct *pos;
3235 /* Attempt to start with the pid of a thread */
3236 if (tid && (nr > 0)) {
3237 pos = find_task_by_pid_ns(tid, ns);
3238 if (pos && (pos->group_leader == leader))
3242 /* If nr exceeds the number of threads there is nothing todo */
3244 if (nr && nr >= get_nr_threads(leader))
3247 /* If we haven't found our starting place yet start
3248 * with the leader and walk nr threads forward.
3250 for (pos = leader; nr > 0; --nr) {
3251 pos = next_thread(pos);
3252 if (pos == leader) {
3258 get_task_struct(pos);
3265 * Find the next thread in the thread list.
3266 * Return NULL if there is an error or no next thread.
3268 * The reference to the input task_struct is released.
3270 static struct task_struct *next_tid(struct task_struct *start)
3272 struct task_struct *pos = NULL;
3274 if (pid_alive(start)) {
3275 pos = next_thread(start);
3276 if (thread_group_leader(pos))
3279 get_task_struct(pos);
3282 put_task_struct(start);
3286 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3287 struct task_struct *task, int tid)
3289 char name[PROC_NUMBUF];
3290 int len = snprintf(name, sizeof(name), "%d", tid);
3291 return proc_fill_cache(filp, dirent, filldir, name, len,
3292 proc_task_instantiate, task, NULL);
3295 /* for the /proc/TGID/task/ directories */
3296 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3298 struct dentry *dentry = filp->f_path.dentry;
3299 struct inode *inode = dentry->d_inode;
3300 struct task_struct *leader = NULL;
3301 struct task_struct *task;
3302 int retval = -ENOENT;
3305 struct pid_namespace *ns;
3307 task = get_proc_task(inode);
3311 if (pid_alive(task)) {
3312 leader = task->group_leader;
3313 get_task_struct(leader);
3316 put_task_struct(task);
3321 switch ((unsigned long)filp->f_pos) {
3324 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3329 ino = parent_ino(dentry);
3330 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3336 /* f_version caches the tgid value that the last readdir call couldn't
3337 * return. lseek aka telldir automagically resets f_version to 0.
3339 ns = filp->f_dentry->d_sb->s_fs_info;
3340 tid = (int)filp->f_version;
3341 filp->f_version = 0;
3342 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3344 task = next_tid(task), filp->f_pos++) {
3345 tid = task_pid_nr_ns(task, ns);
3346 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3347 /* returning this tgid failed, save it as the first
3348 * pid for the next readir call */
3349 filp->f_version = (u64)tid;
3350 put_task_struct(task);
3355 put_task_struct(leader);
3360 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3362 struct inode *inode = dentry->d_inode;
3363 struct task_struct *p = get_proc_task(inode);
3364 generic_fillattr(inode, stat);
3367 stat->nlink += get_nr_threads(p);
3374 static const struct inode_operations proc_task_inode_operations = {
3375 .lookup = proc_task_lookup,
3376 .getattr = proc_task_getattr,
3377 .setattr = proc_setattr,
3380 static const struct file_operations proc_task_operations = {
3381 .read = generic_read_dir,
3382 .readdir = proc_task_readdir,
3383 .llseek = default_llseek,