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>
77 * Implementing inode permission operations in /proc is almost
78 * certainly an error. Permission checks need to happen during
79 * each system call not at open time. The reason is that most of
80 * what we wish to check for permissions in /proc varies at runtime.
82 * The classic example of a problem is opening file descriptors
83 * in /proc for a task before it execs a suid executable.
87 * For hysterical raisins we keep the same inumbers as in the old procfs.
88 * Feel free to change the macro below - just keep the range distinct from
89 * inumbers of the rest of procfs (currently those are in 0x0000--0xffff).
90 * As soon as we'll get a separate superblock we will be able to forget
91 * about magical ranges too.
94 #define fake_ino(pid,ino) (((pid)<<16)|(ino))
96 enum pid_directory_inos
{
101 #ifdef CONFIG_SECCOMP
116 PROC_TGID_MOUNTSTATS
,
121 #ifdef CONFIG_SCHEDSTATS
124 #ifdef CONFIG_CPUSETS
127 #ifdef CONFIG_SECURITY
129 PROC_TGID_ATTR_CURRENT
,
132 PROC_TGID_ATTR_FSCREATE
,
133 PROC_TGID_ATTR_KEYCREATE
,
134 PROC_TGID_ATTR_SOCKCREATE
,
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
,
177 PROC_TID_ATTR_SOCKCREATE
,
179 #ifdef CONFIG_AUDITSYSCALL
185 /* Add new entries before this */
186 PROC_TID_FD_DIR
= 0x8000, /* 0x8000-0xffff */
189 /* Worst case buffer size needed for holding an integer. */
190 #define PROC_NUMBUF 10
197 struct inode_operations
*iop
;
198 struct file_operations
*fop
;
202 #define NOD(TYPE, NAME, MODE, IOP, FOP, OP) { \
204 .len = sizeof(NAME) - 1, \
212 #define DIR(TYPE, NAME, MODE, OTYPE) \
213 NOD(TYPE, NAME, (S_IFDIR|(MODE)), \
214 &proc_##OTYPE##_inode_operations, &proc_##OTYPE##_operations, \
216 #define LNK(TYPE, NAME, OTYPE) \
217 NOD(TYPE, NAME, (S_IFLNK|S_IRWXUGO), \
218 &proc_pid_link_inode_operations, NULL, \
219 { .proc_get_link = &proc_##OTYPE##_link } )
220 #define REG(TYPE, NAME, MODE, OTYPE) \
221 NOD(TYPE, NAME, (S_IFREG|(MODE)), NULL, \
222 &proc_##OTYPE##_operations, {})
223 #define INF(TYPE, NAME, MODE, OTYPE) \
224 NOD(TYPE, NAME, (S_IFREG|(MODE)), \
225 NULL, &proc_info_file_operations, \
226 { .proc_read = &proc_##OTYPE } )
228 static struct fs_struct
*get_fs_struct(struct task_struct
*task
)
230 struct fs_struct
*fs
;
234 atomic_inc(&fs
->count
);
239 static int get_nr_threads(struct task_struct
*tsk
)
241 /* Must be called with the rcu_read_lock held */
245 if (lock_task_sighand(tsk
, &flags
)) {
246 count
= atomic_read(&tsk
->signal
->count
);
247 unlock_task_sighand(tsk
, &flags
);
252 static int proc_cwd_link(struct inode
*inode
, struct dentry
**dentry
, struct vfsmount
**mnt
)
254 struct task_struct
*task
= get_proc_task(inode
);
255 struct fs_struct
*fs
= NULL
;
256 int result
= -ENOENT
;
259 fs
= get_fs_struct(task
);
260 put_task_struct(task
);
263 read_lock(&fs
->lock
);
264 *mnt
= mntget(fs
->pwdmnt
);
265 *dentry
= dget(fs
->pwd
);
266 read_unlock(&fs
->lock
);
273 static int proc_root_link(struct inode
*inode
, struct dentry
**dentry
, struct vfsmount
**mnt
)
275 struct task_struct
*task
= get_proc_task(inode
);
276 struct fs_struct
*fs
= NULL
;
277 int result
= -ENOENT
;
280 fs
= get_fs_struct(task
);
281 put_task_struct(task
);
284 read_lock(&fs
->lock
);
285 *mnt
= mntget(fs
->rootmnt
);
286 *dentry
= dget(fs
->root
);
287 read_unlock(&fs
->lock
);
294 #define MAY_PTRACE(task) \
295 (task == current || \
296 (task->parent == current && \
297 (task->ptrace & PT_PTRACED) && \
298 (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
299 security_ptrace(current,task) == 0))
301 static int proc_pid_environ(struct task_struct
*task
, char * buffer
)
304 struct mm_struct
*mm
= get_task_mm(task
);
306 unsigned int len
= mm
->env_end
- mm
->env_start
;
309 res
= access_process_vm(task
, mm
->env_start
, buffer
, len
, 0);
310 if (!ptrace_may_attach(task
))
317 static int proc_pid_cmdline(struct task_struct
*task
, char * buffer
)
321 struct mm_struct
*mm
= get_task_mm(task
);
325 goto out_mm
; /* Shh! No looking before we're done */
327 len
= mm
->arg_end
- mm
->arg_start
;
332 res
= access_process_vm(task
, mm
->arg_start
, buffer
, len
, 0);
334 // If the nul at the end of args has been overwritten, then
335 // assume application is using setproctitle(3).
336 if (res
> 0 && buffer
[res
-1] != '\0' && len
< PAGE_SIZE
) {
337 len
= strnlen(buffer
, res
);
341 len
= mm
->env_end
- mm
->env_start
;
342 if (len
> PAGE_SIZE
- res
)
343 len
= PAGE_SIZE
- res
;
344 res
+= access_process_vm(task
, mm
->env_start
, buffer
+res
, len
, 0);
345 res
= strnlen(buffer
, res
);
354 static int proc_pid_auxv(struct task_struct
*task
, char *buffer
)
357 struct mm_struct
*mm
= get_task_mm(task
);
359 unsigned int nwords
= 0;
362 while (mm
->saved_auxv
[nwords
- 2] != 0); /* AT_NULL */
363 res
= nwords
* sizeof(mm
->saved_auxv
[0]);
366 memcpy(buffer
, mm
->saved_auxv
, res
);
373 #ifdef CONFIG_KALLSYMS
375 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
376 * Returns the resolved symbol. If that fails, simply return the address.
378 static int proc_pid_wchan(struct task_struct
*task
, char *buffer
)
381 const char *sym_name
;
382 unsigned long wchan
, size
, offset
;
383 char namebuf
[KSYM_NAME_LEN
+1];
385 wchan
= get_wchan(task
);
387 sym_name
= kallsyms_lookup(wchan
, &size
, &offset
, &modname
, namebuf
);
389 return sprintf(buffer
, "%s", sym_name
);
390 return sprintf(buffer
, "%lu", wchan
);
392 #endif /* CONFIG_KALLSYMS */
394 #ifdef CONFIG_SCHEDSTATS
396 * Provides /proc/PID/schedstat
398 static int proc_pid_schedstat(struct task_struct
*task
, char *buffer
)
400 return sprintf(buffer
, "%lu %lu %lu\n",
401 task
->sched_info
.cpu_time
,
402 task
->sched_info
.run_delay
,
403 task
->sched_info
.pcnt
);
407 /* The badness from the OOM killer */
408 unsigned long badness(struct task_struct
*p
, unsigned long uptime
);
409 static int proc_oom_score(struct task_struct
*task
, char *buffer
)
411 unsigned long points
;
412 struct timespec uptime
;
414 do_posix_clock_monotonic_gettime(&uptime
);
415 points
= badness(task
, uptime
.tv_sec
);
416 return sprintf(buffer
, "%lu\n", points
);
419 /************************************************************************/
420 /* Here the fs part begins */
421 /************************************************************************/
423 /* permission checks */
424 static int proc_fd_access_allowed(struct inode
*inode
)
426 struct task_struct
*task
;
428 /* Allow access to a task's file descriptors if it is us or we
429 * may use ptrace attach to the process and find out that
432 task
= get_proc_task(inode
);
434 allowed
= ptrace_may_attach(task
);
435 put_task_struct(task
);
440 static int proc_setattr(struct dentry
*dentry
, struct iattr
*attr
)
443 struct inode
*inode
= dentry
->d_inode
;
445 if (attr
->ia_valid
& ATTR_MODE
)
448 error
= inode_change_ok(inode
, attr
);
450 error
= security_inode_setattr(dentry
, attr
);
452 error
= inode_setattr(inode
, attr
);
457 static struct inode_operations proc_def_inode_operations
= {
458 .setattr
= proc_setattr
,
461 extern struct seq_operations mounts_op
;
467 static int mounts_open(struct inode
*inode
, struct file
*file
)
469 struct task_struct
*task
= get_proc_task(inode
);
470 struct namespace *namespace = NULL
;
471 struct proc_mounts
*p
;
476 namespace = task
->namespace;
478 get_namespace(namespace);
480 put_task_struct(task
);
485 p
= kmalloc(sizeof(struct proc_mounts
), GFP_KERNEL
);
487 file
->private_data
= &p
->m
;
488 ret
= seq_open(file
, &mounts_op
);
490 p
->m
.private = namespace;
491 p
->event
= namespace->event
;
496 put_namespace(namespace);
501 static int mounts_release(struct inode
*inode
, struct file
*file
)
503 struct seq_file
*m
= file
->private_data
;
504 struct namespace *namespace = m
->private;
505 put_namespace(namespace);
506 return seq_release(inode
, file
);
509 static unsigned mounts_poll(struct file
*file
, poll_table
*wait
)
511 struct proc_mounts
*p
= file
->private_data
;
512 struct namespace *ns
= p
->m
.private;
515 poll_wait(file
, &ns
->poll
, wait
);
517 spin_lock(&vfsmount_lock
);
518 if (p
->event
!= ns
->event
) {
519 p
->event
= ns
->event
;
522 spin_unlock(&vfsmount_lock
);
527 static struct file_operations proc_mounts_operations
= {
531 .release
= mounts_release
,
535 extern struct seq_operations mountstats_op
;
536 static int mountstats_open(struct inode
*inode
, struct file
*file
)
538 int ret
= seq_open(file
, &mountstats_op
);
541 struct seq_file
*m
= file
->private_data
;
542 struct namespace *namespace = NULL
;
543 struct task_struct
*task
= get_proc_task(inode
);
547 namespace = task
->namespace;
549 get_namespace(namespace);
551 put_task_struct(task
);
555 m
->private = namespace;
557 seq_release(inode
, file
);
564 static struct file_operations proc_mountstats_operations
= {
565 .open
= mountstats_open
,
568 .release
= mounts_release
,
571 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
573 static ssize_t
proc_info_read(struct file
* file
, char __user
* buf
,
574 size_t count
, loff_t
*ppos
)
576 struct inode
* inode
= file
->f_dentry
->d_inode
;
579 struct task_struct
*task
= get_proc_task(inode
);
585 if (count
> PROC_BLOCK_SIZE
)
586 count
= PROC_BLOCK_SIZE
;
589 if (!(page
= __get_free_page(GFP_KERNEL
)))
592 length
= PROC_I(inode
)->op
.proc_read(task
, (char*)page
);
595 length
= simple_read_from_buffer(buf
, count
, ppos
, (char *)page
, length
);
598 put_task_struct(task
);
603 static struct file_operations proc_info_file_operations
= {
604 .read
= proc_info_read
,
607 static int mem_open(struct inode
* inode
, struct file
* file
)
609 file
->private_data
= (void*)((long)current
->self_exec_id
);
613 static ssize_t
mem_read(struct file
* file
, char __user
* buf
,
614 size_t count
, loff_t
*ppos
)
616 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
618 unsigned long src
= *ppos
;
620 struct mm_struct
*mm
;
625 if (!MAY_PTRACE(task
) || !ptrace_may_attach(task
))
629 page
= (char *)__get_free_page(GFP_USER
);
635 mm
= get_task_mm(task
);
641 if (file
->private_data
!= (void*)((long)current
->self_exec_id
))
647 int this_len
, retval
;
649 this_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
650 retval
= access_process_vm(task
, src
, page
, this_len
, 0);
651 if (!retval
|| !MAY_PTRACE(task
) || !ptrace_may_attach(task
)) {
657 if (copy_to_user(buf
, page
, retval
)) {
672 free_page((unsigned long) page
);
674 put_task_struct(task
);
679 #define mem_write NULL
682 /* This is a security hazard */
683 static ssize_t
mem_write(struct file
* file
, const char * buf
,
684 size_t count
, loff_t
*ppos
)
688 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
689 unsigned long dst
= *ppos
;
695 if (!MAY_PTRACE(task
) || !ptrace_may_attach(task
))
699 page
= (char *)__get_free_page(GFP_USER
);
705 int this_len
, retval
;
707 this_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
708 if (copy_from_user(page
, buf
, this_len
)) {
712 retval
= access_process_vm(task
, dst
, page
, this_len
, 1);
724 free_page((unsigned long) page
);
726 put_task_struct(task
);
732 static loff_t
mem_lseek(struct file
* file
, loff_t offset
, int orig
)
736 file
->f_pos
= offset
;
739 file
->f_pos
+= offset
;
744 force_successful_syscall_return();
748 static struct file_operations proc_mem_operations
= {
755 static ssize_t
oom_adjust_read(struct file
*file
, char __user
*buf
,
756 size_t count
, loff_t
*ppos
)
758 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
759 char buffer
[PROC_NUMBUF
];
762 loff_t __ppos
= *ppos
;
766 oom_adjust
= task
->oomkilladj
;
767 put_task_struct(task
);
769 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", oom_adjust
);
772 if (count
> len
-__ppos
)
774 if (copy_to_user(buf
, buffer
+ __ppos
, count
))
776 *ppos
= __ppos
+ count
;
780 static ssize_t
oom_adjust_write(struct file
*file
, const char __user
*buf
,
781 size_t count
, loff_t
*ppos
)
783 struct task_struct
*task
;
784 char buffer
[PROC_NUMBUF
], *end
;
787 if (!capable(CAP_SYS_RESOURCE
))
789 memset(buffer
, 0, sizeof(buffer
));
790 if (count
> sizeof(buffer
) - 1)
791 count
= sizeof(buffer
) - 1;
792 if (copy_from_user(buffer
, buf
, count
))
794 oom_adjust
= simple_strtol(buffer
, &end
, 0);
795 if ((oom_adjust
< -16 || oom_adjust
> 15) && oom_adjust
!= OOM_DISABLE
)
799 task
= get_proc_task(file
->f_dentry
->d_inode
);
802 task
->oomkilladj
= oom_adjust
;
803 put_task_struct(task
);
804 if (end
- buffer
== 0)
809 static struct file_operations proc_oom_adjust_operations
= {
810 .read
= oom_adjust_read
,
811 .write
= oom_adjust_write
,
814 #ifdef CONFIG_AUDITSYSCALL
816 static ssize_t
proc_loginuid_read(struct file
* file
, char __user
* buf
,
817 size_t count
, loff_t
*ppos
)
819 struct inode
* inode
= file
->f_dentry
->d_inode
;
820 struct task_struct
*task
= get_proc_task(inode
);
822 char tmpbuf
[TMPBUFLEN
];
826 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
827 audit_get_loginuid(task
->audit_context
));
828 put_task_struct(task
);
829 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
832 static ssize_t
proc_loginuid_write(struct file
* file
, const char __user
* buf
,
833 size_t count
, loff_t
*ppos
)
835 struct inode
* inode
= file
->f_dentry
->d_inode
;
840 if (!capable(CAP_AUDIT_CONTROL
))
843 if (current
!= pid_task(proc_pid(inode
), PIDTYPE_PID
))
846 if (count
>= PAGE_SIZE
)
847 count
= PAGE_SIZE
- 1;
850 /* No partial writes. */
853 page
= (char*)__get_free_page(GFP_USER
);
857 if (copy_from_user(page
, buf
, count
))
861 loginuid
= simple_strtoul(page
, &tmp
, 10);
867 length
= audit_set_loginuid(current
, loginuid
);
868 if (likely(length
== 0))
872 free_page((unsigned long) page
);
876 static struct file_operations proc_loginuid_operations
= {
877 .read
= proc_loginuid_read
,
878 .write
= proc_loginuid_write
,
882 #ifdef CONFIG_SECCOMP
883 static ssize_t
seccomp_read(struct file
*file
, char __user
*buf
,
884 size_t count
, loff_t
*ppos
)
886 struct task_struct
*tsk
= get_proc_task(file
->f_dentry
->d_inode
);
888 loff_t __ppos
= *ppos
;
893 /* no need to print the trailing zero, so use only len */
894 len
= sprintf(__buf
, "%u\n", tsk
->seccomp
.mode
);
895 put_task_struct(tsk
);
898 if (count
> len
- __ppos
)
899 count
= len
- __ppos
;
900 if (copy_to_user(buf
, __buf
+ __ppos
, count
))
902 *ppos
= __ppos
+ count
;
906 static ssize_t
seccomp_write(struct file
*file
, const char __user
*buf
,
907 size_t count
, loff_t
*ppos
)
909 struct task_struct
*tsk
= get_proc_task(file
->f_dentry
->d_inode
);
910 char __buf
[20], *end
;
911 unsigned int seccomp_mode
;
918 /* can set it only once to be even more secure */
920 if (unlikely(tsk
->seccomp
.mode
))
924 memset(__buf
, 0, sizeof(__buf
));
925 count
= min(count
, sizeof(__buf
) - 1);
926 if (copy_from_user(__buf
, buf
, count
))
929 seccomp_mode
= simple_strtoul(__buf
, &end
, 0);
933 if (seccomp_mode
&& seccomp_mode
<= NR_SECCOMP_MODES
) {
934 tsk
->seccomp
.mode
= seccomp_mode
;
935 set_tsk_thread_flag(tsk
, TIF_SECCOMP
);
939 if (unlikely(!(end
- __buf
)))
941 result
= end
- __buf
;
943 put_task_struct(tsk
);
948 static struct file_operations proc_seccomp_operations
= {
949 .read
= seccomp_read
,
950 .write
= seccomp_write
,
952 #endif /* CONFIG_SECCOMP */
954 static void *proc_pid_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
956 struct inode
*inode
= dentry
->d_inode
;
959 /* We don't need a base pointer in the /proc filesystem */
962 /* Are we allowed to snoop on the tasks file descriptors? */
963 if (!proc_fd_access_allowed(inode
))
966 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &nd
->dentry
, &nd
->mnt
);
967 nd
->last_type
= LAST_BIND
;
969 return ERR_PTR(error
);
972 static int do_proc_readlink(struct dentry
*dentry
, struct vfsmount
*mnt
,
973 char __user
*buffer
, int buflen
)
975 struct inode
* inode
;
976 char *tmp
= (char*)__get_free_page(GFP_KERNEL
), *path
;
982 inode
= dentry
->d_inode
;
983 path
= d_path(dentry
, mnt
, tmp
, PAGE_SIZE
);
987 len
= tmp
+ PAGE_SIZE
- 1 - path
;
991 if (copy_to_user(buffer
, path
, len
))
994 free_page((unsigned long)tmp
);
998 static int proc_pid_readlink(struct dentry
* dentry
, char __user
* buffer
, int buflen
)
1000 int error
= -EACCES
;
1001 struct inode
*inode
= dentry
->d_inode
;
1003 struct vfsmount
*mnt
= NULL
;
1005 /* Are we allowed to snoop on the tasks file descriptors? */
1006 if (!proc_fd_access_allowed(inode
))
1009 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &de
, &mnt
);
1013 error
= do_proc_readlink(de
, mnt
, buffer
, buflen
);
1020 static struct inode_operations proc_pid_link_inode_operations
= {
1021 .readlink
= proc_pid_readlink
,
1022 .follow_link
= proc_pid_follow_link
,
1023 .setattr
= proc_setattr
,
1027 /* building an inode */
1029 static int task_dumpable(struct task_struct
*task
)
1032 struct mm_struct
*mm
;
1037 dumpable
= mm
->dumpable
;
1045 static struct inode
*proc_pid_make_inode(struct super_block
* sb
, struct task_struct
*task
, int ino
)
1047 struct inode
* inode
;
1048 struct proc_inode
*ei
;
1050 /* We need a new inode */
1052 inode
= new_inode(sb
);
1058 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1059 inode
->i_ino
= fake_ino(task
->pid
, ino
);
1060 inode
->i_op
= &proc_def_inode_operations
;
1063 * grab the reference to task.
1065 ei
->pid
= get_pid(task
->pids
[PIDTYPE_PID
].pid
);
1071 if (task_dumpable(task
)) {
1072 inode
->i_uid
= task
->euid
;
1073 inode
->i_gid
= task
->egid
;
1075 security_task_to_inode(task
, inode
);
1085 static int pid_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
1087 struct inode
*inode
= dentry
->d_inode
;
1088 struct task_struct
*task
;
1089 generic_fillattr(inode
, stat
);
1094 task
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
1096 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1097 task_dumpable(task
)) {
1098 stat
->uid
= task
->euid
;
1099 stat
->gid
= task
->egid
;
1109 * Exceptional case: normally we are not allowed to unhash a busy
1110 * directory. In this case, however, we can do it - no aliasing problems
1111 * due to the way we treat inodes.
1113 * Rewrite the inode's ownerships here because the owning task may have
1114 * performed a setuid(), etc.
1116 * Before the /proc/pid/status file was created the only way to read
1117 * the effective uid of a /process was to stat /proc/pid. Reading
1118 * /proc/pid/status is slow enough that procps and other packages
1119 * kept stating /proc/pid. To keep the rules in /proc simple I have
1120 * made this apply to all per process world readable and executable
1123 static int pid_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1125 struct inode
*inode
= dentry
->d_inode
;
1126 struct task_struct
*task
= get_proc_task(inode
);
1128 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1129 task_dumpable(task
)) {
1130 inode
->i_uid
= task
->euid
;
1131 inode
->i_gid
= task
->egid
;
1136 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1137 security_task_to_inode(task
, inode
);
1138 put_task_struct(task
);
1145 static int pid_delete_dentry(struct dentry
* dentry
)
1147 /* Is the task we represent dead?
1148 * If so, then don't put the dentry on the lru list,
1149 * kill it immediately.
1151 return !proc_pid(dentry
->d_inode
)->tasks
[PIDTYPE_PID
].first
;
1154 static struct dentry_operations pid_dentry_operations
=
1156 .d_revalidate
= pid_revalidate
,
1157 .d_delete
= pid_delete_dentry
,
1162 static unsigned name_to_int(struct dentry
*dentry
)
1164 const char *name
= dentry
->d_name
.name
;
1165 int len
= dentry
->d_name
.len
;
1168 if (len
> 1 && *name
== '0')
1171 unsigned c
= *name
++ - '0';
1174 if (n
>= (~0U-9)/10)
1184 static int proc_fd_link(struct inode
*inode
, struct dentry
**dentry
, struct vfsmount
**mnt
)
1186 struct task_struct
*task
= get_proc_task(inode
);
1187 struct files_struct
*files
= NULL
;
1189 int fd
= proc_fd(inode
);
1192 files
= get_files_struct(task
);
1193 put_task_struct(task
);
1197 * We are not taking a ref to the file structure, so we must
1200 spin_lock(&files
->file_lock
);
1201 file
= fcheck_files(files
, fd
);
1203 *mnt
= mntget(file
->f_vfsmnt
);
1204 *dentry
= dget(file
->f_dentry
);
1205 spin_unlock(&files
->file_lock
);
1206 put_files_struct(files
);
1209 spin_unlock(&files
->file_lock
);
1210 put_files_struct(files
);
1215 static int tid_fd_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1217 struct inode
*inode
= dentry
->d_inode
;
1218 struct task_struct
*task
= get_proc_task(inode
);
1219 int fd
= proc_fd(inode
);
1220 struct files_struct
*files
;
1223 files
= get_files_struct(task
);
1226 if (fcheck_files(files
, fd
)) {
1228 put_files_struct(files
);
1229 if (task_dumpable(task
)) {
1230 inode
->i_uid
= task
->euid
;
1231 inode
->i_gid
= task
->egid
;
1236 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1237 security_task_to_inode(task
, inode
);
1238 put_task_struct(task
);
1242 put_files_struct(files
);
1244 put_task_struct(task
);
1250 static struct dentry_operations tid_fd_dentry_operations
=
1252 .d_revalidate
= tid_fd_revalidate
,
1253 .d_delete
= pid_delete_dentry
,
1257 static struct dentry
*proc_lookupfd(struct inode
* dir
, struct dentry
* dentry
, struct nameidata
*nd
)
1259 struct task_struct
*task
= get_proc_task(dir
);
1260 unsigned fd
= name_to_int(dentry
);
1261 struct dentry
*result
= ERR_PTR(-ENOENT
);
1263 struct files_struct
* files
;
1264 struct inode
*inode
;
1265 struct proc_inode
*ei
;
1272 inode
= proc_pid_make_inode(dir
->i_sb
, task
, PROC_TID_FD_DIR
+fd
);
1277 files
= get_files_struct(task
);
1280 inode
->i_mode
= S_IFLNK
;
1283 * We are not taking a ref to the file structure, so we must
1286 spin_lock(&files
->file_lock
);
1287 file
= fcheck_files(files
, fd
);
1290 if (file
->f_mode
& 1)
1291 inode
->i_mode
|= S_IRUSR
| S_IXUSR
;
1292 if (file
->f_mode
& 2)
1293 inode
->i_mode
|= S_IWUSR
| S_IXUSR
;
1294 spin_unlock(&files
->file_lock
);
1295 put_files_struct(files
);
1296 inode
->i_op
= &proc_pid_link_inode_operations
;
1298 ei
->op
.proc_get_link
= proc_fd_link
;
1299 dentry
->d_op
= &tid_fd_dentry_operations
;
1300 d_add(dentry
, inode
);
1301 /* Close the race of the process dying before we return the dentry */
1302 if (tid_fd_revalidate(dentry
, NULL
))
1305 put_task_struct(task
);
1310 spin_unlock(&files
->file_lock
);
1311 put_files_struct(files
);
1317 static int proc_readfd(struct file
* filp
, void * dirent
, filldir_t filldir
)
1319 struct dentry
*dentry
= filp
->f_dentry
;
1320 struct inode
*inode
= dentry
->d_inode
;
1321 struct task_struct
*p
= get_proc_task(inode
);
1322 unsigned int fd
, tid
, ino
;
1324 char buf
[PROC_NUMBUF
];
1325 struct files_struct
* files
;
1326 struct fdtable
*fdt
;
1337 if (filldir(dirent
, ".", 1, 0, inode
->i_ino
, DT_DIR
) < 0)
1341 ino
= parent_ino(dentry
);
1342 if (filldir(dirent
, "..", 2, 1, ino
, DT_DIR
) < 0)
1346 files
= get_files_struct(p
);
1350 fdt
= files_fdtable(files
);
1351 for (fd
= filp
->f_pos
-2;
1353 fd
++, filp
->f_pos
++) {
1356 if (!fcheck_files(files
, fd
))
1364 buf
[j
] = '0' + (i
% 10);
1368 ino
= fake_ino(tid
, PROC_TID_FD_DIR
+ fd
);
1369 if (filldir(dirent
, buf
+j
, PROC_NUMBUF
-j
, fd
+2, ino
, DT_LNK
) < 0) {
1376 put_files_struct(files
);
1384 static struct file_operations proc_fd_operations
= {
1385 .read
= generic_read_dir
,
1386 .readdir
= proc_readfd
,
1390 * proc directories can do almost nothing..
1392 static struct inode_operations proc_fd_inode_operations
= {
1393 .lookup
= proc_lookupfd
,
1394 .setattr
= proc_setattr
,
1398 static struct dentry
*proc_pident_lookup(struct inode
*dir
,
1399 struct dentry
*dentry
,
1400 struct pid_entry
*ents
)
1402 struct inode
*inode
;
1403 struct dentry
*error
;
1404 struct task_struct
*task
= get_proc_task(dir
);
1405 struct pid_entry
*p
;
1406 struct proc_inode
*ei
;
1408 error
= ERR_PTR(-ENOENT
);
1415 * Yes, it does not scale. And it should not. Don't add
1416 * new entries into /proc/<tgid>/ without very good reasons.
1418 for (p
= ents
; p
->name
; p
++) {
1419 if (p
->len
!= dentry
->d_name
.len
)
1421 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
1427 error
= ERR_PTR(-EINVAL
);
1428 inode
= proc_pid_make_inode(dir
->i_sb
, task
, p
->type
);
1433 inode
->i_mode
= p
->mode
;
1434 if (S_ISDIR(inode
->i_mode
))
1435 inode
->i_nlink
= 2; /* Use getattr to fix if necessary */
1437 inode
->i_op
= p
->iop
;
1439 inode
->i_fop
= p
->fop
;
1441 dentry
->d_op
= &pid_dentry_operations
;
1442 d_add(dentry
, inode
);
1443 /* Close the race of the process dying before we return the dentry */
1444 if (pid_revalidate(dentry
, NULL
))
1447 put_task_struct(task
);
1452 static int proc_pident_readdir(struct file
*filp
,
1453 void *dirent
, filldir_t filldir
,
1454 struct pid_entry
*ents
, unsigned int nents
)
1458 struct dentry
*dentry
= filp
->f_dentry
;
1459 struct inode
*inode
= dentry
->d_inode
;
1460 struct task_struct
*task
= get_proc_task(inode
);
1461 struct pid_entry
*p
;
1471 put_task_struct(task
);
1476 if (filldir(dirent
, ".", 1, i
, ino
, DT_DIR
) < 0)
1482 ino
= parent_ino(dentry
);
1483 if (filldir(dirent
, "..", 2, i
, ino
, DT_DIR
) < 0)
1496 if (filldir(dirent
, p
->name
, p
->len
, filp
->f_pos
,
1497 fake_ino(pid
, p
->type
), p
->mode
>> 12) < 0)
1509 #ifdef CONFIG_SECURITY
1510 static ssize_t
proc_pid_attr_read(struct file
* file
, char __user
* buf
,
1511 size_t count
, loff_t
*ppos
)
1513 struct inode
* inode
= file
->f_dentry
->d_inode
;
1516 struct task_struct
*task
= get_proc_task(inode
);
1522 if (count
> PAGE_SIZE
)
1525 if (!(page
= __get_free_page(GFP_KERNEL
)))
1528 length
= security_getprocattr(task
,
1529 (char*)file
->f_dentry
->d_name
.name
,
1530 (void*)page
, count
);
1532 length
= simple_read_from_buffer(buf
, count
, ppos
, (char *)page
, length
);
1535 put_task_struct(task
);
1540 static ssize_t
proc_pid_attr_write(struct file
* file
, const char __user
* buf
,
1541 size_t count
, loff_t
*ppos
)
1543 struct inode
* inode
= file
->f_dentry
->d_inode
;
1546 struct task_struct
*task
= get_proc_task(inode
);
1551 if (count
> PAGE_SIZE
)
1554 /* No partial writes. */
1560 page
= (char*)__get_free_page(GFP_USER
);
1565 if (copy_from_user(page
, buf
, count
))
1568 length
= security_setprocattr(task
,
1569 (char*)file
->f_dentry
->d_name
.name
,
1570 (void*)page
, count
);
1572 free_page((unsigned long) page
);
1574 put_task_struct(task
);
1579 static struct file_operations proc_pid_attr_operations
= {
1580 .read
= proc_pid_attr_read
,
1581 .write
= proc_pid_attr_write
,
1584 static struct pid_entry tgid_attr_stuff
[] = {
1585 REG(PROC_TGID_ATTR_CURRENT
, "current", S_IRUGO
|S_IWUGO
, pid_attr
),
1586 REG(PROC_TGID_ATTR_PREV
, "prev", S_IRUGO
, pid_attr
),
1587 REG(PROC_TGID_ATTR_EXEC
, "exec", S_IRUGO
|S_IWUGO
, pid_attr
),
1588 REG(PROC_TGID_ATTR_FSCREATE
, "fscreate", S_IRUGO
|S_IWUGO
, pid_attr
),
1589 REG(PROC_TGID_ATTR_KEYCREATE
, "keycreate", S_IRUGO
|S_IWUGO
, pid_attr
),
1590 REG(PROC_TGID_ATTR_SOCKCREATE
, "sockcreate", S_IRUGO
|S_IWUGO
, pid_attr
),
1593 static struct pid_entry tid_attr_stuff
[] = {
1594 REG(PROC_TID_ATTR_CURRENT
, "current", S_IRUGO
|S_IWUGO
, pid_attr
),
1595 REG(PROC_TID_ATTR_PREV
, "prev", S_IRUGO
, pid_attr
),
1596 REG(PROC_TID_ATTR_EXEC
, "exec", S_IRUGO
|S_IWUGO
, pid_attr
),
1597 REG(PROC_TID_ATTR_FSCREATE
, "fscreate", S_IRUGO
|S_IWUGO
, pid_attr
),
1598 REG(PROC_TID_ATTR_KEYCREATE
, "keycreate", S_IRUGO
|S_IWUGO
, pid_attr
),
1599 REG(PROC_TID_ATTR_SOCKCREATE
, "sockcreate", S_IRUGO
|S_IWUGO
, pid_attr
),
1603 static int proc_tgid_attr_readdir(struct file
* filp
,
1604 void * dirent
, filldir_t filldir
)
1606 return proc_pident_readdir(filp
,dirent
,filldir
,
1607 tgid_attr_stuff
,ARRAY_SIZE(tgid_attr_stuff
));
1610 static int proc_tid_attr_readdir(struct file
* filp
,
1611 void * dirent
, filldir_t filldir
)
1613 return proc_pident_readdir(filp
,dirent
,filldir
,
1614 tid_attr_stuff
,ARRAY_SIZE(tid_attr_stuff
));
1617 static struct file_operations proc_tgid_attr_operations
= {
1618 .read
= generic_read_dir
,
1619 .readdir
= proc_tgid_attr_readdir
,
1622 static struct file_operations proc_tid_attr_operations
= {
1623 .read
= generic_read_dir
,
1624 .readdir
= proc_tid_attr_readdir
,
1627 static struct dentry
*proc_tgid_attr_lookup(struct inode
*dir
,
1628 struct dentry
*dentry
, struct nameidata
*nd
)
1630 return proc_pident_lookup(dir
, dentry
, tgid_attr_stuff
);
1633 static struct dentry
*proc_tid_attr_lookup(struct inode
*dir
,
1634 struct dentry
*dentry
, struct nameidata
*nd
)
1636 return proc_pident_lookup(dir
, dentry
, tid_attr_stuff
);
1639 static struct inode_operations proc_tgid_attr_inode_operations
= {
1640 .lookup
= proc_tgid_attr_lookup
,
1641 .getattr
= pid_getattr
,
1642 .setattr
= proc_setattr
,
1645 static struct inode_operations proc_tid_attr_inode_operations
= {
1646 .lookup
= proc_tid_attr_lookup
,
1647 .getattr
= pid_getattr
,
1648 .setattr
= proc_setattr
,
1655 static int proc_self_readlink(struct dentry
*dentry
, char __user
*buffer
,
1658 char tmp
[PROC_NUMBUF
];
1659 sprintf(tmp
, "%d", current
->tgid
);
1660 return vfs_readlink(dentry
,buffer
,buflen
,tmp
);
1663 static void *proc_self_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
1665 char tmp
[PROC_NUMBUF
];
1666 sprintf(tmp
, "%d", current
->tgid
);
1667 return ERR_PTR(vfs_follow_link(nd
,tmp
));
1670 static struct inode_operations proc_self_inode_operations
= {
1671 .readlink
= proc_self_readlink
,
1672 .follow_link
= proc_self_follow_link
,
1678 static struct file_operations proc_task_operations
;
1679 static struct inode_operations proc_task_inode_operations
;
1681 static struct pid_entry tgid_base_stuff
[] = {
1682 DIR(PROC_TGID_TASK
, "task", S_IRUGO
|S_IXUGO
, task
),
1683 DIR(PROC_TGID_FD
, "fd", S_IRUSR
|S_IXUSR
, fd
),
1684 INF(PROC_TGID_ENVIRON
, "environ", S_IRUSR
, pid_environ
),
1685 INF(PROC_TGID_AUXV
, "auxv", S_IRUSR
, pid_auxv
),
1686 INF(PROC_TGID_STATUS
, "status", S_IRUGO
, pid_status
),
1687 INF(PROC_TGID_CMDLINE
, "cmdline", S_IRUGO
, pid_cmdline
),
1688 INF(PROC_TGID_STAT
, "stat", S_IRUGO
, tgid_stat
),
1689 INF(PROC_TGID_STATM
, "statm", S_IRUGO
, pid_statm
),
1690 REG(PROC_TGID_MAPS
, "maps", S_IRUGO
, maps
),
1692 REG(PROC_TGID_NUMA_MAPS
, "numa_maps", S_IRUGO
, numa_maps
),
1694 REG(PROC_TGID_MEM
, "mem", S_IRUSR
|S_IWUSR
, mem
),
1695 #ifdef CONFIG_SECCOMP
1696 REG(PROC_TGID_SECCOMP
, "seccomp", S_IRUSR
|S_IWUSR
, seccomp
),
1698 LNK(PROC_TGID_CWD
, "cwd", cwd
),
1699 LNK(PROC_TGID_ROOT
, "root", root
),
1700 LNK(PROC_TGID_EXE
, "exe", exe
),
1701 REG(PROC_TGID_MOUNTS
, "mounts", S_IRUGO
, mounts
),
1702 REG(PROC_TGID_MOUNTSTATS
, "mountstats", S_IRUSR
, mountstats
),
1704 REG(PROC_TGID_SMAPS
, "smaps", S_IRUGO
, smaps
),
1706 #ifdef CONFIG_SECURITY
1707 DIR(PROC_TGID_ATTR
, "attr", S_IRUGO
|S_IXUGO
, tgid_attr
),
1709 #ifdef CONFIG_KALLSYMS
1710 INF(PROC_TGID_WCHAN
, "wchan", S_IRUGO
, pid_wchan
),
1712 #ifdef CONFIG_SCHEDSTATS
1713 INF(PROC_TGID_SCHEDSTAT
, "schedstat", S_IRUGO
, pid_schedstat
),
1715 #ifdef CONFIG_CPUSETS
1716 REG(PROC_TGID_CPUSET
, "cpuset", S_IRUGO
, cpuset
),
1718 INF(PROC_TGID_OOM_SCORE
, "oom_score", S_IRUGO
, oom_score
),
1719 REG(PROC_TGID_OOM_ADJUST
, "oom_adj", S_IRUGO
|S_IWUSR
, oom_adjust
),
1720 #ifdef CONFIG_AUDITSYSCALL
1721 REG(PROC_TGID_LOGINUID
, "loginuid", S_IWUSR
|S_IRUGO
, loginuid
),
1726 static int proc_tgid_base_readdir(struct file
* filp
,
1727 void * dirent
, filldir_t filldir
)
1729 return proc_pident_readdir(filp
,dirent
,filldir
,
1730 tgid_base_stuff
,ARRAY_SIZE(tgid_base_stuff
));
1733 static struct file_operations proc_tgid_base_operations
= {
1734 .read
= generic_read_dir
,
1735 .readdir
= proc_tgid_base_readdir
,
1738 static struct dentry
*proc_tgid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
1739 return proc_pident_lookup(dir
, dentry
, tgid_base_stuff
);
1742 static struct inode_operations proc_tgid_base_inode_operations
= {
1743 .lookup
= proc_tgid_base_lookup
,
1744 .getattr
= pid_getattr
,
1745 .setattr
= proc_setattr
,
1749 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
1751 * @task: task that should be flushed.
1753 * Looks in the dcache for
1755 * /proc/@tgid/task/@pid
1756 * if either directory is present flushes it and all of it'ts children
1759 * It is safe and reasonable to cache /proc entries for a task until
1760 * that task exits. After that they just clog up the dcache with
1761 * useless entries, possibly causing useful dcache entries to be
1762 * flushed instead. This routine is proved to flush those useless
1763 * dcache entries at process exit time.
1765 * NOTE: This routine is just an optimization so it does not guarantee
1766 * that no dcache entries will exist at process exit time it
1767 * just makes it very unlikely that any will persist.
1769 void proc_flush_task(struct task_struct
*task
)
1771 struct dentry
*dentry
, *leader
, *dir
;
1772 char buf
[PROC_NUMBUF
];
1776 name
.len
= snprintf(buf
, sizeof(buf
), "%d", task
->pid
);
1777 dentry
= d_hash_and_lookup(proc_mnt
->mnt_root
, &name
);
1779 shrink_dcache_parent(dentry
);
1784 if (thread_group_leader(task
))
1788 name
.len
= snprintf(buf
, sizeof(buf
), "%d", task
->tgid
);
1789 leader
= d_hash_and_lookup(proc_mnt
->mnt_root
, &name
);
1794 name
.len
= strlen(name
.name
);
1795 dir
= d_hash_and_lookup(leader
, &name
);
1797 goto out_put_leader
;
1800 name
.len
= snprintf(buf
, sizeof(buf
), "%d", task
->pid
);
1801 dentry
= d_hash_and_lookup(dir
, &name
);
1803 shrink_dcache_parent(dentry
);
1816 struct dentry
*proc_pid_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
1818 struct dentry
*result
= ERR_PTR(-ENOENT
);
1819 struct task_struct
*task
;
1820 struct inode
*inode
;
1821 struct proc_inode
*ei
;
1824 if (dentry
->d_name
.len
== 4 && !memcmp(dentry
->d_name
.name
,"self",4)) {
1825 inode
= new_inode(dir
->i_sb
);
1827 return ERR_PTR(-ENOMEM
);
1829 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1830 inode
->i_ino
= fake_ino(0, PROC_TGID_INO
);
1832 inode
->i_mode
= S_IFLNK
|S_IRWXUGO
;
1833 inode
->i_uid
= inode
->i_gid
= 0;
1835 inode
->i_op
= &proc_self_inode_operations
;
1836 d_add(dentry
, inode
);
1839 tgid
= name_to_int(dentry
);
1844 task
= find_task_by_pid(tgid
);
1846 get_task_struct(task
);
1851 inode
= proc_pid_make_inode(dir
->i_sb
, task
, PROC_TGID_INO
);
1855 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
1856 inode
->i_op
= &proc_tgid_base_inode_operations
;
1857 inode
->i_fop
= &proc_tgid_base_operations
;
1858 inode
->i_flags
|=S_IMMUTABLE
;
1859 #ifdef CONFIG_SECURITY
1865 dentry
->d_op
= &pid_dentry_operations
;
1867 d_add(dentry
, inode
);
1868 /* Close the race of the process dying before we return the dentry */
1869 if (pid_revalidate(dentry
, NULL
))
1873 put_task_struct(task
);
1879 * Find the first task with tgid >= tgid
1882 static struct task_struct
*next_tgid(unsigned int tgid
)
1884 struct task_struct
*task
;
1890 pid
= find_ge_pid(tgid
);
1893 task
= pid_task(pid
, PIDTYPE_PID
);
1894 /* What we to know is if the pid we have find is the
1895 * pid of a thread_group_leader. Testing for task
1896 * being a thread_group_leader is the obvious thing
1897 * todo but there is a window when it fails, due to
1898 * the pid transfer logic in de_thread.
1900 * So we perform the straight forward test of seeing
1901 * if the pid we have found is the pid of a thread
1902 * group leader, and don't worry if the task we have
1903 * found doesn't happen to be a thread group leader.
1904 * As we don't care in the case of readdir.
1906 if (!task
|| !has_group_leader_pid(task
))
1908 get_task_struct(task
);
1914 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + (1 /* /proc/self */))
1916 /* for the /proc/ directory itself, after non-process stuff has been done */
1917 int proc_pid_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
1919 char buf
[PROC_NUMBUF
];
1920 unsigned int nr
= filp
->f_pos
- FIRST_PROCESS_ENTRY
;
1921 struct task_struct
*task
;
1925 ino_t ino
= fake_ino(0,PROC_TGID_INO
);
1926 if (filldir(dirent
, "self", 4, filp
->f_pos
, ino
, DT_LNK
) < 0)
1932 tgid
= filp
->f_pos
- TGID_OFFSET
;
1933 for (task
= next_tgid(tgid
);
1935 put_task_struct(task
), task
= next_tgid(tgid
+ 1)) {
1939 filp
->f_pos
= tgid
+ TGID_OFFSET
;
1940 len
= snprintf(buf
, sizeof(buf
), "%d", tgid
);
1941 ino
= fake_ino(tgid
, PROC_TGID_INO
);
1942 if (filldir(dirent
, buf
, len
, filp
->f_pos
, ino
, DT_DIR
) < 0) {
1943 put_task_struct(task
);
1947 filp
->f_pos
= PID_MAX_LIMIT
+ TGID_OFFSET
;
1955 static struct pid_entry tid_base_stuff
[] = {
1956 DIR(PROC_TID_FD
, "fd", S_IRUSR
|S_IXUSR
, fd
),
1957 INF(PROC_TID_ENVIRON
, "environ", S_IRUSR
, pid_environ
),
1958 INF(PROC_TID_AUXV
, "auxv", S_IRUSR
, pid_auxv
),
1959 INF(PROC_TID_STATUS
, "status", S_IRUGO
, pid_status
),
1960 INF(PROC_TID_CMDLINE
, "cmdline", S_IRUGO
, pid_cmdline
),
1961 INF(PROC_TID_STAT
, "stat", S_IRUGO
, tid_stat
),
1962 INF(PROC_TID_STATM
, "statm", S_IRUGO
, pid_statm
),
1963 REG(PROC_TID_MAPS
, "maps", S_IRUGO
, maps
),
1965 REG(PROC_TID_NUMA_MAPS
, "numa_maps", S_IRUGO
, numa_maps
),
1967 REG(PROC_TID_MEM
, "mem", S_IRUSR
|S_IWUSR
, mem
),
1968 #ifdef CONFIG_SECCOMP
1969 REG(PROC_TID_SECCOMP
, "seccomp", S_IRUSR
|S_IWUSR
, seccomp
),
1971 LNK(PROC_TID_CWD
, "cwd", cwd
),
1972 LNK(PROC_TID_ROOT
, "root", root
),
1973 LNK(PROC_TID_EXE
, "exe", exe
),
1974 REG(PROC_TID_MOUNTS
, "mounts", S_IRUGO
, mounts
),
1976 REG(PROC_TID_SMAPS
, "smaps", S_IRUGO
, smaps
),
1978 #ifdef CONFIG_SECURITY
1979 DIR(PROC_TID_ATTR
, "attr", S_IRUGO
|S_IXUGO
, tid_attr
),
1981 #ifdef CONFIG_KALLSYMS
1982 INF(PROC_TID_WCHAN
, "wchan", S_IRUGO
, pid_wchan
),
1984 #ifdef CONFIG_SCHEDSTATS
1985 INF(PROC_TID_SCHEDSTAT
, "schedstat", S_IRUGO
, pid_schedstat
),
1987 #ifdef CONFIG_CPUSETS
1988 REG(PROC_TID_CPUSET
, "cpuset", S_IRUGO
, cpuset
),
1990 INF(PROC_TID_OOM_SCORE
, "oom_score", S_IRUGO
, oom_score
),
1991 REG(PROC_TID_OOM_ADJUST
, "oom_adj", S_IRUGO
|S_IWUSR
, oom_adjust
),
1992 #ifdef CONFIG_AUDITSYSCALL
1993 REG(PROC_TID_LOGINUID
, "loginuid", S_IWUSR
|S_IRUGO
, loginuid
),
1998 static int proc_tid_base_readdir(struct file
* filp
,
1999 void * dirent
, filldir_t filldir
)
2001 return proc_pident_readdir(filp
,dirent
,filldir
,
2002 tid_base_stuff
,ARRAY_SIZE(tid_base_stuff
));
2005 static struct dentry
*proc_tid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
2006 return proc_pident_lookup(dir
, dentry
, tid_base_stuff
);
2009 static struct file_operations proc_tid_base_operations
= {
2010 .read
= generic_read_dir
,
2011 .readdir
= proc_tid_base_readdir
,
2014 static struct inode_operations proc_tid_base_inode_operations
= {
2015 .lookup
= proc_tid_base_lookup
,
2016 .getattr
= pid_getattr
,
2017 .setattr
= proc_setattr
,
2021 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
2023 struct dentry
*result
= ERR_PTR(-ENOENT
);
2024 struct task_struct
*task
;
2025 struct task_struct
*leader
= get_proc_task(dir
);
2026 struct inode
*inode
;
2032 tid
= name_to_int(dentry
);
2037 task
= find_task_by_pid(tid
);
2039 get_task_struct(task
);
2043 if (leader
->tgid
!= task
->tgid
)
2046 inode
= proc_pid_make_inode(dir
->i_sb
, task
, PROC_TID_INO
);
2051 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
2052 inode
->i_op
= &proc_tid_base_inode_operations
;
2053 inode
->i_fop
= &proc_tid_base_operations
;
2054 inode
->i_flags
|=S_IMMUTABLE
;
2055 #ifdef CONFIG_SECURITY
2061 dentry
->d_op
= &pid_dentry_operations
;
2063 d_add(dentry
, inode
);
2064 /* Close the race of the process dying before we return the dentry */
2065 if (pid_revalidate(dentry
, NULL
))
2069 put_task_struct(task
);
2071 put_task_struct(leader
);
2077 * Find the first tid of a thread group to return to user space.
2079 * Usually this is just the thread group leader, but if the users
2080 * buffer was too small or there was a seek into the middle of the
2081 * directory we have more work todo.
2083 * In the case of a short read we start with find_task_by_pid.
2085 * In the case of a seek we start with the leader and walk nr
2088 static struct task_struct
*first_tid(struct task_struct
*leader
,
2091 struct task_struct
*pos
;
2094 /* Attempt to start with the pid of a thread */
2095 if (tid
&& (nr
> 0)) {
2096 pos
= find_task_by_pid(tid
);
2097 if (pos
&& (pos
->group_leader
== leader
))
2101 /* If nr exceeds the number of threads there is nothing todo */
2103 if (nr
&& nr
>= get_nr_threads(leader
))
2106 /* If we haven't found our starting place yet start
2107 * with the leader and walk nr threads forward.
2109 for (pos
= leader
; nr
> 0; --nr
) {
2110 pos
= next_thread(pos
);
2111 if (pos
== leader
) {
2117 get_task_struct(pos
);
2124 * Find the next thread in the thread list.
2125 * Return NULL if there is an error or no next thread.
2127 * The reference to the input task_struct is released.
2129 static struct task_struct
*next_tid(struct task_struct
*start
)
2131 struct task_struct
*pos
= NULL
;
2133 if (pid_alive(start
)) {
2134 pos
= next_thread(start
);
2135 if (thread_group_leader(pos
))
2138 get_task_struct(pos
);
2141 put_task_struct(start
);
2145 /* for the /proc/TGID/task/ directories */
2146 static int proc_task_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
2148 char buf
[PROC_NUMBUF
];
2149 struct dentry
*dentry
= filp
->f_dentry
;
2150 struct inode
*inode
= dentry
->d_inode
;
2151 struct task_struct
*leader
= get_proc_task(inode
);
2152 struct task_struct
*task
;
2153 int retval
= -ENOENT
;
2156 unsigned long pos
= filp
->f_pos
; /* avoiding "long long" filp->f_pos */
2165 if (filldir(dirent
, ".", 1, pos
, ino
, DT_DIR
) < 0)
2170 ino
= parent_ino(dentry
);
2171 if (filldir(dirent
, "..", 2, pos
, ino
, DT_DIR
) < 0)
2177 /* f_version caches the tgid value that the last readdir call couldn't
2178 * return. lseek aka telldir automagically resets f_version to 0.
2180 tid
= filp
->f_version
;
2181 filp
->f_version
= 0;
2182 for (task
= first_tid(leader
, tid
, pos
- 2);
2184 task
= next_tid(task
), pos
++) {
2187 len
= snprintf(buf
, sizeof(buf
), "%d", tid
);
2188 ino
= fake_ino(tid
, PROC_TID_INO
);
2189 if (filldir(dirent
, buf
, len
, pos
, ino
, DT_DIR
< 0)) {
2190 /* returning this tgid failed, save it as the first
2191 * pid for the next readir call */
2192 filp
->f_version
= tid
;
2193 put_task_struct(task
);
2199 put_task_struct(leader
);
2204 static int proc_task_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
2206 struct inode
*inode
= dentry
->d_inode
;
2207 struct task_struct
*p
= get_proc_task(inode
);
2208 generic_fillattr(inode
, stat
);
2212 stat
->nlink
+= get_nr_threads(p
);
2220 static struct inode_operations proc_task_inode_operations
= {
2221 .lookup
= proc_task_lookup
,
2222 .getattr
= proc_task_getattr
,
2223 .setattr
= proc_setattr
,
2226 static struct file_operations proc_task_operations
= {
2227 .read
= generic_read_dir
,
2228 .readdir
= proc_task_readdir
,