1 /* $NetBSD: nfs_serv.c,v 1.87 2004/01/07 12:17:10 yamt 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 1.87 2004/01/07 12:17:10 yamt 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 <ufs/ufs/dir.h> 74 75 #include <uvm/uvm.h> 76 77 #include <nfs/nfsproto.h> 78 #include <nfs/rpcv2.h> 79 #include <nfs/nfs.h> 80 #include <nfs/xdr_subs.h> 81 #include <nfs/nfsm_subs.h> 82 #include <nfs/nqnfs.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 /* 96 * nfs v3 access service 97 */ 98 int 99 nfsrv3_access(nfsd, slp, procp, mrq) 100 struct nfsrv_descript *nfsd; 101 struct nfssvc_sock *slp; 102 struct proc *procp; 103 struct mbuf **mrq; 104 { 105 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 106 struct mbuf *nam = nfsd->nd_nam; 107 caddr_t dpos = nfsd->nd_dpos; 108 struct ucred *cred = &nfsd->nd_cr; 109 struct vnode *vp; 110 nfsfh_t nfh; 111 fhandle_t *fhp; 112 u_int32_t *tl; 113 int32_t t1; 114 caddr_t bpos; 115 int error = 0, rdonly, cache = 0, getret; 116 char *cp2; 117 struct mbuf *mb, *mreq; 118 struct vattr va; 119 u_long inmode, testmode, outmode; 120 u_quad_t frev; 121 122 fhp = &nfh.fh_generic; 123 nfsm_srvmtofh(fhp); 124 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 125 error = nfsrv_fhtovp(fhp, 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, procp, 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, procp, 0) == 0) 141 outmode |= testmode; 142 if ((inmode & NFSV3ACCESS_EXECUTE) && 143 nfsrv_access(vp, VEXEC, cred, rdonly, procp, 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, procp, 0) == 0) 150 outmode |= testmode; 151 if ((inmode & NFSV3ACCESS_LOOKUP) && 152 nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0) == 0) 153 outmode |= NFSV3ACCESS_LOOKUP; 154 } 155 getret = VOP_GETATTR(vp, &va, cred, procp); 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, procp, mrq) 169 struct nfsrv_descript *nfsd; 170 struct nfssvc_sock *slp; 171 struct proc *procp; 172 struct mbuf **mrq; 173 { 174 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 175 struct mbuf *nam = nfsd->nd_nam; 176 caddr_t dpos = nfsd->nd_dpos; 177 struct ucred *cred = &nfsd->nd_cr; 178 struct nfs_fattr *fp; 179 struct vattr va; 180 struct vnode *vp; 181 nfsfh_t nfh; 182 fhandle_t *fhp; 183 u_int32_t *tl; 184 int32_t t1; 185 caddr_t bpos; 186 int error = 0, rdonly, cache; 187 char *cp2; 188 struct mbuf *mb, *mreq; 189 u_quad_t frev; 190 191 fhp = &nfh.fh_generic; 192 nfsm_srvmtofh(fhp); 193 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, 194 (nfsd->nd_flag & ND_KERBAUTH), FALSE); 195 if (error) { 196 nfsm_reply(0); 197 return (0); 198 } 199 nqsrv_getl(vp, ND_READ); 200 error = VOP_GETATTR(vp, &va, cred, procp); 201 vput(vp); 202 nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); 203 if (error) 204 return (0); 205 nfsm_build(fp, struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); 206 nfsm_srvfillattr(&va, fp); 207 nfsm_srvdone; 208 } 209 210 /* 211 * nfs setattr service 212 */ 213 int 214 nfsrv_setattr(nfsd, slp, procp, mrq) 215 struct nfsrv_descript *nfsd; 216 struct nfssvc_sock *slp; 217 struct proc *procp; 218 struct mbuf **mrq; 219 { 220 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 221 struct mbuf *nam = nfsd->nd_nam; 222 caddr_t dpos = nfsd->nd_dpos; 223 struct ucred *cred = &nfsd->nd_cr; 224 struct vattr va, preat; 225 struct nfsv2_sattr *sp; 226 struct nfs_fattr *fp; 227 struct vnode *vp; 228 nfsfh_t nfh; 229 fhandle_t *fhp; 230 u_int32_t *tl; 231 int32_t t1; 232 caddr_t bpos; 233 int error = 0, rdonly, cache, preat_ret = 1, postat_ret = 1; 234 int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0; 235 char *cp2; 236 struct mbuf *mb, *mreq; 237 u_quad_t frev; 238 struct timespec guard; 239 struct mount *mp = NULL; 240 241 fhp = &nfh.fh_generic; 242 nfsm_srvmtofh(fhp); 243 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 244 return (ESTALE); 245 vn_start_write(NULL, &mp, V_WAIT); 246 VATTR_NULL(&va); 247 if (v3) { 248 nfsm_srvsattr(&va); 249 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 250 gcheck = fxdr_unsigned(int, *tl); 251 if (gcheck) { 252 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 253 fxdr_nfsv3time(tl, &guard); 254 } 255 } else { 256 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 257 /* 258 * Nah nah nah nah na nah 259 * There is a bug in the Sun client that puts 0xffff in the mode 260 * field of sattr when it should put in 0xffffffff. The u_short 261 * doesn't sign extend. 262 * --> check the low order 2 bytes for 0xffff 263 */ 264 if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff) 265 va.va_mode = nfstov_mode(sp->sa_mode); 266 if (sp->sa_uid != nfs_xdrneg1) 267 va.va_uid = fxdr_unsigned(uid_t, sp->sa_uid); 268 if (sp->sa_gid != nfs_xdrneg1) 269 va.va_gid = fxdr_unsigned(gid_t, sp->sa_gid); 270 if (sp->sa_size != nfs_xdrneg1) 271 va.va_size = fxdr_unsigned(u_quad_t, sp->sa_size); 272 if (sp->sa_atime.nfsv2_sec != nfs_xdrneg1) { 273 #ifdef notyet 274 fxdr_nfsv2time(&sp->sa_atime, &va.va_atime); 275 #else 276 va.va_atime.tv_sec = 277 fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec); 278 va.va_atime.tv_nsec = 0; 279 #endif 280 } 281 if (sp->sa_mtime.nfsv2_sec != nfs_xdrneg1) 282 fxdr_nfsv2time(&sp->sa_mtime, &va.va_mtime); 283 284 } 285 286 /* 287 * Now that we have all the fields, lets do it. 288 */ 289 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, 290 (nfsd->nd_flag & ND_KERBAUTH), FALSE); 291 if (error) { 292 nfsm_reply(2 * NFSX_UNSIGNED); 293 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 294 vn_finished_write(mp, 0); 295 return (0); 296 } 297 nqsrv_getl(vp, ND_WRITE); 298 if (v3) { 299 error = preat_ret = VOP_GETATTR(vp, &preat, cred, procp); 300 if (!error && gcheck && 301 (preat.va_ctime.tv_sec != guard.tv_sec || 302 preat.va_ctime.tv_nsec != guard.tv_nsec)) 303 error = NFSERR_NOT_SYNC; 304 if (error) { 305 vput(vp); 306 nfsm_reply(NFSX_WCCDATA(v3)); 307 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 308 vn_finished_write(mp, 0); 309 return (0); 310 } 311 } 312 313 /* 314 * If the size is being changed write acces is required, otherwise 315 * just check for a read only file system. 316 */ 317 if (va.va_size == ((u_quad_t)((quad_t) -1))) { 318 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 319 error = EROFS; 320 goto out; 321 } 322 } else { 323 if (vp->v_type == VDIR) { 324 error = EISDIR; 325 goto out; 326 } else if ((error = nfsrv_access(vp, VWRITE, cred, rdonly, 327 procp, 0)) != 0) 328 goto out; 329 } 330 error = VOP_SETATTR(vp, &va, cred, procp); 331 postat_ret = VOP_GETATTR(vp, &va, cred, procp); 332 if (!error) 333 error = postat_ret; 334 out: 335 vput(vp); 336 vn_finished_write(mp, 0); 337 nfsm_reply(NFSX_WCCORFATTR(v3)); 338 if (v3) { 339 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 340 return (0); 341 } else { 342 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 343 nfsm_srvfillattr(&va, fp); 344 } 345 nfsm_srvdone; 346 } 347 348 /* 349 * nfs lookup rpc 350 */ 351 int 352 nfsrv_lookup(nfsd, slp, procp, mrq) 353 struct nfsrv_descript *nfsd; 354 struct nfssvc_sock *slp; 355 struct proc *procp; 356 struct mbuf **mrq; 357 { 358 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 359 struct mbuf *nam = nfsd->nd_nam; 360 caddr_t dpos = nfsd->nd_dpos; 361 struct ucred *cred = &nfsd->nd_cr; 362 struct nfs_fattr *fp; 363 struct nameidata nd, ind, *ndp = &nd; 364 struct vnode *vp, *dirp; 365 nfsfh_t nfh; 366 fhandle_t *fhp; 367 caddr_t cp; 368 u_int32_t *tl; 369 int32_t t1; 370 caddr_t bpos; 371 int error = 0, cache, dirattr_ret = 1; 372 uint32_t len; 373 int v3 = (nfsd->nd_flag & ND_NFSV3), pubflag; 374 char *cp2; 375 struct mbuf *mb, *mreq; 376 struct vattr va, dirattr; 377 u_quad_t frev; 378 379 fhp = &nfh.fh_generic; 380 nfsm_srvmtofh(fhp); 381 nfsm_srvnamesiz(len); 382 383 pubflag = nfs_ispublicfh(fhp); 384 385 nd.ni_cnd.cn_cred = cred; 386 nd.ni_cnd.cn_nameiop = LOOKUP; 387 nd.ni_cnd.cn_flags = LOCKLEAF | SAVESTART; 388 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 389 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), pubflag); 390 391 if (!error && pubflag) { 392 if (nd.ni_vp->v_type == VDIR && nfs_pub.np_index != NULL) { 393 /* 394 * Setup call to lookup() to see if we can find 395 * the index file. Arguably, this doesn't belong 396 * in a kernel.. Ugh. 397 */ 398 ind = nd; 399 VOP_UNLOCK(nd.ni_vp, 0); 400 ind.ni_pathlen = strlen(nfs_pub.np_index); 401 ind.ni_cnd.cn_nameptr = ind.ni_cnd.cn_pnbuf = 402 nfs_pub.np_index; 403 ind.ni_startdir = nd.ni_vp; 404 VREF(ind.ni_startdir); 405 error = lookup(&ind); 406 if (!error) { 407 /* 408 * Found an index file. Get rid of 409 * the old references. 410 */ 411 if (dirp) 412 vrele(dirp); 413 dirp = nd.ni_vp; 414 vrele(nd.ni_startdir); 415 ndp = &ind; 416 } else 417 error = 0; 418 } 419 /* 420 * If the public filehandle was used, check that this lookup 421 * didn't result in a filehandle outside the publicly exported 422 * filesystem. 423 */ 424 425 if (!error && ndp->ni_vp->v_mount != nfs_pub.np_mount) { 426 vput(nd.ni_vp); 427 error = EPERM; 428 } 429 } 430 431 if (dirp) { 432 if (v3) 433 dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred, 434 procp); 435 vrele(dirp); 436 } 437 438 if (error) { 439 nfsm_reply(NFSX_POSTOPATTR(v3)); 440 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 441 return (0); 442 } 443 444 nqsrv_getl(ndp->ni_startdir, ND_READ); 445 vrele(ndp->ni_startdir); 446 PNBUF_PUT(nd.ni_cnd.cn_pnbuf); 447 vp = ndp->ni_vp; 448 memset((caddr_t)fhp, 0, sizeof(nfh)); 449 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; 450 error = VFS_VPTOFH(vp, &fhp->fh_fid); 451 if (!error) 452 error = VOP_GETATTR(vp, &va, cred, procp); 453 vput(vp); 454 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPORFATTR(v3) + NFSX_POSTOPATTR(v3)); 455 if (error) { 456 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 457 return (0); 458 } 459 nfsm_srvfhtom(fhp, v3); 460 if (v3) { 461 nfsm_srvpostop_attr(0, &va); 462 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 463 } else { 464 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 465 nfsm_srvfillattr(&va, fp); 466 } 467 nfsm_srvdone; 468 } 469 470 /* 471 * nfs readlink service 472 */ 473 int 474 nfsrv_readlink(nfsd, slp, procp, mrq) 475 struct nfsrv_descript *nfsd; 476 struct nfssvc_sock *slp; 477 struct proc *procp; 478 struct mbuf **mrq; 479 { 480 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 481 struct mbuf *nam = nfsd->nd_nam; 482 caddr_t dpos = nfsd->nd_dpos; 483 struct ucred *cred = &nfsd->nd_cr; 484 struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN]; 485 struct iovec *ivp = iv; 486 struct mbuf *mp; 487 u_int32_t *tl; 488 int32_t t1; 489 caddr_t bpos; 490 int error = 0, rdonly, cache, i, padlen, getret; 491 uint32_t len; 492 int v3 = (nfsd->nd_flag & ND_NFSV3); 493 char *cp2; 494 struct mbuf *mb, *mp2 = NULL, *mp3 = NULL, *mreq; 495 struct vnode *vp; 496 struct vattr attr; 497 nfsfh_t nfh; 498 fhandle_t *fhp; 499 struct uio io, *uiop = &io; 500 u_quad_t frev; 501 502 fhp = &nfh.fh_generic; 503 nfsm_srvmtofh(fhp); 504 len = 0; 505 i = 0; 506 while (len < NFS_MAXPATHLEN) { 507 mp = m_get(M_WAIT, MT_DATA); 508 MCLAIM(mp, &nfs_mowner); 509 m_clget(mp, M_WAIT); 510 mp->m_len = NFSMSIZ(mp); 511 if (len == 0) 512 mp3 = mp2 = mp; 513 else { 514 mp2->m_next = mp; 515 mp2 = mp; 516 } 517 if ((len+mp->m_len) > NFS_MAXPATHLEN) { 518 mp->m_len = NFS_MAXPATHLEN-len; 519 len = NFS_MAXPATHLEN; 520 } else 521 len += mp->m_len; 522 ivp->iov_base = mtod(mp, caddr_t); 523 ivp->iov_len = mp->m_len; 524 i++; 525 ivp++; 526 } 527 uiop->uio_iov = iv; 528 uiop->uio_iovcnt = i; 529 uiop->uio_offset = 0; 530 uiop->uio_resid = len; 531 uiop->uio_rw = UIO_READ; 532 uiop->uio_segflg = UIO_SYSSPACE; 533 uiop->uio_procp = (struct proc *)0; 534 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 535 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 536 if (error) { 537 m_freem(mp3); 538 nfsm_reply(2 * NFSX_UNSIGNED); 539 nfsm_srvpostop_attr(1, (struct vattr *)0); 540 return (0); 541 } 542 if (vp->v_type != VLNK) { 543 if (v3) 544 error = EINVAL; 545 else 546 error = ENXIO; 547 goto out; 548 } 549 nqsrv_getl(vp, ND_READ); 550 error = VOP_READLINK(vp, uiop, cred); 551 out: 552 getret = VOP_GETATTR(vp, &attr, cred, procp); 553 vput(vp); 554 if (error) 555 m_freem(mp3); 556 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED); 557 if (v3) { 558 nfsm_srvpostop_attr(getret, &attr); 559 if (error) 560 return (0); 561 } 562 len -= uiop->uio_resid; 563 padlen = nfsm_padlen(len); 564 if (uiop->uio_resid || padlen) 565 nfs_zeropad(mp3, uiop->uio_resid, padlen); 566 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 567 *tl = txdr_unsigned(len); 568 mb->m_next = mp3; 569 nfsm_srvdone; 570 } 571 572 /* 573 * nfs read service 574 */ 575 int 576 nfsrv_read(nfsd, slp, procp, mrq) 577 struct nfsrv_descript *nfsd; 578 struct nfssvc_sock *slp; 579 struct proc *procp; 580 struct mbuf **mrq; 581 { 582 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 583 struct mbuf *nam = nfsd->nd_nam; 584 caddr_t dpos = nfsd->nd_dpos; 585 struct ucred *cred = &nfsd->nd_cr; 586 struct mbuf *m; 587 struct nfs_fattr *fp; 588 u_int32_t *tl; 589 int32_t t1; 590 int i; 591 caddr_t bpos; 592 int error = 0, rdonly, cache, getret; 593 int v3 = (nfsd->nd_flag & ND_NFSV3); 594 uint32_t reqlen, len, cnt, left; 595 int padlen; 596 char *cp2; 597 struct mbuf *mb, *mreq; 598 struct vnode *vp; 599 nfsfh_t nfh; 600 fhandle_t *fhp; 601 struct uio io, *uiop = &io; 602 struct vattr va; 603 off_t off; 604 u_quad_t frev; 605 606 fhp = &nfh.fh_generic; 607 nfsm_srvmtofh(fhp); 608 if (v3) { 609 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 610 off = fxdr_hyper(tl); 611 } else { 612 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 613 off = (off_t)fxdr_unsigned(u_int32_t, *tl); 614 } 615 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 616 reqlen = fxdr_unsigned(uint32_t, *tl); 617 reqlen = MIN(reqlen, NFS_SRVMAXDATA(nfsd)); 618 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 619 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 620 if (error) { 621 nfsm_reply(2 * NFSX_UNSIGNED); 622 nfsm_srvpostop_attr(1, (struct vattr *)0); 623 return (0); 624 } 625 if (vp->v_type != VREG) { 626 if (v3) 627 error = EINVAL; 628 else 629 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 630 } 631 if (!error) { 632 nqsrv_getl(vp, ND_READ); 633 if ((error = nfsrv_access(vp, VREAD, cred, rdonly, procp, 1)) != 0) 634 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 1); 635 } 636 getret = VOP_GETATTR(vp, &va, cred, procp); 637 if (!error) 638 error = getret; 639 if (error) { 640 vput(vp); 641 nfsm_reply(NFSX_POSTOPATTR(v3)); 642 nfsm_srvpostop_attr(getret, &va); 643 return (0); 644 } 645 if (off >= va.va_size) 646 cnt = 0; 647 else if ((off + reqlen) > va.va_size) 648 cnt = va.va_size - off; 649 else 650 cnt = reqlen; 651 nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt)); 652 if (v3) { 653 nfsm_build(tl, u_int32_t *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED); 654 *tl++ = nfs_true; 655 fp = (struct nfs_fattr *)tl; 656 tl += (NFSX_V3FATTR / sizeof (u_int32_t)); 657 } else { 658 nfsm_build(tl, u_int32_t *, NFSX_V2FATTR + NFSX_UNSIGNED); 659 fp = (struct nfs_fattr *)tl; 660 tl += (NFSX_V2FATTR / sizeof (u_int32_t)); 661 } 662 len = left = cnt; 663 if (cnt > 0) { 664 if (nfsd_use_loan) { 665 struct vm_page **pgpp; 666 voff_t pgoff = trunc_page(off); 667 int npages; 668 vaddr_t lva; 669 670 npages = (round_page(off + cnt) - pgoff) >> PAGE_SHIFT; 671 KASSERT(npages <= M_EXT_MAXPAGES); /* XXX */ 672 673 /* allocate kva for mbuf data */ 674 lva = sokvaalloc(npages << PAGE_SHIFT, slp->ns_so); 675 if (lva == 0) { 676 /* fall back to VOP_READ */ 677 goto loan_fail; 678 } 679 680 /* allocate mbuf */ 681 m = m_get(M_WAIT, MT_DATA); 682 MCLAIM(m, &nfs_mowner); 683 pgpp = m->m_ext.ext_pgs; 684 685 /* loan pages */ 686 error = uvm_loanuobjpages(&vp->v_uobj, pgoff, npages, 687 pgpp); 688 if (error) { 689 sokvafree(lva, npages << PAGE_SHIFT); 690 m_free(m); 691 goto read_error; 692 } 693 694 /* associate kva to mbuf */ 695 MEXTADD(m, (void *)(lva + ((vaddr_t)off & PAGE_MASK)), 696 cnt, M_MBUF, soloanfree, slp->ns_so); 697 m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP; 698 m->m_len = cnt; 699 700 /* map pages */ 701 for (i = 0; i < npages; i++) { 702 pmap_kenter_pa(lva, VM_PAGE_TO_PHYS(pgpp[i]), 703 VM_PROT_READ); 704 lva += PAGE_SIZE; 705 } 706 707 pmap_update(pmap_kernel()); 708 709 mb->m_next = m; 710 mb = m; 711 error = 0; 712 uiop->uio_resid = 0; 713 } else { 714 struct iovec *iv; 715 struct iovec *iv2; 716 struct mbuf *m2; 717 int siz; 718 loan_fail: 719 /* 720 * Generate the mbuf list with the uio_iov ref. to it. 721 */ 722 i = 0; 723 m = m2 = mb; 724 while (left > 0) { 725 siz = min(M_TRAILINGSPACE(m), left); 726 if (siz > 0) { 727 left -= siz; 728 i++; 729 } 730 if (left > 0) { 731 m = m_get(M_WAIT, MT_DATA); 732 MCLAIM(m, &nfs_mowner); 733 m_clget(m, M_WAIT); 734 m->m_len = 0; 735 m2->m_next = m; 736 m2 = m; 737 } 738 } 739 iv = malloc(i * sizeof(struct iovec), M_TEMP, M_WAITOK); 740 uiop->uio_iov = iv2 = iv; 741 m = mb; 742 left = cnt; 743 i = 0; 744 while (left > 0) { 745 if (m == NULL) 746 panic("nfsrv_read iov"); 747 siz = min(M_TRAILINGSPACE(m), left); 748 if (siz > 0) { 749 iv->iov_base = mtod(m, caddr_t) + 750 m->m_len; 751 iv->iov_len = siz; 752 m->m_len += siz; 753 left -= siz; 754 iv++; 755 i++; 756 } 757 m = m->m_next; 758 } 759 uiop->uio_iovcnt = i; 760 uiop->uio_offset = off; 761 uiop->uio_resid = cnt; 762 uiop->uio_rw = UIO_READ; 763 uiop->uio_segflg = UIO_SYSSPACE; 764 error = VOP_READ(vp, uiop, IO_NODELOCKED, cred); 765 free((caddr_t)iv2, M_TEMP); 766 } 767 read_error: 768 if (error || (getret = VOP_GETATTR(vp, &va, cred, procp)) != 0){ 769 if (!error) 770 error = getret; 771 m_freem(mreq); 772 vput(vp); 773 nfsm_reply(NFSX_POSTOPATTR(v3)); 774 nfsm_srvpostop_attr(getret, &va); 775 return (0); 776 } 777 } else { 778 uiop->uio_resid = 0; 779 } 780 vput(vp); 781 nfsm_srvfillattr(&va, fp); 782 len -= uiop->uio_resid; 783 padlen = nfsm_padlen(len); 784 if (uiop->uio_resid || padlen) 785 nfs_zeropad(mb, uiop->uio_resid, padlen); 786 if (v3) { 787 /* count */ 788 *tl++ = txdr_unsigned(len); 789 /* eof */ 790 if (off + len >= va.va_size) 791 *tl++ = nfs_true; 792 else 793 *tl++ = nfs_false; 794 } 795 *tl = txdr_unsigned(len); 796 nfsm_srvdone; 797 } 798 799 /* 800 * nfs write service 801 */ 802 int 803 nfsrv_write(nfsd, slp, procp, mrq) 804 struct nfsrv_descript *nfsd; 805 struct nfssvc_sock *slp; 806 struct proc *procp; 807 struct mbuf **mrq; 808 { 809 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 810 struct mbuf *nam = nfsd->nd_nam; 811 caddr_t dpos = nfsd->nd_dpos; 812 struct ucred *cred = &nfsd->nd_cr; 813 struct iovec *ivp; 814 int i, cnt; 815 struct mbuf *mp; 816 struct nfs_fattr *fp; 817 struct iovec *iv; 818 struct vattr va, forat; 819 u_int32_t *tl; 820 int32_t t1; 821 caddr_t bpos; 822 int error = 0, rdonly, cache, len, forat_ret = 1; 823 int ioflags, aftat_ret = 1, retlen, zeroing, adjust; 824 int stable = NFSV3WRITE_FILESYNC; 825 int v3 = (nfsd->nd_flag & ND_NFSV3); 826 char *cp2; 827 struct mbuf *mb, *mreq; 828 struct vnode *vp; 829 nfsfh_t nfh; 830 fhandle_t *fhp; 831 struct uio io, *uiop = &io; 832 off_t off; 833 u_quad_t frev; 834 struct mount *mntp = NULL; 835 836 if (mrep == NULL) { 837 *mrq = NULL; 838 return (0); 839 } 840 fhp = &nfh.fh_generic; 841 nfsm_srvmtofh(fhp); 842 if ((mntp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 843 return (ESTALE); 844 vn_start_write(NULL, &mntp, V_WAIT); 845 if (v3) { 846 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 847 off = fxdr_hyper(tl); 848 tl += 3; 849 stable = fxdr_unsigned(int, *tl++); 850 } else { 851 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 852 off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 853 tl += 2; 854 } 855 retlen = len = fxdr_unsigned(int32_t, *tl); 856 cnt = i = 0; 857 858 /* 859 * For NFS Version 2, it is not obvious what a write of zero length 860 * should do, but I might as well be consistent with Version 3, 861 * which is to return ok so long as there are no permission problems. 862 */ 863 if (len > 0) { 864 zeroing = 1; 865 mp = mrep; 866 while (mp) { 867 if (mp == md) { 868 zeroing = 0; 869 adjust = dpos - mtod(mp, caddr_t); 870 mp->m_len -= adjust; 871 if (mp->m_len > 0 && adjust > 0) 872 NFSMADV(mp, adjust); 873 } 874 if (zeroing) 875 mp->m_len = 0; 876 else if (mp->m_len > 0) { 877 i += mp->m_len; 878 if (i > len) { 879 mp->m_len -= (i - len); 880 zeroing = 1; 881 } 882 if (mp->m_len > 0) 883 cnt++; 884 } 885 mp = mp->m_next; 886 } 887 } 888 if (len > NFS_MAXDATA || len < 0 || i < len) { 889 error = EIO; 890 nfsm_reply(2 * NFSX_UNSIGNED); 891 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 892 vn_finished_write(mntp, 0); 893 return (0); 894 } 895 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 896 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 897 if (error) { 898 nfsm_reply(2 * NFSX_UNSIGNED); 899 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 900 vn_finished_write(mntp, 0); 901 return (0); 902 } 903 if (v3) 904 forat_ret = VOP_GETATTR(vp, &forat, cred, procp); 905 if (vp->v_type != VREG) { 906 if (v3) 907 error = EINVAL; 908 else 909 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 910 } 911 if (!error) { 912 nqsrv_getl(vp, ND_WRITE); 913 error = nfsrv_access(vp, VWRITE, cred, rdonly, procp, 1); 914 } 915 if (error) { 916 vput(vp); 917 nfsm_reply(NFSX_WCCDATA(v3)); 918 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 919 vn_finished_write(mntp, 0); 920 return (0); 921 } 922 923 if (len > 0) { 924 ivp = malloc(cnt * sizeof (struct iovec), M_TEMP, M_WAITOK); 925 uiop->uio_iov = iv = ivp; 926 uiop->uio_iovcnt = cnt; 927 mp = mrep; 928 while (mp) { 929 if (mp->m_len > 0) { 930 ivp->iov_base = mtod(mp, caddr_t); 931 ivp->iov_len = mp->m_len; 932 ivp++; 933 } 934 mp = mp->m_next; 935 } 936 937 /* 938 * XXX 939 * The IO_METASYNC flag indicates that all metadata (and not 940 * just enough to ensure data integrity) must be written to 941 * stable storage synchronously. 942 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.) 943 */ 944 if (stable == NFSV3WRITE_UNSTABLE) 945 ioflags = IO_NODELOCKED; 946 else if (stable == NFSV3WRITE_DATASYNC) 947 ioflags = (IO_SYNC | IO_NODELOCKED); 948 else 949 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 950 uiop->uio_resid = len; 951 uiop->uio_rw = UIO_WRITE; 952 uiop->uio_segflg = UIO_SYSSPACE; 953 uiop->uio_procp = (struct proc *)0; 954 uiop->uio_offset = off; 955 error = VOP_WRITE(vp, uiop, ioflags, cred); 956 nfsstats.srvvop_writes++; 957 free(iv, M_TEMP); 958 } 959 aftat_ret = VOP_GETATTR(vp, &va, cred, procp); 960 vput(vp); 961 vn_finished_write(mntp, 0); 962 if (!error) 963 error = aftat_ret; 964 nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) + 965 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3)); 966 if (v3) { 967 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 968 if (error) 969 return (0); 970 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 971 *tl++ = txdr_unsigned(retlen); 972 if (stable == NFSV3WRITE_UNSTABLE) 973 *tl++ = txdr_unsigned(stable); 974 else 975 *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC); 976 /* 977 * Actually, there is no need to txdr these fields, 978 * but it may make the values more human readable, 979 * for debugging purposes. 980 */ 981 *tl++ = txdr_unsigned(boottime.tv_sec); 982 *tl = txdr_unsigned(boottime.tv_usec); 983 } else { 984 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 985 nfsm_srvfillattr(&va, fp); 986 } 987 nfsm_srvdone; 988 } 989 990 /* 991 * NFS write service with write gathering support. Called when 992 * nfsrvw_procrastinate > 0. 993 * See: Chet Juszczak, "Improving the Write Performance of an NFS Server", 994 * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco, 995 * Jan. 1994. 996 */ 997 int 998 nfsrv_writegather(ndp, slp, procp, mrq) 999 struct nfsrv_descript **ndp; 1000 struct nfssvc_sock *slp; 1001 struct proc *procp; 1002 struct mbuf **mrq; 1003 { 1004 struct iovec *ivp; 1005 struct mbuf *mp; 1006 struct nfsrv_descript *wp, *nfsd, *owp, *swp; 1007 struct nfs_fattr *fp; 1008 int i = 0; 1009 struct iovec *iov; 1010 struct nfsrvw_delayhash *wpp; 1011 struct ucred *cred; 1012 struct vattr va, forat; 1013 u_int32_t *tl; 1014 int32_t t1; 1015 caddr_t bpos, dpos; 1016 int error = 0, rdonly, cache, len = 0, forat_ret = 1; 1017 int ioflags, aftat_ret = 1, s, adjust, v3, zeroing; 1018 char *cp2; 1019 struct mbuf *mb, *mreq, *mrep, *md; 1020 struct vnode *vp; 1021 struct uio io, *uiop = &io; 1022 u_quad_t frev, cur_usec; 1023 struct mount *mntp = NULL; 1024 1025 *mrq = NULL; 1026 if (*ndp) { 1027 nfsd = *ndp; 1028 *ndp = NULL; 1029 mrep = nfsd->nd_mrep; 1030 md = nfsd->nd_md; 1031 dpos = nfsd->nd_dpos; 1032 cred = &nfsd->nd_cr; 1033 v3 = (nfsd->nd_flag & ND_NFSV3); 1034 LIST_INIT(&nfsd->nd_coalesce); 1035 nfsd->nd_mreq = NULL; 1036 nfsd->nd_stable = NFSV3WRITE_FILESYNC; 1037 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec; 1038 nfsd->nd_time = cur_usec + nfsrvw_procrastinate; 1039 1040 /* 1041 * Now, get the write header.. 1042 */ 1043 nfsm_srvmtofh(&nfsd->nd_fh); 1044 if (v3) { 1045 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 1046 nfsd->nd_off = fxdr_hyper(tl); 1047 tl += 3; 1048 nfsd->nd_stable = fxdr_unsigned(int, *tl++); 1049 } else { 1050 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1051 nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 1052 tl += 2; 1053 } 1054 len = fxdr_unsigned(int32_t, *tl); 1055 nfsd->nd_len = len; 1056 nfsd->nd_eoff = nfsd->nd_off + len; 1057 1058 /* 1059 * Trim the header out of the mbuf list and trim off any trailing 1060 * junk so that the mbuf list has only the write data. 1061 */ 1062 zeroing = 1; 1063 i = 0; 1064 mp = mrep; 1065 while (mp) { 1066 if (mp == md) { 1067 zeroing = 0; 1068 adjust = dpos - mtod(mp, caddr_t); 1069 mp->m_len -= adjust; 1070 if (mp->m_len > 0 && adjust > 0) 1071 NFSMADV(mp, adjust); 1072 } 1073 if (zeroing) 1074 mp->m_len = 0; 1075 else { 1076 i += mp->m_len; 1077 if (i > len) { 1078 mp->m_len -= (i - len); 1079 zeroing = 1; 1080 } 1081 } 1082 mp = mp->m_next; 1083 } 1084 if (len > NFS_MAXDATA || len < 0 || i < len) { 1085 nfsmout: 1086 m_freem(mrep); 1087 error = EIO; 1088 nfsm_writereply(2 * NFSX_UNSIGNED, v3); 1089 if (v3) 1090 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1091 nfsd->nd_mreq = mreq; 1092 nfsd->nd_mrep = NULL; 1093 nfsd->nd_time = 0; 1094 } 1095 1096 /* 1097 * Add this entry to the hash and time queues. 1098 */ 1099 s = splsoftclock(); 1100 owp = NULL; 1101 wp = LIST_FIRST(&slp->ns_tq); 1102 while (wp && wp->nd_time < nfsd->nd_time) { 1103 owp = wp; 1104 wp = LIST_NEXT(wp, nd_tq); 1105 } 1106 if (owp) { 1107 LIST_INSERT_AFTER(owp, nfsd, nd_tq); 1108 } else { 1109 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1110 } 1111 if (nfsd->nd_mrep) { 1112 wpp = NWDELAYHASH(slp, nfsd->nd_fh.fh_fid.fid_data); 1113 owp = NULL; 1114 wp = LIST_FIRST(wpp); 1115 while (wp && 1116 memcmp((caddr_t)&nfsd->nd_fh, (caddr_t)&wp->nd_fh, NFSX_V3FH)) { 1117 owp = wp; 1118 wp = LIST_NEXT(wp, nd_hash); 1119 } 1120 while (wp && wp->nd_off < nfsd->nd_off && 1121 !memcmp((caddr_t)&nfsd->nd_fh, (caddr_t)&wp->nd_fh, NFSX_V3FH)) { 1122 owp = wp; 1123 wp = LIST_NEXT(wp, nd_hash); 1124 } 1125 if (owp) { 1126 LIST_INSERT_AFTER(owp, nfsd, nd_hash); 1127 1128 /* 1129 * Search the hash list for overlapping entries and 1130 * coalesce. 1131 */ 1132 for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) { 1133 wp = LIST_NEXT(nfsd, nd_hash); 1134 if (NFSW_SAMECRED(owp, nfsd)) 1135 nfsrvw_coalesce(owp, nfsd); 1136 } 1137 } else { 1138 LIST_INSERT_HEAD(wpp, nfsd, nd_hash); 1139 } 1140 } 1141 splx(s); 1142 } 1143 1144 /* 1145 * Now, do VOP_WRITE()s for any one(s) that need to be done now 1146 * and generate the associated reply mbuf list(s). 1147 */ 1148 loop1: 1149 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec; 1150 s = splsoftclock(); 1151 for (nfsd = LIST_FIRST(&slp->ns_tq); nfsd; nfsd = owp) { 1152 owp = LIST_NEXT(nfsd, nd_tq); 1153 if (nfsd->nd_time > cur_usec) 1154 break; 1155 if (nfsd->nd_mreq) 1156 continue; 1157 LIST_REMOVE(nfsd, nd_tq); 1158 LIST_REMOVE(nfsd, nd_hash); 1159 splx(s); 1160 mrep = nfsd->nd_mrep; 1161 nfsd->nd_mrep = NULL; 1162 cred = &nfsd->nd_cr; 1163 v3 = (nfsd->nd_flag & ND_NFSV3); 1164 forat_ret = aftat_ret = 1; 1165 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp, 1166 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH), 1167 FALSE); 1168 if (!error) { 1169 if (v3) 1170 forat_ret = VOP_GETATTR(vp, &forat, cred, procp); 1171 if (vp->v_type != VREG) { 1172 if (v3) 1173 error = EINVAL; 1174 else 1175 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 1176 } 1177 } else 1178 vp = NULL; 1179 if (!error) { 1180 nqsrv_getl(vp, ND_WRITE); 1181 error = nfsrv_access(vp, VWRITE, cred, rdonly, procp, 1); 1182 } 1183 1184 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE) 1185 ioflags = IO_NODELOCKED; 1186 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC) 1187 ioflags = (IO_SYNC | IO_NODELOCKED); 1188 else 1189 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 1190 uiop->uio_rw = UIO_WRITE; 1191 uiop->uio_segflg = UIO_SYSSPACE; 1192 uiop->uio_procp = (struct proc *)0; 1193 uiop->uio_offset = nfsd->nd_off; 1194 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off; 1195 if (uiop->uio_resid > 0) { 1196 mp = mrep; 1197 i = 0; 1198 while (mp) { 1199 if (mp->m_len > 0) 1200 i++; 1201 mp = mp->m_next; 1202 } 1203 uiop->uio_iovcnt = i; 1204 iov = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK); 1205 uiop->uio_iov = ivp = iov; 1206 mp = mrep; 1207 while (mp) { 1208 if (mp->m_len > 0) { 1209 ivp->iov_base = mtod(mp, caddr_t); 1210 ivp->iov_len = mp->m_len; 1211 ivp++; 1212 } 1213 mp = mp->m_next; 1214 } 1215 if (!error) { 1216 if (vn_start_write(vp, &mntp, V_NOWAIT) != 0) { 1217 VOP_UNLOCK(vp, 0); 1218 vn_start_write(NULL, &mntp, V_WAIT); 1219 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1220 } 1221 if (!error) { 1222 error = VOP_WRITE(vp, uiop, ioflags, cred); 1223 nfsstats.srvvop_writes++; 1224 vn_finished_write(mntp, 0); 1225 } 1226 } 1227 free((caddr_t)iov, M_TEMP); 1228 } 1229 m_freem(mrep); 1230 if (vp) { 1231 aftat_ret = VOP_GETATTR(vp, &va, cred, procp); 1232 vput(vp); 1233 } 1234 1235 /* 1236 * Loop around generating replies for all write rpcs that have 1237 * now been completed. 1238 */ 1239 swp = nfsd; 1240 do { 1241 if (error) { 1242 nfsm_writereply(NFSX_WCCDATA(v3), v3); 1243 if (v3) { 1244 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1245 } 1246 } else { 1247 nfsm_writereply(NFSX_PREOPATTR(v3) + 1248 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED + 1249 NFSX_WRITEVERF(v3), v3); 1250 if (v3) { 1251 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1252 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1253 *tl++ = txdr_unsigned(nfsd->nd_len); 1254 *tl++ = txdr_unsigned(swp->nd_stable); 1255 /* 1256 * Actually, there is no need to txdr these fields, 1257 * but it may make the values more human readable, 1258 * for debugging purposes. 1259 */ 1260 *tl++ = txdr_unsigned(boottime.tv_sec); 1261 *tl = txdr_unsigned(boottime.tv_usec); 1262 } else { 1263 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1264 nfsm_srvfillattr(&va, fp); 1265 } 1266 } 1267 nfsd->nd_mreq = mreq; 1268 if (nfsd->nd_mrep) 1269 panic("nfsrv_write: nd_mrep not free"); 1270 1271 /* 1272 * Done. Put it at the head of the timer queue so that 1273 * the final phase can return the reply. 1274 */ 1275 s = splsoftclock(); 1276 if (nfsd != swp) { 1277 nfsd->nd_time = 0; 1278 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1279 } 1280 nfsd = LIST_FIRST(&swp->nd_coalesce); 1281 if (nfsd) { 1282 LIST_REMOVE(nfsd, nd_tq); 1283 } 1284 splx(s); 1285 } while (nfsd); 1286 s = splsoftclock(); 1287 swp->nd_time = 0; 1288 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq); 1289 splx(s); 1290 goto loop1; 1291 } 1292 splx(s); 1293 1294 /* 1295 * Search for a reply to return. 1296 */ 1297 s = splsoftclock(); 1298 LIST_FOREACH(nfsd, &slp->ns_tq, nd_tq) { 1299 if (nfsd->nd_mreq) { 1300 LIST_REMOVE(nfsd, nd_tq); 1301 *mrq = nfsd->nd_mreq; 1302 *ndp = nfsd; 1303 break; 1304 } 1305 } 1306 splx(s); 1307 return (0); 1308 } 1309 1310 /* 1311 * Coalesce the write request nfsd into owp. To do this we must: 1312 * - remove nfsd from the queues 1313 * - merge nfsd->nd_mrep into owp->nd_mrep 1314 * - update the nd_eoff and nd_stable for owp 1315 * - put nfsd on owp's nd_coalesce list 1316 * NB: Must be called at splsoftclock(). 1317 */ 1318 void 1319 nfsrvw_coalesce(owp, nfsd) 1320 struct nfsrv_descript *owp; 1321 struct nfsrv_descript *nfsd; 1322 { 1323 int overlap; 1324 struct mbuf *mp; 1325 struct nfsrv_descript *m; 1326 1327 LIST_REMOVE(nfsd, nd_hash); 1328 LIST_REMOVE(nfsd, nd_tq); 1329 if (owp->nd_eoff < nfsd->nd_eoff) { 1330 overlap = owp->nd_eoff - nfsd->nd_off; 1331 if (overlap < 0) 1332 panic("nfsrv_coalesce: bad off"); 1333 if (overlap > 0) 1334 m_adj(nfsd->nd_mrep, overlap); 1335 mp = owp->nd_mrep; 1336 while (mp->m_next) 1337 mp = mp->m_next; 1338 mp->m_next = nfsd->nd_mrep; 1339 owp->nd_eoff = nfsd->nd_eoff; 1340 } else 1341 m_freem(nfsd->nd_mrep); 1342 nfsd->nd_mrep = NULL; 1343 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC) 1344 owp->nd_stable = NFSV3WRITE_FILESYNC; 1345 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC && 1346 owp->nd_stable == NFSV3WRITE_UNSTABLE) 1347 owp->nd_stable = NFSV3WRITE_DATASYNC; 1348 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq); 1349 /* 1350 * nfsd might hold coalesce elements! Move them to owp. 1351 * Otherwise, requests may be lost and clients will be stuck. 1352 */ 1353 while ((m = LIST_FIRST(&nfsd->nd_coalesce)) != NULL) { 1354 LIST_REMOVE(m, nd_tq); 1355 LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq); 1356 } 1357 } 1358 1359 /* 1360 * nfs create service 1361 * now does a truncate to 0 length via. setattr if it already exists 1362 */ 1363 int 1364 nfsrv_create(nfsd, slp, procp, mrq) 1365 struct nfsrv_descript *nfsd; 1366 struct nfssvc_sock *slp; 1367 struct proc *procp; 1368 struct mbuf **mrq; 1369 { 1370 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1371 struct mbuf *nam = nfsd->nd_nam; 1372 caddr_t dpos = nfsd->nd_dpos; 1373 struct ucred *cred = &nfsd->nd_cr; 1374 struct nfs_fattr *fp; 1375 struct vattr va, dirfor, diraft; 1376 struct nfsv2_sattr *sp; 1377 u_int32_t *tl; 1378 struct nameidata nd; 1379 caddr_t cp; 1380 int32_t t1; 1381 caddr_t bpos; 1382 int error = 0, cache, len, tsize, dirfor_ret = 1, diraft_ret = 1; 1383 int rdev = 0; 1384 int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0; 1385 char *cp2; 1386 struct mbuf *mb, *mreq; 1387 struct vnode *vp = NULL, *dirp = NULL; 1388 nfsfh_t nfh; 1389 fhandle_t *fhp; 1390 u_quad_t frev, tempsize; 1391 u_char cverf[NFSX_V3CREATEVERF]; 1392 struct mount *mp = NULL; 1393 1394 nd.ni_cnd.cn_nameiop = 0; 1395 fhp = &nfh.fh_generic; 1396 nfsm_srvmtofh(fhp); 1397 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1398 return (ESTALE); 1399 vn_start_write(NULL, &mp, V_WAIT); 1400 nfsm_srvnamesiz(len); 1401 nd.ni_cnd.cn_cred = cred; 1402 nd.ni_cnd.cn_nameiop = CREATE; 1403 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1404 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1405 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1406 if (dirp) { 1407 if (v3) 1408 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, 1409 procp); 1410 else { 1411 vrele(dirp); 1412 dirp = (struct vnode *)0; 1413 } 1414 } 1415 if (error) { 1416 nfsm_reply(NFSX_WCCDATA(v3)); 1417 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1418 if (dirp) 1419 vrele(dirp); 1420 vn_finished_write(mp, 0); 1421 return (0); 1422 } 1423 VATTR_NULL(&va); 1424 if (v3) { 1425 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1426 how = fxdr_unsigned(int, *tl); 1427 switch (how) { 1428 case NFSV3CREATE_GUARDED: 1429 if (nd.ni_vp) { 1430 error = EEXIST; 1431 break; 1432 } 1433 case NFSV3CREATE_UNCHECKED: 1434 nfsm_srvsattr(&va); 1435 break; 1436 case NFSV3CREATE_EXCLUSIVE: 1437 nfsm_dissect(cp, caddr_t, NFSX_V3CREATEVERF); 1438 memcpy(cverf, cp, NFSX_V3CREATEVERF); 1439 exclusive_flag = 1; 1440 if (nd.ni_vp == NULL) 1441 va.va_mode = 0; 1442 break; 1443 }; 1444 va.va_type = VREG; 1445 } else { 1446 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 1447 va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode)); 1448 if (va.va_type == VNON) 1449 va.va_type = VREG; 1450 va.va_mode = nfstov_mode(sp->sa_mode); 1451 switch (va.va_type) { 1452 case VREG: 1453 tsize = fxdr_unsigned(int32_t, sp->sa_size); 1454 if (tsize != -1) 1455 va.va_size = (u_quad_t)tsize; 1456 break; 1457 case VCHR: 1458 case VBLK: 1459 case VFIFO: 1460 rdev = fxdr_unsigned(int32_t, sp->sa_size); 1461 break; 1462 default: 1463 break; 1464 }; 1465 } 1466 1467 /* 1468 * Iff doesn't exist, create it 1469 * otherwise just truncate to 0 length 1470 * should I set the mode too ?? 1471 */ 1472 if (nd.ni_vp == NULL) { 1473 if (va.va_type == VREG || va.va_type == VSOCK) { 1474 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1475 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1476 if (!error) { 1477 if (exclusive_flag) { 1478 exclusive_flag = 0; 1479 VATTR_NULL(&va); 1480 memcpy((caddr_t)&va.va_atime, cverf, 1481 NFSX_V3CREATEVERF); 1482 error = VOP_SETATTR(nd.ni_vp, &va, cred, 1483 procp); 1484 } 1485 } 1486 } else if (va.va_type == VCHR || va.va_type == VBLK || 1487 va.va_type == VFIFO) { 1488 if (va.va_type == VCHR && rdev == 0xffffffff) 1489 va.va_type = VFIFO; 1490 if (va.va_type != VFIFO && 1491 (error = suser(cred, (u_short *)0))) { 1492 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1493 vput(nd.ni_dvp); 1494 nfsm_reply(0); 1495 vn_finished_write(mp, 0); 1496 return (error); 1497 } else 1498 va.va_rdev = (dev_t)rdev; 1499 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1500 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, 1501 &va); 1502 if (error) { 1503 nfsm_reply(0); 1504 } 1505 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1506 vrele(nd.ni_dvp); 1507 vput(nd.ni_vp); 1508 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1509 error = EINVAL; 1510 nfsm_reply(0); 1511 } 1512 } else { 1513 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1514 vput(nd.ni_dvp); 1515 error = ENXIO; 1516 } 1517 vp = nd.ni_vp; 1518 } else { 1519 vp = nd.ni_vp; 1520 if (nd.ni_dvp == vp) 1521 vrele(nd.ni_dvp); 1522 else 1523 vput(nd.ni_dvp); 1524 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1525 if (!error && va.va_size != -1) { 1526 error = nfsrv_access(vp, VWRITE, cred, 1527 (nd.ni_cnd.cn_flags & RDONLY), procp, 0); 1528 if (!error) { 1529 nqsrv_getl(vp, ND_WRITE); 1530 tempsize = va.va_size; 1531 VATTR_NULL(&va); 1532 va.va_size = tempsize; 1533 error = VOP_SETATTR(vp, &va, cred, 1534 procp); 1535 } 1536 } 1537 if (error) 1538 vput(vp); 1539 } 1540 if (!error) { 1541 memset((caddr_t)fhp, 0, sizeof(nfh)); 1542 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; 1543 error = VFS_VPTOFH(vp, &fhp->fh_fid); 1544 if (!error) 1545 error = VOP_GETATTR(vp, &va, cred, procp); 1546 vput(vp); 1547 } 1548 if (v3) { 1549 if (exclusive_flag && !error && 1550 memcmp(cverf, (caddr_t)&va.va_atime, NFSX_V3CREATEVERF)) 1551 error = EEXIST; 1552 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); 1553 vrele(dirp); 1554 } 1555 nfsm_reply(NFSX_SRVFH(v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3)); 1556 if (v3) { 1557 if (!error) { 1558 nfsm_srvpostop_fh(fhp); 1559 nfsm_srvpostop_attr(0, &va); 1560 } 1561 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1562 } else { 1563 nfsm_srvfhtom(fhp, v3); 1564 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1565 nfsm_srvfillattr(&va, fp); 1566 } 1567 vn_finished_write(mp, 0); 1568 return (0); 1569 nfsmout: 1570 if (dirp) 1571 vrele(dirp); 1572 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1573 if (nd.ni_dvp == nd.ni_vp) 1574 vrele(nd.ni_dvp); 1575 else 1576 vput(nd.ni_dvp); 1577 if (nd.ni_vp) 1578 vput(nd.ni_vp); 1579 vn_finished_write(mp, 0); 1580 return (error); 1581 } 1582 1583 /* 1584 * nfs v3 mknod service 1585 */ 1586 int 1587 nfsrv_mknod(nfsd, slp, procp, mrq) 1588 struct nfsrv_descript *nfsd; 1589 struct nfssvc_sock *slp; 1590 struct proc *procp; 1591 struct mbuf **mrq; 1592 { 1593 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1594 struct mbuf *nam = nfsd->nd_nam; 1595 caddr_t dpos = nfsd->nd_dpos; 1596 struct ucred *cred = &nfsd->nd_cr; 1597 struct vattr va, dirfor, diraft; 1598 u_int32_t *tl; 1599 struct nameidata nd; 1600 int32_t t1; 1601 caddr_t bpos; 1602 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 1603 u_int32_t major, minor; 1604 enum vtype vtyp; 1605 char *cp2; 1606 struct mbuf *mb, *mreq; 1607 struct vnode *vp, *dirp = (struct vnode *)0; 1608 nfsfh_t nfh; 1609 fhandle_t *fhp; 1610 u_quad_t frev; 1611 struct mount *mp = NULL; 1612 1613 nd.ni_cnd.cn_nameiop = 0; 1614 fhp = &nfh.fh_generic; 1615 nfsm_srvmtofh(fhp); 1616 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1617 return (ESTALE); 1618 vn_start_write(NULL, &mp, V_WAIT); 1619 nfsm_srvnamesiz(len); 1620 nd.ni_cnd.cn_cred = cred; 1621 nd.ni_cnd.cn_nameiop = CREATE; 1622 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1623 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1624 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1625 if (dirp) 1626 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, procp); 1627 if (error) { 1628 nfsm_reply(NFSX_WCCDATA(1)); 1629 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1630 if (dirp) 1631 vrele(dirp); 1632 vn_finished_write(mp, 0); 1633 return (0); 1634 } 1635 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1636 vtyp = nfsv3tov_type(*tl); 1637 if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) { 1638 error = NFSERR_BADTYPE; 1639 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1640 if (nd.ni_dvp == nd.ni_vp) 1641 vrele(nd.ni_dvp); 1642 else 1643 vput(nd.ni_dvp); 1644 if (nd.ni_vp) 1645 vput(nd.ni_vp); 1646 goto out; 1647 } 1648 VATTR_NULL(&va); 1649 nfsm_srvsattr(&va); 1650 if (vtyp == VCHR || vtyp == VBLK) { 1651 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1652 major = fxdr_unsigned(u_int32_t, *tl++); 1653 minor = fxdr_unsigned(u_int32_t, *tl); 1654 va.va_rdev = makedev(major, minor); 1655 } 1656 1657 /* 1658 * Iff doesn't exist, create it. 1659 */ 1660 if (nd.ni_vp) { 1661 error = EEXIST; 1662 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1663 if (nd.ni_dvp == nd.ni_vp) 1664 vrele(nd.ni_dvp); 1665 else 1666 vput(nd.ni_dvp); 1667 vput(nd.ni_vp); 1668 goto out; 1669 } 1670 va.va_type = vtyp; 1671 if (vtyp == VSOCK) { 1672 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1673 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1674 } else { 1675 if (va.va_type != VFIFO && 1676 (error = suser(cred, (u_short *)0))) { 1677 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1678 vput(nd.ni_dvp); 1679 goto out; 1680 } 1681 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1682 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1683 if (error) { 1684 goto out; 1685 } 1686 if (error) 1687 goto out; 1688 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1689 vput(nd.ni_vp); 1690 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1691 error = EINVAL; 1692 } 1693 } 1694 out: 1695 vp = nd.ni_vp; 1696 if (!error) { 1697 memset((caddr_t)fhp, 0, sizeof(nfh)); 1698 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; 1699 error = VFS_VPTOFH(vp, &fhp->fh_fid); 1700 if (!error) 1701 error = VOP_GETATTR(vp, &va, cred, procp); 1702 vput(vp); 1703 } 1704 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); 1705 vrele(dirp); 1706 nfsm_reply(NFSX_SRVFH(1) + NFSX_POSTOPATTR(1) + NFSX_WCCDATA(1)); 1707 if (!error) { 1708 nfsm_srvpostop_fh(fhp); 1709 nfsm_srvpostop_attr(0, &va); 1710 } 1711 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1712 vn_finished_write(mp, 0); 1713 return (0); 1714 nfsmout: 1715 if (dirp) 1716 vrele(dirp); 1717 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1718 if (nd.ni_dvp == nd.ni_vp) 1719 vrele(nd.ni_dvp); 1720 else 1721 vput(nd.ni_dvp); 1722 if (nd.ni_vp) 1723 vput(nd.ni_vp); 1724 vn_finished_write(mp, 0); 1725 return (error); 1726 } 1727 1728 /* 1729 * nfs remove service 1730 */ 1731 int 1732 nfsrv_remove(nfsd, slp, procp, mrq) 1733 struct nfsrv_descript *nfsd; 1734 struct nfssvc_sock *slp; 1735 struct proc *procp; 1736 struct mbuf **mrq; 1737 { 1738 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1739 struct mbuf *nam = nfsd->nd_nam; 1740 caddr_t dpos = nfsd->nd_dpos; 1741 struct ucred *cred = &nfsd->nd_cr; 1742 struct nameidata nd; 1743 u_int32_t *tl; 1744 int32_t t1; 1745 caddr_t bpos; 1746 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 1747 int v3 = (nfsd->nd_flag & ND_NFSV3); 1748 char *cp2; 1749 struct mbuf *mb, *mreq; 1750 struct vnode *vp, *dirp; 1751 struct vattr dirfor, diraft; 1752 nfsfh_t nfh; 1753 fhandle_t *fhp; 1754 u_quad_t frev; 1755 struct mount *mp = NULL; 1756 1757 #ifndef nolint 1758 vp = (struct vnode *)0; 1759 #endif 1760 fhp = &nfh.fh_generic; 1761 nfsm_srvmtofh(fhp); 1762 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1763 return (ESTALE); 1764 vn_start_write(NULL, &mp, V_WAIT); 1765 nfsm_srvnamesiz(len); 1766 nd.ni_cnd.cn_cred = cred; 1767 nd.ni_cnd.cn_nameiop = DELETE; 1768 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1769 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1770 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1771 if (dirp) { 1772 if (v3) 1773 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, 1774 procp); 1775 else 1776 vrele(dirp); 1777 } 1778 if (!error) { 1779 vp = nd.ni_vp; 1780 if (vp->v_type == VDIR && 1781 (error = suser(cred, (u_short *)0)) != 0) 1782 goto out; 1783 /* 1784 * The root of a mounted filesystem cannot be deleted. 1785 */ 1786 if (vp->v_flag & VROOT) { 1787 error = EBUSY; 1788 goto out; 1789 } 1790 out: 1791 if (!error) { 1792 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1793 nqsrv_getl(vp, ND_WRITE); 1794 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1795 } else { 1796 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1797 if (nd.ni_dvp == vp) 1798 vrele(nd.ni_dvp); 1799 else 1800 vput(nd.ni_dvp); 1801 vput(vp); 1802 } 1803 } 1804 if (dirp && v3) { 1805 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); 1806 vrele(dirp); 1807 } 1808 nfsm_reply(NFSX_WCCDATA(v3)); 1809 if (v3) { 1810 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1811 vn_finished_write(mp, 0); 1812 return (0); 1813 } 1814 vn_finished_write(mp, 0); 1815 nfsm_srvdone; 1816 } 1817 1818 /* 1819 * nfs rename service 1820 */ 1821 int 1822 nfsrv_rename(nfsd, slp, procp, mrq) 1823 struct nfsrv_descript *nfsd; 1824 struct nfssvc_sock *slp; 1825 struct proc *procp; 1826 struct mbuf **mrq; 1827 { 1828 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1829 struct mbuf *nam = nfsd->nd_nam; 1830 caddr_t dpos = nfsd->nd_dpos; 1831 struct ucred *cred = &nfsd->nd_cr; 1832 u_int32_t *tl; 1833 int32_t t1; 1834 caddr_t bpos; 1835 int error = 0, cache, fdirfor_ret = 1, fdiraft_ret = 1; 1836 uint32_t len, len2; 1837 int tdirfor_ret = 1, tdiraft_ret = 1; 1838 int v3 = (nfsd->nd_flag & ND_NFSV3); 1839 char *cp2; 1840 struct mbuf *mb, *mreq; 1841 struct nameidata fromnd, tond; 1842 struct vnode *fvp, *tvp, *tdvp, *fdirp = (struct vnode *)0; 1843 struct vnode *tdirp = (struct vnode *)0; 1844 struct vattr fdirfor, fdiraft, tdirfor, tdiraft; 1845 nfsfh_t fnfh, tnfh; 1846 fhandle_t *ffhp, *tfhp; 1847 u_quad_t frev; 1848 uid_t saved_uid; 1849 struct mount *mp = NULL; 1850 1851 #ifndef nolint 1852 fvp = (struct vnode *)0; 1853 #endif 1854 ffhp = &fnfh.fh_generic; 1855 tfhp = &tnfh.fh_generic; 1856 fromnd.ni_cnd.cn_nameiop = 0; 1857 tond.ni_cnd.cn_nameiop = 0; 1858 nfsm_srvmtofh(ffhp); 1859 if ((mp = vfs_getvfs(&ffhp->fh_fsid)) == NULL) 1860 return (ESTALE); 1861 vn_start_write(NULL, &mp, V_WAIT); 1862 nfsm_srvnamesiz(len); 1863 /* 1864 * Remember our original uid so that we can reset cr_uid before 1865 * the second nfs_namei() call, in case it is remapped. 1866 */ 1867 saved_uid = cred->cr_uid; 1868 fromnd.ni_cnd.cn_cred = cred; 1869 fromnd.ni_cnd.cn_nameiop = DELETE; 1870 fromnd.ni_cnd.cn_flags = WANTPARENT | SAVESTART; 1871 error = nfs_namei(&fromnd, ffhp, len, slp, nam, &md, 1872 &dpos, &fdirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1873 if (fdirp) { 1874 if (v3) 1875 fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred, 1876 procp); 1877 else { 1878 vrele(fdirp); 1879 fdirp = (struct vnode *)0; 1880 } 1881 } 1882 if (error) { 1883 nfsm_reply(2 * NFSX_WCCDATA(v3)); 1884 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 1885 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 1886 if (fdirp) 1887 vrele(fdirp); 1888 vn_finished_write(mp, 0); 1889 return (0); 1890 } 1891 fvp = fromnd.ni_vp; 1892 nfsm_srvmtofh(tfhp); 1893 if (v3) { 1894 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 1895 len2 = fxdr_unsigned(uint32_t, *tl); 1896 /* len2 will be checked by nfs_namei */ 1897 } 1898 else { 1899 /* NFSv2 */ 1900 nfsm_strsiz(len2, NFS_MAXNAMLEN); 1901 } 1902 cred->cr_uid = saved_uid; 1903 tond.ni_cnd.cn_cred = cred; 1904 tond.ni_cnd.cn_nameiop = RENAME; 1905 tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART; 1906 error = nfs_namei(&tond, tfhp, len2, slp, nam, &md, 1907 &dpos, &tdirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1908 if (tdirp) { 1909 if (v3) 1910 tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred, 1911 procp); 1912 else { 1913 vrele(tdirp); 1914 tdirp = (struct vnode *)0; 1915 } 1916 } 1917 if (error) { 1918 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1919 vrele(fromnd.ni_dvp); 1920 vrele(fvp); 1921 goto out1; 1922 } 1923 tdvp = tond.ni_dvp; 1924 tvp = tond.ni_vp; 1925 if (tvp != NULL) { 1926 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 1927 if (v3) 1928 error = EEXIST; 1929 else 1930 error = EISDIR; 1931 goto out; 1932 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 1933 if (v3) 1934 error = EEXIST; 1935 else 1936 error = ENOTDIR; 1937 goto out; 1938 } 1939 if (tvp->v_type == VDIR && tvp->v_mountedhere) { 1940 if (v3) 1941 error = EXDEV; 1942 else 1943 error = ENOTEMPTY; 1944 goto out; 1945 } 1946 } 1947 if (fvp->v_type == VDIR && fvp->v_mountedhere) { 1948 if (v3) 1949 error = EXDEV; 1950 else 1951 error = ENOTEMPTY; 1952 goto out; 1953 } 1954 if (fvp->v_mount != tdvp->v_mount) { 1955 if (v3) 1956 error = EXDEV; 1957 else 1958 error = ENOTEMPTY; 1959 goto out; 1960 } 1961 if (fvp == tdvp) { 1962 if (v3) 1963 error = EINVAL; 1964 else 1965 error = ENOTEMPTY; 1966 } 1967 /* 1968 * If source is the same as the destination (that is the 1969 * same vnode with the same name in the same directory), 1970 * then there is nothing to do. 1971 */ 1972 if (fvp == tvp && fromnd.ni_dvp == tdvp && 1973 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 1974 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr, 1975 fromnd.ni_cnd.cn_namelen)) 1976 error = -1; 1977 out: 1978 if (!error) { 1979 nqsrv_getl(fromnd.ni_dvp, ND_WRITE); 1980 nqsrv_getl(tdvp, ND_WRITE); 1981 if (tvp) { 1982 nqsrv_getl(tvp, ND_WRITE); 1983 } 1984 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 1985 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 1986 } else { 1987 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 1988 if (tdvp == tvp) 1989 vrele(tdvp); 1990 else 1991 vput(tdvp); 1992 if (tvp) 1993 vput(tvp); 1994 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1995 vrele(fromnd.ni_dvp); 1996 vrele(fvp); 1997 if (error == -1) 1998 error = 0; 1999 } 2000 vrele(tond.ni_startdir); 2001 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 2002 out1: 2003 if (fdirp) { 2004 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred, procp); 2005 vrele(fdirp); 2006 } 2007 if (tdirp) { 2008 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred, procp); 2009 vrele(tdirp); 2010 } 2011 vrele(fromnd.ni_startdir); 2012 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2013 nfsm_reply(2 * NFSX_WCCDATA(v3)); 2014 if (v3) { 2015 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 2016 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 2017 } 2018 vn_finished_write(mp, 0); 2019 return (0); 2020 2021 nfsmout: 2022 if (fdirp) 2023 vrele(fdirp); 2024 if (tdirp) 2025 vrele(tdirp); 2026 if (tond.ni_cnd.cn_nameiop) { 2027 vrele(tond.ni_startdir); 2028 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 2029 } 2030 if (fromnd.ni_cnd.cn_nameiop) { 2031 vrele(fromnd.ni_startdir); 2032 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2033 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2034 vrele(fromnd.ni_dvp); 2035 vrele(fvp); 2036 } 2037 return (error); 2038 } 2039 2040 /* 2041 * nfs link service 2042 */ 2043 int 2044 nfsrv_link(nfsd, slp, procp, mrq) 2045 struct nfsrv_descript *nfsd; 2046 struct nfssvc_sock *slp; 2047 struct proc *procp; 2048 struct mbuf **mrq; 2049 { 2050 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2051 struct mbuf *nam = nfsd->nd_nam; 2052 caddr_t dpos = nfsd->nd_dpos; 2053 struct ucred *cred = &nfsd->nd_cr; 2054 struct nameidata nd; 2055 u_int32_t *tl; 2056 int32_t t1; 2057 caddr_t bpos; 2058 int error = 0, rdonly, cache, len, dirfor_ret = 1, diraft_ret = 1; 2059 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3); 2060 char *cp2; 2061 struct mbuf *mb, *mreq; 2062 struct vnode *vp, *xp, *dirp = (struct vnode *)0; 2063 struct vattr dirfor, diraft, at; 2064 nfsfh_t nfh, dnfh; 2065 fhandle_t *fhp, *dfhp; 2066 u_quad_t frev; 2067 struct mount *mp = NULL; 2068 2069 fhp = &nfh.fh_generic; 2070 dfhp = &dnfh.fh_generic; 2071 nfsm_srvmtofh(fhp); 2072 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2073 return (ESTALE); 2074 vn_start_write(NULL, &mp, V_WAIT); 2075 nfsm_srvmtofh(dfhp); 2076 nfsm_srvnamesiz(len); 2077 error = nfsrv_fhtovp(fhp, FALSE, &vp, cred, slp, nam, 2078 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2079 if (error) { 2080 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2081 nfsm_srvpostop_attr(getret, &at); 2082 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2083 vn_finished_write(mp, 0); 2084 return (0); 2085 } 2086 if (vp->v_type == VDIR && (error = suser(cred, (u_short *)0)) != 0) 2087 goto out1; 2088 nd.ni_cnd.cn_cred = cred; 2089 nd.ni_cnd.cn_nameiop = CREATE; 2090 nd.ni_cnd.cn_flags = LOCKPARENT; 2091 error = nfs_namei(&nd, dfhp, len, slp, nam, &md, &dpos, 2092 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2093 if (dirp) { 2094 if (v3) 2095 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, 2096 procp); 2097 else { 2098 vrele(dirp); 2099 dirp = (struct vnode *)0; 2100 } 2101 } 2102 if (error) 2103 goto out1; 2104 xp = nd.ni_vp; 2105 if (xp != NULL) { 2106 error = EEXIST; 2107 goto out; 2108 } 2109 xp = nd.ni_dvp; 2110 if (vp->v_mount != xp->v_mount) 2111 error = EXDEV; 2112 out: 2113 if (!error) { 2114 nqsrv_getl(vp, ND_WRITE); 2115 nqsrv_getl(xp, ND_WRITE); 2116 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2117 } else { 2118 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2119 if (nd.ni_dvp == nd.ni_vp) 2120 vrele(nd.ni_dvp); 2121 else 2122 vput(nd.ni_dvp); 2123 if (nd.ni_vp) 2124 vrele(nd.ni_vp); 2125 } 2126 out1: 2127 if (v3) 2128 getret = VOP_GETATTR(vp, &at, cred, procp); 2129 if (dirp) { 2130 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); 2131 vrele(dirp); 2132 } 2133 vrele(vp); 2134 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2135 if (v3) { 2136 nfsm_srvpostop_attr(getret, &at); 2137 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2138 vn_finished_write(mp, 0); 2139 return (0); 2140 } 2141 vn_finished_write(mp, 0); 2142 nfsm_srvdone; 2143 } 2144 2145 /* 2146 * nfs symbolic link service 2147 */ 2148 int 2149 nfsrv_symlink(nfsd, slp, procp, mrq) 2150 struct nfsrv_descript *nfsd; 2151 struct nfssvc_sock *slp; 2152 struct proc *procp; 2153 struct mbuf **mrq; 2154 { 2155 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2156 struct mbuf *nam = nfsd->nd_nam; 2157 caddr_t dpos = nfsd->nd_dpos; 2158 struct ucred *cred = &nfsd->nd_cr; 2159 struct vattr va, dirfor, diraft; 2160 struct nameidata nd; 2161 u_int32_t *tl; 2162 int32_t t1; 2163 struct nfsv2_sattr *sp; 2164 char *bpos, *pathcp = NULL, *cp2; 2165 struct uio io; 2166 struct iovec iv; 2167 int error = 0, cache, dirfor_ret = 1, diraft_ret = 1; 2168 uint32_t len, len2; 2169 int v3 = (nfsd->nd_flag & ND_NFSV3); 2170 struct mbuf *mb, *mreq; 2171 struct vnode *dirp = (struct vnode *)0; 2172 nfsfh_t nfh; 2173 fhandle_t *fhp; 2174 u_quad_t frev; 2175 struct mount *mp = NULL; 2176 2177 nd.ni_cnd.cn_nameiop = 0; 2178 fhp = &nfh.fh_generic; 2179 nfsm_srvmtofh(fhp); 2180 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2181 return (ESTALE); 2182 vn_start_write(NULL, &mp, V_WAIT); 2183 nfsm_srvnamesiz(len); 2184 nd.ni_cnd.cn_cred = cred; 2185 nd.ni_cnd.cn_nameiop = CREATE; 2186 nd.ni_cnd.cn_flags = LOCKPARENT; 2187 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2188 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2189 if (dirp) { 2190 if (v3) 2191 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, 2192 procp); 2193 else { 2194 vrele(dirp); 2195 dirp = (struct vnode *)0; 2196 } 2197 } 2198 if (error) 2199 goto out; 2200 VATTR_NULL(&va); 2201 if (v3) { 2202 nfsm_srvsattr(&va); 2203 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 2204 len2 = fxdr_unsigned(uint32_t, *tl); 2205 if (len2 > PATH_MAX) { 2206 /* XXX should check _PC_NO_TRUNC */ 2207 error = ENAMETOOLONG; 2208 goto abortop; 2209 } 2210 } 2211 else { 2212 /* NFSv2 */ 2213 nfsm_strsiz(len2, NFS_MAXPATHLEN); 2214 } 2215 pathcp = malloc(len2 + 1, M_TEMP, M_WAITOK); 2216 iv.iov_base = pathcp; 2217 iv.iov_len = len2; 2218 io.uio_resid = len2; 2219 io.uio_offset = 0; 2220 io.uio_iov = &iv; 2221 io.uio_iovcnt = 1; 2222 io.uio_segflg = UIO_SYSSPACE; 2223 io.uio_rw = UIO_READ; 2224 io.uio_procp = (struct proc *)0; 2225 nfsm_mtouio(&io, len2); 2226 if (!v3) { 2227 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 2228 va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode); 2229 } 2230 *(pathcp + len2) = '\0'; 2231 if (nd.ni_vp) { 2232 error = EEXIST; 2233 abortop: 2234 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2235 if (nd.ni_dvp == nd.ni_vp) 2236 vrele(nd.ni_dvp); 2237 else 2238 vput(nd.ni_dvp); 2239 if (nd.ni_vp) 2240 vrele(nd.ni_vp); 2241 goto out; 2242 } 2243 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2244 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp); 2245 if (!error) { 2246 if (v3) { 2247 memset((caddr_t)fhp, 0, sizeof(nfh)); 2248 fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsid; 2249 error = VFS_VPTOFH(nd.ni_vp, &fhp->fh_fid); 2250 if (!error) 2251 error = VOP_GETATTR(nd.ni_vp, &va, cred, 2252 procp); 2253 vput(nd.ni_vp); 2254 } else { 2255 vput(nd.ni_vp); 2256 } 2257 } 2258 out: 2259 if (pathcp) 2260 free(pathcp, M_TEMP); 2261 if (dirp) { 2262 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); 2263 vrele(dirp); 2264 } 2265 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2266 if (v3) { 2267 if (!error) { 2268 nfsm_srvpostop_fh(fhp); 2269 nfsm_srvpostop_attr(0, &va); 2270 } 2271 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2272 } 2273 vn_finished_write(mp, 0); 2274 return (0); 2275 nfsmout: 2276 if (dirp) 2277 vrele(dirp); 2278 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2279 if (nd.ni_dvp == nd.ni_vp) 2280 vrele(nd.ni_dvp); 2281 else 2282 vput(nd.ni_dvp); 2283 if (nd.ni_vp) 2284 vrele(nd.ni_vp); 2285 if (pathcp) 2286 free(pathcp, M_TEMP); 2287 vn_finished_write(mp, 0); 2288 return (error); 2289 } 2290 2291 /* 2292 * nfs mkdir service 2293 */ 2294 int 2295 nfsrv_mkdir(nfsd, slp, procp, mrq) 2296 struct nfsrv_descript *nfsd; 2297 struct nfssvc_sock *slp; 2298 struct proc *procp; 2299 struct mbuf **mrq; 2300 { 2301 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2302 struct mbuf *nam = nfsd->nd_nam; 2303 caddr_t dpos = nfsd->nd_dpos; 2304 struct ucred *cred = &nfsd->nd_cr; 2305 struct vattr va, dirfor, diraft; 2306 struct nfs_fattr *fp; 2307 struct nameidata nd; 2308 caddr_t cp; 2309 u_int32_t *tl; 2310 int32_t t1; 2311 caddr_t bpos; 2312 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 2313 int v3 = (nfsd->nd_flag & ND_NFSV3); 2314 char *cp2; 2315 struct mbuf *mb, *mreq; 2316 struct vnode *vp, *dirp = (struct vnode *)0; 2317 nfsfh_t nfh; 2318 fhandle_t *fhp; 2319 u_quad_t frev; 2320 struct mount *mp = NULL; 2321 2322 fhp = &nfh.fh_generic; 2323 nfsm_srvmtofh(fhp); 2324 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2325 return (ESTALE); 2326 vn_start_write(NULL, &mp, V_WAIT); 2327 nfsm_srvnamesiz(len); 2328 nd.ni_cnd.cn_cred = cred; 2329 nd.ni_cnd.cn_nameiop = CREATE; 2330 nd.ni_cnd.cn_flags = LOCKPARENT; 2331 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2332 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2333 if (dirp) { 2334 if (v3) 2335 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, 2336 procp); 2337 else { 2338 vrele(dirp); 2339 dirp = (struct vnode *)0; 2340 } 2341 } 2342 if (error) { 2343 nfsm_reply(NFSX_WCCDATA(v3)); 2344 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2345 if (dirp) 2346 vrele(dirp); 2347 vn_finished_write(mp, 0); 2348 return (0); 2349 } 2350 VATTR_NULL(&va); 2351 if (v3) { 2352 nfsm_srvsattr(&va); 2353 } else { 2354 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 2355 va.va_mode = nfstov_mode(*tl++); 2356 } 2357 va.va_type = VDIR; 2358 vp = nd.ni_vp; 2359 if (vp != NULL) { 2360 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2361 if (nd.ni_dvp == vp) 2362 vrele(nd.ni_dvp); 2363 else 2364 vput(nd.ni_dvp); 2365 vrele(vp); 2366 error = EEXIST; 2367 goto out; 2368 } 2369 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2370 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 2371 if (!error) { 2372 vp = nd.ni_vp; 2373 memset((caddr_t)fhp, 0, sizeof(nfh)); 2374 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; 2375 error = VFS_VPTOFH(vp, &fhp->fh_fid); 2376 if (!error) 2377 error = VOP_GETATTR(vp, &va, cred, procp); 2378 vput(vp); 2379 } 2380 out: 2381 if (dirp) { 2382 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); 2383 vrele(dirp); 2384 } 2385 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2386 if (v3) { 2387 if (!error) { 2388 nfsm_srvpostop_fh(fhp); 2389 nfsm_srvpostop_attr(0, &va); 2390 } 2391 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2392 } else { 2393 nfsm_srvfhtom(fhp, v3); 2394 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 2395 nfsm_srvfillattr(&va, fp); 2396 } 2397 vn_finished_write(mp, 0); 2398 return (0); 2399 nfsmout: 2400 if (dirp) 2401 vrele(dirp); 2402 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2403 if (nd.ni_dvp == nd.ni_vp) 2404 vrele(nd.ni_dvp); 2405 else 2406 vput(nd.ni_dvp); 2407 if (nd.ni_vp) 2408 vrele(nd.ni_vp); 2409 vn_finished_write(mp, 0); 2410 return (error); 2411 } 2412 2413 /* 2414 * nfs rmdir service 2415 */ 2416 int 2417 nfsrv_rmdir(nfsd, slp, procp, mrq) 2418 struct nfsrv_descript *nfsd; 2419 struct nfssvc_sock *slp; 2420 struct proc *procp; 2421 struct mbuf **mrq; 2422 { 2423 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2424 struct mbuf *nam = nfsd->nd_nam; 2425 caddr_t dpos = nfsd->nd_dpos; 2426 struct ucred *cred = &nfsd->nd_cr; 2427 u_int32_t *tl; 2428 int32_t t1; 2429 caddr_t bpos; 2430 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 2431 int v3 = (nfsd->nd_flag & ND_NFSV3); 2432 char *cp2; 2433 struct mbuf *mb, *mreq; 2434 struct vnode *vp, *dirp = (struct vnode *)0; 2435 struct vattr dirfor, diraft; 2436 nfsfh_t nfh; 2437 fhandle_t *fhp; 2438 struct nameidata nd; 2439 u_quad_t frev; 2440 struct mount *mp = NULL; 2441 2442 fhp = &nfh.fh_generic; 2443 nfsm_srvmtofh(fhp); 2444 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2445 return (ESTALE); 2446 vn_start_write(NULL, &mp, V_WAIT); 2447 nfsm_srvnamesiz(len); 2448 nd.ni_cnd.cn_cred = cred; 2449 nd.ni_cnd.cn_nameiop = DELETE; 2450 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 2451 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2452 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2453 if (dirp) { 2454 if (v3) 2455 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, 2456 procp); 2457 else { 2458 vrele(dirp); 2459 dirp = (struct vnode *)0; 2460 } 2461 } 2462 if (error) { 2463 nfsm_reply(NFSX_WCCDATA(v3)); 2464 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2465 if (dirp) 2466 vrele(dirp); 2467 vn_finished_write(mp, 0); 2468 return (0); 2469 } 2470 vp = nd.ni_vp; 2471 if (vp->v_type != VDIR) { 2472 error = ENOTDIR; 2473 goto out; 2474 } 2475 /* 2476 * No rmdir "." please. 2477 */ 2478 if (nd.ni_dvp == vp) { 2479 error = EINVAL; 2480 goto out; 2481 } 2482 /* 2483 * The root of a mounted filesystem cannot be deleted. 2484 */ 2485 if (vp->v_flag & VROOT) 2486 error = EBUSY; 2487 out: 2488 if (!error) { 2489 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2490 nqsrv_getl(vp, ND_WRITE); 2491 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2492 } else { 2493 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2494 if (nd.ni_dvp == nd.ni_vp) 2495 vrele(nd.ni_dvp); 2496 else 2497 vput(nd.ni_dvp); 2498 vput(vp); 2499 } 2500 if (dirp) { 2501 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); 2502 vrele(dirp); 2503 } 2504 nfsm_reply(NFSX_WCCDATA(v3)); 2505 if (v3) { 2506 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2507 vn_finished_write(mp, 0); 2508 return (0); 2509 } 2510 vn_finished_write(mp, 0); 2511 nfsm_srvdone; 2512 } 2513 2514 /* 2515 * nfs readdir service 2516 * - mallocs what it thinks is enough to read 2517 * count rounded up to a multiple of NFS_DIRBLKSIZ <= NFS_MAXREADDIR 2518 * - calls VOP_READDIR() 2519 * - loops around building the reply 2520 * if the output generated exceeds count break out of loop 2521 * The nfsm_clget macro is used here so that the reply will be packed 2522 * tightly in mbuf clusters. 2523 * - it only knows that it has encountered eof when the VOP_READDIR() 2524 * reads nothing 2525 * - as such one readdir rpc will return eof false although you are there 2526 * and then the next will return eof 2527 * - it trims out records with d_fileno == 0 2528 * this doesn't matter for Unix clients, but they might confuse clients 2529 * for other os'. 2530 * - it trims out records with d_type == DT_WHT 2531 * these cannot be seen through NFS (unless we extend the protocol) 2532 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less 2533 * than requested, but this may not apply to all filesystems. For 2534 * example, client NFS does not { although it is never remote mounted 2535 * anyhow } 2536 * The alternate call nfsrv_readdirplus() does lookups as well. 2537 * PS: The NFS protocol spec. does not clarify what the "count" byte 2538 * argument is a count of.. just name strings and file id's or the 2539 * entire reply rpc or ... 2540 * I tried just file name and id sizes and it confused the Sun client, 2541 * so I am using the full rpc size now. The "paranoia.." comment refers 2542 * to including the status longwords that are not a part of the dir. 2543 * "entry" structures, but are in the rpc. 2544 */ 2545 struct flrep { 2546 nfsuint64 fl_off; 2547 u_int32_t fl_postopok; 2548 u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)]; 2549 u_int32_t fl_fhok; 2550 u_int32_t fl_fhsize; 2551 u_int32_t fl_nfh[NFSX_V3FH / sizeof (u_int32_t)]; 2552 }; 2553 2554 int 2555 nfsrv_readdir(nfsd, slp, procp, mrq) 2556 struct nfsrv_descript *nfsd; 2557 struct nfssvc_sock *slp; 2558 struct proc *procp; 2559 struct mbuf **mrq; 2560 { 2561 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2562 struct mbuf *nam = nfsd->nd_nam; 2563 caddr_t dpos = nfsd->nd_dpos; 2564 struct ucred *cred = &nfsd->nd_cr; 2565 char *bp, *be; 2566 struct mbuf *mp; 2567 struct dirent *dp; 2568 caddr_t cp; 2569 u_int32_t *tl; 2570 int32_t t1; 2571 caddr_t bpos; 2572 struct mbuf *mb, *mreq, *mp2; 2573 char *cpos, *cend, *cp2, *rbuf; 2574 struct vnode *vp; 2575 struct vattr at; 2576 nfsfh_t nfh; 2577 fhandle_t *fhp; 2578 struct uio io; 2579 struct iovec iv; 2580 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2581 int siz, cnt, fullsiz, eofflag, rdonly, cache, ncookies; 2582 int v3 = (nfsd->nd_flag & ND_NFSV3); 2583 u_quad_t frev, off, toff, verf; 2584 off_t *cookies = NULL, *cookiep; 2585 nfsuint64 jar; 2586 2587 fhp = &nfh.fh_generic; 2588 nfsm_srvmtofh(fhp); 2589 if (v3) { 2590 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 2591 toff = fxdr_hyper(tl); 2592 tl += 2; 2593 verf = fxdr_hyper(tl); 2594 tl += 2; 2595 } else { 2596 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2597 toff = fxdr_unsigned(u_quad_t, *tl++); 2598 } 2599 off = toff; 2600 cnt = fxdr_unsigned(int, *tl); 2601 siz = ((cnt + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1)); 2602 xfer = NFS_SRVMAXDATA(nfsd); 2603 if (siz > xfer) 2604 siz = xfer; 2605 fullsiz = siz; 2606 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 2607 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2608 if (!error && vp->v_type != VDIR) { 2609 error = ENOTDIR; 2610 vput(vp); 2611 } 2612 if (error) { 2613 nfsm_reply(NFSX_UNSIGNED); 2614 nfsm_srvpostop_attr(getret, &at); 2615 return (0); 2616 } 2617 nqsrv_getl(vp, ND_READ); 2618 if (v3) { 2619 error = getret = VOP_GETATTR(vp, &at, cred, procp); 2620 #ifdef NFS3_STRICTVERF 2621 /* 2622 * XXX This check is too strict for Solaris 2.5 clients. 2623 */ 2624 if (!error && toff && verf != at.va_filerev) 2625 error = NFSERR_BAD_COOKIE; 2626 #endif 2627 } 2628 if (!error) 2629 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0); 2630 if (error) { 2631 vput(vp); 2632 nfsm_reply(NFSX_POSTOPATTR(v3)); 2633 nfsm_srvpostop_attr(getret, &at); 2634 return (0); 2635 } 2636 VOP_UNLOCK(vp, 0); 2637 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2638 again: 2639 iv.iov_base = rbuf; 2640 iv.iov_len = fullsiz; 2641 io.uio_iov = &iv; 2642 io.uio_iovcnt = 1; 2643 io.uio_offset = (off_t)off; 2644 io.uio_resid = fullsiz; 2645 io.uio_segflg = UIO_SYSSPACE; 2646 io.uio_rw = UIO_READ; 2647 io.uio_procp = (struct proc *)0; 2648 eofflag = 0; 2649 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2650 2651 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2652 2653 off = (off_t)io.uio_offset; 2654 if (!cookies && !error) 2655 error = NFSERR_PERM; 2656 if (v3) { 2657 getret = VOP_GETATTR(vp, &at, cred, procp); 2658 if (!error) 2659 error = getret; 2660 } 2661 2662 VOP_UNLOCK(vp, 0); 2663 if (error) { 2664 vrele(vp); 2665 free((caddr_t)rbuf, M_TEMP); 2666 if (cookies) 2667 free((caddr_t)cookies, M_TEMP); 2668 nfsm_reply(NFSX_POSTOPATTR(v3)); 2669 nfsm_srvpostop_attr(getret, &at); 2670 return (0); 2671 } 2672 if (io.uio_resid) { 2673 siz -= io.uio_resid; 2674 2675 /* 2676 * If nothing read, return eof 2677 * rpc reply 2678 */ 2679 if (siz == 0) { 2680 vrele(vp); 2681 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + 2682 2 * NFSX_UNSIGNED); 2683 if (v3) { 2684 nfsm_srvpostop_attr(getret, &at); 2685 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2686 txdr_hyper(at.va_filerev, tl); 2687 tl += 2; 2688 } else 2689 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2690 *tl++ = nfs_false; 2691 *tl = nfs_true; 2692 free((caddr_t)rbuf, M_TEMP); 2693 free((caddr_t)cookies, M_TEMP); 2694 return (0); 2695 } 2696 } 2697 2698 /* 2699 * Check for degenerate cases of nothing useful read. 2700 * If so go try again 2701 */ 2702 cpos = rbuf; 2703 cend = rbuf + siz; 2704 dp = (struct dirent *)cpos; 2705 cookiep = cookies; 2706 2707 while (cpos < cend && ncookies > 0 && 2708 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2709 cpos += dp->d_reclen; 2710 dp = (struct dirent *)cpos; 2711 cookiep++; 2712 ncookies--; 2713 } 2714 if (cpos >= cend || ncookies == 0) { 2715 toff = off; 2716 siz = fullsiz; 2717 goto again; 2718 } 2719 2720 len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */ 2721 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz); 2722 if (v3) { 2723 nfsm_srvpostop_attr(getret, &at); 2724 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2725 txdr_hyper(at.va_filerev, tl); 2726 } 2727 mp = mp2 = mb; 2728 bp = bpos; 2729 be = bp + M_TRAILINGSPACE(mp); 2730 2731 /* Loop through the records and build reply */ 2732 while (cpos < cend && ncookies > 0) { 2733 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 2734 nlen = dp->d_namlen; 2735 rem = nfsm_rndup(nlen)-nlen; 2736 len += (4 * NFSX_UNSIGNED + nlen + rem); 2737 if (v3) 2738 len += 2 * NFSX_UNSIGNED; 2739 if (len > cnt) { 2740 eofflag = 0; 2741 break; 2742 } 2743 /* 2744 * Build the directory record xdr from 2745 * the dirent entry. 2746 */ 2747 nfsm_clget; 2748 *tl = nfs_true; 2749 bp += NFSX_UNSIGNED; 2750 if (v3) { 2751 nfsm_clget; 2752 *tl = 0; 2753 bp += NFSX_UNSIGNED; 2754 } 2755 nfsm_clget; 2756 *tl = txdr_unsigned(dp->d_fileno); 2757 bp += NFSX_UNSIGNED; 2758 nfsm_clget; 2759 *tl = txdr_unsigned(nlen); 2760 bp += NFSX_UNSIGNED; 2761 2762 /* And loop around copying the name */ 2763 xfer = nlen; 2764 cp = dp->d_name; 2765 while (xfer > 0) { 2766 nfsm_clget; 2767 if ((bp+xfer) > be) 2768 tsiz = be-bp; 2769 else 2770 tsiz = xfer; 2771 memcpy(bp, cp, tsiz); 2772 bp += tsiz; 2773 xfer -= tsiz; 2774 if (xfer > 0) 2775 cp += tsiz; 2776 } 2777 /* And null pad to an int32_t boundary */ 2778 for (i = 0; i < rem; i++) 2779 *bp++ = '\0'; 2780 nfsm_clget; 2781 2782 /* Finish off the record */ 2783 txdr_hyper(*cookiep, &jar); 2784 if (v3) { 2785 *tl = jar.nfsuquad[0]; 2786 bp += NFSX_UNSIGNED; 2787 nfsm_clget; 2788 } 2789 *tl = jar.nfsuquad[1]; 2790 bp += NFSX_UNSIGNED; 2791 } 2792 cpos += dp->d_reclen; 2793 dp = (struct dirent *)cpos; 2794 cookiep++; 2795 ncookies--; 2796 } 2797 vrele(vp); 2798 nfsm_clget; 2799 *tl = nfs_false; 2800 bp += NFSX_UNSIGNED; 2801 nfsm_clget; 2802 if (eofflag) 2803 *tl = nfs_true; 2804 else 2805 *tl = nfs_false; 2806 bp += NFSX_UNSIGNED; 2807 if (mp != mb) { 2808 if (bp < be) 2809 mp->m_len = bp - mtod(mp, caddr_t); 2810 } else 2811 mp->m_len += bp - bpos; 2812 free((caddr_t)rbuf, M_TEMP); 2813 free((caddr_t)cookies, M_TEMP); 2814 nfsm_srvdone; 2815 } 2816 2817 int 2818 nfsrv_readdirplus(nfsd, slp, procp, mrq) 2819 struct nfsrv_descript *nfsd; 2820 struct nfssvc_sock *slp; 2821 struct proc *procp; 2822 struct mbuf **mrq; 2823 { 2824 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2825 struct mbuf *nam = nfsd->nd_nam; 2826 caddr_t dpos = nfsd->nd_dpos; 2827 struct ucred *cred = &nfsd->nd_cr; 2828 char *bp, *be; 2829 struct mbuf *mp; 2830 struct dirent *dp; 2831 caddr_t cp; 2832 u_int32_t *tl; 2833 int32_t t1; 2834 caddr_t bpos; 2835 struct mbuf *mb, *mreq, *mp2; 2836 char *cpos, *cend, *cp2, *rbuf; 2837 struct vnode *vp, *nvp; 2838 struct flrep fl; 2839 nfsfh_t nfh; 2840 fhandle_t *fhp, *nfhp = (fhandle_t *)fl.fl_nfh; 2841 struct uio io; 2842 struct iovec iv; 2843 struct vattr va, at, *vap = &va; 2844 struct nfs_fattr *fp; 2845 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2846 int siz, cnt, fullsiz, eofflag, rdonly, cache, dirlen, ncookies; 2847 u_quad_t frev, off, toff, verf; 2848 off_t *cookies = NULL, *cookiep; 2849 2850 fhp = &nfh.fh_generic; 2851 nfsm_srvmtofh(fhp); 2852 nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED); 2853 toff = fxdr_hyper(tl); 2854 tl += 2; 2855 verf = fxdr_hyper(tl); 2856 tl += 2; 2857 siz = fxdr_unsigned(int, *tl++); 2858 cnt = fxdr_unsigned(int, *tl); 2859 off = toff; 2860 siz = ((siz + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1)); 2861 xfer = NFS_SRVMAXDATA(nfsd); 2862 if (siz > xfer) 2863 siz = xfer; 2864 fullsiz = siz; 2865 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 2866 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2867 if (!error && vp->v_type != VDIR) { 2868 error = ENOTDIR; 2869 vput(vp); 2870 } 2871 if (error) { 2872 nfsm_reply(NFSX_UNSIGNED); 2873 nfsm_srvpostop_attr(getret, &at); 2874 return (0); 2875 } 2876 error = getret = VOP_GETATTR(vp, &at, cred, procp); 2877 #ifdef NFS3_STRICTVERF 2878 /* 2879 * XXX This check is too strict for Solaris 2.5 clients. 2880 */ 2881 if (!error && toff && verf != at.va_filerev) 2882 error = NFSERR_BAD_COOKIE; 2883 #endif 2884 if (!error) { 2885 nqsrv_getl(vp, ND_READ); 2886 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0); 2887 } 2888 if (error) { 2889 vput(vp); 2890 nfsm_reply(NFSX_V3POSTOPATTR); 2891 nfsm_srvpostop_attr(getret, &at); 2892 return (0); 2893 } 2894 VOP_UNLOCK(vp, 0); 2895 2896 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2897 again: 2898 iv.iov_base = rbuf; 2899 iv.iov_len = fullsiz; 2900 io.uio_iov = &iv; 2901 io.uio_iovcnt = 1; 2902 io.uio_offset = (off_t)off; 2903 io.uio_resid = fullsiz; 2904 io.uio_segflg = UIO_SYSSPACE; 2905 io.uio_rw = UIO_READ; 2906 io.uio_procp = (struct proc *)0; 2907 eofflag = 0; 2908 2909 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2910 2911 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2912 2913 off = (u_quad_t)io.uio_offset; 2914 getret = VOP_GETATTR(vp, &at, cred, procp); 2915 2916 VOP_UNLOCK(vp, 0); 2917 2918 /* 2919 * If the VGET operation doesn't work for this filesystem, 2920 * we can't support readdirplus. Returning NOTSUPP should 2921 * make clients fall back to plain readdir. 2922 * There's no need to check for VPTOFH as well, we wouldn't 2923 * even be here otherwise. 2924 */ 2925 if (!getret) { 2926 if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp))) 2927 getret = (getret == EOPNOTSUPP) ? 2928 NFSERR_NOTSUPP : NFSERR_IO; 2929 else 2930 vput(nvp); 2931 } 2932 2933 if (!cookies && !error) 2934 error = NFSERR_PERM; 2935 if (!error) 2936 error = getret; 2937 if (error) { 2938 vrele(vp); 2939 if (cookies) 2940 free((caddr_t)cookies, M_TEMP); 2941 free((caddr_t)rbuf, M_TEMP); 2942 nfsm_reply(NFSX_V3POSTOPATTR); 2943 nfsm_srvpostop_attr(getret, &at); 2944 return (0); 2945 } 2946 if (io.uio_resid) { 2947 siz -= io.uio_resid; 2948 2949 /* 2950 * If nothing read, return eof 2951 * rpc reply 2952 */ 2953 if (siz == 0) { 2954 vrele(vp); 2955 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2956 2 * NFSX_UNSIGNED); 2957 nfsm_srvpostop_attr(getret, &at); 2958 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2959 txdr_hyper(at.va_filerev, tl); 2960 tl += 2; 2961 *tl++ = nfs_false; 2962 *tl = nfs_true; 2963 free((caddr_t)cookies, M_TEMP); 2964 free((caddr_t)rbuf, M_TEMP); 2965 return (0); 2966 } 2967 } 2968 2969 /* 2970 * Check for degenerate cases of nothing useful read. 2971 * If so go try again 2972 */ 2973 cpos = rbuf; 2974 cend = rbuf + siz; 2975 dp = (struct dirent *)cpos; 2976 cookiep = cookies; 2977 2978 while (cpos < cend && ncookies > 0 && 2979 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2980 cpos += dp->d_reclen; 2981 dp = (struct dirent *)cpos; 2982 cookiep++; 2983 ncookies--; 2984 } 2985 if (cpos >= cend || ncookies == 0) { 2986 toff = off; 2987 siz = fullsiz; 2988 goto again; 2989 } 2990 2991 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED; 2992 nfsm_reply(cnt); 2993 nfsm_srvpostop_attr(getret, &at); 2994 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2995 txdr_hyper(at.va_filerev, tl); 2996 mp = mp2 = mb; 2997 bp = bpos; 2998 be = bp + M_TRAILINGSPACE(mp); 2999 3000 /* Loop through the records and build reply */ 3001 while (cpos < cend && ncookies > 0) { 3002 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 3003 nlen = dp->d_namlen; 3004 rem = nfsm_rndup(nlen)-nlen; 3005 3006 /* 3007 * For readdir_and_lookup get the vnode using 3008 * the file number. 3009 */ 3010 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp)) 3011 goto invalid; 3012 memset((caddr_t)nfhp, 0, NFSX_V3FH); 3013 nfhp->fh_fsid = 3014 nvp->v_mount->mnt_stat.f_fsid; 3015 if (VFS_VPTOFH(nvp, &nfhp->fh_fid)) { 3016 vput(nvp); 3017 goto invalid; 3018 } 3019 if (VOP_GETATTR(nvp, vap, cred, procp)) { 3020 vput(nvp); 3021 goto invalid; 3022 } 3023 vput(nvp); 3024 3025 /* 3026 * If either the dircount or maxcount will be 3027 * exceeded, get out now. Both of these lengths 3028 * are calculated conservatively, including all 3029 * XDR overheads. 3030 */ 3031 len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH + 3032 NFSX_V3POSTOPATTR); 3033 dirlen += (6 * NFSX_UNSIGNED + nlen + rem); 3034 if (len > cnt || dirlen > fullsiz) { 3035 eofflag = 0; 3036 break; 3037 } 3038 3039 /* 3040 * Build the directory record xdr from 3041 * the dirent entry. 3042 */ 3043 fp = (struct nfs_fattr *)&fl.fl_fattr; 3044 nfsm_srvfillattr(vap, fp); 3045 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH); 3046 fl.fl_fhok = nfs_true; 3047 fl.fl_postopok = nfs_true; 3048 txdr_hyper(*cookiep, fl.fl_off.nfsuquad); 3049 3050 nfsm_clget; 3051 *tl = nfs_true; 3052 bp += NFSX_UNSIGNED; 3053 nfsm_clget; 3054 *tl = 0; 3055 bp += NFSX_UNSIGNED; 3056 nfsm_clget; 3057 *tl = txdr_unsigned(dp->d_fileno); 3058 bp += NFSX_UNSIGNED; 3059 nfsm_clget; 3060 *tl = txdr_unsigned(nlen); 3061 bp += NFSX_UNSIGNED; 3062 3063 /* And loop around copying the name */ 3064 xfer = nlen; 3065 cp = dp->d_name; 3066 while (xfer > 0) { 3067 nfsm_clget; 3068 if ((bp + xfer) > be) 3069 tsiz = be - bp; 3070 else 3071 tsiz = xfer; 3072 memcpy(bp, cp, tsiz); 3073 bp += tsiz; 3074 xfer -= tsiz; 3075 if (xfer > 0) 3076 cp += tsiz; 3077 } 3078 /* And null pad to an int32_t boundary */ 3079 for (i = 0; i < rem; i++) 3080 *bp++ = '\0'; 3081 3082 /* 3083 * Now copy the flrep structure out. 3084 */ 3085 xfer = sizeof (struct flrep); 3086 cp = (caddr_t)&fl; 3087 while (xfer > 0) { 3088 nfsm_clget; 3089 if ((bp + xfer) > be) 3090 tsiz = be - bp; 3091 else 3092 tsiz = xfer; 3093 memcpy(bp, cp, tsiz); 3094 bp += tsiz; 3095 xfer -= tsiz; 3096 if (xfer > 0) 3097 cp += tsiz; 3098 } 3099 } 3100 invalid: 3101 cpos += dp->d_reclen; 3102 dp = (struct dirent *)cpos; 3103 cookiep++; 3104 ncookies--; 3105 } 3106 vrele(vp); 3107 nfsm_clget; 3108 *tl = nfs_false; 3109 bp += NFSX_UNSIGNED; 3110 nfsm_clget; 3111 if (eofflag) 3112 *tl = nfs_true; 3113 else 3114 *tl = nfs_false; 3115 bp += NFSX_UNSIGNED; 3116 if (mp != mb) { 3117 if (bp < be) 3118 mp->m_len = bp - mtod(mp, caddr_t); 3119 } else 3120 mp->m_len += bp - bpos; 3121 free((caddr_t)cookies, M_TEMP); 3122 free((caddr_t)rbuf, M_TEMP); 3123 nfsm_srvdone; 3124 } 3125 3126 /* 3127 * nfs commit service 3128 */ 3129 int 3130 nfsrv_commit(nfsd, slp, procp, mrq) 3131 struct nfsrv_descript *nfsd; 3132 struct nfssvc_sock *slp; 3133 struct proc *procp; 3134 struct mbuf **mrq; 3135 { 3136 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3137 struct mbuf *nam = nfsd->nd_nam; 3138 caddr_t dpos = nfsd->nd_dpos; 3139 struct ucred *cred = &nfsd->nd_cr; 3140 struct vattr bfor, aft; 3141 struct vnode *vp; 3142 nfsfh_t nfh; 3143 fhandle_t *fhp; 3144 u_int32_t *tl; 3145 int32_t t1; 3146 caddr_t bpos; 3147 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cnt, cache; 3148 char *cp2; 3149 struct mbuf *mb, *mreq; 3150 u_quad_t frev, off, end; 3151 struct mount *mp = NULL; 3152 3153 #ifndef nolint 3154 cache = 0; 3155 #endif 3156 fhp = &nfh.fh_generic; 3157 nfsm_srvmtofh(fhp); 3158 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 3159 return (ESTALE); 3160 vn_start_write(NULL, &mp, V_WAIT); 3161 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED); 3162 3163 off = fxdr_hyper(tl); 3164 tl += 2; 3165 cnt = fxdr_unsigned(int, *tl); 3166 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3167 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3168 if (error) { 3169 nfsm_reply(2 * NFSX_UNSIGNED); 3170 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3171 vn_finished_write(mp, 0); 3172 return (0); 3173 } 3174 for_ret = VOP_GETATTR(vp, &bfor, cred, procp); 3175 end = (cnt > 0) ? off + cnt : vp->v_size; 3176 if (end < off || end > vp->v_size) 3177 end = vp->v_size; 3178 if (off < vp->v_size) 3179 error = VOP_FSYNC(vp, cred, FSYNC_WAIT, off, end, procp); 3180 /* else error == 0, from nfsrv_fhtovp() */ 3181 aft_ret = VOP_GETATTR(vp, &aft, cred, procp); 3182 vput(vp); 3183 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF); 3184 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3185 if (!error) { 3186 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF); 3187 *tl++ = txdr_unsigned(boottime.tv_sec); 3188 *tl = txdr_unsigned(boottime.tv_usec); 3189 } else { 3190 vn_finished_write(mp, 0); 3191 return (0); 3192 } 3193 vn_finished_write(mp, 0); 3194 nfsm_srvdone; 3195 } 3196 3197 /* 3198 * nfs statfs service 3199 */ 3200 int 3201 nfsrv_statfs(nfsd, slp, procp, mrq) 3202 struct nfsrv_descript *nfsd; 3203 struct nfssvc_sock *slp; 3204 struct proc *procp; 3205 struct mbuf **mrq; 3206 { 3207 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3208 struct mbuf *nam = nfsd->nd_nam; 3209 caddr_t dpos = nfsd->nd_dpos; 3210 struct ucred *cred = &nfsd->nd_cr; 3211 struct statfs *sf; 3212 struct nfs_statfs *sfp; 3213 u_int32_t *tl; 3214 int32_t t1; 3215 caddr_t bpos; 3216 int error = 0, rdonly, cache, getret = 1; 3217 int v3 = (nfsd->nd_flag & ND_NFSV3); 3218 char *cp2; 3219 struct mbuf *mb, *mreq; 3220 struct vnode *vp; 3221 struct vattr at; 3222 nfsfh_t nfh; 3223 fhandle_t *fhp; 3224 struct statfs statfs; 3225 u_quad_t frev, tval; 3226 3227 #ifndef nolint 3228 cache = 0; 3229 #endif 3230 fhp = &nfh.fh_generic; 3231 nfsm_srvmtofh(fhp); 3232 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3233 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3234 if (error) { 3235 nfsm_reply(NFSX_UNSIGNED); 3236 nfsm_srvpostop_attr(getret, &at); 3237 return (0); 3238 } 3239 sf = &statfs; 3240 error = VFS_STATFS(vp->v_mount, sf, procp); 3241 getret = VOP_GETATTR(vp, &at, cred, procp); 3242 vput(vp); 3243 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3)); 3244 if (v3) 3245 nfsm_srvpostop_attr(getret, &at); 3246 if (error) 3247 return (0); 3248 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3)); 3249 if (v3) { 3250 tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_bsize); 3251 txdr_hyper(tval, &sfp->sf_tbytes); 3252 tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_bsize); 3253 txdr_hyper(tval, &sfp->sf_fbytes); 3254 tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_bsize); 3255 txdr_hyper(tval, &sfp->sf_abytes); 3256 tval = (u_quad_t)sf->f_files; 3257 txdr_hyper(tval, &sfp->sf_tfiles); 3258 tval = (u_quad_t)sf->f_ffree; 3259 txdr_hyper(tval, &sfp->sf_ffiles); 3260 txdr_hyper(tval, &sfp->sf_afiles); 3261 sfp->sf_invarsec = 0; 3262 } else { 3263 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA); 3264 sfp->sf_bsize = txdr_unsigned(sf->f_bsize); 3265 sfp->sf_blocks = txdr_unsigned(sf->f_blocks); 3266 sfp->sf_bfree = txdr_unsigned(sf->f_bfree); 3267 sfp->sf_bavail = txdr_unsigned(sf->f_bavail); 3268 } 3269 nfsm_srvdone; 3270 } 3271 3272 /* 3273 * nfs fsinfo service 3274 */ 3275 int 3276 nfsrv_fsinfo(nfsd, slp, procp, mrq) 3277 struct nfsrv_descript *nfsd; 3278 struct nfssvc_sock *slp; 3279 struct proc *procp; 3280 struct mbuf **mrq; 3281 { 3282 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3283 struct mbuf *nam = nfsd->nd_nam; 3284 caddr_t dpos = nfsd->nd_dpos; 3285 struct ucred *cred = &nfsd->nd_cr; 3286 u_int32_t *tl; 3287 struct nfsv3_fsinfo *sip; 3288 int32_t t1; 3289 caddr_t bpos; 3290 int error = 0, rdonly, cache, getret = 1; 3291 uint32_t maxdata; 3292 char *cp2; 3293 struct mbuf *mb, *mreq; 3294 struct vnode *vp; 3295 struct vattr at; 3296 nfsfh_t nfh; 3297 fhandle_t *fhp; 3298 u_quad_t frev, maxfsize; 3299 struct statfs sb; 3300 3301 #ifndef nolint 3302 cache = 0; 3303 #endif 3304 fhp = &nfh.fh_generic; 3305 nfsm_srvmtofh(fhp); 3306 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3307 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3308 if (error) { 3309 nfsm_reply(NFSX_UNSIGNED); 3310 nfsm_srvpostop_attr(getret, &at); 3311 return (0); 3312 } 3313 3314 /* XXX Try to make a guess on the max file size. */ 3315 VFS_STATFS(vp->v_mount, &sb, (struct proc *)0); 3316 maxfsize = (u_quad_t)0x80000000 * sb.f_bsize - 1; 3317 3318 getret = VOP_GETATTR(vp, &at, cred, procp); 3319 vput(vp); 3320 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO); 3321 nfsm_srvpostop_attr(getret, &at); 3322 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO); 3323 3324 /* 3325 * XXX 3326 * There should be file system VFS OP(s) to get this information. 3327 * For now, assume ufs. 3328 */ 3329 if (slp->ns_so->so_type == SOCK_DGRAM) 3330 maxdata = NFS_MAXDGRAMDATA; 3331 else 3332 maxdata = NFS_MAXDATA; 3333 sip->fs_rtmax = txdr_unsigned(maxdata); 3334 sip->fs_rtpref = txdr_unsigned(maxdata); 3335 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE); 3336 sip->fs_wtmax = txdr_unsigned(maxdata); 3337 sip->fs_wtpref = txdr_unsigned(maxdata); 3338 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE); 3339 sip->fs_dtpref = txdr_unsigned(maxdata); 3340 txdr_hyper(maxfsize, &sip->fs_maxfilesize); 3341 sip->fs_timedelta.nfsv3_sec = 0; 3342 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1); 3343 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK | 3344 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS | 3345 NFSV3FSINFO_CANSETTIME); 3346 nfsm_srvdone; 3347 } 3348 3349 /* 3350 * nfs pathconf service 3351 */ 3352 int 3353 nfsrv_pathconf(nfsd, slp, procp, mrq) 3354 struct nfsrv_descript *nfsd; 3355 struct nfssvc_sock *slp; 3356 struct proc *procp; 3357 struct mbuf **mrq; 3358 { 3359 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3360 struct mbuf *nam = nfsd->nd_nam; 3361 caddr_t dpos = nfsd->nd_dpos; 3362 struct ucred *cred = &nfsd->nd_cr; 3363 u_int32_t *tl; 3364 struct nfsv3_pathconf *pc; 3365 int32_t t1; 3366 caddr_t bpos; 3367 int error = 0, rdonly, cache, getret = 1; 3368 register_t linkmax, namemax, chownres, notrunc; 3369 char *cp2; 3370 struct mbuf *mb, *mreq; 3371 struct vnode *vp; 3372 struct vattr at; 3373 nfsfh_t nfh; 3374 fhandle_t *fhp; 3375 u_quad_t frev; 3376 3377 #ifndef nolint 3378 cache = 0; 3379 #endif 3380 fhp = &nfh.fh_generic; 3381 nfsm_srvmtofh(fhp); 3382 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3383 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3384 if (error) { 3385 nfsm_reply(NFSX_UNSIGNED); 3386 nfsm_srvpostop_attr(getret, &at); 3387 return (0); 3388 } 3389 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax); 3390 if (!error) 3391 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax); 3392 if (!error) 3393 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres); 3394 if (!error) 3395 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc); 3396 getret = VOP_GETATTR(vp, &at, cred, procp); 3397 vput(vp); 3398 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF); 3399 nfsm_srvpostop_attr(getret, &at); 3400 if (error) 3401 return (0); 3402 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF); 3403 3404 pc->pc_linkmax = txdr_unsigned(linkmax); 3405 pc->pc_namemax = txdr_unsigned(namemax); 3406 pc->pc_notrunc = txdr_unsigned(notrunc); 3407 pc->pc_chownrestricted = txdr_unsigned(chownres); 3408 3409 /* 3410 * These should probably be supported by VOP_PATHCONF(), but 3411 * until msdosfs is exportable (why would you want to?), the 3412 * Unix defaults should be ok. 3413 */ 3414 pc->pc_caseinsensitive = nfs_false; 3415 pc->pc_casepreserving = nfs_true; 3416 nfsm_srvdone; 3417 } 3418 3419 /* 3420 * Null operation, used by clients to ping server 3421 */ 3422 /* ARGSUSED */ 3423 int 3424 nfsrv_null(nfsd, slp, procp, mrq) 3425 struct nfsrv_descript *nfsd; 3426 struct nfssvc_sock *slp; 3427 struct proc *procp; 3428 struct mbuf **mrq; 3429 { 3430 struct mbuf *mrep = nfsd->nd_mrep; 3431 caddr_t bpos; 3432 int error = NFSERR_RETVOID, cache = 0; 3433 struct mbuf *mb, *mreq; 3434 u_quad_t frev; 3435 3436 nfsm_reply(0); 3437 return (0); 3438 } 3439 3440 /* 3441 * No operation, used for obsolete procedures 3442 */ 3443 /* ARGSUSED */ 3444 int 3445 nfsrv_noop(nfsd, slp, procp, mrq) 3446 struct nfsrv_descript *nfsd; 3447 struct nfssvc_sock *slp; 3448 struct proc *procp; 3449 struct mbuf **mrq; 3450 { 3451 struct mbuf *mrep = nfsd->nd_mrep; 3452 caddr_t bpos; 3453 int error, cache = 0; 3454 struct mbuf *mb, *mreq; 3455 u_quad_t frev; 3456 3457 if (nfsd->nd_repstat) 3458 error = nfsd->nd_repstat; 3459 else 3460 error = EPROCUNAVAIL; 3461 nfsm_reply(0); 3462 return (0); 3463 } 3464 3465 /* 3466 * Perform access checking for vnodes obtained from file handles that would 3467 * refer to files already opened by a Unix client. You cannot just use 3468 * vn_writechk() and VOP_ACCESS() for two reasons. 3469 * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case 3470 * 2 - The owner is to be given access irrespective of mode bits for some 3471 * operations, so that processes that chmod after opening a file don't 3472 * break. I don't like this because it opens a security hole, but since 3473 * the nfs server opens a security hole the size of a barn door anyhow, 3474 * what the heck. 3475 * 3476 * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS() 3477 * will return EPERM instead of EACCESS. EPERM is always an error. 3478 */ 3479 int 3480 nfsrv_access(vp, flags, cred, rdonly, p, override) 3481 struct vnode *vp; 3482 int flags; 3483 struct ucred *cred; 3484 int rdonly; 3485 struct proc *p; 3486 { 3487 struct vattr vattr; 3488 int error; 3489 if (flags & VWRITE) { 3490 /* Just vn_writechk() changed to check rdonly */ 3491 /* 3492 * Disallow write attempts on read-only file systems; 3493 * unless the file is a socket or a block or character 3494 * device resident on the file system. 3495 */ 3496 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 3497 switch (vp->v_type) { 3498 case VREG: 3499 case VDIR: 3500 case VLNK: 3501 return (EROFS); 3502 default: 3503 break; 3504 } 3505 } 3506 3507 /* 3508 * If the vnode is in use as a process's text, 3509 * we can't allow writing. 3510 */ 3511 if (vp->v_flag & VTEXT) 3512 return (ETXTBSY); 3513 } 3514 error = VOP_GETATTR(vp, &vattr, cred, p); 3515 if (error) 3516 return (error); 3517 error = VOP_ACCESS(vp, flags, cred, p); 3518 /* 3519 * Allow certain operations for the owner (reads and writes 3520 * on files that are already open). 3521 */ 3522 if (override && error == EACCES && cred->cr_uid == vattr.va_uid) 3523 error = 0; 3524 return error; 3525 } 3526