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 } )
142 /* ANDROID is for special files in /proc. */
143 #define ANDROID(NAME, MODE, OTYPE) \
144 NOD(NAME, (S_IFREG|(MODE)), \
145 &proc_##OTYPE##_inode_operations, \
146 &proc_##OTYPE##_operations, {})
149 * Count the number of hardlinks for the pid_entry table, excluding the .
152 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
159 for (i = 0; i < n; ++i) {
160 if (S_ISDIR(entries[i].mode))
167 static int get_task_root(struct task_struct *task, struct path *root)
169 int result = -ENOENT;
173 get_fs_root(task->fs, root);
180 static int proc_cwd_link(struct dentry *dentry, struct path *path)
182 struct task_struct *task = get_proc_task(dentry->d_inode);
183 int result = -ENOENT;
188 get_fs_pwd(task->fs, path);
192 put_task_struct(task);
197 static int proc_root_link(struct dentry *dentry, struct path *path)
199 struct task_struct *task = get_proc_task(dentry->d_inode);
200 int result = -ENOENT;
203 result = get_task_root(task, path);
204 put_task_struct(task);
209 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
213 struct mm_struct *mm = get_task_mm(task);
217 goto out_mm; /* Shh! No looking before we're done */
219 len = mm->arg_end - mm->arg_start;
224 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
226 // If the nul at the end of args has been overwritten, then
227 // assume application is using setproctitle(3).
228 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
229 len = strnlen(buffer, res);
233 len = mm->env_end - mm->env_start;
234 if (len > PAGE_SIZE - res)
235 len = PAGE_SIZE - res;
236 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
237 res = strnlen(buffer, res);
246 static int proc_pid_auxv(struct task_struct *task, char *buffer)
248 struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ);
249 int res = PTR_ERR(mm);
250 if (mm && !IS_ERR(mm)) {
251 unsigned int nwords = 0;
254 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
255 res = nwords * sizeof(mm->saved_auxv[0]);
258 memcpy(buffer, mm->saved_auxv, res);
265 #ifdef CONFIG_KALLSYMS
267 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
268 * Returns the resolved symbol. If that fails, simply return the address.
270 static int proc_pid_wchan(struct task_struct *task, char *buffer)
273 char symname[KSYM_NAME_LEN];
275 wchan = get_wchan(task);
277 if (lookup_symbol_name(wchan, symname) < 0)
278 if (!ptrace_may_access(task, PTRACE_MODE_READ))
281 return sprintf(buffer, "%lu", wchan);
283 return sprintf(buffer, "%s", symname);
285 #endif /* CONFIG_KALLSYMS */
287 static int lock_trace(struct task_struct *task)
289 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
292 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
293 mutex_unlock(&task->signal->cred_guard_mutex);
299 static void unlock_trace(struct task_struct *task)
301 mutex_unlock(&task->signal->cred_guard_mutex);
304 #ifdef CONFIG_STACKTRACE
306 #define MAX_STACK_TRACE_DEPTH 64
308 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
309 struct pid *pid, struct task_struct *task)
311 struct stack_trace trace;
312 unsigned long *entries;
316 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
320 trace.nr_entries = 0;
321 trace.max_entries = MAX_STACK_TRACE_DEPTH;
322 trace.entries = entries;
325 err = lock_trace(task);
327 save_stack_trace_tsk(task, &trace);
329 for (i = 0; i < trace.nr_entries; i++) {
330 seq_printf(m, "[<%pK>] %pS\n",
331 (void *)entries[i], (void *)entries[i]);
341 #ifdef CONFIG_SCHEDSTATS
343 * Provides /proc/PID/schedstat
345 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
347 return sprintf(buffer, "%llu %llu %lu\n",
348 (unsigned long long)task->se.sum_exec_runtime,
349 (unsigned long long)task->sched_info.run_delay,
350 task->sched_info.pcount);
354 #ifdef CONFIG_LATENCYTOP
355 static int lstats_show_proc(struct seq_file *m, void *v)
358 struct inode *inode = m->private;
359 struct task_struct *task = get_proc_task(inode);
363 seq_puts(m, "Latency Top version : v0.1\n");
364 for (i = 0; i < 32; i++) {
365 struct latency_record *lr = &task->latency_record[i];
366 if (lr->backtrace[0]) {
368 seq_printf(m, "%i %li %li",
369 lr->count, lr->time, lr->max);
370 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
371 unsigned long bt = lr->backtrace[q];
376 seq_printf(m, " %ps", (void *)bt);
382 put_task_struct(task);
386 static int lstats_open(struct inode *inode, struct file *file)
388 return single_open(file, lstats_show_proc, inode);
391 static ssize_t lstats_write(struct file *file, const char __user *buf,
392 size_t count, loff_t *offs)
394 struct task_struct *task = get_proc_task(file_inode(file));
398 clear_all_latency_tracing(task);
399 put_task_struct(task);
404 static const struct file_operations proc_lstats_operations = {
407 .write = lstats_write,
409 .release = single_release,
414 #ifdef CONFIG_CGROUPS
415 static int cgroup_open(struct inode *inode, struct file *file)
417 struct pid *pid = PROC_I(inode)->pid;
418 return single_open(file, proc_cgroup_show, pid);
421 static const struct file_operations proc_cgroup_operations = {
425 .release = single_release,
429 #ifdef CONFIG_PROC_PID_CPUSET
431 static int cpuset_open(struct inode *inode, struct file *file)
433 struct pid *pid = PROC_I(inode)->pid;
434 return single_open(file, proc_cpuset_show, pid);
437 static const struct file_operations proc_cpuset_operations = {
441 .release = single_release,
445 static int proc_oom_score(struct task_struct *task, char *buffer)
447 unsigned long totalpages = totalram_pages + total_swap_pages;
448 unsigned long points = 0;
450 read_lock(&tasklist_lock);
452 points = oom_badness(task, NULL, NULL, totalpages) *
454 read_unlock(&tasklist_lock);
455 return sprintf(buffer, "%lu\n", points);
463 static const struct limit_names lnames[RLIM_NLIMITS] = {
464 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
465 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
466 [RLIMIT_DATA] = {"Max data size", "bytes"},
467 [RLIMIT_STACK] = {"Max stack size", "bytes"},
468 [RLIMIT_CORE] = {"Max core file size", "bytes"},
469 [RLIMIT_RSS] = {"Max resident set", "bytes"},
470 [RLIMIT_NPROC] = {"Max processes", "processes"},
471 [RLIMIT_NOFILE] = {"Max open files", "files"},
472 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
473 [RLIMIT_AS] = {"Max address space", "bytes"},
474 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
475 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
476 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
477 [RLIMIT_NICE] = {"Max nice priority", NULL},
478 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
479 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
482 /* Display limits for a process */
483 static int proc_pid_limits(struct task_struct *task, char *buffer)
488 char *bufptr = buffer;
490 struct rlimit rlim[RLIM_NLIMITS];
492 if (!lock_task_sighand(task, &flags))
494 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
495 unlock_task_sighand(task, &flags);
498 * print the file header
500 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
501 "Limit", "Soft Limit", "Hard Limit", "Units");
503 for (i = 0; i < RLIM_NLIMITS; i++) {
504 if (rlim[i].rlim_cur == RLIM_INFINITY)
505 count += sprintf(&bufptr[count], "%-25s %-20s ",
506 lnames[i].name, "unlimited");
508 count += sprintf(&bufptr[count], "%-25s %-20lu ",
509 lnames[i].name, rlim[i].rlim_cur);
511 if (rlim[i].rlim_max == RLIM_INFINITY)
512 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
514 count += sprintf(&bufptr[count], "%-20lu ",
518 count += sprintf(&bufptr[count], "%-10s\n",
521 count += sprintf(&bufptr[count], "\n");
527 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
528 static int proc_pid_syscall(struct task_struct *task, char *buffer)
531 unsigned long args[6], sp, pc;
532 int res = lock_trace(task);
536 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
537 res = sprintf(buffer, "running\n");
539 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
541 res = sprintf(buffer,
542 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
544 args[0], args[1], args[2], args[3], args[4], args[5],
549 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
551 /************************************************************************/
552 /* Here the fs part begins */
553 /************************************************************************/
555 /* permission checks */
556 static int proc_fd_access_allowed(struct inode *inode)
558 struct task_struct *task;
560 /* Allow access to a task's file descriptors if it is us or we
561 * may use ptrace attach to the process and find out that
564 task = get_proc_task(inode);
566 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
567 put_task_struct(task);
572 int proc_setattr(struct dentry *dentry, struct iattr *attr)
575 struct inode *inode = dentry->d_inode;
577 if (attr->ia_valid & ATTR_MODE)
580 error = inode_change_ok(inode, attr);
584 setattr_copy(inode, attr);
585 mark_inode_dirty(inode);
590 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
591 * or euid/egid (for hide_pid_min=2)?
593 static bool has_pid_permissions(struct pid_namespace *pid,
594 struct task_struct *task,
597 if (pid->hide_pid < hide_pid_min)
599 if (in_group_p(pid->pid_gid))
601 return ptrace_may_access(task, PTRACE_MODE_READ);
605 static int proc_pid_permission(struct inode *inode, int mask)
607 struct pid_namespace *pid = inode->i_sb->s_fs_info;
608 struct task_struct *task;
611 task = get_proc_task(inode);
614 has_perms = has_pid_permissions(pid, task, 1);
615 put_task_struct(task);
618 if (pid->hide_pid == 2) {
620 * Let's make getdents(), stat(), and open()
621 * consistent with each other. If a process
622 * may not stat() a file, it shouldn't be seen
630 return generic_permission(inode, mask);
635 static const struct inode_operations proc_def_inode_operations = {
636 .setattr = proc_setattr,
639 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
641 static ssize_t proc_info_read(struct file * file, char __user * buf,
642 size_t count, loff_t *ppos)
644 struct inode * inode = file_inode(file);
647 struct task_struct *task = get_proc_task(inode);
653 if (count > PROC_BLOCK_SIZE)
654 count = PROC_BLOCK_SIZE;
657 #ifndef CONFIG_SPRD_PAGERECORDER
658 if (!(page = __get_free_page(GFP_TEMPORARY)))
661 if (!(page = __get_free_page_nopagedebug(GFP_TEMPORARY)))
665 length = PROC_I(inode)->op.proc_read(task, (char*)page);
668 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
669 #ifndef CONFIG_SPRD_PAGERECORDER
672 free_page_nopagedebug(page);
675 put_task_struct(task);
680 static const struct file_operations proc_info_file_operations = {
681 .read = proc_info_read,
682 .llseek = generic_file_llseek,
685 static int proc_single_show(struct seq_file *m, void *v)
687 struct inode *inode = m->private;
688 struct pid_namespace *ns;
690 struct task_struct *task;
693 ns = inode->i_sb->s_fs_info;
694 pid = proc_pid(inode);
695 task = get_pid_task(pid, PIDTYPE_PID);
699 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
701 put_task_struct(task);
705 static int proc_single_open(struct inode *inode, struct file *filp)
707 return single_open(filp, proc_single_show, inode);
710 static const struct file_operations proc_single_file_operations = {
711 .open = proc_single_open,
714 .release = single_release,
717 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
719 struct task_struct *task = get_proc_task(file_inode(file));
720 struct mm_struct *mm;
725 mm = mm_access(task, mode);
726 put_task_struct(task);
732 /* ensure this mm_struct can't be freed */
733 atomic_inc(&mm->mm_count);
734 /* but do not pin its memory */
738 file->private_data = mm;
743 static int mem_open(struct inode *inode, struct file *file)
745 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
747 /* OK to pass negative loff_t, we can catch out-of-range */
748 file->f_mode |= FMODE_UNSIGNED_OFFSET;
753 static ssize_t mem_rw(struct file *file, char __user *buf,
754 size_t count, loff_t *ppos, int write)
756 struct mm_struct *mm = file->private_data;
757 unsigned long addr = *ppos;
764 page = (char *)__get_free_page(GFP_TEMPORARY);
769 if (!atomic_inc_not_zero(&mm->mm_users))
773 int this_len = min_t(int, count, PAGE_SIZE);
775 if (write && copy_from_user(page, buf, this_len)) {
780 this_len = access_remote_vm(mm, addr, page, this_len, write);
787 if (!write && copy_to_user(buf, page, this_len)) {
801 free_page((unsigned long) page);
805 static ssize_t mem_read(struct file *file, char __user *buf,
806 size_t count, loff_t *ppos)
808 return mem_rw(file, buf, count, ppos, 0);
811 static ssize_t mem_write(struct file *file, const char __user *buf,
812 size_t count, loff_t *ppos)
814 return mem_rw(file, (char __user*)buf, count, ppos, 1);
817 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
821 file->f_pos = offset;
824 file->f_pos += offset;
829 force_successful_syscall_return();
833 static int mem_release(struct inode *inode, struct file *file)
835 struct mm_struct *mm = file->private_data;
841 static const struct file_operations proc_mem_operations = {
846 .release = mem_release,
849 static int environ_open(struct inode *inode, struct file *file)
851 return __mem_open(inode, file, PTRACE_MODE_READ);
854 static ssize_t environ_read(struct file *file, char __user *buf,
855 size_t count, loff_t *ppos)
858 unsigned long src = *ppos;
860 struct mm_struct *mm = file->private_data;
865 page = (char *)__get_free_page(GFP_TEMPORARY);
870 if (!atomic_inc_not_zero(&mm->mm_users))
873 size_t this_len, max_len;
876 if (src >= (mm->env_end - mm->env_start))
879 this_len = mm->env_end - (mm->env_start + src);
881 max_len = min_t(size_t, PAGE_SIZE, count);
882 this_len = min(max_len, this_len);
884 retval = access_remote_vm(mm, (mm->env_start + src),
892 if (copy_to_user(buf, page, retval)) {
906 free_page((unsigned long) page);
910 static const struct file_operations proc_environ_operations = {
911 .open = environ_open,
912 .read = environ_read,
913 .llseek = generic_file_llseek,
914 .release = mem_release,
917 static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
920 struct task_struct *task = get_proc_task(file_inode(file));
921 char buffer[PROC_NUMBUF];
922 int oom_adj = OOM_ADJUST_MIN;
928 if (lock_task_sighand(task, &flags)) {
929 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
930 oom_adj = OOM_ADJUST_MAX;
932 oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
934 unlock_task_sighand(task, &flags);
936 put_task_struct(task);
937 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
938 return simple_read_from_buffer(buf, count, ppos, buffer, len);
941 static ssize_t oom_adj_write(struct file *file, const char __user *buf,
942 size_t count, loff_t *ppos)
944 struct task_struct *task;
945 char buffer[PROC_NUMBUF];
950 memset(buffer, 0, sizeof(buffer));
951 if (count > sizeof(buffer) - 1)
952 count = sizeof(buffer) - 1;
953 if (copy_from_user(buffer, buf, count)) {
958 err = kstrtoint(strstrip(buffer), 0, &oom_adj);
961 if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
962 oom_adj != OOM_DISABLE) {
967 task = get_proc_task(file_inode(file));
979 if (!lock_task_sighand(task, &flags)) {
985 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
986 * value is always attainable.
988 if (oom_adj == OOM_ADJUST_MAX)
989 oom_adj = OOM_SCORE_ADJ_MAX;
991 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
993 if (oom_adj < task->signal->oom_score_adj &&
994 !capable(CAP_SYS_RESOURCE)) {
999 /*ace add for monkey process oom_adj*/
1000 if(!strncmp("commands.monkey", current->comm, 15)) oom_adj = -16;
1003 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1004 * /proc/pid/oom_score_adj instead.
1006 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1007 current->comm, task_pid_nr(current), task_pid_nr(task),
1010 task->signal->oom_score_adj = oom_adj;
1011 trace_oom_score_adj_update(task);
1013 unlock_task_sighand(task, &flags);
1016 put_task_struct(task);
1018 return err < 0 ? err : count;
1021 static int oom_adjust_permission(struct inode *inode, int mask)
1024 struct task_struct *p;
1026 p = get_proc_task(inode);
1033 * System Server (uid == 1000) is granted access to oom_adj of all
1034 * android applications (uid > 10000) as and services (uid >= 1000)
1036 if (p && (current_fsuid() == 1000) && (uid >= 1000)) {
1037 if (inode->i_mode >> 6 & mask) {
1042 /* Fall back to default. */
1043 return generic_permission(inode, mask);
1046 static const struct inode_operations proc_oom_adj_inode_operations = {
1047 .permission = oom_adjust_permission,
1050 static const struct file_operations proc_oom_adj_operations = {
1051 .read = oom_adj_read,
1052 .write = oom_adj_write,
1053 .llseek = generic_file_llseek,
1056 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1057 size_t count, loff_t *ppos)
1059 struct task_struct *task = get_proc_task(file_inode(file));
1060 char buffer[PROC_NUMBUF];
1061 short oom_score_adj = OOM_SCORE_ADJ_MIN;
1062 unsigned long flags;
1067 if (lock_task_sighand(task, &flags)) {
1068 oom_score_adj = task->signal->oom_score_adj;
1069 unlock_task_sighand(task, &flags);
1071 put_task_struct(task);
1072 len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
1073 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1076 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1077 size_t count, loff_t *ppos)
1079 struct task_struct *task;
1080 char buffer[PROC_NUMBUF];
1081 unsigned long flags;
1085 memset(buffer, 0, sizeof(buffer));
1086 if (count > sizeof(buffer) - 1)
1087 count = sizeof(buffer) - 1;
1088 if (copy_from_user(buffer, buf, count)) {
1093 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1096 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1097 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1102 task = get_proc_task(file_inode(file));
1114 if (!lock_task_sighand(task, &flags)) {
1119 if ((short)oom_score_adj < task->signal->oom_score_adj_min &&
1120 !capable(CAP_SYS_RESOURCE)) {
1125 task->signal->oom_score_adj = (short)oom_score_adj;
1126 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1127 task->signal->oom_score_adj_min = (short)oom_score_adj;
1128 trace_oom_score_adj_update(task);
1131 unlock_task_sighand(task, &flags);
1134 put_task_struct(task);
1136 return err < 0 ? err : count;
1139 static const struct file_operations proc_oom_score_adj_operations = {
1140 .read = oom_score_adj_read,
1141 .write = oom_score_adj_write,
1142 .llseek = default_llseek,
1145 #ifdef CONFIG_AUDITSYSCALL
1146 #define TMPBUFLEN 21
1147 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1148 size_t count, loff_t *ppos)
1150 struct inode * inode = file_inode(file);
1151 struct task_struct *task = get_proc_task(inode);
1153 char tmpbuf[TMPBUFLEN];
1157 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1158 from_kuid(file->f_cred->user_ns,
1159 audit_get_loginuid(task)));
1160 put_task_struct(task);
1161 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1164 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1165 size_t count, loff_t *ppos)
1167 struct inode * inode = file_inode(file);
1174 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1180 if (count >= PAGE_SIZE)
1181 count = PAGE_SIZE - 1;
1184 /* No partial writes. */
1187 page = (char*)__get_free_page(GFP_TEMPORARY);
1191 if (copy_from_user(page, buf, count))
1195 loginuid = simple_strtoul(page, &tmp, 10);
1201 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1202 if (!uid_valid(kloginuid)) {
1207 length = audit_set_loginuid(kloginuid);
1208 if (likely(length == 0))
1212 free_page((unsigned long) page);
1216 static const struct file_operations proc_loginuid_operations = {
1217 .read = proc_loginuid_read,
1218 .write = proc_loginuid_write,
1219 .llseek = generic_file_llseek,
1222 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1223 size_t count, loff_t *ppos)
1225 struct inode * inode = file_inode(file);
1226 struct task_struct *task = get_proc_task(inode);
1228 char tmpbuf[TMPBUFLEN];
1232 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1233 audit_get_sessionid(task));
1234 put_task_struct(task);
1235 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1238 static const struct file_operations proc_sessionid_operations = {
1239 .read = proc_sessionid_read,
1240 .llseek = generic_file_llseek,
1244 #ifdef CONFIG_FAULT_INJECTION
1245 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1246 size_t count, loff_t *ppos)
1248 struct task_struct *task = get_proc_task(file_inode(file));
1249 char buffer[PROC_NUMBUF];
1255 make_it_fail = task->make_it_fail;
1256 put_task_struct(task);
1258 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1260 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1263 static ssize_t proc_fault_inject_write(struct file * file,
1264 const char __user * buf, size_t count, loff_t *ppos)
1266 struct task_struct *task;
1267 char buffer[PROC_NUMBUF], *end;
1270 if (!capable(CAP_SYS_RESOURCE))
1272 memset(buffer, 0, sizeof(buffer));
1273 if (count > sizeof(buffer) - 1)
1274 count = sizeof(buffer) - 1;
1275 if (copy_from_user(buffer, buf, count))
1277 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1280 task = get_proc_task(file_inode(file));
1283 task->make_it_fail = make_it_fail;
1284 put_task_struct(task);
1289 static const struct file_operations proc_fault_inject_operations = {
1290 .read = proc_fault_inject_read,
1291 .write = proc_fault_inject_write,
1292 .llseek = generic_file_llseek,
1297 #ifdef CONFIG_SCHED_DEBUG
1299 * Print out various scheduling related per-task fields:
1301 static int sched_show(struct seq_file *m, void *v)
1303 struct inode *inode = m->private;
1304 struct task_struct *p;
1306 p = get_proc_task(inode);
1309 proc_sched_show_task(p, m);
1317 sched_write(struct file *file, const char __user *buf,
1318 size_t count, loff_t *offset)
1320 struct inode *inode = file_inode(file);
1321 struct task_struct *p;
1323 p = get_proc_task(inode);
1326 proc_sched_set_task(p);
1333 static int sched_open(struct inode *inode, struct file *filp)
1335 return single_open(filp, sched_show, inode);
1338 static const struct file_operations proc_pid_sched_operations = {
1341 .write = sched_write,
1342 .llseek = seq_lseek,
1343 .release = single_release,
1348 #ifdef CONFIG_SCHED_AUTOGROUP
1350 * Print out autogroup related information:
1352 static int sched_autogroup_show(struct seq_file *m, void *v)
1354 struct inode *inode = m->private;
1355 struct task_struct *p;
1357 p = get_proc_task(inode);
1360 proc_sched_autogroup_show_task(p, m);
1368 sched_autogroup_write(struct file *file, const char __user *buf,
1369 size_t count, loff_t *offset)
1371 struct inode *inode = file_inode(file);
1372 struct task_struct *p;
1373 char buffer[PROC_NUMBUF];
1377 memset(buffer, 0, sizeof(buffer));
1378 if (count > sizeof(buffer) - 1)
1379 count = sizeof(buffer) - 1;
1380 if (copy_from_user(buffer, buf, count))
1383 err = kstrtoint(strstrip(buffer), 0, &nice);
1387 p = get_proc_task(inode);
1391 err = proc_sched_autogroup_set_nice(p, nice);
1400 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1404 ret = single_open(filp, sched_autogroup_show, NULL);
1406 struct seq_file *m = filp->private_data;
1413 static const struct file_operations proc_pid_sched_autogroup_operations = {
1414 .open = sched_autogroup_open,
1416 .write = sched_autogroup_write,
1417 .llseek = seq_lseek,
1418 .release = single_release,
1421 #endif /* CONFIG_SCHED_AUTOGROUP */
1423 static ssize_t comm_write(struct file *file, const char __user *buf,
1424 size_t count, loff_t *offset)
1426 struct inode *inode = file_inode(file);
1427 struct task_struct *p;
1428 char buffer[TASK_COMM_LEN];
1429 const size_t maxlen = sizeof(buffer) - 1;
1431 memset(buffer, 0, sizeof(buffer));
1432 if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
1435 p = get_proc_task(inode);
1439 if (same_thread_group(current, p))
1440 set_task_comm(p, buffer);
1449 static int comm_show(struct seq_file *m, void *v)
1451 struct inode *inode = m->private;
1452 struct task_struct *p;
1454 p = get_proc_task(inode);
1459 seq_printf(m, "%s\n", p->comm);
1467 static int comm_open(struct inode *inode, struct file *filp)
1469 return single_open(filp, comm_show, inode);
1472 static const struct file_operations proc_pid_set_comm_operations = {
1475 .write = comm_write,
1476 .llseek = seq_lseek,
1477 .release = single_release,
1480 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1482 struct task_struct *task;
1483 struct mm_struct *mm;
1484 struct file *exe_file;
1486 task = get_proc_task(dentry->d_inode);
1489 mm = get_task_mm(task);
1490 put_task_struct(task);
1493 exe_file = get_mm_exe_file(mm);
1496 *exe_path = exe_file->f_path;
1497 path_get(&exe_file->f_path);
1504 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1506 struct inode *inode = dentry->d_inode;
1508 int error = -EACCES;
1510 /* Are we allowed to snoop on the tasks file descriptors? */
1511 if (!proc_fd_access_allowed(inode))
1514 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1518 nd_jump_link(nd, &path);
1521 return ERR_PTR(error);
1524 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1526 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1533 pathname = d_path(path, tmp, PAGE_SIZE);
1534 len = PTR_ERR(pathname);
1535 if (IS_ERR(pathname))
1537 len = tmp + PAGE_SIZE - 1 - pathname;
1541 if (copy_to_user(buffer, pathname, len))
1544 free_page((unsigned long)tmp);
1548 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1550 int error = -EACCES;
1551 struct inode *inode = dentry->d_inode;
1554 /* Are we allowed to snoop on the tasks file descriptors? */
1555 if (!proc_fd_access_allowed(inode))
1558 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1562 error = do_proc_readlink(&path, buffer, buflen);
1568 const struct inode_operations proc_pid_link_inode_operations = {
1569 .readlink = proc_pid_readlink,
1570 .follow_link = proc_pid_follow_link,
1571 .setattr = proc_setattr,
1575 /* building an inode */
1577 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1579 struct inode * inode;
1580 struct proc_inode *ei;
1581 const struct cred *cred;
1583 /* We need a new inode */
1585 inode = new_inode(sb);
1591 inode->i_ino = get_next_ino();
1592 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1593 inode->i_op = &proc_def_inode_operations;
1596 * grab the reference to task.
1598 ei->pid = get_task_pid(task, PIDTYPE_PID);
1602 if (task_dumpable(task)) {
1604 cred = __task_cred(task);
1605 inode->i_uid = cred->euid;
1606 inode->i_gid = cred->egid;
1609 security_task_to_inode(task, inode);
1619 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1621 struct inode *inode = dentry->d_inode;
1622 struct task_struct *task;
1623 const struct cred *cred;
1624 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
1626 generic_fillattr(inode, stat);
1629 stat->uid = GLOBAL_ROOT_UID;
1630 stat->gid = GLOBAL_ROOT_GID;
1631 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1633 if (!has_pid_permissions(pid, task, 2)) {
1636 * This doesn't prevent learning whether PID exists,
1637 * it only makes getattr() consistent with readdir().
1641 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1642 task_dumpable(task)) {
1643 cred = __task_cred(task);
1644 stat->uid = cred->euid;
1645 stat->gid = cred->egid;
1655 * Exceptional case: normally we are not allowed to unhash a busy
1656 * directory. In this case, however, we can do it - no aliasing problems
1657 * due to the way we treat inodes.
1659 * Rewrite the inode's ownerships here because the owning task may have
1660 * performed a setuid(), etc.
1662 * Before the /proc/pid/status file was created the only way to read
1663 * the effective uid of a /process was to stat /proc/pid. Reading
1664 * /proc/pid/status is slow enough that procps and other packages
1665 * kept stating /proc/pid. To keep the rules in /proc simple I have
1666 * made this apply to all per process world readable and executable
1669 int pid_revalidate(struct dentry *dentry, unsigned int flags)
1671 struct inode *inode;
1672 struct task_struct *task;
1673 const struct cred *cred;
1675 if (flags & LOOKUP_RCU)
1678 inode = dentry->d_inode;
1679 task = get_proc_task(inode);
1682 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1683 task_dumpable(task)) {
1685 cred = __task_cred(task);
1686 inode->i_uid = cred->euid;
1687 inode->i_gid = cred->egid;
1690 inode->i_uid = GLOBAL_ROOT_UID;
1691 inode->i_gid = GLOBAL_ROOT_GID;
1693 inode->i_mode &= ~(S_ISUID | S_ISGID);
1694 security_task_to_inode(task, inode);
1695 put_task_struct(task);
1702 int pid_delete_dentry(const struct dentry *dentry)
1704 /* Is the task we represent dead?
1705 * If so, then don't put the dentry on the lru list,
1706 * kill it immediately.
1708 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1711 const struct dentry_operations pid_dentry_operations =
1713 .d_revalidate = pid_revalidate,
1714 .d_delete = pid_delete_dentry,
1720 * Fill a directory entry.
1722 * If possible create the dcache entry and derive our inode number and
1723 * file type from dcache entry.
1725 * Since all of the proc inode numbers are dynamically generated, the inode
1726 * numbers do not exist until the inode is cache. This means creating the
1727 * the dcache entry in readdir is necessary to keep the inode numbers
1728 * reported by readdir in sync with the inode numbers reported
1731 int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1732 const char *name, int len,
1733 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1735 struct dentry *child, *dir = filp->f_path.dentry;
1736 struct inode *inode;
1739 unsigned type = DT_UNKNOWN;
1743 qname.hash = full_name_hash(name, len);
1745 child = d_lookup(dir, &qname);
1748 new = d_alloc(dir, &qname);
1750 child = instantiate(dir->d_inode, new, task, ptr);
1757 if (!child || IS_ERR(child) || !child->d_inode)
1758 goto end_instantiate;
1759 inode = child->d_inode;
1762 type = inode->i_mode >> 12;
1767 ino = find_inode_number(dir, &qname);
1770 return filldir(dirent, name, len, filp->f_pos, ino, type);
1773 #ifdef CONFIG_CHECKPOINT_RESTORE
1776 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1777 * which represent vma start and end addresses.
1779 static int dname_to_vma_addr(struct dentry *dentry,
1780 unsigned long *start, unsigned long *end)
1782 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1788 static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
1790 unsigned long vm_start, vm_end;
1791 bool exact_vma_exists = false;
1792 struct mm_struct *mm = NULL;
1793 struct task_struct *task;
1794 const struct cred *cred;
1795 struct inode *inode;
1798 if (flags & LOOKUP_RCU)
1801 if (!capable(CAP_SYS_ADMIN)) {
1806 inode = dentry->d_inode;
1807 task = get_proc_task(inode);
1811 mm = mm_access(task, PTRACE_MODE_READ);
1812 if (IS_ERR_OR_NULL(mm))
1815 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1816 down_read(&mm->mmap_sem);
1817 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1818 up_read(&mm->mmap_sem);
1823 if (exact_vma_exists) {
1824 if (task_dumpable(task)) {
1826 cred = __task_cred(task);
1827 inode->i_uid = cred->euid;
1828 inode->i_gid = cred->egid;
1831 inode->i_uid = GLOBAL_ROOT_UID;
1832 inode->i_gid = GLOBAL_ROOT_GID;
1834 security_task_to_inode(task, inode);
1839 put_task_struct(task);
1848 static const struct dentry_operations tid_map_files_dentry_operations = {
1849 .d_revalidate = map_files_d_revalidate,
1850 .d_delete = pid_delete_dentry,
1853 static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
1855 unsigned long vm_start, vm_end;
1856 struct vm_area_struct *vma;
1857 struct task_struct *task;
1858 struct mm_struct *mm;
1862 task = get_proc_task(dentry->d_inode);
1866 mm = get_task_mm(task);
1867 put_task_struct(task);
1871 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
1876 down_read(&mm->mmap_sem);
1877 vma = find_exact_vma(mm, vm_start, vm_end);
1878 if (vma && vma->vm_file) {
1879 *path = vma->vm_file->f_path;
1883 up_read(&mm->mmap_sem);
1891 struct map_files_info {
1894 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
1897 static struct dentry *
1898 proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
1899 struct task_struct *task, const void *ptr)
1901 fmode_t mode = (fmode_t)(unsigned long)ptr;
1902 struct proc_inode *ei;
1903 struct inode *inode;
1905 inode = proc_pid_make_inode(dir->i_sb, task);
1907 return ERR_PTR(-ENOENT);
1910 ei->op.proc_get_link = proc_map_files_get_link;
1912 inode->i_op = &proc_pid_link_inode_operations;
1914 inode->i_mode = S_IFLNK;
1916 if (mode & FMODE_READ)
1917 inode->i_mode |= S_IRUSR;
1918 if (mode & FMODE_WRITE)
1919 inode->i_mode |= S_IWUSR;
1921 d_set_d_op(dentry, &tid_map_files_dentry_operations);
1922 d_add(dentry, inode);
1927 static struct dentry *proc_map_files_lookup(struct inode *dir,
1928 struct dentry *dentry, unsigned int flags)
1930 unsigned long vm_start, vm_end;
1931 struct vm_area_struct *vma;
1932 struct task_struct *task;
1933 struct dentry *result;
1934 struct mm_struct *mm;
1936 result = ERR_PTR(-EPERM);
1937 if (!capable(CAP_SYS_ADMIN))
1940 result = ERR_PTR(-ENOENT);
1941 task = get_proc_task(dir);
1945 result = ERR_PTR(-EACCES);
1946 if (!ptrace_may_access(task, PTRACE_MODE_READ))
1949 result = ERR_PTR(-ENOENT);
1950 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
1953 mm = get_task_mm(task);
1957 down_read(&mm->mmap_sem);
1958 vma = find_exact_vma(mm, vm_start, vm_end);
1963 result = proc_map_files_instantiate(dir, dentry, task,
1964 (void *)(unsigned long)vma->vm_file->f_mode);
1967 up_read(&mm->mmap_sem);
1970 put_task_struct(task);
1975 static const struct inode_operations proc_map_files_inode_operations = {
1976 .lookup = proc_map_files_lookup,
1977 .permission = proc_fd_permission,
1978 .setattr = proc_setattr,
1982 proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
1984 struct dentry *dentry = filp->f_path.dentry;
1985 struct inode *inode = dentry->d_inode;
1986 struct vm_area_struct *vma;
1987 struct task_struct *task;
1988 struct mm_struct *mm;
1993 if (!capable(CAP_SYS_ADMIN))
1997 task = get_proc_task(inode);
2002 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2006 switch (filp->f_pos) {
2009 if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0)
2013 ino = parent_ino(dentry);
2014 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2019 unsigned long nr_files, pos, i;
2020 struct flex_array *fa = NULL;
2021 struct map_files_info info;
2022 struct map_files_info *p;
2024 mm = get_task_mm(task);
2027 down_read(&mm->mmap_sem);
2032 * We need two passes here:
2034 * 1) Collect vmas of mapped files with mmap_sem taken
2035 * 2) Release mmap_sem and instantiate entries
2037 * otherwise we get lockdep complained, since filldir()
2038 * routine might require mmap_sem taken in might_fault().
2041 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2042 if (vma->vm_file && ++pos > filp->f_pos)
2047 fa = flex_array_alloc(sizeof(info), nr_files,
2049 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2053 flex_array_free(fa);
2054 up_read(&mm->mmap_sem);
2058 for (i = 0, vma = mm->mmap, pos = 2; vma;
2059 vma = vma->vm_next) {
2062 if (++pos <= filp->f_pos)
2065 info.mode = vma->vm_file->f_mode;
2066 info.len = snprintf(info.name,
2067 sizeof(info.name), "%lx-%lx",
2068 vma->vm_start, vma->vm_end);
2069 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2073 up_read(&mm->mmap_sem);
2075 for (i = 0; i < nr_files; i++) {
2076 p = flex_array_get(fa, i);
2077 ret = proc_fill_cache(filp, dirent, filldir,
2079 proc_map_files_instantiate,
2081 (void *)(unsigned long)p->mode);
2087 flex_array_free(fa);
2093 put_task_struct(task);
2098 static const struct file_operations proc_map_files_operations = {
2099 .read = generic_read_dir,
2100 .readdir = proc_map_files_readdir,
2101 .llseek = default_llseek,
2104 struct timers_private {
2106 struct task_struct *task;
2107 struct sighand_struct *sighand;
2108 struct pid_namespace *ns;
2109 unsigned long flags;
2112 static void *timers_start(struct seq_file *m, loff_t *pos)
2114 struct timers_private *tp = m->private;
2116 tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2118 return ERR_PTR(-ESRCH);
2120 tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2122 return ERR_PTR(-ESRCH);
2124 return seq_list_start(&tp->task->signal->posix_timers, *pos);
2127 static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2129 struct timers_private *tp = m->private;
2130 return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2133 static void timers_stop(struct seq_file *m, void *v)
2135 struct timers_private *tp = m->private;
2138 unlock_task_sighand(tp->task, &tp->flags);
2143 put_task_struct(tp->task);
2148 static int show_timer(struct seq_file *m, void *v)
2150 struct k_itimer *timer;
2151 struct timers_private *tp = m->private;
2153 static char *nstr[] = {
2154 [SIGEV_SIGNAL] = "signal",
2155 [SIGEV_NONE] = "none",
2156 [SIGEV_THREAD] = "thread",
2159 timer = list_entry((struct list_head *)v, struct k_itimer, list);
2160 notify = timer->it_sigev_notify;
2162 seq_printf(m, "ID: %d\n", timer->it_id);
2163 seq_printf(m, "signal: %d/%p\n", timer->sigq->info.si_signo,
2164 timer->sigq->info.si_value.sival_ptr);
2165 seq_printf(m, "notify: %s/%s.%d\n",
2166 nstr[notify & ~SIGEV_THREAD_ID],
2167 (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2168 pid_nr_ns(timer->it_pid, tp->ns));
2169 seq_printf(m, "ClockID: %d\n", timer->it_clock);
2174 static const struct seq_operations proc_timers_seq_ops = {
2175 .start = timers_start,
2176 .next = timers_next,
2177 .stop = timers_stop,
2181 static int proc_timers_open(struct inode *inode, struct file *file)
2183 struct timers_private *tp;
2185 tp = __seq_open_private(file, &proc_timers_seq_ops,
2186 sizeof(struct timers_private));
2190 tp->pid = proc_pid(inode);
2191 tp->ns = inode->i_sb->s_fs_info;
2195 static const struct file_operations proc_timers_operations = {
2196 .open = proc_timers_open,
2198 .llseek = seq_lseek,
2199 .release = seq_release_private,
2201 #endif /* CONFIG_CHECKPOINT_RESTORE */
2203 static struct dentry *proc_pident_instantiate(struct inode *dir,
2204 struct dentry *dentry, struct task_struct *task, const void *ptr)
2206 const struct pid_entry *p = ptr;
2207 struct inode *inode;
2208 struct proc_inode *ei;
2209 struct dentry *error = ERR_PTR(-ENOENT);
2211 inode = proc_pid_make_inode(dir->i_sb, task);
2216 inode->i_mode = p->mode;
2217 if (S_ISDIR(inode->i_mode))
2218 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2220 inode->i_op = p->iop;
2222 inode->i_fop = p->fop;
2224 d_set_d_op(dentry, &pid_dentry_operations);
2225 d_add(dentry, inode);
2226 /* Close the race of the process dying before we return the dentry */
2227 if (pid_revalidate(dentry, 0))
2233 static struct dentry *proc_pident_lookup(struct inode *dir,
2234 struct dentry *dentry,
2235 const struct pid_entry *ents,
2238 struct dentry *error;
2239 struct task_struct *task = get_proc_task(dir);
2240 const struct pid_entry *p, *last;
2242 error = ERR_PTR(-ENOENT);
2248 * Yes, it does not scale. And it should not. Don't add
2249 * new entries into /proc/<tgid>/ without very good reasons.
2251 last = &ents[nents - 1];
2252 for (p = ents; p <= last; p++) {
2253 if (p->len != dentry->d_name.len)
2255 if (!memcmp(dentry->d_name.name, p->name, p->len))
2261 error = proc_pident_instantiate(dir, dentry, task, p);
2263 put_task_struct(task);
2268 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2269 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2271 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2272 proc_pident_instantiate, task, p);
2275 static int proc_pident_readdir(struct file *filp,
2276 void *dirent, filldir_t filldir,
2277 const struct pid_entry *ents, unsigned int nents)
2280 struct dentry *dentry = filp->f_path.dentry;
2281 struct inode *inode = dentry->d_inode;
2282 struct task_struct *task = get_proc_task(inode);
2283 const struct pid_entry *p, *last;
2296 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2302 ino = parent_ino(dentry);
2303 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2315 last = &ents[nents - 1];
2317 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2326 put_task_struct(task);
2331 #ifdef CONFIG_SECURITY
2332 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2333 size_t count, loff_t *ppos)
2335 struct inode * inode = file_inode(file);
2338 struct task_struct *task = get_proc_task(inode);
2343 length = security_getprocattr(task,
2344 (char*)file->f_path.dentry->d_name.name,
2346 put_task_struct(task);
2348 length = simple_read_from_buffer(buf, count, ppos, p, length);
2353 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2354 size_t count, loff_t *ppos)
2356 struct inode * inode = file_inode(file);
2359 struct task_struct *task = get_proc_task(inode);
2364 if (count > PAGE_SIZE)
2367 /* No partial writes. */
2373 page = (char*)__get_free_page(GFP_TEMPORARY);
2378 if (copy_from_user(page, buf, count))
2381 /* Guard against adverse ptrace interaction */
2382 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2386 length = security_setprocattr(task,
2387 (char*)file->f_path.dentry->d_name.name,
2388 (void*)page, count);
2389 mutex_unlock(&task->signal->cred_guard_mutex);
2391 free_page((unsigned long) page);
2393 put_task_struct(task);
2398 static const struct file_operations proc_pid_attr_operations = {
2399 .read = proc_pid_attr_read,
2400 .write = proc_pid_attr_write,
2401 .llseek = generic_file_llseek,
2404 static const struct pid_entry attr_dir_stuff[] = {
2405 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2406 REG("prev", S_IRUGO, proc_pid_attr_operations),
2407 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2408 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2409 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2410 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2413 static int proc_attr_dir_readdir(struct file * filp,
2414 void * dirent, filldir_t filldir)
2416 return proc_pident_readdir(filp,dirent,filldir,
2417 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2420 static const struct file_operations proc_attr_dir_operations = {
2421 .read = generic_read_dir,
2422 .readdir = proc_attr_dir_readdir,
2423 .llseek = default_llseek,
2426 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2427 struct dentry *dentry, unsigned int flags)
2429 return proc_pident_lookup(dir, dentry,
2430 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2433 static const struct inode_operations proc_attr_dir_inode_operations = {
2434 .lookup = proc_attr_dir_lookup,
2435 .getattr = pid_getattr,
2436 .setattr = proc_setattr,
2441 #ifdef CONFIG_ELF_CORE
2442 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2443 size_t count, loff_t *ppos)
2445 struct task_struct *task = get_proc_task(file_inode(file));
2446 struct mm_struct *mm;
2447 char buffer[PROC_NUMBUF];
2455 mm = get_task_mm(task);
2457 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2458 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2459 MMF_DUMP_FILTER_SHIFT));
2461 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2464 put_task_struct(task);
2469 static ssize_t proc_coredump_filter_write(struct file *file,
2470 const char __user *buf,
2474 struct task_struct *task;
2475 struct mm_struct *mm;
2476 char buffer[PROC_NUMBUF], *end;
2483 memset(buffer, 0, sizeof(buffer));
2484 if (count > sizeof(buffer) - 1)
2485 count = sizeof(buffer) - 1;
2486 if (copy_from_user(buffer, buf, count))
2490 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2493 if (end - buffer == 0)
2497 task = get_proc_task(file_inode(file));
2502 mm = get_task_mm(task);
2506 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2508 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2510 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2515 put_task_struct(task);
2520 static const struct file_operations proc_coredump_filter_operations = {
2521 .read = proc_coredump_filter_read,
2522 .write = proc_coredump_filter_write,
2523 .llseek = generic_file_llseek,
2527 #ifdef CONFIG_TASK_IO_ACCOUNTING
2528 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2530 struct task_io_accounting acct = task->ioac;
2531 unsigned long flags;
2534 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2538 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2543 if (whole && lock_task_sighand(task, &flags)) {
2544 struct task_struct *t = task;
2546 task_io_accounting_add(&acct, &task->signal->ioac);
2547 while_each_thread(task, t)
2548 task_io_accounting_add(&acct, &t->ioac);
2550 unlock_task_sighand(task, &flags);
2552 result = sprintf(buffer,
2557 "read_bytes: %llu\n"
2558 "write_bytes: %llu\n"
2559 "cancelled_write_bytes: %llu\n",
2560 (unsigned long long)acct.rchar,
2561 (unsigned long long)acct.wchar,
2562 (unsigned long long)acct.syscr,
2563 (unsigned long long)acct.syscw,
2564 (unsigned long long)acct.read_bytes,
2565 (unsigned long long)acct.write_bytes,
2566 (unsigned long long)acct.cancelled_write_bytes);
2568 mutex_unlock(&task->signal->cred_guard_mutex);
2572 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2574 return do_io_accounting(task, buffer, 0);
2577 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2579 return do_io_accounting(task, buffer, 1);
2581 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2583 #ifdef CONFIG_USER_NS
2584 static int proc_id_map_open(struct inode *inode, struct file *file,
2585 struct seq_operations *seq_ops)
2587 struct user_namespace *ns = NULL;
2588 struct task_struct *task;
2589 struct seq_file *seq;
2592 task = get_proc_task(inode);
2595 ns = get_user_ns(task_cred_xxx(task, user_ns));
2597 put_task_struct(task);
2602 ret = seq_open(file, seq_ops);
2606 seq = file->private_data;
2616 static int proc_id_map_release(struct inode *inode, struct file *file)
2618 struct seq_file *seq = file->private_data;
2619 struct user_namespace *ns = seq->private;
2621 return seq_release(inode, file);
2624 static int proc_uid_map_open(struct inode *inode, struct file *file)
2626 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2629 static int proc_gid_map_open(struct inode *inode, struct file *file)
2631 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2634 static int proc_projid_map_open(struct inode *inode, struct file *file)
2636 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2639 static const struct file_operations proc_uid_map_operations = {
2640 .open = proc_uid_map_open,
2641 .write = proc_uid_map_write,
2643 .llseek = seq_lseek,
2644 .release = proc_id_map_release,
2647 static const struct file_operations proc_gid_map_operations = {
2648 .open = proc_gid_map_open,
2649 .write = proc_gid_map_write,
2651 .llseek = seq_lseek,
2652 .release = proc_id_map_release,
2655 static const struct file_operations proc_projid_map_operations = {
2656 .open = proc_projid_map_open,
2657 .write = proc_projid_map_write,
2659 .llseek = seq_lseek,
2660 .release = proc_id_map_release,
2663 static int proc_setgroups_open(struct inode *inode, struct file *file)
2665 struct user_namespace *ns = NULL;
2666 struct task_struct *task;
2670 task = get_proc_task(inode);
2673 ns = get_user_ns(task_cred_xxx(task, user_ns));
2675 put_task_struct(task);
2680 if (file->f_mode & FMODE_WRITE) {
2682 if (!ns_capable(ns, CAP_SYS_ADMIN))
2686 ret = single_open(file, &proc_setgroups_show, ns);
2697 static int proc_setgroups_release(struct inode *inode, struct file *file)
2699 struct seq_file *seq = file->private_data;
2700 struct user_namespace *ns = seq->private;
2701 int ret = single_release(inode, file);
2706 static const struct file_operations proc_setgroups_operations = {
2707 .open = proc_setgroups_open,
2708 .write = proc_setgroups_write,
2710 .llseek = seq_lseek,
2711 .release = proc_setgroups_release,
2713 #endif /* CONFIG_USER_NS */
2715 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2716 struct pid *pid, struct task_struct *task)
2718 int err = lock_trace(task);
2720 seq_printf(m, "%08x\n", task->personality);
2729 static const struct file_operations proc_task_operations;
2730 static const struct inode_operations proc_task_inode_operations;
2732 static const struct pid_entry tgid_base_stuff[] = {
2733 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2734 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2735 #ifdef CONFIG_CHECKPOINT_RESTORE
2736 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
2738 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2739 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2741 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2743 REG("environ", S_IRUSR, proc_environ_operations),
2744 INF("auxv", S_IRUSR, proc_pid_auxv),
2745 ONE("status", S_IRUGO, proc_pid_status),
2746 ONE("personality", S_IRUGO, proc_pid_personality),
2747 INF("limits", S_IRUGO, proc_pid_limits),
2748 #ifdef CONFIG_SCHED_DEBUG
2749 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2751 #ifdef CONFIG_SCHED_AUTOGROUP
2752 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2754 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2755 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2756 INF("syscall", S_IRUGO, proc_pid_syscall),
2758 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2759 ONE("stat", S_IRUGO, proc_tgid_stat),
2760 ONE("statm", S_IRUGO, proc_pid_statm),
2761 REG("maps", S_IRUGO, proc_pid_maps_operations),
2763 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
2765 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2766 LNK("cwd", proc_cwd_link),
2767 LNK("root", proc_root_link),
2768 LNK("exe", proc_exe_link),
2769 REG("mounts", S_IRUGO, proc_mounts_operations),
2770 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2771 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2772 #ifdef CONFIG_PROC_PAGE_MONITOR
2773 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2774 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
2775 REG("pagemap", S_IRUGO, proc_pagemap_operations),
2777 #ifdef CONFIG_SECURITY
2778 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2780 #ifdef CONFIG_KALLSYMS
2781 INF("wchan", S_IRUGO, proc_pid_wchan),
2783 #ifdef CONFIG_STACKTRACE
2784 ONE("stack", S_IRUGO, proc_pid_stack),
2786 #ifdef CONFIG_SCHEDSTATS
2787 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2789 #ifdef CONFIG_LATENCYTOP
2790 REG("latency", S_IRUGO, proc_lstats_operations),
2792 #ifdef CONFIG_PROC_PID_CPUSET
2793 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2795 #ifdef CONFIG_CGROUPS
2796 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2798 INF("oom_score", S_IRUGO, proc_oom_score),
2799 ANDROID("oom_adj", S_IRUGO|S_IWUSR, oom_adj),
2800 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2801 #ifdef CONFIG_AUDITSYSCALL
2802 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2803 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2805 #ifdef CONFIG_FAULT_INJECTION
2806 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2808 #ifdef CONFIG_ELF_CORE
2809 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2811 #ifdef CONFIG_TASK_IO_ACCOUNTING
2812 INF("io", S_IRUSR, proc_tgid_io_accounting),
2814 #ifdef CONFIG_HARDWALL
2815 INF("hardwall", S_IRUGO, proc_pid_hardwall),
2817 #ifdef CONFIG_USER_NS
2818 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
2819 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
2820 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
2821 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
2823 #ifdef CONFIG_CHECKPOINT_RESTORE
2824 REG("timers", S_IRUGO, proc_timers_operations),
2828 static int proc_tgid_base_readdir(struct file * filp,
2829 void * dirent, filldir_t filldir)
2831 return proc_pident_readdir(filp,dirent,filldir,
2832 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2835 static const struct file_operations proc_tgid_base_operations = {
2836 .read = generic_read_dir,
2837 .readdir = proc_tgid_base_readdir,
2838 .llseek = default_llseek,
2841 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2843 return proc_pident_lookup(dir, dentry,
2844 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2847 static const struct inode_operations proc_tgid_base_inode_operations = {
2848 .lookup = proc_tgid_base_lookup,
2849 .getattr = pid_getattr,
2850 .setattr = proc_setattr,
2851 .permission = proc_pid_permission,
2854 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2856 struct dentry *dentry, *leader, *dir;
2857 char buf[PROC_NUMBUF];
2861 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2862 /* no ->d_hash() rejects on procfs */
2863 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2865 shrink_dcache_parent(dentry);
2871 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2872 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2877 name.len = strlen(name.name);
2878 dir = d_hash_and_lookup(leader, &name);
2880 goto out_put_leader;
2883 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2884 dentry = d_hash_and_lookup(dir, &name);
2886 shrink_dcache_parent(dentry);
2899 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2900 * @task: task that should be flushed.
2902 * When flushing dentries from proc, one needs to flush them from global
2903 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2904 * in. This call is supposed to do all of this job.
2906 * Looks in the dcache for
2908 * /proc/@tgid/task/@pid
2909 * if either directory is present flushes it and all of it'ts children
2912 * It is safe and reasonable to cache /proc entries for a task until
2913 * that task exits. After that they just clog up the dcache with
2914 * useless entries, possibly causing useful dcache entries to be
2915 * flushed instead. This routine is proved to flush those useless
2916 * dcache entries at process exit time.
2918 * NOTE: This routine is just an optimization so it does not guarantee
2919 * that no dcache entries will exist at process exit time it
2920 * just makes it very unlikely that any will persist.
2923 void proc_flush_task(struct task_struct *task)
2926 struct pid *pid, *tgid;
2929 pid = task_pid(task);
2930 tgid = task_tgid(task);
2932 for (i = 0; i <= pid->level; i++) {
2933 upid = &pid->numbers[i];
2934 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2935 tgid->numbers[i].nr);
2939 static struct dentry *proc_pid_instantiate(struct inode *dir,
2940 struct dentry * dentry,
2941 struct task_struct *task, const void *ptr)
2943 struct dentry *error = ERR_PTR(-ENOENT);
2944 struct inode *inode;
2946 inode = proc_pid_make_inode(dir->i_sb, task);
2950 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2951 inode->i_op = &proc_tgid_base_inode_operations;
2952 inode->i_fop = &proc_tgid_base_operations;
2953 inode->i_flags|=S_IMMUTABLE;
2955 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
2956 ARRAY_SIZE(tgid_base_stuff)));
2958 d_set_d_op(dentry, &pid_dentry_operations);
2960 d_add(dentry, inode);
2961 /* Close the race of the process dying before we return the dentry */
2962 if (pid_revalidate(dentry, 0))
2968 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
2970 struct dentry *result = NULL;
2971 struct task_struct *task;
2973 struct pid_namespace *ns;
2975 tgid = name_to_int(dentry);
2979 ns = dentry->d_sb->s_fs_info;
2981 task = find_task_by_pid_ns(tgid, ns);
2983 get_task_struct(task);
2988 result = proc_pid_instantiate(dir, dentry, task, NULL);
2989 put_task_struct(task);
2995 * Find the first task with tgid >= tgid
3000 struct task_struct *task;
3002 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3007 put_task_struct(iter.task);
3011 pid = find_ge_pid(iter.tgid, ns);
3013 iter.tgid = pid_nr_ns(pid, ns);
3014 iter.task = pid_task(pid, PIDTYPE_PID);
3015 /* What we to know is if the pid we have find is the
3016 * pid of a thread_group_leader. Testing for task
3017 * being a thread_group_leader is the obvious thing
3018 * todo but there is a window when it fails, due to
3019 * the pid transfer logic in de_thread.
3021 * So we perform the straight forward test of seeing
3022 * if the pid we have found is the pid of a thread
3023 * group leader, and don't worry if the task we have
3024 * found doesn't happen to be a thread group leader.
3025 * As we don't care in the case of readdir.
3027 if (!iter.task || !has_group_leader_pid(iter.task)) {
3031 get_task_struct(iter.task);
3037 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 1)
3039 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3040 struct tgid_iter iter)
3042 char name[PROC_NUMBUF];
3043 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
3044 return proc_fill_cache(filp, dirent, filldir, name, len,
3045 proc_pid_instantiate, iter.task, NULL);
3048 static int fake_filldir(void *buf, const char *name, int namelen,
3049 loff_t offset, u64 ino, unsigned d_type)
3054 /* for the /proc/ directory itself, after non-process stuff has been done */
3055 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3057 struct tgid_iter iter;
3058 struct pid_namespace *ns;
3059 filldir_t __filldir;
3060 loff_t pos = filp->f_pos;
3062 if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
3065 if (pos == TGID_OFFSET - 1) {
3066 if (proc_fill_cache(filp, dirent, filldir, "self", 4,
3067 NULL, NULL, NULL) < 0)
3071 iter.tgid = pos - TGID_OFFSET;
3074 ns = filp->f_dentry->d_sb->s_fs_info;
3075 for (iter = next_tgid(ns, iter);
3077 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3078 if (has_pid_permissions(ns, iter.task, 2))
3079 __filldir = filldir;
3081 __filldir = fake_filldir;
3083 filp->f_pos = iter.tgid + TGID_OFFSET;
3084 if (proc_pid_fill_cache(filp, dirent, __filldir, iter) < 0) {
3085 put_task_struct(iter.task);
3089 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3097 static const struct pid_entry tid_base_stuff[] = {
3098 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3099 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3100 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3101 REG("environ", S_IRUSR, proc_environ_operations),
3102 INF("auxv", S_IRUSR, proc_pid_auxv),
3103 ONE("status", S_IRUGO, proc_pid_status),
3104 ONE("personality", S_IRUGO, proc_pid_personality),
3105 INF("limits", S_IRUGO, proc_pid_limits),
3106 #ifdef CONFIG_SCHED_DEBUG
3107 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3109 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3110 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3111 INF("syscall", S_IRUGO, proc_pid_syscall),
3113 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3114 ONE("stat", S_IRUGO, proc_tid_stat),
3115 ONE("statm", S_IRUGO, proc_pid_statm),
3116 REG("maps", S_IRUGO, proc_tid_maps_operations),
3117 #ifdef CONFIG_CHECKPOINT_RESTORE
3118 REG("children", S_IRUGO, proc_tid_children_operations),
3121 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
3123 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3124 LNK("cwd", proc_cwd_link),
3125 LNK("root", proc_root_link),
3126 LNK("exe", proc_exe_link),
3127 REG("mounts", S_IRUGO, proc_mounts_operations),
3128 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3129 #ifdef CONFIG_PROC_PAGE_MONITOR
3130 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3131 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
3132 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3134 #ifdef CONFIG_SECURITY
3135 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3137 #ifdef CONFIG_KALLSYMS
3138 INF("wchan", S_IRUGO, proc_pid_wchan),
3140 #ifdef CONFIG_STACKTRACE
3141 ONE("stack", S_IRUGO, proc_pid_stack),
3143 #ifdef CONFIG_SCHEDSTATS
3144 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3146 #ifdef CONFIG_LATENCYTOP
3147 REG("latency", S_IRUGO, proc_lstats_operations),
3149 #ifdef CONFIG_PROC_PID_CPUSET
3150 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3152 #ifdef CONFIG_CGROUPS
3153 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3155 INF("oom_score", S_IRUGO, proc_oom_score),
3156 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3157 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3158 #ifdef CONFIG_AUDITSYSCALL
3159 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3160 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3162 #ifdef CONFIG_FAULT_INJECTION
3163 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3165 #ifdef CONFIG_TASK_IO_ACCOUNTING
3166 INF("io", S_IRUSR, proc_tid_io_accounting),
3168 #ifdef CONFIG_HARDWALL
3169 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3171 #ifdef CONFIG_USER_NS
3172 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3173 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3174 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3175 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
3179 static int proc_tid_base_readdir(struct file * filp,
3180 void * dirent, filldir_t filldir)
3182 return proc_pident_readdir(filp,dirent,filldir,
3183 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3186 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3188 return proc_pident_lookup(dir, dentry,
3189 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3192 static const struct file_operations proc_tid_base_operations = {
3193 .read = generic_read_dir,
3194 .readdir = proc_tid_base_readdir,
3195 .llseek = default_llseek,
3198 static const struct inode_operations proc_tid_base_inode_operations = {
3199 .lookup = proc_tid_base_lookup,
3200 .getattr = pid_getattr,
3201 .setattr = proc_setattr,
3204 static struct dentry *proc_task_instantiate(struct inode *dir,
3205 struct dentry *dentry, struct task_struct *task, const void *ptr)
3207 struct dentry *error = ERR_PTR(-ENOENT);
3208 struct inode *inode;
3209 inode = proc_pid_make_inode(dir->i_sb, task);
3213 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3214 inode->i_op = &proc_tid_base_inode_operations;
3215 inode->i_fop = &proc_tid_base_operations;
3216 inode->i_flags|=S_IMMUTABLE;
3218 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3219 ARRAY_SIZE(tid_base_stuff)));
3221 d_set_d_op(dentry, &pid_dentry_operations);
3223 d_add(dentry, inode);
3224 /* Close the race of the process dying before we return the dentry */
3225 if (pid_revalidate(dentry, 0))
3231 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3233 struct dentry *result = ERR_PTR(-ENOENT);
3234 struct task_struct *task;
3235 struct task_struct *leader = get_proc_task(dir);
3237 struct pid_namespace *ns;
3242 tid = name_to_int(dentry);
3246 ns = dentry->d_sb->s_fs_info;
3248 task = find_task_by_pid_ns(tid, ns);
3250 get_task_struct(task);
3254 if (!same_thread_group(leader, task))
3257 result = proc_task_instantiate(dir, dentry, task, NULL);
3259 put_task_struct(task);
3261 put_task_struct(leader);
3267 * Find the first tid of a thread group to return to user space.
3269 * Usually this is just the thread group leader, but if the users
3270 * buffer was too small or there was a seek into the middle of the
3271 * directory we have more work todo.
3273 * In the case of a short read we start with find_task_by_pid.
3275 * In the case of a seek we start with the leader and walk nr
3278 static struct task_struct *first_tid(struct task_struct *leader,
3279 int tid, int nr, struct pid_namespace *ns)
3281 struct task_struct *pos;
3284 /* Attempt to start with the pid of a thread */
3285 if (tid && (nr > 0)) {
3286 pos = find_task_by_pid_ns(tid, ns);
3287 if (pos && (pos->group_leader == leader))
3291 /* If nr exceeds the number of threads there is nothing todo */
3293 if (nr && nr >= get_nr_threads(leader))
3296 /* If we haven't found our starting place yet start
3297 * with the leader and walk nr threads forward.
3299 for (pos = leader; nr > 0; --nr) {
3300 pos = next_thread(pos);
3301 if (pos == leader) {
3307 get_task_struct(pos);
3314 * Find the next thread in the thread list.
3315 * Return NULL if there is an error or no next thread.
3317 * The reference to the input task_struct is released.
3319 static struct task_struct *next_tid(struct task_struct *start)
3321 struct task_struct *pos = NULL;
3323 if (pid_alive(start)) {
3324 pos = next_thread(start);
3325 if (thread_group_leader(pos))
3328 get_task_struct(pos);
3331 put_task_struct(start);
3335 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3336 struct task_struct *task, int tid)
3338 char name[PROC_NUMBUF];
3339 int len = snprintf(name, sizeof(name), "%d", tid);
3340 return proc_fill_cache(filp, dirent, filldir, name, len,
3341 proc_task_instantiate, task, NULL);
3344 /* for the /proc/TGID/task/ directories */
3345 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3347 struct dentry *dentry = filp->f_path.dentry;
3348 struct inode *inode = dentry->d_inode;
3349 struct task_struct *leader = NULL;
3350 struct task_struct *task;
3351 int retval = -ENOENT;
3354 struct pid_namespace *ns;
3356 task = get_proc_task(inode);
3360 if (pid_alive(task)) {
3361 leader = task->group_leader;
3362 get_task_struct(leader);
3365 put_task_struct(task);
3370 switch ((unsigned long)filp->f_pos) {
3373 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3378 ino = parent_ino(dentry);
3379 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3385 /* f_version caches the tgid value that the last readdir call couldn't
3386 * return. lseek aka telldir automagically resets f_version to 0.
3388 ns = filp->f_dentry->d_sb->s_fs_info;
3389 tid = (int)filp->f_version;
3390 filp->f_version = 0;
3391 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3393 task = next_tid(task), filp->f_pos++) {
3394 tid = task_pid_nr_ns(task, ns);
3395 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3396 /* returning this tgid failed, save it as the first
3397 * pid for the next readir call */
3398 filp->f_version = (u64)tid;
3399 put_task_struct(task);
3404 put_task_struct(leader);
3409 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3411 struct inode *inode = dentry->d_inode;
3412 struct task_struct *p = get_proc_task(inode);
3413 generic_fillattr(inode, stat);
3416 stat->nlink += get_nr_threads(p);
3423 static const struct inode_operations proc_task_inode_operations = {
3424 .lookup = proc_task_lookup,
3425 .getattr = proc_task_getattr,
3426 .setattr = proc_setattr,
3427 .permission = proc_pid_permission,
3430 static const struct file_operations proc_task_operations = {
3431 .read = generic_read_dir,
3432 .readdir = proc_task_readdir,
3433 .llseek = default_llseek,