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/rcupdate.h>
67 #include <linux/kallsyms.h>
68 #include <linux/stacktrace.h>
69 #include <linux/resource.h>
70 #include <linux/module.h>
71 #include <linux/mount.h>
72 #include <linux/security.h>
73 #include <linux/ptrace.h>
74 #include <linux/tracehook.h>
75 #include <linux/cgroup.h>
76 #include <linux/cpuset.h>
77 #include <linux/audit.h>
78 #include <linux/poll.h>
79 #include <linux/nsproxy.h>
80 #include <linux/oom.h>
81 #include <linux/elf.h>
82 #include <linux/pid_namespace.h>
83 #include <linux/fs_struct.h>
87 * Implementing inode permission operations in /proc is almost
88 * certainly an error. Permission checks need to happen during
89 * each system call not at open time. The reason is that most of
90 * what we wish to check for permissions in /proc varies at runtime.
92 * The classic example of a problem is opening file descriptors
93 * in /proc for a task before it execs a suid executable.
100 const struct inode_operations *iop;
101 const struct file_operations *fop;
105 #define NOD(NAME, MODE, IOP, FOP, OP) { \
107 .len = sizeof(NAME) - 1, \
114 #define DIR(NAME, MODE, iops, fops) \
115 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
116 #define LNK(NAME, get_link) \
117 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
118 &proc_pid_link_inode_operations, NULL, \
119 { .proc_get_link = get_link } )
120 #define REG(NAME, MODE, fops) \
121 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
122 #define INF(NAME, MODE, read) \
123 NOD(NAME, (S_IFREG|(MODE)), \
124 NULL, &proc_info_file_operations, \
125 { .proc_read = read } )
126 #define ONE(NAME, MODE, show) \
127 NOD(NAME, (S_IFREG|(MODE)), \
128 NULL, &proc_single_file_operations, \
129 { .proc_show = show } )
132 * Count the number of hardlinks for the pid_entry table, excluding the .
135 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
142 for (i = 0; i < n; ++i) {
143 if (S_ISDIR(entries[i].mode))
150 static int get_fs_path(struct task_struct *task, struct path *path, bool root)
152 struct fs_struct *fs;
153 int result = -ENOENT;
158 read_lock(&fs->lock);
159 *path = root ? fs->root : fs->pwd;
161 read_unlock(&fs->lock);
168 static int get_nr_threads(struct task_struct *tsk)
173 if (lock_task_sighand(tsk, &flags)) {
174 count = atomic_read(&tsk->signal->count);
175 unlock_task_sighand(tsk, &flags);
180 static int proc_cwd_link(struct inode *inode, struct path *path)
182 struct task_struct *task = get_proc_task(inode);
183 int result = -ENOENT;
186 result = get_fs_path(task, path, 0);
187 put_task_struct(task);
192 static int proc_root_link(struct inode *inode, struct path *path)
194 struct task_struct *task = get_proc_task(inode);
195 int result = -ENOENT;
198 result = get_fs_path(task, path, 1);
199 put_task_struct(task);
205 * Return zero if current may access user memory in @task, -error if not.
207 static int check_mem_permission(struct task_struct *task)
210 * A task can always look at itself, in case it chooses
211 * to use system calls instead of load instructions.
217 * If current is actively ptrace'ing, and would also be
218 * permitted to freshly attach with ptrace now, permit it.
220 if (task_is_stopped_or_traced(task)) {
223 match = (tracehook_tracer_task(task) == current);
225 if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH))
230 * Noone else is allowed.
235 struct mm_struct *mm_for_maps(struct task_struct *task)
237 struct mm_struct *mm;
239 if (mutex_lock_killable(&task->cred_guard_mutex))
242 mm = get_task_mm(task);
243 if (mm && mm != current->mm &&
244 !ptrace_may_access(task, PTRACE_MODE_READ)) {
248 mutex_unlock(&task->cred_guard_mutex);
253 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
257 struct mm_struct *mm = get_task_mm(task);
261 goto out_mm; /* Shh! No looking before we're done */
263 len = mm->arg_end - mm->arg_start;
268 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
270 // If the nul at the end of args has been overwritten, then
271 // assume application is using setproctitle(3).
272 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
273 len = strnlen(buffer, res);
277 len = mm->env_end - mm->env_start;
278 if (len > PAGE_SIZE - res)
279 len = PAGE_SIZE - res;
280 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
281 res = strnlen(buffer, res);
290 static int proc_pid_auxv(struct task_struct *task, char *buffer)
293 struct mm_struct *mm = get_task_mm(task);
295 unsigned int nwords = 0;
298 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
299 res = nwords * sizeof(mm->saved_auxv[0]);
302 memcpy(buffer, mm->saved_auxv, res);
309 #ifdef CONFIG_KALLSYMS
311 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
312 * Returns the resolved symbol. If that fails, simply return the address.
314 static int proc_pid_wchan(struct task_struct *task, char *buffer)
317 char symname[KSYM_NAME_LEN];
319 wchan = get_wchan(task);
321 if (lookup_symbol_name(wchan, symname) < 0)
322 if (!ptrace_may_access(task, PTRACE_MODE_READ))
325 return sprintf(buffer, "%lu", wchan);
327 return sprintf(buffer, "%s", symname);
329 #endif /* CONFIG_KALLSYMS */
331 #ifdef CONFIG_STACKTRACE
333 #define MAX_STACK_TRACE_DEPTH 64
335 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
336 struct pid *pid, struct task_struct *task)
338 struct stack_trace trace;
339 unsigned long *entries;
342 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
346 trace.nr_entries = 0;
347 trace.max_entries = MAX_STACK_TRACE_DEPTH;
348 trace.entries = entries;
350 save_stack_trace_tsk(task, &trace);
352 for (i = 0; i < trace.nr_entries; i++) {
353 seq_printf(m, "[<%p>] %pS\n",
354 (void *)entries[i], (void *)entries[i]);
362 #ifdef CONFIG_SCHEDSTATS
364 * Provides /proc/PID/schedstat
366 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
368 return sprintf(buffer, "%llu %llu %lu\n",
369 (unsigned long long)task->se.sum_exec_runtime,
370 (unsigned long long)task->sched_info.run_delay,
371 task->sched_info.pcount);
375 #ifdef CONFIG_LATENCYTOP
376 static int lstats_show_proc(struct seq_file *m, void *v)
379 struct inode *inode = m->private;
380 struct task_struct *task = get_proc_task(inode);
384 seq_puts(m, "Latency Top version : v0.1\n");
385 for (i = 0; i < 32; i++) {
386 if (task->latency_record[i].backtrace[0]) {
388 seq_printf(m, "%i %li %li ",
389 task->latency_record[i].count,
390 task->latency_record[i].time,
391 task->latency_record[i].max);
392 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
393 char sym[KSYM_SYMBOL_LEN];
395 if (!task->latency_record[i].backtrace[q])
397 if (task->latency_record[i].backtrace[q] == ULONG_MAX)
399 sprint_symbol(sym, task->latency_record[i].backtrace[q]);
400 c = strchr(sym, '+');
403 seq_printf(m, "%s ", sym);
409 put_task_struct(task);
413 static int lstats_open(struct inode *inode, struct file *file)
415 return single_open(file, lstats_show_proc, inode);
418 static ssize_t lstats_write(struct file *file, const char __user *buf,
419 size_t count, loff_t *offs)
421 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
425 clear_all_latency_tracing(task);
426 put_task_struct(task);
431 static const struct file_operations proc_lstats_operations = {
434 .write = lstats_write,
436 .release = single_release,
441 /* The badness from the OOM killer */
442 unsigned long badness(struct task_struct *p, unsigned long uptime);
443 static int proc_oom_score(struct task_struct *task, char *buffer)
445 unsigned long points;
446 struct timespec uptime;
448 do_posix_clock_monotonic_gettime(&uptime);
449 read_lock(&tasklist_lock);
450 points = badness(task->group_leader, uptime.tv_sec);
451 read_unlock(&tasklist_lock);
452 return sprintf(buffer, "%lu\n", points);
460 static const struct limit_names lnames[RLIM_NLIMITS] = {
461 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
462 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
463 [RLIMIT_DATA] = {"Max data size", "bytes"},
464 [RLIMIT_STACK] = {"Max stack size", "bytes"},
465 [RLIMIT_CORE] = {"Max core file size", "bytes"},
466 [RLIMIT_RSS] = {"Max resident set", "bytes"},
467 [RLIMIT_NPROC] = {"Max processes", "processes"},
468 [RLIMIT_NOFILE] = {"Max open files", "files"},
469 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
470 [RLIMIT_AS] = {"Max address space", "bytes"},
471 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
472 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
473 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
474 [RLIMIT_NICE] = {"Max nice priority", NULL},
475 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
476 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
479 /* Display limits for a process */
480 static int proc_pid_limits(struct task_struct *task, char *buffer)
485 char *bufptr = buffer;
487 struct rlimit rlim[RLIM_NLIMITS];
489 if (!lock_task_sighand(task, &flags))
491 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
492 unlock_task_sighand(task, &flags);
495 * print the file header
497 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
498 "Limit", "Soft Limit", "Hard Limit", "Units");
500 for (i = 0; i < RLIM_NLIMITS; i++) {
501 if (rlim[i].rlim_cur == RLIM_INFINITY)
502 count += sprintf(&bufptr[count], "%-25s %-20s ",
503 lnames[i].name, "unlimited");
505 count += sprintf(&bufptr[count], "%-25s %-20lu ",
506 lnames[i].name, rlim[i].rlim_cur);
508 if (rlim[i].rlim_max == RLIM_INFINITY)
509 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
511 count += sprintf(&bufptr[count], "%-20lu ",
515 count += sprintf(&bufptr[count], "%-10s\n",
518 count += sprintf(&bufptr[count], "\n");
524 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
525 static int proc_pid_syscall(struct task_struct *task, char *buffer)
528 unsigned long args[6], sp, pc;
530 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
531 return sprintf(buffer, "running\n");
534 return sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
536 return sprintf(buffer,
537 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
539 args[0], args[1], args[2], args[3], args[4], args[5],
542 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
544 /************************************************************************/
545 /* Here the fs part begins */
546 /************************************************************************/
548 /* permission checks */
549 static int proc_fd_access_allowed(struct inode *inode)
551 struct task_struct *task;
553 /* Allow access to a task's file descriptors if it is us or we
554 * may use ptrace attach to the process and find out that
557 task = get_proc_task(inode);
559 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
560 put_task_struct(task);
565 static int proc_setattr(struct dentry *dentry, struct iattr *attr)
568 struct inode *inode = dentry->d_inode;
570 if (attr->ia_valid & ATTR_MODE)
573 error = inode_change_ok(inode, attr);
575 error = inode_setattr(inode, attr);
579 static const struct inode_operations proc_def_inode_operations = {
580 .setattr = proc_setattr,
583 static int mounts_open_common(struct inode *inode, struct file *file,
584 const struct seq_operations *op)
586 struct task_struct *task = get_proc_task(inode);
588 struct mnt_namespace *ns = NULL;
590 struct proc_mounts *p;
595 nsp = task_nsproxy(task);
602 if (ns && get_fs_path(task, &root, 1) == 0)
604 put_task_struct(task);
613 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
617 file->private_data = &p->m;
618 ret = seq_open(file, op);
625 p->event = ns->event;
639 static int mounts_release(struct inode *inode, struct file *file)
641 struct proc_mounts *p = file->private_data;
644 return seq_release(inode, file);
647 static unsigned mounts_poll(struct file *file, poll_table *wait)
649 struct proc_mounts *p = file->private_data;
650 unsigned res = POLLIN | POLLRDNORM;
652 poll_wait(file, &p->ns->poll, wait);
653 if (mnt_had_events(p))
654 res |= POLLERR | POLLPRI;
659 static int mounts_open(struct inode *inode, struct file *file)
661 return mounts_open_common(inode, file, &mounts_op);
664 static const struct file_operations proc_mounts_operations = {
668 .release = mounts_release,
672 static int mountinfo_open(struct inode *inode, struct file *file)
674 return mounts_open_common(inode, file, &mountinfo_op);
677 static const struct file_operations proc_mountinfo_operations = {
678 .open = mountinfo_open,
681 .release = mounts_release,
685 static int mountstats_open(struct inode *inode, struct file *file)
687 return mounts_open_common(inode, file, &mountstats_op);
690 static const struct file_operations proc_mountstats_operations = {
691 .open = mountstats_open,
694 .release = mounts_release,
697 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
699 static ssize_t proc_info_read(struct file * file, char __user * buf,
700 size_t count, loff_t *ppos)
702 struct inode * inode = file->f_path.dentry->d_inode;
705 struct task_struct *task = get_proc_task(inode);
711 if (count > PROC_BLOCK_SIZE)
712 count = PROC_BLOCK_SIZE;
715 if (!(page = __get_free_page(GFP_TEMPORARY)))
718 length = PROC_I(inode)->op.proc_read(task, (char*)page);
721 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
724 put_task_struct(task);
729 static const struct file_operations proc_info_file_operations = {
730 .read = proc_info_read,
733 static int proc_single_show(struct seq_file *m, void *v)
735 struct inode *inode = m->private;
736 struct pid_namespace *ns;
738 struct task_struct *task;
741 ns = inode->i_sb->s_fs_info;
742 pid = proc_pid(inode);
743 task = get_pid_task(pid, PIDTYPE_PID);
747 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
749 put_task_struct(task);
753 static int proc_single_open(struct inode *inode, struct file *filp)
756 ret = single_open(filp, proc_single_show, NULL);
758 struct seq_file *m = filp->private_data;
765 static const struct file_operations proc_single_file_operations = {
766 .open = proc_single_open,
769 .release = single_release,
772 static int mem_open(struct inode* inode, struct file* file)
774 file->private_data = (void*)((long)current->self_exec_id);
778 static ssize_t mem_read(struct file * file, char __user * buf,
779 size_t count, loff_t *ppos)
781 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
783 unsigned long src = *ppos;
785 struct mm_struct *mm;
790 if (check_mem_permission(task))
794 page = (char *)__get_free_page(GFP_TEMPORARY);
800 mm = get_task_mm(task);
806 if (file->private_data != (void*)((long)current->self_exec_id))
812 int this_len, retval;
814 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
815 retval = access_process_vm(task, src, page, this_len, 0);
816 if (!retval || check_mem_permission(task)) {
822 if (copy_to_user(buf, page, retval)) {
837 free_page((unsigned long) page);
839 put_task_struct(task);
844 #define mem_write NULL
847 /* This is a security hazard */
848 static ssize_t mem_write(struct file * file, const char __user *buf,
849 size_t count, loff_t *ppos)
853 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
854 unsigned long dst = *ppos;
860 if (check_mem_permission(task))
864 page = (char *)__get_free_page(GFP_TEMPORARY);
870 int this_len, retval;
872 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
873 if (copy_from_user(page, buf, this_len)) {
877 retval = access_process_vm(task, dst, page, this_len, 1);
889 free_page((unsigned long) page);
891 put_task_struct(task);
897 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
901 file->f_pos = offset;
904 file->f_pos += offset;
909 force_successful_syscall_return();
913 static const struct file_operations proc_mem_operations = {
920 static ssize_t environ_read(struct file *file, char __user *buf,
921 size_t count, loff_t *ppos)
923 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
925 unsigned long src = *ppos;
927 struct mm_struct *mm;
932 if (!ptrace_may_access(task, PTRACE_MODE_READ))
936 page = (char *)__get_free_page(GFP_TEMPORARY);
942 mm = get_task_mm(task);
947 int this_len, retval, max_len;
949 this_len = mm->env_end - (mm->env_start + src);
954 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
955 this_len = (this_len > max_len) ? max_len : this_len;
957 retval = access_process_vm(task, (mm->env_start + src),
965 if (copy_to_user(buf, page, retval)) {
979 free_page((unsigned long) page);
981 put_task_struct(task);
986 static const struct file_operations proc_environ_operations = {
987 .read = environ_read,
990 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
991 size_t count, loff_t *ppos)
993 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
994 char buffer[PROC_NUMBUF];
996 int oom_adjust = OOM_DISABLE;
1002 if (lock_task_sighand(task, &flags)) {
1003 oom_adjust = task->signal->oom_adj;
1004 unlock_task_sighand(task, &flags);
1007 put_task_struct(task);
1009 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
1011 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1014 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1015 size_t count, loff_t *ppos)
1017 struct task_struct *task;
1018 char buffer[PROC_NUMBUF];
1020 unsigned long flags;
1023 memset(buffer, 0, sizeof(buffer));
1024 if (count > sizeof(buffer) - 1)
1025 count = sizeof(buffer) - 1;
1026 if (copy_from_user(buffer, buf, count))
1029 err = strict_strtol(strstrip(buffer), 0, &oom_adjust);
1032 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
1033 oom_adjust != OOM_DISABLE)
1036 task = get_proc_task(file->f_path.dentry->d_inode);
1039 if (!lock_task_sighand(task, &flags)) {
1040 put_task_struct(task);
1044 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
1045 unlock_task_sighand(task, &flags);
1046 put_task_struct(task);
1050 task->signal->oom_adj = oom_adjust;
1052 unlock_task_sighand(task, &flags);
1053 put_task_struct(task);
1058 static const struct file_operations proc_oom_adjust_operations = {
1059 .read = oom_adjust_read,
1060 .write = oom_adjust_write,
1063 #ifdef CONFIG_AUDITSYSCALL
1064 #define TMPBUFLEN 21
1065 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1066 size_t count, loff_t *ppos)
1068 struct inode * inode = file->f_path.dentry->d_inode;
1069 struct task_struct *task = get_proc_task(inode);
1071 char tmpbuf[TMPBUFLEN];
1075 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1076 audit_get_loginuid(task));
1077 put_task_struct(task);
1078 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1081 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1082 size_t count, loff_t *ppos)
1084 struct inode * inode = file->f_path.dentry->d_inode;
1089 if (!capable(CAP_AUDIT_CONTROL))
1092 if (current != pid_task(proc_pid(inode), PIDTYPE_PID))
1095 if (count >= PAGE_SIZE)
1096 count = PAGE_SIZE - 1;
1099 /* No partial writes. */
1102 page = (char*)__get_free_page(GFP_TEMPORARY);
1106 if (copy_from_user(page, buf, count))
1110 loginuid = simple_strtoul(page, &tmp, 10);
1116 length = audit_set_loginuid(current, loginuid);
1117 if (likely(length == 0))
1121 free_page((unsigned long) page);
1125 static const struct file_operations proc_loginuid_operations = {
1126 .read = proc_loginuid_read,
1127 .write = proc_loginuid_write,
1130 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1131 size_t count, loff_t *ppos)
1133 struct inode * inode = file->f_path.dentry->d_inode;
1134 struct task_struct *task = get_proc_task(inode);
1136 char tmpbuf[TMPBUFLEN];
1140 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1141 audit_get_sessionid(task));
1142 put_task_struct(task);
1143 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1146 static const struct file_operations proc_sessionid_operations = {
1147 .read = proc_sessionid_read,
1151 #ifdef CONFIG_FAULT_INJECTION
1152 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1153 size_t count, loff_t *ppos)
1155 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1156 char buffer[PROC_NUMBUF];
1162 make_it_fail = task->make_it_fail;
1163 put_task_struct(task);
1165 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1167 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1170 static ssize_t proc_fault_inject_write(struct file * file,
1171 const char __user * buf, size_t count, loff_t *ppos)
1173 struct task_struct *task;
1174 char buffer[PROC_NUMBUF], *end;
1177 if (!capable(CAP_SYS_RESOURCE))
1179 memset(buffer, 0, sizeof(buffer));
1180 if (count > sizeof(buffer) - 1)
1181 count = sizeof(buffer) - 1;
1182 if (copy_from_user(buffer, buf, count))
1184 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1187 task = get_proc_task(file->f_dentry->d_inode);
1190 task->make_it_fail = make_it_fail;
1191 put_task_struct(task);
1196 static const struct file_operations proc_fault_inject_operations = {
1197 .read = proc_fault_inject_read,
1198 .write = proc_fault_inject_write,
1203 #ifdef CONFIG_SCHED_DEBUG
1205 * Print out various scheduling related per-task fields:
1207 static int sched_show(struct seq_file *m, void *v)
1209 struct inode *inode = m->private;
1210 struct task_struct *p;
1212 p = get_proc_task(inode);
1215 proc_sched_show_task(p, m);
1223 sched_write(struct file *file, const char __user *buf,
1224 size_t count, loff_t *offset)
1226 struct inode *inode = file->f_path.dentry->d_inode;
1227 struct task_struct *p;
1229 p = get_proc_task(inode);
1232 proc_sched_set_task(p);
1239 static int sched_open(struct inode *inode, struct file *filp)
1243 ret = single_open(filp, sched_show, NULL);
1245 struct seq_file *m = filp->private_data;
1252 static const struct file_operations proc_pid_sched_operations = {
1255 .write = sched_write,
1256 .llseek = seq_lseek,
1257 .release = single_release,
1262 static ssize_t comm_write(struct file *file, const char __user *buf,
1263 size_t count, loff_t *offset)
1265 struct inode *inode = file->f_path.dentry->d_inode;
1266 struct task_struct *p;
1267 char buffer[TASK_COMM_LEN];
1269 memset(buffer, 0, sizeof(buffer));
1270 if (count > sizeof(buffer) - 1)
1271 count = sizeof(buffer) - 1;
1272 if (copy_from_user(buffer, buf, count))
1275 p = get_proc_task(inode);
1279 if (same_thread_group(current, p))
1280 set_task_comm(p, buffer);
1289 static int comm_show(struct seq_file *m, void *v)
1291 struct inode *inode = m->private;
1292 struct task_struct *p;
1294 p = get_proc_task(inode);
1299 seq_printf(m, "%s\n", p->comm);
1307 static int comm_open(struct inode *inode, struct file *filp)
1311 ret = single_open(filp, comm_show, NULL);
1313 struct seq_file *m = filp->private_data;
1320 static const struct file_operations proc_pid_set_comm_operations = {
1323 .write = comm_write,
1324 .llseek = seq_lseek,
1325 .release = single_release,
1329 * We added or removed a vma mapping the executable. The vmas are only mapped
1330 * during exec and are not mapped with the mmap system call.
1331 * Callers must hold down_write() on the mm's mmap_sem for these
1333 void added_exe_file_vma(struct mm_struct *mm)
1335 mm->num_exe_file_vmas++;
1338 void removed_exe_file_vma(struct mm_struct *mm)
1340 mm->num_exe_file_vmas--;
1341 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
1343 mm->exe_file = NULL;
1348 void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1351 get_file(new_exe_file);
1354 mm->exe_file = new_exe_file;
1355 mm->num_exe_file_vmas = 0;
1358 struct file *get_mm_exe_file(struct mm_struct *mm)
1360 struct file *exe_file;
1362 /* We need mmap_sem to protect against races with removal of
1363 * VM_EXECUTABLE vmas */
1364 down_read(&mm->mmap_sem);
1365 exe_file = mm->exe_file;
1368 up_read(&mm->mmap_sem);
1372 void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
1374 /* It's safe to write the exe_file pointer without exe_file_lock because
1375 * this is called during fork when the task is not yet in /proc */
1376 newmm->exe_file = get_mm_exe_file(oldmm);
1379 static int proc_exe_link(struct inode *inode, struct path *exe_path)
1381 struct task_struct *task;
1382 struct mm_struct *mm;
1383 struct file *exe_file;
1385 task = get_proc_task(inode);
1388 mm = get_task_mm(task);
1389 put_task_struct(task);
1392 exe_file = get_mm_exe_file(mm);
1395 *exe_path = exe_file->f_path;
1396 path_get(&exe_file->f_path);
1403 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1405 struct inode *inode = dentry->d_inode;
1406 int error = -EACCES;
1408 /* We don't need a base pointer in the /proc filesystem */
1409 path_put(&nd->path);
1411 /* Are we allowed to snoop on the tasks file descriptors? */
1412 if (!proc_fd_access_allowed(inode))
1415 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1417 return ERR_PTR(error);
1420 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1422 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1429 pathname = d_path(path, tmp, PAGE_SIZE);
1430 len = PTR_ERR(pathname);
1431 if (IS_ERR(pathname))
1433 len = tmp + PAGE_SIZE - 1 - pathname;
1437 if (copy_to_user(buffer, pathname, len))
1440 free_page((unsigned long)tmp);
1444 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1446 int error = -EACCES;
1447 struct inode *inode = dentry->d_inode;
1450 /* Are we allowed to snoop on the tasks file descriptors? */
1451 if (!proc_fd_access_allowed(inode))
1454 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1458 error = do_proc_readlink(&path, buffer, buflen);
1464 static const struct inode_operations proc_pid_link_inode_operations = {
1465 .readlink = proc_pid_readlink,
1466 .follow_link = proc_pid_follow_link,
1467 .setattr = proc_setattr,
1471 /* building an inode */
1473 static int task_dumpable(struct task_struct *task)
1476 struct mm_struct *mm;
1481 dumpable = get_dumpable(mm);
1489 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1491 struct inode * inode;
1492 struct proc_inode *ei;
1493 const struct cred *cred;
1495 /* We need a new inode */
1497 inode = new_inode(sb);
1503 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1504 inode->i_op = &proc_def_inode_operations;
1507 * grab the reference to task.
1509 ei->pid = get_task_pid(task, PIDTYPE_PID);
1513 if (task_dumpable(task)) {
1515 cred = __task_cred(task);
1516 inode->i_uid = cred->euid;
1517 inode->i_gid = cred->egid;
1520 security_task_to_inode(task, inode);
1530 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1532 struct inode *inode = dentry->d_inode;
1533 struct task_struct *task;
1534 const struct cred *cred;
1536 generic_fillattr(inode, stat);
1541 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1543 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1544 task_dumpable(task)) {
1545 cred = __task_cred(task);
1546 stat->uid = cred->euid;
1547 stat->gid = cred->egid;
1557 * Exceptional case: normally we are not allowed to unhash a busy
1558 * directory. In this case, however, we can do it - no aliasing problems
1559 * due to the way we treat inodes.
1561 * Rewrite the inode's ownerships here because the owning task may have
1562 * performed a setuid(), etc.
1564 * Before the /proc/pid/status file was created the only way to read
1565 * the effective uid of a /process was to stat /proc/pid. Reading
1566 * /proc/pid/status is slow enough that procps and other packages
1567 * kept stating /proc/pid. To keep the rules in /proc simple I have
1568 * made this apply to all per process world readable and executable
1571 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1573 struct inode *inode = dentry->d_inode;
1574 struct task_struct *task = get_proc_task(inode);
1575 const struct cred *cred;
1578 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1579 task_dumpable(task)) {
1581 cred = __task_cred(task);
1582 inode->i_uid = cred->euid;
1583 inode->i_gid = cred->egid;
1589 inode->i_mode &= ~(S_ISUID | S_ISGID);
1590 security_task_to_inode(task, inode);
1591 put_task_struct(task);
1598 static int pid_delete_dentry(struct dentry * dentry)
1600 /* Is the task we represent dead?
1601 * If so, then don't put the dentry on the lru list,
1602 * kill it immediately.
1604 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1607 static const struct dentry_operations pid_dentry_operations =
1609 .d_revalidate = pid_revalidate,
1610 .d_delete = pid_delete_dentry,
1615 typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1616 struct task_struct *, const void *);
1619 * Fill a directory entry.
1621 * If possible create the dcache entry and derive our inode number and
1622 * file type from dcache entry.
1624 * Since all of the proc inode numbers are dynamically generated, the inode
1625 * numbers do not exist until the inode is cache. This means creating the
1626 * the dcache entry in readdir is necessary to keep the inode numbers
1627 * reported by readdir in sync with the inode numbers reported
1630 static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1631 char *name, int len,
1632 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1634 struct dentry *child, *dir = filp->f_path.dentry;
1635 struct inode *inode;
1638 unsigned type = DT_UNKNOWN;
1642 qname.hash = full_name_hash(name, len);
1644 child = d_lookup(dir, &qname);
1647 new = d_alloc(dir, &qname);
1649 child = instantiate(dir->d_inode, new, task, ptr);
1656 if (!child || IS_ERR(child) || !child->d_inode)
1657 goto end_instantiate;
1658 inode = child->d_inode;
1661 type = inode->i_mode >> 12;
1666 ino = find_inode_number(dir, &qname);
1669 return filldir(dirent, name, len, filp->f_pos, ino, type);
1672 static unsigned name_to_int(struct dentry *dentry)
1674 const char *name = dentry->d_name.name;
1675 int len = dentry->d_name.len;
1678 if (len > 1 && *name == '0')
1681 unsigned c = *name++ - '0';
1684 if (n >= (~0U-9)/10)
1694 #define PROC_FDINFO_MAX 64
1696 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1698 struct task_struct *task = get_proc_task(inode);
1699 struct files_struct *files = NULL;
1701 int fd = proc_fd(inode);
1704 files = get_files_struct(task);
1705 put_task_struct(task);
1709 * We are not taking a ref to the file structure, so we must
1712 spin_lock(&files->file_lock);
1713 file = fcheck_files(files, fd);
1716 *path = file->f_path;
1717 path_get(&file->f_path);
1720 snprintf(info, PROC_FDINFO_MAX,
1723 (long long) file->f_pos,
1725 spin_unlock(&files->file_lock);
1726 put_files_struct(files);
1729 spin_unlock(&files->file_lock);
1730 put_files_struct(files);
1735 static int proc_fd_link(struct inode *inode, struct path *path)
1737 return proc_fd_info(inode, path, NULL);
1740 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1742 struct inode *inode = dentry->d_inode;
1743 struct task_struct *task = get_proc_task(inode);
1744 int fd = proc_fd(inode);
1745 struct files_struct *files;
1746 const struct cred *cred;
1749 files = get_files_struct(task);
1752 if (fcheck_files(files, fd)) {
1754 put_files_struct(files);
1755 if (task_dumpable(task)) {
1757 cred = __task_cred(task);
1758 inode->i_uid = cred->euid;
1759 inode->i_gid = cred->egid;
1765 inode->i_mode &= ~(S_ISUID | S_ISGID);
1766 security_task_to_inode(task, inode);
1767 put_task_struct(task);
1771 put_files_struct(files);
1773 put_task_struct(task);
1779 static const struct dentry_operations tid_fd_dentry_operations =
1781 .d_revalidate = tid_fd_revalidate,
1782 .d_delete = pid_delete_dentry,
1785 static struct dentry *proc_fd_instantiate(struct inode *dir,
1786 struct dentry *dentry, struct task_struct *task, const void *ptr)
1788 unsigned fd = *(const unsigned *)ptr;
1790 struct files_struct *files;
1791 struct inode *inode;
1792 struct proc_inode *ei;
1793 struct dentry *error = ERR_PTR(-ENOENT);
1795 inode = proc_pid_make_inode(dir->i_sb, task);
1800 files = get_files_struct(task);
1803 inode->i_mode = S_IFLNK;
1806 * We are not taking a ref to the file structure, so we must
1809 spin_lock(&files->file_lock);
1810 file = fcheck_files(files, fd);
1813 if (file->f_mode & FMODE_READ)
1814 inode->i_mode |= S_IRUSR | S_IXUSR;
1815 if (file->f_mode & FMODE_WRITE)
1816 inode->i_mode |= S_IWUSR | S_IXUSR;
1817 spin_unlock(&files->file_lock);
1818 put_files_struct(files);
1820 inode->i_op = &proc_pid_link_inode_operations;
1822 ei->op.proc_get_link = proc_fd_link;
1823 dentry->d_op = &tid_fd_dentry_operations;
1824 d_add(dentry, inode);
1825 /* Close the race of the process dying before we return the dentry */
1826 if (tid_fd_revalidate(dentry, NULL))
1832 spin_unlock(&files->file_lock);
1833 put_files_struct(files);
1839 static struct dentry *proc_lookupfd_common(struct inode *dir,
1840 struct dentry *dentry,
1841 instantiate_t instantiate)
1843 struct task_struct *task = get_proc_task(dir);
1844 unsigned fd = name_to_int(dentry);
1845 struct dentry *result = ERR_PTR(-ENOENT);
1852 result = instantiate(dir, dentry, task, &fd);
1854 put_task_struct(task);
1859 static int proc_readfd_common(struct file * filp, void * dirent,
1860 filldir_t filldir, instantiate_t instantiate)
1862 struct dentry *dentry = filp->f_path.dentry;
1863 struct inode *inode = dentry->d_inode;
1864 struct task_struct *p = get_proc_task(inode);
1865 unsigned int fd, ino;
1867 struct files_struct * files;
1877 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1881 ino = parent_ino(dentry);
1882 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1886 files = get_files_struct(p);
1890 for (fd = filp->f_pos-2;
1891 fd < files_fdtable(files)->max_fds;
1892 fd++, filp->f_pos++) {
1893 char name[PROC_NUMBUF];
1896 if (!fcheck_files(files, fd))
1900 len = snprintf(name, sizeof(name), "%d", fd);
1901 if (proc_fill_cache(filp, dirent, filldir,
1902 name, len, instantiate,
1910 put_files_struct(files);
1918 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1919 struct nameidata *nd)
1921 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1924 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1926 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1929 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1930 size_t len, loff_t *ppos)
1932 char tmp[PROC_FDINFO_MAX];
1933 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
1935 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1939 static const struct file_operations proc_fdinfo_file_operations = {
1940 .open = nonseekable_open,
1941 .read = proc_fdinfo_read,
1944 static const struct file_operations proc_fd_operations = {
1945 .read = generic_read_dir,
1946 .readdir = proc_readfd,
1950 * /proc/pid/fd needs a special permission handler so that a process can still
1951 * access /proc/self/fd after it has executed a setuid().
1953 static int proc_fd_permission(struct inode *inode, int mask)
1957 rv = generic_permission(inode, mask, NULL);
1960 if (task_pid(current) == proc_pid(inode))
1966 * proc directories can do almost nothing..
1968 static const struct inode_operations proc_fd_inode_operations = {
1969 .lookup = proc_lookupfd,
1970 .permission = proc_fd_permission,
1971 .setattr = proc_setattr,
1974 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
1975 struct dentry *dentry, struct task_struct *task, const void *ptr)
1977 unsigned fd = *(unsigned *)ptr;
1978 struct inode *inode;
1979 struct proc_inode *ei;
1980 struct dentry *error = ERR_PTR(-ENOENT);
1982 inode = proc_pid_make_inode(dir->i_sb, task);
1987 inode->i_mode = S_IFREG | S_IRUSR;
1988 inode->i_fop = &proc_fdinfo_file_operations;
1989 dentry->d_op = &tid_fd_dentry_operations;
1990 d_add(dentry, inode);
1991 /* Close the race of the process dying before we return the dentry */
1992 if (tid_fd_revalidate(dentry, NULL))
1999 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2000 struct dentry *dentry,
2001 struct nameidata *nd)
2003 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2006 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2008 return proc_readfd_common(filp, dirent, filldir,
2009 proc_fdinfo_instantiate);
2012 static const struct file_operations proc_fdinfo_operations = {
2013 .read = generic_read_dir,
2014 .readdir = proc_readfdinfo,
2018 * proc directories can do almost nothing..
2020 static const struct inode_operations proc_fdinfo_inode_operations = {
2021 .lookup = proc_lookupfdinfo,
2022 .setattr = proc_setattr,
2026 static struct dentry *proc_pident_instantiate(struct inode *dir,
2027 struct dentry *dentry, struct task_struct *task, const void *ptr)
2029 const struct pid_entry *p = ptr;
2030 struct inode *inode;
2031 struct proc_inode *ei;
2032 struct dentry *error = ERR_PTR(-ENOENT);
2034 inode = proc_pid_make_inode(dir->i_sb, task);
2039 inode->i_mode = p->mode;
2040 if (S_ISDIR(inode->i_mode))
2041 inode->i_nlink = 2; /* Use getattr to fix if necessary */
2043 inode->i_op = p->iop;
2045 inode->i_fop = p->fop;
2047 dentry->d_op = &pid_dentry_operations;
2048 d_add(dentry, inode);
2049 /* Close the race of the process dying before we return the dentry */
2050 if (pid_revalidate(dentry, NULL))
2056 static struct dentry *proc_pident_lookup(struct inode *dir,
2057 struct dentry *dentry,
2058 const struct pid_entry *ents,
2061 struct dentry *error;
2062 struct task_struct *task = get_proc_task(dir);
2063 const struct pid_entry *p, *last;
2065 error = ERR_PTR(-ENOENT);
2071 * Yes, it does not scale. And it should not. Don't add
2072 * new entries into /proc/<tgid>/ without very good reasons.
2074 last = &ents[nents - 1];
2075 for (p = ents; p <= last; p++) {
2076 if (p->len != dentry->d_name.len)
2078 if (!memcmp(dentry->d_name.name, p->name, p->len))
2084 error = proc_pident_instantiate(dir, dentry, task, p);
2086 put_task_struct(task);
2091 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2092 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2094 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2095 proc_pident_instantiate, task, p);
2098 static int proc_pident_readdir(struct file *filp,
2099 void *dirent, filldir_t filldir,
2100 const struct pid_entry *ents, unsigned int nents)
2103 struct dentry *dentry = filp->f_path.dentry;
2104 struct inode *inode = dentry->d_inode;
2105 struct task_struct *task = get_proc_task(inode);
2106 const struct pid_entry *p, *last;
2119 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2125 ino = parent_ino(dentry);
2126 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2138 last = &ents[nents - 1];
2140 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2149 put_task_struct(task);
2154 #ifdef CONFIG_SECURITY
2155 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2156 size_t count, loff_t *ppos)
2158 struct inode * inode = file->f_path.dentry->d_inode;
2161 struct task_struct *task = get_proc_task(inode);
2166 length = security_getprocattr(task,
2167 (char*)file->f_path.dentry->d_name.name,
2169 put_task_struct(task);
2171 length = simple_read_from_buffer(buf, count, ppos, p, length);
2176 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2177 size_t count, loff_t *ppos)
2179 struct inode * inode = file->f_path.dentry->d_inode;
2182 struct task_struct *task = get_proc_task(inode);
2187 if (count > PAGE_SIZE)
2190 /* No partial writes. */
2196 page = (char*)__get_free_page(GFP_TEMPORARY);
2201 if (copy_from_user(page, buf, count))
2204 /* Guard against adverse ptrace interaction */
2205 length = mutex_lock_interruptible(&task->cred_guard_mutex);
2209 length = security_setprocattr(task,
2210 (char*)file->f_path.dentry->d_name.name,
2211 (void*)page, count);
2212 mutex_unlock(&task->cred_guard_mutex);
2214 free_page((unsigned long) page);
2216 put_task_struct(task);
2221 static const struct file_operations proc_pid_attr_operations = {
2222 .read = proc_pid_attr_read,
2223 .write = proc_pid_attr_write,
2226 static const struct pid_entry attr_dir_stuff[] = {
2227 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2228 REG("prev", S_IRUGO, proc_pid_attr_operations),
2229 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2230 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2231 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2232 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2235 static int proc_attr_dir_readdir(struct file * filp,
2236 void * dirent, filldir_t filldir)
2238 return proc_pident_readdir(filp,dirent,filldir,
2239 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2242 static const struct file_operations proc_attr_dir_operations = {
2243 .read = generic_read_dir,
2244 .readdir = proc_attr_dir_readdir,
2247 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2248 struct dentry *dentry, struct nameidata *nd)
2250 return proc_pident_lookup(dir, dentry,
2251 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2254 static const struct inode_operations proc_attr_dir_inode_operations = {
2255 .lookup = proc_attr_dir_lookup,
2256 .getattr = pid_getattr,
2257 .setattr = proc_setattr,
2262 #ifdef CONFIG_ELF_CORE
2263 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2264 size_t count, loff_t *ppos)
2266 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2267 struct mm_struct *mm;
2268 char buffer[PROC_NUMBUF];
2276 mm = get_task_mm(task);
2278 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2279 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2280 MMF_DUMP_FILTER_SHIFT));
2282 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2285 put_task_struct(task);
2290 static ssize_t proc_coredump_filter_write(struct file *file,
2291 const char __user *buf,
2295 struct task_struct *task;
2296 struct mm_struct *mm;
2297 char buffer[PROC_NUMBUF], *end;
2304 memset(buffer, 0, sizeof(buffer));
2305 if (count > sizeof(buffer) - 1)
2306 count = sizeof(buffer) - 1;
2307 if (copy_from_user(buffer, buf, count))
2311 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2314 if (end - buffer == 0)
2318 task = get_proc_task(file->f_dentry->d_inode);
2323 mm = get_task_mm(task);
2327 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2329 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2331 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2336 put_task_struct(task);
2341 static const struct file_operations proc_coredump_filter_operations = {
2342 .read = proc_coredump_filter_read,
2343 .write = proc_coredump_filter_write,
2350 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2353 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2354 pid_t tgid = task_tgid_nr_ns(current, ns);
2355 char tmp[PROC_NUMBUF];
2358 sprintf(tmp, "%d", tgid);
2359 return vfs_readlink(dentry,buffer,buflen,tmp);
2362 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2364 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2365 pid_t tgid = task_tgid_nr_ns(current, ns);
2366 char *name = ERR_PTR(-ENOENT);
2370 name = ERR_PTR(-ENOMEM);
2372 sprintf(name, "%d", tgid);
2374 nd_set_link(nd, name);
2378 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2381 char *s = nd_get_link(nd);
2386 static const struct inode_operations proc_self_inode_operations = {
2387 .readlink = proc_self_readlink,
2388 .follow_link = proc_self_follow_link,
2389 .put_link = proc_self_put_link,
2395 * These are the directory entries in the root directory of /proc
2396 * that properly belong to the /proc filesystem, as they describe
2397 * describe something that is process related.
2399 static const struct pid_entry proc_base_stuff[] = {
2400 NOD("self", S_IFLNK|S_IRWXUGO,
2401 &proc_self_inode_operations, NULL, {}),
2405 * Exceptional case: normally we are not allowed to unhash a busy
2406 * directory. In this case, however, we can do it - no aliasing problems
2407 * due to the way we treat inodes.
2409 static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
2411 struct inode *inode = dentry->d_inode;
2412 struct task_struct *task = get_proc_task(inode);
2414 put_task_struct(task);
2421 static const struct dentry_operations proc_base_dentry_operations =
2423 .d_revalidate = proc_base_revalidate,
2424 .d_delete = pid_delete_dentry,
2427 static struct dentry *proc_base_instantiate(struct inode *dir,
2428 struct dentry *dentry, struct task_struct *task, const void *ptr)
2430 const struct pid_entry *p = ptr;
2431 struct inode *inode;
2432 struct proc_inode *ei;
2433 struct dentry *error = ERR_PTR(-EINVAL);
2435 /* Allocate the inode */
2436 error = ERR_PTR(-ENOMEM);
2437 inode = new_inode(dir->i_sb);
2441 /* Initialize the inode */
2443 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2446 * grab the reference to the task.
2448 ei->pid = get_task_pid(task, PIDTYPE_PID);
2452 inode->i_mode = p->mode;
2453 if (S_ISDIR(inode->i_mode))
2455 if (S_ISLNK(inode->i_mode))
2458 inode->i_op = p->iop;
2460 inode->i_fop = p->fop;
2462 dentry->d_op = &proc_base_dentry_operations;
2463 d_add(dentry, inode);
2472 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2474 struct dentry *error;
2475 struct task_struct *task = get_proc_task(dir);
2476 const struct pid_entry *p, *last;
2478 error = ERR_PTR(-ENOENT);
2483 /* Lookup the directory entry */
2484 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2485 for (p = proc_base_stuff; p <= last; p++) {
2486 if (p->len != dentry->d_name.len)
2488 if (!memcmp(dentry->d_name.name, p->name, p->len))
2494 error = proc_base_instantiate(dir, dentry, task, p);
2497 put_task_struct(task);
2502 static int proc_base_fill_cache(struct file *filp, void *dirent,
2503 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2505 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2506 proc_base_instantiate, task, p);
2509 #ifdef CONFIG_TASK_IO_ACCOUNTING
2510 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2512 struct task_io_accounting acct = task->ioac;
2513 unsigned long flags;
2515 if (whole && lock_task_sighand(task, &flags)) {
2516 struct task_struct *t = task;
2518 task_io_accounting_add(&acct, &task->signal->ioac);
2519 while_each_thread(task, t)
2520 task_io_accounting_add(&acct, &t->ioac);
2522 unlock_task_sighand(task, &flags);
2524 return sprintf(buffer,
2529 "read_bytes: %llu\n"
2530 "write_bytes: %llu\n"
2531 "cancelled_write_bytes: %llu\n",
2532 (unsigned long long)acct.rchar,
2533 (unsigned long long)acct.wchar,
2534 (unsigned long long)acct.syscr,
2535 (unsigned long long)acct.syscw,
2536 (unsigned long long)acct.read_bytes,
2537 (unsigned long long)acct.write_bytes,
2538 (unsigned long long)acct.cancelled_write_bytes);
2541 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2543 return do_io_accounting(task, buffer, 0);
2546 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2548 return do_io_accounting(task, buffer, 1);
2550 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2552 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2553 struct pid *pid, struct task_struct *task)
2555 seq_printf(m, "%08x\n", task->personality);
2562 static const struct file_operations proc_task_operations;
2563 static const struct inode_operations proc_task_inode_operations;
2565 static const struct pid_entry tgid_base_stuff[] = {
2566 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2567 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2568 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2570 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2572 REG("environ", S_IRUSR, proc_environ_operations),
2573 INF("auxv", S_IRUSR, proc_pid_auxv),
2574 ONE("status", S_IRUGO, proc_pid_status),
2575 ONE("personality", S_IRUSR, proc_pid_personality),
2576 INF("limits", S_IRUSR, proc_pid_limits),
2577 #ifdef CONFIG_SCHED_DEBUG
2578 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2580 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2581 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2582 INF("syscall", S_IRUSR, proc_pid_syscall),
2584 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2585 ONE("stat", S_IRUGO, proc_tgid_stat),
2586 ONE("statm", S_IRUGO, proc_pid_statm),
2587 REG("maps", S_IRUGO, proc_maps_operations),
2589 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2591 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2592 LNK("cwd", proc_cwd_link),
2593 LNK("root", proc_root_link),
2594 LNK("exe", proc_exe_link),
2595 REG("mounts", S_IRUGO, proc_mounts_operations),
2596 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2597 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2598 #ifdef CONFIG_PROC_PAGE_MONITOR
2599 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2600 REG("smaps", S_IRUGO, proc_smaps_operations),
2601 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2603 #ifdef CONFIG_SECURITY
2604 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2606 #ifdef CONFIG_KALLSYMS
2607 INF("wchan", S_IRUGO, proc_pid_wchan),
2609 #ifdef CONFIG_STACKTRACE
2610 ONE("stack", S_IRUSR, proc_pid_stack),
2612 #ifdef CONFIG_SCHEDSTATS
2613 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2615 #ifdef CONFIG_LATENCYTOP
2616 REG("latency", S_IRUGO, proc_lstats_operations),
2618 #ifdef CONFIG_PROC_PID_CPUSET
2619 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2621 #ifdef CONFIG_CGROUPS
2622 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2624 INF("oom_score", S_IRUGO, proc_oom_score),
2625 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2626 #ifdef CONFIG_AUDITSYSCALL
2627 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2628 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2630 #ifdef CONFIG_FAULT_INJECTION
2631 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2633 #ifdef CONFIG_ELF_CORE
2634 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2636 #ifdef CONFIG_TASK_IO_ACCOUNTING
2637 INF("io", S_IRUGO, proc_tgid_io_accounting),
2641 static int proc_tgid_base_readdir(struct file * filp,
2642 void * dirent, filldir_t filldir)
2644 return proc_pident_readdir(filp,dirent,filldir,
2645 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2648 static const struct file_operations proc_tgid_base_operations = {
2649 .read = generic_read_dir,
2650 .readdir = proc_tgid_base_readdir,
2653 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2654 return proc_pident_lookup(dir, dentry,
2655 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2658 static const struct inode_operations proc_tgid_base_inode_operations = {
2659 .lookup = proc_tgid_base_lookup,
2660 .getattr = pid_getattr,
2661 .setattr = proc_setattr,
2664 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2666 struct dentry *dentry, *leader, *dir;
2667 char buf[PROC_NUMBUF];
2671 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2672 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2674 shrink_dcache_parent(dentry);
2680 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2681 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2686 name.len = strlen(name.name);
2687 dir = d_hash_and_lookup(leader, &name);
2689 goto out_put_leader;
2692 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2693 dentry = d_hash_and_lookup(dir, &name);
2695 shrink_dcache_parent(dentry);
2708 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2709 * @task: task that should be flushed.
2711 * When flushing dentries from proc, one needs to flush them from global
2712 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2713 * in. This call is supposed to do all of this job.
2715 * Looks in the dcache for
2717 * /proc/@tgid/task/@pid
2718 * if either directory is present flushes it and all of it'ts children
2721 * It is safe and reasonable to cache /proc entries for a task until
2722 * that task exits. After that they just clog up the dcache with
2723 * useless entries, possibly causing useful dcache entries to be
2724 * flushed instead. This routine is proved to flush those useless
2725 * dcache entries at process exit time.
2727 * NOTE: This routine is just an optimization so it does not guarantee
2728 * that no dcache entries will exist at process exit time it
2729 * just makes it very unlikely that any will persist.
2732 void proc_flush_task(struct task_struct *task)
2735 struct pid *pid, *tgid;
2738 pid = task_pid(task);
2739 tgid = task_tgid(task);
2741 for (i = 0; i <= pid->level; i++) {
2742 upid = &pid->numbers[i];
2743 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2744 tgid->numbers[i].nr);
2747 upid = &pid->numbers[pid->level];
2749 pid_ns_release_proc(upid->ns);
2752 static struct dentry *proc_pid_instantiate(struct inode *dir,
2753 struct dentry * dentry,
2754 struct task_struct *task, const void *ptr)
2756 struct dentry *error = ERR_PTR(-ENOENT);
2757 struct inode *inode;
2759 inode = proc_pid_make_inode(dir->i_sb, task);
2763 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2764 inode->i_op = &proc_tgid_base_inode_operations;
2765 inode->i_fop = &proc_tgid_base_operations;
2766 inode->i_flags|=S_IMMUTABLE;
2768 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
2769 ARRAY_SIZE(tgid_base_stuff));
2771 dentry->d_op = &pid_dentry_operations;
2773 d_add(dentry, inode);
2774 /* Close the race of the process dying before we return the dentry */
2775 if (pid_revalidate(dentry, NULL))
2781 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2783 struct dentry *result = ERR_PTR(-ENOENT);
2784 struct task_struct *task;
2786 struct pid_namespace *ns;
2788 result = proc_base_lookup(dir, dentry);
2789 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2792 tgid = name_to_int(dentry);
2796 ns = dentry->d_sb->s_fs_info;
2798 task = find_task_by_pid_ns(tgid, ns);
2800 get_task_struct(task);
2805 result = proc_pid_instantiate(dir, dentry, task, NULL);
2806 put_task_struct(task);
2812 * Find the first task with tgid >= tgid
2817 struct task_struct *task;
2819 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2824 put_task_struct(iter.task);
2828 pid = find_ge_pid(iter.tgid, ns);
2830 iter.tgid = pid_nr_ns(pid, ns);
2831 iter.task = pid_task(pid, PIDTYPE_PID);
2832 /* What we to know is if the pid we have find is the
2833 * pid of a thread_group_leader. Testing for task
2834 * being a thread_group_leader is the obvious thing
2835 * todo but there is a window when it fails, due to
2836 * the pid transfer logic in de_thread.
2838 * So we perform the straight forward test of seeing
2839 * if the pid we have found is the pid of a thread
2840 * group leader, and don't worry if the task we have
2841 * found doesn't happen to be a thread group leader.
2842 * As we don't care in the case of readdir.
2844 if (!iter.task || !has_group_leader_pid(iter.task)) {
2848 get_task_struct(iter.task);
2854 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2856 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2857 struct tgid_iter iter)
2859 char name[PROC_NUMBUF];
2860 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
2861 return proc_fill_cache(filp, dirent, filldir, name, len,
2862 proc_pid_instantiate, iter.task, NULL);
2865 /* for the /proc/ directory itself, after non-process stuff has been done */
2866 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2868 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2869 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
2870 struct tgid_iter iter;
2871 struct pid_namespace *ns;
2876 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
2877 const struct pid_entry *p = &proc_base_stuff[nr];
2878 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
2882 ns = filp->f_dentry->d_sb->s_fs_info;
2884 iter.tgid = filp->f_pos - TGID_OFFSET;
2885 for (iter = next_tgid(ns, iter);
2887 iter.tgid += 1, iter = next_tgid(ns, iter)) {
2888 filp->f_pos = iter.tgid + TGID_OFFSET;
2889 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
2890 put_task_struct(iter.task);
2894 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
2896 put_task_struct(reaper);
2904 static const struct pid_entry tid_base_stuff[] = {
2905 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2906 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fd_operations),
2907 REG("environ", S_IRUSR, proc_environ_operations),
2908 INF("auxv", S_IRUSR, proc_pid_auxv),
2909 ONE("status", S_IRUGO, proc_pid_status),
2910 ONE("personality", S_IRUSR, proc_pid_personality),
2911 INF("limits", S_IRUSR, proc_pid_limits),
2912 #ifdef CONFIG_SCHED_DEBUG
2913 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2915 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2916 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2917 INF("syscall", S_IRUSR, proc_pid_syscall),
2919 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2920 ONE("stat", S_IRUGO, proc_tid_stat),
2921 ONE("statm", S_IRUGO, proc_pid_statm),
2922 REG("maps", S_IRUGO, proc_maps_operations),
2924 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2926 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2927 LNK("cwd", proc_cwd_link),
2928 LNK("root", proc_root_link),
2929 LNK("exe", proc_exe_link),
2930 REG("mounts", S_IRUGO, proc_mounts_operations),
2931 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2932 #ifdef CONFIG_PROC_PAGE_MONITOR
2933 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2934 REG("smaps", S_IRUGO, proc_smaps_operations),
2935 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2937 #ifdef CONFIG_SECURITY
2938 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2940 #ifdef CONFIG_KALLSYMS
2941 INF("wchan", S_IRUGO, proc_pid_wchan),
2943 #ifdef CONFIG_STACKTRACE
2944 ONE("stack", S_IRUSR, proc_pid_stack),
2946 #ifdef CONFIG_SCHEDSTATS
2947 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2949 #ifdef CONFIG_LATENCYTOP
2950 REG("latency", S_IRUGO, proc_lstats_operations),
2952 #ifdef CONFIG_PROC_PID_CPUSET
2953 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2955 #ifdef CONFIG_CGROUPS
2956 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2958 INF("oom_score", S_IRUGO, proc_oom_score),
2959 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2960 #ifdef CONFIG_AUDITSYSCALL
2961 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2962 REG("sessionid", S_IRUSR, proc_sessionid_operations),
2964 #ifdef CONFIG_FAULT_INJECTION
2965 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2967 #ifdef CONFIG_TASK_IO_ACCOUNTING
2968 INF("io", S_IRUGO, proc_tid_io_accounting),
2972 static int proc_tid_base_readdir(struct file * filp,
2973 void * dirent, filldir_t filldir)
2975 return proc_pident_readdir(filp,dirent,filldir,
2976 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
2979 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2980 return proc_pident_lookup(dir, dentry,
2981 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
2984 static const struct file_operations proc_tid_base_operations = {
2985 .read = generic_read_dir,
2986 .readdir = proc_tid_base_readdir,
2989 static const struct inode_operations proc_tid_base_inode_operations = {
2990 .lookup = proc_tid_base_lookup,
2991 .getattr = pid_getattr,
2992 .setattr = proc_setattr,
2995 static struct dentry *proc_task_instantiate(struct inode *dir,
2996 struct dentry *dentry, struct task_struct *task, const void *ptr)
2998 struct dentry *error = ERR_PTR(-ENOENT);
2999 struct inode *inode;
3000 inode = proc_pid_make_inode(dir->i_sb, task);
3004 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3005 inode->i_op = &proc_tid_base_inode_operations;
3006 inode->i_fop = &proc_tid_base_operations;
3007 inode->i_flags|=S_IMMUTABLE;
3009 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
3010 ARRAY_SIZE(tid_base_stuff));
3012 dentry->d_op = &pid_dentry_operations;
3014 d_add(dentry, inode);
3015 /* Close the race of the process dying before we return the dentry */
3016 if (pid_revalidate(dentry, NULL))
3022 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3024 struct dentry *result = ERR_PTR(-ENOENT);
3025 struct task_struct *task;
3026 struct task_struct *leader = get_proc_task(dir);
3028 struct pid_namespace *ns;
3033 tid = name_to_int(dentry);
3037 ns = dentry->d_sb->s_fs_info;
3039 task = find_task_by_pid_ns(tid, ns);
3041 get_task_struct(task);
3045 if (!same_thread_group(leader, task))
3048 result = proc_task_instantiate(dir, dentry, task, NULL);
3050 put_task_struct(task);
3052 put_task_struct(leader);
3058 * Find the first tid of a thread group to return to user space.
3060 * Usually this is just the thread group leader, but if the users
3061 * buffer was too small or there was a seek into the middle of the
3062 * directory we have more work todo.
3064 * In the case of a short read we start with find_task_by_pid.
3066 * In the case of a seek we start with the leader and walk nr
3069 static struct task_struct *first_tid(struct task_struct *leader,
3070 int tid, int nr, struct pid_namespace *ns)
3072 struct task_struct *pos;
3075 /* Attempt to start with the pid of a thread */
3076 if (tid && (nr > 0)) {
3077 pos = find_task_by_pid_ns(tid, ns);
3078 if (pos && (pos->group_leader == leader))
3082 /* If nr exceeds the number of threads there is nothing todo */
3084 if (nr && nr >= get_nr_threads(leader))
3087 /* If we haven't found our starting place yet start
3088 * with the leader and walk nr threads forward.
3090 for (pos = leader; nr > 0; --nr) {
3091 pos = next_thread(pos);
3092 if (pos == leader) {
3098 get_task_struct(pos);
3105 * Find the next thread in the thread list.
3106 * Return NULL if there is an error or no next thread.
3108 * The reference to the input task_struct is released.
3110 static struct task_struct *next_tid(struct task_struct *start)
3112 struct task_struct *pos = NULL;
3114 if (pid_alive(start)) {
3115 pos = next_thread(start);
3116 if (thread_group_leader(pos))
3119 get_task_struct(pos);
3122 put_task_struct(start);
3126 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3127 struct task_struct *task, int tid)
3129 char name[PROC_NUMBUF];
3130 int len = snprintf(name, sizeof(name), "%d", tid);
3131 return proc_fill_cache(filp, dirent, filldir, name, len,
3132 proc_task_instantiate, task, NULL);
3135 /* for the /proc/TGID/task/ directories */
3136 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3138 struct dentry *dentry = filp->f_path.dentry;
3139 struct inode *inode = dentry->d_inode;
3140 struct task_struct *leader = NULL;
3141 struct task_struct *task;
3142 int retval = -ENOENT;
3145 struct pid_namespace *ns;
3147 task = get_proc_task(inode);
3151 if (pid_alive(task)) {
3152 leader = task->group_leader;
3153 get_task_struct(leader);
3156 put_task_struct(task);
3161 switch ((unsigned long)filp->f_pos) {
3164 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3169 ino = parent_ino(dentry);
3170 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3176 /* f_version caches the tgid value that the last readdir call couldn't
3177 * return. lseek aka telldir automagically resets f_version to 0.
3179 ns = filp->f_dentry->d_sb->s_fs_info;
3180 tid = (int)filp->f_version;
3181 filp->f_version = 0;
3182 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3184 task = next_tid(task), filp->f_pos++) {
3185 tid = task_pid_nr_ns(task, ns);
3186 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3187 /* returning this tgid failed, save it as the first
3188 * pid for the next readir call */
3189 filp->f_version = (u64)tid;
3190 put_task_struct(task);
3195 put_task_struct(leader);
3200 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3202 struct inode *inode = dentry->d_inode;
3203 struct task_struct *p = get_proc_task(inode);
3204 generic_fillattr(inode, stat);
3207 stat->nlink += get_nr_threads(p);
3214 static const struct inode_operations proc_task_inode_operations = {
3215 .lookup = proc_task_lookup,
3216 .getattr = proc_task_getattr,
3217 .setattr = proc_setattr,
3220 static const struct file_operations proc_task_operations = {
3221 .read = generic_read_dir,
3222 .readdir = proc_task_readdir,