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/capability.h>
59 #include <linux/file.h>
60 #include <linux/string.h>
61 #include <linux/seq_file.h>
62 #include <linux/namei.h>
63 #include <linux/namespace.h>
65 #include <linux/smp_lock.h>
66 #include <linux/rcupdate.h>
67 #include <linux/kallsyms.h>
68 #include <linux/mount.h>
69 #include <linux/security.h>
70 #include <linux/ptrace.h>
71 #include <linux/seccomp.h>
72 #include <linux/cpuset.h>
73 #include <linux/audit.h>
74 #include <linux/poll.h>
78 * Implementing inode permission operations in /proc is almost
79 * certainly an error. Permission checks need to happen during
80 * each system call not at open time. The reason is that most of
81 * what we wish to check for permissions in /proc varies at runtime.
83 * The classic example of a problem is opening file descriptors
84 * in /proc for a task before it execs a suid executable.
88 * For hysterical raisins we keep the same inumbers as in the old procfs.
89 * Feel free to change the macro below - just keep the range distinct from
90 * inumbers of the rest of procfs (currently those are in 0x0000--0xffff).
91 * As soon as we'll get a separate superblock we will be able to forget
92 * about magical ranges too.
95 #define fake_ino(pid,ino) (((pid)<<16)|(ino))
97 enum pid_directory_inos
{
102 #ifdef CONFIG_SECCOMP
117 PROC_TGID_MOUNTSTATS
,
122 #ifdef CONFIG_SCHEDSTATS
125 #ifdef CONFIG_CPUSETS
128 #ifdef CONFIG_SECURITY
130 PROC_TGID_ATTR_CURRENT
,
133 PROC_TGID_ATTR_FSCREATE
,
134 PROC_TGID_ATTR_KEYCREATE
,
136 #ifdef CONFIG_AUDITSYSCALL
140 PROC_TGID_OOM_ADJUST
,
144 #ifdef CONFIG_SECCOMP
164 #ifdef CONFIG_SCHEDSTATS
167 #ifdef CONFIG_CPUSETS
170 #ifdef CONFIG_SECURITY
172 PROC_TID_ATTR_CURRENT
,
175 PROC_TID_ATTR_FSCREATE
,
176 PROC_TID_ATTR_KEYCREATE
,
178 #ifdef CONFIG_AUDITSYSCALL
184 /* Add new entries before this */
185 PROC_TID_FD_DIR
= 0x8000, /* 0x8000-0xffff */
188 /* Worst case buffer size needed for holding an integer. */
189 #define PROC_NUMBUF 10
198 #define E(type,name,mode) {(type),sizeof(name)-1,(name),(mode)}
200 static struct pid_entry tgid_base_stuff
[] = {
201 E(PROC_TGID_TASK
, "task", S_IFDIR
|S_IRUGO
|S_IXUGO
),
202 E(PROC_TGID_FD
, "fd", S_IFDIR
|S_IRUSR
|S_IXUSR
),
203 E(PROC_TGID_ENVIRON
, "environ", S_IFREG
|S_IRUSR
),
204 E(PROC_TGID_AUXV
, "auxv", S_IFREG
|S_IRUSR
),
205 E(PROC_TGID_STATUS
, "status", S_IFREG
|S_IRUGO
),
206 E(PROC_TGID_CMDLINE
, "cmdline", S_IFREG
|S_IRUGO
),
207 E(PROC_TGID_STAT
, "stat", S_IFREG
|S_IRUGO
),
208 E(PROC_TGID_STATM
, "statm", S_IFREG
|S_IRUGO
),
209 E(PROC_TGID_MAPS
, "maps", S_IFREG
|S_IRUGO
),
211 E(PROC_TGID_NUMA_MAPS
, "numa_maps", S_IFREG
|S_IRUGO
),
213 E(PROC_TGID_MEM
, "mem", S_IFREG
|S_IRUSR
|S_IWUSR
),
214 #ifdef CONFIG_SECCOMP
215 E(PROC_TGID_SECCOMP
, "seccomp", S_IFREG
|S_IRUSR
|S_IWUSR
),
217 E(PROC_TGID_CWD
, "cwd", S_IFLNK
|S_IRWXUGO
),
218 E(PROC_TGID_ROOT
, "root", S_IFLNK
|S_IRWXUGO
),
219 E(PROC_TGID_EXE
, "exe", S_IFLNK
|S_IRWXUGO
),
220 E(PROC_TGID_MOUNTS
, "mounts", S_IFREG
|S_IRUGO
),
221 E(PROC_TGID_MOUNTSTATS
, "mountstats", S_IFREG
|S_IRUSR
),
223 E(PROC_TGID_SMAPS
, "smaps", S_IFREG
|S_IRUGO
),
225 #ifdef CONFIG_SECURITY
226 E(PROC_TGID_ATTR
, "attr", S_IFDIR
|S_IRUGO
|S_IXUGO
),
228 #ifdef CONFIG_KALLSYMS
229 E(PROC_TGID_WCHAN
, "wchan", S_IFREG
|S_IRUGO
),
231 #ifdef CONFIG_SCHEDSTATS
232 E(PROC_TGID_SCHEDSTAT
, "schedstat", S_IFREG
|S_IRUGO
),
234 #ifdef CONFIG_CPUSETS
235 E(PROC_TGID_CPUSET
, "cpuset", S_IFREG
|S_IRUGO
),
237 E(PROC_TGID_OOM_SCORE
, "oom_score",S_IFREG
|S_IRUGO
),
238 E(PROC_TGID_OOM_ADJUST
,"oom_adj", S_IFREG
|S_IRUGO
|S_IWUSR
),
239 #ifdef CONFIG_AUDITSYSCALL
240 E(PROC_TGID_LOGINUID
, "loginuid", S_IFREG
|S_IWUSR
|S_IRUGO
),
244 static struct pid_entry tid_base_stuff
[] = {
245 E(PROC_TID_FD
, "fd", S_IFDIR
|S_IRUSR
|S_IXUSR
),
246 E(PROC_TID_ENVIRON
, "environ", S_IFREG
|S_IRUSR
),
247 E(PROC_TID_AUXV
, "auxv", S_IFREG
|S_IRUSR
),
248 E(PROC_TID_STATUS
, "status", S_IFREG
|S_IRUGO
),
249 E(PROC_TID_CMDLINE
, "cmdline", S_IFREG
|S_IRUGO
),
250 E(PROC_TID_STAT
, "stat", S_IFREG
|S_IRUGO
),
251 E(PROC_TID_STATM
, "statm", S_IFREG
|S_IRUGO
),
252 E(PROC_TID_MAPS
, "maps", S_IFREG
|S_IRUGO
),
254 E(PROC_TID_NUMA_MAPS
, "numa_maps", S_IFREG
|S_IRUGO
),
256 E(PROC_TID_MEM
, "mem", S_IFREG
|S_IRUSR
|S_IWUSR
),
257 #ifdef CONFIG_SECCOMP
258 E(PROC_TID_SECCOMP
, "seccomp", S_IFREG
|S_IRUSR
|S_IWUSR
),
260 E(PROC_TID_CWD
, "cwd", S_IFLNK
|S_IRWXUGO
),
261 E(PROC_TID_ROOT
, "root", S_IFLNK
|S_IRWXUGO
),
262 E(PROC_TID_EXE
, "exe", S_IFLNK
|S_IRWXUGO
),
263 E(PROC_TID_MOUNTS
, "mounts", S_IFREG
|S_IRUGO
),
265 E(PROC_TID_SMAPS
, "smaps", S_IFREG
|S_IRUGO
),
267 #ifdef CONFIG_SECURITY
268 E(PROC_TID_ATTR
, "attr", S_IFDIR
|S_IRUGO
|S_IXUGO
),
270 #ifdef CONFIG_KALLSYMS
271 E(PROC_TID_WCHAN
, "wchan", S_IFREG
|S_IRUGO
),
273 #ifdef CONFIG_SCHEDSTATS
274 E(PROC_TID_SCHEDSTAT
, "schedstat",S_IFREG
|S_IRUGO
),
276 #ifdef CONFIG_CPUSETS
277 E(PROC_TID_CPUSET
, "cpuset", S_IFREG
|S_IRUGO
),
279 E(PROC_TID_OOM_SCORE
, "oom_score",S_IFREG
|S_IRUGO
),
280 E(PROC_TID_OOM_ADJUST
, "oom_adj", S_IFREG
|S_IRUGO
|S_IWUSR
),
281 #ifdef CONFIG_AUDITSYSCALL
282 E(PROC_TID_LOGINUID
, "loginuid", S_IFREG
|S_IWUSR
|S_IRUGO
),
287 #ifdef CONFIG_SECURITY
288 static struct pid_entry tgid_attr_stuff
[] = {
289 E(PROC_TGID_ATTR_CURRENT
, "current", S_IFREG
|S_IRUGO
|S_IWUGO
),
290 E(PROC_TGID_ATTR_PREV
, "prev", S_IFREG
|S_IRUGO
),
291 E(PROC_TGID_ATTR_EXEC
, "exec", S_IFREG
|S_IRUGO
|S_IWUGO
),
292 E(PROC_TGID_ATTR_FSCREATE
, "fscreate", S_IFREG
|S_IRUGO
|S_IWUGO
),
293 E(PROC_TGID_ATTR_KEYCREATE
, "keycreate", S_IFREG
|S_IRUGO
|S_IWUGO
),
296 static struct pid_entry tid_attr_stuff
[] = {
297 E(PROC_TID_ATTR_CURRENT
, "current", S_IFREG
|S_IRUGO
|S_IWUGO
),
298 E(PROC_TID_ATTR_PREV
, "prev", S_IFREG
|S_IRUGO
),
299 E(PROC_TID_ATTR_EXEC
, "exec", S_IFREG
|S_IRUGO
|S_IWUGO
),
300 E(PROC_TID_ATTR_FSCREATE
, "fscreate", S_IFREG
|S_IRUGO
|S_IWUGO
),
301 E(PROC_TID_ATTR_KEYCREATE
, "keycreate", S_IFREG
|S_IRUGO
|S_IWUGO
),
308 static int proc_fd_link(struct inode
*inode
, struct dentry
**dentry
, struct vfsmount
**mnt
)
310 struct task_struct
*task
= get_proc_task(inode
);
311 struct files_struct
*files
= NULL
;
313 int fd
= proc_fd(inode
);
316 files
= get_files_struct(task
);
317 put_task_struct(task
);
321 * We are not taking a ref to the file structure, so we must
324 spin_lock(&files
->file_lock
);
325 file
= fcheck_files(files
, fd
);
327 *mnt
= mntget(file
->f_vfsmnt
);
328 *dentry
= dget(file
->f_dentry
);
329 spin_unlock(&files
->file_lock
);
330 put_files_struct(files
);
333 spin_unlock(&files
->file_lock
);
334 put_files_struct(files
);
339 static struct fs_struct
*get_fs_struct(struct task_struct
*task
)
341 struct fs_struct
*fs
;
345 atomic_inc(&fs
->count
);
350 static int get_nr_threads(struct task_struct
*tsk
)
352 /* Must be called with the rcu_read_lock held */
356 if (lock_task_sighand(tsk
, &flags
)) {
357 count
= atomic_read(&tsk
->signal
->count
);
358 unlock_task_sighand(tsk
, &flags
);
363 static int proc_cwd_link(struct inode
*inode
, struct dentry
**dentry
, struct vfsmount
**mnt
)
365 struct task_struct
*task
= get_proc_task(inode
);
366 struct fs_struct
*fs
= NULL
;
367 int result
= -ENOENT
;
370 fs
= get_fs_struct(task
);
371 put_task_struct(task
);
374 read_lock(&fs
->lock
);
375 *mnt
= mntget(fs
->pwdmnt
);
376 *dentry
= dget(fs
->pwd
);
377 read_unlock(&fs
->lock
);
384 static int proc_root_link(struct inode
*inode
, struct dentry
**dentry
, struct vfsmount
**mnt
)
386 struct task_struct
*task
= get_proc_task(inode
);
387 struct fs_struct
*fs
= NULL
;
388 int result
= -ENOENT
;
391 fs
= get_fs_struct(task
);
392 put_task_struct(task
);
395 read_lock(&fs
->lock
);
396 *mnt
= mntget(fs
->rootmnt
);
397 *dentry
= dget(fs
->root
);
398 read_unlock(&fs
->lock
);
405 #define MAY_PTRACE(task) \
406 (task == current || \
407 (task->parent == current && \
408 (task->ptrace & PT_PTRACED) && \
409 (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
410 security_ptrace(current,task) == 0))
412 static int proc_pid_environ(struct task_struct
*task
, char * buffer
)
415 struct mm_struct
*mm
= get_task_mm(task
);
417 unsigned int len
= mm
->env_end
- mm
->env_start
;
420 res
= access_process_vm(task
, mm
->env_start
, buffer
, len
, 0);
421 if (!ptrace_may_attach(task
))
428 static int proc_pid_cmdline(struct task_struct
*task
, char * buffer
)
432 struct mm_struct
*mm
= get_task_mm(task
);
436 goto out_mm
; /* Shh! No looking before we're done */
438 len
= mm
->arg_end
- mm
->arg_start
;
443 res
= access_process_vm(task
, mm
->arg_start
, buffer
, len
, 0);
445 // If the nul at the end of args has been overwritten, then
446 // assume application is using setproctitle(3).
447 if (res
> 0 && buffer
[res
-1] != '\0' && len
< PAGE_SIZE
) {
448 len
= strnlen(buffer
, res
);
452 len
= mm
->env_end
- mm
->env_start
;
453 if (len
> PAGE_SIZE
- res
)
454 len
= PAGE_SIZE
- res
;
455 res
+= access_process_vm(task
, mm
->env_start
, buffer
+res
, len
, 0);
456 res
= strnlen(buffer
, res
);
465 static int proc_pid_auxv(struct task_struct
*task
, char *buffer
)
468 struct mm_struct
*mm
= get_task_mm(task
);
470 unsigned int nwords
= 0;
473 while (mm
->saved_auxv
[nwords
- 2] != 0); /* AT_NULL */
474 res
= nwords
* sizeof(mm
->saved_auxv
[0]);
477 memcpy(buffer
, mm
->saved_auxv
, res
);
484 #ifdef CONFIG_KALLSYMS
486 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
487 * Returns the resolved symbol. If that fails, simply return the address.
489 static int proc_pid_wchan(struct task_struct
*task
, char *buffer
)
492 const char *sym_name
;
493 unsigned long wchan
, size
, offset
;
494 char namebuf
[KSYM_NAME_LEN
+1];
496 wchan
= get_wchan(task
);
498 sym_name
= kallsyms_lookup(wchan
, &size
, &offset
, &modname
, namebuf
);
500 return sprintf(buffer
, "%s", sym_name
);
501 return sprintf(buffer
, "%lu", wchan
);
503 #endif /* CONFIG_KALLSYMS */
505 #ifdef CONFIG_SCHEDSTATS
507 * Provides /proc/PID/schedstat
509 static int proc_pid_schedstat(struct task_struct
*task
, char *buffer
)
511 return sprintf(buffer
, "%lu %lu %lu\n",
512 task
->sched_info
.cpu_time
,
513 task
->sched_info
.run_delay
,
514 task
->sched_info
.pcnt
);
518 /* The badness from the OOM killer */
519 unsigned long badness(struct task_struct
*p
, unsigned long uptime
);
520 static int proc_oom_score(struct task_struct
*task
, char *buffer
)
522 unsigned long points
;
523 struct timespec uptime
;
525 do_posix_clock_monotonic_gettime(&uptime
);
526 points
= badness(task
, uptime
.tv_sec
);
527 return sprintf(buffer
, "%lu\n", points
);
530 /************************************************************************/
531 /* Here the fs part begins */
532 /************************************************************************/
534 /* permission checks */
536 /* If the process being read is separated by chroot from the reading process,
537 * don't let the reader access the threads.
539 static int proc_check_chroot(struct dentry
*de
, struct vfsmount
*mnt
)
542 struct vfsmount
*our_vfsmnt
;
545 read_lock(¤t
->fs
->lock
);
546 our_vfsmnt
= mntget(current
->fs
->rootmnt
);
547 base
= dget(current
->fs
->root
);
548 read_unlock(¤t
->fs
->lock
);
550 spin_lock(&vfsmount_lock
);
552 while (mnt
!= our_vfsmnt
) {
553 if (mnt
== mnt
->mnt_parent
)
555 de
= mnt
->mnt_mountpoint
;
556 mnt
= mnt
->mnt_parent
;
559 if (!is_subdir(de
, base
))
561 spin_unlock(&vfsmount_lock
);
568 spin_unlock(&vfsmount_lock
);
573 extern struct seq_operations mounts_op
;
579 static int mounts_open(struct inode
*inode
, struct file
*file
)
581 struct task_struct
*task
= get_proc_task(inode
);
582 struct namespace *namespace = NULL
;
583 struct proc_mounts
*p
;
588 namespace = task
->namespace;
590 get_namespace(namespace);
592 put_task_struct(task
);
597 p
= kmalloc(sizeof(struct proc_mounts
), GFP_KERNEL
);
599 file
->private_data
= &p
->m
;
600 ret
= seq_open(file
, &mounts_op
);
602 p
->m
.private = namespace;
603 p
->event
= namespace->event
;
608 put_namespace(namespace);
613 static int mounts_release(struct inode
*inode
, struct file
*file
)
615 struct seq_file
*m
= file
->private_data
;
616 struct namespace *namespace = m
->private;
617 put_namespace(namespace);
618 return seq_release(inode
, file
);
621 static unsigned mounts_poll(struct file
*file
, poll_table
*wait
)
623 struct proc_mounts
*p
= file
->private_data
;
624 struct namespace *ns
= p
->m
.private;
627 poll_wait(file
, &ns
->poll
, wait
);
629 spin_lock(&vfsmount_lock
);
630 if (p
->event
!= ns
->event
) {
631 p
->event
= ns
->event
;
634 spin_unlock(&vfsmount_lock
);
639 static struct file_operations proc_mounts_operations
= {
643 .release
= mounts_release
,
647 extern struct seq_operations mountstats_op
;
648 static int mountstats_open(struct inode
*inode
, struct file
*file
)
650 int ret
= seq_open(file
, &mountstats_op
);
653 struct seq_file
*m
= file
->private_data
;
654 struct namespace *namespace = NULL
;
655 struct task_struct
*task
= get_proc_task(inode
);
659 namespace = task
->namespace;
661 get_namespace(namespace);
663 put_task_struct(task
);
667 m
->private = namespace;
669 seq_release(inode
, file
);
676 static struct file_operations proc_mountstats_operations
= {
677 .open
= mountstats_open
,
680 .release
= mounts_release
,
683 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
685 static ssize_t
proc_info_read(struct file
* file
, char __user
* buf
,
686 size_t count
, loff_t
*ppos
)
688 struct inode
* inode
= file
->f_dentry
->d_inode
;
691 struct task_struct
*task
= get_proc_task(inode
);
697 if (count
> PROC_BLOCK_SIZE
)
698 count
= PROC_BLOCK_SIZE
;
701 if (!(page
= __get_free_page(GFP_KERNEL
)))
704 length
= PROC_I(inode
)->op
.proc_read(task
, (char*)page
);
707 length
= simple_read_from_buffer(buf
, count
, ppos
, (char *)page
, length
);
710 put_task_struct(task
);
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
= get_proc_task(file
->f_dentry
->d_inode
);
730 unsigned long src
= *ppos
;
732 struct mm_struct
*mm
;
737 if (!MAY_PTRACE(task
) || !ptrace_may_attach(task
))
741 page
= (char *)__get_free_page(GFP_USER
);
747 mm
= get_task_mm(task
);
753 if (file
->private_data
!= (void*)((long)current
->self_exec_id
))
759 int this_len
, retval
;
761 this_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
762 retval
= access_process_vm(task
, src
, page
, this_len
, 0);
763 if (!retval
|| !MAY_PTRACE(task
) || !ptrace_may_attach(task
)) {
769 if (copy_to_user(buf
, page
, retval
)) {
784 free_page((unsigned long) page
);
786 put_task_struct(task
);
791 #define mem_write NULL
794 /* This is a security hazard */
795 static ssize_t
mem_write(struct file
* file
, const char * buf
,
796 size_t count
, loff_t
*ppos
)
800 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
801 unsigned long dst
= *ppos
;
807 if (!MAY_PTRACE(task
) || !ptrace_may_attach(task
))
811 page
= (char *)__get_free_page(GFP_USER
);
816 int this_len
, retval
;
818 this_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
819 if (copy_from_user(page
, buf
, this_len
)) {
823 retval
= access_process_vm(task
, dst
, page
, this_len
, 1);
835 free_page((unsigned long) page
);
837 put_task_struct(task
);
843 static loff_t
mem_lseek(struct file
* file
, loff_t offset
, int orig
)
847 file
->f_pos
= offset
;
850 file
->f_pos
+= offset
;
855 force_successful_syscall_return();
859 static struct file_operations proc_mem_operations
= {
866 static ssize_t
oom_adjust_read(struct file
*file
, char __user
*buf
,
867 size_t count
, loff_t
*ppos
)
869 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
870 char buffer
[PROC_NUMBUF
];
873 loff_t __ppos
= *ppos
;
877 oom_adjust
= task
->oomkilladj
;
878 put_task_struct(task
);
880 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", oom_adjust
);
883 if (count
> len
-__ppos
)
885 if (copy_to_user(buf
, buffer
+ __ppos
, count
))
887 *ppos
= __ppos
+ count
;
891 static ssize_t
oom_adjust_write(struct file
*file
, const char __user
*buf
,
892 size_t count
, loff_t
*ppos
)
894 struct task_struct
*task
;
895 char buffer
[PROC_NUMBUF
], *end
;
898 if (!capable(CAP_SYS_RESOURCE
))
900 memset(buffer
, 0, sizeof(buffer
));
901 if (count
> sizeof(buffer
) - 1)
902 count
= sizeof(buffer
) - 1;
903 if (copy_from_user(buffer
, buf
, count
))
905 oom_adjust
= simple_strtol(buffer
, &end
, 0);
906 if ((oom_adjust
< -16 || oom_adjust
> 15) && oom_adjust
!= OOM_DISABLE
)
910 task
= get_proc_task(file
->f_dentry
->d_inode
);
913 task
->oomkilladj
= oom_adjust
;
914 put_task_struct(task
);
915 if (end
- buffer
== 0)
920 static struct file_operations proc_oom_adjust_operations
= {
921 .read
= oom_adjust_read
,
922 .write
= oom_adjust_write
,
925 #ifdef CONFIG_AUDITSYSCALL
927 static ssize_t
proc_loginuid_read(struct file
* file
, char __user
* buf
,
928 size_t count
, loff_t
*ppos
)
930 struct inode
* inode
= file
->f_dentry
->d_inode
;
931 struct task_struct
*task
= get_proc_task(inode
);
933 char tmpbuf
[TMPBUFLEN
];
937 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
938 audit_get_loginuid(task
->audit_context
));
939 put_task_struct(task
);
940 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
943 static ssize_t
proc_loginuid_write(struct file
* file
, const char __user
* buf
,
944 size_t count
, loff_t
*ppos
)
946 struct inode
* inode
= file
->f_dentry
->d_inode
;
951 if (!capable(CAP_AUDIT_CONTROL
))
954 if (current
!= pid_task(proc_pid(inode
), PIDTYPE_PID
))
957 if (count
>= PAGE_SIZE
)
958 count
= PAGE_SIZE
- 1;
961 /* No partial writes. */
964 page
= (char*)__get_free_page(GFP_USER
);
968 if (copy_from_user(page
, buf
, count
))
972 loginuid
= simple_strtoul(page
, &tmp
, 10);
978 length
= audit_set_loginuid(current
, loginuid
);
979 if (likely(length
== 0))
983 free_page((unsigned long) page
);
987 static struct file_operations proc_loginuid_operations
= {
988 .read
= proc_loginuid_read
,
989 .write
= proc_loginuid_write
,
993 #ifdef CONFIG_SECCOMP
994 static ssize_t
seccomp_read(struct file
*file
, char __user
*buf
,
995 size_t count
, loff_t
*ppos
)
997 struct task_struct
*tsk
= get_proc_task(file
->f_dentry
->d_inode
);
999 loff_t __ppos
= *ppos
;
1004 /* no need to print the trailing zero, so use only len */
1005 len
= sprintf(__buf
, "%u\n", tsk
->seccomp
.mode
);
1006 put_task_struct(tsk
);
1009 if (count
> len
- __ppos
)
1010 count
= len
- __ppos
;
1011 if (copy_to_user(buf
, __buf
+ __ppos
, count
))
1013 *ppos
= __ppos
+ count
;
1017 static ssize_t
seccomp_write(struct file
*file
, const char __user
*buf
,
1018 size_t count
, loff_t
*ppos
)
1020 struct task_struct
*tsk
= get_proc_task(file
->f_dentry
->d_inode
);
1021 char __buf
[20], *end
;
1022 unsigned int seccomp_mode
;
1029 /* can set it only once to be even more secure */
1031 if (unlikely(tsk
->seccomp
.mode
))
1035 memset(__buf
, 0, sizeof(__buf
));
1036 count
= min(count
, sizeof(__buf
) - 1);
1037 if (copy_from_user(__buf
, buf
, count
))
1040 seccomp_mode
= simple_strtoul(__buf
, &end
, 0);
1044 if (seccomp_mode
&& seccomp_mode
<= NR_SECCOMP_MODES
) {
1045 tsk
->seccomp
.mode
= seccomp_mode
;
1046 set_tsk_thread_flag(tsk
, TIF_SECCOMP
);
1050 if (unlikely(!(end
- __buf
)))
1052 result
= end
- __buf
;
1054 put_task_struct(tsk
);
1059 static struct file_operations proc_seccomp_operations
= {
1060 .read
= seccomp_read
,
1061 .write
= seccomp_write
,
1063 #endif /* CONFIG_SECCOMP */
1065 static int proc_check_dentry_visible(struct inode
*inode
,
1066 struct dentry
*dentry
, struct vfsmount
*mnt
)
1068 /* Verify that the current process can already see the
1069 * file pointed at by the file descriptor.
1070 * This prevents /proc from being an accidental information leak.
1072 * This prevents access to files that are not visible do to
1073 * being on the otherside of a chroot, in a different
1074 * namespace, or are simply process local (like pipes).
1076 struct task_struct
*task
;
1077 int error
= -EACCES
;
1079 /* See if the the two tasks share a commone set of
1080 * file descriptors. If so everything is visible.
1083 task
= tref_task(proc_tref(inode
));
1085 struct files_struct
*task_files
, *files
;
1086 /* This test answeres the question:
1087 * Is there a point in time since we looked up the
1088 * file descriptor where the two tasks share the
1089 * same files struct?
1092 files
= current
->files
;
1093 task_files
= task
->files
;
1094 if (files
&& (files
== task_files
))
1101 /* If the two tasks don't share a common set of file
1102 * descriptors see if the destination dentry is already
1103 * visible in the current tasks filesystem namespace.
1105 error
= proc_check_chroot(dentry
, mnt
);
1111 static void *proc_pid_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
1113 struct inode
*inode
= dentry
->d_inode
;
1114 int error
= -EACCES
;
1116 /* We don't need a base pointer in the /proc filesystem */
1119 if (current
->fsuid
!= inode
->i_uid
&& !capable(CAP_DAC_OVERRIDE
))
1122 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &nd
->dentry
, &nd
->mnt
);
1123 nd
->last_type
= LAST_BIND
;
1127 /* Only return files this task can already see */
1128 error
= proc_check_dentry_visible(inode
, nd
->dentry
, nd
->mnt
);
1132 return ERR_PTR(error
);
1135 static int do_proc_readlink(struct dentry
*dentry
, struct vfsmount
*mnt
,
1136 char __user
*buffer
, int buflen
)
1138 struct inode
* inode
;
1139 char *tmp
= (char*)__get_free_page(GFP_KERNEL
), *path
;
1145 inode
= dentry
->d_inode
;
1146 path
= d_path(dentry
, mnt
, tmp
, PAGE_SIZE
);
1147 len
= PTR_ERR(path
);
1150 len
= tmp
+ PAGE_SIZE
- 1 - path
;
1154 if (copy_to_user(buffer
, path
, len
))
1157 free_page((unsigned long)tmp
);
1161 static int proc_pid_readlink(struct dentry
* dentry
, char __user
* buffer
, int buflen
)
1163 int error
= -EACCES
;
1164 struct inode
*inode
= dentry
->d_inode
;
1166 struct vfsmount
*mnt
= NULL
;
1169 if (current
->fsuid
!= inode
->i_uid
&& !capable(CAP_DAC_OVERRIDE
))
1172 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &de
, &mnt
);
1176 /* Only return files this task can already see */
1177 error
= proc_check_dentry_visible(inode
, de
, mnt
);
1181 error
= do_proc_readlink(de
, mnt
, buffer
, buflen
);
1189 static struct inode_operations proc_pid_link_inode_operations
= {
1190 .readlink
= proc_pid_readlink
,
1191 .follow_link
= proc_pid_follow_link
1194 static int proc_readfd(struct file
* filp
, void * dirent
, filldir_t filldir
)
1196 struct dentry
*dentry
= filp
->f_dentry
;
1197 struct inode
*inode
= dentry
->d_inode
;
1198 struct task_struct
*p
= get_proc_task(inode
);
1199 unsigned int fd
, tid
, ino
;
1201 char buf
[PROC_NUMBUF
];
1202 struct files_struct
* files
;
1203 struct fdtable
*fdt
;
1214 if (filldir(dirent
, ".", 1, 0, inode
->i_ino
, DT_DIR
) < 0)
1218 ino
= parent_ino(dentry
);
1219 if (filldir(dirent
, "..", 2, 1, ino
, DT_DIR
) < 0)
1223 files
= get_files_struct(p
);
1227 fdt
= files_fdtable(files
);
1228 for (fd
= filp
->f_pos
-2;
1230 fd
++, filp
->f_pos
++) {
1233 if (!fcheck_files(files
, fd
))
1241 buf
[j
] = '0' + (i
% 10);
1245 ino
= fake_ino(tid
, PROC_TID_FD_DIR
+ fd
);
1246 if (filldir(dirent
, buf
+j
, PROC_NUMBUF
-j
, fd
+2, ino
, DT_LNK
) < 0) {
1253 put_files_struct(files
);
1261 static int proc_pident_readdir(struct file
*filp
,
1262 void *dirent
, filldir_t filldir
,
1263 struct pid_entry
*ents
, unsigned int nents
)
1267 struct dentry
*dentry
= filp
->f_dentry
;
1268 struct inode
*inode
= dentry
->d_inode
;
1269 struct task_struct
*task
= get_proc_task(inode
);
1270 struct pid_entry
*p
;
1280 put_task_struct(task
);
1285 if (filldir(dirent
, ".", 1, i
, ino
, DT_DIR
) < 0)
1291 ino
= parent_ino(dentry
);
1292 if (filldir(dirent
, "..", 2, i
, ino
, DT_DIR
) < 0)
1305 if (filldir(dirent
, p
->name
, p
->len
, filp
->f_pos
,
1306 fake_ino(pid
, p
->type
), p
->mode
>> 12) < 0)
1318 static int proc_tgid_base_readdir(struct file
* filp
,
1319 void * dirent
, filldir_t filldir
)
1321 return proc_pident_readdir(filp
,dirent
,filldir
,
1322 tgid_base_stuff
,ARRAY_SIZE(tgid_base_stuff
));
1325 static int proc_tid_base_readdir(struct file
* filp
,
1326 void * dirent
, filldir_t filldir
)
1328 return proc_pident_readdir(filp
,dirent
,filldir
,
1329 tid_base_stuff
,ARRAY_SIZE(tid_base_stuff
));
1332 /* building an inode */
1334 static int task_dumpable(struct task_struct
*task
)
1337 struct mm_struct
*mm
;
1342 dumpable
= mm
->dumpable
;
1350 static struct inode
*proc_pid_make_inode(struct super_block
* sb
, struct task_struct
*task
, int ino
)
1352 struct inode
* inode
;
1353 struct proc_inode
*ei
;
1355 /* We need a new inode */
1357 inode
= new_inode(sb
);
1363 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1364 inode
->i_ino
= fake_ino(task
->pid
, ino
);
1367 * grab the reference to task.
1369 ei
->pid
= get_pid(task
->pids
[PIDTYPE_PID
].pid
);
1375 if (task_dumpable(task
)) {
1376 inode
->i_uid
= task
->euid
;
1377 inode
->i_gid
= task
->egid
;
1379 security_task_to_inode(task
, inode
);
1392 * Exceptional case: normally we are not allowed to unhash a busy
1393 * directory. In this case, however, we can do it - no aliasing problems
1394 * due to the way we treat inodes.
1396 * Rewrite the inode's ownerships here because the owning task may have
1397 * performed a setuid(), etc.
1399 * Before the /proc/pid/status file was created the only way to read
1400 * the effective uid of a /process was to stat /proc/pid. Reading
1401 * /proc/pid/status is slow enough that procps and other packages
1402 * kept stating /proc/pid. To keep the rules in /proc simple I have
1403 * made this apply to all per process world readable and executable
1406 static int pid_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1408 struct inode
*inode
= dentry
->d_inode
;
1409 struct task_struct
*task
= get_proc_task(inode
);
1411 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1412 task_dumpable(task
)) {
1413 inode
->i_uid
= task
->euid
;
1414 inode
->i_gid
= task
->egid
;
1419 security_task_to_inode(task
, inode
);
1420 put_task_struct(task
);
1427 static int pid_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
1429 struct inode
*inode
= dentry
->d_inode
;
1430 struct task_struct
*task
;
1431 generic_fillattr(inode
, stat
);
1436 task
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
1438 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1439 task_dumpable(task
)) {
1440 stat
->uid
= task
->euid
;
1441 stat
->gid
= task
->egid
;
1448 static int tid_fd_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1450 struct inode
*inode
= dentry
->d_inode
;
1451 struct task_struct
*task
= get_proc_task(inode
);
1452 int fd
= proc_fd(inode
);
1453 struct files_struct
*files
;
1456 files
= get_files_struct(task
);
1459 if (fcheck_files(files
, fd
)) {
1461 put_files_struct(files
);
1462 if (task_dumpable(task
)) {
1463 inode
->i_uid
= task
->euid
;
1464 inode
->i_gid
= task
->egid
;
1469 security_task_to_inode(task
, inode
);
1470 put_task_struct(task
);
1474 put_files_struct(files
);
1476 put_task_struct(task
);
1482 static int pid_delete_dentry(struct dentry
* dentry
)
1484 /* Is the task we represent dead?
1485 * If so, then don't put the dentry on the lru list,
1486 * kill it immediately.
1488 return !proc_pid(dentry
->d_inode
)->tasks
[PIDTYPE_PID
].first
;
1491 static struct dentry_operations tid_fd_dentry_operations
=
1493 .d_revalidate
= tid_fd_revalidate
,
1494 .d_delete
= pid_delete_dentry
,
1497 static struct dentry_operations pid_dentry_operations
=
1499 .d_revalidate
= pid_revalidate
,
1500 .d_delete
= pid_delete_dentry
,
1505 static unsigned name_to_int(struct dentry
*dentry
)
1507 const char *name
= dentry
->d_name
.name
;
1508 int len
= dentry
->d_name
.len
;
1511 if (len
> 1 && *name
== '0')
1514 unsigned c
= *name
++ - '0';
1517 if (n
>= (~0U-9)/10)
1528 static struct dentry
*proc_lookupfd(struct inode
* dir
, struct dentry
* dentry
, struct nameidata
*nd
)
1530 struct task_struct
*task
= get_proc_task(dir
);
1531 unsigned fd
= name_to_int(dentry
);
1532 struct dentry
*result
= ERR_PTR(-ENOENT
);
1534 struct files_struct
* files
;
1535 struct inode
*inode
;
1536 struct proc_inode
*ei
;
1543 inode
= proc_pid_make_inode(dir
->i_sb
, task
, PROC_TID_FD_DIR
+fd
);
1548 files
= get_files_struct(task
);
1551 inode
->i_mode
= S_IFLNK
;
1554 * We are not taking a ref to the file structure, so we must
1557 spin_lock(&files
->file_lock
);
1558 file
= fcheck_files(files
, fd
);
1561 if (file
->f_mode
& 1)
1562 inode
->i_mode
|= S_IRUSR
| S_IXUSR
;
1563 if (file
->f_mode
& 2)
1564 inode
->i_mode
|= S_IWUSR
| S_IXUSR
;
1565 spin_unlock(&files
->file_lock
);
1566 put_files_struct(files
);
1567 inode
->i_op
= &proc_pid_link_inode_operations
;
1569 ei
->op
.proc_get_link
= proc_fd_link
;
1570 dentry
->d_op
= &tid_fd_dentry_operations
;
1571 d_add(dentry
, inode
);
1572 /* Close the race of the process dying before we return the dentry */
1573 if (tid_fd_revalidate(dentry
, NULL
))
1576 put_task_struct(task
);
1581 spin_unlock(&files
->file_lock
);
1582 put_files_struct(files
);
1588 static int proc_task_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
);
1589 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
);
1590 static int proc_task_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
);
1592 static struct file_operations proc_fd_operations
= {
1593 .read
= generic_read_dir
,
1594 .readdir
= proc_readfd
,
1597 static struct file_operations proc_task_operations
= {
1598 .read
= generic_read_dir
,
1599 .readdir
= proc_task_readdir
,
1603 * proc directories can do almost nothing..
1605 static struct inode_operations proc_fd_inode_operations
= {
1606 .lookup
= proc_lookupfd
,
1609 static struct inode_operations proc_task_inode_operations
= {
1610 .lookup
= proc_task_lookup
,
1611 .getattr
= proc_task_getattr
,
1614 #ifdef CONFIG_SECURITY
1615 static ssize_t
proc_pid_attr_read(struct file
* file
, char __user
* buf
,
1616 size_t count
, loff_t
*ppos
)
1618 struct inode
* inode
= file
->f_dentry
->d_inode
;
1621 struct task_struct
*task
= get_proc_task(inode
);
1627 if (count
> PAGE_SIZE
)
1630 if (!(page
= __get_free_page(GFP_KERNEL
)))
1633 length
= security_getprocattr(task
,
1634 (char*)file
->f_dentry
->d_name
.name
,
1635 (void*)page
, count
);
1637 length
= simple_read_from_buffer(buf
, count
, ppos
, (char *)page
, length
);
1640 put_task_struct(task
);
1645 static ssize_t
proc_pid_attr_write(struct file
* file
, const char __user
* buf
,
1646 size_t count
, loff_t
*ppos
)
1648 struct inode
* inode
= file
->f_dentry
->d_inode
;
1651 struct task_struct
*task
= get_proc_task(inode
);
1656 if (count
> PAGE_SIZE
)
1659 /* No partial writes. */
1665 page
= (char*)__get_free_page(GFP_USER
);
1670 if (copy_from_user(page
, buf
, count
))
1673 length
= security_setprocattr(task
,
1674 (char*)file
->f_dentry
->d_name
.name
,
1675 (void*)page
, count
);
1677 free_page((unsigned long) page
);
1679 put_task_struct(task
);
1684 static struct file_operations proc_pid_attr_operations
= {
1685 .read
= proc_pid_attr_read
,
1686 .write
= proc_pid_attr_write
,
1689 static struct file_operations proc_tid_attr_operations
;
1690 static struct inode_operations proc_tid_attr_inode_operations
;
1691 static struct file_operations proc_tgid_attr_operations
;
1692 static struct inode_operations proc_tgid_attr_inode_operations
;
1696 static struct dentry
*proc_pident_lookup(struct inode
*dir
,
1697 struct dentry
*dentry
,
1698 struct pid_entry
*ents
)
1700 struct inode
*inode
;
1701 struct dentry
*error
;
1702 struct task_struct
*task
= get_proc_task(dir
);
1703 struct pid_entry
*p
;
1704 struct proc_inode
*ei
;
1706 error
= ERR_PTR(-ENOENT
);
1712 for (p
= ents
; p
->name
; p
++) {
1713 if (p
->len
!= dentry
->d_name
.len
)
1715 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
1721 error
= ERR_PTR(-EINVAL
);
1722 inode
= proc_pid_make_inode(dir
->i_sb
, task
, p
->type
);
1727 inode
->i_mode
= p
->mode
;
1729 * Yes, it does not scale. And it should not. Don't add
1730 * new entries into /proc/<tgid>/ without very good reasons.
1733 case PROC_TGID_TASK
:
1735 inode
->i_op
= &proc_task_inode_operations
;
1736 inode
->i_fop
= &proc_task_operations
;
1741 inode
->i_op
= &proc_fd_inode_operations
;
1742 inode
->i_fop
= &proc_fd_operations
;
1746 inode
->i_op
= &proc_pid_link_inode_operations
;
1747 ei
->op
.proc_get_link
= proc_exe_link
;
1751 inode
->i_op
= &proc_pid_link_inode_operations
;
1752 ei
->op
.proc_get_link
= proc_cwd_link
;
1755 case PROC_TGID_ROOT
:
1756 inode
->i_op
= &proc_pid_link_inode_operations
;
1757 ei
->op
.proc_get_link
= proc_root_link
;
1759 case PROC_TID_ENVIRON
:
1760 case PROC_TGID_ENVIRON
:
1761 inode
->i_fop
= &proc_info_file_operations
;
1762 ei
->op
.proc_read
= proc_pid_environ
;
1765 case PROC_TGID_AUXV
:
1766 inode
->i_fop
= &proc_info_file_operations
;
1767 ei
->op
.proc_read
= proc_pid_auxv
;
1769 case PROC_TID_STATUS
:
1770 case PROC_TGID_STATUS
:
1771 inode
->i_fop
= &proc_info_file_operations
;
1772 ei
->op
.proc_read
= proc_pid_status
;
1775 inode
->i_fop
= &proc_info_file_operations
;
1776 ei
->op
.proc_read
= proc_tid_stat
;
1778 case PROC_TGID_STAT
:
1779 inode
->i_fop
= &proc_info_file_operations
;
1780 ei
->op
.proc_read
= proc_tgid_stat
;
1782 case PROC_TID_CMDLINE
:
1783 case PROC_TGID_CMDLINE
:
1784 inode
->i_fop
= &proc_info_file_operations
;
1785 ei
->op
.proc_read
= proc_pid_cmdline
;
1787 case PROC_TID_STATM
:
1788 case PROC_TGID_STATM
:
1789 inode
->i_fop
= &proc_info_file_operations
;
1790 ei
->op
.proc_read
= proc_pid_statm
;
1793 case PROC_TGID_MAPS
:
1794 inode
->i_fop
= &proc_maps_operations
;
1797 case PROC_TID_NUMA_MAPS
:
1798 case PROC_TGID_NUMA_MAPS
:
1799 inode
->i_fop
= &proc_numa_maps_operations
;
1804 inode
->i_fop
= &proc_mem_operations
;
1806 #ifdef CONFIG_SECCOMP
1807 case PROC_TID_SECCOMP
:
1808 case PROC_TGID_SECCOMP
:
1809 inode
->i_fop
= &proc_seccomp_operations
;
1811 #endif /* CONFIG_SECCOMP */
1812 case PROC_TID_MOUNTS
:
1813 case PROC_TGID_MOUNTS
:
1814 inode
->i_fop
= &proc_mounts_operations
;
1817 case PROC_TID_SMAPS
:
1818 case PROC_TGID_SMAPS
:
1819 inode
->i_fop
= &proc_smaps_operations
;
1822 case PROC_TID_MOUNTSTATS
:
1823 case PROC_TGID_MOUNTSTATS
:
1824 inode
->i_fop
= &proc_mountstats_operations
;
1826 #ifdef CONFIG_SECURITY
1829 inode
->i_op
= &proc_tid_attr_inode_operations
;
1830 inode
->i_fop
= &proc_tid_attr_operations
;
1832 case PROC_TGID_ATTR
:
1834 inode
->i_op
= &proc_tgid_attr_inode_operations
;
1835 inode
->i_fop
= &proc_tgid_attr_operations
;
1837 case PROC_TID_ATTR_CURRENT
:
1838 case PROC_TGID_ATTR_CURRENT
:
1839 case PROC_TID_ATTR_PREV
:
1840 case PROC_TGID_ATTR_PREV
:
1841 case PROC_TID_ATTR_EXEC
:
1842 case PROC_TGID_ATTR_EXEC
:
1843 case PROC_TID_ATTR_FSCREATE
:
1844 case PROC_TGID_ATTR_FSCREATE
:
1845 case PROC_TID_ATTR_KEYCREATE
:
1846 case PROC_TGID_ATTR_KEYCREATE
:
1847 inode
->i_fop
= &proc_pid_attr_operations
;
1850 #ifdef CONFIG_KALLSYMS
1851 case PROC_TID_WCHAN
:
1852 case PROC_TGID_WCHAN
:
1853 inode
->i_fop
= &proc_info_file_operations
;
1854 ei
->op
.proc_read
= proc_pid_wchan
;
1857 #ifdef CONFIG_SCHEDSTATS
1858 case PROC_TID_SCHEDSTAT
:
1859 case PROC_TGID_SCHEDSTAT
:
1860 inode
->i_fop
= &proc_info_file_operations
;
1861 ei
->op
.proc_read
= proc_pid_schedstat
;
1864 #ifdef CONFIG_CPUSETS
1865 case PROC_TID_CPUSET
:
1866 case PROC_TGID_CPUSET
:
1867 inode
->i_fop
= &proc_cpuset_operations
;
1870 case PROC_TID_OOM_SCORE
:
1871 case PROC_TGID_OOM_SCORE
:
1872 inode
->i_fop
= &proc_info_file_operations
;
1873 ei
->op
.proc_read
= proc_oom_score
;
1875 case PROC_TID_OOM_ADJUST
:
1876 case PROC_TGID_OOM_ADJUST
:
1877 inode
->i_fop
= &proc_oom_adjust_operations
;
1879 #ifdef CONFIG_AUDITSYSCALL
1880 case PROC_TID_LOGINUID
:
1881 case PROC_TGID_LOGINUID
:
1882 inode
->i_fop
= &proc_loginuid_operations
;
1886 printk("procfs: impossible type (%d)",p
->type
);
1888 error
= ERR_PTR(-EINVAL
);
1891 dentry
->d_op
= &pid_dentry_operations
;
1892 d_add(dentry
, inode
);
1893 /* Close the race of the process dying before we return the dentry */
1894 if (pid_revalidate(dentry
, NULL
))
1897 put_task_struct(task
);
1902 static struct dentry
*proc_tgid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
1903 return proc_pident_lookup(dir
, dentry
, tgid_base_stuff
);
1906 static struct dentry
*proc_tid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
1907 return proc_pident_lookup(dir
, dentry
, tid_base_stuff
);
1910 static struct file_operations proc_tgid_base_operations
= {
1911 .read
= generic_read_dir
,
1912 .readdir
= proc_tgid_base_readdir
,
1915 static struct file_operations proc_tid_base_operations
= {
1916 .read
= generic_read_dir
,
1917 .readdir
= proc_tid_base_readdir
,
1920 static struct inode_operations proc_tgid_base_inode_operations
= {
1921 .lookup
= proc_tgid_base_lookup
,
1922 .getattr
= pid_getattr
,
1925 static struct inode_operations proc_tid_base_inode_operations
= {
1926 .lookup
= proc_tid_base_lookup
,
1927 .getattr
= pid_getattr
,
1930 #ifdef CONFIG_SECURITY
1931 static int proc_tgid_attr_readdir(struct file
* filp
,
1932 void * dirent
, filldir_t filldir
)
1934 return proc_pident_readdir(filp
,dirent
,filldir
,
1935 tgid_attr_stuff
,ARRAY_SIZE(tgid_attr_stuff
));
1938 static int proc_tid_attr_readdir(struct file
* filp
,
1939 void * dirent
, filldir_t filldir
)
1941 return proc_pident_readdir(filp
,dirent
,filldir
,
1942 tid_attr_stuff
,ARRAY_SIZE(tid_attr_stuff
));
1945 static struct file_operations proc_tgid_attr_operations
= {
1946 .read
= generic_read_dir
,
1947 .readdir
= proc_tgid_attr_readdir
,
1950 static struct file_operations proc_tid_attr_operations
= {
1951 .read
= generic_read_dir
,
1952 .readdir
= proc_tid_attr_readdir
,
1955 static struct dentry
*proc_tgid_attr_lookup(struct inode
*dir
,
1956 struct dentry
*dentry
, struct nameidata
*nd
)
1958 return proc_pident_lookup(dir
, dentry
, tgid_attr_stuff
);
1961 static struct dentry
*proc_tid_attr_lookup(struct inode
*dir
,
1962 struct dentry
*dentry
, struct nameidata
*nd
)
1964 return proc_pident_lookup(dir
, dentry
, tid_attr_stuff
);
1967 static struct inode_operations proc_tgid_attr_inode_operations
= {
1968 .lookup
= proc_tgid_attr_lookup
,
1969 .getattr
= pid_getattr
,
1972 static struct inode_operations proc_tid_attr_inode_operations
= {
1973 .lookup
= proc_tid_attr_lookup
,
1974 .getattr
= pid_getattr
,
1981 static int proc_self_readlink(struct dentry
*dentry
, char __user
*buffer
,
1984 char tmp
[PROC_NUMBUF
];
1985 sprintf(tmp
, "%d", current
->tgid
);
1986 return vfs_readlink(dentry
,buffer
,buflen
,tmp
);
1989 static void *proc_self_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
1991 char tmp
[PROC_NUMBUF
];
1992 sprintf(tmp
, "%d", current
->tgid
);
1993 return ERR_PTR(vfs_follow_link(nd
,tmp
));
1996 static struct inode_operations proc_self_inode_operations
= {
1997 .readlink
= proc_self_readlink
,
1998 .follow_link
= proc_self_follow_link
,
2002 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2004 * @task: task that should be flushed.
2006 * Looks in the dcache for
2008 * /proc/@tgid/task/@pid
2009 * if either directory is present flushes it and all of it'ts children
2012 * It is safe and reasonable to cache /proc entries for a task until
2013 * that task exits. After that they just clog up the dcache with
2014 * useless entries, possibly causing useful dcache entries to be
2015 * flushed instead. This routine is proved to flush those useless
2016 * dcache entries at process exit time.
2018 * NOTE: This routine is just an optimization so it does not guarantee
2019 * that no dcache entries will exist at process exit time it
2020 * just makes it very unlikely that any will persist.
2022 void proc_flush_task(struct task_struct
*task
)
2024 struct dentry
*dentry
, *leader
, *dir
;
2025 char buf
[PROC_NUMBUF
];
2029 name
.len
= snprintf(buf
, sizeof(buf
), "%d", task
->pid
);
2030 dentry
= d_hash_and_lookup(proc_mnt
->mnt_root
, &name
);
2032 shrink_dcache_parent(dentry
);
2037 if (thread_group_leader(task
))
2041 name
.len
= snprintf(buf
, sizeof(buf
), "%d", task
->tgid
);
2042 leader
= d_hash_and_lookup(proc_mnt
->mnt_root
, &name
);
2047 name
.len
= strlen(name
.name
);
2048 dir
= d_hash_and_lookup(leader
, &name
);
2050 goto out_put_leader
;
2053 name
.len
= snprintf(buf
, sizeof(buf
), "%d", task
->pid
);
2054 dentry
= d_hash_and_lookup(dir
, &name
);
2056 shrink_dcache_parent(dentry
);
2069 struct dentry
*proc_pid_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
2071 struct dentry
*result
= ERR_PTR(-ENOENT
);
2072 struct task_struct
*task
;
2073 struct inode
*inode
;
2074 struct proc_inode
*ei
;
2077 if (dentry
->d_name
.len
== 4 && !memcmp(dentry
->d_name
.name
,"self",4)) {
2078 inode
= new_inode(dir
->i_sb
);
2080 return ERR_PTR(-ENOMEM
);
2082 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
2083 inode
->i_ino
= fake_ino(0, PROC_TGID_INO
);
2085 inode
->i_mode
= S_IFLNK
|S_IRWXUGO
;
2086 inode
->i_uid
= inode
->i_gid
= 0;
2088 inode
->i_op
= &proc_self_inode_operations
;
2089 d_add(dentry
, inode
);
2092 tgid
= name_to_int(dentry
);
2097 task
= find_task_by_pid(tgid
);
2099 get_task_struct(task
);
2104 inode
= proc_pid_make_inode(dir
->i_sb
, task
, PROC_TGID_INO
);
2108 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
2109 inode
->i_op
= &proc_tgid_base_inode_operations
;
2110 inode
->i_fop
= &proc_tgid_base_operations
;
2111 inode
->i_flags
|=S_IMMUTABLE
;
2112 #ifdef CONFIG_SECURITY
2118 dentry
->d_op
= &pid_dentry_operations
;
2120 d_add(dentry
, inode
);
2121 /* Close the race of the process dying before we return the dentry */
2122 if (pid_revalidate(dentry
, NULL
))
2126 put_task_struct(task
);
2132 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
2134 struct dentry
*result
= ERR_PTR(-ENOENT
);
2135 struct task_struct
*task
;
2136 struct task_struct
*leader
= get_proc_task(dir
);
2137 struct inode
*inode
;
2143 tid
= name_to_int(dentry
);
2148 task
= find_task_by_pid(tid
);
2150 get_task_struct(task
);
2154 if (leader
->tgid
!= task
->tgid
)
2157 inode
= proc_pid_make_inode(dir
->i_sb
, task
, PROC_TID_INO
);
2162 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
2163 inode
->i_op
= &proc_tid_base_inode_operations
;
2164 inode
->i_fop
= &proc_tid_base_operations
;
2165 inode
->i_flags
|=S_IMMUTABLE
;
2166 #ifdef CONFIG_SECURITY
2172 dentry
->d_op
= &pid_dentry_operations
;
2174 d_add(dentry
, inode
);
2175 /* Close the race of the process dying before we return the dentry */
2176 if (pid_revalidate(dentry
, NULL
))
2180 put_task_struct(task
);
2182 put_task_struct(leader
);
2188 * Find the first tgid to return to user space.
2190 * Usually this is just whatever follows &init_task, but if the users
2191 * buffer was too small to hold the full list or there was a seek into
2192 * the middle of the directory we have more work to do.
2194 * In the case of a short read we start with find_task_by_pid.
2196 * In the case of a seek we start with &init_task and walk nr
2199 static struct task_struct
*first_tgid(int tgid
, unsigned int nr
)
2201 struct task_struct
*pos
;
2204 pos
= find_task_by_pid(tgid
);
2205 if (pos
&& thread_group_leader(pos
))
2208 /* If nr exceeds the number of processes get out quickly */
2210 if (nr
&& nr
>= nr_processes())
2213 /* If we haven't found our starting place yet start with
2214 * the init_task and walk nr tasks forward.
2216 for (pos
= next_task(&init_task
); nr
> 0; --nr
) {
2217 pos
= next_task(pos
);
2218 if (pos
== &init_task
) {
2224 get_task_struct(pos
);
2231 * Find the next task in the task list.
2232 * Return NULL if we loop or there is any error.
2234 * The reference to the input task_struct is released.
2236 static struct task_struct
*next_tgid(struct task_struct
*start
)
2238 struct task_struct
*pos
;
2241 if (pid_alive(start
))
2242 pos
= next_task(start
);
2243 if (pid_alive(pos
) && (pos
!= &init_task
)) {
2244 get_task_struct(pos
);
2250 put_task_struct(start
);
2254 /* for the /proc/ directory itself, after non-process stuff has been done */
2255 int proc_pid_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
2257 char buf
[PROC_NUMBUF
];
2258 unsigned int nr
= filp
->f_pos
- FIRST_PROCESS_ENTRY
;
2259 struct task_struct
*task
;
2263 ino_t ino
= fake_ino(0,PROC_TGID_INO
);
2264 if (filldir(dirent
, "self", 4, filp
->f_pos
, ino
, DT_LNK
) < 0)
2271 /* f_version caches the tgid value that the last readdir call couldn't
2272 * return. lseek aka telldir automagically resets f_version to 0.
2274 tgid
= filp
->f_version
;
2275 filp
->f_version
= 0;
2276 for (task
= first_tgid(tgid
, nr
);
2278 task
= next_tgid(task
), filp
->f_pos
++) {
2282 len
= snprintf(buf
, sizeof(buf
), "%d", tgid
);
2283 ino
= fake_ino(tgid
, PROC_TGID_INO
);
2284 if (filldir(dirent
, buf
, len
, filp
->f_pos
, ino
, DT_DIR
) < 0) {
2285 /* returning this tgid failed, save it as the first
2286 * pid for the next readir call */
2287 filp
->f_version
= tgid
;
2288 put_task_struct(task
);
2296 * Find the first tid of a thread group to return to user space.
2298 * Usually this is just the thread group leader, but if the users
2299 * buffer was too small or there was a seek into the middle of the
2300 * directory we have more work todo.
2302 * In the case of a short read we start with find_task_by_pid.
2304 * In the case of a seek we start with the leader and walk nr
2307 static struct task_struct
*first_tid(struct task_struct
*leader
, int tid
, int nr
)
2309 struct task_struct
*pos
= NULL
;
2310 read_lock(&tasklist_lock
);
2312 /* Attempt to start with the pid of a thread */
2313 if (tid
&& (nr
> 0)) {
2314 pos
= find_task_by_pid(tid
);
2315 if (pos
&& (pos
->group_leader
!= leader
))
2321 /* If nr exceeds the number of threads there is nothing todo */
2323 if (nr
>= get_nr_threads(leader
))
2327 /* If we haven't found our starting place yet start with the
2328 * leader and walk nr threads forward.
2330 if (!pos
&& (nr
>= 0))
2333 for (; pos
&& pid_alive(pos
); pos
= next_thread(pos
)) {
2336 get_task_struct(pos
);
2341 read_unlock(&tasklist_lock
);
2346 * Find the next thread in the thread list.
2347 * Return NULL if there is an error or no next thread.
2349 * The reference to the input task_struct is released.
2351 static struct task_struct
*next_tid(struct task_struct
*start
)
2353 struct task_struct
*pos
;
2354 read_lock(&tasklist_lock
);
2356 if (pid_alive(start
))
2357 pos
= next_thread(start
);
2358 if (pid_alive(pos
) && (pos
!= start
->group_leader
))
2359 get_task_struct(pos
);
2362 read_unlock(&tasklist_lock
);
2363 put_task_struct(start
);
2367 /* for the /proc/TGID/task/ directories */
2368 static int proc_task_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
2370 char buf
[PROC_NUMBUF
];
2371 struct dentry
*dentry
= filp
->f_dentry
;
2372 struct inode
*inode
= dentry
->d_inode
;
2373 struct task_struct
*leader
= get_proc_task(inode
);
2374 struct task_struct
*task
;
2375 int retval
= -ENOENT
;
2378 unsigned long pos
= filp
->f_pos
; /* avoiding "long long" filp->f_pos */
2387 if (filldir(dirent
, ".", 1, pos
, ino
, DT_DIR
) < 0)
2392 ino
= parent_ino(dentry
);
2393 if (filldir(dirent
, "..", 2, pos
, ino
, DT_DIR
) < 0)
2399 /* f_version caches the tgid value that the last readdir call couldn't
2400 * return. lseek aka telldir automagically resets f_version to 0.
2402 tid
= filp
->f_version
;
2403 filp
->f_version
= 0;
2404 for (task
= first_tid(leader
, tid
, pos
- 2);
2406 task
= next_tid(task
), pos
++) {
2409 len
= snprintf(buf
, sizeof(buf
), "%d", tid
);
2410 ino
= fake_ino(tid
, PROC_TID_INO
);
2411 if (filldir(dirent
, buf
, len
, pos
, ino
, DT_DIR
< 0)) {
2412 /* returning this tgid failed, save it as the first
2413 * pid for the next readir call */
2414 filp
->f_version
= tid
;
2415 put_task_struct(task
);
2421 put_task_struct(leader
);
2426 static int proc_task_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
2428 struct inode
*inode
= dentry
->d_inode
;
2429 struct task_struct
*p
= get_proc_task(inode
);
2430 generic_fillattr(inode
, stat
);
2434 stat
->nlink
+= get_nr_threads(p
);