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/cgroup.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>
76 #include <linux/elf.h>
80 * Implementing inode permission operations in /proc is almost
81 * certainly an error. Permission checks need to happen during
82 * each system call not at open time. The reason is that most of
83 * what we wish to check for permissions in /proc varies at runtime.
85 * The classic example of a problem is opening file descriptors
86 * in /proc for a task before it execs a suid executable.
90 /* Worst case buffer size needed for holding an integer. */
91 #define PROC_NUMBUF 13
97 const struct inode_operations *iop;
98 const struct file_operations *fop;
102 #define NOD(NAME, MODE, IOP, FOP, OP) { \
104 .len = sizeof(NAME) - 1, \
111 #define DIR(NAME, MODE, OTYPE) \
112 NOD(NAME, (S_IFDIR|(MODE)), \
113 &proc_##OTYPE##_inode_operations, &proc_##OTYPE##_operations, \
115 #define LNK(NAME, OTYPE) \
116 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
117 &proc_pid_link_inode_operations, NULL, \
118 { .proc_get_link = &proc_##OTYPE##_link } )
119 #define REG(NAME, MODE, OTYPE) \
120 NOD(NAME, (S_IFREG|(MODE)), NULL, \
121 &proc_##OTYPE##_operations, {})
122 #define INF(NAME, MODE, OTYPE) \
123 NOD(NAME, (S_IFREG|(MODE)), \
124 NULL, &proc_info_file_operations, \
125 { .proc_read = &proc_##OTYPE } )
128 EXPORT_SYMBOL(maps_protect);
130 static struct fs_struct *get_fs_struct(struct task_struct *task)
132 struct fs_struct *fs;
136 atomic_inc(&fs->count);
141 static int get_nr_threads(struct task_struct *tsk)
143 /* Must be called with the rcu_read_lock held */
147 if (lock_task_sighand(tsk, &flags)) {
148 count = atomic_read(&tsk->signal->count);
149 unlock_task_sighand(tsk, &flags);
154 static int proc_cwd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
156 struct task_struct *task = get_proc_task(inode);
157 struct fs_struct *fs = NULL;
158 int result = -ENOENT;
161 fs = get_fs_struct(task);
162 put_task_struct(task);
165 read_lock(&fs->lock);
166 *mnt = mntget(fs->pwdmnt);
167 *dentry = dget(fs->pwd);
168 read_unlock(&fs->lock);
175 static int proc_root_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
177 struct task_struct *task = get_proc_task(inode);
178 struct fs_struct *fs = NULL;
179 int result = -ENOENT;
182 fs = get_fs_struct(task);
183 put_task_struct(task);
186 read_lock(&fs->lock);
187 *mnt = mntget(fs->rootmnt);
188 *dentry = dget(fs->root);
189 read_unlock(&fs->lock);
196 #define MAY_PTRACE(task) \
197 (task == current || \
198 (task->parent == current && \
199 (task->ptrace & PT_PTRACED) && \
200 (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
201 security_ptrace(current,task) == 0))
203 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
207 struct mm_struct *mm = get_task_mm(task);
211 goto out_mm; /* Shh! No looking before we're done */
213 len = mm->arg_end - mm->arg_start;
218 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
220 // If the nul at the end of args has been overwritten, then
221 // assume application is using setproctitle(3).
222 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
223 len = strnlen(buffer, res);
227 len = mm->env_end - mm->env_start;
228 if (len > PAGE_SIZE - res)
229 len = PAGE_SIZE - res;
230 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
231 res = strnlen(buffer, res);
240 static int proc_pid_auxv(struct task_struct *task, char *buffer)
243 struct mm_struct *mm = get_task_mm(task);
245 unsigned int nwords = 0;
248 while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
249 res = nwords * sizeof(mm->saved_auxv[0]);
252 memcpy(buffer, mm->saved_auxv, res);
259 #ifdef CONFIG_KALLSYMS
261 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
262 * Returns the resolved symbol. If that fails, simply return the address.
264 static int proc_pid_wchan(struct task_struct *task, char *buffer)
267 char symname[KSYM_NAME_LEN];
269 wchan = get_wchan(task);
271 if (lookup_symbol_name(wchan, symname) < 0)
272 return sprintf(buffer, "%lu", wchan);
274 return sprintf(buffer, "%s", symname);
276 #endif /* CONFIG_KALLSYMS */
278 #ifdef CONFIG_SCHEDSTATS
280 * Provides /proc/PID/schedstat
282 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
284 return sprintf(buffer, "%llu %llu %lu\n",
285 task->sched_info.cpu_time,
286 task->sched_info.run_delay,
287 task->sched_info.pcount);
291 /* The badness from the OOM killer */
292 unsigned long badness(struct task_struct *p, unsigned long uptime);
293 static int proc_oom_score(struct task_struct *task, char *buffer)
295 unsigned long points;
296 struct timespec uptime;
298 do_posix_clock_monotonic_gettime(&uptime);
299 read_lock(&tasklist_lock);
300 points = badness(task, uptime.tv_sec);
301 read_unlock(&tasklist_lock);
302 return sprintf(buffer, "%lu\n", points);
305 /************************************************************************/
306 /* Here the fs part begins */
307 /************************************************************************/
309 /* permission checks */
310 static int proc_fd_access_allowed(struct inode *inode)
312 struct task_struct *task;
314 /* Allow access to a task's file descriptors if it is us or we
315 * may use ptrace attach to the process and find out that
318 task = get_proc_task(inode);
320 allowed = ptrace_may_attach(task);
321 put_task_struct(task);
326 static int proc_setattr(struct dentry *dentry, struct iattr *attr)
329 struct inode *inode = dentry->d_inode;
331 if (attr->ia_valid & ATTR_MODE)
334 error = inode_change_ok(inode, attr);
336 error = inode_setattr(inode, attr);
340 static const struct inode_operations proc_def_inode_operations = {
341 .setattr = proc_setattr,
344 extern struct seq_operations mounts_op;
350 static int mounts_open(struct inode *inode, struct file *file)
352 struct task_struct *task = get_proc_task(inode);
354 struct mnt_namespace *ns = NULL;
355 struct proc_mounts *p;
360 nsp = task_nsproxy(task);
368 put_task_struct(task);
373 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
375 file->private_data = &p->m;
376 ret = seq_open(file, &mounts_op);
379 p->event = ns->event;
389 static int mounts_release(struct inode *inode, struct file *file)
391 struct seq_file *m = file->private_data;
392 struct mnt_namespace *ns = m->private;
394 return seq_release(inode, file);
397 static unsigned mounts_poll(struct file *file, poll_table *wait)
399 struct proc_mounts *p = file->private_data;
400 struct mnt_namespace *ns = p->m.private;
403 poll_wait(file, &ns->poll, wait);
405 spin_lock(&vfsmount_lock);
406 if (p->event != ns->event) {
407 p->event = ns->event;
410 spin_unlock(&vfsmount_lock);
415 static const struct file_operations proc_mounts_operations = {
419 .release = mounts_release,
423 extern struct seq_operations mountstats_op;
424 static int mountstats_open(struct inode *inode, struct file *file)
426 int ret = seq_open(file, &mountstats_op);
429 struct seq_file *m = file->private_data;
431 struct mnt_namespace *mnt_ns = NULL;
432 struct task_struct *task = get_proc_task(inode);
436 nsp = task_nsproxy(task);
438 mnt_ns = nsp->mnt_ns;
444 put_task_struct(task);
450 seq_release(inode, file);
457 static const struct file_operations proc_mountstats_operations = {
458 .open = mountstats_open,
461 .release = mounts_release,
464 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
466 static ssize_t proc_info_read(struct file * file, char __user * buf,
467 size_t count, loff_t *ppos)
469 struct inode * inode = file->f_path.dentry->d_inode;
472 struct task_struct *task = get_proc_task(inode);
478 if (count > PROC_BLOCK_SIZE)
479 count = PROC_BLOCK_SIZE;
482 if (!(page = __get_free_page(GFP_TEMPORARY)))
485 length = PROC_I(inode)->op.proc_read(task, (char*)page);
488 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
491 put_task_struct(task);
496 static const struct file_operations proc_info_file_operations = {
497 .read = proc_info_read,
500 static int mem_open(struct inode* inode, struct file* file)
502 file->private_data = (void*)((long)current->self_exec_id);
506 static ssize_t mem_read(struct file * file, char __user * buf,
507 size_t count, loff_t *ppos)
509 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
511 unsigned long src = *ppos;
513 struct mm_struct *mm;
518 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
522 page = (char *)__get_free_page(GFP_TEMPORARY);
528 mm = get_task_mm(task);
534 if (file->private_data != (void*)((long)current->self_exec_id))
540 int this_len, retval;
542 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
543 retval = access_process_vm(task, src, page, this_len, 0);
544 if (!retval || !MAY_PTRACE(task) || !ptrace_may_attach(task)) {
550 if (copy_to_user(buf, page, retval)) {
565 free_page((unsigned long) page);
567 put_task_struct(task);
572 #define mem_write NULL
575 /* This is a security hazard */
576 static ssize_t mem_write(struct file * file, const char __user *buf,
577 size_t count, loff_t *ppos)
581 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
582 unsigned long dst = *ppos;
588 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
592 page = (char *)__get_free_page(GFP_TEMPORARY);
598 int this_len, retval;
600 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
601 if (copy_from_user(page, buf, this_len)) {
605 retval = access_process_vm(task, dst, page, this_len, 1);
617 free_page((unsigned long) page);
619 put_task_struct(task);
625 static loff_t mem_lseek(struct file * file, loff_t offset, int orig)
629 file->f_pos = offset;
632 file->f_pos += offset;
637 force_successful_syscall_return();
641 static const struct file_operations proc_mem_operations = {
648 static ssize_t environ_read(struct file *file, char __user *buf,
649 size_t count, loff_t *ppos)
651 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
653 unsigned long src = *ppos;
655 struct mm_struct *mm;
660 if (!ptrace_may_attach(task))
664 page = (char *)__get_free_page(GFP_TEMPORARY);
670 mm = get_task_mm(task);
675 int this_len, retval, max_len;
677 this_len = mm->env_end - (mm->env_start + src);
682 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
683 this_len = (this_len > max_len) ? max_len : this_len;
685 retval = access_process_vm(task, (mm->env_start + src),
693 if (copy_to_user(buf, page, retval)) {
707 free_page((unsigned long) page);
709 put_task_struct(task);
714 static const struct file_operations proc_environ_operations = {
715 .read = environ_read,
718 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
719 size_t count, loff_t *ppos)
721 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
722 char buffer[PROC_NUMBUF];
728 oom_adjust = task->oomkilladj;
729 put_task_struct(task);
731 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
733 return simple_read_from_buffer(buf, count, ppos, buffer, len);
736 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
737 size_t count, loff_t *ppos)
739 struct task_struct *task;
740 char buffer[PROC_NUMBUF], *end;
743 memset(buffer, 0, sizeof(buffer));
744 if (count > sizeof(buffer) - 1)
745 count = sizeof(buffer) - 1;
746 if (copy_from_user(buffer, buf, count))
748 oom_adjust = simple_strtol(buffer, &end, 0);
749 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
750 oom_adjust != OOM_DISABLE)
754 task = get_proc_task(file->f_path.dentry->d_inode);
757 if (oom_adjust < task->oomkilladj && !capable(CAP_SYS_RESOURCE)) {
758 put_task_struct(task);
761 task->oomkilladj = oom_adjust;
762 put_task_struct(task);
763 if (end - buffer == 0)
768 static const struct file_operations proc_oom_adjust_operations = {
769 .read = oom_adjust_read,
770 .write = oom_adjust_write,
774 static ssize_t clear_refs_write(struct file *file, const char __user *buf,
775 size_t count, loff_t *ppos)
777 struct task_struct *task;
778 char buffer[PROC_NUMBUF], *end;
779 struct mm_struct *mm;
781 memset(buffer, 0, sizeof(buffer));
782 if (count > sizeof(buffer) - 1)
783 count = sizeof(buffer) - 1;
784 if (copy_from_user(buffer, buf, count))
786 if (!simple_strtol(buffer, &end, 0))
790 task = get_proc_task(file->f_path.dentry->d_inode);
793 mm = get_task_mm(task);
798 put_task_struct(task);
799 if (end - buffer == 0)
804 static struct file_operations proc_clear_refs_operations = {
805 .write = clear_refs_write,
809 #ifdef CONFIG_AUDITSYSCALL
811 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
812 size_t count, loff_t *ppos)
814 struct inode * inode = file->f_path.dentry->d_inode;
815 struct task_struct *task = get_proc_task(inode);
817 char tmpbuf[TMPBUFLEN];
821 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
822 audit_get_loginuid(task->audit_context));
823 put_task_struct(task);
824 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
827 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
828 size_t count, loff_t *ppos)
830 struct inode * inode = file->f_path.dentry->d_inode;
835 if (!capable(CAP_AUDIT_CONTROL))
838 if (current != pid_task(proc_pid(inode), PIDTYPE_PID))
841 if (count >= PAGE_SIZE)
842 count = PAGE_SIZE - 1;
845 /* No partial writes. */
848 page = (char*)__get_free_page(GFP_TEMPORARY);
852 if (copy_from_user(page, buf, count))
856 loginuid = simple_strtoul(page, &tmp, 10);
862 length = audit_set_loginuid(current, loginuid);
863 if (likely(length == 0))
867 free_page((unsigned long) page);
871 static const struct file_operations proc_loginuid_operations = {
872 .read = proc_loginuid_read,
873 .write = proc_loginuid_write,
877 #ifdef CONFIG_FAULT_INJECTION
878 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
879 size_t count, loff_t *ppos)
881 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
882 char buffer[PROC_NUMBUF];
888 make_it_fail = task->make_it_fail;
889 put_task_struct(task);
891 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
893 return simple_read_from_buffer(buf, count, ppos, buffer, len);
896 static ssize_t proc_fault_inject_write(struct file * file,
897 const char __user * buf, size_t count, loff_t *ppos)
899 struct task_struct *task;
900 char buffer[PROC_NUMBUF], *end;
903 if (!capable(CAP_SYS_RESOURCE))
905 memset(buffer, 0, sizeof(buffer));
906 if (count > sizeof(buffer) - 1)
907 count = sizeof(buffer) - 1;
908 if (copy_from_user(buffer, buf, count))
910 make_it_fail = simple_strtol(buffer, &end, 0);
913 task = get_proc_task(file->f_dentry->d_inode);
916 task->make_it_fail = make_it_fail;
917 put_task_struct(task);
918 if (end - buffer == 0)
923 static const struct file_operations proc_fault_inject_operations = {
924 .read = proc_fault_inject_read,
925 .write = proc_fault_inject_write,
929 #ifdef CONFIG_SCHED_DEBUG
931 * Print out various scheduling related per-task fields:
933 static int sched_show(struct seq_file *m, void *v)
935 struct inode *inode = m->private;
936 struct task_struct *p;
940 p = get_proc_task(inode);
943 proc_sched_show_task(p, m);
951 sched_write(struct file *file, const char __user *buf,
952 size_t count, loff_t *offset)
954 struct inode *inode = file->f_path.dentry->d_inode;
955 struct task_struct *p;
959 p = get_proc_task(inode);
962 proc_sched_set_task(p);
969 static int sched_open(struct inode *inode, struct file *filp)
973 ret = single_open(filp, sched_show, NULL);
975 struct seq_file *m = filp->private_data;
982 static const struct file_operations proc_pid_sched_operations = {
985 .write = sched_write,
987 .release = single_release,
992 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
994 struct inode *inode = dentry->d_inode;
997 /* We don't need a base pointer in the /proc filesystem */
1000 /* Are we allowed to snoop on the tasks file descriptors? */
1001 if (!proc_fd_access_allowed(inode))
1004 error = PROC_I(inode)->op.proc_get_link(inode, &nd->dentry, &nd->mnt);
1005 nd->last_type = LAST_BIND;
1007 return ERR_PTR(error);
1010 static int do_proc_readlink(struct dentry *dentry, struct vfsmount *mnt,
1011 char __user *buffer, int buflen)
1013 struct inode * inode;
1014 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1021 inode = dentry->d_inode;
1022 path = d_path(dentry, mnt, tmp, PAGE_SIZE);
1023 len = PTR_ERR(path);
1026 len = tmp + PAGE_SIZE - 1 - path;
1030 if (copy_to_user(buffer, path, len))
1033 free_page((unsigned long)tmp);
1037 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1039 int error = -EACCES;
1040 struct inode *inode = dentry->d_inode;
1042 struct vfsmount *mnt = NULL;
1044 /* Are we allowed to snoop on the tasks file descriptors? */
1045 if (!proc_fd_access_allowed(inode))
1048 error = PROC_I(inode)->op.proc_get_link(inode, &de, &mnt);
1052 error = do_proc_readlink(de, mnt, buffer, buflen);
1059 static const struct inode_operations proc_pid_link_inode_operations = {
1060 .readlink = proc_pid_readlink,
1061 .follow_link = proc_pid_follow_link,
1062 .setattr = proc_setattr,
1066 /* building an inode */
1068 static int task_dumpable(struct task_struct *task)
1071 struct mm_struct *mm;
1076 dumpable = get_dumpable(mm);
1084 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1086 struct inode * inode;
1087 struct proc_inode *ei;
1089 /* We need a new inode */
1091 inode = new_inode(sb);
1097 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1098 inode->i_op = &proc_def_inode_operations;
1101 * grab the reference to task.
1103 ei->pid = get_task_pid(task, PIDTYPE_PID);
1109 if (task_dumpable(task)) {
1110 inode->i_uid = task->euid;
1111 inode->i_gid = task->egid;
1113 security_task_to_inode(task, inode);
1123 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1125 struct inode *inode = dentry->d_inode;
1126 struct task_struct *task;
1127 generic_fillattr(inode, stat);
1132 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1134 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1135 task_dumpable(task)) {
1136 stat->uid = task->euid;
1137 stat->gid = task->egid;
1147 * Exceptional case: normally we are not allowed to unhash a busy
1148 * directory. In this case, however, we can do it - no aliasing problems
1149 * due to the way we treat inodes.
1151 * Rewrite the inode's ownerships here because the owning task may have
1152 * performed a setuid(), etc.
1154 * Before the /proc/pid/status file was created the only way to read
1155 * the effective uid of a /process was to stat /proc/pid. Reading
1156 * /proc/pid/status is slow enough that procps and other packages
1157 * kept stating /proc/pid. To keep the rules in /proc simple I have
1158 * made this apply to all per process world readable and executable
1161 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1163 struct inode *inode = dentry->d_inode;
1164 struct task_struct *task = get_proc_task(inode);
1166 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1167 task_dumpable(task)) {
1168 inode->i_uid = task->euid;
1169 inode->i_gid = task->egid;
1174 inode->i_mode &= ~(S_ISUID | S_ISGID);
1175 security_task_to_inode(task, inode);
1176 put_task_struct(task);
1183 static int pid_delete_dentry(struct dentry * dentry)
1185 /* Is the task we represent dead?
1186 * If so, then don't put the dentry on the lru list,
1187 * kill it immediately.
1189 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1192 static struct dentry_operations pid_dentry_operations =
1194 .d_revalidate = pid_revalidate,
1195 .d_delete = pid_delete_dentry,
1200 typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1201 struct task_struct *, const void *);
1204 * Fill a directory entry.
1206 * If possible create the dcache entry and derive our inode number and
1207 * file type from dcache entry.
1209 * Since all of the proc inode numbers are dynamically generated, the inode
1210 * numbers do not exist until the inode is cache. This means creating the
1211 * the dcache entry in readdir is necessary to keep the inode numbers
1212 * reported by readdir in sync with the inode numbers reported
1215 static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1216 char *name, int len,
1217 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1219 struct dentry *child, *dir = filp->f_path.dentry;
1220 struct inode *inode;
1223 unsigned type = DT_UNKNOWN;
1227 qname.hash = full_name_hash(name, len);
1229 child = d_lookup(dir, &qname);
1232 new = d_alloc(dir, &qname);
1234 child = instantiate(dir->d_inode, new, task, ptr);
1241 if (!child || IS_ERR(child) || !child->d_inode)
1242 goto end_instantiate;
1243 inode = child->d_inode;
1246 type = inode->i_mode >> 12;
1251 ino = find_inode_number(dir, &qname);
1254 return filldir(dirent, name, len, filp->f_pos, ino, type);
1257 static unsigned name_to_int(struct dentry *dentry)
1259 const char *name = dentry->d_name.name;
1260 int len = dentry->d_name.len;
1263 if (len > 1 && *name == '0')
1266 unsigned c = *name++ - '0';
1269 if (n >= (~0U-9)/10)
1279 #define PROC_FDINFO_MAX 64
1281 static int proc_fd_info(struct inode *inode, struct dentry **dentry,
1282 struct vfsmount **mnt, char *info)
1284 struct task_struct *task = get_proc_task(inode);
1285 struct files_struct *files = NULL;
1287 int fd = proc_fd(inode);
1290 files = get_files_struct(task);
1291 put_task_struct(task);
1295 * We are not taking a ref to the file structure, so we must
1298 spin_lock(&files->file_lock);
1299 file = fcheck_files(files, fd);
1302 *mnt = mntget(file->f_path.mnt);
1304 *dentry = dget(file->f_path.dentry);
1306 snprintf(info, PROC_FDINFO_MAX,
1309 (long long) file->f_pos,
1311 spin_unlock(&files->file_lock);
1312 put_files_struct(files);
1315 spin_unlock(&files->file_lock);
1316 put_files_struct(files);
1321 static int proc_fd_link(struct inode *inode, struct dentry **dentry,
1322 struct vfsmount **mnt)
1324 return proc_fd_info(inode, dentry, mnt, NULL);
1327 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1329 struct inode *inode = dentry->d_inode;
1330 struct task_struct *task = get_proc_task(inode);
1331 int fd = proc_fd(inode);
1332 struct files_struct *files;
1335 files = get_files_struct(task);
1338 if (fcheck_files(files, fd)) {
1340 put_files_struct(files);
1341 if (task_dumpable(task)) {
1342 inode->i_uid = task->euid;
1343 inode->i_gid = task->egid;
1348 inode->i_mode &= ~(S_ISUID | S_ISGID);
1349 security_task_to_inode(task, inode);
1350 put_task_struct(task);
1354 put_files_struct(files);
1356 put_task_struct(task);
1362 static struct dentry_operations tid_fd_dentry_operations =
1364 .d_revalidate = tid_fd_revalidate,
1365 .d_delete = pid_delete_dentry,
1368 static struct dentry *proc_fd_instantiate(struct inode *dir,
1369 struct dentry *dentry, struct task_struct *task, const void *ptr)
1371 unsigned fd = *(const unsigned *)ptr;
1373 struct files_struct *files;
1374 struct inode *inode;
1375 struct proc_inode *ei;
1376 struct dentry *error = ERR_PTR(-ENOENT);
1378 inode = proc_pid_make_inode(dir->i_sb, task);
1383 files = get_files_struct(task);
1386 inode->i_mode = S_IFLNK;
1389 * We are not taking a ref to the file structure, so we must
1392 spin_lock(&files->file_lock);
1393 file = fcheck_files(files, fd);
1396 if (file->f_mode & 1)
1397 inode->i_mode |= S_IRUSR | S_IXUSR;
1398 if (file->f_mode & 2)
1399 inode->i_mode |= S_IWUSR | S_IXUSR;
1400 spin_unlock(&files->file_lock);
1401 put_files_struct(files);
1403 inode->i_op = &proc_pid_link_inode_operations;
1405 ei->op.proc_get_link = proc_fd_link;
1406 dentry->d_op = &tid_fd_dentry_operations;
1407 d_add(dentry, inode);
1408 /* Close the race of the process dying before we return the dentry */
1409 if (tid_fd_revalidate(dentry, NULL))
1415 spin_unlock(&files->file_lock);
1416 put_files_struct(files);
1422 static struct dentry *proc_lookupfd_common(struct inode *dir,
1423 struct dentry *dentry,
1424 instantiate_t instantiate)
1426 struct task_struct *task = get_proc_task(dir);
1427 unsigned fd = name_to_int(dentry);
1428 struct dentry *result = ERR_PTR(-ENOENT);
1435 result = instantiate(dir, dentry, task, &fd);
1437 put_task_struct(task);
1442 static int proc_readfd_common(struct file * filp, void * dirent,
1443 filldir_t filldir, instantiate_t instantiate)
1445 struct dentry *dentry = filp->f_path.dentry;
1446 struct inode *inode = dentry->d_inode;
1447 struct task_struct *p = get_proc_task(inode);
1448 unsigned int fd, tid, ino;
1450 struct files_struct * files;
1451 struct fdtable *fdt;
1462 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1466 ino = parent_ino(dentry);
1467 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1471 files = get_files_struct(p);
1475 fdt = files_fdtable(files);
1476 for (fd = filp->f_pos-2;
1478 fd++, filp->f_pos++) {
1479 char name[PROC_NUMBUF];
1482 if (!fcheck_files(files, fd))
1486 len = snprintf(name, sizeof(name), "%d", fd);
1487 if (proc_fill_cache(filp, dirent, filldir,
1488 name, len, instantiate,
1496 put_files_struct(files);
1504 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1505 struct nameidata *nd)
1507 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1510 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1512 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1515 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1516 size_t len, loff_t *ppos)
1518 char tmp[PROC_FDINFO_MAX];
1519 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, NULL, tmp);
1521 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1525 static const struct file_operations proc_fdinfo_file_operations = {
1526 .open = nonseekable_open,
1527 .read = proc_fdinfo_read,
1530 static const struct file_operations proc_fd_operations = {
1531 .read = generic_read_dir,
1532 .readdir = proc_readfd,
1536 * /proc/pid/fd needs a special permission handler so that a process can still
1537 * access /proc/self/fd after it has executed a setuid().
1539 static int proc_fd_permission(struct inode *inode, int mask,
1540 struct nameidata *nd)
1544 rv = generic_permission(inode, mask, NULL);
1547 if (task_pid(current) == proc_pid(inode))
1553 * proc directories can do almost nothing..
1555 static const struct inode_operations proc_fd_inode_operations = {
1556 .lookup = proc_lookupfd,
1557 .permission = proc_fd_permission,
1558 .setattr = proc_setattr,
1561 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
1562 struct dentry *dentry, struct task_struct *task, const void *ptr)
1564 unsigned fd = *(unsigned *)ptr;
1565 struct inode *inode;
1566 struct proc_inode *ei;
1567 struct dentry *error = ERR_PTR(-ENOENT);
1569 inode = proc_pid_make_inode(dir->i_sb, task);
1574 inode->i_mode = S_IFREG | S_IRUSR;
1575 inode->i_fop = &proc_fdinfo_file_operations;
1576 dentry->d_op = &tid_fd_dentry_operations;
1577 d_add(dentry, inode);
1578 /* Close the race of the process dying before we return the dentry */
1579 if (tid_fd_revalidate(dentry, NULL))
1586 static struct dentry *proc_lookupfdinfo(struct inode *dir,
1587 struct dentry *dentry,
1588 struct nameidata *nd)
1590 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
1593 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
1595 return proc_readfd_common(filp, dirent, filldir,
1596 proc_fdinfo_instantiate);
1599 static const struct file_operations proc_fdinfo_operations = {
1600 .read = generic_read_dir,
1601 .readdir = proc_readfdinfo,
1605 * proc directories can do almost nothing..
1607 static const struct inode_operations proc_fdinfo_inode_operations = {
1608 .lookup = proc_lookupfdinfo,
1609 .setattr = proc_setattr,
1613 static struct dentry *proc_pident_instantiate(struct inode *dir,
1614 struct dentry *dentry, struct task_struct *task, const void *ptr)
1616 const struct pid_entry *p = ptr;
1617 struct inode *inode;
1618 struct proc_inode *ei;
1619 struct dentry *error = ERR_PTR(-EINVAL);
1621 inode = proc_pid_make_inode(dir->i_sb, task);
1626 inode->i_mode = p->mode;
1627 if (S_ISDIR(inode->i_mode))
1628 inode->i_nlink = 2; /* Use getattr to fix if necessary */
1630 inode->i_op = p->iop;
1632 inode->i_fop = p->fop;
1634 dentry->d_op = &pid_dentry_operations;
1635 d_add(dentry, inode);
1636 /* Close the race of the process dying before we return the dentry */
1637 if (pid_revalidate(dentry, NULL))
1643 static struct dentry *proc_pident_lookup(struct inode *dir,
1644 struct dentry *dentry,
1645 const struct pid_entry *ents,
1648 struct inode *inode;
1649 struct dentry *error;
1650 struct task_struct *task = get_proc_task(dir);
1651 const struct pid_entry *p, *last;
1653 error = ERR_PTR(-ENOENT);
1660 * Yes, it does not scale. And it should not. Don't add
1661 * new entries into /proc/<tgid>/ without very good reasons.
1663 last = &ents[nents - 1];
1664 for (p = ents; p <= last; p++) {
1665 if (p->len != dentry->d_name.len)
1667 if (!memcmp(dentry->d_name.name, p->name, p->len))
1673 error = proc_pident_instantiate(dir, dentry, task, p);
1675 put_task_struct(task);
1680 static int proc_pident_fill_cache(struct file *filp, void *dirent,
1681 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
1683 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
1684 proc_pident_instantiate, task, p);
1687 static int proc_pident_readdir(struct file *filp,
1688 void *dirent, filldir_t filldir,
1689 const struct pid_entry *ents, unsigned int nents)
1693 struct dentry *dentry = filp->f_path.dentry;
1694 struct inode *inode = dentry->d_inode;
1695 struct task_struct *task = get_proc_task(inode);
1696 const struct pid_entry *p, *last;
1710 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
1716 ino = parent_ino(dentry);
1717 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
1729 last = &ents[nents - 1];
1731 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
1740 put_task_struct(task);
1745 #ifdef CONFIG_SECURITY
1746 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
1747 size_t count, loff_t *ppos)
1749 struct inode * inode = file->f_path.dentry->d_inode;
1752 struct task_struct *task = get_proc_task(inode);
1757 length = security_getprocattr(task,
1758 (char*)file->f_path.dentry->d_name.name,
1760 put_task_struct(task);
1762 length = simple_read_from_buffer(buf, count, ppos, p, length);
1767 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
1768 size_t count, loff_t *ppos)
1770 struct inode * inode = file->f_path.dentry->d_inode;
1773 struct task_struct *task = get_proc_task(inode);
1778 if (count > PAGE_SIZE)
1781 /* No partial writes. */
1787 page = (char*)__get_free_page(GFP_TEMPORARY);
1792 if (copy_from_user(page, buf, count))
1795 length = security_setprocattr(task,
1796 (char*)file->f_path.dentry->d_name.name,
1797 (void*)page, count);
1799 free_page((unsigned long) page);
1801 put_task_struct(task);
1806 static const struct file_operations proc_pid_attr_operations = {
1807 .read = proc_pid_attr_read,
1808 .write = proc_pid_attr_write,
1811 static const struct pid_entry attr_dir_stuff[] = {
1812 REG("current", S_IRUGO|S_IWUGO, pid_attr),
1813 REG("prev", S_IRUGO, pid_attr),
1814 REG("exec", S_IRUGO|S_IWUGO, pid_attr),
1815 REG("fscreate", S_IRUGO|S_IWUGO, pid_attr),
1816 REG("keycreate", S_IRUGO|S_IWUGO, pid_attr),
1817 REG("sockcreate", S_IRUGO|S_IWUGO, pid_attr),
1820 static int proc_attr_dir_readdir(struct file * filp,
1821 void * dirent, filldir_t filldir)
1823 return proc_pident_readdir(filp,dirent,filldir,
1824 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
1827 static const struct file_operations proc_attr_dir_operations = {
1828 .read = generic_read_dir,
1829 .readdir = proc_attr_dir_readdir,
1832 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
1833 struct dentry *dentry, struct nameidata *nd)
1835 return proc_pident_lookup(dir, dentry,
1836 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
1839 static const struct inode_operations proc_attr_dir_inode_operations = {
1840 .lookup = proc_attr_dir_lookup,
1841 .getattr = pid_getattr,
1842 .setattr = proc_setattr,
1847 #if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
1848 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
1849 size_t count, loff_t *ppos)
1851 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1852 struct mm_struct *mm;
1853 char buffer[PROC_NUMBUF];
1861 mm = get_task_mm(task);
1863 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
1864 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
1865 MMF_DUMP_FILTER_SHIFT));
1867 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
1870 put_task_struct(task);
1875 static ssize_t proc_coredump_filter_write(struct file *file,
1876 const char __user *buf,
1880 struct task_struct *task;
1881 struct mm_struct *mm;
1882 char buffer[PROC_NUMBUF], *end;
1889 memset(buffer, 0, sizeof(buffer));
1890 if (count > sizeof(buffer) - 1)
1891 count = sizeof(buffer) - 1;
1892 if (copy_from_user(buffer, buf, count))
1896 val = (unsigned int)simple_strtoul(buffer, &end, 0);
1899 if (end - buffer == 0)
1903 task = get_proc_task(file->f_dentry->d_inode);
1908 mm = get_task_mm(task);
1912 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
1914 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
1916 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
1921 put_task_struct(task);
1926 static const struct file_operations proc_coredump_filter_operations = {
1927 .read = proc_coredump_filter_read,
1928 .write = proc_coredump_filter_write,
1935 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
1938 char tmp[PROC_NUMBUF];
1939 sprintf(tmp, "%d", current->tgid);
1940 return vfs_readlink(dentry,buffer,buflen,tmp);
1943 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
1945 char tmp[PROC_NUMBUF];
1946 sprintf(tmp, "%d", current->tgid);
1947 return ERR_PTR(vfs_follow_link(nd,tmp));
1950 static const struct inode_operations proc_self_inode_operations = {
1951 .readlink = proc_self_readlink,
1952 .follow_link = proc_self_follow_link,
1958 * These are the directory entries in the root directory of /proc
1959 * that properly belong to the /proc filesystem, as they describe
1960 * describe something that is process related.
1962 static const struct pid_entry proc_base_stuff[] = {
1963 NOD("self", S_IFLNK|S_IRWXUGO,
1964 &proc_self_inode_operations, NULL, {}),
1968 * Exceptional case: normally we are not allowed to unhash a busy
1969 * directory. In this case, however, we can do it - no aliasing problems
1970 * due to the way we treat inodes.
1972 static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
1974 struct inode *inode = dentry->d_inode;
1975 struct task_struct *task = get_proc_task(inode);
1977 put_task_struct(task);
1984 static struct dentry_operations proc_base_dentry_operations =
1986 .d_revalidate = proc_base_revalidate,
1987 .d_delete = pid_delete_dentry,
1990 static struct dentry *proc_base_instantiate(struct inode *dir,
1991 struct dentry *dentry, struct task_struct *task, const void *ptr)
1993 const struct pid_entry *p = ptr;
1994 struct inode *inode;
1995 struct proc_inode *ei;
1996 struct dentry *error = ERR_PTR(-EINVAL);
1998 /* Allocate the inode */
1999 error = ERR_PTR(-ENOMEM);
2000 inode = new_inode(dir->i_sb);
2004 /* Initialize the inode */
2006 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2009 * grab the reference to the task.
2011 ei->pid = get_task_pid(task, PIDTYPE_PID);
2017 inode->i_mode = p->mode;
2018 if (S_ISDIR(inode->i_mode))
2020 if (S_ISLNK(inode->i_mode))
2023 inode->i_op = p->iop;
2025 inode->i_fop = p->fop;
2027 dentry->d_op = &proc_base_dentry_operations;
2028 d_add(dentry, inode);
2037 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2039 struct dentry *error;
2040 struct task_struct *task = get_proc_task(dir);
2041 const struct pid_entry *p, *last;
2043 error = ERR_PTR(-ENOENT);
2048 /* Lookup the directory entry */
2049 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2050 for (p = proc_base_stuff; p <= last; p++) {
2051 if (p->len != dentry->d_name.len)
2053 if (!memcmp(dentry->d_name.name, p->name, p->len))
2059 error = proc_base_instantiate(dir, dentry, task, p);
2062 put_task_struct(task);
2067 static int proc_base_fill_cache(struct file *filp, void *dirent,
2068 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2070 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2071 proc_base_instantiate, task, p);
2074 #ifdef CONFIG_TASK_IO_ACCOUNTING
2075 static int proc_pid_io_accounting(struct task_struct *task, char *buffer)
2077 return sprintf(buffer,
2078 #ifdef CONFIG_TASK_XACCT
2084 "read_bytes: %llu\n"
2085 "write_bytes: %llu\n"
2086 "cancelled_write_bytes: %llu\n",
2087 #ifdef CONFIG_TASK_XACCT
2088 (unsigned long long)task->rchar,
2089 (unsigned long long)task->wchar,
2090 (unsigned long long)task->syscr,
2091 (unsigned long long)task->syscw,
2093 (unsigned long long)task->ioac.read_bytes,
2094 (unsigned long long)task->ioac.write_bytes,
2095 (unsigned long long)task->ioac.cancelled_write_bytes);
2102 static const struct file_operations proc_task_operations;
2103 static const struct inode_operations proc_task_inode_operations;
2105 static const struct pid_entry tgid_base_stuff[] = {
2106 DIR("task", S_IRUGO|S_IXUGO, task),
2107 DIR("fd", S_IRUSR|S_IXUSR, fd),
2108 DIR("fdinfo", S_IRUSR|S_IXUSR, fdinfo),
2109 REG("environ", S_IRUSR, environ),
2110 INF("auxv", S_IRUSR, pid_auxv),
2111 INF("status", S_IRUGO, pid_status),
2112 #ifdef CONFIG_SCHED_DEBUG
2113 REG("sched", S_IRUGO|S_IWUSR, pid_sched),
2115 INF("cmdline", S_IRUGO, pid_cmdline),
2116 INF("stat", S_IRUGO, tgid_stat),
2117 INF("statm", S_IRUGO, pid_statm),
2118 REG("maps", S_IRUGO, maps),
2120 REG("numa_maps", S_IRUGO, numa_maps),
2122 REG("mem", S_IRUSR|S_IWUSR, mem),
2126 REG("mounts", S_IRUGO, mounts),
2127 REG("mountstats", S_IRUSR, mountstats),
2129 REG("clear_refs", S_IWUSR, clear_refs),
2130 REG("smaps", S_IRUGO, smaps),
2132 #ifdef CONFIG_SECURITY
2133 DIR("attr", S_IRUGO|S_IXUGO, attr_dir),
2135 #ifdef CONFIG_KALLSYMS
2136 INF("wchan", S_IRUGO, pid_wchan),
2138 #ifdef CONFIG_SCHEDSTATS
2139 INF("schedstat", S_IRUGO, pid_schedstat),
2141 #ifdef CONFIG_PROC_PID_CPUSET
2142 REG("cpuset", S_IRUGO, cpuset),
2144 #ifdef CONFIG_CGROUPS
2145 REG("cgroup", S_IRUGO, cgroup),
2147 INF("oom_score", S_IRUGO, oom_score),
2148 REG("oom_adj", S_IRUGO|S_IWUSR, oom_adjust),
2149 #ifdef CONFIG_AUDITSYSCALL
2150 REG("loginuid", S_IWUSR|S_IRUGO, loginuid),
2152 #ifdef CONFIG_FAULT_INJECTION
2153 REG("make-it-fail", S_IRUGO|S_IWUSR, fault_inject),
2155 #if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
2156 REG("coredump_filter", S_IRUGO|S_IWUSR, coredump_filter),
2158 #ifdef CONFIG_TASK_IO_ACCOUNTING
2159 INF("io", S_IRUGO, pid_io_accounting),
2163 static int proc_tgid_base_readdir(struct file * filp,
2164 void * dirent, filldir_t filldir)
2166 return proc_pident_readdir(filp,dirent,filldir,
2167 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2170 static const struct file_operations proc_tgid_base_operations = {
2171 .read = generic_read_dir,
2172 .readdir = proc_tgid_base_readdir,
2175 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2176 return proc_pident_lookup(dir, dentry,
2177 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2180 static const struct inode_operations proc_tgid_base_inode_operations = {
2181 .lookup = proc_tgid_base_lookup,
2182 .getattr = pid_getattr,
2183 .setattr = proc_setattr,
2187 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2189 * @task: task that should be flushed.
2191 * Looks in the dcache for
2193 * /proc/@tgid/task/@pid
2194 * if either directory is present flushes it and all of it'ts children
2197 * It is safe and reasonable to cache /proc entries for a task until
2198 * that task exits. After that they just clog up the dcache with
2199 * useless entries, possibly causing useful dcache entries to be
2200 * flushed instead. This routine is proved to flush those useless
2201 * dcache entries at process exit time.
2203 * NOTE: This routine is just an optimization so it does not guarantee
2204 * that no dcache entries will exist at process exit time it
2205 * just makes it very unlikely that any will persist.
2207 void proc_flush_task(struct task_struct *task)
2209 struct dentry *dentry, *leader, *dir;
2210 char buf[PROC_NUMBUF];
2214 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
2215 dentry = d_hash_and_lookup(proc_mnt->mnt_root, &name);
2217 shrink_dcache_parent(dentry);
2222 if (thread_group_leader(task))
2226 name.len = snprintf(buf, sizeof(buf), "%d", task->tgid);
2227 leader = d_hash_and_lookup(proc_mnt->mnt_root, &name);
2232 name.len = strlen(name.name);
2233 dir = d_hash_and_lookup(leader, &name);
2235 goto out_put_leader;
2238 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
2239 dentry = d_hash_and_lookup(dir, &name);
2241 shrink_dcache_parent(dentry);
2253 static struct dentry *proc_pid_instantiate(struct inode *dir,
2254 struct dentry * dentry,
2255 struct task_struct *task, const void *ptr)
2257 struct dentry *error = ERR_PTR(-ENOENT);
2258 struct inode *inode;
2260 inode = proc_pid_make_inode(dir->i_sb, task);
2264 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2265 inode->i_op = &proc_tgid_base_inode_operations;
2266 inode->i_fop = &proc_tgid_base_operations;
2267 inode->i_flags|=S_IMMUTABLE;
2269 #ifdef CONFIG_SECURITY
2270 inode->i_nlink += 1;
2273 dentry->d_op = &pid_dentry_operations;
2275 d_add(dentry, inode);
2276 /* Close the race of the process dying before we return the dentry */
2277 if (pid_revalidate(dentry, NULL))
2283 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2285 struct dentry *result = ERR_PTR(-ENOENT);
2286 struct task_struct *task;
2289 result = proc_base_lookup(dir, dentry);
2290 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2293 tgid = name_to_int(dentry);
2298 task = find_task_by_pid(tgid);
2300 get_task_struct(task);
2305 result = proc_pid_instantiate(dir, dentry, task, NULL);
2306 put_task_struct(task);
2312 * Find the first task with tgid >= tgid
2315 static struct task_struct *next_tgid(unsigned int tgid)
2317 struct task_struct *task;
2323 pid = find_ge_pid(tgid);
2326 task = pid_task(pid, PIDTYPE_PID);
2327 /* What we to know is if the pid we have find is the
2328 * pid of a thread_group_leader. Testing for task
2329 * being a thread_group_leader is the obvious thing
2330 * todo but there is a window when it fails, due to
2331 * the pid transfer logic in de_thread.
2333 * So we perform the straight forward test of seeing
2334 * if the pid we have found is the pid of a thread
2335 * group leader, and don't worry if the task we have
2336 * found doesn't happen to be a thread group leader.
2337 * As we don't care in the case of readdir.
2339 if (!task || !has_group_leader_pid(task))
2341 get_task_struct(task);
2347 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2349 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2350 struct task_struct *task, int tgid)
2352 char name[PROC_NUMBUF];
2353 int len = snprintf(name, sizeof(name), "%d", tgid);
2354 return proc_fill_cache(filp, dirent, filldir, name, len,
2355 proc_pid_instantiate, task, NULL);
2358 /* for the /proc/ directory itself, after non-process stuff has been done */
2359 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2361 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2362 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
2363 struct task_struct *task;
2369 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
2370 const struct pid_entry *p = &proc_base_stuff[nr];
2371 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
2375 tgid = filp->f_pos - TGID_OFFSET;
2376 for (task = next_tgid(tgid);
2378 put_task_struct(task), task = next_tgid(tgid + 1)) {
2380 filp->f_pos = tgid + TGID_OFFSET;
2381 if (proc_pid_fill_cache(filp, dirent, filldir, task, tgid) < 0) {
2382 put_task_struct(task);
2386 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
2388 put_task_struct(reaper);
2396 static const struct pid_entry tid_base_stuff[] = {
2397 DIR("fd", S_IRUSR|S_IXUSR, fd),
2398 DIR("fdinfo", S_IRUSR|S_IXUSR, fdinfo),
2399 REG("environ", S_IRUSR, environ),
2400 INF("auxv", S_IRUSR, pid_auxv),
2401 INF("status", S_IRUGO, pid_status),
2402 #ifdef CONFIG_SCHED_DEBUG
2403 REG("sched", S_IRUGO|S_IWUSR, pid_sched),
2405 INF("cmdline", S_IRUGO, pid_cmdline),
2406 INF("stat", S_IRUGO, tid_stat),
2407 INF("statm", S_IRUGO, pid_statm),
2408 REG("maps", S_IRUGO, maps),
2410 REG("numa_maps", S_IRUGO, numa_maps),
2412 REG("mem", S_IRUSR|S_IWUSR, mem),
2416 REG("mounts", S_IRUGO, mounts),
2418 REG("clear_refs", S_IWUSR, clear_refs),
2419 REG("smaps", S_IRUGO, smaps),
2421 #ifdef CONFIG_SECURITY
2422 DIR("attr", S_IRUGO|S_IXUGO, attr_dir),
2424 #ifdef CONFIG_KALLSYMS
2425 INF("wchan", S_IRUGO, pid_wchan),
2427 #ifdef CONFIG_SCHEDSTATS
2428 INF("schedstat", S_IRUGO, pid_schedstat),
2430 #ifdef CONFIG_PROC_PID_CPUSET
2431 REG("cpuset", S_IRUGO, cpuset),
2433 #ifdef CONFIG_CGROUPS
2434 REG("cgroup", S_IRUGO, cgroup),
2436 INF("oom_score", S_IRUGO, oom_score),
2437 REG("oom_adj", S_IRUGO|S_IWUSR, oom_adjust),
2438 #ifdef CONFIG_AUDITSYSCALL
2439 REG("loginuid", S_IWUSR|S_IRUGO, loginuid),
2441 #ifdef CONFIG_FAULT_INJECTION
2442 REG("make-it-fail", S_IRUGO|S_IWUSR, fault_inject),
2446 static int proc_tid_base_readdir(struct file * filp,
2447 void * dirent, filldir_t filldir)
2449 return proc_pident_readdir(filp,dirent,filldir,
2450 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
2453 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2454 return proc_pident_lookup(dir, dentry,
2455 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
2458 static const struct file_operations proc_tid_base_operations = {
2459 .read = generic_read_dir,
2460 .readdir = proc_tid_base_readdir,
2463 static const struct inode_operations proc_tid_base_inode_operations = {
2464 .lookup = proc_tid_base_lookup,
2465 .getattr = pid_getattr,
2466 .setattr = proc_setattr,
2469 static struct dentry *proc_task_instantiate(struct inode *dir,
2470 struct dentry *dentry, struct task_struct *task, const void *ptr)
2472 struct dentry *error = ERR_PTR(-ENOENT);
2473 struct inode *inode;
2474 inode = proc_pid_make_inode(dir->i_sb, task);
2478 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2479 inode->i_op = &proc_tid_base_inode_operations;
2480 inode->i_fop = &proc_tid_base_operations;
2481 inode->i_flags|=S_IMMUTABLE;
2483 #ifdef CONFIG_SECURITY
2484 inode->i_nlink += 1;
2487 dentry->d_op = &pid_dentry_operations;
2489 d_add(dentry, inode);
2490 /* Close the race of the process dying before we return the dentry */
2491 if (pid_revalidate(dentry, NULL))
2497 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2499 struct dentry *result = ERR_PTR(-ENOENT);
2500 struct task_struct *task;
2501 struct task_struct *leader = get_proc_task(dir);
2507 tid = name_to_int(dentry);
2512 task = find_task_by_pid(tid);
2514 get_task_struct(task);
2518 if (leader->tgid != task->tgid)
2521 result = proc_task_instantiate(dir, dentry, task, NULL);
2523 put_task_struct(task);
2525 put_task_struct(leader);
2531 * Find the first tid of a thread group to return to user space.
2533 * Usually this is just the thread group leader, but if the users
2534 * buffer was too small or there was a seek into the middle of the
2535 * directory we have more work todo.
2537 * In the case of a short read we start with find_task_by_pid.
2539 * In the case of a seek we start with the leader and walk nr
2542 static struct task_struct *first_tid(struct task_struct *leader,
2545 struct task_struct *pos;
2548 /* Attempt to start with the pid of a thread */
2549 if (tid && (nr > 0)) {
2550 pos = find_task_by_pid(tid);
2551 if (pos && (pos->group_leader == leader))
2555 /* If nr exceeds the number of threads there is nothing todo */
2557 if (nr && nr >= get_nr_threads(leader))
2560 /* If we haven't found our starting place yet start
2561 * with the leader and walk nr threads forward.
2563 for (pos = leader; nr > 0; --nr) {
2564 pos = next_thread(pos);
2565 if (pos == leader) {
2571 get_task_struct(pos);
2578 * Find the next thread in the thread list.
2579 * Return NULL if there is an error or no next thread.
2581 * The reference to the input task_struct is released.
2583 static struct task_struct *next_tid(struct task_struct *start)
2585 struct task_struct *pos = NULL;
2587 if (pid_alive(start)) {
2588 pos = next_thread(start);
2589 if (thread_group_leader(pos))
2592 get_task_struct(pos);
2595 put_task_struct(start);
2599 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2600 struct task_struct *task, int tid)
2602 char name[PROC_NUMBUF];
2603 int len = snprintf(name, sizeof(name), "%d", tid);
2604 return proc_fill_cache(filp, dirent, filldir, name, len,
2605 proc_task_instantiate, task, NULL);
2608 /* for the /proc/TGID/task/ directories */
2609 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
2611 struct dentry *dentry = filp->f_path.dentry;
2612 struct inode *inode = dentry->d_inode;
2613 struct task_struct *leader = NULL;
2614 struct task_struct *task;
2615 int retval = -ENOENT;
2618 unsigned long pos = filp->f_pos; /* avoiding "long long" filp->f_pos */
2620 task = get_proc_task(inode);
2624 if (pid_alive(task)) {
2625 leader = task->group_leader;
2626 get_task_struct(leader);
2629 put_task_struct(task);
2637 if (filldir(dirent, ".", 1, pos, ino, DT_DIR) < 0)
2642 ino = parent_ino(dentry);
2643 if (filldir(dirent, "..", 2, pos, ino, DT_DIR) < 0)
2649 /* f_version caches the tgid value that the last readdir call couldn't
2650 * return. lseek aka telldir automagically resets f_version to 0.
2652 tid = (int)filp->f_version;
2653 filp->f_version = 0;
2654 for (task = first_tid(leader, tid, pos - 2);
2656 task = next_tid(task), pos++) {
2658 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
2659 /* returning this tgid failed, save it as the first
2660 * pid for the next readir call */
2661 filp->f_version = (u64)tid;
2662 put_task_struct(task);
2668 put_task_struct(leader);
2673 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
2675 struct inode *inode = dentry->d_inode;
2676 struct task_struct *p = get_proc_task(inode);
2677 generic_fillattr(inode, stat);
2681 stat->nlink += get_nr_threads(p);
2689 static const struct inode_operations proc_task_inode_operations = {
2690 .lookup = proc_task_lookup,
2691 .getattr = proc_task_getattr,
2692 .setattr = proc_setattr,
2695 static const struct file_operations proc_task_operations = {
2696 .read = generic_read_dir,
2697 .readdir = proc_task_readdir,