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))
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);
1120 length = audit_set_loginuid(current, loginuid);
1121 if (likely(length == 0))
1125 free_page((unsigned long) page);
1129 static const struct file_operations proc_loginuid_operations = {
1130 .read = proc_loginuid_read,
1131 .write = proc_loginuid_write,
1134 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1135 size_t count, loff_t *ppos)
1137 struct inode * inode = file->f_path.dentry->d_inode;
1138 struct task_struct *task = get_proc_task(inode);
1140 char tmpbuf[TMPBUFLEN];
1144 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1145 audit_get_sessionid(task));
1146 put_task_struct(task);
1147 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1150 static const struct file_operations proc_sessionid_operations = {
1151 .read = proc_sessionid_read,
1155 #ifdef CONFIG_FAULT_INJECTION
1156 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1157 size_t count, loff_t *ppos)
1159 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1160 char buffer[PROC_NUMBUF];
1166 make_it_fail = task->make_it_fail;
1167 put_task_struct(task);
1169 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1171 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1174 static ssize_t proc_fault_inject_write(struct file * file,
1175 const char __user * buf, size_t count, loff_t *ppos)
1177 struct task_struct *task;
1178 char buffer[PROC_NUMBUF], *end;
1181 if (!capable(CAP_SYS_RESOURCE))
1183 memset(buffer, 0, sizeof(buffer));
1184 if (count > sizeof(buffer) - 1)
1185 count = sizeof(buffer) - 1;
1186 if (copy_from_user(buffer, buf, count))
1188 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1191 task = get_proc_task(file->f_dentry->d_inode);
1194 task->make_it_fail = make_it_fail;
1195 put_task_struct(task);
1200 static const struct file_operations proc_fault_inject_operations = {
1201 .read = proc_fault_inject_read,
1202 .write = proc_fault_inject_write,
1207 #ifdef CONFIG_SCHED_DEBUG
1209 * Print out various scheduling related per-task fields:
1211 static int sched_show(struct seq_file *m, void *v)
1213 struct inode *inode = m->private;
1214 struct task_struct *p;
1216 p = get_proc_task(inode);
1219 proc_sched_show_task(p, m);
1227 sched_write(struct file *file, const char __user *buf,
1228 size_t count, loff_t *offset)
1230 struct inode *inode = file->f_path.dentry->d_inode;
1231 struct task_struct *p;
1233 p = get_proc_task(inode);
1236 proc_sched_set_task(p);
1243 static int sched_open(struct inode *inode, struct file *filp)
1247 ret = single_open(filp, sched_show, NULL);
1249 struct seq_file *m = filp->private_data;
1256 static const struct file_operations proc_pid_sched_operations = {
1259 .write = sched_write,
1260 .llseek = seq_lseek,
1261 .release = single_release,
1266 static ssize_t comm_write(struct file *file, const char __user *buf,
1267 size_t count, loff_t *offset)
1269 struct inode *inode = file->f_path.dentry->d_inode;
1270 struct task_struct *p;
1271 char buffer[TASK_COMM_LEN];
1273 memset(buffer, 0, sizeof(buffer));
1274 if (count > sizeof(buffer) - 1)
1275 count = sizeof(buffer) - 1;
1276 if (copy_from_user(buffer, buf, count))
1279 p = get_proc_task(inode);
1283 if (same_thread_group(current, p))
1284 set_task_comm(p, buffer);
1293 static int comm_show(struct seq_file *m, void *v)
1295 struct inode *inode = m->private;
1296 struct task_struct *p;
1298 p = get_proc_task(inode);
1303 seq_printf(m, "%s\n", p->comm);
1311 static int comm_open(struct inode *inode, struct file *filp)
1315 ret = single_open(filp, comm_show, NULL);
1317 struct seq_file *m = filp->private_data;
1324 static const struct file_operations proc_pid_set_comm_operations = {
1327 .write = comm_write,
1328 .llseek = seq_lseek,
1329 .release = single_release,
1333 * We added or removed a vma mapping the executable. The vmas are only mapped
1334 * during exec and are not mapped with the mmap system call.
1335 * Callers must hold down_write() on the mm's mmap_sem for these
1337 void added_exe_file_vma(struct mm_struct *mm)
1339 mm->num_exe_file_vmas++;
1342 void removed_exe_file_vma(struct mm_struct *mm)
1344 mm->num_exe_file_vmas--;
1345 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
1347 mm->exe_file = NULL;
1352 void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1355 get_file(new_exe_file);
1358 mm->exe_file = new_exe_file;
1359 mm->num_exe_file_vmas = 0;
1362 struct file *get_mm_exe_file(struct mm_struct *mm)
1364 struct file *exe_file;
1366 /* We need mmap_sem to protect against races with removal of
1367 * VM_EXECUTABLE vmas */
1368 down_read(&mm->mmap_sem);
1369 exe_file = mm->exe_file;
1372 up_read(&mm->mmap_sem);
1376 void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
1378 /* It's safe to write the exe_file pointer without exe_file_lock because
1379 * this is called during fork when the task is not yet in /proc */
1380 newmm->exe_file = get_mm_exe_file(oldmm);
1383 static int proc_exe_link(struct inode *inode, struct path *exe_path)
1385 struct task_struct *task;
1386 struct mm_struct *mm;
1387 struct file *exe_file;
1389 task = get_proc_task(inode);
1392 mm = get_task_mm(task);
1393 put_task_struct(task);
1396 exe_file = get_mm_exe_file(mm);
1399 *exe_path = exe_file->f_path;
1400 path_get(&exe_file->f_path);
1407 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1409 struct inode *inode = dentry->d_inode;
1410 int error = -EACCES;
1412 /* We don't need a base pointer in the /proc filesystem */
1413 path_put(&nd->path);
1415 /* Are we allowed to snoop on the tasks file descriptors? */
1416 if (!proc_fd_access_allowed(inode))
1419 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1421 return ERR_PTR(error);
1424 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1426 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1433 pathname = d_path(path, tmp, PAGE_SIZE);
1434 len = PTR_ERR(pathname);
1435 if (IS_ERR(pathname))
1437 len = tmp + PAGE_SIZE - 1 - pathname;
1441 if (copy_to_user(buffer, pathname, len))
1444 free_page((unsigned long)tmp);
1448 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1450 int error = -EACCES;
1451 struct inode *inode = dentry->d_inode;
1454 /* Are we allowed to snoop on the tasks file descriptors? */
1455 if (!proc_fd_access_allowed(inode))
1458 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1462 error = do_proc_readlink(&path, buffer, buflen);
1468 static const struct inode_operations proc_pid_link_inode_operations = {
1469 .readlink = proc_pid_readlink,
1470 .follow_link = proc_pid_follow_link,
1471 .setattr = proc_setattr,
1475 /* building an inode */
1477 static int task_dumpable(struct task_struct *task)
1480 struct mm_struct *mm;
1485 dumpable = get_dumpable(mm);
1493 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1495 struct inode * inode;
1496 struct proc_inode *ei;
1497 const struct cred *cred;
1499 /* We need a new inode */
1501 inode = new_inode(sb);
1507 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1508 inode->i_op = &proc_def_inode_operations;
1511 * grab the reference to task.
1513 ei->pid = get_task_pid(task, PIDTYPE_PID);
1517 if (task_dumpable(task)) {
1519 cred = __task_cred(task);
1520 inode->i_uid = cred->euid;
1521 inode->i_gid = cred->egid;
1524 security_task_to_inode(task, inode);
1534 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1536 struct inode *inode = dentry->d_inode;
1537 struct task_struct *task;
1538 const struct cred *cred;
1540 generic_fillattr(inode, stat);
1545 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1547 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1548 task_dumpable(task)) {
1549 cred = __task_cred(task);
1550 stat->uid = cred->euid;
1551 stat->gid = cred->egid;
1561 * Exceptional case: normally we are not allowed to unhash a busy
1562 * directory. In this case, however, we can do it - no aliasing problems
1563 * due to the way we treat inodes.
1565 * Rewrite the inode's ownerships here because the owning task may have
1566 * performed a setuid(), etc.
1568 * Before the /proc/pid/status file was created the only way to read
1569 * the effective uid of a /process was to stat /proc/pid. Reading
1570 * /proc/pid/status is slow enough that procps and other packages
1571 * kept stating /proc/pid. To keep the rules in /proc simple I have
1572 * made this apply to all per process world readable and executable
1575 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1577 struct inode *inode = dentry->d_inode;
1578 struct task_struct *task = get_proc_task(inode);
1579 const struct cred *cred;
1582 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1583 task_dumpable(task)) {
1585 cred = __task_cred(task);
1586 inode->i_uid = cred->euid;
1587 inode->i_gid = cred->egid;
1593 inode->i_mode &= ~(S_ISUID | S_ISGID);
1594 security_task_to_inode(task, inode);
1595 put_task_struct(task);
1602 static int pid_delete_dentry(struct dentry * dentry)
1604 /* Is the task we represent dead?
1605 * If so, then don't put the dentry on the lru list,
1606 * kill it immediately.
1608 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1611 static const struct dentry_operations pid_dentry_operations =
1613 .d_revalidate = pid_revalidate,
1614 .d_delete = pid_delete_dentry,
1619 typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1620 struct task_struct *, const void *);
1623 * Fill a directory entry.
1625 * If possible create the dcache entry and derive our inode number and
1626 * file type from dcache entry.
1628 * Since all of the proc inode numbers are dynamically generated, the inode
1629 * numbers do not exist until the inode is cache. This means creating the
1630 * the dcache entry in readdir is necessary to keep the inode numbers
1631 * reported by readdir in sync with the inode numbers reported
1634 static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1635 char *name, int len,
1636 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1638 struct dentry *child, *dir = filp->f_path.dentry;
1639 struct inode *inode;
1642 unsigned type = DT_UNKNOWN;
1646 qname.hash = full_name_hash(name, len);
1648 child = d_lookup(dir, &qname);
1651 new = d_alloc(dir, &qname);
1653 child = instantiate(dir->d_inode, new, task, ptr);
1660 if (!child || IS_ERR(child) || !child->d_inode)
1661 goto end_instantiate;
1662 inode = child->d_inode;
1665 type = inode->i_mode >> 12;
1670 ino = find_inode_number(dir, &qname);
1673 return filldir(dirent, name, len, filp->f_pos, ino, type);
1676 static unsigned name_to_int(struct dentry *dentry)
1678 const char *name = dentry->d_name.name;
1679 int len = dentry->d_name.len;
1682 if (len > 1 && *name == '0')
1685 unsigned c = *name++ - '0';
1688 if (n >= (~0U-9)/10)
1698 #define PROC_FDINFO_MAX 64
1700 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1702 struct task_struct *task = get_proc_task(inode);
1703 struct files_struct *files = NULL;
1705 int fd = proc_fd(inode);
1708 files = get_files_struct(task);
1709 put_task_struct(task);
1713 * We are not taking a ref to the file structure, so we must
1716 spin_lock(&files->file_lock);
1717 file = fcheck_files(files, fd);
1720 *path = file->f_path;
1721 path_get(&file->f_path);
1724 snprintf(info, PROC_FDINFO_MAX,
1727 (long long) file->f_pos,
1729 spin_unlock(&files->file_lock);
1730 put_files_struct(files);
1733 spin_unlock(&files->file_lock);
1734 put_files_struct(files);
1739 static int proc_fd_link(struct inode *inode, struct path *path)
1741 return proc_fd_info(inode, path, NULL);
1744 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1746 struct inode *inode = dentry->d_inode;
1747 struct task_struct *task = get_proc_task(inode);
1748 int fd = proc_fd(inode);
1749 struct files_struct *files;
1750 const struct cred *cred;
1753 files = get_files_struct(task);
1756 if (fcheck_files(files, fd)) {
1758 put_files_struct(files);
1759 if (task_dumpable(task)) {
1761 cred = __task_cred(task);
1762 inode->i_uid = cred->euid;
1763 inode->i_gid = cred->egid;
1769 inode->i_mode &= ~(S_ISUID | S_ISGID);
1770 security_task_to_inode(task, inode);
1771 put_task_struct(task);
1775 put_files_struct(files);
1777 put_task_struct(task);
1783 static const struct dentry_operations tid_fd_dentry_operations =
1785 .d_revalidate = tid_fd_revalidate,
1786 .d_delete = pid_delete_dentry,
1789 static struct dentry *proc_fd_instantiate(struct inode *dir,
1790 struct dentry *dentry, struct task_struct *task, const void *ptr)
1792 unsigned fd = *(const unsigned *)ptr;
1794 struct files_struct *files;
1795 struct inode *inode;
1796 struct proc_inode *ei;
1797 struct dentry *error = ERR_PTR(-ENOENT);
1799 inode = proc_pid_make_inode(dir->i_sb, task);
1804 files = get_files_struct(task);
1807 inode->i_mode = S_IFLNK;
1810 * We are not taking a ref to the file structure, so we must
1813 spin_lock(&files->file_lock);
1814 file = fcheck_files(files, fd);
1817 if (file->f_mode & FMODE_READ)
1818 inode->i_mode |= S_IRUSR | S_IXUSR;
1819 if (file->f_mode & FMODE_WRITE)
1820 inode->i_mode |= S_IWUSR | S_IXUSR;
1821 spin_unlock(&files->file_lock);
1822 put_files_struct(files);
1824 inode->i_op = &proc_pid_link_inode_operations;
1826 ei->op.proc_get_link = proc_fd_link;
1827 dentry->d_op = &tid_fd_dentry_operations;
1828 d_add(dentry, inode);
1829 /* Close the race of the process dying before we return the dentry */
1830 if (tid_fd_revalidate(dentry, NULL))
1836 spin_unlock(&files->file_lock);
1837 put_files_struct(files);
1843 static struct dentry *proc_lookupfd_common(struct inode *dir,
1844 struct dentry *dentry,
1845 instantiate_t instantiate)
1847 struct task_struct *task = get_proc_task(dir);
1848 unsigned fd = name_to_int(dentry);
1849 struct dentry *result = ERR_PTR(-ENOENT);
1856 result = instantiate(dir, dentry, task, &fd);
1858 put_task_struct(task);
1863 static int proc_readfd_common(struct file * filp, void * dirent,
1864 filldir_t filldir, instantiate_t instantiate)
1866 struct dentry *dentry = filp->f_path.dentry;
1867 struct inode *inode = dentry->d_inode;
1868 struct task_struct *p = get_proc_task(inode);
1869 unsigned int fd, ino;
1871 struct files_struct * files;
1881 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1885 ino = parent_ino(dentry);
1886 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1890 files = get_files_struct(p);
1894 for (fd = filp->f_pos-2;
1895 fd < files_fdtable(files)->max_fds;
1896 fd++, filp->f_pos++) {
1897 char name[PROC_NUMBUF];
1900 if (!fcheck_files(files, fd))
1904 len = snprintf(name, sizeof(name), "%d", fd);
1905 if (proc_fill_cache(filp, dirent, filldir,
1906 name, len, instantiate,
1914 put_files_struct(files);
1922 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1923 struct nameidata *nd)
1925 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1928 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1930 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1933 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1934 size_t len, loff_t *ppos)
1936 char tmp[PROC_FDINFO_MAX];
1937 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
1939 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1943 static const struct file_operations proc_fdinfo_file_operations = {
1944 .open = nonseekable_open,
1945 .read = proc_fdinfo_read,
1948 static const struct file_operations proc_fd_operations = {
1949 .read = generic_read_dir,
1950 .readdir = proc_readfd,
1954 * /proc/pid/fd needs a special permission handler so that a process can still
1955 * access /proc/self/fd after it has executed a setuid().
1957 static int proc_fd_permission(struct inode *inode, int mask)
1961 rv = generic_permission(inode, mask, NULL);
1964 if (task_pid(current) == proc_pid(inode))
1970 * proc directories can do almost nothing..
1972 static const struct inode_operations proc_fd_inode_operations = {
1973 .lookup = proc_lookupfd,
1974 .permission = proc_fd_permission,
1975 .setattr = proc_setattr,
1978 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
1979 struct dentry *dentry, struct task_struct *task, const void *ptr)
1981 unsigned fd = *(unsigned *)ptr;
1982 struct inode *inode;
1983 struct proc_inode *ei;
1984 struct dentry *error = ERR_PTR(-ENOENT);
1986 inode = proc_pid_make_inode(dir->i_sb, task);
1991 inode->i_mode = S_IFREG | S_IRUSR;
1992 inode->i_fop = &proc_fdinfo_file_operations;
1993 dentry->d_op = &tid_fd_dentry_operations;
1994 d_add(dentry, inode);
1995 /* Close the race of the process dying before we return the dentry */
1996 if (tid_fd_revalidate(dentry, NULL))
2003 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2004 struct dentry *dentry,
2005 struct nameidata *nd)
2007 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2010 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2012 return proc_readfd_common(filp, dirent, filldir,
2013 proc_fdinfo_instantiate);
2016 static const struct file_operations proc_fdinfo_operations = {
2017 .read = generic_read_dir,
2018 .readdir = proc_readfdinfo,
2022 * proc directories can do almost nothing..
2024 static const struct inode_operations proc_fdinfo_inode_operations = {
2025 .lookup = proc_lookupfdinfo,
2026 .setattr = proc_setattr,
2030 static struct dentry *proc_pident_instantiate(struct inode *dir,
2031 struct dentry *dentry, struct task_struct *task, const void *ptr)
2033 const struct pid_entry *p = ptr;
2034 struct inode *inode;
2035 struct proc_inode *ei;
2036 struct dentry *error = ERR_PTR(-ENOENT);
2038 inode = proc_pid_make_inode(dir->i_sb, task);
2043 inode->i_mode = p->mode;
2044 if (S_ISDIR(inode->i_mode))
2045 inode->i_nlink = 2; /* Use getattr to fix if necessary */
2047 inode->i_op = p->iop;
2049 inode->i_fop = p->fop;
2051 dentry->d_op = &pid_dentry_operations;
2052 d_add(dentry, inode);
2053 /* Close the race of the process dying before we return the dentry */
2054 if (pid_revalidate(dentry, NULL))
2060 static struct dentry *proc_pident_lookup(struct inode *dir,
2061 struct dentry *dentry,
2062 const struct pid_entry *ents,
2065 struct dentry *error;
2066 struct task_struct *task = get_proc_task(dir);
2067 const struct pid_entry *p, *last;
2069 error = ERR_PTR(-ENOENT);
2075 * Yes, it does not scale. And it should not. Don't add
2076 * new entries into /proc/<tgid>/ without very good reasons.
2078 last = &ents[nents - 1];
2079 for (p = ents; p <= last; p++) {
2080 if (p->len != dentry->d_name.len)
2082 if (!memcmp(dentry->d_name.name, p->name, p->len))
2088 error = proc_pident_instantiate(dir, dentry, task, p);
2090 put_task_struct(task);
2095 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2096 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2098 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2099 proc_pident_instantiate, task, p);
2102 static int proc_pident_readdir(struct file *filp,
2103 void *dirent, filldir_t filldir,
2104 const struct pid_entry *ents, unsigned int nents)
2107 struct dentry *dentry = filp->f_path.dentry;
2108 struct inode *inode = dentry->d_inode;
2109 struct task_struct *task = get_proc_task(inode);
2110 const struct pid_entry *p, *last;
2123 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2129 ino = parent_ino(dentry);
2130 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2142 last = &ents[nents - 1];
2144 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2153 put_task_struct(task);
2158 #ifdef CONFIG_SECURITY
2159 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2160 size_t count, loff_t *ppos)
2162 struct inode * inode = file->f_path.dentry->d_inode;
2165 struct task_struct *task = get_proc_task(inode);
2170 length = security_getprocattr(task,
2171 (char*)file->f_path.dentry->d_name.name,
2173 put_task_struct(task);
2175 length = simple_read_from_buffer(buf, count, ppos, p, length);
2180 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2181 size_t count, loff_t *ppos)
2183 struct inode * inode = file->f_path.dentry->d_inode;
2186 struct task_struct *task = get_proc_task(inode);
2191 if (count > PAGE_SIZE)
2194 /* No partial writes. */
2200 page = (char*)__get_free_page(GFP_TEMPORARY);
2205 if (copy_from_user(page, buf, count))
2208 /* Guard against adverse ptrace interaction */
2209 length = mutex_lock_interruptible(&task->cred_guard_mutex);
2213 length = security_setprocattr(task,
2214 (char*)file->f_path.dentry->d_name.name,
2215 (void*)page, count);
2216 mutex_unlock(&task->cred_guard_mutex);
2218 free_page((unsigned long) page);
2220 put_task_struct(task);
2225 static const struct file_operations proc_pid_attr_operations = {
2226 .read = proc_pid_attr_read,
2227 .write = proc_pid_attr_write,
2230 static const struct pid_entry attr_dir_stuff[] = {
2231 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2232 REG("prev", S_IRUGO, proc_pid_attr_operations),
2233 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2234 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2235 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2236 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2239 static int proc_attr_dir_readdir(struct file * filp,
2240 void * dirent, filldir_t filldir)
2242 return proc_pident_readdir(filp,dirent,filldir,
2243 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2246 static const struct file_operations proc_attr_dir_operations = {
2247 .read = generic_read_dir,
2248 .readdir = proc_attr_dir_readdir,
2251 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2252 struct dentry *dentry, struct nameidata *nd)
2254 return proc_pident_lookup(dir, dentry,
2255 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2258 static const struct inode_operations proc_attr_dir_inode_operations = {
2259 .lookup = proc_attr_dir_lookup,
2260 .getattr = pid_getattr,
2261 .setattr = proc_setattr,
2266 #ifdef CONFIG_ELF_CORE
2267 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2268 size_t count, loff_t *ppos)
2270 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2271 struct mm_struct *mm;
2272 char buffer[PROC_NUMBUF];
2280 mm = get_task_mm(task);
2282 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2283 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2284 MMF_DUMP_FILTER_SHIFT));
2286 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2289 put_task_struct(task);
2294 static ssize_t proc_coredump_filter_write(struct file *file,
2295 const char __user *buf,
2299 struct task_struct *task;
2300 struct mm_struct *mm;
2301 char buffer[PROC_NUMBUF], *end;
2308 memset(buffer, 0, sizeof(buffer));
2309 if (count > sizeof(buffer) - 1)
2310 count = sizeof(buffer) - 1;
2311 if (copy_from_user(buffer, buf, count))
2315 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2318 if (end - buffer == 0)
2322 task = get_proc_task(file->f_dentry->d_inode);
2327 mm = get_task_mm(task);
2331 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2333 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2335 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2340 put_task_struct(task);
2345 static const struct file_operations proc_coredump_filter_operations = {
2346 .read = proc_coredump_filter_read,
2347 .write = proc_coredump_filter_write,
2354 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2357 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2358 pid_t tgid = task_tgid_nr_ns(current, ns);
2359 char tmp[PROC_NUMBUF];
2362 sprintf(tmp, "%d", tgid);
2363 return vfs_readlink(dentry,buffer,buflen,tmp);
2366 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2368 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2369 pid_t tgid = task_tgid_nr_ns(current, ns);
2370 char *name = ERR_PTR(-ENOENT);
2374 name = ERR_PTR(-ENOMEM);
2376 sprintf(name, "%d", tgid);
2378 nd_set_link(nd, name);
2382 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2385 char *s = nd_get_link(nd);
2390 static const struct inode_operations proc_self_inode_operations = {
2391 .readlink = proc_self_readlink,
2392 .follow_link = proc_self_follow_link,
2393 .put_link = proc_self_put_link,
2399 * These are the directory entries in the root directory of /proc
2400 * that properly belong to the /proc filesystem, as they describe
2401 * describe something that is process related.
2403 static const struct pid_entry proc_base_stuff[] = {
2404 NOD("self", S_IFLNK|S_IRWXUGO,
2405 &proc_self_inode_operations, NULL, {}),
2409 * Exceptional case: normally we are not allowed to unhash a busy
2410 * directory. In this case, however, we can do it - no aliasing problems
2411 * due to the way we treat inodes.
2413 static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
2415 struct inode *inode = dentry->d_inode;
2416 struct task_struct *task = get_proc_task(inode);
2418 put_task_struct(task);
2425 static const struct dentry_operations proc_base_dentry_operations =
2427 .d_revalidate = proc_base_revalidate,
2428 .d_delete = pid_delete_dentry,
2431 static struct dentry *proc_base_instantiate(struct inode *dir,
2432 struct dentry *dentry, struct task_struct *task, const void *ptr)
2434 const struct pid_entry *p = ptr;
2435 struct inode *inode;
2436 struct proc_inode *ei;
2437 struct dentry *error = ERR_PTR(-EINVAL);
2439 /* Allocate the inode */
2440 error = ERR_PTR(-ENOMEM);
2441 inode = new_inode(dir->i_sb);
2445 /* Initialize the inode */
2447 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2450 * grab the reference to the task.
2452 ei->pid = get_task_pid(task, PIDTYPE_PID);
2456 inode->i_mode = p->mode;
2457 if (S_ISDIR(inode->i_mode))
2459 if (S_ISLNK(inode->i_mode))
2462 inode->i_op = p->iop;
2464 inode->i_fop = p->fop;
2466 dentry->d_op = &proc_base_dentry_operations;
2467 d_add(dentry, inode);
2476 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2478 struct dentry *error;
2479 struct task_struct *task = get_proc_task(dir);
2480 const struct pid_entry *p, *last;
2482 error = ERR_PTR(-ENOENT);
2487 /* Lookup the directory entry */
2488 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2489 for (p = proc_base_stuff; p <= last; p++) {
2490 if (p->len != dentry->d_name.len)
2492 if (!memcmp(dentry->d_name.name, p->name, p->len))
2498 error = proc_base_instantiate(dir, dentry, task, p);
2501 put_task_struct(task);
2506 static int proc_base_fill_cache(struct file *filp, void *dirent,
2507 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2509 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2510 proc_base_instantiate, task, p);
2513 #ifdef CONFIG_TASK_IO_ACCOUNTING
2514 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2516 struct task_io_accounting acct = task->ioac;
2517 unsigned long flags;
2519 if (whole && lock_task_sighand(task, &flags)) {
2520 struct task_struct *t = task;
2522 task_io_accounting_add(&acct, &task->signal->ioac);
2523 while_each_thread(task, t)
2524 task_io_accounting_add(&acct, &t->ioac);
2526 unlock_task_sighand(task, &flags);
2528 return sprintf(buffer,
2533 "read_bytes: %llu\n"
2534 "write_bytes: %llu\n"
2535 "cancelled_write_bytes: %llu\n",
2536 (unsigned long long)acct.rchar,
2537 (unsigned long long)acct.wchar,
2538 (unsigned long long)acct.syscr,
2539 (unsigned long long)acct.syscw,
2540 (unsigned long long)acct.read_bytes,
2541 (unsigned long long)acct.write_bytes,
2542 (unsigned long long)acct.cancelled_write_bytes);
2545 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2547 return do_io_accounting(task, buffer, 0);
2550 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2552 return do_io_accounting(task, buffer, 1);
2554 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2556 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2557 struct pid *pid, struct task_struct *task)
2559 seq_printf(m, "%08x\n", task->personality);
2566 static const struct file_operations proc_task_operations;
2567 static const struct inode_operations proc_task_inode_operations;
2569 static const struct pid_entry tgid_base_stuff[] = {
2570 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2571 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2572 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2574 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2576 REG("environ", S_IRUSR, proc_environ_operations),
2577 INF("auxv", S_IRUSR, proc_pid_auxv),
2578 ONE("status", S_IRUGO, proc_pid_status),
2579 ONE("personality", S_IRUSR, proc_pid_personality),
2580 INF("limits", S_IRUSR, proc_pid_limits),
2581 #ifdef CONFIG_SCHED_DEBUG
2582 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2584 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2585 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2586 INF("syscall", S_IRUSR, proc_pid_syscall),
2588 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2589 ONE("stat", S_IRUGO, proc_tgid_stat),
2590 ONE("statm", S_IRUGO, proc_pid_statm),
2591 REG("maps", S_IRUGO, proc_maps_operations),
2593 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2595 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2596 LNK("cwd", proc_cwd_link),
2597 LNK("root", proc_root_link),
2598 LNK("exe", proc_exe_link),
2599 REG("mounts", S_IRUGO, proc_mounts_operations),
2600 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2601 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2602 #ifdef CONFIG_PROC_PAGE_MONITOR
2603 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2604 REG("smaps", S_IRUGO, proc_smaps_operations),
2605 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2607 #ifdef CONFIG_SECURITY
2608 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2610 #ifdef CONFIG_KALLSYMS
2611 INF("wchan", S_IRUGO, proc_pid_wchan),
2613 #ifdef CONFIG_STACKTRACE
2614 ONE("stack", S_IRUSR, proc_pid_stack),
2616 #ifdef CONFIG_SCHEDSTATS
2617 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2619 #ifdef CONFIG_LATENCYTOP
2620 REG("latency", S_IRUGO, proc_lstats_operations),
2622 #ifdef CONFIG_PROC_PID_CPUSET
2623 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2625 #ifdef CONFIG_CGROUPS
2626 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2628 INF("oom_score", S_IRUGO, proc_oom_score),
2629 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2630 #ifdef CONFIG_AUDITSYSCALL
2631 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2632 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2634 #ifdef CONFIG_FAULT_INJECTION
2635 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2637 #ifdef CONFIG_ELF_CORE
2638 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2640 #ifdef CONFIG_TASK_IO_ACCOUNTING
2641 INF("io", S_IRUGO, proc_tgid_io_accounting),
2645 static int proc_tgid_base_readdir(struct file * filp,
2646 void * dirent, filldir_t filldir)
2648 return proc_pident_readdir(filp,dirent,filldir,
2649 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2652 static const struct file_operations proc_tgid_base_operations = {
2653 .read = generic_read_dir,
2654 .readdir = proc_tgid_base_readdir,
2657 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2658 return proc_pident_lookup(dir, dentry,
2659 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2662 static const struct inode_operations proc_tgid_base_inode_operations = {
2663 .lookup = proc_tgid_base_lookup,
2664 .getattr = pid_getattr,
2665 .setattr = proc_setattr,
2668 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2670 struct dentry *dentry, *leader, *dir;
2671 char buf[PROC_NUMBUF];
2675 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2676 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2678 shrink_dcache_parent(dentry);
2684 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2685 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2690 name.len = strlen(name.name);
2691 dir = d_hash_and_lookup(leader, &name);
2693 goto out_put_leader;
2696 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2697 dentry = d_hash_and_lookup(dir, &name);
2699 shrink_dcache_parent(dentry);
2712 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2713 * @task: task that should be flushed.
2715 * When flushing dentries from proc, one needs to flush them from global
2716 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2717 * in. This call is supposed to do all of this job.
2719 * Looks in the dcache for
2721 * /proc/@tgid/task/@pid
2722 * if either directory is present flushes it and all of it'ts children
2725 * It is safe and reasonable to cache /proc entries for a task until
2726 * that task exits. After that they just clog up the dcache with
2727 * useless entries, possibly causing useful dcache entries to be
2728 * flushed instead. This routine is proved to flush those useless
2729 * dcache entries at process exit time.
2731 * NOTE: This routine is just an optimization so it does not guarantee
2732 * that no dcache entries will exist at process exit time it
2733 * just makes it very unlikely that any will persist.
2736 void proc_flush_task(struct task_struct *task)
2739 struct pid *pid, *tgid;
2742 pid = task_pid(task);
2743 tgid = task_tgid(task);
2745 for (i = 0; i <= pid->level; i++) {
2746 upid = &pid->numbers[i];
2747 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2748 tgid->numbers[i].nr);
2751 upid = &pid->numbers[pid->level];
2753 pid_ns_release_proc(upid->ns);
2756 static struct dentry *proc_pid_instantiate(struct inode *dir,
2757 struct dentry * dentry,
2758 struct task_struct *task, const void *ptr)
2760 struct dentry *error = ERR_PTR(-ENOENT);
2761 struct inode *inode;
2763 inode = proc_pid_make_inode(dir->i_sb, task);
2767 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2768 inode->i_op = &proc_tgid_base_inode_operations;
2769 inode->i_fop = &proc_tgid_base_operations;
2770 inode->i_flags|=S_IMMUTABLE;
2772 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
2773 ARRAY_SIZE(tgid_base_stuff));
2775 dentry->d_op = &pid_dentry_operations;
2777 d_add(dentry, inode);
2778 /* Close the race of the process dying before we return the dentry */
2779 if (pid_revalidate(dentry, NULL))
2785 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2787 struct dentry *result = ERR_PTR(-ENOENT);
2788 struct task_struct *task;
2790 struct pid_namespace *ns;
2792 result = proc_base_lookup(dir, dentry);
2793 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2796 tgid = name_to_int(dentry);
2800 ns = dentry->d_sb->s_fs_info;
2802 task = find_task_by_pid_ns(tgid, ns);
2804 get_task_struct(task);
2809 result = proc_pid_instantiate(dir, dentry, task, NULL);
2810 put_task_struct(task);
2816 * Find the first task with tgid >= tgid
2821 struct task_struct *task;
2823 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2828 put_task_struct(iter.task);
2832 pid = find_ge_pid(iter.tgid, ns);
2834 iter.tgid = pid_nr_ns(pid, ns);
2835 iter.task = pid_task(pid, PIDTYPE_PID);
2836 /* What we to know is if the pid we have find is the
2837 * pid of a thread_group_leader. Testing for task
2838 * being a thread_group_leader is the obvious thing
2839 * todo but there is a window when it fails, due to
2840 * the pid transfer logic in de_thread.
2842 * So we perform the straight forward test of seeing
2843 * if the pid we have found is the pid of a thread
2844 * group leader, and don't worry if the task we have
2845 * found doesn't happen to be a thread group leader.
2846 * As we don't care in the case of readdir.
2848 if (!iter.task || !has_group_leader_pid(iter.task)) {
2852 get_task_struct(iter.task);
2858 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2860 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2861 struct tgid_iter iter)
2863 char name[PROC_NUMBUF];
2864 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
2865 return proc_fill_cache(filp, dirent, filldir, name, len,
2866 proc_pid_instantiate, iter.task, NULL);
2869 /* for the /proc/ directory itself, after non-process stuff has been done */
2870 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2872 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2873 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
2874 struct tgid_iter iter;
2875 struct pid_namespace *ns;
2880 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
2881 const struct pid_entry *p = &proc_base_stuff[nr];
2882 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
2886 ns = filp->f_dentry->d_sb->s_fs_info;
2888 iter.tgid = filp->f_pos - TGID_OFFSET;
2889 for (iter = next_tgid(ns, iter);
2891 iter.tgid += 1, iter = next_tgid(ns, iter)) {
2892 filp->f_pos = iter.tgid + TGID_OFFSET;
2893 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
2894 put_task_struct(iter.task);
2898 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
2900 put_task_struct(reaper);
2908 static const struct pid_entry tid_base_stuff[] = {
2909 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2910 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fd_operations),
2911 REG("environ", S_IRUSR, proc_environ_operations),
2912 INF("auxv", S_IRUSR, proc_pid_auxv),
2913 ONE("status", S_IRUGO, proc_pid_status),
2914 ONE("personality", S_IRUSR, proc_pid_personality),
2915 INF("limits", S_IRUSR, proc_pid_limits),
2916 #ifdef CONFIG_SCHED_DEBUG
2917 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2919 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2920 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2921 INF("syscall", S_IRUSR, proc_pid_syscall),
2923 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2924 ONE("stat", S_IRUGO, proc_tid_stat),
2925 ONE("statm", S_IRUGO, proc_pid_statm),
2926 REG("maps", S_IRUGO, proc_maps_operations),
2928 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2930 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2931 LNK("cwd", proc_cwd_link),
2932 LNK("root", proc_root_link),
2933 LNK("exe", proc_exe_link),
2934 REG("mounts", S_IRUGO, proc_mounts_operations),
2935 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2936 #ifdef CONFIG_PROC_PAGE_MONITOR
2937 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2938 REG("smaps", S_IRUGO, proc_smaps_operations),
2939 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2941 #ifdef CONFIG_SECURITY
2942 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2944 #ifdef CONFIG_KALLSYMS
2945 INF("wchan", S_IRUGO, proc_pid_wchan),
2947 #ifdef CONFIG_STACKTRACE
2948 ONE("stack", S_IRUSR, proc_pid_stack),
2950 #ifdef CONFIG_SCHEDSTATS
2951 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2953 #ifdef CONFIG_LATENCYTOP
2954 REG("latency", S_IRUGO, proc_lstats_operations),
2956 #ifdef CONFIG_PROC_PID_CPUSET
2957 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2959 #ifdef CONFIG_CGROUPS
2960 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2962 INF("oom_score", S_IRUGO, proc_oom_score),
2963 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2964 #ifdef CONFIG_AUDITSYSCALL
2965 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2966 REG("sessionid", S_IRUSR, proc_sessionid_operations),
2968 #ifdef CONFIG_FAULT_INJECTION
2969 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2971 #ifdef CONFIG_TASK_IO_ACCOUNTING
2972 INF("io", S_IRUGO, proc_tid_io_accounting),
2976 static int proc_tid_base_readdir(struct file * filp,
2977 void * dirent, filldir_t filldir)
2979 return proc_pident_readdir(filp,dirent,filldir,
2980 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
2983 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2984 return proc_pident_lookup(dir, dentry,
2985 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
2988 static const struct file_operations proc_tid_base_operations = {
2989 .read = generic_read_dir,
2990 .readdir = proc_tid_base_readdir,
2993 static const struct inode_operations proc_tid_base_inode_operations = {
2994 .lookup = proc_tid_base_lookup,
2995 .getattr = pid_getattr,
2996 .setattr = proc_setattr,
2999 static struct dentry *proc_task_instantiate(struct inode *dir,
3000 struct dentry *dentry, struct task_struct *task, const void *ptr)
3002 struct dentry *error = ERR_PTR(-ENOENT);
3003 struct inode *inode;
3004 inode = proc_pid_make_inode(dir->i_sb, task);
3008 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3009 inode->i_op = &proc_tid_base_inode_operations;
3010 inode->i_fop = &proc_tid_base_operations;
3011 inode->i_flags|=S_IMMUTABLE;
3013 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
3014 ARRAY_SIZE(tid_base_stuff));
3016 dentry->d_op = &pid_dentry_operations;
3018 d_add(dentry, inode);
3019 /* Close the race of the process dying before we return the dentry */
3020 if (pid_revalidate(dentry, NULL))
3026 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3028 struct dentry *result = ERR_PTR(-ENOENT);
3029 struct task_struct *task;
3030 struct task_struct *leader = get_proc_task(dir);
3032 struct pid_namespace *ns;
3037 tid = name_to_int(dentry);
3041 ns = dentry->d_sb->s_fs_info;
3043 task = find_task_by_pid_ns(tid, ns);
3045 get_task_struct(task);
3049 if (!same_thread_group(leader, task))
3052 result = proc_task_instantiate(dir, dentry, task, NULL);
3054 put_task_struct(task);
3056 put_task_struct(leader);
3062 * Find the first tid of a thread group to return to user space.
3064 * Usually this is just the thread group leader, but if the users
3065 * buffer was too small or there was a seek into the middle of the
3066 * directory we have more work todo.
3068 * In the case of a short read we start with find_task_by_pid.
3070 * In the case of a seek we start with the leader and walk nr
3073 static struct task_struct *first_tid(struct task_struct *leader,
3074 int tid, int nr, struct pid_namespace *ns)
3076 struct task_struct *pos;
3079 /* Attempt to start with the pid of a thread */
3080 if (tid && (nr > 0)) {
3081 pos = find_task_by_pid_ns(tid, ns);
3082 if (pos && (pos->group_leader == leader))
3086 /* If nr exceeds the number of threads there is nothing todo */
3088 if (nr && nr >= get_nr_threads(leader))
3091 /* If we haven't found our starting place yet start
3092 * with the leader and walk nr threads forward.
3094 for (pos = leader; nr > 0; --nr) {
3095 pos = next_thread(pos);
3096 if (pos == leader) {
3102 get_task_struct(pos);
3109 * Find the next thread in the thread list.
3110 * Return NULL if there is an error or no next thread.
3112 * The reference to the input task_struct is released.
3114 static struct task_struct *next_tid(struct task_struct *start)
3116 struct task_struct *pos = NULL;
3118 if (pid_alive(start)) {
3119 pos = next_thread(start);
3120 if (thread_group_leader(pos))
3123 get_task_struct(pos);
3126 put_task_struct(start);
3130 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3131 struct task_struct *task, int tid)
3133 char name[PROC_NUMBUF];
3134 int len = snprintf(name, sizeof(name), "%d", tid);
3135 return proc_fill_cache(filp, dirent, filldir, name, len,
3136 proc_task_instantiate, task, NULL);
3139 /* for the /proc/TGID/task/ directories */
3140 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3142 struct dentry *dentry = filp->f_path.dentry;
3143 struct inode *inode = dentry->d_inode;
3144 struct task_struct *leader = NULL;
3145 struct task_struct *task;
3146 int retval = -ENOENT;
3149 struct pid_namespace *ns;
3151 task = get_proc_task(inode);
3155 if (pid_alive(task)) {
3156 leader = task->group_leader;
3157 get_task_struct(leader);
3160 put_task_struct(task);
3165 switch ((unsigned long)filp->f_pos) {
3168 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3173 ino = parent_ino(dentry);
3174 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3180 /* f_version caches the tgid value that the last readdir call couldn't
3181 * return. lseek aka telldir automagically resets f_version to 0.
3183 ns = filp->f_dentry->d_sb->s_fs_info;
3184 tid = (int)filp->f_version;
3185 filp->f_version = 0;
3186 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3188 task = next_tid(task), filp->f_pos++) {
3189 tid = task_pid_nr_ns(task, ns);
3190 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3191 /* returning this tgid failed, save it as the first
3192 * pid for the next readir call */
3193 filp->f_version = (u64)tid;
3194 put_task_struct(task);
3199 put_task_struct(leader);
3204 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3206 struct inode *inode = dentry->d_inode;
3207 struct task_struct *p = get_proc_task(inode);
3208 generic_fillattr(inode, stat);
3211 stat->nlink += get_nr_threads(p);
3218 static const struct inode_operations proc_task_inode_operations = {
3219 .lookup = proc_task_lookup,
3220 .getattr = proc_task_getattr,
3221 .setattr = proc_setattr,
3224 static const struct file_operations proc_task_operations = {
3225 .read = generic_read_dir,
3226 .readdir = proc_task_readdir,