xref: /netbsd-src/sys/miscfs/procfs/procfs_vnops.c (revision 76dfffe33547c37f8bdd446e3e4ab0f3c16cea4b)
1 /*	$NetBSD: procfs_vnops.c,v 1.45 1996/10/25 21:58:03 cgd Exp $	*/
2 
3 /*
4  * Copyright (c) 1993 Jan-Simon Pendry
5  * Copyright (c) 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * This code is derived from software contributed to Berkeley by
9  * Jan-Simon Pendry.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
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.
26  *
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
37  * SUCH DAMAGE.
38  *
39  *	@(#)procfs_vnops.c	8.8 (Berkeley) 6/15/94
40  */
41 
42 /*
43  * procfs vnode interface
44  */
45 
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/time.h>
49 #include <sys/kernel.h>
50 #include <sys/file.h>
51 #include <sys/proc.h>
52 #include <sys/vnode.h>
53 #include <sys/namei.h>
54 #include <sys/malloc.h>
55 #include <sys/dirent.h>
56 #include <sys/resourcevar.h>
57 #include <sys/ptrace.h>
58 #include <vm/vm.h>	/* for PAGE_SIZE */
59 #include <machine/reg.h>
60 
61 #include <miscfs/genfs/genfs.h>
62 #include <miscfs/procfs/procfs.h>
63 
64 /*
65  * Vnode Operations.
66  *
67  */
68 
69 /*
70  * This is a list of the valid names in the
71  * process-specific sub-directories.  It is
72  * used in procfs_lookup and procfs_readdir
73  */
74 struct proc_target {
75 	u_char	pt_type;
76 	u_char	pt_namlen;
77 	char	*pt_name;
78 	pfstype	pt_pfstype;
79 	int	(*pt_valid) __P((struct proc *p));
80 } proc_targets[] = {
81 #define N(s) sizeof(s)-1, s
82 	/*	  name		type		validp */
83 	{ DT_DIR, N("."),	Pproc,		NULL },
84 	{ DT_DIR, N(".."),	Proot,		NULL },
85 	{ DT_REG, N("file"),	Pfile,		procfs_validfile },
86 	{ DT_REG, N("mem"),	Pmem,		NULL },
87 	{ DT_REG, N("regs"),	Pregs,		procfs_validregs },
88 	{ DT_REG, N("fpregs"),	Pfpregs,	procfs_validfpregs },
89 	{ DT_REG, N("ctl"),	Pctl,		NULL },
90 	{ DT_REG, N("status"),	Pstatus,	NULL },
91 	{ DT_REG, N("note"),	Pnote,		NULL },
92 	{ DT_REG, N("notepg"),	Pnotepg,	NULL },
93 #undef N
94 };
95 static int nproc_targets = sizeof(proc_targets) / sizeof(proc_targets[0]);
96 
97 static pid_t atopid __P((const char *, u_int));
98 
99 int	procfs_lookup	__P((void *));
100 #define	procfs_create	genfs_eopnotsupp
101 #define	procfs_mknod	genfs_eopnotsupp
102 int	procfs_open	__P((void *));
103 int	procfs_close	__P((void *));
104 int	procfs_access	__P((void *));
105 int	procfs_getattr	__P((void *));
106 int	procfs_setattr	__P((void *));
107 #define	procfs_read	procfs_rw
108 #define	procfs_write	procfs_rw
109 #define	procfs_ioctl	genfs_eopnotsupp
110 #define	procfs_poll	genfs_poll
111 #define	procfs_mmap	genfs_eopnotsupp
112 #define	procfs_fsync	genfs_nullop
113 #define	procfs_seek	genfs_nullop
114 #define	procfs_remove	genfs_eopnotsupp
115 int	procfs_link	__P((void *));
116 #define	procfs_rename	genfs_eopnotsupp
117 #define	procfs_mkdir	genfs_eopnotsupp
118 #define	procfs_rmdir	genfs_eopnotsupp
119 int	procfs_symlink	__P((void *));
120 int	procfs_readdir	__P((void *));
121 int	procfs_readlink	__P((void *));
122 #define	procfs_abortop	genfs_abortop
123 int	procfs_inactive	__P((void *));
124 int	procfs_reclaim	__P((void *));
125 #define	procfs_lock	genfs_nullop
126 #define	procfs_unlock	genfs_nullop
127 int	procfs_bmap	__P((void *));
128 #define	procfs_strategy	genfs_badop
129 int	procfs_print	__P((void *));
130 int	procfs_pathconf	__P((void *));
131 #define	procfs_islocked	genfs_nullop
132 #define	procfs_advlock	genfs_eopnotsupp
133 #define	procfs_blkatoff	genfs_eopnotsupp
134 #define	procfs_valloc	genfs_eopnotsupp
135 #define	procfs_vfree	genfs_nullop
136 #define	procfs_truncate	genfs_eopnotsupp
137 #define	procfs_update	genfs_nullop
138 #define	procfs_bwrite	genfs_eopnotsupp
139 
140 static pid_t atopid __P((const char *, u_int));
141 
142 /*
143  * procfs vnode operations.
144  */
145 int (**procfs_vnodeop_p) __P((void *));
146 struct vnodeopv_entry_desc procfs_vnodeop_entries[] = {
147 	{ &vop_default_desc, vn_default_error },
148 	{ &vop_lookup_desc, procfs_lookup },		/* lookup */
149 	{ &vop_create_desc, procfs_create },		/* create */
150 	{ &vop_mknod_desc, procfs_mknod },		/* mknod */
151 	{ &vop_open_desc, procfs_open },		/* open */
152 	{ &vop_close_desc, procfs_close },		/* close */
153 	{ &vop_access_desc, procfs_access },		/* access */
154 	{ &vop_getattr_desc, procfs_getattr },		/* getattr */
155 	{ &vop_setattr_desc, procfs_setattr },		/* setattr */
156 	{ &vop_read_desc, procfs_read },		/* read */
157 	{ &vop_write_desc, procfs_write },		/* write */
158 	{ &vop_ioctl_desc, procfs_ioctl },		/* ioctl */
159 	{ &vop_poll_desc, procfs_poll },		/* poll */
160 	{ &vop_mmap_desc, procfs_mmap },		/* mmap */
161 	{ &vop_fsync_desc, procfs_fsync },		/* fsync */
162 	{ &vop_seek_desc, procfs_seek },		/* seek */
163 	{ &vop_remove_desc, procfs_remove },		/* remove */
164 	{ &vop_link_desc, procfs_link },		/* link */
165 	{ &vop_rename_desc, procfs_rename },		/* rename */
166 	{ &vop_mkdir_desc, procfs_mkdir },		/* mkdir */
167 	{ &vop_rmdir_desc, procfs_rmdir },		/* rmdir */
168 	{ &vop_symlink_desc, procfs_symlink },		/* symlink */
169 	{ &vop_readdir_desc, procfs_readdir },		/* readdir */
170 	{ &vop_readlink_desc, procfs_readlink },	/* readlink */
171 	{ &vop_abortop_desc, procfs_abortop },		/* abortop */
172 	{ &vop_inactive_desc, procfs_inactive },	/* inactive */
173 	{ &vop_reclaim_desc, procfs_reclaim },		/* reclaim */
174 	{ &vop_lock_desc, procfs_lock },		/* lock */
175 	{ &vop_unlock_desc, procfs_unlock },		/* unlock */
176 	{ &vop_bmap_desc, procfs_bmap },		/* bmap */
177 	{ &vop_strategy_desc, procfs_strategy },	/* strategy */
178 	{ &vop_print_desc, procfs_print },		/* print */
179 	{ &vop_islocked_desc, procfs_islocked },	/* islocked */
180 	{ &vop_pathconf_desc, procfs_pathconf },	/* pathconf */
181 	{ &vop_advlock_desc, procfs_advlock },		/* advlock */
182 	{ &vop_blkatoff_desc, procfs_blkatoff },	/* blkatoff */
183 	{ &vop_valloc_desc, procfs_valloc },		/* valloc */
184 	{ &vop_vfree_desc, procfs_vfree },		/* vfree */
185 	{ &vop_truncate_desc, procfs_truncate },	/* truncate */
186 	{ &vop_update_desc, procfs_update },		/* update */
187 	{ (struct vnodeop_desc*)NULL, (int(*) __P((void *)))NULL }
188 };
189 struct vnodeopv_desc procfs_vnodeop_opv_desc =
190 	{ &procfs_vnodeop_p, procfs_vnodeop_entries };
191 /*
192  * set things up for doing i/o on
193  * the pfsnode (vp).  (vp) is locked
194  * on entry, and should be left locked
195  * on exit.
196  *
197  * for procfs we don't need to do anything
198  * in particular for i/o.  all that is done
199  * is to support exclusive open on process
200  * memory images.
201  */
202 int
203 procfs_open(v)
204 	void *v;
205 {
206 	struct vop_open_args /* {
207 		struct vnode *a_vp;
208 		int  a_mode;
209 		struct ucred *a_cred;
210 		struct proc *a_p;
211 	} */ *ap = v;
212 	struct pfsnode *pfs = VTOPFS(ap->a_vp);
213 
214 	switch (pfs->pfs_type) {
215 	case Pmem:
216 		if (PFIND(pfs->pfs_pid) == 0)
217 			return (ENOENT);	/* was ESRCH, jsp */
218 
219 		if (((pfs->pfs_flags & FWRITE) && (ap->a_mode & O_EXCL)) ||
220 		    ((pfs->pfs_flags & O_EXCL) && (ap->a_mode & FWRITE)))
221 			return (EBUSY);
222 
223 		if (ap->a_mode & FWRITE)
224 			pfs->pfs_flags = ap->a_mode & (FWRITE|O_EXCL);
225 
226 		return (0);
227 
228 	default:
229 		break;
230 	}
231 
232 	return (0);
233 }
234 
235 /*
236  * close the pfsnode (vp) after doing i/o.
237  * (vp) is not locked on entry or exit.
238  *
239  * nothing to do for procfs other than undo
240  * any exclusive open flag (see _open above).
241  */
242 int
243 procfs_close(v)
244 	void *v;
245 {
246 	struct vop_close_args /* {
247 		struct vnode *a_vp;
248 		int  a_fflag;
249 		struct ucred *a_cred;
250 		struct proc *a_p;
251 	} */ *ap = v;
252 	struct pfsnode *pfs = VTOPFS(ap->a_vp);
253 
254 	switch (pfs->pfs_type) {
255 	case Pmem:
256 		if ((ap->a_fflag & FWRITE) && (pfs->pfs_flags & O_EXCL))
257 			pfs->pfs_flags &= ~(FWRITE|O_EXCL);
258 		break;
259 	case Pctl:
260 	case Pstatus:
261 	case Pnotepg:
262 	case Pnote:
263 	case Proot:
264 	case Pcurproc:
265 	case Pproc:
266 	case Pfile:
267 	case Pregs:
268 	case Pfpregs:
269 		break;
270 	}
271 
272 	return (0);
273 }
274 
275 /*
276  * do block mapping for pfsnode (vp).
277  * since we don't use the buffer cache
278  * for procfs this function should never
279  * be called.  in any case, it's not clear
280  * what part of the kernel ever makes use
281  * of this function.  for sanity, this is the
282  * usual no-op bmap, although returning
283  * (EIO) would be a reasonable alternative.
284  */
285 int
286 procfs_bmap(v)
287 	void *v;
288 {
289 	struct vop_bmap_args /* {
290 		struct vnode *a_vp;
291 		daddr_t  a_bn;
292 		struct vnode **a_vpp;
293 		daddr_t *a_bnp;
294 	} */ *ap = v;
295 
296 	if (ap->a_vpp != NULL)
297 		*ap->a_vpp = ap->a_vp;
298 	if (ap->a_bnp != NULL)
299 		*ap->a_bnp = ap->a_bn;
300 	return (0);
301 }
302 
303 /*
304  * _inactive is called when the pfsnode
305  * is vrele'd and the reference count goes
306  * to zero.  (vp) will be on the vnode free
307  * list, so to get it back vget() must be
308  * used.
309  *
310  * for procfs, check if the process is still
311  * alive and if it isn't then just throw away
312  * the vnode by calling vgone().  this may
313  * be overkill and a waste of time since the
314  * chances are that the process will still be
315  * there and PFIND is not free.
316  *
317  * (vp) is not locked on entry or exit.
318  */
319 int
320 procfs_inactive(v)
321 	void *v;
322 {
323 	struct vop_inactive_args /* {
324 		struct vnode *a_vp;
325 	} */ *ap = v;
326 	struct pfsnode *pfs = VTOPFS(ap->a_vp);
327 
328 	if (PFIND(pfs->pfs_pid) == 0)
329 		vgone(ap->a_vp);
330 
331 	return (0);
332 }
333 
334 /*
335  * _reclaim is called when getnewvnode()
336  * wants to make use of an entry on the vnode
337  * free list.  at this time the filesystem needs
338  * to free any private data and remove the node
339  * from any private lists.
340  */
341 int
342 procfs_reclaim(v)
343 	void *v;
344 {
345 	struct vop_reclaim_args /* {
346 		struct vnode *a_vp;
347 	} */ *ap = v;
348 
349 	return (procfs_freevp(ap->a_vp));
350 }
351 
352 /*
353  * Return POSIX pathconf information applicable to special devices.
354  */
355 int
356 procfs_pathconf(v)
357 	void *v;
358 {
359 	struct vop_pathconf_args /* {
360 		struct vnode *a_vp;
361 		int a_name;
362 		register_t *a_retval;
363 	} */ *ap = v;
364 
365 	switch (ap->a_name) {
366 	case _PC_LINK_MAX:
367 		*ap->a_retval = LINK_MAX;
368 		return (0);
369 	case _PC_MAX_CANON:
370 		*ap->a_retval = MAX_CANON;
371 		return (0);
372 	case _PC_MAX_INPUT:
373 		*ap->a_retval = MAX_INPUT;
374 		return (0);
375 	case _PC_PIPE_BUF:
376 		*ap->a_retval = PIPE_BUF;
377 		return (0);
378 	case _PC_CHOWN_RESTRICTED:
379 		*ap->a_retval = 1;
380 		return (0);
381 	case _PC_VDISABLE:
382 		*ap->a_retval = _POSIX_VDISABLE;
383 		return (0);
384 	default:
385 		return (EINVAL);
386 	}
387 	/* NOTREACHED */
388 }
389 
390 /*
391  * _print is used for debugging.
392  * just print a readable description
393  * of (vp).
394  */
395 int
396 procfs_print(v)
397 	void *v;
398 {
399 	struct vop_print_args /* {
400 		struct vnode *a_vp;
401 	} */ *ap = v;
402 	struct pfsnode *pfs = VTOPFS(ap->a_vp);
403 
404 	printf("tag VT_PROCFS, type %d, pid %d, mode %x, flags %lx\n",
405 	    pfs->pfs_type, pfs->pfs_pid, pfs->pfs_mode, pfs->pfs_flags);
406 	return 0;
407 }
408 
409 int
410 procfs_link(v)
411 	void *v;
412 {
413 	struct vop_link_args /* {
414 		struct vnode *a_dvp;
415 		struct vnode *a_vp;
416 		struct componentname *a_cnp;
417 	} */ *ap = v;
418 
419 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
420 	vput(ap->a_dvp);
421 	return (EROFS);
422 }
423 
424 int
425 procfs_symlink(v)
426 	void *v;
427 {
428 	struct vop_symlink_args /* {
429 		struct vnode *a_dvp;
430 		struct vnode **a_vpp;
431 		struct componentname *a_cnp;
432 		struct vattr *a_vap;
433 		char *a_target;
434 	} */ *ap = v;
435 
436 	VOP_ABORTOP(ap->a_dvp, ap->a_cnp);
437 	vput(ap->a_dvp);
438 	return (EROFS);
439 }
440 
441 /*
442  * Invent attributes for pfsnode (vp) and store
443  * them in (vap).
444  * Directories lengths are returned as zero since
445  * any real length would require the genuine size
446  * to be computed, and nothing cares anyway.
447  *
448  * this is relatively minimal for procfs.
449  */
450 int
451 procfs_getattr(v)
452 	void *v;
453 {
454 	struct vop_getattr_args /* {
455 		struct vnode *a_vp;
456 		struct vattr *a_vap;
457 		struct ucred *a_cred;
458 		struct proc *a_p;
459 	} */ *ap = v;
460 	struct pfsnode *pfs = VTOPFS(ap->a_vp);
461 	struct vattr *vap = ap->a_vap;
462 	struct proc *procp;
463 	struct timeval tv;
464 	int error;
465 
466 	/* first check the process still exists */
467 	switch (pfs->pfs_type) {
468 	case Proot:
469 	case Pcurproc:
470 		procp = 0;
471 		break;
472 
473 	default:
474 		procp = PFIND(pfs->pfs_pid);
475 		if (procp == 0)
476 			return (ENOENT);
477 	}
478 
479 	error = 0;
480 
481 	/* start by zeroing out the attributes */
482 	VATTR_NULL(vap);
483 
484 	/* next do all the common fields */
485 	vap->va_type = ap->a_vp->v_type;
486 	vap->va_mode = pfs->pfs_mode;
487 	vap->va_fileid = pfs->pfs_fileno;
488 	vap->va_flags = 0;
489 	vap->va_blocksize = PAGE_SIZE;
490 	vap->va_bytes = vap->va_size = 0;
491 
492 	/*
493 	 * Make all times be current TOD.
494 	 * It would be possible to get the process start
495 	 * time from the p_stat structure, but there's
496 	 * no "file creation" time stamp anyway, and the
497 	 * p_stat structure is not addressible if u. gets
498 	 * swapped out for that process.
499 	 */
500 	microtime(&tv);
501 	TIMEVAL_TO_TIMESPEC(&tv, &vap->va_ctime);
502 	vap->va_atime = vap->va_mtime = vap->va_ctime;
503 
504 	/*
505 	 * If the process has exercised some setuid or setgid
506 	 * privilege, then rip away read/write permission so
507 	 * that only root can gain access.
508 	 */
509 	switch (pfs->pfs_type) {
510 	case Pmem:
511 	case Pregs:
512 	case Pfpregs:
513 		if (procp->p_flag & P_SUGID)
514 			vap->va_mode &= ~((VREAD|VWRITE)|
515 					  ((VREAD|VWRITE)>>3)|
516 					  ((VREAD|VWRITE)>>6));
517 	case Pctl:
518 	case Pstatus:
519 	case Pnote:
520 	case Pnotepg:
521 		vap->va_nlink = 1;
522 		vap->va_uid = procp->p_ucred->cr_uid;
523 		vap->va_gid = procp->p_ucred->cr_gid;
524 		break;
525 	case Pproc:
526 	case Pfile:
527 	case Proot:
528 	case Pcurproc:
529 		break;
530 	}
531 
532 	/*
533 	 * now do the object specific fields
534 	 *
535 	 * The size could be set from struct reg, but it's hardly
536 	 * worth the trouble, and it puts some (potentially) machine
537 	 * dependent data into this machine-independent code.  If it
538 	 * becomes important then this function should break out into
539 	 * a per-file stat function in the corresponding .c file.
540 	 */
541 
542 	switch (pfs->pfs_type) {
543 	case Proot:
544 		/*
545 		 * Set nlink to 1 to tell fts(3) we don't actually know.
546 		 */
547 		vap->va_nlink = 1;
548 		vap->va_uid = 0;
549 		vap->va_gid = 0;
550 		vap->va_size = vap->va_bytes = DEV_BSIZE;
551 		break;
552 
553 	case Pcurproc: {
554 		char buf[16];		/* should be enough */
555 		vap->va_nlink = 1;
556 		vap->va_uid = 0;
557 		vap->va_gid = 0;
558 		vap->va_size = vap->va_bytes =
559 		    sprintf(buf, "%ld", (long)curproc->p_pid);
560 		break;
561 	}
562 
563 	case Pproc:
564 		vap->va_nlink = 2;
565 		vap->va_uid = procp->p_ucred->cr_uid;
566 		vap->va_gid = procp->p_ucred->cr_gid;
567 		vap->va_size = vap->va_bytes = DEV_BSIZE;
568 		break;
569 
570 	case Pfile:
571 		error = EOPNOTSUPP;
572 		break;
573 
574 	case Pmem:
575 		vap->va_bytes = vap->va_size =
576 			ctob(procp->p_vmspace->vm_tsize +
577 				    procp->p_vmspace->vm_dsize +
578 				    procp->p_vmspace->vm_ssize);
579 		break;
580 
581 #if defined(PT_GETREGS) || defined(PT_SETREGS)
582 	case Pregs:
583 		vap->va_bytes = vap->va_size = sizeof(struct reg);
584 		break;
585 #endif
586 
587 #if defined(PT_GETFPREGS) || defined(PT_SETFPREGS)
588 	case Pfpregs:
589 		vap->va_bytes = vap->va_size = sizeof(struct fpreg);
590 		break;
591 #endif
592 
593 	case Pctl:
594 	case Pstatus:
595 	case Pnote:
596 	case Pnotepg:
597 		break;
598 
599 	default:
600 		panic("procfs_getattr");
601 	}
602 
603 	return (error);
604 }
605 
606 /*ARGSUSED*/
607 int
608 procfs_setattr(v)
609 	void *v;
610 {
611 	/*
612 	 * just fake out attribute setting
613 	 * it's not good to generate an error
614 	 * return, otherwise things like creat()
615 	 * will fail when they try to set the
616 	 * file length to 0.  worse, this means
617 	 * that echo $note > /proc/$pid/note will fail.
618 	 */
619 
620 	return (0);
621 }
622 
623 /*
624  * implement access checking.
625  *
626  * actually, the check for super-user is slightly
627  * broken since it will allow read access to write-only
628  * objects.  this doesn't cause any particular trouble
629  * but does mean that the i/o entry points need to check
630  * that the operation really does make sense.
631  */
632 int
633 procfs_access(v)
634 	void *v;
635 {
636 	struct vop_access_args /* {
637 		struct vnode *a_vp;
638 		int a_mode;
639 		struct ucred *a_cred;
640 		struct proc *a_p;
641 	} */ *ap = v;
642 	struct vattr va;
643 	int error;
644 
645 	if ((error = VOP_GETATTR(ap->a_vp, &va, ap->a_cred, ap->a_p)) != 0)
646 		return (error);
647 
648 	return (vaccess(va.va_mode, va.va_uid, va.va_gid, ap->a_mode,
649 			ap->a_cred));
650 }
651 
652 /*
653  * lookup.  this is incredibly complicated in the
654  * general case, however for most pseudo-filesystems
655  * very little needs to be done.
656  *
657  * unless you want to get a migraine, just make sure your
658  * filesystem doesn't do any locking of its own.  otherwise
659  * read and inwardly digest ufs_lookup().
660  */
661 int
662 procfs_lookup(v)
663 	void *v;
664 {
665 	struct vop_lookup_args /* {
666 		struct vnode * a_dvp;
667 		struct vnode ** a_vpp;
668 		struct componentname * a_cnp;
669 	} */ *ap = v;
670 	struct componentname *cnp = ap->a_cnp;
671 	struct vnode **vpp = ap->a_vpp;
672 	struct vnode *dvp = ap->a_dvp;
673 	const char *pname = cnp->cn_nameptr;
674 	struct proc_target *pt;
675 	struct vnode *fvp;
676 	pid_t pid;
677 	struct pfsnode *pfs;
678 	struct proc *p;
679 	int i;
680 
681 	*vpp = NULL;
682 
683 	if (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)
684 		return (EROFS);
685 
686 	if (cnp->cn_namelen == 1 && *pname == '.') {
687 		*vpp = dvp;
688 		VREF(dvp);
689 		/*VOP_LOCK(dvp);*/
690 		return (0);
691 	}
692 
693 	pfs = VTOPFS(dvp);
694 	switch (pfs->pfs_type) {
695 	case Proot:
696 		if (cnp->cn_flags & ISDOTDOT)
697 			return (EIO);
698 
699 		if (CNEQ(cnp, "curproc", 7))
700 			return (procfs_allocvp(dvp->v_mount, vpp, 0, Pcurproc));
701 
702 		pid = atopid(pname, cnp->cn_namelen);
703 		if (pid == NO_PID)
704 			break;
705 
706 		p = PFIND(pid);
707 		if (p == 0)
708 			break;
709 
710 		return (procfs_allocvp(dvp->v_mount, vpp, pid, Pproc));
711 
712 	case Pproc:
713 		if (cnp->cn_flags & ISDOTDOT)
714 			return (procfs_root(dvp->v_mount, vpp));
715 
716 		p = PFIND(pfs->pfs_pid);
717 		if (p == 0)
718 			break;
719 
720 		for (pt = proc_targets, i = 0; i < nproc_targets; pt++, i++) {
721 			if (cnp->cn_namelen == pt->pt_namlen &&
722 			    bcmp(pt->pt_name, pname, cnp->cn_namelen) == 0 &&
723 			    (pt->pt_valid == NULL || (*pt->pt_valid)(p)))
724 				goto found;
725 		}
726 		break;
727 
728 	found:
729 		if (pt->pt_pfstype == Pfile) {
730 			fvp = procfs_findtextvp(p);
731 			/* We already checked that it exists. */
732 			VREF(fvp);
733 			VOP_LOCK(fvp);
734 			*vpp = fvp;
735 			return (0);
736 		}
737 
738 		return (procfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid,
739 		    pt->pt_pfstype));
740 
741 	default:
742 		return (ENOTDIR);
743 	}
744 
745 	return (cnp->cn_nameiop == LOOKUP ? ENOENT : EROFS);
746 }
747 
748 int
749 procfs_validfile(p)
750 	struct proc *p;
751 {
752 
753 	return (procfs_findtextvp(p) != NULLVP);
754 }
755 
756 /*
757  * readdir returns directory entries from pfsnode (vp).
758  *
759  * the strategy here with procfs is to generate a single
760  * directory entry at a time (struct dirent) and then
761  * copy that out to userland using uiomove.  a more efficent
762  * though more complex implementation, would try to minimize
763  * the number of calls to uiomove().  for procfs, this is
764  * hardly worth the added code complexity.
765  *
766  * this should just be done through read()
767  */
768 int
769 procfs_readdir(v)
770 	void *v;
771 {
772 	struct vop_readdir_args /* {
773 		struct vnode *a_vp;
774 		struct uio *a_uio;
775 		struct ucred *a_cred;
776 		int *a_eofflag;
777 		u_long *a_cookies;
778 		int a_ncookies;
779 	} */ *ap = v;
780 	struct uio *uio = ap->a_uio;
781 	struct dirent d;
782 	struct pfsnode *pfs;
783 	int i;
784 	int error;
785 	u_long *cookies = ap->a_cookies;
786 	int ncookies = ap->a_ncookies;
787 
788 	pfs = VTOPFS(ap->a_vp);
789 
790 	if (uio->uio_resid < UIO_MX)
791 		return (EINVAL);
792 	if (uio->uio_offset < 0)
793 		return (EINVAL);
794 
795 	error = 0;
796 	i = uio->uio_offset;
797 	bzero((caddr_t)&d, UIO_MX);
798 	d.d_reclen = UIO_MX;
799 
800 	switch (pfs->pfs_type) {
801 	/*
802 	 * this is for the process-specific sub-directories.
803 	 * all that is needed to is copy out all the entries
804 	 * from the procent[] table (top of this file).
805 	 */
806 	case Pproc: {
807 		struct proc *p;
808 		struct proc_target *pt;
809 
810 		p = PFIND(pfs->pfs_pid);
811 		if (p == NULL)
812 			break;
813 
814 		for (pt = &proc_targets[i];
815 		     uio->uio_resid >= UIO_MX && i < nproc_targets; pt++, i++) {
816 			if (pt->pt_valid && (*pt->pt_valid)(p) == 0)
817 				continue;
818 
819 			d.d_fileno = PROCFS_FILENO(pfs->pfs_pid, pt->pt_pfstype);
820 			d.d_namlen = pt->pt_namlen;
821 			bcopy(pt->pt_name, d.d_name, pt->pt_namlen + 1);
822 			d.d_type = pt->pt_type;
823 
824 			if ((error = uiomove((caddr_t)&d, UIO_MX, uio)) != 0)
825 				break;
826 			if (ncookies-- > 0)
827 				*cookies++ = i + 1;
828 		}
829 
830 	    	break;
831 	}
832 
833 	/*
834 	 * this is for the root of the procfs filesystem
835 	 * what is needed is a special entry for "curproc"
836 	 * followed by an entry for each process on allproc
837 #ifdef PROCFS_ZOMBIE
838 	 * and zombproc.
839 #endif
840 	 */
841 
842 	case Proot: {
843 #ifdef PROCFS_ZOMBIE
844 		int doingzomb = 0;
845 #endif
846 		int pcnt = i;
847 		volatile struct proc *p = allproc.lh_first;
848 
849 		if (pcnt > 3)
850 			pcnt = 3;
851 #ifdef PROCFS_ZOMBIE
852 	again:
853 #endif
854 		for (; p && uio->uio_resid >= UIO_MX; i++, pcnt++) {
855 			switch (i) {
856 			case 0:		/* `.' */
857 			case 1:		/* `..' */
858 				d.d_fileno = PROCFS_FILENO(0, Proot);
859 				d.d_namlen = i + 1;
860 				bcopy("..", d.d_name, d.d_namlen);
861 				d.d_name[i + 1] = '\0';
862 				d.d_type = DT_DIR;
863 				break;
864 
865 			case 2:
866 				d.d_fileno = PROCFS_FILENO(0, Pcurproc);
867 				d.d_namlen = 7;
868 				bcopy("curproc", d.d_name, 8);
869 				d.d_type = DT_LNK;
870 				break;
871 
872 			default:
873 				while (pcnt < i) {
874 					pcnt++;
875 					p = p->p_list.le_next;
876 					if (!p)
877 						goto done;
878 				}
879 				d.d_fileno = PROCFS_FILENO(p->p_pid, Pproc);
880 				d.d_namlen = sprintf(d.d_name, "%ld",
881 				    (long)p->p_pid);
882 				d.d_type = DT_REG;
883 				p = p->p_list.le_next;
884 				break;
885 			}
886 
887 			if ((error = uiomove((caddr_t)&d, UIO_MX, uio)) != 0)
888 				break;
889 			if (ncookies-- > 0)
890 				*cookies++ = i + 1;
891 		}
892 	done:
893 
894 #ifdef PROCFS_ZOMBIE
895 		if (p == 0 && doingzomb == 0) {
896 			doingzomb = 1;
897 			p = zombproc.lh_first;
898 			goto again;
899 		}
900 #endif
901 
902 		break;
903 
904 	}
905 
906 	default:
907 		error = ENOTDIR;
908 		break;
909 	}
910 
911 	uio->uio_offset = i;
912 	return (error);
913 }
914 
915 /*
916  * readlink reads the link of `curproc'
917  */
918 int
919 procfs_readlink(v)
920 	void *v;
921 {
922 	struct vop_readlink_args *ap = v;
923 	char buf[16];		/* should be enough */
924 	int len;
925 
926 	if (VTOPFS(ap->a_vp)->pfs_fileno != PROCFS_FILENO(0, Pcurproc))
927 		return (EINVAL);
928 
929 	len = sprintf(buf, "%ld", (long)curproc->p_pid);
930 
931 	return (uiomove((caddr_t)buf, len, ap->a_uio));
932 }
933 
934 /*
935  * convert decimal ascii to pid_t
936  */
937 static pid_t
938 atopid(b, len)
939 	const char *b;
940 	u_int len;
941 {
942 	pid_t p = 0;
943 
944 	while (len--) {
945 		char c = *b++;
946 		if (c < '0' || c > '9')
947 			return (NO_PID);
948 		p = 10 * p + (c - '0');
949 		if (p > PID_MAX)
950 			return (NO_PID);
951 	}
952 
953 	return (p);
954 }
955