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
, struct vfsmount
**mnt
)
348 struct fs_struct
*fs
;
349 int result
= -ENOENT
;
350 struct task_struct
*leader
= proc_task(inode
);
355 atomic_inc(&fs
->count
);
358 /* Try to get fs from other threads */
360 struct task_struct
*task
= leader
;
361 read_lock(&tasklist_lock
);
362 if (pid_alive(task
)) {
363 while ((task
= next_thread(task
)) != leader
) {
367 atomic_inc(&fs
->count
);
374 read_unlock(&tasklist_lock
);
378 read_lock(&fs
->lock
);
379 *mnt
= mntget(fs
->rootmnt
);
380 *dentry
= dget(fs
->root
);
381 read_unlock(&fs
->lock
);
389 #define MAY_PTRACE(task) \
390 (task == current || \
391 (task->parent == current && \
392 (task->ptrace & PT_PTRACED) && \
393 (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
394 security_ptrace(current,task) == 0))
396 static int proc_pid_environ(struct task_struct
*task
, char * buffer
)
399 struct mm_struct
*mm
= get_task_mm(task
);
401 unsigned int len
= mm
->env_end
- mm
->env_start
;
404 res
= access_process_vm(task
, mm
->env_start
, buffer
, len
, 0);
405 if (!ptrace_may_attach(task
))
412 static int proc_pid_cmdline(struct task_struct
*task
, char * buffer
)
416 struct mm_struct
*mm
= get_task_mm(task
);
420 goto out_mm
; /* Shh! No looking before we're done */
422 len
= mm
->arg_end
- mm
->arg_start
;
427 res
= access_process_vm(task
, mm
->arg_start
, buffer
, len
, 0);
429 // If the nul at the end of args has been overwritten, then
430 // assume application is using setproctitle(3).
431 if (res
> 0 && buffer
[res
-1] != '\0' && len
< PAGE_SIZE
) {
432 len
= strnlen(buffer
, res
);
436 len
= mm
->env_end
- mm
->env_start
;
437 if (len
> PAGE_SIZE
- res
)
438 len
= PAGE_SIZE
- res
;
439 res
+= access_process_vm(task
, mm
->env_start
, buffer
+res
, len
, 0);
440 res
= strnlen(buffer
, res
);
449 static int proc_pid_auxv(struct task_struct
*task
, char *buffer
)
452 struct mm_struct
*mm
= get_task_mm(task
);
454 unsigned int nwords
= 0;
457 while (mm
->saved_auxv
[nwords
- 2] != 0); /* AT_NULL */
458 res
= nwords
* sizeof(mm
->saved_auxv
[0]);
461 memcpy(buffer
, mm
->saved_auxv
, res
);
468 #ifdef CONFIG_KALLSYMS
470 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
471 * Returns the resolved symbol. If that fails, simply return the address.
473 static int proc_pid_wchan(struct task_struct
*task
, char *buffer
)
476 const char *sym_name
;
477 unsigned long wchan
, size
, offset
;
478 char namebuf
[KSYM_NAME_LEN
+1];
480 wchan
= get_wchan(task
);
482 sym_name
= kallsyms_lookup(wchan
, &size
, &offset
, &modname
, namebuf
);
484 return sprintf(buffer
, "%s", sym_name
);
485 return sprintf(buffer
, "%lu", wchan
);
487 #endif /* CONFIG_KALLSYMS */
489 #ifdef CONFIG_SCHEDSTATS
491 * Provides /proc/PID/schedstat
493 static int proc_pid_schedstat(struct task_struct
*task
, char *buffer
)
495 return sprintf(buffer
, "%lu %lu %lu\n",
496 task
->sched_info
.cpu_time
,
497 task
->sched_info
.run_delay
,
498 task
->sched_info
.pcnt
);
502 /* The badness from the OOM killer */
503 unsigned long badness(struct task_struct
*p
, unsigned long uptime
);
504 static int proc_oom_score(struct task_struct
*task
, char *buffer
)
506 unsigned long points
;
507 struct timespec uptime
;
509 do_posix_clock_monotonic_gettime(&uptime
);
510 points
= badness(task
, uptime
.tv_sec
);
511 return sprintf(buffer
, "%lu\n", points
);
514 /************************************************************************/
515 /* Here the fs part begins */
516 /************************************************************************/
518 /* permission checks */
520 /* If the process being read is separated by chroot from the reading process,
521 * don't let the reader access the threads.
523 static int proc_check_chroot(struct dentry
*root
, struct vfsmount
*vfsmnt
)
525 struct dentry
*de
, *base
;
526 struct vfsmount
*our_vfsmnt
, *mnt
;
528 read_lock(¤t
->fs
->lock
);
529 our_vfsmnt
= mntget(current
->fs
->rootmnt
);
530 base
= dget(current
->fs
->root
);
531 read_unlock(¤t
->fs
->lock
);
533 spin_lock(&vfsmount_lock
);
537 while (vfsmnt
!= our_vfsmnt
) {
538 if (vfsmnt
== vfsmnt
->mnt_parent
)
540 de
= vfsmnt
->mnt_mountpoint
;
541 vfsmnt
= vfsmnt
->mnt_parent
;
544 if (!is_subdir(de
, base
))
546 spin_unlock(&vfsmount_lock
);
555 spin_unlock(&vfsmount_lock
);
560 static int proc_check_root(struct inode
*inode
)
563 struct vfsmount
*vfsmnt
;
565 if (proc_root_link(inode
, &root
, &vfsmnt
)) /* Ewww... */
567 return proc_check_chroot(root
, vfsmnt
);
570 static int proc_permission(struct inode
*inode
, int mask
, struct nameidata
*nd
)
572 if (generic_permission(inode
, mask
, NULL
) != 0)
574 return proc_check_root(inode
);
577 static int proc_task_permission(struct inode
*inode
, int mask
, struct nameidata
*nd
)
580 struct vfsmount
*vfsmnt
;
582 if (generic_permission(inode
, mask
, NULL
) != 0)
585 if (proc_task_root_link(inode
, &root
, &vfsmnt
))
588 return proc_check_chroot(root
, vfsmnt
);
591 extern struct seq_operations proc_pid_maps_op
;
592 static int maps_open(struct inode
*inode
, struct file
*file
)
594 struct task_struct
*task
= proc_task(inode
);
595 int ret
= seq_open(file
, &proc_pid_maps_op
);
597 struct seq_file
*m
= file
->private_data
;
603 static struct file_operations proc_maps_operations
= {
607 .release
= seq_release
,
611 extern struct seq_operations proc_pid_numa_maps_op
;
612 static int numa_maps_open(struct inode
*inode
, struct file
*file
)
614 struct task_struct
*task
= proc_task(inode
);
615 int ret
= seq_open(file
, &proc_pid_numa_maps_op
);
617 struct seq_file
*m
= file
->private_data
;
623 static struct file_operations proc_numa_maps_operations
= {
624 .open
= numa_maps_open
,
627 .release
= seq_release
,
631 extern struct seq_operations proc_pid_smaps_op
;
632 static int smaps_open(struct inode
*inode
, struct file
*file
)
634 struct task_struct
*task
= proc_task(inode
);
635 int ret
= seq_open(file
, &proc_pid_smaps_op
);
637 struct seq_file
*m
= file
->private_data
;
643 static struct file_operations proc_smaps_operations
= {
647 .release
= seq_release
,
650 extern struct seq_operations mounts_op
;
651 static int mounts_open(struct inode
*inode
, struct file
*file
)
653 struct task_struct
*task
= proc_task(inode
);
654 int ret
= seq_open(file
, &mounts_op
);
657 struct seq_file
*m
= file
->private_data
;
658 struct namespace *namespace;
660 namespace = task
->namespace;
662 get_namespace(namespace);
666 m
->private = namespace;
668 seq_release(inode
, file
);
675 static int mounts_release(struct inode
*inode
, struct file
*file
)
677 struct seq_file
*m
= file
->private_data
;
678 struct namespace *namespace = m
->private;
679 put_namespace(namespace);
680 return seq_release(inode
, file
);
683 static struct file_operations proc_mounts_operations
= {
687 .release
= mounts_release
,
690 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
692 static ssize_t
proc_info_read(struct file
* file
, char __user
* buf
,
693 size_t count
, loff_t
*ppos
)
695 struct inode
* inode
= file
->f_dentry
->d_inode
;
698 struct task_struct
*task
= proc_task(inode
);
700 if (count
> PROC_BLOCK_SIZE
)
701 count
= PROC_BLOCK_SIZE
;
702 if (!(page
= __get_free_page(GFP_KERNEL
)))
705 length
= PROC_I(inode
)->op
.proc_read(task
, (char*)page
);
708 length
= simple_read_from_buffer(buf
, count
, ppos
, (char *)page
, length
);
713 static struct file_operations proc_info_file_operations
= {
714 .read
= proc_info_read
,
717 static int mem_open(struct inode
* inode
, struct file
* file
)
719 file
->private_data
= (void*)((long)current
->self_exec_id
);
723 static ssize_t
mem_read(struct file
* file
, char __user
* buf
,
724 size_t count
, loff_t
*ppos
)
726 struct task_struct
*task
= proc_task(file
->f_dentry
->d_inode
);
728 unsigned long src
= *ppos
;
730 struct mm_struct
*mm
;
732 if (!MAY_PTRACE(task
) || !ptrace_may_attach(task
))
736 page
= (char *)__get_free_page(GFP_USER
);
742 mm
= get_task_mm(task
);
748 if (file
->private_data
!= (void*)((long)current
->self_exec_id
))
754 int this_len
, retval
;
756 this_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
757 retval
= access_process_vm(task
, src
, page
, this_len
, 0);
758 if (!retval
|| !MAY_PTRACE(task
) || !ptrace_may_attach(task
)) {
764 if (copy_to_user(buf
, page
, retval
)) {
779 free_page((unsigned long) page
);
784 #define mem_write NULL
787 /* This is a security hazard */
788 static ssize_t
mem_write(struct file
* file
, const char * buf
,
789 size_t count
, loff_t
*ppos
)
793 struct task_struct
*task
= proc_task(file
->f_dentry
->d_inode
);
794 unsigned long dst
= *ppos
;
796 if (!MAY_PTRACE(task
) || !ptrace_may_attach(task
))
799 page
= (char *)__get_free_page(GFP_USER
);
804 int this_len
, retval
;
806 this_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
807 if (copy_from_user(page
, buf
, this_len
)) {
811 retval
= access_process_vm(task
, dst
, page
, this_len
, 1);
823 free_page((unsigned long) page
);
828 static loff_t
mem_lseek(struct file
* file
, loff_t offset
, int orig
)
832 file
->f_pos
= offset
;
835 file
->f_pos
+= offset
;
840 force_successful_syscall_return();
844 static struct file_operations proc_mem_operations
= {
851 static ssize_t
oom_adjust_read(struct file
*file
, char __user
*buf
,
852 size_t count
, loff_t
*ppos
)
854 struct task_struct
*task
= proc_task(file
->f_dentry
->d_inode
);
857 int oom_adjust
= task
->oomkilladj
;
858 loff_t __ppos
= *ppos
;
860 len
= sprintf(buffer
, "%i\n", oom_adjust
);
863 if (count
> len
-__ppos
)
865 if (copy_to_user(buf
, buffer
+ __ppos
, count
))
867 *ppos
= __ppos
+ count
;
871 static ssize_t
oom_adjust_write(struct file
*file
, const char __user
*buf
,
872 size_t count
, loff_t
*ppos
)
874 struct task_struct
*task
= proc_task(file
->f_dentry
->d_inode
);
875 char buffer
[8], *end
;
878 if (!capable(CAP_SYS_RESOURCE
))
880 memset(buffer
, 0, 8);
883 if (copy_from_user(buffer
, buf
, count
))
885 oom_adjust
= simple_strtol(buffer
, &end
, 0);
886 if ((oom_adjust
< -16 || oom_adjust
> 15) && oom_adjust
!= OOM_DISABLE
)
890 task
->oomkilladj
= oom_adjust
;
891 if (end
- buffer
== 0)
896 static struct file_operations proc_oom_adjust_operations
= {
897 .read
= oom_adjust_read
,
898 .write
= oom_adjust_write
,
901 static struct inode_operations proc_mem_inode_operations
= {
902 .permission
= proc_permission
,
905 #ifdef CONFIG_AUDITSYSCALL
907 static ssize_t
proc_loginuid_read(struct file
* file
, char __user
* buf
,
908 size_t count
, loff_t
*ppos
)
910 struct inode
* inode
= file
->f_dentry
->d_inode
;
911 struct task_struct
*task
= proc_task(inode
);
913 char tmpbuf
[TMPBUFLEN
];
915 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
916 audit_get_loginuid(task
->audit_context
));
917 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
920 static ssize_t
proc_loginuid_write(struct file
* file
, const char __user
* buf
,
921 size_t count
, loff_t
*ppos
)
923 struct inode
* inode
= file
->f_dentry
->d_inode
;
926 struct task_struct
*task
= proc_task(inode
);
929 if (!capable(CAP_AUDIT_CONTROL
))
935 if (count
> PAGE_SIZE
)
939 /* No partial writes. */
942 page
= (char*)__get_free_page(GFP_USER
);
946 if (copy_from_user(page
, buf
, count
))
949 loginuid
= simple_strtoul(page
, &tmp
, 10);
955 length
= audit_set_loginuid(task
, loginuid
);
956 if (likely(length
== 0))
960 free_page((unsigned long) page
);
964 static struct file_operations proc_loginuid_operations
= {
965 .read
= proc_loginuid_read
,
966 .write
= proc_loginuid_write
,
970 #ifdef CONFIG_SECCOMP
971 static ssize_t
seccomp_read(struct file
*file
, char __user
*buf
,
972 size_t count
, loff_t
*ppos
)
974 struct task_struct
*tsk
= proc_task(file
->f_dentry
->d_inode
);
976 loff_t __ppos
= *ppos
;
979 /* no need to print the trailing zero, so use only len */
980 len
= sprintf(__buf
, "%u\n", tsk
->seccomp
.mode
);
983 if (count
> len
- __ppos
)
984 count
= len
- __ppos
;
985 if (copy_to_user(buf
, __buf
+ __ppos
, count
))
987 *ppos
= __ppos
+ count
;
991 static ssize_t
seccomp_write(struct file
*file
, const char __user
*buf
,
992 size_t count
, loff_t
*ppos
)
994 struct task_struct
*tsk
= proc_task(file
->f_dentry
->d_inode
);
995 char __buf
[20], *end
;
996 unsigned int seccomp_mode
;
998 /* can set it only once to be even more secure */
999 if (unlikely(tsk
->seccomp
.mode
))
1002 memset(__buf
, 0, sizeof(__buf
));
1003 count
= min(count
, sizeof(__buf
) - 1);
1004 if (copy_from_user(__buf
, buf
, count
))
1006 seccomp_mode
= simple_strtoul(__buf
, &end
, 0);
1009 if (seccomp_mode
&& seccomp_mode
<= NR_SECCOMP_MODES
) {
1010 tsk
->seccomp
.mode
= seccomp_mode
;
1011 set_tsk_thread_flag(tsk
, TIF_SECCOMP
);
1014 if (unlikely(!(end
- __buf
)))
1019 static struct file_operations proc_seccomp_operations
= {
1020 .read
= seccomp_read
,
1021 .write
= seccomp_write
,
1023 #endif /* CONFIG_SECCOMP */
1025 static void *proc_pid_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
1027 struct inode
*inode
= dentry
->d_inode
;
1028 int error
= -EACCES
;
1030 /* We don't need a base pointer in the /proc filesystem */
1033 if (current
->fsuid
!= inode
->i_uid
&& !capable(CAP_DAC_OVERRIDE
))
1035 error
= proc_check_root(inode
);
1039 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &nd
->dentry
, &nd
->mnt
);
1040 nd
->last_type
= LAST_BIND
;
1042 return ERR_PTR(error
);
1045 static int do_proc_readlink(struct dentry
*dentry
, struct vfsmount
*mnt
,
1046 char __user
*buffer
, int buflen
)
1048 struct inode
* inode
;
1049 char *tmp
= (char*)__get_free_page(GFP_KERNEL
), *path
;
1055 inode
= dentry
->d_inode
;
1056 path
= d_path(dentry
, mnt
, tmp
, PAGE_SIZE
);
1057 len
= PTR_ERR(path
);
1060 len
= tmp
+ PAGE_SIZE
- 1 - path
;
1064 if (copy_to_user(buffer
, path
, len
))
1067 free_page((unsigned long)tmp
);
1071 static int proc_pid_readlink(struct dentry
* dentry
, char __user
* buffer
, int buflen
)
1073 int error
= -EACCES
;
1074 struct inode
*inode
= dentry
->d_inode
;
1076 struct vfsmount
*mnt
= NULL
;
1080 if (current
->fsuid
!= inode
->i_uid
&& !capable(CAP_DAC_OVERRIDE
))
1082 error
= proc_check_root(inode
);
1086 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &de
, &mnt
);
1090 error
= do_proc_readlink(de
, mnt
, buffer
, buflen
);
1098 static struct inode_operations proc_pid_link_inode_operations
= {
1099 .readlink
= proc_pid_readlink
,
1100 .follow_link
= proc_pid_follow_link
1105 static int proc_readfd(struct file
* filp
, void * dirent
, filldir_t filldir
)
1107 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1108 struct task_struct
*p
= proc_task(inode
);
1109 unsigned int fd
, tid
, ino
;
1112 struct files_struct
* files
;
1113 struct fdtable
*fdt
;
1124 if (filldir(dirent
, ".", 1, 0, inode
->i_ino
, DT_DIR
) < 0)
1128 ino
= fake_ino(tid
, PROC_TID_INO
);
1129 if (filldir(dirent
, "..", 2, 1, ino
, DT_DIR
) < 0)
1133 files
= get_files_struct(p
);
1137 fdt
= files_fdtable(files
);
1138 for (fd
= filp
->f_pos
-2;
1140 fd
++, filp
->f_pos
++) {
1143 if (!fcheck_files(files
, fd
))
1151 buf
[j
] = '0' + (i
% 10);
1155 ino
= fake_ino(tid
, PROC_TID_FD_DIR
+ fd
);
1156 if (filldir(dirent
, buf
+j
, NUMBUF
-j
, fd
+2, ino
, DT_LNK
) < 0) {
1163 put_files_struct(files
);
1169 static int proc_pident_readdir(struct file
*filp
,
1170 void *dirent
, filldir_t filldir
,
1171 struct pid_entry
*ents
, unsigned int nents
)
1175 struct dentry
*dentry
= filp
->f_dentry
;
1176 struct inode
*inode
= dentry
->d_inode
;
1177 struct pid_entry
*p
;
1182 if (!pid_alive(proc_task(inode
)))
1186 pid
= proc_task(inode
)->pid
;
1191 if (filldir(dirent
, ".", 1, i
, ino
, DT_DIR
) < 0)
1197 ino
= parent_ino(dentry
);
1198 if (filldir(dirent
, "..", 2, i
, ino
, DT_DIR
) < 0)
1211 if (filldir(dirent
, p
->name
, p
->len
, filp
->f_pos
,
1212 fake_ino(pid
, p
->type
), p
->mode
>> 12) < 0)
1224 static int proc_tgid_base_readdir(struct file
* filp
,
1225 void * dirent
, filldir_t filldir
)
1227 return proc_pident_readdir(filp
,dirent
,filldir
,
1228 tgid_base_stuff
,ARRAY_SIZE(tgid_base_stuff
));
1231 static int proc_tid_base_readdir(struct file
* filp
,
1232 void * dirent
, filldir_t filldir
)
1234 return proc_pident_readdir(filp
,dirent
,filldir
,
1235 tid_base_stuff
,ARRAY_SIZE(tid_base_stuff
));
1238 /* building an inode */
1240 static int task_dumpable(struct task_struct
*task
)
1243 struct mm_struct
*mm
;
1248 dumpable
= mm
->dumpable
;
1256 static struct inode
*proc_pid_make_inode(struct super_block
* sb
, struct task_struct
*task
, int ino
)
1258 struct inode
* inode
;
1259 struct proc_inode
*ei
;
1261 /* We need a new inode */
1263 inode
= new_inode(sb
);
1270 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1271 inode
->i_ino
= fake_ino(task
->pid
, ino
);
1273 if (!pid_alive(task
))
1277 * grab the reference to task.
1279 get_task_struct(task
);
1284 if (ino
== PROC_TGID_INO
|| ino
== PROC_TID_INO
|| task_dumpable(task
)) {
1285 inode
->i_uid
= task
->euid
;
1286 inode
->i_gid
= task
->egid
;
1288 security_task_to_inode(task
, inode
);
1302 * Exceptional case: normally we are not allowed to unhash a busy
1303 * directory. In this case, however, we can do it - no aliasing problems
1304 * due to the way we treat inodes.
1306 * Rewrite the inode's ownerships here because the owning task may have
1307 * performed a setuid(), etc.
1309 static int pid_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1311 struct inode
*inode
= dentry
->d_inode
;
1312 struct task_struct
*task
= proc_task(inode
);
1313 if (pid_alive(task
)) {
1314 if (proc_type(inode
) == PROC_TGID_INO
|| proc_type(inode
) == PROC_TID_INO
|| task_dumpable(task
)) {
1315 inode
->i_uid
= task
->euid
;
1316 inode
->i_gid
= task
->egid
;
1321 security_task_to_inode(task
, inode
);
1328 static int tid_fd_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1330 struct inode
*inode
= dentry
->d_inode
;
1331 struct task_struct
*task
= proc_task(inode
);
1332 int fd
= proc_type(inode
) - PROC_TID_FD_DIR
;
1333 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 security_task_to_inode(task
, inode
);
1352 put_files_struct(files
);
1358 static void pid_base_iput(struct dentry
*dentry
, struct inode
*inode
)
1360 struct task_struct
*task
= proc_task(inode
);
1361 spin_lock(&task
->proc_lock
);
1362 if (task
->proc_dentry
== dentry
)
1363 task
->proc_dentry
= NULL
;
1364 spin_unlock(&task
->proc_lock
);
1368 static int pid_delete_dentry(struct dentry
* dentry
)
1370 /* Is the task we represent dead?
1371 * If so, then don't put the dentry on the lru list,
1372 * kill it immediately.
1374 return !pid_alive(proc_task(dentry
->d_inode
));
1377 static struct dentry_operations tid_fd_dentry_operations
=
1379 .d_revalidate
= tid_fd_revalidate
,
1380 .d_delete
= pid_delete_dentry
,
1383 static struct dentry_operations pid_dentry_operations
=
1385 .d_revalidate
= pid_revalidate
,
1386 .d_delete
= pid_delete_dentry
,
1389 static struct dentry_operations pid_base_dentry_operations
=
1391 .d_revalidate
= pid_revalidate
,
1392 .d_iput
= pid_base_iput
,
1393 .d_delete
= pid_delete_dentry
,
1398 static unsigned name_to_int(struct dentry
*dentry
)
1400 const char *name
= dentry
->d_name
.name
;
1401 int len
= dentry
->d_name
.len
;
1404 if (len
> 1 && *name
== '0')
1407 unsigned c
= *name
++ - '0';
1410 if (n
>= (~0U-9)/10)
1421 static struct dentry
*proc_lookupfd(struct inode
* dir
, struct dentry
* dentry
, struct nameidata
*nd
)
1423 struct task_struct
*task
= proc_task(dir
);
1424 unsigned fd
= name_to_int(dentry
);
1426 struct files_struct
* files
;
1427 struct inode
*inode
;
1428 struct proc_inode
*ei
;
1432 if (!pid_alive(task
))
1435 inode
= proc_pid_make_inode(dir
->i_sb
, task
, PROC_TID_FD_DIR
+fd
);
1439 files
= get_files_struct(task
);
1442 inode
->i_mode
= S_IFLNK
;
1444 file
= fcheck_files(files
, fd
);
1447 if (file
->f_mode
& 1)
1448 inode
->i_mode
|= S_IRUSR
| S_IXUSR
;
1449 if (file
->f_mode
& 2)
1450 inode
->i_mode
|= S_IWUSR
| S_IXUSR
;
1452 put_files_struct(files
);
1453 inode
->i_op
= &proc_pid_link_inode_operations
;
1455 ei
->op
.proc_get_link
= proc_fd_link
;
1456 dentry
->d_op
= &tid_fd_dentry_operations
;
1457 d_add(dentry
, inode
);
1462 put_files_struct(files
);
1466 return ERR_PTR(-ENOENT
);
1469 static int proc_task_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
);
1470 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
);
1472 static struct file_operations proc_fd_operations
= {
1473 .read
= generic_read_dir
,
1474 .readdir
= proc_readfd
,
1477 static struct file_operations proc_task_operations
= {
1478 .read
= generic_read_dir
,
1479 .readdir
= proc_task_readdir
,
1483 * proc directories can do almost nothing..
1485 static struct inode_operations proc_fd_inode_operations
= {
1486 .lookup
= proc_lookupfd
,
1487 .permission
= proc_permission
,
1490 static struct inode_operations proc_task_inode_operations
= {
1491 .lookup
= proc_task_lookup
,
1492 .permission
= proc_task_permission
,
1495 #ifdef CONFIG_SECURITY
1496 static ssize_t
proc_pid_attr_read(struct file
* file
, char __user
* buf
,
1497 size_t count
, loff_t
*ppos
)
1499 struct inode
* inode
= file
->f_dentry
->d_inode
;
1502 struct task_struct
*task
= proc_task(inode
);
1504 if (count
> PAGE_SIZE
)
1506 if (!(page
= __get_free_page(GFP_KERNEL
)))
1509 length
= security_getprocattr(task
,
1510 (char*)file
->f_dentry
->d_name
.name
,
1511 (void*)page
, count
);
1513 length
= simple_read_from_buffer(buf
, count
, ppos
, (char *)page
, length
);
1518 static ssize_t
proc_pid_attr_write(struct file
* file
, const char __user
* buf
,
1519 size_t count
, loff_t
*ppos
)
1521 struct inode
* inode
= file
->f_dentry
->d_inode
;
1524 struct task_struct
*task
= proc_task(inode
);
1526 if (count
> PAGE_SIZE
)
1529 /* No partial writes. */
1532 page
= (char*)__get_free_page(GFP_USER
);
1536 if (copy_from_user(page
, buf
, count
))
1539 length
= security_setprocattr(task
,
1540 (char*)file
->f_dentry
->d_name
.name
,
1541 (void*)page
, count
);
1543 free_page((unsigned long) page
);
1547 static struct file_operations proc_pid_attr_operations
= {
1548 .read
= proc_pid_attr_read
,
1549 .write
= proc_pid_attr_write
,
1552 static struct file_operations proc_tid_attr_operations
;
1553 static struct inode_operations proc_tid_attr_inode_operations
;
1554 static struct file_operations proc_tgid_attr_operations
;
1555 static struct inode_operations proc_tgid_attr_inode_operations
;
1558 static int get_tid_list(int index
, unsigned int *tids
, struct inode
*dir
);
1561 static struct dentry
*proc_pident_lookup(struct inode
*dir
,
1562 struct dentry
*dentry
,
1563 struct pid_entry
*ents
)
1565 struct inode
*inode
;
1567 struct task_struct
*task
= proc_task(dir
);
1568 struct pid_entry
*p
;
1569 struct proc_inode
*ei
;
1574 if (!pid_alive(task
))
1577 for (p
= ents
; p
->name
; p
++) {
1578 if (p
->len
!= dentry
->d_name
.len
)
1580 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
1587 inode
= proc_pid_make_inode(dir
->i_sb
, task
, p
->type
);
1592 inode
->i_mode
= p
->mode
;
1594 * Yes, it does not scale. And it should not. Don't add
1595 * new entries into /proc/<tgid>/ without very good reasons.
1598 case PROC_TGID_TASK
:
1599 inode
->i_nlink
= 2 + get_tid_list(2, NULL
, dir
);
1600 inode
->i_op
= &proc_task_inode_operations
;
1601 inode
->i_fop
= &proc_task_operations
;
1606 inode
->i_op
= &proc_fd_inode_operations
;
1607 inode
->i_fop
= &proc_fd_operations
;
1611 inode
->i_op
= &proc_pid_link_inode_operations
;
1612 ei
->op
.proc_get_link
= proc_exe_link
;
1616 inode
->i_op
= &proc_pid_link_inode_operations
;
1617 ei
->op
.proc_get_link
= proc_cwd_link
;
1620 case PROC_TGID_ROOT
:
1621 inode
->i_op
= &proc_pid_link_inode_operations
;
1622 ei
->op
.proc_get_link
= proc_root_link
;
1624 case PROC_TID_ENVIRON
:
1625 case PROC_TGID_ENVIRON
:
1626 inode
->i_fop
= &proc_info_file_operations
;
1627 ei
->op
.proc_read
= proc_pid_environ
;
1630 case PROC_TGID_AUXV
:
1631 inode
->i_fop
= &proc_info_file_operations
;
1632 ei
->op
.proc_read
= proc_pid_auxv
;
1634 case PROC_TID_STATUS
:
1635 case PROC_TGID_STATUS
:
1636 inode
->i_fop
= &proc_info_file_operations
;
1637 ei
->op
.proc_read
= proc_pid_status
;
1640 inode
->i_fop
= &proc_info_file_operations
;
1641 ei
->op
.proc_read
= proc_tid_stat
;
1643 case PROC_TGID_STAT
:
1644 inode
->i_fop
= &proc_info_file_operations
;
1645 ei
->op
.proc_read
= proc_tgid_stat
;
1647 case PROC_TID_CMDLINE
:
1648 case PROC_TGID_CMDLINE
:
1649 inode
->i_fop
= &proc_info_file_operations
;
1650 ei
->op
.proc_read
= proc_pid_cmdline
;
1652 case PROC_TID_STATM
:
1653 case PROC_TGID_STATM
:
1654 inode
->i_fop
= &proc_info_file_operations
;
1655 ei
->op
.proc_read
= proc_pid_statm
;
1658 case PROC_TGID_MAPS
:
1659 inode
->i_fop
= &proc_maps_operations
;
1662 case PROC_TID_NUMA_MAPS
:
1663 case PROC_TGID_NUMA_MAPS
:
1664 inode
->i_fop
= &proc_numa_maps_operations
;
1669 inode
->i_op
= &proc_mem_inode_operations
;
1670 inode
->i_fop
= &proc_mem_operations
;
1672 #ifdef CONFIG_SECCOMP
1673 case PROC_TID_SECCOMP
:
1674 case PROC_TGID_SECCOMP
:
1675 inode
->i_fop
= &proc_seccomp_operations
;
1677 #endif /* CONFIG_SECCOMP */
1678 case PROC_TID_MOUNTS
:
1679 case PROC_TGID_MOUNTS
:
1680 inode
->i_fop
= &proc_mounts_operations
;
1682 case PROC_TID_SMAPS
:
1683 case PROC_TGID_SMAPS
:
1684 inode
->i_fop
= &proc_smaps_operations
;
1686 #ifdef CONFIG_SECURITY
1689 inode
->i_op
= &proc_tid_attr_inode_operations
;
1690 inode
->i_fop
= &proc_tid_attr_operations
;
1692 case PROC_TGID_ATTR
:
1694 inode
->i_op
= &proc_tgid_attr_inode_operations
;
1695 inode
->i_fop
= &proc_tgid_attr_operations
;
1697 case PROC_TID_ATTR_CURRENT
:
1698 case PROC_TGID_ATTR_CURRENT
:
1699 case PROC_TID_ATTR_PREV
:
1700 case PROC_TGID_ATTR_PREV
:
1701 case PROC_TID_ATTR_EXEC
:
1702 case PROC_TGID_ATTR_EXEC
:
1703 case PROC_TID_ATTR_FSCREATE
:
1704 case PROC_TGID_ATTR_FSCREATE
:
1705 inode
->i_fop
= &proc_pid_attr_operations
;
1708 #ifdef CONFIG_KALLSYMS
1709 case PROC_TID_WCHAN
:
1710 case PROC_TGID_WCHAN
:
1711 inode
->i_fop
= &proc_info_file_operations
;
1712 ei
->op
.proc_read
= proc_pid_wchan
;
1715 #ifdef CONFIG_SCHEDSTATS
1716 case PROC_TID_SCHEDSTAT
:
1717 case PROC_TGID_SCHEDSTAT
:
1718 inode
->i_fop
= &proc_info_file_operations
;
1719 ei
->op
.proc_read
= proc_pid_schedstat
;
1722 #ifdef CONFIG_CPUSETS
1723 case PROC_TID_CPUSET
:
1724 case PROC_TGID_CPUSET
:
1725 inode
->i_fop
= &proc_cpuset_operations
;
1728 case PROC_TID_OOM_SCORE
:
1729 case PROC_TGID_OOM_SCORE
:
1730 inode
->i_fop
= &proc_info_file_operations
;
1731 ei
->op
.proc_read
= proc_oom_score
;
1733 case PROC_TID_OOM_ADJUST
:
1734 case PROC_TGID_OOM_ADJUST
:
1735 inode
->i_fop
= &proc_oom_adjust_operations
;
1737 #ifdef CONFIG_AUDITSYSCALL
1738 case PROC_TID_LOGINUID
:
1739 case PROC_TGID_LOGINUID
:
1740 inode
->i_fop
= &proc_loginuid_operations
;
1744 printk("procfs: impossible type (%d)",p
->type
);
1746 return ERR_PTR(-EINVAL
);
1748 dentry
->d_op
= &pid_dentry_operations
;
1749 d_add(dentry
, inode
);
1753 return ERR_PTR(error
);
1756 static struct dentry
*proc_tgid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
1757 return proc_pident_lookup(dir
, dentry
, tgid_base_stuff
);
1760 static struct dentry
*proc_tid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
1761 return proc_pident_lookup(dir
, dentry
, tid_base_stuff
);
1764 static struct file_operations proc_tgid_base_operations
= {
1765 .read
= generic_read_dir
,
1766 .readdir
= proc_tgid_base_readdir
,
1769 static struct file_operations proc_tid_base_operations
= {
1770 .read
= generic_read_dir
,
1771 .readdir
= proc_tid_base_readdir
,
1774 static struct inode_operations proc_tgid_base_inode_operations
= {
1775 .lookup
= proc_tgid_base_lookup
,
1778 static struct inode_operations proc_tid_base_inode_operations
= {
1779 .lookup
= proc_tid_base_lookup
,
1782 #ifdef CONFIG_SECURITY
1783 static int proc_tgid_attr_readdir(struct file
* filp
,
1784 void * dirent
, filldir_t filldir
)
1786 return proc_pident_readdir(filp
,dirent
,filldir
,
1787 tgid_attr_stuff
,ARRAY_SIZE(tgid_attr_stuff
));
1790 static int proc_tid_attr_readdir(struct file
* filp
,
1791 void * dirent
, filldir_t filldir
)
1793 return proc_pident_readdir(filp
,dirent
,filldir
,
1794 tid_attr_stuff
,ARRAY_SIZE(tid_attr_stuff
));
1797 static struct file_operations proc_tgid_attr_operations
= {
1798 .read
= generic_read_dir
,
1799 .readdir
= proc_tgid_attr_readdir
,
1802 static struct file_operations proc_tid_attr_operations
= {
1803 .read
= generic_read_dir
,
1804 .readdir
= proc_tid_attr_readdir
,
1807 static struct dentry
*proc_tgid_attr_lookup(struct inode
*dir
,
1808 struct dentry
*dentry
, struct nameidata
*nd
)
1810 return proc_pident_lookup(dir
, dentry
, tgid_attr_stuff
);
1813 static struct dentry
*proc_tid_attr_lookup(struct inode
*dir
,
1814 struct dentry
*dentry
, struct nameidata
*nd
)
1816 return proc_pident_lookup(dir
, dentry
, tid_attr_stuff
);
1819 static struct inode_operations proc_tgid_attr_inode_operations
= {
1820 .lookup
= proc_tgid_attr_lookup
,
1823 static struct inode_operations proc_tid_attr_inode_operations
= {
1824 .lookup
= proc_tid_attr_lookup
,
1831 static int proc_self_readlink(struct dentry
*dentry
, char __user
*buffer
,
1835 sprintf(tmp
, "%d", current
->tgid
);
1836 return vfs_readlink(dentry
,buffer
,buflen
,tmp
);
1839 static void *proc_self_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
1842 sprintf(tmp
, "%d", current
->tgid
);
1843 return ERR_PTR(vfs_follow_link(nd
,tmp
));
1846 static struct inode_operations proc_self_inode_operations
= {
1847 .readlink
= proc_self_readlink
,
1848 .follow_link
= proc_self_follow_link
,
1852 * proc_pid_unhash - Unhash /proc/@pid entry from the dcache.
1853 * @p: task that should be flushed.
1855 * Drops the /proc/@pid dcache entry from the hash chains.
1857 * Dropping /proc/@pid entries and detach_pid must be synchroneous,
1858 * otherwise e.g. /proc/@pid/exe might point to the wrong executable,
1859 * if the pid value is immediately reused. This is enforced by
1860 * - caller must acquire spin_lock(p->proc_lock)
1861 * - must be called before detach_pid()
1862 * - proc_pid_lookup acquires proc_lock, and checks that
1863 * the target is not dead by looking at the attach count
1867 struct dentry
*proc_pid_unhash(struct task_struct
*p
)
1869 struct dentry
*proc_dentry
;
1871 proc_dentry
= p
->proc_dentry
;
1872 if (proc_dentry
!= NULL
) {
1874 spin_lock(&dcache_lock
);
1875 spin_lock(&proc_dentry
->d_lock
);
1876 if (!d_unhashed(proc_dentry
)) {
1877 dget_locked(proc_dentry
);
1878 __d_drop(proc_dentry
);
1879 spin_unlock(&proc_dentry
->d_lock
);
1881 spin_unlock(&proc_dentry
->d_lock
);
1884 spin_unlock(&dcache_lock
);
1890 * proc_pid_flush - recover memory used by stale /proc/@pid/x entries
1891 * @proc_dentry: directoy to prune.
1893 * Shrink the /proc directory that was used by the just killed thread.
1896 void proc_pid_flush(struct dentry
*proc_dentry
)
1899 if(proc_dentry
!= NULL
) {
1900 shrink_dcache_parent(proc_dentry
);
1906 struct dentry
*proc_pid_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
1908 struct task_struct
*task
;
1909 struct inode
*inode
;
1910 struct proc_inode
*ei
;
1914 if (dentry
->d_name
.len
== 4 && !memcmp(dentry
->d_name
.name
,"self",4)) {
1915 inode
= new_inode(dir
->i_sb
);
1917 return ERR_PTR(-ENOMEM
);
1919 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1920 inode
->i_ino
= fake_ino(0, PROC_TGID_INO
);
1922 inode
->i_mode
= S_IFLNK
|S_IRWXUGO
;
1923 inode
->i_uid
= inode
->i_gid
= 0;
1925 inode
->i_op
= &proc_self_inode_operations
;
1926 d_add(dentry
, inode
);
1929 tgid
= name_to_int(dentry
);
1933 read_lock(&tasklist_lock
);
1934 task
= find_task_by_pid(tgid
);
1936 get_task_struct(task
);
1937 read_unlock(&tasklist_lock
);
1941 inode
= proc_pid_make_inode(dir
->i_sb
, task
, PROC_TGID_INO
);
1945 put_task_struct(task
);
1948 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
1949 inode
->i_op
= &proc_tgid_base_inode_operations
;
1950 inode
->i_fop
= &proc_tgid_base_operations
;
1951 inode
->i_flags
|=S_IMMUTABLE
;
1952 #ifdef CONFIG_SECURITY
1958 dentry
->d_op
= &pid_base_dentry_operations
;
1961 d_add(dentry
, inode
);
1962 spin_lock(&task
->proc_lock
);
1963 task
->proc_dentry
= dentry
;
1964 if (!pid_alive(task
)) {
1965 dentry
= proc_pid_unhash(task
);
1968 spin_unlock(&task
->proc_lock
);
1970 put_task_struct(task
);
1972 proc_pid_flush(dentry
);
1977 return ERR_PTR(-ENOENT
);
1981 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
1983 struct task_struct
*task
;
1984 struct task_struct
*leader
= proc_task(dir
);
1985 struct inode
*inode
;
1988 tid
= name_to_int(dentry
);
1992 read_lock(&tasklist_lock
);
1993 task
= find_task_by_pid(tid
);
1995 get_task_struct(task
);
1996 read_unlock(&tasklist_lock
);
1999 if (leader
->tgid
!= task
->tgid
)
2002 inode
= proc_pid_make_inode(dir
->i_sb
, task
, PROC_TID_INO
);
2007 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
2008 inode
->i_op
= &proc_tid_base_inode_operations
;
2009 inode
->i_fop
= &proc_tid_base_operations
;
2010 inode
->i_flags
|=S_IMMUTABLE
;
2011 #ifdef CONFIG_SECURITY
2017 dentry
->d_op
= &pid_base_dentry_operations
;
2019 d_add(dentry
, inode
);
2021 put_task_struct(task
);
2024 put_task_struct(task
);
2026 return ERR_PTR(-ENOENT
);
2029 #define PROC_NUMBUF 10
2030 #define PROC_MAXPIDS 20
2033 * Get a few tgid's to return for filldir - we need to hold the
2034 * tasklist lock while doing this, and we must release it before
2035 * we actually do the filldir itself, so we use a temp buffer..
2037 static int get_tgid_list(int index
, unsigned long version
, unsigned int *tgids
)
2039 struct task_struct
*p
;
2043 read_lock(&tasklist_lock
);
2046 p
= find_task_by_pid(version
);
2047 if (p
&& !thread_group_leader(p
))
2054 p
= next_task(&init_task
);
2056 for ( ; p
!= &init_task
; p
= next_task(p
)) {
2062 tgids
[nr_tgids
] = tgid
;
2064 if (nr_tgids
>= PROC_MAXPIDS
)
2067 read_unlock(&tasklist_lock
);
2072 * Get a few tid's to return for filldir - we need to hold the
2073 * tasklist lock while doing this, and we must release it before
2074 * we actually do the filldir itself, so we use a temp buffer..
2076 static int get_tid_list(int index
, unsigned int *tids
, struct inode
*dir
)
2078 struct task_struct
*leader_task
= proc_task(dir
);
2079 struct task_struct
*task
= leader_task
;
2083 read_lock(&tasklist_lock
);
2085 * The starting point task (leader_task) might be an already
2086 * unlinked task, which cannot be used to access the task-list
2087 * via next_thread().
2089 if (pid_alive(task
)) do {
2090 int tid
= task
->pid
;
2095 tids
[nr_tids
] = tid
;
2097 if (nr_tids
>= PROC_MAXPIDS
)
2099 } while ((task
= next_thread(task
)) != leader_task
);
2100 read_unlock(&tasklist_lock
);
2104 /* for the /proc/ directory itself, after non-process stuff has been done */
2105 int proc_pid_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
2107 unsigned int tgid_array
[PROC_MAXPIDS
];
2108 char buf
[PROC_NUMBUF
];
2109 unsigned int nr
= filp
->f_pos
- FIRST_PROCESS_ENTRY
;
2110 unsigned int nr_tgids
, i
;
2114 ino_t ino
= fake_ino(0,PROC_TGID_INO
);
2115 if (filldir(dirent
, "self", 4, filp
->f_pos
, ino
, DT_LNK
) < 0)
2121 /* f_version caches the tgid value that the last readdir call couldn't
2122 * return. lseek aka telldir automagically resets f_version to 0.
2124 next_tgid
= filp
->f_version
;
2125 filp
->f_version
= 0;
2127 nr_tgids
= get_tgid_list(nr
, next_tgid
, tgid_array
);
2129 /* no more entries ! */
2134 /* do not use the last found pid, reserve it for next_tgid */
2135 if (nr_tgids
== PROC_MAXPIDS
) {
2137 next_tgid
= tgid_array
[nr_tgids
];
2140 for (i
=0;i
<nr_tgids
;i
++) {
2141 int tgid
= tgid_array
[i
];
2142 ino_t ino
= fake_ino(tgid
,PROC_TGID_INO
);
2143 unsigned long j
= PROC_NUMBUF
;
2146 buf
[--j
] = '0' + (tgid
% 10);
2147 while ((tgid
/= 10) != 0);
2149 if (filldir(dirent
, buf
+j
, PROC_NUMBUF
-j
, filp
->f_pos
, ino
, DT_DIR
) < 0) {
2150 /* returning this tgid failed, save it as the first
2151 * pid for the next readir call */
2152 filp
->f_version
= tgid_array
[i
];
2163 /* for the /proc/TGID/task/ directories */
2164 static int proc_task_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
2166 unsigned int tid_array
[PROC_MAXPIDS
];
2167 char buf
[PROC_NUMBUF
];
2168 unsigned int nr_tids
, i
;
2169 struct dentry
*dentry
= filp
->f_dentry
;
2170 struct inode
*inode
= dentry
->d_inode
;
2171 int retval
= -ENOENT
;
2173 unsigned long pos
= filp
->f_pos
; /* avoiding "long long" filp->f_pos */
2175 if (!pid_alive(proc_task(inode
)))
2182 if (filldir(dirent
, ".", 1, pos
, ino
, DT_DIR
) < 0)
2187 ino
= parent_ino(dentry
);
2188 if (filldir(dirent
, "..", 2, pos
, ino
, DT_DIR
) < 0)
2194 nr_tids
= get_tid_list(pos
, tid_array
, inode
);
2195 inode
->i_nlink
= pos
+ nr_tids
;
2197 for (i
= 0; i
< nr_tids
; i
++) {
2198 unsigned long j
= PROC_NUMBUF
;
2199 int tid
= tid_array
[i
];
2201 ino
= fake_ino(tid
,PROC_TID_INO
);
2204 buf
[--j
] = '0' + (tid
% 10);
2205 while ((tid
/= 10) != 0);
2207 if (filldir(dirent
, buf
+j
, PROC_NUMBUF
-j
, pos
, ino
, DT_DIR
) < 0)