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 <linux/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/sched/autogroup.h>
89 #include <linux/sched/mm.h>
90 #include <linux/sched/coredump.h>
91 #include <linux/sched/debug.h>
92 #include <linux/sched/stat.h>
93 #include <linux/flex_array.h>
94 #include <linux/posix-timers.h>
95 #ifdef CONFIG_HARDWALL
96 #include <asm/hardwall.h>
98 #include <trace/events/oom.h>
103 * Implementing inode permission operations in /proc is almost
104 * certainly an error. Permission checks need to happen during
105 * each system call not at open time. The reason is that most of
106 * what we wish to check for permissions in /proc varies at runtime.
108 * The classic example of a problem is opening file descriptors
109 * in /proc for a task before it execs a suid executable.
113 static u8 nlink_tgid;
119 const struct inode_operations *iop;
120 const struct file_operations *fop;
124 #define NOD(NAME, MODE, IOP, FOP, OP) { \
126 .len = sizeof(NAME) - 1, \
133 #define DIR(NAME, MODE, iops, fops) \
134 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
135 #define LNK(NAME, get_link) \
136 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
137 &proc_pid_link_inode_operations, NULL, \
138 { .proc_get_link = get_link } )
139 #define REG(NAME, MODE, fops) \
140 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
141 #define ONE(NAME, MODE, show) \
142 NOD(NAME, (S_IFREG|(MODE)), \
143 NULL, &proc_single_file_operations, \
144 { .proc_show = show } )
147 * Count the number of hardlinks for the pid_entry table, excluding the .
150 static unsigned int __init pid_entry_nlink(const struct pid_entry *entries,
157 for (i = 0; i < n; ++i) {
158 if (S_ISDIR(entries[i].mode))
165 static int get_task_root(struct task_struct *task, struct path *root)
167 int result = -ENOENT;
171 get_fs_root(task->fs, root);
178 static int proc_cwd_link(struct dentry *dentry, struct path *path)
180 struct task_struct *task = get_proc_task(d_inode(dentry));
181 int result = -ENOENT;
186 get_fs_pwd(task->fs, path);
190 put_task_struct(task);
195 static int proc_root_link(struct dentry *dentry, struct path *path)
197 struct task_struct *task = get_proc_task(d_inode(dentry));
198 int result = -ENOENT;
201 result = get_task_root(task, path);
202 put_task_struct(task);
207 static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
208 size_t _count, loff_t *pos)
210 struct task_struct *tsk;
211 struct mm_struct *mm;
213 unsigned long count = _count;
214 unsigned long arg_start, arg_end, env_start, env_end;
215 unsigned long len1, len2, len;
222 tsk = get_proc_task(file_inode(file));
225 mm = get_task_mm(tsk);
226 put_task_struct(tsk);
229 /* Check if process spawned far enough to have cmdline. */
235 page = (char *)__get_free_page(GFP_TEMPORARY);
241 down_read(&mm->mmap_sem);
242 arg_start = mm->arg_start;
243 arg_end = mm->arg_end;
244 env_start = mm->env_start;
245 env_end = mm->env_end;
246 up_read(&mm->mmap_sem);
248 BUG_ON(arg_start > arg_end);
249 BUG_ON(env_start > env_end);
251 len1 = arg_end - arg_start;
252 len2 = env_end - env_start;
260 * Inherently racy -- command line shares address space
261 * with code and data.
263 rv = access_remote_vm(mm, arg_end - 1, &c, 1, 0);
270 /* Command line (set of strings) occupies whole ARGV. */
274 p = arg_start + *pos;
276 while (count > 0 && len > 0) {
280 _count = min3(count, len, PAGE_SIZE);
281 nr_read = access_remote_vm(mm, p, page, _count, 0);
287 if (copy_to_user(buf, page, nr_read)) {
300 * Command line (1 string) occupies ARGV and
307 { .p = arg_start, .len = len1 },
308 { .p = env_start, .len = len2 },
314 while (i < 2 && pos1 >= cmdline[i].len) {
315 pos1 -= cmdline[i].len;
319 p = cmdline[i].p + pos1;
320 len = cmdline[i].len - pos1;
321 while (count > 0 && len > 0) {
322 unsigned int _count, l;
326 _count = min3(count, len, PAGE_SIZE);
327 nr_read = access_remote_vm(mm, p, page, _count, 0);
334 * Command line can be shorter than whole ARGV
335 * even if last "marker" byte says it is not.
338 l = strnlen(page, nr_read);
344 if (copy_to_user(buf, page, nr_read)) {
359 /* Only first chunk can be read partially. */
366 free_page((unsigned long)page);
374 static const struct file_operations proc_pid_cmdline_ops = {
375 .read = proc_pid_cmdline_read,
376 .llseek = generic_file_llseek,
379 #ifdef CONFIG_KALLSYMS
381 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
382 * Returns the resolved symbol. If that fails, simply return the address.
384 static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
385 struct pid *pid, struct task_struct *task)
388 char symname[KSYM_NAME_LEN];
390 wchan = get_wchan(task);
392 if (wchan && ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)
393 && !lookup_symbol_name(wchan, symname))
394 seq_printf(m, "%s", symname);
400 #endif /* CONFIG_KALLSYMS */
402 static int lock_trace(struct task_struct *task)
404 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
407 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) {
408 mutex_unlock(&task->signal->cred_guard_mutex);
414 static void unlock_trace(struct task_struct *task)
416 mutex_unlock(&task->signal->cred_guard_mutex);
419 #ifdef CONFIG_STACKTRACE
421 #define MAX_STACK_TRACE_DEPTH 64
423 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
424 struct pid *pid, struct task_struct *task)
426 struct stack_trace trace;
427 unsigned long *entries;
431 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
435 trace.nr_entries = 0;
436 trace.max_entries = MAX_STACK_TRACE_DEPTH;
437 trace.entries = entries;
440 err = lock_trace(task);
442 save_stack_trace_tsk(task, &trace);
444 for (i = 0; i < trace.nr_entries; i++) {
445 seq_printf(m, "[<%pK>] %pB\n",
446 (void *)entries[i], (void *)entries[i]);
456 #ifdef CONFIG_SCHED_INFO
458 * Provides /proc/PID/schedstat
460 static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
461 struct pid *pid, struct task_struct *task)
463 if (unlikely(!sched_info_on()))
464 seq_printf(m, "0 0 0\n");
466 seq_printf(m, "%llu %llu %lu\n",
467 (unsigned long long)task->se.sum_exec_runtime,
468 (unsigned long long)task->sched_info.run_delay,
469 task->sched_info.pcount);
475 #ifdef CONFIG_LATENCYTOP
476 static int lstats_show_proc(struct seq_file *m, void *v)
479 struct inode *inode = m->private;
480 struct task_struct *task = get_proc_task(inode);
484 seq_puts(m, "Latency Top version : v0.1\n");
485 for (i = 0; i < 32; i++) {
486 struct latency_record *lr = &task->latency_record[i];
487 if (lr->backtrace[0]) {
489 seq_printf(m, "%i %li %li",
490 lr->count, lr->time, lr->max);
491 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
492 unsigned long bt = lr->backtrace[q];
497 seq_printf(m, " %ps", (void *)bt);
503 put_task_struct(task);
507 static int lstats_open(struct inode *inode, struct file *file)
509 return single_open(file, lstats_show_proc, inode);
512 static ssize_t lstats_write(struct file *file, const char __user *buf,
513 size_t count, loff_t *offs)
515 struct task_struct *task = get_proc_task(file_inode(file));
519 clear_all_latency_tracing(task);
520 put_task_struct(task);
525 static const struct file_operations proc_lstats_operations = {
528 .write = lstats_write,
530 .release = single_release,
535 static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
536 struct pid *pid, struct task_struct *task)
538 unsigned long totalpages = totalram_pages + total_swap_pages;
539 unsigned long points = 0;
541 points = oom_badness(task, NULL, NULL, totalpages) *
543 seq_printf(m, "%lu\n", points);
553 static const struct limit_names lnames[RLIM_NLIMITS] = {
554 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
555 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
556 [RLIMIT_DATA] = {"Max data size", "bytes"},
557 [RLIMIT_STACK] = {"Max stack size", "bytes"},
558 [RLIMIT_CORE] = {"Max core file size", "bytes"},
559 [RLIMIT_RSS] = {"Max resident set", "bytes"},
560 [RLIMIT_NPROC] = {"Max processes", "processes"},
561 [RLIMIT_NOFILE] = {"Max open files", "files"},
562 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
563 [RLIMIT_AS] = {"Max address space", "bytes"},
564 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
565 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
566 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
567 [RLIMIT_NICE] = {"Max nice priority", NULL},
568 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
569 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
572 /* Display limits for a process */
573 static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
574 struct pid *pid, struct task_struct *task)
579 struct rlimit rlim[RLIM_NLIMITS];
581 if (!lock_task_sighand(task, &flags))
583 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
584 unlock_task_sighand(task, &flags);
587 * print the file header
589 seq_printf(m, "%-25s %-20s %-20s %-10s\n",
590 "Limit", "Soft Limit", "Hard Limit", "Units");
592 for (i = 0; i < RLIM_NLIMITS; i++) {
593 if (rlim[i].rlim_cur == RLIM_INFINITY)
594 seq_printf(m, "%-25s %-20s ",
595 lnames[i].name, "unlimited");
597 seq_printf(m, "%-25s %-20lu ",
598 lnames[i].name, rlim[i].rlim_cur);
600 if (rlim[i].rlim_max == RLIM_INFINITY)
601 seq_printf(m, "%-20s ", "unlimited");
603 seq_printf(m, "%-20lu ", rlim[i].rlim_max);
606 seq_printf(m, "%-10s\n", lnames[i].unit);
614 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
615 static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
616 struct pid *pid, struct task_struct *task)
619 unsigned long args[6], sp, pc;
622 res = lock_trace(task);
626 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
627 seq_puts(m, "running\n");
629 seq_printf(m, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
632 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
634 args[0], args[1], args[2], args[3], args[4], args[5],
640 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
642 /************************************************************************/
643 /* Here the fs part begins */
644 /************************************************************************/
646 /* permission checks */
647 static int proc_fd_access_allowed(struct inode *inode)
649 struct task_struct *task;
651 /* Allow access to a task's file descriptors if it is us or we
652 * may use ptrace attach to the process and find out that
655 task = get_proc_task(inode);
657 allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
658 put_task_struct(task);
663 int proc_setattr(struct dentry *dentry, struct iattr *attr)
666 struct inode *inode = d_inode(dentry);
668 if (attr->ia_valid & ATTR_MODE)
671 error = setattr_prepare(dentry, attr);
675 setattr_copy(inode, attr);
676 mark_inode_dirty(inode);
681 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
682 * or euid/egid (for hide_pid_min=2)?
684 static bool has_pid_permissions(struct pid_namespace *pid,
685 struct task_struct *task,
688 if (pid->hide_pid < hide_pid_min)
690 if (in_group_p(pid->pid_gid))
692 return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
696 static int proc_pid_permission(struct inode *inode, int mask)
698 struct pid_namespace *pid = inode->i_sb->s_fs_info;
699 struct task_struct *task;
702 task = get_proc_task(inode);
705 has_perms = has_pid_permissions(pid, task, HIDEPID_NO_ACCESS);
706 put_task_struct(task);
709 if (pid->hide_pid == HIDEPID_INVISIBLE) {
711 * Let's make getdents(), stat(), and open()
712 * consistent with each other. If a process
713 * may not stat() a file, it shouldn't be seen
721 return generic_permission(inode, mask);
726 static const struct inode_operations proc_def_inode_operations = {
727 .setattr = proc_setattr,
730 static int proc_single_show(struct seq_file *m, void *v)
732 struct inode *inode = m->private;
733 struct pid_namespace *ns;
735 struct task_struct *task;
738 ns = inode->i_sb->s_fs_info;
739 pid = proc_pid(inode);
740 task = get_pid_task(pid, PIDTYPE_PID);
744 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
746 put_task_struct(task);
750 static int proc_single_open(struct inode *inode, struct file *filp)
752 return single_open(filp, proc_single_show, inode);
755 static const struct file_operations proc_single_file_operations = {
756 .open = proc_single_open,
759 .release = single_release,
763 struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
765 struct task_struct *task = get_proc_task(inode);
766 struct mm_struct *mm = ERR_PTR(-ESRCH);
769 mm = mm_access(task, mode | PTRACE_MODE_FSCREDS);
770 put_task_struct(task);
772 if (!IS_ERR_OR_NULL(mm)) {
773 /* ensure this mm_struct can't be freed */
775 /* but do not pin its memory */
783 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
785 struct mm_struct *mm = proc_mem_open(inode, mode);
790 file->private_data = mm;
794 static int mem_open(struct inode *inode, struct file *file)
796 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
798 /* OK to pass negative loff_t, we can catch out-of-range */
799 file->f_mode |= FMODE_UNSIGNED_OFFSET;
804 static ssize_t mem_rw(struct file *file, char __user *buf,
805 size_t count, loff_t *ppos, int write)
807 struct mm_struct *mm = file->private_data;
808 unsigned long addr = *ppos;
816 page = (char *)__get_free_page(GFP_TEMPORARY);
821 if (!mmget_not_zero(mm))
824 flags = FOLL_FORCE | (write ? FOLL_WRITE : 0);
827 int this_len = min_t(int, count, PAGE_SIZE);
829 if (write && copy_from_user(page, buf, this_len)) {
834 this_len = access_remote_vm(mm, addr, page, this_len, flags);
841 if (!write && copy_to_user(buf, page, this_len)) {
855 free_page((unsigned long) page);
859 static ssize_t mem_read(struct file *file, char __user *buf,
860 size_t count, loff_t *ppos)
862 return mem_rw(file, buf, count, ppos, 0);
865 static ssize_t mem_write(struct file *file, const char __user *buf,
866 size_t count, loff_t *ppos)
868 return mem_rw(file, (char __user*)buf, count, ppos, 1);
871 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
875 file->f_pos = offset;
878 file->f_pos += offset;
883 force_successful_syscall_return();
887 static int mem_release(struct inode *inode, struct file *file)
889 struct mm_struct *mm = file->private_data;
895 static const struct file_operations proc_mem_operations = {
900 .release = mem_release,
903 static int environ_open(struct inode *inode, struct file *file)
905 return __mem_open(inode, file, PTRACE_MODE_READ);
908 static ssize_t environ_read(struct file *file, char __user *buf,
909 size_t count, loff_t *ppos)
912 unsigned long src = *ppos;
914 struct mm_struct *mm = file->private_data;
915 unsigned long env_start, env_end;
917 /* Ensure the process spawned far enough to have an environment. */
918 if (!mm || !mm->env_end)
921 page = (char *)__get_free_page(GFP_TEMPORARY);
926 if (!mmget_not_zero(mm))
929 down_read(&mm->mmap_sem);
930 env_start = mm->env_start;
931 env_end = mm->env_end;
932 up_read(&mm->mmap_sem);
935 size_t this_len, max_len;
938 if (src >= (env_end - env_start))
941 this_len = env_end - (env_start + src);
943 max_len = min_t(size_t, PAGE_SIZE, count);
944 this_len = min(max_len, this_len);
946 retval = access_remote_vm(mm, (env_start + src), page, this_len, 0);
953 if (copy_to_user(buf, page, retval)) {
967 free_page((unsigned long) page);
971 static const struct file_operations proc_environ_operations = {
972 .open = environ_open,
973 .read = environ_read,
974 .llseek = generic_file_llseek,
975 .release = mem_release,
978 static int auxv_open(struct inode *inode, struct file *file)
980 return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
983 static ssize_t auxv_read(struct file *file, char __user *buf,
984 size_t count, loff_t *ppos)
986 struct mm_struct *mm = file->private_data;
987 unsigned int nwords = 0;
993 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
994 return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv,
995 nwords * sizeof(mm->saved_auxv[0]));
998 static const struct file_operations proc_auxv_operations = {
1001 .llseek = generic_file_llseek,
1002 .release = mem_release,
1005 static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1008 struct task_struct *task = get_proc_task(file_inode(file));
1009 char buffer[PROC_NUMBUF];
1010 int oom_adj = OOM_ADJUST_MIN;
1015 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1016 oom_adj = OOM_ADJUST_MAX;
1018 oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1020 put_task_struct(task);
1021 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1022 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1025 static int __set_oom_adj(struct file *file, int oom_adj, bool legacy)
1027 static DEFINE_MUTEX(oom_adj_mutex);
1028 struct mm_struct *mm = NULL;
1029 struct task_struct *task;
1032 task = get_proc_task(file_inode(file));
1036 mutex_lock(&oom_adj_mutex);
1038 if (oom_adj < task->signal->oom_score_adj &&
1039 !capable(CAP_SYS_RESOURCE)) {
1044 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1045 * /proc/pid/oom_score_adj instead.
1047 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1048 current->comm, task_pid_nr(current), task_pid_nr(task),
1051 if ((short)oom_adj < task->signal->oom_score_adj_min &&
1052 !capable(CAP_SYS_RESOURCE)) {
1059 * Make sure we will check other processes sharing the mm if this is
1060 * not vfrok which wants its own oom_score_adj.
1061 * pin the mm so it doesn't go away and get reused after task_unlock
1063 if (!task->vfork_done) {
1064 struct task_struct *p = find_lock_task_mm(task);
1067 if (atomic_read(&p->mm->mm_users) > 1) {
1075 task->signal->oom_score_adj = oom_adj;
1076 if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1077 task->signal->oom_score_adj_min = (short)oom_adj;
1078 trace_oom_score_adj_update(task);
1081 struct task_struct *p;
1084 for_each_process(p) {
1085 if (same_thread_group(task, p))
1088 /* do not touch kernel threads or the global init */
1089 if (p->flags & PF_KTHREAD || is_global_init(p))
1093 if (!p->vfork_done && process_shares_mm(p, mm)) {
1094 pr_info("updating oom_score_adj for %d (%s) from %d to %d because it shares mm with %d (%s). Report if this is unexpected.\n",
1095 task_pid_nr(p), p->comm,
1096 p->signal->oom_score_adj, oom_adj,
1097 task_pid_nr(task), task->comm);
1098 p->signal->oom_score_adj = oom_adj;
1099 if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1100 p->signal->oom_score_adj_min = (short)oom_adj;
1108 mutex_unlock(&oom_adj_mutex);
1109 put_task_struct(task);
1114 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1115 * kernels. The effective policy is defined by oom_score_adj, which has a
1116 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1117 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1118 * Processes that become oom disabled via oom_adj will still be oom disabled
1119 * with this implementation.
1121 * oom_adj cannot be removed since existing userspace binaries use it.
1123 static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1124 size_t count, loff_t *ppos)
1126 char buffer[PROC_NUMBUF];
1130 memset(buffer, 0, sizeof(buffer));
1131 if (count > sizeof(buffer) - 1)
1132 count = sizeof(buffer) - 1;
1133 if (copy_from_user(buffer, buf, count)) {
1138 err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1141 if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1142 oom_adj != OOM_DISABLE) {
1148 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1149 * value is always attainable.
1151 if (oom_adj == OOM_ADJUST_MAX)
1152 oom_adj = OOM_SCORE_ADJ_MAX;
1154 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1156 err = __set_oom_adj(file, oom_adj, true);
1158 return err < 0 ? err : count;
1161 static const struct file_operations proc_oom_adj_operations = {
1162 .read = oom_adj_read,
1163 .write = oom_adj_write,
1164 .llseek = generic_file_llseek,
1167 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1168 size_t count, loff_t *ppos)
1170 struct task_struct *task = get_proc_task(file_inode(file));
1171 char buffer[PROC_NUMBUF];
1172 short oom_score_adj = OOM_SCORE_ADJ_MIN;
1177 oom_score_adj = task->signal->oom_score_adj;
1178 put_task_struct(task);
1179 len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
1180 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1183 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1184 size_t count, loff_t *ppos)
1186 char buffer[PROC_NUMBUF];
1190 memset(buffer, 0, sizeof(buffer));
1191 if (count > sizeof(buffer) - 1)
1192 count = sizeof(buffer) - 1;
1193 if (copy_from_user(buffer, buf, count)) {
1198 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1201 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1202 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1207 err = __set_oom_adj(file, oom_score_adj, false);
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 11
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,
1359 static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf,
1360 size_t count, loff_t *ppos)
1362 struct task_struct *task;
1365 task = get_proc_task(file_inode(file));
1368 put_task_struct(task);
1369 if (task != current)
1371 err = kstrtoint_from_user(buf, count, 10, &n);
1374 if (n < 0 || n == INT_MAX)
1376 current->fail_nth = n + 1;
1380 static ssize_t proc_fail_nth_read(struct file *file, char __user *buf,
1381 size_t count, loff_t *ppos)
1383 struct task_struct *task;
1386 task = get_proc_task(file_inode(file));
1389 put_task_struct(task);
1390 if (task != current)
1394 err = put_user((char)(current->fail_nth ? 'N' : 'Y'), buf);
1397 current->fail_nth = 0;
1401 static const struct file_operations proc_fail_nth_operations = {
1402 .read = proc_fail_nth_read,
1403 .write = proc_fail_nth_write,
1408 #ifdef CONFIG_SCHED_DEBUG
1410 * Print out various scheduling related per-task fields:
1412 static int sched_show(struct seq_file *m, void *v)
1414 struct inode *inode = m->private;
1415 struct task_struct *p;
1417 p = get_proc_task(inode);
1420 proc_sched_show_task(p, m);
1428 sched_write(struct file *file, const char __user *buf,
1429 size_t count, loff_t *offset)
1431 struct inode *inode = file_inode(file);
1432 struct task_struct *p;
1434 p = get_proc_task(inode);
1437 proc_sched_set_task(p);
1444 static int sched_open(struct inode *inode, struct file *filp)
1446 return single_open(filp, sched_show, inode);
1449 static const struct file_operations proc_pid_sched_operations = {
1452 .write = sched_write,
1453 .llseek = seq_lseek,
1454 .release = single_release,
1459 #ifdef CONFIG_SCHED_AUTOGROUP
1461 * Print out autogroup related information:
1463 static int sched_autogroup_show(struct seq_file *m, void *v)
1465 struct inode *inode = m->private;
1466 struct task_struct *p;
1468 p = get_proc_task(inode);
1471 proc_sched_autogroup_show_task(p, m);
1479 sched_autogroup_write(struct file *file, const char __user *buf,
1480 size_t count, loff_t *offset)
1482 struct inode *inode = file_inode(file);
1483 struct task_struct *p;
1484 char buffer[PROC_NUMBUF];
1488 memset(buffer, 0, sizeof(buffer));
1489 if (count > sizeof(buffer) - 1)
1490 count = sizeof(buffer) - 1;
1491 if (copy_from_user(buffer, buf, count))
1494 err = kstrtoint(strstrip(buffer), 0, &nice);
1498 p = get_proc_task(inode);
1502 err = proc_sched_autogroup_set_nice(p, nice);
1511 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1515 ret = single_open(filp, sched_autogroup_show, NULL);
1517 struct seq_file *m = filp->private_data;
1524 static const struct file_operations proc_pid_sched_autogroup_operations = {
1525 .open = sched_autogroup_open,
1527 .write = sched_autogroup_write,
1528 .llseek = seq_lseek,
1529 .release = single_release,
1532 #endif /* CONFIG_SCHED_AUTOGROUP */
1534 static ssize_t comm_write(struct file *file, const char __user *buf,
1535 size_t count, loff_t *offset)
1537 struct inode *inode = file_inode(file);
1538 struct task_struct *p;
1539 char buffer[TASK_COMM_LEN];
1540 const size_t maxlen = sizeof(buffer) - 1;
1542 memset(buffer, 0, sizeof(buffer));
1543 if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
1546 p = get_proc_task(inode);
1550 if (same_thread_group(current, p))
1551 set_task_comm(p, buffer);
1560 static int comm_show(struct seq_file *m, void *v)
1562 struct inode *inode = m->private;
1563 struct task_struct *p;
1565 p = get_proc_task(inode);
1570 seq_printf(m, "%s\n", p->comm);
1578 static int comm_open(struct inode *inode, struct file *filp)
1580 return single_open(filp, comm_show, inode);
1583 static const struct file_operations proc_pid_set_comm_operations = {
1586 .write = comm_write,
1587 .llseek = seq_lseek,
1588 .release = single_release,
1591 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1593 struct task_struct *task;
1594 struct file *exe_file;
1596 task = get_proc_task(d_inode(dentry));
1599 exe_file = get_task_exe_file(task);
1600 put_task_struct(task);
1602 *exe_path = exe_file->f_path;
1603 path_get(&exe_file->f_path);
1610 static const char *proc_pid_get_link(struct dentry *dentry,
1611 struct inode *inode,
1612 struct delayed_call *done)
1615 int error = -EACCES;
1618 return ERR_PTR(-ECHILD);
1620 /* Are we allowed to snoop on the tasks file descriptors? */
1621 if (!proc_fd_access_allowed(inode))
1624 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1628 nd_jump_link(&path);
1631 return ERR_PTR(error);
1634 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1636 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1643 pathname = d_path(path, tmp, PAGE_SIZE);
1644 len = PTR_ERR(pathname);
1645 if (IS_ERR(pathname))
1647 len = tmp + PAGE_SIZE - 1 - pathname;
1651 if (copy_to_user(buffer, pathname, len))
1654 free_page((unsigned long)tmp);
1658 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1660 int error = -EACCES;
1661 struct inode *inode = d_inode(dentry);
1664 /* Are we allowed to snoop on the tasks file descriptors? */
1665 if (!proc_fd_access_allowed(inode))
1668 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1672 error = do_proc_readlink(&path, buffer, buflen);
1678 const struct inode_operations proc_pid_link_inode_operations = {
1679 .readlink = proc_pid_readlink,
1680 .get_link = proc_pid_get_link,
1681 .setattr = proc_setattr,
1685 /* building an inode */
1687 void task_dump_owner(struct task_struct *task, mode_t mode,
1688 kuid_t *ruid, kgid_t *rgid)
1690 /* Depending on the state of dumpable compute who should own a
1691 * proc file for a task.
1693 const struct cred *cred;
1697 /* Default to the tasks effective ownership */
1699 cred = __task_cred(task);
1705 * Before the /proc/pid/status file was created the only way to read
1706 * the effective uid of a /process was to stat /proc/pid. Reading
1707 * /proc/pid/status is slow enough that procps and other packages
1708 * kept stating /proc/pid. To keep the rules in /proc simple I have
1709 * made this apply to all per process world readable and executable
1712 if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) {
1713 struct mm_struct *mm;
1716 /* Make non-dumpable tasks owned by some root */
1718 if (get_dumpable(mm) != SUID_DUMP_USER) {
1719 struct user_namespace *user_ns = mm->user_ns;
1721 uid = make_kuid(user_ns, 0);
1722 if (!uid_valid(uid))
1723 uid = GLOBAL_ROOT_UID;
1725 gid = make_kgid(user_ns, 0);
1726 if (!gid_valid(gid))
1727 gid = GLOBAL_ROOT_GID;
1730 uid = GLOBAL_ROOT_UID;
1731 gid = GLOBAL_ROOT_GID;
1739 struct inode *proc_pid_make_inode(struct super_block * sb,
1740 struct task_struct *task, umode_t mode)
1742 struct inode * inode;
1743 struct proc_inode *ei;
1745 /* We need a new inode */
1747 inode = new_inode(sb);
1753 inode->i_mode = mode;
1754 inode->i_ino = get_next_ino();
1755 inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode);
1756 inode->i_op = &proc_def_inode_operations;
1759 * grab the reference to task.
1761 ei->pid = get_task_pid(task, PIDTYPE_PID);
1765 task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
1766 security_task_to_inode(task, inode);
1776 int pid_getattr(const struct path *path, struct kstat *stat,
1777 u32 request_mask, unsigned int query_flags)
1779 struct inode *inode = d_inode(path->dentry);
1780 struct task_struct *task;
1781 struct pid_namespace *pid = path->dentry->d_sb->s_fs_info;
1783 generic_fillattr(inode, stat);
1786 stat->uid = GLOBAL_ROOT_UID;
1787 stat->gid = GLOBAL_ROOT_GID;
1788 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1790 if (!has_pid_permissions(pid, task, HIDEPID_INVISIBLE)) {
1793 * This doesn't prevent learning whether PID exists,
1794 * it only makes getattr() consistent with readdir().
1798 task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid);
1807 * Exceptional case: normally we are not allowed to unhash a busy
1808 * directory. In this case, however, we can do it - no aliasing problems
1809 * due to the way we treat inodes.
1811 * Rewrite the inode's ownerships here because the owning task may have
1812 * performed a setuid(), etc.
1815 int pid_revalidate(struct dentry *dentry, unsigned int flags)
1817 struct inode *inode;
1818 struct task_struct *task;
1820 if (flags & LOOKUP_RCU)
1823 inode = d_inode(dentry);
1824 task = get_proc_task(inode);
1827 task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid);
1829 inode->i_mode &= ~(S_ISUID | S_ISGID);
1830 security_task_to_inode(task, inode);
1831 put_task_struct(task);
1837 static inline bool proc_inode_is_dead(struct inode *inode)
1839 return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1842 int pid_delete_dentry(const struct dentry *dentry)
1844 /* Is the task we represent dead?
1845 * If so, then don't put the dentry on the lru list,
1846 * kill it immediately.
1848 return proc_inode_is_dead(d_inode(dentry));
1851 const struct dentry_operations pid_dentry_operations =
1853 .d_revalidate = pid_revalidate,
1854 .d_delete = pid_delete_dentry,
1860 * Fill a directory entry.
1862 * If possible create the dcache entry and derive our inode number and
1863 * file type from dcache entry.
1865 * Since all of the proc inode numbers are dynamically generated, the inode
1866 * numbers do not exist until the inode is cache. This means creating the
1867 * the dcache entry in readdir is necessary to keep the inode numbers
1868 * reported by readdir in sync with the inode numbers reported
1871 bool proc_fill_cache(struct file *file, struct dir_context *ctx,
1872 const char *name, int len,
1873 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1875 struct dentry *child, *dir = file->f_path.dentry;
1876 struct qstr qname = QSTR_INIT(name, len);
1877 struct inode *inode;
1881 child = d_hash_and_lookup(dir, &qname);
1883 DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq);
1884 child = d_alloc_parallel(dir, &qname, &wq);
1886 goto end_instantiate;
1887 if (d_in_lookup(child)) {
1888 int err = instantiate(d_inode(dir), child, task, ptr);
1889 d_lookup_done(child);
1892 goto end_instantiate;
1896 inode = d_inode(child);
1898 type = inode->i_mode >> 12;
1900 return dir_emit(ctx, name, len, ino, type);
1903 return dir_emit(ctx, name, len, 1, DT_UNKNOWN);
1907 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1908 * which represent vma start and end addresses.
1910 static int dname_to_vma_addr(struct dentry *dentry,
1911 unsigned long *start, unsigned long *end)
1913 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1919 static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
1921 unsigned long vm_start, vm_end;
1922 bool exact_vma_exists = false;
1923 struct mm_struct *mm = NULL;
1924 struct task_struct *task;
1925 struct inode *inode;
1928 if (flags & LOOKUP_RCU)
1931 inode = d_inode(dentry);
1932 task = get_proc_task(inode);
1936 mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
1937 if (IS_ERR_OR_NULL(mm))
1940 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1941 down_read(&mm->mmap_sem);
1942 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1943 up_read(&mm->mmap_sem);
1948 if (exact_vma_exists) {
1949 task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
1951 security_task_to_inode(task, inode);
1956 put_task_struct(task);
1962 static const struct dentry_operations tid_map_files_dentry_operations = {
1963 .d_revalidate = map_files_d_revalidate,
1964 .d_delete = pid_delete_dentry,
1967 static int map_files_get_link(struct dentry *dentry, struct path *path)
1969 unsigned long vm_start, vm_end;
1970 struct vm_area_struct *vma;
1971 struct task_struct *task;
1972 struct mm_struct *mm;
1976 task = get_proc_task(d_inode(dentry));
1980 mm = get_task_mm(task);
1981 put_task_struct(task);
1985 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
1990 down_read(&mm->mmap_sem);
1991 vma = find_exact_vma(mm, vm_start, vm_end);
1992 if (vma && vma->vm_file) {
1993 *path = vma->vm_file->f_path;
1997 up_read(&mm->mmap_sem);
2005 struct map_files_info {
2008 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
2012 * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the
2013 * symlinks may be used to bypass permissions on ancestor directories in the
2014 * path to the file in question.
2017 proc_map_files_get_link(struct dentry *dentry,
2018 struct inode *inode,
2019 struct delayed_call *done)
2021 if (!capable(CAP_SYS_ADMIN))
2022 return ERR_PTR(-EPERM);
2024 return proc_pid_get_link(dentry, inode, done);
2028 * Identical to proc_pid_link_inode_operations except for get_link()
2030 static const struct inode_operations proc_map_files_link_inode_operations = {
2031 .readlink = proc_pid_readlink,
2032 .get_link = proc_map_files_get_link,
2033 .setattr = proc_setattr,
2037 proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
2038 struct task_struct *task, const void *ptr)
2040 fmode_t mode = (fmode_t)(unsigned long)ptr;
2041 struct proc_inode *ei;
2042 struct inode *inode;
2044 inode = proc_pid_make_inode(dir->i_sb, task, S_IFLNK |
2045 ((mode & FMODE_READ ) ? S_IRUSR : 0) |
2046 ((mode & FMODE_WRITE) ? S_IWUSR : 0));
2051 ei->op.proc_get_link = map_files_get_link;
2053 inode->i_op = &proc_map_files_link_inode_operations;
2056 d_set_d_op(dentry, &tid_map_files_dentry_operations);
2057 d_add(dentry, inode);
2062 static struct dentry *proc_map_files_lookup(struct inode *dir,
2063 struct dentry *dentry, unsigned int flags)
2065 unsigned long vm_start, vm_end;
2066 struct vm_area_struct *vma;
2067 struct task_struct *task;
2069 struct mm_struct *mm;
2072 task = get_proc_task(dir);
2077 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2081 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2084 mm = get_task_mm(task);
2088 down_read(&mm->mmap_sem);
2089 vma = find_exact_vma(mm, vm_start, vm_end);
2094 result = proc_map_files_instantiate(dir, dentry, task,
2095 (void *)(unsigned long)vma->vm_file->f_mode);
2098 up_read(&mm->mmap_sem);
2101 put_task_struct(task);
2103 return ERR_PTR(result);
2106 static const struct inode_operations proc_map_files_inode_operations = {
2107 .lookup = proc_map_files_lookup,
2108 .permission = proc_fd_permission,
2109 .setattr = proc_setattr,
2113 proc_map_files_readdir(struct file *file, struct dir_context *ctx)
2115 struct vm_area_struct *vma;
2116 struct task_struct *task;
2117 struct mm_struct *mm;
2118 unsigned long nr_files, pos, i;
2119 struct flex_array *fa = NULL;
2120 struct map_files_info info;
2121 struct map_files_info *p;
2125 task = get_proc_task(file_inode(file));
2130 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2134 if (!dir_emit_dots(file, ctx))
2137 mm = get_task_mm(task);
2140 down_read(&mm->mmap_sem);
2145 * We need two passes here:
2147 * 1) Collect vmas of mapped files with mmap_sem taken
2148 * 2) Release mmap_sem and instantiate entries
2150 * otherwise we get lockdep complained, since filldir()
2151 * routine might require mmap_sem taken in might_fault().
2154 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2155 if (vma->vm_file && ++pos > ctx->pos)
2160 fa = flex_array_alloc(sizeof(info), nr_files,
2162 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2166 flex_array_free(fa);
2167 up_read(&mm->mmap_sem);
2171 for (i = 0, vma = mm->mmap, pos = 2; vma;
2172 vma = vma->vm_next) {
2175 if (++pos <= ctx->pos)
2178 info.mode = vma->vm_file->f_mode;
2179 info.len = snprintf(info.name,
2180 sizeof(info.name), "%lx-%lx",
2181 vma->vm_start, vma->vm_end);
2182 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2186 up_read(&mm->mmap_sem);
2188 for (i = 0; i < nr_files; i++) {
2189 p = flex_array_get(fa, i);
2190 if (!proc_fill_cache(file, ctx,
2192 proc_map_files_instantiate,
2194 (void *)(unsigned long)p->mode))
2199 flex_array_free(fa);
2203 put_task_struct(task);
2208 static const struct file_operations proc_map_files_operations = {
2209 .read = generic_read_dir,
2210 .iterate_shared = proc_map_files_readdir,
2211 .llseek = generic_file_llseek,
2214 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
2215 struct timers_private {
2217 struct task_struct *task;
2218 struct sighand_struct *sighand;
2219 struct pid_namespace *ns;
2220 unsigned long flags;
2223 static void *timers_start(struct seq_file *m, loff_t *pos)
2225 struct timers_private *tp = m->private;
2227 tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2229 return ERR_PTR(-ESRCH);
2231 tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2233 return ERR_PTR(-ESRCH);
2235 return seq_list_start(&tp->task->signal->posix_timers, *pos);
2238 static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2240 struct timers_private *tp = m->private;
2241 return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2244 static void timers_stop(struct seq_file *m, void *v)
2246 struct timers_private *tp = m->private;
2249 unlock_task_sighand(tp->task, &tp->flags);
2254 put_task_struct(tp->task);
2259 static int show_timer(struct seq_file *m, void *v)
2261 struct k_itimer *timer;
2262 struct timers_private *tp = m->private;
2264 static const char * const nstr[] = {
2265 [SIGEV_SIGNAL] = "signal",
2266 [SIGEV_NONE] = "none",
2267 [SIGEV_THREAD] = "thread",
2270 timer = list_entry((struct list_head *)v, struct k_itimer, list);
2271 notify = timer->it_sigev_notify;
2273 seq_printf(m, "ID: %d\n", timer->it_id);
2274 seq_printf(m, "signal: %d/%p\n",
2275 timer->sigq->info.si_signo,
2276 timer->sigq->info.si_value.sival_ptr);
2277 seq_printf(m, "notify: %s/%s.%d\n",
2278 nstr[notify & ~SIGEV_THREAD_ID],
2279 (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2280 pid_nr_ns(timer->it_pid, tp->ns));
2281 seq_printf(m, "ClockID: %d\n", timer->it_clock);
2286 static const struct seq_operations proc_timers_seq_ops = {
2287 .start = timers_start,
2288 .next = timers_next,
2289 .stop = timers_stop,
2293 static int proc_timers_open(struct inode *inode, struct file *file)
2295 struct timers_private *tp;
2297 tp = __seq_open_private(file, &proc_timers_seq_ops,
2298 sizeof(struct timers_private));
2302 tp->pid = proc_pid(inode);
2303 tp->ns = inode->i_sb->s_fs_info;
2307 static const struct file_operations proc_timers_operations = {
2308 .open = proc_timers_open,
2310 .llseek = seq_lseek,
2311 .release = seq_release_private,
2315 static ssize_t timerslack_ns_write(struct file *file, const char __user *buf,
2316 size_t count, loff_t *offset)
2318 struct inode *inode = file_inode(file);
2319 struct task_struct *p;
2323 err = kstrtoull_from_user(buf, count, 10, &slack_ns);
2327 p = get_proc_task(inode);
2332 if (!capable(CAP_SYS_NICE)) {
2337 err = security_task_setscheduler(p);
2346 p->timer_slack_ns = p->default_timer_slack_ns;
2348 p->timer_slack_ns = slack_ns;
2357 static int timerslack_ns_show(struct seq_file *m, void *v)
2359 struct inode *inode = m->private;
2360 struct task_struct *p;
2363 p = get_proc_task(inode);
2369 if (!capable(CAP_SYS_NICE)) {
2373 err = security_task_getscheduler(p);
2379 seq_printf(m, "%llu\n", p->timer_slack_ns);
2388 static int timerslack_ns_open(struct inode *inode, struct file *filp)
2390 return single_open(filp, timerslack_ns_show, inode);
2393 static const struct file_operations proc_pid_set_timerslack_ns_operations = {
2394 .open = timerslack_ns_open,
2396 .write = timerslack_ns_write,
2397 .llseek = seq_lseek,
2398 .release = single_release,
2401 static int proc_pident_instantiate(struct inode *dir,
2402 struct dentry *dentry, struct task_struct *task, const void *ptr)
2404 const struct pid_entry *p = ptr;
2405 struct inode *inode;
2406 struct proc_inode *ei;
2408 inode = proc_pid_make_inode(dir->i_sb, task, p->mode);
2413 if (S_ISDIR(inode->i_mode))
2414 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2416 inode->i_op = p->iop;
2418 inode->i_fop = p->fop;
2420 d_set_d_op(dentry, &pid_dentry_operations);
2421 d_add(dentry, inode);
2422 /* Close the race of the process dying before we return the dentry */
2423 if (pid_revalidate(dentry, 0))
2429 static struct dentry *proc_pident_lookup(struct inode *dir,
2430 struct dentry *dentry,
2431 const struct pid_entry *ents,
2435 struct task_struct *task = get_proc_task(dir);
2436 const struct pid_entry *p, *last;
2444 * Yes, it does not scale. And it should not. Don't add
2445 * new entries into /proc/<tgid>/ without very good reasons.
2447 last = &ents[nents];
2448 for (p = ents; p < last; p++) {
2449 if (p->len != dentry->d_name.len)
2451 if (!memcmp(dentry->d_name.name, p->name, p->len))
2457 error = proc_pident_instantiate(dir, dentry, task, p);
2459 put_task_struct(task);
2461 return ERR_PTR(error);
2464 static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
2465 const struct pid_entry *ents, unsigned int nents)
2467 struct task_struct *task = get_proc_task(file_inode(file));
2468 const struct pid_entry *p;
2473 if (!dir_emit_dots(file, ctx))
2476 if (ctx->pos >= nents + 2)
2479 for (p = ents + (ctx->pos - 2); p < ents + nents; p++) {
2480 if (!proc_fill_cache(file, ctx, p->name, p->len,
2481 proc_pident_instantiate, task, p))
2486 put_task_struct(task);
2490 #ifdef CONFIG_SECURITY
2491 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2492 size_t count, loff_t *ppos)
2494 struct inode * inode = file_inode(file);
2497 struct task_struct *task = get_proc_task(inode);
2502 length = security_getprocattr(task,
2503 (char*)file->f_path.dentry->d_name.name,
2505 put_task_struct(task);
2507 length = simple_read_from_buffer(buf, count, ppos, p, length);
2512 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2513 size_t count, loff_t *ppos)
2515 struct inode * inode = file_inode(file);
2518 struct task_struct *task = get_proc_task(inode);
2524 /* A task may only write its own attributes. */
2526 if (current != task)
2529 if (count > PAGE_SIZE)
2532 /* No partial writes. */
2537 page = memdup_user(buf, count);
2539 length = PTR_ERR(page);
2543 /* Guard against adverse ptrace interaction */
2544 length = mutex_lock_interruptible(¤t->signal->cred_guard_mutex);
2548 length = security_setprocattr(file->f_path.dentry->d_name.name,
2550 mutex_unlock(¤t->signal->cred_guard_mutex);
2554 put_task_struct(task);
2559 static const struct file_operations proc_pid_attr_operations = {
2560 .read = proc_pid_attr_read,
2561 .write = proc_pid_attr_write,
2562 .llseek = generic_file_llseek,
2565 static const struct pid_entry attr_dir_stuff[] = {
2566 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2567 REG("prev", S_IRUGO, proc_pid_attr_operations),
2568 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2569 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2570 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2571 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2574 static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
2576 return proc_pident_readdir(file, ctx,
2577 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2580 static const struct file_operations proc_attr_dir_operations = {
2581 .read = generic_read_dir,
2582 .iterate_shared = proc_attr_dir_readdir,
2583 .llseek = generic_file_llseek,
2586 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2587 struct dentry *dentry, unsigned int flags)
2589 return proc_pident_lookup(dir, dentry,
2590 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2593 static const struct inode_operations proc_attr_dir_inode_operations = {
2594 .lookup = proc_attr_dir_lookup,
2595 .getattr = pid_getattr,
2596 .setattr = proc_setattr,
2601 #ifdef CONFIG_ELF_CORE
2602 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2603 size_t count, loff_t *ppos)
2605 struct task_struct *task = get_proc_task(file_inode(file));
2606 struct mm_struct *mm;
2607 char buffer[PROC_NUMBUF];
2615 mm = get_task_mm(task);
2617 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2618 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2619 MMF_DUMP_FILTER_SHIFT));
2621 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2624 put_task_struct(task);
2629 static ssize_t proc_coredump_filter_write(struct file *file,
2630 const char __user *buf,
2634 struct task_struct *task;
2635 struct mm_struct *mm;
2641 ret = kstrtouint_from_user(buf, count, 0, &val);
2646 task = get_proc_task(file_inode(file));
2650 mm = get_task_mm(task);
2655 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2657 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2659 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2664 put_task_struct(task);
2671 static const struct file_operations proc_coredump_filter_operations = {
2672 .read = proc_coredump_filter_read,
2673 .write = proc_coredump_filter_write,
2674 .llseek = generic_file_llseek,
2678 #ifdef CONFIG_TASK_IO_ACCOUNTING
2679 static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
2681 struct task_io_accounting acct = task->ioac;
2682 unsigned long flags;
2685 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2689 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) {
2694 if (whole && lock_task_sighand(task, &flags)) {
2695 struct task_struct *t = task;
2697 task_io_accounting_add(&acct, &task->signal->ioac);
2698 while_each_thread(task, t)
2699 task_io_accounting_add(&acct, &t->ioac);
2701 unlock_task_sighand(task, &flags);
2708 "read_bytes: %llu\n"
2709 "write_bytes: %llu\n"
2710 "cancelled_write_bytes: %llu\n",
2711 (unsigned long long)acct.rchar,
2712 (unsigned long long)acct.wchar,
2713 (unsigned long long)acct.syscr,
2714 (unsigned long long)acct.syscw,
2715 (unsigned long long)acct.read_bytes,
2716 (unsigned long long)acct.write_bytes,
2717 (unsigned long long)acct.cancelled_write_bytes);
2721 mutex_unlock(&task->signal->cred_guard_mutex);
2725 static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2726 struct pid *pid, struct task_struct *task)
2728 return do_io_accounting(task, m, 0);
2731 static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2732 struct pid *pid, struct task_struct *task)
2734 return do_io_accounting(task, m, 1);
2736 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2738 #ifdef CONFIG_USER_NS
2739 static int proc_id_map_open(struct inode *inode, struct file *file,
2740 const struct seq_operations *seq_ops)
2742 struct user_namespace *ns = NULL;
2743 struct task_struct *task;
2744 struct seq_file *seq;
2747 task = get_proc_task(inode);
2750 ns = get_user_ns(task_cred_xxx(task, user_ns));
2752 put_task_struct(task);
2757 ret = seq_open(file, seq_ops);
2761 seq = file->private_data;
2771 static int proc_id_map_release(struct inode *inode, struct file *file)
2773 struct seq_file *seq = file->private_data;
2774 struct user_namespace *ns = seq->private;
2776 return seq_release(inode, file);
2779 static int proc_uid_map_open(struct inode *inode, struct file *file)
2781 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2784 static int proc_gid_map_open(struct inode *inode, struct file *file)
2786 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2789 static int proc_projid_map_open(struct inode *inode, struct file *file)
2791 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2794 static const struct file_operations proc_uid_map_operations = {
2795 .open = proc_uid_map_open,
2796 .write = proc_uid_map_write,
2798 .llseek = seq_lseek,
2799 .release = proc_id_map_release,
2802 static const struct file_operations proc_gid_map_operations = {
2803 .open = proc_gid_map_open,
2804 .write = proc_gid_map_write,
2806 .llseek = seq_lseek,
2807 .release = proc_id_map_release,
2810 static const struct file_operations proc_projid_map_operations = {
2811 .open = proc_projid_map_open,
2812 .write = proc_projid_map_write,
2814 .llseek = seq_lseek,
2815 .release = proc_id_map_release,
2818 static int proc_setgroups_open(struct inode *inode, struct file *file)
2820 struct user_namespace *ns = NULL;
2821 struct task_struct *task;
2825 task = get_proc_task(inode);
2828 ns = get_user_ns(task_cred_xxx(task, user_ns));
2830 put_task_struct(task);
2835 if (file->f_mode & FMODE_WRITE) {
2837 if (!ns_capable(ns, CAP_SYS_ADMIN))
2841 ret = single_open(file, &proc_setgroups_show, ns);
2852 static int proc_setgroups_release(struct inode *inode, struct file *file)
2854 struct seq_file *seq = file->private_data;
2855 struct user_namespace *ns = seq->private;
2856 int ret = single_release(inode, file);
2861 static const struct file_operations proc_setgroups_operations = {
2862 .open = proc_setgroups_open,
2863 .write = proc_setgroups_write,
2865 .llseek = seq_lseek,
2866 .release = proc_setgroups_release,
2868 #endif /* CONFIG_USER_NS */
2870 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2871 struct pid *pid, struct task_struct *task)
2873 int err = lock_trace(task);
2875 seq_printf(m, "%08x\n", task->personality);
2881 #ifdef CONFIG_LIVEPATCH
2882 static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns,
2883 struct pid *pid, struct task_struct *task)
2885 seq_printf(m, "%d\n", task->patch_state);
2888 #endif /* CONFIG_LIVEPATCH */
2893 static const struct file_operations proc_task_operations;
2894 static const struct inode_operations proc_task_inode_operations;
2896 static const struct pid_entry tgid_base_stuff[] = {
2897 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2898 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2899 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
2900 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2901 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2903 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2905 REG("environ", S_IRUSR, proc_environ_operations),
2906 REG("auxv", S_IRUSR, proc_auxv_operations),
2907 ONE("status", S_IRUGO, proc_pid_status),
2908 ONE("personality", S_IRUSR, proc_pid_personality),
2909 ONE("limits", S_IRUGO, proc_pid_limits),
2910 #ifdef CONFIG_SCHED_DEBUG
2911 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2913 #ifdef CONFIG_SCHED_AUTOGROUP
2914 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2916 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2917 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2918 ONE("syscall", S_IRUSR, proc_pid_syscall),
2920 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
2921 ONE("stat", S_IRUGO, proc_tgid_stat),
2922 ONE("statm", S_IRUGO, proc_pid_statm),
2923 REG("maps", S_IRUGO, proc_pid_maps_operations),
2925 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
2927 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2928 LNK("cwd", proc_cwd_link),
2929 LNK("root", proc_root_link),
2930 LNK("exe", proc_exe_link),
2931 REG("mounts", S_IRUGO, proc_mounts_operations),
2932 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2933 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2934 #ifdef CONFIG_PROC_PAGE_MONITOR
2935 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2936 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
2937 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2939 #ifdef CONFIG_SECURITY
2940 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2942 #ifdef CONFIG_KALLSYMS
2943 ONE("wchan", S_IRUGO, proc_pid_wchan),
2945 #ifdef CONFIG_STACKTRACE
2946 ONE("stack", S_IRUSR, proc_pid_stack),
2948 #ifdef CONFIG_SCHED_INFO
2949 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
2951 #ifdef CONFIG_LATENCYTOP
2952 REG("latency", S_IRUGO, proc_lstats_operations),
2954 #ifdef CONFIG_PROC_PID_CPUSET
2955 ONE("cpuset", S_IRUGO, proc_cpuset_show),
2957 #ifdef CONFIG_CGROUPS
2958 ONE("cgroup", S_IRUGO, proc_cgroup_show),
2960 ONE("oom_score", S_IRUGO, proc_oom_score),
2961 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
2962 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2963 #ifdef CONFIG_AUDITSYSCALL
2964 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2965 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2967 #ifdef CONFIG_FAULT_INJECTION
2968 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2970 #ifdef CONFIG_ELF_CORE
2971 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2973 #ifdef CONFIG_TASK_IO_ACCOUNTING
2974 ONE("io", S_IRUSR, proc_tgid_io_accounting),
2976 #ifdef CONFIG_HARDWALL
2977 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
2979 #ifdef CONFIG_USER_NS
2980 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
2981 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
2982 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
2983 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
2985 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
2986 REG("timers", S_IRUGO, proc_timers_operations),
2988 REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations),
2989 #ifdef CONFIG_LIVEPATCH
2990 ONE("patch_state", S_IRUSR, proc_pid_patch_state),
2994 static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
2996 return proc_pident_readdir(file, ctx,
2997 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3000 static const struct file_operations proc_tgid_base_operations = {
3001 .read = generic_read_dir,
3002 .iterate_shared = proc_tgid_base_readdir,
3003 .llseek = generic_file_llseek,
3006 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3008 return proc_pident_lookup(dir, dentry,
3009 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3012 static const struct inode_operations proc_tgid_base_inode_operations = {
3013 .lookup = proc_tgid_base_lookup,
3014 .getattr = pid_getattr,
3015 .setattr = proc_setattr,
3016 .permission = proc_pid_permission,
3019 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
3021 struct dentry *dentry, *leader, *dir;
3022 char buf[PROC_NUMBUF];
3026 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3027 /* no ->d_hash() rejects on procfs */
3028 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
3030 d_invalidate(dentry);
3038 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
3039 leader = d_hash_and_lookup(mnt->mnt_root, &name);
3044 name.len = strlen(name.name);
3045 dir = d_hash_and_lookup(leader, &name);
3047 goto out_put_leader;
3050 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3051 dentry = d_hash_and_lookup(dir, &name);
3053 d_invalidate(dentry);
3065 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3066 * @task: task that should be flushed.
3068 * When flushing dentries from proc, one needs to flush them from global
3069 * proc (proc_mnt) and from all the namespaces' procs this task was seen
3070 * in. This call is supposed to do all of this job.
3072 * Looks in the dcache for
3074 * /proc/@tgid/task/@pid
3075 * if either directory is present flushes it and all of it'ts children
3078 * It is safe and reasonable to cache /proc entries for a task until
3079 * that task exits. After that they just clog up the dcache with
3080 * useless entries, possibly causing useful dcache entries to be
3081 * flushed instead. This routine is proved to flush those useless
3082 * dcache entries at process exit time.
3084 * NOTE: This routine is just an optimization so it does not guarantee
3085 * that no dcache entries will exist at process exit time it
3086 * just makes it very unlikely that any will persist.
3089 void proc_flush_task(struct task_struct *task)
3092 struct pid *pid, *tgid;
3095 pid = task_pid(task);
3096 tgid = task_tgid(task);
3098 for (i = 0; i <= pid->level; i++) {
3099 upid = &pid->numbers[i];
3100 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
3101 tgid->numbers[i].nr);
3105 static int proc_pid_instantiate(struct inode *dir,
3106 struct dentry * dentry,
3107 struct task_struct *task, const void *ptr)
3109 struct inode *inode;
3111 inode = proc_pid_make_inode(dir->i_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
3115 inode->i_op = &proc_tgid_base_inode_operations;
3116 inode->i_fop = &proc_tgid_base_operations;
3117 inode->i_flags|=S_IMMUTABLE;
3119 set_nlink(inode, nlink_tgid);
3121 d_set_d_op(dentry, &pid_dentry_operations);
3123 d_add(dentry, inode);
3124 /* Close the race of the process dying before we return the dentry */
3125 if (pid_revalidate(dentry, 0))
3131 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3133 int result = -ENOENT;
3134 struct task_struct *task;
3136 struct pid_namespace *ns;
3138 tgid = name_to_int(&dentry->d_name);
3142 ns = dentry->d_sb->s_fs_info;
3144 task = find_task_by_pid_ns(tgid, ns);
3146 get_task_struct(task);
3151 result = proc_pid_instantiate(dir, dentry, task, NULL);
3152 put_task_struct(task);
3154 return ERR_PTR(result);
3158 * Find the first task with tgid >= tgid
3163 struct task_struct *task;
3165 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3170 put_task_struct(iter.task);
3174 pid = find_ge_pid(iter.tgid, ns);
3176 iter.tgid = pid_nr_ns(pid, ns);
3177 iter.task = pid_task(pid, PIDTYPE_PID);
3178 /* What we to know is if the pid we have find is the
3179 * pid of a thread_group_leader. Testing for task
3180 * being a thread_group_leader is the obvious thing
3181 * todo but there is a window when it fails, due to
3182 * the pid transfer logic in de_thread.
3184 * So we perform the straight forward test of seeing
3185 * if the pid we have found is the pid of a thread
3186 * group leader, and don't worry if the task we have
3187 * found doesn't happen to be a thread group leader.
3188 * As we don't care in the case of readdir.
3190 if (!iter.task || !has_group_leader_pid(iter.task)) {
3194 get_task_struct(iter.task);
3200 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3202 /* for the /proc/ directory itself, after non-process stuff has been done */
3203 int proc_pid_readdir(struct file *file, struct dir_context *ctx)
3205 struct tgid_iter iter;
3206 struct pid_namespace *ns = file_inode(file)->i_sb->s_fs_info;
3207 loff_t pos = ctx->pos;
3209 if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
3212 if (pos == TGID_OFFSET - 2) {
3213 struct inode *inode = d_inode(ns->proc_self);
3214 if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
3216 ctx->pos = pos = pos + 1;
3218 if (pos == TGID_OFFSET - 1) {
3219 struct inode *inode = d_inode(ns->proc_thread_self);
3220 if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3222 ctx->pos = pos = pos + 1;
3224 iter.tgid = pos - TGID_OFFSET;
3226 for (iter = next_tgid(ns, iter);
3228 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3229 char name[PROC_NUMBUF];
3233 if (!has_pid_permissions(ns, iter.task, HIDEPID_INVISIBLE))
3236 len = snprintf(name, sizeof(name), "%d", iter.tgid);
3237 ctx->pos = iter.tgid + TGID_OFFSET;
3238 if (!proc_fill_cache(file, ctx, name, len,
3239 proc_pid_instantiate, iter.task, NULL)) {
3240 put_task_struct(iter.task);
3244 ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
3249 * proc_tid_comm_permission is a special permission function exclusively
3250 * used for the node /proc/<pid>/task/<tid>/comm.
3251 * It bypasses generic permission checks in the case where a task of the same
3252 * task group attempts to access the node.
3253 * The rationale behind this is that glibc and bionic access this node for
3254 * cross thread naming (pthread_set/getname_np(!self)). However, if
3255 * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3256 * which locks out the cross thread naming implementation.
3257 * This function makes sure that the node is always accessible for members of
3258 * same thread group.
3260 static int proc_tid_comm_permission(struct inode *inode, int mask)
3262 bool is_same_tgroup;
3263 struct task_struct *task;
3265 task = get_proc_task(inode);
3268 is_same_tgroup = same_thread_group(current, task);
3269 put_task_struct(task);
3271 if (likely(is_same_tgroup && !(mask & MAY_EXEC))) {
3272 /* This file (/proc/<pid>/task/<tid>/comm) can always be
3273 * read or written by the members of the corresponding
3279 return generic_permission(inode, mask);
3282 static const struct inode_operations proc_tid_comm_inode_operations = {
3283 .permission = proc_tid_comm_permission,
3289 static const struct pid_entry tid_base_stuff[] = {
3290 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3291 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3292 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3294 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3296 REG("environ", S_IRUSR, proc_environ_operations),
3297 REG("auxv", S_IRUSR, proc_auxv_operations),
3298 ONE("status", S_IRUGO, proc_pid_status),
3299 ONE("personality", S_IRUSR, proc_pid_personality),
3300 ONE("limits", S_IRUGO, proc_pid_limits),
3301 #ifdef CONFIG_SCHED_DEBUG
3302 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3304 NOD("comm", S_IFREG|S_IRUGO|S_IWUSR,
3305 &proc_tid_comm_inode_operations,
3306 &proc_pid_set_comm_operations, {}),
3307 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3308 ONE("syscall", S_IRUSR, proc_pid_syscall),
3310 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
3311 ONE("stat", S_IRUGO, proc_tid_stat),
3312 ONE("statm", S_IRUGO, proc_pid_statm),
3313 REG("maps", S_IRUGO, proc_tid_maps_operations),
3314 #ifdef CONFIG_PROC_CHILDREN
3315 REG("children", S_IRUGO, proc_tid_children_operations),
3318 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
3320 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3321 LNK("cwd", proc_cwd_link),
3322 LNK("root", proc_root_link),
3323 LNK("exe", proc_exe_link),
3324 REG("mounts", S_IRUGO, proc_mounts_operations),
3325 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3326 #ifdef CONFIG_PROC_PAGE_MONITOR
3327 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3328 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
3329 REG("pagemap", S_IRUSR, proc_pagemap_operations),
3331 #ifdef CONFIG_SECURITY
3332 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3334 #ifdef CONFIG_KALLSYMS
3335 ONE("wchan", S_IRUGO, proc_pid_wchan),
3337 #ifdef CONFIG_STACKTRACE
3338 ONE("stack", S_IRUSR, proc_pid_stack),
3340 #ifdef CONFIG_SCHED_INFO
3341 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
3343 #ifdef CONFIG_LATENCYTOP
3344 REG("latency", S_IRUGO, proc_lstats_operations),
3346 #ifdef CONFIG_PROC_PID_CPUSET
3347 ONE("cpuset", S_IRUGO, proc_cpuset_show),
3349 #ifdef CONFIG_CGROUPS
3350 ONE("cgroup", S_IRUGO, proc_cgroup_show),
3352 ONE("oom_score", S_IRUGO, proc_oom_score),
3353 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3354 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3355 #ifdef CONFIG_AUDITSYSCALL
3356 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3357 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3359 #ifdef CONFIG_FAULT_INJECTION
3360 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3362 * Operations on the file check that the task is current,
3363 * so we create it with 0666 to support testing under unprivileged user.
3365 REG("fail-nth", 0666, proc_fail_nth_operations),
3367 #ifdef CONFIG_TASK_IO_ACCOUNTING
3368 ONE("io", S_IRUSR, proc_tid_io_accounting),
3370 #ifdef CONFIG_HARDWALL
3371 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
3373 #ifdef CONFIG_USER_NS
3374 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3375 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3376 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3377 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
3379 #ifdef CONFIG_LIVEPATCH
3380 ONE("patch_state", S_IRUSR, proc_pid_patch_state),
3384 static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
3386 return proc_pident_readdir(file, ctx,
3387 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3390 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3392 return proc_pident_lookup(dir, dentry,
3393 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3396 static const struct file_operations proc_tid_base_operations = {
3397 .read = generic_read_dir,
3398 .iterate_shared = proc_tid_base_readdir,
3399 .llseek = generic_file_llseek,
3402 static const struct inode_operations proc_tid_base_inode_operations = {
3403 .lookup = proc_tid_base_lookup,
3404 .getattr = pid_getattr,
3405 .setattr = proc_setattr,
3408 static int proc_task_instantiate(struct inode *dir,
3409 struct dentry *dentry, struct task_struct *task, const void *ptr)
3411 struct inode *inode;
3412 inode = proc_pid_make_inode(dir->i_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
3416 inode->i_op = &proc_tid_base_inode_operations;
3417 inode->i_fop = &proc_tid_base_operations;
3418 inode->i_flags|=S_IMMUTABLE;
3420 set_nlink(inode, nlink_tid);
3422 d_set_d_op(dentry, &pid_dentry_operations);
3424 d_add(dentry, inode);
3425 /* Close the race of the process dying before we return the dentry */
3426 if (pid_revalidate(dentry, 0))
3432 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3434 int result = -ENOENT;
3435 struct task_struct *task;
3436 struct task_struct *leader = get_proc_task(dir);
3438 struct pid_namespace *ns;
3443 tid = name_to_int(&dentry->d_name);
3447 ns = dentry->d_sb->s_fs_info;
3449 task = find_task_by_pid_ns(tid, ns);
3451 get_task_struct(task);
3455 if (!same_thread_group(leader, task))
3458 result = proc_task_instantiate(dir, dentry, task, NULL);
3460 put_task_struct(task);
3462 put_task_struct(leader);
3464 return ERR_PTR(result);
3468 * Find the first tid of a thread group to return to user space.
3470 * Usually this is just the thread group leader, but if the users
3471 * buffer was too small or there was a seek into the middle of the
3472 * directory we have more work todo.
3474 * In the case of a short read we start with find_task_by_pid.
3476 * In the case of a seek we start with the leader and walk nr
3479 static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3480 struct pid_namespace *ns)
3482 struct task_struct *pos, *task;
3483 unsigned long nr = f_pos;
3485 if (nr != f_pos) /* 32bit overflow? */
3489 task = pid_task(pid, PIDTYPE_PID);
3493 /* Attempt to start with the tid of a thread */
3495 pos = find_task_by_pid_ns(tid, ns);
3496 if (pos && same_thread_group(pos, task))
3500 /* If nr exceeds the number of threads there is nothing todo */
3501 if (nr >= get_nr_threads(task))
3504 /* If we haven't found our starting place yet start
3505 * with the leader and walk nr threads forward.
3507 pos = task = task->group_leader;
3511 } while_each_thread(task, pos);
3516 get_task_struct(pos);
3523 * Find the next thread in the thread list.
3524 * Return NULL if there is an error or no next thread.
3526 * The reference to the input task_struct is released.
3528 static struct task_struct *next_tid(struct task_struct *start)
3530 struct task_struct *pos = NULL;
3532 if (pid_alive(start)) {
3533 pos = next_thread(start);
3534 if (thread_group_leader(pos))
3537 get_task_struct(pos);
3540 put_task_struct(start);
3544 /* for the /proc/TGID/task/ directories */
3545 static int proc_task_readdir(struct file *file, struct dir_context *ctx)
3547 struct inode *inode = file_inode(file);
3548 struct task_struct *task;
3549 struct pid_namespace *ns;
3552 if (proc_inode_is_dead(inode))
3555 if (!dir_emit_dots(file, ctx))
3558 /* f_version caches the tgid value that the last readdir call couldn't
3559 * return. lseek aka telldir automagically resets f_version to 0.
3561 ns = inode->i_sb->s_fs_info;
3562 tid = (int)file->f_version;
3563 file->f_version = 0;
3564 for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
3566 task = next_tid(task), ctx->pos++) {
3567 char name[PROC_NUMBUF];
3569 tid = task_pid_nr_ns(task, ns);
3570 len = snprintf(name, sizeof(name), "%d", tid);
3571 if (!proc_fill_cache(file, ctx, name, len,
3572 proc_task_instantiate, task, NULL)) {
3573 /* returning this tgid failed, save it as the first
3574 * pid for the next readir call */
3575 file->f_version = (u64)tid;
3576 put_task_struct(task);
3584 static int proc_task_getattr(const struct path *path, struct kstat *stat,
3585 u32 request_mask, unsigned int query_flags)
3587 struct inode *inode = d_inode(path->dentry);
3588 struct task_struct *p = get_proc_task(inode);
3589 generic_fillattr(inode, stat);
3592 stat->nlink += get_nr_threads(p);
3599 static const struct inode_operations proc_task_inode_operations = {
3600 .lookup = proc_task_lookup,
3601 .getattr = proc_task_getattr,
3602 .setattr = proc_setattr,
3603 .permission = proc_pid_permission,
3606 static const struct file_operations proc_task_operations = {
3607 .read = generic_read_dir,
3608 .iterate_shared = proc_task_readdir,
3609 .llseek = generic_file_llseek,
3612 void __init set_proc_pid_nlink(void)
3614 nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3615 nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));