Home | History | Annotate | Download | only in nfs
      1 /*
      2  * CDDL HEADER START
      3  *
      4  * The contents of this file are subject to the terms of the
      5  * Common Development and Distribution License (the "License").
      6  * You may not use this file except in compliance with the License.
      7  *
      8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
      9  * or http://www.opensolaris.org/os/licensing.
     10  * See the License for the specific language governing permissions
     11  * and limitations under the License.
     12  *
     13  * When distributing Covered Code, include this CDDL HEADER in each
     14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
     15  * If applicable, add the following below this CDDL HEADER, with the
     16  * fields enclosed by brackets "[]" replaced with your own identifying
     17  * information: Portions Copyright [yyyy] [name of copyright owner]
     18  *
     19  * CDDL HEADER END
     20  */
     21 /*
     22  * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
     23  * Use is subject to license terms.
     24  */
     25 
     26 #include <sys/param.h>
     27 #include <sys/types.h>
     28 #include <sys/systm.h>
     29 #include <sys/cred.h>
     30 #include <sys/buf.h>
     31 #include <sys/vfs.h>
     32 #include <sys/vnode.h>
     33 #include <sys/uio.h>
     34 #include <sys/errno.h>
     35 #include <sys/sysmacros.h>
     36 #include <sys/statvfs.h>
     37 #include <sys/kmem.h>
     38 #include <sys/dirent.h>
     39 #include <rpc/types.h>
     40 #include <rpc/auth.h>
     41 #include <rpc/rpcsec_gss.h>
     42 #include <rpc/svc.h>
     43 #include <sys/strsubr.h>
     44 #include <sys/strsun.h>
     45 #include <sys/sdt.h>
     46 
     47 #include <nfs/nfs.h>
     48 #include <nfs/export.h>
     49 #include <nfs/nfs4.h>
     50 #include <nfs/nfs_cmd.h>
     51 
     52 
     53 /*
     54  * RFS4_MINLEN_ENTRY4: XDR-encoded size of smallest possible dirent.
     55  *	This is used to return NFS4ERR_TOOSMALL when clients specify
     56  *	maxcount that isn't large enough to hold the smallest possible
     57  *	XDR encoded dirent.
     58  *
     59  *	    sizeof cookie (8 bytes) +
     60  *	    sizeof name_len (4 bytes) +
     61  *	    sizeof smallest (padded) name (4 bytes) +
     62  *	    sizeof bitmap4_len (12 bytes) +   NOTE: we always encode len=2 bm4
     63  *	    sizeof attrlist4_len (4 bytes) +
     64  *	    sizeof next boolean (4 bytes)
     65  *
     66  * RFS4_MINLEN_RDDIR4: XDR-encoded size of READDIR op reply containing
     67  * the smallest possible entry4 (assumes no attrs requested).
     68  *	sizeof nfsstat4 (4 bytes) +
     69  *	sizeof verifier4 (8 bytes) +
     70  *	sizeof entsecond_to_ry4list bool (4 bytes) +
     71  *	sizeof entry4 	(36 bytes) +
     72  *	sizeof eof bool  (4 bytes)
     73  *
     74  * RFS4_MINLEN_RDDIR_BUF: minimum length of buffer server will provide to
     75  *	VOP_READDIR.  Its value is the size of the maximum possible dirent
     76  *	for solaris.  The DIRENT64_RECLEN macro returns	the size of dirent
     77  *	required for a given name length.  MAXNAMELEN is the maximum
     78  *	filename length allowed in Solaris.  The first two DIRENT64_RECLEN()
     79  *	macros are to allow for . and .. entries -- just a minor tweak to try
     80  *	and guarantee that buffer we give to VOP_READDIR will be large enough
     81  *	to hold ., .., and the largest possible solaris dirent64.
     82  */
     83 #define	RFS4_MINLEN_ENTRY4 36
     84 #define	RFS4_MINLEN_RDDIR4 (4 + NFS4_VERIFIER_SIZE + 4 + RFS4_MINLEN_ENTRY4 + 4)
     85 #define	RFS4_MINLEN_RDDIR_BUF \
     86 	(DIRENT64_RECLEN(1) + DIRENT64_RECLEN(2) + DIRENT64_RECLEN(MAXNAMELEN))
     87 
     88 
     89 #ifdef	nextdp
     90 #undef nextdp
     91 #endif
     92 #define	nextdp(dp)	((struct dirent64 *)((char *)(dp) + (dp)->d_reclen))
     93 
     94 verifier4	Readdir4verf = 0x0;
     95 
     96 static nfs_ftype4 vt_to_nf4[] = {
     97 	0, NF4REG, NF4DIR, NF4BLK, NF4CHR, NF4LNK, NF4FIFO, 0, 0, NF4SOCK, 0
     98 };
     99 
    100 int
    101 nfs4_readdir_getvp(vnode_t *dvp, char *d_name, vnode_t **vpp,
    102 		struct exportinfo **exi, struct svc_req *req,
    103 		struct compound_state *cs, int expseudo)
    104 {
    105 	int error;
    106 	int ismntpt;
    107 	fid_t fid;
    108 	vnode_t *vp, *pre_tvp;
    109 	nfsstat4 status;
    110 	struct exportinfo *newexi, *saveexi;
    111 	cred_t *scr;
    112 
    113 	*vpp = vp = NULL;
    114 
    115 	if (error = VOP_LOOKUP(dvp, d_name, &vp, NULL, 0, NULL, cs->cr,
    116 	    NULL, NULL, NULL))
    117 		return (error);
    118 
    119 	/*
    120 	 * If the directory is a referral point, don't return the
    121 	 * attrs, instead set rdattr_error to MOVED.
    122 	 */
    123 	if (vn_is_nfs_reparse(vp, cs->cr) && !client_is_downrev(req)) {
    124 		VN_RELE(vp);
    125 		DTRACE_PROBE2(nfs4serv__func__referral__moved,
    126 		    vnode_t *, vp, char *, "nfs4_readdir_getvp");
    127 		return (NFS4ERR_MOVED);
    128 	}
    129 
    130 	/* Is this object mounted upon? */
    131 	ismntpt = vn_ismntpt(vp);
    132 
    133 	/*
    134 	 * Nothing more to do if object is not a mount point or
    135 	 * a possible LOFS shadow of an LOFS mount (which won't
    136 	 * have v_vfsmountedhere set)
    137 	 */
    138 	if (ismntpt == 0 && dvp->v_vfsp == vp->v_vfsp && expseudo == 0) {
    139 		*vpp = vp;
    140 		return (0);
    141 	}
    142 
    143 	if (ismntpt) {
    144 		/*
    145 		 * Something is mounted here. Traverse and manage the
    146 		 * namespace
    147 		 */
    148 		pre_tvp = vp;
    149 		VN_HOLD(pre_tvp);
    150 
    151 		if ((error = traverse(&vp)) != 0) {
    152 			VN_RELE(pre_tvp);
    153 			return (error);
    154 		}
    155 		if (vn_is_nfs_reparse(vp, cs->cr)) {
    156 			VN_RELE(vp);
    157 			VN_RELE(pre_tvp);
    158 			DTRACE_PROBE2(nfs4serv__func__referral__moved,
    159 			    vnode_t *, vp, char *, "nfs4_readdir_getvp");
    160 			return (NFS4ERR_MOVED);
    161 		}
    162 	}
    163 
    164 	bzero(&fid, sizeof (fid));
    165 	fid.fid_len = MAXFIDSZ;
    166 
    167 	/*
    168 	 * If VOP_FID not supported by underlying fs (mntfs, procfs,
    169 	 * etc.), then return attrs for stub instead of VROOT object.
    170 	 * If it fails for any other reason, then return the error.
    171 	 */
    172 	if (error = VOP_FID(vp, &fid, NULL)) {
    173 		if (ismntpt == 0) {
    174 			VN_RELE(vp);
    175 			return (error);
    176 		}
    177 
    178 		if (error != ENOSYS && error != ENOTSUP) {
    179 			VN_RELE(vp);
    180 			VN_RELE(pre_tvp);
    181 			return (error);
    182 		}
    183 		/* go back to vnode that is "under" mount */
    184 		VN_RELE(vp);
    185 		*vpp = pre_tvp;
    186 		return (0);
    187 	}
    188 
    189 	newexi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp);
    190 	if (newexi == NULL) {
    191 		if (ismntpt == 0) {
    192 			*vpp = vp;
    193 		} else {
    194 			VN_RELE(vp);
    195 			*vpp = pre_tvp;
    196 		}
    197 		return (0);
    198 	}
    199 
    200 	if (ismntpt)
    201 		VN_RELE(pre_tvp);
    202 
    203 	/* Save the exi and present the new one to checkauth4() */
    204 	saveexi = cs->exi;
    205 	cs->exi = newexi;
    206 
    207 	/* Get the right cred like lookup does */
    208 	scr = cs->cr;
    209 	cs->cr = crdup(cs->basecr);
    210 
    211 	status = call_checkauth4(cs, req);
    212 
    213 	crfree(cs->cr);
    214 	cs->cr = scr;
    215 	cs->exi = saveexi;
    216 
    217 	/* Reset what call_checkauth4() may have set */
    218 	*cs->statusp = NFS4_OK;
    219 
    220 	if (status != NFS4_OK) {
    221 		VN_RELE(vp);
    222 		if (status == NFS4ERR_DELAY)
    223 			status = NFS4ERR_ACCESS;
    224 		return (status);
    225 	}
    226 	*vpp = vp;
    227 	*exi = newexi;
    228 
    229 	return (0);
    230 }
    231 
    232 /* This is the set of pathconf data for vfs */
    233 typedef struct {
    234 	uint64_t maxfilesize;
    235 	uint32_t maxlink;
    236 	uint32_t maxname;
    237 } rfs4_pc_encode_t;
    238 
    239 
    240 static int
    241 rfs4_get_pc_encode(vnode_t *vp, rfs4_pc_encode_t *pce, bitmap4 ar, cred_t *cr)
    242 {
    243 	int error;
    244 	ulong_t pc_val;
    245 
    246 	pce->maxfilesize = 0;
    247 	pce->maxlink = 0;
    248 	pce->maxname = 0;
    249 
    250 	if (ar & FATTR4_MAXFILESIZE_MASK) {
    251 		/* Maximum File Size */
    252 		error = VOP_PATHCONF(vp, _PC_FILESIZEBITS, &pc_val, cr, NULL);
    253 		if (error)
    254 			return (error);
    255 
    256 		/*
    257 		 * If the underlying file system does not support
    258 		 * _PC_FILESIZEBITS, return a reasonable default. Note that
    259 		 * error code on VOP_PATHCONF will be 0, even if the underlying
    260 		 * file system does not support _PC_FILESIZEBITS.
    261 		 */
    262 		if (pc_val == (ulong_t)-1) {
    263 			pce->maxfilesize = MAXOFF32_T;
    264 		} else {
    265 			if (pc_val >= (sizeof (uint64_t) * 8))
    266 				pce->maxfilesize = INT64_MAX;
    267 			else
    268 				pce->maxfilesize = ((1LL << (pc_val - 1)) - 1);
    269 		}
    270 	}
    271 
    272 	if (ar & FATTR4_MAXLINK_MASK) {
    273 		/* Maximum Link Count */
    274 		error = VOP_PATHCONF(vp, _PC_LINK_MAX, &pc_val, cr, NULL);
    275 		if (error)
    276 			return (error);
    277 
    278 		pce->maxlink = pc_val;
    279 	}
    280 
    281 	if (ar & FATTR4_MAXNAME_MASK) {
    282 		/* Maximum Name Length */
    283 		error = VOP_PATHCONF(vp, _PC_NAME_MAX, &pc_val, cr, NULL);
    284 		if (error)
    285 			return (error);
    286 
    287 		pce->maxname = pc_val;
    288 	}
    289 
    290 	return (0);
    291 }
    292 
    293 /* This is the set of statvfs data that is ready for encoding */
    294 typedef struct {
    295 	uint64_t space_avail;
    296 	uint64_t space_free;
    297 	uint64_t space_total;
    298 	u_longlong_t fa;
    299 	u_longlong_t ff;
    300 	u_longlong_t ft;
    301 } rfs4_sb_encode_t;
    302 
    303 static int
    304 rfs4_get_sb_encode(vfs_t *vfsp, rfs4_sb_encode_t *psbe)
    305 {
    306 	int error;
    307 	struct statvfs64 sb;
    308 
    309 	/* Grab the per filesystem info */
    310 	if (error = VFS_STATVFS(vfsp, &sb)) {
    311 		return (error);
    312 	}
    313 
    314 	/* Calculate space available */
    315 	if (sb.f_bavail != (fsblkcnt64_t)-1) {
    316 		psbe->space_avail =
    317 		    (fattr4_space_avail) sb.f_frsize *
    318 		    (fattr4_space_avail) sb.f_bavail;
    319 	} else {
    320 		psbe->space_avail =
    321 		    (fattr4_space_avail) sb.f_bavail;
    322 	}
    323 
    324 	/* Calculate space free */
    325 	if (sb.f_bfree != (fsblkcnt64_t)-1) {
    326 		psbe->space_free =
    327 		    (fattr4_space_free) sb.f_frsize *
    328 		    (fattr4_space_free) sb.f_bfree;
    329 	} else {
    330 		psbe->space_free =
    331 		    (fattr4_space_free) sb.f_bfree;
    332 	}
    333 
    334 	/* Calculate space total */
    335 	if (sb.f_blocks != (fsblkcnt64_t)-1) {
    336 		psbe->space_total =
    337 		    (fattr4_space_total) sb.f_frsize *
    338 		    (fattr4_space_total) sb.f_blocks;
    339 	} else {
    340 		psbe->space_total =
    341 		    (fattr4_space_total) sb.f_blocks;
    342 	}
    343 
    344 	/* For use later on attr encode */
    345 	psbe->fa = sb.f_favail;
    346 	psbe->ff = sb.f_ffree;
    347 	psbe->ft = sb.f_files;
    348 
    349 	return (0);
    350 }
    351 
    352 /*
    353  * Macros to handle if we have don't have enough space for the requested
    354  * attributes and this is the first entry and the
    355  * requested attributes are more than the minimal useful
    356  * set, reset the attributes to the minimal set and
    357  * retry the encoding. If the client has asked for both
    358  * mounted_on_fileid and fileid, prefer mounted_on_fileid.
    359  */
    360 #define	MINIMAL_RD_ATTRS						\
    361 	(FATTR4_MOUNTED_ON_FILEID_MASK|					\
    362 	FATTR4_FILEID_MASK|						\
    363 	FATTR4_RDATTR_ERROR_MASK)
    364 
    365 #define	MINIMIZE_ATTR_MASK(m) {						\
    366 	if ((m) & FATTR4_MOUNTED_ON_FILEID_MASK)			\
    367 	    (m) &= FATTR4_RDATTR_ERROR_MASK|FATTR4_MOUNTED_ON_FILEID_MASK;\
    368 	else								\
    369 	    (m) &= FATTR4_RDATTR_ERROR_MASK|FATTR4_FILEID_MASK;		\
    370 }
    371 
    372 #define	IS_MIN_ATTR_MASK(m)	(((m) & ~MINIMAL_RD_ATTRS) == 0)
    373 /*
    374  * If readdir only needs to return FILEID, we can take it from the
    375  * dirent struct and save doing the lookup.
    376  */
    377 /* ARGSUSED */
    378 void
    379 rfs4_op_readdir(nfs_argop4 *argop, nfs_resop4 *resop,
    380 	struct svc_req *req, struct compound_state *cs)
    381 {
    382 	READDIR4args *args = &argop->nfs_argop4_u.opreaddir;
    383 	READDIR4res *resp = &resop->nfs_resop4_u.opreaddir;
    384 	struct exportinfo *newexi = NULL;
    385 	int error;
    386 	mblk_t *mp;
    387 	uint_t mpcount;
    388 	int alloc_err = 0;
    389 	vnode_t *dvp = cs->vp;
    390 	vnode_t *vp;
    391 	vattr_t va;
    392 	struct dirent64 *dp;
    393 	rfs4_sb_encode_t dsbe, sbe;
    394 	int vfs_different;
    395 	int rddir_data_len, rddir_result_size;
    396 	caddr_t rddir_data;
    397 	offset_t rddir_next_offset;
    398 	int dircount;
    399 	int no_space;
    400 	int iseofdir;
    401 	uint_t eof;
    402 	struct iovec iov;
    403 	struct uio uio;
    404 	int tsize;
    405 	int check_visible;
    406 	int expseudo = 0;
    407 
    408 	uint32_t *ptr, *ptr_redzone;
    409 	uint32_t *beginning_ptr;
    410 	uint32_t *lastentry_ptr;
    411 	uint32_t *attrmask_ptr;
    412 	uint32_t *attr_offset_ptr;
    413 	uint32_t attr_length;
    414 	uint32_t rndup;
    415 	uint32_t namelen;
    416 	uint32_t rddirattr_error = 0;
    417 	int nents;
    418 	bitmap4 ar = args->attr_request & NFS4_SRV_RDDIR_SUPPORTED_ATTRS;
    419 	bitmap4 ae;
    420 	rfs4_pc_encode_t dpce, pce;
    421 	ulong_t pc_val;
    422 	uint64_t maxread;
    423 	uint64_t maxwrite;
    424 	uint_t true = TRUE;
    425 	uint_t false = FALSE;
    426 	uid_t lastuid;
    427 	gid_t lastgid;
    428 	int lu_set, lg_set;
    429 	utf8string owner, group;
    430 	int owner_error, group_error;
    431 	struct sockaddr *ca;
    432 	char *name = NULL;
    433 
    434 	DTRACE_NFSV4_2(op__readdir__start, struct compound_state *, cs,
    435 	    READDIR4args *, args);
    436 
    437 	lu_set = lg_set = 0;
    438 	owner.utf8string_len = group.utf8string_len = 0;
    439 	owner.utf8string_val = group.utf8string_val = NULL;
    440 
    441 	resp->mblk = NULL;
    442 
    443 	/* Maximum read and write size */
    444 	maxread = maxwrite = rfs4_tsize(req);
    445 
    446 	if (dvp == NULL) {
    447 		*cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE;
    448 		goto out;
    449 	}
    450 
    451 	/*
    452 	 * If there is an unshared filesystem mounted on this vnode,
    453 	 * do not allow readdir in this directory.
    454 	 */
    455 	if (vn_ismntpt(dvp)) {
    456 		*cs->statusp = resp->status = NFS4ERR_ACCESS;
    457 		goto out;
    458 	}
    459 
    460 	if (dvp->v_type != VDIR) {
    461 		*cs->statusp = resp->status = NFS4ERR_NOTDIR;
    462 		goto out;
    463 	}
    464 
    465 	if (args->maxcount <= RFS4_MINLEN_RDDIR4) {
    466 		*cs->statusp = resp->status = NFS4ERR_TOOSMALL;
    467 		goto out;
    468 	}
    469 
    470 	/*
    471 	 * If write-only attrs are requested, then fail the readdir op
    472 	 */
    473 	if (args->attr_request &
    474 	    (FATTR4_TIME_MODIFY_SET_MASK | FATTR4_TIME_ACCESS_SET_MASK)) {
    475 		*cs->statusp = resp->status = NFS4ERR_INVAL;
    476 		goto out;
    477 	}
    478 
    479 	error = VOP_ACCESS(dvp, VREAD, 0, cs->cr, NULL);
    480 	if (error) {
    481 		*cs->statusp = resp->status = puterrno4(error);
    482 		goto out;
    483 	}
    484 
    485 	if (args->cookieverf != Readdir4verf) {
    486 		*cs->statusp = resp->status = NFS4ERR_NOT_SAME;
    487 		goto out;
    488 	}
    489 
    490 	/* Is there pseudo-fs work that is needed for this readdir? */
    491 	check_visible = PSEUDO(cs->exi) ||
    492 	    ! is_exported_sec(cs->nfsflavor, cs->exi) ||
    493 	    cs->access & CS_ACCESS_LIMITED;
    494 
    495 	/* Check the requested attributes and only do the work if needed */
    496 
    497 	if (ar & (FATTR4_MAXFILESIZE_MASK |
    498 	    FATTR4_MAXLINK_MASK |
    499 	    FATTR4_MAXNAME_MASK)) {
    500 		if (error = rfs4_get_pc_encode(cs->vp, &dpce, ar, cs->cr)) {
    501 			*cs->statusp = resp->status = puterrno4(error);
    502 			goto out;
    503 		}
    504 		pce = dpce;
    505 	}
    506 
    507 	/* If there is statvfs data requested, pick it up once */
    508 	if (ar &
    509 	    (FATTR4_FILES_AVAIL_MASK |
    510 	    FATTR4_FILES_FREE_MASK |
    511 	    FATTR4_FILES_TOTAL_MASK |
    512 	    FATTR4_FILES_AVAIL_MASK |
    513 	    FATTR4_FILES_FREE_MASK |
    514 	    FATTR4_FILES_TOTAL_MASK)) {
    515 		if (error = rfs4_get_sb_encode(dvp->v_vfsp, &dsbe)) {
    516 			*cs->statusp = resp->status = puterrno4(error);
    517 			goto out;
    518 		}
    519 		sbe = dsbe;
    520 	}
    521 
    522 	/*
    523 	 * Max transfer size of the server is the absolute limite.
    524 	 * If the client has decided to max out with something really
    525 	 * tiny, then return toosmall.  Otherwise, move forward and
    526 	 * see if a single entry can be encoded.
    527 	 */
    528 	tsize = rfs4_tsize(req);
    529 	if (args->maxcount > tsize)
    530 		args->maxcount = tsize;
    531 	else if (args->maxcount < RFS4_MINLEN_RDDIR_BUF) {
    532 		if (args->maxcount < RFS4_MINLEN_ENTRY4) {
    533 			*cs->statusp = resp->status = NFS4ERR_TOOSMALL;
    534 			goto out;
    535 		}
    536 	}
    537 
    538 	/*
    539 	 * How large should the mblk be for outgoing encoding.
    540 	 */
    541 	if (args->maxcount < MAXBSIZE)
    542 		mpcount = MAXBSIZE;
    543 	else
    544 		mpcount = args->maxcount;
    545 
    546 	/*
    547 	 * mp will contain the data to be sent out in the readdir reply.
    548 	 * It will be freed after the reply has been sent.
    549 	 * Let's roundup the data to a BYTES_PER_XDR_UNIX multiple,
    550 	 * so that the call to xdrmblk_putmblk() never fails.
    551 	 */
    552 	mp = allocb(RNDUP(mpcount), BPRI_MED);
    553 
    554 	if (mp == NULL) {
    555 		/*
    556 		 * The allocation of the client's requested size has
    557 		 * failed.  It may be that the size is too large for
    558 		 * current system utilization; step down to a "common"
    559 		 * size and wait for the allocation to occur.
    560 		 */
    561 		if (mpcount > MAXBSIZE)
    562 			args->maxcount = mpcount = MAXBSIZE;
    563 		mp = allocb_wait(RNDUP(mpcount), BPRI_MED,
    564 		    STR_NOSIG, &alloc_err);
    565 	}
    566 
    567 	ASSERT(mp != NULL);
    568 	ASSERT(alloc_err == 0);
    569 
    570 	resp->mblk = mp;
    571 
    572 	ptr = beginning_ptr = (uint32_t *)mp->b_datap->db_base;
    573 
    574 	/*
    575 	 * The "redzone" at the end of the encoding buffer is used
    576 	 * to deal with xdr encoding length.  Instead of checking
    577 	 * each encoding of an attribute value before it is done,
    578 	 * make the assumption that it will fit into the buffer and
    579 	 * check occasionally.
    580 	 *
    581 	 * The largest block of attributes that are encoded without
    582 	 * checking the redzone is 18 * BYTES_PER_XDR_UNIT (72 bytes)
    583 	 * "round" to 128 as the redzone size.
    584 	 */
    585 	if (args->maxcount < (mpcount - 128))
    586 		ptr_redzone =
    587 		    (uint32_t *)(((char *)ptr) + RNDUP(args->maxcount));
    588 	else
    589 		ptr_redzone =
    590 		    (uint32_t *)((((char *)ptr) + RNDUP(mpcount)) - 128);
    591 
    592 	/*
    593 	 * Set the dircount; this will be used as the size for the
    594 	 * readdir of the underlying filesystem.  First make sure
    595 	 * that it is large enough to do a reasonable readdir (client
    596 	 * may have short changed us - it is an advisory number);
    597 	 * then make sure that it isn't too large.
    598 	 * After all of that, if maxcount is "small" then just use
    599 	 * that for the dircount number.
    600 	 */
    601 	dircount = (args->dircount < MAXBSIZE) ? MAXBSIZE : args->dircount;
    602 	dircount = (dircount > tsize) ? tsize : dircount;
    603 	if (dircount > args->maxcount)
    604 		dircount = args->maxcount;
    605 	if (args->maxcount <= MAXBSIZE) {
    606 		if (args->maxcount < RFS4_MINLEN_RDDIR_BUF)
    607 			dircount = RFS4_MINLEN_RDDIR_BUF;
    608 		else
    609 			dircount = args->maxcount;
    610 	}
    611 
    612 	/* number of entries fully encoded in outgoing buffer */
    613 	nents = 0;
    614 
    615 	/* ENCODE READDIR4res.cookieverf */
    616 	IXDR_PUT_HYPER(ptr, Readdir4verf);
    617 
    618 	rddir_data_len = dircount;
    619 	rddir_data = kmem_alloc(rddir_data_len, KM_NOSLEEP);
    620 	if (rddir_data == NULL) {
    621 		/* The allocation failed; downsize and wait for it this time */
    622 		if (rddir_data_len > MAXBSIZE)
    623 			rddir_data_len = dircount = MAXBSIZE;
    624 		rddir_data = kmem_alloc(rddir_data_len, KM_SLEEP);
    625 	}
    626 
    627 	rddir_next_offset = (offset_t)args->cookie;
    628 
    629 	ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
    630 
    631 readagain:
    632 
    633 	no_space = FALSE;
    634 	iseofdir = FALSE;
    635 
    636 	vp = NULL;
    637 
    638 	/* Move on to reading the directory contents */
    639 	iov.iov_base = rddir_data;
    640 	iov.iov_len = rddir_data_len;
    641 	uio.uio_iov = &iov;
    642 	uio.uio_iovcnt = 1;
    643 	uio.uio_segflg = UIO_SYSSPACE;
    644 	uio.uio_extflg = UIO_COPY_CACHED;
    645 	uio.uio_loffset = rddir_next_offset;
    646 	uio.uio_resid = rddir_data_len;
    647 
    648 	(void) VOP_RWLOCK(dvp, V_WRITELOCK_FALSE, NULL);
    649 
    650 	error = VOP_READDIR(dvp, &uio, cs->cr, &iseofdir, NULL, 0);
    651 
    652 	VOP_RWUNLOCK(dvp, V_WRITELOCK_FALSE, NULL);
    653 
    654 	if (error) {
    655 		kmem_free((caddr_t)rddir_data, rddir_data_len);
    656 		freeb(resp->mblk);
    657 		resp->mblk = NULL;
    658 		resp->data_len = 0;
    659 		*cs->statusp = resp->status = puterrno4(error);
    660 		goto out;
    661 	}
    662 
    663 
    664 	rddir_result_size = rddir_data_len - uio.uio_resid;
    665 
    666 	/* No data were read. Check if we reached the end of the directory. */
    667 	if (rddir_result_size == 0) {
    668 		/* encode the BOOLEAN marking no further entries */
    669 		IXDR_PUT_U_INT32(ptr, false);
    670 		/* encode the BOOLEAN signifying end of directory */
    671 		IXDR_PUT_U_INT32(ptr, iseofdir ? true : false);
    672 		resp->data_len = (char *)ptr - (char *)beginning_ptr;
    673 		resp->mblk->b_wptr += resp->data_len;
    674 		kmem_free((caddr_t)rddir_data, rddir_data_len);
    675 		*cs->statusp = resp->status = NFS4_OK;
    676 		goto out;
    677 	}
    678 
    679 	lastentry_ptr = ptr;
    680 	no_space = 0;
    681 	for (dp = (struct dirent64 *)rddir_data;
    682 	    !no_space && rddir_result_size > 0; dp = nextdp(dp)) {
    683 
    684 		/* reset expseudo */
    685 		expseudo = 0;
    686 
    687 		if (vp) {
    688 			VN_RELE(vp);
    689 			vp = NULL;
    690 		}
    691 
    692 		if (newexi)
    693 			newexi = NULL;
    694 
    695 		rddir_result_size -= dp->d_reclen;
    696 
    697 		/* skip "." and ".." entries */
    698 		if (dp->d_ino == 0 || NFS_IS_DOTNAME(dp->d_name)) {
    699 			rddir_next_offset = dp->d_off;
    700 			continue;
    701 		}
    702 
    703 		if (check_visible &&
    704 		    !nfs_visible_inode(cs->exi, dp->d_ino, &expseudo)) {
    705 			rddir_next_offset = dp->d_off;
    706 			continue;
    707 		}
    708 
    709 		/*
    710 		 * Only if the client requested attributes...
    711 		 * If the VOP_LOOKUP fails ENOENT, then skip this entry
    712 		 * for the readdir response.  If there was another error,
    713 		 * then set the rddirattr_error and the error will be
    714 		 * encoded later in the "attributes" section.
    715 		 */
    716 		ae = ar;
    717 		if (ar == 0)
    718 			goto reencode_attrs;
    719 
    720 		error = nfs4_readdir_getvp(dvp, dp->d_name,
    721 		    &vp, &newexi, req, cs, expseudo);
    722 		if (error == ENOENT) {
    723 			rddir_next_offset = dp->d_off;
    724 			continue;
    725 		}
    726 
    727 		rddirattr_error = error;
    728 
    729 		/*
    730 		 * The vp obtained from above may be from a
    731 		 * different filesystem mount and the vfs-like
    732 		 * attributes should be obtained from that
    733 		 * different vfs; only do this if appropriate.
    734 		 */
    735 		if (vp &&
    736 		    (vfs_different = (dvp->v_vfsp != vp->v_vfsp))) {
    737 			if (ar & (FATTR4_FILES_AVAIL_MASK |
    738 			    FATTR4_FILES_FREE_MASK |
    739 			    FATTR4_FILES_TOTAL_MASK |
    740 			    FATTR4_FILES_AVAIL_MASK |
    741 			    FATTR4_FILES_FREE_MASK |
    742 			    FATTR4_FILES_TOTAL_MASK)) {
    743 				if (error =
    744 				    rfs4_get_sb_encode(dvp->v_vfsp,
    745 				    &sbe)) {
    746 					/* Remove attrs from encode */
    747 					ae &= ~(FATTR4_FILES_AVAIL_MASK |
    748 					    FATTR4_FILES_FREE_MASK |
    749 					    FATTR4_FILES_TOTAL_MASK |
    750 					    FATTR4_FILES_AVAIL_MASK |
    751 					    FATTR4_FILES_FREE_MASK |
    752 					    FATTR4_FILES_TOTAL_MASK);
    753 					rddirattr_error = error;
    754 				}
    755 			}
    756 			if (ar & (FATTR4_MAXFILESIZE_MASK |
    757 			    FATTR4_MAXLINK_MASK |
    758 			    FATTR4_MAXNAME_MASK)) {
    759 				if (error = rfs4_get_pc_encode(cs->vp,
    760 				    &pce, ar, cs->cr)) {
    761 					ar &= ~(FATTR4_MAXFILESIZE_MASK |
    762 					    FATTR4_MAXLINK_MASK |
    763 					    FATTR4_MAXNAME_MASK);
    764 					rddirattr_error = error;
    765 				}
    766 			}
    767 		}
    768 
    769 reencode_attrs:
    770 		/* encode the BOOLEAN for the existence of the next entry */
    771 		IXDR_PUT_U_INT32(ptr, true);
    772 		/* encode the COOKIE for the entry */
    773 		IXDR_PUT_U_HYPER(ptr, dp->d_off);
    774 
    775 		name = nfscmd_convname(ca, cs->exi, dp->d_name,
    776 		    NFSCMD_CONV_OUTBOUND, MAXPATHLEN + 1);
    777 
    778 		if (name == NULL) {
    779 			rddir_next_offset = dp->d_off;
    780 			continue;
    781 		}
    782 		/* Calculate the dirent name length */
    783 		namelen = strlen(name);
    784 
    785 		rndup = RNDUP(namelen) / BYTES_PER_XDR_UNIT;
    786 
    787 		/* room for LENGTH + string ? */
    788 		if ((ptr + (1 + rndup)) > ptr_redzone) {
    789 			no_space = TRUE;
    790 			continue;
    791 		}
    792 
    793 		/* encode the LENGTH of the name */
    794 		IXDR_PUT_U_INT32(ptr, namelen);
    795 		/* encode the RNDUP FILL first */
    796 		ptr[rndup - 1] = 0;
    797 		/* encode the NAME of the entry */
    798 		bcopy(name, (char *)ptr, namelen);
    799 		/* now bump the ptr after... */
    800 		ptr += rndup;
    801 
    802 		if (name != dp->d_name)
    803 			kmem_free(name, MAXPATHLEN + 1);
    804 
    805 		/*
    806 		 * Keep checking on the dircount to see if we have
    807 		 * reached the limit; from the RFC, dircount is to be
    808 		 * the XDR encoded limit of the cookie plus name.
    809 		 * So the count is the name, XDR_UNIT of length for
    810 		 * that name and 2 * XDR_UNIT bytes of cookie;
    811 		 * However, use the regular DIRENT64 to match most
    812 		 * client's APIs.
    813 		 */
    814 		dircount -= DIRENT64_RECLEN(namelen);
    815 		if (nents != 0 && dircount < 0) {
    816 			no_space = TRUE;
    817 			continue;
    818 		}
    819 
    820 		/*
    821 		 * Attributes requested?
    822 		 * Gather up the attribute info and the previous VOP_LOOKUP()
    823 		 * succeeded; if an error occurs on the VOP_GETATTR() then
    824 		 * return just the error (again if it is requested).
    825 		 * Note that the previous VOP_LOOKUP() could have failed
    826 		 * itself which leaves this code without anything for
    827 		 * a VOP_GETATTR().
    828 		 * Also note that the readdir_attr_error is left in the
    829 		 * encoding mask if requested and so is the mounted_on_fileid.
    830 		 */
    831 		if (ae != 0) {
    832 			if (!vp) {
    833 				ae = ar & (FATTR4_RDATTR_ERROR_MASK |
    834 				    FATTR4_MOUNTED_ON_FILEID_MASK);
    835 			} else {
    836 				va.va_mask = AT_ALL;
    837 				rddirattr_error =
    838 				    VOP_GETATTR(vp, &va, 0, cs->cr, NULL);
    839 				if (rddirattr_error) {
    840 					ae = ar & (FATTR4_RDATTR_ERROR_MASK |
    841 					    FATTR4_MOUNTED_ON_FILEID_MASK);
    842 				} else {
    843 					/*
    844 					 * We may lie about the object
    845 					 * type for a referral
    846 					 */
    847 					if (vn_is_nfs_reparse(vp, cs->cr) &&
    848 					    client_is_downrev(req))
    849 						va.va_type = VLNK;
    850 				}
    851 			}
    852 		}
    853 
    854 		/* START OF ATTRIBUTE ENCODING */
    855 
    856 		/* encode the LENGTH of the BITMAP4 array */
    857 		IXDR_PUT_U_INT32(ptr, 2);
    858 		/* encode the BITMAP4 */
    859 		attrmask_ptr = ptr;
    860 		IXDR_PUT_HYPER(ptr, ae);
    861 		attr_offset_ptr = ptr;
    862 		/* encode the default LENGTH of the attributes for entry */
    863 		IXDR_PUT_U_INT32(ptr, 0);
    864 
    865 		if (ptr > ptr_redzone) {
    866 			no_space = TRUE;
    867 			continue;
    868 		}
    869 
    870 		/* Check if any of the first 32 attributes are being encoded */
    871 		if (ae & 0xffffffff00000000) {
    872 			/*
    873 			 * Redzone check is done at the end of this section.
    874 			 * This particular section will encode a maximum of
    875 			 * 18 * BYTES_PER_XDR_UNIT of data
    876 			 */
    877 			if (ae &
    878 			    (FATTR4_SUPPORTED_ATTRS_MASK |
    879 			    FATTR4_TYPE_MASK |
    880 			    FATTR4_FH_EXPIRE_TYPE_MASK |
    881 			    FATTR4_CHANGE_MASK |
    882 			    FATTR4_SIZE_MASK |
    883 			    FATTR4_LINK_SUPPORT_MASK |
    884 			    FATTR4_SYMLINK_SUPPORT_MASK |
    885 			    FATTR4_NAMED_ATTR_MASK |
    886 			    FATTR4_FSID_MASK |
    887 			    FATTR4_UNIQUE_HANDLES_MASK |
    888 			    FATTR4_LEASE_TIME_MASK |
    889 			    FATTR4_RDATTR_ERROR_MASK)) {
    890 
    891 				if (ae & FATTR4_SUPPORTED_ATTRS_MASK) {
    892 					IXDR_PUT_INT32(ptr, 2);
    893 					IXDR_PUT_HYPER(ptr,
    894 					    rfs4_supported_attrs);
    895 				}
    896 				if (ae & FATTR4_TYPE_MASK) {
    897 					uint_t ftype = vt_to_nf4[va.va_type];
    898 					if (dvp->v_flag & V_XATTRDIR) {
    899 						if (va.va_type == VDIR)
    900 							ftype = NF4ATTRDIR;
    901 						else
    902 							ftype = NF4NAMEDATTR;
    903 					}
    904 					IXDR_PUT_U_INT32(ptr, ftype);
    905 				}
    906 				if (ae & FATTR4_FH_EXPIRE_TYPE_MASK) {
    907 					uint_t expire_type = FH4_PERSISTENT;
    908 					IXDR_PUT_U_INT32(ptr, expire_type);
    909 				}
    910 				if (ae & FATTR4_CHANGE_MASK) {
    911 					u_longlong_t change;
    912 					NFS4_SET_FATTR4_CHANGE(change,
    913 					    va.va_ctime);
    914 					IXDR_PUT_HYPER(ptr, change);
    915 				}
    916 				if (ae & FATTR4_SIZE_MASK) {
    917 					u_longlong_t size = va.va_size;
    918 					IXDR_PUT_HYPER(ptr, size);
    919 				}
    920 				if (ae & FATTR4_LINK_SUPPORT_MASK) {
    921 					IXDR_PUT_U_INT32(ptr, true);
    922 				}
    923 				if (ae & FATTR4_SYMLINK_SUPPORT_MASK) {
    924 					IXDR_PUT_U_INT32(ptr, true);
    925 				}
    926 				if (ae & FATTR4_NAMED_ATTR_MASK) {
    927 					uint_t isit;
    928 					pc_val = FALSE;
    929 					int sattr_error;
    930 
    931 					if (!(vp->v_vfsp->vfs_flag &
    932 					    VFS_XATTR)) {
    933 						isit = FALSE;
    934 					} else {
    935 						sattr_error = VOP_PATHCONF(vp,
    936 						    _PC_SATTR_EXISTS,
    937 						    &pc_val, cs->cr, NULL);
    938 						if (sattr_error || pc_val == 0)
    939 							(void) VOP_PATHCONF(vp,
    940 							    _PC_XATTR_EXISTS,
    941 							    &pc_val,
    942 							    cs->cr, NULL);
    943 					}
    944 					isit = (pc_val ? TRUE : FALSE);
    945 					IXDR_PUT_U_INT32(ptr, isit);
    946 				}
    947 				if (ae & FATTR4_FSID_MASK) {
    948 					u_longlong_t major, minor;
    949 					struct exportinfo *exi;
    950 
    951 					exi = newexi ? newexi : cs->exi;
    952 					if (exi->exi_volatile_dev) {
    953 						int *pmaj = (int *)&major;
    954 
    955 						pmaj[0] = exi->exi_fsid.val[0];
    956 						pmaj[1] = exi->exi_fsid.val[1];
    957 						minor = 0;
    958 					} else {
    959 						major = getmajor(va.va_fsid);
    960 						minor = getminor(va.va_fsid);
    961 					}
    962 					IXDR_PUT_HYPER(ptr, major);
    963 					IXDR_PUT_HYPER(ptr, minor);
    964 				}
    965 				if (ae & FATTR4_UNIQUE_HANDLES_MASK) {
    966 					IXDR_PUT_U_INT32(ptr, false);
    967 				}
    968 				if (ae & FATTR4_LEASE_TIME_MASK) {
    969 					uint_t lt = rfs4_lease_time;
    970 					IXDR_PUT_U_INT32(ptr, lt);
    971 				}
    972 				if (ae & FATTR4_RDATTR_ERROR_MASK) {
    973 					rddirattr_error =
    974 					    (rddirattr_error == 0 ?
    975 					    0 : puterrno4(rddirattr_error));
    976 					IXDR_PUT_U_INT32(ptr, rddirattr_error);
    977 				}
    978 
    979 				/* Check the redzone boundary */
    980 				if (ptr > ptr_redzone) {
    981 					if (nents || IS_MIN_ATTR_MASK(ar)) {
    982 						no_space = TRUE;
    983 						continue;
    984 					}
    985 					MINIMIZE_ATTR_MASK(ar);
    986 					ae = ar;
    987 					ptr = lastentry_ptr;
    988 					goto reencode_attrs;
    989 				}
    990 			}
    991 			/*
    992 			 * Redzone check is done at the end of this section.
    993 			 * This particular section will encode a maximum of
    994 			 * 4 * BYTES_PER_XDR_UNIT of data.
    995 			 * NOTE: that if ACLs are supported that the
    996 			 * redzone calculations will need to change.
    997 			 */
    998 			if (ae &
    999 			    (FATTR4_ACL_MASK |
   1000 			    FATTR4_ACLSUPPORT_MASK |
   1001 			    FATTR4_ARCHIVE_MASK |
   1002 			    FATTR4_CANSETTIME_MASK |
   1003 			    FATTR4_CASE_INSENSITIVE_MASK |
   1004 			    FATTR4_CASE_PRESERVING_MASK |
   1005 			    FATTR4_CHOWN_RESTRICTED_MASK)) {
   1006 
   1007 				if (ae & FATTR4_ACL_MASK) {
   1008 					ASSERT(0);
   1009 				}
   1010 				if (ae & FATTR4_ACLSUPPORT_MASK) {
   1011 					ASSERT(0);
   1012 				}
   1013 				if (ae & FATTR4_ARCHIVE_MASK) {
   1014 					ASSERT(0);
   1015 				}
   1016 				if (ae & FATTR4_CANSETTIME_MASK) {
   1017 					IXDR_PUT_U_INT32(ptr, true);
   1018 				}
   1019 				if (ae & FATTR4_CASE_INSENSITIVE_MASK) {
   1020 					IXDR_PUT_U_INT32(ptr, false);
   1021 				}
   1022 				if (ae & FATTR4_CASE_PRESERVING_MASK) {
   1023 					IXDR_PUT_U_INT32(ptr, true);
   1024 				}
   1025 				if (ae & FATTR4_CHOWN_RESTRICTED_MASK) {
   1026 					uint_t isit;
   1027 					pc_val = FALSE;
   1028 					(void) VOP_PATHCONF(vp,
   1029 					    _PC_CHOWN_RESTRICTED,
   1030 					    &pc_val, cs->cr, NULL);
   1031 					isit = (pc_val ? TRUE : FALSE);
   1032 					IXDR_PUT_U_INT32(ptr, isit);
   1033 				}
   1034 				/* Check the redzone boundary */
   1035 				if (ptr > ptr_redzone) {
   1036 					if (nents || IS_MIN_ATTR_MASK(ar)) {
   1037 						no_space = TRUE;
   1038 						continue;
   1039 					}
   1040 					MINIMIZE_ATTR_MASK(ar);
   1041 					ae = ar;
   1042 					ptr = lastentry_ptr;
   1043 					goto reencode_attrs;
   1044 				}
   1045 			}
   1046 			/*
   1047 			 * Redzone check is done before the filehandle
   1048 			 * is encoded.
   1049 			 */
   1050 			if (ae &
   1051 			    (FATTR4_FILEHANDLE_MASK |
   1052 			    FATTR4_FILEID_MASK)) {
   1053 
   1054 				if (ae & FATTR4_FILEHANDLE_MASK) {
   1055 					struct {
   1056 						uint_t len;
   1057 						char *val;
   1058 						char fh[NFS_FH4_LEN];
   1059 					} fh;
   1060 					fh.len = 0;
   1061 					fh.val = fh.fh;
   1062 					(void) makefh4((nfs_fh4 *)&fh, vp,
   1063 					    (newexi ? newexi : cs->exi));
   1064 
   1065 					if (dvp->v_flag & V_XATTRDIR)
   1066 						set_fh4_flag((nfs_fh4 *)&fh,
   1067 						    FH4_NAMEDATTR);
   1068 
   1069 					if (!xdr_inline_encode_nfs_fh4(
   1070 					    &ptr, ptr_redzone,
   1071 					    (nfs_fh4_fmt_t *)fh.val)) {
   1072 						if (nents ||
   1073 						    IS_MIN_ATTR_MASK(ar)) {
   1074 							no_space = TRUE;
   1075 							continue;
   1076 						}
   1077 						MINIMIZE_ATTR_MASK(ar);
   1078 						ae = ar;
   1079 						ptr = lastentry_ptr;
   1080 						goto reencode_attrs;
   1081 					}
   1082 				}
   1083 				if (ae & FATTR4_FILEID_MASK) {
   1084 					IXDR_PUT_HYPER(ptr, va.va_nodeid);
   1085 				}
   1086 				/* Check the redzone boundary */
   1087 				if (ptr > ptr_redzone) {
   1088 					if (nents || IS_MIN_ATTR_MASK(ar)) {
   1089 						no_space = TRUE;
   1090 						continue;
   1091 					}
   1092 					MINIMIZE_ATTR_MASK(ar);
   1093 					ae = ar;
   1094 					ptr = lastentry_ptr;
   1095 					goto reencode_attrs;
   1096 				}
   1097 			}
   1098 			/*
   1099 			 * Redzone check is done at the end of this section.
   1100 			 * This particular section will encode a maximum of
   1101 			 * 15 * BYTES_PER_XDR_UNIT of data.
   1102 			 */
   1103 			if (ae &
   1104 			    (FATTR4_FILES_AVAIL_MASK |
   1105 			    FATTR4_FILES_FREE_MASK |
   1106 			    FATTR4_FILES_TOTAL_MASK |
   1107 			    FATTR4_FS_LOCATIONS_MASK |
   1108 			    FATTR4_HIDDEN_MASK |
   1109 			    FATTR4_HOMOGENEOUS_MASK |
   1110 			    FATTR4_MAXFILESIZE_MASK |
   1111 			    FATTR4_MAXLINK_MASK |
   1112 			    FATTR4_MAXNAME_MASK |
   1113 			    FATTR4_MAXREAD_MASK |
   1114 			    FATTR4_MAXWRITE_MASK)) {
   1115 
   1116 				if (ae & FATTR4_FILES_AVAIL_MASK) {
   1117 					IXDR_PUT_HYPER(ptr, sbe.fa);
   1118 				}
   1119 				if (ae & FATTR4_FILES_FREE_MASK) {
   1120 					IXDR_PUT_HYPER(ptr, sbe.ff);
   1121 				}
   1122 				if (ae & FATTR4_FILES_TOTAL_MASK) {
   1123 					IXDR_PUT_HYPER(ptr, sbe.ft);
   1124 				}
   1125 				if (ae & FATTR4_FS_LOCATIONS_MASK) {
   1126 					ASSERT(0);
   1127 				}
   1128 				if (ae & FATTR4_HIDDEN_MASK) {
   1129 					ASSERT(0);
   1130 				}
   1131 				if (ae & FATTR4_HOMOGENEOUS_MASK) {
   1132 					IXDR_PUT_U_INT32(ptr, true);
   1133 				}
   1134 				if (ae & FATTR4_MAXFILESIZE_MASK) {
   1135 					IXDR_PUT_HYPER(ptr, pce.maxfilesize);
   1136 				}
   1137 				if (ae & FATTR4_MAXLINK_MASK) {
   1138 					IXDR_PUT_U_INT32(ptr, pce.maxlink);
   1139 				}
   1140 				if (ae & FATTR4_MAXNAME_MASK) {
   1141 					IXDR_PUT_U_INT32(ptr, pce.maxname);
   1142 				}
   1143 				if (ae & FATTR4_MAXREAD_MASK) {
   1144 					IXDR_PUT_HYPER(ptr, maxread);
   1145 				}
   1146 				if (ae & FATTR4_MAXWRITE_MASK) {
   1147 					IXDR_PUT_HYPER(ptr, maxwrite);
   1148 				}
   1149 				/* Check the redzone boundary */
   1150 				if (ptr > ptr_redzone) {
   1151 					if (nents || IS_MIN_ATTR_MASK(ar)) {
   1152 						no_space = TRUE;
   1153 						continue;
   1154 					}
   1155 					MINIMIZE_ATTR_MASK(ar);
   1156 					ae = ar;
   1157 					ptr = lastentry_ptr;
   1158 					goto reencode_attrs;
   1159 				}
   1160 			}
   1161 		}
   1162 		if (ae & 0x00000000ffffffff) {
   1163 			/*
   1164 			 * Redzone check is done at the end of this section.
   1165 			 * This particular section will encode a maximum of
   1166 			 * 3 * BYTES_PER_XDR_UNIT of data.
   1167 			 */
   1168 			if (ae &
   1169 			    (FATTR4_MIMETYPE_MASK |
   1170 			    FATTR4_MODE_MASK |
   1171 			    FATTR4_NO_TRUNC_MASK |
   1172 			    FATTR4_NUMLINKS_MASK)) {
   1173 
   1174 				if (ae & FATTR4_MIMETYPE_MASK) {
   1175 					ASSERT(0);
   1176 				}
   1177 				if (ae & FATTR4_MODE_MASK) {
   1178 					uint_t m = va.va_mode;
   1179 					IXDR_PUT_U_INT32(ptr, m);
   1180 				}
   1181 				if (ae & FATTR4_NO_TRUNC_MASK) {
   1182 					IXDR_PUT_U_INT32(ptr, true);
   1183 				}
   1184 				if (ae & FATTR4_NUMLINKS_MASK) {
   1185 					IXDR_PUT_U_INT32(ptr, va.va_nlink);
   1186 				}
   1187 				/* Check the redzone boundary */
   1188 				if (ptr > ptr_redzone) {
   1189 					if (nents || IS_MIN_ATTR_MASK(ar)) {
   1190 						no_space = TRUE;
   1191 						continue;
   1192 					}
   1193 					MINIMIZE_ATTR_MASK(ar);
   1194 					ae = ar;
   1195 					ptr = lastentry_ptr;
   1196 					goto reencode_attrs;
   1197 				}
   1198 			}
   1199 			/*
   1200 			 * Redzone check is done before the encoding of the
   1201 			 * owner string since the length is indeterminate.
   1202 			 */
   1203 			if (ae & FATTR4_OWNER_MASK) {
   1204 				if (!lu_set) {
   1205 					owner_error = nfs_idmap_uid_str(
   1206 					    va.va_uid, &owner, TRUE);
   1207 					if (!owner_error) {
   1208 						lu_set = TRUE;
   1209 						lastuid = va.va_uid;
   1210 					}
   1211 				} else 	if (va.va_uid != lastuid) {
   1212 					if (owner.utf8string_len != 0) {
   1213 						kmem_free(owner.utf8string_val,
   1214 						    owner.utf8string_len);
   1215 						owner.utf8string_len = 0;
   1216 						owner.utf8string_val = NULL;
   1217 					}
   1218 					owner_error = nfs_idmap_uid_str(
   1219 					    va.va_uid, &owner, TRUE);
   1220 					if (!owner_error) {
   1221 						lastuid = va.va_uid;
   1222 					} else {
   1223 						lu_set = FALSE;
   1224 					}
   1225 				}
   1226 				if (!owner_error) {
   1227 					if ((ptr +
   1228 					    (owner.utf8string_len /
   1229 					    BYTES_PER_XDR_UNIT)
   1230 					    + 2) > ptr_redzone) {
   1231 						if (nents ||
   1232 						    IS_MIN_ATTR_MASK(ar)) {
   1233 							no_space = TRUE;
   1234 							continue;
   1235 						}
   1236 						MINIMIZE_ATTR_MASK(ar);
   1237 						ae = ar;
   1238 						ptr = lastentry_ptr;
   1239 						goto reencode_attrs;
   1240 					}
   1241 					/* encode the LENGTH of owner string */
   1242 					IXDR_PUT_U_INT32(ptr,
   1243 					    owner.utf8string_len);
   1244 					/* encode the RNDUP FILL first */
   1245 					rndup = RNDUP(owner.utf8string_len) /
   1246 					    BYTES_PER_XDR_UNIT;
   1247 					ptr[rndup - 1] = 0;
   1248 					/* encode the OWNER */
   1249 					bcopy(owner.utf8string_val, ptr,
   1250 					    owner.utf8string_len);
   1251 					ptr += rndup;
   1252 				}
   1253 			}
   1254 			/*
   1255 			 * Redzone check is done before the encoding of the
   1256 			 * group string since the length is indeterminate.
   1257 			 */
   1258 			if (ae & FATTR4_OWNER_GROUP_MASK) {
   1259 				if (!lg_set) {
   1260 					group_error =
   1261 					    nfs_idmap_gid_str(va.va_gid,
   1262 					    &group, TRUE);
   1263 					if (!group_error) {
   1264 						lg_set = TRUE;
   1265 						lastgid = va.va_gid;
   1266 					}
   1267 				} else if (va.va_gid != lastgid) {
   1268 					if (group.utf8string_len != 0) {
   1269 						kmem_free(
   1270 						    group.utf8string_val,
   1271 						    group.utf8string_len);
   1272 						group.utf8string_len = 0;
   1273 						group.utf8string_val = NULL;
   1274 					}
   1275 					group_error =
   1276 					    nfs_idmap_gid_str(va.va_gid,
   1277 					    &group, TRUE);
   1278 					if (!group_error)
   1279 						lastgid = va.va_gid;
   1280 					else
   1281 						lg_set = FALSE;
   1282 				}
   1283 				if (!group_error) {
   1284 					if ((ptr +
   1285 					    (group.utf8string_len /
   1286 					    BYTES_PER_XDR_UNIT)
   1287 					    + 2) > ptr_redzone) {
   1288 						if (nents ||
   1289 						    IS_MIN_ATTR_MASK(ar)) {
   1290 							no_space = TRUE;
   1291 							continue;
   1292 						}
   1293 						MINIMIZE_ATTR_MASK(ar);
   1294 						ae = ar;
   1295 						ptr = lastentry_ptr;
   1296 						goto reencode_attrs;
   1297 					}
   1298 					/* encode the LENGTH of owner string */
   1299 					IXDR_PUT_U_INT32(ptr,
   1300 					    group.utf8string_len);
   1301 					/* encode the RNDUP FILL first */
   1302 					rndup = RNDUP(group.utf8string_len) /
   1303 					    BYTES_PER_XDR_UNIT;
   1304 					ptr[rndup - 1] = 0;
   1305 					/* encode the OWNER */
   1306 					bcopy(group.utf8string_val, ptr,
   1307 					    group.utf8string_len);
   1308 					ptr += rndup;
   1309 				}
   1310 			}
   1311 			if (ae &
   1312 			    (FATTR4_QUOTA_AVAIL_HARD_MASK |
   1313 			    FATTR4_QUOTA_AVAIL_SOFT_MASK |
   1314 			    FATTR4_QUOTA_USED_MASK)) {
   1315 				if (ae & FATTR4_QUOTA_AVAIL_HARD_MASK) {
   1316 					ASSERT(0);
   1317 				}
   1318 				if (ae & FATTR4_QUOTA_AVAIL_SOFT_MASK) {
   1319 					ASSERT(0);
   1320 				}
   1321 				if (ae & FATTR4_QUOTA_USED_MASK) {
   1322 					ASSERT(0);
   1323 				}
   1324 			}
   1325 			/*
   1326 			 * Redzone check is done at the end of this section.
   1327 			 * This particular section will encode a maximum of
   1328 			 * 10 * BYTES_PER_XDR_UNIT of data.
   1329 			 */
   1330 			if (ae &
   1331 			    (FATTR4_RAWDEV_MASK |
   1332 			    FATTR4_SPACE_AVAIL_MASK |
   1333 			    FATTR4_SPACE_FREE_MASK |
   1334 			    FATTR4_SPACE_TOTAL_MASK |
   1335 			    FATTR4_SPACE_USED_MASK |
   1336 			    FATTR4_SYSTEM_MASK)) {
   1337 
   1338 				if (ae & FATTR4_RAWDEV_MASK) {
   1339 					fattr4_rawdev rd;
   1340 					rd.specdata1 =
   1341 					    (uint32)getmajor(va.va_rdev);
   1342 					rd.specdata2 =
   1343 					    (uint32)getminor(va.va_rdev);
   1344 					IXDR_PUT_U_INT32(ptr, rd.specdata1);
   1345 					IXDR_PUT_U_INT32(ptr, rd.specdata2);
   1346 				}
   1347 				if (ae & FATTR4_SPACE_AVAIL_MASK) {
   1348 					IXDR_PUT_HYPER(ptr, sbe.space_avail);
   1349 				}
   1350 				if (ae & FATTR4_SPACE_FREE_MASK) {
   1351 					IXDR_PUT_HYPER(ptr, sbe.space_free);
   1352 				}
   1353 				if (ae & FATTR4_SPACE_TOTAL_MASK) {
   1354 					IXDR_PUT_HYPER(ptr, sbe.space_total);
   1355 				}
   1356 				if (ae & FATTR4_SPACE_USED_MASK) {
   1357 					u_longlong_t su;
   1358 					su = (fattr4_space_used) DEV_BSIZE *
   1359 					    (fattr4_space_used) va.va_nblocks;
   1360 					IXDR_PUT_HYPER(ptr, su);
   1361 				}
   1362 				if (ae & FATTR4_SYSTEM_MASK) {
   1363 					ASSERT(0);
   1364 				}
   1365 				/* Check the redzone boundary */
   1366 				if (ptr > ptr_redzone) {
   1367 					if (nents || IS_MIN_ATTR_MASK(ar)) {
   1368 						no_space = TRUE;
   1369 						continue;
   1370 					}
   1371 					MINIMIZE_ATTR_MASK(ar);
   1372 					ae = ar;
   1373 					ptr = lastentry_ptr;
   1374 					goto reencode_attrs;
   1375 				}
   1376 			}
   1377 			/*
   1378 			 * Redzone check is done at the end of this section.
   1379 			 * This particular section will encode a maximum of
   1380 			 * 14 * BYTES_PER_XDR_UNIT of data.
   1381 			 */
   1382 			if (ae &
   1383 			    (FATTR4_TIME_ACCESS_MASK |
   1384 			    FATTR4_TIME_ACCESS_SET_MASK |
   1385 			    FATTR4_TIME_BACKUP_MASK |
   1386 			    FATTR4_TIME_CREATE_MASK |
   1387 			    FATTR4_TIME_DELTA_MASK |
   1388 			    FATTR4_TIME_METADATA_MASK |
   1389 			    FATTR4_TIME_MODIFY_MASK |
   1390 			    FATTR4_TIME_MODIFY_SET_MASK |
   1391 			    FATTR4_MOUNTED_ON_FILEID_MASK)) {
   1392 
   1393 				if (ae & FATTR4_TIME_ACCESS_MASK) {
   1394 					u_longlong_t sec =
   1395 					    (u_longlong_t)va.va_atime.tv_sec;
   1396 					uint_t nsec =
   1397 					    (uint_t)va.va_atime.tv_nsec;
   1398 					IXDR_PUT_HYPER(ptr, sec);
   1399 					IXDR_PUT_INT32(ptr, nsec);
   1400 				}
   1401 				if (ae & FATTR4_TIME_ACCESS_SET_MASK) {
   1402 					ASSERT(0);
   1403 				}
   1404 				if (ae & FATTR4_TIME_BACKUP_MASK) {
   1405 					ASSERT(0);
   1406 				}
   1407 				if (ae & FATTR4_TIME_CREATE_MASK) {
   1408 					ASSERT(0);
   1409 				}
   1410 				if (ae & FATTR4_TIME_DELTA_MASK) {
   1411 					u_longlong_t sec = 0;
   1412 					uint_t nsec = 1000;
   1413 					IXDR_PUT_HYPER(ptr, sec);
   1414 					IXDR_PUT_INT32(ptr, nsec);
   1415 				}
   1416 				if (ae & FATTR4_TIME_METADATA_MASK) {
   1417 					u_longlong_t sec =
   1418 					    (u_longlong_t)va.va_ctime.tv_sec;
   1419 					uint_t nsec =
   1420 					    (uint_t)va.va_ctime.tv_nsec;
   1421 					IXDR_PUT_HYPER(ptr, sec);
   1422 					IXDR_PUT_INT32(ptr, nsec);
   1423 				}
   1424 				if (ae & FATTR4_TIME_MODIFY_MASK) {
   1425 					u_longlong_t sec =
   1426 					    (u_longlong_t)va.va_mtime.tv_sec;
   1427 					uint_t nsec =
   1428 					    (uint_t)va.va_mtime.tv_nsec;
   1429 					IXDR_PUT_HYPER(ptr, sec);
   1430 					IXDR_PUT_INT32(ptr, nsec);
   1431 				}
   1432 				if (ae & FATTR4_TIME_MODIFY_SET_MASK) {
   1433 					ASSERT(0);
   1434 				}
   1435 				if (ae & FATTR4_MOUNTED_ON_FILEID_MASK) {
   1436 					IXDR_PUT_HYPER(ptr, dp->d_ino);
   1437 				}
   1438 				/* Check the redzone boundary */
   1439 				if (ptr > ptr_redzone) {
   1440 					if (nents || IS_MIN_ATTR_MASK(ar)) {
   1441 						no_space = TRUE;
   1442 						continue;
   1443 					}
   1444 					MINIMIZE_ATTR_MASK(ar);
   1445 					ae = ar;
   1446 					ptr = lastentry_ptr;
   1447 					goto reencode_attrs;
   1448 				}
   1449 			}
   1450 		}
   1451 
   1452 		/* Reset to directory's vfs info when encoding complete */
   1453 		if (vfs_different) {
   1454 			dsbe = sbe;
   1455 			dpce = pce;
   1456 			vfs_different = 0;
   1457 		}
   1458 
   1459 		/* "go back" and encode the attributes' length */
   1460 		attr_length =
   1461 		    (char *)ptr -
   1462 		    (char *)attr_offset_ptr -
   1463 		    BYTES_PER_XDR_UNIT;
   1464 		IXDR_PUT_U_INT32(attr_offset_ptr, attr_length);
   1465 
   1466 		/*
   1467 		 * If there was trouble obtaining a mapping for either
   1468 		 * the owner or group attributes, then remove them from
   1469 		 * bitmap4 for this entry and reset the bitmap value
   1470 		 * in the data stream.
   1471 		 */
   1472 		if (owner_error || group_error) {
   1473 			if (owner_error)
   1474 				ae &= ~FATTR4_OWNER_MASK;
   1475 			if (group_error)
   1476 				ae &= ~FATTR4_OWNER_GROUP_MASK;
   1477 			IXDR_PUT_HYPER(attrmask_ptr, ae);
   1478 		}
   1479 
   1480 		/* END OF ATTRIBUTE ENCODING */
   1481 
   1482 		lastentry_ptr = ptr;
   1483 		nents++;
   1484 		rddir_next_offset = dp->d_off;
   1485 	}
   1486 
   1487 	/*
   1488 	 * Check for the case that another VOP_READDIR() has to be done.
   1489 	 * - no space encoding error
   1490 	 * - no entry successfully encoded
   1491 	 * - still more directory to read
   1492 	 */
   1493 	if (!no_space && nents == 0 && !iseofdir)
   1494 		goto readagain;
   1495 
   1496 	*cs->statusp = resp->status = NFS4_OK;
   1497 
   1498 	/*
   1499 	 * If no_space is set then we terminated prematurely,
   1500 	 * rewind to the last entry and this can never be EOF.
   1501 	 */
   1502 	if (no_space) {
   1503 		ptr = lastentry_ptr;
   1504 		eof = FALSE; /* ended encoded prematurely */
   1505 	} else {
   1506 		eof = (iseofdir ? TRUE : FALSE);
   1507 	}
   1508 
   1509 	/*
   1510 	 * If we have entries, always return them, otherwise only error
   1511 	 * if we ran out of space.
   1512 	 */
   1513 	if (nents || !no_space) {
   1514 		ASSERT(ptr != NULL);
   1515 		/* encode the BOOLEAN marking no further entries */
   1516 		IXDR_PUT_U_INT32(ptr, false);
   1517 		/* encode the BOOLEAN signifying end of directory */
   1518 		IXDR_PUT_U_INT32(ptr, eof);
   1519 
   1520 		resp->data_len = (char *)ptr - (char *)beginning_ptr;
   1521 		resp->mblk->b_wptr += resp->data_len;
   1522 	} else {
   1523 		freeb(mp);
   1524 		resp->mblk = NULL;
   1525 		resp->data_len = 0;
   1526 		*cs->statusp = resp->status = NFS4ERR_TOOSMALL;
   1527 	}
   1528 
   1529 	kmem_free((caddr_t)rddir_data, rddir_data_len);
   1530 	if (vp)
   1531 		VN_RELE(vp);
   1532 	if (owner.utf8string_len != 0)
   1533 		kmem_free(owner.utf8string_val,	owner.utf8string_len);
   1534 	if (group.utf8string_len != 0)
   1535 		kmem_free(group.utf8string_val, group.utf8string_len);
   1536 
   1537 out:
   1538 	DTRACE_NFSV4_2(op__readdir__done, struct compound_state *, cs,
   1539 	    READDIR4res *, resp);
   1540 }
   1541