xref: /netbsd-src/sys/ufs/ffs/ffs_extattr.c (revision aef5eb5f59cdfe8314f1b5f78ac04eb144e44010)
1 /*	$NetBSD: ffs_extattr.c,v 1.8 2021/12/14 11:06:50 chs Exp $	*/
2 
3 /*-
4  * SPDX-License-Identifier: (BSD-2-Clause-FreeBSD AND BSD-3-Clause)
5  *
6  * Copyright (c) 2002, 2003 Networks Associates Technology, Inc.
7  * All rights reserved.
8  *
9  * This software was developed for the FreeBSD Project by Marshall
10  * Kirk McKusick and Network Associates Laboratories, the Security
11  * Research Division of Network Associates, Inc. under DARPA/SPAWAR
12  * contract N66001-01-C-8035 ("CBOSS"), as part of the DARPA CHATS
13  * research program
14  *
15  * Redistribution and use in source and binary forms, with or without
16  * modification, are permitted provided that the following conditions
17  * are met:
18  * 1. Redistributions of source code must retain the above copyright
19  *    notice, this list of conditions and the following disclaimer.
20  * 2. Redistributions in binary form must reproduce the above copyright
21  *    notice, this list of conditions and the following disclaimer in the
22  *    documentation and/or other materials provided with the distribution.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  * Copyright (c) 1982, 1986, 1989, 1993
37  *	The Regents of the University of California.  All rights reserved.
38  *
39  * Redistribution and use in source and binary forms, with or without
40  * modification, are permitted provided that the following conditions
41  * are met:
42  * 1. Redistributions of source code must retain the above copyright
43  *    notice, this list of conditions and the following disclaimer.
44  * 2. Redistributions in binary form must reproduce the above copyright
45  *    notice, this list of conditions and the following disclaimer in the
46  *    documentation and/or other materials provided with the distribution.
47  * 3. Neither the name of the University nor the names of its contributors
48  *    may be used to endorse or promote products derived from this software
49  *    without specific prior written permission.
50  *
51  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
52  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
53  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
54  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
55  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
56  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
57  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
58  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
59  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
60  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
61  * SUCH DAMAGE.
62  *
63  *	from: @(#)ufs_readwrite.c	8.11 (Berkeley) 5/8/95
64  * from: $FreeBSD: .../ufs/ufs_readwrite.c,v 1.96 2002/08/12 09:22:11 phk ...
65  *	@(#)ffs_vnops.c	8.15 (Berkeley) 5/14/95
66  */
67 
68 #include <sys/cdefs.h>
69 __KERNEL_RCSID(0, "$NetBSD: ffs_extattr.c,v 1.8 2021/12/14 11:06:50 chs Exp $");
70 
71 #if defined(_KERNEL_OPT)
72 #include "opt_ffs.h"
73 #include "opt_wapbl.h"
74 #endif
75 
76 #include <sys/param.h>
77 #include <sys/systm.h>
78 #include <sys/resourcevar.h>
79 #include <sys/kernel.h>
80 #include <sys/file.h>
81 #include <sys/stat.h>
82 #include <sys/buf.h>
83 #include <sys/event.h>
84 #include <sys/extattr.h>
85 #include <sys/kauth.h>
86 #include <sys/proc.h>
87 #include <sys/mount.h>
88 #include <sys/vnode.h>
89 #include <sys/malloc.h>
90 #include <sys/pool.h>
91 #include <sys/signalvar.h>
92 #include <sys/kauth.h>
93 #include <sys/wapbl.h>
94 
95 #include <miscfs/fifofs/fifo.h>
96 #include <miscfs/genfs/genfs.h>
97 #include <miscfs/specfs/specdev.h>
98 
99 #include <ufs/ufs/inode.h>
100 #include <ufs/ufs/dir.h>
101 #include <ufs/ufs/ufs_extern.h>
102 #include <ufs/ufs/ufsmount.h>
103 #include <ufs/ufs/ufs_wapbl.h>
104 
105 #include <ufs/ffs/fs.h>
106 #include <ufs/ffs/ffs_extern.h>
107 
108 #define ALIGNED_TO(ptr, s)  \
109     (((uintptr_t)(ptr) & (_Alignof(s) - 1)) == 0)
110 #define uoff_t uintmax_t
111 #define ITOFS(ip) (ip)->i_fs
112 #define i_din2 i_din.ffs2_din
113 #define VI_LOCK(vp)		mutex_enter((vp)->v_interlock)
114 #define VI_UNLOCK(vp)		mutex_exit((vp)->v_interlock)
115 #define UFS_INODE_SET_FLAG(ip, f)	((ip)->i_flag |= (f))
116 #define ASSERT_VOP_ELOCKED(vp, m)	KASSERT(VOP_ISLOCKED(vp))
117 #define I_IS_UFS2(ip)		(ITOFS(ip)->fs_magic == FS_UFS2_MAGIC)
118 #define	lblktosize(fs, o)	ffs_lblktosize(fs, o)
119 #define	lblkno(fs, o)		ffs_lblkno(fs, o)
120 #define	blkoff(fs, o)		ffs_blkoff(fs, o)
121 #define	sblksize(fs, o, lbn)	ffs_sblksize(fs, o, lbn)
122 typedef daddr_t ufs_lbn_t;
123 #define msleep(chan, mtx, pri, wmesg, timeo) \
124     mtsleep((chan), (pri), (wmesg), (timeo), *(mtx))
125 #define vm_page_count_severe()		0
126 #define buf_dirty_count_severe()	0
127 #define BA_CLRBUF B_CLRBUF
128 #define IO_ASYNC 0
129 #define vfs_bio_brelse(bp, ioflag) 	brelse(bp, 0)
130 #define vfs_bio_clrbuf(bp) 		clrbuf(bp)
131 #define vfs_bio_set_flags(bp, ioflag) 	__nothing
132 
133 /*
134  * Extended attribute area reading.
135  */
136 static int
137 ffs_extread(struct vnode *vp, struct uio *uio, int ioflag)
138 {
139 	struct inode *ip;
140 	struct ufs2_dinode *dp;
141 	struct fs *fs;
142 	struct buf *bp;
143 	ufs_lbn_t lbn, nextlbn;
144 	off_t bytesinfile;
145 	long size, xfersize, blkoffset;
146 	ssize_t orig_resid;
147 	int error;
148 
149 	ip = VTOI(vp);
150 	fs = ITOFS(ip);
151 	dp = ip->i_din2;
152 
153 #ifdef INVARIANTS
154 	if (uio->uio_rw != UIO_READ || fs->fs_magic != FS_UFS2_MAGIC)
155 		panic("ffs_extread: mode");
156 
157 #endif
158 	orig_resid = uio->uio_resid;
159 	KASSERT(orig_resid >= 0);
160 	if (orig_resid == 0)
161 		return (0);
162 	KASSERT(uio->uio_offset >= 0);
163 
164 	for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
165 		if ((bytesinfile = dp->di_extsize - uio->uio_offset) <= 0)
166 			break;
167 		lbn = lblkno(fs, uio->uio_offset);
168 		nextlbn = lbn + 1;
169 
170 		/*
171 		 * size of buffer.  The buffer representing the
172 		 * end of the file is rounded up to the size of
173 		 * the block type ( fragment or full block,
174 		 * depending ).
175 		 */
176 		size = sblksize(fs, dp->di_extsize, lbn);
177 		blkoffset = blkoff(fs, uio->uio_offset);
178 
179 		/*
180 		 * The amount we want to transfer in this iteration is
181 		 * one FS block less the amount of the data before
182 		 * our startpoint (duh!)
183 		 */
184 		xfersize = fs->fs_bsize - blkoffset;
185 
186 		/*
187 		 * But if we actually want less than the block,
188 		 * or the file doesn't have a whole block more of data,
189 		 * then use the lesser number.
190 		 */
191 		if (uio->uio_resid < xfersize)
192 			xfersize = uio->uio_resid;
193 		if (bytesinfile < xfersize)
194 			xfersize = bytesinfile;
195 
196 		if (lblktosize(fs, nextlbn) >= dp->di_extsize) {
197 			/*
198 			 * Don't do readahead if this is the end of the info.
199 			 */
200 			error = bread(vp, -1 - lbn, size, 0, &bp);
201 		} else {
202 			/*
203 			 * If we have a second block, then
204 			 * fire off a request for a readahead
205 			 * as well as a read. Note that the 4th and 5th
206 			 * arguments point to arrays of the size specified in
207 			 * the 6th argument.
208 			 */
209 			u_int nextsize = sblksize(fs, dp->di_extsize, nextlbn);
210 
211 			nextlbn = -1 - nextlbn;
212 			error = breadn(vp, -1 - lbn,
213 			    size, &nextlbn, &nextsize, 1, 0, &bp);
214 		}
215 		if (error) {
216 			brelse(bp, 0);
217 			bp = NULL;
218 			break;
219 		}
220 
221 		/*
222 		 * We should only get non-zero b_resid when an I/O error
223 		 * has occurred, which should cause us to break above.
224 		 * However, if the short read did not cause an error,
225 		 * then we want to ensure that we do not uiomove bad
226 		 * or uninitialized data.
227 		 */
228 		size -= bp->b_resid;
229 		if (size < xfersize) {
230 			if (size == 0)
231 				break;
232 			xfersize = size;
233 		}
234 
235 		error = uiomove((char *)bp->b_data + blkoffset,
236 					(int)xfersize, uio);
237 		if (error)
238 			break;
239 		vfs_bio_brelse(bp, ioflag);
240 	}
241 
242 	/*
243 	 * This can only happen in the case of an error
244 	 * because the loop above resets bp to NULL on each iteration
245 	 * and on normal completion has not set a new value into it.
246 	 * so it must have come from a 'break' statement
247 	 */
248 	if (bp != NULL)
249 		vfs_bio_brelse(bp, ioflag);
250 	return (error);
251 }
252 /*
253  * Extended attribute area writing.
254  */
255 static int
256 ffs_extwrite(struct vnode *vp, struct uio *uio, int ioflag, kauth_cred_t ucred)
257 {
258 	struct inode *ip;
259 	struct ufs2_dinode *dp;
260 	struct fs *fs;
261 	struct buf *bp;
262 	ufs_lbn_t lbn;
263 	off_t osize;
264 	ssize_t resid;
265 	int blkoffset, error, flags, size, xfersize;
266 
267 	ip = VTOI(vp);
268 	fs = ITOFS(ip);
269 	dp = ip->i_din2;
270 
271 #ifdef INVARIANTS
272 	if (uio->uio_rw != UIO_WRITE || fs->fs_magic != FS_UFS2_MAGIC)
273 		panic("ffs_extwrite: mode");
274 #endif
275 
276 	if (ioflag & IO_APPEND)
277 		uio->uio_offset = dp->di_extsize;
278 	KASSERT(uio->uio_offset >= 0);
279 	if ((uoff_t)uio->uio_offset + uio->uio_resid >
280 	    UFS_NXADDR * fs->fs_bsize)
281 		return (EFBIG);
282 
283 	resid = uio->uio_resid;
284 	osize = dp->di_extsize;
285 	flags = IO_EXT;
286 	if (ioflag & IO_SYNC)
287 		flags |= IO_SYNC;
288 
289 	if ((error = UFS_WAPBL_BEGIN(vp->v_mount)) != 0)
290 		return error;
291 
292 	for (error = 0; uio->uio_resid > 0;) {
293 		lbn = lblkno(fs, uio->uio_offset);
294 		blkoffset = blkoff(fs, uio->uio_offset);
295 		xfersize = fs->fs_bsize - blkoffset;
296 		if (uio->uio_resid < xfersize)
297 			xfersize = uio->uio_resid;
298 
299 		/*
300 		 * We must perform a read-before-write if the transfer size
301 		 * does not cover the entire buffer.
302 		 */
303 		if (fs->fs_bsize > xfersize)
304 			flags |= BA_CLRBUF;
305 		else
306 			flags &= ~BA_CLRBUF;
307 		error = UFS_BALLOC(vp, uio->uio_offset, xfersize,
308 		    ucred, flags, &bp);
309 		if (error != 0)
310 			break;
311 		/*
312 		 * If the buffer is not valid we have to clear out any
313 		 * garbage data from the pages instantiated for the buffer.
314 		 * If we do not, a failed uiomove() during a write can leave
315 		 * the prior contents of the pages exposed to a userland
316 		 * mmap().  XXX deal with uiomove() errors a better way.
317 		 */
318 		if ((bp->b_flags & BC_NOCACHE) && fs->fs_bsize <= xfersize)
319 			vfs_bio_clrbuf(bp);
320 
321 		if (uio->uio_offset + xfersize > dp->di_extsize)
322 			dp->di_extsize = uio->uio_offset + xfersize;
323 
324 		size = sblksize(fs, dp->di_extsize, lbn) - bp->b_resid;
325 		if (size < xfersize)
326 			xfersize = size;
327 
328 		error =
329 		    uiomove((char *)bp->b_data + blkoffset, (int)xfersize, uio);
330 
331 		vfs_bio_set_flags(bp, ioflag);
332 
333 		/*
334 		 * If IO_SYNC each buffer is written synchronously.  Otherwise
335 		 * if we have a severe page deficiency write the buffer
336 		 * asynchronously.  Otherwise try to cluster, and if that
337 		 * doesn't do it then either do an async write (if O_DIRECT),
338 		 * or a delayed write (if not).
339 		 */
340 		if (ioflag & IO_SYNC) {
341 			(void)bwrite(bp);
342 		} else if (vm_page_count_severe() ||
343 			    buf_dirty_count_severe() ||
344 			    xfersize + blkoffset == fs->fs_bsize ||
345 			    (ioflag & (IO_ASYNC | IO_DIRECT)))
346 			bawrite(bp);
347 		else
348 			bdwrite(bp);
349 		if (error || xfersize == 0)
350 			break;
351 		UFS_INODE_SET_FLAG(ip, IN_CHANGE);
352 	}
353 	/*
354 	 * If we successfully wrote any data, and we are not the superuser
355 	 * we clear the setuid and setgid bits as a precaution against
356 	 * tampering.
357 	 */
358 	if ((ip->i_mode & (ISUID | ISGID)) && resid > uio->uio_resid && ucred) {
359 		ip->i_mode &= ~(ISUID | ISGID);
360 		dp->di_mode = ip->i_mode;
361 	}
362 	if (error) {
363 		if (ioflag & IO_UNIT) {
364 			(void)ffs_truncate(vp, osize,
365 			    IO_EXT | (ioflag&IO_SYNC), ucred);
366 			uio->uio_offset -= resid - uio->uio_resid;
367 			uio->uio_resid = resid;
368 		}
369 	} else if (resid > uio->uio_resid && (ioflag & IO_SYNC))
370 		error = ffs_update(vp, NULL, NULL, UPDATE_WAIT);
371 	UFS_WAPBL_END(vp->v_mount);
372 	return (error);
373 }
374 
375 /*
376  * Vnode operating to retrieve a named extended attribute.
377  *
378  * Locate a particular EA (nspace:name) in the area (ptr:length), and return
379  * the length of the EA, and possibly the pointer to the entry and to the data.
380  */
381 static int
382 ffs_findextattr(u_char *ptr, u_int length, int nspace, const char *name,
383     struct extattr **eapp, u_char **eac)
384 {
385 	struct extattr *eap, *eaend;
386 	size_t nlen;
387 
388 	nlen = strlen(name);
389 	KASSERT(ALIGNED_TO(ptr, struct extattr));
390 	eap = (struct extattr *)ptr;
391 	eaend = (struct extattr *)(ptr + length);
392 	for (; eap < eaend; eap = EXTATTR_NEXT(eap)) {
393 		/* make sure this entry is complete */
394 		if (EXTATTR_NEXT(eap) > eaend)
395 			break;
396 		if (eap->ea_namespace != nspace || eap->ea_namelength != nlen
397 		    || memcmp(eap->ea_name, name, nlen) != 0)
398 			continue;
399 		if (eapp != NULL)
400 			*eapp = eap;
401 		if (eac != NULL)
402 			*eac = EXTATTR_CONTENT(eap);
403 		return (EXTATTR_CONTENT_SIZE(eap));
404 	}
405 	return (-1);
406 }
407 
408 static int
409 ffs_rdextattr(u_char **p, struct vnode *vp, int extra)
410 {
411 	struct inode *ip;
412 	struct ufs2_dinode *dp;
413 	struct fs *fs;
414 	struct uio luio;
415 	struct iovec liovec;
416 	u_int easize;
417 	int error;
418 	u_char *eae;
419 
420 	ip = VTOI(vp);
421 	fs = ITOFS(ip);
422 	dp = ip->i_din2;
423 	easize = dp->di_extsize;
424 	if ((uoff_t)easize + extra > UFS_NXADDR * fs->fs_bsize)
425 		return (EFBIG);
426 
427 	eae = malloc(easize + extra, M_TEMP, M_WAITOK);
428 
429 	liovec.iov_base = eae;
430 	liovec.iov_len = easize;
431 	luio.uio_iov = &liovec;
432 	luio.uio_iovcnt = 1;
433 	luio.uio_offset = 0;
434 	luio.uio_resid = easize;
435 	luio.uio_vmspace = vmspace_kernel();
436 	luio.uio_rw = UIO_READ;
437 
438 	error = ffs_extread(vp, &luio, IO_EXT | IO_SYNC);
439 	if (error) {
440 		free(eae, M_TEMP);
441 		return(error);
442 	}
443 	*p = eae;
444 	return (0);
445 }
446 
447 static void
448 ffs_lock_ea(struct vnode *vp)
449 {
450 	genfs_node_wrlock(vp);
451 }
452 
453 static void
454 ffs_unlock_ea(struct vnode *vp)
455 {
456 	genfs_node_unlock(vp);
457 }
458 
459 static int
460 ffs_open_ea(struct vnode *vp, kauth_cred_t cred)
461 {
462 	struct inode *ip;
463 	struct ufs2_dinode *dp;
464 	int error;
465 
466 	ip = VTOI(vp);
467 
468 	ffs_lock_ea(vp);
469 	if (ip->i_ea_area != NULL) {
470 		ip->i_ea_refs++;
471 		ffs_unlock_ea(vp);
472 		return (0);
473 	}
474 	dp = ip->i_din2;
475 	error = ffs_rdextattr(&ip->i_ea_area, vp, 0);
476 	if (error) {
477 		ffs_unlock_ea(vp);
478 		return (error);
479 	}
480 	ip->i_ea_len = dp->di_extsize;
481 	ip->i_ea_error = 0;
482 	ip->i_ea_refs++;
483 	ffs_unlock_ea(vp);
484 	return (0);
485 }
486 
487 /*
488  * Vnode extattr transaction commit/abort
489  */
490 static int
491 ffs_close_ea(struct vnode *vp, int commit, kauth_cred_t cred)
492 {
493 	struct inode *ip;
494 	struct uio luio;
495 	struct iovec liovec;
496 	int error;
497 	struct ufs2_dinode *dp;
498 
499 	ip = VTOI(vp);
500 
501 	if (commit)
502 		KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
503 	else
504 		KASSERT(VOP_ISLOCKED(vp));
505 	ffs_lock_ea(vp);
506 	if (ip->i_ea_area == NULL) {
507 		ffs_unlock_ea(vp);
508 		return (EINVAL);
509 	}
510 	dp = ip->i_din2;
511 	error = ip->i_ea_error;
512 	if (commit && error == 0) {
513 		ASSERT_VOP_ELOCKED(vp, "ffs_close_ea commit");
514 		if (cred == NOCRED)
515 			cred =  lwp0.l_cred;
516 		liovec.iov_base = ip->i_ea_area;
517 		liovec.iov_len = ip->i_ea_len;
518 		luio.uio_iov = &liovec;
519 		luio.uio_iovcnt = 1;
520 		luio.uio_offset = 0;
521 		luio.uio_resid = ip->i_ea_len;
522 		luio.uio_vmspace = vmspace_kernel();
523 		luio.uio_rw = UIO_WRITE;
524 
525 		/* XXX: I'm not happy about truncating to zero size */
526 		if (ip->i_ea_len < dp->di_extsize) {
527 			if ((error = UFS_WAPBL_BEGIN(vp->v_mount)) != 0) {
528 				ffs_unlock_ea(vp);
529 				return error;
530 			}
531 			error = ffs_truncate(vp, 0, IO_EXT, cred);
532 			UFS_WAPBL_END(vp->v_mount);
533 		}
534 		error = ffs_extwrite(vp, &luio, IO_EXT | IO_SYNC, cred);
535 	}
536 	if (--ip->i_ea_refs == 0) {
537 		free(ip->i_ea_area, M_TEMP);
538 		ip->i_ea_area = NULL;
539 		ip->i_ea_len = 0;
540 		ip->i_ea_error = 0;
541 	}
542 	ffs_unlock_ea(vp);
543 	return (error);
544 }
545 
546 /*
547  * Vnode extattr strategy routine for fifos.
548  *
549  * We need to check for a read or write of the external attributes.
550  * Otherwise we just fall through and do the usual thing.
551  */
552 int
553 ffsext_strategy(void *v)
554 {
555 	struct vop_strategy_args /* {
556 		struct vnodeop_desc *a_desc;
557 		struct vnode *a_vp;
558 		struct buf *a_bp;
559 	} */ *ap = v;
560 	struct vnode *vp;
561 	daddr_t lbn;
562 
563 	vp = ap->a_vp;
564 	lbn = ap->a_bp->b_lblkno;
565 	if (I_IS_UFS2(VTOI(vp)) && lbn < 0 && lbn >= -UFS_NXADDR)
566 		return ufs_strategy(ap);
567 	if (vp->v_type == VFIFO)
568 		return vn_fifo_bypass(ap);
569 	panic("spec nodes went here");
570 }
571 
572 /*
573  * Vnode extattr transaction commit/abort
574  */
575 int
576 ffs_openextattr(void *v)
577 {
578 	struct vop_openextattr_args /* {
579 		struct vnode *a_vp;
580 		kauth_cred_t a_cred;
581 		struct proc *a_p;
582 	} */ *ap = v;
583 	struct inode *ip = VTOI(ap->a_vp);
584 	struct fs *fs = ip->i_fs;
585 
586 	/* Not supported for UFS1 file systems. */
587 	if (fs->fs_magic == FS_UFS1_MAGIC)
588 		return (EOPNOTSUPP);
589 
590 #ifdef __FreeBSD__
591 	if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK)
592 		return (EOPNOTSUPP);
593 #endif
594 
595 	return (ffs_open_ea(ap->a_vp, ap->a_cred));
596 }
597 
598 /*
599  * Vnode extattr transaction commit/abort
600  */
601 int
602 ffs_closeextattr(void *v)
603 {
604 	struct vop_closeextattr_args /* {
605 		struct vnode *a_vp;
606 		int a_commit;
607 		kauth_cred_t a_cred;
608 		struct proc *a_p;
609 	} */ *ap = v;
610 	struct inode *ip = VTOI(ap->a_vp);
611 	struct fs *fs = ip->i_fs;
612 
613 	/* Not supported for UFS1 file systems. */
614 	if (fs->fs_magic == FS_UFS1_MAGIC)
615 		return (EOPNOTSUPP);
616 
617 #ifdef __FreeBSD__
618 	if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK)
619 		return (EOPNOTSUPP);
620 #endif
621 
622 	if (ap->a_commit && (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY))
623 		return (EROFS);
624 
625 	return (ffs_close_ea(ap->a_vp, ap->a_commit, ap->a_cred));
626 }
627 
628 /*
629  * Vnode operation to retrieve a named extended attribute.
630  */
631 int
632 ffs_getextattr(void *v)
633 {
634 	struct vop_getextattr_args /* {
635 		struct vnode *a_vp;
636 		int a_attrnamespace;
637 		const char *a_name;
638 		struct uio *a_uio;
639 		size_t *a_size;
640 		kauth_cred_t a_cred;
641 		struct proc *a_p;
642 	} */ *ap = v;
643 	struct vnode *vp = ap->a_vp;
644 	struct inode *ip = VTOI(vp);
645 	struct fs *fs = ip->i_fs;
646 
647 	KASSERT(VOP_ISLOCKED(vp));
648 	if (fs->fs_magic == FS_UFS1_MAGIC) {
649 		return ufs_getextattr(ap);
650 	}
651 
652 	u_char *eae, *p;
653 	unsigned easize;
654 	int error, ealen;
655 
656 #ifdef __FreeBSD__
657 	if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK)
658 		return (EOPNOTSUPP);
659 #endif
660 
661 	error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
662 	    ap->a_cred, VREAD);
663 	if (error)
664 		return (error);
665 
666 	error = ffs_open_ea(ap->a_vp, ap->a_cred);
667 	if (error)
668 		return (error);
669 
670 	eae = ip->i_ea_area;
671 	easize = ip->i_ea_len;
672 
673 	ealen = ffs_findextattr(eae, easize, ap->a_attrnamespace, ap->a_name,
674 	    NULL, &p);
675 	if (ealen >= 0) {
676 		error = 0;
677 		if (ap->a_size != NULL)
678 			*ap->a_size = ealen;
679 		else if (ap->a_uio != NULL)
680 			error = uiomove(p, ealen, ap->a_uio);
681 	} else
682 		error = ENOATTR;
683 
684 	ffs_close_ea(ap->a_vp, 0, ap->a_cred);
685 	return (error);
686 }
687 
688 /*
689  * Vnode operation to set a named attribute.
690  */
691 int
692 ffs_setextattr(void *v)
693 {
694 	struct vop_setextattr_args /* {
695 		struct vnode *a_vp;
696 		int a_attrnamespace;
697 		const char *a_name;
698 		struct uio *a_uio;
699 		kauth_cred_t a_cred;
700 		struct proc *a_p;
701 	} */ *ap = v;
702 	struct vnode *vp = ap->a_vp;
703 	struct inode *ip = VTOI(vp);
704 	struct fs *fs = ip->i_fs;
705 
706 	KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
707 	if (fs->fs_magic == FS_UFS1_MAGIC) {
708 		return ufs_setextattr(ap);
709 	}
710 
711 	struct extattr *eap;
712 	uint32_t ealength, ul;
713 	ssize_t ealen;
714 	int olen, eapad1, eapad2, error, i, easize;
715 	u_char *eae;
716 	void *tmp;
717 
718 	if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK)
719 		return (EOPNOTSUPP);
720 
721 	if (strlen(ap->a_name) == 0)
722 		return (EINVAL);
723 
724 	/* XXX Now unsupported API to delete EAs using NULL uio. */
725 	if (ap->a_uio == NULL)
726 		return (EOPNOTSUPP);
727 
728 	if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY)
729 		return (EROFS);
730 
731 	ealen = ap->a_uio->uio_resid;
732 	if (ealen < 0 || ealen > lblktosize(fs, UFS_NXADDR))
733 		return (EINVAL);
734 
735 	error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
736 	    ap->a_cred, VWRITE);
737 	if (error) {
738 
739 		/*
740 		 * ffs_lock_ea is not needed there, because the vnode
741 		 * must be exclusively locked.
742 		 */
743 		if (ip->i_ea_area != NULL && ip->i_ea_error == 0)
744 			ip->i_ea_error = error;
745 		return (error);
746 	}
747 
748 	error = ffs_open_ea(ap->a_vp, ap->a_cred);
749 	if (error)
750 		return (error);
751 
752 	ealength = sizeof(uint32_t) + 3 + strlen(ap->a_name);
753 	eapad1 = roundup2(ealength, 8) - ealength;
754 	eapad2 = roundup2(ealen, 8) - ealen;
755 	ealength += eapad1 + ealen + eapad2;
756 
757 	/*
758 	 * CEM: rewrites of the same size or smaller could be done in-place
759 	 * instead.  (We don't acquire any fine-grained locks in here either,
760 	 * so we could also do bigger writes in-place.)
761 	 */
762 	eae = malloc(ip->i_ea_len + ealength, M_TEMP, M_WAITOK);
763 	bcopy(ip->i_ea_area, eae, ip->i_ea_len);
764 	easize = ip->i_ea_len;
765 
766 	olen = ffs_findextattr(eae, easize, ap->a_attrnamespace, ap->a_name,
767 	    &eap, NULL);
768         if (olen == -1) {
769 		/* new, append at end */
770 		KASSERT(ALIGNED_TO(eae + easize, struct extattr));
771 		eap = (struct extattr *)(eae + easize);
772 		easize += ealength;
773 	} else {
774 		ul = eap->ea_length;
775 		i = (u_char *)EXTATTR_NEXT(eap) - eae;
776 		if (ul != ealength) {
777 			bcopy(EXTATTR_NEXT(eap), (u_char *)eap + ealength,
778 			    easize - i);
779 			easize += (ealength - ul);
780 		}
781 	}
782 	if (easize > lblktosize(fs, UFS_NXADDR)) {
783 		free(eae, M_TEMP);
784 		ffs_close_ea(ap->a_vp, 0, ap->a_cred);
785 		if (ip->i_ea_area != NULL && ip->i_ea_error == 0)
786 			ip->i_ea_error = ENOSPC;
787 		return (ENOSPC);
788 	}
789 	eap->ea_length = ealength;
790 	eap->ea_namespace = ap->a_attrnamespace;
791 	eap->ea_contentpadlen = eapad2;
792 	eap->ea_namelength = strlen(ap->a_name);
793 	memcpy(eap->ea_name, ap->a_name, strlen(ap->a_name));
794 	bzero(&eap->ea_name[strlen(ap->a_name)], eapad1);
795 	error = uiomove(EXTATTR_CONTENT(eap), ealen, ap->a_uio);
796 	if (error) {
797 		free(eae, M_TEMP);
798 		ffs_close_ea(ap->a_vp, 0, ap->a_cred);
799 		if (ip->i_ea_area != NULL && ip->i_ea_error == 0)
800 			ip->i_ea_error = error;
801 		return (error);
802 	}
803 	bzero((u_char *)EXTATTR_CONTENT(eap) + ealen, eapad2);
804 
805 	tmp = ip->i_ea_area;
806 	ip->i_ea_area = eae;
807 	ip->i_ea_len = easize;
808 	free(tmp, M_TEMP);
809 	error = ffs_close_ea(ap->a_vp, 1, ap->a_cred);
810 	return (error);
811 }
812 
813 /*
814  * Vnode operation to retrieve extended attributes on a vnode.
815  */
816 int
817 ffs_listextattr(void *v)
818 {
819 	struct vop_listextattr_args /* {
820 		struct vnode *a_vp;
821 		int a_attrnamespace;
822 		struct uio *a_uio;
823 		size_t *a_size;
824 		kauth_cred_t a_cred;
825 		struct proc *a_p;
826 	} */ *ap = v;
827 	struct inode *ip = VTOI(ap->a_vp);
828 	struct fs *fs = ip->i_fs;
829 
830 	if (fs->fs_magic == FS_UFS1_MAGIC) {
831 		return ufs_listextattr(ap);
832 	}
833 
834 	struct extattr *eap, *eaend;
835 	int error, ealen;
836 
837 	if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK)
838 		return (EOPNOTSUPP);
839 
840 	error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
841 	    ap->a_cred, VREAD);
842 	if (error)
843 		return (error);
844 
845 	error = ffs_open_ea(ap->a_vp, ap->a_cred);
846 	if (error)
847 		return (error);
848 
849 	error = 0;
850 	if (ap->a_size != NULL)
851 		*ap->a_size = 0;
852 
853 	KASSERT(ALIGNED_TO(ip->i_ea_area, struct extattr));
854 	eap = (struct extattr *)ip->i_ea_area;
855 	eaend = (struct extattr *)(ip->i_ea_area + ip->i_ea_len);
856 	for (; error == 0 && eap < eaend; eap = EXTATTR_NEXT(eap)) {
857 		/* make sure this entry is complete */
858 		if (EXTATTR_NEXT(eap) > eaend)
859 			break;
860 		if (eap->ea_namespace != ap->a_attrnamespace)
861 			continue;
862 
863 		ealen = eap->ea_namelength;
864 		if (ap->a_size != NULL)
865 			*ap->a_size += ealen + 1;
866 		else if (ap->a_uio != NULL)
867 			error = uiomove(&eap->ea_namelength, ealen + 1,
868 			    ap->a_uio);
869 	}
870 
871 	ffs_close_ea(ap->a_vp, 0, ap->a_cred);
872 	return (error);
873 }
874 
875 /*
876  * Vnode operation to remove a named attribute.
877  */
878 int
879 ffs_deleteextattr(void *v)
880 {
881 	struct vop_deleteextattr_args /* {
882 		struct vnode *a_vp;
883 		int a_attrnamespace;
884 		kauth_cred_t a_cred;
885 		struct proc *a_p;
886 	} */ *ap = v;
887 	struct vnode *vp = ap->a_vp;
888 	struct inode *ip = VTOI(vp);
889 	struct fs *fs = ip->i_fs;
890 
891 	if (fs->fs_magic == FS_UFS1_MAGIC) {
892 		return ufs_deleteextattr(ap);
893 	}
894 
895 	struct extattr *eap;
896 	uint32_t ul;
897 	int olen, error, i, easize;
898 	u_char *eae;
899 	void *tmp;
900 
901 #ifdef __FreeBSD__
902 	if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK)
903 		return (EOPNOTSUPP);
904 #endif
905 
906 	if (strlen(ap->a_name) == 0)
907 		return (EINVAL);
908 
909 	if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY)
910 		return (EROFS);
911 
912 	error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
913 	    ap->a_cred, VWRITE);
914 	if (error) {
915 		/*
916 		 * ffs_lock_ea is not needed there, because the vnode
917 		 * must be exclusively locked.
918 		 */
919 		if (ip->i_ea_area != NULL && ip->i_ea_error == 0)
920 			ip->i_ea_error = error;
921 		return (error);
922 	}
923 
924 	error = ffs_open_ea(ap->a_vp, ap->a_cred);
925 	if (error)
926 		return (error);
927 
928 	/* CEM: delete could be done in-place instead */
929 	eae = malloc(ip->i_ea_len, M_TEMP, M_WAITOK);
930 	bcopy(ip->i_ea_area, eae, ip->i_ea_len);
931 	easize = ip->i_ea_len;
932 
933 	olen = ffs_findextattr(eae, easize, ap->a_attrnamespace, ap->a_name,
934 	    &eap, NULL);
935 	if (olen == -1) {
936 		/* delete but nonexistent */
937 		free(eae, M_TEMP);
938 		ffs_close_ea(ap->a_vp, 0, ap->a_cred);
939 		return (ENOATTR);
940 	}
941 	ul = eap->ea_length;
942 	i = (u_char *)EXTATTR_NEXT(eap) - eae;
943 	bcopy(EXTATTR_NEXT(eap), eap, easize - i);
944 	easize -= ul;
945 
946 	tmp = ip->i_ea_area;
947 	ip->i_ea_area = eae;
948 	ip->i_ea_len = easize;
949 	free(tmp, M_TEMP);
950 	error = ffs_close_ea(ap->a_vp, 1, ap->a_cred);
951 	return error;
952 }
953