4 * Copyright (C) 1991, 1992 Linus Torvalds
6 * proc base directory handling functions
8 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
9 * Instead of using magical inumbers to determine the kind of object
10 * we allocate and fill in-core inodes upon lookup. They don't even
11 * go into icache. We cache the reference to task_struct upon lookup too.
12 * Eventually it should become a filesystem in its own. We don't use the
13 * rest of procfs anymore.
19 * Bruna Moreira <bruna.moreira@indt.org.br>
20 * Edjard Mota <edjard.mota@indt.org.br>
21 * Ilias Biris <ilias.biris@indt.org.br>
22 * Mauricio Lin <mauricio.lin@indt.org.br>
24 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
26 * A new process specific entry (smaps) included in /proc. It shows the
27 * size of rss for each memory area. The maps entry lacks information
28 * about physical memory size (rss) for each mapped file, i.e.,
29 * rss information for executables and library files.
30 * This additional information is useful for any tools that need to know
31 * about physical memory consumption for a process specific library.
35 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36 * Pud inclusion in the page table walking.
40 * 10LE Instituto Nokia de Tecnologia - INdT:
41 * A better way to walks through the page table as suggested by Hugh Dickins.
43 * Simo Piiroinen <simo.piiroinen@nokia.com>:
44 * Smaps information related to shared, private, clean and dirty pages.
46 * Paul Mundt <paul.mundt@nokia.com>:
47 * Overall revision about smaps.
50 #include <asm/uaccess.h>
52 #include <linux/errno.h>
53 #include <linux/time.h>
54 #include <linux/proc_fs.h>
55 #include <linux/stat.h>
56 #include <linux/task_io_accounting_ops.h>
57 #include <linux/init.h>
58 #include <linux/capability.h>
59 #include <linux/file.h>
60 #include <linux/fdtable.h>
61 #include <linux/string.h>
62 #include <linux/seq_file.h>
63 #include <linux/namei.h>
64 #include <linux/mnt_namespace.h>
66 #include <linux/swap.h>
67 #include <linux/rcupdate.h>
68 #include <linux/kallsyms.h>
69 #include <linux/stacktrace.h>
70 #include <linux/resource.h>
71 #include <linux/module.h>
72 #include <linux/mount.h>
73 #include <linux/security.h>
74 #include <linux/ptrace.h>
75 #include <linux/tracehook.h>
76 #include <linux/printk.h>
77 #include <linux/cgroup.h>
78 #include <linux/cpuset.h>
79 #include <linux/audit.h>
80 #include <linux/poll.h>
81 #include <linux/nsproxy.h>
82 #include <linux/oom.h>
83 #include <linux/elf.h>
84 #include <linux/pid_namespace.h>
85 #include <linux/user_namespace.h>
86 #include <linux/fs_struct.h>
87 #include <linux/slab.h>
88 #include <linux/flex_array.h>
89 #include <linux/posix-timers.h>
90 #ifdef CONFIG_HARDWALL
91 #include <asm/hardwall.h>
93 #include <trace/events/oom.h>
98 * Implementing inode permission operations in /proc is almost
99 * certainly an error. Permission checks need to happen during
100 * each system call not at open time. The reason is that most of
101 * what we wish to check for permissions in /proc varies at runtime.
103 * The classic example of a problem is opening file descriptors
104 * in /proc for a task before it execs a suid executable.
111 const struct inode_operations *iop;
112 const struct file_operations *fop;
116 #define NOD(NAME, MODE, IOP, FOP, OP) { \
118 .len = sizeof(NAME) - 1, \
125 #define DIR(NAME, MODE, iops, fops) \
126 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
127 #define LNK(NAME, get_link) \
128 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
129 &proc_pid_link_inode_operations, NULL, \
130 { .proc_get_link = get_link } )
131 #define REG(NAME, MODE, fops) \
132 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
133 #define INF(NAME, MODE, read) \
134 NOD(NAME, (S_IFREG|(MODE)), \
135 NULL, &proc_info_file_operations, \
136 { .proc_read = read } )
137 #define ONE(NAME, MODE, show) \
138 NOD(NAME, (S_IFREG|(MODE)), \
139 NULL, &proc_single_file_operations, \
140 { .proc_show = show } )
143 * Count the number of hardlinks for the pid_entry table, excluding the .
146 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
153 for (i = 0; i < n; ++i) {
154 if (S_ISDIR(entries[i].mode))
161 static int get_task_root(struct task_struct *task, struct path *root)
163 int result = -ENOENT;
167 get_fs_root(task->fs, root);
174 static int proc_cwd_link(struct dentry *dentry, struct path *path)
176 struct task_struct *task = get_proc_task(dentry->d_inode);
177 int result = -ENOENT;
182 get_fs_pwd(task->fs, path);
186 put_task_struct(task);
191 static int proc_root_link(struct dentry *dentry, struct path *path)
193 struct task_struct *task = get_proc_task(dentry->d_inode);
194 int result = -ENOENT;
197 result = get_task_root(task, path);
198 put_task_struct(task);
203 static int proc_pid_cmdline(struct task_struct *task, char *buffer)
205 return get_cmdline(task, buffer, PAGE_SIZE);
208 static int proc_pid_auxv(struct task_struct *task, char *buffer)
210 struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ);
211 int res = PTR_ERR(mm);
212 if (mm && !IS_ERR(mm)) {
213 unsigned int nwords = 0;
216 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
217 res = nwords * sizeof(mm->saved_auxv[0]);
220 memcpy(buffer, mm->saved_auxv, res);
227 #ifdef CONFIG_KALLSYMS
229 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
230 * Returns the resolved symbol. If that fails, simply return the address.
232 static int proc_pid_wchan(struct task_struct *task, char *buffer)
235 char symname[KSYM_NAME_LEN];
237 wchan = get_wchan(task);
239 if (lookup_symbol_name(wchan, symname) < 0)
240 if (!ptrace_may_access(task, PTRACE_MODE_READ))
243 return sprintf(buffer, "%lu", wchan);
245 return sprintf(buffer, "%s", symname);
247 #endif /* CONFIG_KALLSYMS */
249 static int lock_trace(struct task_struct *task)
251 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
254 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
255 mutex_unlock(&task->signal->cred_guard_mutex);
261 static void unlock_trace(struct task_struct *task)
263 mutex_unlock(&task->signal->cred_guard_mutex);
266 #ifdef CONFIG_STACKTRACE
268 #define MAX_STACK_TRACE_DEPTH 64
270 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
271 struct pid *pid, struct task_struct *task)
273 struct stack_trace trace;
274 unsigned long *entries;
278 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
282 trace.nr_entries = 0;
283 trace.max_entries = MAX_STACK_TRACE_DEPTH;
284 trace.entries = entries;
287 err = lock_trace(task);
289 save_stack_trace_tsk(task, &trace);
291 for (i = 0; i < trace.nr_entries; i++) {
292 seq_printf(m, "[<%pK>] %pS\n",
293 (void *)entries[i], (void *)entries[i]);
303 #ifdef CONFIG_SCHEDSTATS
305 * Provides /proc/PID/schedstat
307 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
309 return sprintf(buffer, "%llu %llu %lu\n",
310 (unsigned long long)task->se.sum_exec_runtime,
311 (unsigned long long)task->sched_info.run_delay,
312 task->sched_info.pcount);
316 #ifdef CONFIG_LATENCYTOP
317 static int lstats_show_proc(struct seq_file *m, void *v)
320 struct inode *inode = m->private;
321 struct task_struct *task = get_proc_task(inode);
325 seq_puts(m, "Latency Top version : v0.1\n");
326 for (i = 0; i < 32; i++) {
327 struct latency_record *lr = &task->latency_record[i];
328 if (lr->backtrace[0]) {
330 seq_printf(m, "%i %li %li",
331 lr->count, lr->time, lr->max);
332 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
333 unsigned long bt = lr->backtrace[q];
338 seq_printf(m, " %ps", (void *)bt);
344 put_task_struct(task);
348 static int lstats_open(struct inode *inode, struct file *file)
350 return single_open(file, lstats_show_proc, inode);
353 static ssize_t lstats_write(struct file *file, const char __user *buf,
354 size_t count, loff_t *offs)
356 struct task_struct *task = get_proc_task(file_inode(file));
360 clear_all_latency_tracing(task);
361 put_task_struct(task);
366 static const struct file_operations proc_lstats_operations = {
369 .write = lstats_write,
371 .release = single_release,
376 #ifdef CONFIG_CGROUPS
377 static int cgroup_open(struct inode *inode, struct file *file)
379 struct pid *pid = PROC_I(inode)->pid;
380 return single_open(file, proc_cgroup_show, pid);
383 static const struct file_operations proc_cgroup_operations = {
387 .release = single_release,
391 #ifdef CONFIG_PROC_PID_CPUSET
393 static int cpuset_open(struct inode *inode, struct file *file)
395 struct pid *pid = PROC_I(inode)->pid;
396 return single_open(file, proc_cpuset_show, pid);
399 static const struct file_operations proc_cpuset_operations = {
403 .release = single_release,
407 static int proc_oom_score(struct task_struct *task, char *buffer)
409 unsigned long totalpages = totalram_pages + total_swap_pages;
410 unsigned long points = 0;
412 read_lock(&tasklist_lock);
414 points = oom_badness(task, NULL, NULL, totalpages) *
416 read_unlock(&tasklist_lock);
417 return sprintf(buffer, "%lu\n", points);
425 static const struct limit_names lnames[RLIM_NLIMITS] = {
426 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
427 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
428 [RLIMIT_DATA] = {"Max data size", "bytes"},
429 [RLIMIT_STACK] = {"Max stack size", "bytes"},
430 [RLIMIT_CORE] = {"Max core file size", "bytes"},
431 [RLIMIT_RSS] = {"Max resident set", "bytes"},
432 [RLIMIT_NPROC] = {"Max processes", "processes"},
433 [RLIMIT_NOFILE] = {"Max open files", "files"},
434 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
435 [RLIMIT_AS] = {"Max address space", "bytes"},
436 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
437 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
438 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
439 [RLIMIT_NICE] = {"Max nice priority", NULL},
440 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
441 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
444 /* Display limits for a process */
445 static int proc_pid_limits(struct task_struct *task, char *buffer)
450 char *bufptr = buffer;
452 struct rlimit rlim[RLIM_NLIMITS];
454 if (!lock_task_sighand(task, &flags))
456 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
457 unlock_task_sighand(task, &flags);
460 * print the file header
462 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
463 "Limit", "Soft Limit", "Hard Limit", "Units");
465 for (i = 0; i < RLIM_NLIMITS; i++) {
466 if (rlim[i].rlim_cur == RLIM_INFINITY)
467 count += sprintf(&bufptr[count], "%-25s %-20s ",
468 lnames[i].name, "unlimited");
470 count += sprintf(&bufptr[count], "%-25s %-20lu ",
471 lnames[i].name, rlim[i].rlim_cur);
473 if (rlim[i].rlim_max == RLIM_INFINITY)
474 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
476 count += sprintf(&bufptr[count], "%-20lu ",
480 count += sprintf(&bufptr[count], "%-10s\n",
483 count += sprintf(&bufptr[count], "\n");
489 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
490 static int proc_pid_syscall(struct task_struct *task, char *buffer)
493 unsigned long args[6], sp, pc;
494 int res = lock_trace(task);
498 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
499 res = sprintf(buffer, "running\n");
501 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
503 res = sprintf(buffer,
504 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
506 args[0], args[1], args[2], args[3], args[4], args[5],
511 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
513 /************************************************************************/
514 /* Here the fs part begins */
515 /************************************************************************/
517 /* permission checks */
518 static int proc_fd_access_allowed(struct inode *inode)
520 struct task_struct *task;
522 /* Allow access to a task's file descriptors if it is us or we
523 * may use ptrace attach to the process and find out that
526 task = get_proc_task(inode);
528 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
529 put_task_struct(task);
534 int proc_setattr(struct dentry *dentry, struct iattr *attr)
537 struct inode *inode = dentry->d_inode;
539 if (attr->ia_valid & ATTR_MODE)
542 error = inode_change_ok(inode, attr);
546 setattr_copy(inode, attr);
547 mark_inode_dirty(inode);
552 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
553 * or euid/egid (for hide_pid_min=2)?
555 static bool has_pid_permissions(struct pid_namespace *pid,
556 struct task_struct *task,
559 if (pid->hide_pid < hide_pid_min)
561 if (in_group_p(pid->pid_gid))
563 return ptrace_may_access(task, PTRACE_MODE_READ);
567 static int proc_pid_permission(struct inode *inode, int mask)
569 struct pid_namespace *pid = inode->i_sb->s_fs_info;
570 struct task_struct *task;
573 task = get_proc_task(inode);
576 has_perms = has_pid_permissions(pid, task, 1);
577 put_task_struct(task);
580 if (pid->hide_pid == 2) {
582 * Let's make getdents(), stat(), and open()
583 * consistent with each other. If a process
584 * may not stat() a file, it shouldn't be seen
592 return generic_permission(inode, mask);
597 static const struct inode_operations proc_def_inode_operations = {
598 .setattr = proc_setattr,
601 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
603 static ssize_t proc_info_read(struct file * file, char __user * buf,
604 size_t count, loff_t *ppos)
606 struct inode * inode = file_inode(file);
609 struct task_struct *task = get_proc_task(inode);
615 if (count > PROC_BLOCK_SIZE)
616 count = PROC_BLOCK_SIZE;
619 if (!(page = __get_free_page(GFP_TEMPORARY)))
622 length = PROC_I(inode)->op.proc_read(task, (char*)page);
625 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
628 put_task_struct(task);
633 static const struct file_operations proc_info_file_operations = {
634 .read = proc_info_read,
635 .llseek = generic_file_llseek,
638 static int proc_single_show(struct seq_file *m, void *v)
640 struct inode *inode = m->private;
641 struct pid_namespace *ns;
643 struct task_struct *task;
646 ns = inode->i_sb->s_fs_info;
647 pid = proc_pid(inode);
648 task = get_pid_task(pid, PIDTYPE_PID);
652 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
654 put_task_struct(task);
658 static int proc_single_open(struct inode *inode, struct file *filp)
660 return single_open(filp, proc_single_show, inode);
663 static const struct file_operations proc_single_file_operations = {
664 .open = proc_single_open,
667 .release = single_release,
670 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
672 struct task_struct *task = get_proc_task(file_inode(file));
673 struct mm_struct *mm;
678 mm = mm_access(task, mode);
679 put_task_struct(task);
685 /* ensure this mm_struct can't be freed */
686 atomic_inc(&mm->mm_count);
687 /* but do not pin its memory */
691 file->private_data = mm;
696 static int mem_open(struct inode *inode, struct file *file)
698 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
700 /* OK to pass negative loff_t, we can catch out-of-range */
701 file->f_mode |= FMODE_UNSIGNED_OFFSET;
706 static ssize_t mem_rw(struct file *file, char __user *buf,
707 size_t count, loff_t *ppos, int write)
709 struct mm_struct *mm = file->private_data;
710 unsigned long addr = *ppos;
717 page = (char *)__get_free_page(GFP_TEMPORARY);
722 if (!atomic_inc_not_zero(&mm->mm_users))
726 int this_len = min_t(int, count, PAGE_SIZE);
728 if (write && copy_from_user(page, buf, this_len)) {
733 this_len = access_remote_vm(mm, addr, page, this_len, write);
740 if (!write && copy_to_user(buf, page, this_len)) {
754 free_page((unsigned long) page);
758 static ssize_t mem_read(struct file *file, char __user *buf,
759 size_t count, loff_t *ppos)
761 return mem_rw(file, buf, count, ppos, 0);
764 static ssize_t mem_write(struct file *file, const char __user *buf,
765 size_t count, loff_t *ppos)
767 return mem_rw(file, (char __user*)buf, count, ppos, 1);
770 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
774 file->f_pos = offset;
777 file->f_pos += offset;
782 force_successful_syscall_return();
786 static int mem_release(struct inode *inode, struct file *file)
788 struct mm_struct *mm = file->private_data;
794 static const struct file_operations proc_mem_operations = {
799 .release = mem_release,
802 static int environ_open(struct inode *inode, struct file *file)
804 return __mem_open(inode, file, PTRACE_MODE_READ);
807 static ssize_t environ_read(struct file *file, char __user *buf,
808 size_t count, loff_t *ppos)
811 unsigned long src = *ppos;
813 struct mm_struct *mm = file->private_data;
818 page = (char *)__get_free_page(GFP_TEMPORARY);
823 if (!atomic_inc_not_zero(&mm->mm_users))
826 size_t this_len, max_len;
829 if (src >= (mm->env_end - mm->env_start))
832 this_len = mm->env_end - (mm->env_start + src);
834 max_len = min_t(size_t, PAGE_SIZE, count);
835 this_len = min(max_len, this_len);
837 retval = access_remote_vm(mm, (mm->env_start + src),
845 if (copy_to_user(buf, page, retval)) {
859 free_page((unsigned long) page);
863 static const struct file_operations proc_environ_operations = {
864 .open = environ_open,
865 .read = environ_read,
866 .llseek = generic_file_llseek,
867 .release = mem_release,
870 static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
873 struct task_struct *task = get_proc_task(file_inode(file));
874 char buffer[PROC_NUMBUF];
875 int oom_adj = OOM_ADJUST_MIN;
881 if (lock_task_sighand(task, &flags)) {
882 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
883 oom_adj = OOM_ADJUST_MAX;
885 oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
887 unlock_task_sighand(task, &flags);
889 put_task_struct(task);
890 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
891 return simple_read_from_buffer(buf, count, ppos, buffer, len);
894 static ssize_t oom_adj_write(struct file *file, const char __user *buf,
895 size_t count, loff_t *ppos)
897 struct task_struct *task;
898 char buffer[PROC_NUMBUF];
903 memset(buffer, 0, sizeof(buffer));
904 if (count > sizeof(buffer) - 1)
905 count = sizeof(buffer) - 1;
906 if (copy_from_user(buffer, buf, count)) {
911 err = kstrtoint(strstrip(buffer), 0, &oom_adj);
914 if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
915 oom_adj != OOM_DISABLE) {
920 task = get_proc_task(file_inode(file));
932 if (!lock_task_sighand(task, &flags)) {
938 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
939 * value is always attainable.
941 if (oom_adj == OOM_ADJUST_MAX)
942 oom_adj = OOM_SCORE_ADJ_MAX;
944 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
946 if (oom_adj < task->signal->oom_score_adj &&
947 !capable(CAP_SYS_RESOURCE)) {
953 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
954 * /proc/pid/oom_score_adj instead.
956 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
957 current->comm, task_pid_nr(current), task_pid_nr(task),
960 task->signal->oom_score_adj = oom_adj;
961 trace_oom_score_adj_update(task);
963 unlock_task_sighand(task, &flags);
966 put_task_struct(task);
968 return err < 0 ? err : count;
971 static const struct file_operations proc_oom_adj_operations = {
972 .read = oom_adj_read,
973 .write = oom_adj_write,
974 .llseek = generic_file_llseek,
977 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
978 size_t count, loff_t *ppos)
980 struct task_struct *task = get_proc_task(file_inode(file));
981 char buffer[PROC_NUMBUF];
982 short oom_score_adj = OOM_SCORE_ADJ_MIN;
988 if (lock_task_sighand(task, &flags)) {
989 oom_score_adj = task->signal->oom_score_adj;
990 unlock_task_sighand(task, &flags);
992 put_task_struct(task);
993 len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
994 return simple_read_from_buffer(buf, count, ppos, buffer, len);
997 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
998 size_t count, loff_t *ppos)
1000 struct task_struct *task;
1001 char buffer[PROC_NUMBUF];
1002 unsigned long flags;
1006 memset(buffer, 0, sizeof(buffer));
1007 if (count > sizeof(buffer) - 1)
1008 count = sizeof(buffer) - 1;
1009 if (copy_from_user(buffer, buf, count)) {
1014 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1017 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1018 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1023 task = get_proc_task(file_inode(file));
1035 if (!lock_task_sighand(task, &flags)) {
1040 if ((short)oom_score_adj < task->signal->oom_score_adj_min &&
1041 !capable(CAP_SYS_RESOURCE)) {
1046 task->signal->oom_score_adj = (short)oom_score_adj;
1047 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1048 task->signal->oom_score_adj_min = (short)oom_score_adj;
1049 trace_oom_score_adj_update(task);
1052 unlock_task_sighand(task, &flags);
1055 put_task_struct(task);
1057 return err < 0 ? err : count;
1060 static const struct file_operations proc_oom_score_adj_operations = {
1061 .read = oom_score_adj_read,
1062 .write = oom_score_adj_write,
1063 .llseek = default_llseek,
1066 #ifdef CONFIG_AUDITSYSCALL
1067 #define TMPBUFLEN 21
1068 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1069 size_t count, loff_t *ppos)
1071 struct inode * inode = file_inode(file);
1072 struct task_struct *task = get_proc_task(inode);
1074 char tmpbuf[TMPBUFLEN];
1078 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1079 from_kuid(file->f_cred->user_ns,
1080 audit_get_loginuid(task)));
1081 put_task_struct(task);
1082 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1085 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1086 size_t count, loff_t *ppos)
1088 struct inode * inode = file_inode(file);
1095 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1101 if (count >= PAGE_SIZE)
1102 count = PAGE_SIZE - 1;
1105 /* No partial writes. */
1108 page = (char*)__get_free_page(GFP_TEMPORARY);
1112 if (copy_from_user(page, buf, count))
1116 loginuid = simple_strtoul(page, &tmp, 10);
1123 /* is userspace tring to explicitly UNSET the loginuid? */
1124 if (loginuid == AUDIT_UID_UNSET) {
1125 kloginuid = INVALID_UID;
1127 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1128 if (!uid_valid(kloginuid)) {
1134 length = audit_set_loginuid(kloginuid);
1135 if (likely(length == 0))
1139 free_page((unsigned long) page);
1143 static const struct file_operations proc_loginuid_operations = {
1144 .read = proc_loginuid_read,
1145 .write = proc_loginuid_write,
1146 .llseek = generic_file_llseek,
1149 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1150 size_t count, loff_t *ppos)
1152 struct inode * inode = file_inode(file);
1153 struct task_struct *task = get_proc_task(inode);
1155 char tmpbuf[TMPBUFLEN];
1159 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1160 audit_get_sessionid(task));
1161 put_task_struct(task);
1162 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1165 static const struct file_operations proc_sessionid_operations = {
1166 .read = proc_sessionid_read,
1167 .llseek = generic_file_llseek,
1171 #ifdef CONFIG_FAULT_INJECTION
1172 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1173 size_t count, loff_t *ppos)
1175 struct task_struct *task = get_proc_task(file_inode(file));
1176 char buffer[PROC_NUMBUF];
1182 make_it_fail = task->make_it_fail;
1183 put_task_struct(task);
1185 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1187 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1190 static ssize_t proc_fault_inject_write(struct file * file,
1191 const char __user * buf, size_t count, loff_t *ppos)
1193 struct task_struct *task;
1194 char buffer[PROC_NUMBUF], *end;
1197 if (!capable(CAP_SYS_RESOURCE))
1199 memset(buffer, 0, sizeof(buffer));
1200 if (count > sizeof(buffer) - 1)
1201 count = sizeof(buffer) - 1;
1202 if (copy_from_user(buffer, buf, count))
1204 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1207 task = get_proc_task(file_inode(file));
1210 task->make_it_fail = make_it_fail;
1211 put_task_struct(task);
1216 static const struct file_operations proc_fault_inject_operations = {
1217 .read = proc_fault_inject_read,
1218 .write = proc_fault_inject_write,
1219 .llseek = generic_file_llseek,
1224 #ifdef CONFIG_SCHED_DEBUG
1226 * Print out various scheduling related per-task fields:
1228 static int sched_show(struct seq_file *m, void *v)
1230 struct inode *inode = m->private;
1231 struct task_struct *p;
1233 p = get_proc_task(inode);
1236 proc_sched_show_task(p, m);
1244 sched_write(struct file *file, const char __user *buf,
1245 size_t count, loff_t *offset)
1247 struct inode *inode = file_inode(file);
1248 struct task_struct *p;
1250 p = get_proc_task(inode);
1253 proc_sched_set_task(p);
1260 static int sched_open(struct inode *inode, struct file *filp)
1262 return single_open(filp, sched_show, inode);
1265 static const struct file_operations proc_pid_sched_operations = {
1268 .write = sched_write,
1269 .llseek = seq_lseek,
1270 .release = single_release,
1275 #ifdef CONFIG_SCHED_AUTOGROUP
1277 * Print out autogroup related information:
1279 static int sched_autogroup_show(struct seq_file *m, void *v)
1281 struct inode *inode = m->private;
1282 struct task_struct *p;
1284 p = get_proc_task(inode);
1287 proc_sched_autogroup_show_task(p, m);
1295 sched_autogroup_write(struct file *file, const char __user *buf,
1296 size_t count, loff_t *offset)
1298 struct inode *inode = file_inode(file);
1299 struct task_struct *p;
1300 char buffer[PROC_NUMBUF];
1304 memset(buffer, 0, sizeof(buffer));
1305 if (count > sizeof(buffer) - 1)
1306 count = sizeof(buffer) - 1;
1307 if (copy_from_user(buffer, buf, count))
1310 err = kstrtoint(strstrip(buffer), 0, &nice);
1314 p = get_proc_task(inode);
1318 err = proc_sched_autogroup_set_nice(p, nice);
1327 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1331 ret = single_open(filp, sched_autogroup_show, NULL);
1333 struct seq_file *m = filp->private_data;
1340 static const struct file_operations proc_pid_sched_autogroup_operations = {
1341 .open = sched_autogroup_open,
1343 .write = sched_autogroup_write,
1344 .llseek = seq_lseek,
1345 .release = single_release,
1348 #endif /* CONFIG_SCHED_AUTOGROUP */
1350 static ssize_t comm_write(struct file *file, const char __user *buf,
1351 size_t count, loff_t *offset)
1353 struct inode *inode = file_inode(file);
1354 struct task_struct *p;
1355 char buffer[TASK_COMM_LEN];
1356 const size_t maxlen = sizeof(buffer) - 1;
1358 memset(buffer, 0, sizeof(buffer));
1359 if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
1362 p = get_proc_task(inode);
1366 if (same_thread_group(current, p))
1367 set_task_comm(p, buffer);
1376 static int comm_show(struct seq_file *m, void *v)
1378 struct inode *inode = m->private;
1379 struct task_struct *p;
1381 p = get_proc_task(inode);
1386 seq_printf(m, "%s\n", p->comm);
1394 static int comm_open(struct inode *inode, struct file *filp)
1396 return single_open(filp, comm_show, inode);
1399 static const struct file_operations proc_pid_set_comm_operations = {
1402 .write = comm_write,
1403 .llseek = seq_lseek,
1404 .release = single_release,
1407 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1409 struct task_struct *task;
1410 struct mm_struct *mm;
1411 struct file *exe_file;
1413 task = get_proc_task(dentry->d_inode);
1416 mm = get_task_mm(task);
1417 put_task_struct(task);
1420 exe_file = get_mm_exe_file(mm);
1423 *exe_path = exe_file->f_path;
1424 path_get(&exe_file->f_path);
1431 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1433 struct inode *inode = dentry->d_inode;
1435 int error = -EACCES;
1437 /* Are we allowed to snoop on the tasks file descriptors? */
1438 if (!proc_fd_access_allowed(inode))
1441 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1445 nd_jump_link(nd, &path);
1448 return ERR_PTR(error);
1451 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1453 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1460 pathname = d_path(path, tmp, PAGE_SIZE);
1461 len = PTR_ERR(pathname);
1462 if (IS_ERR(pathname))
1464 len = tmp + PAGE_SIZE - 1 - pathname;
1468 if (copy_to_user(buffer, pathname, len))
1471 free_page((unsigned long)tmp);
1475 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1477 int error = -EACCES;
1478 struct inode *inode = dentry->d_inode;
1481 /* Are we allowed to snoop on the tasks file descriptors? */
1482 if (!proc_fd_access_allowed(inode))
1485 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1489 error = do_proc_readlink(&path, buffer, buflen);
1495 const struct inode_operations proc_pid_link_inode_operations = {
1496 .readlink = proc_pid_readlink,
1497 .follow_link = proc_pid_follow_link,
1498 .setattr = proc_setattr,
1502 /* building an inode */
1504 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1506 struct inode * inode;
1507 struct proc_inode *ei;
1508 const struct cred *cred;
1510 /* We need a new inode */
1512 inode = new_inode(sb);
1518 inode->i_ino = get_next_ino();
1519 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1520 inode->i_op = &proc_def_inode_operations;
1523 * grab the reference to task.
1525 ei->pid = get_task_pid(task, PIDTYPE_PID);
1529 if (task_dumpable(task)) {
1531 cred = __task_cred(task);
1532 inode->i_uid = cred->euid;
1533 inode->i_gid = cred->egid;
1536 security_task_to_inode(task, inode);
1546 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1548 struct inode *inode = dentry->d_inode;
1549 struct task_struct *task;
1550 const struct cred *cred;
1551 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
1553 generic_fillattr(inode, stat);
1556 stat->uid = GLOBAL_ROOT_UID;
1557 stat->gid = GLOBAL_ROOT_GID;
1558 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1560 if (!has_pid_permissions(pid, task, 2)) {
1563 * This doesn't prevent learning whether PID exists,
1564 * it only makes getattr() consistent with readdir().
1568 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1569 task_dumpable(task)) {
1570 cred = __task_cred(task);
1571 stat->uid = cred->euid;
1572 stat->gid = cred->egid;
1582 * Exceptional case: normally we are not allowed to unhash a busy
1583 * directory. In this case, however, we can do it - no aliasing problems
1584 * due to the way we treat inodes.
1586 * Rewrite the inode's ownerships here because the owning task may have
1587 * performed a setuid(), etc.
1589 * Before the /proc/pid/status file was created the only way to read
1590 * the effective uid of a /process was to stat /proc/pid. Reading
1591 * /proc/pid/status is slow enough that procps and other packages
1592 * kept stating /proc/pid. To keep the rules in /proc simple I have
1593 * made this apply to all per process world readable and executable
1596 int pid_revalidate(struct dentry *dentry, unsigned int flags)
1598 struct inode *inode;
1599 struct task_struct *task;
1600 const struct cred *cred;
1602 if (flags & LOOKUP_RCU)
1605 inode = dentry->d_inode;
1606 task = get_proc_task(inode);
1609 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1610 task_dumpable(task)) {
1612 cred = __task_cred(task);
1613 inode->i_uid = cred->euid;
1614 inode->i_gid = cred->egid;
1617 inode->i_uid = GLOBAL_ROOT_UID;
1618 inode->i_gid = GLOBAL_ROOT_GID;
1620 inode->i_mode &= ~(S_ISUID | S_ISGID);
1621 security_task_to_inode(task, inode);
1622 put_task_struct(task);
1629 static inline bool proc_inode_is_dead(struct inode *inode)
1631 return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1634 int pid_delete_dentry(const struct dentry *dentry)
1636 /* Is the task we represent dead?
1637 * If so, then don't put the dentry on the lru list,
1638 * kill it immediately.
1640 return proc_inode_is_dead(dentry->d_inode);
1643 const struct dentry_operations pid_dentry_operations =
1645 .d_revalidate = pid_revalidate,
1646 .d_delete = pid_delete_dentry,
1652 * Fill a directory entry.
1654 * If possible create the dcache entry and derive our inode number and
1655 * file type from dcache entry.
1657 * Since all of the proc inode numbers are dynamically generated, the inode
1658 * numbers do not exist until the inode is cache. This means creating the
1659 * the dcache entry in readdir is necessary to keep the inode numbers
1660 * reported by readdir in sync with the inode numbers reported
1663 bool proc_fill_cache(struct file *file, struct dir_context *ctx,
1664 const char *name, int len,
1665 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1667 struct dentry *child, *dir = file->f_path.dentry;
1668 struct qstr qname = QSTR_INIT(name, len);
1669 struct inode *inode;
1673 child = d_hash_and_lookup(dir, &qname);
1675 child = d_alloc(dir, &qname);
1677 goto end_instantiate;
1678 if (instantiate(dir->d_inode, child, task, ptr) < 0) {
1680 goto end_instantiate;
1683 inode = child->d_inode;
1685 type = inode->i_mode >> 12;
1687 return dir_emit(ctx, name, len, ino, type);
1690 return dir_emit(ctx, name, len, 1, DT_UNKNOWN);
1693 #ifdef CONFIG_CHECKPOINT_RESTORE
1696 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1697 * which represent vma start and end addresses.
1699 static int dname_to_vma_addr(struct dentry *dentry,
1700 unsigned long *start, unsigned long *end)
1702 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1708 static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
1710 unsigned long vm_start, vm_end;
1711 bool exact_vma_exists = false;
1712 struct mm_struct *mm = NULL;
1713 struct task_struct *task;
1714 const struct cred *cred;
1715 struct inode *inode;
1718 if (flags & LOOKUP_RCU)
1721 if (!capable(CAP_SYS_ADMIN)) {
1726 inode = dentry->d_inode;
1727 task = get_proc_task(inode);
1731 mm = mm_access(task, PTRACE_MODE_READ);
1732 if (IS_ERR_OR_NULL(mm))
1735 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1736 down_read(&mm->mmap_sem);
1737 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1738 up_read(&mm->mmap_sem);
1743 if (exact_vma_exists) {
1744 if (task_dumpable(task)) {
1746 cred = __task_cred(task);
1747 inode->i_uid = cred->euid;
1748 inode->i_gid = cred->egid;
1751 inode->i_uid = GLOBAL_ROOT_UID;
1752 inode->i_gid = GLOBAL_ROOT_GID;
1754 security_task_to_inode(task, inode);
1759 put_task_struct(task);
1768 static const struct dentry_operations tid_map_files_dentry_operations = {
1769 .d_revalidate = map_files_d_revalidate,
1770 .d_delete = pid_delete_dentry,
1773 static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
1775 unsigned long vm_start, vm_end;
1776 struct vm_area_struct *vma;
1777 struct task_struct *task;
1778 struct mm_struct *mm;
1782 task = get_proc_task(dentry->d_inode);
1786 mm = get_task_mm(task);
1787 put_task_struct(task);
1791 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
1796 down_read(&mm->mmap_sem);
1797 vma = find_exact_vma(mm, vm_start, vm_end);
1798 if (vma && vma->vm_file) {
1799 *path = vma->vm_file->f_path;
1803 up_read(&mm->mmap_sem);
1811 struct map_files_info {
1814 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
1818 proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
1819 struct task_struct *task, const void *ptr)
1821 fmode_t mode = (fmode_t)(unsigned long)ptr;
1822 struct proc_inode *ei;
1823 struct inode *inode;
1825 inode = proc_pid_make_inode(dir->i_sb, task);
1830 ei->op.proc_get_link = proc_map_files_get_link;
1832 inode->i_op = &proc_pid_link_inode_operations;
1834 inode->i_mode = S_IFLNK;
1836 if (mode & FMODE_READ)
1837 inode->i_mode |= S_IRUSR;
1838 if (mode & FMODE_WRITE)
1839 inode->i_mode |= S_IWUSR;
1841 d_set_d_op(dentry, &tid_map_files_dentry_operations);
1842 d_add(dentry, inode);
1847 static struct dentry *proc_map_files_lookup(struct inode *dir,
1848 struct dentry *dentry, unsigned int flags)
1850 unsigned long vm_start, vm_end;
1851 struct vm_area_struct *vma;
1852 struct task_struct *task;
1854 struct mm_struct *mm;
1857 if (!capable(CAP_SYS_ADMIN))
1861 task = get_proc_task(dir);
1866 if (!ptrace_may_access(task, PTRACE_MODE_READ))
1870 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
1873 mm = get_task_mm(task);
1877 down_read(&mm->mmap_sem);
1878 vma = find_exact_vma(mm, vm_start, vm_end);
1883 result = proc_map_files_instantiate(dir, dentry, task,
1884 (void *)(unsigned long)vma->vm_file->f_mode);
1887 up_read(&mm->mmap_sem);
1890 put_task_struct(task);
1892 return ERR_PTR(result);
1895 static const struct inode_operations proc_map_files_inode_operations = {
1896 .lookup = proc_map_files_lookup,
1897 .permission = proc_fd_permission,
1898 .setattr = proc_setattr,
1902 proc_map_files_readdir(struct file *file, struct dir_context *ctx)
1904 struct vm_area_struct *vma;
1905 struct task_struct *task;
1906 struct mm_struct *mm;
1907 unsigned long nr_files, pos, i;
1908 struct flex_array *fa = NULL;
1909 struct map_files_info info;
1910 struct map_files_info *p;
1914 if (!capable(CAP_SYS_ADMIN))
1918 task = get_proc_task(file_inode(file));
1923 if (!ptrace_may_access(task, PTRACE_MODE_READ))
1927 if (!dir_emit_dots(file, ctx))
1930 mm = get_task_mm(task);
1933 down_read(&mm->mmap_sem);
1938 * We need two passes here:
1940 * 1) Collect vmas of mapped files with mmap_sem taken
1941 * 2) Release mmap_sem and instantiate entries
1943 * otherwise we get lockdep complained, since filldir()
1944 * routine might require mmap_sem taken in might_fault().
1947 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
1948 if (vma->vm_file && ++pos > ctx->pos)
1953 fa = flex_array_alloc(sizeof(info), nr_files,
1955 if (!fa || flex_array_prealloc(fa, 0, nr_files,
1959 flex_array_free(fa);
1960 up_read(&mm->mmap_sem);
1964 for (i = 0, vma = mm->mmap, pos = 2; vma;
1965 vma = vma->vm_next) {
1968 if (++pos <= ctx->pos)
1971 info.mode = vma->vm_file->f_mode;
1972 info.len = snprintf(info.name,
1973 sizeof(info.name), "%lx-%lx",
1974 vma->vm_start, vma->vm_end);
1975 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
1979 up_read(&mm->mmap_sem);
1981 for (i = 0; i < nr_files; i++) {
1982 p = flex_array_get(fa, i);
1983 if (!proc_fill_cache(file, ctx,
1985 proc_map_files_instantiate,
1987 (void *)(unsigned long)p->mode))
1992 flex_array_free(fa);
1996 put_task_struct(task);
2001 static const struct file_operations proc_map_files_operations = {
2002 .read = generic_read_dir,
2003 .iterate = proc_map_files_readdir,
2004 .llseek = default_llseek,
2007 struct timers_private {
2009 struct task_struct *task;
2010 struct sighand_struct *sighand;
2011 struct pid_namespace *ns;
2012 unsigned long flags;
2015 static void *timers_start(struct seq_file *m, loff_t *pos)
2017 struct timers_private *tp = m->private;
2019 tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2021 return ERR_PTR(-ESRCH);
2023 tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2025 return ERR_PTR(-ESRCH);
2027 return seq_list_start(&tp->task->signal->posix_timers, *pos);
2030 static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2032 struct timers_private *tp = m->private;
2033 return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2036 static void timers_stop(struct seq_file *m, void *v)
2038 struct timers_private *tp = m->private;
2041 unlock_task_sighand(tp->task, &tp->flags);
2046 put_task_struct(tp->task);
2051 static int show_timer(struct seq_file *m, void *v)
2053 struct k_itimer *timer;
2054 struct timers_private *tp = m->private;
2056 static char *nstr[] = {
2057 [SIGEV_SIGNAL] = "signal",
2058 [SIGEV_NONE] = "none",
2059 [SIGEV_THREAD] = "thread",
2062 timer = list_entry((struct list_head *)v, struct k_itimer, list);
2063 notify = timer->it_sigev_notify;
2065 seq_printf(m, "ID: %d\n", timer->it_id);
2066 seq_printf(m, "signal: %d/%p\n", timer->sigq->info.si_signo,
2067 timer->sigq->info.si_value.sival_ptr);
2068 seq_printf(m, "notify: %s/%s.%d\n",
2069 nstr[notify & ~SIGEV_THREAD_ID],
2070 (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2071 pid_nr_ns(timer->it_pid, tp->ns));
2072 seq_printf(m, "ClockID: %d\n", timer->it_clock);
2077 static const struct seq_operations proc_timers_seq_ops = {
2078 .start = timers_start,
2079 .next = timers_next,
2080 .stop = timers_stop,
2084 static int proc_timers_open(struct inode *inode, struct file *file)
2086 struct timers_private *tp;
2088 tp = __seq_open_private(file, &proc_timers_seq_ops,
2089 sizeof(struct timers_private));
2093 tp->pid = proc_pid(inode);
2094 tp->ns = inode->i_sb->s_fs_info;
2098 static const struct file_operations proc_timers_operations = {
2099 .open = proc_timers_open,
2101 .llseek = seq_lseek,
2102 .release = seq_release_private,
2104 #endif /* CONFIG_CHECKPOINT_RESTORE */
2106 static int proc_pident_instantiate(struct inode *dir,
2107 struct dentry *dentry, struct task_struct *task, const void *ptr)
2109 const struct pid_entry *p = ptr;
2110 struct inode *inode;
2111 struct proc_inode *ei;
2113 inode = proc_pid_make_inode(dir->i_sb, task);
2118 inode->i_mode = p->mode;
2119 if (S_ISDIR(inode->i_mode))
2120 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2122 inode->i_op = p->iop;
2124 inode->i_fop = p->fop;
2126 d_set_d_op(dentry, &pid_dentry_operations);
2127 d_add(dentry, inode);
2128 /* Close the race of the process dying before we return the dentry */
2129 if (pid_revalidate(dentry, 0))
2135 static struct dentry *proc_pident_lookup(struct inode *dir,
2136 struct dentry *dentry,
2137 const struct pid_entry *ents,
2141 struct task_struct *task = get_proc_task(dir);
2142 const struct pid_entry *p, *last;
2150 * Yes, it does not scale. And it should not. Don't add
2151 * new entries into /proc/<tgid>/ without very good reasons.
2153 last = &ents[nents - 1];
2154 for (p = ents; p <= last; p++) {
2155 if (p->len != dentry->d_name.len)
2157 if (!memcmp(dentry->d_name.name, p->name, p->len))
2163 error = proc_pident_instantiate(dir, dentry, task, p);
2165 put_task_struct(task);
2167 return ERR_PTR(error);
2170 static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
2171 const struct pid_entry *ents, unsigned int nents)
2173 struct task_struct *task = get_proc_task(file_inode(file));
2174 const struct pid_entry *p;
2179 if (!dir_emit_dots(file, ctx))
2182 if (ctx->pos >= nents + 2)
2185 for (p = ents + (ctx->pos - 2); p <= ents + nents - 1; p++) {
2186 if (!proc_fill_cache(file, ctx, p->name, p->len,
2187 proc_pident_instantiate, task, p))
2192 put_task_struct(task);
2196 #ifdef CONFIG_SECURITY
2197 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2198 size_t count, loff_t *ppos)
2200 struct inode * inode = file_inode(file);
2203 struct task_struct *task = get_proc_task(inode);
2208 length = security_getprocattr(task,
2209 (char*)file->f_path.dentry->d_name.name,
2211 put_task_struct(task);
2213 length = simple_read_from_buffer(buf, count, ppos, p, length);
2218 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2219 size_t count, loff_t *ppos)
2221 struct inode * inode = file_inode(file);
2224 struct task_struct *task = get_proc_task(inode);
2229 if (count > PAGE_SIZE)
2232 /* No partial writes. */
2238 page = (char*)__get_free_page(GFP_TEMPORARY);
2243 if (copy_from_user(page, buf, count))
2246 /* Guard against adverse ptrace interaction */
2247 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2251 length = security_setprocattr(task,
2252 (char*)file->f_path.dentry->d_name.name,
2253 (void*)page, count);
2254 mutex_unlock(&task->signal->cred_guard_mutex);
2256 free_page((unsigned long) page);
2258 put_task_struct(task);
2263 static const struct file_operations proc_pid_attr_operations = {
2264 .read = proc_pid_attr_read,
2265 .write = proc_pid_attr_write,
2266 .llseek = generic_file_llseek,
2269 static const struct pid_entry attr_dir_stuff[] = {
2270 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2271 REG("prev", S_IRUGO, proc_pid_attr_operations),
2272 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2273 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2274 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2275 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2278 static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
2280 return proc_pident_readdir(file, ctx,
2281 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2284 static const struct file_operations proc_attr_dir_operations = {
2285 .read = generic_read_dir,
2286 .iterate = proc_attr_dir_readdir,
2287 .llseek = default_llseek,
2290 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2291 struct dentry *dentry, unsigned int flags)
2293 return proc_pident_lookup(dir, dentry,
2294 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2297 static const struct inode_operations proc_attr_dir_inode_operations = {
2298 .lookup = proc_attr_dir_lookup,
2299 .getattr = pid_getattr,
2300 .setattr = proc_setattr,
2305 #ifdef CONFIG_ELF_CORE
2306 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2307 size_t count, loff_t *ppos)
2309 struct task_struct *task = get_proc_task(file_inode(file));
2310 struct mm_struct *mm;
2311 char buffer[PROC_NUMBUF];
2319 mm = get_task_mm(task);
2321 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2322 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2323 MMF_DUMP_FILTER_SHIFT));
2325 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2328 put_task_struct(task);
2333 static ssize_t proc_coredump_filter_write(struct file *file,
2334 const char __user *buf,
2338 struct task_struct *task;
2339 struct mm_struct *mm;
2340 char buffer[PROC_NUMBUF], *end;
2347 memset(buffer, 0, sizeof(buffer));
2348 if (count > sizeof(buffer) - 1)
2349 count = sizeof(buffer) - 1;
2350 if (copy_from_user(buffer, buf, count))
2354 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2357 if (end - buffer == 0)
2361 task = get_proc_task(file_inode(file));
2366 mm = get_task_mm(task);
2370 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2372 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2374 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2379 put_task_struct(task);
2384 static const struct file_operations proc_coredump_filter_operations = {
2385 .read = proc_coredump_filter_read,
2386 .write = proc_coredump_filter_write,
2387 .llseek = generic_file_llseek,
2391 #ifdef CONFIG_TASK_IO_ACCOUNTING
2392 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2394 struct task_io_accounting acct = task->ioac;
2395 unsigned long flags;
2398 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2402 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2407 if (whole && lock_task_sighand(task, &flags)) {
2408 struct task_struct *t = task;
2410 task_io_accounting_add(&acct, &task->signal->ioac);
2411 while_each_thread(task, t)
2412 task_io_accounting_add(&acct, &t->ioac);
2414 unlock_task_sighand(task, &flags);
2416 result = sprintf(buffer,
2421 "read_bytes: %llu\n"
2422 "write_bytes: %llu\n"
2423 "cancelled_write_bytes: %llu\n",
2424 (unsigned long long)acct.rchar,
2425 (unsigned long long)acct.wchar,
2426 (unsigned long long)acct.syscr,
2427 (unsigned long long)acct.syscw,
2428 (unsigned long long)acct.read_bytes,
2429 (unsigned long long)acct.write_bytes,
2430 (unsigned long long)acct.cancelled_write_bytes);
2432 mutex_unlock(&task->signal->cred_guard_mutex);
2436 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2438 return do_io_accounting(task, buffer, 0);
2441 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2443 return do_io_accounting(task, buffer, 1);
2445 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2447 #ifdef CONFIG_USER_NS
2448 static int proc_id_map_open(struct inode *inode, struct file *file,
2449 struct seq_operations *seq_ops)
2451 struct user_namespace *ns = NULL;
2452 struct task_struct *task;
2453 struct seq_file *seq;
2456 task = get_proc_task(inode);
2459 ns = get_user_ns(task_cred_xxx(task, user_ns));
2461 put_task_struct(task);
2466 ret = seq_open(file, seq_ops);
2470 seq = file->private_data;
2480 static int proc_id_map_release(struct inode *inode, struct file *file)
2482 struct seq_file *seq = file->private_data;
2483 struct user_namespace *ns = seq->private;
2485 return seq_release(inode, file);
2488 static int proc_uid_map_open(struct inode *inode, struct file *file)
2490 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2493 static int proc_gid_map_open(struct inode *inode, struct file *file)
2495 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2498 static int proc_projid_map_open(struct inode *inode, struct file *file)
2500 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2503 static const struct file_operations proc_uid_map_operations = {
2504 .open = proc_uid_map_open,
2505 .write = proc_uid_map_write,
2507 .llseek = seq_lseek,
2508 .release = proc_id_map_release,
2511 static const struct file_operations proc_gid_map_operations = {
2512 .open = proc_gid_map_open,
2513 .write = proc_gid_map_write,
2515 .llseek = seq_lseek,
2516 .release = proc_id_map_release,
2519 static const struct file_operations proc_projid_map_operations = {
2520 .open = proc_projid_map_open,
2521 .write = proc_projid_map_write,
2523 .llseek = seq_lseek,
2524 .release = proc_id_map_release,
2526 #endif /* CONFIG_USER_NS */
2528 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2529 struct pid *pid, struct task_struct *task)
2531 int err = lock_trace(task);
2533 seq_printf(m, "%08x\n", task->personality);
2542 static const struct file_operations proc_task_operations;
2543 static const struct inode_operations proc_task_inode_operations;
2545 static const struct pid_entry tgid_base_stuff[] = {
2546 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2547 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2548 #ifdef CONFIG_CHECKPOINT_RESTORE
2549 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
2551 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2552 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2554 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2556 REG("environ", S_IRUSR, proc_environ_operations),
2557 INF("auxv", S_IRUSR, proc_pid_auxv),
2558 ONE("status", S_IRUGO, proc_pid_status),
2559 ONE("personality", S_IRUGO, proc_pid_personality),
2560 INF("limits", S_IRUGO, proc_pid_limits),
2561 #ifdef CONFIG_SCHED_DEBUG
2562 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2564 #ifdef CONFIG_SCHED_AUTOGROUP
2565 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2567 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2568 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2569 INF("syscall", S_IRUGO, proc_pid_syscall),
2571 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2572 ONE("stat", S_IRUGO, proc_tgid_stat),
2573 ONE("statm", S_IRUGO, proc_pid_statm),
2574 REG("maps", S_IRUGO, proc_pid_maps_operations),
2576 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
2578 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2579 LNK("cwd", proc_cwd_link),
2580 LNK("root", proc_root_link),
2581 LNK("exe", proc_exe_link),
2582 REG("mounts", S_IRUGO, proc_mounts_operations),
2583 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2584 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2585 #ifdef CONFIG_PROC_PAGE_MONITOR
2586 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2587 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
2588 REG("pagemap", S_IRUGO, proc_pagemap_operations),
2590 #ifdef CONFIG_SECURITY
2591 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2593 #ifdef CONFIG_KALLSYMS
2594 INF("wchan", S_IRUGO, proc_pid_wchan),
2596 #ifdef CONFIG_STACKTRACE
2597 ONE("stack", S_IRUGO, proc_pid_stack),
2599 #ifdef CONFIG_SCHEDSTATS
2600 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2602 #ifdef CONFIG_LATENCYTOP
2603 REG("latency", S_IRUGO, proc_lstats_operations),
2605 #ifdef CONFIG_PROC_PID_CPUSET
2606 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2608 #ifdef CONFIG_CGROUPS
2609 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2611 INF("oom_score", S_IRUGO, proc_oom_score),
2612 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
2613 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2614 #ifdef CONFIG_AUDITSYSCALL
2615 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2616 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2618 #ifdef CONFIG_FAULT_INJECTION
2619 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2621 #ifdef CONFIG_ELF_CORE
2622 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2624 #ifdef CONFIG_TASK_IO_ACCOUNTING
2625 INF("io", S_IRUSR, proc_tgid_io_accounting),
2627 #ifdef CONFIG_HARDWALL
2628 INF("hardwall", S_IRUGO, proc_pid_hardwall),
2630 #ifdef CONFIG_USER_NS
2631 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
2632 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
2633 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
2635 #ifdef CONFIG_CHECKPOINT_RESTORE
2636 REG("timers", S_IRUGO, proc_timers_operations),
2640 static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
2642 return proc_pident_readdir(file, ctx,
2643 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2646 static const struct file_operations proc_tgid_base_operations = {
2647 .read = generic_read_dir,
2648 .iterate = proc_tgid_base_readdir,
2649 .llseek = default_llseek,
2652 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2654 return proc_pident_lookup(dir, dentry,
2655 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2658 static const struct inode_operations proc_tgid_base_inode_operations = {
2659 .lookup = proc_tgid_base_lookup,
2660 .getattr = pid_getattr,
2661 .setattr = proc_setattr,
2662 .permission = proc_pid_permission,
2665 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2667 struct dentry *dentry, *leader, *dir;
2668 char buf[PROC_NUMBUF];
2672 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2673 /* no ->d_hash() rejects on procfs */
2674 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2676 shrink_dcache_parent(dentry);
2682 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2683 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2688 name.len = strlen(name.name);
2689 dir = d_hash_and_lookup(leader, &name);
2691 goto out_put_leader;
2694 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2695 dentry = d_hash_and_lookup(dir, &name);
2697 shrink_dcache_parent(dentry);
2710 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2711 * @task: task that should be flushed.
2713 * When flushing dentries from proc, one needs to flush them from global
2714 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2715 * in. This call is supposed to do all of this job.
2717 * Looks in the dcache for
2719 * /proc/@tgid/task/@pid
2720 * if either directory is present flushes it and all of it'ts children
2723 * It is safe and reasonable to cache /proc entries for a task until
2724 * that task exits. After that they just clog up the dcache with
2725 * useless entries, possibly causing useful dcache entries to be
2726 * flushed instead. This routine is proved to flush those useless
2727 * dcache entries at process exit time.
2729 * NOTE: This routine is just an optimization so it does not guarantee
2730 * that no dcache entries will exist at process exit time it
2731 * just makes it very unlikely that any will persist.
2734 void proc_flush_task(struct task_struct *task)
2737 struct pid *pid, *tgid;
2740 pid = task_pid(task);
2741 tgid = task_tgid(task);
2743 for (i = 0; i <= pid->level; i++) {
2744 upid = &pid->numbers[i];
2745 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2746 tgid->numbers[i].nr);
2750 static int proc_pid_instantiate(struct inode *dir,
2751 struct dentry * dentry,
2752 struct task_struct *task, const void *ptr)
2754 struct inode *inode;
2756 inode = proc_pid_make_inode(dir->i_sb, task);
2760 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2761 inode->i_op = &proc_tgid_base_inode_operations;
2762 inode->i_fop = &proc_tgid_base_operations;
2763 inode->i_flags|=S_IMMUTABLE;
2765 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
2766 ARRAY_SIZE(tgid_base_stuff)));
2768 d_set_d_op(dentry, &pid_dentry_operations);
2770 d_add(dentry, inode);
2771 /* Close the race of the process dying before we return the dentry */
2772 if (pid_revalidate(dentry, 0))
2778 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
2781 struct task_struct *task;
2783 struct pid_namespace *ns;
2785 tgid = name_to_int(dentry);
2789 ns = dentry->d_sb->s_fs_info;
2791 task = find_task_by_pid_ns(tgid, ns);
2793 get_task_struct(task);
2798 result = proc_pid_instantiate(dir, dentry, task, NULL);
2799 put_task_struct(task);
2801 return ERR_PTR(result);
2805 * Find the first task with tgid >= tgid
2810 struct task_struct *task;
2812 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2817 put_task_struct(iter.task);
2821 pid = find_ge_pid(iter.tgid, ns);
2823 iter.tgid = pid_nr_ns(pid, ns);
2824 iter.task = pid_task(pid, PIDTYPE_PID);
2825 /* What we to know is if the pid we have find is the
2826 * pid of a thread_group_leader. Testing for task
2827 * being a thread_group_leader is the obvious thing
2828 * todo but there is a window when it fails, due to
2829 * the pid transfer logic in de_thread.
2831 * So we perform the straight forward test of seeing
2832 * if the pid we have found is the pid of a thread
2833 * group leader, and don't worry if the task we have
2834 * found doesn't happen to be a thread group leader.
2835 * As we don't care in the case of readdir.
2837 if (!iter.task || !has_group_leader_pid(iter.task)) {
2841 get_task_struct(iter.task);
2847 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 1)
2849 /* for the /proc/ directory itself, after non-process stuff has been done */
2850 int proc_pid_readdir(struct file *file, struct dir_context *ctx)
2852 struct tgid_iter iter;
2853 struct pid_namespace *ns = file->f_dentry->d_sb->s_fs_info;
2854 loff_t pos = ctx->pos;
2856 if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
2859 if (pos == TGID_OFFSET - 1) {
2860 struct inode *inode = ns->proc_self->d_inode;
2861 if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
2865 iter.tgid = pos - TGID_OFFSET;
2868 for (iter = next_tgid(ns, iter);
2870 iter.tgid += 1, iter = next_tgid(ns, iter)) {
2871 char name[PROC_NUMBUF];
2873 if (!has_pid_permissions(ns, iter.task, 2))
2876 len = snprintf(name, sizeof(name), "%d", iter.tgid);
2877 ctx->pos = iter.tgid + TGID_OFFSET;
2878 if (!proc_fill_cache(file, ctx, name, len,
2879 proc_pid_instantiate, iter.task, NULL)) {
2880 put_task_struct(iter.task);
2884 ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
2891 static const struct pid_entry tid_base_stuff[] = {
2892 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2893 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2894 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2895 REG("environ", S_IRUSR, proc_environ_operations),
2896 INF("auxv", S_IRUSR, proc_pid_auxv),
2897 ONE("status", S_IRUGO, proc_pid_status),
2898 ONE("personality", S_IRUGO, proc_pid_personality),
2899 INF("limits", S_IRUGO, proc_pid_limits),
2900 #ifdef CONFIG_SCHED_DEBUG
2901 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2903 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2904 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2905 INF("syscall", S_IRUGO, proc_pid_syscall),
2907 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2908 ONE("stat", S_IRUGO, proc_tid_stat),
2909 ONE("statm", S_IRUGO, proc_pid_statm),
2910 REG("maps", S_IRUGO, proc_tid_maps_operations),
2911 #ifdef CONFIG_CHECKPOINT_RESTORE
2912 REG("children", S_IRUGO, proc_tid_children_operations),
2915 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
2917 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2918 LNK("cwd", proc_cwd_link),
2919 LNK("root", proc_root_link),
2920 LNK("exe", proc_exe_link),
2921 REG("mounts", S_IRUGO, proc_mounts_operations),
2922 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2923 #ifdef CONFIG_PROC_PAGE_MONITOR
2924 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2925 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
2926 REG("pagemap", S_IRUGO, proc_pagemap_operations),
2928 #ifdef CONFIG_SECURITY
2929 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2931 #ifdef CONFIG_KALLSYMS
2932 INF("wchan", S_IRUGO, proc_pid_wchan),
2934 #ifdef CONFIG_STACKTRACE
2935 ONE("stack", S_IRUGO, proc_pid_stack),
2937 #ifdef CONFIG_SCHEDSTATS
2938 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2940 #ifdef CONFIG_LATENCYTOP
2941 REG("latency", S_IRUGO, proc_lstats_operations),
2943 #ifdef CONFIG_PROC_PID_CPUSET
2944 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2946 #ifdef CONFIG_CGROUPS
2947 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2949 INF("oom_score", S_IRUGO, proc_oom_score),
2950 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
2951 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2952 #ifdef CONFIG_AUDITSYSCALL
2953 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2954 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2956 #ifdef CONFIG_FAULT_INJECTION
2957 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2959 #ifdef CONFIG_TASK_IO_ACCOUNTING
2960 INF("io", S_IRUSR, proc_tid_io_accounting),
2962 #ifdef CONFIG_HARDWALL
2963 INF("hardwall", S_IRUGO, proc_pid_hardwall),
2965 #ifdef CONFIG_USER_NS
2966 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
2967 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
2968 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
2972 static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
2974 return proc_pident_readdir(file, ctx,
2975 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
2978 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2980 return proc_pident_lookup(dir, dentry,
2981 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
2984 static const struct file_operations proc_tid_base_operations = {
2985 .read = generic_read_dir,
2986 .iterate = proc_tid_base_readdir,
2987 .llseek = default_llseek,
2990 static const struct inode_operations proc_tid_base_inode_operations = {
2991 .lookup = proc_tid_base_lookup,
2992 .getattr = pid_getattr,
2993 .setattr = proc_setattr,
2996 static int proc_task_instantiate(struct inode *dir,
2997 struct dentry *dentry, struct task_struct *task, const void *ptr)
2999 struct inode *inode;
3000 inode = proc_pid_make_inode(dir->i_sb, task);
3004 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3005 inode->i_op = &proc_tid_base_inode_operations;
3006 inode->i_fop = &proc_tid_base_operations;
3007 inode->i_flags|=S_IMMUTABLE;
3009 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3010 ARRAY_SIZE(tid_base_stuff)));
3012 d_set_d_op(dentry, &pid_dentry_operations);
3014 d_add(dentry, inode);
3015 /* Close the race of the process dying before we return the dentry */
3016 if (pid_revalidate(dentry, 0))
3022 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3024 int result = -ENOENT;
3025 struct task_struct *task;
3026 struct task_struct *leader = get_proc_task(dir);
3028 struct pid_namespace *ns;
3033 tid = name_to_int(dentry);
3037 ns = dentry->d_sb->s_fs_info;
3039 task = find_task_by_pid_ns(tid, ns);
3041 get_task_struct(task);
3045 if (!same_thread_group(leader, task))
3048 result = proc_task_instantiate(dir, dentry, task, NULL);
3050 put_task_struct(task);
3052 put_task_struct(leader);
3054 return ERR_PTR(result);
3058 * Find the first tid of a thread group to return to user space.
3060 * Usually this is just the thread group leader, but if the users
3061 * buffer was too small or there was a seek into the middle of the
3062 * directory we have more work todo.
3064 * In the case of a short read we start with find_task_by_pid.
3066 * In the case of a seek we start with the leader and walk nr
3069 static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3070 struct pid_namespace *ns)
3072 struct task_struct *pos, *task;
3073 unsigned long nr = f_pos;
3075 if (nr != f_pos) /* 32bit overflow? */
3079 task = pid_task(pid, PIDTYPE_PID);
3083 /* Attempt to start with the tid of a thread */
3085 pos = find_task_by_pid_ns(tid, ns);
3086 if (pos && same_thread_group(pos, task))
3090 /* If nr exceeds the number of threads there is nothing todo */
3091 if (nr >= get_nr_threads(task))
3094 /* If we haven't found our starting place yet start
3095 * with the leader and walk nr threads forward.
3097 pos = task = task->group_leader;
3101 } while_each_thread(task, pos);
3106 get_task_struct(pos);
3113 * Find the next thread in the thread list.
3114 * Return NULL if there is an error or no next thread.
3116 * The reference to the input task_struct is released.
3118 static struct task_struct *next_tid(struct task_struct *start)
3120 struct task_struct *pos = NULL;
3122 if (pid_alive(start)) {
3123 pos = next_thread(start);
3124 if (thread_group_leader(pos))
3127 get_task_struct(pos);
3130 put_task_struct(start);
3134 /* for the /proc/TGID/task/ directories */
3135 static int proc_task_readdir(struct file *file, struct dir_context *ctx)
3137 struct inode *inode = file_inode(file);
3138 struct task_struct *task;
3139 struct pid_namespace *ns;
3142 if (proc_inode_is_dead(inode))
3145 if (!dir_emit_dots(file, ctx))
3148 /* f_version caches the tgid value that the last readdir call couldn't
3149 * return. lseek aka telldir automagically resets f_version to 0.
3151 ns = file->f_dentry->d_sb->s_fs_info;
3152 tid = (int)file->f_version;
3153 file->f_version = 0;
3154 for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
3156 task = next_tid(task), ctx->pos++) {
3157 char name[PROC_NUMBUF];
3159 tid = task_pid_nr_ns(task, ns);
3160 len = snprintf(name, sizeof(name), "%d", tid);
3161 if (!proc_fill_cache(file, ctx, name, len,
3162 proc_task_instantiate, task, NULL)) {
3163 /* returning this tgid failed, save it as the first
3164 * pid for the next readir call */
3165 file->f_version = (u64)tid;
3166 put_task_struct(task);
3174 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3176 struct inode *inode = dentry->d_inode;
3177 struct task_struct *p = get_proc_task(inode);
3178 generic_fillattr(inode, stat);
3181 stat->nlink += get_nr_threads(p);
3188 static const struct inode_operations proc_task_inode_operations = {
3189 .lookup = proc_task_lookup,
3190 .getattr = proc_task_getattr,
3191 .setattr = proc_setattr,
3192 .permission = proc_pid_permission,
3195 static const struct file_operations proc_task_operations = {
3196 .read = generic_read_dir,
3197 .iterate = proc_task_readdir,
3198 .llseek = default_llseek,