4 * Copyright (C) 1991, 1992 Linus Torvalds
6 * proc base directory handling functions
8 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
9 * Instead of using magical inumbers to determine the kind of object
10 * we allocate and fill in-core inodes upon lookup. They don't even
11 * go into icache. We cache the reference to task_struct upon lookup too.
12 * Eventually it should become a filesystem in its own. We don't use the
13 * rest of procfs anymore.
19 * Bruna Moreira <bruna.moreira@indt.org.br>
20 * Edjard Mota <edjard.mota@indt.org.br>
21 * Ilias Biris <ilias.biris@indt.org.br>
22 * Mauricio Lin <mauricio.lin@indt.org.br>
24 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
26 * A new process specific entry (smaps) included in /proc. It shows the
27 * size of rss for each memory area. The maps entry lacks information
28 * about physical memory size (rss) for each mapped file, i.e.,
29 * rss information for executables and library files.
30 * This additional information is useful for any tools that need to know
31 * about physical memory consumption for a process specific library.
35 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36 * Pud inclusion in the page table walking.
40 * 10LE Instituto Nokia de Tecnologia - INdT:
41 * A better way to walks through the page table as suggested by Hugh Dickins.
43 * Simo Piiroinen <simo.piiroinen@nokia.com>:
44 * Smaps information related to shared, private, clean and dirty pages.
46 * Paul Mundt <paul.mundt@nokia.com>:
47 * Overall revision about smaps.
50 #include <asm/uaccess.h>
52 #include <linux/errno.h>
53 #include <linux/time.h>
54 #include <linux/proc_fs.h>
55 #include <linux/stat.h>
56 #include <linux/task_io_accounting_ops.h>
57 #include <linux/init.h>
58 #include <linux/capability.h>
59 #include <linux/file.h>
60 #include <linux/fdtable.h>
61 #include <linux/string.h>
62 #include <linux/seq_file.h>
63 #include <linux/namei.h>
64 #include <linux/mnt_namespace.h>
66 #include <linux/swap.h>
67 #include <linux/rcupdate.h>
68 #include <linux/kallsyms.h>
69 #include <linux/stacktrace.h>
70 #include <linux/resource.h>
71 #include <linux/module.h>
72 #include <linux/mount.h>
73 #include <linux/security.h>
74 #include <linux/ptrace.h>
75 #include <linux/tracehook.h>
76 #include <linux/cgroup.h>
77 #include <linux/cpuset.h>
78 #include <linux/audit.h>
79 #include <linux/poll.h>
80 #include <linux/nsproxy.h>
81 #include <linux/oom.h>
82 #include <linux/elf.h>
83 #include <linux/pid_namespace.h>
84 #include <linux/user_namespace.h>
85 #include <linux/fs_struct.h>
86 #include <linux/slab.h>
87 #include <linux/flex_array.h>
88 #ifdef CONFIG_HARDWALL
89 #include <asm/hardwall.h>
91 #include <trace/events/oom.h>
95 * Implementing inode permission operations in /proc is almost
96 * certainly an error. Permission checks need to happen during
97 * each system call not at open time. The reason is that most of
98 * what we wish to check for permissions in /proc varies at runtime.
100 * The classic example of a problem is opening file descriptors
101 * in /proc for a task before it execs a suid executable.
108 const struct inode_operations *iop;
109 const struct file_operations *fop;
113 #define NOD(NAME, MODE, IOP, FOP, OP) { \
115 .len = sizeof(NAME) - 1, \
122 #define DIR(NAME, MODE, iops, fops) \
123 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
124 #define LNK(NAME, get_link) \
125 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
126 &proc_pid_link_inode_operations, NULL, \
127 { .proc_get_link = get_link } )
128 #define REG(NAME, MODE, fops) \
129 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
130 #define INF(NAME, MODE, read) \
131 NOD(NAME, (S_IFREG|(MODE)), \
132 NULL, &proc_info_file_operations, \
133 { .proc_read = read } )
134 #define ONE(NAME, MODE, show) \
135 NOD(NAME, (S_IFREG|(MODE)), \
136 NULL, &proc_single_file_operations, \
137 { .proc_show = show } )
139 static int proc_fd_permission(struct inode *inode, int mask);
142 * Count the number of hardlinks for the pid_entry table, excluding the .
145 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
152 for (i = 0; i < n; ++i) {
153 if (S_ISDIR(entries[i].mode))
160 static int get_task_root(struct task_struct *task, struct path *root)
162 int result = -ENOENT;
166 get_fs_root(task->fs, root);
173 static int proc_cwd_link(struct dentry *dentry, struct path *path)
175 struct task_struct *task = get_proc_task(dentry->d_inode);
176 int result = -ENOENT;
181 get_fs_pwd(task->fs, path);
185 put_task_struct(task);
190 static int proc_root_link(struct dentry *dentry, struct path *path)
192 struct task_struct *task = get_proc_task(dentry->d_inode);
193 int result = -ENOENT;
196 result = get_task_root(task, path);
197 put_task_struct(task);
202 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
206 struct mm_struct *mm = get_task_mm(task);
210 goto out_mm; /* Shh! No looking before we're done */
212 len = mm->arg_end - mm->arg_start;
217 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
219 // If the nul at the end of args has been overwritten, then
220 // assume application is using setproctitle(3).
221 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
222 len = strnlen(buffer, res);
226 len = mm->env_end - mm->env_start;
227 if (len > PAGE_SIZE - res)
228 len = PAGE_SIZE - res;
229 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
230 res = strnlen(buffer, res);
239 static int proc_pid_auxv(struct task_struct *task, char *buffer)
241 struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ);
242 int res = PTR_ERR(mm);
243 if (mm && !IS_ERR(mm)) {
244 unsigned int nwords = 0;
247 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
248 res = nwords * sizeof(mm->saved_auxv[0]);
251 memcpy(buffer, mm->saved_auxv, res);
258 #ifdef CONFIG_KALLSYMS
260 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
261 * Returns the resolved symbol. If that fails, simply return the address.
263 static int proc_pid_wchan(struct task_struct *task, char *buffer)
266 char symname[KSYM_NAME_LEN];
268 wchan = get_wchan(task);
270 if (lookup_symbol_name(wchan, symname) < 0)
271 if (!ptrace_may_access(task, PTRACE_MODE_READ))
274 return sprintf(buffer, "%lu", wchan);
276 return sprintf(buffer, "%s", symname);
278 #endif /* CONFIG_KALLSYMS */
280 static int lock_trace(struct task_struct *task)
282 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
285 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
286 mutex_unlock(&task->signal->cred_guard_mutex);
292 static void unlock_trace(struct task_struct *task)
294 mutex_unlock(&task->signal->cred_guard_mutex);
297 #ifdef CONFIG_STACKTRACE
299 #define MAX_STACK_TRACE_DEPTH 64
301 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
302 struct pid *pid, struct task_struct *task)
304 struct stack_trace trace;
305 unsigned long *entries;
309 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
313 trace.nr_entries = 0;
314 trace.max_entries = MAX_STACK_TRACE_DEPTH;
315 trace.entries = entries;
318 err = lock_trace(task);
320 save_stack_trace_tsk(task, &trace);
322 for (i = 0; i < trace.nr_entries; i++) {
323 seq_printf(m, "[<%pK>] %pS\n",
324 (void *)entries[i], (void *)entries[i]);
334 #ifdef CONFIG_SCHEDSTATS
336 * Provides /proc/PID/schedstat
338 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
340 return sprintf(buffer, "%llu %llu %lu\n",
341 (unsigned long long)task->se.sum_exec_runtime,
342 (unsigned long long)task->sched_info.run_delay,
343 task->sched_info.pcount);
347 #ifdef CONFIG_LATENCYTOP
348 static int lstats_show_proc(struct seq_file *m, void *v)
351 struct inode *inode = m->private;
352 struct task_struct *task = get_proc_task(inode);
356 seq_puts(m, "Latency Top version : v0.1\n");
357 for (i = 0; i < 32; i++) {
358 struct latency_record *lr = &task->latency_record[i];
359 if (lr->backtrace[0]) {
361 seq_printf(m, "%i %li %li",
362 lr->count, lr->time, lr->max);
363 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
364 unsigned long bt = lr->backtrace[q];
369 seq_printf(m, " %ps", (void *)bt);
375 put_task_struct(task);
379 static int lstats_open(struct inode *inode, struct file *file)
381 return single_open(file, lstats_show_proc, inode);
384 static ssize_t lstats_write(struct file *file, const char __user *buf,
385 size_t count, loff_t *offs)
387 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
391 clear_all_latency_tracing(task);
392 put_task_struct(task);
397 static const struct file_operations proc_lstats_operations = {
400 .write = lstats_write,
402 .release = single_release,
407 static int proc_oom_score(struct task_struct *task, char *buffer)
409 unsigned long totalpages = totalram_pages + total_swap_pages;
410 unsigned long points = 0;
412 read_lock(&tasklist_lock);
414 points = oom_badness(task, NULL, NULL, totalpages) *
416 read_unlock(&tasklist_lock);
417 return sprintf(buffer, "%lu\n", points);
425 static const struct limit_names lnames[RLIM_NLIMITS] = {
426 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
427 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
428 [RLIMIT_DATA] = {"Max data size", "bytes"},
429 [RLIMIT_STACK] = {"Max stack size", "bytes"},
430 [RLIMIT_CORE] = {"Max core file size", "bytes"},
431 [RLIMIT_RSS] = {"Max resident set", "bytes"},
432 [RLIMIT_NPROC] = {"Max processes", "processes"},
433 [RLIMIT_NOFILE] = {"Max open files", "files"},
434 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
435 [RLIMIT_AS] = {"Max address space", "bytes"},
436 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
437 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
438 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
439 [RLIMIT_NICE] = {"Max nice priority", NULL},
440 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
441 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
444 /* Display limits for a process */
445 static int proc_pid_limits(struct task_struct *task, char *buffer)
450 char *bufptr = buffer;
452 struct rlimit rlim[RLIM_NLIMITS];
454 if (!lock_task_sighand(task, &flags))
456 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
457 unlock_task_sighand(task, &flags);
460 * print the file header
462 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
463 "Limit", "Soft Limit", "Hard Limit", "Units");
465 for (i = 0; i < RLIM_NLIMITS; i++) {
466 if (rlim[i].rlim_cur == RLIM_INFINITY)
467 count += sprintf(&bufptr[count], "%-25s %-20s ",
468 lnames[i].name, "unlimited");
470 count += sprintf(&bufptr[count], "%-25s %-20lu ",
471 lnames[i].name, rlim[i].rlim_cur);
473 if (rlim[i].rlim_max == RLIM_INFINITY)
474 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
476 count += sprintf(&bufptr[count], "%-20lu ",
480 count += sprintf(&bufptr[count], "%-10s\n",
483 count += sprintf(&bufptr[count], "\n");
489 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
490 static int proc_pid_syscall(struct task_struct *task, char *buffer)
493 unsigned long args[6], sp, pc;
494 int res = lock_trace(task);
498 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
499 res = sprintf(buffer, "running\n");
501 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
503 res = sprintf(buffer,
504 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
506 args[0], args[1], args[2], args[3], args[4], args[5],
511 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
513 /************************************************************************/
514 /* Here the fs part begins */
515 /************************************************************************/
517 /* permission checks */
518 static int proc_fd_access_allowed(struct inode *inode)
520 struct task_struct *task;
522 /* Allow access to a task's file descriptors if it is us or we
523 * may use ptrace attach to the process and find out that
526 task = get_proc_task(inode);
528 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
529 put_task_struct(task);
534 int proc_setattr(struct dentry *dentry, struct iattr *attr)
537 struct inode *inode = dentry->d_inode;
539 if (attr->ia_valid & ATTR_MODE)
542 error = inode_change_ok(inode, attr);
546 if ((attr->ia_valid & ATTR_SIZE) &&
547 attr->ia_size != i_size_read(inode)) {
548 error = vmtruncate(inode, attr->ia_size);
553 setattr_copy(inode, attr);
554 mark_inode_dirty(inode);
559 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
560 * or euid/egid (for hide_pid_min=2)?
562 static bool has_pid_permissions(struct pid_namespace *pid,
563 struct task_struct *task,
566 if (pid->hide_pid < hide_pid_min)
568 if (in_group_p(pid->pid_gid))
570 return ptrace_may_access(task, PTRACE_MODE_READ);
574 static int proc_pid_permission(struct inode *inode, int mask)
576 struct pid_namespace *pid = inode->i_sb->s_fs_info;
577 struct task_struct *task;
580 task = get_proc_task(inode);
583 has_perms = has_pid_permissions(pid, task, 1);
584 put_task_struct(task);
587 if (pid->hide_pid == 2) {
589 * Let's make getdents(), stat(), and open()
590 * consistent with each other. If a process
591 * may not stat() a file, it shouldn't be seen
599 return generic_permission(inode, mask);
604 static const struct inode_operations proc_def_inode_operations = {
605 .setattr = proc_setattr,
608 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
610 static ssize_t proc_info_read(struct file * file, char __user * buf,
611 size_t count, loff_t *ppos)
613 struct inode * inode = file->f_path.dentry->d_inode;
616 struct task_struct *task = get_proc_task(inode);
622 if (count > PROC_BLOCK_SIZE)
623 count = PROC_BLOCK_SIZE;
626 if (!(page = __get_free_page(GFP_TEMPORARY)))
629 length = PROC_I(inode)->op.proc_read(task, (char*)page);
632 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
635 put_task_struct(task);
640 static const struct file_operations proc_info_file_operations = {
641 .read = proc_info_read,
642 .llseek = generic_file_llseek,
645 static int proc_single_show(struct seq_file *m, void *v)
647 struct inode *inode = m->private;
648 struct pid_namespace *ns;
650 struct task_struct *task;
653 ns = inode->i_sb->s_fs_info;
654 pid = proc_pid(inode);
655 task = get_pid_task(pid, PIDTYPE_PID);
659 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
661 put_task_struct(task);
665 static int proc_single_open(struct inode *inode, struct file *filp)
667 return single_open(filp, proc_single_show, inode);
670 static const struct file_operations proc_single_file_operations = {
671 .open = proc_single_open,
674 .release = single_release,
677 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
679 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
680 struct mm_struct *mm;
685 mm = mm_access(task, mode);
686 put_task_struct(task);
692 /* ensure this mm_struct can't be freed */
693 atomic_inc(&mm->mm_count);
694 /* but do not pin its memory */
698 file->private_data = mm;
703 static int mem_open(struct inode *inode, struct file *file)
705 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
707 /* OK to pass negative loff_t, we can catch out-of-range */
708 file->f_mode |= FMODE_UNSIGNED_OFFSET;
713 static ssize_t mem_rw(struct file *file, char __user *buf,
714 size_t count, loff_t *ppos, int write)
716 struct mm_struct *mm = file->private_data;
717 unsigned long addr = *ppos;
724 page = (char *)__get_free_page(GFP_TEMPORARY);
729 if (!atomic_inc_not_zero(&mm->mm_users))
733 int this_len = min_t(int, count, PAGE_SIZE);
735 if (write && copy_from_user(page, buf, this_len)) {
740 this_len = access_remote_vm(mm, addr, page, this_len, write);
747 if (!write && copy_to_user(buf, page, this_len)) {
761 free_page((unsigned long) page);
765 static ssize_t mem_read(struct file *file, char __user *buf,
766 size_t count, loff_t *ppos)
768 return mem_rw(file, buf, count, ppos, 0);
771 static ssize_t mem_write(struct file *file, const char __user *buf,
772 size_t count, loff_t *ppos)
774 return mem_rw(file, (char __user*)buf, count, ppos, 1);
777 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
781 file->f_pos = offset;
784 file->f_pos += offset;
789 force_successful_syscall_return();
793 static int mem_release(struct inode *inode, struct file *file)
795 struct mm_struct *mm = file->private_data;
801 static const struct file_operations proc_mem_operations = {
806 .release = mem_release,
809 static int environ_open(struct inode *inode, struct file *file)
811 return __mem_open(inode, file, PTRACE_MODE_READ);
814 static ssize_t environ_read(struct file *file, char __user *buf,
815 size_t count, loff_t *ppos)
818 unsigned long src = *ppos;
820 struct mm_struct *mm = file->private_data;
825 page = (char *)__get_free_page(GFP_TEMPORARY);
830 if (!atomic_inc_not_zero(&mm->mm_users))
833 size_t this_len, max_len;
836 if (src >= (mm->env_end - mm->env_start))
839 this_len = mm->env_end - (mm->env_start + src);
841 max_len = min_t(size_t, PAGE_SIZE, count);
842 this_len = min(max_len, this_len);
844 retval = access_remote_vm(mm, (mm->env_start + src),
852 if (copy_to_user(buf, page, retval)) {
866 free_page((unsigned long) page);
870 static const struct file_operations proc_environ_operations = {
871 .open = environ_open,
872 .read = environ_read,
873 .llseek = generic_file_llseek,
874 .release = mem_release,
877 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
878 size_t count, loff_t *ppos)
880 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
881 char buffer[PROC_NUMBUF];
883 int oom_adjust = OOM_DISABLE;
889 if (lock_task_sighand(task, &flags)) {
890 oom_adjust = task->signal->oom_adj;
891 unlock_task_sighand(task, &flags);
894 put_task_struct(task);
896 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
898 return simple_read_from_buffer(buf, count, ppos, buffer, len);
901 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
902 size_t count, loff_t *ppos)
904 struct task_struct *task;
905 char buffer[PROC_NUMBUF];
910 memset(buffer, 0, sizeof(buffer));
911 if (count > sizeof(buffer) - 1)
912 count = sizeof(buffer) - 1;
913 if (copy_from_user(buffer, buf, count)) {
918 err = kstrtoint(strstrip(buffer), 0, &oom_adjust);
921 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
922 oom_adjust != OOM_DISABLE) {
927 task = get_proc_task(file->f_path.dentry->d_inode);
939 if (!lock_task_sighand(task, &flags)) {
944 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
950 * Warn that /proc/pid/oom_adj is deprecated, see
951 * Documentation/feature-removal-schedule.txt.
953 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
954 current->comm, task_pid_nr(current), task_pid_nr(task),
956 task->signal->oom_adj = oom_adjust;
958 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
959 * value is always attainable.
961 if (task->signal->oom_adj == OOM_ADJUST_MAX)
962 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
964 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
966 trace_oom_score_adj_update(task);
968 unlock_task_sighand(task, &flags);
971 put_task_struct(task);
973 return err < 0 ? err : count;
976 static const struct file_operations proc_oom_adjust_operations = {
977 .read = oom_adjust_read,
978 .write = oom_adjust_write,
979 .llseek = generic_file_llseek,
982 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
983 size_t count, loff_t *ppos)
985 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
986 char buffer[PROC_NUMBUF];
987 int oom_score_adj = OOM_SCORE_ADJ_MIN;
993 if (lock_task_sighand(task, &flags)) {
994 oom_score_adj = task->signal->oom_score_adj;
995 unlock_task_sighand(task, &flags);
997 put_task_struct(task);
998 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
999 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1002 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1003 size_t count, loff_t *ppos)
1005 struct task_struct *task;
1006 char buffer[PROC_NUMBUF];
1007 unsigned long flags;
1011 memset(buffer, 0, sizeof(buffer));
1012 if (count > sizeof(buffer) - 1)
1013 count = sizeof(buffer) - 1;
1014 if (copy_from_user(buffer, buf, count)) {
1019 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1022 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1023 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1028 task = get_proc_task(file->f_path.dentry->d_inode);
1040 if (!lock_task_sighand(task, &flags)) {
1045 if (oom_score_adj < task->signal->oom_score_adj_min &&
1046 !capable(CAP_SYS_RESOURCE)) {
1051 task->signal->oom_score_adj = oom_score_adj;
1052 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1053 task->signal->oom_score_adj_min = oom_score_adj;
1054 trace_oom_score_adj_update(task);
1056 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1057 * always attainable.
1059 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1060 task->signal->oom_adj = OOM_DISABLE;
1062 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1065 unlock_task_sighand(task, &flags);
1068 put_task_struct(task);
1070 return err < 0 ? err : count;
1073 static const struct file_operations proc_oom_score_adj_operations = {
1074 .read = oom_score_adj_read,
1075 .write = oom_score_adj_write,
1076 .llseek = default_llseek,
1079 #ifdef CONFIG_AUDITSYSCALL
1080 #define TMPBUFLEN 21
1081 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1082 size_t count, loff_t *ppos)
1084 struct inode * inode = file->f_path.dentry->d_inode;
1085 struct task_struct *task = get_proc_task(inode);
1087 char tmpbuf[TMPBUFLEN];
1091 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1092 audit_get_loginuid(task));
1093 put_task_struct(task);
1094 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1097 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1098 size_t count, loff_t *ppos)
1100 struct inode * inode = file->f_path.dentry->d_inode;
1106 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1112 if (count >= PAGE_SIZE)
1113 count = PAGE_SIZE - 1;
1116 /* No partial writes. */
1119 page = (char*)__get_free_page(GFP_TEMPORARY);
1123 if (copy_from_user(page, buf, count))
1127 loginuid = simple_strtoul(page, &tmp, 10);
1133 length = audit_set_loginuid(loginuid);
1134 if (likely(length == 0))
1138 free_page((unsigned long) page);
1142 static const struct file_operations proc_loginuid_operations = {
1143 .read = proc_loginuid_read,
1144 .write = proc_loginuid_write,
1145 .llseek = generic_file_llseek,
1148 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1149 size_t count, loff_t *ppos)
1151 struct inode * inode = file->f_path.dentry->d_inode;
1152 struct task_struct *task = get_proc_task(inode);
1154 char tmpbuf[TMPBUFLEN];
1158 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1159 audit_get_sessionid(task));
1160 put_task_struct(task);
1161 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1164 static const struct file_operations proc_sessionid_operations = {
1165 .read = proc_sessionid_read,
1166 .llseek = generic_file_llseek,
1170 #ifdef CONFIG_FAULT_INJECTION
1171 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1172 size_t count, loff_t *ppos)
1174 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1175 char buffer[PROC_NUMBUF];
1181 make_it_fail = task->make_it_fail;
1182 put_task_struct(task);
1184 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1186 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1189 static ssize_t proc_fault_inject_write(struct file * file,
1190 const char __user * buf, size_t count, loff_t *ppos)
1192 struct task_struct *task;
1193 char buffer[PROC_NUMBUF], *end;
1196 if (!capable(CAP_SYS_RESOURCE))
1198 memset(buffer, 0, sizeof(buffer));
1199 if (count > sizeof(buffer) - 1)
1200 count = sizeof(buffer) - 1;
1201 if (copy_from_user(buffer, buf, count))
1203 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1206 task = get_proc_task(file->f_dentry->d_inode);
1209 task->make_it_fail = make_it_fail;
1210 put_task_struct(task);
1215 static const struct file_operations proc_fault_inject_operations = {
1216 .read = proc_fault_inject_read,
1217 .write = proc_fault_inject_write,
1218 .llseek = generic_file_llseek,
1223 #ifdef CONFIG_SCHED_DEBUG
1225 * Print out various scheduling related per-task fields:
1227 static int sched_show(struct seq_file *m, void *v)
1229 struct inode *inode = m->private;
1230 struct task_struct *p;
1232 p = get_proc_task(inode);
1235 proc_sched_show_task(p, m);
1243 sched_write(struct file *file, const char __user *buf,
1244 size_t count, loff_t *offset)
1246 struct inode *inode = file->f_path.dentry->d_inode;
1247 struct task_struct *p;
1249 p = get_proc_task(inode);
1252 proc_sched_set_task(p);
1259 static int sched_open(struct inode *inode, struct file *filp)
1261 return single_open(filp, sched_show, inode);
1264 static const struct file_operations proc_pid_sched_operations = {
1267 .write = sched_write,
1268 .llseek = seq_lseek,
1269 .release = single_release,
1274 static ssize_t comm_write(struct file *file, const char __user *buf,
1275 size_t count, loff_t *offset)
1277 struct inode *inode = file->f_path.dentry->d_inode;
1278 struct task_struct *p;
1279 char buffer[TASK_COMM_LEN];
1281 memset(buffer, 0, sizeof(buffer));
1282 if (count > sizeof(buffer) - 1)
1283 count = sizeof(buffer) - 1;
1284 if (copy_from_user(buffer, buf, count))
1287 p = get_proc_task(inode);
1291 if (same_thread_group(current, p))
1292 set_task_comm(p, buffer);
1301 static int comm_show(struct seq_file *m, void *v)
1303 struct inode *inode = m->private;
1304 struct task_struct *p;
1306 p = get_proc_task(inode);
1311 seq_printf(m, "%s\n", p->comm);
1319 static int comm_open(struct inode *inode, struct file *filp)
1321 return single_open(filp, comm_show, inode);
1324 static const struct file_operations proc_pid_set_comm_operations = {
1327 .write = comm_write,
1328 .llseek = seq_lseek,
1329 .release = single_release,
1332 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1334 struct task_struct *task;
1335 struct mm_struct *mm;
1336 struct file *exe_file;
1338 task = get_proc_task(dentry->d_inode);
1341 mm = get_task_mm(task);
1342 put_task_struct(task);
1345 exe_file = get_mm_exe_file(mm);
1348 *exe_path = exe_file->f_path;
1349 path_get(&exe_file->f_path);
1356 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1358 struct inode *inode = dentry->d_inode;
1360 int error = -EACCES;
1362 /* Are we allowed to snoop on the tasks file descriptors? */
1363 if (!proc_fd_access_allowed(inode))
1366 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1370 nd_jump_link(nd, &path);
1373 return ERR_PTR(error);
1376 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1378 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1385 pathname = d_path(path, tmp, PAGE_SIZE);
1386 len = PTR_ERR(pathname);
1387 if (IS_ERR(pathname))
1389 len = tmp + PAGE_SIZE - 1 - pathname;
1393 if (copy_to_user(buffer, pathname, len))
1396 free_page((unsigned long)tmp);
1400 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1402 int error = -EACCES;
1403 struct inode *inode = dentry->d_inode;
1406 /* Are we allowed to snoop on the tasks file descriptors? */
1407 if (!proc_fd_access_allowed(inode))
1410 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1414 error = do_proc_readlink(&path, buffer, buflen);
1420 static const struct inode_operations proc_pid_link_inode_operations = {
1421 .readlink = proc_pid_readlink,
1422 .follow_link = proc_pid_follow_link,
1423 .setattr = proc_setattr,
1427 /* building an inode */
1429 static int task_dumpable(struct task_struct *task)
1432 struct mm_struct *mm;
1437 dumpable = get_dumpable(mm);
1444 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1446 struct inode * inode;
1447 struct proc_inode *ei;
1448 const struct cred *cred;
1450 /* We need a new inode */
1452 inode = new_inode(sb);
1458 inode->i_ino = get_next_ino();
1459 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1460 inode->i_op = &proc_def_inode_operations;
1463 * grab the reference to task.
1465 ei->pid = get_task_pid(task, PIDTYPE_PID);
1469 if (task_dumpable(task)) {
1471 cred = __task_cred(task);
1472 inode->i_uid = cred->euid;
1473 inode->i_gid = cred->egid;
1476 security_task_to_inode(task, inode);
1486 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1488 struct inode *inode = dentry->d_inode;
1489 struct task_struct *task;
1490 const struct cred *cred;
1491 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
1493 generic_fillattr(inode, stat);
1496 stat->uid = GLOBAL_ROOT_UID;
1497 stat->gid = GLOBAL_ROOT_GID;
1498 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1500 if (!has_pid_permissions(pid, task, 2)) {
1503 * This doesn't prevent learning whether PID exists,
1504 * it only makes getattr() consistent with readdir().
1508 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1509 task_dumpable(task)) {
1510 cred = __task_cred(task);
1511 stat->uid = cred->euid;
1512 stat->gid = cred->egid;
1522 * Exceptional case: normally we are not allowed to unhash a busy
1523 * directory. In this case, however, we can do it - no aliasing problems
1524 * due to the way we treat inodes.
1526 * Rewrite the inode's ownerships here because the owning task may have
1527 * performed a setuid(), etc.
1529 * Before the /proc/pid/status file was created the only way to read
1530 * the effective uid of a /process was to stat /proc/pid. Reading
1531 * /proc/pid/status is slow enough that procps and other packages
1532 * kept stating /proc/pid. To keep the rules in /proc simple I have
1533 * made this apply to all per process world readable and executable
1536 int pid_revalidate(struct dentry *dentry, unsigned int flags)
1538 struct inode *inode;
1539 struct task_struct *task;
1540 const struct cred *cred;
1542 if (flags & LOOKUP_RCU)
1545 inode = dentry->d_inode;
1546 task = get_proc_task(inode);
1549 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1550 task_dumpable(task)) {
1552 cred = __task_cred(task);
1553 inode->i_uid = cred->euid;
1554 inode->i_gid = cred->egid;
1557 inode->i_uid = GLOBAL_ROOT_UID;
1558 inode->i_gid = GLOBAL_ROOT_GID;
1560 inode->i_mode &= ~(S_ISUID | S_ISGID);
1561 security_task_to_inode(task, inode);
1562 put_task_struct(task);
1569 static int pid_delete_dentry(const struct dentry * dentry)
1571 /* Is the task we represent dead?
1572 * If so, then don't put the dentry on the lru list,
1573 * kill it immediately.
1575 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1578 const struct dentry_operations pid_dentry_operations =
1580 .d_revalidate = pid_revalidate,
1581 .d_delete = pid_delete_dentry,
1587 * Fill a directory entry.
1589 * If possible create the dcache entry and derive our inode number and
1590 * file type from dcache entry.
1592 * Since all of the proc inode numbers are dynamically generated, the inode
1593 * numbers do not exist until the inode is cache. This means creating the
1594 * the dcache entry in readdir is necessary to keep the inode numbers
1595 * reported by readdir in sync with the inode numbers reported
1598 int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1599 const char *name, int len,
1600 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1602 struct dentry *child, *dir = filp->f_path.dentry;
1603 struct inode *inode;
1606 unsigned type = DT_UNKNOWN;
1610 qname.hash = full_name_hash(name, len);
1612 child = d_lookup(dir, &qname);
1615 new = d_alloc(dir, &qname);
1617 child = instantiate(dir->d_inode, new, task, ptr);
1624 if (!child || IS_ERR(child) || !child->d_inode)
1625 goto end_instantiate;
1626 inode = child->d_inode;
1629 type = inode->i_mode >> 12;
1634 ino = find_inode_number(dir, &qname);
1637 return filldir(dirent, name, len, filp->f_pos, ino, type);
1640 static unsigned name_to_int(struct dentry *dentry)
1642 const char *name = dentry->d_name.name;
1643 int len = dentry->d_name.len;
1646 if (len > 1 && *name == '0')
1649 unsigned c = *name++ - '0';
1652 if (n >= (~0U-9)/10)
1662 #define PROC_FDINFO_MAX 64
1664 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1666 struct task_struct *task = get_proc_task(inode);
1667 struct files_struct *files = NULL;
1669 int fd = proc_fd(inode);
1672 files = get_files_struct(task);
1673 put_task_struct(task);
1677 * We are not taking a ref to the file structure, so we must
1680 spin_lock(&files->file_lock);
1681 file = fcheck_files(files, fd);
1683 unsigned int f_flags;
1684 struct fdtable *fdt;
1686 fdt = files_fdtable(files);
1687 f_flags = file->f_flags & ~O_CLOEXEC;
1688 if (close_on_exec(fd, fdt))
1689 f_flags |= O_CLOEXEC;
1692 *path = file->f_path;
1693 path_get(&file->f_path);
1696 snprintf(info, PROC_FDINFO_MAX,
1699 (long long) file->f_pos,
1701 spin_unlock(&files->file_lock);
1702 put_files_struct(files);
1705 spin_unlock(&files->file_lock);
1706 put_files_struct(files);
1711 static int proc_fd_link(struct dentry *dentry, struct path *path)
1713 return proc_fd_info(dentry->d_inode, path, NULL);
1716 static int tid_fd_revalidate(struct dentry *dentry, unsigned int flags)
1718 struct inode *inode;
1719 struct task_struct *task;
1721 struct files_struct *files;
1722 const struct cred *cred;
1724 if (flags & LOOKUP_RCU)
1727 inode = dentry->d_inode;
1728 task = get_proc_task(inode);
1729 fd = proc_fd(inode);
1732 files = get_files_struct(task);
1736 file = fcheck_files(files, fd);
1738 unsigned f_mode = file->f_mode;
1741 put_files_struct(files);
1743 if (task_dumpable(task)) {
1745 cred = __task_cred(task);
1746 inode->i_uid = cred->euid;
1747 inode->i_gid = cred->egid;
1750 inode->i_uid = GLOBAL_ROOT_UID;
1751 inode->i_gid = GLOBAL_ROOT_GID;
1754 if (S_ISLNK(inode->i_mode)) {
1755 unsigned i_mode = S_IFLNK;
1756 if (f_mode & FMODE_READ)
1757 i_mode |= S_IRUSR | S_IXUSR;
1758 if (f_mode & FMODE_WRITE)
1759 i_mode |= S_IWUSR | S_IXUSR;
1760 inode->i_mode = i_mode;
1763 security_task_to_inode(task, inode);
1764 put_task_struct(task);
1768 put_files_struct(files);
1770 put_task_struct(task);
1776 static const struct dentry_operations tid_fd_dentry_operations =
1778 .d_revalidate = tid_fd_revalidate,
1779 .d_delete = pid_delete_dentry,
1782 static struct dentry *proc_fd_instantiate(struct inode *dir,
1783 struct dentry *dentry, struct task_struct *task, const void *ptr)
1785 unsigned fd = (unsigned long)ptr;
1786 struct inode *inode;
1787 struct proc_inode *ei;
1788 struct dentry *error = ERR_PTR(-ENOENT);
1790 inode = proc_pid_make_inode(dir->i_sb, task);
1796 inode->i_mode = S_IFLNK;
1797 inode->i_op = &proc_pid_link_inode_operations;
1799 ei->op.proc_get_link = proc_fd_link;
1800 d_set_d_op(dentry, &tid_fd_dentry_operations);
1801 d_add(dentry, inode);
1802 /* Close the race of the process dying before we return the dentry */
1803 if (tid_fd_revalidate(dentry, 0))
1810 static struct dentry *proc_lookupfd_common(struct inode *dir,
1811 struct dentry *dentry,
1812 instantiate_t instantiate)
1814 struct task_struct *task = get_proc_task(dir);
1815 unsigned fd = name_to_int(dentry);
1816 struct dentry *result = ERR_PTR(-ENOENT);
1823 result = instantiate(dir, dentry, task, (void *)(unsigned long)fd);
1825 put_task_struct(task);
1830 static int proc_readfd_common(struct file * filp, void * dirent,
1831 filldir_t filldir, instantiate_t instantiate)
1833 struct dentry *dentry = filp->f_path.dentry;
1834 struct inode *inode = dentry->d_inode;
1835 struct task_struct *p = get_proc_task(inode);
1836 unsigned int fd, ino;
1838 struct files_struct * files;
1848 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1852 ino = parent_ino(dentry);
1853 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1857 files = get_files_struct(p);
1861 for (fd = filp->f_pos-2;
1862 fd < files_fdtable(files)->max_fds;
1863 fd++, filp->f_pos++) {
1864 char name[PROC_NUMBUF];
1868 if (!fcheck_files(files, fd))
1872 len = snprintf(name, sizeof(name), "%d", fd);
1873 rv = proc_fill_cache(filp, dirent, filldir,
1874 name, len, instantiate, p,
1875 (void *)(unsigned long)fd);
1882 put_files_struct(files);
1890 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1893 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1896 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1898 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1901 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1902 size_t len, loff_t *ppos)
1904 char tmp[PROC_FDINFO_MAX];
1905 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
1907 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1911 static const struct file_operations proc_fdinfo_file_operations = {
1912 .open = nonseekable_open,
1913 .read = proc_fdinfo_read,
1914 .llseek = no_llseek,
1917 static const struct file_operations proc_fd_operations = {
1918 .read = generic_read_dir,
1919 .readdir = proc_readfd,
1920 .llseek = default_llseek,
1923 #ifdef CONFIG_CHECKPOINT_RESTORE
1926 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1927 * which represent vma start and end addresses.
1929 static int dname_to_vma_addr(struct dentry *dentry,
1930 unsigned long *start, unsigned long *end)
1932 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1938 static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
1940 unsigned long vm_start, vm_end;
1941 bool exact_vma_exists = false;
1942 struct mm_struct *mm = NULL;
1943 struct task_struct *task;
1944 const struct cred *cred;
1945 struct inode *inode;
1948 if (flags & LOOKUP_RCU)
1951 if (!capable(CAP_SYS_ADMIN)) {
1956 inode = dentry->d_inode;
1957 task = get_proc_task(inode);
1961 mm = mm_access(task, PTRACE_MODE_READ);
1962 if (IS_ERR_OR_NULL(mm))
1965 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1966 down_read(&mm->mmap_sem);
1967 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1968 up_read(&mm->mmap_sem);
1973 if (exact_vma_exists) {
1974 if (task_dumpable(task)) {
1976 cred = __task_cred(task);
1977 inode->i_uid = cred->euid;
1978 inode->i_gid = cred->egid;
1981 inode->i_uid = GLOBAL_ROOT_UID;
1982 inode->i_gid = GLOBAL_ROOT_GID;
1984 security_task_to_inode(task, inode);
1989 put_task_struct(task);
1998 static const struct dentry_operations tid_map_files_dentry_operations = {
1999 .d_revalidate = map_files_d_revalidate,
2000 .d_delete = pid_delete_dentry,
2003 static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
2005 unsigned long vm_start, vm_end;
2006 struct vm_area_struct *vma;
2007 struct task_struct *task;
2008 struct mm_struct *mm;
2012 task = get_proc_task(dentry->d_inode);
2016 mm = get_task_mm(task);
2017 put_task_struct(task);
2021 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2025 down_read(&mm->mmap_sem);
2026 vma = find_exact_vma(mm, vm_start, vm_end);
2027 if (vma && vma->vm_file) {
2028 *path = vma->vm_file->f_path;
2032 up_read(&mm->mmap_sem);
2040 struct map_files_info {
2043 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
2046 static struct dentry *
2047 proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
2048 struct task_struct *task, const void *ptr)
2050 const struct file *file = ptr;
2051 struct proc_inode *ei;
2052 struct inode *inode;
2055 return ERR_PTR(-ENOENT);
2057 inode = proc_pid_make_inode(dir->i_sb, task);
2059 return ERR_PTR(-ENOENT);
2062 ei->op.proc_get_link = proc_map_files_get_link;
2064 inode->i_op = &proc_pid_link_inode_operations;
2066 inode->i_mode = S_IFLNK;
2068 if (file->f_mode & FMODE_READ)
2069 inode->i_mode |= S_IRUSR;
2070 if (file->f_mode & FMODE_WRITE)
2071 inode->i_mode |= S_IWUSR;
2073 d_set_d_op(dentry, &tid_map_files_dentry_operations);
2074 d_add(dentry, inode);
2079 static struct dentry *proc_map_files_lookup(struct inode *dir,
2080 struct dentry *dentry, unsigned int flags)
2082 unsigned long vm_start, vm_end;
2083 struct vm_area_struct *vma;
2084 struct task_struct *task;
2085 struct dentry *result;
2086 struct mm_struct *mm;
2088 result = ERR_PTR(-EACCES);
2089 if (!capable(CAP_SYS_ADMIN))
2092 result = ERR_PTR(-ENOENT);
2093 task = get_proc_task(dir);
2097 result = ERR_PTR(-EACCES);
2098 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2101 result = ERR_PTR(-ENOENT);
2102 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2105 mm = get_task_mm(task);
2109 down_read(&mm->mmap_sem);
2110 vma = find_exact_vma(mm, vm_start, vm_end);
2114 result = proc_map_files_instantiate(dir, dentry, task, vma->vm_file);
2117 up_read(&mm->mmap_sem);
2120 put_task_struct(task);
2125 static const struct inode_operations proc_map_files_inode_operations = {
2126 .lookup = proc_map_files_lookup,
2127 .permission = proc_fd_permission,
2128 .setattr = proc_setattr,
2132 proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
2134 struct dentry *dentry = filp->f_path.dentry;
2135 struct inode *inode = dentry->d_inode;
2136 struct vm_area_struct *vma;
2137 struct task_struct *task;
2138 struct mm_struct *mm;
2143 if (!capable(CAP_SYS_ADMIN))
2147 task = get_proc_task(inode);
2152 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2156 switch (filp->f_pos) {
2159 if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0)
2163 ino = parent_ino(dentry);
2164 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2169 unsigned long nr_files, pos, i;
2170 struct flex_array *fa = NULL;
2171 struct map_files_info info;
2172 struct map_files_info *p;
2174 mm = get_task_mm(task);
2177 down_read(&mm->mmap_sem);
2182 * We need two passes here:
2184 * 1) Collect vmas of mapped files with mmap_sem taken
2185 * 2) Release mmap_sem and instantiate entries
2187 * otherwise we get lockdep complained, since filldir()
2188 * routine might require mmap_sem taken in might_fault().
2191 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2192 if (vma->vm_file && ++pos > filp->f_pos)
2197 fa = flex_array_alloc(sizeof(info), nr_files,
2199 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2203 flex_array_free(fa);
2204 up_read(&mm->mmap_sem);
2208 for (i = 0, vma = mm->mmap, pos = 2; vma;
2209 vma = vma->vm_next) {
2212 if (++pos <= filp->f_pos)
2215 get_file(vma->vm_file);
2216 info.file = vma->vm_file;
2217 info.len = snprintf(info.name,
2218 sizeof(info.name), "%lx-%lx",
2219 vma->vm_start, vma->vm_end);
2220 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2224 up_read(&mm->mmap_sem);
2226 for (i = 0; i < nr_files; i++) {
2227 p = flex_array_get(fa, i);
2228 ret = proc_fill_cache(filp, dirent, filldir,
2230 proc_map_files_instantiate,
2237 for (; i < nr_files; i++) {
2239 * In case of error don't forget
2240 * to put rest of file refs.
2242 p = flex_array_get(fa, i);
2246 flex_array_free(fa);
2252 put_task_struct(task);
2257 static const struct file_operations proc_map_files_operations = {
2258 .read = generic_read_dir,
2259 .readdir = proc_map_files_readdir,
2260 .llseek = default_llseek,
2263 #endif /* CONFIG_CHECKPOINT_RESTORE */
2266 * /proc/pid/fd needs a special permission handler so that a process can still
2267 * access /proc/self/fd after it has executed a setuid().
2269 static int proc_fd_permission(struct inode *inode, int mask)
2271 int rv = generic_permission(inode, mask);
2274 if (task_pid(current) == proc_pid(inode))
2280 * proc directories can do almost nothing..
2282 static const struct inode_operations proc_fd_inode_operations = {
2283 .lookup = proc_lookupfd,
2284 .permission = proc_fd_permission,
2285 .setattr = proc_setattr,
2288 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2289 struct dentry *dentry, struct task_struct *task, const void *ptr)
2291 unsigned fd = (unsigned long)ptr;
2292 struct inode *inode;
2293 struct proc_inode *ei;
2294 struct dentry *error = ERR_PTR(-ENOENT);
2296 inode = proc_pid_make_inode(dir->i_sb, task);
2301 inode->i_mode = S_IFREG | S_IRUSR;
2302 inode->i_fop = &proc_fdinfo_file_operations;
2303 d_set_d_op(dentry, &tid_fd_dentry_operations);
2304 d_add(dentry, inode);
2305 /* Close the race of the process dying before we return the dentry */
2306 if (tid_fd_revalidate(dentry, 0))
2313 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2314 struct dentry *dentry,
2317 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2320 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2322 return proc_readfd_common(filp, dirent, filldir,
2323 proc_fdinfo_instantiate);
2326 static const struct file_operations proc_fdinfo_operations = {
2327 .read = generic_read_dir,
2328 .readdir = proc_readfdinfo,
2329 .llseek = default_llseek,
2333 * proc directories can do almost nothing..
2335 static const struct inode_operations proc_fdinfo_inode_operations = {
2336 .lookup = proc_lookupfdinfo,
2337 .setattr = proc_setattr,
2341 static struct dentry *proc_pident_instantiate(struct inode *dir,
2342 struct dentry *dentry, struct task_struct *task, const void *ptr)
2344 const struct pid_entry *p = ptr;
2345 struct inode *inode;
2346 struct proc_inode *ei;
2347 struct dentry *error = ERR_PTR(-ENOENT);
2349 inode = proc_pid_make_inode(dir->i_sb, task);
2354 inode->i_mode = p->mode;
2355 if (S_ISDIR(inode->i_mode))
2356 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2358 inode->i_op = p->iop;
2360 inode->i_fop = p->fop;
2362 d_set_d_op(dentry, &pid_dentry_operations);
2363 d_add(dentry, inode);
2364 /* Close the race of the process dying before we return the dentry */
2365 if (pid_revalidate(dentry, 0))
2371 static struct dentry *proc_pident_lookup(struct inode *dir,
2372 struct dentry *dentry,
2373 const struct pid_entry *ents,
2376 struct dentry *error;
2377 struct task_struct *task = get_proc_task(dir);
2378 const struct pid_entry *p, *last;
2380 error = ERR_PTR(-ENOENT);
2386 * Yes, it does not scale. And it should not. Don't add
2387 * new entries into /proc/<tgid>/ without very good reasons.
2389 last = &ents[nents - 1];
2390 for (p = ents; p <= last; p++) {
2391 if (p->len != dentry->d_name.len)
2393 if (!memcmp(dentry->d_name.name, p->name, p->len))
2399 error = proc_pident_instantiate(dir, dentry, task, p);
2401 put_task_struct(task);
2406 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2407 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2409 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2410 proc_pident_instantiate, task, p);
2413 static int proc_pident_readdir(struct file *filp,
2414 void *dirent, filldir_t filldir,
2415 const struct pid_entry *ents, unsigned int nents)
2418 struct dentry *dentry = filp->f_path.dentry;
2419 struct inode *inode = dentry->d_inode;
2420 struct task_struct *task = get_proc_task(inode);
2421 const struct pid_entry *p, *last;
2434 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2440 ino = parent_ino(dentry);
2441 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2453 last = &ents[nents - 1];
2455 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2464 put_task_struct(task);
2469 #ifdef CONFIG_SECURITY
2470 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2471 size_t count, loff_t *ppos)
2473 struct inode * inode = file->f_path.dentry->d_inode;
2476 struct task_struct *task = get_proc_task(inode);
2481 length = security_getprocattr(task,
2482 (char*)file->f_path.dentry->d_name.name,
2484 put_task_struct(task);
2486 length = simple_read_from_buffer(buf, count, ppos, p, length);
2491 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2492 size_t count, loff_t *ppos)
2494 struct inode * inode = file->f_path.dentry->d_inode;
2497 struct task_struct *task = get_proc_task(inode);
2502 if (count > PAGE_SIZE)
2505 /* No partial writes. */
2511 page = (char*)__get_free_page(GFP_TEMPORARY);
2516 if (copy_from_user(page, buf, count))
2519 /* Guard against adverse ptrace interaction */
2520 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2524 length = security_setprocattr(task,
2525 (char*)file->f_path.dentry->d_name.name,
2526 (void*)page, count);
2527 mutex_unlock(&task->signal->cred_guard_mutex);
2529 free_page((unsigned long) page);
2531 put_task_struct(task);
2536 static const struct file_operations proc_pid_attr_operations = {
2537 .read = proc_pid_attr_read,
2538 .write = proc_pid_attr_write,
2539 .llseek = generic_file_llseek,
2542 static const struct pid_entry attr_dir_stuff[] = {
2543 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2544 REG("prev", S_IRUGO, proc_pid_attr_operations),
2545 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2546 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2547 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2548 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2551 static int proc_attr_dir_readdir(struct file * filp,
2552 void * dirent, filldir_t filldir)
2554 return proc_pident_readdir(filp,dirent,filldir,
2555 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2558 static const struct file_operations proc_attr_dir_operations = {
2559 .read = generic_read_dir,
2560 .readdir = proc_attr_dir_readdir,
2561 .llseek = default_llseek,
2564 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2565 struct dentry *dentry, unsigned int flags)
2567 return proc_pident_lookup(dir, dentry,
2568 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2571 static const struct inode_operations proc_attr_dir_inode_operations = {
2572 .lookup = proc_attr_dir_lookup,
2573 .getattr = pid_getattr,
2574 .setattr = proc_setattr,
2579 #ifdef CONFIG_ELF_CORE
2580 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2581 size_t count, loff_t *ppos)
2583 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2584 struct mm_struct *mm;
2585 char buffer[PROC_NUMBUF];
2593 mm = get_task_mm(task);
2595 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2596 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2597 MMF_DUMP_FILTER_SHIFT));
2599 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2602 put_task_struct(task);
2607 static ssize_t proc_coredump_filter_write(struct file *file,
2608 const char __user *buf,
2612 struct task_struct *task;
2613 struct mm_struct *mm;
2614 char buffer[PROC_NUMBUF], *end;
2621 memset(buffer, 0, sizeof(buffer));
2622 if (count > sizeof(buffer) - 1)
2623 count = sizeof(buffer) - 1;
2624 if (copy_from_user(buffer, buf, count))
2628 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2631 if (end - buffer == 0)
2635 task = get_proc_task(file->f_dentry->d_inode);
2640 mm = get_task_mm(task);
2644 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2646 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2648 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2653 put_task_struct(task);
2658 static const struct file_operations proc_coredump_filter_operations = {
2659 .read = proc_coredump_filter_read,
2660 .write = proc_coredump_filter_write,
2661 .llseek = generic_file_llseek,
2668 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2671 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2672 pid_t tgid = task_tgid_nr_ns(current, ns);
2673 char tmp[PROC_NUMBUF];
2676 sprintf(tmp, "%d", tgid);
2677 return vfs_readlink(dentry,buffer,buflen,tmp);
2680 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2682 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2683 pid_t tgid = task_tgid_nr_ns(current, ns);
2684 char *name = ERR_PTR(-ENOENT);
2688 name = ERR_PTR(-ENOMEM);
2690 sprintf(name, "%d", tgid);
2692 nd_set_link(nd, name);
2696 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2699 char *s = nd_get_link(nd);
2704 static const struct inode_operations proc_self_inode_operations = {
2705 .readlink = proc_self_readlink,
2706 .follow_link = proc_self_follow_link,
2707 .put_link = proc_self_put_link,
2713 * These are the directory entries in the root directory of /proc
2714 * that properly belong to the /proc filesystem, as they describe
2715 * describe something that is process related.
2717 static const struct pid_entry proc_base_stuff[] = {
2718 NOD("self", S_IFLNK|S_IRWXUGO,
2719 &proc_self_inode_operations, NULL, {}),
2722 static struct dentry *proc_base_instantiate(struct inode *dir,
2723 struct dentry *dentry, struct task_struct *task, const void *ptr)
2725 const struct pid_entry *p = ptr;
2726 struct inode *inode;
2727 struct proc_inode *ei;
2728 struct dentry *error;
2730 /* Allocate the inode */
2731 error = ERR_PTR(-ENOMEM);
2732 inode = new_inode(dir->i_sb);
2736 /* Initialize the inode */
2738 inode->i_ino = get_next_ino();
2739 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2742 * grab the reference to the task.
2744 ei->pid = get_task_pid(task, PIDTYPE_PID);
2748 inode->i_mode = p->mode;
2749 if (S_ISDIR(inode->i_mode))
2750 set_nlink(inode, 2);
2751 if (S_ISLNK(inode->i_mode))
2754 inode->i_op = p->iop;
2756 inode->i_fop = p->fop;
2758 d_add(dentry, inode);
2767 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2769 struct dentry *error;
2770 struct task_struct *task = get_proc_task(dir);
2771 const struct pid_entry *p, *last;
2773 error = ERR_PTR(-ENOENT);
2778 /* Lookup the directory entry */
2779 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2780 for (p = proc_base_stuff; p <= last; p++) {
2781 if (p->len != dentry->d_name.len)
2783 if (!memcmp(dentry->d_name.name, p->name, p->len))
2789 error = proc_base_instantiate(dir, dentry, task, p);
2792 put_task_struct(task);
2797 static int proc_base_fill_cache(struct file *filp, void *dirent,
2798 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2800 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2801 proc_base_instantiate, task, p);
2804 #ifdef CONFIG_TASK_IO_ACCOUNTING
2805 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2807 struct task_io_accounting acct = task->ioac;
2808 unsigned long flags;
2811 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2815 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2820 if (whole && lock_task_sighand(task, &flags)) {
2821 struct task_struct *t = task;
2823 task_io_accounting_add(&acct, &task->signal->ioac);
2824 while_each_thread(task, t)
2825 task_io_accounting_add(&acct, &t->ioac);
2827 unlock_task_sighand(task, &flags);
2829 result = sprintf(buffer,
2834 "read_bytes: %llu\n"
2835 "write_bytes: %llu\n"
2836 "cancelled_write_bytes: %llu\n",
2837 (unsigned long long)acct.rchar,
2838 (unsigned long long)acct.wchar,
2839 (unsigned long long)acct.syscr,
2840 (unsigned long long)acct.syscw,
2841 (unsigned long long)acct.read_bytes,
2842 (unsigned long long)acct.write_bytes,
2843 (unsigned long long)acct.cancelled_write_bytes);
2845 mutex_unlock(&task->signal->cred_guard_mutex);
2849 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2851 return do_io_accounting(task, buffer, 0);
2854 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2856 return do_io_accounting(task, buffer, 1);
2858 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2860 #ifdef CONFIG_USER_NS
2861 static int proc_id_map_open(struct inode *inode, struct file *file,
2862 struct seq_operations *seq_ops)
2864 struct user_namespace *ns = NULL;
2865 struct task_struct *task;
2866 struct seq_file *seq;
2869 task = get_proc_task(inode);
2872 ns = get_user_ns(task_cred_xxx(task, user_ns));
2874 put_task_struct(task);
2879 ret = seq_open(file, seq_ops);
2883 seq = file->private_data;
2893 static int proc_id_map_release(struct inode *inode, struct file *file)
2895 struct seq_file *seq = file->private_data;
2896 struct user_namespace *ns = seq->private;
2898 return seq_release(inode, file);
2901 static int proc_uid_map_open(struct inode *inode, struct file *file)
2903 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2906 static int proc_gid_map_open(struct inode *inode, struct file *file)
2908 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2911 static const struct file_operations proc_uid_map_operations = {
2912 .open = proc_uid_map_open,
2913 .write = proc_uid_map_write,
2915 .llseek = seq_lseek,
2916 .release = proc_id_map_release,
2919 static const struct file_operations proc_gid_map_operations = {
2920 .open = proc_gid_map_open,
2921 .write = proc_gid_map_write,
2923 .llseek = seq_lseek,
2924 .release = proc_id_map_release,
2926 #endif /* CONFIG_USER_NS */
2928 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2929 struct pid *pid, struct task_struct *task)
2931 int err = lock_trace(task);
2933 seq_printf(m, "%08x\n", task->personality);
2942 static const struct file_operations proc_task_operations;
2943 static const struct inode_operations proc_task_inode_operations;
2945 static const struct pid_entry tgid_base_stuff[] = {
2946 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2947 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2948 #ifdef CONFIG_CHECKPOINT_RESTORE
2949 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
2951 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2952 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2954 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2956 REG("environ", S_IRUSR, proc_environ_operations),
2957 INF("auxv", S_IRUSR, proc_pid_auxv),
2958 ONE("status", S_IRUGO, proc_pid_status),
2959 ONE("personality", S_IRUGO, proc_pid_personality),
2960 INF("limits", S_IRUGO, proc_pid_limits),
2961 #ifdef CONFIG_SCHED_DEBUG
2962 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2964 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2965 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2966 INF("syscall", S_IRUGO, proc_pid_syscall),
2968 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2969 ONE("stat", S_IRUGO, proc_tgid_stat),
2970 ONE("statm", S_IRUGO, proc_pid_statm),
2971 REG("maps", S_IRUGO, proc_pid_maps_operations),
2973 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
2975 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2976 LNK("cwd", proc_cwd_link),
2977 LNK("root", proc_root_link),
2978 LNK("exe", proc_exe_link),
2979 REG("mounts", S_IRUGO, proc_mounts_operations),
2980 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2981 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2982 #ifdef CONFIG_PROC_PAGE_MONITOR
2983 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2984 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
2985 REG("pagemap", S_IRUGO, proc_pagemap_operations),
2987 #ifdef CONFIG_SECURITY
2988 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2990 #ifdef CONFIG_KALLSYMS
2991 INF("wchan", S_IRUGO, proc_pid_wchan),
2993 #ifdef CONFIG_STACKTRACE
2994 ONE("stack", S_IRUGO, proc_pid_stack),
2996 #ifdef CONFIG_SCHEDSTATS
2997 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2999 #ifdef CONFIG_LATENCYTOP
3000 REG("latency", S_IRUGO, proc_lstats_operations),
3002 #ifdef CONFIG_PROC_PID_CPUSET
3003 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3005 #ifdef CONFIG_CGROUPS
3006 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3008 INF("oom_score", S_IRUGO, proc_oom_score),
3009 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3010 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3011 #ifdef CONFIG_AUDITSYSCALL
3012 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3013 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3015 #ifdef CONFIG_FAULT_INJECTION
3016 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3018 #ifdef CONFIG_ELF_CORE
3019 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3021 #ifdef CONFIG_TASK_IO_ACCOUNTING
3022 INF("io", S_IRUSR, proc_tgid_io_accounting),
3024 #ifdef CONFIG_HARDWALL
3025 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3027 #ifdef CONFIG_USER_NS
3028 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3029 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3033 static int proc_tgid_base_readdir(struct file * filp,
3034 void * dirent, filldir_t filldir)
3036 return proc_pident_readdir(filp,dirent,filldir,
3037 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
3040 static const struct file_operations proc_tgid_base_operations = {
3041 .read = generic_read_dir,
3042 .readdir = proc_tgid_base_readdir,
3043 .llseek = default_llseek,
3046 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3048 return proc_pident_lookup(dir, dentry,
3049 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3052 static const struct inode_operations proc_tgid_base_inode_operations = {
3053 .lookup = proc_tgid_base_lookup,
3054 .getattr = pid_getattr,
3055 .setattr = proc_setattr,
3056 .permission = proc_pid_permission,
3059 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
3061 struct dentry *dentry, *leader, *dir;
3062 char buf[PROC_NUMBUF];
3066 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3067 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
3069 shrink_dcache_parent(dentry);
3075 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
3076 leader = d_hash_and_lookup(mnt->mnt_root, &name);
3081 name.len = strlen(name.name);
3082 dir = d_hash_and_lookup(leader, &name);
3084 goto out_put_leader;
3087 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3088 dentry = d_hash_and_lookup(dir, &name);
3090 shrink_dcache_parent(dentry);
3103 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3104 * @task: task that should be flushed.
3106 * When flushing dentries from proc, one needs to flush them from global
3107 * proc (proc_mnt) and from all the namespaces' procs this task was seen
3108 * in. This call is supposed to do all of this job.
3110 * Looks in the dcache for
3112 * /proc/@tgid/task/@pid
3113 * if either directory is present flushes it and all of it'ts children
3116 * It is safe and reasonable to cache /proc entries for a task until
3117 * that task exits. After that they just clog up the dcache with
3118 * useless entries, possibly causing useful dcache entries to be
3119 * flushed instead. This routine is proved to flush those useless
3120 * dcache entries at process exit time.
3122 * NOTE: This routine is just an optimization so it does not guarantee
3123 * that no dcache entries will exist at process exit time it
3124 * just makes it very unlikely that any will persist.
3127 void proc_flush_task(struct task_struct *task)
3130 struct pid *pid, *tgid;
3133 pid = task_pid(task);
3134 tgid = task_tgid(task);
3136 for (i = 0; i <= pid->level; i++) {
3137 upid = &pid->numbers[i];
3138 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
3139 tgid->numbers[i].nr);
3142 upid = &pid->numbers[pid->level];
3144 pid_ns_release_proc(upid->ns);
3147 static struct dentry *proc_pid_instantiate(struct inode *dir,
3148 struct dentry * dentry,
3149 struct task_struct *task, const void *ptr)
3151 struct dentry *error = ERR_PTR(-ENOENT);
3152 struct inode *inode;
3154 inode = proc_pid_make_inode(dir->i_sb, task);
3158 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3159 inode->i_op = &proc_tgid_base_inode_operations;
3160 inode->i_fop = &proc_tgid_base_operations;
3161 inode->i_flags|=S_IMMUTABLE;
3163 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
3164 ARRAY_SIZE(tgid_base_stuff)));
3166 d_set_d_op(dentry, &pid_dentry_operations);
3168 d_add(dentry, inode);
3169 /* Close the race of the process dying before we return the dentry */
3170 if (pid_revalidate(dentry, 0))
3176 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3178 struct dentry *result;
3179 struct task_struct *task;
3181 struct pid_namespace *ns;
3183 result = proc_base_lookup(dir, dentry);
3184 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3187 tgid = name_to_int(dentry);
3191 ns = dentry->d_sb->s_fs_info;
3193 task = find_task_by_pid_ns(tgid, ns);
3195 get_task_struct(task);
3200 result = proc_pid_instantiate(dir, dentry, task, NULL);
3201 put_task_struct(task);
3207 * Find the first task with tgid >= tgid
3212 struct task_struct *task;
3214 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3219 put_task_struct(iter.task);
3223 pid = find_ge_pid(iter.tgid, ns);
3225 iter.tgid = pid_nr_ns(pid, ns);
3226 iter.task = pid_task(pid, PIDTYPE_PID);
3227 /* What we to know is if the pid we have find is the
3228 * pid of a thread_group_leader. Testing for task
3229 * being a thread_group_leader is the obvious thing
3230 * todo but there is a window when it fails, due to
3231 * the pid transfer logic in de_thread.
3233 * So we perform the straight forward test of seeing
3234 * if the pid we have found is the pid of a thread
3235 * group leader, and don't worry if the task we have
3236 * found doesn't happen to be a thread group leader.
3237 * As we don't care in the case of readdir.
3239 if (!iter.task || !has_group_leader_pid(iter.task)) {
3243 get_task_struct(iter.task);
3249 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
3251 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3252 struct tgid_iter iter)
3254 char name[PROC_NUMBUF];
3255 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
3256 return proc_fill_cache(filp, dirent, filldir, name, len,
3257 proc_pid_instantiate, iter.task, NULL);
3260 static int fake_filldir(void *buf, const char *name, int namelen,
3261 loff_t offset, u64 ino, unsigned d_type)
3266 /* for the /proc/ directory itself, after non-process stuff has been done */
3267 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3270 struct task_struct *reaper;
3271 struct tgid_iter iter;
3272 struct pid_namespace *ns;
3273 filldir_t __filldir;
3275 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3277 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3279 reaper = get_proc_task(filp->f_path.dentry->d_inode);
3283 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
3284 const struct pid_entry *p = &proc_base_stuff[nr];
3285 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
3289 ns = filp->f_dentry->d_sb->s_fs_info;
3291 iter.tgid = filp->f_pos - TGID_OFFSET;
3292 for (iter = next_tgid(ns, iter);
3294 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3295 if (has_pid_permissions(ns, iter.task, 2))
3296 __filldir = filldir;
3298 __filldir = fake_filldir;
3300 filp->f_pos = iter.tgid + TGID_OFFSET;
3301 if (proc_pid_fill_cache(filp, dirent, __filldir, iter) < 0) {
3302 put_task_struct(iter.task);
3306 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3308 put_task_struct(reaper);
3316 static const struct pid_entry tid_base_stuff[] = {
3317 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3318 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3319 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3320 REG("environ", S_IRUSR, proc_environ_operations),
3321 INF("auxv", S_IRUSR, proc_pid_auxv),
3322 ONE("status", S_IRUGO, proc_pid_status),
3323 ONE("personality", S_IRUGO, proc_pid_personality),
3324 INF("limits", S_IRUGO, proc_pid_limits),
3325 #ifdef CONFIG_SCHED_DEBUG
3326 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3328 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3329 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3330 INF("syscall", S_IRUGO, proc_pid_syscall),
3332 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3333 ONE("stat", S_IRUGO, proc_tid_stat),
3334 ONE("statm", S_IRUGO, proc_pid_statm),
3335 REG("maps", S_IRUGO, proc_tid_maps_operations),
3336 #ifdef CONFIG_CHECKPOINT_RESTORE
3337 REG("children", S_IRUGO, proc_tid_children_operations),
3340 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
3342 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3343 LNK("cwd", proc_cwd_link),
3344 LNK("root", proc_root_link),
3345 LNK("exe", proc_exe_link),
3346 REG("mounts", S_IRUGO, proc_mounts_operations),
3347 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3348 #ifdef CONFIG_PROC_PAGE_MONITOR
3349 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3350 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
3351 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3353 #ifdef CONFIG_SECURITY
3354 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3356 #ifdef CONFIG_KALLSYMS
3357 INF("wchan", S_IRUGO, proc_pid_wchan),
3359 #ifdef CONFIG_STACKTRACE
3360 ONE("stack", S_IRUGO, proc_pid_stack),
3362 #ifdef CONFIG_SCHEDSTATS
3363 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3365 #ifdef CONFIG_LATENCYTOP
3366 REG("latency", S_IRUGO, proc_lstats_operations),
3368 #ifdef CONFIG_PROC_PID_CPUSET
3369 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3371 #ifdef CONFIG_CGROUPS
3372 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3374 INF("oom_score", S_IRUGO, proc_oom_score),
3375 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3376 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3377 #ifdef CONFIG_AUDITSYSCALL
3378 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3379 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3381 #ifdef CONFIG_FAULT_INJECTION
3382 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3384 #ifdef CONFIG_TASK_IO_ACCOUNTING
3385 INF("io", S_IRUSR, proc_tid_io_accounting),
3387 #ifdef CONFIG_HARDWALL
3388 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3390 #ifdef CONFIG_USER_NS
3391 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3392 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3396 static int proc_tid_base_readdir(struct file * filp,
3397 void * dirent, filldir_t filldir)
3399 return proc_pident_readdir(filp,dirent,filldir,
3400 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3403 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3405 return proc_pident_lookup(dir, dentry,
3406 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3409 static const struct file_operations proc_tid_base_operations = {
3410 .read = generic_read_dir,
3411 .readdir = proc_tid_base_readdir,
3412 .llseek = default_llseek,
3415 static const struct inode_operations proc_tid_base_inode_operations = {
3416 .lookup = proc_tid_base_lookup,
3417 .getattr = pid_getattr,
3418 .setattr = proc_setattr,
3421 static struct dentry *proc_task_instantiate(struct inode *dir,
3422 struct dentry *dentry, struct task_struct *task, const void *ptr)
3424 struct dentry *error = ERR_PTR(-ENOENT);
3425 struct inode *inode;
3426 inode = proc_pid_make_inode(dir->i_sb, task);
3430 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3431 inode->i_op = &proc_tid_base_inode_operations;
3432 inode->i_fop = &proc_tid_base_operations;
3433 inode->i_flags|=S_IMMUTABLE;
3435 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3436 ARRAY_SIZE(tid_base_stuff)));
3438 d_set_d_op(dentry, &pid_dentry_operations);
3440 d_add(dentry, inode);
3441 /* Close the race of the process dying before we return the dentry */
3442 if (pid_revalidate(dentry, 0))
3448 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3450 struct dentry *result = ERR_PTR(-ENOENT);
3451 struct task_struct *task;
3452 struct task_struct *leader = get_proc_task(dir);
3454 struct pid_namespace *ns;
3459 tid = name_to_int(dentry);
3463 ns = dentry->d_sb->s_fs_info;
3465 task = find_task_by_pid_ns(tid, ns);
3467 get_task_struct(task);
3471 if (!same_thread_group(leader, task))
3474 result = proc_task_instantiate(dir, dentry, task, NULL);
3476 put_task_struct(task);
3478 put_task_struct(leader);
3484 * Find the first tid of a thread group to return to user space.
3486 * Usually this is just the thread group leader, but if the users
3487 * buffer was too small or there was a seek into the middle of the
3488 * directory we have more work todo.
3490 * In the case of a short read we start with find_task_by_pid.
3492 * In the case of a seek we start with the leader and walk nr
3495 static struct task_struct *first_tid(struct task_struct *leader,
3496 int tid, int nr, struct pid_namespace *ns)
3498 struct task_struct *pos;
3501 /* Attempt to start with the pid of a thread */
3502 if (tid && (nr > 0)) {
3503 pos = find_task_by_pid_ns(tid, ns);
3504 if (pos && (pos->group_leader == leader))
3508 /* If nr exceeds the number of threads there is nothing todo */
3510 if (nr && nr >= get_nr_threads(leader))
3513 /* If we haven't found our starting place yet start
3514 * with the leader and walk nr threads forward.
3516 for (pos = leader; nr > 0; --nr) {
3517 pos = next_thread(pos);
3518 if (pos == leader) {
3524 get_task_struct(pos);
3531 * Find the next thread in the thread list.
3532 * Return NULL if there is an error or no next thread.
3534 * The reference to the input task_struct is released.
3536 static struct task_struct *next_tid(struct task_struct *start)
3538 struct task_struct *pos = NULL;
3540 if (pid_alive(start)) {
3541 pos = next_thread(start);
3542 if (thread_group_leader(pos))
3545 get_task_struct(pos);
3548 put_task_struct(start);
3552 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3553 struct task_struct *task, int tid)
3555 char name[PROC_NUMBUF];
3556 int len = snprintf(name, sizeof(name), "%d", tid);
3557 return proc_fill_cache(filp, dirent, filldir, name, len,
3558 proc_task_instantiate, task, NULL);
3561 /* for the /proc/TGID/task/ directories */
3562 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3564 struct dentry *dentry = filp->f_path.dentry;
3565 struct inode *inode = dentry->d_inode;
3566 struct task_struct *leader = NULL;
3567 struct task_struct *task;
3568 int retval = -ENOENT;
3571 struct pid_namespace *ns;
3573 task = get_proc_task(inode);
3577 if (pid_alive(task)) {
3578 leader = task->group_leader;
3579 get_task_struct(leader);
3582 put_task_struct(task);
3587 switch ((unsigned long)filp->f_pos) {
3590 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3595 ino = parent_ino(dentry);
3596 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3602 /* f_version caches the tgid value that the last readdir call couldn't
3603 * return. lseek aka telldir automagically resets f_version to 0.
3605 ns = filp->f_dentry->d_sb->s_fs_info;
3606 tid = (int)filp->f_version;
3607 filp->f_version = 0;
3608 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3610 task = next_tid(task), filp->f_pos++) {
3611 tid = task_pid_nr_ns(task, ns);
3612 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3613 /* returning this tgid failed, save it as the first
3614 * pid for the next readir call */
3615 filp->f_version = (u64)tid;
3616 put_task_struct(task);
3621 put_task_struct(leader);
3626 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3628 struct inode *inode = dentry->d_inode;
3629 struct task_struct *p = get_proc_task(inode);
3630 generic_fillattr(inode, stat);
3633 stat->nlink += get_nr_threads(p);
3640 static const struct inode_operations proc_task_inode_operations = {
3641 .lookup = proc_task_lookup,
3642 .getattr = proc_task_getattr,
3643 .setattr = proc_setattr,
3644 .permission = proc_pid_permission,
3647 static const struct file_operations proc_task_operations = {
3648 .read = generic_read_dir,
3649 .readdir = proc_task_readdir,
3650 .llseek = default_llseek,