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 static struct mm_struct *__check_mem_permission(struct task_struct *task)
203 struct mm_struct *mm;
205 mm = get_task_mm(task);
207 return ERR_PTR(-EINVAL);
210 * A task can always look at itself, in case it chooses
211 * to use system calls instead of load instructions.
217 * If current is actively ptrace'ing, and would also be
218 * permitted to freshly attach with ptrace now, permit it.
220 if (task_is_stopped_or_traced(task)) {
223 match = (ptrace_parent(task) == current);
225 if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH))
230 * No one else is allowed.
233 return ERR_PTR(-EPERM);
237 * If current may access user memory in @task return a reference to the
238 * corresponding mm, otherwise ERR_PTR.
240 static struct mm_struct *check_mem_permission(struct task_struct *task)
242 struct mm_struct *mm;
246 * Avoid racing if task exec's as we might get a new mm but validate
247 * against old credentials.
249 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
253 mm = __check_mem_permission(task);
254 mutex_unlock(&task->signal->cred_guard_mutex);
259 struct mm_struct *mm_for_maps(struct task_struct *task)
261 struct mm_struct *mm;
264 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
268 mm = get_task_mm(task);
269 if (mm && mm != current->mm &&
270 !ptrace_may_access(task, PTRACE_MODE_READ)) {
272 mm = ERR_PTR(-EACCES);
274 mutex_unlock(&task->signal->cred_guard_mutex);
279 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
283 struct mm_struct *mm = get_task_mm(task);
287 goto out_mm; /* Shh! No looking before we're done */
289 len = mm->arg_end - mm->arg_start;
294 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
296 // If the nul at the end of args has been overwritten, then
297 // assume application is using setproctitle(3).
298 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
299 len = strnlen(buffer, res);
303 len = mm->env_end - mm->env_start;
304 if (len > PAGE_SIZE - res)
305 len = PAGE_SIZE - res;
306 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
307 res = strnlen(buffer, res);
316 static int proc_pid_auxv(struct task_struct *task, char *buffer)
318 struct mm_struct *mm = mm_for_maps(task);
319 int res = PTR_ERR(mm);
320 if (mm && !IS_ERR(mm)) {
321 unsigned int nwords = 0;
324 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
325 res = nwords * sizeof(mm->saved_auxv[0]);
328 memcpy(buffer, mm->saved_auxv, res);
335 #ifdef CONFIG_KALLSYMS
337 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
338 * Returns the resolved symbol. If that fails, simply return the address.
340 static int proc_pid_wchan(struct task_struct *task, char *buffer)
343 char symname[KSYM_NAME_LEN];
345 wchan = get_wchan(task);
347 if (lookup_symbol_name(wchan, symname) < 0)
348 if (!ptrace_may_access(task, PTRACE_MODE_READ))
351 return sprintf(buffer, "%lu", wchan);
353 return sprintf(buffer, "%s", symname);
355 #endif /* CONFIG_KALLSYMS */
357 static int lock_trace(struct task_struct *task)
359 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
362 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
363 mutex_unlock(&task->signal->cred_guard_mutex);
369 static void unlock_trace(struct task_struct *task)
371 mutex_unlock(&task->signal->cred_guard_mutex);
374 #ifdef CONFIG_STACKTRACE
376 #define MAX_STACK_TRACE_DEPTH 64
378 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
379 struct pid *pid, struct task_struct *task)
381 struct stack_trace trace;
382 unsigned long *entries;
386 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
390 trace.nr_entries = 0;
391 trace.max_entries = MAX_STACK_TRACE_DEPTH;
392 trace.entries = entries;
395 err = lock_trace(task);
397 save_stack_trace_tsk(task, &trace);
399 for (i = 0; i < trace.nr_entries; i++) {
400 seq_printf(m, "[<%pK>] %pS\n",
401 (void *)entries[i], (void *)entries[i]);
411 #ifdef CONFIG_SCHEDSTATS
413 * Provides /proc/PID/schedstat
415 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
417 return sprintf(buffer, "%llu %llu %lu\n",
418 (unsigned long long)task->se.sum_exec_runtime,
419 (unsigned long long)task->sched_info.run_delay,
420 task->sched_info.pcount);
424 #ifdef CONFIG_LATENCYTOP
425 static int lstats_show_proc(struct seq_file *m, void *v)
428 struct inode *inode = m->private;
429 struct task_struct *task = get_proc_task(inode);
433 seq_puts(m, "Latency Top version : v0.1\n");
434 for (i = 0; i < 32; i++) {
435 struct latency_record *lr = &task->latency_record[i];
436 if (lr->backtrace[0]) {
438 seq_printf(m, "%i %li %li",
439 lr->count, lr->time, lr->max);
440 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
441 unsigned long bt = lr->backtrace[q];
446 seq_printf(m, " %ps", (void *)bt);
452 put_task_struct(task);
456 static int lstats_open(struct inode *inode, struct file *file)
458 return single_open(file, lstats_show_proc, inode);
461 static ssize_t lstats_write(struct file *file, const char __user *buf,
462 size_t count, loff_t *offs)
464 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
468 clear_all_latency_tracing(task);
469 put_task_struct(task);
474 static const struct file_operations proc_lstats_operations = {
477 .write = lstats_write,
479 .release = single_release,
484 static int proc_oom_score(struct task_struct *task, char *buffer)
486 unsigned long points = 0;
488 read_lock(&tasklist_lock);
490 points = oom_badness(task, NULL, NULL,
491 totalram_pages + total_swap_pages);
492 read_unlock(&tasklist_lock);
493 return sprintf(buffer, "%lu\n", points);
501 static const struct limit_names lnames[RLIM_NLIMITS] = {
502 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
503 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
504 [RLIMIT_DATA] = {"Max data size", "bytes"},
505 [RLIMIT_STACK] = {"Max stack size", "bytes"},
506 [RLIMIT_CORE] = {"Max core file size", "bytes"},
507 [RLIMIT_RSS] = {"Max resident set", "bytes"},
508 [RLIMIT_NPROC] = {"Max processes", "processes"},
509 [RLIMIT_NOFILE] = {"Max open files", "files"},
510 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
511 [RLIMIT_AS] = {"Max address space", "bytes"},
512 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
513 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
514 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
515 [RLIMIT_NICE] = {"Max nice priority", NULL},
516 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
517 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
520 /* Display limits for a process */
521 static int proc_pid_limits(struct task_struct *task, char *buffer)
526 char *bufptr = buffer;
528 struct rlimit rlim[RLIM_NLIMITS];
530 if (!lock_task_sighand(task, &flags))
532 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
533 unlock_task_sighand(task, &flags);
536 * print the file header
538 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
539 "Limit", "Soft Limit", "Hard Limit", "Units");
541 for (i = 0; i < RLIM_NLIMITS; i++) {
542 if (rlim[i].rlim_cur == RLIM_INFINITY)
543 count += sprintf(&bufptr[count], "%-25s %-20s ",
544 lnames[i].name, "unlimited");
546 count += sprintf(&bufptr[count], "%-25s %-20lu ",
547 lnames[i].name, rlim[i].rlim_cur);
549 if (rlim[i].rlim_max == RLIM_INFINITY)
550 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
552 count += sprintf(&bufptr[count], "%-20lu ",
556 count += sprintf(&bufptr[count], "%-10s\n",
559 count += sprintf(&bufptr[count], "\n");
565 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
566 static int proc_pid_syscall(struct task_struct *task, char *buffer)
569 unsigned long args[6], sp, pc;
570 int res = lock_trace(task);
574 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
575 res = sprintf(buffer, "running\n");
577 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
579 res = sprintf(buffer,
580 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
582 args[0], args[1], args[2], args[3], args[4], args[5],
587 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
589 /************************************************************************/
590 /* Here the fs part begins */
591 /************************************************************************/
593 /* permission checks */
594 static int proc_fd_access_allowed(struct inode *inode)
596 struct task_struct *task;
598 /* Allow access to a task's file descriptors if it is us or we
599 * may use ptrace attach to the process and find out that
602 task = get_proc_task(inode);
604 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
605 put_task_struct(task);
610 int proc_setattr(struct dentry *dentry, struct iattr *attr)
613 struct inode *inode = dentry->d_inode;
615 if (attr->ia_valid & ATTR_MODE)
618 error = inode_change_ok(inode, attr);
622 if ((attr->ia_valid & ATTR_SIZE) &&
623 attr->ia_size != i_size_read(inode)) {
624 error = vmtruncate(inode, attr->ia_size);
629 setattr_copy(inode, attr);
630 mark_inode_dirty(inode);
635 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
636 * or euid/egid (for hide_pid_min=2)?
638 static bool has_pid_permissions(struct pid_namespace *pid,
639 struct task_struct *task,
642 if (pid->hide_pid < hide_pid_min)
644 if (in_group_p(pid->pid_gid))
646 return ptrace_may_access(task, PTRACE_MODE_READ);
650 static int proc_pid_permission(struct inode *inode, int mask)
652 struct pid_namespace *pid = inode->i_sb->s_fs_info;
653 struct task_struct *task;
656 task = get_proc_task(inode);
657 has_perms = has_pid_permissions(pid, task, 1);
658 put_task_struct(task);
661 if (pid->hide_pid == 2) {
663 * Let's make getdents(), stat(), and open()
664 * consistent with each other. If a process
665 * may not stat() a file, it shouldn't be seen
673 return generic_permission(inode, mask);
678 static const struct inode_operations proc_def_inode_operations = {
679 .setattr = proc_setattr,
682 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
684 static ssize_t proc_info_read(struct file * file, char __user * buf,
685 size_t count, loff_t *ppos)
687 struct inode * inode = file->f_path.dentry->d_inode;
690 struct task_struct *task = get_proc_task(inode);
696 if (count > PROC_BLOCK_SIZE)
697 count = PROC_BLOCK_SIZE;
700 if (!(page = __get_free_page(GFP_TEMPORARY)))
703 length = PROC_I(inode)->op.proc_read(task, (char*)page);
706 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
709 put_task_struct(task);
714 static const struct file_operations proc_info_file_operations = {
715 .read = proc_info_read,
716 .llseek = generic_file_llseek,
719 static int proc_single_show(struct seq_file *m, void *v)
721 struct inode *inode = m->private;
722 struct pid_namespace *ns;
724 struct task_struct *task;
727 ns = inode->i_sb->s_fs_info;
728 pid = proc_pid(inode);
729 task = get_pid_task(pid, PIDTYPE_PID);
733 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
735 put_task_struct(task);
739 static int proc_single_open(struct inode *inode, struct file *filp)
741 return single_open(filp, proc_single_show, inode);
744 static const struct file_operations proc_single_file_operations = {
745 .open = proc_single_open,
748 .release = single_release,
751 static int mem_open(struct inode* inode, struct file* file)
753 file->private_data = (void*)((long)current->self_exec_id);
754 /* OK to pass negative loff_t, we can catch out-of-range */
755 file->f_mode |= FMODE_UNSIGNED_OFFSET;
759 static ssize_t mem_read(struct file * file, char __user * buf,
760 size_t count, loff_t *ppos)
762 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
764 unsigned long src = *ppos;
766 struct mm_struct *mm;
772 page = (char *)__get_free_page(GFP_TEMPORARY);
776 mm = check_mem_permission(task);
783 if (file->private_data != (void*)((long)current->self_exec_id))
789 int this_len, retval;
791 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
792 retval = access_remote_vm(mm, src, page, this_len, 0);
799 if (copy_to_user(buf, page, retval)) {
814 free_page((unsigned long) page);
816 put_task_struct(task);
821 static ssize_t mem_write(struct file * file, const char __user *buf,
822 size_t count, loff_t *ppos)
826 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
827 unsigned long dst = *ppos;
828 struct mm_struct *mm;
835 page = (char *)__get_free_page(GFP_TEMPORARY);
839 mm = check_mem_permission(task);
840 copied = PTR_ERR(mm);
845 if (file->private_data != (void *)((long)current->self_exec_id))
850 int this_len, retval;
852 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
853 if (copy_from_user(page, buf, this_len)) {
857 retval = access_remote_vm(mm, dst, page, this_len, 1);
873 free_page((unsigned long) page);
875 put_task_struct(task);
880 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
884 file->f_pos = offset;
887 file->f_pos += offset;
892 force_successful_syscall_return();
896 static const struct file_operations proc_mem_operations = {
903 static ssize_t environ_read(struct file *file, char __user *buf,
904 size_t count, loff_t *ppos)
906 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
908 unsigned long src = *ppos;
910 struct mm_struct *mm;
916 page = (char *)__get_free_page(GFP_TEMPORARY);
921 mm = mm_for_maps(task);
923 if (!mm || IS_ERR(mm))
928 int this_len, retval, max_len;
930 this_len = mm->env_end - (mm->env_start + src);
935 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
936 this_len = (this_len > max_len) ? max_len : this_len;
938 retval = access_process_vm(task, (mm->env_start + src),
946 if (copy_to_user(buf, page, retval)) {
960 free_page((unsigned long) page);
962 put_task_struct(task);
967 static const struct file_operations proc_environ_operations = {
968 .read = environ_read,
969 .llseek = generic_file_llseek,
972 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
973 size_t count, loff_t *ppos)
975 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
976 char buffer[PROC_NUMBUF];
978 int oom_adjust = OOM_DISABLE;
984 if (lock_task_sighand(task, &flags)) {
985 oom_adjust = task->signal->oom_adj;
986 unlock_task_sighand(task, &flags);
989 put_task_struct(task);
991 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
993 return simple_read_from_buffer(buf, count, ppos, buffer, len);
996 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
997 size_t count, loff_t *ppos)
999 struct task_struct *task;
1000 char buffer[PROC_NUMBUF];
1002 unsigned long flags;
1005 memset(buffer, 0, sizeof(buffer));
1006 if (count > sizeof(buffer) - 1)
1007 count = sizeof(buffer) - 1;
1008 if (copy_from_user(buffer, buf, count)) {
1013 err = kstrtoint(strstrip(buffer), 0, &oom_adjust);
1016 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
1017 oom_adjust != OOM_DISABLE) {
1022 task = get_proc_task(file->f_path.dentry->d_inode);
1034 if (!lock_task_sighand(task, &flags)) {
1039 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
1045 * Warn that /proc/pid/oom_adj is deprecated, see
1046 * Documentation/feature-removal-schedule.txt.
1048 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1049 current->comm, task_pid_nr(current), task_pid_nr(task),
1051 task->signal->oom_adj = oom_adjust;
1053 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1054 * value is always attainable.
1056 if (task->signal->oom_adj == OOM_ADJUST_MAX)
1057 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
1059 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
1061 trace_oom_score_adj_update(task);
1063 unlock_task_sighand(task, &flags);
1066 put_task_struct(task);
1068 return err < 0 ? err : count;
1071 static const struct file_operations proc_oom_adjust_operations = {
1072 .read = oom_adjust_read,
1073 .write = oom_adjust_write,
1074 .llseek = generic_file_llseek,
1077 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1078 size_t count, loff_t *ppos)
1080 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1081 char buffer[PROC_NUMBUF];
1082 int oom_score_adj = OOM_SCORE_ADJ_MIN;
1083 unsigned long flags;
1088 if (lock_task_sighand(task, &flags)) {
1089 oom_score_adj = task->signal->oom_score_adj;
1090 unlock_task_sighand(task, &flags);
1092 put_task_struct(task);
1093 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1094 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1097 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1098 size_t count, loff_t *ppos)
1100 struct task_struct *task;
1101 char buffer[PROC_NUMBUF];
1102 unsigned long flags;
1106 memset(buffer, 0, sizeof(buffer));
1107 if (count > sizeof(buffer) - 1)
1108 count = sizeof(buffer) - 1;
1109 if (copy_from_user(buffer, buf, count)) {
1114 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1117 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1118 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1123 task = get_proc_task(file->f_path.dentry->d_inode);
1135 if (!lock_task_sighand(task, &flags)) {
1140 if (oom_score_adj < task->signal->oom_score_adj_min &&
1141 !capable(CAP_SYS_RESOURCE)) {
1146 task->signal->oom_score_adj = oom_score_adj;
1147 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1148 task->signal->oom_score_adj_min = oom_score_adj;
1149 trace_oom_score_adj_update(task);
1151 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1152 * always attainable.
1154 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1155 task->signal->oom_adj = OOM_DISABLE;
1157 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1160 unlock_task_sighand(task, &flags);
1163 put_task_struct(task);
1165 return err < 0 ? err : count;
1168 static const struct file_operations proc_oom_score_adj_operations = {
1169 .read = oom_score_adj_read,
1170 .write = oom_score_adj_write,
1171 .llseek = default_llseek,
1174 #ifdef CONFIG_AUDITSYSCALL
1175 #define TMPBUFLEN 21
1176 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1177 size_t count, loff_t *ppos)
1179 struct inode * inode = file->f_path.dentry->d_inode;
1180 struct task_struct *task = get_proc_task(inode);
1182 char tmpbuf[TMPBUFLEN];
1186 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1187 audit_get_loginuid(task));
1188 put_task_struct(task);
1189 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1192 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1193 size_t count, loff_t *ppos)
1195 struct inode * inode = file->f_path.dentry->d_inode;
1201 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1207 if (count >= PAGE_SIZE)
1208 count = PAGE_SIZE - 1;
1211 /* No partial writes. */
1214 page = (char*)__get_free_page(GFP_TEMPORARY);
1218 if (copy_from_user(page, buf, count))
1222 loginuid = simple_strtoul(page, &tmp, 10);
1228 length = audit_set_loginuid(loginuid);
1229 if (likely(length == 0))
1233 free_page((unsigned long) page);
1237 static const struct file_operations proc_loginuid_operations = {
1238 .read = proc_loginuid_read,
1239 .write = proc_loginuid_write,
1240 .llseek = generic_file_llseek,
1243 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1244 size_t count, loff_t *ppos)
1246 struct inode * inode = file->f_path.dentry->d_inode;
1247 struct task_struct *task = get_proc_task(inode);
1249 char tmpbuf[TMPBUFLEN];
1253 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1254 audit_get_sessionid(task));
1255 put_task_struct(task);
1256 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1259 static const struct file_operations proc_sessionid_operations = {
1260 .read = proc_sessionid_read,
1261 .llseek = generic_file_llseek,
1265 #ifdef CONFIG_FAULT_INJECTION
1266 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1267 size_t count, loff_t *ppos)
1269 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1270 char buffer[PROC_NUMBUF];
1276 make_it_fail = task->make_it_fail;
1277 put_task_struct(task);
1279 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1281 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1284 static ssize_t proc_fault_inject_write(struct file * file,
1285 const char __user * buf, size_t count, loff_t *ppos)
1287 struct task_struct *task;
1288 char buffer[PROC_NUMBUF], *end;
1291 if (!capable(CAP_SYS_RESOURCE))
1293 memset(buffer, 0, sizeof(buffer));
1294 if (count > sizeof(buffer) - 1)
1295 count = sizeof(buffer) - 1;
1296 if (copy_from_user(buffer, buf, count))
1298 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1301 task = get_proc_task(file->f_dentry->d_inode);
1304 task->make_it_fail = make_it_fail;
1305 put_task_struct(task);
1310 static const struct file_operations proc_fault_inject_operations = {
1311 .read = proc_fault_inject_read,
1312 .write = proc_fault_inject_write,
1313 .llseek = generic_file_llseek,
1318 #ifdef CONFIG_SCHED_DEBUG
1320 * Print out various scheduling related per-task fields:
1322 static int sched_show(struct seq_file *m, void *v)
1324 struct inode *inode = m->private;
1325 struct task_struct *p;
1327 p = get_proc_task(inode);
1330 proc_sched_show_task(p, m);
1338 sched_write(struct file *file, const char __user *buf,
1339 size_t count, loff_t *offset)
1341 struct inode *inode = file->f_path.dentry->d_inode;
1342 struct task_struct *p;
1344 p = get_proc_task(inode);
1347 proc_sched_set_task(p);
1354 static int sched_open(struct inode *inode, struct file *filp)
1356 return single_open(filp, sched_show, inode);
1359 static const struct file_operations proc_pid_sched_operations = {
1362 .write = sched_write,
1363 .llseek = seq_lseek,
1364 .release = single_release,
1369 #ifdef CONFIG_SCHED_AUTOGROUP
1371 * Print out autogroup related information:
1373 static int sched_autogroup_show(struct seq_file *m, void *v)
1375 struct inode *inode = m->private;
1376 struct task_struct *p;
1378 p = get_proc_task(inode);
1381 proc_sched_autogroup_show_task(p, m);
1389 sched_autogroup_write(struct file *file, const char __user *buf,
1390 size_t count, loff_t *offset)
1392 struct inode *inode = file->f_path.dentry->d_inode;
1393 struct task_struct *p;
1394 char buffer[PROC_NUMBUF];
1398 memset(buffer, 0, sizeof(buffer));
1399 if (count > sizeof(buffer) - 1)
1400 count = sizeof(buffer) - 1;
1401 if (copy_from_user(buffer, buf, count))
1404 err = kstrtoint(strstrip(buffer), 0, &nice);
1408 p = get_proc_task(inode);
1413 err = proc_sched_autogroup_set_nice(p, &err);
1422 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1426 ret = single_open(filp, sched_autogroup_show, NULL);
1428 struct seq_file *m = filp->private_data;
1435 static const struct file_operations proc_pid_sched_autogroup_operations = {
1436 .open = sched_autogroup_open,
1438 .write = sched_autogroup_write,
1439 .llseek = seq_lseek,
1440 .release = single_release,
1443 #endif /* CONFIG_SCHED_AUTOGROUP */
1445 static ssize_t comm_write(struct file *file, const char __user *buf,
1446 size_t count, loff_t *offset)
1448 struct inode *inode = file->f_path.dentry->d_inode;
1449 struct task_struct *p;
1450 char buffer[TASK_COMM_LEN];
1452 memset(buffer, 0, sizeof(buffer));
1453 if (count > sizeof(buffer) - 1)
1454 count = sizeof(buffer) - 1;
1455 if (copy_from_user(buffer, buf, count))
1458 p = get_proc_task(inode);
1462 if (same_thread_group(current, p))
1463 set_task_comm(p, buffer);
1472 static int comm_show(struct seq_file *m, void *v)
1474 struct inode *inode = m->private;
1475 struct task_struct *p;
1477 p = get_proc_task(inode);
1482 seq_printf(m, "%s\n", p->comm);
1490 static int comm_open(struct inode *inode, struct file *filp)
1492 return single_open(filp, comm_show, inode);
1495 static const struct file_operations proc_pid_set_comm_operations = {
1498 .write = comm_write,
1499 .llseek = seq_lseek,
1500 .release = single_release,
1503 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1505 struct task_struct *task;
1506 struct mm_struct *mm;
1507 struct file *exe_file;
1509 task = get_proc_task(dentry->d_inode);
1512 mm = get_task_mm(task);
1513 put_task_struct(task);
1516 exe_file = get_mm_exe_file(mm);
1519 *exe_path = exe_file->f_path;
1520 path_get(&exe_file->f_path);
1527 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1529 struct inode *inode = dentry->d_inode;
1530 int error = -EACCES;
1532 /* We don't need a base pointer in the /proc filesystem */
1533 path_put(&nd->path);
1535 /* Are we allowed to snoop on the tasks file descriptors? */
1536 if (!proc_fd_access_allowed(inode))
1539 error = PROC_I(inode)->op.proc_get_link(dentry, &nd->path);
1541 return ERR_PTR(error);
1544 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1546 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1553 pathname = d_path(path, tmp, PAGE_SIZE);
1554 len = PTR_ERR(pathname);
1555 if (IS_ERR(pathname))
1557 len = tmp + PAGE_SIZE - 1 - pathname;
1561 if (copy_to_user(buffer, pathname, len))
1564 free_page((unsigned long)tmp);
1568 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1570 int error = -EACCES;
1571 struct inode *inode = dentry->d_inode;
1574 /* Are we allowed to snoop on the tasks file descriptors? */
1575 if (!proc_fd_access_allowed(inode))
1578 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1582 error = do_proc_readlink(&path, buffer, buflen);
1588 static const struct inode_operations proc_pid_link_inode_operations = {
1589 .readlink = proc_pid_readlink,
1590 .follow_link = proc_pid_follow_link,
1591 .setattr = proc_setattr,
1595 /* building an inode */
1597 static int task_dumpable(struct task_struct *task)
1600 struct mm_struct *mm;
1605 dumpable = get_dumpable(mm);
1612 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1614 struct inode * inode;
1615 struct proc_inode *ei;
1616 const struct cred *cred;
1618 /* We need a new inode */
1620 inode = new_inode(sb);
1626 inode->i_ino = get_next_ino();
1627 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1628 inode->i_op = &proc_def_inode_operations;
1631 * grab the reference to task.
1633 ei->pid = get_task_pid(task, PIDTYPE_PID);
1637 if (task_dumpable(task)) {
1639 cred = __task_cred(task);
1640 inode->i_uid = cred->euid;
1641 inode->i_gid = cred->egid;
1644 security_task_to_inode(task, inode);
1654 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1656 struct inode *inode = dentry->d_inode;
1657 struct task_struct *task;
1658 const struct cred *cred;
1659 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
1661 generic_fillattr(inode, stat);
1666 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1668 if (!has_pid_permissions(pid, task, 2)) {
1671 * This doesn't prevent learning whether PID exists,
1672 * it only makes getattr() consistent with readdir().
1676 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1677 task_dumpable(task)) {
1678 cred = __task_cred(task);
1679 stat->uid = cred->euid;
1680 stat->gid = cred->egid;
1690 * Exceptional case: normally we are not allowed to unhash a busy
1691 * directory. In this case, however, we can do it - no aliasing problems
1692 * due to the way we treat inodes.
1694 * Rewrite the inode's ownerships here because the owning task may have
1695 * performed a setuid(), etc.
1697 * Before the /proc/pid/status file was created the only way to read
1698 * the effective uid of a /process was to stat /proc/pid. Reading
1699 * /proc/pid/status is slow enough that procps and other packages
1700 * kept stating /proc/pid. To keep the rules in /proc simple I have
1701 * made this apply to all per process world readable and executable
1704 int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1706 struct inode *inode;
1707 struct task_struct *task;
1708 const struct cred *cred;
1710 if (nd && nd->flags & LOOKUP_RCU)
1713 inode = dentry->d_inode;
1714 task = get_proc_task(inode);
1717 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1718 task_dumpable(task)) {
1720 cred = __task_cred(task);
1721 inode->i_uid = cred->euid;
1722 inode->i_gid = cred->egid;
1728 inode->i_mode &= ~(S_ISUID | S_ISGID);
1729 security_task_to_inode(task, inode);
1730 put_task_struct(task);
1737 static int pid_delete_dentry(const struct dentry * dentry)
1739 /* Is the task we represent dead?
1740 * If so, then don't put the dentry on the lru list,
1741 * kill it immediately.
1743 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1746 const struct dentry_operations pid_dentry_operations =
1748 .d_revalidate = pid_revalidate,
1749 .d_delete = pid_delete_dentry,
1755 * Fill a directory entry.
1757 * If possible create the dcache entry and derive our inode number and
1758 * file type from dcache entry.
1760 * Since all of the proc inode numbers are dynamically generated, the inode
1761 * numbers do not exist until the inode is cache. This means creating the
1762 * the dcache entry in readdir is necessary to keep the inode numbers
1763 * reported by readdir in sync with the inode numbers reported
1766 int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1767 const char *name, int len,
1768 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1770 struct dentry *child, *dir = filp->f_path.dentry;
1771 struct inode *inode;
1774 unsigned type = DT_UNKNOWN;
1778 qname.hash = full_name_hash(name, len);
1780 child = d_lookup(dir, &qname);
1783 new = d_alloc(dir, &qname);
1785 child = instantiate(dir->d_inode, new, task, ptr);
1792 if (!child || IS_ERR(child) || !child->d_inode)
1793 goto end_instantiate;
1794 inode = child->d_inode;
1797 type = inode->i_mode >> 12;
1802 ino = find_inode_number(dir, &qname);
1805 return filldir(dirent, name, len, filp->f_pos, ino, type);
1808 static unsigned name_to_int(struct dentry *dentry)
1810 const char *name = dentry->d_name.name;
1811 int len = dentry->d_name.len;
1814 if (len > 1 && *name == '0')
1817 unsigned c = *name++ - '0';
1820 if (n >= (~0U-9)/10)
1830 #define PROC_FDINFO_MAX 64
1832 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1834 struct task_struct *task = get_proc_task(inode);
1835 struct files_struct *files = NULL;
1837 int fd = proc_fd(inode);
1840 files = get_files_struct(task);
1841 put_task_struct(task);
1845 * We are not taking a ref to the file structure, so we must
1848 spin_lock(&files->file_lock);
1849 file = fcheck_files(files, fd);
1851 unsigned int f_flags;
1852 struct fdtable *fdt;
1854 fdt = files_fdtable(files);
1855 f_flags = file->f_flags & ~O_CLOEXEC;
1856 if (FD_ISSET(fd, fdt->close_on_exec))
1857 f_flags |= O_CLOEXEC;
1860 *path = file->f_path;
1861 path_get(&file->f_path);
1864 snprintf(info, PROC_FDINFO_MAX,
1867 (long long) file->f_pos,
1869 spin_unlock(&files->file_lock);
1870 put_files_struct(files);
1873 spin_unlock(&files->file_lock);
1874 put_files_struct(files);
1879 static int proc_fd_link(struct dentry *dentry, struct path *path)
1881 return proc_fd_info(dentry->d_inode, path, NULL);
1884 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1886 struct inode *inode;
1887 struct task_struct *task;
1889 struct files_struct *files;
1890 const struct cred *cred;
1892 if (nd && nd->flags & LOOKUP_RCU)
1895 inode = dentry->d_inode;
1896 task = get_proc_task(inode);
1897 fd = proc_fd(inode);
1900 files = get_files_struct(task);
1903 if (fcheck_files(files, fd)) {
1905 put_files_struct(files);
1906 if (task_dumpable(task)) {
1908 cred = __task_cred(task);
1909 inode->i_uid = cred->euid;
1910 inode->i_gid = cred->egid;
1916 inode->i_mode &= ~(S_ISUID | S_ISGID);
1917 security_task_to_inode(task, inode);
1918 put_task_struct(task);
1922 put_files_struct(files);
1924 put_task_struct(task);
1930 static const struct dentry_operations tid_fd_dentry_operations =
1932 .d_revalidate = tid_fd_revalidate,
1933 .d_delete = pid_delete_dentry,
1936 static struct dentry *proc_fd_instantiate(struct inode *dir,
1937 struct dentry *dentry, struct task_struct *task, const void *ptr)
1939 unsigned fd = *(const unsigned *)ptr;
1941 struct files_struct *files;
1942 struct inode *inode;
1943 struct proc_inode *ei;
1944 struct dentry *error = ERR_PTR(-ENOENT);
1946 inode = proc_pid_make_inode(dir->i_sb, task);
1951 files = get_files_struct(task);
1954 inode->i_mode = S_IFLNK;
1957 * We are not taking a ref to the file structure, so we must
1960 spin_lock(&files->file_lock);
1961 file = fcheck_files(files, fd);
1964 if (file->f_mode & FMODE_READ)
1965 inode->i_mode |= S_IRUSR | S_IXUSR;
1966 if (file->f_mode & FMODE_WRITE)
1967 inode->i_mode |= S_IWUSR | S_IXUSR;
1968 spin_unlock(&files->file_lock);
1969 put_files_struct(files);
1971 inode->i_op = &proc_pid_link_inode_operations;
1973 ei->op.proc_get_link = proc_fd_link;
1974 d_set_d_op(dentry, &tid_fd_dentry_operations);
1975 d_add(dentry, inode);
1976 /* Close the race of the process dying before we return the dentry */
1977 if (tid_fd_revalidate(dentry, NULL))
1983 spin_unlock(&files->file_lock);
1984 put_files_struct(files);
1990 static struct dentry *proc_lookupfd_common(struct inode *dir,
1991 struct dentry *dentry,
1992 instantiate_t instantiate)
1994 struct task_struct *task = get_proc_task(dir);
1995 unsigned fd = name_to_int(dentry);
1996 struct dentry *result = ERR_PTR(-ENOENT);
2003 result = instantiate(dir, dentry, task, &fd);
2005 put_task_struct(task);
2010 static int proc_readfd_common(struct file * filp, void * dirent,
2011 filldir_t filldir, instantiate_t instantiate)
2013 struct dentry *dentry = filp->f_path.dentry;
2014 struct inode *inode = dentry->d_inode;
2015 struct task_struct *p = get_proc_task(inode);
2016 unsigned int fd, ino;
2018 struct files_struct * files;
2028 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
2032 ino = parent_ino(dentry);
2033 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2037 files = get_files_struct(p);
2041 for (fd = filp->f_pos-2;
2042 fd < files_fdtable(files)->max_fds;
2043 fd++, filp->f_pos++) {
2044 char name[PROC_NUMBUF];
2047 if (!fcheck_files(files, fd))
2051 len = snprintf(name, sizeof(name), "%d", fd);
2052 if (proc_fill_cache(filp, dirent, filldir,
2053 name, len, instantiate,
2061 put_files_struct(files);
2069 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
2070 struct nameidata *nd)
2072 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
2075 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
2077 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
2080 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
2081 size_t len, loff_t *ppos)
2083 char tmp[PROC_FDINFO_MAX];
2084 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
2086 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
2090 static const struct file_operations proc_fdinfo_file_operations = {
2091 .open = nonseekable_open,
2092 .read = proc_fdinfo_read,
2093 .llseek = no_llseek,
2096 static const struct file_operations proc_fd_operations = {
2097 .read = generic_read_dir,
2098 .readdir = proc_readfd,
2099 .llseek = default_llseek,
2102 #ifdef CONFIG_CHECKPOINT_RESTORE
2105 * dname_to_vma_addr - maps a dentry name into two unsigned longs
2106 * which represent vma start and end addresses.
2108 static int dname_to_vma_addr(struct dentry *dentry,
2109 unsigned long *start, unsigned long *end)
2111 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
2117 static int map_files_d_revalidate(struct dentry *dentry, struct nameidata *nd)
2119 unsigned long vm_start, vm_end;
2120 bool exact_vma_exists = false;
2121 struct mm_struct *mm = NULL;
2122 struct task_struct *task;
2123 const struct cred *cred;
2124 struct inode *inode;
2127 if (nd && nd->flags & LOOKUP_RCU)
2130 if (!capable(CAP_SYS_ADMIN)) {
2135 inode = dentry->d_inode;
2136 task = get_proc_task(inode);
2140 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2143 mm = get_task_mm(task);
2147 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
2148 down_read(&mm->mmap_sem);
2149 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
2150 up_read(&mm->mmap_sem);
2155 if (exact_vma_exists) {
2156 if (task_dumpable(task)) {
2158 cred = __task_cred(task);
2159 inode->i_uid = cred->euid;
2160 inode->i_gid = cred->egid;
2166 security_task_to_inode(task, inode);
2171 put_task_struct(task);
2180 static const struct dentry_operations tid_map_files_dentry_operations = {
2181 .d_revalidate = map_files_d_revalidate,
2182 .d_delete = pid_delete_dentry,
2185 static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
2187 unsigned long vm_start, vm_end;
2188 struct vm_area_struct *vma;
2189 struct task_struct *task;
2190 struct mm_struct *mm;
2194 task = get_proc_task(dentry->d_inode);
2198 mm = get_task_mm(task);
2199 put_task_struct(task);
2203 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2207 down_read(&mm->mmap_sem);
2208 vma = find_exact_vma(mm, vm_start, vm_end);
2209 if (vma && vma->vm_file) {
2210 *path = vma->vm_file->f_path;
2214 up_read(&mm->mmap_sem);
2222 struct map_files_info {
2225 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
2228 static struct dentry *
2229 proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
2230 struct task_struct *task, const void *ptr)
2232 const struct file *file = ptr;
2233 struct proc_inode *ei;
2234 struct inode *inode;
2237 return ERR_PTR(-ENOENT);
2239 inode = proc_pid_make_inode(dir->i_sb, task);
2241 return ERR_PTR(-ENOENT);
2244 ei->op.proc_get_link = proc_map_files_get_link;
2246 inode->i_op = &proc_pid_link_inode_operations;
2248 inode->i_mode = S_IFLNK;
2250 if (file->f_mode & FMODE_READ)
2251 inode->i_mode |= S_IRUSR;
2252 if (file->f_mode & FMODE_WRITE)
2253 inode->i_mode |= S_IWUSR;
2255 d_set_d_op(dentry, &tid_map_files_dentry_operations);
2256 d_add(dentry, inode);
2261 static struct dentry *proc_map_files_lookup(struct inode *dir,
2262 struct dentry *dentry, struct nameidata *nd)
2264 unsigned long vm_start, vm_end;
2265 struct vm_area_struct *vma;
2266 struct task_struct *task;
2267 struct dentry *result;
2268 struct mm_struct *mm;
2270 result = ERR_PTR(-EACCES);
2271 if (!capable(CAP_SYS_ADMIN))
2274 result = ERR_PTR(-ENOENT);
2275 task = get_proc_task(dir);
2279 result = ERR_PTR(-EACCES);
2280 if (lock_trace(task))
2283 result = ERR_PTR(-ENOENT);
2284 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2287 mm = get_task_mm(task);
2291 down_read(&mm->mmap_sem);
2292 vma = find_exact_vma(mm, vm_start, vm_end);
2296 result = proc_map_files_instantiate(dir, dentry, task, vma->vm_file);
2299 up_read(&mm->mmap_sem);
2304 put_task_struct(task);
2309 static const struct inode_operations proc_map_files_inode_operations = {
2310 .lookup = proc_map_files_lookup,
2311 .permission = proc_fd_permission,
2312 .setattr = proc_setattr,
2316 proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
2318 struct dentry *dentry = filp->f_path.dentry;
2319 struct inode *inode = dentry->d_inode;
2320 struct vm_area_struct *vma;
2321 struct task_struct *task;
2322 struct mm_struct *mm;
2327 if (!capable(CAP_SYS_ADMIN))
2331 task = get_proc_task(inode);
2336 if (lock_trace(task))
2340 switch (filp->f_pos) {
2343 if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0)
2347 ino = parent_ino(dentry);
2348 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2353 unsigned long nr_files, pos, i;
2354 struct flex_array *fa = NULL;
2355 struct map_files_info info;
2356 struct map_files_info *p;
2358 mm = get_task_mm(task);
2361 down_read(&mm->mmap_sem);
2366 * We need two passes here:
2368 * 1) Collect vmas of mapped files with mmap_sem taken
2369 * 2) Release mmap_sem and instantiate entries
2371 * otherwise we get lockdep complained, since filldir()
2372 * routine might require mmap_sem taken in might_fault().
2375 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2376 if (vma->vm_file && ++pos > filp->f_pos)
2381 fa = flex_array_alloc(sizeof(info), nr_files,
2383 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2387 flex_array_free(fa);
2388 up_read(&mm->mmap_sem);
2392 for (i = 0, vma = mm->mmap, pos = 2; vma;
2393 vma = vma->vm_next) {
2396 if (++pos <= filp->f_pos)
2399 get_file(vma->vm_file);
2400 info.file = vma->vm_file;
2401 info.len = snprintf(info.name,
2402 sizeof(info.name), "%lx-%lx",
2403 vma->vm_start, vma->vm_end);
2404 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2408 up_read(&mm->mmap_sem);
2410 for (i = 0; i < nr_files; i++) {
2411 p = flex_array_get(fa, i);
2412 ret = proc_fill_cache(filp, dirent, filldir,
2414 proc_map_files_instantiate,
2421 for (; i < nr_files; i++) {
2423 * In case of error don't forget
2424 * to put rest of file refs.
2426 p = flex_array_get(fa, i);
2430 flex_array_free(fa);
2438 put_task_struct(task);
2443 static const struct file_operations proc_map_files_operations = {
2444 .read = generic_read_dir,
2445 .readdir = proc_map_files_readdir,
2446 .llseek = default_llseek,
2449 #endif /* CONFIG_CHECKPOINT_RESTORE */
2452 * /proc/pid/fd needs a special permission handler so that a process can still
2453 * access /proc/self/fd after it has executed a setuid().
2455 static int proc_fd_permission(struct inode *inode, int mask)
2457 int rv = generic_permission(inode, mask);
2460 if (task_pid(current) == proc_pid(inode))
2466 * proc directories can do almost nothing..
2468 static const struct inode_operations proc_fd_inode_operations = {
2469 .lookup = proc_lookupfd,
2470 .permission = proc_fd_permission,
2471 .setattr = proc_setattr,
2474 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2475 struct dentry *dentry, struct task_struct *task, const void *ptr)
2477 unsigned fd = *(unsigned *)ptr;
2478 struct inode *inode;
2479 struct proc_inode *ei;
2480 struct dentry *error = ERR_PTR(-ENOENT);
2482 inode = proc_pid_make_inode(dir->i_sb, task);
2487 inode->i_mode = S_IFREG | S_IRUSR;
2488 inode->i_fop = &proc_fdinfo_file_operations;
2489 d_set_d_op(dentry, &tid_fd_dentry_operations);
2490 d_add(dentry, inode);
2491 /* Close the race of the process dying before we return the dentry */
2492 if (tid_fd_revalidate(dentry, NULL))
2499 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2500 struct dentry *dentry,
2501 struct nameidata *nd)
2503 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2506 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2508 return proc_readfd_common(filp, dirent, filldir,
2509 proc_fdinfo_instantiate);
2512 static const struct file_operations proc_fdinfo_operations = {
2513 .read = generic_read_dir,
2514 .readdir = proc_readfdinfo,
2515 .llseek = default_llseek,
2519 * proc directories can do almost nothing..
2521 static const struct inode_operations proc_fdinfo_inode_operations = {
2522 .lookup = proc_lookupfdinfo,
2523 .setattr = proc_setattr,
2527 static struct dentry *proc_pident_instantiate(struct inode *dir,
2528 struct dentry *dentry, struct task_struct *task, const void *ptr)
2530 const struct pid_entry *p = ptr;
2531 struct inode *inode;
2532 struct proc_inode *ei;
2533 struct dentry *error = ERR_PTR(-ENOENT);
2535 inode = proc_pid_make_inode(dir->i_sb, task);
2540 inode->i_mode = p->mode;
2541 if (S_ISDIR(inode->i_mode))
2542 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2544 inode->i_op = p->iop;
2546 inode->i_fop = p->fop;
2548 d_set_d_op(dentry, &pid_dentry_operations);
2549 d_add(dentry, inode);
2550 /* Close the race of the process dying before we return the dentry */
2551 if (pid_revalidate(dentry, NULL))
2557 static struct dentry *proc_pident_lookup(struct inode *dir,
2558 struct dentry *dentry,
2559 const struct pid_entry *ents,
2562 struct dentry *error;
2563 struct task_struct *task = get_proc_task(dir);
2564 const struct pid_entry *p, *last;
2566 error = ERR_PTR(-ENOENT);
2572 * Yes, it does not scale. And it should not. Don't add
2573 * new entries into /proc/<tgid>/ without very good reasons.
2575 last = &ents[nents - 1];
2576 for (p = ents; p <= last; p++) {
2577 if (p->len != dentry->d_name.len)
2579 if (!memcmp(dentry->d_name.name, p->name, p->len))
2585 error = proc_pident_instantiate(dir, dentry, task, p);
2587 put_task_struct(task);
2592 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2593 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2595 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2596 proc_pident_instantiate, task, p);
2599 static int proc_pident_readdir(struct file *filp,
2600 void *dirent, filldir_t filldir,
2601 const struct pid_entry *ents, unsigned int nents)
2604 struct dentry *dentry = filp->f_path.dentry;
2605 struct inode *inode = dentry->d_inode;
2606 struct task_struct *task = get_proc_task(inode);
2607 const struct pid_entry *p, *last;
2620 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2626 ino = parent_ino(dentry);
2627 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2639 last = &ents[nents - 1];
2641 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2650 put_task_struct(task);
2655 #ifdef CONFIG_SECURITY
2656 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2657 size_t count, loff_t *ppos)
2659 struct inode * inode = file->f_path.dentry->d_inode;
2662 struct task_struct *task = get_proc_task(inode);
2667 length = security_getprocattr(task,
2668 (char*)file->f_path.dentry->d_name.name,
2670 put_task_struct(task);
2672 length = simple_read_from_buffer(buf, count, ppos, p, length);
2677 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2678 size_t count, loff_t *ppos)
2680 struct inode * inode = file->f_path.dentry->d_inode;
2683 struct task_struct *task = get_proc_task(inode);
2688 if (count > PAGE_SIZE)
2691 /* No partial writes. */
2697 page = (char*)__get_free_page(GFP_TEMPORARY);
2702 if (copy_from_user(page, buf, count))
2705 /* Guard against adverse ptrace interaction */
2706 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2710 length = security_setprocattr(task,
2711 (char*)file->f_path.dentry->d_name.name,
2712 (void*)page, count);
2713 mutex_unlock(&task->signal->cred_guard_mutex);
2715 free_page((unsigned long) page);
2717 put_task_struct(task);
2722 static const struct file_operations proc_pid_attr_operations = {
2723 .read = proc_pid_attr_read,
2724 .write = proc_pid_attr_write,
2725 .llseek = generic_file_llseek,
2728 static const struct pid_entry attr_dir_stuff[] = {
2729 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2730 REG("prev", S_IRUGO, proc_pid_attr_operations),
2731 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2732 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2733 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2734 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2737 static int proc_attr_dir_readdir(struct file * filp,
2738 void * dirent, filldir_t filldir)
2740 return proc_pident_readdir(filp,dirent,filldir,
2741 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2744 static const struct file_operations proc_attr_dir_operations = {
2745 .read = generic_read_dir,
2746 .readdir = proc_attr_dir_readdir,
2747 .llseek = default_llseek,
2750 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2751 struct dentry *dentry, struct nameidata *nd)
2753 return proc_pident_lookup(dir, dentry,
2754 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2757 static const struct inode_operations proc_attr_dir_inode_operations = {
2758 .lookup = proc_attr_dir_lookup,
2759 .getattr = pid_getattr,
2760 .setattr = proc_setattr,
2765 #ifdef CONFIG_ELF_CORE
2766 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2767 size_t count, loff_t *ppos)
2769 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2770 struct mm_struct *mm;
2771 char buffer[PROC_NUMBUF];
2779 mm = get_task_mm(task);
2781 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2782 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2783 MMF_DUMP_FILTER_SHIFT));
2785 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2788 put_task_struct(task);
2793 static ssize_t proc_coredump_filter_write(struct file *file,
2794 const char __user *buf,
2798 struct task_struct *task;
2799 struct mm_struct *mm;
2800 char buffer[PROC_NUMBUF], *end;
2807 memset(buffer, 0, sizeof(buffer));
2808 if (count > sizeof(buffer) - 1)
2809 count = sizeof(buffer) - 1;
2810 if (copy_from_user(buffer, buf, count))
2814 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2817 if (end - buffer == 0)
2821 task = get_proc_task(file->f_dentry->d_inode);
2826 mm = get_task_mm(task);
2830 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2832 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2834 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2839 put_task_struct(task);
2844 static const struct file_operations proc_coredump_filter_operations = {
2845 .read = proc_coredump_filter_read,
2846 .write = proc_coredump_filter_write,
2847 .llseek = generic_file_llseek,
2854 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2857 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2858 pid_t tgid = task_tgid_nr_ns(current, ns);
2859 char tmp[PROC_NUMBUF];
2862 sprintf(tmp, "%d", tgid);
2863 return vfs_readlink(dentry,buffer,buflen,tmp);
2866 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2868 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2869 pid_t tgid = task_tgid_nr_ns(current, ns);
2870 char *name = ERR_PTR(-ENOENT);
2874 name = ERR_PTR(-ENOMEM);
2876 sprintf(name, "%d", tgid);
2878 nd_set_link(nd, name);
2882 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2885 char *s = nd_get_link(nd);
2890 static const struct inode_operations proc_self_inode_operations = {
2891 .readlink = proc_self_readlink,
2892 .follow_link = proc_self_follow_link,
2893 .put_link = proc_self_put_link,
2899 * These are the directory entries in the root directory of /proc
2900 * that properly belong to the /proc filesystem, as they describe
2901 * describe something that is process related.
2903 static const struct pid_entry proc_base_stuff[] = {
2904 NOD("self", S_IFLNK|S_IRWXUGO,
2905 &proc_self_inode_operations, NULL, {}),
2908 static struct dentry *proc_base_instantiate(struct inode *dir,
2909 struct dentry *dentry, struct task_struct *task, const void *ptr)
2911 const struct pid_entry *p = ptr;
2912 struct inode *inode;
2913 struct proc_inode *ei;
2914 struct dentry *error;
2916 /* Allocate the inode */
2917 error = ERR_PTR(-ENOMEM);
2918 inode = new_inode(dir->i_sb);
2922 /* Initialize the inode */
2924 inode->i_ino = get_next_ino();
2925 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2928 * grab the reference to the task.
2930 ei->pid = get_task_pid(task, PIDTYPE_PID);
2934 inode->i_mode = p->mode;
2935 if (S_ISDIR(inode->i_mode))
2936 set_nlink(inode, 2);
2937 if (S_ISLNK(inode->i_mode))
2940 inode->i_op = p->iop;
2942 inode->i_fop = p->fop;
2944 d_add(dentry, inode);
2953 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2955 struct dentry *error;
2956 struct task_struct *task = get_proc_task(dir);
2957 const struct pid_entry *p, *last;
2959 error = ERR_PTR(-ENOENT);
2964 /* Lookup the directory entry */
2965 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2966 for (p = proc_base_stuff; p <= last; p++) {
2967 if (p->len != dentry->d_name.len)
2969 if (!memcmp(dentry->d_name.name, p->name, p->len))
2975 error = proc_base_instantiate(dir, dentry, task, p);
2978 put_task_struct(task);
2983 static int proc_base_fill_cache(struct file *filp, void *dirent,
2984 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2986 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2987 proc_base_instantiate, task, p);
2990 #ifdef CONFIG_TASK_IO_ACCOUNTING
2991 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2993 struct task_io_accounting acct = task->ioac;
2994 unsigned long flags;
2997 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
3001 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
3006 if (whole && lock_task_sighand(task, &flags)) {
3007 struct task_struct *t = task;
3009 task_io_accounting_add(&acct, &task->signal->ioac);
3010 while_each_thread(task, t)
3011 task_io_accounting_add(&acct, &t->ioac);
3013 unlock_task_sighand(task, &flags);
3015 result = sprintf(buffer,
3020 "read_bytes: %llu\n"
3021 "write_bytes: %llu\n"
3022 "cancelled_write_bytes: %llu\n",
3023 (unsigned long long)acct.rchar,
3024 (unsigned long long)acct.wchar,
3025 (unsigned long long)acct.syscr,
3026 (unsigned long long)acct.syscw,
3027 (unsigned long long)acct.read_bytes,
3028 (unsigned long long)acct.write_bytes,
3029 (unsigned long long)acct.cancelled_write_bytes);
3031 mutex_unlock(&task->signal->cred_guard_mutex);
3035 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
3037 return do_io_accounting(task, buffer, 0);
3040 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
3042 return do_io_accounting(task, buffer, 1);
3044 #endif /* CONFIG_TASK_IO_ACCOUNTING */
3046 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
3047 struct pid *pid, struct task_struct *task)
3049 int err = lock_trace(task);
3051 seq_printf(m, "%08x\n", task->personality);
3060 static const struct file_operations proc_task_operations;
3061 static const struct inode_operations proc_task_inode_operations;
3063 static const struct pid_entry tgid_base_stuff[] = {
3064 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
3065 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3066 #ifdef CONFIG_CHECKPOINT_RESTORE
3067 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
3069 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3070 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3072 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3074 REG("environ", S_IRUSR, proc_environ_operations),
3075 INF("auxv", S_IRUSR, proc_pid_auxv),
3076 ONE("status", S_IRUGO, proc_pid_status),
3077 ONE("personality", S_IRUGO, proc_pid_personality),
3078 INF("limits", S_IRUGO, proc_pid_limits),
3079 #ifdef CONFIG_SCHED_DEBUG
3080 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3082 #ifdef CONFIG_SCHED_AUTOGROUP
3083 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
3085 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3086 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3087 INF("syscall", S_IRUGO, proc_pid_syscall),
3089 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3090 ONE("stat", S_IRUGO, proc_tgid_stat),
3091 ONE("statm", S_IRUGO, proc_pid_statm),
3092 REG("maps", S_IRUGO, proc_maps_operations),
3094 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
3096 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3097 LNK("cwd", proc_cwd_link),
3098 LNK("root", proc_root_link),
3099 LNK("exe", proc_exe_link),
3100 REG("mounts", S_IRUGO, proc_mounts_operations),
3101 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3102 REG("mountstats", S_IRUSR, proc_mountstats_operations),
3103 #ifdef CONFIG_PROC_PAGE_MONITOR
3104 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3105 REG("smaps", S_IRUGO, proc_smaps_operations),
3106 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3108 #ifdef CONFIG_SECURITY
3109 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3111 #ifdef CONFIG_KALLSYMS
3112 INF("wchan", S_IRUGO, proc_pid_wchan),
3114 #ifdef CONFIG_STACKTRACE
3115 ONE("stack", S_IRUGO, proc_pid_stack),
3117 #ifdef CONFIG_SCHEDSTATS
3118 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3120 #ifdef CONFIG_LATENCYTOP
3121 REG("latency", S_IRUGO, proc_lstats_operations),
3123 #ifdef CONFIG_PROC_PID_CPUSET
3124 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3126 #ifdef CONFIG_CGROUPS
3127 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3129 INF("oom_score", S_IRUGO, proc_oom_score),
3130 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3131 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3132 #ifdef CONFIG_AUDITSYSCALL
3133 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3134 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3136 #ifdef CONFIG_FAULT_INJECTION
3137 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3139 #ifdef CONFIG_ELF_CORE
3140 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3142 #ifdef CONFIG_TASK_IO_ACCOUNTING
3143 INF("io", S_IRUSR, proc_tgid_io_accounting),
3145 #ifdef CONFIG_HARDWALL
3146 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3150 static int proc_tgid_base_readdir(struct file * filp,
3151 void * dirent, filldir_t filldir)
3153 return proc_pident_readdir(filp,dirent,filldir,
3154 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
3157 static const struct file_operations proc_tgid_base_operations = {
3158 .read = generic_read_dir,
3159 .readdir = proc_tgid_base_readdir,
3160 .llseek = default_llseek,
3163 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3164 return proc_pident_lookup(dir, dentry,
3165 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3168 static const struct inode_operations proc_tgid_base_inode_operations = {
3169 .lookup = proc_tgid_base_lookup,
3170 .getattr = pid_getattr,
3171 .setattr = proc_setattr,
3172 .permission = proc_pid_permission,
3175 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
3177 struct dentry *dentry, *leader, *dir;
3178 char buf[PROC_NUMBUF];
3182 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3183 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
3185 shrink_dcache_parent(dentry);
3191 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
3192 leader = d_hash_and_lookup(mnt->mnt_root, &name);
3197 name.len = strlen(name.name);
3198 dir = d_hash_and_lookup(leader, &name);
3200 goto out_put_leader;
3203 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3204 dentry = d_hash_and_lookup(dir, &name);
3206 shrink_dcache_parent(dentry);
3219 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3220 * @task: task that should be flushed.
3222 * When flushing dentries from proc, one needs to flush them from global
3223 * proc (proc_mnt) and from all the namespaces' procs this task was seen
3224 * in. This call is supposed to do all of this job.
3226 * Looks in the dcache for
3228 * /proc/@tgid/task/@pid
3229 * if either directory is present flushes it and all of it'ts children
3232 * It is safe and reasonable to cache /proc entries for a task until
3233 * that task exits. After that they just clog up the dcache with
3234 * useless entries, possibly causing useful dcache entries to be
3235 * flushed instead. This routine is proved to flush those useless
3236 * dcache entries at process exit time.
3238 * NOTE: This routine is just an optimization so it does not guarantee
3239 * that no dcache entries will exist at process exit time it
3240 * just makes it very unlikely that any will persist.
3243 void proc_flush_task(struct task_struct *task)
3246 struct pid *pid, *tgid;
3249 pid = task_pid(task);
3250 tgid = task_tgid(task);
3252 for (i = 0; i <= pid->level; i++) {
3253 upid = &pid->numbers[i];
3254 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
3255 tgid->numbers[i].nr);
3258 upid = &pid->numbers[pid->level];
3260 pid_ns_release_proc(upid->ns);
3263 static struct dentry *proc_pid_instantiate(struct inode *dir,
3264 struct dentry * dentry,
3265 struct task_struct *task, const void *ptr)
3267 struct dentry *error = ERR_PTR(-ENOENT);
3268 struct inode *inode;
3270 inode = proc_pid_make_inode(dir->i_sb, task);
3274 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3275 inode->i_op = &proc_tgid_base_inode_operations;
3276 inode->i_fop = &proc_tgid_base_operations;
3277 inode->i_flags|=S_IMMUTABLE;
3279 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
3280 ARRAY_SIZE(tgid_base_stuff)));
3282 d_set_d_op(dentry, &pid_dentry_operations);
3284 d_add(dentry, inode);
3285 /* Close the race of the process dying before we return the dentry */
3286 if (pid_revalidate(dentry, NULL))
3292 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3294 struct dentry *result;
3295 struct task_struct *task;
3297 struct pid_namespace *ns;
3299 result = proc_base_lookup(dir, dentry);
3300 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3303 tgid = name_to_int(dentry);
3307 ns = dentry->d_sb->s_fs_info;
3309 task = find_task_by_pid_ns(tgid, ns);
3311 get_task_struct(task);
3316 result = proc_pid_instantiate(dir, dentry, task, NULL);
3317 put_task_struct(task);
3323 * Find the first task with tgid >= tgid
3328 struct task_struct *task;
3330 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3335 put_task_struct(iter.task);
3339 pid = find_ge_pid(iter.tgid, ns);
3341 iter.tgid = pid_nr_ns(pid, ns);
3342 iter.task = pid_task(pid, PIDTYPE_PID);
3343 /* What we to know is if the pid we have find is the
3344 * pid of a thread_group_leader. Testing for task
3345 * being a thread_group_leader is the obvious thing
3346 * todo but there is a window when it fails, due to
3347 * the pid transfer logic in de_thread.
3349 * So we perform the straight forward test of seeing
3350 * if the pid we have found is the pid of a thread
3351 * group leader, and don't worry if the task we have
3352 * found doesn't happen to be a thread group leader.
3353 * As we don't care in the case of readdir.
3355 if (!iter.task || !has_group_leader_pid(iter.task)) {
3359 get_task_struct(iter.task);
3365 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
3367 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3368 struct tgid_iter iter)
3370 char name[PROC_NUMBUF];
3371 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
3372 return proc_fill_cache(filp, dirent, filldir, name, len,
3373 proc_pid_instantiate, iter.task, NULL);
3376 static int fake_filldir(void *buf, const char *name, int namelen,
3377 loff_t offset, u64 ino, unsigned d_type)
3382 /* for the /proc/ directory itself, after non-process stuff has been done */
3383 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3386 struct task_struct *reaper;
3387 struct tgid_iter iter;
3388 struct pid_namespace *ns;
3389 filldir_t __filldir;
3391 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3393 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3395 reaper = get_proc_task(filp->f_path.dentry->d_inode);
3399 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
3400 const struct pid_entry *p = &proc_base_stuff[nr];
3401 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
3405 ns = filp->f_dentry->d_sb->s_fs_info;
3407 iter.tgid = filp->f_pos - TGID_OFFSET;
3408 for (iter = next_tgid(ns, iter);
3410 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3411 if (has_pid_permissions(ns, iter.task, 2))
3412 __filldir = filldir;
3414 __filldir = fake_filldir;
3416 filp->f_pos = iter.tgid + TGID_OFFSET;
3417 if (proc_pid_fill_cache(filp, dirent, __filldir, iter) < 0) {
3418 put_task_struct(iter.task);
3422 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3424 put_task_struct(reaper);
3432 static const struct pid_entry tid_base_stuff[] = {
3433 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3434 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3435 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3436 REG("environ", S_IRUSR, proc_environ_operations),
3437 INF("auxv", S_IRUSR, proc_pid_auxv),
3438 ONE("status", S_IRUGO, proc_pid_status),
3439 ONE("personality", S_IRUGO, proc_pid_personality),
3440 INF("limits", S_IRUGO, proc_pid_limits),
3441 #ifdef CONFIG_SCHED_DEBUG
3442 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3444 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3445 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3446 INF("syscall", S_IRUGO, proc_pid_syscall),
3448 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3449 ONE("stat", S_IRUGO, proc_tid_stat),
3450 ONE("statm", S_IRUGO, proc_pid_statm),
3451 REG("maps", S_IRUGO, proc_maps_operations),
3453 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
3455 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3456 LNK("cwd", proc_cwd_link),
3457 LNK("root", proc_root_link),
3458 LNK("exe", proc_exe_link),
3459 REG("mounts", S_IRUGO, proc_mounts_operations),
3460 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3461 #ifdef CONFIG_PROC_PAGE_MONITOR
3462 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3463 REG("smaps", S_IRUGO, proc_smaps_operations),
3464 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3466 #ifdef CONFIG_SECURITY
3467 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3469 #ifdef CONFIG_KALLSYMS
3470 INF("wchan", S_IRUGO, proc_pid_wchan),
3472 #ifdef CONFIG_STACKTRACE
3473 ONE("stack", S_IRUGO, proc_pid_stack),
3475 #ifdef CONFIG_SCHEDSTATS
3476 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3478 #ifdef CONFIG_LATENCYTOP
3479 REG("latency", S_IRUGO, proc_lstats_operations),
3481 #ifdef CONFIG_PROC_PID_CPUSET
3482 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3484 #ifdef CONFIG_CGROUPS
3485 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3487 INF("oom_score", S_IRUGO, proc_oom_score),
3488 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3489 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3490 #ifdef CONFIG_AUDITSYSCALL
3491 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3492 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3494 #ifdef CONFIG_FAULT_INJECTION
3495 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3497 #ifdef CONFIG_TASK_IO_ACCOUNTING
3498 INF("io", S_IRUSR, proc_tid_io_accounting),
3500 #ifdef CONFIG_HARDWALL
3501 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3505 static int proc_tid_base_readdir(struct file * filp,
3506 void * dirent, filldir_t filldir)
3508 return proc_pident_readdir(filp,dirent,filldir,
3509 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3512 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3513 return proc_pident_lookup(dir, dentry,
3514 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3517 static const struct file_operations proc_tid_base_operations = {
3518 .read = generic_read_dir,
3519 .readdir = proc_tid_base_readdir,
3520 .llseek = default_llseek,
3523 static const struct inode_operations proc_tid_base_inode_operations = {
3524 .lookup = proc_tid_base_lookup,
3525 .getattr = pid_getattr,
3526 .setattr = proc_setattr,
3529 static struct dentry *proc_task_instantiate(struct inode *dir,
3530 struct dentry *dentry, struct task_struct *task, const void *ptr)
3532 struct dentry *error = ERR_PTR(-ENOENT);
3533 struct inode *inode;
3534 inode = proc_pid_make_inode(dir->i_sb, task);
3538 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3539 inode->i_op = &proc_tid_base_inode_operations;
3540 inode->i_fop = &proc_tid_base_operations;
3541 inode->i_flags|=S_IMMUTABLE;
3543 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3544 ARRAY_SIZE(tid_base_stuff)));
3546 d_set_d_op(dentry, &pid_dentry_operations);
3548 d_add(dentry, inode);
3549 /* Close the race of the process dying before we return the dentry */
3550 if (pid_revalidate(dentry, NULL))
3556 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3558 struct dentry *result = ERR_PTR(-ENOENT);
3559 struct task_struct *task;
3560 struct task_struct *leader = get_proc_task(dir);
3562 struct pid_namespace *ns;
3567 tid = name_to_int(dentry);
3571 ns = dentry->d_sb->s_fs_info;
3573 task = find_task_by_pid_ns(tid, ns);
3575 get_task_struct(task);
3579 if (!same_thread_group(leader, task))
3582 result = proc_task_instantiate(dir, dentry, task, NULL);
3584 put_task_struct(task);
3586 put_task_struct(leader);
3592 * Find the first tid of a thread group to return to user space.
3594 * Usually this is just the thread group leader, but if the users
3595 * buffer was too small or there was a seek into the middle of the
3596 * directory we have more work todo.
3598 * In the case of a short read we start with find_task_by_pid.
3600 * In the case of a seek we start with the leader and walk nr
3603 static struct task_struct *first_tid(struct task_struct *leader,
3604 int tid, int nr, struct pid_namespace *ns)
3606 struct task_struct *pos;
3609 /* Attempt to start with the pid of a thread */
3610 if (tid && (nr > 0)) {
3611 pos = find_task_by_pid_ns(tid, ns);
3612 if (pos && (pos->group_leader == leader))
3616 /* If nr exceeds the number of threads there is nothing todo */
3618 if (nr && nr >= get_nr_threads(leader))
3621 /* If we haven't found our starting place yet start
3622 * with the leader and walk nr threads forward.
3624 for (pos = leader; nr > 0; --nr) {
3625 pos = next_thread(pos);
3626 if (pos == leader) {
3632 get_task_struct(pos);
3639 * Find the next thread in the thread list.
3640 * Return NULL if there is an error or no next thread.
3642 * The reference to the input task_struct is released.
3644 static struct task_struct *next_tid(struct task_struct *start)
3646 struct task_struct *pos = NULL;
3648 if (pid_alive(start)) {
3649 pos = next_thread(start);
3650 if (thread_group_leader(pos))
3653 get_task_struct(pos);
3656 put_task_struct(start);
3660 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3661 struct task_struct *task, int tid)
3663 char name[PROC_NUMBUF];
3664 int len = snprintf(name, sizeof(name), "%d", tid);
3665 return proc_fill_cache(filp, dirent, filldir, name, len,
3666 proc_task_instantiate, task, NULL);
3669 /* for the /proc/TGID/task/ directories */
3670 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3672 struct dentry *dentry = filp->f_path.dentry;
3673 struct inode *inode = dentry->d_inode;
3674 struct task_struct *leader = NULL;
3675 struct task_struct *task;
3676 int retval = -ENOENT;
3679 struct pid_namespace *ns;
3681 task = get_proc_task(inode);
3685 if (pid_alive(task)) {
3686 leader = task->group_leader;
3687 get_task_struct(leader);
3690 put_task_struct(task);
3695 switch ((unsigned long)filp->f_pos) {
3698 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3703 ino = parent_ino(dentry);
3704 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3710 /* f_version caches the tgid value that the last readdir call couldn't
3711 * return. lseek aka telldir automagically resets f_version to 0.
3713 ns = filp->f_dentry->d_sb->s_fs_info;
3714 tid = (int)filp->f_version;
3715 filp->f_version = 0;
3716 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3718 task = next_tid(task), filp->f_pos++) {
3719 tid = task_pid_nr_ns(task, ns);
3720 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3721 /* returning this tgid failed, save it as the first
3722 * pid for the next readir call */
3723 filp->f_version = (u64)tid;
3724 put_task_struct(task);
3729 put_task_struct(leader);
3734 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3736 struct inode *inode = dentry->d_inode;
3737 struct task_struct *p = get_proc_task(inode);
3738 generic_fillattr(inode, stat);
3741 stat->nlink += get_nr_threads(p);
3748 static const struct inode_operations proc_task_inode_operations = {
3749 .lookup = proc_task_lookup,
3750 .getattr = proc_task_getattr,
3751 .setattr = proc_setattr,
3752 .permission = proc_pid_permission,
3755 static const struct file_operations proc_task_operations = {
3756 .read = generic_read_dir,
3757 .readdir = proc_task_readdir,
3758 .llseek = default_llseek,