4 * Copyright (C) 1991, 1992 Linus Torvalds
6 * proc base directory handling functions
8 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
9 * Instead of using magical inumbers to determine the kind of object
10 * we allocate and fill in-core inodes upon lookup. They don't even
11 * go into icache. We cache the reference to task_struct upon lookup too.
12 * Eventually it should become a filesystem in its own. We don't use the
13 * rest of procfs anymore.
19 * Bruna Moreira <bruna.moreira@indt.org.br>
20 * Edjard Mota <edjard.mota@indt.org.br>
21 * Ilias Biris <ilias.biris@indt.org.br>
22 * Mauricio Lin <mauricio.lin@indt.org.br>
24 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
26 * A new process specific entry (smaps) included in /proc. It shows the
27 * size of rss for each memory area. The maps entry lacks information
28 * about physical memory size (rss) for each mapped file, i.e.,
29 * rss information for executables and library files.
30 * This additional information is useful for any tools that need to know
31 * about physical memory consumption for a process specific library.
35 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36 * Pud inclusion in the page table walking.
40 * 10LE Instituto Nokia de Tecnologia - INdT:
41 * A better way to walks through the page table as suggested by Hugh Dickins.
43 * Simo Piiroinen <simo.piiroinen@nokia.com>:
44 * Smaps information related to shared, private, clean and dirty pages.
46 * Paul Mundt <paul.mundt@nokia.com>:
47 * Overall revision about smaps.
50 #include <asm/uaccess.h>
52 #include <linux/errno.h>
53 #include <linux/time.h>
54 #include <linux/proc_fs.h>
55 #include <linux/stat.h>
56 #include <linux/init.h>
57 #include <linux/capability.h>
58 #include <linux/file.h>
59 #include <linux/string.h>
60 #include <linux/seq_file.h>
61 #include <linux/namei.h>
62 #include <linux/mnt_namespace.h>
64 #include <linux/rcupdate.h>
65 #include <linux/kallsyms.h>
66 #include <linux/module.h>
67 #include <linux/mount.h>
68 #include <linux/security.h>
69 #include <linux/ptrace.h>
70 #include <linux/seccomp.h>
71 #include <linux/cpuset.h>
72 #include <linux/audit.h>
73 #include <linux/poll.h>
74 #include <linux/nsproxy.h>
75 #include <linux/oom.h>
79 * Implementing inode permission operations in /proc is almost
80 * certainly an error. Permission checks need to happen during
81 * each system call not at open time. The reason is that most of
82 * what we wish to check for permissions in /proc varies at runtime.
84 * The classic example of a problem is opening file descriptors
85 * in /proc for a task before it execs a suid executable.
89 /* Worst case buffer size needed for holding an integer. */
90 #define PROC_NUMBUF 13
96 const struct inode_operations *iop;
97 const struct file_operations *fop;
101 #define NOD(NAME, MODE, IOP, FOP, OP) { \
103 .len = sizeof(NAME) - 1, \
110 #define DIR(NAME, MODE, OTYPE) \
111 NOD(NAME, (S_IFDIR|(MODE)), \
112 &proc_##OTYPE##_inode_operations, &proc_##OTYPE##_operations, \
114 #define LNK(NAME, OTYPE) \
115 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
116 &proc_pid_link_inode_operations, NULL, \
117 { .proc_get_link = &proc_##OTYPE##_link } )
118 #define REG(NAME, MODE, OTYPE) \
119 NOD(NAME, (S_IFREG|(MODE)), NULL, \
120 &proc_##OTYPE##_operations, {})
121 #define INF(NAME, MODE, OTYPE) \
122 NOD(NAME, (S_IFREG|(MODE)), \
123 NULL, &proc_info_file_operations, \
124 { .proc_read = &proc_##OTYPE } )
127 EXPORT_SYMBOL(maps_protect);
129 static struct fs_struct *get_fs_struct(struct task_struct *task)
131 struct fs_struct *fs;
135 atomic_inc(&fs->count);
140 static int get_nr_threads(struct task_struct *tsk)
142 /* Must be called with the rcu_read_lock held */
146 if (lock_task_sighand(tsk, &flags)) {
147 count = atomic_read(&tsk->signal->count);
148 unlock_task_sighand(tsk, &flags);
153 static int proc_cwd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
155 struct task_struct *task = get_proc_task(inode);
156 struct fs_struct *fs = NULL;
157 int result = -ENOENT;
160 fs = get_fs_struct(task);
161 put_task_struct(task);
164 read_lock(&fs->lock);
165 *mnt = mntget(fs->pwdmnt);
166 *dentry = dget(fs->pwd);
167 read_unlock(&fs->lock);
174 static int proc_root_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
176 struct task_struct *task = get_proc_task(inode);
177 struct fs_struct *fs = NULL;
178 int result = -ENOENT;
181 fs = get_fs_struct(task);
182 put_task_struct(task);
185 read_lock(&fs->lock);
186 *mnt = mntget(fs->rootmnt);
187 *dentry = dget(fs->root);
188 read_unlock(&fs->lock);
195 #define MAY_PTRACE(task) \
196 (task == current || \
197 (task->parent == current && \
198 (task->ptrace & PT_PTRACED) && \
199 (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
200 security_ptrace(current,task) == 0))
202 static int proc_pid_environ(struct task_struct *task, char * buffer)
205 struct mm_struct *mm = get_task_mm(task);
207 unsigned int len = mm->env_end - mm->env_start;
210 res = access_process_vm(task, mm->env_start, buffer, len, 0);
211 if (!ptrace_may_attach(task))
218 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
222 struct mm_struct *mm = get_task_mm(task);
226 goto out_mm; /* Shh! No looking before we're done */
228 len = mm->arg_end - mm->arg_start;
233 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
235 // If the nul at the end of args has been overwritten, then
236 // assume application is using setproctitle(3).
237 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
238 len = strnlen(buffer, res);
242 len = mm->env_end - mm->env_start;
243 if (len > PAGE_SIZE - res)
244 len = PAGE_SIZE - res;
245 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
246 res = strnlen(buffer, res);
255 static int proc_pid_auxv(struct task_struct *task, char *buffer)
258 struct mm_struct *mm = get_task_mm(task);
260 unsigned int nwords = 0;
263 while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
264 res = nwords * sizeof(mm->saved_auxv[0]);
267 memcpy(buffer, mm->saved_auxv, res);
274 #ifdef CONFIG_KALLSYMS
276 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
277 * Returns the resolved symbol. If that fails, simply return the address.
279 static int proc_pid_wchan(struct task_struct *task, char *buffer)
282 char symname[KSYM_NAME_LEN+1];
284 wchan = get_wchan(task);
286 if (lookup_symbol_name(wchan, symname) < 0)
287 return sprintf(buffer, "%lu", wchan);
289 return sprintf(buffer, "%s", symname);
291 #endif /* CONFIG_KALLSYMS */
293 #ifdef CONFIG_SCHEDSTATS
295 * Provides /proc/PID/schedstat
297 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
299 return sprintf(buffer, "%lu %lu %lu\n",
300 task->sched_info.cpu_time,
301 task->sched_info.run_delay,
302 task->sched_info.pcnt);
306 /* The badness from the OOM killer */
307 unsigned long badness(struct task_struct *p, unsigned long uptime);
308 static int proc_oom_score(struct task_struct *task, char *buffer)
310 unsigned long points;
311 struct timespec uptime;
313 do_posix_clock_monotonic_gettime(&uptime);
314 read_lock(&tasklist_lock);
315 points = badness(task, uptime.tv_sec);
316 read_unlock(&tasklist_lock);
317 return sprintf(buffer, "%lu\n", points);
320 /************************************************************************/
321 /* Here the fs part begins */
322 /************************************************************************/
324 /* permission checks */
325 static int proc_fd_access_allowed(struct inode *inode)
327 struct task_struct *task;
329 /* Allow access to a task's file descriptors if it is us or we
330 * may use ptrace attach to the process and find out that
333 task = get_proc_task(inode);
335 allowed = ptrace_may_attach(task);
336 put_task_struct(task);
341 static int proc_setattr(struct dentry *dentry, struct iattr *attr)
344 struct inode *inode = dentry->d_inode;
346 if (attr->ia_valid & ATTR_MODE)
349 error = inode_change_ok(inode, attr);
351 error = inode_setattr(inode, attr);
355 static const struct inode_operations proc_def_inode_operations = {
356 .setattr = proc_setattr,
359 extern struct seq_operations mounts_op;
365 static int mounts_open(struct inode *inode, struct file *file)
367 struct task_struct *task = get_proc_task(inode);
368 struct mnt_namespace *ns = NULL;
369 struct proc_mounts *p;
375 ns = task->nsproxy->mnt_ns;
380 put_task_struct(task);
385 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
387 file->private_data = &p->m;
388 ret = seq_open(file, &mounts_op);
391 p->event = ns->event;
401 static int mounts_release(struct inode *inode, struct file *file)
403 struct seq_file *m = file->private_data;
404 struct mnt_namespace *ns = m->private;
406 return seq_release(inode, file);
409 static unsigned mounts_poll(struct file *file, poll_table *wait)
411 struct proc_mounts *p = file->private_data;
412 struct mnt_namespace *ns = p->m.private;
415 poll_wait(file, &ns->poll, wait);
417 spin_lock(&vfsmount_lock);
418 if (p->event != ns->event) {
419 p->event = ns->event;
422 spin_unlock(&vfsmount_lock);
427 static const struct file_operations proc_mounts_operations = {
431 .release = mounts_release,
435 extern struct seq_operations mountstats_op;
436 static int mountstats_open(struct inode *inode, struct file *file)
438 int ret = seq_open(file, &mountstats_op);
441 struct seq_file *m = file->private_data;
442 struct mnt_namespace *mnt_ns = NULL;
443 struct task_struct *task = get_proc_task(inode);
448 mnt_ns = task->nsproxy->mnt_ns;
452 put_task_struct(task);
458 seq_release(inode, file);
465 static const struct file_operations proc_mountstats_operations = {
466 .open = mountstats_open,
469 .release = mounts_release,
472 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
474 static ssize_t proc_info_read(struct file * file, char __user * buf,
475 size_t count, loff_t *ppos)
477 struct inode * inode = file->f_path.dentry->d_inode;
480 struct task_struct *task = get_proc_task(inode);
486 if (count > PROC_BLOCK_SIZE)
487 count = PROC_BLOCK_SIZE;
490 if (!(page = __get_free_page(GFP_KERNEL)))
493 length = PROC_I(inode)->op.proc_read(task, (char*)page);
496 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
499 put_task_struct(task);
504 static const struct file_operations proc_info_file_operations = {
505 .read = proc_info_read,
508 static int mem_open(struct inode* inode, struct file* file)
510 file->private_data = (void*)((long)current->self_exec_id);
514 static ssize_t mem_read(struct file * file, char __user * buf,
515 size_t count, loff_t *ppos)
517 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
519 unsigned long src = *ppos;
521 struct mm_struct *mm;
526 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
530 page = (char *)__get_free_page(GFP_USER);
536 mm = get_task_mm(task);
542 if (file->private_data != (void*)((long)current->self_exec_id))
548 int this_len, retval;
550 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
551 retval = access_process_vm(task, src, page, this_len, 0);
552 if (!retval || !MAY_PTRACE(task) || !ptrace_may_attach(task)) {
558 if (copy_to_user(buf, page, retval)) {
573 free_page((unsigned long) page);
575 put_task_struct(task);
580 #define mem_write NULL
583 /* This is a security hazard */
584 static ssize_t mem_write(struct file * file, const char __user *buf,
585 size_t count, loff_t *ppos)
589 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
590 unsigned long dst = *ppos;
596 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
600 page = (char *)__get_free_page(GFP_USER);
606 int this_len, retval;
608 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
609 if (copy_from_user(page, buf, this_len)) {
613 retval = access_process_vm(task, dst, page, this_len, 1);
625 free_page((unsigned long) page);
627 put_task_struct(task);
633 static loff_t mem_lseek(struct file * file, loff_t offset, int orig)
637 file->f_pos = offset;
640 file->f_pos += offset;
645 force_successful_syscall_return();
649 static const struct file_operations proc_mem_operations = {
656 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
657 size_t count, loff_t *ppos)
659 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
660 char buffer[PROC_NUMBUF];
666 oom_adjust = task->oomkilladj;
667 put_task_struct(task);
669 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
671 return simple_read_from_buffer(buf, count, ppos, buffer, len);
674 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
675 size_t count, loff_t *ppos)
677 struct task_struct *task;
678 char buffer[PROC_NUMBUF], *end;
681 memset(buffer, 0, sizeof(buffer));
682 if (count > sizeof(buffer) - 1)
683 count = sizeof(buffer) - 1;
684 if (copy_from_user(buffer, buf, count))
686 oom_adjust = simple_strtol(buffer, &end, 0);
687 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
688 oom_adjust != OOM_DISABLE)
692 task = get_proc_task(file->f_path.dentry->d_inode);
695 if (oom_adjust < task->oomkilladj && !capable(CAP_SYS_RESOURCE)) {
696 put_task_struct(task);
699 task->oomkilladj = oom_adjust;
700 put_task_struct(task);
701 if (end - buffer == 0)
706 static const struct file_operations proc_oom_adjust_operations = {
707 .read = oom_adjust_read,
708 .write = oom_adjust_write,
711 static ssize_t clear_refs_write(struct file *file, const char __user *buf,
712 size_t count, loff_t *ppos)
714 struct task_struct *task;
715 char buffer[PROC_NUMBUF], *end;
716 struct mm_struct *mm;
718 memset(buffer, 0, sizeof(buffer));
719 if (count > sizeof(buffer) - 1)
720 count = sizeof(buffer) - 1;
721 if (copy_from_user(buffer, buf, count))
723 if (!simple_strtol(buffer, &end, 0))
727 task = get_proc_task(file->f_path.dentry->d_inode);
730 mm = get_task_mm(task);
735 put_task_struct(task);
736 if (end - buffer == 0)
741 static struct file_operations proc_clear_refs_operations = {
742 .write = clear_refs_write,
745 #ifdef CONFIG_AUDITSYSCALL
747 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
748 size_t count, loff_t *ppos)
750 struct inode * inode = file->f_path.dentry->d_inode;
751 struct task_struct *task = get_proc_task(inode);
753 char tmpbuf[TMPBUFLEN];
757 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
758 audit_get_loginuid(task->audit_context));
759 put_task_struct(task);
760 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
763 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
764 size_t count, loff_t *ppos)
766 struct inode * inode = file->f_path.dentry->d_inode;
771 if (!capable(CAP_AUDIT_CONTROL))
774 if (current != pid_task(proc_pid(inode), PIDTYPE_PID))
777 if (count >= PAGE_SIZE)
778 count = PAGE_SIZE - 1;
781 /* No partial writes. */
784 page = (char*)__get_free_page(GFP_USER);
788 if (copy_from_user(page, buf, count))
792 loginuid = simple_strtoul(page, &tmp, 10);
798 length = audit_set_loginuid(current, loginuid);
799 if (likely(length == 0))
803 free_page((unsigned long) page);
807 static const struct file_operations proc_loginuid_operations = {
808 .read = proc_loginuid_read,
809 .write = proc_loginuid_write,
813 #ifdef CONFIG_SECCOMP
814 static ssize_t seccomp_read(struct file *file, char __user *buf,
815 size_t count, loff_t *ppos)
817 struct task_struct *tsk = get_proc_task(file->f_dentry->d_inode);
823 /* no need to print the trailing zero, so use only len */
824 len = sprintf(__buf, "%u\n", tsk->seccomp.mode);
825 put_task_struct(tsk);
827 return simple_read_from_buffer(buf, count, ppos, __buf, len);
830 static ssize_t seccomp_write(struct file *file, const char __user *buf,
831 size_t count, loff_t *ppos)
833 struct task_struct *tsk = get_proc_task(file->f_dentry->d_inode);
834 char __buf[20], *end;
835 unsigned int seccomp_mode;
842 /* can set it only once to be even more secure */
844 if (unlikely(tsk->seccomp.mode))
848 memset(__buf, 0, sizeof(__buf));
849 count = min(count, sizeof(__buf) - 1);
850 if (copy_from_user(__buf, buf, count))
853 seccomp_mode = simple_strtoul(__buf, &end, 0);
857 if (seccomp_mode && seccomp_mode <= NR_SECCOMP_MODES) {
858 tsk->seccomp.mode = seccomp_mode;
859 set_tsk_thread_flag(tsk, TIF_SECCOMP);
863 if (unlikely(!(end - __buf)))
865 result = end - __buf;
867 put_task_struct(tsk);
872 static const struct file_operations proc_seccomp_operations = {
873 .read = seccomp_read,
874 .write = seccomp_write,
876 #endif /* CONFIG_SECCOMP */
878 #ifdef CONFIG_FAULT_INJECTION
879 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
880 size_t count, loff_t *ppos)
882 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
883 char buffer[PROC_NUMBUF];
889 make_it_fail = task->make_it_fail;
890 put_task_struct(task);
892 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
894 return simple_read_from_buffer(buf, count, ppos, buffer, len);
897 static ssize_t proc_fault_inject_write(struct file * file,
898 const char __user * buf, size_t count, loff_t *ppos)
900 struct task_struct *task;
901 char buffer[PROC_NUMBUF], *end;
904 if (!capable(CAP_SYS_RESOURCE))
906 memset(buffer, 0, sizeof(buffer));
907 if (count > sizeof(buffer) - 1)
908 count = sizeof(buffer) - 1;
909 if (copy_from_user(buffer, buf, count))
911 make_it_fail = simple_strtol(buffer, &end, 0);
914 task = get_proc_task(file->f_dentry->d_inode);
917 task->make_it_fail = make_it_fail;
918 put_task_struct(task);
919 if (end - buffer == 0)
924 static const struct file_operations proc_fault_inject_operations = {
925 .read = proc_fault_inject_read,
926 .write = proc_fault_inject_write,
930 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
932 struct inode *inode = dentry->d_inode;
935 /* We don't need a base pointer in the /proc filesystem */
938 /* Are we allowed to snoop on the tasks file descriptors? */
939 if (!proc_fd_access_allowed(inode))
942 error = PROC_I(inode)->op.proc_get_link(inode, &nd->dentry, &nd->mnt);
943 nd->last_type = LAST_BIND;
945 return ERR_PTR(error);
948 static int do_proc_readlink(struct dentry *dentry, struct vfsmount *mnt,
949 char __user *buffer, int buflen)
951 struct inode * inode;
952 char *tmp = (char*)__get_free_page(GFP_KERNEL), *path;
958 inode = dentry->d_inode;
959 path = d_path(dentry, mnt, tmp, PAGE_SIZE);
963 len = tmp + PAGE_SIZE - 1 - path;
967 if (copy_to_user(buffer, path, len))
970 free_page((unsigned long)tmp);
974 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
977 struct inode *inode = dentry->d_inode;
979 struct vfsmount *mnt = NULL;
981 /* Are we allowed to snoop on the tasks file descriptors? */
982 if (!proc_fd_access_allowed(inode))
985 error = PROC_I(inode)->op.proc_get_link(inode, &de, &mnt);
989 error = do_proc_readlink(de, mnt, buffer, buflen);
996 static const struct inode_operations proc_pid_link_inode_operations = {
997 .readlink = proc_pid_readlink,
998 .follow_link = proc_pid_follow_link,
999 .setattr = proc_setattr,
1003 /* building an inode */
1005 static int task_dumpable(struct task_struct *task)
1008 struct mm_struct *mm;
1013 dumpable = mm->dumpable;
1021 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1023 struct inode * inode;
1024 struct proc_inode *ei;
1026 /* We need a new inode */
1028 inode = new_inode(sb);
1034 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1035 inode->i_op = &proc_def_inode_operations;
1038 * grab the reference to task.
1040 ei->pid = get_task_pid(task, PIDTYPE_PID);
1046 if (task_dumpable(task)) {
1047 inode->i_uid = task->euid;
1048 inode->i_gid = task->egid;
1050 security_task_to_inode(task, inode);
1060 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1062 struct inode *inode = dentry->d_inode;
1063 struct task_struct *task;
1064 generic_fillattr(inode, stat);
1069 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1071 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1072 task_dumpable(task)) {
1073 stat->uid = task->euid;
1074 stat->gid = task->egid;
1084 * Exceptional case: normally we are not allowed to unhash a busy
1085 * directory. In this case, however, we can do it - no aliasing problems
1086 * due to the way we treat inodes.
1088 * Rewrite the inode's ownerships here because the owning task may have
1089 * performed a setuid(), etc.
1091 * Before the /proc/pid/status file was created the only way to read
1092 * the effective uid of a /process was to stat /proc/pid. Reading
1093 * /proc/pid/status is slow enough that procps and other packages
1094 * kept stating /proc/pid. To keep the rules in /proc simple I have
1095 * made this apply to all per process world readable and executable
1098 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1100 struct inode *inode = dentry->d_inode;
1101 struct task_struct *task = get_proc_task(inode);
1103 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1104 task_dumpable(task)) {
1105 inode->i_uid = task->euid;
1106 inode->i_gid = task->egid;
1111 inode->i_mode &= ~(S_ISUID | S_ISGID);
1112 security_task_to_inode(task, inode);
1113 put_task_struct(task);
1120 static int pid_delete_dentry(struct dentry * dentry)
1122 /* Is the task we represent dead?
1123 * If so, then don't put the dentry on the lru list,
1124 * kill it immediately.
1126 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1129 static struct dentry_operations pid_dentry_operations =
1131 .d_revalidate = pid_revalidate,
1132 .d_delete = pid_delete_dentry,
1137 typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1138 struct task_struct *, const void *);
1141 * Fill a directory entry.
1143 * If possible create the dcache entry and derive our inode number and
1144 * file type from dcache entry.
1146 * Since all of the proc inode numbers are dynamically generated, the inode
1147 * numbers do not exist until the inode is cache. This means creating the
1148 * the dcache entry in readdir is necessary to keep the inode numbers
1149 * reported by readdir in sync with the inode numbers reported
1152 static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1153 char *name, int len,
1154 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1156 struct dentry *child, *dir = filp->f_path.dentry;
1157 struct inode *inode;
1160 unsigned type = DT_UNKNOWN;
1164 qname.hash = full_name_hash(name, len);
1166 child = d_lookup(dir, &qname);
1169 new = d_alloc(dir, &qname);
1171 child = instantiate(dir->d_inode, new, task, ptr);
1178 if (!child || IS_ERR(child) || !child->d_inode)
1179 goto end_instantiate;
1180 inode = child->d_inode;
1183 type = inode->i_mode >> 12;
1188 ino = find_inode_number(dir, &qname);
1191 return filldir(dirent, name, len, filp->f_pos, ino, type);
1194 static unsigned name_to_int(struct dentry *dentry)
1196 const char *name = dentry->d_name.name;
1197 int len = dentry->d_name.len;
1200 if (len > 1 && *name == '0')
1203 unsigned c = *name++ - '0';
1206 if (n >= (~0U-9)/10)
1216 #define PROC_FDINFO_MAX 64
1218 static int proc_fd_info(struct inode *inode, struct dentry **dentry,
1219 struct vfsmount **mnt, char *info)
1221 struct task_struct *task = get_proc_task(inode);
1222 struct files_struct *files = NULL;
1224 int fd = proc_fd(inode);
1227 files = get_files_struct(task);
1228 put_task_struct(task);
1232 * We are not taking a ref to the file structure, so we must
1235 spin_lock(&files->file_lock);
1236 file = fcheck_files(files, fd);
1239 *mnt = mntget(file->f_path.mnt);
1241 *dentry = dget(file->f_path.dentry);
1243 snprintf(info, PROC_FDINFO_MAX,
1246 (long long) file->f_pos,
1248 spin_unlock(&files->file_lock);
1249 put_files_struct(files);
1252 spin_unlock(&files->file_lock);
1253 put_files_struct(files);
1258 static int proc_fd_link(struct inode *inode, struct dentry **dentry,
1259 struct vfsmount **mnt)
1261 return proc_fd_info(inode, dentry, mnt, NULL);
1264 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1266 struct inode *inode = dentry->d_inode;
1267 struct task_struct *task = get_proc_task(inode);
1268 int fd = proc_fd(inode);
1269 struct files_struct *files;
1272 files = get_files_struct(task);
1275 if (fcheck_files(files, fd)) {
1277 put_files_struct(files);
1278 if (task_dumpable(task)) {
1279 inode->i_uid = task->euid;
1280 inode->i_gid = task->egid;
1285 inode->i_mode &= ~(S_ISUID | S_ISGID);
1286 security_task_to_inode(task, inode);
1287 put_task_struct(task);
1291 put_files_struct(files);
1293 put_task_struct(task);
1299 static struct dentry_operations tid_fd_dentry_operations =
1301 .d_revalidate = tid_fd_revalidate,
1302 .d_delete = pid_delete_dentry,
1305 static struct dentry *proc_fd_instantiate(struct inode *dir,
1306 struct dentry *dentry, struct task_struct *task, const void *ptr)
1308 unsigned fd = *(const unsigned *)ptr;
1310 struct files_struct *files;
1311 struct inode *inode;
1312 struct proc_inode *ei;
1313 struct dentry *error = ERR_PTR(-ENOENT);
1315 inode = proc_pid_make_inode(dir->i_sb, task);
1320 files = get_files_struct(task);
1323 inode->i_mode = S_IFLNK;
1326 * We are not taking a ref to the file structure, so we must
1329 spin_lock(&files->file_lock);
1330 file = fcheck_files(files, fd);
1333 if (file->f_mode & 1)
1334 inode->i_mode |= S_IRUSR | S_IXUSR;
1335 if (file->f_mode & 2)
1336 inode->i_mode |= S_IWUSR | S_IXUSR;
1337 spin_unlock(&files->file_lock);
1338 put_files_struct(files);
1340 inode->i_op = &proc_pid_link_inode_operations;
1342 ei->op.proc_get_link = proc_fd_link;
1343 dentry->d_op = &tid_fd_dentry_operations;
1344 d_add(dentry, inode);
1345 /* Close the race of the process dying before we return the dentry */
1346 if (tid_fd_revalidate(dentry, NULL))
1352 spin_unlock(&files->file_lock);
1353 put_files_struct(files);
1359 static struct dentry *proc_lookupfd_common(struct inode *dir,
1360 struct dentry *dentry,
1361 instantiate_t instantiate)
1363 struct task_struct *task = get_proc_task(dir);
1364 unsigned fd = name_to_int(dentry);
1365 struct dentry *result = ERR_PTR(-ENOENT);
1372 result = instantiate(dir, dentry, task, &fd);
1374 put_task_struct(task);
1379 static int proc_readfd_common(struct file * filp, void * dirent,
1380 filldir_t filldir, instantiate_t instantiate)
1382 struct dentry *dentry = filp->f_path.dentry;
1383 struct inode *inode = dentry->d_inode;
1384 struct task_struct *p = get_proc_task(inode);
1385 unsigned int fd, tid, ino;
1387 struct files_struct * files;
1388 struct fdtable *fdt;
1399 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1403 ino = parent_ino(dentry);
1404 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1408 files = get_files_struct(p);
1412 fdt = files_fdtable(files);
1413 for (fd = filp->f_pos-2;
1415 fd++, filp->f_pos++) {
1416 char name[PROC_NUMBUF];
1419 if (!fcheck_files(files, fd))
1423 len = snprintf(name, sizeof(name), "%d", fd);
1424 if (proc_fill_cache(filp, dirent, filldir,
1425 name, len, instantiate,
1433 put_files_struct(files);
1441 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1442 struct nameidata *nd)
1444 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1447 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1449 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1452 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1453 size_t len, loff_t *ppos)
1455 char tmp[PROC_FDINFO_MAX];
1456 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, NULL, tmp);
1458 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1462 static const struct file_operations proc_fdinfo_file_operations = {
1463 .open = nonseekable_open,
1464 .read = proc_fdinfo_read,
1467 static const struct file_operations proc_fd_operations = {
1468 .read = generic_read_dir,
1469 .readdir = proc_readfd,
1473 * /proc/pid/fd needs a special permission handler so that a process can still
1474 * access /proc/self/fd after it has executed a setuid().
1476 static int proc_fd_permission(struct inode *inode, int mask,
1477 struct nameidata *nd)
1481 rv = generic_permission(inode, mask, NULL);
1484 if (task_pid(current) == proc_pid(inode))
1490 * proc directories can do almost nothing..
1492 static const struct inode_operations proc_fd_inode_operations = {
1493 .lookup = proc_lookupfd,
1494 .permission = proc_fd_permission,
1495 .setattr = proc_setattr,
1498 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
1499 struct dentry *dentry, struct task_struct *task, const void *ptr)
1501 unsigned fd = *(unsigned *)ptr;
1502 struct inode *inode;
1503 struct proc_inode *ei;
1504 struct dentry *error = ERR_PTR(-ENOENT);
1506 inode = proc_pid_make_inode(dir->i_sb, task);
1511 inode->i_mode = S_IFREG | S_IRUSR;
1512 inode->i_fop = &proc_fdinfo_file_operations;
1513 dentry->d_op = &tid_fd_dentry_operations;
1514 d_add(dentry, inode);
1515 /* Close the race of the process dying before we return the dentry */
1516 if (tid_fd_revalidate(dentry, NULL))
1523 static struct dentry *proc_lookupfdinfo(struct inode *dir,
1524 struct dentry *dentry,
1525 struct nameidata *nd)
1527 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
1530 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
1532 return proc_readfd_common(filp, dirent, filldir,
1533 proc_fdinfo_instantiate);
1536 static const struct file_operations proc_fdinfo_operations = {
1537 .read = generic_read_dir,
1538 .readdir = proc_readfdinfo,
1542 * proc directories can do almost nothing..
1544 static const struct inode_operations proc_fdinfo_inode_operations = {
1545 .lookup = proc_lookupfdinfo,
1546 .setattr = proc_setattr,
1550 static struct dentry *proc_pident_instantiate(struct inode *dir,
1551 struct dentry *dentry, struct task_struct *task, const void *ptr)
1553 const struct pid_entry *p = ptr;
1554 struct inode *inode;
1555 struct proc_inode *ei;
1556 struct dentry *error = ERR_PTR(-EINVAL);
1558 inode = proc_pid_make_inode(dir->i_sb, task);
1563 inode->i_mode = p->mode;
1564 if (S_ISDIR(inode->i_mode))
1565 inode->i_nlink = 2; /* Use getattr to fix if necessary */
1567 inode->i_op = p->iop;
1569 inode->i_fop = p->fop;
1571 dentry->d_op = &pid_dentry_operations;
1572 d_add(dentry, inode);
1573 /* Close the race of the process dying before we return the dentry */
1574 if (pid_revalidate(dentry, NULL))
1580 static struct dentry *proc_pident_lookup(struct inode *dir,
1581 struct dentry *dentry,
1582 const struct pid_entry *ents,
1585 struct inode *inode;
1586 struct dentry *error;
1587 struct task_struct *task = get_proc_task(dir);
1588 const struct pid_entry *p, *last;
1590 error = ERR_PTR(-ENOENT);
1597 * Yes, it does not scale. And it should not. Don't add
1598 * new entries into /proc/<tgid>/ without very good reasons.
1600 last = &ents[nents - 1];
1601 for (p = ents; p <= last; p++) {
1602 if (p->len != dentry->d_name.len)
1604 if (!memcmp(dentry->d_name.name, p->name, p->len))
1610 error = proc_pident_instantiate(dir, dentry, task, p);
1612 put_task_struct(task);
1617 static int proc_pident_fill_cache(struct file *filp, void *dirent,
1618 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
1620 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
1621 proc_pident_instantiate, task, p);
1624 static int proc_pident_readdir(struct file *filp,
1625 void *dirent, filldir_t filldir,
1626 const struct pid_entry *ents, unsigned int nents)
1630 struct dentry *dentry = filp->f_path.dentry;
1631 struct inode *inode = dentry->d_inode;
1632 struct task_struct *task = get_proc_task(inode);
1633 const struct pid_entry *p, *last;
1647 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
1653 ino = parent_ino(dentry);
1654 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
1666 last = &ents[nents - 1];
1668 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
1677 put_task_struct(task);
1682 #ifdef CONFIG_SECURITY
1683 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
1684 size_t count, loff_t *ppos)
1686 struct inode * inode = file->f_path.dentry->d_inode;
1689 struct task_struct *task = get_proc_task(inode);
1694 length = security_getprocattr(task,
1695 (char*)file->f_path.dentry->d_name.name,
1697 put_task_struct(task);
1699 length = simple_read_from_buffer(buf, count, ppos, p, length);
1704 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
1705 size_t count, loff_t *ppos)
1707 struct inode * inode = file->f_path.dentry->d_inode;
1710 struct task_struct *task = get_proc_task(inode);
1715 if (count > PAGE_SIZE)
1718 /* No partial writes. */
1724 page = (char*)__get_free_page(GFP_USER);
1729 if (copy_from_user(page, buf, count))
1732 length = security_setprocattr(task,
1733 (char*)file->f_path.dentry->d_name.name,
1734 (void*)page, count);
1736 free_page((unsigned long) page);
1738 put_task_struct(task);
1743 static const struct file_operations proc_pid_attr_operations = {
1744 .read = proc_pid_attr_read,
1745 .write = proc_pid_attr_write,
1748 static const struct pid_entry attr_dir_stuff[] = {
1749 REG("current", S_IRUGO|S_IWUGO, pid_attr),
1750 REG("prev", S_IRUGO, pid_attr),
1751 REG("exec", S_IRUGO|S_IWUGO, pid_attr),
1752 REG("fscreate", S_IRUGO|S_IWUGO, pid_attr),
1753 REG("keycreate", S_IRUGO|S_IWUGO, pid_attr),
1754 REG("sockcreate", S_IRUGO|S_IWUGO, pid_attr),
1757 static int proc_attr_dir_readdir(struct file * filp,
1758 void * dirent, filldir_t filldir)
1760 return proc_pident_readdir(filp,dirent,filldir,
1761 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
1764 static const struct file_operations proc_attr_dir_operations = {
1765 .read = generic_read_dir,
1766 .readdir = proc_attr_dir_readdir,
1769 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
1770 struct dentry *dentry, struct nameidata *nd)
1772 return proc_pident_lookup(dir, dentry,
1773 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
1776 static const struct inode_operations proc_attr_dir_inode_operations = {
1777 .lookup = proc_attr_dir_lookup,
1778 .getattr = pid_getattr,
1779 .setattr = proc_setattr,
1787 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
1790 char tmp[PROC_NUMBUF];
1791 sprintf(tmp, "%d", current->tgid);
1792 return vfs_readlink(dentry,buffer,buflen,tmp);
1795 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
1797 char tmp[PROC_NUMBUF];
1798 sprintf(tmp, "%d", current->tgid);
1799 return ERR_PTR(vfs_follow_link(nd,tmp));
1802 static const struct inode_operations proc_self_inode_operations = {
1803 .readlink = proc_self_readlink,
1804 .follow_link = proc_self_follow_link,
1810 * These are the directory entries in the root directory of /proc
1811 * that properly belong to the /proc filesystem, as they describe
1812 * describe something that is process related.
1814 static const struct pid_entry proc_base_stuff[] = {
1815 NOD("self", S_IFLNK|S_IRWXUGO,
1816 &proc_self_inode_operations, NULL, {}),
1820 * Exceptional case: normally we are not allowed to unhash a busy
1821 * directory. In this case, however, we can do it - no aliasing problems
1822 * due to the way we treat inodes.
1824 static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
1826 struct inode *inode = dentry->d_inode;
1827 struct task_struct *task = get_proc_task(inode);
1829 put_task_struct(task);
1836 static struct dentry_operations proc_base_dentry_operations =
1838 .d_revalidate = proc_base_revalidate,
1839 .d_delete = pid_delete_dentry,
1842 static struct dentry *proc_base_instantiate(struct inode *dir,
1843 struct dentry *dentry, struct task_struct *task, const void *ptr)
1845 const struct pid_entry *p = ptr;
1846 struct inode *inode;
1847 struct proc_inode *ei;
1848 struct dentry *error = ERR_PTR(-EINVAL);
1850 /* Allocate the inode */
1851 error = ERR_PTR(-ENOMEM);
1852 inode = new_inode(dir->i_sb);
1856 /* Initialize the inode */
1858 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1861 * grab the reference to the task.
1863 ei->pid = get_task_pid(task, PIDTYPE_PID);
1869 inode->i_mode = p->mode;
1870 if (S_ISDIR(inode->i_mode))
1872 if (S_ISLNK(inode->i_mode))
1875 inode->i_op = p->iop;
1877 inode->i_fop = p->fop;
1879 dentry->d_op = &proc_base_dentry_operations;
1880 d_add(dentry, inode);
1889 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
1891 struct dentry *error;
1892 struct task_struct *task = get_proc_task(dir);
1893 const struct pid_entry *p, *last;
1895 error = ERR_PTR(-ENOENT);
1900 /* Lookup the directory entry */
1901 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
1902 for (p = proc_base_stuff; p <= last; p++) {
1903 if (p->len != dentry->d_name.len)
1905 if (!memcmp(dentry->d_name.name, p->name, p->len))
1911 error = proc_base_instantiate(dir, dentry, task, p);
1914 put_task_struct(task);
1919 static int proc_base_fill_cache(struct file *filp, void *dirent,
1920 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
1922 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
1923 proc_base_instantiate, task, p);
1926 #ifdef CONFIG_TASK_IO_ACCOUNTING
1927 static int proc_pid_io_accounting(struct task_struct *task, char *buffer)
1929 return sprintf(buffer,
1930 #ifdef CONFIG_TASK_XACCT
1936 "read_bytes: %llu\n"
1937 "write_bytes: %llu\n"
1938 "cancelled_write_bytes: %llu\n",
1939 #ifdef CONFIG_TASK_XACCT
1940 (unsigned long long)task->rchar,
1941 (unsigned long long)task->wchar,
1942 (unsigned long long)task->syscr,
1943 (unsigned long long)task->syscw,
1945 (unsigned long long)task->ioac.read_bytes,
1946 (unsigned long long)task->ioac.write_bytes,
1947 (unsigned long long)task->ioac.cancelled_write_bytes);
1954 static const struct file_operations proc_task_operations;
1955 static const struct inode_operations proc_task_inode_operations;
1957 static const struct pid_entry tgid_base_stuff[] = {
1958 DIR("task", S_IRUGO|S_IXUGO, task),
1959 DIR("fd", S_IRUSR|S_IXUSR, fd),
1960 DIR("fdinfo", S_IRUSR|S_IXUSR, fdinfo),
1961 INF("environ", S_IRUSR, pid_environ),
1962 INF("auxv", S_IRUSR, pid_auxv),
1963 INF("status", S_IRUGO, pid_status),
1964 INF("cmdline", S_IRUGO, pid_cmdline),
1965 INF("stat", S_IRUGO, tgid_stat),
1966 INF("statm", S_IRUGO, pid_statm),
1967 REG("maps", S_IRUGO, maps),
1969 REG("numa_maps", S_IRUGO, numa_maps),
1971 REG("mem", S_IRUSR|S_IWUSR, mem),
1972 #ifdef CONFIG_SECCOMP
1973 REG("seccomp", S_IRUSR|S_IWUSR, seccomp),
1978 REG("mounts", S_IRUGO, mounts),
1979 REG("mountstats", S_IRUSR, mountstats),
1981 REG("clear_refs", S_IWUSR, clear_refs),
1982 REG("smaps", S_IRUGO, smaps),
1984 #ifdef CONFIG_SECURITY
1985 DIR("attr", S_IRUGO|S_IXUGO, attr_dir),
1987 #ifdef CONFIG_KALLSYMS
1988 INF("wchan", S_IRUGO, pid_wchan),
1990 #ifdef CONFIG_SCHEDSTATS
1991 INF("schedstat", S_IRUGO, pid_schedstat),
1993 #ifdef CONFIG_CPUSETS
1994 REG("cpuset", S_IRUGO, cpuset),
1996 INF("oom_score", S_IRUGO, oom_score),
1997 REG("oom_adj", S_IRUGO|S_IWUSR, oom_adjust),
1998 #ifdef CONFIG_AUDITSYSCALL
1999 REG("loginuid", S_IWUSR|S_IRUGO, loginuid),
2001 #ifdef CONFIG_FAULT_INJECTION
2002 REG("make-it-fail", S_IRUGO|S_IWUSR, fault_inject),
2004 #ifdef CONFIG_TASK_IO_ACCOUNTING
2005 INF("io", S_IRUGO, pid_io_accounting),
2009 static int proc_tgid_base_readdir(struct file * filp,
2010 void * dirent, filldir_t filldir)
2012 return proc_pident_readdir(filp,dirent,filldir,
2013 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2016 static const struct file_operations proc_tgid_base_operations = {
2017 .read = generic_read_dir,
2018 .readdir = proc_tgid_base_readdir,
2021 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2022 return proc_pident_lookup(dir, dentry,
2023 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2026 static const struct inode_operations proc_tgid_base_inode_operations = {
2027 .lookup = proc_tgid_base_lookup,
2028 .getattr = pid_getattr,
2029 .setattr = proc_setattr,
2033 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2035 * @task: task that should be flushed.
2037 * Looks in the dcache for
2039 * /proc/@tgid/task/@pid
2040 * if either directory is present flushes it and all of it'ts children
2043 * It is safe and reasonable to cache /proc entries for a task until
2044 * that task exits. After that they just clog up the dcache with
2045 * useless entries, possibly causing useful dcache entries to be
2046 * flushed instead. This routine is proved to flush those useless
2047 * dcache entries at process exit time.
2049 * NOTE: This routine is just an optimization so it does not guarantee
2050 * that no dcache entries will exist at process exit time it
2051 * just makes it very unlikely that any will persist.
2053 void proc_flush_task(struct task_struct *task)
2055 struct dentry *dentry, *leader, *dir;
2056 char buf[PROC_NUMBUF];
2060 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
2061 dentry = d_hash_and_lookup(proc_mnt->mnt_root, &name);
2063 shrink_dcache_parent(dentry);
2068 if (thread_group_leader(task))
2072 name.len = snprintf(buf, sizeof(buf), "%d", task->tgid);
2073 leader = d_hash_and_lookup(proc_mnt->mnt_root, &name);
2078 name.len = strlen(name.name);
2079 dir = d_hash_and_lookup(leader, &name);
2081 goto out_put_leader;
2084 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
2085 dentry = d_hash_and_lookup(dir, &name);
2087 shrink_dcache_parent(dentry);
2099 static struct dentry *proc_pid_instantiate(struct inode *dir,
2100 struct dentry * dentry,
2101 struct task_struct *task, const void *ptr)
2103 struct dentry *error = ERR_PTR(-ENOENT);
2104 struct inode *inode;
2106 inode = proc_pid_make_inode(dir->i_sb, task);
2110 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2111 inode->i_op = &proc_tgid_base_inode_operations;
2112 inode->i_fop = &proc_tgid_base_operations;
2113 inode->i_flags|=S_IMMUTABLE;
2115 #ifdef CONFIG_SECURITY
2116 inode->i_nlink += 1;
2119 dentry->d_op = &pid_dentry_operations;
2121 d_add(dentry, inode);
2122 /* Close the race of the process dying before we return the dentry */
2123 if (pid_revalidate(dentry, NULL))
2129 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2131 struct dentry *result = ERR_PTR(-ENOENT);
2132 struct task_struct *task;
2135 result = proc_base_lookup(dir, dentry);
2136 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2139 tgid = name_to_int(dentry);
2144 task = find_task_by_pid(tgid);
2146 get_task_struct(task);
2151 result = proc_pid_instantiate(dir, dentry, task, NULL);
2152 put_task_struct(task);
2158 * Find the first task with tgid >= tgid
2161 static struct task_struct *next_tgid(unsigned int tgid)
2163 struct task_struct *task;
2169 pid = find_ge_pid(tgid);
2172 task = pid_task(pid, PIDTYPE_PID);
2173 /* What we to know is if the pid we have find is the
2174 * pid of a thread_group_leader. Testing for task
2175 * being a thread_group_leader is the obvious thing
2176 * todo but there is a window when it fails, due to
2177 * the pid transfer logic in de_thread.
2179 * So we perform the straight forward test of seeing
2180 * if the pid we have found is the pid of a thread
2181 * group leader, and don't worry if the task we have
2182 * found doesn't happen to be a thread group leader.
2183 * As we don't care in the case of readdir.
2185 if (!task || !has_group_leader_pid(task))
2187 get_task_struct(task);
2193 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2195 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2196 struct task_struct *task, int tgid)
2198 char name[PROC_NUMBUF];
2199 int len = snprintf(name, sizeof(name), "%d", tgid);
2200 return proc_fill_cache(filp, dirent, filldir, name, len,
2201 proc_pid_instantiate, task, NULL);
2204 /* for the /proc/ directory itself, after non-process stuff has been done */
2205 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2207 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2208 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
2209 struct task_struct *task;
2215 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
2216 const struct pid_entry *p = &proc_base_stuff[nr];
2217 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
2221 tgid = filp->f_pos - TGID_OFFSET;
2222 for (task = next_tgid(tgid);
2224 put_task_struct(task), task = next_tgid(tgid + 1)) {
2226 filp->f_pos = tgid + TGID_OFFSET;
2227 if (proc_pid_fill_cache(filp, dirent, filldir, task, tgid) < 0) {
2228 put_task_struct(task);
2232 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
2234 put_task_struct(reaper);
2242 static const struct pid_entry tid_base_stuff[] = {
2243 DIR("fd", S_IRUSR|S_IXUSR, fd),
2244 DIR("fdinfo", S_IRUSR|S_IXUSR, fdinfo),
2245 INF("environ", S_IRUSR, pid_environ),
2246 INF("auxv", S_IRUSR, pid_auxv),
2247 INF("status", S_IRUGO, pid_status),
2248 INF("cmdline", S_IRUGO, pid_cmdline),
2249 INF("stat", S_IRUGO, tid_stat),
2250 INF("statm", S_IRUGO, pid_statm),
2251 REG("maps", S_IRUGO, maps),
2253 REG("numa_maps", S_IRUGO, numa_maps),
2255 REG("mem", S_IRUSR|S_IWUSR, mem),
2256 #ifdef CONFIG_SECCOMP
2257 REG("seccomp", S_IRUSR|S_IWUSR, seccomp),
2262 REG("mounts", S_IRUGO, mounts),
2264 REG("clear_refs", S_IWUSR, clear_refs),
2265 REG("smaps", S_IRUGO, smaps),
2267 #ifdef CONFIG_SECURITY
2268 DIR("attr", S_IRUGO|S_IXUGO, attr_dir),
2270 #ifdef CONFIG_KALLSYMS
2271 INF("wchan", S_IRUGO, pid_wchan),
2273 #ifdef CONFIG_SCHEDSTATS
2274 INF("schedstat", S_IRUGO, pid_schedstat),
2276 #ifdef CONFIG_CPUSETS
2277 REG("cpuset", S_IRUGO, cpuset),
2279 INF("oom_score", S_IRUGO, oom_score),
2280 REG("oom_adj", S_IRUGO|S_IWUSR, oom_adjust),
2281 #ifdef CONFIG_AUDITSYSCALL
2282 REG("loginuid", S_IWUSR|S_IRUGO, loginuid),
2284 #ifdef CONFIG_FAULT_INJECTION
2285 REG("make-it-fail", S_IRUGO|S_IWUSR, fault_inject),
2289 static int proc_tid_base_readdir(struct file * filp,
2290 void * dirent, filldir_t filldir)
2292 return proc_pident_readdir(filp,dirent,filldir,
2293 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
2296 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2297 return proc_pident_lookup(dir, dentry,
2298 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
2301 static const struct file_operations proc_tid_base_operations = {
2302 .read = generic_read_dir,
2303 .readdir = proc_tid_base_readdir,
2306 static const struct inode_operations proc_tid_base_inode_operations = {
2307 .lookup = proc_tid_base_lookup,
2308 .getattr = pid_getattr,
2309 .setattr = proc_setattr,
2312 static struct dentry *proc_task_instantiate(struct inode *dir,
2313 struct dentry *dentry, struct task_struct *task, const void *ptr)
2315 struct dentry *error = ERR_PTR(-ENOENT);
2316 struct inode *inode;
2317 inode = proc_pid_make_inode(dir->i_sb, task);
2321 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2322 inode->i_op = &proc_tid_base_inode_operations;
2323 inode->i_fop = &proc_tid_base_operations;
2324 inode->i_flags|=S_IMMUTABLE;
2326 #ifdef CONFIG_SECURITY
2327 inode->i_nlink += 1;
2330 dentry->d_op = &pid_dentry_operations;
2332 d_add(dentry, inode);
2333 /* Close the race of the process dying before we return the dentry */
2334 if (pid_revalidate(dentry, NULL))
2340 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2342 struct dentry *result = ERR_PTR(-ENOENT);
2343 struct task_struct *task;
2344 struct task_struct *leader = get_proc_task(dir);
2350 tid = name_to_int(dentry);
2355 task = find_task_by_pid(tid);
2357 get_task_struct(task);
2361 if (leader->tgid != task->tgid)
2364 result = proc_task_instantiate(dir, dentry, task, NULL);
2366 put_task_struct(task);
2368 put_task_struct(leader);
2374 * Find the first tid of a thread group to return to user space.
2376 * Usually this is just the thread group leader, but if the users
2377 * buffer was too small or there was a seek into the middle of the
2378 * directory we have more work todo.
2380 * In the case of a short read we start with find_task_by_pid.
2382 * In the case of a seek we start with the leader and walk nr
2385 static struct task_struct *first_tid(struct task_struct *leader,
2388 struct task_struct *pos;
2391 /* Attempt to start with the pid of a thread */
2392 if (tid && (nr > 0)) {
2393 pos = find_task_by_pid(tid);
2394 if (pos && (pos->group_leader == leader))
2398 /* If nr exceeds the number of threads there is nothing todo */
2400 if (nr && nr >= get_nr_threads(leader))
2403 /* If we haven't found our starting place yet start
2404 * with the leader and walk nr threads forward.
2406 for (pos = leader; nr > 0; --nr) {
2407 pos = next_thread(pos);
2408 if (pos == leader) {
2414 get_task_struct(pos);
2421 * Find the next thread in the thread list.
2422 * Return NULL if there is an error or no next thread.
2424 * The reference to the input task_struct is released.
2426 static struct task_struct *next_tid(struct task_struct *start)
2428 struct task_struct *pos = NULL;
2430 if (pid_alive(start)) {
2431 pos = next_thread(start);
2432 if (thread_group_leader(pos))
2435 get_task_struct(pos);
2438 put_task_struct(start);
2442 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2443 struct task_struct *task, int tid)
2445 char name[PROC_NUMBUF];
2446 int len = snprintf(name, sizeof(name), "%d", tid);
2447 return proc_fill_cache(filp, dirent, filldir, name, len,
2448 proc_task_instantiate, task, NULL);
2451 /* for the /proc/TGID/task/ directories */
2452 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
2454 struct dentry *dentry = filp->f_path.dentry;
2455 struct inode *inode = dentry->d_inode;
2456 struct task_struct *leader = NULL;
2457 struct task_struct *task;
2458 int retval = -ENOENT;
2461 unsigned long pos = filp->f_pos; /* avoiding "long long" filp->f_pos */
2463 task = get_proc_task(inode);
2467 if (pid_alive(task)) {
2468 leader = task->group_leader;
2469 get_task_struct(leader);
2472 put_task_struct(task);
2480 if (filldir(dirent, ".", 1, pos, ino, DT_DIR) < 0)
2485 ino = parent_ino(dentry);
2486 if (filldir(dirent, "..", 2, pos, ino, DT_DIR) < 0)
2492 /* f_version caches the tgid value that the last readdir call couldn't
2493 * return. lseek aka telldir automagically resets f_version to 0.
2495 tid = filp->f_version;
2496 filp->f_version = 0;
2497 for (task = first_tid(leader, tid, pos - 2);
2499 task = next_tid(task), pos++) {
2501 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
2502 /* returning this tgid failed, save it as the first
2503 * pid for the next readir call */
2504 filp->f_version = tid;
2505 put_task_struct(task);
2511 put_task_struct(leader);
2516 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
2518 struct inode *inode = dentry->d_inode;
2519 struct task_struct *p = get_proc_task(inode);
2520 generic_fillattr(inode, stat);
2524 stat->nlink += get_nr_threads(p);
2532 static const struct inode_operations proc_task_inode_operations = {
2533 .lookup = proc_task_lookup,
2534 .getattr = proc_task_getattr,
2535 .setattr = proc_setattr,
2538 static const struct file_operations proc_task_operations = {
2539 .read = generic_read_dir,
2540 .readdir = proc_task_readdir,