4 * Copyright (C) 1991, 1992 Linus Torvalds
6 * proc base directory handling functions
8 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
9 * Instead of using magical inumbers to determine the kind of object
10 * we allocate and fill in-core inodes upon lookup. They don't even
11 * go into icache. We cache the reference to task_struct upon lookup too.
12 * Eventually it should become a filesystem in its own. We don't use the
13 * rest of procfs anymore.
19 * Bruna Moreira <bruna.moreira@indt.org.br>
20 * Edjard Mota <edjard.mota@indt.org.br>
21 * Ilias Biris <ilias.biris@indt.org.br>
22 * Mauricio Lin <mauricio.lin@indt.org.br>
24 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
26 * A new process specific entry (smaps) included in /proc. It shows the
27 * size of rss for each memory area. The maps entry lacks information
28 * about physical memory size (rss) for each mapped file, i.e.,
29 * rss information for executables and library files.
30 * This additional information is useful for any tools that need to know
31 * about physical memory consumption for a process specific library.
35 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36 * Pud inclusion in the page table walking.
40 * 10LE Instituto Nokia de Tecnologia - INdT:
41 * A better way to walks through the page table as suggested by Hugh Dickins.
43 * Simo Piiroinen <simo.piiroinen@nokia.com>:
44 * Smaps information related to shared, private, clean and dirty pages.
46 * Paul Mundt <paul.mundt@nokia.com>:
47 * Overall revision about smaps.
50 #include <asm/uaccess.h>
52 #include <linux/errno.h>
53 #include <linux/time.h>
54 #include <linux/proc_fs.h>
55 #include <linux/stat.h>
56 #include <linux/init.h>
57 #include <linux/capability.h>
58 #include <linux/file.h>
59 #include <linux/string.h>
60 #include <linux/seq_file.h>
61 #include <linux/namei.h>
62 #include <linux/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>
73 #include <linux/poll.h>
74 #include <linux/nsproxy.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 /* Worst case buffer size needed for holding an integer. */
89 #define PROC_NUMBUF 10
95 struct inode_operations
*iop
;
96 struct file_operations
*fop
;
100 #define NOD(NAME, MODE, IOP, FOP, OP) { \
101 .len = sizeof(NAME) - 1, \
109 #define DIR(NAME, MODE, OTYPE) \
110 NOD(NAME, (S_IFDIR|(MODE)), \
111 &proc_##OTYPE##_inode_operations, &proc_##OTYPE##_operations, \
113 #define LNK(NAME, OTYPE) \
114 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
115 &proc_pid_link_inode_operations, NULL, \
116 { .proc_get_link = &proc_##OTYPE##_link } )
117 #define REG(NAME, MODE, OTYPE) \
118 NOD(NAME, (S_IFREG|(MODE)), NULL, \
119 &proc_##OTYPE##_operations, {})
120 #define INF(NAME, MODE, OTYPE) \
121 NOD(NAME, (S_IFREG|(MODE)), \
122 NULL, &proc_info_file_operations, \
123 { .proc_read = &proc_##OTYPE } )
125 static struct fs_struct
*get_fs_struct(struct task_struct
*task
)
127 struct fs_struct
*fs
;
131 atomic_inc(&fs
->count
);
136 static int get_nr_threads(struct task_struct
*tsk
)
138 /* Must be called with the rcu_read_lock held */
142 if (lock_task_sighand(tsk
, &flags
)) {
143 count
= atomic_read(&tsk
->signal
->count
);
144 unlock_task_sighand(tsk
, &flags
);
149 static int proc_cwd_link(struct inode
*inode
, struct dentry
**dentry
, struct vfsmount
**mnt
)
151 struct task_struct
*task
= get_proc_task(inode
);
152 struct fs_struct
*fs
= NULL
;
153 int result
= -ENOENT
;
156 fs
= get_fs_struct(task
);
157 put_task_struct(task
);
160 read_lock(&fs
->lock
);
161 *mnt
= mntget(fs
->pwdmnt
);
162 *dentry
= dget(fs
->pwd
);
163 read_unlock(&fs
->lock
);
170 static int proc_root_link(struct inode
*inode
, struct dentry
**dentry
, struct vfsmount
**mnt
)
172 struct task_struct
*task
= get_proc_task(inode
);
173 struct fs_struct
*fs
= NULL
;
174 int result
= -ENOENT
;
177 fs
= get_fs_struct(task
);
178 put_task_struct(task
);
181 read_lock(&fs
->lock
);
182 *mnt
= mntget(fs
->rootmnt
);
183 *dentry
= dget(fs
->root
);
184 read_unlock(&fs
->lock
);
191 #define MAY_PTRACE(task) \
192 (task == current || \
193 (task->parent == current && \
194 (task->ptrace & PT_PTRACED) && \
195 (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
196 security_ptrace(current,task) == 0))
198 static int proc_pid_environ(struct task_struct
*task
, char * buffer
)
201 struct mm_struct
*mm
= get_task_mm(task
);
203 unsigned int len
= mm
->env_end
- mm
->env_start
;
206 res
= access_process_vm(task
, mm
->env_start
, buffer
, len
, 0);
207 if (!ptrace_may_attach(task
))
214 static int proc_pid_cmdline(struct task_struct
*task
, char * buffer
)
218 struct mm_struct
*mm
= get_task_mm(task
);
222 goto out_mm
; /* Shh! No looking before we're done */
224 len
= mm
->arg_end
- mm
->arg_start
;
229 res
= access_process_vm(task
, mm
->arg_start
, buffer
, len
, 0);
231 // If the nul at the end of args has been overwritten, then
232 // assume application is using setproctitle(3).
233 if (res
> 0 && buffer
[res
-1] != '\0' && len
< PAGE_SIZE
) {
234 len
= strnlen(buffer
, res
);
238 len
= mm
->env_end
- mm
->env_start
;
239 if (len
> PAGE_SIZE
- res
)
240 len
= PAGE_SIZE
- res
;
241 res
+= access_process_vm(task
, mm
->env_start
, buffer
+res
, len
, 0);
242 res
= strnlen(buffer
, res
);
251 static int proc_pid_auxv(struct task_struct
*task
, char *buffer
)
254 struct mm_struct
*mm
= get_task_mm(task
);
256 unsigned int nwords
= 0;
259 while (mm
->saved_auxv
[nwords
- 2] != 0); /* AT_NULL */
260 res
= nwords
* sizeof(mm
->saved_auxv
[0]);
263 memcpy(buffer
, mm
->saved_auxv
, res
);
270 #ifdef CONFIG_KALLSYMS
272 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
273 * Returns the resolved symbol. If that fails, simply return the address.
275 static int proc_pid_wchan(struct task_struct
*task
, char *buffer
)
278 const char *sym_name
;
279 unsigned long wchan
, size
, offset
;
280 char namebuf
[KSYM_NAME_LEN
+1];
282 wchan
= get_wchan(task
);
284 sym_name
= kallsyms_lookup(wchan
, &size
, &offset
, &modname
, namebuf
);
286 return sprintf(buffer
, "%s", sym_name
);
287 return sprintf(buffer
, "%lu", wchan
);
289 #endif /* CONFIG_KALLSYMS */
291 #ifdef CONFIG_SCHEDSTATS
293 * Provides /proc/PID/schedstat
295 static int proc_pid_schedstat(struct task_struct
*task
, char *buffer
)
297 return sprintf(buffer
, "%lu %lu %lu\n",
298 task
->sched_info
.cpu_time
,
299 task
->sched_info
.run_delay
,
300 task
->sched_info
.pcnt
);
304 /* The badness from the OOM killer */
305 unsigned long badness(struct task_struct
*p
, unsigned long uptime
);
306 static int proc_oom_score(struct task_struct
*task
, char *buffer
)
308 unsigned long points
;
309 struct timespec uptime
;
311 do_posix_clock_monotonic_gettime(&uptime
);
312 points
= badness(task
, uptime
.tv_sec
);
313 return sprintf(buffer
, "%lu\n", points
);
316 /************************************************************************/
317 /* Here the fs part begins */
318 /************************************************************************/
320 /* permission checks */
321 static int proc_fd_access_allowed(struct inode
*inode
)
323 struct task_struct
*task
;
325 /* Allow access to a task's file descriptors if it is us or we
326 * may use ptrace attach to the process and find out that
329 task
= get_proc_task(inode
);
331 allowed
= ptrace_may_attach(task
);
332 put_task_struct(task
);
337 static int proc_setattr(struct dentry
*dentry
, struct iattr
*attr
)
340 struct inode
*inode
= dentry
->d_inode
;
342 if (attr
->ia_valid
& ATTR_MODE
)
345 error
= inode_change_ok(inode
, attr
);
347 error
= security_inode_setattr(dentry
, attr
);
349 error
= inode_setattr(inode
, attr
);
354 static struct inode_operations proc_def_inode_operations
= {
355 .setattr
= proc_setattr
,
358 extern struct seq_operations mounts_op
;
364 static int mounts_open(struct inode
*inode
, struct file
*file
)
366 struct task_struct
*task
= get_proc_task(inode
);
367 struct namespace *namespace = NULL
;
368 struct proc_mounts
*p
;
373 namespace = task
->nsproxy
->namespace;
375 get_namespace(namespace);
377 put_task_struct(task
);
382 p
= kmalloc(sizeof(struct proc_mounts
), GFP_KERNEL
);
384 file
->private_data
= &p
->m
;
385 ret
= seq_open(file
, &mounts_op
);
387 p
->m
.private = namespace;
388 p
->event
= namespace->event
;
393 put_namespace(namespace);
398 static int mounts_release(struct inode
*inode
, struct file
*file
)
400 struct seq_file
*m
= file
->private_data
;
401 struct namespace *namespace = m
->private;
402 put_namespace(namespace);
403 return seq_release(inode
, file
);
406 static unsigned mounts_poll(struct file
*file
, poll_table
*wait
)
408 struct proc_mounts
*p
= file
->private_data
;
409 struct namespace *ns
= p
->m
.private;
412 poll_wait(file
, &ns
->poll
, wait
);
414 spin_lock(&vfsmount_lock
);
415 if (p
->event
!= ns
->event
) {
416 p
->event
= ns
->event
;
419 spin_unlock(&vfsmount_lock
);
424 static struct file_operations proc_mounts_operations
= {
428 .release
= mounts_release
,
432 extern struct seq_operations mountstats_op
;
433 static int mountstats_open(struct inode
*inode
, struct file
*file
)
435 int ret
= seq_open(file
, &mountstats_op
);
438 struct seq_file
*m
= file
->private_data
;
439 struct namespace *namespace = NULL
;
440 struct task_struct
*task
= get_proc_task(inode
);
444 namespace = task
->nsproxy
->namespace;
446 get_namespace(namespace);
448 put_task_struct(task
);
452 m
->private = namespace;
454 seq_release(inode
, file
);
461 static struct file_operations proc_mountstats_operations
= {
462 .open
= mountstats_open
,
465 .release
= mounts_release
,
468 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
470 static ssize_t
proc_info_read(struct file
* file
, char __user
* buf
,
471 size_t count
, loff_t
*ppos
)
473 struct inode
* inode
= file
->f_dentry
->d_inode
;
476 struct task_struct
*task
= get_proc_task(inode
);
482 if (count
> PROC_BLOCK_SIZE
)
483 count
= PROC_BLOCK_SIZE
;
486 if (!(page
= __get_free_page(GFP_KERNEL
)))
489 length
= PROC_I(inode
)->op
.proc_read(task
, (char*)page
);
492 length
= simple_read_from_buffer(buf
, count
, ppos
, (char *)page
, length
);
495 put_task_struct(task
);
500 static struct file_operations proc_info_file_operations
= {
501 .read
= proc_info_read
,
504 static int mem_open(struct inode
* inode
, struct file
* file
)
506 file
->private_data
= (void*)((long)current
->self_exec_id
);
510 static ssize_t
mem_read(struct file
* file
, char __user
* buf
,
511 size_t count
, loff_t
*ppos
)
513 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
515 unsigned long src
= *ppos
;
517 struct mm_struct
*mm
;
522 if (!MAY_PTRACE(task
) || !ptrace_may_attach(task
))
526 page
= (char *)__get_free_page(GFP_USER
);
532 mm
= get_task_mm(task
);
538 if (file
->private_data
!= (void*)((long)current
->self_exec_id
))
544 int this_len
, retval
;
546 this_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
547 retval
= access_process_vm(task
, src
, page
, this_len
, 0);
548 if (!retval
|| !MAY_PTRACE(task
) || !ptrace_may_attach(task
)) {
554 if (copy_to_user(buf
, page
, retval
)) {
569 free_page((unsigned long) page
);
571 put_task_struct(task
);
576 #define mem_write NULL
579 /* This is a security hazard */
580 static ssize_t
mem_write(struct file
* file
, const char * buf
,
581 size_t count
, loff_t
*ppos
)
585 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
586 unsigned long dst
= *ppos
;
592 if (!MAY_PTRACE(task
) || !ptrace_may_attach(task
))
596 page
= (char *)__get_free_page(GFP_USER
);
602 int this_len
, retval
;
604 this_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
605 if (copy_from_user(page
, buf
, this_len
)) {
609 retval
= access_process_vm(task
, dst
, page
, this_len
, 1);
621 free_page((unsigned long) page
);
623 put_task_struct(task
);
629 static loff_t
mem_lseek(struct file
* file
, loff_t offset
, int orig
)
633 file
->f_pos
= offset
;
636 file
->f_pos
+= offset
;
641 force_successful_syscall_return();
645 static struct file_operations proc_mem_operations
= {
652 static ssize_t
oom_adjust_read(struct file
*file
, char __user
*buf
,
653 size_t count
, loff_t
*ppos
)
655 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
656 char buffer
[PROC_NUMBUF
];
659 loff_t __ppos
= *ppos
;
663 oom_adjust
= task
->oomkilladj
;
664 put_task_struct(task
);
666 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", oom_adjust
);
669 if (count
> len
-__ppos
)
671 if (copy_to_user(buf
, buffer
+ __ppos
, count
))
673 *ppos
= __ppos
+ count
;
677 static ssize_t
oom_adjust_write(struct file
*file
, const char __user
*buf
,
678 size_t count
, loff_t
*ppos
)
680 struct task_struct
*task
;
681 char buffer
[PROC_NUMBUF
], *end
;
684 if (!capable(CAP_SYS_RESOURCE
))
686 memset(buffer
, 0, sizeof(buffer
));
687 if (count
> sizeof(buffer
) - 1)
688 count
= sizeof(buffer
) - 1;
689 if (copy_from_user(buffer
, buf
, count
))
691 oom_adjust
= simple_strtol(buffer
, &end
, 0);
692 if ((oom_adjust
< -16 || oom_adjust
> 15) && oom_adjust
!= OOM_DISABLE
)
696 task
= get_proc_task(file
->f_dentry
->d_inode
);
699 task
->oomkilladj
= oom_adjust
;
700 put_task_struct(task
);
701 if (end
- buffer
== 0)
706 static struct file_operations proc_oom_adjust_operations
= {
707 .read
= oom_adjust_read
,
708 .write
= oom_adjust_write
,
711 #ifdef CONFIG_AUDITSYSCALL
713 static ssize_t
proc_loginuid_read(struct file
* file
, char __user
* buf
,
714 size_t count
, loff_t
*ppos
)
716 struct inode
* inode
= file
->f_dentry
->d_inode
;
717 struct task_struct
*task
= get_proc_task(inode
);
719 char tmpbuf
[TMPBUFLEN
];
723 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
724 audit_get_loginuid(task
->audit_context
));
725 put_task_struct(task
);
726 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
729 static ssize_t
proc_loginuid_write(struct file
* file
, const char __user
* buf
,
730 size_t count
, loff_t
*ppos
)
732 struct inode
* inode
= file
->f_dentry
->d_inode
;
737 if (!capable(CAP_AUDIT_CONTROL
))
740 if (current
!= pid_task(proc_pid(inode
), PIDTYPE_PID
))
743 if (count
>= PAGE_SIZE
)
744 count
= PAGE_SIZE
- 1;
747 /* No partial writes. */
750 page
= (char*)__get_free_page(GFP_USER
);
754 if (copy_from_user(page
, buf
, count
))
758 loginuid
= simple_strtoul(page
, &tmp
, 10);
764 length
= audit_set_loginuid(current
, loginuid
);
765 if (likely(length
== 0))
769 free_page((unsigned long) page
);
773 static struct file_operations proc_loginuid_operations
= {
774 .read
= proc_loginuid_read
,
775 .write
= proc_loginuid_write
,
779 #ifdef CONFIG_SECCOMP
780 static ssize_t
seccomp_read(struct file
*file
, char __user
*buf
,
781 size_t count
, loff_t
*ppos
)
783 struct task_struct
*tsk
= get_proc_task(file
->f_dentry
->d_inode
);
785 loff_t __ppos
= *ppos
;
790 /* no need to print the trailing zero, so use only len */
791 len
= sprintf(__buf
, "%u\n", tsk
->seccomp
.mode
);
792 put_task_struct(tsk
);
795 if (count
> len
- __ppos
)
796 count
= len
- __ppos
;
797 if (copy_to_user(buf
, __buf
+ __ppos
, count
))
799 *ppos
= __ppos
+ count
;
803 static ssize_t
seccomp_write(struct file
*file
, const char __user
*buf
,
804 size_t count
, loff_t
*ppos
)
806 struct task_struct
*tsk
= get_proc_task(file
->f_dentry
->d_inode
);
807 char __buf
[20], *end
;
808 unsigned int seccomp_mode
;
815 /* can set it only once to be even more secure */
817 if (unlikely(tsk
->seccomp
.mode
))
821 memset(__buf
, 0, sizeof(__buf
));
822 count
= min(count
, sizeof(__buf
) - 1);
823 if (copy_from_user(__buf
, buf
, count
))
826 seccomp_mode
= simple_strtoul(__buf
, &end
, 0);
830 if (seccomp_mode
&& seccomp_mode
<= NR_SECCOMP_MODES
) {
831 tsk
->seccomp
.mode
= seccomp_mode
;
832 set_tsk_thread_flag(tsk
, TIF_SECCOMP
);
836 if (unlikely(!(end
- __buf
)))
838 result
= end
- __buf
;
840 put_task_struct(tsk
);
845 static struct file_operations proc_seccomp_operations
= {
846 .read
= seccomp_read
,
847 .write
= seccomp_write
,
849 #endif /* CONFIG_SECCOMP */
851 static void *proc_pid_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
853 struct inode
*inode
= dentry
->d_inode
;
856 /* We don't need a base pointer in the /proc filesystem */
859 /* Are we allowed to snoop on the tasks file descriptors? */
860 if (!proc_fd_access_allowed(inode
))
863 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &nd
->dentry
, &nd
->mnt
);
864 nd
->last_type
= LAST_BIND
;
866 return ERR_PTR(error
);
869 static int do_proc_readlink(struct dentry
*dentry
, struct vfsmount
*mnt
,
870 char __user
*buffer
, int buflen
)
872 struct inode
* inode
;
873 char *tmp
= (char*)__get_free_page(GFP_KERNEL
), *path
;
879 inode
= dentry
->d_inode
;
880 path
= d_path(dentry
, mnt
, tmp
, PAGE_SIZE
);
884 len
= tmp
+ PAGE_SIZE
- 1 - path
;
888 if (copy_to_user(buffer
, path
, len
))
891 free_page((unsigned long)tmp
);
895 static int proc_pid_readlink(struct dentry
* dentry
, char __user
* buffer
, int buflen
)
898 struct inode
*inode
= dentry
->d_inode
;
900 struct vfsmount
*mnt
= NULL
;
902 /* Are we allowed to snoop on the tasks file descriptors? */
903 if (!proc_fd_access_allowed(inode
))
906 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &de
, &mnt
);
910 error
= do_proc_readlink(de
, mnt
, buffer
, buflen
);
917 static struct inode_operations proc_pid_link_inode_operations
= {
918 .readlink
= proc_pid_readlink
,
919 .follow_link
= proc_pid_follow_link
,
920 .setattr
= proc_setattr
,
924 /* building an inode */
926 static int task_dumpable(struct task_struct
*task
)
929 struct mm_struct
*mm
;
934 dumpable
= mm
->dumpable
;
942 static struct inode
*proc_pid_make_inode(struct super_block
* sb
, struct task_struct
*task
)
944 struct inode
* inode
;
945 struct proc_inode
*ei
;
947 /* We need a new inode */
949 inode
= new_inode(sb
);
955 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
956 inode
->i_op
= &proc_def_inode_operations
;
959 * grab the reference to task.
961 ei
->pid
= get_pid(task
->pids
[PIDTYPE_PID
].pid
);
967 if (task_dumpable(task
)) {
968 inode
->i_uid
= task
->euid
;
969 inode
->i_gid
= task
->egid
;
971 security_task_to_inode(task
, inode
);
981 static int pid_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
983 struct inode
*inode
= dentry
->d_inode
;
984 struct task_struct
*task
;
985 generic_fillattr(inode
, stat
);
990 task
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
992 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
993 task_dumpable(task
)) {
994 stat
->uid
= task
->euid
;
995 stat
->gid
= task
->egid
;
1005 * Exceptional case: normally we are not allowed to unhash a busy
1006 * directory. In this case, however, we can do it - no aliasing problems
1007 * due to the way we treat inodes.
1009 * Rewrite the inode's ownerships here because the owning task may have
1010 * performed a setuid(), etc.
1012 * Before the /proc/pid/status file was created the only way to read
1013 * the effective uid of a /process was to stat /proc/pid. Reading
1014 * /proc/pid/status is slow enough that procps and other packages
1015 * kept stating /proc/pid. To keep the rules in /proc simple I have
1016 * made this apply to all per process world readable and executable
1019 static int pid_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1021 struct inode
*inode
= dentry
->d_inode
;
1022 struct task_struct
*task
= get_proc_task(inode
);
1024 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1025 task_dumpable(task
)) {
1026 inode
->i_uid
= task
->euid
;
1027 inode
->i_gid
= task
->egid
;
1032 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1033 security_task_to_inode(task
, inode
);
1034 put_task_struct(task
);
1041 static int pid_delete_dentry(struct dentry
* dentry
)
1043 /* Is the task we represent dead?
1044 * If so, then don't put the dentry on the lru list,
1045 * kill it immediately.
1047 return !proc_pid(dentry
->d_inode
)->tasks
[PIDTYPE_PID
].first
;
1050 static struct dentry_operations pid_dentry_operations
=
1052 .d_revalidate
= pid_revalidate
,
1053 .d_delete
= pid_delete_dentry
,
1058 typedef struct dentry
*instantiate_t(struct inode
*, struct dentry
*, struct task_struct
*, void *);
1060 static int proc_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
1061 char *name
, int len
,
1062 instantiate_t instantiate
, struct task_struct
*task
, void *ptr
)
1064 struct dentry
*child
, *dir
= filp
->f_dentry
;
1065 struct inode
*inode
;
1068 unsigned type
= DT_UNKNOWN
;
1072 qname
.hash
= full_name_hash(name
, len
);
1074 child
= d_lookup(dir
, &qname
);
1077 new = d_alloc(dir
, &qname
);
1079 child
= instantiate(dir
->d_inode
, new, task
, ptr
);
1086 if (!child
|| IS_ERR(child
) || !child
->d_inode
)
1087 goto end_instantiate
;
1088 inode
= child
->d_inode
;
1091 type
= inode
->i_mode
>> 12;
1096 ino
= find_inode_number(dir
, &qname
);
1099 return filldir(dirent
, name
, len
, filp
->f_pos
, ino
, type
);
1102 static unsigned name_to_int(struct dentry
*dentry
)
1104 const char *name
= dentry
->d_name
.name
;
1105 int len
= dentry
->d_name
.len
;
1108 if (len
> 1 && *name
== '0')
1111 unsigned c
= *name
++ - '0';
1114 if (n
>= (~0U-9)/10)
1124 static int proc_fd_link(struct inode
*inode
, struct dentry
**dentry
, struct vfsmount
**mnt
)
1126 struct task_struct
*task
= get_proc_task(inode
);
1127 struct files_struct
*files
= NULL
;
1129 int fd
= proc_fd(inode
);
1132 files
= get_files_struct(task
);
1133 put_task_struct(task
);
1137 * We are not taking a ref to the file structure, so we must
1140 spin_lock(&files
->file_lock
);
1141 file
= fcheck_files(files
, fd
);
1143 *mnt
= mntget(file
->f_vfsmnt
);
1144 *dentry
= dget(file
->f_dentry
);
1145 spin_unlock(&files
->file_lock
);
1146 put_files_struct(files
);
1149 spin_unlock(&files
->file_lock
);
1150 put_files_struct(files
);
1155 static int tid_fd_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1157 struct inode
*inode
= dentry
->d_inode
;
1158 struct task_struct
*task
= get_proc_task(inode
);
1159 int fd
= proc_fd(inode
);
1160 struct files_struct
*files
;
1163 files
= get_files_struct(task
);
1166 if (fcheck_files(files
, fd
)) {
1168 put_files_struct(files
);
1169 if (task_dumpable(task
)) {
1170 inode
->i_uid
= task
->euid
;
1171 inode
->i_gid
= task
->egid
;
1176 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1177 security_task_to_inode(task
, inode
);
1178 put_task_struct(task
);
1182 put_files_struct(files
);
1184 put_task_struct(task
);
1190 static struct dentry_operations tid_fd_dentry_operations
=
1192 .d_revalidate
= tid_fd_revalidate
,
1193 .d_delete
= pid_delete_dentry
,
1196 static struct dentry
*proc_fd_instantiate(struct inode
*dir
,
1197 struct dentry
*dentry
, struct task_struct
*task
, void *ptr
)
1199 unsigned fd
= *(unsigned *)ptr
;
1201 struct files_struct
*files
;
1202 struct inode
*inode
;
1203 struct proc_inode
*ei
;
1204 struct dentry
*error
= ERR_PTR(-ENOENT
);
1206 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1211 files
= get_files_struct(task
);
1214 inode
->i_mode
= S_IFLNK
;
1217 * We are not taking a ref to the file structure, so we must
1220 spin_lock(&files
->file_lock
);
1221 file
= fcheck_files(files
, fd
);
1224 if (file
->f_mode
& 1)
1225 inode
->i_mode
|= S_IRUSR
| S_IXUSR
;
1226 if (file
->f_mode
& 2)
1227 inode
->i_mode
|= S_IWUSR
| S_IXUSR
;
1228 spin_unlock(&files
->file_lock
);
1229 put_files_struct(files
);
1231 inode
->i_op
= &proc_pid_link_inode_operations
;
1233 ei
->op
.proc_get_link
= proc_fd_link
;
1234 dentry
->d_op
= &tid_fd_dentry_operations
;
1235 d_add(dentry
, inode
);
1236 /* Close the race of the process dying before we return the dentry */
1237 if (tid_fd_revalidate(dentry
, NULL
))
1243 spin_unlock(&files
->file_lock
);
1244 put_files_struct(files
);
1251 static struct dentry
*proc_lookupfd(struct inode
* dir
, struct dentry
* dentry
, struct nameidata
*nd
)
1253 struct task_struct
*task
= get_proc_task(dir
);
1254 unsigned fd
= name_to_int(dentry
);
1255 struct dentry
*result
= ERR_PTR(-ENOENT
);
1262 result
= proc_fd_instantiate(dir
, dentry
, task
, &fd
);
1264 put_task_struct(task
);
1269 static int proc_fd_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
1270 struct task_struct
*task
, int fd
)
1272 char name
[PROC_NUMBUF
];
1273 int len
= snprintf(name
, sizeof(name
), "%d", fd
);
1274 return proc_fill_cache(filp
, dirent
, filldir
, name
, len
,
1275 proc_fd_instantiate
, task
, &fd
);
1278 static int proc_readfd(struct file
* filp
, void * dirent
, filldir_t filldir
)
1280 struct dentry
*dentry
= filp
->f_dentry
;
1281 struct inode
*inode
= dentry
->d_inode
;
1282 struct task_struct
*p
= get_proc_task(inode
);
1283 unsigned int fd
, tid
, ino
;
1285 struct files_struct
* files
;
1286 struct fdtable
*fdt
;
1297 if (filldir(dirent
, ".", 1, 0, inode
->i_ino
, DT_DIR
) < 0)
1301 ino
= parent_ino(dentry
);
1302 if (filldir(dirent
, "..", 2, 1, ino
, DT_DIR
) < 0)
1306 files
= get_files_struct(p
);
1310 fdt
= files_fdtable(files
);
1311 for (fd
= filp
->f_pos
-2;
1313 fd
++, filp
->f_pos
++) {
1315 if (!fcheck_files(files
, fd
))
1319 if (proc_fd_fill_cache(filp
, dirent
, filldir
, p
, fd
) < 0) {
1326 put_files_struct(files
);
1334 static struct file_operations proc_fd_operations
= {
1335 .read
= generic_read_dir
,
1336 .readdir
= proc_readfd
,
1340 * proc directories can do almost nothing..
1342 static struct inode_operations proc_fd_inode_operations
= {
1343 .lookup
= proc_lookupfd
,
1344 .setattr
= proc_setattr
,
1347 static struct dentry
*proc_pident_instantiate(struct inode
*dir
,
1348 struct dentry
*dentry
, struct task_struct
*task
, void *ptr
)
1350 struct pid_entry
*p
= ptr
;
1351 struct inode
*inode
;
1352 struct proc_inode
*ei
;
1353 struct dentry
*error
= ERR_PTR(-EINVAL
);
1355 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1360 inode
->i_mode
= p
->mode
;
1361 if (S_ISDIR(inode
->i_mode
))
1362 inode
->i_nlink
= 2; /* Use getattr to fix if necessary */
1364 inode
->i_op
= p
->iop
;
1366 inode
->i_fop
= p
->fop
;
1368 dentry
->d_op
= &pid_dentry_operations
;
1369 d_add(dentry
, inode
);
1370 /* Close the race of the process dying before we return the dentry */
1371 if (pid_revalidate(dentry
, NULL
))
1378 static struct dentry
*proc_pident_lookup(struct inode
*dir
,
1379 struct dentry
*dentry
,
1380 struct pid_entry
*ents
)
1382 struct inode
*inode
;
1383 struct dentry
*error
;
1384 struct task_struct
*task
= get_proc_task(dir
);
1385 struct pid_entry
*p
;
1387 error
= ERR_PTR(-ENOENT
);
1394 * Yes, it does not scale. And it should not. Don't add
1395 * new entries into /proc/<tgid>/ without very good reasons.
1397 for (p
= ents
; p
->name
; p
++) {
1398 if (p
->len
!= dentry
->d_name
.len
)
1400 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
1406 error
= proc_pident_instantiate(dir
, dentry
, task
, p
);
1408 put_task_struct(task
);
1413 static int proc_pident_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
1414 struct task_struct
*task
, struct pid_entry
*p
)
1416 return proc_fill_cache(filp
, dirent
, filldir
, p
->name
, p
->len
,
1417 proc_pident_instantiate
, task
, p
);
1420 static int proc_pident_readdir(struct file
*filp
,
1421 void *dirent
, filldir_t filldir
,
1422 struct pid_entry
*ents
, unsigned int nents
)
1426 struct dentry
*dentry
= filp
->f_dentry
;
1427 struct inode
*inode
= dentry
->d_inode
;
1428 struct task_struct
*task
= get_proc_task(inode
);
1429 struct pid_entry
*p
;
1443 if (filldir(dirent
, ".", 1, i
, ino
, DT_DIR
) < 0)
1449 ino
= parent_ino(dentry
);
1450 if (filldir(dirent
, "..", 2, i
, ino
, DT_DIR
) < 0)
1463 if (proc_pident_fill_cache(filp
, dirent
, filldir
, task
, p
) < 0)
1472 put_task_struct(task
);
1477 #ifdef CONFIG_SECURITY
1478 static ssize_t
proc_pid_attr_read(struct file
* file
, char __user
* buf
,
1479 size_t count
, loff_t
*ppos
)
1481 struct inode
* inode
= file
->f_dentry
->d_inode
;
1484 struct task_struct
*task
= get_proc_task(inode
);
1490 if (count
> PAGE_SIZE
)
1493 if (!(page
= __get_free_page(GFP_KERNEL
)))
1496 length
= security_getprocattr(task
,
1497 (char*)file
->f_dentry
->d_name
.name
,
1498 (void*)page
, count
);
1500 length
= simple_read_from_buffer(buf
, count
, ppos
, (char *)page
, length
);
1503 put_task_struct(task
);
1508 static ssize_t
proc_pid_attr_write(struct file
* file
, const char __user
* buf
,
1509 size_t count
, loff_t
*ppos
)
1511 struct inode
* inode
= file
->f_dentry
->d_inode
;
1514 struct task_struct
*task
= get_proc_task(inode
);
1519 if (count
> PAGE_SIZE
)
1522 /* No partial writes. */
1528 page
= (char*)__get_free_page(GFP_USER
);
1533 if (copy_from_user(page
, buf
, count
))
1536 length
= security_setprocattr(task
,
1537 (char*)file
->f_dentry
->d_name
.name
,
1538 (void*)page
, count
);
1540 free_page((unsigned long) page
);
1542 put_task_struct(task
);
1547 static struct file_operations proc_pid_attr_operations
= {
1548 .read
= proc_pid_attr_read
,
1549 .write
= proc_pid_attr_write
,
1552 static struct pid_entry tgid_attr_stuff
[] = {
1553 REG("current", S_IRUGO
|S_IWUGO
, pid_attr
),
1554 REG("prev", S_IRUGO
, pid_attr
),
1555 REG("exec", S_IRUGO
|S_IWUGO
, pid_attr
),
1556 REG("fscreate", S_IRUGO
|S_IWUGO
, pid_attr
),
1557 REG("keycreate", S_IRUGO
|S_IWUGO
, pid_attr
),
1558 REG("sockcreate", S_IRUGO
|S_IWUGO
, pid_attr
),
1561 static struct pid_entry tid_attr_stuff
[] = {
1562 REG("current", S_IRUGO
|S_IWUGO
, pid_attr
),
1563 REG("prev", S_IRUGO
, pid_attr
),
1564 REG("exec", S_IRUGO
|S_IWUGO
, pid_attr
),
1565 REG("fscreate", S_IRUGO
|S_IWUGO
, pid_attr
),
1566 REG("keycreate", S_IRUGO
|S_IWUGO
, pid_attr
),
1567 REG("sockcreate", S_IRUGO
|S_IWUGO
, pid_attr
),
1571 static int proc_tgid_attr_readdir(struct file
* filp
,
1572 void * dirent
, filldir_t filldir
)
1574 return proc_pident_readdir(filp
,dirent
,filldir
,
1575 tgid_attr_stuff
,ARRAY_SIZE(tgid_attr_stuff
));
1578 static int proc_tid_attr_readdir(struct file
* filp
,
1579 void * dirent
, filldir_t filldir
)
1581 return proc_pident_readdir(filp
,dirent
,filldir
,
1582 tid_attr_stuff
,ARRAY_SIZE(tid_attr_stuff
));
1585 static struct file_operations proc_tgid_attr_operations
= {
1586 .read
= generic_read_dir
,
1587 .readdir
= proc_tgid_attr_readdir
,
1590 static struct file_operations proc_tid_attr_operations
= {
1591 .read
= generic_read_dir
,
1592 .readdir
= proc_tid_attr_readdir
,
1595 static struct dentry
*proc_tgid_attr_lookup(struct inode
*dir
,
1596 struct dentry
*dentry
, struct nameidata
*nd
)
1598 return proc_pident_lookup(dir
, dentry
, tgid_attr_stuff
);
1601 static struct dentry
*proc_tid_attr_lookup(struct inode
*dir
,
1602 struct dentry
*dentry
, struct nameidata
*nd
)
1604 return proc_pident_lookup(dir
, dentry
, tid_attr_stuff
);
1607 static struct inode_operations proc_tgid_attr_inode_operations
= {
1608 .lookup
= proc_tgid_attr_lookup
,
1609 .getattr
= pid_getattr
,
1610 .setattr
= proc_setattr
,
1613 static struct inode_operations proc_tid_attr_inode_operations
= {
1614 .lookup
= proc_tid_attr_lookup
,
1615 .getattr
= pid_getattr
,
1616 .setattr
= proc_setattr
,
1623 static int proc_self_readlink(struct dentry
*dentry
, char __user
*buffer
,
1626 char tmp
[PROC_NUMBUF
];
1627 sprintf(tmp
, "%d", current
->tgid
);
1628 return vfs_readlink(dentry
,buffer
,buflen
,tmp
);
1631 static void *proc_self_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
1633 char tmp
[PROC_NUMBUF
];
1634 sprintf(tmp
, "%d", current
->tgid
);
1635 return ERR_PTR(vfs_follow_link(nd
,tmp
));
1638 static struct inode_operations proc_self_inode_operations
= {
1639 .readlink
= proc_self_readlink
,
1640 .follow_link
= proc_self_follow_link
,
1646 * These are the directory entries in the root directory of /proc
1647 * that properly belong to the /proc filesystem, as they describe
1648 * describe something that is process related.
1650 static struct pid_entry proc_base_stuff
[] = {
1651 NOD("self", S_IFLNK
|S_IRWXUGO
,
1652 &proc_self_inode_operations
, NULL
, {}),
1657 * Exceptional case: normally we are not allowed to unhash a busy
1658 * directory. In this case, however, we can do it - no aliasing problems
1659 * due to the way we treat inodes.
1661 static int proc_base_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1663 struct inode
*inode
= dentry
->d_inode
;
1664 struct task_struct
*task
= get_proc_task(inode
);
1666 put_task_struct(task
);
1673 static struct dentry_operations proc_base_dentry_operations
=
1675 .d_revalidate
= proc_base_revalidate
,
1676 .d_delete
= pid_delete_dentry
,
1679 static struct dentry
*proc_base_instantiate(struct inode
*dir
,
1680 struct dentry
*dentry
, struct task_struct
*task
, void *ptr
)
1682 struct pid_entry
*p
= ptr
;
1683 struct inode
*inode
;
1684 struct proc_inode
*ei
;
1685 struct dentry
*error
= ERR_PTR(-EINVAL
);
1687 /* Allocate the inode */
1688 error
= ERR_PTR(-ENOMEM
);
1689 inode
= new_inode(dir
->i_sb
);
1693 /* Initialize the inode */
1695 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1698 * grab the reference to the task.
1700 ei
->pid
= get_pid(task_pid(task
));
1706 inode
->i_mode
= p
->mode
;
1707 if (S_ISDIR(inode
->i_mode
))
1709 if (S_ISLNK(inode
->i_mode
))
1712 inode
->i_op
= p
->iop
;
1714 inode
->i_fop
= p
->fop
;
1716 dentry
->d_op
= &proc_base_dentry_operations
;
1717 d_add(dentry
, inode
);
1726 static struct dentry
*proc_base_lookup(struct inode
*dir
, struct dentry
*dentry
)
1728 struct dentry
*error
;
1729 struct task_struct
*task
= get_proc_task(dir
);
1730 struct pid_entry
*p
;
1732 error
= ERR_PTR(-ENOENT
);
1737 /* Lookup the directory entry */
1738 for (p
= proc_base_stuff
; p
->name
; p
++) {
1739 if (p
->len
!= dentry
->d_name
.len
)
1741 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
1747 error
= proc_base_instantiate(dir
, dentry
, task
, p
);
1750 put_task_struct(task
);
1755 static int proc_base_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
1756 struct task_struct
*task
, struct pid_entry
*p
)
1758 return proc_fill_cache(filp
, dirent
, filldir
, p
->name
, p
->len
,
1759 proc_base_instantiate
, task
, p
);
1765 static struct file_operations proc_task_operations
;
1766 static struct inode_operations proc_task_inode_operations
;
1768 static struct pid_entry tgid_base_stuff
[] = {
1769 DIR("task", S_IRUGO
|S_IXUGO
, task
),
1770 DIR("fd", S_IRUSR
|S_IXUSR
, fd
),
1771 INF("environ", S_IRUSR
, pid_environ
),
1772 INF("auxv", S_IRUSR
, pid_auxv
),
1773 INF("status", S_IRUGO
, pid_status
),
1774 INF("cmdline", S_IRUGO
, pid_cmdline
),
1775 INF("stat", S_IRUGO
, tgid_stat
),
1776 INF("statm", S_IRUGO
, pid_statm
),
1777 REG("maps", S_IRUGO
, maps
),
1779 REG("numa_maps", S_IRUGO
, numa_maps
),
1781 REG("mem", S_IRUSR
|S_IWUSR
, mem
),
1782 #ifdef CONFIG_SECCOMP
1783 REG("seccomp", S_IRUSR
|S_IWUSR
, seccomp
),
1788 REG("mounts", S_IRUGO
, mounts
),
1789 REG("mountstats", S_IRUSR
, mountstats
),
1791 REG("smaps", S_IRUGO
, smaps
),
1793 #ifdef CONFIG_SECURITY
1794 DIR("attr", S_IRUGO
|S_IXUGO
, tgid_attr
),
1796 #ifdef CONFIG_KALLSYMS
1797 INF("wchan", S_IRUGO
, pid_wchan
),
1799 #ifdef CONFIG_SCHEDSTATS
1800 INF("schedstat", S_IRUGO
, pid_schedstat
),
1802 #ifdef CONFIG_CPUSETS
1803 REG("cpuset", S_IRUGO
, cpuset
),
1805 INF("oom_score", S_IRUGO
, oom_score
),
1806 REG("oom_adj", S_IRUGO
|S_IWUSR
, oom_adjust
),
1807 #ifdef CONFIG_AUDITSYSCALL
1808 REG("loginuid", S_IWUSR
|S_IRUGO
, loginuid
),
1813 static int proc_tgid_base_readdir(struct file
* filp
,
1814 void * dirent
, filldir_t filldir
)
1816 return proc_pident_readdir(filp
,dirent
,filldir
,
1817 tgid_base_stuff
,ARRAY_SIZE(tgid_base_stuff
));
1820 static struct file_operations proc_tgid_base_operations
= {
1821 .read
= generic_read_dir
,
1822 .readdir
= proc_tgid_base_readdir
,
1825 static struct dentry
*proc_tgid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
1826 return proc_pident_lookup(dir
, dentry
, tgid_base_stuff
);
1829 static struct inode_operations proc_tgid_base_inode_operations
= {
1830 .lookup
= proc_tgid_base_lookup
,
1831 .getattr
= pid_getattr
,
1832 .setattr
= proc_setattr
,
1836 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
1838 * @task: task that should be flushed.
1840 * Looks in the dcache for
1842 * /proc/@tgid/task/@pid
1843 * if either directory is present flushes it and all of it'ts children
1846 * It is safe and reasonable to cache /proc entries for a task until
1847 * that task exits. After that they just clog up the dcache with
1848 * useless entries, possibly causing useful dcache entries to be
1849 * flushed instead. This routine is proved to flush those useless
1850 * dcache entries at process exit time.
1852 * NOTE: This routine is just an optimization so it does not guarantee
1853 * that no dcache entries will exist at process exit time it
1854 * just makes it very unlikely that any will persist.
1856 void proc_flush_task(struct task_struct
*task
)
1858 struct dentry
*dentry
, *leader
, *dir
;
1859 char buf
[PROC_NUMBUF
];
1863 name
.len
= snprintf(buf
, sizeof(buf
), "%d", task
->pid
);
1864 dentry
= d_hash_and_lookup(proc_mnt
->mnt_root
, &name
);
1866 shrink_dcache_parent(dentry
);
1871 if (thread_group_leader(task
))
1875 name
.len
= snprintf(buf
, sizeof(buf
), "%d", task
->tgid
);
1876 leader
= d_hash_and_lookup(proc_mnt
->mnt_root
, &name
);
1881 name
.len
= strlen(name
.name
);
1882 dir
= d_hash_and_lookup(leader
, &name
);
1884 goto out_put_leader
;
1887 name
.len
= snprintf(buf
, sizeof(buf
), "%d", task
->pid
);
1888 dentry
= d_hash_and_lookup(dir
, &name
);
1890 shrink_dcache_parent(dentry
);
1902 struct dentry
*proc_pid_instantiate(struct inode
*dir
,
1903 struct dentry
* dentry
, struct task_struct
*task
, void *ptr
)
1905 struct dentry
*error
= ERR_PTR(-ENOENT
);
1906 struct inode
*inode
;
1908 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1912 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
1913 inode
->i_op
= &proc_tgid_base_inode_operations
;
1914 inode
->i_fop
= &proc_tgid_base_operations
;
1915 inode
->i_flags
|=S_IMMUTABLE
;
1917 #ifdef CONFIG_SECURITY
1918 inode
->i_nlink
+= 1;
1921 dentry
->d_op
= &pid_dentry_operations
;
1923 d_add(dentry
, inode
);
1924 /* Close the race of the process dying before we return the dentry */
1925 if (pid_revalidate(dentry
, NULL
))
1932 struct dentry
*proc_pid_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
1934 struct dentry
*result
= ERR_PTR(-ENOENT
);
1935 struct task_struct
*task
;
1938 result
= proc_base_lookup(dir
, dentry
);
1939 if (!IS_ERR(result
) || PTR_ERR(result
) != -ENOENT
)
1942 tgid
= name_to_int(dentry
);
1947 task
= find_task_by_pid(tgid
);
1949 get_task_struct(task
);
1954 result
= proc_pid_instantiate(dir
, dentry
, task
, NULL
);
1955 put_task_struct(task
);
1961 * Find the first task with tgid >= tgid
1964 static struct task_struct
*next_tgid(unsigned int tgid
)
1966 struct task_struct
*task
;
1972 pid
= find_ge_pid(tgid
);
1975 task
= pid_task(pid
, PIDTYPE_PID
);
1976 /* What we to know is if the pid we have find is the
1977 * pid of a thread_group_leader. Testing for task
1978 * being a thread_group_leader is the obvious thing
1979 * todo but there is a window when it fails, due to
1980 * the pid transfer logic in de_thread.
1982 * So we perform the straight forward test of seeing
1983 * if the pid we have found is the pid of a thread
1984 * group leader, and don't worry if the task we have
1985 * found doesn't happen to be a thread group leader.
1986 * As we don't care in the case of readdir.
1988 if (!task
|| !has_group_leader_pid(task
))
1990 get_task_struct(task
);
1996 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + (1 /* /proc/self */))
1998 static int proc_pid_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
1999 struct task_struct
*task
, int tgid
)
2001 char name
[PROC_NUMBUF
];
2002 int len
= snprintf(name
, sizeof(name
), "%d", tgid
);
2003 return proc_fill_cache(filp
, dirent
, filldir
, name
, len
,
2004 proc_pid_instantiate
, task
, NULL
);
2007 /* for the /proc/ directory itself, after non-process stuff has been done */
2008 int proc_pid_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
2010 unsigned int nr
= filp
->f_pos
- FIRST_PROCESS_ENTRY
;
2011 struct task_struct
*reaper
= get_proc_task(filp
->f_dentry
->d_inode
);
2012 struct task_struct
*task
;
2018 for (; nr
< (ARRAY_SIZE(proc_base_stuff
) - 1); filp
->f_pos
++, nr
++) {
2019 struct pid_entry
*p
= &proc_base_stuff
[nr
];
2020 if (proc_base_fill_cache(filp
, dirent
, filldir
, reaper
, p
) < 0)
2024 tgid
= filp
->f_pos
- TGID_OFFSET
;
2025 for (task
= next_tgid(tgid
);
2027 put_task_struct(task
), task
= next_tgid(tgid
+ 1)) {
2029 filp
->f_pos
= tgid
+ TGID_OFFSET
;
2030 if (proc_pid_fill_cache(filp
, dirent
, filldir
, task
, tgid
) < 0) {
2031 put_task_struct(task
);
2035 filp
->f_pos
= PID_MAX_LIMIT
+ TGID_OFFSET
;
2037 put_task_struct(reaper
);
2045 static struct pid_entry tid_base_stuff
[] = {
2046 DIR("fd", S_IRUSR
|S_IXUSR
, fd
),
2047 INF("environ", S_IRUSR
, pid_environ
),
2048 INF("auxv", S_IRUSR
, pid_auxv
),
2049 INF("status", S_IRUGO
, pid_status
),
2050 INF("cmdline", S_IRUGO
, pid_cmdline
),
2051 INF("stat", S_IRUGO
, tid_stat
),
2052 INF("statm", S_IRUGO
, pid_statm
),
2053 REG("maps", S_IRUGO
, maps
),
2055 REG("numa_maps", S_IRUGO
, numa_maps
),
2057 REG("mem", S_IRUSR
|S_IWUSR
, mem
),
2058 #ifdef CONFIG_SECCOMP
2059 REG("seccomp", S_IRUSR
|S_IWUSR
, seccomp
),
2064 REG("mounts", S_IRUGO
, mounts
),
2066 REG("smaps", S_IRUGO
, smaps
),
2068 #ifdef CONFIG_SECURITY
2069 DIR("attr", S_IRUGO
|S_IXUGO
, tid_attr
),
2071 #ifdef CONFIG_KALLSYMS
2072 INF("wchan", S_IRUGO
, pid_wchan
),
2074 #ifdef CONFIG_SCHEDSTATS
2075 INF("schedstat", S_IRUGO
, pid_schedstat
),
2077 #ifdef CONFIG_CPUSETS
2078 REG("cpuset", S_IRUGO
, cpuset
),
2080 INF("oom_score", S_IRUGO
, oom_score
),
2081 REG("oom_adj", S_IRUGO
|S_IWUSR
, oom_adjust
),
2082 #ifdef CONFIG_AUDITSYSCALL
2083 REG("loginuid", S_IWUSR
|S_IRUGO
, loginuid
),
2088 static int proc_tid_base_readdir(struct file
* filp
,
2089 void * dirent
, filldir_t filldir
)
2091 return proc_pident_readdir(filp
,dirent
,filldir
,
2092 tid_base_stuff
,ARRAY_SIZE(tid_base_stuff
));
2095 static struct dentry
*proc_tid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
2096 return proc_pident_lookup(dir
, dentry
, tid_base_stuff
);
2099 static struct file_operations proc_tid_base_operations
= {
2100 .read
= generic_read_dir
,
2101 .readdir
= proc_tid_base_readdir
,
2104 static struct inode_operations proc_tid_base_inode_operations
= {
2105 .lookup
= proc_tid_base_lookup
,
2106 .getattr
= pid_getattr
,
2107 .setattr
= proc_setattr
,
2110 static struct dentry
*proc_task_instantiate(struct inode
*dir
,
2111 struct dentry
*dentry
, struct task_struct
*task
, void *ptr
)
2113 struct dentry
*error
= ERR_PTR(-ENOENT
);
2114 struct inode
*inode
;
2115 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2119 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
2120 inode
->i_op
= &proc_tid_base_inode_operations
;
2121 inode
->i_fop
= &proc_tid_base_operations
;
2122 inode
->i_flags
|=S_IMMUTABLE
;
2124 #ifdef CONFIG_SECURITY
2125 inode
->i_nlink
+= 1;
2128 dentry
->d_op
= &pid_dentry_operations
;
2130 d_add(dentry
, inode
);
2131 /* Close the race of the process dying before we return the dentry */
2132 if (pid_revalidate(dentry
, NULL
))
2139 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
2141 struct dentry
*result
= ERR_PTR(-ENOENT
);
2142 struct task_struct
*task
;
2143 struct task_struct
*leader
= get_proc_task(dir
);
2149 tid
= name_to_int(dentry
);
2154 task
= find_task_by_pid(tid
);
2156 get_task_struct(task
);
2160 if (leader
->tgid
!= task
->tgid
)
2163 result
= proc_task_instantiate(dir
, dentry
, task
, NULL
);
2165 put_task_struct(task
);
2167 put_task_struct(leader
);
2173 * Find the first tid of a thread group to return to user space.
2175 * Usually this is just the thread group leader, but if the users
2176 * buffer was too small or there was a seek into the middle of the
2177 * directory we have more work todo.
2179 * In the case of a short read we start with find_task_by_pid.
2181 * In the case of a seek we start with the leader and walk nr
2184 static struct task_struct
*first_tid(struct task_struct
*leader
,
2187 struct task_struct
*pos
;
2190 /* Attempt to start with the pid of a thread */
2191 if (tid
&& (nr
> 0)) {
2192 pos
= find_task_by_pid(tid
);
2193 if (pos
&& (pos
->group_leader
== leader
))
2197 /* If nr exceeds the number of threads there is nothing todo */
2199 if (nr
&& nr
>= get_nr_threads(leader
))
2202 /* If we haven't found our starting place yet start
2203 * with the leader and walk nr threads forward.
2205 for (pos
= leader
; nr
> 0; --nr
) {
2206 pos
= next_thread(pos
);
2207 if (pos
== leader
) {
2213 get_task_struct(pos
);
2220 * Find the next thread in the thread list.
2221 * Return NULL if there is an error or no next thread.
2223 * The reference to the input task_struct is released.
2225 static struct task_struct
*next_tid(struct task_struct
*start
)
2227 struct task_struct
*pos
= NULL
;
2229 if (pid_alive(start
)) {
2230 pos
= next_thread(start
);
2231 if (thread_group_leader(pos
))
2234 get_task_struct(pos
);
2237 put_task_struct(start
);
2241 static int proc_task_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
2242 struct task_struct
*task
, int tid
)
2244 char name
[PROC_NUMBUF
];
2245 int len
= snprintf(name
, sizeof(name
), "%d", tid
);
2246 return proc_fill_cache(filp
, dirent
, filldir
, name
, len
,
2247 proc_task_instantiate
, task
, NULL
);
2250 /* for the /proc/TGID/task/ directories */
2251 static int proc_task_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
2253 struct dentry
*dentry
= filp
->f_dentry
;
2254 struct inode
*inode
= dentry
->d_inode
;
2255 struct task_struct
*leader
= get_proc_task(inode
);
2256 struct task_struct
*task
;
2257 int retval
= -ENOENT
;
2260 unsigned long pos
= filp
->f_pos
; /* avoiding "long long" filp->f_pos */
2269 if (filldir(dirent
, ".", 1, pos
, ino
, DT_DIR
) < 0)
2274 ino
= parent_ino(dentry
);
2275 if (filldir(dirent
, "..", 2, pos
, ino
, DT_DIR
) < 0)
2281 /* f_version caches the tgid value that the last readdir call couldn't
2282 * return. lseek aka telldir automagically resets f_version to 0.
2284 tid
= filp
->f_version
;
2285 filp
->f_version
= 0;
2286 for (task
= first_tid(leader
, tid
, pos
- 2);
2288 task
= next_tid(task
), pos
++) {
2290 if (proc_task_fill_cache(filp
, dirent
, filldir
, task
, tid
) < 0) {
2291 /* returning this tgid failed, save it as the first
2292 * pid for the next readir call */
2293 filp
->f_version
= tid
;
2294 put_task_struct(task
);
2300 put_task_struct(leader
);
2305 static int proc_task_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
2307 struct inode
*inode
= dentry
->d_inode
;
2308 struct task_struct
*p
= get_proc_task(inode
);
2309 generic_fillattr(inode
, stat
);
2313 stat
->nlink
+= get_nr_threads(p
);
2321 static struct inode_operations proc_task_inode_operations
= {
2322 .lookup
= proc_task_lookup
,
2323 .getattr
= proc_task_getattr
,
2324 .setattr
= proc_setattr
,
2327 static struct file_operations proc_task_operations
= {
2328 .read
= generic_read_dir
,
2329 .readdir
= proc_task_readdir
,