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_FSCREDS);
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_FSCREDS)
434 && !lookup_symbol_name(wchan, symname))
435 seq_printf(m, "%s", symname);
441 #endif /* CONFIG_KALLSYMS */
443 static int lock_trace(struct task_struct *task)
445 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
448 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) {
449 mutex_unlock(&task->signal->cred_guard_mutex);
455 static void unlock_trace(struct task_struct *task)
457 mutex_unlock(&task->signal->cred_guard_mutex);
460 #ifdef CONFIG_STACKTRACE
462 #define MAX_STACK_TRACE_DEPTH 64
464 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
465 struct pid *pid, struct task_struct *task)
467 struct stack_trace trace;
468 unsigned long *entries;
472 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
476 trace.nr_entries = 0;
477 trace.max_entries = MAX_STACK_TRACE_DEPTH;
478 trace.entries = entries;
481 err = lock_trace(task);
483 save_stack_trace_tsk(task, &trace);
485 for (i = 0; i < trace.nr_entries; i++) {
486 seq_printf(m, "[<%pK>] %pS\n",
487 (void *)entries[i], (void *)entries[i]);
497 #ifdef CONFIG_SCHED_INFO
499 * Provides /proc/PID/schedstat
501 static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
502 struct pid *pid, struct task_struct *task)
504 if (unlikely(!sched_info_on()))
505 seq_printf(m, "0 0 0\n");
507 seq_printf(m, "%llu %llu %lu\n",
508 (unsigned long long)task->se.sum_exec_runtime,
509 (unsigned long long)task->sched_info.run_delay,
510 task->sched_info.pcount);
516 #ifdef CONFIG_LATENCYTOP
517 static int lstats_show_proc(struct seq_file *m, void *v)
520 struct inode *inode = m->private;
521 struct task_struct *task = get_proc_task(inode);
525 seq_puts(m, "Latency Top version : v0.1\n");
526 for (i = 0; i < 32; i++) {
527 struct latency_record *lr = &task->latency_record[i];
528 if (lr->backtrace[0]) {
530 seq_printf(m, "%i %li %li",
531 lr->count, lr->time, lr->max);
532 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
533 unsigned long bt = lr->backtrace[q];
538 seq_printf(m, " %ps", (void *)bt);
544 put_task_struct(task);
548 static int lstats_open(struct inode *inode, struct file *file)
550 return single_open(file, lstats_show_proc, inode);
553 static ssize_t lstats_write(struct file *file, const char __user *buf,
554 size_t count, loff_t *offs)
556 struct task_struct *task = get_proc_task(file_inode(file));
560 clear_all_latency_tracing(task);
561 put_task_struct(task);
566 static const struct file_operations proc_lstats_operations = {
569 .write = lstats_write,
571 .release = single_release,
576 static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
577 struct pid *pid, struct task_struct *task)
579 unsigned long totalpages = totalram_pages + total_swap_pages;
580 unsigned long points = 0;
582 read_lock(&tasklist_lock);
584 points = oom_badness(task, NULL, NULL, totalpages) *
586 read_unlock(&tasklist_lock);
587 seq_printf(m, "%lu\n", points);
597 static const struct limit_names lnames[RLIM_NLIMITS] = {
598 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
599 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
600 [RLIMIT_DATA] = {"Max data size", "bytes"},
601 [RLIMIT_STACK] = {"Max stack size", "bytes"},
602 [RLIMIT_CORE] = {"Max core file size", "bytes"},
603 [RLIMIT_RSS] = {"Max resident set", "bytes"},
604 [RLIMIT_NPROC] = {"Max processes", "processes"},
605 [RLIMIT_NOFILE] = {"Max open files", "files"},
606 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
607 [RLIMIT_AS] = {"Max address space", "bytes"},
608 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
609 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
610 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
611 [RLIMIT_NICE] = {"Max nice priority", NULL},
612 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
613 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
616 /* Display limits for a process */
617 static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
618 struct pid *pid, struct task_struct *task)
623 struct rlimit rlim[RLIM_NLIMITS];
625 if (!lock_task_sighand(task, &flags))
627 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
628 unlock_task_sighand(task, &flags);
631 * print the file header
633 seq_printf(m, "%-25s %-20s %-20s %-10s\n",
634 "Limit", "Soft Limit", "Hard Limit", "Units");
636 for (i = 0; i < RLIM_NLIMITS; i++) {
637 if (rlim[i].rlim_cur == RLIM_INFINITY)
638 seq_printf(m, "%-25s %-20s ",
639 lnames[i].name, "unlimited");
641 seq_printf(m, "%-25s %-20lu ",
642 lnames[i].name, rlim[i].rlim_cur);
644 if (rlim[i].rlim_max == RLIM_INFINITY)
645 seq_printf(m, "%-20s ", "unlimited");
647 seq_printf(m, "%-20lu ", rlim[i].rlim_max);
650 seq_printf(m, "%-10s\n", lnames[i].unit);
658 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
659 static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
660 struct pid *pid, struct task_struct *task)
663 unsigned long args[6], sp, pc;
666 res = lock_trace(task);
670 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
671 seq_puts(m, "running\n");
673 seq_printf(m, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
676 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
678 args[0], args[1], args[2], args[3], args[4], args[5],
684 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
686 /************************************************************************/
687 /* Here the fs part begins */
688 /************************************************************************/
690 /* permission checks */
691 static int proc_fd_access_allowed(struct inode *inode)
693 struct task_struct *task;
695 /* Allow access to a task's file descriptors if it is us or we
696 * may use ptrace attach to the process and find out that
699 task = get_proc_task(inode);
701 allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
702 put_task_struct(task);
707 int proc_setattr(struct dentry *dentry, struct iattr *attr)
710 struct inode *inode = d_inode(dentry);
712 if (attr->ia_valid & ATTR_MODE)
715 error = inode_change_ok(inode, attr);
719 setattr_copy(inode, attr);
720 mark_inode_dirty(inode);
725 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
726 * or euid/egid (for hide_pid_min=2)?
728 static bool has_pid_permissions(struct pid_namespace *pid,
729 struct task_struct *task,
732 if (pid->hide_pid < hide_pid_min)
734 if (in_group_p(pid->pid_gid))
736 return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
740 static int proc_pid_permission(struct inode *inode, int mask)
742 struct pid_namespace *pid = inode->i_sb->s_fs_info;
743 struct task_struct *task;
746 task = get_proc_task(inode);
749 has_perms = has_pid_permissions(pid, task, 1);
750 put_task_struct(task);
753 if (pid->hide_pid == 2) {
755 * Let's make getdents(), stat(), and open()
756 * consistent with each other. If a process
757 * may not stat() a file, it shouldn't be seen
765 return generic_permission(inode, mask);
770 static const struct inode_operations proc_def_inode_operations = {
771 .setattr = proc_setattr,
774 static int proc_single_show(struct seq_file *m, void *v)
776 struct inode *inode = m->private;
777 struct pid_namespace *ns;
779 struct task_struct *task;
782 ns = inode->i_sb->s_fs_info;
783 pid = proc_pid(inode);
784 task = get_pid_task(pid, PIDTYPE_PID);
788 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
790 put_task_struct(task);
794 static int proc_single_open(struct inode *inode, struct file *filp)
796 return single_open(filp, proc_single_show, inode);
799 static const struct file_operations proc_single_file_operations = {
800 .open = proc_single_open,
803 .release = single_release,
807 struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
809 struct task_struct *task = get_proc_task(inode);
810 struct mm_struct *mm = ERR_PTR(-ESRCH);
813 mm = mm_access(task, mode | PTRACE_MODE_FSCREDS);
814 put_task_struct(task);
816 if (!IS_ERR_OR_NULL(mm)) {
817 /* ensure this mm_struct can't be freed */
818 atomic_inc(&mm->mm_count);
819 /* but do not pin its memory */
827 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
829 struct mm_struct *mm = proc_mem_open(inode, mode);
834 file->private_data = mm;
838 static int mem_open(struct inode *inode, struct file *file)
840 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
842 /* OK to pass negative loff_t, we can catch out-of-range */
843 file->f_mode |= FMODE_UNSIGNED_OFFSET;
848 static ssize_t mem_rw(struct file *file, char __user *buf,
849 size_t count, loff_t *ppos, int write)
851 struct mm_struct *mm = file->private_data;
852 unsigned long addr = *ppos;
859 page = (char *)__get_free_page(GFP_TEMPORARY);
864 if (!atomic_inc_not_zero(&mm->mm_users))
868 int this_len = min_t(int, count, PAGE_SIZE);
870 if (write && copy_from_user(page, buf, this_len)) {
875 this_len = access_remote_vm(mm, addr, page, this_len, write);
882 if (!write && copy_to_user(buf, page, this_len)) {
896 free_page((unsigned long) page);
900 static ssize_t mem_read(struct file *file, char __user *buf,
901 size_t count, loff_t *ppos)
903 return mem_rw(file, buf, count, ppos, 0);
906 static ssize_t mem_write(struct file *file, const char __user *buf,
907 size_t count, loff_t *ppos)
909 return mem_rw(file, (char __user*)buf, count, ppos, 1);
912 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
916 file->f_pos = offset;
919 file->f_pos += offset;
924 force_successful_syscall_return();
928 static int mem_release(struct inode *inode, struct file *file)
930 struct mm_struct *mm = file->private_data;
936 static const struct file_operations proc_mem_operations = {
941 .release = mem_release,
944 static int environ_open(struct inode *inode, struct file *file)
946 return __mem_open(inode, file, PTRACE_MODE_READ);
949 static ssize_t environ_read(struct file *file, char __user *buf,
950 size_t count, loff_t *ppos)
953 unsigned long src = *ppos;
955 struct mm_struct *mm = file->private_data;
960 page = (char *)__get_free_page(GFP_TEMPORARY);
965 if (!atomic_inc_not_zero(&mm->mm_users))
968 size_t this_len, max_len;
971 if (src >= (mm->env_end - mm->env_start))
974 this_len = mm->env_end - (mm->env_start + src);
976 max_len = min_t(size_t, PAGE_SIZE, count);
977 this_len = min(max_len, this_len);
979 retval = access_remote_vm(mm, (mm->env_start + src),
987 if (copy_to_user(buf, page, retval)) {
1001 free_page((unsigned long) page);
1005 static const struct file_operations proc_environ_operations = {
1006 .open = environ_open,
1007 .read = environ_read,
1008 .llseek = generic_file_llseek,
1009 .release = mem_release,
1012 static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1015 struct task_struct *task = get_proc_task(file_inode(file));
1016 char buffer[PROC_NUMBUF];
1017 int oom_adj = OOM_ADJUST_MIN;
1019 unsigned long flags;
1023 if (lock_task_sighand(task, &flags)) {
1024 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1025 oom_adj = OOM_ADJUST_MAX;
1027 oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1029 unlock_task_sighand(task, &flags);
1031 put_task_struct(task);
1032 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1033 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1037 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1038 * kernels. The effective policy is defined by oom_score_adj, which has a
1039 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1040 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1041 * Processes that become oom disabled via oom_adj will still be oom disabled
1042 * with this implementation.
1044 * oom_adj cannot be removed since existing userspace binaries use it.
1046 static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1047 size_t count, loff_t *ppos)
1049 struct task_struct *task;
1050 char buffer[PROC_NUMBUF];
1052 unsigned long flags;
1055 memset(buffer, 0, sizeof(buffer));
1056 if (count > sizeof(buffer) - 1)
1057 count = sizeof(buffer) - 1;
1058 if (copy_from_user(buffer, buf, count)) {
1063 err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1066 if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1067 oom_adj != OOM_DISABLE) {
1072 task = get_proc_task(file_inode(file));
1084 if (!lock_task_sighand(task, &flags)) {
1090 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1091 * value is always attainable.
1093 if (oom_adj == OOM_ADJUST_MAX)
1094 oom_adj = OOM_SCORE_ADJ_MAX;
1096 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1098 if (oom_adj < task->signal->oom_score_adj &&
1099 !capable(CAP_SYS_RESOURCE)) {
1105 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1106 * /proc/pid/oom_score_adj instead.
1108 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1109 current->comm, task_pid_nr(current), task_pid_nr(task),
1112 task->signal->oom_score_adj = oom_adj;
1113 trace_oom_score_adj_update(task);
1115 unlock_task_sighand(task, &flags);
1118 put_task_struct(task);
1120 return err < 0 ? err : count;
1123 static const struct file_operations proc_oom_adj_operations = {
1124 .read = oom_adj_read,
1125 .write = oom_adj_write,
1126 .llseek = generic_file_llseek,
1129 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1130 size_t count, loff_t *ppos)
1132 struct task_struct *task = get_proc_task(file_inode(file));
1133 char buffer[PROC_NUMBUF];
1134 short oom_score_adj = OOM_SCORE_ADJ_MIN;
1135 unsigned long flags;
1140 if (lock_task_sighand(task, &flags)) {
1141 oom_score_adj = task->signal->oom_score_adj;
1142 unlock_task_sighand(task, &flags);
1144 put_task_struct(task);
1145 len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
1146 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1149 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1150 size_t count, loff_t *ppos)
1152 struct task_struct *task;
1153 char buffer[PROC_NUMBUF];
1154 unsigned long flags;
1158 memset(buffer, 0, sizeof(buffer));
1159 if (count > sizeof(buffer) - 1)
1160 count = sizeof(buffer) - 1;
1161 if (copy_from_user(buffer, buf, count)) {
1166 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1169 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1170 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1175 task = get_proc_task(file_inode(file));
1187 if (!lock_task_sighand(task, &flags)) {
1192 if ((short)oom_score_adj < task->signal->oom_score_adj_min &&
1193 !capable(CAP_SYS_RESOURCE)) {
1198 task->signal->oom_score_adj = (short)oom_score_adj;
1199 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1200 task->signal->oom_score_adj_min = (short)oom_score_adj;
1201 trace_oom_score_adj_update(task);
1204 unlock_task_sighand(task, &flags);
1207 put_task_struct(task);
1209 return err < 0 ? err : count;
1212 static const struct file_operations proc_oom_score_adj_operations = {
1213 .read = oom_score_adj_read,
1214 .write = oom_score_adj_write,
1215 .llseek = default_llseek,
1218 #ifdef CONFIG_AUDITSYSCALL
1219 #define TMPBUFLEN 21
1220 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1221 size_t count, loff_t *ppos)
1223 struct inode * inode = file_inode(file);
1224 struct task_struct *task = get_proc_task(inode);
1226 char tmpbuf[TMPBUFLEN];
1230 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1231 from_kuid(file->f_cred->user_ns,
1232 audit_get_loginuid(task)));
1233 put_task_struct(task);
1234 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1237 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1238 size_t count, loff_t *ppos)
1240 struct inode * inode = file_inode(file);
1246 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1253 /* No partial writes. */
1257 rv = kstrtou32_from_user(buf, count, 10, &loginuid);
1261 /* is userspace tring to explicitly UNSET the loginuid? */
1262 if (loginuid == AUDIT_UID_UNSET) {
1263 kloginuid = INVALID_UID;
1265 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1266 if (!uid_valid(kloginuid))
1270 rv = audit_set_loginuid(kloginuid);
1276 static const struct file_operations proc_loginuid_operations = {
1277 .read = proc_loginuid_read,
1278 .write = proc_loginuid_write,
1279 .llseek = generic_file_llseek,
1282 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1283 size_t count, loff_t *ppos)
1285 struct inode * inode = file_inode(file);
1286 struct task_struct *task = get_proc_task(inode);
1288 char tmpbuf[TMPBUFLEN];
1292 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1293 audit_get_sessionid(task));
1294 put_task_struct(task);
1295 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1298 static const struct file_operations proc_sessionid_operations = {
1299 .read = proc_sessionid_read,
1300 .llseek = generic_file_llseek,
1304 #ifdef CONFIG_FAULT_INJECTION
1305 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1306 size_t count, loff_t *ppos)
1308 struct task_struct *task = get_proc_task(file_inode(file));
1309 char buffer[PROC_NUMBUF];
1315 make_it_fail = task->make_it_fail;
1316 put_task_struct(task);
1318 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1320 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1323 static ssize_t proc_fault_inject_write(struct file * file,
1324 const char __user * buf, size_t count, loff_t *ppos)
1326 struct task_struct *task;
1327 char buffer[PROC_NUMBUF];
1331 if (!capable(CAP_SYS_RESOURCE))
1333 memset(buffer, 0, sizeof(buffer));
1334 if (count > sizeof(buffer) - 1)
1335 count = sizeof(buffer) - 1;
1336 if (copy_from_user(buffer, buf, count))
1338 rv = kstrtoint(strstrip(buffer), 0, &make_it_fail);
1341 if (make_it_fail < 0 || make_it_fail > 1)
1344 task = get_proc_task(file_inode(file));
1347 task->make_it_fail = make_it_fail;
1348 put_task_struct(task);
1353 static const struct file_operations proc_fault_inject_operations = {
1354 .read = proc_fault_inject_read,
1355 .write = proc_fault_inject_write,
1356 .llseek = generic_file_llseek,
1361 #ifdef CONFIG_SCHED_DEBUG
1363 * Print out various scheduling related per-task fields:
1365 static int sched_show(struct seq_file *m, void *v)
1367 struct inode *inode = m->private;
1368 struct task_struct *p;
1370 p = get_proc_task(inode);
1373 proc_sched_show_task(p, m);
1381 sched_write(struct file *file, const char __user *buf,
1382 size_t count, loff_t *offset)
1384 struct inode *inode = file_inode(file);
1385 struct task_struct *p;
1387 p = get_proc_task(inode);
1390 proc_sched_set_task(p);
1397 static int sched_open(struct inode *inode, struct file *filp)
1399 return single_open(filp, sched_show, inode);
1402 static const struct file_operations proc_pid_sched_operations = {
1405 .write = sched_write,
1406 .llseek = seq_lseek,
1407 .release = single_release,
1412 #ifdef CONFIG_SCHED_AUTOGROUP
1414 * Print out autogroup related information:
1416 static int sched_autogroup_show(struct seq_file *m, void *v)
1418 struct inode *inode = m->private;
1419 struct task_struct *p;
1421 p = get_proc_task(inode);
1424 proc_sched_autogroup_show_task(p, m);
1432 sched_autogroup_write(struct file *file, const char __user *buf,
1433 size_t count, loff_t *offset)
1435 struct inode *inode = file_inode(file);
1436 struct task_struct *p;
1437 char buffer[PROC_NUMBUF];
1441 memset(buffer, 0, sizeof(buffer));
1442 if (count > sizeof(buffer) - 1)
1443 count = sizeof(buffer) - 1;
1444 if (copy_from_user(buffer, buf, count))
1447 err = kstrtoint(strstrip(buffer), 0, &nice);
1451 p = get_proc_task(inode);
1455 err = proc_sched_autogroup_set_nice(p, nice);
1464 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1468 ret = single_open(filp, sched_autogroup_show, NULL);
1470 struct seq_file *m = filp->private_data;
1477 static const struct file_operations proc_pid_sched_autogroup_operations = {
1478 .open = sched_autogroup_open,
1480 .write = sched_autogroup_write,
1481 .llseek = seq_lseek,
1482 .release = single_release,
1485 #endif /* CONFIG_SCHED_AUTOGROUP */
1487 static ssize_t comm_write(struct file *file, const char __user *buf,
1488 size_t count, loff_t *offset)
1490 struct inode *inode = file_inode(file);
1491 struct task_struct *p;
1492 char buffer[TASK_COMM_LEN];
1493 const size_t maxlen = sizeof(buffer) - 1;
1495 memset(buffer, 0, sizeof(buffer));
1496 if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
1499 p = get_proc_task(inode);
1503 if (same_thread_group(current, p))
1504 set_task_comm(p, buffer);
1513 static int comm_show(struct seq_file *m, void *v)
1515 struct inode *inode = m->private;
1516 struct task_struct *p;
1518 p = get_proc_task(inode);
1523 seq_printf(m, "%s\n", p->comm);
1531 static int comm_open(struct inode *inode, struct file *filp)
1533 return single_open(filp, comm_show, inode);
1536 static const struct file_operations proc_pid_set_comm_operations = {
1539 .write = comm_write,
1540 .llseek = seq_lseek,
1541 .release = single_release,
1544 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1546 struct task_struct *task;
1547 struct mm_struct *mm;
1548 struct file *exe_file;
1550 task = get_proc_task(d_inode(dentry));
1553 mm = get_task_mm(task);
1554 put_task_struct(task);
1557 exe_file = get_mm_exe_file(mm);
1560 *exe_path = exe_file->f_path;
1561 path_get(&exe_file->f_path);
1568 static const char *proc_pid_follow_link(struct dentry *dentry, void **cookie)
1570 struct inode *inode = d_inode(dentry);
1572 int error = -EACCES;
1574 /* Are we allowed to snoop on the tasks file descriptors? */
1575 if (!proc_fd_access_allowed(inode))
1578 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1582 nd_jump_link(&path);
1585 return ERR_PTR(error);
1588 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1590 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1597 pathname = d_path(path, tmp, PAGE_SIZE);
1598 len = PTR_ERR(pathname);
1599 if (IS_ERR(pathname))
1601 len = tmp + PAGE_SIZE - 1 - pathname;
1605 if (copy_to_user(buffer, pathname, len))
1608 free_page((unsigned long)tmp);
1612 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1614 int error = -EACCES;
1615 struct inode *inode = d_inode(dentry);
1618 /* Are we allowed to snoop on the tasks file descriptors? */
1619 if (!proc_fd_access_allowed(inode))
1622 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1626 error = do_proc_readlink(&path, buffer, buflen);
1632 const struct inode_operations proc_pid_link_inode_operations = {
1633 .readlink = proc_pid_readlink,
1634 .follow_link = proc_pid_follow_link,
1635 .setattr = proc_setattr,
1639 /* building an inode */
1641 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1643 struct inode * inode;
1644 struct proc_inode *ei;
1645 const struct cred *cred;
1647 /* We need a new inode */
1649 inode = new_inode(sb);
1655 inode->i_ino = get_next_ino();
1656 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1657 inode->i_op = &proc_def_inode_operations;
1660 * grab the reference to task.
1662 ei->pid = get_task_pid(task, PIDTYPE_PID);
1666 if (task_dumpable(task)) {
1668 cred = __task_cred(task);
1669 inode->i_uid = cred->euid;
1670 inode->i_gid = cred->egid;
1673 security_task_to_inode(task, inode);
1683 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1685 struct inode *inode = d_inode(dentry);
1686 struct task_struct *task;
1687 const struct cred *cred;
1688 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
1690 generic_fillattr(inode, stat);
1693 stat->uid = GLOBAL_ROOT_UID;
1694 stat->gid = GLOBAL_ROOT_GID;
1695 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1697 if (!has_pid_permissions(pid, task, 2)) {
1700 * This doesn't prevent learning whether PID exists,
1701 * it only makes getattr() consistent with readdir().
1705 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1706 task_dumpable(task)) {
1707 cred = __task_cred(task);
1708 stat->uid = cred->euid;
1709 stat->gid = cred->egid;
1719 * Exceptional case: normally we are not allowed to unhash a busy
1720 * directory. In this case, however, we can do it - no aliasing problems
1721 * due to the way we treat inodes.
1723 * Rewrite the inode's ownerships here because the owning task may have
1724 * performed a setuid(), etc.
1726 * Before the /proc/pid/status file was created the only way to read
1727 * the effective uid of a /process was to stat /proc/pid. Reading
1728 * /proc/pid/status is slow enough that procps and other packages
1729 * kept stating /proc/pid. To keep the rules in /proc simple I have
1730 * made this apply to all per process world readable and executable
1733 int pid_revalidate(struct dentry *dentry, unsigned int flags)
1735 struct inode *inode;
1736 struct task_struct *task;
1737 const struct cred *cred;
1739 if (flags & LOOKUP_RCU)
1742 inode = d_inode(dentry);
1743 task = get_proc_task(inode);
1746 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1747 task_dumpable(task)) {
1749 cred = __task_cred(task);
1750 inode->i_uid = cred->euid;
1751 inode->i_gid = cred->egid;
1754 inode->i_uid = GLOBAL_ROOT_UID;
1755 inode->i_gid = GLOBAL_ROOT_GID;
1757 inode->i_mode &= ~(S_ISUID | S_ISGID);
1758 security_task_to_inode(task, inode);
1759 put_task_struct(task);
1765 static inline bool proc_inode_is_dead(struct inode *inode)
1767 return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1770 int pid_delete_dentry(const struct dentry *dentry)
1772 /* Is the task we represent dead?
1773 * If so, then don't put the dentry on the lru list,
1774 * kill it immediately.
1776 return proc_inode_is_dead(d_inode(dentry));
1779 const struct dentry_operations pid_dentry_operations =
1781 .d_revalidate = pid_revalidate,
1782 .d_delete = pid_delete_dentry,
1788 * Fill a directory entry.
1790 * If possible create the dcache entry and derive our inode number and
1791 * file type from dcache entry.
1793 * Since all of the proc inode numbers are dynamically generated, the inode
1794 * numbers do not exist until the inode is cache. This means creating the
1795 * the dcache entry in readdir is necessary to keep the inode numbers
1796 * reported by readdir in sync with the inode numbers reported
1799 bool proc_fill_cache(struct file *file, struct dir_context *ctx,
1800 const char *name, int len,
1801 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1803 struct dentry *child, *dir = file->f_path.dentry;
1804 struct qstr qname = QSTR_INIT(name, len);
1805 struct inode *inode;
1809 child = d_hash_and_lookup(dir, &qname);
1811 child = d_alloc(dir, &qname);
1813 goto end_instantiate;
1814 if (instantiate(d_inode(dir), child, task, ptr) < 0) {
1816 goto end_instantiate;
1819 inode = d_inode(child);
1821 type = inode->i_mode >> 12;
1823 return dir_emit(ctx, name, len, ino, type);
1826 return dir_emit(ctx, name, len, 1, DT_UNKNOWN);
1830 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1831 * which represent vma start and end addresses.
1833 static int dname_to_vma_addr(struct dentry *dentry,
1834 unsigned long *start, unsigned long *end)
1836 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1842 static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
1844 unsigned long vm_start, vm_end;
1845 bool exact_vma_exists = false;
1846 struct mm_struct *mm = NULL;
1847 struct task_struct *task;
1848 const struct cred *cred;
1849 struct inode *inode;
1852 if (flags & LOOKUP_RCU)
1855 inode = d_inode(dentry);
1856 task = get_proc_task(inode);
1860 mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
1861 if (IS_ERR_OR_NULL(mm))
1864 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1865 down_read(&mm->mmap_sem);
1866 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1867 up_read(&mm->mmap_sem);
1872 if (exact_vma_exists) {
1873 if (task_dumpable(task)) {
1875 cred = __task_cred(task);
1876 inode->i_uid = cred->euid;
1877 inode->i_gid = cred->egid;
1880 inode->i_uid = GLOBAL_ROOT_UID;
1881 inode->i_gid = GLOBAL_ROOT_GID;
1883 security_task_to_inode(task, inode);
1888 put_task_struct(task);
1894 static const struct dentry_operations tid_map_files_dentry_operations = {
1895 .d_revalidate = map_files_d_revalidate,
1896 .d_delete = pid_delete_dentry,
1899 static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
1901 unsigned long vm_start, vm_end;
1902 struct vm_area_struct *vma;
1903 struct task_struct *task;
1904 struct mm_struct *mm;
1908 task = get_proc_task(d_inode(dentry));
1912 mm = get_task_mm(task);
1913 put_task_struct(task);
1917 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
1922 down_read(&mm->mmap_sem);
1923 vma = find_exact_vma(mm, vm_start, vm_end);
1924 if (vma && vma->vm_file) {
1925 *path = vma->vm_file->f_path;
1929 up_read(&mm->mmap_sem);
1937 struct map_files_info {
1940 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
1944 * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the
1945 * symlinks may be used to bypass permissions on ancestor directories in the
1946 * path to the file in question.
1949 proc_map_files_follow_link(struct dentry *dentry, void **cookie)
1951 if (!capable(CAP_SYS_ADMIN))
1952 return ERR_PTR(-EPERM);
1954 return proc_pid_follow_link(dentry, NULL);
1958 * Identical to proc_pid_link_inode_operations except for follow_link()
1960 static const struct inode_operations proc_map_files_link_inode_operations = {
1961 .readlink = proc_pid_readlink,
1962 .follow_link = proc_map_files_follow_link,
1963 .setattr = proc_setattr,
1967 proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
1968 struct task_struct *task, const void *ptr)
1970 fmode_t mode = (fmode_t)(unsigned long)ptr;
1971 struct proc_inode *ei;
1972 struct inode *inode;
1974 inode = proc_pid_make_inode(dir->i_sb, task);
1979 ei->op.proc_get_link = proc_map_files_get_link;
1981 inode->i_op = &proc_map_files_link_inode_operations;
1983 inode->i_mode = S_IFLNK;
1985 if (mode & FMODE_READ)
1986 inode->i_mode |= S_IRUSR;
1987 if (mode & FMODE_WRITE)
1988 inode->i_mode |= S_IWUSR;
1990 d_set_d_op(dentry, &tid_map_files_dentry_operations);
1991 d_add(dentry, inode);
1996 static struct dentry *proc_map_files_lookup(struct inode *dir,
1997 struct dentry *dentry, unsigned int flags)
1999 unsigned long vm_start, vm_end;
2000 struct vm_area_struct *vma;
2001 struct task_struct *task;
2003 struct mm_struct *mm;
2006 task = get_proc_task(dir);
2011 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2015 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2018 mm = get_task_mm(task);
2022 down_read(&mm->mmap_sem);
2023 vma = find_exact_vma(mm, vm_start, vm_end);
2028 result = proc_map_files_instantiate(dir, dentry, task,
2029 (void *)(unsigned long)vma->vm_file->f_mode);
2032 up_read(&mm->mmap_sem);
2035 put_task_struct(task);
2037 return ERR_PTR(result);
2040 static const struct inode_operations proc_map_files_inode_operations = {
2041 .lookup = proc_map_files_lookup,
2042 .permission = proc_fd_permission,
2043 .setattr = proc_setattr,
2047 proc_map_files_readdir(struct file *file, struct dir_context *ctx)
2049 struct vm_area_struct *vma;
2050 struct task_struct *task;
2051 struct mm_struct *mm;
2052 unsigned long nr_files, pos, i;
2053 struct flex_array *fa = NULL;
2054 struct map_files_info info;
2055 struct map_files_info *p;
2059 task = get_proc_task(file_inode(file));
2064 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2068 if (!dir_emit_dots(file, ctx))
2071 mm = get_task_mm(task);
2074 down_read(&mm->mmap_sem);
2079 * We need two passes here:
2081 * 1) Collect vmas of mapped files with mmap_sem taken
2082 * 2) Release mmap_sem and instantiate entries
2084 * otherwise we get lockdep complained, since filldir()
2085 * routine might require mmap_sem taken in might_fault().
2088 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2089 if (vma->vm_file && ++pos > ctx->pos)
2094 fa = flex_array_alloc(sizeof(info), nr_files,
2096 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2100 flex_array_free(fa);
2101 up_read(&mm->mmap_sem);
2105 for (i = 0, vma = mm->mmap, pos = 2; vma;
2106 vma = vma->vm_next) {
2109 if (++pos <= ctx->pos)
2112 info.mode = vma->vm_file->f_mode;
2113 info.len = snprintf(info.name,
2114 sizeof(info.name), "%lx-%lx",
2115 vma->vm_start, vma->vm_end);
2116 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2120 up_read(&mm->mmap_sem);
2122 for (i = 0; i < nr_files; i++) {
2123 p = flex_array_get(fa, i);
2124 if (!proc_fill_cache(file, ctx,
2126 proc_map_files_instantiate,
2128 (void *)(unsigned long)p->mode))
2133 flex_array_free(fa);
2137 put_task_struct(task);
2142 static const struct file_operations proc_map_files_operations = {
2143 .read = generic_read_dir,
2144 .iterate = proc_map_files_readdir,
2145 .llseek = default_llseek,
2148 struct timers_private {
2150 struct task_struct *task;
2151 struct sighand_struct *sighand;
2152 struct pid_namespace *ns;
2153 unsigned long flags;
2156 static void *timers_start(struct seq_file *m, loff_t *pos)
2158 struct timers_private *tp = m->private;
2160 tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2162 return ERR_PTR(-ESRCH);
2164 tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2166 return ERR_PTR(-ESRCH);
2168 return seq_list_start(&tp->task->signal->posix_timers, *pos);
2171 static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2173 struct timers_private *tp = m->private;
2174 return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2177 static void timers_stop(struct seq_file *m, void *v)
2179 struct timers_private *tp = m->private;
2182 unlock_task_sighand(tp->task, &tp->flags);
2187 put_task_struct(tp->task);
2192 static int show_timer(struct seq_file *m, void *v)
2194 struct k_itimer *timer;
2195 struct timers_private *tp = m->private;
2197 static const char * const nstr[] = {
2198 [SIGEV_SIGNAL] = "signal",
2199 [SIGEV_NONE] = "none",
2200 [SIGEV_THREAD] = "thread",
2203 timer = list_entry((struct list_head *)v, struct k_itimer, list);
2204 notify = timer->it_sigev_notify;
2206 seq_printf(m, "ID: %d\n", timer->it_id);
2207 seq_printf(m, "signal: %d/%p\n",
2208 timer->sigq->info.si_signo,
2209 timer->sigq->info.si_value.sival_ptr);
2210 seq_printf(m, "notify: %s/%s.%d\n",
2211 nstr[notify & ~SIGEV_THREAD_ID],
2212 (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2213 pid_nr_ns(timer->it_pid, tp->ns));
2214 seq_printf(m, "ClockID: %d\n", timer->it_clock);
2219 static const struct seq_operations proc_timers_seq_ops = {
2220 .start = timers_start,
2221 .next = timers_next,
2222 .stop = timers_stop,
2226 static int proc_timers_open(struct inode *inode, struct file *file)
2228 struct timers_private *tp;
2230 tp = __seq_open_private(file, &proc_timers_seq_ops,
2231 sizeof(struct timers_private));
2235 tp->pid = proc_pid(inode);
2236 tp->ns = inode->i_sb->s_fs_info;
2240 static const struct file_operations proc_timers_operations = {
2241 .open = proc_timers_open,
2243 .llseek = seq_lseek,
2244 .release = seq_release_private,
2247 static int proc_pident_instantiate(struct inode *dir,
2248 struct dentry *dentry, struct task_struct *task, const void *ptr)
2250 const struct pid_entry *p = ptr;
2251 struct inode *inode;
2252 struct proc_inode *ei;
2254 inode = proc_pid_make_inode(dir->i_sb, task);
2259 inode->i_mode = p->mode;
2260 if (S_ISDIR(inode->i_mode))
2261 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2263 inode->i_op = p->iop;
2265 inode->i_fop = p->fop;
2267 d_set_d_op(dentry, &pid_dentry_operations);
2268 d_add(dentry, inode);
2269 /* Close the race of the process dying before we return the dentry */
2270 if (pid_revalidate(dentry, 0))
2276 static struct dentry *proc_pident_lookup(struct inode *dir,
2277 struct dentry *dentry,
2278 const struct pid_entry *ents,
2282 struct task_struct *task = get_proc_task(dir);
2283 const struct pid_entry *p, *last;
2291 * Yes, it does not scale. And it should not. Don't add
2292 * new entries into /proc/<tgid>/ without very good reasons.
2294 last = &ents[nents - 1];
2295 for (p = ents; p <= last; p++) {
2296 if (p->len != dentry->d_name.len)
2298 if (!memcmp(dentry->d_name.name, p->name, p->len))
2304 error = proc_pident_instantiate(dir, dentry, task, p);
2306 put_task_struct(task);
2308 return ERR_PTR(error);
2311 static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
2312 const struct pid_entry *ents, unsigned int nents)
2314 struct task_struct *task = get_proc_task(file_inode(file));
2315 const struct pid_entry *p;
2320 if (!dir_emit_dots(file, ctx))
2323 if (ctx->pos >= nents + 2)
2326 for (p = ents + (ctx->pos - 2); p <= ents + nents - 1; p++) {
2327 if (!proc_fill_cache(file, ctx, p->name, p->len,
2328 proc_pident_instantiate, task, p))
2333 put_task_struct(task);
2337 #ifdef CONFIG_SECURITY
2338 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2339 size_t count, loff_t *ppos)
2341 struct inode * inode = file_inode(file);
2344 struct task_struct *task = get_proc_task(inode);
2349 length = security_getprocattr(task,
2350 (char*)file->f_path.dentry->d_name.name,
2352 put_task_struct(task);
2354 length = simple_read_from_buffer(buf, count, ppos, p, length);
2359 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2360 size_t count, loff_t *ppos)
2362 struct inode * inode = file_inode(file);
2365 struct task_struct *task = get_proc_task(inode);
2370 if (count > PAGE_SIZE)
2373 /* No partial writes. */
2379 page = (char*)__get_free_page(GFP_TEMPORARY);
2384 if (copy_from_user(page, buf, count))
2387 /* Guard against adverse ptrace interaction */
2388 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2392 length = security_setprocattr(task,
2393 (char*)file->f_path.dentry->d_name.name,
2394 (void*)page, count);
2395 mutex_unlock(&task->signal->cred_guard_mutex);
2397 free_page((unsigned long) page);
2399 put_task_struct(task);
2404 static const struct file_operations proc_pid_attr_operations = {
2405 .read = proc_pid_attr_read,
2406 .write = proc_pid_attr_write,
2407 .llseek = generic_file_llseek,
2410 static const struct pid_entry attr_dir_stuff[] = {
2411 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2412 REG("prev", S_IRUGO, proc_pid_attr_operations),
2413 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2414 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2415 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2416 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2419 static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
2421 return proc_pident_readdir(file, ctx,
2422 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2425 static const struct file_operations proc_attr_dir_operations = {
2426 .read = generic_read_dir,
2427 .iterate = proc_attr_dir_readdir,
2428 .llseek = default_llseek,
2431 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2432 struct dentry *dentry, unsigned int flags)
2434 return proc_pident_lookup(dir, dentry,
2435 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2438 static const struct inode_operations proc_attr_dir_inode_operations = {
2439 .lookup = proc_attr_dir_lookup,
2440 .getattr = pid_getattr,
2441 .setattr = proc_setattr,
2446 #ifdef CONFIG_ELF_CORE
2447 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2448 size_t count, loff_t *ppos)
2450 struct task_struct *task = get_proc_task(file_inode(file));
2451 struct mm_struct *mm;
2452 char buffer[PROC_NUMBUF];
2460 mm = get_task_mm(task);
2462 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2463 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2464 MMF_DUMP_FILTER_SHIFT));
2466 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2469 put_task_struct(task);
2474 static ssize_t proc_coredump_filter_write(struct file *file,
2475 const char __user *buf,
2479 struct task_struct *task;
2480 struct mm_struct *mm;
2486 ret = kstrtouint_from_user(buf, count, 0, &val);
2491 task = get_proc_task(file_inode(file));
2495 mm = get_task_mm(task);
2500 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2502 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2504 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2509 put_task_struct(task);
2516 static const struct file_operations proc_coredump_filter_operations = {
2517 .read = proc_coredump_filter_read,
2518 .write = proc_coredump_filter_write,
2519 .llseek = generic_file_llseek,
2523 #ifdef CONFIG_TASK_IO_ACCOUNTING
2524 static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
2526 struct task_io_accounting acct = task->ioac;
2527 unsigned long flags;
2530 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2534 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) {
2539 if (whole && lock_task_sighand(task, &flags)) {
2540 struct task_struct *t = task;
2542 task_io_accounting_add(&acct, &task->signal->ioac);
2543 while_each_thread(task, t)
2544 task_io_accounting_add(&acct, &t->ioac);
2546 unlock_task_sighand(task, &flags);
2553 "read_bytes: %llu\n"
2554 "write_bytes: %llu\n"
2555 "cancelled_write_bytes: %llu\n",
2556 (unsigned long long)acct.rchar,
2557 (unsigned long long)acct.wchar,
2558 (unsigned long long)acct.syscr,
2559 (unsigned long long)acct.syscw,
2560 (unsigned long long)acct.read_bytes,
2561 (unsigned long long)acct.write_bytes,
2562 (unsigned long long)acct.cancelled_write_bytes);
2566 mutex_unlock(&task->signal->cred_guard_mutex);
2570 static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2571 struct pid *pid, struct task_struct *task)
2573 return do_io_accounting(task, m, 0);
2576 static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2577 struct pid *pid, struct task_struct *task)
2579 return do_io_accounting(task, m, 1);
2581 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2583 #ifdef CONFIG_USER_NS
2584 static int proc_id_map_open(struct inode *inode, struct file *file,
2585 const struct seq_operations *seq_ops)
2587 struct user_namespace *ns = NULL;
2588 struct task_struct *task;
2589 struct seq_file *seq;
2592 task = get_proc_task(inode);
2595 ns = get_user_ns(task_cred_xxx(task, user_ns));
2597 put_task_struct(task);
2602 ret = seq_open(file, seq_ops);
2606 seq = file->private_data;
2616 static int proc_id_map_release(struct inode *inode, struct file *file)
2618 struct seq_file *seq = file->private_data;
2619 struct user_namespace *ns = seq->private;
2621 return seq_release(inode, file);
2624 static int proc_uid_map_open(struct inode *inode, struct file *file)
2626 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2629 static int proc_gid_map_open(struct inode *inode, struct file *file)
2631 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2634 static int proc_projid_map_open(struct inode *inode, struct file *file)
2636 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2639 static const struct file_operations proc_uid_map_operations = {
2640 .open = proc_uid_map_open,
2641 .write = proc_uid_map_write,
2643 .llseek = seq_lseek,
2644 .release = proc_id_map_release,
2647 static const struct file_operations proc_gid_map_operations = {
2648 .open = proc_gid_map_open,
2649 .write = proc_gid_map_write,
2651 .llseek = seq_lseek,
2652 .release = proc_id_map_release,
2655 static const struct file_operations proc_projid_map_operations = {
2656 .open = proc_projid_map_open,
2657 .write = proc_projid_map_write,
2659 .llseek = seq_lseek,
2660 .release = proc_id_map_release,
2663 static int proc_setgroups_open(struct inode *inode, struct file *file)
2665 struct user_namespace *ns = NULL;
2666 struct task_struct *task;
2670 task = get_proc_task(inode);
2673 ns = get_user_ns(task_cred_xxx(task, user_ns));
2675 put_task_struct(task);
2680 if (file->f_mode & FMODE_WRITE) {
2682 if (!ns_capable(ns, CAP_SYS_ADMIN))
2686 ret = single_open(file, &proc_setgroups_show, ns);
2697 static int proc_setgroups_release(struct inode *inode, struct file *file)
2699 struct seq_file *seq = file->private_data;
2700 struct user_namespace *ns = seq->private;
2701 int ret = single_release(inode, file);
2706 static const struct file_operations proc_setgroups_operations = {
2707 .open = proc_setgroups_open,
2708 .write = proc_setgroups_write,
2710 .llseek = seq_lseek,
2711 .release = proc_setgroups_release,
2713 #endif /* CONFIG_USER_NS */
2715 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2716 struct pid *pid, struct task_struct *task)
2718 int err = lock_trace(task);
2720 seq_printf(m, "%08x\n", task->personality);
2729 static const struct file_operations proc_task_operations;
2730 static const struct inode_operations proc_task_inode_operations;
2732 static const struct pid_entry tgid_base_stuff[] = {
2733 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2734 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2735 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
2736 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2737 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2739 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2741 REG("environ", S_IRUSR, proc_environ_operations),
2742 ONE("auxv", S_IRUSR, proc_pid_auxv),
2743 ONE("status", S_IRUGO, proc_pid_status),
2744 ONE("personality", S_IRUSR, proc_pid_personality),
2745 ONE("limits", S_IRUGO, proc_pid_limits),
2746 #ifdef CONFIG_SCHED_DEBUG
2747 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2749 #ifdef CONFIG_SCHED_AUTOGROUP
2750 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2752 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2753 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2754 ONE("syscall", S_IRUSR, proc_pid_syscall),
2756 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
2757 ONE("stat", S_IRUGO, proc_tgid_stat),
2758 ONE("statm", S_IRUGO, proc_pid_statm),
2759 REG("maps", S_IRUGO, proc_pid_maps_operations),
2761 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
2763 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2764 LNK("cwd", proc_cwd_link),
2765 LNK("root", proc_root_link),
2766 LNK("exe", proc_exe_link),
2767 REG("mounts", S_IRUGO, proc_mounts_operations),
2768 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2769 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2770 #ifdef CONFIG_PROC_PAGE_MONITOR
2771 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2772 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
2773 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2775 #ifdef CONFIG_SECURITY
2776 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2778 #ifdef CONFIG_KALLSYMS
2779 ONE("wchan", S_IRUGO, proc_pid_wchan),
2781 #ifdef CONFIG_STACKTRACE
2782 ONE("stack", S_IRUSR, proc_pid_stack),
2784 #ifdef CONFIG_SCHED_INFO
2785 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
2787 #ifdef CONFIG_LATENCYTOP
2788 REG("latency", S_IRUGO, proc_lstats_operations),
2790 #ifdef CONFIG_PROC_PID_CPUSET
2791 ONE("cpuset", S_IRUGO, proc_cpuset_show),
2793 #ifdef CONFIG_CGROUPS
2794 ONE("cgroup", S_IRUGO, proc_cgroup_show),
2796 ONE("oom_score", S_IRUGO, proc_oom_score),
2797 REG("oom_adj", S_IRUSR, proc_oom_adj_operations),
2798 REG("oom_score_adj", S_IRUSR, proc_oom_score_adj_operations),
2799 #ifdef CONFIG_AUDITSYSCALL
2800 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2801 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2803 #ifdef CONFIG_FAULT_INJECTION
2804 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2806 #ifdef CONFIG_ELF_CORE
2807 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2809 #ifdef CONFIG_TASK_IO_ACCOUNTING
2810 ONE("io", S_IRUSR, proc_tgid_io_accounting),
2812 #ifdef CONFIG_HARDWALL
2813 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
2815 #ifdef CONFIG_USER_NS
2816 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
2817 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
2818 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
2819 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
2821 #ifdef CONFIG_CHECKPOINT_RESTORE
2822 REG("timers", S_IRUGO, proc_timers_operations),
2826 static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
2828 return proc_pident_readdir(file, ctx,
2829 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2832 static const struct file_operations proc_tgid_base_operations = {
2833 .read = generic_read_dir,
2834 .iterate = proc_tgid_base_readdir,
2835 .llseek = default_llseek,
2838 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2840 return proc_pident_lookup(dir, dentry,
2841 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2844 static const struct inode_operations proc_tgid_base_inode_operations = {
2845 .lookup = proc_tgid_base_lookup,
2846 .getattr = pid_getattr,
2847 .setattr = proc_setattr,
2848 .permission = proc_pid_permission,
2851 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2853 struct dentry *dentry, *leader, *dir;
2854 char buf[PROC_NUMBUF];
2858 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2859 /* no ->d_hash() rejects on procfs */
2860 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2862 d_invalidate(dentry);
2870 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2871 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2876 name.len = strlen(name.name);
2877 dir = d_hash_and_lookup(leader, &name);
2879 goto out_put_leader;
2882 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2883 dentry = d_hash_and_lookup(dir, &name);
2885 d_invalidate(dentry);
2897 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2898 * @task: task that should be flushed.
2900 * When flushing dentries from proc, one needs to flush them from global
2901 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2902 * in. This call is supposed to do all of this job.
2904 * Looks in the dcache for
2906 * /proc/@tgid/task/@pid
2907 * if either directory is present flushes it and all of it'ts children
2910 * It is safe and reasonable to cache /proc entries for a task until
2911 * that task exits. After that they just clog up the dcache with
2912 * useless entries, possibly causing useful dcache entries to be
2913 * flushed instead. This routine is proved to flush those useless
2914 * dcache entries at process exit time.
2916 * NOTE: This routine is just an optimization so it does not guarantee
2917 * that no dcache entries will exist at process exit time it
2918 * just makes it very unlikely that any will persist.
2921 void proc_flush_task(struct task_struct *task)
2924 struct pid *pid, *tgid;
2927 pid = task_pid(task);
2928 tgid = task_tgid(task);
2930 for (i = 0; i <= pid->level; i++) {
2931 upid = &pid->numbers[i];
2932 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2933 tgid->numbers[i].nr);
2937 static int proc_pid_instantiate(struct inode *dir,
2938 struct dentry * dentry,
2939 struct task_struct *task, const void *ptr)
2941 struct inode *inode;
2943 inode = proc_pid_make_inode(dir->i_sb, task);
2947 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2948 inode->i_op = &proc_tgid_base_inode_operations;
2949 inode->i_fop = &proc_tgid_base_operations;
2950 inode->i_flags|=S_IMMUTABLE;
2952 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
2953 ARRAY_SIZE(tgid_base_stuff)));
2955 d_set_d_op(dentry, &pid_dentry_operations);
2957 d_add(dentry, inode);
2958 /* Close the race of the process dying before we return the dentry */
2959 if (pid_revalidate(dentry, 0))
2965 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
2967 int result = -ENOENT;
2968 struct task_struct *task;
2970 struct pid_namespace *ns;
2972 tgid = name_to_int(&dentry->d_name);
2976 ns = dentry->d_sb->s_fs_info;
2978 task = find_task_by_pid_ns(tgid, ns);
2980 get_task_struct(task);
2985 result = proc_pid_instantiate(dir, dentry, task, NULL);
2986 put_task_struct(task);
2988 return ERR_PTR(result);
2992 * Find the first task with tgid >= tgid
2997 struct task_struct *task;
2999 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3004 put_task_struct(iter.task);
3008 pid = find_ge_pid(iter.tgid, ns);
3010 iter.tgid = pid_nr_ns(pid, ns);
3011 iter.task = pid_task(pid, PIDTYPE_PID);
3012 /* What we to know is if the pid we have find is the
3013 * pid of a thread_group_leader. Testing for task
3014 * being a thread_group_leader is the obvious thing
3015 * todo but there is a window when it fails, due to
3016 * the pid transfer logic in de_thread.
3018 * So we perform the straight forward test of seeing
3019 * if the pid we have found is the pid of a thread
3020 * group leader, and don't worry if the task we have
3021 * found doesn't happen to be a thread group leader.
3022 * As we don't care in the case of readdir.
3024 if (!iter.task || !has_group_leader_pid(iter.task)) {
3028 get_task_struct(iter.task);
3034 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3036 /* for the /proc/ directory itself, after non-process stuff has been done */
3037 int proc_pid_readdir(struct file *file, struct dir_context *ctx)
3039 struct tgid_iter iter;
3040 struct pid_namespace *ns = file_inode(file)->i_sb->s_fs_info;
3041 loff_t pos = ctx->pos;
3043 if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
3046 if (pos == TGID_OFFSET - 2) {
3047 struct inode *inode = d_inode(ns->proc_self);
3048 if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
3050 ctx->pos = pos = pos + 1;
3052 if (pos == TGID_OFFSET - 1) {
3053 struct inode *inode = d_inode(ns->proc_thread_self);
3054 if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3056 ctx->pos = pos = pos + 1;
3058 iter.tgid = pos - TGID_OFFSET;
3060 for (iter = next_tgid(ns, iter);
3062 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3063 char name[PROC_NUMBUF];
3065 if (!has_pid_permissions(ns, iter.task, 2))
3068 len = snprintf(name, sizeof(name), "%d", iter.tgid);
3069 ctx->pos = iter.tgid + TGID_OFFSET;
3070 if (!proc_fill_cache(file, ctx, name, len,
3071 proc_pid_instantiate, iter.task, NULL)) {
3072 put_task_struct(iter.task);
3076 ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
3083 static const struct pid_entry tid_base_stuff[] = {
3084 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3085 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3086 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3088 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3090 REG("environ", S_IRUSR, proc_environ_operations),
3091 ONE("auxv", S_IRUSR, proc_pid_auxv),
3092 ONE("status", S_IRUGO, proc_pid_status),
3093 ONE("personality", S_IRUSR, proc_pid_personality),
3094 ONE("limits", S_IRUGO, proc_pid_limits),
3095 #ifdef CONFIG_SCHED_DEBUG
3096 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3098 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3099 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3100 ONE("syscall", S_IRUSR, proc_pid_syscall),
3102 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
3103 ONE("stat", S_IRUGO, proc_tid_stat),
3104 ONE("statm", S_IRUGO, proc_pid_statm),
3105 REG("maps", S_IRUGO, proc_tid_maps_operations),
3106 #ifdef CONFIG_PROC_CHILDREN
3107 REG("children", S_IRUGO, proc_tid_children_operations),
3110 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
3112 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3113 LNK("cwd", proc_cwd_link),
3114 LNK("root", proc_root_link),
3115 LNK("exe", proc_exe_link),
3116 REG("mounts", S_IRUGO, proc_mounts_operations),
3117 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3118 #ifdef CONFIG_PROC_PAGE_MONITOR
3119 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3120 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
3121 REG("pagemap", S_IRUSR, proc_pagemap_operations),
3123 #ifdef CONFIG_SECURITY
3124 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3126 #ifdef CONFIG_KALLSYMS
3127 ONE("wchan", S_IRUGO, proc_pid_wchan),
3129 #ifdef CONFIG_STACKTRACE
3130 ONE("stack", S_IRUSR, proc_pid_stack),
3132 #ifdef CONFIG_SCHED_INFO
3133 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
3135 #ifdef CONFIG_LATENCYTOP
3136 REG("latency", S_IRUGO, proc_lstats_operations),
3138 #ifdef CONFIG_PROC_PID_CPUSET
3139 ONE("cpuset", S_IRUGO, proc_cpuset_show),
3141 #ifdef CONFIG_CGROUPS
3142 ONE("cgroup", S_IRUGO, proc_cgroup_show),
3144 ONE("oom_score", S_IRUGO, proc_oom_score),
3145 REG("oom_adj", S_IRUSR, proc_oom_adj_operations),
3146 REG("oom_score_adj", S_IRUSR, proc_oom_score_adj_operations),
3147 #ifdef CONFIG_AUDITSYSCALL
3148 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3149 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3151 #ifdef CONFIG_FAULT_INJECTION
3152 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3154 #ifdef CONFIG_TASK_IO_ACCOUNTING
3155 ONE("io", S_IRUSR, proc_tid_io_accounting),
3157 #ifdef CONFIG_HARDWALL
3158 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
3160 #ifdef CONFIG_USER_NS
3161 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3162 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3163 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3164 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
3168 static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
3170 return proc_pident_readdir(file, ctx,
3171 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3174 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3176 return proc_pident_lookup(dir, dentry,
3177 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3180 static const struct file_operations proc_tid_base_operations = {
3181 .read = generic_read_dir,
3182 .iterate = proc_tid_base_readdir,
3183 .llseek = default_llseek,
3186 static const struct inode_operations proc_tid_base_inode_operations = {
3187 .lookup = proc_tid_base_lookup,
3188 .getattr = pid_getattr,
3189 .setattr = proc_setattr,
3192 static int proc_task_instantiate(struct inode *dir,
3193 struct dentry *dentry, struct task_struct *task, const void *ptr)
3195 struct inode *inode;
3196 inode = proc_pid_make_inode(dir->i_sb, task);
3200 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3201 inode->i_op = &proc_tid_base_inode_operations;
3202 inode->i_fop = &proc_tid_base_operations;
3203 inode->i_flags|=S_IMMUTABLE;
3205 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3206 ARRAY_SIZE(tid_base_stuff)));
3208 d_set_d_op(dentry, &pid_dentry_operations);
3210 d_add(dentry, inode);
3211 /* Close the race of the process dying before we return the dentry */
3212 if (pid_revalidate(dentry, 0))
3218 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3220 int result = -ENOENT;
3221 struct task_struct *task;
3222 struct task_struct *leader = get_proc_task(dir);
3224 struct pid_namespace *ns;
3229 tid = name_to_int(&dentry->d_name);
3233 ns = dentry->d_sb->s_fs_info;
3235 task = find_task_by_pid_ns(tid, ns);
3237 get_task_struct(task);
3241 if (!same_thread_group(leader, task))
3244 result = proc_task_instantiate(dir, dentry, task, NULL);
3246 put_task_struct(task);
3248 put_task_struct(leader);
3250 return ERR_PTR(result);
3254 * Find the first tid of a thread group to return to user space.
3256 * Usually this is just the thread group leader, but if the users
3257 * buffer was too small or there was a seek into the middle of the
3258 * directory we have more work todo.
3260 * In the case of a short read we start with find_task_by_pid.
3262 * In the case of a seek we start with the leader and walk nr
3265 static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3266 struct pid_namespace *ns)
3268 struct task_struct *pos, *task;
3269 unsigned long nr = f_pos;
3271 if (nr != f_pos) /* 32bit overflow? */
3275 task = pid_task(pid, PIDTYPE_PID);
3279 /* Attempt to start with the tid of a thread */
3281 pos = find_task_by_pid_ns(tid, ns);
3282 if (pos && same_thread_group(pos, task))
3286 /* If nr exceeds the number of threads there is nothing todo */
3287 if (nr >= get_nr_threads(task))
3290 /* If we haven't found our starting place yet start
3291 * with the leader and walk nr threads forward.
3293 pos = task = task->group_leader;
3297 } while_each_thread(task, pos);
3302 get_task_struct(pos);
3309 * Find the next thread in the thread list.
3310 * Return NULL if there is an error or no next thread.
3312 * The reference to the input task_struct is released.
3314 static struct task_struct *next_tid(struct task_struct *start)
3316 struct task_struct *pos = NULL;
3318 if (pid_alive(start)) {
3319 pos = next_thread(start);
3320 if (thread_group_leader(pos))
3323 get_task_struct(pos);
3326 put_task_struct(start);
3330 /* for the /proc/TGID/task/ directories */
3331 static int proc_task_readdir(struct file *file, struct dir_context *ctx)
3333 struct inode *inode = file_inode(file);
3334 struct task_struct *task;
3335 struct pid_namespace *ns;
3338 if (proc_inode_is_dead(inode))
3341 if (!dir_emit_dots(file, ctx))
3344 /* f_version caches the tgid value that the last readdir call couldn't
3345 * return. lseek aka telldir automagically resets f_version to 0.
3347 ns = inode->i_sb->s_fs_info;
3348 tid = (int)file->f_version;
3349 file->f_version = 0;
3350 for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
3352 task = next_tid(task), ctx->pos++) {
3353 char name[PROC_NUMBUF];
3355 tid = task_pid_nr_ns(task, ns);
3356 len = snprintf(name, sizeof(name), "%d", tid);
3357 if (!proc_fill_cache(file, ctx, name, len,
3358 proc_task_instantiate, task, NULL)) {
3359 /* returning this tgid failed, save it as the first
3360 * pid for the next readir call */
3361 file->f_version = (u64)tid;
3362 put_task_struct(task);
3370 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3372 struct inode *inode = d_inode(dentry);
3373 struct task_struct *p = get_proc_task(inode);
3374 generic_fillattr(inode, stat);
3377 stat->nlink += get_nr_threads(p);
3384 static const struct inode_operations proc_task_inode_operations = {
3385 .lookup = proc_task_lookup,
3386 .getattr = proc_task_getattr,
3387 .setattr = proc_setattr,
3388 .permission = proc_pid_permission,
3391 static const struct file_operations proc_task_operations = {
3392 .read = generic_read_dir,
3393 .iterate = proc_task_readdir,
3394 .llseek = default_llseek,