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 */
195 #define E(type,name,mode) {(type),sizeof(name)-1,(name),(mode)}
197 static struct pid_entry tgid_base_stuff[] = {
198 E(PROC_TGID_TASK, "task", S_IFDIR|S_IRUGO|S_IXUGO),
199 E(PROC_TGID_FD, "fd", S_IFDIR|S_IRUSR|S_IXUSR),
200 E(PROC_TGID_ENVIRON, "environ", S_IFREG|S_IRUSR),
201 E(PROC_TGID_AUXV, "auxv", S_IFREG|S_IRUSR),
202 E(PROC_TGID_STATUS, "status", S_IFREG|S_IRUGO),
203 E(PROC_TGID_CMDLINE, "cmdline", S_IFREG|S_IRUGO),
204 E(PROC_TGID_STAT, "stat", S_IFREG|S_IRUGO),
205 E(PROC_TGID_STATM, "statm", S_IFREG|S_IRUGO),
206 E(PROC_TGID_MAPS, "maps", S_IFREG|S_IRUGO),
208 E(PROC_TGID_NUMA_MAPS, "numa_maps", S_IFREG|S_IRUGO),
210 E(PROC_TGID_MEM, "mem", S_IFREG|S_IRUSR|S_IWUSR),
211 #ifdef CONFIG_SECCOMP
212 E(PROC_TGID_SECCOMP, "seccomp", S_IFREG|S_IRUSR|S_IWUSR),
214 E(PROC_TGID_CWD, "cwd", S_IFLNK|S_IRWXUGO),
215 E(PROC_TGID_ROOT, "root", S_IFLNK|S_IRWXUGO),
216 E(PROC_TGID_EXE, "exe", S_IFLNK|S_IRWXUGO),
217 E(PROC_TGID_MOUNTS, "mounts", S_IFREG|S_IRUGO),
218 E(PROC_TGID_MOUNTSTATS, "mountstats", S_IFREG|S_IRUSR),
220 E(PROC_TGID_SMAPS, "smaps", S_IFREG|S_IRUGO),
222 #ifdef CONFIG_SECURITY
223 E(PROC_TGID_ATTR, "attr", S_IFDIR|S_IRUGO|S_IXUGO),
225 #ifdef CONFIG_KALLSYMS
226 E(PROC_TGID_WCHAN, "wchan", S_IFREG|S_IRUGO),
228 #ifdef CONFIG_SCHEDSTATS
229 E(PROC_TGID_SCHEDSTAT, "schedstat", S_IFREG|S_IRUGO),
231 #ifdef CONFIG_CPUSETS
232 E(PROC_TGID_CPUSET, "cpuset", S_IFREG|S_IRUGO),
234 E(PROC_TGID_OOM_SCORE, "oom_score",S_IFREG|S_IRUGO),
235 E(PROC_TGID_OOM_ADJUST,"oom_adj", S_IFREG|S_IRUGO|S_IWUSR),
236 #ifdef CONFIG_AUDITSYSCALL
237 E(PROC_TGID_LOGINUID, "loginuid", S_IFREG|S_IWUSR|S_IRUGO),
241 static struct pid_entry tid_base_stuff[] = {
242 E(PROC_TID_FD, "fd", S_IFDIR|S_IRUSR|S_IXUSR),
243 E(PROC_TID_ENVIRON, "environ", S_IFREG|S_IRUSR),
244 E(PROC_TID_AUXV, "auxv", S_IFREG|S_IRUSR),
245 E(PROC_TID_STATUS, "status", S_IFREG|S_IRUGO),
246 E(PROC_TID_CMDLINE, "cmdline", S_IFREG|S_IRUGO),
247 E(PROC_TID_STAT, "stat", S_IFREG|S_IRUGO),
248 E(PROC_TID_STATM, "statm", S_IFREG|S_IRUGO),
249 E(PROC_TID_MAPS, "maps", S_IFREG|S_IRUGO),
251 E(PROC_TID_NUMA_MAPS, "numa_maps", S_IFREG|S_IRUGO),
253 E(PROC_TID_MEM, "mem", S_IFREG|S_IRUSR|S_IWUSR),
254 #ifdef CONFIG_SECCOMP
255 E(PROC_TID_SECCOMP, "seccomp", S_IFREG|S_IRUSR|S_IWUSR),
257 E(PROC_TID_CWD, "cwd", S_IFLNK|S_IRWXUGO),
258 E(PROC_TID_ROOT, "root", S_IFLNK|S_IRWXUGO),
259 E(PROC_TID_EXE, "exe", S_IFLNK|S_IRWXUGO),
260 E(PROC_TID_MOUNTS, "mounts", S_IFREG|S_IRUGO),
262 E(PROC_TID_SMAPS, "smaps", S_IFREG|S_IRUGO),
264 #ifdef CONFIG_SECURITY
265 E(PROC_TID_ATTR, "attr", S_IFDIR|S_IRUGO|S_IXUGO),
267 #ifdef CONFIG_KALLSYMS
268 E(PROC_TID_WCHAN, "wchan", S_IFREG|S_IRUGO),
270 #ifdef CONFIG_SCHEDSTATS
271 E(PROC_TID_SCHEDSTAT, "schedstat",S_IFREG|S_IRUGO),
273 #ifdef CONFIG_CPUSETS
274 E(PROC_TID_CPUSET, "cpuset", S_IFREG|S_IRUGO),
276 E(PROC_TID_OOM_SCORE, "oom_score",S_IFREG|S_IRUGO),
277 E(PROC_TID_OOM_ADJUST, "oom_adj", S_IFREG|S_IRUGO|S_IWUSR),
278 #ifdef CONFIG_AUDITSYSCALL
279 E(PROC_TID_LOGINUID, "loginuid", S_IFREG|S_IWUSR|S_IRUGO),
284 #ifdef CONFIG_SECURITY
285 static struct pid_entry tgid_attr_stuff[] = {
286 E(PROC_TGID_ATTR_CURRENT, "current", S_IFREG|S_IRUGO|S_IWUGO),
287 E(PROC_TGID_ATTR_PREV, "prev", S_IFREG|S_IRUGO),
288 E(PROC_TGID_ATTR_EXEC, "exec", S_IFREG|S_IRUGO|S_IWUGO),
289 E(PROC_TGID_ATTR_FSCREATE, "fscreate", S_IFREG|S_IRUGO|S_IWUGO),
290 E(PROC_TGID_ATTR_KEYCREATE, "keycreate", S_IFREG|S_IRUGO|S_IWUGO),
293 static struct pid_entry tid_attr_stuff[] = {
294 E(PROC_TID_ATTR_CURRENT, "current", S_IFREG|S_IRUGO|S_IWUGO),
295 E(PROC_TID_ATTR_PREV, "prev", S_IFREG|S_IRUGO),
296 E(PROC_TID_ATTR_EXEC, "exec", S_IFREG|S_IRUGO|S_IWUGO),
297 E(PROC_TID_ATTR_FSCREATE, "fscreate", S_IFREG|S_IRUGO|S_IWUGO),
298 E(PROC_TID_ATTR_KEYCREATE, "keycreate", S_IFREG|S_IRUGO|S_IWUGO),
305 static int proc_fd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
307 struct task_struct *task = proc_task(inode);
308 struct files_struct *files;
310 int fd = proc_fd(inode);
312 files = get_files_struct(task);
315 * We are not taking a ref to the file structure, so we must
318 spin_lock(&files->file_lock);
319 file = fcheck_files(files, fd);
321 *mnt = mntget(file->f_vfsmnt);
322 *dentry = dget(file->f_dentry);
323 spin_unlock(&files->file_lock);
324 put_files_struct(files);
327 spin_unlock(&files->file_lock);
328 put_files_struct(files);
333 static struct fs_struct *get_fs_struct(struct task_struct *task)
335 struct fs_struct *fs;
339 atomic_inc(&fs->count);
344 static int proc_cwd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
346 struct fs_struct *fs = get_fs_struct(proc_task(inode));
347 int result = -ENOENT;
349 read_lock(&fs->lock);
350 *mnt = mntget(fs->pwdmnt);
351 *dentry = dget(fs->pwd);
352 read_unlock(&fs->lock);
359 static int proc_root_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
361 struct fs_struct *fs = get_fs_struct(proc_task(inode));
362 int result = -ENOENT;
364 read_lock(&fs->lock);
365 *mnt = mntget(fs->rootmnt);
366 *dentry = dget(fs->root);
367 read_unlock(&fs->lock);
374 #define MAY_PTRACE(task) \
375 (task == current || \
376 (task->parent == current && \
377 (task->ptrace & PT_PTRACED) && \
378 (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
379 security_ptrace(current,task) == 0))
381 static int proc_pid_environ(struct task_struct *task, char * buffer)
384 struct mm_struct *mm = get_task_mm(task);
386 unsigned int len = mm->env_end - mm->env_start;
389 res = access_process_vm(task, mm->env_start, buffer, len, 0);
390 if (!ptrace_may_attach(task))
397 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
401 struct mm_struct *mm = get_task_mm(task);
405 goto out_mm; /* Shh! No looking before we're done */
407 len = mm->arg_end - mm->arg_start;
412 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
414 // If the nul at the end of args has been overwritten, then
415 // assume application is using setproctitle(3).
416 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
417 len = strnlen(buffer, res);
421 len = mm->env_end - mm->env_start;
422 if (len > PAGE_SIZE - res)
423 len = PAGE_SIZE - res;
424 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
425 res = strnlen(buffer, res);
434 static int proc_pid_auxv(struct task_struct *task, char *buffer)
437 struct mm_struct *mm = get_task_mm(task);
439 unsigned int nwords = 0;
442 while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
443 res = nwords * sizeof(mm->saved_auxv[0]);
446 memcpy(buffer, mm->saved_auxv, res);
453 #ifdef CONFIG_KALLSYMS
455 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
456 * Returns the resolved symbol. If that fails, simply return the address.
458 static int proc_pid_wchan(struct task_struct *task, char *buffer)
461 const char *sym_name;
462 unsigned long wchan, size, offset;
463 char namebuf[KSYM_NAME_LEN+1];
465 wchan = get_wchan(task);
467 sym_name = kallsyms_lookup(wchan, &size, &offset, &modname, namebuf);
469 return sprintf(buffer, "%s", sym_name);
470 return sprintf(buffer, "%lu", wchan);
472 #endif /* CONFIG_KALLSYMS */
474 #ifdef CONFIG_SCHEDSTATS
476 * Provides /proc/PID/schedstat
478 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
480 return sprintf(buffer, "%lu %lu %lu\n",
481 task->sched_info.cpu_time,
482 task->sched_info.run_delay,
483 task->sched_info.pcnt);
487 /* The badness from the OOM killer */
488 unsigned long badness(struct task_struct *p, unsigned long uptime);
489 static int proc_oom_score(struct task_struct *task, char *buffer)
491 unsigned long points;
492 struct timespec uptime;
494 do_posix_clock_monotonic_gettime(&uptime);
495 points = badness(task, uptime.tv_sec);
496 return sprintf(buffer, "%lu\n", points);
499 /************************************************************************/
500 /* Here the fs part begins */
501 /************************************************************************/
503 /* permission checks */
505 /* If the process being read is separated by chroot from the reading process,
506 * don't let the reader access the threads.
508 static int proc_check_chroot(struct dentry *de, struct vfsmount *mnt)
511 struct vfsmount *our_vfsmnt;
514 read_lock(¤t->fs->lock);
515 our_vfsmnt = mntget(current->fs->rootmnt);
516 base = dget(current->fs->root);
517 read_unlock(¤t->fs->lock);
519 spin_lock(&vfsmount_lock);
521 while (mnt != our_vfsmnt) {
522 if (mnt == mnt->mnt_parent)
524 de = mnt->mnt_mountpoint;
525 mnt = mnt->mnt_parent;
528 if (!is_subdir(de, base))
530 spin_unlock(&vfsmount_lock);
537 spin_unlock(&vfsmount_lock);
542 extern struct seq_operations proc_pid_maps_op;
543 static int maps_open(struct inode *inode, struct file *file)
545 struct task_struct *task = proc_task(inode);
546 int ret = seq_open(file, &proc_pid_maps_op);
548 struct seq_file *m = file->private_data;
554 static struct file_operations proc_maps_operations = {
558 .release = seq_release,
562 extern struct seq_operations proc_pid_numa_maps_op;
563 static int numa_maps_open(struct inode *inode, struct file *file)
565 struct task_struct *task = proc_task(inode);
566 int ret = seq_open(file, &proc_pid_numa_maps_op);
568 struct seq_file *m = file->private_data;
574 static struct file_operations proc_numa_maps_operations = {
575 .open = numa_maps_open,
578 .release = seq_release,
583 extern struct seq_operations proc_pid_smaps_op;
584 static int smaps_open(struct inode *inode, struct file *file)
586 struct task_struct *task = proc_task(inode);
587 int ret = seq_open(file, &proc_pid_smaps_op);
589 struct seq_file *m = file->private_data;
595 static struct file_operations proc_smaps_operations = {
599 .release = seq_release,
603 extern struct seq_operations mounts_op;
609 static int mounts_open(struct inode *inode, struct file *file)
611 struct task_struct *task = proc_task(inode);
612 struct namespace *namespace;
613 struct proc_mounts *p;
617 namespace = task->namespace;
619 get_namespace(namespace);
624 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
626 file->private_data = &p->m;
627 ret = seq_open(file, &mounts_op);
629 p->m.private = namespace;
630 p->event = namespace->event;
635 put_namespace(namespace);
640 static int mounts_release(struct inode *inode, struct file *file)
642 struct seq_file *m = file->private_data;
643 struct namespace *namespace = m->private;
644 put_namespace(namespace);
645 return seq_release(inode, file);
648 static unsigned mounts_poll(struct file *file, poll_table *wait)
650 struct proc_mounts *p = file->private_data;
651 struct namespace *ns = p->m.private;
654 poll_wait(file, &ns->poll, wait);
656 spin_lock(&vfsmount_lock);
657 if (p->event != ns->event) {
658 p->event = ns->event;
661 spin_unlock(&vfsmount_lock);
666 static struct file_operations proc_mounts_operations = {
670 .release = mounts_release,
674 extern struct seq_operations mountstats_op;
675 static int mountstats_open(struct inode *inode, struct file *file)
677 struct task_struct *task = proc_task(inode);
678 int ret = seq_open(file, &mountstats_op);
681 struct seq_file *m = file->private_data;
682 struct namespace *namespace;
684 namespace = task->namespace;
686 get_namespace(namespace);
690 m->private = namespace;
692 seq_release(inode, file);
699 static struct file_operations proc_mountstats_operations = {
700 .open = mountstats_open,
703 .release = mounts_release,
706 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
708 static ssize_t proc_info_read(struct file * file, char __user * buf,
709 size_t count, loff_t *ppos)
711 struct inode * inode = file->f_dentry->d_inode;
714 struct task_struct *task = proc_task(inode);
716 if (count > PROC_BLOCK_SIZE)
717 count = PROC_BLOCK_SIZE;
718 if (!(page = __get_free_page(GFP_KERNEL)))
721 length = PROC_I(inode)->op.proc_read(task, (char*)page);
724 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
729 static struct file_operations proc_info_file_operations = {
730 .read = proc_info_read,
733 static int mem_open(struct inode* inode, struct file* file)
735 file->private_data = (void*)((long)current->self_exec_id);
739 static ssize_t mem_read(struct file * file, char __user * buf,
740 size_t count, loff_t *ppos)
742 struct task_struct *task = proc_task(file->f_dentry->d_inode);
744 unsigned long src = *ppos;
746 struct mm_struct *mm;
748 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
752 page = (char *)__get_free_page(GFP_USER);
758 mm = get_task_mm(task);
764 if (file->private_data != (void*)((long)current->self_exec_id))
770 int this_len, retval;
772 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
773 retval = access_process_vm(task, src, page, this_len, 0);
774 if (!retval || !MAY_PTRACE(task) || !ptrace_may_attach(task)) {
780 if (copy_to_user(buf, page, retval)) {
795 free_page((unsigned long) page);
800 #define mem_write NULL
803 /* This is a security hazard */
804 static ssize_t mem_write(struct file * file, const char * buf,
805 size_t count, loff_t *ppos)
809 struct task_struct *task = proc_task(file->f_dentry->d_inode);
810 unsigned long dst = *ppos;
812 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
815 page = (char *)__get_free_page(GFP_USER);
820 int this_len, retval;
822 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
823 if (copy_from_user(page, buf, this_len)) {
827 retval = access_process_vm(task, dst, page, this_len, 1);
839 free_page((unsigned long) page);
844 static loff_t mem_lseek(struct file * file, loff_t offset, int orig)
848 file->f_pos = offset;
851 file->f_pos += offset;
856 force_successful_syscall_return();
860 static struct file_operations proc_mem_operations = {
867 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
868 size_t count, loff_t *ppos)
870 struct task_struct *task = proc_task(file->f_dentry->d_inode);
873 int oom_adjust = task->oomkilladj;
874 loff_t __ppos = *ppos;
876 len = sprintf(buffer, "%i\n", oom_adjust);
879 if (count > len-__ppos)
881 if (copy_to_user(buf, buffer + __ppos, count))
883 *ppos = __ppos + count;
887 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
888 size_t count, loff_t *ppos)
890 struct task_struct *task = proc_task(file->f_dentry->d_inode);
891 char buffer[8], *end;
894 if (!capable(CAP_SYS_RESOURCE))
896 memset(buffer, 0, 8);
899 if (copy_from_user(buffer, buf, count))
901 oom_adjust = simple_strtol(buffer, &end, 0);
902 if ((oom_adjust < -16 || oom_adjust > 15) && oom_adjust != OOM_DISABLE)
906 task->oomkilladj = oom_adjust;
907 if (end - buffer == 0)
912 static struct file_operations proc_oom_adjust_operations = {
913 .read = oom_adjust_read,
914 .write = oom_adjust_write,
917 #ifdef CONFIG_AUDITSYSCALL
919 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
920 size_t count, loff_t *ppos)
922 struct inode * inode = file->f_dentry->d_inode;
923 struct task_struct *task = proc_task(inode);
925 char tmpbuf[TMPBUFLEN];
927 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
928 audit_get_loginuid(task->audit_context));
929 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
932 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
933 size_t count, loff_t *ppos)
935 struct inode * inode = file->f_dentry->d_inode;
938 struct task_struct *task = proc_task(inode);
941 if (!capable(CAP_AUDIT_CONTROL))
947 if (count >= PAGE_SIZE)
948 count = PAGE_SIZE - 1;
951 /* No partial writes. */
954 page = (char*)__get_free_page(GFP_USER);
958 if (copy_from_user(page, buf, count))
962 loginuid = simple_strtoul(page, &tmp, 10);
968 length = audit_set_loginuid(task, loginuid);
969 if (likely(length == 0))
973 free_page((unsigned long) page);
977 static struct file_operations proc_loginuid_operations = {
978 .read = proc_loginuid_read,
979 .write = proc_loginuid_write,
983 #ifdef CONFIG_SECCOMP
984 static ssize_t seccomp_read(struct file *file, char __user *buf,
985 size_t count, loff_t *ppos)
987 struct task_struct *tsk = proc_task(file->f_dentry->d_inode);
989 loff_t __ppos = *ppos;
992 /* no need to print the trailing zero, so use only len */
993 len = sprintf(__buf, "%u\n", tsk->seccomp.mode);
996 if (count > len - __ppos)
997 count = len - __ppos;
998 if (copy_to_user(buf, __buf + __ppos, count))
1000 *ppos = __ppos + count;
1004 static ssize_t seccomp_write(struct file *file, const char __user *buf,
1005 size_t count, loff_t *ppos)
1007 struct task_struct *tsk = proc_task(file->f_dentry->d_inode);
1008 char __buf[20], *end;
1009 unsigned int seccomp_mode;
1011 /* can set it only once to be even more secure */
1012 if (unlikely(tsk->seccomp.mode))
1015 memset(__buf, 0, sizeof(__buf));
1016 count = min(count, sizeof(__buf) - 1);
1017 if (copy_from_user(__buf, buf, count))
1019 seccomp_mode = simple_strtoul(__buf, &end, 0);
1022 if (seccomp_mode && seccomp_mode <= NR_SECCOMP_MODES) {
1023 tsk->seccomp.mode = seccomp_mode;
1024 set_tsk_thread_flag(tsk, TIF_SECCOMP);
1027 if (unlikely(!(end - __buf)))
1032 static struct file_operations proc_seccomp_operations = {
1033 .read = seccomp_read,
1034 .write = seccomp_write,
1036 #endif /* CONFIG_SECCOMP */
1038 static int proc_check_dentry_visible(struct inode *inode,
1039 struct dentry *dentry, struct vfsmount *mnt)
1041 /* Verify that the current process can already see the
1042 * file pointed at by the file descriptor.
1043 * This prevents /proc from being an accidental information leak.
1045 * This prevents access to files that are not visible do to
1046 * being on the otherside of a chroot, in a different
1047 * namespace, or are simply process local (like pipes).
1049 struct task_struct *task;
1050 struct files_struct *task_files, *files;
1051 int error = -EACCES;
1053 /* See if the the two tasks share a commone set of
1054 * file descriptors. If so everything is visible.
1056 task = proc_task(inode);
1059 files = get_files_struct(current);
1060 task_files = get_files_struct(task);
1061 if (files && task_files && (files == task_files))
1064 put_files_struct(task_files);
1066 put_files_struct(files);
1070 /* If the two tasks don't share a common set of file
1071 * descriptors see if the destination dentry is already
1072 * visible in the current tasks filesystem namespace.
1074 error = proc_check_chroot(dentry, mnt);
1080 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1082 struct inode *inode = dentry->d_inode;
1083 int error = -EACCES;
1085 /* We don't need a base pointer in the /proc filesystem */
1088 if (current->fsuid != inode->i_uid && !capable(CAP_DAC_OVERRIDE))
1091 error = PROC_I(inode)->op.proc_get_link(inode, &nd->dentry, &nd->mnt);
1092 nd->last_type = LAST_BIND;
1096 /* Only return files this task can already see */
1097 error = proc_check_dentry_visible(inode, nd->dentry, nd->mnt);
1101 return ERR_PTR(error);
1104 static int do_proc_readlink(struct dentry *dentry, struct vfsmount *mnt,
1105 char __user *buffer, int buflen)
1107 struct inode * inode;
1108 char *tmp = (char*)__get_free_page(GFP_KERNEL), *path;
1114 inode = dentry->d_inode;
1115 path = d_path(dentry, mnt, tmp, PAGE_SIZE);
1116 len = PTR_ERR(path);
1119 len = tmp + PAGE_SIZE - 1 - path;
1123 if (copy_to_user(buffer, path, len))
1126 free_page((unsigned long)tmp);
1130 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1132 int error = -EACCES;
1133 struct inode *inode = dentry->d_inode;
1135 struct vfsmount *mnt = NULL;
1138 if (current->fsuid != inode->i_uid && !capable(CAP_DAC_OVERRIDE))
1141 error = PROC_I(inode)->op.proc_get_link(inode, &de, &mnt);
1145 /* Only return files this task can already see */
1146 error = proc_check_dentry_visible(inode, de, mnt);
1150 error = do_proc_readlink(de, mnt, buffer, buflen);
1158 static struct inode_operations proc_pid_link_inode_operations = {
1159 .readlink = proc_pid_readlink,
1160 .follow_link = proc_pid_follow_link
1165 static int proc_readfd(struct file * filp, void * dirent, filldir_t filldir)
1167 struct dentry *dentry = filp->f_dentry;
1168 struct inode *inode = dentry->d_inode;
1169 struct task_struct *p = proc_task(inode);
1170 unsigned int fd, tid, ino;
1173 struct files_struct * files;
1174 struct fdtable *fdt;
1185 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1189 ino = parent_ino(dentry);
1190 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1194 files = get_files_struct(p);
1198 fdt = files_fdtable(files);
1199 for (fd = filp->f_pos-2;
1201 fd++, filp->f_pos++) {
1204 if (!fcheck_files(files, fd))
1212 buf[j] = '0' + (i % 10);
1216 ino = fake_ino(tid, PROC_TID_FD_DIR + fd);
1217 if (filldir(dirent, buf+j, NUMBUF-j, fd+2, ino, DT_LNK) < 0) {
1224 put_files_struct(files);
1230 static int proc_pident_readdir(struct file *filp,
1231 void *dirent, filldir_t filldir,
1232 struct pid_entry *ents, unsigned int nents)
1236 struct dentry *dentry = filp->f_dentry;
1237 struct inode *inode = dentry->d_inode;
1238 struct pid_entry *p;
1243 if (!pid_alive(proc_task(inode)))
1247 pid = proc_task(inode)->pid;
1252 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
1258 ino = parent_ino(dentry);
1259 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
1272 if (filldir(dirent, p->name, p->len, filp->f_pos,
1273 fake_ino(pid, p->type), p->mode >> 12) < 0)
1285 static int proc_tgid_base_readdir(struct file * filp,
1286 void * dirent, filldir_t filldir)
1288 return proc_pident_readdir(filp,dirent,filldir,
1289 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1292 static int proc_tid_base_readdir(struct file * filp,
1293 void * dirent, filldir_t filldir)
1295 return proc_pident_readdir(filp,dirent,filldir,
1296 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
1299 /* building an inode */
1301 static int task_dumpable(struct task_struct *task)
1304 struct mm_struct *mm;
1309 dumpable = mm->dumpable;
1317 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task, int ino)
1319 struct inode * inode;
1320 struct proc_inode *ei;
1322 /* We need a new inode */
1324 inode = new_inode(sb);
1330 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1331 inode->i_ino = fake_ino(task->pid, ino);
1333 if (!pid_alive(task))
1337 * grab the reference to task.
1339 get_task_struct(task);
1343 if (task_dumpable(task)) {
1344 inode->i_uid = task->euid;
1345 inode->i_gid = task->egid;
1347 security_task_to_inode(task, inode);
1360 * Exceptional case: normally we are not allowed to unhash a busy
1361 * directory. In this case, however, we can do it - no aliasing problems
1362 * due to the way we treat inodes.
1364 * Rewrite the inode's ownerships here because the owning task may have
1365 * performed a setuid(), etc.
1367 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1369 struct inode *inode = dentry->d_inode;
1370 struct task_struct *task = proc_task(inode);
1371 if (pid_alive(task)) {
1372 if (task_dumpable(task)) {
1373 inode->i_uid = task->euid;
1374 inode->i_gid = task->egid;
1379 security_task_to_inode(task, inode);
1386 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1388 struct inode *inode = dentry->d_inode;
1389 struct task_struct *task = proc_task(inode);
1390 int fd = proc_fd(inode);
1391 struct files_struct *files;
1393 files = get_files_struct(task);
1396 if (fcheck_files(files, fd)) {
1398 put_files_struct(files);
1399 if (task_dumpable(task)) {
1400 inode->i_uid = task->euid;
1401 inode->i_gid = task->egid;
1406 security_task_to_inode(task, inode);
1410 put_files_struct(files);
1416 static void pid_base_iput(struct dentry *dentry, struct inode *inode)
1418 struct task_struct *task = proc_task(inode);
1419 spin_lock(&task->proc_lock);
1420 if (task->proc_dentry == dentry)
1421 task->proc_dentry = NULL;
1422 spin_unlock(&task->proc_lock);
1426 static int pid_delete_dentry(struct dentry * dentry)
1428 /* Is the task we represent dead?
1429 * If so, then don't put the dentry on the lru list,
1430 * kill it immediately.
1432 return !pid_alive(proc_task(dentry->d_inode));
1435 static struct dentry_operations tid_fd_dentry_operations =
1437 .d_revalidate = tid_fd_revalidate,
1438 .d_delete = pid_delete_dentry,
1441 static struct dentry_operations pid_dentry_operations =
1443 .d_revalidate = pid_revalidate,
1444 .d_delete = pid_delete_dentry,
1447 static struct dentry_operations pid_base_dentry_operations =
1449 .d_revalidate = pid_revalidate,
1450 .d_iput = pid_base_iput,
1451 .d_delete = pid_delete_dentry,
1456 static unsigned name_to_int(struct dentry *dentry)
1458 const char *name = dentry->d_name.name;
1459 int len = dentry->d_name.len;
1462 if (len > 1 && *name == '0')
1465 unsigned c = *name++ - '0';
1468 if (n >= (~0U-9)/10)
1479 static struct dentry *proc_lookupfd(struct inode * dir, struct dentry * dentry, struct nameidata *nd)
1481 struct task_struct *task = proc_task(dir);
1482 unsigned fd = name_to_int(dentry);
1484 struct files_struct * files;
1485 struct inode *inode;
1486 struct proc_inode *ei;
1490 if (!pid_alive(task))
1493 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TID_FD_DIR+fd);
1498 files = get_files_struct(task);
1501 inode->i_mode = S_IFLNK;
1504 * We are not taking a ref to the file structure, so we must
1507 spin_lock(&files->file_lock);
1508 file = fcheck_files(files, fd);
1511 if (file->f_mode & 1)
1512 inode->i_mode |= S_IRUSR | S_IXUSR;
1513 if (file->f_mode & 2)
1514 inode->i_mode |= S_IWUSR | S_IXUSR;
1515 spin_unlock(&files->file_lock);
1516 put_files_struct(files);
1517 inode->i_op = &proc_pid_link_inode_operations;
1519 ei->op.proc_get_link = proc_fd_link;
1520 dentry->d_op = &tid_fd_dentry_operations;
1521 d_add(dentry, inode);
1525 spin_unlock(&files->file_lock);
1526 put_files_struct(files);
1530 return ERR_PTR(-ENOENT);
1533 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir);
1534 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd);
1536 static struct file_operations proc_fd_operations = {
1537 .read = generic_read_dir,
1538 .readdir = proc_readfd,
1541 static struct file_operations proc_task_operations = {
1542 .read = generic_read_dir,
1543 .readdir = proc_task_readdir,
1547 * proc directories can do almost nothing..
1549 static struct inode_operations proc_fd_inode_operations = {
1550 .lookup = proc_lookupfd,
1553 static struct inode_operations proc_task_inode_operations = {
1554 .lookup = proc_task_lookup,
1557 #ifdef CONFIG_SECURITY
1558 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
1559 size_t count, loff_t *ppos)
1561 struct inode * inode = file->f_dentry->d_inode;
1564 struct task_struct *task = proc_task(inode);
1566 if (count > PAGE_SIZE)
1568 if (!(page = __get_free_page(GFP_KERNEL)))
1571 length = security_getprocattr(task,
1572 (char*)file->f_dentry->d_name.name,
1573 (void*)page, count);
1575 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
1580 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
1581 size_t count, loff_t *ppos)
1583 struct inode * inode = file->f_dentry->d_inode;
1586 struct task_struct *task = proc_task(inode);
1588 if (count > PAGE_SIZE)
1591 /* No partial writes. */
1594 page = (char*)__get_free_page(GFP_USER);
1598 if (copy_from_user(page, buf, count))
1601 length = security_setprocattr(task,
1602 (char*)file->f_dentry->d_name.name,
1603 (void*)page, count);
1605 free_page((unsigned long) page);
1609 static struct file_operations proc_pid_attr_operations = {
1610 .read = proc_pid_attr_read,
1611 .write = proc_pid_attr_write,
1614 static struct file_operations proc_tid_attr_operations;
1615 static struct inode_operations proc_tid_attr_inode_operations;
1616 static struct file_operations proc_tgid_attr_operations;
1617 static struct inode_operations proc_tgid_attr_inode_operations;
1620 static int get_tid_list(int index, unsigned int *tids, struct inode *dir);
1623 static struct dentry *proc_pident_lookup(struct inode *dir,
1624 struct dentry *dentry,
1625 struct pid_entry *ents)
1627 struct inode *inode;
1629 struct task_struct *task = proc_task(dir);
1630 struct pid_entry *p;
1631 struct proc_inode *ei;
1636 if (!pid_alive(task))
1639 for (p = ents; p->name; p++) {
1640 if (p->len != dentry->d_name.len)
1642 if (!memcmp(dentry->d_name.name, p->name, p->len))
1649 inode = proc_pid_make_inode(dir->i_sb, task, p->type);
1654 inode->i_mode = p->mode;
1656 * Yes, it does not scale. And it should not. Don't add
1657 * new entries into /proc/<tgid>/ without very good reasons.
1660 case PROC_TGID_TASK:
1661 inode->i_nlink = 2 + get_tid_list(2, NULL, dir);
1662 inode->i_op = &proc_task_inode_operations;
1663 inode->i_fop = &proc_task_operations;
1668 inode->i_op = &proc_fd_inode_operations;
1669 inode->i_fop = &proc_fd_operations;
1673 inode->i_op = &proc_pid_link_inode_operations;
1674 ei->op.proc_get_link = proc_exe_link;
1678 inode->i_op = &proc_pid_link_inode_operations;
1679 ei->op.proc_get_link = proc_cwd_link;
1682 case PROC_TGID_ROOT:
1683 inode->i_op = &proc_pid_link_inode_operations;
1684 ei->op.proc_get_link = proc_root_link;
1686 case PROC_TID_ENVIRON:
1687 case PROC_TGID_ENVIRON:
1688 inode->i_fop = &proc_info_file_operations;
1689 ei->op.proc_read = proc_pid_environ;
1692 case PROC_TGID_AUXV:
1693 inode->i_fop = &proc_info_file_operations;
1694 ei->op.proc_read = proc_pid_auxv;
1696 case PROC_TID_STATUS:
1697 case PROC_TGID_STATUS:
1698 inode->i_fop = &proc_info_file_operations;
1699 ei->op.proc_read = proc_pid_status;
1702 inode->i_fop = &proc_info_file_operations;
1703 ei->op.proc_read = proc_tid_stat;
1705 case PROC_TGID_STAT:
1706 inode->i_fop = &proc_info_file_operations;
1707 ei->op.proc_read = proc_tgid_stat;
1709 case PROC_TID_CMDLINE:
1710 case PROC_TGID_CMDLINE:
1711 inode->i_fop = &proc_info_file_operations;
1712 ei->op.proc_read = proc_pid_cmdline;
1714 case PROC_TID_STATM:
1715 case PROC_TGID_STATM:
1716 inode->i_fop = &proc_info_file_operations;
1717 ei->op.proc_read = proc_pid_statm;
1720 case PROC_TGID_MAPS:
1721 inode->i_fop = &proc_maps_operations;
1724 case PROC_TID_NUMA_MAPS:
1725 case PROC_TGID_NUMA_MAPS:
1726 inode->i_fop = &proc_numa_maps_operations;
1731 inode->i_fop = &proc_mem_operations;
1733 #ifdef CONFIG_SECCOMP
1734 case PROC_TID_SECCOMP:
1735 case PROC_TGID_SECCOMP:
1736 inode->i_fop = &proc_seccomp_operations;
1738 #endif /* CONFIG_SECCOMP */
1739 case PROC_TID_MOUNTS:
1740 case PROC_TGID_MOUNTS:
1741 inode->i_fop = &proc_mounts_operations;
1744 case PROC_TID_SMAPS:
1745 case PROC_TGID_SMAPS:
1746 inode->i_fop = &proc_smaps_operations;
1749 case PROC_TID_MOUNTSTATS:
1750 case PROC_TGID_MOUNTSTATS:
1751 inode->i_fop = &proc_mountstats_operations;
1753 #ifdef CONFIG_SECURITY
1756 inode->i_op = &proc_tid_attr_inode_operations;
1757 inode->i_fop = &proc_tid_attr_operations;
1759 case PROC_TGID_ATTR:
1761 inode->i_op = &proc_tgid_attr_inode_operations;
1762 inode->i_fop = &proc_tgid_attr_operations;
1764 case PROC_TID_ATTR_CURRENT:
1765 case PROC_TGID_ATTR_CURRENT:
1766 case PROC_TID_ATTR_PREV:
1767 case PROC_TGID_ATTR_PREV:
1768 case PROC_TID_ATTR_EXEC:
1769 case PROC_TGID_ATTR_EXEC:
1770 case PROC_TID_ATTR_FSCREATE:
1771 case PROC_TGID_ATTR_FSCREATE:
1772 case PROC_TID_ATTR_KEYCREATE:
1773 case PROC_TGID_ATTR_KEYCREATE:
1774 inode->i_fop = &proc_pid_attr_operations;
1777 #ifdef CONFIG_KALLSYMS
1778 case PROC_TID_WCHAN:
1779 case PROC_TGID_WCHAN:
1780 inode->i_fop = &proc_info_file_operations;
1781 ei->op.proc_read = proc_pid_wchan;
1784 #ifdef CONFIG_SCHEDSTATS
1785 case PROC_TID_SCHEDSTAT:
1786 case PROC_TGID_SCHEDSTAT:
1787 inode->i_fop = &proc_info_file_operations;
1788 ei->op.proc_read = proc_pid_schedstat;
1791 #ifdef CONFIG_CPUSETS
1792 case PROC_TID_CPUSET:
1793 case PROC_TGID_CPUSET:
1794 inode->i_fop = &proc_cpuset_operations;
1797 case PROC_TID_OOM_SCORE:
1798 case PROC_TGID_OOM_SCORE:
1799 inode->i_fop = &proc_info_file_operations;
1800 ei->op.proc_read = proc_oom_score;
1802 case PROC_TID_OOM_ADJUST:
1803 case PROC_TGID_OOM_ADJUST:
1804 inode->i_fop = &proc_oom_adjust_operations;
1806 #ifdef CONFIG_AUDITSYSCALL
1807 case PROC_TID_LOGINUID:
1808 case PROC_TGID_LOGINUID:
1809 inode->i_fop = &proc_loginuid_operations;
1813 printk("procfs: impossible type (%d)",p->type);
1815 return ERR_PTR(-EINVAL);
1817 dentry->d_op = &pid_dentry_operations;
1818 d_add(dentry, inode);
1822 return ERR_PTR(error);
1825 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
1826 return proc_pident_lookup(dir, dentry, tgid_base_stuff);
1829 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
1830 return proc_pident_lookup(dir, dentry, tid_base_stuff);
1833 static struct file_operations proc_tgid_base_operations = {
1834 .read = generic_read_dir,
1835 .readdir = proc_tgid_base_readdir,
1838 static struct file_operations proc_tid_base_operations = {
1839 .read = generic_read_dir,
1840 .readdir = proc_tid_base_readdir,
1843 static struct inode_operations proc_tgid_base_inode_operations = {
1844 .lookup = proc_tgid_base_lookup,
1847 static struct inode_operations proc_tid_base_inode_operations = {
1848 .lookup = proc_tid_base_lookup,
1851 #ifdef CONFIG_SECURITY
1852 static int proc_tgid_attr_readdir(struct file * filp,
1853 void * dirent, filldir_t filldir)
1855 return proc_pident_readdir(filp,dirent,filldir,
1856 tgid_attr_stuff,ARRAY_SIZE(tgid_attr_stuff));
1859 static int proc_tid_attr_readdir(struct file * filp,
1860 void * dirent, filldir_t filldir)
1862 return proc_pident_readdir(filp,dirent,filldir,
1863 tid_attr_stuff,ARRAY_SIZE(tid_attr_stuff));
1866 static struct file_operations proc_tgid_attr_operations = {
1867 .read = generic_read_dir,
1868 .readdir = proc_tgid_attr_readdir,
1871 static struct file_operations proc_tid_attr_operations = {
1872 .read = generic_read_dir,
1873 .readdir = proc_tid_attr_readdir,
1876 static struct dentry *proc_tgid_attr_lookup(struct inode *dir,
1877 struct dentry *dentry, struct nameidata *nd)
1879 return proc_pident_lookup(dir, dentry, tgid_attr_stuff);
1882 static struct dentry *proc_tid_attr_lookup(struct inode *dir,
1883 struct dentry *dentry, struct nameidata *nd)
1885 return proc_pident_lookup(dir, dentry, tid_attr_stuff);
1888 static struct inode_operations proc_tgid_attr_inode_operations = {
1889 .lookup = proc_tgid_attr_lookup,
1892 static struct inode_operations proc_tid_attr_inode_operations = {
1893 .lookup = proc_tid_attr_lookup,
1900 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
1904 sprintf(tmp, "%d", current->tgid);
1905 return vfs_readlink(dentry,buffer,buflen,tmp);
1908 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
1911 sprintf(tmp, "%d", current->tgid);
1912 return ERR_PTR(vfs_follow_link(nd,tmp));
1915 static struct inode_operations proc_self_inode_operations = {
1916 .readlink = proc_self_readlink,
1917 .follow_link = proc_self_follow_link,
1921 * proc_pid_unhash - Unhash /proc/@pid entry from the dcache.
1922 * @p: task that should be flushed.
1924 * Drops the /proc/@pid dcache entry from the hash chains.
1926 * Dropping /proc/@pid entries and detach_pid must be synchroneous,
1927 * otherwise e.g. /proc/@pid/exe might point to the wrong executable,
1928 * if the pid value is immediately reused. This is enforced by
1929 * - caller must acquire spin_lock(p->proc_lock)
1930 * - must be called before detach_pid()
1931 * - proc_pid_lookup acquires proc_lock, and checks that
1932 * the target is not dead by looking at the attach count
1936 struct dentry *proc_pid_unhash(struct task_struct *p)
1938 struct dentry *proc_dentry;
1940 proc_dentry = p->proc_dentry;
1941 if (proc_dentry != NULL) {
1943 spin_lock(&dcache_lock);
1944 spin_lock(&proc_dentry->d_lock);
1945 if (!d_unhashed(proc_dentry)) {
1946 dget_locked(proc_dentry);
1947 __d_drop(proc_dentry);
1948 spin_unlock(&proc_dentry->d_lock);
1950 spin_unlock(&proc_dentry->d_lock);
1953 spin_unlock(&dcache_lock);
1959 * proc_pid_flush - recover memory used by stale /proc/@pid/x entries
1960 * @proc_dentry: directoy to prune.
1962 * Shrink the /proc directory that was used by the just killed thread.
1965 void proc_pid_flush(struct dentry *proc_dentry)
1968 if(proc_dentry != NULL) {
1969 shrink_dcache_parent(proc_dentry);
1975 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
1977 struct task_struct *task;
1978 struct inode *inode;
1979 struct proc_inode *ei;
1983 if (dentry->d_name.len == 4 && !memcmp(dentry->d_name.name,"self",4)) {
1984 inode = new_inode(dir->i_sb);
1986 return ERR_PTR(-ENOMEM);
1988 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1989 inode->i_ino = fake_ino(0, PROC_TGID_INO);
1991 inode->i_mode = S_IFLNK|S_IRWXUGO;
1992 inode->i_uid = inode->i_gid = 0;
1994 inode->i_op = &proc_self_inode_operations;
1995 d_add(dentry, inode);
1998 tgid = name_to_int(dentry);
2002 read_lock(&tasklist_lock);
2003 task = find_task_by_pid(tgid);
2005 get_task_struct(task);
2006 read_unlock(&tasklist_lock);
2010 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TGID_INO);
2014 put_task_struct(task);
2017 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2018 inode->i_op = &proc_tgid_base_inode_operations;
2019 inode->i_fop = &proc_tgid_base_operations;
2020 inode->i_flags|=S_IMMUTABLE;
2021 #ifdef CONFIG_SECURITY
2027 dentry->d_op = &pid_base_dentry_operations;
2030 d_add(dentry, inode);
2031 spin_lock(&task->proc_lock);
2032 task->proc_dentry = dentry;
2033 if (!pid_alive(task)) {
2034 dentry = proc_pid_unhash(task);
2037 spin_unlock(&task->proc_lock);
2039 put_task_struct(task);
2041 proc_pid_flush(dentry);
2046 return ERR_PTR(-ENOENT);
2050 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2052 struct task_struct *task;
2053 struct task_struct *leader = proc_task(dir);
2054 struct inode *inode;
2057 tid = name_to_int(dentry);
2061 read_lock(&tasklist_lock);
2062 task = find_task_by_pid(tid);
2064 get_task_struct(task);
2065 read_unlock(&tasklist_lock);
2068 if (leader->tgid != task->tgid)
2071 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TID_INO);
2076 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2077 inode->i_op = &proc_tid_base_inode_operations;
2078 inode->i_fop = &proc_tid_base_operations;
2079 inode->i_flags|=S_IMMUTABLE;
2080 #ifdef CONFIG_SECURITY
2086 dentry->d_op = &pid_base_dentry_operations;
2088 d_add(dentry, inode);
2090 put_task_struct(task);
2093 put_task_struct(task);
2095 return ERR_PTR(-ENOENT);
2098 #define PROC_NUMBUF 10
2099 #define PROC_MAXPIDS 20
2102 * Get a few tgid's to return for filldir - we need to hold the
2103 * tasklist lock while doing this, and we must release it before
2104 * we actually do the filldir itself, so we use a temp buffer..
2106 static int get_tgid_list(int index, unsigned long version, unsigned int *tgids)
2108 struct task_struct *p;
2112 read_lock(&tasklist_lock);
2115 p = find_task_by_pid(version);
2116 if (p && !thread_group_leader(p))
2123 p = next_task(&init_task);
2125 for ( ; p != &init_task; p = next_task(p)) {
2131 tgids[nr_tgids] = tgid;
2133 if (nr_tgids >= PROC_MAXPIDS)
2136 read_unlock(&tasklist_lock);
2141 * Get a few tid's to return for filldir - we need to hold the
2142 * tasklist lock while doing this, and we must release it before
2143 * we actually do the filldir itself, so we use a temp buffer..
2145 static int get_tid_list(int index, unsigned int *tids, struct inode *dir)
2147 struct task_struct *leader_task = proc_task(dir);
2148 struct task_struct *task = leader_task;
2152 read_lock(&tasklist_lock);
2154 * The starting point task (leader_task) might be an already
2155 * unlinked task, which cannot be used to access the task-list
2156 * via next_thread().
2158 if (pid_alive(task)) do {
2159 int tid = task->pid;
2164 tids[nr_tids] = tid;
2166 if (nr_tids >= PROC_MAXPIDS)
2168 } while ((task = next_thread(task)) != leader_task);
2169 read_unlock(&tasklist_lock);
2173 /* for the /proc/ directory itself, after non-process stuff has been done */
2174 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2176 unsigned int tgid_array[PROC_MAXPIDS];
2177 char buf[PROC_NUMBUF];
2178 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2179 unsigned int nr_tgids, i;
2183 ino_t ino = fake_ino(0,PROC_TGID_INO);
2184 if (filldir(dirent, "self", 4, filp->f_pos, ino, DT_LNK) < 0)
2190 /* f_version caches the tgid value that the last readdir call couldn't
2191 * return. lseek aka telldir automagically resets f_version to 0.
2193 next_tgid = filp->f_version;
2194 filp->f_version = 0;
2196 nr_tgids = get_tgid_list(nr, next_tgid, tgid_array);
2198 /* no more entries ! */
2203 /* do not use the last found pid, reserve it for next_tgid */
2204 if (nr_tgids == PROC_MAXPIDS) {
2206 next_tgid = tgid_array[nr_tgids];
2209 for (i=0;i<nr_tgids;i++) {
2210 int tgid = tgid_array[i];
2211 ino_t ino = fake_ino(tgid,PROC_TGID_INO);
2212 unsigned long j = PROC_NUMBUF;
2215 buf[--j] = '0' + (tgid % 10);
2216 while ((tgid /= 10) != 0);
2218 if (filldir(dirent, buf+j, PROC_NUMBUF-j, filp->f_pos, ino, DT_DIR) < 0) {
2219 /* returning this tgid failed, save it as the first
2220 * pid for the next readir call */
2221 filp->f_version = tgid_array[i];
2232 /* for the /proc/TGID/task/ directories */
2233 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
2235 unsigned int tid_array[PROC_MAXPIDS];
2236 char buf[PROC_NUMBUF];
2237 unsigned int nr_tids, i;
2238 struct dentry *dentry = filp->f_dentry;
2239 struct inode *inode = dentry->d_inode;
2240 int retval = -ENOENT;
2242 unsigned long pos = filp->f_pos; /* avoiding "long long" filp->f_pos */
2244 if (!pid_alive(proc_task(inode)))
2251 if (filldir(dirent, ".", 1, pos, ino, DT_DIR) < 0)
2256 ino = parent_ino(dentry);
2257 if (filldir(dirent, "..", 2, pos, ino, DT_DIR) < 0)
2263 nr_tids = get_tid_list(pos, tid_array, inode);
2264 inode->i_nlink = pos + nr_tids;
2266 for (i = 0; i < nr_tids; i++) {
2267 unsigned long j = PROC_NUMBUF;
2268 int tid = tid_array[i];
2270 ino = fake_ino(tid,PROC_TID_INO);
2273 buf[--j] = '0' + (tid % 10);
2274 while ((tid /= 10) != 0);
2276 if (filldir(dirent, buf+j, PROC_NUMBUF-j, pos, ino, DT_DIR) < 0)