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/printk.h>
77 #include <linux/cgroup.h>
78 #include <linux/cpuset.h>
79 #include <linux/audit.h>
80 #include <linux/poll.h>
81 #include <linux/nsproxy.h>
82 #include <linux/oom.h>
83 #include <linux/elf.h>
84 #include <linux/pid_namespace.h>
85 #include <linux/user_namespace.h>
86 #include <linux/fs_struct.h>
87 #include <linux/slab.h>
88 #include <linux/flex_array.h>
89 #include <linux/posix-timers.h>
90 #ifdef CONFIG_HARDWALL
91 #include <asm/hardwall.h>
93 #include <trace/events/oom.h>
98 * Implementing inode permission operations in /proc is almost
99 * certainly an error. Permission checks need to happen during
100 * each system call not at open time. The reason is that most of
101 * what we wish to check for permissions in /proc varies at runtime.
103 * The classic example of a problem is opening file descriptors
104 * in /proc for a task before it execs a suid executable.
111 const struct inode_operations *iop;
112 const struct file_operations *fop;
116 #define NOD(NAME, MODE, IOP, FOP, OP) { \
118 .len = sizeof(NAME) - 1, \
125 #define DIR(NAME, MODE, iops, fops) \
126 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
127 #define LNK(NAME, get_link) \
128 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
129 &proc_pid_link_inode_operations, NULL, \
130 { .proc_get_link = get_link } )
131 #define REG(NAME, MODE, fops) \
132 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
133 #define ONE(NAME, MODE, show) \
134 NOD(NAME, (S_IFREG|(MODE)), \
135 NULL, &proc_single_file_operations, \
136 { .proc_show = show } )
139 * Count the number of hardlinks for the pid_entry table, excluding the .
142 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
149 for (i = 0; i < n; ++i) {
150 if (S_ISDIR(entries[i].mode))
157 static int get_task_root(struct task_struct *task, struct path *root)
159 int result = -ENOENT;
163 get_fs_root(task->fs, root);
170 static int proc_cwd_link(struct dentry *dentry, struct path *path)
172 struct task_struct *task = get_proc_task(d_inode(dentry));
173 int result = -ENOENT;
178 get_fs_pwd(task->fs, path);
182 put_task_struct(task);
187 static int proc_root_link(struct dentry *dentry, struct path *path)
189 struct task_struct *task = get_proc_task(d_inode(dentry));
190 int result = -ENOENT;
193 result = get_task_root(task, path);
194 put_task_struct(task);
199 static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
200 size_t _count, loff_t *pos)
202 struct task_struct *tsk;
203 struct mm_struct *mm;
205 unsigned long count = _count;
206 unsigned long arg_start, arg_end, env_start, env_end;
207 unsigned long len1, len2, len;
214 tsk = get_proc_task(file_inode(file));
217 mm = get_task_mm(tsk);
218 put_task_struct(tsk);
221 /* Check if process spawned far enough to have cmdline. */
227 page = (char *)__get_free_page(GFP_TEMPORARY);
233 down_read(&mm->mmap_sem);
234 arg_start = mm->arg_start;
235 arg_end = mm->arg_end;
236 env_start = mm->env_start;
237 env_end = mm->env_end;
238 up_read(&mm->mmap_sem);
240 BUG_ON(arg_start > arg_end);
241 BUG_ON(env_start > env_end);
243 len1 = arg_end - arg_start;
244 len2 = env_end - env_start;
252 * Inherently racy -- command line shares address space
253 * with code and data.
255 rv = access_remote_vm(mm, arg_end - 1, &c, 1, 0);
262 /* Command line (set of strings) occupies whole ARGV. */
266 p = arg_start + *pos;
268 while (count > 0 && len > 0) {
272 _count = min3(count, len, PAGE_SIZE);
273 nr_read = access_remote_vm(mm, p, page, _count, 0);
279 if (copy_to_user(buf, page, nr_read)) {
292 * Command line (1 string) occupies ARGV and maybe
295 if (len1 + len2 <= *pos)
300 p = arg_start + *pos;
302 while (count > 0 && len > 0) {
303 unsigned int _count, l;
307 _count = min3(count, len, PAGE_SIZE);
308 nr_read = access_remote_vm(mm, p, page, _count, 0);
315 * Command line can be shorter than whole ARGV
316 * even if last "marker" byte says it is not.
319 l = strnlen(page, nr_read);
325 if (copy_to_user(buf, page, nr_read)) {
341 * Command line (1 string) occupies ARGV and
345 p = env_start + *pos - len1;
346 len = len1 + len2 - *pos;
351 while (count > 0 && len > 0) {
352 unsigned int _count, l;
356 _count = min3(count, len, PAGE_SIZE);
357 nr_read = access_remote_vm(mm, p, page, _count, 0);
365 l = strnlen(page, nr_read);
371 if (copy_to_user(buf, page, nr_read)) {
390 free_page((unsigned long)page);
398 static const struct file_operations proc_pid_cmdline_ops = {
399 .read = proc_pid_cmdline_read,
400 .llseek = generic_file_llseek,
403 static int proc_pid_auxv(struct seq_file *m, struct pid_namespace *ns,
404 struct pid *pid, struct task_struct *task)
406 struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ);
407 if (mm && !IS_ERR(mm)) {
408 unsigned int nwords = 0;
411 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
412 seq_write(m, mm->saved_auxv, nwords * sizeof(mm->saved_auxv[0]));
420 #ifdef CONFIG_KALLSYMS
422 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
423 * Returns the resolved symbol. If that fails, simply return the address.
425 static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
426 struct pid *pid, struct task_struct *task)
429 char symname[KSYM_NAME_LEN];
431 wchan = get_wchan(task);
433 if (wchan && ptrace_may_access(task, PTRACE_MODE_READ) && !lookup_symbol_name(wchan, symname))
434 seq_printf(m, "%s", symname);
440 #endif /* CONFIG_KALLSYMS */
442 static int lock_trace(struct task_struct *task)
444 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
447 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
448 mutex_unlock(&task->signal->cred_guard_mutex);
454 static void unlock_trace(struct task_struct *task)
456 mutex_unlock(&task->signal->cred_guard_mutex);
459 #ifdef CONFIG_STACKTRACE
461 #define MAX_STACK_TRACE_DEPTH 64
463 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
464 struct pid *pid, struct task_struct *task)
466 struct stack_trace trace;
467 unsigned long *entries;
471 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
475 trace.nr_entries = 0;
476 trace.max_entries = MAX_STACK_TRACE_DEPTH;
477 trace.entries = entries;
480 err = lock_trace(task);
482 save_stack_trace_tsk(task, &trace);
484 for (i = 0; i < trace.nr_entries; i++) {
485 seq_printf(m, "[<%pK>] %pS\n",
486 (void *)entries[i], (void *)entries[i]);
496 #ifdef CONFIG_SCHED_INFO
498 * Provides /proc/PID/schedstat
500 static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
501 struct pid *pid, struct task_struct *task)
503 if (unlikely(!sched_info_on()))
504 seq_printf(m, "0 0 0\n");
506 seq_printf(m, "%llu %llu %lu\n",
507 (unsigned long long)task->se.sum_exec_runtime,
508 (unsigned long long)task->sched_info.run_delay,
509 task->sched_info.pcount);
515 #ifdef CONFIG_LATENCYTOP
516 static int lstats_show_proc(struct seq_file *m, void *v)
519 struct inode *inode = m->private;
520 struct task_struct *task = get_proc_task(inode);
524 seq_puts(m, "Latency Top version : v0.1\n");
525 for (i = 0; i < 32; i++) {
526 struct latency_record *lr = &task->latency_record[i];
527 if (lr->backtrace[0]) {
529 seq_printf(m, "%i %li %li",
530 lr->count, lr->time, lr->max);
531 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
532 unsigned long bt = lr->backtrace[q];
537 seq_printf(m, " %ps", (void *)bt);
543 put_task_struct(task);
547 static int lstats_open(struct inode *inode, struct file *file)
549 return single_open(file, lstats_show_proc, inode);
552 static ssize_t lstats_write(struct file *file, const char __user *buf,
553 size_t count, loff_t *offs)
555 struct task_struct *task = get_proc_task(file_inode(file));
559 clear_all_latency_tracing(task);
560 put_task_struct(task);
565 static const struct file_operations proc_lstats_operations = {
568 .write = lstats_write,
570 .release = single_release,
575 static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
576 struct pid *pid, struct task_struct *task)
578 unsigned long totalpages = totalram_pages + total_swap_pages;
579 unsigned long points = 0;
581 read_lock(&tasklist_lock);
583 points = oom_badness(task, NULL, NULL, totalpages) *
585 read_unlock(&tasklist_lock);
586 seq_printf(m, "%lu\n", points);
596 static const struct limit_names lnames[RLIM_NLIMITS] = {
597 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
598 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
599 [RLIMIT_DATA] = {"Max data size", "bytes"},
600 [RLIMIT_STACK] = {"Max stack size", "bytes"},
601 [RLIMIT_CORE] = {"Max core file size", "bytes"},
602 [RLIMIT_RSS] = {"Max resident set", "bytes"},
603 [RLIMIT_NPROC] = {"Max processes", "processes"},
604 [RLIMIT_NOFILE] = {"Max open files", "files"},
605 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
606 [RLIMIT_AS] = {"Max address space", "bytes"},
607 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
608 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
609 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
610 [RLIMIT_NICE] = {"Max nice priority", NULL},
611 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
612 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
615 /* Display limits for a process */
616 static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
617 struct pid *pid, struct task_struct *task)
622 struct rlimit rlim[RLIM_NLIMITS];
624 if (!lock_task_sighand(task, &flags))
626 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
627 unlock_task_sighand(task, &flags);
630 * print the file header
632 seq_printf(m, "%-25s %-20s %-20s %-10s\n",
633 "Limit", "Soft Limit", "Hard Limit", "Units");
635 for (i = 0; i < RLIM_NLIMITS; i++) {
636 if (rlim[i].rlim_cur == RLIM_INFINITY)
637 seq_printf(m, "%-25s %-20s ",
638 lnames[i].name, "unlimited");
640 seq_printf(m, "%-25s %-20lu ",
641 lnames[i].name, rlim[i].rlim_cur);
643 if (rlim[i].rlim_max == RLIM_INFINITY)
644 seq_printf(m, "%-20s ", "unlimited");
646 seq_printf(m, "%-20lu ", rlim[i].rlim_max);
649 seq_printf(m, "%-10s\n", lnames[i].unit);
657 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
658 static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
659 struct pid *pid, struct task_struct *task)
662 unsigned long args[6], sp, pc;
665 res = lock_trace(task);
669 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
670 seq_puts(m, "running\n");
672 seq_printf(m, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
675 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
677 args[0], args[1], args[2], args[3], args[4], args[5],
683 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
685 /************************************************************************/
686 /* Here the fs part begins */
687 /************************************************************************/
689 /* permission checks */
690 static int proc_fd_access_allowed(struct inode *inode)
692 struct task_struct *task;
694 /* Allow access to a task's file descriptors if it is us or we
695 * may use ptrace attach to the process and find out that
698 task = get_proc_task(inode);
700 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
701 put_task_struct(task);
706 int proc_setattr(struct dentry *dentry, struct iattr *attr)
709 struct inode *inode = d_inode(dentry);
711 if (attr->ia_valid & ATTR_MODE)
714 error = inode_change_ok(inode, attr);
718 setattr_copy(inode, attr);
719 mark_inode_dirty(inode);
724 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
725 * or euid/egid (for hide_pid_min=2)?
727 static bool has_pid_permissions(struct pid_namespace *pid,
728 struct task_struct *task,
731 if (pid->hide_pid < hide_pid_min)
733 if (in_group_p(pid->pid_gid))
735 return ptrace_may_access(task, PTRACE_MODE_READ);
739 static int proc_pid_permission(struct inode *inode, int mask)
741 struct pid_namespace *pid = inode->i_sb->s_fs_info;
742 struct task_struct *task;
745 task = get_proc_task(inode);
748 has_perms = has_pid_permissions(pid, task, 1);
749 put_task_struct(task);
752 if (pid->hide_pid == 2) {
754 * Let's make getdents(), stat(), and open()
755 * consistent with each other. If a process
756 * may not stat() a file, it shouldn't be seen
764 return generic_permission(inode, mask);
769 static const struct inode_operations proc_def_inode_operations = {
770 .setattr = proc_setattr,
773 static int proc_single_show(struct seq_file *m, void *v)
775 struct inode *inode = m->private;
776 struct pid_namespace *ns;
778 struct task_struct *task;
781 ns = inode->i_sb->s_fs_info;
782 pid = proc_pid(inode);
783 task = get_pid_task(pid, PIDTYPE_PID);
787 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
789 put_task_struct(task);
793 static int proc_single_open(struct inode *inode, struct file *filp)
795 return single_open(filp, proc_single_show, inode);
798 static const struct file_operations proc_single_file_operations = {
799 .open = proc_single_open,
802 .release = single_release,
806 struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
808 struct task_struct *task = get_proc_task(inode);
809 struct mm_struct *mm = ERR_PTR(-ESRCH);
812 mm = mm_access(task, mode);
813 put_task_struct(task);
815 if (!IS_ERR_OR_NULL(mm)) {
816 /* ensure this mm_struct can't be freed */
817 atomic_inc(&mm->mm_count);
818 /* but do not pin its memory */
826 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
828 struct mm_struct *mm = proc_mem_open(inode, mode);
833 file->private_data = mm;
837 static int mem_open(struct inode *inode, struct file *file)
839 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
841 /* OK to pass negative loff_t, we can catch out-of-range */
842 file->f_mode |= FMODE_UNSIGNED_OFFSET;
847 static ssize_t mem_rw(struct file *file, char __user *buf,
848 size_t count, loff_t *ppos, int write)
850 struct mm_struct *mm = file->private_data;
851 unsigned long addr = *ppos;
858 page = (char *)__get_free_page(GFP_TEMPORARY);
863 if (!atomic_inc_not_zero(&mm->mm_users))
867 int this_len = min_t(int, count, PAGE_SIZE);
869 if (write && copy_from_user(page, buf, this_len)) {
874 this_len = access_remote_vm(mm, addr, page, this_len, write);
881 if (!write && copy_to_user(buf, page, this_len)) {
895 free_page((unsigned long) page);
899 static ssize_t mem_read(struct file *file, char __user *buf,
900 size_t count, loff_t *ppos)
902 return mem_rw(file, buf, count, ppos, 0);
905 static ssize_t mem_write(struct file *file, const char __user *buf,
906 size_t count, loff_t *ppos)
908 return mem_rw(file, (char __user*)buf, count, ppos, 1);
911 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
915 file->f_pos = offset;
918 file->f_pos += offset;
923 force_successful_syscall_return();
927 static int mem_release(struct inode *inode, struct file *file)
929 struct mm_struct *mm = file->private_data;
935 static const struct file_operations proc_mem_operations = {
940 .release = mem_release,
943 static int environ_open(struct inode *inode, struct file *file)
945 return __mem_open(inode, file, PTRACE_MODE_READ);
948 static ssize_t environ_read(struct file *file, char __user *buf,
949 size_t count, loff_t *ppos)
952 unsigned long src = *ppos;
954 struct mm_struct *mm = file->private_data;
959 page = (char *)__get_free_page(GFP_TEMPORARY);
964 if (!atomic_inc_not_zero(&mm->mm_users))
967 size_t this_len, max_len;
970 if (src >= (mm->env_end - mm->env_start))
973 this_len = mm->env_end - (mm->env_start + src);
975 max_len = min_t(size_t, PAGE_SIZE, count);
976 this_len = min(max_len, this_len);
978 retval = access_remote_vm(mm, (mm->env_start + src),
986 if (copy_to_user(buf, page, retval)) {
1000 free_page((unsigned long) page);
1004 static const struct file_operations proc_environ_operations = {
1005 .open = environ_open,
1006 .read = environ_read,
1007 .llseek = generic_file_llseek,
1008 .release = mem_release,
1011 static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1014 struct task_struct *task = get_proc_task(file_inode(file));
1015 char buffer[PROC_NUMBUF];
1016 int oom_adj = OOM_ADJUST_MIN;
1018 unsigned long flags;
1022 if (lock_task_sighand(task, &flags)) {
1023 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1024 oom_adj = OOM_ADJUST_MAX;
1026 oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1028 unlock_task_sighand(task, &flags);
1030 put_task_struct(task);
1031 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1032 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1035 static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1036 size_t count, loff_t *ppos)
1038 struct task_struct *task;
1039 char buffer[PROC_NUMBUF];
1041 unsigned long flags;
1044 memset(buffer, 0, sizeof(buffer));
1045 if (count > sizeof(buffer) - 1)
1046 count = sizeof(buffer) - 1;
1047 if (copy_from_user(buffer, buf, count)) {
1052 err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1055 if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1056 oom_adj != OOM_DISABLE) {
1061 task = get_proc_task(file_inode(file));
1073 if (!lock_task_sighand(task, &flags)) {
1079 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1080 * value is always attainable.
1082 if (oom_adj == OOM_ADJUST_MAX)
1083 oom_adj = OOM_SCORE_ADJ_MAX;
1085 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1087 if (oom_adj < task->signal->oom_score_adj &&
1088 !capable(CAP_SYS_RESOURCE)) {
1094 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1095 * /proc/pid/oom_score_adj instead.
1097 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1098 current->comm, task_pid_nr(current), task_pid_nr(task),
1101 task->signal->oom_score_adj = oom_adj;
1102 trace_oom_score_adj_update(task);
1104 unlock_task_sighand(task, &flags);
1107 put_task_struct(task);
1109 return err < 0 ? err : count;
1112 static const struct file_operations proc_oom_adj_operations = {
1113 .read = oom_adj_read,
1114 .write = oom_adj_write,
1115 .llseek = generic_file_llseek,
1118 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1119 size_t count, loff_t *ppos)
1121 struct task_struct *task = get_proc_task(file_inode(file));
1122 char buffer[PROC_NUMBUF];
1123 short oom_score_adj = OOM_SCORE_ADJ_MIN;
1124 unsigned long flags;
1129 if (lock_task_sighand(task, &flags)) {
1130 oom_score_adj = task->signal->oom_score_adj;
1131 unlock_task_sighand(task, &flags);
1133 put_task_struct(task);
1134 len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
1135 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1138 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1139 size_t count, loff_t *ppos)
1141 struct task_struct *task;
1142 char buffer[PROC_NUMBUF];
1143 unsigned long flags;
1147 memset(buffer, 0, sizeof(buffer));
1148 if (count > sizeof(buffer) - 1)
1149 count = sizeof(buffer) - 1;
1150 if (copy_from_user(buffer, buf, count)) {
1155 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1158 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1159 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1164 task = get_proc_task(file_inode(file));
1176 if (!lock_task_sighand(task, &flags)) {
1181 if ((short)oom_score_adj < task->signal->oom_score_adj_min &&
1182 !capable(CAP_SYS_RESOURCE)) {
1187 task->signal->oom_score_adj = (short)oom_score_adj;
1188 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1189 task->signal->oom_score_adj_min = (short)oom_score_adj;
1190 trace_oom_score_adj_update(task);
1193 unlock_task_sighand(task, &flags);
1196 put_task_struct(task);
1198 return err < 0 ? err : count;
1201 static const struct file_operations proc_oom_score_adj_operations = {
1202 .read = oom_score_adj_read,
1203 .write = oom_score_adj_write,
1204 .llseek = default_llseek,
1207 #ifdef CONFIG_AUDITSYSCALL
1208 #define TMPBUFLEN 21
1209 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1210 size_t count, loff_t *ppos)
1212 struct inode * inode = file_inode(file);
1213 struct task_struct *task = get_proc_task(inode);
1215 char tmpbuf[TMPBUFLEN];
1219 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1220 from_kuid(file->f_cred->user_ns,
1221 audit_get_loginuid(task)));
1222 put_task_struct(task);
1223 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1226 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1227 size_t count, loff_t *ppos)
1229 struct inode * inode = file_inode(file);
1235 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1242 /* No partial writes. */
1246 rv = kstrtou32_from_user(buf, count, 10, &loginuid);
1250 /* is userspace tring to explicitly UNSET the loginuid? */
1251 if (loginuid == AUDIT_UID_UNSET) {
1252 kloginuid = INVALID_UID;
1254 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1255 if (!uid_valid(kloginuid))
1259 rv = audit_set_loginuid(kloginuid);
1265 static const struct file_operations proc_loginuid_operations = {
1266 .read = proc_loginuid_read,
1267 .write = proc_loginuid_write,
1268 .llseek = generic_file_llseek,
1271 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1272 size_t count, loff_t *ppos)
1274 struct inode * inode = file_inode(file);
1275 struct task_struct *task = get_proc_task(inode);
1277 char tmpbuf[TMPBUFLEN];
1281 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1282 audit_get_sessionid(task));
1283 put_task_struct(task);
1284 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1287 static const struct file_operations proc_sessionid_operations = {
1288 .read = proc_sessionid_read,
1289 .llseek = generic_file_llseek,
1293 #ifdef CONFIG_FAULT_INJECTION
1294 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1295 size_t count, loff_t *ppos)
1297 struct task_struct *task = get_proc_task(file_inode(file));
1298 char buffer[PROC_NUMBUF];
1304 make_it_fail = task->make_it_fail;
1305 put_task_struct(task);
1307 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1309 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1312 static ssize_t proc_fault_inject_write(struct file * file,
1313 const char __user * buf, size_t count, loff_t *ppos)
1315 struct task_struct *task;
1316 char buffer[PROC_NUMBUF];
1320 if (!capable(CAP_SYS_RESOURCE))
1322 memset(buffer, 0, sizeof(buffer));
1323 if (count > sizeof(buffer) - 1)
1324 count = sizeof(buffer) - 1;
1325 if (copy_from_user(buffer, buf, count))
1327 rv = kstrtoint(strstrip(buffer), 0, &make_it_fail);
1330 if (make_it_fail < 0 || make_it_fail > 1)
1333 task = get_proc_task(file_inode(file));
1336 task->make_it_fail = make_it_fail;
1337 put_task_struct(task);
1342 static const struct file_operations proc_fault_inject_operations = {
1343 .read = proc_fault_inject_read,
1344 .write = proc_fault_inject_write,
1345 .llseek = generic_file_llseek,
1350 #ifdef CONFIG_SCHED_DEBUG
1352 * Print out various scheduling related per-task fields:
1354 static int sched_show(struct seq_file *m, void *v)
1356 struct inode *inode = m->private;
1357 struct task_struct *p;
1359 p = get_proc_task(inode);
1362 proc_sched_show_task(p, m);
1370 sched_write(struct file *file, const char __user *buf,
1371 size_t count, loff_t *offset)
1373 struct inode *inode = file_inode(file);
1374 struct task_struct *p;
1376 p = get_proc_task(inode);
1379 proc_sched_set_task(p);
1386 static int sched_open(struct inode *inode, struct file *filp)
1388 return single_open(filp, sched_show, inode);
1391 static const struct file_operations proc_pid_sched_operations = {
1394 .write = sched_write,
1395 .llseek = seq_lseek,
1396 .release = single_release,
1401 #ifdef CONFIG_SCHED_AUTOGROUP
1403 * Print out autogroup related information:
1405 static int sched_autogroup_show(struct seq_file *m, void *v)
1407 struct inode *inode = m->private;
1408 struct task_struct *p;
1410 p = get_proc_task(inode);
1413 proc_sched_autogroup_show_task(p, m);
1421 sched_autogroup_write(struct file *file, const char __user *buf,
1422 size_t count, loff_t *offset)
1424 struct inode *inode = file_inode(file);
1425 struct task_struct *p;
1426 char buffer[PROC_NUMBUF];
1430 memset(buffer, 0, sizeof(buffer));
1431 if (count > sizeof(buffer) - 1)
1432 count = sizeof(buffer) - 1;
1433 if (copy_from_user(buffer, buf, count))
1436 err = kstrtoint(strstrip(buffer), 0, &nice);
1440 p = get_proc_task(inode);
1444 err = proc_sched_autogroup_set_nice(p, nice);
1453 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1457 ret = single_open(filp, sched_autogroup_show, NULL);
1459 struct seq_file *m = filp->private_data;
1466 static const struct file_operations proc_pid_sched_autogroup_operations = {
1467 .open = sched_autogroup_open,
1469 .write = sched_autogroup_write,
1470 .llseek = seq_lseek,
1471 .release = single_release,
1474 #endif /* CONFIG_SCHED_AUTOGROUP */
1476 static ssize_t comm_write(struct file *file, const char __user *buf,
1477 size_t count, loff_t *offset)
1479 struct inode *inode = file_inode(file);
1480 struct task_struct *p;
1481 char buffer[TASK_COMM_LEN];
1482 const size_t maxlen = sizeof(buffer) - 1;
1484 memset(buffer, 0, sizeof(buffer));
1485 if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
1488 p = get_proc_task(inode);
1492 if (same_thread_group(current, p))
1493 set_task_comm(p, buffer);
1502 static int comm_show(struct seq_file *m, void *v)
1504 struct inode *inode = m->private;
1505 struct task_struct *p;
1507 p = get_proc_task(inode);
1512 seq_printf(m, "%s\n", p->comm);
1520 static int comm_open(struct inode *inode, struct file *filp)
1522 return single_open(filp, comm_show, inode);
1525 static const struct file_operations proc_pid_set_comm_operations = {
1528 .write = comm_write,
1529 .llseek = seq_lseek,
1530 .release = single_release,
1533 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1535 struct task_struct *task;
1536 struct mm_struct *mm;
1537 struct file *exe_file;
1539 task = get_proc_task(d_inode(dentry));
1542 mm = get_task_mm(task);
1543 put_task_struct(task);
1546 exe_file = get_mm_exe_file(mm);
1549 *exe_path = exe_file->f_path;
1550 path_get(&exe_file->f_path);
1557 static const char *proc_pid_follow_link(struct dentry *dentry, void **cookie)
1559 struct inode *inode = d_inode(dentry);
1561 int error = -EACCES;
1563 /* Are we allowed to snoop on the tasks file descriptors? */
1564 if (!proc_fd_access_allowed(inode))
1567 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1571 nd_jump_link(&path);
1574 return ERR_PTR(error);
1577 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1579 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1586 pathname = d_path(path, tmp, PAGE_SIZE);
1587 len = PTR_ERR(pathname);
1588 if (IS_ERR(pathname))
1590 len = tmp + PAGE_SIZE - 1 - pathname;
1594 if (copy_to_user(buffer, pathname, len))
1597 free_page((unsigned long)tmp);
1601 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1603 int error = -EACCES;
1604 struct inode *inode = d_inode(dentry);
1607 /* Are we allowed to snoop on the tasks file descriptors? */
1608 if (!proc_fd_access_allowed(inode))
1611 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1615 error = do_proc_readlink(&path, buffer, buflen);
1621 const struct inode_operations proc_pid_link_inode_operations = {
1622 .readlink = proc_pid_readlink,
1623 .follow_link = proc_pid_follow_link,
1624 .setattr = proc_setattr,
1628 /* building an inode */
1630 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1632 struct inode * inode;
1633 struct proc_inode *ei;
1634 const struct cred *cred;
1636 /* We need a new inode */
1638 inode = new_inode(sb);
1644 inode->i_ino = get_next_ino();
1645 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1646 inode->i_op = &proc_def_inode_operations;
1649 * grab the reference to task.
1651 ei->pid = get_task_pid(task, PIDTYPE_PID);
1655 if (task_dumpable(task)) {
1657 cred = __task_cred(task);
1658 inode->i_uid = cred->euid;
1659 inode->i_gid = cred->egid;
1662 security_task_to_inode(task, inode);
1672 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1674 struct inode *inode = d_inode(dentry);
1675 struct task_struct *task;
1676 const struct cred *cred;
1677 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
1679 generic_fillattr(inode, stat);
1682 stat->uid = GLOBAL_ROOT_UID;
1683 stat->gid = GLOBAL_ROOT_GID;
1684 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1686 if (!has_pid_permissions(pid, task, 2)) {
1689 * This doesn't prevent learning whether PID exists,
1690 * it only makes getattr() consistent with readdir().
1694 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1695 task_dumpable(task)) {
1696 cred = __task_cred(task);
1697 stat->uid = cred->euid;
1698 stat->gid = cred->egid;
1708 * Exceptional case: normally we are not allowed to unhash a busy
1709 * directory. In this case, however, we can do it - no aliasing problems
1710 * due to the way we treat inodes.
1712 * Rewrite the inode's ownerships here because the owning task may have
1713 * performed a setuid(), etc.
1715 * Before the /proc/pid/status file was created the only way to read
1716 * the effective uid of a /process was to stat /proc/pid. Reading
1717 * /proc/pid/status is slow enough that procps and other packages
1718 * kept stating /proc/pid. To keep the rules in /proc simple I have
1719 * made this apply to all per process world readable and executable
1722 int pid_revalidate(struct dentry *dentry, unsigned int flags)
1724 struct inode *inode;
1725 struct task_struct *task;
1726 const struct cred *cred;
1728 if (flags & LOOKUP_RCU)
1731 inode = d_inode(dentry);
1732 task = get_proc_task(inode);
1735 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1736 task_dumpable(task)) {
1738 cred = __task_cred(task);
1739 inode->i_uid = cred->euid;
1740 inode->i_gid = cred->egid;
1743 inode->i_uid = GLOBAL_ROOT_UID;
1744 inode->i_gid = GLOBAL_ROOT_GID;
1746 inode->i_mode &= ~(S_ISUID | S_ISGID);
1747 security_task_to_inode(task, inode);
1748 put_task_struct(task);
1754 static inline bool proc_inode_is_dead(struct inode *inode)
1756 return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1759 int pid_delete_dentry(const struct dentry *dentry)
1761 /* Is the task we represent dead?
1762 * If so, then don't put the dentry on the lru list,
1763 * kill it immediately.
1765 return proc_inode_is_dead(d_inode(dentry));
1768 const struct dentry_operations pid_dentry_operations =
1770 .d_revalidate = pid_revalidate,
1771 .d_delete = pid_delete_dentry,
1777 * Fill a directory entry.
1779 * If possible create the dcache entry and derive our inode number and
1780 * file type from dcache entry.
1782 * Since all of the proc inode numbers are dynamically generated, the inode
1783 * numbers do not exist until the inode is cache. This means creating the
1784 * the dcache entry in readdir is necessary to keep the inode numbers
1785 * reported by readdir in sync with the inode numbers reported
1788 bool proc_fill_cache(struct file *file, struct dir_context *ctx,
1789 const char *name, int len,
1790 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1792 struct dentry *child, *dir = file->f_path.dentry;
1793 struct qstr qname = QSTR_INIT(name, len);
1794 struct inode *inode;
1798 child = d_hash_and_lookup(dir, &qname);
1800 child = d_alloc(dir, &qname);
1802 goto end_instantiate;
1803 if (instantiate(d_inode(dir), child, task, ptr) < 0) {
1805 goto end_instantiate;
1808 inode = d_inode(child);
1810 type = inode->i_mode >> 12;
1812 return dir_emit(ctx, name, len, ino, type);
1815 return dir_emit(ctx, name, len, 1, DT_UNKNOWN);
1819 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1820 * which represent vma start and end addresses.
1822 static int dname_to_vma_addr(struct dentry *dentry,
1823 unsigned long *start, unsigned long *end)
1825 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1831 static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
1833 unsigned long vm_start, vm_end;
1834 bool exact_vma_exists = false;
1835 struct mm_struct *mm = NULL;
1836 struct task_struct *task;
1837 const struct cred *cred;
1838 struct inode *inode;
1841 if (flags & LOOKUP_RCU)
1844 inode = d_inode(dentry);
1845 task = get_proc_task(inode);
1849 mm = mm_access(task, PTRACE_MODE_READ);
1850 if (IS_ERR_OR_NULL(mm))
1853 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1854 down_read(&mm->mmap_sem);
1855 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1856 up_read(&mm->mmap_sem);
1861 if (exact_vma_exists) {
1862 if (task_dumpable(task)) {
1864 cred = __task_cred(task);
1865 inode->i_uid = cred->euid;
1866 inode->i_gid = cred->egid;
1869 inode->i_uid = GLOBAL_ROOT_UID;
1870 inode->i_gid = GLOBAL_ROOT_GID;
1872 security_task_to_inode(task, inode);
1877 put_task_struct(task);
1883 static const struct dentry_operations tid_map_files_dentry_operations = {
1884 .d_revalidate = map_files_d_revalidate,
1885 .d_delete = pid_delete_dentry,
1888 static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
1890 unsigned long vm_start, vm_end;
1891 struct vm_area_struct *vma;
1892 struct task_struct *task;
1893 struct mm_struct *mm;
1897 task = get_proc_task(d_inode(dentry));
1901 mm = get_task_mm(task);
1902 put_task_struct(task);
1906 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
1911 down_read(&mm->mmap_sem);
1912 vma = find_exact_vma(mm, vm_start, vm_end);
1913 if (vma && vma->vm_file) {
1914 *path = vma->vm_file->f_path;
1918 up_read(&mm->mmap_sem);
1926 struct map_files_info {
1929 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
1933 * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the
1934 * symlinks may be used to bypass permissions on ancestor directories in the
1935 * path to the file in question.
1938 proc_map_files_follow_link(struct dentry *dentry, void **cookie)
1940 if (!capable(CAP_SYS_ADMIN))
1941 return ERR_PTR(-EPERM);
1943 return proc_pid_follow_link(dentry, NULL);
1947 * Identical to proc_pid_link_inode_operations except for follow_link()
1949 static const struct inode_operations proc_map_files_link_inode_operations = {
1950 .readlink = proc_pid_readlink,
1951 .follow_link = proc_map_files_follow_link,
1952 .setattr = proc_setattr,
1956 proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
1957 struct task_struct *task, const void *ptr)
1959 fmode_t mode = (fmode_t)(unsigned long)ptr;
1960 struct proc_inode *ei;
1961 struct inode *inode;
1963 inode = proc_pid_make_inode(dir->i_sb, task);
1968 ei->op.proc_get_link = proc_map_files_get_link;
1970 inode->i_op = &proc_map_files_link_inode_operations;
1972 inode->i_mode = S_IFLNK;
1974 if (mode & FMODE_READ)
1975 inode->i_mode |= S_IRUSR;
1976 if (mode & FMODE_WRITE)
1977 inode->i_mode |= S_IWUSR;
1979 d_set_d_op(dentry, &tid_map_files_dentry_operations);
1980 d_add(dentry, inode);
1985 static struct dentry *proc_map_files_lookup(struct inode *dir,
1986 struct dentry *dentry, unsigned int flags)
1988 unsigned long vm_start, vm_end;
1989 struct vm_area_struct *vma;
1990 struct task_struct *task;
1992 struct mm_struct *mm;
1995 task = get_proc_task(dir);
2000 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2004 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2007 mm = get_task_mm(task);
2011 down_read(&mm->mmap_sem);
2012 vma = find_exact_vma(mm, vm_start, vm_end);
2017 result = proc_map_files_instantiate(dir, dentry, task,
2018 (void *)(unsigned long)vma->vm_file->f_mode);
2021 up_read(&mm->mmap_sem);
2024 put_task_struct(task);
2026 return ERR_PTR(result);
2029 static const struct inode_operations proc_map_files_inode_operations = {
2030 .lookup = proc_map_files_lookup,
2031 .permission = proc_fd_permission,
2032 .setattr = proc_setattr,
2036 proc_map_files_readdir(struct file *file, struct dir_context *ctx)
2038 struct vm_area_struct *vma;
2039 struct task_struct *task;
2040 struct mm_struct *mm;
2041 unsigned long nr_files, pos, i;
2042 struct flex_array *fa = NULL;
2043 struct map_files_info info;
2044 struct map_files_info *p;
2048 task = get_proc_task(file_inode(file));
2053 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2057 if (!dir_emit_dots(file, ctx))
2060 mm = get_task_mm(task);
2063 down_read(&mm->mmap_sem);
2068 * We need two passes here:
2070 * 1) Collect vmas of mapped files with mmap_sem taken
2071 * 2) Release mmap_sem and instantiate entries
2073 * otherwise we get lockdep complained, since filldir()
2074 * routine might require mmap_sem taken in might_fault().
2077 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2078 if (vma->vm_file && ++pos > ctx->pos)
2083 fa = flex_array_alloc(sizeof(info), nr_files,
2085 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2089 flex_array_free(fa);
2090 up_read(&mm->mmap_sem);
2094 for (i = 0, vma = mm->mmap, pos = 2; vma;
2095 vma = vma->vm_next) {
2098 if (++pos <= ctx->pos)
2101 info.mode = vma->vm_file->f_mode;
2102 info.len = snprintf(info.name,
2103 sizeof(info.name), "%lx-%lx",
2104 vma->vm_start, vma->vm_end);
2105 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2109 up_read(&mm->mmap_sem);
2111 for (i = 0; i < nr_files; i++) {
2112 p = flex_array_get(fa, i);
2113 if (!proc_fill_cache(file, ctx,
2115 proc_map_files_instantiate,
2117 (void *)(unsigned long)p->mode))
2122 flex_array_free(fa);
2126 put_task_struct(task);
2131 static const struct file_operations proc_map_files_operations = {
2132 .read = generic_read_dir,
2133 .iterate = proc_map_files_readdir,
2134 .llseek = default_llseek,
2137 struct timers_private {
2139 struct task_struct *task;
2140 struct sighand_struct *sighand;
2141 struct pid_namespace *ns;
2142 unsigned long flags;
2145 static void *timers_start(struct seq_file *m, loff_t *pos)
2147 struct timers_private *tp = m->private;
2149 tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2151 return ERR_PTR(-ESRCH);
2153 tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2155 return ERR_PTR(-ESRCH);
2157 return seq_list_start(&tp->task->signal->posix_timers, *pos);
2160 static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2162 struct timers_private *tp = m->private;
2163 return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2166 static void timers_stop(struct seq_file *m, void *v)
2168 struct timers_private *tp = m->private;
2171 unlock_task_sighand(tp->task, &tp->flags);
2176 put_task_struct(tp->task);
2181 static int show_timer(struct seq_file *m, void *v)
2183 struct k_itimer *timer;
2184 struct timers_private *tp = m->private;
2186 static const char * const nstr[] = {
2187 [SIGEV_SIGNAL] = "signal",
2188 [SIGEV_NONE] = "none",
2189 [SIGEV_THREAD] = "thread",
2192 timer = list_entry((struct list_head *)v, struct k_itimer, list);
2193 notify = timer->it_sigev_notify;
2195 seq_printf(m, "ID: %d\n", timer->it_id);
2196 seq_printf(m, "signal: %d/%p\n",
2197 timer->sigq->info.si_signo,
2198 timer->sigq->info.si_value.sival_ptr);
2199 seq_printf(m, "notify: %s/%s.%d\n",
2200 nstr[notify & ~SIGEV_THREAD_ID],
2201 (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2202 pid_nr_ns(timer->it_pid, tp->ns));
2203 seq_printf(m, "ClockID: %d\n", timer->it_clock);
2208 static const struct seq_operations proc_timers_seq_ops = {
2209 .start = timers_start,
2210 .next = timers_next,
2211 .stop = timers_stop,
2215 static int proc_timers_open(struct inode *inode, struct file *file)
2217 struct timers_private *tp;
2219 tp = __seq_open_private(file, &proc_timers_seq_ops,
2220 sizeof(struct timers_private));
2224 tp->pid = proc_pid(inode);
2225 tp->ns = inode->i_sb->s_fs_info;
2229 static const struct file_operations proc_timers_operations = {
2230 .open = proc_timers_open,
2232 .llseek = seq_lseek,
2233 .release = seq_release_private,
2236 static int proc_pident_instantiate(struct inode *dir,
2237 struct dentry *dentry, struct task_struct *task, const void *ptr)
2239 const struct pid_entry *p = ptr;
2240 struct inode *inode;
2241 struct proc_inode *ei;
2243 inode = proc_pid_make_inode(dir->i_sb, task);
2248 inode->i_mode = p->mode;
2249 if (S_ISDIR(inode->i_mode))
2250 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2252 inode->i_op = p->iop;
2254 inode->i_fop = p->fop;
2256 d_set_d_op(dentry, &pid_dentry_operations);
2257 d_add(dentry, inode);
2258 /* Close the race of the process dying before we return the dentry */
2259 if (pid_revalidate(dentry, 0))
2265 static struct dentry *proc_pident_lookup(struct inode *dir,
2266 struct dentry *dentry,
2267 const struct pid_entry *ents,
2271 struct task_struct *task = get_proc_task(dir);
2272 const struct pid_entry *p, *last;
2280 * Yes, it does not scale. And it should not. Don't add
2281 * new entries into /proc/<tgid>/ without very good reasons.
2283 last = &ents[nents - 1];
2284 for (p = ents; p <= last; p++) {
2285 if (p->len != dentry->d_name.len)
2287 if (!memcmp(dentry->d_name.name, p->name, p->len))
2293 error = proc_pident_instantiate(dir, dentry, task, p);
2295 put_task_struct(task);
2297 return ERR_PTR(error);
2300 static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
2301 const struct pid_entry *ents, unsigned int nents)
2303 struct task_struct *task = get_proc_task(file_inode(file));
2304 const struct pid_entry *p;
2309 if (!dir_emit_dots(file, ctx))
2312 if (ctx->pos >= nents + 2)
2315 for (p = ents + (ctx->pos - 2); p <= ents + nents - 1; p++) {
2316 if (!proc_fill_cache(file, ctx, p->name, p->len,
2317 proc_pident_instantiate, task, p))
2322 put_task_struct(task);
2326 #ifdef CONFIG_SECURITY
2327 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2328 size_t count, loff_t *ppos)
2330 struct inode * inode = file_inode(file);
2333 struct task_struct *task = get_proc_task(inode);
2338 length = security_getprocattr(task,
2339 (char*)file->f_path.dentry->d_name.name,
2341 put_task_struct(task);
2343 length = simple_read_from_buffer(buf, count, ppos, p, length);
2348 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2349 size_t count, loff_t *ppos)
2351 struct inode * inode = file_inode(file);
2354 struct task_struct *task = get_proc_task(inode);
2359 if (count > PAGE_SIZE)
2362 /* No partial writes. */
2368 page = (char*)__get_free_page(GFP_TEMPORARY);
2373 if (copy_from_user(page, buf, count))
2376 /* Guard against adverse ptrace interaction */
2377 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2381 length = security_setprocattr(task,
2382 (char*)file->f_path.dentry->d_name.name,
2383 (void*)page, count);
2384 mutex_unlock(&task->signal->cred_guard_mutex);
2386 free_page((unsigned long) page);
2388 put_task_struct(task);
2393 static const struct file_operations proc_pid_attr_operations = {
2394 .read = proc_pid_attr_read,
2395 .write = proc_pid_attr_write,
2396 .llseek = generic_file_llseek,
2399 static const struct pid_entry attr_dir_stuff[] = {
2400 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2401 REG("prev", S_IRUGO, proc_pid_attr_operations),
2402 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2403 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2404 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2405 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2408 static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
2410 return proc_pident_readdir(file, ctx,
2411 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2414 static const struct file_operations proc_attr_dir_operations = {
2415 .read = generic_read_dir,
2416 .iterate = proc_attr_dir_readdir,
2417 .llseek = default_llseek,
2420 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2421 struct dentry *dentry, unsigned int flags)
2423 return proc_pident_lookup(dir, dentry,
2424 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2427 static const struct inode_operations proc_attr_dir_inode_operations = {
2428 .lookup = proc_attr_dir_lookup,
2429 .getattr = pid_getattr,
2430 .setattr = proc_setattr,
2435 #ifdef CONFIG_ELF_CORE
2436 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2437 size_t count, loff_t *ppos)
2439 struct task_struct *task = get_proc_task(file_inode(file));
2440 struct mm_struct *mm;
2441 char buffer[PROC_NUMBUF];
2449 mm = get_task_mm(task);
2451 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2452 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2453 MMF_DUMP_FILTER_SHIFT));
2455 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2458 put_task_struct(task);
2463 static ssize_t proc_coredump_filter_write(struct file *file,
2464 const char __user *buf,
2468 struct task_struct *task;
2469 struct mm_struct *mm;
2475 ret = kstrtouint_from_user(buf, count, 0, &val);
2480 task = get_proc_task(file_inode(file));
2484 mm = get_task_mm(task);
2488 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2490 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2492 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2497 put_task_struct(task);
2504 static const struct file_operations proc_coredump_filter_operations = {
2505 .read = proc_coredump_filter_read,
2506 .write = proc_coredump_filter_write,
2507 .llseek = generic_file_llseek,
2511 #ifdef CONFIG_TASK_IO_ACCOUNTING
2512 static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
2514 struct task_io_accounting acct = task->ioac;
2515 unsigned long flags;
2518 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2522 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2527 if (whole && lock_task_sighand(task, &flags)) {
2528 struct task_struct *t = task;
2530 task_io_accounting_add(&acct, &task->signal->ioac);
2531 while_each_thread(task, t)
2532 task_io_accounting_add(&acct, &t->ioac);
2534 unlock_task_sighand(task, &flags);
2541 "read_bytes: %llu\n"
2542 "write_bytes: %llu\n"
2543 "cancelled_write_bytes: %llu\n",
2544 (unsigned long long)acct.rchar,
2545 (unsigned long long)acct.wchar,
2546 (unsigned long long)acct.syscr,
2547 (unsigned long long)acct.syscw,
2548 (unsigned long long)acct.read_bytes,
2549 (unsigned long long)acct.write_bytes,
2550 (unsigned long long)acct.cancelled_write_bytes);
2554 mutex_unlock(&task->signal->cred_guard_mutex);
2558 static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2559 struct pid *pid, struct task_struct *task)
2561 return do_io_accounting(task, m, 0);
2564 static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2565 struct pid *pid, struct task_struct *task)
2567 return do_io_accounting(task, m, 1);
2569 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2571 #ifdef CONFIG_USER_NS
2572 static int proc_id_map_open(struct inode *inode, struct file *file,
2573 const struct seq_operations *seq_ops)
2575 struct user_namespace *ns = NULL;
2576 struct task_struct *task;
2577 struct seq_file *seq;
2580 task = get_proc_task(inode);
2583 ns = get_user_ns(task_cred_xxx(task, user_ns));
2585 put_task_struct(task);
2590 ret = seq_open(file, seq_ops);
2594 seq = file->private_data;
2604 static int proc_id_map_release(struct inode *inode, struct file *file)
2606 struct seq_file *seq = file->private_data;
2607 struct user_namespace *ns = seq->private;
2609 return seq_release(inode, file);
2612 static int proc_uid_map_open(struct inode *inode, struct file *file)
2614 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2617 static int proc_gid_map_open(struct inode *inode, struct file *file)
2619 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2622 static int proc_projid_map_open(struct inode *inode, struct file *file)
2624 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2627 static const struct file_operations proc_uid_map_operations = {
2628 .open = proc_uid_map_open,
2629 .write = proc_uid_map_write,
2631 .llseek = seq_lseek,
2632 .release = proc_id_map_release,
2635 static const struct file_operations proc_gid_map_operations = {
2636 .open = proc_gid_map_open,
2637 .write = proc_gid_map_write,
2639 .llseek = seq_lseek,
2640 .release = proc_id_map_release,
2643 static const struct file_operations proc_projid_map_operations = {
2644 .open = proc_projid_map_open,
2645 .write = proc_projid_map_write,
2647 .llseek = seq_lseek,
2648 .release = proc_id_map_release,
2651 static int proc_setgroups_open(struct inode *inode, struct file *file)
2653 struct user_namespace *ns = NULL;
2654 struct task_struct *task;
2658 task = get_proc_task(inode);
2661 ns = get_user_ns(task_cred_xxx(task, user_ns));
2663 put_task_struct(task);
2668 if (file->f_mode & FMODE_WRITE) {
2670 if (!ns_capable(ns, CAP_SYS_ADMIN))
2674 ret = single_open(file, &proc_setgroups_show, ns);
2685 static int proc_setgroups_release(struct inode *inode, struct file *file)
2687 struct seq_file *seq = file->private_data;
2688 struct user_namespace *ns = seq->private;
2689 int ret = single_release(inode, file);
2694 static const struct file_operations proc_setgroups_operations = {
2695 .open = proc_setgroups_open,
2696 .write = proc_setgroups_write,
2698 .llseek = seq_lseek,
2699 .release = proc_setgroups_release,
2701 #endif /* CONFIG_USER_NS */
2703 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2704 struct pid *pid, struct task_struct *task)
2706 int err = lock_trace(task);
2708 seq_printf(m, "%08x\n", task->personality);
2717 static const struct file_operations proc_task_operations;
2718 static const struct inode_operations proc_task_inode_operations;
2720 static const struct pid_entry tgid_base_stuff[] = {
2721 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2722 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2723 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
2724 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2725 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2727 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2729 REG("environ", S_IRUSR, proc_environ_operations),
2730 ONE("auxv", S_IRUSR, proc_pid_auxv),
2731 ONE("status", S_IRUGO, proc_pid_status),
2732 ONE("personality", S_IRUSR, proc_pid_personality),
2733 ONE("limits", S_IRUGO, proc_pid_limits),
2734 #ifdef CONFIG_SCHED_DEBUG
2735 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2737 #ifdef CONFIG_SCHED_AUTOGROUP
2738 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2740 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2741 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2742 ONE("syscall", S_IRUSR, proc_pid_syscall),
2744 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
2745 ONE("stat", S_IRUGO, proc_tgid_stat),
2746 ONE("statm", S_IRUGO, proc_pid_statm),
2747 REG("maps", S_IRUGO, proc_pid_maps_operations),
2749 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
2751 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2752 LNK("cwd", proc_cwd_link),
2753 LNK("root", proc_root_link),
2754 LNK("exe", proc_exe_link),
2755 REG("mounts", S_IRUGO, proc_mounts_operations),
2756 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2757 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2758 #ifdef CONFIG_PROC_PAGE_MONITOR
2759 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2760 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
2761 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2763 #ifdef CONFIG_SECURITY
2764 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2766 #ifdef CONFIG_KALLSYMS
2767 ONE("wchan", S_IRUGO, proc_pid_wchan),
2769 #ifdef CONFIG_STACKTRACE
2770 ONE("stack", S_IRUSR, proc_pid_stack),
2772 #ifdef CONFIG_SCHED_INFO
2773 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
2775 #ifdef CONFIG_LATENCYTOP
2776 REG("latency", S_IRUGO, proc_lstats_operations),
2778 #ifdef CONFIG_PROC_PID_CPUSET
2779 ONE("cpuset", S_IRUGO, proc_cpuset_show),
2781 #ifdef CONFIG_CGROUPS
2782 ONE("cgroup", S_IRUGO, proc_cgroup_show),
2784 ONE("oom_score", S_IRUGO, proc_oom_score),
2785 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
2786 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2787 #ifdef CONFIG_AUDITSYSCALL
2788 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2789 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2791 #ifdef CONFIG_FAULT_INJECTION
2792 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2794 #ifdef CONFIG_ELF_CORE
2795 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2797 #ifdef CONFIG_TASK_IO_ACCOUNTING
2798 ONE("io", S_IRUSR, proc_tgid_io_accounting),
2800 #ifdef CONFIG_HARDWALL
2801 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
2803 #ifdef CONFIG_USER_NS
2804 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
2805 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
2806 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
2807 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
2809 #ifdef CONFIG_CHECKPOINT_RESTORE
2810 REG("timers", S_IRUGO, proc_timers_operations),
2814 static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
2816 return proc_pident_readdir(file, ctx,
2817 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2820 static const struct file_operations proc_tgid_base_operations = {
2821 .read = generic_read_dir,
2822 .iterate = proc_tgid_base_readdir,
2823 .llseek = default_llseek,
2826 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2828 return proc_pident_lookup(dir, dentry,
2829 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2832 static const struct inode_operations proc_tgid_base_inode_operations = {
2833 .lookup = proc_tgid_base_lookup,
2834 .getattr = pid_getattr,
2835 .setattr = proc_setattr,
2836 .permission = proc_pid_permission,
2839 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2841 struct dentry *dentry, *leader, *dir;
2842 char buf[PROC_NUMBUF];
2846 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2847 /* no ->d_hash() rejects on procfs */
2848 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2850 d_invalidate(dentry);
2858 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2859 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2864 name.len = strlen(name.name);
2865 dir = d_hash_and_lookup(leader, &name);
2867 goto out_put_leader;
2870 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2871 dentry = d_hash_and_lookup(dir, &name);
2873 d_invalidate(dentry);
2885 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2886 * @task: task that should be flushed.
2888 * When flushing dentries from proc, one needs to flush them from global
2889 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2890 * in. This call is supposed to do all of this job.
2892 * Looks in the dcache for
2894 * /proc/@tgid/task/@pid
2895 * if either directory is present flushes it and all of it'ts children
2898 * It is safe and reasonable to cache /proc entries for a task until
2899 * that task exits. After that they just clog up the dcache with
2900 * useless entries, possibly causing useful dcache entries to be
2901 * flushed instead. This routine is proved to flush those useless
2902 * dcache entries at process exit time.
2904 * NOTE: This routine is just an optimization so it does not guarantee
2905 * that no dcache entries will exist at process exit time it
2906 * just makes it very unlikely that any will persist.
2909 void proc_flush_task(struct task_struct *task)
2912 struct pid *pid, *tgid;
2915 pid = task_pid(task);
2916 tgid = task_tgid(task);
2918 for (i = 0; i <= pid->level; i++) {
2919 upid = &pid->numbers[i];
2920 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2921 tgid->numbers[i].nr);
2925 static int proc_pid_instantiate(struct inode *dir,
2926 struct dentry * dentry,
2927 struct task_struct *task, const void *ptr)
2929 struct inode *inode;
2931 inode = proc_pid_make_inode(dir->i_sb, task);
2935 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2936 inode->i_op = &proc_tgid_base_inode_operations;
2937 inode->i_fop = &proc_tgid_base_operations;
2938 inode->i_flags|=S_IMMUTABLE;
2940 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
2941 ARRAY_SIZE(tgid_base_stuff)));
2943 d_set_d_op(dentry, &pid_dentry_operations);
2945 d_add(dentry, inode);
2946 /* Close the race of the process dying before we return the dentry */
2947 if (pid_revalidate(dentry, 0))
2953 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
2955 int result = -ENOENT;
2956 struct task_struct *task;
2958 struct pid_namespace *ns;
2960 tgid = name_to_int(&dentry->d_name);
2964 ns = dentry->d_sb->s_fs_info;
2966 task = find_task_by_pid_ns(tgid, ns);
2968 get_task_struct(task);
2973 result = proc_pid_instantiate(dir, dentry, task, NULL);
2974 put_task_struct(task);
2976 return ERR_PTR(result);
2980 * Find the first task with tgid >= tgid
2985 struct task_struct *task;
2987 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2992 put_task_struct(iter.task);
2996 pid = find_ge_pid(iter.tgid, ns);
2998 iter.tgid = pid_nr_ns(pid, ns);
2999 iter.task = pid_task(pid, PIDTYPE_PID);
3000 /* What we to know is if the pid we have find is the
3001 * pid of a thread_group_leader. Testing for task
3002 * being a thread_group_leader is the obvious thing
3003 * todo but there is a window when it fails, due to
3004 * the pid transfer logic in de_thread.
3006 * So we perform the straight forward test of seeing
3007 * if the pid we have found is the pid of a thread
3008 * group leader, and don't worry if the task we have
3009 * found doesn't happen to be a thread group leader.
3010 * As we don't care in the case of readdir.
3012 if (!iter.task || !has_group_leader_pid(iter.task)) {
3016 get_task_struct(iter.task);
3022 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3024 /* for the /proc/ directory itself, after non-process stuff has been done */
3025 int proc_pid_readdir(struct file *file, struct dir_context *ctx)
3027 struct tgid_iter iter;
3028 struct pid_namespace *ns = file_inode(file)->i_sb->s_fs_info;
3029 loff_t pos = ctx->pos;
3031 if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
3034 if (pos == TGID_OFFSET - 2) {
3035 struct inode *inode = d_inode(ns->proc_self);
3036 if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
3038 ctx->pos = pos = pos + 1;
3040 if (pos == TGID_OFFSET - 1) {
3041 struct inode *inode = d_inode(ns->proc_thread_self);
3042 if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3044 ctx->pos = pos = pos + 1;
3046 iter.tgid = pos - TGID_OFFSET;
3048 for (iter = next_tgid(ns, iter);
3050 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3051 char name[PROC_NUMBUF];
3053 if (!has_pid_permissions(ns, iter.task, 2))
3056 len = snprintf(name, sizeof(name), "%d", iter.tgid);
3057 ctx->pos = iter.tgid + TGID_OFFSET;
3058 if (!proc_fill_cache(file, ctx, name, len,
3059 proc_pid_instantiate, iter.task, NULL)) {
3060 put_task_struct(iter.task);
3064 ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
3071 static const struct pid_entry tid_base_stuff[] = {
3072 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3073 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3074 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3076 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3078 REG("environ", S_IRUSR, proc_environ_operations),
3079 ONE("auxv", S_IRUSR, proc_pid_auxv),
3080 ONE("status", S_IRUGO, proc_pid_status),
3081 ONE("personality", S_IRUSR, proc_pid_personality),
3082 ONE("limits", S_IRUGO, proc_pid_limits),
3083 #ifdef CONFIG_SCHED_DEBUG
3084 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3086 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3087 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3088 ONE("syscall", S_IRUSR, proc_pid_syscall),
3090 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
3091 ONE("stat", S_IRUGO, proc_tid_stat),
3092 ONE("statm", S_IRUGO, proc_pid_statm),
3093 REG("maps", S_IRUGO, proc_tid_maps_operations),
3094 #ifdef CONFIG_PROC_CHILDREN
3095 REG("children", S_IRUGO, proc_tid_children_operations),
3098 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
3100 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3101 LNK("cwd", proc_cwd_link),
3102 LNK("root", proc_root_link),
3103 LNK("exe", proc_exe_link),
3104 REG("mounts", S_IRUGO, proc_mounts_operations),
3105 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3106 #ifdef CONFIG_PROC_PAGE_MONITOR
3107 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3108 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
3109 REG("pagemap", S_IRUSR, proc_pagemap_operations),
3111 #ifdef CONFIG_SECURITY
3112 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3114 #ifdef CONFIG_KALLSYMS
3115 ONE("wchan", S_IRUGO, proc_pid_wchan),
3117 #ifdef CONFIG_STACKTRACE
3118 ONE("stack", S_IRUSR, proc_pid_stack),
3120 #ifdef CONFIG_SCHED_INFO
3121 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
3123 #ifdef CONFIG_LATENCYTOP
3124 REG("latency", S_IRUGO, proc_lstats_operations),
3126 #ifdef CONFIG_PROC_PID_CPUSET
3127 ONE("cpuset", S_IRUGO, proc_cpuset_show),
3129 #ifdef CONFIG_CGROUPS
3130 ONE("cgroup", S_IRUGO, proc_cgroup_show),
3132 ONE("oom_score", S_IRUGO, proc_oom_score),
3133 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3134 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3135 #ifdef CONFIG_AUDITSYSCALL
3136 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3137 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3139 #ifdef CONFIG_FAULT_INJECTION
3140 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3142 #ifdef CONFIG_TASK_IO_ACCOUNTING
3143 ONE("io", S_IRUSR, proc_tid_io_accounting),
3145 #ifdef CONFIG_HARDWALL
3146 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
3148 #ifdef CONFIG_USER_NS
3149 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3150 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3151 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3152 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
3156 static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
3158 return proc_pident_readdir(file, ctx,
3159 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3162 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3164 return proc_pident_lookup(dir, dentry,
3165 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3168 static const struct file_operations proc_tid_base_operations = {
3169 .read = generic_read_dir,
3170 .iterate = proc_tid_base_readdir,
3171 .llseek = default_llseek,
3174 static const struct inode_operations proc_tid_base_inode_operations = {
3175 .lookup = proc_tid_base_lookup,
3176 .getattr = pid_getattr,
3177 .setattr = proc_setattr,
3180 static int proc_task_instantiate(struct inode *dir,
3181 struct dentry *dentry, struct task_struct *task, const void *ptr)
3183 struct inode *inode;
3184 inode = proc_pid_make_inode(dir->i_sb, task);
3188 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3189 inode->i_op = &proc_tid_base_inode_operations;
3190 inode->i_fop = &proc_tid_base_operations;
3191 inode->i_flags|=S_IMMUTABLE;
3193 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3194 ARRAY_SIZE(tid_base_stuff)));
3196 d_set_d_op(dentry, &pid_dentry_operations);
3198 d_add(dentry, inode);
3199 /* Close the race of the process dying before we return the dentry */
3200 if (pid_revalidate(dentry, 0))
3206 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3208 int result = -ENOENT;
3209 struct task_struct *task;
3210 struct task_struct *leader = get_proc_task(dir);
3212 struct pid_namespace *ns;
3217 tid = name_to_int(&dentry->d_name);
3221 ns = dentry->d_sb->s_fs_info;
3223 task = find_task_by_pid_ns(tid, ns);
3225 get_task_struct(task);
3229 if (!same_thread_group(leader, task))
3232 result = proc_task_instantiate(dir, dentry, task, NULL);
3234 put_task_struct(task);
3236 put_task_struct(leader);
3238 return ERR_PTR(result);
3242 * Find the first tid of a thread group to return to user space.
3244 * Usually this is just the thread group leader, but if the users
3245 * buffer was too small or there was a seek into the middle of the
3246 * directory we have more work todo.
3248 * In the case of a short read we start with find_task_by_pid.
3250 * In the case of a seek we start with the leader and walk nr
3253 static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3254 struct pid_namespace *ns)
3256 struct task_struct *pos, *task;
3257 unsigned long nr = f_pos;
3259 if (nr != f_pos) /* 32bit overflow? */
3263 task = pid_task(pid, PIDTYPE_PID);
3267 /* Attempt to start with the tid of a thread */
3269 pos = find_task_by_pid_ns(tid, ns);
3270 if (pos && same_thread_group(pos, task))
3274 /* If nr exceeds the number of threads there is nothing todo */
3275 if (nr >= get_nr_threads(task))
3278 /* If we haven't found our starting place yet start
3279 * with the leader and walk nr threads forward.
3281 pos = task = task->group_leader;
3285 } while_each_thread(task, pos);
3290 get_task_struct(pos);
3297 * Find the next thread in the thread list.
3298 * Return NULL if there is an error or no next thread.
3300 * The reference to the input task_struct is released.
3302 static struct task_struct *next_tid(struct task_struct *start)
3304 struct task_struct *pos = NULL;
3306 if (pid_alive(start)) {
3307 pos = next_thread(start);
3308 if (thread_group_leader(pos))
3311 get_task_struct(pos);
3314 put_task_struct(start);
3318 /* for the /proc/TGID/task/ directories */
3319 static int proc_task_readdir(struct file *file, struct dir_context *ctx)
3321 struct inode *inode = file_inode(file);
3322 struct task_struct *task;
3323 struct pid_namespace *ns;
3326 if (proc_inode_is_dead(inode))
3329 if (!dir_emit_dots(file, ctx))
3332 /* f_version caches the tgid value that the last readdir call couldn't
3333 * return. lseek aka telldir automagically resets f_version to 0.
3335 ns = inode->i_sb->s_fs_info;
3336 tid = (int)file->f_version;
3337 file->f_version = 0;
3338 for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
3340 task = next_tid(task), ctx->pos++) {
3341 char name[PROC_NUMBUF];
3343 tid = task_pid_nr_ns(task, ns);
3344 len = snprintf(name, sizeof(name), "%d", tid);
3345 if (!proc_fill_cache(file, ctx, name, len,
3346 proc_task_instantiate, task, NULL)) {
3347 /* returning this tgid failed, save it as the first
3348 * pid for the next readir call */
3349 file->f_version = (u64)tid;
3350 put_task_struct(task);
3358 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3360 struct inode *inode = d_inode(dentry);
3361 struct task_struct *p = get_proc_task(inode);
3362 generic_fillattr(inode, stat);
3365 stat->nlink += get_nr_threads(p);
3372 static const struct inode_operations proc_task_inode_operations = {
3373 .lookup = proc_task_lookup,
3374 .getattr = proc_task_getattr,
3375 .setattr = proc_setattr,
3376 .permission = proc_pid_permission,
3379 static const struct file_operations proc_task_operations = {
3380 .read = generic_read_dir,
3381 .iterate = proc_task_readdir,
3382 .llseek = default_llseek,