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/fs_struct.h>
85 #include <linux/slab.h>
86 #include <linux/flex_array.h>
87 #ifdef CONFIG_HARDWALL
88 #include <asm/hardwall.h>
90 #include <trace/events/oom.h>
94 * Implementing inode permission operations in /proc is almost
95 * certainly an error. Permission checks need to happen during
96 * each system call not at open time. The reason is that most of
97 * what we wish to check for permissions in /proc varies at runtime.
99 * The classic example of a problem is opening file descriptors
100 * in /proc for a task before it execs a suid executable.
107 const struct inode_operations *iop;
108 const struct file_operations *fop;
112 #define NOD(NAME, MODE, IOP, FOP, OP) { \
114 .len = sizeof(NAME) - 1, \
121 #define DIR(NAME, MODE, iops, fops) \
122 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
123 #define LNK(NAME, get_link) \
124 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
125 &proc_pid_link_inode_operations, NULL, \
126 { .proc_get_link = get_link } )
127 #define REG(NAME, MODE, fops) \
128 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
129 #define INF(NAME, MODE, read) \
130 NOD(NAME, (S_IFREG|(MODE)), \
131 NULL, &proc_info_file_operations, \
132 { .proc_read = read } )
133 #define ONE(NAME, MODE, show) \
134 NOD(NAME, (S_IFREG|(MODE)), \
135 NULL, &proc_single_file_operations, \
136 { .proc_show = show } )
138 static int proc_fd_permission(struct inode *inode, int mask);
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 struct mm_struct *mm_for_maps(struct task_struct *task)
203 return mm_access(task, PTRACE_MODE_READ);
206 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
210 struct mm_struct *mm = get_task_mm(task);
214 goto out_mm; /* Shh! No looking before we're done */
216 len = mm->arg_end - mm->arg_start;
221 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
223 // If the nul at the end of args has been overwritten, then
224 // assume application is using setproctitle(3).
225 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
226 len = strnlen(buffer, res);
230 len = mm->env_end - mm->env_start;
231 if (len > PAGE_SIZE - res)
232 len = PAGE_SIZE - res;
233 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
234 res = strnlen(buffer, res);
243 static int proc_pid_auxv(struct task_struct *task, char *buffer)
245 struct mm_struct *mm = mm_for_maps(task);
246 int res = PTR_ERR(mm);
247 if (mm && !IS_ERR(mm)) {
248 unsigned int nwords = 0;
251 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
252 res = nwords * sizeof(mm->saved_auxv[0]);
255 memcpy(buffer, mm->saved_auxv, res);
262 #ifdef CONFIG_KALLSYMS
264 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
265 * Returns the resolved symbol. If that fails, simply return the address.
267 static int proc_pid_wchan(struct task_struct *task, char *buffer)
270 char symname[KSYM_NAME_LEN];
272 wchan = get_wchan(task);
274 if (lookup_symbol_name(wchan, symname) < 0)
275 if (!ptrace_may_access(task, PTRACE_MODE_READ))
278 return sprintf(buffer, "%lu", wchan);
280 return sprintf(buffer, "%s", symname);
282 #endif /* CONFIG_KALLSYMS */
284 static int lock_trace(struct task_struct *task)
286 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
289 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
290 mutex_unlock(&task->signal->cred_guard_mutex);
296 static void unlock_trace(struct task_struct *task)
298 mutex_unlock(&task->signal->cred_guard_mutex);
301 #ifdef CONFIG_STACKTRACE
303 #define MAX_STACK_TRACE_DEPTH 64
305 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
306 struct pid *pid, struct task_struct *task)
308 struct stack_trace trace;
309 unsigned long *entries;
313 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
317 trace.nr_entries = 0;
318 trace.max_entries = MAX_STACK_TRACE_DEPTH;
319 trace.entries = entries;
322 err = lock_trace(task);
324 save_stack_trace_tsk(task, &trace);
326 for (i = 0; i < trace.nr_entries; i++) {
327 seq_printf(m, "[<%pK>] %pS\n",
328 (void *)entries[i], (void *)entries[i]);
338 #ifdef CONFIG_SCHEDSTATS
340 * Provides /proc/PID/schedstat
342 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
344 return sprintf(buffer, "%llu %llu %lu\n",
345 (unsigned long long)task->se.sum_exec_runtime,
346 (unsigned long long)task->sched_info.run_delay,
347 task->sched_info.pcount);
351 #ifdef CONFIG_LATENCYTOP
352 static int lstats_show_proc(struct seq_file *m, void *v)
355 struct inode *inode = m->private;
356 struct task_struct *task = get_proc_task(inode);
360 seq_puts(m, "Latency Top version : v0.1\n");
361 for (i = 0; i < 32; i++) {
362 struct latency_record *lr = &task->latency_record[i];
363 if (lr->backtrace[0]) {
365 seq_printf(m, "%i %li %li",
366 lr->count, lr->time, lr->max);
367 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
368 unsigned long bt = lr->backtrace[q];
373 seq_printf(m, " %ps", (void *)bt);
379 put_task_struct(task);
383 static int lstats_open(struct inode *inode, struct file *file)
385 return single_open(file, lstats_show_proc, inode);
388 static ssize_t lstats_write(struct file *file, const char __user *buf,
389 size_t count, loff_t *offs)
391 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
395 clear_all_latency_tracing(task);
396 put_task_struct(task);
401 static const struct file_operations proc_lstats_operations = {
404 .write = lstats_write,
406 .release = single_release,
411 static int proc_oom_score(struct task_struct *task, char *buffer)
413 unsigned long points = 0;
415 read_lock(&tasklist_lock);
417 points = oom_badness(task, NULL, NULL,
418 totalram_pages + total_swap_pages);
419 read_unlock(&tasklist_lock);
420 return sprintf(buffer, "%lu\n", points);
428 static const struct limit_names lnames[RLIM_NLIMITS] = {
429 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
430 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
431 [RLIMIT_DATA] = {"Max data size", "bytes"},
432 [RLIMIT_STACK] = {"Max stack size", "bytes"},
433 [RLIMIT_CORE] = {"Max core file size", "bytes"},
434 [RLIMIT_RSS] = {"Max resident set", "bytes"},
435 [RLIMIT_NPROC] = {"Max processes", "processes"},
436 [RLIMIT_NOFILE] = {"Max open files", "files"},
437 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
438 [RLIMIT_AS] = {"Max address space", "bytes"},
439 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
440 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
441 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
442 [RLIMIT_NICE] = {"Max nice priority", NULL},
443 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
444 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
447 /* Display limits for a process */
448 static int proc_pid_limits(struct task_struct *task, char *buffer)
453 char *bufptr = buffer;
455 struct rlimit rlim[RLIM_NLIMITS];
457 if (!lock_task_sighand(task, &flags))
459 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
460 unlock_task_sighand(task, &flags);
463 * print the file header
465 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
466 "Limit", "Soft Limit", "Hard Limit", "Units");
468 for (i = 0; i < RLIM_NLIMITS; i++) {
469 if (rlim[i].rlim_cur == RLIM_INFINITY)
470 count += sprintf(&bufptr[count], "%-25s %-20s ",
471 lnames[i].name, "unlimited");
473 count += sprintf(&bufptr[count], "%-25s %-20lu ",
474 lnames[i].name, rlim[i].rlim_cur);
476 if (rlim[i].rlim_max == RLIM_INFINITY)
477 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
479 count += sprintf(&bufptr[count], "%-20lu ",
483 count += sprintf(&bufptr[count], "%-10s\n",
486 count += sprintf(&bufptr[count], "\n");
492 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
493 static int proc_pid_syscall(struct task_struct *task, char *buffer)
496 unsigned long args[6], sp, pc;
497 int res = lock_trace(task);
501 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
502 res = sprintf(buffer, "running\n");
504 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
506 res = sprintf(buffer,
507 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
509 args[0], args[1], args[2], args[3], args[4], args[5],
514 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
516 /************************************************************************/
517 /* Here the fs part begins */
518 /************************************************************************/
520 /* permission checks */
521 static int proc_fd_access_allowed(struct inode *inode)
523 struct task_struct *task;
525 /* Allow access to a task's file descriptors if it is us or we
526 * may use ptrace attach to the process and find out that
529 task = get_proc_task(inode);
531 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
532 put_task_struct(task);
537 int proc_setattr(struct dentry *dentry, struct iattr *attr)
540 struct inode *inode = dentry->d_inode;
542 if (attr->ia_valid & ATTR_MODE)
545 error = inode_change_ok(inode, attr);
549 if ((attr->ia_valid & ATTR_SIZE) &&
550 attr->ia_size != i_size_read(inode)) {
551 error = vmtruncate(inode, attr->ia_size);
556 setattr_copy(inode, attr);
557 mark_inode_dirty(inode);
562 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
563 * or euid/egid (for hide_pid_min=2)?
565 static bool has_pid_permissions(struct pid_namespace *pid,
566 struct task_struct *task,
569 if (pid->hide_pid < hide_pid_min)
571 if (in_group_p(pid->pid_gid))
573 return ptrace_may_access(task, PTRACE_MODE_READ);
577 static int proc_pid_permission(struct inode *inode, int mask)
579 struct pid_namespace *pid = inode->i_sb->s_fs_info;
580 struct task_struct *task;
583 task = get_proc_task(inode);
586 has_perms = has_pid_permissions(pid, task, 1);
587 put_task_struct(task);
590 if (pid->hide_pid == 2) {
592 * Let's make getdents(), stat(), and open()
593 * consistent with each other. If a process
594 * may not stat() a file, it shouldn't be seen
602 return generic_permission(inode, mask);
607 static const struct inode_operations proc_def_inode_operations = {
608 .setattr = proc_setattr,
611 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
613 static ssize_t proc_info_read(struct file * file, char __user * buf,
614 size_t count, loff_t *ppos)
616 struct inode * inode = file->f_path.dentry->d_inode;
619 struct task_struct *task = get_proc_task(inode);
625 if (count > PROC_BLOCK_SIZE)
626 count = PROC_BLOCK_SIZE;
629 if (!(page = __get_free_page(GFP_TEMPORARY)))
632 length = PROC_I(inode)->op.proc_read(task, (char*)page);
635 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
638 put_task_struct(task);
643 static const struct file_operations proc_info_file_operations = {
644 .read = proc_info_read,
645 .llseek = generic_file_llseek,
648 static int proc_single_show(struct seq_file *m, void *v)
650 struct inode *inode = m->private;
651 struct pid_namespace *ns;
653 struct task_struct *task;
656 ns = inode->i_sb->s_fs_info;
657 pid = proc_pid(inode);
658 task = get_pid_task(pid, PIDTYPE_PID);
662 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
664 put_task_struct(task);
668 static int proc_single_open(struct inode *inode, struct file *filp)
670 return single_open(filp, proc_single_show, inode);
673 static const struct file_operations proc_single_file_operations = {
674 .open = proc_single_open,
677 .release = single_release,
680 static int mem_open(struct inode* inode, struct file* file)
682 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
683 struct mm_struct *mm;
688 mm = mm_access(task, PTRACE_MODE_ATTACH);
689 put_task_struct(task);
695 /* ensure this mm_struct can't be freed */
696 atomic_inc(&mm->mm_count);
697 /* but do not pin its memory */
701 /* OK to pass negative loff_t, we can catch out-of-range */
702 file->f_mode |= FMODE_UNSIGNED_OFFSET;
703 file->private_data = mm;
708 static ssize_t mem_rw(struct file *file, char __user *buf,
709 size_t count, loff_t *ppos, int write)
711 struct mm_struct *mm = file->private_data;
712 unsigned long addr = *ppos;
719 page = (char *)__get_free_page(GFP_TEMPORARY);
724 if (!atomic_inc_not_zero(&mm->mm_users))
728 int this_len = min_t(int, count, PAGE_SIZE);
730 if (write && copy_from_user(page, buf, this_len)) {
735 this_len = access_remote_vm(mm, addr, page, this_len, write);
742 if (!write && copy_to_user(buf, page, this_len)) {
756 free_page((unsigned long) page);
760 static ssize_t mem_read(struct file *file, char __user *buf,
761 size_t count, loff_t *ppos)
763 return mem_rw(file, buf, count, ppos, 0);
766 static ssize_t mem_write(struct file *file, const char __user *buf,
767 size_t count, loff_t *ppos)
769 return mem_rw(file, (char __user*)buf, count, ppos, 1);
772 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
776 file->f_pos = offset;
779 file->f_pos += offset;
784 force_successful_syscall_return();
788 static int mem_release(struct inode *inode, struct file *file)
790 struct mm_struct *mm = file->private_data;
796 static const struct file_operations proc_mem_operations = {
801 .release = mem_release,
804 static ssize_t environ_read(struct file *file, char __user *buf,
805 size_t count, loff_t *ppos)
807 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
809 unsigned long src = *ppos;
811 struct mm_struct *mm;
817 page = (char *)__get_free_page(GFP_TEMPORARY);
822 mm = mm_for_maps(task);
824 if (!mm || IS_ERR(mm))
829 int this_len, retval, max_len;
831 this_len = mm->env_end - (mm->env_start + src);
836 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
837 this_len = (this_len > max_len) ? max_len : this_len;
839 retval = access_process_vm(task, (mm->env_start + src),
847 if (copy_to_user(buf, page, retval)) {
861 free_page((unsigned long) page);
863 put_task_struct(task);
868 static const struct file_operations proc_environ_operations = {
869 .read = environ_read,
870 .llseek = generic_file_llseek,
873 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
874 size_t count, loff_t *ppos)
876 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
877 char buffer[PROC_NUMBUF];
879 int oom_adjust = OOM_DISABLE;
885 if (lock_task_sighand(task, &flags)) {
886 oom_adjust = task->signal->oom_adj;
887 unlock_task_sighand(task, &flags);
890 put_task_struct(task);
892 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
894 return simple_read_from_buffer(buf, count, ppos, buffer, len);
897 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
898 size_t count, loff_t *ppos)
900 struct task_struct *task;
901 char buffer[PROC_NUMBUF];
906 memset(buffer, 0, sizeof(buffer));
907 if (count > sizeof(buffer) - 1)
908 count = sizeof(buffer) - 1;
909 if (copy_from_user(buffer, buf, count)) {
914 err = kstrtoint(strstrip(buffer), 0, &oom_adjust);
917 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
918 oom_adjust != OOM_DISABLE) {
923 task = get_proc_task(file->f_path.dentry->d_inode);
935 if (!lock_task_sighand(task, &flags)) {
940 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
946 * Warn that /proc/pid/oom_adj is deprecated, see
947 * Documentation/feature-removal-schedule.txt.
949 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
950 current->comm, task_pid_nr(current), task_pid_nr(task),
952 task->signal->oom_adj = oom_adjust;
954 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
955 * value is always attainable.
957 if (task->signal->oom_adj == OOM_ADJUST_MAX)
958 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
960 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
962 trace_oom_score_adj_update(task);
964 unlock_task_sighand(task, &flags);
967 put_task_struct(task);
969 return err < 0 ? err : count;
972 static const struct file_operations proc_oom_adjust_operations = {
973 .read = oom_adjust_read,
974 .write = oom_adjust_write,
975 .llseek = generic_file_llseek,
978 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
979 size_t count, loff_t *ppos)
981 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
982 char buffer[PROC_NUMBUF];
983 int oom_score_adj = OOM_SCORE_ADJ_MIN;
989 if (lock_task_sighand(task, &flags)) {
990 oom_score_adj = task->signal->oom_score_adj;
991 unlock_task_sighand(task, &flags);
993 put_task_struct(task);
994 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
995 return simple_read_from_buffer(buf, count, ppos, buffer, len);
998 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
999 size_t count, loff_t *ppos)
1001 struct task_struct *task;
1002 char buffer[PROC_NUMBUF];
1003 unsigned long flags;
1007 memset(buffer, 0, sizeof(buffer));
1008 if (count > sizeof(buffer) - 1)
1009 count = sizeof(buffer) - 1;
1010 if (copy_from_user(buffer, buf, count)) {
1015 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1018 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1019 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1024 task = get_proc_task(file->f_path.dentry->d_inode);
1036 if (!lock_task_sighand(task, &flags)) {
1041 if (oom_score_adj < task->signal->oom_score_adj_min &&
1042 !capable(CAP_SYS_RESOURCE)) {
1047 task->signal->oom_score_adj = oom_score_adj;
1048 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1049 task->signal->oom_score_adj_min = oom_score_adj;
1050 trace_oom_score_adj_update(task);
1052 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1053 * always attainable.
1055 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1056 task->signal->oom_adj = OOM_DISABLE;
1058 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1061 unlock_task_sighand(task, &flags);
1064 put_task_struct(task);
1066 return err < 0 ? err : count;
1069 static const struct file_operations proc_oom_score_adj_operations = {
1070 .read = oom_score_adj_read,
1071 .write = oom_score_adj_write,
1072 .llseek = default_llseek,
1075 #ifdef CONFIG_AUDITSYSCALL
1076 #define TMPBUFLEN 21
1077 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1078 size_t count, loff_t *ppos)
1080 struct inode * inode = file->f_path.dentry->d_inode;
1081 struct task_struct *task = get_proc_task(inode);
1083 char tmpbuf[TMPBUFLEN];
1087 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1088 audit_get_loginuid(task));
1089 put_task_struct(task);
1090 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1093 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1094 size_t count, loff_t *ppos)
1096 struct inode * inode = file->f_path.dentry->d_inode;
1102 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1108 if (count >= PAGE_SIZE)
1109 count = PAGE_SIZE - 1;
1112 /* No partial writes. */
1115 page = (char*)__get_free_page(GFP_TEMPORARY);
1119 if (copy_from_user(page, buf, count))
1123 loginuid = simple_strtoul(page, &tmp, 10);
1129 length = audit_set_loginuid(loginuid);
1130 if (likely(length == 0))
1134 free_page((unsigned long) page);
1138 static const struct file_operations proc_loginuid_operations = {
1139 .read = proc_loginuid_read,
1140 .write = proc_loginuid_write,
1141 .llseek = generic_file_llseek,
1144 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1145 size_t count, loff_t *ppos)
1147 struct inode * inode = file->f_path.dentry->d_inode;
1148 struct task_struct *task = get_proc_task(inode);
1150 char tmpbuf[TMPBUFLEN];
1154 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1155 audit_get_sessionid(task));
1156 put_task_struct(task);
1157 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1160 static const struct file_operations proc_sessionid_operations = {
1161 .read = proc_sessionid_read,
1162 .llseek = generic_file_llseek,
1166 #ifdef CONFIG_FAULT_INJECTION
1167 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1168 size_t count, loff_t *ppos)
1170 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1171 char buffer[PROC_NUMBUF];
1177 make_it_fail = task->make_it_fail;
1178 put_task_struct(task);
1180 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1182 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1185 static ssize_t proc_fault_inject_write(struct file * file,
1186 const char __user * buf, size_t count, loff_t *ppos)
1188 struct task_struct *task;
1189 char buffer[PROC_NUMBUF], *end;
1192 if (!capable(CAP_SYS_RESOURCE))
1194 memset(buffer, 0, sizeof(buffer));
1195 if (count > sizeof(buffer) - 1)
1196 count = sizeof(buffer) - 1;
1197 if (copy_from_user(buffer, buf, count))
1199 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1202 task = get_proc_task(file->f_dentry->d_inode);
1205 task->make_it_fail = make_it_fail;
1206 put_task_struct(task);
1211 static const struct file_operations proc_fault_inject_operations = {
1212 .read = proc_fault_inject_read,
1213 .write = proc_fault_inject_write,
1214 .llseek = generic_file_llseek,
1219 #ifdef CONFIG_SCHED_DEBUG
1221 * Print out various scheduling related per-task fields:
1223 static int sched_show(struct seq_file *m, void *v)
1225 struct inode *inode = m->private;
1226 struct task_struct *p;
1228 p = get_proc_task(inode);
1231 proc_sched_show_task(p, m);
1239 sched_write(struct file *file, const char __user *buf,
1240 size_t count, loff_t *offset)
1242 struct inode *inode = file->f_path.dentry->d_inode;
1243 struct task_struct *p;
1245 p = get_proc_task(inode);
1248 proc_sched_set_task(p);
1255 static int sched_open(struct inode *inode, struct file *filp)
1257 return single_open(filp, sched_show, inode);
1260 static const struct file_operations proc_pid_sched_operations = {
1263 .write = sched_write,
1264 .llseek = seq_lseek,
1265 .release = single_release,
1270 #ifdef CONFIG_SCHED_AUTOGROUP
1272 * Print out autogroup related information:
1274 static int sched_autogroup_show(struct seq_file *m, void *v)
1276 struct inode *inode = m->private;
1277 struct task_struct *p;
1279 p = get_proc_task(inode);
1282 proc_sched_autogroup_show_task(p, m);
1290 sched_autogroup_write(struct file *file, const char __user *buf,
1291 size_t count, loff_t *offset)
1293 struct inode *inode = file->f_path.dentry->d_inode;
1294 struct task_struct *p;
1295 char buffer[PROC_NUMBUF];
1299 memset(buffer, 0, sizeof(buffer));
1300 if (count > sizeof(buffer) - 1)
1301 count = sizeof(buffer) - 1;
1302 if (copy_from_user(buffer, buf, count))
1305 err = kstrtoint(strstrip(buffer), 0, &nice);
1309 p = get_proc_task(inode);
1313 err = proc_sched_autogroup_set_nice(p, nice);
1322 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1326 ret = single_open(filp, sched_autogroup_show, NULL);
1328 struct seq_file *m = filp->private_data;
1335 static const struct file_operations proc_pid_sched_autogroup_operations = {
1336 .open = sched_autogroup_open,
1338 .write = sched_autogroup_write,
1339 .llseek = seq_lseek,
1340 .release = single_release,
1343 #endif /* CONFIG_SCHED_AUTOGROUP */
1345 static ssize_t comm_write(struct file *file, const char __user *buf,
1346 size_t count, loff_t *offset)
1348 struct inode *inode = file->f_path.dentry->d_inode;
1349 struct task_struct *p;
1350 char buffer[TASK_COMM_LEN];
1352 memset(buffer, 0, sizeof(buffer));
1353 if (count > sizeof(buffer) - 1)
1354 count = sizeof(buffer) - 1;
1355 if (copy_from_user(buffer, buf, count))
1358 p = get_proc_task(inode);
1362 if (same_thread_group(current, p))
1363 set_task_comm(p, buffer);
1372 static int comm_show(struct seq_file *m, void *v)
1374 struct inode *inode = m->private;
1375 struct task_struct *p;
1377 p = get_proc_task(inode);
1382 seq_printf(m, "%s\n", p->comm);
1390 static int comm_open(struct inode *inode, struct file *filp)
1392 return single_open(filp, comm_show, inode);
1395 static const struct file_operations proc_pid_set_comm_operations = {
1398 .write = comm_write,
1399 .llseek = seq_lseek,
1400 .release = single_release,
1403 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1405 struct task_struct *task;
1406 struct mm_struct *mm;
1407 struct file *exe_file;
1409 task = get_proc_task(dentry->d_inode);
1412 mm = get_task_mm(task);
1413 put_task_struct(task);
1416 exe_file = get_mm_exe_file(mm);
1419 *exe_path = exe_file->f_path;
1420 path_get(&exe_file->f_path);
1427 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1429 struct inode *inode = dentry->d_inode;
1430 int error = -EACCES;
1432 /* We don't need a base pointer in the /proc filesystem */
1433 path_put(&nd->path);
1435 /* Are we allowed to snoop on the tasks file descriptors? */
1436 if (!proc_fd_access_allowed(inode))
1439 error = PROC_I(inode)->op.proc_get_link(dentry, &nd->path);
1441 return ERR_PTR(error);
1444 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1446 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1453 pathname = d_path(path, tmp, PAGE_SIZE);
1454 len = PTR_ERR(pathname);
1455 if (IS_ERR(pathname))
1457 len = tmp + PAGE_SIZE - 1 - pathname;
1461 if (copy_to_user(buffer, pathname, len))
1464 free_page((unsigned long)tmp);
1468 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1470 int error = -EACCES;
1471 struct inode *inode = dentry->d_inode;
1474 /* Are we allowed to snoop on the tasks file descriptors? */
1475 if (!proc_fd_access_allowed(inode))
1478 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1482 error = do_proc_readlink(&path, buffer, buflen);
1488 static const struct inode_operations proc_pid_link_inode_operations = {
1489 .readlink = proc_pid_readlink,
1490 .follow_link = proc_pid_follow_link,
1491 .setattr = proc_setattr,
1495 /* building an inode */
1497 static int task_dumpable(struct task_struct *task)
1500 struct mm_struct *mm;
1505 dumpable = get_dumpable(mm);
1512 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1514 struct inode * inode;
1515 struct proc_inode *ei;
1516 const struct cred *cred;
1518 /* We need a new inode */
1520 inode = new_inode(sb);
1526 inode->i_ino = get_next_ino();
1527 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1528 inode->i_op = &proc_def_inode_operations;
1531 * grab the reference to task.
1533 ei->pid = get_task_pid(task, PIDTYPE_PID);
1537 if (task_dumpable(task)) {
1539 cred = __task_cred(task);
1540 inode->i_uid = cred->euid;
1541 inode->i_gid = cred->egid;
1544 security_task_to_inode(task, inode);
1554 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1556 struct inode *inode = dentry->d_inode;
1557 struct task_struct *task;
1558 const struct cred *cred;
1559 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
1561 generic_fillattr(inode, stat);
1566 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1568 if (!has_pid_permissions(pid, task, 2)) {
1571 * This doesn't prevent learning whether PID exists,
1572 * it only makes getattr() consistent with readdir().
1576 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1577 task_dumpable(task)) {
1578 cred = __task_cred(task);
1579 stat->uid = cred->euid;
1580 stat->gid = cred->egid;
1590 * Exceptional case: normally we are not allowed to unhash a busy
1591 * directory. In this case, however, we can do it - no aliasing problems
1592 * due to the way we treat inodes.
1594 * Rewrite the inode's ownerships here because the owning task may have
1595 * performed a setuid(), etc.
1597 * Before the /proc/pid/status file was created the only way to read
1598 * the effective uid of a /process was to stat /proc/pid. Reading
1599 * /proc/pid/status is slow enough that procps and other packages
1600 * kept stating /proc/pid. To keep the rules in /proc simple I have
1601 * made this apply to all per process world readable and executable
1604 int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1606 struct inode *inode;
1607 struct task_struct *task;
1608 const struct cred *cred;
1610 if (nd && nd->flags & LOOKUP_RCU)
1613 inode = dentry->d_inode;
1614 task = get_proc_task(inode);
1617 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1618 task_dumpable(task)) {
1620 cred = __task_cred(task);
1621 inode->i_uid = cred->euid;
1622 inode->i_gid = cred->egid;
1628 inode->i_mode &= ~(S_ISUID | S_ISGID);
1629 security_task_to_inode(task, inode);
1630 put_task_struct(task);
1637 static int pid_delete_dentry(const struct dentry * dentry)
1639 /* Is the task we represent dead?
1640 * If so, then don't put the dentry on the lru list,
1641 * kill it immediately.
1643 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1646 const struct dentry_operations pid_dentry_operations =
1648 .d_revalidate = pid_revalidate,
1649 .d_delete = pid_delete_dentry,
1655 * Fill a directory entry.
1657 * If possible create the dcache entry and derive our inode number and
1658 * file type from dcache entry.
1660 * Since all of the proc inode numbers are dynamically generated, the inode
1661 * numbers do not exist until the inode is cache. This means creating the
1662 * the dcache entry in readdir is necessary to keep the inode numbers
1663 * reported by readdir in sync with the inode numbers reported
1666 int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1667 const char *name, int len,
1668 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1670 struct dentry *child, *dir = filp->f_path.dentry;
1671 struct inode *inode;
1674 unsigned type = DT_UNKNOWN;
1678 qname.hash = full_name_hash(name, len);
1680 child = d_lookup(dir, &qname);
1683 new = d_alloc(dir, &qname);
1685 child = instantiate(dir->d_inode, new, task, ptr);
1692 if (!child || IS_ERR(child) || !child->d_inode)
1693 goto end_instantiate;
1694 inode = child->d_inode;
1697 type = inode->i_mode >> 12;
1702 ino = find_inode_number(dir, &qname);
1705 return filldir(dirent, name, len, filp->f_pos, ino, type);
1708 static unsigned name_to_int(struct dentry *dentry)
1710 const char *name = dentry->d_name.name;
1711 int len = dentry->d_name.len;
1714 if (len > 1 && *name == '0')
1717 unsigned c = *name++ - '0';
1720 if (n >= (~0U-9)/10)
1730 #define PROC_FDINFO_MAX 64
1732 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1734 struct task_struct *task = get_proc_task(inode);
1735 struct files_struct *files = NULL;
1737 int fd = proc_fd(inode);
1740 files = get_files_struct(task);
1741 put_task_struct(task);
1745 * We are not taking a ref to the file structure, so we must
1748 spin_lock(&files->file_lock);
1749 file = fcheck_files(files, fd);
1751 unsigned int f_flags;
1752 struct fdtable *fdt;
1754 fdt = files_fdtable(files);
1755 f_flags = file->f_flags & ~O_CLOEXEC;
1756 if (close_on_exec(fd, fdt))
1757 f_flags |= O_CLOEXEC;
1760 *path = file->f_path;
1761 path_get(&file->f_path);
1764 snprintf(info, PROC_FDINFO_MAX,
1767 (long long) file->f_pos,
1769 spin_unlock(&files->file_lock);
1770 put_files_struct(files);
1773 spin_unlock(&files->file_lock);
1774 put_files_struct(files);
1779 static int proc_fd_link(struct dentry *dentry, struct path *path)
1781 return proc_fd_info(dentry->d_inode, path, NULL);
1784 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1786 struct inode *inode;
1787 struct task_struct *task;
1789 struct files_struct *files;
1790 const struct cred *cred;
1792 if (nd && nd->flags & LOOKUP_RCU)
1795 inode = dentry->d_inode;
1796 task = get_proc_task(inode);
1797 fd = proc_fd(inode);
1800 files = get_files_struct(task);
1804 file = fcheck_files(files, fd);
1806 unsigned i_mode, f_mode = file->f_mode;
1809 put_files_struct(files);
1811 if (task_dumpable(task)) {
1813 cred = __task_cred(task);
1814 inode->i_uid = cred->euid;
1815 inode->i_gid = cred->egid;
1823 if (f_mode & FMODE_READ)
1824 i_mode |= S_IRUSR | S_IXUSR;
1825 if (f_mode & FMODE_WRITE)
1826 i_mode |= S_IWUSR | S_IXUSR;
1827 inode->i_mode = i_mode;
1829 security_task_to_inode(task, inode);
1830 put_task_struct(task);
1834 put_files_struct(files);
1836 put_task_struct(task);
1842 static const struct dentry_operations tid_fd_dentry_operations =
1844 .d_revalidate = tid_fd_revalidate,
1845 .d_delete = pid_delete_dentry,
1848 static struct dentry *proc_fd_instantiate(struct inode *dir,
1849 struct dentry *dentry, struct task_struct *task, const void *ptr)
1851 unsigned fd = *(const unsigned *)ptr;
1852 struct inode *inode;
1853 struct proc_inode *ei;
1854 struct dentry *error = ERR_PTR(-ENOENT);
1856 inode = proc_pid_make_inode(dir->i_sb, task);
1862 inode->i_op = &proc_pid_link_inode_operations;
1864 ei->op.proc_get_link = proc_fd_link;
1865 d_set_d_op(dentry, &tid_fd_dentry_operations);
1866 d_add(dentry, inode);
1867 /* Close the race of the process dying before we return the dentry */
1868 if (tid_fd_revalidate(dentry, NULL))
1875 static struct dentry *proc_lookupfd_common(struct inode *dir,
1876 struct dentry *dentry,
1877 instantiate_t instantiate)
1879 struct task_struct *task = get_proc_task(dir);
1880 unsigned fd = name_to_int(dentry);
1881 struct dentry *result = ERR_PTR(-ENOENT);
1888 result = instantiate(dir, dentry, task, &fd);
1890 put_task_struct(task);
1895 static int proc_readfd_common(struct file * filp, void * dirent,
1896 filldir_t filldir, instantiate_t instantiate)
1898 struct dentry *dentry = filp->f_path.dentry;
1899 struct inode *inode = dentry->d_inode;
1900 struct task_struct *p = get_proc_task(inode);
1901 unsigned int fd, ino;
1903 struct files_struct * files;
1913 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1917 ino = parent_ino(dentry);
1918 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1922 files = get_files_struct(p);
1926 for (fd = filp->f_pos-2;
1927 fd < files_fdtable(files)->max_fds;
1928 fd++, filp->f_pos++) {
1929 char name[PROC_NUMBUF];
1932 if (!fcheck_files(files, fd))
1936 len = snprintf(name, sizeof(name), "%d", fd);
1937 if (proc_fill_cache(filp, dirent, filldir,
1938 name, len, instantiate,
1946 put_files_struct(files);
1954 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1955 struct nameidata *nd)
1957 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1960 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1962 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1965 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1966 size_t len, loff_t *ppos)
1968 char tmp[PROC_FDINFO_MAX];
1969 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
1971 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1975 static const struct file_operations proc_fdinfo_file_operations = {
1976 .open = nonseekable_open,
1977 .read = proc_fdinfo_read,
1978 .llseek = no_llseek,
1981 static const struct file_operations proc_fd_operations = {
1982 .read = generic_read_dir,
1983 .readdir = proc_readfd,
1984 .llseek = default_llseek,
1987 #ifdef CONFIG_CHECKPOINT_RESTORE
1990 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1991 * which represent vma start and end addresses.
1993 static int dname_to_vma_addr(struct dentry *dentry,
1994 unsigned long *start, unsigned long *end)
1996 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
2002 static int map_files_d_revalidate(struct dentry *dentry, struct nameidata *nd)
2004 unsigned long vm_start, vm_end;
2005 bool exact_vma_exists = false;
2006 struct mm_struct *mm = NULL;
2007 struct task_struct *task;
2008 const struct cred *cred;
2009 struct inode *inode;
2012 if (nd && nd->flags & LOOKUP_RCU)
2015 if (!capable(CAP_SYS_ADMIN)) {
2020 inode = dentry->d_inode;
2021 task = get_proc_task(inode);
2025 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2028 mm = get_task_mm(task);
2032 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
2033 down_read(&mm->mmap_sem);
2034 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
2035 up_read(&mm->mmap_sem);
2040 if (exact_vma_exists) {
2041 if (task_dumpable(task)) {
2043 cred = __task_cred(task);
2044 inode->i_uid = cred->euid;
2045 inode->i_gid = cred->egid;
2051 security_task_to_inode(task, inode);
2056 put_task_struct(task);
2065 static const struct dentry_operations tid_map_files_dentry_operations = {
2066 .d_revalidate = map_files_d_revalidate,
2067 .d_delete = pid_delete_dentry,
2070 static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
2072 unsigned long vm_start, vm_end;
2073 struct vm_area_struct *vma;
2074 struct task_struct *task;
2075 struct mm_struct *mm;
2079 task = get_proc_task(dentry->d_inode);
2083 mm = get_task_mm(task);
2084 put_task_struct(task);
2088 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2092 down_read(&mm->mmap_sem);
2093 vma = find_exact_vma(mm, vm_start, vm_end);
2094 if (vma && vma->vm_file) {
2095 *path = vma->vm_file->f_path;
2099 up_read(&mm->mmap_sem);
2107 struct map_files_info {
2110 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
2113 static struct dentry *
2114 proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
2115 struct task_struct *task, const void *ptr)
2117 const struct file *file = ptr;
2118 struct proc_inode *ei;
2119 struct inode *inode;
2122 return ERR_PTR(-ENOENT);
2124 inode = proc_pid_make_inode(dir->i_sb, task);
2126 return ERR_PTR(-ENOENT);
2129 ei->op.proc_get_link = proc_map_files_get_link;
2131 inode->i_op = &proc_pid_link_inode_operations;
2133 inode->i_mode = S_IFLNK;
2135 if (file->f_mode & FMODE_READ)
2136 inode->i_mode |= S_IRUSR;
2137 if (file->f_mode & FMODE_WRITE)
2138 inode->i_mode |= S_IWUSR;
2140 d_set_d_op(dentry, &tid_map_files_dentry_operations);
2141 d_add(dentry, inode);
2146 static struct dentry *proc_map_files_lookup(struct inode *dir,
2147 struct dentry *dentry, struct nameidata *nd)
2149 unsigned long vm_start, vm_end;
2150 struct vm_area_struct *vma;
2151 struct task_struct *task;
2152 struct dentry *result;
2153 struct mm_struct *mm;
2155 result = ERR_PTR(-EACCES);
2156 if (!capable(CAP_SYS_ADMIN))
2159 result = ERR_PTR(-ENOENT);
2160 task = get_proc_task(dir);
2164 result = ERR_PTR(-EACCES);
2165 if (lock_trace(task))
2168 result = ERR_PTR(-ENOENT);
2169 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2172 mm = get_task_mm(task);
2176 down_read(&mm->mmap_sem);
2177 vma = find_exact_vma(mm, vm_start, vm_end);
2181 result = proc_map_files_instantiate(dir, dentry, task, vma->vm_file);
2184 up_read(&mm->mmap_sem);
2189 put_task_struct(task);
2194 static const struct inode_operations proc_map_files_inode_operations = {
2195 .lookup = proc_map_files_lookup,
2196 .permission = proc_fd_permission,
2197 .setattr = proc_setattr,
2201 proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
2203 struct dentry *dentry = filp->f_path.dentry;
2204 struct inode *inode = dentry->d_inode;
2205 struct vm_area_struct *vma;
2206 struct task_struct *task;
2207 struct mm_struct *mm;
2212 if (!capable(CAP_SYS_ADMIN))
2216 task = get_proc_task(inode);
2221 if (lock_trace(task))
2225 switch (filp->f_pos) {
2228 if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0)
2232 ino = parent_ino(dentry);
2233 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2238 unsigned long nr_files, pos, i;
2239 struct flex_array *fa = NULL;
2240 struct map_files_info info;
2241 struct map_files_info *p;
2243 mm = get_task_mm(task);
2246 down_read(&mm->mmap_sem);
2251 * We need two passes here:
2253 * 1) Collect vmas of mapped files with mmap_sem taken
2254 * 2) Release mmap_sem and instantiate entries
2256 * otherwise we get lockdep complained, since filldir()
2257 * routine might require mmap_sem taken in might_fault().
2260 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2261 if (vma->vm_file && ++pos > filp->f_pos)
2266 fa = flex_array_alloc(sizeof(info), nr_files,
2268 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2272 flex_array_free(fa);
2273 up_read(&mm->mmap_sem);
2277 for (i = 0, vma = mm->mmap, pos = 2; vma;
2278 vma = vma->vm_next) {
2281 if (++pos <= filp->f_pos)
2284 get_file(vma->vm_file);
2285 info.file = vma->vm_file;
2286 info.len = snprintf(info.name,
2287 sizeof(info.name), "%lx-%lx",
2288 vma->vm_start, vma->vm_end);
2289 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2293 up_read(&mm->mmap_sem);
2295 for (i = 0; i < nr_files; i++) {
2296 p = flex_array_get(fa, i);
2297 ret = proc_fill_cache(filp, dirent, filldir,
2299 proc_map_files_instantiate,
2306 for (; i < nr_files; i++) {
2308 * In case of error don't forget
2309 * to put rest of file refs.
2311 p = flex_array_get(fa, i);
2315 flex_array_free(fa);
2323 put_task_struct(task);
2328 static const struct file_operations proc_map_files_operations = {
2329 .read = generic_read_dir,
2330 .readdir = proc_map_files_readdir,
2331 .llseek = default_llseek,
2334 #endif /* CONFIG_CHECKPOINT_RESTORE */
2337 * /proc/pid/fd needs a special permission handler so that a process can still
2338 * access /proc/self/fd after it has executed a setuid().
2340 static int proc_fd_permission(struct inode *inode, int mask)
2342 int rv = generic_permission(inode, mask);
2345 if (task_pid(current) == proc_pid(inode))
2351 * proc directories can do almost nothing..
2353 static const struct inode_operations proc_fd_inode_operations = {
2354 .lookup = proc_lookupfd,
2355 .permission = proc_fd_permission,
2356 .setattr = proc_setattr,
2359 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2360 struct dentry *dentry, struct task_struct *task, const void *ptr)
2362 unsigned fd = *(unsigned *)ptr;
2363 struct inode *inode;
2364 struct proc_inode *ei;
2365 struct dentry *error = ERR_PTR(-ENOENT);
2367 inode = proc_pid_make_inode(dir->i_sb, task);
2372 inode->i_mode = S_IFREG | S_IRUSR;
2373 inode->i_fop = &proc_fdinfo_file_operations;
2374 d_set_d_op(dentry, &tid_fd_dentry_operations);
2375 d_add(dentry, inode);
2376 /* Close the race of the process dying before we return the dentry */
2377 if (tid_fd_revalidate(dentry, NULL))
2384 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2385 struct dentry *dentry,
2386 struct nameidata *nd)
2388 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2391 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2393 return proc_readfd_common(filp, dirent, filldir,
2394 proc_fdinfo_instantiate);
2397 static const struct file_operations proc_fdinfo_operations = {
2398 .read = generic_read_dir,
2399 .readdir = proc_readfdinfo,
2400 .llseek = default_llseek,
2404 * proc directories can do almost nothing..
2406 static const struct inode_operations proc_fdinfo_inode_operations = {
2407 .lookup = proc_lookupfdinfo,
2408 .setattr = proc_setattr,
2412 static struct dentry *proc_pident_instantiate(struct inode *dir,
2413 struct dentry *dentry, struct task_struct *task, const void *ptr)
2415 const struct pid_entry *p = ptr;
2416 struct inode *inode;
2417 struct proc_inode *ei;
2418 struct dentry *error = ERR_PTR(-ENOENT);
2420 inode = proc_pid_make_inode(dir->i_sb, task);
2425 inode->i_mode = p->mode;
2426 if (S_ISDIR(inode->i_mode))
2427 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2429 inode->i_op = p->iop;
2431 inode->i_fop = p->fop;
2433 d_set_d_op(dentry, &pid_dentry_operations);
2434 d_add(dentry, inode);
2435 /* Close the race of the process dying before we return the dentry */
2436 if (pid_revalidate(dentry, NULL))
2442 static struct dentry *proc_pident_lookup(struct inode *dir,
2443 struct dentry *dentry,
2444 const struct pid_entry *ents,
2447 struct dentry *error;
2448 struct task_struct *task = get_proc_task(dir);
2449 const struct pid_entry *p, *last;
2451 error = ERR_PTR(-ENOENT);
2457 * Yes, it does not scale. And it should not. Don't add
2458 * new entries into /proc/<tgid>/ without very good reasons.
2460 last = &ents[nents - 1];
2461 for (p = ents; p <= last; p++) {
2462 if (p->len != dentry->d_name.len)
2464 if (!memcmp(dentry->d_name.name, p->name, p->len))
2470 error = proc_pident_instantiate(dir, dentry, task, p);
2472 put_task_struct(task);
2477 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2478 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2480 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2481 proc_pident_instantiate, task, p);
2484 static int proc_pident_readdir(struct file *filp,
2485 void *dirent, filldir_t filldir,
2486 const struct pid_entry *ents, unsigned int nents)
2489 struct dentry *dentry = filp->f_path.dentry;
2490 struct inode *inode = dentry->d_inode;
2491 struct task_struct *task = get_proc_task(inode);
2492 const struct pid_entry *p, *last;
2505 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2511 ino = parent_ino(dentry);
2512 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2524 last = &ents[nents - 1];
2526 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2535 put_task_struct(task);
2540 #ifdef CONFIG_SECURITY
2541 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2542 size_t count, loff_t *ppos)
2544 struct inode * inode = file->f_path.dentry->d_inode;
2547 struct task_struct *task = get_proc_task(inode);
2552 length = security_getprocattr(task,
2553 (char*)file->f_path.dentry->d_name.name,
2555 put_task_struct(task);
2557 length = simple_read_from_buffer(buf, count, ppos, p, length);
2562 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2563 size_t count, loff_t *ppos)
2565 struct inode * inode = file->f_path.dentry->d_inode;
2568 struct task_struct *task = get_proc_task(inode);
2573 if (count > PAGE_SIZE)
2576 /* No partial writes. */
2582 page = (char*)__get_free_page(GFP_TEMPORARY);
2587 if (copy_from_user(page, buf, count))
2590 /* Guard against adverse ptrace interaction */
2591 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2595 length = security_setprocattr(task,
2596 (char*)file->f_path.dentry->d_name.name,
2597 (void*)page, count);
2598 mutex_unlock(&task->signal->cred_guard_mutex);
2600 free_page((unsigned long) page);
2602 put_task_struct(task);
2607 static const struct file_operations proc_pid_attr_operations = {
2608 .read = proc_pid_attr_read,
2609 .write = proc_pid_attr_write,
2610 .llseek = generic_file_llseek,
2613 static const struct pid_entry attr_dir_stuff[] = {
2614 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2615 REG("prev", S_IRUGO, proc_pid_attr_operations),
2616 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2617 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2618 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2619 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2622 static int proc_attr_dir_readdir(struct file * filp,
2623 void * dirent, filldir_t filldir)
2625 return proc_pident_readdir(filp,dirent,filldir,
2626 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2629 static const struct file_operations proc_attr_dir_operations = {
2630 .read = generic_read_dir,
2631 .readdir = proc_attr_dir_readdir,
2632 .llseek = default_llseek,
2635 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2636 struct dentry *dentry, struct nameidata *nd)
2638 return proc_pident_lookup(dir, dentry,
2639 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2642 static const struct inode_operations proc_attr_dir_inode_operations = {
2643 .lookup = proc_attr_dir_lookup,
2644 .getattr = pid_getattr,
2645 .setattr = proc_setattr,
2650 #ifdef CONFIG_ELF_CORE
2651 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2652 size_t count, loff_t *ppos)
2654 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2655 struct mm_struct *mm;
2656 char buffer[PROC_NUMBUF];
2664 mm = get_task_mm(task);
2666 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2667 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2668 MMF_DUMP_FILTER_SHIFT));
2670 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2673 put_task_struct(task);
2678 static ssize_t proc_coredump_filter_write(struct file *file,
2679 const char __user *buf,
2683 struct task_struct *task;
2684 struct mm_struct *mm;
2685 char buffer[PROC_NUMBUF], *end;
2692 memset(buffer, 0, sizeof(buffer));
2693 if (count > sizeof(buffer) - 1)
2694 count = sizeof(buffer) - 1;
2695 if (copy_from_user(buffer, buf, count))
2699 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2702 if (end - buffer == 0)
2706 task = get_proc_task(file->f_dentry->d_inode);
2711 mm = get_task_mm(task);
2715 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2717 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2719 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2724 put_task_struct(task);
2729 static const struct file_operations proc_coredump_filter_operations = {
2730 .read = proc_coredump_filter_read,
2731 .write = proc_coredump_filter_write,
2732 .llseek = generic_file_llseek,
2739 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2742 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2743 pid_t tgid = task_tgid_nr_ns(current, ns);
2744 char tmp[PROC_NUMBUF];
2747 sprintf(tmp, "%d", tgid);
2748 return vfs_readlink(dentry,buffer,buflen,tmp);
2751 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2753 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2754 pid_t tgid = task_tgid_nr_ns(current, ns);
2755 char *name = ERR_PTR(-ENOENT);
2759 name = ERR_PTR(-ENOMEM);
2761 sprintf(name, "%d", tgid);
2763 nd_set_link(nd, name);
2767 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2770 char *s = nd_get_link(nd);
2775 static const struct inode_operations proc_self_inode_operations = {
2776 .readlink = proc_self_readlink,
2777 .follow_link = proc_self_follow_link,
2778 .put_link = proc_self_put_link,
2784 * These are the directory entries in the root directory of /proc
2785 * that properly belong to the /proc filesystem, as they describe
2786 * describe something that is process related.
2788 static const struct pid_entry proc_base_stuff[] = {
2789 NOD("self", S_IFLNK|S_IRWXUGO,
2790 &proc_self_inode_operations, NULL, {}),
2793 static struct dentry *proc_base_instantiate(struct inode *dir,
2794 struct dentry *dentry, struct task_struct *task, const void *ptr)
2796 const struct pid_entry *p = ptr;
2797 struct inode *inode;
2798 struct proc_inode *ei;
2799 struct dentry *error;
2801 /* Allocate the inode */
2802 error = ERR_PTR(-ENOMEM);
2803 inode = new_inode(dir->i_sb);
2807 /* Initialize the inode */
2809 inode->i_ino = get_next_ino();
2810 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2813 * grab the reference to the task.
2815 ei->pid = get_task_pid(task, PIDTYPE_PID);
2819 inode->i_mode = p->mode;
2820 if (S_ISDIR(inode->i_mode))
2821 set_nlink(inode, 2);
2822 if (S_ISLNK(inode->i_mode))
2825 inode->i_op = p->iop;
2827 inode->i_fop = p->fop;
2829 d_add(dentry, inode);
2838 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2840 struct dentry *error;
2841 struct task_struct *task = get_proc_task(dir);
2842 const struct pid_entry *p, *last;
2844 error = ERR_PTR(-ENOENT);
2849 /* Lookup the directory entry */
2850 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2851 for (p = proc_base_stuff; p <= last; p++) {
2852 if (p->len != dentry->d_name.len)
2854 if (!memcmp(dentry->d_name.name, p->name, p->len))
2860 error = proc_base_instantiate(dir, dentry, task, p);
2863 put_task_struct(task);
2868 static int proc_base_fill_cache(struct file *filp, void *dirent,
2869 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2871 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2872 proc_base_instantiate, task, p);
2875 #ifdef CONFIG_TASK_IO_ACCOUNTING
2876 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2878 struct task_io_accounting acct = task->ioac;
2879 unsigned long flags;
2882 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2886 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2891 if (whole && lock_task_sighand(task, &flags)) {
2892 struct task_struct *t = task;
2894 task_io_accounting_add(&acct, &task->signal->ioac);
2895 while_each_thread(task, t)
2896 task_io_accounting_add(&acct, &t->ioac);
2898 unlock_task_sighand(task, &flags);
2900 result = sprintf(buffer,
2905 "read_bytes: %llu\n"
2906 "write_bytes: %llu\n"
2907 "cancelled_write_bytes: %llu\n",
2908 (unsigned long long)acct.rchar,
2909 (unsigned long long)acct.wchar,
2910 (unsigned long long)acct.syscr,
2911 (unsigned long long)acct.syscw,
2912 (unsigned long long)acct.read_bytes,
2913 (unsigned long long)acct.write_bytes,
2914 (unsigned long long)acct.cancelled_write_bytes);
2916 mutex_unlock(&task->signal->cred_guard_mutex);
2920 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2922 return do_io_accounting(task, buffer, 0);
2925 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2927 return do_io_accounting(task, buffer, 1);
2929 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2931 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2932 struct pid *pid, struct task_struct *task)
2934 int err = lock_trace(task);
2936 seq_printf(m, "%08x\n", task->personality);
2945 static const struct file_operations proc_task_operations;
2946 static const struct inode_operations proc_task_inode_operations;
2948 static const struct pid_entry tgid_base_stuff[] = {
2949 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2950 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2951 #ifdef CONFIG_CHECKPOINT_RESTORE
2952 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
2954 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2955 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2957 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2959 REG("environ", S_IRUSR, proc_environ_operations),
2960 INF("auxv", S_IRUSR, proc_pid_auxv),
2961 ONE("status", S_IRUGO, proc_pid_status),
2962 ONE("personality", S_IRUGO, proc_pid_personality),
2963 INF("limits", S_IRUGO, proc_pid_limits),
2964 #ifdef CONFIG_SCHED_DEBUG
2965 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2967 #ifdef CONFIG_SCHED_AUTOGROUP
2968 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2970 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2971 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2972 INF("syscall", S_IRUGO, proc_pid_syscall),
2974 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2975 ONE("stat", S_IRUGO, proc_tgid_stat),
2976 ONE("statm", S_IRUGO, proc_pid_statm),
2977 REG("maps", S_IRUGO, proc_pid_maps_operations),
2979 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
2981 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2982 LNK("cwd", proc_cwd_link),
2983 LNK("root", proc_root_link),
2984 LNK("exe", proc_exe_link),
2985 REG("mounts", S_IRUGO, proc_mounts_operations),
2986 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2987 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2988 #ifdef CONFIG_PROC_PAGE_MONITOR
2989 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2990 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
2991 REG("pagemap", S_IRUGO, proc_pagemap_operations),
2993 #ifdef CONFIG_SECURITY
2994 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2996 #ifdef CONFIG_KALLSYMS
2997 INF("wchan", S_IRUGO, proc_pid_wchan),
2999 #ifdef CONFIG_STACKTRACE
3000 ONE("stack", S_IRUGO, proc_pid_stack),
3002 #ifdef CONFIG_SCHEDSTATS
3003 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3005 #ifdef CONFIG_LATENCYTOP
3006 REG("latency", S_IRUGO, proc_lstats_operations),
3008 #ifdef CONFIG_PROC_PID_CPUSET
3009 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3011 #ifdef CONFIG_CGROUPS
3012 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3014 INF("oom_score", S_IRUGO, proc_oom_score),
3015 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3016 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3017 #ifdef CONFIG_AUDITSYSCALL
3018 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3019 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3021 #ifdef CONFIG_FAULT_INJECTION
3022 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3024 #ifdef CONFIG_ELF_CORE
3025 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3027 #ifdef CONFIG_TASK_IO_ACCOUNTING
3028 INF("io", S_IRUSR, proc_tgid_io_accounting),
3030 #ifdef CONFIG_HARDWALL
3031 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3035 static int proc_tgid_base_readdir(struct file * filp,
3036 void * dirent, filldir_t filldir)
3038 return proc_pident_readdir(filp,dirent,filldir,
3039 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
3042 static const struct file_operations proc_tgid_base_operations = {
3043 .read = generic_read_dir,
3044 .readdir = proc_tgid_base_readdir,
3045 .llseek = default_llseek,
3048 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3049 return proc_pident_lookup(dir, dentry,
3050 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3053 static const struct inode_operations proc_tgid_base_inode_operations = {
3054 .lookup = proc_tgid_base_lookup,
3055 .getattr = pid_getattr,
3056 .setattr = proc_setattr,
3057 .permission = proc_pid_permission,
3060 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
3062 struct dentry *dentry, *leader, *dir;
3063 char buf[PROC_NUMBUF];
3067 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3068 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
3070 shrink_dcache_parent(dentry);
3076 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
3077 leader = d_hash_and_lookup(mnt->mnt_root, &name);
3082 name.len = strlen(name.name);
3083 dir = d_hash_and_lookup(leader, &name);
3085 goto out_put_leader;
3088 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3089 dentry = d_hash_and_lookup(dir, &name);
3091 shrink_dcache_parent(dentry);
3104 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3105 * @task: task that should be flushed.
3107 * When flushing dentries from proc, one needs to flush them from global
3108 * proc (proc_mnt) and from all the namespaces' procs this task was seen
3109 * in. This call is supposed to do all of this job.
3111 * Looks in the dcache for
3113 * /proc/@tgid/task/@pid
3114 * if either directory is present flushes it and all of it'ts children
3117 * It is safe and reasonable to cache /proc entries for a task until
3118 * that task exits. After that they just clog up the dcache with
3119 * useless entries, possibly causing useful dcache entries to be
3120 * flushed instead. This routine is proved to flush those useless
3121 * dcache entries at process exit time.
3123 * NOTE: This routine is just an optimization so it does not guarantee
3124 * that no dcache entries will exist at process exit time it
3125 * just makes it very unlikely that any will persist.
3128 void proc_flush_task(struct task_struct *task)
3131 struct pid *pid, *tgid;
3134 pid = task_pid(task);
3135 tgid = task_tgid(task);
3137 for (i = 0; i <= pid->level; i++) {
3138 upid = &pid->numbers[i];
3139 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
3140 tgid->numbers[i].nr);
3143 upid = &pid->numbers[pid->level];
3145 pid_ns_release_proc(upid->ns);
3148 static struct dentry *proc_pid_instantiate(struct inode *dir,
3149 struct dentry * dentry,
3150 struct task_struct *task, const void *ptr)
3152 struct dentry *error = ERR_PTR(-ENOENT);
3153 struct inode *inode;
3155 inode = proc_pid_make_inode(dir->i_sb, task);
3159 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3160 inode->i_op = &proc_tgid_base_inode_operations;
3161 inode->i_fop = &proc_tgid_base_operations;
3162 inode->i_flags|=S_IMMUTABLE;
3164 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
3165 ARRAY_SIZE(tgid_base_stuff)));
3167 d_set_d_op(dentry, &pid_dentry_operations);
3169 d_add(dentry, inode);
3170 /* Close the race of the process dying before we return the dentry */
3171 if (pid_revalidate(dentry, NULL))
3177 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3179 struct dentry *result;
3180 struct task_struct *task;
3182 struct pid_namespace *ns;
3184 result = proc_base_lookup(dir, dentry);
3185 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3188 tgid = name_to_int(dentry);
3192 ns = dentry->d_sb->s_fs_info;
3194 task = find_task_by_pid_ns(tgid, ns);
3196 get_task_struct(task);
3201 result = proc_pid_instantiate(dir, dentry, task, NULL);
3202 put_task_struct(task);
3208 * Find the first task with tgid >= tgid
3213 struct task_struct *task;
3215 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3220 put_task_struct(iter.task);
3224 pid = find_ge_pid(iter.tgid, ns);
3226 iter.tgid = pid_nr_ns(pid, ns);
3227 iter.task = pid_task(pid, PIDTYPE_PID);
3228 /* What we to know is if the pid we have find is the
3229 * pid of a thread_group_leader. Testing for task
3230 * being a thread_group_leader is the obvious thing
3231 * todo but there is a window when it fails, due to
3232 * the pid transfer logic in de_thread.
3234 * So we perform the straight forward test of seeing
3235 * if the pid we have found is the pid of a thread
3236 * group leader, and don't worry if the task we have
3237 * found doesn't happen to be a thread group leader.
3238 * As we don't care in the case of readdir.
3240 if (!iter.task || !has_group_leader_pid(iter.task)) {
3244 get_task_struct(iter.task);
3250 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
3252 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3253 struct tgid_iter iter)
3255 char name[PROC_NUMBUF];
3256 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
3257 return proc_fill_cache(filp, dirent, filldir, name, len,
3258 proc_pid_instantiate, iter.task, NULL);
3261 static int fake_filldir(void *buf, const char *name, int namelen,
3262 loff_t offset, u64 ino, unsigned d_type)
3267 /* for the /proc/ directory itself, after non-process stuff has been done */
3268 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3271 struct task_struct *reaper;
3272 struct tgid_iter iter;
3273 struct pid_namespace *ns;
3274 filldir_t __filldir;
3276 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3278 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3280 reaper = get_proc_task(filp->f_path.dentry->d_inode);
3284 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
3285 const struct pid_entry *p = &proc_base_stuff[nr];
3286 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
3290 ns = filp->f_dentry->d_sb->s_fs_info;
3292 iter.tgid = filp->f_pos - TGID_OFFSET;
3293 for (iter = next_tgid(ns, iter);
3295 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3296 if (has_pid_permissions(ns, iter.task, 2))
3297 __filldir = filldir;
3299 __filldir = fake_filldir;
3301 filp->f_pos = iter.tgid + TGID_OFFSET;
3302 if (proc_pid_fill_cache(filp, dirent, __filldir, iter) < 0) {
3303 put_task_struct(iter.task);
3307 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3309 put_task_struct(reaper);
3317 static const struct pid_entry tid_base_stuff[] = {
3318 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3319 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3320 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3321 REG("environ", S_IRUSR, proc_environ_operations),
3322 INF("auxv", S_IRUSR, proc_pid_auxv),
3323 ONE("status", S_IRUGO, proc_pid_status),
3324 ONE("personality", S_IRUGO, proc_pid_personality),
3325 INF("limits", S_IRUGO, proc_pid_limits),
3326 #ifdef CONFIG_SCHED_DEBUG
3327 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3329 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3330 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3331 INF("syscall", S_IRUGO, proc_pid_syscall),
3333 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3334 ONE("stat", S_IRUGO, proc_tid_stat),
3335 ONE("statm", S_IRUGO, proc_pid_statm),
3336 REG("maps", S_IRUGO, proc_tid_maps_operations),
3338 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
3340 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3341 LNK("cwd", proc_cwd_link),
3342 LNK("root", proc_root_link),
3343 LNK("exe", proc_exe_link),
3344 REG("mounts", S_IRUGO, proc_mounts_operations),
3345 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3346 #ifdef CONFIG_PROC_PAGE_MONITOR
3347 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3348 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
3349 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3351 #ifdef CONFIG_SECURITY
3352 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3354 #ifdef CONFIG_KALLSYMS
3355 INF("wchan", S_IRUGO, proc_pid_wchan),
3357 #ifdef CONFIG_STACKTRACE
3358 ONE("stack", S_IRUGO, proc_pid_stack),
3360 #ifdef CONFIG_SCHEDSTATS
3361 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3363 #ifdef CONFIG_LATENCYTOP
3364 REG("latency", S_IRUGO, proc_lstats_operations),
3366 #ifdef CONFIG_PROC_PID_CPUSET
3367 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3369 #ifdef CONFIG_CGROUPS
3370 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3372 INF("oom_score", S_IRUGO, proc_oom_score),
3373 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3374 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3375 #ifdef CONFIG_AUDITSYSCALL
3376 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3377 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3379 #ifdef CONFIG_FAULT_INJECTION
3380 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3382 #ifdef CONFIG_TASK_IO_ACCOUNTING
3383 INF("io", S_IRUSR, proc_tid_io_accounting),
3385 #ifdef CONFIG_HARDWALL
3386 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3390 static int proc_tid_base_readdir(struct file * filp,
3391 void * dirent, filldir_t filldir)
3393 return proc_pident_readdir(filp,dirent,filldir,
3394 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3397 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3398 return proc_pident_lookup(dir, dentry,
3399 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3402 static const struct file_operations proc_tid_base_operations = {
3403 .read = generic_read_dir,
3404 .readdir = proc_tid_base_readdir,
3405 .llseek = default_llseek,
3408 static const struct inode_operations proc_tid_base_inode_operations = {
3409 .lookup = proc_tid_base_lookup,
3410 .getattr = pid_getattr,
3411 .setattr = proc_setattr,
3414 static struct dentry *proc_task_instantiate(struct inode *dir,
3415 struct dentry *dentry, struct task_struct *task, const void *ptr)
3417 struct dentry *error = ERR_PTR(-ENOENT);
3418 struct inode *inode;
3419 inode = proc_pid_make_inode(dir->i_sb, task);
3423 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3424 inode->i_op = &proc_tid_base_inode_operations;
3425 inode->i_fop = &proc_tid_base_operations;
3426 inode->i_flags|=S_IMMUTABLE;
3428 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3429 ARRAY_SIZE(tid_base_stuff)));
3431 d_set_d_op(dentry, &pid_dentry_operations);
3433 d_add(dentry, inode);
3434 /* Close the race of the process dying before we return the dentry */
3435 if (pid_revalidate(dentry, NULL))
3441 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3443 struct dentry *result = ERR_PTR(-ENOENT);
3444 struct task_struct *task;
3445 struct task_struct *leader = get_proc_task(dir);
3447 struct pid_namespace *ns;
3452 tid = name_to_int(dentry);
3456 ns = dentry->d_sb->s_fs_info;
3458 task = find_task_by_pid_ns(tid, ns);
3460 get_task_struct(task);
3464 if (!same_thread_group(leader, task))
3467 result = proc_task_instantiate(dir, dentry, task, NULL);
3469 put_task_struct(task);
3471 put_task_struct(leader);
3477 * Find the first tid of a thread group to return to user space.
3479 * Usually this is just the thread group leader, but if the users
3480 * buffer was too small or there was a seek into the middle of the
3481 * directory we have more work todo.
3483 * In the case of a short read we start with find_task_by_pid.
3485 * In the case of a seek we start with the leader and walk nr
3488 static struct task_struct *first_tid(struct task_struct *leader,
3489 int tid, int nr, struct pid_namespace *ns)
3491 struct task_struct *pos;
3494 /* Attempt to start with the pid of a thread */
3495 if (tid && (nr > 0)) {
3496 pos = find_task_by_pid_ns(tid, ns);
3497 if (pos && (pos->group_leader == leader))
3501 /* If nr exceeds the number of threads there is nothing todo */
3503 if (nr && nr >= get_nr_threads(leader))
3506 /* If we haven't found our starting place yet start
3507 * with the leader and walk nr threads forward.
3509 for (pos = leader; nr > 0; --nr) {
3510 pos = next_thread(pos);
3511 if (pos == leader) {
3517 get_task_struct(pos);
3524 * Find the next thread in the thread list.
3525 * Return NULL if there is an error or no next thread.
3527 * The reference to the input task_struct is released.
3529 static struct task_struct *next_tid(struct task_struct *start)
3531 struct task_struct *pos = NULL;
3533 if (pid_alive(start)) {
3534 pos = next_thread(start);
3535 if (thread_group_leader(pos))
3538 get_task_struct(pos);
3541 put_task_struct(start);
3545 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3546 struct task_struct *task, int tid)
3548 char name[PROC_NUMBUF];
3549 int len = snprintf(name, sizeof(name), "%d", tid);
3550 return proc_fill_cache(filp, dirent, filldir, name, len,
3551 proc_task_instantiate, task, NULL);
3554 /* for the /proc/TGID/task/ directories */
3555 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3557 struct dentry *dentry = filp->f_path.dentry;
3558 struct inode *inode = dentry->d_inode;
3559 struct task_struct *leader = NULL;
3560 struct task_struct *task;
3561 int retval = -ENOENT;
3564 struct pid_namespace *ns;
3566 task = get_proc_task(inode);
3570 if (pid_alive(task)) {
3571 leader = task->group_leader;
3572 get_task_struct(leader);
3575 put_task_struct(task);
3580 switch ((unsigned long)filp->f_pos) {
3583 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3588 ino = parent_ino(dentry);
3589 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3595 /* f_version caches the tgid value that the last readdir call couldn't
3596 * return. lseek aka telldir automagically resets f_version to 0.
3598 ns = filp->f_dentry->d_sb->s_fs_info;
3599 tid = (int)filp->f_version;
3600 filp->f_version = 0;
3601 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3603 task = next_tid(task), filp->f_pos++) {
3604 tid = task_pid_nr_ns(task, ns);
3605 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3606 /* returning this tgid failed, save it as the first
3607 * pid for the next readir call */
3608 filp->f_version = (u64)tid;
3609 put_task_struct(task);
3614 put_task_struct(leader);
3619 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3621 struct inode *inode = dentry->d_inode;
3622 struct task_struct *p = get_proc_task(inode);
3623 generic_fillattr(inode, stat);
3626 stat->nlink += get_nr_threads(p);
3633 static const struct inode_operations proc_task_inode_operations = {
3634 .lookup = proc_task_lookup,
3635 .getattr = proc_task_getattr,
3636 .setattr = proc_setattr,
3637 .permission = proc_pid_permission,
3640 static const struct file_operations proc_task_operations = {
3641 .read = generic_read_dir,
3642 .readdir = proc_task_readdir,
3643 .llseek = default_llseek,