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 (lookup_symbol_name(wchan, symname) < 0) {
434 if (!ptrace_may_access(task, PTRACE_MODE_READ))
436 seq_printf(m, "%lu", wchan);
438 seq_printf(m, "%s", symname);
443 #endif /* CONFIG_KALLSYMS */
445 static int lock_trace(struct task_struct *task)
447 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
450 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
451 mutex_unlock(&task->signal->cred_guard_mutex);
457 static void unlock_trace(struct task_struct *task)
459 mutex_unlock(&task->signal->cred_guard_mutex);
462 #ifdef CONFIG_STACKTRACE
464 #define MAX_STACK_TRACE_DEPTH 64
466 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
467 struct pid *pid, struct task_struct *task)
469 struct stack_trace trace;
470 unsigned long *entries;
474 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
478 trace.nr_entries = 0;
479 trace.max_entries = MAX_STACK_TRACE_DEPTH;
480 trace.entries = entries;
483 err = lock_trace(task);
485 save_stack_trace_tsk(task, &trace);
487 for (i = 0; i < trace.nr_entries; i++) {
488 seq_printf(m, "[<%pK>] %pS\n",
489 (void *)entries[i], (void *)entries[i]);
499 #ifdef CONFIG_SCHED_INFO
501 * Provides /proc/PID/schedstat
503 static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
504 struct pid *pid, struct task_struct *task)
506 if (unlikely(!sched_info_on()))
507 seq_printf(m, "0 0 0\n");
509 seq_printf(m, "%llu %llu %lu\n",
510 (unsigned long long)task->se.sum_exec_runtime,
511 (unsigned long long)task->sched_info.run_delay,
512 task->sched_info.pcount);
518 #ifdef CONFIG_LATENCYTOP
519 static int lstats_show_proc(struct seq_file *m, void *v)
522 struct inode *inode = m->private;
523 struct task_struct *task = get_proc_task(inode);
527 seq_puts(m, "Latency Top version : v0.1\n");
528 for (i = 0; i < 32; i++) {
529 struct latency_record *lr = &task->latency_record[i];
530 if (lr->backtrace[0]) {
532 seq_printf(m, "%i %li %li",
533 lr->count, lr->time, lr->max);
534 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
535 unsigned long bt = lr->backtrace[q];
540 seq_printf(m, " %ps", (void *)bt);
546 put_task_struct(task);
550 static int lstats_open(struct inode *inode, struct file *file)
552 return single_open(file, lstats_show_proc, inode);
555 static ssize_t lstats_write(struct file *file, const char __user *buf,
556 size_t count, loff_t *offs)
558 struct task_struct *task = get_proc_task(file_inode(file));
562 clear_all_latency_tracing(task);
563 put_task_struct(task);
568 static const struct file_operations proc_lstats_operations = {
571 .write = lstats_write,
573 .release = single_release,
578 static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
579 struct pid *pid, struct task_struct *task)
581 unsigned long totalpages = totalram_pages + total_swap_pages;
582 unsigned long points = 0;
584 read_lock(&tasklist_lock);
586 points = oom_badness(task, NULL, NULL, totalpages) *
588 read_unlock(&tasklist_lock);
589 seq_printf(m, "%lu\n", points);
599 static const struct limit_names lnames[RLIM_NLIMITS] = {
600 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
601 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
602 [RLIMIT_DATA] = {"Max data size", "bytes"},
603 [RLIMIT_STACK] = {"Max stack size", "bytes"},
604 [RLIMIT_CORE] = {"Max core file size", "bytes"},
605 [RLIMIT_RSS] = {"Max resident set", "bytes"},
606 [RLIMIT_NPROC] = {"Max processes", "processes"},
607 [RLIMIT_NOFILE] = {"Max open files", "files"},
608 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
609 [RLIMIT_AS] = {"Max address space", "bytes"},
610 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
611 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
612 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
613 [RLIMIT_NICE] = {"Max nice priority", NULL},
614 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
615 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
618 /* Display limits for a process */
619 static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
620 struct pid *pid, struct task_struct *task)
625 struct rlimit rlim[RLIM_NLIMITS];
627 if (!lock_task_sighand(task, &flags))
629 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
630 unlock_task_sighand(task, &flags);
633 * print the file header
635 seq_printf(m, "%-25s %-20s %-20s %-10s\n",
636 "Limit", "Soft Limit", "Hard Limit", "Units");
638 for (i = 0; i < RLIM_NLIMITS; i++) {
639 if (rlim[i].rlim_cur == RLIM_INFINITY)
640 seq_printf(m, "%-25s %-20s ",
641 lnames[i].name, "unlimited");
643 seq_printf(m, "%-25s %-20lu ",
644 lnames[i].name, rlim[i].rlim_cur);
646 if (rlim[i].rlim_max == RLIM_INFINITY)
647 seq_printf(m, "%-20s ", "unlimited");
649 seq_printf(m, "%-20lu ", rlim[i].rlim_max);
652 seq_printf(m, "%-10s\n", lnames[i].unit);
660 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
661 static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
662 struct pid *pid, struct task_struct *task)
665 unsigned long args[6], sp, pc;
668 res = lock_trace(task);
672 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
673 seq_puts(m, "running\n");
675 seq_printf(m, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
678 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
680 args[0], args[1], args[2], args[3], args[4], args[5],
686 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
688 /************************************************************************/
689 /* Here the fs part begins */
690 /************************************************************************/
692 /* permission checks */
693 static int proc_fd_access_allowed(struct inode *inode)
695 struct task_struct *task;
697 /* Allow access to a task's file descriptors if it is us or we
698 * may use ptrace attach to the process and find out that
701 task = get_proc_task(inode);
703 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
704 put_task_struct(task);
709 int proc_setattr(struct dentry *dentry, struct iattr *attr)
712 struct inode *inode = d_inode(dentry);
714 if (attr->ia_valid & ATTR_MODE)
717 error = inode_change_ok(inode, attr);
721 setattr_copy(inode, attr);
722 mark_inode_dirty(inode);
727 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
728 * or euid/egid (for hide_pid_min=2)?
730 static bool has_pid_permissions(struct pid_namespace *pid,
731 struct task_struct *task,
734 if (pid->hide_pid < hide_pid_min)
736 if (in_group_p(pid->pid_gid))
738 return ptrace_may_access(task, PTRACE_MODE_READ);
742 static int proc_pid_permission(struct inode *inode, int mask)
744 struct pid_namespace *pid = inode->i_sb->s_fs_info;
745 struct task_struct *task;
748 task = get_proc_task(inode);
751 has_perms = has_pid_permissions(pid, task, 1);
752 put_task_struct(task);
755 if (pid->hide_pid == 2) {
757 * Let's make getdents(), stat(), and open()
758 * consistent with each other. If a process
759 * may not stat() a file, it shouldn't be seen
767 return generic_permission(inode, mask);
772 static const struct inode_operations proc_def_inode_operations = {
773 .setattr = proc_setattr,
776 static int proc_single_show(struct seq_file *m, void *v)
778 struct inode *inode = m->private;
779 struct pid_namespace *ns;
781 struct task_struct *task;
784 ns = inode->i_sb->s_fs_info;
785 pid = proc_pid(inode);
786 task = get_pid_task(pid, PIDTYPE_PID);
790 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
792 put_task_struct(task);
796 static int proc_single_open(struct inode *inode, struct file *filp)
798 return single_open(filp, proc_single_show, inode);
801 static const struct file_operations proc_single_file_operations = {
802 .open = proc_single_open,
805 .release = single_release,
809 struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
811 struct task_struct *task = get_proc_task(inode);
812 struct mm_struct *mm = ERR_PTR(-ESRCH);
815 mm = mm_access(task, mode);
816 put_task_struct(task);
818 if (!IS_ERR_OR_NULL(mm)) {
819 /* ensure this mm_struct can't be freed */
820 atomic_inc(&mm->mm_count);
821 /* but do not pin its memory */
829 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
831 struct mm_struct *mm = proc_mem_open(inode, mode);
836 file->private_data = mm;
840 static int mem_open(struct inode *inode, struct file *file)
842 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
844 /* OK to pass negative loff_t, we can catch out-of-range */
845 file->f_mode |= FMODE_UNSIGNED_OFFSET;
850 static ssize_t mem_rw(struct file *file, char __user *buf,
851 size_t count, loff_t *ppos, int write)
853 struct mm_struct *mm = file->private_data;
854 unsigned long addr = *ppos;
861 page = (char *)__get_free_page(GFP_TEMPORARY);
866 if (!atomic_inc_not_zero(&mm->mm_users))
870 int this_len = min_t(int, count, PAGE_SIZE);
872 if (write && copy_from_user(page, buf, this_len)) {
877 this_len = access_remote_vm(mm, addr, page, this_len, write);
884 if (!write && copy_to_user(buf, page, this_len)) {
898 free_page((unsigned long) page);
902 static ssize_t mem_read(struct file *file, char __user *buf,
903 size_t count, loff_t *ppos)
905 return mem_rw(file, buf, count, ppos, 0);
908 static ssize_t mem_write(struct file *file, const char __user *buf,
909 size_t count, loff_t *ppos)
911 return mem_rw(file, (char __user*)buf, count, ppos, 1);
914 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
918 file->f_pos = offset;
921 file->f_pos += offset;
926 force_successful_syscall_return();
930 static int mem_release(struct inode *inode, struct file *file)
932 struct mm_struct *mm = file->private_data;
938 static const struct file_operations proc_mem_operations = {
943 .release = mem_release,
946 static int environ_open(struct inode *inode, struct file *file)
948 return __mem_open(inode, file, PTRACE_MODE_READ);
951 static ssize_t environ_read(struct file *file, char __user *buf,
952 size_t count, loff_t *ppos)
955 unsigned long src = *ppos;
957 struct mm_struct *mm = file->private_data;
962 page = (char *)__get_free_page(GFP_TEMPORARY);
967 if (!atomic_inc_not_zero(&mm->mm_users))
970 size_t this_len, max_len;
973 if (src >= (mm->env_end - mm->env_start))
976 this_len = mm->env_end - (mm->env_start + src);
978 max_len = min_t(size_t, PAGE_SIZE, count);
979 this_len = min(max_len, this_len);
981 retval = access_remote_vm(mm, (mm->env_start + src),
989 if (copy_to_user(buf, page, retval)) {
1003 free_page((unsigned long) page);
1007 static const struct file_operations proc_environ_operations = {
1008 .open = environ_open,
1009 .read = environ_read,
1010 .llseek = generic_file_llseek,
1011 .release = mem_release,
1014 static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1017 struct task_struct *task = get_proc_task(file_inode(file));
1018 char buffer[PROC_NUMBUF];
1019 int oom_adj = OOM_ADJUST_MIN;
1021 unsigned long flags;
1025 if (lock_task_sighand(task, &flags)) {
1026 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1027 oom_adj = OOM_ADJUST_MAX;
1029 oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1031 unlock_task_sighand(task, &flags);
1033 put_task_struct(task);
1034 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1035 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1038 static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1039 size_t count, loff_t *ppos)
1041 struct task_struct *task;
1042 char buffer[PROC_NUMBUF];
1044 unsigned long flags;
1047 memset(buffer, 0, sizeof(buffer));
1048 if (count > sizeof(buffer) - 1)
1049 count = sizeof(buffer) - 1;
1050 if (copy_from_user(buffer, buf, count)) {
1055 err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1058 if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1059 oom_adj != OOM_DISABLE) {
1064 task = get_proc_task(file_inode(file));
1076 if (!lock_task_sighand(task, &flags)) {
1082 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1083 * value is always attainable.
1085 if (oom_adj == OOM_ADJUST_MAX)
1086 oom_adj = OOM_SCORE_ADJ_MAX;
1088 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1090 if (oom_adj < task->signal->oom_score_adj &&
1091 !capable(CAP_SYS_RESOURCE)) {
1097 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1098 * /proc/pid/oom_score_adj instead.
1100 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1101 current->comm, task_pid_nr(current), task_pid_nr(task),
1104 task->signal->oom_score_adj = oom_adj;
1105 trace_oom_score_adj_update(task);
1107 unlock_task_sighand(task, &flags);
1110 put_task_struct(task);
1112 return err < 0 ? err : count;
1115 static const struct file_operations proc_oom_adj_operations = {
1116 .read = oom_adj_read,
1117 .write = oom_adj_write,
1118 .llseek = generic_file_llseek,
1121 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1122 size_t count, loff_t *ppos)
1124 struct task_struct *task = get_proc_task(file_inode(file));
1125 char buffer[PROC_NUMBUF];
1126 short oom_score_adj = OOM_SCORE_ADJ_MIN;
1127 unsigned long flags;
1132 if (lock_task_sighand(task, &flags)) {
1133 oom_score_adj = task->signal->oom_score_adj;
1134 unlock_task_sighand(task, &flags);
1136 put_task_struct(task);
1137 len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
1138 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1141 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1142 size_t count, loff_t *ppos)
1144 struct task_struct *task;
1145 char buffer[PROC_NUMBUF];
1146 unsigned long flags;
1150 memset(buffer, 0, sizeof(buffer));
1151 if (count > sizeof(buffer) - 1)
1152 count = sizeof(buffer) - 1;
1153 if (copy_from_user(buffer, buf, count)) {
1158 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1161 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1162 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1167 task = get_proc_task(file_inode(file));
1179 if (!lock_task_sighand(task, &flags)) {
1184 if ((short)oom_score_adj < task->signal->oom_score_adj_min &&
1185 !capable(CAP_SYS_RESOURCE)) {
1190 task->signal->oom_score_adj = (short)oom_score_adj;
1191 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1192 task->signal->oom_score_adj_min = (short)oom_score_adj;
1193 trace_oom_score_adj_update(task);
1196 unlock_task_sighand(task, &flags);
1199 put_task_struct(task);
1201 return err < 0 ? err : count;
1204 static const struct file_operations proc_oom_score_adj_operations = {
1205 .read = oom_score_adj_read,
1206 .write = oom_score_adj_write,
1207 .llseek = default_llseek,
1210 #ifdef CONFIG_AUDITSYSCALL
1211 #define TMPBUFLEN 21
1212 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1213 size_t count, loff_t *ppos)
1215 struct inode * inode = file_inode(file);
1216 struct task_struct *task = get_proc_task(inode);
1218 char tmpbuf[TMPBUFLEN];
1222 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1223 from_kuid(file->f_cred->user_ns,
1224 audit_get_loginuid(task)));
1225 put_task_struct(task);
1226 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1229 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1230 size_t count, loff_t *ppos)
1232 struct inode * inode = file_inode(file);
1238 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1245 /* No partial writes. */
1249 rv = kstrtou32_from_user(buf, count, 10, &loginuid);
1253 /* is userspace tring to explicitly UNSET the loginuid? */
1254 if (loginuid == AUDIT_UID_UNSET) {
1255 kloginuid = INVALID_UID;
1257 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1258 if (!uid_valid(kloginuid))
1262 rv = audit_set_loginuid(kloginuid);
1268 static const struct file_operations proc_loginuid_operations = {
1269 .read = proc_loginuid_read,
1270 .write = proc_loginuid_write,
1271 .llseek = generic_file_llseek,
1274 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1275 size_t count, loff_t *ppos)
1277 struct inode * inode = file_inode(file);
1278 struct task_struct *task = get_proc_task(inode);
1280 char tmpbuf[TMPBUFLEN];
1284 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1285 audit_get_sessionid(task));
1286 put_task_struct(task);
1287 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1290 static const struct file_operations proc_sessionid_operations = {
1291 .read = proc_sessionid_read,
1292 .llseek = generic_file_llseek,
1296 #ifdef CONFIG_FAULT_INJECTION
1297 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1298 size_t count, loff_t *ppos)
1300 struct task_struct *task = get_proc_task(file_inode(file));
1301 char buffer[PROC_NUMBUF];
1307 make_it_fail = task->make_it_fail;
1308 put_task_struct(task);
1310 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1312 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1315 static ssize_t proc_fault_inject_write(struct file * file,
1316 const char __user * buf, size_t count, loff_t *ppos)
1318 struct task_struct *task;
1319 char buffer[PROC_NUMBUF];
1323 if (!capable(CAP_SYS_RESOURCE))
1325 memset(buffer, 0, sizeof(buffer));
1326 if (count > sizeof(buffer) - 1)
1327 count = sizeof(buffer) - 1;
1328 if (copy_from_user(buffer, buf, count))
1330 rv = kstrtoint(strstrip(buffer), 0, &make_it_fail);
1333 if (make_it_fail < 0 || make_it_fail > 1)
1336 task = get_proc_task(file_inode(file));
1339 task->make_it_fail = make_it_fail;
1340 put_task_struct(task);
1345 static const struct file_operations proc_fault_inject_operations = {
1346 .read = proc_fault_inject_read,
1347 .write = proc_fault_inject_write,
1348 .llseek = generic_file_llseek,
1353 #ifdef CONFIG_SCHED_DEBUG
1355 * Print out various scheduling related per-task fields:
1357 static int sched_show(struct seq_file *m, void *v)
1359 struct inode *inode = m->private;
1360 struct task_struct *p;
1362 p = get_proc_task(inode);
1365 proc_sched_show_task(p, m);
1373 sched_write(struct file *file, const char __user *buf,
1374 size_t count, loff_t *offset)
1376 struct inode *inode = file_inode(file);
1377 struct task_struct *p;
1379 p = get_proc_task(inode);
1382 proc_sched_set_task(p);
1389 static int sched_open(struct inode *inode, struct file *filp)
1391 return single_open(filp, sched_show, inode);
1394 static const struct file_operations proc_pid_sched_operations = {
1397 .write = sched_write,
1398 .llseek = seq_lseek,
1399 .release = single_release,
1404 #ifdef CONFIG_SCHED_AUTOGROUP
1406 * Print out autogroup related information:
1408 static int sched_autogroup_show(struct seq_file *m, void *v)
1410 struct inode *inode = m->private;
1411 struct task_struct *p;
1413 p = get_proc_task(inode);
1416 proc_sched_autogroup_show_task(p, m);
1424 sched_autogroup_write(struct file *file, const char __user *buf,
1425 size_t count, loff_t *offset)
1427 struct inode *inode = file_inode(file);
1428 struct task_struct *p;
1429 char buffer[PROC_NUMBUF];
1433 memset(buffer, 0, sizeof(buffer));
1434 if (count > sizeof(buffer) - 1)
1435 count = sizeof(buffer) - 1;
1436 if (copy_from_user(buffer, buf, count))
1439 err = kstrtoint(strstrip(buffer), 0, &nice);
1443 p = get_proc_task(inode);
1447 err = proc_sched_autogroup_set_nice(p, nice);
1456 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1460 ret = single_open(filp, sched_autogroup_show, NULL);
1462 struct seq_file *m = filp->private_data;
1469 static const struct file_operations proc_pid_sched_autogroup_operations = {
1470 .open = sched_autogroup_open,
1472 .write = sched_autogroup_write,
1473 .llseek = seq_lseek,
1474 .release = single_release,
1477 #endif /* CONFIG_SCHED_AUTOGROUP */
1479 static ssize_t comm_write(struct file *file, const char __user *buf,
1480 size_t count, loff_t *offset)
1482 struct inode *inode = file_inode(file);
1483 struct task_struct *p;
1484 char buffer[TASK_COMM_LEN];
1485 const size_t maxlen = sizeof(buffer) - 1;
1487 memset(buffer, 0, sizeof(buffer));
1488 if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
1491 p = get_proc_task(inode);
1495 if (same_thread_group(current, p))
1496 set_task_comm(p, buffer);
1505 static int comm_show(struct seq_file *m, void *v)
1507 struct inode *inode = m->private;
1508 struct task_struct *p;
1510 p = get_proc_task(inode);
1515 seq_printf(m, "%s\n", p->comm);
1523 static int comm_open(struct inode *inode, struct file *filp)
1525 return single_open(filp, comm_show, inode);
1528 static const struct file_operations proc_pid_set_comm_operations = {
1531 .write = comm_write,
1532 .llseek = seq_lseek,
1533 .release = single_release,
1536 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1538 struct task_struct *task;
1539 struct mm_struct *mm;
1540 struct file *exe_file;
1542 task = get_proc_task(d_inode(dentry));
1545 mm = get_task_mm(task);
1546 put_task_struct(task);
1549 exe_file = get_mm_exe_file(mm);
1552 *exe_path = exe_file->f_path;
1553 path_get(&exe_file->f_path);
1560 static const char *proc_pid_follow_link(struct dentry *dentry, void **cookie)
1562 struct inode *inode = d_inode(dentry);
1564 int error = -EACCES;
1566 /* Are we allowed to snoop on the tasks file descriptors? */
1567 if (!proc_fd_access_allowed(inode))
1570 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1574 nd_jump_link(&path);
1577 return ERR_PTR(error);
1580 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1582 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1589 pathname = d_path(path, tmp, PAGE_SIZE);
1590 len = PTR_ERR(pathname);
1591 if (IS_ERR(pathname))
1593 len = tmp + PAGE_SIZE - 1 - pathname;
1597 if (copy_to_user(buffer, pathname, len))
1600 free_page((unsigned long)tmp);
1604 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1606 int error = -EACCES;
1607 struct inode *inode = d_inode(dentry);
1610 /* Are we allowed to snoop on the tasks file descriptors? */
1611 if (!proc_fd_access_allowed(inode))
1614 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1618 error = do_proc_readlink(&path, buffer, buflen);
1624 const struct inode_operations proc_pid_link_inode_operations = {
1625 .readlink = proc_pid_readlink,
1626 .follow_link = proc_pid_follow_link,
1627 .setattr = proc_setattr,
1631 /* building an inode */
1633 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1635 struct inode * inode;
1636 struct proc_inode *ei;
1637 const struct cred *cred;
1639 /* We need a new inode */
1641 inode = new_inode(sb);
1647 inode->i_ino = get_next_ino();
1648 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1649 inode->i_op = &proc_def_inode_operations;
1652 * grab the reference to task.
1654 ei->pid = get_task_pid(task, PIDTYPE_PID);
1658 if (task_dumpable(task)) {
1660 cred = __task_cred(task);
1661 inode->i_uid = cred->euid;
1662 inode->i_gid = cred->egid;
1665 security_task_to_inode(task, inode);
1675 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1677 struct inode *inode = d_inode(dentry);
1678 struct task_struct *task;
1679 const struct cred *cred;
1680 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
1682 generic_fillattr(inode, stat);
1685 stat->uid = GLOBAL_ROOT_UID;
1686 stat->gid = GLOBAL_ROOT_GID;
1687 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1689 if (!has_pid_permissions(pid, task, 2)) {
1692 * This doesn't prevent learning whether PID exists,
1693 * it only makes getattr() consistent with readdir().
1697 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1698 task_dumpable(task)) {
1699 cred = __task_cred(task);
1700 stat->uid = cred->euid;
1701 stat->gid = cred->egid;
1711 * Exceptional case: normally we are not allowed to unhash a busy
1712 * directory. In this case, however, we can do it - no aliasing problems
1713 * due to the way we treat inodes.
1715 * Rewrite the inode's ownerships here because the owning task may have
1716 * performed a setuid(), etc.
1718 * Before the /proc/pid/status file was created the only way to read
1719 * the effective uid of a /process was to stat /proc/pid. Reading
1720 * /proc/pid/status is slow enough that procps and other packages
1721 * kept stating /proc/pid. To keep the rules in /proc simple I have
1722 * made this apply to all per process world readable and executable
1725 int pid_revalidate(struct dentry *dentry, unsigned int flags)
1727 struct inode *inode;
1728 struct task_struct *task;
1729 const struct cred *cred;
1731 if (flags & LOOKUP_RCU)
1734 inode = d_inode(dentry);
1735 task = get_proc_task(inode);
1738 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1739 task_dumpable(task)) {
1741 cred = __task_cred(task);
1742 inode->i_uid = cred->euid;
1743 inode->i_gid = cred->egid;
1746 inode->i_uid = GLOBAL_ROOT_UID;
1747 inode->i_gid = GLOBAL_ROOT_GID;
1749 inode->i_mode &= ~(S_ISUID | S_ISGID);
1750 security_task_to_inode(task, inode);
1751 put_task_struct(task);
1757 static inline bool proc_inode_is_dead(struct inode *inode)
1759 return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1762 int pid_delete_dentry(const struct dentry *dentry)
1764 /* Is the task we represent dead?
1765 * If so, then don't put the dentry on the lru list,
1766 * kill it immediately.
1768 return proc_inode_is_dead(d_inode(dentry));
1771 const struct dentry_operations pid_dentry_operations =
1773 .d_revalidate = pid_revalidate,
1774 .d_delete = pid_delete_dentry,
1780 * Fill a directory entry.
1782 * If possible create the dcache entry and derive our inode number and
1783 * file type from dcache entry.
1785 * Since all of the proc inode numbers are dynamically generated, the inode
1786 * numbers do not exist until the inode is cache. This means creating the
1787 * the dcache entry in readdir is necessary to keep the inode numbers
1788 * reported by readdir in sync with the inode numbers reported
1791 bool proc_fill_cache(struct file *file, struct dir_context *ctx,
1792 const char *name, int len,
1793 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1795 struct dentry *child, *dir = file->f_path.dentry;
1796 struct qstr qname = QSTR_INIT(name, len);
1797 struct inode *inode;
1801 child = d_hash_and_lookup(dir, &qname);
1803 child = d_alloc(dir, &qname);
1805 goto end_instantiate;
1806 if (instantiate(d_inode(dir), child, task, ptr) < 0) {
1808 goto end_instantiate;
1811 inode = d_inode(child);
1813 type = inode->i_mode >> 12;
1815 return dir_emit(ctx, name, len, ino, type);
1818 return dir_emit(ctx, name, len, 1, DT_UNKNOWN);
1822 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1823 * which represent vma start and end addresses.
1825 static int dname_to_vma_addr(struct dentry *dentry,
1826 unsigned long *start, unsigned long *end)
1828 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1834 static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
1836 unsigned long vm_start, vm_end;
1837 bool exact_vma_exists = false;
1838 struct mm_struct *mm = NULL;
1839 struct task_struct *task;
1840 const struct cred *cred;
1841 struct inode *inode;
1844 if (flags & LOOKUP_RCU)
1847 inode = d_inode(dentry);
1848 task = get_proc_task(inode);
1852 mm = mm_access(task, PTRACE_MODE_READ);
1853 if (IS_ERR_OR_NULL(mm))
1856 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1857 down_read(&mm->mmap_sem);
1858 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1859 up_read(&mm->mmap_sem);
1864 if (exact_vma_exists) {
1865 if (task_dumpable(task)) {
1867 cred = __task_cred(task);
1868 inode->i_uid = cred->euid;
1869 inode->i_gid = cred->egid;
1872 inode->i_uid = GLOBAL_ROOT_UID;
1873 inode->i_gid = GLOBAL_ROOT_GID;
1875 security_task_to_inode(task, inode);
1880 put_task_struct(task);
1886 static const struct dentry_operations tid_map_files_dentry_operations = {
1887 .d_revalidate = map_files_d_revalidate,
1888 .d_delete = pid_delete_dentry,
1891 static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
1893 unsigned long vm_start, vm_end;
1894 struct vm_area_struct *vma;
1895 struct task_struct *task;
1896 struct mm_struct *mm;
1900 task = get_proc_task(d_inode(dentry));
1904 mm = get_task_mm(task);
1905 put_task_struct(task);
1909 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
1914 down_read(&mm->mmap_sem);
1915 vma = find_exact_vma(mm, vm_start, vm_end);
1916 if (vma && vma->vm_file) {
1917 *path = vma->vm_file->f_path;
1921 up_read(&mm->mmap_sem);
1929 struct map_files_info {
1932 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
1936 * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the
1937 * symlinks may be used to bypass permissions on ancestor directories in the
1938 * path to the file in question.
1941 proc_map_files_follow_link(struct dentry *dentry, void **cookie)
1943 if (!capable(CAP_SYS_ADMIN))
1944 return ERR_PTR(-EPERM);
1946 return proc_pid_follow_link(dentry, NULL);
1950 * Identical to proc_pid_link_inode_operations except for follow_link()
1952 static const struct inode_operations proc_map_files_link_inode_operations = {
1953 .readlink = proc_pid_readlink,
1954 .follow_link = proc_map_files_follow_link,
1955 .setattr = proc_setattr,
1959 proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
1960 struct task_struct *task, const void *ptr)
1962 fmode_t mode = (fmode_t)(unsigned long)ptr;
1963 struct proc_inode *ei;
1964 struct inode *inode;
1966 inode = proc_pid_make_inode(dir->i_sb, task);
1971 ei->op.proc_get_link = proc_map_files_get_link;
1973 inode->i_op = &proc_map_files_link_inode_operations;
1975 inode->i_mode = S_IFLNK;
1977 if (mode & FMODE_READ)
1978 inode->i_mode |= S_IRUSR;
1979 if (mode & FMODE_WRITE)
1980 inode->i_mode |= S_IWUSR;
1982 d_set_d_op(dentry, &tid_map_files_dentry_operations);
1983 d_add(dentry, inode);
1988 static struct dentry *proc_map_files_lookup(struct inode *dir,
1989 struct dentry *dentry, unsigned int flags)
1991 unsigned long vm_start, vm_end;
1992 struct vm_area_struct *vma;
1993 struct task_struct *task;
1995 struct mm_struct *mm;
1998 task = get_proc_task(dir);
2003 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2007 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2010 mm = get_task_mm(task);
2014 down_read(&mm->mmap_sem);
2015 vma = find_exact_vma(mm, vm_start, vm_end);
2020 result = proc_map_files_instantiate(dir, dentry, task,
2021 (void *)(unsigned long)vma->vm_file->f_mode);
2024 up_read(&mm->mmap_sem);
2027 put_task_struct(task);
2029 return ERR_PTR(result);
2032 static const struct inode_operations proc_map_files_inode_operations = {
2033 .lookup = proc_map_files_lookup,
2034 .permission = proc_fd_permission,
2035 .setattr = proc_setattr,
2039 proc_map_files_readdir(struct file *file, struct dir_context *ctx)
2041 struct vm_area_struct *vma;
2042 struct task_struct *task;
2043 struct mm_struct *mm;
2044 unsigned long nr_files, pos, i;
2045 struct flex_array *fa = NULL;
2046 struct map_files_info info;
2047 struct map_files_info *p;
2051 task = get_proc_task(file_inode(file));
2056 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2060 if (!dir_emit_dots(file, ctx))
2063 mm = get_task_mm(task);
2066 down_read(&mm->mmap_sem);
2071 * We need two passes here:
2073 * 1) Collect vmas of mapped files with mmap_sem taken
2074 * 2) Release mmap_sem and instantiate entries
2076 * otherwise we get lockdep complained, since filldir()
2077 * routine might require mmap_sem taken in might_fault().
2080 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2081 if (vma->vm_file && ++pos > ctx->pos)
2086 fa = flex_array_alloc(sizeof(info), nr_files,
2088 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2092 flex_array_free(fa);
2093 up_read(&mm->mmap_sem);
2097 for (i = 0, vma = mm->mmap, pos = 2; vma;
2098 vma = vma->vm_next) {
2101 if (++pos <= ctx->pos)
2104 info.mode = vma->vm_file->f_mode;
2105 info.len = snprintf(info.name,
2106 sizeof(info.name), "%lx-%lx",
2107 vma->vm_start, vma->vm_end);
2108 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2112 up_read(&mm->mmap_sem);
2114 for (i = 0; i < nr_files; i++) {
2115 p = flex_array_get(fa, i);
2116 if (!proc_fill_cache(file, ctx,
2118 proc_map_files_instantiate,
2120 (void *)(unsigned long)p->mode))
2125 flex_array_free(fa);
2129 put_task_struct(task);
2134 static const struct file_operations proc_map_files_operations = {
2135 .read = generic_read_dir,
2136 .iterate = proc_map_files_readdir,
2137 .llseek = default_llseek,
2140 struct timers_private {
2142 struct task_struct *task;
2143 struct sighand_struct *sighand;
2144 struct pid_namespace *ns;
2145 unsigned long flags;
2148 static void *timers_start(struct seq_file *m, loff_t *pos)
2150 struct timers_private *tp = m->private;
2152 tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2154 return ERR_PTR(-ESRCH);
2156 tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2158 return ERR_PTR(-ESRCH);
2160 return seq_list_start(&tp->task->signal->posix_timers, *pos);
2163 static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2165 struct timers_private *tp = m->private;
2166 return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2169 static void timers_stop(struct seq_file *m, void *v)
2171 struct timers_private *tp = m->private;
2174 unlock_task_sighand(tp->task, &tp->flags);
2179 put_task_struct(tp->task);
2184 static int show_timer(struct seq_file *m, void *v)
2186 struct k_itimer *timer;
2187 struct timers_private *tp = m->private;
2189 static const char * const nstr[] = {
2190 [SIGEV_SIGNAL] = "signal",
2191 [SIGEV_NONE] = "none",
2192 [SIGEV_THREAD] = "thread",
2195 timer = list_entry((struct list_head *)v, struct k_itimer, list);
2196 notify = timer->it_sigev_notify;
2198 seq_printf(m, "ID: %d\n", timer->it_id);
2199 seq_printf(m, "signal: %d/%p\n",
2200 timer->sigq->info.si_signo,
2201 timer->sigq->info.si_value.sival_ptr);
2202 seq_printf(m, "notify: %s/%s.%d\n",
2203 nstr[notify & ~SIGEV_THREAD_ID],
2204 (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2205 pid_nr_ns(timer->it_pid, tp->ns));
2206 seq_printf(m, "ClockID: %d\n", timer->it_clock);
2211 static const struct seq_operations proc_timers_seq_ops = {
2212 .start = timers_start,
2213 .next = timers_next,
2214 .stop = timers_stop,
2218 static int proc_timers_open(struct inode *inode, struct file *file)
2220 struct timers_private *tp;
2222 tp = __seq_open_private(file, &proc_timers_seq_ops,
2223 sizeof(struct timers_private));
2227 tp->pid = proc_pid(inode);
2228 tp->ns = inode->i_sb->s_fs_info;
2232 static const struct file_operations proc_timers_operations = {
2233 .open = proc_timers_open,
2235 .llseek = seq_lseek,
2236 .release = seq_release_private,
2239 static int proc_pident_instantiate(struct inode *dir,
2240 struct dentry *dentry, struct task_struct *task, const void *ptr)
2242 const struct pid_entry *p = ptr;
2243 struct inode *inode;
2244 struct proc_inode *ei;
2246 inode = proc_pid_make_inode(dir->i_sb, task);
2251 inode->i_mode = p->mode;
2252 if (S_ISDIR(inode->i_mode))
2253 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2255 inode->i_op = p->iop;
2257 inode->i_fop = p->fop;
2259 d_set_d_op(dentry, &pid_dentry_operations);
2260 d_add(dentry, inode);
2261 /* Close the race of the process dying before we return the dentry */
2262 if (pid_revalidate(dentry, 0))
2268 static struct dentry *proc_pident_lookup(struct inode *dir,
2269 struct dentry *dentry,
2270 const struct pid_entry *ents,
2274 struct task_struct *task = get_proc_task(dir);
2275 const struct pid_entry *p, *last;
2283 * Yes, it does not scale. And it should not. Don't add
2284 * new entries into /proc/<tgid>/ without very good reasons.
2286 last = &ents[nents - 1];
2287 for (p = ents; p <= last; p++) {
2288 if (p->len != dentry->d_name.len)
2290 if (!memcmp(dentry->d_name.name, p->name, p->len))
2296 error = proc_pident_instantiate(dir, dentry, task, p);
2298 put_task_struct(task);
2300 return ERR_PTR(error);
2303 static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
2304 const struct pid_entry *ents, unsigned int nents)
2306 struct task_struct *task = get_proc_task(file_inode(file));
2307 const struct pid_entry *p;
2312 if (!dir_emit_dots(file, ctx))
2315 if (ctx->pos >= nents + 2)
2318 for (p = ents + (ctx->pos - 2); p <= ents + nents - 1; p++) {
2319 if (!proc_fill_cache(file, ctx, p->name, p->len,
2320 proc_pident_instantiate, task, p))
2325 put_task_struct(task);
2329 #ifdef CONFIG_SECURITY
2330 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2331 size_t count, loff_t *ppos)
2333 struct inode * inode = file_inode(file);
2336 struct task_struct *task = get_proc_task(inode);
2341 length = security_getprocattr(task,
2342 (char*)file->f_path.dentry->d_name.name,
2344 put_task_struct(task);
2346 length = simple_read_from_buffer(buf, count, ppos, p, length);
2351 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2352 size_t count, loff_t *ppos)
2354 struct inode * inode = file_inode(file);
2357 struct task_struct *task = get_proc_task(inode);
2362 if (count > PAGE_SIZE)
2365 /* No partial writes. */
2371 page = (char*)__get_free_page(GFP_TEMPORARY);
2376 if (copy_from_user(page, buf, count))
2379 /* Guard against adverse ptrace interaction */
2380 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2384 length = security_setprocattr(task,
2385 (char*)file->f_path.dentry->d_name.name,
2386 (void*)page, count);
2387 mutex_unlock(&task->signal->cred_guard_mutex);
2389 free_page((unsigned long) page);
2391 put_task_struct(task);
2396 static const struct file_operations proc_pid_attr_operations = {
2397 .read = proc_pid_attr_read,
2398 .write = proc_pid_attr_write,
2399 .llseek = generic_file_llseek,
2402 static const struct pid_entry attr_dir_stuff[] = {
2403 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2404 REG("prev", S_IRUGO, proc_pid_attr_operations),
2405 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2406 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2407 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2408 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2411 static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
2413 return proc_pident_readdir(file, ctx,
2414 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2417 static const struct file_operations proc_attr_dir_operations = {
2418 .read = generic_read_dir,
2419 .iterate = proc_attr_dir_readdir,
2420 .llseek = default_llseek,
2423 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2424 struct dentry *dentry, unsigned int flags)
2426 return proc_pident_lookup(dir, dentry,
2427 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2430 static const struct inode_operations proc_attr_dir_inode_operations = {
2431 .lookup = proc_attr_dir_lookup,
2432 .getattr = pid_getattr,
2433 .setattr = proc_setattr,
2438 #ifdef CONFIG_ELF_CORE
2439 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2440 size_t count, loff_t *ppos)
2442 struct task_struct *task = get_proc_task(file_inode(file));
2443 struct mm_struct *mm;
2444 char buffer[PROC_NUMBUF];
2452 mm = get_task_mm(task);
2454 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2455 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2456 MMF_DUMP_FILTER_SHIFT));
2458 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2461 put_task_struct(task);
2466 static ssize_t proc_coredump_filter_write(struct file *file,
2467 const char __user *buf,
2471 struct task_struct *task;
2472 struct mm_struct *mm;
2478 ret = kstrtouint_from_user(buf, count, 0, &val);
2483 task = get_proc_task(file_inode(file));
2487 mm = get_task_mm(task);
2491 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2493 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2495 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2500 put_task_struct(task);
2507 static const struct file_operations proc_coredump_filter_operations = {
2508 .read = proc_coredump_filter_read,
2509 .write = proc_coredump_filter_write,
2510 .llseek = generic_file_llseek,
2514 #ifdef CONFIG_TASK_IO_ACCOUNTING
2515 static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
2517 struct task_io_accounting acct = task->ioac;
2518 unsigned long flags;
2521 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2525 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2530 if (whole && lock_task_sighand(task, &flags)) {
2531 struct task_struct *t = task;
2533 task_io_accounting_add(&acct, &task->signal->ioac);
2534 while_each_thread(task, t)
2535 task_io_accounting_add(&acct, &t->ioac);
2537 unlock_task_sighand(task, &flags);
2544 "read_bytes: %llu\n"
2545 "write_bytes: %llu\n"
2546 "cancelled_write_bytes: %llu\n",
2547 (unsigned long long)acct.rchar,
2548 (unsigned long long)acct.wchar,
2549 (unsigned long long)acct.syscr,
2550 (unsigned long long)acct.syscw,
2551 (unsigned long long)acct.read_bytes,
2552 (unsigned long long)acct.write_bytes,
2553 (unsigned long long)acct.cancelled_write_bytes);
2557 mutex_unlock(&task->signal->cred_guard_mutex);
2561 static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2562 struct pid *pid, struct task_struct *task)
2564 return do_io_accounting(task, m, 0);
2567 static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2568 struct pid *pid, struct task_struct *task)
2570 return do_io_accounting(task, m, 1);
2572 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2574 #ifdef CONFIG_USER_NS
2575 static int proc_id_map_open(struct inode *inode, struct file *file,
2576 const struct seq_operations *seq_ops)
2578 struct user_namespace *ns = NULL;
2579 struct task_struct *task;
2580 struct seq_file *seq;
2583 task = get_proc_task(inode);
2586 ns = get_user_ns(task_cred_xxx(task, user_ns));
2588 put_task_struct(task);
2593 ret = seq_open(file, seq_ops);
2597 seq = file->private_data;
2607 static int proc_id_map_release(struct inode *inode, struct file *file)
2609 struct seq_file *seq = file->private_data;
2610 struct user_namespace *ns = seq->private;
2612 return seq_release(inode, file);
2615 static int proc_uid_map_open(struct inode *inode, struct file *file)
2617 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2620 static int proc_gid_map_open(struct inode *inode, struct file *file)
2622 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2625 static int proc_projid_map_open(struct inode *inode, struct file *file)
2627 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2630 static const struct file_operations proc_uid_map_operations = {
2631 .open = proc_uid_map_open,
2632 .write = proc_uid_map_write,
2634 .llseek = seq_lseek,
2635 .release = proc_id_map_release,
2638 static const struct file_operations proc_gid_map_operations = {
2639 .open = proc_gid_map_open,
2640 .write = proc_gid_map_write,
2642 .llseek = seq_lseek,
2643 .release = proc_id_map_release,
2646 static const struct file_operations proc_projid_map_operations = {
2647 .open = proc_projid_map_open,
2648 .write = proc_projid_map_write,
2650 .llseek = seq_lseek,
2651 .release = proc_id_map_release,
2654 static int proc_setgroups_open(struct inode *inode, struct file *file)
2656 struct user_namespace *ns = NULL;
2657 struct task_struct *task;
2661 task = get_proc_task(inode);
2664 ns = get_user_ns(task_cred_xxx(task, user_ns));
2666 put_task_struct(task);
2671 if (file->f_mode & FMODE_WRITE) {
2673 if (!ns_capable(ns, CAP_SYS_ADMIN))
2677 ret = single_open(file, &proc_setgroups_show, ns);
2688 static int proc_setgroups_release(struct inode *inode, struct file *file)
2690 struct seq_file *seq = file->private_data;
2691 struct user_namespace *ns = seq->private;
2692 int ret = single_release(inode, file);
2697 static const struct file_operations proc_setgroups_operations = {
2698 .open = proc_setgroups_open,
2699 .write = proc_setgroups_write,
2701 .llseek = seq_lseek,
2702 .release = proc_setgroups_release,
2704 #endif /* CONFIG_USER_NS */
2706 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2707 struct pid *pid, struct task_struct *task)
2709 int err = lock_trace(task);
2711 seq_printf(m, "%08x\n", task->personality);
2720 static const struct file_operations proc_task_operations;
2721 static const struct inode_operations proc_task_inode_operations;
2723 static const struct pid_entry tgid_base_stuff[] = {
2724 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2725 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2726 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
2727 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2728 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2730 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2732 REG("environ", S_IRUSR, proc_environ_operations),
2733 ONE("auxv", S_IRUSR, proc_pid_auxv),
2734 ONE("status", S_IRUGO, proc_pid_status),
2735 ONE("personality", S_IRUSR, proc_pid_personality),
2736 ONE("limits", S_IRUGO, proc_pid_limits),
2737 #ifdef CONFIG_SCHED_DEBUG
2738 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2740 #ifdef CONFIG_SCHED_AUTOGROUP
2741 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2743 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2744 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2745 ONE("syscall", S_IRUSR, proc_pid_syscall),
2747 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
2748 ONE("stat", S_IRUGO, proc_tgid_stat),
2749 ONE("statm", S_IRUGO, proc_pid_statm),
2750 REG("maps", S_IRUGO, proc_pid_maps_operations),
2752 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
2754 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2755 LNK("cwd", proc_cwd_link),
2756 LNK("root", proc_root_link),
2757 LNK("exe", proc_exe_link),
2758 REG("mounts", S_IRUGO, proc_mounts_operations),
2759 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2760 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2761 #ifdef CONFIG_PROC_PAGE_MONITOR
2762 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2763 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
2764 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2766 #ifdef CONFIG_SECURITY
2767 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2769 #ifdef CONFIG_KALLSYMS
2770 ONE("wchan", S_IRUGO, proc_pid_wchan),
2772 #ifdef CONFIG_STACKTRACE
2773 ONE("stack", S_IRUSR, proc_pid_stack),
2775 #ifdef CONFIG_SCHED_INFO
2776 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
2778 #ifdef CONFIG_LATENCYTOP
2779 REG("latency", S_IRUGO, proc_lstats_operations),
2781 #ifdef CONFIG_PROC_PID_CPUSET
2782 ONE("cpuset", S_IRUGO, proc_cpuset_show),
2784 #ifdef CONFIG_CGROUPS
2785 ONE("cgroup", S_IRUGO, proc_cgroup_show),
2787 ONE("oom_score", S_IRUGO, proc_oom_score),
2788 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
2789 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2790 #ifdef CONFIG_AUDITSYSCALL
2791 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2792 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2794 #ifdef CONFIG_FAULT_INJECTION
2795 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2797 #ifdef CONFIG_ELF_CORE
2798 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2800 #ifdef CONFIG_TASK_IO_ACCOUNTING
2801 ONE("io", S_IRUSR, proc_tgid_io_accounting),
2803 #ifdef CONFIG_HARDWALL
2804 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
2806 #ifdef CONFIG_USER_NS
2807 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
2808 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
2809 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
2810 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
2812 #ifdef CONFIG_CHECKPOINT_RESTORE
2813 REG("timers", S_IRUGO, proc_timers_operations),
2817 static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
2819 return proc_pident_readdir(file, ctx,
2820 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2823 static const struct file_operations proc_tgid_base_operations = {
2824 .read = generic_read_dir,
2825 .iterate = proc_tgid_base_readdir,
2826 .llseek = default_llseek,
2829 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2831 return proc_pident_lookup(dir, dentry,
2832 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2835 static const struct inode_operations proc_tgid_base_inode_operations = {
2836 .lookup = proc_tgid_base_lookup,
2837 .getattr = pid_getattr,
2838 .setattr = proc_setattr,
2839 .permission = proc_pid_permission,
2842 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2844 struct dentry *dentry, *leader, *dir;
2845 char buf[PROC_NUMBUF];
2849 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2850 /* no ->d_hash() rejects on procfs */
2851 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2853 d_invalidate(dentry);
2861 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2862 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2867 name.len = strlen(name.name);
2868 dir = d_hash_and_lookup(leader, &name);
2870 goto out_put_leader;
2873 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2874 dentry = d_hash_and_lookup(dir, &name);
2876 d_invalidate(dentry);
2888 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2889 * @task: task that should be flushed.
2891 * When flushing dentries from proc, one needs to flush them from global
2892 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2893 * in. This call is supposed to do all of this job.
2895 * Looks in the dcache for
2897 * /proc/@tgid/task/@pid
2898 * if either directory is present flushes it and all of it'ts children
2901 * It is safe and reasonable to cache /proc entries for a task until
2902 * that task exits. After that they just clog up the dcache with
2903 * useless entries, possibly causing useful dcache entries to be
2904 * flushed instead. This routine is proved to flush those useless
2905 * dcache entries at process exit time.
2907 * NOTE: This routine is just an optimization so it does not guarantee
2908 * that no dcache entries will exist at process exit time it
2909 * just makes it very unlikely that any will persist.
2912 void proc_flush_task(struct task_struct *task)
2915 struct pid *pid, *tgid;
2918 pid = task_pid(task);
2919 tgid = task_tgid(task);
2921 for (i = 0; i <= pid->level; i++) {
2922 upid = &pid->numbers[i];
2923 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2924 tgid->numbers[i].nr);
2928 static int proc_pid_instantiate(struct inode *dir,
2929 struct dentry * dentry,
2930 struct task_struct *task, const void *ptr)
2932 struct inode *inode;
2934 inode = proc_pid_make_inode(dir->i_sb, task);
2938 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2939 inode->i_op = &proc_tgid_base_inode_operations;
2940 inode->i_fop = &proc_tgid_base_operations;
2941 inode->i_flags|=S_IMMUTABLE;
2943 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
2944 ARRAY_SIZE(tgid_base_stuff)));
2946 d_set_d_op(dentry, &pid_dentry_operations);
2948 d_add(dentry, inode);
2949 /* Close the race of the process dying before we return the dentry */
2950 if (pid_revalidate(dentry, 0))
2956 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
2958 int result = -ENOENT;
2959 struct task_struct *task;
2961 struct pid_namespace *ns;
2963 tgid = name_to_int(&dentry->d_name);
2967 ns = dentry->d_sb->s_fs_info;
2969 task = find_task_by_pid_ns(tgid, ns);
2971 get_task_struct(task);
2976 result = proc_pid_instantiate(dir, dentry, task, NULL);
2977 put_task_struct(task);
2979 return ERR_PTR(result);
2983 * Find the first task with tgid >= tgid
2988 struct task_struct *task;
2990 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2995 put_task_struct(iter.task);
2999 pid = find_ge_pid(iter.tgid, ns);
3001 iter.tgid = pid_nr_ns(pid, ns);
3002 iter.task = pid_task(pid, PIDTYPE_PID);
3003 /* What we to know is if the pid we have find is the
3004 * pid of a thread_group_leader. Testing for task
3005 * being a thread_group_leader is the obvious thing
3006 * todo but there is a window when it fails, due to
3007 * the pid transfer logic in de_thread.
3009 * So we perform the straight forward test of seeing
3010 * if the pid we have found is the pid of a thread
3011 * group leader, and don't worry if the task we have
3012 * found doesn't happen to be a thread group leader.
3013 * As we don't care in the case of readdir.
3015 if (!iter.task || !has_group_leader_pid(iter.task)) {
3019 get_task_struct(iter.task);
3025 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3027 /* for the /proc/ directory itself, after non-process stuff has been done */
3028 int proc_pid_readdir(struct file *file, struct dir_context *ctx)
3030 struct tgid_iter iter;
3031 struct pid_namespace *ns = file_inode(file)->i_sb->s_fs_info;
3032 loff_t pos = ctx->pos;
3034 if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
3037 if (pos == TGID_OFFSET - 2) {
3038 struct inode *inode = d_inode(ns->proc_self);
3039 if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
3041 ctx->pos = pos = pos + 1;
3043 if (pos == TGID_OFFSET - 1) {
3044 struct inode *inode = d_inode(ns->proc_thread_self);
3045 if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3047 ctx->pos = pos = pos + 1;
3049 iter.tgid = pos - TGID_OFFSET;
3051 for (iter = next_tgid(ns, iter);
3053 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3054 char name[PROC_NUMBUF];
3056 if (!has_pid_permissions(ns, iter.task, 2))
3059 len = snprintf(name, sizeof(name), "%d", iter.tgid);
3060 ctx->pos = iter.tgid + TGID_OFFSET;
3061 if (!proc_fill_cache(file, ctx, name, len,
3062 proc_pid_instantiate, iter.task, NULL)) {
3063 put_task_struct(iter.task);
3067 ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
3074 static const struct pid_entry tid_base_stuff[] = {
3075 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3076 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3077 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3079 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3081 REG("environ", S_IRUSR, proc_environ_operations),
3082 ONE("auxv", S_IRUSR, proc_pid_auxv),
3083 ONE("status", S_IRUGO, proc_pid_status),
3084 ONE("personality", S_IRUSR, proc_pid_personality),
3085 ONE("limits", S_IRUGO, proc_pid_limits),
3086 #ifdef CONFIG_SCHED_DEBUG
3087 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3089 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3090 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3091 ONE("syscall", S_IRUSR, proc_pid_syscall),
3093 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
3094 ONE("stat", S_IRUGO, proc_tid_stat),
3095 ONE("statm", S_IRUGO, proc_pid_statm),
3096 REG("maps", S_IRUGO, proc_tid_maps_operations),
3097 #ifdef CONFIG_PROC_CHILDREN
3098 REG("children", S_IRUGO, proc_tid_children_operations),
3101 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
3103 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3104 LNK("cwd", proc_cwd_link),
3105 LNK("root", proc_root_link),
3106 LNK("exe", proc_exe_link),
3107 REG("mounts", S_IRUGO, proc_mounts_operations),
3108 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3109 #ifdef CONFIG_PROC_PAGE_MONITOR
3110 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3111 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
3112 REG("pagemap", S_IRUSR, proc_pagemap_operations),
3114 #ifdef CONFIG_SECURITY
3115 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3117 #ifdef CONFIG_KALLSYMS
3118 ONE("wchan", S_IRUGO, proc_pid_wchan),
3120 #ifdef CONFIG_STACKTRACE
3121 ONE("stack", S_IRUSR, proc_pid_stack),
3123 #ifdef CONFIG_SCHED_INFO
3124 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
3126 #ifdef CONFIG_LATENCYTOP
3127 REG("latency", S_IRUGO, proc_lstats_operations),
3129 #ifdef CONFIG_PROC_PID_CPUSET
3130 ONE("cpuset", S_IRUGO, proc_cpuset_show),
3132 #ifdef CONFIG_CGROUPS
3133 ONE("cgroup", S_IRUGO, proc_cgroup_show),
3135 ONE("oom_score", S_IRUGO, proc_oom_score),
3136 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3137 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3138 #ifdef CONFIG_AUDITSYSCALL
3139 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3140 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3142 #ifdef CONFIG_FAULT_INJECTION
3143 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3145 #ifdef CONFIG_TASK_IO_ACCOUNTING
3146 ONE("io", S_IRUSR, proc_tid_io_accounting),
3148 #ifdef CONFIG_HARDWALL
3149 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
3151 #ifdef CONFIG_USER_NS
3152 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3153 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3154 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3155 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
3159 static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
3161 return proc_pident_readdir(file, ctx,
3162 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3165 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3167 return proc_pident_lookup(dir, dentry,
3168 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3171 static const struct file_operations proc_tid_base_operations = {
3172 .read = generic_read_dir,
3173 .iterate = proc_tid_base_readdir,
3174 .llseek = default_llseek,
3177 static const struct inode_operations proc_tid_base_inode_operations = {
3178 .lookup = proc_tid_base_lookup,
3179 .getattr = pid_getattr,
3180 .setattr = proc_setattr,
3183 static int proc_task_instantiate(struct inode *dir,
3184 struct dentry *dentry, struct task_struct *task, const void *ptr)
3186 struct inode *inode;
3187 inode = proc_pid_make_inode(dir->i_sb, task);
3191 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3192 inode->i_op = &proc_tid_base_inode_operations;
3193 inode->i_fop = &proc_tid_base_operations;
3194 inode->i_flags|=S_IMMUTABLE;
3196 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3197 ARRAY_SIZE(tid_base_stuff)));
3199 d_set_d_op(dentry, &pid_dentry_operations);
3201 d_add(dentry, inode);
3202 /* Close the race of the process dying before we return the dentry */
3203 if (pid_revalidate(dentry, 0))
3209 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3211 int result = -ENOENT;
3212 struct task_struct *task;
3213 struct task_struct *leader = get_proc_task(dir);
3215 struct pid_namespace *ns;
3220 tid = name_to_int(&dentry->d_name);
3224 ns = dentry->d_sb->s_fs_info;
3226 task = find_task_by_pid_ns(tid, ns);
3228 get_task_struct(task);
3232 if (!same_thread_group(leader, task))
3235 result = proc_task_instantiate(dir, dentry, task, NULL);
3237 put_task_struct(task);
3239 put_task_struct(leader);
3241 return ERR_PTR(result);
3245 * Find the first tid of a thread group to return to user space.
3247 * Usually this is just the thread group leader, but if the users
3248 * buffer was too small or there was a seek into the middle of the
3249 * directory we have more work todo.
3251 * In the case of a short read we start with find_task_by_pid.
3253 * In the case of a seek we start with the leader and walk nr
3256 static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3257 struct pid_namespace *ns)
3259 struct task_struct *pos, *task;
3260 unsigned long nr = f_pos;
3262 if (nr != f_pos) /* 32bit overflow? */
3266 task = pid_task(pid, PIDTYPE_PID);
3270 /* Attempt to start with the tid of a thread */
3272 pos = find_task_by_pid_ns(tid, ns);
3273 if (pos && same_thread_group(pos, task))
3277 /* If nr exceeds the number of threads there is nothing todo */
3278 if (nr >= get_nr_threads(task))
3281 /* If we haven't found our starting place yet start
3282 * with the leader and walk nr threads forward.
3284 pos = task = task->group_leader;
3288 } while_each_thread(task, pos);
3293 get_task_struct(pos);
3300 * Find the next thread in the thread list.
3301 * Return NULL if there is an error or no next thread.
3303 * The reference to the input task_struct is released.
3305 static struct task_struct *next_tid(struct task_struct *start)
3307 struct task_struct *pos = NULL;
3309 if (pid_alive(start)) {
3310 pos = next_thread(start);
3311 if (thread_group_leader(pos))
3314 get_task_struct(pos);
3317 put_task_struct(start);
3321 /* for the /proc/TGID/task/ directories */
3322 static int proc_task_readdir(struct file *file, struct dir_context *ctx)
3324 struct inode *inode = file_inode(file);
3325 struct task_struct *task;
3326 struct pid_namespace *ns;
3329 if (proc_inode_is_dead(inode))
3332 if (!dir_emit_dots(file, ctx))
3335 /* f_version caches the tgid value that the last readdir call couldn't
3336 * return. lseek aka telldir automagically resets f_version to 0.
3338 ns = inode->i_sb->s_fs_info;
3339 tid = (int)file->f_version;
3340 file->f_version = 0;
3341 for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
3343 task = next_tid(task), ctx->pos++) {
3344 char name[PROC_NUMBUF];
3346 tid = task_pid_nr_ns(task, ns);
3347 len = snprintf(name, sizeof(name), "%d", tid);
3348 if (!proc_fill_cache(file, ctx, name, len,
3349 proc_task_instantiate, task, NULL)) {
3350 /* returning this tgid failed, save it as the first
3351 * pid for the next readir call */
3352 file->f_version = (u64)tid;
3353 put_task_struct(task);
3361 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3363 struct inode *inode = d_inode(dentry);
3364 struct task_struct *p = get_proc_task(inode);
3365 generic_fillattr(inode, stat);
3368 stat->nlink += get_nr_threads(p);
3375 static const struct inode_operations proc_task_inode_operations = {
3376 .lookup = proc_task_lookup,
3377 .getattr = proc_task_getattr,
3378 .setattr = proc_setattr,
3379 .permission = proc_pid_permission,
3382 static const struct file_operations proc_task_operations = {
3383 .read = generic_read_dir,
3384 .iterate = proc_task_readdir,
3385 .llseek = default_llseek,