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/printk.h>
78 #include <linux/cache.h>
79 #include <linux/cgroup.h>
80 #include <linux/cpuset.h>
81 #include <linux/audit.h>
82 #include <linux/poll.h>
83 #include <linux/nsproxy.h>
84 #include <linux/oom.h>
85 #include <linux/elf.h>
86 #include <linux/pid_namespace.h>
87 #include <linux/user_namespace.h>
88 #include <linux/fs_struct.h>
89 #include <linux/slab.h>
90 #include <linux/sched/autogroup.h>
91 #include <linux/sched/mm.h>
92 #include <linux/sched/coredump.h>
93 #include <linux/sched/debug.h>
94 #include <linux/sched/stat.h>
95 #include <linux/posix-timers.h>
96 #include <linux/time_namespace.h>
97 #include <linux/resctrl.h>
98 #include <linux/cn_proc.h>
99 #include <linux/ksm.h>
100 #include <uapi/linux/lsm.h>
101 #include <trace/events/oom.h>
102 #include "internal.h"
105 #include "../../lib/kstrtox.h"
108 * Implementing inode permission operations in /proc is almost
109 * certainly an error. Permission checks need to happen during
110 * each system call not at open time. The reason is that most of
111 * what we wish to check for permissions in /proc varies at runtime.
113 * The classic example of a problem is opening file descriptors
114 * in /proc for a task before it execs a suid executable.
117 static u8 nlink_tid __ro_after_init;
118 static u8 nlink_tgid __ro_after_init;
124 const struct inode_operations *iop;
125 const struct file_operations *fop;
129 #define NOD(NAME, MODE, IOP, FOP, OP) { \
131 .len = sizeof(NAME) - 1, \
138 #define DIR(NAME, MODE, iops, fops) \
139 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
140 #define LNK(NAME, get_link) \
141 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
142 &proc_pid_link_inode_operations, NULL, \
143 { .proc_get_link = get_link } )
144 #define REG(NAME, MODE, fops) \
145 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
146 #define ONE(NAME, MODE, show) \
147 NOD(NAME, (S_IFREG|(MODE)), \
148 NULL, &proc_single_file_operations, \
149 { .proc_show = show } )
150 #define ATTR(LSMID, NAME, MODE) \
151 NOD(NAME, (S_IFREG|(MODE)), \
152 NULL, &proc_pid_attr_operations, \
156 * Count the number of hardlinks for the pid_entry table, excluding the .
159 static unsigned int __init pid_entry_nlink(const struct pid_entry *entries,
166 for (i = 0; i < n; ++i) {
167 if (S_ISDIR(entries[i].mode))
174 static int get_task_root(struct task_struct *task, struct path *root)
176 int result = -ENOENT;
180 get_fs_root(task->fs, root);
187 static int proc_cwd_link(struct dentry *dentry, struct path *path)
189 struct task_struct *task = get_proc_task(d_inode(dentry));
190 int result = -ENOENT;
195 get_fs_pwd(task->fs, path);
199 put_task_struct(task);
204 static int proc_root_link(struct dentry *dentry, struct path *path)
206 struct task_struct *task = get_proc_task(d_inode(dentry));
207 int result = -ENOENT;
210 result = get_task_root(task, path);
211 put_task_struct(task);
217 * If the user used setproctitle(), we just get the string from
218 * user space at arg_start, and limit it to a maximum of one page.
220 static ssize_t get_mm_proctitle(struct mm_struct *mm, char __user *buf,
221 size_t count, unsigned long pos,
222 unsigned long arg_start)
227 if (pos >= PAGE_SIZE)
230 page = (char *)__get_free_page(GFP_KERNEL);
235 got = access_remote_vm(mm, arg_start, page, PAGE_SIZE, FOLL_ANON);
237 int len = strnlen(page, got);
239 /* Include the NUL character if it was found */
247 len -= copy_to_user(buf, page+pos, len);
253 free_page((unsigned long)page);
257 static ssize_t get_mm_cmdline(struct mm_struct *mm, char __user *buf,
258 size_t count, loff_t *ppos)
260 unsigned long arg_start, arg_end, env_start, env_end;
261 unsigned long pos, len;
264 /* Check if process spawned far enough to have cmdline. */
268 spin_lock(&mm->arg_lock);
269 arg_start = mm->arg_start;
270 arg_end = mm->arg_end;
271 env_start = mm->env_start;
272 env_end = mm->env_end;
273 spin_unlock(&mm->arg_lock);
275 if (arg_start >= arg_end)
279 * We allow setproctitle() to overwrite the argument
280 * strings, and overflow past the original end. But
281 * only when it overflows into the environment area.
283 if (env_start != arg_end || env_end < env_start)
284 env_start = env_end = arg_end;
285 len = env_end - arg_start;
287 /* We're not going to care if "*ppos" has high bits set */
291 if (count > len - pos)
297 * Magical special case: if the argv[] end byte is not
298 * zero, the user has overwritten it with setproctitle(3).
300 * Possible future enhancement: do this only once when
301 * pos is 0, and set a flag in the 'struct file'.
303 if (access_remote_vm(mm, arg_end-1, &c, 1, FOLL_ANON) == 1 && c)
304 return get_mm_proctitle(mm, buf, count, pos, arg_start);
307 * For the non-setproctitle() case we limit things strictly
308 * to the [arg_start, arg_end[ range.
311 if (pos < arg_start || pos >= arg_end)
313 if (count > arg_end - pos)
314 count = arg_end - pos;
316 page = (char *)__get_free_page(GFP_KERNEL);
323 size_t size = min_t(size_t, PAGE_SIZE, count);
325 got = access_remote_vm(mm, pos, page, size, FOLL_ANON);
328 got -= copy_to_user(buf, page, got);
329 if (unlikely(!got)) {
340 free_page((unsigned long)page);
344 static ssize_t get_task_cmdline(struct task_struct *tsk, char __user *buf,
345 size_t count, loff_t *pos)
347 struct mm_struct *mm;
350 mm = get_task_mm(tsk);
354 ret = get_mm_cmdline(mm, buf, count, pos);
359 static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
360 size_t count, loff_t *pos)
362 struct task_struct *tsk;
367 tsk = get_proc_task(file_inode(file));
370 ret = get_task_cmdline(tsk, buf, count, pos);
371 put_task_struct(tsk);
377 static const struct file_operations proc_pid_cmdline_ops = {
378 .read = proc_pid_cmdline_read,
379 .llseek = generic_file_llseek,
382 #ifdef CONFIG_KALLSYMS
384 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
385 * Returns the resolved symbol. If that fails, simply return the address.
387 static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
388 struct pid *pid, struct task_struct *task)
391 char symname[KSYM_NAME_LEN];
393 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
396 wchan = get_wchan(task);
397 if (wchan && !lookup_symbol_name(wchan, symname)) {
398 seq_puts(m, symname);
406 #endif /* CONFIG_KALLSYMS */
408 static int lock_trace(struct task_struct *task)
410 int err = down_read_killable(&task->signal->exec_update_lock);
413 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) {
414 up_read(&task->signal->exec_update_lock);
420 static void unlock_trace(struct task_struct *task)
422 up_read(&task->signal->exec_update_lock);
425 #ifdef CONFIG_STACKTRACE
427 #define MAX_STACK_TRACE_DEPTH 64
429 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
430 struct pid *pid, struct task_struct *task)
432 unsigned long *entries;
436 * The ability to racily run the kernel stack unwinder on a running task
437 * and then observe the unwinder output is scary; while it is useful for
438 * debugging kernel issues, it can also allow an attacker to leak kernel
440 * Doing this in a manner that is at least safe from races would require
441 * some work to ensure that the remote task can not be scheduled; and
442 * even then, this would still expose the unwinder as local attack
444 * Therefore, this interface is restricted to root.
446 if (!file_ns_capable(m->file, &init_user_ns, CAP_SYS_ADMIN))
449 entries = kmalloc_array(MAX_STACK_TRACE_DEPTH, sizeof(*entries),
454 err = lock_trace(task);
456 unsigned int i, nr_entries;
458 nr_entries = stack_trace_save_tsk(task, entries,
459 MAX_STACK_TRACE_DEPTH, 0);
461 for (i = 0; i < nr_entries; i++) {
462 seq_printf(m, "[<0>] %pB\n", (void *)entries[i]);
473 #ifdef CONFIG_SCHED_INFO
475 * Provides /proc/PID/schedstat
477 static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
478 struct pid *pid, struct task_struct *task)
480 if (unlikely(!sched_info_on()))
481 seq_puts(m, "0 0 0\n");
483 seq_printf(m, "%llu %llu %lu\n",
484 (unsigned long long)task->se.sum_exec_runtime,
485 (unsigned long long)task->sched_info.run_delay,
486 task->sched_info.pcount);
492 #ifdef CONFIG_LATENCYTOP
493 static int lstats_show_proc(struct seq_file *m, void *v)
496 struct inode *inode = m->private;
497 struct task_struct *task = get_proc_task(inode);
501 seq_puts(m, "Latency Top version : v0.1\n");
502 for (i = 0; i < LT_SAVECOUNT; i++) {
503 struct latency_record *lr = &task->latency_record[i];
504 if (lr->backtrace[0]) {
506 seq_printf(m, "%i %li %li",
507 lr->count, lr->time, lr->max);
508 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
509 unsigned long bt = lr->backtrace[q];
513 seq_printf(m, " %ps", (void *)bt);
519 put_task_struct(task);
523 static int lstats_open(struct inode *inode, struct file *file)
525 return single_open(file, lstats_show_proc, inode);
528 static ssize_t lstats_write(struct file *file, const char __user *buf,
529 size_t count, loff_t *offs)
531 struct task_struct *task = get_proc_task(file_inode(file));
535 clear_tsk_latency_tracing(task);
536 put_task_struct(task);
541 static const struct file_operations proc_lstats_operations = {
544 .write = lstats_write,
546 .release = single_release,
551 static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
552 struct pid *pid, struct task_struct *task)
554 unsigned long totalpages = totalram_pages() + total_swap_pages;
555 unsigned long points = 0;
558 badness = oom_badness(task, totalpages);
560 * Special case OOM_SCORE_ADJ_MIN for all others scale the
561 * badness value into [0, 2000] range which we have been
562 * exporting for a long time so userspace might depend on it.
564 if (badness != LONG_MIN)
565 points = (1000 + badness * 1000 / (long)totalpages) * 2 / 3;
567 seq_printf(m, "%lu\n", points);
577 static const struct limit_names lnames[RLIM_NLIMITS] = {
578 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
579 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
580 [RLIMIT_DATA] = {"Max data size", "bytes"},
581 [RLIMIT_STACK] = {"Max stack size", "bytes"},
582 [RLIMIT_CORE] = {"Max core file size", "bytes"},
583 [RLIMIT_RSS] = {"Max resident set", "bytes"},
584 [RLIMIT_NPROC] = {"Max processes", "processes"},
585 [RLIMIT_NOFILE] = {"Max open files", "files"},
586 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
587 [RLIMIT_AS] = {"Max address space", "bytes"},
588 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
589 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
590 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
591 [RLIMIT_NICE] = {"Max nice priority", NULL},
592 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
593 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
596 /* Display limits for a process */
597 static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
598 struct pid *pid, struct task_struct *task)
603 struct rlimit rlim[RLIM_NLIMITS];
605 if (!lock_task_sighand(task, &flags))
607 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
608 unlock_task_sighand(task, &flags);
611 * print the file header
618 for (i = 0; i < RLIM_NLIMITS; i++) {
619 if (rlim[i].rlim_cur == RLIM_INFINITY)
620 seq_printf(m, "%-25s %-20s ",
621 lnames[i].name, "unlimited");
623 seq_printf(m, "%-25s %-20lu ",
624 lnames[i].name, rlim[i].rlim_cur);
626 if (rlim[i].rlim_max == RLIM_INFINITY)
627 seq_printf(m, "%-20s ", "unlimited");
629 seq_printf(m, "%-20lu ", rlim[i].rlim_max);
632 seq_printf(m, "%-10s\n", lnames[i].unit);
640 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
641 static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
642 struct pid *pid, struct task_struct *task)
644 struct syscall_info info;
645 u64 *args = &info.data.args[0];
648 res = lock_trace(task);
652 if (task_current_syscall(task, &info))
653 seq_puts(m, "running\n");
654 else if (info.data.nr < 0)
655 seq_printf(m, "%d 0x%llx 0x%llx\n",
656 info.data.nr, info.sp, info.data.instruction_pointer);
659 "%d 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx\n",
661 args[0], args[1], args[2], args[3], args[4], args[5],
662 info.sp, info.data.instruction_pointer);
667 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
669 /************************************************************************/
670 /* Here the fs part begins */
671 /************************************************************************/
673 /* permission checks */
674 static bool proc_fd_access_allowed(struct inode *inode)
676 struct task_struct *task;
677 bool allowed = false;
678 /* Allow access to a task's file descriptors if it is us or we
679 * may use ptrace attach to the process and find out that
682 task = get_proc_task(inode);
684 allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
685 put_task_struct(task);
690 int proc_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
694 struct inode *inode = d_inode(dentry);
696 if (attr->ia_valid & ATTR_MODE)
699 error = setattr_prepare(&nop_mnt_idmap, dentry, attr);
703 setattr_copy(&nop_mnt_idmap, inode, attr);
708 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
709 * or euid/egid (for hide_pid_min=2)?
711 static bool has_pid_permissions(struct proc_fs_info *fs_info,
712 struct task_struct *task,
713 enum proc_hidepid hide_pid_min)
716 * If 'hidpid' mount option is set force a ptrace check,
717 * we indicate that we are using a filesystem syscall
718 * by passing PTRACE_MODE_READ_FSCREDS
720 if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE)
721 return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
723 if (fs_info->hide_pid < hide_pid_min)
725 if (in_group_p(fs_info->pid_gid))
727 return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
731 static int proc_pid_permission(struct mnt_idmap *idmap,
732 struct inode *inode, int mask)
734 struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
735 struct task_struct *task;
738 task = get_proc_task(inode);
741 has_perms = has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS);
742 put_task_struct(task);
745 if (fs_info->hide_pid == HIDEPID_INVISIBLE) {
747 * Let's make getdents(), stat(), and open()
748 * consistent with each other. If a process
749 * may not stat() a file, it shouldn't be seen
757 return generic_permission(&nop_mnt_idmap, inode, mask);
762 static const struct inode_operations proc_def_inode_operations = {
763 .setattr = proc_setattr,
766 static int proc_single_show(struct seq_file *m, void *v)
768 struct inode *inode = m->private;
769 struct pid_namespace *ns = proc_pid_ns(inode->i_sb);
770 struct pid *pid = proc_pid(inode);
771 struct task_struct *task;
774 task = get_pid_task(pid, PIDTYPE_PID);
778 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
780 put_task_struct(task);
784 static int proc_single_open(struct inode *inode, struct file *filp)
786 return single_open(filp, proc_single_show, inode);
789 static const struct file_operations proc_single_file_operations = {
790 .open = proc_single_open,
793 .release = single_release,
797 struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
799 struct task_struct *task = get_proc_task(inode);
800 struct mm_struct *mm = ERR_PTR(-ESRCH);
803 mm = mm_access(task, mode | PTRACE_MODE_FSCREDS);
804 put_task_struct(task);
806 if (!IS_ERR_OR_NULL(mm)) {
807 /* ensure this mm_struct can't be freed */
809 /* but do not pin its memory */
817 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
819 struct mm_struct *mm = proc_mem_open(inode, mode);
824 file->private_data = mm;
828 static int mem_open(struct inode *inode, struct file *file)
830 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
832 /* OK to pass negative loff_t, we can catch out-of-range */
833 file->f_mode |= FMODE_UNSIGNED_OFFSET;
838 static ssize_t mem_rw(struct file *file, char __user *buf,
839 size_t count, loff_t *ppos, int write)
841 struct mm_struct *mm = file->private_data;
842 unsigned long addr = *ppos;
850 page = (char *)__get_free_page(GFP_KERNEL);
855 if (!mmget_not_zero(mm))
858 flags = FOLL_FORCE | (write ? FOLL_WRITE : 0);
861 size_t this_len = min_t(size_t, count, PAGE_SIZE);
863 if (write && copy_from_user(page, buf, this_len)) {
868 this_len = access_remote_vm(mm, addr, page, this_len, flags);
875 if (!write && copy_to_user(buf, page, this_len)) {
889 free_page((unsigned long) page);
893 static ssize_t mem_read(struct file *file, char __user *buf,
894 size_t count, loff_t *ppos)
896 return mem_rw(file, buf, count, ppos, 0);
899 static ssize_t mem_write(struct file *file, const char __user *buf,
900 size_t count, loff_t *ppos)
902 return mem_rw(file, (char __user*)buf, count, ppos, 1);
905 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
909 file->f_pos = offset;
912 file->f_pos += offset;
917 force_successful_syscall_return();
921 static int mem_release(struct inode *inode, struct file *file)
923 struct mm_struct *mm = file->private_data;
929 static const struct file_operations proc_mem_operations = {
934 .release = mem_release,
937 static int environ_open(struct inode *inode, struct file *file)
939 return __mem_open(inode, file, PTRACE_MODE_READ);
942 static ssize_t environ_read(struct file *file, char __user *buf,
943 size_t count, loff_t *ppos)
946 unsigned long src = *ppos;
948 struct mm_struct *mm = file->private_data;
949 unsigned long env_start, env_end;
951 /* Ensure the process spawned far enough to have an environment. */
952 if (!mm || !mm->env_end)
955 page = (char *)__get_free_page(GFP_KERNEL);
960 if (!mmget_not_zero(mm))
963 spin_lock(&mm->arg_lock);
964 env_start = mm->env_start;
965 env_end = mm->env_end;
966 spin_unlock(&mm->arg_lock);
969 size_t this_len, max_len;
972 if (src >= (env_end - env_start))
975 this_len = env_end - (env_start + src);
977 max_len = min_t(size_t, PAGE_SIZE, count);
978 this_len = min(max_len, this_len);
980 retval = access_remote_vm(mm, (env_start + src), page, this_len, FOLL_ANON);
987 if (copy_to_user(buf, page, retval)) {
1001 free_page((unsigned long) page);
1005 static const struct file_operations proc_environ_operations = {
1006 .open = environ_open,
1007 .read = environ_read,
1008 .llseek = generic_file_llseek,
1009 .release = mem_release,
1012 static int auxv_open(struct inode *inode, struct file *file)
1014 return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
1017 static ssize_t auxv_read(struct file *file, char __user *buf,
1018 size_t count, loff_t *ppos)
1020 struct mm_struct *mm = file->private_data;
1021 unsigned int nwords = 0;
1027 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
1028 return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv,
1029 nwords * sizeof(mm->saved_auxv[0]));
1032 static const struct file_operations proc_auxv_operations = {
1035 .llseek = generic_file_llseek,
1036 .release = mem_release,
1039 static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1042 struct task_struct *task = get_proc_task(file_inode(file));
1043 char buffer[PROC_NUMBUF];
1044 int oom_adj = OOM_ADJUST_MIN;
1049 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1050 oom_adj = OOM_ADJUST_MAX;
1052 oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1054 put_task_struct(task);
1055 if (oom_adj > OOM_ADJUST_MAX)
1056 oom_adj = OOM_ADJUST_MAX;
1057 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1058 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1061 static int __set_oom_adj(struct file *file, int oom_adj, bool legacy)
1063 struct mm_struct *mm = NULL;
1064 struct task_struct *task;
1067 task = get_proc_task(file_inode(file));
1071 mutex_lock(&oom_adj_mutex);
1073 if (oom_adj < task->signal->oom_score_adj &&
1074 !capable(CAP_SYS_RESOURCE)) {
1079 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1080 * /proc/pid/oom_score_adj instead.
1082 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1083 current->comm, task_pid_nr(current), task_pid_nr(task),
1086 if ((short)oom_adj < task->signal->oom_score_adj_min &&
1087 !capable(CAP_SYS_RESOURCE)) {
1094 * Make sure we will check other processes sharing the mm if this is
1095 * not vfrok which wants its own oom_score_adj.
1096 * pin the mm so it doesn't go away and get reused after task_unlock
1098 if (!task->vfork_done) {
1099 struct task_struct *p = find_lock_task_mm(task);
1102 if (test_bit(MMF_MULTIPROCESS, &p->mm->flags)) {
1110 task->signal->oom_score_adj = oom_adj;
1111 if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1112 task->signal->oom_score_adj_min = (short)oom_adj;
1113 trace_oom_score_adj_update(task);
1116 struct task_struct *p;
1119 for_each_process(p) {
1120 if (same_thread_group(task, p))
1123 /* do not touch kernel threads or the global init */
1124 if (p->flags & PF_KTHREAD || is_global_init(p))
1128 if (!p->vfork_done && process_shares_mm(p, mm)) {
1129 p->signal->oom_score_adj = oom_adj;
1130 if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1131 p->signal->oom_score_adj_min = (short)oom_adj;
1139 mutex_unlock(&oom_adj_mutex);
1140 put_task_struct(task);
1145 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1146 * kernels. The effective policy is defined by oom_score_adj, which has a
1147 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1148 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1149 * Processes that become oom disabled via oom_adj will still be oom disabled
1150 * with this implementation.
1152 * oom_adj cannot be removed since existing userspace binaries use it.
1154 static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1155 size_t count, loff_t *ppos)
1157 char buffer[PROC_NUMBUF] = {};
1161 if (count > sizeof(buffer) - 1)
1162 count = sizeof(buffer) - 1;
1163 if (copy_from_user(buffer, buf, count)) {
1168 err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1171 if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1172 oom_adj != OOM_DISABLE) {
1178 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1179 * value is always attainable.
1181 if (oom_adj == OOM_ADJUST_MAX)
1182 oom_adj = OOM_SCORE_ADJ_MAX;
1184 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1186 err = __set_oom_adj(file, oom_adj, true);
1188 return err < 0 ? err : count;
1191 static const struct file_operations proc_oom_adj_operations = {
1192 .read = oom_adj_read,
1193 .write = oom_adj_write,
1194 .llseek = generic_file_llseek,
1197 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1198 size_t count, loff_t *ppos)
1200 struct task_struct *task = get_proc_task(file_inode(file));
1201 char buffer[PROC_NUMBUF];
1202 short oom_score_adj = OOM_SCORE_ADJ_MIN;
1207 oom_score_adj = task->signal->oom_score_adj;
1208 put_task_struct(task);
1209 len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
1210 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1213 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1214 size_t count, loff_t *ppos)
1216 char buffer[PROC_NUMBUF] = {};
1220 if (count > sizeof(buffer) - 1)
1221 count = sizeof(buffer) - 1;
1222 if (copy_from_user(buffer, buf, count)) {
1227 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1230 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1231 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1236 err = __set_oom_adj(file, oom_score_adj, false);
1238 return err < 0 ? err : count;
1241 static const struct file_operations proc_oom_score_adj_operations = {
1242 .read = oom_score_adj_read,
1243 .write = oom_score_adj_write,
1244 .llseek = default_llseek,
1248 #define TMPBUFLEN 11
1249 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1250 size_t count, loff_t *ppos)
1252 struct inode * inode = file_inode(file);
1253 struct task_struct *task = get_proc_task(inode);
1255 char tmpbuf[TMPBUFLEN];
1259 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1260 from_kuid(file->f_cred->user_ns,
1261 audit_get_loginuid(task)));
1262 put_task_struct(task);
1263 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1266 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1267 size_t count, loff_t *ppos)
1269 struct inode * inode = file_inode(file);
1274 /* Don't let kthreads write their own loginuid */
1275 if (current->flags & PF_KTHREAD)
1279 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1286 /* No partial writes. */
1290 rv = kstrtou32_from_user(buf, count, 10, &loginuid);
1294 /* is userspace tring to explicitly UNSET the loginuid? */
1295 if (loginuid == AUDIT_UID_UNSET) {
1296 kloginuid = INVALID_UID;
1298 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1299 if (!uid_valid(kloginuid))
1303 rv = audit_set_loginuid(kloginuid);
1309 static const struct file_operations proc_loginuid_operations = {
1310 .read = proc_loginuid_read,
1311 .write = proc_loginuid_write,
1312 .llseek = generic_file_llseek,
1315 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1316 size_t count, loff_t *ppos)
1318 struct inode * inode = file_inode(file);
1319 struct task_struct *task = get_proc_task(inode);
1321 char tmpbuf[TMPBUFLEN];
1325 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1326 audit_get_sessionid(task));
1327 put_task_struct(task);
1328 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1331 static const struct file_operations proc_sessionid_operations = {
1332 .read = proc_sessionid_read,
1333 .llseek = generic_file_llseek,
1337 #ifdef CONFIG_FAULT_INJECTION
1338 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1339 size_t count, loff_t *ppos)
1341 struct task_struct *task = get_proc_task(file_inode(file));
1342 char buffer[PROC_NUMBUF];
1348 make_it_fail = task->make_it_fail;
1349 put_task_struct(task);
1351 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1353 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1356 static ssize_t proc_fault_inject_write(struct file * file,
1357 const char __user * buf, size_t count, loff_t *ppos)
1359 struct task_struct *task;
1360 char buffer[PROC_NUMBUF] = {};
1364 if (!capable(CAP_SYS_RESOURCE))
1367 if (count > sizeof(buffer) - 1)
1368 count = sizeof(buffer) - 1;
1369 if (copy_from_user(buffer, buf, count))
1371 rv = kstrtoint(strstrip(buffer), 0, &make_it_fail);
1374 if (make_it_fail < 0 || make_it_fail > 1)
1377 task = get_proc_task(file_inode(file));
1380 task->make_it_fail = make_it_fail;
1381 put_task_struct(task);
1386 static const struct file_operations proc_fault_inject_operations = {
1387 .read = proc_fault_inject_read,
1388 .write = proc_fault_inject_write,
1389 .llseek = generic_file_llseek,
1392 static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf,
1393 size_t count, loff_t *ppos)
1395 struct task_struct *task;
1399 err = kstrtouint_from_user(buf, count, 0, &n);
1403 task = get_proc_task(file_inode(file));
1407 put_task_struct(task);
1412 static ssize_t proc_fail_nth_read(struct file *file, char __user *buf,
1413 size_t count, loff_t *ppos)
1415 struct task_struct *task;
1416 char numbuf[PROC_NUMBUF];
1419 task = get_proc_task(file_inode(file));
1422 len = snprintf(numbuf, sizeof(numbuf), "%u\n", task->fail_nth);
1423 put_task_struct(task);
1424 return simple_read_from_buffer(buf, count, ppos, numbuf, len);
1427 static const struct file_operations proc_fail_nth_operations = {
1428 .read = proc_fail_nth_read,
1429 .write = proc_fail_nth_write,
1434 #ifdef CONFIG_SCHED_DEBUG
1436 * Print out various scheduling related per-task fields:
1438 static int sched_show(struct seq_file *m, void *v)
1440 struct inode *inode = m->private;
1441 struct pid_namespace *ns = proc_pid_ns(inode->i_sb);
1442 struct task_struct *p;
1444 p = get_proc_task(inode);
1447 proc_sched_show_task(p, ns, m);
1455 sched_write(struct file *file, const char __user *buf,
1456 size_t count, loff_t *offset)
1458 struct inode *inode = file_inode(file);
1459 struct task_struct *p;
1461 p = get_proc_task(inode);
1464 proc_sched_set_task(p);
1471 static int sched_open(struct inode *inode, struct file *filp)
1473 return single_open(filp, sched_show, inode);
1476 static const struct file_operations proc_pid_sched_operations = {
1479 .write = sched_write,
1480 .llseek = seq_lseek,
1481 .release = single_release,
1486 #ifdef CONFIG_SCHED_AUTOGROUP
1488 * Print out autogroup related information:
1490 static int sched_autogroup_show(struct seq_file *m, void *v)
1492 struct inode *inode = m->private;
1493 struct task_struct *p;
1495 p = get_proc_task(inode);
1498 proc_sched_autogroup_show_task(p, m);
1506 sched_autogroup_write(struct file *file, const char __user *buf,
1507 size_t count, loff_t *offset)
1509 struct inode *inode = file_inode(file);
1510 struct task_struct *p;
1511 char buffer[PROC_NUMBUF] = {};
1515 if (count > sizeof(buffer) - 1)
1516 count = sizeof(buffer) - 1;
1517 if (copy_from_user(buffer, buf, count))
1520 err = kstrtoint(strstrip(buffer), 0, &nice);
1524 p = get_proc_task(inode);
1528 err = proc_sched_autogroup_set_nice(p, nice);
1537 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1541 ret = single_open(filp, sched_autogroup_show, NULL);
1543 struct seq_file *m = filp->private_data;
1550 static const struct file_operations proc_pid_sched_autogroup_operations = {
1551 .open = sched_autogroup_open,
1553 .write = sched_autogroup_write,
1554 .llseek = seq_lseek,
1555 .release = single_release,
1558 #endif /* CONFIG_SCHED_AUTOGROUP */
1560 #ifdef CONFIG_TIME_NS
1561 static int timens_offsets_show(struct seq_file *m, void *v)
1563 struct task_struct *p;
1565 p = get_proc_task(file_inode(m->file));
1568 proc_timens_show_offsets(p, m);
1575 static ssize_t timens_offsets_write(struct file *file, const char __user *buf,
1576 size_t count, loff_t *ppos)
1578 struct inode *inode = file_inode(file);
1579 struct proc_timens_offset offsets[2];
1580 char *kbuf = NULL, *pos, *next_line;
1581 struct task_struct *p;
1584 /* Only allow < page size writes at the beginning of the file */
1585 if ((*ppos != 0) || (count >= PAGE_SIZE))
1588 /* Slurp in the user data */
1589 kbuf = memdup_user_nul(buf, count);
1591 return PTR_ERR(kbuf);
1593 /* Parse the user data */
1596 for (pos = kbuf; pos; pos = next_line) {
1597 struct proc_timens_offset *off = &offsets[noffsets];
1601 /* Find the end of line and ensure we don't look past it */
1602 next_line = strchr(pos, '\n');
1606 if (*next_line == '\0')
1610 err = sscanf(pos, "%9s %lld %lu", clock,
1611 &off->val.tv_sec, &off->val.tv_nsec);
1612 if (err != 3 || off->val.tv_nsec >= NSEC_PER_SEC)
1615 clock[sizeof(clock) - 1] = 0;
1616 if (strcmp(clock, "monotonic") == 0 ||
1617 strcmp(clock, __stringify(CLOCK_MONOTONIC)) == 0)
1618 off->clockid = CLOCK_MONOTONIC;
1619 else if (strcmp(clock, "boottime") == 0 ||
1620 strcmp(clock, __stringify(CLOCK_BOOTTIME)) == 0)
1621 off->clockid = CLOCK_BOOTTIME;
1626 if (noffsets == ARRAY_SIZE(offsets)) {
1628 count = next_line - kbuf;
1634 p = get_proc_task(inode);
1637 ret = proc_timens_set_offset(file, p, offsets, noffsets);
1648 static int timens_offsets_open(struct inode *inode, struct file *filp)
1650 return single_open(filp, timens_offsets_show, inode);
1653 static const struct file_operations proc_timens_offsets_operations = {
1654 .open = timens_offsets_open,
1656 .write = timens_offsets_write,
1657 .llseek = seq_lseek,
1658 .release = single_release,
1660 #endif /* CONFIG_TIME_NS */
1662 static ssize_t comm_write(struct file *file, const char __user *buf,
1663 size_t count, loff_t *offset)
1665 struct inode *inode = file_inode(file);
1666 struct task_struct *p;
1667 char buffer[TASK_COMM_LEN] = {};
1668 const size_t maxlen = sizeof(buffer) - 1;
1670 if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
1673 p = get_proc_task(inode);
1677 if (same_thread_group(current, p)) {
1678 set_task_comm(p, buffer);
1679 proc_comm_connector(p);
1689 static int comm_show(struct seq_file *m, void *v)
1691 struct inode *inode = m->private;
1692 struct task_struct *p;
1694 p = get_proc_task(inode);
1698 proc_task_name(m, p, false);
1706 static int comm_open(struct inode *inode, struct file *filp)
1708 return single_open(filp, comm_show, inode);
1711 static const struct file_operations proc_pid_set_comm_operations = {
1714 .write = comm_write,
1715 .llseek = seq_lseek,
1716 .release = single_release,
1719 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1721 struct task_struct *task;
1722 struct file *exe_file;
1724 task = get_proc_task(d_inode(dentry));
1727 exe_file = get_task_exe_file(task);
1728 put_task_struct(task);
1730 *exe_path = exe_file->f_path;
1731 path_get(&exe_file->f_path);
1738 static const char *proc_pid_get_link(struct dentry *dentry,
1739 struct inode *inode,
1740 struct delayed_call *done)
1743 int error = -EACCES;
1746 return ERR_PTR(-ECHILD);
1748 /* Are we allowed to snoop on the tasks file descriptors? */
1749 if (!proc_fd_access_allowed(inode))
1752 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1756 error = nd_jump_link(&path);
1758 return ERR_PTR(error);
1761 static int do_proc_readlink(const struct path *path, char __user *buffer, int buflen)
1763 char *tmp = kmalloc(PATH_MAX, GFP_KERNEL);
1770 pathname = d_path(path, tmp, PATH_MAX);
1771 len = PTR_ERR(pathname);
1772 if (IS_ERR(pathname))
1774 len = tmp + PATH_MAX - 1 - pathname;
1778 if (copy_to_user(buffer, pathname, len))
1785 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1787 int error = -EACCES;
1788 struct inode *inode = d_inode(dentry);
1791 /* Are we allowed to snoop on the tasks file descriptors? */
1792 if (!proc_fd_access_allowed(inode))
1795 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1799 error = do_proc_readlink(&path, buffer, buflen);
1805 const struct inode_operations proc_pid_link_inode_operations = {
1806 .readlink = proc_pid_readlink,
1807 .get_link = proc_pid_get_link,
1808 .setattr = proc_setattr,
1812 /* building an inode */
1814 void task_dump_owner(struct task_struct *task, umode_t mode,
1815 kuid_t *ruid, kgid_t *rgid)
1817 /* Depending on the state of dumpable compute who should own a
1818 * proc file for a task.
1820 const struct cred *cred;
1824 if (unlikely(task->flags & PF_KTHREAD)) {
1825 *ruid = GLOBAL_ROOT_UID;
1826 *rgid = GLOBAL_ROOT_GID;
1830 /* Default to the tasks effective ownership */
1832 cred = __task_cred(task);
1838 * Before the /proc/pid/status file was created the only way to read
1839 * the effective uid of a /process was to stat /proc/pid. Reading
1840 * /proc/pid/status is slow enough that procps and other packages
1841 * kept stating /proc/pid. To keep the rules in /proc simple I have
1842 * made this apply to all per process world readable and executable
1845 if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) {
1846 struct mm_struct *mm;
1849 /* Make non-dumpable tasks owned by some root */
1851 if (get_dumpable(mm) != SUID_DUMP_USER) {
1852 struct user_namespace *user_ns = mm->user_ns;
1854 uid = make_kuid(user_ns, 0);
1855 if (!uid_valid(uid))
1856 uid = GLOBAL_ROOT_UID;
1858 gid = make_kgid(user_ns, 0);
1859 if (!gid_valid(gid))
1860 gid = GLOBAL_ROOT_GID;
1863 uid = GLOBAL_ROOT_UID;
1864 gid = GLOBAL_ROOT_GID;
1872 void proc_pid_evict_inode(struct proc_inode *ei)
1874 struct pid *pid = ei->pid;
1876 if (S_ISDIR(ei->vfs_inode.i_mode)) {
1877 spin_lock(&pid->lock);
1878 hlist_del_init_rcu(&ei->sibling_inodes);
1879 spin_unlock(&pid->lock);
1885 struct inode *proc_pid_make_inode(struct super_block *sb,
1886 struct task_struct *task, umode_t mode)
1888 struct inode * inode;
1889 struct proc_inode *ei;
1892 /* We need a new inode */
1894 inode = new_inode(sb);
1900 inode->i_mode = mode;
1901 inode->i_ino = get_next_ino();
1902 simple_inode_init_ts(inode);
1903 inode->i_op = &proc_def_inode_operations;
1906 * grab the reference to task.
1908 pid = get_task_pid(task, PIDTYPE_PID);
1912 /* Let the pid remember us for quick removal */
1915 task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
1916 security_task_to_inode(task, inode);
1927 * Generating an inode and adding it into @pid->inodes, so that task will
1928 * invalidate inode's dentry before being released.
1930 * This helper is used for creating dir-type entries under '/proc' and
1931 * '/proc/<tgid>/task'. Other entries(eg. fd, stat) under '/proc/<tgid>'
1932 * can be released by invalidating '/proc/<tgid>' dentry.
1933 * In theory, dentries under '/proc/<tgid>/task' can also be released by
1934 * invalidating '/proc/<tgid>' dentry, we reserve it to handle single
1935 * thread exiting situation: Any one of threads should invalidate its
1936 * '/proc/<tgid>/task/<pid>' dentry before released.
1938 static struct inode *proc_pid_make_base_inode(struct super_block *sb,
1939 struct task_struct *task, umode_t mode)
1941 struct inode *inode;
1942 struct proc_inode *ei;
1945 inode = proc_pid_make_inode(sb, task, mode);
1949 /* Let proc_flush_pid find this directory inode */
1952 spin_lock(&pid->lock);
1953 hlist_add_head_rcu(&ei->sibling_inodes, &pid->inodes);
1954 spin_unlock(&pid->lock);
1959 int pid_getattr(struct mnt_idmap *idmap, const struct path *path,
1960 struct kstat *stat, u32 request_mask, unsigned int query_flags)
1962 struct inode *inode = d_inode(path->dentry);
1963 struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
1964 struct task_struct *task;
1966 generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat);
1968 stat->uid = GLOBAL_ROOT_UID;
1969 stat->gid = GLOBAL_ROOT_GID;
1971 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1973 if (!has_pid_permissions(fs_info, task, HIDEPID_INVISIBLE)) {
1976 * This doesn't prevent learning whether PID exists,
1977 * it only makes getattr() consistent with readdir().
1981 task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid);
1990 * Set <pid>/... inode ownership (can change due to setuid(), etc.)
1992 void pid_update_inode(struct task_struct *task, struct inode *inode)
1994 task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid);
1996 inode->i_mode &= ~(S_ISUID | S_ISGID);
1997 security_task_to_inode(task, inode);
2001 * Rewrite the inode's ownerships here because the owning task may have
2002 * performed a setuid(), etc.
2005 static int pid_revalidate(struct dentry *dentry, unsigned int flags)
2007 struct inode *inode;
2008 struct task_struct *task;
2012 inode = d_inode_rcu(dentry);
2015 task = pid_task(proc_pid(inode), PIDTYPE_PID);
2018 pid_update_inode(task, inode);
2026 static inline bool proc_inode_is_dead(struct inode *inode)
2028 return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
2031 int pid_delete_dentry(const struct dentry *dentry)
2033 /* Is the task we represent dead?
2034 * If so, then don't put the dentry on the lru list,
2035 * kill it immediately.
2037 return proc_inode_is_dead(d_inode(dentry));
2040 const struct dentry_operations pid_dentry_operations =
2042 .d_revalidate = pid_revalidate,
2043 .d_delete = pid_delete_dentry,
2049 * Fill a directory entry.
2051 * If possible create the dcache entry and derive our inode number and
2052 * file type from dcache entry.
2054 * Since all of the proc inode numbers are dynamically generated, the inode
2055 * numbers do not exist until the inode is cache. This means creating
2056 * the dcache entry in readdir is necessary to keep the inode numbers
2057 * reported by readdir in sync with the inode numbers reported
2060 bool proc_fill_cache(struct file *file, struct dir_context *ctx,
2061 const char *name, unsigned int len,
2062 instantiate_t instantiate, struct task_struct *task, const void *ptr)
2064 struct dentry *child, *dir = file->f_path.dentry;
2065 struct qstr qname = QSTR_INIT(name, len);
2066 struct inode *inode;
2067 unsigned type = DT_UNKNOWN;
2070 child = d_hash_and_lookup(dir, &qname);
2072 DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq);
2073 child = d_alloc_parallel(dir, &qname, &wq);
2075 goto end_instantiate;
2076 if (d_in_lookup(child)) {
2078 res = instantiate(child, task, ptr);
2079 d_lookup_done(child);
2080 if (unlikely(res)) {
2084 goto end_instantiate;
2088 inode = d_inode(child);
2090 type = inode->i_mode >> 12;
2093 return dir_emit(ctx, name, len, ino, type);
2097 * dname_to_vma_addr - maps a dentry name into two unsigned longs
2098 * which represent vma start and end addresses.
2100 static int dname_to_vma_addr(struct dentry *dentry,
2101 unsigned long *start, unsigned long *end)
2103 const char *str = dentry->d_name.name;
2104 unsigned long long sval, eval;
2107 if (str[0] == '0' && str[1] != '-')
2109 len = _parse_integer(str, 16, &sval);
2110 if (len & KSTRTOX_OVERFLOW)
2112 if (sval != (unsigned long)sval)
2120 if (str[0] == '0' && str[1])
2122 len = _parse_integer(str, 16, &eval);
2123 if (len & KSTRTOX_OVERFLOW)
2125 if (eval != (unsigned long)eval)
2138 static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
2140 unsigned long vm_start, vm_end;
2141 bool exact_vma_exists = false;
2142 struct mm_struct *mm = NULL;
2143 struct task_struct *task;
2144 struct inode *inode;
2147 if (flags & LOOKUP_RCU)
2150 inode = d_inode(dentry);
2151 task = get_proc_task(inode);
2155 mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
2156 if (IS_ERR_OR_NULL(mm))
2159 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
2160 status = mmap_read_lock_killable(mm);
2162 exact_vma_exists = !!find_exact_vma(mm, vm_start,
2164 mmap_read_unlock(mm);
2170 if (exact_vma_exists) {
2171 task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
2173 security_task_to_inode(task, inode);
2178 put_task_struct(task);
2184 static const struct dentry_operations tid_map_files_dentry_operations = {
2185 .d_revalidate = map_files_d_revalidate,
2186 .d_delete = pid_delete_dentry,
2189 static int map_files_get_link(struct dentry *dentry, struct path *path)
2191 unsigned long vm_start, vm_end;
2192 struct vm_area_struct *vma;
2193 struct task_struct *task;
2194 struct mm_struct *mm;
2198 task = get_proc_task(d_inode(dentry));
2202 mm = get_task_mm(task);
2203 put_task_struct(task);
2207 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2211 rc = mmap_read_lock_killable(mm);
2216 vma = find_exact_vma(mm, vm_start, vm_end);
2217 if (vma && vma->vm_file) {
2218 *path = *file_user_path(vma->vm_file);
2222 mmap_read_unlock(mm);
2230 struct map_files_info {
2231 unsigned long start;
2237 * Only allow CAP_SYS_ADMIN and CAP_CHECKPOINT_RESTORE to follow the links, due
2238 * to concerns about how the symlinks may be used to bypass permissions on
2239 * ancestor directories in the path to the file in question.
2242 proc_map_files_get_link(struct dentry *dentry,
2243 struct inode *inode,
2244 struct delayed_call *done)
2246 if (!checkpoint_restore_ns_capable(&init_user_ns))
2247 return ERR_PTR(-EPERM);
2249 return proc_pid_get_link(dentry, inode, done);
2253 * Identical to proc_pid_link_inode_operations except for get_link()
2255 static const struct inode_operations proc_map_files_link_inode_operations = {
2256 .readlink = proc_pid_readlink,
2257 .get_link = proc_map_files_get_link,
2258 .setattr = proc_setattr,
2261 static struct dentry *
2262 proc_map_files_instantiate(struct dentry *dentry,
2263 struct task_struct *task, const void *ptr)
2265 fmode_t mode = (fmode_t)(unsigned long)ptr;
2266 struct proc_inode *ei;
2267 struct inode *inode;
2269 inode = proc_pid_make_inode(dentry->d_sb, task, S_IFLNK |
2270 ((mode & FMODE_READ ) ? S_IRUSR : 0) |
2271 ((mode & FMODE_WRITE) ? S_IWUSR : 0));
2273 return ERR_PTR(-ENOENT);
2276 ei->op.proc_get_link = map_files_get_link;
2278 inode->i_op = &proc_map_files_link_inode_operations;
2281 d_set_d_op(dentry, &tid_map_files_dentry_operations);
2282 return d_splice_alias(inode, dentry);
2285 static struct dentry *proc_map_files_lookup(struct inode *dir,
2286 struct dentry *dentry, unsigned int flags)
2288 unsigned long vm_start, vm_end;
2289 struct vm_area_struct *vma;
2290 struct task_struct *task;
2291 struct dentry *result;
2292 struct mm_struct *mm;
2294 result = ERR_PTR(-ENOENT);
2295 task = get_proc_task(dir);
2299 result = ERR_PTR(-EACCES);
2300 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2303 result = ERR_PTR(-ENOENT);
2304 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2307 mm = get_task_mm(task);
2311 result = ERR_PTR(-EINTR);
2312 if (mmap_read_lock_killable(mm))
2315 result = ERR_PTR(-ENOENT);
2316 vma = find_exact_vma(mm, vm_start, vm_end);
2321 result = proc_map_files_instantiate(dentry, task,
2322 (void *)(unsigned long)vma->vm_file->f_mode);
2325 mmap_read_unlock(mm);
2329 put_task_struct(task);
2334 static const struct inode_operations proc_map_files_inode_operations = {
2335 .lookup = proc_map_files_lookup,
2336 .permission = proc_fd_permission,
2337 .setattr = proc_setattr,
2341 proc_map_files_readdir(struct file *file, struct dir_context *ctx)
2343 struct vm_area_struct *vma;
2344 struct task_struct *task;
2345 struct mm_struct *mm;
2346 unsigned long nr_files, pos, i;
2347 GENRADIX(struct map_files_info) fa;
2348 struct map_files_info *p;
2350 struct vma_iterator vmi;
2355 task = get_proc_task(file_inode(file));
2360 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2364 if (!dir_emit_dots(file, ctx))
2367 mm = get_task_mm(task);
2371 ret = mmap_read_lock_killable(mm);
2380 * We need two passes here:
2382 * 1) Collect vmas of mapped files with mmap_lock taken
2383 * 2) Release mmap_lock and instantiate entries
2385 * otherwise we get lockdep complained, since filldir()
2386 * routine might require mmap_lock taken in might_fault().
2390 vma_iter_init(&vmi, mm, 0);
2391 for_each_vma(vmi, vma) {
2394 if (++pos <= ctx->pos)
2397 p = genradix_ptr_alloc(&fa, nr_files++, GFP_KERNEL);
2400 mmap_read_unlock(mm);
2405 p->start = vma->vm_start;
2406 p->end = vma->vm_end;
2407 p->mode = vma->vm_file->f_mode;
2409 mmap_read_unlock(mm);
2412 for (i = 0; i < nr_files; i++) {
2413 char buf[4 * sizeof(long) + 2]; /* max: %lx-%lx\0 */
2416 p = genradix_ptr(&fa, i);
2417 len = snprintf(buf, sizeof(buf), "%lx-%lx", p->start, p->end);
2418 if (!proc_fill_cache(file, ctx,
2420 proc_map_files_instantiate,
2422 (void *)(unsigned long)p->mode))
2428 put_task_struct(task);
2434 static const struct file_operations proc_map_files_operations = {
2435 .read = generic_read_dir,
2436 .iterate_shared = proc_map_files_readdir,
2437 .llseek = generic_file_llseek,
2440 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
2441 struct timers_private {
2443 struct task_struct *task;
2444 struct sighand_struct *sighand;
2445 struct pid_namespace *ns;
2446 unsigned long flags;
2449 static void *timers_start(struct seq_file *m, loff_t *pos)
2451 struct timers_private *tp = m->private;
2453 tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2455 return ERR_PTR(-ESRCH);
2457 tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2459 return ERR_PTR(-ESRCH);
2461 return seq_list_start(&tp->task->signal->posix_timers, *pos);
2464 static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2466 struct timers_private *tp = m->private;
2467 return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2470 static void timers_stop(struct seq_file *m, void *v)
2472 struct timers_private *tp = m->private;
2475 unlock_task_sighand(tp->task, &tp->flags);
2480 put_task_struct(tp->task);
2485 static int show_timer(struct seq_file *m, void *v)
2487 struct k_itimer *timer;
2488 struct timers_private *tp = m->private;
2490 static const char * const nstr[] = {
2491 [SIGEV_SIGNAL] = "signal",
2492 [SIGEV_NONE] = "none",
2493 [SIGEV_THREAD] = "thread",
2496 timer = list_entry((struct list_head *)v, struct k_itimer, list);
2497 notify = timer->it_sigev_notify;
2499 seq_printf(m, "ID: %d\n", timer->it_id);
2500 seq_printf(m, "signal: %d/%px\n",
2501 timer->sigq->info.si_signo,
2502 timer->sigq->info.si_value.sival_ptr);
2503 seq_printf(m, "notify: %s/%s.%d\n",
2504 nstr[notify & ~SIGEV_THREAD_ID],
2505 (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2506 pid_nr_ns(timer->it_pid, tp->ns));
2507 seq_printf(m, "ClockID: %d\n", timer->it_clock);
2512 static const struct seq_operations proc_timers_seq_ops = {
2513 .start = timers_start,
2514 .next = timers_next,
2515 .stop = timers_stop,
2519 static int proc_timers_open(struct inode *inode, struct file *file)
2521 struct timers_private *tp;
2523 tp = __seq_open_private(file, &proc_timers_seq_ops,
2524 sizeof(struct timers_private));
2528 tp->pid = proc_pid(inode);
2529 tp->ns = proc_pid_ns(inode->i_sb);
2533 static const struct file_operations proc_timers_operations = {
2534 .open = proc_timers_open,
2536 .llseek = seq_lseek,
2537 .release = seq_release_private,
2541 static ssize_t timerslack_ns_write(struct file *file, const char __user *buf,
2542 size_t count, loff_t *offset)
2544 struct inode *inode = file_inode(file);
2545 struct task_struct *p;
2549 err = kstrtoull_from_user(buf, count, 10, &slack_ns);
2553 p = get_proc_task(inode);
2559 if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
2566 err = security_task_setscheduler(p);
2575 p->timer_slack_ns = p->default_timer_slack_ns;
2577 p->timer_slack_ns = slack_ns;
2586 static int timerslack_ns_show(struct seq_file *m, void *v)
2588 struct inode *inode = m->private;
2589 struct task_struct *p;
2592 p = get_proc_task(inode);
2598 if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
2605 err = security_task_getscheduler(p);
2611 seq_printf(m, "%llu\n", p->timer_slack_ns);
2620 static int timerslack_ns_open(struct inode *inode, struct file *filp)
2622 return single_open(filp, timerslack_ns_show, inode);
2625 static const struct file_operations proc_pid_set_timerslack_ns_operations = {
2626 .open = timerslack_ns_open,
2628 .write = timerslack_ns_write,
2629 .llseek = seq_lseek,
2630 .release = single_release,
2633 static struct dentry *proc_pident_instantiate(struct dentry *dentry,
2634 struct task_struct *task, const void *ptr)
2636 const struct pid_entry *p = ptr;
2637 struct inode *inode;
2638 struct proc_inode *ei;
2640 inode = proc_pid_make_inode(dentry->d_sb, task, p->mode);
2642 return ERR_PTR(-ENOENT);
2645 if (S_ISDIR(inode->i_mode))
2646 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2648 inode->i_op = p->iop;
2650 inode->i_fop = p->fop;
2652 pid_update_inode(task, inode);
2653 d_set_d_op(dentry, &pid_dentry_operations);
2654 return d_splice_alias(inode, dentry);
2657 static struct dentry *proc_pident_lookup(struct inode *dir,
2658 struct dentry *dentry,
2659 const struct pid_entry *p,
2660 const struct pid_entry *end)
2662 struct task_struct *task = get_proc_task(dir);
2663 struct dentry *res = ERR_PTR(-ENOENT);
2669 * Yes, it does not scale. And it should not. Don't add
2670 * new entries into /proc/<tgid>/ without very good reasons.
2672 for (; p < end; p++) {
2673 if (p->len != dentry->d_name.len)
2675 if (!memcmp(dentry->d_name.name, p->name, p->len)) {
2676 res = proc_pident_instantiate(dentry, task, p);
2680 put_task_struct(task);
2685 static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
2686 const struct pid_entry *ents, unsigned int nents)
2688 struct task_struct *task = get_proc_task(file_inode(file));
2689 const struct pid_entry *p;
2694 if (!dir_emit_dots(file, ctx))
2697 if (ctx->pos >= nents + 2)
2700 for (p = ents + (ctx->pos - 2); p < ents + nents; p++) {
2701 if (!proc_fill_cache(file, ctx, p->name, p->len,
2702 proc_pident_instantiate, task, p))
2707 put_task_struct(task);
2711 #ifdef CONFIG_SECURITY
2712 static int proc_pid_attr_open(struct inode *inode, struct file *file)
2714 file->private_data = NULL;
2715 __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
2719 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2720 size_t count, loff_t *ppos)
2722 struct inode * inode = file_inode(file);
2725 struct task_struct *task = get_proc_task(inode);
2730 length = security_getprocattr(task, PROC_I(inode)->op.lsmid,
2731 file->f_path.dentry->d_name.name,
2733 put_task_struct(task);
2735 length = simple_read_from_buffer(buf, count, ppos, p, length);
2740 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2741 size_t count, loff_t *ppos)
2743 struct inode * inode = file_inode(file);
2744 struct task_struct *task;
2748 /* A task may only write when it was the opener. */
2749 if (file->private_data != current->mm)
2753 task = pid_task(proc_pid(inode), PIDTYPE_PID);
2758 /* A task may only write its own attributes. */
2759 if (current != task) {
2763 /* Prevent changes to overridden credentials. */
2764 if (current_cred() != current_real_cred()) {
2770 if (count > PAGE_SIZE)
2773 /* No partial writes. */
2777 page = memdup_user(buf, count);
2783 /* Guard against adverse ptrace interaction */
2784 rv = mutex_lock_interruptible(¤t->signal->cred_guard_mutex);
2788 rv = security_setprocattr(PROC_I(inode)->op.lsmid,
2789 file->f_path.dentry->d_name.name, page,
2791 mutex_unlock(¤t->signal->cred_guard_mutex);
2798 static const struct file_operations proc_pid_attr_operations = {
2799 .open = proc_pid_attr_open,
2800 .read = proc_pid_attr_read,
2801 .write = proc_pid_attr_write,
2802 .llseek = generic_file_llseek,
2803 .release = mem_release,
2806 #define LSM_DIR_OPS(LSM) \
2807 static int proc_##LSM##_attr_dir_iterate(struct file *filp, \
2808 struct dir_context *ctx) \
2810 return proc_pident_readdir(filp, ctx, \
2811 LSM##_attr_dir_stuff, \
2812 ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2815 static const struct file_operations proc_##LSM##_attr_dir_ops = { \
2816 .read = generic_read_dir, \
2817 .iterate_shared = proc_##LSM##_attr_dir_iterate, \
2818 .llseek = default_llseek, \
2821 static struct dentry *proc_##LSM##_attr_dir_lookup(struct inode *dir, \
2822 struct dentry *dentry, unsigned int flags) \
2824 return proc_pident_lookup(dir, dentry, \
2825 LSM##_attr_dir_stuff, \
2826 LSM##_attr_dir_stuff + ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2829 static const struct inode_operations proc_##LSM##_attr_dir_inode_ops = { \
2830 .lookup = proc_##LSM##_attr_dir_lookup, \
2831 .getattr = pid_getattr, \
2832 .setattr = proc_setattr, \
2835 #ifdef CONFIG_SECURITY_SMACK
2836 static const struct pid_entry smack_attr_dir_stuff[] = {
2837 ATTR(LSM_ID_SMACK, "current", 0666),
2842 #ifdef CONFIG_SECURITY_APPARMOR
2843 static const struct pid_entry apparmor_attr_dir_stuff[] = {
2844 ATTR(LSM_ID_APPARMOR, "current", 0666),
2845 ATTR(LSM_ID_APPARMOR, "prev", 0444),
2846 ATTR(LSM_ID_APPARMOR, "exec", 0666),
2848 LSM_DIR_OPS(apparmor);
2851 static const struct pid_entry attr_dir_stuff[] = {
2852 ATTR(LSM_ID_UNDEF, "current", 0666),
2853 ATTR(LSM_ID_UNDEF, "prev", 0444),
2854 ATTR(LSM_ID_UNDEF, "exec", 0666),
2855 ATTR(LSM_ID_UNDEF, "fscreate", 0666),
2856 ATTR(LSM_ID_UNDEF, "keycreate", 0666),
2857 ATTR(LSM_ID_UNDEF, "sockcreate", 0666),
2858 #ifdef CONFIG_SECURITY_SMACK
2860 proc_smack_attr_dir_inode_ops, proc_smack_attr_dir_ops),
2862 #ifdef CONFIG_SECURITY_APPARMOR
2863 DIR("apparmor", 0555,
2864 proc_apparmor_attr_dir_inode_ops, proc_apparmor_attr_dir_ops),
2868 static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
2870 return proc_pident_readdir(file, ctx,
2871 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2874 static const struct file_operations proc_attr_dir_operations = {
2875 .read = generic_read_dir,
2876 .iterate_shared = proc_attr_dir_readdir,
2877 .llseek = generic_file_llseek,
2880 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2881 struct dentry *dentry, unsigned int flags)
2883 return proc_pident_lookup(dir, dentry,
2885 attr_dir_stuff + ARRAY_SIZE(attr_dir_stuff));
2888 static const struct inode_operations proc_attr_dir_inode_operations = {
2889 .lookup = proc_attr_dir_lookup,
2890 .getattr = pid_getattr,
2891 .setattr = proc_setattr,
2896 #ifdef CONFIG_ELF_CORE
2897 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2898 size_t count, loff_t *ppos)
2900 struct task_struct *task = get_proc_task(file_inode(file));
2901 struct mm_struct *mm;
2902 char buffer[PROC_NUMBUF];
2910 mm = get_task_mm(task);
2912 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2913 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2914 MMF_DUMP_FILTER_SHIFT));
2916 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2919 put_task_struct(task);
2924 static ssize_t proc_coredump_filter_write(struct file *file,
2925 const char __user *buf,
2929 struct task_struct *task;
2930 struct mm_struct *mm;
2936 ret = kstrtouint_from_user(buf, count, 0, &val);
2941 task = get_proc_task(file_inode(file));
2945 mm = get_task_mm(task);
2950 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2952 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2954 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2959 put_task_struct(task);
2966 static const struct file_operations proc_coredump_filter_operations = {
2967 .read = proc_coredump_filter_read,
2968 .write = proc_coredump_filter_write,
2969 .llseek = generic_file_llseek,
2973 #ifdef CONFIG_TASK_IO_ACCOUNTING
2974 static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
2976 struct task_io_accounting acct;
2979 result = down_read_killable(&task->signal->exec_update_lock);
2983 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) {
2989 struct signal_struct *sig = task->signal;
2990 struct task_struct *t;
2991 unsigned int seq = 1;
2992 unsigned long flags;
2996 seq++; /* 2 on the 1st/lockless path, otherwise odd */
2997 flags = read_seqbegin_or_lock_irqsave(&sig->stats_lock, &seq);
3000 __for_each_thread(sig, t)
3001 task_io_accounting_add(&acct, &t->ioac);
3003 } while (need_seqretry(&sig->stats_lock, seq));
3004 done_seqretry_irqrestore(&sig->stats_lock, seq, flags);
3015 "read_bytes: %llu\n"
3016 "write_bytes: %llu\n"
3017 "cancelled_write_bytes: %llu\n",
3018 (unsigned long long)acct.rchar,
3019 (unsigned long long)acct.wchar,
3020 (unsigned long long)acct.syscr,
3021 (unsigned long long)acct.syscw,
3022 (unsigned long long)acct.read_bytes,
3023 (unsigned long long)acct.write_bytes,
3024 (unsigned long long)acct.cancelled_write_bytes);
3028 up_read(&task->signal->exec_update_lock);
3032 static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
3033 struct pid *pid, struct task_struct *task)
3035 return do_io_accounting(task, m, 0);
3038 static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
3039 struct pid *pid, struct task_struct *task)
3041 return do_io_accounting(task, m, 1);
3043 #endif /* CONFIG_TASK_IO_ACCOUNTING */
3045 #ifdef CONFIG_USER_NS
3046 static int proc_id_map_open(struct inode *inode, struct file *file,
3047 const struct seq_operations *seq_ops)
3049 struct user_namespace *ns = NULL;
3050 struct task_struct *task;
3051 struct seq_file *seq;
3054 task = get_proc_task(inode);
3057 ns = get_user_ns(task_cred_xxx(task, user_ns));
3059 put_task_struct(task);
3064 ret = seq_open(file, seq_ops);
3068 seq = file->private_data;
3078 static int proc_id_map_release(struct inode *inode, struct file *file)
3080 struct seq_file *seq = file->private_data;
3081 struct user_namespace *ns = seq->private;
3083 return seq_release(inode, file);
3086 static int proc_uid_map_open(struct inode *inode, struct file *file)
3088 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
3091 static int proc_gid_map_open(struct inode *inode, struct file *file)
3093 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
3096 static int proc_projid_map_open(struct inode *inode, struct file *file)
3098 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
3101 static const struct file_operations proc_uid_map_operations = {
3102 .open = proc_uid_map_open,
3103 .write = proc_uid_map_write,
3105 .llseek = seq_lseek,
3106 .release = proc_id_map_release,
3109 static const struct file_operations proc_gid_map_operations = {
3110 .open = proc_gid_map_open,
3111 .write = proc_gid_map_write,
3113 .llseek = seq_lseek,
3114 .release = proc_id_map_release,
3117 static const struct file_operations proc_projid_map_operations = {
3118 .open = proc_projid_map_open,
3119 .write = proc_projid_map_write,
3121 .llseek = seq_lseek,
3122 .release = proc_id_map_release,
3125 static int proc_setgroups_open(struct inode *inode, struct file *file)
3127 struct user_namespace *ns = NULL;
3128 struct task_struct *task;
3132 task = get_proc_task(inode);
3135 ns = get_user_ns(task_cred_xxx(task, user_ns));
3137 put_task_struct(task);
3142 if (file->f_mode & FMODE_WRITE) {
3144 if (!ns_capable(ns, CAP_SYS_ADMIN))
3148 ret = single_open(file, &proc_setgroups_show, ns);
3159 static int proc_setgroups_release(struct inode *inode, struct file *file)
3161 struct seq_file *seq = file->private_data;
3162 struct user_namespace *ns = seq->private;
3163 int ret = single_release(inode, file);
3168 static const struct file_operations proc_setgroups_operations = {
3169 .open = proc_setgroups_open,
3170 .write = proc_setgroups_write,
3172 .llseek = seq_lseek,
3173 .release = proc_setgroups_release,
3175 #endif /* CONFIG_USER_NS */
3177 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
3178 struct pid *pid, struct task_struct *task)
3180 int err = lock_trace(task);
3182 seq_printf(m, "%08x\n", task->personality);
3188 #ifdef CONFIG_LIVEPATCH
3189 static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns,
3190 struct pid *pid, struct task_struct *task)
3192 seq_printf(m, "%d\n", task->patch_state);
3195 #endif /* CONFIG_LIVEPATCH */
3198 static int proc_pid_ksm_merging_pages(struct seq_file *m, struct pid_namespace *ns,
3199 struct pid *pid, struct task_struct *task)
3201 struct mm_struct *mm;
3203 mm = get_task_mm(task);
3205 seq_printf(m, "%lu\n", mm->ksm_merging_pages);
3211 static int proc_pid_ksm_stat(struct seq_file *m, struct pid_namespace *ns,
3212 struct pid *pid, struct task_struct *task)
3214 struct mm_struct *mm;
3216 mm = get_task_mm(task);
3218 seq_printf(m, "ksm_rmap_items %lu\n", mm->ksm_rmap_items);
3219 seq_printf(m, "ksm_zero_pages %lu\n", mm->ksm_zero_pages);
3220 seq_printf(m, "ksm_merging_pages %lu\n", mm->ksm_merging_pages);
3221 seq_printf(m, "ksm_process_profit %ld\n", ksm_process_profit(mm));
3227 #endif /* CONFIG_KSM */
3229 #ifdef CONFIG_STACKLEAK_METRICS
3230 static int proc_stack_depth(struct seq_file *m, struct pid_namespace *ns,
3231 struct pid *pid, struct task_struct *task)
3233 unsigned long prev_depth = THREAD_SIZE -
3234 (task->prev_lowest_stack & (THREAD_SIZE - 1));
3235 unsigned long depth = THREAD_SIZE -
3236 (task->lowest_stack & (THREAD_SIZE - 1));
3238 seq_printf(m, "previous stack depth: %lu\nstack depth: %lu\n",
3242 #endif /* CONFIG_STACKLEAK_METRICS */
3247 static const struct file_operations proc_task_operations;
3248 static const struct inode_operations proc_task_inode_operations;
3250 static const struct pid_entry tgid_base_stuff[] = {
3251 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
3252 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3253 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
3254 DIR("fdinfo", S_IRUGO|S_IXUGO, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3255 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3257 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3259 REG("environ", S_IRUSR, proc_environ_operations),
3260 REG("auxv", S_IRUSR, proc_auxv_operations),
3261 ONE("status", S_IRUGO, proc_pid_status),
3262 ONE("personality", S_IRUSR, proc_pid_personality),
3263 ONE("limits", S_IRUGO, proc_pid_limits),
3264 #ifdef CONFIG_SCHED_DEBUG
3265 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3267 #ifdef CONFIG_SCHED_AUTOGROUP
3268 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
3270 #ifdef CONFIG_TIME_NS
3271 REG("timens_offsets", S_IRUGO|S_IWUSR, proc_timens_offsets_operations),
3273 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3274 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3275 ONE("syscall", S_IRUSR, proc_pid_syscall),
3277 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
3278 ONE("stat", S_IRUGO, proc_tgid_stat),
3279 ONE("statm", S_IRUGO, proc_pid_statm),
3280 REG("maps", S_IRUGO, proc_pid_maps_operations),
3282 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
3284 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3285 LNK("cwd", proc_cwd_link),
3286 LNK("root", proc_root_link),
3287 LNK("exe", proc_exe_link),
3288 REG("mounts", S_IRUGO, proc_mounts_operations),
3289 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3290 REG("mountstats", S_IRUSR, proc_mountstats_operations),
3291 #ifdef CONFIG_PROC_PAGE_MONITOR
3292 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3293 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
3294 REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
3295 REG("pagemap", S_IRUSR, proc_pagemap_operations),
3297 #ifdef CONFIG_SECURITY
3298 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3300 #ifdef CONFIG_KALLSYMS
3301 ONE("wchan", S_IRUGO, proc_pid_wchan),
3303 #ifdef CONFIG_STACKTRACE
3304 ONE("stack", S_IRUSR, proc_pid_stack),
3306 #ifdef CONFIG_SCHED_INFO
3307 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
3309 #ifdef CONFIG_LATENCYTOP
3310 REG("latency", S_IRUGO, proc_lstats_operations),
3312 #ifdef CONFIG_PROC_PID_CPUSET
3313 ONE("cpuset", S_IRUGO, proc_cpuset_show),
3315 #ifdef CONFIG_CGROUPS
3316 ONE("cgroup", S_IRUGO, proc_cgroup_show),
3318 #ifdef CONFIG_PROC_CPU_RESCTRL
3319 ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show),
3321 ONE("oom_score", S_IRUGO, proc_oom_score),
3322 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3323 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3325 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3326 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3328 #ifdef CONFIG_FAULT_INJECTION
3329 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3330 REG("fail-nth", 0644, proc_fail_nth_operations),
3332 #ifdef CONFIG_ELF_CORE
3333 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3335 #ifdef CONFIG_TASK_IO_ACCOUNTING
3336 ONE("io", S_IRUSR, proc_tgid_io_accounting),
3338 #ifdef CONFIG_USER_NS
3339 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3340 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3341 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3342 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
3344 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
3345 REG("timers", S_IRUGO, proc_timers_operations),
3347 REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations),
3348 #ifdef CONFIG_LIVEPATCH
3349 ONE("patch_state", S_IRUSR, proc_pid_patch_state),
3351 #ifdef CONFIG_STACKLEAK_METRICS
3352 ONE("stack_depth", S_IRUGO, proc_stack_depth),
3354 #ifdef CONFIG_PROC_PID_ARCH_STATUS
3355 ONE("arch_status", S_IRUGO, proc_pid_arch_status),
3357 #ifdef CONFIG_SECCOMP_CACHE_DEBUG
3358 ONE("seccomp_cache", S_IRUSR, proc_pid_seccomp_cache),
3361 ONE("ksm_merging_pages", S_IRUSR, proc_pid_ksm_merging_pages),
3362 ONE("ksm_stat", S_IRUSR, proc_pid_ksm_stat),
3366 static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
3368 return proc_pident_readdir(file, ctx,
3369 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3372 static const struct file_operations proc_tgid_base_operations = {
3373 .read = generic_read_dir,
3374 .iterate_shared = proc_tgid_base_readdir,
3375 .llseek = generic_file_llseek,
3378 struct pid *tgid_pidfd_to_pid(const struct file *file)
3380 if (file->f_op != &proc_tgid_base_operations)
3381 return ERR_PTR(-EBADF);
3383 return proc_pid(file_inode(file));
3386 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3388 return proc_pident_lookup(dir, dentry,
3390 tgid_base_stuff + ARRAY_SIZE(tgid_base_stuff));
3393 static const struct inode_operations proc_tgid_base_inode_operations = {
3394 .lookup = proc_tgid_base_lookup,
3395 .getattr = pid_getattr,
3396 .setattr = proc_setattr,
3397 .permission = proc_pid_permission,
3401 * proc_flush_pid - Remove dcache entries for @pid from the /proc dcache.
3402 * @pid: pid that should be flushed.
3404 * This function walks a list of inodes (that belong to any proc
3405 * filesystem) that are attached to the pid and flushes them from
3408 * It is safe and reasonable to cache /proc entries for a task until
3409 * that task exits. After that they just clog up the dcache with
3410 * useless entries, possibly causing useful dcache entries to be
3411 * flushed instead. This routine is provided to flush those useless
3412 * dcache entries when a process is reaped.
3414 * NOTE: This routine is just an optimization so it does not guarantee
3415 * that no dcache entries will exist after a process is reaped
3416 * it just makes it very unlikely that any will persist.
3419 void proc_flush_pid(struct pid *pid)
3421 proc_invalidate_siblings_dcache(&pid->inodes, &pid->lock);
3424 static struct dentry *proc_pid_instantiate(struct dentry * dentry,
3425 struct task_struct *task, const void *ptr)
3427 struct inode *inode;
3429 inode = proc_pid_make_base_inode(dentry->d_sb, task,
3430 S_IFDIR | S_IRUGO | S_IXUGO);
3432 return ERR_PTR(-ENOENT);
3434 inode->i_op = &proc_tgid_base_inode_operations;
3435 inode->i_fop = &proc_tgid_base_operations;
3436 inode->i_flags|=S_IMMUTABLE;
3438 set_nlink(inode, nlink_tgid);
3439 pid_update_inode(task, inode);
3441 d_set_d_op(dentry, &pid_dentry_operations);
3442 return d_splice_alias(inode, dentry);
3445 struct dentry *proc_pid_lookup(struct dentry *dentry, unsigned int flags)
3447 struct task_struct *task;
3449 struct proc_fs_info *fs_info;
3450 struct pid_namespace *ns;
3451 struct dentry *result = ERR_PTR(-ENOENT);
3453 tgid = name_to_int(&dentry->d_name);
3457 fs_info = proc_sb_info(dentry->d_sb);
3458 ns = fs_info->pid_ns;
3460 task = find_task_by_pid_ns(tgid, ns);
3462 get_task_struct(task);
3467 /* Limit procfs to only ptraceable tasks */
3468 if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE) {
3469 if (!has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS))
3473 result = proc_pid_instantiate(dentry, task, NULL);
3475 put_task_struct(task);
3481 * Find the first task with tgid >= tgid
3486 struct task_struct *task;
3488 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3493 put_task_struct(iter.task);
3497 pid = find_ge_pid(iter.tgid, ns);
3499 iter.tgid = pid_nr_ns(pid, ns);
3500 iter.task = pid_task(pid, PIDTYPE_TGID);
3505 get_task_struct(iter.task);
3511 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3513 /* for the /proc/ directory itself, after non-process stuff has been done */
3514 int proc_pid_readdir(struct file *file, struct dir_context *ctx)
3516 struct tgid_iter iter;
3517 struct proc_fs_info *fs_info = proc_sb_info(file_inode(file)->i_sb);
3518 struct pid_namespace *ns = proc_pid_ns(file_inode(file)->i_sb);
3519 loff_t pos = ctx->pos;
3521 if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
3524 if (pos == TGID_OFFSET - 2) {
3525 struct inode *inode = d_inode(fs_info->proc_self);
3526 if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
3528 ctx->pos = pos = pos + 1;
3530 if (pos == TGID_OFFSET - 1) {
3531 struct inode *inode = d_inode(fs_info->proc_thread_self);
3532 if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3534 ctx->pos = pos = pos + 1;
3536 iter.tgid = pos - TGID_OFFSET;
3538 for (iter = next_tgid(ns, iter);
3540 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3545 if (!has_pid_permissions(fs_info, iter.task, HIDEPID_INVISIBLE))
3548 len = snprintf(name, sizeof(name), "%u", iter.tgid);
3549 ctx->pos = iter.tgid + TGID_OFFSET;
3550 if (!proc_fill_cache(file, ctx, name, len,
3551 proc_pid_instantiate, iter.task, NULL)) {
3552 put_task_struct(iter.task);
3556 ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
3561 * proc_tid_comm_permission is a special permission function exclusively
3562 * used for the node /proc/<pid>/task/<tid>/comm.
3563 * It bypasses generic permission checks in the case where a task of the same
3564 * task group attempts to access the node.
3565 * The rationale behind this is that glibc and bionic access this node for
3566 * cross thread naming (pthread_set/getname_np(!self)). However, if
3567 * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3568 * which locks out the cross thread naming implementation.
3569 * This function makes sure that the node is always accessible for members of
3570 * same thread group.
3572 static int proc_tid_comm_permission(struct mnt_idmap *idmap,
3573 struct inode *inode, int mask)
3575 bool is_same_tgroup;
3576 struct task_struct *task;
3578 task = get_proc_task(inode);
3581 is_same_tgroup = same_thread_group(current, task);
3582 put_task_struct(task);
3584 if (likely(is_same_tgroup && !(mask & MAY_EXEC))) {
3585 /* This file (/proc/<pid>/task/<tid>/comm) can always be
3586 * read or written by the members of the corresponding
3592 return generic_permission(&nop_mnt_idmap, inode, mask);
3595 static const struct inode_operations proc_tid_comm_inode_operations = {
3596 .setattr = proc_setattr,
3597 .permission = proc_tid_comm_permission,
3603 static const struct pid_entry tid_base_stuff[] = {
3604 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3605 DIR("fdinfo", S_IRUGO|S_IXUGO, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3606 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3608 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3610 REG("environ", S_IRUSR, proc_environ_operations),
3611 REG("auxv", S_IRUSR, proc_auxv_operations),
3612 ONE("status", S_IRUGO, proc_pid_status),
3613 ONE("personality", S_IRUSR, proc_pid_personality),
3614 ONE("limits", S_IRUGO, proc_pid_limits),
3615 #ifdef CONFIG_SCHED_DEBUG
3616 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3618 NOD("comm", S_IFREG|S_IRUGO|S_IWUSR,
3619 &proc_tid_comm_inode_operations,
3620 &proc_pid_set_comm_operations, {}),
3621 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3622 ONE("syscall", S_IRUSR, proc_pid_syscall),
3624 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
3625 ONE("stat", S_IRUGO, proc_tid_stat),
3626 ONE("statm", S_IRUGO, proc_pid_statm),
3627 REG("maps", S_IRUGO, proc_pid_maps_operations),
3628 #ifdef CONFIG_PROC_CHILDREN
3629 REG("children", S_IRUGO, proc_tid_children_operations),
3632 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
3634 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3635 LNK("cwd", proc_cwd_link),
3636 LNK("root", proc_root_link),
3637 LNK("exe", proc_exe_link),
3638 REG("mounts", S_IRUGO, proc_mounts_operations),
3639 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3640 #ifdef CONFIG_PROC_PAGE_MONITOR
3641 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3642 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
3643 REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
3644 REG("pagemap", S_IRUSR, proc_pagemap_operations),
3646 #ifdef CONFIG_SECURITY
3647 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3649 #ifdef CONFIG_KALLSYMS
3650 ONE("wchan", S_IRUGO, proc_pid_wchan),
3652 #ifdef CONFIG_STACKTRACE
3653 ONE("stack", S_IRUSR, proc_pid_stack),
3655 #ifdef CONFIG_SCHED_INFO
3656 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
3658 #ifdef CONFIG_LATENCYTOP
3659 REG("latency", S_IRUGO, proc_lstats_operations),
3661 #ifdef CONFIG_PROC_PID_CPUSET
3662 ONE("cpuset", S_IRUGO, proc_cpuset_show),
3664 #ifdef CONFIG_CGROUPS
3665 ONE("cgroup", S_IRUGO, proc_cgroup_show),
3667 #ifdef CONFIG_PROC_CPU_RESCTRL
3668 ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show),
3670 ONE("oom_score", S_IRUGO, proc_oom_score),
3671 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3672 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3674 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3675 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3677 #ifdef CONFIG_FAULT_INJECTION
3678 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3679 REG("fail-nth", 0644, proc_fail_nth_operations),
3681 #ifdef CONFIG_TASK_IO_ACCOUNTING
3682 ONE("io", S_IRUSR, proc_tid_io_accounting),
3684 #ifdef CONFIG_USER_NS
3685 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3686 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3687 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3688 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
3690 #ifdef CONFIG_LIVEPATCH
3691 ONE("patch_state", S_IRUSR, proc_pid_patch_state),
3693 #ifdef CONFIG_PROC_PID_ARCH_STATUS
3694 ONE("arch_status", S_IRUGO, proc_pid_arch_status),
3696 #ifdef CONFIG_SECCOMP_CACHE_DEBUG
3697 ONE("seccomp_cache", S_IRUSR, proc_pid_seccomp_cache),
3700 ONE("ksm_merging_pages", S_IRUSR, proc_pid_ksm_merging_pages),
3701 ONE("ksm_stat", S_IRUSR, proc_pid_ksm_stat),
3705 static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
3707 return proc_pident_readdir(file, ctx,
3708 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3711 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3713 return proc_pident_lookup(dir, dentry,
3715 tid_base_stuff + ARRAY_SIZE(tid_base_stuff));
3718 static const struct file_operations proc_tid_base_operations = {
3719 .read = generic_read_dir,
3720 .iterate_shared = proc_tid_base_readdir,
3721 .llseek = generic_file_llseek,
3724 static const struct inode_operations proc_tid_base_inode_operations = {
3725 .lookup = proc_tid_base_lookup,
3726 .getattr = pid_getattr,
3727 .setattr = proc_setattr,
3730 static struct dentry *proc_task_instantiate(struct dentry *dentry,
3731 struct task_struct *task, const void *ptr)
3733 struct inode *inode;
3734 inode = proc_pid_make_base_inode(dentry->d_sb, task,
3735 S_IFDIR | S_IRUGO | S_IXUGO);
3737 return ERR_PTR(-ENOENT);
3739 inode->i_op = &proc_tid_base_inode_operations;
3740 inode->i_fop = &proc_tid_base_operations;
3741 inode->i_flags |= S_IMMUTABLE;
3743 set_nlink(inode, nlink_tid);
3744 pid_update_inode(task, inode);
3746 d_set_d_op(dentry, &pid_dentry_operations);
3747 return d_splice_alias(inode, dentry);
3750 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3752 struct task_struct *task;
3753 struct task_struct *leader = get_proc_task(dir);
3755 struct proc_fs_info *fs_info;
3756 struct pid_namespace *ns;
3757 struct dentry *result = ERR_PTR(-ENOENT);
3762 tid = name_to_int(&dentry->d_name);
3766 fs_info = proc_sb_info(dentry->d_sb);
3767 ns = fs_info->pid_ns;
3769 task = find_task_by_pid_ns(tid, ns);
3771 get_task_struct(task);
3775 if (!same_thread_group(leader, task))
3778 result = proc_task_instantiate(dentry, task, NULL);
3780 put_task_struct(task);
3782 put_task_struct(leader);
3788 * Find the first tid of a thread group to return to user space.
3790 * Usually this is just the thread group leader, but if the users
3791 * buffer was too small or there was a seek into the middle of the
3792 * directory we have more work todo.
3794 * In the case of a short read we start with find_task_by_pid.
3796 * In the case of a seek we start with the leader and walk nr
3799 static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3800 struct pid_namespace *ns)
3802 struct task_struct *pos, *task;
3803 unsigned long nr = f_pos;
3805 if (nr != f_pos) /* 32bit overflow? */
3809 task = pid_task(pid, PIDTYPE_PID);
3813 /* Attempt to start with the tid of a thread */
3815 pos = find_task_by_pid_ns(tid, ns);
3816 if (pos && same_thread_group(pos, task))
3820 /* If nr exceeds the number of threads there is nothing todo */
3821 if (nr >= get_nr_threads(task))
3824 /* If we haven't found our starting place yet start
3825 * with the leader and walk nr threads forward.
3827 for_each_thread(task, pos) {
3835 get_task_struct(pos);
3842 * Find the next thread in the thread list.
3843 * Return NULL if there is an error or no next thread.
3845 * The reference to the input task_struct is released.
3847 static struct task_struct *next_tid(struct task_struct *start)
3849 struct task_struct *pos = NULL;
3851 if (pid_alive(start)) {
3852 pos = __next_thread(start);
3854 get_task_struct(pos);
3857 put_task_struct(start);
3861 /* for the /proc/TGID/task/ directories */
3862 static int proc_task_readdir(struct file *file, struct dir_context *ctx)
3864 struct inode *inode = file_inode(file);
3865 struct task_struct *task;
3866 struct pid_namespace *ns;
3869 if (proc_inode_is_dead(inode))
3872 if (!dir_emit_dots(file, ctx))
3875 /* f_version caches the tgid value that the last readdir call couldn't
3876 * return. lseek aka telldir automagically resets f_version to 0.
3878 ns = proc_pid_ns(inode->i_sb);
3879 tid = (int)file->f_version;
3880 file->f_version = 0;
3881 for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
3883 task = next_tid(task), ctx->pos++) {
3887 tid = task_pid_nr_ns(task, ns);
3889 continue; /* The task has just exited. */
3890 len = snprintf(name, sizeof(name), "%u", tid);
3891 if (!proc_fill_cache(file, ctx, name, len,
3892 proc_task_instantiate, task, NULL)) {
3893 /* returning this tgid failed, save it as the first
3894 * pid for the next readir call */
3895 file->f_version = (u64)tid;
3896 put_task_struct(task);
3904 static int proc_task_getattr(struct mnt_idmap *idmap,
3905 const struct path *path, struct kstat *stat,
3906 u32 request_mask, unsigned int query_flags)
3908 struct inode *inode = d_inode(path->dentry);
3909 struct task_struct *p = get_proc_task(inode);
3910 generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat);
3913 stat->nlink += get_nr_threads(p);
3920 static const struct inode_operations proc_task_inode_operations = {
3921 .lookup = proc_task_lookup,
3922 .getattr = proc_task_getattr,
3923 .setattr = proc_setattr,
3924 .permission = proc_pid_permission,
3927 static const struct file_operations proc_task_operations = {
3928 .read = generic_read_dir,
3929 .iterate_shared = proc_task_readdir,
3930 .llseek = generic_file_llseek,
3933 void __init set_proc_pid_nlink(void)
3935 nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3936 nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));