Home | History | Annotate | Line # | Download | only in ffs
ffs_extattr.c revision 1.5
      1 /*	$NetBSD: ffs_extattr.c,v 1.5 2020/05/16 18:31:53 christos 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 1.5 2020/05/16 18:31:53 christos 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 #include <uvm/uvm.h>
    109 
    110 #define ALIGNED_TO(ptr, s)  \
    111     (((uintptr_t)(ptr) & (_Alignof(s) - 1)) == 0)
    112 #define uoff_t uintmax_t
    113 #define ITOFS(ip) (ip)->i_fs
    114 #define i_din2 i_din.ffs2_din
    115 #define VI_LOCK(vp)		mutex_enter((vp)->v_interlock)
    116 #define VI_UNLOCK(vp)		mutex_exit((vp)->v_interlock)
    117 #define UFS_INODE_SET_FLAG(ip, f)	((ip)->i_flag |= (f))
    118 #define ASSERT_VOP_ELOCKED(vp, m)	KASSERT(VOP_ISLOCKED(vp))
    119 #define I_IS_UFS2(ip)		(ITOFS(ip)->fs_magic == FS_UFS2_MAGIC)
    120 #define	lblktosize(fs, o)	ffs_lblktosize(fs, o)
    121 #define	lblkno(fs, o)		ffs_lblkno(fs, o)
    122 #define	blkoff(fs, o)		ffs_blkoff(fs, o)
    123 #define	sblksize(fs, o, lbn)	ffs_sblksize(fs, o, lbn)
    124 typedef mode_t accmode_t;	/* so that it breaks soon */
    125 typedef daddr_t ufs_lbn_t;
    126 #define msleep(chan, mtx, pri, wmesg, timeo) \
    127     mtsleep((chan), (pri), (wmesg), (timeo), *(mtx))
    128 #define vm_page_count_severe()		0
    129 #define buf_dirty_count_severe()	0
    130 #define BA_CLRBUF B_CLRBUF
    131 #define IO_ASYNC 0
    132 #define vfs_bio_brelse(bp, ioflag) 	brelse(bp, 0)
    133 #define vfs_bio_clrbuf(bp) 		clrbuf(bp)
    134 #define vfs_bio_set_flags(bp, ioflag) 	__nothing
    135 
    136 /*
    137  * Extended attribute area reading.
    138  */
    139 static int
    140 ffs_extread(struct vnode *vp, struct uio *uio, int ioflag)
    141 {
    142 	struct inode *ip;
    143 	struct ufs2_dinode *dp;
    144 	struct fs *fs;
    145 	struct buf *bp;
    146 	ufs_lbn_t lbn, nextlbn;
    147 	off_t bytesinfile;
    148 	long size, xfersize, blkoffset;
    149 	ssize_t orig_resid;
    150 	int error;
    151 
    152 	ip = VTOI(vp);
    153 	fs = ITOFS(ip);
    154 	dp = ip->i_din2;
    155 
    156 #ifdef INVARIANTS
    157 	if (uio->uio_rw != UIO_READ || fs->fs_magic != FS_UFS2_MAGIC)
    158 		panic("ffs_extread: mode");
    159 
    160 #endif
    161 	orig_resid = uio->uio_resid;
    162 	KASSERT(orig_resid >= 0);
    163 	if (orig_resid == 0)
    164 		return (0);
    165 	KASSERT(uio->uio_offset >= 0);
    166 
    167 	for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
    168 		if ((bytesinfile = dp->di_extsize - uio->uio_offset) <= 0)
    169 			break;
    170 		lbn = lblkno(fs, uio->uio_offset);
    171 		nextlbn = lbn + 1;
    172 
    173 		/*
    174 		 * size of buffer.  The buffer representing the
    175 		 * end of the file is rounded up to the size of
    176 		 * the block type ( fragment or full block,
    177 		 * depending ).
    178 		 */
    179 		size = sblksize(fs, dp->di_extsize, lbn);
    180 		blkoffset = blkoff(fs, uio->uio_offset);
    181 
    182 		/*
    183 		 * The amount we want to transfer in this iteration is
    184 		 * one FS block less the amount of the data before
    185 		 * our startpoint (duh!)
    186 		 */
    187 		xfersize = fs->fs_bsize - blkoffset;
    188 
    189 		/*
    190 		 * But if we actually want less than the block,
    191 		 * or the file doesn't have a whole block more of data,
    192 		 * then use the lesser number.
    193 		 */
    194 		if (uio->uio_resid < xfersize)
    195 			xfersize = uio->uio_resid;
    196 		if (bytesinfile < xfersize)
    197 			xfersize = bytesinfile;
    198 
    199 		if (lblktosize(fs, nextlbn) >= dp->di_extsize) {
    200 			/*
    201 			 * Don't do readahead if this is the end of the info.
    202 			 */
    203 			error = bread(vp, -1 - lbn, size, 0, &bp);
    204 		} else {
    205 			/*
    206 			 * If we have a second block, then
    207 			 * fire off a request for a readahead
    208 			 * as well as a read. Note that the 4th and 5th
    209 			 * arguments point to arrays of the size specified in
    210 			 * the 6th argument.
    211 			 */
    212 			u_int nextsize = sblksize(fs, dp->di_extsize, nextlbn);
    213 
    214 			nextlbn = -1 - nextlbn;
    215 			error = breadn(vp, -1 - lbn,
    216 			    size, &nextlbn, &nextsize, 1, 0, &bp);
    217 		}
    218 		if (error) {
    219 			brelse(bp, 0);
    220 			bp = NULL;
    221 			break;
    222 		}
    223 
    224 		/*
    225 		 * We should only get non-zero b_resid when an I/O error
    226 		 * has occurred, which should cause us to break above.
    227 		 * However, if the short read did not cause an error,
    228 		 * then we want to ensure that we do not uiomove bad
    229 		 * or uninitialized data.
    230 		 */
    231 		size -= bp->b_resid;
    232 		if (size < xfersize) {
    233 			if (size == 0)
    234 				break;
    235 			xfersize = size;
    236 		}
    237 
    238 		error = uiomove((char *)bp->b_data + blkoffset,
    239 					(int)xfersize, uio);
    240 		if (error)
    241 			break;
    242 		vfs_bio_brelse(bp, ioflag);
    243 	}
    244 
    245 	/*
    246 	 * This can only happen in the case of an error
    247 	 * because the loop above resets bp to NULL on each iteration
    248 	 * and on normal completion has not set a new value into it.
    249 	 * so it must have come from a 'break' statement
    250 	 */
    251 	if (bp != NULL)
    252 		vfs_bio_brelse(bp, ioflag);
    253 	return (error);
    254 }
    255 /*
    256  * Extended attribute area writing.
    257  */
    258 static int
    259 ffs_extwrite(struct vnode *vp, struct uio *uio, int ioflag, kauth_cred_t ucred)
    260 {
    261 	struct inode *ip;
    262 	struct ufs2_dinode *dp;
    263 	struct fs *fs;
    264 	struct buf *bp;
    265 	ufs_lbn_t lbn;
    266 	off_t osize;
    267 	ssize_t resid;
    268 	int blkoffset, error, flags, size, xfersize;
    269 
    270 	ip = VTOI(vp);
    271 	fs = ITOFS(ip);
    272 	dp = ip->i_din2;
    273 
    274 #ifdef INVARIANTS
    275 	if (uio->uio_rw != UIO_WRITE || fs->fs_magic != FS_UFS2_MAGIC)
    276 		panic("ffs_extwrite: mode");
    277 #endif
    278 
    279 	if (ioflag & IO_APPEND)
    280 		uio->uio_offset = dp->di_extsize;
    281 	KASSERT(uio->uio_offset >= 0);
    282 	if ((uoff_t)uio->uio_offset + uio->uio_resid >
    283 	    UFS_NXADDR * fs->fs_bsize)
    284 		return (EFBIG);
    285 
    286 	resid = uio->uio_resid;
    287 	osize = dp->di_extsize;
    288 	flags = IO_EXT;
    289 	if (ioflag & IO_SYNC)
    290 		flags |= IO_SYNC;
    291 
    292 	if ((error = UFS_WAPBL_BEGIN(vp->v_mount)) != 0)
    293 		return error;
    294 
    295 	for (error = 0; uio->uio_resid > 0;) {
    296 		lbn = lblkno(fs, uio->uio_offset);
    297 		blkoffset = blkoff(fs, uio->uio_offset);
    298 		xfersize = fs->fs_bsize - blkoffset;
    299 		if (uio->uio_resid < xfersize)
    300 			xfersize = uio->uio_resid;
    301 
    302 		/*
    303 		 * We must perform a read-before-write if the transfer size
    304 		 * does not cover the entire buffer.
    305 		 */
    306 		if (fs->fs_bsize > xfersize)
    307 			flags |= BA_CLRBUF;
    308 		else
    309 			flags &= ~BA_CLRBUF;
    310 		error = UFS_BALLOC(vp, uio->uio_offset, xfersize,
    311 		    ucred, flags, &bp);
    312 		if (error != 0)
    313 			break;
    314 		/*
    315 		 * If the buffer is not valid we have to clear out any
    316 		 * garbage data from the pages instantiated for the buffer.
    317 		 * If we do not, a failed uiomove() during a write can leave
    318 		 * the prior contents of the pages exposed to a userland
    319 		 * mmap().  XXX deal with uiomove() errors a better way.
    320 		 */
    321 		if ((bp->b_flags & BC_NOCACHE) && fs->fs_bsize <= xfersize)
    322 			vfs_bio_clrbuf(bp);
    323 
    324 		if (uio->uio_offset + xfersize > dp->di_extsize)
    325 			dp->di_extsize = uio->uio_offset + xfersize;
    326 
    327 		size = sblksize(fs, dp->di_extsize, lbn) - bp->b_resid;
    328 		if (size < xfersize)
    329 			xfersize = size;
    330 
    331 		error =
    332 		    uiomove((char *)bp->b_data + blkoffset, (int)xfersize, uio);
    333 
    334 		vfs_bio_set_flags(bp, ioflag);
    335 
    336 		/*
    337 		 * If IO_SYNC each buffer is written synchronously.  Otherwise
    338 		 * if we have a severe page deficiency write the buffer
    339 		 * asynchronously.  Otherwise try to cluster, and if that
    340 		 * doesn't do it then either do an async write (if O_DIRECT),
    341 		 * or a delayed write (if not).
    342 		 */
    343 		if (ioflag & IO_SYNC) {
    344 			(void)bwrite(bp);
    345 		} else if (vm_page_count_severe() ||
    346 			    buf_dirty_count_severe() ||
    347 			    xfersize + blkoffset == fs->fs_bsize ||
    348 			    (ioflag & (IO_ASYNC | IO_DIRECT)))
    349 			bawrite(bp);
    350 		else
    351 			bdwrite(bp);
    352 		if (error || xfersize == 0)
    353 			break;
    354 		UFS_INODE_SET_FLAG(ip, IN_CHANGE);
    355 	}
    356 	/*
    357 	 * If we successfully wrote any data, and we are not the superuser
    358 	 * we clear the setuid and setgid bits as a precaution against
    359 	 * tampering.
    360 	 */
    361 	if ((ip->i_mode & (ISUID | ISGID)) && resid > uio->uio_resid && ucred) {
    362 		ip->i_mode &= ~(ISUID | ISGID);
    363 		dp->di_mode = ip->i_mode;
    364 	}
    365 	if (error) {
    366 		if (ioflag & IO_UNIT) {
    367 			(void)ffs_truncate(vp, osize,
    368 			    IO_EXT | (ioflag&IO_SYNC), ucred);
    369 			uio->uio_offset -= resid - uio->uio_resid;
    370 			uio->uio_resid = resid;
    371 		}
    372 	} else if (resid > uio->uio_resid && (ioflag & IO_SYNC))
    373 		error = ffs_update(vp, NULL, NULL, UPDATE_WAIT);
    374 	UFS_WAPBL_END(vp->v_mount);
    375 	return (error);
    376 }
    377 
    378 /*
    379  * Vnode operating to retrieve a named extended attribute.
    380  *
    381  * Locate a particular EA (nspace:name) in the area (ptr:length), and return
    382  * the length of the EA, and possibly the pointer to the entry and to the data.
    383  */
    384 static int
    385 ffs_findextattr(u_char *ptr, u_int length, int nspace, const char *name,
    386     struct extattr **eapp, u_char **eac)
    387 {
    388 	struct extattr *eap, *eaend;
    389 	size_t nlen;
    390 
    391 	nlen = strlen(name);
    392 	KASSERT(ALIGNED_TO(ptr, struct extattr));
    393 	eap = (struct extattr *)ptr;
    394 	eaend = (struct extattr *)(ptr + length);
    395 	for (; eap < eaend; eap = EXTATTR_NEXT(eap)) {
    396 		/* make sure this entry is complete */
    397 		if (EXTATTR_NEXT(eap) > eaend)
    398 			break;
    399 		if (eap->ea_namespace != nspace || eap->ea_namelength != nlen
    400 		    || memcmp(eap->ea_name, name, nlen) != 0)
    401 			continue;
    402 		if (eapp != NULL)
    403 			*eapp = eap;
    404 		if (eac != NULL)
    405 			*eac = EXTATTR_CONTENT(eap);
    406 		return (EXTATTR_CONTENT_SIZE(eap));
    407 	}
    408 	return (-1);
    409 }
    410 
    411 static int
    412 ffs_rdextattr(u_char **p, struct vnode *vp, int extra)
    413 {
    414 	struct inode *ip;
    415 	struct ufs2_dinode *dp;
    416 	struct fs *fs;
    417 	struct uio luio;
    418 	struct iovec liovec;
    419 	u_int easize;
    420 	int error;
    421 	u_char *eae;
    422 
    423 	ip = VTOI(vp);
    424 	fs = ITOFS(ip);
    425 	dp = ip->i_din2;
    426 	easize = dp->di_extsize;
    427 	if ((uoff_t)easize + extra > UFS_NXADDR * fs->fs_bsize)
    428 		return (EFBIG);
    429 
    430 	eae = malloc(easize + extra, M_TEMP, M_WAITOK);
    431 
    432 	liovec.iov_base = eae;
    433 	liovec.iov_len = easize;
    434 	luio.uio_iov = &liovec;
    435 	luio.uio_iovcnt = 1;
    436 	luio.uio_offset = 0;
    437 	luio.uio_resid = easize;
    438 	luio.uio_vmspace = vmspace_kernel();
    439 	luio.uio_rw = UIO_READ;
    440 
    441 	error = ffs_extread(vp, &luio, IO_EXT | IO_SYNC);
    442 	if (error) {
    443 		free(eae, M_TEMP);
    444 		return(error);
    445 	}
    446 	*p = eae;
    447 	return (0);
    448 }
    449 
    450 static void
    451 ffs_lock_ea(struct vnode *vp)
    452 {
    453 	genfs_node_wrlock(vp);
    454 }
    455 
    456 static void
    457 ffs_unlock_ea(struct vnode *vp)
    458 {
    459 	genfs_node_unlock(vp);
    460 }
    461 
    462 static int
    463 ffs_open_ea(struct vnode *vp, kauth_cred_t cred)
    464 {
    465 	struct inode *ip;
    466 	struct ufs2_dinode *dp;
    467 	int error;
    468 
    469 	ip = VTOI(vp);
    470 
    471 	ffs_lock_ea(vp);
    472 	if (ip->i_ea_area != NULL) {
    473 		ip->i_ea_refs++;
    474 		ffs_unlock_ea(vp);
    475 		return (0);
    476 	}
    477 	dp = ip->i_din2;
    478 	error = ffs_rdextattr(&ip->i_ea_area, vp, 0);
    479 	if (error) {
    480 		ffs_unlock_ea(vp);
    481 		return (error);
    482 	}
    483 	ip->i_ea_len = dp->di_extsize;
    484 	ip->i_ea_error = 0;
    485 	ip->i_ea_refs++;
    486 	ffs_unlock_ea(vp);
    487 	return (0);
    488 }
    489 
    490 /*
    491  * Vnode extattr transaction commit/abort
    492  */
    493 static int
    494 ffs_close_ea(struct vnode *vp, int commit, kauth_cred_t cred)
    495 {
    496 	struct inode *ip;
    497 	struct uio luio;
    498 	struct iovec liovec;
    499 	int error;
    500 	struct ufs2_dinode *dp;
    501 
    502 	ip = VTOI(vp);
    503 
    504 	if (commit)
    505 		KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
    506 	else
    507 		KASSERT(VOP_ISLOCKED(vp));
    508 	ffs_lock_ea(vp);
    509 	if (ip->i_ea_area == NULL) {
    510 		ffs_unlock_ea(vp);
    511 		return (EINVAL);
    512 	}
    513 	dp = ip->i_din2;
    514 	error = ip->i_ea_error;
    515 	if (commit && error == 0) {
    516 		ASSERT_VOP_ELOCKED(vp, "ffs_close_ea commit");
    517 		if (cred == NOCRED)
    518 			cred =  lwp0.l_cred;
    519 		liovec.iov_base = ip->i_ea_area;
    520 		liovec.iov_len = ip->i_ea_len;
    521 		luio.uio_iov = &liovec;
    522 		luio.uio_iovcnt = 1;
    523 		luio.uio_offset = 0;
    524 		luio.uio_resid = ip->i_ea_len;
    525 		luio.uio_vmspace = vmspace_kernel();
    526 		luio.uio_rw = UIO_WRITE;
    527 
    528 		/* XXX: I'm not happy about truncating to zero size */
    529 		if (ip->i_ea_len < dp->di_extsize) {
    530 			if ((error = UFS_WAPBL_BEGIN(vp->v_mount)) != 0) {
    531 				ffs_unlock_ea(vp);
    532 				return error;
    533 			}
    534 			error = ffs_truncate(vp, 0, IO_EXT, cred);
    535 			UFS_WAPBL_END(vp->v_mount);
    536 		}
    537 		error = ffs_extwrite(vp, &luio, IO_EXT | IO_SYNC, cred);
    538 	}
    539 	if (--ip->i_ea_refs == 0) {
    540 		free(ip->i_ea_area, M_TEMP);
    541 		ip->i_ea_area = NULL;
    542 		ip->i_ea_len = 0;
    543 		ip->i_ea_error = 0;
    544 	}
    545 	ffs_unlock_ea(vp);
    546 	return (error);
    547 }
    548 
    549 /*
    550  * Vnode extattr strategy routine for fifos.
    551  *
    552  * We need to check for a read or write of the external attributes.
    553  * Otherwise we just fall through and do the usual thing.
    554  */
    555 int
    556 ffsext_strategy(void *v)
    557 {
    558 	struct vop_strategy_args /* {
    559 		struct vnodeop_desc *a_desc;
    560 		struct vnode *a_vp;
    561 		struct buf *a_bp;
    562 	} */ *ap = v;
    563 	struct vnode *vp;
    564 	daddr_t lbn;
    565 
    566 	vp = ap->a_vp;
    567 	lbn = ap->a_bp->b_lblkno;
    568 	if (I_IS_UFS2(VTOI(vp)) && lbn < 0 && lbn >= -UFS_NXADDR)
    569 		return ufs_strategy(ap);
    570 	if (vp->v_type == VFIFO)
    571 		return vn_fifo_bypass(ap);
    572 	panic("spec nodes went here");
    573 }
    574 
    575 /*
    576  * Vnode extattr transaction commit/abort
    577  */
    578 int
    579 ffs_openextattr(void *v)
    580 {
    581 	struct vop_openextattr_args /* {
    582 		struct vnode *a_vp;
    583 		kauth_cred_t a_cred;
    584 		struct proc *a_p;
    585 	} */ *ap = v;
    586 	struct inode *ip = VTOI(ap->a_vp);
    587 	struct fs *fs = ip->i_fs;
    588 
    589 	/* Not supported for UFS1 file systems. */
    590 	if (fs->fs_magic == FS_UFS1_MAGIC)
    591 		return (EOPNOTSUPP);
    592 
    593 	if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK)
    594 		return (EOPNOTSUPP);
    595 
    596 	return (ffs_open_ea(ap->a_vp, ap->a_cred));
    597 }
    598 
    599 /*
    600  * Vnode extattr transaction commit/abort
    601  */
    602 int
    603 ffs_closeextattr(void *v)
    604 {
    605 	struct vop_closeextattr_args /* {
    606 		struct vnode *a_vp;
    607 		int a_commit;
    608 		kauth_cred_t a_cred;
    609 		struct proc *a_p;
    610 	} */ *ap = v;
    611 	struct inode *ip = VTOI(ap->a_vp);
    612 	struct fs *fs = ip->i_fs;
    613 
    614 	/* Not supported for UFS1 file systems. */
    615 	if (fs->fs_magic == FS_UFS1_MAGIC)
    616 		return (EOPNOTSUPP);
    617 
    618 	if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK)
    619 		return (EOPNOTSUPP);
    620 
    621 	if (ap->a_commit && (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY))
    622 		return (EROFS);
    623 
    624 	return (ffs_close_ea(ap->a_vp, ap->a_commit, ap->a_cred));
    625 }
    626 
    627 /*
    628  * Vnode operation to retrieve a named extended attribute.
    629  */
    630 int
    631 ffs_getextattr(void *v)
    632 {
    633 	struct vop_getextattr_args /* {
    634 		struct vnode *a_vp;
    635 		int a_attrnamespace;
    636 		const char *a_name;
    637 		struct uio *a_uio;
    638 		size_t *a_size;
    639 		kauth_cred_t a_cred;
    640 		struct proc *a_p;
    641 	} */ *ap = v;
    642 	struct vnode *vp = ap->a_vp;
    643 	struct inode *ip = VTOI(vp);
    644 	struct fs *fs = ip->i_fs;
    645 
    646 	KASSERT(VOP_ISLOCKED(vp));
    647 	if (fs->fs_magic == FS_UFS1_MAGIC) {
    648 		return ufs_getextattr(ap);
    649 	}
    650 
    651 	u_char *eae, *p;
    652 	unsigned easize;
    653 	int error, ealen;
    654 
    655 	if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK)
    656 		return (EOPNOTSUPP);
    657 
    658 	error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
    659 	    ap->a_cred, VREAD);
    660 	if (error)
    661 		return (error);
    662 
    663 	error = ffs_open_ea(ap->a_vp, ap->a_cred);
    664 	if (error)
    665 		return (error);
    666 
    667 	eae = ip->i_ea_area;
    668 	easize = ip->i_ea_len;
    669 
    670 	ealen = ffs_findextattr(eae, easize, ap->a_attrnamespace, ap->a_name,
    671 	    NULL, &p);
    672 	if (ealen >= 0) {
    673 		error = 0;
    674 		if (ap->a_size != NULL)
    675 			*ap->a_size = ealen;
    676 		else if (ap->a_uio != NULL)
    677 			error = uiomove(p, ealen, ap->a_uio);
    678 	} else
    679 		error = ENOATTR;
    680 
    681 	ffs_close_ea(ap->a_vp, 0, ap->a_cred);
    682 	return (error);
    683 }
    684 
    685 /*
    686  * Vnode operation to set a named attribute.
    687  */
    688 int
    689 ffs_setextattr(void *v)
    690 {
    691 	struct vop_setextattr_args /* {
    692 		struct vnode *a_vp;
    693 		int a_attrnamespace;
    694 		const char *a_name;
    695 		struct uio *a_uio;
    696 		kauth_cred_t a_cred;
    697 		struct proc *a_p;
    698 	} */ *ap = v;
    699 	struct vnode *vp = ap->a_vp;
    700 	struct inode *ip = VTOI(vp);
    701 	struct fs *fs = ip->i_fs;
    702 
    703 	KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
    704 	if (fs->fs_magic == FS_UFS1_MAGIC) {
    705 		return ufs_setextattr(ap);
    706 	}
    707 
    708 	struct extattr *eap;
    709 	uint32_t ealength, ul;
    710 	ssize_t ealen;
    711 	int olen, eapad1, eapad2, error, i, easize;
    712 	u_char *eae;
    713 	void *tmp;
    714 
    715 	if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK)
    716 		return (EOPNOTSUPP);
    717 
    718 	if (strlen(ap->a_name) == 0)
    719 		return (EINVAL);
    720 
    721 	/* XXX Now unsupported API to delete EAs using NULL uio. */
    722 	if (ap->a_uio == NULL)
    723 		return (EOPNOTSUPP);
    724 
    725 	if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY)
    726 		return (EROFS);
    727 
    728 	ealen = ap->a_uio->uio_resid;
    729 	if (ealen < 0 || ealen > lblktosize(fs, UFS_NXADDR))
    730 		return (EINVAL);
    731 
    732 	error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
    733 	    ap->a_cred, VWRITE);
    734 	if (error) {
    735 
    736 		/*
    737 		 * ffs_lock_ea is not needed there, because the vnode
    738 		 * must be exclusively locked.
    739 		 */
    740 		if (ip->i_ea_area != NULL && ip->i_ea_error == 0)
    741 			ip->i_ea_error = error;
    742 		return (error);
    743 	}
    744 
    745 	error = ffs_open_ea(ap->a_vp, ap->a_cred);
    746 	if (error)
    747 		return (error);
    748 
    749 	ealength = sizeof(uint32_t) + 3 + strlen(ap->a_name);
    750 	eapad1 = roundup2(ealength, 8) - ealength;
    751 	eapad2 = roundup2(ealen, 8) - ealen;
    752 	ealength += eapad1 + ealen + eapad2;
    753 
    754 	/*
    755 	 * CEM: rewrites of the same size or smaller could be done in-place
    756 	 * instead.  (We don't acquire any fine-grained locks in here either,
    757 	 * so we could also do bigger writes in-place.)
    758 	 */
    759 	eae = malloc(ip->i_ea_len + ealength, M_TEMP, M_WAITOK);
    760 	bcopy(ip->i_ea_area, eae, ip->i_ea_len);
    761 	easize = ip->i_ea_len;
    762 
    763 	olen = ffs_findextattr(eae, easize, ap->a_attrnamespace, ap->a_name,
    764 	    &eap, NULL);
    765         if (olen == -1) {
    766 		/* new, append at end */
    767 		KASSERT(ALIGNED_TO(eae + easize, struct extattr));
    768 		eap = (struct extattr *)(eae + easize);
    769 		easize += ealength;
    770 	} else {
    771 		ul = eap->ea_length;
    772 		i = (u_char *)EXTATTR_NEXT(eap) - eae;
    773 		if (ul != ealength) {
    774 			bcopy(EXTATTR_NEXT(eap), (u_char *)eap + ealength,
    775 			    easize - i);
    776 			easize += (ealength - ul);
    777 		}
    778 	}
    779 	if (easize > lblktosize(fs, UFS_NXADDR)) {
    780 		free(eae, M_TEMP);
    781 		ffs_close_ea(ap->a_vp, 0, ap->a_cred);
    782 		if (ip->i_ea_area != NULL && ip->i_ea_error == 0)
    783 			ip->i_ea_error = ENOSPC;
    784 		return (ENOSPC);
    785 	}
    786 	eap->ea_length = ealength;
    787 	eap->ea_namespace = ap->a_attrnamespace;
    788 	eap->ea_contentpadlen = eapad2;
    789 	eap->ea_namelength = strlen(ap->a_name);
    790 	memcpy(eap->ea_name, ap->a_name, strlen(ap->a_name));
    791 	bzero(&eap->ea_name[strlen(ap->a_name)], eapad1);
    792 	error = uiomove(EXTATTR_CONTENT(eap), ealen, ap->a_uio);
    793 	if (error) {
    794 		free(eae, M_TEMP);
    795 		ffs_close_ea(ap->a_vp, 0, ap->a_cred);
    796 		if (ip->i_ea_area != NULL && ip->i_ea_error == 0)
    797 			ip->i_ea_error = error;
    798 		return (error);
    799 	}
    800 	bzero((u_char *)EXTATTR_CONTENT(eap) + ealen, eapad2);
    801 
    802 	tmp = ip->i_ea_area;
    803 	ip->i_ea_area = eae;
    804 	ip->i_ea_len = easize;
    805 	free(tmp, M_TEMP);
    806 	error = ffs_close_ea(ap->a_vp, 1, ap->a_cred);
    807 	return (error);
    808 }
    809 
    810 /*
    811  * Vnode operation to retrieve extended attributes on a vnode.
    812  */
    813 int
    814 ffs_listextattr(void *v)
    815 {
    816 	struct vop_listextattr_args /* {
    817 		struct vnode *a_vp;
    818 		int a_attrnamespace;
    819 		struct uio *a_uio;
    820 		size_t *a_size;
    821 		kauth_cred_t a_cred;
    822 		struct proc *a_p;
    823 	} */ *ap = v;
    824 	struct inode *ip = VTOI(ap->a_vp);
    825 	struct fs *fs = ip->i_fs;
    826 
    827 	if (fs->fs_magic == FS_UFS1_MAGIC) {
    828 		return ufs_listextattr(ap);
    829 	}
    830 
    831 	struct extattr *eap, *eaend;
    832 	int error, ealen;
    833 
    834 	if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK)
    835 		return (EOPNOTSUPP);
    836 
    837 	error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
    838 	    ap->a_cred, VREAD);
    839 	if (error)
    840 		return (error);
    841 
    842 	error = ffs_open_ea(ap->a_vp, ap->a_cred);
    843 	if (error)
    844 		return (error);
    845 
    846 	error = 0;
    847 	if (ap->a_size != NULL)
    848 		*ap->a_size = 0;
    849 
    850 	KASSERT(ALIGNED_TO(ip->i_ea_area, struct extattr));
    851 	eap = (struct extattr *)ip->i_ea_area;
    852 	eaend = (struct extattr *)(ip->i_ea_area + ip->i_ea_len);
    853 	for (; error == 0 && eap < eaend; eap = EXTATTR_NEXT(eap)) {
    854 		/* make sure this entry is complete */
    855 		if (EXTATTR_NEXT(eap) > eaend)
    856 			break;
    857 		if (eap->ea_namespace != ap->a_attrnamespace)
    858 			continue;
    859 
    860 		ealen = eap->ea_namelength;
    861 		if (ap->a_size != NULL)
    862 			*ap->a_size += ealen + 1;
    863 		else if (ap->a_uio != NULL)
    864 			error = uiomove(&eap->ea_namelength, ealen + 1,
    865 			    ap->a_uio);
    866 	}
    867 
    868 	ffs_close_ea(ap->a_vp, 0, ap->a_cred);
    869 	return (error);
    870 }
    871 
    872 /*
    873  * Vnode operation to remove a named attribute.
    874  */
    875 int
    876 ffs_deleteextattr(void *v)
    877 {
    878 	struct vop_deleteextattr_args /* {
    879 		struct vnode *a_vp;
    880 		int a_attrnamespace;
    881 		kauth_cred_t a_cred;
    882 		struct proc *a_p;
    883 	} */ *ap = v;
    884 	struct vnode *vp = ap->a_vp;
    885 	struct inode *ip = VTOI(vp);
    886 	struct fs *fs = ip->i_fs;
    887 
    888 	if (fs->fs_magic == FS_UFS1_MAGIC) {
    889 		return ufs_deleteextattr(ap);
    890 	}
    891 
    892 	struct extattr *eap;
    893 	uint32_t ul;
    894 	int olen, error, i, easize;
    895 	u_char *eae;
    896 	void *tmp;
    897 
    898 	if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK)
    899 		return (EOPNOTSUPP);
    900 
    901 	if (strlen(ap->a_name) == 0)
    902 		return (EINVAL);
    903 
    904 	if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY)
    905 		return (EROFS);
    906 
    907 	error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
    908 	    ap->a_cred, VWRITE);
    909 	if (error) {
    910 		/*
    911 		 * ffs_lock_ea is not needed there, because the vnode
    912 		 * must be exclusively locked.
    913 		 */
    914 		if (ip->i_ea_area != NULL && ip->i_ea_error == 0)
    915 			ip->i_ea_error = error;
    916 		return (error);
    917 	}
    918 
    919 	error = ffs_open_ea(ap->a_vp, ap->a_cred);
    920 	if (error)
    921 		return (error);
    922 
    923 	/* CEM: delete could be done in-place instead */
    924 	eae = malloc(ip->i_ea_len, M_TEMP, M_WAITOK);
    925 	bcopy(ip->i_ea_area, eae, ip->i_ea_len);
    926 	easize = ip->i_ea_len;
    927 
    928 	olen = ffs_findextattr(eae, easize, ap->a_attrnamespace, ap->a_name,
    929 	    &eap, NULL);
    930 	if (olen == -1) {
    931 		/* delete but nonexistent */
    932 		free(eae, M_TEMP);
    933 		ffs_close_ea(ap->a_vp, 0, ap->a_cred);
    934 		return (ENOATTR);
    935 	}
    936 	ul = eap->ea_length;
    937 	i = (u_char *)EXTATTR_NEXT(eap) - eae;
    938 	bcopy(EXTATTR_NEXT(eap), eap, easize - i);
    939 	easize -= ul;
    940 
    941 	tmp = ip->i_ea_area;
    942 	ip->i_ea_area = eae;
    943 	ip->i_ea_len = easize;
    944 	free(tmp, M_TEMP);
    945 	error = ffs_close_ea(ap->a_vp, 1, ap->a_cred);
    946 	return error;
    947 }
    948