4 * Copyright (C) 1991, 1992 Linus Torvalds
6 * proc base directory handling functions
8 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
9 * Instead of using magical inumbers to determine the kind of object
10 * we allocate and fill in-core inodes upon lookup. They don't even
11 * go into icache. We cache the reference to task_struct upon lookup too.
12 * Eventually it should become a filesystem in its own. We don't use the
13 * rest of procfs anymore.
19 * Bruna Moreira <bruna.moreira@indt.org.br>
20 * Edjard Mota <edjard.mota@indt.org.br>
21 * Ilias Biris <ilias.biris@indt.org.br>
22 * Mauricio Lin <mauricio.lin@indt.org.br>
24 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
26 * A new process specific entry (smaps) included in /proc. It shows the
27 * size of rss for each memory area. The maps entry lacks information
28 * about physical memory size (rss) for each mapped file, i.e.,
29 * rss information for executables and library files.
30 * This additional information is useful for any tools that need to know
31 * about physical memory consumption for a process specific library.
35 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36 * Pud inclusion in the page table walking.
40 * 10LE Instituto Nokia de Tecnologia - INdT:
41 * A better way to walks through the page table as suggested by Hugh Dickins.
43 * Simo Piiroinen <simo.piiroinen@nokia.com>:
44 * Smaps information related to shared, private, clean and dirty pages.
46 * Paul Mundt <paul.mundt@nokia.com>:
47 * Overall revision about smaps.
50 #include <asm/uaccess.h>
52 #include <linux/errno.h>
53 #include <linux/time.h>
54 #include <linux/proc_fs.h>
55 #include <linux/stat.h>
56 #include <linux/task_io_accounting_ops.h>
57 #include <linux/init.h>
58 #include <linux/capability.h>
59 #include <linux/file.h>
60 #include <linux/fdtable.h>
61 #include <linux/string.h>
62 #include <linux/seq_file.h>
63 #include <linux/namei.h>
64 #include <linux/mnt_namespace.h>
66 #include <linux/swap.h>
67 #include <linux/rcupdate.h>
68 #include <linux/kallsyms.h>
69 #include <linux/stacktrace.h>
70 #include <linux/resource.h>
71 #include <linux/module.h>
72 #include <linux/mount.h>
73 #include <linux/security.h>
74 #include <linux/ptrace.h>
75 #include <linux/tracehook.h>
76 #include <linux/cgroup.h>
77 #include <linux/cpuset.h>
78 #include <linux/audit.h>
79 #include <linux/poll.h>
80 #include <linux/nsproxy.h>
81 #include <linux/oom.h>
82 #include <linux/elf.h>
83 #include <linux/pid_namespace.h>
84 #include <linux/fs_struct.h>
85 #include <linux/slab.h>
89 * Implementing inode permission operations in /proc is almost
90 * certainly an error. Permission checks need to happen during
91 * each system call not at open time. The reason is that most of
92 * what we wish to check for permissions in /proc varies at runtime.
94 * The classic example of a problem is opening file descriptors
95 * in /proc for a task before it execs a suid executable.
102 const struct inode_operations *iop;
103 const struct file_operations *fop;
107 #define NOD(NAME, MODE, IOP, FOP, OP) { \
109 .len = sizeof(NAME) - 1, \
116 #define DIR(NAME, MODE, iops, fops) \
117 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
118 #define LNK(NAME, get_link) \
119 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
120 &proc_pid_link_inode_operations, NULL, \
121 { .proc_get_link = get_link } )
122 #define REG(NAME, MODE, fops) \
123 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
124 #define INF(NAME, MODE, read) \
125 NOD(NAME, (S_IFREG|(MODE)), \
126 NULL, &proc_info_file_operations, \
127 { .proc_read = read } )
128 #define ONE(NAME, MODE, show) \
129 NOD(NAME, (S_IFREG|(MODE)), \
130 NULL, &proc_single_file_operations, \
131 { .proc_show = show } )
134 * Count the number of hardlinks for the pid_entry table, excluding the .
137 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
144 for (i = 0; i < n; ++i) {
145 if (S_ISDIR(entries[i].mode))
152 static int get_task_root(struct task_struct *task, struct path *root)
154 int result = -ENOENT;
158 get_fs_root(task->fs, root);
165 static int proc_cwd_link(struct inode *inode, struct path *path)
167 struct task_struct *task = get_proc_task(inode);
168 int result = -ENOENT;
173 get_fs_pwd(task->fs, path);
177 put_task_struct(task);
182 static int proc_root_link(struct inode *inode, struct path *path)
184 struct task_struct *task = get_proc_task(inode);
185 int result = -ENOENT;
188 result = get_task_root(task, path);
189 put_task_struct(task);
195 * Return zero if current may access user memory in @task, -error if not.
197 static int check_mem_permission(struct task_struct *task)
200 * A task can always look at itself, in case it chooses
201 * to use system calls instead of load instructions.
207 * If current is actively ptrace'ing, and would also be
208 * permitted to freshly attach with ptrace now, permit it.
210 if (task_is_stopped_or_traced(task)) {
213 match = (tracehook_tracer_task(task) == current);
215 if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH))
220 * Noone else is allowed.
225 struct mm_struct *mm_for_maps(struct task_struct *task)
227 struct mm_struct *mm;
230 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
234 mm = get_task_mm(task);
235 if (mm && mm != current->mm &&
236 !ptrace_may_access(task, PTRACE_MODE_READ)) {
238 mm = ERR_PTR(-EACCES);
240 mutex_unlock(&task->signal->cred_guard_mutex);
245 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
249 struct mm_struct *mm = get_task_mm(task);
253 goto out_mm; /* Shh! No looking before we're done */
255 len = mm->arg_end - mm->arg_start;
260 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
262 // If the nul at the end of args has been overwritten, then
263 // assume application is using setproctitle(3).
264 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
265 len = strnlen(buffer, res);
269 len = mm->env_end - mm->env_start;
270 if (len > PAGE_SIZE - res)
271 len = PAGE_SIZE - res;
272 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
273 res = strnlen(buffer, res);
282 static int proc_pid_auxv(struct task_struct *task, char *buffer)
285 struct mm_struct *mm = get_task_mm(task);
287 unsigned int nwords = 0;
290 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
291 res = nwords * sizeof(mm->saved_auxv[0]);
294 memcpy(buffer, mm->saved_auxv, res);
301 #ifdef CONFIG_KALLSYMS
303 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
304 * Returns the resolved symbol. If that fails, simply return the address.
306 static int proc_pid_wchan(struct task_struct *task, char *buffer)
309 char symname[KSYM_NAME_LEN];
311 wchan = get_wchan(task);
313 if (lookup_symbol_name(wchan, symname) < 0)
314 if (!ptrace_may_access(task, PTRACE_MODE_READ))
317 return sprintf(buffer, "%lu", wchan);
319 return sprintf(buffer, "%s", symname);
321 #endif /* CONFIG_KALLSYMS */
323 #ifdef CONFIG_STACKTRACE
325 #define MAX_STACK_TRACE_DEPTH 64
327 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
328 struct pid *pid, struct task_struct *task)
330 struct stack_trace trace;
331 unsigned long *entries;
334 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
338 trace.nr_entries = 0;
339 trace.max_entries = MAX_STACK_TRACE_DEPTH;
340 trace.entries = entries;
342 save_stack_trace_tsk(task, &trace);
344 for (i = 0; i < trace.nr_entries; i++) {
345 seq_printf(m, "[<%pK>] %pS\n",
346 (void *)entries[i], (void *)entries[i]);
354 #ifdef CONFIG_SCHEDSTATS
356 * Provides /proc/PID/schedstat
358 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
360 return sprintf(buffer, "%llu %llu %lu\n",
361 (unsigned long long)task->se.sum_exec_runtime,
362 (unsigned long long)task->sched_info.run_delay,
363 task->sched_info.pcount);
367 #ifdef CONFIG_LATENCYTOP
368 static int lstats_show_proc(struct seq_file *m, void *v)
371 struct inode *inode = m->private;
372 struct task_struct *task = get_proc_task(inode);
376 seq_puts(m, "Latency Top version : v0.1\n");
377 for (i = 0; i < 32; i++) {
378 struct latency_record *lr = &task->latency_record[i];
379 if (lr->backtrace[0]) {
381 seq_printf(m, "%i %li %li",
382 lr->count, lr->time, lr->max);
383 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
384 unsigned long bt = lr->backtrace[q];
389 seq_printf(m, " %ps", (void *)bt);
395 put_task_struct(task);
399 static int lstats_open(struct inode *inode, struct file *file)
401 return single_open(file, lstats_show_proc, inode);
404 static ssize_t lstats_write(struct file *file, const char __user *buf,
405 size_t count, loff_t *offs)
407 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
411 clear_all_latency_tracing(task);
412 put_task_struct(task);
417 static const struct file_operations proc_lstats_operations = {
420 .write = lstats_write,
422 .release = single_release,
427 static int proc_oom_score(struct task_struct *task, char *buffer)
429 unsigned long points = 0;
431 read_lock(&tasklist_lock);
433 points = oom_badness(task, NULL, NULL,
434 totalram_pages + total_swap_pages);
435 read_unlock(&tasklist_lock);
436 return sprintf(buffer, "%lu\n", points);
444 static const struct limit_names lnames[RLIM_NLIMITS] = {
445 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
446 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
447 [RLIMIT_DATA] = {"Max data size", "bytes"},
448 [RLIMIT_STACK] = {"Max stack size", "bytes"},
449 [RLIMIT_CORE] = {"Max core file size", "bytes"},
450 [RLIMIT_RSS] = {"Max resident set", "bytes"},
451 [RLIMIT_NPROC] = {"Max processes", "processes"},
452 [RLIMIT_NOFILE] = {"Max open files", "files"},
453 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
454 [RLIMIT_AS] = {"Max address space", "bytes"},
455 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
456 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
457 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
458 [RLIMIT_NICE] = {"Max nice priority", NULL},
459 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
460 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
463 /* Display limits for a process */
464 static int proc_pid_limits(struct task_struct *task, char *buffer)
469 char *bufptr = buffer;
471 struct rlimit rlim[RLIM_NLIMITS];
473 if (!lock_task_sighand(task, &flags))
475 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
476 unlock_task_sighand(task, &flags);
479 * print the file header
481 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
482 "Limit", "Soft Limit", "Hard Limit", "Units");
484 for (i = 0; i < RLIM_NLIMITS; i++) {
485 if (rlim[i].rlim_cur == RLIM_INFINITY)
486 count += sprintf(&bufptr[count], "%-25s %-20s ",
487 lnames[i].name, "unlimited");
489 count += sprintf(&bufptr[count], "%-25s %-20lu ",
490 lnames[i].name, rlim[i].rlim_cur);
492 if (rlim[i].rlim_max == RLIM_INFINITY)
493 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
495 count += sprintf(&bufptr[count], "%-20lu ",
499 count += sprintf(&bufptr[count], "%-10s\n",
502 count += sprintf(&bufptr[count], "\n");
508 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
509 static int proc_pid_syscall(struct task_struct *task, char *buffer)
512 unsigned long args[6], sp, pc;
514 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
515 return sprintf(buffer, "running\n");
518 return sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
520 return sprintf(buffer,
521 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
523 args[0], args[1], args[2], args[3], args[4], args[5],
526 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
528 /************************************************************************/
529 /* Here the fs part begins */
530 /************************************************************************/
532 /* permission checks */
533 static int proc_fd_access_allowed(struct inode *inode)
535 struct task_struct *task;
537 /* Allow access to a task's file descriptors if it is us or we
538 * may use ptrace attach to the process and find out that
541 task = get_proc_task(inode);
543 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
544 put_task_struct(task);
549 static int proc_setattr(struct dentry *dentry, struct iattr *attr)
552 struct inode *inode = dentry->d_inode;
554 if (attr->ia_valid & ATTR_MODE)
557 error = inode_change_ok(inode, attr);
561 if ((attr->ia_valid & ATTR_SIZE) &&
562 attr->ia_size != i_size_read(inode)) {
563 error = vmtruncate(inode, attr->ia_size);
568 setattr_copy(inode, attr);
569 mark_inode_dirty(inode);
573 static const struct inode_operations proc_def_inode_operations = {
574 .setattr = proc_setattr,
577 static int mounts_open_common(struct inode *inode, struct file *file,
578 const struct seq_operations *op)
580 struct task_struct *task = get_proc_task(inode);
582 struct mnt_namespace *ns = NULL;
584 struct proc_mounts *p;
589 nsp = task_nsproxy(task);
596 if (ns && get_task_root(task, &root) == 0)
598 put_task_struct(task);
607 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
611 file->private_data = &p->m;
612 ret = seq_open(file, op);
619 p->event = ns->event;
633 static int mounts_release(struct inode *inode, struct file *file)
635 struct proc_mounts *p = file->private_data;
638 return seq_release(inode, file);
641 static unsigned mounts_poll(struct file *file, poll_table *wait)
643 struct proc_mounts *p = file->private_data;
644 unsigned res = POLLIN | POLLRDNORM;
646 poll_wait(file, &p->ns->poll, wait);
647 if (mnt_had_events(p))
648 res |= POLLERR | POLLPRI;
653 static int mounts_open(struct inode *inode, struct file *file)
655 return mounts_open_common(inode, file, &mounts_op);
658 static const struct file_operations proc_mounts_operations = {
662 .release = mounts_release,
666 static int mountinfo_open(struct inode *inode, struct file *file)
668 return mounts_open_common(inode, file, &mountinfo_op);
671 static const struct file_operations proc_mountinfo_operations = {
672 .open = mountinfo_open,
675 .release = mounts_release,
679 static int mountstats_open(struct inode *inode, struct file *file)
681 return mounts_open_common(inode, file, &mountstats_op);
684 static const struct file_operations proc_mountstats_operations = {
685 .open = mountstats_open,
688 .release = mounts_release,
691 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
693 static ssize_t proc_info_read(struct file * file, char __user * buf,
694 size_t count, loff_t *ppos)
696 struct inode * inode = file->f_path.dentry->d_inode;
699 struct task_struct *task = get_proc_task(inode);
705 if (count > PROC_BLOCK_SIZE)
706 count = PROC_BLOCK_SIZE;
709 if (!(page = __get_free_page(GFP_TEMPORARY)))
712 length = PROC_I(inode)->op.proc_read(task, (char*)page);
715 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
718 put_task_struct(task);
723 static const struct file_operations proc_info_file_operations = {
724 .read = proc_info_read,
725 .llseek = generic_file_llseek,
728 static int proc_single_show(struct seq_file *m, void *v)
730 struct inode *inode = m->private;
731 struct pid_namespace *ns;
733 struct task_struct *task;
736 ns = inode->i_sb->s_fs_info;
737 pid = proc_pid(inode);
738 task = get_pid_task(pid, PIDTYPE_PID);
742 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
744 put_task_struct(task);
748 static int proc_single_open(struct inode *inode, struct file *filp)
750 return single_open(filp, proc_single_show, inode);
753 static const struct file_operations proc_single_file_operations = {
754 .open = proc_single_open,
757 .release = single_release,
760 static int mem_open(struct inode* inode, struct file* file)
762 file->private_data = (void*)((long)current->self_exec_id);
763 /* OK to pass negative loff_t, we can catch out-of-range */
764 file->f_mode |= FMODE_UNSIGNED_OFFSET;
768 static ssize_t mem_read(struct file * file, char __user * buf,
769 size_t count, loff_t *ppos)
771 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
773 unsigned long src = *ppos;
775 struct mm_struct *mm;
780 if (check_mem_permission(task))
784 page = (char *)__get_free_page(GFP_TEMPORARY);
790 mm = get_task_mm(task);
796 if (file->private_data != (void*)((long)current->self_exec_id))
802 int this_len, retval;
804 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
805 retval = access_process_vm(task, src, page, this_len, 0);
806 if (!retval || check_mem_permission(task)) {
812 if (copy_to_user(buf, page, retval)) {
827 free_page((unsigned long) page);
829 put_task_struct(task);
834 #define mem_write NULL
837 /* This is a security hazard */
838 static ssize_t mem_write(struct file * file, const char __user *buf,
839 size_t count, loff_t *ppos)
843 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
844 unsigned long dst = *ppos;
850 if (check_mem_permission(task))
854 page = (char *)__get_free_page(GFP_TEMPORARY);
860 int this_len, retval;
862 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
863 if (copy_from_user(page, buf, this_len)) {
867 retval = access_process_vm(task, dst, page, this_len, 1);
879 free_page((unsigned long) page);
881 put_task_struct(task);
887 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
891 file->f_pos = offset;
894 file->f_pos += offset;
899 force_successful_syscall_return();
903 static const struct file_operations proc_mem_operations = {
910 static ssize_t environ_read(struct file *file, char __user *buf,
911 size_t count, loff_t *ppos)
913 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
915 unsigned long src = *ppos;
917 struct mm_struct *mm;
922 if (!ptrace_may_access(task, PTRACE_MODE_READ))
926 page = (char *)__get_free_page(GFP_TEMPORARY);
932 mm = get_task_mm(task);
937 int this_len, retval, max_len;
939 this_len = mm->env_end - (mm->env_start + src);
944 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
945 this_len = (this_len > max_len) ? max_len : this_len;
947 retval = access_process_vm(task, (mm->env_start + src),
955 if (copy_to_user(buf, page, retval)) {
969 free_page((unsigned long) page);
971 put_task_struct(task);
976 static const struct file_operations proc_environ_operations = {
977 .read = environ_read,
978 .llseek = generic_file_llseek,
981 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
982 size_t count, loff_t *ppos)
984 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
985 char buffer[PROC_NUMBUF];
987 int oom_adjust = OOM_DISABLE;
993 if (lock_task_sighand(task, &flags)) {
994 oom_adjust = task->signal->oom_adj;
995 unlock_task_sighand(task, &flags);
998 put_task_struct(task);
1000 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
1002 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1005 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1006 size_t count, loff_t *ppos)
1008 struct task_struct *task;
1009 char buffer[PROC_NUMBUF];
1011 unsigned long flags;
1014 memset(buffer, 0, sizeof(buffer));
1015 if (count > sizeof(buffer) - 1)
1016 count = sizeof(buffer) - 1;
1017 if (copy_from_user(buffer, buf, count)) {
1022 err = strict_strtol(strstrip(buffer), 0, &oom_adjust);
1025 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
1026 oom_adjust != OOM_DISABLE) {
1031 task = get_proc_task(file->f_path.dentry->d_inode);
1043 if (!lock_task_sighand(task, &flags)) {
1048 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
1053 if (oom_adjust != task->signal->oom_adj) {
1054 if (oom_adjust == OOM_DISABLE)
1055 atomic_inc(&task->mm->oom_disable_count);
1056 if (task->signal->oom_adj == OOM_DISABLE)
1057 atomic_dec(&task->mm->oom_disable_count);
1061 * Warn that /proc/pid/oom_adj is deprecated, see
1062 * Documentation/feature-removal-schedule.txt.
1064 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, "
1065 "please use /proc/%d/oom_score_adj instead.\n",
1066 current->comm, task_pid_nr(current),
1067 task_pid_nr(task), task_pid_nr(task));
1068 task->signal->oom_adj = oom_adjust;
1070 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1071 * value is always attainable.
1073 if (task->signal->oom_adj == OOM_ADJUST_MAX)
1074 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
1076 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
1079 unlock_task_sighand(task, &flags);
1082 put_task_struct(task);
1084 return err < 0 ? err : count;
1087 static const struct file_operations proc_oom_adjust_operations = {
1088 .read = oom_adjust_read,
1089 .write = oom_adjust_write,
1090 .llseek = generic_file_llseek,
1093 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1094 size_t count, loff_t *ppos)
1096 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1097 char buffer[PROC_NUMBUF];
1098 int oom_score_adj = OOM_SCORE_ADJ_MIN;
1099 unsigned long flags;
1104 if (lock_task_sighand(task, &flags)) {
1105 oom_score_adj = task->signal->oom_score_adj;
1106 unlock_task_sighand(task, &flags);
1108 put_task_struct(task);
1109 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1110 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1113 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1114 size_t count, loff_t *ppos)
1116 struct task_struct *task;
1117 char buffer[PROC_NUMBUF];
1118 unsigned long flags;
1122 memset(buffer, 0, sizeof(buffer));
1123 if (count > sizeof(buffer) - 1)
1124 count = sizeof(buffer) - 1;
1125 if (copy_from_user(buffer, buf, count)) {
1130 err = strict_strtol(strstrip(buffer), 0, &oom_score_adj);
1133 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1134 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1139 task = get_proc_task(file->f_path.dentry->d_inode);
1151 if (!lock_task_sighand(task, &flags)) {
1156 if (oom_score_adj < task->signal->oom_score_adj_min &&
1157 !capable(CAP_SYS_RESOURCE)) {
1162 if (oom_score_adj != task->signal->oom_score_adj) {
1163 if (oom_score_adj == OOM_SCORE_ADJ_MIN)
1164 atomic_inc(&task->mm->oom_disable_count);
1165 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1166 atomic_dec(&task->mm->oom_disable_count);
1168 task->signal->oom_score_adj = oom_score_adj;
1169 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1170 task->signal->oom_score_adj_min = oom_score_adj;
1172 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1173 * always attainable.
1175 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1176 task->signal->oom_adj = OOM_DISABLE;
1178 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1181 unlock_task_sighand(task, &flags);
1184 put_task_struct(task);
1186 return err < 0 ? err : count;
1189 static const struct file_operations proc_oom_score_adj_operations = {
1190 .read = oom_score_adj_read,
1191 .write = oom_score_adj_write,
1192 .llseek = default_llseek,
1195 #ifdef CONFIG_AUDITSYSCALL
1196 #define TMPBUFLEN 21
1197 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1198 size_t count, loff_t *ppos)
1200 struct inode * inode = file->f_path.dentry->d_inode;
1201 struct task_struct *task = get_proc_task(inode);
1203 char tmpbuf[TMPBUFLEN];
1207 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1208 audit_get_loginuid(task));
1209 put_task_struct(task);
1210 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1213 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1214 size_t count, loff_t *ppos)
1216 struct inode * inode = file->f_path.dentry->d_inode;
1221 if (!capable(CAP_AUDIT_CONTROL))
1225 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1231 if (count >= PAGE_SIZE)
1232 count = PAGE_SIZE - 1;
1235 /* No partial writes. */
1238 page = (char*)__get_free_page(GFP_TEMPORARY);
1242 if (copy_from_user(page, buf, count))
1246 loginuid = simple_strtoul(page, &tmp, 10);
1252 length = audit_set_loginuid(current, loginuid);
1253 if (likely(length == 0))
1257 free_page((unsigned long) page);
1261 static const struct file_operations proc_loginuid_operations = {
1262 .read = proc_loginuid_read,
1263 .write = proc_loginuid_write,
1264 .llseek = generic_file_llseek,
1267 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1268 size_t count, loff_t *ppos)
1270 struct inode * inode = file->f_path.dentry->d_inode;
1271 struct task_struct *task = get_proc_task(inode);
1273 char tmpbuf[TMPBUFLEN];
1277 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1278 audit_get_sessionid(task));
1279 put_task_struct(task);
1280 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1283 static const struct file_operations proc_sessionid_operations = {
1284 .read = proc_sessionid_read,
1285 .llseek = generic_file_llseek,
1289 #ifdef CONFIG_FAULT_INJECTION
1290 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1291 size_t count, loff_t *ppos)
1293 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1294 char buffer[PROC_NUMBUF];
1300 make_it_fail = task->make_it_fail;
1301 put_task_struct(task);
1303 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1305 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1308 static ssize_t proc_fault_inject_write(struct file * file,
1309 const char __user * buf, size_t count, loff_t *ppos)
1311 struct task_struct *task;
1312 char buffer[PROC_NUMBUF], *end;
1315 if (!capable(CAP_SYS_RESOURCE))
1317 memset(buffer, 0, sizeof(buffer));
1318 if (count > sizeof(buffer) - 1)
1319 count = sizeof(buffer) - 1;
1320 if (copy_from_user(buffer, buf, count))
1322 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1325 task = get_proc_task(file->f_dentry->d_inode);
1328 task->make_it_fail = make_it_fail;
1329 put_task_struct(task);
1334 static const struct file_operations proc_fault_inject_operations = {
1335 .read = proc_fault_inject_read,
1336 .write = proc_fault_inject_write,
1337 .llseek = generic_file_llseek,
1342 #ifdef CONFIG_SCHED_DEBUG
1344 * Print out various scheduling related per-task fields:
1346 static int sched_show(struct seq_file *m, void *v)
1348 struct inode *inode = m->private;
1349 struct task_struct *p;
1351 p = get_proc_task(inode);
1354 proc_sched_show_task(p, m);
1362 sched_write(struct file *file, const char __user *buf,
1363 size_t count, loff_t *offset)
1365 struct inode *inode = file->f_path.dentry->d_inode;
1366 struct task_struct *p;
1368 p = get_proc_task(inode);
1371 proc_sched_set_task(p);
1378 static int sched_open(struct inode *inode, struct file *filp)
1380 return single_open(filp, sched_show, inode);
1383 static const struct file_operations proc_pid_sched_operations = {
1386 .write = sched_write,
1387 .llseek = seq_lseek,
1388 .release = single_release,
1393 #ifdef CONFIG_SCHED_AUTOGROUP
1395 * Print out autogroup related information:
1397 static int sched_autogroup_show(struct seq_file *m, void *v)
1399 struct inode *inode = m->private;
1400 struct task_struct *p;
1402 p = get_proc_task(inode);
1405 proc_sched_autogroup_show_task(p, m);
1413 sched_autogroup_write(struct file *file, const char __user *buf,
1414 size_t count, loff_t *offset)
1416 struct inode *inode = file->f_path.dentry->d_inode;
1417 struct task_struct *p;
1418 char buffer[PROC_NUMBUF];
1422 memset(buffer, 0, sizeof(buffer));
1423 if (count > sizeof(buffer) - 1)
1424 count = sizeof(buffer) - 1;
1425 if (copy_from_user(buffer, buf, count))
1428 err = strict_strtol(strstrip(buffer), 0, &nice);
1432 p = get_proc_task(inode);
1437 err = proc_sched_autogroup_set_nice(p, &err);
1446 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1450 ret = single_open(filp, sched_autogroup_show, NULL);
1452 struct seq_file *m = filp->private_data;
1459 static const struct file_operations proc_pid_sched_autogroup_operations = {
1460 .open = sched_autogroup_open,
1462 .write = sched_autogroup_write,
1463 .llseek = seq_lseek,
1464 .release = single_release,
1467 #endif /* CONFIG_SCHED_AUTOGROUP */
1469 static ssize_t comm_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[TASK_COMM_LEN];
1476 memset(buffer, 0, sizeof(buffer));
1477 if (count > sizeof(buffer) - 1)
1478 count = sizeof(buffer) - 1;
1479 if (copy_from_user(buffer, buf, count))
1482 p = get_proc_task(inode);
1486 if (same_thread_group(current, p))
1487 set_task_comm(p, buffer);
1496 static int comm_show(struct seq_file *m, void *v)
1498 struct inode *inode = m->private;
1499 struct task_struct *p;
1501 p = get_proc_task(inode);
1506 seq_printf(m, "%s\n", p->comm);
1514 static int comm_open(struct inode *inode, struct file *filp)
1516 return single_open(filp, comm_show, inode);
1519 static const struct file_operations proc_pid_set_comm_operations = {
1522 .write = comm_write,
1523 .llseek = seq_lseek,
1524 .release = single_release,
1528 * We added or removed a vma mapping the executable. The vmas are only mapped
1529 * during exec and are not mapped with the mmap system call.
1530 * Callers must hold down_write() on the mm's mmap_sem for these
1532 void added_exe_file_vma(struct mm_struct *mm)
1534 mm->num_exe_file_vmas++;
1537 void removed_exe_file_vma(struct mm_struct *mm)
1539 mm->num_exe_file_vmas--;
1540 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
1542 mm->exe_file = NULL;
1547 void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1550 get_file(new_exe_file);
1553 mm->exe_file = new_exe_file;
1554 mm->num_exe_file_vmas = 0;
1557 struct file *get_mm_exe_file(struct mm_struct *mm)
1559 struct file *exe_file;
1561 /* We need mmap_sem to protect against races with removal of
1562 * VM_EXECUTABLE vmas */
1563 down_read(&mm->mmap_sem);
1564 exe_file = mm->exe_file;
1567 up_read(&mm->mmap_sem);
1571 void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
1573 /* It's safe to write the exe_file pointer without exe_file_lock because
1574 * this is called during fork when the task is not yet in /proc */
1575 newmm->exe_file = get_mm_exe_file(oldmm);
1578 static int proc_exe_link(struct inode *inode, struct path *exe_path)
1580 struct task_struct *task;
1581 struct mm_struct *mm;
1582 struct file *exe_file;
1584 task = get_proc_task(inode);
1587 mm = get_task_mm(task);
1588 put_task_struct(task);
1591 exe_file = get_mm_exe_file(mm);
1594 *exe_path = exe_file->f_path;
1595 path_get(&exe_file->f_path);
1602 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1604 struct inode *inode = dentry->d_inode;
1605 int error = -EACCES;
1607 /* We don't need a base pointer in the /proc filesystem */
1608 path_put(&nd->path);
1610 /* Are we allowed to snoop on the tasks file descriptors? */
1611 if (!proc_fd_access_allowed(inode))
1614 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1616 return ERR_PTR(error);
1619 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1621 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1628 pathname = d_path(path, tmp, PAGE_SIZE);
1629 len = PTR_ERR(pathname);
1630 if (IS_ERR(pathname))
1632 len = tmp + PAGE_SIZE - 1 - pathname;
1636 if (copy_to_user(buffer, pathname, len))
1639 free_page((unsigned long)tmp);
1643 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1645 int error = -EACCES;
1646 struct inode *inode = dentry->d_inode;
1649 /* Are we allowed to snoop on the tasks file descriptors? */
1650 if (!proc_fd_access_allowed(inode))
1653 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1657 error = do_proc_readlink(&path, buffer, buflen);
1663 static const struct inode_operations proc_pid_link_inode_operations = {
1664 .readlink = proc_pid_readlink,
1665 .follow_link = proc_pid_follow_link,
1666 .setattr = proc_setattr,
1670 /* building an inode */
1672 static int task_dumpable(struct task_struct *task)
1675 struct mm_struct *mm;
1680 dumpable = get_dumpable(mm);
1688 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1690 struct inode * inode;
1691 struct proc_inode *ei;
1692 const struct cred *cred;
1694 /* We need a new inode */
1696 inode = new_inode(sb);
1702 inode->i_ino = get_next_ino();
1703 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1704 inode->i_op = &proc_def_inode_operations;
1707 * grab the reference to task.
1709 ei->pid = get_task_pid(task, PIDTYPE_PID);
1713 if (task_dumpable(task)) {
1715 cred = __task_cred(task);
1716 inode->i_uid = cred->euid;
1717 inode->i_gid = cred->egid;
1720 security_task_to_inode(task, inode);
1730 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1732 struct inode *inode = dentry->d_inode;
1733 struct task_struct *task;
1734 const struct cred *cred;
1736 generic_fillattr(inode, stat);
1741 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1743 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1744 task_dumpable(task)) {
1745 cred = __task_cred(task);
1746 stat->uid = cred->euid;
1747 stat->gid = cred->egid;
1757 * Exceptional case: normally we are not allowed to unhash a busy
1758 * directory. In this case, however, we can do it - no aliasing problems
1759 * due to the way we treat inodes.
1761 * Rewrite the inode's ownerships here because the owning task may have
1762 * performed a setuid(), etc.
1764 * Before the /proc/pid/status file was created the only way to read
1765 * the effective uid of a /process was to stat /proc/pid. Reading
1766 * /proc/pid/status is slow enough that procps and other packages
1767 * kept stating /proc/pid. To keep the rules in /proc simple I have
1768 * made this apply to all per process world readable and executable
1771 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1773 struct inode *inode;
1774 struct task_struct *task;
1775 const struct cred *cred;
1777 if (nd && nd->flags & LOOKUP_RCU)
1780 inode = dentry->d_inode;
1781 task = get_proc_task(inode);
1784 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1785 task_dumpable(task)) {
1787 cred = __task_cred(task);
1788 inode->i_uid = cred->euid;
1789 inode->i_gid = cred->egid;
1795 inode->i_mode &= ~(S_ISUID | S_ISGID);
1796 security_task_to_inode(task, inode);
1797 put_task_struct(task);
1804 static int pid_delete_dentry(const struct dentry * dentry)
1806 /* Is the task we represent dead?
1807 * If so, then don't put the dentry on the lru list,
1808 * kill it immediately.
1810 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1813 static const struct dentry_operations pid_dentry_operations =
1815 .d_revalidate = pid_revalidate,
1816 .d_delete = pid_delete_dentry,
1821 typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1822 struct task_struct *, const void *);
1825 * Fill a directory entry.
1827 * If possible create the dcache entry and derive our inode number and
1828 * file type from dcache entry.
1830 * Since all of the proc inode numbers are dynamically generated, the inode
1831 * numbers do not exist until the inode is cache. This means creating the
1832 * the dcache entry in readdir is necessary to keep the inode numbers
1833 * reported by readdir in sync with the inode numbers reported
1836 static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1837 char *name, int len,
1838 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1840 struct dentry *child, *dir = filp->f_path.dentry;
1841 struct inode *inode;
1844 unsigned type = DT_UNKNOWN;
1848 qname.hash = full_name_hash(name, len);
1850 child = d_lookup(dir, &qname);
1853 new = d_alloc(dir, &qname);
1855 child = instantiate(dir->d_inode, new, task, ptr);
1862 if (!child || IS_ERR(child) || !child->d_inode)
1863 goto end_instantiate;
1864 inode = child->d_inode;
1867 type = inode->i_mode >> 12;
1872 ino = find_inode_number(dir, &qname);
1875 return filldir(dirent, name, len, filp->f_pos, ino, type);
1878 static unsigned name_to_int(struct dentry *dentry)
1880 const char *name = dentry->d_name.name;
1881 int len = dentry->d_name.len;
1884 if (len > 1 && *name == '0')
1887 unsigned c = *name++ - '0';
1890 if (n >= (~0U-9)/10)
1900 #define PROC_FDINFO_MAX 64
1902 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1904 struct task_struct *task = get_proc_task(inode);
1905 struct files_struct *files = NULL;
1907 int fd = proc_fd(inode);
1910 files = get_files_struct(task);
1911 put_task_struct(task);
1915 * We are not taking a ref to the file structure, so we must
1918 spin_lock(&files->file_lock);
1919 file = fcheck_files(files, fd);
1922 *path = file->f_path;
1923 path_get(&file->f_path);
1926 snprintf(info, PROC_FDINFO_MAX,
1929 (long long) file->f_pos,
1931 spin_unlock(&files->file_lock);
1932 put_files_struct(files);
1935 spin_unlock(&files->file_lock);
1936 put_files_struct(files);
1941 static int proc_fd_link(struct inode *inode, struct path *path)
1943 return proc_fd_info(inode, path, NULL);
1946 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1948 struct inode *inode;
1949 struct task_struct *task;
1951 struct files_struct *files;
1952 const struct cred *cred;
1954 if (nd && nd->flags & LOOKUP_RCU)
1957 inode = dentry->d_inode;
1958 task = get_proc_task(inode);
1959 fd = proc_fd(inode);
1962 files = get_files_struct(task);
1965 if (fcheck_files(files, fd)) {
1967 put_files_struct(files);
1968 if (task_dumpable(task)) {
1970 cred = __task_cred(task);
1971 inode->i_uid = cred->euid;
1972 inode->i_gid = cred->egid;
1978 inode->i_mode &= ~(S_ISUID | S_ISGID);
1979 security_task_to_inode(task, inode);
1980 put_task_struct(task);
1984 put_files_struct(files);
1986 put_task_struct(task);
1992 static const struct dentry_operations tid_fd_dentry_operations =
1994 .d_revalidate = tid_fd_revalidate,
1995 .d_delete = pid_delete_dentry,
1998 static struct dentry *proc_fd_instantiate(struct inode *dir,
1999 struct dentry *dentry, struct task_struct *task, const void *ptr)
2001 unsigned fd = *(const unsigned *)ptr;
2003 struct files_struct *files;
2004 struct inode *inode;
2005 struct proc_inode *ei;
2006 struct dentry *error = ERR_PTR(-ENOENT);
2008 inode = proc_pid_make_inode(dir->i_sb, task);
2013 files = get_files_struct(task);
2016 inode->i_mode = S_IFLNK;
2019 * We are not taking a ref to the file structure, so we must
2022 spin_lock(&files->file_lock);
2023 file = fcheck_files(files, fd);
2026 if (file->f_mode & FMODE_READ)
2027 inode->i_mode |= S_IRUSR | S_IXUSR;
2028 if (file->f_mode & FMODE_WRITE)
2029 inode->i_mode |= S_IWUSR | S_IXUSR;
2030 spin_unlock(&files->file_lock);
2031 put_files_struct(files);
2033 inode->i_op = &proc_pid_link_inode_operations;
2035 ei->op.proc_get_link = proc_fd_link;
2036 d_set_d_op(dentry, &tid_fd_dentry_operations);
2037 d_add(dentry, inode);
2038 /* Close the race of the process dying before we return the dentry */
2039 if (tid_fd_revalidate(dentry, NULL))
2045 spin_unlock(&files->file_lock);
2046 put_files_struct(files);
2052 static struct dentry *proc_lookupfd_common(struct inode *dir,
2053 struct dentry *dentry,
2054 instantiate_t instantiate)
2056 struct task_struct *task = get_proc_task(dir);
2057 unsigned fd = name_to_int(dentry);
2058 struct dentry *result = ERR_PTR(-ENOENT);
2065 result = instantiate(dir, dentry, task, &fd);
2067 put_task_struct(task);
2072 static int proc_readfd_common(struct file * filp, void * dirent,
2073 filldir_t filldir, instantiate_t instantiate)
2075 struct dentry *dentry = filp->f_path.dentry;
2076 struct inode *inode = dentry->d_inode;
2077 struct task_struct *p = get_proc_task(inode);
2078 unsigned int fd, ino;
2080 struct files_struct * files;
2090 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
2094 ino = parent_ino(dentry);
2095 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2099 files = get_files_struct(p);
2103 for (fd = filp->f_pos-2;
2104 fd < files_fdtable(files)->max_fds;
2105 fd++, filp->f_pos++) {
2106 char name[PROC_NUMBUF];
2109 if (!fcheck_files(files, fd))
2113 len = snprintf(name, sizeof(name), "%d", fd);
2114 if (proc_fill_cache(filp, dirent, filldir,
2115 name, len, instantiate,
2123 put_files_struct(files);
2131 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
2132 struct nameidata *nd)
2134 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
2137 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
2139 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
2142 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
2143 size_t len, loff_t *ppos)
2145 char tmp[PROC_FDINFO_MAX];
2146 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
2148 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
2152 static const struct file_operations proc_fdinfo_file_operations = {
2153 .open = nonseekable_open,
2154 .read = proc_fdinfo_read,
2155 .llseek = no_llseek,
2158 static const struct file_operations proc_fd_operations = {
2159 .read = generic_read_dir,
2160 .readdir = proc_readfd,
2161 .llseek = default_llseek,
2165 * /proc/pid/fd needs a special permission handler so that a process can still
2166 * access /proc/self/fd after it has executed a setuid().
2168 static int proc_fd_permission(struct inode *inode, int mask, unsigned int flags)
2172 if (flags & IPERM_FLAG_RCU)
2174 rv = generic_permission(inode, mask, flags, NULL);
2177 if (task_pid(current) == proc_pid(inode))
2183 * proc directories can do almost nothing..
2185 static const struct inode_operations proc_fd_inode_operations = {
2186 .lookup = proc_lookupfd,
2187 .permission = proc_fd_permission,
2188 .setattr = proc_setattr,
2191 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2192 struct dentry *dentry, struct task_struct *task, const void *ptr)
2194 unsigned fd = *(unsigned *)ptr;
2195 struct inode *inode;
2196 struct proc_inode *ei;
2197 struct dentry *error = ERR_PTR(-ENOENT);
2199 inode = proc_pid_make_inode(dir->i_sb, task);
2204 inode->i_mode = S_IFREG | S_IRUSR;
2205 inode->i_fop = &proc_fdinfo_file_operations;
2206 d_set_d_op(dentry, &tid_fd_dentry_operations);
2207 d_add(dentry, inode);
2208 /* Close the race of the process dying before we return the dentry */
2209 if (tid_fd_revalidate(dentry, NULL))
2216 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2217 struct dentry *dentry,
2218 struct nameidata *nd)
2220 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2223 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2225 return proc_readfd_common(filp, dirent, filldir,
2226 proc_fdinfo_instantiate);
2229 static const struct file_operations proc_fdinfo_operations = {
2230 .read = generic_read_dir,
2231 .readdir = proc_readfdinfo,
2232 .llseek = default_llseek,
2236 * proc directories can do almost nothing..
2238 static const struct inode_operations proc_fdinfo_inode_operations = {
2239 .lookup = proc_lookupfdinfo,
2240 .setattr = proc_setattr,
2244 static struct dentry *proc_pident_instantiate(struct inode *dir,
2245 struct dentry *dentry, struct task_struct *task, const void *ptr)
2247 const struct pid_entry *p = ptr;
2248 struct inode *inode;
2249 struct proc_inode *ei;
2250 struct dentry *error = ERR_PTR(-ENOENT);
2252 inode = proc_pid_make_inode(dir->i_sb, task);
2257 inode->i_mode = p->mode;
2258 if (S_ISDIR(inode->i_mode))
2259 inode->i_nlink = 2; /* Use getattr to fix if necessary */
2261 inode->i_op = p->iop;
2263 inode->i_fop = p->fop;
2265 d_set_d_op(dentry, &pid_dentry_operations);
2266 d_add(dentry, inode);
2267 /* Close the race of the process dying before we return the dentry */
2268 if (pid_revalidate(dentry, NULL))
2274 static struct dentry *proc_pident_lookup(struct inode *dir,
2275 struct dentry *dentry,
2276 const struct pid_entry *ents,
2279 struct dentry *error;
2280 struct task_struct *task = get_proc_task(dir);
2281 const struct pid_entry *p, *last;
2283 error = ERR_PTR(-ENOENT);
2289 * Yes, it does not scale. And it should not. Don't add
2290 * new entries into /proc/<tgid>/ without very good reasons.
2292 last = &ents[nents - 1];
2293 for (p = ents; p <= last; p++) {
2294 if (p->len != dentry->d_name.len)
2296 if (!memcmp(dentry->d_name.name, p->name, p->len))
2302 error = proc_pident_instantiate(dir, dentry, task, p);
2304 put_task_struct(task);
2309 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2310 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2312 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2313 proc_pident_instantiate, task, p);
2316 static int proc_pident_readdir(struct file *filp,
2317 void *dirent, filldir_t filldir,
2318 const struct pid_entry *ents, unsigned int nents)
2321 struct dentry *dentry = filp->f_path.dentry;
2322 struct inode *inode = dentry->d_inode;
2323 struct task_struct *task = get_proc_task(inode);
2324 const struct pid_entry *p, *last;
2337 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2343 ino = parent_ino(dentry);
2344 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2356 last = &ents[nents - 1];
2358 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2367 put_task_struct(task);
2372 #ifdef CONFIG_SECURITY
2373 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2374 size_t count, loff_t *ppos)
2376 struct inode * inode = file->f_path.dentry->d_inode;
2379 struct task_struct *task = get_proc_task(inode);
2384 length = security_getprocattr(task,
2385 (char*)file->f_path.dentry->d_name.name,
2387 put_task_struct(task);
2389 length = simple_read_from_buffer(buf, count, ppos, p, length);
2394 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2395 size_t count, loff_t *ppos)
2397 struct inode * inode = file->f_path.dentry->d_inode;
2400 struct task_struct *task = get_proc_task(inode);
2405 if (count > PAGE_SIZE)
2408 /* No partial writes. */
2414 page = (char*)__get_free_page(GFP_TEMPORARY);
2419 if (copy_from_user(page, buf, count))
2422 /* Guard against adverse ptrace interaction */
2423 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2427 length = security_setprocattr(task,
2428 (char*)file->f_path.dentry->d_name.name,
2429 (void*)page, count);
2430 mutex_unlock(&task->signal->cred_guard_mutex);
2432 free_page((unsigned long) page);
2434 put_task_struct(task);
2439 static const struct file_operations proc_pid_attr_operations = {
2440 .read = proc_pid_attr_read,
2441 .write = proc_pid_attr_write,
2442 .llseek = generic_file_llseek,
2445 static const struct pid_entry attr_dir_stuff[] = {
2446 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2447 REG("prev", S_IRUGO, proc_pid_attr_operations),
2448 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2449 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2450 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2451 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2454 static int proc_attr_dir_readdir(struct file * filp,
2455 void * dirent, filldir_t filldir)
2457 return proc_pident_readdir(filp,dirent,filldir,
2458 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2461 static const struct file_operations proc_attr_dir_operations = {
2462 .read = generic_read_dir,
2463 .readdir = proc_attr_dir_readdir,
2464 .llseek = default_llseek,
2467 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2468 struct dentry *dentry, struct nameidata *nd)
2470 return proc_pident_lookup(dir, dentry,
2471 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2474 static const struct inode_operations proc_attr_dir_inode_operations = {
2475 .lookup = proc_attr_dir_lookup,
2476 .getattr = pid_getattr,
2477 .setattr = proc_setattr,
2482 #ifdef CONFIG_ELF_CORE
2483 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2484 size_t count, loff_t *ppos)
2486 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2487 struct mm_struct *mm;
2488 char buffer[PROC_NUMBUF];
2496 mm = get_task_mm(task);
2498 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2499 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2500 MMF_DUMP_FILTER_SHIFT));
2502 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2505 put_task_struct(task);
2510 static ssize_t proc_coredump_filter_write(struct file *file,
2511 const char __user *buf,
2515 struct task_struct *task;
2516 struct mm_struct *mm;
2517 char buffer[PROC_NUMBUF], *end;
2524 memset(buffer, 0, sizeof(buffer));
2525 if (count > sizeof(buffer) - 1)
2526 count = sizeof(buffer) - 1;
2527 if (copy_from_user(buffer, buf, count))
2531 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2534 if (end - buffer == 0)
2538 task = get_proc_task(file->f_dentry->d_inode);
2543 mm = get_task_mm(task);
2547 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2549 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2551 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2556 put_task_struct(task);
2561 static const struct file_operations proc_coredump_filter_operations = {
2562 .read = proc_coredump_filter_read,
2563 .write = proc_coredump_filter_write,
2564 .llseek = generic_file_llseek,
2571 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2574 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2575 pid_t tgid = task_tgid_nr_ns(current, ns);
2576 char tmp[PROC_NUMBUF];
2579 sprintf(tmp, "%d", tgid);
2580 return vfs_readlink(dentry,buffer,buflen,tmp);
2583 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2585 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2586 pid_t tgid = task_tgid_nr_ns(current, ns);
2587 char *name = ERR_PTR(-ENOENT);
2591 name = ERR_PTR(-ENOMEM);
2593 sprintf(name, "%d", tgid);
2595 nd_set_link(nd, name);
2599 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2602 char *s = nd_get_link(nd);
2607 static const struct inode_operations proc_self_inode_operations = {
2608 .readlink = proc_self_readlink,
2609 .follow_link = proc_self_follow_link,
2610 .put_link = proc_self_put_link,
2616 * These are the directory entries in the root directory of /proc
2617 * that properly belong to the /proc filesystem, as they describe
2618 * describe something that is process related.
2620 static const struct pid_entry proc_base_stuff[] = {
2621 NOD("self", S_IFLNK|S_IRWXUGO,
2622 &proc_self_inode_operations, NULL, {}),
2625 static struct dentry *proc_base_instantiate(struct inode *dir,
2626 struct dentry *dentry, struct task_struct *task, const void *ptr)
2628 const struct pid_entry *p = ptr;
2629 struct inode *inode;
2630 struct proc_inode *ei;
2631 struct dentry *error;
2633 /* Allocate the inode */
2634 error = ERR_PTR(-ENOMEM);
2635 inode = new_inode(dir->i_sb);
2639 /* Initialize the inode */
2641 inode->i_ino = get_next_ino();
2642 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2645 * grab the reference to the task.
2647 ei->pid = get_task_pid(task, PIDTYPE_PID);
2651 inode->i_mode = p->mode;
2652 if (S_ISDIR(inode->i_mode))
2654 if (S_ISLNK(inode->i_mode))
2657 inode->i_op = p->iop;
2659 inode->i_fop = p->fop;
2661 d_add(dentry, inode);
2670 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2672 struct dentry *error;
2673 struct task_struct *task = get_proc_task(dir);
2674 const struct pid_entry *p, *last;
2676 error = ERR_PTR(-ENOENT);
2681 /* Lookup the directory entry */
2682 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2683 for (p = proc_base_stuff; p <= last; p++) {
2684 if (p->len != dentry->d_name.len)
2686 if (!memcmp(dentry->d_name.name, p->name, p->len))
2692 error = proc_base_instantiate(dir, dentry, task, p);
2695 put_task_struct(task);
2700 static int proc_base_fill_cache(struct file *filp, void *dirent,
2701 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2703 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2704 proc_base_instantiate, task, p);
2707 #ifdef CONFIG_TASK_IO_ACCOUNTING
2708 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2710 struct task_io_accounting acct = task->ioac;
2711 unsigned long flags;
2713 if (whole && lock_task_sighand(task, &flags)) {
2714 struct task_struct *t = task;
2716 task_io_accounting_add(&acct, &task->signal->ioac);
2717 while_each_thread(task, t)
2718 task_io_accounting_add(&acct, &t->ioac);
2720 unlock_task_sighand(task, &flags);
2722 return sprintf(buffer,
2727 "read_bytes: %llu\n"
2728 "write_bytes: %llu\n"
2729 "cancelled_write_bytes: %llu\n",
2730 (unsigned long long)acct.rchar,
2731 (unsigned long long)acct.wchar,
2732 (unsigned long long)acct.syscr,
2733 (unsigned long long)acct.syscw,
2734 (unsigned long long)acct.read_bytes,
2735 (unsigned long long)acct.write_bytes,
2736 (unsigned long long)acct.cancelled_write_bytes);
2739 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2741 return do_io_accounting(task, buffer, 0);
2744 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2746 return do_io_accounting(task, buffer, 1);
2748 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2750 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2751 struct pid *pid, struct task_struct *task)
2753 seq_printf(m, "%08x\n", task->personality);
2760 static const struct file_operations proc_task_operations;
2761 static const struct inode_operations proc_task_inode_operations;
2763 static const struct pid_entry tgid_base_stuff[] = {
2764 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2765 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2766 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2768 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2770 REG("environ", S_IRUSR, proc_environ_operations),
2771 INF("auxv", S_IRUSR, proc_pid_auxv),
2772 ONE("status", S_IRUGO, proc_pid_status),
2773 ONE("personality", S_IRUSR, proc_pid_personality),
2774 INF("limits", S_IRUGO, proc_pid_limits),
2775 #ifdef CONFIG_SCHED_DEBUG
2776 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2778 #ifdef CONFIG_SCHED_AUTOGROUP
2779 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2781 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2782 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2783 INF("syscall", S_IRUSR, proc_pid_syscall),
2785 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2786 ONE("stat", S_IRUGO, proc_tgid_stat),
2787 ONE("statm", S_IRUGO, proc_pid_statm),
2788 REG("maps", S_IRUGO, proc_maps_operations),
2790 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2792 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2793 LNK("cwd", proc_cwd_link),
2794 LNK("root", proc_root_link),
2795 LNK("exe", proc_exe_link),
2796 REG("mounts", S_IRUGO, proc_mounts_operations),
2797 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2798 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2799 #ifdef CONFIG_PROC_PAGE_MONITOR
2800 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2801 REG("smaps", S_IRUGO, proc_smaps_operations),
2802 REG("pagemap", S_IRUGO, proc_pagemap_operations),
2804 #ifdef CONFIG_SECURITY
2805 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2807 #ifdef CONFIG_KALLSYMS
2808 INF("wchan", S_IRUGO, proc_pid_wchan),
2810 #ifdef CONFIG_STACKTRACE
2811 ONE("stack", S_IRUSR, proc_pid_stack),
2813 #ifdef CONFIG_SCHEDSTATS
2814 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2816 #ifdef CONFIG_LATENCYTOP
2817 REG("latency", S_IRUGO, proc_lstats_operations),
2819 #ifdef CONFIG_PROC_PID_CPUSET
2820 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2822 #ifdef CONFIG_CGROUPS
2823 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2825 INF("oom_score", S_IRUGO, proc_oom_score),
2826 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2827 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2828 #ifdef CONFIG_AUDITSYSCALL
2829 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2830 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2832 #ifdef CONFIG_FAULT_INJECTION
2833 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2835 #ifdef CONFIG_ELF_CORE
2836 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2838 #ifdef CONFIG_TASK_IO_ACCOUNTING
2839 INF("io", S_IRUGO, proc_tgid_io_accounting),
2843 static int proc_tgid_base_readdir(struct file * filp,
2844 void * dirent, filldir_t filldir)
2846 return proc_pident_readdir(filp,dirent,filldir,
2847 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2850 static const struct file_operations proc_tgid_base_operations = {
2851 .read = generic_read_dir,
2852 .readdir = proc_tgid_base_readdir,
2853 .llseek = default_llseek,
2856 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2857 return proc_pident_lookup(dir, dentry,
2858 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2861 static const struct inode_operations proc_tgid_base_inode_operations = {
2862 .lookup = proc_tgid_base_lookup,
2863 .getattr = pid_getattr,
2864 .setattr = proc_setattr,
2867 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2869 struct dentry *dentry, *leader, *dir;
2870 char buf[PROC_NUMBUF];
2874 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2875 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2877 shrink_dcache_parent(dentry);
2883 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2884 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2889 name.len = strlen(name.name);
2890 dir = d_hash_and_lookup(leader, &name);
2892 goto out_put_leader;
2895 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2896 dentry = d_hash_and_lookup(dir, &name);
2898 shrink_dcache_parent(dentry);
2911 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2912 * @task: task that should be flushed.
2914 * When flushing dentries from proc, one needs to flush them from global
2915 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2916 * in. This call is supposed to do all of this job.
2918 * Looks in the dcache for
2920 * /proc/@tgid/task/@pid
2921 * if either directory is present flushes it and all of it'ts children
2924 * It is safe and reasonable to cache /proc entries for a task until
2925 * that task exits. After that they just clog up the dcache with
2926 * useless entries, possibly causing useful dcache entries to be
2927 * flushed instead. This routine is proved to flush those useless
2928 * dcache entries at process exit time.
2930 * NOTE: This routine is just an optimization so it does not guarantee
2931 * that no dcache entries will exist at process exit time it
2932 * just makes it very unlikely that any will persist.
2935 void proc_flush_task(struct task_struct *task)
2938 struct pid *pid, *tgid;
2941 pid = task_pid(task);
2942 tgid = task_tgid(task);
2944 for (i = 0; i <= pid->level; i++) {
2945 upid = &pid->numbers[i];
2946 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2947 tgid->numbers[i].nr);
2950 upid = &pid->numbers[pid->level];
2952 pid_ns_release_proc(upid->ns);
2955 static struct dentry *proc_pid_instantiate(struct inode *dir,
2956 struct dentry * dentry,
2957 struct task_struct *task, const void *ptr)
2959 struct dentry *error = ERR_PTR(-ENOENT);
2960 struct inode *inode;
2962 inode = proc_pid_make_inode(dir->i_sb, task);
2966 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2967 inode->i_op = &proc_tgid_base_inode_operations;
2968 inode->i_fop = &proc_tgid_base_operations;
2969 inode->i_flags|=S_IMMUTABLE;
2971 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
2972 ARRAY_SIZE(tgid_base_stuff));
2974 d_set_d_op(dentry, &pid_dentry_operations);
2976 d_add(dentry, inode);
2977 /* Close the race of the process dying before we return the dentry */
2978 if (pid_revalidate(dentry, NULL))
2984 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2986 struct dentry *result;
2987 struct task_struct *task;
2989 struct pid_namespace *ns;
2991 result = proc_base_lookup(dir, dentry);
2992 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2995 tgid = name_to_int(dentry);
2999 ns = dentry->d_sb->s_fs_info;
3001 task = find_task_by_pid_ns(tgid, ns);
3003 get_task_struct(task);
3008 result = proc_pid_instantiate(dir, dentry, task, NULL);
3009 put_task_struct(task);
3015 * Find the first task with tgid >= tgid
3020 struct task_struct *task;
3022 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3027 put_task_struct(iter.task);
3031 pid = find_ge_pid(iter.tgid, ns);
3033 iter.tgid = pid_nr_ns(pid, ns);
3034 iter.task = pid_task(pid, PIDTYPE_PID);
3035 /* What we to know is if the pid we have find is the
3036 * pid of a thread_group_leader. Testing for task
3037 * being a thread_group_leader is the obvious thing
3038 * todo but there is a window when it fails, due to
3039 * the pid transfer logic in de_thread.
3041 * So we perform the straight forward test of seeing
3042 * if the pid we have found is the pid of a thread
3043 * group leader, and don't worry if the task we have
3044 * found doesn't happen to be a thread group leader.
3045 * As we don't care in the case of readdir.
3047 if (!iter.task || !has_group_leader_pid(iter.task)) {
3051 get_task_struct(iter.task);
3057 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
3059 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3060 struct tgid_iter iter)
3062 char name[PROC_NUMBUF];
3063 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
3064 return proc_fill_cache(filp, dirent, filldir, name, len,
3065 proc_pid_instantiate, iter.task, NULL);
3068 /* for the /proc/ directory itself, after non-process stuff has been done */
3069 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3072 struct task_struct *reaper;
3073 struct tgid_iter iter;
3074 struct pid_namespace *ns;
3076 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3078 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3080 reaper = get_proc_task(filp->f_path.dentry->d_inode);
3084 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
3085 const struct pid_entry *p = &proc_base_stuff[nr];
3086 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
3090 ns = filp->f_dentry->d_sb->s_fs_info;
3092 iter.tgid = filp->f_pos - TGID_OFFSET;
3093 for (iter = next_tgid(ns, iter);
3095 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3096 filp->f_pos = iter.tgid + TGID_OFFSET;
3097 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
3098 put_task_struct(iter.task);
3102 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3104 put_task_struct(reaper);
3112 static const struct pid_entry tid_base_stuff[] = {
3113 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3114 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3115 REG("environ", S_IRUSR, proc_environ_operations),
3116 INF("auxv", S_IRUSR, proc_pid_auxv),
3117 ONE("status", S_IRUGO, proc_pid_status),
3118 ONE("personality", S_IRUSR, proc_pid_personality),
3119 INF("limits", S_IRUGO, proc_pid_limits),
3120 #ifdef CONFIG_SCHED_DEBUG
3121 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3123 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3124 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3125 INF("syscall", S_IRUSR, proc_pid_syscall),
3127 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3128 ONE("stat", S_IRUGO, proc_tid_stat),
3129 ONE("statm", S_IRUGO, proc_pid_statm),
3130 REG("maps", S_IRUGO, proc_maps_operations),
3132 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
3134 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3135 LNK("cwd", proc_cwd_link),
3136 LNK("root", proc_root_link),
3137 LNK("exe", proc_exe_link),
3138 REG("mounts", S_IRUGO, proc_mounts_operations),
3139 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3140 #ifdef CONFIG_PROC_PAGE_MONITOR
3141 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3142 REG("smaps", S_IRUGO, proc_smaps_operations),
3143 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3145 #ifdef CONFIG_SECURITY
3146 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3148 #ifdef CONFIG_KALLSYMS
3149 INF("wchan", S_IRUGO, proc_pid_wchan),
3151 #ifdef CONFIG_STACKTRACE
3152 ONE("stack", S_IRUSR, proc_pid_stack),
3154 #ifdef CONFIG_SCHEDSTATS
3155 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3157 #ifdef CONFIG_LATENCYTOP
3158 REG("latency", S_IRUGO, proc_lstats_operations),
3160 #ifdef CONFIG_PROC_PID_CPUSET
3161 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3163 #ifdef CONFIG_CGROUPS
3164 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3166 INF("oom_score", S_IRUGO, proc_oom_score),
3167 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3168 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3169 #ifdef CONFIG_AUDITSYSCALL
3170 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3171 REG("sessionid", S_IRUSR, proc_sessionid_operations),
3173 #ifdef CONFIG_FAULT_INJECTION
3174 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3176 #ifdef CONFIG_TASK_IO_ACCOUNTING
3177 INF("io", S_IRUGO, proc_tid_io_accounting),
3181 static int proc_tid_base_readdir(struct file * filp,
3182 void * dirent, filldir_t filldir)
3184 return proc_pident_readdir(filp,dirent,filldir,
3185 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3188 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3189 return proc_pident_lookup(dir, dentry,
3190 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3193 static const struct file_operations proc_tid_base_operations = {
3194 .read = generic_read_dir,
3195 .readdir = proc_tid_base_readdir,
3196 .llseek = default_llseek,
3199 static const struct inode_operations proc_tid_base_inode_operations = {
3200 .lookup = proc_tid_base_lookup,
3201 .getattr = pid_getattr,
3202 .setattr = proc_setattr,
3205 static struct dentry *proc_task_instantiate(struct inode *dir,
3206 struct dentry *dentry, struct task_struct *task, const void *ptr)
3208 struct dentry *error = ERR_PTR(-ENOENT);
3209 struct inode *inode;
3210 inode = proc_pid_make_inode(dir->i_sb, task);
3214 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3215 inode->i_op = &proc_tid_base_inode_operations;
3216 inode->i_fop = &proc_tid_base_operations;
3217 inode->i_flags|=S_IMMUTABLE;
3219 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
3220 ARRAY_SIZE(tid_base_stuff));
3222 d_set_d_op(dentry, &pid_dentry_operations);
3224 d_add(dentry, inode);
3225 /* Close the race of the process dying before we return the dentry */
3226 if (pid_revalidate(dentry, NULL))
3232 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3234 struct dentry *result = ERR_PTR(-ENOENT);
3235 struct task_struct *task;
3236 struct task_struct *leader = get_proc_task(dir);
3238 struct pid_namespace *ns;
3243 tid = name_to_int(dentry);
3247 ns = dentry->d_sb->s_fs_info;
3249 task = find_task_by_pid_ns(tid, ns);
3251 get_task_struct(task);
3255 if (!same_thread_group(leader, task))
3258 result = proc_task_instantiate(dir, dentry, task, NULL);
3260 put_task_struct(task);
3262 put_task_struct(leader);
3268 * Find the first tid of a thread group to return to user space.
3270 * Usually this is just the thread group leader, but if the users
3271 * buffer was too small or there was a seek into the middle of the
3272 * directory we have more work todo.
3274 * In the case of a short read we start with find_task_by_pid.
3276 * In the case of a seek we start with the leader and walk nr
3279 static struct task_struct *first_tid(struct task_struct *leader,
3280 int tid, int nr, struct pid_namespace *ns)
3282 struct task_struct *pos;
3285 /* Attempt to start with the pid of a thread */
3286 if (tid && (nr > 0)) {
3287 pos = find_task_by_pid_ns(tid, ns);
3288 if (pos && (pos->group_leader == leader))
3292 /* If nr exceeds the number of threads there is nothing todo */
3294 if (nr && nr >= get_nr_threads(leader))
3297 /* If we haven't found our starting place yet start
3298 * with the leader and walk nr threads forward.
3300 for (pos = leader; nr > 0; --nr) {
3301 pos = next_thread(pos);
3302 if (pos == leader) {
3308 get_task_struct(pos);
3315 * Find the next thread in the thread list.
3316 * Return NULL if there is an error or no next thread.
3318 * The reference to the input task_struct is released.
3320 static struct task_struct *next_tid(struct task_struct *start)
3322 struct task_struct *pos = NULL;
3324 if (pid_alive(start)) {
3325 pos = next_thread(start);
3326 if (thread_group_leader(pos))
3329 get_task_struct(pos);
3332 put_task_struct(start);
3336 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3337 struct task_struct *task, int tid)
3339 char name[PROC_NUMBUF];
3340 int len = snprintf(name, sizeof(name), "%d", tid);
3341 return proc_fill_cache(filp, dirent, filldir, name, len,
3342 proc_task_instantiate, task, NULL);
3345 /* for the /proc/TGID/task/ directories */
3346 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3348 struct dentry *dentry = filp->f_path.dentry;
3349 struct inode *inode = dentry->d_inode;
3350 struct task_struct *leader = NULL;
3351 struct task_struct *task;
3352 int retval = -ENOENT;
3355 struct pid_namespace *ns;
3357 task = get_proc_task(inode);
3361 if (pid_alive(task)) {
3362 leader = task->group_leader;
3363 get_task_struct(leader);
3366 put_task_struct(task);
3371 switch ((unsigned long)filp->f_pos) {
3374 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3379 ino = parent_ino(dentry);
3380 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3386 /* f_version caches the tgid value that the last readdir call couldn't
3387 * return. lseek aka telldir automagically resets f_version to 0.
3389 ns = filp->f_dentry->d_sb->s_fs_info;
3390 tid = (int)filp->f_version;
3391 filp->f_version = 0;
3392 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3394 task = next_tid(task), filp->f_pos++) {
3395 tid = task_pid_nr_ns(task, ns);
3396 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3397 /* returning this tgid failed, save it as the first
3398 * pid for the next readir call */
3399 filp->f_version = (u64)tid;
3400 put_task_struct(task);
3405 put_task_struct(leader);
3410 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3412 struct inode *inode = dentry->d_inode;
3413 struct task_struct *p = get_proc_task(inode);
3414 generic_fillattr(inode, stat);
3417 stat->nlink += get_nr_threads(p);
3424 static const struct inode_operations proc_task_inode_operations = {
3425 .lookup = proc_task_lookup,
3426 .getattr = proc_task_getattr,
3427 .setattr = proc_setattr,
3430 static const struct file_operations proc_task_operations = {
3431 .read = generic_read_dir,
3432 .readdir = proc_task_readdir,
3433 .llseek = default_llseek,