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 file->private_data = mm;
703 static int mem_open(struct inode *inode, struct file *file)
705 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
707 /* OK to pass negative loff_t, we can catch out-of-range */
708 file->f_mode |= FMODE_UNSIGNED_OFFSET;
713 static ssize_t mem_rw(struct file *file, char __user *buf,
714 size_t count, loff_t *ppos, int write)
716 struct mm_struct *mm = file->private_data;
717 unsigned long addr = *ppos;
724 page = (char *)__get_free_page(GFP_TEMPORARY);
729 if (!atomic_inc_not_zero(&mm->mm_users))
733 int this_len = min_t(int, count, PAGE_SIZE);
735 if (write && copy_from_user(page, buf, this_len)) {
740 this_len = access_remote_vm(mm, addr, page, this_len, write);
747 if (!write && copy_to_user(buf, page, this_len)) {
761 free_page((unsigned long) page);
765 static ssize_t mem_read(struct file *file, char __user *buf,
766 size_t count, loff_t *ppos)
768 return mem_rw(file, buf, count, ppos, 0);
771 static ssize_t mem_write(struct file *file, const char __user *buf,
772 size_t count, loff_t *ppos)
774 return mem_rw(file, (char __user*)buf, count, ppos, 1);
777 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
781 file->f_pos = offset;
784 file->f_pos += offset;
789 force_successful_syscall_return();
793 static int mem_release(struct inode *inode, struct file *file)
795 struct mm_struct *mm = file->private_data;
801 static const struct file_operations proc_mem_operations = {
806 .release = mem_release,
809 static int environ_open(struct inode *inode, struct file *file)
811 return __mem_open(inode, file, PTRACE_MODE_READ);
814 static ssize_t environ_read(struct file *file, char __user *buf,
815 size_t count, loff_t *ppos)
818 unsigned long src = *ppos;
820 struct mm_struct *mm = file->private_data;
825 page = (char *)__get_free_page(GFP_TEMPORARY);
830 if (!atomic_inc_not_zero(&mm->mm_users))
833 size_t this_len, max_len;
836 if (src >= (mm->env_end - mm->env_start))
839 this_len = mm->env_end - (mm->env_start + src);
841 max_len = min_t(size_t, PAGE_SIZE, count);
842 this_len = min(max_len, this_len);
844 retval = access_remote_vm(mm, (mm->env_start + src),
852 if (copy_to_user(buf, page, retval)) {
866 free_page((unsigned long) page);
870 static const struct file_operations proc_environ_operations = {
871 .open = environ_open,
872 .read = environ_read,
873 .llseek = generic_file_llseek,
874 .release = mem_release,
877 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
878 size_t count, loff_t *ppos)
880 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
881 char buffer[PROC_NUMBUF];
883 int oom_adjust = OOM_DISABLE;
889 if (lock_task_sighand(task, &flags)) {
890 oom_adjust = task->signal->oom_adj;
891 unlock_task_sighand(task, &flags);
894 put_task_struct(task);
896 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
898 return simple_read_from_buffer(buf, count, ppos, buffer, len);
901 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
902 size_t count, loff_t *ppos)
904 struct task_struct *task;
905 char buffer[PROC_NUMBUF];
910 memset(buffer, 0, sizeof(buffer));
911 if (count > sizeof(buffer) - 1)
912 count = sizeof(buffer) - 1;
913 if (copy_from_user(buffer, buf, count)) {
918 err = kstrtoint(strstrip(buffer), 0, &oom_adjust);
921 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
922 oom_adjust != OOM_DISABLE) {
927 task = get_proc_task(file->f_path.dentry->d_inode);
939 if (!lock_task_sighand(task, &flags)) {
944 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
950 * Warn that /proc/pid/oom_adj is deprecated, see
951 * Documentation/feature-removal-schedule.txt.
953 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
954 current->comm, task_pid_nr(current), task_pid_nr(task),
956 task->signal->oom_adj = oom_adjust;
958 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
959 * value is always attainable.
961 if (task->signal->oom_adj == OOM_ADJUST_MAX)
962 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
964 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
966 trace_oom_score_adj_update(task);
968 unlock_task_sighand(task, &flags);
971 put_task_struct(task);
973 return err < 0 ? err : count;
976 static const struct file_operations proc_oom_adjust_operations = {
977 .read = oom_adjust_read,
978 .write = oom_adjust_write,
979 .llseek = generic_file_llseek,
982 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
983 size_t count, loff_t *ppos)
985 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
986 char buffer[PROC_NUMBUF];
987 int oom_score_adj = OOM_SCORE_ADJ_MIN;
993 if (lock_task_sighand(task, &flags)) {
994 oom_score_adj = task->signal->oom_score_adj;
995 unlock_task_sighand(task, &flags);
997 put_task_struct(task);
998 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
999 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1002 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1003 size_t count, loff_t *ppos)
1005 struct task_struct *task;
1006 char buffer[PROC_NUMBUF];
1007 unsigned long flags;
1011 memset(buffer, 0, sizeof(buffer));
1012 if (count > sizeof(buffer) - 1)
1013 count = sizeof(buffer) - 1;
1014 if (copy_from_user(buffer, buf, count)) {
1019 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1022 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1023 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1028 task = get_proc_task(file->f_path.dentry->d_inode);
1040 if (!lock_task_sighand(task, &flags)) {
1045 if (oom_score_adj < task->signal->oom_score_adj_min &&
1046 !capable(CAP_SYS_RESOURCE)) {
1051 task->signal->oom_score_adj = oom_score_adj;
1052 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1053 task->signal->oom_score_adj_min = oom_score_adj;
1054 trace_oom_score_adj_update(task);
1056 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1057 * always attainable.
1059 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1060 task->signal->oom_adj = OOM_DISABLE;
1062 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1065 unlock_task_sighand(task, &flags);
1068 put_task_struct(task);
1070 return err < 0 ? err : count;
1073 static const struct file_operations proc_oom_score_adj_operations = {
1074 .read = oom_score_adj_read,
1075 .write = oom_score_adj_write,
1076 .llseek = default_llseek,
1079 #ifdef CONFIG_AUDITSYSCALL
1080 #define TMPBUFLEN 21
1081 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1082 size_t count, loff_t *ppos)
1084 struct inode * inode = file->f_path.dentry->d_inode;
1085 struct task_struct *task = get_proc_task(inode);
1087 char tmpbuf[TMPBUFLEN];
1091 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1092 from_kuid(file->f_cred->user_ns,
1093 audit_get_loginuid(task)));
1094 put_task_struct(task);
1095 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1098 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1099 size_t count, loff_t *ppos)
1101 struct inode * inode = file->f_path.dentry->d_inode;
1108 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1114 if (count >= PAGE_SIZE)
1115 count = PAGE_SIZE - 1;
1118 /* No partial writes. */
1121 page = (char*)__get_free_page(GFP_TEMPORARY);
1125 if (copy_from_user(page, buf, count))
1129 loginuid = simple_strtoul(page, &tmp, 10);
1135 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1136 if (!uid_valid(kloginuid)) {
1141 length = audit_set_loginuid(kloginuid);
1142 if (likely(length == 0))
1146 free_page((unsigned long) page);
1150 static const struct file_operations proc_loginuid_operations = {
1151 .read = proc_loginuid_read,
1152 .write = proc_loginuid_write,
1153 .llseek = generic_file_llseek,
1156 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1157 size_t count, loff_t *ppos)
1159 struct inode * inode = file->f_path.dentry->d_inode;
1160 struct task_struct *task = get_proc_task(inode);
1162 char tmpbuf[TMPBUFLEN];
1166 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1167 audit_get_sessionid(task));
1168 put_task_struct(task);
1169 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1172 static const struct file_operations proc_sessionid_operations = {
1173 .read = proc_sessionid_read,
1174 .llseek = generic_file_llseek,
1178 #ifdef CONFIG_FAULT_INJECTION
1179 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1180 size_t count, loff_t *ppos)
1182 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1183 char buffer[PROC_NUMBUF];
1189 make_it_fail = task->make_it_fail;
1190 put_task_struct(task);
1192 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1194 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1197 static ssize_t proc_fault_inject_write(struct file * file,
1198 const char __user * buf, size_t count, loff_t *ppos)
1200 struct task_struct *task;
1201 char buffer[PROC_NUMBUF], *end;
1204 if (!capable(CAP_SYS_RESOURCE))
1206 memset(buffer, 0, sizeof(buffer));
1207 if (count > sizeof(buffer) - 1)
1208 count = sizeof(buffer) - 1;
1209 if (copy_from_user(buffer, buf, count))
1211 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1214 task = get_proc_task(file->f_dentry->d_inode);
1217 task->make_it_fail = make_it_fail;
1218 put_task_struct(task);
1223 static const struct file_operations proc_fault_inject_operations = {
1224 .read = proc_fault_inject_read,
1225 .write = proc_fault_inject_write,
1226 .llseek = generic_file_llseek,
1231 #ifdef CONFIG_SCHED_DEBUG
1233 * Print out various scheduling related per-task fields:
1235 static int sched_show(struct seq_file *m, void *v)
1237 struct inode *inode = m->private;
1238 struct task_struct *p;
1240 p = get_proc_task(inode);
1243 proc_sched_show_task(p, m);
1251 sched_write(struct file *file, const char __user *buf,
1252 size_t count, loff_t *offset)
1254 struct inode *inode = file->f_path.dentry->d_inode;
1255 struct task_struct *p;
1257 p = get_proc_task(inode);
1260 proc_sched_set_task(p);
1267 static int sched_open(struct inode *inode, struct file *filp)
1269 return single_open(filp, sched_show, inode);
1272 static const struct file_operations proc_pid_sched_operations = {
1275 .write = sched_write,
1276 .llseek = seq_lseek,
1277 .release = single_release,
1282 #ifdef CONFIG_SCHED_AUTOGROUP
1284 * Print out autogroup related information:
1286 static int sched_autogroup_show(struct seq_file *m, void *v)
1288 struct inode *inode = m->private;
1289 struct task_struct *p;
1291 p = get_proc_task(inode);
1294 proc_sched_autogroup_show_task(p, m);
1302 sched_autogroup_write(struct file *file, const char __user *buf,
1303 size_t count, loff_t *offset)
1305 struct inode *inode = file->f_path.dentry->d_inode;
1306 struct task_struct *p;
1307 char buffer[PROC_NUMBUF];
1311 memset(buffer, 0, sizeof(buffer));
1312 if (count > sizeof(buffer) - 1)
1313 count = sizeof(buffer) - 1;
1314 if (copy_from_user(buffer, buf, count))
1317 err = kstrtoint(strstrip(buffer), 0, &nice);
1321 p = get_proc_task(inode);
1325 err = proc_sched_autogroup_set_nice(p, nice);
1334 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1338 ret = single_open(filp, sched_autogroup_show, NULL);
1340 struct seq_file *m = filp->private_data;
1347 static const struct file_operations proc_pid_sched_autogroup_operations = {
1348 .open = sched_autogroup_open,
1350 .write = sched_autogroup_write,
1351 .llseek = seq_lseek,
1352 .release = single_release,
1355 #endif /* CONFIG_SCHED_AUTOGROUP */
1357 static ssize_t comm_write(struct file *file, const char __user *buf,
1358 size_t count, loff_t *offset)
1360 struct inode *inode = file->f_path.dentry->d_inode;
1361 struct task_struct *p;
1362 char buffer[TASK_COMM_LEN];
1364 memset(buffer, 0, sizeof(buffer));
1365 if (count > sizeof(buffer) - 1)
1366 count = sizeof(buffer) - 1;
1367 if (copy_from_user(buffer, buf, count))
1370 p = get_proc_task(inode);
1374 if (same_thread_group(current, p))
1375 set_task_comm(p, buffer);
1384 static int comm_show(struct seq_file *m, void *v)
1386 struct inode *inode = m->private;
1387 struct task_struct *p;
1389 p = get_proc_task(inode);
1394 seq_printf(m, "%s\n", p->comm);
1402 static int comm_open(struct inode *inode, struct file *filp)
1404 return single_open(filp, comm_show, inode);
1407 static const struct file_operations proc_pid_set_comm_operations = {
1410 .write = comm_write,
1411 .llseek = seq_lseek,
1412 .release = single_release,
1415 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1417 struct task_struct *task;
1418 struct mm_struct *mm;
1419 struct file *exe_file;
1421 task = get_proc_task(dentry->d_inode);
1424 mm = get_task_mm(task);
1425 put_task_struct(task);
1428 exe_file = get_mm_exe_file(mm);
1431 *exe_path = exe_file->f_path;
1432 path_get(&exe_file->f_path);
1439 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1441 struct inode *inode = dentry->d_inode;
1443 int error = -EACCES;
1445 /* Are we allowed to snoop on the tasks file descriptors? */
1446 if (!proc_fd_access_allowed(inode))
1449 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1453 nd_jump_link(nd, &path);
1456 return ERR_PTR(error);
1459 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1461 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1468 pathname = d_path(path, tmp, PAGE_SIZE);
1469 len = PTR_ERR(pathname);
1470 if (IS_ERR(pathname))
1472 len = tmp + PAGE_SIZE - 1 - pathname;
1476 if (copy_to_user(buffer, pathname, len))
1479 free_page((unsigned long)tmp);
1483 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1485 int error = -EACCES;
1486 struct inode *inode = dentry->d_inode;
1489 /* Are we allowed to snoop on the tasks file descriptors? */
1490 if (!proc_fd_access_allowed(inode))
1493 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1497 error = do_proc_readlink(&path, buffer, buflen);
1503 static const struct inode_operations proc_pid_link_inode_operations = {
1504 .readlink = proc_pid_readlink,
1505 .follow_link = proc_pid_follow_link,
1506 .setattr = proc_setattr,
1510 /* building an inode */
1512 static int task_dumpable(struct task_struct *task)
1515 struct mm_struct *mm;
1520 dumpable = get_dumpable(mm);
1527 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1529 struct inode * inode;
1530 struct proc_inode *ei;
1531 const struct cred *cred;
1533 /* We need a new inode */
1535 inode = new_inode(sb);
1541 inode->i_ino = get_next_ino();
1542 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1543 inode->i_op = &proc_def_inode_operations;
1546 * grab the reference to task.
1548 ei->pid = get_task_pid(task, PIDTYPE_PID);
1552 if (task_dumpable(task)) {
1554 cred = __task_cred(task);
1555 inode->i_uid = cred->euid;
1556 inode->i_gid = cred->egid;
1559 security_task_to_inode(task, inode);
1569 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1571 struct inode *inode = dentry->d_inode;
1572 struct task_struct *task;
1573 const struct cred *cred;
1574 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
1576 generic_fillattr(inode, stat);
1579 stat->uid = GLOBAL_ROOT_UID;
1580 stat->gid = GLOBAL_ROOT_GID;
1581 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1583 if (!has_pid_permissions(pid, task, 2)) {
1586 * This doesn't prevent learning whether PID exists,
1587 * it only makes getattr() consistent with readdir().
1591 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1592 task_dumpable(task)) {
1593 cred = __task_cred(task);
1594 stat->uid = cred->euid;
1595 stat->gid = cred->egid;
1605 * Exceptional case: normally we are not allowed to unhash a busy
1606 * directory. In this case, however, we can do it - no aliasing problems
1607 * due to the way we treat inodes.
1609 * Rewrite the inode's ownerships here because the owning task may have
1610 * performed a setuid(), etc.
1612 * Before the /proc/pid/status file was created the only way to read
1613 * the effective uid of a /process was to stat /proc/pid. Reading
1614 * /proc/pid/status is slow enough that procps and other packages
1615 * kept stating /proc/pid. To keep the rules in /proc simple I have
1616 * made this apply to all per process world readable and executable
1619 int pid_revalidate(struct dentry *dentry, unsigned int flags)
1621 struct inode *inode;
1622 struct task_struct *task;
1623 const struct cred *cred;
1625 if (flags & LOOKUP_RCU)
1628 inode = dentry->d_inode;
1629 task = get_proc_task(inode);
1632 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1633 task_dumpable(task)) {
1635 cred = __task_cred(task);
1636 inode->i_uid = cred->euid;
1637 inode->i_gid = cred->egid;
1640 inode->i_uid = GLOBAL_ROOT_UID;
1641 inode->i_gid = GLOBAL_ROOT_GID;
1643 inode->i_mode &= ~(S_ISUID | S_ISGID);
1644 security_task_to_inode(task, inode);
1645 put_task_struct(task);
1652 static int pid_delete_dentry(const struct dentry * dentry)
1654 /* Is the task we represent dead?
1655 * If so, then don't put the dentry on the lru list,
1656 * kill it immediately.
1658 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1661 const struct dentry_operations pid_dentry_operations =
1663 .d_revalidate = pid_revalidate,
1664 .d_delete = pid_delete_dentry,
1670 * Fill a directory entry.
1672 * If possible create the dcache entry and derive our inode number and
1673 * file type from dcache entry.
1675 * Since all of the proc inode numbers are dynamically generated, the inode
1676 * numbers do not exist until the inode is cache. This means creating the
1677 * the dcache entry in readdir is necessary to keep the inode numbers
1678 * reported by readdir in sync with the inode numbers reported
1681 int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1682 const char *name, int len,
1683 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1685 struct dentry *child, *dir = filp->f_path.dentry;
1686 struct inode *inode;
1689 unsigned type = DT_UNKNOWN;
1693 qname.hash = full_name_hash(name, len);
1695 child = d_lookup(dir, &qname);
1698 new = d_alloc(dir, &qname);
1700 child = instantiate(dir->d_inode, new, task, ptr);
1707 if (!child || IS_ERR(child) || !child->d_inode)
1708 goto end_instantiate;
1709 inode = child->d_inode;
1712 type = inode->i_mode >> 12;
1717 ino = find_inode_number(dir, &qname);
1720 return filldir(dirent, name, len, filp->f_pos, ino, type);
1723 static unsigned name_to_int(struct dentry *dentry)
1725 const char *name = dentry->d_name.name;
1726 int len = dentry->d_name.len;
1729 if (len > 1 && *name == '0')
1732 unsigned c = *name++ - '0';
1735 if (n >= (~0U-9)/10)
1745 #define PROC_FDINFO_MAX 64
1747 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1749 struct task_struct *task = get_proc_task(inode);
1750 struct files_struct *files = NULL;
1752 int fd = proc_fd(inode);
1755 files = get_files_struct(task);
1756 put_task_struct(task);
1760 * We are not taking a ref to the file structure, so we must
1763 spin_lock(&files->file_lock);
1764 file = fcheck_files(files, fd);
1766 unsigned int f_flags;
1767 struct fdtable *fdt;
1769 fdt = files_fdtable(files);
1770 f_flags = file->f_flags & ~O_CLOEXEC;
1771 if (close_on_exec(fd, fdt))
1772 f_flags |= O_CLOEXEC;
1775 *path = file->f_path;
1776 path_get(&file->f_path);
1779 snprintf(info, PROC_FDINFO_MAX,
1782 (long long) file->f_pos,
1784 spin_unlock(&files->file_lock);
1785 put_files_struct(files);
1788 spin_unlock(&files->file_lock);
1789 put_files_struct(files);
1794 static int proc_fd_link(struct dentry *dentry, struct path *path)
1796 return proc_fd_info(dentry->d_inode, path, NULL);
1799 static int tid_fd_revalidate(struct dentry *dentry, unsigned int flags)
1801 struct inode *inode;
1802 struct task_struct *task;
1804 struct files_struct *files;
1805 const struct cred *cred;
1807 if (flags & LOOKUP_RCU)
1810 inode = dentry->d_inode;
1811 task = get_proc_task(inode);
1812 fd = proc_fd(inode);
1815 files = get_files_struct(task);
1819 file = fcheck_files(files, fd);
1821 unsigned f_mode = file->f_mode;
1824 put_files_struct(files);
1826 if (task_dumpable(task)) {
1828 cred = __task_cred(task);
1829 inode->i_uid = cred->euid;
1830 inode->i_gid = cred->egid;
1833 inode->i_uid = GLOBAL_ROOT_UID;
1834 inode->i_gid = GLOBAL_ROOT_GID;
1837 if (S_ISLNK(inode->i_mode)) {
1838 unsigned i_mode = S_IFLNK;
1839 if (f_mode & FMODE_READ)
1840 i_mode |= S_IRUSR | S_IXUSR;
1841 if (f_mode & FMODE_WRITE)
1842 i_mode |= S_IWUSR | S_IXUSR;
1843 inode->i_mode = i_mode;
1846 security_task_to_inode(task, inode);
1847 put_task_struct(task);
1851 put_files_struct(files);
1853 put_task_struct(task);
1859 static const struct dentry_operations tid_fd_dentry_operations =
1861 .d_revalidate = tid_fd_revalidate,
1862 .d_delete = pid_delete_dentry,
1865 static struct dentry *proc_fd_instantiate(struct inode *dir,
1866 struct dentry *dentry, struct task_struct *task, const void *ptr)
1868 unsigned fd = (unsigned long)ptr;
1869 struct inode *inode;
1870 struct proc_inode *ei;
1871 struct dentry *error = ERR_PTR(-ENOENT);
1873 inode = proc_pid_make_inode(dir->i_sb, task);
1879 inode->i_mode = S_IFLNK;
1880 inode->i_op = &proc_pid_link_inode_operations;
1882 ei->op.proc_get_link = proc_fd_link;
1883 d_set_d_op(dentry, &tid_fd_dentry_operations);
1884 d_add(dentry, inode);
1885 /* Close the race of the process dying before we return the dentry */
1886 if (tid_fd_revalidate(dentry, 0))
1893 static struct dentry *proc_lookupfd_common(struct inode *dir,
1894 struct dentry *dentry,
1895 instantiate_t instantiate)
1897 struct task_struct *task = get_proc_task(dir);
1898 unsigned fd = name_to_int(dentry);
1899 struct dentry *result = ERR_PTR(-ENOENT);
1906 result = instantiate(dir, dentry, task, (void *)(unsigned long)fd);
1908 put_task_struct(task);
1913 static int proc_readfd_common(struct file * filp, void * dirent,
1914 filldir_t filldir, instantiate_t instantiate)
1916 struct dentry *dentry = filp->f_path.dentry;
1917 struct inode *inode = dentry->d_inode;
1918 struct task_struct *p = get_proc_task(inode);
1919 unsigned int fd, ino;
1921 struct files_struct * files;
1931 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1935 ino = parent_ino(dentry);
1936 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1940 files = get_files_struct(p);
1944 for (fd = filp->f_pos-2;
1945 fd < files_fdtable(files)->max_fds;
1946 fd++, filp->f_pos++) {
1947 char name[PROC_NUMBUF];
1951 if (!fcheck_files(files, fd))
1955 len = snprintf(name, sizeof(name), "%d", fd);
1956 rv = proc_fill_cache(filp, dirent, filldir,
1957 name, len, instantiate, p,
1958 (void *)(unsigned long)fd);
1965 put_files_struct(files);
1973 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1976 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1979 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1981 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1984 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1985 size_t len, loff_t *ppos)
1987 char tmp[PROC_FDINFO_MAX];
1988 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
1990 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1994 static const struct file_operations proc_fdinfo_file_operations = {
1995 .open = nonseekable_open,
1996 .read = proc_fdinfo_read,
1997 .llseek = no_llseek,
2000 static const struct file_operations proc_fd_operations = {
2001 .read = generic_read_dir,
2002 .readdir = proc_readfd,
2003 .llseek = default_llseek,
2006 #ifdef CONFIG_CHECKPOINT_RESTORE
2009 * dname_to_vma_addr - maps a dentry name into two unsigned longs
2010 * which represent vma start and end addresses.
2012 static int dname_to_vma_addr(struct dentry *dentry,
2013 unsigned long *start, unsigned long *end)
2015 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
2021 static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
2023 unsigned long vm_start, vm_end;
2024 bool exact_vma_exists = false;
2025 struct mm_struct *mm = NULL;
2026 struct task_struct *task;
2027 const struct cred *cred;
2028 struct inode *inode;
2031 if (flags & LOOKUP_RCU)
2034 if (!capable(CAP_SYS_ADMIN)) {
2039 inode = dentry->d_inode;
2040 task = get_proc_task(inode);
2044 mm = mm_access(task, PTRACE_MODE_READ);
2045 if (IS_ERR_OR_NULL(mm))
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, unsigned int flags)
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 (!ptrace_may_access(task, PTRACE_MODE_READ))
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);
2203 put_task_struct(task);
2208 static const struct inode_operations proc_map_files_inode_operations = {
2209 .lookup = proc_map_files_lookup,
2210 .permission = proc_fd_permission,
2211 .setattr = proc_setattr,
2215 proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
2217 struct dentry *dentry = filp->f_path.dentry;
2218 struct inode *inode = dentry->d_inode;
2219 struct vm_area_struct *vma;
2220 struct task_struct *task;
2221 struct mm_struct *mm;
2226 if (!capable(CAP_SYS_ADMIN))
2230 task = get_proc_task(inode);
2235 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2239 switch (filp->f_pos) {
2242 if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0)
2246 ino = parent_ino(dentry);
2247 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2252 unsigned long nr_files, pos, i;
2253 struct flex_array *fa = NULL;
2254 struct map_files_info info;
2255 struct map_files_info *p;
2257 mm = get_task_mm(task);
2260 down_read(&mm->mmap_sem);
2265 * We need two passes here:
2267 * 1) Collect vmas of mapped files with mmap_sem taken
2268 * 2) Release mmap_sem and instantiate entries
2270 * otherwise we get lockdep complained, since filldir()
2271 * routine might require mmap_sem taken in might_fault().
2274 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2275 if (vma->vm_file && ++pos > filp->f_pos)
2280 fa = flex_array_alloc(sizeof(info), nr_files,
2282 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2286 flex_array_free(fa);
2287 up_read(&mm->mmap_sem);
2291 for (i = 0, vma = mm->mmap, pos = 2; vma;
2292 vma = vma->vm_next) {
2295 if (++pos <= filp->f_pos)
2298 get_file(vma->vm_file);
2299 info.file = vma->vm_file;
2300 info.len = snprintf(info.name,
2301 sizeof(info.name), "%lx-%lx",
2302 vma->vm_start, vma->vm_end);
2303 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2307 up_read(&mm->mmap_sem);
2309 for (i = 0; i < nr_files; i++) {
2310 p = flex_array_get(fa, i);
2311 ret = proc_fill_cache(filp, dirent, filldir,
2313 proc_map_files_instantiate,
2320 for (; i < nr_files; i++) {
2322 * In case of error don't forget
2323 * to put rest of file refs.
2325 p = flex_array_get(fa, i);
2329 flex_array_free(fa);
2335 put_task_struct(task);
2340 static const struct file_operations proc_map_files_operations = {
2341 .read = generic_read_dir,
2342 .readdir = proc_map_files_readdir,
2343 .llseek = default_llseek,
2346 #endif /* CONFIG_CHECKPOINT_RESTORE */
2349 * /proc/pid/fd needs a special permission handler so that a process can still
2350 * access /proc/self/fd after it has executed a setuid().
2352 static int proc_fd_permission(struct inode *inode, int mask)
2354 int rv = generic_permission(inode, mask);
2357 if (task_pid(current) == proc_pid(inode))
2363 * proc directories can do almost nothing..
2365 static const struct inode_operations proc_fd_inode_operations = {
2366 .lookup = proc_lookupfd,
2367 .permission = proc_fd_permission,
2368 .setattr = proc_setattr,
2371 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2372 struct dentry *dentry, struct task_struct *task, const void *ptr)
2374 unsigned fd = (unsigned long)ptr;
2375 struct inode *inode;
2376 struct proc_inode *ei;
2377 struct dentry *error = ERR_PTR(-ENOENT);
2379 inode = proc_pid_make_inode(dir->i_sb, task);
2384 inode->i_mode = S_IFREG | S_IRUSR;
2385 inode->i_fop = &proc_fdinfo_file_operations;
2386 d_set_d_op(dentry, &tid_fd_dentry_operations);
2387 d_add(dentry, inode);
2388 /* Close the race of the process dying before we return the dentry */
2389 if (tid_fd_revalidate(dentry, 0))
2396 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2397 struct dentry *dentry,
2400 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2403 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2405 return proc_readfd_common(filp, dirent, filldir,
2406 proc_fdinfo_instantiate);
2409 static const struct file_operations proc_fdinfo_operations = {
2410 .read = generic_read_dir,
2411 .readdir = proc_readfdinfo,
2412 .llseek = default_llseek,
2416 * proc directories can do almost nothing..
2418 static const struct inode_operations proc_fdinfo_inode_operations = {
2419 .lookup = proc_lookupfdinfo,
2420 .setattr = proc_setattr,
2424 static struct dentry *proc_pident_instantiate(struct inode *dir,
2425 struct dentry *dentry, struct task_struct *task, const void *ptr)
2427 const struct pid_entry *p = ptr;
2428 struct inode *inode;
2429 struct proc_inode *ei;
2430 struct dentry *error = ERR_PTR(-ENOENT);
2432 inode = proc_pid_make_inode(dir->i_sb, task);
2437 inode->i_mode = p->mode;
2438 if (S_ISDIR(inode->i_mode))
2439 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2441 inode->i_op = p->iop;
2443 inode->i_fop = p->fop;
2445 d_set_d_op(dentry, &pid_dentry_operations);
2446 d_add(dentry, inode);
2447 /* Close the race of the process dying before we return the dentry */
2448 if (pid_revalidate(dentry, 0))
2454 static struct dentry *proc_pident_lookup(struct inode *dir,
2455 struct dentry *dentry,
2456 const struct pid_entry *ents,
2459 struct dentry *error;
2460 struct task_struct *task = get_proc_task(dir);
2461 const struct pid_entry *p, *last;
2463 error = ERR_PTR(-ENOENT);
2469 * Yes, it does not scale. And it should not. Don't add
2470 * new entries into /proc/<tgid>/ without very good reasons.
2472 last = &ents[nents - 1];
2473 for (p = ents; p <= last; p++) {
2474 if (p->len != dentry->d_name.len)
2476 if (!memcmp(dentry->d_name.name, p->name, p->len))
2482 error = proc_pident_instantiate(dir, dentry, task, p);
2484 put_task_struct(task);
2489 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2490 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2492 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2493 proc_pident_instantiate, task, p);
2496 static int proc_pident_readdir(struct file *filp,
2497 void *dirent, filldir_t filldir,
2498 const struct pid_entry *ents, unsigned int nents)
2501 struct dentry *dentry = filp->f_path.dentry;
2502 struct inode *inode = dentry->d_inode;
2503 struct task_struct *task = get_proc_task(inode);
2504 const struct pid_entry *p, *last;
2517 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2523 ino = parent_ino(dentry);
2524 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2536 last = &ents[nents - 1];
2538 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2547 put_task_struct(task);
2552 #ifdef CONFIG_SECURITY
2553 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2554 size_t count, loff_t *ppos)
2556 struct inode * inode = file->f_path.dentry->d_inode;
2559 struct task_struct *task = get_proc_task(inode);
2564 length = security_getprocattr(task,
2565 (char*)file->f_path.dentry->d_name.name,
2567 put_task_struct(task);
2569 length = simple_read_from_buffer(buf, count, ppos, p, length);
2574 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2575 size_t count, loff_t *ppos)
2577 struct inode * inode = file->f_path.dentry->d_inode;
2580 struct task_struct *task = get_proc_task(inode);
2585 if (count > PAGE_SIZE)
2588 /* No partial writes. */
2594 page = (char*)__get_free_page(GFP_TEMPORARY);
2599 if (copy_from_user(page, buf, count))
2602 /* Guard against adverse ptrace interaction */
2603 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2607 length = security_setprocattr(task,
2608 (char*)file->f_path.dentry->d_name.name,
2609 (void*)page, count);
2610 mutex_unlock(&task->signal->cred_guard_mutex);
2612 free_page((unsigned long) page);
2614 put_task_struct(task);
2619 static const struct file_operations proc_pid_attr_operations = {
2620 .read = proc_pid_attr_read,
2621 .write = proc_pid_attr_write,
2622 .llseek = generic_file_llseek,
2625 static const struct pid_entry attr_dir_stuff[] = {
2626 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2627 REG("prev", S_IRUGO, proc_pid_attr_operations),
2628 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2629 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2630 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2631 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2634 static int proc_attr_dir_readdir(struct file * filp,
2635 void * dirent, filldir_t filldir)
2637 return proc_pident_readdir(filp,dirent,filldir,
2638 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2641 static const struct file_operations proc_attr_dir_operations = {
2642 .read = generic_read_dir,
2643 .readdir = proc_attr_dir_readdir,
2644 .llseek = default_llseek,
2647 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2648 struct dentry *dentry, unsigned int flags)
2650 return proc_pident_lookup(dir, dentry,
2651 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2654 static const struct inode_operations proc_attr_dir_inode_operations = {
2655 .lookup = proc_attr_dir_lookup,
2656 .getattr = pid_getattr,
2657 .setattr = proc_setattr,
2662 #ifdef CONFIG_ELF_CORE
2663 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2664 size_t count, loff_t *ppos)
2666 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2667 struct mm_struct *mm;
2668 char buffer[PROC_NUMBUF];
2676 mm = get_task_mm(task);
2678 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2679 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2680 MMF_DUMP_FILTER_SHIFT));
2682 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2685 put_task_struct(task);
2690 static ssize_t proc_coredump_filter_write(struct file *file,
2691 const char __user *buf,
2695 struct task_struct *task;
2696 struct mm_struct *mm;
2697 char buffer[PROC_NUMBUF], *end;
2704 memset(buffer, 0, sizeof(buffer));
2705 if (count > sizeof(buffer) - 1)
2706 count = sizeof(buffer) - 1;
2707 if (copy_from_user(buffer, buf, count))
2711 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2714 if (end - buffer == 0)
2718 task = get_proc_task(file->f_dentry->d_inode);
2723 mm = get_task_mm(task);
2727 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2729 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2731 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2736 put_task_struct(task);
2741 static const struct file_operations proc_coredump_filter_operations = {
2742 .read = proc_coredump_filter_read,
2743 .write = proc_coredump_filter_write,
2744 .llseek = generic_file_llseek,
2751 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2754 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2755 pid_t tgid = task_tgid_nr_ns(current, ns);
2756 char tmp[PROC_NUMBUF];
2759 sprintf(tmp, "%d", tgid);
2760 return vfs_readlink(dentry,buffer,buflen,tmp);
2763 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2765 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2766 pid_t tgid = task_tgid_nr_ns(current, ns);
2767 char *name = ERR_PTR(-ENOENT);
2771 name = ERR_PTR(-ENOMEM);
2773 sprintf(name, "%d", tgid);
2775 nd_set_link(nd, name);
2779 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2782 char *s = nd_get_link(nd);
2787 static const struct inode_operations proc_self_inode_operations = {
2788 .readlink = proc_self_readlink,
2789 .follow_link = proc_self_follow_link,
2790 .put_link = proc_self_put_link,
2796 * These are the directory entries in the root directory of /proc
2797 * that properly belong to the /proc filesystem, as they describe
2798 * describe something that is process related.
2800 static const struct pid_entry proc_base_stuff[] = {
2801 NOD("self", S_IFLNK|S_IRWXUGO,
2802 &proc_self_inode_operations, NULL, {}),
2805 static struct dentry *proc_base_instantiate(struct inode *dir,
2806 struct dentry *dentry, struct task_struct *task, const void *ptr)
2808 const struct pid_entry *p = ptr;
2809 struct inode *inode;
2810 struct proc_inode *ei;
2811 struct dentry *error;
2813 /* Allocate the inode */
2814 error = ERR_PTR(-ENOMEM);
2815 inode = new_inode(dir->i_sb);
2819 /* Initialize the inode */
2821 inode->i_ino = get_next_ino();
2822 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2825 * grab the reference to the task.
2827 ei->pid = get_task_pid(task, PIDTYPE_PID);
2831 inode->i_mode = p->mode;
2832 if (S_ISDIR(inode->i_mode))
2833 set_nlink(inode, 2);
2834 if (S_ISLNK(inode->i_mode))
2837 inode->i_op = p->iop;
2839 inode->i_fop = p->fop;
2841 d_add(dentry, inode);
2850 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2852 struct dentry *error;
2853 struct task_struct *task = get_proc_task(dir);
2854 const struct pid_entry *p, *last;
2856 error = ERR_PTR(-ENOENT);
2861 /* Lookup the directory entry */
2862 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2863 for (p = proc_base_stuff; p <= last; p++) {
2864 if (p->len != dentry->d_name.len)
2866 if (!memcmp(dentry->d_name.name, p->name, p->len))
2872 error = proc_base_instantiate(dir, dentry, task, p);
2875 put_task_struct(task);
2880 static int proc_base_fill_cache(struct file *filp, void *dirent,
2881 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2883 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2884 proc_base_instantiate, task, p);
2887 #ifdef CONFIG_TASK_IO_ACCOUNTING
2888 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2890 struct task_io_accounting acct = task->ioac;
2891 unsigned long flags;
2894 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2898 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2903 if (whole && lock_task_sighand(task, &flags)) {
2904 struct task_struct *t = task;
2906 task_io_accounting_add(&acct, &task->signal->ioac);
2907 while_each_thread(task, t)
2908 task_io_accounting_add(&acct, &t->ioac);
2910 unlock_task_sighand(task, &flags);
2912 result = sprintf(buffer,
2917 "read_bytes: %llu\n"
2918 "write_bytes: %llu\n"
2919 "cancelled_write_bytes: %llu\n",
2920 (unsigned long long)acct.rchar,
2921 (unsigned long long)acct.wchar,
2922 (unsigned long long)acct.syscr,
2923 (unsigned long long)acct.syscw,
2924 (unsigned long long)acct.read_bytes,
2925 (unsigned long long)acct.write_bytes,
2926 (unsigned long long)acct.cancelled_write_bytes);
2928 mutex_unlock(&task->signal->cred_guard_mutex);
2932 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2934 return do_io_accounting(task, buffer, 0);
2937 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2939 return do_io_accounting(task, buffer, 1);
2941 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2943 #ifdef CONFIG_USER_NS
2944 static int proc_id_map_open(struct inode *inode, struct file *file,
2945 struct seq_operations *seq_ops)
2947 struct user_namespace *ns = NULL;
2948 struct task_struct *task;
2949 struct seq_file *seq;
2952 task = get_proc_task(inode);
2955 ns = get_user_ns(task_cred_xxx(task, user_ns));
2957 put_task_struct(task);
2962 ret = seq_open(file, seq_ops);
2966 seq = file->private_data;
2976 static int proc_id_map_release(struct inode *inode, struct file *file)
2978 struct seq_file *seq = file->private_data;
2979 struct user_namespace *ns = seq->private;
2981 return seq_release(inode, file);
2984 static int proc_uid_map_open(struct inode *inode, struct file *file)
2986 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2989 static int proc_gid_map_open(struct inode *inode, struct file *file)
2991 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2994 static int proc_projid_map_open(struct inode *inode, struct file *file)
2996 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2999 static const struct file_operations proc_uid_map_operations = {
3000 .open = proc_uid_map_open,
3001 .write = proc_uid_map_write,
3003 .llseek = seq_lseek,
3004 .release = proc_id_map_release,
3007 static const struct file_operations proc_gid_map_operations = {
3008 .open = proc_gid_map_open,
3009 .write = proc_gid_map_write,
3011 .llseek = seq_lseek,
3012 .release = proc_id_map_release,
3015 static const struct file_operations proc_projid_map_operations = {
3016 .open = proc_projid_map_open,
3017 .write = proc_projid_map_write,
3019 .llseek = seq_lseek,
3020 .release = proc_id_map_release,
3022 #endif /* CONFIG_USER_NS */
3024 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
3025 struct pid *pid, struct task_struct *task)
3027 int err = lock_trace(task);
3029 seq_printf(m, "%08x\n", task->personality);
3038 static const struct file_operations proc_task_operations;
3039 static const struct inode_operations proc_task_inode_operations;
3041 static const struct pid_entry tgid_base_stuff[] = {
3042 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
3043 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3044 #ifdef CONFIG_CHECKPOINT_RESTORE
3045 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
3047 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3048 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3050 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3052 REG("environ", S_IRUSR, proc_environ_operations),
3053 INF("auxv", S_IRUSR, proc_pid_auxv),
3054 ONE("status", S_IRUGO, proc_pid_status),
3055 ONE("personality", S_IRUGO, proc_pid_personality),
3056 INF("limits", S_IRUGO, proc_pid_limits),
3057 #ifdef CONFIG_SCHED_DEBUG
3058 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3060 #ifdef CONFIG_SCHED_AUTOGROUP
3061 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
3063 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3064 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3065 INF("syscall", S_IRUGO, proc_pid_syscall),
3067 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3068 ONE("stat", S_IRUGO, proc_tgid_stat),
3069 ONE("statm", S_IRUGO, proc_pid_statm),
3070 REG("maps", S_IRUGO, proc_pid_maps_operations),
3072 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
3074 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3075 LNK("cwd", proc_cwd_link),
3076 LNK("root", proc_root_link),
3077 LNK("exe", proc_exe_link),
3078 REG("mounts", S_IRUGO, proc_mounts_operations),
3079 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3080 REG("mountstats", S_IRUSR, proc_mountstats_operations),
3081 #ifdef CONFIG_PROC_PAGE_MONITOR
3082 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3083 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
3084 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3086 #ifdef CONFIG_SECURITY
3087 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3089 #ifdef CONFIG_KALLSYMS
3090 INF("wchan", S_IRUGO, proc_pid_wchan),
3092 #ifdef CONFIG_STACKTRACE
3093 ONE("stack", S_IRUGO, proc_pid_stack),
3095 #ifdef CONFIG_SCHEDSTATS
3096 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3098 #ifdef CONFIG_LATENCYTOP
3099 REG("latency", S_IRUGO, proc_lstats_operations),
3101 #ifdef CONFIG_PROC_PID_CPUSET
3102 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3104 #ifdef CONFIG_CGROUPS
3105 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3107 INF("oom_score", S_IRUGO, proc_oom_score),
3108 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3109 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3110 #ifdef CONFIG_AUDITSYSCALL
3111 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3112 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3114 #ifdef CONFIG_FAULT_INJECTION
3115 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3117 #ifdef CONFIG_ELF_CORE
3118 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3120 #ifdef CONFIG_TASK_IO_ACCOUNTING
3121 INF("io", S_IRUSR, proc_tgid_io_accounting),
3123 #ifdef CONFIG_HARDWALL
3124 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3126 #ifdef CONFIG_USER_NS
3127 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3128 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3129 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3133 static int proc_tgid_base_readdir(struct file * filp,
3134 void * dirent, filldir_t filldir)
3136 return proc_pident_readdir(filp,dirent,filldir,
3137 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
3140 static const struct file_operations proc_tgid_base_operations = {
3141 .read = generic_read_dir,
3142 .readdir = proc_tgid_base_readdir,
3143 .llseek = default_llseek,
3146 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3148 return proc_pident_lookup(dir, dentry,
3149 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3152 static const struct inode_operations proc_tgid_base_inode_operations = {
3153 .lookup = proc_tgid_base_lookup,
3154 .getattr = pid_getattr,
3155 .setattr = proc_setattr,
3156 .permission = proc_pid_permission,
3159 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
3161 struct dentry *dentry, *leader, *dir;
3162 char buf[PROC_NUMBUF];
3166 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3167 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
3169 shrink_dcache_parent(dentry);
3175 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
3176 leader = d_hash_and_lookup(mnt->mnt_root, &name);
3181 name.len = strlen(name.name);
3182 dir = d_hash_and_lookup(leader, &name);
3184 goto out_put_leader;
3187 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3188 dentry = d_hash_and_lookup(dir, &name);
3190 shrink_dcache_parent(dentry);
3203 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3204 * @task: task that should be flushed.
3206 * When flushing dentries from proc, one needs to flush them from global
3207 * proc (proc_mnt) and from all the namespaces' procs this task was seen
3208 * in. This call is supposed to do all of this job.
3210 * Looks in the dcache for
3212 * /proc/@tgid/task/@pid
3213 * if either directory is present flushes it and all of it'ts children
3216 * It is safe and reasonable to cache /proc entries for a task until
3217 * that task exits. After that they just clog up the dcache with
3218 * useless entries, possibly causing useful dcache entries to be
3219 * flushed instead. This routine is proved to flush those useless
3220 * dcache entries at process exit time.
3222 * NOTE: This routine is just an optimization so it does not guarantee
3223 * that no dcache entries will exist at process exit time it
3224 * just makes it very unlikely that any will persist.
3227 void proc_flush_task(struct task_struct *task)
3230 struct pid *pid, *tgid;
3233 pid = task_pid(task);
3234 tgid = task_tgid(task);
3236 for (i = 0; i <= pid->level; i++) {
3237 upid = &pid->numbers[i];
3238 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
3239 tgid->numbers[i].nr);
3242 upid = &pid->numbers[pid->level];
3244 pid_ns_release_proc(upid->ns);
3247 static struct dentry *proc_pid_instantiate(struct inode *dir,
3248 struct dentry * dentry,
3249 struct task_struct *task, const void *ptr)
3251 struct dentry *error = ERR_PTR(-ENOENT);
3252 struct inode *inode;
3254 inode = proc_pid_make_inode(dir->i_sb, task);
3258 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3259 inode->i_op = &proc_tgid_base_inode_operations;
3260 inode->i_fop = &proc_tgid_base_operations;
3261 inode->i_flags|=S_IMMUTABLE;
3263 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
3264 ARRAY_SIZE(tgid_base_stuff)));
3266 d_set_d_op(dentry, &pid_dentry_operations);
3268 d_add(dentry, inode);
3269 /* Close the race of the process dying before we return the dentry */
3270 if (pid_revalidate(dentry, 0))
3276 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3278 struct dentry *result;
3279 struct task_struct *task;
3281 struct pid_namespace *ns;
3283 result = proc_base_lookup(dir, dentry);
3284 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3287 tgid = name_to_int(dentry);
3291 ns = dentry->d_sb->s_fs_info;
3293 task = find_task_by_pid_ns(tgid, ns);
3295 get_task_struct(task);
3300 result = proc_pid_instantiate(dir, dentry, task, NULL);
3301 put_task_struct(task);
3307 * Find the first task with tgid >= tgid
3312 struct task_struct *task;
3314 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3319 put_task_struct(iter.task);
3323 pid = find_ge_pid(iter.tgid, ns);
3325 iter.tgid = pid_nr_ns(pid, ns);
3326 iter.task = pid_task(pid, PIDTYPE_PID);
3327 /* What we to know is if the pid we have find is the
3328 * pid of a thread_group_leader. Testing for task
3329 * being a thread_group_leader is the obvious thing
3330 * todo but there is a window when it fails, due to
3331 * the pid transfer logic in de_thread.
3333 * So we perform the straight forward test of seeing
3334 * if the pid we have found is the pid of a thread
3335 * group leader, and don't worry if the task we have
3336 * found doesn't happen to be a thread group leader.
3337 * As we don't care in the case of readdir.
3339 if (!iter.task || !has_group_leader_pid(iter.task)) {
3343 get_task_struct(iter.task);
3349 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
3351 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3352 struct tgid_iter iter)
3354 char name[PROC_NUMBUF];
3355 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
3356 return proc_fill_cache(filp, dirent, filldir, name, len,
3357 proc_pid_instantiate, iter.task, NULL);
3360 static int fake_filldir(void *buf, const char *name, int namelen,
3361 loff_t offset, u64 ino, unsigned d_type)
3366 /* for the /proc/ directory itself, after non-process stuff has been done */
3367 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3370 struct task_struct *reaper;
3371 struct tgid_iter iter;
3372 struct pid_namespace *ns;
3373 filldir_t __filldir;
3375 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3377 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3379 reaper = get_proc_task(filp->f_path.dentry->d_inode);
3383 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
3384 const struct pid_entry *p = &proc_base_stuff[nr];
3385 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
3389 ns = filp->f_dentry->d_sb->s_fs_info;
3391 iter.tgid = filp->f_pos - TGID_OFFSET;
3392 for (iter = next_tgid(ns, iter);
3394 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3395 if (has_pid_permissions(ns, iter.task, 2))
3396 __filldir = filldir;
3398 __filldir = fake_filldir;
3400 filp->f_pos = iter.tgid + TGID_OFFSET;
3401 if (proc_pid_fill_cache(filp, dirent, __filldir, iter) < 0) {
3402 put_task_struct(iter.task);
3406 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3408 put_task_struct(reaper);
3416 static const struct pid_entry tid_base_stuff[] = {
3417 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3418 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3419 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3420 REG("environ", S_IRUSR, proc_environ_operations),
3421 INF("auxv", S_IRUSR, proc_pid_auxv),
3422 ONE("status", S_IRUGO, proc_pid_status),
3423 ONE("personality", S_IRUGO, proc_pid_personality),
3424 INF("limits", S_IRUGO, proc_pid_limits),
3425 #ifdef CONFIG_SCHED_DEBUG
3426 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3428 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3429 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3430 INF("syscall", S_IRUGO, proc_pid_syscall),
3432 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3433 ONE("stat", S_IRUGO, proc_tid_stat),
3434 ONE("statm", S_IRUGO, proc_pid_statm),
3435 REG("maps", S_IRUGO, proc_tid_maps_operations),
3436 #ifdef CONFIG_CHECKPOINT_RESTORE
3437 REG("children", S_IRUGO, proc_tid_children_operations),
3440 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
3442 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3443 LNK("cwd", proc_cwd_link),
3444 LNK("root", proc_root_link),
3445 LNK("exe", proc_exe_link),
3446 REG("mounts", S_IRUGO, proc_mounts_operations),
3447 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3448 #ifdef CONFIG_PROC_PAGE_MONITOR
3449 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3450 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
3451 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3453 #ifdef CONFIG_SECURITY
3454 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3456 #ifdef CONFIG_KALLSYMS
3457 INF("wchan", S_IRUGO, proc_pid_wchan),
3459 #ifdef CONFIG_STACKTRACE
3460 ONE("stack", S_IRUGO, proc_pid_stack),
3462 #ifdef CONFIG_SCHEDSTATS
3463 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3465 #ifdef CONFIG_LATENCYTOP
3466 REG("latency", S_IRUGO, proc_lstats_operations),
3468 #ifdef CONFIG_PROC_PID_CPUSET
3469 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3471 #ifdef CONFIG_CGROUPS
3472 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3474 INF("oom_score", S_IRUGO, proc_oom_score),
3475 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3476 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3477 #ifdef CONFIG_AUDITSYSCALL
3478 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3479 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3481 #ifdef CONFIG_FAULT_INJECTION
3482 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3484 #ifdef CONFIG_TASK_IO_ACCOUNTING
3485 INF("io", S_IRUSR, proc_tid_io_accounting),
3487 #ifdef CONFIG_HARDWALL
3488 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3490 #ifdef CONFIG_USER_NS
3491 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3492 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3493 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3497 static int proc_tid_base_readdir(struct file * filp,
3498 void * dirent, filldir_t filldir)
3500 return proc_pident_readdir(filp,dirent,filldir,
3501 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3504 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3506 return proc_pident_lookup(dir, dentry,
3507 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3510 static const struct file_operations proc_tid_base_operations = {
3511 .read = generic_read_dir,
3512 .readdir = proc_tid_base_readdir,
3513 .llseek = default_llseek,
3516 static const struct inode_operations proc_tid_base_inode_operations = {
3517 .lookup = proc_tid_base_lookup,
3518 .getattr = pid_getattr,
3519 .setattr = proc_setattr,
3522 static struct dentry *proc_task_instantiate(struct inode *dir,
3523 struct dentry *dentry, struct task_struct *task, const void *ptr)
3525 struct dentry *error = ERR_PTR(-ENOENT);
3526 struct inode *inode;
3527 inode = proc_pid_make_inode(dir->i_sb, task);
3531 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3532 inode->i_op = &proc_tid_base_inode_operations;
3533 inode->i_fop = &proc_tid_base_operations;
3534 inode->i_flags|=S_IMMUTABLE;
3536 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3537 ARRAY_SIZE(tid_base_stuff)));
3539 d_set_d_op(dentry, &pid_dentry_operations);
3541 d_add(dentry, inode);
3542 /* Close the race of the process dying before we return the dentry */
3543 if (pid_revalidate(dentry, 0))
3549 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3551 struct dentry *result = ERR_PTR(-ENOENT);
3552 struct task_struct *task;
3553 struct task_struct *leader = get_proc_task(dir);
3555 struct pid_namespace *ns;
3560 tid = name_to_int(dentry);
3564 ns = dentry->d_sb->s_fs_info;
3566 task = find_task_by_pid_ns(tid, ns);
3568 get_task_struct(task);
3572 if (!same_thread_group(leader, task))
3575 result = proc_task_instantiate(dir, dentry, task, NULL);
3577 put_task_struct(task);
3579 put_task_struct(leader);
3585 * Find the first tid of a thread group to return to user space.
3587 * Usually this is just the thread group leader, but if the users
3588 * buffer was too small or there was a seek into the middle of the
3589 * directory we have more work todo.
3591 * In the case of a short read we start with find_task_by_pid.
3593 * In the case of a seek we start with the leader and walk nr
3596 static struct task_struct *first_tid(struct task_struct *leader,
3597 int tid, int nr, struct pid_namespace *ns)
3599 struct task_struct *pos;
3602 /* Attempt to start with the pid of a thread */
3603 if (tid && (nr > 0)) {
3604 pos = find_task_by_pid_ns(tid, ns);
3605 if (pos && (pos->group_leader == leader))
3609 /* If nr exceeds the number of threads there is nothing todo */
3611 if (nr && nr >= get_nr_threads(leader))
3614 /* If we haven't found our starting place yet start
3615 * with the leader and walk nr threads forward.
3617 for (pos = leader; nr > 0; --nr) {
3618 pos = next_thread(pos);
3619 if (pos == leader) {
3625 get_task_struct(pos);
3632 * Find the next thread in the thread list.
3633 * Return NULL if there is an error or no next thread.
3635 * The reference to the input task_struct is released.
3637 static struct task_struct *next_tid(struct task_struct *start)
3639 struct task_struct *pos = NULL;
3641 if (pid_alive(start)) {
3642 pos = next_thread(start);
3643 if (thread_group_leader(pos))
3646 get_task_struct(pos);
3649 put_task_struct(start);
3653 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3654 struct task_struct *task, int tid)
3656 char name[PROC_NUMBUF];
3657 int len = snprintf(name, sizeof(name), "%d", tid);
3658 return proc_fill_cache(filp, dirent, filldir, name, len,
3659 proc_task_instantiate, task, NULL);
3662 /* for the /proc/TGID/task/ directories */
3663 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3665 struct dentry *dentry = filp->f_path.dentry;
3666 struct inode *inode = dentry->d_inode;
3667 struct task_struct *leader = NULL;
3668 struct task_struct *task;
3669 int retval = -ENOENT;
3672 struct pid_namespace *ns;
3674 task = get_proc_task(inode);
3678 if (pid_alive(task)) {
3679 leader = task->group_leader;
3680 get_task_struct(leader);
3683 put_task_struct(task);
3688 switch ((unsigned long)filp->f_pos) {
3691 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3696 ino = parent_ino(dentry);
3697 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3703 /* f_version caches the tgid value that the last readdir call couldn't
3704 * return. lseek aka telldir automagically resets f_version to 0.
3706 ns = filp->f_dentry->d_sb->s_fs_info;
3707 tid = (int)filp->f_version;
3708 filp->f_version = 0;
3709 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3711 task = next_tid(task), filp->f_pos++) {
3712 tid = task_pid_nr_ns(task, ns);
3713 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3714 /* returning this tgid failed, save it as the first
3715 * pid for the next readir call */
3716 filp->f_version = (u64)tid;
3717 put_task_struct(task);
3722 put_task_struct(leader);
3727 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3729 struct inode *inode = dentry->d_inode;
3730 struct task_struct *p = get_proc_task(inode);
3731 generic_fillattr(inode, stat);
3734 stat->nlink += get_nr_threads(p);
3741 static const struct inode_operations proc_task_inode_operations = {
3742 .lookup = proc_task_lookup,
3743 .getattr = proc_task_getattr,
3744 .setattr = proc_setattr,
3745 .permission = proc_pid_permission,
3748 static const struct file_operations proc_task_operations = {
3749 .read = generic_read_dir,
3750 .readdir = proc_task_readdir,
3751 .llseek = default_llseek,