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/namespace.h>
64 #include <linux/smp_lock.h>
65 #include <linux/rcupdate.h>
66 #include <linux/kallsyms.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 struct inode_operations *iop;
97 struct file_operations *fop;
101 #define NOD(NAME, MODE, IOP, FOP, OP) { \
102 .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 } )
126 static struct fs_struct *get_fs_struct(struct task_struct *task)
128 struct fs_struct *fs;
132 atomic_inc(&fs->count);
137 static int get_nr_threads(struct task_struct *tsk)
139 /* Must be called with the rcu_read_lock held */
143 if (lock_task_sighand(tsk, &flags)) {
144 count = atomic_read(&tsk->signal->count);
145 unlock_task_sighand(tsk, &flags);
150 static int proc_cwd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
152 struct task_struct *task = get_proc_task(inode);
153 struct fs_struct *fs = NULL;
154 int result = -ENOENT;
157 fs = get_fs_struct(task);
158 put_task_struct(task);
161 read_lock(&fs->lock);
162 *mnt = mntget(fs->pwdmnt);
163 *dentry = dget(fs->pwd);
164 read_unlock(&fs->lock);
171 static int proc_root_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
173 struct task_struct *task = get_proc_task(inode);
174 struct fs_struct *fs = NULL;
175 int result = -ENOENT;
178 fs = get_fs_struct(task);
179 put_task_struct(task);
182 read_lock(&fs->lock);
183 *mnt = mntget(fs->rootmnt);
184 *dentry = dget(fs->root);
185 read_unlock(&fs->lock);
192 #define MAY_PTRACE(task) \
193 (task == current || \
194 (task->parent == current && \
195 (task->ptrace & PT_PTRACED) && \
196 (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
197 security_ptrace(current,task) == 0))
199 static int proc_pid_environ(struct task_struct *task, char * buffer)
202 struct mm_struct *mm = get_task_mm(task);
204 unsigned int len = mm->env_end - mm->env_start;
207 res = access_process_vm(task, mm->env_start, buffer, len, 0);
208 if (!ptrace_may_attach(task))
215 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
219 struct mm_struct *mm = get_task_mm(task);
223 goto out_mm; /* Shh! No looking before we're done */
225 len = mm->arg_end - mm->arg_start;
230 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
232 // If the nul at the end of args has been overwritten, then
233 // assume application is using setproctitle(3).
234 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
235 len = strnlen(buffer, res);
239 len = mm->env_end - mm->env_start;
240 if (len > PAGE_SIZE - res)
241 len = PAGE_SIZE - res;
242 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
243 res = strnlen(buffer, res);
252 static int proc_pid_auxv(struct task_struct *task, char *buffer)
255 struct mm_struct *mm = get_task_mm(task);
257 unsigned int nwords = 0;
260 while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
261 res = nwords * sizeof(mm->saved_auxv[0]);
264 memcpy(buffer, mm->saved_auxv, res);
271 #ifdef CONFIG_KALLSYMS
273 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
274 * Returns the resolved symbol. If that fails, simply return the address.
276 static int proc_pid_wchan(struct task_struct *task, char *buffer)
279 const char *sym_name;
280 unsigned long wchan, size, offset;
281 char namebuf[KSYM_NAME_LEN+1];
283 wchan = get_wchan(task);
285 sym_name = kallsyms_lookup(wchan, &size, &offset, &modname, namebuf);
287 return sprintf(buffer, "%s", sym_name);
288 return sprintf(buffer, "%lu", wchan);
290 #endif /* CONFIG_KALLSYMS */
292 #ifdef CONFIG_SCHEDSTATS
294 * Provides /proc/PID/schedstat
296 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
298 return sprintf(buffer, "%lu %lu %lu\n",
299 task->sched_info.cpu_time,
300 task->sched_info.run_delay,
301 task->sched_info.pcnt);
305 /* The badness from the OOM killer */
306 unsigned long badness(struct task_struct *p, unsigned long uptime);
307 static int proc_oom_score(struct task_struct *task, char *buffer)
309 unsigned long points;
310 struct timespec uptime;
312 do_posix_clock_monotonic_gettime(&uptime);
313 points = badness(task, uptime.tv_sec);
314 return sprintf(buffer, "%lu\n", points);
317 /************************************************************************/
318 /* Here the fs part begins */
319 /************************************************************************/
321 /* permission checks */
322 static int proc_fd_access_allowed(struct inode *inode)
324 struct task_struct *task;
326 /* Allow access to a task's file descriptors if it is us or we
327 * may use ptrace attach to the process and find out that
330 task = get_proc_task(inode);
332 allowed = ptrace_may_attach(task);
333 put_task_struct(task);
338 static int proc_setattr(struct dentry *dentry, struct iattr *attr)
341 struct inode *inode = dentry->d_inode;
343 if (attr->ia_valid & ATTR_MODE)
346 error = inode_change_ok(inode, attr);
348 error = security_inode_setattr(dentry, attr);
350 error = inode_setattr(inode, attr);
355 static 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 namespace *namespace = NULL;
369 struct proc_mounts *p;
374 namespace = task->nsproxy->namespace;
376 get_namespace(namespace);
378 put_task_struct(task);
383 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
385 file->private_data = &p->m;
386 ret = seq_open(file, &mounts_op);
388 p->m.private = namespace;
389 p->event = namespace->event;
394 put_namespace(namespace);
399 static int mounts_release(struct inode *inode, struct file *file)
401 struct seq_file *m = file->private_data;
402 struct namespace *namespace = m->private;
403 put_namespace(namespace);
404 return seq_release(inode, file);
407 static unsigned mounts_poll(struct file *file, poll_table *wait)
409 struct proc_mounts *p = file->private_data;
410 struct namespace *ns = p->m.private;
413 poll_wait(file, &ns->poll, wait);
415 spin_lock(&vfsmount_lock);
416 if (p->event != ns->event) {
417 p->event = ns->event;
420 spin_unlock(&vfsmount_lock);
425 static struct file_operations proc_mounts_operations = {
429 .release = mounts_release,
433 extern struct seq_operations mountstats_op;
434 static int mountstats_open(struct inode *inode, struct file *file)
436 int ret = seq_open(file, &mountstats_op);
439 struct seq_file *m = file->private_data;
440 struct namespace *namespace = NULL;
441 struct task_struct *task = get_proc_task(inode);
446 namespace = task->nsproxy->namespace;
448 get_namespace(namespace);
450 put_task_struct(task);
454 m->private = namespace;
456 seq_release(inode, file);
463 static struct file_operations proc_mountstats_operations = {
464 .open = mountstats_open,
467 .release = mounts_release,
470 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
472 static ssize_t proc_info_read(struct file * file, char __user * buf,
473 size_t count, loff_t *ppos)
475 struct inode * inode = file->f_dentry->d_inode;
478 struct task_struct *task = get_proc_task(inode);
484 if (count > PROC_BLOCK_SIZE)
485 count = PROC_BLOCK_SIZE;
488 if (!(page = __get_free_page(GFP_KERNEL)))
491 length = PROC_I(inode)->op.proc_read(task, (char*)page);
494 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
497 put_task_struct(task);
502 static struct file_operations proc_info_file_operations = {
503 .read = proc_info_read,
506 static int mem_open(struct inode* inode, struct file* file)
508 file->private_data = (void*)((long)current->self_exec_id);
512 static ssize_t mem_read(struct file * file, char __user * buf,
513 size_t count, loff_t *ppos)
515 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
517 unsigned long src = *ppos;
519 struct mm_struct *mm;
524 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
528 page = (char *)__get_free_page(GFP_USER);
534 mm = get_task_mm(task);
540 if (file->private_data != (void*)((long)current->self_exec_id))
546 int this_len, retval;
548 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
549 retval = access_process_vm(task, src, page, this_len, 0);
550 if (!retval || !MAY_PTRACE(task) || !ptrace_may_attach(task)) {
556 if (copy_to_user(buf, page, retval)) {
571 free_page((unsigned long) page);
573 put_task_struct(task);
578 #define mem_write NULL
581 /* This is a security hazard */
582 static ssize_t mem_write(struct file * file, const char * buf,
583 size_t count, loff_t *ppos)
587 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
588 unsigned long dst = *ppos;
594 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
598 page = (char *)__get_free_page(GFP_USER);
604 int this_len, retval;
606 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
607 if (copy_from_user(page, buf, this_len)) {
611 retval = access_process_vm(task, dst, page, this_len, 1);
623 free_page((unsigned long) page);
625 put_task_struct(task);
631 static loff_t mem_lseek(struct file * file, loff_t offset, int orig)
635 file->f_pos = offset;
638 file->f_pos += offset;
643 force_successful_syscall_return();
647 static struct file_operations proc_mem_operations = {
654 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
655 size_t count, loff_t *ppos)
657 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
658 char buffer[PROC_NUMBUF];
661 loff_t __ppos = *ppos;
665 oom_adjust = task->oomkilladj;
666 put_task_struct(task);
668 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
671 if (count > len-__ppos)
673 if (copy_to_user(buf, buffer + __ppos, count))
675 *ppos = __ppos + count;
679 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
680 size_t count, loff_t *ppos)
682 struct task_struct *task;
683 char buffer[PROC_NUMBUF], *end;
686 memset(buffer, 0, sizeof(buffer));
687 if (count > sizeof(buffer) - 1)
688 count = sizeof(buffer) - 1;
689 if (copy_from_user(buffer, buf, count))
691 oom_adjust = simple_strtol(buffer, &end, 0);
692 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
693 oom_adjust != OOM_DISABLE)
697 task = get_proc_task(file->f_dentry->d_inode);
700 if (oom_adjust < task->oomkilladj && !capable(CAP_SYS_RESOURCE)) {
701 put_task_struct(task);
704 task->oomkilladj = oom_adjust;
705 put_task_struct(task);
706 if (end - buffer == 0)
711 static struct file_operations proc_oom_adjust_operations = {
712 .read = oom_adjust_read,
713 .write = oom_adjust_write,
716 #ifdef CONFIG_AUDITSYSCALL
718 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
719 size_t count, loff_t *ppos)
721 struct inode * inode = file->f_dentry->d_inode;
722 struct task_struct *task = get_proc_task(inode);
724 char tmpbuf[TMPBUFLEN];
728 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
729 audit_get_loginuid(task->audit_context));
730 put_task_struct(task);
731 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
734 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
735 size_t count, loff_t *ppos)
737 struct inode * inode = file->f_dentry->d_inode;
742 if (!capable(CAP_AUDIT_CONTROL))
745 if (current != pid_task(proc_pid(inode), PIDTYPE_PID))
748 if (count >= PAGE_SIZE)
749 count = PAGE_SIZE - 1;
752 /* No partial writes. */
755 page = (char*)__get_free_page(GFP_USER);
759 if (copy_from_user(page, buf, count))
763 loginuid = simple_strtoul(page, &tmp, 10);
769 length = audit_set_loginuid(current, loginuid);
770 if (likely(length == 0))
774 free_page((unsigned long) page);
778 static struct file_operations proc_loginuid_operations = {
779 .read = proc_loginuid_read,
780 .write = proc_loginuid_write,
784 #ifdef CONFIG_SECCOMP
785 static ssize_t seccomp_read(struct file *file, char __user *buf,
786 size_t count, loff_t *ppos)
788 struct task_struct *tsk = get_proc_task(file->f_dentry->d_inode);
790 loff_t __ppos = *ppos;
795 /* no need to print the trailing zero, so use only len */
796 len = sprintf(__buf, "%u\n", tsk->seccomp.mode);
797 put_task_struct(tsk);
800 if (count > len - __ppos)
801 count = len - __ppos;
802 if (copy_to_user(buf, __buf + __ppos, count))
804 *ppos = __ppos + count;
808 static ssize_t seccomp_write(struct file *file, const char __user *buf,
809 size_t count, loff_t *ppos)
811 struct task_struct *tsk = get_proc_task(file->f_dentry->d_inode);
812 char __buf[20], *end;
813 unsigned int seccomp_mode;
820 /* can set it only once to be even more secure */
822 if (unlikely(tsk->seccomp.mode))
826 memset(__buf, 0, sizeof(__buf));
827 count = min(count, sizeof(__buf) - 1);
828 if (copy_from_user(__buf, buf, count))
831 seccomp_mode = simple_strtoul(__buf, &end, 0);
835 if (seccomp_mode && seccomp_mode <= NR_SECCOMP_MODES) {
836 tsk->seccomp.mode = seccomp_mode;
837 set_tsk_thread_flag(tsk, TIF_SECCOMP);
841 if (unlikely(!(end - __buf)))
843 result = end - __buf;
845 put_task_struct(tsk);
850 static struct file_operations proc_seccomp_operations = {
851 .read = seccomp_read,
852 .write = seccomp_write,
854 #endif /* CONFIG_SECCOMP */
856 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
858 struct inode *inode = dentry->d_inode;
861 /* We don't need a base pointer in the /proc filesystem */
864 /* Are we allowed to snoop on the tasks file descriptors? */
865 if (!proc_fd_access_allowed(inode))
868 error = PROC_I(inode)->op.proc_get_link(inode, &nd->dentry, &nd->mnt);
869 nd->last_type = LAST_BIND;
871 return ERR_PTR(error);
874 static int do_proc_readlink(struct dentry *dentry, struct vfsmount *mnt,
875 char __user *buffer, int buflen)
877 struct inode * inode;
878 char *tmp = (char*)__get_free_page(GFP_KERNEL), *path;
884 inode = dentry->d_inode;
885 path = d_path(dentry, mnt, tmp, PAGE_SIZE);
889 len = tmp + PAGE_SIZE - 1 - path;
893 if (copy_to_user(buffer, path, len))
896 free_page((unsigned long)tmp);
900 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
903 struct inode *inode = dentry->d_inode;
905 struct vfsmount *mnt = NULL;
907 /* Are we allowed to snoop on the tasks file descriptors? */
908 if (!proc_fd_access_allowed(inode))
911 error = PROC_I(inode)->op.proc_get_link(inode, &de, &mnt);
915 error = do_proc_readlink(de, mnt, buffer, buflen);
922 static struct inode_operations proc_pid_link_inode_operations = {
923 .readlink = proc_pid_readlink,
924 .follow_link = proc_pid_follow_link,
925 .setattr = proc_setattr,
929 /* building an inode */
931 static int task_dumpable(struct task_struct *task)
934 struct mm_struct *mm;
939 dumpable = mm->dumpable;
947 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
949 struct inode * inode;
950 struct proc_inode *ei;
952 /* We need a new inode */
954 inode = new_inode(sb);
960 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
961 inode->i_op = &proc_def_inode_operations;
964 * grab the reference to task.
966 ei->pid = get_task_pid(task, PIDTYPE_PID);
972 if (task_dumpable(task)) {
973 inode->i_uid = task->euid;
974 inode->i_gid = task->egid;
976 security_task_to_inode(task, inode);
986 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
988 struct inode *inode = dentry->d_inode;
989 struct task_struct *task;
990 generic_fillattr(inode, stat);
995 task = pid_task(proc_pid(inode), PIDTYPE_PID);
997 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
998 task_dumpable(task)) {
999 stat->uid = task->euid;
1000 stat->gid = task->egid;
1010 * Exceptional case: normally we are not allowed to unhash a busy
1011 * directory. In this case, however, we can do it - no aliasing problems
1012 * due to the way we treat inodes.
1014 * Rewrite the inode's ownerships here because the owning task may have
1015 * performed a setuid(), etc.
1017 * Before the /proc/pid/status file was created the only way to read
1018 * the effective uid of a /process was to stat /proc/pid. Reading
1019 * /proc/pid/status is slow enough that procps and other packages
1020 * kept stating /proc/pid. To keep the rules in /proc simple I have
1021 * made this apply to all per process world readable and executable
1024 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1026 struct inode *inode = dentry->d_inode;
1027 struct task_struct *task = get_proc_task(inode);
1029 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1030 task_dumpable(task)) {
1031 inode->i_uid = task->euid;
1032 inode->i_gid = task->egid;
1037 inode->i_mode &= ~(S_ISUID | S_ISGID);
1038 security_task_to_inode(task, inode);
1039 put_task_struct(task);
1046 static int pid_delete_dentry(struct dentry * dentry)
1048 /* Is the task we represent dead?
1049 * If so, then don't put the dentry on the lru list,
1050 * kill it immediately.
1052 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1055 static struct dentry_operations pid_dentry_operations =
1057 .d_revalidate = pid_revalidate,
1058 .d_delete = pid_delete_dentry,
1063 typedef struct dentry *instantiate_t(struct inode *, struct dentry *, struct task_struct *, void *);
1066 * Fill a directory entry.
1068 * If possible create the dcache entry and derive our inode number and
1069 * file type from dcache entry.
1071 * Since all of the proc inode numbers are dynamically generated, the inode
1072 * numbers do not exist until the inode is cache. This means creating the
1073 * the dcache entry in readdir is necessary to keep the inode numbers
1074 * reported by readdir in sync with the inode numbers reported
1077 static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1078 char *name, int len,
1079 instantiate_t instantiate, struct task_struct *task, void *ptr)
1081 struct dentry *child, *dir = filp->f_dentry;
1082 struct inode *inode;
1085 unsigned type = DT_UNKNOWN;
1089 qname.hash = full_name_hash(name, len);
1091 child = d_lookup(dir, &qname);
1094 new = d_alloc(dir, &qname);
1096 child = instantiate(dir->d_inode, new, task, ptr);
1103 if (!child || IS_ERR(child) || !child->d_inode)
1104 goto end_instantiate;
1105 inode = child->d_inode;
1108 type = inode->i_mode >> 12;
1113 ino = find_inode_number(dir, &qname);
1116 return filldir(dirent, name, len, filp->f_pos, ino, type);
1119 static unsigned name_to_int(struct dentry *dentry)
1121 const char *name = dentry->d_name.name;
1122 int len = dentry->d_name.len;
1125 if (len > 1 && *name == '0')
1128 unsigned c = *name++ - '0';
1131 if (n >= (~0U-9)/10)
1141 static int proc_fd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
1143 struct task_struct *task = get_proc_task(inode);
1144 struct files_struct *files = NULL;
1146 int fd = proc_fd(inode);
1149 files = get_files_struct(task);
1150 put_task_struct(task);
1154 * We are not taking a ref to the file structure, so we must
1157 spin_lock(&files->file_lock);
1158 file = fcheck_files(files, fd);
1160 *mnt = mntget(file->f_vfsmnt);
1161 *dentry = dget(file->f_dentry);
1162 spin_unlock(&files->file_lock);
1163 put_files_struct(files);
1166 spin_unlock(&files->file_lock);
1167 put_files_struct(files);
1172 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1174 struct inode *inode = dentry->d_inode;
1175 struct task_struct *task = get_proc_task(inode);
1176 int fd = proc_fd(inode);
1177 struct files_struct *files;
1180 files = get_files_struct(task);
1183 if (fcheck_files(files, fd)) {
1185 put_files_struct(files);
1186 if (task_dumpable(task)) {
1187 inode->i_uid = task->euid;
1188 inode->i_gid = task->egid;
1193 inode->i_mode &= ~(S_ISUID | S_ISGID);
1194 security_task_to_inode(task, inode);
1195 put_task_struct(task);
1199 put_files_struct(files);
1201 put_task_struct(task);
1207 static struct dentry_operations tid_fd_dentry_operations =
1209 .d_revalidate = tid_fd_revalidate,
1210 .d_delete = pid_delete_dentry,
1213 static struct dentry *proc_fd_instantiate(struct inode *dir,
1214 struct dentry *dentry, struct task_struct *task, void *ptr)
1216 unsigned fd = *(unsigned *)ptr;
1218 struct files_struct *files;
1219 struct inode *inode;
1220 struct proc_inode *ei;
1221 struct dentry *error = ERR_PTR(-ENOENT);
1223 inode = proc_pid_make_inode(dir->i_sb, task);
1228 files = get_files_struct(task);
1231 inode->i_mode = S_IFLNK;
1234 * We are not taking a ref to the file structure, so we must
1237 spin_lock(&files->file_lock);
1238 file = fcheck_files(files, fd);
1241 if (file->f_mode & 1)
1242 inode->i_mode |= S_IRUSR | S_IXUSR;
1243 if (file->f_mode & 2)
1244 inode->i_mode |= S_IWUSR | S_IXUSR;
1245 spin_unlock(&files->file_lock);
1246 put_files_struct(files);
1248 inode->i_op = &proc_pid_link_inode_operations;
1250 ei->op.proc_get_link = proc_fd_link;
1251 dentry->d_op = &tid_fd_dentry_operations;
1252 d_add(dentry, inode);
1253 /* Close the race of the process dying before we return the dentry */
1254 if (tid_fd_revalidate(dentry, NULL))
1260 spin_unlock(&files->file_lock);
1261 put_files_struct(files);
1267 static struct dentry *proc_lookupfd(struct inode * dir, struct dentry * dentry, struct nameidata *nd)
1269 struct task_struct *task = get_proc_task(dir);
1270 unsigned fd = name_to_int(dentry);
1271 struct dentry *result = ERR_PTR(-ENOENT);
1278 result = proc_fd_instantiate(dir, dentry, task, &fd);
1280 put_task_struct(task);
1285 static int proc_fd_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1286 struct task_struct *task, int fd)
1288 char name[PROC_NUMBUF];
1289 int len = snprintf(name, sizeof(name), "%d", fd);
1290 return proc_fill_cache(filp, dirent, filldir, name, len,
1291 proc_fd_instantiate, task, &fd);
1294 static int proc_readfd(struct file * filp, void * dirent, filldir_t filldir)
1296 struct dentry *dentry = filp->f_dentry;
1297 struct inode *inode = dentry->d_inode;
1298 struct task_struct *p = get_proc_task(inode);
1299 unsigned int fd, tid, ino;
1301 struct files_struct * files;
1302 struct fdtable *fdt;
1313 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1317 ino = parent_ino(dentry);
1318 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1322 files = get_files_struct(p);
1326 fdt = files_fdtable(files);
1327 for (fd = filp->f_pos-2;
1329 fd++, filp->f_pos++) {
1331 if (!fcheck_files(files, fd))
1335 if (proc_fd_fill_cache(filp, dirent, filldir, p, fd) < 0) {
1342 put_files_struct(files);
1350 static struct file_operations proc_fd_operations = {
1351 .read = generic_read_dir,
1352 .readdir = proc_readfd,
1356 * proc directories can do almost nothing..
1358 static struct inode_operations proc_fd_inode_operations = {
1359 .lookup = proc_lookupfd,
1360 .setattr = proc_setattr,
1363 static struct dentry *proc_pident_instantiate(struct inode *dir,
1364 struct dentry *dentry, struct task_struct *task, void *ptr)
1366 struct pid_entry *p = ptr;
1367 struct inode *inode;
1368 struct proc_inode *ei;
1369 struct dentry *error = ERR_PTR(-EINVAL);
1371 inode = proc_pid_make_inode(dir->i_sb, task);
1376 inode->i_mode = p->mode;
1377 if (S_ISDIR(inode->i_mode))
1378 inode->i_nlink = 2; /* Use getattr to fix if necessary */
1380 inode->i_op = p->iop;
1382 inode->i_fop = p->fop;
1384 dentry->d_op = &pid_dentry_operations;
1385 d_add(dentry, inode);
1386 /* Close the race of the process dying before we return the dentry */
1387 if (pid_revalidate(dentry, NULL))
1393 static struct dentry *proc_pident_lookup(struct inode *dir,
1394 struct dentry *dentry,
1395 struct pid_entry *ents,
1398 struct inode *inode;
1399 struct dentry *error;
1400 struct task_struct *task = get_proc_task(dir);
1401 struct pid_entry *p, *last;
1403 error = ERR_PTR(-ENOENT);
1410 * Yes, it does not scale. And it should not. Don't add
1411 * new entries into /proc/<tgid>/ without very good reasons.
1413 last = &ents[nents - 1];
1414 for (p = ents; p <= last; p++) {
1415 if (p->len != dentry->d_name.len)
1417 if (!memcmp(dentry->d_name.name, p->name, p->len))
1423 error = proc_pident_instantiate(dir, dentry, task, p);
1425 put_task_struct(task);
1430 static int proc_pident_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1431 struct task_struct *task, struct pid_entry *p)
1433 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
1434 proc_pident_instantiate, task, p);
1437 static int proc_pident_readdir(struct file *filp,
1438 void *dirent, filldir_t filldir,
1439 struct pid_entry *ents, unsigned int nents)
1443 struct dentry *dentry = filp->f_dentry;
1444 struct inode *inode = dentry->d_inode;
1445 struct task_struct *task = get_proc_task(inode);
1446 struct pid_entry *p, *last;
1460 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
1466 ino = parent_ino(dentry);
1467 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
1479 last = &ents[nents - 1];
1481 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
1490 put_task_struct(task);
1495 #ifdef CONFIG_SECURITY
1496 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
1497 size_t count, loff_t *ppos)
1499 struct inode * inode = file->f_dentry->d_inode;
1502 struct task_struct *task = get_proc_task(inode);
1508 if (count > PAGE_SIZE)
1511 if (!(page = __get_free_page(GFP_KERNEL)))
1514 length = security_getprocattr(task,
1515 (char*)file->f_dentry->d_name.name,
1516 (void*)page, count);
1518 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
1521 put_task_struct(task);
1526 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
1527 size_t count, loff_t *ppos)
1529 struct inode * inode = file->f_dentry->d_inode;
1532 struct task_struct *task = get_proc_task(inode);
1537 if (count > PAGE_SIZE)
1540 /* No partial writes. */
1546 page = (char*)__get_free_page(GFP_USER);
1551 if (copy_from_user(page, buf, count))
1554 length = security_setprocattr(task,
1555 (char*)file->f_dentry->d_name.name,
1556 (void*)page, count);
1558 free_page((unsigned long) page);
1560 put_task_struct(task);
1565 static struct file_operations proc_pid_attr_operations = {
1566 .read = proc_pid_attr_read,
1567 .write = proc_pid_attr_write,
1570 static struct pid_entry attr_dir_stuff[] = {
1571 REG("current", S_IRUGO|S_IWUGO, pid_attr),
1572 REG("prev", S_IRUGO, pid_attr),
1573 REG("exec", S_IRUGO|S_IWUGO, pid_attr),
1574 REG("fscreate", S_IRUGO|S_IWUGO, pid_attr),
1575 REG("keycreate", S_IRUGO|S_IWUGO, pid_attr),
1576 REG("sockcreate", S_IRUGO|S_IWUGO, pid_attr),
1579 static int proc_attr_dir_readdir(struct file * filp,
1580 void * dirent, filldir_t filldir)
1582 return proc_pident_readdir(filp,dirent,filldir,
1583 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
1586 static struct file_operations proc_attr_dir_operations = {
1587 .read = generic_read_dir,
1588 .readdir = proc_attr_dir_readdir,
1591 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
1592 struct dentry *dentry, struct nameidata *nd)
1594 return proc_pident_lookup(dir, dentry,
1595 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
1598 static struct inode_operations proc_attr_dir_inode_operations = {
1599 .lookup = proc_attr_dir_lookup,
1600 .getattr = pid_getattr,
1601 .setattr = proc_setattr,
1609 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
1612 char tmp[PROC_NUMBUF];
1613 sprintf(tmp, "%d", current->tgid);
1614 return vfs_readlink(dentry,buffer,buflen,tmp);
1617 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
1619 char tmp[PROC_NUMBUF];
1620 sprintf(tmp, "%d", current->tgid);
1621 return ERR_PTR(vfs_follow_link(nd,tmp));
1624 static struct inode_operations proc_self_inode_operations = {
1625 .readlink = proc_self_readlink,
1626 .follow_link = proc_self_follow_link,
1632 * These are the directory entries in the root directory of /proc
1633 * that properly belong to the /proc filesystem, as they describe
1634 * describe something that is process related.
1636 static struct pid_entry proc_base_stuff[] = {
1637 NOD("self", S_IFLNK|S_IRWXUGO,
1638 &proc_self_inode_operations, NULL, {}),
1642 * Exceptional case: normally we are not allowed to unhash a busy
1643 * directory. In this case, however, we can do it - no aliasing problems
1644 * due to the way we treat inodes.
1646 static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
1648 struct inode *inode = dentry->d_inode;
1649 struct task_struct *task = get_proc_task(inode);
1651 put_task_struct(task);
1658 static struct dentry_operations proc_base_dentry_operations =
1660 .d_revalidate = proc_base_revalidate,
1661 .d_delete = pid_delete_dentry,
1664 static struct dentry *proc_base_instantiate(struct inode *dir,
1665 struct dentry *dentry, struct task_struct *task, void *ptr)
1667 struct pid_entry *p = ptr;
1668 struct inode *inode;
1669 struct proc_inode *ei;
1670 struct dentry *error = ERR_PTR(-EINVAL);
1672 /* Allocate the inode */
1673 error = ERR_PTR(-ENOMEM);
1674 inode = new_inode(dir->i_sb);
1678 /* Initialize the inode */
1680 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1683 * grab the reference to the task.
1685 ei->pid = get_task_pid(task, PIDTYPE_PID);
1691 inode->i_mode = p->mode;
1692 if (S_ISDIR(inode->i_mode))
1694 if (S_ISLNK(inode->i_mode))
1697 inode->i_op = p->iop;
1699 inode->i_fop = p->fop;
1701 dentry->d_op = &proc_base_dentry_operations;
1702 d_add(dentry, inode);
1711 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
1713 struct dentry *error;
1714 struct task_struct *task = get_proc_task(dir);
1715 struct pid_entry *p, *last;
1717 error = ERR_PTR(-ENOENT);
1722 /* Lookup the directory entry */
1723 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
1724 for (p = proc_base_stuff; p <= last; p++) {
1725 if (p->len != dentry->d_name.len)
1727 if (!memcmp(dentry->d_name.name, p->name, p->len))
1733 error = proc_base_instantiate(dir, dentry, task, p);
1736 put_task_struct(task);
1741 static int proc_base_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1742 struct task_struct *task, struct pid_entry *p)
1744 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
1745 proc_base_instantiate, task, p);
1751 static struct file_operations proc_task_operations;
1752 static struct inode_operations proc_task_inode_operations;
1754 static struct pid_entry tgid_base_stuff[] = {
1755 DIR("task", S_IRUGO|S_IXUGO, task),
1756 DIR("fd", S_IRUSR|S_IXUSR, fd),
1757 INF("environ", S_IRUSR, pid_environ),
1758 INF("auxv", S_IRUSR, pid_auxv),
1759 INF("status", S_IRUGO, pid_status),
1760 INF("cmdline", S_IRUGO, pid_cmdline),
1761 INF("stat", S_IRUGO, tgid_stat),
1762 INF("statm", S_IRUGO, pid_statm),
1763 REG("maps", S_IRUGO, maps),
1765 REG("numa_maps", S_IRUGO, numa_maps),
1767 REG("mem", S_IRUSR|S_IWUSR, mem),
1768 #ifdef CONFIG_SECCOMP
1769 REG("seccomp", S_IRUSR|S_IWUSR, seccomp),
1774 REG("mounts", S_IRUGO, mounts),
1775 REG("mountstats", S_IRUSR, mountstats),
1777 REG("smaps", S_IRUGO, smaps),
1779 #ifdef CONFIG_SECURITY
1780 DIR("attr", S_IRUGO|S_IXUGO, attr_dir),
1782 #ifdef CONFIG_KALLSYMS
1783 INF("wchan", S_IRUGO, pid_wchan),
1785 #ifdef CONFIG_SCHEDSTATS
1786 INF("schedstat", S_IRUGO, pid_schedstat),
1788 #ifdef CONFIG_CPUSETS
1789 REG("cpuset", S_IRUGO, cpuset),
1791 INF("oom_score", S_IRUGO, oom_score),
1792 REG("oom_adj", S_IRUGO|S_IWUSR, oom_adjust),
1793 #ifdef CONFIG_AUDITSYSCALL
1794 REG("loginuid", S_IWUSR|S_IRUGO, loginuid),
1798 static int proc_tgid_base_readdir(struct file * filp,
1799 void * dirent, filldir_t filldir)
1801 return proc_pident_readdir(filp,dirent,filldir,
1802 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1805 static struct file_operations proc_tgid_base_operations = {
1806 .read = generic_read_dir,
1807 .readdir = proc_tgid_base_readdir,
1810 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
1811 return proc_pident_lookup(dir, dentry,
1812 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1815 static struct inode_operations proc_tgid_base_inode_operations = {
1816 .lookup = proc_tgid_base_lookup,
1817 .getattr = pid_getattr,
1818 .setattr = proc_setattr,
1822 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
1824 * @task: task that should be flushed.
1826 * Looks in the dcache for
1828 * /proc/@tgid/task/@pid
1829 * if either directory is present flushes it and all of it'ts children
1832 * It is safe and reasonable to cache /proc entries for a task until
1833 * that task exits. After that they just clog up the dcache with
1834 * useless entries, possibly causing useful dcache entries to be
1835 * flushed instead. This routine is proved to flush those useless
1836 * dcache entries at process exit time.
1838 * NOTE: This routine is just an optimization so it does not guarantee
1839 * that no dcache entries will exist at process exit time it
1840 * just makes it very unlikely that any will persist.
1842 void proc_flush_task(struct task_struct *task)
1844 struct dentry *dentry, *leader, *dir;
1845 char buf[PROC_NUMBUF];
1849 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
1850 dentry = d_hash_and_lookup(proc_mnt->mnt_root, &name);
1852 shrink_dcache_parent(dentry);
1857 if (thread_group_leader(task))
1861 name.len = snprintf(buf, sizeof(buf), "%d", task->tgid);
1862 leader = d_hash_and_lookup(proc_mnt->mnt_root, &name);
1867 name.len = strlen(name.name);
1868 dir = d_hash_and_lookup(leader, &name);
1870 goto out_put_leader;
1873 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
1874 dentry = d_hash_and_lookup(dir, &name);
1876 shrink_dcache_parent(dentry);
1888 static struct dentry *proc_pid_instantiate(struct inode *dir,
1889 struct dentry * dentry,
1890 struct task_struct *task, void *ptr)
1892 struct dentry *error = ERR_PTR(-ENOENT);
1893 struct inode *inode;
1895 inode = proc_pid_make_inode(dir->i_sb, task);
1899 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
1900 inode->i_op = &proc_tgid_base_inode_operations;
1901 inode->i_fop = &proc_tgid_base_operations;
1902 inode->i_flags|=S_IMMUTABLE;
1904 #ifdef CONFIG_SECURITY
1905 inode->i_nlink += 1;
1908 dentry->d_op = &pid_dentry_operations;
1910 d_add(dentry, inode);
1911 /* Close the race of the process dying before we return the dentry */
1912 if (pid_revalidate(dentry, NULL))
1918 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
1920 struct dentry *result = ERR_PTR(-ENOENT);
1921 struct task_struct *task;
1924 result = proc_base_lookup(dir, dentry);
1925 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
1928 tgid = name_to_int(dentry);
1933 task = find_task_by_pid(tgid);
1935 get_task_struct(task);
1940 result = proc_pid_instantiate(dir, dentry, task, NULL);
1941 put_task_struct(task);
1947 * Find the first task with tgid >= tgid
1950 static struct task_struct *next_tgid(unsigned int tgid)
1952 struct task_struct *task;
1958 pid = find_ge_pid(tgid);
1961 task = pid_task(pid, PIDTYPE_PID);
1962 /* What we to know is if the pid we have find is the
1963 * pid of a thread_group_leader. Testing for task
1964 * being a thread_group_leader is the obvious thing
1965 * todo but there is a window when it fails, due to
1966 * the pid transfer logic in de_thread.
1968 * So we perform the straight forward test of seeing
1969 * if the pid we have found is the pid of a thread
1970 * group leader, and don't worry if the task we have
1971 * found doesn't happen to be a thread group leader.
1972 * As we don't care in the case of readdir.
1974 if (!task || !has_group_leader_pid(task))
1976 get_task_struct(task);
1982 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
1984 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1985 struct task_struct *task, int tgid)
1987 char name[PROC_NUMBUF];
1988 int len = snprintf(name, sizeof(name), "%d", tgid);
1989 return proc_fill_cache(filp, dirent, filldir, name, len,
1990 proc_pid_instantiate, task, NULL);
1993 /* for the /proc/ directory itself, after non-process stuff has been done */
1994 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
1996 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
1997 struct task_struct *reaper = get_proc_task(filp->f_dentry->d_inode);
1998 struct task_struct *task;
2004 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
2005 struct pid_entry *p = &proc_base_stuff[nr];
2006 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
2010 tgid = filp->f_pos - TGID_OFFSET;
2011 for (task = next_tgid(tgid);
2013 put_task_struct(task), task = next_tgid(tgid + 1)) {
2015 filp->f_pos = tgid + TGID_OFFSET;
2016 if (proc_pid_fill_cache(filp, dirent, filldir, task, tgid) < 0) {
2017 put_task_struct(task);
2021 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
2023 put_task_struct(reaper);
2031 static struct pid_entry tid_base_stuff[] = {
2032 DIR("fd", S_IRUSR|S_IXUSR, fd),
2033 INF("environ", S_IRUSR, pid_environ),
2034 INF("auxv", S_IRUSR, pid_auxv),
2035 INF("status", S_IRUGO, pid_status),
2036 INF("cmdline", S_IRUGO, pid_cmdline),
2037 INF("stat", S_IRUGO, tid_stat),
2038 INF("statm", S_IRUGO, pid_statm),
2039 REG("maps", S_IRUGO, maps),
2041 REG("numa_maps", S_IRUGO, numa_maps),
2043 REG("mem", S_IRUSR|S_IWUSR, mem),
2044 #ifdef CONFIG_SECCOMP
2045 REG("seccomp", S_IRUSR|S_IWUSR, seccomp),
2050 REG("mounts", S_IRUGO, mounts),
2052 REG("smaps", S_IRUGO, smaps),
2054 #ifdef CONFIG_SECURITY
2055 DIR("attr", S_IRUGO|S_IXUGO, attr_dir),
2057 #ifdef CONFIG_KALLSYMS
2058 INF("wchan", S_IRUGO, pid_wchan),
2060 #ifdef CONFIG_SCHEDSTATS
2061 INF("schedstat", S_IRUGO, pid_schedstat),
2063 #ifdef CONFIG_CPUSETS
2064 REG("cpuset", S_IRUGO, cpuset),
2066 INF("oom_score", S_IRUGO, oom_score),
2067 REG("oom_adj", S_IRUGO|S_IWUSR, oom_adjust),
2068 #ifdef CONFIG_AUDITSYSCALL
2069 REG("loginuid", S_IWUSR|S_IRUGO, loginuid),
2073 static int proc_tid_base_readdir(struct file * filp,
2074 void * dirent, filldir_t filldir)
2076 return proc_pident_readdir(filp,dirent,filldir,
2077 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
2080 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2081 return proc_pident_lookup(dir, dentry,
2082 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
2085 static struct file_operations proc_tid_base_operations = {
2086 .read = generic_read_dir,
2087 .readdir = proc_tid_base_readdir,
2090 static struct inode_operations proc_tid_base_inode_operations = {
2091 .lookup = proc_tid_base_lookup,
2092 .getattr = pid_getattr,
2093 .setattr = proc_setattr,
2096 static struct dentry *proc_task_instantiate(struct inode *dir,
2097 struct dentry *dentry, struct task_struct *task, void *ptr)
2099 struct dentry *error = ERR_PTR(-ENOENT);
2100 struct inode *inode;
2101 inode = proc_pid_make_inode(dir->i_sb, task);
2105 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2106 inode->i_op = &proc_tid_base_inode_operations;
2107 inode->i_fop = &proc_tid_base_operations;
2108 inode->i_flags|=S_IMMUTABLE;
2110 #ifdef CONFIG_SECURITY
2111 inode->i_nlink += 1;
2114 dentry->d_op = &pid_dentry_operations;
2116 d_add(dentry, inode);
2117 /* Close the race of the process dying before we return the dentry */
2118 if (pid_revalidate(dentry, NULL))
2124 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2126 struct dentry *result = ERR_PTR(-ENOENT);
2127 struct task_struct *task;
2128 struct task_struct *leader = get_proc_task(dir);
2134 tid = name_to_int(dentry);
2139 task = find_task_by_pid(tid);
2141 get_task_struct(task);
2145 if (leader->tgid != task->tgid)
2148 result = proc_task_instantiate(dir, dentry, task, NULL);
2150 put_task_struct(task);
2152 put_task_struct(leader);
2158 * Find the first tid of a thread group to return to user space.
2160 * Usually this is just the thread group leader, but if the users
2161 * buffer was too small or there was a seek into the middle of the
2162 * directory we have more work todo.
2164 * In the case of a short read we start with find_task_by_pid.
2166 * In the case of a seek we start with the leader and walk nr
2169 static struct task_struct *first_tid(struct task_struct *leader,
2172 struct task_struct *pos;
2175 /* Attempt to start with the pid of a thread */
2176 if (tid && (nr > 0)) {
2177 pos = find_task_by_pid(tid);
2178 if (pos && (pos->group_leader == leader))
2182 /* If nr exceeds the number of threads there is nothing todo */
2184 if (nr && nr >= get_nr_threads(leader))
2187 /* If we haven't found our starting place yet start
2188 * with the leader and walk nr threads forward.
2190 for (pos = leader; nr > 0; --nr) {
2191 pos = next_thread(pos);
2192 if (pos == leader) {
2198 get_task_struct(pos);
2205 * Find the next thread in the thread list.
2206 * Return NULL if there is an error or no next thread.
2208 * The reference to the input task_struct is released.
2210 static struct task_struct *next_tid(struct task_struct *start)
2212 struct task_struct *pos = NULL;
2214 if (pid_alive(start)) {
2215 pos = next_thread(start);
2216 if (thread_group_leader(pos))
2219 get_task_struct(pos);
2222 put_task_struct(start);
2226 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2227 struct task_struct *task, int tid)
2229 char name[PROC_NUMBUF];
2230 int len = snprintf(name, sizeof(name), "%d", tid);
2231 return proc_fill_cache(filp, dirent, filldir, name, len,
2232 proc_task_instantiate, task, NULL);
2235 /* for the /proc/TGID/task/ directories */
2236 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
2238 struct dentry *dentry = filp->f_dentry;
2239 struct inode *inode = dentry->d_inode;
2240 struct task_struct *leader = get_proc_task(inode);
2241 struct task_struct *task;
2242 int retval = -ENOENT;
2245 unsigned long pos = filp->f_pos; /* avoiding "long long" filp->f_pos */
2254 if (filldir(dirent, ".", 1, pos, ino, DT_DIR) < 0)
2259 ino = parent_ino(dentry);
2260 if (filldir(dirent, "..", 2, pos, ino, DT_DIR) < 0)
2266 /* f_version caches the tgid value that the last readdir call couldn't
2267 * return. lseek aka telldir automagically resets f_version to 0.
2269 tid = filp->f_version;
2270 filp->f_version = 0;
2271 for (task = first_tid(leader, tid, pos - 2);
2273 task = next_tid(task), pos++) {
2275 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
2276 /* returning this tgid failed, save it as the first
2277 * pid for the next readir call */
2278 filp->f_version = tid;
2279 put_task_struct(task);
2285 put_task_struct(leader);
2290 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
2292 struct inode *inode = dentry->d_inode;
2293 struct task_struct *p = get_proc_task(inode);
2294 generic_fillattr(inode, stat);
2298 stat->nlink += get_nr_threads(p);
2306 static struct inode_operations proc_task_inode_operations = {
2307 .lookup = proc_task_lookup,
2308 .getattr = proc_task_getattr,
2309 .setattr = proc_setattr,
2312 static struct file_operations proc_task_operations = {
2313 .read = generic_read_dir,
2314 .readdir = proc_task_readdir,