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/cpuset.h>
71 #include <linux/audit.h>
72 #include <linux/poll.h>
73 #include <linux/nsproxy.h>
74 #include <linux/oom.h>
75 #include <linux/elf.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);
210 if (!ptrace_may_attach(task))
213 len = mm->env_end - mm->env_start;
216 res = access_process_vm(task, mm->env_start, buffer, len, 0);
223 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
227 struct mm_struct *mm = get_task_mm(task);
231 goto out_mm; /* Shh! No looking before we're done */
233 len = mm->arg_end - mm->arg_start;
238 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
240 // If the nul at the end of args has been overwritten, then
241 // assume application is using setproctitle(3).
242 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
243 len = strnlen(buffer, res);
247 len = mm->env_end - mm->env_start;
248 if (len > PAGE_SIZE - res)
249 len = PAGE_SIZE - res;
250 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
251 res = strnlen(buffer, res);
260 static int proc_pid_auxv(struct task_struct *task, char *buffer)
263 struct mm_struct *mm = get_task_mm(task);
265 unsigned int nwords = 0;
268 while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
269 res = nwords * sizeof(mm->saved_auxv[0]);
272 memcpy(buffer, mm->saved_auxv, res);
279 #ifdef CONFIG_KALLSYMS
281 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
282 * Returns the resolved symbol. If that fails, simply return the address.
284 static int proc_pid_wchan(struct task_struct *task, char *buffer)
287 char symname[KSYM_NAME_LEN];
289 wchan = get_wchan(task);
291 if (lookup_symbol_name(wchan, symname) < 0)
292 return sprintf(buffer, "%lu", wchan);
294 return sprintf(buffer, "%s", symname);
296 #endif /* CONFIG_KALLSYMS */
298 #ifdef CONFIG_SCHEDSTATS
300 * Provides /proc/PID/schedstat
302 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
304 return sprintf(buffer, "%llu %llu %lu\n",
305 task->sched_info.cpu_time,
306 task->sched_info.run_delay,
307 task->sched_info.pcount);
311 /* The badness from the OOM killer */
312 unsigned long badness(struct task_struct *p, unsigned long uptime);
313 static int proc_oom_score(struct task_struct *task, char *buffer)
315 unsigned long points;
316 struct timespec uptime;
318 do_posix_clock_monotonic_gettime(&uptime);
319 read_lock(&tasklist_lock);
320 points = badness(task, uptime.tv_sec);
321 read_unlock(&tasklist_lock);
322 return sprintf(buffer, "%lu\n", points);
325 /************************************************************************/
326 /* Here the fs part begins */
327 /************************************************************************/
329 /* permission checks */
330 static int proc_fd_access_allowed(struct inode *inode)
332 struct task_struct *task;
334 /* Allow access to a task's file descriptors if it is us or we
335 * may use ptrace attach to the process and find out that
338 task = get_proc_task(inode);
340 allowed = ptrace_may_attach(task);
341 put_task_struct(task);
346 static int proc_setattr(struct dentry *dentry, struct iattr *attr)
349 struct inode *inode = dentry->d_inode;
351 if (attr->ia_valid & ATTR_MODE)
354 error = inode_change_ok(inode, attr);
356 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_TEMPORARY)))
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_TEMPORARY);
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_TEMPORARY);
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];
671 oom_adjust = task->oomkilladj;
672 put_task_struct(task);
674 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
676 return simple_read_from_buffer(buf, count, ppos, buffer, len);
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_path.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 const struct file_operations proc_oom_adjust_operations = {
712 .read = oom_adjust_read,
713 .write = oom_adjust_write,
717 static ssize_t clear_refs_write(struct file *file, const char __user *buf,
718 size_t count, loff_t *ppos)
720 struct task_struct *task;
721 char buffer[PROC_NUMBUF], *end;
722 struct mm_struct *mm;
724 memset(buffer, 0, sizeof(buffer));
725 if (count > sizeof(buffer) - 1)
726 count = sizeof(buffer) - 1;
727 if (copy_from_user(buffer, buf, count))
729 if (!simple_strtol(buffer, &end, 0))
733 task = get_proc_task(file->f_path.dentry->d_inode);
736 mm = get_task_mm(task);
741 put_task_struct(task);
742 if (end - buffer == 0)
747 static struct file_operations proc_clear_refs_operations = {
748 .write = clear_refs_write,
752 #ifdef CONFIG_AUDITSYSCALL
754 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
755 size_t count, loff_t *ppos)
757 struct inode * inode = file->f_path.dentry->d_inode;
758 struct task_struct *task = get_proc_task(inode);
760 char tmpbuf[TMPBUFLEN];
764 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
765 audit_get_loginuid(task->audit_context));
766 put_task_struct(task);
767 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
770 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
771 size_t count, loff_t *ppos)
773 struct inode * inode = file->f_path.dentry->d_inode;
778 if (!capable(CAP_AUDIT_CONTROL))
781 if (current != pid_task(proc_pid(inode), PIDTYPE_PID))
784 if (count >= PAGE_SIZE)
785 count = PAGE_SIZE - 1;
788 /* No partial writes. */
791 page = (char*)__get_free_page(GFP_TEMPORARY);
795 if (copy_from_user(page, buf, count))
799 loginuid = simple_strtoul(page, &tmp, 10);
805 length = audit_set_loginuid(current, loginuid);
806 if (likely(length == 0))
810 free_page((unsigned long) page);
814 static const struct file_operations proc_loginuid_operations = {
815 .read = proc_loginuid_read,
816 .write = proc_loginuid_write,
820 #ifdef CONFIG_FAULT_INJECTION
821 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
822 size_t count, loff_t *ppos)
824 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
825 char buffer[PROC_NUMBUF];
831 make_it_fail = task->make_it_fail;
832 put_task_struct(task);
834 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
836 return simple_read_from_buffer(buf, count, ppos, buffer, len);
839 static ssize_t proc_fault_inject_write(struct file * file,
840 const char __user * buf, size_t count, loff_t *ppos)
842 struct task_struct *task;
843 char buffer[PROC_NUMBUF], *end;
846 if (!capable(CAP_SYS_RESOURCE))
848 memset(buffer, 0, sizeof(buffer));
849 if (count > sizeof(buffer) - 1)
850 count = sizeof(buffer) - 1;
851 if (copy_from_user(buffer, buf, count))
853 make_it_fail = simple_strtol(buffer, &end, 0);
856 task = get_proc_task(file->f_dentry->d_inode);
859 task->make_it_fail = make_it_fail;
860 put_task_struct(task);
861 if (end - buffer == 0)
866 static const struct file_operations proc_fault_inject_operations = {
867 .read = proc_fault_inject_read,
868 .write = proc_fault_inject_write,
872 #ifdef CONFIG_SCHED_DEBUG
874 * Print out various scheduling related per-task fields:
876 static int sched_show(struct seq_file *m, void *v)
878 struct inode *inode = m->private;
879 struct task_struct *p;
883 p = get_proc_task(inode);
886 proc_sched_show_task(p, m);
894 sched_write(struct file *file, const char __user *buf,
895 size_t count, loff_t *offset)
897 struct inode *inode = file->f_path.dentry->d_inode;
898 struct task_struct *p;
902 p = get_proc_task(inode);
905 proc_sched_set_task(p);
912 static int sched_open(struct inode *inode, struct file *filp)
916 ret = single_open(filp, sched_show, NULL);
918 struct seq_file *m = filp->private_data;
925 static const struct file_operations proc_pid_sched_operations = {
928 .write = sched_write,
930 .release = single_release,
935 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
937 struct inode *inode = dentry->d_inode;
940 /* We don't need a base pointer in the /proc filesystem */
943 /* Are we allowed to snoop on the tasks file descriptors? */
944 if (!proc_fd_access_allowed(inode))
947 error = PROC_I(inode)->op.proc_get_link(inode, &nd->dentry, &nd->mnt);
948 nd->last_type = LAST_BIND;
950 return ERR_PTR(error);
953 static int do_proc_readlink(struct dentry *dentry, struct vfsmount *mnt,
954 char __user *buffer, int buflen)
956 struct inode * inode;
957 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
964 inode = dentry->d_inode;
965 path = d_path(dentry, mnt, tmp, PAGE_SIZE);
969 len = tmp + PAGE_SIZE - 1 - path;
973 if (copy_to_user(buffer, path, len))
976 free_page((unsigned long)tmp);
980 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
983 struct inode *inode = dentry->d_inode;
985 struct vfsmount *mnt = NULL;
987 /* Are we allowed to snoop on the tasks file descriptors? */
988 if (!proc_fd_access_allowed(inode))
991 error = PROC_I(inode)->op.proc_get_link(inode, &de, &mnt);
995 error = do_proc_readlink(de, mnt, buffer, buflen);
1002 static const struct inode_operations proc_pid_link_inode_operations = {
1003 .readlink = proc_pid_readlink,
1004 .follow_link = proc_pid_follow_link,
1005 .setattr = proc_setattr,
1009 /* building an inode */
1011 static int task_dumpable(struct task_struct *task)
1014 struct mm_struct *mm;
1019 dumpable = get_dumpable(mm);
1027 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1029 struct inode * inode;
1030 struct proc_inode *ei;
1032 /* We need a new inode */
1034 inode = new_inode(sb);
1040 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1041 inode->i_op = &proc_def_inode_operations;
1044 * grab the reference to task.
1046 ei->pid = get_task_pid(task, PIDTYPE_PID);
1052 if (task_dumpable(task)) {
1053 inode->i_uid = task->euid;
1054 inode->i_gid = task->egid;
1056 security_task_to_inode(task, inode);
1066 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1068 struct inode *inode = dentry->d_inode;
1069 struct task_struct *task;
1070 generic_fillattr(inode, stat);
1075 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1077 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1078 task_dumpable(task)) {
1079 stat->uid = task->euid;
1080 stat->gid = task->egid;
1090 * Exceptional case: normally we are not allowed to unhash a busy
1091 * directory. In this case, however, we can do it - no aliasing problems
1092 * due to the way we treat inodes.
1094 * Rewrite the inode's ownerships here because the owning task may have
1095 * performed a setuid(), etc.
1097 * Before the /proc/pid/status file was created the only way to read
1098 * the effective uid of a /process was to stat /proc/pid. Reading
1099 * /proc/pid/status is slow enough that procps and other packages
1100 * kept stating /proc/pid. To keep the rules in /proc simple I have
1101 * made this apply to all per process world readable and executable
1104 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1106 struct inode *inode = dentry->d_inode;
1107 struct task_struct *task = get_proc_task(inode);
1109 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1110 task_dumpable(task)) {
1111 inode->i_uid = task->euid;
1112 inode->i_gid = task->egid;
1117 inode->i_mode &= ~(S_ISUID | S_ISGID);
1118 security_task_to_inode(task, inode);
1119 put_task_struct(task);
1126 static int pid_delete_dentry(struct dentry * dentry)
1128 /* Is the task we represent dead?
1129 * If so, then don't put the dentry on the lru list,
1130 * kill it immediately.
1132 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1135 static struct dentry_operations pid_dentry_operations =
1137 .d_revalidate = pid_revalidate,
1138 .d_delete = pid_delete_dentry,
1143 typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1144 struct task_struct *, const void *);
1147 * Fill a directory entry.
1149 * If possible create the dcache entry and derive our inode number and
1150 * file type from dcache entry.
1152 * Since all of the proc inode numbers are dynamically generated, the inode
1153 * numbers do not exist until the inode is cache. This means creating the
1154 * the dcache entry in readdir is necessary to keep the inode numbers
1155 * reported by readdir in sync with the inode numbers reported
1158 static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1159 char *name, int len,
1160 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1162 struct dentry *child, *dir = filp->f_path.dentry;
1163 struct inode *inode;
1166 unsigned type = DT_UNKNOWN;
1170 qname.hash = full_name_hash(name, len);
1172 child = d_lookup(dir, &qname);
1175 new = d_alloc(dir, &qname);
1177 child = instantiate(dir->d_inode, new, task, ptr);
1184 if (!child || IS_ERR(child) || !child->d_inode)
1185 goto end_instantiate;
1186 inode = child->d_inode;
1189 type = inode->i_mode >> 12;
1194 ino = find_inode_number(dir, &qname);
1197 return filldir(dirent, name, len, filp->f_pos, ino, type);
1200 static unsigned name_to_int(struct dentry *dentry)
1202 const char *name = dentry->d_name.name;
1203 int len = dentry->d_name.len;
1206 if (len > 1 && *name == '0')
1209 unsigned c = *name++ - '0';
1212 if (n >= (~0U-9)/10)
1222 #define PROC_FDINFO_MAX 64
1224 static int proc_fd_info(struct inode *inode, struct dentry **dentry,
1225 struct vfsmount **mnt, char *info)
1227 struct task_struct *task = get_proc_task(inode);
1228 struct files_struct *files = NULL;
1230 int fd = proc_fd(inode);
1233 files = get_files_struct(task);
1234 put_task_struct(task);
1238 * We are not taking a ref to the file structure, so we must
1241 spin_lock(&files->file_lock);
1242 file = fcheck_files(files, fd);
1245 *mnt = mntget(file->f_path.mnt);
1247 *dentry = dget(file->f_path.dentry);
1249 snprintf(info, PROC_FDINFO_MAX,
1252 (long long) file->f_pos,
1254 spin_unlock(&files->file_lock);
1255 put_files_struct(files);
1258 spin_unlock(&files->file_lock);
1259 put_files_struct(files);
1264 static int proc_fd_link(struct inode *inode, struct dentry **dentry,
1265 struct vfsmount **mnt)
1267 return proc_fd_info(inode, dentry, mnt, NULL);
1270 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1272 struct inode *inode = dentry->d_inode;
1273 struct task_struct *task = get_proc_task(inode);
1274 int fd = proc_fd(inode);
1275 struct files_struct *files;
1278 files = get_files_struct(task);
1281 if (fcheck_files(files, fd)) {
1283 put_files_struct(files);
1284 if (task_dumpable(task)) {
1285 inode->i_uid = task->euid;
1286 inode->i_gid = task->egid;
1291 inode->i_mode &= ~(S_ISUID | S_ISGID);
1292 security_task_to_inode(task, inode);
1293 put_task_struct(task);
1297 put_files_struct(files);
1299 put_task_struct(task);
1305 static struct dentry_operations tid_fd_dentry_operations =
1307 .d_revalidate = tid_fd_revalidate,
1308 .d_delete = pid_delete_dentry,
1311 static struct dentry *proc_fd_instantiate(struct inode *dir,
1312 struct dentry *dentry, struct task_struct *task, const void *ptr)
1314 unsigned fd = *(const unsigned *)ptr;
1316 struct files_struct *files;
1317 struct inode *inode;
1318 struct proc_inode *ei;
1319 struct dentry *error = ERR_PTR(-ENOENT);
1321 inode = proc_pid_make_inode(dir->i_sb, task);
1326 files = get_files_struct(task);
1329 inode->i_mode = S_IFLNK;
1332 * We are not taking a ref to the file structure, so we must
1335 spin_lock(&files->file_lock);
1336 file = fcheck_files(files, fd);
1339 if (file->f_mode & 1)
1340 inode->i_mode |= S_IRUSR | S_IXUSR;
1341 if (file->f_mode & 2)
1342 inode->i_mode |= S_IWUSR | S_IXUSR;
1343 spin_unlock(&files->file_lock);
1344 put_files_struct(files);
1346 inode->i_op = &proc_pid_link_inode_operations;
1348 ei->op.proc_get_link = proc_fd_link;
1349 dentry->d_op = &tid_fd_dentry_operations;
1350 d_add(dentry, inode);
1351 /* Close the race of the process dying before we return the dentry */
1352 if (tid_fd_revalidate(dentry, NULL))
1358 spin_unlock(&files->file_lock);
1359 put_files_struct(files);
1365 static struct dentry *proc_lookupfd_common(struct inode *dir,
1366 struct dentry *dentry,
1367 instantiate_t instantiate)
1369 struct task_struct *task = get_proc_task(dir);
1370 unsigned fd = name_to_int(dentry);
1371 struct dentry *result = ERR_PTR(-ENOENT);
1378 result = instantiate(dir, dentry, task, &fd);
1380 put_task_struct(task);
1385 static int proc_readfd_common(struct file * filp, void * dirent,
1386 filldir_t filldir, instantiate_t instantiate)
1388 struct dentry *dentry = filp->f_path.dentry;
1389 struct inode *inode = dentry->d_inode;
1390 struct task_struct *p = get_proc_task(inode);
1391 unsigned int fd, tid, ino;
1393 struct files_struct * files;
1394 struct fdtable *fdt;
1405 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1409 ino = parent_ino(dentry);
1410 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1414 files = get_files_struct(p);
1418 fdt = files_fdtable(files);
1419 for (fd = filp->f_pos-2;
1421 fd++, filp->f_pos++) {
1422 char name[PROC_NUMBUF];
1425 if (!fcheck_files(files, fd))
1429 len = snprintf(name, sizeof(name), "%d", fd);
1430 if (proc_fill_cache(filp, dirent, filldir,
1431 name, len, instantiate,
1439 put_files_struct(files);
1447 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1448 struct nameidata *nd)
1450 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1453 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1455 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1458 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1459 size_t len, loff_t *ppos)
1461 char tmp[PROC_FDINFO_MAX];
1462 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, NULL, tmp);
1464 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1468 static const struct file_operations proc_fdinfo_file_operations = {
1469 .open = nonseekable_open,
1470 .read = proc_fdinfo_read,
1473 static const struct file_operations proc_fd_operations = {
1474 .read = generic_read_dir,
1475 .readdir = proc_readfd,
1479 * /proc/pid/fd needs a special permission handler so that a process can still
1480 * access /proc/self/fd after it has executed a setuid().
1482 static int proc_fd_permission(struct inode *inode, int mask,
1483 struct nameidata *nd)
1487 rv = generic_permission(inode, mask, NULL);
1490 if (task_pid(current) == proc_pid(inode))
1496 * proc directories can do almost nothing..
1498 static const struct inode_operations proc_fd_inode_operations = {
1499 .lookup = proc_lookupfd,
1500 .permission = proc_fd_permission,
1501 .setattr = proc_setattr,
1504 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
1505 struct dentry *dentry, struct task_struct *task, const void *ptr)
1507 unsigned fd = *(unsigned *)ptr;
1508 struct inode *inode;
1509 struct proc_inode *ei;
1510 struct dentry *error = ERR_PTR(-ENOENT);
1512 inode = proc_pid_make_inode(dir->i_sb, task);
1517 inode->i_mode = S_IFREG | S_IRUSR;
1518 inode->i_fop = &proc_fdinfo_file_operations;
1519 dentry->d_op = &tid_fd_dentry_operations;
1520 d_add(dentry, inode);
1521 /* Close the race of the process dying before we return the dentry */
1522 if (tid_fd_revalidate(dentry, NULL))
1529 static struct dentry *proc_lookupfdinfo(struct inode *dir,
1530 struct dentry *dentry,
1531 struct nameidata *nd)
1533 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
1536 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
1538 return proc_readfd_common(filp, dirent, filldir,
1539 proc_fdinfo_instantiate);
1542 static const struct file_operations proc_fdinfo_operations = {
1543 .read = generic_read_dir,
1544 .readdir = proc_readfdinfo,
1548 * proc directories can do almost nothing..
1550 static const struct inode_operations proc_fdinfo_inode_operations = {
1551 .lookup = proc_lookupfdinfo,
1552 .setattr = proc_setattr,
1556 static struct dentry *proc_pident_instantiate(struct inode *dir,
1557 struct dentry *dentry, struct task_struct *task, const void *ptr)
1559 const struct pid_entry *p = ptr;
1560 struct inode *inode;
1561 struct proc_inode *ei;
1562 struct dentry *error = ERR_PTR(-EINVAL);
1564 inode = proc_pid_make_inode(dir->i_sb, task);
1569 inode->i_mode = p->mode;
1570 if (S_ISDIR(inode->i_mode))
1571 inode->i_nlink = 2; /* Use getattr to fix if necessary */
1573 inode->i_op = p->iop;
1575 inode->i_fop = p->fop;
1577 dentry->d_op = &pid_dentry_operations;
1578 d_add(dentry, inode);
1579 /* Close the race of the process dying before we return the dentry */
1580 if (pid_revalidate(dentry, NULL))
1586 static struct dentry *proc_pident_lookup(struct inode *dir,
1587 struct dentry *dentry,
1588 const struct pid_entry *ents,
1591 struct inode *inode;
1592 struct dentry *error;
1593 struct task_struct *task = get_proc_task(dir);
1594 const struct pid_entry *p, *last;
1596 error = ERR_PTR(-ENOENT);
1603 * Yes, it does not scale. And it should not. Don't add
1604 * new entries into /proc/<tgid>/ without very good reasons.
1606 last = &ents[nents - 1];
1607 for (p = ents; p <= last; p++) {
1608 if (p->len != dentry->d_name.len)
1610 if (!memcmp(dentry->d_name.name, p->name, p->len))
1616 error = proc_pident_instantiate(dir, dentry, task, p);
1618 put_task_struct(task);
1623 static int proc_pident_fill_cache(struct file *filp, void *dirent,
1624 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
1626 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
1627 proc_pident_instantiate, task, p);
1630 static int proc_pident_readdir(struct file *filp,
1631 void *dirent, filldir_t filldir,
1632 const struct pid_entry *ents, unsigned int nents)
1636 struct dentry *dentry = filp->f_path.dentry;
1637 struct inode *inode = dentry->d_inode;
1638 struct task_struct *task = get_proc_task(inode);
1639 const struct pid_entry *p, *last;
1653 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
1659 ino = parent_ino(dentry);
1660 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
1672 last = &ents[nents - 1];
1674 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
1683 put_task_struct(task);
1688 #ifdef CONFIG_SECURITY
1689 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
1690 size_t count, loff_t *ppos)
1692 struct inode * inode = file->f_path.dentry->d_inode;
1695 struct task_struct *task = get_proc_task(inode);
1700 length = security_getprocattr(task,
1701 (char*)file->f_path.dentry->d_name.name,
1703 put_task_struct(task);
1705 length = simple_read_from_buffer(buf, count, ppos, p, length);
1710 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
1711 size_t count, loff_t *ppos)
1713 struct inode * inode = file->f_path.dentry->d_inode;
1716 struct task_struct *task = get_proc_task(inode);
1721 if (count > PAGE_SIZE)
1724 /* No partial writes. */
1730 page = (char*)__get_free_page(GFP_TEMPORARY);
1735 if (copy_from_user(page, buf, count))
1738 length = security_setprocattr(task,
1739 (char*)file->f_path.dentry->d_name.name,
1740 (void*)page, count);
1742 free_page((unsigned long) page);
1744 put_task_struct(task);
1749 static const struct file_operations proc_pid_attr_operations = {
1750 .read = proc_pid_attr_read,
1751 .write = proc_pid_attr_write,
1754 static const struct pid_entry attr_dir_stuff[] = {
1755 REG("current", S_IRUGO|S_IWUGO, pid_attr),
1756 REG("prev", S_IRUGO, pid_attr),
1757 REG("exec", S_IRUGO|S_IWUGO, pid_attr),
1758 REG("fscreate", S_IRUGO|S_IWUGO, pid_attr),
1759 REG("keycreate", S_IRUGO|S_IWUGO, pid_attr),
1760 REG("sockcreate", S_IRUGO|S_IWUGO, pid_attr),
1763 static int proc_attr_dir_readdir(struct file * filp,
1764 void * dirent, filldir_t filldir)
1766 return proc_pident_readdir(filp,dirent,filldir,
1767 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
1770 static const struct file_operations proc_attr_dir_operations = {
1771 .read = generic_read_dir,
1772 .readdir = proc_attr_dir_readdir,
1775 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
1776 struct dentry *dentry, struct nameidata *nd)
1778 return proc_pident_lookup(dir, dentry,
1779 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
1782 static const struct inode_operations proc_attr_dir_inode_operations = {
1783 .lookup = proc_attr_dir_lookup,
1784 .getattr = pid_getattr,
1785 .setattr = proc_setattr,
1790 #if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
1791 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
1792 size_t count, loff_t *ppos)
1794 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1795 struct mm_struct *mm;
1796 char buffer[PROC_NUMBUF];
1804 mm = get_task_mm(task);
1806 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
1807 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
1808 MMF_DUMP_FILTER_SHIFT));
1810 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
1813 put_task_struct(task);
1818 static ssize_t proc_coredump_filter_write(struct file *file,
1819 const char __user *buf,
1823 struct task_struct *task;
1824 struct mm_struct *mm;
1825 char buffer[PROC_NUMBUF], *end;
1832 memset(buffer, 0, sizeof(buffer));
1833 if (count > sizeof(buffer) - 1)
1834 count = sizeof(buffer) - 1;
1835 if (copy_from_user(buffer, buf, count))
1839 val = (unsigned int)simple_strtoul(buffer, &end, 0);
1842 if (end - buffer == 0)
1846 task = get_proc_task(file->f_dentry->d_inode);
1851 mm = get_task_mm(task);
1855 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
1857 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
1859 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
1864 put_task_struct(task);
1869 static const struct file_operations proc_coredump_filter_operations = {
1870 .read = proc_coredump_filter_read,
1871 .write = proc_coredump_filter_write,
1878 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
1881 char tmp[PROC_NUMBUF];
1882 sprintf(tmp, "%d", current->tgid);
1883 return vfs_readlink(dentry,buffer,buflen,tmp);
1886 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
1888 char tmp[PROC_NUMBUF];
1889 sprintf(tmp, "%d", current->tgid);
1890 return ERR_PTR(vfs_follow_link(nd,tmp));
1893 static const struct inode_operations proc_self_inode_operations = {
1894 .readlink = proc_self_readlink,
1895 .follow_link = proc_self_follow_link,
1901 * These are the directory entries in the root directory of /proc
1902 * that properly belong to the /proc filesystem, as they describe
1903 * describe something that is process related.
1905 static const struct pid_entry proc_base_stuff[] = {
1906 NOD("self", S_IFLNK|S_IRWXUGO,
1907 &proc_self_inode_operations, NULL, {}),
1911 * Exceptional case: normally we are not allowed to unhash a busy
1912 * directory. In this case, however, we can do it - no aliasing problems
1913 * due to the way we treat inodes.
1915 static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
1917 struct inode *inode = dentry->d_inode;
1918 struct task_struct *task = get_proc_task(inode);
1920 put_task_struct(task);
1927 static struct dentry_operations proc_base_dentry_operations =
1929 .d_revalidate = proc_base_revalidate,
1930 .d_delete = pid_delete_dentry,
1933 static struct dentry *proc_base_instantiate(struct inode *dir,
1934 struct dentry *dentry, struct task_struct *task, const void *ptr)
1936 const struct pid_entry *p = ptr;
1937 struct inode *inode;
1938 struct proc_inode *ei;
1939 struct dentry *error = ERR_PTR(-EINVAL);
1941 /* Allocate the inode */
1942 error = ERR_PTR(-ENOMEM);
1943 inode = new_inode(dir->i_sb);
1947 /* Initialize the inode */
1949 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1952 * grab the reference to the task.
1954 ei->pid = get_task_pid(task, PIDTYPE_PID);
1960 inode->i_mode = p->mode;
1961 if (S_ISDIR(inode->i_mode))
1963 if (S_ISLNK(inode->i_mode))
1966 inode->i_op = p->iop;
1968 inode->i_fop = p->fop;
1970 dentry->d_op = &proc_base_dentry_operations;
1971 d_add(dentry, inode);
1980 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
1982 struct dentry *error;
1983 struct task_struct *task = get_proc_task(dir);
1984 const struct pid_entry *p, *last;
1986 error = ERR_PTR(-ENOENT);
1991 /* Lookup the directory entry */
1992 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
1993 for (p = proc_base_stuff; p <= last; p++) {
1994 if (p->len != dentry->d_name.len)
1996 if (!memcmp(dentry->d_name.name, p->name, p->len))
2002 error = proc_base_instantiate(dir, dentry, task, p);
2005 put_task_struct(task);
2010 static int proc_base_fill_cache(struct file *filp, void *dirent,
2011 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2013 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2014 proc_base_instantiate, task, p);
2017 #ifdef CONFIG_TASK_IO_ACCOUNTING
2018 static int proc_pid_io_accounting(struct task_struct *task, char *buffer)
2020 return sprintf(buffer,
2021 #ifdef CONFIG_TASK_XACCT
2027 "read_bytes: %llu\n"
2028 "write_bytes: %llu\n"
2029 "cancelled_write_bytes: %llu\n",
2030 #ifdef CONFIG_TASK_XACCT
2031 (unsigned long long)task->rchar,
2032 (unsigned long long)task->wchar,
2033 (unsigned long long)task->syscr,
2034 (unsigned long long)task->syscw,
2036 (unsigned long long)task->ioac.read_bytes,
2037 (unsigned long long)task->ioac.write_bytes,
2038 (unsigned long long)task->ioac.cancelled_write_bytes);
2045 static const struct file_operations proc_task_operations;
2046 static const struct inode_operations proc_task_inode_operations;
2048 static const struct pid_entry tgid_base_stuff[] = {
2049 DIR("task", S_IRUGO|S_IXUGO, task),
2050 DIR("fd", S_IRUSR|S_IXUSR, fd),
2051 DIR("fdinfo", S_IRUSR|S_IXUSR, fdinfo),
2052 INF("environ", S_IRUSR, pid_environ),
2053 INF("auxv", S_IRUSR, pid_auxv),
2054 INF("status", S_IRUGO, pid_status),
2055 #ifdef CONFIG_SCHED_DEBUG
2056 REG("sched", S_IRUGO|S_IWUSR, pid_sched),
2058 INF("cmdline", S_IRUGO, pid_cmdline),
2059 INF("stat", S_IRUGO, tgid_stat),
2060 INF("statm", S_IRUGO, pid_statm),
2061 REG("maps", S_IRUGO, maps),
2063 REG("numa_maps", S_IRUGO, numa_maps),
2065 REG("mem", S_IRUSR|S_IWUSR, mem),
2069 REG("mounts", S_IRUGO, mounts),
2070 REG("mountstats", S_IRUSR, mountstats),
2072 REG("clear_refs", S_IWUSR, clear_refs),
2073 REG("smaps", S_IRUGO, smaps),
2075 #ifdef CONFIG_SECURITY
2076 DIR("attr", S_IRUGO|S_IXUGO, attr_dir),
2078 #ifdef CONFIG_KALLSYMS
2079 INF("wchan", S_IRUGO, pid_wchan),
2081 #ifdef CONFIG_SCHEDSTATS
2082 INF("schedstat", S_IRUGO, pid_schedstat),
2084 #ifdef CONFIG_CPUSETS
2085 REG("cpuset", S_IRUGO, cpuset),
2087 INF("oom_score", S_IRUGO, oom_score),
2088 REG("oom_adj", S_IRUGO|S_IWUSR, oom_adjust),
2089 #ifdef CONFIG_AUDITSYSCALL
2090 REG("loginuid", S_IWUSR|S_IRUGO, loginuid),
2092 #ifdef CONFIG_FAULT_INJECTION
2093 REG("make-it-fail", S_IRUGO|S_IWUSR, fault_inject),
2095 #if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
2096 REG("coredump_filter", S_IRUGO|S_IWUSR, coredump_filter),
2098 #ifdef CONFIG_TASK_IO_ACCOUNTING
2099 INF("io", S_IRUGO, pid_io_accounting),
2103 static int proc_tgid_base_readdir(struct file * filp,
2104 void * dirent, filldir_t filldir)
2106 return proc_pident_readdir(filp,dirent,filldir,
2107 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2110 static const struct file_operations proc_tgid_base_operations = {
2111 .read = generic_read_dir,
2112 .readdir = proc_tgid_base_readdir,
2115 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2116 return proc_pident_lookup(dir, dentry,
2117 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2120 static const struct inode_operations proc_tgid_base_inode_operations = {
2121 .lookup = proc_tgid_base_lookup,
2122 .getattr = pid_getattr,
2123 .setattr = proc_setattr,
2127 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2129 * @task: task that should be flushed.
2131 * Looks in the dcache for
2133 * /proc/@tgid/task/@pid
2134 * if either directory is present flushes it and all of it'ts children
2137 * It is safe and reasonable to cache /proc entries for a task until
2138 * that task exits. After that they just clog up the dcache with
2139 * useless entries, possibly causing useful dcache entries to be
2140 * flushed instead. This routine is proved to flush those useless
2141 * dcache entries at process exit time.
2143 * NOTE: This routine is just an optimization so it does not guarantee
2144 * that no dcache entries will exist at process exit time it
2145 * just makes it very unlikely that any will persist.
2147 void proc_flush_task(struct task_struct *task)
2149 struct dentry *dentry, *leader, *dir;
2150 char buf[PROC_NUMBUF];
2154 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
2155 dentry = d_hash_and_lookup(proc_mnt->mnt_root, &name);
2157 shrink_dcache_parent(dentry);
2162 if (thread_group_leader(task))
2166 name.len = snprintf(buf, sizeof(buf), "%d", task->tgid);
2167 leader = d_hash_and_lookup(proc_mnt->mnt_root, &name);
2172 name.len = strlen(name.name);
2173 dir = d_hash_and_lookup(leader, &name);
2175 goto out_put_leader;
2178 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
2179 dentry = d_hash_and_lookup(dir, &name);
2181 shrink_dcache_parent(dentry);
2193 static struct dentry *proc_pid_instantiate(struct inode *dir,
2194 struct dentry * dentry,
2195 struct task_struct *task, const void *ptr)
2197 struct dentry *error = ERR_PTR(-ENOENT);
2198 struct inode *inode;
2200 inode = proc_pid_make_inode(dir->i_sb, task);
2204 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2205 inode->i_op = &proc_tgid_base_inode_operations;
2206 inode->i_fop = &proc_tgid_base_operations;
2207 inode->i_flags|=S_IMMUTABLE;
2209 #ifdef CONFIG_SECURITY
2210 inode->i_nlink += 1;
2213 dentry->d_op = &pid_dentry_operations;
2215 d_add(dentry, inode);
2216 /* Close the race of the process dying before we return the dentry */
2217 if (pid_revalidate(dentry, NULL))
2223 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2225 struct dentry *result = ERR_PTR(-ENOENT);
2226 struct task_struct *task;
2229 result = proc_base_lookup(dir, dentry);
2230 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2233 tgid = name_to_int(dentry);
2238 task = find_task_by_pid(tgid);
2240 get_task_struct(task);
2245 result = proc_pid_instantiate(dir, dentry, task, NULL);
2246 put_task_struct(task);
2252 * Find the first task with tgid >= tgid
2255 static struct task_struct *next_tgid(unsigned int tgid)
2257 struct task_struct *task;
2263 pid = find_ge_pid(tgid);
2266 task = pid_task(pid, PIDTYPE_PID);
2267 /* What we to know is if the pid we have find is the
2268 * pid of a thread_group_leader. Testing for task
2269 * being a thread_group_leader is the obvious thing
2270 * todo but there is a window when it fails, due to
2271 * the pid transfer logic in de_thread.
2273 * So we perform the straight forward test of seeing
2274 * if the pid we have found is the pid of a thread
2275 * group leader, and don't worry if the task we have
2276 * found doesn't happen to be a thread group leader.
2277 * As we don't care in the case of readdir.
2279 if (!task || !has_group_leader_pid(task))
2281 get_task_struct(task);
2287 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2289 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2290 struct task_struct *task, int tgid)
2292 char name[PROC_NUMBUF];
2293 int len = snprintf(name, sizeof(name), "%d", tgid);
2294 return proc_fill_cache(filp, dirent, filldir, name, len,
2295 proc_pid_instantiate, task, NULL);
2298 /* for the /proc/ directory itself, after non-process stuff has been done */
2299 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2301 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2302 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
2303 struct task_struct *task;
2309 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
2310 const struct pid_entry *p = &proc_base_stuff[nr];
2311 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
2315 tgid = filp->f_pos - TGID_OFFSET;
2316 for (task = next_tgid(tgid);
2318 put_task_struct(task), task = next_tgid(tgid + 1)) {
2320 filp->f_pos = tgid + TGID_OFFSET;
2321 if (proc_pid_fill_cache(filp, dirent, filldir, task, tgid) < 0) {
2322 put_task_struct(task);
2326 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
2328 put_task_struct(reaper);
2336 static const struct pid_entry tid_base_stuff[] = {
2337 DIR("fd", S_IRUSR|S_IXUSR, fd),
2338 DIR("fdinfo", S_IRUSR|S_IXUSR, fdinfo),
2339 INF("environ", S_IRUSR, pid_environ),
2340 INF("auxv", S_IRUSR, pid_auxv),
2341 INF("status", S_IRUGO, pid_status),
2342 #ifdef CONFIG_SCHED_DEBUG
2343 REG("sched", S_IRUGO|S_IWUSR, pid_sched),
2345 INF("cmdline", S_IRUGO, pid_cmdline),
2346 INF("stat", S_IRUGO, tid_stat),
2347 INF("statm", S_IRUGO, pid_statm),
2348 REG("maps", S_IRUGO, maps),
2350 REG("numa_maps", S_IRUGO, numa_maps),
2352 REG("mem", S_IRUSR|S_IWUSR, mem),
2356 REG("mounts", S_IRUGO, mounts),
2358 REG("clear_refs", S_IWUSR, clear_refs),
2359 REG("smaps", S_IRUGO, smaps),
2361 #ifdef CONFIG_SECURITY
2362 DIR("attr", S_IRUGO|S_IXUGO, attr_dir),
2364 #ifdef CONFIG_KALLSYMS
2365 INF("wchan", S_IRUGO, pid_wchan),
2367 #ifdef CONFIG_SCHEDSTATS
2368 INF("schedstat", S_IRUGO, pid_schedstat),
2370 #ifdef CONFIG_CPUSETS
2371 REG("cpuset", S_IRUGO, cpuset),
2373 INF("oom_score", S_IRUGO, oom_score),
2374 REG("oom_adj", S_IRUGO|S_IWUSR, oom_adjust),
2375 #ifdef CONFIG_AUDITSYSCALL
2376 REG("loginuid", S_IWUSR|S_IRUGO, loginuid),
2378 #ifdef CONFIG_FAULT_INJECTION
2379 REG("make-it-fail", S_IRUGO|S_IWUSR, fault_inject),
2383 static int proc_tid_base_readdir(struct file * filp,
2384 void * dirent, filldir_t filldir)
2386 return proc_pident_readdir(filp,dirent,filldir,
2387 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
2390 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2391 return proc_pident_lookup(dir, dentry,
2392 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
2395 static const struct file_operations proc_tid_base_operations = {
2396 .read = generic_read_dir,
2397 .readdir = proc_tid_base_readdir,
2400 static const struct inode_operations proc_tid_base_inode_operations = {
2401 .lookup = proc_tid_base_lookup,
2402 .getattr = pid_getattr,
2403 .setattr = proc_setattr,
2406 static struct dentry *proc_task_instantiate(struct inode *dir,
2407 struct dentry *dentry, struct task_struct *task, const void *ptr)
2409 struct dentry *error = ERR_PTR(-ENOENT);
2410 struct inode *inode;
2411 inode = proc_pid_make_inode(dir->i_sb, task);
2415 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2416 inode->i_op = &proc_tid_base_inode_operations;
2417 inode->i_fop = &proc_tid_base_operations;
2418 inode->i_flags|=S_IMMUTABLE;
2420 #ifdef CONFIG_SECURITY
2421 inode->i_nlink += 1;
2424 dentry->d_op = &pid_dentry_operations;
2426 d_add(dentry, inode);
2427 /* Close the race of the process dying before we return the dentry */
2428 if (pid_revalidate(dentry, NULL))
2434 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2436 struct dentry *result = ERR_PTR(-ENOENT);
2437 struct task_struct *task;
2438 struct task_struct *leader = get_proc_task(dir);
2444 tid = name_to_int(dentry);
2449 task = find_task_by_pid(tid);
2451 get_task_struct(task);
2455 if (leader->tgid != task->tgid)
2458 result = proc_task_instantiate(dir, dentry, task, NULL);
2460 put_task_struct(task);
2462 put_task_struct(leader);
2468 * Find the first tid of a thread group to return to user space.
2470 * Usually this is just the thread group leader, but if the users
2471 * buffer was too small or there was a seek into the middle of the
2472 * directory we have more work todo.
2474 * In the case of a short read we start with find_task_by_pid.
2476 * In the case of a seek we start with the leader and walk nr
2479 static struct task_struct *first_tid(struct task_struct *leader,
2482 struct task_struct *pos;
2485 /* Attempt to start with the pid of a thread */
2486 if (tid && (nr > 0)) {
2487 pos = find_task_by_pid(tid);
2488 if (pos && (pos->group_leader == leader))
2492 /* If nr exceeds the number of threads there is nothing todo */
2494 if (nr && nr >= get_nr_threads(leader))
2497 /* If we haven't found our starting place yet start
2498 * with the leader and walk nr threads forward.
2500 for (pos = leader; nr > 0; --nr) {
2501 pos = next_thread(pos);
2502 if (pos == leader) {
2508 get_task_struct(pos);
2515 * Find the next thread in the thread list.
2516 * Return NULL if there is an error or no next thread.
2518 * The reference to the input task_struct is released.
2520 static struct task_struct *next_tid(struct task_struct *start)
2522 struct task_struct *pos = NULL;
2524 if (pid_alive(start)) {
2525 pos = next_thread(start);
2526 if (thread_group_leader(pos))
2529 get_task_struct(pos);
2532 put_task_struct(start);
2536 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2537 struct task_struct *task, int tid)
2539 char name[PROC_NUMBUF];
2540 int len = snprintf(name, sizeof(name), "%d", tid);
2541 return proc_fill_cache(filp, dirent, filldir, name, len,
2542 proc_task_instantiate, task, NULL);
2545 /* for the /proc/TGID/task/ directories */
2546 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
2548 struct dentry *dentry = filp->f_path.dentry;
2549 struct inode *inode = dentry->d_inode;
2550 struct task_struct *leader = NULL;
2551 struct task_struct *task;
2552 int retval = -ENOENT;
2555 unsigned long pos = filp->f_pos; /* avoiding "long long" filp->f_pos */
2557 task = get_proc_task(inode);
2561 if (pid_alive(task)) {
2562 leader = task->group_leader;
2563 get_task_struct(leader);
2566 put_task_struct(task);
2574 if (filldir(dirent, ".", 1, pos, ino, DT_DIR) < 0)
2579 ino = parent_ino(dentry);
2580 if (filldir(dirent, "..", 2, pos, ino, DT_DIR) < 0)
2586 /* f_version caches the tgid value that the last readdir call couldn't
2587 * return. lseek aka telldir automagically resets f_version to 0.
2589 tid = (int)filp->f_version;
2590 filp->f_version = 0;
2591 for (task = first_tid(leader, tid, pos - 2);
2593 task = next_tid(task), pos++) {
2595 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
2596 /* returning this tgid failed, save it as the first
2597 * pid for the next readir call */
2598 filp->f_version = (u64)tid;
2599 put_task_struct(task);
2605 put_task_struct(leader);
2610 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
2612 struct inode *inode = dentry->d_inode;
2613 struct task_struct *p = get_proc_task(inode);
2614 generic_fillattr(inode, stat);
2618 stat->nlink += get_nr_threads(p);
2626 static const struct inode_operations proc_task_inode_operations = {
2627 .lookup = proc_task_lookup,
2628 .getattr = proc_task_getattr,
2629 .setattr = proc_setattr,
2632 static const struct file_operations proc_task_operations = {
2633 .read = generic_read_dir,
2634 .readdir = proc_task_readdir,