2 * Copyright (c) 2000 Dag-Erling Coïdan Smørgrav
3 * Copyright (c) 1999 Pierre Beyssac
4 * Copyright (c) 1993, 1995 Jan-Simon Pendry
5 * Copyright (c) 1993, 1995
6 * The Regents of the University of California. All rights reserved.
8 * This code is derived from software contributed to Berkeley by
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. All advertising materials mentioning features or use of this software
20 * must display the following acknowledgement:
21 * This product includes software developed by the University of
22 * California, Berkeley and its contributors.
23 * 4. Neither the name of the University nor the names of its contributors
24 * may be used to endorse or promote products derived from this software
25 * without specific prior written permission.
27 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
28 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
29 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
30 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
31 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
32 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
33 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
34 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
35 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
36 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
39 * @(#)procfs_vnops.c 8.18 (Berkeley) 5/21/95
41 * $FreeBSD: src/sys/i386/linux/linprocfs/linprocfs_vnops.c,v 1.3.2.5 2001/08/12 14:29:19 rwatson Exp $
42 * $DragonFly: src/sys/emulation/linux/i386/linprocfs/linprocfs_vnops.c,v 1.42 2007/11/20 21:03:48 dillon Exp $
46 * procfs vnode interface
49 #include <sys/param.h>
50 #include <sys/systm.h>
52 #include <sys/kernel.h>
54 #include <sys/fcntl.h>
57 #include <sys/signalvar.h>
58 #include <sys/vnode.h>
59 #include <sys/mount.h>
60 #include <sys/namei.h>
61 #include <sys/dirent.h>
62 #include <sys/malloc.h>
65 #include <vm/vm_zone.h>
66 #include "linprocfs.h"
67 #include <sys/pioctl.h>
69 #include <machine/limits.h>
71 extern struct vnode
*procfs_findtextvp (struct proc
*);
73 static int linprocfs_access (struct vop_access_args
*);
74 static int linprocfs_badop (struct vop_generic_args
*);
75 static int linprocfs_bmap (struct vop_bmap_args
*);
76 static int linprocfs_close (struct vop_close_args
*);
77 static int linprocfs_getattr (struct vop_getattr_args
*);
78 static int linprocfs_inactive (struct vop_inactive_args
*);
79 static int linprocfs_ioctl (struct vop_ioctl_args
*);
80 static int linprocfs_lookup (struct vop_old_lookup_args
*);
81 static int linprocfs_open (struct vop_open_args
*);
82 static int linprocfs_print (struct vop_print_args
*);
83 static int linprocfs_readdir (struct vop_readdir_args
*);
84 static int linprocfs_readlink (struct vop_readlink_args
*);
85 static int linprocfs_reclaim (struct vop_reclaim_args
*);
86 static int linprocfs_setattr (struct vop_setattr_args
*);
88 static int linprocfs_readdir_proc(struct vop_readdir_args
*);
89 static int linprocfs_readdir_root(struct vop_readdir_args
*);
90 static int linprocfs_readdir_net(struct vop_readdir_args
*ap
);
91 static int linprocfs_readdir_sys(struct vop_readdir_args
*ap
);
92 static int linprocfs_readdir_syskernel(struct vop_readdir_args
*ap
);
95 * procfs vnode operations.
97 struct vop_ops linprocfs_vnode_vops
= {
98 .vop_default
= vop_defaultop
,
99 .vop_access
= linprocfs_access
,
100 .vop_advlock
= (void *)linprocfs_badop
,
101 .vop_bmap
= linprocfs_bmap
,
102 .vop_close
= linprocfs_close
,
103 .vop_old_create
= (void *)linprocfs_badop
,
104 .vop_getattr
= linprocfs_getattr
,
105 .vop_inactive
= linprocfs_inactive
,
106 .vop_old_link
= (void *)linprocfs_badop
,
107 .vop_old_lookup
= linprocfs_lookup
,
108 .vop_old_mkdir
= (void *)linprocfs_badop
,
109 .vop_old_mknod
= (void *)linprocfs_badop
,
110 .vop_open
= linprocfs_open
,
111 .vop_pathconf
= vop_stdpathconf
,
112 .vop_print
= linprocfs_print
,
113 .vop_read
= (void *)linprocfs_rw
,
114 .vop_readdir
= linprocfs_readdir
,
115 .vop_readlink
= linprocfs_readlink
,
116 .vop_reclaim
= linprocfs_reclaim
,
117 .vop_old_remove
= (void *)linprocfs_badop
,
118 .vop_old_rename
= (void *)linprocfs_badop
,
119 .vop_old_rmdir
= (void *)linprocfs_badop
,
120 .vop_setattr
= linprocfs_setattr
,
121 .vop_old_symlink
= (void *)linprocfs_badop
,
122 .vop_write
= (void *)linprocfs_rw
,
123 .vop_ioctl
= linprocfs_ioctl
127 * This is a list of the valid names in the
128 * process-specific sub-directories. It is
129 * used in linprocfs_lookup and linprocfs_readdir
131 static struct proc_target
{
136 int (*pt_valid
) (struct proc
*p
);
138 #define N(s) sizeof(s)-1, s
139 /* name type validp */
140 { DT_DIR
, N("."), Pproc
, NULL
},
141 { DT_DIR
, N(".."), Proot
, NULL
},
142 { DT_REG
, N("mem"), Pmem
, NULL
},
144 { DT_LNK
, N("exe"), Pexe
, NULL
},
145 { DT_LNK
, N("cwd"), Pcwd
, NULL
},
146 { DT_LNK
, N("root"), Pprocroot
, NULL
},
147 { DT_LNK
, N("fd"), Pfd
, NULL
},
149 { DT_REG
, N("stat"), Pprocstat
, NULL
},
150 { DT_REG
, N("status"), Pprocstatus
, NULL
},
151 { DT_REG
, N("maps"), Pmaps
, NULL
},
152 { DT_REG
, N("statm"), Pstatm
, NULL
},
154 { DT_REG
, N("cmdline"), Pcmdline
, NULL
},
155 { DT_REG
, N("environ"), Penviron
, NULL
},
159 static const int nproc_targets
= sizeof(proc_targets
) / sizeof(proc_targets
[0]);
161 static pid_t
atopid (const char *, u_int
);
164 * set things up for doing i/o on
165 * the pfsnode (vp). (vp) is locked
166 * on entry, and should be left locked
169 * for procfs we don't need to do anything
170 * in particular for i/o. all that is done
171 * is to support exclusive open on process
175 linprocfs_open(struct vop_open_args
*ap
)
177 struct pfsnode
*pfs
= VTOPFS(ap
->a_vp
);
180 p2
= PFIND(pfs
->pfs_pid
);
183 if (pfs
->pfs_pid
&& !PRISON_CHECK(ap
->a_cred
, p2
->p_ucred
))
186 switch (pfs
->pfs_type
) {
188 if (((pfs
->pfs_flags
& FWRITE
) && (ap
->a_mode
& O_EXCL
)) ||
189 ((pfs
->pfs_flags
& O_EXCL
) && (ap
->a_mode
& FWRITE
)))
192 if (p_trespass(ap
->a_cred
, p2
->p_ucred
))
195 if (ap
->a_mode
& FWRITE
)
196 pfs
->pfs_flags
= ap
->a_mode
& (FWRITE
|O_EXCL
);
203 return (vop_stdopen(ap
));
207 * close the pfsnode (vp) after doing i/o.
208 * (vp) is not locked on entry or exit.
210 * nothing to do for procfs other than undo
211 * any exclusive open flag (see _open above).
214 linprocfs_close(struct vop_close_args
*ap
)
216 struct pfsnode
*pfs
= VTOPFS(ap
->a_vp
);
219 switch (pfs
->pfs_type
) {
221 if ((ap
->a_fflag
& FWRITE
) && (pfs
->pfs_flags
& O_EXCL
))
222 pfs
->pfs_flags
&= ~(FWRITE
|O_EXCL
);
224 * If this is the last close, then it checks to see if
225 * the target process has PF_LINGER set in p_pfsflags,
226 * if this is *not* the case, then the process' stop flags
227 * are cleared, and the process is woken up. This is
228 * to help prevent the case where a process has been
229 * told to stop on an event, but then the requesting process
230 * has gone away or forgotten about it.
232 if ((ap
->a_vp
->v_opencount
< 2)
233 && (p
= pfind(pfs
->pfs_pid
))
234 && !(p
->p_pfsflags
& PF_LINGER
)) {
243 return (vop_stdclose(ap
));
247 * do an ioctl operation on a pfsnode (vp).
248 * (vp) is not locked on entry or exit.
251 linprocfs_ioctl(struct vop_ioctl_args
*ap
)
253 struct pfsnode
*pfs
= VTOPFS(ap
->a_vp
);
257 struct procfs_status
*psp
;
260 procp
= pfind(pfs
->pfs_pid
);
265 if (p_trespass(ap
->a_cred
, procp
->p_ucred
))
268 switch (ap
->a_command
) {
270 procp
->p_stops
|= *(unsigned int*)ap
->a_data
;
273 procp
->p_stops
&= ~*(unsigned int*)ap
->a_data
;
277 * NFLAGS is "non-suser_xxx flags" -- currently, only
278 * PFS_ISUGID ("ignore set u/g id");
280 #define NFLAGS (PF_ISUGID)
281 flags
= (unsigned char)*(unsigned int*)ap
->a_data
;
282 if (flags
& NFLAGS
&& (error
= priv_check_cred(ap
->a_cred
, PRIV_ROOT
, 0)))
284 procp
->p_pfsflags
= flags
;
287 *(unsigned int*)ap
->a_data
= (unsigned int)procp
->p_pfsflags
;
289 psp
= (struct procfs_status
*)ap
->a_data
;
290 psp
->state
= (procp
->p_step
== 0);
291 psp
->flags
= procp
->p_pfsflags
;
292 psp
->events
= procp
->p_stops
;
294 psp
->why
= procp
->p_stype
;
295 psp
->val
= procp
->p_xstat
;
297 psp
->why
= psp
->val
= 0; /* Not defined values */
301 psp
= (struct procfs_status
*)ap
->a_data
;
302 if (procp
->p_step
== 0) {
303 error
= tsleep(&procp
->p_stype
, PCATCH
, "piocwait", 0);
307 psp
->state
= 1; /* It stopped */
308 psp
->flags
= procp
->p_pfsflags
;
309 psp
->events
= procp
->p_stops
;
310 psp
->why
= procp
->p_stype
; /* why it stopped */
311 psp
->val
= procp
->p_xstat
; /* any extra info */
313 case PIOCCONT
: /* Restart a proc */
314 if (procp
->p_step
== 0)
315 return EINVAL
; /* Can only start a stopped process */
316 if ((signo
= *(int*)ap
->a_data
) != 0) {
317 if (signo
>= NSIG
|| signo
<= 0)
319 ksignal(procp
, signo
);
322 wakeup(&procp
->p_step
);
331 * do block mapping for pfsnode (vp).
332 * since we don't use the buffer cache
333 * for procfs this function should never
334 * be called. in any case, it's not clear
335 * what part of the kernel ever makes use
336 * of this function. for sanity, this is the
337 * usual no-op bmap, although returning
338 * (EIO) would be a reasonable alternative.
341 linprocfs_bmap(struct vop_bmap_args
*ap
)
343 if (ap
->a_doffsetp
!= NULL
)
344 *ap
->a_doffsetp
= ap
->a_loffset
;
345 if (ap
->a_runp
!= NULL
)
347 if (ap
->a_runb
!= NULL
)
353 * linprocfs_inactive is called when the pfsnode
354 * is vrele'd and the reference count is about
355 * to go to zero. (vp) will be on the vnode free
356 * list, so to get it back vget() must be
359 * (vp) is locked on entry and must remain locked
363 linprocfs_inactive(struct vop_inactive_args
*ap
)
365 /*struct vnode *vp = ap->a_vp;*/
371 * _reclaim is called when getnewvnode()
372 * wants to make use of an entry on the vnode
373 * free list. at this time the filesystem needs
374 * to free any private data and remove the node
375 * from any private lists.
378 linprocfs_reclaim(struct vop_reclaim_args
*ap
)
380 return (linprocfs_freevp(ap
->a_vp
));
384 * _print is used for debugging.
385 * just print a readable description
389 linprocfs_print(struct vop_print_args
*ap
)
391 struct pfsnode
*pfs
= VTOPFS(ap
->a_vp
);
393 kprintf("tag VT_PROCFS, type %d, pid %ld, mode %x, flags %lx\n",
394 pfs
->pfs_type
, (long)pfs
->pfs_pid
, pfs
->pfs_mode
, pfs
->pfs_flags
);
399 * generic entry point for unsupported operations
402 linprocfs_badop(struct vop_generic_args
*ap __unused
)
409 * Invent attributes for pfsnode (vp) and store
411 * Directories lengths are returned as zero since
412 * any real length would require the genuine size
413 * to be computed, and nothing cares anyway.
415 * this is relatively minimal for procfs.
418 linprocfs_getattr(struct vop_getattr_args
*ap
)
420 struct pfsnode
*pfs
= VTOPFS(ap
->a_vp
);
421 struct vattr
*vap
= ap
->a_vap
;
426 * First make sure that the process and its credentials
429 switch (pfs
->pfs_type
) {
436 procp
= PFIND(pfs
->pfs_pid
);
437 if (procp
== 0 || procp
->p_ucred
== NULL
)
443 /* start by zeroing out the attributes */
446 /* next do all the common fields */
447 vap
->va_type
= ap
->a_vp
->v_type
;
448 vap
->va_mode
= pfs
->pfs_mode
;
449 vap
->va_fileid
= pfs
->pfs_fileno
;
451 vap
->va_blocksize
= PAGE_SIZE
;
452 vap
->va_bytes
= vap
->va_size
= 0;
453 vap
->va_fsid
= ap
->a_vp
->v_mount
->mnt_stat
.f_fsid
.val
[0];
456 * Make all times be current TOD.
457 * It would be possible to get the process start
458 * time from the p_stat structure, but there's
459 * no "file creation" time stamp anyway, and the
460 * p_stat structure is not addressible if u. gets
461 * swapped out for that process.
463 nanotime(&vap
->va_ctime
);
464 vap
->va_atime
= vap
->va_mtime
= vap
->va_ctime
;
467 * now do the object specific fields
469 * The size could be set from struct reg, but it's hardly
470 * worth the trouble, and it puts some (potentially) machine
471 * dependent data into this machine-independent code. If it
472 * becomes important then this function should break out into
473 * a per-file stat function in the corresponding .c file.
478 vap
->va_uid
= procp
->p_ucred
->cr_uid
;
479 vap
->va_gid
= procp
->p_ucred
->cr_gid
;
482 switch (pfs
->pfs_type
) {
488 * Set nlink to 1 to tell fts(3) we don't actually know.
493 vap
->va_size
= vap
->va_bytes
= DEV_BSIZE
;
497 char buf
[16]; /* should be enough */
500 vap
->va_size
= vap
->va_bytes
=
501 ksnprintf(buf
, sizeof(buf
), "%ld", (long)curproc
->p_pid
);
506 vap
->va_nlink
= nproc_targets
;
507 vap
->va_size
= vap
->va_bytes
= DEV_BSIZE
;
511 char *fullpath
, *freepath
;
512 error
= cache_fullpath(procp
, &procp
->p_textnch
, &fullpath
, &freepath
);
513 /* error = vn_fullpath(procp, NULL, &fullpath, &freepath); */
515 vap
->va_size
= strlen(fullpath
);
516 kfree(freepath
, M_TEMP
);
518 vap
->va_size
= sizeof("unknown") - 1;
521 vap
->va_bytes
= vap
->va_size
;
525 char *fullpath
, *freepath
;
526 error
= cache_fullpath(procp
, &procp
->p_fd
->fd_ncdir
, &fullpath
, &freepath
);
528 vap
->va_size
= strlen(fullpath
);
529 kfree(freepath
, M_TEMP
);
531 vap
->va_size
= sizeof("unknown") - 1;
534 vap
->va_bytes
= vap
->va_size
;
538 struct nchandle
*nchp
;
539 char *fullpath
, *freepath
;
540 nchp
= jailed(procp
->p_ucred
) ? &procp
->p_fd
->fd_njdir
: &procp
->p_fd
->fd_nrdir
;
541 error
= cache_fullpath(procp
, nchp
, &fullpath
, &freepath
);
543 vap
->va_size
= strlen(fullpath
);
544 kfree(freepath
, M_TEMP
);
546 vap
->va_size
= sizeof("unknown") - 1;
549 vap
->va_bytes
= vap
->va_size
;
553 if (procp
== curproc
) {
554 vap
->va_size
= sizeof("/dev/fd") - 1;
557 vap
->va_size
= sizeof("unknown") - 1;
560 vap
->va_bytes
= vap
->va_size
;
576 vap
->va_bytes
= vap
->va_size
= 0;
583 * If we denied owner access earlier, then we have to
584 * change the owner to root - otherwise 'ps' and friends
585 * will break even though they are setgid kmem. *SIGH*
587 if (procp
->p_flag
& P_SUGID
)
590 vap
->va_uid
= procp
->p_ucred
->cr_uid
;
599 vap
->va_bytes
= vap
->va_size
= 0;
600 /* uid, gid are already set */
604 panic("linprocfs_getattr");
611 linprocfs_setattr(struct vop_setattr_args
*ap
)
614 if (ap
->a_vap
->va_flags
!= VNOVAL
)
618 * just fake out attribute setting
619 * it's not good to generate an error
620 * return, otherwise things like creat()
621 * will fail when they try to set the
622 * file length to 0. worse, this means
623 * that echo $note > /proc/$pid/note will fail.
630 * implement access checking.
632 * something very similar to this code is duplicated
633 * throughout the 4bsd kernel and should be moved
634 * into kern/vfs_subr.c sometime.
636 * actually, the check for super-user is slightly
637 * broken since it will allow read access to write-only
638 * objects. this doesn't cause any particular trouble
639 * but does mean that the i/o entry points need to check
640 * that the operation really does make sense.
643 linprocfs_access(struct vop_access_args
*ap
)
650 * If you're the super-user,
651 * you always get access.
653 if (ap
->a_cred
->cr_uid
== 0)
657 error
= VOP_GETATTR(ap
->a_vp
, vap
);
662 * Access check is based on only one of owner, group, public.
663 * If not owner, then check group. If not a member of the
664 * group, then check public access.
666 if (ap
->a_cred
->cr_uid
!= vap
->va_uid
) {
671 gp
= ap
->a_cred
->cr_groups
;
672 for (i
= 0; i
< ap
->a_cred
->cr_ngroups
; i
++, gp
++)
673 if (vap
->va_gid
== *gp
)
680 if ((vap
->va_mode
& ap
->a_mode
) == ap
->a_mode
)
687 * lookup. this is incredibly complicated in the general case, however
688 * for most pseudo-filesystems very little needs to be done.
691 linprocfs_lookup(struct vop_old_lookup_args
*ap
)
693 struct componentname
*cnp
= ap
->a_cnp
;
694 struct vnode
**vpp
= ap
->a_vpp
;
695 struct vnode
*dvp
= ap
->a_dvp
;
696 char *pname
= cnp
->cn_nameptr
;
697 struct proc_target
*pt
;
706 if (cnp
->cn_nameiop
== NAMEI_DELETE
||
707 cnp
->cn_nameiop
== NAMEI_RENAME
||
708 cnp
->cn_nameiop
== NAMEI_CREATE
) {
714 if (cnp
->cn_namelen
== 1 && *pname
== '.') {
721 switch (pfs
->pfs_type
) {
723 if (cnp
->cn_flags
& CNP_ISDOTDOT
) {
724 error
= linprocfs_root(dvp
->v_mount
, vpp
);
727 if (CNEQ(cnp
, "kernel", 6)) {
728 error
= linprocfs_allocvp(dvp
->v_mount
, vpp
, 0, Psyskernel
);
733 if (cnp
->cn_flags
& CNP_ISDOTDOT
) {
734 error
= linprocfs_root(dvp
->v_mount
, vpp
);
737 if (CNEQ(cnp
, "dev", 3)) {
738 error
= linprocfs_allocvp(dvp
->v_mount
, vpp
, 0, Pnetdev
);
743 if (cnp
->cn_flags
& CNP_ISDOTDOT
) {
744 /* XXX: this is wrong, wrong, wrong. */
745 error
= linprocfs_root(dvp
->v_mount
, vpp
);
748 if (CNEQ(cnp
, "osrelease", 9)) {
749 error
= linprocfs_allocvp(dvp
->v_mount
, vpp
, 0, Posrelease
);
752 if (CNEQ(cnp
, "ostype", 6)) {
753 error
= linprocfs_allocvp(dvp
->v_mount
, vpp
, 0, Postype
);
756 if (CNEQ(cnp
, "pid_max", 7)) {
757 error
= linprocfs_allocvp(dvp
->v_mount
, vpp
, 0, Ppidmax
);
760 if (CNEQ(cnp
, "version", 7)) {
761 error
= linprocfs_allocvp(dvp
->v_mount
, vpp
, 0, Pversion
);
767 if (cnp
->cn_flags
& CNP_ISDOTDOT
)
770 if (CNEQ(cnp
, "self", 4)) {
771 error
= linprocfs_allocvp(dvp
->v_mount
, vpp
, 0, Pself
);
774 if (CNEQ(cnp
, "meminfo", 7)) {
775 error
= linprocfs_allocvp(dvp
->v_mount
, vpp
, 0, Pmeminfo
);
778 if (CNEQ(cnp
, "cpuinfo", 7)) {
779 error
= linprocfs_allocvp(dvp
->v_mount
, vpp
, 0, Pcpuinfo
);
782 if (CNEQ(cnp
, "mounts", 6)) {
783 error
= linprocfs_allocvp(dvp
->v_mount
, vpp
, 0, Pmounts
);
786 if (CNEQ(cnp
, "stat", 4)) {
787 error
= linprocfs_allocvp(dvp
->v_mount
, vpp
, 0, Pstat
);
790 if (CNEQ(cnp
, "uptime", 6)) {
791 error
= linprocfs_allocvp(dvp
->v_mount
, vpp
, 0, Puptime
);
794 if (CNEQ(cnp
, "version", 7)) {
795 error
= linprocfs_allocvp(dvp
->v_mount
, vpp
, 0, Pversion
);
798 if (CNEQ(cnp
, "loadavg", 7)) {
799 error
= linprocfs_allocvp(dvp
->v_mount
, vpp
, 0, Ploadavg
);
802 if (CNEQ(cnp
, "net", 3)) {
803 error
= linprocfs_allocvp(dvp
->v_mount
, vpp
, 0, Pnet
);
806 if (CNEQ(cnp
, "sys", 3)) {
807 error
= linprocfs_allocvp(dvp
->v_mount
, vpp
, 0, Psys
);
811 pid
= atopid(pname
, cnp
->cn_namelen
);
819 if (!PRISON_CHECK(ap
->a_cnp
->cn_cred
, p
->p_ucred
))
822 if (ps_showallprocs
== 0 && ap
->a_cnp
->cn_cred
->cr_uid
!= 0 &&
823 ap
->a_cnp
->cn_cred
->cr_uid
!= p
->p_ucred
->cr_uid
)
826 error
= linprocfs_allocvp(dvp
->v_mount
, vpp
, pid
, Pproc
);
830 if (cnp
->cn_flags
& CNP_ISDOTDOT
) {
831 error
= linprocfs_root(dvp
->v_mount
, vpp
);
835 p
= PFIND(pfs
->pfs_pid
);
839 if (!PRISON_CHECK(ap
->a_cnp
->cn_cred
, p
->p_ucred
))
842 if (ps_showallprocs
== 0 && ap
->a_cnp
->cn_cred
->cr_uid
!= 0 &&
843 ap
->a_cnp
->cn_cred
->cr_uid
!= p
->p_ucred
->cr_uid
)
846 for (pt
= proc_targets
, i
= 0; i
< nproc_targets
; pt
++, i
++) {
847 if (cnp
->cn_namelen
== pt
->pt_namlen
&&
848 bcmp(pt
->pt_name
, pname
, cnp
->cn_namelen
) == 0 &&
849 (pt
->pt_valid
== NULL
|| (*pt
->pt_valid
)(p
)))
855 error
= linprocfs_allocvp(dvp
->v_mount
, vpp
, pfs
->pfs_pid
,
864 if (cnp
->cn_nameiop
== NAMEI_LOOKUP
)
870 * If no error occured *vpp will hold a referenced locked vnode.
871 * dvp was passed to us locked and *vpp must be returned locked
872 * so if dvp != *vpp and CNP_LOCKPARENT is not set, unlock dvp.
876 if (*vpp
!= dvp
&& (cnp
->cn_flags
& CNP_LOCKPARENT
) == 0) {
877 cnp
->cn_flags
|= CNP_PDIRUNLOCK
;
885 * Does this process have a text file?
888 linprocfs_validfile(struct proc
*p
)
891 return (procfs_findtextvp(p
) != NULLVP
);
895 * readdir() returns directory entries from pfsnode (vp).
897 * We generate just one directory entry at a time, as it would probably
898 * not pay off to buffer several entries locally to save uiomove calls.
900 * linprocfs_readdir(struct vnode *a_vp, struct uio *a_uio,
901 * struct ucred *a_cred, int *a_eofflag,
902 * int *a_ncookies, off_t **a_cookies)
905 linprocfs_readdir(struct vop_readdir_args
*ap
)
910 if (ap
->a_uio
->uio_offset
< 0 || ap
->a_uio
->uio_offset
> INT_MAX
)
913 pfs
= VTOPFS(ap
->a_vp
);
914 if ((error
= vn_lock(ap
->a_vp
, LK_EXCLUSIVE
| LK_RETRY
)) != 0)
917 switch (pfs
->pfs_type
) {
920 * This is for the process-specific sub-directories.
921 * all that is needed to is copy out all the entries
922 * from the procent[] table (top of this file).
924 error
= linprocfs_readdir_proc(ap
);
928 * This is for the root of the procfs filesystem
929 * what is needed is a special entry for "self"
930 * followed by an entry for each process on allproc
932 error
= linprocfs_readdir_root(ap
);
935 error
= linprocfs_readdir_net(ap
);
938 error
= linprocfs_readdir_sys(ap
);
941 error
= linprocfs_readdir_syskernel(ap
);
953 linprocfs_readdir_proc(struct vop_readdir_args
*ap
)
956 int error
, i
, retval
;
958 struct proc_target
*pt
;
959 struct uio
*uio
= ap
->a_uio
;
961 pfs
= VTOPFS(ap
->a_vp
);
962 p
= PFIND(pfs
->pfs_pid
);
965 if (!PRISON_CHECK(ap
->a_cred
, p
->p_ucred
))
971 for (pt
= &proc_targets
[i
];
972 !error
&& uio
->uio_resid
> 0 && i
< nproc_targets
; pt
++, i
++) {
973 if (pt
->pt_valid
&& (*pt
->pt_valid
)(p
) == 0)
976 retval
= vop_write_dirent(&error
, uio
,
977 PROCFS_FILENO(pfs
->pfs_pid
, pt
->pt_pfstype
), pt
->pt_type
,
978 pt
->pt_namlen
, pt
->pt_name
);
988 struct linprocfs_readdir_root_info
{
997 * Scan the root directory by scanning all process
999 static int linprocfs_readdir_root_callback(struct proc
*p
, void *data
);
1002 linprocfs_readdir_root(struct vop_readdir_args
*ap
)
1004 struct linprocfs_readdir_root_info info
;
1005 struct uio
*uio
= ap
->a_uio
;
1009 info
.i
= uio
->uio_offset
;
1012 info
.cred
= ap
->a_cred
;
1014 while (info
.pcnt
< 12) {
1015 res
= linprocfs_readdir_root_callback(NULL
, &info
);
1020 allproc_scan(linprocfs_readdir_root_callback
, &info
);
1022 uio
->uio_offset
= info
.i
;
1027 linprocfs_readdir_root_callback(struct proc
*p
, void *data
)
1029 struct linprocfs_readdir_root_info
*info
= data
;
1031 struct uio
*uio
= info
->uio
;
1034 char d_name_pid
[20];
1038 switch (info
->pcnt
) {
1040 d_ino
= PROCFS_FILENO(0, Proot
);
1046 d_ino
= PROCFS_FILENO(0, Proot
);
1053 d_ino
= PROCFS_FILENO(0, Proot
);
1060 d_ino
= PROCFS_FILENO(0, Pmeminfo
);
1067 d_ino
= PROCFS_FILENO(0, Pcpuinfo
);
1074 d_ino
= PROCFS_FILENO(0, Pstat
);
1081 d_ino
= PROCFS_FILENO(0, Puptime
);
1088 d_ino
= PROCFS_FILENO(0, Pversion
);
1095 d_ino
= PROCFS_FILENO(0, Ploadavg
);
1101 d_ino
= PROCFS_FILENO(0, Pnet
);
1107 d_ino
= PROCFS_FILENO(0, Psys
);
1113 d_ino
= PROCFS_FILENO(0, Pmounts
);
1120 d_ino
= PROCFS_FILENO(0, Pdevices
);
1128 * Ignore processes that aren't in our prison
1130 if (PRISON_CHECK(info
->cred
, p
->p_ucred
) == 0)
1134 * Ignore processes that we do not want to be visible.
1136 if (ps_showallprocs
== 0 &&
1137 info
->cred
->cr_uid
!= 0 &&
1138 info
->cred
->cr_uid
!= p
->p_ucred
->cr_uid
) {
1143 * Skip processes we have already read (optimization)
1145 if (info
->pcnt
< info
->i
) {
1149 d_ino
= PROCFS_FILENO(p
->p_pid
, Pproc
);
1150 d_namlen
= ksnprintf(d_name_pid
, sizeof(d_name_pid
),
1151 "%ld", (long)p
->p_pid
);
1152 d_name
= d_name_pid
;
1158 * Skip processes we have already read
1160 if (info
->pcnt
< info
->i
) {
1164 retval
= vop_write_dirent(&info
->error
, info
->uio
,
1165 d_ino
, d_type
, d_namlen
, d_name
);
1167 ++info
->pcnt
; /* iterate proc candidates scanned */
1168 ++info
->i
; /* iterate entries written */
1170 if (retval
|| info
->error
|| uio
->uio_resid
<= 0)
1176 * Scan the root directory by scanning all process
1178 static int linprocfs_readdir_net_callback(struct proc
*p
, void *data
);
1181 linprocfs_readdir_net(struct vop_readdir_args
*ap
)
1183 struct linprocfs_readdir_root_info info
;
1184 struct uio
*uio
= ap
->a_uio
;
1188 info
.i
= uio
->uio_offset
;
1191 info
.cred
= ap
->a_cred
;
1193 while (info
.pcnt
< 3) {
1194 res
= linprocfs_readdir_net_callback(NULL
, &info
);
1199 uio
->uio_offset
= info
.i
;
1204 linprocfs_readdir_net_callback(struct proc
*p
, void *data
)
1206 struct linprocfs_readdir_root_info
*info
= data
;
1208 struct uio
*uio
= info
->uio
;
1214 switch (info
->pcnt
) {
1216 d_ino
= PROCFS_FILENO(0, Pnet
);
1222 d_ino
= PROCFS_FILENO(0, Proot
);
1229 d_ino
= PROCFS_FILENO(0, Pnet
);
1243 * Skip processes we have already read
1245 if (info
->pcnt
< info
->i
) {
1249 retval
= vop_write_dirent(&info
->error
, info
->uio
,
1250 d_ino
, d_type
, d_namlen
, d_name
);
1252 ++info
->pcnt
; /* iterate proc candidates scanned */
1253 ++info
->i
; /* iterate entries written */
1255 if (retval
|| info
->error
|| uio
->uio_resid
<= 0)
1267 * Scan the root directory by scanning all process
1269 static int linprocfs_readdir_sys_callback(struct proc
*p
, void *data
);
1272 linprocfs_readdir_sys(struct vop_readdir_args
*ap
)
1274 struct linprocfs_readdir_root_info info
;
1275 struct uio
*uio
= ap
->a_uio
;
1279 info
.i
= uio
->uio_offset
;
1282 info
.cred
= ap
->a_cred
;
1284 while (info
.pcnt
< 3) {
1285 res
= linprocfs_readdir_sys_callback(NULL
, &info
);
1290 uio
->uio_offset
= info
.i
;
1295 linprocfs_readdir_sys_callback(struct proc
*p
, void *data
)
1297 struct linprocfs_readdir_root_info
*info
= data
;
1299 struct uio
*uio
= info
->uio
;
1305 switch (info
->pcnt
) {
1307 d_ino
= PROCFS_FILENO(0, Psys
);
1313 d_ino
= PROCFS_FILENO(0, Proot
);
1320 d_ino
= PROCFS_FILENO(0, Psyskernel
);
1334 * Skip processes we have already read
1336 if (info
->pcnt
< info
->i
) {
1340 retval
= vop_write_dirent(&info
->error
, info
->uio
,
1341 d_ino
, d_type
, d_namlen
, d_name
);
1343 ++info
->pcnt
; /* iterate proc candidates scanned */
1344 ++info
->i
; /* iterate entries written */
1346 if (retval
|| info
->error
|| uio
->uio_resid
<= 0)
1356 * Scan the root directory by scanning all process
1358 static int linprocfs_readdir_syskernel_callback(struct proc
*p
, void *data
);
1361 linprocfs_readdir_syskernel(struct vop_readdir_args
*ap
)
1363 struct linprocfs_readdir_root_info info
;
1364 struct uio
*uio
= ap
->a_uio
;
1368 info
.i
= uio
->uio_offset
;
1371 info
.cred
= ap
->a_cred
;
1373 while (info
.pcnt
< 6) {
1374 res
= linprocfs_readdir_syskernel_callback(NULL
, &info
);
1379 uio
->uio_offset
= info
.i
;
1384 linprocfs_readdir_syskernel_callback(struct proc
*p
, void *data
)
1386 struct linprocfs_readdir_root_info
*info
= data
;
1388 struct uio
*uio
= info
->uio
;
1394 switch (info
->pcnt
) {
1396 d_ino
= PROCFS_FILENO(0, Psyskernel
);
1402 d_ino
= PROCFS_FILENO(0, Psys
);
1409 d_ino
= PROCFS_FILENO(0, Posrelease
);
1411 d_name
= "osrelease";
1416 d_ino
= PROCFS_FILENO(0, Postype
);
1423 d_ino
= PROCFS_FILENO(0, Pversion
);
1430 d_ino
= PROCFS_FILENO(0, Ppidmax
);
1444 * Skip processes we have already read
1446 if (info
->pcnt
< info
->i
) {
1450 retval
= vop_write_dirent(&info
->error
, info
->uio
,
1451 d_ino
, d_type
, d_namlen
, d_name
);
1453 ++info
->pcnt
; /* iterate proc candidates scanned */
1454 ++info
->i
; /* iterate entries written */
1456 if (retval
|| info
->error
|| uio
->uio_resid
<= 0)
1462 * readlink reads the link of `self' or `exe'
1465 linprocfs_readlink(struct vop_readlink_args
*ap
)
1467 char buf
[16]; /* should be enough */
1469 struct vnode
*vp
= ap
->a_vp
;
1470 struct nchandle
*nchp
;
1471 struct pfsnode
*pfs
= VTOPFS(vp
);
1472 char *fullpath
, *freepath
;
1475 switch (pfs
->pfs_type
) {
1477 if (pfs
->pfs_fileno
!= PROCFS_FILENO(0, Pself
))
1480 len
= ksnprintf(buf
, sizeof(buf
), "%ld", (long)curproc
->p_pid
);
1482 return (uiomove(buf
, len
, ap
->a_uio
));
1484 * There _should_ be no way for an entire process to disappear
1488 procp
= PFIND(pfs
->pfs_pid
);
1489 if (procp
== NULL
|| procp
->p_ucred
== NULL
) {
1490 kprintf("linprocfs_readlink: pid %d disappeared\n",
1492 return (uiomove("unknown", sizeof("unknown") - 1,
1495 error
= cache_fullpath(procp
, &procp
->p_textnch
, &fullpath
, &freepath
);
1497 return (uiomove("unknown", sizeof("unknown") - 1,
1499 error
= uiomove(fullpath
, strlen(fullpath
), ap
->a_uio
);
1500 kfree(freepath
, M_TEMP
);
1503 procp
= PFIND(pfs
->pfs_pid
);
1504 if (procp
== NULL
|| procp
->p_ucred
== NULL
) {
1505 kprintf("linprocfs_readlink: pid %d disappeared\n",
1507 return (uiomove("unknown", sizeof("unknown") - 1,
1510 error
= cache_fullpath(procp
, &procp
->p_fd
->fd_ncdir
, &fullpath
, &freepath
);
1512 return (uiomove("unknown", sizeof("unknown") - 1,
1514 error
= uiomove(fullpath
, strlen(fullpath
), ap
->a_uio
);
1515 kfree(freepath
, M_TEMP
);
1518 procp
= PFIND(pfs
->pfs_pid
);
1519 if (procp
== NULL
|| procp
->p_ucred
== NULL
) {
1520 kprintf("linprocfs_readlink: pid %d disappeared\n",
1522 return (uiomove("unknown", sizeof("unknown") - 1,
1525 nchp
= jailed(procp
->p_ucred
) ? &procp
->p_fd
->fd_njdir
: &procp
->p_fd
->fd_nrdir
;
1526 error
= cache_fullpath(procp
, nchp
, &fullpath
, &freepath
);
1528 return (uiomove("unknown", sizeof("unknown") - 1,
1530 error
= uiomove(fullpath
, strlen(fullpath
), ap
->a_uio
);
1531 kfree(freepath
, M_TEMP
);
1534 procp
= PFIND(pfs
->pfs_pid
);
1535 if (procp
== NULL
|| procp
->p_ucred
== NULL
) {
1536 kprintf("linprocfs_readlink: pid %d disappeared\n",
1538 return (uiomove("unknown", sizeof("unknown") - 1,
1541 if (procp
== curproc
) {
1542 return (uiomove("/dev/fd", sizeof("/dev/fd") - 1,
1545 return (uiomove("unknown", sizeof("unknown") - 1,
1556 * convert decimal ascii to pid_t
1559 atopid(const char *b
, u_int len
)
1565 if (c
< '0' || c
> '9')
1567 p
= 10 * p
+ (c
- '0');