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/printk.h>
77 #include <linux/cgroup.h>
78 #include <linux/cpuset.h>
79 #include <linux/audit.h>
80 #include <linux/poll.h>
81 #include <linux/nsproxy.h>
82 #include <linux/oom.h>
83 #include <linux/elf.h>
84 #include <linux/pid_namespace.h>
85 #include <linux/user_namespace.h>
86 #include <linux/fs_struct.h>
87 #include <linux/slab.h>
88 #include <linux/flex_array.h>
89 #include <linux/posix-timers.h>
90 #ifdef CONFIG_HARDWALL
91 #include <asm/hardwall.h>
93 #include <trace/events/oom.h>
98 * Implementing inode permission operations in /proc is almost
99 * certainly an error. Permission checks need to happen during
100 * each system call not at open time. The reason is that most of
101 * what we wish to check for permissions in /proc varies at runtime.
103 * The classic example of a problem is opening file descriptors
104 * in /proc for a task before it execs a suid executable.
111 const struct inode_operations *iop;
112 const struct file_operations *fop;
116 #define NOD(NAME, MODE, IOP, FOP, OP) { \
118 .len = sizeof(NAME) - 1, \
125 #define DIR(NAME, MODE, iops, fops) \
126 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
127 #define LNK(NAME, get_link) \
128 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
129 &proc_pid_link_inode_operations, NULL, \
130 { .proc_get_link = get_link } )
131 #define REG(NAME, MODE, fops) \
132 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
133 #define INF(NAME, MODE, read) \
134 NOD(NAME, (S_IFREG|(MODE)), \
135 NULL, &proc_info_file_operations, \
136 { .proc_read = read } )
137 #define ONE(NAME, MODE, show) \
138 NOD(NAME, (S_IFREG|(MODE)), \
139 NULL, &proc_single_file_operations, \
140 { .proc_show = show } )
143 * Count the number of hardlinks for the pid_entry table, excluding the .
146 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
153 for (i = 0; i < n; ++i) {
154 if (S_ISDIR(entries[i].mode))
161 static int get_task_root(struct task_struct *task, struct path *root)
163 int result = -ENOENT;
167 get_fs_root(task->fs, root);
174 static int proc_cwd_link(struct dentry *dentry, struct path *path)
176 struct task_struct *task = get_proc_task(dentry->d_inode);
177 int result = -ENOENT;
182 get_fs_pwd(task->fs, path);
186 put_task_struct(task);
191 static int proc_root_link(struct dentry *dentry, struct path *path)
193 struct task_struct *task = get_proc_task(dentry->d_inode);
194 int result = -ENOENT;
197 result = get_task_root(task, path);
198 put_task_struct(task);
203 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
207 struct mm_struct *mm = get_task_mm(task);
211 goto out_mm; /* Shh! No looking before we're done */
213 len = mm->arg_end - mm->arg_start;
218 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
220 // If the nul at the end of args has been overwritten, then
221 // assume application is using setproctitle(3).
222 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
223 len = strnlen(buffer, res);
227 len = mm->env_end - mm->env_start;
228 if (len > PAGE_SIZE - res)
229 len = PAGE_SIZE - res;
230 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
231 res = strnlen(buffer, res);
240 static int proc_pid_auxv(struct task_struct *task, char *buffer)
242 struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ);
243 int res = PTR_ERR(mm);
244 if (mm && !IS_ERR(mm)) {
245 unsigned int nwords = 0;
248 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
249 res = nwords * sizeof(mm->saved_auxv[0]);
252 memcpy(buffer, mm->saved_auxv, res);
259 #ifdef CONFIG_KALLSYMS
261 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
262 * Returns the resolved symbol. If that fails, simply return the address.
264 static int proc_pid_wchan(struct task_struct *task, char *buffer)
267 char symname[KSYM_NAME_LEN];
269 wchan = get_wchan(task);
271 if (lookup_symbol_name(wchan, symname) < 0)
272 if (!ptrace_may_access(task, PTRACE_MODE_READ))
275 return sprintf(buffer, "%lu", wchan);
277 return sprintf(buffer, "%s", symname);
279 #endif /* CONFIG_KALLSYMS */
281 static int lock_trace(struct task_struct *task)
283 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
286 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
287 mutex_unlock(&task->signal->cred_guard_mutex);
293 static void unlock_trace(struct task_struct *task)
295 mutex_unlock(&task->signal->cred_guard_mutex);
298 #ifdef CONFIG_STACKTRACE
300 #define MAX_STACK_TRACE_DEPTH 64
302 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
303 struct pid *pid, struct task_struct *task)
305 struct stack_trace trace;
306 unsigned long *entries;
310 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
314 trace.nr_entries = 0;
315 trace.max_entries = MAX_STACK_TRACE_DEPTH;
316 trace.entries = entries;
319 err = lock_trace(task);
321 save_stack_trace_tsk(task, &trace);
323 for (i = 0; i < trace.nr_entries; i++) {
324 seq_printf(m, "[<%pK>] %pS\n",
325 (void *)entries[i], (void *)entries[i]);
335 #ifdef CONFIG_SCHEDSTATS
337 * Provides /proc/PID/schedstat
339 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
341 return sprintf(buffer, "%llu %llu %lu\n",
342 (unsigned long long)task->se.sum_exec_runtime,
343 (unsigned long long)task->sched_info.run_delay,
344 task->sched_info.pcount);
348 #ifdef CONFIG_LATENCYTOP
349 static int lstats_show_proc(struct seq_file *m, void *v)
352 struct inode *inode = m->private;
353 struct task_struct *task = get_proc_task(inode);
357 seq_puts(m, "Latency Top version : v0.1\n");
358 for (i = 0; i < 32; i++) {
359 struct latency_record *lr = &task->latency_record[i];
360 if (lr->backtrace[0]) {
362 seq_printf(m, "%i %li %li",
363 lr->count, lr->time, lr->max);
364 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
365 unsigned long bt = lr->backtrace[q];
370 seq_printf(m, " %ps", (void *)bt);
376 put_task_struct(task);
380 static int lstats_open(struct inode *inode, struct file *file)
382 return single_open(file, lstats_show_proc, inode);
385 static ssize_t lstats_write(struct file *file, const char __user *buf,
386 size_t count, loff_t *offs)
388 struct task_struct *task = get_proc_task(file_inode(file));
392 clear_all_latency_tracing(task);
393 put_task_struct(task);
398 static const struct file_operations proc_lstats_operations = {
401 .write = lstats_write,
403 .release = single_release,
408 #ifdef CONFIG_CGROUPS
409 static int cgroup_open(struct inode *inode, struct file *file)
411 struct pid *pid = PROC_I(inode)->pid;
412 return single_open(file, proc_cgroup_show, pid);
415 static const struct file_operations proc_cgroup_operations = {
419 .release = single_release,
423 #ifdef CONFIG_PROC_PID_CPUSET
425 static int cpuset_open(struct inode *inode, struct file *file)
427 struct pid *pid = PROC_I(inode)->pid;
428 return single_open(file, proc_cpuset_show, pid);
431 static const struct file_operations proc_cpuset_operations = {
435 .release = single_release,
439 static int proc_oom_score(struct task_struct *task, char *buffer)
441 unsigned long totalpages = totalram_pages + total_swap_pages;
442 unsigned long points = 0;
444 read_lock(&tasklist_lock);
446 points = oom_badness(task, NULL, NULL, totalpages) *
448 read_unlock(&tasklist_lock);
449 return sprintf(buffer, "%lu\n", points);
457 static const struct limit_names lnames[RLIM_NLIMITS] = {
458 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
459 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
460 [RLIMIT_DATA] = {"Max data size", "bytes"},
461 [RLIMIT_STACK] = {"Max stack size", "bytes"},
462 [RLIMIT_CORE] = {"Max core file size", "bytes"},
463 [RLIMIT_RSS] = {"Max resident set", "bytes"},
464 [RLIMIT_NPROC] = {"Max processes", "processes"},
465 [RLIMIT_NOFILE] = {"Max open files", "files"},
466 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
467 [RLIMIT_AS] = {"Max address space", "bytes"},
468 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
469 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
470 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
471 [RLIMIT_NICE] = {"Max nice priority", NULL},
472 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
473 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
476 /* Display limits for a process */
477 static int proc_pid_limits(struct task_struct *task, char *buffer)
482 char *bufptr = buffer;
484 struct rlimit rlim[RLIM_NLIMITS];
486 if (!lock_task_sighand(task, &flags))
488 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
489 unlock_task_sighand(task, &flags);
492 * print the file header
494 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
495 "Limit", "Soft Limit", "Hard Limit", "Units");
497 for (i = 0; i < RLIM_NLIMITS; i++) {
498 if (rlim[i].rlim_cur == RLIM_INFINITY)
499 count += sprintf(&bufptr[count], "%-25s %-20s ",
500 lnames[i].name, "unlimited");
502 count += sprintf(&bufptr[count], "%-25s %-20lu ",
503 lnames[i].name, rlim[i].rlim_cur);
505 if (rlim[i].rlim_max == RLIM_INFINITY)
506 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
508 count += sprintf(&bufptr[count], "%-20lu ",
512 count += sprintf(&bufptr[count], "%-10s\n",
515 count += sprintf(&bufptr[count], "\n");
521 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
522 static int proc_pid_syscall(struct task_struct *task, char *buffer)
525 unsigned long args[6], sp, pc;
526 int res = lock_trace(task);
530 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
531 res = sprintf(buffer, "running\n");
533 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
535 res = sprintf(buffer,
536 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
538 args[0], args[1], args[2], args[3], args[4], args[5],
543 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
545 /************************************************************************/
546 /* Here the fs part begins */
547 /************************************************************************/
549 /* permission checks */
550 static int proc_fd_access_allowed(struct inode *inode)
552 struct task_struct *task;
554 /* Allow access to a task's file descriptors if it is us or we
555 * may use ptrace attach to the process and find out that
558 task = get_proc_task(inode);
560 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
561 put_task_struct(task);
566 int proc_setattr(struct dentry *dentry, struct iattr *attr)
569 struct inode *inode = dentry->d_inode;
571 if (attr->ia_valid & ATTR_MODE)
574 error = inode_change_ok(inode, attr);
578 setattr_copy(inode, attr);
579 mark_inode_dirty(inode);
584 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
585 * or euid/egid (for hide_pid_min=2)?
587 static bool has_pid_permissions(struct pid_namespace *pid,
588 struct task_struct *task,
591 if (pid->hide_pid < hide_pid_min)
593 if (in_group_p(pid->pid_gid))
595 return ptrace_may_access(task, PTRACE_MODE_READ);
599 static int proc_pid_permission(struct inode *inode, int mask)
601 struct pid_namespace *pid = inode->i_sb->s_fs_info;
602 struct task_struct *task;
605 task = get_proc_task(inode);
608 has_perms = has_pid_permissions(pid, task, 1);
609 put_task_struct(task);
612 if (pid->hide_pid == 2) {
614 * Let's make getdents(), stat(), and open()
615 * consistent with each other. If a process
616 * may not stat() a file, it shouldn't be seen
624 return generic_permission(inode, mask);
629 static const struct inode_operations proc_def_inode_operations = {
630 .setattr = proc_setattr,
633 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
635 static ssize_t proc_info_read(struct file * file, char __user * buf,
636 size_t count, loff_t *ppos)
638 struct inode * inode = file_inode(file);
641 struct task_struct *task = get_proc_task(inode);
647 if (count > PROC_BLOCK_SIZE)
648 count = PROC_BLOCK_SIZE;
651 if (!(page = __get_free_page(GFP_TEMPORARY)))
654 length = PROC_I(inode)->op.proc_read(task, (char*)page);
657 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
660 put_task_struct(task);
665 static const struct file_operations proc_info_file_operations = {
666 .read = proc_info_read,
667 .llseek = generic_file_llseek,
670 static int proc_single_show(struct seq_file *m, void *v)
672 struct inode *inode = m->private;
673 struct pid_namespace *ns;
675 struct task_struct *task;
678 ns = inode->i_sb->s_fs_info;
679 pid = proc_pid(inode);
680 task = get_pid_task(pid, PIDTYPE_PID);
684 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
686 put_task_struct(task);
690 static int proc_single_open(struct inode *inode, struct file *filp)
692 return single_open(filp, proc_single_show, inode);
695 static const struct file_operations proc_single_file_operations = {
696 .open = proc_single_open,
699 .release = single_release,
702 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
704 struct task_struct *task = get_proc_task(file_inode(file));
705 struct mm_struct *mm;
710 mm = mm_access(task, mode);
711 put_task_struct(task);
717 /* ensure this mm_struct can't be freed */
718 atomic_inc(&mm->mm_count);
719 /* but do not pin its memory */
723 file->private_data = mm;
728 static int mem_open(struct inode *inode, struct file *file)
730 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
732 /* OK to pass negative loff_t, we can catch out-of-range */
733 file->f_mode |= FMODE_UNSIGNED_OFFSET;
738 static ssize_t mem_rw(struct file *file, char __user *buf,
739 size_t count, loff_t *ppos, int write)
741 struct mm_struct *mm = file->private_data;
742 unsigned long addr = *ppos;
749 page = (char *)__get_free_page(GFP_TEMPORARY);
754 if (!atomic_inc_not_zero(&mm->mm_users))
758 int this_len = min_t(int, count, PAGE_SIZE);
760 if (write && copy_from_user(page, buf, this_len)) {
765 this_len = access_remote_vm(mm, addr, page, this_len, write);
772 if (!write && copy_to_user(buf, page, this_len)) {
786 free_page((unsigned long) page);
790 static ssize_t mem_read(struct file *file, char __user *buf,
791 size_t count, loff_t *ppos)
793 return mem_rw(file, buf, count, ppos, 0);
796 static ssize_t mem_write(struct file *file, const char __user *buf,
797 size_t count, loff_t *ppos)
799 return mem_rw(file, (char __user*)buf, count, ppos, 1);
802 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
806 file->f_pos = offset;
809 file->f_pos += offset;
814 force_successful_syscall_return();
818 static int mem_release(struct inode *inode, struct file *file)
820 struct mm_struct *mm = file->private_data;
826 static const struct file_operations proc_mem_operations = {
831 .release = mem_release,
834 static int environ_open(struct inode *inode, struct file *file)
836 return __mem_open(inode, file, PTRACE_MODE_READ);
839 static ssize_t environ_read(struct file *file, char __user *buf,
840 size_t count, loff_t *ppos)
843 unsigned long src = *ppos;
845 struct mm_struct *mm = file->private_data;
850 page = (char *)__get_free_page(GFP_TEMPORARY);
855 if (!atomic_inc_not_zero(&mm->mm_users))
858 size_t this_len, max_len;
861 if (src >= (mm->env_end - mm->env_start))
864 this_len = mm->env_end - (mm->env_start + src);
866 max_len = min_t(size_t, PAGE_SIZE, count);
867 this_len = min(max_len, this_len);
869 retval = access_remote_vm(mm, (mm->env_start + src),
877 if (copy_to_user(buf, page, retval)) {
891 free_page((unsigned long) page);
895 static const struct file_operations proc_environ_operations = {
896 .open = environ_open,
897 .read = environ_read,
898 .llseek = generic_file_llseek,
899 .release = mem_release,
902 static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
905 struct task_struct *task = get_proc_task(file_inode(file));
906 char buffer[PROC_NUMBUF];
907 int oom_adj = OOM_ADJUST_MIN;
913 if (lock_task_sighand(task, &flags)) {
914 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
915 oom_adj = OOM_ADJUST_MAX;
917 oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
919 unlock_task_sighand(task, &flags);
921 put_task_struct(task);
922 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
923 return simple_read_from_buffer(buf, count, ppos, buffer, len);
926 static ssize_t oom_adj_write(struct file *file, const char __user *buf,
927 size_t count, loff_t *ppos)
929 struct task_struct *task;
930 char buffer[PROC_NUMBUF];
935 memset(buffer, 0, sizeof(buffer));
936 if (count > sizeof(buffer) - 1)
937 count = sizeof(buffer) - 1;
938 if (copy_from_user(buffer, buf, count)) {
943 err = kstrtoint(strstrip(buffer), 0, &oom_adj);
946 if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
947 oom_adj != OOM_DISABLE) {
952 task = get_proc_task(file_inode(file));
964 if (!lock_task_sighand(task, &flags)) {
970 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
971 * value is always attainable.
973 if (oom_adj == OOM_ADJUST_MAX)
974 oom_adj = OOM_SCORE_ADJ_MAX;
976 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
978 if (oom_adj < task->signal->oom_score_adj &&
979 !capable(CAP_SYS_RESOURCE)) {
985 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
986 * /proc/pid/oom_score_adj instead.
988 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
989 current->comm, task_pid_nr(current), task_pid_nr(task),
992 task->signal->oom_score_adj = oom_adj;
993 trace_oom_score_adj_update(task);
995 unlock_task_sighand(task, &flags);
998 put_task_struct(task);
1000 return err < 0 ? err : count;
1003 static const struct file_operations proc_oom_adj_operations = {
1004 .read = oom_adj_read,
1005 .write = oom_adj_write,
1006 .llseek = generic_file_llseek,
1009 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1010 size_t count, loff_t *ppos)
1012 struct task_struct *task = get_proc_task(file_inode(file));
1013 char buffer[PROC_NUMBUF];
1014 short oom_score_adj = OOM_SCORE_ADJ_MIN;
1015 unsigned long flags;
1020 if (lock_task_sighand(task, &flags)) {
1021 oom_score_adj = task->signal->oom_score_adj;
1022 unlock_task_sighand(task, &flags);
1024 put_task_struct(task);
1025 len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
1026 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1029 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1030 size_t count, loff_t *ppos)
1032 struct task_struct *task;
1033 char buffer[PROC_NUMBUF];
1034 unsigned long flags;
1038 memset(buffer, 0, sizeof(buffer));
1039 if (count > sizeof(buffer) - 1)
1040 count = sizeof(buffer) - 1;
1041 if (copy_from_user(buffer, buf, count)) {
1046 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1049 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1050 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1055 task = get_proc_task(file_inode(file));
1067 if (!lock_task_sighand(task, &flags)) {
1072 if ((short)oom_score_adj < task->signal->oom_score_adj_min &&
1073 !capable(CAP_SYS_RESOURCE)) {
1078 task->signal->oom_score_adj = (short)oom_score_adj;
1079 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1080 task->signal->oom_score_adj_min = (short)oom_score_adj;
1081 trace_oom_score_adj_update(task);
1084 unlock_task_sighand(task, &flags);
1087 put_task_struct(task);
1089 return err < 0 ? err : count;
1092 static const struct file_operations proc_oom_score_adj_operations = {
1093 .read = oom_score_adj_read,
1094 .write = oom_score_adj_write,
1095 .llseek = default_llseek,
1098 #ifdef CONFIG_AUDITSYSCALL
1099 #define TMPBUFLEN 21
1100 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1101 size_t count, loff_t *ppos)
1103 struct inode * inode = file_inode(file);
1104 struct task_struct *task = get_proc_task(inode);
1106 char tmpbuf[TMPBUFLEN];
1110 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1111 from_kuid(file->f_cred->user_ns,
1112 audit_get_loginuid(task)));
1113 put_task_struct(task);
1114 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1117 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1118 size_t count, loff_t *ppos)
1120 struct inode * inode = file_inode(file);
1127 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1133 if (count >= PAGE_SIZE)
1134 count = PAGE_SIZE - 1;
1137 /* No partial writes. */
1140 page = (char*)__get_free_page(GFP_TEMPORARY);
1144 if (copy_from_user(page, buf, count))
1148 loginuid = simple_strtoul(page, &tmp, 10);
1154 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1155 if (!uid_valid(kloginuid)) {
1160 length = audit_set_loginuid(kloginuid);
1161 if (likely(length == 0))
1165 free_page((unsigned long) page);
1169 static const struct file_operations proc_loginuid_operations = {
1170 .read = proc_loginuid_read,
1171 .write = proc_loginuid_write,
1172 .llseek = generic_file_llseek,
1175 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1176 size_t count, loff_t *ppos)
1178 struct inode * inode = file_inode(file);
1179 struct task_struct *task = get_proc_task(inode);
1181 char tmpbuf[TMPBUFLEN];
1185 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1186 audit_get_sessionid(task));
1187 put_task_struct(task);
1188 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1191 static const struct file_operations proc_sessionid_operations = {
1192 .read = proc_sessionid_read,
1193 .llseek = generic_file_llseek,
1197 #ifdef CONFIG_FAULT_INJECTION
1198 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1199 size_t count, loff_t *ppos)
1201 struct task_struct *task = get_proc_task(file_inode(file));
1202 char buffer[PROC_NUMBUF];
1208 make_it_fail = task->make_it_fail;
1209 put_task_struct(task);
1211 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1213 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1216 static ssize_t proc_fault_inject_write(struct file * file,
1217 const char __user * buf, size_t count, loff_t *ppos)
1219 struct task_struct *task;
1220 char buffer[PROC_NUMBUF], *end;
1223 if (!capable(CAP_SYS_RESOURCE))
1225 memset(buffer, 0, sizeof(buffer));
1226 if (count > sizeof(buffer) - 1)
1227 count = sizeof(buffer) - 1;
1228 if (copy_from_user(buffer, buf, count))
1230 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1233 task = get_proc_task(file_inode(file));
1236 task->make_it_fail = make_it_fail;
1237 put_task_struct(task);
1242 static const struct file_operations proc_fault_inject_operations = {
1243 .read = proc_fault_inject_read,
1244 .write = proc_fault_inject_write,
1245 .llseek = generic_file_llseek,
1250 #ifdef CONFIG_SCHED_DEBUG
1252 * Print out various scheduling related per-task fields:
1254 static int sched_show(struct seq_file *m, void *v)
1256 struct inode *inode = m->private;
1257 struct task_struct *p;
1259 p = get_proc_task(inode);
1262 proc_sched_show_task(p, m);
1270 sched_write(struct file *file, const char __user *buf,
1271 size_t count, loff_t *offset)
1273 struct inode *inode = file_inode(file);
1274 struct task_struct *p;
1276 p = get_proc_task(inode);
1279 proc_sched_set_task(p);
1286 static int sched_open(struct inode *inode, struct file *filp)
1288 return single_open(filp, sched_show, inode);
1291 static const struct file_operations proc_pid_sched_operations = {
1294 .write = sched_write,
1295 .llseek = seq_lseek,
1296 .release = single_release,
1301 #ifdef CONFIG_SCHED_AUTOGROUP
1303 * Print out autogroup related information:
1305 static int sched_autogroup_show(struct seq_file *m, void *v)
1307 struct inode *inode = m->private;
1308 struct task_struct *p;
1310 p = get_proc_task(inode);
1313 proc_sched_autogroup_show_task(p, m);
1321 sched_autogroup_write(struct file *file, const char __user *buf,
1322 size_t count, loff_t *offset)
1324 struct inode *inode = file_inode(file);
1325 struct task_struct *p;
1326 char buffer[PROC_NUMBUF];
1330 memset(buffer, 0, sizeof(buffer));
1331 if (count > sizeof(buffer) - 1)
1332 count = sizeof(buffer) - 1;
1333 if (copy_from_user(buffer, buf, count))
1336 err = kstrtoint(strstrip(buffer), 0, &nice);
1340 p = get_proc_task(inode);
1344 err = proc_sched_autogroup_set_nice(p, nice);
1353 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1357 ret = single_open(filp, sched_autogroup_show, NULL);
1359 struct seq_file *m = filp->private_data;
1366 static const struct file_operations proc_pid_sched_autogroup_operations = {
1367 .open = sched_autogroup_open,
1369 .write = sched_autogroup_write,
1370 .llseek = seq_lseek,
1371 .release = single_release,
1374 #endif /* CONFIG_SCHED_AUTOGROUP */
1376 static ssize_t comm_write(struct file *file, const char __user *buf,
1377 size_t count, loff_t *offset)
1379 struct inode *inode = file_inode(file);
1380 struct task_struct *p;
1381 char buffer[TASK_COMM_LEN];
1382 const size_t maxlen = sizeof(buffer) - 1;
1384 memset(buffer, 0, sizeof(buffer));
1385 if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
1388 p = get_proc_task(inode);
1392 if (same_thread_group(current, p))
1393 set_task_comm(p, buffer);
1402 static int comm_show(struct seq_file *m, void *v)
1404 struct inode *inode = m->private;
1405 struct task_struct *p;
1407 p = get_proc_task(inode);
1412 seq_printf(m, "%s\n", p->comm);
1420 static int comm_open(struct inode *inode, struct file *filp)
1422 return single_open(filp, comm_show, inode);
1425 static const struct file_operations proc_pid_set_comm_operations = {
1428 .write = comm_write,
1429 .llseek = seq_lseek,
1430 .release = single_release,
1433 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1435 struct task_struct *task;
1436 struct mm_struct *mm;
1437 struct file *exe_file;
1439 task = get_proc_task(dentry->d_inode);
1442 mm = get_task_mm(task);
1443 put_task_struct(task);
1446 exe_file = get_mm_exe_file(mm);
1449 *exe_path = exe_file->f_path;
1450 path_get(&exe_file->f_path);
1457 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1459 struct inode *inode = dentry->d_inode;
1461 int error = -EACCES;
1463 /* Are we allowed to snoop on the tasks file descriptors? */
1464 if (!proc_fd_access_allowed(inode))
1467 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1471 nd_jump_link(nd, &path);
1474 return ERR_PTR(error);
1477 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1479 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1486 pathname = d_path(path, tmp, PAGE_SIZE);
1487 len = PTR_ERR(pathname);
1488 if (IS_ERR(pathname))
1490 len = tmp + PAGE_SIZE - 1 - pathname;
1494 if (copy_to_user(buffer, pathname, len))
1497 free_page((unsigned long)tmp);
1501 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1503 int error = -EACCES;
1504 struct inode *inode = dentry->d_inode;
1507 /* Are we allowed to snoop on the tasks file descriptors? */
1508 if (!proc_fd_access_allowed(inode))
1511 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1515 error = do_proc_readlink(&path, buffer, buflen);
1521 const struct inode_operations proc_pid_link_inode_operations = {
1522 .readlink = proc_pid_readlink,
1523 .follow_link = proc_pid_follow_link,
1524 .setattr = proc_setattr,
1528 /* building an inode */
1530 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1532 struct inode * inode;
1533 struct proc_inode *ei;
1534 const struct cred *cred;
1536 /* We need a new inode */
1538 inode = new_inode(sb);
1544 inode->i_ino = get_next_ino();
1545 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1546 inode->i_op = &proc_def_inode_operations;
1549 * grab the reference to task.
1551 ei->pid = get_task_pid(task, PIDTYPE_PID);
1555 if (task_dumpable(task)) {
1557 cred = __task_cred(task);
1558 inode->i_uid = cred->euid;
1559 inode->i_gid = cred->egid;
1562 security_task_to_inode(task, inode);
1572 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1574 struct inode *inode = dentry->d_inode;
1575 struct task_struct *task;
1576 const struct cred *cred;
1577 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
1579 generic_fillattr(inode, stat);
1582 stat->uid = GLOBAL_ROOT_UID;
1583 stat->gid = GLOBAL_ROOT_GID;
1584 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1586 if (!has_pid_permissions(pid, task, 2)) {
1589 * This doesn't prevent learning whether PID exists,
1590 * it only makes getattr() consistent with readdir().
1594 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1595 task_dumpable(task)) {
1596 cred = __task_cred(task);
1597 stat->uid = cred->euid;
1598 stat->gid = cred->egid;
1608 * Exceptional case: normally we are not allowed to unhash a busy
1609 * directory. In this case, however, we can do it - no aliasing problems
1610 * due to the way we treat inodes.
1612 * Rewrite the inode's ownerships here because the owning task may have
1613 * performed a setuid(), etc.
1615 * Before the /proc/pid/status file was created the only way to read
1616 * the effective uid of a /process was to stat /proc/pid. Reading
1617 * /proc/pid/status is slow enough that procps and other packages
1618 * kept stating /proc/pid. To keep the rules in /proc simple I have
1619 * made this apply to all per process world readable and executable
1622 int pid_revalidate(struct dentry *dentry, unsigned int flags)
1624 struct inode *inode;
1625 struct task_struct *task;
1626 const struct cred *cred;
1628 if (flags & LOOKUP_RCU)
1631 inode = dentry->d_inode;
1632 task = get_proc_task(inode);
1635 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1636 task_dumpable(task)) {
1638 cred = __task_cred(task);
1639 inode->i_uid = cred->euid;
1640 inode->i_gid = cred->egid;
1643 inode->i_uid = GLOBAL_ROOT_UID;
1644 inode->i_gid = GLOBAL_ROOT_GID;
1646 inode->i_mode &= ~(S_ISUID | S_ISGID);
1647 security_task_to_inode(task, inode);
1648 put_task_struct(task);
1655 int pid_delete_dentry(const struct dentry *dentry)
1657 /* Is the task we represent dead?
1658 * If so, then don't put the dentry on the lru list,
1659 * kill it immediately.
1661 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1664 const struct dentry_operations pid_dentry_operations =
1666 .d_revalidate = pid_revalidate,
1667 .d_delete = pid_delete_dentry,
1673 * Fill a directory entry.
1675 * If possible create the dcache entry and derive our inode number and
1676 * file type from dcache entry.
1678 * Since all of the proc inode numbers are dynamically generated, the inode
1679 * numbers do not exist until the inode is cache. This means creating the
1680 * the dcache entry in readdir is necessary to keep the inode numbers
1681 * reported by readdir in sync with the inode numbers reported
1684 bool proc_fill_cache(struct file *file, struct dir_context *ctx,
1685 const char *name, int len,
1686 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1688 struct dentry *child, *dir = file->f_path.dentry;
1689 struct inode *inode;
1692 unsigned type = DT_UNKNOWN;
1696 qname.hash = full_name_hash(name, len);
1698 child = d_lookup(dir, &qname);
1701 new = d_alloc(dir, &qname);
1703 child = instantiate(dir->d_inode, new, task, ptr);
1710 if (!child || IS_ERR(child) || !child->d_inode)
1711 goto end_instantiate;
1712 inode = child->d_inode;
1714 type = inode->i_mode >> 12;
1718 ino = find_inode_number(dir, &qname);
1721 return dir_emit(ctx, name, len, ino, type);
1724 #ifdef CONFIG_CHECKPOINT_RESTORE
1727 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1728 * which represent vma start and end addresses.
1730 static int dname_to_vma_addr(struct dentry *dentry,
1731 unsigned long *start, unsigned long *end)
1733 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1739 static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
1741 unsigned long vm_start, vm_end;
1742 bool exact_vma_exists = false;
1743 struct mm_struct *mm = NULL;
1744 struct task_struct *task;
1745 const struct cred *cred;
1746 struct inode *inode;
1749 if (flags & LOOKUP_RCU)
1752 if (!capable(CAP_SYS_ADMIN)) {
1757 inode = dentry->d_inode;
1758 task = get_proc_task(inode);
1762 mm = mm_access(task, PTRACE_MODE_READ);
1763 if (IS_ERR_OR_NULL(mm))
1766 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1767 down_read(&mm->mmap_sem);
1768 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1769 up_read(&mm->mmap_sem);
1774 if (exact_vma_exists) {
1775 if (task_dumpable(task)) {
1777 cred = __task_cred(task);
1778 inode->i_uid = cred->euid;
1779 inode->i_gid = cred->egid;
1782 inode->i_uid = GLOBAL_ROOT_UID;
1783 inode->i_gid = GLOBAL_ROOT_GID;
1785 security_task_to_inode(task, inode);
1790 put_task_struct(task);
1799 static const struct dentry_operations tid_map_files_dentry_operations = {
1800 .d_revalidate = map_files_d_revalidate,
1801 .d_delete = pid_delete_dentry,
1804 static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
1806 unsigned long vm_start, vm_end;
1807 struct vm_area_struct *vma;
1808 struct task_struct *task;
1809 struct mm_struct *mm;
1813 task = get_proc_task(dentry->d_inode);
1817 mm = get_task_mm(task);
1818 put_task_struct(task);
1822 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
1826 down_read(&mm->mmap_sem);
1827 vma = find_exact_vma(mm, vm_start, vm_end);
1828 if (vma && vma->vm_file) {
1829 *path = vma->vm_file->f_path;
1833 up_read(&mm->mmap_sem);
1841 struct map_files_info {
1844 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
1847 static struct dentry *
1848 proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
1849 struct task_struct *task, const void *ptr)
1851 fmode_t mode = (fmode_t)(unsigned long)ptr;
1852 struct proc_inode *ei;
1853 struct inode *inode;
1855 inode = proc_pid_make_inode(dir->i_sb, task);
1857 return ERR_PTR(-ENOENT);
1860 ei->op.proc_get_link = proc_map_files_get_link;
1862 inode->i_op = &proc_pid_link_inode_operations;
1864 inode->i_mode = S_IFLNK;
1866 if (mode & FMODE_READ)
1867 inode->i_mode |= S_IRUSR;
1868 if (mode & FMODE_WRITE)
1869 inode->i_mode |= S_IWUSR;
1871 d_set_d_op(dentry, &tid_map_files_dentry_operations);
1872 d_add(dentry, inode);
1877 static struct dentry *proc_map_files_lookup(struct inode *dir,
1878 struct dentry *dentry, unsigned int flags)
1880 unsigned long vm_start, vm_end;
1881 struct vm_area_struct *vma;
1882 struct task_struct *task;
1883 struct dentry *result;
1884 struct mm_struct *mm;
1886 result = ERR_PTR(-EPERM);
1887 if (!capable(CAP_SYS_ADMIN))
1890 result = ERR_PTR(-ENOENT);
1891 task = get_proc_task(dir);
1895 result = ERR_PTR(-EACCES);
1896 if (!ptrace_may_access(task, PTRACE_MODE_READ))
1899 result = ERR_PTR(-ENOENT);
1900 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
1903 mm = get_task_mm(task);
1907 down_read(&mm->mmap_sem);
1908 vma = find_exact_vma(mm, vm_start, vm_end);
1913 result = proc_map_files_instantiate(dir, dentry, task,
1914 (void *)(unsigned long)vma->vm_file->f_mode);
1917 up_read(&mm->mmap_sem);
1920 put_task_struct(task);
1925 static const struct inode_operations proc_map_files_inode_operations = {
1926 .lookup = proc_map_files_lookup,
1927 .permission = proc_fd_permission,
1928 .setattr = proc_setattr,
1932 proc_map_files_readdir(struct file *file, struct dir_context *ctx)
1934 struct vm_area_struct *vma;
1935 struct task_struct *task;
1936 struct mm_struct *mm;
1937 unsigned long nr_files, pos, i;
1938 struct flex_array *fa = NULL;
1939 struct map_files_info info;
1940 struct map_files_info *p;
1944 if (!capable(CAP_SYS_ADMIN))
1948 task = get_proc_task(file_inode(file));
1953 if (!ptrace_may_access(task, PTRACE_MODE_READ))
1957 if (!dir_emit_dots(file, ctx))
1960 mm = get_task_mm(task);
1963 down_read(&mm->mmap_sem);
1968 * We need two passes here:
1970 * 1) Collect vmas of mapped files with mmap_sem taken
1971 * 2) Release mmap_sem and instantiate entries
1973 * otherwise we get lockdep complained, since filldir()
1974 * routine might require mmap_sem taken in might_fault().
1977 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
1978 if (vma->vm_file && ++pos > ctx->pos)
1983 fa = flex_array_alloc(sizeof(info), nr_files,
1985 if (!fa || flex_array_prealloc(fa, 0, nr_files,
1989 flex_array_free(fa);
1990 up_read(&mm->mmap_sem);
1994 for (i = 0, vma = mm->mmap, pos = 2; vma;
1995 vma = vma->vm_next) {
1998 if (++pos <= ctx->pos)
2001 info.mode = vma->vm_file->f_mode;
2002 info.len = snprintf(info.name,
2003 sizeof(info.name), "%lx-%lx",
2004 vma->vm_start, vma->vm_end);
2005 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2009 up_read(&mm->mmap_sem);
2011 for (i = 0; i < nr_files; i++) {
2012 p = flex_array_get(fa, i);
2013 if (!proc_fill_cache(file, ctx,
2015 proc_map_files_instantiate,
2017 (void *)(unsigned long)p->mode))
2022 flex_array_free(fa);
2026 put_task_struct(task);
2031 static const struct file_operations proc_map_files_operations = {
2032 .read = generic_read_dir,
2033 .iterate = proc_map_files_readdir,
2034 .llseek = default_llseek,
2037 struct timers_private {
2039 struct task_struct *task;
2040 struct sighand_struct *sighand;
2041 struct pid_namespace *ns;
2042 unsigned long flags;
2045 static void *timers_start(struct seq_file *m, loff_t *pos)
2047 struct timers_private *tp = m->private;
2049 tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2051 return ERR_PTR(-ESRCH);
2053 tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2055 return ERR_PTR(-ESRCH);
2057 return seq_list_start(&tp->task->signal->posix_timers, *pos);
2060 static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2062 struct timers_private *tp = m->private;
2063 return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2066 static void timers_stop(struct seq_file *m, void *v)
2068 struct timers_private *tp = m->private;
2071 unlock_task_sighand(tp->task, &tp->flags);
2076 put_task_struct(tp->task);
2081 static int show_timer(struct seq_file *m, void *v)
2083 struct k_itimer *timer;
2084 struct timers_private *tp = m->private;
2086 static char *nstr[] = {
2087 [SIGEV_SIGNAL] = "signal",
2088 [SIGEV_NONE] = "none",
2089 [SIGEV_THREAD] = "thread",
2092 timer = list_entry((struct list_head *)v, struct k_itimer, list);
2093 notify = timer->it_sigev_notify;
2095 seq_printf(m, "ID: %d\n", timer->it_id);
2096 seq_printf(m, "signal: %d/%p\n", timer->sigq->info.si_signo,
2097 timer->sigq->info.si_value.sival_ptr);
2098 seq_printf(m, "notify: %s/%s.%d\n",
2099 nstr[notify & ~SIGEV_THREAD_ID],
2100 (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2101 pid_nr_ns(timer->it_pid, tp->ns));
2102 seq_printf(m, "ClockID: %d\n", timer->it_clock);
2107 static const struct seq_operations proc_timers_seq_ops = {
2108 .start = timers_start,
2109 .next = timers_next,
2110 .stop = timers_stop,
2114 static int proc_timers_open(struct inode *inode, struct file *file)
2116 struct timers_private *tp;
2118 tp = __seq_open_private(file, &proc_timers_seq_ops,
2119 sizeof(struct timers_private));
2123 tp->pid = proc_pid(inode);
2124 tp->ns = inode->i_sb->s_fs_info;
2128 static const struct file_operations proc_timers_operations = {
2129 .open = proc_timers_open,
2131 .llseek = seq_lseek,
2132 .release = seq_release_private,
2134 #endif /* CONFIG_CHECKPOINT_RESTORE */
2136 static struct dentry *proc_pident_instantiate(struct inode *dir,
2137 struct dentry *dentry, struct task_struct *task, const void *ptr)
2139 const struct pid_entry *p = ptr;
2140 struct inode *inode;
2141 struct proc_inode *ei;
2142 struct dentry *error = ERR_PTR(-ENOENT);
2144 inode = proc_pid_make_inode(dir->i_sb, task);
2149 inode->i_mode = p->mode;
2150 if (S_ISDIR(inode->i_mode))
2151 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2153 inode->i_op = p->iop;
2155 inode->i_fop = p->fop;
2157 d_set_d_op(dentry, &pid_dentry_operations);
2158 d_add(dentry, inode);
2159 /* Close the race of the process dying before we return the dentry */
2160 if (pid_revalidate(dentry, 0))
2166 static struct dentry *proc_pident_lookup(struct inode *dir,
2167 struct dentry *dentry,
2168 const struct pid_entry *ents,
2171 struct dentry *error;
2172 struct task_struct *task = get_proc_task(dir);
2173 const struct pid_entry *p, *last;
2175 error = ERR_PTR(-ENOENT);
2181 * Yes, it does not scale. And it should not. Don't add
2182 * new entries into /proc/<tgid>/ without very good reasons.
2184 last = &ents[nents - 1];
2185 for (p = ents; p <= last; p++) {
2186 if (p->len != dentry->d_name.len)
2188 if (!memcmp(dentry->d_name.name, p->name, p->len))
2194 error = proc_pident_instantiate(dir, dentry, task, p);
2196 put_task_struct(task);
2201 static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
2202 const struct pid_entry *ents, unsigned int nents)
2204 struct task_struct *task = get_proc_task(file_inode(file));
2205 const struct pid_entry *p;
2210 if (!dir_emit_dots(file, ctx))
2213 if (ctx->pos >= nents + 2)
2216 for (p = ents + (ctx->pos - 2); p <= ents + nents - 1; p++) {
2217 if (!proc_fill_cache(file, ctx, p->name, p->len,
2218 proc_pident_instantiate, task, p))
2223 put_task_struct(task);
2227 #ifdef CONFIG_SECURITY
2228 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2229 size_t count, loff_t *ppos)
2231 struct inode * inode = file_inode(file);
2234 struct task_struct *task = get_proc_task(inode);
2239 length = security_getprocattr(task,
2240 (char*)file->f_path.dentry->d_name.name,
2242 put_task_struct(task);
2244 length = simple_read_from_buffer(buf, count, ppos, p, length);
2249 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2250 size_t count, loff_t *ppos)
2252 struct inode * inode = file_inode(file);
2255 struct task_struct *task = get_proc_task(inode);
2260 if (count > PAGE_SIZE)
2263 /* No partial writes. */
2269 page = (char*)__get_free_page(GFP_TEMPORARY);
2274 if (copy_from_user(page, buf, count))
2277 /* Guard against adverse ptrace interaction */
2278 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2282 length = security_setprocattr(task,
2283 (char*)file->f_path.dentry->d_name.name,
2284 (void*)page, count);
2285 mutex_unlock(&task->signal->cred_guard_mutex);
2287 free_page((unsigned long) page);
2289 put_task_struct(task);
2294 static const struct file_operations proc_pid_attr_operations = {
2295 .read = proc_pid_attr_read,
2296 .write = proc_pid_attr_write,
2297 .llseek = generic_file_llseek,
2300 static const struct pid_entry attr_dir_stuff[] = {
2301 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2302 REG("prev", S_IRUGO, proc_pid_attr_operations),
2303 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2304 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2305 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2306 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2309 static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
2311 return proc_pident_readdir(file, ctx,
2312 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2315 static const struct file_operations proc_attr_dir_operations = {
2316 .read = generic_read_dir,
2317 .iterate = proc_attr_dir_readdir,
2318 .llseek = default_llseek,
2321 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2322 struct dentry *dentry, unsigned int flags)
2324 return proc_pident_lookup(dir, dentry,
2325 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2328 static const struct inode_operations proc_attr_dir_inode_operations = {
2329 .lookup = proc_attr_dir_lookup,
2330 .getattr = pid_getattr,
2331 .setattr = proc_setattr,
2336 #ifdef CONFIG_ELF_CORE
2337 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2338 size_t count, loff_t *ppos)
2340 struct task_struct *task = get_proc_task(file_inode(file));
2341 struct mm_struct *mm;
2342 char buffer[PROC_NUMBUF];
2350 mm = get_task_mm(task);
2352 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2353 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2354 MMF_DUMP_FILTER_SHIFT));
2356 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2359 put_task_struct(task);
2364 static ssize_t proc_coredump_filter_write(struct file *file,
2365 const char __user *buf,
2369 struct task_struct *task;
2370 struct mm_struct *mm;
2371 char buffer[PROC_NUMBUF], *end;
2378 memset(buffer, 0, sizeof(buffer));
2379 if (count > sizeof(buffer) - 1)
2380 count = sizeof(buffer) - 1;
2381 if (copy_from_user(buffer, buf, count))
2385 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2388 if (end - buffer == 0)
2392 task = get_proc_task(file_inode(file));
2397 mm = get_task_mm(task);
2401 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2403 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2405 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2410 put_task_struct(task);
2415 static const struct file_operations proc_coredump_filter_operations = {
2416 .read = proc_coredump_filter_read,
2417 .write = proc_coredump_filter_write,
2418 .llseek = generic_file_llseek,
2422 #ifdef CONFIG_TASK_IO_ACCOUNTING
2423 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2425 struct task_io_accounting acct = task->ioac;
2426 unsigned long flags;
2429 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2433 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2438 if (whole && lock_task_sighand(task, &flags)) {
2439 struct task_struct *t = task;
2441 task_io_accounting_add(&acct, &task->signal->ioac);
2442 while_each_thread(task, t)
2443 task_io_accounting_add(&acct, &t->ioac);
2445 unlock_task_sighand(task, &flags);
2447 result = sprintf(buffer,
2452 "read_bytes: %llu\n"
2453 "write_bytes: %llu\n"
2454 "cancelled_write_bytes: %llu\n",
2455 (unsigned long long)acct.rchar,
2456 (unsigned long long)acct.wchar,
2457 (unsigned long long)acct.syscr,
2458 (unsigned long long)acct.syscw,
2459 (unsigned long long)acct.read_bytes,
2460 (unsigned long long)acct.write_bytes,
2461 (unsigned long long)acct.cancelled_write_bytes);
2463 mutex_unlock(&task->signal->cred_guard_mutex);
2467 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2469 return do_io_accounting(task, buffer, 0);
2472 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2474 return do_io_accounting(task, buffer, 1);
2476 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2478 #ifdef CONFIG_USER_NS
2479 static int proc_id_map_open(struct inode *inode, struct file *file,
2480 struct seq_operations *seq_ops)
2482 struct user_namespace *ns = NULL;
2483 struct task_struct *task;
2484 struct seq_file *seq;
2487 task = get_proc_task(inode);
2490 ns = get_user_ns(task_cred_xxx(task, user_ns));
2492 put_task_struct(task);
2497 ret = seq_open(file, seq_ops);
2501 seq = file->private_data;
2511 static int proc_id_map_release(struct inode *inode, struct file *file)
2513 struct seq_file *seq = file->private_data;
2514 struct user_namespace *ns = seq->private;
2516 return seq_release(inode, file);
2519 static int proc_uid_map_open(struct inode *inode, struct file *file)
2521 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2524 static int proc_gid_map_open(struct inode *inode, struct file *file)
2526 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2529 static int proc_projid_map_open(struct inode *inode, struct file *file)
2531 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2534 static const struct file_operations proc_uid_map_operations = {
2535 .open = proc_uid_map_open,
2536 .write = proc_uid_map_write,
2538 .llseek = seq_lseek,
2539 .release = proc_id_map_release,
2542 static const struct file_operations proc_gid_map_operations = {
2543 .open = proc_gid_map_open,
2544 .write = proc_gid_map_write,
2546 .llseek = seq_lseek,
2547 .release = proc_id_map_release,
2550 static const struct file_operations proc_projid_map_operations = {
2551 .open = proc_projid_map_open,
2552 .write = proc_projid_map_write,
2554 .llseek = seq_lseek,
2555 .release = proc_id_map_release,
2557 #endif /* CONFIG_USER_NS */
2559 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2560 struct pid *pid, struct task_struct *task)
2562 int err = lock_trace(task);
2564 seq_printf(m, "%08x\n", task->personality);
2573 static const struct file_operations proc_task_operations;
2574 static const struct inode_operations proc_task_inode_operations;
2576 static const struct pid_entry tgid_base_stuff[] = {
2577 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2578 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2579 #ifdef CONFIG_CHECKPOINT_RESTORE
2580 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
2582 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2583 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2585 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2587 REG("environ", S_IRUSR, proc_environ_operations),
2588 INF("auxv", S_IRUSR, proc_pid_auxv),
2589 ONE("status", S_IRUGO, proc_pid_status),
2590 ONE("personality", S_IRUGO, proc_pid_personality),
2591 INF("limits", S_IRUGO, proc_pid_limits),
2592 #ifdef CONFIG_SCHED_DEBUG
2593 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2595 #ifdef CONFIG_SCHED_AUTOGROUP
2596 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2598 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2599 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2600 INF("syscall", S_IRUGO, proc_pid_syscall),
2602 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2603 ONE("stat", S_IRUGO, proc_tgid_stat),
2604 ONE("statm", S_IRUGO, proc_pid_statm),
2605 REG("maps", S_IRUGO, proc_pid_maps_operations),
2607 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
2609 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2610 LNK("cwd", proc_cwd_link),
2611 LNK("root", proc_root_link),
2612 LNK("exe", proc_exe_link),
2613 REG("mounts", S_IRUGO, proc_mounts_operations),
2614 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2615 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2616 #ifdef CONFIG_PROC_PAGE_MONITOR
2617 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2618 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
2619 REG("pagemap", S_IRUGO, proc_pagemap_operations),
2621 #ifdef CONFIG_SECURITY
2622 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2624 #ifdef CONFIG_KALLSYMS
2625 INF("wchan", S_IRUGO, proc_pid_wchan),
2627 #ifdef CONFIG_STACKTRACE
2628 ONE("stack", S_IRUGO, proc_pid_stack),
2630 #ifdef CONFIG_SCHEDSTATS
2631 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2633 #ifdef CONFIG_LATENCYTOP
2634 REG("latency", S_IRUGO, proc_lstats_operations),
2636 #ifdef CONFIG_PROC_PID_CPUSET
2637 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2639 #ifdef CONFIG_CGROUPS
2640 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2642 INF("oom_score", S_IRUGO, proc_oom_score),
2643 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
2644 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2645 #ifdef CONFIG_AUDITSYSCALL
2646 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2647 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2649 #ifdef CONFIG_FAULT_INJECTION
2650 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2652 #ifdef CONFIG_ELF_CORE
2653 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2655 #ifdef CONFIG_TASK_IO_ACCOUNTING
2656 INF("io", S_IRUSR, proc_tgid_io_accounting),
2658 #ifdef CONFIG_HARDWALL
2659 INF("hardwall", S_IRUGO, proc_pid_hardwall),
2661 #ifdef CONFIG_USER_NS
2662 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
2663 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
2664 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
2666 #ifdef CONFIG_CHECKPOINT_RESTORE
2667 REG("timers", S_IRUGO, proc_timers_operations),
2671 static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
2673 return proc_pident_readdir(file, ctx,
2674 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2677 static const struct file_operations proc_tgid_base_operations = {
2678 .read = generic_read_dir,
2679 .iterate = proc_tgid_base_readdir,
2680 .llseek = default_llseek,
2683 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2685 return proc_pident_lookup(dir, dentry,
2686 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2689 static const struct inode_operations proc_tgid_base_inode_operations = {
2690 .lookup = proc_tgid_base_lookup,
2691 .getattr = pid_getattr,
2692 .setattr = proc_setattr,
2693 .permission = proc_pid_permission,
2696 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2698 struct dentry *dentry, *leader, *dir;
2699 char buf[PROC_NUMBUF];
2703 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2704 /* no ->d_hash() rejects on procfs */
2705 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2707 shrink_dcache_parent(dentry);
2713 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2714 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2719 name.len = strlen(name.name);
2720 dir = d_hash_and_lookup(leader, &name);
2722 goto out_put_leader;
2725 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2726 dentry = d_hash_and_lookup(dir, &name);
2728 shrink_dcache_parent(dentry);
2741 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2742 * @task: task that should be flushed.
2744 * When flushing dentries from proc, one needs to flush them from global
2745 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2746 * in. This call is supposed to do all of this job.
2748 * Looks in the dcache for
2750 * /proc/@tgid/task/@pid
2751 * if either directory is present flushes it and all of it'ts children
2754 * It is safe and reasonable to cache /proc entries for a task until
2755 * that task exits. After that they just clog up the dcache with
2756 * useless entries, possibly causing useful dcache entries to be
2757 * flushed instead. This routine is proved to flush those useless
2758 * dcache entries at process exit time.
2760 * NOTE: This routine is just an optimization so it does not guarantee
2761 * that no dcache entries will exist at process exit time it
2762 * just makes it very unlikely that any will persist.
2765 void proc_flush_task(struct task_struct *task)
2768 struct pid *pid, *tgid;
2771 pid = task_pid(task);
2772 tgid = task_tgid(task);
2774 for (i = 0; i <= pid->level; i++) {
2775 upid = &pid->numbers[i];
2776 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2777 tgid->numbers[i].nr);
2781 static struct dentry *proc_pid_instantiate(struct inode *dir,
2782 struct dentry * dentry,
2783 struct task_struct *task, const void *ptr)
2785 struct dentry *error = ERR_PTR(-ENOENT);
2786 struct inode *inode;
2788 inode = proc_pid_make_inode(dir->i_sb, task);
2792 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2793 inode->i_op = &proc_tgid_base_inode_operations;
2794 inode->i_fop = &proc_tgid_base_operations;
2795 inode->i_flags|=S_IMMUTABLE;
2797 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
2798 ARRAY_SIZE(tgid_base_stuff)));
2800 d_set_d_op(dentry, &pid_dentry_operations);
2802 d_add(dentry, inode);
2803 /* Close the race of the process dying before we return the dentry */
2804 if (pid_revalidate(dentry, 0))
2810 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
2812 struct dentry *result = NULL;
2813 struct task_struct *task;
2815 struct pid_namespace *ns;
2817 tgid = name_to_int(dentry);
2821 ns = dentry->d_sb->s_fs_info;
2823 task = find_task_by_pid_ns(tgid, ns);
2825 get_task_struct(task);
2830 result = proc_pid_instantiate(dir, dentry, task, NULL);
2831 put_task_struct(task);
2837 * Find the first task with tgid >= tgid
2842 struct task_struct *task;
2844 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2849 put_task_struct(iter.task);
2853 pid = find_ge_pid(iter.tgid, ns);
2855 iter.tgid = pid_nr_ns(pid, ns);
2856 iter.task = pid_task(pid, PIDTYPE_PID);
2857 /* What we to know is if the pid we have find is the
2858 * pid of a thread_group_leader. Testing for task
2859 * being a thread_group_leader is the obvious thing
2860 * todo but there is a window when it fails, due to
2861 * the pid transfer logic in de_thread.
2863 * So we perform the straight forward test of seeing
2864 * if the pid we have found is the pid of a thread
2865 * group leader, and don't worry if the task we have
2866 * found doesn't happen to be a thread group leader.
2867 * As we don't care in the case of readdir.
2869 if (!iter.task || !has_group_leader_pid(iter.task)) {
2873 get_task_struct(iter.task);
2879 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 1)
2881 /* for the /proc/ directory itself, after non-process stuff has been done */
2882 int proc_pid_readdir(struct file *file, struct dir_context *ctx)
2884 struct tgid_iter iter;
2885 struct pid_namespace *ns = file->f_dentry->d_sb->s_fs_info;
2886 loff_t pos = ctx->pos;
2888 if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
2891 if (pos == TGID_OFFSET - 1) {
2892 struct inode *inode = ns->proc_self->d_inode;
2893 if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
2897 iter.tgid = pos - TGID_OFFSET;
2900 for (iter = next_tgid(ns, iter);
2902 iter.tgid += 1, iter = next_tgid(ns, iter)) {
2903 char name[PROC_NUMBUF];
2905 if (!has_pid_permissions(ns, iter.task, 2))
2908 len = snprintf(name, sizeof(name), "%d", iter.tgid);
2909 ctx->pos = iter.tgid + TGID_OFFSET;
2910 if (!proc_fill_cache(file, ctx, name, len,
2911 proc_pid_instantiate, iter.task, NULL)) {
2912 put_task_struct(iter.task);
2916 ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
2923 static const struct pid_entry tid_base_stuff[] = {
2924 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2925 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2926 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2927 REG("environ", S_IRUSR, proc_environ_operations),
2928 INF("auxv", S_IRUSR, proc_pid_auxv),
2929 ONE("status", S_IRUGO, proc_pid_status),
2930 ONE("personality", S_IRUGO, proc_pid_personality),
2931 INF("limits", S_IRUGO, proc_pid_limits),
2932 #ifdef CONFIG_SCHED_DEBUG
2933 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2935 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2936 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2937 INF("syscall", S_IRUGO, proc_pid_syscall),
2939 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2940 ONE("stat", S_IRUGO, proc_tid_stat),
2941 ONE("statm", S_IRUGO, proc_pid_statm),
2942 REG("maps", S_IRUGO, proc_tid_maps_operations),
2943 #ifdef CONFIG_CHECKPOINT_RESTORE
2944 REG("children", S_IRUGO, proc_tid_children_operations),
2947 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
2949 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2950 LNK("cwd", proc_cwd_link),
2951 LNK("root", proc_root_link),
2952 LNK("exe", proc_exe_link),
2953 REG("mounts", S_IRUGO, proc_mounts_operations),
2954 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2955 #ifdef CONFIG_PROC_PAGE_MONITOR
2956 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2957 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
2958 REG("pagemap", S_IRUGO, proc_pagemap_operations),
2960 #ifdef CONFIG_SECURITY
2961 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2963 #ifdef CONFIG_KALLSYMS
2964 INF("wchan", S_IRUGO, proc_pid_wchan),
2966 #ifdef CONFIG_STACKTRACE
2967 ONE("stack", S_IRUGO, proc_pid_stack),
2969 #ifdef CONFIG_SCHEDSTATS
2970 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2972 #ifdef CONFIG_LATENCYTOP
2973 REG("latency", S_IRUGO, proc_lstats_operations),
2975 #ifdef CONFIG_PROC_PID_CPUSET
2976 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2978 #ifdef CONFIG_CGROUPS
2979 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2981 INF("oom_score", S_IRUGO, proc_oom_score),
2982 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
2983 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2984 #ifdef CONFIG_AUDITSYSCALL
2985 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2986 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2988 #ifdef CONFIG_FAULT_INJECTION
2989 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2991 #ifdef CONFIG_TASK_IO_ACCOUNTING
2992 INF("io", S_IRUSR, proc_tid_io_accounting),
2994 #ifdef CONFIG_HARDWALL
2995 INF("hardwall", S_IRUGO, proc_pid_hardwall),
2997 #ifdef CONFIG_USER_NS
2998 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
2999 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3000 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3004 static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
3006 return proc_pident_readdir(file, ctx,
3007 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3010 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3012 return proc_pident_lookup(dir, dentry,
3013 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3016 static const struct file_operations proc_tid_base_operations = {
3017 .read = generic_read_dir,
3018 .iterate = proc_tid_base_readdir,
3019 .llseek = default_llseek,
3022 static const struct inode_operations proc_tid_base_inode_operations = {
3023 .lookup = proc_tid_base_lookup,
3024 .getattr = pid_getattr,
3025 .setattr = proc_setattr,
3028 static struct dentry *proc_task_instantiate(struct inode *dir,
3029 struct dentry *dentry, struct task_struct *task, const void *ptr)
3031 struct dentry *error = ERR_PTR(-ENOENT);
3032 struct inode *inode;
3033 inode = proc_pid_make_inode(dir->i_sb, task);
3037 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3038 inode->i_op = &proc_tid_base_inode_operations;
3039 inode->i_fop = &proc_tid_base_operations;
3040 inode->i_flags|=S_IMMUTABLE;
3042 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3043 ARRAY_SIZE(tid_base_stuff)));
3045 d_set_d_op(dentry, &pid_dentry_operations);
3047 d_add(dentry, inode);
3048 /* Close the race of the process dying before we return the dentry */
3049 if (pid_revalidate(dentry, 0))
3055 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3057 struct dentry *result = ERR_PTR(-ENOENT);
3058 struct task_struct *task;
3059 struct task_struct *leader = get_proc_task(dir);
3061 struct pid_namespace *ns;
3066 tid = name_to_int(dentry);
3070 ns = dentry->d_sb->s_fs_info;
3072 task = find_task_by_pid_ns(tid, ns);
3074 get_task_struct(task);
3078 if (!same_thread_group(leader, task))
3081 result = proc_task_instantiate(dir, dentry, task, NULL);
3083 put_task_struct(task);
3085 put_task_struct(leader);
3091 * Find the first tid of a thread group to return to user space.
3093 * Usually this is just the thread group leader, but if the users
3094 * buffer was too small or there was a seek into the middle of the
3095 * directory we have more work todo.
3097 * In the case of a short read we start with find_task_by_pid.
3099 * In the case of a seek we start with the leader and walk nr
3102 static struct task_struct *first_tid(struct task_struct *leader,
3103 int tid, int nr, struct pid_namespace *ns)
3105 struct task_struct *pos;
3108 /* Attempt to start with the pid of a thread */
3109 if (tid && (nr > 0)) {
3110 pos = find_task_by_pid_ns(tid, ns);
3111 if (pos && (pos->group_leader == leader))
3115 /* If nr exceeds the number of threads there is nothing todo */
3117 if (nr && nr >= get_nr_threads(leader))
3120 /* If we haven't found our starting place yet start
3121 * with the leader and walk nr threads forward.
3123 for (pos = leader; nr > 0; --nr) {
3124 pos = next_thread(pos);
3125 if (pos == leader) {
3131 get_task_struct(pos);
3138 * Find the next thread in the thread list.
3139 * Return NULL if there is an error or no next thread.
3141 * The reference to the input task_struct is released.
3143 static struct task_struct *next_tid(struct task_struct *start)
3145 struct task_struct *pos = NULL;
3147 if (pid_alive(start)) {
3148 pos = next_thread(start);
3149 if (thread_group_leader(pos))
3152 get_task_struct(pos);
3155 put_task_struct(start);
3159 /* for the /proc/TGID/task/ directories */
3160 static int proc_task_readdir(struct file *file, struct dir_context *ctx)
3162 struct task_struct *leader = NULL;
3163 struct task_struct *task = get_proc_task(file_inode(file));
3164 struct pid_namespace *ns;
3170 if (pid_alive(task)) {
3171 leader = task->group_leader;
3172 get_task_struct(leader);
3175 put_task_struct(task);
3179 if (!dir_emit_dots(file, ctx))
3182 /* f_version caches the tgid value that the last readdir call couldn't
3183 * return. lseek aka telldir automagically resets f_version to 0.
3185 ns = file->f_dentry->d_sb->s_fs_info;
3186 tid = (int)file->f_version;
3187 file->f_version = 0;
3188 for (task = first_tid(leader, tid, ctx->pos - 2, ns);
3190 task = next_tid(task), ctx->pos++) {
3191 char name[PROC_NUMBUF];
3193 tid = task_pid_nr_ns(task, ns);
3194 len = snprintf(name, sizeof(name), "%d", tid);
3195 if (!proc_fill_cache(file, ctx, name, len,
3196 proc_task_instantiate, task, NULL)) {
3197 /* returning this tgid failed, save it as the first
3198 * pid for the next readir call */
3199 file->f_version = (u64)tid;
3200 put_task_struct(task);
3205 put_task_struct(leader);
3209 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3211 struct inode *inode = dentry->d_inode;
3212 struct task_struct *p = get_proc_task(inode);
3213 generic_fillattr(inode, stat);
3216 stat->nlink += get_nr_threads(p);
3223 static const struct inode_operations proc_task_inode_operations = {
3224 .lookup = proc_task_lookup,
3225 .getattr = proc_task_getattr,
3226 .setattr = proc_setattr,
3227 .permission = proc_pid_permission,
3230 static const struct file_operations proc_task_operations = {
3231 .read = generic_read_dir,
3232 .iterate = proc_task_readdir,
3233 .llseek = default_llseek,