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 const char *sym_name;
283 unsigned long wchan, size, offset;
284 char namebuf[KSYM_NAME_LEN+1];
286 wchan = get_wchan(task);
288 sym_name = kallsyms_lookup(wchan, &size, &offset, &modname, namebuf);
290 return sprintf(buffer, "%s", sym_name);
291 return sprintf(buffer, "%lu", wchan);
293 #endif /* CONFIG_KALLSYMS */
295 #ifdef CONFIG_SCHEDSTATS
297 * Provides /proc/PID/schedstat
299 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
301 return sprintf(buffer, "%lu %lu %lu\n",
302 task->sched_info.cpu_time,
303 task->sched_info.run_delay,
304 task->sched_info.pcnt);
308 /* The badness from the OOM killer */
309 unsigned long badness(struct task_struct *p, unsigned long uptime);
310 static int proc_oom_score(struct task_struct *task, char *buffer)
312 unsigned long points;
313 struct timespec uptime;
315 do_posix_clock_monotonic_gettime(&uptime);
316 read_lock(&tasklist_lock);
317 points = badness(task, uptime.tv_sec);
318 read_unlock(&tasklist_lock);
319 return sprintf(buffer, "%lu\n", points);
322 /************************************************************************/
323 /* Here the fs part begins */
324 /************************************************************************/
326 /* permission checks */
327 static int proc_fd_access_allowed(struct inode *inode)
329 struct task_struct *task;
331 /* Allow access to a task's file descriptors if it is us or we
332 * may use ptrace attach to the process and find out that
335 task = get_proc_task(inode);
337 allowed = ptrace_may_attach(task);
338 put_task_struct(task);
343 static int proc_setattr(struct dentry *dentry, struct iattr *attr)
346 struct inode *inode = dentry->d_inode;
348 if (attr->ia_valid & ATTR_MODE)
351 error = inode_change_ok(inode, attr);
353 error = security_inode_setattr(dentry, attr);
355 error = inode_setattr(inode, attr);
360 static const struct inode_operations proc_def_inode_operations = {
361 .setattr = proc_setattr,
364 extern struct seq_operations mounts_op;
370 static int mounts_open(struct inode *inode, struct file *file)
372 struct task_struct *task = get_proc_task(inode);
373 struct mnt_namespace *ns = NULL;
374 struct proc_mounts *p;
380 ns = task->nsproxy->mnt_ns;
385 put_task_struct(task);
390 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
392 file->private_data = &p->m;
393 ret = seq_open(file, &mounts_op);
396 p->event = ns->event;
406 static int mounts_release(struct inode *inode, struct file *file)
408 struct seq_file *m = file->private_data;
409 struct mnt_namespace *ns = m->private;
411 return seq_release(inode, file);
414 static unsigned mounts_poll(struct file *file, poll_table *wait)
416 struct proc_mounts *p = file->private_data;
417 struct mnt_namespace *ns = p->m.private;
420 poll_wait(file, &ns->poll, wait);
422 spin_lock(&vfsmount_lock);
423 if (p->event != ns->event) {
424 p->event = ns->event;
427 spin_unlock(&vfsmount_lock);
432 static const struct file_operations proc_mounts_operations = {
436 .release = mounts_release,
440 extern struct seq_operations mountstats_op;
441 static int mountstats_open(struct inode *inode, struct file *file)
443 int ret = seq_open(file, &mountstats_op);
446 struct seq_file *m = file->private_data;
447 struct mnt_namespace *mnt_ns = NULL;
448 struct task_struct *task = get_proc_task(inode);
453 mnt_ns = task->nsproxy->mnt_ns;
457 put_task_struct(task);
463 seq_release(inode, file);
470 static const struct file_operations proc_mountstats_operations = {
471 .open = mountstats_open,
474 .release = mounts_release,
477 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
479 static ssize_t proc_info_read(struct file * file, char __user * buf,
480 size_t count, loff_t *ppos)
482 struct inode * inode = file->f_path.dentry->d_inode;
485 struct task_struct *task = get_proc_task(inode);
491 if (count > PROC_BLOCK_SIZE)
492 count = PROC_BLOCK_SIZE;
495 if (!(page = __get_free_page(GFP_KERNEL)))
498 length = PROC_I(inode)->op.proc_read(task, (char*)page);
501 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
504 put_task_struct(task);
509 static const struct file_operations proc_info_file_operations = {
510 .read = proc_info_read,
513 static int mem_open(struct inode* inode, struct file* file)
515 file->private_data = (void*)((long)current->self_exec_id);
519 static ssize_t mem_read(struct file * file, char __user * buf,
520 size_t count, loff_t *ppos)
522 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
524 unsigned long src = *ppos;
526 struct mm_struct *mm;
531 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
535 page = (char *)__get_free_page(GFP_USER);
541 mm = get_task_mm(task);
547 if (file->private_data != (void*)((long)current->self_exec_id))
553 int this_len, retval;
555 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
556 retval = access_process_vm(task, src, page, this_len, 0);
557 if (!retval || !MAY_PTRACE(task) || !ptrace_may_attach(task)) {
563 if (copy_to_user(buf, page, retval)) {
578 free_page((unsigned long) page);
580 put_task_struct(task);
585 #define mem_write NULL
588 /* This is a security hazard */
589 static ssize_t mem_write(struct file * file, const char __user *buf,
590 size_t count, loff_t *ppos)
594 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
595 unsigned long dst = *ppos;
601 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
605 page = (char *)__get_free_page(GFP_USER);
611 int this_len, retval;
613 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
614 if (copy_from_user(page, buf, this_len)) {
618 retval = access_process_vm(task, dst, page, this_len, 1);
630 free_page((unsigned long) page);
632 put_task_struct(task);
638 static loff_t mem_lseek(struct file * file, loff_t offset, int orig)
642 file->f_pos = offset;
645 file->f_pos += offset;
650 force_successful_syscall_return();
654 static const struct file_operations proc_mem_operations = {
661 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
662 size_t count, loff_t *ppos)
664 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
665 char buffer[PROC_NUMBUF];
668 loff_t __ppos = *ppos;
672 oom_adjust = task->oomkilladj;
673 put_task_struct(task);
675 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
678 if (count > len-__ppos)
680 if (copy_to_user(buf, buffer + __ppos, count))
682 *ppos = __ppos + count;
686 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
687 size_t count, loff_t *ppos)
689 struct task_struct *task;
690 char buffer[PROC_NUMBUF], *end;
693 memset(buffer, 0, sizeof(buffer));
694 if (count > sizeof(buffer) - 1)
695 count = sizeof(buffer) - 1;
696 if (copy_from_user(buffer, buf, count))
698 oom_adjust = simple_strtol(buffer, &end, 0);
699 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
700 oom_adjust != OOM_DISABLE)
704 task = get_proc_task(file->f_path.dentry->d_inode);
707 if (oom_adjust < task->oomkilladj && !capable(CAP_SYS_RESOURCE)) {
708 put_task_struct(task);
711 task->oomkilladj = oom_adjust;
712 put_task_struct(task);
713 if (end - buffer == 0)
718 static const struct file_operations proc_oom_adjust_operations = {
719 .read = oom_adjust_read,
720 .write = oom_adjust_write,
723 static ssize_t clear_refs_write(struct file *file, const char __user *buf,
724 size_t count, loff_t *ppos)
726 struct task_struct *task;
727 char buffer[PROC_NUMBUF], *end;
728 struct mm_struct *mm;
730 memset(buffer, 0, sizeof(buffer));
731 if (count > sizeof(buffer) - 1)
732 count = sizeof(buffer) - 1;
733 if (copy_from_user(buffer, buf, count))
735 if (!simple_strtol(buffer, &end, 0))
739 task = get_proc_task(file->f_path.dentry->d_inode);
742 mm = get_task_mm(task);
747 put_task_struct(task);
748 if (end - buffer == 0)
753 static struct file_operations proc_clear_refs_operations = {
754 .write = clear_refs_write,
757 #ifdef CONFIG_AUDITSYSCALL
759 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
760 size_t count, loff_t *ppos)
762 struct inode * inode = file->f_path.dentry->d_inode;
763 struct task_struct *task = get_proc_task(inode);
765 char tmpbuf[TMPBUFLEN];
769 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
770 audit_get_loginuid(task->audit_context));
771 put_task_struct(task);
772 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
775 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
776 size_t count, loff_t *ppos)
778 struct inode * inode = file->f_path.dentry->d_inode;
783 if (!capable(CAP_AUDIT_CONTROL))
786 if (current != pid_task(proc_pid(inode), PIDTYPE_PID))
789 if (count >= PAGE_SIZE)
790 count = PAGE_SIZE - 1;
793 /* No partial writes. */
796 page = (char*)__get_free_page(GFP_USER);
800 if (copy_from_user(page, buf, count))
804 loginuid = simple_strtoul(page, &tmp, 10);
810 length = audit_set_loginuid(current, loginuid);
811 if (likely(length == 0))
815 free_page((unsigned long) page);
819 static const struct file_operations proc_loginuid_operations = {
820 .read = proc_loginuid_read,
821 .write = proc_loginuid_write,
825 #ifdef CONFIG_SECCOMP
826 static ssize_t seccomp_read(struct file *file, char __user *buf,
827 size_t count, loff_t *ppos)
829 struct task_struct *tsk = get_proc_task(file->f_dentry->d_inode);
831 loff_t __ppos = *ppos;
836 /* no need to print the trailing zero, so use only len */
837 len = sprintf(__buf, "%u\n", tsk->seccomp.mode);
838 put_task_struct(tsk);
841 if (count > len - __ppos)
842 count = len - __ppos;
843 if (copy_to_user(buf, __buf + __ppos, count))
845 *ppos = __ppos + count;
849 static ssize_t seccomp_write(struct file *file, const char __user *buf,
850 size_t count, loff_t *ppos)
852 struct task_struct *tsk = get_proc_task(file->f_dentry->d_inode);
853 char __buf[20], *end;
854 unsigned int seccomp_mode;
861 /* can set it only once to be even more secure */
863 if (unlikely(tsk->seccomp.mode))
867 memset(__buf, 0, sizeof(__buf));
868 count = min(count, sizeof(__buf) - 1);
869 if (copy_from_user(__buf, buf, count))
872 seccomp_mode = simple_strtoul(__buf, &end, 0);
876 if (seccomp_mode && seccomp_mode <= NR_SECCOMP_MODES) {
877 tsk->seccomp.mode = seccomp_mode;
878 set_tsk_thread_flag(tsk, TIF_SECCOMP);
882 if (unlikely(!(end - __buf)))
884 result = end - __buf;
886 put_task_struct(tsk);
891 static const struct file_operations proc_seccomp_operations = {
892 .read = seccomp_read,
893 .write = seccomp_write,
895 #endif /* CONFIG_SECCOMP */
897 #ifdef CONFIG_FAULT_INJECTION
898 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
899 size_t count, loff_t *ppos)
901 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
902 char buffer[PROC_NUMBUF];
905 loff_t __ppos = *ppos;
909 make_it_fail = task->make_it_fail;
910 put_task_struct(task);
912 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
915 if (count > len-__ppos)
917 if (copy_to_user(buf, buffer + __ppos, count))
919 *ppos = __ppos + count;
923 static ssize_t proc_fault_inject_write(struct file * file,
924 const char __user * buf, size_t count, loff_t *ppos)
926 struct task_struct *task;
927 char buffer[PROC_NUMBUF], *end;
930 if (!capable(CAP_SYS_RESOURCE))
932 memset(buffer, 0, sizeof(buffer));
933 if (count > sizeof(buffer) - 1)
934 count = sizeof(buffer) - 1;
935 if (copy_from_user(buffer, buf, count))
937 make_it_fail = simple_strtol(buffer, &end, 0);
940 task = get_proc_task(file->f_dentry->d_inode);
943 task->make_it_fail = make_it_fail;
944 put_task_struct(task);
945 if (end - buffer == 0)
950 static const struct file_operations proc_fault_inject_operations = {
951 .read = proc_fault_inject_read,
952 .write = proc_fault_inject_write,
956 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
958 struct inode *inode = dentry->d_inode;
961 /* We don't need a base pointer in the /proc filesystem */
964 /* Are we allowed to snoop on the tasks file descriptors? */
965 if (!proc_fd_access_allowed(inode))
968 error = PROC_I(inode)->op.proc_get_link(inode, &nd->dentry, &nd->mnt);
969 nd->last_type = LAST_BIND;
971 return ERR_PTR(error);
974 static int do_proc_readlink(struct dentry *dentry, struct vfsmount *mnt,
975 char __user *buffer, int buflen)
977 struct inode * inode;
978 char *tmp = (char*)__get_free_page(GFP_KERNEL), *path;
984 inode = dentry->d_inode;
985 path = d_path(dentry, mnt, tmp, PAGE_SIZE);
989 len = tmp + PAGE_SIZE - 1 - path;
993 if (copy_to_user(buffer, path, len))
996 free_page((unsigned long)tmp);
1000 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1002 int error = -EACCES;
1003 struct inode *inode = dentry->d_inode;
1005 struct vfsmount *mnt = NULL;
1007 /* Are we allowed to snoop on the tasks file descriptors? */
1008 if (!proc_fd_access_allowed(inode))
1011 error = PROC_I(inode)->op.proc_get_link(inode, &de, &mnt);
1015 error = do_proc_readlink(de, mnt, buffer, buflen);
1022 static const struct inode_operations proc_pid_link_inode_operations = {
1023 .readlink = proc_pid_readlink,
1024 .follow_link = proc_pid_follow_link,
1025 .setattr = proc_setattr,
1029 /* building an inode */
1031 static int task_dumpable(struct task_struct *task)
1034 struct mm_struct *mm;
1039 dumpable = mm->dumpable;
1047 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1049 struct inode * inode;
1050 struct proc_inode *ei;
1052 /* We need a new inode */
1054 inode = new_inode(sb);
1060 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1061 inode->i_op = &proc_def_inode_operations;
1064 * grab the reference to task.
1066 ei->pid = get_task_pid(task, PIDTYPE_PID);
1072 if (task_dumpable(task)) {
1073 inode->i_uid = task->euid;
1074 inode->i_gid = task->egid;
1076 security_task_to_inode(task, inode);
1086 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1088 struct inode *inode = dentry->d_inode;
1089 struct task_struct *task;
1090 generic_fillattr(inode, stat);
1095 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1097 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1098 task_dumpable(task)) {
1099 stat->uid = task->euid;
1100 stat->gid = task->egid;
1110 * Exceptional case: normally we are not allowed to unhash a busy
1111 * directory. In this case, however, we can do it - no aliasing problems
1112 * due to the way we treat inodes.
1114 * Rewrite the inode's ownerships here because the owning task may have
1115 * performed a setuid(), etc.
1117 * Before the /proc/pid/status file was created the only way to read
1118 * the effective uid of a /process was to stat /proc/pid. Reading
1119 * /proc/pid/status is slow enough that procps and other packages
1120 * kept stating /proc/pid. To keep the rules in /proc simple I have
1121 * made this apply to all per process world readable and executable
1124 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1126 struct inode *inode = dentry->d_inode;
1127 struct task_struct *task = get_proc_task(inode);
1129 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1130 task_dumpable(task)) {
1131 inode->i_uid = task->euid;
1132 inode->i_gid = task->egid;
1137 inode->i_mode &= ~(S_ISUID | S_ISGID);
1138 security_task_to_inode(task, inode);
1139 put_task_struct(task);
1146 static int pid_delete_dentry(struct dentry * dentry)
1148 /* Is the task we represent dead?
1149 * If so, then don't put the dentry on the lru list,
1150 * kill it immediately.
1152 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1155 static struct dentry_operations pid_dentry_operations =
1157 .d_revalidate = pid_revalidate,
1158 .d_delete = pid_delete_dentry,
1163 typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1164 struct task_struct *, const void *);
1167 * Fill a directory entry.
1169 * If possible create the dcache entry and derive our inode number and
1170 * file type from dcache entry.
1172 * Since all of the proc inode numbers are dynamically generated, the inode
1173 * numbers do not exist until the inode is cache. This means creating the
1174 * the dcache entry in readdir is necessary to keep the inode numbers
1175 * reported by readdir in sync with the inode numbers reported
1178 static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1179 char *name, int len,
1180 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1182 struct dentry *child, *dir = filp->f_path.dentry;
1183 struct inode *inode;
1186 unsigned type = DT_UNKNOWN;
1190 qname.hash = full_name_hash(name, len);
1192 child = d_lookup(dir, &qname);
1195 new = d_alloc(dir, &qname);
1197 child = instantiate(dir->d_inode, new, task, ptr);
1204 if (!child || IS_ERR(child) || !child->d_inode)
1205 goto end_instantiate;
1206 inode = child->d_inode;
1209 type = inode->i_mode >> 12;
1214 ino = find_inode_number(dir, &qname);
1217 return filldir(dirent, name, len, filp->f_pos, ino, type);
1220 static unsigned name_to_int(struct dentry *dentry)
1222 const char *name = dentry->d_name.name;
1223 int len = dentry->d_name.len;
1226 if (len > 1 && *name == '0')
1229 unsigned c = *name++ - '0';
1232 if (n >= (~0U-9)/10)
1242 #define PROC_FDINFO_MAX 64
1244 static int proc_fd_info(struct inode *inode, struct dentry **dentry,
1245 struct vfsmount **mnt, char *info)
1247 struct task_struct *task = get_proc_task(inode);
1248 struct files_struct *files = NULL;
1250 int fd = proc_fd(inode);
1253 files = get_files_struct(task);
1254 put_task_struct(task);
1258 * We are not taking a ref to the file structure, so we must
1261 spin_lock(&files->file_lock);
1262 file = fcheck_files(files, fd);
1265 *mnt = mntget(file->f_path.mnt);
1267 *dentry = dget(file->f_path.dentry);
1269 snprintf(info, PROC_FDINFO_MAX,
1272 (long long) file->f_pos,
1274 spin_unlock(&files->file_lock);
1275 put_files_struct(files);
1278 spin_unlock(&files->file_lock);
1279 put_files_struct(files);
1284 static int proc_fd_link(struct inode *inode, struct dentry **dentry,
1285 struct vfsmount **mnt)
1287 return proc_fd_info(inode, dentry, mnt, NULL);
1290 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1292 struct inode *inode = dentry->d_inode;
1293 struct task_struct *task = get_proc_task(inode);
1294 int fd = proc_fd(inode);
1295 struct files_struct *files;
1298 files = get_files_struct(task);
1301 if (fcheck_files(files, fd)) {
1303 put_files_struct(files);
1304 if (task_dumpable(task)) {
1305 inode->i_uid = task->euid;
1306 inode->i_gid = task->egid;
1311 inode->i_mode &= ~(S_ISUID | S_ISGID);
1312 security_task_to_inode(task, inode);
1313 put_task_struct(task);
1317 put_files_struct(files);
1319 put_task_struct(task);
1325 static struct dentry_operations tid_fd_dentry_operations =
1327 .d_revalidate = tid_fd_revalidate,
1328 .d_delete = pid_delete_dentry,
1331 static struct dentry *proc_fd_instantiate(struct inode *dir,
1332 struct dentry *dentry, struct task_struct *task, const void *ptr)
1334 unsigned fd = *(const unsigned *)ptr;
1336 struct files_struct *files;
1337 struct inode *inode;
1338 struct proc_inode *ei;
1339 struct dentry *error = ERR_PTR(-ENOENT);
1341 inode = proc_pid_make_inode(dir->i_sb, task);
1346 files = get_files_struct(task);
1349 inode->i_mode = S_IFLNK;
1352 * We are not taking a ref to the file structure, so we must
1355 spin_lock(&files->file_lock);
1356 file = fcheck_files(files, fd);
1359 if (file->f_mode & 1)
1360 inode->i_mode |= S_IRUSR | S_IXUSR;
1361 if (file->f_mode & 2)
1362 inode->i_mode |= S_IWUSR | S_IXUSR;
1363 spin_unlock(&files->file_lock);
1364 put_files_struct(files);
1366 inode->i_op = &proc_pid_link_inode_operations;
1368 ei->op.proc_get_link = proc_fd_link;
1369 dentry->d_op = &tid_fd_dentry_operations;
1370 d_add(dentry, inode);
1371 /* Close the race of the process dying before we return the dentry */
1372 if (tid_fd_revalidate(dentry, NULL))
1378 spin_unlock(&files->file_lock);
1379 put_files_struct(files);
1385 static struct dentry *proc_lookupfd_common(struct inode *dir,
1386 struct dentry *dentry,
1387 instantiate_t instantiate)
1389 struct task_struct *task = get_proc_task(dir);
1390 unsigned fd = name_to_int(dentry);
1391 struct dentry *result = ERR_PTR(-ENOENT);
1398 result = instantiate(dir, dentry, task, &fd);
1400 put_task_struct(task);
1405 static int proc_readfd_common(struct file * filp, void * dirent,
1406 filldir_t filldir, instantiate_t instantiate)
1408 struct dentry *dentry = filp->f_path.dentry;
1409 struct inode *inode = dentry->d_inode;
1410 struct task_struct *p = get_proc_task(inode);
1411 unsigned int fd, tid, ino;
1413 struct files_struct * files;
1414 struct fdtable *fdt;
1425 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1429 ino = parent_ino(dentry);
1430 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1434 files = get_files_struct(p);
1438 fdt = files_fdtable(files);
1439 for (fd = filp->f_pos-2;
1441 fd++, filp->f_pos++) {
1442 char name[PROC_NUMBUF];
1445 if (!fcheck_files(files, fd))
1449 len = snprintf(name, sizeof(name), "%d", fd);
1450 if (proc_fill_cache(filp, dirent, filldir,
1451 name, len, instantiate,
1459 put_files_struct(files);
1467 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1468 struct nameidata *nd)
1470 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1473 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1475 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1478 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1479 size_t len, loff_t *ppos)
1481 char tmp[PROC_FDINFO_MAX];
1482 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, NULL, tmp);
1484 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1488 static const struct file_operations proc_fdinfo_file_operations = {
1489 .open = nonseekable_open,
1490 .read = proc_fdinfo_read,
1493 static const struct file_operations proc_fd_operations = {
1494 .read = generic_read_dir,
1495 .readdir = proc_readfd,
1499 * /proc/pid/fd needs a special permission handler so that a process can still
1500 * access /proc/self/fd after it has executed a setuid().
1502 static int proc_fd_permission(struct inode *inode, int mask,
1503 struct nameidata *nd)
1507 rv = generic_permission(inode, mask, NULL);
1510 if (task_pid(current) == proc_pid(inode))
1516 * proc directories can do almost nothing..
1518 static const struct inode_operations proc_fd_inode_operations = {
1519 .lookup = proc_lookupfd,
1520 .permission = proc_fd_permission,
1521 .setattr = proc_setattr,
1524 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
1525 struct dentry *dentry, struct task_struct *task, const void *ptr)
1527 unsigned fd = *(unsigned *)ptr;
1528 struct inode *inode;
1529 struct proc_inode *ei;
1530 struct dentry *error = ERR_PTR(-ENOENT);
1532 inode = proc_pid_make_inode(dir->i_sb, task);
1537 inode->i_mode = S_IFREG | S_IRUSR;
1538 inode->i_fop = &proc_fdinfo_file_operations;
1539 dentry->d_op = &tid_fd_dentry_operations;
1540 d_add(dentry, inode);
1541 /* Close the race of the process dying before we return the dentry */
1542 if (tid_fd_revalidate(dentry, NULL))
1549 static struct dentry *proc_lookupfdinfo(struct inode *dir,
1550 struct dentry *dentry,
1551 struct nameidata *nd)
1553 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
1556 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
1558 return proc_readfd_common(filp, dirent, filldir,
1559 proc_fdinfo_instantiate);
1562 static const struct file_operations proc_fdinfo_operations = {
1563 .read = generic_read_dir,
1564 .readdir = proc_readfdinfo,
1568 * proc directories can do almost nothing..
1570 static const struct inode_operations proc_fdinfo_inode_operations = {
1571 .lookup = proc_lookupfdinfo,
1572 .setattr = proc_setattr,
1576 static struct dentry *proc_pident_instantiate(struct inode *dir,
1577 struct dentry *dentry, struct task_struct *task, const void *ptr)
1579 const struct pid_entry *p = ptr;
1580 struct inode *inode;
1581 struct proc_inode *ei;
1582 struct dentry *error = ERR_PTR(-EINVAL);
1584 inode = proc_pid_make_inode(dir->i_sb, task);
1589 inode->i_mode = p->mode;
1590 if (S_ISDIR(inode->i_mode))
1591 inode->i_nlink = 2; /* Use getattr to fix if necessary */
1593 inode->i_op = p->iop;
1595 inode->i_fop = p->fop;
1597 dentry->d_op = &pid_dentry_operations;
1598 d_add(dentry, inode);
1599 /* Close the race of the process dying before we return the dentry */
1600 if (pid_revalidate(dentry, NULL))
1606 static struct dentry *proc_pident_lookup(struct inode *dir,
1607 struct dentry *dentry,
1608 const struct pid_entry *ents,
1611 struct inode *inode;
1612 struct dentry *error;
1613 struct task_struct *task = get_proc_task(dir);
1614 const struct pid_entry *p, *last;
1616 error = ERR_PTR(-ENOENT);
1623 * Yes, it does not scale. And it should not. Don't add
1624 * new entries into /proc/<tgid>/ without very good reasons.
1626 last = &ents[nents - 1];
1627 for (p = ents; p <= last; p++) {
1628 if (p->len != dentry->d_name.len)
1630 if (!memcmp(dentry->d_name.name, p->name, p->len))
1636 error = proc_pident_instantiate(dir, dentry, task, p);
1638 put_task_struct(task);
1643 static int proc_pident_fill_cache(struct file *filp, void *dirent,
1644 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
1646 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
1647 proc_pident_instantiate, task, p);
1650 static int proc_pident_readdir(struct file *filp,
1651 void *dirent, filldir_t filldir,
1652 const struct pid_entry *ents, unsigned int nents)
1656 struct dentry *dentry = filp->f_path.dentry;
1657 struct inode *inode = dentry->d_inode;
1658 struct task_struct *task = get_proc_task(inode);
1659 const struct pid_entry *p, *last;
1673 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
1679 ino = parent_ino(dentry);
1680 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
1692 last = &ents[nents - 1];
1694 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
1703 put_task_struct(task);
1708 #ifdef CONFIG_SECURITY
1709 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
1710 size_t count, loff_t *ppos)
1712 struct inode * inode = file->f_path.dentry->d_inode;
1715 struct task_struct *task = get_proc_task(inode);
1720 length = security_getprocattr(task,
1721 (char*)file->f_path.dentry->d_name.name,
1723 put_task_struct(task);
1725 length = simple_read_from_buffer(buf, count, ppos, p, length);
1730 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
1731 size_t count, loff_t *ppos)
1733 struct inode * inode = file->f_path.dentry->d_inode;
1736 struct task_struct *task = get_proc_task(inode);
1741 if (count > PAGE_SIZE)
1744 /* No partial writes. */
1750 page = (char*)__get_free_page(GFP_USER);
1755 if (copy_from_user(page, buf, count))
1758 length = security_setprocattr(task,
1759 (char*)file->f_path.dentry->d_name.name,
1760 (void*)page, count);
1762 free_page((unsigned long) page);
1764 put_task_struct(task);
1769 static const struct file_operations proc_pid_attr_operations = {
1770 .read = proc_pid_attr_read,
1771 .write = proc_pid_attr_write,
1774 static const struct pid_entry attr_dir_stuff[] = {
1775 REG("current", S_IRUGO|S_IWUGO, pid_attr),
1776 REG("prev", S_IRUGO, pid_attr),
1777 REG("exec", S_IRUGO|S_IWUGO, pid_attr),
1778 REG("fscreate", S_IRUGO|S_IWUGO, pid_attr),
1779 REG("keycreate", S_IRUGO|S_IWUGO, pid_attr),
1780 REG("sockcreate", S_IRUGO|S_IWUGO, pid_attr),
1783 static int proc_attr_dir_readdir(struct file * filp,
1784 void * dirent, filldir_t filldir)
1786 return proc_pident_readdir(filp,dirent,filldir,
1787 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
1790 static const struct file_operations proc_attr_dir_operations = {
1791 .read = generic_read_dir,
1792 .readdir = proc_attr_dir_readdir,
1795 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
1796 struct dentry *dentry, struct nameidata *nd)
1798 return proc_pident_lookup(dir, dentry,
1799 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
1802 static const struct inode_operations proc_attr_dir_inode_operations = {
1803 .lookup = proc_attr_dir_lookup,
1804 .getattr = pid_getattr,
1805 .setattr = proc_setattr,
1813 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
1816 char tmp[PROC_NUMBUF];
1817 sprintf(tmp, "%d", current->tgid);
1818 return vfs_readlink(dentry,buffer,buflen,tmp);
1821 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
1823 char tmp[PROC_NUMBUF];
1824 sprintf(tmp, "%d", current->tgid);
1825 return ERR_PTR(vfs_follow_link(nd,tmp));
1828 static const struct inode_operations proc_self_inode_operations = {
1829 .readlink = proc_self_readlink,
1830 .follow_link = proc_self_follow_link,
1836 * These are the directory entries in the root directory of /proc
1837 * that properly belong to the /proc filesystem, as they describe
1838 * describe something that is process related.
1840 static const struct pid_entry proc_base_stuff[] = {
1841 NOD("self", S_IFLNK|S_IRWXUGO,
1842 &proc_self_inode_operations, NULL, {}),
1846 * Exceptional case: normally we are not allowed to unhash a busy
1847 * directory. In this case, however, we can do it - no aliasing problems
1848 * due to the way we treat inodes.
1850 static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
1852 struct inode *inode = dentry->d_inode;
1853 struct task_struct *task = get_proc_task(inode);
1855 put_task_struct(task);
1862 static struct dentry_operations proc_base_dentry_operations =
1864 .d_revalidate = proc_base_revalidate,
1865 .d_delete = pid_delete_dentry,
1868 static struct dentry *proc_base_instantiate(struct inode *dir,
1869 struct dentry *dentry, struct task_struct *task, const void *ptr)
1871 const struct pid_entry *p = ptr;
1872 struct inode *inode;
1873 struct proc_inode *ei;
1874 struct dentry *error = ERR_PTR(-EINVAL);
1876 /* Allocate the inode */
1877 error = ERR_PTR(-ENOMEM);
1878 inode = new_inode(dir->i_sb);
1882 /* Initialize the inode */
1884 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1887 * grab the reference to the task.
1889 ei->pid = get_task_pid(task, PIDTYPE_PID);
1895 inode->i_mode = p->mode;
1896 if (S_ISDIR(inode->i_mode))
1898 if (S_ISLNK(inode->i_mode))
1901 inode->i_op = p->iop;
1903 inode->i_fop = p->fop;
1905 dentry->d_op = &proc_base_dentry_operations;
1906 d_add(dentry, inode);
1915 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
1917 struct dentry *error;
1918 struct task_struct *task = get_proc_task(dir);
1919 const struct pid_entry *p, *last;
1921 error = ERR_PTR(-ENOENT);
1926 /* Lookup the directory entry */
1927 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
1928 for (p = proc_base_stuff; p <= last; p++) {
1929 if (p->len != dentry->d_name.len)
1931 if (!memcmp(dentry->d_name.name, p->name, p->len))
1937 error = proc_base_instantiate(dir, dentry, task, p);
1940 put_task_struct(task);
1945 static int proc_base_fill_cache(struct file *filp, void *dirent,
1946 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
1948 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
1949 proc_base_instantiate, task, p);
1952 #ifdef CONFIG_TASK_IO_ACCOUNTING
1953 static int proc_pid_io_accounting(struct task_struct *task, char *buffer)
1955 return sprintf(buffer,
1956 #ifdef CONFIG_TASK_XACCT
1962 "read_bytes: %llu\n"
1963 "write_bytes: %llu\n"
1964 "cancelled_write_bytes: %llu\n",
1965 #ifdef CONFIG_TASK_XACCT
1966 (unsigned long long)task->rchar,
1967 (unsigned long long)task->wchar,
1968 (unsigned long long)task->syscr,
1969 (unsigned long long)task->syscw,
1971 (unsigned long long)task->ioac.read_bytes,
1972 (unsigned long long)task->ioac.write_bytes,
1973 (unsigned long long)task->ioac.cancelled_write_bytes);
1980 static const struct file_operations proc_task_operations;
1981 static const struct inode_operations proc_task_inode_operations;
1983 static const struct pid_entry tgid_base_stuff[] = {
1984 DIR("task", S_IRUGO|S_IXUGO, task),
1985 DIR("fd", S_IRUSR|S_IXUSR, fd),
1986 DIR("fdinfo", S_IRUSR|S_IXUSR, fdinfo),
1987 INF("environ", S_IRUSR, pid_environ),
1988 INF("auxv", S_IRUSR, pid_auxv),
1989 INF("status", S_IRUGO, pid_status),
1990 INF("cmdline", S_IRUGO, pid_cmdline),
1991 INF("stat", S_IRUGO, tgid_stat),
1992 INF("statm", S_IRUGO, pid_statm),
1993 REG("maps", S_IRUGO, maps),
1995 REG("numa_maps", S_IRUGO, numa_maps),
1997 REG("mem", S_IRUSR|S_IWUSR, mem),
1998 #ifdef CONFIG_SECCOMP
1999 REG("seccomp", S_IRUSR|S_IWUSR, seccomp),
2004 REG("mounts", S_IRUGO, mounts),
2005 REG("mountstats", S_IRUSR, mountstats),
2007 REG("clear_refs", S_IWUSR, clear_refs),
2008 REG("smaps", S_IRUGO, smaps),
2010 #ifdef CONFIG_SECURITY
2011 DIR("attr", S_IRUGO|S_IXUGO, attr_dir),
2013 #ifdef CONFIG_KALLSYMS
2014 INF("wchan", S_IRUGO, pid_wchan),
2016 #ifdef CONFIG_SCHEDSTATS
2017 INF("schedstat", S_IRUGO, pid_schedstat),
2019 #ifdef CONFIG_CPUSETS
2020 REG("cpuset", S_IRUGO, cpuset),
2022 INF("oom_score", S_IRUGO, oom_score),
2023 REG("oom_adj", S_IRUGO|S_IWUSR, oom_adjust),
2024 #ifdef CONFIG_AUDITSYSCALL
2025 REG("loginuid", S_IWUSR|S_IRUGO, loginuid),
2027 #ifdef CONFIG_FAULT_INJECTION
2028 REG("make-it-fail", S_IRUGO|S_IWUSR, fault_inject),
2030 #ifdef CONFIG_TASK_IO_ACCOUNTING
2031 INF("io", S_IRUGO, pid_io_accounting),
2035 static int proc_tgid_base_readdir(struct file * filp,
2036 void * dirent, filldir_t filldir)
2038 return proc_pident_readdir(filp,dirent,filldir,
2039 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2042 static const struct file_operations proc_tgid_base_operations = {
2043 .read = generic_read_dir,
2044 .readdir = proc_tgid_base_readdir,
2047 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2048 return proc_pident_lookup(dir, dentry,
2049 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2052 static const struct inode_operations proc_tgid_base_inode_operations = {
2053 .lookup = proc_tgid_base_lookup,
2054 .getattr = pid_getattr,
2055 .setattr = proc_setattr,
2059 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2061 * @task: task that should be flushed.
2063 * Looks in the dcache for
2065 * /proc/@tgid/task/@pid
2066 * if either directory is present flushes it and all of it'ts children
2069 * It is safe and reasonable to cache /proc entries for a task until
2070 * that task exits. After that they just clog up the dcache with
2071 * useless entries, possibly causing useful dcache entries to be
2072 * flushed instead. This routine is proved to flush those useless
2073 * dcache entries at process exit time.
2075 * NOTE: This routine is just an optimization so it does not guarantee
2076 * that no dcache entries will exist at process exit time it
2077 * just makes it very unlikely that any will persist.
2079 void proc_flush_task(struct task_struct *task)
2081 struct dentry *dentry, *leader, *dir;
2082 char buf[PROC_NUMBUF];
2086 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
2087 dentry = d_hash_and_lookup(proc_mnt->mnt_root, &name);
2089 shrink_dcache_parent(dentry);
2094 if (thread_group_leader(task))
2098 name.len = snprintf(buf, sizeof(buf), "%d", task->tgid);
2099 leader = d_hash_and_lookup(proc_mnt->mnt_root, &name);
2104 name.len = strlen(name.name);
2105 dir = d_hash_and_lookup(leader, &name);
2107 goto out_put_leader;
2110 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
2111 dentry = d_hash_and_lookup(dir, &name);
2113 shrink_dcache_parent(dentry);
2125 static struct dentry *proc_pid_instantiate(struct inode *dir,
2126 struct dentry * dentry,
2127 struct task_struct *task, const void *ptr)
2129 struct dentry *error = ERR_PTR(-ENOENT);
2130 struct inode *inode;
2132 inode = proc_pid_make_inode(dir->i_sb, task);
2136 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2137 inode->i_op = &proc_tgid_base_inode_operations;
2138 inode->i_fop = &proc_tgid_base_operations;
2139 inode->i_flags|=S_IMMUTABLE;
2141 #ifdef CONFIG_SECURITY
2142 inode->i_nlink += 1;
2145 dentry->d_op = &pid_dentry_operations;
2147 d_add(dentry, inode);
2148 /* Close the race of the process dying before we return the dentry */
2149 if (pid_revalidate(dentry, NULL))
2155 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2157 struct dentry *result = ERR_PTR(-ENOENT);
2158 struct task_struct *task;
2161 result = proc_base_lookup(dir, dentry);
2162 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2165 tgid = name_to_int(dentry);
2170 task = find_task_by_pid(tgid);
2172 get_task_struct(task);
2177 result = proc_pid_instantiate(dir, dentry, task, NULL);
2178 put_task_struct(task);
2184 * Find the first task with tgid >= tgid
2187 static struct task_struct *next_tgid(unsigned int tgid)
2189 struct task_struct *task;
2195 pid = find_ge_pid(tgid);
2198 task = pid_task(pid, PIDTYPE_PID);
2199 /* What we to know is if the pid we have find is the
2200 * pid of a thread_group_leader. Testing for task
2201 * being a thread_group_leader is the obvious thing
2202 * todo but there is a window when it fails, due to
2203 * the pid transfer logic in de_thread.
2205 * So we perform the straight forward test of seeing
2206 * if the pid we have found is the pid of a thread
2207 * group leader, and don't worry if the task we have
2208 * found doesn't happen to be a thread group leader.
2209 * As we don't care in the case of readdir.
2211 if (!task || !has_group_leader_pid(task))
2213 get_task_struct(task);
2219 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2221 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2222 struct task_struct *task, int tgid)
2224 char name[PROC_NUMBUF];
2225 int len = snprintf(name, sizeof(name), "%d", tgid);
2226 return proc_fill_cache(filp, dirent, filldir, name, len,
2227 proc_pid_instantiate, task, NULL);
2230 /* for the /proc/ directory itself, after non-process stuff has been done */
2231 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2233 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2234 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
2235 struct task_struct *task;
2241 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
2242 const struct pid_entry *p = &proc_base_stuff[nr];
2243 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
2247 tgid = filp->f_pos - TGID_OFFSET;
2248 for (task = next_tgid(tgid);
2250 put_task_struct(task), task = next_tgid(tgid + 1)) {
2252 filp->f_pos = tgid + TGID_OFFSET;
2253 if (proc_pid_fill_cache(filp, dirent, filldir, task, tgid) < 0) {
2254 put_task_struct(task);
2258 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
2260 put_task_struct(reaper);
2268 static const struct pid_entry tid_base_stuff[] = {
2269 DIR("fd", S_IRUSR|S_IXUSR, fd),
2270 DIR("fdinfo", S_IRUSR|S_IXUSR, fdinfo),
2271 INF("environ", S_IRUSR, pid_environ),
2272 INF("auxv", S_IRUSR, pid_auxv),
2273 INF("status", S_IRUGO, pid_status),
2274 INF("cmdline", S_IRUGO, pid_cmdline),
2275 INF("stat", S_IRUGO, tid_stat),
2276 INF("statm", S_IRUGO, pid_statm),
2277 REG("maps", S_IRUGO, maps),
2279 REG("numa_maps", S_IRUGO, numa_maps),
2281 REG("mem", S_IRUSR|S_IWUSR, mem),
2282 #ifdef CONFIG_SECCOMP
2283 REG("seccomp", S_IRUSR|S_IWUSR, seccomp),
2288 REG("mounts", S_IRUGO, mounts),
2290 REG("clear_refs", S_IWUSR, clear_refs),
2291 REG("smaps", S_IRUGO, smaps),
2293 #ifdef CONFIG_SECURITY
2294 DIR("attr", S_IRUGO|S_IXUGO, attr_dir),
2296 #ifdef CONFIG_KALLSYMS
2297 INF("wchan", S_IRUGO, pid_wchan),
2299 #ifdef CONFIG_SCHEDSTATS
2300 INF("schedstat", S_IRUGO, pid_schedstat),
2302 #ifdef CONFIG_CPUSETS
2303 REG("cpuset", S_IRUGO, cpuset),
2305 INF("oom_score", S_IRUGO, oom_score),
2306 REG("oom_adj", S_IRUGO|S_IWUSR, oom_adjust),
2307 #ifdef CONFIG_AUDITSYSCALL
2308 REG("loginuid", S_IWUSR|S_IRUGO, loginuid),
2310 #ifdef CONFIG_FAULT_INJECTION
2311 REG("make-it-fail", S_IRUGO|S_IWUSR, fault_inject),
2315 static int proc_tid_base_readdir(struct file * filp,
2316 void * dirent, filldir_t filldir)
2318 return proc_pident_readdir(filp,dirent,filldir,
2319 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
2322 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2323 return proc_pident_lookup(dir, dentry,
2324 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
2327 static const struct file_operations proc_tid_base_operations = {
2328 .read = generic_read_dir,
2329 .readdir = proc_tid_base_readdir,
2332 static const struct inode_operations proc_tid_base_inode_operations = {
2333 .lookup = proc_tid_base_lookup,
2334 .getattr = pid_getattr,
2335 .setattr = proc_setattr,
2338 static struct dentry *proc_task_instantiate(struct inode *dir,
2339 struct dentry *dentry, struct task_struct *task, const void *ptr)
2341 struct dentry *error = ERR_PTR(-ENOENT);
2342 struct inode *inode;
2343 inode = proc_pid_make_inode(dir->i_sb, task);
2347 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2348 inode->i_op = &proc_tid_base_inode_operations;
2349 inode->i_fop = &proc_tid_base_operations;
2350 inode->i_flags|=S_IMMUTABLE;
2352 #ifdef CONFIG_SECURITY
2353 inode->i_nlink += 1;
2356 dentry->d_op = &pid_dentry_operations;
2358 d_add(dentry, inode);
2359 /* Close the race of the process dying before we return the dentry */
2360 if (pid_revalidate(dentry, NULL))
2366 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2368 struct dentry *result = ERR_PTR(-ENOENT);
2369 struct task_struct *task;
2370 struct task_struct *leader = get_proc_task(dir);
2376 tid = name_to_int(dentry);
2381 task = find_task_by_pid(tid);
2383 get_task_struct(task);
2387 if (leader->tgid != task->tgid)
2390 result = proc_task_instantiate(dir, dentry, task, NULL);
2392 put_task_struct(task);
2394 put_task_struct(leader);
2400 * Find the first tid of a thread group to return to user space.
2402 * Usually this is just the thread group leader, but if the users
2403 * buffer was too small or there was a seek into the middle of the
2404 * directory we have more work todo.
2406 * In the case of a short read we start with find_task_by_pid.
2408 * In the case of a seek we start with the leader and walk nr
2411 static struct task_struct *first_tid(struct task_struct *leader,
2414 struct task_struct *pos;
2417 /* Attempt to start with the pid of a thread */
2418 if (tid && (nr > 0)) {
2419 pos = find_task_by_pid(tid);
2420 if (pos && (pos->group_leader == leader))
2424 /* If nr exceeds the number of threads there is nothing todo */
2426 if (nr && nr >= get_nr_threads(leader))
2429 /* If we haven't found our starting place yet start
2430 * with the leader and walk nr threads forward.
2432 for (pos = leader; nr > 0; --nr) {
2433 pos = next_thread(pos);
2434 if (pos == leader) {
2440 get_task_struct(pos);
2447 * Find the next thread in the thread list.
2448 * Return NULL if there is an error or no next thread.
2450 * The reference to the input task_struct is released.
2452 static struct task_struct *next_tid(struct task_struct *start)
2454 struct task_struct *pos = NULL;
2456 if (pid_alive(start)) {
2457 pos = next_thread(start);
2458 if (thread_group_leader(pos))
2461 get_task_struct(pos);
2464 put_task_struct(start);
2468 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2469 struct task_struct *task, int tid)
2471 char name[PROC_NUMBUF];
2472 int len = snprintf(name, sizeof(name), "%d", tid);
2473 return proc_fill_cache(filp, dirent, filldir, name, len,
2474 proc_task_instantiate, task, NULL);
2477 /* for the /proc/TGID/task/ directories */
2478 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
2480 struct dentry *dentry = filp->f_path.dentry;
2481 struct inode *inode = dentry->d_inode;
2482 struct task_struct *leader = NULL;
2483 struct task_struct *task;
2484 int retval = -ENOENT;
2487 unsigned long pos = filp->f_pos; /* avoiding "long long" filp->f_pos */
2489 task = get_proc_task(inode);
2493 if (pid_alive(task)) {
2494 leader = task->group_leader;
2495 get_task_struct(leader);
2498 put_task_struct(task);
2506 if (filldir(dirent, ".", 1, pos, ino, DT_DIR) < 0)
2511 ino = parent_ino(dentry);
2512 if (filldir(dirent, "..", 2, pos, ino, DT_DIR) < 0)
2518 /* f_version caches the tgid value that the last readdir call couldn't
2519 * return. lseek aka telldir automagically resets f_version to 0.
2521 tid = filp->f_version;
2522 filp->f_version = 0;
2523 for (task = first_tid(leader, tid, pos - 2);
2525 task = next_tid(task), pos++) {
2527 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
2528 /* returning this tgid failed, save it as the first
2529 * pid for the next readir call */
2530 filp->f_version = tid;
2531 put_task_struct(task);
2537 put_task_struct(leader);
2542 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
2544 struct inode *inode = dentry->d_inode;
2545 struct task_struct *p = get_proc_task(inode);
2546 generic_fillattr(inode, stat);
2550 stat->nlink += get_nr_threads(p);
2558 static const struct inode_operations proc_task_inode_operations = {
2559 .lookup = proc_task_lookup,
2560 .getattr = proc_task_getattr,
2561 .setattr = proc_setattr,
2564 static const struct file_operations proc_task_operations = {
2565 .read = generic_read_dir,
2566 .readdir = proc_task_readdir,