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 = (tracehook_tracer_task(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->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 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, "
1122 "please use /proc/%d/oom_score_adj instead.\n",
1123 current->comm, task_pid_nr(current),
1124 task_pid_nr(task), task_pid_nr(task));
1125 task->signal->oom_adj = oom_adjust;
1127 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1128 * value is always attainable.
1130 if (task->signal->oom_adj == OOM_ADJUST_MAX)
1131 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
1133 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
1136 unlock_task_sighand(task, &flags);
1139 put_task_struct(task);
1141 return err < 0 ? err : count;
1144 static const struct file_operations proc_oom_adjust_operations = {
1145 .read = oom_adjust_read,
1146 .write = oom_adjust_write,
1147 .llseek = generic_file_llseek,
1150 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1151 size_t count, loff_t *ppos)
1153 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1154 char buffer[PROC_NUMBUF];
1155 int oom_score_adj = OOM_SCORE_ADJ_MIN;
1156 unsigned long flags;
1161 if (lock_task_sighand(task, &flags)) {
1162 oom_score_adj = task->signal->oom_score_adj;
1163 unlock_task_sighand(task, &flags);
1165 put_task_struct(task);
1166 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1167 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1170 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1171 size_t count, loff_t *ppos)
1173 struct task_struct *task;
1174 char buffer[PROC_NUMBUF];
1175 unsigned long flags;
1179 memset(buffer, 0, sizeof(buffer));
1180 if (count > sizeof(buffer) - 1)
1181 count = sizeof(buffer) - 1;
1182 if (copy_from_user(buffer, buf, count)) {
1187 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1190 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1191 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1196 task = get_proc_task(file->f_path.dentry->d_inode);
1208 if (!lock_task_sighand(task, &flags)) {
1213 if (oom_score_adj < task->signal->oom_score_adj_min &&
1214 !capable(CAP_SYS_RESOURCE)) {
1219 if (oom_score_adj != task->signal->oom_score_adj) {
1220 if (oom_score_adj == OOM_SCORE_ADJ_MIN)
1221 atomic_inc(&task->mm->oom_disable_count);
1222 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1223 atomic_dec(&task->mm->oom_disable_count);
1225 task->signal->oom_score_adj = oom_score_adj;
1226 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1227 task->signal->oom_score_adj_min = oom_score_adj;
1229 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1230 * always attainable.
1232 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1233 task->signal->oom_adj = OOM_DISABLE;
1235 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1238 unlock_task_sighand(task, &flags);
1241 put_task_struct(task);
1243 return err < 0 ? err : count;
1246 static const struct file_operations proc_oom_score_adj_operations = {
1247 .read = oom_score_adj_read,
1248 .write = oom_score_adj_write,
1249 .llseek = default_llseek,
1252 #ifdef CONFIG_AUDITSYSCALL
1253 #define TMPBUFLEN 21
1254 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1255 size_t count, loff_t *ppos)
1257 struct inode * inode = file->f_path.dentry->d_inode;
1258 struct task_struct *task = get_proc_task(inode);
1260 char tmpbuf[TMPBUFLEN];
1264 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1265 audit_get_loginuid(task));
1266 put_task_struct(task);
1267 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1270 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1271 size_t count, loff_t *ppos)
1273 struct inode * inode = file->f_path.dentry->d_inode;
1278 if (!capable(CAP_AUDIT_CONTROL))
1282 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1288 if (count >= PAGE_SIZE)
1289 count = PAGE_SIZE - 1;
1292 /* No partial writes. */
1295 page = (char*)__get_free_page(GFP_TEMPORARY);
1299 if (copy_from_user(page, buf, count))
1303 loginuid = simple_strtoul(page, &tmp, 10);
1309 length = audit_set_loginuid(current, loginuid);
1310 if (likely(length == 0))
1314 free_page((unsigned long) page);
1318 static const struct file_operations proc_loginuid_operations = {
1319 .read = proc_loginuid_read,
1320 .write = proc_loginuid_write,
1321 .llseek = generic_file_llseek,
1324 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1325 size_t count, loff_t *ppos)
1327 struct inode * inode = file->f_path.dentry->d_inode;
1328 struct task_struct *task = get_proc_task(inode);
1330 char tmpbuf[TMPBUFLEN];
1334 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1335 audit_get_sessionid(task));
1336 put_task_struct(task);
1337 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1340 static const struct file_operations proc_sessionid_operations = {
1341 .read = proc_sessionid_read,
1342 .llseek = generic_file_llseek,
1346 #ifdef CONFIG_FAULT_INJECTION
1347 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1348 size_t count, loff_t *ppos)
1350 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1351 char buffer[PROC_NUMBUF];
1357 make_it_fail = task->make_it_fail;
1358 put_task_struct(task);
1360 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1362 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1365 static ssize_t proc_fault_inject_write(struct file * file,
1366 const char __user * buf, size_t count, loff_t *ppos)
1368 struct task_struct *task;
1369 char buffer[PROC_NUMBUF], *end;
1372 if (!capable(CAP_SYS_RESOURCE))
1374 memset(buffer, 0, sizeof(buffer));
1375 if (count > sizeof(buffer) - 1)
1376 count = sizeof(buffer) - 1;
1377 if (copy_from_user(buffer, buf, count))
1379 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1382 task = get_proc_task(file->f_dentry->d_inode);
1385 task->make_it_fail = make_it_fail;
1386 put_task_struct(task);
1391 static const struct file_operations proc_fault_inject_operations = {
1392 .read = proc_fault_inject_read,
1393 .write = proc_fault_inject_write,
1394 .llseek = generic_file_llseek,
1399 #ifdef CONFIG_SCHED_DEBUG
1401 * Print out various scheduling related per-task fields:
1403 static int sched_show(struct seq_file *m, void *v)
1405 struct inode *inode = m->private;
1406 struct task_struct *p;
1408 p = get_proc_task(inode);
1411 proc_sched_show_task(p, m);
1419 sched_write(struct file *file, const char __user *buf,
1420 size_t count, loff_t *offset)
1422 struct inode *inode = file->f_path.dentry->d_inode;
1423 struct task_struct *p;
1425 p = get_proc_task(inode);
1428 proc_sched_set_task(p);
1435 static int sched_open(struct inode *inode, struct file *filp)
1437 return single_open(filp, sched_show, inode);
1440 static const struct file_operations proc_pid_sched_operations = {
1443 .write = sched_write,
1444 .llseek = seq_lseek,
1445 .release = single_release,
1450 #ifdef CONFIG_SCHED_AUTOGROUP
1452 * Print out autogroup related information:
1454 static int sched_autogroup_show(struct seq_file *m, void *v)
1456 struct inode *inode = m->private;
1457 struct task_struct *p;
1459 p = get_proc_task(inode);
1462 proc_sched_autogroup_show_task(p, m);
1470 sched_autogroup_write(struct file *file, const char __user *buf,
1471 size_t count, loff_t *offset)
1473 struct inode *inode = file->f_path.dentry->d_inode;
1474 struct task_struct *p;
1475 char buffer[PROC_NUMBUF];
1479 memset(buffer, 0, sizeof(buffer));
1480 if (count > sizeof(buffer) - 1)
1481 count = sizeof(buffer) - 1;
1482 if (copy_from_user(buffer, buf, count))
1485 err = kstrtoint(strstrip(buffer), 0, &nice);
1489 p = get_proc_task(inode);
1494 err = proc_sched_autogroup_set_nice(p, &err);
1503 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1507 ret = single_open(filp, sched_autogroup_show, NULL);
1509 struct seq_file *m = filp->private_data;
1516 static const struct file_operations proc_pid_sched_autogroup_operations = {
1517 .open = sched_autogroup_open,
1519 .write = sched_autogroup_write,
1520 .llseek = seq_lseek,
1521 .release = single_release,
1524 #endif /* CONFIG_SCHED_AUTOGROUP */
1526 static ssize_t comm_write(struct file *file, const char __user *buf,
1527 size_t count, loff_t *offset)
1529 struct inode *inode = file->f_path.dentry->d_inode;
1530 struct task_struct *p;
1531 char buffer[TASK_COMM_LEN];
1533 memset(buffer, 0, sizeof(buffer));
1534 if (count > sizeof(buffer) - 1)
1535 count = sizeof(buffer) - 1;
1536 if (copy_from_user(buffer, buf, count))
1539 p = get_proc_task(inode);
1543 if (same_thread_group(current, p))
1544 set_task_comm(p, buffer);
1553 static int comm_show(struct seq_file *m, void *v)
1555 struct inode *inode = m->private;
1556 struct task_struct *p;
1558 p = get_proc_task(inode);
1563 seq_printf(m, "%s\n", p->comm);
1571 static int comm_open(struct inode *inode, struct file *filp)
1573 return single_open(filp, comm_show, inode);
1576 static const struct file_operations proc_pid_set_comm_operations = {
1579 .write = comm_write,
1580 .llseek = seq_lseek,
1581 .release = single_release,
1584 static int proc_exe_link(struct inode *inode, struct path *exe_path)
1586 struct task_struct *task;
1587 struct mm_struct *mm;
1588 struct file *exe_file;
1590 task = get_proc_task(inode);
1593 mm = get_task_mm(task);
1594 put_task_struct(task);
1597 exe_file = get_mm_exe_file(mm);
1600 *exe_path = exe_file->f_path;
1601 path_get(&exe_file->f_path);
1608 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1610 struct inode *inode = dentry->d_inode;
1611 int error = -EACCES;
1613 /* We don't need a base pointer in the /proc filesystem */
1614 path_put(&nd->path);
1616 /* Are we allowed to snoop on the tasks file descriptors? */
1617 if (!proc_fd_access_allowed(inode))
1620 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1622 return ERR_PTR(error);
1625 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1627 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1634 pathname = d_path(path, tmp, PAGE_SIZE);
1635 len = PTR_ERR(pathname);
1636 if (IS_ERR(pathname))
1638 len = tmp + PAGE_SIZE - 1 - pathname;
1642 if (copy_to_user(buffer, pathname, len))
1645 free_page((unsigned long)tmp);
1649 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1651 int error = -EACCES;
1652 struct inode *inode = dentry->d_inode;
1655 /* Are we allowed to snoop on the tasks file descriptors? */
1656 if (!proc_fd_access_allowed(inode))
1659 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1663 error = do_proc_readlink(&path, buffer, buflen);
1669 static const struct inode_operations proc_pid_link_inode_operations = {
1670 .readlink = proc_pid_readlink,
1671 .follow_link = proc_pid_follow_link,
1672 .setattr = proc_setattr,
1676 /* building an inode */
1678 static int task_dumpable(struct task_struct *task)
1681 struct mm_struct *mm;
1686 dumpable = get_dumpable(mm);
1693 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1695 struct inode * inode;
1696 struct proc_inode *ei;
1697 const struct cred *cred;
1699 /* We need a new inode */
1701 inode = new_inode(sb);
1707 inode->i_ino = get_next_ino();
1708 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1709 inode->i_op = &proc_def_inode_operations;
1712 * grab the reference to task.
1714 ei->pid = get_task_pid(task, PIDTYPE_PID);
1718 if (task_dumpable(task)) {
1720 cred = __task_cred(task);
1721 inode->i_uid = cred->euid;
1722 inode->i_gid = cred->egid;
1725 security_task_to_inode(task, inode);
1735 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1737 struct inode *inode = dentry->d_inode;
1738 struct task_struct *task;
1739 const struct cred *cred;
1741 generic_fillattr(inode, stat);
1746 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1748 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1749 task_dumpable(task)) {
1750 cred = __task_cred(task);
1751 stat->uid = cred->euid;
1752 stat->gid = cred->egid;
1762 * Exceptional case: normally we are not allowed to unhash a busy
1763 * directory. In this case, however, we can do it - no aliasing problems
1764 * due to the way we treat inodes.
1766 * Rewrite the inode's ownerships here because the owning task may have
1767 * performed a setuid(), etc.
1769 * Before the /proc/pid/status file was created the only way to read
1770 * the effective uid of a /process was to stat /proc/pid. Reading
1771 * /proc/pid/status is slow enough that procps and other packages
1772 * kept stating /proc/pid. To keep the rules in /proc simple I have
1773 * made this apply to all per process world readable and executable
1776 int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1778 struct inode *inode;
1779 struct task_struct *task;
1780 const struct cred *cred;
1782 if (nd && nd->flags & LOOKUP_RCU)
1785 inode = dentry->d_inode;
1786 task = get_proc_task(inode);
1789 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1790 task_dumpable(task)) {
1792 cred = __task_cred(task);
1793 inode->i_uid = cred->euid;
1794 inode->i_gid = cred->egid;
1800 inode->i_mode &= ~(S_ISUID | S_ISGID);
1801 security_task_to_inode(task, inode);
1802 put_task_struct(task);
1809 static int pid_delete_dentry(const struct dentry * dentry)
1811 /* Is the task we represent dead?
1812 * If so, then don't put the dentry on the lru list,
1813 * kill it immediately.
1815 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1818 const struct dentry_operations pid_dentry_operations =
1820 .d_revalidate = pid_revalidate,
1821 .d_delete = pid_delete_dentry,
1827 * Fill a directory entry.
1829 * If possible create the dcache entry and derive our inode number and
1830 * file type from dcache entry.
1832 * Since all of the proc inode numbers are dynamically generated, the inode
1833 * numbers do not exist until the inode is cache. This means creating the
1834 * the dcache entry in readdir is necessary to keep the inode numbers
1835 * reported by readdir in sync with the inode numbers reported
1838 int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1839 const char *name, int len,
1840 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1842 struct dentry *child, *dir = filp->f_path.dentry;
1843 struct inode *inode;
1846 unsigned type = DT_UNKNOWN;
1850 qname.hash = full_name_hash(name, len);
1852 child = d_lookup(dir, &qname);
1855 new = d_alloc(dir, &qname);
1857 child = instantiate(dir->d_inode, new, task, ptr);
1864 if (!child || IS_ERR(child) || !child->d_inode)
1865 goto end_instantiate;
1866 inode = child->d_inode;
1869 type = inode->i_mode >> 12;
1874 ino = find_inode_number(dir, &qname);
1877 return filldir(dirent, name, len, filp->f_pos, ino, type);
1880 static unsigned name_to_int(struct dentry *dentry)
1882 const char *name = dentry->d_name.name;
1883 int len = dentry->d_name.len;
1886 if (len > 1 && *name == '0')
1889 unsigned c = *name++ - '0';
1892 if (n >= (~0U-9)/10)
1902 #define PROC_FDINFO_MAX 64
1904 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1906 struct task_struct *task = get_proc_task(inode);
1907 struct files_struct *files = NULL;
1909 int fd = proc_fd(inode);
1912 files = get_files_struct(task);
1913 put_task_struct(task);
1917 * We are not taking a ref to the file structure, so we must
1920 spin_lock(&files->file_lock);
1921 file = fcheck_files(files, fd);
1924 *path = file->f_path;
1925 path_get(&file->f_path);
1928 snprintf(info, PROC_FDINFO_MAX,
1931 (long long) file->f_pos,
1933 spin_unlock(&files->file_lock);
1934 put_files_struct(files);
1937 spin_unlock(&files->file_lock);
1938 put_files_struct(files);
1943 static int proc_fd_link(struct inode *inode, struct path *path)
1945 return proc_fd_info(inode, path, NULL);
1948 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1950 struct inode *inode;
1951 struct task_struct *task;
1953 struct files_struct *files;
1954 const struct cred *cred;
1956 if (nd && nd->flags & LOOKUP_RCU)
1959 inode = dentry->d_inode;
1960 task = get_proc_task(inode);
1961 fd = proc_fd(inode);
1964 files = get_files_struct(task);
1967 if (fcheck_files(files, fd)) {
1969 put_files_struct(files);
1970 if (task_dumpable(task)) {
1972 cred = __task_cred(task);
1973 inode->i_uid = cred->euid;
1974 inode->i_gid = cred->egid;
1980 inode->i_mode &= ~(S_ISUID | S_ISGID);
1981 security_task_to_inode(task, inode);
1982 put_task_struct(task);
1986 put_files_struct(files);
1988 put_task_struct(task);
1994 static const struct dentry_operations tid_fd_dentry_operations =
1996 .d_revalidate = tid_fd_revalidate,
1997 .d_delete = pid_delete_dentry,
2000 static struct dentry *proc_fd_instantiate(struct inode *dir,
2001 struct dentry *dentry, struct task_struct *task, const void *ptr)
2003 unsigned fd = *(const unsigned *)ptr;
2005 struct files_struct *files;
2006 struct inode *inode;
2007 struct proc_inode *ei;
2008 struct dentry *error = ERR_PTR(-ENOENT);
2010 inode = proc_pid_make_inode(dir->i_sb, task);
2015 files = get_files_struct(task);
2018 inode->i_mode = S_IFLNK;
2021 * We are not taking a ref to the file structure, so we must
2024 spin_lock(&files->file_lock);
2025 file = fcheck_files(files, fd);
2028 if (file->f_mode & FMODE_READ)
2029 inode->i_mode |= S_IRUSR | S_IXUSR;
2030 if (file->f_mode & FMODE_WRITE)
2031 inode->i_mode |= S_IWUSR | S_IXUSR;
2032 spin_unlock(&files->file_lock);
2033 put_files_struct(files);
2035 inode->i_op = &proc_pid_link_inode_operations;
2037 ei->op.proc_get_link = proc_fd_link;
2038 d_set_d_op(dentry, &tid_fd_dentry_operations);
2039 d_add(dentry, inode);
2040 /* Close the race of the process dying before we return the dentry */
2041 if (tid_fd_revalidate(dentry, NULL))
2047 spin_unlock(&files->file_lock);
2048 put_files_struct(files);
2054 static struct dentry *proc_lookupfd_common(struct inode *dir,
2055 struct dentry *dentry,
2056 instantiate_t instantiate)
2058 struct task_struct *task = get_proc_task(dir);
2059 unsigned fd = name_to_int(dentry);
2060 struct dentry *result = ERR_PTR(-ENOENT);
2067 result = instantiate(dir, dentry, task, &fd);
2069 put_task_struct(task);
2074 static int proc_readfd_common(struct file * filp, void * dirent,
2075 filldir_t filldir, instantiate_t instantiate)
2077 struct dentry *dentry = filp->f_path.dentry;
2078 struct inode *inode = dentry->d_inode;
2079 struct task_struct *p = get_proc_task(inode);
2080 unsigned int fd, ino;
2082 struct files_struct * files;
2092 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
2096 ino = parent_ino(dentry);
2097 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2101 files = get_files_struct(p);
2105 for (fd = filp->f_pos-2;
2106 fd < files_fdtable(files)->max_fds;
2107 fd++, filp->f_pos++) {
2108 char name[PROC_NUMBUF];
2111 if (!fcheck_files(files, fd))
2115 len = snprintf(name, sizeof(name), "%d", fd);
2116 if (proc_fill_cache(filp, dirent, filldir,
2117 name, len, instantiate,
2125 put_files_struct(files);
2133 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
2134 struct nameidata *nd)
2136 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
2139 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
2141 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
2144 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
2145 size_t len, loff_t *ppos)
2147 char tmp[PROC_FDINFO_MAX];
2148 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
2150 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
2154 static const struct file_operations proc_fdinfo_file_operations = {
2155 .open = nonseekable_open,
2156 .read = proc_fdinfo_read,
2157 .llseek = no_llseek,
2160 static const struct file_operations proc_fd_operations = {
2161 .read = generic_read_dir,
2162 .readdir = proc_readfd,
2163 .llseek = default_llseek,
2167 * /proc/pid/fd needs a special permission handler so that a process can still
2168 * access /proc/self/fd after it has executed a setuid().
2170 static int proc_fd_permission(struct inode *inode, int mask, unsigned int flags)
2174 if (flags & IPERM_FLAG_RCU)
2176 rv = generic_permission(inode, mask, flags, NULL);
2179 if (task_pid(current) == proc_pid(inode))
2185 * proc directories can do almost nothing..
2187 static const struct inode_operations proc_fd_inode_operations = {
2188 .lookup = proc_lookupfd,
2189 .permission = proc_fd_permission,
2190 .setattr = proc_setattr,
2193 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2194 struct dentry *dentry, struct task_struct *task, const void *ptr)
2196 unsigned fd = *(unsigned *)ptr;
2197 struct inode *inode;
2198 struct proc_inode *ei;
2199 struct dentry *error = ERR_PTR(-ENOENT);
2201 inode = proc_pid_make_inode(dir->i_sb, task);
2206 inode->i_mode = S_IFREG | S_IRUSR;
2207 inode->i_fop = &proc_fdinfo_file_operations;
2208 d_set_d_op(dentry, &tid_fd_dentry_operations);
2209 d_add(dentry, inode);
2210 /* Close the race of the process dying before we return the dentry */
2211 if (tid_fd_revalidate(dentry, NULL))
2218 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2219 struct dentry *dentry,
2220 struct nameidata *nd)
2222 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2225 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2227 return proc_readfd_common(filp, dirent, filldir,
2228 proc_fdinfo_instantiate);
2231 static const struct file_operations proc_fdinfo_operations = {
2232 .read = generic_read_dir,
2233 .readdir = proc_readfdinfo,
2234 .llseek = default_llseek,
2238 * proc directories can do almost nothing..
2240 static const struct inode_operations proc_fdinfo_inode_operations = {
2241 .lookup = proc_lookupfdinfo,
2242 .setattr = proc_setattr,
2246 static struct dentry *proc_pident_instantiate(struct inode *dir,
2247 struct dentry *dentry, struct task_struct *task, const void *ptr)
2249 const struct pid_entry *p = ptr;
2250 struct inode *inode;
2251 struct proc_inode *ei;
2252 struct dentry *error = ERR_PTR(-ENOENT);
2254 inode = proc_pid_make_inode(dir->i_sb, task);
2259 inode->i_mode = p->mode;
2260 if (S_ISDIR(inode->i_mode))
2261 inode->i_nlink = 2; /* Use getattr to fix if necessary */
2263 inode->i_op = p->iop;
2265 inode->i_fop = p->fop;
2267 d_set_d_op(dentry, &pid_dentry_operations);
2268 d_add(dentry, inode);
2269 /* Close the race of the process dying before we return the dentry */
2270 if (pid_revalidate(dentry, NULL))
2276 static struct dentry *proc_pident_lookup(struct inode *dir,
2277 struct dentry *dentry,
2278 const struct pid_entry *ents,
2281 struct dentry *error;
2282 struct task_struct *task = get_proc_task(dir);
2283 const struct pid_entry *p, *last;
2285 error = ERR_PTR(-ENOENT);
2291 * Yes, it does not scale. And it should not. Don't add
2292 * new entries into /proc/<tgid>/ without very good reasons.
2294 last = &ents[nents - 1];
2295 for (p = ents; p <= last; p++) {
2296 if (p->len != dentry->d_name.len)
2298 if (!memcmp(dentry->d_name.name, p->name, p->len))
2304 error = proc_pident_instantiate(dir, dentry, task, p);
2306 put_task_struct(task);
2311 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2312 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2314 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2315 proc_pident_instantiate, task, p);
2318 static int proc_pident_readdir(struct file *filp,
2319 void *dirent, filldir_t filldir,
2320 const struct pid_entry *ents, unsigned int nents)
2323 struct dentry *dentry = filp->f_path.dentry;
2324 struct inode *inode = dentry->d_inode;
2325 struct task_struct *task = get_proc_task(inode);
2326 const struct pid_entry *p, *last;
2339 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2345 ino = parent_ino(dentry);
2346 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2358 last = &ents[nents - 1];
2360 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2369 put_task_struct(task);
2374 #ifdef CONFIG_SECURITY
2375 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2376 size_t count, loff_t *ppos)
2378 struct inode * inode = file->f_path.dentry->d_inode;
2381 struct task_struct *task = get_proc_task(inode);
2386 length = security_getprocattr(task,
2387 (char*)file->f_path.dentry->d_name.name,
2389 put_task_struct(task);
2391 length = simple_read_from_buffer(buf, count, ppos, p, length);
2396 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2397 size_t count, loff_t *ppos)
2399 struct inode * inode = file->f_path.dentry->d_inode;
2402 struct task_struct *task = get_proc_task(inode);
2407 if (count > PAGE_SIZE)
2410 /* No partial writes. */
2416 page = (char*)__get_free_page(GFP_TEMPORARY);
2421 if (copy_from_user(page, buf, count))
2424 /* Guard against adverse ptrace interaction */
2425 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2429 length = security_setprocattr(task,
2430 (char*)file->f_path.dentry->d_name.name,
2431 (void*)page, count);
2432 mutex_unlock(&task->signal->cred_guard_mutex);
2434 free_page((unsigned long) page);
2436 put_task_struct(task);
2441 static const struct file_operations proc_pid_attr_operations = {
2442 .read = proc_pid_attr_read,
2443 .write = proc_pid_attr_write,
2444 .llseek = generic_file_llseek,
2447 static const struct pid_entry attr_dir_stuff[] = {
2448 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2449 REG("prev", S_IRUGO, proc_pid_attr_operations),
2450 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2451 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2452 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2453 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2456 static int proc_attr_dir_readdir(struct file * filp,
2457 void * dirent, filldir_t filldir)
2459 return proc_pident_readdir(filp,dirent,filldir,
2460 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2463 static const struct file_operations proc_attr_dir_operations = {
2464 .read = generic_read_dir,
2465 .readdir = proc_attr_dir_readdir,
2466 .llseek = default_llseek,
2469 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2470 struct dentry *dentry, struct nameidata *nd)
2472 return proc_pident_lookup(dir, dentry,
2473 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2476 static const struct inode_operations proc_attr_dir_inode_operations = {
2477 .lookup = proc_attr_dir_lookup,
2478 .getattr = pid_getattr,
2479 .setattr = proc_setattr,
2484 #ifdef CONFIG_ELF_CORE
2485 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2486 size_t count, loff_t *ppos)
2488 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2489 struct mm_struct *mm;
2490 char buffer[PROC_NUMBUF];
2498 mm = get_task_mm(task);
2500 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2501 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2502 MMF_DUMP_FILTER_SHIFT));
2504 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2507 put_task_struct(task);
2512 static ssize_t proc_coredump_filter_write(struct file *file,
2513 const char __user *buf,
2517 struct task_struct *task;
2518 struct mm_struct *mm;
2519 char buffer[PROC_NUMBUF], *end;
2526 memset(buffer, 0, sizeof(buffer));
2527 if (count > sizeof(buffer) - 1)
2528 count = sizeof(buffer) - 1;
2529 if (copy_from_user(buffer, buf, count))
2533 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2536 if (end - buffer == 0)
2540 task = get_proc_task(file->f_dentry->d_inode);
2545 mm = get_task_mm(task);
2549 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2551 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2553 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2558 put_task_struct(task);
2563 static const struct file_operations proc_coredump_filter_operations = {
2564 .read = proc_coredump_filter_read,
2565 .write = proc_coredump_filter_write,
2566 .llseek = generic_file_llseek,
2573 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2576 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2577 pid_t tgid = task_tgid_nr_ns(current, ns);
2578 char tmp[PROC_NUMBUF];
2581 sprintf(tmp, "%d", tgid);
2582 return vfs_readlink(dentry,buffer,buflen,tmp);
2585 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2587 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2588 pid_t tgid = task_tgid_nr_ns(current, ns);
2589 char *name = ERR_PTR(-ENOENT);
2593 name = ERR_PTR(-ENOMEM);
2595 sprintf(name, "%d", tgid);
2597 nd_set_link(nd, name);
2601 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2604 char *s = nd_get_link(nd);
2609 static const struct inode_operations proc_self_inode_operations = {
2610 .readlink = proc_self_readlink,
2611 .follow_link = proc_self_follow_link,
2612 .put_link = proc_self_put_link,
2618 * These are the directory entries in the root directory of /proc
2619 * that properly belong to the /proc filesystem, as they describe
2620 * describe something that is process related.
2622 static const struct pid_entry proc_base_stuff[] = {
2623 NOD("self", S_IFLNK|S_IRWXUGO,
2624 &proc_self_inode_operations, NULL, {}),
2627 static struct dentry *proc_base_instantiate(struct inode *dir,
2628 struct dentry *dentry, struct task_struct *task, const void *ptr)
2630 const struct pid_entry *p = ptr;
2631 struct inode *inode;
2632 struct proc_inode *ei;
2633 struct dentry *error;
2635 /* Allocate the inode */
2636 error = ERR_PTR(-ENOMEM);
2637 inode = new_inode(dir->i_sb);
2641 /* Initialize the inode */
2643 inode->i_ino = get_next_ino();
2644 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2647 * grab the reference to the task.
2649 ei->pid = get_task_pid(task, PIDTYPE_PID);
2653 inode->i_mode = p->mode;
2654 if (S_ISDIR(inode->i_mode))
2656 if (S_ISLNK(inode->i_mode))
2659 inode->i_op = p->iop;
2661 inode->i_fop = p->fop;
2663 d_add(dentry, inode);
2672 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2674 struct dentry *error;
2675 struct task_struct *task = get_proc_task(dir);
2676 const struct pid_entry *p, *last;
2678 error = ERR_PTR(-ENOENT);
2683 /* Lookup the directory entry */
2684 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2685 for (p = proc_base_stuff; p <= last; p++) {
2686 if (p->len != dentry->d_name.len)
2688 if (!memcmp(dentry->d_name.name, p->name, p->len))
2694 error = proc_base_instantiate(dir, dentry, task, p);
2697 put_task_struct(task);
2702 static int proc_base_fill_cache(struct file *filp, void *dirent,
2703 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2705 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2706 proc_base_instantiate, task, p);
2709 #ifdef CONFIG_TASK_IO_ACCOUNTING
2710 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2712 struct task_io_accounting acct = task->ioac;
2713 unsigned long flags;
2715 if (whole && lock_task_sighand(task, &flags)) {
2716 struct task_struct *t = task;
2718 task_io_accounting_add(&acct, &task->signal->ioac);
2719 while_each_thread(task, t)
2720 task_io_accounting_add(&acct, &t->ioac);
2722 unlock_task_sighand(task, &flags);
2724 return sprintf(buffer,
2729 "read_bytes: %llu\n"
2730 "write_bytes: %llu\n"
2731 "cancelled_write_bytes: %llu\n",
2732 (unsigned long long)acct.rchar,
2733 (unsigned long long)acct.wchar,
2734 (unsigned long long)acct.syscr,
2735 (unsigned long long)acct.syscw,
2736 (unsigned long long)acct.read_bytes,
2737 (unsigned long long)acct.write_bytes,
2738 (unsigned long long)acct.cancelled_write_bytes);
2741 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2743 return do_io_accounting(task, buffer, 0);
2746 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2748 return do_io_accounting(task, buffer, 1);
2750 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2752 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2753 struct pid *pid, struct task_struct *task)
2755 int err = lock_trace(task);
2757 seq_printf(m, "%08x\n", task->personality);
2766 static const struct file_operations proc_task_operations;
2767 static const struct inode_operations proc_task_inode_operations;
2769 static const struct pid_entry tgid_base_stuff[] = {
2770 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2771 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2772 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2773 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2775 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2777 REG("environ", S_IRUSR, proc_environ_operations),
2778 INF("auxv", S_IRUSR, proc_pid_auxv),
2779 ONE("status", S_IRUGO, proc_pid_status),
2780 ONE("personality", S_IRUGO, proc_pid_personality),
2781 INF("limits", S_IRUGO, proc_pid_limits),
2782 #ifdef CONFIG_SCHED_DEBUG
2783 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2785 #ifdef CONFIG_SCHED_AUTOGROUP
2786 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2788 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2789 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2790 INF("syscall", S_IRUGO, proc_pid_syscall),
2792 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2793 ONE("stat", S_IRUGO, proc_tgid_stat),
2794 ONE("statm", S_IRUGO, proc_pid_statm),
2795 REG("maps", S_IRUGO, proc_maps_operations),
2797 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2799 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2800 LNK("cwd", proc_cwd_link),
2801 LNK("root", proc_root_link),
2802 LNK("exe", proc_exe_link),
2803 REG("mounts", S_IRUGO, proc_mounts_operations),
2804 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2805 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2806 #ifdef CONFIG_PROC_PAGE_MONITOR
2807 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2808 REG("smaps", S_IRUGO, proc_smaps_operations),
2809 REG("pagemap", S_IRUGO, proc_pagemap_operations),
2811 #ifdef CONFIG_SECURITY
2812 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2814 #ifdef CONFIG_KALLSYMS
2815 INF("wchan", S_IRUGO, proc_pid_wchan),
2817 #ifdef CONFIG_STACKTRACE
2818 ONE("stack", S_IRUGO, proc_pid_stack),
2820 #ifdef CONFIG_SCHEDSTATS
2821 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2823 #ifdef CONFIG_LATENCYTOP
2824 REG("latency", S_IRUGO, proc_lstats_operations),
2826 #ifdef CONFIG_PROC_PID_CPUSET
2827 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2829 #ifdef CONFIG_CGROUPS
2830 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2832 INF("oom_score", S_IRUGO, proc_oom_score),
2833 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2834 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2835 #ifdef CONFIG_AUDITSYSCALL
2836 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2837 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2839 #ifdef CONFIG_FAULT_INJECTION
2840 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2842 #ifdef CONFIG_ELF_CORE
2843 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2845 #ifdef CONFIG_TASK_IO_ACCOUNTING
2846 INF("io", S_IRUGO, proc_tgid_io_accounting),
2848 #ifdef CONFIG_HARDWALL
2849 INF("hardwall", S_IRUGO, proc_pid_hardwall),
2853 static int proc_tgid_base_readdir(struct file * filp,
2854 void * dirent, filldir_t filldir)
2856 return proc_pident_readdir(filp,dirent,filldir,
2857 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2860 static const struct file_operations proc_tgid_base_operations = {
2861 .read = generic_read_dir,
2862 .readdir = proc_tgid_base_readdir,
2863 .llseek = default_llseek,
2866 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2867 return proc_pident_lookup(dir, dentry,
2868 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2871 static const struct inode_operations proc_tgid_base_inode_operations = {
2872 .lookup = proc_tgid_base_lookup,
2873 .getattr = pid_getattr,
2874 .setattr = proc_setattr,
2877 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2879 struct dentry *dentry, *leader, *dir;
2880 char buf[PROC_NUMBUF];
2884 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2885 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2887 shrink_dcache_parent(dentry);
2893 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2894 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2899 name.len = strlen(name.name);
2900 dir = d_hash_and_lookup(leader, &name);
2902 goto out_put_leader;
2905 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2906 dentry = d_hash_and_lookup(dir, &name);
2908 shrink_dcache_parent(dentry);
2921 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2922 * @task: task that should be flushed.
2924 * When flushing dentries from proc, one needs to flush them from global
2925 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2926 * in. This call is supposed to do all of this job.
2928 * Looks in the dcache for
2930 * /proc/@tgid/task/@pid
2931 * if either directory is present flushes it and all of it'ts children
2934 * It is safe and reasonable to cache /proc entries for a task until
2935 * that task exits. After that they just clog up the dcache with
2936 * useless entries, possibly causing useful dcache entries to be
2937 * flushed instead. This routine is proved to flush those useless
2938 * dcache entries at process exit time.
2940 * NOTE: This routine is just an optimization so it does not guarantee
2941 * that no dcache entries will exist at process exit time it
2942 * just makes it very unlikely that any will persist.
2945 void proc_flush_task(struct task_struct *task)
2948 struct pid *pid, *tgid;
2951 pid = task_pid(task);
2952 tgid = task_tgid(task);
2954 for (i = 0; i <= pid->level; i++) {
2955 upid = &pid->numbers[i];
2956 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2957 tgid->numbers[i].nr);
2960 upid = &pid->numbers[pid->level];
2962 pid_ns_release_proc(upid->ns);
2965 static struct dentry *proc_pid_instantiate(struct inode *dir,
2966 struct dentry * dentry,
2967 struct task_struct *task, const void *ptr)
2969 struct dentry *error = ERR_PTR(-ENOENT);
2970 struct inode *inode;
2972 inode = proc_pid_make_inode(dir->i_sb, task);
2976 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2977 inode->i_op = &proc_tgid_base_inode_operations;
2978 inode->i_fop = &proc_tgid_base_operations;
2979 inode->i_flags|=S_IMMUTABLE;
2981 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
2982 ARRAY_SIZE(tgid_base_stuff));
2984 d_set_d_op(dentry, &pid_dentry_operations);
2986 d_add(dentry, inode);
2987 /* Close the race of the process dying before we return the dentry */
2988 if (pid_revalidate(dentry, NULL))
2994 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2996 struct dentry *result;
2997 struct task_struct *task;
2999 struct pid_namespace *ns;
3001 result = proc_base_lookup(dir, dentry);
3002 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3005 tgid = name_to_int(dentry);
3009 ns = dentry->d_sb->s_fs_info;
3011 task = find_task_by_pid_ns(tgid, ns);
3013 get_task_struct(task);
3018 result = proc_pid_instantiate(dir, dentry, task, NULL);
3019 put_task_struct(task);
3025 * Find the first task with tgid >= tgid
3030 struct task_struct *task;
3032 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3037 put_task_struct(iter.task);
3041 pid = find_ge_pid(iter.tgid, ns);
3043 iter.tgid = pid_nr_ns(pid, ns);
3044 iter.task = pid_task(pid, PIDTYPE_PID);
3045 /* What we to know is if the pid we have find is the
3046 * pid of a thread_group_leader. Testing for task
3047 * being a thread_group_leader is the obvious thing
3048 * todo but there is a window when it fails, due to
3049 * the pid transfer logic in de_thread.
3051 * So we perform the straight forward test of seeing
3052 * if the pid we have found is the pid of a thread
3053 * group leader, and don't worry if the task we have
3054 * found doesn't happen to be a thread group leader.
3055 * As we don't care in the case of readdir.
3057 if (!iter.task || !has_group_leader_pid(iter.task)) {
3061 get_task_struct(iter.task);
3067 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
3069 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3070 struct tgid_iter iter)
3072 char name[PROC_NUMBUF];
3073 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
3074 return proc_fill_cache(filp, dirent, filldir, name, len,
3075 proc_pid_instantiate, iter.task, NULL);
3078 /* for the /proc/ directory itself, after non-process stuff has been done */
3079 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3082 struct task_struct *reaper;
3083 struct tgid_iter iter;
3084 struct pid_namespace *ns;
3086 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3088 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3090 reaper = get_proc_task(filp->f_path.dentry->d_inode);
3094 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
3095 const struct pid_entry *p = &proc_base_stuff[nr];
3096 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
3100 ns = filp->f_dentry->d_sb->s_fs_info;
3102 iter.tgid = filp->f_pos - TGID_OFFSET;
3103 for (iter = next_tgid(ns, iter);
3105 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3106 filp->f_pos = iter.tgid + TGID_OFFSET;
3107 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
3108 put_task_struct(iter.task);
3112 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3114 put_task_struct(reaper);
3122 static const struct pid_entry tid_base_stuff[] = {
3123 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3124 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3125 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3126 REG("environ", S_IRUSR, proc_environ_operations),
3127 INF("auxv", S_IRUSR, proc_pid_auxv),
3128 ONE("status", S_IRUGO, proc_pid_status),
3129 ONE("personality", S_IRUGO, proc_pid_personality),
3130 INF("limits", S_IRUGO, proc_pid_limits),
3131 #ifdef CONFIG_SCHED_DEBUG
3132 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3134 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3135 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3136 INF("syscall", S_IRUGO, proc_pid_syscall),
3138 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3139 ONE("stat", S_IRUGO, proc_tid_stat),
3140 ONE("statm", S_IRUGO, proc_pid_statm),
3141 REG("maps", S_IRUGO, proc_maps_operations),
3143 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
3145 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3146 LNK("cwd", proc_cwd_link),
3147 LNK("root", proc_root_link),
3148 LNK("exe", proc_exe_link),
3149 REG("mounts", S_IRUGO, proc_mounts_operations),
3150 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3151 #ifdef CONFIG_PROC_PAGE_MONITOR
3152 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3153 REG("smaps", S_IRUGO, proc_smaps_operations),
3154 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3156 #ifdef CONFIG_SECURITY
3157 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3159 #ifdef CONFIG_KALLSYMS
3160 INF("wchan", S_IRUGO, proc_pid_wchan),
3162 #ifdef CONFIG_STACKTRACE
3163 ONE("stack", S_IRUGO, proc_pid_stack),
3165 #ifdef CONFIG_SCHEDSTATS
3166 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3168 #ifdef CONFIG_LATENCYTOP
3169 REG("latency", S_IRUGO, proc_lstats_operations),
3171 #ifdef CONFIG_PROC_PID_CPUSET
3172 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3174 #ifdef CONFIG_CGROUPS
3175 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3177 INF("oom_score", S_IRUGO, proc_oom_score),
3178 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3179 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3180 #ifdef CONFIG_AUDITSYSCALL
3181 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3182 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3184 #ifdef CONFIG_FAULT_INJECTION
3185 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3187 #ifdef CONFIG_TASK_IO_ACCOUNTING
3188 INF("io", S_IRUGO, proc_tid_io_accounting),
3190 #ifdef CONFIG_HARDWALL
3191 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3195 static int proc_tid_base_readdir(struct file * filp,
3196 void * dirent, filldir_t filldir)
3198 return proc_pident_readdir(filp,dirent,filldir,
3199 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3202 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3203 return proc_pident_lookup(dir, dentry,
3204 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3207 static const struct file_operations proc_tid_base_operations = {
3208 .read = generic_read_dir,
3209 .readdir = proc_tid_base_readdir,
3210 .llseek = default_llseek,
3213 static const struct inode_operations proc_tid_base_inode_operations = {
3214 .lookup = proc_tid_base_lookup,
3215 .getattr = pid_getattr,
3216 .setattr = proc_setattr,
3219 static struct dentry *proc_task_instantiate(struct inode *dir,
3220 struct dentry *dentry, struct task_struct *task, const void *ptr)
3222 struct dentry *error = ERR_PTR(-ENOENT);
3223 struct inode *inode;
3224 inode = proc_pid_make_inode(dir->i_sb, task);
3228 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3229 inode->i_op = &proc_tid_base_inode_operations;
3230 inode->i_fop = &proc_tid_base_operations;
3231 inode->i_flags|=S_IMMUTABLE;
3233 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
3234 ARRAY_SIZE(tid_base_stuff));
3236 d_set_d_op(dentry, &pid_dentry_operations);
3238 d_add(dentry, inode);
3239 /* Close the race of the process dying before we return the dentry */
3240 if (pid_revalidate(dentry, NULL))
3246 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3248 struct dentry *result = ERR_PTR(-ENOENT);
3249 struct task_struct *task;
3250 struct task_struct *leader = get_proc_task(dir);
3252 struct pid_namespace *ns;
3257 tid = name_to_int(dentry);
3261 ns = dentry->d_sb->s_fs_info;
3263 task = find_task_by_pid_ns(tid, ns);
3265 get_task_struct(task);
3269 if (!same_thread_group(leader, task))
3272 result = proc_task_instantiate(dir, dentry, task, NULL);
3274 put_task_struct(task);
3276 put_task_struct(leader);
3282 * Find the first tid of a thread group to return to user space.
3284 * Usually this is just the thread group leader, but if the users
3285 * buffer was too small or there was a seek into the middle of the
3286 * directory we have more work todo.
3288 * In the case of a short read we start with find_task_by_pid.
3290 * In the case of a seek we start with the leader and walk nr
3293 static struct task_struct *first_tid(struct task_struct *leader,
3294 int tid, int nr, struct pid_namespace *ns)
3296 struct task_struct *pos;
3299 /* Attempt to start with the pid of a thread */
3300 if (tid && (nr > 0)) {
3301 pos = find_task_by_pid_ns(tid, ns);
3302 if (pos && (pos->group_leader == leader))
3306 /* If nr exceeds the number of threads there is nothing todo */
3308 if (nr && nr >= get_nr_threads(leader))
3311 /* If we haven't found our starting place yet start
3312 * with the leader and walk nr threads forward.
3314 for (pos = leader; nr > 0; --nr) {
3315 pos = next_thread(pos);
3316 if (pos == leader) {
3322 get_task_struct(pos);
3329 * Find the next thread in the thread list.
3330 * Return NULL if there is an error or no next thread.
3332 * The reference to the input task_struct is released.
3334 static struct task_struct *next_tid(struct task_struct *start)
3336 struct task_struct *pos = NULL;
3338 if (pid_alive(start)) {
3339 pos = next_thread(start);
3340 if (thread_group_leader(pos))
3343 get_task_struct(pos);
3346 put_task_struct(start);
3350 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3351 struct task_struct *task, int tid)
3353 char name[PROC_NUMBUF];
3354 int len = snprintf(name, sizeof(name), "%d", tid);
3355 return proc_fill_cache(filp, dirent, filldir, name, len,
3356 proc_task_instantiate, task, NULL);
3359 /* for the /proc/TGID/task/ directories */
3360 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3362 struct dentry *dentry = filp->f_path.dentry;
3363 struct inode *inode = dentry->d_inode;
3364 struct task_struct *leader = NULL;
3365 struct task_struct *task;
3366 int retval = -ENOENT;
3369 struct pid_namespace *ns;
3371 task = get_proc_task(inode);
3375 if (pid_alive(task)) {
3376 leader = task->group_leader;
3377 get_task_struct(leader);
3380 put_task_struct(task);
3385 switch ((unsigned long)filp->f_pos) {
3388 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3393 ino = parent_ino(dentry);
3394 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3400 /* f_version caches the tgid value that the last readdir call couldn't
3401 * return. lseek aka telldir automagically resets f_version to 0.
3403 ns = filp->f_dentry->d_sb->s_fs_info;
3404 tid = (int)filp->f_version;
3405 filp->f_version = 0;
3406 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3408 task = next_tid(task), filp->f_pos++) {
3409 tid = task_pid_nr_ns(task, ns);
3410 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3411 /* returning this tgid failed, save it as the first
3412 * pid for the next readir call */
3413 filp->f_version = (u64)tid;
3414 put_task_struct(task);
3419 put_task_struct(leader);
3424 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3426 struct inode *inode = dentry->d_inode;
3427 struct task_struct *p = get_proc_task(inode);
3428 generic_fillattr(inode, stat);
3431 stat->nlink += get_nr_threads(p);
3438 static const struct inode_operations proc_task_inode_operations = {
3439 .lookup = proc_task_lookup,
3440 .getattr = proc_task_getattr,
3441 .setattr = proc_setattr,
3444 static const struct file_operations proc_task_operations = {
3445 .read = generic_read_dir,
3446 .readdir = proc_task_readdir,
3447 .llseek = default_llseek,