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>
84 #include <linux/slab.h>
88 * Implementing inode permission operations in /proc is almost
89 * certainly an error. Permission checks need to happen during
90 * each system call not at open time. The reason is that most of
91 * what we wish to check for permissions in /proc varies at runtime.
93 * The classic example of a problem is opening file descriptors
94 * in /proc for a task before it execs a suid executable.
101 const struct inode_operations *iop;
102 const struct file_operations *fop;
106 #define NOD(NAME, MODE, IOP, FOP, OP) { \
108 .len = sizeof(NAME) - 1, \
115 #define DIR(NAME, MODE, iops, fops) \
116 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
117 #define LNK(NAME, get_link) \
118 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
119 &proc_pid_link_inode_operations, NULL, \
120 { .proc_get_link = get_link } )
121 #define REG(NAME, MODE, fops) \
122 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
123 #define INF(NAME, MODE, read) \
124 NOD(NAME, (S_IFREG|(MODE)), \
125 NULL, &proc_info_file_operations, \
126 { .proc_read = read } )
127 #define ONE(NAME, MODE, show) \
128 NOD(NAME, (S_IFREG|(MODE)), \
129 NULL, &proc_single_file_operations, \
130 { .proc_show = show } )
133 * Count the number of hardlinks for the pid_entry table, excluding the .
136 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
143 for (i = 0; i < n; ++i) {
144 if (S_ISDIR(entries[i].mode))
151 static int get_fs_path(struct task_struct *task, struct path *path, bool root)
153 struct fs_struct *fs;
154 int result = -ENOENT;
159 read_lock(&fs->lock);
160 *path = root ? fs->root : fs->pwd;
162 read_unlock(&fs->lock);
169 static int get_nr_threads(struct task_struct *tsk)
174 if (lock_task_sighand(tsk, &flags)) {
175 count = atomic_read(&tsk->signal->count);
176 unlock_task_sighand(tsk, &flags);
181 static int proc_cwd_link(struct inode *inode, struct path *path)
183 struct task_struct *task = get_proc_task(inode);
184 int result = -ENOENT;
187 result = get_fs_path(task, path, 0);
188 put_task_struct(task);
193 static int proc_root_link(struct inode *inode, struct path *path)
195 struct task_struct *task = get_proc_task(inode);
196 int result = -ENOENT;
199 result = get_fs_path(task, path, 1);
200 put_task_struct(task);
206 * Return zero if current may access user memory in @task, -error if not.
208 static int check_mem_permission(struct task_struct *task)
211 * A task can always look at itself, in case it chooses
212 * to use system calls instead of load instructions.
218 * If current is actively ptrace'ing, and would also be
219 * permitted to freshly attach with ptrace now, permit it.
221 if (task_is_stopped_or_traced(task)) {
224 match = (tracehook_tracer_task(task) == current);
226 if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH))
231 * Noone else is allowed.
236 struct mm_struct *mm_for_maps(struct task_struct *task)
238 struct mm_struct *mm;
240 if (mutex_lock_killable(&task->cred_guard_mutex))
243 mm = get_task_mm(task);
244 if (mm && mm != current->mm &&
245 !ptrace_may_access(task, PTRACE_MODE_READ)) {
249 mutex_unlock(&task->cred_guard_mutex);
254 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
258 struct mm_struct *mm = get_task_mm(task);
262 goto out_mm; /* Shh! No looking before we're done */
264 len = mm->arg_end - mm->arg_start;
269 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
271 // If the nul at the end of args has been overwritten, then
272 // assume application is using setproctitle(3).
273 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
274 len = strnlen(buffer, res);
278 len = mm->env_end - mm->env_start;
279 if (len > PAGE_SIZE - res)
280 len = PAGE_SIZE - res;
281 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
282 res = strnlen(buffer, res);
291 static int proc_pid_auxv(struct task_struct *task, char *buffer)
294 struct mm_struct *mm = get_task_mm(task);
296 unsigned int nwords = 0;
299 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
300 res = nwords * sizeof(mm->saved_auxv[0]);
303 memcpy(buffer, mm->saved_auxv, res);
310 #ifdef CONFIG_KALLSYMS
312 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
313 * Returns the resolved symbol. If that fails, simply return the address.
315 static int proc_pid_wchan(struct task_struct *task, char *buffer)
318 char symname[KSYM_NAME_LEN];
320 wchan = get_wchan(task);
322 if (lookup_symbol_name(wchan, symname) < 0)
323 if (!ptrace_may_access(task, PTRACE_MODE_READ))
326 return sprintf(buffer, "%lu", wchan);
328 return sprintf(buffer, "%s", symname);
330 #endif /* CONFIG_KALLSYMS */
332 #ifdef CONFIG_STACKTRACE
334 #define MAX_STACK_TRACE_DEPTH 64
336 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
337 struct pid *pid, struct task_struct *task)
339 struct stack_trace trace;
340 unsigned long *entries;
343 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
347 trace.nr_entries = 0;
348 trace.max_entries = MAX_STACK_TRACE_DEPTH;
349 trace.entries = entries;
351 save_stack_trace_tsk(task, &trace);
353 for (i = 0; i < trace.nr_entries; i++) {
354 seq_printf(m, "[<%p>] %pS\n",
355 (void *)entries[i], (void *)entries[i]);
363 #ifdef CONFIG_SCHEDSTATS
365 * Provides /proc/PID/schedstat
367 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
369 return sprintf(buffer, "%llu %llu %lu\n",
370 (unsigned long long)task->se.sum_exec_runtime,
371 (unsigned long long)task->sched_info.run_delay,
372 task->sched_info.pcount);
376 #ifdef CONFIG_LATENCYTOP
377 static int lstats_show_proc(struct seq_file *m, void *v)
380 struct inode *inode = m->private;
381 struct task_struct *task = get_proc_task(inode);
385 seq_puts(m, "Latency Top version : v0.1\n");
386 for (i = 0; i < 32; i++) {
387 if (task->latency_record[i].backtrace[0]) {
389 seq_printf(m, "%i %li %li ",
390 task->latency_record[i].count,
391 task->latency_record[i].time,
392 task->latency_record[i].max);
393 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
394 char sym[KSYM_SYMBOL_LEN];
396 if (!task->latency_record[i].backtrace[q])
398 if (task->latency_record[i].backtrace[q] == ULONG_MAX)
400 sprint_symbol(sym, task->latency_record[i].backtrace[q]);
401 c = strchr(sym, '+');
404 seq_printf(m, "%s ", sym);
410 put_task_struct(task);
414 static int lstats_open(struct inode *inode, struct file *file)
416 return single_open(file, lstats_show_proc, inode);
419 static ssize_t lstats_write(struct file *file, const char __user *buf,
420 size_t count, loff_t *offs)
422 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
426 clear_all_latency_tracing(task);
427 put_task_struct(task);
432 static const struct file_operations proc_lstats_operations = {
435 .write = lstats_write,
437 .release = single_release,
442 /* The badness from the OOM killer */
443 unsigned long badness(struct task_struct *p, unsigned long uptime);
444 static int proc_oom_score(struct task_struct *task, char *buffer)
446 unsigned long points;
447 struct timespec uptime;
449 do_posix_clock_monotonic_gettime(&uptime);
450 read_lock(&tasklist_lock);
451 points = badness(task->group_leader, uptime.tv_sec);
452 read_unlock(&tasklist_lock);
453 return sprintf(buffer, "%lu\n", points);
461 static const struct limit_names lnames[RLIM_NLIMITS] = {
462 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
463 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
464 [RLIMIT_DATA] = {"Max data size", "bytes"},
465 [RLIMIT_STACK] = {"Max stack size", "bytes"},
466 [RLIMIT_CORE] = {"Max core file size", "bytes"},
467 [RLIMIT_RSS] = {"Max resident set", "bytes"},
468 [RLIMIT_NPROC] = {"Max processes", "processes"},
469 [RLIMIT_NOFILE] = {"Max open files", "files"},
470 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
471 [RLIMIT_AS] = {"Max address space", "bytes"},
472 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
473 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
474 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
475 [RLIMIT_NICE] = {"Max nice priority", NULL},
476 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
477 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
480 /* Display limits for a process */
481 static int proc_pid_limits(struct task_struct *task, char *buffer)
486 char *bufptr = buffer;
488 struct rlimit rlim[RLIM_NLIMITS];
490 if (!lock_task_sighand(task, &flags))
492 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
493 unlock_task_sighand(task, &flags);
496 * print the file header
498 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
499 "Limit", "Soft Limit", "Hard Limit", "Units");
501 for (i = 0; i < RLIM_NLIMITS; i++) {
502 if (rlim[i].rlim_cur == RLIM_INFINITY)
503 count += sprintf(&bufptr[count], "%-25s %-20s ",
504 lnames[i].name, "unlimited");
506 count += sprintf(&bufptr[count], "%-25s %-20lu ",
507 lnames[i].name, rlim[i].rlim_cur);
509 if (rlim[i].rlim_max == RLIM_INFINITY)
510 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
512 count += sprintf(&bufptr[count], "%-20lu ",
516 count += sprintf(&bufptr[count], "%-10s\n",
519 count += sprintf(&bufptr[count], "\n");
525 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
526 static int proc_pid_syscall(struct task_struct *task, char *buffer)
529 unsigned long args[6], sp, pc;
531 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
532 return sprintf(buffer, "running\n");
535 return sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
537 return sprintf(buffer,
538 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
540 args[0], args[1], args[2], args[3], args[4], args[5],
543 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
545 /************************************************************************/
546 /* Here the fs part begins */
547 /************************************************************************/
549 /* permission checks */
550 static int proc_fd_access_allowed(struct inode *inode)
552 struct task_struct *task;
554 /* Allow access to a task's file descriptors if it is us or we
555 * may use ptrace attach to the process and find out that
558 task = get_proc_task(inode);
560 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
561 put_task_struct(task);
566 static int proc_setattr(struct dentry *dentry, struct iattr *attr)
569 struct inode *inode = dentry->d_inode;
571 if (attr->ia_valid & ATTR_MODE)
574 error = inode_change_ok(inode, attr);
576 error = inode_setattr(inode, attr);
580 static const struct inode_operations proc_def_inode_operations = {
581 .setattr = proc_setattr,
584 static int mounts_open_common(struct inode *inode, struct file *file,
585 const struct seq_operations *op)
587 struct task_struct *task = get_proc_task(inode);
589 struct mnt_namespace *ns = NULL;
591 struct proc_mounts *p;
596 nsp = task_nsproxy(task);
603 if (ns && get_fs_path(task, &root, 1) == 0)
605 put_task_struct(task);
614 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
618 file->private_data = &p->m;
619 ret = seq_open(file, op);
626 p->event = ns->event;
640 static int mounts_release(struct inode *inode, struct file *file)
642 struct proc_mounts *p = file->private_data;
645 return seq_release(inode, file);
648 static unsigned mounts_poll(struct file *file, poll_table *wait)
650 struct proc_mounts *p = file->private_data;
651 unsigned res = POLLIN | POLLRDNORM;
653 poll_wait(file, &p->ns->poll, wait);
654 if (mnt_had_events(p))
655 res |= POLLERR | POLLPRI;
660 static int mounts_open(struct inode *inode, struct file *file)
662 return mounts_open_common(inode, file, &mounts_op);
665 static const struct file_operations proc_mounts_operations = {
669 .release = mounts_release,
673 static int mountinfo_open(struct inode *inode, struct file *file)
675 return mounts_open_common(inode, file, &mountinfo_op);
678 static const struct file_operations proc_mountinfo_operations = {
679 .open = mountinfo_open,
682 .release = mounts_release,
686 static int mountstats_open(struct inode *inode, struct file *file)
688 return mounts_open_common(inode, file, &mountstats_op);
691 static const struct file_operations proc_mountstats_operations = {
692 .open = mountstats_open,
695 .release = mounts_release,
698 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
700 static ssize_t proc_info_read(struct file * file, char __user * buf,
701 size_t count, loff_t *ppos)
703 struct inode * inode = file->f_path.dentry->d_inode;
706 struct task_struct *task = get_proc_task(inode);
712 if (count > PROC_BLOCK_SIZE)
713 count = PROC_BLOCK_SIZE;
716 if (!(page = __get_free_page(GFP_TEMPORARY)))
719 length = PROC_I(inode)->op.proc_read(task, (char*)page);
722 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
725 put_task_struct(task);
730 static const struct file_operations proc_info_file_operations = {
731 .read = proc_info_read,
734 static int proc_single_show(struct seq_file *m, void *v)
736 struct inode *inode = m->private;
737 struct pid_namespace *ns;
739 struct task_struct *task;
742 ns = inode->i_sb->s_fs_info;
743 pid = proc_pid(inode);
744 task = get_pid_task(pid, PIDTYPE_PID);
748 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
750 put_task_struct(task);
754 static int proc_single_open(struct inode *inode, struct file *filp)
757 ret = single_open(filp, proc_single_show, NULL);
759 struct seq_file *m = filp->private_data;
766 static const struct file_operations proc_single_file_operations = {
767 .open = proc_single_open,
770 .release = single_release,
773 static int mem_open(struct inode* inode, struct file* file)
775 file->private_data = (void*)((long)current->self_exec_id);
779 static ssize_t mem_read(struct file * file, char __user * buf,
780 size_t count, loff_t *ppos)
782 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
784 unsigned long src = *ppos;
786 struct mm_struct *mm;
791 if (check_mem_permission(task))
795 page = (char *)__get_free_page(GFP_TEMPORARY);
801 mm = get_task_mm(task);
807 if (file->private_data != (void*)((long)current->self_exec_id))
813 int this_len, retval;
815 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
816 retval = access_process_vm(task, src, page, this_len, 0);
817 if (!retval || check_mem_permission(task)) {
823 if (copy_to_user(buf, page, retval)) {
838 free_page((unsigned long) page);
840 put_task_struct(task);
845 #define mem_write NULL
848 /* This is a security hazard */
849 static ssize_t mem_write(struct file * file, const char __user *buf,
850 size_t count, loff_t *ppos)
854 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
855 unsigned long dst = *ppos;
861 if (check_mem_permission(task))
865 page = (char *)__get_free_page(GFP_TEMPORARY);
871 int this_len, retval;
873 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
874 if (copy_from_user(page, buf, this_len)) {
878 retval = access_process_vm(task, dst, page, this_len, 1);
890 free_page((unsigned long) page);
892 put_task_struct(task);
898 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
902 file->f_pos = offset;
905 file->f_pos += offset;
910 force_successful_syscall_return();
914 static const struct file_operations proc_mem_operations = {
921 static ssize_t environ_read(struct file *file, char __user *buf,
922 size_t count, loff_t *ppos)
924 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
926 unsigned long src = *ppos;
928 struct mm_struct *mm;
933 if (!ptrace_may_access(task, PTRACE_MODE_READ))
937 page = (char *)__get_free_page(GFP_TEMPORARY);
943 mm = get_task_mm(task);
948 int this_len, retval, max_len;
950 this_len = mm->env_end - (mm->env_start + src);
955 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
956 this_len = (this_len > max_len) ? max_len : this_len;
958 retval = access_process_vm(task, (mm->env_start + src),
966 if (copy_to_user(buf, page, retval)) {
980 free_page((unsigned long) page);
982 put_task_struct(task);
987 static const struct file_operations proc_environ_operations = {
988 .read = environ_read,
991 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
992 size_t count, loff_t *ppos)
994 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
995 char buffer[PROC_NUMBUF];
997 int oom_adjust = OOM_DISABLE;
1003 if (lock_task_sighand(task, &flags)) {
1004 oom_adjust = task->signal->oom_adj;
1005 unlock_task_sighand(task, &flags);
1008 put_task_struct(task);
1010 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
1012 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1015 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1016 size_t count, loff_t *ppos)
1018 struct task_struct *task;
1019 char buffer[PROC_NUMBUF];
1021 unsigned long flags;
1024 memset(buffer, 0, sizeof(buffer));
1025 if (count > sizeof(buffer) - 1)
1026 count = sizeof(buffer) - 1;
1027 if (copy_from_user(buffer, buf, count))
1030 err = strict_strtol(strstrip(buffer), 0, &oom_adjust);
1033 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
1034 oom_adjust != OOM_DISABLE)
1037 task = get_proc_task(file->f_path.dentry->d_inode);
1040 if (!lock_task_sighand(task, &flags)) {
1041 put_task_struct(task);
1045 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
1046 unlock_task_sighand(task, &flags);
1047 put_task_struct(task);
1051 task->signal->oom_adj = oom_adjust;
1053 unlock_task_sighand(task, &flags);
1054 put_task_struct(task);
1059 static const struct file_operations proc_oom_adjust_operations = {
1060 .read = oom_adjust_read,
1061 .write = oom_adjust_write,
1064 #ifdef CONFIG_AUDITSYSCALL
1065 #define TMPBUFLEN 21
1066 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1067 size_t count, loff_t *ppos)
1069 struct inode * inode = file->f_path.dentry->d_inode;
1070 struct task_struct *task = get_proc_task(inode);
1072 char tmpbuf[TMPBUFLEN];
1076 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1077 audit_get_loginuid(task));
1078 put_task_struct(task);
1079 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1082 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1083 size_t count, loff_t *ppos)
1085 struct inode * inode = file->f_path.dentry->d_inode;
1090 if (!capable(CAP_AUDIT_CONTROL))
1094 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1100 if (count >= PAGE_SIZE)
1101 count = PAGE_SIZE - 1;
1104 /* No partial writes. */
1107 page = (char*)__get_free_page(GFP_TEMPORARY);
1111 if (copy_from_user(page, buf, count))
1115 loginuid = simple_strtoul(page, &tmp, 10);
1121 length = audit_set_loginuid(current, loginuid);
1122 if (likely(length == 0))
1126 free_page((unsigned long) page);
1130 static const struct file_operations proc_loginuid_operations = {
1131 .read = proc_loginuid_read,
1132 .write = proc_loginuid_write,
1135 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1136 size_t count, loff_t *ppos)
1138 struct inode * inode = file->f_path.dentry->d_inode;
1139 struct task_struct *task = get_proc_task(inode);
1141 char tmpbuf[TMPBUFLEN];
1145 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1146 audit_get_sessionid(task));
1147 put_task_struct(task);
1148 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1151 static const struct file_operations proc_sessionid_operations = {
1152 .read = proc_sessionid_read,
1156 #ifdef CONFIG_FAULT_INJECTION
1157 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1158 size_t count, loff_t *ppos)
1160 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1161 char buffer[PROC_NUMBUF];
1167 make_it_fail = task->make_it_fail;
1168 put_task_struct(task);
1170 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1172 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1175 static ssize_t proc_fault_inject_write(struct file * file,
1176 const char __user * buf, size_t count, loff_t *ppos)
1178 struct task_struct *task;
1179 char buffer[PROC_NUMBUF], *end;
1182 if (!capable(CAP_SYS_RESOURCE))
1184 memset(buffer, 0, sizeof(buffer));
1185 if (count > sizeof(buffer) - 1)
1186 count = sizeof(buffer) - 1;
1187 if (copy_from_user(buffer, buf, count))
1189 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1192 task = get_proc_task(file->f_dentry->d_inode);
1195 task->make_it_fail = make_it_fail;
1196 put_task_struct(task);
1201 static const struct file_operations proc_fault_inject_operations = {
1202 .read = proc_fault_inject_read,
1203 .write = proc_fault_inject_write,
1208 #ifdef CONFIG_SCHED_DEBUG
1210 * Print out various scheduling related per-task fields:
1212 static int sched_show(struct seq_file *m, void *v)
1214 struct inode *inode = m->private;
1215 struct task_struct *p;
1217 p = get_proc_task(inode);
1220 proc_sched_show_task(p, m);
1228 sched_write(struct file *file, const char __user *buf,
1229 size_t count, loff_t *offset)
1231 struct inode *inode = file->f_path.dentry->d_inode;
1232 struct task_struct *p;
1234 p = get_proc_task(inode);
1237 proc_sched_set_task(p);
1244 static int sched_open(struct inode *inode, struct file *filp)
1248 ret = single_open(filp, sched_show, NULL);
1250 struct seq_file *m = filp->private_data;
1257 static const struct file_operations proc_pid_sched_operations = {
1260 .write = sched_write,
1261 .llseek = seq_lseek,
1262 .release = single_release,
1267 static ssize_t comm_write(struct file *file, const char __user *buf,
1268 size_t count, loff_t *offset)
1270 struct inode *inode = file->f_path.dentry->d_inode;
1271 struct task_struct *p;
1272 char buffer[TASK_COMM_LEN];
1274 memset(buffer, 0, sizeof(buffer));
1275 if (count > sizeof(buffer) - 1)
1276 count = sizeof(buffer) - 1;
1277 if (copy_from_user(buffer, buf, count))
1280 p = get_proc_task(inode);
1284 if (same_thread_group(current, p))
1285 set_task_comm(p, buffer);
1294 static int comm_show(struct seq_file *m, void *v)
1296 struct inode *inode = m->private;
1297 struct task_struct *p;
1299 p = get_proc_task(inode);
1304 seq_printf(m, "%s\n", p->comm);
1312 static int comm_open(struct inode *inode, struct file *filp)
1316 ret = single_open(filp, comm_show, NULL);
1318 struct seq_file *m = filp->private_data;
1325 static const struct file_operations proc_pid_set_comm_operations = {
1328 .write = comm_write,
1329 .llseek = seq_lseek,
1330 .release = single_release,
1334 * We added or removed a vma mapping the executable. The vmas are only mapped
1335 * during exec and are not mapped with the mmap system call.
1336 * Callers must hold down_write() on the mm's mmap_sem for these
1338 void added_exe_file_vma(struct mm_struct *mm)
1340 mm->num_exe_file_vmas++;
1343 void removed_exe_file_vma(struct mm_struct *mm)
1345 mm->num_exe_file_vmas--;
1346 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
1348 mm->exe_file = NULL;
1353 void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1356 get_file(new_exe_file);
1359 mm->exe_file = new_exe_file;
1360 mm->num_exe_file_vmas = 0;
1363 struct file *get_mm_exe_file(struct mm_struct *mm)
1365 struct file *exe_file;
1367 /* We need mmap_sem to protect against races with removal of
1368 * VM_EXECUTABLE vmas */
1369 down_read(&mm->mmap_sem);
1370 exe_file = mm->exe_file;
1373 up_read(&mm->mmap_sem);
1377 void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
1379 /* It's safe to write the exe_file pointer without exe_file_lock because
1380 * this is called during fork when the task is not yet in /proc */
1381 newmm->exe_file = get_mm_exe_file(oldmm);
1384 static int proc_exe_link(struct inode *inode, struct path *exe_path)
1386 struct task_struct *task;
1387 struct mm_struct *mm;
1388 struct file *exe_file;
1390 task = get_proc_task(inode);
1393 mm = get_task_mm(task);
1394 put_task_struct(task);
1397 exe_file = get_mm_exe_file(mm);
1400 *exe_path = exe_file->f_path;
1401 path_get(&exe_file->f_path);
1408 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1410 struct inode *inode = dentry->d_inode;
1411 int error = -EACCES;
1413 /* We don't need a base pointer in the /proc filesystem */
1414 path_put(&nd->path);
1416 /* Are we allowed to snoop on the tasks file descriptors? */
1417 if (!proc_fd_access_allowed(inode))
1420 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1422 return ERR_PTR(error);
1425 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1427 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1434 pathname = d_path(path, tmp, PAGE_SIZE);
1435 len = PTR_ERR(pathname);
1436 if (IS_ERR(pathname))
1438 len = tmp + PAGE_SIZE - 1 - pathname;
1442 if (copy_to_user(buffer, pathname, len))
1445 free_page((unsigned long)tmp);
1449 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1451 int error = -EACCES;
1452 struct inode *inode = dentry->d_inode;
1455 /* Are we allowed to snoop on the tasks file descriptors? */
1456 if (!proc_fd_access_allowed(inode))
1459 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1463 error = do_proc_readlink(&path, buffer, buflen);
1469 static const struct inode_operations proc_pid_link_inode_operations = {
1470 .readlink = proc_pid_readlink,
1471 .follow_link = proc_pid_follow_link,
1472 .setattr = proc_setattr,
1476 /* building an inode */
1478 static int task_dumpable(struct task_struct *task)
1481 struct mm_struct *mm;
1486 dumpable = get_dumpable(mm);
1494 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1496 struct inode * inode;
1497 struct proc_inode *ei;
1498 const struct cred *cred;
1500 /* We need a new inode */
1502 inode = new_inode(sb);
1508 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1509 inode->i_op = &proc_def_inode_operations;
1512 * grab the reference to task.
1514 ei->pid = get_task_pid(task, PIDTYPE_PID);
1518 if (task_dumpable(task)) {
1520 cred = __task_cred(task);
1521 inode->i_uid = cred->euid;
1522 inode->i_gid = cred->egid;
1525 security_task_to_inode(task, inode);
1535 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1537 struct inode *inode = dentry->d_inode;
1538 struct task_struct *task;
1539 const struct cred *cred;
1541 generic_fillattr(inode, stat);
1546 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1548 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1549 task_dumpable(task)) {
1550 cred = __task_cred(task);
1551 stat->uid = cred->euid;
1552 stat->gid = cred->egid;
1562 * Exceptional case: normally we are not allowed to unhash a busy
1563 * directory. In this case, however, we can do it - no aliasing problems
1564 * due to the way we treat inodes.
1566 * Rewrite the inode's ownerships here because the owning task may have
1567 * performed a setuid(), etc.
1569 * Before the /proc/pid/status file was created the only way to read
1570 * the effective uid of a /process was to stat /proc/pid. Reading
1571 * /proc/pid/status is slow enough that procps and other packages
1572 * kept stating /proc/pid. To keep the rules in /proc simple I have
1573 * made this apply to all per process world readable and executable
1576 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1578 struct inode *inode = dentry->d_inode;
1579 struct task_struct *task = get_proc_task(inode);
1580 const struct cred *cred;
1583 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1584 task_dumpable(task)) {
1586 cred = __task_cred(task);
1587 inode->i_uid = cred->euid;
1588 inode->i_gid = cred->egid;
1594 inode->i_mode &= ~(S_ISUID | S_ISGID);
1595 security_task_to_inode(task, inode);
1596 put_task_struct(task);
1603 static int pid_delete_dentry(struct dentry * dentry)
1605 /* Is the task we represent dead?
1606 * If so, then don't put the dentry on the lru list,
1607 * kill it immediately.
1609 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1612 static const struct dentry_operations pid_dentry_operations =
1614 .d_revalidate = pid_revalidate,
1615 .d_delete = pid_delete_dentry,
1620 typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1621 struct task_struct *, const void *);
1624 * Fill a directory entry.
1626 * If possible create the dcache entry and derive our inode number and
1627 * file type from dcache entry.
1629 * Since all of the proc inode numbers are dynamically generated, the inode
1630 * numbers do not exist until the inode is cache. This means creating the
1631 * the dcache entry in readdir is necessary to keep the inode numbers
1632 * reported by readdir in sync with the inode numbers reported
1635 static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1636 char *name, int len,
1637 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1639 struct dentry *child, *dir = filp->f_path.dentry;
1640 struct inode *inode;
1643 unsigned type = DT_UNKNOWN;
1647 qname.hash = full_name_hash(name, len);
1649 child = d_lookup(dir, &qname);
1652 new = d_alloc(dir, &qname);
1654 child = instantiate(dir->d_inode, new, task, ptr);
1661 if (!child || IS_ERR(child) || !child->d_inode)
1662 goto end_instantiate;
1663 inode = child->d_inode;
1666 type = inode->i_mode >> 12;
1671 ino = find_inode_number(dir, &qname);
1674 return filldir(dirent, name, len, filp->f_pos, ino, type);
1677 static unsigned name_to_int(struct dentry *dentry)
1679 const char *name = dentry->d_name.name;
1680 int len = dentry->d_name.len;
1683 if (len > 1 && *name == '0')
1686 unsigned c = *name++ - '0';
1689 if (n >= (~0U-9)/10)
1699 #define PROC_FDINFO_MAX 64
1701 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1703 struct task_struct *task = get_proc_task(inode);
1704 struct files_struct *files = NULL;
1706 int fd = proc_fd(inode);
1709 files = get_files_struct(task);
1710 put_task_struct(task);
1714 * We are not taking a ref to the file structure, so we must
1717 spin_lock(&files->file_lock);
1718 file = fcheck_files(files, fd);
1721 *path = file->f_path;
1722 path_get(&file->f_path);
1725 snprintf(info, PROC_FDINFO_MAX,
1728 (long long) file->f_pos,
1730 spin_unlock(&files->file_lock);
1731 put_files_struct(files);
1734 spin_unlock(&files->file_lock);
1735 put_files_struct(files);
1740 static int proc_fd_link(struct inode *inode, struct path *path)
1742 return proc_fd_info(inode, path, NULL);
1745 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1747 struct inode *inode = dentry->d_inode;
1748 struct task_struct *task = get_proc_task(inode);
1749 int fd = proc_fd(inode);
1750 struct files_struct *files;
1751 const struct cred *cred;
1754 files = get_files_struct(task);
1757 if (fcheck_files(files, fd)) {
1759 put_files_struct(files);
1760 if (task_dumpable(task)) {
1762 cred = __task_cred(task);
1763 inode->i_uid = cred->euid;
1764 inode->i_gid = cred->egid;
1770 inode->i_mode &= ~(S_ISUID | S_ISGID);
1771 security_task_to_inode(task, inode);
1772 put_task_struct(task);
1776 put_files_struct(files);
1778 put_task_struct(task);
1784 static const struct dentry_operations tid_fd_dentry_operations =
1786 .d_revalidate = tid_fd_revalidate,
1787 .d_delete = pid_delete_dentry,
1790 static struct dentry *proc_fd_instantiate(struct inode *dir,
1791 struct dentry *dentry, struct task_struct *task, const void *ptr)
1793 unsigned fd = *(const unsigned *)ptr;
1795 struct files_struct *files;
1796 struct inode *inode;
1797 struct proc_inode *ei;
1798 struct dentry *error = ERR_PTR(-ENOENT);
1800 inode = proc_pid_make_inode(dir->i_sb, task);
1805 files = get_files_struct(task);
1808 inode->i_mode = S_IFLNK;
1811 * We are not taking a ref to the file structure, so we must
1814 spin_lock(&files->file_lock);
1815 file = fcheck_files(files, fd);
1818 if (file->f_mode & FMODE_READ)
1819 inode->i_mode |= S_IRUSR | S_IXUSR;
1820 if (file->f_mode & FMODE_WRITE)
1821 inode->i_mode |= S_IWUSR | S_IXUSR;
1822 spin_unlock(&files->file_lock);
1823 put_files_struct(files);
1825 inode->i_op = &proc_pid_link_inode_operations;
1827 ei->op.proc_get_link = proc_fd_link;
1828 dentry->d_op = &tid_fd_dentry_operations;
1829 d_add(dentry, inode);
1830 /* Close the race of the process dying before we return the dentry */
1831 if (tid_fd_revalidate(dentry, NULL))
1837 spin_unlock(&files->file_lock);
1838 put_files_struct(files);
1844 static struct dentry *proc_lookupfd_common(struct inode *dir,
1845 struct dentry *dentry,
1846 instantiate_t instantiate)
1848 struct task_struct *task = get_proc_task(dir);
1849 unsigned fd = name_to_int(dentry);
1850 struct dentry *result = ERR_PTR(-ENOENT);
1857 result = instantiate(dir, dentry, task, &fd);
1859 put_task_struct(task);
1864 static int proc_readfd_common(struct file * filp, void * dirent,
1865 filldir_t filldir, instantiate_t instantiate)
1867 struct dentry *dentry = filp->f_path.dentry;
1868 struct inode *inode = dentry->d_inode;
1869 struct task_struct *p = get_proc_task(inode);
1870 unsigned int fd, ino;
1872 struct files_struct * files;
1882 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1886 ino = parent_ino(dentry);
1887 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1891 files = get_files_struct(p);
1895 for (fd = filp->f_pos-2;
1896 fd < files_fdtable(files)->max_fds;
1897 fd++, filp->f_pos++) {
1898 char name[PROC_NUMBUF];
1901 if (!fcheck_files(files, fd))
1905 len = snprintf(name, sizeof(name), "%d", fd);
1906 if (proc_fill_cache(filp, dirent, filldir,
1907 name, len, instantiate,
1915 put_files_struct(files);
1923 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1924 struct nameidata *nd)
1926 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1929 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1931 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1934 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1935 size_t len, loff_t *ppos)
1937 char tmp[PROC_FDINFO_MAX];
1938 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
1940 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1944 static const struct file_operations proc_fdinfo_file_operations = {
1945 .open = nonseekable_open,
1946 .read = proc_fdinfo_read,
1949 static const struct file_operations proc_fd_operations = {
1950 .read = generic_read_dir,
1951 .readdir = proc_readfd,
1955 * /proc/pid/fd needs a special permission handler so that a process can still
1956 * access /proc/self/fd after it has executed a setuid().
1958 static int proc_fd_permission(struct inode *inode, int mask)
1962 rv = generic_permission(inode, mask, NULL);
1965 if (task_pid(current) == proc_pid(inode))
1971 * proc directories can do almost nothing..
1973 static const struct inode_operations proc_fd_inode_operations = {
1974 .lookup = proc_lookupfd,
1975 .permission = proc_fd_permission,
1976 .setattr = proc_setattr,
1979 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
1980 struct dentry *dentry, struct task_struct *task, const void *ptr)
1982 unsigned fd = *(unsigned *)ptr;
1983 struct inode *inode;
1984 struct proc_inode *ei;
1985 struct dentry *error = ERR_PTR(-ENOENT);
1987 inode = proc_pid_make_inode(dir->i_sb, task);
1992 inode->i_mode = S_IFREG | S_IRUSR;
1993 inode->i_fop = &proc_fdinfo_file_operations;
1994 dentry->d_op = &tid_fd_dentry_operations;
1995 d_add(dentry, inode);
1996 /* Close the race of the process dying before we return the dentry */
1997 if (tid_fd_revalidate(dentry, NULL))
2004 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2005 struct dentry *dentry,
2006 struct nameidata *nd)
2008 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2011 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2013 return proc_readfd_common(filp, dirent, filldir,
2014 proc_fdinfo_instantiate);
2017 static const struct file_operations proc_fdinfo_operations = {
2018 .read = generic_read_dir,
2019 .readdir = proc_readfdinfo,
2023 * proc directories can do almost nothing..
2025 static const struct inode_operations proc_fdinfo_inode_operations = {
2026 .lookup = proc_lookupfdinfo,
2027 .setattr = proc_setattr,
2031 static struct dentry *proc_pident_instantiate(struct inode *dir,
2032 struct dentry *dentry, struct task_struct *task, const void *ptr)
2034 const struct pid_entry *p = ptr;
2035 struct inode *inode;
2036 struct proc_inode *ei;
2037 struct dentry *error = ERR_PTR(-ENOENT);
2039 inode = proc_pid_make_inode(dir->i_sb, task);
2044 inode->i_mode = p->mode;
2045 if (S_ISDIR(inode->i_mode))
2046 inode->i_nlink = 2; /* Use getattr to fix if necessary */
2048 inode->i_op = p->iop;
2050 inode->i_fop = p->fop;
2052 dentry->d_op = &pid_dentry_operations;
2053 d_add(dentry, inode);
2054 /* Close the race of the process dying before we return the dentry */
2055 if (pid_revalidate(dentry, NULL))
2061 static struct dentry *proc_pident_lookup(struct inode *dir,
2062 struct dentry *dentry,
2063 const struct pid_entry *ents,
2066 struct dentry *error;
2067 struct task_struct *task = get_proc_task(dir);
2068 const struct pid_entry *p, *last;
2070 error = ERR_PTR(-ENOENT);
2076 * Yes, it does not scale. And it should not. Don't add
2077 * new entries into /proc/<tgid>/ without very good reasons.
2079 last = &ents[nents - 1];
2080 for (p = ents; p <= last; p++) {
2081 if (p->len != dentry->d_name.len)
2083 if (!memcmp(dentry->d_name.name, p->name, p->len))
2089 error = proc_pident_instantiate(dir, dentry, task, p);
2091 put_task_struct(task);
2096 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2097 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2099 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2100 proc_pident_instantiate, task, p);
2103 static int proc_pident_readdir(struct file *filp,
2104 void *dirent, filldir_t filldir,
2105 const struct pid_entry *ents, unsigned int nents)
2108 struct dentry *dentry = filp->f_path.dentry;
2109 struct inode *inode = dentry->d_inode;
2110 struct task_struct *task = get_proc_task(inode);
2111 const struct pid_entry *p, *last;
2124 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2130 ino = parent_ino(dentry);
2131 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2143 last = &ents[nents - 1];
2145 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2154 put_task_struct(task);
2159 #ifdef CONFIG_SECURITY
2160 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2161 size_t count, loff_t *ppos)
2163 struct inode * inode = file->f_path.dentry->d_inode;
2166 struct task_struct *task = get_proc_task(inode);
2171 length = security_getprocattr(task,
2172 (char*)file->f_path.dentry->d_name.name,
2174 put_task_struct(task);
2176 length = simple_read_from_buffer(buf, count, ppos, p, length);
2181 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2182 size_t count, loff_t *ppos)
2184 struct inode * inode = file->f_path.dentry->d_inode;
2187 struct task_struct *task = get_proc_task(inode);
2192 if (count > PAGE_SIZE)
2195 /* No partial writes. */
2201 page = (char*)__get_free_page(GFP_TEMPORARY);
2206 if (copy_from_user(page, buf, count))
2209 /* Guard against adverse ptrace interaction */
2210 length = mutex_lock_interruptible(&task->cred_guard_mutex);
2214 length = security_setprocattr(task,
2215 (char*)file->f_path.dentry->d_name.name,
2216 (void*)page, count);
2217 mutex_unlock(&task->cred_guard_mutex);
2219 free_page((unsigned long) page);
2221 put_task_struct(task);
2226 static const struct file_operations proc_pid_attr_operations = {
2227 .read = proc_pid_attr_read,
2228 .write = proc_pid_attr_write,
2231 static const struct pid_entry attr_dir_stuff[] = {
2232 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2233 REG("prev", S_IRUGO, proc_pid_attr_operations),
2234 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2235 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2236 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2237 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2240 static int proc_attr_dir_readdir(struct file * filp,
2241 void * dirent, filldir_t filldir)
2243 return proc_pident_readdir(filp,dirent,filldir,
2244 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2247 static const struct file_operations proc_attr_dir_operations = {
2248 .read = generic_read_dir,
2249 .readdir = proc_attr_dir_readdir,
2252 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2253 struct dentry *dentry, struct nameidata *nd)
2255 return proc_pident_lookup(dir, dentry,
2256 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2259 static const struct inode_operations proc_attr_dir_inode_operations = {
2260 .lookup = proc_attr_dir_lookup,
2261 .getattr = pid_getattr,
2262 .setattr = proc_setattr,
2267 #ifdef CONFIG_ELF_CORE
2268 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2269 size_t count, loff_t *ppos)
2271 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2272 struct mm_struct *mm;
2273 char buffer[PROC_NUMBUF];
2281 mm = get_task_mm(task);
2283 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2284 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2285 MMF_DUMP_FILTER_SHIFT));
2287 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2290 put_task_struct(task);
2295 static ssize_t proc_coredump_filter_write(struct file *file,
2296 const char __user *buf,
2300 struct task_struct *task;
2301 struct mm_struct *mm;
2302 char buffer[PROC_NUMBUF], *end;
2309 memset(buffer, 0, sizeof(buffer));
2310 if (count > sizeof(buffer) - 1)
2311 count = sizeof(buffer) - 1;
2312 if (copy_from_user(buffer, buf, count))
2316 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2319 if (end - buffer == 0)
2323 task = get_proc_task(file->f_dentry->d_inode);
2328 mm = get_task_mm(task);
2332 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2334 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2336 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2341 put_task_struct(task);
2346 static const struct file_operations proc_coredump_filter_operations = {
2347 .read = proc_coredump_filter_read,
2348 .write = proc_coredump_filter_write,
2355 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2358 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2359 pid_t tgid = task_tgid_nr_ns(current, ns);
2360 char tmp[PROC_NUMBUF];
2363 sprintf(tmp, "%d", tgid);
2364 return vfs_readlink(dentry,buffer,buflen,tmp);
2367 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2369 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2370 pid_t tgid = task_tgid_nr_ns(current, ns);
2371 char *name = ERR_PTR(-ENOENT);
2375 name = ERR_PTR(-ENOMEM);
2377 sprintf(name, "%d", tgid);
2379 nd_set_link(nd, name);
2383 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2386 char *s = nd_get_link(nd);
2391 static const struct inode_operations proc_self_inode_operations = {
2392 .readlink = proc_self_readlink,
2393 .follow_link = proc_self_follow_link,
2394 .put_link = proc_self_put_link,
2400 * These are the directory entries in the root directory of /proc
2401 * that properly belong to the /proc filesystem, as they describe
2402 * describe something that is process related.
2404 static const struct pid_entry proc_base_stuff[] = {
2405 NOD("self", S_IFLNK|S_IRWXUGO,
2406 &proc_self_inode_operations, NULL, {}),
2410 * Exceptional case: normally we are not allowed to unhash a busy
2411 * directory. In this case, however, we can do it - no aliasing problems
2412 * due to the way we treat inodes.
2414 static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
2416 struct inode *inode = dentry->d_inode;
2417 struct task_struct *task = get_proc_task(inode);
2419 put_task_struct(task);
2426 static const struct dentry_operations proc_base_dentry_operations =
2428 .d_revalidate = proc_base_revalidate,
2429 .d_delete = pid_delete_dentry,
2432 static struct dentry *proc_base_instantiate(struct inode *dir,
2433 struct dentry *dentry, struct task_struct *task, const void *ptr)
2435 const struct pid_entry *p = ptr;
2436 struct inode *inode;
2437 struct proc_inode *ei;
2438 struct dentry *error = ERR_PTR(-EINVAL);
2440 /* Allocate the inode */
2441 error = ERR_PTR(-ENOMEM);
2442 inode = new_inode(dir->i_sb);
2446 /* Initialize the inode */
2448 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2451 * grab the reference to the task.
2453 ei->pid = get_task_pid(task, PIDTYPE_PID);
2457 inode->i_mode = p->mode;
2458 if (S_ISDIR(inode->i_mode))
2460 if (S_ISLNK(inode->i_mode))
2463 inode->i_op = p->iop;
2465 inode->i_fop = p->fop;
2467 dentry->d_op = &proc_base_dentry_operations;
2468 d_add(dentry, inode);
2477 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2479 struct dentry *error;
2480 struct task_struct *task = get_proc_task(dir);
2481 const struct pid_entry *p, *last;
2483 error = ERR_PTR(-ENOENT);
2488 /* Lookup the directory entry */
2489 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2490 for (p = proc_base_stuff; p <= last; p++) {
2491 if (p->len != dentry->d_name.len)
2493 if (!memcmp(dentry->d_name.name, p->name, p->len))
2499 error = proc_base_instantiate(dir, dentry, task, p);
2502 put_task_struct(task);
2507 static int proc_base_fill_cache(struct file *filp, void *dirent,
2508 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2510 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2511 proc_base_instantiate, task, p);
2514 #ifdef CONFIG_TASK_IO_ACCOUNTING
2515 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2517 struct task_io_accounting acct = task->ioac;
2518 unsigned long flags;
2520 if (whole && lock_task_sighand(task, &flags)) {
2521 struct task_struct *t = task;
2523 task_io_accounting_add(&acct, &task->signal->ioac);
2524 while_each_thread(task, t)
2525 task_io_accounting_add(&acct, &t->ioac);
2527 unlock_task_sighand(task, &flags);
2529 return sprintf(buffer,
2534 "read_bytes: %llu\n"
2535 "write_bytes: %llu\n"
2536 "cancelled_write_bytes: %llu\n",
2537 (unsigned long long)acct.rchar,
2538 (unsigned long long)acct.wchar,
2539 (unsigned long long)acct.syscr,
2540 (unsigned long long)acct.syscw,
2541 (unsigned long long)acct.read_bytes,
2542 (unsigned long long)acct.write_bytes,
2543 (unsigned long long)acct.cancelled_write_bytes);
2546 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2548 return do_io_accounting(task, buffer, 0);
2551 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2553 return do_io_accounting(task, buffer, 1);
2555 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2557 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2558 struct pid *pid, struct task_struct *task)
2560 seq_printf(m, "%08x\n", task->personality);
2567 static const struct file_operations proc_task_operations;
2568 static const struct inode_operations proc_task_inode_operations;
2570 static const struct pid_entry tgid_base_stuff[] = {
2571 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2572 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2573 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2575 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2577 REG("environ", S_IRUSR, proc_environ_operations),
2578 INF("auxv", S_IRUSR, proc_pid_auxv),
2579 ONE("status", S_IRUGO, proc_pid_status),
2580 ONE("personality", S_IRUSR, proc_pid_personality),
2581 INF("limits", S_IRUSR, proc_pid_limits),
2582 #ifdef CONFIG_SCHED_DEBUG
2583 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2585 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2586 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2587 INF("syscall", S_IRUSR, proc_pid_syscall),
2589 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2590 ONE("stat", S_IRUGO, proc_tgid_stat),
2591 ONE("statm", S_IRUGO, proc_pid_statm),
2592 REG("maps", S_IRUGO, proc_maps_operations),
2594 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2596 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2597 LNK("cwd", proc_cwd_link),
2598 LNK("root", proc_root_link),
2599 LNK("exe", proc_exe_link),
2600 REG("mounts", S_IRUGO, proc_mounts_operations),
2601 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2602 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2603 #ifdef CONFIG_PROC_PAGE_MONITOR
2604 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2605 REG("smaps", S_IRUGO, proc_smaps_operations),
2606 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2608 #ifdef CONFIG_SECURITY
2609 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2611 #ifdef CONFIG_KALLSYMS
2612 INF("wchan", S_IRUGO, proc_pid_wchan),
2614 #ifdef CONFIG_STACKTRACE
2615 ONE("stack", S_IRUSR, proc_pid_stack),
2617 #ifdef CONFIG_SCHEDSTATS
2618 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2620 #ifdef CONFIG_LATENCYTOP
2621 REG("latency", S_IRUGO, proc_lstats_operations),
2623 #ifdef CONFIG_PROC_PID_CPUSET
2624 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2626 #ifdef CONFIG_CGROUPS
2627 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2629 INF("oom_score", S_IRUGO, proc_oom_score),
2630 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2631 #ifdef CONFIG_AUDITSYSCALL
2632 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2633 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2635 #ifdef CONFIG_FAULT_INJECTION
2636 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2638 #ifdef CONFIG_ELF_CORE
2639 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2641 #ifdef CONFIG_TASK_IO_ACCOUNTING
2642 INF("io", S_IRUGO, proc_tgid_io_accounting),
2646 static int proc_tgid_base_readdir(struct file * filp,
2647 void * dirent, filldir_t filldir)
2649 return proc_pident_readdir(filp,dirent,filldir,
2650 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2653 static const struct file_operations proc_tgid_base_operations = {
2654 .read = generic_read_dir,
2655 .readdir = proc_tgid_base_readdir,
2658 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2659 return proc_pident_lookup(dir, dentry,
2660 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2663 static const struct inode_operations proc_tgid_base_inode_operations = {
2664 .lookup = proc_tgid_base_lookup,
2665 .getattr = pid_getattr,
2666 .setattr = proc_setattr,
2669 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2671 struct dentry *dentry, *leader, *dir;
2672 char buf[PROC_NUMBUF];
2676 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2677 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2679 shrink_dcache_parent(dentry);
2685 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2686 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2691 name.len = strlen(name.name);
2692 dir = d_hash_and_lookup(leader, &name);
2694 goto out_put_leader;
2697 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2698 dentry = d_hash_and_lookup(dir, &name);
2700 shrink_dcache_parent(dentry);
2713 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2714 * @task: task that should be flushed.
2716 * When flushing dentries from proc, one needs to flush them from global
2717 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2718 * in. This call is supposed to do all of this job.
2720 * Looks in the dcache for
2722 * /proc/@tgid/task/@pid
2723 * if either directory is present flushes it and all of it'ts children
2726 * It is safe and reasonable to cache /proc entries for a task until
2727 * that task exits. After that they just clog up the dcache with
2728 * useless entries, possibly causing useful dcache entries to be
2729 * flushed instead. This routine is proved to flush those useless
2730 * dcache entries at process exit time.
2732 * NOTE: This routine is just an optimization so it does not guarantee
2733 * that no dcache entries will exist at process exit time it
2734 * just makes it very unlikely that any will persist.
2737 void proc_flush_task(struct task_struct *task)
2740 struct pid *pid, *tgid;
2743 pid = task_pid(task);
2744 tgid = task_tgid(task);
2746 for (i = 0; i <= pid->level; i++) {
2747 upid = &pid->numbers[i];
2748 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2749 tgid->numbers[i].nr);
2752 upid = &pid->numbers[pid->level];
2754 pid_ns_release_proc(upid->ns);
2757 static struct dentry *proc_pid_instantiate(struct inode *dir,
2758 struct dentry * dentry,
2759 struct task_struct *task, const void *ptr)
2761 struct dentry *error = ERR_PTR(-ENOENT);
2762 struct inode *inode;
2764 inode = proc_pid_make_inode(dir->i_sb, task);
2768 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2769 inode->i_op = &proc_tgid_base_inode_operations;
2770 inode->i_fop = &proc_tgid_base_operations;
2771 inode->i_flags|=S_IMMUTABLE;
2773 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
2774 ARRAY_SIZE(tgid_base_stuff));
2776 dentry->d_op = &pid_dentry_operations;
2778 d_add(dentry, inode);
2779 /* Close the race of the process dying before we return the dentry */
2780 if (pid_revalidate(dentry, NULL))
2786 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2788 struct dentry *result = ERR_PTR(-ENOENT);
2789 struct task_struct *task;
2791 struct pid_namespace *ns;
2793 result = proc_base_lookup(dir, dentry);
2794 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2797 tgid = name_to_int(dentry);
2801 ns = dentry->d_sb->s_fs_info;
2803 task = find_task_by_pid_ns(tgid, ns);
2805 get_task_struct(task);
2810 result = proc_pid_instantiate(dir, dentry, task, NULL);
2811 put_task_struct(task);
2817 * Find the first task with tgid >= tgid
2822 struct task_struct *task;
2824 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2829 put_task_struct(iter.task);
2833 pid = find_ge_pid(iter.tgid, ns);
2835 iter.tgid = pid_nr_ns(pid, ns);
2836 iter.task = pid_task(pid, PIDTYPE_PID);
2837 /* What we to know is if the pid we have find is the
2838 * pid of a thread_group_leader. Testing for task
2839 * being a thread_group_leader is the obvious thing
2840 * todo but there is a window when it fails, due to
2841 * the pid transfer logic in de_thread.
2843 * So we perform the straight forward test of seeing
2844 * if the pid we have found is the pid of a thread
2845 * group leader, and don't worry if the task we have
2846 * found doesn't happen to be a thread group leader.
2847 * As we don't care in the case of readdir.
2849 if (!iter.task || !has_group_leader_pid(iter.task)) {
2853 get_task_struct(iter.task);
2859 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2861 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2862 struct tgid_iter iter)
2864 char name[PROC_NUMBUF];
2865 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
2866 return proc_fill_cache(filp, dirent, filldir, name, len,
2867 proc_pid_instantiate, iter.task, NULL);
2870 /* for the /proc/ directory itself, after non-process stuff has been done */
2871 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2873 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2874 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
2875 struct tgid_iter iter;
2876 struct pid_namespace *ns;
2881 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
2882 const struct pid_entry *p = &proc_base_stuff[nr];
2883 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
2887 ns = filp->f_dentry->d_sb->s_fs_info;
2889 iter.tgid = filp->f_pos - TGID_OFFSET;
2890 for (iter = next_tgid(ns, iter);
2892 iter.tgid += 1, iter = next_tgid(ns, iter)) {
2893 filp->f_pos = iter.tgid + TGID_OFFSET;
2894 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
2895 put_task_struct(iter.task);
2899 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
2901 put_task_struct(reaper);
2909 static const struct pid_entry tid_base_stuff[] = {
2910 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2911 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fd_operations),
2912 REG("environ", S_IRUSR, proc_environ_operations),
2913 INF("auxv", S_IRUSR, proc_pid_auxv),
2914 ONE("status", S_IRUGO, proc_pid_status),
2915 ONE("personality", S_IRUSR, proc_pid_personality),
2916 INF("limits", S_IRUSR, proc_pid_limits),
2917 #ifdef CONFIG_SCHED_DEBUG
2918 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2920 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2921 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2922 INF("syscall", S_IRUSR, proc_pid_syscall),
2924 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2925 ONE("stat", S_IRUGO, proc_tid_stat),
2926 ONE("statm", S_IRUGO, proc_pid_statm),
2927 REG("maps", S_IRUGO, proc_maps_operations),
2929 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
2931 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2932 LNK("cwd", proc_cwd_link),
2933 LNK("root", proc_root_link),
2934 LNK("exe", proc_exe_link),
2935 REG("mounts", S_IRUGO, proc_mounts_operations),
2936 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2937 #ifdef CONFIG_PROC_PAGE_MONITOR
2938 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2939 REG("smaps", S_IRUGO, proc_smaps_operations),
2940 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2942 #ifdef CONFIG_SECURITY
2943 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2945 #ifdef CONFIG_KALLSYMS
2946 INF("wchan", S_IRUGO, proc_pid_wchan),
2948 #ifdef CONFIG_STACKTRACE
2949 ONE("stack", S_IRUSR, proc_pid_stack),
2951 #ifdef CONFIG_SCHEDSTATS
2952 INF("schedstat", S_IRUGO, proc_pid_schedstat),
2954 #ifdef CONFIG_LATENCYTOP
2955 REG("latency", S_IRUGO, proc_lstats_operations),
2957 #ifdef CONFIG_PROC_PID_CPUSET
2958 REG("cpuset", S_IRUGO, proc_cpuset_operations),
2960 #ifdef CONFIG_CGROUPS
2961 REG("cgroup", S_IRUGO, proc_cgroup_operations),
2963 INF("oom_score", S_IRUGO, proc_oom_score),
2964 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2965 #ifdef CONFIG_AUDITSYSCALL
2966 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2967 REG("sessionid", S_IRUSR, proc_sessionid_operations),
2969 #ifdef CONFIG_FAULT_INJECTION
2970 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2972 #ifdef CONFIG_TASK_IO_ACCOUNTING
2973 INF("io", S_IRUGO, proc_tid_io_accounting),
2977 static int proc_tid_base_readdir(struct file * filp,
2978 void * dirent, filldir_t filldir)
2980 return proc_pident_readdir(filp,dirent,filldir,
2981 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
2984 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2985 return proc_pident_lookup(dir, dentry,
2986 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
2989 static const struct file_operations proc_tid_base_operations = {
2990 .read = generic_read_dir,
2991 .readdir = proc_tid_base_readdir,
2994 static const struct inode_operations proc_tid_base_inode_operations = {
2995 .lookup = proc_tid_base_lookup,
2996 .getattr = pid_getattr,
2997 .setattr = proc_setattr,
3000 static struct dentry *proc_task_instantiate(struct inode *dir,
3001 struct dentry *dentry, struct task_struct *task, const void *ptr)
3003 struct dentry *error = ERR_PTR(-ENOENT);
3004 struct inode *inode;
3005 inode = proc_pid_make_inode(dir->i_sb, task);
3009 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3010 inode->i_op = &proc_tid_base_inode_operations;
3011 inode->i_fop = &proc_tid_base_operations;
3012 inode->i_flags|=S_IMMUTABLE;
3014 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
3015 ARRAY_SIZE(tid_base_stuff));
3017 dentry->d_op = &pid_dentry_operations;
3019 d_add(dentry, inode);
3020 /* Close the race of the process dying before we return the dentry */
3021 if (pid_revalidate(dentry, NULL))
3027 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3029 struct dentry *result = ERR_PTR(-ENOENT);
3030 struct task_struct *task;
3031 struct task_struct *leader = get_proc_task(dir);
3033 struct pid_namespace *ns;
3038 tid = name_to_int(dentry);
3042 ns = dentry->d_sb->s_fs_info;
3044 task = find_task_by_pid_ns(tid, ns);
3046 get_task_struct(task);
3050 if (!same_thread_group(leader, task))
3053 result = proc_task_instantiate(dir, dentry, task, NULL);
3055 put_task_struct(task);
3057 put_task_struct(leader);
3063 * Find the first tid of a thread group to return to user space.
3065 * Usually this is just the thread group leader, but if the users
3066 * buffer was too small or there was a seek into the middle of the
3067 * directory we have more work todo.
3069 * In the case of a short read we start with find_task_by_pid.
3071 * In the case of a seek we start with the leader and walk nr
3074 static struct task_struct *first_tid(struct task_struct *leader,
3075 int tid, int nr, struct pid_namespace *ns)
3077 struct task_struct *pos;
3080 /* Attempt to start with the pid of a thread */
3081 if (tid && (nr > 0)) {
3082 pos = find_task_by_pid_ns(tid, ns);
3083 if (pos && (pos->group_leader == leader))
3087 /* If nr exceeds the number of threads there is nothing todo */
3089 if (nr && nr >= get_nr_threads(leader))
3092 /* If we haven't found our starting place yet start
3093 * with the leader and walk nr threads forward.
3095 for (pos = leader; nr > 0; --nr) {
3096 pos = next_thread(pos);
3097 if (pos == leader) {
3103 get_task_struct(pos);
3110 * Find the next thread in the thread list.
3111 * Return NULL if there is an error or no next thread.
3113 * The reference to the input task_struct is released.
3115 static struct task_struct *next_tid(struct task_struct *start)
3117 struct task_struct *pos = NULL;
3119 if (pid_alive(start)) {
3120 pos = next_thread(start);
3121 if (thread_group_leader(pos))
3124 get_task_struct(pos);
3127 put_task_struct(start);
3131 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3132 struct task_struct *task, int tid)
3134 char name[PROC_NUMBUF];
3135 int len = snprintf(name, sizeof(name), "%d", tid);
3136 return proc_fill_cache(filp, dirent, filldir, name, len,
3137 proc_task_instantiate, task, NULL);
3140 /* for the /proc/TGID/task/ directories */
3141 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3143 struct dentry *dentry = filp->f_path.dentry;
3144 struct inode *inode = dentry->d_inode;
3145 struct task_struct *leader = NULL;
3146 struct task_struct *task;
3147 int retval = -ENOENT;
3150 struct pid_namespace *ns;
3152 task = get_proc_task(inode);
3156 if (pid_alive(task)) {
3157 leader = task->group_leader;
3158 get_task_struct(leader);
3161 put_task_struct(task);
3166 switch ((unsigned long)filp->f_pos) {
3169 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3174 ino = parent_ino(dentry);
3175 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3181 /* f_version caches the tgid value that the last readdir call couldn't
3182 * return. lseek aka telldir automagically resets f_version to 0.
3184 ns = filp->f_dentry->d_sb->s_fs_info;
3185 tid = (int)filp->f_version;
3186 filp->f_version = 0;
3187 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3189 task = next_tid(task), filp->f_pos++) {
3190 tid = task_pid_nr_ns(task, ns);
3191 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3192 /* returning this tgid failed, save it as the first
3193 * pid for the next readir call */
3194 filp->f_version = (u64)tid;
3195 put_task_struct(task);
3200 put_task_struct(leader);
3205 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3207 struct inode *inode = dentry->d_inode;
3208 struct task_struct *p = get_proc_task(inode);
3209 generic_fillattr(inode, stat);
3212 stat->nlink += get_nr_threads(p);
3219 static const struct inode_operations proc_task_inode_operations = {
3220 .lookup = proc_task_lookup,
3221 .getattr = proc_task_getattr,
3222 .setattr = proc_setattr,
3225 static const struct file_operations proc_task_operations = {
3226 .read = generic_read_dir,
3227 .readdir = proc_task_readdir,