4 * Copyright (C) 1991, 1992 Linus Torvalds
6 * proc base directory handling functions
8 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
9 * Instead of using magical inumbers to determine the kind of object
10 * we allocate and fill in-core inodes upon lookup. They don't even
11 * go into icache. We cache the reference to task_struct upon lookup too.
12 * Eventually it should become a filesystem in its own. We don't use the
13 * rest of procfs anymore.
19 * Bruna Moreira <bruna.moreira@indt.org.br>
20 * Edjard Mota <edjard.mota@indt.org.br>
21 * Ilias Biris <ilias.biris@indt.org.br>
22 * Mauricio Lin <mauricio.lin@indt.org.br>
24 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
26 * A new process specific entry (smaps) included in /proc. It shows the
27 * size of rss for each memory area. The maps entry lacks information
28 * about physical memory size (rss) for each mapped file, i.e.,
29 * rss information for executables and library files.
30 * This additional information is useful for any tools that need to know
31 * about physical memory consumption for a process specific library.
35 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36 * Pud inclusion in the page table walking.
40 * 10LE Instituto Nokia de Tecnologia - INdT:
41 * A better way to walks through the page table as suggested by Hugh Dickins.
43 * Simo Piiroinen <simo.piiroinen@nokia.com>:
44 * Smaps information related to shared, private, clean and dirty pages.
46 * Paul Mundt <paul.mundt@nokia.com>:
47 * Overall revision about smaps.
50 #include <asm/uaccess.h>
52 #include <linux/errno.h>
53 #include <linux/time.h>
54 #include <linux/proc_fs.h>
55 #include <linux/stat.h>
56 #include <linux/task_io_accounting_ops.h>
57 #include <linux/init.h>
58 #include <linux/capability.h>
59 #include <linux/file.h>
60 #include <linux/fdtable.h>
61 #include <linux/string.h>
62 #include <linux/seq_file.h>
63 #include <linux/namei.h>
64 #include <linux/mnt_namespace.h>
66 #include <linux/swap.h>
67 #include <linux/rcupdate.h>
68 #include <linux/kallsyms.h>
69 #include <linux/stacktrace.h>
70 #include <linux/resource.h>
71 #include <linux/module.h>
72 #include <linux/mount.h>
73 #include <linux/security.h>
74 #include <linux/ptrace.h>
75 #include <linux/tracehook.h>
76 #include <linux/cgroup.h>
77 #include <linux/cpuset.h>
78 #include <linux/audit.h>
79 #include <linux/poll.h>
80 #include <linux/nsproxy.h>
81 #include <linux/oom.h>
82 #include <linux/elf.h>
83 #include <linux/pid_namespace.h>
84 #include <linux/fs_struct.h>
85 #include <linux/slab.h>
86 #ifdef CONFIG_HARDWALL
87 #include <asm/hardwall.h>
92 * Implementing inode permission operations in /proc is almost
93 * certainly an error. Permission checks need to happen during
94 * each system call not at open time. The reason is that most of
95 * what we wish to check for permissions in /proc varies at runtime.
97 * The classic example of a problem is opening file descriptors
98 * in /proc for a task before it execs a suid executable.
105 const struct inode_operations *iop;
106 const struct file_operations *fop;
110 #define NOD(NAME, MODE, IOP, FOP, OP) { \
112 .len = sizeof(NAME) - 1, \
119 #define DIR(NAME, MODE, iops, fops) \
120 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
121 #define LNK(NAME, get_link) \
122 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
123 &proc_pid_link_inode_operations, NULL, \
124 { .proc_get_link = get_link } )
125 #define REG(NAME, MODE, fops) \
126 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
127 #define INF(NAME, MODE, read) \
128 NOD(NAME, (S_IFREG|(MODE)), \
129 NULL, &proc_info_file_operations, \
130 { .proc_read = read } )
131 #define ONE(NAME, MODE, show) \
132 NOD(NAME, (S_IFREG|(MODE)), \
133 NULL, &proc_single_file_operations, \
134 { .proc_show = show } )
137 * Count the number of hardlinks for the pid_entry table, excluding the .
140 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
147 for (i = 0; i < n; ++i) {
148 if (S_ISDIR(entries[i].mode))
155 static int get_task_root(struct task_struct *task, struct path *root)
157 int result = -ENOENT;
161 get_fs_root(task->fs, root);
168 static int proc_cwd_link(struct inode *inode, struct path *path)
170 struct task_struct *task = get_proc_task(inode);
171 int result = -ENOENT;
176 get_fs_pwd(task->fs, path);
180 put_task_struct(task);
185 static int proc_root_link(struct inode *inode, struct path *path)
187 struct task_struct *task = get_proc_task(inode);
188 int result = -ENOENT;
191 result = get_task_root(task, path);
192 put_task_struct(task);
197 static struct mm_struct *__check_mem_permission(struct task_struct *task)
199 struct mm_struct *mm;
201 mm = get_task_mm(task);
203 return ERR_PTR(-EINVAL);
206 * A task can always look at itself, in case it chooses
207 * to use system calls instead of load instructions.
213 * If current is actively ptrace'ing, and would also be
214 * permitted to freshly attach with ptrace now, permit it.
216 if (task_is_stopped_or_traced(task)) {
219 match = (ptrace_parent(task) == current);
221 if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH))
226 * No one else is allowed.
229 return ERR_PTR(-EPERM);
233 * If current may access user memory in @task return a reference to the
234 * corresponding mm, otherwise ERR_PTR.
236 static struct mm_struct *check_mem_permission(struct task_struct *task)
238 struct mm_struct *mm;
242 * Avoid racing if task exec's as we might get a new mm but validate
243 * against old credentials.
245 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
249 mm = __check_mem_permission(task);
250 mutex_unlock(&task->signal->cred_guard_mutex);
255 struct mm_struct *mm_for_maps(struct task_struct *task)
257 struct mm_struct *mm;
260 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
264 mm = get_task_mm(task);
265 if (mm && mm != current->mm &&
266 !ptrace_may_access(task, PTRACE_MODE_READ)) {
268 mm = ERR_PTR(-EACCES);
270 mutex_unlock(&task->signal->cred_guard_mutex);
275 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
279 struct mm_struct *mm = get_task_mm(task);
283 goto out_mm; /* Shh! No looking before we're done */
285 len = mm->arg_end - mm->arg_start;
290 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
292 // If the nul at the end of args has been overwritten, then
293 // assume application is using setproctitle(3).
294 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
295 len = strnlen(buffer, res);
299 len = mm->env_end - mm->env_start;
300 if (len > PAGE_SIZE - res)
301 len = PAGE_SIZE - res;
302 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
303 res = strnlen(buffer, res);
312 static int proc_pid_auxv(struct task_struct *task, char *buffer)
314 struct mm_struct *mm = mm_for_maps(task);
315 int res = PTR_ERR(mm);
316 if (mm && !IS_ERR(mm)) {
317 unsigned int nwords = 0;
320 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
321 res = nwords * sizeof(mm->saved_auxv[0]);
324 memcpy(buffer, mm->saved_auxv, res);
331 #ifdef CONFIG_KALLSYMS
333 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
334 * Returns the resolved symbol. If that fails, simply return the address.
336 static int proc_pid_wchan(struct task_struct *task, char *buffer)
339 char symname[KSYM_NAME_LEN];
341 wchan = get_wchan(task);
343 if (lookup_symbol_name(wchan, symname) < 0)
344 if (!ptrace_may_access(task, PTRACE_MODE_READ))
347 return sprintf(buffer, "%lu", wchan);
349 return sprintf(buffer, "%s", symname);
351 #endif /* CONFIG_KALLSYMS */
353 static int lock_trace(struct task_struct *task)
355 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
358 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
359 mutex_unlock(&task->signal->cred_guard_mutex);
365 static void unlock_trace(struct task_struct *task)
367 mutex_unlock(&task->signal->cred_guard_mutex);
370 #ifdef CONFIG_STACKTRACE
372 #define MAX_STACK_TRACE_DEPTH 64
374 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
375 struct pid *pid, struct task_struct *task)
377 struct stack_trace trace;
378 unsigned long *entries;
382 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
386 trace.nr_entries = 0;
387 trace.max_entries = MAX_STACK_TRACE_DEPTH;
388 trace.entries = entries;
391 err = lock_trace(task);
393 save_stack_trace_tsk(task, &trace);
395 for (i = 0; i < trace.nr_entries; i++) {
396 seq_printf(m, "[<%pK>] %pS\n",
397 (void *)entries[i], (void *)entries[i]);
407 #ifdef CONFIG_SCHEDSTATS
409 * Provides /proc/PID/schedstat
411 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
413 return sprintf(buffer, "%llu %llu %lu\n",
414 (unsigned long long)task->se.sum_exec_runtime,
415 (unsigned long long)task->sched_info.run_delay,
416 task->sched_info.pcount);
420 #ifdef CONFIG_LATENCYTOP
421 static int lstats_show_proc(struct seq_file *m, void *v)
424 struct inode *inode = m->private;
425 struct task_struct *task = get_proc_task(inode);
429 seq_puts(m, "Latency Top version : v0.1\n");
430 for (i = 0; i < 32; i++) {
431 struct latency_record *lr = &task->latency_record[i];
432 if (lr->backtrace[0]) {
434 seq_printf(m, "%i %li %li",
435 lr->count, lr->time, lr->max);
436 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
437 unsigned long bt = lr->backtrace[q];
442 seq_printf(m, " %ps", (void *)bt);
448 put_task_struct(task);
452 static int lstats_open(struct inode *inode, struct file *file)
454 return single_open(file, lstats_show_proc, inode);
457 static ssize_t lstats_write(struct file *file, const char __user *buf,
458 size_t count, loff_t *offs)
460 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
464 clear_all_latency_tracing(task);
465 put_task_struct(task);
470 static const struct file_operations proc_lstats_operations = {
473 .write = lstats_write,
475 .release = single_release,
480 static int proc_oom_score(struct task_struct *task, char *buffer)
482 unsigned long points = 0;
484 read_lock(&tasklist_lock);
486 points = oom_badness(task, NULL, NULL,
487 totalram_pages + total_swap_pages);
488 read_unlock(&tasklist_lock);
489 return sprintf(buffer, "%lu\n", points);
497 static const struct limit_names lnames[RLIM_NLIMITS] = {
498 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
499 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
500 [RLIMIT_DATA] = {"Max data size", "bytes"},
501 [RLIMIT_STACK] = {"Max stack size", "bytes"},
502 [RLIMIT_CORE] = {"Max core file size", "bytes"},
503 [RLIMIT_RSS] = {"Max resident set", "bytes"},
504 [RLIMIT_NPROC] = {"Max processes", "processes"},
505 [RLIMIT_NOFILE] = {"Max open files", "files"},
506 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
507 [RLIMIT_AS] = {"Max address space", "bytes"},
508 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
509 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
510 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
511 [RLIMIT_NICE] = {"Max nice priority", NULL},
512 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
513 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
516 /* Display limits for a process */
517 static int proc_pid_limits(struct task_struct *task, char *buffer)
522 char *bufptr = buffer;
524 struct rlimit rlim[RLIM_NLIMITS];
526 if (!lock_task_sighand(task, &flags))
528 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
529 unlock_task_sighand(task, &flags);
532 * print the file header
534 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
535 "Limit", "Soft Limit", "Hard Limit", "Units");
537 for (i = 0; i < RLIM_NLIMITS; i++) {
538 if (rlim[i].rlim_cur == RLIM_INFINITY)
539 count += sprintf(&bufptr[count], "%-25s %-20s ",
540 lnames[i].name, "unlimited");
542 count += sprintf(&bufptr[count], "%-25s %-20lu ",
543 lnames[i].name, rlim[i].rlim_cur);
545 if (rlim[i].rlim_max == RLIM_INFINITY)
546 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
548 count += sprintf(&bufptr[count], "%-20lu ",
552 count += sprintf(&bufptr[count], "%-10s\n",
555 count += sprintf(&bufptr[count], "\n");
561 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
562 static int proc_pid_syscall(struct task_struct *task, char *buffer)
565 unsigned long args[6], sp, pc;
566 int res = lock_trace(task);
570 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
571 res = sprintf(buffer, "running\n");
573 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
575 res = sprintf(buffer,
576 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
578 args[0], args[1], args[2], args[3], args[4], args[5],
583 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
585 /************************************************************************/
586 /* Here the fs part begins */
587 /************************************************************************/
589 /* permission checks */
590 static int proc_fd_access_allowed(struct inode *inode)
592 struct task_struct *task;
594 /* Allow access to a task's file descriptors if it is us or we
595 * may use ptrace attach to the process and find out that
598 task = get_proc_task(inode);
600 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
601 put_task_struct(task);
606 int proc_setattr(struct dentry *dentry, struct iattr *attr)
609 struct inode *inode = dentry->d_inode;
611 if (attr->ia_valid & ATTR_MODE)
614 error = inode_change_ok(inode, attr);
618 if ((attr->ia_valid & ATTR_SIZE) &&
619 attr->ia_size != i_size_read(inode)) {
620 error = vmtruncate(inode, attr->ia_size);
625 setattr_copy(inode, attr);
626 mark_inode_dirty(inode);
630 static const struct inode_operations proc_def_inode_operations = {
631 .setattr = proc_setattr,
634 static int mounts_open_common(struct inode *inode, struct file *file,
635 const struct seq_operations *op)
637 struct task_struct *task = get_proc_task(inode);
639 struct mnt_namespace *ns = NULL;
641 struct proc_mounts *p;
646 nsp = task_nsproxy(task);
653 if (ns && get_task_root(task, &root) == 0)
655 put_task_struct(task);
664 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
668 file->private_data = &p->m;
669 ret = seq_open(file, op);
676 p->m.poll_event = ns->event;
690 static int mounts_release(struct inode *inode, struct file *file)
692 struct proc_mounts *p = file->private_data;
695 return seq_release(inode, file);
698 static unsigned mounts_poll(struct file *file, poll_table *wait)
700 struct proc_mounts *p = file->private_data;
701 unsigned res = POLLIN | POLLRDNORM;
703 poll_wait(file, &p->ns->poll, wait);
704 if (mnt_had_events(p))
705 res |= POLLERR | POLLPRI;
710 static int mounts_open(struct inode *inode, struct file *file)
712 return mounts_open_common(inode, file, &mounts_op);
715 static const struct file_operations proc_mounts_operations = {
719 .release = mounts_release,
723 static int mountinfo_open(struct inode *inode, struct file *file)
725 return mounts_open_common(inode, file, &mountinfo_op);
728 static const struct file_operations proc_mountinfo_operations = {
729 .open = mountinfo_open,
732 .release = mounts_release,
736 static int mountstats_open(struct inode *inode, struct file *file)
738 return mounts_open_common(inode, file, &mountstats_op);
741 static const struct file_operations proc_mountstats_operations = {
742 .open = mountstats_open,
745 .release = mounts_release,
748 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
750 static ssize_t proc_info_read(struct file * file, char __user * buf,
751 size_t count, loff_t *ppos)
753 struct inode * inode = file->f_path.dentry->d_inode;
756 struct task_struct *task = get_proc_task(inode);
762 if (count > PROC_BLOCK_SIZE)
763 count = PROC_BLOCK_SIZE;
766 if (!(page = __get_free_page(GFP_TEMPORARY)))
769 length = PROC_I(inode)->op.proc_read(task, (char*)page);
772 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
775 put_task_struct(task);
780 static const struct file_operations proc_info_file_operations = {
781 .read = proc_info_read,
782 .llseek = generic_file_llseek,
785 static int proc_single_show(struct seq_file *m, void *v)
787 struct inode *inode = m->private;
788 struct pid_namespace *ns;
790 struct task_struct *task;
793 ns = inode->i_sb->s_fs_info;
794 pid = proc_pid(inode);
795 task = get_pid_task(pid, PIDTYPE_PID);
799 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
801 put_task_struct(task);
805 static int proc_single_open(struct inode *inode, struct file *filp)
807 return single_open(filp, proc_single_show, inode);
810 static const struct file_operations proc_single_file_operations = {
811 .open = proc_single_open,
814 .release = single_release,
817 static int mem_open(struct inode* inode, struct file* file)
819 file->private_data = (void*)((long)current->self_exec_id);
820 /* OK to pass negative loff_t, we can catch out-of-range */
821 file->f_mode |= FMODE_UNSIGNED_OFFSET;
825 static ssize_t mem_read(struct file * file, char __user * buf,
826 size_t count, loff_t *ppos)
828 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
830 unsigned long src = *ppos;
832 struct mm_struct *mm;
838 page = (char *)__get_free_page(GFP_TEMPORARY);
842 mm = check_mem_permission(task);
849 if (file->private_data != (void*)((long)current->self_exec_id))
855 int this_len, retval;
857 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
858 retval = access_remote_vm(mm, src, page, this_len, 0);
865 if (copy_to_user(buf, page, retval)) {
880 free_page((unsigned long) page);
882 put_task_struct(task);
887 static ssize_t mem_write(struct file * file, const char __user *buf,
888 size_t count, loff_t *ppos)
892 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
893 unsigned long dst = *ppos;
894 struct mm_struct *mm;
901 page = (char *)__get_free_page(GFP_TEMPORARY);
905 mm = check_mem_permission(task);
906 copied = PTR_ERR(mm);
911 if (file->private_data != (void *)((long)current->self_exec_id))
916 int this_len, retval;
918 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
919 if (copy_from_user(page, buf, this_len)) {
923 retval = access_remote_vm(mm, dst, page, this_len, 1);
939 free_page((unsigned long) page);
941 put_task_struct(task);
946 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
950 file->f_pos = offset;
953 file->f_pos += offset;
958 force_successful_syscall_return();
962 static const struct file_operations proc_mem_operations = {
969 static ssize_t environ_read(struct file *file, char __user *buf,
970 size_t count, loff_t *ppos)
972 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
974 unsigned long src = *ppos;
976 struct mm_struct *mm;
982 page = (char *)__get_free_page(GFP_TEMPORARY);
987 mm = mm_for_maps(task);
989 if (!mm || IS_ERR(mm))
994 int this_len, retval, max_len;
996 this_len = mm->env_end - (mm->env_start + src);
1001 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
1002 this_len = (this_len > max_len) ? max_len : this_len;
1004 retval = access_process_vm(task, (mm->env_start + src),
1012 if (copy_to_user(buf, page, retval)) {
1026 free_page((unsigned long) page);
1028 put_task_struct(task);
1033 static const struct file_operations proc_environ_operations = {
1034 .read = environ_read,
1035 .llseek = generic_file_llseek,
1038 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
1039 size_t count, loff_t *ppos)
1041 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1042 char buffer[PROC_NUMBUF];
1044 int oom_adjust = OOM_DISABLE;
1045 unsigned long flags;
1050 if (lock_task_sighand(task, &flags)) {
1051 oom_adjust = task->signal->oom_adj;
1052 unlock_task_sighand(task, &flags);
1055 put_task_struct(task);
1057 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
1059 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1062 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1063 size_t count, loff_t *ppos)
1065 struct task_struct *task;
1066 char buffer[PROC_NUMBUF];
1068 unsigned long flags;
1071 memset(buffer, 0, sizeof(buffer));
1072 if (count > sizeof(buffer) - 1)
1073 count = sizeof(buffer) - 1;
1074 if (copy_from_user(buffer, buf, count)) {
1079 err = kstrtoint(strstrip(buffer), 0, &oom_adjust);
1082 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
1083 oom_adjust != OOM_DISABLE) {
1088 task = get_proc_task(file->f_path.dentry->d_inode);
1100 if (!lock_task_sighand(task, &flags)) {
1105 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
1110 if (oom_adjust != task->signal->oom_adj) {
1111 if (oom_adjust == OOM_DISABLE)
1112 atomic_inc(&task->mm->oom_disable_count);
1113 if (task->signal->oom_adj == OOM_DISABLE)
1114 atomic_dec(&task->mm->oom_disable_count);
1118 * Warn that /proc/pid/oom_adj is deprecated, see
1119 * Documentation/feature-removal-schedule.txt.
1121 WARN_ONCE(1, "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1122 current->comm, task_pid_nr(current), task_pid_nr(task),
1124 task->signal->oom_adj = oom_adjust;
1126 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1127 * value is always attainable.
1129 if (task->signal->oom_adj == OOM_ADJUST_MAX)
1130 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
1132 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
1135 unlock_task_sighand(task, &flags);
1138 put_task_struct(task);
1140 return err < 0 ? err : count;
1143 static const struct file_operations proc_oom_adjust_operations = {
1144 .read = oom_adjust_read,
1145 .write = oom_adjust_write,
1146 .llseek = generic_file_llseek,
1149 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1150 size_t count, loff_t *ppos)
1152 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1153 char buffer[PROC_NUMBUF];
1154 int oom_score_adj = OOM_SCORE_ADJ_MIN;
1155 unsigned long flags;
1160 if (lock_task_sighand(task, &flags)) {
1161 oom_score_adj = task->signal->oom_score_adj;
1162 unlock_task_sighand(task, &flags);
1164 put_task_struct(task);
1165 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1166 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1169 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1170 size_t count, loff_t *ppos)
1172 struct task_struct *task;
1173 char buffer[PROC_NUMBUF];
1174 unsigned long flags;
1178 memset(buffer, 0, sizeof(buffer));
1179 if (count > sizeof(buffer) - 1)
1180 count = sizeof(buffer) - 1;
1181 if (copy_from_user(buffer, buf, count)) {
1186 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1189 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1190 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1195 task = get_proc_task(file->f_path.dentry->d_inode);
1207 if (!lock_task_sighand(task, &flags)) {
1212 if (oom_score_adj < task->signal->oom_score_adj_min &&
1213 !capable(CAP_SYS_RESOURCE)) {
1218 if (oom_score_adj != task->signal->oom_score_adj) {
1219 if (oom_score_adj == OOM_SCORE_ADJ_MIN)
1220 atomic_inc(&task->mm->oom_disable_count);
1221 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1222 atomic_dec(&task->mm->oom_disable_count);
1224 task->signal->oom_score_adj = oom_score_adj;
1225 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1226 task->signal->oom_score_adj_min = oom_score_adj;
1228 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1229 * always attainable.
1231 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1232 task->signal->oom_adj = OOM_DISABLE;
1234 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1237 unlock_task_sighand(task, &flags);
1240 put_task_struct(task);
1242 return err < 0 ? err : count;
1245 static const struct file_operations proc_oom_score_adj_operations = {
1246 .read = oom_score_adj_read,
1247 .write = oom_score_adj_write,
1248 .llseek = default_llseek,
1251 #ifdef CONFIG_AUDITSYSCALL
1252 #define TMPBUFLEN 21
1253 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1254 size_t count, loff_t *ppos)
1256 struct inode * inode = file->f_path.dentry->d_inode;
1257 struct task_struct *task = get_proc_task(inode);
1259 char tmpbuf[TMPBUFLEN];
1263 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1264 audit_get_loginuid(task));
1265 put_task_struct(task);
1266 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1269 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1270 size_t count, loff_t *ppos)
1272 struct inode * inode = file->f_path.dentry->d_inode;
1277 if (!capable(CAP_AUDIT_CONTROL))
1281 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1287 if (count >= PAGE_SIZE)
1288 count = PAGE_SIZE - 1;
1291 /* No partial writes. */
1294 page = (char*)__get_free_page(GFP_TEMPORARY);
1298 if (copy_from_user(page, buf, count))
1302 loginuid = simple_strtoul(page, &tmp, 10);
1308 length = audit_set_loginuid(current, loginuid);
1309 if (likely(length == 0))
1313 free_page((unsigned long) page);
1317 static const struct file_operations proc_loginuid_operations = {
1318 .read = proc_loginuid_read,
1319 .write = proc_loginuid_write,
1320 .llseek = generic_file_llseek,
1323 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1324 size_t count, loff_t *ppos)
1326 struct inode * inode = file->f_path.dentry->d_inode;
1327 struct task_struct *task = get_proc_task(inode);
1329 char tmpbuf[TMPBUFLEN];
1333 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1334 audit_get_sessionid(task));
1335 put_task_struct(task);
1336 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1339 static const struct file_operations proc_sessionid_operations = {
1340 .read = proc_sessionid_read,
1341 .llseek = generic_file_llseek,
1345 #ifdef CONFIG_FAULT_INJECTION
1346 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1347 size_t count, loff_t *ppos)
1349 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1350 char buffer[PROC_NUMBUF];
1356 make_it_fail = task->make_it_fail;
1357 put_task_struct(task);
1359 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1361 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1364 static ssize_t proc_fault_inject_write(struct file * file,
1365 const char __user * buf, size_t count, loff_t *ppos)
1367 struct task_struct *task;
1368 char buffer[PROC_NUMBUF], *end;
1371 if (!capable(CAP_SYS_RESOURCE))
1373 memset(buffer, 0, sizeof(buffer));
1374 if (count > sizeof(buffer) - 1)
1375 count = sizeof(buffer) - 1;
1376 if (copy_from_user(buffer, buf, count))
1378 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1381 task = get_proc_task(file->f_dentry->d_inode);
1384 task->make_it_fail = make_it_fail;
1385 put_task_struct(task);
1390 static const struct file_operations proc_fault_inject_operations = {
1391 .read = proc_fault_inject_read,
1392 .write = proc_fault_inject_write,
1393 .llseek = generic_file_llseek,
1398 #ifdef CONFIG_SCHED_DEBUG
1400 * Print out various scheduling related per-task fields:
1402 static int sched_show(struct seq_file *m, void *v)
1404 struct inode *inode = m->private;
1405 struct task_struct *p;
1407 p = get_proc_task(inode);
1410 proc_sched_show_task(p, m);
1418 sched_write(struct file *file, const char __user *buf,
1419 size_t count, loff_t *offset)
1421 struct inode *inode = file->f_path.dentry->d_inode;
1422 struct task_struct *p;
1424 p = get_proc_task(inode);
1427 proc_sched_set_task(p);
1434 static int sched_open(struct inode *inode, struct file *filp)
1436 return single_open(filp, sched_show, inode);
1439 static const struct file_operations proc_pid_sched_operations = {
1442 .write = sched_write,
1443 .llseek = seq_lseek,
1444 .release = single_release,
1449 #ifdef CONFIG_SCHED_AUTOGROUP
1451 * Print out autogroup related information:
1453 static int sched_autogroup_show(struct seq_file *m, void *v)
1455 struct inode *inode = m->private;
1456 struct task_struct *p;
1458 p = get_proc_task(inode);
1461 proc_sched_autogroup_show_task(p, m);
1469 sched_autogroup_write(struct file *file, const char __user *buf,
1470 size_t count, loff_t *offset)
1472 struct inode *inode = file->f_path.dentry->d_inode;
1473 struct task_struct *p;
1474 char buffer[PROC_NUMBUF];
1478 memset(buffer, 0, sizeof(buffer));
1479 if (count > sizeof(buffer) - 1)
1480 count = sizeof(buffer) - 1;
1481 if (copy_from_user(buffer, buf, count))
1484 err = kstrtoint(strstrip(buffer), 0, &nice);
1488 p = get_proc_task(inode);
1493 err = proc_sched_autogroup_set_nice(p, &err);
1502 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1506 ret = single_open(filp, sched_autogroup_show, NULL);
1508 struct seq_file *m = filp->private_data;
1515 static const struct file_operations proc_pid_sched_autogroup_operations = {
1516 .open = sched_autogroup_open,
1518 .write = sched_autogroup_write,
1519 .llseek = seq_lseek,
1520 .release = single_release,
1523 #endif /* CONFIG_SCHED_AUTOGROUP */
1525 static ssize_t comm_write(struct file *file, const char __user *buf,
1526 size_t count, loff_t *offset)
1528 struct inode *inode = file->f_path.dentry->d_inode;
1529 struct task_struct *p;
1530 char buffer[TASK_COMM_LEN];
1532 memset(buffer, 0, sizeof(buffer));
1533 if (count > sizeof(buffer) - 1)
1534 count = sizeof(buffer) - 1;
1535 if (copy_from_user(buffer, buf, count))
1538 p = get_proc_task(inode);
1542 if (same_thread_group(current, p))
1543 set_task_comm(p, buffer);
1552 static int comm_show(struct seq_file *m, void *v)
1554 struct inode *inode = m->private;
1555 struct task_struct *p;
1557 p = get_proc_task(inode);
1562 seq_printf(m, "%s\n", p->comm);
1570 static int comm_open(struct inode *inode, struct file *filp)
1572 return single_open(filp, comm_show, inode);
1575 static const struct file_operations proc_pid_set_comm_operations = {
1578 .write = comm_write,
1579 .llseek = seq_lseek,
1580 .release = single_release,
1583 static int proc_exe_link(struct inode *inode, struct path *exe_path)
1585 struct task_struct *task;
1586 struct mm_struct *mm;
1587 struct file *exe_file;
1589 task = get_proc_task(inode);
1592 mm = get_task_mm(task);
1593 put_task_struct(task);
1596 exe_file = get_mm_exe_file(mm);
1599 *exe_path = exe_file->f_path;
1600 path_get(&exe_file->f_path);
1607 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1609 struct inode *inode = dentry->d_inode;
1610 int error = -EACCES;
1612 /* We don't need a base pointer in the /proc filesystem */
1613 path_put(&nd->path);
1615 /* Are we allowed to snoop on the tasks file descriptors? */
1616 if (!proc_fd_access_allowed(inode))
1619 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1621 return ERR_PTR(error);
1624 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1626 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1633 pathname = d_path(path, tmp, PAGE_SIZE);
1634 len = PTR_ERR(pathname);
1635 if (IS_ERR(pathname))
1637 len = tmp + PAGE_SIZE - 1 - pathname;
1641 if (copy_to_user(buffer, pathname, len))
1644 free_page((unsigned long)tmp);
1648 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1650 int error = -EACCES;
1651 struct inode *inode = dentry->d_inode;
1654 /* Are we allowed to snoop on the tasks file descriptors? */
1655 if (!proc_fd_access_allowed(inode))
1658 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1662 error = do_proc_readlink(&path, buffer, buflen);
1668 static const struct inode_operations proc_pid_link_inode_operations = {
1669 .readlink = proc_pid_readlink,
1670 .follow_link = proc_pid_follow_link,
1671 .setattr = proc_setattr,
1675 /* building an inode */
1677 static int task_dumpable(struct task_struct *task)
1680 struct mm_struct *mm;
1685 dumpable = get_dumpable(mm);
1692 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1694 struct inode * inode;
1695 struct proc_inode *ei;
1696 const struct cred *cred;
1698 /* We need a new inode */
1700 inode = new_inode(sb);
1706 inode->i_ino = get_next_ino();
1707 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1708 inode->i_op = &proc_def_inode_operations;
1711 * grab the reference to task.
1713 ei->pid = get_task_pid(task, PIDTYPE_PID);
1717 if (task_dumpable(task)) {
1719 cred = __task_cred(task);
1720 inode->i_uid = cred->euid;
1721 inode->i_gid = cred->egid;
1724 security_task_to_inode(task, inode);
1734 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1736 struct inode *inode = dentry->d_inode;
1737 struct task_struct *task;
1738 const struct cred *cred;
1740 generic_fillattr(inode, stat);
1745 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1747 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1748 task_dumpable(task)) {
1749 cred = __task_cred(task);
1750 stat->uid = cred->euid;
1751 stat->gid = cred->egid;
1761 * Exceptional case: normally we are not allowed to unhash a busy
1762 * directory. In this case, however, we can do it - no aliasing problems
1763 * due to the way we treat inodes.
1765 * Rewrite the inode's ownerships here because the owning task may have
1766 * performed a setuid(), etc.
1768 * Before the /proc/pid/status file was created the only way to read
1769 * the effective uid of a /process was to stat /proc/pid. Reading
1770 * /proc/pid/status is slow enough that procps and other packages
1771 * kept stating /proc/pid. To keep the rules in /proc simple I have
1772 * made this apply to all per process world readable and executable
1775 int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1777 struct inode *inode;
1778 struct task_struct *task;
1779 const struct cred *cred;
1781 if (nd && nd->flags & LOOKUP_RCU)
1784 inode = dentry->d_inode;
1785 task = get_proc_task(inode);
1788 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1789 task_dumpable(task)) {
1791 cred = __task_cred(task);
1792 inode->i_uid = cred->euid;
1793 inode->i_gid = cred->egid;
1799 inode->i_mode &= ~(S_ISUID | S_ISGID);
1800 security_task_to_inode(task, inode);
1801 put_task_struct(task);
1808 static int pid_delete_dentry(const struct dentry * dentry)
1810 /* Is the task we represent dead?
1811 * If so, then don't put the dentry on the lru list,
1812 * kill it immediately.
1814 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1817 const struct dentry_operations pid_dentry_operations =
1819 .d_revalidate = pid_revalidate,
1820 .d_delete = pid_delete_dentry,
1826 * Fill a directory entry.
1828 * If possible create the dcache entry and derive our inode number and
1829 * file type from dcache entry.
1831 * Since all of the proc inode numbers are dynamically generated, the inode
1832 * numbers do not exist until the inode is cache. This means creating the
1833 * the dcache entry in readdir is necessary to keep the inode numbers
1834 * reported by readdir in sync with the inode numbers reported
1837 int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1838 const char *name, int len,
1839 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1841 struct dentry *child, *dir = filp->f_path.dentry;
1842 struct inode *inode;
1845 unsigned type = DT_UNKNOWN;
1849 qname.hash = full_name_hash(name, len);
1851 child = d_lookup(dir, &qname);
1854 new = d_alloc(dir, &qname);
1856 child = instantiate(dir->d_inode, new, task, ptr);
1863 if (!child || IS_ERR(child) || !child->d_inode)
1864 goto end_instantiate;
1865 inode = child->d_inode;
1868 type = inode->i_mode >> 12;
1873 ino = find_inode_number(dir, &qname);
1876 return filldir(dirent, name, len, filp->f_pos, ino, type);
1879 static unsigned name_to_int(struct dentry *dentry)
1881 const char *name = dentry->d_name.name;
1882 int len = dentry->d_name.len;
1885 if (len > 1 && *name == '0')
1888 unsigned c = *name++ - '0';
1891 if (n >= (~0U-9)/10)
1901 #define PROC_FDINFO_MAX 64
1903 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1905 struct task_struct *task = get_proc_task(inode);
1906 struct files_struct *files = NULL;
1908 int fd = proc_fd(inode);
1911 files = get_files_struct(task);
1912 put_task_struct(task);
1916 * We are not taking a ref to the file structure, so we must
1919 spin_lock(&files->file_lock);
1920 file = fcheck_files(files, fd);
1923 *path = file->f_path;
1924 path_get(&file->f_path);
1927 snprintf(info, PROC_FDINFO_MAX,
1930 (long long) file->f_pos,
1932 spin_unlock(&files->file_lock);
1933 put_files_struct(files);
1936 spin_unlock(&files->file_lock);
1937 put_files_struct(files);
1942 static int proc_fd_link(struct inode *inode, struct path *path)
1944 return proc_fd_info(inode, path, NULL);
1947 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1949 struct inode *inode;
1950 struct task_struct *task;
1952 struct files_struct *files;
1953 const struct cred *cred;
1955 if (nd && nd->flags & LOOKUP_RCU)
1958 inode = dentry->d_inode;
1959 task = get_proc_task(inode);
1960 fd = proc_fd(inode);
1963 files = get_files_struct(task);
1966 if (fcheck_files(files, fd)) {
1968 put_files_struct(files);
1969 if (task_dumpable(task)) {
1971 cred = __task_cred(task);
1972 inode->i_uid = cred->euid;
1973 inode->i_gid = cred->egid;
1979 inode->i_mode &= ~(S_ISUID | S_ISGID);
1980 security_task_to_inode(task, inode);
1981 put_task_struct(task);
1985 put_files_struct(files);
1987 put_task_struct(task);
1993 static const struct dentry_operations tid_fd_dentry_operations =
1995 .d_revalidate = tid_fd_revalidate,
1996 .d_delete = pid_delete_dentry,
1999 static struct dentry *proc_fd_instantiate(struct inode *dir,
2000 struct dentry *dentry, struct task_struct *task, const void *ptr)
2002 unsigned fd = *(const unsigned *)ptr;
2004 struct files_struct *files;
2005 struct inode *inode;
2006 struct proc_inode *ei;
2007 struct dentry *error = ERR_PTR(-ENOENT);
2009 inode = proc_pid_make_inode(dir->i_sb, task);
2014 files = get_files_struct(task);
2017 inode->i_mode = S_IFLNK;
2020 * We are not taking a ref to the file structure, so we must
2023 spin_lock(&files->file_lock);
2024 file = fcheck_files(files, fd);
2027 if (file->f_mode & FMODE_READ)
2028 inode->i_mode |= S_IRUSR | S_IXUSR;
2029 if (file->f_mode & FMODE_WRITE)
2030 inode->i_mode |= S_IWUSR | S_IXUSR;
2031 spin_unlock(&files->file_lock);
2032 put_files_struct(files);
2034 inode->i_op = &proc_pid_link_inode_operations;
2036 ei->op.proc_get_link = proc_fd_link;
2037 d_set_d_op(dentry, &tid_fd_dentry_operations);
2038 d_add(dentry, inode);
2039 /* Close the race of the process dying before we return the dentry */
2040 if (tid_fd_revalidate(dentry, NULL))
2046 spin_unlock(&files->file_lock);
2047 put_files_struct(files);
2053 static struct dentry *proc_lookupfd_common(struct inode *dir,
2054 struct dentry *dentry,
2055 instantiate_t instantiate)
2057 struct task_struct *task = get_proc_task(dir);
2058 unsigned fd = name_to_int(dentry);
2059 struct dentry *result = ERR_PTR(-ENOENT);
2066 result = instantiate(dir, dentry, task, &fd);
2068 put_task_struct(task);
2073 static int proc_readfd_common(struct file * filp, void * dirent,
2074 filldir_t filldir, instantiate_t instantiate)
2076 struct dentry *dentry = filp->f_path.dentry;
2077 struct inode *inode = dentry->d_inode;
2078 struct task_struct *p = get_proc_task(inode);
2079 unsigned int fd, ino;
2081 struct files_struct * files;
2091 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
2095 ino = parent_ino(dentry);
2096 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2100 files = get_files_struct(p);
2104 for (fd = filp->f_pos-2;
2105 fd < files_fdtable(files)->max_fds;
2106 fd++, filp->f_pos++) {
2107 char name[PROC_NUMBUF];
2110 if (!fcheck_files(files, fd))
2114 len = snprintf(name, sizeof(name), "%d", fd);
2115 if (proc_fill_cache(filp, dirent, filldir,
2116 name, len, instantiate,
2124 put_files_struct(files);
2132 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
2133 struct nameidata *nd)
2135 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
2138 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
2140 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
2143 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
2144 size_t len, loff_t *ppos)
2146 char tmp[PROC_FDINFO_MAX];
2147 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
2149 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
2153 static const struct file_operations proc_fdinfo_file_operations = {
2154 .open = nonseekable_open,
2155 .read = proc_fdinfo_read,
2156 .llseek = no_llseek,
2159 static const struct file_operations proc_fd_operations = {
2160 .read = generic_read_dir,
2161 .readdir = proc_readfd,
2162 .llseek = default_llseek,
2166 * /proc/pid/fd needs a special permission handler so that a process can still
2167 * access /proc/self/fd after it has executed a setuid().
2169 static int proc_fd_permission(struct inode *inode, int mask)
2171 int rv = generic_permission(inode, mask);
2174 if (task_pid(current) == proc_pid(inode))
2180 * proc directories can do almost nothing..
2182 static const struct inode_operations proc_fd_inode_operations = {
2183 .lookup = proc_lookupfd,
2184 .permission = proc_fd_permission,
2185 .setattr = proc_setattr,
2188 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2189 struct dentry *dentry, struct task_struct *task, const void *ptr)
2191 unsigned fd = *(unsigned *)ptr;
2192 struct inode *inode;
2193 struct proc_inode *ei;
2194 struct dentry *error = ERR_PTR(-ENOENT);
2196 inode = proc_pid_make_inode(dir->i_sb, task);
2201 inode->i_mode = S_IFREG | S_IRUSR;
2202 inode->i_fop = &proc_fdinfo_file_operations;
2203 d_set_d_op(dentry, &tid_fd_dentry_operations);
2204 d_add(dentry, inode);
2205 /* Close the race of the process dying before we return the dentry */
2206 if (tid_fd_revalidate(dentry, NULL))
2213 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2214 struct dentry *dentry,
2215 struct nameidata *nd)
2217 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2220 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2222 return proc_readfd_common(filp, dirent, filldir,
2223 proc_fdinfo_instantiate);
2226 static const struct file_operations proc_fdinfo_operations = {
2227 .read = generic_read_dir,
2228 .readdir = proc_readfdinfo,
2229 .llseek = default_llseek,
2233 * proc directories can do almost nothing..
2235 static const struct inode_operations proc_fdinfo_inode_operations = {
2236 .lookup = proc_lookupfdinfo,
2237 .setattr = proc_setattr,
2241 static struct dentry *proc_pident_instantiate(struct inode *dir,
2242 struct dentry *dentry, struct task_struct *task, const void *ptr)
2244 const struct pid_entry *p = ptr;
2245 struct inode *inode;
2246 struct proc_inode *ei;
2247 struct dentry *error = ERR_PTR(-ENOENT);
2249 inode = proc_pid_make_inode(dir->i_sb, task);
2254 inode->i_mode = p->mode;
2255 if (S_ISDIR(inode->i_mode))
2256 inode->i_nlink = 2; /* Use getattr to fix if necessary */
2258 inode->i_op = p->iop;
2260 inode->i_fop = p->fop;
2262 d_set_d_op(dentry, &pid_dentry_operations);
2263 d_add(dentry, inode);
2264 /* Close the race of the process dying before we return the dentry */
2265 if (pid_revalidate(dentry, NULL))
2271 static struct dentry *proc_pident_lookup(struct inode *dir,
2272 struct dentry *dentry,
2273 const struct pid_entry *ents,
2276 struct dentry *error;
2277 struct task_struct *task = get_proc_task(dir);
2278 const struct pid_entry *p, *last;
2280 error = ERR_PTR(-ENOENT);
2286 * Yes, it does not scale. And it should not. Don't add
2287 * new entries into /proc/<tgid>/ without very good reasons.
2289 last = &ents[nents - 1];
2290 for (p = ents; p <= last; p++) {
2291 if (p->len != dentry->d_name.len)
2293 if (!memcmp(dentry->d_name.name, p->name, p->len))
2299 error = proc_pident_instantiate(dir, dentry, task, p);
2301 put_task_struct(task);
2306 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2307 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2309 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2310 proc_pident_instantiate, task, p);
2313 static int proc_pident_readdir(struct file *filp,
2314 void *dirent, filldir_t filldir,
2315 const struct pid_entry *ents, unsigned int nents)
2318 struct dentry *dentry = filp->f_path.dentry;
2319 struct inode *inode = dentry->d_inode;
2320 struct task_struct *task = get_proc_task(inode);
2321 const struct pid_entry *p, *last;
2334 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2340 ino = parent_ino(dentry);
2341 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2353 last = &ents[nents - 1];
2355 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2364 put_task_struct(task);
2369 #ifdef CONFIG_SECURITY
2370 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2371 size_t count, loff_t *ppos)
2373 struct inode * inode = file->f_path.dentry->d_inode;
2376 struct task_struct *task = get_proc_task(inode);
2381 length = security_getprocattr(task,
2382 (char*)file->f_path.dentry->d_name.name,
2384 put_task_struct(task);
2386 length = simple_read_from_buffer(buf, count, ppos, p, length);
2391 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2392 size_t count, loff_t *ppos)
2394 struct inode * inode = file->f_path.dentry->d_inode;
2397 struct task_struct *task = get_proc_task(inode);
2402 if (count > PAGE_SIZE)
2405 /* No partial writes. */
2411 page = (char*)__get_free_page(GFP_TEMPORARY);
2416 if (copy_from_user(page, buf, count))
2419 /* Guard against adverse ptrace interaction */
2420 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2424 length = security_setprocattr(task,
2425 (char*)file->f_path.dentry->d_name.name,
2426 (void*)page, count);
2427 mutex_unlock(&task->signal->cred_guard_mutex);
2429 free_page((unsigned long) page);
2431 put_task_struct(task);
2436 static const struct file_operations proc_pid_attr_operations = {
2437 .read = proc_pid_attr_read,
2438 .write = proc_pid_attr_write,
2439 .llseek = generic_file_llseek,
2442 static const struct pid_entry attr_dir_stuff[] = {
2443 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2444 REG("prev", S_IRUGO, proc_pid_attr_operations),
2445 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2446 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2447 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2448 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2451 static int proc_attr_dir_readdir(struct file * filp,
2452 void * dirent, filldir_t filldir)
2454 return proc_pident_readdir(filp,dirent,filldir,
2455 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2458 static const struct file_operations proc_attr_dir_operations = {
2459 .read = generic_read_dir,
2460 .readdir = proc_attr_dir_readdir,
2461 .llseek = default_llseek,
2464 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2465 struct dentry *dentry, struct nameidata *nd)
2467 return proc_pident_lookup(dir, dentry,
2468 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2471 static const struct inode_operations proc_attr_dir_inode_operations = {
2472 .lookup = proc_attr_dir_lookup,
2473 .getattr = pid_getattr,
2474 .setattr = proc_setattr,
2479 #ifdef CONFIG_ELF_CORE
2480 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2481 size_t count, loff_t *ppos)
2483 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2484 struct mm_struct *mm;
2485 char buffer[PROC_NUMBUF];
2493 mm = get_task_mm(task);
2495 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2496 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2497 MMF_DUMP_FILTER_SHIFT));
2499 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2502 put_task_struct(task);
2507 static ssize_t proc_coredump_filter_write(struct file *file,
2508 const char __user *buf,
2512 struct task_struct *task;
2513 struct mm_struct *mm;
2514 char buffer[PROC_NUMBUF], *end;
2521 memset(buffer, 0, sizeof(buffer));
2522 if (count > sizeof(buffer) - 1)
2523 count = sizeof(buffer) - 1;
2524 if (copy_from_user(buffer, buf, count))
2528 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2531 if (end - buffer == 0)
2535 task = get_proc_task(file->f_dentry->d_inode);
2540 mm = get_task_mm(task);
2544 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2546 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2548 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2553 put_task_struct(task);
2558 static const struct file_operations proc_coredump_filter_operations = {
2559 .read = proc_coredump_filter_read,
2560 .write = proc_coredump_filter_write,
2561 .llseek = generic_file_llseek,
2568 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2571 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2572 pid_t tgid = task_tgid_nr_ns(current, ns);
2573 char tmp[PROC_NUMBUF];
2576 sprintf(tmp, "%d", tgid);
2577 return vfs_readlink(dentry,buffer,buflen,tmp);
2580 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2582 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2583 pid_t tgid = task_tgid_nr_ns(current, ns);
2584 char *name = ERR_PTR(-ENOENT);
2588 name = ERR_PTR(-ENOMEM);
2590 sprintf(name, "%d", tgid);
2592 nd_set_link(nd, name);
2596 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2599 char *s = nd_get_link(nd);
2604 static const struct inode_operations proc_self_inode_operations = {
2605 .readlink = proc_self_readlink,
2606 .follow_link = proc_self_follow_link,
2607 .put_link = proc_self_put_link,
2613 * These are the directory entries in the root directory of /proc
2614 * that properly belong to the /proc filesystem, as they describe
2615 * describe something that is process related.
2617 static const struct pid_entry proc_base_stuff[] = {
2618 NOD("self", S_IFLNK|S_IRWXUGO,
2619 &proc_self_inode_operations, NULL, {}),
2622 static struct dentry *proc_base_instantiate(struct inode *dir,
2623 struct dentry *dentry, struct task_struct *task, const void *ptr)
2625 const struct pid_entry *p = ptr;
2626 struct inode *inode;
2627 struct proc_inode *ei;
2628 struct dentry *error;
2630 /* Allocate the inode */
2631 error = ERR_PTR(-ENOMEM);
2632 inode = new_inode(dir->i_sb);
2636 /* Initialize the inode */
2638 inode->i_ino = get_next_ino();
2639 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2642 * grab the reference to the task.
2644 ei->pid = get_task_pid(task, PIDTYPE_PID);
2648 inode->i_mode = p->mode;
2649 if (S_ISDIR(inode->i_mode))
2651 if (S_ISLNK(inode->i_mode))
2654 inode->i_op = p->iop;
2656 inode->i_fop = p->fop;
2658 d_add(dentry, inode);
2667 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2669 struct dentry *error;
2670 struct task_struct *task = get_proc_task(dir);
2671 const struct pid_entry *p, *last;
2673 error = ERR_PTR(-ENOENT);
2678 /* Lookup the directory entry */
2679 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2680 for (p = proc_base_stuff; p <= last; p++) {
2681 if (p->len != dentry->d_name.len)
2683 if (!memcmp(dentry->d_name.name, p->name, p->len))
2689 error = proc_base_instantiate(dir, dentry, task, p);
2692 put_task_struct(task);
2697 static int proc_base_fill_cache(struct file *filp, void *dirent,
2698 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2700 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2701 proc_base_instantiate, task, p);
2704 #ifdef CONFIG_TASK_IO_ACCOUNTING
2705 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2707 struct task_io_accounting acct = task->ioac;
2708 unsigned long flags;
2711 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2715 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2720 if (whole && lock_task_sighand(task, &flags)) {
2721 struct task_struct *t = task;
2723 task_io_accounting_add(&acct, &task->signal->ioac);
2724 while_each_thread(task, t)
2725 task_io_accounting_add(&acct, &t->ioac);
2727 unlock_task_sighand(task, &flags);
2729 result = sprintf(buffer,
2734 "read_bytes: %llu\n"
2735 "write_bytes: %llu\n"
2736 "cancelled_write_bytes: %llu\n",
2737 (unsigned long long)acct.rchar,
2738 (unsigned long long)acct.wchar,
2739 (unsigned long long)acct.syscr,
2740 (unsigned long long)acct.syscw,
2741 (unsigned long long)acct.read_bytes,
2742 (unsigned long long)acct.write_bytes,
2743 (unsigned long long)acct.cancelled_write_bytes);
2745 mutex_unlock(&task->signal->cred_guard_mutex);
2749 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2751 return do_io_accounting(task, buffer, 0);
2754 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2756 return do_io_accounting(task, buffer, 1);
2758 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2760 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2761 struct pid *pid, struct task_struct *task)
2763 int err = lock_trace(task);
2765 seq_printf(m, "%08x\n", task->personality);
2774 static const struct file_operations proc_task_operations;
2775 static const struct inode_operations proc_task_inode_operations;
2777 static const struct pid_entry tgid_base_stuff[] = {
2778 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2779 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2780 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2781 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2783 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2785 REG("environ", S_IRUSR, proc_environ_operations),
2786 INF("auxv", S_IRUSR, proc_pid_auxv),
2787 ONE("status", S_IRUGO, proc_pid_status),
2788 ONE("personality", S_IRUGO, proc_pid_personality),
2789 INF("limits", S_IRUGO, proc_pid_limits),
2790 #ifdef CONFIG_SCHED_DEBUG
2791 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2793 #ifdef CONFIG_SCHED_AUTOGROUP
2794 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2796 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2797 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2798 INF("syscall", S_IRUGO, proc_pid_syscall),
2800 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2801 ONE("stat", S_IRUGO, proc_tgid_stat),
2802 ONE("statm", S_IRUGO, proc_pid_statm),
2803 REG("maps", S_IRUGO, proc_maps_operations),
2805 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2807 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2808 LNK("cwd", proc_cwd_link),
2809 LNK("root", proc_root_link),
2810 LNK("exe", proc_exe_link),
2811 REG("mounts", S_IRUGO, proc_mounts_operations),
2812 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2813 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2814 #ifdef CONFIG_PROC_PAGE_MONITOR
2815 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2816 REG("smaps", S_IRUGO, proc_smaps_operations),
2817 REG("pagemap", S_IRUGO, proc_pagemap_operations),
2819 #ifdef CONFIG_SECURITY
2820 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2822 #ifdef CONFIG_KALLSYMS
2823 INF("wchan", S_IRUGO, proc_pid_wchan),
2825 #ifdef CONFIG_STACKTRACE
2826 ONE("stack", S_IRUGO, proc_pid_stack),
2828 #ifdef CONFIG_SCHEDSTATS
2829 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2831 #ifdef CONFIG_LATENCYTOP
2832 REG("latency", S_IRUGO, proc_lstats_operations),
2834 #ifdef CONFIG_PROC_PID_CPUSET
2835 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2837 #ifdef CONFIG_CGROUPS
2838 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2840 INF("oom_score", S_IRUGO, proc_oom_score),
2841 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2842 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2843 #ifdef CONFIG_AUDITSYSCALL
2844 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2845 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2847 #ifdef CONFIG_FAULT_INJECTION
2848 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2850 #ifdef CONFIG_ELF_CORE
2851 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2853 #ifdef CONFIG_TASK_IO_ACCOUNTING
2854 INF("io", S_IRUSR, proc_tgid_io_accounting),
2856 #ifdef CONFIG_HARDWALL
2857 INF("hardwall", S_IRUGO, proc_pid_hardwall),
2861 static int proc_tgid_base_readdir(struct file * filp,
2862 void * dirent, filldir_t filldir)
2864 return proc_pident_readdir(filp,dirent,filldir,
2865 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2868 static const struct file_operations proc_tgid_base_operations = {
2869 .read = generic_read_dir,
2870 .readdir = proc_tgid_base_readdir,
2871 .llseek = default_llseek,
2874 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2875 return proc_pident_lookup(dir, dentry,
2876 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2879 static const struct inode_operations proc_tgid_base_inode_operations = {
2880 .lookup = proc_tgid_base_lookup,
2881 .getattr = pid_getattr,
2882 .setattr = proc_setattr,
2885 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2887 struct dentry *dentry, *leader, *dir;
2888 char buf[PROC_NUMBUF];
2892 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2893 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2895 shrink_dcache_parent(dentry);
2901 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2902 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2907 name.len = strlen(name.name);
2908 dir = d_hash_and_lookup(leader, &name);
2910 goto out_put_leader;
2913 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2914 dentry = d_hash_and_lookup(dir, &name);
2916 shrink_dcache_parent(dentry);
2929 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2930 * @task: task that should be flushed.
2932 * When flushing dentries from proc, one needs to flush them from global
2933 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2934 * in. This call is supposed to do all of this job.
2936 * Looks in the dcache for
2938 * /proc/@tgid/task/@pid
2939 * if either directory is present flushes it and all of it'ts children
2942 * It is safe and reasonable to cache /proc entries for a task until
2943 * that task exits. After that they just clog up the dcache with
2944 * useless entries, possibly causing useful dcache entries to be
2945 * flushed instead. This routine is proved to flush those useless
2946 * dcache entries at process exit time.
2948 * NOTE: This routine is just an optimization so it does not guarantee
2949 * that no dcache entries will exist at process exit time it
2950 * just makes it very unlikely that any will persist.
2953 void proc_flush_task(struct task_struct *task)
2956 struct pid *pid, *tgid;
2959 pid = task_pid(task);
2960 tgid = task_tgid(task);
2962 for (i = 0; i <= pid->level; i++) {
2963 upid = &pid->numbers[i];
2964 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2965 tgid->numbers[i].nr);
2968 upid = &pid->numbers[pid->level];
2970 pid_ns_release_proc(upid->ns);
2973 static struct dentry *proc_pid_instantiate(struct inode *dir,
2974 struct dentry * dentry,
2975 struct task_struct *task, const void *ptr)
2977 struct dentry *error = ERR_PTR(-ENOENT);
2978 struct inode *inode;
2980 inode = proc_pid_make_inode(dir->i_sb, task);
2984 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2985 inode->i_op = &proc_tgid_base_inode_operations;
2986 inode->i_fop = &proc_tgid_base_operations;
2987 inode->i_flags|=S_IMMUTABLE;
2989 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
2990 ARRAY_SIZE(tgid_base_stuff));
2992 d_set_d_op(dentry, &pid_dentry_operations);
2994 d_add(dentry, inode);
2995 /* Close the race of the process dying before we return the dentry */
2996 if (pid_revalidate(dentry, NULL))
3002 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3004 struct dentry *result;
3005 struct task_struct *task;
3007 struct pid_namespace *ns;
3009 result = proc_base_lookup(dir, dentry);
3010 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3013 tgid = name_to_int(dentry);
3017 ns = dentry->d_sb->s_fs_info;
3019 task = find_task_by_pid_ns(tgid, ns);
3021 get_task_struct(task);
3026 result = proc_pid_instantiate(dir, dentry, task, NULL);
3027 put_task_struct(task);
3033 * Find the first task with tgid >= tgid
3038 struct task_struct *task;
3040 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3045 put_task_struct(iter.task);
3049 pid = find_ge_pid(iter.tgid, ns);
3051 iter.tgid = pid_nr_ns(pid, ns);
3052 iter.task = pid_task(pid, PIDTYPE_PID);
3053 /* What we to know is if the pid we have find is the
3054 * pid of a thread_group_leader. Testing for task
3055 * being a thread_group_leader is the obvious thing
3056 * todo but there is a window when it fails, due to
3057 * the pid transfer logic in de_thread.
3059 * So we perform the straight forward test of seeing
3060 * if the pid we have found is the pid of a thread
3061 * group leader, and don't worry if the task we have
3062 * found doesn't happen to be a thread group leader.
3063 * As we don't care in the case of readdir.
3065 if (!iter.task || !has_group_leader_pid(iter.task)) {
3069 get_task_struct(iter.task);
3075 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
3077 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3078 struct tgid_iter iter)
3080 char name[PROC_NUMBUF];
3081 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
3082 return proc_fill_cache(filp, dirent, filldir, name, len,
3083 proc_pid_instantiate, iter.task, NULL);
3086 /* for the /proc/ directory itself, after non-process stuff has been done */
3087 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3090 struct task_struct *reaper;
3091 struct tgid_iter iter;
3092 struct pid_namespace *ns;
3094 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3096 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3098 reaper = get_proc_task(filp->f_path.dentry->d_inode);
3102 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
3103 const struct pid_entry *p = &proc_base_stuff[nr];
3104 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
3108 ns = filp->f_dentry->d_sb->s_fs_info;
3110 iter.tgid = filp->f_pos - TGID_OFFSET;
3111 for (iter = next_tgid(ns, iter);
3113 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3114 filp->f_pos = iter.tgid + TGID_OFFSET;
3115 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
3116 put_task_struct(iter.task);
3120 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3122 put_task_struct(reaper);
3130 static const struct pid_entry tid_base_stuff[] = {
3131 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3132 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3133 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3134 REG("environ", S_IRUSR, proc_environ_operations),
3135 INF("auxv", S_IRUSR, proc_pid_auxv),
3136 ONE("status", S_IRUGO, proc_pid_status),
3137 ONE("personality", S_IRUGO, proc_pid_personality),
3138 INF("limits", S_IRUGO, proc_pid_limits),
3139 #ifdef CONFIG_SCHED_DEBUG
3140 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3142 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3143 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3144 INF("syscall", S_IRUGO, proc_pid_syscall),
3146 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3147 ONE("stat", S_IRUGO, proc_tid_stat),
3148 ONE("statm", S_IRUGO, proc_pid_statm),
3149 REG("maps", S_IRUGO, proc_maps_operations),
3151 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
3153 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3154 LNK("cwd", proc_cwd_link),
3155 LNK("root", proc_root_link),
3156 LNK("exe", proc_exe_link),
3157 REG("mounts", S_IRUGO, proc_mounts_operations),
3158 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3159 #ifdef CONFIG_PROC_PAGE_MONITOR
3160 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3161 REG("smaps", S_IRUGO, proc_smaps_operations),
3162 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3164 #ifdef CONFIG_SECURITY
3165 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3167 #ifdef CONFIG_KALLSYMS
3168 INF("wchan", S_IRUGO, proc_pid_wchan),
3170 #ifdef CONFIG_STACKTRACE
3171 ONE("stack", S_IRUGO, proc_pid_stack),
3173 #ifdef CONFIG_SCHEDSTATS
3174 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3176 #ifdef CONFIG_LATENCYTOP
3177 REG("latency", S_IRUGO, proc_lstats_operations),
3179 #ifdef CONFIG_PROC_PID_CPUSET
3180 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3182 #ifdef CONFIG_CGROUPS
3183 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3185 INF("oom_score", S_IRUGO, proc_oom_score),
3186 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3187 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3188 #ifdef CONFIG_AUDITSYSCALL
3189 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3190 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3192 #ifdef CONFIG_FAULT_INJECTION
3193 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3195 #ifdef CONFIG_TASK_IO_ACCOUNTING
3196 INF("io", S_IRUSR, proc_tid_io_accounting),
3198 #ifdef CONFIG_HARDWALL
3199 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3203 static int proc_tid_base_readdir(struct file * filp,
3204 void * dirent, filldir_t filldir)
3206 return proc_pident_readdir(filp,dirent,filldir,
3207 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3210 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3211 return proc_pident_lookup(dir, dentry,
3212 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3215 static const struct file_operations proc_tid_base_operations = {
3216 .read = generic_read_dir,
3217 .readdir = proc_tid_base_readdir,
3218 .llseek = default_llseek,
3221 static const struct inode_operations proc_tid_base_inode_operations = {
3222 .lookup = proc_tid_base_lookup,
3223 .getattr = pid_getattr,
3224 .setattr = proc_setattr,
3227 static struct dentry *proc_task_instantiate(struct inode *dir,
3228 struct dentry *dentry, struct task_struct *task, const void *ptr)
3230 struct dentry *error = ERR_PTR(-ENOENT);
3231 struct inode *inode;
3232 inode = proc_pid_make_inode(dir->i_sb, task);
3236 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3237 inode->i_op = &proc_tid_base_inode_operations;
3238 inode->i_fop = &proc_tid_base_operations;
3239 inode->i_flags|=S_IMMUTABLE;
3241 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
3242 ARRAY_SIZE(tid_base_stuff));
3244 d_set_d_op(dentry, &pid_dentry_operations);
3246 d_add(dentry, inode);
3247 /* Close the race of the process dying before we return the dentry */
3248 if (pid_revalidate(dentry, NULL))
3254 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3256 struct dentry *result = ERR_PTR(-ENOENT);
3257 struct task_struct *task;
3258 struct task_struct *leader = get_proc_task(dir);
3260 struct pid_namespace *ns;
3265 tid = name_to_int(dentry);
3269 ns = dentry->d_sb->s_fs_info;
3271 task = find_task_by_pid_ns(tid, ns);
3273 get_task_struct(task);
3277 if (!same_thread_group(leader, task))
3280 result = proc_task_instantiate(dir, dentry, task, NULL);
3282 put_task_struct(task);
3284 put_task_struct(leader);
3290 * Find the first tid of a thread group to return to user space.
3292 * Usually this is just the thread group leader, but if the users
3293 * buffer was too small or there was a seek into the middle of the
3294 * directory we have more work todo.
3296 * In the case of a short read we start with find_task_by_pid.
3298 * In the case of a seek we start with the leader and walk nr
3301 static struct task_struct *first_tid(struct task_struct *leader,
3302 int tid, int nr, struct pid_namespace *ns)
3304 struct task_struct *pos;
3307 /* Attempt to start with the pid of a thread */
3308 if (tid && (nr > 0)) {
3309 pos = find_task_by_pid_ns(tid, ns);
3310 if (pos && (pos->group_leader == leader))
3314 /* If nr exceeds the number of threads there is nothing todo */
3316 if (nr && nr >= get_nr_threads(leader))
3319 /* If we haven't found our starting place yet start
3320 * with the leader and walk nr threads forward.
3322 for (pos = leader; nr > 0; --nr) {
3323 pos = next_thread(pos);
3324 if (pos == leader) {
3330 get_task_struct(pos);
3337 * Find the next thread in the thread list.
3338 * Return NULL if there is an error or no next thread.
3340 * The reference to the input task_struct is released.
3342 static struct task_struct *next_tid(struct task_struct *start)
3344 struct task_struct *pos = NULL;
3346 if (pid_alive(start)) {
3347 pos = next_thread(start);
3348 if (thread_group_leader(pos))
3351 get_task_struct(pos);
3354 put_task_struct(start);
3358 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3359 struct task_struct *task, int tid)
3361 char name[PROC_NUMBUF];
3362 int len = snprintf(name, sizeof(name), "%d", tid);
3363 return proc_fill_cache(filp, dirent, filldir, name, len,
3364 proc_task_instantiate, task, NULL);
3367 /* for the /proc/TGID/task/ directories */
3368 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3370 struct dentry *dentry = filp->f_path.dentry;
3371 struct inode *inode = dentry->d_inode;
3372 struct task_struct *leader = NULL;
3373 struct task_struct *task;
3374 int retval = -ENOENT;
3377 struct pid_namespace *ns;
3379 task = get_proc_task(inode);
3383 if (pid_alive(task)) {
3384 leader = task->group_leader;
3385 get_task_struct(leader);
3388 put_task_struct(task);
3393 switch ((unsigned long)filp->f_pos) {
3396 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3401 ino = parent_ino(dentry);
3402 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3408 /* f_version caches the tgid value that the last readdir call couldn't
3409 * return. lseek aka telldir automagically resets f_version to 0.
3411 ns = filp->f_dentry->d_sb->s_fs_info;
3412 tid = (int)filp->f_version;
3413 filp->f_version = 0;
3414 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3416 task = next_tid(task), filp->f_pos++) {
3417 tid = task_pid_nr_ns(task, ns);
3418 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3419 /* returning this tgid failed, save it as the first
3420 * pid for the next readir call */
3421 filp->f_version = (u64)tid;
3422 put_task_struct(task);
3427 put_task_struct(leader);
3432 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3434 struct inode *inode = dentry->d_inode;
3435 struct task_struct *p = get_proc_task(inode);
3436 generic_fillattr(inode, stat);
3439 stat->nlink += get_nr_threads(p);
3446 static const struct inode_operations proc_task_inode_operations = {
3447 .lookup = proc_task_lookup,
3448 .getattr = proc_task_getattr,
3449 .setattr = proc_setattr,
3452 static const struct file_operations proc_task_operations = {
3453 .read = generic_read_dir,
3454 .readdir = proc_task_readdir,
3455 .llseek = default_llseek,