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)
205 return get_cmdline(task, buffer, PAGE_SIZE);
208 static int proc_pid_auxv(struct seq_file *m, struct pid_namespace *ns,
209 struct pid *pid, struct task_struct *task)
211 struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ);
212 if (mm && !IS_ERR(mm)) {
213 unsigned int nwords = 0;
216 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
217 seq_write(m, mm->saved_auxv, nwords * sizeof(mm->saved_auxv[0]));
225 #ifdef CONFIG_KALLSYMS
227 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
228 * Returns the resolved symbol. If that fails, simply return the address.
230 static int proc_pid_wchan(struct task_struct *task, char *buffer)
233 char symname[KSYM_NAME_LEN];
235 wchan = get_wchan(task);
237 if (lookup_symbol_name(wchan, symname) < 0)
238 if (!ptrace_may_access(task, PTRACE_MODE_READ))
241 return sprintf(buffer, "%lu", wchan);
243 return sprintf(buffer, "%s", symname);
245 #endif /* CONFIG_KALLSYMS */
247 static int lock_trace(struct task_struct *task)
249 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
252 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
253 mutex_unlock(&task->signal->cred_guard_mutex);
259 static void unlock_trace(struct task_struct *task)
261 mutex_unlock(&task->signal->cred_guard_mutex);
264 #ifdef CONFIG_STACKTRACE
266 #define MAX_STACK_TRACE_DEPTH 64
268 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
269 struct pid *pid, struct task_struct *task)
271 struct stack_trace trace;
272 unsigned long *entries;
276 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
280 trace.nr_entries = 0;
281 trace.max_entries = MAX_STACK_TRACE_DEPTH;
282 trace.entries = entries;
285 err = lock_trace(task);
287 save_stack_trace_tsk(task, &trace);
289 for (i = 0; i < trace.nr_entries; i++) {
290 seq_printf(m, "[<%pK>] %pS\n",
291 (void *)entries[i], (void *)entries[i]);
301 #ifdef CONFIG_SCHEDSTATS
303 * Provides /proc/PID/schedstat
305 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
307 return sprintf(buffer, "%llu %llu %lu\n",
308 (unsigned long long)task->se.sum_exec_runtime,
309 (unsigned long long)task->sched_info.run_delay,
310 task->sched_info.pcount);
314 #ifdef CONFIG_LATENCYTOP
315 static int lstats_show_proc(struct seq_file *m, void *v)
318 struct inode *inode = m->private;
319 struct task_struct *task = get_proc_task(inode);
323 seq_puts(m, "Latency Top version : v0.1\n");
324 for (i = 0; i < 32; i++) {
325 struct latency_record *lr = &task->latency_record[i];
326 if (lr->backtrace[0]) {
328 seq_printf(m, "%i %li %li",
329 lr->count, lr->time, lr->max);
330 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
331 unsigned long bt = lr->backtrace[q];
336 seq_printf(m, " %ps", (void *)bt);
342 put_task_struct(task);
346 static int lstats_open(struct inode *inode, struct file *file)
348 return single_open(file, lstats_show_proc, inode);
351 static ssize_t lstats_write(struct file *file, const char __user *buf,
352 size_t count, loff_t *offs)
354 struct task_struct *task = get_proc_task(file_inode(file));
358 clear_all_latency_tracing(task);
359 put_task_struct(task);
364 static const struct file_operations proc_lstats_operations = {
367 .write = lstats_write,
369 .release = single_release,
374 #ifdef CONFIG_CGROUPS
375 static int cgroup_open(struct inode *inode, struct file *file)
377 struct pid *pid = PROC_I(inode)->pid;
378 return single_open(file, proc_cgroup_show, pid);
381 static const struct file_operations proc_cgroup_operations = {
385 .release = single_release,
389 #ifdef CONFIG_PROC_PID_CPUSET
391 static int cpuset_open(struct inode *inode, struct file *file)
393 struct pid *pid = PROC_I(inode)->pid;
394 return single_open(file, proc_cpuset_show, pid);
397 static const struct file_operations proc_cpuset_operations = {
401 .release = single_release,
405 static int proc_oom_score(struct task_struct *task, char *buffer)
407 unsigned long totalpages = totalram_pages + total_swap_pages;
408 unsigned long points = 0;
410 read_lock(&tasklist_lock);
412 points = oom_badness(task, NULL, NULL, totalpages) *
414 read_unlock(&tasklist_lock);
415 return sprintf(buffer, "%lu\n", points);
423 static const struct limit_names lnames[RLIM_NLIMITS] = {
424 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
425 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
426 [RLIMIT_DATA] = {"Max data size", "bytes"},
427 [RLIMIT_STACK] = {"Max stack size", "bytes"},
428 [RLIMIT_CORE] = {"Max core file size", "bytes"},
429 [RLIMIT_RSS] = {"Max resident set", "bytes"},
430 [RLIMIT_NPROC] = {"Max processes", "processes"},
431 [RLIMIT_NOFILE] = {"Max open files", "files"},
432 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
433 [RLIMIT_AS] = {"Max address space", "bytes"},
434 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
435 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
436 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
437 [RLIMIT_NICE] = {"Max nice priority", NULL},
438 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
439 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
442 /* Display limits for a process */
443 static int proc_pid_limits(struct task_struct *task, char *buffer)
448 char *bufptr = buffer;
450 struct rlimit rlim[RLIM_NLIMITS];
452 if (!lock_task_sighand(task, &flags))
454 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
455 unlock_task_sighand(task, &flags);
458 * print the file header
460 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
461 "Limit", "Soft Limit", "Hard Limit", "Units");
463 for (i = 0; i < RLIM_NLIMITS; i++) {
464 if (rlim[i].rlim_cur == RLIM_INFINITY)
465 count += sprintf(&bufptr[count], "%-25s %-20s ",
466 lnames[i].name, "unlimited");
468 count += sprintf(&bufptr[count], "%-25s %-20lu ",
469 lnames[i].name, rlim[i].rlim_cur);
471 if (rlim[i].rlim_max == RLIM_INFINITY)
472 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
474 count += sprintf(&bufptr[count], "%-20lu ",
478 count += sprintf(&bufptr[count], "%-10s\n",
481 count += sprintf(&bufptr[count], "\n");
487 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
488 static int proc_pid_syscall(struct task_struct *task, char *buffer)
491 unsigned long args[6], sp, pc;
492 int res = lock_trace(task);
496 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
497 res = sprintf(buffer, "running\n");
499 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
501 res = sprintf(buffer,
502 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
504 args[0], args[1], args[2], args[3], args[4], args[5],
509 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
511 /************************************************************************/
512 /* Here the fs part begins */
513 /************************************************************************/
515 /* permission checks */
516 static int proc_fd_access_allowed(struct inode *inode)
518 struct task_struct *task;
520 /* Allow access to a task's file descriptors if it is us or we
521 * may use ptrace attach to the process and find out that
524 task = get_proc_task(inode);
526 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
527 put_task_struct(task);
532 int proc_setattr(struct dentry *dentry, struct iattr *attr)
535 struct inode *inode = dentry->d_inode;
537 if (attr->ia_valid & ATTR_MODE)
540 error = inode_change_ok(inode, attr);
544 setattr_copy(inode, attr);
545 mark_inode_dirty(inode);
550 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
551 * or euid/egid (for hide_pid_min=2)?
553 static bool has_pid_permissions(struct pid_namespace *pid,
554 struct task_struct *task,
557 if (pid->hide_pid < hide_pid_min)
559 if (in_group_p(pid->pid_gid))
561 return ptrace_may_access(task, PTRACE_MODE_READ);
565 static int proc_pid_permission(struct inode *inode, int mask)
567 struct pid_namespace *pid = inode->i_sb->s_fs_info;
568 struct task_struct *task;
571 task = get_proc_task(inode);
574 has_perms = has_pid_permissions(pid, task, 1);
575 put_task_struct(task);
578 if (pid->hide_pid == 2) {
580 * Let's make getdents(), stat(), and open()
581 * consistent with each other. If a process
582 * may not stat() a file, it shouldn't be seen
590 return generic_permission(inode, mask);
595 static const struct inode_operations proc_def_inode_operations = {
596 .setattr = proc_setattr,
599 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
601 static ssize_t proc_info_read(struct file * file, char __user * buf,
602 size_t count, loff_t *ppos)
604 struct inode * inode = file_inode(file);
607 struct task_struct *task = get_proc_task(inode);
613 if (count > PROC_BLOCK_SIZE)
614 count = PROC_BLOCK_SIZE;
617 if (!(page = __get_free_page(GFP_TEMPORARY)))
620 length = PROC_I(inode)->op.proc_read(task, (char*)page);
623 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
626 put_task_struct(task);
631 static const struct file_operations proc_info_file_operations = {
632 .read = proc_info_read,
633 .llseek = generic_file_llseek,
636 static int proc_single_show(struct seq_file *m, void *v)
638 struct inode *inode = m->private;
639 struct pid_namespace *ns;
641 struct task_struct *task;
644 ns = inode->i_sb->s_fs_info;
645 pid = proc_pid(inode);
646 task = get_pid_task(pid, PIDTYPE_PID);
650 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
652 put_task_struct(task);
656 static int proc_single_open(struct inode *inode, struct file *filp)
658 return single_open(filp, proc_single_show, inode);
661 static const struct file_operations proc_single_file_operations = {
662 .open = proc_single_open,
665 .release = single_release,
668 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
670 struct task_struct *task = get_proc_task(file_inode(file));
671 struct mm_struct *mm;
676 mm = mm_access(task, mode);
677 put_task_struct(task);
683 /* ensure this mm_struct can't be freed */
684 atomic_inc(&mm->mm_count);
685 /* but do not pin its memory */
689 file->private_data = mm;
694 static int mem_open(struct inode *inode, struct file *file)
696 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
698 /* OK to pass negative loff_t, we can catch out-of-range */
699 file->f_mode |= FMODE_UNSIGNED_OFFSET;
704 static ssize_t mem_rw(struct file *file, char __user *buf,
705 size_t count, loff_t *ppos, int write)
707 struct mm_struct *mm = file->private_data;
708 unsigned long addr = *ppos;
715 page = (char *)__get_free_page(GFP_TEMPORARY);
720 if (!atomic_inc_not_zero(&mm->mm_users))
724 int this_len = min_t(int, count, PAGE_SIZE);
726 if (write && copy_from_user(page, buf, this_len)) {
731 this_len = access_remote_vm(mm, addr, page, this_len, write);
738 if (!write && copy_to_user(buf, page, this_len)) {
752 free_page((unsigned long) page);
756 static ssize_t mem_read(struct file *file, char __user *buf,
757 size_t count, loff_t *ppos)
759 return mem_rw(file, buf, count, ppos, 0);
762 static ssize_t mem_write(struct file *file, const char __user *buf,
763 size_t count, loff_t *ppos)
765 return mem_rw(file, (char __user*)buf, count, ppos, 1);
768 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
772 file->f_pos = offset;
775 file->f_pos += offset;
780 force_successful_syscall_return();
784 static int mem_release(struct inode *inode, struct file *file)
786 struct mm_struct *mm = file->private_data;
792 static const struct file_operations proc_mem_operations = {
797 .release = mem_release,
800 static int environ_open(struct inode *inode, struct file *file)
802 return __mem_open(inode, file, PTRACE_MODE_READ);
805 static ssize_t environ_read(struct file *file, char __user *buf,
806 size_t count, loff_t *ppos)
809 unsigned long src = *ppos;
811 struct mm_struct *mm = file->private_data;
816 page = (char *)__get_free_page(GFP_TEMPORARY);
821 if (!atomic_inc_not_zero(&mm->mm_users))
824 size_t this_len, max_len;
827 if (src >= (mm->env_end - mm->env_start))
830 this_len = mm->env_end - (mm->env_start + src);
832 max_len = min_t(size_t, PAGE_SIZE, count);
833 this_len = min(max_len, this_len);
835 retval = access_remote_vm(mm, (mm->env_start + src),
843 if (copy_to_user(buf, page, retval)) {
857 free_page((unsigned long) page);
861 static const struct file_operations proc_environ_operations = {
862 .open = environ_open,
863 .read = environ_read,
864 .llseek = generic_file_llseek,
865 .release = mem_release,
868 static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
871 struct task_struct *task = get_proc_task(file_inode(file));
872 char buffer[PROC_NUMBUF];
873 int oom_adj = OOM_ADJUST_MIN;
879 if (lock_task_sighand(task, &flags)) {
880 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
881 oom_adj = OOM_ADJUST_MAX;
883 oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
885 unlock_task_sighand(task, &flags);
887 put_task_struct(task);
888 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
889 return simple_read_from_buffer(buf, count, ppos, buffer, len);
892 static ssize_t oom_adj_write(struct file *file, const char __user *buf,
893 size_t count, loff_t *ppos)
895 struct task_struct *task;
896 char buffer[PROC_NUMBUF];
901 memset(buffer, 0, sizeof(buffer));
902 if (count > sizeof(buffer) - 1)
903 count = sizeof(buffer) - 1;
904 if (copy_from_user(buffer, buf, count)) {
909 err = kstrtoint(strstrip(buffer), 0, &oom_adj);
912 if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
913 oom_adj != OOM_DISABLE) {
918 task = get_proc_task(file_inode(file));
930 if (!lock_task_sighand(task, &flags)) {
936 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
937 * value is always attainable.
939 if (oom_adj == OOM_ADJUST_MAX)
940 oom_adj = OOM_SCORE_ADJ_MAX;
942 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
944 if (oom_adj < task->signal->oom_score_adj &&
945 !capable(CAP_SYS_RESOURCE)) {
951 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
952 * /proc/pid/oom_score_adj instead.
954 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
955 current->comm, task_pid_nr(current), task_pid_nr(task),
958 task->signal->oom_score_adj = oom_adj;
959 trace_oom_score_adj_update(task);
961 unlock_task_sighand(task, &flags);
964 put_task_struct(task);
966 return err < 0 ? err : count;
969 static const struct file_operations proc_oom_adj_operations = {
970 .read = oom_adj_read,
971 .write = oom_adj_write,
972 .llseek = generic_file_llseek,
975 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
976 size_t count, loff_t *ppos)
978 struct task_struct *task = get_proc_task(file_inode(file));
979 char buffer[PROC_NUMBUF];
980 short oom_score_adj = OOM_SCORE_ADJ_MIN;
986 if (lock_task_sighand(task, &flags)) {
987 oom_score_adj = task->signal->oom_score_adj;
988 unlock_task_sighand(task, &flags);
990 put_task_struct(task);
991 len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
992 return simple_read_from_buffer(buf, count, ppos, buffer, len);
995 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
996 size_t count, loff_t *ppos)
998 struct task_struct *task;
999 char buffer[PROC_NUMBUF];
1000 unsigned long flags;
1004 memset(buffer, 0, sizeof(buffer));
1005 if (count > sizeof(buffer) - 1)
1006 count = sizeof(buffer) - 1;
1007 if (copy_from_user(buffer, buf, count)) {
1012 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1015 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1016 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1021 task = get_proc_task(file_inode(file));
1033 if (!lock_task_sighand(task, &flags)) {
1038 if ((short)oom_score_adj < task->signal->oom_score_adj_min &&
1039 !capable(CAP_SYS_RESOURCE)) {
1044 task->signal->oom_score_adj = (short)oom_score_adj;
1045 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1046 task->signal->oom_score_adj_min = (short)oom_score_adj;
1047 trace_oom_score_adj_update(task);
1050 unlock_task_sighand(task, &flags);
1053 put_task_struct(task);
1055 return err < 0 ? err : count;
1058 static const struct file_operations proc_oom_score_adj_operations = {
1059 .read = oom_score_adj_read,
1060 .write = oom_score_adj_write,
1061 .llseek = default_llseek,
1064 #ifdef CONFIG_AUDITSYSCALL
1065 #define TMPBUFLEN 21
1066 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1067 size_t count, loff_t *ppos)
1069 struct inode * inode = file_inode(file);
1070 struct task_struct *task = get_proc_task(inode);
1072 char tmpbuf[TMPBUFLEN];
1076 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1077 from_kuid(file->f_cred->user_ns,
1078 audit_get_loginuid(task)));
1079 put_task_struct(task);
1080 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1083 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1084 size_t count, loff_t *ppos)
1086 struct inode * inode = file_inode(file);
1093 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1099 if (count >= PAGE_SIZE)
1100 count = PAGE_SIZE - 1;
1103 /* No partial writes. */
1106 page = (char*)__get_free_page(GFP_TEMPORARY);
1110 if (copy_from_user(page, buf, count))
1114 loginuid = simple_strtoul(page, &tmp, 10);
1121 /* is userspace tring to explicitly UNSET the loginuid? */
1122 if (loginuid == AUDIT_UID_UNSET) {
1123 kloginuid = INVALID_UID;
1125 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1126 if (!uid_valid(kloginuid)) {
1132 length = audit_set_loginuid(kloginuid);
1133 if (likely(length == 0))
1137 free_page((unsigned long) page);
1141 static const struct file_operations proc_loginuid_operations = {
1142 .read = proc_loginuid_read,
1143 .write = proc_loginuid_write,
1144 .llseek = generic_file_llseek,
1147 static ssize_t proc_sessionid_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 audit_get_sessionid(task));
1159 put_task_struct(task);
1160 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1163 static const struct file_operations proc_sessionid_operations = {
1164 .read = proc_sessionid_read,
1165 .llseek = generic_file_llseek,
1169 #ifdef CONFIG_FAULT_INJECTION
1170 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1171 size_t count, loff_t *ppos)
1173 struct task_struct *task = get_proc_task(file_inode(file));
1174 char buffer[PROC_NUMBUF];
1180 make_it_fail = task->make_it_fail;
1181 put_task_struct(task);
1183 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1185 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1188 static ssize_t proc_fault_inject_write(struct file * file,
1189 const char __user * buf, size_t count, loff_t *ppos)
1191 struct task_struct *task;
1192 char buffer[PROC_NUMBUF], *end;
1195 if (!capable(CAP_SYS_RESOURCE))
1197 memset(buffer, 0, sizeof(buffer));
1198 if (count > sizeof(buffer) - 1)
1199 count = sizeof(buffer) - 1;
1200 if (copy_from_user(buffer, buf, count))
1202 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1205 if (make_it_fail < 0 || make_it_fail > 1)
1208 task = get_proc_task(file_inode(file));
1211 task->make_it_fail = make_it_fail;
1212 put_task_struct(task);
1217 static const struct file_operations proc_fault_inject_operations = {
1218 .read = proc_fault_inject_read,
1219 .write = proc_fault_inject_write,
1220 .llseek = generic_file_llseek,
1225 #ifdef CONFIG_SCHED_DEBUG
1227 * Print out various scheduling related per-task fields:
1229 static int sched_show(struct seq_file *m, void *v)
1231 struct inode *inode = m->private;
1232 struct task_struct *p;
1234 p = get_proc_task(inode);
1237 proc_sched_show_task(p, m);
1245 sched_write(struct file *file, const char __user *buf,
1246 size_t count, loff_t *offset)
1248 struct inode *inode = file_inode(file);
1249 struct task_struct *p;
1251 p = get_proc_task(inode);
1254 proc_sched_set_task(p);
1261 static int sched_open(struct inode *inode, struct file *filp)
1263 return single_open(filp, sched_show, inode);
1266 static const struct file_operations proc_pid_sched_operations = {
1269 .write = sched_write,
1270 .llseek = seq_lseek,
1271 .release = single_release,
1276 #ifdef CONFIG_SCHED_AUTOGROUP
1278 * Print out autogroup related information:
1280 static int sched_autogroup_show(struct seq_file *m, void *v)
1282 struct inode *inode = m->private;
1283 struct task_struct *p;
1285 p = get_proc_task(inode);
1288 proc_sched_autogroup_show_task(p, m);
1296 sched_autogroup_write(struct file *file, const char __user *buf,
1297 size_t count, loff_t *offset)
1299 struct inode *inode = file_inode(file);
1300 struct task_struct *p;
1301 char buffer[PROC_NUMBUF];
1305 memset(buffer, 0, sizeof(buffer));
1306 if (count > sizeof(buffer) - 1)
1307 count = sizeof(buffer) - 1;
1308 if (copy_from_user(buffer, buf, count))
1311 err = kstrtoint(strstrip(buffer), 0, &nice);
1315 p = get_proc_task(inode);
1319 err = proc_sched_autogroup_set_nice(p, nice);
1328 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1332 ret = single_open(filp, sched_autogroup_show, NULL);
1334 struct seq_file *m = filp->private_data;
1341 static const struct file_operations proc_pid_sched_autogroup_operations = {
1342 .open = sched_autogroup_open,
1344 .write = sched_autogroup_write,
1345 .llseek = seq_lseek,
1346 .release = single_release,
1349 #endif /* CONFIG_SCHED_AUTOGROUP */
1351 static ssize_t comm_write(struct file *file, const char __user *buf,
1352 size_t count, loff_t *offset)
1354 struct inode *inode = file_inode(file);
1355 struct task_struct *p;
1356 char buffer[TASK_COMM_LEN];
1357 const size_t maxlen = sizeof(buffer) - 1;
1359 memset(buffer, 0, sizeof(buffer));
1360 if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
1363 p = get_proc_task(inode);
1367 if (same_thread_group(current, p))
1368 set_task_comm(p, buffer);
1377 static int comm_show(struct seq_file *m, void *v)
1379 struct inode *inode = m->private;
1380 struct task_struct *p;
1382 p = get_proc_task(inode);
1387 seq_printf(m, "%s\n", p->comm);
1395 static int comm_open(struct inode *inode, struct file *filp)
1397 return single_open(filp, comm_show, inode);
1400 static const struct file_operations proc_pid_set_comm_operations = {
1403 .write = comm_write,
1404 .llseek = seq_lseek,
1405 .release = single_release,
1408 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1410 struct task_struct *task;
1411 struct mm_struct *mm;
1412 struct file *exe_file;
1414 task = get_proc_task(dentry->d_inode);
1417 mm = get_task_mm(task);
1418 put_task_struct(task);
1421 exe_file = get_mm_exe_file(mm);
1424 *exe_path = exe_file->f_path;
1425 path_get(&exe_file->f_path);
1432 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1434 struct inode *inode = dentry->d_inode;
1436 int error = -EACCES;
1438 /* Are we allowed to snoop on the tasks file descriptors? */
1439 if (!proc_fd_access_allowed(inode))
1442 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1446 nd_jump_link(nd, &path);
1449 return ERR_PTR(error);
1452 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1454 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1461 pathname = d_path(path, tmp, PAGE_SIZE);
1462 len = PTR_ERR(pathname);
1463 if (IS_ERR(pathname))
1465 len = tmp + PAGE_SIZE - 1 - pathname;
1469 if (copy_to_user(buffer, pathname, len))
1472 free_page((unsigned long)tmp);
1476 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1478 int error = -EACCES;
1479 struct inode *inode = dentry->d_inode;
1482 /* Are we allowed to snoop on the tasks file descriptors? */
1483 if (!proc_fd_access_allowed(inode))
1486 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1490 error = do_proc_readlink(&path, buffer, buflen);
1496 const struct inode_operations proc_pid_link_inode_operations = {
1497 .readlink = proc_pid_readlink,
1498 .follow_link = proc_pid_follow_link,
1499 .setattr = proc_setattr,
1503 /* building an inode */
1505 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1507 struct inode * inode;
1508 struct proc_inode *ei;
1509 const struct cred *cred;
1511 /* We need a new inode */
1513 inode = new_inode(sb);
1519 inode->i_ino = get_next_ino();
1520 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1521 inode->i_op = &proc_def_inode_operations;
1524 * grab the reference to task.
1526 ei->pid = get_task_pid(task, PIDTYPE_PID);
1530 if (task_dumpable(task)) {
1532 cred = __task_cred(task);
1533 inode->i_uid = cred->euid;
1534 inode->i_gid = cred->egid;
1537 security_task_to_inode(task, inode);
1547 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1549 struct inode *inode = dentry->d_inode;
1550 struct task_struct *task;
1551 const struct cred *cred;
1552 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
1554 generic_fillattr(inode, stat);
1557 stat->uid = GLOBAL_ROOT_UID;
1558 stat->gid = GLOBAL_ROOT_GID;
1559 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1561 if (!has_pid_permissions(pid, task, 2)) {
1564 * This doesn't prevent learning whether PID exists,
1565 * it only makes getattr() consistent with readdir().
1569 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1570 task_dumpable(task)) {
1571 cred = __task_cred(task);
1572 stat->uid = cred->euid;
1573 stat->gid = cred->egid;
1583 * Exceptional case: normally we are not allowed to unhash a busy
1584 * directory. In this case, however, we can do it - no aliasing problems
1585 * due to the way we treat inodes.
1587 * Rewrite the inode's ownerships here because the owning task may have
1588 * performed a setuid(), etc.
1590 * Before the /proc/pid/status file was created the only way to read
1591 * the effective uid of a /process was to stat /proc/pid. Reading
1592 * /proc/pid/status is slow enough that procps and other packages
1593 * kept stating /proc/pid. To keep the rules in /proc simple I have
1594 * made this apply to all per process world readable and executable
1597 int pid_revalidate(struct dentry *dentry, unsigned int flags)
1599 struct inode *inode;
1600 struct task_struct *task;
1601 const struct cred *cred;
1603 if (flags & LOOKUP_RCU)
1606 inode = dentry->d_inode;
1607 task = get_proc_task(inode);
1610 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1611 task_dumpable(task)) {
1613 cred = __task_cred(task);
1614 inode->i_uid = cred->euid;
1615 inode->i_gid = cred->egid;
1618 inode->i_uid = GLOBAL_ROOT_UID;
1619 inode->i_gid = GLOBAL_ROOT_GID;
1621 inode->i_mode &= ~(S_ISUID | S_ISGID);
1622 security_task_to_inode(task, inode);
1623 put_task_struct(task);
1630 static inline bool proc_inode_is_dead(struct inode *inode)
1632 return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1635 int pid_delete_dentry(const struct dentry *dentry)
1637 /* Is the task we represent dead?
1638 * If so, then don't put the dentry on the lru list,
1639 * kill it immediately.
1641 return proc_inode_is_dead(dentry->d_inode);
1644 const struct dentry_operations pid_dentry_operations =
1646 .d_revalidate = pid_revalidate,
1647 .d_delete = pid_delete_dentry,
1653 * Fill a directory entry.
1655 * If possible create the dcache entry and derive our inode number and
1656 * file type from dcache entry.
1658 * Since all of the proc inode numbers are dynamically generated, the inode
1659 * numbers do not exist until the inode is cache. This means creating the
1660 * the dcache entry in readdir is necessary to keep the inode numbers
1661 * reported by readdir in sync with the inode numbers reported
1664 bool proc_fill_cache(struct file *file, struct dir_context *ctx,
1665 const char *name, int len,
1666 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1668 struct dentry *child, *dir = file->f_path.dentry;
1669 struct qstr qname = QSTR_INIT(name, len);
1670 struct inode *inode;
1674 child = d_hash_and_lookup(dir, &qname);
1676 child = d_alloc(dir, &qname);
1678 goto end_instantiate;
1679 if (instantiate(dir->d_inode, child, task, ptr) < 0) {
1681 goto end_instantiate;
1684 inode = child->d_inode;
1686 type = inode->i_mode >> 12;
1688 return dir_emit(ctx, name, len, ino, type);
1691 return dir_emit(ctx, name, len, 1, DT_UNKNOWN);
1694 #ifdef CONFIG_CHECKPOINT_RESTORE
1697 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1698 * which represent vma start and end addresses.
1700 static int dname_to_vma_addr(struct dentry *dentry,
1701 unsigned long *start, unsigned long *end)
1703 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1709 static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
1711 unsigned long vm_start, vm_end;
1712 bool exact_vma_exists = false;
1713 struct mm_struct *mm = NULL;
1714 struct task_struct *task;
1715 const struct cred *cred;
1716 struct inode *inode;
1719 if (flags & LOOKUP_RCU)
1722 if (!capable(CAP_SYS_ADMIN)) {
1727 inode = dentry->d_inode;
1728 task = get_proc_task(inode);
1732 mm = mm_access(task, PTRACE_MODE_READ);
1733 if (IS_ERR_OR_NULL(mm))
1736 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1737 down_read(&mm->mmap_sem);
1738 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1739 up_read(&mm->mmap_sem);
1744 if (exact_vma_exists) {
1745 if (task_dumpable(task)) {
1747 cred = __task_cred(task);
1748 inode->i_uid = cred->euid;
1749 inode->i_gid = cred->egid;
1752 inode->i_uid = GLOBAL_ROOT_UID;
1753 inode->i_gid = GLOBAL_ROOT_GID;
1755 security_task_to_inode(task, inode);
1760 put_task_struct(task);
1769 static const struct dentry_operations tid_map_files_dentry_operations = {
1770 .d_revalidate = map_files_d_revalidate,
1771 .d_delete = pid_delete_dentry,
1774 static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
1776 unsigned long vm_start, vm_end;
1777 struct vm_area_struct *vma;
1778 struct task_struct *task;
1779 struct mm_struct *mm;
1783 task = get_proc_task(dentry->d_inode);
1787 mm = get_task_mm(task);
1788 put_task_struct(task);
1792 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
1797 down_read(&mm->mmap_sem);
1798 vma = find_exact_vma(mm, vm_start, vm_end);
1799 if (vma && vma->vm_file) {
1800 *path = vma->vm_file->f_path;
1804 up_read(&mm->mmap_sem);
1812 struct map_files_info {
1815 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
1819 proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
1820 struct task_struct *task, const void *ptr)
1822 fmode_t mode = (fmode_t)(unsigned long)ptr;
1823 struct proc_inode *ei;
1824 struct inode *inode;
1826 inode = proc_pid_make_inode(dir->i_sb, task);
1831 ei->op.proc_get_link = proc_map_files_get_link;
1833 inode->i_op = &proc_pid_link_inode_operations;
1835 inode->i_mode = S_IFLNK;
1837 if (mode & FMODE_READ)
1838 inode->i_mode |= S_IRUSR;
1839 if (mode & FMODE_WRITE)
1840 inode->i_mode |= S_IWUSR;
1842 d_set_d_op(dentry, &tid_map_files_dentry_operations);
1843 d_add(dentry, inode);
1848 static struct dentry *proc_map_files_lookup(struct inode *dir,
1849 struct dentry *dentry, unsigned int flags)
1851 unsigned long vm_start, vm_end;
1852 struct vm_area_struct *vma;
1853 struct task_struct *task;
1855 struct mm_struct *mm;
1858 if (!capable(CAP_SYS_ADMIN))
1862 task = get_proc_task(dir);
1867 if (!ptrace_may_access(task, PTRACE_MODE_READ))
1871 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
1874 mm = get_task_mm(task);
1878 down_read(&mm->mmap_sem);
1879 vma = find_exact_vma(mm, vm_start, vm_end);
1884 result = proc_map_files_instantiate(dir, dentry, task,
1885 (void *)(unsigned long)vma->vm_file->f_mode);
1888 up_read(&mm->mmap_sem);
1891 put_task_struct(task);
1893 return ERR_PTR(result);
1896 static const struct inode_operations proc_map_files_inode_operations = {
1897 .lookup = proc_map_files_lookup,
1898 .permission = proc_fd_permission,
1899 .setattr = proc_setattr,
1903 proc_map_files_readdir(struct file *file, struct dir_context *ctx)
1905 struct vm_area_struct *vma;
1906 struct task_struct *task;
1907 struct mm_struct *mm;
1908 unsigned long nr_files, pos, i;
1909 struct flex_array *fa = NULL;
1910 struct map_files_info info;
1911 struct map_files_info *p;
1915 if (!capable(CAP_SYS_ADMIN))
1919 task = get_proc_task(file_inode(file));
1924 if (!ptrace_may_access(task, PTRACE_MODE_READ))
1928 if (!dir_emit_dots(file, ctx))
1931 mm = get_task_mm(task);
1934 down_read(&mm->mmap_sem);
1939 * We need two passes here:
1941 * 1) Collect vmas of mapped files with mmap_sem taken
1942 * 2) Release mmap_sem and instantiate entries
1944 * otherwise we get lockdep complained, since filldir()
1945 * routine might require mmap_sem taken in might_fault().
1948 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
1949 if (vma->vm_file && ++pos > ctx->pos)
1954 fa = flex_array_alloc(sizeof(info), nr_files,
1956 if (!fa || flex_array_prealloc(fa, 0, nr_files,
1960 flex_array_free(fa);
1961 up_read(&mm->mmap_sem);
1965 for (i = 0, vma = mm->mmap, pos = 2; vma;
1966 vma = vma->vm_next) {
1969 if (++pos <= ctx->pos)
1972 info.mode = vma->vm_file->f_mode;
1973 info.len = snprintf(info.name,
1974 sizeof(info.name), "%lx-%lx",
1975 vma->vm_start, vma->vm_end);
1976 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
1980 up_read(&mm->mmap_sem);
1982 for (i = 0; i < nr_files; i++) {
1983 p = flex_array_get(fa, i);
1984 if (!proc_fill_cache(file, ctx,
1986 proc_map_files_instantiate,
1988 (void *)(unsigned long)p->mode))
1993 flex_array_free(fa);
1997 put_task_struct(task);
2002 static const struct file_operations proc_map_files_operations = {
2003 .read = generic_read_dir,
2004 .iterate = proc_map_files_readdir,
2005 .llseek = default_llseek,
2008 struct timers_private {
2010 struct task_struct *task;
2011 struct sighand_struct *sighand;
2012 struct pid_namespace *ns;
2013 unsigned long flags;
2016 static void *timers_start(struct seq_file *m, loff_t *pos)
2018 struct timers_private *tp = m->private;
2020 tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2022 return ERR_PTR(-ESRCH);
2024 tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2026 return ERR_PTR(-ESRCH);
2028 return seq_list_start(&tp->task->signal->posix_timers, *pos);
2031 static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2033 struct timers_private *tp = m->private;
2034 return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2037 static void timers_stop(struct seq_file *m, void *v)
2039 struct timers_private *tp = m->private;
2042 unlock_task_sighand(tp->task, &tp->flags);
2047 put_task_struct(tp->task);
2052 static int show_timer(struct seq_file *m, void *v)
2054 struct k_itimer *timer;
2055 struct timers_private *tp = m->private;
2057 static const char * const nstr[] = {
2058 [SIGEV_SIGNAL] = "signal",
2059 [SIGEV_NONE] = "none",
2060 [SIGEV_THREAD] = "thread",
2063 timer = list_entry((struct list_head *)v, struct k_itimer, list);
2064 notify = timer->it_sigev_notify;
2066 seq_printf(m, "ID: %d\n", timer->it_id);
2067 seq_printf(m, "signal: %d/%p\n", timer->sigq->info.si_signo,
2068 timer->sigq->info.si_value.sival_ptr);
2069 seq_printf(m, "notify: %s/%s.%d\n",
2070 nstr[notify & ~SIGEV_THREAD_ID],
2071 (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2072 pid_nr_ns(timer->it_pid, tp->ns));
2073 seq_printf(m, "ClockID: %d\n", timer->it_clock);
2078 static const struct seq_operations proc_timers_seq_ops = {
2079 .start = timers_start,
2080 .next = timers_next,
2081 .stop = timers_stop,
2085 static int proc_timers_open(struct inode *inode, struct file *file)
2087 struct timers_private *tp;
2089 tp = __seq_open_private(file, &proc_timers_seq_ops,
2090 sizeof(struct timers_private));
2094 tp->pid = proc_pid(inode);
2095 tp->ns = inode->i_sb->s_fs_info;
2099 static const struct file_operations proc_timers_operations = {
2100 .open = proc_timers_open,
2102 .llseek = seq_lseek,
2103 .release = seq_release_private,
2105 #endif /* CONFIG_CHECKPOINT_RESTORE */
2107 static int proc_pident_instantiate(struct inode *dir,
2108 struct dentry *dentry, struct task_struct *task, const void *ptr)
2110 const struct pid_entry *p = ptr;
2111 struct inode *inode;
2112 struct proc_inode *ei;
2114 inode = proc_pid_make_inode(dir->i_sb, task);
2119 inode->i_mode = p->mode;
2120 if (S_ISDIR(inode->i_mode))
2121 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2123 inode->i_op = p->iop;
2125 inode->i_fop = p->fop;
2127 d_set_d_op(dentry, &pid_dentry_operations);
2128 d_add(dentry, inode);
2129 /* Close the race of the process dying before we return the dentry */
2130 if (pid_revalidate(dentry, 0))
2136 static struct dentry *proc_pident_lookup(struct inode *dir,
2137 struct dentry *dentry,
2138 const struct pid_entry *ents,
2142 struct task_struct *task = get_proc_task(dir);
2143 const struct pid_entry *p, *last;
2151 * Yes, it does not scale. And it should not. Don't add
2152 * new entries into /proc/<tgid>/ without very good reasons.
2154 last = &ents[nents - 1];
2155 for (p = ents; p <= last; p++) {
2156 if (p->len != dentry->d_name.len)
2158 if (!memcmp(dentry->d_name.name, p->name, p->len))
2164 error = proc_pident_instantiate(dir, dentry, task, p);
2166 put_task_struct(task);
2168 return ERR_PTR(error);
2171 static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
2172 const struct pid_entry *ents, unsigned int nents)
2174 struct task_struct *task = get_proc_task(file_inode(file));
2175 const struct pid_entry *p;
2180 if (!dir_emit_dots(file, ctx))
2183 if (ctx->pos >= nents + 2)
2186 for (p = ents + (ctx->pos - 2); p <= ents + nents - 1; p++) {
2187 if (!proc_fill_cache(file, ctx, p->name, p->len,
2188 proc_pident_instantiate, task, p))
2193 put_task_struct(task);
2197 #ifdef CONFIG_SECURITY
2198 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2199 size_t count, loff_t *ppos)
2201 struct inode * inode = file_inode(file);
2204 struct task_struct *task = get_proc_task(inode);
2209 length = security_getprocattr(task,
2210 (char*)file->f_path.dentry->d_name.name,
2212 put_task_struct(task);
2214 length = simple_read_from_buffer(buf, count, ppos, p, length);
2219 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2220 size_t count, loff_t *ppos)
2222 struct inode * inode = file_inode(file);
2225 struct task_struct *task = get_proc_task(inode);
2230 if (count > PAGE_SIZE)
2233 /* No partial writes. */
2239 page = (char*)__get_free_page(GFP_TEMPORARY);
2244 if (copy_from_user(page, buf, count))
2247 /* Guard against adverse ptrace interaction */
2248 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2252 length = security_setprocattr(task,
2253 (char*)file->f_path.dentry->d_name.name,
2254 (void*)page, count);
2255 mutex_unlock(&task->signal->cred_guard_mutex);
2257 free_page((unsigned long) page);
2259 put_task_struct(task);
2264 static const struct file_operations proc_pid_attr_operations = {
2265 .read = proc_pid_attr_read,
2266 .write = proc_pid_attr_write,
2267 .llseek = generic_file_llseek,
2270 static const struct pid_entry attr_dir_stuff[] = {
2271 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2272 REG("prev", S_IRUGO, proc_pid_attr_operations),
2273 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2274 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2275 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2276 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2279 static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
2281 return proc_pident_readdir(file, ctx,
2282 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2285 static const struct file_operations proc_attr_dir_operations = {
2286 .read = generic_read_dir,
2287 .iterate = proc_attr_dir_readdir,
2288 .llseek = default_llseek,
2291 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2292 struct dentry *dentry, unsigned int flags)
2294 return proc_pident_lookup(dir, dentry,
2295 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2298 static const struct inode_operations proc_attr_dir_inode_operations = {
2299 .lookup = proc_attr_dir_lookup,
2300 .getattr = pid_getattr,
2301 .setattr = proc_setattr,
2306 #ifdef CONFIG_ELF_CORE
2307 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2308 size_t count, loff_t *ppos)
2310 struct task_struct *task = get_proc_task(file_inode(file));
2311 struct mm_struct *mm;
2312 char buffer[PROC_NUMBUF];
2320 mm = get_task_mm(task);
2322 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2323 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2324 MMF_DUMP_FILTER_SHIFT));
2326 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2329 put_task_struct(task);
2334 static ssize_t proc_coredump_filter_write(struct file *file,
2335 const char __user *buf,
2339 struct task_struct *task;
2340 struct mm_struct *mm;
2341 char buffer[PROC_NUMBUF], *end;
2348 memset(buffer, 0, sizeof(buffer));
2349 if (count > sizeof(buffer) - 1)
2350 count = sizeof(buffer) - 1;
2351 if (copy_from_user(buffer, buf, count))
2355 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2358 if (end - buffer == 0)
2362 task = get_proc_task(file_inode(file));
2367 mm = get_task_mm(task);
2371 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2373 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2375 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2380 put_task_struct(task);
2385 static const struct file_operations proc_coredump_filter_operations = {
2386 .read = proc_coredump_filter_read,
2387 .write = proc_coredump_filter_write,
2388 .llseek = generic_file_llseek,
2392 #ifdef CONFIG_TASK_IO_ACCOUNTING
2393 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2395 struct task_io_accounting acct = task->ioac;
2396 unsigned long flags;
2399 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2403 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2408 if (whole && lock_task_sighand(task, &flags)) {
2409 struct task_struct *t = task;
2411 task_io_accounting_add(&acct, &task->signal->ioac);
2412 while_each_thread(task, t)
2413 task_io_accounting_add(&acct, &t->ioac);
2415 unlock_task_sighand(task, &flags);
2417 result = sprintf(buffer,
2422 "read_bytes: %llu\n"
2423 "write_bytes: %llu\n"
2424 "cancelled_write_bytes: %llu\n",
2425 (unsigned long long)acct.rchar,
2426 (unsigned long long)acct.wchar,
2427 (unsigned long long)acct.syscr,
2428 (unsigned long long)acct.syscw,
2429 (unsigned long long)acct.read_bytes,
2430 (unsigned long long)acct.write_bytes,
2431 (unsigned long long)acct.cancelled_write_bytes);
2433 mutex_unlock(&task->signal->cred_guard_mutex);
2437 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2439 return do_io_accounting(task, buffer, 0);
2442 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2444 return do_io_accounting(task, buffer, 1);
2446 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2448 #ifdef CONFIG_USER_NS
2449 static int proc_id_map_open(struct inode *inode, struct file *file,
2450 const struct seq_operations *seq_ops)
2452 struct user_namespace *ns = NULL;
2453 struct task_struct *task;
2454 struct seq_file *seq;
2457 task = get_proc_task(inode);
2460 ns = get_user_ns(task_cred_xxx(task, user_ns));
2462 put_task_struct(task);
2467 ret = seq_open(file, seq_ops);
2471 seq = file->private_data;
2481 static int proc_id_map_release(struct inode *inode, struct file *file)
2483 struct seq_file *seq = file->private_data;
2484 struct user_namespace *ns = seq->private;
2486 return seq_release(inode, file);
2489 static int proc_uid_map_open(struct inode *inode, struct file *file)
2491 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2494 static int proc_gid_map_open(struct inode *inode, struct file *file)
2496 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2499 static int proc_projid_map_open(struct inode *inode, struct file *file)
2501 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2504 static const struct file_operations proc_uid_map_operations = {
2505 .open = proc_uid_map_open,
2506 .write = proc_uid_map_write,
2508 .llseek = seq_lseek,
2509 .release = proc_id_map_release,
2512 static const struct file_operations proc_gid_map_operations = {
2513 .open = proc_gid_map_open,
2514 .write = proc_gid_map_write,
2516 .llseek = seq_lseek,
2517 .release = proc_id_map_release,
2520 static const struct file_operations proc_projid_map_operations = {
2521 .open = proc_projid_map_open,
2522 .write = proc_projid_map_write,
2524 .llseek = seq_lseek,
2525 .release = proc_id_map_release,
2527 #endif /* CONFIG_USER_NS */
2529 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2530 struct pid *pid, struct task_struct *task)
2532 int err = lock_trace(task);
2534 seq_printf(m, "%08x\n", task->personality);
2543 static const struct file_operations proc_task_operations;
2544 static const struct inode_operations proc_task_inode_operations;
2546 static const struct pid_entry tgid_base_stuff[] = {
2547 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2548 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2549 #ifdef CONFIG_CHECKPOINT_RESTORE
2550 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
2552 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2553 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2555 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2557 REG("environ", S_IRUSR, proc_environ_operations),
2558 ONE("auxv", S_IRUSR, proc_pid_auxv),
2559 ONE("status", S_IRUGO, proc_pid_status),
2560 ONE("personality", S_IRUSR, proc_pid_personality),
2561 INF("limits", S_IRUGO, proc_pid_limits),
2562 #ifdef CONFIG_SCHED_DEBUG
2563 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2565 #ifdef CONFIG_SCHED_AUTOGROUP
2566 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2568 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2569 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2570 INF("syscall", S_IRUSR, proc_pid_syscall),
2572 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2573 ONE("stat", S_IRUGO, proc_tgid_stat),
2574 ONE("statm", S_IRUGO, proc_pid_statm),
2575 REG("maps", S_IRUGO, proc_pid_maps_operations),
2577 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
2579 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2580 LNK("cwd", proc_cwd_link),
2581 LNK("root", proc_root_link),
2582 LNK("exe", proc_exe_link),
2583 REG("mounts", S_IRUGO, proc_mounts_operations),
2584 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2585 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2586 #ifdef CONFIG_PROC_PAGE_MONITOR
2587 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2588 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
2589 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2591 #ifdef CONFIG_SECURITY
2592 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2594 #ifdef CONFIG_KALLSYMS
2595 INF("wchan", S_IRUGO, proc_pid_wchan),
2597 #ifdef CONFIG_STACKTRACE
2598 ONE("stack", S_IRUSR, proc_pid_stack),
2600 #ifdef CONFIG_SCHEDSTATS
2601 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2603 #ifdef CONFIG_LATENCYTOP
2604 REG("latency", S_IRUGO, proc_lstats_operations),
2606 #ifdef CONFIG_PROC_PID_CPUSET
2607 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2609 #ifdef CONFIG_CGROUPS
2610 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2612 INF("oom_score", S_IRUGO, proc_oom_score),
2613 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
2614 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2615 #ifdef CONFIG_AUDITSYSCALL
2616 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2617 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2619 #ifdef CONFIG_FAULT_INJECTION
2620 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2622 #ifdef CONFIG_ELF_CORE
2623 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2625 #ifdef CONFIG_TASK_IO_ACCOUNTING
2626 INF("io", S_IRUSR, proc_tgid_io_accounting),
2628 #ifdef CONFIG_HARDWALL
2629 INF("hardwall", S_IRUGO, proc_pid_hardwall),
2631 #ifdef CONFIG_USER_NS
2632 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
2633 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
2634 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
2636 #ifdef CONFIG_CHECKPOINT_RESTORE
2637 REG("timers", S_IRUGO, proc_timers_operations),
2641 static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
2643 return proc_pident_readdir(file, ctx,
2644 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2647 static const struct file_operations proc_tgid_base_operations = {
2648 .read = generic_read_dir,
2649 .iterate = proc_tgid_base_readdir,
2650 .llseek = default_llseek,
2653 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2655 return proc_pident_lookup(dir, dentry,
2656 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2659 static const struct inode_operations proc_tgid_base_inode_operations = {
2660 .lookup = proc_tgid_base_lookup,
2661 .getattr = pid_getattr,
2662 .setattr = proc_setattr,
2663 .permission = proc_pid_permission,
2666 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2668 struct dentry *dentry, *leader, *dir;
2669 char buf[PROC_NUMBUF];
2673 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2674 /* no ->d_hash() rejects on procfs */
2675 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2677 shrink_dcache_parent(dentry);
2683 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2684 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2689 name.len = strlen(name.name);
2690 dir = d_hash_and_lookup(leader, &name);
2692 goto out_put_leader;
2695 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2696 dentry = d_hash_and_lookup(dir, &name);
2698 shrink_dcache_parent(dentry);
2711 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2712 * @task: task that should be flushed.
2714 * When flushing dentries from proc, one needs to flush them from global
2715 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2716 * in. This call is supposed to do all of this job.
2718 * Looks in the dcache for
2720 * /proc/@tgid/task/@pid
2721 * if either directory is present flushes it and all of it'ts children
2724 * It is safe and reasonable to cache /proc entries for a task until
2725 * that task exits. After that they just clog up the dcache with
2726 * useless entries, possibly causing useful dcache entries to be
2727 * flushed instead. This routine is proved to flush those useless
2728 * dcache entries at process exit time.
2730 * NOTE: This routine is just an optimization so it does not guarantee
2731 * that no dcache entries will exist at process exit time it
2732 * just makes it very unlikely that any will persist.
2735 void proc_flush_task(struct task_struct *task)
2738 struct pid *pid, *tgid;
2741 pid = task_pid(task);
2742 tgid = task_tgid(task);
2744 for (i = 0; i <= pid->level; i++) {
2745 upid = &pid->numbers[i];
2746 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2747 tgid->numbers[i].nr);
2751 static int proc_pid_instantiate(struct inode *dir,
2752 struct dentry * dentry,
2753 struct task_struct *task, const void *ptr)
2755 struct inode *inode;
2757 inode = proc_pid_make_inode(dir->i_sb, task);
2761 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2762 inode->i_op = &proc_tgid_base_inode_operations;
2763 inode->i_fop = &proc_tgid_base_operations;
2764 inode->i_flags|=S_IMMUTABLE;
2766 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
2767 ARRAY_SIZE(tgid_base_stuff)));
2769 d_set_d_op(dentry, &pid_dentry_operations);
2771 d_add(dentry, inode);
2772 /* Close the race of the process dying before we return the dentry */
2773 if (pid_revalidate(dentry, 0))
2779 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
2781 int result = -ENOENT;
2782 struct task_struct *task;
2784 struct pid_namespace *ns;
2786 tgid = name_to_int(&dentry->d_name);
2790 ns = dentry->d_sb->s_fs_info;
2792 task = find_task_by_pid_ns(tgid, ns);
2794 get_task_struct(task);
2799 result = proc_pid_instantiate(dir, dentry, task, NULL);
2800 put_task_struct(task);
2802 return ERR_PTR(result);
2806 * Find the first task with tgid >= tgid
2811 struct task_struct *task;
2813 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2818 put_task_struct(iter.task);
2822 pid = find_ge_pid(iter.tgid, ns);
2824 iter.tgid = pid_nr_ns(pid, ns);
2825 iter.task = pid_task(pid, PIDTYPE_PID);
2826 /* What we to know is if the pid we have find is the
2827 * pid of a thread_group_leader. Testing for task
2828 * being a thread_group_leader is the obvious thing
2829 * todo but there is a window when it fails, due to
2830 * the pid transfer logic in de_thread.
2832 * So we perform the straight forward test of seeing
2833 * if the pid we have found is the pid of a thread
2834 * group leader, and don't worry if the task we have
2835 * found doesn't happen to be a thread group leader.
2836 * As we don't care in the case of readdir.
2838 if (!iter.task || !has_group_leader_pid(iter.task)) {
2842 get_task_struct(iter.task);
2848 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 1)
2850 /* for the /proc/ directory itself, after non-process stuff has been done */
2851 int proc_pid_readdir(struct file *file, struct dir_context *ctx)
2853 struct tgid_iter iter;
2854 struct pid_namespace *ns = file->f_dentry->d_sb->s_fs_info;
2855 loff_t pos = ctx->pos;
2857 if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
2860 if (pos == TGID_OFFSET - 1) {
2861 struct inode *inode = ns->proc_self->d_inode;
2862 if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
2866 iter.tgid = pos - TGID_OFFSET;
2869 for (iter = next_tgid(ns, iter);
2871 iter.tgid += 1, iter = next_tgid(ns, iter)) {
2872 char name[PROC_NUMBUF];
2874 if (!has_pid_permissions(ns, iter.task, 2))
2877 len = snprintf(name, sizeof(name), "%d", iter.tgid);
2878 ctx->pos = iter.tgid + TGID_OFFSET;
2879 if (!proc_fill_cache(file, ctx, name, len,
2880 proc_pid_instantiate, iter.task, NULL)) {
2881 put_task_struct(iter.task);
2885 ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
2892 static const struct pid_entry tid_base_stuff[] = {
2893 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2894 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2895 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2896 REG("environ", S_IRUSR, proc_environ_operations),
2897 ONE("auxv", S_IRUSR, proc_pid_auxv),
2898 ONE("status", S_IRUGO, proc_pid_status),
2899 ONE("personality", S_IRUSR, proc_pid_personality),
2900 INF("limits", S_IRUGO, proc_pid_limits),
2901 #ifdef CONFIG_SCHED_DEBUG
2902 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2904 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2905 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2906 INF("syscall", S_IRUSR, proc_pid_syscall),
2908 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2909 ONE("stat", S_IRUGO, proc_tid_stat),
2910 ONE("statm", S_IRUGO, proc_pid_statm),
2911 REG("maps", S_IRUGO, proc_tid_maps_operations),
2912 #ifdef CONFIG_CHECKPOINT_RESTORE
2913 REG("children", S_IRUGO, proc_tid_children_operations),
2916 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
2918 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2919 LNK("cwd", proc_cwd_link),
2920 LNK("root", proc_root_link),
2921 LNK("exe", proc_exe_link),
2922 REG("mounts", S_IRUGO, proc_mounts_operations),
2923 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2924 #ifdef CONFIG_PROC_PAGE_MONITOR
2925 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2926 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
2927 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2929 #ifdef CONFIG_SECURITY
2930 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2932 #ifdef CONFIG_KALLSYMS
2933 INF("wchan", S_IRUGO, proc_pid_wchan),
2935 #ifdef CONFIG_STACKTRACE
2936 ONE("stack", S_IRUSR, proc_pid_stack),
2938 #ifdef CONFIG_SCHEDSTATS
2939 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2941 #ifdef CONFIG_LATENCYTOP
2942 REG("latency", S_IRUGO, proc_lstats_operations),
2944 #ifdef CONFIG_PROC_PID_CPUSET
2945 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2947 #ifdef CONFIG_CGROUPS
2948 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2950 INF("oom_score", S_IRUGO, proc_oom_score),
2951 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
2952 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2953 #ifdef CONFIG_AUDITSYSCALL
2954 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2955 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2957 #ifdef CONFIG_FAULT_INJECTION
2958 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2960 #ifdef CONFIG_TASK_IO_ACCOUNTING
2961 INF("io", S_IRUSR, proc_tid_io_accounting),
2963 #ifdef CONFIG_HARDWALL
2964 INF("hardwall", S_IRUGO, proc_pid_hardwall),
2966 #ifdef CONFIG_USER_NS
2967 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
2968 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
2969 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
2973 static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
2975 return proc_pident_readdir(file, ctx,
2976 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
2979 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2981 return proc_pident_lookup(dir, dentry,
2982 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
2985 static const struct file_operations proc_tid_base_operations = {
2986 .read = generic_read_dir,
2987 .iterate = proc_tid_base_readdir,
2988 .llseek = default_llseek,
2991 static const struct inode_operations proc_tid_base_inode_operations = {
2992 .lookup = proc_tid_base_lookup,
2993 .getattr = pid_getattr,
2994 .setattr = proc_setattr,
2997 static int proc_task_instantiate(struct inode *dir,
2998 struct dentry *dentry, struct task_struct *task, const void *ptr)
3000 struct inode *inode;
3001 inode = proc_pid_make_inode(dir->i_sb, task);
3005 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3006 inode->i_op = &proc_tid_base_inode_operations;
3007 inode->i_fop = &proc_tid_base_operations;
3008 inode->i_flags|=S_IMMUTABLE;
3010 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3011 ARRAY_SIZE(tid_base_stuff)));
3013 d_set_d_op(dentry, &pid_dentry_operations);
3015 d_add(dentry, inode);
3016 /* Close the race of the process dying before we return the dentry */
3017 if (pid_revalidate(dentry, 0))
3023 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3025 int result = -ENOENT;
3026 struct task_struct *task;
3027 struct task_struct *leader = get_proc_task(dir);
3029 struct pid_namespace *ns;
3034 tid = name_to_int(&dentry->d_name);
3038 ns = dentry->d_sb->s_fs_info;
3040 task = find_task_by_pid_ns(tid, ns);
3042 get_task_struct(task);
3046 if (!same_thread_group(leader, task))
3049 result = proc_task_instantiate(dir, dentry, task, NULL);
3051 put_task_struct(task);
3053 put_task_struct(leader);
3055 return ERR_PTR(result);
3059 * Find the first tid of a thread group to return to user space.
3061 * Usually this is just the thread group leader, but if the users
3062 * buffer was too small or there was a seek into the middle of the
3063 * directory we have more work todo.
3065 * In the case of a short read we start with find_task_by_pid.
3067 * In the case of a seek we start with the leader and walk nr
3070 static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3071 struct pid_namespace *ns)
3073 struct task_struct *pos, *task;
3074 unsigned long nr = f_pos;
3076 if (nr != f_pos) /* 32bit overflow? */
3080 task = pid_task(pid, PIDTYPE_PID);
3084 /* Attempt to start with the tid of a thread */
3086 pos = find_task_by_pid_ns(tid, ns);
3087 if (pos && same_thread_group(pos, task))
3091 /* If nr exceeds the number of threads there is nothing todo */
3092 if (nr >= get_nr_threads(task))
3095 /* If we haven't found our starting place yet start
3096 * with the leader and walk nr threads forward.
3098 pos = task = task->group_leader;
3102 } while_each_thread(task, pos);
3107 get_task_struct(pos);
3114 * Find the next thread in the thread list.
3115 * Return NULL if there is an error or no next thread.
3117 * The reference to the input task_struct is released.
3119 static struct task_struct *next_tid(struct task_struct *start)
3121 struct task_struct *pos = NULL;
3123 if (pid_alive(start)) {
3124 pos = next_thread(start);
3125 if (thread_group_leader(pos))
3128 get_task_struct(pos);
3131 put_task_struct(start);
3135 /* for the /proc/TGID/task/ directories */
3136 static int proc_task_readdir(struct file *file, struct dir_context *ctx)
3138 struct inode *inode = file_inode(file);
3139 struct task_struct *task;
3140 struct pid_namespace *ns;
3143 if (proc_inode_is_dead(inode))
3146 if (!dir_emit_dots(file, ctx))
3149 /* f_version caches the tgid value that the last readdir call couldn't
3150 * return. lseek aka telldir automagically resets f_version to 0.
3152 ns = file->f_dentry->d_sb->s_fs_info;
3153 tid = (int)file->f_version;
3154 file->f_version = 0;
3155 for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
3157 task = next_tid(task), ctx->pos++) {
3158 char name[PROC_NUMBUF];
3160 tid = task_pid_nr_ns(task, ns);
3161 len = snprintf(name, sizeof(name), "%d", tid);
3162 if (!proc_fill_cache(file, ctx, name, len,
3163 proc_task_instantiate, task, NULL)) {
3164 /* returning this tgid failed, save it as the first
3165 * pid for the next readir call */
3166 file->f_version = (u64)tid;
3167 put_task_struct(task);
3175 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3177 struct inode *inode = dentry->d_inode;
3178 struct task_struct *p = get_proc_task(inode);
3179 generic_fillattr(inode, stat);
3182 stat->nlink += get_nr_threads(p);
3189 static const struct inode_operations proc_task_inode_operations = {
3190 .lookup = proc_task_lookup,
3191 .getattr = proc_task_getattr,
3192 .setattr = proc_setattr,
3193 .permission = proc_pid_permission,
3196 static const struct file_operations proc_task_operations = {
3197 .read = generic_read_dir,
3198 .iterate = proc_task_readdir,
3199 .llseek = default_llseek,