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>
96 * Implementing inode permission operations in /proc is almost
97 * certainly an error. Permission checks need to happen during
98 * each system call not at open time. The reason is that most of
99 * what we wish to check for permissions in /proc varies at runtime.
101 * The classic example of a problem is opening file descriptors
102 * in /proc for a task before it execs a suid executable.
109 const struct inode_operations *iop;
110 const struct file_operations *fop;
114 #define NOD(NAME, MODE, IOP, FOP, OP) { \
116 .len = sizeof(NAME) - 1, \
123 #define DIR(NAME, MODE, iops, fops) \
124 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
125 #define LNK(NAME, get_link) \
126 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
127 &proc_pid_link_inode_operations, NULL, \
128 { .proc_get_link = get_link } )
129 #define REG(NAME, MODE, fops) \
130 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
131 #define INF(NAME, MODE, read) \
132 NOD(NAME, (S_IFREG|(MODE)), \
133 NULL, &proc_info_file_operations, \
134 { .proc_read = read } )
135 #define ONE(NAME, MODE, show) \
136 NOD(NAME, (S_IFREG|(MODE)), \
137 NULL, &proc_single_file_operations, \
138 { .proc_show = show } )
141 * Count the number of hardlinks for the pid_entry table, excluding the .
144 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
151 for (i = 0; i < n; ++i) {
152 if (S_ISDIR(entries[i].mode))
159 static int get_task_root(struct task_struct *task, struct path *root)
161 int result = -ENOENT;
165 get_fs_root(task->fs, root);
172 static int proc_cwd_link(struct dentry *dentry, struct path *path)
174 struct task_struct *task = get_proc_task(dentry->d_inode);
175 int result = -ENOENT;
180 get_fs_pwd(task->fs, path);
184 put_task_struct(task);
189 static int proc_root_link(struct dentry *dentry, struct path *path)
191 struct task_struct *task = get_proc_task(dentry->d_inode);
192 int result = -ENOENT;
195 result = get_task_root(task, path);
196 put_task_struct(task);
201 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
205 struct mm_struct *mm = get_task_mm(task);
209 goto out_mm; /* Shh! No looking before we're done */
211 len = mm->arg_end - mm->arg_start;
216 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
218 // If the nul at the end of args has been overwritten, then
219 // assume application is using setproctitle(3).
220 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
221 len = strnlen(buffer, res);
225 len = mm->env_end - mm->env_start;
226 if (len > PAGE_SIZE - res)
227 len = PAGE_SIZE - res;
228 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
229 res = strnlen(buffer, res);
238 static int proc_pid_auxv(struct task_struct *task, char *buffer)
240 struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ);
241 int res = PTR_ERR(mm);
242 if (mm && !IS_ERR(mm)) {
243 unsigned int nwords = 0;
246 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
247 res = nwords * sizeof(mm->saved_auxv[0]);
250 memcpy(buffer, mm->saved_auxv, res);
257 #ifdef CONFIG_KALLSYMS
259 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
260 * Returns the resolved symbol. If that fails, simply return the address.
262 static int proc_pid_wchan(struct task_struct *task, char *buffer)
265 char symname[KSYM_NAME_LEN];
267 wchan = get_wchan(task);
269 if (lookup_symbol_name(wchan, symname) < 0)
270 if (!ptrace_may_access(task, PTRACE_MODE_READ))
273 return sprintf(buffer, "%lu", wchan);
275 return sprintf(buffer, "%s", symname);
277 #endif /* CONFIG_KALLSYMS */
279 static int lock_trace(struct task_struct *task)
281 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
284 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
285 mutex_unlock(&task->signal->cred_guard_mutex);
291 static void unlock_trace(struct task_struct *task)
293 mutex_unlock(&task->signal->cred_guard_mutex);
296 #ifdef CONFIG_STACKTRACE
298 #define MAX_STACK_TRACE_DEPTH 64
300 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
301 struct pid *pid, struct task_struct *task)
303 struct stack_trace trace;
304 unsigned long *entries;
308 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
312 trace.nr_entries = 0;
313 trace.max_entries = MAX_STACK_TRACE_DEPTH;
314 trace.entries = entries;
317 err = lock_trace(task);
319 save_stack_trace_tsk(task, &trace);
321 for (i = 0; i < trace.nr_entries; i++) {
322 seq_printf(m, "[<%pK>] %pS\n",
323 (void *)entries[i], (void *)entries[i]);
333 #ifdef CONFIG_SCHEDSTATS
335 * Provides /proc/PID/schedstat
337 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
339 return sprintf(buffer, "%llu %llu %lu\n",
340 (unsigned long long)task->se.sum_exec_runtime,
341 (unsigned long long)task->sched_info.run_delay,
342 task->sched_info.pcount);
346 #ifdef CONFIG_LATENCYTOP
347 static int lstats_show_proc(struct seq_file *m, void *v)
350 struct inode *inode = m->private;
351 struct task_struct *task = get_proc_task(inode);
355 seq_puts(m, "Latency Top version : v0.1\n");
356 for (i = 0; i < 32; i++) {
357 struct latency_record *lr = &task->latency_record[i];
358 if (lr->backtrace[0]) {
360 seq_printf(m, "%i %li %li",
361 lr->count, lr->time, lr->max);
362 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
363 unsigned long bt = lr->backtrace[q];
368 seq_printf(m, " %ps", (void *)bt);
374 put_task_struct(task);
378 static int lstats_open(struct inode *inode, struct file *file)
380 return single_open(file, lstats_show_proc, inode);
383 static ssize_t lstats_write(struct file *file, const char __user *buf,
384 size_t count, loff_t *offs)
386 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
390 clear_all_latency_tracing(task);
391 put_task_struct(task);
396 static const struct file_operations proc_lstats_operations = {
399 .write = lstats_write,
401 .release = single_release,
406 static int proc_oom_score(struct task_struct *task, char *buffer)
408 unsigned long totalpages = totalram_pages + total_swap_pages;
409 unsigned long points = 0;
411 read_lock(&tasklist_lock);
413 points = oom_badness(task, NULL, NULL, totalpages) *
415 read_unlock(&tasklist_lock);
416 return sprintf(buffer, "%lu\n", points);
424 static const struct limit_names lnames[RLIM_NLIMITS] = {
425 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
426 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
427 [RLIMIT_DATA] = {"Max data size", "bytes"},
428 [RLIMIT_STACK] = {"Max stack size", "bytes"},
429 [RLIMIT_CORE] = {"Max core file size", "bytes"},
430 [RLIMIT_RSS] = {"Max resident set", "bytes"},
431 [RLIMIT_NPROC] = {"Max processes", "processes"},
432 [RLIMIT_NOFILE] = {"Max open files", "files"},
433 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
434 [RLIMIT_AS] = {"Max address space", "bytes"},
435 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
436 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
437 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
438 [RLIMIT_NICE] = {"Max nice priority", NULL},
439 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
440 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
443 /* Display limits for a process */
444 static int proc_pid_limits(struct task_struct *task, char *buffer)
449 char *bufptr = buffer;
451 struct rlimit rlim[RLIM_NLIMITS];
453 if (!lock_task_sighand(task, &flags))
455 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
456 unlock_task_sighand(task, &flags);
459 * print the file header
461 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
462 "Limit", "Soft Limit", "Hard Limit", "Units");
464 for (i = 0; i < RLIM_NLIMITS; i++) {
465 if (rlim[i].rlim_cur == RLIM_INFINITY)
466 count += sprintf(&bufptr[count], "%-25s %-20s ",
467 lnames[i].name, "unlimited");
469 count += sprintf(&bufptr[count], "%-25s %-20lu ",
470 lnames[i].name, rlim[i].rlim_cur);
472 if (rlim[i].rlim_max == RLIM_INFINITY)
473 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
475 count += sprintf(&bufptr[count], "%-20lu ",
479 count += sprintf(&bufptr[count], "%-10s\n",
482 count += sprintf(&bufptr[count], "\n");
488 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
489 static int proc_pid_syscall(struct task_struct *task, char *buffer)
492 unsigned long args[6], sp, pc;
493 int res = lock_trace(task);
497 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
498 res = sprintf(buffer, "running\n");
500 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
502 res = sprintf(buffer,
503 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
505 args[0], args[1], args[2], args[3], args[4], args[5],
510 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
512 /************************************************************************/
513 /* Here the fs part begins */
514 /************************************************************************/
516 /* permission checks */
517 static int proc_fd_access_allowed(struct inode *inode)
519 struct task_struct *task;
521 /* Allow access to a task's file descriptors if it is us or we
522 * may use ptrace attach to the process and find out that
525 task = get_proc_task(inode);
527 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
528 put_task_struct(task);
533 int proc_setattr(struct dentry *dentry, struct iattr *attr)
536 struct inode *inode = dentry->d_inode;
538 if (attr->ia_valid & ATTR_MODE)
541 error = inode_change_ok(inode, attr);
545 if ((attr->ia_valid & ATTR_SIZE) &&
546 attr->ia_size != i_size_read(inode)) {
547 error = vmtruncate(inode, attr->ia_size);
552 setattr_copy(inode, attr);
553 mark_inode_dirty(inode);
558 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
559 * or euid/egid (for hide_pid_min=2)?
561 static bool has_pid_permissions(struct pid_namespace *pid,
562 struct task_struct *task,
565 if (pid->hide_pid < hide_pid_min)
567 if (in_group_p(pid->pid_gid))
569 return ptrace_may_access(task, PTRACE_MODE_READ);
573 static int proc_pid_permission(struct inode *inode, int mask)
575 struct pid_namespace *pid = inode->i_sb->s_fs_info;
576 struct task_struct *task;
579 task = get_proc_task(inode);
582 has_perms = has_pid_permissions(pid, task, 1);
583 put_task_struct(task);
586 if (pid->hide_pid == 2) {
588 * Let's make getdents(), stat(), and open()
589 * consistent with each other. If a process
590 * may not stat() a file, it shouldn't be seen
598 return generic_permission(inode, mask);
603 static const struct inode_operations proc_def_inode_operations = {
604 .setattr = proc_setattr,
607 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
609 static ssize_t proc_info_read(struct file * file, char __user * buf,
610 size_t count, loff_t *ppos)
612 struct inode * inode = file->f_path.dentry->d_inode;
615 struct task_struct *task = get_proc_task(inode);
621 if (count > PROC_BLOCK_SIZE)
622 count = PROC_BLOCK_SIZE;
625 if (!(page = __get_free_page(GFP_TEMPORARY)))
628 length = PROC_I(inode)->op.proc_read(task, (char*)page);
631 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
634 put_task_struct(task);
639 static const struct file_operations proc_info_file_operations = {
640 .read = proc_info_read,
641 .llseek = generic_file_llseek,
644 static int proc_single_show(struct seq_file *m, void *v)
646 struct inode *inode = m->private;
647 struct pid_namespace *ns;
649 struct task_struct *task;
652 ns = inode->i_sb->s_fs_info;
653 pid = proc_pid(inode);
654 task = get_pid_task(pid, PIDTYPE_PID);
658 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
660 put_task_struct(task);
664 static int proc_single_open(struct inode *inode, struct file *filp)
666 return single_open(filp, proc_single_show, inode);
669 static const struct file_operations proc_single_file_operations = {
670 .open = proc_single_open,
673 .release = single_release,
676 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
678 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
679 struct mm_struct *mm;
684 mm = mm_access(task, mode);
685 put_task_struct(task);
691 /* ensure this mm_struct can't be freed */
692 atomic_inc(&mm->mm_count);
693 /* but do not pin its memory */
697 file->private_data = mm;
702 static int mem_open(struct inode *inode, struct file *file)
704 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
706 /* OK to pass negative loff_t, we can catch out-of-range */
707 file->f_mode |= FMODE_UNSIGNED_OFFSET;
712 static ssize_t mem_rw(struct file *file, char __user *buf,
713 size_t count, loff_t *ppos, int write)
715 struct mm_struct *mm = file->private_data;
716 unsigned long addr = *ppos;
723 page = (char *)__get_free_page(GFP_TEMPORARY);
728 if (!atomic_inc_not_zero(&mm->mm_users))
732 int this_len = min_t(int, count, PAGE_SIZE);
734 if (write && copy_from_user(page, buf, this_len)) {
739 this_len = access_remote_vm(mm, addr, page, this_len, write);
746 if (!write && copy_to_user(buf, page, this_len)) {
760 free_page((unsigned long) page);
764 static ssize_t mem_read(struct file *file, char __user *buf,
765 size_t count, loff_t *ppos)
767 return mem_rw(file, buf, count, ppos, 0);
770 static ssize_t mem_write(struct file *file, const char __user *buf,
771 size_t count, loff_t *ppos)
773 return mem_rw(file, (char __user*)buf, count, ppos, 1);
776 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
780 file->f_pos = offset;
783 file->f_pos += offset;
788 force_successful_syscall_return();
792 static int mem_release(struct inode *inode, struct file *file)
794 struct mm_struct *mm = file->private_data;
800 static const struct file_operations proc_mem_operations = {
805 .release = mem_release,
808 static int environ_open(struct inode *inode, struct file *file)
810 return __mem_open(inode, file, PTRACE_MODE_READ);
813 static ssize_t environ_read(struct file *file, char __user *buf,
814 size_t count, loff_t *ppos)
817 unsigned long src = *ppos;
819 struct mm_struct *mm = file->private_data;
824 page = (char *)__get_free_page(GFP_TEMPORARY);
829 if (!atomic_inc_not_zero(&mm->mm_users))
832 size_t this_len, max_len;
835 if (src >= (mm->env_end - mm->env_start))
838 this_len = mm->env_end - (mm->env_start + src);
840 max_len = min_t(size_t, PAGE_SIZE, count);
841 this_len = min(max_len, this_len);
843 retval = access_remote_vm(mm, (mm->env_start + src),
851 if (copy_to_user(buf, page, retval)) {
865 free_page((unsigned long) page);
869 static const struct file_operations proc_environ_operations = {
870 .open = environ_open,
871 .read = environ_read,
872 .llseek = generic_file_llseek,
873 .release = mem_release,
876 static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
879 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
880 char buffer[PROC_NUMBUF];
881 int oom_adj = OOM_ADJUST_MIN;
887 if (lock_task_sighand(task, &flags)) {
888 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
889 oom_adj = OOM_ADJUST_MAX;
891 oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
893 unlock_task_sighand(task, &flags);
895 put_task_struct(task);
896 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
897 return simple_read_from_buffer(buf, count, ppos, buffer, len);
900 static ssize_t oom_adj_write(struct file *file, const char __user *buf,
901 size_t count, loff_t *ppos)
903 struct task_struct *task;
904 char buffer[PROC_NUMBUF];
909 memset(buffer, 0, sizeof(buffer));
910 if (count > sizeof(buffer) - 1)
911 count = sizeof(buffer) - 1;
912 if (copy_from_user(buffer, buf, count)) {
917 err = kstrtoint(strstrip(buffer), 0, &oom_adj);
920 if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
921 oom_adj != OOM_DISABLE) {
926 task = get_proc_task(file->f_path.dentry->d_inode);
938 if (!lock_task_sighand(task, &flags)) {
944 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
945 * value is always attainable.
947 if (oom_adj == OOM_ADJUST_MAX)
948 oom_adj = OOM_SCORE_ADJ_MAX;
950 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
952 if (oom_adj < task->signal->oom_score_adj &&
953 !capable(CAP_SYS_RESOURCE)) {
959 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
960 * /proc/pid/oom_score_adj instead.
962 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
963 current->comm, task_pid_nr(current), task_pid_nr(task),
966 task->signal->oom_score_adj = oom_adj;
967 trace_oom_score_adj_update(task);
969 unlock_task_sighand(task, &flags);
972 put_task_struct(task);
974 return err < 0 ? err : count;
977 static const struct file_operations proc_oom_adj_operations = {
978 .read = oom_adj_read,
979 .write = oom_adj_write,
980 .llseek = generic_file_llseek,
983 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
984 size_t count, loff_t *ppos)
986 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
987 char buffer[PROC_NUMBUF];
988 int oom_score_adj = OOM_SCORE_ADJ_MIN;
994 if (lock_task_sighand(task, &flags)) {
995 oom_score_adj = task->signal->oom_score_adj;
996 unlock_task_sighand(task, &flags);
998 put_task_struct(task);
999 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1000 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1003 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1004 size_t count, loff_t *ppos)
1006 struct task_struct *task;
1007 char buffer[PROC_NUMBUF];
1008 unsigned long flags;
1012 memset(buffer, 0, sizeof(buffer));
1013 if (count > sizeof(buffer) - 1)
1014 count = sizeof(buffer) - 1;
1015 if (copy_from_user(buffer, buf, count)) {
1020 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1023 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1024 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1029 task = get_proc_task(file->f_path.dentry->d_inode);
1041 if (!lock_task_sighand(task, &flags)) {
1046 if (oom_score_adj < task->signal->oom_score_adj_min &&
1047 !capable(CAP_SYS_RESOURCE)) {
1052 task->signal->oom_score_adj = oom_score_adj;
1053 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1054 task->signal->oom_score_adj_min = oom_score_adj;
1055 trace_oom_score_adj_update(task);
1058 unlock_task_sighand(task, &flags);
1061 put_task_struct(task);
1063 return err < 0 ? err : count;
1066 static const struct file_operations proc_oom_score_adj_operations = {
1067 .read = oom_score_adj_read,
1068 .write = oom_score_adj_write,
1069 .llseek = default_llseek,
1072 #ifdef CONFIG_AUDITSYSCALL
1073 #define TMPBUFLEN 21
1074 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1075 size_t count, loff_t *ppos)
1077 struct inode * inode = file->f_path.dentry->d_inode;
1078 struct task_struct *task = get_proc_task(inode);
1080 char tmpbuf[TMPBUFLEN];
1084 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1085 from_kuid(file->f_cred->user_ns,
1086 audit_get_loginuid(task)));
1087 put_task_struct(task);
1088 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1091 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1092 size_t count, loff_t *ppos)
1094 struct inode * inode = file->f_path.dentry->d_inode;
1101 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1107 if (count >= PAGE_SIZE)
1108 count = PAGE_SIZE - 1;
1111 /* No partial writes. */
1114 page = (char*)__get_free_page(GFP_TEMPORARY);
1118 if (copy_from_user(page, buf, count))
1122 loginuid = simple_strtoul(page, &tmp, 10);
1128 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1129 if (!uid_valid(kloginuid)) {
1134 length = audit_set_loginuid(kloginuid);
1135 if (likely(length == 0))
1139 free_page((unsigned long) page);
1143 static const struct file_operations proc_loginuid_operations = {
1144 .read = proc_loginuid_read,
1145 .write = proc_loginuid_write,
1146 .llseek = generic_file_llseek,
1149 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1150 size_t count, loff_t *ppos)
1152 struct inode * inode = file->f_path.dentry->d_inode;
1153 struct task_struct *task = get_proc_task(inode);
1155 char tmpbuf[TMPBUFLEN];
1159 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1160 audit_get_sessionid(task));
1161 put_task_struct(task);
1162 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1165 static const struct file_operations proc_sessionid_operations = {
1166 .read = proc_sessionid_read,
1167 .llseek = generic_file_llseek,
1171 #ifdef CONFIG_FAULT_INJECTION
1172 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1173 size_t count, loff_t *ppos)
1175 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1176 char buffer[PROC_NUMBUF];
1182 make_it_fail = task->make_it_fail;
1183 put_task_struct(task);
1185 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1187 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1190 static ssize_t proc_fault_inject_write(struct file * file,
1191 const char __user * buf, size_t count, loff_t *ppos)
1193 struct task_struct *task;
1194 char buffer[PROC_NUMBUF], *end;
1197 if (!capable(CAP_SYS_RESOURCE))
1199 memset(buffer, 0, sizeof(buffer));
1200 if (count > sizeof(buffer) - 1)
1201 count = sizeof(buffer) - 1;
1202 if (copy_from_user(buffer, buf, count))
1204 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1207 task = get_proc_task(file->f_dentry->d_inode);
1210 task->make_it_fail = make_it_fail;
1211 put_task_struct(task);
1216 static const struct file_operations proc_fault_inject_operations = {
1217 .read = proc_fault_inject_read,
1218 .write = proc_fault_inject_write,
1219 .llseek = generic_file_llseek,
1224 #ifdef CONFIG_SCHED_DEBUG
1226 * Print out various scheduling related per-task fields:
1228 static int sched_show(struct seq_file *m, void *v)
1230 struct inode *inode = m->private;
1231 struct task_struct *p;
1233 p = get_proc_task(inode);
1236 proc_sched_show_task(p, m);
1244 sched_write(struct file *file, const char __user *buf,
1245 size_t count, loff_t *offset)
1247 struct inode *inode = file->f_path.dentry->d_inode;
1248 struct task_struct *p;
1250 p = get_proc_task(inode);
1253 proc_sched_set_task(p);
1260 static int sched_open(struct inode *inode, struct file *filp)
1262 return single_open(filp, sched_show, inode);
1265 static const struct file_operations proc_pid_sched_operations = {
1268 .write = sched_write,
1269 .llseek = seq_lseek,
1270 .release = single_release,
1275 #ifdef CONFIG_SCHED_AUTOGROUP
1277 * Print out autogroup related information:
1279 static int sched_autogroup_show(struct seq_file *m, void *v)
1281 struct inode *inode = m->private;
1282 struct task_struct *p;
1284 p = get_proc_task(inode);
1287 proc_sched_autogroup_show_task(p, m);
1295 sched_autogroup_write(struct file *file, const char __user *buf,
1296 size_t count, loff_t *offset)
1298 struct inode *inode = file->f_path.dentry->d_inode;
1299 struct task_struct *p;
1300 char buffer[PROC_NUMBUF];
1304 memset(buffer, 0, sizeof(buffer));
1305 if (count > sizeof(buffer) - 1)
1306 count = sizeof(buffer) - 1;
1307 if (copy_from_user(buffer, buf, count))
1310 err = kstrtoint(strstrip(buffer), 0, &nice);
1314 p = get_proc_task(inode);
1318 err = proc_sched_autogroup_set_nice(p, nice);
1327 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1331 ret = single_open(filp, sched_autogroup_show, NULL);
1333 struct seq_file *m = filp->private_data;
1340 static const struct file_operations proc_pid_sched_autogroup_operations = {
1341 .open = sched_autogroup_open,
1343 .write = sched_autogroup_write,
1344 .llseek = seq_lseek,
1345 .release = single_release,
1348 #endif /* CONFIG_SCHED_AUTOGROUP */
1350 static ssize_t comm_write(struct file *file, const char __user *buf,
1351 size_t count, loff_t *offset)
1353 struct inode *inode = file->f_path.dentry->d_inode;
1354 struct task_struct *p;
1355 char buffer[TASK_COMM_LEN];
1357 memset(buffer, 0, sizeof(buffer));
1358 if (count > sizeof(buffer) - 1)
1359 count = sizeof(buffer) - 1;
1360 if (copy_from_user(buffer, buf, count))
1363 p = get_proc_task(inode);
1367 if (same_thread_group(current, p))
1368 set_task_comm(p, buffer);
1377 static int comm_show(struct seq_file *m, void *v)
1379 struct inode *inode = m->private;
1380 struct task_struct *p;
1382 p = get_proc_task(inode);
1387 seq_printf(m, "%s\n", p->comm);
1395 static int comm_open(struct inode *inode, struct file *filp)
1397 return single_open(filp, comm_show, inode);
1400 static const struct file_operations proc_pid_set_comm_operations = {
1403 .write = comm_write,
1404 .llseek = seq_lseek,
1405 .release = single_release,
1408 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1410 struct task_struct *task;
1411 struct mm_struct *mm;
1412 struct file *exe_file;
1414 task = get_proc_task(dentry->d_inode);
1417 mm = get_task_mm(task);
1418 put_task_struct(task);
1421 exe_file = get_mm_exe_file(mm);
1424 *exe_path = exe_file->f_path;
1425 path_get(&exe_file->f_path);
1432 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1434 struct inode *inode = dentry->d_inode;
1436 int error = -EACCES;
1438 /* Are we allowed to snoop on the tasks file descriptors? */
1439 if (!proc_fd_access_allowed(inode))
1442 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1446 nd_jump_link(nd, &path);
1449 return ERR_PTR(error);
1452 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1454 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1461 pathname = d_path(path, tmp, PAGE_SIZE);
1462 len = PTR_ERR(pathname);
1463 if (IS_ERR(pathname))
1465 len = tmp + PAGE_SIZE - 1 - pathname;
1469 if (copy_to_user(buffer, pathname, len))
1472 free_page((unsigned long)tmp);
1476 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1478 int error = -EACCES;
1479 struct inode *inode = dentry->d_inode;
1482 /* Are we allowed to snoop on the tasks file descriptors? */
1483 if (!proc_fd_access_allowed(inode))
1486 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1490 error = do_proc_readlink(&path, buffer, buflen);
1496 const struct inode_operations proc_pid_link_inode_operations = {
1497 .readlink = proc_pid_readlink,
1498 .follow_link = proc_pid_follow_link,
1499 .setattr = proc_setattr,
1503 /* building an inode */
1505 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1507 struct inode * inode;
1508 struct proc_inode *ei;
1509 const struct cred *cred;
1511 /* We need a new inode */
1513 inode = new_inode(sb);
1519 inode->i_ino = get_next_ino();
1520 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1521 inode->i_op = &proc_def_inode_operations;
1524 * grab the reference to task.
1526 ei->pid = get_task_pid(task, PIDTYPE_PID);
1530 if (task_dumpable(task)) {
1532 cred = __task_cred(task);
1533 inode->i_uid = cred->euid;
1534 inode->i_gid = cred->egid;
1537 security_task_to_inode(task, inode);
1547 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1549 struct inode *inode = dentry->d_inode;
1550 struct task_struct *task;
1551 const struct cred *cred;
1552 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
1554 generic_fillattr(inode, stat);
1557 stat->uid = GLOBAL_ROOT_UID;
1558 stat->gid = GLOBAL_ROOT_GID;
1559 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1561 if (!has_pid_permissions(pid, task, 2)) {
1564 * This doesn't prevent learning whether PID exists,
1565 * it only makes getattr() consistent with readdir().
1569 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1570 task_dumpable(task)) {
1571 cred = __task_cred(task);
1572 stat->uid = cred->euid;
1573 stat->gid = cred->egid;
1583 * Exceptional case: normally we are not allowed to unhash a busy
1584 * directory. In this case, however, we can do it - no aliasing problems
1585 * due to the way we treat inodes.
1587 * Rewrite the inode's ownerships here because the owning task may have
1588 * performed a setuid(), etc.
1590 * Before the /proc/pid/status file was created the only way to read
1591 * the effective uid of a /process was to stat /proc/pid. Reading
1592 * /proc/pid/status is slow enough that procps and other packages
1593 * kept stating /proc/pid. To keep the rules in /proc simple I have
1594 * made this apply to all per process world readable and executable
1597 int pid_revalidate(struct dentry *dentry, unsigned int flags)
1599 struct inode *inode;
1600 struct task_struct *task;
1601 const struct cred *cred;
1603 if (flags & LOOKUP_RCU)
1606 inode = dentry->d_inode;
1607 task = get_proc_task(inode);
1610 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1611 task_dumpable(task)) {
1613 cred = __task_cred(task);
1614 inode->i_uid = cred->euid;
1615 inode->i_gid = cred->egid;
1618 inode->i_uid = GLOBAL_ROOT_UID;
1619 inode->i_gid = GLOBAL_ROOT_GID;
1621 inode->i_mode &= ~(S_ISUID | S_ISGID);
1622 security_task_to_inode(task, inode);
1623 put_task_struct(task);
1630 const struct dentry_operations pid_dentry_operations =
1632 .d_revalidate = pid_revalidate,
1633 .d_delete = pid_delete_dentry,
1639 * Fill a directory entry.
1641 * If possible create the dcache entry and derive our inode number and
1642 * file type from dcache entry.
1644 * Since all of the proc inode numbers are dynamically generated, the inode
1645 * numbers do not exist until the inode is cache. This means creating the
1646 * the dcache entry in readdir is necessary to keep the inode numbers
1647 * reported by readdir in sync with the inode numbers reported
1650 int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1651 const char *name, int len,
1652 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1654 struct dentry *child, *dir = filp->f_path.dentry;
1655 struct inode *inode;
1658 unsigned type = DT_UNKNOWN;
1662 qname.hash = full_name_hash(name, len);
1664 child = d_lookup(dir, &qname);
1667 new = d_alloc(dir, &qname);
1669 child = instantiate(dir->d_inode, new, task, ptr);
1676 if (!child || IS_ERR(child) || !child->d_inode)
1677 goto end_instantiate;
1678 inode = child->d_inode;
1681 type = inode->i_mode >> 12;
1686 ino = find_inode_number(dir, &qname);
1689 return filldir(dirent, name, len, filp->f_pos, ino, type);
1692 #ifdef CONFIG_CHECKPOINT_RESTORE
1695 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1696 * which represent vma start and end addresses.
1698 static int dname_to_vma_addr(struct dentry *dentry,
1699 unsigned long *start, unsigned long *end)
1701 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1707 static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
1709 unsigned long vm_start, vm_end;
1710 bool exact_vma_exists = false;
1711 struct mm_struct *mm = NULL;
1712 struct task_struct *task;
1713 const struct cred *cred;
1714 struct inode *inode;
1717 if (flags & LOOKUP_RCU)
1720 if (!capable(CAP_SYS_ADMIN)) {
1725 inode = dentry->d_inode;
1726 task = get_proc_task(inode);
1730 mm = mm_access(task, PTRACE_MODE_READ);
1731 if (IS_ERR_OR_NULL(mm))
1734 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1735 down_read(&mm->mmap_sem);
1736 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1737 up_read(&mm->mmap_sem);
1742 if (exact_vma_exists) {
1743 if (task_dumpable(task)) {
1745 cred = __task_cred(task);
1746 inode->i_uid = cred->euid;
1747 inode->i_gid = cred->egid;
1750 inode->i_uid = GLOBAL_ROOT_UID;
1751 inode->i_gid = GLOBAL_ROOT_GID;
1753 security_task_to_inode(task, inode);
1758 put_task_struct(task);
1767 static const struct dentry_operations tid_map_files_dentry_operations = {
1768 .d_revalidate = map_files_d_revalidate,
1769 .d_delete = pid_delete_dentry,
1772 static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
1774 unsigned long vm_start, vm_end;
1775 struct vm_area_struct *vma;
1776 struct task_struct *task;
1777 struct mm_struct *mm;
1781 task = get_proc_task(dentry->d_inode);
1785 mm = get_task_mm(task);
1786 put_task_struct(task);
1790 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
1794 down_read(&mm->mmap_sem);
1795 vma = find_exact_vma(mm, vm_start, vm_end);
1796 if (vma && vma->vm_file) {
1797 *path = vma->vm_file->f_path;
1801 up_read(&mm->mmap_sem);
1809 struct map_files_info {
1812 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
1815 static struct dentry *
1816 proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
1817 struct task_struct *task, const void *ptr)
1819 fmode_t mode = (fmode_t)(unsigned long)ptr;
1820 struct proc_inode *ei;
1821 struct inode *inode;
1823 inode = proc_pid_make_inode(dir->i_sb, task);
1825 return ERR_PTR(-ENOENT);
1828 ei->op.proc_get_link = proc_map_files_get_link;
1830 inode->i_op = &proc_pid_link_inode_operations;
1832 inode->i_mode = S_IFLNK;
1834 if (mode & FMODE_READ)
1835 inode->i_mode |= S_IRUSR;
1836 if (mode & FMODE_WRITE)
1837 inode->i_mode |= S_IWUSR;
1839 d_set_d_op(dentry, &tid_map_files_dentry_operations);
1840 d_add(dentry, inode);
1845 static struct dentry *proc_map_files_lookup(struct inode *dir,
1846 struct dentry *dentry, unsigned int flags)
1848 unsigned long vm_start, vm_end;
1849 struct vm_area_struct *vma;
1850 struct task_struct *task;
1851 struct dentry *result;
1852 struct mm_struct *mm;
1854 result = ERR_PTR(-EACCES);
1855 if (!capable(CAP_SYS_ADMIN))
1858 result = ERR_PTR(-ENOENT);
1859 task = get_proc_task(dir);
1863 result = ERR_PTR(-EACCES);
1864 if (!ptrace_may_access(task, PTRACE_MODE_READ))
1867 result = ERR_PTR(-ENOENT);
1868 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
1871 mm = get_task_mm(task);
1875 down_read(&mm->mmap_sem);
1876 vma = find_exact_vma(mm, vm_start, vm_end);
1881 result = proc_map_files_instantiate(dir, dentry, task,
1882 (void *)(unsigned long)vma->vm_file->f_mode);
1885 up_read(&mm->mmap_sem);
1888 put_task_struct(task);
1893 static const struct inode_operations proc_map_files_inode_operations = {
1894 .lookup = proc_map_files_lookup,
1895 .permission = proc_fd_permission,
1896 .setattr = proc_setattr,
1900 proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
1902 struct dentry *dentry = filp->f_path.dentry;
1903 struct inode *inode = dentry->d_inode;
1904 struct vm_area_struct *vma;
1905 struct task_struct *task;
1906 struct mm_struct *mm;
1911 if (!capable(CAP_SYS_ADMIN))
1915 task = get_proc_task(inode);
1920 if (!ptrace_may_access(task, PTRACE_MODE_READ))
1924 switch (filp->f_pos) {
1927 if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0)
1931 ino = parent_ino(dentry);
1932 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1937 unsigned long nr_files, pos, i;
1938 struct flex_array *fa = NULL;
1939 struct map_files_info info;
1940 struct map_files_info *p;
1942 mm = get_task_mm(task);
1945 down_read(&mm->mmap_sem);
1950 * We need two passes here:
1952 * 1) Collect vmas of mapped files with mmap_sem taken
1953 * 2) Release mmap_sem and instantiate entries
1955 * otherwise we get lockdep complained, since filldir()
1956 * routine might require mmap_sem taken in might_fault().
1959 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
1960 if (vma->vm_file && ++pos > filp->f_pos)
1965 fa = flex_array_alloc(sizeof(info), nr_files,
1967 if (!fa || flex_array_prealloc(fa, 0, nr_files,
1971 flex_array_free(fa);
1972 up_read(&mm->mmap_sem);
1976 for (i = 0, vma = mm->mmap, pos = 2; vma;
1977 vma = vma->vm_next) {
1980 if (++pos <= filp->f_pos)
1983 info.mode = vma->vm_file->f_mode;
1984 info.len = snprintf(info.name,
1985 sizeof(info.name), "%lx-%lx",
1986 vma->vm_start, vma->vm_end);
1987 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
1991 up_read(&mm->mmap_sem);
1993 for (i = 0; i < nr_files; i++) {
1994 p = flex_array_get(fa, i);
1995 ret = proc_fill_cache(filp, dirent, filldir,
1997 proc_map_files_instantiate,
1999 (void *)(unsigned long)p->mode);
2005 flex_array_free(fa);
2011 put_task_struct(task);
2016 static const struct file_operations proc_map_files_operations = {
2017 .read = generic_read_dir,
2018 .readdir = proc_map_files_readdir,
2019 .llseek = default_llseek,
2022 #endif /* CONFIG_CHECKPOINT_RESTORE */
2024 static struct dentry *proc_pident_instantiate(struct inode *dir,
2025 struct dentry *dentry, struct task_struct *task, const void *ptr)
2027 const struct pid_entry *p = ptr;
2028 struct inode *inode;
2029 struct proc_inode *ei;
2030 struct dentry *error = ERR_PTR(-ENOENT);
2032 inode = proc_pid_make_inode(dir->i_sb, task);
2037 inode->i_mode = p->mode;
2038 if (S_ISDIR(inode->i_mode))
2039 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2041 inode->i_op = p->iop;
2043 inode->i_fop = p->fop;
2045 d_set_d_op(dentry, &pid_dentry_operations);
2046 d_add(dentry, inode);
2047 /* Close the race of the process dying before we return the dentry */
2048 if (pid_revalidate(dentry, 0))
2054 static struct dentry *proc_pident_lookup(struct inode *dir,
2055 struct dentry *dentry,
2056 const struct pid_entry *ents,
2059 struct dentry *error;
2060 struct task_struct *task = get_proc_task(dir);
2061 const struct pid_entry *p, *last;
2063 error = ERR_PTR(-ENOENT);
2069 * Yes, it does not scale. And it should not. Don't add
2070 * new entries into /proc/<tgid>/ without very good reasons.
2072 last = &ents[nents - 1];
2073 for (p = ents; p <= last; p++) {
2074 if (p->len != dentry->d_name.len)
2076 if (!memcmp(dentry->d_name.name, p->name, p->len))
2082 error = proc_pident_instantiate(dir, dentry, task, p);
2084 put_task_struct(task);
2089 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2090 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2092 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2093 proc_pident_instantiate, task, p);
2096 static int proc_pident_readdir(struct file *filp,
2097 void *dirent, filldir_t filldir,
2098 const struct pid_entry *ents, unsigned int nents)
2101 struct dentry *dentry = filp->f_path.dentry;
2102 struct inode *inode = dentry->d_inode;
2103 struct task_struct *task = get_proc_task(inode);
2104 const struct pid_entry *p, *last;
2117 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2123 ino = parent_ino(dentry);
2124 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2136 last = &ents[nents - 1];
2138 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2147 put_task_struct(task);
2152 #ifdef CONFIG_SECURITY
2153 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2154 size_t count, loff_t *ppos)
2156 struct inode * inode = file->f_path.dentry->d_inode;
2159 struct task_struct *task = get_proc_task(inode);
2164 length = security_getprocattr(task,
2165 (char*)file->f_path.dentry->d_name.name,
2167 put_task_struct(task);
2169 length = simple_read_from_buffer(buf, count, ppos, p, length);
2174 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2175 size_t count, loff_t *ppos)
2177 struct inode * inode = file->f_path.dentry->d_inode;
2180 struct task_struct *task = get_proc_task(inode);
2185 if (count > PAGE_SIZE)
2188 /* No partial writes. */
2194 page = (char*)__get_free_page(GFP_TEMPORARY);
2199 if (copy_from_user(page, buf, count))
2202 /* Guard against adverse ptrace interaction */
2203 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2207 length = security_setprocattr(task,
2208 (char*)file->f_path.dentry->d_name.name,
2209 (void*)page, count);
2210 mutex_unlock(&task->signal->cred_guard_mutex);
2212 free_page((unsigned long) page);
2214 put_task_struct(task);
2219 static const struct file_operations proc_pid_attr_operations = {
2220 .read = proc_pid_attr_read,
2221 .write = proc_pid_attr_write,
2222 .llseek = generic_file_llseek,
2225 static const struct pid_entry attr_dir_stuff[] = {
2226 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2227 REG("prev", S_IRUGO, proc_pid_attr_operations),
2228 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2229 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2230 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2231 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2234 static int proc_attr_dir_readdir(struct file * filp,
2235 void * dirent, filldir_t filldir)
2237 return proc_pident_readdir(filp,dirent,filldir,
2238 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2241 static const struct file_operations proc_attr_dir_operations = {
2242 .read = generic_read_dir,
2243 .readdir = proc_attr_dir_readdir,
2244 .llseek = default_llseek,
2247 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2248 struct dentry *dentry, unsigned int flags)
2250 return proc_pident_lookup(dir, dentry,
2251 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2254 static const struct inode_operations proc_attr_dir_inode_operations = {
2255 .lookup = proc_attr_dir_lookup,
2256 .getattr = pid_getattr,
2257 .setattr = proc_setattr,
2262 #ifdef CONFIG_ELF_CORE
2263 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2264 size_t count, loff_t *ppos)
2266 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2267 struct mm_struct *mm;
2268 char buffer[PROC_NUMBUF];
2276 mm = get_task_mm(task);
2278 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2279 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2280 MMF_DUMP_FILTER_SHIFT));
2282 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2285 put_task_struct(task);
2290 static ssize_t proc_coredump_filter_write(struct file *file,
2291 const char __user *buf,
2295 struct task_struct *task;
2296 struct mm_struct *mm;
2297 char buffer[PROC_NUMBUF], *end;
2304 memset(buffer, 0, sizeof(buffer));
2305 if (count > sizeof(buffer) - 1)
2306 count = sizeof(buffer) - 1;
2307 if (copy_from_user(buffer, buf, count))
2311 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2314 if (end - buffer == 0)
2318 task = get_proc_task(file->f_dentry->d_inode);
2323 mm = get_task_mm(task);
2327 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2329 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2331 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2336 put_task_struct(task);
2341 static const struct file_operations proc_coredump_filter_operations = {
2342 .read = proc_coredump_filter_read,
2343 .write = proc_coredump_filter_write,
2344 .llseek = generic_file_llseek,
2351 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2354 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2355 pid_t tgid = task_tgid_nr_ns(current, ns);
2356 char tmp[PROC_NUMBUF];
2359 sprintf(tmp, "%d", tgid);
2360 return vfs_readlink(dentry,buffer,buflen,tmp);
2363 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2365 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2366 pid_t tgid = task_tgid_nr_ns(current, ns);
2367 char *name = ERR_PTR(-ENOENT);
2369 /* 11 for max length of signed int in decimal + NULL term */
2370 name = kmalloc(12, GFP_KERNEL);
2372 name = ERR_PTR(-ENOMEM);
2374 sprintf(name, "%d", tgid);
2376 nd_set_link(nd, name);
2380 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2383 char *s = nd_get_link(nd);
2388 static const struct inode_operations proc_self_inode_operations = {
2389 .readlink = proc_self_readlink,
2390 .follow_link = proc_self_follow_link,
2391 .put_link = proc_self_put_link,
2397 * These are the directory entries in the root directory of /proc
2398 * that properly belong to the /proc filesystem, as they describe
2399 * describe something that is process related.
2401 static const struct pid_entry proc_base_stuff[] = {
2402 NOD("self", S_IFLNK|S_IRWXUGO,
2403 &proc_self_inode_operations, NULL, {}),
2406 static struct dentry *proc_base_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;
2414 /* Allocate the inode */
2415 error = ERR_PTR(-ENOMEM);
2416 inode = new_inode(dir->i_sb);
2420 /* Initialize the inode */
2422 inode->i_ino = get_next_ino();
2423 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2426 * grab the reference to the task.
2428 ei->pid = get_task_pid(task, PIDTYPE_PID);
2432 inode->i_mode = p->mode;
2433 if (S_ISDIR(inode->i_mode))
2434 set_nlink(inode, 2);
2435 if (S_ISLNK(inode->i_mode))
2438 inode->i_op = p->iop;
2440 inode->i_fop = p->fop;
2442 d_add(dentry, inode);
2451 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2453 struct dentry *error;
2454 struct task_struct *task = get_proc_task(dir);
2455 const struct pid_entry *p, *last;
2457 error = ERR_PTR(-ENOENT);
2462 /* Lookup the directory entry */
2463 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2464 for (p = proc_base_stuff; p <= last; p++) {
2465 if (p->len != dentry->d_name.len)
2467 if (!memcmp(dentry->d_name.name, p->name, p->len))
2473 error = proc_base_instantiate(dir, dentry, task, p);
2476 put_task_struct(task);
2481 static int proc_base_fill_cache(struct file *filp, void *dirent,
2482 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2484 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2485 proc_base_instantiate, task, p);
2488 #ifdef CONFIG_TASK_IO_ACCOUNTING
2489 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2491 struct task_io_accounting acct = task->ioac;
2492 unsigned long flags;
2495 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2499 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2504 if (whole && lock_task_sighand(task, &flags)) {
2505 struct task_struct *t = task;
2507 task_io_accounting_add(&acct, &task->signal->ioac);
2508 while_each_thread(task, t)
2509 task_io_accounting_add(&acct, &t->ioac);
2511 unlock_task_sighand(task, &flags);
2513 result = sprintf(buffer,
2518 "read_bytes: %llu\n"
2519 "write_bytes: %llu\n"
2520 "cancelled_write_bytes: %llu\n",
2521 (unsigned long long)acct.rchar,
2522 (unsigned long long)acct.wchar,
2523 (unsigned long long)acct.syscr,
2524 (unsigned long long)acct.syscw,
2525 (unsigned long long)acct.read_bytes,
2526 (unsigned long long)acct.write_bytes,
2527 (unsigned long long)acct.cancelled_write_bytes);
2529 mutex_unlock(&task->signal->cred_guard_mutex);
2533 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2535 return do_io_accounting(task, buffer, 0);
2538 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2540 return do_io_accounting(task, buffer, 1);
2542 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2544 #ifdef CONFIG_USER_NS
2545 static int proc_id_map_open(struct inode *inode, struct file *file,
2546 struct seq_operations *seq_ops)
2548 struct user_namespace *ns = NULL;
2549 struct task_struct *task;
2550 struct seq_file *seq;
2553 task = get_proc_task(inode);
2556 ns = get_user_ns(task_cred_xxx(task, user_ns));
2558 put_task_struct(task);
2563 ret = seq_open(file, seq_ops);
2567 seq = file->private_data;
2577 static int proc_id_map_release(struct inode *inode, struct file *file)
2579 struct seq_file *seq = file->private_data;
2580 struct user_namespace *ns = seq->private;
2582 return seq_release(inode, file);
2585 static int proc_uid_map_open(struct inode *inode, struct file *file)
2587 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2590 static int proc_gid_map_open(struct inode *inode, struct file *file)
2592 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2595 static int proc_projid_map_open(struct inode *inode, struct file *file)
2597 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2600 static const struct file_operations proc_uid_map_operations = {
2601 .open = proc_uid_map_open,
2602 .write = proc_uid_map_write,
2604 .llseek = seq_lseek,
2605 .release = proc_id_map_release,
2608 static const struct file_operations proc_gid_map_operations = {
2609 .open = proc_gid_map_open,
2610 .write = proc_gid_map_write,
2612 .llseek = seq_lseek,
2613 .release = proc_id_map_release,
2616 static const struct file_operations proc_projid_map_operations = {
2617 .open = proc_projid_map_open,
2618 .write = proc_projid_map_write,
2620 .llseek = seq_lseek,
2621 .release = proc_id_map_release,
2623 #endif /* CONFIG_USER_NS */
2625 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2626 struct pid *pid, struct task_struct *task)
2628 int err = lock_trace(task);
2630 seq_printf(m, "%08x\n", task->personality);
2639 static const struct file_operations proc_task_operations;
2640 static const struct inode_operations proc_task_inode_operations;
2642 static const struct pid_entry tgid_base_stuff[] = {
2643 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2644 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2645 #ifdef CONFIG_CHECKPOINT_RESTORE
2646 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
2648 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2649 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2651 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2653 REG("environ", S_IRUSR, proc_environ_operations),
2654 INF("auxv", S_IRUSR, proc_pid_auxv),
2655 ONE("status", S_IRUGO, proc_pid_status),
2656 ONE("personality", S_IRUGO, proc_pid_personality),
2657 INF("limits", S_IRUGO, proc_pid_limits),
2658 #ifdef CONFIG_SCHED_DEBUG
2659 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2661 #ifdef CONFIG_SCHED_AUTOGROUP
2662 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2664 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2665 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2666 INF("syscall", S_IRUGO, proc_pid_syscall),
2668 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2669 ONE("stat", S_IRUGO, proc_tgid_stat),
2670 ONE("statm", S_IRUGO, proc_pid_statm),
2671 REG("maps", S_IRUGO, proc_pid_maps_operations),
2673 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
2675 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2676 LNK("cwd", proc_cwd_link),
2677 LNK("root", proc_root_link),
2678 LNK("exe", proc_exe_link),
2679 REG("mounts", S_IRUGO, proc_mounts_operations),
2680 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2681 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2682 #ifdef CONFIG_PROC_PAGE_MONITOR
2683 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2684 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
2685 REG("pagemap", S_IRUGO, proc_pagemap_operations),
2687 #ifdef CONFIG_SECURITY
2688 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2690 #ifdef CONFIG_KALLSYMS
2691 INF("wchan", S_IRUGO, proc_pid_wchan),
2693 #ifdef CONFIG_STACKTRACE
2694 ONE("stack", S_IRUGO, proc_pid_stack),
2696 #ifdef CONFIG_SCHEDSTATS
2697 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2699 #ifdef CONFIG_LATENCYTOP
2700 REG("latency", S_IRUGO, proc_lstats_operations),
2702 #ifdef CONFIG_PROC_PID_CPUSET
2703 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2705 #ifdef CONFIG_CGROUPS
2706 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2708 INF("oom_score", S_IRUGO, proc_oom_score),
2709 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
2710 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2711 #ifdef CONFIG_AUDITSYSCALL
2712 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2713 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2715 #ifdef CONFIG_FAULT_INJECTION
2716 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2718 #ifdef CONFIG_ELF_CORE
2719 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2721 #ifdef CONFIG_TASK_IO_ACCOUNTING
2722 INF("io", S_IRUSR, proc_tgid_io_accounting),
2724 #ifdef CONFIG_HARDWALL
2725 INF("hardwall", S_IRUGO, proc_pid_hardwall),
2727 #ifdef CONFIG_USER_NS
2728 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
2729 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
2730 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
2734 static int proc_tgid_base_readdir(struct file * filp,
2735 void * dirent, filldir_t filldir)
2737 return proc_pident_readdir(filp,dirent,filldir,
2738 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2741 static const struct file_operations proc_tgid_base_operations = {
2742 .read = generic_read_dir,
2743 .readdir = proc_tgid_base_readdir,
2744 .llseek = default_llseek,
2747 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2749 return proc_pident_lookup(dir, dentry,
2750 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2753 static const struct inode_operations proc_tgid_base_inode_operations = {
2754 .lookup = proc_tgid_base_lookup,
2755 .getattr = pid_getattr,
2756 .setattr = proc_setattr,
2757 .permission = proc_pid_permission,
2760 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2762 struct dentry *dentry, *leader, *dir;
2763 char buf[PROC_NUMBUF];
2767 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2768 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2770 shrink_dcache_parent(dentry);
2776 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2777 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2782 name.len = strlen(name.name);
2783 dir = d_hash_and_lookup(leader, &name);
2785 goto out_put_leader;
2788 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2789 dentry = d_hash_and_lookup(dir, &name);
2791 shrink_dcache_parent(dentry);
2804 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2805 * @task: task that should be flushed.
2807 * When flushing dentries from proc, one needs to flush them from global
2808 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2809 * in. This call is supposed to do all of this job.
2811 * Looks in the dcache for
2813 * /proc/@tgid/task/@pid
2814 * if either directory is present flushes it and all of it'ts children
2817 * It is safe and reasonable to cache /proc entries for a task until
2818 * that task exits. After that they just clog up the dcache with
2819 * useless entries, possibly causing useful dcache entries to be
2820 * flushed instead. This routine is proved to flush those useless
2821 * dcache entries at process exit time.
2823 * NOTE: This routine is just an optimization so it does not guarantee
2824 * that no dcache entries will exist at process exit time it
2825 * just makes it very unlikely that any will persist.
2828 void proc_flush_task(struct task_struct *task)
2831 struct pid *pid, *tgid;
2834 pid = task_pid(task);
2835 tgid = task_tgid(task);
2837 for (i = 0; i <= pid->level; i++) {
2838 upid = &pid->numbers[i];
2839 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2840 tgid->numbers[i].nr);
2843 upid = &pid->numbers[pid->level];
2845 pid_ns_release_proc(upid->ns);
2848 static struct dentry *proc_pid_instantiate(struct inode *dir,
2849 struct dentry * dentry,
2850 struct task_struct *task, const void *ptr)
2852 struct dentry *error = ERR_PTR(-ENOENT);
2853 struct inode *inode;
2855 inode = proc_pid_make_inode(dir->i_sb, task);
2859 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2860 inode->i_op = &proc_tgid_base_inode_operations;
2861 inode->i_fop = &proc_tgid_base_operations;
2862 inode->i_flags|=S_IMMUTABLE;
2864 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
2865 ARRAY_SIZE(tgid_base_stuff)));
2867 d_set_d_op(dentry, &pid_dentry_operations);
2869 d_add(dentry, inode);
2870 /* Close the race of the process dying before we return the dentry */
2871 if (pid_revalidate(dentry, 0))
2877 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
2879 struct dentry *result;
2880 struct task_struct *task;
2882 struct pid_namespace *ns;
2884 result = proc_base_lookup(dir, dentry);
2885 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2888 tgid = name_to_int(dentry);
2892 ns = dentry->d_sb->s_fs_info;
2894 task = find_task_by_pid_ns(tgid, ns);
2896 get_task_struct(task);
2901 result = proc_pid_instantiate(dir, dentry, task, NULL);
2902 put_task_struct(task);
2908 * Find the first task with tgid >= tgid
2913 struct task_struct *task;
2915 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2920 put_task_struct(iter.task);
2924 pid = find_ge_pid(iter.tgid, ns);
2926 iter.tgid = pid_nr_ns(pid, ns);
2927 iter.task = pid_task(pid, PIDTYPE_PID);
2928 /* What we to know is if the pid we have find is the
2929 * pid of a thread_group_leader. Testing for task
2930 * being a thread_group_leader is the obvious thing
2931 * todo but there is a window when it fails, due to
2932 * the pid transfer logic in de_thread.
2934 * So we perform the straight forward test of seeing
2935 * if the pid we have found is the pid of a thread
2936 * group leader, and don't worry if the task we have
2937 * found doesn't happen to be a thread group leader.
2938 * As we don't care in the case of readdir.
2940 if (!iter.task || !has_group_leader_pid(iter.task)) {
2944 get_task_struct(iter.task);
2950 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2952 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2953 struct tgid_iter iter)
2955 char name[PROC_NUMBUF];
2956 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
2957 return proc_fill_cache(filp, dirent, filldir, name, len,
2958 proc_pid_instantiate, iter.task, NULL);
2961 static int fake_filldir(void *buf, const char *name, int namelen,
2962 loff_t offset, u64 ino, unsigned d_type)
2967 /* for the /proc/ directory itself, after non-process stuff has been done */
2968 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2971 struct task_struct *reaper;
2972 struct tgid_iter iter;
2973 struct pid_namespace *ns;
2974 filldir_t __filldir;
2976 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
2978 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2980 reaper = get_proc_task(filp->f_path.dentry->d_inode);
2984 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
2985 const struct pid_entry *p = &proc_base_stuff[nr];
2986 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
2990 ns = filp->f_dentry->d_sb->s_fs_info;
2992 iter.tgid = filp->f_pos - TGID_OFFSET;
2993 for (iter = next_tgid(ns, iter);
2995 iter.tgid += 1, iter = next_tgid(ns, iter)) {
2996 if (has_pid_permissions(ns, iter.task, 2))
2997 __filldir = filldir;
2999 __filldir = fake_filldir;
3001 filp->f_pos = iter.tgid + TGID_OFFSET;
3002 if (proc_pid_fill_cache(filp, dirent, __filldir, iter) < 0) {
3003 put_task_struct(iter.task);
3007 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3009 put_task_struct(reaper);
3017 static const struct pid_entry tid_base_stuff[] = {
3018 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3019 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3020 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_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 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3030 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3031 INF("syscall", S_IRUGO, proc_pid_syscall),
3033 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3034 ONE("stat", S_IRUGO, proc_tid_stat),
3035 ONE("statm", S_IRUGO, proc_pid_statm),
3036 REG("maps", S_IRUGO, proc_tid_maps_operations),
3037 #ifdef CONFIG_CHECKPOINT_RESTORE
3038 REG("children", S_IRUGO, proc_tid_children_operations),
3041 REG("numa_maps", S_IRUGO, proc_tid_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 #ifdef CONFIG_PROC_PAGE_MONITOR
3050 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3051 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
3052 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3054 #ifdef CONFIG_SECURITY
3055 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3057 #ifdef CONFIG_KALLSYMS
3058 INF("wchan", S_IRUGO, proc_pid_wchan),
3060 #ifdef CONFIG_STACKTRACE
3061 ONE("stack", S_IRUGO, proc_pid_stack),
3063 #ifdef CONFIG_SCHEDSTATS
3064 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3066 #ifdef CONFIG_LATENCYTOP
3067 REG("latency", S_IRUGO, proc_lstats_operations),
3069 #ifdef CONFIG_PROC_PID_CPUSET
3070 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3072 #ifdef CONFIG_CGROUPS
3073 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3075 INF("oom_score", S_IRUGO, proc_oom_score),
3076 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3077 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3078 #ifdef CONFIG_AUDITSYSCALL
3079 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3080 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3082 #ifdef CONFIG_FAULT_INJECTION
3083 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3085 #ifdef CONFIG_TASK_IO_ACCOUNTING
3086 INF("io", S_IRUSR, proc_tid_io_accounting),
3088 #ifdef CONFIG_HARDWALL
3089 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3091 #ifdef CONFIG_USER_NS
3092 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3093 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3094 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3098 static int proc_tid_base_readdir(struct file * filp,
3099 void * dirent, filldir_t filldir)
3101 return proc_pident_readdir(filp,dirent,filldir,
3102 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3105 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3107 return proc_pident_lookup(dir, dentry,
3108 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3111 static const struct file_operations proc_tid_base_operations = {
3112 .read = generic_read_dir,
3113 .readdir = proc_tid_base_readdir,
3114 .llseek = default_llseek,
3117 static const struct inode_operations proc_tid_base_inode_operations = {
3118 .lookup = proc_tid_base_lookup,
3119 .getattr = pid_getattr,
3120 .setattr = proc_setattr,
3123 static struct dentry *proc_task_instantiate(struct inode *dir,
3124 struct dentry *dentry, struct task_struct *task, const void *ptr)
3126 struct dentry *error = ERR_PTR(-ENOENT);
3127 struct inode *inode;
3128 inode = proc_pid_make_inode(dir->i_sb, task);
3132 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3133 inode->i_op = &proc_tid_base_inode_operations;
3134 inode->i_fop = &proc_tid_base_operations;
3135 inode->i_flags|=S_IMMUTABLE;
3137 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3138 ARRAY_SIZE(tid_base_stuff)));
3140 d_set_d_op(dentry, &pid_dentry_operations);
3142 d_add(dentry, inode);
3143 /* Close the race of the process dying before we return the dentry */
3144 if (pid_revalidate(dentry, 0))
3150 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3152 struct dentry *result = ERR_PTR(-ENOENT);
3153 struct task_struct *task;
3154 struct task_struct *leader = get_proc_task(dir);
3156 struct pid_namespace *ns;
3161 tid = name_to_int(dentry);
3165 ns = dentry->d_sb->s_fs_info;
3167 task = find_task_by_pid_ns(tid, ns);
3169 get_task_struct(task);
3173 if (!same_thread_group(leader, task))
3176 result = proc_task_instantiate(dir, dentry, task, NULL);
3178 put_task_struct(task);
3180 put_task_struct(leader);
3186 * Find the first tid of a thread group to return to user space.
3188 * Usually this is just the thread group leader, but if the users
3189 * buffer was too small or there was a seek into the middle of the
3190 * directory we have more work todo.
3192 * In the case of a short read we start with find_task_by_pid.
3194 * In the case of a seek we start with the leader and walk nr
3197 static struct task_struct *first_tid(struct task_struct *leader,
3198 int tid, int nr, struct pid_namespace *ns)
3200 struct task_struct *pos;
3203 /* Attempt to start with the pid of a thread */
3204 if (tid && (nr > 0)) {
3205 pos = find_task_by_pid_ns(tid, ns);
3206 if (pos && (pos->group_leader == leader))
3210 /* If nr exceeds the number of threads there is nothing todo */
3212 if (nr && nr >= get_nr_threads(leader))
3215 /* If we haven't found our starting place yet start
3216 * with the leader and walk nr threads forward.
3218 for (pos = leader; nr > 0; --nr) {
3219 pos = next_thread(pos);
3220 if (pos == leader) {
3226 get_task_struct(pos);
3233 * Find the next thread in the thread list.
3234 * Return NULL if there is an error or no next thread.
3236 * The reference to the input task_struct is released.
3238 static struct task_struct *next_tid(struct task_struct *start)
3240 struct task_struct *pos = NULL;
3242 if (pid_alive(start)) {
3243 pos = next_thread(start);
3244 if (thread_group_leader(pos))
3247 get_task_struct(pos);
3250 put_task_struct(start);
3254 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3255 struct task_struct *task, int tid)
3257 char name[PROC_NUMBUF];
3258 int len = snprintf(name, sizeof(name), "%d", tid);
3259 return proc_fill_cache(filp, dirent, filldir, name, len,
3260 proc_task_instantiate, task, NULL);
3263 /* for the /proc/TGID/task/ directories */
3264 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3266 struct dentry *dentry = filp->f_path.dentry;
3267 struct inode *inode = dentry->d_inode;
3268 struct task_struct *leader = NULL;
3269 struct task_struct *task;
3270 int retval = -ENOENT;
3273 struct pid_namespace *ns;
3275 task = get_proc_task(inode);
3279 if (pid_alive(task)) {
3280 leader = task->group_leader;
3281 get_task_struct(leader);
3284 put_task_struct(task);
3289 switch ((unsigned long)filp->f_pos) {
3292 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3297 ino = parent_ino(dentry);
3298 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3304 /* f_version caches the tgid value that the last readdir call couldn't
3305 * return. lseek aka telldir automagically resets f_version to 0.
3307 ns = filp->f_dentry->d_sb->s_fs_info;
3308 tid = (int)filp->f_version;
3309 filp->f_version = 0;
3310 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3312 task = next_tid(task), filp->f_pos++) {
3313 tid = task_pid_nr_ns(task, ns);
3314 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3315 /* returning this tgid failed, save it as the first
3316 * pid for the next readir call */
3317 filp->f_version = (u64)tid;
3318 put_task_struct(task);
3323 put_task_struct(leader);
3328 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3330 struct inode *inode = dentry->d_inode;
3331 struct task_struct *p = get_proc_task(inode);
3332 generic_fillattr(inode, stat);
3335 stat->nlink += get_nr_threads(p);
3342 static const struct inode_operations proc_task_inode_operations = {
3343 .lookup = proc_task_lookup,
3344 .getattr = proc_task_getattr,
3345 .setattr = proc_setattr,
3346 .permission = proc_pid_permission,
3349 static const struct file_operations proc_task_operations = {
3350 .read = generic_read_dir,
3351 .readdir = proc_task_readdir,
3352 .llseek = default_llseek,