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 #ifdef CONFIG_HARDWALL
87 #include <asm/hardwall.h>
89 #include <trace/events/oom.h>
93 * Implementing inode permission operations in /proc is almost
94 * certainly an error. Permission checks need to happen during
95 * each system call not at open time. The reason is that most of
96 * what we wish to check for permissions in /proc varies at runtime.
98 * The classic example of a problem is opening file descriptors
99 * in /proc for a task before it execs a suid executable.
106 const struct inode_operations *iop;
107 const struct file_operations *fop;
111 #define NOD(NAME, MODE, IOP, FOP, OP) { \
113 .len = sizeof(NAME) - 1, \
120 #define DIR(NAME, MODE, iops, fops) \
121 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
122 #define LNK(NAME, get_link) \
123 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
124 &proc_pid_link_inode_operations, NULL, \
125 { .proc_get_link = get_link } )
126 #define REG(NAME, MODE, fops) \
127 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
128 #define INF(NAME, MODE, read) \
129 NOD(NAME, (S_IFREG|(MODE)), \
130 NULL, &proc_info_file_operations, \
131 { .proc_read = read } )
132 #define ONE(NAME, MODE, show) \
133 NOD(NAME, (S_IFREG|(MODE)), \
134 NULL, &proc_single_file_operations, \
135 { .proc_show = show } )
138 * Count the number of hardlinks for the pid_entry table, excluding the .
141 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
148 for (i = 0; i < n; ++i) {
149 if (S_ISDIR(entries[i].mode))
156 static int get_task_root(struct task_struct *task, struct path *root)
158 int result = -ENOENT;
162 get_fs_root(task->fs, root);
169 static int proc_cwd_link(struct inode *inode, struct path *path)
171 struct task_struct *task = get_proc_task(inode);
172 int result = -ENOENT;
177 get_fs_pwd(task->fs, path);
181 put_task_struct(task);
186 static int proc_root_link(struct inode *inode, struct path *path)
188 struct task_struct *task = get_proc_task(inode);
189 int result = -ENOENT;
192 result = get_task_root(task, path);
193 put_task_struct(task);
198 static struct mm_struct *__check_mem_permission(struct task_struct *task)
200 struct mm_struct *mm;
202 mm = get_task_mm(task);
204 return ERR_PTR(-EINVAL);
207 * A task can always look at itself, in case it chooses
208 * to use system calls instead of load instructions.
214 * If current is actively ptrace'ing, and would also be
215 * permitted to freshly attach with ptrace now, permit it.
217 if (task_is_stopped_or_traced(task)) {
220 match = (ptrace_parent(task) == current);
222 if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH))
227 * No one else is allowed.
230 return ERR_PTR(-EPERM);
234 * If current may access user memory in @task return a reference to the
235 * corresponding mm, otherwise ERR_PTR.
237 static struct mm_struct *check_mem_permission(struct task_struct *task)
239 struct mm_struct *mm;
243 * Avoid racing if task exec's as we might get a new mm but validate
244 * against old credentials.
246 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
250 mm = __check_mem_permission(task);
251 mutex_unlock(&task->signal->cred_guard_mutex);
256 struct mm_struct *mm_for_maps(struct task_struct *task)
258 struct mm_struct *mm;
261 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
265 mm = get_task_mm(task);
266 if (mm && mm != current->mm &&
267 !ptrace_may_access(task, PTRACE_MODE_READ)) {
269 mm = ERR_PTR(-EACCES);
271 mutex_unlock(&task->signal->cred_guard_mutex);
276 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
280 struct mm_struct *mm = get_task_mm(task);
284 goto out_mm; /* Shh! No looking before we're done */
286 len = mm->arg_end - mm->arg_start;
291 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
293 // If the nul at the end of args has been overwritten, then
294 // assume application is using setproctitle(3).
295 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
296 len = strnlen(buffer, res);
300 len = mm->env_end - mm->env_start;
301 if (len > PAGE_SIZE - res)
302 len = PAGE_SIZE - res;
303 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
304 res = strnlen(buffer, res);
313 static int proc_pid_auxv(struct task_struct *task, char *buffer)
315 struct mm_struct *mm = mm_for_maps(task);
316 int res = PTR_ERR(mm);
317 if (mm && !IS_ERR(mm)) {
318 unsigned int nwords = 0;
321 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
322 res = nwords * sizeof(mm->saved_auxv[0]);
325 memcpy(buffer, mm->saved_auxv, res);
332 #ifdef CONFIG_KALLSYMS
334 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
335 * Returns the resolved symbol. If that fails, simply return the address.
337 static int proc_pid_wchan(struct task_struct *task, char *buffer)
340 char symname[KSYM_NAME_LEN];
342 wchan = get_wchan(task);
344 if (lookup_symbol_name(wchan, symname) < 0)
345 if (!ptrace_may_access(task, PTRACE_MODE_READ))
348 return sprintf(buffer, "%lu", wchan);
350 return sprintf(buffer, "%s", symname);
352 #endif /* CONFIG_KALLSYMS */
354 static int lock_trace(struct task_struct *task)
356 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
359 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
360 mutex_unlock(&task->signal->cred_guard_mutex);
366 static void unlock_trace(struct task_struct *task)
368 mutex_unlock(&task->signal->cred_guard_mutex);
371 #ifdef CONFIG_STACKTRACE
373 #define MAX_STACK_TRACE_DEPTH 64
375 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
376 struct pid *pid, struct task_struct *task)
378 struct stack_trace trace;
379 unsigned long *entries;
383 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
387 trace.nr_entries = 0;
388 trace.max_entries = MAX_STACK_TRACE_DEPTH;
389 trace.entries = entries;
392 err = lock_trace(task);
394 save_stack_trace_tsk(task, &trace);
396 for (i = 0; i < trace.nr_entries; i++) {
397 seq_printf(m, "[<%pK>] %pS\n",
398 (void *)entries[i], (void *)entries[i]);
408 #ifdef CONFIG_SCHEDSTATS
410 * Provides /proc/PID/schedstat
412 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
414 return sprintf(buffer, "%llu %llu %lu\n",
415 (unsigned long long)task->se.sum_exec_runtime,
416 (unsigned long long)task->sched_info.run_delay,
417 task->sched_info.pcount);
421 #ifdef CONFIG_LATENCYTOP
422 static int lstats_show_proc(struct seq_file *m, void *v)
425 struct inode *inode = m->private;
426 struct task_struct *task = get_proc_task(inode);
430 seq_puts(m, "Latency Top version : v0.1\n");
431 for (i = 0; i < 32; i++) {
432 struct latency_record *lr = &task->latency_record[i];
433 if (lr->backtrace[0]) {
435 seq_printf(m, "%i %li %li",
436 lr->count, lr->time, lr->max);
437 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
438 unsigned long bt = lr->backtrace[q];
443 seq_printf(m, " %ps", (void *)bt);
449 put_task_struct(task);
453 static int lstats_open(struct inode *inode, struct file *file)
455 return single_open(file, lstats_show_proc, inode);
458 static ssize_t lstats_write(struct file *file, const char __user *buf,
459 size_t count, loff_t *offs)
461 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
465 clear_all_latency_tracing(task);
466 put_task_struct(task);
471 static const struct file_operations proc_lstats_operations = {
474 .write = lstats_write,
476 .release = single_release,
481 static int proc_oom_score(struct task_struct *task, char *buffer)
483 unsigned long points = 0;
485 read_lock(&tasklist_lock);
487 points = oom_badness(task, NULL, NULL,
488 totalram_pages + total_swap_pages);
489 read_unlock(&tasklist_lock);
490 return sprintf(buffer, "%lu\n", points);
498 static const struct limit_names lnames[RLIM_NLIMITS] = {
499 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
500 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
501 [RLIMIT_DATA] = {"Max data size", "bytes"},
502 [RLIMIT_STACK] = {"Max stack size", "bytes"},
503 [RLIMIT_CORE] = {"Max core file size", "bytes"},
504 [RLIMIT_RSS] = {"Max resident set", "bytes"},
505 [RLIMIT_NPROC] = {"Max processes", "processes"},
506 [RLIMIT_NOFILE] = {"Max open files", "files"},
507 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
508 [RLIMIT_AS] = {"Max address space", "bytes"},
509 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
510 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
511 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
512 [RLIMIT_NICE] = {"Max nice priority", NULL},
513 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
514 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
517 /* Display limits for a process */
518 static int proc_pid_limits(struct task_struct *task, char *buffer)
523 char *bufptr = buffer;
525 struct rlimit rlim[RLIM_NLIMITS];
527 if (!lock_task_sighand(task, &flags))
529 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
530 unlock_task_sighand(task, &flags);
533 * print the file header
535 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
536 "Limit", "Soft Limit", "Hard Limit", "Units");
538 for (i = 0; i < RLIM_NLIMITS; i++) {
539 if (rlim[i].rlim_cur == RLIM_INFINITY)
540 count += sprintf(&bufptr[count], "%-25s %-20s ",
541 lnames[i].name, "unlimited");
543 count += sprintf(&bufptr[count], "%-25s %-20lu ",
544 lnames[i].name, rlim[i].rlim_cur);
546 if (rlim[i].rlim_max == RLIM_INFINITY)
547 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
549 count += sprintf(&bufptr[count], "%-20lu ",
553 count += sprintf(&bufptr[count], "%-10s\n",
556 count += sprintf(&bufptr[count], "\n");
562 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
563 static int proc_pid_syscall(struct task_struct *task, char *buffer)
566 unsigned long args[6], sp, pc;
567 int res = lock_trace(task);
571 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
572 res = sprintf(buffer, "running\n");
574 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
576 res = sprintf(buffer,
577 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
579 args[0], args[1], args[2], args[3], args[4], args[5],
584 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
586 /************************************************************************/
587 /* Here the fs part begins */
588 /************************************************************************/
590 /* permission checks */
591 static int proc_fd_access_allowed(struct inode *inode)
593 struct task_struct *task;
595 /* Allow access to a task's file descriptors if it is us or we
596 * may use ptrace attach to the process and find out that
599 task = get_proc_task(inode);
601 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
602 put_task_struct(task);
607 int proc_setattr(struct dentry *dentry, struct iattr *attr)
610 struct inode *inode = dentry->d_inode;
612 if (attr->ia_valid & ATTR_MODE)
615 error = inode_change_ok(inode, attr);
619 if ((attr->ia_valid & ATTR_SIZE) &&
620 attr->ia_size != i_size_read(inode)) {
621 error = vmtruncate(inode, attr->ia_size);
626 setattr_copy(inode, attr);
627 mark_inode_dirty(inode);
631 static const struct inode_operations proc_def_inode_operations = {
632 .setattr = proc_setattr,
635 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
637 static ssize_t proc_info_read(struct file * file, char __user * buf,
638 size_t count, loff_t *ppos)
640 struct inode * inode = file->f_path.dentry->d_inode;
643 struct task_struct *task = get_proc_task(inode);
649 if (count > PROC_BLOCK_SIZE)
650 count = PROC_BLOCK_SIZE;
653 if (!(page = __get_free_page(GFP_TEMPORARY)))
656 length = PROC_I(inode)->op.proc_read(task, (char*)page);
659 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
662 put_task_struct(task);
667 static const struct file_operations proc_info_file_operations = {
668 .read = proc_info_read,
669 .llseek = generic_file_llseek,
672 static int proc_single_show(struct seq_file *m, void *v)
674 struct inode *inode = m->private;
675 struct pid_namespace *ns;
677 struct task_struct *task;
680 ns = inode->i_sb->s_fs_info;
681 pid = proc_pid(inode);
682 task = get_pid_task(pid, PIDTYPE_PID);
686 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
688 put_task_struct(task);
692 static int proc_single_open(struct inode *inode, struct file *filp)
694 return single_open(filp, proc_single_show, inode);
697 static const struct file_operations proc_single_file_operations = {
698 .open = proc_single_open,
701 .release = single_release,
704 static int mem_open(struct inode* inode, struct file* file)
706 file->private_data = (void*)((long)current->self_exec_id);
707 /* OK to pass negative loff_t, we can catch out-of-range */
708 file->f_mode |= FMODE_UNSIGNED_OFFSET;
712 static ssize_t mem_read(struct file * file, char __user * buf,
713 size_t count, loff_t *ppos)
715 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
717 unsigned long src = *ppos;
719 struct mm_struct *mm;
725 page = (char *)__get_free_page(GFP_TEMPORARY);
729 mm = check_mem_permission(task);
736 if (file->private_data != (void*)((long)current->self_exec_id))
742 int this_len, retval;
744 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
745 retval = access_remote_vm(mm, src, page, this_len, 0);
752 if (copy_to_user(buf, page, retval)) {
767 free_page((unsigned long) page);
769 put_task_struct(task);
774 static ssize_t mem_write(struct file * file, const char __user *buf,
775 size_t count, loff_t *ppos)
779 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
780 unsigned long dst = *ppos;
781 struct mm_struct *mm;
788 page = (char *)__get_free_page(GFP_TEMPORARY);
792 mm = check_mem_permission(task);
793 copied = PTR_ERR(mm);
798 if (file->private_data != (void *)((long)current->self_exec_id))
803 int this_len, retval;
805 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
806 if (copy_from_user(page, buf, this_len)) {
810 retval = access_remote_vm(mm, dst, page, this_len, 1);
826 free_page((unsigned long) page);
828 put_task_struct(task);
833 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
837 file->f_pos = offset;
840 file->f_pos += offset;
845 force_successful_syscall_return();
849 static const struct file_operations proc_mem_operations = {
856 static ssize_t environ_read(struct file *file, char __user *buf,
857 size_t count, loff_t *ppos)
859 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
861 unsigned long src = *ppos;
863 struct mm_struct *mm;
869 page = (char *)__get_free_page(GFP_TEMPORARY);
874 mm = mm_for_maps(task);
876 if (!mm || IS_ERR(mm))
881 int this_len, retval, max_len;
883 this_len = mm->env_end - (mm->env_start + src);
888 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
889 this_len = (this_len > max_len) ? max_len : this_len;
891 retval = access_process_vm(task, (mm->env_start + src),
899 if (copy_to_user(buf, page, retval)) {
913 free_page((unsigned long) page);
915 put_task_struct(task);
920 static const struct file_operations proc_environ_operations = {
921 .read = environ_read,
922 .llseek = generic_file_llseek,
925 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
926 size_t count, loff_t *ppos)
928 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
929 char buffer[PROC_NUMBUF];
931 int oom_adjust = OOM_DISABLE;
937 if (lock_task_sighand(task, &flags)) {
938 oom_adjust = task->signal->oom_adj;
939 unlock_task_sighand(task, &flags);
942 put_task_struct(task);
944 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
946 return simple_read_from_buffer(buf, count, ppos, buffer, len);
949 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
950 size_t count, loff_t *ppos)
952 struct task_struct *task;
953 char buffer[PROC_NUMBUF];
958 memset(buffer, 0, sizeof(buffer));
959 if (count > sizeof(buffer) - 1)
960 count = sizeof(buffer) - 1;
961 if (copy_from_user(buffer, buf, count)) {
966 err = kstrtoint(strstrip(buffer), 0, &oom_adjust);
969 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
970 oom_adjust != OOM_DISABLE) {
975 task = get_proc_task(file->f_path.dentry->d_inode);
987 if (!lock_task_sighand(task, &flags)) {
992 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
998 * Warn that /proc/pid/oom_adj is deprecated, see
999 * Documentation/feature-removal-schedule.txt.
1001 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1002 current->comm, task_pid_nr(current), task_pid_nr(task),
1004 task->signal->oom_adj = oom_adjust;
1006 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1007 * value is always attainable.
1009 if (task->signal->oom_adj == OOM_ADJUST_MAX)
1010 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
1012 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
1014 trace_oom_score_adj_update(task);
1016 unlock_task_sighand(task, &flags);
1019 put_task_struct(task);
1021 return err < 0 ? err : count;
1024 static const struct file_operations proc_oom_adjust_operations = {
1025 .read = oom_adjust_read,
1026 .write = oom_adjust_write,
1027 .llseek = generic_file_llseek,
1030 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1031 size_t count, loff_t *ppos)
1033 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1034 char buffer[PROC_NUMBUF];
1035 int oom_score_adj = OOM_SCORE_ADJ_MIN;
1036 unsigned long flags;
1041 if (lock_task_sighand(task, &flags)) {
1042 oom_score_adj = task->signal->oom_score_adj;
1043 unlock_task_sighand(task, &flags);
1045 put_task_struct(task);
1046 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1047 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1050 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1051 size_t count, loff_t *ppos)
1053 struct task_struct *task;
1054 char buffer[PROC_NUMBUF];
1055 unsigned long flags;
1059 memset(buffer, 0, sizeof(buffer));
1060 if (count > sizeof(buffer) - 1)
1061 count = sizeof(buffer) - 1;
1062 if (copy_from_user(buffer, buf, count)) {
1067 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1070 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1071 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1076 task = get_proc_task(file->f_path.dentry->d_inode);
1088 if (!lock_task_sighand(task, &flags)) {
1093 if (oom_score_adj < task->signal->oom_score_adj_min &&
1094 !capable(CAP_SYS_RESOURCE)) {
1099 task->signal->oom_score_adj = oom_score_adj;
1100 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1101 task->signal->oom_score_adj_min = oom_score_adj;
1102 trace_oom_score_adj_update(task);
1104 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1105 * always attainable.
1107 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1108 task->signal->oom_adj = OOM_DISABLE;
1110 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1113 unlock_task_sighand(task, &flags);
1116 put_task_struct(task);
1118 return err < 0 ? err : count;
1121 static const struct file_operations proc_oom_score_adj_operations = {
1122 .read = oom_score_adj_read,
1123 .write = oom_score_adj_write,
1124 .llseek = default_llseek,
1127 #ifdef CONFIG_AUDITSYSCALL
1128 #define TMPBUFLEN 21
1129 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1130 size_t count, loff_t *ppos)
1132 struct inode * inode = file->f_path.dentry->d_inode;
1133 struct task_struct *task = get_proc_task(inode);
1135 char tmpbuf[TMPBUFLEN];
1139 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1140 audit_get_loginuid(task));
1141 put_task_struct(task);
1142 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1145 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1146 size_t count, loff_t *ppos)
1148 struct inode * inode = file->f_path.dentry->d_inode;
1153 if (!capable(CAP_AUDIT_CONTROL))
1157 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1163 if (count >= PAGE_SIZE)
1164 count = PAGE_SIZE - 1;
1167 /* No partial writes. */
1170 page = (char*)__get_free_page(GFP_TEMPORARY);
1174 if (copy_from_user(page, buf, count))
1178 loginuid = simple_strtoul(page, &tmp, 10);
1184 length = audit_set_loginuid(current, loginuid);
1185 if (likely(length == 0))
1189 free_page((unsigned long) page);
1193 static const struct file_operations proc_loginuid_operations = {
1194 .read = proc_loginuid_read,
1195 .write = proc_loginuid_write,
1196 .llseek = generic_file_llseek,
1199 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1200 size_t count, loff_t *ppos)
1202 struct inode * inode = file->f_path.dentry->d_inode;
1203 struct task_struct *task = get_proc_task(inode);
1205 char tmpbuf[TMPBUFLEN];
1209 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1210 audit_get_sessionid(task));
1211 put_task_struct(task);
1212 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1215 static const struct file_operations proc_sessionid_operations = {
1216 .read = proc_sessionid_read,
1217 .llseek = generic_file_llseek,
1221 #ifdef CONFIG_FAULT_INJECTION
1222 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1223 size_t count, loff_t *ppos)
1225 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1226 char buffer[PROC_NUMBUF];
1232 make_it_fail = task->make_it_fail;
1233 put_task_struct(task);
1235 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1237 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1240 static ssize_t proc_fault_inject_write(struct file * file,
1241 const char __user * buf, size_t count, loff_t *ppos)
1243 struct task_struct *task;
1244 char buffer[PROC_NUMBUF], *end;
1247 if (!capable(CAP_SYS_RESOURCE))
1249 memset(buffer, 0, sizeof(buffer));
1250 if (count > sizeof(buffer) - 1)
1251 count = sizeof(buffer) - 1;
1252 if (copy_from_user(buffer, buf, count))
1254 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1257 task = get_proc_task(file->f_dentry->d_inode);
1260 task->make_it_fail = make_it_fail;
1261 put_task_struct(task);
1266 static const struct file_operations proc_fault_inject_operations = {
1267 .read = proc_fault_inject_read,
1268 .write = proc_fault_inject_write,
1269 .llseek = generic_file_llseek,
1274 #ifdef CONFIG_SCHED_DEBUG
1276 * Print out various scheduling related per-task fields:
1278 static int sched_show(struct seq_file *m, void *v)
1280 struct inode *inode = m->private;
1281 struct task_struct *p;
1283 p = get_proc_task(inode);
1286 proc_sched_show_task(p, m);
1294 sched_write(struct file *file, const char __user *buf,
1295 size_t count, loff_t *offset)
1297 struct inode *inode = file->f_path.dentry->d_inode;
1298 struct task_struct *p;
1300 p = get_proc_task(inode);
1303 proc_sched_set_task(p);
1310 static int sched_open(struct inode *inode, struct file *filp)
1312 return single_open(filp, sched_show, inode);
1315 static const struct file_operations proc_pid_sched_operations = {
1318 .write = sched_write,
1319 .llseek = seq_lseek,
1320 .release = single_release,
1325 #ifdef CONFIG_SCHED_AUTOGROUP
1327 * Print out autogroup related information:
1329 static int sched_autogroup_show(struct seq_file *m, void *v)
1331 struct inode *inode = m->private;
1332 struct task_struct *p;
1334 p = get_proc_task(inode);
1337 proc_sched_autogroup_show_task(p, m);
1345 sched_autogroup_write(struct file *file, const char __user *buf,
1346 size_t count, loff_t *offset)
1348 struct inode *inode = file->f_path.dentry->d_inode;
1349 struct task_struct *p;
1350 char buffer[PROC_NUMBUF];
1354 memset(buffer, 0, sizeof(buffer));
1355 if (count > sizeof(buffer) - 1)
1356 count = sizeof(buffer) - 1;
1357 if (copy_from_user(buffer, buf, count))
1360 err = kstrtoint(strstrip(buffer), 0, &nice);
1364 p = get_proc_task(inode);
1369 err = proc_sched_autogroup_set_nice(p, &err);
1378 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1382 ret = single_open(filp, sched_autogroup_show, NULL);
1384 struct seq_file *m = filp->private_data;
1391 static const struct file_operations proc_pid_sched_autogroup_operations = {
1392 .open = sched_autogroup_open,
1394 .write = sched_autogroup_write,
1395 .llseek = seq_lseek,
1396 .release = single_release,
1399 #endif /* CONFIG_SCHED_AUTOGROUP */
1401 static ssize_t comm_write(struct file *file, const char __user *buf,
1402 size_t count, loff_t *offset)
1404 struct inode *inode = file->f_path.dentry->d_inode;
1405 struct task_struct *p;
1406 char buffer[TASK_COMM_LEN];
1408 memset(buffer, 0, sizeof(buffer));
1409 if (count > sizeof(buffer) - 1)
1410 count = sizeof(buffer) - 1;
1411 if (copy_from_user(buffer, buf, count))
1414 p = get_proc_task(inode);
1418 if (same_thread_group(current, p))
1419 set_task_comm(p, buffer);
1428 static int comm_show(struct seq_file *m, void *v)
1430 struct inode *inode = m->private;
1431 struct task_struct *p;
1433 p = get_proc_task(inode);
1438 seq_printf(m, "%s\n", p->comm);
1446 static int comm_open(struct inode *inode, struct file *filp)
1448 return single_open(filp, comm_show, inode);
1451 static const struct file_operations proc_pid_set_comm_operations = {
1454 .write = comm_write,
1455 .llseek = seq_lseek,
1456 .release = single_release,
1459 static int proc_exe_link(struct inode *inode, struct path *exe_path)
1461 struct task_struct *task;
1462 struct mm_struct *mm;
1463 struct file *exe_file;
1465 task = get_proc_task(inode);
1468 mm = get_task_mm(task);
1469 put_task_struct(task);
1472 exe_file = get_mm_exe_file(mm);
1475 *exe_path = exe_file->f_path;
1476 path_get(&exe_file->f_path);
1483 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1485 struct inode *inode = dentry->d_inode;
1486 int error = -EACCES;
1488 /* We don't need a base pointer in the /proc filesystem */
1489 path_put(&nd->path);
1491 /* Are we allowed to snoop on the tasks file descriptors? */
1492 if (!proc_fd_access_allowed(inode))
1495 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1497 return ERR_PTR(error);
1500 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1502 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1509 pathname = d_path(path, tmp, PAGE_SIZE);
1510 len = PTR_ERR(pathname);
1511 if (IS_ERR(pathname))
1513 len = tmp + PAGE_SIZE - 1 - pathname;
1517 if (copy_to_user(buffer, pathname, len))
1520 free_page((unsigned long)tmp);
1524 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1526 int error = -EACCES;
1527 struct inode *inode = dentry->d_inode;
1530 /* Are we allowed to snoop on the tasks file descriptors? */
1531 if (!proc_fd_access_allowed(inode))
1534 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1538 error = do_proc_readlink(&path, buffer, buflen);
1544 static const struct inode_operations proc_pid_link_inode_operations = {
1545 .readlink = proc_pid_readlink,
1546 .follow_link = proc_pid_follow_link,
1547 .setattr = proc_setattr,
1551 /* building an inode */
1553 static int task_dumpable(struct task_struct *task)
1556 struct mm_struct *mm;
1561 dumpable = get_dumpable(mm);
1568 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1570 struct inode * inode;
1571 struct proc_inode *ei;
1572 const struct cred *cred;
1574 /* We need a new inode */
1576 inode = new_inode(sb);
1582 inode->i_ino = get_next_ino();
1583 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1584 inode->i_op = &proc_def_inode_operations;
1587 * grab the reference to task.
1589 ei->pid = get_task_pid(task, PIDTYPE_PID);
1593 if (task_dumpable(task)) {
1595 cred = __task_cred(task);
1596 inode->i_uid = cred->euid;
1597 inode->i_gid = cred->egid;
1600 security_task_to_inode(task, inode);
1610 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1612 struct inode *inode = dentry->d_inode;
1613 struct task_struct *task;
1614 const struct cred *cred;
1616 generic_fillattr(inode, stat);
1621 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1623 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1624 task_dumpable(task)) {
1625 cred = __task_cred(task);
1626 stat->uid = cred->euid;
1627 stat->gid = cred->egid;
1637 * Exceptional case: normally we are not allowed to unhash a busy
1638 * directory. In this case, however, we can do it - no aliasing problems
1639 * due to the way we treat inodes.
1641 * Rewrite the inode's ownerships here because the owning task may have
1642 * performed a setuid(), etc.
1644 * Before the /proc/pid/status file was created the only way to read
1645 * the effective uid of a /process was to stat /proc/pid. Reading
1646 * /proc/pid/status is slow enough that procps and other packages
1647 * kept stating /proc/pid. To keep the rules in /proc simple I have
1648 * made this apply to all per process world readable and executable
1651 int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1653 struct inode *inode;
1654 struct task_struct *task;
1655 const struct cred *cred;
1657 if (nd && nd->flags & LOOKUP_RCU)
1660 inode = dentry->d_inode;
1661 task = get_proc_task(inode);
1664 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1665 task_dumpable(task)) {
1667 cred = __task_cred(task);
1668 inode->i_uid = cred->euid;
1669 inode->i_gid = cred->egid;
1675 inode->i_mode &= ~(S_ISUID | S_ISGID);
1676 security_task_to_inode(task, inode);
1677 put_task_struct(task);
1684 static int pid_delete_dentry(const struct dentry * dentry)
1686 /* Is the task we represent dead?
1687 * If so, then don't put the dentry on the lru list,
1688 * kill it immediately.
1690 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1693 const struct dentry_operations pid_dentry_operations =
1695 .d_revalidate = pid_revalidate,
1696 .d_delete = pid_delete_dentry,
1702 * Fill a directory entry.
1704 * If possible create the dcache entry and derive our inode number and
1705 * file type from dcache entry.
1707 * Since all of the proc inode numbers are dynamically generated, the inode
1708 * numbers do not exist until the inode is cache. This means creating the
1709 * the dcache entry in readdir is necessary to keep the inode numbers
1710 * reported by readdir in sync with the inode numbers reported
1713 int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1714 const char *name, int len,
1715 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1717 struct dentry *child, *dir = filp->f_path.dentry;
1718 struct inode *inode;
1721 unsigned type = DT_UNKNOWN;
1725 qname.hash = full_name_hash(name, len);
1727 child = d_lookup(dir, &qname);
1730 new = d_alloc(dir, &qname);
1732 child = instantiate(dir->d_inode, new, task, ptr);
1739 if (!child || IS_ERR(child) || !child->d_inode)
1740 goto end_instantiate;
1741 inode = child->d_inode;
1744 type = inode->i_mode >> 12;
1749 ino = find_inode_number(dir, &qname);
1752 return filldir(dirent, name, len, filp->f_pos, ino, type);
1755 static unsigned name_to_int(struct dentry *dentry)
1757 const char *name = dentry->d_name.name;
1758 int len = dentry->d_name.len;
1761 if (len > 1 && *name == '0')
1764 unsigned c = *name++ - '0';
1767 if (n >= (~0U-9)/10)
1777 #define PROC_FDINFO_MAX 64
1779 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1781 struct task_struct *task = get_proc_task(inode);
1782 struct files_struct *files = NULL;
1784 int fd = proc_fd(inode);
1787 files = get_files_struct(task);
1788 put_task_struct(task);
1792 * We are not taking a ref to the file structure, so we must
1795 spin_lock(&files->file_lock);
1796 file = fcheck_files(files, fd);
1798 unsigned int f_flags;
1799 struct fdtable *fdt;
1801 fdt = files_fdtable(files);
1802 f_flags = file->f_flags & ~O_CLOEXEC;
1803 if (FD_ISSET(fd, fdt->close_on_exec))
1804 f_flags |= O_CLOEXEC;
1807 *path = file->f_path;
1808 path_get(&file->f_path);
1811 snprintf(info, PROC_FDINFO_MAX,
1814 (long long) file->f_pos,
1816 spin_unlock(&files->file_lock);
1817 put_files_struct(files);
1820 spin_unlock(&files->file_lock);
1821 put_files_struct(files);
1826 static int proc_fd_link(struct inode *inode, struct path *path)
1828 return proc_fd_info(inode, path, NULL);
1831 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1833 struct inode *inode;
1834 struct task_struct *task;
1836 struct files_struct *files;
1837 const struct cred *cred;
1839 if (nd && nd->flags & LOOKUP_RCU)
1842 inode = dentry->d_inode;
1843 task = get_proc_task(inode);
1844 fd = proc_fd(inode);
1847 files = get_files_struct(task);
1850 if (fcheck_files(files, fd)) {
1852 put_files_struct(files);
1853 if (task_dumpable(task)) {
1855 cred = __task_cred(task);
1856 inode->i_uid = cred->euid;
1857 inode->i_gid = cred->egid;
1863 inode->i_mode &= ~(S_ISUID | S_ISGID);
1864 security_task_to_inode(task, inode);
1865 put_task_struct(task);
1869 put_files_struct(files);
1871 put_task_struct(task);
1877 static const struct dentry_operations tid_fd_dentry_operations =
1879 .d_revalidate = tid_fd_revalidate,
1880 .d_delete = pid_delete_dentry,
1883 static struct dentry *proc_fd_instantiate(struct inode *dir,
1884 struct dentry *dentry, struct task_struct *task, const void *ptr)
1886 unsigned fd = *(const unsigned *)ptr;
1888 struct files_struct *files;
1889 struct inode *inode;
1890 struct proc_inode *ei;
1891 struct dentry *error = ERR_PTR(-ENOENT);
1893 inode = proc_pid_make_inode(dir->i_sb, task);
1898 files = get_files_struct(task);
1901 inode->i_mode = S_IFLNK;
1904 * We are not taking a ref to the file structure, so we must
1907 spin_lock(&files->file_lock);
1908 file = fcheck_files(files, fd);
1911 if (file->f_mode & FMODE_READ)
1912 inode->i_mode |= S_IRUSR | S_IXUSR;
1913 if (file->f_mode & FMODE_WRITE)
1914 inode->i_mode |= S_IWUSR | S_IXUSR;
1915 spin_unlock(&files->file_lock);
1916 put_files_struct(files);
1918 inode->i_op = &proc_pid_link_inode_operations;
1920 ei->op.proc_get_link = proc_fd_link;
1921 d_set_d_op(dentry, &tid_fd_dentry_operations);
1922 d_add(dentry, inode);
1923 /* Close the race of the process dying before we return the dentry */
1924 if (tid_fd_revalidate(dentry, NULL))
1930 spin_unlock(&files->file_lock);
1931 put_files_struct(files);
1937 static struct dentry *proc_lookupfd_common(struct inode *dir,
1938 struct dentry *dentry,
1939 instantiate_t instantiate)
1941 struct task_struct *task = get_proc_task(dir);
1942 unsigned fd = name_to_int(dentry);
1943 struct dentry *result = ERR_PTR(-ENOENT);
1950 result = instantiate(dir, dentry, task, &fd);
1952 put_task_struct(task);
1957 static int proc_readfd_common(struct file * filp, void * dirent,
1958 filldir_t filldir, instantiate_t instantiate)
1960 struct dentry *dentry = filp->f_path.dentry;
1961 struct inode *inode = dentry->d_inode;
1962 struct task_struct *p = get_proc_task(inode);
1963 unsigned int fd, ino;
1965 struct files_struct * files;
1975 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1979 ino = parent_ino(dentry);
1980 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1984 files = get_files_struct(p);
1988 for (fd = filp->f_pos-2;
1989 fd < files_fdtable(files)->max_fds;
1990 fd++, filp->f_pos++) {
1991 char name[PROC_NUMBUF];
1994 if (!fcheck_files(files, fd))
1998 len = snprintf(name, sizeof(name), "%d", fd);
1999 if (proc_fill_cache(filp, dirent, filldir,
2000 name, len, instantiate,
2008 put_files_struct(files);
2016 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
2017 struct nameidata *nd)
2019 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
2022 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
2024 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
2027 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
2028 size_t len, loff_t *ppos)
2030 char tmp[PROC_FDINFO_MAX];
2031 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
2033 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
2037 static const struct file_operations proc_fdinfo_file_operations = {
2038 .open = nonseekable_open,
2039 .read = proc_fdinfo_read,
2040 .llseek = no_llseek,
2043 static const struct file_operations proc_fd_operations = {
2044 .read = generic_read_dir,
2045 .readdir = proc_readfd,
2046 .llseek = default_llseek,
2050 * /proc/pid/fd needs a special permission handler so that a process can still
2051 * access /proc/self/fd after it has executed a setuid().
2053 static int proc_fd_permission(struct inode *inode, int mask)
2055 int rv = generic_permission(inode, mask);
2058 if (task_pid(current) == proc_pid(inode))
2064 * proc directories can do almost nothing..
2066 static const struct inode_operations proc_fd_inode_operations = {
2067 .lookup = proc_lookupfd,
2068 .permission = proc_fd_permission,
2069 .setattr = proc_setattr,
2072 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2073 struct dentry *dentry, struct task_struct *task, const void *ptr)
2075 unsigned fd = *(unsigned *)ptr;
2076 struct inode *inode;
2077 struct proc_inode *ei;
2078 struct dentry *error = ERR_PTR(-ENOENT);
2080 inode = proc_pid_make_inode(dir->i_sb, task);
2085 inode->i_mode = S_IFREG | S_IRUSR;
2086 inode->i_fop = &proc_fdinfo_file_operations;
2087 d_set_d_op(dentry, &tid_fd_dentry_operations);
2088 d_add(dentry, inode);
2089 /* Close the race of the process dying before we return the dentry */
2090 if (tid_fd_revalidate(dentry, NULL))
2097 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2098 struct dentry *dentry,
2099 struct nameidata *nd)
2101 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2104 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2106 return proc_readfd_common(filp, dirent, filldir,
2107 proc_fdinfo_instantiate);
2110 static const struct file_operations proc_fdinfo_operations = {
2111 .read = generic_read_dir,
2112 .readdir = proc_readfdinfo,
2113 .llseek = default_llseek,
2117 * proc directories can do almost nothing..
2119 static const struct inode_operations proc_fdinfo_inode_operations = {
2120 .lookup = proc_lookupfdinfo,
2121 .setattr = proc_setattr,
2125 static struct dentry *proc_pident_instantiate(struct inode *dir,
2126 struct dentry *dentry, struct task_struct *task, const void *ptr)
2128 const struct pid_entry *p = ptr;
2129 struct inode *inode;
2130 struct proc_inode *ei;
2131 struct dentry *error = ERR_PTR(-ENOENT);
2133 inode = proc_pid_make_inode(dir->i_sb, task);
2138 inode->i_mode = p->mode;
2139 if (S_ISDIR(inode->i_mode))
2140 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2142 inode->i_op = p->iop;
2144 inode->i_fop = p->fop;
2146 d_set_d_op(dentry, &pid_dentry_operations);
2147 d_add(dentry, inode);
2148 /* Close the race of the process dying before we return the dentry */
2149 if (pid_revalidate(dentry, NULL))
2155 static struct dentry *proc_pident_lookup(struct inode *dir,
2156 struct dentry *dentry,
2157 const struct pid_entry *ents,
2160 struct dentry *error;
2161 struct task_struct *task = get_proc_task(dir);
2162 const struct pid_entry *p, *last;
2164 error = ERR_PTR(-ENOENT);
2170 * Yes, it does not scale. And it should not. Don't add
2171 * new entries into /proc/<tgid>/ without very good reasons.
2173 last = &ents[nents - 1];
2174 for (p = ents; p <= last; p++) {
2175 if (p->len != dentry->d_name.len)
2177 if (!memcmp(dentry->d_name.name, p->name, p->len))
2183 error = proc_pident_instantiate(dir, dentry, task, p);
2185 put_task_struct(task);
2190 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2191 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2193 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2194 proc_pident_instantiate, task, p);
2197 static int proc_pident_readdir(struct file *filp,
2198 void *dirent, filldir_t filldir,
2199 const struct pid_entry *ents, unsigned int nents)
2202 struct dentry *dentry = filp->f_path.dentry;
2203 struct inode *inode = dentry->d_inode;
2204 struct task_struct *task = get_proc_task(inode);
2205 const struct pid_entry *p, *last;
2218 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2224 ino = parent_ino(dentry);
2225 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2237 last = &ents[nents - 1];
2239 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2248 put_task_struct(task);
2253 #ifdef CONFIG_SECURITY
2254 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2255 size_t count, loff_t *ppos)
2257 struct inode * inode = file->f_path.dentry->d_inode;
2260 struct task_struct *task = get_proc_task(inode);
2265 length = security_getprocattr(task,
2266 (char*)file->f_path.dentry->d_name.name,
2268 put_task_struct(task);
2270 length = simple_read_from_buffer(buf, count, ppos, p, length);
2275 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2276 size_t count, loff_t *ppos)
2278 struct inode * inode = file->f_path.dentry->d_inode;
2281 struct task_struct *task = get_proc_task(inode);
2286 if (count > PAGE_SIZE)
2289 /* No partial writes. */
2295 page = (char*)__get_free_page(GFP_TEMPORARY);
2300 if (copy_from_user(page, buf, count))
2303 /* Guard against adverse ptrace interaction */
2304 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2308 length = security_setprocattr(task,
2309 (char*)file->f_path.dentry->d_name.name,
2310 (void*)page, count);
2311 mutex_unlock(&task->signal->cred_guard_mutex);
2313 free_page((unsigned long) page);
2315 put_task_struct(task);
2320 static const struct file_operations proc_pid_attr_operations = {
2321 .read = proc_pid_attr_read,
2322 .write = proc_pid_attr_write,
2323 .llseek = generic_file_llseek,
2326 static const struct pid_entry attr_dir_stuff[] = {
2327 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2328 REG("prev", S_IRUGO, proc_pid_attr_operations),
2329 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2330 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2331 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2332 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2335 static int proc_attr_dir_readdir(struct file * filp,
2336 void * dirent, filldir_t filldir)
2338 return proc_pident_readdir(filp,dirent,filldir,
2339 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2342 static const struct file_operations proc_attr_dir_operations = {
2343 .read = generic_read_dir,
2344 .readdir = proc_attr_dir_readdir,
2345 .llseek = default_llseek,
2348 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2349 struct dentry *dentry, struct nameidata *nd)
2351 return proc_pident_lookup(dir, dentry,
2352 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2355 static const struct inode_operations proc_attr_dir_inode_operations = {
2356 .lookup = proc_attr_dir_lookup,
2357 .getattr = pid_getattr,
2358 .setattr = proc_setattr,
2363 #ifdef CONFIG_ELF_CORE
2364 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2365 size_t count, loff_t *ppos)
2367 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2368 struct mm_struct *mm;
2369 char buffer[PROC_NUMBUF];
2377 mm = get_task_mm(task);
2379 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2380 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2381 MMF_DUMP_FILTER_SHIFT));
2383 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2386 put_task_struct(task);
2391 static ssize_t proc_coredump_filter_write(struct file *file,
2392 const char __user *buf,
2396 struct task_struct *task;
2397 struct mm_struct *mm;
2398 char buffer[PROC_NUMBUF], *end;
2405 memset(buffer, 0, sizeof(buffer));
2406 if (count > sizeof(buffer) - 1)
2407 count = sizeof(buffer) - 1;
2408 if (copy_from_user(buffer, buf, count))
2412 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2415 if (end - buffer == 0)
2419 task = get_proc_task(file->f_dentry->d_inode);
2424 mm = get_task_mm(task);
2428 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2430 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2432 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2437 put_task_struct(task);
2442 static const struct file_operations proc_coredump_filter_operations = {
2443 .read = proc_coredump_filter_read,
2444 .write = proc_coredump_filter_write,
2445 .llseek = generic_file_llseek,
2452 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2455 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2456 pid_t tgid = task_tgid_nr_ns(current, ns);
2457 char tmp[PROC_NUMBUF];
2460 sprintf(tmp, "%d", tgid);
2461 return vfs_readlink(dentry,buffer,buflen,tmp);
2464 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2466 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2467 pid_t tgid = task_tgid_nr_ns(current, ns);
2468 char *name = ERR_PTR(-ENOENT);
2472 name = ERR_PTR(-ENOMEM);
2474 sprintf(name, "%d", tgid);
2476 nd_set_link(nd, name);
2480 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2483 char *s = nd_get_link(nd);
2488 static const struct inode_operations proc_self_inode_operations = {
2489 .readlink = proc_self_readlink,
2490 .follow_link = proc_self_follow_link,
2491 .put_link = proc_self_put_link,
2497 * These are the directory entries in the root directory of /proc
2498 * that properly belong to the /proc filesystem, as they describe
2499 * describe something that is process related.
2501 static const struct pid_entry proc_base_stuff[] = {
2502 NOD("self", S_IFLNK|S_IRWXUGO,
2503 &proc_self_inode_operations, NULL, {}),
2506 static struct dentry *proc_base_instantiate(struct inode *dir,
2507 struct dentry *dentry, struct task_struct *task, const void *ptr)
2509 const struct pid_entry *p = ptr;
2510 struct inode *inode;
2511 struct proc_inode *ei;
2512 struct dentry *error;
2514 /* Allocate the inode */
2515 error = ERR_PTR(-ENOMEM);
2516 inode = new_inode(dir->i_sb);
2520 /* Initialize the inode */
2522 inode->i_ino = get_next_ino();
2523 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2526 * grab the reference to the task.
2528 ei->pid = get_task_pid(task, PIDTYPE_PID);
2532 inode->i_mode = p->mode;
2533 if (S_ISDIR(inode->i_mode))
2534 set_nlink(inode, 2);
2535 if (S_ISLNK(inode->i_mode))
2538 inode->i_op = p->iop;
2540 inode->i_fop = p->fop;
2542 d_add(dentry, inode);
2551 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2553 struct dentry *error;
2554 struct task_struct *task = get_proc_task(dir);
2555 const struct pid_entry *p, *last;
2557 error = ERR_PTR(-ENOENT);
2562 /* Lookup the directory entry */
2563 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2564 for (p = proc_base_stuff; p <= last; p++) {
2565 if (p->len != dentry->d_name.len)
2567 if (!memcmp(dentry->d_name.name, p->name, p->len))
2573 error = proc_base_instantiate(dir, dentry, task, p);
2576 put_task_struct(task);
2581 static int proc_base_fill_cache(struct file *filp, void *dirent,
2582 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2584 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2585 proc_base_instantiate, task, p);
2588 #ifdef CONFIG_TASK_IO_ACCOUNTING
2589 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2591 struct task_io_accounting acct = task->ioac;
2592 unsigned long flags;
2595 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2599 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2604 if (whole && lock_task_sighand(task, &flags)) {
2605 struct task_struct *t = task;
2607 task_io_accounting_add(&acct, &task->signal->ioac);
2608 while_each_thread(task, t)
2609 task_io_accounting_add(&acct, &t->ioac);
2611 unlock_task_sighand(task, &flags);
2613 result = sprintf(buffer,
2618 "read_bytes: %llu\n"
2619 "write_bytes: %llu\n"
2620 "cancelled_write_bytes: %llu\n",
2621 (unsigned long long)acct.rchar,
2622 (unsigned long long)acct.wchar,
2623 (unsigned long long)acct.syscr,
2624 (unsigned long long)acct.syscw,
2625 (unsigned long long)acct.read_bytes,
2626 (unsigned long long)acct.write_bytes,
2627 (unsigned long long)acct.cancelled_write_bytes);
2629 mutex_unlock(&task->signal->cred_guard_mutex);
2633 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2635 return do_io_accounting(task, buffer, 0);
2638 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2640 return do_io_accounting(task, buffer, 1);
2642 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2644 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2645 struct pid *pid, struct task_struct *task)
2647 int err = lock_trace(task);
2649 seq_printf(m, "%08x\n", task->personality);
2658 static const struct file_operations proc_task_operations;
2659 static const struct inode_operations proc_task_inode_operations;
2661 static const struct pid_entry tgid_base_stuff[] = {
2662 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2663 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2664 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2665 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2667 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2669 REG("environ", S_IRUSR, proc_environ_operations),
2670 INF("auxv", S_IRUSR, proc_pid_auxv),
2671 ONE("status", S_IRUGO, proc_pid_status),
2672 ONE("personality", S_IRUGO, proc_pid_personality),
2673 INF("limits", S_IRUGO, proc_pid_limits),
2674 #ifdef CONFIG_SCHED_DEBUG
2675 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2677 #ifdef CONFIG_SCHED_AUTOGROUP
2678 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2680 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2681 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2682 INF("syscall", S_IRUGO, proc_pid_syscall),
2684 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2685 ONE("stat", S_IRUGO, proc_tgid_stat),
2686 ONE("statm", S_IRUGO, proc_pid_statm),
2687 REG("maps", S_IRUGO, proc_maps_operations),
2689 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2691 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2692 LNK("cwd", proc_cwd_link),
2693 LNK("root", proc_root_link),
2694 LNK("exe", proc_exe_link),
2695 REG("mounts", S_IRUGO, proc_mounts_operations),
2696 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2697 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2698 #ifdef CONFIG_PROC_PAGE_MONITOR
2699 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2700 REG("smaps", S_IRUGO, proc_smaps_operations),
2701 REG("pagemap", S_IRUGO, proc_pagemap_operations),
2703 #ifdef CONFIG_SECURITY
2704 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2706 #ifdef CONFIG_KALLSYMS
2707 INF("wchan", S_IRUGO, proc_pid_wchan),
2709 #ifdef CONFIG_STACKTRACE
2710 ONE("stack", S_IRUGO, proc_pid_stack),
2712 #ifdef CONFIG_SCHEDSTATS
2713 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2715 #ifdef CONFIG_LATENCYTOP
2716 REG("latency", S_IRUGO, proc_lstats_operations),
2718 #ifdef CONFIG_PROC_PID_CPUSET
2719 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2721 #ifdef CONFIG_CGROUPS
2722 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2724 INF("oom_score", S_IRUGO, proc_oom_score),
2725 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2726 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2727 #ifdef CONFIG_AUDITSYSCALL
2728 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2729 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2731 #ifdef CONFIG_FAULT_INJECTION
2732 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2734 #ifdef CONFIG_ELF_CORE
2735 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2737 #ifdef CONFIG_TASK_IO_ACCOUNTING
2738 INF("io", S_IRUSR, proc_tgid_io_accounting),
2740 #ifdef CONFIG_HARDWALL
2741 INF("hardwall", S_IRUGO, proc_pid_hardwall),
2745 static int proc_tgid_base_readdir(struct file * filp,
2746 void * dirent, filldir_t filldir)
2748 return proc_pident_readdir(filp,dirent,filldir,
2749 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2752 static const struct file_operations proc_tgid_base_operations = {
2753 .read = generic_read_dir,
2754 .readdir = proc_tgid_base_readdir,
2755 .llseek = default_llseek,
2758 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2759 return proc_pident_lookup(dir, dentry,
2760 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2763 static const struct inode_operations proc_tgid_base_inode_operations = {
2764 .lookup = proc_tgid_base_lookup,
2765 .getattr = pid_getattr,
2766 .setattr = proc_setattr,
2769 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2771 struct dentry *dentry, *leader, *dir;
2772 char buf[PROC_NUMBUF];
2776 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2777 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2779 shrink_dcache_parent(dentry);
2785 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2786 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2791 name.len = strlen(name.name);
2792 dir = d_hash_and_lookup(leader, &name);
2794 goto out_put_leader;
2797 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2798 dentry = d_hash_and_lookup(dir, &name);
2800 shrink_dcache_parent(dentry);
2813 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2814 * @task: task that should be flushed.
2816 * When flushing dentries from proc, one needs to flush them from global
2817 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2818 * in. This call is supposed to do all of this job.
2820 * Looks in the dcache for
2822 * /proc/@tgid/task/@pid
2823 * if either directory is present flushes it and all of it'ts children
2826 * It is safe and reasonable to cache /proc entries for a task until
2827 * that task exits. After that they just clog up the dcache with
2828 * useless entries, possibly causing useful dcache entries to be
2829 * flushed instead. This routine is proved to flush those useless
2830 * dcache entries at process exit time.
2832 * NOTE: This routine is just an optimization so it does not guarantee
2833 * that no dcache entries will exist at process exit time it
2834 * just makes it very unlikely that any will persist.
2837 void proc_flush_task(struct task_struct *task)
2840 struct pid *pid, *tgid;
2843 pid = task_pid(task);
2844 tgid = task_tgid(task);
2846 for (i = 0; i <= pid->level; i++) {
2847 upid = &pid->numbers[i];
2848 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2849 tgid->numbers[i].nr);
2852 upid = &pid->numbers[pid->level];
2854 pid_ns_release_proc(upid->ns);
2857 static struct dentry *proc_pid_instantiate(struct inode *dir,
2858 struct dentry * dentry,
2859 struct task_struct *task, const void *ptr)
2861 struct dentry *error = ERR_PTR(-ENOENT);
2862 struct inode *inode;
2864 inode = proc_pid_make_inode(dir->i_sb, task);
2868 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2869 inode->i_op = &proc_tgid_base_inode_operations;
2870 inode->i_fop = &proc_tgid_base_operations;
2871 inode->i_flags|=S_IMMUTABLE;
2873 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
2874 ARRAY_SIZE(tgid_base_stuff)));
2876 d_set_d_op(dentry, &pid_dentry_operations);
2878 d_add(dentry, inode);
2879 /* Close the race of the process dying before we return the dentry */
2880 if (pid_revalidate(dentry, NULL))
2886 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2888 struct dentry *result;
2889 struct task_struct *task;
2891 struct pid_namespace *ns;
2893 result = proc_base_lookup(dir, dentry);
2894 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2897 tgid = name_to_int(dentry);
2901 ns = dentry->d_sb->s_fs_info;
2903 task = find_task_by_pid_ns(tgid, ns);
2905 get_task_struct(task);
2910 result = proc_pid_instantiate(dir, dentry, task, NULL);
2911 put_task_struct(task);
2917 * Find the first task with tgid >= tgid
2922 struct task_struct *task;
2924 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2929 put_task_struct(iter.task);
2933 pid = find_ge_pid(iter.tgid, ns);
2935 iter.tgid = pid_nr_ns(pid, ns);
2936 iter.task = pid_task(pid, PIDTYPE_PID);
2937 /* What we to know is if the pid we have find is the
2938 * pid of a thread_group_leader. Testing for task
2939 * being a thread_group_leader is the obvious thing
2940 * todo but there is a window when it fails, due to
2941 * the pid transfer logic in de_thread.
2943 * So we perform the straight forward test of seeing
2944 * if the pid we have found is the pid of a thread
2945 * group leader, and don't worry if the task we have
2946 * found doesn't happen to be a thread group leader.
2947 * As we don't care in the case of readdir.
2949 if (!iter.task || !has_group_leader_pid(iter.task)) {
2953 get_task_struct(iter.task);
2959 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2961 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2962 struct tgid_iter iter)
2964 char name[PROC_NUMBUF];
2965 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
2966 return proc_fill_cache(filp, dirent, filldir, name, len,
2967 proc_pid_instantiate, iter.task, NULL);
2970 /* for the /proc/ directory itself, after non-process stuff has been done */
2971 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2974 struct task_struct *reaper;
2975 struct tgid_iter iter;
2976 struct pid_namespace *ns;
2978 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
2980 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2982 reaper = get_proc_task(filp->f_path.dentry->d_inode);
2986 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
2987 const struct pid_entry *p = &proc_base_stuff[nr];
2988 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
2992 ns = filp->f_dentry->d_sb->s_fs_info;
2994 iter.tgid = filp->f_pos - TGID_OFFSET;
2995 for (iter = next_tgid(ns, iter);
2997 iter.tgid += 1, iter = next_tgid(ns, iter)) {
2998 filp->f_pos = iter.tgid + TGID_OFFSET;
2999 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
3000 put_task_struct(iter.task);
3004 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3006 put_task_struct(reaper);
3014 static const struct pid_entry tid_base_stuff[] = {
3015 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3016 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3017 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3018 REG("environ", S_IRUSR, proc_environ_operations),
3019 INF("auxv", S_IRUSR, proc_pid_auxv),
3020 ONE("status", S_IRUGO, proc_pid_status),
3021 ONE("personality", S_IRUGO, proc_pid_personality),
3022 INF("limits", S_IRUGO, proc_pid_limits),
3023 #ifdef CONFIG_SCHED_DEBUG
3024 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3026 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3027 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3028 INF("syscall", S_IRUGO, proc_pid_syscall),
3030 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3031 ONE("stat", S_IRUGO, proc_tid_stat),
3032 ONE("statm", S_IRUGO, proc_pid_statm),
3033 REG("maps", S_IRUGO, proc_maps_operations),
3035 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
3037 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3038 LNK("cwd", proc_cwd_link),
3039 LNK("root", proc_root_link),
3040 LNK("exe", proc_exe_link),
3041 REG("mounts", S_IRUGO, proc_mounts_operations),
3042 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3043 #ifdef CONFIG_PROC_PAGE_MONITOR
3044 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3045 REG("smaps", S_IRUGO, proc_smaps_operations),
3046 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3048 #ifdef CONFIG_SECURITY
3049 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3051 #ifdef CONFIG_KALLSYMS
3052 INF("wchan", S_IRUGO, proc_pid_wchan),
3054 #ifdef CONFIG_STACKTRACE
3055 ONE("stack", S_IRUGO, proc_pid_stack),
3057 #ifdef CONFIG_SCHEDSTATS
3058 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3060 #ifdef CONFIG_LATENCYTOP
3061 REG("latency", S_IRUGO, proc_lstats_operations),
3063 #ifdef CONFIG_PROC_PID_CPUSET
3064 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3066 #ifdef CONFIG_CGROUPS
3067 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3069 INF("oom_score", S_IRUGO, proc_oom_score),
3070 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3071 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3072 #ifdef CONFIG_AUDITSYSCALL
3073 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3074 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3076 #ifdef CONFIG_FAULT_INJECTION
3077 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3079 #ifdef CONFIG_TASK_IO_ACCOUNTING
3080 INF("io", S_IRUSR, proc_tid_io_accounting),
3082 #ifdef CONFIG_HARDWALL
3083 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3087 static int proc_tid_base_readdir(struct file * filp,
3088 void * dirent, filldir_t filldir)
3090 return proc_pident_readdir(filp,dirent,filldir,
3091 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3094 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3095 return proc_pident_lookup(dir, dentry,
3096 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3099 static const struct file_operations proc_tid_base_operations = {
3100 .read = generic_read_dir,
3101 .readdir = proc_tid_base_readdir,
3102 .llseek = default_llseek,
3105 static const struct inode_operations proc_tid_base_inode_operations = {
3106 .lookup = proc_tid_base_lookup,
3107 .getattr = pid_getattr,
3108 .setattr = proc_setattr,
3111 static struct dentry *proc_task_instantiate(struct inode *dir,
3112 struct dentry *dentry, struct task_struct *task, const void *ptr)
3114 struct dentry *error = ERR_PTR(-ENOENT);
3115 struct inode *inode;
3116 inode = proc_pid_make_inode(dir->i_sb, task);
3120 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3121 inode->i_op = &proc_tid_base_inode_operations;
3122 inode->i_fop = &proc_tid_base_operations;
3123 inode->i_flags|=S_IMMUTABLE;
3125 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3126 ARRAY_SIZE(tid_base_stuff)));
3128 d_set_d_op(dentry, &pid_dentry_operations);
3130 d_add(dentry, inode);
3131 /* Close the race of the process dying before we return the dentry */
3132 if (pid_revalidate(dentry, NULL))
3138 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3140 struct dentry *result = ERR_PTR(-ENOENT);
3141 struct task_struct *task;
3142 struct task_struct *leader = get_proc_task(dir);
3144 struct pid_namespace *ns;
3149 tid = name_to_int(dentry);
3153 ns = dentry->d_sb->s_fs_info;
3155 task = find_task_by_pid_ns(tid, ns);
3157 get_task_struct(task);
3161 if (!same_thread_group(leader, task))
3164 result = proc_task_instantiate(dir, dentry, task, NULL);
3166 put_task_struct(task);
3168 put_task_struct(leader);
3174 * Find the first tid of a thread group to return to user space.
3176 * Usually this is just the thread group leader, but if the users
3177 * buffer was too small or there was a seek into the middle of the
3178 * directory we have more work todo.
3180 * In the case of a short read we start with find_task_by_pid.
3182 * In the case of a seek we start with the leader and walk nr
3185 static struct task_struct *first_tid(struct task_struct *leader,
3186 int tid, int nr, struct pid_namespace *ns)
3188 struct task_struct *pos;
3191 /* Attempt to start with the pid of a thread */
3192 if (tid && (nr > 0)) {
3193 pos = find_task_by_pid_ns(tid, ns);
3194 if (pos && (pos->group_leader == leader))
3198 /* If nr exceeds the number of threads there is nothing todo */
3200 if (nr && nr >= get_nr_threads(leader))
3203 /* If we haven't found our starting place yet start
3204 * with the leader and walk nr threads forward.
3206 for (pos = leader; nr > 0; --nr) {
3207 pos = next_thread(pos);
3208 if (pos == leader) {
3214 get_task_struct(pos);
3221 * Find the next thread in the thread list.
3222 * Return NULL if there is an error or no next thread.
3224 * The reference to the input task_struct is released.
3226 static struct task_struct *next_tid(struct task_struct *start)
3228 struct task_struct *pos = NULL;
3230 if (pid_alive(start)) {
3231 pos = next_thread(start);
3232 if (thread_group_leader(pos))
3235 get_task_struct(pos);
3238 put_task_struct(start);
3242 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3243 struct task_struct *task, int tid)
3245 char name[PROC_NUMBUF];
3246 int len = snprintf(name, sizeof(name), "%d", tid);
3247 return proc_fill_cache(filp, dirent, filldir, name, len,
3248 proc_task_instantiate, task, NULL);
3251 /* for the /proc/TGID/task/ directories */
3252 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3254 struct dentry *dentry = filp->f_path.dentry;
3255 struct inode *inode = dentry->d_inode;
3256 struct task_struct *leader = NULL;
3257 struct task_struct *task;
3258 int retval = -ENOENT;
3261 struct pid_namespace *ns;
3263 task = get_proc_task(inode);
3267 if (pid_alive(task)) {
3268 leader = task->group_leader;
3269 get_task_struct(leader);
3272 put_task_struct(task);
3277 switch ((unsigned long)filp->f_pos) {
3280 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3285 ino = parent_ino(dentry);
3286 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3292 /* f_version caches the tgid value that the last readdir call couldn't
3293 * return. lseek aka telldir automagically resets f_version to 0.
3295 ns = filp->f_dentry->d_sb->s_fs_info;
3296 tid = (int)filp->f_version;
3297 filp->f_version = 0;
3298 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3300 task = next_tid(task), filp->f_pos++) {
3301 tid = task_pid_nr_ns(task, ns);
3302 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3303 /* returning this tgid failed, save it as the first
3304 * pid for the next readir call */
3305 filp->f_version = (u64)tid;
3306 put_task_struct(task);
3311 put_task_struct(leader);
3316 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3318 struct inode *inode = dentry->d_inode;
3319 struct task_struct *p = get_proc_task(inode);
3320 generic_fillattr(inode, stat);
3323 stat->nlink += get_nr_threads(p);
3330 static const struct inode_operations proc_task_inode_operations = {
3331 .lookup = proc_task_lookup,
3332 .getattr = proc_task_getattr,
3333 .setattr = proc_setattr,
3336 static const struct file_operations proc_task_operations = {
3337 .read = generic_read_dir,
3338 .readdir = proc_task_readdir,
3339 .llseek = default_llseek,