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 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
206 struct mm_struct *mm = get_task_mm(task);
210 goto out_mm; /* Shh! No looking before we're done */
212 len = mm->arg_end - mm->arg_start;
217 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
219 // If the nul at the end of args has been overwritten, then
220 // assume application is using setproctitle(3).
221 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
222 len = strnlen(buffer, res);
226 len = mm->env_end - mm->env_start;
227 if (len > PAGE_SIZE - res)
228 len = PAGE_SIZE - res;
229 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
230 res = strnlen(buffer, res);
239 static int proc_pid_auxv(struct task_struct *task, char *buffer)
241 struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ);
242 int res = PTR_ERR(mm);
243 if (mm && !IS_ERR(mm)) {
244 unsigned int nwords = 0;
247 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
248 res = nwords * sizeof(mm->saved_auxv[0]);
251 memcpy(buffer, mm->saved_auxv, res);
258 #ifdef CONFIG_KALLSYMS
260 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
261 * Returns the resolved symbol. If that fails, simply return the address.
263 static int proc_pid_wchan(struct task_struct *task, char *buffer)
266 char symname[KSYM_NAME_LEN];
268 wchan = get_wchan(task);
270 if (lookup_symbol_name(wchan, symname) < 0)
271 if (!ptrace_may_access(task, PTRACE_MODE_READ))
274 return sprintf(buffer, "%lu", wchan);
276 return sprintf(buffer, "%s", symname);
278 #endif /* CONFIG_KALLSYMS */
280 static int lock_trace(struct task_struct *task)
282 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
285 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
286 mutex_unlock(&task->signal->cred_guard_mutex);
292 static void unlock_trace(struct task_struct *task)
294 mutex_unlock(&task->signal->cred_guard_mutex);
297 #ifdef CONFIG_STACKTRACE
299 #define MAX_STACK_TRACE_DEPTH 64
301 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
302 struct pid *pid, struct task_struct *task)
304 struct stack_trace trace;
305 unsigned long *entries;
309 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
313 trace.nr_entries = 0;
314 trace.max_entries = MAX_STACK_TRACE_DEPTH;
315 trace.entries = entries;
318 err = lock_trace(task);
320 save_stack_trace_tsk(task, &trace);
322 for (i = 0; i < trace.nr_entries; i++) {
323 seq_printf(m, "[<%pK>] %pS\n",
324 (void *)entries[i], (void *)entries[i]);
334 #ifdef CONFIG_SCHEDSTATS
336 * Provides /proc/PID/schedstat
338 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
340 return sprintf(buffer, "%llu %llu %lu\n",
341 (unsigned long long)task->se.sum_exec_runtime,
342 (unsigned long long)task->sched_info.run_delay,
343 task->sched_info.pcount);
347 #ifdef CONFIG_LATENCYTOP
348 static int lstats_show_proc(struct seq_file *m, void *v)
351 struct inode *inode = m->private;
352 struct task_struct *task = get_proc_task(inode);
356 seq_puts(m, "Latency Top version : v0.1\n");
357 for (i = 0; i < 32; i++) {
358 struct latency_record *lr = &task->latency_record[i];
359 if (lr->backtrace[0]) {
361 seq_printf(m, "%i %li %li",
362 lr->count, lr->time, lr->max);
363 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
364 unsigned long bt = lr->backtrace[q];
369 seq_printf(m, " %ps", (void *)bt);
375 put_task_struct(task);
379 static int lstats_open(struct inode *inode, struct file *file)
381 return single_open(file, lstats_show_proc, inode);
384 static ssize_t lstats_write(struct file *file, const char __user *buf,
385 size_t count, loff_t *offs)
387 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
391 clear_all_latency_tracing(task);
392 put_task_struct(task);
397 static const struct file_operations proc_lstats_operations = {
400 .write = lstats_write,
402 .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 if ((attr->ia_valid & ATTR_SIZE) &&
547 attr->ia_size != i_size_read(inode)) {
548 error = vmtruncate(inode, attr->ia_size);
553 setattr_copy(inode, attr);
554 mark_inode_dirty(inode);
559 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
560 * or euid/egid (for hide_pid_min=2)?
562 static bool has_pid_permissions(struct pid_namespace *pid,
563 struct task_struct *task,
566 if (pid->hide_pid < hide_pid_min)
568 if (in_group_p(pid->pid_gid))
570 return ptrace_may_access(task, PTRACE_MODE_READ);
574 static int proc_pid_permission(struct inode *inode, int mask)
576 struct pid_namespace *pid = inode->i_sb->s_fs_info;
577 struct task_struct *task;
580 task = get_proc_task(inode);
583 has_perms = has_pid_permissions(pid, task, 1);
584 put_task_struct(task);
587 if (pid->hide_pid == 2) {
589 * Let's make getdents(), stat(), and open()
590 * consistent with each other. If a process
591 * may not stat() a file, it shouldn't be seen
599 return generic_permission(inode, mask);
604 static const struct inode_operations proc_def_inode_operations = {
605 .setattr = proc_setattr,
608 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
610 static ssize_t proc_info_read(struct file * file, char __user * buf,
611 size_t count, loff_t *ppos)
613 struct inode * inode = file->f_path.dentry->d_inode;
616 struct task_struct *task = get_proc_task(inode);
622 if (count > PROC_BLOCK_SIZE)
623 count = PROC_BLOCK_SIZE;
626 if (!(page = __get_free_page(GFP_TEMPORARY)))
629 length = PROC_I(inode)->op.proc_read(task, (char*)page);
632 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
635 put_task_struct(task);
640 static const struct file_operations proc_info_file_operations = {
641 .read = proc_info_read,
642 .llseek = generic_file_llseek,
645 static int proc_single_show(struct seq_file *m, void *v)
647 struct inode *inode = m->private;
648 struct pid_namespace *ns;
650 struct task_struct *task;
653 ns = inode->i_sb->s_fs_info;
654 pid = proc_pid(inode);
655 task = get_pid_task(pid, PIDTYPE_PID);
659 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
661 put_task_struct(task);
665 static int proc_single_open(struct inode *inode, struct file *filp)
667 return single_open(filp, proc_single_show, inode);
670 static const struct file_operations proc_single_file_operations = {
671 .open = proc_single_open,
674 .release = single_release,
677 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
679 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
680 struct mm_struct *mm;
685 mm = mm_access(task, mode);
686 put_task_struct(task);
692 /* ensure this mm_struct can't be freed */
693 atomic_inc(&mm->mm_count);
694 /* but do not pin its memory */
698 /* OK to pass negative loff_t, we can catch out-of-range */
699 file->f_mode |= FMODE_UNSIGNED_OFFSET;
700 file->private_data = mm;
705 static int mem_open(struct inode *inode, struct file *file)
707 return __mem_open(inode, file, PTRACE_MODE_ATTACH);
710 static ssize_t mem_rw(struct file *file, char __user *buf,
711 size_t count, loff_t *ppos, int write)
713 struct mm_struct *mm = file->private_data;
714 unsigned long addr = *ppos;
721 page = (char *)__get_free_page(GFP_TEMPORARY);
726 if (!atomic_inc_not_zero(&mm->mm_users))
730 int this_len = min_t(int, count, PAGE_SIZE);
732 if (write && copy_from_user(page, buf, this_len)) {
737 this_len = access_remote_vm(mm, addr, page, this_len, write);
744 if (!write && copy_to_user(buf, page, this_len)) {
758 free_page((unsigned long) page);
762 static ssize_t mem_read(struct file *file, char __user *buf,
763 size_t count, loff_t *ppos)
765 return mem_rw(file, buf, count, ppos, 0);
768 static ssize_t mem_write(struct file *file, const char __user *buf,
769 size_t count, loff_t *ppos)
771 return mem_rw(file, (char __user*)buf, count, ppos, 1);
774 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
778 file->f_pos = offset;
781 file->f_pos += offset;
786 force_successful_syscall_return();
790 static int mem_release(struct inode *inode, struct file *file)
792 struct mm_struct *mm = file->private_data;
798 static const struct file_operations proc_mem_operations = {
803 .release = mem_release,
806 static int environ_open(struct inode *inode, struct file *file)
808 return __mem_open(inode, file, PTRACE_MODE_READ);
811 static ssize_t environ_read(struct file *file, char __user *buf,
812 size_t count, loff_t *ppos)
815 unsigned long src = *ppos;
817 struct mm_struct *mm = file->private_data;
822 page = (char *)__get_free_page(GFP_TEMPORARY);
827 if (!atomic_inc_not_zero(&mm->mm_users))
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_remote_vm(mm, (mm->env_start + src),
848 if (copy_to_user(buf, page, retval)) {
862 free_page((unsigned long) page);
866 static const struct file_operations proc_environ_operations = {
867 .open = environ_open,
868 .read = environ_read,
869 .llseek = generic_file_llseek,
870 .release = mem_release,
873 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
874 size_t count, loff_t *ppos)
876 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
877 char buffer[PROC_NUMBUF];
879 int oom_adjust = OOM_DISABLE;
885 if (lock_task_sighand(task, &flags)) {
886 oom_adjust = task->signal->oom_adj;
887 unlock_task_sighand(task, &flags);
890 put_task_struct(task);
892 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
894 return simple_read_from_buffer(buf, count, ppos, buffer, len);
897 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
898 size_t count, loff_t *ppos)
900 struct task_struct *task;
901 char buffer[PROC_NUMBUF];
906 memset(buffer, 0, sizeof(buffer));
907 if (count > sizeof(buffer) - 1)
908 count = sizeof(buffer) - 1;
909 if (copy_from_user(buffer, buf, count)) {
914 err = kstrtoint(strstrip(buffer), 0, &oom_adjust);
917 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
918 oom_adjust != OOM_DISABLE) {
923 task = get_proc_task(file->f_path.dentry->d_inode);
935 if (!lock_task_sighand(task, &flags)) {
940 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
946 * Warn that /proc/pid/oom_adj is deprecated, see
947 * Documentation/feature-removal-schedule.txt.
949 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
950 current->comm, task_pid_nr(current), task_pid_nr(task),
952 task->signal->oom_adj = oom_adjust;
954 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
955 * value is always attainable.
957 if (task->signal->oom_adj == OOM_ADJUST_MAX)
958 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
960 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
962 trace_oom_score_adj_update(task);
964 unlock_task_sighand(task, &flags);
967 put_task_struct(task);
969 return err < 0 ? err : count;
972 static const struct file_operations proc_oom_adjust_operations = {
973 .read = oom_adjust_read,
974 .write = oom_adjust_write,
975 .llseek = generic_file_llseek,
978 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
979 size_t count, loff_t *ppos)
981 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
982 char buffer[PROC_NUMBUF];
983 int oom_score_adj = OOM_SCORE_ADJ_MIN;
989 if (lock_task_sighand(task, &flags)) {
990 oom_score_adj = task->signal->oom_score_adj;
991 unlock_task_sighand(task, &flags);
993 put_task_struct(task);
994 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
995 return simple_read_from_buffer(buf, count, ppos, buffer, len);
998 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
999 size_t count, loff_t *ppos)
1001 struct task_struct *task;
1002 char buffer[PROC_NUMBUF];
1003 unsigned long flags;
1007 memset(buffer, 0, sizeof(buffer));
1008 if (count > sizeof(buffer) - 1)
1009 count = sizeof(buffer) - 1;
1010 if (copy_from_user(buffer, buf, count)) {
1015 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1018 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1019 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1024 task = get_proc_task(file->f_path.dentry->d_inode);
1036 if (!lock_task_sighand(task, &flags)) {
1041 if (oom_score_adj < task->signal->oom_score_adj_min &&
1042 !capable(CAP_SYS_RESOURCE)) {
1047 task->signal->oom_score_adj = oom_score_adj;
1048 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1049 task->signal->oom_score_adj_min = oom_score_adj;
1050 trace_oom_score_adj_update(task);
1052 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1053 * always attainable.
1055 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1056 task->signal->oom_adj = OOM_DISABLE;
1058 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1061 unlock_task_sighand(task, &flags);
1064 put_task_struct(task);
1066 return err < 0 ? err : count;
1069 static const struct file_operations proc_oom_score_adj_operations = {
1070 .read = oom_score_adj_read,
1071 .write = oom_score_adj_write,
1072 .llseek = default_llseek,
1075 #ifdef CONFIG_AUDITSYSCALL
1076 #define TMPBUFLEN 21
1077 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1078 size_t count, loff_t *ppos)
1080 struct inode * inode = file->f_path.dentry->d_inode;
1081 struct task_struct *task = get_proc_task(inode);
1083 char tmpbuf[TMPBUFLEN];
1087 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1088 audit_get_loginuid(task));
1089 put_task_struct(task);
1090 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1093 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1094 size_t count, loff_t *ppos)
1096 struct inode * inode = file->f_path.dentry->d_inode;
1102 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1108 if (count >= PAGE_SIZE)
1109 count = PAGE_SIZE - 1;
1112 /* No partial writes. */
1115 page = (char*)__get_free_page(GFP_TEMPORARY);
1119 if (copy_from_user(page, buf, count))
1123 loginuid = simple_strtoul(page, &tmp, 10);
1129 length = audit_set_loginuid(loginuid);
1130 if (likely(length == 0))
1134 free_page((unsigned long) page);
1138 static const struct file_operations proc_loginuid_operations = {
1139 .read = proc_loginuid_read,
1140 .write = proc_loginuid_write,
1141 .llseek = generic_file_llseek,
1144 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1145 size_t count, loff_t *ppos)
1147 struct inode * inode = file->f_path.dentry->d_inode;
1148 struct task_struct *task = get_proc_task(inode);
1150 char tmpbuf[TMPBUFLEN];
1154 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1155 audit_get_sessionid(task));
1156 put_task_struct(task);
1157 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1160 static const struct file_operations proc_sessionid_operations = {
1161 .read = proc_sessionid_read,
1162 .llseek = generic_file_llseek,
1166 #ifdef CONFIG_FAULT_INJECTION
1167 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1168 size_t count, loff_t *ppos)
1170 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1171 char buffer[PROC_NUMBUF];
1177 make_it_fail = task->make_it_fail;
1178 put_task_struct(task);
1180 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1182 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1185 static ssize_t proc_fault_inject_write(struct file * file,
1186 const char __user * buf, size_t count, loff_t *ppos)
1188 struct task_struct *task;
1189 char buffer[PROC_NUMBUF], *end;
1192 if (!capable(CAP_SYS_RESOURCE))
1194 memset(buffer, 0, sizeof(buffer));
1195 if (count > sizeof(buffer) - 1)
1196 count = sizeof(buffer) - 1;
1197 if (copy_from_user(buffer, buf, count))
1199 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1202 task = get_proc_task(file->f_dentry->d_inode);
1205 task->make_it_fail = make_it_fail;
1206 put_task_struct(task);
1211 static const struct file_operations proc_fault_inject_operations = {
1212 .read = proc_fault_inject_read,
1213 .write = proc_fault_inject_write,
1214 .llseek = generic_file_llseek,
1219 #ifdef CONFIG_SCHED_DEBUG
1221 * Print out various scheduling related per-task fields:
1223 static int sched_show(struct seq_file *m, void *v)
1225 struct inode *inode = m->private;
1226 struct task_struct *p;
1228 p = get_proc_task(inode);
1231 proc_sched_show_task(p, m);
1239 sched_write(struct file *file, const char __user *buf,
1240 size_t count, loff_t *offset)
1242 struct inode *inode = file->f_path.dentry->d_inode;
1243 struct task_struct *p;
1245 p = get_proc_task(inode);
1248 proc_sched_set_task(p);
1255 static int sched_open(struct inode *inode, struct file *filp)
1257 return single_open(filp, sched_show, inode);
1260 static const struct file_operations proc_pid_sched_operations = {
1263 .write = sched_write,
1264 .llseek = seq_lseek,
1265 .release = single_release,
1270 #ifdef CONFIG_SCHED_AUTOGROUP
1272 * Print out autogroup related information:
1274 static int sched_autogroup_show(struct seq_file *m, void *v)
1276 struct inode *inode = m->private;
1277 struct task_struct *p;
1279 p = get_proc_task(inode);
1282 proc_sched_autogroup_show_task(p, m);
1290 sched_autogroup_write(struct file *file, const char __user *buf,
1291 size_t count, loff_t *offset)
1293 struct inode *inode = file->f_path.dentry->d_inode;
1294 struct task_struct *p;
1295 char buffer[PROC_NUMBUF];
1299 memset(buffer, 0, sizeof(buffer));
1300 if (count > sizeof(buffer) - 1)
1301 count = sizeof(buffer) - 1;
1302 if (copy_from_user(buffer, buf, count))
1305 err = kstrtoint(strstrip(buffer), 0, &nice);
1309 p = get_proc_task(inode);
1313 err = proc_sched_autogroup_set_nice(p, nice);
1322 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1326 ret = single_open(filp, sched_autogroup_show, NULL);
1328 struct seq_file *m = filp->private_data;
1335 static const struct file_operations proc_pid_sched_autogroup_operations = {
1336 .open = sched_autogroup_open,
1338 .write = sched_autogroup_write,
1339 .llseek = seq_lseek,
1340 .release = single_release,
1343 #endif /* CONFIG_SCHED_AUTOGROUP */
1345 static ssize_t comm_write(struct file *file, const char __user *buf,
1346 size_t count, loff_t *offset)
1348 struct inode *inode = file->f_path.dentry->d_inode;
1349 struct task_struct *p;
1350 char buffer[TASK_COMM_LEN];
1352 memset(buffer, 0, sizeof(buffer));
1353 if (count > sizeof(buffer) - 1)
1354 count = sizeof(buffer) - 1;
1355 if (copy_from_user(buffer, buf, count))
1358 p = get_proc_task(inode);
1362 if (same_thread_group(current, p))
1363 set_task_comm(p, buffer);
1372 static int comm_show(struct seq_file *m, void *v)
1374 struct inode *inode = m->private;
1375 struct task_struct *p;
1377 p = get_proc_task(inode);
1382 seq_printf(m, "%s\n", p->comm);
1390 static int comm_open(struct inode *inode, struct file *filp)
1392 return single_open(filp, comm_show, inode);
1395 static const struct file_operations proc_pid_set_comm_operations = {
1398 .write = comm_write,
1399 .llseek = seq_lseek,
1400 .release = single_release,
1403 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1405 struct task_struct *task;
1406 struct mm_struct *mm;
1407 struct file *exe_file;
1409 task = get_proc_task(dentry->d_inode);
1412 mm = get_task_mm(task);
1413 put_task_struct(task);
1416 exe_file = get_mm_exe_file(mm);
1419 *exe_path = exe_file->f_path;
1420 path_get(&exe_file->f_path);
1427 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1429 struct inode *inode = dentry->d_inode;
1430 int error = -EACCES;
1432 /* We don't need a base pointer in the /proc filesystem */
1433 path_put(&nd->path);
1435 /* Are we allowed to snoop on the tasks file descriptors? */
1436 if (!proc_fd_access_allowed(inode))
1439 error = PROC_I(inode)->op.proc_get_link(dentry, &nd->path);
1441 return ERR_PTR(error);
1444 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1446 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1453 pathname = d_path(path, tmp, PAGE_SIZE);
1454 len = PTR_ERR(pathname);
1455 if (IS_ERR(pathname))
1457 len = tmp + PAGE_SIZE - 1 - pathname;
1461 if (copy_to_user(buffer, pathname, len))
1464 free_page((unsigned long)tmp);
1468 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1470 int error = -EACCES;
1471 struct inode *inode = dentry->d_inode;
1474 /* Are we allowed to snoop on the tasks file descriptors? */
1475 if (!proc_fd_access_allowed(inode))
1478 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1482 error = do_proc_readlink(&path, buffer, buflen);
1488 static const struct inode_operations proc_pid_link_inode_operations = {
1489 .readlink = proc_pid_readlink,
1490 .follow_link = proc_pid_follow_link,
1491 .setattr = proc_setattr,
1495 /* building an inode */
1497 static int task_dumpable(struct task_struct *task)
1500 struct mm_struct *mm;
1505 dumpable = get_dumpable(mm);
1512 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1514 struct inode * inode;
1515 struct proc_inode *ei;
1516 const struct cred *cred;
1518 /* We need a new inode */
1520 inode = new_inode(sb);
1526 inode->i_ino = get_next_ino();
1527 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1528 inode->i_op = &proc_def_inode_operations;
1531 * grab the reference to task.
1533 ei->pid = get_task_pid(task, PIDTYPE_PID);
1537 if (task_dumpable(task)) {
1539 cred = __task_cred(task);
1540 inode->i_uid = cred->euid;
1541 inode->i_gid = cred->egid;
1544 security_task_to_inode(task, inode);
1554 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1556 struct inode *inode = dentry->d_inode;
1557 struct task_struct *task;
1558 const struct cred *cred;
1559 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
1561 generic_fillattr(inode, stat);
1564 stat->uid = GLOBAL_ROOT_UID;
1565 stat->gid = GLOBAL_ROOT_GID;
1566 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1568 if (!has_pid_permissions(pid, task, 2)) {
1571 * This doesn't prevent learning whether PID exists,
1572 * it only makes getattr() consistent with readdir().
1576 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1577 task_dumpable(task)) {
1578 cred = __task_cred(task);
1579 stat->uid = cred->euid;
1580 stat->gid = cred->egid;
1590 * Exceptional case: normally we are not allowed to unhash a busy
1591 * directory. In this case, however, we can do it - no aliasing problems
1592 * due to the way we treat inodes.
1594 * Rewrite the inode's ownerships here because the owning task may have
1595 * performed a setuid(), etc.
1597 * Before the /proc/pid/status file was created the only way to read
1598 * the effective uid of a /process was to stat /proc/pid. Reading
1599 * /proc/pid/status is slow enough that procps and other packages
1600 * kept stating /proc/pid. To keep the rules in /proc simple I have
1601 * made this apply to all per process world readable and executable
1604 int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1606 struct inode *inode;
1607 struct task_struct *task;
1608 const struct cred *cred;
1610 if (nd && nd->flags & LOOKUP_RCU)
1613 inode = dentry->d_inode;
1614 task = get_proc_task(inode);
1617 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1618 task_dumpable(task)) {
1620 cred = __task_cred(task);
1621 inode->i_uid = cred->euid;
1622 inode->i_gid = cred->egid;
1625 inode->i_uid = GLOBAL_ROOT_UID;
1626 inode->i_gid = GLOBAL_ROOT_GID;
1628 inode->i_mode &= ~(S_ISUID | S_ISGID);
1629 security_task_to_inode(task, inode);
1630 put_task_struct(task);
1637 static int pid_delete_dentry(const struct dentry * dentry)
1639 /* Is the task we represent dead?
1640 * If so, then don't put the dentry on the lru list,
1641 * kill it immediately.
1643 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1646 const struct dentry_operations pid_dentry_operations =
1648 .d_revalidate = pid_revalidate,
1649 .d_delete = pid_delete_dentry,
1655 * Fill a directory entry.
1657 * If possible create the dcache entry and derive our inode number and
1658 * file type from dcache entry.
1660 * Since all of the proc inode numbers are dynamically generated, the inode
1661 * numbers do not exist until the inode is cache. This means creating the
1662 * the dcache entry in readdir is necessary to keep the inode numbers
1663 * reported by readdir in sync with the inode numbers reported
1666 int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1667 const char *name, int len,
1668 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1670 struct dentry *child, *dir = filp->f_path.dentry;
1671 struct inode *inode;
1674 unsigned type = DT_UNKNOWN;
1678 qname.hash = full_name_hash(name, len);
1680 child = d_lookup(dir, &qname);
1683 new = d_alloc(dir, &qname);
1685 child = instantiate(dir->d_inode, new, task, ptr);
1692 if (!child || IS_ERR(child) || !child->d_inode)
1693 goto end_instantiate;
1694 inode = child->d_inode;
1697 type = inode->i_mode >> 12;
1702 ino = find_inode_number(dir, &qname);
1705 return filldir(dirent, name, len, filp->f_pos, ino, type);
1708 static unsigned name_to_int(struct dentry *dentry)
1710 const char *name = dentry->d_name.name;
1711 int len = dentry->d_name.len;
1714 if (len > 1 && *name == '0')
1717 unsigned c = *name++ - '0';
1720 if (n >= (~0U-9)/10)
1730 #define PROC_FDINFO_MAX 64
1732 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1734 struct task_struct *task = get_proc_task(inode);
1735 struct files_struct *files = NULL;
1737 int fd = proc_fd(inode);
1740 files = get_files_struct(task);
1741 put_task_struct(task);
1745 * We are not taking a ref to the file structure, so we must
1748 spin_lock(&files->file_lock);
1749 file = fcheck_files(files, fd);
1751 unsigned int f_flags;
1752 struct fdtable *fdt;
1754 fdt = files_fdtable(files);
1755 f_flags = file->f_flags & ~O_CLOEXEC;
1756 if (close_on_exec(fd, fdt))
1757 f_flags |= O_CLOEXEC;
1760 *path = file->f_path;
1761 path_get(&file->f_path);
1764 snprintf(info, PROC_FDINFO_MAX,
1767 (long long) file->f_pos,
1769 spin_unlock(&files->file_lock);
1770 put_files_struct(files);
1773 spin_unlock(&files->file_lock);
1774 put_files_struct(files);
1779 static int proc_fd_link(struct dentry *dentry, struct path *path)
1781 return proc_fd_info(dentry->d_inode, path, NULL);
1784 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1786 struct inode *inode;
1787 struct task_struct *task;
1789 struct files_struct *files;
1790 const struct cred *cred;
1792 if (nd && nd->flags & LOOKUP_RCU)
1795 inode = dentry->d_inode;
1796 task = get_proc_task(inode);
1797 fd = proc_fd(inode);
1800 files = get_files_struct(task);
1804 file = fcheck_files(files, fd);
1806 unsigned i_mode, f_mode = file->f_mode;
1809 put_files_struct(files);
1811 if (task_dumpable(task)) {
1813 cred = __task_cred(task);
1814 inode->i_uid = cred->euid;
1815 inode->i_gid = cred->egid;
1818 inode->i_uid = GLOBAL_ROOT_UID;
1819 inode->i_gid = GLOBAL_ROOT_GID;
1823 if (f_mode & FMODE_READ)
1824 i_mode |= S_IRUSR | S_IXUSR;
1825 if (f_mode & FMODE_WRITE)
1826 i_mode |= S_IWUSR | S_IXUSR;
1827 inode->i_mode = i_mode;
1829 security_task_to_inode(task, inode);
1830 put_task_struct(task);
1834 put_files_struct(files);
1836 put_task_struct(task);
1842 static const struct dentry_operations tid_fd_dentry_operations =
1844 .d_revalidate = tid_fd_revalidate,
1845 .d_delete = pid_delete_dentry,
1848 static struct dentry *proc_fd_instantiate(struct inode *dir,
1849 struct dentry *dentry, struct task_struct *task, const void *ptr)
1851 unsigned fd = (unsigned long)ptr;
1852 struct inode *inode;
1853 struct proc_inode *ei;
1854 struct dentry *error = ERR_PTR(-ENOENT);
1856 inode = proc_pid_make_inode(dir->i_sb, task);
1862 inode->i_op = &proc_pid_link_inode_operations;
1864 ei->op.proc_get_link = proc_fd_link;
1865 d_set_d_op(dentry, &tid_fd_dentry_operations);
1866 d_add(dentry, inode);
1867 /* Close the race of the process dying before we return the dentry */
1868 if (tid_fd_revalidate(dentry, NULL))
1875 static struct dentry *proc_lookupfd_common(struct inode *dir,
1876 struct dentry *dentry,
1877 instantiate_t instantiate)
1879 struct task_struct *task = get_proc_task(dir);
1880 unsigned fd = name_to_int(dentry);
1881 struct dentry *result = ERR_PTR(-ENOENT);
1888 result = instantiate(dir, dentry, task, (void *)(unsigned long)fd);
1890 put_task_struct(task);
1895 static int proc_readfd_common(struct file * filp, void * dirent,
1896 filldir_t filldir, instantiate_t instantiate)
1898 struct dentry *dentry = filp->f_path.dentry;
1899 struct inode *inode = dentry->d_inode;
1900 struct task_struct *p = get_proc_task(inode);
1901 unsigned int fd, ino;
1903 struct files_struct * files;
1913 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1917 ino = parent_ino(dentry);
1918 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1922 files = get_files_struct(p);
1926 for (fd = filp->f_pos-2;
1927 fd < files_fdtable(files)->max_fds;
1928 fd++, filp->f_pos++) {
1929 char name[PROC_NUMBUF];
1933 if (!fcheck_files(files, fd))
1937 len = snprintf(name, sizeof(name), "%d", fd);
1938 rv = proc_fill_cache(filp, dirent, filldir,
1939 name, len, instantiate, p,
1940 (void *)(unsigned long)fd);
1947 put_files_struct(files);
1955 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1956 struct nameidata *nd)
1958 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1961 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1963 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1966 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1967 size_t len, loff_t *ppos)
1969 char tmp[PROC_FDINFO_MAX];
1970 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
1972 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1976 static const struct file_operations proc_fdinfo_file_operations = {
1977 .open = nonseekable_open,
1978 .read = proc_fdinfo_read,
1979 .llseek = no_llseek,
1982 static const struct file_operations proc_fd_operations = {
1983 .read = generic_read_dir,
1984 .readdir = proc_readfd,
1985 .llseek = default_llseek,
1988 #ifdef CONFIG_CHECKPOINT_RESTORE
1991 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1992 * which represent vma start and end addresses.
1994 static int dname_to_vma_addr(struct dentry *dentry,
1995 unsigned long *start, unsigned long *end)
1997 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
2003 static int map_files_d_revalidate(struct dentry *dentry, struct nameidata *nd)
2005 unsigned long vm_start, vm_end;
2006 bool exact_vma_exists = false;
2007 struct mm_struct *mm = NULL;
2008 struct task_struct *task;
2009 const struct cred *cred;
2010 struct inode *inode;
2013 if (nd && nd->flags & LOOKUP_RCU)
2016 if (!capable(CAP_SYS_ADMIN)) {
2021 inode = dentry->d_inode;
2022 task = get_proc_task(inode);
2026 mm = mm_access(task, PTRACE_MODE_READ);
2027 if (IS_ERR_OR_NULL(mm))
2030 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
2031 down_read(&mm->mmap_sem);
2032 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
2033 up_read(&mm->mmap_sem);
2038 if (exact_vma_exists) {
2039 if (task_dumpable(task)) {
2041 cred = __task_cred(task);
2042 inode->i_uid = cred->euid;
2043 inode->i_gid = cred->egid;
2046 inode->i_uid = GLOBAL_ROOT_UID;
2047 inode->i_gid = GLOBAL_ROOT_GID;
2049 security_task_to_inode(task, inode);
2054 put_task_struct(task);
2063 static const struct dentry_operations tid_map_files_dentry_operations = {
2064 .d_revalidate = map_files_d_revalidate,
2065 .d_delete = pid_delete_dentry,
2068 static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
2070 unsigned long vm_start, vm_end;
2071 struct vm_area_struct *vma;
2072 struct task_struct *task;
2073 struct mm_struct *mm;
2077 task = get_proc_task(dentry->d_inode);
2081 mm = get_task_mm(task);
2082 put_task_struct(task);
2086 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2090 down_read(&mm->mmap_sem);
2091 vma = find_exact_vma(mm, vm_start, vm_end);
2092 if (vma && vma->vm_file) {
2093 *path = vma->vm_file->f_path;
2097 up_read(&mm->mmap_sem);
2105 struct map_files_info {
2108 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
2111 static struct dentry *
2112 proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
2113 struct task_struct *task, const void *ptr)
2115 const struct file *file = ptr;
2116 struct proc_inode *ei;
2117 struct inode *inode;
2120 return ERR_PTR(-ENOENT);
2122 inode = proc_pid_make_inode(dir->i_sb, task);
2124 return ERR_PTR(-ENOENT);
2127 ei->op.proc_get_link = proc_map_files_get_link;
2129 inode->i_op = &proc_pid_link_inode_operations;
2131 inode->i_mode = S_IFLNK;
2133 if (file->f_mode & FMODE_READ)
2134 inode->i_mode |= S_IRUSR;
2135 if (file->f_mode & FMODE_WRITE)
2136 inode->i_mode |= S_IWUSR;
2138 d_set_d_op(dentry, &tid_map_files_dentry_operations);
2139 d_add(dentry, inode);
2144 static struct dentry *proc_map_files_lookup(struct inode *dir,
2145 struct dentry *dentry, struct nameidata *nd)
2147 unsigned long vm_start, vm_end;
2148 struct vm_area_struct *vma;
2149 struct task_struct *task;
2150 struct dentry *result;
2151 struct mm_struct *mm;
2153 result = ERR_PTR(-EACCES);
2154 if (!capable(CAP_SYS_ADMIN))
2157 result = ERR_PTR(-ENOENT);
2158 task = get_proc_task(dir);
2162 result = ERR_PTR(-EACCES);
2163 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2166 result = ERR_PTR(-ENOENT);
2167 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2170 mm = get_task_mm(task);
2174 down_read(&mm->mmap_sem);
2175 vma = find_exact_vma(mm, vm_start, vm_end);
2179 result = proc_map_files_instantiate(dir, dentry, task, vma->vm_file);
2182 up_read(&mm->mmap_sem);
2185 put_task_struct(task);
2190 static const struct inode_operations proc_map_files_inode_operations = {
2191 .lookup = proc_map_files_lookup,
2192 .permission = proc_fd_permission,
2193 .setattr = proc_setattr,
2197 proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
2199 struct dentry *dentry = filp->f_path.dentry;
2200 struct inode *inode = dentry->d_inode;
2201 struct vm_area_struct *vma;
2202 struct task_struct *task;
2203 struct mm_struct *mm;
2208 if (!capable(CAP_SYS_ADMIN))
2212 task = get_proc_task(inode);
2217 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2221 switch (filp->f_pos) {
2224 if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0)
2228 ino = parent_ino(dentry);
2229 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2234 unsigned long nr_files, pos, i;
2235 struct flex_array *fa = NULL;
2236 struct map_files_info info;
2237 struct map_files_info *p;
2239 mm = get_task_mm(task);
2242 down_read(&mm->mmap_sem);
2247 * We need two passes here:
2249 * 1) Collect vmas of mapped files with mmap_sem taken
2250 * 2) Release mmap_sem and instantiate entries
2252 * otherwise we get lockdep complained, since filldir()
2253 * routine might require mmap_sem taken in might_fault().
2256 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2257 if (vma->vm_file && ++pos > filp->f_pos)
2262 fa = flex_array_alloc(sizeof(info), nr_files,
2264 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2268 flex_array_free(fa);
2269 up_read(&mm->mmap_sem);
2273 for (i = 0, vma = mm->mmap, pos = 2; vma;
2274 vma = vma->vm_next) {
2277 if (++pos <= filp->f_pos)
2280 get_file(vma->vm_file);
2281 info.file = vma->vm_file;
2282 info.len = snprintf(info.name,
2283 sizeof(info.name), "%lx-%lx",
2284 vma->vm_start, vma->vm_end);
2285 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2289 up_read(&mm->mmap_sem);
2291 for (i = 0; i < nr_files; i++) {
2292 p = flex_array_get(fa, i);
2293 ret = proc_fill_cache(filp, dirent, filldir,
2295 proc_map_files_instantiate,
2302 for (; i < nr_files; i++) {
2304 * In case of error don't forget
2305 * to put rest of file refs.
2307 p = flex_array_get(fa, i);
2311 flex_array_free(fa);
2317 put_task_struct(task);
2322 static const struct file_operations proc_map_files_operations = {
2323 .read = generic_read_dir,
2324 .readdir = proc_map_files_readdir,
2325 .llseek = default_llseek,
2328 #endif /* CONFIG_CHECKPOINT_RESTORE */
2331 * /proc/pid/fd needs a special permission handler so that a process can still
2332 * access /proc/self/fd after it has executed a setuid().
2334 static int proc_fd_permission(struct inode *inode, int mask)
2336 int rv = generic_permission(inode, mask);
2339 if (task_pid(current) == proc_pid(inode))
2345 * proc directories can do almost nothing..
2347 static const struct inode_operations proc_fd_inode_operations = {
2348 .lookup = proc_lookupfd,
2349 .permission = proc_fd_permission,
2350 .setattr = proc_setattr,
2353 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2354 struct dentry *dentry, struct task_struct *task, const void *ptr)
2356 unsigned fd = (unsigned long)ptr;
2357 struct inode *inode;
2358 struct proc_inode *ei;
2359 struct dentry *error = ERR_PTR(-ENOENT);
2361 inode = proc_pid_make_inode(dir->i_sb, task);
2366 inode->i_mode = S_IFREG | S_IRUSR;
2367 inode->i_fop = &proc_fdinfo_file_operations;
2368 d_set_d_op(dentry, &tid_fd_dentry_operations);
2369 d_add(dentry, inode);
2370 /* Close the race of the process dying before we return the dentry */
2371 if (tid_fd_revalidate(dentry, NULL))
2378 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2379 struct dentry *dentry,
2380 struct nameidata *nd)
2382 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2385 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2387 return proc_readfd_common(filp, dirent, filldir,
2388 proc_fdinfo_instantiate);
2391 static const struct file_operations proc_fdinfo_operations = {
2392 .read = generic_read_dir,
2393 .readdir = proc_readfdinfo,
2394 .llseek = default_llseek,
2398 * proc directories can do almost nothing..
2400 static const struct inode_operations proc_fdinfo_inode_operations = {
2401 .lookup = proc_lookupfdinfo,
2402 .setattr = proc_setattr,
2406 static struct dentry *proc_pident_instantiate(struct inode *dir,
2407 struct dentry *dentry, struct task_struct *task, const void *ptr)
2409 const struct pid_entry *p = ptr;
2410 struct inode *inode;
2411 struct proc_inode *ei;
2412 struct dentry *error = ERR_PTR(-ENOENT);
2414 inode = proc_pid_make_inode(dir->i_sb, task);
2419 inode->i_mode = p->mode;
2420 if (S_ISDIR(inode->i_mode))
2421 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2423 inode->i_op = p->iop;
2425 inode->i_fop = p->fop;
2427 d_set_d_op(dentry, &pid_dentry_operations);
2428 d_add(dentry, inode);
2429 /* Close the race of the process dying before we return the dentry */
2430 if (pid_revalidate(dentry, NULL))
2436 static struct dentry *proc_pident_lookup(struct inode *dir,
2437 struct dentry *dentry,
2438 const struct pid_entry *ents,
2441 struct dentry *error;
2442 struct task_struct *task = get_proc_task(dir);
2443 const struct pid_entry *p, *last;
2445 error = ERR_PTR(-ENOENT);
2451 * Yes, it does not scale. And it should not. Don't add
2452 * new entries into /proc/<tgid>/ without very good reasons.
2454 last = &ents[nents - 1];
2455 for (p = ents; p <= last; p++) {
2456 if (p->len != dentry->d_name.len)
2458 if (!memcmp(dentry->d_name.name, p->name, p->len))
2464 error = proc_pident_instantiate(dir, dentry, task, p);
2466 put_task_struct(task);
2471 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2472 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2474 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2475 proc_pident_instantiate, task, p);
2478 static int proc_pident_readdir(struct file *filp,
2479 void *dirent, filldir_t filldir,
2480 const struct pid_entry *ents, unsigned int nents)
2483 struct dentry *dentry = filp->f_path.dentry;
2484 struct inode *inode = dentry->d_inode;
2485 struct task_struct *task = get_proc_task(inode);
2486 const struct pid_entry *p, *last;
2499 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2505 ino = parent_ino(dentry);
2506 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2518 last = &ents[nents - 1];
2520 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2529 put_task_struct(task);
2534 #ifdef CONFIG_SECURITY
2535 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2536 size_t count, loff_t *ppos)
2538 struct inode * inode = file->f_path.dentry->d_inode;
2541 struct task_struct *task = get_proc_task(inode);
2546 length = security_getprocattr(task,
2547 (char*)file->f_path.dentry->d_name.name,
2549 put_task_struct(task);
2551 length = simple_read_from_buffer(buf, count, ppos, p, length);
2556 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2557 size_t count, loff_t *ppos)
2559 struct inode * inode = file->f_path.dentry->d_inode;
2562 struct task_struct *task = get_proc_task(inode);
2567 if (count > PAGE_SIZE)
2570 /* No partial writes. */
2576 page = (char*)__get_free_page(GFP_TEMPORARY);
2581 if (copy_from_user(page, buf, count))
2584 /* Guard against adverse ptrace interaction */
2585 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2589 length = security_setprocattr(task,
2590 (char*)file->f_path.dentry->d_name.name,
2591 (void*)page, count);
2592 mutex_unlock(&task->signal->cred_guard_mutex);
2594 free_page((unsigned long) page);
2596 put_task_struct(task);
2601 static const struct file_operations proc_pid_attr_operations = {
2602 .read = proc_pid_attr_read,
2603 .write = proc_pid_attr_write,
2604 .llseek = generic_file_llseek,
2607 static const struct pid_entry attr_dir_stuff[] = {
2608 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2609 REG("prev", S_IRUGO, proc_pid_attr_operations),
2610 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2611 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2612 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2613 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2616 static int proc_attr_dir_readdir(struct file * filp,
2617 void * dirent, filldir_t filldir)
2619 return proc_pident_readdir(filp,dirent,filldir,
2620 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2623 static const struct file_operations proc_attr_dir_operations = {
2624 .read = generic_read_dir,
2625 .readdir = proc_attr_dir_readdir,
2626 .llseek = default_llseek,
2629 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2630 struct dentry *dentry, struct nameidata *nd)
2632 return proc_pident_lookup(dir, dentry,
2633 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2636 static const struct inode_operations proc_attr_dir_inode_operations = {
2637 .lookup = proc_attr_dir_lookup,
2638 .getattr = pid_getattr,
2639 .setattr = proc_setattr,
2644 #ifdef CONFIG_ELF_CORE
2645 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2646 size_t count, loff_t *ppos)
2648 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2649 struct mm_struct *mm;
2650 char buffer[PROC_NUMBUF];
2658 mm = get_task_mm(task);
2660 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2661 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2662 MMF_DUMP_FILTER_SHIFT));
2664 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2667 put_task_struct(task);
2672 static ssize_t proc_coredump_filter_write(struct file *file,
2673 const char __user *buf,
2677 struct task_struct *task;
2678 struct mm_struct *mm;
2679 char buffer[PROC_NUMBUF], *end;
2686 memset(buffer, 0, sizeof(buffer));
2687 if (count > sizeof(buffer) - 1)
2688 count = sizeof(buffer) - 1;
2689 if (copy_from_user(buffer, buf, count))
2693 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2696 if (end - buffer == 0)
2700 task = get_proc_task(file->f_dentry->d_inode);
2705 mm = get_task_mm(task);
2709 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2711 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2713 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2718 put_task_struct(task);
2723 static const struct file_operations proc_coredump_filter_operations = {
2724 .read = proc_coredump_filter_read,
2725 .write = proc_coredump_filter_write,
2726 .llseek = generic_file_llseek,
2733 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2736 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2737 pid_t tgid = task_tgid_nr_ns(current, ns);
2738 char tmp[PROC_NUMBUF];
2741 sprintf(tmp, "%d", tgid);
2742 return vfs_readlink(dentry,buffer,buflen,tmp);
2745 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2747 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2748 pid_t tgid = task_tgid_nr_ns(current, ns);
2749 char *name = ERR_PTR(-ENOENT);
2753 name = ERR_PTR(-ENOMEM);
2755 sprintf(name, "%d", tgid);
2757 nd_set_link(nd, name);
2761 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2764 char *s = nd_get_link(nd);
2769 static const struct inode_operations proc_self_inode_operations = {
2770 .readlink = proc_self_readlink,
2771 .follow_link = proc_self_follow_link,
2772 .put_link = proc_self_put_link,
2778 * These are the directory entries in the root directory of /proc
2779 * that properly belong to the /proc filesystem, as they describe
2780 * describe something that is process related.
2782 static const struct pid_entry proc_base_stuff[] = {
2783 NOD("self", S_IFLNK|S_IRWXUGO,
2784 &proc_self_inode_operations, NULL, {}),
2787 static struct dentry *proc_base_instantiate(struct inode *dir,
2788 struct dentry *dentry, struct task_struct *task, const void *ptr)
2790 const struct pid_entry *p = ptr;
2791 struct inode *inode;
2792 struct proc_inode *ei;
2793 struct dentry *error;
2795 /* Allocate the inode */
2796 error = ERR_PTR(-ENOMEM);
2797 inode = new_inode(dir->i_sb);
2801 /* Initialize the inode */
2803 inode->i_ino = get_next_ino();
2804 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2807 * grab the reference to the task.
2809 ei->pid = get_task_pid(task, PIDTYPE_PID);
2813 inode->i_mode = p->mode;
2814 if (S_ISDIR(inode->i_mode))
2815 set_nlink(inode, 2);
2816 if (S_ISLNK(inode->i_mode))
2819 inode->i_op = p->iop;
2821 inode->i_fop = p->fop;
2823 d_add(dentry, inode);
2832 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2834 struct dentry *error;
2835 struct task_struct *task = get_proc_task(dir);
2836 const struct pid_entry *p, *last;
2838 error = ERR_PTR(-ENOENT);
2843 /* Lookup the directory entry */
2844 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2845 for (p = proc_base_stuff; p <= last; p++) {
2846 if (p->len != dentry->d_name.len)
2848 if (!memcmp(dentry->d_name.name, p->name, p->len))
2854 error = proc_base_instantiate(dir, dentry, task, p);
2857 put_task_struct(task);
2862 static int proc_base_fill_cache(struct file *filp, void *dirent,
2863 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2865 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2866 proc_base_instantiate, task, p);
2869 #ifdef CONFIG_TASK_IO_ACCOUNTING
2870 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2872 struct task_io_accounting acct = task->ioac;
2873 unsigned long flags;
2876 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2880 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2885 if (whole && lock_task_sighand(task, &flags)) {
2886 struct task_struct *t = task;
2888 task_io_accounting_add(&acct, &task->signal->ioac);
2889 while_each_thread(task, t)
2890 task_io_accounting_add(&acct, &t->ioac);
2892 unlock_task_sighand(task, &flags);
2894 result = sprintf(buffer,
2899 "read_bytes: %llu\n"
2900 "write_bytes: %llu\n"
2901 "cancelled_write_bytes: %llu\n",
2902 (unsigned long long)acct.rchar,
2903 (unsigned long long)acct.wchar,
2904 (unsigned long long)acct.syscr,
2905 (unsigned long long)acct.syscw,
2906 (unsigned long long)acct.read_bytes,
2907 (unsigned long long)acct.write_bytes,
2908 (unsigned long long)acct.cancelled_write_bytes);
2910 mutex_unlock(&task->signal->cred_guard_mutex);
2914 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2916 return do_io_accounting(task, buffer, 0);
2919 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2921 return do_io_accounting(task, buffer, 1);
2923 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2925 #ifdef CONFIG_USER_NS
2926 static int proc_id_map_open(struct inode *inode, struct file *file,
2927 struct seq_operations *seq_ops)
2929 struct user_namespace *ns = NULL;
2930 struct task_struct *task;
2931 struct seq_file *seq;
2934 task = get_proc_task(inode);
2937 ns = get_user_ns(task_cred_xxx(task, user_ns));
2939 put_task_struct(task);
2944 ret = seq_open(file, seq_ops);
2948 seq = file->private_data;
2958 static int proc_id_map_release(struct inode *inode, struct file *file)
2960 struct seq_file *seq = file->private_data;
2961 struct user_namespace *ns = seq->private;
2963 return seq_release(inode, file);
2966 static int proc_uid_map_open(struct inode *inode, struct file *file)
2968 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2971 static int proc_gid_map_open(struct inode *inode, struct file *file)
2973 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2976 static const struct file_operations proc_uid_map_operations = {
2977 .open = proc_uid_map_open,
2978 .write = proc_uid_map_write,
2980 .llseek = seq_lseek,
2981 .release = proc_id_map_release,
2984 static const struct file_operations proc_gid_map_operations = {
2985 .open = proc_gid_map_open,
2986 .write = proc_gid_map_write,
2988 .llseek = seq_lseek,
2989 .release = proc_id_map_release,
2991 #endif /* CONFIG_USER_NS */
2993 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2994 struct pid *pid, struct task_struct *task)
2996 int err = lock_trace(task);
2998 seq_printf(m, "%08x\n", task->personality);
3007 static const struct file_operations proc_task_operations;
3008 static const struct inode_operations proc_task_inode_operations;
3010 static const struct pid_entry tgid_base_stuff[] = {
3011 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
3012 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3013 #ifdef CONFIG_CHECKPOINT_RESTORE
3014 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
3016 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3017 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3019 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3021 REG("environ", S_IRUSR, proc_environ_operations),
3022 INF("auxv", S_IRUSR, proc_pid_auxv),
3023 ONE("status", S_IRUGO, proc_pid_status),
3024 ONE("personality", S_IRUGO, proc_pid_personality),
3025 INF("limits", S_IRUGO, proc_pid_limits),
3026 #ifdef CONFIG_SCHED_DEBUG
3027 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3029 #ifdef CONFIG_SCHED_AUTOGROUP
3030 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
3032 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3033 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3034 INF("syscall", S_IRUGO, proc_pid_syscall),
3036 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3037 ONE("stat", S_IRUGO, proc_tgid_stat),
3038 ONE("statm", S_IRUGO, proc_pid_statm),
3039 REG("maps", S_IRUGO, proc_pid_maps_operations),
3041 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
3043 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3044 LNK("cwd", proc_cwd_link),
3045 LNK("root", proc_root_link),
3046 LNK("exe", proc_exe_link),
3047 REG("mounts", S_IRUGO, proc_mounts_operations),
3048 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3049 REG("mountstats", S_IRUSR, proc_mountstats_operations),
3050 #ifdef CONFIG_PROC_PAGE_MONITOR
3051 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3052 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
3053 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3055 #ifdef CONFIG_SECURITY
3056 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3058 #ifdef CONFIG_KALLSYMS
3059 INF("wchan", S_IRUGO, proc_pid_wchan),
3061 #ifdef CONFIG_STACKTRACE
3062 ONE("stack", S_IRUGO, proc_pid_stack),
3064 #ifdef CONFIG_SCHEDSTATS
3065 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3067 #ifdef CONFIG_LATENCYTOP
3068 REG("latency", S_IRUGO, proc_lstats_operations),
3070 #ifdef CONFIG_PROC_PID_CPUSET
3071 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3073 #ifdef CONFIG_CGROUPS
3074 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3076 INF("oom_score", S_IRUGO, proc_oom_score),
3077 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3078 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3079 #ifdef CONFIG_AUDITSYSCALL
3080 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3081 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3083 #ifdef CONFIG_FAULT_INJECTION
3084 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3086 #ifdef CONFIG_ELF_CORE
3087 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3089 #ifdef CONFIG_TASK_IO_ACCOUNTING
3090 INF("io", S_IRUSR, proc_tgid_io_accounting),
3092 #ifdef CONFIG_HARDWALL
3093 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3095 #ifdef CONFIG_USER_NS
3096 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3097 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3101 static int proc_tgid_base_readdir(struct file * filp,
3102 void * dirent, filldir_t filldir)
3104 return proc_pident_readdir(filp,dirent,filldir,
3105 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
3108 static const struct file_operations proc_tgid_base_operations = {
3109 .read = generic_read_dir,
3110 .readdir = proc_tgid_base_readdir,
3111 .llseek = default_llseek,
3114 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3115 return proc_pident_lookup(dir, dentry,
3116 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3119 static const struct inode_operations proc_tgid_base_inode_operations = {
3120 .lookup = proc_tgid_base_lookup,
3121 .getattr = pid_getattr,
3122 .setattr = proc_setattr,
3123 .permission = proc_pid_permission,
3126 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
3128 struct dentry *dentry, *leader, *dir;
3129 char buf[PROC_NUMBUF];
3133 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3134 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
3136 shrink_dcache_parent(dentry);
3142 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
3143 leader = d_hash_and_lookup(mnt->mnt_root, &name);
3148 name.len = strlen(name.name);
3149 dir = d_hash_and_lookup(leader, &name);
3151 goto out_put_leader;
3154 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3155 dentry = d_hash_and_lookup(dir, &name);
3157 shrink_dcache_parent(dentry);
3170 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3171 * @task: task that should be flushed.
3173 * When flushing dentries from proc, one needs to flush them from global
3174 * proc (proc_mnt) and from all the namespaces' procs this task was seen
3175 * in. This call is supposed to do all of this job.
3177 * Looks in the dcache for
3179 * /proc/@tgid/task/@pid
3180 * if either directory is present flushes it and all of it'ts children
3183 * It is safe and reasonable to cache /proc entries for a task until
3184 * that task exits. After that they just clog up the dcache with
3185 * useless entries, possibly causing useful dcache entries to be
3186 * flushed instead. This routine is proved to flush those useless
3187 * dcache entries at process exit time.
3189 * NOTE: This routine is just an optimization so it does not guarantee
3190 * that no dcache entries will exist at process exit time it
3191 * just makes it very unlikely that any will persist.
3194 void proc_flush_task(struct task_struct *task)
3197 struct pid *pid, *tgid;
3200 pid = task_pid(task);
3201 tgid = task_tgid(task);
3203 for (i = 0; i <= pid->level; i++) {
3204 upid = &pid->numbers[i];
3205 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
3206 tgid->numbers[i].nr);
3209 upid = &pid->numbers[pid->level];
3211 pid_ns_release_proc(upid->ns);
3214 static struct dentry *proc_pid_instantiate(struct inode *dir,
3215 struct dentry * dentry,
3216 struct task_struct *task, const void *ptr)
3218 struct dentry *error = ERR_PTR(-ENOENT);
3219 struct inode *inode;
3221 inode = proc_pid_make_inode(dir->i_sb, task);
3225 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3226 inode->i_op = &proc_tgid_base_inode_operations;
3227 inode->i_fop = &proc_tgid_base_operations;
3228 inode->i_flags|=S_IMMUTABLE;
3230 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
3231 ARRAY_SIZE(tgid_base_stuff)));
3233 d_set_d_op(dentry, &pid_dentry_operations);
3235 d_add(dentry, inode);
3236 /* Close the race of the process dying before we return the dentry */
3237 if (pid_revalidate(dentry, NULL))
3243 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3245 struct dentry *result;
3246 struct task_struct *task;
3248 struct pid_namespace *ns;
3250 result = proc_base_lookup(dir, dentry);
3251 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3254 tgid = name_to_int(dentry);
3258 ns = dentry->d_sb->s_fs_info;
3260 task = find_task_by_pid_ns(tgid, ns);
3262 get_task_struct(task);
3267 result = proc_pid_instantiate(dir, dentry, task, NULL);
3268 put_task_struct(task);
3274 * Find the first task with tgid >= tgid
3279 struct task_struct *task;
3281 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3286 put_task_struct(iter.task);
3290 pid = find_ge_pid(iter.tgid, ns);
3292 iter.tgid = pid_nr_ns(pid, ns);
3293 iter.task = pid_task(pid, PIDTYPE_PID);
3294 /* What we to know is if the pid we have find is the
3295 * pid of a thread_group_leader. Testing for task
3296 * being a thread_group_leader is the obvious thing
3297 * todo but there is a window when it fails, due to
3298 * the pid transfer logic in de_thread.
3300 * So we perform the straight forward test of seeing
3301 * if the pid we have found is the pid of a thread
3302 * group leader, and don't worry if the task we have
3303 * found doesn't happen to be a thread group leader.
3304 * As we don't care in the case of readdir.
3306 if (!iter.task || !has_group_leader_pid(iter.task)) {
3310 get_task_struct(iter.task);
3316 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
3318 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3319 struct tgid_iter iter)
3321 char name[PROC_NUMBUF];
3322 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
3323 return proc_fill_cache(filp, dirent, filldir, name, len,
3324 proc_pid_instantiate, iter.task, NULL);
3327 static int fake_filldir(void *buf, const char *name, int namelen,
3328 loff_t offset, u64 ino, unsigned d_type)
3333 /* for the /proc/ directory itself, after non-process stuff has been done */
3334 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3337 struct task_struct *reaper;
3338 struct tgid_iter iter;
3339 struct pid_namespace *ns;
3340 filldir_t __filldir;
3342 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3344 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3346 reaper = get_proc_task(filp->f_path.dentry->d_inode);
3350 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
3351 const struct pid_entry *p = &proc_base_stuff[nr];
3352 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
3356 ns = filp->f_dentry->d_sb->s_fs_info;
3358 iter.tgid = filp->f_pos - TGID_OFFSET;
3359 for (iter = next_tgid(ns, iter);
3361 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3362 if (has_pid_permissions(ns, iter.task, 2))
3363 __filldir = filldir;
3365 __filldir = fake_filldir;
3367 filp->f_pos = iter.tgid + TGID_OFFSET;
3368 if (proc_pid_fill_cache(filp, dirent, __filldir, iter) < 0) {
3369 put_task_struct(iter.task);
3373 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3375 put_task_struct(reaper);
3383 static const struct pid_entry tid_base_stuff[] = {
3384 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3385 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3386 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3387 REG("environ", S_IRUSR, proc_environ_operations),
3388 INF("auxv", S_IRUSR, proc_pid_auxv),
3389 ONE("status", S_IRUGO, proc_pid_status),
3390 ONE("personality", S_IRUGO, proc_pid_personality),
3391 INF("limits", S_IRUGO, proc_pid_limits),
3392 #ifdef CONFIG_SCHED_DEBUG
3393 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3395 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3396 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3397 INF("syscall", S_IRUGO, proc_pid_syscall),
3399 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3400 ONE("stat", S_IRUGO, proc_tid_stat),
3401 ONE("statm", S_IRUGO, proc_pid_statm),
3402 REG("maps", S_IRUGO, proc_tid_maps_operations),
3403 #ifdef CONFIG_CHECKPOINT_RESTORE
3404 REG("children", S_IRUGO, proc_tid_children_operations),
3407 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
3409 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3410 LNK("cwd", proc_cwd_link),
3411 LNK("root", proc_root_link),
3412 LNK("exe", proc_exe_link),
3413 REG("mounts", S_IRUGO, proc_mounts_operations),
3414 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3415 #ifdef CONFIG_PROC_PAGE_MONITOR
3416 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3417 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
3418 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3420 #ifdef CONFIG_SECURITY
3421 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3423 #ifdef CONFIG_KALLSYMS
3424 INF("wchan", S_IRUGO, proc_pid_wchan),
3426 #ifdef CONFIG_STACKTRACE
3427 ONE("stack", S_IRUGO, proc_pid_stack),
3429 #ifdef CONFIG_SCHEDSTATS
3430 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3432 #ifdef CONFIG_LATENCYTOP
3433 REG("latency", S_IRUGO, proc_lstats_operations),
3435 #ifdef CONFIG_PROC_PID_CPUSET
3436 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3438 #ifdef CONFIG_CGROUPS
3439 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3441 INF("oom_score", S_IRUGO, proc_oom_score),
3442 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3443 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3444 #ifdef CONFIG_AUDITSYSCALL
3445 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3446 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3448 #ifdef CONFIG_FAULT_INJECTION
3449 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3451 #ifdef CONFIG_TASK_IO_ACCOUNTING
3452 INF("io", S_IRUSR, proc_tid_io_accounting),
3454 #ifdef CONFIG_HARDWALL
3455 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3457 #ifdef CONFIG_USER_NS
3458 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3459 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3463 static int proc_tid_base_readdir(struct file * filp,
3464 void * dirent, filldir_t filldir)
3466 return proc_pident_readdir(filp,dirent,filldir,
3467 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3470 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3471 return proc_pident_lookup(dir, dentry,
3472 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3475 static const struct file_operations proc_tid_base_operations = {
3476 .read = generic_read_dir,
3477 .readdir = proc_tid_base_readdir,
3478 .llseek = default_llseek,
3481 static const struct inode_operations proc_tid_base_inode_operations = {
3482 .lookup = proc_tid_base_lookup,
3483 .getattr = pid_getattr,
3484 .setattr = proc_setattr,
3487 static struct dentry *proc_task_instantiate(struct inode *dir,
3488 struct dentry *dentry, struct task_struct *task, const void *ptr)
3490 struct dentry *error = ERR_PTR(-ENOENT);
3491 struct inode *inode;
3492 inode = proc_pid_make_inode(dir->i_sb, task);
3496 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3497 inode->i_op = &proc_tid_base_inode_operations;
3498 inode->i_fop = &proc_tid_base_operations;
3499 inode->i_flags|=S_IMMUTABLE;
3501 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3502 ARRAY_SIZE(tid_base_stuff)));
3504 d_set_d_op(dentry, &pid_dentry_operations);
3506 d_add(dentry, inode);
3507 /* Close the race of the process dying before we return the dentry */
3508 if (pid_revalidate(dentry, NULL))
3514 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3516 struct dentry *result = ERR_PTR(-ENOENT);
3517 struct task_struct *task;
3518 struct task_struct *leader = get_proc_task(dir);
3520 struct pid_namespace *ns;
3525 tid = name_to_int(dentry);
3529 ns = dentry->d_sb->s_fs_info;
3531 task = find_task_by_pid_ns(tid, ns);
3533 get_task_struct(task);
3537 if (!same_thread_group(leader, task))
3540 result = proc_task_instantiate(dir, dentry, task, NULL);
3542 put_task_struct(task);
3544 put_task_struct(leader);
3550 * Find the first tid of a thread group to return to user space.
3552 * Usually this is just the thread group leader, but if the users
3553 * buffer was too small or there was a seek into the middle of the
3554 * directory we have more work todo.
3556 * In the case of a short read we start with find_task_by_pid.
3558 * In the case of a seek we start with the leader and walk nr
3561 static struct task_struct *first_tid(struct task_struct *leader,
3562 int tid, int nr, struct pid_namespace *ns)
3564 struct task_struct *pos;
3567 /* Attempt to start with the pid of a thread */
3568 if (tid && (nr > 0)) {
3569 pos = find_task_by_pid_ns(tid, ns);
3570 if (pos && (pos->group_leader == leader))
3574 /* If nr exceeds the number of threads there is nothing todo */
3576 if (nr && nr >= get_nr_threads(leader))
3579 /* If we haven't found our starting place yet start
3580 * with the leader and walk nr threads forward.
3582 for (pos = leader; nr > 0; --nr) {
3583 pos = next_thread(pos);
3584 if (pos == leader) {
3590 get_task_struct(pos);
3597 * Find the next thread in the thread list.
3598 * Return NULL if there is an error or no next thread.
3600 * The reference to the input task_struct is released.
3602 static struct task_struct *next_tid(struct task_struct *start)
3604 struct task_struct *pos = NULL;
3606 if (pid_alive(start)) {
3607 pos = next_thread(start);
3608 if (thread_group_leader(pos))
3611 get_task_struct(pos);
3614 put_task_struct(start);
3618 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3619 struct task_struct *task, int tid)
3621 char name[PROC_NUMBUF];
3622 int len = snprintf(name, sizeof(name), "%d", tid);
3623 return proc_fill_cache(filp, dirent, filldir, name, len,
3624 proc_task_instantiate, task, NULL);
3627 /* for the /proc/TGID/task/ directories */
3628 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3630 struct dentry *dentry = filp->f_path.dentry;
3631 struct inode *inode = dentry->d_inode;
3632 struct task_struct *leader = NULL;
3633 struct task_struct *task;
3634 int retval = -ENOENT;
3637 struct pid_namespace *ns;
3639 task = get_proc_task(inode);
3643 if (pid_alive(task)) {
3644 leader = task->group_leader;
3645 get_task_struct(leader);
3648 put_task_struct(task);
3653 switch ((unsigned long)filp->f_pos) {
3656 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3661 ino = parent_ino(dentry);
3662 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3668 /* f_version caches the tgid value that the last readdir call couldn't
3669 * return. lseek aka telldir automagically resets f_version to 0.
3671 ns = filp->f_dentry->d_sb->s_fs_info;
3672 tid = (int)filp->f_version;
3673 filp->f_version = 0;
3674 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3676 task = next_tid(task), filp->f_pos++) {
3677 tid = task_pid_nr_ns(task, ns);
3678 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3679 /* returning this tgid failed, save it as the first
3680 * pid for the next readir call */
3681 filp->f_version = (u64)tid;
3682 put_task_struct(task);
3687 put_task_struct(leader);
3692 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3694 struct inode *inode = dentry->d_inode;
3695 struct task_struct *p = get_proc_task(inode);
3696 generic_fillattr(inode, stat);
3699 stat->nlink += get_nr_threads(p);
3706 static const struct inode_operations proc_task_inode_operations = {
3707 .lookup = proc_task_lookup,
3708 .getattr = proc_task_getattr,
3709 .setattr = proc_setattr,
3710 .permission = proc_pid_permission,
3713 static const struct file_operations proc_task_operations = {
3714 .read = generic_read_dir,
3715 .readdir = proc_task_readdir,
3716 .llseek = default_llseek,