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/config.h>
53 #include <linux/errno.h>
54 #include <linux/time.h>
55 #include <linux/proc_fs.h>
56 #include <linux/stat.h>
57 #include <linux/init.h>
58 #include <linux/file.h>
59 #include <linux/string.h>
60 #include <linux/seq_file.h>
61 #include <linux/namei.h>
62 #include <linux/namespace.h>
64 #include <linux/smp_lock.h>
65 #include <linux/rcupdate.h>
66 #include <linux/kallsyms.h>
67 #include <linux/mount.h>
68 #include <linux/security.h>
69 #include <linux/ptrace.h>
70 #include <linux/seccomp.h>
71 #include <linux/cpuset.h>
72 #include <linux/audit.h>
76 * For hysterical raisins we keep the same inumbers as in the old procfs.
77 * Feel free to change the macro below - just keep the range distinct from
78 * inumbers of the rest of procfs (currently those are in 0x0000--0xffff).
79 * As soon as we'll get a separate superblock we will be able to forget
80 * about magical ranges too.
83 #define fake_ino(pid,ino) (((pid)<<16)|(ino))
85 enum pid_directory_inos
{
107 #ifdef CONFIG_SCHEDSTATS
110 #ifdef CONFIG_CPUSETS
113 #ifdef CONFIG_SECURITY
115 PROC_TGID_ATTR_CURRENT
,
118 PROC_TGID_ATTR_FSCREATE
,
120 #ifdef CONFIG_AUDITSYSCALL
124 PROC_TGID_OOM_ADJUST
,
128 #ifdef CONFIG_SECCOMP
145 #ifdef CONFIG_SCHEDSTATS
148 #ifdef CONFIG_CPUSETS
151 #ifdef CONFIG_SECURITY
153 PROC_TID_ATTR_CURRENT
,
156 PROC_TID_ATTR_FSCREATE
,
158 #ifdef CONFIG_AUDITSYSCALL
164 /* Add new entries before this */
165 PROC_TID_FD_DIR
= 0x8000, /* 0x8000-0xffff */
175 #define E(type,name,mode) {(type),sizeof(name)-1,(name),(mode)}
177 static struct pid_entry tgid_base_stuff
[] = {
178 E(PROC_TGID_TASK
, "task", S_IFDIR
|S_IRUGO
|S_IXUGO
),
179 E(PROC_TGID_FD
, "fd", S_IFDIR
|S_IRUSR
|S_IXUSR
),
180 E(PROC_TGID_ENVIRON
, "environ", S_IFREG
|S_IRUSR
),
181 E(PROC_TGID_AUXV
, "auxv", S_IFREG
|S_IRUSR
),
182 E(PROC_TGID_STATUS
, "status", S_IFREG
|S_IRUGO
),
183 E(PROC_TGID_CMDLINE
, "cmdline", S_IFREG
|S_IRUGO
),
184 E(PROC_TGID_STAT
, "stat", S_IFREG
|S_IRUGO
),
185 E(PROC_TGID_STATM
, "statm", S_IFREG
|S_IRUGO
),
186 E(PROC_TGID_MAPS
, "maps", S_IFREG
|S_IRUGO
),
188 E(PROC_TGID_NUMA_MAPS
, "numa_maps", S_IFREG
|S_IRUGO
),
190 E(PROC_TGID_MEM
, "mem", S_IFREG
|S_IRUSR
|S_IWUSR
),
191 #ifdef CONFIG_SECCOMP
192 E(PROC_TGID_SECCOMP
, "seccomp", S_IFREG
|S_IRUSR
|S_IWUSR
),
194 E(PROC_TGID_CWD
, "cwd", S_IFLNK
|S_IRWXUGO
),
195 E(PROC_TGID_ROOT
, "root", S_IFLNK
|S_IRWXUGO
),
196 E(PROC_TGID_EXE
, "exe", S_IFLNK
|S_IRWXUGO
),
197 E(PROC_TGID_MOUNTS
, "mounts", S_IFREG
|S_IRUGO
),
198 E(PROC_TGID_SMAPS
, "smaps", S_IFREG
|S_IRUGO
),
199 #ifdef CONFIG_SECURITY
200 E(PROC_TGID_ATTR
, "attr", S_IFDIR
|S_IRUGO
|S_IXUGO
),
202 #ifdef CONFIG_KALLSYMS
203 E(PROC_TGID_WCHAN
, "wchan", S_IFREG
|S_IRUGO
),
205 #ifdef CONFIG_SCHEDSTATS
206 E(PROC_TGID_SCHEDSTAT
, "schedstat", S_IFREG
|S_IRUGO
),
208 #ifdef CONFIG_CPUSETS
209 E(PROC_TGID_CPUSET
, "cpuset", S_IFREG
|S_IRUGO
),
211 E(PROC_TGID_OOM_SCORE
, "oom_score",S_IFREG
|S_IRUGO
),
212 E(PROC_TGID_OOM_ADJUST
,"oom_adj", S_IFREG
|S_IRUGO
|S_IWUSR
),
213 #ifdef CONFIG_AUDITSYSCALL
214 E(PROC_TGID_LOGINUID
, "loginuid", S_IFREG
|S_IWUSR
|S_IRUGO
),
218 static struct pid_entry tid_base_stuff
[] = {
219 E(PROC_TID_FD
, "fd", S_IFDIR
|S_IRUSR
|S_IXUSR
),
220 E(PROC_TID_ENVIRON
, "environ", S_IFREG
|S_IRUSR
),
221 E(PROC_TID_AUXV
, "auxv", S_IFREG
|S_IRUSR
),
222 E(PROC_TID_STATUS
, "status", S_IFREG
|S_IRUGO
),
223 E(PROC_TID_CMDLINE
, "cmdline", S_IFREG
|S_IRUGO
),
224 E(PROC_TID_STAT
, "stat", S_IFREG
|S_IRUGO
),
225 E(PROC_TID_STATM
, "statm", S_IFREG
|S_IRUGO
),
226 E(PROC_TID_MAPS
, "maps", S_IFREG
|S_IRUGO
),
228 E(PROC_TID_NUMA_MAPS
, "numa_maps", S_IFREG
|S_IRUGO
),
230 E(PROC_TID_MEM
, "mem", S_IFREG
|S_IRUSR
|S_IWUSR
),
231 #ifdef CONFIG_SECCOMP
232 E(PROC_TID_SECCOMP
, "seccomp", S_IFREG
|S_IRUSR
|S_IWUSR
),
234 E(PROC_TID_CWD
, "cwd", S_IFLNK
|S_IRWXUGO
),
235 E(PROC_TID_ROOT
, "root", S_IFLNK
|S_IRWXUGO
),
236 E(PROC_TID_EXE
, "exe", S_IFLNK
|S_IRWXUGO
),
237 E(PROC_TID_MOUNTS
, "mounts", S_IFREG
|S_IRUGO
),
238 E(PROC_TID_SMAPS
, "smaps", S_IFREG
|S_IRUGO
),
239 #ifdef CONFIG_SECURITY
240 E(PROC_TID_ATTR
, "attr", S_IFDIR
|S_IRUGO
|S_IXUGO
),
242 #ifdef CONFIG_KALLSYMS
243 E(PROC_TID_WCHAN
, "wchan", S_IFREG
|S_IRUGO
),
245 #ifdef CONFIG_SCHEDSTATS
246 E(PROC_TID_SCHEDSTAT
, "schedstat",S_IFREG
|S_IRUGO
),
248 #ifdef CONFIG_CPUSETS
249 E(PROC_TID_CPUSET
, "cpuset", S_IFREG
|S_IRUGO
),
251 E(PROC_TID_OOM_SCORE
, "oom_score",S_IFREG
|S_IRUGO
),
252 E(PROC_TID_OOM_ADJUST
, "oom_adj", S_IFREG
|S_IRUGO
|S_IWUSR
),
253 #ifdef CONFIG_AUDITSYSCALL
254 E(PROC_TID_LOGINUID
, "loginuid", S_IFREG
|S_IWUSR
|S_IRUGO
),
259 #ifdef CONFIG_SECURITY
260 static struct pid_entry tgid_attr_stuff
[] = {
261 E(PROC_TGID_ATTR_CURRENT
, "current", S_IFREG
|S_IRUGO
|S_IWUGO
),
262 E(PROC_TGID_ATTR_PREV
, "prev", S_IFREG
|S_IRUGO
),
263 E(PROC_TGID_ATTR_EXEC
, "exec", S_IFREG
|S_IRUGO
|S_IWUGO
),
264 E(PROC_TGID_ATTR_FSCREATE
, "fscreate", S_IFREG
|S_IRUGO
|S_IWUGO
),
267 static struct pid_entry tid_attr_stuff
[] = {
268 E(PROC_TID_ATTR_CURRENT
, "current", S_IFREG
|S_IRUGO
|S_IWUGO
),
269 E(PROC_TID_ATTR_PREV
, "prev", S_IFREG
|S_IRUGO
),
270 E(PROC_TID_ATTR_EXEC
, "exec", S_IFREG
|S_IRUGO
|S_IWUGO
),
271 E(PROC_TID_ATTR_FSCREATE
, "fscreate", S_IFREG
|S_IRUGO
|S_IWUGO
),
278 static int proc_fd_link(struct inode
*inode
, struct dentry
**dentry
, struct vfsmount
**mnt
)
280 struct task_struct
*task
= proc_task(inode
);
281 struct files_struct
*files
;
283 int fd
= proc_type(inode
) - PROC_TID_FD_DIR
;
285 files
= get_files_struct(task
);
288 file
= fcheck_files(files
, fd
);
290 *mnt
= mntget(file
->f_vfsmnt
);
291 *dentry
= dget(file
->f_dentry
);
293 put_files_struct(files
);
297 put_files_struct(files
);
302 static struct fs_struct
*get_fs_struct(struct task_struct
*task
)
304 struct fs_struct
*fs
;
308 atomic_inc(&fs
->count
);
313 static int proc_cwd_link(struct inode
*inode
, struct dentry
**dentry
, struct vfsmount
**mnt
)
315 struct fs_struct
*fs
= get_fs_struct(proc_task(inode
));
316 int result
= -ENOENT
;
318 read_lock(&fs
->lock
);
319 *mnt
= mntget(fs
->pwdmnt
);
320 *dentry
= dget(fs
->pwd
);
321 read_unlock(&fs
->lock
);
328 static int proc_root_link(struct inode
*inode
, struct dentry
**dentry
, struct vfsmount
**mnt
)
330 struct fs_struct
*fs
= get_fs_struct(proc_task(inode
));
331 int result
= -ENOENT
;
333 read_lock(&fs
->lock
);
334 *mnt
= mntget(fs
->rootmnt
);
335 *dentry
= dget(fs
->root
);
336 read_unlock(&fs
->lock
);
344 /* Same as proc_root_link, but this addionally tries to get fs from other
345 * threads in the group */
346 static int proc_task_root_link(struct inode
*inode
, struct dentry
**dentry
,
347 struct vfsmount
**mnt
)
349 struct fs_struct
*fs
;
350 int result
= -ENOENT
;
351 struct task_struct
*leader
= proc_task(inode
);
356 atomic_inc(&fs
->count
);
359 /* Try to get fs from other threads */
361 read_lock(&tasklist_lock
);
362 if (pid_alive(leader
)) {
363 struct task_struct
*task
= leader
;
365 while ((task
= next_thread(task
)) != leader
) {
369 atomic_inc(&fs
->count
);
376 read_unlock(&tasklist_lock
);
380 read_lock(&fs
->lock
);
381 *mnt
= mntget(fs
->rootmnt
);
382 *dentry
= dget(fs
->root
);
383 read_unlock(&fs
->lock
);
391 #define MAY_PTRACE(task) \
392 (task == current || \
393 (task->parent == current && \
394 (task->ptrace & PT_PTRACED) && \
395 (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
396 security_ptrace(current,task) == 0))
398 static int proc_pid_environ(struct task_struct
*task
, char * buffer
)
401 struct mm_struct
*mm
= get_task_mm(task
);
403 unsigned int len
= mm
->env_end
- mm
->env_start
;
406 res
= access_process_vm(task
, mm
->env_start
, buffer
, len
, 0);
407 if (!ptrace_may_attach(task
))
414 static int proc_pid_cmdline(struct task_struct
*task
, char * buffer
)
418 struct mm_struct
*mm
= get_task_mm(task
);
422 goto out_mm
; /* Shh! No looking before we're done */
424 len
= mm
->arg_end
- mm
->arg_start
;
429 res
= access_process_vm(task
, mm
->arg_start
, buffer
, len
, 0);
431 // If the nul at the end of args has been overwritten, then
432 // assume application is using setproctitle(3).
433 if (res
> 0 && buffer
[res
-1] != '\0' && len
< PAGE_SIZE
) {
434 len
= strnlen(buffer
, res
);
438 len
= mm
->env_end
- mm
->env_start
;
439 if (len
> PAGE_SIZE
- res
)
440 len
= PAGE_SIZE
- res
;
441 res
+= access_process_vm(task
, mm
->env_start
, buffer
+res
, len
, 0);
442 res
= strnlen(buffer
, res
);
451 static int proc_pid_auxv(struct task_struct
*task
, char *buffer
)
454 struct mm_struct
*mm
= get_task_mm(task
);
456 unsigned int nwords
= 0;
459 while (mm
->saved_auxv
[nwords
- 2] != 0); /* AT_NULL */
460 res
= nwords
* sizeof(mm
->saved_auxv
[0]);
463 memcpy(buffer
, mm
->saved_auxv
, res
);
470 #ifdef CONFIG_KALLSYMS
472 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
473 * Returns the resolved symbol. If that fails, simply return the address.
475 static int proc_pid_wchan(struct task_struct
*task
, char *buffer
)
478 const char *sym_name
;
479 unsigned long wchan
, size
, offset
;
480 char namebuf
[KSYM_NAME_LEN
+1];
482 wchan
= get_wchan(task
);
484 sym_name
= kallsyms_lookup(wchan
, &size
, &offset
, &modname
, namebuf
);
486 return sprintf(buffer
, "%s", sym_name
);
487 return sprintf(buffer
, "%lu", wchan
);
489 #endif /* CONFIG_KALLSYMS */
491 #ifdef CONFIG_SCHEDSTATS
493 * Provides /proc/PID/schedstat
495 static int proc_pid_schedstat(struct task_struct
*task
, char *buffer
)
497 return sprintf(buffer
, "%lu %lu %lu\n",
498 task
->sched_info
.cpu_time
,
499 task
->sched_info
.run_delay
,
500 task
->sched_info
.pcnt
);
504 /* The badness from the OOM killer */
505 unsigned long badness(struct task_struct
*p
, unsigned long uptime
);
506 static int proc_oom_score(struct task_struct
*task
, char *buffer
)
508 unsigned long points
;
509 struct timespec uptime
;
511 do_posix_clock_monotonic_gettime(&uptime
);
512 points
= badness(task
, uptime
.tv_sec
);
513 return sprintf(buffer
, "%lu\n", points
);
516 /************************************************************************/
517 /* Here the fs part begins */
518 /************************************************************************/
520 /* permission checks */
522 /* If the process being read is separated by chroot from the reading process,
523 * don't let the reader access the threads.
525 static int proc_check_chroot(struct dentry
*root
, struct vfsmount
*vfsmnt
)
527 struct dentry
*de
, *base
;
528 struct vfsmount
*our_vfsmnt
, *mnt
;
530 read_lock(¤t
->fs
->lock
);
531 our_vfsmnt
= mntget(current
->fs
->rootmnt
);
532 base
= dget(current
->fs
->root
);
533 read_unlock(¤t
->fs
->lock
);
535 spin_lock(&vfsmount_lock
);
539 while (vfsmnt
!= our_vfsmnt
) {
540 if (vfsmnt
== vfsmnt
->mnt_parent
)
542 de
= vfsmnt
->mnt_mountpoint
;
543 vfsmnt
= vfsmnt
->mnt_parent
;
546 if (!is_subdir(de
, base
))
548 spin_unlock(&vfsmount_lock
);
557 spin_unlock(&vfsmount_lock
);
562 static int proc_check_root(struct inode
*inode
)
565 struct vfsmount
*vfsmnt
;
567 if (proc_root_link(inode
, &root
, &vfsmnt
)) /* Ewww... */
569 return proc_check_chroot(root
, vfsmnt
);
572 static int proc_permission(struct inode
*inode
, int mask
, struct nameidata
*nd
)
574 if (generic_permission(inode
, mask
, NULL
) != 0)
576 return proc_check_root(inode
);
579 static int proc_task_permission(struct inode
*inode
, int mask
, struct nameidata
*nd
)
582 struct vfsmount
*vfsmnt
;
584 if (generic_permission(inode
, mask
, NULL
) != 0)
587 if (proc_task_root_link(inode
, &root
, &vfsmnt
))
590 return proc_check_chroot(root
, vfsmnt
);
593 extern struct seq_operations proc_pid_maps_op
;
594 static int maps_open(struct inode
*inode
, struct file
*file
)
596 struct task_struct
*task
= proc_task(inode
);
597 int ret
= seq_open(file
, &proc_pid_maps_op
);
599 struct seq_file
*m
= file
->private_data
;
605 static struct file_operations proc_maps_operations
= {
609 .release
= seq_release
,
613 extern struct seq_operations proc_pid_numa_maps_op
;
614 static int numa_maps_open(struct inode
*inode
, struct file
*file
)
616 struct task_struct
*task
= proc_task(inode
);
617 int ret
= seq_open(file
, &proc_pid_numa_maps_op
);
619 struct seq_file
*m
= file
->private_data
;
625 static struct file_operations proc_numa_maps_operations
= {
626 .open
= numa_maps_open
,
629 .release
= seq_release
,
633 extern struct seq_operations proc_pid_smaps_op
;
634 static int smaps_open(struct inode
*inode
, struct file
*file
)
636 struct task_struct
*task
= proc_task(inode
);
637 int ret
= seq_open(file
, &proc_pid_smaps_op
);
639 struct seq_file
*m
= file
->private_data
;
645 static struct file_operations proc_smaps_operations
= {
649 .release
= seq_release
,
652 extern struct seq_operations mounts_op
;
653 static int mounts_open(struct inode
*inode
, struct file
*file
)
655 struct task_struct
*task
= proc_task(inode
);
656 int ret
= seq_open(file
, &mounts_op
);
659 struct seq_file
*m
= file
->private_data
;
660 struct namespace *namespace;
662 namespace = task
->namespace;
664 get_namespace(namespace);
668 m
->private = namespace;
670 seq_release(inode
, file
);
677 static int mounts_release(struct inode
*inode
, struct file
*file
)
679 struct seq_file
*m
= file
->private_data
;
680 struct namespace *namespace = m
->private;
681 put_namespace(namespace);
682 return seq_release(inode
, file
);
685 static struct file_operations proc_mounts_operations
= {
689 .release
= mounts_release
,
692 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
694 static ssize_t
proc_info_read(struct file
* file
, char __user
* buf
,
695 size_t count
, loff_t
*ppos
)
697 struct inode
* inode
= file
->f_dentry
->d_inode
;
700 struct task_struct
*task
= proc_task(inode
);
702 if (count
> PROC_BLOCK_SIZE
)
703 count
= PROC_BLOCK_SIZE
;
704 if (!(page
= __get_free_page(GFP_KERNEL
)))
707 length
= PROC_I(inode
)->op
.proc_read(task
, (char*)page
);
710 length
= simple_read_from_buffer(buf
, count
, ppos
, (char *)page
, length
);
715 static struct file_operations proc_info_file_operations
= {
716 .read
= proc_info_read
,
719 static int mem_open(struct inode
* inode
, struct file
* file
)
721 file
->private_data
= (void*)((long)current
->self_exec_id
);
725 static ssize_t
mem_read(struct file
* file
, char __user
* buf
,
726 size_t count
, loff_t
*ppos
)
728 struct task_struct
*task
= proc_task(file
->f_dentry
->d_inode
);
730 unsigned long src
= *ppos
;
732 struct mm_struct
*mm
;
734 if (!MAY_PTRACE(task
) || !ptrace_may_attach(task
))
738 page
= (char *)__get_free_page(GFP_USER
);
744 mm
= get_task_mm(task
);
750 if (file
->private_data
!= (void*)((long)current
->self_exec_id
))
756 int this_len
, retval
;
758 this_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
759 retval
= access_process_vm(task
, src
, page
, this_len
, 0);
760 if (!retval
|| !MAY_PTRACE(task
) || !ptrace_may_attach(task
)) {
766 if (copy_to_user(buf
, page
, retval
)) {
781 free_page((unsigned long) page
);
786 #define mem_write NULL
789 /* This is a security hazard */
790 static ssize_t
mem_write(struct file
* file
, const char * buf
,
791 size_t count
, loff_t
*ppos
)
795 struct task_struct
*task
= proc_task(file
->f_dentry
->d_inode
);
796 unsigned long dst
= *ppos
;
798 if (!MAY_PTRACE(task
) || !ptrace_may_attach(task
))
801 page
= (char *)__get_free_page(GFP_USER
);
806 int this_len
, retval
;
808 this_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
809 if (copy_from_user(page
, buf
, this_len
)) {
813 retval
= access_process_vm(task
, dst
, page
, this_len
, 1);
825 free_page((unsigned long) page
);
830 static loff_t
mem_lseek(struct file
* file
, loff_t offset
, int orig
)
834 file
->f_pos
= offset
;
837 file
->f_pos
+= offset
;
842 force_successful_syscall_return();
846 static struct file_operations proc_mem_operations
= {
853 static ssize_t
oom_adjust_read(struct file
*file
, char __user
*buf
,
854 size_t count
, loff_t
*ppos
)
856 struct task_struct
*task
= proc_task(file
->f_dentry
->d_inode
);
859 int oom_adjust
= task
->oomkilladj
;
860 loff_t __ppos
= *ppos
;
862 len
= sprintf(buffer
, "%i\n", oom_adjust
);
865 if (count
> len
-__ppos
)
867 if (copy_to_user(buf
, buffer
+ __ppos
, count
))
869 *ppos
= __ppos
+ count
;
873 static ssize_t
oom_adjust_write(struct file
*file
, const char __user
*buf
,
874 size_t count
, loff_t
*ppos
)
876 struct task_struct
*task
= proc_task(file
->f_dentry
->d_inode
);
877 char buffer
[8], *end
;
880 if (!capable(CAP_SYS_RESOURCE
))
882 memset(buffer
, 0, 8);
885 if (copy_from_user(buffer
, buf
, count
))
887 oom_adjust
= simple_strtol(buffer
, &end
, 0);
888 if ((oom_adjust
< -16 || oom_adjust
> 15) && oom_adjust
!= OOM_DISABLE
)
892 task
->oomkilladj
= oom_adjust
;
893 if (end
- buffer
== 0)
898 static struct file_operations proc_oom_adjust_operations
= {
899 .read
= oom_adjust_read
,
900 .write
= oom_adjust_write
,
903 static struct inode_operations proc_mem_inode_operations
= {
904 .permission
= proc_permission
,
907 #ifdef CONFIG_AUDITSYSCALL
909 static ssize_t
proc_loginuid_read(struct file
* file
, char __user
* buf
,
910 size_t count
, loff_t
*ppos
)
912 struct inode
* inode
= file
->f_dentry
->d_inode
;
913 struct task_struct
*task
= proc_task(inode
);
915 char tmpbuf
[TMPBUFLEN
];
917 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
918 audit_get_loginuid(task
->audit_context
));
919 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
922 static ssize_t
proc_loginuid_write(struct file
* file
, const char __user
* buf
,
923 size_t count
, loff_t
*ppos
)
925 struct inode
* inode
= file
->f_dentry
->d_inode
;
928 struct task_struct
*task
= proc_task(inode
);
931 if (!capable(CAP_AUDIT_CONTROL
))
937 if (count
> PAGE_SIZE
)
941 /* No partial writes. */
944 page
= (char*)__get_free_page(GFP_USER
);
948 if (copy_from_user(page
, buf
, count
))
951 loginuid
= simple_strtoul(page
, &tmp
, 10);
957 length
= audit_set_loginuid(task
, loginuid
);
958 if (likely(length
== 0))
962 free_page((unsigned long) page
);
966 static struct file_operations proc_loginuid_operations
= {
967 .read
= proc_loginuid_read
,
968 .write
= proc_loginuid_write
,
972 #ifdef CONFIG_SECCOMP
973 static ssize_t
seccomp_read(struct file
*file
, char __user
*buf
,
974 size_t count
, loff_t
*ppos
)
976 struct task_struct
*tsk
= proc_task(file
->f_dentry
->d_inode
);
978 loff_t __ppos
= *ppos
;
981 /* no need to print the trailing zero, so use only len */
982 len
= sprintf(__buf
, "%u\n", tsk
->seccomp
.mode
);
985 if (count
> len
- __ppos
)
986 count
= len
- __ppos
;
987 if (copy_to_user(buf
, __buf
+ __ppos
, count
))
989 *ppos
= __ppos
+ count
;
993 static ssize_t
seccomp_write(struct file
*file
, const char __user
*buf
,
994 size_t count
, loff_t
*ppos
)
996 struct task_struct
*tsk
= proc_task(file
->f_dentry
->d_inode
);
997 char __buf
[20], *end
;
998 unsigned int seccomp_mode
;
1000 /* can set it only once to be even more secure */
1001 if (unlikely(tsk
->seccomp
.mode
))
1004 memset(__buf
, 0, sizeof(__buf
));
1005 count
= min(count
, sizeof(__buf
) - 1);
1006 if (copy_from_user(__buf
, buf
, count
))
1008 seccomp_mode
= simple_strtoul(__buf
, &end
, 0);
1011 if (seccomp_mode
&& seccomp_mode
<= NR_SECCOMP_MODES
) {
1012 tsk
->seccomp
.mode
= seccomp_mode
;
1013 set_tsk_thread_flag(tsk
, TIF_SECCOMP
);
1016 if (unlikely(!(end
- __buf
)))
1021 static struct file_operations proc_seccomp_operations
= {
1022 .read
= seccomp_read
,
1023 .write
= seccomp_write
,
1025 #endif /* CONFIG_SECCOMP */
1027 static void *proc_pid_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
1029 struct inode
*inode
= dentry
->d_inode
;
1030 int error
= -EACCES
;
1032 /* We don't need a base pointer in the /proc filesystem */
1035 if (current
->fsuid
!= inode
->i_uid
&& !capable(CAP_DAC_OVERRIDE
))
1037 error
= proc_check_root(inode
);
1041 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &nd
->dentry
, &nd
->mnt
);
1042 nd
->last_type
= LAST_BIND
;
1044 return ERR_PTR(error
);
1047 static int do_proc_readlink(struct dentry
*dentry
, struct vfsmount
*mnt
,
1048 char __user
*buffer
, int buflen
)
1050 struct inode
* inode
;
1051 char *tmp
= (char*)__get_free_page(GFP_KERNEL
), *path
;
1057 inode
= dentry
->d_inode
;
1058 path
= d_path(dentry
, mnt
, tmp
, PAGE_SIZE
);
1059 len
= PTR_ERR(path
);
1062 len
= tmp
+ PAGE_SIZE
- 1 - path
;
1066 if (copy_to_user(buffer
, path
, len
))
1069 free_page((unsigned long)tmp
);
1073 static int proc_pid_readlink(struct dentry
* dentry
, char __user
* buffer
, int buflen
)
1075 int error
= -EACCES
;
1076 struct inode
*inode
= dentry
->d_inode
;
1078 struct vfsmount
*mnt
= NULL
;
1082 if (current
->fsuid
!= inode
->i_uid
&& !capable(CAP_DAC_OVERRIDE
))
1084 error
= proc_check_root(inode
);
1088 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &de
, &mnt
);
1092 error
= do_proc_readlink(de
, mnt
, buffer
, buflen
);
1100 static struct inode_operations proc_pid_link_inode_operations
= {
1101 .readlink
= proc_pid_readlink
,
1102 .follow_link
= proc_pid_follow_link
1107 static int proc_readfd(struct file
* filp
, void * dirent
, filldir_t filldir
)
1109 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1110 struct task_struct
*p
= proc_task(inode
);
1111 unsigned int fd
, tid
, ino
;
1114 struct files_struct
* files
;
1115 struct fdtable
*fdt
;
1126 if (filldir(dirent
, ".", 1, 0, inode
->i_ino
, DT_DIR
) < 0)
1130 ino
= fake_ino(tid
, PROC_TID_INO
);
1131 if (filldir(dirent
, "..", 2, 1, ino
, DT_DIR
) < 0)
1135 files
= get_files_struct(p
);
1139 fdt
= files_fdtable(files
);
1140 for (fd
= filp
->f_pos
-2;
1142 fd
++, filp
->f_pos
++) {
1145 if (!fcheck_files(files
, fd
))
1153 buf
[j
] = '0' + (i
% 10);
1157 ino
= fake_ino(tid
, PROC_TID_FD_DIR
+ fd
);
1158 if (filldir(dirent
, buf
+j
, NUMBUF
-j
, fd
+2, ino
, DT_LNK
) < 0) {
1165 put_files_struct(files
);
1171 static int proc_pident_readdir(struct file
*filp
,
1172 void *dirent
, filldir_t filldir
,
1173 struct pid_entry
*ents
, unsigned int nents
)
1177 struct dentry
*dentry
= filp
->f_dentry
;
1178 struct inode
*inode
= dentry
->d_inode
;
1179 struct pid_entry
*p
;
1184 if (!pid_alive(proc_task(inode
)))
1188 pid
= proc_task(inode
)->pid
;
1193 if (filldir(dirent
, ".", 1, i
, ino
, DT_DIR
) < 0)
1199 ino
= parent_ino(dentry
);
1200 if (filldir(dirent
, "..", 2, i
, ino
, DT_DIR
) < 0)
1213 if (filldir(dirent
, p
->name
, p
->len
, filp
->f_pos
,
1214 fake_ino(pid
, p
->type
), p
->mode
>> 12) < 0)
1226 static int proc_tgid_base_readdir(struct file
* filp
,
1227 void * dirent
, filldir_t filldir
)
1229 return proc_pident_readdir(filp
,dirent
,filldir
,
1230 tgid_base_stuff
,ARRAY_SIZE(tgid_base_stuff
));
1233 static int proc_tid_base_readdir(struct file
* filp
,
1234 void * dirent
, filldir_t filldir
)
1236 return proc_pident_readdir(filp
,dirent
,filldir
,
1237 tid_base_stuff
,ARRAY_SIZE(tid_base_stuff
));
1240 /* building an inode */
1242 static int task_dumpable(struct task_struct
*task
)
1245 struct mm_struct
*mm
;
1250 dumpable
= mm
->dumpable
;
1258 static struct inode
*proc_pid_make_inode(struct super_block
* sb
, struct task_struct
*task
, int ino
)
1260 struct inode
* inode
;
1261 struct proc_inode
*ei
;
1263 /* We need a new inode */
1265 inode
= new_inode(sb
);
1272 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1273 inode
->i_ino
= fake_ino(task
->pid
, ino
);
1275 if (!pid_alive(task
))
1279 * grab the reference to task.
1281 get_task_struct(task
);
1286 if (ino
== PROC_TGID_INO
|| ino
== PROC_TID_INO
|| task_dumpable(task
)) {
1287 inode
->i_uid
= task
->euid
;
1288 inode
->i_gid
= task
->egid
;
1290 security_task_to_inode(task
, inode
);
1304 * Exceptional case: normally we are not allowed to unhash a busy
1305 * directory. In this case, however, we can do it - no aliasing problems
1306 * due to the way we treat inodes.
1308 * Rewrite the inode's ownerships here because the owning task may have
1309 * performed a setuid(), etc.
1311 static int pid_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1313 struct inode
*inode
= dentry
->d_inode
;
1314 struct task_struct
*task
= proc_task(inode
);
1315 if (pid_alive(task
)) {
1316 if (proc_type(inode
) == PROC_TGID_INO
|| proc_type(inode
) == PROC_TID_INO
|| task_dumpable(task
)) {
1317 inode
->i_uid
= task
->euid
;
1318 inode
->i_gid
= task
->egid
;
1323 security_task_to_inode(task
, inode
);
1330 static int tid_fd_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1332 struct inode
*inode
= dentry
->d_inode
;
1333 struct task_struct
*task
= proc_task(inode
);
1334 int fd
= proc_type(inode
) - PROC_TID_FD_DIR
;
1335 struct files_struct
*files
;
1337 files
= get_files_struct(task
);
1340 if (fcheck_files(files
, fd
)) {
1342 put_files_struct(files
);
1343 if (task_dumpable(task
)) {
1344 inode
->i_uid
= task
->euid
;
1345 inode
->i_gid
= task
->egid
;
1350 security_task_to_inode(task
, inode
);
1354 put_files_struct(files
);
1360 static void pid_base_iput(struct dentry
*dentry
, struct inode
*inode
)
1362 struct task_struct
*task
= proc_task(inode
);
1363 spin_lock(&task
->proc_lock
);
1364 if (task
->proc_dentry
== dentry
)
1365 task
->proc_dentry
= NULL
;
1366 spin_unlock(&task
->proc_lock
);
1370 static int pid_delete_dentry(struct dentry
* dentry
)
1372 /* Is the task we represent dead?
1373 * If so, then don't put the dentry on the lru list,
1374 * kill it immediately.
1376 return !pid_alive(proc_task(dentry
->d_inode
));
1379 static struct dentry_operations tid_fd_dentry_operations
=
1381 .d_revalidate
= tid_fd_revalidate
,
1382 .d_delete
= pid_delete_dentry
,
1385 static struct dentry_operations pid_dentry_operations
=
1387 .d_revalidate
= pid_revalidate
,
1388 .d_delete
= pid_delete_dentry
,
1391 static struct dentry_operations pid_base_dentry_operations
=
1393 .d_revalidate
= pid_revalidate
,
1394 .d_iput
= pid_base_iput
,
1395 .d_delete
= pid_delete_dentry
,
1400 static unsigned name_to_int(struct dentry
*dentry
)
1402 const char *name
= dentry
->d_name
.name
;
1403 int len
= dentry
->d_name
.len
;
1406 if (len
> 1 && *name
== '0')
1409 unsigned c
= *name
++ - '0';
1412 if (n
>= (~0U-9)/10)
1423 static struct dentry
*proc_lookupfd(struct inode
* dir
, struct dentry
* dentry
, struct nameidata
*nd
)
1425 struct task_struct
*task
= proc_task(dir
);
1426 unsigned fd
= name_to_int(dentry
);
1428 struct files_struct
* files
;
1429 struct inode
*inode
;
1430 struct proc_inode
*ei
;
1434 if (!pid_alive(task
))
1437 inode
= proc_pid_make_inode(dir
->i_sb
, task
, PROC_TID_FD_DIR
+fd
);
1441 files
= get_files_struct(task
);
1444 inode
->i_mode
= S_IFLNK
;
1446 file
= fcheck_files(files
, fd
);
1449 if (file
->f_mode
& 1)
1450 inode
->i_mode
|= S_IRUSR
| S_IXUSR
;
1451 if (file
->f_mode
& 2)
1452 inode
->i_mode
|= S_IWUSR
| S_IXUSR
;
1454 put_files_struct(files
);
1455 inode
->i_op
= &proc_pid_link_inode_operations
;
1457 ei
->op
.proc_get_link
= proc_fd_link
;
1458 dentry
->d_op
= &tid_fd_dentry_operations
;
1459 d_add(dentry
, inode
);
1464 put_files_struct(files
);
1468 return ERR_PTR(-ENOENT
);
1471 static int proc_task_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
);
1472 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
);
1474 static struct file_operations proc_fd_operations
= {
1475 .read
= generic_read_dir
,
1476 .readdir
= proc_readfd
,
1479 static struct file_operations proc_task_operations
= {
1480 .read
= generic_read_dir
,
1481 .readdir
= proc_task_readdir
,
1485 * proc directories can do almost nothing..
1487 static struct inode_operations proc_fd_inode_operations
= {
1488 .lookup
= proc_lookupfd
,
1489 .permission
= proc_permission
,
1492 static struct inode_operations proc_task_inode_operations
= {
1493 .lookup
= proc_task_lookup
,
1494 .permission
= proc_task_permission
,
1497 #ifdef CONFIG_SECURITY
1498 static ssize_t
proc_pid_attr_read(struct file
* file
, char __user
* buf
,
1499 size_t count
, loff_t
*ppos
)
1501 struct inode
* inode
= file
->f_dentry
->d_inode
;
1504 struct task_struct
*task
= proc_task(inode
);
1506 if (count
> PAGE_SIZE
)
1508 if (!(page
= __get_free_page(GFP_KERNEL
)))
1511 length
= security_getprocattr(task
,
1512 (char*)file
->f_dentry
->d_name
.name
,
1513 (void*)page
, count
);
1515 length
= simple_read_from_buffer(buf
, count
, ppos
, (char *)page
, length
);
1520 static ssize_t
proc_pid_attr_write(struct file
* file
, const char __user
* buf
,
1521 size_t count
, loff_t
*ppos
)
1523 struct inode
* inode
= file
->f_dentry
->d_inode
;
1526 struct task_struct
*task
= proc_task(inode
);
1528 if (count
> PAGE_SIZE
)
1531 /* No partial writes. */
1534 page
= (char*)__get_free_page(GFP_USER
);
1538 if (copy_from_user(page
, buf
, count
))
1541 length
= security_setprocattr(task
,
1542 (char*)file
->f_dentry
->d_name
.name
,
1543 (void*)page
, count
);
1545 free_page((unsigned long) page
);
1549 static struct file_operations proc_pid_attr_operations
= {
1550 .read
= proc_pid_attr_read
,
1551 .write
= proc_pid_attr_write
,
1554 static struct file_operations proc_tid_attr_operations
;
1555 static struct inode_operations proc_tid_attr_inode_operations
;
1556 static struct file_operations proc_tgid_attr_operations
;
1557 static struct inode_operations proc_tgid_attr_inode_operations
;
1560 static int get_tid_list(int index
, unsigned int *tids
, struct inode
*dir
);
1563 static struct dentry
*proc_pident_lookup(struct inode
*dir
,
1564 struct dentry
*dentry
,
1565 struct pid_entry
*ents
)
1567 struct inode
*inode
;
1569 struct task_struct
*task
= proc_task(dir
);
1570 struct pid_entry
*p
;
1571 struct proc_inode
*ei
;
1576 if (!pid_alive(task
))
1579 for (p
= ents
; p
->name
; p
++) {
1580 if (p
->len
!= dentry
->d_name
.len
)
1582 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
1589 inode
= proc_pid_make_inode(dir
->i_sb
, task
, p
->type
);
1594 inode
->i_mode
= p
->mode
;
1596 * Yes, it does not scale. And it should not. Don't add
1597 * new entries into /proc/<tgid>/ without very good reasons.
1600 case PROC_TGID_TASK
:
1601 inode
->i_nlink
= 2 + get_tid_list(2, NULL
, dir
);
1602 inode
->i_op
= &proc_task_inode_operations
;
1603 inode
->i_fop
= &proc_task_operations
;
1608 inode
->i_op
= &proc_fd_inode_operations
;
1609 inode
->i_fop
= &proc_fd_operations
;
1613 inode
->i_op
= &proc_pid_link_inode_operations
;
1614 ei
->op
.proc_get_link
= proc_exe_link
;
1618 inode
->i_op
= &proc_pid_link_inode_operations
;
1619 ei
->op
.proc_get_link
= proc_cwd_link
;
1622 case PROC_TGID_ROOT
:
1623 inode
->i_op
= &proc_pid_link_inode_operations
;
1624 ei
->op
.proc_get_link
= proc_root_link
;
1626 case PROC_TID_ENVIRON
:
1627 case PROC_TGID_ENVIRON
:
1628 inode
->i_fop
= &proc_info_file_operations
;
1629 ei
->op
.proc_read
= proc_pid_environ
;
1632 case PROC_TGID_AUXV
:
1633 inode
->i_fop
= &proc_info_file_operations
;
1634 ei
->op
.proc_read
= proc_pid_auxv
;
1636 case PROC_TID_STATUS
:
1637 case PROC_TGID_STATUS
:
1638 inode
->i_fop
= &proc_info_file_operations
;
1639 ei
->op
.proc_read
= proc_pid_status
;
1642 inode
->i_fop
= &proc_info_file_operations
;
1643 ei
->op
.proc_read
= proc_tid_stat
;
1645 case PROC_TGID_STAT
:
1646 inode
->i_fop
= &proc_info_file_operations
;
1647 ei
->op
.proc_read
= proc_tgid_stat
;
1649 case PROC_TID_CMDLINE
:
1650 case PROC_TGID_CMDLINE
:
1651 inode
->i_fop
= &proc_info_file_operations
;
1652 ei
->op
.proc_read
= proc_pid_cmdline
;
1654 case PROC_TID_STATM
:
1655 case PROC_TGID_STATM
:
1656 inode
->i_fop
= &proc_info_file_operations
;
1657 ei
->op
.proc_read
= proc_pid_statm
;
1660 case PROC_TGID_MAPS
:
1661 inode
->i_fop
= &proc_maps_operations
;
1664 case PROC_TID_NUMA_MAPS
:
1665 case PROC_TGID_NUMA_MAPS
:
1666 inode
->i_fop
= &proc_numa_maps_operations
;
1671 inode
->i_op
= &proc_mem_inode_operations
;
1672 inode
->i_fop
= &proc_mem_operations
;
1674 #ifdef CONFIG_SECCOMP
1675 case PROC_TID_SECCOMP
:
1676 case PROC_TGID_SECCOMP
:
1677 inode
->i_fop
= &proc_seccomp_operations
;
1679 #endif /* CONFIG_SECCOMP */
1680 case PROC_TID_MOUNTS
:
1681 case PROC_TGID_MOUNTS
:
1682 inode
->i_fop
= &proc_mounts_operations
;
1684 case PROC_TID_SMAPS
:
1685 case PROC_TGID_SMAPS
:
1686 inode
->i_fop
= &proc_smaps_operations
;
1688 #ifdef CONFIG_SECURITY
1691 inode
->i_op
= &proc_tid_attr_inode_operations
;
1692 inode
->i_fop
= &proc_tid_attr_operations
;
1694 case PROC_TGID_ATTR
:
1696 inode
->i_op
= &proc_tgid_attr_inode_operations
;
1697 inode
->i_fop
= &proc_tgid_attr_operations
;
1699 case PROC_TID_ATTR_CURRENT
:
1700 case PROC_TGID_ATTR_CURRENT
:
1701 case PROC_TID_ATTR_PREV
:
1702 case PROC_TGID_ATTR_PREV
:
1703 case PROC_TID_ATTR_EXEC
:
1704 case PROC_TGID_ATTR_EXEC
:
1705 case PROC_TID_ATTR_FSCREATE
:
1706 case PROC_TGID_ATTR_FSCREATE
:
1707 inode
->i_fop
= &proc_pid_attr_operations
;
1710 #ifdef CONFIG_KALLSYMS
1711 case PROC_TID_WCHAN
:
1712 case PROC_TGID_WCHAN
:
1713 inode
->i_fop
= &proc_info_file_operations
;
1714 ei
->op
.proc_read
= proc_pid_wchan
;
1717 #ifdef CONFIG_SCHEDSTATS
1718 case PROC_TID_SCHEDSTAT
:
1719 case PROC_TGID_SCHEDSTAT
:
1720 inode
->i_fop
= &proc_info_file_operations
;
1721 ei
->op
.proc_read
= proc_pid_schedstat
;
1724 #ifdef CONFIG_CPUSETS
1725 case PROC_TID_CPUSET
:
1726 case PROC_TGID_CPUSET
:
1727 inode
->i_fop
= &proc_cpuset_operations
;
1730 case PROC_TID_OOM_SCORE
:
1731 case PROC_TGID_OOM_SCORE
:
1732 inode
->i_fop
= &proc_info_file_operations
;
1733 ei
->op
.proc_read
= proc_oom_score
;
1735 case PROC_TID_OOM_ADJUST
:
1736 case PROC_TGID_OOM_ADJUST
:
1737 inode
->i_fop
= &proc_oom_adjust_operations
;
1739 #ifdef CONFIG_AUDITSYSCALL
1740 case PROC_TID_LOGINUID
:
1741 case PROC_TGID_LOGINUID
:
1742 inode
->i_fop
= &proc_loginuid_operations
;
1746 printk("procfs: impossible type (%d)",p
->type
);
1748 return ERR_PTR(-EINVAL
);
1750 dentry
->d_op
= &pid_dentry_operations
;
1751 d_add(dentry
, inode
);
1755 return ERR_PTR(error
);
1758 static struct dentry
*proc_tgid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
1759 return proc_pident_lookup(dir
, dentry
, tgid_base_stuff
);
1762 static struct dentry
*proc_tid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
1763 return proc_pident_lookup(dir
, dentry
, tid_base_stuff
);
1766 static struct file_operations proc_tgid_base_operations
= {
1767 .read
= generic_read_dir
,
1768 .readdir
= proc_tgid_base_readdir
,
1771 static struct file_operations proc_tid_base_operations
= {
1772 .read
= generic_read_dir
,
1773 .readdir
= proc_tid_base_readdir
,
1776 static struct inode_operations proc_tgid_base_inode_operations
= {
1777 .lookup
= proc_tgid_base_lookup
,
1780 static struct inode_operations proc_tid_base_inode_operations
= {
1781 .lookup
= proc_tid_base_lookup
,
1784 #ifdef CONFIG_SECURITY
1785 static int proc_tgid_attr_readdir(struct file
* filp
,
1786 void * dirent
, filldir_t filldir
)
1788 return proc_pident_readdir(filp
,dirent
,filldir
,
1789 tgid_attr_stuff
,ARRAY_SIZE(tgid_attr_stuff
));
1792 static int proc_tid_attr_readdir(struct file
* filp
,
1793 void * dirent
, filldir_t filldir
)
1795 return proc_pident_readdir(filp
,dirent
,filldir
,
1796 tid_attr_stuff
,ARRAY_SIZE(tid_attr_stuff
));
1799 static struct file_operations proc_tgid_attr_operations
= {
1800 .read
= generic_read_dir
,
1801 .readdir
= proc_tgid_attr_readdir
,
1804 static struct file_operations proc_tid_attr_operations
= {
1805 .read
= generic_read_dir
,
1806 .readdir
= proc_tid_attr_readdir
,
1809 static struct dentry
*proc_tgid_attr_lookup(struct inode
*dir
,
1810 struct dentry
*dentry
, struct nameidata
*nd
)
1812 return proc_pident_lookup(dir
, dentry
, tgid_attr_stuff
);
1815 static struct dentry
*proc_tid_attr_lookup(struct inode
*dir
,
1816 struct dentry
*dentry
, struct nameidata
*nd
)
1818 return proc_pident_lookup(dir
, dentry
, tid_attr_stuff
);
1821 static struct inode_operations proc_tgid_attr_inode_operations
= {
1822 .lookup
= proc_tgid_attr_lookup
,
1825 static struct inode_operations proc_tid_attr_inode_operations
= {
1826 .lookup
= proc_tid_attr_lookup
,
1833 static int proc_self_readlink(struct dentry
*dentry
, char __user
*buffer
,
1837 sprintf(tmp
, "%d", current
->tgid
);
1838 return vfs_readlink(dentry
,buffer
,buflen
,tmp
);
1841 static void *proc_self_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
1844 sprintf(tmp
, "%d", current
->tgid
);
1845 return ERR_PTR(vfs_follow_link(nd
,tmp
));
1848 static struct inode_operations proc_self_inode_operations
= {
1849 .readlink
= proc_self_readlink
,
1850 .follow_link
= proc_self_follow_link
,
1854 * proc_pid_unhash - Unhash /proc/@pid entry from the dcache.
1855 * @p: task that should be flushed.
1857 * Drops the /proc/@pid dcache entry from the hash chains.
1859 * Dropping /proc/@pid entries and detach_pid must be synchroneous,
1860 * otherwise e.g. /proc/@pid/exe might point to the wrong executable,
1861 * if the pid value is immediately reused. This is enforced by
1862 * - caller must acquire spin_lock(p->proc_lock)
1863 * - must be called before detach_pid()
1864 * - proc_pid_lookup acquires proc_lock, and checks that
1865 * the target is not dead by looking at the attach count
1869 struct dentry
*proc_pid_unhash(struct task_struct
*p
)
1871 struct dentry
*proc_dentry
;
1873 proc_dentry
= p
->proc_dentry
;
1874 if (proc_dentry
!= NULL
) {
1876 spin_lock(&dcache_lock
);
1877 spin_lock(&proc_dentry
->d_lock
);
1878 if (!d_unhashed(proc_dentry
)) {
1879 dget_locked(proc_dentry
);
1880 __d_drop(proc_dentry
);
1881 spin_unlock(&proc_dentry
->d_lock
);
1883 spin_unlock(&proc_dentry
->d_lock
);
1886 spin_unlock(&dcache_lock
);
1892 * proc_pid_flush - recover memory used by stale /proc/@pid/x entries
1893 * @proc_dentry: directoy to prune.
1895 * Shrink the /proc directory that was used by the just killed thread.
1898 void proc_pid_flush(struct dentry
*proc_dentry
)
1901 if(proc_dentry
!= NULL
) {
1902 shrink_dcache_parent(proc_dentry
);
1908 struct dentry
*proc_pid_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
1910 struct task_struct
*task
;
1911 struct inode
*inode
;
1912 struct proc_inode
*ei
;
1916 if (dentry
->d_name
.len
== 4 && !memcmp(dentry
->d_name
.name
,"self",4)) {
1917 inode
= new_inode(dir
->i_sb
);
1919 return ERR_PTR(-ENOMEM
);
1921 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1922 inode
->i_ino
= fake_ino(0, PROC_TGID_INO
);
1924 inode
->i_mode
= S_IFLNK
|S_IRWXUGO
;
1925 inode
->i_uid
= inode
->i_gid
= 0;
1927 inode
->i_op
= &proc_self_inode_operations
;
1928 d_add(dentry
, inode
);
1931 tgid
= name_to_int(dentry
);
1935 read_lock(&tasklist_lock
);
1936 task
= find_task_by_pid(tgid
);
1938 get_task_struct(task
);
1939 read_unlock(&tasklist_lock
);
1943 inode
= proc_pid_make_inode(dir
->i_sb
, task
, PROC_TGID_INO
);
1947 put_task_struct(task
);
1950 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
1951 inode
->i_op
= &proc_tgid_base_inode_operations
;
1952 inode
->i_fop
= &proc_tgid_base_operations
;
1953 inode
->i_flags
|=S_IMMUTABLE
;
1954 #ifdef CONFIG_SECURITY
1960 dentry
->d_op
= &pid_base_dentry_operations
;
1963 d_add(dentry
, inode
);
1964 spin_lock(&task
->proc_lock
);
1965 task
->proc_dentry
= dentry
;
1966 if (!pid_alive(task
)) {
1967 dentry
= proc_pid_unhash(task
);
1970 spin_unlock(&task
->proc_lock
);
1972 put_task_struct(task
);
1974 proc_pid_flush(dentry
);
1979 return ERR_PTR(-ENOENT
);
1983 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
1985 struct task_struct
*task
;
1986 struct task_struct
*leader
= proc_task(dir
);
1987 struct inode
*inode
;
1990 tid
= name_to_int(dentry
);
1994 read_lock(&tasklist_lock
);
1995 task
= find_task_by_pid(tid
);
1997 get_task_struct(task
);
1998 read_unlock(&tasklist_lock
);
2001 if (leader
->tgid
!= task
->tgid
)
2004 inode
= proc_pid_make_inode(dir
->i_sb
, task
, PROC_TID_INO
);
2009 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
2010 inode
->i_op
= &proc_tid_base_inode_operations
;
2011 inode
->i_fop
= &proc_tid_base_operations
;
2012 inode
->i_flags
|=S_IMMUTABLE
;
2013 #ifdef CONFIG_SECURITY
2019 dentry
->d_op
= &pid_base_dentry_operations
;
2021 d_add(dentry
, inode
);
2023 put_task_struct(task
);
2026 put_task_struct(task
);
2028 return ERR_PTR(-ENOENT
);
2031 #define PROC_NUMBUF 10
2032 #define PROC_MAXPIDS 20
2035 * Get a few tgid's to return for filldir - we need to hold the
2036 * tasklist lock while doing this, and we must release it before
2037 * we actually do the filldir itself, so we use a temp buffer..
2039 static int get_tgid_list(int index
, unsigned long version
, unsigned int *tgids
)
2041 struct task_struct
*p
;
2045 read_lock(&tasklist_lock
);
2048 p
= find_task_by_pid(version
);
2049 if (p
&& !thread_group_leader(p
))
2056 p
= next_task(&init_task
);
2058 for ( ; p
!= &init_task
; p
= next_task(p
)) {
2064 tgids
[nr_tgids
] = tgid
;
2066 if (nr_tgids
>= PROC_MAXPIDS
)
2069 read_unlock(&tasklist_lock
);
2074 * Get a few tid's to return for filldir - we need to hold the
2075 * tasklist lock while doing this, and we must release it before
2076 * we actually do the filldir itself, so we use a temp buffer..
2078 static int get_tid_list(int index
, unsigned int *tids
, struct inode
*dir
)
2080 struct task_struct
*leader_task
= proc_task(dir
);
2081 struct task_struct
*task
= leader_task
;
2085 read_lock(&tasklist_lock
);
2087 * The starting point task (leader_task) might be an already
2088 * unlinked task, which cannot be used to access the task-list
2089 * via next_thread().
2091 if (pid_alive(task
)) do {
2092 int tid
= task
->pid
;
2097 tids
[nr_tids
] = tid
;
2099 if (nr_tids
>= PROC_MAXPIDS
)
2101 } while ((task
= next_thread(task
)) != leader_task
);
2102 read_unlock(&tasklist_lock
);
2106 /* for the /proc/ directory itself, after non-process stuff has been done */
2107 int proc_pid_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
2109 unsigned int tgid_array
[PROC_MAXPIDS
];
2110 char buf
[PROC_NUMBUF
];
2111 unsigned int nr
= filp
->f_pos
- FIRST_PROCESS_ENTRY
;
2112 unsigned int nr_tgids
, i
;
2116 ino_t ino
= fake_ino(0,PROC_TGID_INO
);
2117 if (filldir(dirent
, "self", 4, filp
->f_pos
, ino
, DT_LNK
) < 0)
2123 /* f_version caches the tgid value that the last readdir call couldn't
2124 * return. lseek aka telldir automagically resets f_version to 0.
2126 next_tgid
= filp
->f_version
;
2127 filp
->f_version
= 0;
2129 nr_tgids
= get_tgid_list(nr
, next_tgid
, tgid_array
);
2131 /* no more entries ! */
2136 /* do not use the last found pid, reserve it for next_tgid */
2137 if (nr_tgids
== PROC_MAXPIDS
) {
2139 next_tgid
= tgid_array
[nr_tgids
];
2142 for (i
=0;i
<nr_tgids
;i
++) {
2143 int tgid
= tgid_array
[i
];
2144 ino_t ino
= fake_ino(tgid
,PROC_TGID_INO
);
2145 unsigned long j
= PROC_NUMBUF
;
2148 buf
[--j
] = '0' + (tgid
% 10);
2149 while ((tgid
/= 10) != 0);
2151 if (filldir(dirent
, buf
+j
, PROC_NUMBUF
-j
, filp
->f_pos
, ino
, DT_DIR
) < 0) {
2152 /* returning this tgid failed, save it as the first
2153 * pid for the next readir call */
2154 filp
->f_version
= tgid_array
[i
];
2165 /* for the /proc/TGID/task/ directories */
2166 static int proc_task_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
2168 unsigned int tid_array
[PROC_MAXPIDS
];
2169 char buf
[PROC_NUMBUF
];
2170 unsigned int nr_tids
, i
;
2171 struct dentry
*dentry
= filp
->f_dentry
;
2172 struct inode
*inode
= dentry
->d_inode
;
2173 int retval
= -ENOENT
;
2175 unsigned long pos
= filp
->f_pos
; /* avoiding "long long" filp->f_pos */
2177 if (!pid_alive(proc_task(inode
)))
2184 if (filldir(dirent
, ".", 1, pos
, ino
, DT_DIR
) < 0)
2189 ino
= parent_ino(dentry
);
2190 if (filldir(dirent
, "..", 2, pos
, ino
, DT_DIR
) < 0)
2196 nr_tids
= get_tid_list(pos
, tid_array
, inode
);
2197 inode
->i_nlink
= pos
+ nr_tids
;
2199 for (i
= 0; i
< nr_tids
; i
++) {
2200 unsigned long j
= PROC_NUMBUF
;
2201 int tid
= tid_array
[i
];
2203 ino
= fake_ino(tid
,PROC_TID_INO
);
2206 buf
[--j
] = '0' + (tid
% 10);
2207 while ((tid
/= 10) != 0);
2209 if (filldir(dirent
, buf
+j
, PROC_NUMBUF
-j
, pos
, ino
, DT_DIR
) < 0)