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->signal->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->signal->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);
774 /* OK to pass negative loff_t, we can catch out-of-range */
775 file->f_mode |= FMODE_UNSIGNED_OFFSET;
779 static ssize_t mem_read(struct file * file, char __user * buf,
780 size_t count, loff_t *ppos)
782 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
784 unsigned long src = *ppos;
786 struct mm_struct *mm;
791 if (check_mem_permission(task))
795 page = (char *)__get_free_page(GFP_TEMPORARY);
801 mm = get_task_mm(task);
807 if (file->private_data != (void*)((long)current->self_exec_id))
813 int this_len, retval;
815 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
816 retval = access_process_vm(task, src, page, this_len, 0);
817 if (!retval || check_mem_permission(task)) {
823 if (copy_to_user(buf, page, retval)) {
838 free_page((unsigned long) page);
840 put_task_struct(task);
845 #define mem_write NULL
848 /* This is a security hazard */
849 static ssize_t mem_write(struct file * file, const char __user *buf,
850 size_t count, loff_t *ppos)
854 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
855 unsigned long dst = *ppos;
861 if (check_mem_permission(task))
865 page = (char *)__get_free_page(GFP_TEMPORARY);
871 int this_len, retval;
873 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
874 if (copy_from_user(page, buf, this_len)) {
878 retval = access_process_vm(task, dst, page, this_len, 1);
890 free_page((unsigned long) page);
892 put_task_struct(task);
898 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
902 file->f_pos = offset;
905 file->f_pos += offset;
910 force_successful_syscall_return();
914 static const struct file_operations proc_mem_operations = {
921 static ssize_t environ_read(struct file *file, char __user *buf,
922 size_t count, loff_t *ppos)
924 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
926 unsigned long src = *ppos;
928 struct mm_struct *mm;
933 if (!ptrace_may_access(task, PTRACE_MODE_READ))
937 page = (char *)__get_free_page(GFP_TEMPORARY);
943 mm = get_task_mm(task);
948 int this_len, retval, max_len;
950 this_len = mm->env_end - (mm->env_start + src);
955 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
956 this_len = (this_len > max_len) ? max_len : this_len;
958 retval = access_process_vm(task, (mm->env_start + src),
966 if (copy_to_user(buf, page, retval)) {
980 free_page((unsigned long) page);
982 put_task_struct(task);
987 static const struct file_operations proc_environ_operations = {
988 .read = environ_read,
989 .llseek = generic_file_llseek,
992 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
993 size_t count, loff_t *ppos)
995 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
996 char buffer[PROC_NUMBUF];
998 int oom_adjust = OOM_DISABLE;
1004 if (lock_task_sighand(task, &flags)) {
1005 oom_adjust = task->signal->oom_adj;
1006 unlock_task_sighand(task, &flags);
1009 put_task_struct(task);
1011 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
1013 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1016 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1017 size_t count, loff_t *ppos)
1019 struct task_struct *task;
1020 char buffer[PROC_NUMBUF];
1022 unsigned long flags;
1025 memset(buffer, 0, sizeof(buffer));
1026 if (count > sizeof(buffer) - 1)
1027 count = sizeof(buffer) - 1;
1028 if (copy_from_user(buffer, buf, count)) {
1033 err = strict_strtol(strstrip(buffer), 0, &oom_adjust);
1036 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
1037 oom_adjust != OOM_DISABLE) {
1042 task = get_proc_task(file->f_path.dentry->d_inode);
1054 if (!lock_task_sighand(task, &flags)) {
1059 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
1064 if (oom_adjust != task->signal->oom_adj) {
1065 if (oom_adjust == OOM_DISABLE)
1066 atomic_inc(&task->mm->oom_disable_count);
1067 if (task->signal->oom_adj == OOM_DISABLE)
1068 atomic_dec(&task->mm->oom_disable_count);
1072 * Warn that /proc/pid/oom_adj is deprecated, see
1073 * Documentation/feature-removal-schedule.txt.
1075 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, "
1076 "please use /proc/%d/oom_score_adj instead.\n",
1077 current->comm, task_pid_nr(current),
1078 task_pid_nr(task), task_pid_nr(task));
1079 task->signal->oom_adj = oom_adjust;
1081 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1082 * value is always attainable.
1084 if (task->signal->oom_adj == OOM_ADJUST_MAX)
1085 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
1087 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
1090 unlock_task_sighand(task, &flags);
1093 put_task_struct(task);
1095 return err < 0 ? err : count;
1098 static const struct file_operations proc_oom_adjust_operations = {
1099 .read = oom_adjust_read,
1100 .write = oom_adjust_write,
1101 .llseek = generic_file_llseek,
1104 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1105 size_t count, loff_t *ppos)
1107 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1108 char buffer[PROC_NUMBUF];
1109 int oom_score_adj = OOM_SCORE_ADJ_MIN;
1110 unsigned long flags;
1115 if (lock_task_sighand(task, &flags)) {
1116 oom_score_adj = task->signal->oom_score_adj;
1117 unlock_task_sighand(task, &flags);
1119 put_task_struct(task);
1120 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1121 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1124 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1125 size_t count, loff_t *ppos)
1127 struct task_struct *task;
1128 char buffer[PROC_NUMBUF];
1129 unsigned long flags;
1133 memset(buffer, 0, sizeof(buffer));
1134 if (count > sizeof(buffer) - 1)
1135 count = sizeof(buffer) - 1;
1136 if (copy_from_user(buffer, buf, count)) {
1141 err = strict_strtol(strstrip(buffer), 0, &oom_score_adj);
1144 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1145 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1150 task = get_proc_task(file->f_path.dentry->d_inode);
1162 if (!lock_task_sighand(task, &flags)) {
1167 if (oom_score_adj < task->signal->oom_score_adj &&
1168 !capable(CAP_SYS_RESOURCE)) {
1173 if (oom_score_adj != task->signal->oom_score_adj) {
1174 if (oom_score_adj == OOM_SCORE_ADJ_MIN)
1175 atomic_inc(&task->mm->oom_disable_count);
1176 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1177 atomic_dec(&task->mm->oom_disable_count);
1179 task->signal->oom_score_adj = oom_score_adj;
1181 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1182 * always attainable.
1184 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1185 task->signal->oom_adj = OOM_DISABLE;
1187 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1190 unlock_task_sighand(task, &flags);
1193 put_task_struct(task);
1195 return err < 0 ? err : count;
1198 static const struct file_operations proc_oom_score_adj_operations = {
1199 .read = oom_score_adj_read,
1200 .write = oom_score_adj_write,
1201 .llseek = default_llseek,
1204 #ifdef CONFIG_AUDITSYSCALL
1205 #define TMPBUFLEN 21
1206 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1207 size_t count, loff_t *ppos)
1209 struct inode * inode = file->f_path.dentry->d_inode;
1210 struct task_struct *task = get_proc_task(inode);
1212 char tmpbuf[TMPBUFLEN];
1216 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1217 audit_get_loginuid(task));
1218 put_task_struct(task);
1219 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1222 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1223 size_t count, loff_t *ppos)
1225 struct inode * inode = file->f_path.dentry->d_inode;
1230 if (!capable(CAP_AUDIT_CONTROL))
1234 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1240 if (count >= PAGE_SIZE)
1241 count = PAGE_SIZE - 1;
1244 /* No partial writes. */
1247 page = (char*)__get_free_page(GFP_TEMPORARY);
1251 if (copy_from_user(page, buf, count))
1255 loginuid = simple_strtoul(page, &tmp, 10);
1261 length = audit_set_loginuid(current, loginuid);
1262 if (likely(length == 0))
1266 free_page((unsigned long) page);
1270 static const struct file_operations proc_loginuid_operations = {
1271 .read = proc_loginuid_read,
1272 .write = proc_loginuid_write,
1273 .llseek = generic_file_llseek,
1276 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1277 size_t count, loff_t *ppos)
1279 struct inode * inode = file->f_path.dentry->d_inode;
1280 struct task_struct *task = get_proc_task(inode);
1282 char tmpbuf[TMPBUFLEN];
1286 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1287 audit_get_sessionid(task));
1288 put_task_struct(task);
1289 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1292 static const struct file_operations proc_sessionid_operations = {
1293 .read = proc_sessionid_read,
1294 .llseek = generic_file_llseek,
1298 #ifdef CONFIG_FAULT_INJECTION
1299 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1300 size_t count, loff_t *ppos)
1302 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1303 char buffer[PROC_NUMBUF];
1309 make_it_fail = task->make_it_fail;
1310 put_task_struct(task);
1312 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1314 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1317 static ssize_t proc_fault_inject_write(struct file * file,
1318 const char __user * buf, size_t count, loff_t *ppos)
1320 struct task_struct *task;
1321 char buffer[PROC_NUMBUF], *end;
1324 if (!capable(CAP_SYS_RESOURCE))
1326 memset(buffer, 0, sizeof(buffer));
1327 if (count > sizeof(buffer) - 1)
1328 count = sizeof(buffer) - 1;
1329 if (copy_from_user(buffer, buf, count))
1331 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1334 task = get_proc_task(file->f_dentry->d_inode);
1337 task->make_it_fail = make_it_fail;
1338 put_task_struct(task);
1343 static const struct file_operations proc_fault_inject_operations = {
1344 .read = proc_fault_inject_read,
1345 .write = proc_fault_inject_write,
1346 .llseek = generic_file_llseek,
1351 #ifdef CONFIG_SCHED_DEBUG
1353 * Print out various scheduling related per-task fields:
1355 static int sched_show(struct seq_file *m, void *v)
1357 struct inode *inode = m->private;
1358 struct task_struct *p;
1360 p = get_proc_task(inode);
1363 proc_sched_show_task(p, m);
1371 sched_write(struct file *file, const char __user *buf,
1372 size_t count, loff_t *offset)
1374 struct inode *inode = file->f_path.dentry->d_inode;
1375 struct task_struct *p;
1377 p = get_proc_task(inode);
1380 proc_sched_set_task(p);
1387 static int sched_open(struct inode *inode, struct file *filp)
1391 ret = single_open(filp, sched_show, NULL);
1393 struct seq_file *m = filp->private_data;
1400 static const struct file_operations proc_pid_sched_operations = {
1403 .write = sched_write,
1404 .llseek = seq_lseek,
1405 .release = single_release,
1410 static ssize_t comm_write(struct file *file, const char __user *buf,
1411 size_t count, loff_t *offset)
1413 struct inode *inode = file->f_path.dentry->d_inode;
1414 struct task_struct *p;
1415 char buffer[TASK_COMM_LEN];
1417 memset(buffer, 0, sizeof(buffer));
1418 if (count > sizeof(buffer) - 1)
1419 count = sizeof(buffer) - 1;
1420 if (copy_from_user(buffer, buf, count))
1423 p = get_proc_task(inode);
1427 if (same_thread_group(current, p))
1428 set_task_comm(p, buffer);
1437 static int comm_show(struct seq_file *m, void *v)
1439 struct inode *inode = m->private;
1440 struct task_struct *p;
1442 p = get_proc_task(inode);
1447 seq_printf(m, "%s\n", p->comm);
1455 static int comm_open(struct inode *inode, struct file *filp)
1459 ret = single_open(filp, comm_show, NULL);
1461 struct seq_file *m = filp->private_data;
1468 static const struct file_operations proc_pid_set_comm_operations = {
1471 .write = comm_write,
1472 .llseek = seq_lseek,
1473 .release = single_release,
1477 * We added or removed a vma mapping the executable. The vmas are only mapped
1478 * during exec and are not mapped with the mmap system call.
1479 * Callers must hold down_write() on the mm's mmap_sem for these
1481 void added_exe_file_vma(struct mm_struct *mm)
1483 mm->num_exe_file_vmas++;
1486 void removed_exe_file_vma(struct mm_struct *mm)
1488 mm->num_exe_file_vmas--;
1489 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
1491 mm->exe_file = NULL;
1496 void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1499 get_file(new_exe_file);
1502 mm->exe_file = new_exe_file;
1503 mm->num_exe_file_vmas = 0;
1506 struct file *get_mm_exe_file(struct mm_struct *mm)
1508 struct file *exe_file;
1510 /* We need mmap_sem to protect against races with removal of
1511 * VM_EXECUTABLE vmas */
1512 down_read(&mm->mmap_sem);
1513 exe_file = mm->exe_file;
1516 up_read(&mm->mmap_sem);
1520 void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
1522 /* It's safe to write the exe_file pointer without exe_file_lock because
1523 * this is called during fork when the task is not yet in /proc */
1524 newmm->exe_file = get_mm_exe_file(oldmm);
1527 static int proc_exe_link(struct inode *inode, struct path *exe_path)
1529 struct task_struct *task;
1530 struct mm_struct *mm;
1531 struct file *exe_file;
1533 task = get_proc_task(inode);
1536 mm = get_task_mm(task);
1537 put_task_struct(task);
1540 exe_file = get_mm_exe_file(mm);
1543 *exe_path = exe_file->f_path;
1544 path_get(&exe_file->f_path);
1551 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1553 struct inode *inode = dentry->d_inode;
1554 int error = -EACCES;
1556 /* We don't need a base pointer in the /proc filesystem */
1557 path_put(&nd->path);
1559 /* Are we allowed to snoop on the tasks file descriptors? */
1560 if (!proc_fd_access_allowed(inode))
1563 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1565 return ERR_PTR(error);
1568 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1570 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1577 pathname = d_path(path, tmp, PAGE_SIZE);
1578 len = PTR_ERR(pathname);
1579 if (IS_ERR(pathname))
1581 len = tmp + PAGE_SIZE - 1 - pathname;
1585 if (copy_to_user(buffer, pathname, len))
1588 free_page((unsigned long)tmp);
1592 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1594 int error = -EACCES;
1595 struct inode *inode = dentry->d_inode;
1598 /* Are we allowed to snoop on the tasks file descriptors? */
1599 if (!proc_fd_access_allowed(inode))
1602 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1606 error = do_proc_readlink(&path, buffer, buflen);
1612 static const struct inode_operations proc_pid_link_inode_operations = {
1613 .readlink = proc_pid_readlink,
1614 .follow_link = proc_pid_follow_link,
1615 .setattr = proc_setattr,
1619 /* building an inode */
1621 static int task_dumpable(struct task_struct *task)
1624 struct mm_struct *mm;
1629 dumpable = get_dumpable(mm);
1637 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1639 struct inode * inode;
1640 struct proc_inode *ei;
1641 const struct cred *cred;
1643 /* We need a new inode */
1645 inode = new_inode(sb);
1651 inode->i_ino = get_next_ino();
1652 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1653 inode->i_op = &proc_def_inode_operations;
1656 * grab the reference to task.
1658 ei->pid = get_task_pid(task, PIDTYPE_PID);
1662 if (task_dumpable(task)) {
1664 cred = __task_cred(task);
1665 inode->i_uid = cred->euid;
1666 inode->i_gid = cred->egid;
1669 security_task_to_inode(task, inode);
1679 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1681 struct inode *inode = dentry->d_inode;
1682 struct task_struct *task;
1683 const struct cred *cred;
1685 generic_fillattr(inode, stat);
1690 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1692 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1693 task_dumpable(task)) {
1694 cred = __task_cred(task);
1695 stat->uid = cred->euid;
1696 stat->gid = cred->egid;
1706 * Exceptional case: normally we are not allowed to unhash a busy
1707 * directory. In this case, however, we can do it - no aliasing problems
1708 * due to the way we treat inodes.
1710 * Rewrite the inode's ownerships here because the owning task may have
1711 * performed a setuid(), etc.
1713 * Before the /proc/pid/status file was created the only way to read
1714 * the effective uid of a /process was to stat /proc/pid. Reading
1715 * /proc/pid/status is slow enough that procps and other packages
1716 * kept stating /proc/pid. To keep the rules in /proc simple I have
1717 * made this apply to all per process world readable and executable
1720 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1722 struct inode *inode;
1723 struct task_struct *task;
1724 const struct cred *cred;
1726 if (nd && nd->flags & LOOKUP_RCU)
1729 inode = dentry->d_inode;
1730 task = get_proc_task(inode);
1733 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1734 task_dumpable(task)) {
1736 cred = __task_cred(task);
1737 inode->i_uid = cred->euid;
1738 inode->i_gid = cred->egid;
1744 inode->i_mode &= ~(S_ISUID | S_ISGID);
1745 security_task_to_inode(task, inode);
1746 put_task_struct(task);
1753 static int pid_delete_dentry(const struct dentry * dentry)
1755 /* Is the task we represent dead?
1756 * If so, then don't put the dentry on the lru list,
1757 * kill it immediately.
1759 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1762 static const struct dentry_operations pid_dentry_operations =
1764 .d_revalidate = pid_revalidate,
1765 .d_delete = pid_delete_dentry,
1770 typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1771 struct task_struct *, const void *);
1774 * Fill a directory entry.
1776 * If possible create the dcache entry and derive our inode number and
1777 * file type from dcache entry.
1779 * Since all of the proc inode numbers are dynamically generated, the inode
1780 * numbers do not exist until the inode is cache. This means creating the
1781 * the dcache entry in readdir is necessary to keep the inode numbers
1782 * reported by readdir in sync with the inode numbers reported
1785 static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1786 char *name, int len,
1787 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1789 struct dentry *child, *dir = filp->f_path.dentry;
1790 struct inode *inode;
1793 unsigned type = DT_UNKNOWN;
1797 qname.hash = full_name_hash(name, len);
1799 child = d_lookup(dir, &qname);
1802 new = d_alloc(dir, &qname);
1804 child = instantiate(dir->d_inode, new, task, ptr);
1811 if (!child || IS_ERR(child) || !child->d_inode)
1812 goto end_instantiate;
1813 inode = child->d_inode;
1816 type = inode->i_mode >> 12;
1821 ino = find_inode_number(dir, &qname);
1824 return filldir(dirent, name, len, filp->f_pos, ino, type);
1827 static unsigned name_to_int(struct dentry *dentry)
1829 const char *name = dentry->d_name.name;
1830 int len = dentry->d_name.len;
1833 if (len > 1 && *name == '0')
1836 unsigned c = *name++ - '0';
1839 if (n >= (~0U-9)/10)
1849 #define PROC_FDINFO_MAX 64
1851 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1853 struct task_struct *task = get_proc_task(inode);
1854 struct files_struct *files = NULL;
1856 int fd = proc_fd(inode);
1859 files = get_files_struct(task);
1860 put_task_struct(task);
1864 * We are not taking a ref to the file structure, so we must
1867 spin_lock(&files->file_lock);
1868 file = fcheck_files(files, fd);
1871 *path = file->f_path;
1872 path_get(&file->f_path);
1875 snprintf(info, PROC_FDINFO_MAX,
1878 (long long) file->f_pos,
1880 spin_unlock(&files->file_lock);
1881 put_files_struct(files);
1884 spin_unlock(&files->file_lock);
1885 put_files_struct(files);
1890 static int proc_fd_link(struct inode *inode, struct path *path)
1892 return proc_fd_info(inode, path, NULL);
1895 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1897 struct inode *inode;
1898 struct task_struct *task;
1900 struct files_struct *files;
1901 const struct cred *cred;
1903 if (nd && nd->flags & LOOKUP_RCU)
1906 inode = dentry->d_inode;
1907 task = get_proc_task(inode);
1908 fd = proc_fd(inode);
1911 files = get_files_struct(task);
1914 if (fcheck_files(files, fd)) {
1916 put_files_struct(files);
1917 if (task_dumpable(task)) {
1919 cred = __task_cred(task);
1920 inode->i_uid = cred->euid;
1921 inode->i_gid = cred->egid;
1927 inode->i_mode &= ~(S_ISUID | S_ISGID);
1928 security_task_to_inode(task, inode);
1929 put_task_struct(task);
1933 put_files_struct(files);
1935 put_task_struct(task);
1941 static const struct dentry_operations tid_fd_dentry_operations =
1943 .d_revalidate = tid_fd_revalidate,
1944 .d_delete = pid_delete_dentry,
1947 static struct dentry *proc_fd_instantiate(struct inode *dir,
1948 struct dentry *dentry, struct task_struct *task, const void *ptr)
1950 unsigned fd = *(const unsigned *)ptr;
1952 struct files_struct *files;
1953 struct inode *inode;
1954 struct proc_inode *ei;
1955 struct dentry *error = ERR_PTR(-ENOENT);
1957 inode = proc_pid_make_inode(dir->i_sb, task);
1962 files = get_files_struct(task);
1965 inode->i_mode = S_IFLNK;
1968 * We are not taking a ref to the file structure, so we must
1971 spin_lock(&files->file_lock);
1972 file = fcheck_files(files, fd);
1975 if (file->f_mode & FMODE_READ)
1976 inode->i_mode |= S_IRUSR | S_IXUSR;
1977 if (file->f_mode & FMODE_WRITE)
1978 inode->i_mode |= S_IWUSR | S_IXUSR;
1979 spin_unlock(&files->file_lock);
1980 put_files_struct(files);
1982 inode->i_op = &proc_pid_link_inode_operations;
1984 ei->op.proc_get_link = proc_fd_link;
1985 d_set_d_op(dentry, &tid_fd_dentry_operations);
1986 d_add(dentry, inode);
1987 /* Close the race of the process dying before we return the dentry */
1988 if (tid_fd_revalidate(dentry, NULL))
1994 spin_unlock(&files->file_lock);
1995 put_files_struct(files);
2001 static struct dentry *proc_lookupfd_common(struct inode *dir,
2002 struct dentry *dentry,
2003 instantiate_t instantiate)
2005 struct task_struct *task = get_proc_task(dir);
2006 unsigned fd = name_to_int(dentry);
2007 struct dentry *result = ERR_PTR(-ENOENT);
2014 result = instantiate(dir, dentry, task, &fd);
2016 put_task_struct(task);
2021 static int proc_readfd_common(struct file * filp, void * dirent,
2022 filldir_t filldir, instantiate_t instantiate)
2024 struct dentry *dentry = filp->f_path.dentry;
2025 struct inode *inode = dentry->d_inode;
2026 struct task_struct *p = get_proc_task(inode);
2027 unsigned int fd, ino;
2029 struct files_struct * files;
2039 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
2043 ino = parent_ino(dentry);
2044 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2048 files = get_files_struct(p);
2052 for (fd = filp->f_pos-2;
2053 fd < files_fdtable(files)->max_fds;
2054 fd++, filp->f_pos++) {
2055 char name[PROC_NUMBUF];
2058 if (!fcheck_files(files, fd))
2062 len = snprintf(name, sizeof(name), "%d", fd);
2063 if (proc_fill_cache(filp, dirent, filldir,
2064 name, len, instantiate,
2072 put_files_struct(files);
2080 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
2081 struct nameidata *nd)
2083 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
2086 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
2088 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
2091 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
2092 size_t len, loff_t *ppos)
2094 char tmp[PROC_FDINFO_MAX];
2095 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
2097 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
2101 static const struct file_operations proc_fdinfo_file_operations = {
2102 .open = nonseekable_open,
2103 .read = proc_fdinfo_read,
2104 .llseek = no_llseek,
2107 static const struct file_operations proc_fd_operations = {
2108 .read = generic_read_dir,
2109 .readdir = proc_readfd,
2110 .llseek = default_llseek,
2114 * /proc/pid/fd needs a special permission handler so that a process can still
2115 * access /proc/self/fd after it has executed a setuid().
2117 static int proc_fd_permission(struct inode *inode, int mask, unsigned int flags)
2121 if (flags & IPERM_FLAG_RCU)
2123 rv = generic_permission(inode, mask, flags, NULL);
2126 if (task_pid(current) == proc_pid(inode))
2132 * proc directories can do almost nothing..
2134 static const struct inode_operations proc_fd_inode_operations = {
2135 .lookup = proc_lookupfd,
2136 .permission = proc_fd_permission,
2137 .setattr = proc_setattr,
2140 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2141 struct dentry *dentry, struct task_struct *task, const void *ptr)
2143 unsigned fd = *(unsigned *)ptr;
2144 struct inode *inode;
2145 struct proc_inode *ei;
2146 struct dentry *error = ERR_PTR(-ENOENT);
2148 inode = proc_pid_make_inode(dir->i_sb, task);
2153 inode->i_mode = S_IFREG | S_IRUSR;
2154 inode->i_fop = &proc_fdinfo_file_operations;
2155 d_set_d_op(dentry, &tid_fd_dentry_operations);
2156 d_add(dentry, inode);
2157 /* Close the race of the process dying before we return the dentry */
2158 if (tid_fd_revalidate(dentry, NULL))
2165 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2166 struct dentry *dentry,
2167 struct nameidata *nd)
2169 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2172 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2174 return proc_readfd_common(filp, dirent, filldir,
2175 proc_fdinfo_instantiate);
2178 static const struct file_operations proc_fdinfo_operations = {
2179 .read = generic_read_dir,
2180 .readdir = proc_readfdinfo,
2181 .llseek = default_llseek,
2185 * proc directories can do almost nothing..
2187 static const struct inode_operations proc_fdinfo_inode_operations = {
2188 .lookup = proc_lookupfdinfo,
2189 .setattr = proc_setattr,
2193 static struct dentry *proc_pident_instantiate(struct inode *dir,
2194 struct dentry *dentry, struct task_struct *task, const void *ptr)
2196 const struct pid_entry *p = ptr;
2197 struct inode *inode;
2198 struct proc_inode *ei;
2199 struct dentry *error = ERR_PTR(-ENOENT);
2201 inode = proc_pid_make_inode(dir->i_sb, task);
2206 inode->i_mode = p->mode;
2207 if (S_ISDIR(inode->i_mode))
2208 inode->i_nlink = 2; /* Use getattr to fix if necessary */
2210 inode->i_op = p->iop;
2212 inode->i_fop = p->fop;
2214 d_set_d_op(dentry, &pid_dentry_operations);
2215 d_add(dentry, inode);
2216 /* Close the race of the process dying before we return the dentry */
2217 if (pid_revalidate(dentry, NULL))
2223 static struct dentry *proc_pident_lookup(struct inode *dir,
2224 struct dentry *dentry,
2225 const struct pid_entry *ents,
2228 struct dentry *error;
2229 struct task_struct *task = get_proc_task(dir);
2230 const struct pid_entry *p, *last;
2232 error = ERR_PTR(-ENOENT);
2238 * Yes, it does not scale. And it should not. Don't add
2239 * new entries into /proc/<tgid>/ without very good reasons.
2241 last = &ents[nents - 1];
2242 for (p = ents; p <= last; p++) {
2243 if (p->len != dentry->d_name.len)
2245 if (!memcmp(dentry->d_name.name, p->name, p->len))
2251 error = proc_pident_instantiate(dir, dentry, task, p);
2253 put_task_struct(task);
2258 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2259 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2261 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2262 proc_pident_instantiate, task, p);
2265 static int proc_pident_readdir(struct file *filp,
2266 void *dirent, filldir_t filldir,
2267 const struct pid_entry *ents, unsigned int nents)
2270 struct dentry *dentry = filp->f_path.dentry;
2271 struct inode *inode = dentry->d_inode;
2272 struct task_struct *task = get_proc_task(inode);
2273 const struct pid_entry *p, *last;
2286 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2292 ino = parent_ino(dentry);
2293 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2305 last = &ents[nents - 1];
2307 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2316 put_task_struct(task);
2321 #ifdef CONFIG_SECURITY
2322 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2323 size_t count, loff_t *ppos)
2325 struct inode * inode = file->f_path.dentry->d_inode;
2328 struct task_struct *task = get_proc_task(inode);
2333 length = security_getprocattr(task,
2334 (char*)file->f_path.dentry->d_name.name,
2336 put_task_struct(task);
2338 length = simple_read_from_buffer(buf, count, ppos, p, length);
2343 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2344 size_t count, loff_t *ppos)
2346 struct inode * inode = file->f_path.dentry->d_inode;
2349 struct task_struct *task = get_proc_task(inode);
2354 if (count > PAGE_SIZE)
2357 /* No partial writes. */
2363 page = (char*)__get_free_page(GFP_TEMPORARY);
2368 if (copy_from_user(page, buf, count))
2371 /* Guard against adverse ptrace interaction */
2372 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2376 length = security_setprocattr(task,
2377 (char*)file->f_path.dentry->d_name.name,
2378 (void*)page, count);
2379 mutex_unlock(&task->signal->cred_guard_mutex);
2381 free_page((unsigned long) page);
2383 put_task_struct(task);
2388 static const struct file_operations proc_pid_attr_operations = {
2389 .read = proc_pid_attr_read,
2390 .write = proc_pid_attr_write,
2391 .llseek = generic_file_llseek,
2394 static const struct pid_entry attr_dir_stuff[] = {
2395 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2396 REG("prev", S_IRUGO, proc_pid_attr_operations),
2397 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2398 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2399 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2400 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2403 static int proc_attr_dir_readdir(struct file * filp,
2404 void * dirent, filldir_t filldir)
2406 return proc_pident_readdir(filp,dirent,filldir,
2407 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2410 static const struct file_operations proc_attr_dir_operations = {
2411 .read = generic_read_dir,
2412 .readdir = proc_attr_dir_readdir,
2413 .llseek = default_llseek,
2416 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2417 struct dentry *dentry, struct nameidata *nd)
2419 return proc_pident_lookup(dir, dentry,
2420 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2423 static const struct inode_operations proc_attr_dir_inode_operations = {
2424 .lookup = proc_attr_dir_lookup,
2425 .getattr = pid_getattr,
2426 .setattr = proc_setattr,
2431 #ifdef CONFIG_ELF_CORE
2432 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2433 size_t count, loff_t *ppos)
2435 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2436 struct mm_struct *mm;
2437 char buffer[PROC_NUMBUF];
2445 mm = get_task_mm(task);
2447 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2448 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2449 MMF_DUMP_FILTER_SHIFT));
2451 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2454 put_task_struct(task);
2459 static ssize_t proc_coredump_filter_write(struct file *file,
2460 const char __user *buf,
2464 struct task_struct *task;
2465 struct mm_struct *mm;
2466 char buffer[PROC_NUMBUF], *end;
2473 memset(buffer, 0, sizeof(buffer));
2474 if (count > sizeof(buffer) - 1)
2475 count = sizeof(buffer) - 1;
2476 if (copy_from_user(buffer, buf, count))
2480 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2483 if (end - buffer == 0)
2487 task = get_proc_task(file->f_dentry->d_inode);
2492 mm = get_task_mm(task);
2496 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2498 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2500 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2505 put_task_struct(task);
2510 static const struct file_operations proc_coredump_filter_operations = {
2511 .read = proc_coredump_filter_read,
2512 .write = proc_coredump_filter_write,
2513 .llseek = generic_file_llseek,
2520 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2523 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2524 pid_t tgid = task_tgid_nr_ns(current, ns);
2525 char tmp[PROC_NUMBUF];
2528 sprintf(tmp, "%d", tgid);
2529 return vfs_readlink(dentry,buffer,buflen,tmp);
2532 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2534 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2535 pid_t tgid = task_tgid_nr_ns(current, ns);
2536 char *name = ERR_PTR(-ENOENT);
2540 name = ERR_PTR(-ENOMEM);
2542 sprintf(name, "%d", tgid);
2544 nd_set_link(nd, name);
2548 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2551 char *s = nd_get_link(nd);
2556 static const struct inode_operations proc_self_inode_operations = {
2557 .readlink = proc_self_readlink,
2558 .follow_link = proc_self_follow_link,
2559 .put_link = proc_self_put_link,
2565 * These are the directory entries in the root directory of /proc
2566 * that properly belong to the /proc filesystem, as they describe
2567 * describe something that is process related.
2569 static const struct pid_entry proc_base_stuff[] = {
2570 NOD("self", S_IFLNK|S_IRWXUGO,
2571 &proc_self_inode_operations, NULL, {}),
2575 * Exceptional case: normally we are not allowed to unhash a busy
2576 * directory. In this case, however, we can do it - no aliasing problems
2577 * due to the way we treat inodes.
2579 static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
2581 struct inode *inode;
2582 struct task_struct *task;
2584 if (nd->flags & LOOKUP_RCU)
2587 inode = dentry->d_inode;
2588 task = get_proc_task(inode);
2590 put_task_struct(task);
2597 static const struct dentry_operations proc_base_dentry_operations =
2599 .d_revalidate = proc_base_revalidate,
2600 .d_delete = pid_delete_dentry,
2603 static struct dentry *proc_base_instantiate(struct inode *dir,
2604 struct dentry *dentry, struct task_struct *task, const void *ptr)
2606 const struct pid_entry *p = ptr;
2607 struct inode *inode;
2608 struct proc_inode *ei;
2609 struct dentry *error;
2611 /* Allocate the inode */
2612 error = ERR_PTR(-ENOMEM);
2613 inode = new_inode(dir->i_sb);
2617 /* Initialize the inode */
2619 inode->i_ino = get_next_ino();
2620 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2623 * grab the reference to the task.
2625 ei->pid = get_task_pid(task, PIDTYPE_PID);
2629 inode->i_mode = p->mode;
2630 if (S_ISDIR(inode->i_mode))
2632 if (S_ISLNK(inode->i_mode))
2635 inode->i_op = p->iop;
2637 inode->i_fop = p->fop;
2639 d_set_d_op(dentry, &proc_base_dentry_operations);
2640 d_add(dentry, inode);
2649 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2651 struct dentry *error;
2652 struct task_struct *task = get_proc_task(dir);
2653 const struct pid_entry *p, *last;
2655 error = ERR_PTR(-ENOENT);
2660 /* Lookup the directory entry */
2661 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2662 for (p = proc_base_stuff; p <= last; p++) {
2663 if (p->len != dentry->d_name.len)
2665 if (!memcmp(dentry->d_name.name, p->name, p->len))
2671 error = proc_base_instantiate(dir, dentry, task, p);
2674 put_task_struct(task);
2679 static int proc_base_fill_cache(struct file *filp, void *dirent,
2680 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2682 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2683 proc_base_instantiate, task, p);
2686 #ifdef CONFIG_TASK_IO_ACCOUNTING
2687 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2689 struct task_io_accounting acct = task->ioac;
2690 unsigned long flags;
2692 if (whole && lock_task_sighand(task, &flags)) {
2693 struct task_struct *t = task;
2695 task_io_accounting_add(&acct, &task->signal->ioac);
2696 while_each_thread(task, t)
2697 task_io_accounting_add(&acct, &t->ioac);
2699 unlock_task_sighand(task, &flags);
2701 return sprintf(buffer,
2706 "read_bytes: %llu\n"
2707 "write_bytes: %llu\n"
2708 "cancelled_write_bytes: %llu\n",
2709 (unsigned long long)acct.rchar,
2710 (unsigned long long)acct.wchar,
2711 (unsigned long long)acct.syscr,
2712 (unsigned long long)acct.syscw,
2713 (unsigned long long)acct.read_bytes,
2714 (unsigned long long)acct.write_bytes,
2715 (unsigned long long)acct.cancelled_write_bytes);
2718 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2720 return do_io_accounting(task, buffer, 0);
2723 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2725 return do_io_accounting(task, buffer, 1);
2727 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2729 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2730 struct pid *pid, struct task_struct *task)
2732 seq_printf(m, "%08x\n", task->personality);
2739 static const struct file_operations proc_task_operations;
2740 static const struct inode_operations proc_task_inode_operations;
2742 static const struct pid_entry tgid_base_stuff[] = {
2743 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2744 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2745 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2747 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2749 REG("environ", S_IRUSR, proc_environ_operations),
2750 INF("auxv", S_IRUSR, proc_pid_auxv),
2751 ONE("status", S_IRUGO, proc_pid_status),
2752 ONE("personality", S_IRUSR, proc_pid_personality),
2753 INF("limits", S_IRUGO, proc_pid_limits),
2754 #ifdef CONFIG_SCHED_DEBUG
2755 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2757 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2758 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2759 INF("syscall", S_IRUSR, proc_pid_syscall),
2761 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2762 ONE("stat", S_IRUGO, proc_tgid_stat),
2763 ONE("statm", S_IRUGO, proc_pid_statm),
2764 REG("maps", S_IRUGO, proc_maps_operations),
2766 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2768 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2769 LNK("cwd", proc_cwd_link),
2770 LNK("root", proc_root_link),
2771 LNK("exe", proc_exe_link),
2772 REG("mounts", S_IRUGO, proc_mounts_operations),
2773 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2774 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2775 #ifdef CONFIG_PROC_PAGE_MONITOR
2776 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2777 REG("smaps", S_IRUGO, proc_smaps_operations),
2778 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2780 #ifdef CONFIG_SECURITY
2781 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2783 #ifdef CONFIG_KALLSYMS
2784 INF("wchan", S_IRUGO, proc_pid_wchan),
2786 #ifdef CONFIG_STACKTRACE
2787 ONE("stack", S_IRUSR, proc_pid_stack),
2789 #ifdef CONFIG_SCHEDSTATS
2790 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2792 #ifdef CONFIG_LATENCYTOP
2793 REG("latency", S_IRUGO, proc_lstats_operations),
2795 #ifdef CONFIG_PROC_PID_CPUSET
2796 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2798 #ifdef CONFIG_CGROUPS
2799 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2801 INF("oom_score", S_IRUGO, proc_oom_score),
2802 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2803 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2804 #ifdef CONFIG_AUDITSYSCALL
2805 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2806 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2808 #ifdef CONFIG_FAULT_INJECTION
2809 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2811 #ifdef CONFIG_ELF_CORE
2812 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2814 #ifdef CONFIG_TASK_IO_ACCOUNTING
2815 INF("io", S_IRUGO, proc_tgid_io_accounting),
2819 static int proc_tgid_base_readdir(struct file * filp,
2820 void * dirent, filldir_t filldir)
2822 return proc_pident_readdir(filp,dirent,filldir,
2823 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2826 static const struct file_operations proc_tgid_base_operations = {
2827 .read = generic_read_dir,
2828 .readdir = proc_tgid_base_readdir,
2829 .llseek = default_llseek,
2832 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2833 return proc_pident_lookup(dir, dentry,
2834 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2837 static const struct inode_operations proc_tgid_base_inode_operations = {
2838 .lookup = proc_tgid_base_lookup,
2839 .getattr = pid_getattr,
2840 .setattr = proc_setattr,
2843 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2845 struct dentry *dentry, *leader, *dir;
2846 char buf[PROC_NUMBUF];
2850 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2851 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2853 shrink_dcache_parent(dentry);
2859 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2860 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2865 name.len = strlen(name.name);
2866 dir = d_hash_and_lookup(leader, &name);
2868 goto out_put_leader;
2871 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2872 dentry = d_hash_and_lookup(dir, &name);
2874 shrink_dcache_parent(dentry);
2887 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2888 * @task: task that should be flushed.
2890 * When flushing dentries from proc, one needs to flush them from global
2891 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2892 * in. This call is supposed to do all of this job.
2894 * Looks in the dcache for
2896 * /proc/@tgid/task/@pid
2897 * if either directory is present flushes it and all of it'ts children
2900 * It is safe and reasonable to cache /proc entries for a task until
2901 * that task exits. After that they just clog up the dcache with
2902 * useless entries, possibly causing useful dcache entries to be
2903 * flushed instead. This routine is proved to flush those useless
2904 * dcache entries at process exit time.
2906 * NOTE: This routine is just an optimization so it does not guarantee
2907 * that no dcache entries will exist at process exit time it
2908 * just makes it very unlikely that any will persist.
2911 void proc_flush_task(struct task_struct *task)
2914 struct pid *pid, *tgid;
2917 pid = task_pid(task);
2918 tgid = task_tgid(task);
2920 for (i = 0; i <= pid->level; i++) {
2921 upid = &pid->numbers[i];
2922 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2923 tgid->numbers[i].nr);
2926 upid = &pid->numbers[pid->level];
2928 pid_ns_release_proc(upid->ns);
2931 static struct dentry *proc_pid_instantiate(struct inode *dir,
2932 struct dentry * dentry,
2933 struct task_struct *task, const void *ptr)
2935 struct dentry *error = ERR_PTR(-ENOENT);
2936 struct inode *inode;
2938 inode = proc_pid_make_inode(dir->i_sb, task);
2942 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2943 inode->i_op = &proc_tgid_base_inode_operations;
2944 inode->i_fop = &proc_tgid_base_operations;
2945 inode->i_flags|=S_IMMUTABLE;
2947 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
2948 ARRAY_SIZE(tgid_base_stuff));
2950 d_set_d_op(dentry, &pid_dentry_operations);
2952 d_add(dentry, inode);
2953 /* Close the race of the process dying before we return the dentry */
2954 if (pid_revalidate(dentry, NULL))
2960 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2962 struct dentry *result;
2963 struct task_struct *task;
2965 struct pid_namespace *ns;
2967 result = proc_base_lookup(dir, dentry);
2968 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2971 tgid = name_to_int(dentry);
2975 ns = dentry->d_sb->s_fs_info;
2977 task = find_task_by_pid_ns(tgid, ns);
2979 get_task_struct(task);
2984 result = proc_pid_instantiate(dir, dentry, task, NULL);
2985 put_task_struct(task);
2991 * Find the first task with tgid >= tgid
2996 struct task_struct *task;
2998 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3003 put_task_struct(iter.task);
3007 pid = find_ge_pid(iter.tgid, ns);
3009 iter.tgid = pid_nr_ns(pid, ns);
3010 iter.task = pid_task(pid, PIDTYPE_PID);
3011 /* What we to know is if the pid we have find is the
3012 * pid of a thread_group_leader. Testing for task
3013 * being a thread_group_leader is the obvious thing
3014 * todo but there is a window when it fails, due to
3015 * the pid transfer logic in de_thread.
3017 * So we perform the straight forward test of seeing
3018 * if the pid we have found is the pid of a thread
3019 * group leader, and don't worry if the task we have
3020 * found doesn't happen to be a thread group leader.
3021 * As we don't care in the case of readdir.
3023 if (!iter.task || !has_group_leader_pid(iter.task)) {
3027 get_task_struct(iter.task);
3033 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
3035 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3036 struct tgid_iter iter)
3038 char name[PROC_NUMBUF];
3039 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
3040 return proc_fill_cache(filp, dirent, filldir, name, len,
3041 proc_pid_instantiate, iter.task, NULL);
3044 /* for the /proc/ directory itself, after non-process stuff has been done */
3045 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3047 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3048 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
3049 struct tgid_iter iter;
3050 struct pid_namespace *ns;
3055 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
3056 const struct pid_entry *p = &proc_base_stuff[nr];
3057 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
3061 ns = filp->f_dentry->d_sb->s_fs_info;
3063 iter.tgid = filp->f_pos - TGID_OFFSET;
3064 for (iter = next_tgid(ns, iter);
3066 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3067 filp->f_pos = iter.tgid + TGID_OFFSET;
3068 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
3069 put_task_struct(iter.task);
3073 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3075 put_task_struct(reaper);
3083 static const struct pid_entry tid_base_stuff[] = {
3084 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3085 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3086 REG("environ", S_IRUSR, proc_environ_operations),
3087 INF("auxv", S_IRUSR, proc_pid_auxv),
3088 ONE("status", S_IRUGO, proc_pid_status),
3089 ONE("personality", S_IRUSR, proc_pid_personality),
3090 INF("limits", S_IRUGO, proc_pid_limits),
3091 #ifdef CONFIG_SCHED_DEBUG
3092 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3094 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3095 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3096 INF("syscall", S_IRUSR, proc_pid_syscall),
3098 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3099 ONE("stat", S_IRUGO, proc_tid_stat),
3100 ONE("statm", S_IRUGO, proc_pid_statm),
3101 REG("maps", S_IRUGO, proc_maps_operations),
3103 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
3105 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3106 LNK("cwd", proc_cwd_link),
3107 LNK("root", proc_root_link),
3108 LNK("exe", proc_exe_link),
3109 REG("mounts", S_IRUGO, proc_mounts_operations),
3110 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3111 #ifdef CONFIG_PROC_PAGE_MONITOR
3112 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3113 REG("smaps", S_IRUGO, proc_smaps_operations),
3114 REG("pagemap", S_IRUSR, proc_pagemap_operations),
3116 #ifdef CONFIG_SECURITY
3117 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3119 #ifdef CONFIG_KALLSYMS
3120 INF("wchan", S_IRUGO, proc_pid_wchan),
3122 #ifdef CONFIG_STACKTRACE
3123 ONE("stack", S_IRUSR, proc_pid_stack),
3125 #ifdef CONFIG_SCHEDSTATS
3126 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3128 #ifdef CONFIG_LATENCYTOP
3129 REG("latency", S_IRUGO, proc_lstats_operations),
3131 #ifdef CONFIG_PROC_PID_CPUSET
3132 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3134 #ifdef CONFIG_CGROUPS
3135 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3137 INF("oom_score", S_IRUGO, proc_oom_score),
3138 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3139 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3140 #ifdef CONFIG_AUDITSYSCALL
3141 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3142 REG("sessionid", S_IRUSR, proc_sessionid_operations),
3144 #ifdef CONFIG_FAULT_INJECTION
3145 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3147 #ifdef CONFIG_TASK_IO_ACCOUNTING
3148 INF("io", S_IRUGO, proc_tid_io_accounting),
3152 static int proc_tid_base_readdir(struct file * filp,
3153 void * dirent, filldir_t filldir)
3155 return proc_pident_readdir(filp,dirent,filldir,
3156 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3159 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3160 return proc_pident_lookup(dir, dentry,
3161 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3164 static const struct file_operations proc_tid_base_operations = {
3165 .read = generic_read_dir,
3166 .readdir = proc_tid_base_readdir,
3167 .llseek = default_llseek,
3170 static const struct inode_operations proc_tid_base_inode_operations = {
3171 .lookup = proc_tid_base_lookup,
3172 .getattr = pid_getattr,
3173 .setattr = proc_setattr,
3176 static struct dentry *proc_task_instantiate(struct inode *dir,
3177 struct dentry *dentry, struct task_struct *task, const void *ptr)
3179 struct dentry *error = ERR_PTR(-ENOENT);
3180 struct inode *inode;
3181 inode = proc_pid_make_inode(dir->i_sb, task);
3185 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3186 inode->i_op = &proc_tid_base_inode_operations;
3187 inode->i_fop = &proc_tid_base_operations;
3188 inode->i_flags|=S_IMMUTABLE;
3190 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
3191 ARRAY_SIZE(tid_base_stuff));
3193 d_set_d_op(dentry, &pid_dentry_operations);
3195 d_add(dentry, inode);
3196 /* Close the race of the process dying before we return the dentry */
3197 if (pid_revalidate(dentry, NULL))
3203 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3205 struct dentry *result = ERR_PTR(-ENOENT);
3206 struct task_struct *task;
3207 struct task_struct *leader = get_proc_task(dir);
3209 struct pid_namespace *ns;
3214 tid = name_to_int(dentry);
3218 ns = dentry->d_sb->s_fs_info;
3220 task = find_task_by_pid_ns(tid, ns);
3222 get_task_struct(task);
3226 if (!same_thread_group(leader, task))
3229 result = proc_task_instantiate(dir, dentry, task, NULL);
3231 put_task_struct(task);
3233 put_task_struct(leader);
3239 * Find the first tid of a thread group to return to user space.
3241 * Usually this is just the thread group leader, but if the users
3242 * buffer was too small or there was a seek into the middle of the
3243 * directory we have more work todo.
3245 * In the case of a short read we start with find_task_by_pid.
3247 * In the case of a seek we start with the leader and walk nr
3250 static struct task_struct *first_tid(struct task_struct *leader,
3251 int tid, int nr, struct pid_namespace *ns)
3253 struct task_struct *pos;
3256 /* Attempt to start with the pid of a thread */
3257 if (tid && (nr > 0)) {
3258 pos = find_task_by_pid_ns(tid, ns);
3259 if (pos && (pos->group_leader == leader))
3263 /* If nr exceeds the number of threads there is nothing todo */
3265 if (nr && nr >= get_nr_threads(leader))
3268 /* If we haven't found our starting place yet start
3269 * with the leader and walk nr threads forward.
3271 for (pos = leader; nr > 0; --nr) {
3272 pos = next_thread(pos);
3273 if (pos == leader) {
3279 get_task_struct(pos);
3286 * Find the next thread in the thread list.
3287 * Return NULL if there is an error or no next thread.
3289 * The reference to the input task_struct is released.
3291 static struct task_struct *next_tid(struct task_struct *start)
3293 struct task_struct *pos = NULL;
3295 if (pid_alive(start)) {
3296 pos = next_thread(start);
3297 if (thread_group_leader(pos))
3300 get_task_struct(pos);
3303 put_task_struct(start);
3307 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3308 struct task_struct *task, int tid)
3310 char name[PROC_NUMBUF];
3311 int len = snprintf(name, sizeof(name), "%d", tid);
3312 return proc_fill_cache(filp, dirent, filldir, name, len,
3313 proc_task_instantiate, task, NULL);
3316 /* for the /proc/TGID/task/ directories */
3317 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3319 struct dentry *dentry = filp->f_path.dentry;
3320 struct inode *inode = dentry->d_inode;
3321 struct task_struct *leader = NULL;
3322 struct task_struct *task;
3323 int retval = -ENOENT;
3326 struct pid_namespace *ns;
3328 task = get_proc_task(inode);
3332 if (pid_alive(task)) {
3333 leader = task->group_leader;
3334 get_task_struct(leader);
3337 put_task_struct(task);
3342 switch ((unsigned long)filp->f_pos) {
3345 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3350 ino = parent_ino(dentry);
3351 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3357 /* f_version caches the tgid value that the last readdir call couldn't
3358 * return. lseek aka telldir automagically resets f_version to 0.
3360 ns = filp->f_dentry->d_sb->s_fs_info;
3361 tid = (int)filp->f_version;
3362 filp->f_version = 0;
3363 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3365 task = next_tid(task), filp->f_pos++) {
3366 tid = task_pid_nr_ns(task, ns);
3367 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3368 /* returning this tgid failed, save it as the first
3369 * pid for the next readir call */
3370 filp->f_version = (u64)tid;
3371 put_task_struct(task);
3376 put_task_struct(leader);
3381 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3383 struct inode *inode = dentry->d_inode;
3384 struct task_struct *p = get_proc_task(inode);
3385 generic_fillattr(inode, stat);
3388 stat->nlink += get_nr_threads(p);
3395 static const struct inode_operations proc_task_inode_operations = {
3396 .lookup = proc_task_lookup,
3397 .getattr = proc_task_getattr,
3398 .setattr = proc_setattr,
3401 static const struct file_operations proc_task_operations = {
3402 .read = generic_read_dir,
3403 .readdir = proc_task_readdir,
3404 .llseek = default_llseek,