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/cgroup.h>
77 #include <linux/cpuset.h>
78 #include <linux/audit.h>
79 #include <linux/poll.h>
80 #include <linux/nsproxy.h>
81 #include <linux/oom.h>
82 #include <linux/elf.h>
83 #include <linux/pid_namespace.h>
84 #include <linux/user_namespace.h>
85 #include <linux/fs_struct.h>
86 #include <linux/slab.h>
87 #include <linux/flex_array.h>
88 #ifdef CONFIG_HARDWALL
89 #include <asm/hardwall.h>
91 #include <trace/events/oom.h>
95 * Implementing inode permission operations in /proc is almost
96 * certainly an error. Permission checks need to happen during
97 * each system call not at open time. The reason is that most of
98 * what we wish to check for permissions in /proc varies at runtime.
100 * The classic example of a problem is opening file descriptors
101 * in /proc for a task before it execs a suid executable.
108 const struct inode_operations *iop;
109 const struct file_operations *fop;
113 #define NOD(NAME, MODE, IOP, FOP, OP) { \
115 .len = sizeof(NAME) - 1, \
122 #define DIR(NAME, MODE, iops, fops) \
123 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
124 #define LNK(NAME, get_link) \
125 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
126 &proc_pid_link_inode_operations, NULL, \
127 { .proc_get_link = get_link } )
128 #define REG(NAME, MODE, fops) \
129 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
130 #define INF(NAME, MODE, read) \
131 NOD(NAME, (S_IFREG|(MODE)), \
132 NULL, &proc_info_file_operations, \
133 { .proc_read = read } )
134 #define ONE(NAME, MODE, show) \
135 NOD(NAME, (S_IFREG|(MODE)), \
136 NULL, &proc_single_file_operations, \
137 { .proc_show = show } )
139 static int proc_fd_permission(struct inode *inode, int mask);
142 * Count the number of hardlinks for the pid_entry table, excluding the .
145 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
152 for (i = 0; i < n; ++i) {
153 if (S_ISDIR(entries[i].mode))
160 static int get_task_root(struct task_struct *task, struct path *root)
162 int result = -ENOENT;
166 get_fs_root(task->fs, root);
173 static int proc_cwd_link(struct dentry *dentry, struct path *path)
175 struct task_struct *task = get_proc_task(dentry->d_inode);
176 int result = -ENOENT;
181 get_fs_pwd(task->fs, path);
185 put_task_struct(task);
190 static int proc_root_link(struct dentry *dentry, struct path *path)
192 struct task_struct *task = get_proc_task(dentry->d_inode);
193 int result = -ENOENT;
196 result = get_task_root(task, path);
197 put_task_struct(task);
202 struct mm_struct *mm_for_maps(struct task_struct *task)
204 return mm_access(task, PTRACE_MODE_READ);
207 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
211 struct mm_struct *mm = get_task_mm(task);
215 goto out_mm; /* Shh! No looking before we're done */
217 len = mm->arg_end - mm->arg_start;
222 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
224 // If the nul at the end of args has been overwritten, then
225 // assume application is using setproctitle(3).
226 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
227 len = strnlen(buffer, res);
231 len = mm->env_end - mm->env_start;
232 if (len > PAGE_SIZE - res)
233 len = PAGE_SIZE - res;
234 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
235 res = strnlen(buffer, res);
244 static int proc_pid_auxv(struct task_struct *task, char *buffer)
246 struct mm_struct *mm = mm_for_maps(task);
247 int res = PTR_ERR(mm);
248 if (mm && !IS_ERR(mm)) {
249 unsigned int nwords = 0;
252 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
253 res = nwords * sizeof(mm->saved_auxv[0]);
256 memcpy(buffer, mm->saved_auxv, res);
263 #ifdef CONFIG_KALLSYMS
265 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
266 * Returns the resolved symbol. If that fails, simply return the address.
268 static int proc_pid_wchan(struct task_struct *task, char *buffer)
271 char symname[KSYM_NAME_LEN];
273 wchan = get_wchan(task);
275 if (lookup_symbol_name(wchan, symname) < 0)
276 if (!ptrace_may_access(task, PTRACE_MODE_READ))
279 return sprintf(buffer, "%lu", wchan);
281 return sprintf(buffer, "%s", symname);
283 #endif /* CONFIG_KALLSYMS */
285 static int lock_trace(struct task_struct *task)
287 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
290 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
291 mutex_unlock(&task->signal->cred_guard_mutex);
297 static void unlock_trace(struct task_struct *task)
299 mutex_unlock(&task->signal->cred_guard_mutex);
302 #ifdef CONFIG_STACKTRACE
304 #define MAX_STACK_TRACE_DEPTH 64
306 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
307 struct pid *pid, struct task_struct *task)
309 struct stack_trace trace;
310 unsigned long *entries;
314 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
318 trace.nr_entries = 0;
319 trace.max_entries = MAX_STACK_TRACE_DEPTH;
320 trace.entries = entries;
323 err = lock_trace(task);
325 save_stack_trace_tsk(task, &trace);
327 for (i = 0; i < trace.nr_entries; i++) {
328 seq_printf(m, "[<%pK>] %pS\n",
329 (void *)entries[i], (void *)entries[i]);
339 #ifdef CONFIG_SCHEDSTATS
341 * Provides /proc/PID/schedstat
343 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
345 return sprintf(buffer, "%llu %llu %lu\n",
346 (unsigned long long)task->se.sum_exec_runtime,
347 (unsigned long long)task->sched_info.run_delay,
348 task->sched_info.pcount);
352 #ifdef CONFIG_LATENCYTOP
353 static int lstats_show_proc(struct seq_file *m, void *v)
356 struct inode *inode = m->private;
357 struct task_struct *task = get_proc_task(inode);
361 seq_puts(m, "Latency Top version : v0.1\n");
362 for (i = 0; i < 32; i++) {
363 struct latency_record *lr = &task->latency_record[i];
364 if (lr->backtrace[0]) {
366 seq_printf(m, "%i %li %li",
367 lr->count, lr->time, lr->max);
368 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
369 unsigned long bt = lr->backtrace[q];
374 seq_printf(m, " %ps", (void *)bt);
380 put_task_struct(task);
384 static int lstats_open(struct inode *inode, struct file *file)
386 return single_open(file, lstats_show_proc, inode);
389 static ssize_t lstats_write(struct file *file, const char __user *buf,
390 size_t count, loff_t *offs)
392 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
396 clear_all_latency_tracing(task);
397 put_task_struct(task);
402 static const struct file_operations proc_lstats_operations = {
405 .write = lstats_write,
407 .release = single_release,
412 static int proc_oom_score(struct task_struct *task, char *buffer)
414 unsigned long points = 0;
416 read_lock(&tasklist_lock);
418 points = oom_badness(task, NULL, NULL,
419 totalram_pages + total_swap_pages);
420 read_unlock(&tasklist_lock);
421 return sprintf(buffer, "%lu\n", points);
429 static const struct limit_names lnames[RLIM_NLIMITS] = {
430 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
431 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
432 [RLIMIT_DATA] = {"Max data size", "bytes"},
433 [RLIMIT_STACK] = {"Max stack size", "bytes"},
434 [RLIMIT_CORE] = {"Max core file size", "bytes"},
435 [RLIMIT_RSS] = {"Max resident set", "bytes"},
436 [RLIMIT_NPROC] = {"Max processes", "processes"},
437 [RLIMIT_NOFILE] = {"Max open files", "files"},
438 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
439 [RLIMIT_AS] = {"Max address space", "bytes"},
440 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
441 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
442 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
443 [RLIMIT_NICE] = {"Max nice priority", NULL},
444 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
445 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
448 /* Display limits for a process */
449 static int proc_pid_limits(struct task_struct *task, char *buffer)
454 char *bufptr = buffer;
456 struct rlimit rlim[RLIM_NLIMITS];
458 if (!lock_task_sighand(task, &flags))
460 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
461 unlock_task_sighand(task, &flags);
464 * print the file header
466 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
467 "Limit", "Soft Limit", "Hard Limit", "Units");
469 for (i = 0; i < RLIM_NLIMITS; i++) {
470 if (rlim[i].rlim_cur == RLIM_INFINITY)
471 count += sprintf(&bufptr[count], "%-25s %-20s ",
472 lnames[i].name, "unlimited");
474 count += sprintf(&bufptr[count], "%-25s %-20lu ",
475 lnames[i].name, rlim[i].rlim_cur);
477 if (rlim[i].rlim_max == RLIM_INFINITY)
478 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
480 count += sprintf(&bufptr[count], "%-20lu ",
484 count += sprintf(&bufptr[count], "%-10s\n",
487 count += sprintf(&bufptr[count], "\n");
493 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
494 static int proc_pid_syscall(struct task_struct *task, char *buffer)
497 unsigned long args[6], sp, pc;
498 int res = lock_trace(task);
502 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
503 res = sprintf(buffer, "running\n");
505 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
507 res = sprintf(buffer,
508 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
510 args[0], args[1], args[2], args[3], args[4], args[5],
515 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
517 /************************************************************************/
518 /* Here the fs part begins */
519 /************************************************************************/
521 /* permission checks */
522 static int proc_fd_access_allowed(struct inode *inode)
524 struct task_struct *task;
526 /* Allow access to a task's file descriptors if it is us or we
527 * may use ptrace attach to the process and find out that
530 task = get_proc_task(inode);
532 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
533 put_task_struct(task);
538 int proc_setattr(struct dentry *dentry, struct iattr *attr)
541 struct inode *inode = dentry->d_inode;
543 if (attr->ia_valid & ATTR_MODE)
546 error = inode_change_ok(inode, attr);
550 if ((attr->ia_valid & ATTR_SIZE) &&
551 attr->ia_size != i_size_read(inode)) {
552 error = vmtruncate(inode, attr->ia_size);
557 setattr_copy(inode, attr);
558 mark_inode_dirty(inode);
563 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
564 * or euid/egid (for hide_pid_min=2)?
566 static bool has_pid_permissions(struct pid_namespace *pid,
567 struct task_struct *task,
570 if (pid->hide_pid < hide_pid_min)
572 if (in_group_p(pid->pid_gid))
574 return ptrace_may_access(task, PTRACE_MODE_READ);
578 static int proc_pid_permission(struct inode *inode, int mask)
580 struct pid_namespace *pid = inode->i_sb->s_fs_info;
581 struct task_struct *task;
584 task = get_proc_task(inode);
587 has_perms = has_pid_permissions(pid, task, 1);
588 put_task_struct(task);
591 if (pid->hide_pid == 2) {
593 * Let's make getdents(), stat(), and open()
594 * consistent with each other. If a process
595 * may not stat() a file, it shouldn't be seen
603 return generic_permission(inode, mask);
608 static const struct inode_operations proc_def_inode_operations = {
609 .setattr = proc_setattr,
612 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
614 static ssize_t proc_info_read(struct file * file, char __user * buf,
615 size_t count, loff_t *ppos)
617 struct inode * inode = file->f_path.dentry->d_inode;
620 struct task_struct *task = get_proc_task(inode);
626 if (count > PROC_BLOCK_SIZE)
627 count = PROC_BLOCK_SIZE;
630 if (!(page = __get_free_page(GFP_TEMPORARY)))
633 length = PROC_I(inode)->op.proc_read(task, (char*)page);
636 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
639 put_task_struct(task);
644 static const struct file_operations proc_info_file_operations = {
645 .read = proc_info_read,
646 .llseek = generic_file_llseek,
649 static int proc_single_show(struct seq_file *m, void *v)
651 struct inode *inode = m->private;
652 struct pid_namespace *ns;
654 struct task_struct *task;
657 ns = inode->i_sb->s_fs_info;
658 pid = proc_pid(inode);
659 task = get_pid_task(pid, PIDTYPE_PID);
663 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
665 put_task_struct(task);
669 static int proc_single_open(struct inode *inode, struct file *filp)
671 return single_open(filp, proc_single_show, inode);
674 static const struct file_operations proc_single_file_operations = {
675 .open = proc_single_open,
678 .release = single_release,
681 static int mem_open(struct inode* inode, struct file* file)
683 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
684 struct mm_struct *mm;
689 mm = mm_access(task, PTRACE_MODE_ATTACH);
690 put_task_struct(task);
696 /* ensure this mm_struct can't be freed */
697 atomic_inc(&mm->mm_count);
698 /* but do not pin its memory */
702 /* OK to pass negative loff_t, we can catch out-of-range */
703 file->f_mode |= FMODE_UNSIGNED_OFFSET;
704 file->private_data = mm;
709 static ssize_t mem_rw(struct file *file, char __user *buf,
710 size_t count, loff_t *ppos, int write)
712 struct mm_struct *mm = file->private_data;
713 unsigned long addr = *ppos;
720 page = (char *)__get_free_page(GFP_TEMPORARY);
725 if (!atomic_inc_not_zero(&mm->mm_users))
729 int this_len = min_t(int, count, PAGE_SIZE);
731 if (write && copy_from_user(page, buf, this_len)) {
736 this_len = access_remote_vm(mm, addr, page, this_len, write);
743 if (!write && copy_to_user(buf, page, this_len)) {
757 free_page((unsigned long) page);
761 static ssize_t mem_read(struct file *file, char __user *buf,
762 size_t count, loff_t *ppos)
764 return mem_rw(file, buf, count, ppos, 0);
767 static ssize_t mem_write(struct file *file, const char __user *buf,
768 size_t count, loff_t *ppos)
770 return mem_rw(file, (char __user*)buf, count, ppos, 1);
773 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
777 file->f_pos = offset;
780 file->f_pos += offset;
785 force_successful_syscall_return();
789 static int mem_release(struct inode *inode, struct file *file)
791 struct mm_struct *mm = file->private_data;
797 static const struct file_operations proc_mem_operations = {
802 .release = mem_release,
805 static ssize_t environ_read(struct file *file, char __user *buf,
806 size_t count, loff_t *ppos)
808 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
810 unsigned long src = *ppos;
812 struct mm_struct *mm;
818 page = (char *)__get_free_page(GFP_TEMPORARY);
823 mm = mm_for_maps(task);
825 if (!mm || IS_ERR(mm))
830 int this_len, retval, max_len;
832 this_len = mm->env_end - (mm->env_start + src);
837 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
838 this_len = (this_len > max_len) ? max_len : this_len;
840 retval = access_process_vm(task, (mm->env_start + src),
848 if (copy_to_user(buf, page, retval)) {
862 free_page((unsigned long) page);
864 put_task_struct(task);
869 static const struct file_operations proc_environ_operations = {
870 .read = environ_read,
871 .llseek = generic_file_llseek,
874 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
875 size_t count, loff_t *ppos)
877 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
878 char buffer[PROC_NUMBUF];
880 int oom_adjust = OOM_DISABLE;
886 if (lock_task_sighand(task, &flags)) {
887 oom_adjust = task->signal->oom_adj;
888 unlock_task_sighand(task, &flags);
891 put_task_struct(task);
893 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
895 return simple_read_from_buffer(buf, count, ppos, buffer, len);
898 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
899 size_t count, loff_t *ppos)
901 struct task_struct *task;
902 char buffer[PROC_NUMBUF];
907 memset(buffer, 0, sizeof(buffer));
908 if (count > sizeof(buffer) - 1)
909 count = sizeof(buffer) - 1;
910 if (copy_from_user(buffer, buf, count)) {
915 err = kstrtoint(strstrip(buffer), 0, &oom_adjust);
918 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
919 oom_adjust != OOM_DISABLE) {
924 task = get_proc_task(file->f_path.dentry->d_inode);
936 if (!lock_task_sighand(task, &flags)) {
941 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
947 * Warn that /proc/pid/oom_adj is deprecated, see
948 * Documentation/feature-removal-schedule.txt.
950 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
951 current->comm, task_pid_nr(current), task_pid_nr(task),
953 task->signal->oom_adj = oom_adjust;
955 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
956 * value is always attainable.
958 if (task->signal->oom_adj == OOM_ADJUST_MAX)
959 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
961 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
963 trace_oom_score_adj_update(task);
965 unlock_task_sighand(task, &flags);
968 put_task_struct(task);
970 return err < 0 ? err : count;
973 static const struct file_operations proc_oom_adjust_operations = {
974 .read = oom_adjust_read,
975 .write = oom_adjust_write,
976 .llseek = generic_file_llseek,
979 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
980 size_t count, loff_t *ppos)
982 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
983 char buffer[PROC_NUMBUF];
984 int oom_score_adj = OOM_SCORE_ADJ_MIN;
990 if (lock_task_sighand(task, &flags)) {
991 oom_score_adj = task->signal->oom_score_adj;
992 unlock_task_sighand(task, &flags);
994 put_task_struct(task);
995 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
996 return simple_read_from_buffer(buf, count, ppos, buffer, len);
999 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1000 size_t count, loff_t *ppos)
1002 struct task_struct *task;
1003 char buffer[PROC_NUMBUF];
1004 unsigned long flags;
1008 memset(buffer, 0, sizeof(buffer));
1009 if (count > sizeof(buffer) - 1)
1010 count = sizeof(buffer) - 1;
1011 if (copy_from_user(buffer, buf, count)) {
1016 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1019 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1020 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1025 task = get_proc_task(file->f_path.dentry->d_inode);
1037 if (!lock_task_sighand(task, &flags)) {
1042 if (oom_score_adj < task->signal->oom_score_adj_min &&
1043 !capable(CAP_SYS_RESOURCE)) {
1048 task->signal->oom_score_adj = oom_score_adj;
1049 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1050 task->signal->oom_score_adj_min = oom_score_adj;
1051 trace_oom_score_adj_update(task);
1053 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1054 * always attainable.
1056 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1057 task->signal->oom_adj = OOM_DISABLE;
1059 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1062 unlock_task_sighand(task, &flags);
1065 put_task_struct(task);
1067 return err < 0 ? err : count;
1070 static const struct file_operations proc_oom_score_adj_operations = {
1071 .read = oom_score_adj_read,
1072 .write = oom_score_adj_write,
1073 .llseek = default_llseek,
1076 #ifdef CONFIG_AUDITSYSCALL
1077 #define TMPBUFLEN 21
1078 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1079 size_t count, loff_t *ppos)
1081 struct inode * inode = file->f_path.dentry->d_inode;
1082 struct task_struct *task = get_proc_task(inode);
1084 char tmpbuf[TMPBUFLEN];
1088 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1089 audit_get_loginuid(task));
1090 put_task_struct(task);
1091 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1094 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1095 size_t count, loff_t *ppos)
1097 struct inode * inode = file->f_path.dentry->d_inode;
1103 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1109 if (count >= PAGE_SIZE)
1110 count = PAGE_SIZE - 1;
1113 /* No partial writes. */
1116 page = (char*)__get_free_page(GFP_TEMPORARY);
1120 if (copy_from_user(page, buf, count))
1124 loginuid = simple_strtoul(page, &tmp, 10);
1130 length = audit_set_loginuid(loginuid);
1131 if (likely(length == 0))
1135 free_page((unsigned long) page);
1139 static const struct file_operations proc_loginuid_operations = {
1140 .read = proc_loginuid_read,
1141 .write = proc_loginuid_write,
1142 .llseek = generic_file_llseek,
1145 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1146 size_t count, loff_t *ppos)
1148 struct inode * inode = file->f_path.dentry->d_inode;
1149 struct task_struct *task = get_proc_task(inode);
1151 char tmpbuf[TMPBUFLEN];
1155 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1156 audit_get_sessionid(task));
1157 put_task_struct(task);
1158 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1161 static const struct file_operations proc_sessionid_operations = {
1162 .read = proc_sessionid_read,
1163 .llseek = generic_file_llseek,
1167 #ifdef CONFIG_FAULT_INJECTION
1168 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1169 size_t count, loff_t *ppos)
1171 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1172 char buffer[PROC_NUMBUF];
1178 make_it_fail = task->make_it_fail;
1179 put_task_struct(task);
1181 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1183 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1186 static ssize_t proc_fault_inject_write(struct file * file,
1187 const char __user * buf, size_t count, loff_t *ppos)
1189 struct task_struct *task;
1190 char buffer[PROC_NUMBUF], *end;
1193 if (!capable(CAP_SYS_RESOURCE))
1195 memset(buffer, 0, sizeof(buffer));
1196 if (count > sizeof(buffer) - 1)
1197 count = sizeof(buffer) - 1;
1198 if (copy_from_user(buffer, buf, count))
1200 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1203 task = get_proc_task(file->f_dentry->d_inode);
1206 task->make_it_fail = make_it_fail;
1207 put_task_struct(task);
1212 static const struct file_operations proc_fault_inject_operations = {
1213 .read = proc_fault_inject_read,
1214 .write = proc_fault_inject_write,
1215 .llseek = generic_file_llseek,
1220 #ifdef CONFIG_SCHED_DEBUG
1222 * Print out various scheduling related per-task fields:
1224 static int sched_show(struct seq_file *m, void *v)
1226 struct inode *inode = m->private;
1227 struct task_struct *p;
1229 p = get_proc_task(inode);
1232 proc_sched_show_task(p, m);
1240 sched_write(struct file *file, const char __user *buf,
1241 size_t count, loff_t *offset)
1243 struct inode *inode = file->f_path.dentry->d_inode;
1244 struct task_struct *p;
1246 p = get_proc_task(inode);
1249 proc_sched_set_task(p);
1256 static int sched_open(struct inode *inode, struct file *filp)
1258 return single_open(filp, sched_show, inode);
1261 static const struct file_operations proc_pid_sched_operations = {
1264 .write = sched_write,
1265 .llseek = seq_lseek,
1266 .release = single_release,
1271 #ifdef CONFIG_SCHED_AUTOGROUP
1273 * Print out autogroup related information:
1275 static int sched_autogroup_show(struct seq_file *m, void *v)
1277 struct inode *inode = m->private;
1278 struct task_struct *p;
1280 p = get_proc_task(inode);
1283 proc_sched_autogroup_show_task(p, m);
1291 sched_autogroup_write(struct file *file, const char __user *buf,
1292 size_t count, loff_t *offset)
1294 struct inode *inode = file->f_path.dentry->d_inode;
1295 struct task_struct *p;
1296 char buffer[PROC_NUMBUF];
1300 memset(buffer, 0, sizeof(buffer));
1301 if (count > sizeof(buffer) - 1)
1302 count = sizeof(buffer) - 1;
1303 if (copy_from_user(buffer, buf, count))
1306 err = kstrtoint(strstrip(buffer), 0, &nice);
1310 p = get_proc_task(inode);
1314 err = proc_sched_autogroup_set_nice(p, nice);
1323 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1327 ret = single_open(filp, sched_autogroup_show, NULL);
1329 struct seq_file *m = filp->private_data;
1336 static const struct file_operations proc_pid_sched_autogroup_operations = {
1337 .open = sched_autogroup_open,
1339 .write = sched_autogroup_write,
1340 .llseek = seq_lseek,
1341 .release = single_release,
1344 #endif /* CONFIG_SCHED_AUTOGROUP */
1346 static ssize_t comm_write(struct file *file, const char __user *buf,
1347 size_t count, loff_t *offset)
1349 struct inode *inode = file->f_path.dentry->d_inode;
1350 struct task_struct *p;
1351 char buffer[TASK_COMM_LEN];
1353 memset(buffer, 0, sizeof(buffer));
1354 if (count > sizeof(buffer) - 1)
1355 count = sizeof(buffer) - 1;
1356 if (copy_from_user(buffer, buf, count))
1359 p = get_proc_task(inode);
1363 if (same_thread_group(current, p))
1364 set_task_comm(p, buffer);
1373 static int comm_show(struct seq_file *m, void *v)
1375 struct inode *inode = m->private;
1376 struct task_struct *p;
1378 p = get_proc_task(inode);
1383 seq_printf(m, "%s\n", p->comm);
1391 static int comm_open(struct inode *inode, struct file *filp)
1393 return single_open(filp, comm_show, inode);
1396 static const struct file_operations proc_pid_set_comm_operations = {
1399 .write = comm_write,
1400 .llseek = seq_lseek,
1401 .release = single_release,
1404 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1406 struct task_struct *task;
1407 struct mm_struct *mm;
1408 struct file *exe_file;
1410 task = get_proc_task(dentry->d_inode);
1413 mm = get_task_mm(task);
1414 put_task_struct(task);
1417 exe_file = get_mm_exe_file(mm);
1420 *exe_path = exe_file->f_path;
1421 path_get(&exe_file->f_path);
1428 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1430 struct inode *inode = dentry->d_inode;
1431 int error = -EACCES;
1433 /* We don't need a base pointer in the /proc filesystem */
1434 path_put(&nd->path);
1436 /* Are we allowed to snoop on the tasks file descriptors? */
1437 if (!proc_fd_access_allowed(inode))
1440 error = PROC_I(inode)->op.proc_get_link(dentry, &nd->path);
1442 return ERR_PTR(error);
1445 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1447 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1454 pathname = d_path(path, tmp, PAGE_SIZE);
1455 len = PTR_ERR(pathname);
1456 if (IS_ERR(pathname))
1458 len = tmp + PAGE_SIZE - 1 - pathname;
1462 if (copy_to_user(buffer, pathname, len))
1465 free_page((unsigned long)tmp);
1469 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1471 int error = -EACCES;
1472 struct inode *inode = dentry->d_inode;
1475 /* Are we allowed to snoop on the tasks file descriptors? */
1476 if (!proc_fd_access_allowed(inode))
1479 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1483 error = do_proc_readlink(&path, buffer, buflen);
1489 static const struct inode_operations proc_pid_link_inode_operations = {
1490 .readlink = proc_pid_readlink,
1491 .follow_link = proc_pid_follow_link,
1492 .setattr = proc_setattr,
1496 /* building an inode */
1498 static int task_dumpable(struct task_struct *task)
1501 struct mm_struct *mm;
1506 dumpable = get_dumpable(mm);
1513 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1515 struct inode * inode;
1516 struct proc_inode *ei;
1517 const struct cred *cred;
1519 /* We need a new inode */
1521 inode = new_inode(sb);
1527 inode->i_ino = get_next_ino();
1528 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1529 inode->i_op = &proc_def_inode_operations;
1532 * grab the reference to task.
1534 ei->pid = get_task_pid(task, PIDTYPE_PID);
1538 if (task_dumpable(task)) {
1540 cred = __task_cred(task);
1541 inode->i_uid = cred->euid;
1542 inode->i_gid = cred->egid;
1545 security_task_to_inode(task, inode);
1555 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1557 struct inode *inode = dentry->d_inode;
1558 struct task_struct *task;
1559 const struct cred *cred;
1560 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
1562 generic_fillattr(inode, stat);
1565 stat->uid = GLOBAL_ROOT_UID;
1566 stat->gid = GLOBAL_ROOT_GID;
1567 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1569 if (!has_pid_permissions(pid, task, 2)) {
1572 * This doesn't prevent learning whether PID exists,
1573 * it only makes getattr() consistent with readdir().
1577 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1578 task_dumpable(task)) {
1579 cred = __task_cred(task);
1580 stat->uid = cred->euid;
1581 stat->gid = cred->egid;
1591 * Exceptional case: normally we are not allowed to unhash a busy
1592 * directory. In this case, however, we can do it - no aliasing problems
1593 * due to the way we treat inodes.
1595 * Rewrite the inode's ownerships here because the owning task may have
1596 * performed a setuid(), etc.
1598 * Before the /proc/pid/status file was created the only way to read
1599 * the effective uid of a /process was to stat /proc/pid. Reading
1600 * /proc/pid/status is slow enough that procps and other packages
1601 * kept stating /proc/pid. To keep the rules in /proc simple I have
1602 * made this apply to all per process world readable and executable
1605 int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1607 struct inode *inode;
1608 struct task_struct *task;
1609 const struct cred *cred;
1611 if (nd && nd->flags & LOOKUP_RCU)
1614 inode = dentry->d_inode;
1615 task = get_proc_task(inode);
1618 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1619 task_dumpable(task)) {
1621 cred = __task_cred(task);
1622 inode->i_uid = cred->euid;
1623 inode->i_gid = cred->egid;
1626 inode->i_uid = GLOBAL_ROOT_UID;
1627 inode->i_gid = GLOBAL_ROOT_GID;
1629 inode->i_mode &= ~(S_ISUID | S_ISGID);
1630 security_task_to_inode(task, inode);
1631 put_task_struct(task);
1638 static int pid_delete_dentry(const struct dentry * dentry)
1640 /* Is the task we represent dead?
1641 * If so, then don't put the dentry on the lru list,
1642 * kill it immediately.
1644 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1647 const struct dentry_operations pid_dentry_operations =
1649 .d_revalidate = pid_revalidate,
1650 .d_delete = pid_delete_dentry,
1656 * Fill a directory entry.
1658 * If possible create the dcache entry and derive our inode number and
1659 * file type from dcache entry.
1661 * Since all of the proc inode numbers are dynamically generated, the inode
1662 * numbers do not exist until the inode is cache. This means creating the
1663 * the dcache entry in readdir is necessary to keep the inode numbers
1664 * reported by readdir in sync with the inode numbers reported
1667 int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1668 const char *name, int len,
1669 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1671 struct dentry *child, *dir = filp->f_path.dentry;
1672 struct inode *inode;
1675 unsigned type = DT_UNKNOWN;
1679 qname.hash = full_name_hash(name, len);
1681 child = d_lookup(dir, &qname);
1684 new = d_alloc(dir, &qname);
1686 child = instantiate(dir->d_inode, new, task, ptr);
1693 if (!child || IS_ERR(child) || !child->d_inode)
1694 goto end_instantiate;
1695 inode = child->d_inode;
1698 type = inode->i_mode >> 12;
1703 ino = find_inode_number(dir, &qname);
1706 return filldir(dirent, name, len, filp->f_pos, ino, type);
1709 static unsigned name_to_int(struct dentry *dentry)
1711 const char *name = dentry->d_name.name;
1712 int len = dentry->d_name.len;
1715 if (len > 1 && *name == '0')
1718 unsigned c = *name++ - '0';
1721 if (n >= (~0U-9)/10)
1731 #define PROC_FDINFO_MAX 64
1733 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1735 struct task_struct *task = get_proc_task(inode);
1736 struct files_struct *files = NULL;
1738 int fd = proc_fd(inode);
1741 files = get_files_struct(task);
1742 put_task_struct(task);
1746 * We are not taking a ref to the file structure, so we must
1749 spin_lock(&files->file_lock);
1750 file = fcheck_files(files, fd);
1752 unsigned int f_flags;
1753 struct fdtable *fdt;
1755 fdt = files_fdtable(files);
1756 f_flags = file->f_flags & ~O_CLOEXEC;
1757 if (close_on_exec(fd, fdt))
1758 f_flags |= O_CLOEXEC;
1761 *path = file->f_path;
1762 path_get(&file->f_path);
1765 snprintf(info, PROC_FDINFO_MAX,
1768 (long long) file->f_pos,
1770 spin_unlock(&files->file_lock);
1771 put_files_struct(files);
1774 spin_unlock(&files->file_lock);
1775 put_files_struct(files);
1780 static int proc_fd_link(struct dentry *dentry, struct path *path)
1782 return proc_fd_info(dentry->d_inode, path, NULL);
1785 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1787 struct inode *inode;
1788 struct task_struct *task;
1790 struct files_struct *files;
1791 const struct cred *cred;
1793 if (nd && nd->flags & LOOKUP_RCU)
1796 inode = dentry->d_inode;
1797 task = get_proc_task(inode);
1798 fd = proc_fd(inode);
1801 files = get_files_struct(task);
1804 if (fcheck_files(files, fd)) {
1806 put_files_struct(files);
1807 if (task_dumpable(task)) {
1809 cred = __task_cred(task);
1810 inode->i_uid = cred->euid;
1811 inode->i_gid = cred->egid;
1814 inode->i_uid = GLOBAL_ROOT_UID;
1815 inode->i_gid = GLOBAL_ROOT_GID;
1817 inode->i_mode &= ~(S_ISUID | S_ISGID);
1818 security_task_to_inode(task, inode);
1819 put_task_struct(task);
1823 put_files_struct(files);
1825 put_task_struct(task);
1831 static const struct dentry_operations tid_fd_dentry_operations =
1833 .d_revalidate = tid_fd_revalidate,
1834 .d_delete = pid_delete_dentry,
1837 static struct dentry *proc_fd_instantiate(struct inode *dir,
1838 struct dentry *dentry, struct task_struct *task, const void *ptr)
1840 unsigned fd = *(const unsigned *)ptr;
1842 struct files_struct *files;
1843 struct inode *inode;
1844 struct proc_inode *ei;
1845 struct dentry *error = ERR_PTR(-ENOENT);
1847 inode = proc_pid_make_inode(dir->i_sb, task);
1852 files = get_files_struct(task);
1855 inode->i_mode = S_IFLNK;
1858 * We are not taking a ref to the file structure, so we must
1861 spin_lock(&files->file_lock);
1862 file = fcheck_files(files, fd);
1865 if (file->f_mode & FMODE_READ)
1866 inode->i_mode |= S_IRUSR | S_IXUSR;
1867 if (file->f_mode & FMODE_WRITE)
1868 inode->i_mode |= S_IWUSR | S_IXUSR;
1869 spin_unlock(&files->file_lock);
1870 put_files_struct(files);
1872 inode->i_op = &proc_pid_link_inode_operations;
1874 ei->op.proc_get_link = proc_fd_link;
1875 d_set_d_op(dentry, &tid_fd_dentry_operations);
1876 d_add(dentry, inode);
1877 /* Close the race of the process dying before we return the dentry */
1878 if (tid_fd_revalidate(dentry, NULL))
1884 spin_unlock(&files->file_lock);
1885 put_files_struct(files);
1891 static struct dentry *proc_lookupfd_common(struct inode *dir,
1892 struct dentry *dentry,
1893 instantiate_t instantiate)
1895 struct task_struct *task = get_proc_task(dir);
1896 unsigned fd = name_to_int(dentry);
1897 struct dentry *result = ERR_PTR(-ENOENT);
1904 result = instantiate(dir, dentry, task, &fd);
1906 put_task_struct(task);
1911 static int proc_readfd_common(struct file * filp, void * dirent,
1912 filldir_t filldir, instantiate_t instantiate)
1914 struct dentry *dentry = filp->f_path.dentry;
1915 struct inode *inode = dentry->d_inode;
1916 struct task_struct *p = get_proc_task(inode);
1917 unsigned int fd, ino;
1919 struct files_struct * files;
1929 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1933 ino = parent_ino(dentry);
1934 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1938 files = get_files_struct(p);
1942 for (fd = filp->f_pos-2;
1943 fd < files_fdtable(files)->max_fds;
1944 fd++, filp->f_pos++) {
1945 char name[PROC_NUMBUF];
1948 if (!fcheck_files(files, fd))
1952 len = snprintf(name, sizeof(name), "%d", fd);
1953 if (proc_fill_cache(filp, dirent, filldir,
1954 name, len, instantiate,
1962 put_files_struct(files);
1970 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1971 struct nameidata *nd)
1973 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1976 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1978 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1981 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1982 size_t len, loff_t *ppos)
1984 char tmp[PROC_FDINFO_MAX];
1985 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
1987 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1991 static const struct file_operations proc_fdinfo_file_operations = {
1992 .open = nonseekable_open,
1993 .read = proc_fdinfo_read,
1994 .llseek = no_llseek,
1997 static const struct file_operations proc_fd_operations = {
1998 .read = generic_read_dir,
1999 .readdir = proc_readfd,
2000 .llseek = default_llseek,
2003 #ifdef CONFIG_CHECKPOINT_RESTORE
2006 * dname_to_vma_addr - maps a dentry name into two unsigned longs
2007 * which represent vma start and end addresses.
2009 static int dname_to_vma_addr(struct dentry *dentry,
2010 unsigned long *start, unsigned long *end)
2012 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
2018 static int map_files_d_revalidate(struct dentry *dentry, struct nameidata *nd)
2020 unsigned long vm_start, vm_end;
2021 bool exact_vma_exists = false;
2022 struct mm_struct *mm = NULL;
2023 struct task_struct *task;
2024 const struct cred *cred;
2025 struct inode *inode;
2028 if (nd && nd->flags & LOOKUP_RCU)
2031 if (!capable(CAP_SYS_ADMIN)) {
2036 inode = dentry->d_inode;
2037 task = get_proc_task(inode);
2041 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2044 mm = get_task_mm(task);
2048 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
2049 down_read(&mm->mmap_sem);
2050 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
2051 up_read(&mm->mmap_sem);
2056 if (exact_vma_exists) {
2057 if (task_dumpable(task)) {
2059 cred = __task_cred(task);
2060 inode->i_uid = cred->euid;
2061 inode->i_gid = cred->egid;
2064 inode->i_uid = GLOBAL_ROOT_UID;
2065 inode->i_gid = GLOBAL_ROOT_GID;
2067 security_task_to_inode(task, inode);
2072 put_task_struct(task);
2081 static const struct dentry_operations tid_map_files_dentry_operations = {
2082 .d_revalidate = map_files_d_revalidate,
2083 .d_delete = pid_delete_dentry,
2086 static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
2088 unsigned long vm_start, vm_end;
2089 struct vm_area_struct *vma;
2090 struct task_struct *task;
2091 struct mm_struct *mm;
2095 task = get_proc_task(dentry->d_inode);
2099 mm = get_task_mm(task);
2100 put_task_struct(task);
2104 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2108 down_read(&mm->mmap_sem);
2109 vma = find_exact_vma(mm, vm_start, vm_end);
2110 if (vma && vma->vm_file) {
2111 *path = vma->vm_file->f_path;
2115 up_read(&mm->mmap_sem);
2123 struct map_files_info {
2126 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
2129 static struct dentry *
2130 proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
2131 struct task_struct *task, const void *ptr)
2133 const struct file *file = ptr;
2134 struct proc_inode *ei;
2135 struct inode *inode;
2138 return ERR_PTR(-ENOENT);
2140 inode = proc_pid_make_inode(dir->i_sb, task);
2142 return ERR_PTR(-ENOENT);
2145 ei->op.proc_get_link = proc_map_files_get_link;
2147 inode->i_op = &proc_pid_link_inode_operations;
2149 inode->i_mode = S_IFLNK;
2151 if (file->f_mode & FMODE_READ)
2152 inode->i_mode |= S_IRUSR;
2153 if (file->f_mode & FMODE_WRITE)
2154 inode->i_mode |= S_IWUSR;
2156 d_set_d_op(dentry, &tid_map_files_dentry_operations);
2157 d_add(dentry, inode);
2162 static struct dentry *proc_map_files_lookup(struct inode *dir,
2163 struct dentry *dentry, struct nameidata *nd)
2165 unsigned long vm_start, vm_end;
2166 struct vm_area_struct *vma;
2167 struct task_struct *task;
2168 struct dentry *result;
2169 struct mm_struct *mm;
2171 result = ERR_PTR(-EACCES);
2172 if (!capable(CAP_SYS_ADMIN))
2175 result = ERR_PTR(-ENOENT);
2176 task = get_proc_task(dir);
2180 result = ERR_PTR(-EACCES);
2181 if (lock_trace(task))
2184 result = ERR_PTR(-ENOENT);
2185 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2188 mm = get_task_mm(task);
2192 down_read(&mm->mmap_sem);
2193 vma = find_exact_vma(mm, vm_start, vm_end);
2197 result = proc_map_files_instantiate(dir, dentry, task, vma->vm_file);
2200 up_read(&mm->mmap_sem);
2205 put_task_struct(task);
2210 static const struct inode_operations proc_map_files_inode_operations = {
2211 .lookup = proc_map_files_lookup,
2212 .permission = proc_fd_permission,
2213 .setattr = proc_setattr,
2217 proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
2219 struct dentry *dentry = filp->f_path.dentry;
2220 struct inode *inode = dentry->d_inode;
2221 struct vm_area_struct *vma;
2222 struct task_struct *task;
2223 struct mm_struct *mm;
2228 if (!capable(CAP_SYS_ADMIN))
2232 task = get_proc_task(inode);
2237 if (lock_trace(task))
2241 switch (filp->f_pos) {
2244 if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0)
2248 ino = parent_ino(dentry);
2249 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2254 unsigned long nr_files, pos, i;
2255 struct flex_array *fa = NULL;
2256 struct map_files_info info;
2257 struct map_files_info *p;
2259 mm = get_task_mm(task);
2262 down_read(&mm->mmap_sem);
2267 * We need two passes here:
2269 * 1) Collect vmas of mapped files with mmap_sem taken
2270 * 2) Release mmap_sem and instantiate entries
2272 * otherwise we get lockdep complained, since filldir()
2273 * routine might require mmap_sem taken in might_fault().
2276 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2277 if (vma->vm_file && ++pos > filp->f_pos)
2282 fa = flex_array_alloc(sizeof(info), nr_files,
2284 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2288 flex_array_free(fa);
2289 up_read(&mm->mmap_sem);
2293 for (i = 0, vma = mm->mmap, pos = 2; vma;
2294 vma = vma->vm_next) {
2297 if (++pos <= filp->f_pos)
2300 get_file(vma->vm_file);
2301 info.file = vma->vm_file;
2302 info.len = snprintf(info.name,
2303 sizeof(info.name), "%lx-%lx",
2304 vma->vm_start, vma->vm_end);
2305 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2309 up_read(&mm->mmap_sem);
2311 for (i = 0; i < nr_files; i++) {
2312 p = flex_array_get(fa, i);
2313 ret = proc_fill_cache(filp, dirent, filldir,
2315 proc_map_files_instantiate,
2322 for (; i < nr_files; i++) {
2324 * In case of error don't forget
2325 * to put rest of file refs.
2327 p = flex_array_get(fa, i);
2331 flex_array_free(fa);
2339 put_task_struct(task);
2344 static const struct file_operations proc_map_files_operations = {
2345 .read = generic_read_dir,
2346 .readdir = proc_map_files_readdir,
2347 .llseek = default_llseek,
2350 #endif /* CONFIG_CHECKPOINT_RESTORE */
2353 * /proc/pid/fd needs a special permission handler so that a process can still
2354 * access /proc/self/fd after it has executed a setuid().
2356 static int proc_fd_permission(struct inode *inode, int mask)
2358 int rv = generic_permission(inode, mask);
2361 if (task_pid(current) == proc_pid(inode))
2367 * proc directories can do almost nothing..
2369 static const struct inode_operations proc_fd_inode_operations = {
2370 .lookup = proc_lookupfd,
2371 .permission = proc_fd_permission,
2372 .setattr = proc_setattr,
2375 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2376 struct dentry *dentry, struct task_struct *task, const void *ptr)
2378 unsigned fd = *(unsigned *)ptr;
2379 struct inode *inode;
2380 struct proc_inode *ei;
2381 struct dentry *error = ERR_PTR(-ENOENT);
2383 inode = proc_pid_make_inode(dir->i_sb, task);
2388 inode->i_mode = S_IFREG | S_IRUSR;
2389 inode->i_fop = &proc_fdinfo_file_operations;
2390 d_set_d_op(dentry, &tid_fd_dentry_operations);
2391 d_add(dentry, inode);
2392 /* Close the race of the process dying before we return the dentry */
2393 if (tid_fd_revalidate(dentry, NULL))
2400 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2401 struct dentry *dentry,
2402 struct nameidata *nd)
2404 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2407 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2409 return proc_readfd_common(filp, dirent, filldir,
2410 proc_fdinfo_instantiate);
2413 static const struct file_operations proc_fdinfo_operations = {
2414 .read = generic_read_dir,
2415 .readdir = proc_readfdinfo,
2416 .llseek = default_llseek,
2420 * proc directories can do almost nothing..
2422 static const struct inode_operations proc_fdinfo_inode_operations = {
2423 .lookup = proc_lookupfdinfo,
2424 .setattr = proc_setattr,
2428 static struct dentry *proc_pident_instantiate(struct inode *dir,
2429 struct dentry *dentry, struct task_struct *task, const void *ptr)
2431 const struct pid_entry *p = ptr;
2432 struct inode *inode;
2433 struct proc_inode *ei;
2434 struct dentry *error = ERR_PTR(-ENOENT);
2436 inode = proc_pid_make_inode(dir->i_sb, task);
2441 inode->i_mode = p->mode;
2442 if (S_ISDIR(inode->i_mode))
2443 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2445 inode->i_op = p->iop;
2447 inode->i_fop = p->fop;
2449 d_set_d_op(dentry, &pid_dentry_operations);
2450 d_add(dentry, inode);
2451 /* Close the race of the process dying before we return the dentry */
2452 if (pid_revalidate(dentry, NULL))
2458 static struct dentry *proc_pident_lookup(struct inode *dir,
2459 struct dentry *dentry,
2460 const struct pid_entry *ents,
2463 struct dentry *error;
2464 struct task_struct *task = get_proc_task(dir);
2465 const struct pid_entry *p, *last;
2467 error = ERR_PTR(-ENOENT);
2473 * Yes, it does not scale. And it should not. Don't add
2474 * new entries into /proc/<tgid>/ without very good reasons.
2476 last = &ents[nents - 1];
2477 for (p = ents; p <= last; p++) {
2478 if (p->len != dentry->d_name.len)
2480 if (!memcmp(dentry->d_name.name, p->name, p->len))
2486 error = proc_pident_instantiate(dir, dentry, task, p);
2488 put_task_struct(task);
2493 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2494 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2496 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2497 proc_pident_instantiate, task, p);
2500 static int proc_pident_readdir(struct file *filp,
2501 void *dirent, filldir_t filldir,
2502 const struct pid_entry *ents, unsigned int nents)
2505 struct dentry *dentry = filp->f_path.dentry;
2506 struct inode *inode = dentry->d_inode;
2507 struct task_struct *task = get_proc_task(inode);
2508 const struct pid_entry *p, *last;
2521 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2527 ino = parent_ino(dentry);
2528 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2540 last = &ents[nents - 1];
2542 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2551 put_task_struct(task);
2556 #ifdef CONFIG_SECURITY
2557 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2558 size_t count, loff_t *ppos)
2560 struct inode * inode = file->f_path.dentry->d_inode;
2563 struct task_struct *task = get_proc_task(inode);
2568 length = security_getprocattr(task,
2569 (char*)file->f_path.dentry->d_name.name,
2571 put_task_struct(task);
2573 length = simple_read_from_buffer(buf, count, ppos, p, length);
2578 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2579 size_t count, loff_t *ppos)
2581 struct inode * inode = file->f_path.dentry->d_inode;
2584 struct task_struct *task = get_proc_task(inode);
2589 if (count > PAGE_SIZE)
2592 /* No partial writes. */
2598 page = (char*)__get_free_page(GFP_TEMPORARY);
2603 if (copy_from_user(page, buf, count))
2606 /* Guard against adverse ptrace interaction */
2607 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2611 length = security_setprocattr(task,
2612 (char*)file->f_path.dentry->d_name.name,
2613 (void*)page, count);
2614 mutex_unlock(&task->signal->cred_guard_mutex);
2616 free_page((unsigned long) page);
2618 put_task_struct(task);
2623 static const struct file_operations proc_pid_attr_operations = {
2624 .read = proc_pid_attr_read,
2625 .write = proc_pid_attr_write,
2626 .llseek = generic_file_llseek,
2629 static const struct pid_entry attr_dir_stuff[] = {
2630 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2631 REG("prev", S_IRUGO, proc_pid_attr_operations),
2632 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2633 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2634 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2635 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2638 static int proc_attr_dir_readdir(struct file * filp,
2639 void * dirent, filldir_t filldir)
2641 return proc_pident_readdir(filp,dirent,filldir,
2642 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2645 static const struct file_operations proc_attr_dir_operations = {
2646 .read = generic_read_dir,
2647 .readdir = proc_attr_dir_readdir,
2648 .llseek = default_llseek,
2651 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2652 struct dentry *dentry, struct nameidata *nd)
2654 return proc_pident_lookup(dir, dentry,
2655 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2658 static const struct inode_operations proc_attr_dir_inode_operations = {
2659 .lookup = proc_attr_dir_lookup,
2660 .getattr = pid_getattr,
2661 .setattr = proc_setattr,
2666 #ifdef CONFIG_ELF_CORE
2667 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2668 size_t count, loff_t *ppos)
2670 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2671 struct mm_struct *mm;
2672 char buffer[PROC_NUMBUF];
2680 mm = get_task_mm(task);
2682 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2683 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2684 MMF_DUMP_FILTER_SHIFT));
2686 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2689 put_task_struct(task);
2694 static ssize_t proc_coredump_filter_write(struct file *file,
2695 const char __user *buf,
2699 struct task_struct *task;
2700 struct mm_struct *mm;
2701 char buffer[PROC_NUMBUF], *end;
2708 memset(buffer, 0, sizeof(buffer));
2709 if (count > sizeof(buffer) - 1)
2710 count = sizeof(buffer) - 1;
2711 if (copy_from_user(buffer, buf, count))
2715 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2718 if (end - buffer == 0)
2722 task = get_proc_task(file->f_dentry->d_inode);
2727 mm = get_task_mm(task);
2731 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2733 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2735 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2740 put_task_struct(task);
2745 static const struct file_operations proc_coredump_filter_operations = {
2746 .read = proc_coredump_filter_read,
2747 .write = proc_coredump_filter_write,
2748 .llseek = generic_file_llseek,
2755 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2758 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2759 pid_t tgid = task_tgid_nr_ns(current, ns);
2760 char tmp[PROC_NUMBUF];
2763 sprintf(tmp, "%d", tgid);
2764 return vfs_readlink(dentry,buffer,buflen,tmp);
2767 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2769 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2770 pid_t tgid = task_tgid_nr_ns(current, ns);
2771 char *name = ERR_PTR(-ENOENT);
2775 name = ERR_PTR(-ENOMEM);
2777 sprintf(name, "%d", tgid);
2779 nd_set_link(nd, name);
2783 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2786 char *s = nd_get_link(nd);
2791 static const struct inode_operations proc_self_inode_operations = {
2792 .readlink = proc_self_readlink,
2793 .follow_link = proc_self_follow_link,
2794 .put_link = proc_self_put_link,
2800 * These are the directory entries in the root directory of /proc
2801 * that properly belong to the /proc filesystem, as they describe
2802 * describe something that is process related.
2804 static const struct pid_entry proc_base_stuff[] = {
2805 NOD("self", S_IFLNK|S_IRWXUGO,
2806 &proc_self_inode_operations, NULL, {}),
2809 static struct dentry *proc_base_instantiate(struct inode *dir,
2810 struct dentry *dentry, struct task_struct *task, const void *ptr)
2812 const struct pid_entry *p = ptr;
2813 struct inode *inode;
2814 struct proc_inode *ei;
2815 struct dentry *error;
2817 /* Allocate the inode */
2818 error = ERR_PTR(-ENOMEM);
2819 inode = new_inode(dir->i_sb);
2823 /* Initialize the inode */
2825 inode->i_ino = get_next_ino();
2826 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2829 * grab the reference to the task.
2831 ei->pid = get_task_pid(task, PIDTYPE_PID);
2835 inode->i_mode = p->mode;
2836 if (S_ISDIR(inode->i_mode))
2837 set_nlink(inode, 2);
2838 if (S_ISLNK(inode->i_mode))
2841 inode->i_op = p->iop;
2843 inode->i_fop = p->fop;
2845 d_add(dentry, inode);
2854 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2856 struct dentry *error;
2857 struct task_struct *task = get_proc_task(dir);
2858 const struct pid_entry *p, *last;
2860 error = ERR_PTR(-ENOENT);
2865 /* Lookup the directory entry */
2866 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2867 for (p = proc_base_stuff; p <= last; p++) {
2868 if (p->len != dentry->d_name.len)
2870 if (!memcmp(dentry->d_name.name, p->name, p->len))
2876 error = proc_base_instantiate(dir, dentry, task, p);
2879 put_task_struct(task);
2884 static int proc_base_fill_cache(struct file *filp, void *dirent,
2885 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2887 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2888 proc_base_instantiate, task, p);
2891 #ifdef CONFIG_TASK_IO_ACCOUNTING
2892 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2894 struct task_io_accounting acct = task->ioac;
2895 unsigned long flags;
2898 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2902 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2907 if (whole && lock_task_sighand(task, &flags)) {
2908 struct task_struct *t = task;
2910 task_io_accounting_add(&acct, &task->signal->ioac);
2911 while_each_thread(task, t)
2912 task_io_accounting_add(&acct, &t->ioac);
2914 unlock_task_sighand(task, &flags);
2916 result = sprintf(buffer,
2921 "read_bytes: %llu\n"
2922 "write_bytes: %llu\n"
2923 "cancelled_write_bytes: %llu\n",
2924 (unsigned long long)acct.rchar,
2925 (unsigned long long)acct.wchar,
2926 (unsigned long long)acct.syscr,
2927 (unsigned long long)acct.syscw,
2928 (unsigned long long)acct.read_bytes,
2929 (unsigned long long)acct.write_bytes,
2930 (unsigned long long)acct.cancelled_write_bytes);
2932 mutex_unlock(&task->signal->cred_guard_mutex);
2936 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2938 return do_io_accounting(task, buffer, 0);
2941 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2943 return do_io_accounting(task, buffer, 1);
2945 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2947 #ifdef CONFIG_USER_NS
2948 static int proc_id_map_open(struct inode *inode, struct file *file,
2949 struct seq_operations *seq_ops)
2951 struct user_namespace *ns = NULL;
2952 struct task_struct *task;
2953 struct seq_file *seq;
2956 task = get_proc_task(inode);
2959 ns = get_user_ns(task_cred_xxx(task, user_ns));
2961 put_task_struct(task);
2966 ret = seq_open(file, seq_ops);
2970 seq = file->private_data;
2980 static int proc_id_map_release(struct inode *inode, struct file *file)
2982 struct seq_file *seq = file->private_data;
2983 struct user_namespace *ns = seq->private;
2985 return seq_release(inode, file);
2988 static int proc_uid_map_open(struct inode *inode, struct file *file)
2990 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2993 static int proc_gid_map_open(struct inode *inode, struct file *file)
2995 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2998 static const struct file_operations proc_uid_map_operations = {
2999 .open = proc_uid_map_open,
3000 .write = proc_uid_map_write,
3002 .llseek = seq_lseek,
3003 .release = proc_id_map_release,
3006 static const struct file_operations proc_gid_map_operations = {
3007 .open = proc_gid_map_open,
3008 .write = proc_gid_map_write,
3010 .llseek = seq_lseek,
3011 .release = proc_id_map_release,
3013 #endif /* CONFIG_USER_NS */
3015 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
3016 struct pid *pid, struct task_struct *task)
3018 int err = lock_trace(task);
3020 seq_printf(m, "%08x\n", task->personality);
3029 static const struct file_operations proc_task_operations;
3030 static const struct inode_operations proc_task_inode_operations;
3032 static const struct pid_entry tgid_base_stuff[] = {
3033 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
3034 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3035 #ifdef CONFIG_CHECKPOINT_RESTORE
3036 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
3038 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3039 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3041 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3043 REG("environ", S_IRUSR, proc_environ_operations),
3044 INF("auxv", S_IRUSR, proc_pid_auxv),
3045 ONE("status", S_IRUGO, proc_pid_status),
3046 ONE("personality", S_IRUGO, proc_pid_personality),
3047 INF("limits", S_IRUGO, proc_pid_limits),
3048 #ifdef CONFIG_SCHED_DEBUG
3049 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3051 #ifdef CONFIG_SCHED_AUTOGROUP
3052 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
3054 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3055 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3056 INF("syscall", S_IRUGO, proc_pid_syscall),
3058 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3059 ONE("stat", S_IRUGO, proc_tgid_stat),
3060 ONE("statm", S_IRUGO, proc_pid_statm),
3061 REG("maps", S_IRUGO, proc_pid_maps_operations),
3063 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
3065 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3066 LNK("cwd", proc_cwd_link),
3067 LNK("root", proc_root_link),
3068 LNK("exe", proc_exe_link),
3069 REG("mounts", S_IRUGO, proc_mounts_operations),
3070 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3071 REG("mountstats", S_IRUSR, proc_mountstats_operations),
3072 #ifdef CONFIG_PROC_PAGE_MONITOR
3073 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3074 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
3075 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3077 #ifdef CONFIG_SECURITY
3078 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3080 #ifdef CONFIG_KALLSYMS
3081 INF("wchan", S_IRUGO, proc_pid_wchan),
3083 #ifdef CONFIG_STACKTRACE
3084 ONE("stack", S_IRUGO, proc_pid_stack),
3086 #ifdef CONFIG_SCHEDSTATS
3087 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3089 #ifdef CONFIG_LATENCYTOP
3090 REG("latency", S_IRUGO, proc_lstats_operations),
3092 #ifdef CONFIG_PROC_PID_CPUSET
3093 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3095 #ifdef CONFIG_CGROUPS
3096 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3098 INF("oom_score", S_IRUGO, proc_oom_score),
3099 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3100 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3101 #ifdef CONFIG_AUDITSYSCALL
3102 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3103 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3105 #ifdef CONFIG_FAULT_INJECTION
3106 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3108 #ifdef CONFIG_ELF_CORE
3109 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3111 #ifdef CONFIG_TASK_IO_ACCOUNTING
3112 INF("io", S_IRUSR, proc_tgid_io_accounting),
3114 #ifdef CONFIG_HARDWALL
3115 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3117 #ifdef CONFIG_USER_NS
3118 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3119 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3123 static int proc_tgid_base_readdir(struct file * filp,
3124 void * dirent, filldir_t filldir)
3126 return proc_pident_readdir(filp,dirent,filldir,
3127 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
3130 static const struct file_operations proc_tgid_base_operations = {
3131 .read = generic_read_dir,
3132 .readdir = proc_tgid_base_readdir,
3133 .llseek = default_llseek,
3136 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3137 return proc_pident_lookup(dir, dentry,
3138 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3141 static const struct inode_operations proc_tgid_base_inode_operations = {
3142 .lookup = proc_tgid_base_lookup,
3143 .getattr = pid_getattr,
3144 .setattr = proc_setattr,
3145 .permission = proc_pid_permission,
3148 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
3150 struct dentry *dentry, *leader, *dir;
3151 char buf[PROC_NUMBUF];
3155 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3156 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
3158 shrink_dcache_parent(dentry);
3164 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
3165 leader = d_hash_and_lookup(mnt->mnt_root, &name);
3170 name.len = strlen(name.name);
3171 dir = d_hash_and_lookup(leader, &name);
3173 goto out_put_leader;
3176 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3177 dentry = d_hash_and_lookup(dir, &name);
3179 shrink_dcache_parent(dentry);
3192 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3193 * @task: task that should be flushed.
3195 * When flushing dentries from proc, one needs to flush them from global
3196 * proc (proc_mnt) and from all the namespaces' procs this task was seen
3197 * in. This call is supposed to do all of this job.
3199 * Looks in the dcache for
3201 * /proc/@tgid/task/@pid
3202 * if either directory is present flushes it and all of it'ts children
3205 * It is safe and reasonable to cache /proc entries for a task until
3206 * that task exits. After that they just clog up the dcache with
3207 * useless entries, possibly causing useful dcache entries to be
3208 * flushed instead. This routine is proved to flush those useless
3209 * dcache entries at process exit time.
3211 * NOTE: This routine is just an optimization so it does not guarantee
3212 * that no dcache entries will exist at process exit time it
3213 * just makes it very unlikely that any will persist.
3216 void proc_flush_task(struct task_struct *task)
3219 struct pid *pid, *tgid;
3222 pid = task_pid(task);
3223 tgid = task_tgid(task);
3225 for (i = 0; i <= pid->level; i++) {
3226 upid = &pid->numbers[i];
3227 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
3228 tgid->numbers[i].nr);
3231 upid = &pid->numbers[pid->level];
3233 pid_ns_release_proc(upid->ns);
3236 static struct dentry *proc_pid_instantiate(struct inode *dir,
3237 struct dentry * dentry,
3238 struct task_struct *task, const void *ptr)
3240 struct dentry *error = ERR_PTR(-ENOENT);
3241 struct inode *inode;
3243 inode = proc_pid_make_inode(dir->i_sb, task);
3247 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3248 inode->i_op = &proc_tgid_base_inode_operations;
3249 inode->i_fop = &proc_tgid_base_operations;
3250 inode->i_flags|=S_IMMUTABLE;
3252 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
3253 ARRAY_SIZE(tgid_base_stuff)));
3255 d_set_d_op(dentry, &pid_dentry_operations);
3257 d_add(dentry, inode);
3258 /* Close the race of the process dying before we return the dentry */
3259 if (pid_revalidate(dentry, NULL))
3265 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3267 struct dentry *result;
3268 struct task_struct *task;
3270 struct pid_namespace *ns;
3272 result = proc_base_lookup(dir, dentry);
3273 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3276 tgid = name_to_int(dentry);
3280 ns = dentry->d_sb->s_fs_info;
3282 task = find_task_by_pid_ns(tgid, ns);
3284 get_task_struct(task);
3289 result = proc_pid_instantiate(dir, dentry, task, NULL);
3290 put_task_struct(task);
3296 * Find the first task with tgid >= tgid
3301 struct task_struct *task;
3303 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3308 put_task_struct(iter.task);
3312 pid = find_ge_pid(iter.tgid, ns);
3314 iter.tgid = pid_nr_ns(pid, ns);
3315 iter.task = pid_task(pid, PIDTYPE_PID);
3316 /* What we to know is if the pid we have find is the
3317 * pid of a thread_group_leader. Testing for task
3318 * being a thread_group_leader is the obvious thing
3319 * todo but there is a window when it fails, due to
3320 * the pid transfer logic in de_thread.
3322 * So we perform the straight forward test of seeing
3323 * if the pid we have found is the pid of a thread
3324 * group leader, and don't worry if the task we have
3325 * found doesn't happen to be a thread group leader.
3326 * As we don't care in the case of readdir.
3328 if (!iter.task || !has_group_leader_pid(iter.task)) {
3332 get_task_struct(iter.task);
3338 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
3340 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3341 struct tgid_iter iter)
3343 char name[PROC_NUMBUF];
3344 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
3345 return proc_fill_cache(filp, dirent, filldir, name, len,
3346 proc_pid_instantiate, iter.task, NULL);
3349 static int fake_filldir(void *buf, const char *name, int namelen,
3350 loff_t offset, u64 ino, unsigned d_type)
3355 /* for the /proc/ directory itself, after non-process stuff has been done */
3356 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3359 struct task_struct *reaper;
3360 struct tgid_iter iter;
3361 struct pid_namespace *ns;
3362 filldir_t __filldir;
3364 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3366 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3368 reaper = get_proc_task(filp->f_path.dentry->d_inode);
3372 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
3373 const struct pid_entry *p = &proc_base_stuff[nr];
3374 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
3378 ns = filp->f_dentry->d_sb->s_fs_info;
3380 iter.tgid = filp->f_pos - TGID_OFFSET;
3381 for (iter = next_tgid(ns, iter);
3383 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3384 if (has_pid_permissions(ns, iter.task, 2))
3385 __filldir = filldir;
3387 __filldir = fake_filldir;
3389 filp->f_pos = iter.tgid + TGID_OFFSET;
3390 if (proc_pid_fill_cache(filp, dirent, __filldir, iter) < 0) {
3391 put_task_struct(iter.task);
3395 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3397 put_task_struct(reaper);
3405 static const struct pid_entry tid_base_stuff[] = {
3406 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3407 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3408 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3409 REG("environ", S_IRUSR, proc_environ_operations),
3410 INF("auxv", S_IRUSR, proc_pid_auxv),
3411 ONE("status", S_IRUGO, proc_pid_status),
3412 ONE("personality", S_IRUGO, proc_pid_personality),
3413 INF("limits", S_IRUGO, proc_pid_limits),
3414 #ifdef CONFIG_SCHED_DEBUG
3415 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3417 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3418 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3419 INF("syscall", S_IRUGO, proc_pid_syscall),
3421 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3422 ONE("stat", S_IRUGO, proc_tid_stat),
3423 ONE("statm", S_IRUGO, proc_pid_statm),
3424 REG("maps", S_IRUGO, proc_tid_maps_operations),
3426 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
3428 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3429 LNK("cwd", proc_cwd_link),
3430 LNK("root", proc_root_link),
3431 LNK("exe", proc_exe_link),
3432 REG("mounts", S_IRUGO, proc_mounts_operations),
3433 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3434 #ifdef CONFIG_PROC_PAGE_MONITOR
3435 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3436 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
3437 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3439 #ifdef CONFIG_SECURITY
3440 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3442 #ifdef CONFIG_KALLSYMS
3443 INF("wchan", S_IRUGO, proc_pid_wchan),
3445 #ifdef CONFIG_STACKTRACE
3446 ONE("stack", S_IRUGO, proc_pid_stack),
3448 #ifdef CONFIG_SCHEDSTATS
3449 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3451 #ifdef CONFIG_LATENCYTOP
3452 REG("latency", S_IRUGO, proc_lstats_operations),
3454 #ifdef CONFIG_PROC_PID_CPUSET
3455 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3457 #ifdef CONFIG_CGROUPS
3458 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3460 INF("oom_score", S_IRUGO, proc_oom_score),
3461 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3462 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3463 #ifdef CONFIG_AUDITSYSCALL
3464 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3465 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3467 #ifdef CONFIG_FAULT_INJECTION
3468 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3470 #ifdef CONFIG_TASK_IO_ACCOUNTING
3471 INF("io", S_IRUSR, proc_tid_io_accounting),
3473 #ifdef CONFIG_HARDWALL
3474 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3476 #ifdef CONFIG_USER_NS
3477 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3478 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3482 static int proc_tid_base_readdir(struct file * filp,
3483 void * dirent, filldir_t filldir)
3485 return proc_pident_readdir(filp,dirent,filldir,
3486 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3489 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3490 return proc_pident_lookup(dir, dentry,
3491 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3494 static const struct file_operations proc_tid_base_operations = {
3495 .read = generic_read_dir,
3496 .readdir = proc_tid_base_readdir,
3497 .llseek = default_llseek,
3500 static const struct inode_operations proc_tid_base_inode_operations = {
3501 .lookup = proc_tid_base_lookup,
3502 .getattr = pid_getattr,
3503 .setattr = proc_setattr,
3506 static struct dentry *proc_task_instantiate(struct inode *dir,
3507 struct dentry *dentry, struct task_struct *task, const void *ptr)
3509 struct dentry *error = ERR_PTR(-ENOENT);
3510 struct inode *inode;
3511 inode = proc_pid_make_inode(dir->i_sb, task);
3515 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3516 inode->i_op = &proc_tid_base_inode_operations;
3517 inode->i_fop = &proc_tid_base_operations;
3518 inode->i_flags|=S_IMMUTABLE;
3520 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3521 ARRAY_SIZE(tid_base_stuff)));
3523 d_set_d_op(dentry, &pid_dentry_operations);
3525 d_add(dentry, inode);
3526 /* Close the race of the process dying before we return the dentry */
3527 if (pid_revalidate(dentry, NULL))
3533 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3535 struct dentry *result = ERR_PTR(-ENOENT);
3536 struct task_struct *task;
3537 struct task_struct *leader = get_proc_task(dir);
3539 struct pid_namespace *ns;
3544 tid = name_to_int(dentry);
3548 ns = dentry->d_sb->s_fs_info;
3550 task = find_task_by_pid_ns(tid, ns);
3552 get_task_struct(task);
3556 if (!same_thread_group(leader, task))
3559 result = proc_task_instantiate(dir, dentry, task, NULL);
3561 put_task_struct(task);
3563 put_task_struct(leader);
3569 * Find the first tid of a thread group to return to user space.
3571 * Usually this is just the thread group leader, but if the users
3572 * buffer was too small or there was a seek into the middle of the
3573 * directory we have more work todo.
3575 * In the case of a short read we start with find_task_by_pid.
3577 * In the case of a seek we start with the leader and walk nr
3580 static struct task_struct *first_tid(struct task_struct *leader,
3581 int tid, int nr, struct pid_namespace *ns)
3583 struct task_struct *pos;
3586 /* Attempt to start with the pid of a thread */
3587 if (tid && (nr > 0)) {
3588 pos = find_task_by_pid_ns(tid, ns);
3589 if (pos && (pos->group_leader == leader))
3593 /* If nr exceeds the number of threads there is nothing todo */
3595 if (nr && nr >= get_nr_threads(leader))
3598 /* If we haven't found our starting place yet start
3599 * with the leader and walk nr threads forward.
3601 for (pos = leader; nr > 0; --nr) {
3602 pos = next_thread(pos);
3603 if (pos == leader) {
3609 get_task_struct(pos);
3616 * Find the next thread in the thread list.
3617 * Return NULL if there is an error or no next thread.
3619 * The reference to the input task_struct is released.
3621 static struct task_struct *next_tid(struct task_struct *start)
3623 struct task_struct *pos = NULL;
3625 if (pid_alive(start)) {
3626 pos = next_thread(start);
3627 if (thread_group_leader(pos))
3630 get_task_struct(pos);
3633 put_task_struct(start);
3637 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3638 struct task_struct *task, int tid)
3640 char name[PROC_NUMBUF];
3641 int len = snprintf(name, sizeof(name), "%d", tid);
3642 return proc_fill_cache(filp, dirent, filldir, name, len,
3643 proc_task_instantiate, task, NULL);
3646 /* for the /proc/TGID/task/ directories */
3647 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3649 struct dentry *dentry = filp->f_path.dentry;
3650 struct inode *inode = dentry->d_inode;
3651 struct task_struct *leader = NULL;
3652 struct task_struct *task;
3653 int retval = -ENOENT;
3656 struct pid_namespace *ns;
3658 task = get_proc_task(inode);
3662 if (pid_alive(task)) {
3663 leader = task->group_leader;
3664 get_task_struct(leader);
3667 put_task_struct(task);
3672 switch ((unsigned long)filp->f_pos) {
3675 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3680 ino = parent_ino(dentry);
3681 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3687 /* f_version caches the tgid value that the last readdir call couldn't
3688 * return. lseek aka telldir automagically resets f_version to 0.
3690 ns = filp->f_dentry->d_sb->s_fs_info;
3691 tid = (int)filp->f_version;
3692 filp->f_version = 0;
3693 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3695 task = next_tid(task), filp->f_pos++) {
3696 tid = task_pid_nr_ns(task, ns);
3697 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3698 /* returning this tgid failed, save it as the first
3699 * pid for the next readir call */
3700 filp->f_version = (u64)tid;
3701 put_task_struct(task);
3706 put_task_struct(leader);
3711 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3713 struct inode *inode = dentry->d_inode;
3714 struct task_struct *p = get_proc_task(inode);
3715 generic_fillattr(inode, stat);
3718 stat->nlink += get_nr_threads(p);
3725 static const struct inode_operations proc_task_inode_operations = {
3726 .lookup = proc_task_lookup,
3727 .getattr = proc_task_getattr,
3728 .setattr = proc_setattr,
3729 .permission = proc_pid_permission,
3732 static const struct file_operations proc_task_operations = {
3733 .read = generic_read_dir,
3734 .readdir = proc_task_readdir,
3735 .llseek = default_llseek,