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);
195 * Return zero if current may access user memory in @task, -error if not.
197 static int check_mem_permission(struct task_struct *task)
200 * A task can always look at itself, in case it chooses
201 * to use system calls instead of load instructions.
207 * If current is actively ptrace'ing, and would also be
208 * permitted to freshly attach with ptrace now, permit it.
210 if (task_is_stopped_or_traced(task)) {
213 match = (tracehook_tracer_task(task) == current);
215 if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH))
220 * Noone else is allowed.
225 struct mm_struct *mm_for_maps(struct task_struct *task)
227 struct mm_struct *mm;
230 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
234 mm = get_task_mm(task);
235 if (mm && mm != current->mm &&
236 !ptrace_may_access(task, PTRACE_MODE_READ)) {
238 mm = ERR_PTR(-EACCES);
240 mutex_unlock(&task->signal->cred_guard_mutex);
245 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
249 struct mm_struct *mm = get_task_mm(task);
253 goto out_mm; /* Shh! No looking before we're done */
255 len = mm->arg_end - mm->arg_start;
260 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
262 // If the nul at the end of args has been overwritten, then
263 // assume application is using setproctitle(3).
264 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
265 len = strnlen(buffer, res);
269 len = mm->env_end - mm->env_start;
270 if (len > PAGE_SIZE - res)
271 len = PAGE_SIZE - res;
272 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
273 res = strnlen(buffer, res);
282 static int proc_pid_auxv(struct task_struct *task, char *buffer)
284 struct mm_struct *mm = mm_for_maps(task);
285 int res = PTR_ERR(mm);
286 if (mm && !IS_ERR(mm)) {
287 unsigned int nwords = 0;
290 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
291 res = nwords * sizeof(mm->saved_auxv[0]);
294 memcpy(buffer, mm->saved_auxv, res);
301 #ifdef CONFIG_KALLSYMS
303 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
304 * Returns the resolved symbol. If that fails, simply return the address.
306 static int proc_pid_wchan(struct task_struct *task, char *buffer)
309 char symname[KSYM_NAME_LEN];
311 wchan = get_wchan(task);
313 if (lookup_symbol_name(wchan, symname) < 0)
314 if (!ptrace_may_access(task, PTRACE_MODE_READ))
317 return sprintf(buffer, "%lu", wchan);
319 return sprintf(buffer, "%s", symname);
321 #endif /* CONFIG_KALLSYMS */
323 static int lock_trace(struct task_struct *task)
325 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
328 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
329 mutex_unlock(&task->signal->cred_guard_mutex);
335 static void unlock_trace(struct task_struct *task)
337 mutex_unlock(&task->signal->cred_guard_mutex);
340 #ifdef CONFIG_STACKTRACE
342 #define MAX_STACK_TRACE_DEPTH 64
344 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
345 struct pid *pid, struct task_struct *task)
347 struct stack_trace trace;
348 unsigned long *entries;
352 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
356 trace.nr_entries = 0;
357 trace.max_entries = MAX_STACK_TRACE_DEPTH;
358 trace.entries = entries;
361 err = lock_trace(task);
363 save_stack_trace_tsk(task, &trace);
365 for (i = 0; i < trace.nr_entries; i++) {
366 seq_printf(m, "[<%pK>] %pS\n",
367 (void *)entries[i], (void *)entries[i]);
377 #ifdef CONFIG_SCHEDSTATS
379 * Provides /proc/PID/schedstat
381 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
383 return sprintf(buffer, "%llu %llu %lu\n",
384 (unsigned long long)task->se.sum_exec_runtime,
385 (unsigned long long)task->sched_info.run_delay,
386 task->sched_info.pcount);
390 #ifdef CONFIG_LATENCYTOP
391 static int lstats_show_proc(struct seq_file *m, void *v)
394 struct inode *inode = m->private;
395 struct task_struct *task = get_proc_task(inode);
399 seq_puts(m, "Latency Top version : v0.1\n");
400 for (i = 0; i < 32; i++) {
401 struct latency_record *lr = &task->latency_record[i];
402 if (lr->backtrace[0]) {
404 seq_printf(m, "%i %li %li",
405 lr->count, lr->time, lr->max);
406 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
407 unsigned long bt = lr->backtrace[q];
412 seq_printf(m, " %ps", (void *)bt);
418 put_task_struct(task);
422 static int lstats_open(struct inode *inode, struct file *file)
424 return single_open(file, lstats_show_proc, inode);
427 static ssize_t lstats_write(struct file *file, const char __user *buf,
428 size_t count, loff_t *offs)
430 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
434 clear_all_latency_tracing(task);
435 put_task_struct(task);
440 static const struct file_operations proc_lstats_operations = {
443 .write = lstats_write,
445 .release = single_release,
450 static int proc_oom_score(struct task_struct *task, char *buffer)
452 unsigned long points = 0;
454 read_lock(&tasklist_lock);
456 points = oom_badness(task, NULL, NULL,
457 totalram_pages + total_swap_pages);
458 read_unlock(&tasklist_lock);
459 return sprintf(buffer, "%lu\n", points);
467 static const struct limit_names lnames[RLIM_NLIMITS] = {
468 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
469 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
470 [RLIMIT_DATA] = {"Max data size", "bytes"},
471 [RLIMIT_STACK] = {"Max stack size", "bytes"},
472 [RLIMIT_CORE] = {"Max core file size", "bytes"},
473 [RLIMIT_RSS] = {"Max resident set", "bytes"},
474 [RLIMIT_NPROC] = {"Max processes", "processes"},
475 [RLIMIT_NOFILE] = {"Max open files", "files"},
476 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
477 [RLIMIT_AS] = {"Max address space", "bytes"},
478 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
479 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
480 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
481 [RLIMIT_NICE] = {"Max nice priority", NULL},
482 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
483 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
486 /* Display limits for a process */
487 static int proc_pid_limits(struct task_struct *task, char *buffer)
492 char *bufptr = buffer;
494 struct rlimit rlim[RLIM_NLIMITS];
496 if (!lock_task_sighand(task, &flags))
498 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
499 unlock_task_sighand(task, &flags);
502 * print the file header
504 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
505 "Limit", "Soft Limit", "Hard Limit", "Units");
507 for (i = 0; i < RLIM_NLIMITS; i++) {
508 if (rlim[i].rlim_cur == RLIM_INFINITY)
509 count += sprintf(&bufptr[count], "%-25s %-20s ",
510 lnames[i].name, "unlimited");
512 count += sprintf(&bufptr[count], "%-25s %-20lu ",
513 lnames[i].name, rlim[i].rlim_cur);
515 if (rlim[i].rlim_max == RLIM_INFINITY)
516 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
518 count += sprintf(&bufptr[count], "%-20lu ",
522 count += sprintf(&bufptr[count], "%-10s\n",
525 count += sprintf(&bufptr[count], "\n");
531 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
532 static int proc_pid_syscall(struct task_struct *task, char *buffer)
535 unsigned long args[6], sp, pc;
536 int res = lock_trace(task);
540 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
541 res = sprintf(buffer, "running\n");
543 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
545 res = sprintf(buffer,
546 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
548 args[0], args[1], args[2], args[3], args[4], args[5],
553 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
555 /************************************************************************/
556 /* Here the fs part begins */
557 /************************************************************************/
559 /* permission checks */
560 static int proc_fd_access_allowed(struct inode *inode)
562 struct task_struct *task;
564 /* Allow access to a task's file descriptors if it is us or we
565 * may use ptrace attach to the process and find out that
568 task = get_proc_task(inode);
570 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
571 put_task_struct(task);
576 static int proc_setattr(struct dentry *dentry, struct iattr *attr)
579 struct inode *inode = dentry->d_inode;
581 if (attr->ia_valid & ATTR_MODE)
584 error = inode_change_ok(inode, attr);
588 if ((attr->ia_valid & ATTR_SIZE) &&
589 attr->ia_size != i_size_read(inode)) {
590 error = vmtruncate(inode, attr->ia_size);
595 setattr_copy(inode, attr);
596 mark_inode_dirty(inode);
600 static const struct inode_operations proc_def_inode_operations = {
601 .setattr = proc_setattr,
604 static int mounts_open_common(struct inode *inode, struct file *file,
605 const struct seq_operations *op)
607 struct task_struct *task = get_proc_task(inode);
609 struct mnt_namespace *ns = NULL;
611 struct proc_mounts *p;
616 nsp = task_nsproxy(task);
623 if (ns && get_task_root(task, &root) == 0)
625 put_task_struct(task);
634 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
638 file->private_data = &p->m;
639 ret = seq_open(file, op);
646 p->event = ns->event;
660 static int mounts_release(struct inode *inode, struct file *file)
662 struct proc_mounts *p = file->private_data;
665 return seq_release(inode, file);
668 static unsigned mounts_poll(struct file *file, poll_table *wait)
670 struct proc_mounts *p = file->private_data;
671 unsigned res = POLLIN | POLLRDNORM;
673 poll_wait(file, &p->ns->poll, wait);
674 if (mnt_had_events(p))
675 res |= POLLERR | POLLPRI;
680 static int mounts_open(struct inode *inode, struct file *file)
682 return mounts_open_common(inode, file, &mounts_op);
685 static const struct file_operations proc_mounts_operations = {
689 .release = mounts_release,
693 static int mountinfo_open(struct inode *inode, struct file *file)
695 return mounts_open_common(inode, file, &mountinfo_op);
698 static const struct file_operations proc_mountinfo_operations = {
699 .open = mountinfo_open,
702 .release = mounts_release,
706 static int mountstats_open(struct inode *inode, struct file *file)
708 return mounts_open_common(inode, file, &mountstats_op);
711 static const struct file_operations proc_mountstats_operations = {
712 .open = mountstats_open,
715 .release = mounts_release,
718 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
720 static ssize_t proc_info_read(struct file * file, char __user * buf,
721 size_t count, loff_t *ppos)
723 struct inode * inode = file->f_path.dentry->d_inode;
726 struct task_struct *task = get_proc_task(inode);
732 if (count > PROC_BLOCK_SIZE)
733 count = PROC_BLOCK_SIZE;
736 if (!(page = __get_free_page(GFP_TEMPORARY)))
739 length = PROC_I(inode)->op.proc_read(task, (char*)page);
742 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
745 put_task_struct(task);
750 static const struct file_operations proc_info_file_operations = {
751 .read = proc_info_read,
752 .llseek = generic_file_llseek,
755 static int proc_single_show(struct seq_file *m, void *v)
757 struct inode *inode = m->private;
758 struct pid_namespace *ns;
760 struct task_struct *task;
763 ns = inode->i_sb->s_fs_info;
764 pid = proc_pid(inode);
765 task = get_pid_task(pid, PIDTYPE_PID);
769 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
771 put_task_struct(task);
775 static int proc_single_open(struct inode *inode, struct file *filp)
777 return single_open(filp, proc_single_show, inode);
780 static const struct file_operations proc_single_file_operations = {
781 .open = proc_single_open,
784 .release = single_release,
787 static int mem_open(struct inode* inode, struct file* file)
789 file->private_data = (void*)((long)current->self_exec_id);
790 /* OK to pass negative loff_t, we can catch out-of-range */
791 file->f_mode |= FMODE_UNSIGNED_OFFSET;
795 static ssize_t mem_read(struct file * file, char __user * buf,
796 size_t count, loff_t *ppos)
798 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
800 unsigned long src = *ppos;
802 struct mm_struct *mm;
807 if (check_mem_permission(task))
811 page = (char *)__get_free_page(GFP_TEMPORARY);
817 mm = get_task_mm(task);
823 if (file->private_data != (void*)((long)current->self_exec_id))
829 int this_len, retval;
831 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
832 retval = access_process_vm(task, src, page, this_len, 0);
833 if (!retval || check_mem_permission(task)) {
839 if (copy_to_user(buf, page, retval)) {
854 free_page((unsigned long) page);
856 put_task_struct(task);
861 #define mem_write NULL
864 /* This is a security hazard */
865 static ssize_t mem_write(struct file * file, const char __user *buf,
866 size_t count, loff_t *ppos)
870 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
871 unsigned long dst = *ppos;
877 if (check_mem_permission(task))
881 page = (char *)__get_free_page(GFP_TEMPORARY);
887 int this_len, retval;
889 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
890 if (copy_from_user(page, buf, this_len)) {
894 retval = access_process_vm(task, dst, page, this_len, 1);
906 free_page((unsigned long) page);
908 put_task_struct(task);
914 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
918 file->f_pos = offset;
921 file->f_pos += offset;
926 force_successful_syscall_return();
930 static const struct file_operations proc_mem_operations = {
937 static ssize_t environ_read(struct file *file, char __user *buf,
938 size_t count, loff_t *ppos)
940 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
942 unsigned long src = *ppos;
944 struct mm_struct *mm;
950 page = (char *)__get_free_page(GFP_TEMPORARY);
955 mm = mm_for_maps(task);
957 if (!mm || IS_ERR(mm))
962 int this_len, retval, max_len;
964 this_len = mm->env_end - (mm->env_start + src);
969 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
970 this_len = (this_len > max_len) ? max_len : this_len;
972 retval = access_process_vm(task, (mm->env_start + src),
980 if (copy_to_user(buf, page, retval)) {
994 free_page((unsigned long) page);
996 put_task_struct(task);
1001 static const struct file_operations proc_environ_operations = {
1002 .read = environ_read,
1003 .llseek = generic_file_llseek,
1006 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
1007 size_t count, loff_t *ppos)
1009 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1010 char buffer[PROC_NUMBUF];
1012 int oom_adjust = OOM_DISABLE;
1013 unsigned long flags;
1018 if (lock_task_sighand(task, &flags)) {
1019 oom_adjust = task->signal->oom_adj;
1020 unlock_task_sighand(task, &flags);
1023 put_task_struct(task);
1025 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
1027 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1030 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1031 size_t count, loff_t *ppos)
1033 struct task_struct *task;
1034 char buffer[PROC_NUMBUF];
1036 unsigned long flags;
1039 memset(buffer, 0, sizeof(buffer));
1040 if (count > sizeof(buffer) - 1)
1041 count = sizeof(buffer) - 1;
1042 if (copy_from_user(buffer, buf, count)) {
1047 err = strict_strtol(strstrip(buffer), 0, &oom_adjust);
1050 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
1051 oom_adjust != OOM_DISABLE) {
1056 task = get_proc_task(file->f_path.dentry->d_inode);
1068 if (!lock_task_sighand(task, &flags)) {
1073 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
1078 if (oom_adjust != task->signal->oom_adj) {
1079 if (oom_adjust == OOM_DISABLE)
1080 atomic_inc(&task->mm->oom_disable_count);
1081 if (task->signal->oom_adj == OOM_DISABLE)
1082 atomic_dec(&task->mm->oom_disable_count);
1086 * Warn that /proc/pid/oom_adj is deprecated, see
1087 * Documentation/feature-removal-schedule.txt.
1089 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, "
1090 "please use /proc/%d/oom_score_adj instead.\n",
1091 current->comm, task_pid_nr(current),
1092 task_pid_nr(task), task_pid_nr(task));
1093 task->signal->oom_adj = oom_adjust;
1095 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1096 * value is always attainable.
1098 if (task->signal->oom_adj == OOM_ADJUST_MAX)
1099 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
1101 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
1104 unlock_task_sighand(task, &flags);
1107 put_task_struct(task);
1109 return err < 0 ? err : count;
1112 static const struct file_operations proc_oom_adjust_operations = {
1113 .read = oom_adjust_read,
1114 .write = oom_adjust_write,
1115 .llseek = generic_file_llseek,
1118 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1119 size_t count, loff_t *ppos)
1121 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1122 char buffer[PROC_NUMBUF];
1123 int oom_score_adj = OOM_SCORE_ADJ_MIN;
1124 unsigned long flags;
1129 if (lock_task_sighand(task, &flags)) {
1130 oom_score_adj = task->signal->oom_score_adj;
1131 unlock_task_sighand(task, &flags);
1133 put_task_struct(task);
1134 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1135 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1138 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1139 size_t count, loff_t *ppos)
1141 struct task_struct *task;
1142 char buffer[PROC_NUMBUF];
1143 unsigned long flags;
1147 memset(buffer, 0, sizeof(buffer));
1148 if (count > sizeof(buffer) - 1)
1149 count = sizeof(buffer) - 1;
1150 if (copy_from_user(buffer, buf, count)) {
1155 err = strict_strtol(strstrip(buffer), 0, &oom_score_adj);
1158 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1159 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1164 task = get_proc_task(file->f_path.dentry->d_inode);
1176 if (!lock_task_sighand(task, &flags)) {
1181 if (oom_score_adj < task->signal->oom_score_adj_min &&
1182 !capable(CAP_SYS_RESOURCE)) {
1187 if (oom_score_adj != task->signal->oom_score_adj) {
1188 if (oom_score_adj == OOM_SCORE_ADJ_MIN)
1189 atomic_inc(&task->mm->oom_disable_count);
1190 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1191 atomic_dec(&task->mm->oom_disable_count);
1193 task->signal->oom_score_adj = oom_score_adj;
1194 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1195 task->signal->oom_score_adj_min = oom_score_adj;
1197 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1198 * always attainable.
1200 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1201 task->signal->oom_adj = OOM_DISABLE;
1203 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1206 unlock_task_sighand(task, &flags);
1209 put_task_struct(task);
1211 return err < 0 ? err : count;
1214 static const struct file_operations proc_oom_score_adj_operations = {
1215 .read = oom_score_adj_read,
1216 .write = oom_score_adj_write,
1217 .llseek = default_llseek,
1220 #ifdef CONFIG_AUDITSYSCALL
1221 #define TMPBUFLEN 21
1222 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1223 size_t count, loff_t *ppos)
1225 struct inode * inode = file->f_path.dentry->d_inode;
1226 struct task_struct *task = get_proc_task(inode);
1228 char tmpbuf[TMPBUFLEN];
1232 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1233 audit_get_loginuid(task));
1234 put_task_struct(task);
1235 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1238 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1239 size_t count, loff_t *ppos)
1241 struct inode * inode = file->f_path.dentry->d_inode;
1246 if (!capable(CAP_AUDIT_CONTROL))
1250 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1256 if (count >= PAGE_SIZE)
1257 count = PAGE_SIZE - 1;
1260 /* No partial writes. */
1263 page = (char*)__get_free_page(GFP_TEMPORARY);
1267 if (copy_from_user(page, buf, count))
1271 loginuid = simple_strtoul(page, &tmp, 10);
1277 length = audit_set_loginuid(current, loginuid);
1278 if (likely(length == 0))
1282 free_page((unsigned long) page);
1286 static const struct file_operations proc_loginuid_operations = {
1287 .read = proc_loginuid_read,
1288 .write = proc_loginuid_write,
1289 .llseek = generic_file_llseek,
1292 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1293 size_t count, loff_t *ppos)
1295 struct inode * inode = file->f_path.dentry->d_inode;
1296 struct task_struct *task = get_proc_task(inode);
1298 char tmpbuf[TMPBUFLEN];
1302 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1303 audit_get_sessionid(task));
1304 put_task_struct(task);
1305 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1308 static const struct file_operations proc_sessionid_operations = {
1309 .read = proc_sessionid_read,
1310 .llseek = generic_file_llseek,
1314 #ifdef CONFIG_FAULT_INJECTION
1315 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1316 size_t count, loff_t *ppos)
1318 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1319 char buffer[PROC_NUMBUF];
1325 make_it_fail = task->make_it_fail;
1326 put_task_struct(task);
1328 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1330 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1333 static ssize_t proc_fault_inject_write(struct file * file,
1334 const char __user * buf, size_t count, loff_t *ppos)
1336 struct task_struct *task;
1337 char buffer[PROC_NUMBUF], *end;
1340 if (!capable(CAP_SYS_RESOURCE))
1342 memset(buffer, 0, sizeof(buffer));
1343 if (count > sizeof(buffer) - 1)
1344 count = sizeof(buffer) - 1;
1345 if (copy_from_user(buffer, buf, count))
1347 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1350 task = get_proc_task(file->f_dentry->d_inode);
1353 task->make_it_fail = make_it_fail;
1354 put_task_struct(task);
1359 static const struct file_operations proc_fault_inject_operations = {
1360 .read = proc_fault_inject_read,
1361 .write = proc_fault_inject_write,
1362 .llseek = generic_file_llseek,
1367 #ifdef CONFIG_SCHED_DEBUG
1369 * Print out various scheduling related per-task fields:
1371 static int sched_show(struct seq_file *m, void *v)
1373 struct inode *inode = m->private;
1374 struct task_struct *p;
1376 p = get_proc_task(inode);
1379 proc_sched_show_task(p, m);
1387 sched_write(struct file *file, const char __user *buf,
1388 size_t count, loff_t *offset)
1390 struct inode *inode = file->f_path.dentry->d_inode;
1391 struct task_struct *p;
1393 p = get_proc_task(inode);
1396 proc_sched_set_task(p);
1403 static int sched_open(struct inode *inode, struct file *filp)
1405 return single_open(filp, sched_show, inode);
1408 static const struct file_operations proc_pid_sched_operations = {
1411 .write = sched_write,
1412 .llseek = seq_lseek,
1413 .release = single_release,
1418 #ifdef CONFIG_SCHED_AUTOGROUP
1420 * Print out autogroup related information:
1422 static int sched_autogroup_show(struct seq_file *m, void *v)
1424 struct inode *inode = m->private;
1425 struct task_struct *p;
1427 p = get_proc_task(inode);
1430 proc_sched_autogroup_show_task(p, m);
1438 sched_autogroup_write(struct file *file, const char __user *buf,
1439 size_t count, loff_t *offset)
1441 struct inode *inode = file->f_path.dentry->d_inode;
1442 struct task_struct *p;
1443 char buffer[PROC_NUMBUF];
1447 memset(buffer, 0, sizeof(buffer));
1448 if (count > sizeof(buffer) - 1)
1449 count = sizeof(buffer) - 1;
1450 if (copy_from_user(buffer, buf, count))
1453 err = strict_strtol(strstrip(buffer), 0, &nice);
1457 p = get_proc_task(inode);
1462 err = proc_sched_autogroup_set_nice(p, &err);
1471 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1475 ret = single_open(filp, sched_autogroup_show, NULL);
1477 struct seq_file *m = filp->private_data;
1484 static const struct file_operations proc_pid_sched_autogroup_operations = {
1485 .open = sched_autogroup_open,
1487 .write = sched_autogroup_write,
1488 .llseek = seq_lseek,
1489 .release = single_release,
1492 #endif /* CONFIG_SCHED_AUTOGROUP */
1494 static ssize_t comm_write(struct file *file, const char __user *buf,
1495 size_t count, loff_t *offset)
1497 struct inode *inode = file->f_path.dentry->d_inode;
1498 struct task_struct *p;
1499 char buffer[TASK_COMM_LEN];
1501 memset(buffer, 0, sizeof(buffer));
1502 if (count > sizeof(buffer) - 1)
1503 count = sizeof(buffer) - 1;
1504 if (copy_from_user(buffer, buf, count))
1507 p = get_proc_task(inode);
1511 if (same_thread_group(current, p))
1512 set_task_comm(p, buffer);
1521 static int comm_show(struct seq_file *m, void *v)
1523 struct inode *inode = m->private;
1524 struct task_struct *p;
1526 p = get_proc_task(inode);
1531 seq_printf(m, "%s\n", p->comm);
1539 static int comm_open(struct inode *inode, struct file *filp)
1541 return single_open(filp, comm_show, inode);
1544 static const struct file_operations proc_pid_set_comm_operations = {
1547 .write = comm_write,
1548 .llseek = seq_lseek,
1549 .release = single_release,
1553 * We added or removed a vma mapping the executable. The vmas are only mapped
1554 * during exec and are not mapped with the mmap system call.
1555 * Callers must hold down_write() on the mm's mmap_sem for these
1557 void added_exe_file_vma(struct mm_struct *mm)
1559 mm->num_exe_file_vmas++;
1562 void removed_exe_file_vma(struct mm_struct *mm)
1564 mm->num_exe_file_vmas--;
1565 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
1567 mm->exe_file = NULL;
1572 void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1575 get_file(new_exe_file);
1578 mm->exe_file = new_exe_file;
1579 mm->num_exe_file_vmas = 0;
1582 struct file *get_mm_exe_file(struct mm_struct *mm)
1584 struct file *exe_file;
1586 /* We need mmap_sem to protect against races with removal of
1587 * VM_EXECUTABLE vmas */
1588 down_read(&mm->mmap_sem);
1589 exe_file = mm->exe_file;
1592 up_read(&mm->mmap_sem);
1596 void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
1598 /* It's safe to write the exe_file pointer without exe_file_lock because
1599 * this is called during fork when the task is not yet in /proc */
1600 newmm->exe_file = get_mm_exe_file(oldmm);
1603 static int proc_exe_link(struct inode *inode, struct path *exe_path)
1605 struct task_struct *task;
1606 struct mm_struct *mm;
1607 struct file *exe_file;
1609 task = get_proc_task(inode);
1612 mm = get_task_mm(task);
1613 put_task_struct(task);
1616 exe_file = get_mm_exe_file(mm);
1619 *exe_path = exe_file->f_path;
1620 path_get(&exe_file->f_path);
1627 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1629 struct inode *inode = dentry->d_inode;
1630 int error = -EACCES;
1632 /* We don't need a base pointer in the /proc filesystem */
1633 path_put(&nd->path);
1635 /* Are we allowed to snoop on the tasks file descriptors? */
1636 if (!proc_fd_access_allowed(inode))
1639 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1641 return ERR_PTR(error);
1644 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1646 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1653 pathname = d_path(path, tmp, PAGE_SIZE);
1654 len = PTR_ERR(pathname);
1655 if (IS_ERR(pathname))
1657 len = tmp + PAGE_SIZE - 1 - pathname;
1661 if (copy_to_user(buffer, pathname, len))
1664 free_page((unsigned long)tmp);
1668 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1670 int error = -EACCES;
1671 struct inode *inode = dentry->d_inode;
1674 /* Are we allowed to snoop on the tasks file descriptors? */
1675 if (!proc_fd_access_allowed(inode))
1678 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1682 error = do_proc_readlink(&path, buffer, buflen);
1688 static const struct inode_operations proc_pid_link_inode_operations = {
1689 .readlink = proc_pid_readlink,
1690 .follow_link = proc_pid_follow_link,
1691 .setattr = proc_setattr,
1695 /* building an inode */
1697 static int task_dumpable(struct task_struct *task)
1700 struct mm_struct *mm;
1705 dumpable = get_dumpable(mm);
1713 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1715 struct inode * inode;
1716 struct proc_inode *ei;
1717 const struct cred *cred;
1719 /* We need a new inode */
1721 inode = new_inode(sb);
1727 inode->i_ino = get_next_ino();
1728 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1729 inode->i_op = &proc_def_inode_operations;
1732 * grab the reference to task.
1734 ei->pid = get_task_pid(task, PIDTYPE_PID);
1738 if (task_dumpable(task)) {
1740 cred = __task_cred(task);
1741 inode->i_uid = cred->euid;
1742 inode->i_gid = cred->egid;
1745 security_task_to_inode(task, inode);
1755 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1757 struct inode *inode = dentry->d_inode;
1758 struct task_struct *task;
1759 const struct cred *cred;
1761 generic_fillattr(inode, stat);
1766 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1768 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1769 task_dumpable(task)) {
1770 cred = __task_cred(task);
1771 stat->uid = cred->euid;
1772 stat->gid = cred->egid;
1782 * Exceptional case: normally we are not allowed to unhash a busy
1783 * directory. In this case, however, we can do it - no aliasing problems
1784 * due to the way we treat inodes.
1786 * Rewrite the inode's ownerships here because the owning task may have
1787 * performed a setuid(), etc.
1789 * Before the /proc/pid/status file was created the only way to read
1790 * the effective uid of a /process was to stat /proc/pid. Reading
1791 * /proc/pid/status is slow enough that procps and other packages
1792 * kept stating /proc/pid. To keep the rules in /proc simple I have
1793 * made this apply to all per process world readable and executable
1796 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1798 struct inode *inode;
1799 struct task_struct *task;
1800 const struct cred *cred;
1802 if (nd && nd->flags & LOOKUP_RCU)
1805 inode = dentry->d_inode;
1806 task = get_proc_task(inode);
1809 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1810 task_dumpable(task)) {
1812 cred = __task_cred(task);
1813 inode->i_uid = cred->euid;
1814 inode->i_gid = cred->egid;
1820 inode->i_mode &= ~(S_ISUID | S_ISGID);
1821 security_task_to_inode(task, inode);
1822 put_task_struct(task);
1829 static int pid_delete_dentry(const struct dentry * dentry)
1831 /* Is the task we represent dead?
1832 * If so, then don't put the dentry on the lru list,
1833 * kill it immediately.
1835 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1838 static const struct dentry_operations pid_dentry_operations =
1840 .d_revalidate = pid_revalidate,
1841 .d_delete = pid_delete_dentry,
1846 typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1847 struct task_struct *, const void *);
1850 * Fill a directory entry.
1852 * If possible create the dcache entry and derive our inode number and
1853 * file type from dcache entry.
1855 * Since all of the proc inode numbers are dynamically generated, the inode
1856 * numbers do not exist until the inode is cache. This means creating the
1857 * the dcache entry in readdir is necessary to keep the inode numbers
1858 * reported by readdir in sync with the inode numbers reported
1861 static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1862 char *name, int len,
1863 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1865 struct dentry *child, *dir = filp->f_path.dentry;
1866 struct inode *inode;
1869 unsigned type = DT_UNKNOWN;
1873 qname.hash = full_name_hash(name, len);
1875 child = d_lookup(dir, &qname);
1878 new = d_alloc(dir, &qname);
1880 child = instantiate(dir->d_inode, new, task, ptr);
1887 if (!child || IS_ERR(child) || !child->d_inode)
1888 goto end_instantiate;
1889 inode = child->d_inode;
1892 type = inode->i_mode >> 12;
1897 ino = find_inode_number(dir, &qname);
1900 return filldir(dirent, name, len, filp->f_pos, ino, type);
1903 static unsigned name_to_int(struct dentry *dentry)
1905 const char *name = dentry->d_name.name;
1906 int len = dentry->d_name.len;
1909 if (len > 1 && *name == '0')
1912 unsigned c = *name++ - '0';
1915 if (n >= (~0U-9)/10)
1925 #define PROC_FDINFO_MAX 64
1927 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1929 struct task_struct *task = get_proc_task(inode);
1930 struct files_struct *files = NULL;
1932 int fd = proc_fd(inode);
1935 files = get_files_struct(task);
1936 put_task_struct(task);
1940 * We are not taking a ref to the file structure, so we must
1943 spin_lock(&files->file_lock);
1944 file = fcheck_files(files, fd);
1947 *path = file->f_path;
1948 path_get(&file->f_path);
1951 snprintf(info, PROC_FDINFO_MAX,
1954 (long long) file->f_pos,
1956 spin_unlock(&files->file_lock);
1957 put_files_struct(files);
1960 spin_unlock(&files->file_lock);
1961 put_files_struct(files);
1966 static int proc_fd_link(struct inode *inode, struct path *path)
1968 return proc_fd_info(inode, path, NULL);
1971 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1973 struct inode *inode;
1974 struct task_struct *task;
1976 struct files_struct *files;
1977 const struct cred *cred;
1979 if (nd && nd->flags & LOOKUP_RCU)
1982 inode = dentry->d_inode;
1983 task = get_proc_task(inode);
1984 fd = proc_fd(inode);
1987 files = get_files_struct(task);
1990 if (fcheck_files(files, fd)) {
1992 put_files_struct(files);
1993 if (task_dumpable(task)) {
1995 cred = __task_cred(task);
1996 inode->i_uid = cred->euid;
1997 inode->i_gid = cred->egid;
2003 inode->i_mode &= ~(S_ISUID | S_ISGID);
2004 security_task_to_inode(task, inode);
2005 put_task_struct(task);
2009 put_files_struct(files);
2011 put_task_struct(task);
2017 static const struct dentry_operations tid_fd_dentry_operations =
2019 .d_revalidate = tid_fd_revalidate,
2020 .d_delete = pid_delete_dentry,
2023 static struct dentry *proc_fd_instantiate(struct inode *dir,
2024 struct dentry *dentry, struct task_struct *task, const void *ptr)
2026 unsigned fd = *(const unsigned *)ptr;
2028 struct files_struct *files;
2029 struct inode *inode;
2030 struct proc_inode *ei;
2031 struct dentry *error = ERR_PTR(-ENOENT);
2033 inode = proc_pid_make_inode(dir->i_sb, task);
2038 files = get_files_struct(task);
2041 inode->i_mode = S_IFLNK;
2044 * We are not taking a ref to the file structure, so we must
2047 spin_lock(&files->file_lock);
2048 file = fcheck_files(files, fd);
2051 if (file->f_mode & FMODE_READ)
2052 inode->i_mode |= S_IRUSR | S_IXUSR;
2053 if (file->f_mode & FMODE_WRITE)
2054 inode->i_mode |= S_IWUSR | S_IXUSR;
2055 spin_unlock(&files->file_lock);
2056 put_files_struct(files);
2058 inode->i_op = &proc_pid_link_inode_operations;
2060 ei->op.proc_get_link = proc_fd_link;
2061 d_set_d_op(dentry, &tid_fd_dentry_operations);
2062 d_add(dentry, inode);
2063 /* Close the race of the process dying before we return the dentry */
2064 if (tid_fd_revalidate(dentry, NULL))
2070 spin_unlock(&files->file_lock);
2071 put_files_struct(files);
2077 static struct dentry *proc_lookupfd_common(struct inode *dir,
2078 struct dentry *dentry,
2079 instantiate_t instantiate)
2081 struct task_struct *task = get_proc_task(dir);
2082 unsigned fd = name_to_int(dentry);
2083 struct dentry *result = ERR_PTR(-ENOENT);
2090 result = instantiate(dir, dentry, task, &fd);
2092 put_task_struct(task);
2097 static int proc_readfd_common(struct file * filp, void * dirent,
2098 filldir_t filldir, instantiate_t instantiate)
2100 struct dentry *dentry = filp->f_path.dentry;
2101 struct inode *inode = dentry->d_inode;
2102 struct task_struct *p = get_proc_task(inode);
2103 unsigned int fd, ino;
2105 struct files_struct * files;
2115 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
2119 ino = parent_ino(dentry);
2120 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2124 files = get_files_struct(p);
2128 for (fd = filp->f_pos-2;
2129 fd < files_fdtable(files)->max_fds;
2130 fd++, filp->f_pos++) {
2131 char name[PROC_NUMBUF];
2134 if (!fcheck_files(files, fd))
2138 len = snprintf(name, sizeof(name), "%d", fd);
2139 if (proc_fill_cache(filp, dirent, filldir,
2140 name, len, instantiate,
2148 put_files_struct(files);
2156 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
2157 struct nameidata *nd)
2159 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
2162 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
2164 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
2167 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
2168 size_t len, loff_t *ppos)
2170 char tmp[PROC_FDINFO_MAX];
2171 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
2173 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
2177 static const struct file_operations proc_fdinfo_file_operations = {
2178 .open = nonseekable_open,
2179 .read = proc_fdinfo_read,
2180 .llseek = no_llseek,
2183 static const struct file_operations proc_fd_operations = {
2184 .read = generic_read_dir,
2185 .readdir = proc_readfd,
2186 .llseek = default_llseek,
2190 * /proc/pid/fd needs a special permission handler so that a process can still
2191 * access /proc/self/fd after it has executed a setuid().
2193 static int proc_fd_permission(struct inode *inode, int mask, unsigned int flags)
2197 if (flags & IPERM_FLAG_RCU)
2199 rv = generic_permission(inode, mask, flags, NULL);
2202 if (task_pid(current) == proc_pid(inode))
2208 * proc directories can do almost nothing..
2210 static const struct inode_operations proc_fd_inode_operations = {
2211 .lookup = proc_lookupfd,
2212 .permission = proc_fd_permission,
2213 .setattr = proc_setattr,
2216 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2217 struct dentry *dentry, struct task_struct *task, const void *ptr)
2219 unsigned fd = *(unsigned *)ptr;
2220 struct inode *inode;
2221 struct proc_inode *ei;
2222 struct dentry *error = ERR_PTR(-ENOENT);
2224 inode = proc_pid_make_inode(dir->i_sb, task);
2229 inode->i_mode = S_IFREG | S_IRUSR;
2230 inode->i_fop = &proc_fdinfo_file_operations;
2231 d_set_d_op(dentry, &tid_fd_dentry_operations);
2232 d_add(dentry, inode);
2233 /* Close the race of the process dying before we return the dentry */
2234 if (tid_fd_revalidate(dentry, NULL))
2241 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2242 struct dentry *dentry,
2243 struct nameidata *nd)
2245 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2248 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2250 return proc_readfd_common(filp, dirent, filldir,
2251 proc_fdinfo_instantiate);
2254 static const struct file_operations proc_fdinfo_operations = {
2255 .read = generic_read_dir,
2256 .readdir = proc_readfdinfo,
2257 .llseek = default_llseek,
2261 * proc directories can do almost nothing..
2263 static const struct inode_operations proc_fdinfo_inode_operations = {
2264 .lookup = proc_lookupfdinfo,
2265 .setattr = proc_setattr,
2269 static struct dentry *proc_pident_instantiate(struct inode *dir,
2270 struct dentry *dentry, struct task_struct *task, const void *ptr)
2272 const struct pid_entry *p = ptr;
2273 struct inode *inode;
2274 struct proc_inode *ei;
2275 struct dentry *error = ERR_PTR(-ENOENT);
2277 inode = proc_pid_make_inode(dir->i_sb, task);
2282 inode->i_mode = p->mode;
2283 if (S_ISDIR(inode->i_mode))
2284 inode->i_nlink = 2; /* Use getattr to fix if necessary */
2286 inode->i_op = p->iop;
2288 inode->i_fop = p->fop;
2290 d_set_d_op(dentry, &pid_dentry_operations);
2291 d_add(dentry, inode);
2292 /* Close the race of the process dying before we return the dentry */
2293 if (pid_revalidate(dentry, NULL))
2299 static struct dentry *proc_pident_lookup(struct inode *dir,
2300 struct dentry *dentry,
2301 const struct pid_entry *ents,
2304 struct dentry *error;
2305 struct task_struct *task = get_proc_task(dir);
2306 const struct pid_entry *p, *last;
2308 error = ERR_PTR(-ENOENT);
2314 * Yes, it does not scale. And it should not. Don't add
2315 * new entries into /proc/<tgid>/ without very good reasons.
2317 last = &ents[nents - 1];
2318 for (p = ents; p <= last; p++) {
2319 if (p->len != dentry->d_name.len)
2321 if (!memcmp(dentry->d_name.name, p->name, p->len))
2327 error = proc_pident_instantiate(dir, dentry, task, p);
2329 put_task_struct(task);
2334 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2335 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2337 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2338 proc_pident_instantiate, task, p);
2341 static int proc_pident_readdir(struct file *filp,
2342 void *dirent, filldir_t filldir,
2343 const struct pid_entry *ents, unsigned int nents)
2346 struct dentry *dentry = filp->f_path.dentry;
2347 struct inode *inode = dentry->d_inode;
2348 struct task_struct *task = get_proc_task(inode);
2349 const struct pid_entry *p, *last;
2362 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2368 ino = parent_ino(dentry);
2369 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2381 last = &ents[nents - 1];
2383 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2392 put_task_struct(task);
2397 #ifdef CONFIG_SECURITY
2398 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2399 size_t count, loff_t *ppos)
2401 struct inode * inode = file->f_path.dentry->d_inode;
2404 struct task_struct *task = get_proc_task(inode);
2409 length = security_getprocattr(task,
2410 (char*)file->f_path.dentry->d_name.name,
2412 put_task_struct(task);
2414 length = simple_read_from_buffer(buf, count, ppos, p, length);
2419 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2420 size_t count, loff_t *ppos)
2422 struct inode * inode = file->f_path.dentry->d_inode;
2425 struct task_struct *task = get_proc_task(inode);
2430 if (count > PAGE_SIZE)
2433 /* No partial writes. */
2439 page = (char*)__get_free_page(GFP_TEMPORARY);
2444 if (copy_from_user(page, buf, count))
2447 /* Guard against adverse ptrace interaction */
2448 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2452 length = security_setprocattr(task,
2453 (char*)file->f_path.dentry->d_name.name,
2454 (void*)page, count);
2455 mutex_unlock(&task->signal->cred_guard_mutex);
2457 free_page((unsigned long) page);
2459 put_task_struct(task);
2464 static const struct file_operations proc_pid_attr_operations = {
2465 .read = proc_pid_attr_read,
2466 .write = proc_pid_attr_write,
2467 .llseek = generic_file_llseek,
2470 static const struct pid_entry attr_dir_stuff[] = {
2471 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2472 REG("prev", S_IRUGO, proc_pid_attr_operations),
2473 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2474 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2475 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2476 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2479 static int proc_attr_dir_readdir(struct file * filp,
2480 void * dirent, filldir_t filldir)
2482 return proc_pident_readdir(filp,dirent,filldir,
2483 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2486 static const struct file_operations proc_attr_dir_operations = {
2487 .read = generic_read_dir,
2488 .readdir = proc_attr_dir_readdir,
2489 .llseek = default_llseek,
2492 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2493 struct dentry *dentry, struct nameidata *nd)
2495 return proc_pident_lookup(dir, dentry,
2496 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2499 static const struct inode_operations proc_attr_dir_inode_operations = {
2500 .lookup = proc_attr_dir_lookup,
2501 .getattr = pid_getattr,
2502 .setattr = proc_setattr,
2507 #ifdef CONFIG_ELF_CORE
2508 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2509 size_t count, loff_t *ppos)
2511 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2512 struct mm_struct *mm;
2513 char buffer[PROC_NUMBUF];
2521 mm = get_task_mm(task);
2523 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2524 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2525 MMF_DUMP_FILTER_SHIFT));
2527 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2530 put_task_struct(task);
2535 static ssize_t proc_coredump_filter_write(struct file *file,
2536 const char __user *buf,
2540 struct task_struct *task;
2541 struct mm_struct *mm;
2542 char buffer[PROC_NUMBUF], *end;
2549 memset(buffer, 0, sizeof(buffer));
2550 if (count > sizeof(buffer) - 1)
2551 count = sizeof(buffer) - 1;
2552 if (copy_from_user(buffer, buf, count))
2556 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2559 if (end - buffer == 0)
2563 task = get_proc_task(file->f_dentry->d_inode);
2568 mm = get_task_mm(task);
2572 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2574 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2576 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2581 put_task_struct(task);
2586 static const struct file_operations proc_coredump_filter_operations = {
2587 .read = proc_coredump_filter_read,
2588 .write = proc_coredump_filter_write,
2589 .llseek = generic_file_llseek,
2596 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2599 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2600 pid_t tgid = task_tgid_nr_ns(current, ns);
2601 char tmp[PROC_NUMBUF];
2604 sprintf(tmp, "%d", tgid);
2605 return vfs_readlink(dentry,buffer,buflen,tmp);
2608 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2610 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2611 pid_t tgid = task_tgid_nr_ns(current, ns);
2612 char *name = ERR_PTR(-ENOENT);
2616 name = ERR_PTR(-ENOMEM);
2618 sprintf(name, "%d", tgid);
2620 nd_set_link(nd, name);
2624 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2627 char *s = nd_get_link(nd);
2632 static const struct inode_operations proc_self_inode_operations = {
2633 .readlink = proc_self_readlink,
2634 .follow_link = proc_self_follow_link,
2635 .put_link = proc_self_put_link,
2641 * These are the directory entries in the root directory of /proc
2642 * that properly belong to the /proc filesystem, as they describe
2643 * describe something that is process related.
2645 static const struct pid_entry proc_base_stuff[] = {
2646 NOD("self", S_IFLNK|S_IRWXUGO,
2647 &proc_self_inode_operations, NULL, {}),
2650 static struct dentry *proc_base_instantiate(struct inode *dir,
2651 struct dentry *dentry, struct task_struct *task, const void *ptr)
2653 const struct pid_entry *p = ptr;
2654 struct inode *inode;
2655 struct proc_inode *ei;
2656 struct dentry *error;
2658 /* Allocate the inode */
2659 error = ERR_PTR(-ENOMEM);
2660 inode = new_inode(dir->i_sb);
2664 /* Initialize the inode */
2666 inode->i_ino = get_next_ino();
2667 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2670 * grab the reference to the task.
2672 ei->pid = get_task_pid(task, PIDTYPE_PID);
2676 inode->i_mode = p->mode;
2677 if (S_ISDIR(inode->i_mode))
2679 if (S_ISLNK(inode->i_mode))
2682 inode->i_op = p->iop;
2684 inode->i_fop = p->fop;
2686 d_add(dentry, inode);
2695 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2697 struct dentry *error;
2698 struct task_struct *task = get_proc_task(dir);
2699 const struct pid_entry *p, *last;
2701 error = ERR_PTR(-ENOENT);
2706 /* Lookup the directory entry */
2707 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2708 for (p = proc_base_stuff; p <= last; p++) {
2709 if (p->len != dentry->d_name.len)
2711 if (!memcmp(dentry->d_name.name, p->name, p->len))
2717 error = proc_base_instantiate(dir, dentry, task, p);
2720 put_task_struct(task);
2725 static int proc_base_fill_cache(struct file *filp, void *dirent,
2726 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2728 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2729 proc_base_instantiate, task, p);
2732 #ifdef CONFIG_TASK_IO_ACCOUNTING
2733 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2735 struct task_io_accounting acct = task->ioac;
2736 unsigned long flags;
2738 if (whole && lock_task_sighand(task, &flags)) {
2739 struct task_struct *t = task;
2741 task_io_accounting_add(&acct, &task->signal->ioac);
2742 while_each_thread(task, t)
2743 task_io_accounting_add(&acct, &t->ioac);
2745 unlock_task_sighand(task, &flags);
2747 return sprintf(buffer,
2752 "read_bytes: %llu\n"
2753 "write_bytes: %llu\n"
2754 "cancelled_write_bytes: %llu\n",
2755 (unsigned long long)acct.rchar,
2756 (unsigned long long)acct.wchar,
2757 (unsigned long long)acct.syscr,
2758 (unsigned long long)acct.syscw,
2759 (unsigned long long)acct.read_bytes,
2760 (unsigned long long)acct.write_bytes,
2761 (unsigned long long)acct.cancelled_write_bytes);
2764 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2766 return do_io_accounting(task, buffer, 0);
2769 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2771 return do_io_accounting(task, buffer, 1);
2773 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2775 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2776 struct pid *pid, struct task_struct *task)
2778 int err = lock_trace(task);
2780 seq_printf(m, "%08x\n", task->personality);
2789 static const struct file_operations proc_task_operations;
2790 static const struct inode_operations proc_task_inode_operations;
2792 static const struct pid_entry tgid_base_stuff[] = {
2793 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2794 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2795 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2797 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2799 REG("environ", S_IRUSR, proc_environ_operations),
2800 INF("auxv", S_IRUSR, proc_pid_auxv),
2801 ONE("status", S_IRUGO, proc_pid_status),
2802 ONE("personality", S_IRUGO, proc_pid_personality),
2803 INF("limits", S_IRUGO, proc_pid_limits),
2804 #ifdef CONFIG_SCHED_DEBUG
2805 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2807 #ifdef CONFIG_SCHED_AUTOGROUP
2808 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2810 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2811 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2812 INF("syscall", S_IRUGO, proc_pid_syscall),
2814 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2815 ONE("stat", S_IRUGO, proc_tgid_stat),
2816 ONE("statm", S_IRUGO, proc_pid_statm),
2817 REG("maps", S_IRUGO, proc_maps_operations),
2819 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2821 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2822 LNK("cwd", proc_cwd_link),
2823 LNK("root", proc_root_link),
2824 LNK("exe", proc_exe_link),
2825 REG("mounts", S_IRUGO, proc_mounts_operations),
2826 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2827 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2828 #ifdef CONFIG_PROC_PAGE_MONITOR
2829 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2830 REG("smaps", S_IRUGO, proc_smaps_operations),
2831 REG("pagemap", S_IRUGO, proc_pagemap_operations),
2833 #ifdef CONFIG_SECURITY
2834 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2836 #ifdef CONFIG_KALLSYMS
2837 INF("wchan", S_IRUGO, proc_pid_wchan),
2839 #ifdef CONFIG_STACKTRACE
2840 ONE("stack", S_IRUGO, proc_pid_stack),
2842 #ifdef CONFIG_SCHEDSTATS
2843 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2845 #ifdef CONFIG_LATENCYTOP
2846 REG("latency", S_IRUGO, proc_lstats_operations),
2848 #ifdef CONFIG_PROC_PID_CPUSET
2849 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2851 #ifdef CONFIG_CGROUPS
2852 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2854 INF("oom_score", S_IRUGO, proc_oom_score),
2855 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2856 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2857 #ifdef CONFIG_AUDITSYSCALL
2858 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2859 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2861 #ifdef CONFIG_FAULT_INJECTION
2862 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2864 #ifdef CONFIG_ELF_CORE
2865 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2867 #ifdef CONFIG_TASK_IO_ACCOUNTING
2868 INF("io", S_IRUGO, proc_tgid_io_accounting),
2872 static int proc_tgid_base_readdir(struct file * filp,
2873 void * dirent, filldir_t filldir)
2875 return proc_pident_readdir(filp,dirent,filldir,
2876 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2879 static const struct file_operations proc_tgid_base_operations = {
2880 .read = generic_read_dir,
2881 .readdir = proc_tgid_base_readdir,
2882 .llseek = default_llseek,
2885 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2886 return proc_pident_lookup(dir, dentry,
2887 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2890 static const struct inode_operations proc_tgid_base_inode_operations = {
2891 .lookup = proc_tgid_base_lookup,
2892 .getattr = pid_getattr,
2893 .setattr = proc_setattr,
2896 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2898 struct dentry *dentry, *leader, *dir;
2899 char buf[PROC_NUMBUF];
2903 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2904 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2906 shrink_dcache_parent(dentry);
2912 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2913 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2918 name.len = strlen(name.name);
2919 dir = d_hash_and_lookup(leader, &name);
2921 goto out_put_leader;
2924 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2925 dentry = d_hash_and_lookup(dir, &name);
2927 shrink_dcache_parent(dentry);
2940 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2941 * @task: task that should be flushed.
2943 * When flushing dentries from proc, one needs to flush them from global
2944 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2945 * in. This call is supposed to do all of this job.
2947 * Looks in the dcache for
2949 * /proc/@tgid/task/@pid
2950 * if either directory is present flushes it and all of it'ts children
2953 * It is safe and reasonable to cache /proc entries for a task until
2954 * that task exits. After that they just clog up the dcache with
2955 * useless entries, possibly causing useful dcache entries to be
2956 * flushed instead. This routine is proved to flush those useless
2957 * dcache entries at process exit time.
2959 * NOTE: This routine is just an optimization so it does not guarantee
2960 * that no dcache entries will exist at process exit time it
2961 * just makes it very unlikely that any will persist.
2964 void proc_flush_task(struct task_struct *task)
2967 struct pid *pid, *tgid;
2970 pid = task_pid(task);
2971 tgid = task_tgid(task);
2973 for (i = 0; i <= pid->level; i++) {
2974 upid = &pid->numbers[i];
2975 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2976 tgid->numbers[i].nr);
2979 upid = &pid->numbers[pid->level];
2981 pid_ns_release_proc(upid->ns);
2984 static struct dentry *proc_pid_instantiate(struct inode *dir,
2985 struct dentry * dentry,
2986 struct task_struct *task, const void *ptr)
2988 struct dentry *error = ERR_PTR(-ENOENT);
2989 struct inode *inode;
2991 inode = proc_pid_make_inode(dir->i_sb, task);
2995 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2996 inode->i_op = &proc_tgid_base_inode_operations;
2997 inode->i_fop = &proc_tgid_base_operations;
2998 inode->i_flags|=S_IMMUTABLE;
3000 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
3001 ARRAY_SIZE(tgid_base_stuff));
3003 d_set_d_op(dentry, &pid_dentry_operations);
3005 d_add(dentry, inode);
3006 /* Close the race of the process dying before we return the dentry */
3007 if (pid_revalidate(dentry, NULL))
3013 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3015 struct dentry *result;
3016 struct task_struct *task;
3018 struct pid_namespace *ns;
3020 result = proc_base_lookup(dir, dentry);
3021 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3024 tgid = name_to_int(dentry);
3028 ns = dentry->d_sb->s_fs_info;
3030 task = find_task_by_pid_ns(tgid, ns);
3032 get_task_struct(task);
3037 result = proc_pid_instantiate(dir, dentry, task, NULL);
3038 put_task_struct(task);
3044 * Find the first task with tgid >= tgid
3049 struct task_struct *task;
3051 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3056 put_task_struct(iter.task);
3060 pid = find_ge_pid(iter.tgid, ns);
3062 iter.tgid = pid_nr_ns(pid, ns);
3063 iter.task = pid_task(pid, PIDTYPE_PID);
3064 /* What we to know is if the pid we have find is the
3065 * pid of a thread_group_leader. Testing for task
3066 * being a thread_group_leader is the obvious thing
3067 * todo but there is a window when it fails, due to
3068 * the pid transfer logic in de_thread.
3070 * So we perform the straight forward test of seeing
3071 * if the pid we have found is the pid of a thread
3072 * group leader, and don't worry if the task we have
3073 * found doesn't happen to be a thread group leader.
3074 * As we don't care in the case of readdir.
3076 if (!iter.task || !has_group_leader_pid(iter.task)) {
3080 get_task_struct(iter.task);
3086 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
3088 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3089 struct tgid_iter iter)
3091 char name[PROC_NUMBUF];
3092 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
3093 return proc_fill_cache(filp, dirent, filldir, name, len,
3094 proc_pid_instantiate, iter.task, NULL);
3097 /* for the /proc/ directory itself, after non-process stuff has been done */
3098 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3101 struct task_struct *reaper;
3102 struct tgid_iter iter;
3103 struct pid_namespace *ns;
3105 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3107 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3109 reaper = get_proc_task(filp->f_path.dentry->d_inode);
3113 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
3114 const struct pid_entry *p = &proc_base_stuff[nr];
3115 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
3119 ns = filp->f_dentry->d_sb->s_fs_info;
3121 iter.tgid = filp->f_pos - TGID_OFFSET;
3122 for (iter = next_tgid(ns, iter);
3124 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3125 filp->f_pos = iter.tgid + TGID_OFFSET;
3126 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
3127 put_task_struct(iter.task);
3131 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3133 put_task_struct(reaper);
3141 static const struct pid_entry tid_base_stuff[] = {
3142 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3143 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3144 REG("environ", S_IRUSR, proc_environ_operations),
3145 INF("auxv", S_IRUSR, proc_pid_auxv),
3146 ONE("status", S_IRUGO, proc_pid_status),
3147 ONE("personality", S_IRUGO, proc_pid_personality),
3148 INF("limits", S_IRUGO, proc_pid_limits),
3149 #ifdef CONFIG_SCHED_DEBUG
3150 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3152 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3153 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3154 INF("syscall", S_IRUGO, proc_pid_syscall),
3156 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3157 ONE("stat", S_IRUGO, proc_tid_stat),
3158 ONE("statm", S_IRUGO, proc_pid_statm),
3159 REG("maps", S_IRUGO, proc_maps_operations),
3161 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
3163 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3164 LNK("cwd", proc_cwd_link),
3165 LNK("root", proc_root_link),
3166 LNK("exe", proc_exe_link),
3167 REG("mounts", S_IRUGO, proc_mounts_operations),
3168 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3169 #ifdef CONFIG_PROC_PAGE_MONITOR
3170 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3171 REG("smaps", S_IRUGO, proc_smaps_operations),
3172 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3174 #ifdef CONFIG_SECURITY
3175 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3177 #ifdef CONFIG_KALLSYMS
3178 INF("wchan", S_IRUGO, proc_pid_wchan),
3180 #ifdef CONFIG_STACKTRACE
3181 ONE("stack", S_IRUGO, proc_pid_stack),
3183 #ifdef CONFIG_SCHEDSTATS
3184 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3186 #ifdef CONFIG_LATENCYTOP
3187 REG("latency", S_IRUGO, proc_lstats_operations),
3189 #ifdef CONFIG_PROC_PID_CPUSET
3190 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3192 #ifdef CONFIG_CGROUPS
3193 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3195 INF("oom_score", S_IRUGO, proc_oom_score),
3196 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3197 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3198 #ifdef CONFIG_AUDITSYSCALL
3199 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3200 REG("sessionid", S_IRUSR, proc_sessionid_operations),
3202 #ifdef CONFIG_FAULT_INJECTION
3203 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3205 #ifdef CONFIG_TASK_IO_ACCOUNTING
3206 INF("io", S_IRUGO, proc_tid_io_accounting),
3210 static int proc_tid_base_readdir(struct file * filp,
3211 void * dirent, filldir_t filldir)
3213 return proc_pident_readdir(filp,dirent,filldir,
3214 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3217 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3218 return proc_pident_lookup(dir, dentry,
3219 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3222 static const struct file_operations proc_tid_base_operations = {
3223 .read = generic_read_dir,
3224 .readdir = proc_tid_base_readdir,
3225 .llseek = default_llseek,
3228 static const struct inode_operations proc_tid_base_inode_operations = {
3229 .lookup = proc_tid_base_lookup,
3230 .getattr = pid_getattr,
3231 .setattr = proc_setattr,
3234 static struct dentry *proc_task_instantiate(struct inode *dir,
3235 struct dentry *dentry, struct task_struct *task, const void *ptr)
3237 struct dentry *error = ERR_PTR(-ENOENT);
3238 struct inode *inode;
3239 inode = proc_pid_make_inode(dir->i_sb, task);
3243 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3244 inode->i_op = &proc_tid_base_inode_operations;
3245 inode->i_fop = &proc_tid_base_operations;
3246 inode->i_flags|=S_IMMUTABLE;
3248 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
3249 ARRAY_SIZE(tid_base_stuff));
3251 d_set_d_op(dentry, &pid_dentry_operations);
3253 d_add(dentry, inode);
3254 /* Close the race of the process dying before we return the dentry */
3255 if (pid_revalidate(dentry, NULL))
3261 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3263 struct dentry *result = ERR_PTR(-ENOENT);
3264 struct task_struct *task;
3265 struct task_struct *leader = get_proc_task(dir);
3267 struct pid_namespace *ns;
3272 tid = name_to_int(dentry);
3276 ns = dentry->d_sb->s_fs_info;
3278 task = find_task_by_pid_ns(tid, ns);
3280 get_task_struct(task);
3284 if (!same_thread_group(leader, task))
3287 result = proc_task_instantiate(dir, dentry, task, NULL);
3289 put_task_struct(task);
3291 put_task_struct(leader);
3297 * Find the first tid of a thread group to return to user space.
3299 * Usually this is just the thread group leader, but if the users
3300 * buffer was too small or there was a seek into the middle of the
3301 * directory we have more work todo.
3303 * In the case of a short read we start with find_task_by_pid.
3305 * In the case of a seek we start with the leader and walk nr
3308 static struct task_struct *first_tid(struct task_struct *leader,
3309 int tid, int nr, struct pid_namespace *ns)
3311 struct task_struct *pos;
3314 /* Attempt to start with the pid of a thread */
3315 if (tid && (nr > 0)) {
3316 pos = find_task_by_pid_ns(tid, ns);
3317 if (pos && (pos->group_leader == leader))
3321 /* If nr exceeds the number of threads there is nothing todo */
3323 if (nr && nr >= get_nr_threads(leader))
3326 /* If we haven't found our starting place yet start
3327 * with the leader and walk nr threads forward.
3329 for (pos = leader; nr > 0; --nr) {
3330 pos = next_thread(pos);
3331 if (pos == leader) {
3337 get_task_struct(pos);
3344 * Find the next thread in the thread list.
3345 * Return NULL if there is an error or no next thread.
3347 * The reference to the input task_struct is released.
3349 static struct task_struct *next_tid(struct task_struct *start)
3351 struct task_struct *pos = NULL;
3353 if (pid_alive(start)) {
3354 pos = next_thread(start);
3355 if (thread_group_leader(pos))
3358 get_task_struct(pos);
3361 put_task_struct(start);
3365 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3366 struct task_struct *task, int tid)
3368 char name[PROC_NUMBUF];
3369 int len = snprintf(name, sizeof(name), "%d", tid);
3370 return proc_fill_cache(filp, dirent, filldir, name, len,
3371 proc_task_instantiate, task, NULL);
3374 /* for the /proc/TGID/task/ directories */
3375 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3377 struct dentry *dentry = filp->f_path.dentry;
3378 struct inode *inode = dentry->d_inode;
3379 struct task_struct *leader = NULL;
3380 struct task_struct *task;
3381 int retval = -ENOENT;
3384 struct pid_namespace *ns;
3386 task = get_proc_task(inode);
3390 if (pid_alive(task)) {
3391 leader = task->group_leader;
3392 get_task_struct(leader);
3395 put_task_struct(task);
3400 switch ((unsigned long)filp->f_pos) {
3403 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3408 ino = parent_ino(dentry);
3409 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3415 /* f_version caches the tgid value that the last readdir call couldn't
3416 * return. lseek aka telldir automagically resets f_version to 0.
3418 ns = filp->f_dentry->d_sb->s_fs_info;
3419 tid = (int)filp->f_version;
3420 filp->f_version = 0;
3421 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3423 task = next_tid(task), filp->f_pos++) {
3424 tid = task_pid_nr_ns(task, ns);
3425 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3426 /* returning this tgid failed, save it as the first
3427 * pid for the next readir call */
3428 filp->f_version = (u64)tid;
3429 put_task_struct(task);
3434 put_task_struct(leader);
3439 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3441 struct inode *inode = dentry->d_inode;
3442 struct task_struct *p = get_proc_task(inode);
3443 generic_fillattr(inode, stat);
3446 stat->nlink += get_nr_threads(p);
3453 static const struct inode_operations proc_task_inode_operations = {
3454 .lookup = proc_task_lookup,
3455 .getattr = proc_task_getattr,
3456 .setattr = proc_setattr,
3459 static const struct file_operations proc_task_operations = {
3460 .read = generic_read_dir,
3461 .readdir = proc_task_readdir,
3462 .llseek = default_llseek,