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 * For hysterical raisins we keep the same inumbers as in the old procfs.
79 * Feel free to change the macro below - just keep the range distinct from
80 * inumbers of the rest of procfs (currently those are in 0x0000--0xffff).
81 * As soon as we'll get a separate superblock we will be able to forget
82 * about magical ranges too.
85 #define fake_ino(pid,ino) (((pid)<<16)|(ino))
87 enum pid_directory_inos {
107 PROC_TGID_MOUNTSTATS,
112 #ifdef CONFIG_SCHEDSTATS
115 #ifdef CONFIG_CPUSETS
118 #ifdef CONFIG_SECURITY
120 PROC_TGID_ATTR_CURRENT,
123 PROC_TGID_ATTR_FSCREATE,
124 PROC_TGID_ATTR_KEYCREATE,
126 #ifdef CONFIG_AUDITSYSCALL
130 PROC_TGID_OOM_ADJUST,
134 #ifdef CONFIG_SECCOMP
154 #ifdef CONFIG_SCHEDSTATS
157 #ifdef CONFIG_CPUSETS
160 #ifdef CONFIG_SECURITY
162 PROC_TID_ATTR_CURRENT,
165 PROC_TID_ATTR_FSCREATE,
166 PROC_TID_ATTR_KEYCREATE,
168 #ifdef CONFIG_AUDITSYSCALL
174 /* Add new entries before this */
175 PROC_TID_FD_DIR = 0x8000, /* 0x8000-0xffff */
185 #define E(type,name,mode) {(type),sizeof(name)-1,(name),(mode)}
187 static struct pid_entry tgid_base_stuff[] = {
188 E(PROC_TGID_TASK, "task", S_IFDIR|S_IRUGO|S_IXUGO),
189 E(PROC_TGID_FD, "fd", S_IFDIR|S_IRUSR|S_IXUSR),
190 E(PROC_TGID_ENVIRON, "environ", S_IFREG|S_IRUSR),
191 E(PROC_TGID_AUXV, "auxv", S_IFREG|S_IRUSR),
192 E(PROC_TGID_STATUS, "status", S_IFREG|S_IRUGO),
193 E(PROC_TGID_CMDLINE, "cmdline", S_IFREG|S_IRUGO),
194 E(PROC_TGID_STAT, "stat", S_IFREG|S_IRUGO),
195 E(PROC_TGID_STATM, "statm", S_IFREG|S_IRUGO),
196 E(PROC_TGID_MAPS, "maps", S_IFREG|S_IRUGO),
198 E(PROC_TGID_NUMA_MAPS, "numa_maps", S_IFREG|S_IRUGO),
200 E(PROC_TGID_MEM, "mem", S_IFREG|S_IRUSR|S_IWUSR),
201 #ifdef CONFIG_SECCOMP
202 E(PROC_TGID_SECCOMP, "seccomp", S_IFREG|S_IRUSR|S_IWUSR),
204 E(PROC_TGID_CWD, "cwd", S_IFLNK|S_IRWXUGO),
205 E(PROC_TGID_ROOT, "root", S_IFLNK|S_IRWXUGO),
206 E(PROC_TGID_EXE, "exe", S_IFLNK|S_IRWXUGO),
207 E(PROC_TGID_MOUNTS, "mounts", S_IFREG|S_IRUGO),
208 E(PROC_TGID_MOUNTSTATS, "mountstats", S_IFREG|S_IRUSR),
210 E(PROC_TGID_SMAPS, "smaps", S_IFREG|S_IRUGO),
212 #ifdef CONFIG_SECURITY
213 E(PROC_TGID_ATTR, "attr", S_IFDIR|S_IRUGO|S_IXUGO),
215 #ifdef CONFIG_KALLSYMS
216 E(PROC_TGID_WCHAN, "wchan", S_IFREG|S_IRUGO),
218 #ifdef CONFIG_SCHEDSTATS
219 E(PROC_TGID_SCHEDSTAT, "schedstat", S_IFREG|S_IRUGO),
221 #ifdef CONFIG_CPUSETS
222 E(PROC_TGID_CPUSET, "cpuset", S_IFREG|S_IRUGO),
224 E(PROC_TGID_OOM_SCORE, "oom_score",S_IFREG|S_IRUGO),
225 E(PROC_TGID_OOM_ADJUST,"oom_adj", S_IFREG|S_IRUGO|S_IWUSR),
226 #ifdef CONFIG_AUDITSYSCALL
227 E(PROC_TGID_LOGINUID, "loginuid", S_IFREG|S_IWUSR|S_IRUGO),
231 static struct pid_entry tid_base_stuff[] = {
232 E(PROC_TID_FD, "fd", S_IFDIR|S_IRUSR|S_IXUSR),
233 E(PROC_TID_ENVIRON, "environ", S_IFREG|S_IRUSR),
234 E(PROC_TID_AUXV, "auxv", S_IFREG|S_IRUSR),
235 E(PROC_TID_STATUS, "status", S_IFREG|S_IRUGO),
236 E(PROC_TID_CMDLINE, "cmdline", S_IFREG|S_IRUGO),
237 E(PROC_TID_STAT, "stat", S_IFREG|S_IRUGO),
238 E(PROC_TID_STATM, "statm", S_IFREG|S_IRUGO),
239 E(PROC_TID_MAPS, "maps", S_IFREG|S_IRUGO),
241 E(PROC_TID_NUMA_MAPS, "numa_maps", S_IFREG|S_IRUGO),
243 E(PROC_TID_MEM, "mem", S_IFREG|S_IRUSR|S_IWUSR),
244 #ifdef CONFIG_SECCOMP
245 E(PROC_TID_SECCOMP, "seccomp", S_IFREG|S_IRUSR|S_IWUSR),
247 E(PROC_TID_CWD, "cwd", S_IFLNK|S_IRWXUGO),
248 E(PROC_TID_ROOT, "root", S_IFLNK|S_IRWXUGO),
249 E(PROC_TID_EXE, "exe", S_IFLNK|S_IRWXUGO),
250 E(PROC_TID_MOUNTS, "mounts", S_IFREG|S_IRUGO),
252 E(PROC_TID_SMAPS, "smaps", S_IFREG|S_IRUGO),
254 #ifdef CONFIG_SECURITY
255 E(PROC_TID_ATTR, "attr", S_IFDIR|S_IRUGO|S_IXUGO),
257 #ifdef CONFIG_KALLSYMS
258 E(PROC_TID_WCHAN, "wchan", S_IFREG|S_IRUGO),
260 #ifdef CONFIG_SCHEDSTATS
261 E(PROC_TID_SCHEDSTAT, "schedstat",S_IFREG|S_IRUGO),
263 #ifdef CONFIG_CPUSETS
264 E(PROC_TID_CPUSET, "cpuset", S_IFREG|S_IRUGO),
266 E(PROC_TID_OOM_SCORE, "oom_score",S_IFREG|S_IRUGO),
267 E(PROC_TID_OOM_ADJUST, "oom_adj", S_IFREG|S_IRUGO|S_IWUSR),
268 #ifdef CONFIG_AUDITSYSCALL
269 E(PROC_TID_LOGINUID, "loginuid", S_IFREG|S_IWUSR|S_IRUGO),
274 #ifdef CONFIG_SECURITY
275 static struct pid_entry tgid_attr_stuff[] = {
276 E(PROC_TGID_ATTR_CURRENT, "current", S_IFREG|S_IRUGO|S_IWUGO),
277 E(PROC_TGID_ATTR_PREV, "prev", S_IFREG|S_IRUGO),
278 E(PROC_TGID_ATTR_EXEC, "exec", S_IFREG|S_IRUGO|S_IWUGO),
279 E(PROC_TGID_ATTR_FSCREATE, "fscreate", S_IFREG|S_IRUGO|S_IWUGO),
280 E(PROC_TGID_ATTR_KEYCREATE, "keycreate", S_IFREG|S_IRUGO|S_IWUGO),
283 static struct pid_entry tid_attr_stuff[] = {
284 E(PROC_TID_ATTR_CURRENT, "current", S_IFREG|S_IRUGO|S_IWUGO),
285 E(PROC_TID_ATTR_PREV, "prev", S_IFREG|S_IRUGO),
286 E(PROC_TID_ATTR_EXEC, "exec", S_IFREG|S_IRUGO|S_IWUGO),
287 E(PROC_TID_ATTR_FSCREATE, "fscreate", S_IFREG|S_IRUGO|S_IWUGO),
288 E(PROC_TID_ATTR_KEYCREATE, "keycreate", S_IFREG|S_IRUGO|S_IWUGO),
295 static int proc_fd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
297 struct task_struct *task = proc_task(inode);
298 struct files_struct *files;
300 int fd = proc_type(inode) - PROC_TID_FD_DIR;
302 files = get_files_struct(task);
305 * We are not taking a ref to the file structure, so we must
308 spin_lock(&files->file_lock);
309 file = fcheck_files(files, fd);
311 *mnt = mntget(file->f_vfsmnt);
312 *dentry = dget(file->f_dentry);
313 spin_unlock(&files->file_lock);
314 put_files_struct(files);
317 spin_unlock(&files->file_lock);
318 put_files_struct(files);
323 static struct fs_struct *get_fs_struct(struct task_struct *task)
325 struct fs_struct *fs;
329 atomic_inc(&fs->count);
334 static int proc_cwd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
336 struct fs_struct *fs = get_fs_struct(proc_task(inode));
337 int result = -ENOENT;
339 read_lock(&fs->lock);
340 *mnt = mntget(fs->pwdmnt);
341 *dentry = dget(fs->pwd);
342 read_unlock(&fs->lock);
349 static int proc_root_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
351 struct fs_struct *fs = get_fs_struct(proc_task(inode));
352 int result = -ENOENT;
354 read_lock(&fs->lock);
355 *mnt = mntget(fs->rootmnt);
356 *dentry = dget(fs->root);
357 read_unlock(&fs->lock);
365 /* Same as proc_root_link, but this addionally tries to get fs from other
366 * threads in the group */
367 static int proc_task_root_link(struct inode *inode, struct dentry **dentry,
368 struct vfsmount **mnt)
370 struct fs_struct *fs;
371 int result = -ENOENT;
372 struct task_struct *leader = proc_task(inode);
377 atomic_inc(&fs->count);
380 /* Try to get fs from other threads */
382 read_lock(&tasklist_lock);
383 if (pid_alive(leader)) {
384 struct task_struct *task = leader;
386 while ((task = next_thread(task)) != leader) {
390 atomic_inc(&fs->count);
397 read_unlock(&tasklist_lock);
401 read_lock(&fs->lock);
402 *mnt = mntget(fs->rootmnt);
403 *dentry = dget(fs->root);
404 read_unlock(&fs->lock);
412 #define MAY_PTRACE(task) \
413 (task == current || \
414 (task->parent == current && \
415 (task->ptrace & PT_PTRACED) && \
416 (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
417 security_ptrace(current,task) == 0))
419 static int proc_pid_environ(struct task_struct *task, char * buffer)
422 struct mm_struct *mm = get_task_mm(task);
424 unsigned int len = mm->env_end - mm->env_start;
427 res = access_process_vm(task, mm->env_start, buffer, len, 0);
428 if (!ptrace_may_attach(task))
435 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
439 struct mm_struct *mm = get_task_mm(task);
443 goto out_mm; /* Shh! No looking before we're done */
445 len = mm->arg_end - mm->arg_start;
450 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
452 // If the nul at the end of args has been overwritten, then
453 // assume application is using setproctitle(3).
454 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
455 len = strnlen(buffer, res);
459 len = mm->env_end - mm->env_start;
460 if (len > PAGE_SIZE - res)
461 len = PAGE_SIZE - res;
462 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
463 res = strnlen(buffer, res);
472 static int proc_pid_auxv(struct task_struct *task, char *buffer)
475 struct mm_struct *mm = get_task_mm(task);
477 unsigned int nwords = 0;
480 while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
481 res = nwords * sizeof(mm->saved_auxv[0]);
484 memcpy(buffer, mm->saved_auxv, res);
491 #ifdef CONFIG_KALLSYMS
493 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
494 * Returns the resolved symbol. If that fails, simply return the address.
496 static int proc_pid_wchan(struct task_struct *task, char *buffer)
499 const char *sym_name;
500 unsigned long wchan, size, offset;
501 char namebuf[KSYM_NAME_LEN+1];
503 wchan = get_wchan(task);
505 sym_name = kallsyms_lookup(wchan, &size, &offset, &modname, namebuf);
507 return sprintf(buffer, "%s", sym_name);
508 return sprintf(buffer, "%lu", wchan);
510 #endif /* CONFIG_KALLSYMS */
512 #ifdef CONFIG_SCHEDSTATS
514 * Provides /proc/PID/schedstat
516 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
518 return sprintf(buffer, "%lu %lu %lu\n",
519 task->sched_info.cpu_time,
520 task->sched_info.run_delay,
521 task->sched_info.pcnt);
525 /* The badness from the OOM killer */
526 unsigned long badness(struct task_struct *p, unsigned long uptime);
527 static int proc_oom_score(struct task_struct *task, char *buffer)
529 unsigned long points;
530 struct timespec uptime;
532 do_posix_clock_monotonic_gettime(&uptime);
533 points = badness(task, uptime.tv_sec);
534 return sprintf(buffer, "%lu\n", points);
537 /************************************************************************/
538 /* Here the fs part begins */
539 /************************************************************************/
541 /* permission checks */
543 /* If the process being read is separated by chroot from the reading process,
544 * don't let the reader access the threads.
546 * note: this does dput(root) and mntput(vfsmnt) on exit.
548 static int proc_check_chroot(struct dentry *root, struct vfsmount *vfsmnt)
550 struct dentry *de, *base;
551 struct vfsmount *our_vfsmnt, *mnt;
554 read_lock(¤t->fs->lock);
555 our_vfsmnt = mntget(current->fs->rootmnt);
556 base = dget(current->fs->root);
557 read_unlock(¤t->fs->lock);
559 spin_lock(&vfsmount_lock);
563 while (mnt != our_vfsmnt) {
564 if (mnt == mnt->mnt_parent)
566 de = mnt->mnt_mountpoint;
567 mnt = mnt->mnt_parent;
570 if (!is_subdir(de, base))
572 spin_unlock(&vfsmount_lock);
581 spin_unlock(&vfsmount_lock);
586 static int proc_check_root(struct inode *inode)
589 struct vfsmount *vfsmnt;
591 if (proc_root_link(inode, &root, &vfsmnt)) /* Ewww... */
593 return proc_check_chroot(root, vfsmnt);
596 static int proc_permission(struct inode *inode, int mask, struct nameidata *nd)
598 if (generic_permission(inode, mask, NULL) != 0)
600 return proc_check_root(inode);
603 static int proc_task_permission(struct inode *inode, int mask, struct nameidata *nd)
606 struct vfsmount *vfsmnt;
608 if (generic_permission(inode, mask, NULL) != 0)
611 if (proc_task_root_link(inode, &root, &vfsmnt))
614 return proc_check_chroot(root, vfsmnt);
617 extern struct seq_operations proc_pid_maps_op;
618 static int maps_open(struct inode *inode, struct file *file)
620 struct task_struct *task = proc_task(inode);
621 int ret = seq_open(file, &proc_pid_maps_op);
623 struct seq_file *m = file->private_data;
629 static struct file_operations proc_maps_operations = {
633 .release = seq_release,
637 extern struct seq_operations proc_pid_numa_maps_op;
638 static int numa_maps_open(struct inode *inode, struct file *file)
640 struct task_struct *task = proc_task(inode);
641 int ret = seq_open(file, &proc_pid_numa_maps_op);
643 struct seq_file *m = file->private_data;
649 static struct file_operations proc_numa_maps_operations = {
650 .open = numa_maps_open,
653 .release = seq_release,
658 extern struct seq_operations proc_pid_smaps_op;
659 static int smaps_open(struct inode *inode, struct file *file)
661 struct task_struct *task = proc_task(inode);
662 int ret = seq_open(file, &proc_pid_smaps_op);
664 struct seq_file *m = file->private_data;
670 static struct file_operations proc_smaps_operations = {
674 .release = seq_release,
678 extern struct seq_operations mounts_op;
684 static int mounts_open(struct inode *inode, struct file *file)
686 struct task_struct *task = proc_task(inode);
687 struct namespace *namespace;
688 struct proc_mounts *p;
692 namespace = task->namespace;
694 get_namespace(namespace);
699 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
701 file->private_data = &p->m;
702 ret = seq_open(file, &mounts_op);
704 p->m.private = namespace;
705 p->event = namespace->event;
710 put_namespace(namespace);
715 static int mounts_release(struct inode *inode, struct file *file)
717 struct seq_file *m = file->private_data;
718 struct namespace *namespace = m->private;
719 put_namespace(namespace);
720 return seq_release(inode, file);
723 static unsigned mounts_poll(struct file *file, poll_table *wait)
725 struct proc_mounts *p = file->private_data;
726 struct namespace *ns = p->m.private;
729 poll_wait(file, &ns->poll, wait);
731 spin_lock(&vfsmount_lock);
732 if (p->event != ns->event) {
733 p->event = ns->event;
736 spin_unlock(&vfsmount_lock);
741 static struct file_operations proc_mounts_operations = {
745 .release = mounts_release,
749 extern struct seq_operations mountstats_op;
750 static int mountstats_open(struct inode *inode, struct file *file)
752 struct task_struct *task = proc_task(inode);
753 int ret = seq_open(file, &mountstats_op);
756 struct seq_file *m = file->private_data;
757 struct namespace *namespace;
759 namespace = task->namespace;
761 get_namespace(namespace);
765 m->private = namespace;
767 seq_release(inode, file);
774 static struct file_operations proc_mountstats_operations = {
775 .open = mountstats_open,
778 .release = mounts_release,
781 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
783 static ssize_t proc_info_read(struct file * file, char __user * buf,
784 size_t count, loff_t *ppos)
786 struct inode * inode = file->f_dentry->d_inode;
789 struct task_struct *task = proc_task(inode);
791 if (count > PROC_BLOCK_SIZE)
792 count = PROC_BLOCK_SIZE;
793 if (!(page = __get_free_page(GFP_KERNEL)))
796 length = PROC_I(inode)->op.proc_read(task, (char*)page);
799 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
804 static struct file_operations proc_info_file_operations = {
805 .read = proc_info_read,
808 static int mem_open(struct inode* inode, struct file* file)
810 file->private_data = (void*)((long)current->self_exec_id);
814 static ssize_t mem_read(struct file * file, char __user * buf,
815 size_t count, loff_t *ppos)
817 struct task_struct *task = proc_task(file->f_dentry->d_inode);
819 unsigned long src = *ppos;
821 struct mm_struct *mm;
823 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
827 page = (char *)__get_free_page(GFP_USER);
833 mm = get_task_mm(task);
839 if (file->private_data != (void*)((long)current->self_exec_id))
845 int this_len, retval;
847 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
848 retval = access_process_vm(task, src, page, this_len, 0);
849 if (!retval || !MAY_PTRACE(task) || !ptrace_may_attach(task)) {
855 if (copy_to_user(buf, page, retval)) {
870 free_page((unsigned long) page);
875 #define mem_write NULL
878 /* This is a security hazard */
879 static ssize_t mem_write(struct file * file, const char * buf,
880 size_t count, loff_t *ppos)
884 struct task_struct *task = proc_task(file->f_dentry->d_inode);
885 unsigned long dst = *ppos;
887 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
890 page = (char *)__get_free_page(GFP_USER);
895 int this_len, retval;
897 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
898 if (copy_from_user(page, buf, this_len)) {
902 retval = access_process_vm(task, dst, page, this_len, 1);
914 free_page((unsigned long) page);
919 static loff_t mem_lseek(struct file * file, loff_t offset, int orig)
923 file->f_pos = offset;
926 file->f_pos += offset;
931 force_successful_syscall_return();
935 static struct file_operations proc_mem_operations = {
942 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
943 size_t count, loff_t *ppos)
945 struct task_struct *task = proc_task(file->f_dentry->d_inode);
948 int oom_adjust = task->oomkilladj;
949 loff_t __ppos = *ppos;
951 len = sprintf(buffer, "%i\n", oom_adjust);
954 if (count > len-__ppos)
956 if (copy_to_user(buf, buffer + __ppos, count))
958 *ppos = __ppos + count;
962 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
963 size_t count, loff_t *ppos)
965 struct task_struct *task = proc_task(file->f_dentry->d_inode);
966 char buffer[8], *end;
969 if (!capable(CAP_SYS_RESOURCE))
971 memset(buffer, 0, 8);
974 if (copy_from_user(buffer, buf, count))
976 oom_adjust = simple_strtol(buffer, &end, 0);
977 if ((oom_adjust < -16 || oom_adjust > 15) && oom_adjust != OOM_DISABLE)
981 task->oomkilladj = oom_adjust;
982 if (end - buffer == 0)
987 static struct file_operations proc_oom_adjust_operations = {
988 .read = oom_adjust_read,
989 .write = oom_adjust_write,
992 static struct inode_operations proc_mem_inode_operations = {
993 .permission = proc_permission,
996 #ifdef CONFIG_AUDITSYSCALL
998 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
999 size_t count, loff_t *ppos)
1001 struct inode * inode = file->f_dentry->d_inode;
1002 struct task_struct *task = proc_task(inode);
1004 char tmpbuf[TMPBUFLEN];
1006 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1007 audit_get_loginuid(task->audit_context));
1008 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1011 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1012 size_t count, loff_t *ppos)
1014 struct inode * inode = file->f_dentry->d_inode;
1017 struct task_struct *task = proc_task(inode);
1020 if (!capable(CAP_AUDIT_CONTROL))
1023 if (current != task)
1026 if (count >= PAGE_SIZE)
1027 count = PAGE_SIZE - 1;
1030 /* No partial writes. */
1033 page = (char*)__get_free_page(GFP_USER);
1037 if (copy_from_user(page, buf, count))
1041 loginuid = simple_strtoul(page, &tmp, 10);
1047 length = audit_set_loginuid(task, loginuid);
1048 if (likely(length == 0))
1052 free_page((unsigned long) page);
1056 static struct file_operations proc_loginuid_operations = {
1057 .read = proc_loginuid_read,
1058 .write = proc_loginuid_write,
1062 #ifdef CONFIG_SECCOMP
1063 static ssize_t seccomp_read(struct file *file, char __user *buf,
1064 size_t count, loff_t *ppos)
1066 struct task_struct *tsk = proc_task(file->f_dentry->d_inode);
1068 loff_t __ppos = *ppos;
1071 /* no need to print the trailing zero, so use only len */
1072 len = sprintf(__buf, "%u\n", tsk->seccomp.mode);
1075 if (count > len - __ppos)
1076 count = len - __ppos;
1077 if (copy_to_user(buf, __buf + __ppos, count))
1079 *ppos = __ppos + count;
1083 static ssize_t seccomp_write(struct file *file, const char __user *buf,
1084 size_t count, loff_t *ppos)
1086 struct task_struct *tsk = proc_task(file->f_dentry->d_inode);
1087 char __buf[20], *end;
1088 unsigned int seccomp_mode;
1090 /* can set it only once to be even more secure */
1091 if (unlikely(tsk->seccomp.mode))
1094 memset(__buf, 0, sizeof(__buf));
1095 count = min(count, sizeof(__buf) - 1);
1096 if (copy_from_user(__buf, buf, count))
1098 seccomp_mode = simple_strtoul(__buf, &end, 0);
1101 if (seccomp_mode && seccomp_mode <= NR_SECCOMP_MODES) {
1102 tsk->seccomp.mode = seccomp_mode;
1103 set_tsk_thread_flag(tsk, TIF_SECCOMP);
1106 if (unlikely(!(end - __buf)))
1111 static struct file_operations proc_seccomp_operations = {
1112 .read = seccomp_read,
1113 .write = seccomp_write,
1115 #endif /* CONFIG_SECCOMP */
1117 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1119 struct inode *inode = dentry->d_inode;
1120 int error = -EACCES;
1122 /* We don't need a base pointer in the /proc filesystem */
1125 if (current->fsuid != inode->i_uid && !capable(CAP_DAC_OVERRIDE))
1127 error = proc_check_root(inode);
1131 error = PROC_I(inode)->op.proc_get_link(inode, &nd->dentry, &nd->mnt);
1132 nd->last_type = LAST_BIND;
1134 return ERR_PTR(error);
1137 static int do_proc_readlink(struct dentry *dentry, struct vfsmount *mnt,
1138 char __user *buffer, int buflen)
1140 struct inode * inode;
1141 char *tmp = (char*)__get_free_page(GFP_KERNEL), *path;
1147 inode = dentry->d_inode;
1148 path = d_path(dentry, mnt, tmp, PAGE_SIZE);
1149 len = PTR_ERR(path);
1152 len = tmp + PAGE_SIZE - 1 - path;
1156 if (copy_to_user(buffer, path, len))
1159 free_page((unsigned long)tmp);
1163 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1165 int error = -EACCES;
1166 struct inode *inode = dentry->d_inode;
1168 struct vfsmount *mnt = NULL;
1171 if (current->fsuid != inode->i_uid && !capable(CAP_DAC_OVERRIDE))
1173 error = proc_check_root(inode);
1177 error = PROC_I(inode)->op.proc_get_link(inode, &de, &mnt);
1181 error = do_proc_readlink(de, mnt, buffer, buflen);
1188 static struct inode_operations proc_pid_link_inode_operations = {
1189 .readlink = proc_pid_readlink,
1190 .follow_link = proc_pid_follow_link
1195 static int proc_readfd(struct file * filp, void * dirent, filldir_t filldir)
1197 struct dentry *dentry = filp->f_dentry;
1198 struct inode *inode = dentry->d_inode;
1199 struct task_struct *p = proc_task(inode);
1200 unsigned int fd, tid, ino;
1203 struct files_struct * files;
1204 struct fdtable *fdt;
1215 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1219 ino = parent_ino(dentry);
1220 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1224 files = get_files_struct(p);
1228 fdt = files_fdtable(files);
1229 for (fd = filp->f_pos-2;
1231 fd++, filp->f_pos++) {
1234 if (!fcheck_files(files, fd))
1242 buf[j] = '0' + (i % 10);
1246 ino = fake_ino(tid, PROC_TID_FD_DIR + fd);
1247 if (filldir(dirent, buf+j, NUMBUF-j, fd+2, ino, DT_LNK) < 0) {
1254 put_files_struct(files);
1260 static int proc_pident_readdir(struct file *filp,
1261 void *dirent, filldir_t filldir,
1262 struct pid_entry *ents, unsigned int nents)
1266 struct dentry *dentry = filp->f_dentry;
1267 struct inode *inode = dentry->d_inode;
1268 struct pid_entry *p;
1273 if (!pid_alive(proc_task(inode)))
1277 pid = proc_task(inode)->pid;
1282 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
1288 ino = parent_ino(dentry);
1289 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
1302 if (filldir(dirent, p->name, p->len, filp->f_pos,
1303 fake_ino(pid, p->type), p->mode >> 12) < 0)
1315 static int proc_tgid_base_readdir(struct file * filp,
1316 void * dirent, filldir_t filldir)
1318 return proc_pident_readdir(filp,dirent,filldir,
1319 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1322 static int proc_tid_base_readdir(struct file * filp,
1323 void * dirent, filldir_t filldir)
1325 return proc_pident_readdir(filp,dirent,filldir,
1326 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
1329 /* building an inode */
1331 static int task_dumpable(struct task_struct *task)
1334 struct mm_struct *mm;
1339 dumpable = mm->dumpable;
1347 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task, int ino)
1349 struct inode * inode;
1350 struct proc_inode *ei;
1352 /* We need a new inode */
1354 inode = new_inode(sb);
1361 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1362 inode->i_ino = fake_ino(task->pid, ino);
1364 if (!pid_alive(task))
1368 * grab the reference to task.
1370 get_task_struct(task);
1375 if (ino == PROC_TGID_INO || ino == PROC_TID_INO || task_dumpable(task)) {
1376 inode->i_uid = task->euid;
1377 inode->i_gid = task->egid;
1379 security_task_to_inode(task, inode);
1393 * Exceptional case: normally we are not allowed to unhash a busy
1394 * directory. In this case, however, we can do it - no aliasing problems
1395 * due to the way we treat inodes.
1397 * Rewrite the inode's ownerships here because the owning task may have
1398 * performed a setuid(), etc.
1400 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1402 struct inode *inode = dentry->d_inode;
1403 struct task_struct *task = proc_task(inode);
1404 if (pid_alive(task)) {
1405 if (proc_type(inode) == PROC_TGID_INO || proc_type(inode) == PROC_TID_INO || task_dumpable(task)) {
1406 inode->i_uid = task->euid;
1407 inode->i_gid = task->egid;
1412 security_task_to_inode(task, inode);
1419 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1421 struct inode *inode = dentry->d_inode;
1422 struct task_struct *task = proc_task(inode);
1423 int fd = proc_type(inode) - PROC_TID_FD_DIR;
1424 struct files_struct *files;
1426 files = get_files_struct(task);
1429 if (fcheck_files(files, fd)) {
1431 put_files_struct(files);
1432 if (task_dumpable(task)) {
1433 inode->i_uid = task->euid;
1434 inode->i_gid = task->egid;
1439 security_task_to_inode(task, inode);
1443 put_files_struct(files);
1449 static void pid_base_iput(struct dentry *dentry, struct inode *inode)
1451 struct task_struct *task = proc_task(inode);
1452 spin_lock(&task->proc_lock);
1453 if (task->proc_dentry == dentry)
1454 task->proc_dentry = NULL;
1455 spin_unlock(&task->proc_lock);
1459 static int pid_delete_dentry(struct dentry * dentry)
1461 /* Is the task we represent dead?
1462 * If so, then don't put the dentry on the lru list,
1463 * kill it immediately.
1465 return !pid_alive(proc_task(dentry->d_inode));
1468 static struct dentry_operations tid_fd_dentry_operations =
1470 .d_revalidate = tid_fd_revalidate,
1471 .d_delete = pid_delete_dentry,
1474 static struct dentry_operations pid_dentry_operations =
1476 .d_revalidate = pid_revalidate,
1477 .d_delete = pid_delete_dentry,
1480 static struct dentry_operations pid_base_dentry_operations =
1482 .d_revalidate = pid_revalidate,
1483 .d_iput = pid_base_iput,
1484 .d_delete = pid_delete_dentry,
1489 static unsigned name_to_int(struct dentry *dentry)
1491 const char *name = dentry->d_name.name;
1492 int len = dentry->d_name.len;
1495 if (len > 1 && *name == '0')
1498 unsigned c = *name++ - '0';
1501 if (n >= (~0U-9)/10)
1512 static struct dentry *proc_lookupfd(struct inode * dir, struct dentry * dentry, struct nameidata *nd)
1514 struct task_struct *task = proc_task(dir);
1515 unsigned fd = name_to_int(dentry);
1517 struct files_struct * files;
1518 struct inode *inode;
1519 struct proc_inode *ei;
1523 if (!pid_alive(task))
1526 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TID_FD_DIR+fd);
1530 files = get_files_struct(task);
1533 inode->i_mode = S_IFLNK;
1536 * We are not taking a ref to the file structure, so we must
1539 spin_lock(&files->file_lock);
1540 file = fcheck_files(files, fd);
1543 if (file->f_mode & 1)
1544 inode->i_mode |= S_IRUSR | S_IXUSR;
1545 if (file->f_mode & 2)
1546 inode->i_mode |= S_IWUSR | S_IXUSR;
1547 spin_unlock(&files->file_lock);
1548 put_files_struct(files);
1549 inode->i_op = &proc_pid_link_inode_operations;
1551 ei->op.proc_get_link = proc_fd_link;
1552 dentry->d_op = &tid_fd_dentry_operations;
1553 d_add(dentry, inode);
1557 spin_unlock(&files->file_lock);
1558 put_files_struct(files);
1562 return ERR_PTR(-ENOENT);
1565 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir);
1566 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd);
1568 static struct file_operations proc_fd_operations = {
1569 .read = generic_read_dir,
1570 .readdir = proc_readfd,
1573 static struct file_operations proc_task_operations = {
1574 .read = generic_read_dir,
1575 .readdir = proc_task_readdir,
1579 * proc directories can do almost nothing..
1581 static struct inode_operations proc_fd_inode_operations = {
1582 .lookup = proc_lookupfd,
1583 .permission = proc_permission,
1586 static struct inode_operations proc_task_inode_operations = {
1587 .lookup = proc_task_lookup,
1588 .permission = proc_task_permission,
1591 #ifdef CONFIG_SECURITY
1592 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
1593 size_t count, loff_t *ppos)
1595 struct inode * inode = file->f_dentry->d_inode;
1598 struct task_struct *task = proc_task(inode);
1600 if (count > PAGE_SIZE)
1602 if (!(page = __get_free_page(GFP_KERNEL)))
1605 length = security_getprocattr(task,
1606 (char*)file->f_dentry->d_name.name,
1607 (void*)page, count);
1609 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
1614 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
1615 size_t count, loff_t *ppos)
1617 struct inode * inode = file->f_dentry->d_inode;
1620 struct task_struct *task = proc_task(inode);
1622 if (count > PAGE_SIZE)
1625 /* No partial writes. */
1628 page = (char*)__get_free_page(GFP_USER);
1632 if (copy_from_user(page, buf, count))
1635 length = security_setprocattr(task,
1636 (char*)file->f_dentry->d_name.name,
1637 (void*)page, count);
1639 free_page((unsigned long) page);
1643 static struct file_operations proc_pid_attr_operations = {
1644 .read = proc_pid_attr_read,
1645 .write = proc_pid_attr_write,
1648 static struct file_operations proc_tid_attr_operations;
1649 static struct inode_operations proc_tid_attr_inode_operations;
1650 static struct file_operations proc_tgid_attr_operations;
1651 static struct inode_operations proc_tgid_attr_inode_operations;
1654 static int get_tid_list(int index, unsigned int *tids, struct inode *dir);
1657 static struct dentry *proc_pident_lookup(struct inode *dir,
1658 struct dentry *dentry,
1659 struct pid_entry *ents)
1661 struct inode *inode;
1663 struct task_struct *task = proc_task(dir);
1664 struct pid_entry *p;
1665 struct proc_inode *ei;
1670 if (!pid_alive(task))
1673 for (p = ents; p->name; p++) {
1674 if (p->len != dentry->d_name.len)
1676 if (!memcmp(dentry->d_name.name, p->name, p->len))
1683 inode = proc_pid_make_inode(dir->i_sb, task, p->type);
1688 inode->i_mode = p->mode;
1690 * Yes, it does not scale. And it should not. Don't add
1691 * new entries into /proc/<tgid>/ without very good reasons.
1694 case PROC_TGID_TASK:
1695 inode->i_nlink = 2 + get_tid_list(2, NULL, dir);
1696 inode->i_op = &proc_task_inode_operations;
1697 inode->i_fop = &proc_task_operations;
1702 inode->i_op = &proc_fd_inode_operations;
1703 inode->i_fop = &proc_fd_operations;
1707 inode->i_op = &proc_pid_link_inode_operations;
1708 ei->op.proc_get_link = proc_exe_link;
1712 inode->i_op = &proc_pid_link_inode_operations;
1713 ei->op.proc_get_link = proc_cwd_link;
1716 case PROC_TGID_ROOT:
1717 inode->i_op = &proc_pid_link_inode_operations;
1718 ei->op.proc_get_link = proc_root_link;
1720 case PROC_TID_ENVIRON:
1721 case PROC_TGID_ENVIRON:
1722 inode->i_fop = &proc_info_file_operations;
1723 ei->op.proc_read = proc_pid_environ;
1726 case PROC_TGID_AUXV:
1727 inode->i_fop = &proc_info_file_operations;
1728 ei->op.proc_read = proc_pid_auxv;
1730 case PROC_TID_STATUS:
1731 case PROC_TGID_STATUS:
1732 inode->i_fop = &proc_info_file_operations;
1733 ei->op.proc_read = proc_pid_status;
1736 inode->i_fop = &proc_info_file_operations;
1737 ei->op.proc_read = proc_tid_stat;
1739 case PROC_TGID_STAT:
1740 inode->i_fop = &proc_info_file_operations;
1741 ei->op.proc_read = proc_tgid_stat;
1743 case PROC_TID_CMDLINE:
1744 case PROC_TGID_CMDLINE:
1745 inode->i_fop = &proc_info_file_operations;
1746 ei->op.proc_read = proc_pid_cmdline;
1748 case PROC_TID_STATM:
1749 case PROC_TGID_STATM:
1750 inode->i_fop = &proc_info_file_operations;
1751 ei->op.proc_read = proc_pid_statm;
1754 case PROC_TGID_MAPS:
1755 inode->i_fop = &proc_maps_operations;
1758 case PROC_TID_NUMA_MAPS:
1759 case PROC_TGID_NUMA_MAPS:
1760 inode->i_fop = &proc_numa_maps_operations;
1765 inode->i_op = &proc_mem_inode_operations;
1766 inode->i_fop = &proc_mem_operations;
1768 #ifdef CONFIG_SECCOMP
1769 case PROC_TID_SECCOMP:
1770 case PROC_TGID_SECCOMP:
1771 inode->i_fop = &proc_seccomp_operations;
1773 #endif /* CONFIG_SECCOMP */
1774 case PROC_TID_MOUNTS:
1775 case PROC_TGID_MOUNTS:
1776 inode->i_fop = &proc_mounts_operations;
1779 case PROC_TID_SMAPS:
1780 case PROC_TGID_SMAPS:
1781 inode->i_fop = &proc_smaps_operations;
1784 case PROC_TID_MOUNTSTATS:
1785 case PROC_TGID_MOUNTSTATS:
1786 inode->i_fop = &proc_mountstats_operations;
1788 #ifdef CONFIG_SECURITY
1791 inode->i_op = &proc_tid_attr_inode_operations;
1792 inode->i_fop = &proc_tid_attr_operations;
1794 case PROC_TGID_ATTR:
1796 inode->i_op = &proc_tgid_attr_inode_operations;
1797 inode->i_fop = &proc_tgid_attr_operations;
1799 case PROC_TID_ATTR_CURRENT:
1800 case PROC_TGID_ATTR_CURRENT:
1801 case PROC_TID_ATTR_PREV:
1802 case PROC_TGID_ATTR_PREV:
1803 case PROC_TID_ATTR_EXEC:
1804 case PROC_TGID_ATTR_EXEC:
1805 case PROC_TID_ATTR_FSCREATE:
1806 case PROC_TGID_ATTR_FSCREATE:
1807 case PROC_TID_ATTR_KEYCREATE:
1808 case PROC_TGID_ATTR_KEYCREATE:
1809 inode->i_fop = &proc_pid_attr_operations;
1812 #ifdef CONFIG_KALLSYMS
1813 case PROC_TID_WCHAN:
1814 case PROC_TGID_WCHAN:
1815 inode->i_fop = &proc_info_file_operations;
1816 ei->op.proc_read = proc_pid_wchan;
1819 #ifdef CONFIG_SCHEDSTATS
1820 case PROC_TID_SCHEDSTAT:
1821 case PROC_TGID_SCHEDSTAT:
1822 inode->i_fop = &proc_info_file_operations;
1823 ei->op.proc_read = proc_pid_schedstat;
1826 #ifdef CONFIG_CPUSETS
1827 case PROC_TID_CPUSET:
1828 case PROC_TGID_CPUSET:
1829 inode->i_fop = &proc_cpuset_operations;
1832 case PROC_TID_OOM_SCORE:
1833 case PROC_TGID_OOM_SCORE:
1834 inode->i_fop = &proc_info_file_operations;
1835 ei->op.proc_read = proc_oom_score;
1837 case PROC_TID_OOM_ADJUST:
1838 case PROC_TGID_OOM_ADJUST:
1839 inode->i_fop = &proc_oom_adjust_operations;
1841 #ifdef CONFIG_AUDITSYSCALL
1842 case PROC_TID_LOGINUID:
1843 case PROC_TGID_LOGINUID:
1844 inode->i_fop = &proc_loginuid_operations;
1848 printk("procfs: impossible type (%d)",p->type);
1850 return ERR_PTR(-EINVAL);
1852 dentry->d_op = &pid_dentry_operations;
1853 d_add(dentry, inode);
1857 return ERR_PTR(error);
1860 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
1861 return proc_pident_lookup(dir, dentry, tgid_base_stuff);
1864 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
1865 return proc_pident_lookup(dir, dentry, tid_base_stuff);
1868 static struct file_operations proc_tgid_base_operations = {
1869 .read = generic_read_dir,
1870 .readdir = proc_tgid_base_readdir,
1873 static struct file_operations proc_tid_base_operations = {
1874 .read = generic_read_dir,
1875 .readdir = proc_tid_base_readdir,
1878 static struct inode_operations proc_tgid_base_inode_operations = {
1879 .lookup = proc_tgid_base_lookup,
1882 static struct inode_operations proc_tid_base_inode_operations = {
1883 .lookup = proc_tid_base_lookup,
1886 #ifdef CONFIG_SECURITY
1887 static int proc_tgid_attr_readdir(struct file * filp,
1888 void * dirent, filldir_t filldir)
1890 return proc_pident_readdir(filp,dirent,filldir,
1891 tgid_attr_stuff,ARRAY_SIZE(tgid_attr_stuff));
1894 static int proc_tid_attr_readdir(struct file * filp,
1895 void * dirent, filldir_t filldir)
1897 return proc_pident_readdir(filp,dirent,filldir,
1898 tid_attr_stuff,ARRAY_SIZE(tid_attr_stuff));
1901 static struct file_operations proc_tgid_attr_operations = {
1902 .read = generic_read_dir,
1903 .readdir = proc_tgid_attr_readdir,
1906 static struct file_operations proc_tid_attr_operations = {
1907 .read = generic_read_dir,
1908 .readdir = proc_tid_attr_readdir,
1911 static struct dentry *proc_tgid_attr_lookup(struct inode *dir,
1912 struct dentry *dentry, struct nameidata *nd)
1914 return proc_pident_lookup(dir, dentry, tgid_attr_stuff);
1917 static struct dentry *proc_tid_attr_lookup(struct inode *dir,
1918 struct dentry *dentry, struct nameidata *nd)
1920 return proc_pident_lookup(dir, dentry, tid_attr_stuff);
1923 static struct inode_operations proc_tgid_attr_inode_operations = {
1924 .lookup = proc_tgid_attr_lookup,
1927 static struct inode_operations proc_tid_attr_inode_operations = {
1928 .lookup = proc_tid_attr_lookup,
1935 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
1939 sprintf(tmp, "%d", current->tgid);
1940 return vfs_readlink(dentry,buffer,buflen,tmp);
1943 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
1946 sprintf(tmp, "%d", current->tgid);
1947 return ERR_PTR(vfs_follow_link(nd,tmp));
1950 static struct inode_operations proc_self_inode_operations = {
1951 .readlink = proc_self_readlink,
1952 .follow_link = proc_self_follow_link,
1956 * proc_pid_unhash - Unhash /proc/@pid entry from the dcache.
1957 * @p: task that should be flushed.
1959 * Drops the /proc/@pid dcache entry from the hash chains.
1961 * Dropping /proc/@pid entries and detach_pid must be synchroneous,
1962 * otherwise e.g. /proc/@pid/exe might point to the wrong executable,
1963 * if the pid value is immediately reused. This is enforced by
1964 * - caller must acquire spin_lock(p->proc_lock)
1965 * - must be called before detach_pid()
1966 * - proc_pid_lookup acquires proc_lock, and checks that
1967 * the target is not dead by looking at the attach count
1971 struct dentry *proc_pid_unhash(struct task_struct *p)
1973 struct dentry *proc_dentry;
1975 proc_dentry = p->proc_dentry;
1976 if (proc_dentry != NULL) {
1978 spin_lock(&dcache_lock);
1979 spin_lock(&proc_dentry->d_lock);
1980 if (!d_unhashed(proc_dentry)) {
1981 dget_locked(proc_dentry);
1982 __d_drop(proc_dentry);
1983 spin_unlock(&proc_dentry->d_lock);
1985 spin_unlock(&proc_dentry->d_lock);
1988 spin_unlock(&dcache_lock);
1994 * proc_pid_flush - recover memory used by stale /proc/@pid/x entries
1995 * @proc_dentry: directoy to prune.
1997 * Shrink the /proc directory that was used by the just killed thread.
2000 void proc_pid_flush(struct dentry *proc_dentry)
2003 if(proc_dentry != NULL) {
2004 shrink_dcache_parent(proc_dentry);
2010 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2012 struct task_struct *task;
2013 struct inode *inode;
2014 struct proc_inode *ei;
2018 if (dentry->d_name.len == 4 && !memcmp(dentry->d_name.name,"self",4)) {
2019 inode = new_inode(dir->i_sb);
2021 return ERR_PTR(-ENOMEM);
2023 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2024 inode->i_ino = fake_ino(0, PROC_TGID_INO);
2026 inode->i_mode = S_IFLNK|S_IRWXUGO;
2027 inode->i_uid = inode->i_gid = 0;
2029 inode->i_op = &proc_self_inode_operations;
2030 d_add(dentry, inode);
2033 tgid = name_to_int(dentry);
2037 read_lock(&tasklist_lock);
2038 task = find_task_by_pid(tgid);
2040 get_task_struct(task);
2041 read_unlock(&tasklist_lock);
2045 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TGID_INO);
2049 put_task_struct(task);
2052 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2053 inode->i_op = &proc_tgid_base_inode_operations;
2054 inode->i_fop = &proc_tgid_base_operations;
2055 inode->i_flags|=S_IMMUTABLE;
2056 #ifdef CONFIG_SECURITY
2062 dentry->d_op = &pid_base_dentry_operations;
2065 d_add(dentry, inode);
2066 spin_lock(&task->proc_lock);
2067 task->proc_dentry = dentry;
2068 if (!pid_alive(task)) {
2069 dentry = proc_pid_unhash(task);
2072 spin_unlock(&task->proc_lock);
2074 put_task_struct(task);
2076 proc_pid_flush(dentry);
2081 return ERR_PTR(-ENOENT);
2085 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2087 struct task_struct *task;
2088 struct task_struct *leader = proc_task(dir);
2089 struct inode *inode;
2092 tid = name_to_int(dentry);
2096 read_lock(&tasklist_lock);
2097 task = find_task_by_pid(tid);
2099 get_task_struct(task);
2100 read_unlock(&tasklist_lock);
2103 if (leader->tgid != task->tgid)
2106 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TID_INO);
2111 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2112 inode->i_op = &proc_tid_base_inode_operations;
2113 inode->i_fop = &proc_tid_base_operations;
2114 inode->i_flags|=S_IMMUTABLE;
2115 #ifdef CONFIG_SECURITY
2121 dentry->d_op = &pid_base_dentry_operations;
2123 d_add(dentry, inode);
2125 put_task_struct(task);
2128 put_task_struct(task);
2130 return ERR_PTR(-ENOENT);
2133 #define PROC_NUMBUF 10
2134 #define PROC_MAXPIDS 20
2137 * Get a few tgid's to return for filldir - we need to hold the
2138 * tasklist lock while doing this, and we must release it before
2139 * we actually do the filldir itself, so we use a temp buffer..
2141 static int get_tgid_list(int index, unsigned long version, unsigned int *tgids)
2143 struct task_struct *p;
2147 read_lock(&tasklist_lock);
2150 p = find_task_by_pid(version);
2151 if (p && !thread_group_leader(p))
2158 p = next_task(&init_task);
2160 for ( ; p != &init_task; p = next_task(p)) {
2166 tgids[nr_tgids] = tgid;
2168 if (nr_tgids >= PROC_MAXPIDS)
2171 read_unlock(&tasklist_lock);
2176 * Get a few tid's to return for filldir - we need to hold the
2177 * tasklist lock while doing this, and we must release it before
2178 * we actually do the filldir itself, so we use a temp buffer..
2180 static int get_tid_list(int index, unsigned int *tids, struct inode *dir)
2182 struct task_struct *leader_task = proc_task(dir);
2183 struct task_struct *task = leader_task;
2187 read_lock(&tasklist_lock);
2189 * The starting point task (leader_task) might be an already
2190 * unlinked task, which cannot be used to access the task-list
2191 * via next_thread().
2193 if (pid_alive(task)) do {
2194 int tid = task->pid;
2199 tids[nr_tids] = tid;
2201 if (nr_tids >= PROC_MAXPIDS)
2203 } while ((task = next_thread(task)) != leader_task);
2204 read_unlock(&tasklist_lock);
2208 /* for the /proc/ directory itself, after non-process stuff has been done */
2209 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2211 unsigned int tgid_array[PROC_MAXPIDS];
2212 char buf[PROC_NUMBUF];
2213 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2214 unsigned int nr_tgids, i;
2218 ino_t ino = fake_ino(0,PROC_TGID_INO);
2219 if (filldir(dirent, "self", 4, filp->f_pos, ino, DT_LNK) < 0)
2225 /* f_version caches the tgid value that the last readdir call couldn't
2226 * return. lseek aka telldir automagically resets f_version to 0.
2228 next_tgid = filp->f_version;
2229 filp->f_version = 0;
2231 nr_tgids = get_tgid_list(nr, next_tgid, tgid_array);
2233 /* no more entries ! */
2238 /* do not use the last found pid, reserve it for next_tgid */
2239 if (nr_tgids == PROC_MAXPIDS) {
2241 next_tgid = tgid_array[nr_tgids];
2244 for (i=0;i<nr_tgids;i++) {
2245 int tgid = tgid_array[i];
2246 ino_t ino = fake_ino(tgid,PROC_TGID_INO);
2247 unsigned long j = PROC_NUMBUF;
2250 buf[--j] = '0' + (tgid % 10);
2251 while ((tgid /= 10) != 0);
2253 if (filldir(dirent, buf+j, PROC_NUMBUF-j, filp->f_pos, ino, DT_DIR) < 0) {
2254 /* returning this tgid failed, save it as the first
2255 * pid for the next readir call */
2256 filp->f_version = tgid_array[i];
2267 /* for the /proc/TGID/task/ directories */
2268 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
2270 unsigned int tid_array[PROC_MAXPIDS];
2271 char buf[PROC_NUMBUF];
2272 unsigned int nr_tids, i;
2273 struct dentry *dentry = filp->f_dentry;
2274 struct inode *inode = dentry->d_inode;
2275 int retval = -ENOENT;
2277 unsigned long pos = filp->f_pos; /* avoiding "long long" filp->f_pos */
2279 if (!pid_alive(proc_task(inode)))
2286 if (filldir(dirent, ".", 1, pos, ino, DT_DIR) < 0)
2291 ino = parent_ino(dentry);
2292 if (filldir(dirent, "..", 2, pos, ino, DT_DIR) < 0)
2298 nr_tids = get_tid_list(pos, tid_array, inode);
2299 inode->i_nlink = pos + nr_tids;
2301 for (i = 0; i < nr_tids; i++) {
2302 unsigned long j = PROC_NUMBUF;
2303 int tid = tid_array[i];
2305 ino = fake_ino(tid,PROC_TID_INO);
2308 buf[--j] = '0' + (tid % 10);
2309 while ((tid /= 10) != 0);
2311 if (filldir(dirent, buf+j, PROC_NUMBUF-j, pos, ino, DT_DIR) < 0)