1 // SPDX-License-Identifier: GPL-2.0
5 * Copyright (C) 1991, 1992 Linus Torvalds
7 * proc base directory handling functions
9 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
10 * Instead of using magical inumbers to determine the kind of object
11 * we allocate and fill in-core inodes upon lookup. They don't even
12 * go into icache. We cache the reference to task_struct upon lookup too.
13 * Eventually it should become a filesystem in its own. We don't use the
14 * rest of procfs anymore.
20 * Bruna Moreira <bruna.moreira@indt.org.br>
21 * Edjard Mota <edjard.mota@indt.org.br>
22 * Ilias Biris <ilias.biris@indt.org.br>
23 * Mauricio Lin <mauricio.lin@indt.org.br>
25 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
27 * A new process specific entry (smaps) included in /proc. It shows the
28 * size of rss for each memory area. The maps entry lacks information
29 * about physical memory size (rss) for each mapped file, i.e.,
30 * rss information for executables and library files.
31 * This additional information is useful for any tools that need to know
32 * about physical memory consumption for a process specific library.
36 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
37 * Pud inclusion in the page table walking.
41 * 10LE Instituto Nokia de Tecnologia - INdT:
42 * A better way to walks through the page table as suggested by Hugh Dickins.
44 * Simo Piiroinen <simo.piiroinen@nokia.com>:
45 * Smaps information related to shared, private, clean and dirty pages.
47 * Paul Mundt <paul.mundt@nokia.com>:
48 * Overall revision about smaps.
51 #include <linux/uaccess.h>
53 #include <linux/errno.h>
54 #include <linux/time.h>
55 #include <linux/proc_fs.h>
56 #include <linux/stat.h>
57 #include <linux/task_io_accounting_ops.h>
58 #include <linux/init.h>
59 #include <linux/capability.h>
60 #include <linux/file.h>
61 #include <linux/fdtable.h>
62 #include <linux/generic-radix-tree.h>
63 #include <linux/string.h>
64 #include <linux/seq_file.h>
65 #include <linux/namei.h>
66 #include <linux/mnt_namespace.h>
68 #include <linux/swap.h>
69 #include <linux/rcupdate.h>
70 #include <linux/kallsyms.h>
71 #include <linux/stacktrace.h>
72 #include <linux/resource.h>
73 #include <linux/module.h>
74 #include <linux/mount.h>
75 #include <linux/security.h>
76 #include <linux/ptrace.h>
77 #include <linux/tracehook.h>
78 #include <linux/printk.h>
79 #include <linux/cache.h>
80 #include <linux/cgroup.h>
81 #include <linux/cpuset.h>
82 #include <linux/audit.h>
83 #include <linux/poll.h>
84 #include <linux/nsproxy.h>
85 #include <linux/oom.h>
86 #include <linux/elf.h>
87 #include <linux/pid_namespace.h>
88 #include <linux/user_namespace.h>
89 #include <linux/fs_struct.h>
90 #include <linux/slab.h>
91 #include <linux/sched/autogroup.h>
92 #include <linux/sched/mm.h>
93 #include <linux/sched/coredump.h>
94 #include <linux/sched/debug.h>
95 #include <linux/sched/stat.h>
96 #include <linux/posix-timers.h>
97 #include <linux/time_namespace.h>
98 #include <linux/resctrl.h>
99 #include <trace/events/oom.h>
100 #include "internal.h"
103 #include "../../lib/kstrtox.h"
106 * Implementing inode permission operations in /proc is almost
107 * certainly an error. Permission checks need to happen during
108 * each system call not at open time. The reason is that most of
109 * what we wish to check for permissions in /proc varies at runtime.
111 * The classic example of a problem is opening file descriptors
112 * in /proc for a task before it execs a suid executable.
115 static u8 nlink_tid __ro_after_init;
116 static u8 nlink_tgid __ro_after_init;
122 const struct inode_operations *iop;
123 const struct file_operations *fop;
127 #define NOD(NAME, MODE, IOP, FOP, OP) { \
129 .len = sizeof(NAME) - 1, \
136 #define DIR(NAME, MODE, iops, fops) \
137 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
138 #define LNK(NAME, get_link) \
139 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
140 &proc_pid_link_inode_operations, NULL, \
141 { .proc_get_link = get_link } )
142 #define REG(NAME, MODE, fops) \
143 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
144 #define ONE(NAME, MODE, show) \
145 NOD(NAME, (S_IFREG|(MODE)), \
146 NULL, &proc_single_file_operations, \
147 { .proc_show = show } )
148 #define ATTR(LSM, NAME, MODE) \
149 NOD(NAME, (S_IFREG|(MODE)), \
150 NULL, &proc_pid_attr_operations, \
154 * Count the number of hardlinks for the pid_entry table, excluding the .
157 static unsigned int __init pid_entry_nlink(const struct pid_entry *entries,
164 for (i = 0; i < n; ++i) {
165 if (S_ISDIR(entries[i].mode))
172 static int get_task_root(struct task_struct *task, struct path *root)
174 int result = -ENOENT;
178 get_fs_root(task->fs, root);
185 static int proc_cwd_link(struct dentry *dentry, struct path *path)
187 struct task_struct *task = get_proc_task(d_inode(dentry));
188 int result = -ENOENT;
193 get_fs_pwd(task->fs, path);
197 put_task_struct(task);
202 static int proc_root_link(struct dentry *dentry, struct path *path)
204 struct task_struct *task = get_proc_task(d_inode(dentry));
205 int result = -ENOENT;
208 result = get_task_root(task, path);
209 put_task_struct(task);
215 * If the user used setproctitle(), we just get the string from
216 * user space at arg_start, and limit it to a maximum of one page.
218 static ssize_t get_mm_proctitle(struct mm_struct *mm, char __user *buf,
219 size_t count, unsigned long pos,
220 unsigned long arg_start)
225 if (pos >= PAGE_SIZE)
228 page = (char *)__get_free_page(GFP_KERNEL);
233 got = access_remote_vm(mm, arg_start, page, PAGE_SIZE, FOLL_ANON);
235 int len = strnlen(page, got);
237 /* Include the NUL character if it was found */
245 len -= copy_to_user(buf, page+pos, len);
251 free_page((unsigned long)page);
255 static ssize_t get_mm_cmdline(struct mm_struct *mm, char __user *buf,
256 size_t count, loff_t *ppos)
258 unsigned long arg_start, arg_end, env_start, env_end;
259 unsigned long pos, len;
262 /* Check if process spawned far enough to have cmdline. */
266 spin_lock(&mm->arg_lock);
267 arg_start = mm->arg_start;
268 arg_end = mm->arg_end;
269 env_start = mm->env_start;
270 env_end = mm->env_end;
271 spin_unlock(&mm->arg_lock);
273 if (arg_start >= arg_end)
277 * We allow setproctitle() to overwrite the argument
278 * strings, and overflow past the original end. But
279 * only when it overflows into the environment area.
281 if (env_start != arg_end || env_end < env_start)
282 env_start = env_end = arg_end;
283 len = env_end - arg_start;
285 /* We're not going to care if "*ppos" has high bits set */
289 if (count > len - pos)
295 * Magical special case: if the argv[] end byte is not
296 * zero, the user has overwritten it with setproctitle(3).
298 * Possible future enhancement: do this only once when
299 * pos is 0, and set a flag in the 'struct file'.
301 if (access_remote_vm(mm, arg_end-1, &c, 1, FOLL_ANON) == 1 && c)
302 return get_mm_proctitle(mm, buf, count, pos, arg_start);
305 * For the non-setproctitle() case we limit things strictly
306 * to the [arg_start, arg_end[ range.
309 if (pos < arg_start || pos >= arg_end)
311 if (count > arg_end - pos)
312 count = arg_end - pos;
314 page = (char *)__get_free_page(GFP_KERNEL);
321 size_t size = min_t(size_t, PAGE_SIZE, count);
323 got = access_remote_vm(mm, pos, page, size, FOLL_ANON);
326 got -= copy_to_user(buf, page, got);
327 if (unlikely(!got)) {
338 free_page((unsigned long)page);
342 static ssize_t get_task_cmdline(struct task_struct *tsk, char __user *buf,
343 size_t count, loff_t *pos)
345 struct mm_struct *mm;
348 mm = get_task_mm(tsk);
352 ret = get_mm_cmdline(mm, buf, count, pos);
357 static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
358 size_t count, loff_t *pos)
360 struct task_struct *tsk;
365 tsk = get_proc_task(file_inode(file));
368 ret = get_task_cmdline(tsk, buf, count, pos);
369 put_task_struct(tsk);
375 static const struct file_operations proc_pid_cmdline_ops = {
376 .read = proc_pid_cmdline_read,
377 .llseek = generic_file_llseek,
380 #ifdef CONFIG_KALLSYMS
382 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
383 * Returns the resolved symbol. If that fails, simply return the address.
385 static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
386 struct pid *pid, struct task_struct *task)
389 char symname[KSYM_NAME_LEN];
391 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
394 wchan = get_wchan(task);
395 if (wchan && !lookup_symbol_name(wchan, symname)) {
396 seq_puts(m, symname);
404 #endif /* CONFIG_KALLSYMS */
406 static int lock_trace(struct task_struct *task)
408 int err = mutex_lock_killable(&task->signal->exec_update_mutex);
411 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) {
412 mutex_unlock(&task->signal->exec_update_mutex);
418 static void unlock_trace(struct task_struct *task)
420 mutex_unlock(&task->signal->exec_update_mutex);
423 #ifdef CONFIG_STACKTRACE
425 #define MAX_STACK_TRACE_DEPTH 64
427 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
428 struct pid *pid, struct task_struct *task)
430 unsigned long *entries;
434 * The ability to racily run the kernel stack unwinder on a running task
435 * and then observe the unwinder output is scary; while it is useful for
436 * debugging kernel issues, it can also allow an attacker to leak kernel
438 * Doing this in a manner that is at least safe from races would require
439 * some work to ensure that the remote task can not be scheduled; and
440 * even then, this would still expose the unwinder as local attack
442 * Therefore, this interface is restricted to root.
444 if (!file_ns_capable(m->file, &init_user_ns, CAP_SYS_ADMIN))
447 entries = kmalloc_array(MAX_STACK_TRACE_DEPTH, sizeof(*entries),
452 err = lock_trace(task);
454 unsigned int i, nr_entries;
456 nr_entries = stack_trace_save_tsk(task, entries,
457 MAX_STACK_TRACE_DEPTH, 0);
459 for (i = 0; i < nr_entries; i++) {
460 seq_printf(m, "[<0>] %pB\n", (void *)entries[i]);
471 #ifdef CONFIG_SCHED_INFO
473 * Provides /proc/PID/schedstat
475 static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
476 struct pid *pid, struct task_struct *task)
478 if (unlikely(!sched_info_on()))
479 seq_puts(m, "0 0 0\n");
481 seq_printf(m, "%llu %llu %lu\n",
482 (unsigned long long)task->se.sum_exec_runtime,
483 (unsigned long long)task->sched_info.run_delay,
484 task->sched_info.pcount);
490 #ifdef CONFIG_LATENCYTOP
491 static int lstats_show_proc(struct seq_file *m, void *v)
494 struct inode *inode = m->private;
495 struct task_struct *task = get_proc_task(inode);
499 seq_puts(m, "Latency Top version : v0.1\n");
500 for (i = 0; i < LT_SAVECOUNT; i++) {
501 struct latency_record *lr = &task->latency_record[i];
502 if (lr->backtrace[0]) {
504 seq_printf(m, "%i %li %li",
505 lr->count, lr->time, lr->max);
506 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
507 unsigned long bt = lr->backtrace[q];
511 seq_printf(m, " %ps", (void *)bt);
517 put_task_struct(task);
521 static int lstats_open(struct inode *inode, struct file *file)
523 return single_open(file, lstats_show_proc, inode);
526 static ssize_t lstats_write(struct file *file, const char __user *buf,
527 size_t count, loff_t *offs)
529 struct task_struct *task = get_proc_task(file_inode(file));
533 clear_tsk_latency_tracing(task);
534 put_task_struct(task);
539 static const struct file_operations proc_lstats_operations = {
542 .write = lstats_write,
544 .release = single_release,
549 static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
550 struct pid *pid, struct task_struct *task)
552 unsigned long totalpages = totalram_pages() + total_swap_pages;
553 unsigned long points = 0;
555 points = oom_badness(task, totalpages) * 1000 / totalpages;
556 seq_printf(m, "%lu\n", points);
566 static const struct limit_names lnames[RLIM_NLIMITS] = {
567 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
568 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
569 [RLIMIT_DATA] = {"Max data size", "bytes"},
570 [RLIMIT_STACK] = {"Max stack size", "bytes"},
571 [RLIMIT_CORE] = {"Max core file size", "bytes"},
572 [RLIMIT_RSS] = {"Max resident set", "bytes"},
573 [RLIMIT_NPROC] = {"Max processes", "processes"},
574 [RLIMIT_NOFILE] = {"Max open files", "files"},
575 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
576 [RLIMIT_AS] = {"Max address space", "bytes"},
577 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
578 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
579 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
580 [RLIMIT_NICE] = {"Max nice priority", NULL},
581 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
582 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
585 /* Display limits for a process */
586 static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
587 struct pid *pid, struct task_struct *task)
592 struct rlimit rlim[RLIM_NLIMITS];
594 if (!lock_task_sighand(task, &flags))
596 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
597 unlock_task_sighand(task, &flags);
600 * print the file header
607 for (i = 0; i < RLIM_NLIMITS; i++) {
608 if (rlim[i].rlim_cur == RLIM_INFINITY)
609 seq_printf(m, "%-25s %-20s ",
610 lnames[i].name, "unlimited");
612 seq_printf(m, "%-25s %-20lu ",
613 lnames[i].name, rlim[i].rlim_cur);
615 if (rlim[i].rlim_max == RLIM_INFINITY)
616 seq_printf(m, "%-20s ", "unlimited");
618 seq_printf(m, "%-20lu ", rlim[i].rlim_max);
621 seq_printf(m, "%-10s\n", lnames[i].unit);
629 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
630 static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
631 struct pid *pid, struct task_struct *task)
633 struct syscall_info info;
634 u64 *args = &info.data.args[0];
637 res = lock_trace(task);
641 if (task_current_syscall(task, &info))
642 seq_puts(m, "running\n");
643 else if (info.data.nr < 0)
644 seq_printf(m, "%d 0x%llx 0x%llx\n",
645 info.data.nr, info.sp, info.data.instruction_pointer);
648 "%d 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx\n",
650 args[0], args[1], args[2], args[3], args[4], args[5],
651 info.sp, info.data.instruction_pointer);
656 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
658 /************************************************************************/
659 /* Here the fs part begins */
660 /************************************************************************/
662 /* permission checks */
663 static int proc_fd_access_allowed(struct inode *inode)
665 struct task_struct *task;
667 /* Allow access to a task's file descriptors if it is us or we
668 * may use ptrace attach to the process and find out that
671 task = get_proc_task(inode);
673 allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
674 put_task_struct(task);
679 int proc_setattr(struct dentry *dentry, struct iattr *attr)
682 struct inode *inode = d_inode(dentry);
684 if (attr->ia_valid & ATTR_MODE)
687 error = setattr_prepare(dentry, attr);
691 setattr_copy(inode, attr);
692 mark_inode_dirty(inode);
697 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
698 * or euid/egid (for hide_pid_min=2)?
700 static bool has_pid_permissions(struct proc_fs_info *fs_info,
701 struct task_struct *task,
702 enum proc_hidepid hide_pid_min)
705 * If 'hidpid' mount option is set force a ptrace check,
706 * we indicate that we are using a filesystem syscall
707 * by passing PTRACE_MODE_READ_FSCREDS
709 if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE)
710 return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
712 if (fs_info->hide_pid < hide_pid_min)
714 if (in_group_p(fs_info->pid_gid))
716 return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
720 static int proc_pid_permission(struct inode *inode, int mask)
722 struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
723 struct task_struct *task;
726 task = get_proc_task(inode);
729 has_perms = has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS);
730 put_task_struct(task);
733 if (fs_info->hide_pid == HIDEPID_INVISIBLE) {
735 * Let's make getdents(), stat(), and open()
736 * consistent with each other. If a process
737 * may not stat() a file, it shouldn't be seen
745 return generic_permission(inode, mask);
750 static const struct inode_operations proc_def_inode_operations = {
751 .setattr = proc_setattr,
754 static int proc_single_show(struct seq_file *m, void *v)
756 struct inode *inode = m->private;
757 struct pid_namespace *ns = proc_pid_ns(inode);
758 struct pid *pid = proc_pid(inode);
759 struct task_struct *task;
762 task = get_pid_task(pid, PIDTYPE_PID);
766 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
768 put_task_struct(task);
772 static int proc_single_open(struct inode *inode, struct file *filp)
774 return single_open(filp, proc_single_show, inode);
777 static const struct file_operations proc_single_file_operations = {
778 .open = proc_single_open,
781 .release = single_release,
785 struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
787 struct task_struct *task = get_proc_task(inode);
788 struct mm_struct *mm = ERR_PTR(-ESRCH);
791 mm = mm_access(task, mode | PTRACE_MODE_FSCREDS);
792 put_task_struct(task);
794 if (!IS_ERR_OR_NULL(mm)) {
795 /* ensure this mm_struct can't be freed */
797 /* but do not pin its memory */
805 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
807 struct mm_struct *mm = proc_mem_open(inode, mode);
812 file->private_data = mm;
816 static int mem_open(struct inode *inode, struct file *file)
818 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
820 /* OK to pass negative loff_t, we can catch out-of-range */
821 file->f_mode |= FMODE_UNSIGNED_OFFSET;
826 static ssize_t mem_rw(struct file *file, char __user *buf,
827 size_t count, loff_t *ppos, int write)
829 struct mm_struct *mm = file->private_data;
830 unsigned long addr = *ppos;
838 page = (char *)__get_free_page(GFP_KERNEL);
843 if (!mmget_not_zero(mm))
846 flags = FOLL_FORCE | (write ? FOLL_WRITE : 0);
849 int this_len = min_t(int, count, PAGE_SIZE);
851 if (write && copy_from_user(page, buf, this_len)) {
856 this_len = access_remote_vm(mm, addr, page, this_len, flags);
863 if (!write && copy_to_user(buf, page, this_len)) {
877 free_page((unsigned long) page);
881 static ssize_t mem_read(struct file *file, char __user *buf,
882 size_t count, loff_t *ppos)
884 return mem_rw(file, buf, count, ppos, 0);
887 static ssize_t mem_write(struct file *file, const char __user *buf,
888 size_t count, loff_t *ppos)
890 return mem_rw(file, (char __user*)buf, count, ppos, 1);
893 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
897 file->f_pos = offset;
900 file->f_pos += offset;
905 force_successful_syscall_return();
909 static int mem_release(struct inode *inode, struct file *file)
911 struct mm_struct *mm = file->private_data;
917 static const struct file_operations proc_mem_operations = {
922 .release = mem_release,
925 static int environ_open(struct inode *inode, struct file *file)
927 return __mem_open(inode, file, PTRACE_MODE_READ);
930 static ssize_t environ_read(struct file *file, char __user *buf,
931 size_t count, loff_t *ppos)
934 unsigned long src = *ppos;
936 struct mm_struct *mm = file->private_data;
937 unsigned long env_start, env_end;
939 /* Ensure the process spawned far enough to have an environment. */
940 if (!mm || !mm->env_end)
943 page = (char *)__get_free_page(GFP_KERNEL);
948 if (!mmget_not_zero(mm))
951 spin_lock(&mm->arg_lock);
952 env_start = mm->env_start;
953 env_end = mm->env_end;
954 spin_unlock(&mm->arg_lock);
957 size_t this_len, max_len;
960 if (src >= (env_end - env_start))
963 this_len = env_end - (env_start + src);
965 max_len = min_t(size_t, PAGE_SIZE, count);
966 this_len = min(max_len, this_len);
968 retval = access_remote_vm(mm, (env_start + src), page, this_len, FOLL_ANON);
975 if (copy_to_user(buf, page, retval)) {
989 free_page((unsigned long) page);
993 static const struct file_operations proc_environ_operations = {
994 .open = environ_open,
995 .read = environ_read,
996 .llseek = generic_file_llseek,
997 .release = mem_release,
1000 static int auxv_open(struct inode *inode, struct file *file)
1002 return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
1005 static ssize_t auxv_read(struct file *file, char __user *buf,
1006 size_t count, loff_t *ppos)
1008 struct mm_struct *mm = file->private_data;
1009 unsigned int nwords = 0;
1015 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
1016 return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv,
1017 nwords * sizeof(mm->saved_auxv[0]));
1020 static const struct file_operations proc_auxv_operations = {
1023 .llseek = generic_file_llseek,
1024 .release = mem_release,
1027 static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1030 struct task_struct *task = get_proc_task(file_inode(file));
1031 char buffer[PROC_NUMBUF];
1032 int oom_adj = OOM_ADJUST_MIN;
1037 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1038 oom_adj = OOM_ADJUST_MAX;
1040 oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1042 put_task_struct(task);
1043 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1044 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1047 static int __set_oom_adj(struct file *file, int oom_adj, bool legacy)
1049 static DEFINE_MUTEX(oom_adj_mutex);
1050 struct mm_struct *mm = NULL;
1051 struct task_struct *task;
1054 task = get_proc_task(file_inode(file));
1058 mutex_lock(&oom_adj_mutex);
1060 if (oom_adj < task->signal->oom_score_adj &&
1061 !capable(CAP_SYS_RESOURCE)) {
1066 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1067 * /proc/pid/oom_score_adj instead.
1069 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1070 current->comm, task_pid_nr(current), task_pid_nr(task),
1073 if ((short)oom_adj < task->signal->oom_score_adj_min &&
1074 !capable(CAP_SYS_RESOURCE)) {
1081 * Make sure we will check other processes sharing the mm if this is
1082 * not vfrok which wants its own oom_score_adj.
1083 * pin the mm so it doesn't go away and get reused after task_unlock
1085 if (!task->vfork_done) {
1086 struct task_struct *p = find_lock_task_mm(task);
1089 if (atomic_read(&p->mm->mm_users) > 1) {
1097 task->signal->oom_score_adj = oom_adj;
1098 if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1099 task->signal->oom_score_adj_min = (short)oom_adj;
1100 trace_oom_score_adj_update(task);
1103 struct task_struct *p;
1106 for_each_process(p) {
1107 if (same_thread_group(task, p))
1110 /* do not touch kernel threads or the global init */
1111 if (p->flags & PF_KTHREAD || is_global_init(p))
1115 if (!p->vfork_done && process_shares_mm(p, mm)) {
1116 p->signal->oom_score_adj = oom_adj;
1117 if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1118 p->signal->oom_score_adj_min = (short)oom_adj;
1126 mutex_unlock(&oom_adj_mutex);
1127 put_task_struct(task);
1132 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1133 * kernels. The effective policy is defined by oom_score_adj, which has a
1134 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1135 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1136 * Processes that become oom disabled via oom_adj will still be oom disabled
1137 * with this implementation.
1139 * oom_adj cannot be removed since existing userspace binaries use it.
1141 static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1142 size_t count, loff_t *ppos)
1144 char buffer[PROC_NUMBUF];
1148 memset(buffer, 0, sizeof(buffer));
1149 if (count > sizeof(buffer) - 1)
1150 count = sizeof(buffer) - 1;
1151 if (copy_from_user(buffer, buf, count)) {
1156 err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1159 if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1160 oom_adj != OOM_DISABLE) {
1166 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1167 * value is always attainable.
1169 if (oom_adj == OOM_ADJUST_MAX)
1170 oom_adj = OOM_SCORE_ADJ_MAX;
1172 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1174 err = __set_oom_adj(file, oom_adj, true);
1176 return err < 0 ? err : count;
1179 static const struct file_operations proc_oom_adj_operations = {
1180 .read = oom_adj_read,
1181 .write = oom_adj_write,
1182 .llseek = generic_file_llseek,
1185 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1186 size_t count, loff_t *ppos)
1188 struct task_struct *task = get_proc_task(file_inode(file));
1189 char buffer[PROC_NUMBUF];
1190 short oom_score_adj = OOM_SCORE_ADJ_MIN;
1195 oom_score_adj = task->signal->oom_score_adj;
1196 put_task_struct(task);
1197 len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
1198 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1201 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1202 size_t count, loff_t *ppos)
1204 char buffer[PROC_NUMBUF];
1208 memset(buffer, 0, sizeof(buffer));
1209 if (count > sizeof(buffer) - 1)
1210 count = sizeof(buffer) - 1;
1211 if (copy_from_user(buffer, buf, count)) {
1216 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1219 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1220 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1225 err = __set_oom_adj(file, oom_score_adj, false);
1227 return err < 0 ? err : count;
1230 static const struct file_operations proc_oom_score_adj_operations = {
1231 .read = oom_score_adj_read,
1232 .write = oom_score_adj_write,
1233 .llseek = default_llseek,
1237 #define TMPBUFLEN 11
1238 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1239 size_t count, loff_t *ppos)
1241 struct inode * inode = file_inode(file);
1242 struct task_struct *task = get_proc_task(inode);
1244 char tmpbuf[TMPBUFLEN];
1248 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1249 from_kuid(file->f_cred->user_ns,
1250 audit_get_loginuid(task)));
1251 put_task_struct(task);
1252 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1255 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1256 size_t count, loff_t *ppos)
1258 struct inode * inode = file_inode(file);
1264 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1271 /* No partial writes. */
1275 rv = kstrtou32_from_user(buf, count, 10, &loginuid);
1279 /* is userspace tring to explicitly UNSET the loginuid? */
1280 if (loginuid == AUDIT_UID_UNSET) {
1281 kloginuid = INVALID_UID;
1283 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1284 if (!uid_valid(kloginuid))
1288 rv = audit_set_loginuid(kloginuid);
1294 static const struct file_operations proc_loginuid_operations = {
1295 .read = proc_loginuid_read,
1296 .write = proc_loginuid_write,
1297 .llseek = generic_file_llseek,
1300 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1301 size_t count, loff_t *ppos)
1303 struct inode * inode = file_inode(file);
1304 struct task_struct *task = get_proc_task(inode);
1306 char tmpbuf[TMPBUFLEN];
1310 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1311 audit_get_sessionid(task));
1312 put_task_struct(task);
1313 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1316 static const struct file_operations proc_sessionid_operations = {
1317 .read = proc_sessionid_read,
1318 .llseek = generic_file_llseek,
1322 #ifdef CONFIG_FAULT_INJECTION
1323 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1324 size_t count, loff_t *ppos)
1326 struct task_struct *task = get_proc_task(file_inode(file));
1327 char buffer[PROC_NUMBUF];
1333 make_it_fail = task->make_it_fail;
1334 put_task_struct(task);
1336 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1338 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1341 static ssize_t proc_fault_inject_write(struct file * file,
1342 const char __user * buf, size_t count, loff_t *ppos)
1344 struct task_struct *task;
1345 char buffer[PROC_NUMBUF];
1349 if (!capable(CAP_SYS_RESOURCE))
1351 memset(buffer, 0, sizeof(buffer));
1352 if (count > sizeof(buffer) - 1)
1353 count = sizeof(buffer) - 1;
1354 if (copy_from_user(buffer, buf, count))
1356 rv = kstrtoint(strstrip(buffer), 0, &make_it_fail);
1359 if (make_it_fail < 0 || make_it_fail > 1)
1362 task = get_proc_task(file_inode(file));
1365 task->make_it_fail = make_it_fail;
1366 put_task_struct(task);
1371 static const struct file_operations proc_fault_inject_operations = {
1372 .read = proc_fault_inject_read,
1373 .write = proc_fault_inject_write,
1374 .llseek = generic_file_llseek,
1377 static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf,
1378 size_t count, loff_t *ppos)
1380 struct task_struct *task;
1384 err = kstrtouint_from_user(buf, count, 0, &n);
1388 task = get_proc_task(file_inode(file));
1392 put_task_struct(task);
1397 static ssize_t proc_fail_nth_read(struct file *file, char __user *buf,
1398 size_t count, loff_t *ppos)
1400 struct task_struct *task;
1401 char numbuf[PROC_NUMBUF];
1404 task = get_proc_task(file_inode(file));
1407 len = snprintf(numbuf, sizeof(numbuf), "%u\n", task->fail_nth);
1408 put_task_struct(task);
1409 return simple_read_from_buffer(buf, count, ppos, numbuf, len);
1412 static const struct file_operations proc_fail_nth_operations = {
1413 .read = proc_fail_nth_read,
1414 .write = proc_fail_nth_write,
1419 #ifdef CONFIG_SCHED_DEBUG
1421 * Print out various scheduling related per-task fields:
1423 static int sched_show(struct seq_file *m, void *v)
1425 struct inode *inode = m->private;
1426 struct pid_namespace *ns = proc_pid_ns(inode);
1427 struct task_struct *p;
1429 p = get_proc_task(inode);
1432 proc_sched_show_task(p, ns, m);
1440 sched_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;
1446 p = get_proc_task(inode);
1449 proc_sched_set_task(p);
1456 static int sched_open(struct inode *inode, struct file *filp)
1458 return single_open(filp, sched_show, inode);
1461 static const struct file_operations proc_pid_sched_operations = {
1464 .write = sched_write,
1465 .llseek = seq_lseek,
1466 .release = single_release,
1471 #ifdef CONFIG_SCHED_AUTOGROUP
1473 * Print out autogroup related information:
1475 static int sched_autogroup_show(struct seq_file *m, void *v)
1477 struct inode *inode = m->private;
1478 struct task_struct *p;
1480 p = get_proc_task(inode);
1483 proc_sched_autogroup_show_task(p, m);
1491 sched_autogroup_write(struct file *file, const char __user *buf,
1492 size_t count, loff_t *offset)
1494 struct inode *inode = file_inode(file);
1495 struct task_struct *p;
1496 char buffer[PROC_NUMBUF];
1500 memset(buffer, 0, sizeof(buffer));
1501 if (count > sizeof(buffer) - 1)
1502 count = sizeof(buffer) - 1;
1503 if (copy_from_user(buffer, buf, count))
1506 err = kstrtoint(strstrip(buffer), 0, &nice);
1510 p = get_proc_task(inode);
1514 err = proc_sched_autogroup_set_nice(p, nice);
1523 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1527 ret = single_open(filp, sched_autogroup_show, NULL);
1529 struct seq_file *m = filp->private_data;
1536 static const struct file_operations proc_pid_sched_autogroup_operations = {
1537 .open = sched_autogroup_open,
1539 .write = sched_autogroup_write,
1540 .llseek = seq_lseek,
1541 .release = single_release,
1544 #endif /* CONFIG_SCHED_AUTOGROUP */
1546 #ifdef CONFIG_TIME_NS
1547 static int timens_offsets_show(struct seq_file *m, void *v)
1549 struct task_struct *p;
1551 p = get_proc_task(file_inode(m->file));
1554 proc_timens_show_offsets(p, m);
1561 static ssize_t timens_offsets_write(struct file *file, const char __user *buf,
1562 size_t count, loff_t *ppos)
1564 struct inode *inode = file_inode(file);
1565 struct proc_timens_offset offsets[2];
1566 char *kbuf = NULL, *pos, *next_line;
1567 struct task_struct *p;
1570 /* Only allow < page size writes at the beginning of the file */
1571 if ((*ppos != 0) || (count >= PAGE_SIZE))
1574 /* Slurp in the user data */
1575 kbuf = memdup_user_nul(buf, count);
1577 return PTR_ERR(kbuf);
1579 /* Parse the user data */
1582 for (pos = kbuf; pos; pos = next_line) {
1583 struct proc_timens_offset *off = &offsets[noffsets];
1586 /* Find the end of line and ensure we don't look past it */
1587 next_line = strchr(pos, '\n');
1591 if (*next_line == '\0')
1595 err = sscanf(pos, "%u %lld %lu", &off->clockid,
1596 &off->val.tv_sec, &off->val.tv_nsec);
1597 if (err != 3 || off->val.tv_nsec >= NSEC_PER_SEC)
1600 if (noffsets == ARRAY_SIZE(offsets)) {
1602 count = next_line - kbuf;
1608 p = get_proc_task(inode);
1611 ret = proc_timens_set_offset(file, p, offsets, noffsets);
1622 static int timens_offsets_open(struct inode *inode, struct file *filp)
1624 return single_open(filp, timens_offsets_show, inode);
1627 static const struct file_operations proc_timens_offsets_operations = {
1628 .open = timens_offsets_open,
1630 .write = timens_offsets_write,
1631 .llseek = seq_lseek,
1632 .release = single_release,
1634 #endif /* CONFIG_TIME_NS */
1636 static ssize_t comm_write(struct file *file, const char __user *buf,
1637 size_t count, loff_t *offset)
1639 struct inode *inode = file_inode(file);
1640 struct task_struct *p;
1641 char buffer[TASK_COMM_LEN];
1642 const size_t maxlen = sizeof(buffer) - 1;
1644 memset(buffer, 0, sizeof(buffer));
1645 if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
1648 p = get_proc_task(inode);
1652 if (same_thread_group(current, p))
1653 set_task_comm(p, buffer);
1662 static int comm_show(struct seq_file *m, void *v)
1664 struct inode *inode = m->private;
1665 struct task_struct *p;
1667 p = get_proc_task(inode);
1671 proc_task_name(m, p, false);
1679 static int comm_open(struct inode *inode, struct file *filp)
1681 return single_open(filp, comm_show, inode);
1684 static const struct file_operations proc_pid_set_comm_operations = {
1687 .write = comm_write,
1688 .llseek = seq_lseek,
1689 .release = single_release,
1692 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1694 struct task_struct *task;
1695 struct file *exe_file;
1697 task = get_proc_task(d_inode(dentry));
1700 exe_file = get_task_exe_file(task);
1701 put_task_struct(task);
1703 *exe_path = exe_file->f_path;
1704 path_get(&exe_file->f_path);
1711 static const char *proc_pid_get_link(struct dentry *dentry,
1712 struct inode *inode,
1713 struct delayed_call *done)
1716 int error = -EACCES;
1719 return ERR_PTR(-ECHILD);
1721 /* Are we allowed to snoop on the tasks file descriptors? */
1722 if (!proc_fd_access_allowed(inode))
1725 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1729 error = nd_jump_link(&path);
1731 return ERR_PTR(error);
1734 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1736 char *tmp = (char *)__get_free_page(GFP_KERNEL);
1743 pathname = d_path(path, tmp, PAGE_SIZE);
1744 len = PTR_ERR(pathname);
1745 if (IS_ERR(pathname))
1747 len = tmp + PAGE_SIZE - 1 - pathname;
1751 if (copy_to_user(buffer, pathname, len))
1754 free_page((unsigned long)tmp);
1758 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1760 int error = -EACCES;
1761 struct inode *inode = d_inode(dentry);
1764 /* Are we allowed to snoop on the tasks file descriptors? */
1765 if (!proc_fd_access_allowed(inode))
1768 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1772 error = do_proc_readlink(&path, buffer, buflen);
1778 const struct inode_operations proc_pid_link_inode_operations = {
1779 .readlink = proc_pid_readlink,
1780 .get_link = proc_pid_get_link,
1781 .setattr = proc_setattr,
1785 /* building an inode */
1787 void task_dump_owner(struct task_struct *task, umode_t mode,
1788 kuid_t *ruid, kgid_t *rgid)
1790 /* Depending on the state of dumpable compute who should own a
1791 * proc file for a task.
1793 const struct cred *cred;
1797 if (unlikely(task->flags & PF_KTHREAD)) {
1798 *ruid = GLOBAL_ROOT_UID;
1799 *rgid = GLOBAL_ROOT_GID;
1803 /* Default to the tasks effective ownership */
1805 cred = __task_cred(task);
1811 * Before the /proc/pid/status file was created the only way to read
1812 * the effective uid of a /process was to stat /proc/pid. Reading
1813 * /proc/pid/status is slow enough that procps and other packages
1814 * kept stating /proc/pid. To keep the rules in /proc simple I have
1815 * made this apply to all per process world readable and executable
1818 if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) {
1819 struct mm_struct *mm;
1822 /* Make non-dumpable tasks owned by some root */
1824 if (get_dumpable(mm) != SUID_DUMP_USER) {
1825 struct user_namespace *user_ns = mm->user_ns;
1827 uid = make_kuid(user_ns, 0);
1828 if (!uid_valid(uid))
1829 uid = GLOBAL_ROOT_UID;
1831 gid = make_kgid(user_ns, 0);
1832 if (!gid_valid(gid))
1833 gid = GLOBAL_ROOT_GID;
1836 uid = GLOBAL_ROOT_UID;
1837 gid = GLOBAL_ROOT_GID;
1845 void proc_pid_evict_inode(struct proc_inode *ei)
1847 struct pid *pid = ei->pid;
1849 if (S_ISDIR(ei->vfs_inode.i_mode)) {
1850 spin_lock(&pid->lock);
1851 hlist_del_init_rcu(&ei->sibling_inodes);
1852 spin_unlock(&pid->lock);
1858 struct inode *proc_pid_make_inode(struct super_block * sb,
1859 struct task_struct *task, umode_t mode)
1861 struct inode * inode;
1862 struct proc_inode *ei;
1865 /* We need a new inode */
1867 inode = new_inode(sb);
1873 inode->i_mode = mode;
1874 inode->i_ino = get_next_ino();
1875 inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode);
1876 inode->i_op = &proc_def_inode_operations;
1879 * grab the reference to task.
1881 pid = get_task_pid(task, PIDTYPE_PID);
1885 /* Let the pid remember us for quick removal */
1887 if (S_ISDIR(mode)) {
1888 spin_lock(&pid->lock);
1889 hlist_add_head_rcu(&ei->sibling_inodes, &pid->inodes);
1890 spin_unlock(&pid->lock);
1893 task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
1894 security_task_to_inode(task, inode);
1904 int pid_getattr(const struct path *path, struct kstat *stat,
1905 u32 request_mask, unsigned int query_flags)
1907 struct inode *inode = d_inode(path->dentry);
1908 struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
1909 struct task_struct *task;
1911 generic_fillattr(inode, stat);
1913 stat->uid = GLOBAL_ROOT_UID;
1914 stat->gid = GLOBAL_ROOT_GID;
1916 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1918 if (!has_pid_permissions(fs_info, task, HIDEPID_INVISIBLE)) {
1921 * This doesn't prevent learning whether PID exists,
1922 * it only makes getattr() consistent with readdir().
1926 task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid);
1935 * Set <pid>/... inode ownership (can change due to setuid(), etc.)
1937 void pid_update_inode(struct task_struct *task, struct inode *inode)
1939 task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid);
1941 inode->i_mode &= ~(S_ISUID | S_ISGID);
1942 security_task_to_inode(task, inode);
1946 * Rewrite the inode's ownerships here because the owning task may have
1947 * performed a setuid(), etc.
1950 static int pid_revalidate(struct dentry *dentry, unsigned int flags)
1952 struct inode *inode;
1953 struct task_struct *task;
1955 if (flags & LOOKUP_RCU)
1958 inode = d_inode(dentry);
1959 task = get_proc_task(inode);
1962 pid_update_inode(task, inode);
1963 put_task_struct(task);
1969 static inline bool proc_inode_is_dead(struct inode *inode)
1971 return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1974 int pid_delete_dentry(const struct dentry *dentry)
1976 /* Is the task we represent dead?
1977 * If so, then don't put the dentry on the lru list,
1978 * kill it immediately.
1980 return proc_inode_is_dead(d_inode(dentry));
1983 const struct dentry_operations pid_dentry_operations =
1985 .d_revalidate = pid_revalidate,
1986 .d_delete = pid_delete_dentry,
1992 * Fill a directory entry.
1994 * If possible create the dcache entry and derive our inode number and
1995 * file type from dcache entry.
1997 * Since all of the proc inode numbers are dynamically generated, the inode
1998 * numbers do not exist until the inode is cache. This means creating the
1999 * the dcache entry in readdir is necessary to keep the inode numbers
2000 * reported by readdir in sync with the inode numbers reported
2003 bool proc_fill_cache(struct file *file, struct dir_context *ctx,
2004 const char *name, unsigned int len,
2005 instantiate_t instantiate, struct task_struct *task, const void *ptr)
2007 struct dentry *child, *dir = file->f_path.dentry;
2008 struct qstr qname = QSTR_INIT(name, len);
2009 struct inode *inode;
2010 unsigned type = DT_UNKNOWN;
2013 child = d_hash_and_lookup(dir, &qname);
2015 DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq);
2016 child = d_alloc_parallel(dir, &qname, &wq);
2018 goto end_instantiate;
2019 if (d_in_lookup(child)) {
2021 res = instantiate(child, task, ptr);
2022 d_lookup_done(child);
2023 if (unlikely(res)) {
2027 goto end_instantiate;
2031 inode = d_inode(child);
2033 type = inode->i_mode >> 12;
2036 return dir_emit(ctx, name, len, ino, type);
2040 * dname_to_vma_addr - maps a dentry name into two unsigned longs
2041 * which represent vma start and end addresses.
2043 static int dname_to_vma_addr(struct dentry *dentry,
2044 unsigned long *start, unsigned long *end)
2046 const char *str = dentry->d_name.name;
2047 unsigned long long sval, eval;
2050 if (str[0] == '0' && str[1] != '-')
2052 len = _parse_integer(str, 16, &sval);
2053 if (len & KSTRTOX_OVERFLOW)
2055 if (sval != (unsigned long)sval)
2063 if (str[0] == '0' && str[1])
2065 len = _parse_integer(str, 16, &eval);
2066 if (len & KSTRTOX_OVERFLOW)
2068 if (eval != (unsigned long)eval)
2081 static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
2083 unsigned long vm_start, vm_end;
2084 bool exact_vma_exists = false;
2085 struct mm_struct *mm = NULL;
2086 struct task_struct *task;
2087 struct inode *inode;
2090 if (flags & LOOKUP_RCU)
2093 inode = d_inode(dentry);
2094 task = get_proc_task(inode);
2098 mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
2099 if (IS_ERR_OR_NULL(mm))
2102 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
2103 status = down_read_killable(&mm->mmap_sem);
2105 exact_vma_exists = !!find_exact_vma(mm, vm_start,
2107 up_read(&mm->mmap_sem);
2113 if (exact_vma_exists) {
2114 task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
2116 security_task_to_inode(task, inode);
2121 put_task_struct(task);
2127 static const struct dentry_operations tid_map_files_dentry_operations = {
2128 .d_revalidate = map_files_d_revalidate,
2129 .d_delete = pid_delete_dentry,
2132 static int map_files_get_link(struct dentry *dentry, struct path *path)
2134 unsigned long vm_start, vm_end;
2135 struct vm_area_struct *vma;
2136 struct task_struct *task;
2137 struct mm_struct *mm;
2141 task = get_proc_task(d_inode(dentry));
2145 mm = get_task_mm(task);
2146 put_task_struct(task);
2150 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2154 rc = down_read_killable(&mm->mmap_sem);
2159 vma = find_exact_vma(mm, vm_start, vm_end);
2160 if (vma && vma->vm_file) {
2161 *path = vma->vm_file->f_path;
2165 up_read(&mm->mmap_sem);
2173 struct map_files_info {
2174 unsigned long start;
2180 * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the
2181 * symlinks may be used to bypass permissions on ancestor directories in the
2182 * path to the file in question.
2185 proc_map_files_get_link(struct dentry *dentry,
2186 struct inode *inode,
2187 struct delayed_call *done)
2189 if (!capable(CAP_SYS_ADMIN))
2190 return ERR_PTR(-EPERM);
2192 return proc_pid_get_link(dentry, inode, done);
2196 * Identical to proc_pid_link_inode_operations except for get_link()
2198 static const struct inode_operations proc_map_files_link_inode_operations = {
2199 .readlink = proc_pid_readlink,
2200 .get_link = proc_map_files_get_link,
2201 .setattr = proc_setattr,
2204 static struct dentry *
2205 proc_map_files_instantiate(struct dentry *dentry,
2206 struct task_struct *task, const void *ptr)
2208 fmode_t mode = (fmode_t)(unsigned long)ptr;
2209 struct proc_inode *ei;
2210 struct inode *inode;
2212 inode = proc_pid_make_inode(dentry->d_sb, task, S_IFLNK |
2213 ((mode & FMODE_READ ) ? S_IRUSR : 0) |
2214 ((mode & FMODE_WRITE) ? S_IWUSR : 0));
2216 return ERR_PTR(-ENOENT);
2219 ei->op.proc_get_link = map_files_get_link;
2221 inode->i_op = &proc_map_files_link_inode_operations;
2224 d_set_d_op(dentry, &tid_map_files_dentry_operations);
2225 return d_splice_alias(inode, dentry);
2228 static struct dentry *proc_map_files_lookup(struct inode *dir,
2229 struct dentry *dentry, unsigned int flags)
2231 unsigned long vm_start, vm_end;
2232 struct vm_area_struct *vma;
2233 struct task_struct *task;
2234 struct dentry *result;
2235 struct mm_struct *mm;
2237 result = ERR_PTR(-ENOENT);
2238 task = get_proc_task(dir);
2242 result = ERR_PTR(-EACCES);
2243 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2246 result = ERR_PTR(-ENOENT);
2247 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2250 mm = get_task_mm(task);
2254 result = ERR_PTR(-EINTR);
2255 if (down_read_killable(&mm->mmap_sem))
2258 result = ERR_PTR(-ENOENT);
2259 vma = find_exact_vma(mm, vm_start, vm_end);
2264 result = proc_map_files_instantiate(dentry, task,
2265 (void *)(unsigned long)vma->vm_file->f_mode);
2268 up_read(&mm->mmap_sem);
2272 put_task_struct(task);
2277 static const struct inode_operations proc_map_files_inode_operations = {
2278 .lookup = proc_map_files_lookup,
2279 .permission = proc_fd_permission,
2280 .setattr = proc_setattr,
2284 proc_map_files_readdir(struct file *file, struct dir_context *ctx)
2286 struct vm_area_struct *vma;
2287 struct task_struct *task;
2288 struct mm_struct *mm;
2289 unsigned long nr_files, pos, i;
2290 GENRADIX(struct map_files_info) fa;
2291 struct map_files_info *p;
2297 task = get_proc_task(file_inode(file));
2302 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2306 if (!dir_emit_dots(file, ctx))
2309 mm = get_task_mm(task);
2313 ret = down_read_killable(&mm->mmap_sem);
2322 * We need two passes here:
2324 * 1) Collect vmas of mapped files with mmap_sem taken
2325 * 2) Release mmap_sem and instantiate entries
2327 * otherwise we get lockdep complained, since filldir()
2328 * routine might require mmap_sem taken in might_fault().
2331 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2334 if (++pos <= ctx->pos)
2337 p = genradix_ptr_alloc(&fa, nr_files++, GFP_KERNEL);
2340 up_read(&mm->mmap_sem);
2345 p->start = vma->vm_start;
2346 p->end = vma->vm_end;
2347 p->mode = vma->vm_file->f_mode;
2349 up_read(&mm->mmap_sem);
2352 for (i = 0; i < nr_files; i++) {
2353 char buf[4 * sizeof(long) + 2]; /* max: %lx-%lx\0 */
2356 p = genradix_ptr(&fa, i);
2357 len = snprintf(buf, sizeof(buf), "%lx-%lx", p->start, p->end);
2358 if (!proc_fill_cache(file, ctx,
2360 proc_map_files_instantiate,
2362 (void *)(unsigned long)p->mode))
2368 put_task_struct(task);
2374 static const struct file_operations proc_map_files_operations = {
2375 .read = generic_read_dir,
2376 .iterate_shared = proc_map_files_readdir,
2377 .llseek = generic_file_llseek,
2380 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
2381 struct timers_private {
2383 struct task_struct *task;
2384 struct sighand_struct *sighand;
2385 struct pid_namespace *ns;
2386 unsigned long flags;
2389 static void *timers_start(struct seq_file *m, loff_t *pos)
2391 struct timers_private *tp = m->private;
2393 tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2395 return ERR_PTR(-ESRCH);
2397 tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2399 return ERR_PTR(-ESRCH);
2401 return seq_list_start(&tp->task->signal->posix_timers, *pos);
2404 static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2406 struct timers_private *tp = m->private;
2407 return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2410 static void timers_stop(struct seq_file *m, void *v)
2412 struct timers_private *tp = m->private;
2415 unlock_task_sighand(tp->task, &tp->flags);
2420 put_task_struct(tp->task);
2425 static int show_timer(struct seq_file *m, void *v)
2427 struct k_itimer *timer;
2428 struct timers_private *tp = m->private;
2430 static const char * const nstr[] = {
2431 [SIGEV_SIGNAL] = "signal",
2432 [SIGEV_NONE] = "none",
2433 [SIGEV_THREAD] = "thread",
2436 timer = list_entry((struct list_head *)v, struct k_itimer, list);
2437 notify = timer->it_sigev_notify;
2439 seq_printf(m, "ID: %d\n", timer->it_id);
2440 seq_printf(m, "signal: %d/%px\n",
2441 timer->sigq->info.si_signo,
2442 timer->sigq->info.si_value.sival_ptr);
2443 seq_printf(m, "notify: %s/%s.%d\n",
2444 nstr[notify & ~SIGEV_THREAD_ID],
2445 (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2446 pid_nr_ns(timer->it_pid, tp->ns));
2447 seq_printf(m, "ClockID: %d\n", timer->it_clock);
2452 static const struct seq_operations proc_timers_seq_ops = {
2453 .start = timers_start,
2454 .next = timers_next,
2455 .stop = timers_stop,
2459 static int proc_timers_open(struct inode *inode, struct file *file)
2461 struct timers_private *tp;
2463 tp = __seq_open_private(file, &proc_timers_seq_ops,
2464 sizeof(struct timers_private));
2468 tp->pid = proc_pid(inode);
2469 tp->ns = proc_pid_ns(inode);
2473 static const struct file_operations proc_timers_operations = {
2474 .open = proc_timers_open,
2476 .llseek = seq_lseek,
2477 .release = seq_release_private,
2481 static ssize_t timerslack_ns_write(struct file *file, const char __user *buf,
2482 size_t count, loff_t *offset)
2484 struct inode *inode = file_inode(file);
2485 struct task_struct *p;
2489 err = kstrtoull_from_user(buf, count, 10, &slack_ns);
2493 p = get_proc_task(inode);
2499 if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
2506 err = security_task_setscheduler(p);
2515 p->timer_slack_ns = p->default_timer_slack_ns;
2517 p->timer_slack_ns = slack_ns;
2526 static int timerslack_ns_show(struct seq_file *m, void *v)
2528 struct inode *inode = m->private;
2529 struct task_struct *p;
2532 p = get_proc_task(inode);
2538 if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
2545 err = security_task_getscheduler(p);
2551 seq_printf(m, "%llu\n", p->timer_slack_ns);
2560 static int timerslack_ns_open(struct inode *inode, struct file *filp)
2562 return single_open(filp, timerslack_ns_show, inode);
2565 static const struct file_operations proc_pid_set_timerslack_ns_operations = {
2566 .open = timerslack_ns_open,
2568 .write = timerslack_ns_write,
2569 .llseek = seq_lseek,
2570 .release = single_release,
2573 static struct dentry *proc_pident_instantiate(struct dentry *dentry,
2574 struct task_struct *task, const void *ptr)
2576 const struct pid_entry *p = ptr;
2577 struct inode *inode;
2578 struct proc_inode *ei;
2580 inode = proc_pid_make_inode(dentry->d_sb, task, p->mode);
2582 return ERR_PTR(-ENOENT);
2585 if (S_ISDIR(inode->i_mode))
2586 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2588 inode->i_op = p->iop;
2590 inode->i_fop = p->fop;
2592 pid_update_inode(task, inode);
2593 d_set_d_op(dentry, &pid_dentry_operations);
2594 return d_splice_alias(inode, dentry);
2597 static struct dentry *proc_pident_lookup(struct inode *dir,
2598 struct dentry *dentry,
2599 const struct pid_entry *p,
2600 const struct pid_entry *end)
2602 struct task_struct *task = get_proc_task(dir);
2603 struct dentry *res = ERR_PTR(-ENOENT);
2609 * Yes, it does not scale. And it should not. Don't add
2610 * new entries into /proc/<tgid>/ without very good reasons.
2612 for (; p < end; p++) {
2613 if (p->len != dentry->d_name.len)
2615 if (!memcmp(dentry->d_name.name, p->name, p->len)) {
2616 res = proc_pident_instantiate(dentry, task, p);
2620 put_task_struct(task);
2625 static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
2626 const struct pid_entry *ents, unsigned int nents)
2628 struct task_struct *task = get_proc_task(file_inode(file));
2629 const struct pid_entry *p;
2634 if (!dir_emit_dots(file, ctx))
2637 if (ctx->pos >= nents + 2)
2640 for (p = ents + (ctx->pos - 2); p < ents + nents; p++) {
2641 if (!proc_fill_cache(file, ctx, p->name, p->len,
2642 proc_pident_instantiate, task, p))
2647 put_task_struct(task);
2651 #ifdef CONFIG_SECURITY
2652 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2653 size_t count, loff_t *ppos)
2655 struct inode * inode = file_inode(file);
2658 struct task_struct *task = get_proc_task(inode);
2663 length = security_getprocattr(task, PROC_I(inode)->op.lsm,
2664 (char*)file->f_path.dentry->d_name.name,
2666 put_task_struct(task);
2668 length = simple_read_from_buffer(buf, count, ppos, p, length);
2673 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2674 size_t count, loff_t *ppos)
2676 struct inode * inode = file_inode(file);
2677 struct task_struct *task;
2682 task = pid_task(proc_pid(inode), PIDTYPE_PID);
2687 /* A task may only write its own attributes. */
2688 if (current != task) {
2692 /* Prevent changes to overridden credentials. */
2693 if (current_cred() != current_real_cred()) {
2699 if (count > PAGE_SIZE)
2702 /* No partial writes. */
2706 page = memdup_user(buf, count);
2712 /* Guard against adverse ptrace interaction */
2713 rv = mutex_lock_interruptible(¤t->signal->cred_guard_mutex);
2717 rv = security_setprocattr(PROC_I(inode)->op.lsm,
2718 file->f_path.dentry->d_name.name, page,
2720 mutex_unlock(¤t->signal->cred_guard_mutex);
2727 static const struct file_operations proc_pid_attr_operations = {
2728 .read = proc_pid_attr_read,
2729 .write = proc_pid_attr_write,
2730 .llseek = generic_file_llseek,
2733 #define LSM_DIR_OPS(LSM) \
2734 static int proc_##LSM##_attr_dir_iterate(struct file *filp, \
2735 struct dir_context *ctx) \
2737 return proc_pident_readdir(filp, ctx, \
2738 LSM##_attr_dir_stuff, \
2739 ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2742 static const struct file_operations proc_##LSM##_attr_dir_ops = { \
2743 .read = generic_read_dir, \
2744 .iterate = proc_##LSM##_attr_dir_iterate, \
2745 .llseek = default_llseek, \
2748 static struct dentry *proc_##LSM##_attr_dir_lookup(struct inode *dir, \
2749 struct dentry *dentry, unsigned int flags) \
2751 return proc_pident_lookup(dir, dentry, \
2752 LSM##_attr_dir_stuff, \
2753 LSM##_attr_dir_stuff + ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2756 static const struct inode_operations proc_##LSM##_attr_dir_inode_ops = { \
2757 .lookup = proc_##LSM##_attr_dir_lookup, \
2758 .getattr = pid_getattr, \
2759 .setattr = proc_setattr, \
2762 #ifdef CONFIG_SECURITY_SMACK
2763 static const struct pid_entry smack_attr_dir_stuff[] = {
2764 ATTR("smack", "current", 0666),
2769 static const struct pid_entry attr_dir_stuff[] = {
2770 ATTR(NULL, "current", 0666),
2771 ATTR(NULL, "prev", 0444),
2772 ATTR(NULL, "exec", 0666),
2773 ATTR(NULL, "fscreate", 0666),
2774 ATTR(NULL, "keycreate", 0666),
2775 ATTR(NULL, "sockcreate", 0666),
2776 #ifdef CONFIG_SECURITY_SMACK
2778 proc_smack_attr_dir_inode_ops, proc_smack_attr_dir_ops),
2782 static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
2784 return proc_pident_readdir(file, ctx,
2785 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2788 static const struct file_operations proc_attr_dir_operations = {
2789 .read = generic_read_dir,
2790 .iterate_shared = proc_attr_dir_readdir,
2791 .llseek = generic_file_llseek,
2794 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2795 struct dentry *dentry, unsigned int flags)
2797 return proc_pident_lookup(dir, dentry,
2799 attr_dir_stuff + ARRAY_SIZE(attr_dir_stuff));
2802 static const struct inode_operations proc_attr_dir_inode_operations = {
2803 .lookup = proc_attr_dir_lookup,
2804 .getattr = pid_getattr,
2805 .setattr = proc_setattr,
2810 #ifdef CONFIG_ELF_CORE
2811 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2812 size_t count, loff_t *ppos)
2814 struct task_struct *task = get_proc_task(file_inode(file));
2815 struct mm_struct *mm;
2816 char buffer[PROC_NUMBUF];
2824 mm = get_task_mm(task);
2826 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2827 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2828 MMF_DUMP_FILTER_SHIFT));
2830 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2833 put_task_struct(task);
2838 static ssize_t proc_coredump_filter_write(struct file *file,
2839 const char __user *buf,
2843 struct task_struct *task;
2844 struct mm_struct *mm;
2850 ret = kstrtouint_from_user(buf, count, 0, &val);
2855 task = get_proc_task(file_inode(file));
2859 mm = get_task_mm(task);
2864 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2866 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2868 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2873 put_task_struct(task);
2880 static const struct file_operations proc_coredump_filter_operations = {
2881 .read = proc_coredump_filter_read,
2882 .write = proc_coredump_filter_write,
2883 .llseek = generic_file_llseek,
2887 #ifdef CONFIG_TASK_IO_ACCOUNTING
2888 static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
2890 struct task_io_accounting acct = task->ioac;
2891 unsigned long flags;
2894 result = mutex_lock_killable(&task->signal->exec_update_mutex);
2898 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) {
2903 if (whole && lock_task_sighand(task, &flags)) {
2904 struct task_struct *t = task;
2906 task_io_accounting_add(&acct, &task->signal->ioac);
2907 while_each_thread(task, t)
2908 task_io_accounting_add(&acct, &t->ioac);
2910 unlock_task_sighand(task, &flags);
2917 "read_bytes: %llu\n"
2918 "write_bytes: %llu\n"
2919 "cancelled_write_bytes: %llu\n",
2920 (unsigned long long)acct.rchar,
2921 (unsigned long long)acct.wchar,
2922 (unsigned long long)acct.syscr,
2923 (unsigned long long)acct.syscw,
2924 (unsigned long long)acct.read_bytes,
2925 (unsigned long long)acct.write_bytes,
2926 (unsigned long long)acct.cancelled_write_bytes);
2930 mutex_unlock(&task->signal->exec_update_mutex);
2934 static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2935 struct pid *pid, struct task_struct *task)
2937 return do_io_accounting(task, m, 0);
2940 static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2941 struct pid *pid, struct task_struct *task)
2943 return do_io_accounting(task, m, 1);
2945 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2947 #ifdef CONFIG_USER_NS
2948 static int proc_id_map_open(struct inode *inode, struct file *file,
2949 const struct seq_operations *seq_ops)
2951 struct user_namespace *ns = NULL;
2952 struct task_struct *task;
2953 struct seq_file *seq;
2956 task = get_proc_task(inode);
2959 ns = get_user_ns(task_cred_xxx(task, user_ns));
2961 put_task_struct(task);
2966 ret = seq_open(file, seq_ops);
2970 seq = file->private_data;
2980 static int proc_id_map_release(struct inode *inode, struct file *file)
2982 struct seq_file *seq = file->private_data;
2983 struct user_namespace *ns = seq->private;
2985 return seq_release(inode, file);
2988 static int proc_uid_map_open(struct inode *inode, struct file *file)
2990 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2993 static int proc_gid_map_open(struct inode *inode, struct file *file)
2995 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2998 static int proc_projid_map_open(struct inode *inode, struct file *file)
3000 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
3003 static const struct file_operations proc_uid_map_operations = {
3004 .open = proc_uid_map_open,
3005 .write = proc_uid_map_write,
3007 .llseek = seq_lseek,
3008 .release = proc_id_map_release,
3011 static const struct file_operations proc_gid_map_operations = {
3012 .open = proc_gid_map_open,
3013 .write = proc_gid_map_write,
3015 .llseek = seq_lseek,
3016 .release = proc_id_map_release,
3019 static const struct file_operations proc_projid_map_operations = {
3020 .open = proc_projid_map_open,
3021 .write = proc_projid_map_write,
3023 .llseek = seq_lseek,
3024 .release = proc_id_map_release,
3027 static int proc_setgroups_open(struct inode *inode, struct file *file)
3029 struct user_namespace *ns = NULL;
3030 struct task_struct *task;
3034 task = get_proc_task(inode);
3037 ns = get_user_ns(task_cred_xxx(task, user_ns));
3039 put_task_struct(task);
3044 if (file->f_mode & FMODE_WRITE) {
3046 if (!ns_capable(ns, CAP_SYS_ADMIN))
3050 ret = single_open(file, &proc_setgroups_show, ns);
3061 static int proc_setgroups_release(struct inode *inode, struct file *file)
3063 struct seq_file *seq = file->private_data;
3064 struct user_namespace *ns = seq->private;
3065 int ret = single_release(inode, file);
3070 static const struct file_operations proc_setgroups_operations = {
3071 .open = proc_setgroups_open,
3072 .write = proc_setgroups_write,
3074 .llseek = seq_lseek,
3075 .release = proc_setgroups_release,
3077 #endif /* CONFIG_USER_NS */
3079 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
3080 struct pid *pid, struct task_struct *task)
3082 int err = lock_trace(task);
3084 seq_printf(m, "%08x\n", task->personality);
3090 #ifdef CONFIG_LIVEPATCH
3091 static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns,
3092 struct pid *pid, struct task_struct *task)
3094 seq_printf(m, "%d\n", task->patch_state);
3097 #endif /* CONFIG_LIVEPATCH */
3099 #ifdef CONFIG_STACKLEAK_METRICS
3100 static int proc_stack_depth(struct seq_file *m, struct pid_namespace *ns,
3101 struct pid *pid, struct task_struct *task)
3103 unsigned long prev_depth = THREAD_SIZE -
3104 (task->prev_lowest_stack & (THREAD_SIZE - 1));
3105 unsigned long depth = THREAD_SIZE -
3106 (task->lowest_stack & (THREAD_SIZE - 1));
3108 seq_printf(m, "previous stack depth: %lu\nstack depth: %lu\n",
3112 #endif /* CONFIG_STACKLEAK_METRICS */
3117 static const struct file_operations proc_task_operations;
3118 static const struct inode_operations proc_task_inode_operations;
3120 static const struct pid_entry tgid_base_stuff[] = {
3121 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
3122 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3123 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
3124 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3125 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3127 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3129 REG("environ", S_IRUSR, proc_environ_operations),
3130 REG("auxv", S_IRUSR, proc_auxv_operations),
3131 ONE("status", S_IRUGO, proc_pid_status),
3132 ONE("personality", S_IRUSR, proc_pid_personality),
3133 ONE("limits", S_IRUGO, proc_pid_limits),
3134 #ifdef CONFIG_SCHED_DEBUG
3135 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3137 #ifdef CONFIG_SCHED_AUTOGROUP
3138 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
3140 #ifdef CONFIG_TIME_NS
3141 REG("timens_offsets", S_IRUGO|S_IWUSR, proc_timens_offsets_operations),
3143 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3144 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3145 ONE("syscall", S_IRUSR, proc_pid_syscall),
3147 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
3148 ONE("stat", S_IRUGO, proc_tgid_stat),
3149 ONE("statm", S_IRUGO, proc_pid_statm),
3150 REG("maps", S_IRUGO, proc_pid_maps_operations),
3152 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
3154 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3155 LNK("cwd", proc_cwd_link),
3156 LNK("root", proc_root_link),
3157 LNK("exe", proc_exe_link),
3158 REG("mounts", S_IRUGO, proc_mounts_operations),
3159 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3160 REG("mountstats", S_IRUSR, proc_mountstats_operations),
3161 #ifdef CONFIG_PROC_PAGE_MONITOR
3162 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3163 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
3164 REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
3165 REG("pagemap", S_IRUSR, proc_pagemap_operations),
3167 #ifdef CONFIG_SECURITY
3168 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3170 #ifdef CONFIG_KALLSYMS
3171 ONE("wchan", S_IRUGO, proc_pid_wchan),
3173 #ifdef CONFIG_STACKTRACE
3174 ONE("stack", S_IRUSR, proc_pid_stack),
3176 #ifdef CONFIG_SCHED_INFO
3177 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
3179 #ifdef CONFIG_LATENCYTOP
3180 REG("latency", S_IRUGO, proc_lstats_operations),
3182 #ifdef CONFIG_PROC_PID_CPUSET
3183 ONE("cpuset", S_IRUGO, proc_cpuset_show),
3185 #ifdef CONFIG_CGROUPS
3186 ONE("cgroup", S_IRUGO, proc_cgroup_show),
3188 #ifdef CONFIG_PROC_CPU_RESCTRL
3189 ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show),
3191 ONE("oom_score", S_IRUGO, proc_oom_score),
3192 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3193 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3195 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3196 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3198 #ifdef CONFIG_FAULT_INJECTION
3199 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3200 REG("fail-nth", 0644, proc_fail_nth_operations),
3202 #ifdef CONFIG_ELF_CORE
3203 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3205 #ifdef CONFIG_TASK_IO_ACCOUNTING
3206 ONE("io", S_IRUSR, proc_tgid_io_accounting),
3208 #ifdef CONFIG_USER_NS
3209 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3210 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3211 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3212 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
3214 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
3215 REG("timers", S_IRUGO, proc_timers_operations),
3217 REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations),
3218 #ifdef CONFIG_LIVEPATCH
3219 ONE("patch_state", S_IRUSR, proc_pid_patch_state),
3221 #ifdef CONFIG_STACKLEAK_METRICS
3222 ONE("stack_depth", S_IRUGO, proc_stack_depth),
3224 #ifdef CONFIG_PROC_PID_ARCH_STATUS
3225 ONE("arch_status", S_IRUGO, proc_pid_arch_status),
3229 static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
3231 return proc_pident_readdir(file, ctx,
3232 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3235 static const struct file_operations proc_tgid_base_operations = {
3236 .read = generic_read_dir,
3237 .iterate_shared = proc_tgid_base_readdir,
3238 .llseek = generic_file_llseek,
3241 struct pid *tgid_pidfd_to_pid(const struct file *file)
3243 if (file->f_op != &proc_tgid_base_operations)
3244 return ERR_PTR(-EBADF);
3246 return proc_pid(file_inode(file));
3249 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3251 return proc_pident_lookup(dir, dentry,
3253 tgid_base_stuff + ARRAY_SIZE(tgid_base_stuff));
3256 static const struct inode_operations proc_tgid_base_inode_operations = {
3257 .lookup = proc_tgid_base_lookup,
3258 .getattr = pid_getattr,
3259 .setattr = proc_setattr,
3260 .permission = proc_pid_permission,
3264 * proc_flush_pid - Remove dcache entries for @pid from the /proc dcache.
3265 * @pid: pid that should be flushed.
3267 * This function walks a list of inodes (that belong to any proc
3268 * filesystem) that are attached to the pid and flushes them from
3271 * It is safe and reasonable to cache /proc entries for a task until
3272 * that task exits. After that they just clog up the dcache with
3273 * useless entries, possibly causing useful dcache entries to be
3274 * flushed instead. This routine is provided to flush those useless
3275 * dcache entries when a process is reaped.
3277 * NOTE: This routine is just an optimization so it does not guarantee
3278 * that no dcache entries will exist after a process is reaped
3279 * it just makes it very unlikely that any will persist.
3282 void proc_flush_pid(struct pid *pid)
3284 proc_invalidate_siblings_dcache(&pid->inodes, &pid->lock);
3287 static struct dentry *proc_pid_instantiate(struct dentry * dentry,
3288 struct task_struct *task, const void *ptr)
3290 struct inode *inode;
3292 inode = proc_pid_make_inode(dentry->d_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
3294 return ERR_PTR(-ENOENT);
3296 inode->i_op = &proc_tgid_base_inode_operations;
3297 inode->i_fop = &proc_tgid_base_operations;
3298 inode->i_flags|=S_IMMUTABLE;
3300 set_nlink(inode, nlink_tgid);
3301 pid_update_inode(task, inode);
3303 d_set_d_op(dentry, &pid_dentry_operations);
3304 return d_splice_alias(inode, dentry);
3307 struct dentry *proc_pid_lookup(struct dentry *dentry, unsigned int flags)
3309 struct task_struct *task;
3311 struct proc_fs_info *fs_info;
3312 struct pid_namespace *ns;
3313 struct dentry *result = ERR_PTR(-ENOENT);
3315 tgid = name_to_int(&dentry->d_name);
3319 fs_info = proc_sb_info(dentry->d_sb);
3320 ns = fs_info->pid_ns;
3322 task = find_task_by_pid_ns(tgid, ns);
3324 get_task_struct(task);
3329 /* Limit procfs to only ptraceable tasks */
3330 if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE) {
3331 if (!has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS))
3335 result = proc_pid_instantiate(dentry, task, NULL);
3337 put_task_struct(task);
3343 * Find the first task with tgid >= tgid
3348 struct task_struct *task;
3350 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3355 put_task_struct(iter.task);
3359 pid = find_ge_pid(iter.tgid, ns);
3361 iter.tgid = pid_nr_ns(pid, ns);
3362 iter.task = pid_task(pid, PIDTYPE_TGID);
3367 get_task_struct(iter.task);
3373 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3375 /* for the /proc/ directory itself, after non-process stuff has been done */
3376 int proc_pid_readdir(struct file *file, struct dir_context *ctx)
3378 struct tgid_iter iter;
3379 struct proc_fs_info *fs_info = proc_sb_info(file_inode(file)->i_sb);
3380 struct pid_namespace *ns = proc_pid_ns(file_inode(file));
3381 loff_t pos = ctx->pos;
3383 if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
3386 if (pos == TGID_OFFSET - 2) {
3387 struct inode *inode = d_inode(fs_info->proc_self);
3388 if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
3390 ctx->pos = pos = pos + 1;
3392 if (pos == TGID_OFFSET - 1) {
3393 struct inode *inode = d_inode(fs_info->proc_thread_self);
3394 if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3396 ctx->pos = pos = pos + 1;
3398 iter.tgid = pos - TGID_OFFSET;
3400 for (iter = next_tgid(ns, iter);
3402 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3407 if (!has_pid_permissions(fs_info, iter.task, HIDEPID_INVISIBLE))
3410 len = snprintf(name, sizeof(name), "%u", iter.tgid);
3411 ctx->pos = iter.tgid + TGID_OFFSET;
3412 if (!proc_fill_cache(file, ctx, name, len,
3413 proc_pid_instantiate, iter.task, NULL)) {
3414 put_task_struct(iter.task);
3418 ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
3423 * proc_tid_comm_permission is a special permission function exclusively
3424 * used for the node /proc/<pid>/task/<tid>/comm.
3425 * It bypasses generic permission checks in the case where a task of the same
3426 * task group attempts to access the node.
3427 * The rationale behind this is that glibc and bionic access this node for
3428 * cross thread naming (pthread_set/getname_np(!self)). However, if
3429 * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3430 * which locks out the cross thread naming implementation.
3431 * This function makes sure that the node is always accessible for members of
3432 * same thread group.
3434 static int proc_tid_comm_permission(struct inode *inode, int mask)
3436 bool is_same_tgroup;
3437 struct task_struct *task;
3439 task = get_proc_task(inode);
3442 is_same_tgroup = same_thread_group(current, task);
3443 put_task_struct(task);
3445 if (likely(is_same_tgroup && !(mask & MAY_EXEC))) {
3446 /* This file (/proc/<pid>/task/<tid>/comm) can always be
3447 * read or written by the members of the corresponding
3453 return generic_permission(inode, mask);
3456 static const struct inode_operations proc_tid_comm_inode_operations = {
3457 .permission = proc_tid_comm_permission,
3463 static const struct pid_entry tid_base_stuff[] = {
3464 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3465 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3466 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3468 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3470 REG("environ", S_IRUSR, proc_environ_operations),
3471 REG("auxv", S_IRUSR, proc_auxv_operations),
3472 ONE("status", S_IRUGO, proc_pid_status),
3473 ONE("personality", S_IRUSR, proc_pid_personality),
3474 ONE("limits", S_IRUGO, proc_pid_limits),
3475 #ifdef CONFIG_SCHED_DEBUG
3476 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3478 NOD("comm", S_IFREG|S_IRUGO|S_IWUSR,
3479 &proc_tid_comm_inode_operations,
3480 &proc_pid_set_comm_operations, {}),
3481 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3482 ONE("syscall", S_IRUSR, proc_pid_syscall),
3484 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
3485 ONE("stat", S_IRUGO, proc_tid_stat),
3486 ONE("statm", S_IRUGO, proc_pid_statm),
3487 REG("maps", S_IRUGO, proc_pid_maps_operations),
3488 #ifdef CONFIG_PROC_CHILDREN
3489 REG("children", S_IRUGO, proc_tid_children_operations),
3492 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
3494 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3495 LNK("cwd", proc_cwd_link),
3496 LNK("root", proc_root_link),
3497 LNK("exe", proc_exe_link),
3498 REG("mounts", S_IRUGO, proc_mounts_operations),
3499 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3500 #ifdef CONFIG_PROC_PAGE_MONITOR
3501 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3502 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
3503 REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
3504 REG("pagemap", S_IRUSR, proc_pagemap_operations),
3506 #ifdef CONFIG_SECURITY
3507 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3509 #ifdef CONFIG_KALLSYMS
3510 ONE("wchan", S_IRUGO, proc_pid_wchan),
3512 #ifdef CONFIG_STACKTRACE
3513 ONE("stack", S_IRUSR, proc_pid_stack),
3515 #ifdef CONFIG_SCHED_INFO
3516 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
3518 #ifdef CONFIG_LATENCYTOP
3519 REG("latency", S_IRUGO, proc_lstats_operations),
3521 #ifdef CONFIG_PROC_PID_CPUSET
3522 ONE("cpuset", S_IRUGO, proc_cpuset_show),
3524 #ifdef CONFIG_CGROUPS
3525 ONE("cgroup", S_IRUGO, proc_cgroup_show),
3527 #ifdef CONFIG_PROC_CPU_RESCTRL
3528 ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show),
3530 ONE("oom_score", S_IRUGO, proc_oom_score),
3531 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3532 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3534 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3535 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3537 #ifdef CONFIG_FAULT_INJECTION
3538 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3539 REG("fail-nth", 0644, proc_fail_nth_operations),
3541 #ifdef CONFIG_TASK_IO_ACCOUNTING
3542 ONE("io", S_IRUSR, proc_tid_io_accounting),
3544 #ifdef CONFIG_USER_NS
3545 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3546 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3547 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3548 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
3550 #ifdef CONFIG_LIVEPATCH
3551 ONE("patch_state", S_IRUSR, proc_pid_patch_state),
3553 #ifdef CONFIG_PROC_PID_ARCH_STATUS
3554 ONE("arch_status", S_IRUGO, proc_pid_arch_status),
3558 static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
3560 return proc_pident_readdir(file, ctx,
3561 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3564 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3566 return proc_pident_lookup(dir, dentry,
3568 tid_base_stuff + ARRAY_SIZE(tid_base_stuff));
3571 static const struct file_operations proc_tid_base_operations = {
3572 .read = generic_read_dir,
3573 .iterate_shared = proc_tid_base_readdir,
3574 .llseek = generic_file_llseek,
3577 static const struct inode_operations proc_tid_base_inode_operations = {
3578 .lookup = proc_tid_base_lookup,
3579 .getattr = pid_getattr,
3580 .setattr = proc_setattr,
3583 static struct dentry *proc_task_instantiate(struct dentry *dentry,
3584 struct task_struct *task, const void *ptr)
3586 struct inode *inode;
3587 inode = proc_pid_make_inode(dentry->d_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
3589 return ERR_PTR(-ENOENT);
3591 inode->i_op = &proc_tid_base_inode_operations;
3592 inode->i_fop = &proc_tid_base_operations;
3593 inode->i_flags |= S_IMMUTABLE;
3595 set_nlink(inode, nlink_tid);
3596 pid_update_inode(task, inode);
3598 d_set_d_op(dentry, &pid_dentry_operations);
3599 return d_splice_alias(inode, dentry);
3602 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3604 struct task_struct *task;
3605 struct task_struct *leader = get_proc_task(dir);
3607 struct proc_fs_info *fs_info;
3608 struct pid_namespace *ns;
3609 struct dentry *result = ERR_PTR(-ENOENT);
3614 tid = name_to_int(&dentry->d_name);
3618 fs_info = proc_sb_info(dentry->d_sb);
3619 ns = fs_info->pid_ns;
3621 task = find_task_by_pid_ns(tid, ns);
3623 get_task_struct(task);
3627 if (!same_thread_group(leader, task))
3630 result = proc_task_instantiate(dentry, task, NULL);
3632 put_task_struct(task);
3634 put_task_struct(leader);
3640 * Find the first tid of a thread group to return to user space.
3642 * Usually this is just the thread group leader, but if the users
3643 * buffer was too small or there was a seek into the middle of the
3644 * directory we have more work todo.
3646 * In the case of a short read we start with find_task_by_pid.
3648 * In the case of a seek we start with the leader and walk nr
3651 static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3652 struct pid_namespace *ns)
3654 struct task_struct *pos, *task;
3655 unsigned long nr = f_pos;
3657 if (nr != f_pos) /* 32bit overflow? */
3661 task = pid_task(pid, PIDTYPE_PID);
3665 /* Attempt to start with the tid of a thread */
3667 pos = find_task_by_pid_ns(tid, ns);
3668 if (pos && same_thread_group(pos, task))
3672 /* If nr exceeds the number of threads there is nothing todo */
3673 if (nr >= get_nr_threads(task))
3676 /* If we haven't found our starting place yet start
3677 * with the leader and walk nr threads forward.
3679 pos = task = task->group_leader;
3683 } while_each_thread(task, pos);
3688 get_task_struct(pos);
3695 * Find the next thread in the thread list.
3696 * Return NULL if there is an error or no next thread.
3698 * The reference to the input task_struct is released.
3700 static struct task_struct *next_tid(struct task_struct *start)
3702 struct task_struct *pos = NULL;
3704 if (pid_alive(start)) {
3705 pos = next_thread(start);
3706 if (thread_group_leader(pos))
3709 get_task_struct(pos);
3712 put_task_struct(start);
3716 /* for the /proc/TGID/task/ directories */
3717 static int proc_task_readdir(struct file *file, struct dir_context *ctx)
3719 struct inode *inode = file_inode(file);
3720 struct task_struct *task;
3721 struct pid_namespace *ns;
3724 if (proc_inode_is_dead(inode))
3727 if (!dir_emit_dots(file, ctx))
3730 /* f_version caches the tgid value that the last readdir call couldn't
3731 * return. lseek aka telldir automagically resets f_version to 0.
3733 ns = proc_pid_ns(inode);
3734 tid = (int)file->f_version;
3735 file->f_version = 0;
3736 for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
3738 task = next_tid(task), ctx->pos++) {
3741 tid = task_pid_nr_ns(task, ns);
3742 len = snprintf(name, sizeof(name), "%u", tid);
3743 if (!proc_fill_cache(file, ctx, name, len,
3744 proc_task_instantiate, task, NULL)) {
3745 /* returning this tgid failed, save it as the first
3746 * pid for the next readir call */
3747 file->f_version = (u64)tid;
3748 put_task_struct(task);
3756 static int proc_task_getattr(const struct path *path, struct kstat *stat,
3757 u32 request_mask, unsigned int query_flags)
3759 struct inode *inode = d_inode(path->dentry);
3760 struct task_struct *p = get_proc_task(inode);
3761 generic_fillattr(inode, stat);
3764 stat->nlink += get_nr_threads(p);
3771 static const struct inode_operations proc_task_inode_operations = {
3772 .lookup = proc_task_lookup,
3773 .getattr = proc_task_getattr,
3774 .setattr = proc_setattr,
3775 .permission = proc_pid_permission,
3778 static const struct file_operations proc_task_operations = {
3779 .read = generic_read_dir,
3780 .iterate_shared = proc_task_readdir,
3781 .llseek = generic_file_llseek,
3784 void __init set_proc_pid_nlink(void)
3786 nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3787 nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));