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/config.h>
53 #include <linux/errno.h>
54 #include <linux/time.h>
55 #include <linux/proc_fs.h>
56 #include <linux/stat.h>
57 #include <linux/init.h>
58 #include <linux/capability.h>
59 #include <linux/file.h>
60 #include <linux/string.h>
61 #include <linux/seq_file.h>
62 #include <linux/namei.h>
63 #include <linux/namespace.h>
65 #include <linux/smp_lock.h>
66 #include <linux/rcupdate.h>
67 #include <linux/kallsyms.h>
68 #include <linux/mount.h>
69 #include <linux/security.h>
70 #include <linux/ptrace.h>
71 #include <linux/seccomp.h>
72 #include <linux/cpuset.h>
73 #include <linux/audit.h>
74 #include <linux/poll.h>
78 * Implementing inode permission operations in /proc is almost
79 * certainly an error. Permission checks need to happen during
80 * each system call not at open time. The reason is that most of
81 * what we wish to check for permissions in /proc varies at runtime.
83 * The classic example of a problem is opening file descriptors
84 * in /proc for a task before it execs a suid executable.
88 * For hysterical raisins we keep the same inumbers as in the old procfs.
89 * Feel free to change the macro below - just keep the range distinct from
90 * inumbers of the rest of procfs (currently those are in 0x0000--0xffff).
91 * As soon as we'll get a separate superblock we will be able to forget
92 * about magical ranges too.
95 #define fake_ino(pid,ino) (((pid)<<16)|(ino))
97 enum pid_directory_inos {
102 #ifdef CONFIG_SECCOMP
117 PROC_TGID_MOUNTSTATS,
122 #ifdef CONFIG_SCHEDSTATS
125 #ifdef CONFIG_CPUSETS
128 #ifdef CONFIG_SECURITY
130 PROC_TGID_ATTR_CURRENT,
133 PROC_TGID_ATTR_FSCREATE,
134 PROC_TGID_ATTR_KEYCREATE,
136 #ifdef CONFIG_AUDITSYSCALL
140 PROC_TGID_OOM_ADJUST,
144 #ifdef CONFIG_SECCOMP
164 #ifdef CONFIG_SCHEDSTATS
167 #ifdef CONFIG_CPUSETS
170 #ifdef CONFIG_SECURITY
172 PROC_TID_ATTR_CURRENT,
175 PROC_TID_ATTR_FSCREATE,
176 PROC_TID_ATTR_KEYCREATE,
178 #ifdef CONFIG_AUDITSYSCALL
184 /* Add new entries before this */
185 PROC_TID_FD_DIR = 0x8000, /* 0x8000-0xffff */
188 /* Worst case buffer size needed for holding an integer. */
189 #define PROC_NUMBUF 10
198 #define E(type,name,mode) {(type),sizeof(name)-1,(name),(mode)}
200 static struct pid_entry tgid_base_stuff[] = {
201 E(PROC_TGID_TASK, "task", S_IFDIR|S_IRUGO|S_IXUGO),
202 E(PROC_TGID_FD, "fd", S_IFDIR|S_IRUSR|S_IXUSR),
203 E(PROC_TGID_ENVIRON, "environ", S_IFREG|S_IRUSR),
204 E(PROC_TGID_AUXV, "auxv", S_IFREG|S_IRUSR),
205 E(PROC_TGID_STATUS, "status", S_IFREG|S_IRUGO),
206 E(PROC_TGID_CMDLINE, "cmdline", S_IFREG|S_IRUGO),
207 E(PROC_TGID_STAT, "stat", S_IFREG|S_IRUGO),
208 E(PROC_TGID_STATM, "statm", S_IFREG|S_IRUGO),
209 E(PROC_TGID_MAPS, "maps", S_IFREG|S_IRUGO),
211 E(PROC_TGID_NUMA_MAPS, "numa_maps", S_IFREG|S_IRUGO),
213 E(PROC_TGID_MEM, "mem", S_IFREG|S_IRUSR|S_IWUSR),
214 #ifdef CONFIG_SECCOMP
215 E(PROC_TGID_SECCOMP, "seccomp", S_IFREG|S_IRUSR|S_IWUSR),
217 E(PROC_TGID_CWD, "cwd", S_IFLNK|S_IRWXUGO),
218 E(PROC_TGID_ROOT, "root", S_IFLNK|S_IRWXUGO),
219 E(PROC_TGID_EXE, "exe", S_IFLNK|S_IRWXUGO),
220 E(PROC_TGID_MOUNTS, "mounts", S_IFREG|S_IRUGO),
221 E(PROC_TGID_MOUNTSTATS, "mountstats", S_IFREG|S_IRUSR),
223 E(PROC_TGID_SMAPS, "smaps", S_IFREG|S_IRUGO),
225 #ifdef CONFIG_SECURITY
226 E(PROC_TGID_ATTR, "attr", S_IFDIR|S_IRUGO|S_IXUGO),
228 #ifdef CONFIG_KALLSYMS
229 E(PROC_TGID_WCHAN, "wchan", S_IFREG|S_IRUGO),
231 #ifdef CONFIG_SCHEDSTATS
232 E(PROC_TGID_SCHEDSTAT, "schedstat", S_IFREG|S_IRUGO),
234 #ifdef CONFIG_CPUSETS
235 E(PROC_TGID_CPUSET, "cpuset", S_IFREG|S_IRUGO),
237 E(PROC_TGID_OOM_SCORE, "oom_score",S_IFREG|S_IRUGO),
238 E(PROC_TGID_OOM_ADJUST,"oom_adj", S_IFREG|S_IRUGO|S_IWUSR),
239 #ifdef CONFIG_AUDITSYSCALL
240 E(PROC_TGID_LOGINUID, "loginuid", S_IFREG|S_IWUSR|S_IRUGO),
244 static struct pid_entry tid_base_stuff[] = {
245 E(PROC_TID_FD, "fd", S_IFDIR|S_IRUSR|S_IXUSR),
246 E(PROC_TID_ENVIRON, "environ", S_IFREG|S_IRUSR),
247 E(PROC_TID_AUXV, "auxv", S_IFREG|S_IRUSR),
248 E(PROC_TID_STATUS, "status", S_IFREG|S_IRUGO),
249 E(PROC_TID_CMDLINE, "cmdline", S_IFREG|S_IRUGO),
250 E(PROC_TID_STAT, "stat", S_IFREG|S_IRUGO),
251 E(PROC_TID_STATM, "statm", S_IFREG|S_IRUGO),
252 E(PROC_TID_MAPS, "maps", S_IFREG|S_IRUGO),
254 E(PROC_TID_NUMA_MAPS, "numa_maps", S_IFREG|S_IRUGO),
256 E(PROC_TID_MEM, "mem", S_IFREG|S_IRUSR|S_IWUSR),
257 #ifdef CONFIG_SECCOMP
258 E(PROC_TID_SECCOMP, "seccomp", S_IFREG|S_IRUSR|S_IWUSR),
260 E(PROC_TID_CWD, "cwd", S_IFLNK|S_IRWXUGO),
261 E(PROC_TID_ROOT, "root", S_IFLNK|S_IRWXUGO),
262 E(PROC_TID_EXE, "exe", S_IFLNK|S_IRWXUGO),
263 E(PROC_TID_MOUNTS, "mounts", S_IFREG|S_IRUGO),
265 E(PROC_TID_SMAPS, "smaps", S_IFREG|S_IRUGO),
267 #ifdef CONFIG_SECURITY
268 E(PROC_TID_ATTR, "attr", S_IFDIR|S_IRUGO|S_IXUGO),
270 #ifdef CONFIG_KALLSYMS
271 E(PROC_TID_WCHAN, "wchan", S_IFREG|S_IRUGO),
273 #ifdef CONFIG_SCHEDSTATS
274 E(PROC_TID_SCHEDSTAT, "schedstat",S_IFREG|S_IRUGO),
276 #ifdef CONFIG_CPUSETS
277 E(PROC_TID_CPUSET, "cpuset", S_IFREG|S_IRUGO),
279 E(PROC_TID_OOM_SCORE, "oom_score",S_IFREG|S_IRUGO),
280 E(PROC_TID_OOM_ADJUST, "oom_adj", S_IFREG|S_IRUGO|S_IWUSR),
281 #ifdef CONFIG_AUDITSYSCALL
282 E(PROC_TID_LOGINUID, "loginuid", S_IFREG|S_IWUSR|S_IRUGO),
287 #ifdef CONFIG_SECURITY
288 static struct pid_entry tgid_attr_stuff[] = {
289 E(PROC_TGID_ATTR_CURRENT, "current", S_IFREG|S_IRUGO|S_IWUGO),
290 E(PROC_TGID_ATTR_PREV, "prev", S_IFREG|S_IRUGO),
291 E(PROC_TGID_ATTR_EXEC, "exec", S_IFREG|S_IRUGO|S_IWUGO),
292 E(PROC_TGID_ATTR_FSCREATE, "fscreate", S_IFREG|S_IRUGO|S_IWUGO),
293 E(PROC_TGID_ATTR_KEYCREATE, "keycreate", S_IFREG|S_IRUGO|S_IWUGO),
296 static struct pid_entry tid_attr_stuff[] = {
297 E(PROC_TID_ATTR_CURRENT, "current", S_IFREG|S_IRUGO|S_IWUGO),
298 E(PROC_TID_ATTR_PREV, "prev", S_IFREG|S_IRUGO),
299 E(PROC_TID_ATTR_EXEC, "exec", S_IFREG|S_IRUGO|S_IWUGO),
300 E(PROC_TID_ATTR_FSCREATE, "fscreate", S_IFREG|S_IRUGO|S_IWUGO),
301 E(PROC_TID_ATTR_KEYCREATE, "keycreate", S_IFREG|S_IRUGO|S_IWUGO),
308 static int proc_fd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
310 struct task_struct *task = proc_task(inode);
311 struct files_struct *files;
313 int fd = proc_fd(inode);
315 files = get_files_struct(task);
318 * We are not taking a ref to the file structure, so we must
321 spin_lock(&files->file_lock);
322 file = fcheck_files(files, fd);
324 *mnt = mntget(file->f_vfsmnt);
325 *dentry = dget(file->f_dentry);
326 spin_unlock(&files->file_lock);
327 put_files_struct(files);
330 spin_unlock(&files->file_lock);
331 put_files_struct(files);
336 static struct fs_struct *get_fs_struct(struct task_struct *task)
338 struct fs_struct *fs;
342 atomic_inc(&fs->count);
347 static int proc_cwd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
349 struct fs_struct *fs = get_fs_struct(proc_task(inode));
350 int result = -ENOENT;
352 read_lock(&fs->lock);
353 *mnt = mntget(fs->pwdmnt);
354 *dentry = dget(fs->pwd);
355 read_unlock(&fs->lock);
362 static int proc_root_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
364 struct fs_struct *fs = get_fs_struct(proc_task(inode));
365 int result = -ENOENT;
367 read_lock(&fs->lock);
368 *mnt = mntget(fs->rootmnt);
369 *dentry = dget(fs->root);
370 read_unlock(&fs->lock);
377 #define MAY_PTRACE(task) \
378 (task == current || \
379 (task->parent == current && \
380 (task->ptrace & PT_PTRACED) && \
381 (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
382 security_ptrace(current,task) == 0))
384 static int proc_pid_environ(struct task_struct *task, char * buffer)
387 struct mm_struct *mm = get_task_mm(task);
389 unsigned int len = mm->env_end - mm->env_start;
392 res = access_process_vm(task, mm->env_start, buffer, len, 0);
393 if (!ptrace_may_attach(task))
400 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
404 struct mm_struct *mm = get_task_mm(task);
408 goto out_mm; /* Shh! No looking before we're done */
410 len = mm->arg_end - mm->arg_start;
415 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
417 // If the nul at the end of args has been overwritten, then
418 // assume application is using setproctitle(3).
419 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
420 len = strnlen(buffer, res);
424 len = mm->env_end - mm->env_start;
425 if (len > PAGE_SIZE - res)
426 len = PAGE_SIZE - res;
427 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
428 res = strnlen(buffer, res);
437 static int proc_pid_auxv(struct task_struct *task, char *buffer)
440 struct mm_struct *mm = get_task_mm(task);
442 unsigned int nwords = 0;
445 while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
446 res = nwords * sizeof(mm->saved_auxv[0]);
449 memcpy(buffer, mm->saved_auxv, res);
456 #ifdef CONFIG_KALLSYMS
458 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
459 * Returns the resolved symbol. If that fails, simply return the address.
461 static int proc_pid_wchan(struct task_struct *task, char *buffer)
464 const char *sym_name;
465 unsigned long wchan, size, offset;
466 char namebuf[KSYM_NAME_LEN+1];
468 wchan = get_wchan(task);
470 sym_name = kallsyms_lookup(wchan, &size, &offset, &modname, namebuf);
472 return sprintf(buffer, "%s", sym_name);
473 return sprintf(buffer, "%lu", wchan);
475 #endif /* CONFIG_KALLSYMS */
477 #ifdef CONFIG_SCHEDSTATS
479 * Provides /proc/PID/schedstat
481 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
483 return sprintf(buffer, "%lu %lu %lu\n",
484 task->sched_info.cpu_time,
485 task->sched_info.run_delay,
486 task->sched_info.pcnt);
490 /* The badness from the OOM killer */
491 unsigned long badness(struct task_struct *p, unsigned long uptime);
492 static int proc_oom_score(struct task_struct *task, char *buffer)
494 unsigned long points;
495 struct timespec uptime;
497 do_posix_clock_monotonic_gettime(&uptime);
498 points = badness(task, uptime.tv_sec);
499 return sprintf(buffer, "%lu\n", points);
502 /************************************************************************/
503 /* Here the fs part begins */
504 /************************************************************************/
506 /* permission checks */
508 /* If the process being read is separated by chroot from the reading process,
509 * don't let the reader access the threads.
511 static int proc_check_chroot(struct dentry *de, struct vfsmount *mnt)
514 struct vfsmount *our_vfsmnt;
517 read_lock(¤t->fs->lock);
518 our_vfsmnt = mntget(current->fs->rootmnt);
519 base = dget(current->fs->root);
520 read_unlock(¤t->fs->lock);
522 spin_lock(&vfsmount_lock);
524 while (mnt != our_vfsmnt) {
525 if (mnt == mnt->mnt_parent)
527 de = mnt->mnt_mountpoint;
528 mnt = mnt->mnt_parent;
531 if (!is_subdir(de, base))
533 spin_unlock(&vfsmount_lock);
540 spin_unlock(&vfsmount_lock);
545 extern struct seq_operations mounts_op;
551 static int mounts_open(struct inode *inode, struct file *file)
553 struct task_struct *task = proc_task(inode);
554 struct namespace *namespace;
555 struct proc_mounts *p;
559 namespace = task->namespace;
561 get_namespace(namespace);
566 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
568 file->private_data = &p->m;
569 ret = seq_open(file, &mounts_op);
571 p->m.private = namespace;
572 p->event = namespace->event;
577 put_namespace(namespace);
582 static int mounts_release(struct inode *inode, struct file *file)
584 struct seq_file *m = file->private_data;
585 struct namespace *namespace = m->private;
586 put_namespace(namespace);
587 return seq_release(inode, file);
590 static unsigned mounts_poll(struct file *file, poll_table *wait)
592 struct proc_mounts *p = file->private_data;
593 struct namespace *ns = p->m.private;
596 poll_wait(file, &ns->poll, wait);
598 spin_lock(&vfsmount_lock);
599 if (p->event != ns->event) {
600 p->event = ns->event;
603 spin_unlock(&vfsmount_lock);
608 static struct file_operations proc_mounts_operations = {
612 .release = mounts_release,
616 extern struct seq_operations mountstats_op;
617 static int mountstats_open(struct inode *inode, struct file *file)
619 struct task_struct *task = proc_task(inode);
620 int ret = seq_open(file, &mountstats_op);
623 struct seq_file *m = file->private_data;
624 struct namespace *namespace;
626 namespace = task->namespace;
628 get_namespace(namespace);
632 m->private = namespace;
634 seq_release(inode, file);
641 static struct file_operations proc_mountstats_operations = {
642 .open = mountstats_open,
645 .release = mounts_release,
648 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
650 static ssize_t proc_info_read(struct file * file, char __user * buf,
651 size_t count, loff_t *ppos)
653 struct inode * inode = file->f_dentry->d_inode;
656 struct task_struct *task = proc_task(inode);
658 if (count > PROC_BLOCK_SIZE)
659 count = PROC_BLOCK_SIZE;
660 if (!(page = __get_free_page(GFP_KERNEL)))
663 length = PROC_I(inode)->op.proc_read(task, (char*)page);
666 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
671 static struct file_operations proc_info_file_operations = {
672 .read = proc_info_read,
675 static int mem_open(struct inode* inode, struct file* file)
677 file->private_data = (void*)((long)current->self_exec_id);
681 static ssize_t mem_read(struct file * file, char __user * buf,
682 size_t count, loff_t *ppos)
684 struct task_struct *task = proc_task(file->f_dentry->d_inode);
686 unsigned long src = *ppos;
688 struct mm_struct *mm;
690 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
694 page = (char *)__get_free_page(GFP_USER);
700 mm = get_task_mm(task);
706 if (file->private_data != (void*)((long)current->self_exec_id))
712 int this_len, retval;
714 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
715 retval = access_process_vm(task, src, page, this_len, 0);
716 if (!retval || !MAY_PTRACE(task) || !ptrace_may_attach(task)) {
722 if (copy_to_user(buf, page, retval)) {
737 free_page((unsigned long) page);
742 #define mem_write NULL
745 /* This is a security hazard */
746 static ssize_t mem_write(struct file * file, const char * buf,
747 size_t count, loff_t *ppos)
751 struct task_struct *task = proc_task(file->f_dentry->d_inode);
752 unsigned long dst = *ppos;
754 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
757 page = (char *)__get_free_page(GFP_USER);
762 int this_len, retval;
764 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
765 if (copy_from_user(page, buf, this_len)) {
769 retval = access_process_vm(task, dst, page, this_len, 1);
781 free_page((unsigned long) page);
786 static loff_t mem_lseek(struct file * file, loff_t offset, int orig)
790 file->f_pos = offset;
793 file->f_pos += offset;
798 force_successful_syscall_return();
802 static struct file_operations proc_mem_operations = {
809 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
810 size_t count, loff_t *ppos)
812 struct task_struct *task = proc_task(file->f_dentry->d_inode);
813 char buffer[PROC_NUMBUF];
815 int oom_adjust = task->oomkilladj;
816 loff_t __ppos = *ppos;
818 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
821 if (count > len-__ppos)
823 if (copy_to_user(buf, buffer + __ppos, count))
825 *ppos = __ppos + count;
829 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
830 size_t count, loff_t *ppos)
832 struct task_struct *task = proc_task(file->f_dentry->d_inode);
833 char buffer[PROC_NUMBUF], *end;
836 if (!capable(CAP_SYS_RESOURCE))
838 memset(buffer, 0, sizeof(buffer));
839 if (count > sizeof(buffer) - 1)
840 count = sizeof(buffer) - 1;
841 if (copy_from_user(buffer, buf, count))
843 oom_adjust = simple_strtol(buffer, &end, 0);
844 if ((oom_adjust < -16 || oom_adjust > 15) && oom_adjust != OOM_DISABLE)
848 task->oomkilladj = oom_adjust;
849 if (end - buffer == 0)
854 static struct file_operations proc_oom_adjust_operations = {
855 .read = oom_adjust_read,
856 .write = oom_adjust_write,
859 #ifdef CONFIG_AUDITSYSCALL
861 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
862 size_t count, loff_t *ppos)
864 struct inode * inode = file->f_dentry->d_inode;
865 struct task_struct *task = proc_task(inode);
867 char tmpbuf[TMPBUFLEN];
869 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
870 audit_get_loginuid(task->audit_context));
871 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
874 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
875 size_t count, loff_t *ppos)
877 struct inode * inode = file->f_dentry->d_inode;
880 struct task_struct *task = proc_task(inode);
883 if (!capable(CAP_AUDIT_CONTROL))
889 if (count >= PAGE_SIZE)
890 count = PAGE_SIZE - 1;
893 /* No partial writes. */
896 page = (char*)__get_free_page(GFP_USER);
900 if (copy_from_user(page, buf, count))
904 loginuid = simple_strtoul(page, &tmp, 10);
910 length = audit_set_loginuid(task, loginuid);
911 if (likely(length == 0))
915 free_page((unsigned long) page);
919 static struct file_operations proc_loginuid_operations = {
920 .read = proc_loginuid_read,
921 .write = proc_loginuid_write,
925 #ifdef CONFIG_SECCOMP
926 static ssize_t seccomp_read(struct file *file, char __user *buf,
927 size_t count, loff_t *ppos)
929 struct task_struct *tsk = proc_task(file->f_dentry->d_inode);
931 loff_t __ppos = *ppos;
934 /* no need to print the trailing zero, so use only len */
935 len = sprintf(__buf, "%u\n", tsk->seccomp.mode);
938 if (count > len - __ppos)
939 count = len - __ppos;
940 if (copy_to_user(buf, __buf + __ppos, count))
942 *ppos = __ppos + count;
946 static ssize_t seccomp_write(struct file *file, const char __user *buf,
947 size_t count, loff_t *ppos)
949 struct task_struct *tsk = proc_task(file->f_dentry->d_inode);
950 char __buf[20], *end;
951 unsigned int seccomp_mode;
953 /* can set it only once to be even more secure */
954 if (unlikely(tsk->seccomp.mode))
957 memset(__buf, 0, sizeof(__buf));
958 count = min(count, sizeof(__buf) - 1);
959 if (copy_from_user(__buf, buf, count))
961 seccomp_mode = simple_strtoul(__buf, &end, 0);
964 if (seccomp_mode && seccomp_mode <= NR_SECCOMP_MODES) {
965 tsk->seccomp.mode = seccomp_mode;
966 set_tsk_thread_flag(tsk, TIF_SECCOMP);
969 if (unlikely(!(end - __buf)))
974 static struct file_operations proc_seccomp_operations = {
975 .read = seccomp_read,
976 .write = seccomp_write,
978 #endif /* CONFIG_SECCOMP */
980 static int proc_check_dentry_visible(struct inode *inode,
981 struct dentry *dentry, struct vfsmount *mnt)
983 /* Verify that the current process can already see the
984 * file pointed at by the file descriptor.
985 * This prevents /proc from being an accidental information leak.
987 * This prevents access to files that are not visible do to
988 * being on the otherside of a chroot, in a different
989 * namespace, or are simply process local (like pipes).
991 struct task_struct *task;
992 struct files_struct *task_files, *files;
995 /* See if the the two tasks share a commone set of
996 * file descriptors. If so everything is visible.
998 task = proc_task(inode);
1001 files = get_files_struct(current);
1002 task_files = get_files_struct(task);
1003 if (files && task_files && (files == task_files))
1006 put_files_struct(task_files);
1008 put_files_struct(files);
1012 /* If the two tasks don't share a common set of file
1013 * descriptors see if the destination dentry is already
1014 * visible in the current tasks filesystem namespace.
1016 error = proc_check_chroot(dentry, mnt);
1022 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1024 struct inode *inode = dentry->d_inode;
1025 int error = -EACCES;
1027 /* We don't need a base pointer in the /proc filesystem */
1030 if (current->fsuid != inode->i_uid && !capable(CAP_DAC_OVERRIDE))
1033 error = PROC_I(inode)->op.proc_get_link(inode, &nd->dentry, &nd->mnt);
1034 nd->last_type = LAST_BIND;
1038 /* Only return files this task can already see */
1039 error = proc_check_dentry_visible(inode, nd->dentry, nd->mnt);
1043 return ERR_PTR(error);
1046 static int do_proc_readlink(struct dentry *dentry, struct vfsmount *mnt,
1047 char __user *buffer, int buflen)
1049 struct inode * inode;
1050 char *tmp = (char*)__get_free_page(GFP_KERNEL), *path;
1056 inode = dentry->d_inode;
1057 path = d_path(dentry, mnt, tmp, PAGE_SIZE);
1058 len = PTR_ERR(path);
1061 len = tmp + PAGE_SIZE - 1 - path;
1065 if (copy_to_user(buffer, path, len))
1068 free_page((unsigned long)tmp);
1072 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1074 int error = -EACCES;
1075 struct inode *inode = dentry->d_inode;
1077 struct vfsmount *mnt = NULL;
1080 if (current->fsuid != inode->i_uid && !capable(CAP_DAC_OVERRIDE))
1083 error = PROC_I(inode)->op.proc_get_link(inode, &de, &mnt);
1087 /* Only return files this task can already see */
1088 error = proc_check_dentry_visible(inode, de, mnt);
1092 error = do_proc_readlink(de, mnt, buffer, buflen);
1100 static struct inode_operations proc_pid_link_inode_operations = {
1101 .readlink = proc_pid_readlink,
1102 .follow_link = proc_pid_follow_link
1105 static int proc_readfd(struct file * filp, void * dirent, filldir_t filldir)
1107 struct dentry *dentry = filp->f_dentry;
1108 struct inode *inode = dentry->d_inode;
1109 struct task_struct *p = proc_task(inode);
1110 unsigned int fd, tid, ino;
1112 char buf[PROC_NUMBUF];
1113 struct files_struct * files;
1114 struct fdtable *fdt;
1125 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1129 ino = parent_ino(dentry);
1130 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1134 files = get_files_struct(p);
1138 fdt = files_fdtable(files);
1139 for (fd = filp->f_pos-2;
1141 fd++, filp->f_pos++) {
1144 if (!fcheck_files(files, fd))
1152 buf[j] = '0' + (i % 10);
1156 ino = fake_ino(tid, PROC_TID_FD_DIR + fd);
1157 if (filldir(dirent, buf+j, PROC_NUMBUF-j, fd+2, ino, DT_LNK) < 0) {
1164 put_files_struct(files);
1170 static int proc_pident_readdir(struct file *filp,
1171 void *dirent, filldir_t filldir,
1172 struct pid_entry *ents, unsigned int nents)
1176 struct dentry *dentry = filp->f_dentry;
1177 struct inode *inode = dentry->d_inode;
1178 struct pid_entry *p;
1183 if (!pid_alive(proc_task(inode)))
1187 pid = proc_task(inode)->pid;
1192 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
1198 ino = parent_ino(dentry);
1199 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
1212 if (filldir(dirent, p->name, p->len, filp->f_pos,
1213 fake_ino(pid, p->type), p->mode >> 12) < 0)
1225 static int proc_tgid_base_readdir(struct file * filp,
1226 void * dirent, filldir_t filldir)
1228 return proc_pident_readdir(filp,dirent,filldir,
1229 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1232 static int proc_tid_base_readdir(struct file * filp,
1233 void * dirent, filldir_t filldir)
1235 return proc_pident_readdir(filp,dirent,filldir,
1236 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
1239 /* building an inode */
1241 static int task_dumpable(struct task_struct *task)
1244 struct mm_struct *mm;
1249 dumpable = mm->dumpable;
1257 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task, int ino)
1259 struct inode * inode;
1260 struct proc_inode *ei;
1262 /* We need a new inode */
1264 inode = new_inode(sb);
1270 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1271 inode->i_ino = fake_ino(task->pid, ino);
1273 if (!pid_alive(task))
1277 * grab the reference to task.
1279 get_task_struct(task);
1283 if (task_dumpable(task)) {
1284 inode->i_uid = task->euid;
1285 inode->i_gid = task->egid;
1287 security_task_to_inode(task, inode);
1300 * Exceptional case: normally we are not allowed to unhash a busy
1301 * directory. In this case, however, we can do it - no aliasing problems
1302 * due to the way we treat inodes.
1304 * Rewrite the inode's ownerships here because the owning task may have
1305 * performed a setuid(), etc.
1307 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1309 struct inode *inode = dentry->d_inode;
1310 struct task_struct *task = proc_task(inode);
1311 if (pid_alive(task)) {
1312 if (task_dumpable(task)) {
1313 inode->i_uid = task->euid;
1314 inode->i_gid = task->egid;
1319 security_task_to_inode(task, inode);
1326 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1328 struct inode *inode = dentry->d_inode;
1329 struct task_struct *task = proc_task(inode);
1330 int fd = proc_fd(inode);
1331 struct files_struct *files;
1333 files = get_files_struct(task);
1336 if (fcheck_files(files, fd)) {
1338 put_files_struct(files);
1339 if (task_dumpable(task)) {
1340 inode->i_uid = task->euid;
1341 inode->i_gid = task->egid;
1346 security_task_to_inode(task, inode);
1350 put_files_struct(files);
1356 static int pid_delete_dentry(struct dentry * dentry)
1358 /* Is the task we represent dead?
1359 * If so, then don't put the dentry on the lru list,
1360 * kill it immediately.
1362 return !pid_alive(proc_task(dentry->d_inode));
1365 static struct dentry_operations tid_fd_dentry_operations =
1367 .d_revalidate = tid_fd_revalidate,
1368 .d_delete = pid_delete_dentry,
1371 static struct dentry_operations pid_dentry_operations =
1373 .d_revalidate = pid_revalidate,
1374 .d_delete = pid_delete_dentry,
1379 static unsigned name_to_int(struct dentry *dentry)
1381 const char *name = dentry->d_name.name;
1382 int len = dentry->d_name.len;
1385 if (len > 1 && *name == '0')
1388 unsigned c = *name++ - '0';
1391 if (n >= (~0U-9)/10)
1402 static struct dentry *proc_lookupfd(struct inode * dir, struct dentry * dentry, struct nameidata *nd)
1404 struct task_struct *task = proc_task(dir);
1405 unsigned fd = name_to_int(dentry);
1406 struct dentry *result = ERR_PTR(-ENOENT);
1408 struct files_struct * files;
1409 struct inode *inode;
1410 struct proc_inode *ei;
1414 if (!pid_alive(task))
1417 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TID_FD_DIR+fd);
1422 files = get_files_struct(task);
1425 inode->i_mode = S_IFLNK;
1428 * We are not taking a ref to the file structure, so we must
1431 spin_lock(&files->file_lock);
1432 file = fcheck_files(files, fd);
1435 if (file->f_mode & 1)
1436 inode->i_mode |= S_IRUSR | S_IXUSR;
1437 if (file->f_mode & 2)
1438 inode->i_mode |= S_IWUSR | S_IXUSR;
1439 spin_unlock(&files->file_lock);
1440 put_files_struct(files);
1441 inode->i_op = &proc_pid_link_inode_operations;
1443 ei->op.proc_get_link = proc_fd_link;
1444 dentry->d_op = &tid_fd_dentry_operations;
1445 d_add(dentry, inode);
1446 /* Close the race of the process dying before we return the dentry */
1447 if (tid_fd_revalidate(dentry, NULL))
1453 spin_unlock(&files->file_lock);
1454 put_files_struct(files);
1460 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir);
1461 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd);
1462 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat);
1464 static struct file_operations proc_fd_operations = {
1465 .read = generic_read_dir,
1466 .readdir = proc_readfd,
1469 static struct file_operations proc_task_operations = {
1470 .read = generic_read_dir,
1471 .readdir = proc_task_readdir,
1475 * proc directories can do almost nothing..
1477 static struct inode_operations proc_fd_inode_operations = {
1478 .lookup = proc_lookupfd,
1481 static struct inode_operations proc_task_inode_operations = {
1482 .lookup = proc_task_lookup,
1483 .getattr = proc_task_getattr,
1486 #ifdef CONFIG_SECURITY
1487 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
1488 size_t count, loff_t *ppos)
1490 struct inode * inode = file->f_dentry->d_inode;
1493 struct task_struct *task = proc_task(inode);
1495 if (count > PAGE_SIZE)
1497 if (!(page = __get_free_page(GFP_KERNEL)))
1500 length = security_getprocattr(task,
1501 (char*)file->f_dentry->d_name.name,
1502 (void*)page, count);
1504 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
1509 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
1510 size_t count, loff_t *ppos)
1512 struct inode * inode = file->f_dentry->d_inode;
1515 struct task_struct *task = proc_task(inode);
1517 if (count > PAGE_SIZE)
1520 /* No partial writes. */
1523 page = (char*)__get_free_page(GFP_USER);
1527 if (copy_from_user(page, buf, count))
1530 length = security_setprocattr(task,
1531 (char*)file->f_dentry->d_name.name,
1532 (void*)page, count);
1534 free_page((unsigned long) page);
1538 static struct file_operations proc_pid_attr_operations = {
1539 .read = proc_pid_attr_read,
1540 .write = proc_pid_attr_write,
1543 static struct file_operations proc_tid_attr_operations;
1544 static struct inode_operations proc_tid_attr_inode_operations;
1545 static struct file_operations proc_tgid_attr_operations;
1546 static struct inode_operations proc_tgid_attr_inode_operations;
1550 static struct dentry *proc_pident_lookup(struct inode *dir,
1551 struct dentry *dentry,
1552 struct pid_entry *ents)
1554 struct inode *inode;
1555 struct dentry *error;
1556 struct task_struct *task = proc_task(dir);
1557 struct pid_entry *p;
1558 struct proc_inode *ei;
1560 error = ERR_PTR(-ENOENT);
1563 if (!pid_alive(task))
1566 for (p = ents; p->name; p++) {
1567 if (p->len != dentry->d_name.len)
1569 if (!memcmp(dentry->d_name.name, p->name, p->len))
1575 error = ERR_PTR(-EINVAL);
1576 inode = proc_pid_make_inode(dir->i_sb, task, p->type);
1581 inode->i_mode = p->mode;
1583 * Yes, it does not scale. And it should not. Don't add
1584 * new entries into /proc/<tgid>/ without very good reasons.
1587 case PROC_TGID_TASK:
1589 inode->i_op = &proc_task_inode_operations;
1590 inode->i_fop = &proc_task_operations;
1595 inode->i_op = &proc_fd_inode_operations;
1596 inode->i_fop = &proc_fd_operations;
1600 inode->i_op = &proc_pid_link_inode_operations;
1601 ei->op.proc_get_link = proc_exe_link;
1605 inode->i_op = &proc_pid_link_inode_operations;
1606 ei->op.proc_get_link = proc_cwd_link;
1609 case PROC_TGID_ROOT:
1610 inode->i_op = &proc_pid_link_inode_operations;
1611 ei->op.proc_get_link = proc_root_link;
1613 case PROC_TID_ENVIRON:
1614 case PROC_TGID_ENVIRON:
1615 inode->i_fop = &proc_info_file_operations;
1616 ei->op.proc_read = proc_pid_environ;
1619 case PROC_TGID_AUXV:
1620 inode->i_fop = &proc_info_file_operations;
1621 ei->op.proc_read = proc_pid_auxv;
1623 case PROC_TID_STATUS:
1624 case PROC_TGID_STATUS:
1625 inode->i_fop = &proc_info_file_operations;
1626 ei->op.proc_read = proc_pid_status;
1629 inode->i_fop = &proc_info_file_operations;
1630 ei->op.proc_read = proc_tid_stat;
1632 case PROC_TGID_STAT:
1633 inode->i_fop = &proc_info_file_operations;
1634 ei->op.proc_read = proc_tgid_stat;
1636 case PROC_TID_CMDLINE:
1637 case PROC_TGID_CMDLINE:
1638 inode->i_fop = &proc_info_file_operations;
1639 ei->op.proc_read = proc_pid_cmdline;
1641 case PROC_TID_STATM:
1642 case PROC_TGID_STATM:
1643 inode->i_fop = &proc_info_file_operations;
1644 ei->op.proc_read = proc_pid_statm;
1647 case PROC_TGID_MAPS:
1648 inode->i_fop = &proc_maps_operations;
1651 case PROC_TID_NUMA_MAPS:
1652 case PROC_TGID_NUMA_MAPS:
1653 inode->i_fop = &proc_numa_maps_operations;
1658 inode->i_fop = &proc_mem_operations;
1660 #ifdef CONFIG_SECCOMP
1661 case PROC_TID_SECCOMP:
1662 case PROC_TGID_SECCOMP:
1663 inode->i_fop = &proc_seccomp_operations;
1665 #endif /* CONFIG_SECCOMP */
1666 case PROC_TID_MOUNTS:
1667 case PROC_TGID_MOUNTS:
1668 inode->i_fop = &proc_mounts_operations;
1671 case PROC_TID_SMAPS:
1672 case PROC_TGID_SMAPS:
1673 inode->i_fop = &proc_smaps_operations;
1676 case PROC_TID_MOUNTSTATS:
1677 case PROC_TGID_MOUNTSTATS:
1678 inode->i_fop = &proc_mountstats_operations;
1680 #ifdef CONFIG_SECURITY
1683 inode->i_op = &proc_tid_attr_inode_operations;
1684 inode->i_fop = &proc_tid_attr_operations;
1686 case PROC_TGID_ATTR:
1688 inode->i_op = &proc_tgid_attr_inode_operations;
1689 inode->i_fop = &proc_tgid_attr_operations;
1691 case PROC_TID_ATTR_CURRENT:
1692 case PROC_TGID_ATTR_CURRENT:
1693 case PROC_TID_ATTR_PREV:
1694 case PROC_TGID_ATTR_PREV:
1695 case PROC_TID_ATTR_EXEC:
1696 case PROC_TGID_ATTR_EXEC:
1697 case PROC_TID_ATTR_FSCREATE:
1698 case PROC_TGID_ATTR_FSCREATE:
1699 case PROC_TID_ATTR_KEYCREATE:
1700 case PROC_TGID_ATTR_KEYCREATE:
1701 inode->i_fop = &proc_pid_attr_operations;
1704 #ifdef CONFIG_KALLSYMS
1705 case PROC_TID_WCHAN:
1706 case PROC_TGID_WCHAN:
1707 inode->i_fop = &proc_info_file_operations;
1708 ei->op.proc_read = proc_pid_wchan;
1711 #ifdef CONFIG_SCHEDSTATS
1712 case PROC_TID_SCHEDSTAT:
1713 case PROC_TGID_SCHEDSTAT:
1714 inode->i_fop = &proc_info_file_operations;
1715 ei->op.proc_read = proc_pid_schedstat;
1718 #ifdef CONFIG_CPUSETS
1719 case PROC_TID_CPUSET:
1720 case PROC_TGID_CPUSET:
1721 inode->i_fop = &proc_cpuset_operations;
1724 case PROC_TID_OOM_SCORE:
1725 case PROC_TGID_OOM_SCORE:
1726 inode->i_fop = &proc_info_file_operations;
1727 ei->op.proc_read = proc_oom_score;
1729 case PROC_TID_OOM_ADJUST:
1730 case PROC_TGID_OOM_ADJUST:
1731 inode->i_fop = &proc_oom_adjust_operations;
1733 #ifdef CONFIG_AUDITSYSCALL
1734 case PROC_TID_LOGINUID:
1735 case PROC_TGID_LOGINUID:
1736 inode->i_fop = &proc_loginuid_operations;
1740 printk("procfs: impossible type (%d)",p->type);
1742 error = ERR_PTR(-EINVAL);
1745 dentry->d_op = &pid_dentry_operations;
1746 d_add(dentry, inode);
1747 /* Close the race of the process dying before we return the dentry */
1748 if (pid_revalidate(dentry, NULL))
1754 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
1755 return proc_pident_lookup(dir, dentry, tgid_base_stuff);
1758 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
1759 return proc_pident_lookup(dir, dentry, tid_base_stuff);
1762 static struct file_operations proc_tgid_base_operations = {
1763 .read = generic_read_dir,
1764 .readdir = proc_tgid_base_readdir,
1767 static struct file_operations proc_tid_base_operations = {
1768 .read = generic_read_dir,
1769 .readdir = proc_tid_base_readdir,
1772 static struct inode_operations proc_tgid_base_inode_operations = {
1773 .lookup = proc_tgid_base_lookup,
1776 static struct inode_operations proc_tid_base_inode_operations = {
1777 .lookup = proc_tid_base_lookup,
1780 #ifdef CONFIG_SECURITY
1781 static int proc_tgid_attr_readdir(struct file * filp,
1782 void * dirent, filldir_t filldir)
1784 return proc_pident_readdir(filp,dirent,filldir,
1785 tgid_attr_stuff,ARRAY_SIZE(tgid_attr_stuff));
1788 static int proc_tid_attr_readdir(struct file * filp,
1789 void * dirent, filldir_t filldir)
1791 return proc_pident_readdir(filp,dirent,filldir,
1792 tid_attr_stuff,ARRAY_SIZE(tid_attr_stuff));
1795 static struct file_operations proc_tgid_attr_operations = {
1796 .read = generic_read_dir,
1797 .readdir = proc_tgid_attr_readdir,
1800 static struct file_operations proc_tid_attr_operations = {
1801 .read = generic_read_dir,
1802 .readdir = proc_tid_attr_readdir,
1805 static struct dentry *proc_tgid_attr_lookup(struct inode *dir,
1806 struct dentry *dentry, struct nameidata *nd)
1808 return proc_pident_lookup(dir, dentry, tgid_attr_stuff);
1811 static struct dentry *proc_tid_attr_lookup(struct inode *dir,
1812 struct dentry *dentry, struct nameidata *nd)
1814 return proc_pident_lookup(dir, dentry, tid_attr_stuff);
1817 static struct inode_operations proc_tgid_attr_inode_operations = {
1818 .lookup = proc_tgid_attr_lookup,
1821 static struct inode_operations proc_tid_attr_inode_operations = {
1822 .lookup = proc_tid_attr_lookup,
1829 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
1832 char tmp[PROC_NUMBUF];
1833 sprintf(tmp, "%d", current->tgid);
1834 return vfs_readlink(dentry,buffer,buflen,tmp);
1837 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
1839 char tmp[PROC_NUMBUF];
1840 sprintf(tmp, "%d", current->tgid);
1841 return ERR_PTR(vfs_follow_link(nd,tmp));
1844 static struct inode_operations proc_self_inode_operations = {
1845 .readlink = proc_self_readlink,
1846 .follow_link = proc_self_follow_link,
1850 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
1852 * @task: task that should be flushed.
1854 * Looks in the dcache for
1856 * /proc/@tgid/task/@pid
1857 * if either directory is present flushes it and all of it'ts children
1860 * It is safe and reasonable to cache /proc entries for a task until
1861 * that task exits. After that they just clog up the dcache with
1862 * useless entries, possibly causing useful dcache entries to be
1863 * flushed instead. This routine is proved to flush those useless
1864 * dcache entries at process exit time.
1866 * NOTE: This routine is just an optimization so it does not guarantee
1867 * that no dcache entries will exist at process exit time it
1868 * just makes it very unlikely that any will persist.
1870 void proc_flush_task(struct task_struct *task)
1872 struct dentry *dentry, *leader, *dir;
1873 char buf[PROC_NUMBUF];
1877 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
1878 dentry = d_hash_and_lookup(proc_mnt->mnt_root, &name);
1880 shrink_dcache_parent(dentry);
1885 if (thread_group_leader(task))
1889 name.len = snprintf(buf, sizeof(buf), "%d", task->tgid);
1890 leader = d_hash_and_lookup(proc_mnt->mnt_root, &name);
1895 name.len = strlen(name.name);
1896 dir = d_hash_and_lookup(leader, &name);
1898 goto out_put_leader;
1901 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
1902 dentry = d_hash_and_lookup(dir, &name);
1904 shrink_dcache_parent(dentry);
1917 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
1919 struct dentry *result = ERR_PTR(-ENOENT);
1920 struct task_struct *task;
1921 struct inode *inode;
1922 struct proc_inode *ei;
1925 if (dentry->d_name.len == 4 && !memcmp(dentry->d_name.name,"self",4)) {
1926 inode = new_inode(dir->i_sb);
1928 return ERR_PTR(-ENOMEM);
1930 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1931 inode->i_ino = fake_ino(0, PROC_TGID_INO);
1933 inode->i_mode = S_IFLNK|S_IRWXUGO;
1934 inode->i_uid = inode->i_gid = 0;
1936 inode->i_op = &proc_self_inode_operations;
1937 d_add(dentry, inode);
1940 tgid = name_to_int(dentry);
1945 task = find_task_by_pid(tgid);
1947 get_task_struct(task);
1952 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TGID_INO);
1956 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
1957 inode->i_op = &proc_tgid_base_inode_operations;
1958 inode->i_fop = &proc_tgid_base_operations;
1959 inode->i_flags|=S_IMMUTABLE;
1960 #ifdef CONFIG_SECURITY
1966 dentry->d_op = &pid_dentry_operations;
1968 d_add(dentry, inode);
1969 /* Close the race of the process dying before we return the dentry */
1970 if (pid_revalidate(dentry, NULL))
1974 put_task_struct(task);
1980 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
1982 struct dentry *result = ERR_PTR(-ENOENT);
1983 struct task_struct *task;
1984 struct task_struct *leader = proc_task(dir);
1985 struct inode *inode;
1988 tid = name_to_int(dentry);
1993 task = find_task_by_pid(tid);
1995 get_task_struct(task);
1999 if (leader->tgid != task->tgid)
2002 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TID_INO);
2007 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2008 inode->i_op = &proc_tid_base_inode_operations;
2009 inode->i_fop = &proc_tid_base_operations;
2010 inode->i_flags|=S_IMMUTABLE;
2011 #ifdef CONFIG_SECURITY
2017 dentry->d_op = &pid_dentry_operations;
2019 d_add(dentry, inode);
2020 /* Close the race of the process dying before we return the dentry */
2021 if (pid_revalidate(dentry, NULL))
2025 put_task_struct(task);
2031 * Find the first tgid to return to user space.
2033 * Usually this is just whatever follows &init_task, but if the users
2034 * buffer was too small to hold the full list or there was a seek into
2035 * the middle of the directory we have more work to do.
2037 * In the case of a short read we start with find_task_by_pid.
2039 * In the case of a seek we start with &init_task and walk nr
2042 static struct task_struct *first_tgid(int tgid, unsigned int nr)
2044 struct task_struct *pos;
2047 pos = find_task_by_pid(tgid);
2048 if (pos && thread_group_leader(pos))
2051 /* If nr exceeds the number of processes get out quickly */
2053 if (nr && nr >= nr_processes())
2056 /* If we haven't found our starting place yet start with
2057 * the init_task and walk nr tasks forward.
2059 for (pos = next_task(&init_task); nr > 0; --nr) {
2060 pos = next_task(pos);
2061 if (pos == &init_task) {
2067 get_task_struct(pos);
2074 * Find the next task in the task list.
2075 * Return NULL if we loop or there is any error.
2077 * The reference to the input task_struct is released.
2079 static struct task_struct *next_tgid(struct task_struct *start)
2081 struct task_struct *pos;
2084 if (pid_alive(start))
2085 pos = next_task(start);
2086 if (pid_alive(pos) && (pos != &init_task)) {
2087 get_task_struct(pos);
2093 put_task_struct(start);
2097 /* for the /proc/ directory itself, after non-process stuff has been done */
2098 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2100 char buf[PROC_NUMBUF];
2101 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2102 struct task_struct *task;
2106 ino_t ino = fake_ino(0,PROC_TGID_INO);
2107 if (filldir(dirent, "self", 4, filp->f_pos, ino, DT_LNK) < 0)
2114 /* f_version caches the tgid value that the last readdir call couldn't
2115 * return. lseek aka telldir automagically resets f_version to 0.
2117 tgid = filp->f_version;
2118 filp->f_version = 0;
2119 for (task = first_tgid(tgid, nr);
2121 task = next_tgid(task), filp->f_pos++) {
2125 len = snprintf(buf, sizeof(buf), "%d", tgid);
2126 ino = fake_ino(tgid, PROC_TGID_INO);
2127 if (filldir(dirent, buf, len, filp->f_pos, ino, DT_DIR) < 0) {
2128 /* returning this tgid failed, save it as the first
2129 * pid for the next readir call */
2130 filp->f_version = tgid;
2131 put_task_struct(task);
2139 * Find the first tid of a thread group to return to user space.
2141 * Usually this is just the thread group leader, but if the users
2142 * buffer was too small or there was a seek into the middle of the
2143 * directory we have more work todo.
2145 * In the case of a short read we start with find_task_by_pid.
2147 * In the case of a seek we start with the leader and walk nr
2150 static struct task_struct *first_tid(struct task_struct *leader, int tid, int nr)
2152 struct task_struct *pos = NULL;
2153 read_lock(&tasklist_lock);
2155 /* Attempt to start with the pid of a thread */
2156 if (tid && (nr > 0)) {
2157 pos = find_task_by_pid(tid);
2158 if (pos && (pos->group_leader != leader))
2164 /* If nr exceeds the number of threads there is nothing todo */
2169 threads = atomic_read(&leader->signal->count);
2170 task_unlock(leader);
2175 /* If we haven't found our starting place yet start with the
2176 * leader and walk nr threads forward.
2178 if (!pos && (nr >= 0))
2181 for (; pos && pid_alive(pos); pos = next_thread(pos)) {
2184 get_task_struct(pos);
2189 read_unlock(&tasklist_lock);
2194 * Find the next thread in the thread list.
2195 * Return NULL if there is an error or no next thread.
2197 * The reference to the input task_struct is released.
2199 static struct task_struct *next_tid(struct task_struct *start)
2201 struct task_struct *pos;
2202 read_lock(&tasklist_lock);
2204 if (pid_alive(start))
2205 pos = next_thread(start);
2206 if (pid_alive(pos) && (pos != start->group_leader))
2207 get_task_struct(pos);
2210 read_unlock(&tasklist_lock);
2211 put_task_struct(start);
2215 /* for the /proc/TGID/task/ directories */
2216 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
2218 char buf[PROC_NUMBUF];
2219 struct dentry *dentry = filp->f_dentry;
2220 struct inode *inode = dentry->d_inode;
2221 struct task_struct *leader = proc_task(inode);
2222 struct task_struct *task;
2223 int retval = -ENOENT;
2226 unsigned long pos = filp->f_pos; /* avoiding "long long" filp->f_pos */
2228 if (!pid_alive(leader))
2235 if (filldir(dirent, ".", 1, pos, ino, DT_DIR) < 0)
2240 ino = parent_ino(dentry);
2241 if (filldir(dirent, "..", 2, pos, ino, DT_DIR) < 0)
2247 /* f_version caches the tgid value that the last readdir call couldn't
2248 * return. lseek aka telldir automagically resets f_version to 0.
2250 tid = filp->f_version;
2251 filp->f_version = 0;
2252 for (task = first_tid(leader, tid, pos - 2);
2254 task = next_tid(task), pos++) {
2257 len = snprintf(buf, sizeof(buf), "%d", tid);
2258 ino = fake_ino(tid, PROC_TID_INO);
2259 if (filldir(dirent, buf, len, pos, ino, DT_DIR < 0)) {
2260 /* returning this tgid failed, save it as the first
2261 * pid for the next readir call */
2262 filp->f_version = tid;
2263 put_task_struct(task);
2272 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
2274 struct inode *inode = dentry->d_inode;
2275 struct task_struct *p = proc_task(inode);
2276 generic_fillattr(inode, stat);
2281 stat->nlink += atomic_read(&p->signal->count);