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>
89 * Implementing inode permission operations in /proc is almost
90 * certainly an error. Permission checks need to happen during
91 * each system call not at open time. The reason is that most of
92 * what we wish to check for permissions in /proc varies at runtime.
94 * The classic example of a problem is opening file descriptors
95 * in /proc for a task before it execs a suid executable.
102 const struct inode_operations *iop;
103 const struct file_operations *fop;
107 #define NOD(NAME, MODE, IOP, FOP, OP) { \
109 .len = sizeof(NAME) - 1, \
116 #define DIR(NAME, MODE, iops, fops) \
117 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
118 #define LNK(NAME, get_link) \
119 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
120 &proc_pid_link_inode_operations, NULL, \
121 { .proc_get_link = get_link } )
122 #define REG(NAME, MODE, fops) \
123 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
124 #define INF(NAME, MODE, read) \
125 NOD(NAME, (S_IFREG|(MODE)), \
126 NULL, &proc_info_file_operations, \
127 { .proc_read = read } )
128 #define ONE(NAME, MODE, show) \
129 NOD(NAME, (S_IFREG|(MODE)), \
130 NULL, &proc_single_file_operations, \
131 { .proc_show = show } )
134 * Count the number of hardlinks for the pid_entry table, excluding the .
137 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
144 for (i = 0; i < n; ++i) {
145 if (S_ISDIR(entries[i].mode))
152 static int get_task_root(struct task_struct *task, struct path *root)
154 int result = -ENOENT;
158 get_fs_root(task->fs, root);
165 static int proc_cwd_link(struct inode *inode, struct path *path)
167 struct task_struct *task = get_proc_task(inode);
168 int result = -ENOENT;
173 get_fs_pwd(task->fs, path);
177 put_task_struct(task);
182 static int proc_root_link(struct inode *inode, struct path *path)
184 struct task_struct *task = get_proc_task(inode);
185 int result = -ENOENT;
188 result = get_task_root(task, path);
189 put_task_struct(task);
194 static struct mm_struct *__check_mem_permission(struct task_struct *task)
196 struct mm_struct *mm;
198 mm = get_task_mm(task);
200 return ERR_PTR(-EINVAL);
203 * A task can always look at itself, in case it chooses
204 * to use system calls instead of load instructions.
210 * If current is actively ptrace'ing, and would also be
211 * permitted to freshly attach with ptrace now, permit it.
213 if (task_is_stopped_or_traced(task)) {
216 match = (tracehook_tracer_task(task) == current);
218 if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH))
223 * No one else is allowed.
226 return ERR_PTR(-EPERM);
230 * If current may access user memory in @task return a reference to the
231 * corresponding mm, otherwise ERR_PTR.
233 static struct mm_struct *check_mem_permission(struct task_struct *task)
235 struct mm_struct *mm;
239 * Avoid racing if task exec's as we might get a new mm but validate
240 * against old credentials.
242 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
246 mm = __check_mem_permission(task);
247 mutex_unlock(&task->signal->cred_guard_mutex);
252 struct mm_struct *mm_for_maps(struct task_struct *task)
254 struct mm_struct *mm;
257 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
261 mm = get_task_mm(task);
262 if (mm && mm != current->mm &&
263 !ptrace_may_access(task, PTRACE_MODE_READ)) {
265 mm = ERR_PTR(-EACCES);
267 mutex_unlock(&task->signal->cred_guard_mutex);
272 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
276 struct mm_struct *mm = get_task_mm(task);
280 goto out_mm; /* Shh! No looking before we're done */
282 len = mm->arg_end - mm->arg_start;
287 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
289 // If the nul at the end of args has been overwritten, then
290 // assume application is using setproctitle(3).
291 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
292 len = strnlen(buffer, res);
296 len = mm->env_end - mm->env_start;
297 if (len > PAGE_SIZE - res)
298 len = PAGE_SIZE - res;
299 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
300 res = strnlen(buffer, res);
309 static int proc_pid_auxv(struct task_struct *task, char *buffer)
311 struct mm_struct *mm = mm_for_maps(task);
312 int res = PTR_ERR(mm);
313 if (mm && !IS_ERR(mm)) {
314 unsigned int nwords = 0;
317 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
318 res = nwords * sizeof(mm->saved_auxv[0]);
321 memcpy(buffer, mm->saved_auxv, res);
328 #ifdef CONFIG_KALLSYMS
330 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
331 * Returns the resolved symbol. If that fails, simply return the address.
333 static int proc_pid_wchan(struct task_struct *task, char *buffer)
336 char symname[KSYM_NAME_LEN];
338 wchan = get_wchan(task);
340 if (lookup_symbol_name(wchan, symname) < 0)
341 if (!ptrace_may_access(task, PTRACE_MODE_READ))
344 return sprintf(buffer, "%lu", wchan);
346 return sprintf(buffer, "%s", symname);
348 #endif /* CONFIG_KALLSYMS */
350 static int lock_trace(struct task_struct *task)
352 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
355 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
356 mutex_unlock(&task->signal->cred_guard_mutex);
362 static void unlock_trace(struct task_struct *task)
364 mutex_unlock(&task->signal->cred_guard_mutex);
367 #ifdef CONFIG_STACKTRACE
369 #define MAX_STACK_TRACE_DEPTH 64
371 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
372 struct pid *pid, struct task_struct *task)
374 struct stack_trace trace;
375 unsigned long *entries;
379 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
383 trace.nr_entries = 0;
384 trace.max_entries = MAX_STACK_TRACE_DEPTH;
385 trace.entries = entries;
388 err = lock_trace(task);
390 save_stack_trace_tsk(task, &trace);
392 for (i = 0; i < trace.nr_entries; i++) {
393 seq_printf(m, "[<%pK>] %pS\n",
394 (void *)entries[i], (void *)entries[i]);
404 #ifdef CONFIG_SCHEDSTATS
406 * Provides /proc/PID/schedstat
408 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
410 return sprintf(buffer, "%llu %llu %lu\n",
411 (unsigned long long)task->se.sum_exec_runtime,
412 (unsigned long long)task->sched_info.run_delay,
413 task->sched_info.pcount);
417 #ifdef CONFIG_LATENCYTOP
418 static int lstats_show_proc(struct seq_file *m, void *v)
421 struct inode *inode = m->private;
422 struct task_struct *task = get_proc_task(inode);
426 seq_puts(m, "Latency Top version : v0.1\n");
427 for (i = 0; i < 32; i++) {
428 struct latency_record *lr = &task->latency_record[i];
429 if (lr->backtrace[0]) {
431 seq_printf(m, "%i %li %li",
432 lr->count, lr->time, lr->max);
433 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
434 unsigned long bt = lr->backtrace[q];
439 seq_printf(m, " %ps", (void *)bt);
445 put_task_struct(task);
449 static int lstats_open(struct inode *inode, struct file *file)
451 return single_open(file, lstats_show_proc, inode);
454 static ssize_t lstats_write(struct file *file, const char __user *buf,
455 size_t count, loff_t *offs)
457 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
461 clear_all_latency_tracing(task);
462 put_task_struct(task);
467 static const struct file_operations proc_lstats_operations = {
470 .write = lstats_write,
472 .release = single_release,
477 static int proc_oom_score(struct task_struct *task, char *buffer)
479 unsigned long points = 0;
481 read_lock(&tasklist_lock);
483 points = oom_badness(task, NULL, NULL,
484 totalram_pages + total_swap_pages);
485 read_unlock(&tasklist_lock);
486 return sprintf(buffer, "%lu\n", points);
494 static const struct limit_names lnames[RLIM_NLIMITS] = {
495 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
496 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
497 [RLIMIT_DATA] = {"Max data size", "bytes"},
498 [RLIMIT_STACK] = {"Max stack size", "bytes"},
499 [RLIMIT_CORE] = {"Max core file size", "bytes"},
500 [RLIMIT_RSS] = {"Max resident set", "bytes"},
501 [RLIMIT_NPROC] = {"Max processes", "processes"},
502 [RLIMIT_NOFILE] = {"Max open files", "files"},
503 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
504 [RLIMIT_AS] = {"Max address space", "bytes"},
505 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
506 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
507 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
508 [RLIMIT_NICE] = {"Max nice priority", NULL},
509 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
510 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
513 /* Display limits for a process */
514 static int proc_pid_limits(struct task_struct *task, char *buffer)
519 char *bufptr = buffer;
521 struct rlimit rlim[RLIM_NLIMITS];
523 if (!lock_task_sighand(task, &flags))
525 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
526 unlock_task_sighand(task, &flags);
529 * print the file header
531 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
532 "Limit", "Soft Limit", "Hard Limit", "Units");
534 for (i = 0; i < RLIM_NLIMITS; i++) {
535 if (rlim[i].rlim_cur == RLIM_INFINITY)
536 count += sprintf(&bufptr[count], "%-25s %-20s ",
537 lnames[i].name, "unlimited");
539 count += sprintf(&bufptr[count], "%-25s %-20lu ",
540 lnames[i].name, rlim[i].rlim_cur);
542 if (rlim[i].rlim_max == RLIM_INFINITY)
543 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
545 count += sprintf(&bufptr[count], "%-20lu ",
549 count += sprintf(&bufptr[count], "%-10s\n",
552 count += sprintf(&bufptr[count], "\n");
558 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
559 static int proc_pid_syscall(struct task_struct *task, char *buffer)
562 unsigned long args[6], sp, pc;
563 int res = lock_trace(task);
567 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
568 res = sprintf(buffer, "running\n");
570 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
572 res = sprintf(buffer,
573 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
575 args[0], args[1], args[2], args[3], args[4], args[5],
580 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
582 /************************************************************************/
583 /* Here the fs part begins */
584 /************************************************************************/
586 /* permission checks */
587 static int proc_fd_access_allowed(struct inode *inode)
589 struct task_struct *task;
591 /* Allow access to a task's file descriptors if it is us or we
592 * may use ptrace attach to the process and find out that
595 task = get_proc_task(inode);
597 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
598 put_task_struct(task);
603 int proc_setattr(struct dentry *dentry, struct iattr *attr)
606 struct inode *inode = dentry->d_inode;
608 if (attr->ia_valid & ATTR_MODE)
611 error = inode_change_ok(inode, attr);
615 if ((attr->ia_valid & ATTR_SIZE) &&
616 attr->ia_size != i_size_read(inode)) {
617 error = vmtruncate(inode, attr->ia_size);
622 setattr_copy(inode, attr);
623 mark_inode_dirty(inode);
627 static const struct inode_operations proc_def_inode_operations = {
628 .setattr = proc_setattr,
631 static int mounts_open_common(struct inode *inode, struct file *file,
632 const struct seq_operations *op)
634 struct task_struct *task = get_proc_task(inode);
636 struct mnt_namespace *ns = NULL;
638 struct proc_mounts *p;
643 nsp = task_nsproxy(task);
650 if (ns && get_task_root(task, &root) == 0)
652 put_task_struct(task);
661 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
665 file->private_data = &p->m;
666 ret = seq_open(file, op);
673 p->event = ns->event;
687 static int mounts_release(struct inode *inode, struct file *file)
689 struct proc_mounts *p = file->private_data;
692 return seq_release(inode, file);
695 static unsigned mounts_poll(struct file *file, poll_table *wait)
697 struct proc_mounts *p = file->private_data;
698 unsigned res = POLLIN | POLLRDNORM;
700 poll_wait(file, &p->ns->poll, wait);
701 if (mnt_had_events(p))
702 res |= POLLERR | POLLPRI;
707 static int mounts_open(struct inode *inode, struct file *file)
709 return mounts_open_common(inode, file, &mounts_op);
712 static const struct file_operations proc_mounts_operations = {
716 .release = mounts_release,
720 static int mountinfo_open(struct inode *inode, struct file *file)
722 return mounts_open_common(inode, file, &mountinfo_op);
725 static const struct file_operations proc_mountinfo_operations = {
726 .open = mountinfo_open,
729 .release = mounts_release,
733 static int mountstats_open(struct inode *inode, struct file *file)
735 return mounts_open_common(inode, file, &mountstats_op);
738 static const struct file_operations proc_mountstats_operations = {
739 .open = mountstats_open,
742 .release = mounts_release,
745 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
747 static ssize_t proc_info_read(struct file * file, char __user * buf,
748 size_t count, loff_t *ppos)
750 struct inode * inode = file->f_path.dentry->d_inode;
753 struct task_struct *task = get_proc_task(inode);
759 if (count > PROC_BLOCK_SIZE)
760 count = PROC_BLOCK_SIZE;
763 if (!(page = __get_free_page(GFP_TEMPORARY)))
766 length = PROC_I(inode)->op.proc_read(task, (char*)page);
769 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
772 put_task_struct(task);
777 static const struct file_operations proc_info_file_operations = {
778 .read = proc_info_read,
779 .llseek = generic_file_llseek,
782 static int proc_single_show(struct seq_file *m, void *v)
784 struct inode *inode = m->private;
785 struct pid_namespace *ns;
787 struct task_struct *task;
790 ns = inode->i_sb->s_fs_info;
791 pid = proc_pid(inode);
792 task = get_pid_task(pid, PIDTYPE_PID);
796 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
798 put_task_struct(task);
802 static int proc_single_open(struct inode *inode, struct file *filp)
804 return single_open(filp, proc_single_show, inode);
807 static const struct file_operations proc_single_file_operations = {
808 .open = proc_single_open,
811 .release = single_release,
814 static int mem_open(struct inode* inode, struct file* file)
816 file->private_data = (void*)((long)current->self_exec_id);
817 /* OK to pass negative loff_t, we can catch out-of-range */
818 file->f_mode |= FMODE_UNSIGNED_OFFSET;
822 static ssize_t mem_read(struct file * file, char __user * buf,
823 size_t count, loff_t *ppos)
825 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
827 unsigned long src = *ppos;
829 struct mm_struct *mm;
835 page = (char *)__get_free_page(GFP_TEMPORARY);
839 mm = check_mem_permission(task);
846 if (file->private_data != (void*)((long)current->self_exec_id))
852 int this_len, retval;
854 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
855 retval = access_remote_vm(mm, src, page, this_len, 0);
862 if (copy_to_user(buf, page, retval)) {
877 free_page((unsigned long) page);
879 put_task_struct(task);
884 static ssize_t mem_write(struct file * file, const char __user *buf,
885 size_t count, loff_t *ppos)
889 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
890 unsigned long dst = *ppos;
891 struct mm_struct *mm;
897 mm = check_mem_permission(task);
898 copied = PTR_ERR(mm);
903 if (file->private_data != (void *)((long)current->self_exec_id))
907 page = (char *)__get_free_page(GFP_TEMPORARY);
913 int this_len, retval;
915 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
916 if (copy_from_user(page, buf, this_len)) {
920 retval = access_remote_vm(mm, dst, page, this_len, 1);
932 free_page((unsigned long) page);
936 put_task_struct(task);
941 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
945 file->f_pos = offset;
948 file->f_pos += offset;
953 force_successful_syscall_return();
957 static const struct file_operations proc_mem_operations = {
964 static ssize_t environ_read(struct file *file, char __user *buf,
965 size_t count, loff_t *ppos)
967 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
969 unsigned long src = *ppos;
971 struct mm_struct *mm;
977 page = (char *)__get_free_page(GFP_TEMPORARY);
982 mm = mm_for_maps(task);
984 if (!mm || IS_ERR(mm))
989 int this_len, retval, max_len;
991 this_len = mm->env_end - (mm->env_start + src);
996 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
997 this_len = (this_len > max_len) ? max_len : this_len;
999 retval = access_process_vm(task, (mm->env_start + src),
1007 if (copy_to_user(buf, page, retval)) {
1021 free_page((unsigned long) page);
1023 put_task_struct(task);
1028 static const struct file_operations proc_environ_operations = {
1029 .read = environ_read,
1030 .llseek = generic_file_llseek,
1033 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
1034 size_t count, loff_t *ppos)
1036 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1037 char buffer[PROC_NUMBUF];
1039 int oom_adjust = OOM_DISABLE;
1040 unsigned long flags;
1045 if (lock_task_sighand(task, &flags)) {
1046 oom_adjust = task->signal->oom_adj;
1047 unlock_task_sighand(task, &flags);
1050 put_task_struct(task);
1052 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
1054 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1057 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1058 size_t count, loff_t *ppos)
1060 struct task_struct *task;
1061 char buffer[PROC_NUMBUF];
1063 unsigned long flags;
1066 memset(buffer, 0, sizeof(buffer));
1067 if (count > sizeof(buffer) - 1)
1068 count = sizeof(buffer) - 1;
1069 if (copy_from_user(buffer, buf, count)) {
1074 err = strict_strtol(strstrip(buffer), 0, &oom_adjust);
1077 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
1078 oom_adjust != OOM_DISABLE) {
1083 task = get_proc_task(file->f_path.dentry->d_inode);
1095 if (!lock_task_sighand(task, &flags)) {
1100 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
1105 if (oom_adjust != task->signal->oom_adj) {
1106 if (oom_adjust == OOM_DISABLE)
1107 atomic_inc(&task->mm->oom_disable_count);
1108 if (task->signal->oom_adj == OOM_DISABLE)
1109 atomic_dec(&task->mm->oom_disable_count);
1113 * Warn that /proc/pid/oom_adj is deprecated, see
1114 * Documentation/feature-removal-schedule.txt.
1116 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, "
1117 "please use /proc/%d/oom_score_adj instead.\n",
1118 current->comm, task_pid_nr(current),
1119 task_pid_nr(task), task_pid_nr(task));
1120 task->signal->oom_adj = oom_adjust;
1122 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1123 * value is always attainable.
1125 if (task->signal->oom_adj == OOM_ADJUST_MAX)
1126 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
1128 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
1131 unlock_task_sighand(task, &flags);
1134 put_task_struct(task);
1136 return err < 0 ? err : count;
1139 static const struct file_operations proc_oom_adjust_operations = {
1140 .read = oom_adjust_read,
1141 .write = oom_adjust_write,
1142 .llseek = generic_file_llseek,
1145 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1146 size_t count, loff_t *ppos)
1148 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1149 char buffer[PROC_NUMBUF];
1150 int oom_score_adj = OOM_SCORE_ADJ_MIN;
1151 unsigned long flags;
1156 if (lock_task_sighand(task, &flags)) {
1157 oom_score_adj = task->signal->oom_score_adj;
1158 unlock_task_sighand(task, &flags);
1160 put_task_struct(task);
1161 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1162 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1165 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1166 size_t count, loff_t *ppos)
1168 struct task_struct *task;
1169 char buffer[PROC_NUMBUF];
1170 unsigned long flags;
1174 memset(buffer, 0, sizeof(buffer));
1175 if (count > sizeof(buffer) - 1)
1176 count = sizeof(buffer) - 1;
1177 if (copy_from_user(buffer, buf, count)) {
1182 err = strict_strtol(strstrip(buffer), 0, &oom_score_adj);
1185 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1186 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1191 task = get_proc_task(file->f_path.dentry->d_inode);
1203 if (!lock_task_sighand(task, &flags)) {
1208 if (oom_score_adj < task->signal->oom_score_adj_min &&
1209 !capable(CAP_SYS_RESOURCE)) {
1214 if (oom_score_adj != task->signal->oom_score_adj) {
1215 if (oom_score_adj == OOM_SCORE_ADJ_MIN)
1216 atomic_inc(&task->mm->oom_disable_count);
1217 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1218 atomic_dec(&task->mm->oom_disable_count);
1220 task->signal->oom_score_adj = oom_score_adj;
1221 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1222 task->signal->oom_score_adj_min = oom_score_adj;
1224 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1225 * always attainable.
1227 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1228 task->signal->oom_adj = OOM_DISABLE;
1230 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1233 unlock_task_sighand(task, &flags);
1236 put_task_struct(task);
1238 return err < 0 ? err : count;
1241 static const struct file_operations proc_oom_score_adj_operations = {
1242 .read = oom_score_adj_read,
1243 .write = oom_score_adj_write,
1244 .llseek = default_llseek,
1247 #ifdef CONFIG_AUDITSYSCALL
1248 #define TMPBUFLEN 21
1249 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1250 size_t count, loff_t *ppos)
1252 struct inode * inode = file->f_path.dentry->d_inode;
1253 struct task_struct *task = get_proc_task(inode);
1255 char tmpbuf[TMPBUFLEN];
1259 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1260 audit_get_loginuid(task));
1261 put_task_struct(task);
1262 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1265 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1266 size_t count, loff_t *ppos)
1268 struct inode * inode = file->f_path.dentry->d_inode;
1273 if (!capable(CAP_AUDIT_CONTROL))
1277 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1283 if (count >= PAGE_SIZE)
1284 count = PAGE_SIZE - 1;
1287 /* No partial writes. */
1290 page = (char*)__get_free_page(GFP_TEMPORARY);
1294 if (copy_from_user(page, buf, count))
1298 loginuid = simple_strtoul(page, &tmp, 10);
1304 length = audit_set_loginuid(current, loginuid);
1305 if (likely(length == 0))
1309 free_page((unsigned long) page);
1313 static const struct file_operations proc_loginuid_operations = {
1314 .read = proc_loginuid_read,
1315 .write = proc_loginuid_write,
1316 .llseek = generic_file_llseek,
1319 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1320 size_t count, loff_t *ppos)
1322 struct inode * inode = file->f_path.dentry->d_inode;
1323 struct task_struct *task = get_proc_task(inode);
1325 char tmpbuf[TMPBUFLEN];
1329 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1330 audit_get_sessionid(task));
1331 put_task_struct(task);
1332 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1335 static const struct file_operations proc_sessionid_operations = {
1336 .read = proc_sessionid_read,
1337 .llseek = generic_file_llseek,
1341 #ifdef CONFIG_FAULT_INJECTION
1342 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1343 size_t count, loff_t *ppos)
1345 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1346 char buffer[PROC_NUMBUF];
1352 make_it_fail = task->make_it_fail;
1353 put_task_struct(task);
1355 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1357 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1360 static ssize_t proc_fault_inject_write(struct file * file,
1361 const char __user * buf, size_t count, loff_t *ppos)
1363 struct task_struct *task;
1364 char buffer[PROC_NUMBUF], *end;
1367 if (!capable(CAP_SYS_RESOURCE))
1369 memset(buffer, 0, sizeof(buffer));
1370 if (count > sizeof(buffer) - 1)
1371 count = sizeof(buffer) - 1;
1372 if (copy_from_user(buffer, buf, count))
1374 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1377 task = get_proc_task(file->f_dentry->d_inode);
1380 task->make_it_fail = make_it_fail;
1381 put_task_struct(task);
1386 static const struct file_operations proc_fault_inject_operations = {
1387 .read = proc_fault_inject_read,
1388 .write = proc_fault_inject_write,
1389 .llseek = generic_file_llseek,
1394 #ifdef CONFIG_SCHED_DEBUG
1396 * Print out various scheduling related per-task fields:
1398 static int sched_show(struct seq_file *m, void *v)
1400 struct inode *inode = m->private;
1401 struct task_struct *p;
1403 p = get_proc_task(inode);
1406 proc_sched_show_task(p, m);
1414 sched_write(struct file *file, const char __user *buf,
1415 size_t count, loff_t *offset)
1417 struct inode *inode = file->f_path.dentry->d_inode;
1418 struct task_struct *p;
1420 p = get_proc_task(inode);
1423 proc_sched_set_task(p);
1430 static int sched_open(struct inode *inode, struct file *filp)
1432 return single_open(filp, sched_show, inode);
1435 static const struct file_operations proc_pid_sched_operations = {
1438 .write = sched_write,
1439 .llseek = seq_lseek,
1440 .release = single_release,
1445 #ifdef CONFIG_SCHED_AUTOGROUP
1447 * Print out autogroup related information:
1449 static int sched_autogroup_show(struct seq_file *m, void *v)
1451 struct inode *inode = m->private;
1452 struct task_struct *p;
1454 p = get_proc_task(inode);
1457 proc_sched_autogroup_show_task(p, m);
1465 sched_autogroup_write(struct file *file, const char __user *buf,
1466 size_t count, loff_t *offset)
1468 struct inode *inode = file->f_path.dentry->d_inode;
1469 struct task_struct *p;
1470 char buffer[PROC_NUMBUF];
1474 memset(buffer, 0, sizeof(buffer));
1475 if (count > sizeof(buffer) - 1)
1476 count = sizeof(buffer) - 1;
1477 if (copy_from_user(buffer, buf, count))
1480 err = strict_strtol(strstrip(buffer), 0, &nice);
1484 p = get_proc_task(inode);
1489 err = proc_sched_autogroup_set_nice(p, &err);
1498 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1502 ret = single_open(filp, sched_autogroup_show, NULL);
1504 struct seq_file *m = filp->private_data;
1511 static const struct file_operations proc_pid_sched_autogroup_operations = {
1512 .open = sched_autogroup_open,
1514 .write = sched_autogroup_write,
1515 .llseek = seq_lseek,
1516 .release = single_release,
1519 #endif /* CONFIG_SCHED_AUTOGROUP */
1521 static ssize_t comm_write(struct file *file, const char __user *buf,
1522 size_t count, loff_t *offset)
1524 struct inode *inode = file->f_path.dentry->d_inode;
1525 struct task_struct *p;
1526 char buffer[TASK_COMM_LEN];
1528 memset(buffer, 0, sizeof(buffer));
1529 if (count > sizeof(buffer) - 1)
1530 count = sizeof(buffer) - 1;
1531 if (copy_from_user(buffer, buf, count))
1534 p = get_proc_task(inode);
1538 if (same_thread_group(current, p))
1539 set_task_comm(p, buffer);
1548 static int comm_show(struct seq_file *m, void *v)
1550 struct inode *inode = m->private;
1551 struct task_struct *p;
1553 p = get_proc_task(inode);
1558 seq_printf(m, "%s\n", p->comm);
1566 static int comm_open(struct inode *inode, struct file *filp)
1568 return single_open(filp, comm_show, inode);
1571 static const struct file_operations proc_pid_set_comm_operations = {
1574 .write = comm_write,
1575 .llseek = seq_lseek,
1576 .release = single_release,
1580 * We added or removed a vma mapping the executable. The vmas are only mapped
1581 * during exec and are not mapped with the mmap system call.
1582 * Callers must hold down_write() on the mm's mmap_sem for these
1584 void added_exe_file_vma(struct mm_struct *mm)
1586 mm->num_exe_file_vmas++;
1589 void removed_exe_file_vma(struct mm_struct *mm)
1591 mm->num_exe_file_vmas--;
1592 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
1594 mm->exe_file = NULL;
1599 void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1602 get_file(new_exe_file);
1605 mm->exe_file = new_exe_file;
1606 mm->num_exe_file_vmas = 0;
1609 struct file *get_mm_exe_file(struct mm_struct *mm)
1611 struct file *exe_file;
1613 /* We need mmap_sem to protect against races with removal of
1614 * VM_EXECUTABLE vmas */
1615 down_read(&mm->mmap_sem);
1616 exe_file = mm->exe_file;
1619 up_read(&mm->mmap_sem);
1623 void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
1625 /* It's safe to write the exe_file pointer without exe_file_lock because
1626 * this is called during fork when the task is not yet in /proc */
1627 newmm->exe_file = get_mm_exe_file(oldmm);
1630 static int proc_exe_link(struct inode *inode, struct path *exe_path)
1632 struct task_struct *task;
1633 struct mm_struct *mm;
1634 struct file *exe_file;
1636 task = get_proc_task(inode);
1639 mm = get_task_mm(task);
1640 put_task_struct(task);
1643 exe_file = get_mm_exe_file(mm);
1646 *exe_path = exe_file->f_path;
1647 path_get(&exe_file->f_path);
1654 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1656 struct inode *inode = dentry->d_inode;
1657 int error = -EACCES;
1659 /* We don't need a base pointer in the /proc filesystem */
1660 path_put(&nd->path);
1662 /* Are we allowed to snoop on the tasks file descriptors? */
1663 if (!proc_fd_access_allowed(inode))
1666 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1668 return ERR_PTR(error);
1671 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1673 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1680 pathname = d_path(path, tmp, PAGE_SIZE);
1681 len = PTR_ERR(pathname);
1682 if (IS_ERR(pathname))
1684 len = tmp + PAGE_SIZE - 1 - pathname;
1688 if (copy_to_user(buffer, pathname, len))
1691 free_page((unsigned long)tmp);
1695 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1697 int error = -EACCES;
1698 struct inode *inode = dentry->d_inode;
1701 /* Are we allowed to snoop on the tasks file descriptors? */
1702 if (!proc_fd_access_allowed(inode))
1705 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1709 error = do_proc_readlink(&path, buffer, buflen);
1715 static const struct inode_operations proc_pid_link_inode_operations = {
1716 .readlink = proc_pid_readlink,
1717 .follow_link = proc_pid_follow_link,
1718 .setattr = proc_setattr,
1722 /* building an inode */
1724 static int task_dumpable(struct task_struct *task)
1727 struct mm_struct *mm;
1732 dumpable = get_dumpable(mm);
1739 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1741 struct inode * inode;
1742 struct proc_inode *ei;
1743 const struct cred *cred;
1745 /* We need a new inode */
1747 inode = new_inode(sb);
1753 inode->i_ino = get_next_ino();
1754 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1755 inode->i_op = &proc_def_inode_operations;
1758 * grab the reference to task.
1760 ei->pid = get_task_pid(task, PIDTYPE_PID);
1764 if (task_dumpable(task)) {
1766 cred = __task_cred(task);
1767 inode->i_uid = cred->euid;
1768 inode->i_gid = cred->egid;
1771 security_task_to_inode(task, inode);
1781 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1783 struct inode *inode = dentry->d_inode;
1784 struct task_struct *task;
1785 const struct cred *cred;
1787 generic_fillattr(inode, stat);
1792 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1794 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1795 task_dumpable(task)) {
1796 cred = __task_cred(task);
1797 stat->uid = cred->euid;
1798 stat->gid = cred->egid;
1808 * Exceptional case: normally we are not allowed to unhash a busy
1809 * directory. In this case, however, we can do it - no aliasing problems
1810 * due to the way we treat inodes.
1812 * Rewrite the inode's ownerships here because the owning task may have
1813 * performed a setuid(), etc.
1815 * Before the /proc/pid/status file was created the only way to read
1816 * the effective uid of a /process was to stat /proc/pid. Reading
1817 * /proc/pid/status is slow enough that procps and other packages
1818 * kept stating /proc/pid. To keep the rules in /proc simple I have
1819 * made this apply to all per process world readable and executable
1822 int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1824 struct inode *inode;
1825 struct task_struct *task;
1826 const struct cred *cred;
1828 if (nd && nd->flags & LOOKUP_RCU)
1831 inode = dentry->d_inode;
1832 task = get_proc_task(inode);
1835 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1836 task_dumpable(task)) {
1838 cred = __task_cred(task);
1839 inode->i_uid = cred->euid;
1840 inode->i_gid = cred->egid;
1846 inode->i_mode &= ~(S_ISUID | S_ISGID);
1847 security_task_to_inode(task, inode);
1848 put_task_struct(task);
1855 static int pid_delete_dentry(const struct dentry * dentry)
1857 /* Is the task we represent dead?
1858 * If so, then don't put the dentry on the lru list,
1859 * kill it immediately.
1861 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1864 const struct dentry_operations pid_dentry_operations =
1866 .d_revalidate = pid_revalidate,
1867 .d_delete = pid_delete_dentry,
1873 * Fill a directory entry.
1875 * If possible create the dcache entry and derive our inode number and
1876 * file type from dcache entry.
1878 * Since all of the proc inode numbers are dynamically generated, the inode
1879 * numbers do not exist until the inode is cache. This means creating the
1880 * the dcache entry in readdir is necessary to keep the inode numbers
1881 * reported by readdir in sync with the inode numbers reported
1884 int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1885 const char *name, int len,
1886 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1888 struct dentry *child, *dir = filp->f_path.dentry;
1889 struct inode *inode;
1892 unsigned type = DT_UNKNOWN;
1896 qname.hash = full_name_hash(name, len);
1898 child = d_lookup(dir, &qname);
1901 new = d_alloc(dir, &qname);
1903 child = instantiate(dir->d_inode, new, task, ptr);
1910 if (!child || IS_ERR(child) || !child->d_inode)
1911 goto end_instantiate;
1912 inode = child->d_inode;
1915 type = inode->i_mode >> 12;
1920 ino = find_inode_number(dir, &qname);
1923 return filldir(dirent, name, len, filp->f_pos, ino, type);
1926 static unsigned name_to_int(struct dentry *dentry)
1928 const char *name = dentry->d_name.name;
1929 int len = dentry->d_name.len;
1932 if (len > 1 && *name == '0')
1935 unsigned c = *name++ - '0';
1938 if (n >= (~0U-9)/10)
1948 #define PROC_FDINFO_MAX 64
1950 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1952 struct task_struct *task = get_proc_task(inode);
1953 struct files_struct *files = NULL;
1955 int fd = proc_fd(inode);
1958 files = get_files_struct(task);
1959 put_task_struct(task);
1963 * We are not taking a ref to the file structure, so we must
1966 spin_lock(&files->file_lock);
1967 file = fcheck_files(files, fd);
1970 *path = file->f_path;
1971 path_get(&file->f_path);
1974 snprintf(info, PROC_FDINFO_MAX,
1977 (long long) file->f_pos,
1979 spin_unlock(&files->file_lock);
1980 put_files_struct(files);
1983 spin_unlock(&files->file_lock);
1984 put_files_struct(files);
1989 static int proc_fd_link(struct inode *inode, struct path *path)
1991 return proc_fd_info(inode, path, NULL);
1994 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1996 struct inode *inode;
1997 struct task_struct *task;
1999 struct files_struct *files;
2000 const struct cred *cred;
2002 if (nd && nd->flags & LOOKUP_RCU)
2005 inode = dentry->d_inode;
2006 task = get_proc_task(inode);
2007 fd = proc_fd(inode);
2010 files = get_files_struct(task);
2013 if (fcheck_files(files, fd)) {
2015 put_files_struct(files);
2016 if (task_dumpable(task)) {
2018 cred = __task_cred(task);
2019 inode->i_uid = cred->euid;
2020 inode->i_gid = cred->egid;
2026 inode->i_mode &= ~(S_ISUID | S_ISGID);
2027 security_task_to_inode(task, inode);
2028 put_task_struct(task);
2032 put_files_struct(files);
2034 put_task_struct(task);
2040 static const struct dentry_operations tid_fd_dentry_operations =
2042 .d_revalidate = tid_fd_revalidate,
2043 .d_delete = pid_delete_dentry,
2046 static struct dentry *proc_fd_instantiate(struct inode *dir,
2047 struct dentry *dentry, struct task_struct *task, const void *ptr)
2049 unsigned fd = *(const unsigned *)ptr;
2051 struct files_struct *files;
2052 struct inode *inode;
2053 struct proc_inode *ei;
2054 struct dentry *error = ERR_PTR(-ENOENT);
2056 inode = proc_pid_make_inode(dir->i_sb, task);
2061 files = get_files_struct(task);
2064 inode->i_mode = S_IFLNK;
2067 * We are not taking a ref to the file structure, so we must
2070 spin_lock(&files->file_lock);
2071 file = fcheck_files(files, fd);
2074 if (file->f_mode & FMODE_READ)
2075 inode->i_mode |= S_IRUSR | S_IXUSR;
2076 if (file->f_mode & FMODE_WRITE)
2077 inode->i_mode |= S_IWUSR | S_IXUSR;
2078 spin_unlock(&files->file_lock);
2079 put_files_struct(files);
2081 inode->i_op = &proc_pid_link_inode_operations;
2083 ei->op.proc_get_link = proc_fd_link;
2084 d_set_d_op(dentry, &tid_fd_dentry_operations);
2085 d_add(dentry, inode);
2086 /* Close the race of the process dying before we return the dentry */
2087 if (tid_fd_revalidate(dentry, NULL))
2093 spin_unlock(&files->file_lock);
2094 put_files_struct(files);
2100 static struct dentry *proc_lookupfd_common(struct inode *dir,
2101 struct dentry *dentry,
2102 instantiate_t instantiate)
2104 struct task_struct *task = get_proc_task(dir);
2105 unsigned fd = name_to_int(dentry);
2106 struct dentry *result = ERR_PTR(-ENOENT);
2113 result = instantiate(dir, dentry, task, &fd);
2115 put_task_struct(task);
2120 static int proc_readfd_common(struct file * filp, void * dirent,
2121 filldir_t filldir, instantiate_t instantiate)
2123 struct dentry *dentry = filp->f_path.dentry;
2124 struct inode *inode = dentry->d_inode;
2125 struct task_struct *p = get_proc_task(inode);
2126 unsigned int fd, ino;
2128 struct files_struct * files;
2138 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
2142 ino = parent_ino(dentry);
2143 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2147 files = get_files_struct(p);
2151 for (fd = filp->f_pos-2;
2152 fd < files_fdtable(files)->max_fds;
2153 fd++, filp->f_pos++) {
2154 char name[PROC_NUMBUF];
2157 if (!fcheck_files(files, fd))
2161 len = snprintf(name, sizeof(name), "%d", fd);
2162 if (proc_fill_cache(filp, dirent, filldir,
2163 name, len, instantiate,
2171 put_files_struct(files);
2179 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
2180 struct nameidata *nd)
2182 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
2185 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
2187 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
2190 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
2191 size_t len, loff_t *ppos)
2193 char tmp[PROC_FDINFO_MAX];
2194 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
2196 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
2200 static const struct file_operations proc_fdinfo_file_operations = {
2201 .open = nonseekable_open,
2202 .read = proc_fdinfo_read,
2203 .llseek = no_llseek,
2206 static const struct file_operations proc_fd_operations = {
2207 .read = generic_read_dir,
2208 .readdir = proc_readfd,
2209 .llseek = default_llseek,
2213 * /proc/pid/fd needs a special permission handler so that a process can still
2214 * access /proc/self/fd after it has executed a setuid().
2216 static int proc_fd_permission(struct inode *inode, int mask, unsigned int flags)
2220 if (flags & IPERM_FLAG_RCU)
2222 rv = generic_permission(inode, mask, flags, NULL);
2225 if (task_pid(current) == proc_pid(inode))
2231 * proc directories can do almost nothing..
2233 static const struct inode_operations proc_fd_inode_operations = {
2234 .lookup = proc_lookupfd,
2235 .permission = proc_fd_permission,
2236 .setattr = proc_setattr,
2239 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2240 struct dentry *dentry, struct task_struct *task, const void *ptr)
2242 unsigned fd = *(unsigned *)ptr;
2243 struct inode *inode;
2244 struct proc_inode *ei;
2245 struct dentry *error = ERR_PTR(-ENOENT);
2247 inode = proc_pid_make_inode(dir->i_sb, task);
2252 inode->i_mode = S_IFREG | S_IRUSR;
2253 inode->i_fop = &proc_fdinfo_file_operations;
2254 d_set_d_op(dentry, &tid_fd_dentry_operations);
2255 d_add(dentry, inode);
2256 /* Close the race of the process dying before we return the dentry */
2257 if (tid_fd_revalidate(dentry, NULL))
2264 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2265 struct dentry *dentry,
2266 struct nameidata *nd)
2268 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2271 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2273 return proc_readfd_common(filp, dirent, filldir,
2274 proc_fdinfo_instantiate);
2277 static const struct file_operations proc_fdinfo_operations = {
2278 .read = generic_read_dir,
2279 .readdir = proc_readfdinfo,
2280 .llseek = default_llseek,
2284 * proc directories can do almost nothing..
2286 static const struct inode_operations proc_fdinfo_inode_operations = {
2287 .lookup = proc_lookupfdinfo,
2288 .setattr = proc_setattr,
2292 static struct dentry *proc_pident_instantiate(struct inode *dir,
2293 struct dentry *dentry, struct task_struct *task, const void *ptr)
2295 const struct pid_entry *p = ptr;
2296 struct inode *inode;
2297 struct proc_inode *ei;
2298 struct dentry *error = ERR_PTR(-ENOENT);
2300 inode = proc_pid_make_inode(dir->i_sb, task);
2305 inode->i_mode = p->mode;
2306 if (S_ISDIR(inode->i_mode))
2307 inode->i_nlink = 2; /* Use getattr to fix if necessary */
2309 inode->i_op = p->iop;
2311 inode->i_fop = p->fop;
2313 d_set_d_op(dentry, &pid_dentry_operations);
2314 d_add(dentry, inode);
2315 /* Close the race of the process dying before we return the dentry */
2316 if (pid_revalidate(dentry, NULL))
2322 static struct dentry *proc_pident_lookup(struct inode *dir,
2323 struct dentry *dentry,
2324 const struct pid_entry *ents,
2327 struct dentry *error;
2328 struct task_struct *task = get_proc_task(dir);
2329 const struct pid_entry *p, *last;
2331 error = ERR_PTR(-ENOENT);
2337 * Yes, it does not scale. And it should not. Don't add
2338 * new entries into /proc/<tgid>/ without very good reasons.
2340 last = &ents[nents - 1];
2341 for (p = ents; p <= last; p++) {
2342 if (p->len != dentry->d_name.len)
2344 if (!memcmp(dentry->d_name.name, p->name, p->len))
2350 error = proc_pident_instantiate(dir, dentry, task, p);
2352 put_task_struct(task);
2357 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2358 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2360 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2361 proc_pident_instantiate, task, p);
2364 static int proc_pident_readdir(struct file *filp,
2365 void *dirent, filldir_t filldir,
2366 const struct pid_entry *ents, unsigned int nents)
2369 struct dentry *dentry = filp->f_path.dentry;
2370 struct inode *inode = dentry->d_inode;
2371 struct task_struct *task = get_proc_task(inode);
2372 const struct pid_entry *p, *last;
2385 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2391 ino = parent_ino(dentry);
2392 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2404 last = &ents[nents - 1];
2406 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2415 put_task_struct(task);
2420 #ifdef CONFIG_SECURITY
2421 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2422 size_t count, loff_t *ppos)
2424 struct inode * inode = file->f_path.dentry->d_inode;
2427 struct task_struct *task = get_proc_task(inode);
2432 length = security_getprocattr(task,
2433 (char*)file->f_path.dentry->d_name.name,
2435 put_task_struct(task);
2437 length = simple_read_from_buffer(buf, count, ppos, p, length);
2442 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2443 size_t count, loff_t *ppos)
2445 struct inode * inode = file->f_path.dentry->d_inode;
2448 struct task_struct *task = get_proc_task(inode);
2453 if (count > PAGE_SIZE)
2456 /* No partial writes. */
2462 page = (char*)__get_free_page(GFP_TEMPORARY);
2467 if (copy_from_user(page, buf, count))
2470 /* Guard against adverse ptrace interaction */
2471 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2475 length = security_setprocattr(task,
2476 (char*)file->f_path.dentry->d_name.name,
2477 (void*)page, count);
2478 mutex_unlock(&task->signal->cred_guard_mutex);
2480 free_page((unsigned long) page);
2482 put_task_struct(task);
2487 static const struct file_operations proc_pid_attr_operations = {
2488 .read = proc_pid_attr_read,
2489 .write = proc_pid_attr_write,
2490 .llseek = generic_file_llseek,
2493 static const struct pid_entry attr_dir_stuff[] = {
2494 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2495 REG("prev", S_IRUGO, proc_pid_attr_operations),
2496 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2497 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2498 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2499 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2502 static int proc_attr_dir_readdir(struct file * filp,
2503 void * dirent, filldir_t filldir)
2505 return proc_pident_readdir(filp,dirent,filldir,
2506 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2509 static const struct file_operations proc_attr_dir_operations = {
2510 .read = generic_read_dir,
2511 .readdir = proc_attr_dir_readdir,
2512 .llseek = default_llseek,
2515 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2516 struct dentry *dentry, struct nameidata *nd)
2518 return proc_pident_lookup(dir, dentry,
2519 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2522 static const struct inode_operations proc_attr_dir_inode_operations = {
2523 .lookup = proc_attr_dir_lookup,
2524 .getattr = pid_getattr,
2525 .setattr = proc_setattr,
2530 #ifdef CONFIG_ELF_CORE
2531 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2532 size_t count, loff_t *ppos)
2534 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2535 struct mm_struct *mm;
2536 char buffer[PROC_NUMBUF];
2544 mm = get_task_mm(task);
2546 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2547 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2548 MMF_DUMP_FILTER_SHIFT));
2550 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2553 put_task_struct(task);
2558 static ssize_t proc_coredump_filter_write(struct file *file,
2559 const char __user *buf,
2563 struct task_struct *task;
2564 struct mm_struct *mm;
2565 char buffer[PROC_NUMBUF], *end;
2572 memset(buffer, 0, sizeof(buffer));
2573 if (count > sizeof(buffer) - 1)
2574 count = sizeof(buffer) - 1;
2575 if (copy_from_user(buffer, buf, count))
2579 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2582 if (end - buffer == 0)
2586 task = get_proc_task(file->f_dentry->d_inode);
2591 mm = get_task_mm(task);
2595 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2597 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2599 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2604 put_task_struct(task);
2609 static const struct file_operations proc_coredump_filter_operations = {
2610 .read = proc_coredump_filter_read,
2611 .write = proc_coredump_filter_write,
2612 .llseek = generic_file_llseek,
2619 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2622 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2623 pid_t tgid = task_tgid_nr_ns(current, ns);
2624 char tmp[PROC_NUMBUF];
2627 sprintf(tmp, "%d", tgid);
2628 return vfs_readlink(dentry,buffer,buflen,tmp);
2631 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2633 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2634 pid_t tgid = task_tgid_nr_ns(current, ns);
2635 char *name = ERR_PTR(-ENOENT);
2639 name = ERR_PTR(-ENOMEM);
2641 sprintf(name, "%d", tgid);
2643 nd_set_link(nd, name);
2647 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2650 char *s = nd_get_link(nd);
2655 static const struct inode_operations proc_self_inode_operations = {
2656 .readlink = proc_self_readlink,
2657 .follow_link = proc_self_follow_link,
2658 .put_link = proc_self_put_link,
2664 * These are the directory entries in the root directory of /proc
2665 * that properly belong to the /proc filesystem, as they describe
2666 * describe something that is process related.
2668 static const struct pid_entry proc_base_stuff[] = {
2669 NOD("self", S_IFLNK|S_IRWXUGO,
2670 &proc_self_inode_operations, NULL, {}),
2673 static struct dentry *proc_base_instantiate(struct inode *dir,
2674 struct dentry *dentry, struct task_struct *task, const void *ptr)
2676 const struct pid_entry *p = ptr;
2677 struct inode *inode;
2678 struct proc_inode *ei;
2679 struct dentry *error;
2681 /* Allocate the inode */
2682 error = ERR_PTR(-ENOMEM);
2683 inode = new_inode(dir->i_sb);
2687 /* Initialize the inode */
2689 inode->i_ino = get_next_ino();
2690 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2693 * grab the reference to the task.
2695 ei->pid = get_task_pid(task, PIDTYPE_PID);
2699 inode->i_mode = p->mode;
2700 if (S_ISDIR(inode->i_mode))
2702 if (S_ISLNK(inode->i_mode))
2705 inode->i_op = p->iop;
2707 inode->i_fop = p->fop;
2709 d_add(dentry, inode);
2718 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2720 struct dentry *error;
2721 struct task_struct *task = get_proc_task(dir);
2722 const struct pid_entry *p, *last;
2724 error = ERR_PTR(-ENOENT);
2729 /* Lookup the directory entry */
2730 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2731 for (p = proc_base_stuff; p <= last; p++) {
2732 if (p->len != dentry->d_name.len)
2734 if (!memcmp(dentry->d_name.name, p->name, p->len))
2740 error = proc_base_instantiate(dir, dentry, task, p);
2743 put_task_struct(task);
2748 static int proc_base_fill_cache(struct file *filp, void *dirent,
2749 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2751 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2752 proc_base_instantiate, task, p);
2755 #ifdef CONFIG_TASK_IO_ACCOUNTING
2756 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2758 struct task_io_accounting acct = task->ioac;
2759 unsigned long flags;
2761 if (whole && lock_task_sighand(task, &flags)) {
2762 struct task_struct *t = task;
2764 task_io_accounting_add(&acct, &task->signal->ioac);
2765 while_each_thread(task, t)
2766 task_io_accounting_add(&acct, &t->ioac);
2768 unlock_task_sighand(task, &flags);
2770 return sprintf(buffer,
2775 "read_bytes: %llu\n"
2776 "write_bytes: %llu\n"
2777 "cancelled_write_bytes: %llu\n",
2778 (unsigned long long)acct.rchar,
2779 (unsigned long long)acct.wchar,
2780 (unsigned long long)acct.syscr,
2781 (unsigned long long)acct.syscw,
2782 (unsigned long long)acct.read_bytes,
2783 (unsigned long long)acct.write_bytes,
2784 (unsigned long long)acct.cancelled_write_bytes);
2787 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2789 return do_io_accounting(task, buffer, 0);
2792 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2794 return do_io_accounting(task, buffer, 1);
2796 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2798 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2799 struct pid *pid, struct task_struct *task)
2801 int err = lock_trace(task);
2803 seq_printf(m, "%08x\n", task->personality);
2812 static const struct file_operations proc_task_operations;
2813 static const struct inode_operations proc_task_inode_operations;
2815 static const struct pid_entry tgid_base_stuff[] = {
2816 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2817 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2818 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2819 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2821 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2823 REG("environ", S_IRUSR, proc_environ_operations),
2824 INF("auxv", S_IRUSR, proc_pid_auxv),
2825 ONE("status", S_IRUGO, proc_pid_status),
2826 ONE("personality", S_IRUGO, proc_pid_personality),
2827 INF("limits", S_IRUGO, proc_pid_limits),
2828 #ifdef CONFIG_SCHED_DEBUG
2829 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2831 #ifdef CONFIG_SCHED_AUTOGROUP
2832 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2834 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2835 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2836 INF("syscall", S_IRUGO, proc_pid_syscall),
2838 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2839 ONE("stat", S_IRUGO, proc_tgid_stat),
2840 ONE("statm", S_IRUGO, proc_pid_statm),
2841 REG("maps", S_IRUGO, proc_maps_operations),
2843 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2845 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2846 LNK("cwd", proc_cwd_link),
2847 LNK("root", proc_root_link),
2848 LNK("exe", proc_exe_link),
2849 REG("mounts", S_IRUGO, proc_mounts_operations),
2850 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2851 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2852 #ifdef CONFIG_PROC_PAGE_MONITOR
2853 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2854 REG("smaps", S_IRUGO, proc_smaps_operations),
2855 REG("pagemap", S_IRUGO, proc_pagemap_operations),
2857 #ifdef CONFIG_SECURITY
2858 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2860 #ifdef CONFIG_KALLSYMS
2861 INF("wchan", S_IRUGO, proc_pid_wchan),
2863 #ifdef CONFIG_STACKTRACE
2864 ONE("stack", S_IRUGO, proc_pid_stack),
2866 #ifdef CONFIG_SCHEDSTATS
2867 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2869 #ifdef CONFIG_LATENCYTOP
2870 REG("latency", S_IRUGO, proc_lstats_operations),
2872 #ifdef CONFIG_PROC_PID_CPUSET
2873 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2875 #ifdef CONFIG_CGROUPS
2876 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2878 INF("oom_score", S_IRUGO, proc_oom_score),
2879 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2880 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2881 #ifdef CONFIG_AUDITSYSCALL
2882 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2883 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2885 #ifdef CONFIG_FAULT_INJECTION
2886 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2888 #ifdef CONFIG_ELF_CORE
2889 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2891 #ifdef CONFIG_TASK_IO_ACCOUNTING
2892 INF("io", S_IRUGO, proc_tgid_io_accounting),
2896 static int proc_tgid_base_readdir(struct file * filp,
2897 void * dirent, filldir_t filldir)
2899 return proc_pident_readdir(filp,dirent,filldir,
2900 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2903 static const struct file_operations proc_tgid_base_operations = {
2904 .read = generic_read_dir,
2905 .readdir = proc_tgid_base_readdir,
2906 .llseek = default_llseek,
2909 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2910 return proc_pident_lookup(dir, dentry,
2911 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2914 static const struct inode_operations proc_tgid_base_inode_operations = {
2915 .lookup = proc_tgid_base_lookup,
2916 .getattr = pid_getattr,
2917 .setattr = proc_setattr,
2920 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2922 struct dentry *dentry, *leader, *dir;
2923 char buf[PROC_NUMBUF];
2927 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2928 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2930 shrink_dcache_parent(dentry);
2936 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2937 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2942 name.len = strlen(name.name);
2943 dir = d_hash_and_lookup(leader, &name);
2945 goto out_put_leader;
2948 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2949 dentry = d_hash_and_lookup(dir, &name);
2951 shrink_dcache_parent(dentry);
2964 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2965 * @task: task that should be flushed.
2967 * When flushing dentries from proc, one needs to flush them from global
2968 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2969 * in. This call is supposed to do all of this job.
2971 * Looks in the dcache for
2973 * /proc/@tgid/task/@pid
2974 * if either directory is present flushes it and all of it'ts children
2977 * It is safe and reasonable to cache /proc entries for a task until
2978 * that task exits. After that they just clog up the dcache with
2979 * useless entries, possibly causing useful dcache entries to be
2980 * flushed instead. This routine is proved to flush those useless
2981 * dcache entries at process exit time.
2983 * NOTE: This routine is just an optimization so it does not guarantee
2984 * that no dcache entries will exist at process exit time it
2985 * just makes it very unlikely that any will persist.
2988 void proc_flush_task(struct task_struct *task)
2991 struct pid *pid, *tgid;
2994 pid = task_pid(task);
2995 tgid = task_tgid(task);
2997 for (i = 0; i <= pid->level; i++) {
2998 upid = &pid->numbers[i];
2999 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
3000 tgid->numbers[i].nr);
3003 upid = &pid->numbers[pid->level];
3005 pid_ns_release_proc(upid->ns);
3008 static struct dentry *proc_pid_instantiate(struct inode *dir,
3009 struct dentry * dentry,
3010 struct task_struct *task, const void *ptr)
3012 struct dentry *error = ERR_PTR(-ENOENT);
3013 struct inode *inode;
3015 inode = proc_pid_make_inode(dir->i_sb, task);
3019 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3020 inode->i_op = &proc_tgid_base_inode_operations;
3021 inode->i_fop = &proc_tgid_base_operations;
3022 inode->i_flags|=S_IMMUTABLE;
3024 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
3025 ARRAY_SIZE(tgid_base_stuff));
3027 d_set_d_op(dentry, &pid_dentry_operations);
3029 d_add(dentry, inode);
3030 /* Close the race of the process dying before we return the dentry */
3031 if (pid_revalidate(dentry, NULL))
3037 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3039 struct dentry *result;
3040 struct task_struct *task;
3042 struct pid_namespace *ns;
3044 result = proc_base_lookup(dir, dentry);
3045 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3048 tgid = name_to_int(dentry);
3052 ns = dentry->d_sb->s_fs_info;
3054 task = find_task_by_pid_ns(tgid, ns);
3056 get_task_struct(task);
3061 result = proc_pid_instantiate(dir, dentry, task, NULL);
3062 put_task_struct(task);
3068 * Find the first task with tgid >= tgid
3073 struct task_struct *task;
3075 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3080 put_task_struct(iter.task);
3084 pid = find_ge_pid(iter.tgid, ns);
3086 iter.tgid = pid_nr_ns(pid, ns);
3087 iter.task = pid_task(pid, PIDTYPE_PID);
3088 /* What we to know is if the pid we have find is the
3089 * pid of a thread_group_leader. Testing for task
3090 * being a thread_group_leader is the obvious thing
3091 * todo but there is a window when it fails, due to
3092 * the pid transfer logic in de_thread.
3094 * So we perform the straight forward test of seeing
3095 * if the pid we have found is the pid of a thread
3096 * group leader, and don't worry if the task we have
3097 * found doesn't happen to be a thread group leader.
3098 * As we don't care in the case of readdir.
3100 if (!iter.task || !has_group_leader_pid(iter.task)) {
3104 get_task_struct(iter.task);
3110 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
3112 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3113 struct tgid_iter iter)
3115 char name[PROC_NUMBUF];
3116 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
3117 return proc_fill_cache(filp, dirent, filldir, name, len,
3118 proc_pid_instantiate, iter.task, NULL);
3121 /* for the /proc/ directory itself, after non-process stuff has been done */
3122 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3125 struct task_struct *reaper;
3126 struct tgid_iter iter;
3127 struct pid_namespace *ns;
3129 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3131 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3133 reaper = get_proc_task(filp->f_path.dentry->d_inode);
3137 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
3138 const struct pid_entry *p = &proc_base_stuff[nr];
3139 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
3143 ns = filp->f_dentry->d_sb->s_fs_info;
3145 iter.tgid = filp->f_pos - TGID_OFFSET;
3146 for (iter = next_tgid(ns, iter);
3148 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3149 filp->f_pos = iter.tgid + TGID_OFFSET;
3150 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
3151 put_task_struct(iter.task);
3155 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3157 put_task_struct(reaper);
3165 static const struct pid_entry tid_base_stuff[] = {
3166 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3167 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3168 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3169 REG("environ", S_IRUSR, proc_environ_operations),
3170 INF("auxv", S_IRUSR, proc_pid_auxv),
3171 ONE("status", S_IRUGO, proc_pid_status),
3172 ONE("personality", S_IRUGO, proc_pid_personality),
3173 INF("limits", S_IRUGO, proc_pid_limits),
3174 #ifdef CONFIG_SCHED_DEBUG
3175 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3177 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3178 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3179 INF("syscall", S_IRUGO, proc_pid_syscall),
3181 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3182 ONE("stat", S_IRUGO, proc_tid_stat),
3183 ONE("statm", S_IRUGO, proc_pid_statm),
3184 REG("maps", S_IRUGO, proc_maps_operations),
3186 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
3188 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3189 LNK("cwd", proc_cwd_link),
3190 LNK("root", proc_root_link),
3191 LNK("exe", proc_exe_link),
3192 REG("mounts", S_IRUGO, proc_mounts_operations),
3193 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3194 #ifdef CONFIG_PROC_PAGE_MONITOR
3195 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3196 REG("smaps", S_IRUGO, proc_smaps_operations),
3197 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3199 #ifdef CONFIG_SECURITY
3200 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3202 #ifdef CONFIG_KALLSYMS
3203 INF("wchan", S_IRUGO, proc_pid_wchan),
3205 #ifdef CONFIG_STACKTRACE
3206 ONE("stack", S_IRUGO, proc_pid_stack),
3208 #ifdef CONFIG_SCHEDSTATS
3209 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3211 #ifdef CONFIG_LATENCYTOP
3212 REG("latency", S_IRUGO, proc_lstats_operations),
3214 #ifdef CONFIG_PROC_PID_CPUSET
3215 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3217 #ifdef CONFIG_CGROUPS
3218 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3220 INF("oom_score", S_IRUGO, proc_oom_score),
3221 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3222 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3223 #ifdef CONFIG_AUDITSYSCALL
3224 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3225 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3227 #ifdef CONFIG_FAULT_INJECTION
3228 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3230 #ifdef CONFIG_TASK_IO_ACCOUNTING
3231 INF("io", S_IRUGO, proc_tid_io_accounting),
3235 static int proc_tid_base_readdir(struct file * filp,
3236 void * dirent, filldir_t filldir)
3238 return proc_pident_readdir(filp,dirent,filldir,
3239 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3242 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3243 return proc_pident_lookup(dir, dentry,
3244 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3247 static const struct file_operations proc_tid_base_operations = {
3248 .read = generic_read_dir,
3249 .readdir = proc_tid_base_readdir,
3250 .llseek = default_llseek,
3253 static const struct inode_operations proc_tid_base_inode_operations = {
3254 .lookup = proc_tid_base_lookup,
3255 .getattr = pid_getattr,
3256 .setattr = proc_setattr,
3259 static struct dentry *proc_task_instantiate(struct inode *dir,
3260 struct dentry *dentry, struct task_struct *task, const void *ptr)
3262 struct dentry *error = ERR_PTR(-ENOENT);
3263 struct inode *inode;
3264 inode = proc_pid_make_inode(dir->i_sb, task);
3268 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3269 inode->i_op = &proc_tid_base_inode_operations;
3270 inode->i_fop = &proc_tid_base_operations;
3271 inode->i_flags|=S_IMMUTABLE;
3273 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
3274 ARRAY_SIZE(tid_base_stuff));
3276 d_set_d_op(dentry, &pid_dentry_operations);
3278 d_add(dentry, inode);
3279 /* Close the race of the process dying before we return the dentry */
3280 if (pid_revalidate(dentry, NULL))
3286 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3288 struct dentry *result = ERR_PTR(-ENOENT);
3289 struct task_struct *task;
3290 struct task_struct *leader = get_proc_task(dir);
3292 struct pid_namespace *ns;
3297 tid = name_to_int(dentry);
3301 ns = dentry->d_sb->s_fs_info;
3303 task = find_task_by_pid_ns(tid, ns);
3305 get_task_struct(task);
3309 if (!same_thread_group(leader, task))
3312 result = proc_task_instantiate(dir, dentry, task, NULL);
3314 put_task_struct(task);
3316 put_task_struct(leader);
3322 * Find the first tid of a thread group to return to user space.
3324 * Usually this is just the thread group leader, but if the users
3325 * buffer was too small or there was a seek into the middle of the
3326 * directory we have more work todo.
3328 * In the case of a short read we start with find_task_by_pid.
3330 * In the case of a seek we start with the leader and walk nr
3333 static struct task_struct *first_tid(struct task_struct *leader,
3334 int tid, int nr, struct pid_namespace *ns)
3336 struct task_struct *pos;
3339 /* Attempt to start with the pid of a thread */
3340 if (tid && (nr > 0)) {
3341 pos = find_task_by_pid_ns(tid, ns);
3342 if (pos && (pos->group_leader == leader))
3346 /* If nr exceeds the number of threads there is nothing todo */
3348 if (nr && nr >= get_nr_threads(leader))
3351 /* If we haven't found our starting place yet start
3352 * with the leader and walk nr threads forward.
3354 for (pos = leader; nr > 0; --nr) {
3355 pos = next_thread(pos);
3356 if (pos == leader) {
3362 get_task_struct(pos);
3369 * Find the next thread in the thread list.
3370 * Return NULL if there is an error or no next thread.
3372 * The reference to the input task_struct is released.
3374 static struct task_struct *next_tid(struct task_struct *start)
3376 struct task_struct *pos = NULL;
3378 if (pid_alive(start)) {
3379 pos = next_thread(start);
3380 if (thread_group_leader(pos))
3383 get_task_struct(pos);
3386 put_task_struct(start);
3390 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3391 struct task_struct *task, int tid)
3393 char name[PROC_NUMBUF];
3394 int len = snprintf(name, sizeof(name), "%d", tid);
3395 return proc_fill_cache(filp, dirent, filldir, name, len,
3396 proc_task_instantiate, task, NULL);
3399 /* for the /proc/TGID/task/ directories */
3400 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3402 struct dentry *dentry = filp->f_path.dentry;
3403 struct inode *inode = dentry->d_inode;
3404 struct task_struct *leader = NULL;
3405 struct task_struct *task;
3406 int retval = -ENOENT;
3409 struct pid_namespace *ns;
3411 task = get_proc_task(inode);
3415 if (pid_alive(task)) {
3416 leader = task->group_leader;
3417 get_task_struct(leader);
3420 put_task_struct(task);
3425 switch ((unsigned long)filp->f_pos) {
3428 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3433 ino = parent_ino(dentry);
3434 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3440 /* f_version caches the tgid value that the last readdir call couldn't
3441 * return. lseek aka telldir automagically resets f_version to 0.
3443 ns = filp->f_dentry->d_sb->s_fs_info;
3444 tid = (int)filp->f_version;
3445 filp->f_version = 0;
3446 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3448 task = next_tid(task), filp->f_pos++) {
3449 tid = task_pid_nr_ns(task, ns);
3450 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3451 /* returning this tgid failed, save it as the first
3452 * pid for the next readir call */
3453 filp->f_version = (u64)tid;
3454 put_task_struct(task);
3459 put_task_struct(leader);
3464 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3466 struct inode *inode = dentry->d_inode;
3467 struct task_struct *p = get_proc_task(inode);
3468 generic_fillattr(inode, stat);
3471 stat->nlink += get_nr_threads(p);
3478 static const struct inode_operations proc_task_inode_operations = {
3479 .lookup = proc_task_lookup,
3480 .getattr = proc_task_getattr,
3481 .setattr = proc_setattr,
3484 static const struct file_operations proc_task_operations = {
3485 .read = generic_read_dir,
3486 .readdir = proc_task_readdir,
3487 .llseek = default_llseek,