1 /* $NetBSD: nfs_serv.c,v 1.130 2007/10/10 20:42:31 ad Exp $ */ 2 3 /* 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * This code is derived from software contributed to Berkeley by 8 * Rick Macklem at The University of Guelph. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * @(#)nfs_serv.c 8.8 (Berkeley) 7/31/95 35 */ 36 37 /* 38 * nfs version 2 and 3 server calls to vnode ops 39 * - these routines generally have 3 phases 40 * 1 - break down and validate rpc request in mbuf list 41 * 2 - do the vnode ops for the request 42 * (surprisingly ?? many are very similar to syscalls in vfs_syscalls.c) 43 * 3 - build the rpc reply in an mbuf list 44 * nb: 45 * - do not mix the phases, since the nfsm_?? macros can return failures 46 * on a bad rpc or similar and do not do any vrele() or vput()'s 47 * 48 * - the nfsm_reply() macro generates an nfs rpc reply with the nfs 49 * error number iff error != 0 whereas 50 * returning an error from the server function implies a fatal error 51 * such as a badly constructed rpc request that should be dropped without 52 * a reply. 53 * For Version 3, nfsm_reply() does not return for the error case, since 54 * most version 3 rpcs return more than the status for error cases. 55 */ 56 57 #include <sys/cdefs.h> 58 __KERNEL_RCSID(0, "$NetBSD: nfs_serv.c,v 1.130 2007/10/10 20:42:31 ad Exp $"); 59 60 #include <sys/param.h> 61 #include <sys/systm.h> 62 #include <sys/proc.h> 63 #include <sys/file.h> 64 #include <sys/namei.h> 65 #include <sys/vnode.h> 66 #include <sys/mount.h> 67 #include <sys/socket.h> 68 #include <sys/socketvar.h> 69 #include <sys/mbuf.h> 70 #include <sys/dirent.h> 71 #include <sys/stat.h> 72 #include <sys/kernel.h> 73 #include <sys/hash.h> 74 #include <sys/kauth.h> 75 76 #include <uvm/uvm.h> 77 78 #include <nfs/nfsproto.h> 79 #include <nfs/rpcv2.h> 80 #include <nfs/nfs.h> 81 #include <nfs/xdr_subs.h> 82 #include <nfs/nfsm_subs.h> 83 #include <nfs/nfs_var.h> 84 85 /* Global vars */ 86 extern u_int32_t nfs_xdrneg1; 87 extern u_int32_t nfs_false, nfs_true; 88 extern enum vtype nv3tov_type[8]; 89 extern struct nfsstats nfsstats; 90 extern const nfstype nfsv2_type[9]; 91 extern const nfstype nfsv3_type[9]; 92 int nfsrvw_procrastinate = NFS_GATHERDELAY * 1000; 93 int nfsd_use_loan = 1; /* use page-loan for READ OP */ 94 95 #define nqsrv_getl(vp, rw) /* nothing */ 96 97 /* 98 * nfs v3 access service 99 */ 100 int 101 nfsrv3_access(nfsd, slp, lwp, mrq) 102 struct nfsrv_descript *nfsd; 103 struct nfssvc_sock *slp; 104 struct lwp *lwp; 105 struct mbuf **mrq; 106 { 107 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 108 struct mbuf *nam = nfsd->nd_nam; 109 char *dpos = nfsd->nd_dpos; 110 kauth_cred_t cred = nfsd->nd_cr; 111 struct vnode *vp; 112 nfsrvfh_t nsfh; 113 u_int32_t *tl; 114 int32_t t1; 115 char *bpos; 116 int error = 0, rdonly, cache = 0, getret; 117 char *cp2; 118 struct mbuf *mb, *mreq; 119 struct vattr va; 120 u_long inmode, testmode, outmode; 121 u_quad_t frev; 122 123 nfsm_srvmtofh(&nsfh); 124 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 125 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, &rdonly, 126 (nfsd->nd_flag & ND_KERBAUTH), false); 127 if (error) { 128 nfsm_reply(NFSX_UNSIGNED); 129 nfsm_srvpostop_attr(1, (struct vattr *)0); 130 return (0); 131 } 132 inmode = fxdr_unsigned(u_int32_t, *tl); 133 outmode = 0; 134 if ((inmode & NFSV3ACCESS_READ) && 135 nfsrv_access(vp, VREAD, cred, rdonly, lwp, 0) == 0) 136 outmode |= NFSV3ACCESS_READ; 137 if (vp->v_type != VDIR) { 138 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND); 139 if (testmode && 140 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0) 141 outmode |= testmode; 142 if ((inmode & NFSV3ACCESS_EXECUTE) && 143 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0) 144 outmode |= NFSV3ACCESS_EXECUTE; 145 } else { 146 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND | 147 NFSV3ACCESS_DELETE); 148 if (testmode && 149 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0) 150 outmode |= testmode; 151 if ((inmode & NFSV3ACCESS_LOOKUP) && 152 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0) 153 outmode |= NFSV3ACCESS_LOOKUP; 154 } 155 getret = VOP_GETATTR(vp, &va, cred, lwp); 156 vput(vp); 157 nfsm_reply(NFSX_POSTOPATTR(1) + NFSX_UNSIGNED); 158 nfsm_srvpostop_attr(getret, &va); 159 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 160 *tl = txdr_unsigned(outmode); 161 nfsm_srvdone; 162 } 163 164 /* 165 * nfs getattr service 166 */ 167 int 168 nfsrv_getattr(nfsd, slp, lwp, mrq) 169 struct nfsrv_descript *nfsd; 170 struct nfssvc_sock *slp; 171 struct lwp *lwp; 172 struct mbuf **mrq; 173 { 174 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 175 struct mbuf *nam = nfsd->nd_nam; 176 char *dpos = nfsd->nd_dpos; 177 kauth_cred_t cred = nfsd->nd_cr; 178 struct nfs_fattr *fp; 179 struct vattr va; 180 struct vnode *vp; 181 nfsrvfh_t nsfh; 182 u_int32_t *tl; 183 int32_t t1; 184 char *bpos; 185 int error = 0, rdonly, cache = 0; 186 char *cp2; 187 struct mbuf *mb, *mreq; 188 u_quad_t frev; 189 190 nfsm_srvmtofh(&nsfh); 191 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, &rdonly, 192 (nfsd->nd_flag & ND_KERBAUTH), false); 193 if (error) { 194 nfsm_reply(0); 195 return (0); 196 } 197 nqsrv_getl(vp, ND_READ); 198 error = VOP_GETATTR(vp, &va, cred, lwp); 199 vput(vp); 200 nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); 201 if (error) 202 return (0); 203 nfsm_build(fp, struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); 204 nfsm_srvfillattr(&va, fp); 205 nfsm_srvdone; 206 } 207 208 /* 209 * nfs setattr service 210 */ 211 int 212 nfsrv_setattr(nfsd, slp, lwp, mrq) 213 struct nfsrv_descript *nfsd; 214 struct nfssvc_sock *slp; 215 struct lwp *lwp; 216 struct mbuf **mrq; 217 { 218 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 219 struct mbuf *nam = nfsd->nd_nam; 220 char *dpos = nfsd->nd_dpos; 221 kauth_cred_t cred = nfsd->nd_cr; 222 struct vattr va, preat; 223 struct nfsv2_sattr *sp; 224 struct nfs_fattr *fp; 225 struct vnode *vp; 226 nfsrvfh_t nsfh; 227 u_int32_t *tl; 228 int32_t t1; 229 char *bpos; 230 int error = 0, rdonly, cache = 0, preat_ret = 1, postat_ret = 1; 231 int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0; 232 char *cp2; 233 struct mbuf *mb, *mreq; 234 u_quad_t frev; 235 struct timespec guard; 236 237 memset(&guard, 0, sizeof guard); /* XXX gcc */ 238 239 nfsm_srvmtofh(&nsfh); 240 VATTR_NULL(&va); 241 if (v3) { 242 nfsm_srvsattr(&va); 243 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 244 gcheck = fxdr_unsigned(int, *tl); 245 if (gcheck) { 246 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 247 fxdr_nfsv3time(tl, &guard); 248 } 249 } else { 250 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 251 /* 252 * Nah nah nah nah na nah 253 * There is a bug in the Sun client that puts 0xffff in the mode 254 * field of sattr when it should put in 0xffffffff. The u_short 255 * doesn't sign extend. 256 * --> check the low order 2 bytes for 0xffff 257 */ 258 if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff) 259 va.va_mode = nfstov_mode(sp->sa_mode); 260 if (sp->sa_uid != nfs_xdrneg1) 261 va.va_uid = fxdr_unsigned(uid_t, sp->sa_uid); 262 if (sp->sa_gid != nfs_xdrneg1) 263 va.va_gid = fxdr_unsigned(gid_t, sp->sa_gid); 264 if (sp->sa_size != nfs_xdrneg1) 265 va.va_size = fxdr_unsigned(u_quad_t, sp->sa_size); 266 if (sp->sa_atime.nfsv2_sec != nfs_xdrneg1) { 267 #ifdef notyet 268 fxdr_nfsv2time(&sp->sa_atime, &va.va_atime); 269 #else 270 va.va_atime.tv_sec = 271 fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec); 272 va.va_atime.tv_nsec = 0; 273 #endif 274 } 275 if (sp->sa_mtime.nfsv2_sec != nfs_xdrneg1) 276 fxdr_nfsv2time(&sp->sa_mtime, &va.va_mtime); 277 278 } 279 280 /* 281 * Now that we have all the fields, lets do it. 282 */ 283 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, &rdonly, 284 (nfsd->nd_flag & ND_KERBAUTH), false); 285 if (error) { 286 nfsm_reply(2 * NFSX_UNSIGNED); 287 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 288 return (0); 289 } 290 nqsrv_getl(vp, ND_WRITE); 291 if (v3) { 292 error = preat_ret = VOP_GETATTR(vp, &preat, cred, lwp); 293 if (!error && gcheck && 294 (preat.va_ctime.tv_sec != guard.tv_sec || 295 preat.va_ctime.tv_nsec != guard.tv_nsec)) 296 error = NFSERR_NOT_SYNC; 297 if (error) { 298 vput(vp); 299 nfsm_reply(NFSX_WCCDATA(v3)); 300 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 301 return (0); 302 } 303 } 304 305 /* 306 * If the size is being changed write acces is required, otherwise 307 * just check for a read only file system. 308 */ 309 if (va.va_size == ((u_quad_t)((quad_t) -1))) { 310 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 311 error = EROFS; 312 goto out; 313 } 314 } else { 315 if (vp->v_type == VDIR) { 316 error = EISDIR; 317 goto out; 318 } else if ((error = nfsrv_access(vp, VWRITE, cred, rdonly, 319 lwp, 0)) != 0) 320 goto out; 321 } 322 error = VOP_SETATTR(vp, &va, cred, lwp); 323 postat_ret = VOP_GETATTR(vp, &va, cred, lwp); 324 if (!error) 325 error = postat_ret; 326 out: 327 vput(vp); 328 nfsm_reply(NFSX_WCCORFATTR(v3)); 329 if (v3) { 330 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 331 return (0); 332 } else { 333 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 334 nfsm_srvfillattr(&va, fp); 335 } 336 nfsm_srvdone; 337 } 338 339 /* 340 * nfs lookup rpc 341 */ 342 int 343 nfsrv_lookup(nfsd, slp, lwp, mrq) 344 struct nfsrv_descript *nfsd; 345 struct nfssvc_sock *slp; 346 struct lwp *lwp; 347 struct mbuf **mrq; 348 { 349 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 350 struct mbuf *nam = nfsd->nd_nam; 351 char *dpos = nfsd->nd_dpos; 352 kauth_cred_t cred = nfsd->nd_cr; 353 struct nfs_fattr *fp; 354 struct nameidata nd, ind, *ndp = &nd; 355 struct vnode *vp, *dirp; 356 nfsrvfh_t nsfh; 357 char *cp; 358 u_int32_t *tl; 359 int32_t t1; 360 char *bpos; 361 int error = 0, cache = 0, dirattr_ret = 1; 362 uint32_t len; 363 int v3 = (nfsd->nd_flag & ND_NFSV3), pubflag; 364 char *cp2; 365 struct mbuf *mb, *mreq; 366 struct vattr va, dirattr; 367 u_quad_t frev; 368 369 nfsm_srvmtofh(&nsfh); 370 nfsm_srvnamesiz(len); 371 372 pubflag = nfs_ispublicfh(&nsfh); 373 374 nd.ni_cnd.cn_cred = cred; 375 nd.ni_cnd.cn_nameiop = LOOKUP; 376 nd.ni_cnd.cn_flags = LOCKLEAF | SAVESTART; 377 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 378 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), pubflag); 379 380 if (!error && pubflag) { 381 if (nd.ni_vp->v_type == VDIR && nfs_pub.np_index != NULL) { 382 /* 383 * Setup call to lookup() to see if we can find 384 * the index file. Arguably, this doesn't belong 385 * in a kernel.. Ugh. 386 */ 387 ind = nd; 388 VOP_UNLOCK(nd.ni_vp, 0); 389 ind.ni_pathlen = strlen(nfs_pub.np_index); 390 ind.ni_cnd.cn_nameptr = ind.ni_cnd.cn_pnbuf = 391 nfs_pub.np_index; 392 ind.ni_startdir = nd.ni_vp; 393 VREF(ind.ni_startdir); 394 error = lookup(&ind); 395 if (!error) { 396 /* 397 * Found an index file. Get rid of 398 * the old references. 399 */ 400 if (dirp) 401 vrele(dirp); 402 dirp = nd.ni_vp; 403 vrele(nd.ni_startdir); 404 ndp = &ind; 405 } else 406 error = 0; 407 } 408 /* 409 * If the public filehandle was used, check that this lookup 410 * didn't result in a filehandle outside the publicly exported 411 * filesystem. 412 */ 413 414 if (!error && ndp->ni_vp->v_mount != nfs_pub.np_mount) { 415 vput(nd.ni_vp); 416 error = EPERM; 417 } 418 } 419 420 if (dirp) { 421 if (v3) 422 dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred, lwp); 423 vrele(dirp); 424 } 425 426 if (error) { 427 nfsm_reply(NFSX_POSTOPATTR(v3)); 428 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 429 return (0); 430 } 431 432 nqsrv_getl(ndp->ni_startdir, ND_READ); 433 PNBUF_PUT(nd.ni_cnd.cn_pnbuf); 434 vp = ndp->ni_vp; 435 error = nfsrv_composefh(vp, &nsfh, v3); 436 if (!error) 437 error = VOP_GETATTR(vp, &va, cred, lwp); 438 vput(vp); 439 vrele(ndp->ni_startdir); 440 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_POSTOPORFATTR(v3) + 441 NFSX_POSTOPATTR(v3)); 442 if (error) { 443 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 444 return (0); 445 } 446 nfsm_srvfhtom(&nsfh, v3); 447 if (v3) { 448 nfsm_srvpostop_attr(0, &va); 449 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 450 } else { 451 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 452 nfsm_srvfillattr(&va, fp); 453 } 454 nfsm_srvdone; 455 } 456 457 /* 458 * nfs readlink service 459 */ 460 int 461 nfsrv_readlink(nfsd, slp, lwp, mrq) 462 struct nfsrv_descript *nfsd; 463 struct nfssvc_sock *slp; 464 struct lwp *lwp; 465 struct mbuf **mrq; 466 { 467 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 468 struct mbuf *nam = nfsd->nd_nam; 469 char *dpos = nfsd->nd_dpos; 470 kauth_cred_t cred = nfsd->nd_cr; 471 struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN]; 472 struct iovec *ivp = iv; 473 struct mbuf *mp; 474 u_int32_t *tl; 475 int32_t t1; 476 char *bpos; 477 int error = 0, rdonly, cache = 0, i, padlen, getret; 478 uint32_t len; 479 int v3 = (nfsd->nd_flag & ND_NFSV3); 480 char *cp2; 481 struct mbuf *mb, *mp2 = NULL, *mp3 = NULL, *mreq; 482 struct vnode *vp; 483 struct vattr attr; 484 nfsrvfh_t nsfh; 485 struct uio io, *uiop = &io; 486 u_quad_t frev; 487 488 nfsm_srvmtofh(&nsfh); 489 len = 0; 490 i = 0; 491 while (len < NFS_MAXPATHLEN) { 492 mp = m_get(M_WAIT, MT_DATA); 493 MCLAIM(mp, &nfs_mowner); 494 m_clget(mp, M_WAIT); 495 mp->m_len = NFSMSIZ(mp); 496 if (len == 0) 497 mp3 = mp2 = mp; 498 else { 499 mp2->m_next = mp; 500 mp2 = mp; 501 } 502 if ((len+mp->m_len) > NFS_MAXPATHLEN) { 503 mp->m_len = NFS_MAXPATHLEN-len; 504 len = NFS_MAXPATHLEN; 505 } else 506 len += mp->m_len; 507 ivp->iov_base = mtod(mp, void *); 508 ivp->iov_len = mp->m_len; 509 i++; 510 ivp++; 511 } 512 uiop->uio_iov = iv; 513 uiop->uio_iovcnt = i; 514 uiop->uio_offset = 0; 515 uiop->uio_resid = len; 516 uiop->uio_rw = UIO_READ; 517 UIO_SETUP_SYSSPACE(uiop); 518 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 519 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 520 if (error) { 521 m_freem(mp3); 522 nfsm_reply(2 * NFSX_UNSIGNED); 523 nfsm_srvpostop_attr(1, (struct vattr *)0); 524 return (0); 525 } 526 if (vp->v_type != VLNK) { 527 if (v3) 528 error = EINVAL; 529 else 530 error = ENXIO; 531 goto out; 532 } 533 nqsrv_getl(vp, ND_READ); 534 error = VOP_READLINK(vp, uiop, cred); 535 out: 536 getret = VOP_GETATTR(vp, &attr, cred, lwp); 537 vput(vp); 538 if (error) 539 m_freem(mp3); 540 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED); 541 if (v3) { 542 nfsm_srvpostop_attr(getret, &attr); 543 if (error) 544 return (0); 545 } 546 len -= uiop->uio_resid; 547 padlen = nfsm_padlen(len); 548 if (uiop->uio_resid || padlen) 549 nfs_zeropad(mp3, uiop->uio_resid, padlen); 550 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 551 *tl = txdr_unsigned(len); 552 mb->m_next = mp3; 553 nfsm_srvdone; 554 } 555 556 /* 557 * nfs read service 558 */ 559 int 560 nfsrv_read(nfsd, slp, lwp, mrq) 561 struct nfsrv_descript *nfsd; 562 struct nfssvc_sock *slp; 563 struct lwp *lwp; 564 struct mbuf **mrq; 565 { 566 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 567 struct mbuf *nam = nfsd->nd_nam; 568 char *dpos = nfsd->nd_dpos; 569 kauth_cred_t cred = nfsd->nd_cr; 570 struct mbuf *m; 571 struct nfs_fattr *fp; 572 u_int32_t *tl; 573 int32_t t1; 574 int i; 575 char *bpos; 576 int error = 0, rdonly, cache = 0, getret; 577 int v3 = (nfsd->nd_flag & ND_NFSV3); 578 uint32_t reqlen, len, cnt, left; 579 int padlen; 580 char *cp2; 581 struct mbuf *mb, *mreq; 582 struct vnode *vp; 583 nfsrvfh_t nsfh; 584 struct uio io, *uiop = &io; 585 struct vattr va; 586 off_t off; 587 u_quad_t frev; 588 589 nfsm_srvmtofh(&nsfh); 590 if (v3) { 591 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 592 off = fxdr_hyper(tl); 593 } else { 594 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 595 off = (off_t)fxdr_unsigned(u_int32_t, *tl); 596 } 597 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 598 reqlen = fxdr_unsigned(uint32_t, *tl); 599 reqlen = MIN(reqlen, NFS_SRVMAXDATA(nfsd)); 600 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 601 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 602 if (error) { 603 nfsm_reply(2 * NFSX_UNSIGNED); 604 nfsm_srvpostop_attr(1, (struct vattr *)0); 605 return (0); 606 } 607 if (vp->v_type != VREG) { 608 if (v3) 609 error = EINVAL; 610 else 611 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 612 } 613 if (!error) { 614 nqsrv_getl(vp, ND_READ); 615 if ((error = nfsrv_access(vp, VREAD, cred, rdonly, lwp, 1)) != 0) 616 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 1); 617 } 618 getret = VOP_GETATTR(vp, &va, cred, lwp); 619 if (!error) 620 error = getret; 621 if (error) { 622 vput(vp); 623 nfsm_reply(NFSX_POSTOPATTR(v3)); 624 nfsm_srvpostop_attr(getret, &va); 625 return (0); 626 } 627 if (off >= va.va_size) 628 cnt = 0; 629 else if ((off + reqlen) > va.va_size) 630 cnt = va.va_size - off; 631 else 632 cnt = reqlen; 633 nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt)); 634 if (v3) { 635 nfsm_build(tl, u_int32_t *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED); 636 *tl++ = nfs_true; 637 fp = (struct nfs_fattr *)tl; 638 tl += (NFSX_V3FATTR / sizeof (u_int32_t)); 639 } else { 640 nfsm_build(tl, u_int32_t *, NFSX_V2FATTR + NFSX_UNSIGNED); 641 fp = (struct nfs_fattr *)tl; 642 tl += (NFSX_V2FATTR / sizeof (u_int32_t)); 643 } 644 len = left = cnt; 645 if (cnt > 0) { 646 if (nfsd_use_loan) { 647 struct vm_page **pgpp; 648 voff_t pgoff = trunc_page(off); 649 int npages; 650 vaddr_t lva; 651 652 npages = (round_page(off + cnt) - pgoff) >> PAGE_SHIFT; 653 KASSERT(npages <= M_EXT_MAXPAGES); /* XXX */ 654 655 /* allocate kva for mbuf data */ 656 lva = sokvaalloc(npages << PAGE_SHIFT, slp->ns_so); 657 if (lva == 0) { 658 /* fall back to VOP_READ */ 659 goto loan_fail; 660 } 661 662 /* allocate mbuf */ 663 m = m_get(M_WAIT, MT_DATA); 664 MCLAIM(m, &nfs_mowner); 665 pgpp = m->m_ext.ext_pgs; 666 667 /* loan pages */ 668 error = uvm_loanuobjpages(&vp->v_uobj, pgoff, npages, 669 pgpp); 670 if (error) { 671 sokvafree(lva, npages << PAGE_SHIFT); 672 m_free(m); 673 if (error == EBUSY) 674 goto loan_fail; 675 goto read_error; 676 } 677 678 /* associate kva to mbuf */ 679 MEXTADD(m, (void *)(lva + ((vaddr_t)off & PAGE_MASK)), 680 cnt, M_MBUF, soloanfree, slp->ns_so); 681 m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP; 682 m->m_len = cnt; 683 684 /* map pages */ 685 for (i = 0; i < npages; i++) { 686 pmap_kenter_pa(lva, VM_PAGE_TO_PHYS(pgpp[i]), 687 VM_PROT_READ); 688 lva += PAGE_SIZE; 689 } 690 691 pmap_update(pmap_kernel()); 692 693 mb->m_next = m; 694 mb = m; 695 error = 0; 696 uiop->uio_resid = 0; 697 } else { 698 struct iovec *iv; 699 struct iovec *iv2; 700 struct mbuf *m2; 701 int siz; 702 loan_fail: 703 /* 704 * Generate the mbuf list with the uio_iov ref. to it. 705 */ 706 i = 0; 707 m = m2 = mb; 708 while (left > 0) { 709 siz = min(M_TRAILINGSPACE(m), left); 710 if (siz > 0) { 711 left -= siz; 712 i++; 713 } 714 if (left > 0) { 715 m = m_get(M_WAIT, MT_DATA); 716 MCLAIM(m, &nfs_mowner); 717 m_clget(m, M_WAIT); 718 m->m_len = 0; 719 m2->m_next = m; 720 m2 = m; 721 } 722 } 723 iv = malloc(i * sizeof(struct iovec), M_TEMP, M_WAITOK); 724 uiop->uio_iov = iv2 = iv; 725 m = mb; 726 left = cnt; 727 i = 0; 728 while (left > 0) { 729 if (m == NULL) 730 panic("nfsrv_read iov"); 731 siz = min(M_TRAILINGSPACE(m), left); 732 if (siz > 0) { 733 iv->iov_base = mtod(m, char *) + 734 m->m_len; 735 iv->iov_len = siz; 736 m->m_len += siz; 737 left -= siz; 738 iv++; 739 i++; 740 } 741 m = m->m_next; 742 } 743 uiop->uio_iovcnt = i; 744 uiop->uio_offset = off; 745 uiop->uio_resid = cnt; 746 uiop->uio_rw = UIO_READ; 747 UIO_SETUP_SYSSPACE(uiop); 748 error = VOP_READ(vp, uiop, IO_NODELOCKED, cred); 749 free((void *)iv2, M_TEMP); 750 } 751 read_error: 752 if (error || (getret = VOP_GETATTR(vp, &va, cred, lwp)) != 0){ 753 if (!error) 754 error = getret; 755 m_freem(mreq); 756 vput(vp); 757 nfsm_reply(NFSX_POSTOPATTR(v3)); 758 nfsm_srvpostop_attr(getret, &va); 759 return (0); 760 } 761 } else { 762 uiop->uio_resid = 0; 763 } 764 vput(vp); 765 nfsm_srvfillattr(&va, fp); 766 len -= uiop->uio_resid; 767 padlen = nfsm_padlen(len); 768 if (uiop->uio_resid || padlen) 769 nfs_zeropad(mb, uiop->uio_resid, padlen); 770 if (v3) { 771 /* count */ 772 *tl++ = txdr_unsigned(len); 773 /* eof */ 774 if (off + len >= va.va_size) 775 *tl++ = nfs_true; 776 else 777 *tl++ = nfs_false; 778 } 779 *tl = txdr_unsigned(len); 780 nfsm_srvdone; 781 } 782 783 /* 784 * nfs write service 785 */ 786 int 787 nfsrv_write(nfsd, slp, lwp, mrq) 788 struct nfsrv_descript *nfsd; 789 struct nfssvc_sock *slp; 790 struct lwp *lwp; 791 struct mbuf **mrq; 792 { 793 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 794 struct mbuf *nam = nfsd->nd_nam; 795 char *dpos = nfsd->nd_dpos; 796 kauth_cred_t cred = nfsd->nd_cr; 797 struct iovec *ivp; 798 int i, cnt; 799 struct mbuf *mp; 800 struct nfs_fattr *fp; 801 struct iovec *iv; 802 struct vattr va, forat; 803 u_int32_t *tl; 804 int32_t t1; 805 char *bpos; 806 int error = 0, rdonly, cache = 0, len, forat_ret = 1; 807 int ioflags, aftat_ret = 1, retlen, zeroing, adjust; 808 int stable = NFSV3WRITE_FILESYNC; 809 int v3 = (nfsd->nd_flag & ND_NFSV3); 810 char *cp2; 811 struct mbuf *mb, *mreq; 812 struct vnode *vp; 813 nfsrvfh_t nsfh; 814 struct uio io, *uiop = &io; 815 off_t off; 816 u_quad_t frev; 817 818 if (mrep == NULL) { 819 *mrq = NULL; 820 return (0); 821 } 822 nfsm_srvmtofh(&nsfh); 823 if (v3) { 824 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 825 off = fxdr_hyper(tl); 826 tl += 3; 827 stable = fxdr_unsigned(int, *tl++); 828 } else { 829 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 830 off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 831 tl += 2; 832 } 833 retlen = len = fxdr_unsigned(int32_t, *tl); 834 cnt = i = 0; 835 836 /* 837 * For NFS Version 2, it is not obvious what a write of zero length 838 * should do, but I might as well be consistent with Version 3, 839 * which is to return ok so long as there are no permission problems. 840 */ 841 if (len > 0) { 842 zeroing = 1; 843 mp = mrep; 844 while (mp) { 845 if (mp == md) { 846 zeroing = 0; 847 adjust = dpos - mtod(mp, char *); 848 mp->m_len -= adjust; 849 if (mp->m_len > 0 && adjust > 0) 850 NFSMADV(mp, adjust); 851 } 852 if (zeroing) 853 mp->m_len = 0; 854 else if (mp->m_len > 0) { 855 i += mp->m_len; 856 if (i > len) { 857 mp->m_len -= (i - len); 858 zeroing = 1; 859 } 860 if (mp->m_len > 0) 861 cnt++; 862 } 863 mp = mp->m_next; 864 } 865 } 866 if (len > NFS_MAXDATA || len < 0 || i < len) { 867 error = EIO; 868 nfsm_reply(2 * NFSX_UNSIGNED); 869 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 870 return (0); 871 } 872 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 873 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 874 if (error) { 875 nfsm_reply(2 * NFSX_UNSIGNED); 876 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 877 return (0); 878 } 879 if (v3) 880 forat_ret = VOP_GETATTR(vp, &forat, cred, lwp); 881 if (vp->v_type != VREG) { 882 if (v3) 883 error = EINVAL; 884 else 885 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 886 } 887 if (!error) { 888 nqsrv_getl(vp, ND_WRITE); 889 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1); 890 } 891 if (error) { 892 vput(vp); 893 nfsm_reply(NFSX_WCCDATA(v3)); 894 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 895 return (0); 896 } 897 898 if (len > 0) { 899 ivp = malloc(cnt * sizeof (struct iovec), M_TEMP, M_WAITOK); 900 uiop->uio_iov = iv = ivp; 901 uiop->uio_iovcnt = cnt; 902 mp = mrep; 903 while (mp) { 904 if (mp->m_len > 0) { 905 ivp->iov_base = mtod(mp, void *); 906 ivp->iov_len = mp->m_len; 907 ivp++; 908 } 909 mp = mp->m_next; 910 } 911 912 /* 913 * XXX 914 * The IO_METASYNC flag indicates that all metadata (and not 915 * just enough to ensure data integrity) must be written to 916 * stable storage synchronously. 917 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.) 918 */ 919 if (stable == NFSV3WRITE_UNSTABLE) 920 ioflags = IO_NODELOCKED; 921 else if (stable == NFSV3WRITE_DATASYNC) 922 ioflags = (IO_SYNC | IO_NODELOCKED); 923 else 924 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 925 uiop->uio_resid = len; 926 uiop->uio_rw = UIO_WRITE; 927 uiop->uio_offset = off; 928 UIO_SETUP_SYSSPACE(uiop); 929 error = VOP_WRITE(vp, uiop, ioflags, cred); 930 nfsstats.srvvop_writes++; 931 free(iv, M_TEMP); 932 } 933 aftat_ret = VOP_GETATTR(vp, &va, cred, lwp); 934 vput(vp); 935 if (!error) 936 error = aftat_ret; 937 nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) + 938 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3)); 939 if (v3) { 940 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 941 if (error) 942 return (0); 943 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 944 *tl++ = txdr_unsigned(retlen); 945 if (stable == NFSV3WRITE_UNSTABLE) 946 *tl++ = txdr_unsigned(stable); 947 else 948 *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC); 949 /* 950 * Actually, there is no need to txdr these fields, 951 * but it may make the values more human readable, 952 * for debugging purposes. 953 */ 954 *tl++ = txdr_unsigned(boottime.tv_sec); 955 *tl = txdr_unsigned(boottime.tv_usec); 956 } else { 957 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 958 nfsm_srvfillattr(&va, fp); 959 } 960 nfsm_srvdone; 961 } 962 963 /* 964 * XXX elad: the original NFSW_SAMECRED() macro also made sure the 965 * two nd_flag fields of the descriptors contained 966 * ND_KERBAUTH. 967 */ 968 static int 969 nfsrv_samecred(kauth_cred_t cred1, kauth_cred_t cred2) 970 { 971 int i, do_ngroups; 972 973 if (kauth_cred_geteuid(cred1) != kauth_cred_geteuid(cred2)) 974 return (0); 975 if (kauth_cred_ngroups(cred1) != kauth_cred_ngroups(cred2)) 976 return (0); 977 do_ngroups = kauth_cred_ngroups(cred1); 978 for (i = 0; i < do_ngroups; i++) 979 if (kauth_cred_group(cred1, i) != 980 kauth_cred_group(cred2, i)) 981 return (0); 982 983 return (1); 984 } 985 986 static struct nfsrvw_delayhash * 987 nfsrv_nwdelayhash(struct nfssvc_sock *slp, const nfsrvfh_t *nsfh) 988 { 989 uint32_t hash; 990 991 hash = hash32_buf(NFSRVFH_DATA(nsfh), NFSRVFH_SIZE(nsfh), 992 HASH32_BUF_INIT); 993 return &slp->ns_wdelayhashtbl[hash % NFS_WDELAYHASHSIZ]; 994 } 995 996 /* 997 * NFS write service with write gathering support. Called when 998 * nfsrvw_procrastinate > 0. 999 * See: Chet Juszczak, "Improving the Write Performance of an NFS Server", 1000 * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco, 1001 * Jan. 1994. 1002 */ 1003 int 1004 nfsrv_writegather(ndp, slp, lwp, mrq) 1005 struct nfsrv_descript **ndp; 1006 struct nfssvc_sock *slp; 1007 struct lwp *lwp; 1008 struct mbuf **mrq; 1009 { 1010 struct timeval now; 1011 struct iovec *ivp; 1012 struct mbuf *mp; 1013 struct nfsrv_descript *wp, *nfsd, *owp, *swp; 1014 struct nfs_fattr *fp; 1015 int i = 0; 1016 struct iovec *iov; 1017 struct nfsrvw_delayhash *wpp; 1018 kauth_cred_t cred; 1019 struct vattr va, forat; 1020 u_int32_t *tl; 1021 int32_t t1; 1022 char *bpos, *dpos; 1023 int error = 0, rdonly, cache = 0, len = 0, forat_ret = 1; 1024 int ioflags, aftat_ret = 1, s, adjust, v3, zeroing; 1025 char *cp2; 1026 struct mbuf *mb, *mreq, *mrep, *md; 1027 struct vnode *vp; 1028 struct uio io, *uiop = &io; 1029 u_quad_t frev, cur_usec; 1030 1031 *mrq = NULL; 1032 if (*ndp) { 1033 nfsd = *ndp; 1034 *ndp = NULL; 1035 mrep = nfsd->nd_mrep; 1036 md = nfsd->nd_md; 1037 dpos = nfsd->nd_dpos; 1038 cred = nfsd->nd_cr; 1039 v3 = (nfsd->nd_flag & ND_NFSV3); 1040 LIST_INIT(&nfsd->nd_coalesce); 1041 nfsd->nd_mreq = NULL; 1042 nfsd->nd_stable = NFSV3WRITE_FILESYNC; 1043 getmicrotime(&now); 1044 cur_usec = (u_quad_t)now.tv_sec * 1000000 + (u_quad_t)now.tv_usec; 1045 nfsd->nd_time = cur_usec + nfsrvw_procrastinate; 1046 1047 /* 1048 * Now, get the write header.. 1049 */ 1050 nfsm_srvmtofh(&nfsd->nd_fh); 1051 if (v3) { 1052 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 1053 nfsd->nd_off = fxdr_hyper(tl); 1054 tl += 3; 1055 nfsd->nd_stable = fxdr_unsigned(int, *tl++); 1056 } else { 1057 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1058 nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 1059 tl += 2; 1060 } 1061 len = fxdr_unsigned(int32_t, *tl); 1062 nfsd->nd_len = len; 1063 nfsd->nd_eoff = nfsd->nd_off + len; 1064 1065 /* 1066 * Trim the header out of the mbuf list and trim off any trailing 1067 * junk so that the mbuf list has only the write data. 1068 */ 1069 zeroing = 1; 1070 i = 0; 1071 mp = mrep; 1072 while (mp) { 1073 if (mp == md) { 1074 zeroing = 0; 1075 adjust = dpos - mtod(mp, char *); 1076 mp->m_len -= adjust; 1077 if (mp->m_len > 0 && adjust > 0) 1078 NFSMADV(mp, adjust); 1079 } 1080 if (zeroing) 1081 mp->m_len = 0; 1082 else { 1083 i += mp->m_len; 1084 if (i > len) { 1085 mp->m_len -= (i - len); 1086 zeroing = 1; 1087 } 1088 } 1089 mp = mp->m_next; 1090 } 1091 if (len > NFS_MAXDATA || len < 0 || i < len) { 1092 nfsmout: 1093 m_freem(mrep); 1094 error = EIO; 1095 nfsm_writereply(2 * NFSX_UNSIGNED, v3); 1096 if (v3) 1097 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1098 nfsd->nd_mreq = mreq; 1099 nfsd->nd_mrep = NULL; 1100 nfsd->nd_time = 0; 1101 } 1102 1103 /* 1104 * Add this entry to the hash and time queues. 1105 */ 1106 s = splsoftclock(); 1107 owp = NULL; 1108 wp = LIST_FIRST(&slp->ns_tq); 1109 while (wp && wp->nd_time < nfsd->nd_time) { 1110 owp = wp; 1111 wp = LIST_NEXT(wp, nd_tq); 1112 } 1113 if (owp) { 1114 LIST_INSERT_AFTER(owp, nfsd, nd_tq); 1115 } else { 1116 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1117 } 1118 if (nfsd->nd_mrep) { 1119 wpp = nfsrv_nwdelayhash(slp, &nfsd->nd_fh); 1120 owp = NULL; 1121 wp = LIST_FIRST(wpp); 1122 while (wp && nfsrv_comparefh(&nfsd->nd_fh, &wp->nd_fh)) { 1123 owp = wp; 1124 wp = LIST_NEXT(wp, nd_hash); 1125 } 1126 while (wp && wp->nd_off < nfsd->nd_off && 1127 !nfsrv_comparefh(&nfsd->nd_fh, &wp->nd_fh)) { 1128 owp = wp; 1129 wp = LIST_NEXT(wp, nd_hash); 1130 } 1131 if (owp) { 1132 LIST_INSERT_AFTER(owp, nfsd, nd_hash); 1133 1134 /* 1135 * Search the hash list for overlapping entries and 1136 * coalesce. 1137 */ 1138 for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) { 1139 wp = LIST_NEXT(nfsd, nd_hash); 1140 if (nfsrv_samecred(owp->nd_cr, nfsd->nd_cr)) 1141 nfsrvw_coalesce(owp, nfsd); 1142 } 1143 } else { 1144 LIST_INSERT_HEAD(wpp, nfsd, nd_hash); 1145 } 1146 } 1147 splx(s); 1148 } 1149 1150 /* 1151 * Now, do VOP_WRITE()s for any one(s) that need to be done now 1152 * and generate the associated reply mbuf list(s). 1153 */ 1154 loop1: 1155 getmicrotime(&now); 1156 cur_usec = (u_quad_t)now.tv_sec * 1000000 + (u_quad_t)now.tv_usec; 1157 s = splsoftclock(); 1158 for (nfsd = LIST_FIRST(&slp->ns_tq); nfsd; nfsd = owp) { 1159 owp = LIST_NEXT(nfsd, nd_tq); 1160 if (nfsd->nd_time > cur_usec) 1161 break; 1162 if (nfsd->nd_mreq) 1163 continue; 1164 LIST_REMOVE(nfsd, nd_tq); 1165 LIST_REMOVE(nfsd, nd_hash); 1166 splx(s); 1167 mrep = nfsd->nd_mrep; 1168 nfsd->nd_mrep = NULL; 1169 cred = nfsd->nd_cr; 1170 v3 = (nfsd->nd_flag & ND_NFSV3); 1171 forat_ret = aftat_ret = 1; 1172 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp, 1173 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH), 1174 false); 1175 if (!error) { 1176 if (v3) 1177 forat_ret = VOP_GETATTR(vp, &forat, cred, lwp); 1178 if (vp->v_type != VREG) { 1179 if (v3) 1180 error = EINVAL; 1181 else 1182 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 1183 } 1184 } else 1185 vp = NULL; 1186 if (!error) { 1187 nqsrv_getl(vp, ND_WRITE); 1188 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1); 1189 } 1190 1191 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE) 1192 ioflags = IO_NODELOCKED; 1193 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC) 1194 ioflags = (IO_SYNC | IO_NODELOCKED); 1195 else 1196 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 1197 uiop->uio_rw = UIO_WRITE; 1198 uiop->uio_offset = nfsd->nd_off; 1199 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off; 1200 UIO_SETUP_SYSSPACE(uiop); 1201 if (uiop->uio_resid > 0) { 1202 mp = mrep; 1203 i = 0; 1204 while (mp) { 1205 if (mp->m_len > 0) 1206 i++; 1207 mp = mp->m_next; 1208 } 1209 uiop->uio_iovcnt = i; 1210 iov = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK); 1211 uiop->uio_iov = ivp = iov; 1212 mp = mrep; 1213 while (mp) { 1214 if (mp->m_len > 0) { 1215 ivp->iov_base = mtod(mp, void *); 1216 ivp->iov_len = mp->m_len; 1217 ivp++; 1218 } 1219 mp = mp->m_next; 1220 } 1221 if (!error) { 1222 error = VOP_WRITE(vp, uiop, ioflags, cred); 1223 nfsstats.srvvop_writes++; 1224 } 1225 free((void *)iov, M_TEMP); 1226 } 1227 m_freem(mrep); 1228 if (vp) { 1229 aftat_ret = VOP_GETATTR(vp, &va, cred, lwp); 1230 vput(vp); 1231 } 1232 1233 /* 1234 * Loop around generating replies for all write rpcs that have 1235 * now been completed. 1236 */ 1237 swp = nfsd; 1238 do { 1239 if (error) { 1240 nfsm_writereply(NFSX_WCCDATA(v3), v3); 1241 if (v3) { 1242 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1243 } 1244 } else { 1245 nfsm_writereply(NFSX_PREOPATTR(v3) + 1246 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED + 1247 NFSX_WRITEVERF(v3), v3); 1248 if (v3) { 1249 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1250 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1251 *tl++ = txdr_unsigned(nfsd->nd_len); 1252 *tl++ = txdr_unsigned(swp->nd_stable); 1253 /* 1254 * Actually, there is no need to txdr these fields, 1255 * but it may make the values more human readable, 1256 * for debugging purposes. 1257 */ 1258 *tl++ = txdr_unsigned(boottime.tv_sec); 1259 *tl = txdr_unsigned(boottime.tv_usec); 1260 } else { 1261 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1262 nfsm_srvfillattr(&va, fp); 1263 } 1264 } 1265 nfsd->nd_mreq = mreq; 1266 if (nfsd->nd_mrep) 1267 panic("nfsrv_write: nd_mrep not free"); 1268 1269 /* 1270 * Done. Put it at the head of the timer queue so that 1271 * the final phase can return the reply. 1272 */ 1273 s = splsoftclock(); 1274 if (nfsd != swp) { 1275 nfsd->nd_time = 0; 1276 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1277 } 1278 nfsd = LIST_FIRST(&swp->nd_coalesce); 1279 if (nfsd) { 1280 LIST_REMOVE(nfsd, nd_tq); 1281 } 1282 splx(s); 1283 } while (nfsd); 1284 s = splsoftclock(); 1285 swp->nd_time = 0; 1286 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq); 1287 splx(s); 1288 goto loop1; 1289 } 1290 nfs_timer_start(); 1291 splx(s); 1292 1293 /* 1294 * Search for a reply to return. 1295 */ 1296 s = splsoftclock(); 1297 LIST_FOREACH(nfsd, &slp->ns_tq, nd_tq) { 1298 if (nfsd->nd_mreq) { 1299 LIST_REMOVE(nfsd, nd_tq); 1300 *mrq = nfsd->nd_mreq; 1301 *ndp = nfsd; 1302 break; 1303 } 1304 } 1305 splx(s); 1306 return (0); 1307 } 1308 1309 /* 1310 * Coalesce the write request nfsd into owp. To do this we must: 1311 * - remove nfsd from the queues 1312 * - merge nfsd->nd_mrep into owp->nd_mrep 1313 * - update the nd_eoff and nd_stable for owp 1314 * - put nfsd on owp's nd_coalesce list 1315 * NB: Must be called at splsoftclock(). 1316 */ 1317 void 1318 nfsrvw_coalesce(owp, nfsd) 1319 struct nfsrv_descript *owp; 1320 struct nfsrv_descript *nfsd; 1321 { 1322 int overlap; 1323 struct mbuf *mp; 1324 struct nfsrv_descript *m; 1325 1326 LIST_REMOVE(nfsd, nd_hash); 1327 LIST_REMOVE(nfsd, nd_tq); 1328 if (owp->nd_eoff < nfsd->nd_eoff) { 1329 overlap = owp->nd_eoff - nfsd->nd_off; 1330 if (overlap < 0) 1331 panic("nfsrv_coalesce: bad off"); 1332 if (overlap > 0) 1333 m_adj(nfsd->nd_mrep, overlap); 1334 mp = owp->nd_mrep; 1335 while (mp->m_next) 1336 mp = mp->m_next; 1337 mp->m_next = nfsd->nd_mrep; 1338 owp->nd_eoff = nfsd->nd_eoff; 1339 } else 1340 m_freem(nfsd->nd_mrep); 1341 nfsd->nd_mrep = NULL; 1342 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC) 1343 owp->nd_stable = NFSV3WRITE_FILESYNC; 1344 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC && 1345 owp->nd_stable == NFSV3WRITE_UNSTABLE) 1346 owp->nd_stable = NFSV3WRITE_DATASYNC; 1347 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq); 1348 /* 1349 * nfsd might hold coalesce elements! Move them to owp. 1350 * Otherwise, requests may be lost and clients will be stuck. 1351 */ 1352 while ((m = LIST_FIRST(&nfsd->nd_coalesce)) != NULL) { 1353 LIST_REMOVE(m, nd_tq); 1354 LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq); 1355 } 1356 } 1357 1358 /* 1359 * nfs create service 1360 * now does a truncate to 0 length via. setattr if it already exists 1361 */ 1362 int 1363 nfsrv_create(nfsd, slp, lwp, mrq) 1364 struct nfsrv_descript *nfsd; 1365 struct nfssvc_sock *slp; 1366 struct lwp *lwp; 1367 struct mbuf **mrq; 1368 { 1369 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1370 struct mbuf *nam = nfsd->nd_nam; 1371 char *dpos = nfsd->nd_dpos; 1372 kauth_cred_t cred = nfsd->nd_cr; 1373 struct nfs_fattr *fp; 1374 struct vattr va, dirfor, diraft; 1375 struct nfsv2_sattr *sp; 1376 u_int32_t *tl; 1377 struct nameidata nd; 1378 char *cp; 1379 int32_t t1; 1380 char *bpos; 1381 int error = 0, cache = 0, len, tsize, dirfor_ret = 1, diraft_ret = 1; 1382 int rdev = 0; 1383 int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0; 1384 char *cp2; 1385 struct mbuf *mb, *mreq; 1386 struct vnode *vp = NULL, *dirp = NULL; 1387 nfsrvfh_t nsfh; 1388 u_quad_t frev, tempsize; 1389 u_char cverf[NFSX_V3CREATEVERF]; 1390 1391 nd.ni_cnd.cn_nameiop = 0; 1392 nfsm_srvmtofh(&nsfh); 1393 nfsm_srvnamesiz(len); 1394 nd.ni_cnd.cn_cred = cred; 1395 nd.ni_cnd.cn_nameiop = CREATE; 1396 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1397 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 1398 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1399 if (dirp && v3) { 1400 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1401 } 1402 if (error) { 1403 nfsm_reply(NFSX_WCCDATA(v3)); 1404 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1405 if (dirp) 1406 vrele(dirp); 1407 return (0); 1408 } 1409 VATTR_NULL(&va); 1410 if (v3) { 1411 va.va_mode = 0; 1412 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1413 how = fxdr_unsigned(int, *tl); 1414 switch (how) { 1415 case NFSV3CREATE_GUARDED: 1416 if (nd.ni_vp) { 1417 error = EEXIST; 1418 break; 1419 } 1420 case NFSV3CREATE_UNCHECKED: 1421 nfsm_srvsattr(&va); 1422 break; 1423 case NFSV3CREATE_EXCLUSIVE: 1424 nfsm_dissect(cp, void *, NFSX_V3CREATEVERF); 1425 memcpy(cverf, cp, NFSX_V3CREATEVERF); 1426 exclusive_flag = 1; 1427 break; 1428 }; 1429 va.va_type = VREG; 1430 } else { 1431 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 1432 va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode)); 1433 if (va.va_type == VNON) 1434 va.va_type = VREG; 1435 va.va_mode = nfstov_mode(sp->sa_mode); 1436 switch (va.va_type) { 1437 case VREG: 1438 tsize = fxdr_unsigned(int32_t, sp->sa_size); 1439 if (tsize != -1) 1440 va.va_size = (u_quad_t)tsize; 1441 break; 1442 case VCHR: 1443 case VBLK: 1444 case VFIFO: 1445 rdev = fxdr_unsigned(int32_t, sp->sa_size); 1446 break; 1447 default: 1448 break; 1449 }; 1450 } 1451 1452 /* 1453 * Iff doesn't exist, create it 1454 * otherwise just truncate to 0 length 1455 * should I set the mode too ?? 1456 */ 1457 if (nd.ni_vp == NULL) { 1458 if (va.va_type == VREG || va.va_type == VSOCK) { 1459 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1460 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1461 if (!error) { 1462 if (exclusive_flag) { 1463 exclusive_flag = 0; 1464 VATTR_NULL(&va); 1465 /* 1466 * XXX 1467 * assuming NFSX_V3CREATEVERF 1468 * == sizeof(nfstime3) 1469 */ 1470 fxdr_nfsv3time(cverf, &va.va_atime); 1471 error = VOP_SETATTR(nd.ni_vp, &va, cred, 1472 lwp); 1473 } 1474 } 1475 } else if (va.va_type == VCHR || va.va_type == VBLK || 1476 va.va_type == VFIFO) { 1477 if (va.va_type == VCHR && rdev == 0xffffffff) 1478 va.va_type = VFIFO; 1479 if (va.va_type != VFIFO && 1480 (error = kauth_authorize_generic(cred, 1481 KAUTH_GENERIC_ISSUSER, NULL))) { 1482 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1483 vput(nd.ni_dvp); 1484 nfsm_reply(0); 1485 return (error); 1486 } else 1487 va.va_rdev = (dev_t)rdev; 1488 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1489 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, 1490 &va); 1491 if (error) { 1492 nfsm_reply(0); 1493 } 1494 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1495 vput(nd.ni_vp); 1496 vrele(nd.ni_dvp); 1497 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1498 error = EINVAL; 1499 nfsm_reply(0); 1500 } 1501 } else { 1502 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1503 vput(nd.ni_dvp); 1504 error = ENXIO; 1505 } 1506 vp = nd.ni_vp; 1507 } else { 1508 vp = nd.ni_vp; 1509 if (nd.ni_dvp == vp) 1510 vrele(nd.ni_dvp); 1511 else 1512 vput(nd.ni_dvp); 1513 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1514 if (!error && va.va_size != -1) { 1515 error = nfsrv_access(vp, VWRITE, cred, 1516 (nd.ni_cnd.cn_flags & RDONLY), lwp, 0); 1517 if (!error) { 1518 nqsrv_getl(vp, ND_WRITE); 1519 tempsize = va.va_size; 1520 VATTR_NULL(&va); 1521 va.va_size = tempsize; 1522 error = VOP_SETATTR(vp, &va, cred, lwp); 1523 } 1524 } 1525 if (error) 1526 vput(vp); 1527 } 1528 if (!error) { 1529 error = nfsrv_composefh(vp, &nsfh, v3); 1530 if (!error) 1531 error = VOP_GETATTR(vp, &va, cred, lwp); 1532 vput(vp); 1533 } 1534 if (v3) { 1535 if (exclusive_flag && !error) { 1536 /* 1537 * XXX assuming NFSX_V3CREATEVERF == sizeof(nfstime3) 1538 */ 1539 char oldverf[NFSX_V3CREATEVERF]; 1540 1541 txdr_nfsv3time(&va.va_atime, oldverf); 1542 if (memcmp(cverf, oldverf, NFSX_V3CREATEVERF)) 1543 error = EEXIST; 1544 } 1545 if (dirp) { 1546 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1547 } 1548 } 1549 if (dirp) { 1550 vrele(dirp); 1551 } 1552 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3)); 1553 if (v3) { 1554 if (!error) { 1555 nfsm_srvpostop_fh(&nsfh); 1556 nfsm_srvpostop_attr(0, &va); 1557 } 1558 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1559 } else { 1560 nfsm_srvfhtom(&nsfh, v3); 1561 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1562 nfsm_srvfillattr(&va, fp); 1563 } 1564 return (0); 1565 nfsmout: 1566 if (dirp) 1567 vrele(dirp); 1568 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1569 if (nd.ni_dvp == nd.ni_vp) 1570 vrele(nd.ni_dvp); 1571 else 1572 vput(nd.ni_dvp); 1573 if (nd.ni_vp) 1574 vput(nd.ni_vp); 1575 return (error); 1576 } 1577 1578 /* 1579 * nfs v3 mknod service 1580 */ 1581 int 1582 nfsrv_mknod(nfsd, slp, lwp, mrq) 1583 struct nfsrv_descript *nfsd; 1584 struct nfssvc_sock *slp; 1585 struct lwp *lwp; 1586 struct mbuf **mrq; 1587 { 1588 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1589 struct mbuf *nam = nfsd->nd_nam; 1590 char *dpos = nfsd->nd_dpos; 1591 kauth_cred_t cred = nfsd->nd_cr; 1592 struct vattr va, dirfor, diraft; 1593 u_int32_t *tl; 1594 struct nameidata nd; 1595 int32_t t1; 1596 char *bpos; 1597 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 1598 u_int32_t major, minor; 1599 enum vtype vtyp; 1600 char *cp2; 1601 struct mbuf *mb, *mreq; 1602 struct vnode *vp, *dirp = (struct vnode *)0; 1603 nfsrvfh_t nsfh; 1604 u_quad_t frev; 1605 1606 nd.ni_cnd.cn_nameiop = 0; 1607 nfsm_srvmtofh(&nsfh); 1608 nfsm_srvnamesiz(len); 1609 nd.ni_cnd.cn_cred = cred; 1610 nd.ni_cnd.cn_nameiop = CREATE; 1611 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1612 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 1613 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1614 if (dirp) 1615 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1616 if (error) { 1617 nfsm_reply(NFSX_WCCDATA(1)); 1618 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1619 if (dirp) 1620 vrele(dirp); 1621 return (0); 1622 } 1623 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1624 vtyp = nfsv3tov_type(*tl); 1625 if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) { 1626 error = NFSERR_BADTYPE; 1627 goto abort; 1628 } 1629 VATTR_NULL(&va); 1630 va.va_mode = 0; 1631 nfsm_srvsattr(&va); 1632 if (vtyp == VCHR || vtyp == VBLK) { 1633 dev_t rdev; 1634 1635 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1636 major = fxdr_unsigned(u_int32_t, *tl++); 1637 minor = fxdr_unsigned(u_int32_t, *tl); 1638 rdev = makedev(major, minor); 1639 if (major(rdev) != major || minor(rdev) != minor) { 1640 error = EINVAL; 1641 goto abort; 1642 } 1643 va.va_rdev = rdev; 1644 } 1645 1646 /* 1647 * Iff doesn't exist, create it. 1648 */ 1649 if (nd.ni_vp) { 1650 error = EEXIST; 1651 abort: 1652 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1653 if (nd.ni_dvp == nd.ni_vp) 1654 vrele(nd.ni_dvp); 1655 else 1656 vput(nd.ni_dvp); 1657 if (nd.ni_vp) 1658 vput(nd.ni_vp); 1659 goto out; 1660 } 1661 va.va_type = vtyp; 1662 if (vtyp == VSOCK) { 1663 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1664 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1665 } else { 1666 if (va.va_type != VFIFO && 1667 (error = kauth_authorize_generic(cred, 1668 KAUTH_GENERIC_ISSUSER, NULL))) { 1669 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1670 vput(nd.ni_dvp); 1671 goto out; 1672 } 1673 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1674 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1675 if (error) 1676 goto out; 1677 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1678 vput(nd.ni_vp); 1679 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1680 error = EINVAL; 1681 } 1682 } 1683 out: 1684 vp = nd.ni_vp; 1685 if (!error) { 1686 error = nfsrv_composefh(vp, &nsfh, true); 1687 if (!error) 1688 error = VOP_GETATTR(vp, &va, cred, lwp); 1689 vput(vp); 1690 } 1691 if (dirp) { 1692 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1693 vrele(dirp); 1694 } 1695 nfsm_reply(NFSX_SRVFH(&nsfh, true) + NFSX_POSTOPATTR(1) + 1696 NFSX_WCCDATA(1)); 1697 if (!error) { 1698 nfsm_srvpostop_fh(&nsfh); 1699 nfsm_srvpostop_attr(0, &va); 1700 } 1701 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1702 return (0); 1703 nfsmout: 1704 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1705 if (nd.ni_dvp == nd.ni_vp) 1706 vrele(nd.ni_dvp); 1707 else 1708 vput(nd.ni_dvp); 1709 if (nd.ni_vp) 1710 vput(nd.ni_vp); 1711 if (dirp) 1712 vrele(dirp); 1713 return (error); 1714 } 1715 1716 /* 1717 * nfs remove service 1718 */ 1719 int 1720 nfsrv_remove(nfsd, slp, lwp, mrq) 1721 struct nfsrv_descript *nfsd; 1722 struct nfssvc_sock *slp; 1723 struct lwp *lwp; 1724 struct mbuf **mrq; 1725 { 1726 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1727 struct mbuf *nam = nfsd->nd_nam; 1728 char *dpos = nfsd->nd_dpos; 1729 kauth_cred_t cred = nfsd->nd_cr; 1730 struct nameidata nd; 1731 u_int32_t *tl; 1732 int32_t t1; 1733 char *bpos; 1734 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 1735 int v3 = (nfsd->nd_flag & ND_NFSV3); 1736 char *cp2; 1737 struct mbuf *mb, *mreq; 1738 struct vnode *vp, *dirp; 1739 struct vattr dirfor, diraft; 1740 nfsrvfh_t nsfh; 1741 u_quad_t frev; 1742 1743 #ifndef nolint 1744 vp = (struct vnode *)0; 1745 #endif 1746 nfsm_srvmtofh(&nsfh); 1747 nfsm_srvnamesiz(len); 1748 nd.ni_cnd.cn_cred = cred; 1749 nd.ni_cnd.cn_nameiop = DELETE; 1750 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1751 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 1752 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1753 if (dirp && v3) { 1754 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1755 } 1756 if (!error) { 1757 vp = nd.ni_vp; 1758 if (vp->v_type == VDIR && 1759 (error = kauth_authorize_generic(cred, 1760 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 1761 goto out; 1762 /* 1763 * The root of a mounted filesystem cannot be deleted. 1764 */ 1765 if (vp->v_vflag & VV_ROOT) { 1766 error = EBUSY; 1767 goto out; 1768 } 1769 out: 1770 if (!error) { 1771 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1772 nqsrv_getl(vp, ND_WRITE); 1773 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1774 } else { 1775 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1776 if (nd.ni_dvp == vp) 1777 vrele(nd.ni_dvp); 1778 else 1779 vput(nd.ni_dvp); 1780 vput(vp); 1781 } 1782 } 1783 if (dirp) { 1784 if (v3) { 1785 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1786 } 1787 vrele(dirp); 1788 } 1789 nfsm_reply(NFSX_WCCDATA(v3)); 1790 if (v3) { 1791 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1792 return (0); 1793 } 1794 nfsm_srvdone; 1795 } 1796 1797 /* 1798 * nfs rename service 1799 */ 1800 int 1801 nfsrv_rename(nfsd, slp, lwp, mrq) 1802 struct nfsrv_descript *nfsd; 1803 struct nfssvc_sock *slp; 1804 struct lwp *lwp; 1805 struct mbuf **mrq; 1806 { 1807 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1808 struct mbuf *nam = nfsd->nd_nam; 1809 char *dpos = nfsd->nd_dpos; 1810 kauth_cred_t cred = nfsd->nd_cr; 1811 u_int32_t *tl; 1812 int32_t t1; 1813 char *bpos; 1814 int error = 0, cache = 0, fdirfor_ret = 1, fdiraft_ret = 1; 1815 uint32_t len, len2; 1816 int tdirfor_ret = 1, tdiraft_ret = 1; 1817 int v3 = (nfsd->nd_flag & ND_NFSV3); 1818 char *cp2; 1819 struct mbuf *mb, *mreq; 1820 struct nameidata fromnd, tond; 1821 struct vnode *fvp, *tvp, *tdvp; 1822 struct vnode *fdirp = NULL, *tdirp = NULL; 1823 struct vattr fdirfor, fdiraft, tdirfor, tdiraft; 1824 nfsrvfh_t fnsfh, tnsfh; 1825 u_quad_t frev; 1826 uid_t saved_uid; 1827 1828 #ifndef nolint 1829 fvp = (struct vnode *)0; 1830 #endif 1831 fromnd.ni_cnd.cn_nameiop = 0; 1832 tond.ni_cnd.cn_nameiop = 0; 1833 nfsm_srvmtofh(&fnsfh); 1834 nfsm_srvnamesiz(len); 1835 /* 1836 * Remember our original uid so that we can reset cr_uid before 1837 * the second nfs_namei() call, in case it is remapped. 1838 */ 1839 saved_uid = kauth_cred_geteuid(cred); 1840 fromnd.ni_cnd.cn_cred = cred; 1841 fromnd.ni_cnd.cn_nameiop = DELETE; 1842 fromnd.ni_cnd.cn_flags = LOCKPARENT | SAVESTART; 1843 error = nfs_namei(&fromnd, &fnsfh, len, slp, nam, &md, 1844 &dpos, &fdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1845 if (fdirp && v3) { 1846 fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred, lwp); 1847 } 1848 if (error) { 1849 nfsm_reply(2 * NFSX_WCCDATA(v3)); 1850 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 1851 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 1852 if (fdirp) 1853 vrele(fdirp); 1854 return (0); 1855 } 1856 if (fromnd.ni_dvp != fromnd.ni_vp) { 1857 VOP_UNLOCK(fromnd.ni_dvp, 0); 1858 } 1859 fvp = fromnd.ni_vp; 1860 nfsm_srvmtofh(&tnsfh); 1861 if (v3) { 1862 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 1863 len2 = fxdr_unsigned(uint32_t, *tl); 1864 /* len2 will be checked by nfs_namei */ 1865 } 1866 else { 1867 /* NFSv2 */ 1868 nfsm_strsiz(len2, NFS_MAXNAMLEN); 1869 } 1870 kauth_cred_seteuid(cred, saved_uid); 1871 tond.ni_cnd.cn_cred = cred; 1872 tond.ni_cnd.cn_nameiop = RENAME; 1873 tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART; 1874 error = nfs_namei(&tond, &tnsfh, len2, slp, nam, &md, 1875 &dpos, &tdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1876 if (tdirp && v3) { 1877 tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred, lwp); 1878 } 1879 if (error) { 1880 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1881 vrele(fromnd.ni_dvp); 1882 vrele(fvp); 1883 goto out1; 1884 } 1885 tdvp = tond.ni_dvp; 1886 tvp = tond.ni_vp; 1887 if (tvp != NULL) { 1888 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 1889 if (v3) 1890 error = EEXIST; 1891 else 1892 error = EISDIR; 1893 goto out; 1894 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 1895 if (v3) 1896 error = EEXIST; 1897 else 1898 error = ENOTDIR; 1899 goto out; 1900 } 1901 if (tvp->v_type == VDIR && tvp->v_mountedhere) { 1902 if (v3) 1903 error = EXDEV; 1904 else 1905 error = ENOTEMPTY; 1906 goto out; 1907 } 1908 } 1909 if (fvp->v_type == VDIR && fvp->v_mountedhere) { 1910 if (v3) 1911 error = EXDEV; 1912 else 1913 error = ENOTEMPTY; 1914 goto out; 1915 } 1916 if (fvp->v_mount != tdvp->v_mount) { 1917 if (v3) 1918 error = EXDEV; 1919 else 1920 error = ENOTEMPTY; 1921 goto out; 1922 } 1923 if (fvp == tdvp) { 1924 if (v3) 1925 error = EINVAL; 1926 else 1927 error = ENOTEMPTY; 1928 } 1929 /* 1930 * If source is the same as the destination (that is the 1931 * same vnode with the same name in the same directory), 1932 * then there is nothing to do. 1933 */ 1934 if (fvp == tvp && fromnd.ni_dvp == tdvp && 1935 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 1936 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr, 1937 fromnd.ni_cnd.cn_namelen)) 1938 error = -1; 1939 out: 1940 if (!error) { 1941 nqsrv_getl(fromnd.ni_dvp, ND_WRITE); 1942 nqsrv_getl(tdvp, ND_WRITE); 1943 if (tvp) { 1944 nqsrv_getl(tvp, ND_WRITE); 1945 } 1946 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 1947 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 1948 } else { 1949 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 1950 if (tdvp == tvp) 1951 vrele(tdvp); 1952 else 1953 vput(tdvp); 1954 if (tvp) 1955 vput(tvp); 1956 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1957 vrele(fromnd.ni_dvp); 1958 vrele(fvp); 1959 if (error == -1) 1960 error = 0; 1961 } 1962 vrele(tond.ni_startdir); 1963 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 1964 out1: 1965 if (fdirp) { 1966 if (v3) { 1967 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred, lwp); 1968 } 1969 vrele(fdirp); 1970 } 1971 if (tdirp) { 1972 if (v3) { 1973 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred, lwp); 1974 } 1975 vrele(tdirp); 1976 } 1977 vrele(fromnd.ni_startdir); 1978 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 1979 nfsm_reply(2 * NFSX_WCCDATA(v3)); 1980 if (v3) { 1981 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 1982 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 1983 } 1984 return (0); 1985 1986 nfsmout: 1987 if (fdirp) 1988 vrele(fdirp); 1989 #ifdef notdef 1990 if (tdirp) 1991 vrele(tdirp); 1992 #endif 1993 if (tond.ni_cnd.cn_nameiop) { 1994 vrele(tond.ni_startdir); 1995 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 1996 } 1997 if (fromnd.ni_cnd.cn_nameiop) { 1998 vrele(fromnd.ni_startdir); 1999 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2000 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2001 vrele(fromnd.ni_dvp); 2002 vrele(fvp); 2003 } 2004 return (error); 2005 } 2006 2007 /* 2008 * nfs link service 2009 */ 2010 int 2011 nfsrv_link(nfsd, slp, lwp, mrq) 2012 struct nfsrv_descript *nfsd; 2013 struct nfssvc_sock *slp; 2014 struct lwp *lwp; 2015 struct mbuf **mrq; 2016 { 2017 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2018 struct mbuf *nam = nfsd->nd_nam; 2019 char *dpos = nfsd->nd_dpos; 2020 kauth_cred_t cred = nfsd->nd_cr; 2021 struct nameidata nd; 2022 u_int32_t *tl; 2023 int32_t t1; 2024 char *bpos; 2025 int error = 0, rdonly, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 2026 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3); 2027 char *cp2; 2028 struct mbuf *mb, *mreq; 2029 struct vnode *vp, *xp, *dirp = (struct vnode *)0; 2030 struct vattr dirfor, diraft, at; 2031 nfsrvfh_t nsfh, dnsfh; 2032 u_quad_t frev; 2033 2034 nfsm_srvmtofh(&nsfh); 2035 nfsm_srvmtofh(&dnsfh); 2036 nfsm_srvnamesiz(len); 2037 error = nfsrv_fhtovp(&nsfh, false, &vp, cred, slp, nam, 2038 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 2039 if (error) { 2040 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2041 nfsm_srvpostop_attr(getret, &at); 2042 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2043 return (0); 2044 } 2045 if (vp->v_type == VDIR && (error = kauth_authorize_generic(cred, 2046 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 2047 goto out1; 2048 nd.ni_cnd.cn_cred = cred; 2049 nd.ni_cnd.cn_nameiop = CREATE; 2050 nd.ni_cnd.cn_flags = LOCKPARENT; 2051 error = nfs_namei(&nd, &dnsfh, len, slp, nam, &md, &dpos, 2052 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2053 if (dirp && v3) { 2054 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2055 } 2056 if (error) 2057 goto out1; 2058 xp = nd.ni_vp; 2059 if (xp != NULL) { 2060 error = EEXIST; 2061 goto out; 2062 } 2063 xp = nd.ni_dvp; 2064 if (vp->v_mount != xp->v_mount) 2065 error = EXDEV; 2066 out: 2067 if (!error) { 2068 nqsrv_getl(vp, ND_WRITE); 2069 nqsrv_getl(xp, ND_WRITE); 2070 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2071 } else { 2072 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2073 if (nd.ni_dvp == nd.ni_vp) 2074 vrele(nd.ni_dvp); 2075 else 2076 vput(nd.ni_dvp); 2077 if (nd.ni_vp) 2078 vrele(nd.ni_vp); 2079 } 2080 out1: 2081 if (v3) 2082 getret = VOP_GETATTR(vp, &at, cred, lwp); 2083 if (dirp) { 2084 if (v3) { 2085 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2086 } 2087 vrele(dirp); 2088 } 2089 vrele(vp); 2090 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2091 if (v3) { 2092 nfsm_srvpostop_attr(getret, &at); 2093 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2094 return (0); 2095 } 2096 nfsm_srvdone; 2097 } 2098 2099 /* 2100 * nfs symbolic link service 2101 */ 2102 int 2103 nfsrv_symlink(nfsd, slp, lwp, mrq) 2104 struct nfsrv_descript *nfsd; 2105 struct nfssvc_sock *slp; 2106 struct lwp *lwp; 2107 struct mbuf **mrq; 2108 { 2109 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2110 struct mbuf *nam = nfsd->nd_nam; 2111 char *dpos = nfsd->nd_dpos; 2112 kauth_cred_t cred = nfsd->nd_cr; 2113 struct vattr va, dirfor, diraft; 2114 struct nameidata nd; 2115 u_int32_t *tl; 2116 int32_t t1; 2117 struct nfsv2_sattr *sp; 2118 char *bpos, *pathcp = NULL, *cp2; 2119 struct uio io; 2120 struct iovec iv; 2121 int error = 0, cache = 0, dirfor_ret = 1, diraft_ret = 1; 2122 uint32_t len, len2; 2123 int v3 = (nfsd->nd_flag & ND_NFSV3); 2124 struct mbuf *mb, *mreq; 2125 struct vnode *dirp = (struct vnode *)0; 2126 nfsrvfh_t nsfh; 2127 u_quad_t frev; 2128 2129 nd.ni_cnd.cn_nameiop = 0; 2130 nfsm_srvmtofh(&nsfh); 2131 nfsm_srvnamesiz(len); 2132 nd.ni_cnd.cn_cred = cred; 2133 nd.ni_cnd.cn_nameiop = CREATE; 2134 nd.ni_cnd.cn_flags = LOCKPARENT; 2135 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 2136 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2137 if (dirp && v3) { 2138 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2139 } 2140 if (error) 2141 goto out; 2142 VATTR_NULL(&va); 2143 va.va_type = VLNK; 2144 if (v3) { 2145 va.va_mode = 0; 2146 nfsm_srvsattr(&va); 2147 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 2148 len2 = fxdr_unsigned(uint32_t, *tl); 2149 if (len2 > PATH_MAX) { 2150 /* XXX should check _PC_NO_TRUNC */ 2151 error = ENAMETOOLONG; 2152 goto abortop; 2153 } 2154 } 2155 else { 2156 /* NFSv2 */ 2157 nfsm_strsiz(len2, NFS_MAXPATHLEN); 2158 } 2159 pathcp = malloc(len2 + 1, M_TEMP, M_WAITOK); 2160 iv.iov_base = pathcp; 2161 iv.iov_len = len2; 2162 io.uio_resid = len2; 2163 io.uio_offset = 0; 2164 io.uio_iov = &iv; 2165 io.uio_iovcnt = 1; 2166 io.uio_rw = UIO_READ; 2167 UIO_SETUP_SYSSPACE(&io); 2168 nfsm_mtouio(&io, len2); 2169 if (!v3) { 2170 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 2171 va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode); 2172 } 2173 *(pathcp + len2) = '\0'; 2174 if (nd.ni_vp) { 2175 error = EEXIST; 2176 abortop: 2177 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2178 if (nd.ni_dvp == nd.ni_vp) 2179 vrele(nd.ni_dvp); 2180 else 2181 vput(nd.ni_dvp); 2182 if (nd.ni_vp) 2183 vrele(nd.ni_vp); 2184 goto out; 2185 } 2186 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2187 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp); 2188 if (!error) { 2189 if (v3) { 2190 error = nfsrv_composefh(nd.ni_vp, &nsfh, v3); 2191 if (!error) 2192 error = VOP_GETATTR(nd.ni_vp, &va, cred, lwp); 2193 vput(nd.ni_vp); 2194 } else { 2195 vput(nd.ni_vp); 2196 } 2197 } 2198 out: 2199 if (pathcp) 2200 free(pathcp, M_TEMP); 2201 if (dirp) { 2202 if (v3) { 2203 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2204 } 2205 vrele(dirp); 2206 } 2207 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_POSTOPATTR(v3) + 2208 NFSX_WCCDATA(v3)); 2209 if (v3) { 2210 if (!error) { 2211 nfsm_srvpostop_fh(&nsfh); 2212 nfsm_srvpostop_attr(0, &va); 2213 } 2214 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2215 } 2216 return (0); 2217 nfsmout: 2218 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2219 if (nd.ni_dvp == nd.ni_vp) 2220 vrele(nd.ni_dvp); 2221 else 2222 vput(nd.ni_dvp); 2223 if (nd.ni_vp) 2224 vrele(nd.ni_vp); 2225 if (dirp) 2226 vrele(dirp); 2227 if (pathcp) 2228 free(pathcp, M_TEMP); 2229 return (error); 2230 } 2231 2232 /* 2233 * nfs mkdir service 2234 */ 2235 int 2236 nfsrv_mkdir(nfsd, slp, lwp, mrq) 2237 struct nfsrv_descript *nfsd; 2238 struct nfssvc_sock *slp; 2239 struct lwp *lwp; 2240 struct mbuf **mrq; 2241 { 2242 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2243 struct mbuf *nam = nfsd->nd_nam; 2244 char *dpos = nfsd->nd_dpos; 2245 kauth_cred_t cred = nfsd->nd_cr; 2246 struct vattr va, dirfor, diraft; 2247 struct nfs_fattr *fp; 2248 struct nameidata nd; 2249 char *cp; 2250 u_int32_t *tl; 2251 int32_t t1; 2252 char *bpos; 2253 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 2254 int v3 = (nfsd->nd_flag & ND_NFSV3); 2255 char *cp2; 2256 struct mbuf *mb, *mreq; 2257 struct vnode *vp, *dirp = (struct vnode *)0; 2258 nfsrvfh_t nsfh; 2259 u_quad_t frev; 2260 2261 nfsm_srvmtofh(&nsfh); 2262 nfsm_srvnamesiz(len); 2263 nd.ni_cnd.cn_cred = cred; 2264 nd.ni_cnd.cn_nameiop = CREATE; 2265 nd.ni_cnd.cn_flags = LOCKPARENT; 2266 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 2267 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2268 if (dirp && v3) { 2269 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2270 } 2271 if (error) { 2272 nfsm_reply(NFSX_WCCDATA(v3)); 2273 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2274 if (dirp) 2275 vrele(dirp); 2276 return (0); 2277 } 2278 VATTR_NULL(&va); 2279 if (v3) { 2280 va.va_mode = 0; 2281 nfsm_srvsattr(&va); 2282 } else { 2283 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 2284 va.va_mode = nfstov_mode(*tl++); 2285 } 2286 va.va_type = VDIR; 2287 vp = nd.ni_vp; 2288 if (vp != NULL) { 2289 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2290 if (nd.ni_dvp == vp) 2291 vrele(nd.ni_dvp); 2292 else 2293 vput(nd.ni_dvp); 2294 vrele(vp); 2295 error = EEXIST; 2296 goto out; 2297 } 2298 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2299 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 2300 if (!error) { 2301 vp = nd.ni_vp; 2302 error = nfsrv_composefh(vp, &nsfh, v3); 2303 if (!error) 2304 error = VOP_GETATTR(vp, &va, cred, lwp); 2305 vput(vp); 2306 } 2307 out: 2308 if (dirp) { 2309 if (v3) { 2310 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2311 } 2312 vrele(dirp); 2313 } 2314 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_POSTOPATTR(v3) + 2315 NFSX_WCCDATA(v3)); 2316 if (v3) { 2317 if (!error) { 2318 nfsm_srvpostop_fh(&nsfh); 2319 nfsm_srvpostop_attr(0, &va); 2320 } 2321 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2322 } else { 2323 nfsm_srvfhtom(&nsfh, v3); 2324 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 2325 nfsm_srvfillattr(&va, fp); 2326 } 2327 return (0); 2328 nfsmout: 2329 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2330 if (nd.ni_dvp == nd.ni_vp) 2331 vrele(nd.ni_dvp); 2332 else 2333 vput(nd.ni_dvp); 2334 if (nd.ni_vp) 2335 vrele(nd.ni_vp); 2336 if (dirp) 2337 vrele(dirp); 2338 return (error); 2339 } 2340 2341 /* 2342 * nfs rmdir service 2343 */ 2344 int 2345 nfsrv_rmdir(nfsd, slp, lwp, mrq) 2346 struct nfsrv_descript *nfsd; 2347 struct nfssvc_sock *slp; 2348 struct lwp *lwp; 2349 struct mbuf **mrq; 2350 { 2351 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2352 struct mbuf *nam = nfsd->nd_nam; 2353 char *dpos = nfsd->nd_dpos; 2354 kauth_cred_t cred = nfsd->nd_cr; 2355 u_int32_t *tl; 2356 int32_t t1; 2357 char *bpos; 2358 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 2359 int v3 = (nfsd->nd_flag & ND_NFSV3); 2360 char *cp2; 2361 struct mbuf *mb, *mreq; 2362 struct vnode *vp, *dirp = (struct vnode *)0; 2363 struct vattr dirfor, diraft; 2364 nfsrvfh_t nsfh; 2365 struct nameidata nd; 2366 u_quad_t frev; 2367 2368 nfsm_srvmtofh(&nsfh); 2369 nfsm_srvnamesiz(len); 2370 nd.ni_cnd.cn_cred = cred; 2371 nd.ni_cnd.cn_nameiop = DELETE; 2372 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 2373 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 2374 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2375 if (dirp && v3) { 2376 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2377 } 2378 if (error) { 2379 nfsm_reply(NFSX_WCCDATA(v3)); 2380 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2381 if (dirp) 2382 vrele(dirp); 2383 return (0); 2384 } 2385 vp = nd.ni_vp; 2386 if (vp->v_type != VDIR) { 2387 error = ENOTDIR; 2388 goto out; 2389 } 2390 /* 2391 * No rmdir "." please. 2392 */ 2393 if (nd.ni_dvp == vp) { 2394 error = EINVAL; 2395 goto out; 2396 } 2397 /* 2398 * The root of a mounted filesystem cannot be deleted. 2399 */ 2400 if (vp->v_vflag & VV_ROOT) 2401 error = EBUSY; 2402 out: 2403 if (!error) { 2404 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2405 nqsrv_getl(vp, ND_WRITE); 2406 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2407 } else { 2408 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2409 if (nd.ni_dvp == nd.ni_vp) 2410 vrele(nd.ni_dvp); 2411 else 2412 vput(nd.ni_dvp); 2413 vput(vp); 2414 } 2415 if (dirp) { 2416 if (v3) { 2417 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2418 } 2419 vrele(dirp); 2420 } 2421 nfsm_reply(NFSX_WCCDATA(v3)); 2422 if (v3) { 2423 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2424 return (0); 2425 } 2426 nfsm_srvdone; 2427 } 2428 2429 /* 2430 * nfs readdir service 2431 * - mallocs what it thinks is enough to read 2432 * count rounded up to a multiple of NFS_SRVDIRBLKSIZ <= NFS_MAXREADDIR 2433 * - calls VOP_READDIR() 2434 * - loops around building the reply 2435 * if the output generated exceeds count break out of loop 2436 * The nfsm_clget macro is used here so that the reply will be packed 2437 * tightly in mbuf clusters. 2438 * - it only knows that it has encountered eof when the VOP_READDIR() 2439 * reads nothing 2440 * - as such one readdir rpc will return eof false although you are there 2441 * and then the next will return eof 2442 * - it trims out records with d_fileno == 0 2443 * this doesn't matter for Unix clients, but they might confuse clients 2444 * for other os'. 2445 * - it trims out records with d_type == DT_WHT 2446 * these cannot be seen through NFS (unless we extend the protocol) 2447 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less 2448 * than requested, but this may not apply to all filesystems. For 2449 * example, client NFS does not { although it is never remote mounted 2450 * anyhow } 2451 * The alternate call nfsrv_readdirplus() does lookups as well. 2452 * PS: The NFS protocol spec. does not clarify what the "count" byte 2453 * argument is a count of.. just name strings and file id's or the 2454 * entire reply rpc or ... 2455 * I tried just file name and id sizes and it confused the Sun client, 2456 * so I am using the full rpc size now. The "paranoia.." comment refers 2457 * to including the status longwords that are not a part of the dir. 2458 * "entry" structures, but are in the rpc. 2459 */ 2460 2461 #define NFS_SRVDIRBLKSIZ 1024 2462 2463 struct flrep { 2464 nfsuint64 fl_off; 2465 u_int32_t fl_postopok; 2466 u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)]; 2467 u_int32_t fl_fhok; 2468 u_int32_t fl_fhsize; 2469 }; 2470 2471 int 2472 nfsrv_readdir(nfsd, slp, lwp, mrq) 2473 struct nfsrv_descript *nfsd; 2474 struct nfssvc_sock *slp; 2475 struct lwp *lwp; 2476 struct mbuf **mrq; 2477 { 2478 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2479 struct mbuf *nam = nfsd->nd_nam; 2480 char *dpos = nfsd->nd_dpos; 2481 kauth_cred_t cred = nfsd->nd_cr; 2482 char *bp, *be; 2483 struct mbuf *mp; 2484 struct dirent *dp; 2485 char *cp; 2486 u_int32_t *tl; 2487 int32_t t1; 2488 char *bpos; 2489 struct mbuf *mb, *mreq, *mp2; 2490 char *cpos, *cend, *cp2, *rbuf; 2491 struct vnode *vp; 2492 struct vattr at; 2493 nfsrvfh_t nsfh; 2494 struct uio io; 2495 struct iovec iv; 2496 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2497 int siz, cnt, fullsiz, eofflag, rdonly, cache = 0, ncookies; 2498 int v3 = (nfsd->nd_flag & ND_NFSV3); 2499 u_quad_t frev, off, toff, verf; 2500 off_t *cookies = NULL, *cookiep; 2501 nfsuint64 jar; 2502 2503 nfsm_srvmtofh(&nsfh); 2504 if (v3) { 2505 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 2506 toff = fxdr_hyper(tl); 2507 tl += 2; 2508 verf = fxdr_hyper(tl); 2509 tl += 2; 2510 } else { 2511 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2512 toff = fxdr_unsigned(u_quad_t, *tl++); 2513 } 2514 off = toff; 2515 cnt = fxdr_unsigned(int, *tl); 2516 siz = ((cnt + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1)); 2517 xfer = NFS_SRVMAXDATA(nfsd); 2518 if (siz > xfer) 2519 siz = xfer; 2520 fullsiz = siz; 2521 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 2522 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 2523 if (!error && vp->v_type != VDIR) { 2524 error = ENOTDIR; 2525 vput(vp); 2526 } 2527 if (error) { 2528 nfsm_reply(NFSX_UNSIGNED); 2529 nfsm_srvpostop_attr(getret, &at); 2530 return (0); 2531 } 2532 nqsrv_getl(vp, ND_READ); 2533 if (v3) { 2534 error = getret = VOP_GETATTR(vp, &at, cred, lwp); 2535 #ifdef NFS3_STRICTVERF 2536 /* 2537 * XXX This check is too strict for Solaris 2.5 clients. 2538 */ 2539 if (!error && toff && verf != at.va_filerev) 2540 error = NFSERR_BAD_COOKIE; 2541 #endif 2542 } 2543 if (!error) 2544 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0); 2545 if (error) { 2546 vput(vp); 2547 nfsm_reply(NFSX_POSTOPATTR(v3)); 2548 nfsm_srvpostop_attr(getret, &at); 2549 return (0); 2550 } 2551 VOP_UNLOCK(vp, 0); 2552 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2553 again: 2554 iv.iov_base = rbuf; 2555 iv.iov_len = fullsiz; 2556 io.uio_iov = &iv; 2557 io.uio_iovcnt = 1; 2558 io.uio_offset = (off_t)off; 2559 io.uio_resid = fullsiz; 2560 io.uio_rw = UIO_READ; 2561 UIO_SETUP_SYSSPACE(&io); 2562 eofflag = 0; 2563 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2564 2565 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2566 2567 off = (off_t)io.uio_offset; 2568 if (!cookies && !error) 2569 error = NFSERR_PERM; 2570 if (v3) { 2571 getret = VOP_GETATTR(vp, &at, cred, lwp); 2572 if (!error) 2573 error = getret; 2574 } 2575 2576 VOP_UNLOCK(vp, 0); 2577 if (error) { 2578 vrele(vp); 2579 free((void *)rbuf, M_TEMP); 2580 if (cookies) 2581 free((void *)cookies, M_TEMP); 2582 nfsm_reply(NFSX_POSTOPATTR(v3)); 2583 nfsm_srvpostop_attr(getret, &at); 2584 return (0); 2585 } 2586 if (io.uio_resid) { 2587 siz -= io.uio_resid; 2588 2589 /* 2590 * If nothing read, return eof 2591 * rpc reply 2592 */ 2593 if (siz == 0) { 2594 vrele(vp); 2595 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + 2596 2 * NFSX_UNSIGNED); 2597 if (v3) { 2598 nfsm_srvpostop_attr(getret, &at); 2599 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2600 txdr_hyper(at.va_filerev, tl); 2601 tl += 2; 2602 } else 2603 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2604 *tl++ = nfs_false; 2605 *tl = nfs_true; 2606 free((void *)rbuf, M_TEMP); 2607 free((void *)cookies, M_TEMP); 2608 return (0); 2609 } 2610 } 2611 2612 /* 2613 * Check for degenerate cases of nothing useful read. 2614 * If so go try again 2615 */ 2616 cpos = rbuf; 2617 cend = rbuf + siz; 2618 dp = (struct dirent *)cpos; 2619 cookiep = cookies; 2620 2621 while (cpos < cend && ncookies > 0 && 2622 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2623 cpos += dp->d_reclen; 2624 dp = (struct dirent *)cpos; 2625 cookiep++; 2626 ncookies--; 2627 } 2628 if (cpos >= cend || ncookies == 0) { 2629 toff = off; 2630 siz = fullsiz; 2631 free(cookies, M_TEMP); 2632 cookies = NULL; 2633 goto again; 2634 } 2635 2636 len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */ 2637 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz); 2638 if (v3) { 2639 nfsm_srvpostop_attr(getret, &at); 2640 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2641 txdr_hyper(at.va_filerev, tl); 2642 } 2643 mp = mp2 = mb; 2644 bp = bpos; 2645 be = bp + M_TRAILINGSPACE(mp); 2646 2647 /* Loop through the records and build reply */ 2648 while (cpos < cend && ncookies > 0) { 2649 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 2650 nlen = dp->d_namlen; 2651 rem = nfsm_rndup(nlen)-nlen; 2652 len += (4 * NFSX_UNSIGNED + nlen + rem); 2653 if (v3) 2654 len += 2 * NFSX_UNSIGNED; 2655 if (len > cnt) { 2656 eofflag = 0; 2657 break; 2658 } 2659 /* 2660 * Build the directory record xdr from 2661 * the dirent entry. 2662 */ 2663 nfsm_clget; 2664 *tl = nfs_true; 2665 bp += NFSX_UNSIGNED; 2666 if (v3) { 2667 nfsm_clget; 2668 *tl = txdr_unsigned(dp->d_fileno >> 32); 2669 bp += NFSX_UNSIGNED; 2670 } 2671 nfsm_clget; 2672 *tl = txdr_unsigned(dp->d_fileno); 2673 bp += NFSX_UNSIGNED; 2674 nfsm_clget; 2675 *tl = txdr_unsigned(nlen); 2676 bp += NFSX_UNSIGNED; 2677 2678 /* And loop around copying the name */ 2679 xfer = nlen; 2680 cp = dp->d_name; 2681 while (xfer > 0) { 2682 nfsm_clget; 2683 if ((bp+xfer) > be) 2684 tsiz = be-bp; 2685 else 2686 tsiz = xfer; 2687 memcpy(bp, cp, tsiz); 2688 bp += tsiz; 2689 xfer -= tsiz; 2690 if (xfer > 0) 2691 cp += tsiz; 2692 } 2693 /* And null pad to an int32_t boundary */ 2694 for (i = 0; i < rem; i++) 2695 *bp++ = '\0'; 2696 nfsm_clget; 2697 2698 /* Finish off the record */ 2699 txdr_hyper(*cookiep, &jar); 2700 if (v3) { 2701 *tl = jar.nfsuquad[0]; 2702 bp += NFSX_UNSIGNED; 2703 nfsm_clget; 2704 } 2705 *tl = jar.nfsuquad[1]; 2706 bp += NFSX_UNSIGNED; 2707 } 2708 cpos += dp->d_reclen; 2709 dp = (struct dirent *)cpos; 2710 cookiep++; 2711 ncookies--; 2712 } 2713 vrele(vp); 2714 nfsm_clget; 2715 *tl = nfs_false; 2716 bp += NFSX_UNSIGNED; 2717 nfsm_clget; 2718 if (eofflag) 2719 *tl = nfs_true; 2720 else 2721 *tl = nfs_false; 2722 bp += NFSX_UNSIGNED; 2723 if (mp != mb) { 2724 if (bp < be) 2725 mp->m_len = bp - mtod(mp, char *); 2726 } else 2727 mp->m_len += bp - bpos; 2728 free((void *)rbuf, M_TEMP); 2729 free((void *)cookies, M_TEMP); 2730 nfsm_srvdone; 2731 } 2732 2733 int 2734 nfsrv_readdirplus(nfsd, slp, lwp, mrq) 2735 struct nfsrv_descript *nfsd; 2736 struct nfssvc_sock *slp; 2737 struct lwp *lwp; 2738 struct mbuf **mrq; 2739 { 2740 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2741 struct mbuf *nam = nfsd->nd_nam; 2742 char *dpos = nfsd->nd_dpos; 2743 kauth_cred_t cred = nfsd->nd_cr; 2744 char *bp, *be; 2745 struct mbuf *mp; 2746 struct dirent *dp; 2747 char *cp; 2748 u_int32_t *tl; 2749 int32_t t1; 2750 char *bpos; 2751 struct mbuf *mb, *mreq, *mp2; 2752 char *cpos, *cend, *cp2, *rbuf; 2753 struct vnode *vp, *nvp; 2754 struct flrep fl; 2755 nfsrvfh_t nsfh; 2756 struct uio io; 2757 struct iovec iv; 2758 struct vattr va, at, *vap = &va; 2759 struct nfs_fattr *fp; 2760 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2761 int siz, cnt, fullsiz, eofflag, rdonly, cache = 0, dirlen, ncookies; 2762 u_quad_t frev, off, toff, verf; 2763 off_t *cookies = NULL, *cookiep; 2764 2765 nfsm_srvmtofh(&nsfh); 2766 nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED); 2767 toff = fxdr_hyper(tl); 2768 tl += 2; 2769 verf = fxdr_hyper(tl); 2770 tl += 2; 2771 siz = fxdr_unsigned(int, *tl++); 2772 cnt = fxdr_unsigned(int, *tl); 2773 off = toff; 2774 siz = ((siz + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1)); 2775 xfer = NFS_SRVMAXDATA(nfsd); 2776 if (siz > xfer) 2777 siz = xfer; 2778 fullsiz = siz; 2779 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 2780 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 2781 if (!error && vp->v_type != VDIR) { 2782 error = ENOTDIR; 2783 vput(vp); 2784 } 2785 if (error) { 2786 nfsm_reply(NFSX_UNSIGNED); 2787 nfsm_srvpostop_attr(getret, &at); 2788 return (0); 2789 } 2790 error = getret = VOP_GETATTR(vp, &at, cred, lwp); 2791 #ifdef NFS3_STRICTVERF 2792 /* 2793 * XXX This check is too strict for Solaris 2.5 clients. 2794 */ 2795 if (!error && toff && verf != at.va_filerev) 2796 error = NFSERR_BAD_COOKIE; 2797 #endif 2798 if (!error) { 2799 nqsrv_getl(vp, ND_READ); 2800 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0); 2801 } 2802 if (error) { 2803 vput(vp); 2804 nfsm_reply(NFSX_V3POSTOPATTR); 2805 nfsm_srvpostop_attr(getret, &at); 2806 return (0); 2807 } 2808 VOP_UNLOCK(vp, 0); 2809 2810 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2811 again: 2812 iv.iov_base = rbuf; 2813 iv.iov_len = fullsiz; 2814 io.uio_iov = &iv; 2815 io.uio_iovcnt = 1; 2816 io.uio_offset = (off_t)off; 2817 io.uio_resid = fullsiz; 2818 io.uio_rw = UIO_READ; 2819 UIO_SETUP_SYSSPACE(&io); 2820 eofflag = 0; 2821 2822 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2823 2824 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2825 2826 off = (u_quad_t)io.uio_offset; 2827 getret = VOP_GETATTR(vp, &at, cred, lwp); 2828 2829 VOP_UNLOCK(vp, 0); 2830 2831 /* 2832 * If the VGET operation doesn't work for this filesystem, 2833 * we can't support readdirplus. Returning NOTSUPP should 2834 * make clients fall back to plain readdir. 2835 * There's no need to check for VPTOFH as well, we wouldn't 2836 * even be here otherwise. 2837 */ 2838 if (!getret) { 2839 if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp))) 2840 getret = (getret == EOPNOTSUPP) ? 2841 NFSERR_NOTSUPP : NFSERR_IO; 2842 else 2843 vput(nvp); 2844 } 2845 2846 if (!cookies && !error) 2847 error = NFSERR_PERM; 2848 if (!error) 2849 error = getret; 2850 if (error) { 2851 vrele(vp); 2852 if (cookies) 2853 free((void *)cookies, M_TEMP); 2854 free((void *)rbuf, M_TEMP); 2855 nfsm_reply(NFSX_V3POSTOPATTR); 2856 nfsm_srvpostop_attr(getret, &at); 2857 return (0); 2858 } 2859 if (io.uio_resid) { 2860 siz -= io.uio_resid; 2861 2862 /* 2863 * If nothing read, return eof 2864 * rpc reply 2865 */ 2866 if (siz == 0) { 2867 vrele(vp); 2868 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2869 2 * NFSX_UNSIGNED); 2870 nfsm_srvpostop_attr(getret, &at); 2871 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2872 txdr_hyper(at.va_filerev, tl); 2873 tl += 2; 2874 *tl++ = nfs_false; 2875 *tl = nfs_true; 2876 free((void *)cookies, M_TEMP); 2877 free((void *)rbuf, M_TEMP); 2878 return (0); 2879 } 2880 } 2881 2882 /* 2883 * Check for degenerate cases of nothing useful read. 2884 * If so go try again 2885 */ 2886 cpos = rbuf; 2887 cend = rbuf + siz; 2888 dp = (struct dirent *)cpos; 2889 cookiep = cookies; 2890 2891 while (cpos < cend && ncookies > 0 && 2892 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2893 cpos += dp->d_reclen; 2894 dp = (struct dirent *)cpos; 2895 cookiep++; 2896 ncookies--; 2897 } 2898 if (cpos >= cend || ncookies == 0) { 2899 toff = off; 2900 siz = fullsiz; 2901 free(cookies, M_TEMP); 2902 cookies = NULL; 2903 goto again; 2904 } 2905 2906 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED; 2907 nfsm_reply(cnt); 2908 nfsm_srvpostop_attr(getret, &at); 2909 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2910 txdr_hyper(at.va_filerev, tl); 2911 mp = mp2 = mb; 2912 bp = bpos; 2913 be = bp + M_TRAILINGSPACE(mp); 2914 2915 /* Loop through the records and build reply */ 2916 while (cpos < cend && ncookies > 0) { 2917 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 2918 nfsrvfh_t nnsfh; 2919 2920 nlen = dp->d_namlen; 2921 rem = nfsm_rndup(nlen)-nlen; 2922 2923 /* 2924 * For readdir_and_lookup get the vnode using 2925 * the file number. 2926 */ 2927 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp)) 2928 goto invalid; 2929 if (nfsrv_composefh(nvp, &nnsfh, true)) { 2930 vput(nvp); 2931 goto invalid; 2932 } 2933 if (VOP_GETATTR(nvp, vap, cred, lwp)) { 2934 vput(nvp); 2935 goto invalid; 2936 } 2937 vput(nvp); 2938 2939 /* 2940 * If either the dircount or maxcount will be 2941 * exceeded, get out now. Both of these lengths 2942 * are calculated conservatively, including all 2943 * XDR overheads. 2944 */ 2945 len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH + 2946 NFSX_V3POSTOPATTR); 2947 dirlen += (6 * NFSX_UNSIGNED + nlen + rem); 2948 if (len > cnt || dirlen > fullsiz) { 2949 eofflag = 0; 2950 break; 2951 } 2952 2953 /* 2954 * Build the directory record xdr from 2955 * the dirent entry. 2956 */ 2957 fp = (struct nfs_fattr *)&fl.fl_fattr; 2958 nfsm_srvfillattr(vap, fp); 2959 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH); 2960 fl.fl_fhok = nfs_true; 2961 fl.fl_postopok = nfs_true; 2962 txdr_hyper(*cookiep, fl.fl_off.nfsuquad); 2963 2964 nfsm_clget; 2965 *tl = nfs_true; 2966 bp += NFSX_UNSIGNED; 2967 nfsm_clget; 2968 *tl = txdr_unsigned(dp->d_fileno >> 32); 2969 bp += NFSX_UNSIGNED; 2970 nfsm_clget; 2971 *tl = txdr_unsigned(dp->d_fileno); 2972 bp += NFSX_UNSIGNED; 2973 nfsm_clget; 2974 *tl = txdr_unsigned(nlen); 2975 bp += NFSX_UNSIGNED; 2976 2977 /* And loop around copying the name */ 2978 xfer = nlen; 2979 cp = dp->d_name; 2980 while (xfer > 0) { 2981 nfsm_clget; 2982 if ((bp + xfer) > be) 2983 tsiz = be - bp; 2984 else 2985 tsiz = xfer; 2986 memcpy(bp, cp, tsiz); 2987 bp += tsiz; 2988 xfer -= tsiz; 2989 if (xfer > 0) 2990 cp += tsiz; 2991 } 2992 /* And null pad to an int32_t boundary */ 2993 for (i = 0; i < rem; i++) 2994 *bp++ = '\0'; 2995 2996 /* 2997 * Now copy the flrep structure out. 2998 */ 2999 xfer = sizeof(struct flrep); 3000 cp = (void *)&fl; 3001 while (xfer > 0) { 3002 nfsm_clget; 3003 if ((bp + xfer) > be) 3004 tsiz = be - bp; 3005 else 3006 tsiz = xfer; 3007 memcpy(bp, cp, tsiz); 3008 bp += tsiz; 3009 xfer -= tsiz; 3010 if (xfer > 0) 3011 cp += tsiz; 3012 } 3013 3014 /* 3015 * ... and filehandle. 3016 */ 3017 xfer = NFSRVFH_SIZE(&nnsfh); 3018 cp = NFSRVFH_DATA(&nnsfh); 3019 while (xfer > 0) { 3020 nfsm_clget; 3021 if ((bp + xfer) > be) 3022 tsiz = be - bp; 3023 else 3024 tsiz = xfer; 3025 memcpy(bp, cp, tsiz); 3026 bp += tsiz; 3027 xfer -= tsiz; 3028 if (xfer > 0) 3029 cp += tsiz; 3030 } 3031 } 3032 invalid: 3033 cpos += dp->d_reclen; 3034 dp = (struct dirent *)cpos; 3035 cookiep++; 3036 ncookies--; 3037 } 3038 vrele(vp); 3039 nfsm_clget; 3040 *tl = nfs_false; 3041 bp += NFSX_UNSIGNED; 3042 nfsm_clget; 3043 if (eofflag) 3044 *tl = nfs_true; 3045 else 3046 *tl = nfs_false; 3047 bp += NFSX_UNSIGNED; 3048 if (mp != mb) { 3049 if (bp < be) 3050 mp->m_len = bp - mtod(mp, char *); 3051 } else 3052 mp->m_len += bp - bpos; 3053 free((void *)cookies, M_TEMP); 3054 free((void *)rbuf, M_TEMP); 3055 nfsm_srvdone; 3056 } 3057 3058 /* 3059 * nfs commit service 3060 */ 3061 int 3062 nfsrv_commit(nfsd, slp, lwp, mrq) 3063 struct nfsrv_descript *nfsd; 3064 struct nfssvc_sock *slp; 3065 struct lwp *lwp; 3066 struct mbuf **mrq; 3067 { 3068 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3069 struct mbuf *nam = nfsd->nd_nam; 3070 char *dpos = nfsd->nd_dpos; 3071 kauth_cred_t cred = nfsd->nd_cr; 3072 struct vattr bfor, aft; 3073 struct vnode *vp; 3074 nfsrvfh_t nsfh; 3075 u_int32_t *tl; 3076 int32_t t1; 3077 char *bpos; 3078 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cache = 0; 3079 uint32_t cnt; 3080 char *cp2; 3081 struct mbuf *mb, *mreq; 3082 u_quad_t frev, off, end; 3083 3084 nfsm_srvmtofh(&nsfh); 3085 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED); 3086 3087 off = fxdr_hyper(tl); 3088 tl += 2; 3089 cnt = fxdr_unsigned(uint32_t, *tl); 3090 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3091 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3092 if (error) { 3093 nfsm_reply(2 * NFSX_UNSIGNED); 3094 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3095 return (0); 3096 } 3097 for_ret = VOP_GETATTR(vp, &bfor, cred, lwp); 3098 end = (cnt > 0) ? off + cnt : vp->v_size; 3099 if (end < off || end > vp->v_size) 3100 end = vp->v_size; 3101 if (off < vp->v_size) 3102 error = VOP_FSYNC(vp, cred, FSYNC_WAIT, off, end, lwp); 3103 /* else error == 0, from nfsrv_fhtovp() */ 3104 aft_ret = VOP_GETATTR(vp, &aft, cred, lwp); 3105 vput(vp); 3106 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF); 3107 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3108 if (!error) { 3109 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF); 3110 *tl++ = txdr_unsigned(boottime.tv_sec); 3111 *tl = txdr_unsigned(boottime.tv_usec); 3112 } else { 3113 return (0); 3114 } 3115 nfsm_srvdone; 3116 } 3117 3118 /* 3119 * nfs statfs service 3120 */ 3121 int 3122 nfsrv_statfs(nfsd, slp, lwp, mrq) 3123 struct nfsrv_descript *nfsd; 3124 struct nfssvc_sock *slp; 3125 struct lwp *lwp; 3126 struct mbuf **mrq; 3127 { 3128 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3129 struct mbuf *nam = nfsd->nd_nam; 3130 char *dpos = nfsd->nd_dpos; 3131 kauth_cred_t cred = nfsd->nd_cr; 3132 struct statvfs *sf = NULL; 3133 struct nfs_statfs *sfp; 3134 u_int32_t *tl; 3135 int32_t t1; 3136 char *bpos; 3137 int error = 0, rdonly, cache = 0, getret = 1; 3138 int v3 = (nfsd->nd_flag & ND_NFSV3); 3139 char *cp2; 3140 struct mbuf *mb, *mreq; 3141 struct vnode *vp; 3142 struct vattr at; 3143 nfsrvfh_t nsfh; 3144 u_quad_t frev, tval; 3145 3146 nfsm_srvmtofh(&nsfh); 3147 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3148 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3149 if (error) { 3150 nfsm_reply(NFSX_UNSIGNED); 3151 nfsm_srvpostop_attr(getret, &at); 3152 return (0); 3153 } 3154 sf = malloc(sizeof(*sf), M_TEMP, M_WAITOK); 3155 error = VFS_STATVFS(vp->v_mount, sf, lwp); 3156 getret = VOP_GETATTR(vp, &at, cred, lwp); 3157 vput(vp); 3158 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3)); 3159 if (v3) 3160 nfsm_srvpostop_attr(getret, &at); 3161 if (error) { 3162 free(sf, M_TEMP); 3163 return (0); 3164 } 3165 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3)); 3166 if (v3) { 3167 tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_frsize); 3168 txdr_hyper(tval, &sfp->sf_tbytes); 3169 tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_frsize); 3170 txdr_hyper(tval, &sfp->sf_fbytes); 3171 tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_frsize); 3172 txdr_hyper(tval, &sfp->sf_abytes); 3173 tval = (u_quad_t)sf->f_files; 3174 txdr_hyper(tval, &sfp->sf_tfiles); 3175 tval = (u_quad_t)sf->f_ffree; 3176 txdr_hyper(tval, &sfp->sf_ffiles); 3177 txdr_hyper(tval, &sfp->sf_afiles); 3178 sfp->sf_invarsec = 0; 3179 } else { 3180 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA); 3181 sfp->sf_bsize = txdr_unsigned(sf->f_frsize); 3182 sfp->sf_blocks = txdr_unsigned(sf->f_blocks); 3183 sfp->sf_bfree = txdr_unsigned(sf->f_bfree); 3184 sfp->sf_bavail = txdr_unsigned(sf->f_bavail); 3185 } 3186 nfsmout: 3187 if (sf) 3188 free(sf, M_TEMP); 3189 return error; 3190 } 3191 3192 /* 3193 * nfs fsinfo service 3194 */ 3195 int 3196 nfsrv_fsinfo(nfsd, slp, lwp, mrq) 3197 struct nfsrv_descript *nfsd; 3198 struct nfssvc_sock *slp; 3199 struct lwp *lwp; 3200 struct mbuf **mrq; 3201 { 3202 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3203 struct mbuf *nam = nfsd->nd_nam; 3204 char *dpos = nfsd->nd_dpos; 3205 kauth_cred_t cred = nfsd->nd_cr; 3206 u_int32_t *tl; 3207 struct nfsv3_fsinfo *sip; 3208 int32_t t1; 3209 char *bpos; 3210 int error = 0, rdonly, cache = 0, getret = 1; 3211 uint32_t maxdata; 3212 char *cp2; 3213 struct mbuf *mb, *mreq; 3214 struct vnode *vp; 3215 struct vattr at; 3216 nfsrvfh_t nsfh; 3217 u_quad_t frev, maxfsize; 3218 struct statvfs *sb; 3219 3220 nfsm_srvmtofh(&nsfh); 3221 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3222 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3223 if (error) { 3224 nfsm_reply(NFSX_UNSIGNED); 3225 nfsm_srvpostop_attr(getret, &at); 3226 return (0); 3227 } 3228 3229 /* XXX Try to make a guess on the max file size. */ 3230 sb = malloc(sizeof(*sb), M_TEMP, M_WAITOK); 3231 VFS_STATVFS(vp->v_mount, sb, (struct lwp *)0); 3232 maxfsize = (u_quad_t)0x80000000 * sb->f_frsize - 1; 3233 free(sb, M_TEMP); 3234 3235 getret = VOP_GETATTR(vp, &at, cred, lwp); 3236 vput(vp); 3237 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO); 3238 nfsm_srvpostop_attr(getret, &at); 3239 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO); 3240 3241 /* 3242 * XXX 3243 * There should be file system VFS OP(s) to get this information. 3244 * For now, assume ufs. 3245 */ 3246 if (slp->ns_so->so_type == SOCK_DGRAM) 3247 maxdata = NFS_MAXDGRAMDATA; 3248 else 3249 maxdata = NFS_MAXDATA; 3250 sip->fs_rtmax = txdr_unsigned(maxdata); 3251 sip->fs_rtpref = txdr_unsigned(maxdata); 3252 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE); 3253 sip->fs_wtmax = txdr_unsigned(maxdata); 3254 sip->fs_wtpref = txdr_unsigned(maxdata); 3255 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE); 3256 sip->fs_dtpref = txdr_unsigned(maxdata); 3257 txdr_hyper(maxfsize, &sip->fs_maxfilesize); 3258 sip->fs_timedelta.nfsv3_sec = 0; 3259 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1); 3260 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK | 3261 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS | 3262 NFSV3FSINFO_CANSETTIME); 3263 nfsm_srvdone; 3264 } 3265 3266 /* 3267 * nfs pathconf service 3268 */ 3269 int 3270 nfsrv_pathconf(nfsd, slp, lwp, mrq) 3271 struct nfsrv_descript *nfsd; 3272 struct nfssvc_sock *slp; 3273 struct lwp *lwp; 3274 struct mbuf **mrq; 3275 { 3276 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3277 struct mbuf *nam = nfsd->nd_nam; 3278 char *dpos = nfsd->nd_dpos; 3279 kauth_cred_t cred = nfsd->nd_cr; 3280 u_int32_t *tl; 3281 struct nfsv3_pathconf *pc; 3282 int32_t t1; 3283 char *bpos; 3284 int error = 0, rdonly, cache = 0, getret = 1; 3285 register_t linkmax, namemax, chownres, notrunc; 3286 char *cp2; 3287 struct mbuf *mb, *mreq; 3288 struct vnode *vp; 3289 struct vattr at; 3290 nfsrvfh_t nsfh; 3291 u_quad_t frev; 3292 3293 nfsm_srvmtofh(&nsfh); 3294 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3295 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3296 if (error) { 3297 nfsm_reply(NFSX_UNSIGNED); 3298 nfsm_srvpostop_attr(getret, &at); 3299 return (0); 3300 } 3301 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax); 3302 if (!error) 3303 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax); 3304 if (!error) 3305 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres); 3306 if (!error) 3307 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc); 3308 getret = VOP_GETATTR(vp, &at, cred, lwp); 3309 vput(vp); 3310 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF); 3311 nfsm_srvpostop_attr(getret, &at); 3312 if (error) 3313 return (0); 3314 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF); 3315 3316 pc->pc_linkmax = txdr_unsigned(linkmax); 3317 pc->pc_namemax = txdr_unsigned(namemax); 3318 pc->pc_notrunc = txdr_unsigned(notrunc); 3319 pc->pc_chownrestricted = txdr_unsigned(chownres); 3320 3321 /* 3322 * These should probably be supported by VOP_PATHCONF(), but 3323 * until msdosfs is exportable (why would you want to?), the 3324 * Unix defaults should be ok. 3325 */ 3326 pc->pc_caseinsensitive = nfs_false; 3327 pc->pc_casepreserving = nfs_true; 3328 nfsm_srvdone; 3329 } 3330 3331 /* 3332 * Null operation, used by clients to ping server 3333 */ 3334 /* ARGSUSED */ 3335 int 3336 nfsrv_null(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, 3337 struct lwp *lwp, struct mbuf **mrq) 3338 { 3339 struct mbuf *mrep = nfsd->nd_mrep; 3340 char *bpos; 3341 int error = NFSERR_RETVOID, cache = 0; 3342 struct mbuf *mb, *mreq; 3343 u_quad_t frev; 3344 3345 nfsm_reply(0); 3346 return (0); 3347 } 3348 3349 /* 3350 * No operation, used for obsolete procedures 3351 */ 3352 /* ARGSUSED */ 3353 int 3354 nfsrv_noop(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, 3355 struct lwp *lwp, struct mbuf **mrq) 3356 { 3357 struct mbuf *mrep = nfsd->nd_mrep; 3358 char *bpos; 3359 int error, cache = 0; 3360 struct mbuf *mb, *mreq; 3361 u_quad_t frev; 3362 3363 if (nfsd->nd_repstat) 3364 error = nfsd->nd_repstat; 3365 else 3366 error = EPROCUNAVAIL; 3367 nfsm_reply(0); 3368 return (0); 3369 } 3370 3371 /* 3372 * Perform access checking for vnodes obtained from file handles that would 3373 * refer to files already opened by a Unix client. You cannot just use 3374 * vn_writechk() and VOP_ACCESS() for two reasons. 3375 * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case 3376 * 2 - The owner is to be given access irrespective of mode bits for some 3377 * operations, so that processes that chmod after opening a file don't 3378 * break. I don't like this because it opens a security hole, but since 3379 * the nfs server opens a security hole the size of a barn door anyhow, 3380 * what the heck. 3381 * 3382 * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS() 3383 * will return EPERM instead of EACCESS. EPERM is always an error. 3384 */ 3385 int 3386 nfsrv_access(vp, flags, cred, rdonly, lwp, override) 3387 struct vnode *vp; 3388 int flags; 3389 kauth_cred_t cred; 3390 int rdonly; 3391 struct lwp *lwp; 3392 int override; 3393 { 3394 struct vattr vattr; 3395 int error; 3396 if (flags & VWRITE) { 3397 /* Just vn_writechk() changed to check rdonly */ 3398 /* 3399 * Disallow write attempts on read-only file systems; 3400 * unless the file is a socket or a block or character 3401 * device resident on the file system. 3402 */ 3403 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 3404 switch (vp->v_type) { 3405 case VREG: 3406 case VDIR: 3407 case VLNK: 3408 return (EROFS); 3409 default: 3410 break; 3411 } 3412 } 3413 3414 /* 3415 * If the vnode is in use as a process's text, 3416 * we can't allow writing. 3417 */ 3418 if (vp->v_iflag & VI_TEXT) 3419 return (ETXTBSY); 3420 } 3421 error = VOP_GETATTR(vp, &vattr, cred, lwp); 3422 if (error) 3423 return (error); 3424 error = VOP_ACCESS(vp, flags, cred, lwp); 3425 /* 3426 * Allow certain operations for the owner (reads and writes 3427 * on files that are already open). 3428 */ 3429 if (override && error == EACCES && kauth_cred_geteuid(cred) == vattr.va_uid) 3430 error = 0; 3431 return error; 3432 } 3433