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;
956 unsigned long env_start, env_end;
958 /* Ensure the process spawned far enough to have an environment. */
959 if (!mm || !mm->env_end)
962 page = (char *)__get_free_page(GFP_TEMPORARY);
967 if (!atomic_inc_not_zero(&mm->mm_users))
970 down_read(&mm->mmap_sem);
971 env_start = mm->env_start;
972 env_end = mm->env_end;
973 up_read(&mm->mmap_sem);
976 size_t this_len, max_len;
979 if (src >= (env_end - env_start))
982 this_len = env_end - (env_start + src);
984 max_len = min_t(size_t, PAGE_SIZE, count);
985 this_len = min(max_len, this_len);
987 retval = access_remote_vm(mm, (env_start + src),
995 if (copy_to_user(buf, page, retval)) {
1009 free_page((unsigned long) page);
1013 static const struct file_operations proc_environ_operations = {
1014 .open = environ_open,
1015 .read = environ_read,
1016 .llseek = generic_file_llseek,
1017 .release = mem_release,
1020 static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1023 struct task_struct *task = get_proc_task(file_inode(file));
1024 char buffer[PROC_NUMBUF];
1025 int oom_adj = OOM_ADJUST_MIN;
1027 unsigned long flags;
1031 if (lock_task_sighand(task, &flags)) {
1032 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1033 oom_adj = OOM_ADJUST_MAX;
1035 oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1037 unlock_task_sighand(task, &flags);
1039 put_task_struct(task);
1040 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1041 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1045 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1046 * kernels. The effective policy is defined by oom_score_adj, which has a
1047 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1048 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1049 * Processes that become oom disabled via oom_adj will still be oom disabled
1050 * with this implementation.
1052 * oom_adj cannot be removed since existing userspace binaries use it.
1054 static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1055 size_t count, loff_t *ppos)
1057 struct task_struct *task;
1058 char buffer[PROC_NUMBUF];
1060 unsigned long flags;
1063 memset(buffer, 0, sizeof(buffer));
1064 if (count > sizeof(buffer) - 1)
1065 count = sizeof(buffer) - 1;
1066 if (copy_from_user(buffer, buf, count)) {
1071 err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1074 if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1075 oom_adj != OOM_DISABLE) {
1080 task = get_proc_task(file_inode(file));
1092 if (!lock_task_sighand(task, &flags)) {
1098 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1099 * value is always attainable.
1101 if (oom_adj == OOM_ADJUST_MAX)
1102 oom_adj = OOM_SCORE_ADJ_MAX;
1104 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1106 if (oom_adj < task->signal->oom_score_adj &&
1107 !capable(CAP_SYS_RESOURCE)) {
1113 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1114 * /proc/pid/oom_score_adj instead.
1116 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1117 current->comm, task_pid_nr(current), task_pid_nr(task),
1120 task->signal->oom_score_adj = oom_adj;
1121 trace_oom_score_adj_update(task);
1123 unlock_task_sighand(task, &flags);
1126 put_task_struct(task);
1128 return err < 0 ? err : count;
1131 static const struct file_operations proc_oom_adj_operations = {
1132 .read = oom_adj_read,
1133 .write = oom_adj_write,
1134 .llseek = generic_file_llseek,
1137 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1138 size_t count, loff_t *ppos)
1140 struct task_struct *task = get_proc_task(file_inode(file));
1141 char buffer[PROC_NUMBUF];
1142 short oom_score_adj = OOM_SCORE_ADJ_MIN;
1143 unsigned long flags;
1148 if (lock_task_sighand(task, &flags)) {
1149 oom_score_adj = task->signal->oom_score_adj;
1150 unlock_task_sighand(task, &flags);
1152 put_task_struct(task);
1153 len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
1154 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1157 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1158 size_t count, loff_t *ppos)
1160 struct task_struct *task;
1161 char buffer[PROC_NUMBUF];
1162 unsigned long flags;
1166 memset(buffer, 0, sizeof(buffer));
1167 if (count > sizeof(buffer) - 1)
1168 count = sizeof(buffer) - 1;
1169 if (copy_from_user(buffer, buf, count)) {
1174 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1177 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1178 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1183 task = get_proc_task(file_inode(file));
1195 if (!lock_task_sighand(task, &flags)) {
1200 if ((short)oom_score_adj < task->signal->oom_score_adj_min &&
1201 !capable(CAP_SYS_RESOURCE)) {
1206 task->signal->oom_score_adj = (short)oom_score_adj;
1207 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1208 task->signal->oom_score_adj_min = (short)oom_score_adj;
1209 trace_oom_score_adj_update(task);
1212 unlock_task_sighand(task, &flags);
1215 put_task_struct(task);
1217 return err < 0 ? err : count;
1220 static const struct file_operations proc_oom_score_adj_operations = {
1221 .read = oom_score_adj_read,
1222 .write = oom_score_adj_write,
1223 .llseek = default_llseek,
1226 #ifdef CONFIG_AUDITSYSCALL
1227 #define TMPBUFLEN 21
1228 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1229 size_t count, loff_t *ppos)
1231 struct inode * inode = file_inode(file);
1232 struct task_struct *task = get_proc_task(inode);
1234 char tmpbuf[TMPBUFLEN];
1238 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1239 from_kuid(file->f_cred->user_ns,
1240 audit_get_loginuid(task)));
1241 put_task_struct(task);
1242 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1245 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1246 size_t count, loff_t *ppos)
1248 struct inode * inode = file_inode(file);
1254 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1261 /* No partial writes. */
1265 rv = kstrtou32_from_user(buf, count, 10, &loginuid);
1269 /* is userspace tring to explicitly UNSET the loginuid? */
1270 if (loginuid == AUDIT_UID_UNSET) {
1271 kloginuid = INVALID_UID;
1273 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1274 if (!uid_valid(kloginuid))
1278 rv = audit_set_loginuid(kloginuid);
1284 static const struct file_operations proc_loginuid_operations = {
1285 .read = proc_loginuid_read,
1286 .write = proc_loginuid_write,
1287 .llseek = generic_file_llseek,
1290 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1291 size_t count, loff_t *ppos)
1293 struct inode * inode = file_inode(file);
1294 struct task_struct *task = get_proc_task(inode);
1296 char tmpbuf[TMPBUFLEN];
1300 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1301 audit_get_sessionid(task));
1302 put_task_struct(task);
1303 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1306 static const struct file_operations proc_sessionid_operations = {
1307 .read = proc_sessionid_read,
1308 .llseek = generic_file_llseek,
1312 #ifdef CONFIG_FAULT_INJECTION
1313 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1314 size_t count, loff_t *ppos)
1316 struct task_struct *task = get_proc_task(file_inode(file));
1317 char buffer[PROC_NUMBUF];
1323 make_it_fail = task->make_it_fail;
1324 put_task_struct(task);
1326 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1328 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1331 static ssize_t proc_fault_inject_write(struct file * file,
1332 const char __user * buf, size_t count, loff_t *ppos)
1334 struct task_struct *task;
1335 char buffer[PROC_NUMBUF];
1339 if (!capable(CAP_SYS_RESOURCE))
1341 memset(buffer, 0, sizeof(buffer));
1342 if (count > sizeof(buffer) - 1)
1343 count = sizeof(buffer) - 1;
1344 if (copy_from_user(buffer, buf, count))
1346 rv = kstrtoint(strstrip(buffer), 0, &make_it_fail);
1349 if (make_it_fail < 0 || make_it_fail > 1)
1352 task = get_proc_task(file_inode(file));
1355 task->make_it_fail = make_it_fail;
1356 put_task_struct(task);
1361 static const struct file_operations proc_fault_inject_operations = {
1362 .read = proc_fault_inject_read,
1363 .write = proc_fault_inject_write,
1364 .llseek = generic_file_llseek,
1369 #ifdef CONFIG_SCHED_DEBUG
1371 * Print out various scheduling related per-task fields:
1373 static int sched_show(struct seq_file *m, void *v)
1375 struct inode *inode = m->private;
1376 struct task_struct *p;
1378 p = get_proc_task(inode);
1381 proc_sched_show_task(p, m);
1389 sched_write(struct file *file, const char __user *buf,
1390 size_t count, loff_t *offset)
1392 struct inode *inode = file_inode(file);
1393 struct task_struct *p;
1395 p = get_proc_task(inode);
1398 proc_sched_set_task(p);
1405 static int sched_open(struct inode *inode, struct file *filp)
1407 return single_open(filp, sched_show, inode);
1410 static const struct file_operations proc_pid_sched_operations = {
1413 .write = sched_write,
1414 .llseek = seq_lseek,
1415 .release = single_release,
1420 #ifdef CONFIG_SCHED_AUTOGROUP
1422 * Print out autogroup related information:
1424 static int sched_autogroup_show(struct seq_file *m, void *v)
1426 struct inode *inode = m->private;
1427 struct task_struct *p;
1429 p = get_proc_task(inode);
1432 proc_sched_autogroup_show_task(p, m);
1440 sched_autogroup_write(struct file *file, const char __user *buf,
1441 size_t count, loff_t *offset)
1443 struct inode *inode = file_inode(file);
1444 struct task_struct *p;
1445 char buffer[PROC_NUMBUF];
1449 memset(buffer, 0, sizeof(buffer));
1450 if (count > sizeof(buffer) - 1)
1451 count = sizeof(buffer) - 1;
1452 if (copy_from_user(buffer, buf, count))
1455 err = kstrtoint(strstrip(buffer), 0, &nice);
1459 p = get_proc_task(inode);
1463 err = proc_sched_autogroup_set_nice(p, nice);
1472 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1476 ret = single_open(filp, sched_autogroup_show, NULL);
1478 struct seq_file *m = filp->private_data;
1485 static const struct file_operations proc_pid_sched_autogroup_operations = {
1486 .open = sched_autogroup_open,
1488 .write = sched_autogroup_write,
1489 .llseek = seq_lseek,
1490 .release = single_release,
1493 #endif /* CONFIG_SCHED_AUTOGROUP */
1495 static ssize_t comm_write(struct file *file, const char __user *buf,
1496 size_t count, loff_t *offset)
1498 struct inode *inode = file_inode(file);
1499 struct task_struct *p;
1500 char buffer[TASK_COMM_LEN];
1501 const size_t maxlen = sizeof(buffer) - 1;
1503 memset(buffer, 0, sizeof(buffer));
1504 if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
1507 p = get_proc_task(inode);
1511 if (same_thread_group(current, p))
1512 set_task_comm(p, buffer);
1521 static int comm_show(struct seq_file *m, void *v)
1523 struct inode *inode = m->private;
1524 struct task_struct *p;
1526 p = get_proc_task(inode);
1531 seq_printf(m, "%s\n", p->comm);
1539 static int comm_open(struct inode *inode, struct file *filp)
1541 return single_open(filp, comm_show, inode);
1544 static const struct file_operations proc_pid_set_comm_operations = {
1547 .write = comm_write,
1548 .llseek = seq_lseek,
1549 .release = single_release,
1552 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1554 struct task_struct *task;
1555 struct mm_struct *mm;
1556 struct file *exe_file;
1558 task = get_proc_task(d_inode(dentry));
1561 mm = get_task_mm(task);
1562 put_task_struct(task);
1565 exe_file = get_mm_exe_file(mm);
1568 *exe_path = exe_file->f_path;
1569 path_get(&exe_file->f_path);
1576 static const char *proc_pid_get_link(struct dentry *dentry,
1577 struct inode *inode,
1578 struct delayed_call *done)
1581 int error = -EACCES;
1584 return ERR_PTR(-ECHILD);
1586 /* Are we allowed to snoop on the tasks file descriptors? */
1587 if (!proc_fd_access_allowed(inode))
1590 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1594 nd_jump_link(&path);
1597 return ERR_PTR(error);
1600 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1602 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1609 pathname = d_path(path, tmp, PAGE_SIZE);
1610 len = PTR_ERR(pathname);
1611 if (IS_ERR(pathname))
1613 len = tmp + PAGE_SIZE - 1 - pathname;
1617 if (copy_to_user(buffer, pathname, len))
1620 free_page((unsigned long)tmp);
1624 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1626 int error = -EACCES;
1627 struct inode *inode = d_inode(dentry);
1630 /* Are we allowed to snoop on the tasks file descriptors? */
1631 if (!proc_fd_access_allowed(inode))
1634 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1638 error = do_proc_readlink(&path, buffer, buflen);
1644 const struct inode_operations proc_pid_link_inode_operations = {
1645 .readlink = proc_pid_readlink,
1646 .get_link = proc_pid_get_link,
1647 .setattr = proc_setattr,
1651 /* building an inode */
1653 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1655 struct inode * inode;
1656 struct proc_inode *ei;
1657 const struct cred *cred;
1659 /* We need a new inode */
1661 inode = new_inode(sb);
1667 inode->i_ino = get_next_ino();
1668 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1669 inode->i_op = &proc_def_inode_operations;
1672 * grab the reference to task.
1674 ei->pid = get_task_pid(task, PIDTYPE_PID);
1678 if (task_dumpable(task)) {
1680 cred = __task_cred(task);
1681 inode->i_uid = cred->euid;
1682 inode->i_gid = cred->egid;
1685 security_task_to_inode(task, inode);
1695 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1697 struct inode *inode = d_inode(dentry);
1698 struct task_struct *task;
1699 const struct cred *cred;
1700 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
1702 generic_fillattr(inode, stat);
1705 stat->uid = GLOBAL_ROOT_UID;
1706 stat->gid = GLOBAL_ROOT_GID;
1707 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1709 if (!has_pid_permissions(pid, task, 2)) {
1712 * This doesn't prevent learning whether PID exists,
1713 * it only makes getattr() consistent with readdir().
1717 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1718 task_dumpable(task)) {
1719 cred = __task_cred(task);
1720 stat->uid = cred->euid;
1721 stat->gid = cred->egid;
1731 * Exceptional case: normally we are not allowed to unhash a busy
1732 * directory. In this case, however, we can do it - no aliasing problems
1733 * due to the way we treat inodes.
1735 * Rewrite the inode's ownerships here because the owning task may have
1736 * performed a setuid(), etc.
1738 * Before the /proc/pid/status file was created the only way to read
1739 * the effective uid of a /process was to stat /proc/pid. Reading
1740 * /proc/pid/status is slow enough that procps and other packages
1741 * kept stating /proc/pid. To keep the rules in /proc simple I have
1742 * made this apply to all per process world readable and executable
1745 int pid_revalidate(struct dentry *dentry, unsigned int flags)
1747 struct inode *inode;
1748 struct task_struct *task;
1749 const struct cred *cred;
1751 if (flags & LOOKUP_RCU)
1754 inode = d_inode(dentry);
1755 task = get_proc_task(inode);
1758 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1759 task_dumpable(task)) {
1761 cred = __task_cred(task);
1762 inode->i_uid = cred->euid;
1763 inode->i_gid = cred->egid;
1766 inode->i_uid = GLOBAL_ROOT_UID;
1767 inode->i_gid = GLOBAL_ROOT_GID;
1769 inode->i_mode &= ~(S_ISUID | S_ISGID);
1770 security_task_to_inode(task, inode);
1771 put_task_struct(task);
1777 static inline bool proc_inode_is_dead(struct inode *inode)
1779 return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1782 int pid_delete_dentry(const struct dentry *dentry)
1784 /* Is the task we represent dead?
1785 * If so, then don't put the dentry on the lru list,
1786 * kill it immediately.
1788 return proc_inode_is_dead(d_inode(dentry));
1791 const struct dentry_operations pid_dentry_operations =
1793 .d_revalidate = pid_revalidate,
1794 .d_delete = pid_delete_dentry,
1800 * Fill a directory entry.
1802 * If possible create the dcache entry and derive our inode number and
1803 * file type from dcache entry.
1805 * Since all of the proc inode numbers are dynamically generated, the inode
1806 * numbers do not exist until the inode is cache. This means creating the
1807 * the dcache entry in readdir is necessary to keep the inode numbers
1808 * reported by readdir in sync with the inode numbers reported
1811 bool proc_fill_cache(struct file *file, struct dir_context *ctx,
1812 const char *name, int len,
1813 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1815 struct dentry *child, *dir = file->f_path.dentry;
1816 struct qstr qname = QSTR_INIT(name, len);
1817 struct inode *inode;
1821 child = d_hash_and_lookup(dir, &qname);
1823 child = d_alloc(dir, &qname);
1825 goto end_instantiate;
1826 if (instantiate(d_inode(dir), child, task, ptr) < 0) {
1828 goto end_instantiate;
1831 inode = d_inode(child);
1833 type = inode->i_mode >> 12;
1835 return dir_emit(ctx, name, len, ino, type);
1838 return dir_emit(ctx, name, len, 1, DT_UNKNOWN);
1842 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1843 * which represent vma start and end addresses.
1845 static int dname_to_vma_addr(struct dentry *dentry,
1846 unsigned long *start, unsigned long *end)
1848 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1854 static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
1856 unsigned long vm_start, vm_end;
1857 bool exact_vma_exists = false;
1858 struct mm_struct *mm = NULL;
1859 struct task_struct *task;
1860 const struct cred *cred;
1861 struct inode *inode;
1864 if (flags & LOOKUP_RCU)
1867 inode = d_inode(dentry);
1868 task = get_proc_task(inode);
1872 mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
1873 if (IS_ERR_OR_NULL(mm))
1876 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1877 down_read(&mm->mmap_sem);
1878 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1879 up_read(&mm->mmap_sem);
1884 if (exact_vma_exists) {
1885 if (task_dumpable(task)) {
1887 cred = __task_cred(task);
1888 inode->i_uid = cred->euid;
1889 inode->i_gid = cred->egid;
1892 inode->i_uid = GLOBAL_ROOT_UID;
1893 inode->i_gid = GLOBAL_ROOT_GID;
1895 security_task_to_inode(task, inode);
1900 put_task_struct(task);
1906 static const struct dentry_operations tid_map_files_dentry_operations = {
1907 .d_revalidate = map_files_d_revalidate,
1908 .d_delete = pid_delete_dentry,
1911 static int map_files_get_link(struct dentry *dentry, struct path *path)
1913 unsigned long vm_start, vm_end;
1914 struct vm_area_struct *vma;
1915 struct task_struct *task;
1916 struct mm_struct *mm;
1920 task = get_proc_task(d_inode(dentry));
1924 mm = get_task_mm(task);
1925 put_task_struct(task);
1929 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
1934 down_read(&mm->mmap_sem);
1935 vma = find_exact_vma(mm, vm_start, vm_end);
1936 if (vma && vma->vm_file) {
1937 *path = vma->vm_file->f_path;
1941 up_read(&mm->mmap_sem);
1949 struct map_files_info {
1952 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
1956 * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the
1957 * symlinks may be used to bypass permissions on ancestor directories in the
1958 * path to the file in question.
1961 proc_map_files_get_link(struct dentry *dentry,
1962 struct inode *inode,
1963 struct delayed_call *done)
1965 if (!capable(CAP_SYS_ADMIN))
1966 return ERR_PTR(-EPERM);
1968 return proc_pid_get_link(dentry, inode, done);
1972 * Identical to proc_pid_link_inode_operations except for get_link()
1974 static const struct inode_operations proc_map_files_link_inode_operations = {
1975 .readlink = proc_pid_readlink,
1976 .get_link = proc_map_files_get_link,
1977 .setattr = proc_setattr,
1981 proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
1982 struct task_struct *task, const void *ptr)
1984 fmode_t mode = (fmode_t)(unsigned long)ptr;
1985 struct proc_inode *ei;
1986 struct inode *inode;
1988 inode = proc_pid_make_inode(dir->i_sb, task);
1993 ei->op.proc_get_link = map_files_get_link;
1995 inode->i_op = &proc_map_files_link_inode_operations;
1997 inode->i_mode = S_IFLNK;
1999 if (mode & FMODE_READ)
2000 inode->i_mode |= S_IRUSR;
2001 if (mode & FMODE_WRITE)
2002 inode->i_mode |= S_IWUSR;
2004 d_set_d_op(dentry, &tid_map_files_dentry_operations);
2005 d_add(dentry, inode);
2010 static struct dentry *proc_map_files_lookup(struct inode *dir,
2011 struct dentry *dentry, unsigned int flags)
2013 unsigned long vm_start, vm_end;
2014 struct vm_area_struct *vma;
2015 struct task_struct *task;
2017 struct mm_struct *mm;
2020 task = get_proc_task(dir);
2025 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2029 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2032 mm = get_task_mm(task);
2036 down_read(&mm->mmap_sem);
2037 vma = find_exact_vma(mm, vm_start, vm_end);
2042 result = proc_map_files_instantiate(dir, dentry, task,
2043 (void *)(unsigned long)vma->vm_file->f_mode);
2046 up_read(&mm->mmap_sem);
2049 put_task_struct(task);
2051 return ERR_PTR(result);
2054 static const struct inode_operations proc_map_files_inode_operations = {
2055 .lookup = proc_map_files_lookup,
2056 .permission = proc_fd_permission,
2057 .setattr = proc_setattr,
2061 proc_map_files_readdir(struct file *file, struct dir_context *ctx)
2063 struct vm_area_struct *vma;
2064 struct task_struct *task;
2065 struct mm_struct *mm;
2066 unsigned long nr_files, pos, i;
2067 struct flex_array *fa = NULL;
2068 struct map_files_info info;
2069 struct map_files_info *p;
2073 task = get_proc_task(file_inode(file));
2078 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2082 if (!dir_emit_dots(file, ctx))
2085 mm = get_task_mm(task);
2088 down_read(&mm->mmap_sem);
2093 * We need two passes here:
2095 * 1) Collect vmas of mapped files with mmap_sem taken
2096 * 2) Release mmap_sem and instantiate entries
2098 * otherwise we get lockdep complained, since filldir()
2099 * routine might require mmap_sem taken in might_fault().
2102 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2103 if (vma->vm_file && ++pos > ctx->pos)
2108 fa = flex_array_alloc(sizeof(info), nr_files,
2110 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2114 flex_array_free(fa);
2115 up_read(&mm->mmap_sem);
2119 for (i = 0, vma = mm->mmap, pos = 2; vma;
2120 vma = vma->vm_next) {
2123 if (++pos <= ctx->pos)
2126 info.mode = vma->vm_file->f_mode;
2127 info.len = snprintf(info.name,
2128 sizeof(info.name), "%lx-%lx",
2129 vma->vm_start, vma->vm_end);
2130 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2134 up_read(&mm->mmap_sem);
2136 for (i = 0; i < nr_files; i++) {
2137 p = flex_array_get(fa, i);
2138 if (!proc_fill_cache(file, ctx,
2140 proc_map_files_instantiate,
2142 (void *)(unsigned long)p->mode))
2147 flex_array_free(fa);
2151 put_task_struct(task);
2156 static const struct file_operations proc_map_files_operations = {
2157 .read = generic_read_dir,
2158 .iterate = proc_map_files_readdir,
2159 .llseek = default_llseek,
2162 #ifdef CONFIG_CHECKPOINT_RESTORE
2163 struct timers_private {
2165 struct task_struct *task;
2166 struct sighand_struct *sighand;
2167 struct pid_namespace *ns;
2168 unsigned long flags;
2171 static void *timers_start(struct seq_file *m, loff_t *pos)
2173 struct timers_private *tp = m->private;
2175 tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2177 return ERR_PTR(-ESRCH);
2179 tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2181 return ERR_PTR(-ESRCH);
2183 return seq_list_start(&tp->task->signal->posix_timers, *pos);
2186 static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2188 struct timers_private *tp = m->private;
2189 return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2192 static void timers_stop(struct seq_file *m, void *v)
2194 struct timers_private *tp = m->private;
2197 unlock_task_sighand(tp->task, &tp->flags);
2202 put_task_struct(tp->task);
2207 static int show_timer(struct seq_file *m, void *v)
2209 struct k_itimer *timer;
2210 struct timers_private *tp = m->private;
2212 static const char * const nstr[] = {
2213 [SIGEV_SIGNAL] = "signal",
2214 [SIGEV_NONE] = "none",
2215 [SIGEV_THREAD] = "thread",
2218 timer = list_entry((struct list_head *)v, struct k_itimer, list);
2219 notify = timer->it_sigev_notify;
2221 seq_printf(m, "ID: %d\n", timer->it_id);
2222 seq_printf(m, "signal: %d/%p\n",
2223 timer->sigq->info.si_signo,
2224 timer->sigq->info.si_value.sival_ptr);
2225 seq_printf(m, "notify: %s/%s.%d\n",
2226 nstr[notify & ~SIGEV_THREAD_ID],
2227 (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2228 pid_nr_ns(timer->it_pid, tp->ns));
2229 seq_printf(m, "ClockID: %d\n", timer->it_clock);
2234 static const struct seq_operations proc_timers_seq_ops = {
2235 .start = timers_start,
2236 .next = timers_next,
2237 .stop = timers_stop,
2241 static int proc_timers_open(struct inode *inode, struct file *file)
2243 struct timers_private *tp;
2245 tp = __seq_open_private(file, &proc_timers_seq_ops,
2246 sizeof(struct timers_private));
2250 tp->pid = proc_pid(inode);
2251 tp->ns = inode->i_sb->s_fs_info;
2255 static const struct file_operations proc_timers_operations = {
2256 .open = proc_timers_open,
2258 .llseek = seq_lseek,
2259 .release = seq_release_private,
2263 static ssize_t timerslack_ns_write(struct file *file, const char __user *buf,
2264 size_t count, loff_t *offset)
2266 struct inode *inode = file_inode(file);
2267 struct task_struct *p;
2271 err = kstrtoull_from_user(buf, count, 10, &slack_ns);
2275 p = get_proc_task(inode);
2279 if (ptrace_may_access(p, PTRACE_MODE_ATTACH_FSCREDS)) {
2282 p->timer_slack_ns = p->default_timer_slack_ns;
2284 p->timer_slack_ns = slack_ns;
2294 static int timerslack_ns_show(struct seq_file *m, void *v)
2296 struct inode *inode = m->private;
2297 struct task_struct *p;
2300 p = get_proc_task(inode);
2304 if (ptrace_may_access(p, PTRACE_MODE_ATTACH_FSCREDS)) {
2306 seq_printf(m, "%llu\n", p->timer_slack_ns);
2316 static int timerslack_ns_open(struct inode *inode, struct file *filp)
2318 return single_open(filp, timerslack_ns_show, inode);
2321 static const struct file_operations proc_pid_set_timerslack_ns_operations = {
2322 .open = timerslack_ns_open,
2324 .write = timerslack_ns_write,
2325 .llseek = seq_lseek,
2326 .release = single_release,
2329 static int proc_pident_instantiate(struct inode *dir,
2330 struct dentry *dentry, struct task_struct *task, const void *ptr)
2332 const struct pid_entry *p = ptr;
2333 struct inode *inode;
2334 struct proc_inode *ei;
2336 inode = proc_pid_make_inode(dir->i_sb, task);
2341 inode->i_mode = p->mode;
2342 if (S_ISDIR(inode->i_mode))
2343 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2345 inode->i_op = p->iop;
2347 inode->i_fop = p->fop;
2349 d_set_d_op(dentry, &pid_dentry_operations);
2350 d_add(dentry, inode);
2351 /* Close the race of the process dying before we return the dentry */
2352 if (pid_revalidate(dentry, 0))
2358 static struct dentry *proc_pident_lookup(struct inode *dir,
2359 struct dentry *dentry,
2360 const struct pid_entry *ents,
2364 struct task_struct *task = get_proc_task(dir);
2365 const struct pid_entry *p, *last;
2373 * Yes, it does not scale. And it should not. Don't add
2374 * new entries into /proc/<tgid>/ without very good reasons.
2376 last = &ents[nents - 1];
2377 for (p = ents; p <= last; p++) {
2378 if (p->len != dentry->d_name.len)
2380 if (!memcmp(dentry->d_name.name, p->name, p->len))
2386 error = proc_pident_instantiate(dir, dentry, task, p);
2388 put_task_struct(task);
2390 return ERR_PTR(error);
2393 static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
2394 const struct pid_entry *ents, unsigned int nents)
2396 struct task_struct *task = get_proc_task(file_inode(file));
2397 const struct pid_entry *p;
2402 if (!dir_emit_dots(file, ctx))
2405 if (ctx->pos >= nents + 2)
2408 for (p = ents + (ctx->pos - 2); p <= ents + nents - 1; p++) {
2409 if (!proc_fill_cache(file, ctx, p->name, p->len,
2410 proc_pident_instantiate, task, p))
2415 put_task_struct(task);
2419 #ifdef CONFIG_SECURITY
2420 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2421 size_t count, loff_t *ppos)
2423 struct inode * inode = file_inode(file);
2426 struct task_struct *task = get_proc_task(inode);
2431 length = security_getprocattr(task,
2432 (char*)file->f_path.dentry->d_name.name,
2434 put_task_struct(task);
2436 length = simple_read_from_buffer(buf, count, ppos, p, length);
2441 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2442 size_t count, loff_t *ppos)
2444 struct inode * inode = file_inode(file);
2447 struct task_struct *task = get_proc_task(inode);
2452 if (count > PAGE_SIZE)
2455 /* No partial writes. */
2460 page = memdup_user(buf, count);
2462 length = PTR_ERR(page);
2466 /* Guard against adverse ptrace interaction */
2467 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2471 length = security_setprocattr(task,
2472 (char*)file->f_path.dentry->d_name.name,
2474 mutex_unlock(&task->signal->cred_guard_mutex);
2478 put_task_struct(task);
2483 static const struct file_operations proc_pid_attr_operations = {
2484 .read = proc_pid_attr_read,
2485 .write = proc_pid_attr_write,
2486 .llseek = generic_file_llseek,
2489 static const struct pid_entry attr_dir_stuff[] = {
2490 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2491 REG("prev", S_IRUGO, proc_pid_attr_operations),
2492 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2493 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2494 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2495 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2498 static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
2500 return proc_pident_readdir(file, ctx,
2501 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2504 static const struct file_operations proc_attr_dir_operations = {
2505 .read = generic_read_dir,
2506 .iterate = proc_attr_dir_readdir,
2507 .llseek = default_llseek,
2510 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2511 struct dentry *dentry, unsigned int flags)
2513 return proc_pident_lookup(dir, dentry,
2514 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2517 static const struct inode_operations proc_attr_dir_inode_operations = {
2518 .lookup = proc_attr_dir_lookup,
2519 .getattr = pid_getattr,
2520 .setattr = proc_setattr,
2525 #ifdef CONFIG_ELF_CORE
2526 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2527 size_t count, loff_t *ppos)
2529 struct task_struct *task = get_proc_task(file_inode(file));
2530 struct mm_struct *mm;
2531 char buffer[PROC_NUMBUF];
2539 mm = get_task_mm(task);
2541 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2542 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2543 MMF_DUMP_FILTER_SHIFT));
2545 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2548 put_task_struct(task);
2553 static ssize_t proc_coredump_filter_write(struct file *file,
2554 const char __user *buf,
2558 struct task_struct *task;
2559 struct mm_struct *mm;
2565 ret = kstrtouint_from_user(buf, count, 0, &val);
2570 task = get_proc_task(file_inode(file));
2574 mm = get_task_mm(task);
2579 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2581 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2583 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2588 put_task_struct(task);
2595 static const struct file_operations proc_coredump_filter_operations = {
2596 .read = proc_coredump_filter_read,
2597 .write = proc_coredump_filter_write,
2598 .llseek = generic_file_llseek,
2602 #ifdef CONFIG_TASK_IO_ACCOUNTING
2603 static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
2605 struct task_io_accounting acct = task->ioac;
2606 unsigned long flags;
2609 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2613 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) {
2618 if (whole && lock_task_sighand(task, &flags)) {
2619 struct task_struct *t = task;
2621 task_io_accounting_add(&acct, &task->signal->ioac);
2622 while_each_thread(task, t)
2623 task_io_accounting_add(&acct, &t->ioac);
2625 unlock_task_sighand(task, &flags);
2632 "read_bytes: %llu\n"
2633 "write_bytes: %llu\n"
2634 "cancelled_write_bytes: %llu\n",
2635 (unsigned long long)acct.rchar,
2636 (unsigned long long)acct.wchar,
2637 (unsigned long long)acct.syscr,
2638 (unsigned long long)acct.syscw,
2639 (unsigned long long)acct.read_bytes,
2640 (unsigned long long)acct.write_bytes,
2641 (unsigned long long)acct.cancelled_write_bytes);
2645 mutex_unlock(&task->signal->cred_guard_mutex);
2649 static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2650 struct pid *pid, struct task_struct *task)
2652 return do_io_accounting(task, m, 0);
2655 static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2656 struct pid *pid, struct task_struct *task)
2658 return do_io_accounting(task, m, 1);
2660 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2662 #ifdef CONFIG_USER_NS
2663 static int proc_id_map_open(struct inode *inode, struct file *file,
2664 const struct seq_operations *seq_ops)
2666 struct user_namespace *ns = NULL;
2667 struct task_struct *task;
2668 struct seq_file *seq;
2671 task = get_proc_task(inode);
2674 ns = get_user_ns(task_cred_xxx(task, user_ns));
2676 put_task_struct(task);
2681 ret = seq_open(file, seq_ops);
2685 seq = file->private_data;
2695 static int proc_id_map_release(struct inode *inode, struct file *file)
2697 struct seq_file *seq = file->private_data;
2698 struct user_namespace *ns = seq->private;
2700 return seq_release(inode, file);
2703 static int proc_uid_map_open(struct inode *inode, struct file *file)
2705 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2708 static int proc_gid_map_open(struct inode *inode, struct file *file)
2710 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2713 static int proc_projid_map_open(struct inode *inode, struct file *file)
2715 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2718 static const struct file_operations proc_uid_map_operations = {
2719 .open = proc_uid_map_open,
2720 .write = proc_uid_map_write,
2722 .llseek = seq_lseek,
2723 .release = proc_id_map_release,
2726 static const struct file_operations proc_gid_map_operations = {
2727 .open = proc_gid_map_open,
2728 .write = proc_gid_map_write,
2730 .llseek = seq_lseek,
2731 .release = proc_id_map_release,
2734 static const struct file_operations proc_projid_map_operations = {
2735 .open = proc_projid_map_open,
2736 .write = proc_projid_map_write,
2738 .llseek = seq_lseek,
2739 .release = proc_id_map_release,
2742 static int proc_setgroups_open(struct inode *inode, struct file *file)
2744 struct user_namespace *ns = NULL;
2745 struct task_struct *task;
2749 task = get_proc_task(inode);
2752 ns = get_user_ns(task_cred_xxx(task, user_ns));
2754 put_task_struct(task);
2759 if (file->f_mode & FMODE_WRITE) {
2761 if (!ns_capable(ns, CAP_SYS_ADMIN))
2765 ret = single_open(file, &proc_setgroups_show, ns);
2776 static int proc_setgroups_release(struct inode *inode, struct file *file)
2778 struct seq_file *seq = file->private_data;
2779 struct user_namespace *ns = seq->private;
2780 int ret = single_release(inode, file);
2785 static const struct file_operations proc_setgroups_operations = {
2786 .open = proc_setgroups_open,
2787 .write = proc_setgroups_write,
2789 .llseek = seq_lseek,
2790 .release = proc_setgroups_release,
2792 #endif /* CONFIG_USER_NS */
2794 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2795 struct pid *pid, struct task_struct *task)
2797 int err = lock_trace(task);
2799 seq_printf(m, "%08x\n", task->personality);
2808 static const struct file_operations proc_task_operations;
2809 static const struct inode_operations proc_task_inode_operations;
2811 static const struct pid_entry tgid_base_stuff[] = {
2812 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2813 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2814 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
2815 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2816 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2818 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2820 REG("environ", S_IRUSR, proc_environ_operations),
2821 ONE("auxv", S_IRUSR, proc_pid_auxv),
2822 ONE("status", S_IRUGO, proc_pid_status),
2823 ONE("personality", S_IRUSR, proc_pid_personality),
2824 ONE("limits", S_IRUGO, proc_pid_limits),
2825 #ifdef CONFIG_SCHED_DEBUG
2826 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2828 #ifdef CONFIG_SCHED_AUTOGROUP
2829 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2831 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2832 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2833 ONE("syscall", S_IRUSR, proc_pid_syscall),
2835 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
2836 ONE("stat", S_IRUGO, proc_tgid_stat),
2837 ONE("statm", S_IRUGO, proc_pid_statm),
2838 REG("maps", S_IRUGO, proc_pid_maps_operations),
2840 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
2842 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2843 LNK("cwd", proc_cwd_link),
2844 LNK("root", proc_root_link),
2845 LNK("exe", proc_exe_link),
2846 REG("mounts", S_IRUGO, proc_mounts_operations),
2847 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2848 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2849 #ifdef CONFIG_PROC_PAGE_MONITOR
2850 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2851 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
2852 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2854 #ifdef CONFIG_SECURITY
2855 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2857 #ifdef CONFIG_KALLSYMS
2858 ONE("wchan", S_IRUGO, proc_pid_wchan),
2860 #ifdef CONFIG_STACKTRACE
2861 ONE("stack", S_IRUSR, proc_pid_stack),
2863 #ifdef CONFIG_SCHED_INFO
2864 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
2866 #ifdef CONFIG_LATENCYTOP
2867 REG("latency", S_IRUGO, proc_lstats_operations),
2869 #ifdef CONFIG_PROC_PID_CPUSET
2870 ONE("cpuset", S_IRUGO, proc_cpuset_show),
2872 #ifdef CONFIG_CGROUPS
2873 ONE("cgroup", S_IRUGO, proc_cgroup_show),
2875 ONE("oom_score", S_IRUGO, proc_oom_score),
2876 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
2877 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2878 #ifdef CONFIG_AUDITSYSCALL
2879 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2880 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2882 #ifdef CONFIG_FAULT_INJECTION
2883 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2885 #ifdef CONFIG_ELF_CORE
2886 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2888 #ifdef CONFIG_TASK_IO_ACCOUNTING
2889 ONE("io", S_IRUSR, proc_tgid_io_accounting),
2891 #ifdef CONFIG_HARDWALL
2892 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
2894 #ifdef CONFIG_USER_NS
2895 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
2896 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
2897 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
2898 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
2900 #ifdef CONFIG_CHECKPOINT_RESTORE
2901 REG("timers", S_IRUGO, proc_timers_operations),
2903 REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations),
2906 static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
2908 return proc_pident_readdir(file, ctx,
2909 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2912 static const struct file_operations proc_tgid_base_operations = {
2913 .read = generic_read_dir,
2914 .iterate = proc_tgid_base_readdir,
2915 .llseek = default_llseek,
2918 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2920 return proc_pident_lookup(dir, dentry,
2921 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2924 static const struct inode_operations proc_tgid_base_inode_operations = {
2925 .lookup = proc_tgid_base_lookup,
2926 .getattr = pid_getattr,
2927 .setattr = proc_setattr,
2928 .permission = proc_pid_permission,
2931 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2933 struct dentry *dentry, *leader, *dir;
2934 char buf[PROC_NUMBUF];
2938 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2939 /* no ->d_hash() rejects on procfs */
2940 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2942 d_invalidate(dentry);
2950 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2951 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2956 name.len = strlen(name.name);
2957 dir = d_hash_and_lookup(leader, &name);
2959 goto out_put_leader;
2962 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2963 dentry = d_hash_and_lookup(dir, &name);
2965 d_invalidate(dentry);
2977 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2978 * @task: task that should be flushed.
2980 * When flushing dentries from proc, one needs to flush them from global
2981 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2982 * in. This call is supposed to do all of this job.
2984 * Looks in the dcache for
2986 * /proc/@tgid/task/@pid
2987 * if either directory is present flushes it and all of it'ts children
2990 * It is safe and reasonable to cache /proc entries for a task until
2991 * that task exits. After that they just clog up the dcache with
2992 * useless entries, possibly causing useful dcache entries to be
2993 * flushed instead. This routine is proved to flush those useless
2994 * dcache entries at process exit time.
2996 * NOTE: This routine is just an optimization so it does not guarantee
2997 * that no dcache entries will exist at process exit time it
2998 * just makes it very unlikely that any will persist.
3001 void proc_flush_task(struct task_struct *task)
3004 struct pid *pid, *tgid;
3007 pid = task_pid(task);
3008 tgid = task_tgid(task);
3010 for (i = 0; i <= pid->level; i++) {
3011 upid = &pid->numbers[i];
3012 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
3013 tgid->numbers[i].nr);
3017 static int proc_pid_instantiate(struct inode *dir,
3018 struct dentry * dentry,
3019 struct task_struct *task, const void *ptr)
3021 struct inode *inode;
3023 inode = proc_pid_make_inode(dir->i_sb, task);
3027 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3028 inode->i_op = &proc_tgid_base_inode_operations;
3029 inode->i_fop = &proc_tgid_base_operations;
3030 inode->i_flags|=S_IMMUTABLE;
3032 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
3033 ARRAY_SIZE(tgid_base_stuff)));
3035 d_set_d_op(dentry, &pid_dentry_operations);
3037 d_add(dentry, inode);
3038 /* Close the race of the process dying before we return the dentry */
3039 if (pid_revalidate(dentry, 0))
3045 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3047 int result = -ENOENT;
3048 struct task_struct *task;
3050 struct pid_namespace *ns;
3052 tgid = name_to_int(&dentry->d_name);
3056 ns = dentry->d_sb->s_fs_info;
3058 task = find_task_by_pid_ns(tgid, ns);
3060 get_task_struct(task);
3065 result = proc_pid_instantiate(dir, dentry, task, NULL);
3066 put_task_struct(task);
3068 return ERR_PTR(result);
3072 * Find the first task with tgid >= tgid
3077 struct task_struct *task;
3079 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3084 put_task_struct(iter.task);
3088 pid = find_ge_pid(iter.tgid, ns);
3090 iter.tgid = pid_nr_ns(pid, ns);
3091 iter.task = pid_task(pid, PIDTYPE_PID);
3092 /* What we to know is if the pid we have find is the
3093 * pid of a thread_group_leader. Testing for task
3094 * being a thread_group_leader is the obvious thing
3095 * todo but there is a window when it fails, due to
3096 * the pid transfer logic in de_thread.
3098 * So we perform the straight forward test of seeing
3099 * if the pid we have found is the pid of a thread
3100 * group leader, and don't worry if the task we have
3101 * found doesn't happen to be a thread group leader.
3102 * As we don't care in the case of readdir.
3104 if (!iter.task || !has_group_leader_pid(iter.task)) {
3108 get_task_struct(iter.task);
3114 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3116 /* for the /proc/ directory itself, after non-process stuff has been done */
3117 int proc_pid_readdir(struct file *file, struct dir_context *ctx)
3119 struct tgid_iter iter;
3120 struct pid_namespace *ns = file_inode(file)->i_sb->s_fs_info;
3121 loff_t pos = ctx->pos;
3123 if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
3126 if (pos == TGID_OFFSET - 2) {
3127 struct inode *inode = d_inode(ns->proc_self);
3128 if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
3130 ctx->pos = pos = pos + 1;
3132 if (pos == TGID_OFFSET - 1) {
3133 struct inode *inode = d_inode(ns->proc_thread_self);
3134 if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3136 ctx->pos = pos = pos + 1;
3138 iter.tgid = pos - TGID_OFFSET;
3140 for (iter = next_tgid(ns, iter);
3142 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3143 char name[PROC_NUMBUF];
3145 if (!has_pid_permissions(ns, iter.task, 2))
3148 len = snprintf(name, sizeof(name), "%d", iter.tgid);
3149 ctx->pos = iter.tgid + TGID_OFFSET;
3150 if (!proc_fill_cache(file, ctx, name, len,
3151 proc_pid_instantiate, iter.task, NULL)) {
3152 put_task_struct(iter.task);
3156 ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
3163 static const struct pid_entry tid_base_stuff[] = {
3164 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3165 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3166 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3168 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3170 REG("environ", S_IRUSR, proc_environ_operations),
3171 ONE("auxv", S_IRUSR, proc_pid_auxv),
3172 ONE("status", S_IRUGO, proc_pid_status),
3173 ONE("personality", S_IRUSR, proc_pid_personality),
3174 ONE("limits", S_IRUGO, proc_pid_limits),
3175 #ifdef CONFIG_SCHED_DEBUG
3176 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3178 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3179 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3180 ONE("syscall", S_IRUSR, proc_pid_syscall),
3182 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
3183 ONE("stat", S_IRUGO, proc_tid_stat),
3184 ONE("statm", S_IRUGO, proc_pid_statm),
3185 REG("maps", S_IRUGO, proc_tid_maps_operations),
3186 #ifdef CONFIG_PROC_CHILDREN
3187 REG("children", S_IRUGO, proc_tid_children_operations),
3190 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
3192 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3193 LNK("cwd", proc_cwd_link),
3194 LNK("root", proc_root_link),
3195 LNK("exe", proc_exe_link),
3196 REG("mounts", S_IRUGO, proc_mounts_operations),
3197 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3198 #ifdef CONFIG_PROC_PAGE_MONITOR
3199 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3200 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
3201 REG("pagemap", S_IRUSR, proc_pagemap_operations),
3203 #ifdef CONFIG_SECURITY
3204 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3206 #ifdef CONFIG_KALLSYMS
3207 ONE("wchan", S_IRUGO, proc_pid_wchan),
3209 #ifdef CONFIG_STACKTRACE
3210 ONE("stack", S_IRUSR, proc_pid_stack),
3212 #ifdef CONFIG_SCHED_INFO
3213 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
3215 #ifdef CONFIG_LATENCYTOP
3216 REG("latency", S_IRUGO, proc_lstats_operations),
3218 #ifdef CONFIG_PROC_PID_CPUSET
3219 ONE("cpuset", S_IRUGO, proc_cpuset_show),
3221 #ifdef CONFIG_CGROUPS
3222 ONE("cgroup", S_IRUGO, proc_cgroup_show),
3224 ONE("oom_score", S_IRUGO, proc_oom_score),
3225 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3226 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3227 #ifdef CONFIG_AUDITSYSCALL
3228 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3229 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3231 #ifdef CONFIG_FAULT_INJECTION
3232 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3234 #ifdef CONFIG_TASK_IO_ACCOUNTING
3235 ONE("io", S_IRUSR, proc_tid_io_accounting),
3237 #ifdef CONFIG_HARDWALL
3238 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
3240 #ifdef CONFIG_USER_NS
3241 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3242 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3243 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3244 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
3248 static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
3250 return proc_pident_readdir(file, ctx,
3251 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3254 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3256 return proc_pident_lookup(dir, dentry,
3257 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3260 static const struct file_operations proc_tid_base_operations = {
3261 .read = generic_read_dir,
3262 .iterate = proc_tid_base_readdir,
3263 .llseek = default_llseek,
3266 static const struct inode_operations proc_tid_base_inode_operations = {
3267 .lookup = proc_tid_base_lookup,
3268 .getattr = pid_getattr,
3269 .setattr = proc_setattr,
3272 static int proc_task_instantiate(struct inode *dir,
3273 struct dentry *dentry, struct task_struct *task, const void *ptr)
3275 struct inode *inode;
3276 inode = proc_pid_make_inode(dir->i_sb, task);
3280 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3281 inode->i_op = &proc_tid_base_inode_operations;
3282 inode->i_fop = &proc_tid_base_operations;
3283 inode->i_flags|=S_IMMUTABLE;
3285 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3286 ARRAY_SIZE(tid_base_stuff)));
3288 d_set_d_op(dentry, &pid_dentry_operations);
3290 d_add(dentry, inode);
3291 /* Close the race of the process dying before we return the dentry */
3292 if (pid_revalidate(dentry, 0))
3298 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3300 int result = -ENOENT;
3301 struct task_struct *task;
3302 struct task_struct *leader = get_proc_task(dir);
3304 struct pid_namespace *ns;
3309 tid = name_to_int(&dentry->d_name);
3313 ns = dentry->d_sb->s_fs_info;
3315 task = find_task_by_pid_ns(tid, ns);
3317 get_task_struct(task);
3321 if (!same_thread_group(leader, task))
3324 result = proc_task_instantiate(dir, dentry, task, NULL);
3326 put_task_struct(task);
3328 put_task_struct(leader);
3330 return ERR_PTR(result);
3334 * Find the first tid of a thread group to return to user space.
3336 * Usually this is just the thread group leader, but if the users
3337 * buffer was too small or there was a seek into the middle of the
3338 * directory we have more work todo.
3340 * In the case of a short read we start with find_task_by_pid.
3342 * In the case of a seek we start with the leader and walk nr
3345 static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3346 struct pid_namespace *ns)
3348 struct task_struct *pos, *task;
3349 unsigned long nr = f_pos;
3351 if (nr != f_pos) /* 32bit overflow? */
3355 task = pid_task(pid, PIDTYPE_PID);
3359 /* Attempt to start with the tid of a thread */
3361 pos = find_task_by_pid_ns(tid, ns);
3362 if (pos && same_thread_group(pos, task))
3366 /* If nr exceeds the number of threads there is nothing todo */
3367 if (nr >= get_nr_threads(task))
3370 /* If we haven't found our starting place yet start
3371 * with the leader and walk nr threads forward.
3373 pos = task = task->group_leader;
3377 } while_each_thread(task, pos);
3382 get_task_struct(pos);
3389 * Find the next thread in the thread list.
3390 * Return NULL if there is an error or no next thread.
3392 * The reference to the input task_struct is released.
3394 static struct task_struct *next_tid(struct task_struct *start)
3396 struct task_struct *pos = NULL;
3398 if (pid_alive(start)) {
3399 pos = next_thread(start);
3400 if (thread_group_leader(pos))
3403 get_task_struct(pos);
3406 put_task_struct(start);
3410 /* for the /proc/TGID/task/ directories */
3411 static int proc_task_readdir(struct file *file, struct dir_context *ctx)
3413 struct inode *inode = file_inode(file);
3414 struct task_struct *task;
3415 struct pid_namespace *ns;
3418 if (proc_inode_is_dead(inode))
3421 if (!dir_emit_dots(file, ctx))
3424 /* f_version caches the tgid value that the last readdir call couldn't
3425 * return. lseek aka telldir automagically resets f_version to 0.
3427 ns = inode->i_sb->s_fs_info;
3428 tid = (int)file->f_version;
3429 file->f_version = 0;
3430 for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
3432 task = next_tid(task), ctx->pos++) {
3433 char name[PROC_NUMBUF];
3435 tid = task_pid_nr_ns(task, ns);
3436 len = snprintf(name, sizeof(name), "%d", tid);
3437 if (!proc_fill_cache(file, ctx, name, len,
3438 proc_task_instantiate, task, NULL)) {
3439 /* returning this tgid failed, save it as the first
3440 * pid for the next readir call */
3441 file->f_version = (u64)tid;
3442 put_task_struct(task);
3450 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3452 struct inode *inode = d_inode(dentry);
3453 struct task_struct *p = get_proc_task(inode);
3454 generic_fillattr(inode, stat);
3457 stat->nlink += get_nr_threads(p);
3464 static const struct inode_operations proc_task_inode_operations = {
3465 .lookup = proc_task_lookup,
3466 .getattr = proc_task_getattr,
3467 .setattr = proc_setattr,
3468 .permission = proc_pid_permission,
3471 static const struct file_operations proc_task_operations = {
3472 .read = generic_read_dir,
3473 .iterate = proc_task_readdir,
3474 .llseek = default_llseek,