nfs_serv.c revision 1.100.4.1
1 /* $NetBSD: nfs_serv.c,v 1.100.4.1 2006/02/04 14:12:50 simonb 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 1.100.4.1 2006/02/04 14:12:50 simonb 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 74 #include <uvm/uvm.h> 75 76 #include <nfs/nfsproto.h> 77 #include <nfs/rpcv2.h> 78 #include <nfs/nfs.h> 79 #include <nfs/xdr_subs.h> 80 #include <nfs/nfsm_subs.h> 81 #include <nfs/nqnfs.h> 82 #include <nfs/nfs_var.h> 83 84 /* Global vars */ 85 extern u_int32_t nfs_xdrneg1; 86 extern u_int32_t nfs_false, nfs_true; 87 extern enum vtype nv3tov_type[8]; 88 extern struct nfsstats nfsstats; 89 extern const nfstype nfsv2_type[9]; 90 extern const nfstype nfsv3_type[9]; 91 int nfsrvw_procrastinate = NFS_GATHERDELAY * 1000; 92 int nfsd_use_loan = 1; /* use page-loan for READ OP */ 93 94 /* 95 * nfs v3 access service 96 */ 97 int 98 nfsrv3_access(nfsd, slp, lwp, mrq) 99 struct nfsrv_descript *nfsd; 100 struct nfssvc_sock *slp; 101 struct lwp *lwp; 102 struct mbuf **mrq; 103 { 104 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 105 struct mbuf *nam = nfsd->nd_nam; 106 caddr_t dpos = nfsd->nd_dpos; 107 struct ucred *cred = &nfsd->nd_cr; 108 struct vnode *vp; 109 nfsfh_t nfh; 110 fhandle_t *fhp; 111 u_int32_t *tl; 112 int32_t t1; 113 caddr_t bpos; 114 int error = 0, rdonly, cache = 0, getret; 115 char *cp2; 116 struct mbuf *mb, *mreq; 117 struct vattr va; 118 u_long inmode, testmode, outmode; 119 u_quad_t frev; 120 121 fhp = &nfh.fh_generic; 122 nfsm_srvmtofh(fhp); 123 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 124 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, 125 (nfsd->nd_flag & ND_KERBAUTH), FALSE); 126 if (error) { 127 nfsm_reply(NFSX_UNSIGNED); 128 nfsm_srvpostop_attr(1, (struct vattr *)0); 129 return (0); 130 } 131 inmode = fxdr_unsigned(u_int32_t, *tl); 132 outmode = 0; 133 if ((inmode & NFSV3ACCESS_READ) && 134 nfsrv_access(vp, VREAD, cred, rdonly, lwp, 0) == 0) 135 outmode |= NFSV3ACCESS_READ; 136 if (vp->v_type != VDIR) { 137 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND); 138 if (testmode && 139 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0) 140 outmode |= testmode; 141 if ((inmode & NFSV3ACCESS_EXECUTE) && 142 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0) 143 outmode |= NFSV3ACCESS_EXECUTE; 144 } else { 145 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND | 146 NFSV3ACCESS_DELETE); 147 if (testmode && 148 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0) 149 outmode |= testmode; 150 if ((inmode & NFSV3ACCESS_LOOKUP) && 151 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0) 152 outmode |= NFSV3ACCESS_LOOKUP; 153 } 154 getret = VOP_GETATTR(vp, &va, cred, lwp); 155 vput(vp); 156 nfsm_reply(NFSX_POSTOPATTR(1) + NFSX_UNSIGNED); 157 nfsm_srvpostop_attr(getret, &va); 158 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 159 *tl = txdr_unsigned(outmode); 160 nfsm_srvdone; 161 } 162 163 /* 164 * nfs getattr service 165 */ 166 int 167 nfsrv_getattr(nfsd, slp, lwp, mrq) 168 struct nfsrv_descript *nfsd; 169 struct nfssvc_sock *slp; 170 struct lwp *lwp; 171 struct mbuf **mrq; 172 { 173 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 174 struct mbuf *nam = nfsd->nd_nam; 175 caddr_t dpos = nfsd->nd_dpos; 176 struct ucred *cred = &nfsd->nd_cr; 177 struct nfs_fattr *fp; 178 struct vattr va; 179 struct vnode *vp; 180 nfsfh_t nfh; 181 fhandle_t *fhp; 182 u_int32_t *tl; 183 int32_t t1; 184 caddr_t bpos; 185 int error = 0, rdonly, cache; 186 char *cp2; 187 struct mbuf *mb, *mreq; 188 u_quad_t frev; 189 190 fhp = &nfh.fh_generic; 191 nfsm_srvmtofh(fhp); 192 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, 193 (nfsd->nd_flag & ND_KERBAUTH), FALSE); 194 if (error) { 195 nfsm_reply(0); 196 return (0); 197 } 198 nqsrv_getl(vp, ND_READ); 199 error = VOP_GETATTR(vp, &va, cred, lwp); 200 vput(vp); 201 nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); 202 if (error) 203 return (0); 204 nfsm_build(fp, struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); 205 nfsm_srvfillattr(&va, fp); 206 nfsm_srvdone; 207 } 208 209 /* 210 * nfs setattr service 211 */ 212 int 213 nfsrv_setattr(nfsd, slp, lwp, mrq) 214 struct nfsrv_descript *nfsd; 215 struct nfssvc_sock *slp; 216 struct lwp *lwp; 217 struct mbuf **mrq; 218 { 219 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 220 struct mbuf *nam = nfsd->nd_nam; 221 caddr_t dpos = nfsd->nd_dpos; 222 struct ucred *cred = &nfsd->nd_cr; 223 struct vattr va, preat; 224 struct nfsv2_sattr *sp; 225 struct nfs_fattr *fp; 226 struct vnode *vp; 227 nfsfh_t nfh; 228 fhandle_t *fhp; 229 u_int32_t *tl; 230 int32_t t1; 231 caddr_t bpos; 232 int error = 0, rdonly, cache, preat_ret = 1, postat_ret = 1; 233 int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0; 234 char *cp2; 235 struct mbuf *mb, *mreq; 236 u_quad_t frev; 237 struct timespec guard; 238 struct mount *mp = NULL; 239 240 fhp = &nfh.fh_generic; 241 nfsm_srvmtofh(fhp); 242 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 243 return (ESTALE); 244 vn_start_write(NULL, &mp, V_WAIT); 245 VATTR_NULL(&va); 246 if (v3) { 247 nfsm_srvsattr(&va); 248 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 249 gcheck = fxdr_unsigned(int, *tl); 250 if (gcheck) { 251 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 252 fxdr_nfsv3time(tl, &guard); 253 } 254 } else { 255 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 256 /* 257 * Nah nah nah nah na nah 258 * There is a bug in the Sun client that puts 0xffff in the mode 259 * field of sattr when it should put in 0xffffffff. The u_short 260 * doesn't sign extend. 261 * --> check the low order 2 bytes for 0xffff 262 */ 263 if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff) 264 va.va_mode = nfstov_mode(sp->sa_mode); 265 if (sp->sa_uid != nfs_xdrneg1) 266 va.va_uid = fxdr_unsigned(uid_t, sp->sa_uid); 267 if (sp->sa_gid != nfs_xdrneg1) 268 va.va_gid = fxdr_unsigned(gid_t, sp->sa_gid); 269 if (sp->sa_size != nfs_xdrneg1) 270 va.va_size = fxdr_unsigned(u_quad_t, sp->sa_size); 271 if (sp->sa_atime.nfsv2_sec != nfs_xdrneg1) { 272 #ifdef notyet 273 fxdr_nfsv2time(&sp->sa_atime, &va.va_atime); 274 #else 275 va.va_atime.tv_sec = 276 fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec); 277 va.va_atime.tv_nsec = 0; 278 #endif 279 } 280 if (sp->sa_mtime.nfsv2_sec != nfs_xdrneg1) 281 fxdr_nfsv2time(&sp->sa_mtime, &va.va_mtime); 282 283 } 284 285 /* 286 * Now that we have all the fields, lets do it. 287 */ 288 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, 289 (nfsd->nd_flag & ND_KERBAUTH), FALSE); 290 if (error) { 291 nfsm_reply(2 * NFSX_UNSIGNED); 292 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 293 vn_finished_write(mp, 0); 294 return (0); 295 } 296 nqsrv_getl(vp, ND_WRITE); 297 if (v3) { 298 error = preat_ret = VOP_GETATTR(vp, &preat, cred, lwp); 299 if (!error && gcheck && 300 (preat.va_ctime.tv_sec != guard.tv_sec || 301 preat.va_ctime.tv_nsec != guard.tv_nsec)) 302 error = NFSERR_NOT_SYNC; 303 if (error) { 304 vput(vp); 305 nfsm_reply(NFSX_WCCDATA(v3)); 306 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 307 vn_finished_write(mp, 0); 308 return (0); 309 } 310 } 311 312 /* 313 * If the size is being changed write acces is required, otherwise 314 * just check for a read only file system. 315 */ 316 if (va.va_size == ((u_quad_t)((quad_t) -1))) { 317 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 318 error = EROFS; 319 goto out; 320 } 321 } else { 322 if (vp->v_type == VDIR) { 323 error = EISDIR; 324 goto out; 325 } else if ((error = nfsrv_access(vp, VWRITE, cred, rdonly, 326 lwp, 0)) != 0) 327 goto out; 328 } 329 error = VOP_SETATTR(vp, &va, cred, lwp); 330 postat_ret = VOP_GETATTR(vp, &va, cred, lwp); 331 if (!error) 332 error = postat_ret; 333 out: 334 vput(vp); 335 vn_finished_write(mp, 0); 336 nfsm_reply(NFSX_WCCORFATTR(v3)); 337 if (v3) { 338 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 339 return (0); 340 } else { 341 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 342 nfsm_srvfillattr(&va, fp); 343 } 344 nfsm_srvdone; 345 } 346 347 /* 348 * nfs lookup rpc 349 */ 350 int 351 nfsrv_lookup(nfsd, slp, lwp, mrq) 352 struct nfsrv_descript *nfsd; 353 struct nfssvc_sock *slp; 354 struct lwp *lwp; 355 struct mbuf **mrq; 356 { 357 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 358 struct mbuf *nam = nfsd->nd_nam; 359 caddr_t dpos = nfsd->nd_dpos; 360 struct ucred *cred = &nfsd->nd_cr; 361 struct nfs_fattr *fp; 362 struct nameidata nd, ind, *ndp = &nd; 363 struct vnode *vp, *dirp; 364 nfsfh_t nfh; 365 fhandle_t *fhp; 366 caddr_t cp; 367 u_int32_t *tl; 368 int32_t t1; 369 caddr_t bpos; 370 int error = 0, cache, dirattr_ret = 1; 371 uint32_t len; 372 int v3 = (nfsd->nd_flag & ND_NFSV3), pubflag; 373 char *cp2; 374 struct mbuf *mb, *mreq; 375 struct vattr va, dirattr; 376 u_quad_t frev; 377 378 fhp = &nfh.fh_generic; 379 nfsm_srvmtofh(fhp); 380 nfsm_srvnamesiz(len); 381 382 pubflag = nfs_ispublicfh(fhp); 383 384 nd.ni_cnd.cn_cred = cred; 385 nd.ni_cnd.cn_nameiop = LOOKUP; 386 nd.ni_cnd.cn_flags = LOCKLEAF | SAVESTART; 387 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 388 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), pubflag); 389 390 if (!error && pubflag) { 391 if (nd.ni_vp->v_type == VDIR && nfs_pub.np_index != NULL) { 392 /* 393 * Setup call to lookup() to see if we can find 394 * the index file. Arguably, this doesn't belong 395 * in a kernel.. Ugh. 396 */ 397 ind = nd; 398 VOP_UNLOCK(nd.ni_vp, 0); 399 ind.ni_pathlen = strlen(nfs_pub.np_index); 400 ind.ni_cnd.cn_nameptr = ind.ni_cnd.cn_pnbuf = 401 nfs_pub.np_index; 402 ind.ni_startdir = nd.ni_vp; 403 VREF(ind.ni_startdir); 404 error = lookup(&ind); 405 if (!error) { 406 /* 407 * Found an index file. Get rid of 408 * the old references. 409 */ 410 if (dirp) 411 vrele(dirp); 412 dirp = nd.ni_vp; 413 vrele(nd.ni_startdir); 414 ndp = &ind; 415 } else 416 error = 0; 417 } 418 /* 419 * If the public filehandle was used, check that this lookup 420 * didn't result in a filehandle outside the publicly exported 421 * filesystem. 422 */ 423 424 if (!error && ndp->ni_vp->v_mount != nfs_pub.np_mount) { 425 vput(nd.ni_vp); 426 error = EPERM; 427 } 428 } 429 430 if (dirp) { 431 if (v3) 432 dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred, lwp); 433 vrele(dirp); 434 } 435 436 if (error) { 437 nfsm_reply(NFSX_POSTOPATTR(v3)); 438 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 439 return (0); 440 } 441 442 nqsrv_getl(ndp->ni_startdir, ND_READ); 443 vrele(ndp->ni_startdir); 444 PNBUF_PUT(nd.ni_cnd.cn_pnbuf); 445 vp = ndp->ni_vp; 446 memset((caddr_t)fhp, 0, sizeof(nfh)); 447 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 448 error = VFS_VPTOFH(vp, &fhp->fh_fid); 449 if (!error) 450 error = VOP_GETATTR(vp, &va, cred, lwp); 451 vput(vp); 452 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPORFATTR(v3) + NFSX_POSTOPATTR(v3)); 453 if (error) { 454 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 455 return (0); 456 } 457 nfsm_srvfhtom(fhp, v3); 458 if (v3) { 459 nfsm_srvpostop_attr(0, &va); 460 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 461 } else { 462 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 463 nfsm_srvfillattr(&va, fp); 464 } 465 nfsm_srvdone; 466 } 467 468 /* 469 * nfs readlink service 470 */ 471 int 472 nfsrv_readlink(nfsd, slp, lwp, mrq) 473 struct nfsrv_descript *nfsd; 474 struct nfssvc_sock *slp; 475 struct lwp *lwp; 476 struct mbuf **mrq; 477 { 478 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 479 struct mbuf *nam = nfsd->nd_nam; 480 caddr_t dpos = nfsd->nd_dpos; 481 struct ucred *cred = &nfsd->nd_cr; 482 struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN]; 483 struct iovec *ivp = iv; 484 struct mbuf *mp; 485 u_int32_t *tl; 486 int32_t t1; 487 caddr_t bpos; 488 int error = 0, rdonly, cache, i, padlen, getret; 489 uint32_t len; 490 int v3 = (nfsd->nd_flag & ND_NFSV3); 491 char *cp2; 492 struct mbuf *mb, *mp2 = NULL, *mp3 = NULL, *mreq; 493 struct vnode *vp; 494 struct vattr attr; 495 nfsfh_t nfh; 496 fhandle_t *fhp; 497 struct uio io, *uiop = &io; 498 u_quad_t frev; 499 500 fhp = &nfh.fh_generic; 501 nfsm_srvmtofh(fhp); 502 len = 0; 503 i = 0; 504 while (len < NFS_MAXPATHLEN) { 505 mp = m_get(M_WAIT, MT_DATA); 506 MCLAIM(mp, &nfs_mowner); 507 m_clget(mp, M_WAIT); 508 mp->m_len = NFSMSIZ(mp); 509 if (len == 0) 510 mp3 = mp2 = mp; 511 else { 512 mp2->m_next = mp; 513 mp2 = mp; 514 } 515 if ((len+mp->m_len) > NFS_MAXPATHLEN) { 516 mp->m_len = NFS_MAXPATHLEN-len; 517 len = NFS_MAXPATHLEN; 518 } else 519 len += mp->m_len; 520 ivp->iov_base = mtod(mp, caddr_t); 521 ivp->iov_len = mp->m_len; 522 i++; 523 ivp++; 524 } 525 uiop->uio_iov = iv; 526 uiop->uio_iovcnt = i; 527 uiop->uio_offset = 0; 528 uiop->uio_resid = len; 529 uiop->uio_rw = UIO_READ; 530 uiop->uio_segflg = UIO_SYSSPACE; 531 uiop->uio_lwp = NULL; 532 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 533 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 534 if (error) { 535 m_freem(mp3); 536 nfsm_reply(2 * NFSX_UNSIGNED); 537 nfsm_srvpostop_attr(1, (struct vattr *)0); 538 return (0); 539 } 540 if (vp->v_type != VLNK) { 541 if (v3) 542 error = EINVAL; 543 else 544 error = ENXIO; 545 goto out; 546 } 547 nqsrv_getl(vp, ND_READ); 548 error = VOP_READLINK(vp, uiop, cred); 549 out: 550 getret = VOP_GETATTR(vp, &attr, cred, lwp); 551 vput(vp); 552 if (error) 553 m_freem(mp3); 554 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED); 555 if (v3) { 556 nfsm_srvpostop_attr(getret, &attr); 557 if (error) 558 return (0); 559 } 560 len -= uiop->uio_resid; 561 padlen = nfsm_padlen(len); 562 if (uiop->uio_resid || padlen) 563 nfs_zeropad(mp3, uiop->uio_resid, padlen); 564 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 565 *tl = txdr_unsigned(len); 566 mb->m_next = mp3; 567 nfsm_srvdone; 568 } 569 570 /* 571 * nfs read service 572 */ 573 int 574 nfsrv_read(nfsd, slp, lwp, mrq) 575 struct nfsrv_descript *nfsd; 576 struct nfssvc_sock *slp; 577 struct lwp *lwp; 578 struct mbuf **mrq; 579 { 580 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 581 struct mbuf *nam = nfsd->nd_nam; 582 caddr_t dpos = nfsd->nd_dpos; 583 struct ucred *cred = &nfsd->nd_cr; 584 struct mbuf *m; 585 struct nfs_fattr *fp; 586 u_int32_t *tl; 587 int32_t t1; 588 int i; 589 caddr_t bpos; 590 int error = 0, rdonly, cache, getret; 591 int v3 = (nfsd->nd_flag & ND_NFSV3); 592 uint32_t reqlen, len, cnt, left; 593 int padlen; 594 char *cp2; 595 struct mbuf *mb, *mreq; 596 struct vnode *vp; 597 nfsfh_t nfh; 598 fhandle_t *fhp; 599 struct uio io, *uiop = &io; 600 struct vattr va; 601 off_t off; 602 u_quad_t frev; 603 604 fhp = &nfh.fh_generic; 605 nfsm_srvmtofh(fhp); 606 if (v3) { 607 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 608 off = fxdr_hyper(tl); 609 } else { 610 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 611 off = (off_t)fxdr_unsigned(u_int32_t, *tl); 612 } 613 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 614 reqlen = fxdr_unsigned(uint32_t, *tl); 615 reqlen = MIN(reqlen, NFS_SRVMAXDATA(nfsd)); 616 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 617 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 618 if (error) { 619 nfsm_reply(2 * NFSX_UNSIGNED); 620 nfsm_srvpostop_attr(1, (struct vattr *)0); 621 return (0); 622 } 623 if (vp->v_type != VREG) { 624 if (v3) 625 error = EINVAL; 626 else 627 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 628 } 629 if (!error) { 630 nqsrv_getl(vp, ND_READ); 631 if ((error = nfsrv_access(vp, VREAD, cred, rdonly, lwp, 1)) != 0) 632 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 1); 633 } 634 getret = VOP_GETATTR(vp, &va, cred, lwp); 635 if (!error) 636 error = getret; 637 if (error) { 638 vput(vp); 639 nfsm_reply(NFSX_POSTOPATTR(v3)); 640 nfsm_srvpostop_attr(getret, &va); 641 return (0); 642 } 643 if (off >= va.va_size) 644 cnt = 0; 645 else if ((off + reqlen) > va.va_size) 646 cnt = va.va_size - off; 647 else 648 cnt = reqlen; 649 nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt)); 650 if (v3) { 651 nfsm_build(tl, u_int32_t *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED); 652 *tl++ = nfs_true; 653 fp = (struct nfs_fattr *)tl; 654 tl += (NFSX_V3FATTR / sizeof (u_int32_t)); 655 } else { 656 nfsm_build(tl, u_int32_t *, NFSX_V2FATTR + NFSX_UNSIGNED); 657 fp = (struct nfs_fattr *)tl; 658 tl += (NFSX_V2FATTR / sizeof (u_int32_t)); 659 } 660 len = left = cnt; 661 if (cnt > 0) { 662 if (nfsd_use_loan) { 663 struct vm_page **pgpp; 664 voff_t pgoff = trunc_page(off); 665 int npages; 666 vaddr_t lva; 667 668 npages = (round_page(off + cnt) - pgoff) >> PAGE_SHIFT; 669 KASSERT(npages <= M_EXT_MAXPAGES); /* XXX */ 670 671 /* allocate kva for mbuf data */ 672 lva = sokvaalloc(npages << PAGE_SHIFT, slp->ns_so); 673 if (lva == 0) { 674 /* fall back to VOP_READ */ 675 goto loan_fail; 676 } 677 678 /* allocate mbuf */ 679 m = m_get(M_WAIT, MT_DATA); 680 MCLAIM(m, &nfs_mowner); 681 pgpp = m->m_ext.ext_pgs; 682 683 /* loan pages */ 684 error = uvm_loanuobjpages(&vp->v_uobj, pgoff, npages, 685 pgpp); 686 if (error) { 687 sokvafree(lva, npages << PAGE_SHIFT); 688 m_free(m); 689 if (error == EBUSY) 690 goto loan_fail; 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, lwp)) != 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, lwp, mrq) 804 struct nfsrv_descript *nfsd; 805 struct nfssvc_sock *slp; 806 struct lwp *lwp; 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, lwp); 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, lwp, 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_lwp = NULL; 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, lwp); 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, lwp, mrq) 999 struct nfsrv_descript **ndp; 1000 struct nfssvc_sock *slp; 1001 struct lwp *lwp; 1002 struct mbuf **mrq; 1003 { 1004 struct timeval now; 1005 struct iovec *ivp; 1006 struct mbuf *mp; 1007 struct nfsrv_descript *wp, *nfsd, *owp, *swp; 1008 struct nfs_fattr *fp; 1009 int i = 0; 1010 struct iovec *iov; 1011 struct nfsrvw_delayhash *wpp; 1012 struct ucred *cred; 1013 struct vattr va, forat; 1014 u_int32_t *tl; 1015 int32_t t1; 1016 caddr_t bpos, dpos; 1017 int error = 0, rdonly, cache, len = 0, forat_ret = 1; 1018 int ioflags, aftat_ret = 1, s, adjust, v3, zeroing; 1019 char *cp2; 1020 struct mbuf *mb, *mreq, *mrep, *md; 1021 struct vnode *vp; 1022 struct uio io, *uiop = &io; 1023 u_quad_t frev, cur_usec; 1024 struct mount *mntp = NULL; 1025 1026 *mrq = NULL; 1027 if (*ndp) { 1028 nfsd = *ndp; 1029 *ndp = NULL; 1030 mrep = nfsd->nd_mrep; 1031 md = nfsd->nd_md; 1032 dpos = nfsd->nd_dpos; 1033 cred = &nfsd->nd_cr; 1034 v3 = (nfsd->nd_flag & ND_NFSV3); 1035 LIST_INIT(&nfsd->nd_coalesce); 1036 nfsd->nd_mreq = NULL; 1037 nfsd->nd_stable = NFSV3WRITE_FILESYNC; 1038 getmicrotime(&now); 1039 cur_usec = (u_quad_t)now.tv_sec * 1000000 + (u_quad_t)now.tv_usec; 1040 nfsd->nd_time = cur_usec + nfsrvw_procrastinate; 1041 1042 /* 1043 * Now, get the write header.. 1044 */ 1045 nfsm_srvmtofh(&nfsd->nd_fh); 1046 if (v3) { 1047 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 1048 nfsd->nd_off = fxdr_hyper(tl); 1049 tl += 3; 1050 nfsd->nd_stable = fxdr_unsigned(int, *tl++); 1051 } else { 1052 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1053 nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 1054 tl += 2; 1055 } 1056 len = fxdr_unsigned(int32_t, *tl); 1057 nfsd->nd_len = len; 1058 nfsd->nd_eoff = nfsd->nd_off + len; 1059 1060 /* 1061 * Trim the header out of the mbuf list and trim off any trailing 1062 * junk so that the mbuf list has only the write data. 1063 */ 1064 zeroing = 1; 1065 i = 0; 1066 mp = mrep; 1067 while (mp) { 1068 if (mp == md) { 1069 zeroing = 0; 1070 adjust = dpos - mtod(mp, caddr_t); 1071 mp->m_len -= adjust; 1072 if (mp->m_len > 0 && adjust > 0) 1073 NFSMADV(mp, adjust); 1074 } 1075 if (zeroing) 1076 mp->m_len = 0; 1077 else { 1078 i += mp->m_len; 1079 if (i > len) { 1080 mp->m_len -= (i - len); 1081 zeroing = 1; 1082 } 1083 } 1084 mp = mp->m_next; 1085 } 1086 if (len > NFS_MAXDATA || len < 0 || i < len) { 1087 nfsmout: 1088 m_freem(mrep); 1089 error = EIO; 1090 nfsm_writereply(2 * NFSX_UNSIGNED, v3); 1091 if (v3) 1092 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1093 nfsd->nd_mreq = mreq; 1094 nfsd->nd_mrep = NULL; 1095 nfsd->nd_time = 0; 1096 } 1097 1098 /* 1099 * Add this entry to the hash and time queues. 1100 */ 1101 s = splsoftclock(); 1102 owp = NULL; 1103 wp = LIST_FIRST(&slp->ns_tq); 1104 while (wp && wp->nd_time < nfsd->nd_time) { 1105 owp = wp; 1106 wp = LIST_NEXT(wp, nd_tq); 1107 } 1108 if (owp) { 1109 LIST_INSERT_AFTER(owp, nfsd, nd_tq); 1110 } else { 1111 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1112 } 1113 if (nfsd->nd_mrep) { 1114 wpp = NWDELAYHASH(slp, nfsd->nd_fh.fh_fid.fid_data); 1115 owp = NULL; 1116 wp = LIST_FIRST(wpp); 1117 while (wp && 1118 memcmp(&nfsd->nd_fh, &wp->nd_fh, NFSX_V3FH)) { 1119 owp = wp; 1120 wp = LIST_NEXT(wp, nd_hash); 1121 } 1122 while (wp && wp->nd_off < nfsd->nd_off && 1123 !memcmp(&nfsd->nd_fh, &wp->nd_fh, NFSX_V3FH)) { 1124 owp = wp; 1125 wp = LIST_NEXT(wp, nd_hash); 1126 } 1127 if (owp) { 1128 LIST_INSERT_AFTER(owp, nfsd, nd_hash); 1129 1130 /* 1131 * Search the hash list for overlapping entries and 1132 * coalesce. 1133 */ 1134 for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) { 1135 wp = LIST_NEXT(nfsd, nd_hash); 1136 if (NFSW_SAMECRED(owp, nfsd)) 1137 nfsrvw_coalesce(owp, nfsd); 1138 } 1139 } else { 1140 LIST_INSERT_HEAD(wpp, nfsd, nd_hash); 1141 } 1142 } 1143 splx(s); 1144 } 1145 1146 /* 1147 * Now, do VOP_WRITE()s for any one(s) that need to be done now 1148 * and generate the associated reply mbuf list(s). 1149 */ 1150 loop1: 1151 getmicrotime(&now); 1152 cur_usec = (u_quad_t)now.tv_sec * 1000000 + (u_quad_t)now.tv_usec; 1153 s = splsoftclock(); 1154 for (nfsd = LIST_FIRST(&slp->ns_tq); nfsd; nfsd = owp) { 1155 owp = LIST_NEXT(nfsd, nd_tq); 1156 if (nfsd->nd_time > cur_usec) 1157 break; 1158 if (nfsd->nd_mreq) 1159 continue; 1160 LIST_REMOVE(nfsd, nd_tq); 1161 LIST_REMOVE(nfsd, nd_hash); 1162 splx(s); 1163 mrep = nfsd->nd_mrep; 1164 nfsd->nd_mrep = NULL; 1165 cred = &nfsd->nd_cr; 1166 v3 = (nfsd->nd_flag & ND_NFSV3); 1167 forat_ret = aftat_ret = 1; 1168 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp, 1169 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH), 1170 FALSE); 1171 if (!error) { 1172 if (v3) 1173 forat_ret = VOP_GETATTR(vp, &forat, cred, lwp); 1174 if (vp->v_type != VREG) { 1175 if (v3) 1176 error = EINVAL; 1177 else 1178 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 1179 } 1180 } else 1181 vp = NULL; 1182 if (!error) { 1183 nqsrv_getl(vp, ND_WRITE); 1184 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1); 1185 } 1186 1187 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE) 1188 ioflags = IO_NODELOCKED; 1189 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC) 1190 ioflags = (IO_SYNC | IO_NODELOCKED); 1191 else 1192 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 1193 uiop->uio_rw = UIO_WRITE; 1194 uiop->uio_segflg = UIO_SYSSPACE; 1195 uiop->uio_lwp = NULL; 1196 uiop->uio_offset = nfsd->nd_off; 1197 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off; 1198 if (uiop->uio_resid > 0) { 1199 mp = mrep; 1200 i = 0; 1201 while (mp) { 1202 if (mp->m_len > 0) 1203 i++; 1204 mp = mp->m_next; 1205 } 1206 uiop->uio_iovcnt = i; 1207 iov = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK); 1208 uiop->uio_iov = ivp = iov; 1209 mp = mrep; 1210 while (mp) { 1211 if (mp->m_len > 0) { 1212 ivp->iov_base = mtod(mp, caddr_t); 1213 ivp->iov_len = mp->m_len; 1214 ivp++; 1215 } 1216 mp = mp->m_next; 1217 } 1218 if (!error) { 1219 if (vn_start_write(vp, &mntp, V_NOWAIT) != 0) { 1220 VOP_UNLOCK(vp, 0); 1221 vn_start_write(NULL, &mntp, V_WAIT); 1222 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1223 } 1224 if (!error) { 1225 error = VOP_WRITE(vp, uiop, ioflags, cred); 1226 nfsstats.srvvop_writes++; 1227 vn_finished_write(mntp, 0); 1228 } 1229 } 1230 free((caddr_t)iov, M_TEMP); 1231 } 1232 m_freem(mrep); 1233 if (vp) { 1234 aftat_ret = VOP_GETATTR(vp, &va, cred, lwp); 1235 vput(vp); 1236 } 1237 1238 /* 1239 * Loop around generating replies for all write rpcs that have 1240 * now been completed. 1241 */ 1242 swp = nfsd; 1243 do { 1244 if (error) { 1245 nfsm_writereply(NFSX_WCCDATA(v3), v3); 1246 if (v3) { 1247 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1248 } 1249 } else { 1250 nfsm_writereply(NFSX_PREOPATTR(v3) + 1251 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED + 1252 NFSX_WRITEVERF(v3), v3); 1253 if (v3) { 1254 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1255 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1256 *tl++ = txdr_unsigned(nfsd->nd_len); 1257 *tl++ = txdr_unsigned(swp->nd_stable); 1258 /* 1259 * Actually, there is no need to txdr these fields, 1260 * but it may make the values more human readable, 1261 * for debugging purposes. 1262 */ 1263 *tl++ = txdr_unsigned(boottime.tv_sec); 1264 *tl = txdr_unsigned(boottime.tv_usec); 1265 } else { 1266 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1267 nfsm_srvfillattr(&va, fp); 1268 } 1269 } 1270 nfsd->nd_mreq = mreq; 1271 if (nfsd->nd_mrep) 1272 panic("nfsrv_write: nd_mrep not free"); 1273 1274 /* 1275 * Done. Put it at the head of the timer queue so that 1276 * the final phase can return the reply. 1277 */ 1278 s = splsoftclock(); 1279 if (nfsd != swp) { 1280 nfsd->nd_time = 0; 1281 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1282 } 1283 nfsd = LIST_FIRST(&swp->nd_coalesce); 1284 if (nfsd) { 1285 LIST_REMOVE(nfsd, nd_tq); 1286 } 1287 splx(s); 1288 } while (nfsd); 1289 s = splsoftclock(); 1290 swp->nd_time = 0; 1291 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq); 1292 splx(s); 1293 goto loop1; 1294 } 1295 splx(s); 1296 1297 /* 1298 * Search for a reply to return. 1299 */ 1300 s = splsoftclock(); 1301 LIST_FOREACH(nfsd, &slp->ns_tq, nd_tq) { 1302 if (nfsd->nd_mreq) { 1303 LIST_REMOVE(nfsd, nd_tq); 1304 *mrq = nfsd->nd_mreq; 1305 *ndp = nfsd; 1306 break; 1307 } 1308 } 1309 splx(s); 1310 return (0); 1311 } 1312 1313 /* 1314 * Coalesce the write request nfsd into owp. To do this we must: 1315 * - remove nfsd from the queues 1316 * - merge nfsd->nd_mrep into owp->nd_mrep 1317 * - update the nd_eoff and nd_stable for owp 1318 * - put nfsd on owp's nd_coalesce list 1319 * NB: Must be called at splsoftclock(). 1320 */ 1321 void 1322 nfsrvw_coalesce(owp, nfsd) 1323 struct nfsrv_descript *owp; 1324 struct nfsrv_descript *nfsd; 1325 { 1326 int overlap; 1327 struct mbuf *mp; 1328 struct nfsrv_descript *m; 1329 1330 LIST_REMOVE(nfsd, nd_hash); 1331 LIST_REMOVE(nfsd, nd_tq); 1332 if (owp->nd_eoff < nfsd->nd_eoff) { 1333 overlap = owp->nd_eoff - nfsd->nd_off; 1334 if (overlap < 0) 1335 panic("nfsrv_coalesce: bad off"); 1336 if (overlap > 0) 1337 m_adj(nfsd->nd_mrep, overlap); 1338 mp = owp->nd_mrep; 1339 while (mp->m_next) 1340 mp = mp->m_next; 1341 mp->m_next = nfsd->nd_mrep; 1342 owp->nd_eoff = nfsd->nd_eoff; 1343 } else 1344 m_freem(nfsd->nd_mrep); 1345 nfsd->nd_mrep = NULL; 1346 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC) 1347 owp->nd_stable = NFSV3WRITE_FILESYNC; 1348 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC && 1349 owp->nd_stable == NFSV3WRITE_UNSTABLE) 1350 owp->nd_stable = NFSV3WRITE_DATASYNC; 1351 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq); 1352 /* 1353 * nfsd might hold coalesce elements! Move them to owp. 1354 * Otherwise, requests may be lost and clients will be stuck. 1355 */ 1356 while ((m = LIST_FIRST(&nfsd->nd_coalesce)) != NULL) { 1357 LIST_REMOVE(m, nd_tq); 1358 LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq); 1359 } 1360 } 1361 1362 /* 1363 * nfs create service 1364 * now does a truncate to 0 length via. setattr if it already exists 1365 */ 1366 int 1367 nfsrv_create(nfsd, slp, lwp, mrq) 1368 struct nfsrv_descript *nfsd; 1369 struct nfssvc_sock *slp; 1370 struct lwp *lwp; 1371 struct mbuf **mrq; 1372 { 1373 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1374 struct mbuf *nam = nfsd->nd_nam; 1375 caddr_t dpos = nfsd->nd_dpos; 1376 struct ucred *cred = &nfsd->nd_cr; 1377 struct nfs_fattr *fp; 1378 struct vattr va, dirfor, diraft; 1379 struct nfsv2_sattr *sp; 1380 u_int32_t *tl; 1381 struct nameidata nd; 1382 caddr_t cp; 1383 int32_t t1; 1384 caddr_t bpos; 1385 int error = 0, cache, len, tsize, dirfor_ret = 1, diraft_ret = 1; 1386 int rdev = 0; 1387 int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0; 1388 char *cp2; 1389 struct mbuf *mb, *mreq; 1390 struct vnode *vp = NULL, *dirp = NULL; 1391 nfsfh_t nfh; 1392 fhandle_t *fhp; 1393 u_quad_t frev, tempsize; 1394 u_char cverf[NFSX_V3CREATEVERF]; 1395 struct mount *mp = NULL; 1396 1397 nd.ni_cnd.cn_nameiop = 0; 1398 fhp = &nfh.fh_generic; 1399 nfsm_srvmtofh(fhp); 1400 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1401 return (ESTALE); 1402 vn_start_write(NULL, &mp, V_WAIT); 1403 nfsm_srvnamesiz(len); 1404 nd.ni_cnd.cn_cred = cred; 1405 nd.ni_cnd.cn_nameiop = CREATE; 1406 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1407 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1408 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1409 if (dirp) { 1410 if (v3) 1411 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1412 else { 1413 vrele(dirp); 1414 dirp = (struct vnode *)0; 1415 } 1416 } 1417 if (error) { 1418 nfsm_reply(NFSX_WCCDATA(v3)); 1419 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1420 if (dirp) 1421 vrele(dirp); 1422 vn_finished_write(mp, 0); 1423 return (0); 1424 } 1425 VATTR_NULL(&va); 1426 if (v3) { 1427 va.va_mode = 0; 1428 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1429 how = fxdr_unsigned(int, *tl); 1430 switch (how) { 1431 case NFSV3CREATE_GUARDED: 1432 if (nd.ni_vp) { 1433 error = EEXIST; 1434 break; 1435 } 1436 case NFSV3CREATE_UNCHECKED: 1437 nfsm_srvsattr(&va); 1438 break; 1439 case NFSV3CREATE_EXCLUSIVE: 1440 nfsm_dissect(cp, caddr_t, NFSX_V3CREATEVERF); 1441 memcpy(cverf, cp, NFSX_V3CREATEVERF); 1442 exclusive_flag = 1; 1443 break; 1444 }; 1445 va.va_type = VREG; 1446 } else { 1447 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 1448 va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode)); 1449 if (va.va_type == VNON) 1450 va.va_type = VREG; 1451 va.va_mode = nfstov_mode(sp->sa_mode); 1452 switch (va.va_type) { 1453 case VREG: 1454 tsize = fxdr_unsigned(int32_t, sp->sa_size); 1455 if (tsize != -1) 1456 va.va_size = (u_quad_t)tsize; 1457 break; 1458 case VCHR: 1459 case VBLK: 1460 case VFIFO: 1461 rdev = fxdr_unsigned(int32_t, sp->sa_size); 1462 break; 1463 default: 1464 break; 1465 }; 1466 } 1467 1468 /* 1469 * Iff doesn't exist, create it 1470 * otherwise just truncate to 0 length 1471 * should I set the mode too ?? 1472 */ 1473 if (nd.ni_vp == NULL) { 1474 if (va.va_type == VREG || va.va_type == VSOCK) { 1475 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1476 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1477 if (!error) { 1478 if (exclusive_flag) { 1479 exclusive_flag = 0; 1480 VATTR_NULL(&va); 1481 /* 1482 * XXX 1483 * assuming NFSX_V3CREATEVERF 1484 * == sizeof(nfstime3) 1485 */ 1486 fxdr_nfsv3time(cverf, &va.va_atime); 1487 error = VOP_SETATTR(nd.ni_vp, &va, cred, 1488 lwp); 1489 } 1490 } 1491 } else if (va.va_type == VCHR || va.va_type == VBLK || 1492 va.va_type == VFIFO) { 1493 if (va.va_type == VCHR && rdev == 0xffffffff) 1494 va.va_type = VFIFO; 1495 if (va.va_type != VFIFO && 1496 (error = suser(cred, (u_short *)0))) { 1497 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1498 vput(nd.ni_dvp); 1499 nfsm_reply(0); 1500 vn_finished_write(mp, 0); 1501 return (error); 1502 } else 1503 va.va_rdev = (dev_t)rdev; 1504 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1505 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, 1506 &va); 1507 if (error) { 1508 nfsm_reply(0); 1509 } 1510 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1511 vrele(nd.ni_dvp); 1512 vput(nd.ni_vp); 1513 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1514 error = EINVAL; 1515 nfsm_reply(0); 1516 } 1517 } else { 1518 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1519 vput(nd.ni_dvp); 1520 error = ENXIO; 1521 } 1522 vp = nd.ni_vp; 1523 } else { 1524 vp = nd.ni_vp; 1525 if (nd.ni_dvp == vp) 1526 vrele(nd.ni_dvp); 1527 else 1528 vput(nd.ni_dvp); 1529 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1530 if (!error && va.va_size != -1) { 1531 error = nfsrv_access(vp, VWRITE, cred, 1532 (nd.ni_cnd.cn_flags & RDONLY), lwp, 0); 1533 if (!error) { 1534 nqsrv_getl(vp, ND_WRITE); 1535 tempsize = va.va_size; 1536 VATTR_NULL(&va); 1537 va.va_size = tempsize; 1538 error = VOP_SETATTR(vp, &va, cred, lwp); 1539 } 1540 } 1541 if (error) 1542 vput(vp); 1543 } 1544 if (!error) { 1545 memset((caddr_t)fhp, 0, sizeof(nfh)); 1546 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1547 error = VFS_VPTOFH(vp, &fhp->fh_fid); 1548 if (!error) 1549 error = VOP_GETATTR(vp, &va, cred, lwp); 1550 vput(vp); 1551 } 1552 if (v3) { 1553 if (exclusive_flag && !error) { 1554 /* 1555 * XXX assuming NFSX_V3CREATEVERF == sizeof(nfstime3) 1556 */ 1557 char oldverf[NFSX_V3CREATEVERF]; 1558 1559 txdr_nfsv3time(&va.va_atime, oldverf); 1560 if (memcmp(cverf, oldverf, NFSX_V3CREATEVERF)) 1561 error = EEXIST; 1562 } 1563 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1564 vrele(dirp); 1565 } 1566 nfsm_reply(NFSX_SRVFH(v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3)); 1567 if (v3) { 1568 if (!error) { 1569 nfsm_srvpostop_fh(fhp); 1570 nfsm_srvpostop_attr(0, &va); 1571 } 1572 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1573 } else { 1574 nfsm_srvfhtom(fhp, v3); 1575 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1576 nfsm_srvfillattr(&va, fp); 1577 } 1578 vn_finished_write(mp, 0); 1579 return (0); 1580 nfsmout: 1581 if (dirp) 1582 vrele(dirp); 1583 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1584 if (nd.ni_dvp == nd.ni_vp) 1585 vrele(nd.ni_dvp); 1586 else 1587 vput(nd.ni_dvp); 1588 if (nd.ni_vp) 1589 vput(nd.ni_vp); 1590 vn_finished_write(mp, 0); 1591 return (error); 1592 } 1593 1594 /* 1595 * nfs v3 mknod service 1596 */ 1597 int 1598 nfsrv_mknod(nfsd, slp, lwp, mrq) 1599 struct nfsrv_descript *nfsd; 1600 struct nfssvc_sock *slp; 1601 struct lwp *lwp; 1602 struct mbuf **mrq; 1603 { 1604 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1605 struct mbuf *nam = nfsd->nd_nam; 1606 caddr_t dpos = nfsd->nd_dpos; 1607 struct ucred *cred = &nfsd->nd_cr; 1608 struct vattr va, dirfor, diraft; 1609 u_int32_t *tl; 1610 struct nameidata nd; 1611 int32_t t1; 1612 caddr_t bpos; 1613 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 1614 u_int32_t major, minor; 1615 enum vtype vtyp; 1616 char *cp2; 1617 struct mbuf *mb, *mreq; 1618 struct vnode *vp, *dirp = (struct vnode *)0; 1619 nfsfh_t nfh; 1620 fhandle_t *fhp; 1621 u_quad_t frev; 1622 struct mount *mp = NULL; 1623 1624 nd.ni_cnd.cn_nameiop = 0; 1625 fhp = &nfh.fh_generic; 1626 nfsm_srvmtofh(fhp); 1627 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1628 return (ESTALE); 1629 vn_start_write(NULL, &mp, V_WAIT); 1630 nfsm_srvnamesiz(len); 1631 nd.ni_cnd.cn_cred = cred; 1632 nd.ni_cnd.cn_nameiop = CREATE; 1633 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1634 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1635 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1636 if (dirp) 1637 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1638 if (error) { 1639 nfsm_reply(NFSX_WCCDATA(1)); 1640 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1641 if (dirp) 1642 vrele(dirp); 1643 vn_finished_write(mp, 0); 1644 return (0); 1645 } 1646 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1647 vtyp = nfsv3tov_type(*tl); 1648 if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) { 1649 error = NFSERR_BADTYPE; 1650 goto abort; 1651 } 1652 VATTR_NULL(&va); 1653 va.va_mode = 0; 1654 nfsm_srvsattr(&va); 1655 if (vtyp == VCHR || vtyp == VBLK) { 1656 dev_t rdev; 1657 1658 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1659 major = fxdr_unsigned(u_int32_t, *tl++); 1660 minor = fxdr_unsigned(u_int32_t, *tl); 1661 rdev = makedev(major, minor); 1662 if (major(rdev) != major || minor(rdev) != minor) { 1663 error = EINVAL; 1664 goto abort; 1665 } 1666 va.va_rdev = rdev; 1667 } 1668 1669 /* 1670 * Iff doesn't exist, create it. 1671 */ 1672 if (nd.ni_vp) { 1673 error = EEXIST; 1674 abort: 1675 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1676 if (nd.ni_dvp == nd.ni_vp) 1677 vrele(nd.ni_dvp); 1678 else 1679 vput(nd.ni_dvp); 1680 if (nd.ni_vp) 1681 vput(nd.ni_vp); 1682 goto out; 1683 } 1684 va.va_type = vtyp; 1685 if (vtyp == VSOCK) { 1686 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1687 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1688 } else { 1689 if (va.va_type != VFIFO && 1690 (error = suser(cred, (u_short *)0))) { 1691 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1692 vput(nd.ni_dvp); 1693 goto out; 1694 } 1695 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1696 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1697 if (error) { 1698 goto out; 1699 } 1700 if (error) 1701 goto out; 1702 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1703 vput(nd.ni_vp); 1704 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1705 error = EINVAL; 1706 } 1707 } 1708 out: 1709 vp = nd.ni_vp; 1710 if (!error) { 1711 memset((caddr_t)fhp, 0, sizeof(nfh)); 1712 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1713 error = VFS_VPTOFH(vp, &fhp->fh_fid); 1714 if (!error) 1715 error = VOP_GETATTR(vp, &va, cred, lwp); 1716 vput(vp); 1717 } 1718 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1719 vrele(dirp); 1720 nfsm_reply(NFSX_SRVFH(1) + NFSX_POSTOPATTR(1) + NFSX_WCCDATA(1)); 1721 if (!error) { 1722 nfsm_srvpostop_fh(fhp); 1723 nfsm_srvpostop_attr(0, &va); 1724 } 1725 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1726 vn_finished_write(mp, 0); 1727 return (0); 1728 nfsmout: 1729 if (dirp) 1730 vrele(dirp); 1731 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1732 if (nd.ni_dvp == nd.ni_vp) 1733 vrele(nd.ni_dvp); 1734 else 1735 vput(nd.ni_dvp); 1736 if (nd.ni_vp) 1737 vput(nd.ni_vp); 1738 vn_finished_write(mp, 0); 1739 return (error); 1740 } 1741 1742 /* 1743 * nfs remove service 1744 */ 1745 int 1746 nfsrv_remove(nfsd, slp, lwp, mrq) 1747 struct nfsrv_descript *nfsd; 1748 struct nfssvc_sock *slp; 1749 struct lwp *lwp; 1750 struct mbuf **mrq; 1751 { 1752 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1753 struct mbuf *nam = nfsd->nd_nam; 1754 caddr_t dpos = nfsd->nd_dpos; 1755 struct ucred *cred = &nfsd->nd_cr; 1756 struct nameidata nd; 1757 u_int32_t *tl; 1758 int32_t t1; 1759 caddr_t bpos; 1760 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 1761 int v3 = (nfsd->nd_flag & ND_NFSV3); 1762 char *cp2; 1763 struct mbuf *mb, *mreq; 1764 struct vnode *vp, *dirp; 1765 struct vattr dirfor, diraft; 1766 nfsfh_t nfh; 1767 fhandle_t *fhp; 1768 u_quad_t frev; 1769 struct mount *mp = NULL; 1770 1771 #ifndef nolint 1772 vp = (struct vnode *)0; 1773 #endif 1774 fhp = &nfh.fh_generic; 1775 nfsm_srvmtofh(fhp); 1776 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1777 return (ESTALE); 1778 vn_start_write(NULL, &mp, V_WAIT); 1779 nfsm_srvnamesiz(len); 1780 nd.ni_cnd.cn_cred = cred; 1781 nd.ni_cnd.cn_nameiop = DELETE; 1782 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1783 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1784 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1785 if (dirp) { 1786 if (v3) 1787 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1788 else 1789 vrele(dirp); 1790 } 1791 if (!error) { 1792 vp = nd.ni_vp; 1793 if (vp->v_type == VDIR && 1794 (error = suser(cred, (u_short *)0)) != 0) 1795 goto out; 1796 /* 1797 * The root of a mounted filesystem cannot be deleted. 1798 */ 1799 if (vp->v_flag & VROOT) { 1800 error = EBUSY; 1801 goto out; 1802 } 1803 out: 1804 if (!error) { 1805 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1806 nqsrv_getl(vp, ND_WRITE); 1807 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1808 } else { 1809 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1810 if (nd.ni_dvp == vp) 1811 vrele(nd.ni_dvp); 1812 else 1813 vput(nd.ni_dvp); 1814 vput(vp); 1815 } 1816 } 1817 if (dirp && v3) { 1818 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1819 vrele(dirp); 1820 } 1821 nfsm_reply(NFSX_WCCDATA(v3)); 1822 if (v3) { 1823 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1824 vn_finished_write(mp, 0); 1825 return (0); 1826 } 1827 vn_finished_write(mp, 0); 1828 nfsm_srvdone; 1829 } 1830 1831 /* 1832 * nfs rename service 1833 */ 1834 int 1835 nfsrv_rename(nfsd, slp, lwp, mrq) 1836 struct nfsrv_descript *nfsd; 1837 struct nfssvc_sock *slp; 1838 struct lwp *lwp; 1839 struct mbuf **mrq; 1840 { 1841 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1842 struct mbuf *nam = nfsd->nd_nam; 1843 caddr_t dpos = nfsd->nd_dpos; 1844 struct ucred *cred = &nfsd->nd_cr; 1845 u_int32_t *tl; 1846 int32_t t1; 1847 caddr_t bpos; 1848 int error = 0, cache, fdirfor_ret = 1, fdiraft_ret = 1; 1849 uint32_t len, len2; 1850 int tdirfor_ret = 1, tdiraft_ret = 1; 1851 int v3 = (nfsd->nd_flag & ND_NFSV3); 1852 char *cp2; 1853 struct mbuf *mb, *mreq; 1854 struct nameidata fromnd, tond; 1855 struct vnode *fvp, *tvp, *tdvp, *fdirp = (struct vnode *)0; 1856 struct vnode *tdirp = (struct vnode *)0; 1857 struct vattr fdirfor, fdiraft, tdirfor, tdiraft; 1858 nfsfh_t fnfh, tnfh; 1859 fhandle_t *ffhp, *tfhp; 1860 u_quad_t frev; 1861 uid_t saved_uid; 1862 struct mount *mp = NULL; 1863 1864 #ifndef nolint 1865 fvp = (struct vnode *)0; 1866 #endif 1867 ffhp = &fnfh.fh_generic; 1868 tfhp = &tnfh.fh_generic; 1869 fromnd.ni_cnd.cn_nameiop = 0; 1870 tond.ni_cnd.cn_nameiop = 0; 1871 nfsm_srvmtofh(ffhp); 1872 if ((mp = vfs_getvfs(&ffhp->fh_fsid)) == NULL) 1873 return (ESTALE); 1874 vn_start_write(NULL, &mp, V_WAIT); 1875 nfsm_srvnamesiz(len); 1876 /* 1877 * Remember our original uid so that we can reset cr_uid before 1878 * the second nfs_namei() call, in case it is remapped. 1879 */ 1880 saved_uid = cred->cr_uid; 1881 fromnd.ni_cnd.cn_cred = cred; 1882 fromnd.ni_cnd.cn_nameiop = DELETE; 1883 fromnd.ni_cnd.cn_flags = WANTPARENT | SAVESTART; 1884 error = nfs_namei(&fromnd, ffhp, len, slp, nam, &md, 1885 &dpos, &fdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1886 if (fdirp) { 1887 if (v3) 1888 fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred, lwp); 1889 else { 1890 vrele(fdirp); 1891 fdirp = (struct vnode *)0; 1892 } 1893 } 1894 if (error) { 1895 nfsm_reply(2 * NFSX_WCCDATA(v3)); 1896 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 1897 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 1898 if (fdirp) 1899 vrele(fdirp); 1900 vn_finished_write(mp, 0); 1901 return (0); 1902 } 1903 fvp = fromnd.ni_vp; 1904 nfsm_srvmtofh(tfhp); 1905 if (v3) { 1906 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 1907 len2 = fxdr_unsigned(uint32_t, *tl); 1908 /* len2 will be checked by nfs_namei */ 1909 } 1910 else { 1911 /* NFSv2 */ 1912 nfsm_strsiz(len2, NFS_MAXNAMLEN); 1913 } 1914 cred->cr_uid = saved_uid; 1915 tond.ni_cnd.cn_cred = cred; 1916 tond.ni_cnd.cn_nameiop = RENAME; 1917 tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART; 1918 error = nfs_namei(&tond, tfhp, len2, slp, nam, &md, 1919 &dpos, &tdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1920 if (tdirp) { 1921 if (v3) 1922 tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred, lwp); 1923 else { 1924 vrele(tdirp); 1925 tdirp = (struct vnode *)0; 1926 } 1927 } 1928 if (error) { 1929 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1930 vrele(fromnd.ni_dvp); 1931 vrele(fvp); 1932 goto out1; 1933 } 1934 tdvp = tond.ni_dvp; 1935 tvp = tond.ni_vp; 1936 if (tvp != NULL) { 1937 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 1938 if (v3) 1939 error = EEXIST; 1940 else 1941 error = EISDIR; 1942 goto out; 1943 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 1944 if (v3) 1945 error = EEXIST; 1946 else 1947 error = ENOTDIR; 1948 goto out; 1949 } 1950 if (tvp->v_type == VDIR && tvp->v_mountedhere) { 1951 if (v3) 1952 error = EXDEV; 1953 else 1954 error = ENOTEMPTY; 1955 goto out; 1956 } 1957 } 1958 if (fvp->v_type == VDIR && fvp->v_mountedhere) { 1959 if (v3) 1960 error = EXDEV; 1961 else 1962 error = ENOTEMPTY; 1963 goto out; 1964 } 1965 if (fvp->v_mount != tdvp->v_mount) { 1966 if (v3) 1967 error = EXDEV; 1968 else 1969 error = ENOTEMPTY; 1970 goto out; 1971 } 1972 if (fvp == tdvp) { 1973 if (v3) 1974 error = EINVAL; 1975 else 1976 error = ENOTEMPTY; 1977 } 1978 /* 1979 * If source is the same as the destination (that is the 1980 * same vnode with the same name in the same directory), 1981 * then there is nothing to do. 1982 */ 1983 if (fvp == tvp && fromnd.ni_dvp == tdvp && 1984 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 1985 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr, 1986 fromnd.ni_cnd.cn_namelen)) 1987 error = -1; 1988 out: 1989 if (!error) { 1990 nqsrv_getl(fromnd.ni_dvp, ND_WRITE); 1991 nqsrv_getl(tdvp, ND_WRITE); 1992 if (tvp) { 1993 nqsrv_getl(tvp, ND_WRITE); 1994 } 1995 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 1996 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 1997 } else { 1998 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 1999 if (tdvp == tvp) 2000 vrele(tdvp); 2001 else 2002 vput(tdvp); 2003 if (tvp) 2004 vput(tvp); 2005 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2006 vrele(fromnd.ni_dvp); 2007 vrele(fvp); 2008 if (error == -1) 2009 error = 0; 2010 } 2011 vrele(tond.ni_startdir); 2012 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 2013 out1: 2014 if (fdirp) { 2015 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred, lwp); 2016 vrele(fdirp); 2017 } 2018 if (tdirp) { 2019 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred, lwp); 2020 vrele(tdirp); 2021 } 2022 vrele(fromnd.ni_startdir); 2023 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2024 nfsm_reply(2 * NFSX_WCCDATA(v3)); 2025 if (v3) { 2026 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 2027 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 2028 } 2029 vn_finished_write(mp, 0); 2030 return (0); 2031 2032 nfsmout: 2033 if (fdirp) 2034 vrele(fdirp); 2035 if (tdirp) 2036 vrele(tdirp); 2037 if (tond.ni_cnd.cn_nameiop) { 2038 vrele(tond.ni_startdir); 2039 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 2040 } 2041 if (fromnd.ni_cnd.cn_nameiop) { 2042 vrele(fromnd.ni_startdir); 2043 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2044 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2045 vrele(fromnd.ni_dvp); 2046 vrele(fvp); 2047 } 2048 return (error); 2049 } 2050 2051 /* 2052 * nfs link service 2053 */ 2054 int 2055 nfsrv_link(nfsd, slp, lwp, mrq) 2056 struct nfsrv_descript *nfsd; 2057 struct nfssvc_sock *slp; 2058 struct lwp *lwp; 2059 struct mbuf **mrq; 2060 { 2061 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2062 struct mbuf *nam = nfsd->nd_nam; 2063 caddr_t dpos = nfsd->nd_dpos; 2064 struct ucred *cred = &nfsd->nd_cr; 2065 struct nameidata nd; 2066 u_int32_t *tl; 2067 int32_t t1; 2068 caddr_t bpos; 2069 int error = 0, rdonly, cache, len, dirfor_ret = 1, diraft_ret = 1; 2070 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3); 2071 char *cp2; 2072 struct mbuf *mb, *mreq; 2073 struct vnode *vp, *xp, *dirp = (struct vnode *)0; 2074 struct vattr dirfor, diraft, at; 2075 nfsfh_t nfh, dnfh; 2076 fhandle_t *fhp, *dfhp; 2077 u_quad_t frev; 2078 struct mount *mp = NULL; 2079 2080 fhp = &nfh.fh_generic; 2081 dfhp = &dnfh.fh_generic; 2082 nfsm_srvmtofh(fhp); 2083 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2084 return (ESTALE); 2085 vn_start_write(NULL, &mp, V_WAIT); 2086 nfsm_srvmtofh(dfhp); 2087 nfsm_srvnamesiz(len); 2088 error = nfsrv_fhtovp(fhp, FALSE, &vp, cred, slp, nam, 2089 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2090 if (error) { 2091 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2092 nfsm_srvpostop_attr(getret, &at); 2093 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2094 vn_finished_write(mp, 0); 2095 return (0); 2096 } 2097 if (vp->v_type == VDIR && (error = suser(cred, (u_short *)0)) != 0) 2098 goto out1; 2099 nd.ni_cnd.cn_cred = cred; 2100 nd.ni_cnd.cn_nameiop = CREATE; 2101 nd.ni_cnd.cn_flags = LOCKPARENT; 2102 error = nfs_namei(&nd, dfhp, len, slp, nam, &md, &dpos, 2103 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2104 if (dirp) { 2105 if (v3) 2106 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2107 else { 2108 vrele(dirp); 2109 dirp = (struct vnode *)0; 2110 } 2111 } 2112 if (error) 2113 goto out1; 2114 xp = nd.ni_vp; 2115 if (xp != NULL) { 2116 error = EEXIST; 2117 goto out; 2118 } 2119 xp = nd.ni_dvp; 2120 if (vp->v_mount != xp->v_mount) 2121 error = EXDEV; 2122 out: 2123 if (!error) { 2124 nqsrv_getl(vp, ND_WRITE); 2125 nqsrv_getl(xp, ND_WRITE); 2126 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2127 } else { 2128 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2129 if (nd.ni_dvp == nd.ni_vp) 2130 vrele(nd.ni_dvp); 2131 else 2132 vput(nd.ni_dvp); 2133 if (nd.ni_vp) 2134 vrele(nd.ni_vp); 2135 } 2136 out1: 2137 if (v3) 2138 getret = VOP_GETATTR(vp, &at, cred, lwp); 2139 if (dirp) { 2140 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2141 vrele(dirp); 2142 } 2143 vrele(vp); 2144 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2145 if (v3) { 2146 nfsm_srvpostop_attr(getret, &at); 2147 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2148 vn_finished_write(mp, 0); 2149 return (0); 2150 } 2151 vn_finished_write(mp, 0); 2152 nfsm_srvdone; 2153 } 2154 2155 /* 2156 * nfs symbolic link service 2157 */ 2158 int 2159 nfsrv_symlink(nfsd, slp, lwp, mrq) 2160 struct nfsrv_descript *nfsd; 2161 struct nfssvc_sock *slp; 2162 struct lwp *lwp; 2163 struct mbuf **mrq; 2164 { 2165 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2166 struct mbuf *nam = nfsd->nd_nam; 2167 caddr_t dpos = nfsd->nd_dpos; 2168 struct ucred *cred = &nfsd->nd_cr; 2169 struct vattr va, dirfor, diraft; 2170 struct nameidata nd; 2171 u_int32_t *tl; 2172 int32_t t1; 2173 struct nfsv2_sattr *sp; 2174 char *bpos, *pathcp = NULL, *cp2; 2175 struct uio io; 2176 struct iovec iv; 2177 int error = 0, cache, dirfor_ret = 1, diraft_ret = 1; 2178 uint32_t len, len2; 2179 int v3 = (nfsd->nd_flag & ND_NFSV3); 2180 struct mbuf *mb, *mreq; 2181 struct vnode *dirp = (struct vnode *)0; 2182 nfsfh_t nfh; 2183 fhandle_t *fhp; 2184 u_quad_t frev; 2185 struct mount *mp = NULL; 2186 2187 nd.ni_cnd.cn_nameiop = 0; 2188 fhp = &nfh.fh_generic; 2189 nfsm_srvmtofh(fhp); 2190 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2191 return (ESTALE); 2192 vn_start_write(NULL, &mp, V_WAIT); 2193 nfsm_srvnamesiz(len); 2194 nd.ni_cnd.cn_cred = cred; 2195 nd.ni_cnd.cn_nameiop = CREATE; 2196 nd.ni_cnd.cn_flags = LOCKPARENT; 2197 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2198 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2199 if (dirp) { 2200 if (v3) 2201 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2202 else { 2203 vrele(dirp); 2204 dirp = (struct vnode *)0; 2205 } 2206 } 2207 if (error) 2208 goto out; 2209 VATTR_NULL(&va); 2210 va.va_type = VLNK; 2211 if (v3) { 2212 va.va_mode = 0; 2213 nfsm_srvsattr(&va); 2214 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 2215 len2 = fxdr_unsigned(uint32_t, *tl); 2216 if (len2 > PATH_MAX) { 2217 /* XXX should check _PC_NO_TRUNC */ 2218 error = ENAMETOOLONG; 2219 goto abortop; 2220 } 2221 } 2222 else { 2223 /* NFSv2 */ 2224 nfsm_strsiz(len2, NFS_MAXPATHLEN); 2225 } 2226 pathcp = malloc(len2 + 1, M_TEMP, M_WAITOK); 2227 iv.iov_base = pathcp; 2228 iv.iov_len = len2; 2229 io.uio_resid = len2; 2230 io.uio_offset = 0; 2231 io.uio_iov = &iv; 2232 io.uio_iovcnt = 1; 2233 io.uio_segflg = UIO_SYSSPACE; 2234 io.uio_rw = UIO_READ; 2235 io.uio_lwp = NULL; 2236 nfsm_mtouio(&io, len2); 2237 if (!v3) { 2238 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 2239 va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode); 2240 } 2241 *(pathcp + len2) = '\0'; 2242 if (nd.ni_vp) { 2243 error = EEXIST; 2244 abortop: 2245 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2246 if (nd.ni_dvp == nd.ni_vp) 2247 vrele(nd.ni_dvp); 2248 else 2249 vput(nd.ni_dvp); 2250 if (nd.ni_vp) 2251 vrele(nd.ni_vp); 2252 goto out; 2253 } 2254 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2255 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp); 2256 if (!error) { 2257 if (v3) { 2258 memset((caddr_t)fhp, 0, sizeof(nfh)); 2259 fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsidx; 2260 error = VFS_VPTOFH(nd.ni_vp, &fhp->fh_fid); 2261 if (!error) 2262 error = VOP_GETATTR(nd.ni_vp, &va, cred, lwp); 2263 vput(nd.ni_vp); 2264 } else { 2265 vput(nd.ni_vp); 2266 } 2267 } 2268 out: 2269 if (pathcp) 2270 free(pathcp, M_TEMP); 2271 if (dirp) { 2272 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2273 vrele(dirp); 2274 } 2275 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2276 if (v3) { 2277 if (!error) { 2278 nfsm_srvpostop_fh(fhp); 2279 nfsm_srvpostop_attr(0, &va); 2280 } 2281 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2282 } 2283 vn_finished_write(mp, 0); 2284 return (0); 2285 nfsmout: 2286 if (dirp) 2287 vrele(dirp); 2288 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2289 if (nd.ni_dvp == nd.ni_vp) 2290 vrele(nd.ni_dvp); 2291 else 2292 vput(nd.ni_dvp); 2293 if (nd.ni_vp) 2294 vrele(nd.ni_vp); 2295 if (pathcp) 2296 free(pathcp, M_TEMP); 2297 vn_finished_write(mp, 0); 2298 return (error); 2299 } 2300 2301 /* 2302 * nfs mkdir service 2303 */ 2304 int 2305 nfsrv_mkdir(nfsd, slp, lwp, mrq) 2306 struct nfsrv_descript *nfsd; 2307 struct nfssvc_sock *slp; 2308 struct lwp *lwp; 2309 struct mbuf **mrq; 2310 { 2311 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2312 struct mbuf *nam = nfsd->nd_nam; 2313 caddr_t dpos = nfsd->nd_dpos; 2314 struct ucred *cred = &nfsd->nd_cr; 2315 struct vattr va, dirfor, diraft; 2316 struct nfs_fattr *fp; 2317 struct nameidata nd; 2318 caddr_t cp; 2319 u_int32_t *tl; 2320 int32_t t1; 2321 caddr_t bpos; 2322 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 2323 int v3 = (nfsd->nd_flag & ND_NFSV3); 2324 char *cp2; 2325 struct mbuf *mb, *mreq; 2326 struct vnode *vp, *dirp = (struct vnode *)0; 2327 nfsfh_t nfh; 2328 fhandle_t *fhp; 2329 u_quad_t frev; 2330 struct mount *mp = NULL; 2331 2332 fhp = &nfh.fh_generic; 2333 nfsm_srvmtofh(fhp); 2334 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2335 return (ESTALE); 2336 vn_start_write(NULL, &mp, V_WAIT); 2337 nfsm_srvnamesiz(len); 2338 nd.ni_cnd.cn_cred = cred; 2339 nd.ni_cnd.cn_nameiop = CREATE; 2340 nd.ni_cnd.cn_flags = LOCKPARENT; 2341 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2342 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2343 if (dirp) { 2344 if (v3) 2345 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2346 else { 2347 vrele(dirp); 2348 dirp = (struct vnode *)0; 2349 } 2350 } 2351 if (error) { 2352 nfsm_reply(NFSX_WCCDATA(v3)); 2353 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2354 if (dirp) 2355 vrele(dirp); 2356 vn_finished_write(mp, 0); 2357 return (0); 2358 } 2359 VATTR_NULL(&va); 2360 if (v3) { 2361 va.va_mode = 0; 2362 nfsm_srvsattr(&va); 2363 } else { 2364 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 2365 va.va_mode = nfstov_mode(*tl++); 2366 } 2367 va.va_type = VDIR; 2368 vp = nd.ni_vp; 2369 if (vp != NULL) { 2370 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2371 if (nd.ni_dvp == vp) 2372 vrele(nd.ni_dvp); 2373 else 2374 vput(nd.ni_dvp); 2375 vrele(vp); 2376 error = EEXIST; 2377 goto out; 2378 } 2379 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2380 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 2381 if (!error) { 2382 vp = nd.ni_vp; 2383 memset((caddr_t)fhp, 0, sizeof(nfh)); 2384 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 2385 error = VFS_VPTOFH(vp, &fhp->fh_fid); 2386 if (!error) 2387 error = VOP_GETATTR(vp, &va, cred, lwp); 2388 vput(vp); 2389 } 2390 out: 2391 if (dirp) { 2392 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2393 vrele(dirp); 2394 } 2395 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2396 if (v3) { 2397 if (!error) { 2398 nfsm_srvpostop_fh(fhp); 2399 nfsm_srvpostop_attr(0, &va); 2400 } 2401 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2402 } else { 2403 nfsm_srvfhtom(fhp, v3); 2404 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 2405 nfsm_srvfillattr(&va, fp); 2406 } 2407 vn_finished_write(mp, 0); 2408 return (0); 2409 nfsmout: 2410 if (dirp) 2411 vrele(dirp); 2412 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2413 if (nd.ni_dvp == nd.ni_vp) 2414 vrele(nd.ni_dvp); 2415 else 2416 vput(nd.ni_dvp); 2417 if (nd.ni_vp) 2418 vrele(nd.ni_vp); 2419 vn_finished_write(mp, 0); 2420 return (error); 2421 } 2422 2423 /* 2424 * nfs rmdir service 2425 */ 2426 int 2427 nfsrv_rmdir(nfsd, slp, lwp, mrq) 2428 struct nfsrv_descript *nfsd; 2429 struct nfssvc_sock *slp; 2430 struct lwp *lwp; 2431 struct mbuf **mrq; 2432 { 2433 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2434 struct mbuf *nam = nfsd->nd_nam; 2435 caddr_t dpos = nfsd->nd_dpos; 2436 struct ucred *cred = &nfsd->nd_cr; 2437 u_int32_t *tl; 2438 int32_t t1; 2439 caddr_t bpos; 2440 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 2441 int v3 = (nfsd->nd_flag & ND_NFSV3); 2442 char *cp2; 2443 struct mbuf *mb, *mreq; 2444 struct vnode *vp, *dirp = (struct vnode *)0; 2445 struct vattr dirfor, diraft; 2446 nfsfh_t nfh; 2447 fhandle_t *fhp; 2448 struct nameidata nd; 2449 u_quad_t frev; 2450 struct mount *mp = NULL; 2451 2452 fhp = &nfh.fh_generic; 2453 nfsm_srvmtofh(fhp); 2454 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2455 return (ESTALE); 2456 vn_start_write(NULL, &mp, V_WAIT); 2457 nfsm_srvnamesiz(len); 2458 nd.ni_cnd.cn_cred = cred; 2459 nd.ni_cnd.cn_nameiop = DELETE; 2460 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 2461 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2462 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2463 if (dirp) { 2464 if (v3) 2465 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2466 else { 2467 vrele(dirp); 2468 dirp = (struct vnode *)0; 2469 } 2470 } 2471 if (error) { 2472 nfsm_reply(NFSX_WCCDATA(v3)); 2473 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2474 if (dirp) 2475 vrele(dirp); 2476 vn_finished_write(mp, 0); 2477 return (0); 2478 } 2479 vp = nd.ni_vp; 2480 if (vp->v_type != VDIR) { 2481 error = ENOTDIR; 2482 goto out; 2483 } 2484 /* 2485 * No rmdir "." please. 2486 */ 2487 if (nd.ni_dvp == vp) { 2488 error = EINVAL; 2489 goto out; 2490 } 2491 /* 2492 * The root of a mounted filesystem cannot be deleted. 2493 */ 2494 if (vp->v_flag & VROOT) 2495 error = EBUSY; 2496 out: 2497 if (!error) { 2498 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2499 nqsrv_getl(vp, ND_WRITE); 2500 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2501 } else { 2502 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2503 if (nd.ni_dvp == nd.ni_vp) 2504 vrele(nd.ni_dvp); 2505 else 2506 vput(nd.ni_dvp); 2507 vput(vp); 2508 } 2509 if (dirp) { 2510 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2511 vrele(dirp); 2512 } 2513 nfsm_reply(NFSX_WCCDATA(v3)); 2514 if (v3) { 2515 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2516 vn_finished_write(mp, 0); 2517 return (0); 2518 } 2519 vn_finished_write(mp, 0); 2520 nfsm_srvdone; 2521 } 2522 2523 /* 2524 * nfs readdir service 2525 * - mallocs what it thinks is enough to read 2526 * count rounded up to a multiple of NFS_SRVDIRBLKSIZ <= NFS_MAXREADDIR 2527 * - calls VOP_READDIR() 2528 * - loops around building the reply 2529 * if the output generated exceeds count break out of loop 2530 * The nfsm_clget macro is used here so that the reply will be packed 2531 * tightly in mbuf clusters. 2532 * - it only knows that it has encountered eof when the VOP_READDIR() 2533 * reads nothing 2534 * - as such one readdir rpc will return eof false although you are there 2535 * and then the next will return eof 2536 * - it trims out records with d_fileno == 0 2537 * this doesn't matter for Unix clients, but they might confuse clients 2538 * for other os'. 2539 * - it trims out records with d_type == DT_WHT 2540 * these cannot be seen through NFS (unless we extend the protocol) 2541 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less 2542 * than requested, but this may not apply to all filesystems. For 2543 * example, client NFS does not { although it is never remote mounted 2544 * anyhow } 2545 * The alternate call nfsrv_readdirplus() does lookups as well. 2546 * PS: The NFS protocol spec. does not clarify what the "count" byte 2547 * argument is a count of.. just name strings and file id's or the 2548 * entire reply rpc or ... 2549 * I tried just file name and id sizes and it confused the Sun client, 2550 * so I am using the full rpc size now. The "paranoia.." comment refers 2551 * to including the status longwords that are not a part of the dir. 2552 * "entry" structures, but are in the rpc. 2553 */ 2554 2555 #define NFS_SRVDIRBLKSIZ 1024 2556 2557 struct flrep { 2558 nfsuint64 fl_off; 2559 u_int32_t fl_postopok; 2560 u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)]; 2561 u_int32_t fl_fhok; 2562 u_int32_t fl_fhsize; 2563 u_int32_t fl_nfh[NFSX_V3FH / sizeof (u_int32_t)]; 2564 }; 2565 2566 int 2567 nfsrv_readdir(nfsd, slp, lwp, mrq) 2568 struct nfsrv_descript *nfsd; 2569 struct nfssvc_sock *slp; 2570 struct lwp *lwp; 2571 struct mbuf **mrq; 2572 { 2573 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2574 struct mbuf *nam = nfsd->nd_nam; 2575 caddr_t dpos = nfsd->nd_dpos; 2576 struct ucred *cred = &nfsd->nd_cr; 2577 char *bp, *be; 2578 struct mbuf *mp; 2579 struct dirent *dp; 2580 caddr_t cp; 2581 u_int32_t *tl; 2582 int32_t t1; 2583 caddr_t bpos; 2584 struct mbuf *mb, *mreq, *mp2; 2585 char *cpos, *cend, *cp2, *rbuf; 2586 struct vnode *vp; 2587 struct vattr at; 2588 nfsfh_t nfh; 2589 fhandle_t *fhp; 2590 struct uio io; 2591 struct iovec iv; 2592 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2593 int siz, cnt, fullsiz, eofflag, rdonly, cache, ncookies; 2594 int v3 = (nfsd->nd_flag & ND_NFSV3); 2595 u_quad_t frev, off, toff, verf; 2596 off_t *cookies = NULL, *cookiep; 2597 nfsuint64 jar; 2598 2599 fhp = &nfh.fh_generic; 2600 nfsm_srvmtofh(fhp); 2601 if (v3) { 2602 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 2603 toff = fxdr_hyper(tl); 2604 tl += 2; 2605 verf = fxdr_hyper(tl); 2606 tl += 2; 2607 } else { 2608 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2609 toff = fxdr_unsigned(u_quad_t, *tl++); 2610 } 2611 off = toff; 2612 cnt = fxdr_unsigned(int, *tl); 2613 siz = ((cnt + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1)); 2614 xfer = NFS_SRVMAXDATA(nfsd); 2615 if (siz > xfer) 2616 siz = xfer; 2617 fullsiz = siz; 2618 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 2619 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2620 if (!error && vp->v_type != VDIR) { 2621 error = ENOTDIR; 2622 vput(vp); 2623 } 2624 if (error) { 2625 nfsm_reply(NFSX_UNSIGNED); 2626 nfsm_srvpostop_attr(getret, &at); 2627 return (0); 2628 } 2629 nqsrv_getl(vp, ND_READ); 2630 if (v3) { 2631 error = getret = VOP_GETATTR(vp, &at, cred, lwp); 2632 #ifdef NFS3_STRICTVERF 2633 /* 2634 * XXX This check is too strict for Solaris 2.5 clients. 2635 */ 2636 if (!error && toff && verf != at.va_filerev) 2637 error = NFSERR_BAD_COOKIE; 2638 #endif 2639 } 2640 if (!error) 2641 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0); 2642 if (error) { 2643 vput(vp); 2644 nfsm_reply(NFSX_POSTOPATTR(v3)); 2645 nfsm_srvpostop_attr(getret, &at); 2646 return (0); 2647 } 2648 VOP_UNLOCK(vp, 0); 2649 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2650 again: 2651 iv.iov_base = rbuf; 2652 iv.iov_len = fullsiz; 2653 io.uio_iov = &iv; 2654 io.uio_iovcnt = 1; 2655 io.uio_offset = (off_t)off; 2656 io.uio_resid = fullsiz; 2657 io.uio_segflg = UIO_SYSSPACE; 2658 io.uio_rw = UIO_READ; 2659 io.uio_lwp = NULL; 2660 eofflag = 0; 2661 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2662 2663 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2664 2665 off = (off_t)io.uio_offset; 2666 if (!cookies && !error) 2667 error = NFSERR_PERM; 2668 if (v3) { 2669 getret = VOP_GETATTR(vp, &at, cred, lwp); 2670 if (!error) 2671 error = getret; 2672 } 2673 2674 VOP_UNLOCK(vp, 0); 2675 if (error) { 2676 vrele(vp); 2677 free((caddr_t)rbuf, M_TEMP); 2678 if (cookies) 2679 free((caddr_t)cookies, M_TEMP); 2680 nfsm_reply(NFSX_POSTOPATTR(v3)); 2681 nfsm_srvpostop_attr(getret, &at); 2682 return (0); 2683 } 2684 if (io.uio_resid) { 2685 siz -= io.uio_resid; 2686 2687 /* 2688 * If nothing read, return eof 2689 * rpc reply 2690 */ 2691 if (siz == 0) { 2692 vrele(vp); 2693 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + 2694 2 * NFSX_UNSIGNED); 2695 if (v3) { 2696 nfsm_srvpostop_attr(getret, &at); 2697 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2698 txdr_hyper(at.va_filerev, tl); 2699 tl += 2; 2700 } else 2701 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2702 *tl++ = nfs_false; 2703 *tl = nfs_true; 2704 free((caddr_t)rbuf, M_TEMP); 2705 free((caddr_t)cookies, M_TEMP); 2706 return (0); 2707 } 2708 } 2709 2710 /* 2711 * Check for degenerate cases of nothing useful read. 2712 * If so go try again 2713 */ 2714 cpos = rbuf; 2715 cend = rbuf + siz; 2716 dp = (struct dirent *)cpos; 2717 cookiep = cookies; 2718 2719 while (cpos < cend && ncookies > 0 && 2720 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2721 cpos += dp->d_reclen; 2722 dp = (struct dirent *)cpos; 2723 cookiep++; 2724 ncookies--; 2725 } 2726 if (cpos >= cend || ncookies == 0) { 2727 toff = off; 2728 siz = fullsiz; 2729 goto again; 2730 } 2731 2732 len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */ 2733 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz); 2734 if (v3) { 2735 nfsm_srvpostop_attr(getret, &at); 2736 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2737 txdr_hyper(at.va_filerev, tl); 2738 } 2739 mp = mp2 = mb; 2740 bp = bpos; 2741 be = bp + M_TRAILINGSPACE(mp); 2742 2743 /* Loop through the records and build reply */ 2744 while (cpos < cend && ncookies > 0) { 2745 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 2746 nlen = dp->d_namlen; 2747 rem = nfsm_rndup(nlen)-nlen; 2748 len += (4 * NFSX_UNSIGNED + nlen + rem); 2749 if (v3) 2750 len += 2 * NFSX_UNSIGNED; 2751 if (len > cnt) { 2752 eofflag = 0; 2753 break; 2754 } 2755 /* 2756 * Build the directory record xdr from 2757 * the dirent entry. 2758 */ 2759 nfsm_clget; 2760 *tl = nfs_true; 2761 bp += NFSX_UNSIGNED; 2762 if (v3) { 2763 nfsm_clget; 2764 *tl = txdr_unsigned(dp->d_fileno >> 32); 2765 bp += NFSX_UNSIGNED; 2766 } 2767 nfsm_clget; 2768 *tl = txdr_unsigned(dp->d_fileno); 2769 bp += NFSX_UNSIGNED; 2770 nfsm_clget; 2771 *tl = txdr_unsigned(nlen); 2772 bp += NFSX_UNSIGNED; 2773 2774 /* And loop around copying the name */ 2775 xfer = nlen; 2776 cp = dp->d_name; 2777 while (xfer > 0) { 2778 nfsm_clget; 2779 if ((bp+xfer) > be) 2780 tsiz = be-bp; 2781 else 2782 tsiz = xfer; 2783 memcpy(bp, cp, tsiz); 2784 bp += tsiz; 2785 xfer -= tsiz; 2786 if (xfer > 0) 2787 cp += tsiz; 2788 } 2789 /* And null pad to an int32_t boundary */ 2790 for (i = 0; i < rem; i++) 2791 *bp++ = '\0'; 2792 nfsm_clget; 2793 2794 /* Finish off the record */ 2795 txdr_hyper(*cookiep, &jar); 2796 if (v3) { 2797 *tl = jar.nfsuquad[0]; 2798 bp += NFSX_UNSIGNED; 2799 nfsm_clget; 2800 } 2801 *tl = jar.nfsuquad[1]; 2802 bp += NFSX_UNSIGNED; 2803 } 2804 cpos += dp->d_reclen; 2805 dp = (struct dirent *)cpos; 2806 cookiep++; 2807 ncookies--; 2808 } 2809 vrele(vp); 2810 nfsm_clget; 2811 *tl = nfs_false; 2812 bp += NFSX_UNSIGNED; 2813 nfsm_clget; 2814 if (eofflag) 2815 *tl = nfs_true; 2816 else 2817 *tl = nfs_false; 2818 bp += NFSX_UNSIGNED; 2819 if (mp != mb) { 2820 if (bp < be) 2821 mp->m_len = bp - mtod(mp, caddr_t); 2822 } else 2823 mp->m_len += bp - bpos; 2824 free((caddr_t)rbuf, M_TEMP); 2825 free((caddr_t)cookies, M_TEMP); 2826 nfsm_srvdone; 2827 } 2828 2829 int 2830 nfsrv_readdirplus(nfsd, slp, lwp, mrq) 2831 struct nfsrv_descript *nfsd; 2832 struct nfssvc_sock *slp; 2833 struct lwp *lwp; 2834 struct mbuf **mrq; 2835 { 2836 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2837 struct mbuf *nam = nfsd->nd_nam; 2838 caddr_t dpos = nfsd->nd_dpos; 2839 struct ucred *cred = &nfsd->nd_cr; 2840 char *bp, *be; 2841 struct mbuf *mp; 2842 struct dirent *dp; 2843 caddr_t cp; 2844 u_int32_t *tl; 2845 int32_t t1; 2846 caddr_t bpos; 2847 struct mbuf *mb, *mreq, *mp2; 2848 char *cpos, *cend, *cp2, *rbuf; 2849 struct vnode *vp, *nvp; 2850 struct flrep fl; 2851 nfsfh_t nfh; 2852 fhandle_t *fhp, *nfhp = (fhandle_t *)fl.fl_nfh; 2853 struct uio io; 2854 struct iovec iv; 2855 struct vattr va, at, *vap = &va; 2856 struct nfs_fattr *fp; 2857 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2858 int siz, cnt, fullsiz, eofflag, rdonly, cache, dirlen, ncookies; 2859 u_quad_t frev, off, toff, verf; 2860 off_t *cookies = NULL, *cookiep; 2861 2862 fhp = &nfh.fh_generic; 2863 nfsm_srvmtofh(fhp); 2864 nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED); 2865 toff = fxdr_hyper(tl); 2866 tl += 2; 2867 verf = fxdr_hyper(tl); 2868 tl += 2; 2869 siz = fxdr_unsigned(int, *tl++); 2870 cnt = fxdr_unsigned(int, *tl); 2871 off = toff; 2872 siz = ((siz + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1)); 2873 xfer = NFS_SRVMAXDATA(nfsd); 2874 if (siz > xfer) 2875 siz = xfer; 2876 fullsiz = siz; 2877 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 2878 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2879 if (!error && vp->v_type != VDIR) { 2880 error = ENOTDIR; 2881 vput(vp); 2882 } 2883 if (error) { 2884 nfsm_reply(NFSX_UNSIGNED); 2885 nfsm_srvpostop_attr(getret, &at); 2886 return (0); 2887 } 2888 error = getret = VOP_GETATTR(vp, &at, cred, lwp); 2889 #ifdef NFS3_STRICTVERF 2890 /* 2891 * XXX This check is too strict for Solaris 2.5 clients. 2892 */ 2893 if (!error && toff && verf != at.va_filerev) 2894 error = NFSERR_BAD_COOKIE; 2895 #endif 2896 if (!error) { 2897 nqsrv_getl(vp, ND_READ); 2898 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0); 2899 } 2900 if (error) { 2901 vput(vp); 2902 nfsm_reply(NFSX_V3POSTOPATTR); 2903 nfsm_srvpostop_attr(getret, &at); 2904 return (0); 2905 } 2906 VOP_UNLOCK(vp, 0); 2907 2908 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2909 again: 2910 iv.iov_base = rbuf; 2911 iv.iov_len = fullsiz; 2912 io.uio_iov = &iv; 2913 io.uio_iovcnt = 1; 2914 io.uio_offset = (off_t)off; 2915 io.uio_resid = fullsiz; 2916 io.uio_segflg = UIO_SYSSPACE; 2917 io.uio_rw = UIO_READ; 2918 io.uio_lwp = NULL; 2919 eofflag = 0; 2920 2921 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2922 2923 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2924 2925 off = (u_quad_t)io.uio_offset; 2926 getret = VOP_GETATTR(vp, &at, cred, lwp); 2927 2928 VOP_UNLOCK(vp, 0); 2929 2930 /* 2931 * If the VGET operation doesn't work for this filesystem, 2932 * we can't support readdirplus. Returning NOTSUPP should 2933 * make clients fall back to plain readdir. 2934 * There's no need to check for VPTOFH as well, we wouldn't 2935 * even be here otherwise. 2936 */ 2937 if (!getret) { 2938 if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp))) 2939 getret = (getret == EOPNOTSUPP) ? 2940 NFSERR_NOTSUPP : NFSERR_IO; 2941 else 2942 vput(nvp); 2943 } 2944 2945 if (!cookies && !error) 2946 error = NFSERR_PERM; 2947 if (!error) 2948 error = getret; 2949 if (error) { 2950 vrele(vp); 2951 if (cookies) 2952 free((caddr_t)cookies, M_TEMP); 2953 free((caddr_t)rbuf, M_TEMP); 2954 nfsm_reply(NFSX_V3POSTOPATTR); 2955 nfsm_srvpostop_attr(getret, &at); 2956 return (0); 2957 } 2958 if (io.uio_resid) { 2959 siz -= io.uio_resid; 2960 2961 /* 2962 * If nothing read, return eof 2963 * rpc reply 2964 */ 2965 if (siz == 0) { 2966 vrele(vp); 2967 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2968 2 * NFSX_UNSIGNED); 2969 nfsm_srvpostop_attr(getret, &at); 2970 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2971 txdr_hyper(at.va_filerev, tl); 2972 tl += 2; 2973 *tl++ = nfs_false; 2974 *tl = nfs_true; 2975 free((caddr_t)cookies, M_TEMP); 2976 free((caddr_t)rbuf, M_TEMP); 2977 return (0); 2978 } 2979 } 2980 2981 /* 2982 * Check for degenerate cases of nothing useful read. 2983 * If so go try again 2984 */ 2985 cpos = rbuf; 2986 cend = rbuf + siz; 2987 dp = (struct dirent *)cpos; 2988 cookiep = cookies; 2989 2990 while (cpos < cend && ncookies > 0 && 2991 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2992 cpos += dp->d_reclen; 2993 dp = (struct dirent *)cpos; 2994 cookiep++; 2995 ncookies--; 2996 } 2997 if (cpos >= cend || ncookies == 0) { 2998 toff = off; 2999 siz = fullsiz; 3000 goto again; 3001 } 3002 3003 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED; 3004 nfsm_reply(cnt); 3005 nfsm_srvpostop_attr(getret, &at); 3006 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 3007 txdr_hyper(at.va_filerev, tl); 3008 mp = mp2 = mb; 3009 bp = bpos; 3010 be = bp + M_TRAILINGSPACE(mp); 3011 3012 /* Loop through the records and build reply */ 3013 while (cpos < cend && ncookies > 0) { 3014 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 3015 nlen = dp->d_namlen; 3016 rem = nfsm_rndup(nlen)-nlen; 3017 3018 /* 3019 * For readdir_and_lookup get the vnode using 3020 * the file number. 3021 */ 3022 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp)) 3023 goto invalid; 3024 memset((caddr_t)nfhp, 0, NFSX_V3FH); 3025 nfhp->fh_fsid = 3026 nvp->v_mount->mnt_stat.f_fsidx; 3027 if (VFS_VPTOFH(nvp, &nfhp->fh_fid)) { 3028 vput(nvp); 3029 goto invalid; 3030 } 3031 if (VOP_GETATTR(nvp, vap, cred, lwp)) { 3032 vput(nvp); 3033 goto invalid; 3034 } 3035 vput(nvp); 3036 3037 /* 3038 * If either the dircount or maxcount will be 3039 * exceeded, get out now. Both of these lengths 3040 * are calculated conservatively, including all 3041 * XDR overheads. 3042 */ 3043 len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH + 3044 NFSX_V3POSTOPATTR); 3045 dirlen += (6 * NFSX_UNSIGNED + nlen + rem); 3046 if (len > cnt || dirlen > fullsiz) { 3047 eofflag = 0; 3048 break; 3049 } 3050 3051 /* 3052 * Build the directory record xdr from 3053 * the dirent entry. 3054 */ 3055 fp = (struct nfs_fattr *)&fl.fl_fattr; 3056 nfsm_srvfillattr(vap, fp); 3057 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH); 3058 fl.fl_fhok = nfs_true; 3059 fl.fl_postopok = nfs_true; 3060 txdr_hyper(*cookiep, fl.fl_off.nfsuquad); 3061 3062 nfsm_clget; 3063 *tl = nfs_true; 3064 bp += NFSX_UNSIGNED; 3065 nfsm_clget; 3066 *tl = txdr_unsigned(dp->d_fileno >> 32); 3067 bp += NFSX_UNSIGNED; 3068 nfsm_clget; 3069 *tl = txdr_unsigned(dp->d_fileno); 3070 bp += NFSX_UNSIGNED; 3071 nfsm_clget; 3072 *tl = txdr_unsigned(nlen); 3073 bp += NFSX_UNSIGNED; 3074 3075 /* And loop around copying the name */ 3076 xfer = nlen; 3077 cp = dp->d_name; 3078 while (xfer > 0) { 3079 nfsm_clget; 3080 if ((bp + xfer) > be) 3081 tsiz = be - bp; 3082 else 3083 tsiz = xfer; 3084 memcpy(bp, cp, tsiz); 3085 bp += tsiz; 3086 xfer -= tsiz; 3087 if (xfer > 0) 3088 cp += tsiz; 3089 } 3090 /* And null pad to an int32_t boundary */ 3091 for (i = 0; i < rem; i++) 3092 *bp++ = '\0'; 3093 3094 /* 3095 * Now copy the flrep structure out. 3096 */ 3097 xfer = sizeof (struct flrep); 3098 cp = (caddr_t)&fl; 3099 while (xfer > 0) { 3100 nfsm_clget; 3101 if ((bp + xfer) > be) 3102 tsiz = be - bp; 3103 else 3104 tsiz = xfer; 3105 memcpy(bp, cp, tsiz); 3106 bp += tsiz; 3107 xfer -= tsiz; 3108 if (xfer > 0) 3109 cp += tsiz; 3110 } 3111 } 3112 invalid: 3113 cpos += dp->d_reclen; 3114 dp = (struct dirent *)cpos; 3115 cookiep++; 3116 ncookies--; 3117 } 3118 vrele(vp); 3119 nfsm_clget; 3120 *tl = nfs_false; 3121 bp += NFSX_UNSIGNED; 3122 nfsm_clget; 3123 if (eofflag) 3124 *tl = nfs_true; 3125 else 3126 *tl = nfs_false; 3127 bp += NFSX_UNSIGNED; 3128 if (mp != mb) { 3129 if (bp < be) 3130 mp->m_len = bp - mtod(mp, caddr_t); 3131 } else 3132 mp->m_len += bp - bpos; 3133 free((caddr_t)cookies, M_TEMP); 3134 free((caddr_t)rbuf, M_TEMP); 3135 nfsm_srvdone; 3136 } 3137 3138 /* 3139 * nfs commit service 3140 */ 3141 int 3142 nfsrv_commit(nfsd, slp, lwp, mrq) 3143 struct nfsrv_descript *nfsd; 3144 struct nfssvc_sock *slp; 3145 struct lwp *lwp; 3146 struct mbuf **mrq; 3147 { 3148 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3149 struct mbuf *nam = nfsd->nd_nam; 3150 caddr_t dpos = nfsd->nd_dpos; 3151 struct ucred *cred = &nfsd->nd_cr; 3152 struct vattr bfor, aft; 3153 struct vnode *vp; 3154 nfsfh_t nfh; 3155 fhandle_t *fhp; 3156 u_int32_t *tl; 3157 int32_t t1; 3158 caddr_t bpos; 3159 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cache; 3160 uint32_t cnt; 3161 char *cp2; 3162 struct mbuf *mb, *mreq; 3163 u_quad_t frev, off, end; 3164 struct mount *mp = NULL; 3165 3166 #ifndef nolint 3167 cache = 0; 3168 #endif 3169 fhp = &nfh.fh_generic; 3170 nfsm_srvmtofh(fhp); 3171 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 3172 return (ESTALE); 3173 vn_start_write(NULL, &mp, V_WAIT); 3174 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED); 3175 3176 off = fxdr_hyper(tl); 3177 tl += 2; 3178 cnt = fxdr_unsigned(uint32_t, *tl); 3179 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3180 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3181 if (error) { 3182 nfsm_reply(2 * NFSX_UNSIGNED); 3183 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3184 vn_finished_write(mp, 0); 3185 return (0); 3186 } 3187 for_ret = VOP_GETATTR(vp, &bfor, cred, lwp); 3188 end = (cnt > 0) ? off + cnt : vp->v_size; 3189 if (end < off || end > vp->v_size) 3190 end = vp->v_size; 3191 if (off < vp->v_size) 3192 error = VOP_FSYNC(vp, cred, FSYNC_WAIT, off, end, lwp); 3193 /* else error == 0, from nfsrv_fhtovp() */ 3194 aft_ret = VOP_GETATTR(vp, &aft, cred, lwp); 3195 vput(vp); 3196 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF); 3197 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3198 if (!error) { 3199 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF); 3200 *tl++ = txdr_unsigned(boottime.tv_sec); 3201 *tl = txdr_unsigned(boottime.tv_usec); 3202 } else { 3203 vn_finished_write(mp, 0); 3204 return (0); 3205 } 3206 vn_finished_write(mp, 0); 3207 nfsm_srvdone; 3208 } 3209 3210 /* 3211 * nfs statfs service 3212 */ 3213 int 3214 nfsrv_statfs(nfsd, slp, lwp, mrq) 3215 struct nfsrv_descript *nfsd; 3216 struct nfssvc_sock *slp; 3217 struct lwp *lwp; 3218 struct mbuf **mrq; 3219 { 3220 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3221 struct mbuf *nam = nfsd->nd_nam; 3222 caddr_t dpos = nfsd->nd_dpos; 3223 struct ucred *cred = &nfsd->nd_cr; 3224 struct statvfs *sf; 3225 struct nfs_statfs *sfp; 3226 u_int32_t *tl; 3227 int32_t t1; 3228 caddr_t bpos; 3229 int error = 0, rdonly, cache, getret = 1; 3230 int v3 = (nfsd->nd_flag & ND_NFSV3); 3231 char *cp2; 3232 struct mbuf *mb, *mreq; 3233 struct vnode *vp; 3234 struct vattr at; 3235 nfsfh_t nfh; 3236 fhandle_t *fhp; 3237 struct statvfs statvfs; 3238 u_quad_t frev, tval; 3239 3240 #ifndef nolint 3241 cache = 0; 3242 #endif 3243 fhp = &nfh.fh_generic; 3244 nfsm_srvmtofh(fhp); 3245 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3246 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3247 if (error) { 3248 nfsm_reply(NFSX_UNSIGNED); 3249 nfsm_srvpostop_attr(getret, &at); 3250 return (0); 3251 } 3252 sf = &statvfs; 3253 error = VFS_STATVFS(vp->v_mount, sf, lwp); 3254 getret = VOP_GETATTR(vp, &at, cred, lwp); 3255 vput(vp); 3256 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3)); 3257 if (v3) 3258 nfsm_srvpostop_attr(getret, &at); 3259 if (error) 3260 return (0); 3261 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3)); 3262 if (v3) { 3263 tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_frsize); 3264 txdr_hyper(tval, &sfp->sf_tbytes); 3265 tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_frsize); 3266 txdr_hyper(tval, &sfp->sf_fbytes); 3267 tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_frsize); 3268 txdr_hyper(tval, &sfp->sf_abytes); 3269 tval = (u_quad_t)sf->f_files; 3270 txdr_hyper(tval, &sfp->sf_tfiles); 3271 tval = (u_quad_t)sf->f_ffree; 3272 txdr_hyper(tval, &sfp->sf_ffiles); 3273 txdr_hyper(tval, &sfp->sf_afiles); 3274 sfp->sf_invarsec = 0; 3275 } else { 3276 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA); 3277 sfp->sf_bsize = txdr_unsigned(sf->f_frsize); 3278 sfp->sf_blocks = txdr_unsigned(sf->f_blocks); 3279 sfp->sf_bfree = txdr_unsigned(sf->f_bfree); 3280 sfp->sf_bavail = txdr_unsigned(sf->f_bavail); 3281 } 3282 nfsm_srvdone; 3283 } 3284 3285 /* 3286 * nfs fsinfo service 3287 */ 3288 int 3289 nfsrv_fsinfo(nfsd, slp, lwp, mrq) 3290 struct nfsrv_descript *nfsd; 3291 struct nfssvc_sock *slp; 3292 struct lwp *lwp; 3293 struct mbuf **mrq; 3294 { 3295 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3296 struct mbuf *nam = nfsd->nd_nam; 3297 caddr_t dpos = nfsd->nd_dpos; 3298 struct ucred *cred = &nfsd->nd_cr; 3299 u_int32_t *tl; 3300 struct nfsv3_fsinfo *sip; 3301 int32_t t1; 3302 caddr_t bpos; 3303 int error = 0, rdonly, cache, getret = 1; 3304 uint32_t maxdata; 3305 char *cp2; 3306 struct mbuf *mb, *mreq; 3307 struct vnode *vp; 3308 struct vattr at; 3309 nfsfh_t nfh; 3310 fhandle_t *fhp; 3311 u_quad_t frev, maxfsize; 3312 struct statvfs sb; 3313 3314 #ifndef nolint 3315 cache = 0; 3316 #endif 3317 fhp = &nfh.fh_generic; 3318 nfsm_srvmtofh(fhp); 3319 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3320 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3321 if (error) { 3322 nfsm_reply(NFSX_UNSIGNED); 3323 nfsm_srvpostop_attr(getret, &at); 3324 return (0); 3325 } 3326 3327 /* XXX Try to make a guess on the max file size. */ 3328 VFS_STATVFS(vp->v_mount, &sb, (struct lwp *)0); 3329 maxfsize = (u_quad_t)0x80000000 * sb.f_frsize - 1; 3330 3331 getret = VOP_GETATTR(vp, &at, cred, lwp); 3332 vput(vp); 3333 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO); 3334 nfsm_srvpostop_attr(getret, &at); 3335 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO); 3336 3337 /* 3338 * XXX 3339 * There should be file system VFS OP(s) to get this information. 3340 * For now, assume ufs. 3341 */ 3342 if (slp->ns_so->so_type == SOCK_DGRAM) 3343 maxdata = NFS_MAXDGRAMDATA; 3344 else 3345 maxdata = NFS_MAXDATA; 3346 sip->fs_rtmax = txdr_unsigned(maxdata); 3347 sip->fs_rtpref = txdr_unsigned(maxdata); 3348 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE); 3349 sip->fs_wtmax = txdr_unsigned(maxdata); 3350 sip->fs_wtpref = txdr_unsigned(maxdata); 3351 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE); 3352 sip->fs_dtpref = txdr_unsigned(maxdata); 3353 txdr_hyper(maxfsize, &sip->fs_maxfilesize); 3354 sip->fs_timedelta.nfsv3_sec = 0; 3355 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1); 3356 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK | 3357 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS | 3358 NFSV3FSINFO_CANSETTIME); 3359 nfsm_srvdone; 3360 } 3361 3362 /* 3363 * nfs pathconf service 3364 */ 3365 int 3366 nfsrv_pathconf(nfsd, slp, lwp, mrq) 3367 struct nfsrv_descript *nfsd; 3368 struct nfssvc_sock *slp; 3369 struct lwp *lwp; 3370 struct mbuf **mrq; 3371 { 3372 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3373 struct mbuf *nam = nfsd->nd_nam; 3374 caddr_t dpos = nfsd->nd_dpos; 3375 struct ucred *cred = &nfsd->nd_cr; 3376 u_int32_t *tl; 3377 struct nfsv3_pathconf *pc; 3378 int32_t t1; 3379 caddr_t bpos; 3380 int error = 0, rdonly, cache, getret = 1; 3381 register_t linkmax, namemax, chownres, notrunc; 3382 char *cp2; 3383 struct mbuf *mb, *mreq; 3384 struct vnode *vp; 3385 struct vattr at; 3386 nfsfh_t nfh; 3387 fhandle_t *fhp; 3388 u_quad_t frev; 3389 3390 #ifndef nolint 3391 cache = 0; 3392 #endif 3393 fhp = &nfh.fh_generic; 3394 nfsm_srvmtofh(fhp); 3395 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3396 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3397 if (error) { 3398 nfsm_reply(NFSX_UNSIGNED); 3399 nfsm_srvpostop_attr(getret, &at); 3400 return (0); 3401 } 3402 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax); 3403 if (!error) 3404 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax); 3405 if (!error) 3406 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres); 3407 if (!error) 3408 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc); 3409 getret = VOP_GETATTR(vp, &at, cred, lwp); 3410 vput(vp); 3411 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF); 3412 nfsm_srvpostop_attr(getret, &at); 3413 if (error) 3414 return (0); 3415 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF); 3416 3417 pc->pc_linkmax = txdr_unsigned(linkmax); 3418 pc->pc_namemax = txdr_unsigned(namemax); 3419 pc->pc_notrunc = txdr_unsigned(notrunc); 3420 pc->pc_chownrestricted = txdr_unsigned(chownres); 3421 3422 /* 3423 * These should probably be supported by VOP_PATHCONF(), but 3424 * until msdosfs is exportable (why would you want to?), the 3425 * Unix defaults should be ok. 3426 */ 3427 pc->pc_caseinsensitive = nfs_false; 3428 pc->pc_casepreserving = nfs_true; 3429 nfsm_srvdone; 3430 } 3431 3432 /* 3433 * Null operation, used by clients to ping server 3434 */ 3435 /* ARGSUSED */ 3436 int 3437 nfsrv_null(nfsd, slp, lwp, mrq) 3438 struct nfsrv_descript *nfsd; 3439 struct nfssvc_sock *slp; 3440 struct lwp *lwp; 3441 struct mbuf **mrq; 3442 { 3443 struct mbuf *mrep = nfsd->nd_mrep; 3444 caddr_t bpos; 3445 int error = NFSERR_RETVOID, cache = 0; 3446 struct mbuf *mb, *mreq; 3447 u_quad_t frev; 3448 3449 nfsm_reply(0); 3450 return (0); 3451 } 3452 3453 /* 3454 * No operation, used for obsolete procedures 3455 */ 3456 /* ARGSUSED */ 3457 int 3458 nfsrv_noop(nfsd, slp, lwp, mrq) 3459 struct nfsrv_descript *nfsd; 3460 struct nfssvc_sock *slp; 3461 struct lwp *lwp; 3462 struct mbuf **mrq; 3463 { 3464 struct mbuf *mrep = nfsd->nd_mrep; 3465 caddr_t bpos; 3466 int error, cache = 0; 3467 struct mbuf *mb, *mreq; 3468 u_quad_t frev; 3469 3470 if (nfsd->nd_repstat) 3471 error = nfsd->nd_repstat; 3472 else 3473 error = EPROCUNAVAIL; 3474 nfsm_reply(0); 3475 return (0); 3476 } 3477 3478 /* 3479 * Perform access checking for vnodes obtained from file handles that would 3480 * refer to files already opened by a Unix client. You cannot just use 3481 * vn_writechk() and VOP_ACCESS() for two reasons. 3482 * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case 3483 * 2 - The owner is to be given access irrespective of mode bits for some 3484 * operations, so that processes that chmod after opening a file don't 3485 * break. I don't like this because it opens a security hole, but since 3486 * the nfs server opens a security hole the size of a barn door anyhow, 3487 * what the heck. 3488 * 3489 * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS() 3490 * will return EPERM instead of EACCESS. EPERM is always an error. 3491 */ 3492 int 3493 nfsrv_access(vp, flags, cred, rdonly, lwp, override) 3494 struct vnode *vp; 3495 int flags; 3496 struct ucred *cred; 3497 int rdonly; 3498 struct lwp *lwp; 3499 { 3500 struct vattr vattr; 3501 int error; 3502 if (flags & VWRITE) { 3503 /* Just vn_writechk() changed to check rdonly */ 3504 /* 3505 * Disallow write attempts on read-only file systems; 3506 * unless the file is a socket or a block or character 3507 * device resident on the file system. 3508 */ 3509 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 3510 switch (vp->v_type) { 3511 case VREG: 3512 case VDIR: 3513 case VLNK: 3514 return (EROFS); 3515 default: 3516 break; 3517 } 3518 } 3519 3520 /* 3521 * If the vnode is in use as a process's text, 3522 * we can't allow writing. 3523 */ 3524 if (vp->v_flag & VTEXT) 3525 return (ETXTBSY); 3526 } 3527 error = VOP_GETATTR(vp, &vattr, cred, lwp); 3528 if (error) 3529 return (error); 3530 error = VOP_ACCESS(vp, flags, cred, lwp); 3531 /* 3532 * Allow certain operations for the owner (reads and writes 3533 * on files that are already open). 3534 */ 3535 if (override && error == EACCES && cred->cr_uid == vattr.va_uid) 3536 error = 0; 3537 return error; 3538 } 3539
Indexes created Thu Oct 02 14:10:14 GMT 2025