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;
229 if (mutex_lock_killable(&task->cred_guard_mutex))
232 mm = get_task_mm(task);
233 if (mm && mm != current->mm &&
234 !ptrace_may_access(task, PTRACE_MODE_READ)) {
238 mutex_unlock(&task->cred_guard_mutex);
243 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
247 struct mm_struct *mm = get_task_mm(task);
251 goto out_mm; /* Shh! No looking before we're done */
253 len = mm->arg_end - mm->arg_start;
258 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
260 // If the nul at the end of args has been overwritten, then
261 // assume application is using setproctitle(3).
262 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
263 len = strnlen(buffer, res);
267 len = mm->env_end - mm->env_start;
268 if (len > PAGE_SIZE - res)
269 len = PAGE_SIZE - res;
270 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
271 res = strnlen(buffer, res);
280 static int proc_pid_auxv(struct task_struct *task, char *buffer)
283 struct mm_struct *mm = get_task_mm(task);
285 unsigned int nwords = 0;
288 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
289 res = nwords * sizeof(mm->saved_auxv[0]);
292 memcpy(buffer, mm->saved_auxv, res);
299 #ifdef CONFIG_KALLSYMS
301 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
302 * Returns the resolved symbol. If that fails, simply return the address.
304 static int proc_pid_wchan(struct task_struct *task, char *buffer)
307 char symname[KSYM_NAME_LEN];
309 wchan = get_wchan(task);
311 if (lookup_symbol_name(wchan, symname) < 0)
312 if (!ptrace_may_access(task, PTRACE_MODE_READ))
315 return sprintf(buffer, "%lu", wchan);
317 return sprintf(buffer, "%s", symname);
319 #endif /* CONFIG_KALLSYMS */
321 #ifdef CONFIG_STACKTRACE
323 #define MAX_STACK_TRACE_DEPTH 64
325 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
326 struct pid *pid, struct task_struct *task)
328 struct stack_trace trace;
329 unsigned long *entries;
332 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
336 trace.nr_entries = 0;
337 trace.max_entries = MAX_STACK_TRACE_DEPTH;
338 trace.entries = entries;
340 save_stack_trace_tsk(task, &trace);
342 for (i = 0; i < trace.nr_entries; i++) {
343 seq_printf(m, "[<%p>] %pS\n",
344 (void *)entries[i], (void *)entries[i]);
352 #ifdef CONFIG_SCHEDSTATS
354 * Provides /proc/PID/schedstat
356 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
358 return sprintf(buffer, "%llu %llu %lu\n",
359 (unsigned long long)task->se.sum_exec_runtime,
360 (unsigned long long)task->sched_info.run_delay,
361 task->sched_info.pcount);
365 #ifdef CONFIG_LATENCYTOP
366 static int lstats_show_proc(struct seq_file *m, void *v)
369 struct inode *inode = m->private;
370 struct task_struct *task = get_proc_task(inode);
374 seq_puts(m, "Latency Top version : v0.1\n");
375 for (i = 0; i < 32; i++) {
376 if (task->latency_record[i].backtrace[0]) {
378 seq_printf(m, "%i %li %li ",
379 task->latency_record[i].count,
380 task->latency_record[i].time,
381 task->latency_record[i].max);
382 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
383 char sym[KSYM_SYMBOL_LEN];
385 if (!task->latency_record[i].backtrace[q])
387 if (task->latency_record[i].backtrace[q] == ULONG_MAX)
389 sprint_symbol(sym, task->latency_record[i].backtrace[q]);
390 c = strchr(sym, '+');
393 seq_printf(m, "%s ", sym);
399 put_task_struct(task);
403 static int lstats_open(struct inode *inode, struct file *file)
405 return single_open(file, lstats_show_proc, inode);
408 static ssize_t lstats_write(struct file *file, const char __user *buf,
409 size_t count, loff_t *offs)
411 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
415 clear_all_latency_tracing(task);
416 put_task_struct(task);
421 static const struct file_operations proc_lstats_operations = {
424 .write = lstats_write,
426 .release = single_release,
431 static int proc_oom_score(struct task_struct *task, char *buffer)
433 unsigned long points = 0;
435 read_lock(&tasklist_lock);
437 points = oom_badness(task, NULL, NULL,
438 totalram_pages + total_swap_pages);
439 read_unlock(&tasklist_lock);
440 return sprintf(buffer, "%lu\n", points);
448 static const struct limit_names lnames[RLIM_NLIMITS] = {
449 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
450 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
451 [RLIMIT_DATA] = {"Max data size", "bytes"},
452 [RLIMIT_STACK] = {"Max stack size", "bytes"},
453 [RLIMIT_CORE] = {"Max core file size", "bytes"},
454 [RLIMIT_RSS] = {"Max resident set", "bytes"},
455 [RLIMIT_NPROC] = {"Max processes", "processes"},
456 [RLIMIT_NOFILE] = {"Max open files", "files"},
457 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
458 [RLIMIT_AS] = {"Max address space", "bytes"},
459 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
460 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
461 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
462 [RLIMIT_NICE] = {"Max nice priority", NULL},
463 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
464 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
467 /* Display limits for a process */
468 static int proc_pid_limits(struct task_struct *task, char *buffer)
473 char *bufptr = buffer;
475 struct rlimit rlim[RLIM_NLIMITS];
477 if (!lock_task_sighand(task, &flags))
479 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
480 unlock_task_sighand(task, &flags);
483 * print the file header
485 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
486 "Limit", "Soft Limit", "Hard Limit", "Units");
488 for (i = 0; i < RLIM_NLIMITS; i++) {
489 if (rlim[i].rlim_cur == RLIM_INFINITY)
490 count += sprintf(&bufptr[count], "%-25s %-20s ",
491 lnames[i].name, "unlimited");
493 count += sprintf(&bufptr[count], "%-25s %-20lu ",
494 lnames[i].name, rlim[i].rlim_cur);
496 if (rlim[i].rlim_max == RLIM_INFINITY)
497 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
499 count += sprintf(&bufptr[count], "%-20lu ",
503 count += sprintf(&bufptr[count], "%-10s\n",
506 count += sprintf(&bufptr[count], "\n");
512 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
513 static int proc_pid_syscall(struct task_struct *task, char *buffer)
516 unsigned long args[6], sp, pc;
518 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
519 return sprintf(buffer, "running\n");
522 return sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
524 return sprintf(buffer,
525 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
527 args[0], args[1], args[2], args[3], args[4], args[5],
530 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
532 /************************************************************************/
533 /* Here the fs part begins */
534 /************************************************************************/
536 /* permission checks */
537 static int proc_fd_access_allowed(struct inode *inode)
539 struct task_struct *task;
541 /* Allow access to a task's file descriptors if it is us or we
542 * may use ptrace attach to the process and find out that
545 task = get_proc_task(inode);
547 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
548 put_task_struct(task);
553 static int proc_setattr(struct dentry *dentry, struct iattr *attr)
556 struct inode *inode = dentry->d_inode;
558 if (attr->ia_valid & ATTR_MODE)
561 error = inode_change_ok(inode, attr);
565 if ((attr->ia_valid & ATTR_SIZE) &&
566 attr->ia_size != i_size_read(inode)) {
567 error = vmtruncate(inode, attr->ia_size);
572 setattr_copy(inode, attr);
573 mark_inode_dirty(inode);
577 static const struct inode_operations proc_def_inode_operations = {
578 .setattr = proc_setattr,
581 static int mounts_open_common(struct inode *inode, struct file *file,
582 const struct seq_operations *op)
584 struct task_struct *task = get_proc_task(inode);
586 struct mnt_namespace *ns = NULL;
588 struct proc_mounts *p;
593 nsp = task_nsproxy(task);
600 if (ns && get_task_root(task, &root) == 0)
602 put_task_struct(task);
611 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
615 file->private_data = &p->m;
616 ret = seq_open(file, op);
623 p->event = ns->event;
637 static int mounts_release(struct inode *inode, struct file *file)
639 struct proc_mounts *p = file->private_data;
642 return seq_release(inode, file);
645 static unsigned mounts_poll(struct file *file, poll_table *wait)
647 struct proc_mounts *p = file->private_data;
648 unsigned res = POLLIN | POLLRDNORM;
650 poll_wait(file, &p->ns->poll, wait);
651 if (mnt_had_events(p))
652 res |= POLLERR | POLLPRI;
657 static int mounts_open(struct inode *inode, struct file *file)
659 return mounts_open_common(inode, file, &mounts_op);
662 static const struct file_operations proc_mounts_operations = {
666 .release = mounts_release,
670 static int mountinfo_open(struct inode *inode, struct file *file)
672 return mounts_open_common(inode, file, &mountinfo_op);
675 static const struct file_operations proc_mountinfo_operations = {
676 .open = mountinfo_open,
679 .release = mounts_release,
683 static int mountstats_open(struct inode *inode, struct file *file)
685 return mounts_open_common(inode, file, &mountstats_op);
688 static const struct file_operations proc_mountstats_operations = {
689 .open = mountstats_open,
692 .release = mounts_release,
695 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
697 static ssize_t proc_info_read(struct file * file, char __user * buf,
698 size_t count, loff_t *ppos)
700 struct inode * inode = file->f_path.dentry->d_inode;
703 struct task_struct *task = get_proc_task(inode);
709 if (count > PROC_BLOCK_SIZE)
710 count = PROC_BLOCK_SIZE;
713 if (!(page = __get_free_page(GFP_TEMPORARY)))
716 length = PROC_I(inode)->op.proc_read(task, (char*)page);
719 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
722 put_task_struct(task);
727 static const struct file_operations proc_info_file_operations = {
728 .read = proc_info_read,
729 .llseek = generic_file_llseek,
732 static int proc_single_show(struct seq_file *m, void *v)
734 struct inode *inode = m->private;
735 struct pid_namespace *ns;
737 struct task_struct *task;
740 ns = inode->i_sb->s_fs_info;
741 pid = proc_pid(inode);
742 task = get_pid_task(pid, PIDTYPE_PID);
746 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
748 put_task_struct(task);
752 static int proc_single_open(struct inode *inode, struct file *filp)
755 ret = single_open(filp, proc_single_show, NULL);
757 struct seq_file *m = filp->private_data;
764 static const struct file_operations proc_single_file_operations = {
765 .open = proc_single_open,
768 .release = single_release,
771 static int mem_open(struct inode* inode, struct file* file)
773 file->private_data = (void*)((long)current->self_exec_id);
777 static ssize_t mem_read(struct file * file, char __user * buf,
778 size_t count, loff_t *ppos)
780 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
782 unsigned long src = *ppos;
784 struct mm_struct *mm;
789 if (check_mem_permission(task))
793 page = (char *)__get_free_page(GFP_TEMPORARY);
799 mm = get_task_mm(task);
805 if (file->private_data != (void*)((long)current->self_exec_id))
811 int this_len, retval;
813 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
814 retval = access_process_vm(task, src, page, this_len, 0);
815 if (!retval || check_mem_permission(task)) {
821 if (copy_to_user(buf, page, retval)) {
836 free_page((unsigned long) page);
838 put_task_struct(task);
843 #define mem_write NULL
846 /* This is a security hazard */
847 static ssize_t mem_write(struct file * file, const char __user *buf,
848 size_t count, loff_t *ppos)
852 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
853 unsigned long dst = *ppos;
859 if (check_mem_permission(task))
863 page = (char *)__get_free_page(GFP_TEMPORARY);
869 int this_len, retval;
871 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
872 if (copy_from_user(page, buf, this_len)) {
876 retval = access_process_vm(task, dst, page, this_len, 1);
888 free_page((unsigned long) page);
890 put_task_struct(task);
896 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
900 file->f_pos = offset;
903 file->f_pos += offset;
908 force_successful_syscall_return();
912 static const struct file_operations proc_mem_operations = {
919 static ssize_t environ_read(struct file *file, char __user *buf,
920 size_t count, loff_t *ppos)
922 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
924 unsigned long src = *ppos;
926 struct mm_struct *mm;
931 if (!ptrace_may_access(task, PTRACE_MODE_READ))
935 page = (char *)__get_free_page(GFP_TEMPORARY);
941 mm = get_task_mm(task);
946 int this_len, retval, max_len;
948 this_len = mm->env_end - (mm->env_start + src);
953 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
954 this_len = (this_len > max_len) ? max_len : this_len;
956 retval = access_process_vm(task, (mm->env_start + src),
964 if (copy_to_user(buf, page, retval)) {
978 free_page((unsigned long) page);
980 put_task_struct(task);
985 static const struct file_operations proc_environ_operations = {
986 .read = environ_read,
987 .llseek = generic_file_llseek,
990 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
991 size_t count, loff_t *ppos)
993 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
994 char buffer[PROC_NUMBUF];
996 int oom_adjust = OOM_DISABLE;
1002 if (lock_task_sighand(task, &flags)) {
1003 oom_adjust = task->signal->oom_adj;
1004 unlock_task_sighand(task, &flags);
1007 put_task_struct(task);
1009 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
1011 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1014 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1015 size_t count, loff_t *ppos)
1017 struct task_struct *task;
1018 char buffer[PROC_NUMBUF];
1020 unsigned long flags;
1023 memset(buffer, 0, sizeof(buffer));
1024 if (count > sizeof(buffer) - 1)
1025 count = sizeof(buffer) - 1;
1026 if (copy_from_user(buffer, buf, count))
1029 err = strict_strtol(strstrip(buffer), 0, &oom_adjust);
1032 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
1033 oom_adjust != OOM_DISABLE)
1036 task = get_proc_task(file->f_path.dentry->d_inode);
1039 if (!lock_task_sighand(task, &flags)) {
1040 put_task_struct(task);
1044 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
1045 unlock_task_sighand(task, &flags);
1046 put_task_struct(task);
1053 unlock_task_sighand(task, &flags);
1054 put_task_struct(task);
1058 if (oom_adjust != task->signal->oom_adj) {
1059 if (oom_adjust == OOM_DISABLE)
1060 atomic_inc(&task->mm->oom_disable_count);
1061 if (task->signal->oom_adj == OOM_DISABLE)
1062 atomic_dec(&task->mm->oom_disable_count);
1066 * Warn that /proc/pid/oom_adj is deprecated, see
1067 * Documentation/feature-removal-schedule.txt.
1069 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, "
1070 "please use /proc/%d/oom_score_adj instead.\n",
1071 current->comm, task_pid_nr(current),
1072 task_pid_nr(task), task_pid_nr(task));
1073 task->signal->oom_adj = oom_adjust;
1075 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1076 * value is always attainable.
1078 if (task->signal->oom_adj == OOM_ADJUST_MAX)
1079 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
1081 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
1084 unlock_task_sighand(task, &flags);
1085 put_task_struct(task);
1090 static const struct file_operations proc_oom_adjust_operations = {
1091 .read = oom_adjust_read,
1092 .write = oom_adjust_write,
1093 .llseek = generic_file_llseek,
1096 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1097 size_t count, loff_t *ppos)
1099 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1100 char buffer[PROC_NUMBUF];
1101 int oom_score_adj = OOM_SCORE_ADJ_MIN;
1102 unsigned long flags;
1107 if (lock_task_sighand(task, &flags)) {
1108 oom_score_adj = task->signal->oom_score_adj;
1109 unlock_task_sighand(task, &flags);
1111 put_task_struct(task);
1112 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1113 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1116 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1117 size_t count, loff_t *ppos)
1119 struct task_struct *task;
1120 char buffer[PROC_NUMBUF];
1121 unsigned long flags;
1125 memset(buffer, 0, sizeof(buffer));
1126 if (count > sizeof(buffer) - 1)
1127 count = sizeof(buffer) - 1;
1128 if (copy_from_user(buffer, buf, count))
1131 err = strict_strtol(strstrip(buffer), 0, &oom_score_adj);
1134 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1135 oom_score_adj > OOM_SCORE_ADJ_MAX)
1138 task = get_proc_task(file->f_path.dentry->d_inode);
1141 if (!lock_task_sighand(task, &flags)) {
1142 put_task_struct(task);
1145 if (oom_score_adj < task->signal->oom_score_adj &&
1146 !capable(CAP_SYS_RESOURCE)) {
1147 unlock_task_sighand(task, &flags);
1148 put_task_struct(task);
1155 unlock_task_sighand(task, &flags);
1156 put_task_struct(task);
1159 if (oom_score_adj != task->signal->oom_score_adj) {
1160 if (oom_score_adj == OOM_SCORE_ADJ_MIN)
1161 atomic_inc(&task->mm->oom_disable_count);
1162 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1163 atomic_dec(&task->mm->oom_disable_count);
1165 task->signal->oom_score_adj = oom_score_adj;
1167 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1168 * always attainable.
1170 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1171 task->signal->oom_adj = OOM_DISABLE;
1173 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1176 unlock_task_sighand(task, &flags);
1177 put_task_struct(task);
1181 static const struct file_operations proc_oom_score_adj_operations = {
1182 .read = oom_score_adj_read,
1183 .write = oom_score_adj_write,
1184 .llseek = default_llseek,
1187 #ifdef CONFIG_AUDITSYSCALL
1188 #define TMPBUFLEN 21
1189 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1190 size_t count, loff_t *ppos)
1192 struct inode * inode = file->f_path.dentry->d_inode;
1193 struct task_struct *task = get_proc_task(inode);
1195 char tmpbuf[TMPBUFLEN];
1199 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1200 audit_get_loginuid(task));
1201 put_task_struct(task);
1202 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1205 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1206 size_t count, loff_t *ppos)
1208 struct inode * inode = file->f_path.dentry->d_inode;
1213 if (!capable(CAP_AUDIT_CONTROL))
1217 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1223 if (count >= PAGE_SIZE)
1224 count = PAGE_SIZE - 1;
1227 /* No partial writes. */
1230 page = (char*)__get_free_page(GFP_TEMPORARY);
1234 if (copy_from_user(page, buf, count))
1238 loginuid = simple_strtoul(page, &tmp, 10);
1244 length = audit_set_loginuid(current, loginuid);
1245 if (likely(length == 0))
1249 free_page((unsigned long) page);
1253 static const struct file_operations proc_loginuid_operations = {
1254 .read = proc_loginuid_read,
1255 .write = proc_loginuid_write,
1256 .llseek = generic_file_llseek,
1259 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1260 size_t count, loff_t *ppos)
1262 struct inode * inode = file->f_path.dentry->d_inode;
1263 struct task_struct *task = get_proc_task(inode);
1265 char tmpbuf[TMPBUFLEN];
1269 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1270 audit_get_sessionid(task));
1271 put_task_struct(task);
1272 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1275 static const struct file_operations proc_sessionid_operations = {
1276 .read = proc_sessionid_read,
1277 .llseek = generic_file_llseek,
1281 #ifdef CONFIG_FAULT_INJECTION
1282 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1283 size_t count, loff_t *ppos)
1285 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1286 char buffer[PROC_NUMBUF];
1292 make_it_fail = task->make_it_fail;
1293 put_task_struct(task);
1295 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1297 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1300 static ssize_t proc_fault_inject_write(struct file * file,
1301 const char __user * buf, size_t count, loff_t *ppos)
1303 struct task_struct *task;
1304 char buffer[PROC_NUMBUF], *end;
1307 if (!capable(CAP_SYS_RESOURCE))
1309 memset(buffer, 0, sizeof(buffer));
1310 if (count > sizeof(buffer) - 1)
1311 count = sizeof(buffer) - 1;
1312 if (copy_from_user(buffer, buf, count))
1314 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1317 task = get_proc_task(file->f_dentry->d_inode);
1320 task->make_it_fail = make_it_fail;
1321 put_task_struct(task);
1326 static const struct file_operations proc_fault_inject_operations = {
1327 .read = proc_fault_inject_read,
1328 .write = proc_fault_inject_write,
1329 .llseek = generic_file_llseek,
1334 #ifdef CONFIG_SCHED_DEBUG
1336 * Print out various scheduling related per-task fields:
1338 static int sched_show(struct seq_file *m, void *v)
1340 struct inode *inode = m->private;
1341 struct task_struct *p;
1343 p = get_proc_task(inode);
1346 proc_sched_show_task(p, m);
1354 sched_write(struct file *file, const char __user *buf,
1355 size_t count, loff_t *offset)
1357 struct inode *inode = file->f_path.dentry->d_inode;
1358 struct task_struct *p;
1360 p = get_proc_task(inode);
1363 proc_sched_set_task(p);
1370 static int sched_open(struct inode *inode, struct file *filp)
1374 ret = single_open(filp, sched_show, NULL);
1376 struct seq_file *m = filp->private_data;
1383 static const struct file_operations proc_pid_sched_operations = {
1386 .write = sched_write,
1387 .llseek = seq_lseek,
1388 .release = single_release,
1393 static ssize_t comm_write(struct file *file, const char __user *buf,
1394 size_t count, loff_t *offset)
1396 struct inode *inode = file->f_path.dentry->d_inode;
1397 struct task_struct *p;
1398 char buffer[TASK_COMM_LEN];
1400 memset(buffer, 0, sizeof(buffer));
1401 if (count > sizeof(buffer) - 1)
1402 count = sizeof(buffer) - 1;
1403 if (copy_from_user(buffer, buf, count))
1406 p = get_proc_task(inode);
1410 if (same_thread_group(current, p))
1411 set_task_comm(p, buffer);
1420 static int comm_show(struct seq_file *m, void *v)
1422 struct inode *inode = m->private;
1423 struct task_struct *p;
1425 p = get_proc_task(inode);
1430 seq_printf(m, "%s\n", p->comm);
1438 static int comm_open(struct inode *inode, struct file *filp)
1442 ret = single_open(filp, comm_show, NULL);
1444 struct seq_file *m = filp->private_data;
1451 static const struct file_operations proc_pid_set_comm_operations = {
1454 .write = comm_write,
1455 .llseek = seq_lseek,
1456 .release = single_release,
1460 * We added or removed a vma mapping the executable. The vmas are only mapped
1461 * during exec and are not mapped with the mmap system call.
1462 * Callers must hold down_write() on the mm's mmap_sem for these
1464 void added_exe_file_vma(struct mm_struct *mm)
1466 mm->num_exe_file_vmas++;
1469 void removed_exe_file_vma(struct mm_struct *mm)
1471 mm->num_exe_file_vmas--;
1472 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
1474 mm->exe_file = NULL;
1479 void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1482 get_file(new_exe_file);
1485 mm->exe_file = new_exe_file;
1486 mm->num_exe_file_vmas = 0;
1489 struct file *get_mm_exe_file(struct mm_struct *mm)
1491 struct file *exe_file;
1493 /* We need mmap_sem to protect against races with removal of
1494 * VM_EXECUTABLE vmas */
1495 down_read(&mm->mmap_sem);
1496 exe_file = mm->exe_file;
1499 up_read(&mm->mmap_sem);
1503 void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
1505 /* It's safe to write the exe_file pointer without exe_file_lock because
1506 * this is called during fork when the task is not yet in /proc */
1507 newmm->exe_file = get_mm_exe_file(oldmm);
1510 static int proc_exe_link(struct inode *inode, struct path *exe_path)
1512 struct task_struct *task;
1513 struct mm_struct *mm;
1514 struct file *exe_file;
1516 task = get_proc_task(inode);
1519 mm = get_task_mm(task);
1520 put_task_struct(task);
1523 exe_file = get_mm_exe_file(mm);
1526 *exe_path = exe_file->f_path;
1527 path_get(&exe_file->f_path);
1534 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1536 struct inode *inode = dentry->d_inode;
1537 int error = -EACCES;
1539 /* We don't need a base pointer in the /proc filesystem */
1540 path_put(&nd->path);
1542 /* Are we allowed to snoop on the tasks file descriptors? */
1543 if (!proc_fd_access_allowed(inode))
1546 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1548 return ERR_PTR(error);
1551 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1553 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1560 pathname = d_path_with_unreachable(path, tmp, PAGE_SIZE);
1561 len = PTR_ERR(pathname);
1562 if (IS_ERR(pathname))
1564 len = tmp + PAGE_SIZE - 1 - pathname;
1568 if (copy_to_user(buffer, pathname, len))
1571 free_page((unsigned long)tmp);
1575 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1577 int error = -EACCES;
1578 struct inode *inode = dentry->d_inode;
1581 /* Are we allowed to snoop on the tasks file descriptors? */
1582 if (!proc_fd_access_allowed(inode))
1585 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1589 error = do_proc_readlink(&path, buffer, buflen);
1595 static const struct inode_operations proc_pid_link_inode_operations = {
1596 .readlink = proc_pid_readlink,
1597 .follow_link = proc_pid_follow_link,
1598 .setattr = proc_setattr,
1602 /* building an inode */
1604 static int task_dumpable(struct task_struct *task)
1607 struct mm_struct *mm;
1612 dumpable = get_dumpable(mm);
1620 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1622 struct inode * inode;
1623 struct proc_inode *ei;
1624 const struct cred *cred;
1626 /* We need a new inode */
1628 inode = new_inode(sb);
1634 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1635 inode->i_op = &proc_def_inode_operations;
1638 * grab the reference to task.
1640 ei->pid = get_task_pid(task, PIDTYPE_PID);
1644 if (task_dumpable(task)) {
1646 cred = __task_cred(task);
1647 inode->i_uid = cred->euid;
1648 inode->i_gid = cred->egid;
1651 security_task_to_inode(task, inode);
1661 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1663 struct inode *inode = dentry->d_inode;
1664 struct task_struct *task;
1665 const struct cred *cred;
1667 generic_fillattr(inode, stat);
1672 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1674 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1675 task_dumpable(task)) {
1676 cred = __task_cred(task);
1677 stat->uid = cred->euid;
1678 stat->gid = cred->egid;
1688 * Exceptional case: normally we are not allowed to unhash a busy
1689 * directory. In this case, however, we can do it - no aliasing problems
1690 * due to the way we treat inodes.
1692 * Rewrite the inode's ownerships here because the owning task may have
1693 * performed a setuid(), etc.
1695 * Before the /proc/pid/status file was created the only way to read
1696 * the effective uid of a /process was to stat /proc/pid. Reading
1697 * /proc/pid/status is slow enough that procps and other packages
1698 * kept stating /proc/pid. To keep the rules in /proc simple I have
1699 * made this apply to all per process world readable and executable
1702 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1704 struct inode *inode = dentry->d_inode;
1705 struct task_struct *task = get_proc_task(inode);
1706 const struct cred *cred;
1709 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1710 task_dumpable(task)) {
1712 cred = __task_cred(task);
1713 inode->i_uid = cred->euid;
1714 inode->i_gid = cred->egid;
1720 inode->i_mode &= ~(S_ISUID | S_ISGID);
1721 security_task_to_inode(task, inode);
1722 put_task_struct(task);
1729 static int pid_delete_dentry(struct dentry * dentry)
1731 /* Is the task we represent dead?
1732 * If so, then don't put the dentry on the lru list,
1733 * kill it immediately.
1735 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1738 static const struct dentry_operations pid_dentry_operations =
1740 .d_revalidate = pid_revalidate,
1741 .d_delete = pid_delete_dentry,
1746 typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1747 struct task_struct *, const void *);
1750 * Fill a directory entry.
1752 * If possible create the dcache entry and derive our inode number and
1753 * file type from dcache entry.
1755 * Since all of the proc inode numbers are dynamically generated, the inode
1756 * numbers do not exist until the inode is cache. This means creating the
1757 * the dcache entry in readdir is necessary to keep the inode numbers
1758 * reported by readdir in sync with the inode numbers reported
1761 static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1762 char *name, int len,
1763 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1765 struct dentry *child, *dir = filp->f_path.dentry;
1766 struct inode *inode;
1769 unsigned type = DT_UNKNOWN;
1773 qname.hash = full_name_hash(name, len);
1775 child = d_lookup(dir, &qname);
1778 new = d_alloc(dir, &qname);
1780 child = instantiate(dir->d_inode, new, task, ptr);
1787 if (!child || IS_ERR(child) || !child->d_inode)
1788 goto end_instantiate;
1789 inode = child->d_inode;
1792 type = inode->i_mode >> 12;
1797 ino = find_inode_number(dir, &qname);
1800 return filldir(dirent, name, len, filp->f_pos, ino, type);
1803 static unsigned name_to_int(struct dentry *dentry)
1805 const char *name = dentry->d_name.name;
1806 int len = dentry->d_name.len;
1809 if (len > 1 && *name == '0')
1812 unsigned c = *name++ - '0';
1815 if (n >= (~0U-9)/10)
1825 #define PROC_FDINFO_MAX 64
1827 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1829 struct task_struct *task = get_proc_task(inode);
1830 struct files_struct *files = NULL;
1832 int fd = proc_fd(inode);
1835 files = get_files_struct(task);
1836 put_task_struct(task);
1840 * We are not taking a ref to the file structure, so we must
1843 spin_lock(&files->file_lock);
1844 file = fcheck_files(files, fd);
1847 *path = file->f_path;
1848 path_get(&file->f_path);
1851 snprintf(info, PROC_FDINFO_MAX,
1854 (long long) file->f_pos,
1856 spin_unlock(&files->file_lock);
1857 put_files_struct(files);
1860 spin_unlock(&files->file_lock);
1861 put_files_struct(files);
1866 static int proc_fd_link(struct inode *inode, struct path *path)
1868 return proc_fd_info(inode, path, NULL);
1871 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1873 struct inode *inode = dentry->d_inode;
1874 struct task_struct *task = get_proc_task(inode);
1875 int fd = proc_fd(inode);
1876 struct files_struct *files;
1877 const struct cred *cred;
1880 files = get_files_struct(task);
1883 if (fcheck_files(files, fd)) {
1885 put_files_struct(files);
1886 if (task_dumpable(task)) {
1888 cred = __task_cred(task);
1889 inode->i_uid = cred->euid;
1890 inode->i_gid = cred->egid;
1896 inode->i_mode &= ~(S_ISUID | S_ISGID);
1897 security_task_to_inode(task, inode);
1898 put_task_struct(task);
1902 put_files_struct(files);
1904 put_task_struct(task);
1910 static const struct dentry_operations tid_fd_dentry_operations =
1912 .d_revalidate = tid_fd_revalidate,
1913 .d_delete = pid_delete_dentry,
1916 static struct dentry *proc_fd_instantiate(struct inode *dir,
1917 struct dentry *dentry, struct task_struct *task, const void *ptr)
1919 unsigned fd = *(const unsigned *)ptr;
1921 struct files_struct *files;
1922 struct inode *inode;
1923 struct proc_inode *ei;
1924 struct dentry *error = ERR_PTR(-ENOENT);
1926 inode = proc_pid_make_inode(dir->i_sb, task);
1931 files = get_files_struct(task);
1934 inode->i_mode = S_IFLNK;
1937 * We are not taking a ref to the file structure, so we must
1940 spin_lock(&files->file_lock);
1941 file = fcheck_files(files, fd);
1944 if (file->f_mode & FMODE_READ)
1945 inode->i_mode |= S_IRUSR | S_IXUSR;
1946 if (file->f_mode & FMODE_WRITE)
1947 inode->i_mode |= S_IWUSR | S_IXUSR;
1948 spin_unlock(&files->file_lock);
1949 put_files_struct(files);
1951 inode->i_op = &proc_pid_link_inode_operations;
1953 ei->op.proc_get_link = proc_fd_link;
1954 dentry->d_op = &tid_fd_dentry_operations;
1955 d_add(dentry, inode);
1956 /* Close the race of the process dying before we return the dentry */
1957 if (tid_fd_revalidate(dentry, NULL))
1963 spin_unlock(&files->file_lock);
1964 put_files_struct(files);
1970 static struct dentry *proc_lookupfd_common(struct inode *dir,
1971 struct dentry *dentry,
1972 instantiate_t instantiate)
1974 struct task_struct *task = get_proc_task(dir);
1975 unsigned fd = name_to_int(dentry);
1976 struct dentry *result = ERR_PTR(-ENOENT);
1983 result = instantiate(dir, dentry, task, &fd);
1985 put_task_struct(task);
1990 static int proc_readfd_common(struct file * filp, void * dirent,
1991 filldir_t filldir, instantiate_t instantiate)
1993 struct dentry *dentry = filp->f_path.dentry;
1994 struct inode *inode = dentry->d_inode;
1995 struct task_struct *p = get_proc_task(inode);
1996 unsigned int fd, ino;
1998 struct files_struct * files;
2008 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
2012 ino = parent_ino(dentry);
2013 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2017 files = get_files_struct(p);
2021 for (fd = filp->f_pos-2;
2022 fd < files_fdtable(files)->max_fds;
2023 fd++, filp->f_pos++) {
2024 char name[PROC_NUMBUF];
2027 if (!fcheck_files(files, fd))
2031 len = snprintf(name, sizeof(name), "%d", fd);
2032 if (proc_fill_cache(filp, dirent, filldir,
2033 name, len, instantiate,
2041 put_files_struct(files);
2049 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
2050 struct nameidata *nd)
2052 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
2055 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
2057 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
2060 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
2061 size_t len, loff_t *ppos)
2063 char tmp[PROC_FDINFO_MAX];
2064 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
2066 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
2070 static const struct file_operations proc_fdinfo_file_operations = {
2071 .open = nonseekable_open,
2072 .read = proc_fdinfo_read,
2073 .llseek = no_llseek,
2076 static const struct file_operations proc_fd_operations = {
2077 .read = generic_read_dir,
2078 .readdir = proc_readfd,
2079 .llseek = default_llseek,
2083 * /proc/pid/fd needs a special permission handler so that a process can still
2084 * access /proc/self/fd after it has executed a setuid().
2086 static int proc_fd_permission(struct inode *inode, int mask)
2090 rv = generic_permission(inode, mask, NULL);
2093 if (task_pid(current) == proc_pid(inode))
2099 * proc directories can do almost nothing..
2101 static const struct inode_operations proc_fd_inode_operations = {
2102 .lookup = proc_lookupfd,
2103 .permission = proc_fd_permission,
2104 .setattr = proc_setattr,
2107 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2108 struct dentry *dentry, struct task_struct *task, const void *ptr)
2110 unsigned fd = *(unsigned *)ptr;
2111 struct inode *inode;
2112 struct proc_inode *ei;
2113 struct dentry *error = ERR_PTR(-ENOENT);
2115 inode = proc_pid_make_inode(dir->i_sb, task);
2120 inode->i_mode = S_IFREG | S_IRUSR;
2121 inode->i_fop = &proc_fdinfo_file_operations;
2122 dentry->d_op = &tid_fd_dentry_operations;
2123 d_add(dentry, inode);
2124 /* Close the race of the process dying before we return the dentry */
2125 if (tid_fd_revalidate(dentry, NULL))
2132 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2133 struct dentry *dentry,
2134 struct nameidata *nd)
2136 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2139 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2141 return proc_readfd_common(filp, dirent, filldir,
2142 proc_fdinfo_instantiate);
2145 static const struct file_operations proc_fdinfo_operations = {
2146 .read = generic_read_dir,
2147 .readdir = proc_readfdinfo,
2148 .llseek = default_llseek,
2152 * proc directories can do almost nothing..
2154 static const struct inode_operations proc_fdinfo_inode_operations = {
2155 .lookup = proc_lookupfdinfo,
2156 .setattr = proc_setattr,
2160 static struct dentry *proc_pident_instantiate(struct inode *dir,
2161 struct dentry *dentry, struct task_struct *task, const void *ptr)
2163 const struct pid_entry *p = ptr;
2164 struct inode *inode;
2165 struct proc_inode *ei;
2166 struct dentry *error = ERR_PTR(-ENOENT);
2168 inode = proc_pid_make_inode(dir->i_sb, task);
2173 inode->i_mode = p->mode;
2174 if (S_ISDIR(inode->i_mode))
2175 inode->i_nlink = 2; /* Use getattr to fix if necessary */
2177 inode->i_op = p->iop;
2179 inode->i_fop = p->fop;
2181 dentry->d_op = &pid_dentry_operations;
2182 d_add(dentry, inode);
2183 /* Close the race of the process dying before we return the dentry */
2184 if (pid_revalidate(dentry, NULL))
2190 static struct dentry *proc_pident_lookup(struct inode *dir,
2191 struct dentry *dentry,
2192 const struct pid_entry *ents,
2195 struct dentry *error;
2196 struct task_struct *task = get_proc_task(dir);
2197 const struct pid_entry *p, *last;
2199 error = ERR_PTR(-ENOENT);
2205 * Yes, it does not scale. And it should not. Don't add
2206 * new entries into /proc/<tgid>/ without very good reasons.
2208 last = &ents[nents - 1];
2209 for (p = ents; p <= last; p++) {
2210 if (p->len != dentry->d_name.len)
2212 if (!memcmp(dentry->d_name.name, p->name, p->len))
2218 error = proc_pident_instantiate(dir, dentry, task, p);
2220 put_task_struct(task);
2225 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2226 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2228 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2229 proc_pident_instantiate, task, p);
2232 static int proc_pident_readdir(struct file *filp,
2233 void *dirent, filldir_t filldir,
2234 const struct pid_entry *ents, unsigned int nents)
2237 struct dentry *dentry = filp->f_path.dentry;
2238 struct inode *inode = dentry->d_inode;
2239 struct task_struct *task = get_proc_task(inode);
2240 const struct pid_entry *p, *last;
2253 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2259 ino = parent_ino(dentry);
2260 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2272 last = &ents[nents - 1];
2274 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2283 put_task_struct(task);
2288 #ifdef CONFIG_SECURITY
2289 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2290 size_t count, loff_t *ppos)
2292 struct inode * inode = file->f_path.dentry->d_inode;
2295 struct task_struct *task = get_proc_task(inode);
2300 length = security_getprocattr(task,
2301 (char*)file->f_path.dentry->d_name.name,
2303 put_task_struct(task);
2305 length = simple_read_from_buffer(buf, count, ppos, p, length);
2310 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2311 size_t count, loff_t *ppos)
2313 struct inode * inode = file->f_path.dentry->d_inode;
2316 struct task_struct *task = get_proc_task(inode);
2321 if (count > PAGE_SIZE)
2324 /* No partial writes. */
2330 page = (char*)__get_free_page(GFP_TEMPORARY);
2335 if (copy_from_user(page, buf, count))
2338 /* Guard against adverse ptrace interaction */
2339 length = mutex_lock_interruptible(&task->cred_guard_mutex);
2343 length = security_setprocattr(task,
2344 (char*)file->f_path.dentry->d_name.name,
2345 (void*)page, count);
2346 mutex_unlock(&task->cred_guard_mutex);
2348 free_page((unsigned long) page);
2350 put_task_struct(task);
2355 static const struct file_operations proc_pid_attr_operations = {
2356 .read = proc_pid_attr_read,
2357 .write = proc_pid_attr_write,
2358 .llseek = generic_file_llseek,
2361 static const struct pid_entry attr_dir_stuff[] = {
2362 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2363 REG("prev", S_IRUGO, proc_pid_attr_operations),
2364 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2365 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2366 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2367 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2370 static int proc_attr_dir_readdir(struct file * filp,
2371 void * dirent, filldir_t filldir)
2373 return proc_pident_readdir(filp,dirent,filldir,
2374 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2377 static const struct file_operations proc_attr_dir_operations = {
2378 .read = generic_read_dir,
2379 .readdir = proc_attr_dir_readdir,
2380 .llseek = default_llseek,
2383 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2384 struct dentry *dentry, struct nameidata *nd)
2386 return proc_pident_lookup(dir, dentry,
2387 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2390 static const struct inode_operations proc_attr_dir_inode_operations = {
2391 .lookup = proc_attr_dir_lookup,
2392 .getattr = pid_getattr,
2393 .setattr = proc_setattr,
2398 #ifdef CONFIG_ELF_CORE
2399 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2400 size_t count, loff_t *ppos)
2402 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2403 struct mm_struct *mm;
2404 char buffer[PROC_NUMBUF];
2412 mm = get_task_mm(task);
2414 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2415 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2416 MMF_DUMP_FILTER_SHIFT));
2418 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2421 put_task_struct(task);
2426 static ssize_t proc_coredump_filter_write(struct file *file,
2427 const char __user *buf,
2431 struct task_struct *task;
2432 struct mm_struct *mm;
2433 char buffer[PROC_NUMBUF], *end;
2440 memset(buffer, 0, sizeof(buffer));
2441 if (count > sizeof(buffer) - 1)
2442 count = sizeof(buffer) - 1;
2443 if (copy_from_user(buffer, buf, count))
2447 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2450 if (end - buffer == 0)
2454 task = get_proc_task(file->f_dentry->d_inode);
2459 mm = get_task_mm(task);
2463 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2465 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2467 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2472 put_task_struct(task);
2477 static const struct file_operations proc_coredump_filter_operations = {
2478 .read = proc_coredump_filter_read,
2479 .write = proc_coredump_filter_write,
2480 .llseek = generic_file_llseek,
2487 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2490 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2491 pid_t tgid = task_tgid_nr_ns(current, ns);
2492 char tmp[PROC_NUMBUF];
2495 sprintf(tmp, "%d", tgid);
2496 return vfs_readlink(dentry,buffer,buflen,tmp);
2499 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2501 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2502 pid_t tgid = task_tgid_nr_ns(current, ns);
2503 char *name = ERR_PTR(-ENOENT);
2507 name = ERR_PTR(-ENOMEM);
2509 sprintf(name, "%d", tgid);
2511 nd_set_link(nd, name);
2515 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2518 char *s = nd_get_link(nd);
2523 static const struct inode_operations proc_self_inode_operations = {
2524 .readlink = proc_self_readlink,
2525 .follow_link = proc_self_follow_link,
2526 .put_link = proc_self_put_link,
2532 * These are the directory entries in the root directory of /proc
2533 * that properly belong to the /proc filesystem, as they describe
2534 * describe something that is process related.
2536 static const struct pid_entry proc_base_stuff[] = {
2537 NOD("self", S_IFLNK|S_IRWXUGO,
2538 &proc_self_inode_operations, NULL, {}),
2542 * Exceptional case: normally we are not allowed to unhash a busy
2543 * directory. In this case, however, we can do it - no aliasing problems
2544 * due to the way we treat inodes.
2546 static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
2548 struct inode *inode = dentry->d_inode;
2549 struct task_struct *task = get_proc_task(inode);
2551 put_task_struct(task);
2558 static const struct dentry_operations proc_base_dentry_operations =
2560 .d_revalidate = proc_base_revalidate,
2561 .d_delete = pid_delete_dentry,
2564 static struct dentry *proc_base_instantiate(struct inode *dir,
2565 struct dentry *dentry, struct task_struct *task, const void *ptr)
2567 const struct pid_entry *p = ptr;
2568 struct inode *inode;
2569 struct proc_inode *ei;
2570 struct dentry *error;
2572 /* Allocate the inode */
2573 error = ERR_PTR(-ENOMEM);
2574 inode = new_inode(dir->i_sb);
2578 /* Initialize the inode */
2580 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2583 * grab the reference to the task.
2585 ei->pid = get_task_pid(task, PIDTYPE_PID);
2589 inode->i_mode = p->mode;
2590 if (S_ISDIR(inode->i_mode))
2592 if (S_ISLNK(inode->i_mode))
2595 inode->i_op = p->iop;
2597 inode->i_fop = p->fop;
2599 dentry->d_op = &proc_base_dentry_operations;
2600 d_add(dentry, inode);
2609 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2611 struct dentry *error;
2612 struct task_struct *task = get_proc_task(dir);
2613 const struct pid_entry *p, *last;
2615 error = ERR_PTR(-ENOENT);
2620 /* Lookup the directory entry */
2621 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2622 for (p = proc_base_stuff; p <= last; p++) {
2623 if (p->len != dentry->d_name.len)
2625 if (!memcmp(dentry->d_name.name, p->name, p->len))
2631 error = proc_base_instantiate(dir, dentry, task, p);
2634 put_task_struct(task);
2639 static int proc_base_fill_cache(struct file *filp, void *dirent,
2640 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2642 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2643 proc_base_instantiate, task, p);
2646 #ifdef CONFIG_TASK_IO_ACCOUNTING
2647 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2649 struct task_io_accounting acct = task->ioac;
2650 unsigned long flags;
2652 if (whole && lock_task_sighand(task, &flags)) {
2653 struct task_struct *t = task;
2655 task_io_accounting_add(&acct, &task->signal->ioac);
2656 while_each_thread(task, t)
2657 task_io_accounting_add(&acct, &t->ioac);
2659 unlock_task_sighand(task, &flags);
2661 return sprintf(buffer,
2666 "read_bytes: %llu\n"
2667 "write_bytes: %llu\n"
2668 "cancelled_write_bytes: %llu\n",
2669 (unsigned long long)acct.rchar,
2670 (unsigned long long)acct.wchar,
2671 (unsigned long long)acct.syscr,
2672 (unsigned long long)acct.syscw,
2673 (unsigned long long)acct.read_bytes,
2674 (unsigned long long)acct.write_bytes,
2675 (unsigned long long)acct.cancelled_write_bytes);
2678 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2680 return do_io_accounting(task, buffer, 0);
2683 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2685 return do_io_accounting(task, buffer, 1);
2687 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2689 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2690 struct pid *pid, struct task_struct *task)
2692 seq_printf(m, "%08x\n", task->personality);
2699 static const struct file_operations proc_task_operations;
2700 static const struct inode_operations proc_task_inode_operations;
2702 static const struct pid_entry tgid_base_stuff[] = {
2703 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2704 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2705 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2707 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2709 REG("environ", S_IRUSR, proc_environ_operations),
2710 INF("auxv", S_IRUSR, proc_pid_auxv),
2711 ONE("status", S_IRUGO, proc_pid_status),
2712 ONE("personality", S_IRUSR, proc_pid_personality),
2713 INF("limits", S_IRUGO, proc_pid_limits),
2714 #ifdef CONFIG_SCHED_DEBUG
2715 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2717 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2718 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2719 INF("syscall", S_IRUSR, proc_pid_syscall),
2721 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2722 ONE("stat", S_IRUGO, proc_tgid_stat),
2723 ONE("statm", S_IRUGO, proc_pid_statm),
2724 REG("maps", S_IRUGO, proc_maps_operations),
2726 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2728 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2729 LNK("cwd", proc_cwd_link),
2730 LNK("root", proc_root_link),
2731 LNK("exe", proc_exe_link),
2732 REG("mounts", S_IRUGO, proc_mounts_operations),
2733 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2734 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2735 #ifdef CONFIG_PROC_PAGE_MONITOR
2736 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2737 REG("smaps", S_IRUGO, proc_smaps_operations),
2738 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2740 #ifdef CONFIG_SECURITY
2741 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2743 #ifdef CONFIG_KALLSYMS
2744 INF("wchan", S_IRUGO, proc_pid_wchan),
2746 #ifdef CONFIG_STACKTRACE
2747 ONE("stack", S_IRUSR, proc_pid_stack),
2749 #ifdef CONFIG_SCHEDSTATS
2750 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2752 #ifdef CONFIG_LATENCYTOP
2753 REG("latency", S_IRUGO, proc_lstats_operations),
2755 #ifdef CONFIG_PROC_PID_CPUSET
2756 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2758 #ifdef CONFIG_CGROUPS
2759 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2761 INF("oom_score", S_IRUGO, proc_oom_score),
2762 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2763 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2764 #ifdef CONFIG_AUDITSYSCALL
2765 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2766 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2768 #ifdef CONFIG_FAULT_INJECTION
2769 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2771 #ifdef CONFIG_ELF_CORE
2772 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2774 #ifdef CONFIG_TASK_IO_ACCOUNTING
2775 INF("io", S_IRUGO, proc_tgid_io_accounting),
2779 static int proc_tgid_base_readdir(struct file * filp,
2780 void * dirent, filldir_t filldir)
2782 return proc_pident_readdir(filp,dirent,filldir,
2783 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2786 static const struct file_operations proc_tgid_base_operations = {
2787 .read = generic_read_dir,
2788 .readdir = proc_tgid_base_readdir,
2789 .llseek = default_llseek,
2792 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2793 return proc_pident_lookup(dir, dentry,
2794 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2797 static const struct inode_operations proc_tgid_base_inode_operations = {
2798 .lookup = proc_tgid_base_lookup,
2799 .getattr = pid_getattr,
2800 .setattr = proc_setattr,
2803 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2805 struct dentry *dentry, *leader, *dir;
2806 char buf[PROC_NUMBUF];
2810 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2811 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2813 shrink_dcache_parent(dentry);
2819 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2820 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2825 name.len = strlen(name.name);
2826 dir = d_hash_and_lookup(leader, &name);
2828 goto out_put_leader;
2831 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2832 dentry = d_hash_and_lookup(dir, &name);
2834 shrink_dcache_parent(dentry);
2847 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2848 * @task: task that should be flushed.
2850 * When flushing dentries from proc, one needs to flush them from global
2851 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2852 * in. This call is supposed to do all of this job.
2854 * Looks in the dcache for
2856 * /proc/@tgid/task/@pid
2857 * if either directory is present flushes it and all of it'ts children
2860 * It is safe and reasonable to cache /proc entries for a task until
2861 * that task exits. After that they just clog up the dcache with
2862 * useless entries, possibly causing useful dcache entries to be
2863 * flushed instead. This routine is proved to flush those useless
2864 * dcache entries at process exit time.
2866 * NOTE: This routine is just an optimization so it does not guarantee
2867 * that no dcache entries will exist at process exit time it
2868 * just makes it very unlikely that any will persist.
2871 void proc_flush_task(struct task_struct *task)
2874 struct pid *pid, *tgid;
2877 pid = task_pid(task);
2878 tgid = task_tgid(task);
2880 for (i = 0; i <= pid->level; i++) {
2881 upid = &pid->numbers[i];
2882 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2883 tgid->numbers[i].nr);
2886 upid = &pid->numbers[pid->level];
2888 pid_ns_release_proc(upid->ns);
2891 static struct dentry *proc_pid_instantiate(struct inode *dir,
2892 struct dentry * dentry,
2893 struct task_struct *task, const void *ptr)
2895 struct dentry *error = ERR_PTR(-ENOENT);
2896 struct inode *inode;
2898 inode = proc_pid_make_inode(dir->i_sb, task);
2902 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2903 inode->i_op = &proc_tgid_base_inode_operations;
2904 inode->i_fop = &proc_tgid_base_operations;
2905 inode->i_flags|=S_IMMUTABLE;
2907 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
2908 ARRAY_SIZE(tgid_base_stuff));
2910 dentry->d_op = &pid_dentry_operations;
2912 d_add(dentry, inode);
2913 /* Close the race of the process dying before we return the dentry */
2914 if (pid_revalidate(dentry, NULL))
2920 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2922 struct dentry *result;
2923 struct task_struct *task;
2925 struct pid_namespace *ns;
2927 result = proc_base_lookup(dir, dentry);
2928 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2931 tgid = name_to_int(dentry);
2935 ns = dentry->d_sb->s_fs_info;
2937 task = find_task_by_pid_ns(tgid, ns);
2939 get_task_struct(task);
2944 result = proc_pid_instantiate(dir, dentry, task, NULL);
2945 put_task_struct(task);
2951 * Find the first task with tgid >= tgid
2956 struct task_struct *task;
2958 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2963 put_task_struct(iter.task);
2967 pid = find_ge_pid(iter.tgid, ns);
2969 iter.tgid = pid_nr_ns(pid, ns);
2970 iter.task = pid_task(pid, PIDTYPE_PID);
2971 /* What we to know is if the pid we have find is the
2972 * pid of a thread_group_leader. Testing for task
2973 * being a thread_group_leader is the obvious thing
2974 * todo but there is a window when it fails, due to
2975 * the pid transfer logic in de_thread.
2977 * So we perform the straight forward test of seeing
2978 * if the pid we have found is the pid of a thread
2979 * group leader, and don't worry if the task we have
2980 * found doesn't happen to be a thread group leader.
2981 * As we don't care in the case of readdir.
2983 if (!iter.task || !has_group_leader_pid(iter.task)) {
2987 get_task_struct(iter.task);
2993 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2995 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2996 struct tgid_iter iter)
2998 char name[PROC_NUMBUF];
2999 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
3000 return proc_fill_cache(filp, dirent, filldir, name, len,
3001 proc_pid_instantiate, iter.task, NULL);
3004 /* for the /proc/ directory itself, after non-process stuff has been done */
3005 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3007 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3008 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
3009 struct tgid_iter iter;
3010 struct pid_namespace *ns;
3015 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
3016 const struct pid_entry *p = &proc_base_stuff[nr];
3017 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
3021 ns = filp->f_dentry->d_sb->s_fs_info;
3023 iter.tgid = filp->f_pos - TGID_OFFSET;
3024 for (iter = next_tgid(ns, iter);
3026 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3027 filp->f_pos = iter.tgid + TGID_OFFSET;
3028 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
3029 put_task_struct(iter.task);
3033 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3035 put_task_struct(reaper);
3043 static const struct pid_entry tid_base_stuff[] = {
3044 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3045 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3046 REG("environ", S_IRUSR, proc_environ_operations),
3047 INF("auxv", S_IRUSR, proc_pid_auxv),
3048 ONE("status", S_IRUGO, proc_pid_status),
3049 ONE("personality", S_IRUSR, proc_pid_personality),
3050 INF("limits", S_IRUGO, proc_pid_limits),
3051 #ifdef CONFIG_SCHED_DEBUG
3052 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3054 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3055 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3056 INF("syscall", S_IRUSR, proc_pid_syscall),
3058 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3059 ONE("stat", S_IRUGO, proc_tid_stat),
3060 ONE("statm", S_IRUGO, proc_pid_statm),
3061 REG("maps", S_IRUGO, proc_maps_operations),
3063 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
3065 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3066 LNK("cwd", proc_cwd_link),
3067 LNK("root", proc_root_link),
3068 LNK("exe", proc_exe_link),
3069 REG("mounts", S_IRUGO, proc_mounts_operations),
3070 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3071 #ifdef CONFIG_PROC_PAGE_MONITOR
3072 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3073 REG("smaps", S_IRUGO, proc_smaps_operations),
3074 REG("pagemap", S_IRUSR, proc_pagemap_operations),
3076 #ifdef CONFIG_SECURITY
3077 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3079 #ifdef CONFIG_KALLSYMS
3080 INF("wchan", S_IRUGO, proc_pid_wchan),
3082 #ifdef CONFIG_STACKTRACE
3083 ONE("stack", S_IRUSR, proc_pid_stack),
3085 #ifdef CONFIG_SCHEDSTATS
3086 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3088 #ifdef CONFIG_LATENCYTOP
3089 REG("latency", S_IRUGO, proc_lstats_operations),
3091 #ifdef CONFIG_PROC_PID_CPUSET
3092 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3094 #ifdef CONFIG_CGROUPS
3095 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3097 INF("oom_score", S_IRUGO, proc_oom_score),
3098 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3099 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3100 #ifdef CONFIG_AUDITSYSCALL
3101 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3102 REG("sessionid", S_IRUSR, proc_sessionid_operations),
3104 #ifdef CONFIG_FAULT_INJECTION
3105 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3107 #ifdef CONFIG_TASK_IO_ACCOUNTING
3108 INF("io", S_IRUGO, proc_tid_io_accounting),
3112 static int proc_tid_base_readdir(struct file * filp,
3113 void * dirent, filldir_t filldir)
3115 return proc_pident_readdir(filp,dirent,filldir,
3116 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3119 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3120 return proc_pident_lookup(dir, dentry,
3121 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3124 static const struct file_operations proc_tid_base_operations = {
3125 .read = generic_read_dir,
3126 .readdir = proc_tid_base_readdir,
3127 .llseek = default_llseek,
3130 static const struct inode_operations proc_tid_base_inode_operations = {
3131 .lookup = proc_tid_base_lookup,
3132 .getattr = pid_getattr,
3133 .setattr = proc_setattr,
3136 static struct dentry *proc_task_instantiate(struct inode *dir,
3137 struct dentry *dentry, struct task_struct *task, const void *ptr)
3139 struct dentry *error = ERR_PTR(-ENOENT);
3140 struct inode *inode;
3141 inode = proc_pid_make_inode(dir->i_sb, task);
3145 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3146 inode->i_op = &proc_tid_base_inode_operations;
3147 inode->i_fop = &proc_tid_base_operations;
3148 inode->i_flags|=S_IMMUTABLE;
3150 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
3151 ARRAY_SIZE(tid_base_stuff));
3153 dentry->d_op = &pid_dentry_operations;
3155 d_add(dentry, inode);
3156 /* Close the race of the process dying before we return the dentry */
3157 if (pid_revalidate(dentry, NULL))
3163 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3165 struct dentry *result = ERR_PTR(-ENOENT);
3166 struct task_struct *task;
3167 struct task_struct *leader = get_proc_task(dir);
3169 struct pid_namespace *ns;
3174 tid = name_to_int(dentry);
3178 ns = dentry->d_sb->s_fs_info;
3180 task = find_task_by_pid_ns(tid, ns);
3182 get_task_struct(task);
3186 if (!same_thread_group(leader, task))
3189 result = proc_task_instantiate(dir, dentry, task, NULL);
3191 put_task_struct(task);
3193 put_task_struct(leader);
3199 * Find the first tid of a thread group to return to user space.
3201 * Usually this is just the thread group leader, but if the users
3202 * buffer was too small or there was a seek into the middle of the
3203 * directory we have more work todo.
3205 * In the case of a short read we start with find_task_by_pid.
3207 * In the case of a seek we start with the leader and walk nr
3210 static struct task_struct *first_tid(struct task_struct *leader,
3211 int tid, int nr, struct pid_namespace *ns)
3213 struct task_struct *pos;
3216 /* Attempt to start with the pid of a thread */
3217 if (tid && (nr > 0)) {
3218 pos = find_task_by_pid_ns(tid, ns);
3219 if (pos && (pos->group_leader == leader))
3223 /* If nr exceeds the number of threads there is nothing todo */
3225 if (nr && nr >= get_nr_threads(leader))
3228 /* If we haven't found our starting place yet start
3229 * with the leader and walk nr threads forward.
3231 for (pos = leader; nr > 0; --nr) {
3232 pos = next_thread(pos);
3233 if (pos == leader) {
3239 get_task_struct(pos);
3246 * Find the next thread in the thread list.
3247 * Return NULL if there is an error or no next thread.
3249 * The reference to the input task_struct is released.
3251 static struct task_struct *next_tid(struct task_struct *start)
3253 struct task_struct *pos = NULL;
3255 if (pid_alive(start)) {
3256 pos = next_thread(start);
3257 if (thread_group_leader(pos))
3260 get_task_struct(pos);
3263 put_task_struct(start);
3267 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3268 struct task_struct *task, int tid)
3270 char name[PROC_NUMBUF];
3271 int len = snprintf(name, sizeof(name), "%d", tid);
3272 return proc_fill_cache(filp, dirent, filldir, name, len,
3273 proc_task_instantiate, task, NULL);
3276 /* for the /proc/TGID/task/ directories */
3277 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3279 struct dentry *dentry = filp->f_path.dentry;
3280 struct inode *inode = dentry->d_inode;
3281 struct task_struct *leader = NULL;
3282 struct task_struct *task;
3283 int retval = -ENOENT;
3286 struct pid_namespace *ns;
3288 task = get_proc_task(inode);
3292 if (pid_alive(task)) {
3293 leader = task->group_leader;
3294 get_task_struct(leader);
3297 put_task_struct(task);
3302 switch ((unsigned long)filp->f_pos) {
3305 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3310 ino = parent_ino(dentry);
3311 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3317 /* f_version caches the tgid value that the last readdir call couldn't
3318 * return. lseek aka telldir automagically resets f_version to 0.
3320 ns = filp->f_dentry->d_sb->s_fs_info;
3321 tid = (int)filp->f_version;
3322 filp->f_version = 0;
3323 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3325 task = next_tid(task), filp->f_pos++) {
3326 tid = task_pid_nr_ns(task, ns);
3327 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3328 /* returning this tgid failed, save it as the first
3329 * pid for the next readir call */
3330 filp->f_version = (u64)tid;
3331 put_task_struct(task);
3336 put_task_struct(leader);
3341 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3343 struct inode *inode = dentry->d_inode;
3344 struct task_struct *p = get_proc_task(inode);
3345 generic_fillattr(inode, stat);
3348 stat->nlink += get_nr_threads(p);
3355 static const struct inode_operations proc_task_inode_operations = {
3356 .lookup = proc_task_lookup,
3357 .getattr = proc_task_getattr,
3358 .setattr = proc_setattr,
3361 static const struct file_operations proc_task_operations = {
3362 .read = generic_read_dir,
3363 .readdir = proc_task_readdir,
3364 .llseek = default_llseek,