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/user_namespace.h>
85 #include <linux/fs_struct.h>
86 #include <linux/slab.h>
87 #include <linux/flex_array.h>
88 #ifdef CONFIG_HARDWALL
89 #include <asm/hardwall.h>
91 #include <trace/events/oom.h>
95 * Implementing inode permission operations in /proc is almost
96 * certainly an error. Permission checks need to happen during
97 * each system call not at open time. The reason is that most of
98 * what we wish to check for permissions in /proc varies at runtime.
100 * The classic example of a problem is opening file descriptors
101 * in /proc for a task before it execs a suid executable.
108 const struct inode_operations *iop;
109 const struct file_operations *fop;
113 #define NOD(NAME, MODE, IOP, FOP, OP) { \
115 .len = sizeof(NAME) - 1, \
122 #define DIR(NAME, MODE, iops, fops) \
123 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
124 #define LNK(NAME, get_link) \
125 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
126 &proc_pid_link_inode_operations, NULL, \
127 { .proc_get_link = get_link } )
128 #define REG(NAME, MODE, fops) \
129 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
130 #define INF(NAME, MODE, read) \
131 NOD(NAME, (S_IFREG|(MODE)), \
132 NULL, &proc_info_file_operations, \
133 { .proc_read = read } )
134 #define ONE(NAME, MODE, show) \
135 NOD(NAME, (S_IFREG|(MODE)), \
136 NULL, &proc_single_file_operations, \
137 { .proc_show = show } )
139 static int proc_fd_permission(struct inode *inode, int mask);
142 * Count the number of hardlinks for the pid_entry table, excluding the .
145 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
152 for (i = 0; i < n; ++i) {
153 if (S_ISDIR(entries[i].mode))
160 static int get_task_root(struct task_struct *task, struct path *root)
162 int result = -ENOENT;
166 get_fs_root(task->fs, root);
173 static int proc_cwd_link(struct dentry *dentry, struct path *path)
175 struct task_struct *task = get_proc_task(dentry->d_inode);
176 int result = -ENOENT;
181 get_fs_pwd(task->fs, path);
185 put_task_struct(task);
190 static int proc_root_link(struct dentry *dentry, struct path *path)
192 struct task_struct *task = get_proc_task(dentry->d_inode);
193 int result = -ENOENT;
196 result = get_task_root(task, path);
197 put_task_struct(task);
202 struct mm_struct *mm_for_maps(struct task_struct *task)
204 return mm_access(task, PTRACE_MODE_READ);
207 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
211 struct mm_struct *mm = get_task_mm(task);
215 goto out_mm; /* Shh! No looking before we're done */
217 len = mm->arg_end - mm->arg_start;
222 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
224 // If the nul at the end of args has been overwritten, then
225 // assume application is using setproctitle(3).
226 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
227 len = strnlen(buffer, res);
231 len = mm->env_end - mm->env_start;
232 if (len > PAGE_SIZE - res)
233 len = PAGE_SIZE - res;
234 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
235 res = strnlen(buffer, res);
244 static int proc_pid_auxv(struct task_struct *task, char *buffer)
246 struct mm_struct *mm = mm_for_maps(task);
247 int res = PTR_ERR(mm);
248 if (mm && !IS_ERR(mm)) {
249 unsigned int nwords = 0;
252 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
253 res = nwords * sizeof(mm->saved_auxv[0]);
256 memcpy(buffer, mm->saved_auxv, res);
263 #ifdef CONFIG_KALLSYMS
265 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
266 * Returns the resolved symbol. If that fails, simply return the address.
268 static int proc_pid_wchan(struct task_struct *task, char *buffer)
271 char symname[KSYM_NAME_LEN];
273 wchan = get_wchan(task);
275 if (lookup_symbol_name(wchan, symname) < 0)
276 if (!ptrace_may_access(task, PTRACE_MODE_READ))
279 return sprintf(buffer, "%lu", wchan);
281 return sprintf(buffer, "%s", symname);
283 #endif /* CONFIG_KALLSYMS */
285 static int lock_trace(struct task_struct *task)
287 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
290 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
291 mutex_unlock(&task->signal->cred_guard_mutex);
297 static void unlock_trace(struct task_struct *task)
299 mutex_unlock(&task->signal->cred_guard_mutex);
302 #ifdef CONFIG_STACKTRACE
304 #define MAX_STACK_TRACE_DEPTH 64
306 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
307 struct pid *pid, struct task_struct *task)
309 struct stack_trace trace;
310 unsigned long *entries;
314 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
318 trace.nr_entries = 0;
319 trace.max_entries = MAX_STACK_TRACE_DEPTH;
320 trace.entries = entries;
323 err = lock_trace(task);
325 save_stack_trace_tsk(task, &trace);
327 for (i = 0; i < trace.nr_entries; i++) {
328 seq_printf(m, "[<%pK>] %pS\n",
329 (void *)entries[i], (void *)entries[i]);
339 #ifdef CONFIG_SCHEDSTATS
341 * Provides /proc/PID/schedstat
343 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
345 return sprintf(buffer, "%llu %llu %lu\n",
346 (unsigned long long)task->se.sum_exec_runtime,
347 (unsigned long long)task->sched_info.run_delay,
348 task->sched_info.pcount);
352 #ifdef CONFIG_LATENCYTOP
353 static int lstats_show_proc(struct seq_file *m, void *v)
356 struct inode *inode = m->private;
357 struct task_struct *task = get_proc_task(inode);
361 seq_puts(m, "Latency Top version : v0.1\n");
362 for (i = 0; i < 32; i++) {
363 struct latency_record *lr = &task->latency_record[i];
364 if (lr->backtrace[0]) {
366 seq_printf(m, "%i %li %li",
367 lr->count, lr->time, lr->max);
368 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
369 unsigned long bt = lr->backtrace[q];
374 seq_printf(m, " %ps", (void *)bt);
380 put_task_struct(task);
384 static int lstats_open(struct inode *inode, struct file *file)
386 return single_open(file, lstats_show_proc, inode);
389 static ssize_t lstats_write(struct file *file, const char __user *buf,
390 size_t count, loff_t *offs)
392 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
396 clear_all_latency_tracing(task);
397 put_task_struct(task);
402 static const struct file_operations proc_lstats_operations = {
405 .write = lstats_write,
407 .release = single_release,
412 static int proc_oom_score(struct task_struct *task, char *buffer)
414 unsigned long totalpages = totalram_pages + total_swap_pages;
415 unsigned long points = 0;
417 read_lock(&tasklist_lock);
419 points = oom_badness(task, NULL, NULL, totalpages) *
421 read_unlock(&tasklist_lock);
422 return sprintf(buffer, "%lu\n", points);
430 static const struct limit_names lnames[RLIM_NLIMITS] = {
431 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
432 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
433 [RLIMIT_DATA] = {"Max data size", "bytes"},
434 [RLIMIT_STACK] = {"Max stack size", "bytes"},
435 [RLIMIT_CORE] = {"Max core file size", "bytes"},
436 [RLIMIT_RSS] = {"Max resident set", "bytes"},
437 [RLIMIT_NPROC] = {"Max processes", "processes"},
438 [RLIMIT_NOFILE] = {"Max open files", "files"},
439 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
440 [RLIMIT_AS] = {"Max address space", "bytes"},
441 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
442 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
443 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
444 [RLIMIT_NICE] = {"Max nice priority", NULL},
445 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
446 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
449 /* Display limits for a process */
450 static int proc_pid_limits(struct task_struct *task, char *buffer)
455 char *bufptr = buffer;
457 struct rlimit rlim[RLIM_NLIMITS];
459 if (!lock_task_sighand(task, &flags))
461 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
462 unlock_task_sighand(task, &flags);
465 * print the file header
467 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
468 "Limit", "Soft Limit", "Hard Limit", "Units");
470 for (i = 0; i < RLIM_NLIMITS; i++) {
471 if (rlim[i].rlim_cur == RLIM_INFINITY)
472 count += sprintf(&bufptr[count], "%-25s %-20s ",
473 lnames[i].name, "unlimited");
475 count += sprintf(&bufptr[count], "%-25s %-20lu ",
476 lnames[i].name, rlim[i].rlim_cur);
478 if (rlim[i].rlim_max == RLIM_INFINITY)
479 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
481 count += sprintf(&bufptr[count], "%-20lu ",
485 count += sprintf(&bufptr[count], "%-10s\n",
488 count += sprintf(&bufptr[count], "\n");
494 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
495 static int proc_pid_syscall(struct task_struct *task, char *buffer)
498 unsigned long args[6], sp, pc;
499 int res = lock_trace(task);
503 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
504 res = sprintf(buffer, "running\n");
506 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
508 res = sprintf(buffer,
509 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
511 args[0], args[1], args[2], args[3], args[4], args[5],
516 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
518 /************************************************************************/
519 /* Here the fs part begins */
520 /************************************************************************/
522 /* permission checks */
523 static int proc_fd_access_allowed(struct inode *inode)
525 struct task_struct *task;
527 /* Allow access to a task's file descriptors if it is us or we
528 * may use ptrace attach to the process and find out that
531 task = get_proc_task(inode);
533 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
534 put_task_struct(task);
539 int proc_setattr(struct dentry *dentry, struct iattr *attr)
542 struct inode *inode = dentry->d_inode;
544 if (attr->ia_valid & ATTR_MODE)
547 error = inode_change_ok(inode, attr);
551 if ((attr->ia_valid & ATTR_SIZE) &&
552 attr->ia_size != i_size_read(inode)) {
553 error = vmtruncate(inode, attr->ia_size);
558 setattr_copy(inode, attr);
559 mark_inode_dirty(inode);
564 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
565 * or euid/egid (for hide_pid_min=2)?
567 static bool has_pid_permissions(struct pid_namespace *pid,
568 struct task_struct *task,
571 if (pid->hide_pid < hide_pid_min)
573 if (in_group_p(pid->pid_gid))
575 return ptrace_may_access(task, PTRACE_MODE_READ);
579 static int proc_pid_permission(struct inode *inode, int mask)
581 struct pid_namespace *pid = inode->i_sb->s_fs_info;
582 struct task_struct *task;
585 task = get_proc_task(inode);
588 has_perms = has_pid_permissions(pid, task, 1);
589 put_task_struct(task);
592 if (pid->hide_pid == 2) {
594 * Let's make getdents(), stat(), and open()
595 * consistent with each other. If a process
596 * may not stat() a file, it shouldn't be seen
604 return generic_permission(inode, mask);
609 static const struct inode_operations proc_def_inode_operations = {
610 .setattr = proc_setattr,
613 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
615 static ssize_t proc_info_read(struct file * file, char __user * buf,
616 size_t count, loff_t *ppos)
618 struct inode * inode = file->f_path.dentry->d_inode;
621 struct task_struct *task = get_proc_task(inode);
627 if (count > PROC_BLOCK_SIZE)
628 count = PROC_BLOCK_SIZE;
631 if (!(page = __get_free_page(GFP_TEMPORARY)))
634 length = PROC_I(inode)->op.proc_read(task, (char*)page);
637 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
640 put_task_struct(task);
645 static const struct file_operations proc_info_file_operations = {
646 .read = proc_info_read,
647 .llseek = generic_file_llseek,
650 static int proc_single_show(struct seq_file *m, void *v)
652 struct inode *inode = m->private;
653 struct pid_namespace *ns;
655 struct task_struct *task;
658 ns = inode->i_sb->s_fs_info;
659 pid = proc_pid(inode);
660 task = get_pid_task(pid, PIDTYPE_PID);
664 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
666 put_task_struct(task);
670 static int proc_single_open(struct inode *inode, struct file *filp)
672 return single_open(filp, proc_single_show, inode);
675 static const struct file_operations proc_single_file_operations = {
676 .open = proc_single_open,
679 .release = single_release,
682 static int mem_open(struct inode* inode, struct file* file)
684 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
685 struct mm_struct *mm;
690 mm = mm_access(task, PTRACE_MODE_ATTACH);
691 put_task_struct(task);
697 /* ensure this mm_struct can't be freed */
698 atomic_inc(&mm->mm_count);
699 /* but do not pin its memory */
703 /* OK to pass negative loff_t, we can catch out-of-range */
704 file->f_mode |= FMODE_UNSIGNED_OFFSET;
705 file->private_data = mm;
710 static ssize_t mem_rw(struct file *file, char __user *buf,
711 size_t count, loff_t *ppos, int write)
713 struct mm_struct *mm = file->private_data;
714 unsigned long addr = *ppos;
721 page = (char *)__get_free_page(GFP_TEMPORARY);
726 if (!atomic_inc_not_zero(&mm->mm_users))
730 int this_len = min_t(int, count, PAGE_SIZE);
732 if (write && copy_from_user(page, buf, this_len)) {
737 this_len = access_remote_vm(mm, addr, page, this_len, write);
744 if (!write && copy_to_user(buf, page, this_len)) {
758 free_page((unsigned long) page);
762 static ssize_t mem_read(struct file *file, char __user *buf,
763 size_t count, loff_t *ppos)
765 return mem_rw(file, buf, count, ppos, 0);
768 static ssize_t mem_write(struct file *file, const char __user *buf,
769 size_t count, loff_t *ppos)
771 return mem_rw(file, (char __user*)buf, count, ppos, 1);
774 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
778 file->f_pos = offset;
781 file->f_pos += offset;
786 force_successful_syscall_return();
790 static int mem_release(struct inode *inode, struct file *file)
792 struct mm_struct *mm = file->private_data;
798 static const struct file_operations proc_mem_operations = {
803 .release = mem_release,
806 static ssize_t environ_read(struct file *file, char __user *buf,
807 size_t count, loff_t *ppos)
809 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
811 unsigned long src = *ppos;
813 struct mm_struct *mm;
819 page = (char *)__get_free_page(GFP_TEMPORARY);
824 mm = mm_for_maps(task);
826 if (!mm || IS_ERR(mm))
831 int this_len, retval, max_len;
833 this_len = mm->env_end - (mm->env_start + src);
838 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
839 this_len = (this_len > max_len) ? max_len : this_len;
841 retval = access_process_vm(task, (mm->env_start + src),
849 if (copy_to_user(buf, page, retval)) {
863 free_page((unsigned long) page);
865 put_task_struct(task);
870 static const struct file_operations proc_environ_operations = {
871 .read = environ_read,
872 .llseek = generic_file_llseek,
875 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
876 size_t count, loff_t *ppos)
878 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
879 char buffer[PROC_NUMBUF];
881 int oom_adjust = OOM_DISABLE;
887 if (lock_task_sighand(task, &flags)) {
888 oom_adjust = task->signal->oom_adj;
889 unlock_task_sighand(task, &flags);
892 put_task_struct(task);
894 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
896 return simple_read_from_buffer(buf, count, ppos, buffer, len);
899 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
900 size_t count, loff_t *ppos)
902 struct task_struct *task;
903 char buffer[PROC_NUMBUF];
908 memset(buffer, 0, sizeof(buffer));
909 if (count > sizeof(buffer) - 1)
910 count = sizeof(buffer) - 1;
911 if (copy_from_user(buffer, buf, count)) {
916 err = kstrtoint(strstrip(buffer), 0, &oom_adjust);
919 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
920 oom_adjust != OOM_DISABLE) {
925 task = get_proc_task(file->f_path.dentry->d_inode);
937 if (!lock_task_sighand(task, &flags)) {
942 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
948 * Warn that /proc/pid/oom_adj is deprecated, see
949 * Documentation/feature-removal-schedule.txt.
951 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
952 current->comm, task_pid_nr(current), task_pid_nr(task),
954 task->signal->oom_adj = oom_adjust;
956 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
957 * value is always attainable.
959 if (task->signal->oom_adj == OOM_ADJUST_MAX)
960 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
962 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
964 trace_oom_score_adj_update(task);
966 unlock_task_sighand(task, &flags);
969 put_task_struct(task);
971 return err < 0 ? err : count;
974 static const struct file_operations proc_oom_adjust_operations = {
975 .read = oom_adjust_read,
976 .write = oom_adjust_write,
977 .llseek = generic_file_llseek,
980 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
981 size_t count, loff_t *ppos)
983 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
984 char buffer[PROC_NUMBUF];
985 int oom_score_adj = OOM_SCORE_ADJ_MIN;
991 if (lock_task_sighand(task, &flags)) {
992 oom_score_adj = task->signal->oom_score_adj;
993 unlock_task_sighand(task, &flags);
995 put_task_struct(task);
996 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
997 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1000 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1001 size_t count, loff_t *ppos)
1003 struct task_struct *task;
1004 char buffer[PROC_NUMBUF];
1005 unsigned long flags;
1009 memset(buffer, 0, sizeof(buffer));
1010 if (count > sizeof(buffer) - 1)
1011 count = sizeof(buffer) - 1;
1012 if (copy_from_user(buffer, buf, count)) {
1017 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1020 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1021 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1026 task = get_proc_task(file->f_path.dentry->d_inode);
1038 if (!lock_task_sighand(task, &flags)) {
1043 if (oom_score_adj < task->signal->oom_score_adj_min &&
1044 !capable(CAP_SYS_RESOURCE)) {
1049 task->signal->oom_score_adj = oom_score_adj;
1050 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1051 task->signal->oom_score_adj_min = oom_score_adj;
1052 trace_oom_score_adj_update(task);
1054 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1055 * always attainable.
1057 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1058 task->signal->oom_adj = OOM_DISABLE;
1060 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1063 unlock_task_sighand(task, &flags);
1066 put_task_struct(task);
1068 return err < 0 ? err : count;
1071 static const struct file_operations proc_oom_score_adj_operations = {
1072 .read = oom_score_adj_read,
1073 .write = oom_score_adj_write,
1074 .llseek = default_llseek,
1077 #ifdef CONFIG_AUDITSYSCALL
1078 #define TMPBUFLEN 21
1079 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1080 size_t count, loff_t *ppos)
1082 struct inode * inode = file->f_path.dentry->d_inode;
1083 struct task_struct *task = get_proc_task(inode);
1085 char tmpbuf[TMPBUFLEN];
1089 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1090 audit_get_loginuid(task));
1091 put_task_struct(task);
1092 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1095 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1096 size_t count, loff_t *ppos)
1098 struct inode * inode = file->f_path.dentry->d_inode;
1104 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1110 if (count >= PAGE_SIZE)
1111 count = PAGE_SIZE - 1;
1114 /* No partial writes. */
1117 page = (char*)__get_free_page(GFP_TEMPORARY);
1121 if (copy_from_user(page, buf, count))
1125 loginuid = simple_strtoul(page, &tmp, 10);
1131 length = audit_set_loginuid(loginuid);
1132 if (likely(length == 0))
1136 free_page((unsigned long) page);
1140 static const struct file_operations proc_loginuid_operations = {
1141 .read = proc_loginuid_read,
1142 .write = proc_loginuid_write,
1143 .llseek = generic_file_llseek,
1146 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1147 size_t count, loff_t *ppos)
1149 struct inode * inode = file->f_path.dentry->d_inode;
1150 struct task_struct *task = get_proc_task(inode);
1152 char tmpbuf[TMPBUFLEN];
1156 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1157 audit_get_sessionid(task));
1158 put_task_struct(task);
1159 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1162 static const struct file_operations proc_sessionid_operations = {
1163 .read = proc_sessionid_read,
1164 .llseek = generic_file_llseek,
1168 #ifdef CONFIG_FAULT_INJECTION
1169 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1170 size_t count, loff_t *ppos)
1172 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1173 char buffer[PROC_NUMBUF];
1179 make_it_fail = task->make_it_fail;
1180 put_task_struct(task);
1182 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1184 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1187 static ssize_t proc_fault_inject_write(struct file * file,
1188 const char __user * buf, size_t count, loff_t *ppos)
1190 struct task_struct *task;
1191 char buffer[PROC_NUMBUF], *end;
1194 if (!capable(CAP_SYS_RESOURCE))
1196 memset(buffer, 0, sizeof(buffer));
1197 if (count > sizeof(buffer) - 1)
1198 count = sizeof(buffer) - 1;
1199 if (copy_from_user(buffer, buf, count))
1201 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1204 task = get_proc_task(file->f_dentry->d_inode);
1207 task->make_it_fail = make_it_fail;
1208 put_task_struct(task);
1213 static const struct file_operations proc_fault_inject_operations = {
1214 .read = proc_fault_inject_read,
1215 .write = proc_fault_inject_write,
1216 .llseek = generic_file_llseek,
1221 #ifdef CONFIG_SCHED_DEBUG
1223 * Print out various scheduling related per-task fields:
1225 static int sched_show(struct seq_file *m, void *v)
1227 struct inode *inode = m->private;
1228 struct task_struct *p;
1230 p = get_proc_task(inode);
1233 proc_sched_show_task(p, m);
1241 sched_write(struct file *file, const char __user *buf,
1242 size_t count, loff_t *offset)
1244 struct inode *inode = file->f_path.dentry->d_inode;
1245 struct task_struct *p;
1247 p = get_proc_task(inode);
1250 proc_sched_set_task(p);
1257 static int sched_open(struct inode *inode, struct file *filp)
1259 return single_open(filp, sched_show, inode);
1262 static const struct file_operations proc_pid_sched_operations = {
1265 .write = sched_write,
1266 .llseek = seq_lseek,
1267 .release = single_release,
1272 #ifdef CONFIG_SCHED_AUTOGROUP
1274 * Print out autogroup related information:
1276 static int sched_autogroup_show(struct seq_file *m, void *v)
1278 struct inode *inode = m->private;
1279 struct task_struct *p;
1281 p = get_proc_task(inode);
1284 proc_sched_autogroup_show_task(p, m);
1292 sched_autogroup_write(struct file *file, const char __user *buf,
1293 size_t count, loff_t *offset)
1295 struct inode *inode = file->f_path.dentry->d_inode;
1296 struct task_struct *p;
1297 char buffer[PROC_NUMBUF];
1301 memset(buffer, 0, sizeof(buffer));
1302 if (count > sizeof(buffer) - 1)
1303 count = sizeof(buffer) - 1;
1304 if (copy_from_user(buffer, buf, count))
1307 err = kstrtoint(strstrip(buffer), 0, &nice);
1311 p = get_proc_task(inode);
1315 err = proc_sched_autogroup_set_nice(p, nice);
1324 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1328 ret = single_open(filp, sched_autogroup_show, NULL);
1330 struct seq_file *m = filp->private_data;
1337 static const struct file_operations proc_pid_sched_autogroup_operations = {
1338 .open = sched_autogroup_open,
1340 .write = sched_autogroup_write,
1341 .llseek = seq_lseek,
1342 .release = single_release,
1345 #endif /* CONFIG_SCHED_AUTOGROUP */
1347 static ssize_t comm_write(struct file *file, const char __user *buf,
1348 size_t count, loff_t *offset)
1350 struct inode *inode = file->f_path.dentry->d_inode;
1351 struct task_struct *p;
1352 char buffer[TASK_COMM_LEN];
1354 memset(buffer, 0, sizeof(buffer));
1355 if (count > sizeof(buffer) - 1)
1356 count = sizeof(buffer) - 1;
1357 if (copy_from_user(buffer, buf, count))
1360 p = get_proc_task(inode);
1364 if (same_thread_group(current, p))
1365 set_task_comm(p, buffer);
1374 static int comm_show(struct seq_file *m, void *v)
1376 struct inode *inode = m->private;
1377 struct task_struct *p;
1379 p = get_proc_task(inode);
1384 seq_printf(m, "%s\n", p->comm);
1392 static int comm_open(struct inode *inode, struct file *filp)
1394 return single_open(filp, comm_show, inode);
1397 static const struct file_operations proc_pid_set_comm_operations = {
1400 .write = comm_write,
1401 .llseek = seq_lseek,
1402 .release = single_release,
1405 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1407 struct task_struct *task;
1408 struct mm_struct *mm;
1409 struct file *exe_file;
1411 task = get_proc_task(dentry->d_inode);
1414 mm = get_task_mm(task);
1415 put_task_struct(task);
1418 exe_file = get_mm_exe_file(mm);
1421 *exe_path = exe_file->f_path;
1422 path_get(&exe_file->f_path);
1429 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1431 struct inode *inode = dentry->d_inode;
1432 int error = -EACCES;
1434 /* We don't need a base pointer in the /proc filesystem */
1435 path_put(&nd->path);
1437 /* Are we allowed to snoop on the tasks file descriptors? */
1438 if (!proc_fd_access_allowed(inode))
1441 error = PROC_I(inode)->op.proc_get_link(dentry, &nd->path);
1443 return ERR_PTR(error);
1446 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1448 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1455 pathname = d_path(path, tmp, PAGE_SIZE);
1456 len = PTR_ERR(pathname);
1457 if (IS_ERR(pathname))
1459 len = tmp + PAGE_SIZE - 1 - pathname;
1463 if (copy_to_user(buffer, pathname, len))
1466 free_page((unsigned long)tmp);
1470 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1472 int error = -EACCES;
1473 struct inode *inode = dentry->d_inode;
1476 /* Are we allowed to snoop on the tasks file descriptors? */
1477 if (!proc_fd_access_allowed(inode))
1480 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1484 error = do_proc_readlink(&path, buffer, buflen);
1490 static const struct inode_operations proc_pid_link_inode_operations = {
1491 .readlink = proc_pid_readlink,
1492 .follow_link = proc_pid_follow_link,
1493 .setattr = proc_setattr,
1497 /* building an inode */
1499 static int task_dumpable(struct task_struct *task)
1502 struct mm_struct *mm;
1507 dumpable = get_dumpable(mm);
1514 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1516 struct inode * inode;
1517 struct proc_inode *ei;
1518 const struct cred *cred;
1520 /* We need a new inode */
1522 inode = new_inode(sb);
1528 inode->i_ino = get_next_ino();
1529 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1530 inode->i_op = &proc_def_inode_operations;
1533 * grab the reference to task.
1535 ei->pid = get_task_pid(task, PIDTYPE_PID);
1539 if (task_dumpable(task)) {
1541 cred = __task_cred(task);
1542 inode->i_uid = cred->euid;
1543 inode->i_gid = cred->egid;
1546 security_task_to_inode(task, inode);
1556 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1558 struct inode *inode = dentry->d_inode;
1559 struct task_struct *task;
1560 const struct cred *cred;
1561 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
1563 generic_fillattr(inode, stat);
1566 stat->uid = GLOBAL_ROOT_UID;
1567 stat->gid = GLOBAL_ROOT_GID;
1568 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1570 if (!has_pid_permissions(pid, task, 2)) {
1573 * This doesn't prevent learning whether PID exists,
1574 * it only makes getattr() consistent with readdir().
1578 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1579 task_dumpable(task)) {
1580 cred = __task_cred(task);
1581 stat->uid = cred->euid;
1582 stat->gid = cred->egid;
1592 * Exceptional case: normally we are not allowed to unhash a busy
1593 * directory. In this case, however, we can do it - no aliasing problems
1594 * due to the way we treat inodes.
1596 * Rewrite the inode's ownerships here because the owning task may have
1597 * performed a setuid(), etc.
1599 * Before the /proc/pid/status file was created the only way to read
1600 * the effective uid of a /process was to stat /proc/pid. Reading
1601 * /proc/pid/status is slow enough that procps and other packages
1602 * kept stating /proc/pid. To keep the rules in /proc simple I have
1603 * made this apply to all per process world readable and executable
1606 int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1608 struct inode *inode;
1609 struct task_struct *task;
1610 const struct cred *cred;
1612 if (nd && nd->flags & LOOKUP_RCU)
1615 inode = dentry->d_inode;
1616 task = get_proc_task(inode);
1619 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1620 task_dumpable(task)) {
1622 cred = __task_cred(task);
1623 inode->i_uid = cred->euid;
1624 inode->i_gid = cred->egid;
1627 inode->i_uid = GLOBAL_ROOT_UID;
1628 inode->i_gid = GLOBAL_ROOT_GID;
1630 inode->i_mode &= ~(S_ISUID | S_ISGID);
1631 security_task_to_inode(task, inode);
1632 put_task_struct(task);
1639 static int pid_delete_dentry(const struct dentry * dentry)
1641 /* Is the task we represent dead?
1642 * If so, then don't put the dentry on the lru list,
1643 * kill it immediately.
1645 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1648 const struct dentry_operations pid_dentry_operations =
1650 .d_revalidate = pid_revalidate,
1651 .d_delete = pid_delete_dentry,
1657 * Fill a directory entry.
1659 * If possible create the dcache entry and derive our inode number and
1660 * file type from dcache entry.
1662 * Since all of the proc inode numbers are dynamically generated, the inode
1663 * numbers do not exist until the inode is cache. This means creating the
1664 * the dcache entry in readdir is necessary to keep the inode numbers
1665 * reported by readdir in sync with the inode numbers reported
1668 int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1669 const char *name, int len,
1670 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1672 struct dentry *child, *dir = filp->f_path.dentry;
1673 struct inode *inode;
1676 unsigned type = DT_UNKNOWN;
1680 qname.hash = full_name_hash(name, len);
1682 child = d_lookup(dir, &qname);
1685 new = d_alloc(dir, &qname);
1687 child = instantiate(dir->d_inode, new, task, ptr);
1694 if (!child || IS_ERR(child) || !child->d_inode)
1695 goto end_instantiate;
1696 inode = child->d_inode;
1699 type = inode->i_mode >> 12;
1704 ino = find_inode_number(dir, &qname);
1707 return filldir(dirent, name, len, filp->f_pos, ino, type);
1710 static unsigned name_to_int(struct dentry *dentry)
1712 const char *name = dentry->d_name.name;
1713 int len = dentry->d_name.len;
1716 if (len > 1 && *name == '0')
1719 unsigned c = *name++ - '0';
1722 if (n >= (~0U-9)/10)
1732 #define PROC_FDINFO_MAX 64
1734 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1736 struct task_struct *task = get_proc_task(inode);
1737 struct files_struct *files = NULL;
1739 int fd = proc_fd(inode);
1742 files = get_files_struct(task);
1743 put_task_struct(task);
1747 * We are not taking a ref to the file structure, so we must
1750 spin_lock(&files->file_lock);
1751 file = fcheck_files(files, fd);
1753 unsigned int f_flags;
1754 struct fdtable *fdt;
1756 fdt = files_fdtable(files);
1757 f_flags = file->f_flags & ~O_CLOEXEC;
1758 if (close_on_exec(fd, fdt))
1759 f_flags |= O_CLOEXEC;
1762 *path = file->f_path;
1763 path_get(&file->f_path);
1766 snprintf(info, PROC_FDINFO_MAX,
1769 (long long) file->f_pos,
1771 spin_unlock(&files->file_lock);
1772 put_files_struct(files);
1775 spin_unlock(&files->file_lock);
1776 put_files_struct(files);
1781 static int proc_fd_link(struct dentry *dentry, struct path *path)
1783 return proc_fd_info(dentry->d_inode, path, NULL);
1786 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1788 struct inode *inode;
1789 struct task_struct *task;
1791 struct files_struct *files;
1792 const struct cred *cred;
1794 if (nd && nd->flags & LOOKUP_RCU)
1797 inode = dentry->d_inode;
1798 task = get_proc_task(inode);
1799 fd = proc_fd(inode);
1802 files = get_files_struct(task);
1806 file = fcheck_files(files, fd);
1808 unsigned i_mode, f_mode = file->f_mode;
1811 put_files_struct(files);
1813 if (task_dumpable(task)) {
1815 cred = __task_cred(task);
1816 inode->i_uid = cred->euid;
1817 inode->i_gid = cred->egid;
1820 inode->i_uid = GLOBAL_ROOT_UID;
1821 inode->i_gid = GLOBAL_ROOT_GID;
1825 if (f_mode & FMODE_READ)
1826 i_mode |= S_IRUSR | S_IXUSR;
1827 if (f_mode & FMODE_WRITE)
1828 i_mode |= S_IWUSR | S_IXUSR;
1829 inode->i_mode = i_mode;
1831 security_task_to_inode(task, inode);
1832 put_task_struct(task);
1836 put_files_struct(files);
1838 put_task_struct(task);
1844 static const struct dentry_operations tid_fd_dentry_operations =
1846 .d_revalidate = tid_fd_revalidate,
1847 .d_delete = pid_delete_dentry,
1850 static struct dentry *proc_fd_instantiate(struct inode *dir,
1851 struct dentry *dentry, struct task_struct *task, const void *ptr)
1853 unsigned fd = *(const unsigned *)ptr;
1854 struct inode *inode;
1855 struct proc_inode *ei;
1856 struct dentry *error = ERR_PTR(-ENOENT);
1858 inode = proc_pid_make_inode(dir->i_sb, task);
1864 inode->i_op = &proc_pid_link_inode_operations;
1866 ei->op.proc_get_link = proc_fd_link;
1867 d_set_d_op(dentry, &tid_fd_dentry_operations);
1868 d_add(dentry, inode);
1869 /* Close the race of the process dying before we return the dentry */
1870 if (tid_fd_revalidate(dentry, NULL))
1877 static struct dentry *proc_lookupfd_common(struct inode *dir,
1878 struct dentry *dentry,
1879 instantiate_t instantiate)
1881 struct task_struct *task = get_proc_task(dir);
1882 unsigned fd = name_to_int(dentry);
1883 struct dentry *result = ERR_PTR(-ENOENT);
1890 result = instantiate(dir, dentry, task, &fd);
1892 put_task_struct(task);
1897 static int proc_readfd_common(struct file * filp, void * dirent,
1898 filldir_t filldir, instantiate_t instantiate)
1900 struct dentry *dentry = filp->f_path.dentry;
1901 struct inode *inode = dentry->d_inode;
1902 struct task_struct *p = get_proc_task(inode);
1903 unsigned int fd, ino;
1905 struct files_struct * files;
1915 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1919 ino = parent_ino(dentry);
1920 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1924 files = get_files_struct(p);
1928 for (fd = filp->f_pos-2;
1929 fd < files_fdtable(files)->max_fds;
1930 fd++, filp->f_pos++) {
1931 char name[PROC_NUMBUF];
1934 if (!fcheck_files(files, fd))
1938 len = snprintf(name, sizeof(name), "%d", fd);
1939 if (proc_fill_cache(filp, dirent, filldir,
1940 name, len, instantiate,
1948 put_files_struct(files);
1956 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1957 struct nameidata *nd)
1959 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1962 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1964 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1967 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1968 size_t len, loff_t *ppos)
1970 char tmp[PROC_FDINFO_MAX];
1971 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
1973 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1977 static const struct file_operations proc_fdinfo_file_operations = {
1978 .open = nonseekable_open,
1979 .read = proc_fdinfo_read,
1980 .llseek = no_llseek,
1983 static const struct file_operations proc_fd_operations = {
1984 .read = generic_read_dir,
1985 .readdir = proc_readfd,
1986 .llseek = default_llseek,
1989 #ifdef CONFIG_CHECKPOINT_RESTORE
1992 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1993 * which represent vma start and end addresses.
1995 static int dname_to_vma_addr(struct dentry *dentry,
1996 unsigned long *start, unsigned long *end)
1998 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
2004 static int map_files_d_revalidate(struct dentry *dentry, struct nameidata *nd)
2006 unsigned long vm_start, vm_end;
2007 bool exact_vma_exists = false;
2008 struct mm_struct *mm = NULL;
2009 struct task_struct *task;
2010 const struct cred *cred;
2011 struct inode *inode;
2014 if (nd && nd->flags & LOOKUP_RCU)
2017 if (!capable(CAP_SYS_ADMIN)) {
2022 inode = dentry->d_inode;
2023 task = get_proc_task(inode);
2027 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2030 mm = get_task_mm(task);
2034 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
2035 down_read(&mm->mmap_sem);
2036 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
2037 up_read(&mm->mmap_sem);
2042 if (exact_vma_exists) {
2043 if (task_dumpable(task)) {
2045 cred = __task_cred(task);
2046 inode->i_uid = cred->euid;
2047 inode->i_gid = cred->egid;
2050 inode->i_uid = GLOBAL_ROOT_UID;
2051 inode->i_gid = GLOBAL_ROOT_GID;
2053 security_task_to_inode(task, inode);
2058 put_task_struct(task);
2067 static const struct dentry_operations tid_map_files_dentry_operations = {
2068 .d_revalidate = map_files_d_revalidate,
2069 .d_delete = pid_delete_dentry,
2072 static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
2074 unsigned long vm_start, vm_end;
2075 struct vm_area_struct *vma;
2076 struct task_struct *task;
2077 struct mm_struct *mm;
2081 task = get_proc_task(dentry->d_inode);
2085 mm = get_task_mm(task);
2086 put_task_struct(task);
2090 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2094 down_read(&mm->mmap_sem);
2095 vma = find_exact_vma(mm, vm_start, vm_end);
2096 if (vma && vma->vm_file) {
2097 *path = vma->vm_file->f_path;
2101 up_read(&mm->mmap_sem);
2109 struct map_files_info {
2112 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
2115 static struct dentry *
2116 proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
2117 struct task_struct *task, const void *ptr)
2119 const struct file *file = ptr;
2120 struct proc_inode *ei;
2121 struct inode *inode;
2124 return ERR_PTR(-ENOENT);
2126 inode = proc_pid_make_inode(dir->i_sb, task);
2128 return ERR_PTR(-ENOENT);
2131 ei->op.proc_get_link = proc_map_files_get_link;
2133 inode->i_op = &proc_pid_link_inode_operations;
2135 inode->i_mode = S_IFLNK;
2137 if (file->f_mode & FMODE_READ)
2138 inode->i_mode |= S_IRUSR;
2139 if (file->f_mode & FMODE_WRITE)
2140 inode->i_mode |= S_IWUSR;
2142 d_set_d_op(dentry, &tid_map_files_dentry_operations);
2143 d_add(dentry, inode);
2148 static struct dentry *proc_map_files_lookup(struct inode *dir,
2149 struct dentry *dentry, struct nameidata *nd)
2151 unsigned long vm_start, vm_end;
2152 struct vm_area_struct *vma;
2153 struct task_struct *task;
2154 struct dentry *result;
2155 struct mm_struct *mm;
2157 result = ERR_PTR(-EACCES);
2158 if (!capable(CAP_SYS_ADMIN))
2161 result = ERR_PTR(-ENOENT);
2162 task = get_proc_task(dir);
2166 result = ERR_PTR(-EACCES);
2167 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2170 result = ERR_PTR(-ENOENT);
2171 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2174 mm = get_task_mm(task);
2178 down_read(&mm->mmap_sem);
2179 vma = find_exact_vma(mm, vm_start, vm_end);
2183 result = proc_map_files_instantiate(dir, dentry, task, vma->vm_file);
2186 up_read(&mm->mmap_sem);
2189 put_task_struct(task);
2194 static const struct inode_operations proc_map_files_inode_operations = {
2195 .lookup = proc_map_files_lookup,
2196 .permission = proc_fd_permission,
2197 .setattr = proc_setattr,
2201 proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
2203 struct dentry *dentry = filp->f_path.dentry;
2204 struct inode *inode = dentry->d_inode;
2205 struct vm_area_struct *vma;
2206 struct task_struct *task;
2207 struct mm_struct *mm;
2212 if (!capable(CAP_SYS_ADMIN))
2216 task = get_proc_task(inode);
2221 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2225 switch (filp->f_pos) {
2228 if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0)
2232 ino = parent_ino(dentry);
2233 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2238 unsigned long nr_files, pos, i;
2239 struct flex_array *fa = NULL;
2240 struct map_files_info info;
2241 struct map_files_info *p;
2243 mm = get_task_mm(task);
2246 down_read(&mm->mmap_sem);
2251 * We need two passes here:
2253 * 1) Collect vmas of mapped files with mmap_sem taken
2254 * 2) Release mmap_sem and instantiate entries
2256 * otherwise we get lockdep complained, since filldir()
2257 * routine might require mmap_sem taken in might_fault().
2260 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2261 if (vma->vm_file && ++pos > filp->f_pos)
2266 fa = flex_array_alloc(sizeof(info), nr_files,
2268 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2272 flex_array_free(fa);
2273 up_read(&mm->mmap_sem);
2277 for (i = 0, vma = mm->mmap, pos = 2; vma;
2278 vma = vma->vm_next) {
2281 if (++pos <= filp->f_pos)
2284 get_file(vma->vm_file);
2285 info.file = vma->vm_file;
2286 info.len = snprintf(info.name,
2287 sizeof(info.name), "%lx-%lx",
2288 vma->vm_start, vma->vm_end);
2289 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2293 up_read(&mm->mmap_sem);
2295 for (i = 0; i < nr_files; i++) {
2296 p = flex_array_get(fa, i);
2297 ret = proc_fill_cache(filp, dirent, filldir,
2299 proc_map_files_instantiate,
2306 for (; i < nr_files; i++) {
2308 * In case of error don't forget
2309 * to put rest of file refs.
2311 p = flex_array_get(fa, i);
2315 flex_array_free(fa);
2321 put_task_struct(task);
2326 static const struct file_operations proc_map_files_operations = {
2327 .read = generic_read_dir,
2328 .readdir = proc_map_files_readdir,
2329 .llseek = default_llseek,
2332 #endif /* CONFIG_CHECKPOINT_RESTORE */
2335 * /proc/pid/fd needs a special permission handler so that a process can still
2336 * access /proc/self/fd after it has executed a setuid().
2338 static int proc_fd_permission(struct inode *inode, int mask)
2340 int rv = generic_permission(inode, mask);
2343 if (task_pid(current) == proc_pid(inode))
2349 * proc directories can do almost nothing..
2351 static const struct inode_operations proc_fd_inode_operations = {
2352 .lookup = proc_lookupfd,
2353 .permission = proc_fd_permission,
2354 .setattr = proc_setattr,
2357 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2358 struct dentry *dentry, struct task_struct *task, const void *ptr)
2360 unsigned fd = *(unsigned *)ptr;
2361 struct inode *inode;
2362 struct proc_inode *ei;
2363 struct dentry *error = ERR_PTR(-ENOENT);
2365 inode = proc_pid_make_inode(dir->i_sb, task);
2370 inode->i_mode = S_IFREG | S_IRUSR;
2371 inode->i_fop = &proc_fdinfo_file_operations;
2372 d_set_d_op(dentry, &tid_fd_dentry_operations);
2373 d_add(dentry, inode);
2374 /* Close the race of the process dying before we return the dentry */
2375 if (tid_fd_revalidate(dentry, NULL))
2382 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2383 struct dentry *dentry,
2384 struct nameidata *nd)
2386 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2389 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2391 return proc_readfd_common(filp, dirent, filldir,
2392 proc_fdinfo_instantiate);
2395 static const struct file_operations proc_fdinfo_operations = {
2396 .read = generic_read_dir,
2397 .readdir = proc_readfdinfo,
2398 .llseek = default_llseek,
2402 * proc directories can do almost nothing..
2404 static const struct inode_operations proc_fdinfo_inode_operations = {
2405 .lookup = proc_lookupfdinfo,
2406 .setattr = proc_setattr,
2410 static struct dentry *proc_pident_instantiate(struct inode *dir,
2411 struct dentry *dentry, struct task_struct *task, const void *ptr)
2413 const struct pid_entry *p = ptr;
2414 struct inode *inode;
2415 struct proc_inode *ei;
2416 struct dentry *error = ERR_PTR(-ENOENT);
2418 inode = proc_pid_make_inode(dir->i_sb, task);
2423 inode->i_mode = p->mode;
2424 if (S_ISDIR(inode->i_mode))
2425 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2427 inode->i_op = p->iop;
2429 inode->i_fop = p->fop;
2431 d_set_d_op(dentry, &pid_dentry_operations);
2432 d_add(dentry, inode);
2433 /* Close the race of the process dying before we return the dentry */
2434 if (pid_revalidate(dentry, NULL))
2440 static struct dentry *proc_pident_lookup(struct inode *dir,
2441 struct dentry *dentry,
2442 const struct pid_entry *ents,
2445 struct dentry *error;
2446 struct task_struct *task = get_proc_task(dir);
2447 const struct pid_entry *p, *last;
2449 error = ERR_PTR(-ENOENT);
2455 * Yes, it does not scale. And it should not. Don't add
2456 * new entries into /proc/<tgid>/ without very good reasons.
2458 last = &ents[nents - 1];
2459 for (p = ents; p <= last; p++) {
2460 if (p->len != dentry->d_name.len)
2462 if (!memcmp(dentry->d_name.name, p->name, p->len))
2468 error = proc_pident_instantiate(dir, dentry, task, p);
2470 put_task_struct(task);
2475 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2476 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2478 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2479 proc_pident_instantiate, task, p);
2482 static int proc_pident_readdir(struct file *filp,
2483 void *dirent, filldir_t filldir,
2484 const struct pid_entry *ents, unsigned int nents)
2487 struct dentry *dentry = filp->f_path.dentry;
2488 struct inode *inode = dentry->d_inode;
2489 struct task_struct *task = get_proc_task(inode);
2490 const struct pid_entry *p, *last;
2503 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2509 ino = parent_ino(dentry);
2510 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2522 last = &ents[nents - 1];
2524 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2533 put_task_struct(task);
2538 #ifdef CONFIG_SECURITY
2539 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2540 size_t count, loff_t *ppos)
2542 struct inode * inode = file->f_path.dentry->d_inode;
2545 struct task_struct *task = get_proc_task(inode);
2550 length = security_getprocattr(task,
2551 (char*)file->f_path.dentry->d_name.name,
2553 put_task_struct(task);
2555 length = simple_read_from_buffer(buf, count, ppos, p, length);
2560 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2561 size_t count, loff_t *ppos)
2563 struct inode * inode = file->f_path.dentry->d_inode;
2566 struct task_struct *task = get_proc_task(inode);
2571 if (count > PAGE_SIZE)
2574 /* No partial writes. */
2580 page = (char*)__get_free_page(GFP_TEMPORARY);
2585 if (copy_from_user(page, buf, count))
2588 /* Guard against adverse ptrace interaction */
2589 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2593 length = security_setprocattr(task,
2594 (char*)file->f_path.dentry->d_name.name,
2595 (void*)page, count);
2596 mutex_unlock(&task->signal->cred_guard_mutex);
2598 free_page((unsigned long) page);
2600 put_task_struct(task);
2605 static const struct file_operations proc_pid_attr_operations = {
2606 .read = proc_pid_attr_read,
2607 .write = proc_pid_attr_write,
2608 .llseek = generic_file_llseek,
2611 static const struct pid_entry attr_dir_stuff[] = {
2612 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2613 REG("prev", S_IRUGO, proc_pid_attr_operations),
2614 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2615 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2616 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2617 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2620 static int proc_attr_dir_readdir(struct file * filp,
2621 void * dirent, filldir_t filldir)
2623 return proc_pident_readdir(filp,dirent,filldir,
2624 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2627 static const struct file_operations proc_attr_dir_operations = {
2628 .read = generic_read_dir,
2629 .readdir = proc_attr_dir_readdir,
2630 .llseek = default_llseek,
2633 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2634 struct dentry *dentry, struct nameidata *nd)
2636 return proc_pident_lookup(dir, dentry,
2637 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2640 static const struct inode_operations proc_attr_dir_inode_operations = {
2641 .lookup = proc_attr_dir_lookup,
2642 .getattr = pid_getattr,
2643 .setattr = proc_setattr,
2648 #ifdef CONFIG_ELF_CORE
2649 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2650 size_t count, loff_t *ppos)
2652 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2653 struct mm_struct *mm;
2654 char buffer[PROC_NUMBUF];
2662 mm = get_task_mm(task);
2664 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2665 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2666 MMF_DUMP_FILTER_SHIFT));
2668 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2671 put_task_struct(task);
2676 static ssize_t proc_coredump_filter_write(struct file *file,
2677 const char __user *buf,
2681 struct task_struct *task;
2682 struct mm_struct *mm;
2683 char buffer[PROC_NUMBUF], *end;
2690 memset(buffer, 0, sizeof(buffer));
2691 if (count > sizeof(buffer) - 1)
2692 count = sizeof(buffer) - 1;
2693 if (copy_from_user(buffer, buf, count))
2697 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2700 if (end - buffer == 0)
2704 task = get_proc_task(file->f_dentry->d_inode);
2709 mm = get_task_mm(task);
2713 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2715 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2717 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2722 put_task_struct(task);
2727 static const struct file_operations proc_coredump_filter_operations = {
2728 .read = proc_coredump_filter_read,
2729 .write = proc_coredump_filter_write,
2730 .llseek = generic_file_llseek,
2737 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2740 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2741 pid_t tgid = task_tgid_nr_ns(current, ns);
2742 char tmp[PROC_NUMBUF];
2745 sprintf(tmp, "%d", tgid);
2746 return vfs_readlink(dentry,buffer,buflen,tmp);
2749 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2751 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2752 pid_t tgid = task_tgid_nr_ns(current, ns);
2753 char *name = ERR_PTR(-ENOENT);
2757 name = ERR_PTR(-ENOMEM);
2759 sprintf(name, "%d", tgid);
2761 nd_set_link(nd, name);
2765 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2768 char *s = nd_get_link(nd);
2773 static const struct inode_operations proc_self_inode_operations = {
2774 .readlink = proc_self_readlink,
2775 .follow_link = proc_self_follow_link,
2776 .put_link = proc_self_put_link,
2782 * These are the directory entries in the root directory of /proc
2783 * that properly belong to the /proc filesystem, as they describe
2784 * describe something that is process related.
2786 static const struct pid_entry proc_base_stuff[] = {
2787 NOD("self", S_IFLNK|S_IRWXUGO,
2788 &proc_self_inode_operations, NULL, {}),
2791 static struct dentry *proc_base_instantiate(struct inode *dir,
2792 struct dentry *dentry, struct task_struct *task, const void *ptr)
2794 const struct pid_entry *p = ptr;
2795 struct inode *inode;
2796 struct proc_inode *ei;
2797 struct dentry *error;
2799 /* Allocate the inode */
2800 error = ERR_PTR(-ENOMEM);
2801 inode = new_inode(dir->i_sb);
2805 /* Initialize the inode */
2807 inode->i_ino = get_next_ino();
2808 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2811 * grab the reference to the task.
2813 ei->pid = get_task_pid(task, PIDTYPE_PID);
2817 inode->i_mode = p->mode;
2818 if (S_ISDIR(inode->i_mode))
2819 set_nlink(inode, 2);
2820 if (S_ISLNK(inode->i_mode))
2823 inode->i_op = p->iop;
2825 inode->i_fop = p->fop;
2827 d_add(dentry, inode);
2836 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2838 struct dentry *error;
2839 struct task_struct *task = get_proc_task(dir);
2840 const struct pid_entry *p, *last;
2842 error = ERR_PTR(-ENOENT);
2847 /* Lookup the directory entry */
2848 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2849 for (p = proc_base_stuff; p <= last; p++) {
2850 if (p->len != dentry->d_name.len)
2852 if (!memcmp(dentry->d_name.name, p->name, p->len))
2858 error = proc_base_instantiate(dir, dentry, task, p);
2861 put_task_struct(task);
2866 static int proc_base_fill_cache(struct file *filp, void *dirent,
2867 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2869 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2870 proc_base_instantiate, task, p);
2873 #ifdef CONFIG_TASK_IO_ACCOUNTING
2874 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2876 struct task_io_accounting acct = task->ioac;
2877 unsigned long flags;
2880 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2884 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2889 if (whole && lock_task_sighand(task, &flags)) {
2890 struct task_struct *t = task;
2892 task_io_accounting_add(&acct, &task->signal->ioac);
2893 while_each_thread(task, t)
2894 task_io_accounting_add(&acct, &t->ioac);
2896 unlock_task_sighand(task, &flags);
2898 result = sprintf(buffer,
2903 "read_bytes: %llu\n"
2904 "write_bytes: %llu\n"
2905 "cancelled_write_bytes: %llu\n",
2906 (unsigned long long)acct.rchar,
2907 (unsigned long long)acct.wchar,
2908 (unsigned long long)acct.syscr,
2909 (unsigned long long)acct.syscw,
2910 (unsigned long long)acct.read_bytes,
2911 (unsigned long long)acct.write_bytes,
2912 (unsigned long long)acct.cancelled_write_bytes);
2914 mutex_unlock(&task->signal->cred_guard_mutex);
2918 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2920 return do_io_accounting(task, buffer, 0);
2923 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2925 return do_io_accounting(task, buffer, 1);
2927 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2929 #ifdef CONFIG_USER_NS
2930 static int proc_id_map_open(struct inode *inode, struct file *file,
2931 struct seq_operations *seq_ops)
2933 struct user_namespace *ns = NULL;
2934 struct task_struct *task;
2935 struct seq_file *seq;
2938 task = get_proc_task(inode);
2941 ns = get_user_ns(task_cred_xxx(task, user_ns));
2943 put_task_struct(task);
2948 ret = seq_open(file, seq_ops);
2952 seq = file->private_data;
2962 static int proc_id_map_release(struct inode *inode, struct file *file)
2964 struct seq_file *seq = file->private_data;
2965 struct user_namespace *ns = seq->private;
2967 return seq_release(inode, file);
2970 static int proc_uid_map_open(struct inode *inode, struct file *file)
2972 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2975 static int proc_gid_map_open(struct inode *inode, struct file *file)
2977 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2980 static const struct file_operations proc_uid_map_operations = {
2981 .open = proc_uid_map_open,
2982 .write = proc_uid_map_write,
2984 .llseek = seq_lseek,
2985 .release = proc_id_map_release,
2988 static const struct file_operations proc_gid_map_operations = {
2989 .open = proc_gid_map_open,
2990 .write = proc_gid_map_write,
2992 .llseek = seq_lseek,
2993 .release = proc_id_map_release,
2995 #endif /* CONFIG_USER_NS */
2997 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2998 struct pid *pid, struct task_struct *task)
3000 int err = lock_trace(task);
3002 seq_printf(m, "%08x\n", task->personality);
3011 static const struct file_operations proc_task_operations;
3012 static const struct inode_operations proc_task_inode_operations;
3014 static const struct pid_entry tgid_base_stuff[] = {
3015 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
3016 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3017 #ifdef CONFIG_CHECKPOINT_RESTORE
3018 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
3020 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3021 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3023 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3025 REG("environ", S_IRUSR, proc_environ_operations),
3026 INF("auxv", S_IRUSR, proc_pid_auxv),
3027 ONE("status", S_IRUGO, proc_pid_status),
3028 ONE("personality", S_IRUGO, proc_pid_personality),
3029 INF("limits", S_IRUGO, proc_pid_limits),
3030 #ifdef CONFIG_SCHED_DEBUG
3031 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3033 #ifdef CONFIG_SCHED_AUTOGROUP
3034 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
3036 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3037 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3038 INF("syscall", S_IRUGO, proc_pid_syscall),
3040 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3041 ONE("stat", S_IRUGO, proc_tgid_stat),
3042 ONE("statm", S_IRUGO, proc_pid_statm),
3043 REG("maps", S_IRUGO, proc_pid_maps_operations),
3045 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
3047 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3048 LNK("cwd", proc_cwd_link),
3049 LNK("root", proc_root_link),
3050 LNK("exe", proc_exe_link),
3051 REG("mounts", S_IRUGO, proc_mounts_operations),
3052 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3053 REG("mountstats", S_IRUSR, proc_mountstats_operations),
3054 #ifdef CONFIG_PROC_PAGE_MONITOR
3055 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3056 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
3057 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3059 #ifdef CONFIG_SECURITY
3060 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3062 #ifdef CONFIG_KALLSYMS
3063 INF("wchan", S_IRUGO, proc_pid_wchan),
3065 #ifdef CONFIG_STACKTRACE
3066 ONE("stack", S_IRUGO, proc_pid_stack),
3068 #ifdef CONFIG_SCHEDSTATS
3069 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3071 #ifdef CONFIG_LATENCYTOP
3072 REG("latency", S_IRUGO, proc_lstats_operations),
3074 #ifdef CONFIG_PROC_PID_CPUSET
3075 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3077 #ifdef CONFIG_CGROUPS
3078 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3080 INF("oom_score", S_IRUGO, proc_oom_score),
3081 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3082 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3083 #ifdef CONFIG_AUDITSYSCALL
3084 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3085 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3087 #ifdef CONFIG_FAULT_INJECTION
3088 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3090 #ifdef CONFIG_ELF_CORE
3091 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3093 #ifdef CONFIG_TASK_IO_ACCOUNTING
3094 INF("io", S_IRUSR, proc_tgid_io_accounting),
3096 #ifdef CONFIG_HARDWALL
3097 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3099 #ifdef CONFIG_USER_NS
3100 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3101 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3105 static int proc_tgid_base_readdir(struct file * filp,
3106 void * dirent, filldir_t filldir)
3108 return proc_pident_readdir(filp,dirent,filldir,
3109 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
3112 static const struct file_operations proc_tgid_base_operations = {
3113 .read = generic_read_dir,
3114 .readdir = proc_tgid_base_readdir,
3115 .llseek = default_llseek,
3118 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3119 return proc_pident_lookup(dir, dentry,
3120 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3123 static const struct inode_operations proc_tgid_base_inode_operations = {
3124 .lookup = proc_tgid_base_lookup,
3125 .getattr = pid_getattr,
3126 .setattr = proc_setattr,
3127 .permission = proc_pid_permission,
3130 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
3132 struct dentry *dentry, *leader, *dir;
3133 char buf[PROC_NUMBUF];
3137 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3138 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
3140 shrink_dcache_parent(dentry);
3146 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
3147 leader = d_hash_and_lookup(mnt->mnt_root, &name);
3152 name.len = strlen(name.name);
3153 dir = d_hash_and_lookup(leader, &name);
3155 goto out_put_leader;
3158 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3159 dentry = d_hash_and_lookup(dir, &name);
3161 shrink_dcache_parent(dentry);
3174 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3175 * @task: task that should be flushed.
3177 * When flushing dentries from proc, one needs to flush them from global
3178 * proc (proc_mnt) and from all the namespaces' procs this task was seen
3179 * in. This call is supposed to do all of this job.
3181 * Looks in the dcache for
3183 * /proc/@tgid/task/@pid
3184 * if either directory is present flushes it and all of it'ts children
3187 * It is safe and reasonable to cache /proc entries for a task until
3188 * that task exits. After that they just clog up the dcache with
3189 * useless entries, possibly causing useful dcache entries to be
3190 * flushed instead. This routine is proved to flush those useless
3191 * dcache entries at process exit time.
3193 * NOTE: This routine is just an optimization so it does not guarantee
3194 * that no dcache entries will exist at process exit time it
3195 * just makes it very unlikely that any will persist.
3198 void proc_flush_task(struct task_struct *task)
3201 struct pid *pid, *tgid;
3204 pid = task_pid(task);
3205 tgid = task_tgid(task);
3207 for (i = 0; i <= pid->level; i++) {
3208 upid = &pid->numbers[i];
3209 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
3210 tgid->numbers[i].nr);
3213 upid = &pid->numbers[pid->level];
3215 pid_ns_release_proc(upid->ns);
3218 static struct dentry *proc_pid_instantiate(struct inode *dir,
3219 struct dentry * dentry,
3220 struct task_struct *task, const void *ptr)
3222 struct dentry *error = ERR_PTR(-ENOENT);
3223 struct inode *inode;
3225 inode = proc_pid_make_inode(dir->i_sb, task);
3229 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3230 inode->i_op = &proc_tgid_base_inode_operations;
3231 inode->i_fop = &proc_tgid_base_operations;
3232 inode->i_flags|=S_IMMUTABLE;
3234 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
3235 ARRAY_SIZE(tgid_base_stuff)));
3237 d_set_d_op(dentry, &pid_dentry_operations);
3239 d_add(dentry, inode);
3240 /* Close the race of the process dying before we return the dentry */
3241 if (pid_revalidate(dentry, NULL))
3247 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3249 struct dentry *result;
3250 struct task_struct *task;
3252 struct pid_namespace *ns;
3254 result = proc_base_lookup(dir, dentry);
3255 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3258 tgid = name_to_int(dentry);
3262 ns = dentry->d_sb->s_fs_info;
3264 task = find_task_by_pid_ns(tgid, ns);
3266 get_task_struct(task);
3271 result = proc_pid_instantiate(dir, dentry, task, NULL);
3272 put_task_struct(task);
3278 * Find the first task with tgid >= tgid
3283 struct task_struct *task;
3285 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3290 put_task_struct(iter.task);
3294 pid = find_ge_pid(iter.tgid, ns);
3296 iter.tgid = pid_nr_ns(pid, ns);
3297 iter.task = pid_task(pid, PIDTYPE_PID);
3298 /* What we to know is if the pid we have find is the
3299 * pid of a thread_group_leader. Testing for task
3300 * being a thread_group_leader is the obvious thing
3301 * todo but there is a window when it fails, due to
3302 * the pid transfer logic in de_thread.
3304 * So we perform the straight forward test of seeing
3305 * if the pid we have found is the pid of a thread
3306 * group leader, and don't worry if the task we have
3307 * found doesn't happen to be a thread group leader.
3308 * As we don't care in the case of readdir.
3310 if (!iter.task || !has_group_leader_pid(iter.task)) {
3314 get_task_struct(iter.task);
3320 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
3322 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3323 struct tgid_iter iter)
3325 char name[PROC_NUMBUF];
3326 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
3327 return proc_fill_cache(filp, dirent, filldir, name, len,
3328 proc_pid_instantiate, iter.task, NULL);
3331 static int fake_filldir(void *buf, const char *name, int namelen,
3332 loff_t offset, u64 ino, unsigned d_type)
3337 /* for the /proc/ directory itself, after non-process stuff has been done */
3338 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3341 struct task_struct *reaper;
3342 struct tgid_iter iter;
3343 struct pid_namespace *ns;
3344 filldir_t __filldir;
3346 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3348 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3350 reaper = get_proc_task(filp->f_path.dentry->d_inode);
3354 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
3355 const struct pid_entry *p = &proc_base_stuff[nr];
3356 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
3360 ns = filp->f_dentry->d_sb->s_fs_info;
3362 iter.tgid = filp->f_pos - TGID_OFFSET;
3363 for (iter = next_tgid(ns, iter);
3365 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3366 if (has_pid_permissions(ns, iter.task, 2))
3367 __filldir = filldir;
3369 __filldir = fake_filldir;
3371 filp->f_pos = iter.tgid + TGID_OFFSET;
3372 if (proc_pid_fill_cache(filp, dirent, __filldir, iter) < 0) {
3373 put_task_struct(iter.task);
3377 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3379 put_task_struct(reaper);
3387 static const struct pid_entry tid_base_stuff[] = {
3388 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3389 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3390 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3391 REG("environ", S_IRUSR, proc_environ_operations),
3392 INF("auxv", S_IRUSR, proc_pid_auxv),
3393 ONE("status", S_IRUGO, proc_pid_status),
3394 ONE("personality", S_IRUGO, proc_pid_personality),
3395 INF("limits", S_IRUGO, proc_pid_limits),
3396 #ifdef CONFIG_SCHED_DEBUG
3397 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3399 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3400 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3401 INF("syscall", S_IRUGO, proc_pid_syscall),
3403 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3404 ONE("stat", S_IRUGO, proc_tid_stat),
3405 ONE("statm", S_IRUGO, proc_pid_statm),
3406 REG("maps", S_IRUGO, proc_tid_maps_operations),
3408 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
3410 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3411 LNK("cwd", proc_cwd_link),
3412 LNK("root", proc_root_link),
3413 LNK("exe", proc_exe_link),
3414 REG("mounts", S_IRUGO, proc_mounts_operations),
3415 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3416 #ifdef CONFIG_PROC_PAGE_MONITOR
3417 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3418 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
3419 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3421 #ifdef CONFIG_SECURITY
3422 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3424 #ifdef CONFIG_KALLSYMS
3425 INF("wchan", S_IRUGO, proc_pid_wchan),
3427 #ifdef CONFIG_STACKTRACE
3428 ONE("stack", S_IRUGO, proc_pid_stack),
3430 #ifdef CONFIG_SCHEDSTATS
3431 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3433 #ifdef CONFIG_LATENCYTOP
3434 REG("latency", S_IRUGO, proc_lstats_operations),
3436 #ifdef CONFIG_PROC_PID_CPUSET
3437 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3439 #ifdef CONFIG_CGROUPS
3440 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3442 INF("oom_score", S_IRUGO, proc_oom_score),
3443 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3444 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3445 #ifdef CONFIG_AUDITSYSCALL
3446 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3447 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3449 #ifdef CONFIG_FAULT_INJECTION
3450 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3452 #ifdef CONFIG_TASK_IO_ACCOUNTING
3453 INF("io", S_IRUSR, proc_tid_io_accounting),
3455 #ifdef CONFIG_HARDWALL
3456 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3458 #ifdef CONFIG_USER_NS
3459 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3460 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3464 static int proc_tid_base_readdir(struct file * filp,
3465 void * dirent, filldir_t filldir)
3467 return proc_pident_readdir(filp,dirent,filldir,
3468 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3471 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3472 return proc_pident_lookup(dir, dentry,
3473 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3476 static const struct file_operations proc_tid_base_operations = {
3477 .read = generic_read_dir,
3478 .readdir = proc_tid_base_readdir,
3479 .llseek = default_llseek,
3482 static const struct inode_operations proc_tid_base_inode_operations = {
3483 .lookup = proc_tid_base_lookup,
3484 .getattr = pid_getattr,
3485 .setattr = proc_setattr,
3488 static struct dentry *proc_task_instantiate(struct inode *dir,
3489 struct dentry *dentry, struct task_struct *task, const void *ptr)
3491 struct dentry *error = ERR_PTR(-ENOENT);
3492 struct inode *inode;
3493 inode = proc_pid_make_inode(dir->i_sb, task);
3497 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3498 inode->i_op = &proc_tid_base_inode_operations;
3499 inode->i_fop = &proc_tid_base_operations;
3500 inode->i_flags|=S_IMMUTABLE;
3502 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3503 ARRAY_SIZE(tid_base_stuff)));
3505 d_set_d_op(dentry, &pid_dentry_operations);
3507 d_add(dentry, inode);
3508 /* Close the race of the process dying before we return the dentry */
3509 if (pid_revalidate(dentry, NULL))
3515 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3517 struct dentry *result = ERR_PTR(-ENOENT);
3518 struct task_struct *task;
3519 struct task_struct *leader = get_proc_task(dir);
3521 struct pid_namespace *ns;
3526 tid = name_to_int(dentry);
3530 ns = dentry->d_sb->s_fs_info;
3532 task = find_task_by_pid_ns(tid, ns);
3534 get_task_struct(task);
3538 if (!same_thread_group(leader, task))
3541 result = proc_task_instantiate(dir, dentry, task, NULL);
3543 put_task_struct(task);
3545 put_task_struct(leader);
3551 * Find the first tid of a thread group to return to user space.
3553 * Usually this is just the thread group leader, but if the users
3554 * buffer was too small or there was a seek into the middle of the
3555 * directory we have more work todo.
3557 * In the case of a short read we start with find_task_by_pid.
3559 * In the case of a seek we start with the leader and walk nr
3562 static struct task_struct *first_tid(struct task_struct *leader,
3563 int tid, int nr, struct pid_namespace *ns)
3565 struct task_struct *pos;
3568 /* Attempt to start with the pid of a thread */
3569 if (tid && (nr > 0)) {
3570 pos = find_task_by_pid_ns(tid, ns);
3571 if (pos && (pos->group_leader == leader))
3575 /* If nr exceeds the number of threads there is nothing todo */
3577 if (nr && nr >= get_nr_threads(leader))
3580 /* If we haven't found our starting place yet start
3581 * with the leader and walk nr threads forward.
3583 for (pos = leader; nr > 0; --nr) {
3584 pos = next_thread(pos);
3585 if (pos == leader) {
3591 get_task_struct(pos);
3598 * Find the next thread in the thread list.
3599 * Return NULL if there is an error or no next thread.
3601 * The reference to the input task_struct is released.
3603 static struct task_struct *next_tid(struct task_struct *start)
3605 struct task_struct *pos = NULL;
3607 if (pid_alive(start)) {
3608 pos = next_thread(start);
3609 if (thread_group_leader(pos))
3612 get_task_struct(pos);
3615 put_task_struct(start);
3619 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3620 struct task_struct *task, int tid)
3622 char name[PROC_NUMBUF];
3623 int len = snprintf(name, sizeof(name), "%d", tid);
3624 return proc_fill_cache(filp, dirent, filldir, name, len,
3625 proc_task_instantiate, task, NULL);
3628 /* for the /proc/TGID/task/ directories */
3629 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3631 struct dentry *dentry = filp->f_path.dentry;
3632 struct inode *inode = dentry->d_inode;
3633 struct task_struct *leader = NULL;
3634 struct task_struct *task;
3635 int retval = -ENOENT;
3638 struct pid_namespace *ns;
3640 task = get_proc_task(inode);
3644 if (pid_alive(task)) {
3645 leader = task->group_leader;
3646 get_task_struct(leader);
3649 put_task_struct(task);
3654 switch ((unsigned long)filp->f_pos) {
3657 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3662 ino = parent_ino(dentry);
3663 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3669 /* f_version caches the tgid value that the last readdir call couldn't
3670 * return. lseek aka telldir automagically resets f_version to 0.
3672 ns = filp->f_dentry->d_sb->s_fs_info;
3673 tid = (int)filp->f_version;
3674 filp->f_version = 0;
3675 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3677 task = next_tid(task), filp->f_pos++) {
3678 tid = task_pid_nr_ns(task, ns);
3679 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3680 /* returning this tgid failed, save it as the first
3681 * pid for the next readir call */
3682 filp->f_version = (u64)tid;
3683 put_task_struct(task);
3688 put_task_struct(leader);
3693 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3695 struct inode *inode = dentry->d_inode;
3696 struct task_struct *p = get_proc_task(inode);
3697 generic_fillattr(inode, stat);
3700 stat->nlink += get_nr_threads(p);
3707 static const struct inode_operations proc_task_inode_operations = {
3708 .lookup = proc_task_lookup,
3709 .getattr = proc_task_getattr,
3710 .setattr = proc_setattr,
3711 .permission = proc_pid_permission,
3714 static const struct file_operations proc_task_operations = {
3715 .read = generic_read_dir,
3716 .readdir = proc_task_readdir,
3717 .llseek = default_llseek,