nfs_serv.c revision 1.78
1 /* $NetBSD: nfs_serv.c,v 1.78 2003/06/29 18:43:36 thorpej 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. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the University of 21 * California, Berkeley and its contributors. 22 * 4. Neither the name of the University nor the names of its contributors 23 * may be used to endorse or promote products derived from this software 24 * without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 * 38 * @(#)nfs_serv.c 8.8 (Berkeley) 7/31/95 39 */ 40 41 /* 42 * nfs version 2 and 3 server calls to vnode ops 43 * - these routines generally have 3 phases 44 * 1 - break down and validate rpc request in mbuf list 45 * 2 - do the vnode ops for the request 46 * (surprisingly ?? many are very similar to syscalls in vfs_syscalls.c) 47 * 3 - build the rpc reply in an mbuf list 48 * nb: 49 * - do not mix the phases, since the nfsm_?? macros can return failures 50 * on a bad rpc or similar and do not do any vrele() or vput()'s 51 * 52 * - the nfsm_reply() macro generates an nfs rpc reply with the nfs 53 * error number iff error != 0 whereas 54 * returning an error from the server function implies a fatal error 55 * such as a badly constructed rpc request that should be dropped without 56 * a reply. 57 * For Version 3, nfsm_reply() does not return for the error case, since 58 * most version 3 rpcs return more than the status for error cases. 59 */ 60 61 #include <sys/cdefs.h> 62 __KERNEL_RCSID(0, "$NetBSD: nfs_serv.c,v 1.78 2003/06/29 18:43:36 thorpej Exp $"); 63 64 #include <sys/param.h> 65 #include <sys/systm.h> 66 #include <sys/proc.h> 67 #include <sys/file.h> 68 #include <sys/namei.h> 69 #include <sys/vnode.h> 70 #include <sys/mount.h> 71 #include <sys/socket.h> 72 #include <sys/socketvar.h> 73 #include <sys/mbuf.h> 74 #include <sys/dirent.h> 75 #include <sys/stat.h> 76 #include <sys/kernel.h> 77 #include <ufs/ufs/dir.h> 78 79 #include <uvm/uvm.h> 80 81 #include <nfs/nfsproto.h> 82 #include <nfs/rpcv2.h> 83 #include <nfs/nfs.h> 84 #include <nfs/xdr_subs.h> 85 #include <nfs/nfsm_subs.h> 86 #include <nfs/nqnfs.h> 87 #include <nfs/nfs_var.h> 88 89 /* Global vars */ 90 extern u_int32_t nfs_xdrneg1; 91 extern u_int32_t nfs_false, nfs_true; 92 extern enum vtype nv3tov_type[8]; 93 extern struct nfsstats nfsstats; 94 extern const nfstype nfsv2_type[9]; 95 extern const nfstype nfsv3_type[9]; 96 int nfsrvw_procrastinate = NFS_GATHERDELAY * 1000; 97 int nfsd_use_loan = 1; /* use page-loan for READ OP */ 98 99 /* 100 * nfs v3 access service 101 */ 102 int 103 nfsrv3_access(nfsd, slp, lwp, mrq) 104 struct nfsrv_descript *nfsd; 105 struct nfssvc_sock *slp; 106 struct lwp *lwp; 107 struct mbuf **mrq; 108 { 109 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 110 struct mbuf *nam = nfsd->nd_nam; 111 caddr_t dpos = nfsd->nd_dpos; 112 struct ucred *cred = &nfsd->nd_cr; 113 struct vnode *vp; 114 nfsfh_t nfh; 115 fhandle_t *fhp; 116 u_int32_t *tl; 117 int32_t t1; 118 caddr_t bpos; 119 int error = 0, rdonly, cache = 0, getret; 120 char *cp2; 121 struct mbuf *mb, *mreq; 122 struct vattr va; 123 u_long inmode, testmode, outmode; 124 u_quad_t frev; 125 126 fhp = &nfh.fh_generic; 127 nfsm_srvmtofh(fhp); 128 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 129 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, 130 (nfsd->nd_flag & ND_KERBAUTH), FALSE); 131 if (error) { 132 nfsm_reply(NFSX_UNSIGNED); 133 nfsm_srvpostop_attr(1, (struct vattr *)0); 134 return (0); 135 } 136 inmode = fxdr_unsigned(u_int32_t, *tl); 137 outmode = 0; 138 if ((inmode & NFSV3ACCESS_READ) && 139 nfsrv_access(vp, VREAD, cred, rdonly, lwp, 0) == 0) 140 outmode |= NFSV3ACCESS_READ; 141 if (vp->v_type != VDIR) { 142 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND); 143 if (testmode && 144 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0) 145 outmode |= testmode; 146 if ((inmode & NFSV3ACCESS_EXECUTE) && 147 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0) 148 outmode |= NFSV3ACCESS_EXECUTE; 149 } else { 150 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND | 151 NFSV3ACCESS_DELETE); 152 if (testmode && 153 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0) 154 outmode |= testmode; 155 if ((inmode & NFSV3ACCESS_LOOKUP) && 156 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0) 157 outmode |= NFSV3ACCESS_LOOKUP; 158 } 159 getret = VOP_GETATTR(vp, &va, cred, lwp); 160 vput(vp); 161 nfsm_reply(NFSX_POSTOPATTR(1) + NFSX_UNSIGNED); 162 nfsm_srvpostop_attr(getret, &va); 163 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 164 *tl = txdr_unsigned(outmode); 165 nfsm_srvdone; 166 } 167 168 /* 169 * nfs getattr service 170 */ 171 int 172 nfsrv_getattr(nfsd, slp, lwp, mrq) 173 struct nfsrv_descript *nfsd; 174 struct nfssvc_sock *slp; 175 struct lwp *lwp; 176 struct mbuf **mrq; 177 { 178 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 179 struct mbuf *nam = nfsd->nd_nam; 180 caddr_t dpos = nfsd->nd_dpos; 181 struct ucred *cred = &nfsd->nd_cr; 182 struct nfs_fattr *fp; 183 struct vattr va; 184 struct vnode *vp; 185 nfsfh_t nfh; 186 fhandle_t *fhp; 187 u_int32_t *tl; 188 int32_t t1; 189 caddr_t bpos; 190 int error = 0, rdonly, cache; 191 char *cp2; 192 struct mbuf *mb, *mreq; 193 u_quad_t frev; 194 195 fhp = &nfh.fh_generic; 196 nfsm_srvmtofh(fhp); 197 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, 198 (nfsd->nd_flag & ND_KERBAUTH), FALSE); 199 if (error) { 200 nfsm_reply(0); 201 return (0); 202 } 203 nqsrv_getl(vp, ND_READ); 204 error = VOP_GETATTR(vp, &va, cred, lwp); 205 vput(vp); 206 nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); 207 if (error) 208 return (0); 209 nfsm_build(fp, struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); 210 nfsm_srvfillattr(&va, fp); 211 nfsm_srvdone; 212 } 213 214 /* 215 * nfs setattr service 216 */ 217 int 218 nfsrv_setattr(nfsd, slp, lwp, mrq) 219 struct nfsrv_descript *nfsd; 220 struct nfssvc_sock *slp; 221 struct lwp *lwp; 222 struct mbuf **mrq; 223 { 224 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 225 struct mbuf *nam = nfsd->nd_nam; 226 caddr_t dpos = nfsd->nd_dpos; 227 struct ucred *cred = &nfsd->nd_cr; 228 struct vattr va, preat; 229 struct nfsv2_sattr *sp; 230 struct nfs_fattr *fp; 231 struct vnode *vp; 232 nfsfh_t nfh; 233 fhandle_t *fhp; 234 u_int32_t *tl; 235 int32_t t1; 236 caddr_t bpos; 237 int error = 0, rdonly, cache, preat_ret = 1, postat_ret = 1; 238 int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0; 239 char *cp2; 240 struct mbuf *mb, *mreq; 241 u_quad_t frev; 242 struct timespec guard; 243 244 fhp = &nfh.fh_generic; 245 nfsm_srvmtofh(fhp); 246 VATTR_NULL(&va); 247 if (v3) { 248 nfsm_srvsattr(&va); 249 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 250 gcheck = fxdr_unsigned(int, *tl); 251 if (gcheck) { 252 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 253 fxdr_nfsv3time(tl, &guard); 254 } 255 } else { 256 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 257 /* 258 * Nah nah nah nah na nah 259 * There is a bug in the Sun client that puts 0xffff in the mode 260 * field of sattr when it should put in 0xffffffff. The u_short 261 * doesn't sign extend. 262 * --> check the low order 2 bytes for 0xffff 263 */ 264 if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff) 265 va.va_mode = nfstov_mode(sp->sa_mode); 266 if (sp->sa_uid != nfs_xdrneg1) 267 va.va_uid = fxdr_unsigned(uid_t, sp->sa_uid); 268 if (sp->sa_gid != nfs_xdrneg1) 269 va.va_gid = fxdr_unsigned(gid_t, sp->sa_gid); 270 if (sp->sa_size != nfs_xdrneg1) 271 va.va_size = fxdr_unsigned(u_quad_t, sp->sa_size); 272 if (sp->sa_atime.nfsv2_sec != nfs_xdrneg1) { 273 #ifdef notyet 274 fxdr_nfsv2time(&sp->sa_atime, &va.va_atime); 275 #else 276 va.va_atime.tv_sec = 277 fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec); 278 va.va_atime.tv_nsec = 0; 279 #endif 280 } 281 if (sp->sa_mtime.nfsv2_sec != nfs_xdrneg1) 282 fxdr_nfsv2time(&sp->sa_mtime, &va.va_mtime); 283 284 } 285 286 /* 287 * Now that we have all the fields, lets do it. 288 */ 289 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, 290 (nfsd->nd_flag & ND_KERBAUTH), FALSE); 291 if (error) { 292 nfsm_reply(2 * NFSX_UNSIGNED); 293 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 294 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 return (0); 308 } 309 } 310 311 /* 312 * If the size is being changed write acces is required, otherwise 313 * just check for a read only file system. 314 */ 315 if (va.va_size == ((u_quad_t)((quad_t) -1))) { 316 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 317 error = EROFS; 318 goto out; 319 } 320 } else { 321 if (vp->v_type == VDIR) { 322 error = EISDIR; 323 goto out; 324 } else if ((error = nfsrv_access(vp, VWRITE, cred, rdonly, 325 lwp, 0)) != 0) 326 goto out; 327 } 328 error = VOP_SETATTR(vp, &va, cred, lwp); 329 postat_ret = VOP_GETATTR(vp, &va, cred, lwp); 330 if (!error) 331 error = postat_ret; 332 out: 333 vput(vp); 334 nfsm_reply(NFSX_WCCORFATTR(v3)); 335 if (v3) { 336 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 337 return (0); 338 } else { 339 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 340 nfsm_srvfillattr(&va, fp); 341 } 342 nfsm_srvdone; 343 } 344 345 /* 346 * nfs lookup rpc 347 */ 348 int 349 nfsrv_lookup(nfsd, slp, lwp, mrq) 350 struct nfsrv_descript *nfsd; 351 struct nfssvc_sock *slp; 352 struct lwp *lwp; 353 struct mbuf **mrq; 354 { 355 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 356 struct mbuf *nam = nfsd->nd_nam; 357 caddr_t dpos = nfsd->nd_dpos; 358 struct ucred *cred = &nfsd->nd_cr; 359 struct nfs_fattr *fp; 360 struct nameidata nd, ind, *ndp = &nd; 361 struct vnode *vp, *dirp; 362 nfsfh_t nfh; 363 fhandle_t *fhp; 364 caddr_t cp; 365 u_int32_t *tl; 366 int32_t t1; 367 caddr_t bpos; 368 int error = 0, cache, dirattr_ret = 1; 369 uint32_t len; 370 int v3 = (nfsd->nd_flag & ND_NFSV3), pubflag; 371 char *cp2; 372 struct mbuf *mb, *mreq; 373 struct vattr va, dirattr; 374 u_quad_t frev; 375 376 fhp = &nfh.fh_generic; 377 nfsm_srvmtofh(fhp); 378 nfsm_srvnamesiz(len); 379 380 pubflag = nfs_ispublicfh(fhp); 381 382 nd.ni_cnd.cn_cred = cred; 383 nd.ni_cnd.cn_nameiop = LOOKUP; 384 nd.ni_cnd.cn_flags = LOCKLEAF | SAVESTART; 385 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 386 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), pubflag); 387 388 if (!error && pubflag) { 389 if (nd.ni_vp->v_type == VDIR && nfs_pub.np_index != NULL) { 390 /* 391 * Setup call to lookup() to see if we can find 392 * the index file. Arguably, this doesn't belong 393 * in a kernel.. Ugh. 394 */ 395 ind = nd; 396 VOP_UNLOCK(nd.ni_vp, 0); 397 ind.ni_pathlen = strlen(nfs_pub.np_index); 398 ind.ni_cnd.cn_nameptr = ind.ni_cnd.cn_pnbuf = 399 nfs_pub.np_index; 400 ind.ni_startdir = nd.ni_vp; 401 VREF(ind.ni_startdir); 402 error = lookup(&ind); 403 if (!error) { 404 /* 405 * Found an index file. Get rid of 406 * the old references. 407 */ 408 if (dirp) 409 vrele(dirp); 410 dirp = nd.ni_vp; 411 vrele(nd.ni_startdir); 412 ndp = &ind; 413 } else 414 error = 0; 415 } 416 /* 417 * If the public filehandle was used, check that this lookup 418 * didn't result in a filehandle outside the publicly exported 419 * filesystem. 420 */ 421 422 if (!error && ndp->ni_vp->v_mount != nfs_pub.np_mount) { 423 vput(nd.ni_vp); 424 error = EPERM; 425 } 426 } 427 428 if (dirp) { 429 if (v3) 430 dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred, lwp); 431 vrele(dirp); 432 } 433 434 if (error) { 435 nfsm_reply(NFSX_POSTOPATTR(v3)); 436 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 437 return (0); 438 } 439 440 nqsrv_getl(ndp->ni_startdir, ND_READ); 441 vrele(ndp->ni_startdir); 442 PNBUF_PUT(nd.ni_cnd.cn_pnbuf); 443 vp = ndp->ni_vp; 444 memset((caddr_t)fhp, 0, sizeof(nfh)); 445 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; 446 error = VFS_VPTOFH(vp, &fhp->fh_fid); 447 if (!error) 448 error = VOP_GETATTR(vp, &va, cred, lwp); 449 vput(vp); 450 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPORFATTR(v3) + NFSX_POSTOPATTR(v3)); 451 if (error) { 452 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 453 return (0); 454 } 455 nfsm_srvfhtom(fhp, v3); 456 if (v3) { 457 nfsm_srvpostop_attr(0, &va); 458 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 459 } else { 460 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 461 nfsm_srvfillattr(&va, fp); 462 } 463 nfsm_srvdone; 464 } 465 466 /* 467 * nfs readlink service 468 */ 469 int 470 nfsrv_readlink(nfsd, slp, lwp, mrq) 471 struct nfsrv_descript *nfsd; 472 struct nfssvc_sock *slp; 473 struct lwp *lwp; 474 struct mbuf **mrq; 475 { 476 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 477 struct mbuf *nam = nfsd->nd_nam; 478 caddr_t dpos = nfsd->nd_dpos; 479 struct ucred *cred = &nfsd->nd_cr; 480 struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN]; 481 struct iovec *ivp = iv; 482 struct mbuf *mp; 483 u_int32_t *tl; 484 int32_t t1; 485 caddr_t bpos; 486 int error = 0, rdonly, cache, i, padlen, getret; 487 uint32_t len; 488 int v3 = (nfsd->nd_flag & ND_NFSV3); 489 char *cp2; 490 struct mbuf *mb, *mp2 = NULL, *mp3 = NULL, *mreq; 491 struct vnode *vp; 492 struct vattr attr; 493 nfsfh_t nfh; 494 fhandle_t *fhp; 495 struct uio io, *uiop = &io; 496 u_quad_t frev; 497 498 fhp = &nfh.fh_generic; 499 nfsm_srvmtofh(fhp); 500 len = 0; 501 i = 0; 502 while (len < NFS_MAXPATHLEN) { 503 mp = m_get(M_WAIT, MT_DATA); 504 MCLAIM(mp, &nfs_mowner); 505 m_clget(mp, M_WAIT); 506 mp->m_len = NFSMSIZ(mp); 507 if (len == 0) 508 mp3 = mp2 = mp; 509 else { 510 mp2->m_next = mp; 511 mp2 = mp; 512 } 513 if ((len+mp->m_len) > NFS_MAXPATHLEN) { 514 mp->m_len = NFS_MAXPATHLEN-len; 515 len = NFS_MAXPATHLEN; 516 } else 517 len += mp->m_len; 518 ivp->iov_base = mtod(mp, caddr_t); 519 ivp->iov_len = mp->m_len; 520 i++; 521 ivp++; 522 } 523 uiop->uio_iov = iv; 524 uiop->uio_iovcnt = i; 525 uiop->uio_offset = 0; 526 uiop->uio_resid = len; 527 uiop->uio_rw = UIO_READ; 528 uiop->uio_segflg = UIO_SYSSPACE; 529 uiop->uio_lwp = (struct lwp *)0; 530 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 531 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 532 if (error) { 533 m_freem(mp3); 534 nfsm_reply(2 * NFSX_UNSIGNED); 535 nfsm_srvpostop_attr(1, (struct vattr *)0); 536 return (0); 537 } 538 if (vp->v_type != VLNK) { 539 if (v3) 540 error = EINVAL; 541 else 542 error = ENXIO; 543 goto out; 544 } 545 nqsrv_getl(vp, ND_READ); 546 error = VOP_READLINK(vp, uiop, cred); 547 out: 548 getret = VOP_GETATTR(vp, &attr, cred, lwp); 549 vput(vp); 550 if (error) 551 m_freem(mp3); 552 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED); 553 if (v3) { 554 nfsm_srvpostop_attr(getret, &attr); 555 if (error) 556 return (0); 557 } 558 len -= uiop->uio_resid; 559 padlen = nfsm_padlen(len); 560 if (uiop->uio_resid || padlen) 561 nfs_zeropad(mp3, uiop->uio_resid, padlen); 562 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 563 *tl = txdr_unsigned(len); 564 mb->m_next = mp3; 565 nfsm_srvdone; 566 } 567 568 /* 569 * XXX UBC temp limit 570 * maximum number of pages we can do VOP_GETPAGES and loan-out at once. 571 * should be <= MAX_READ_AHEAD in genfs_vnops.c 572 */ 573 #define NFSD_READ_GETPAGES_CHUNK 16 574 575 /* 576 * nfs read service 577 */ 578 int 579 nfsrv_read(nfsd, slp, lwp, mrq) 580 struct nfsrv_descript *nfsd; 581 struct nfssvc_sock *slp; 582 struct lwp *lwp; 583 struct mbuf **mrq; 584 { 585 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 586 struct mbuf *nam = nfsd->nd_nam; 587 caddr_t dpos = nfsd->nd_dpos; 588 struct ucred *cred = &nfsd->nd_cr; 589 struct mbuf *m; 590 struct nfs_fattr *fp; 591 u_int32_t *tl; 592 int32_t t1; 593 int i; 594 caddr_t bpos; 595 int error = 0, rdonly, cache, getret; 596 int v3 = (nfsd->nd_flag & ND_NFSV3); 597 uint32_t reqlen, len, cnt, left; 598 int padlen; 599 char *cp2; 600 struct mbuf *mb, *mreq; 601 struct vnode *vp; 602 nfsfh_t nfh; 603 fhandle_t *fhp; 604 struct uio io, *uiop = &io; 605 struct vattr va; 606 off_t off; 607 u_quad_t frev; 608 609 fhp = &nfh.fh_generic; 610 nfsm_srvmtofh(fhp); 611 if (v3) { 612 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 613 off = fxdr_hyper(tl); 614 } else { 615 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 616 off = (off_t)fxdr_unsigned(u_int32_t, *tl); 617 } 618 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 619 reqlen = fxdr_unsigned(uint32_t, *tl); 620 reqlen = MIN(reqlen, NFS_SRVMAXDATA(nfsd)); 621 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 622 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 623 if (error) { 624 nfsm_reply(2 * NFSX_UNSIGNED); 625 nfsm_srvpostop_attr(1, (struct vattr *)0); 626 return (0); 627 } 628 if (vp->v_type != VREG) { 629 if (v3) 630 error = EINVAL; 631 else 632 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 633 } 634 if (!error) { 635 nqsrv_getl(vp, ND_READ); 636 if ((error = nfsrv_access(vp, VREAD, cred, rdonly, lwp, 1)) != 0) 637 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 1); 638 } 639 getret = VOP_GETATTR(vp, &va, cred, lwp); 640 if (!error) 641 error = getret; 642 if (error) { 643 vput(vp); 644 nfsm_reply(NFSX_POSTOPATTR(v3)); 645 nfsm_srvpostop_attr(getret, &va); 646 return (0); 647 } 648 if (off >= va.va_size) 649 cnt = 0; 650 else if ((off + reqlen) > va.va_size) 651 cnt = va.va_size - off; 652 else 653 cnt = reqlen; 654 nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt)); 655 if (v3) { 656 nfsm_build(tl, u_int32_t *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED); 657 *tl++ = nfs_true; 658 fp = (struct nfs_fattr *)tl; 659 tl += (NFSX_V3FATTR / sizeof (u_int32_t)); 660 } else { 661 nfsm_build(tl, u_int32_t *, NFSX_V2FATTR + NFSX_UNSIGNED); 662 fp = (struct nfs_fattr *)tl; 663 tl += (NFSX_V2FATTR / sizeof (u_int32_t)); 664 } 665 len = left = cnt; 666 if (cnt > 0) { 667 if (nfsd_use_loan) { 668 struct vm_page **pgpp; 669 voff_t pgoff = trunc_page(off); 670 int orignpages, nleftpages; 671 vaddr_t lva, curlva; 672 673 orignpages = (round_page(off + cnt) - pgoff) 674 >> PAGE_SHIFT; 675 KASSERT(orignpages <= M_EXT_MAXPAGES); /* XXX */ 676 677 lva = sokvaalloc(orignpages << PAGE_SHIFT, slp->ns_so); 678 if (lva == 0) { 679 /* fall back to VOP_READ */ 680 goto loan_fail; 681 } 682 683 m = m_get(M_WAIT, MT_DATA); 684 pgpp = m->m_ext.ext_pgs; 685 686 curlva = lva; 687 nleftpages = orignpages; 688 while (nleftpages > 0) { 689 int npages = nleftpages; 690 if (npages > NFSD_READ_GETPAGES_CHUNK) 691 npages = NFSD_READ_GETPAGES_CHUNK; 692 again: 693 simple_lock(&vp->v_interlock); 694 error = VOP_GETPAGES(vp, pgoff, pgpp, &npages, 695 0, VM_PROT_READ, 0, PGO_SYNCIO); 696 if (error == EAGAIN) { 697 tsleep(&lbolt, PVM, "nfsread", 0); 698 goto again; 699 } 700 if (error) { 701 uvm_unloan(m->m_ext.ext_pgs, 702 orignpages - nleftpages, 703 UVM_LOAN_TOPAGE); 704 sokvafree(lva, 705 orignpages << PAGE_SHIFT); 706 m_free(m); 707 goto read_error; 708 } 709 710 /* loan and unbusy pages */ 711 simple_lock(&vp->v_interlock); 712 for (i = 0; i < npages; i++) { 713 if (pgpp[i]->flags & PG_RELEASED) { 714 uvm_lock_pageq(); 715 uvm_page_unbusy(pgpp, npages); 716 uvm_unlock_pageq(); 717 simple_unlock(&vp->v_interlock); 718 continue; 719 } 720 } 721 uvm_loanuobjpages(pgpp, npages); 722 simple_unlock(&vp->v_interlock); 723 724 /* map pages */ 725 for (i = 0; i < npages; i++) { 726 pmap_kenter_pa(curlva, 727 VM_PAGE_TO_PHYS(pgpp[i]), 728 VM_PROT_READ); 729 curlva += PAGE_SIZE; 730 } 731 nleftpages -= npages; 732 pgpp += npages; 733 pgoff += npages << PAGE_SHIFT; 734 } 735 736 lva += off & PAGE_MASK; 737 738 MCLAIM(m, &nfs_mowner); 739 MEXTADD(m, (void *)lva, cnt, M_MBUF, soloanfree, 740 slp->ns_so); 741 m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP; 742 m->m_len = cnt; 743 744 pmap_update(pmap_kernel()); 745 mb->m_next = m; 746 mb = m; 747 error = 0; 748 uiop->uio_resid = 0; 749 } else { 750 struct iovec *iv; 751 struct iovec *iv2; 752 struct mbuf *m2; 753 int siz; 754 loan_fail: 755 /* 756 * Generate the mbuf list with the uio_iov ref. to it. 757 */ 758 i = 0; 759 m = m2 = mb; 760 while (left > 0) { 761 siz = min(M_TRAILINGSPACE(m), left); 762 if (siz > 0) { 763 left -= siz; 764 i++; 765 } 766 if (left > 0) { 767 m = m_get(M_WAIT, MT_DATA); 768 MCLAIM(m, &nfs_mowner); 769 m_clget(m, M_WAIT); 770 m->m_len = 0; 771 m2->m_next = m; 772 m2 = m; 773 } 774 } 775 iv = malloc(i * sizeof(struct iovec), M_TEMP, M_WAITOK); 776 uiop->uio_iov = iv2 = iv; 777 m = mb; 778 left = cnt; 779 i = 0; 780 while (left > 0) { 781 if (m == NULL) 782 panic("nfsrv_read iov"); 783 siz = min(M_TRAILINGSPACE(m), left); 784 if (siz > 0) { 785 iv->iov_base = mtod(m, caddr_t) + 786 m->m_len; 787 iv->iov_len = siz; 788 m->m_len += siz; 789 left -= siz; 790 iv++; 791 i++; 792 } 793 m = m->m_next; 794 } 795 uiop->uio_iovcnt = i; 796 uiop->uio_offset = off; 797 uiop->uio_resid = cnt; 798 uiop->uio_rw = UIO_READ; 799 uiop->uio_segflg = UIO_SYSSPACE; 800 error = VOP_READ(vp, uiop, IO_NODELOCKED, cred); 801 off = uiop->uio_offset; 802 free((caddr_t)iv2, M_TEMP); 803 } 804 read_error: 805 if (error || (getret = VOP_GETATTR(vp, &va, cred, lwp)) != 0){ 806 if (!error) 807 error = getret; 808 m_freem(mreq); 809 vput(vp); 810 nfsm_reply(NFSX_POSTOPATTR(v3)); 811 nfsm_srvpostop_attr(getret, &va); 812 return (0); 813 } 814 } else { 815 uiop->uio_resid = 0; 816 } 817 vput(vp); 818 nfsm_srvfillattr(&va, fp); 819 len -= uiop->uio_resid; 820 padlen = nfsm_padlen(len); 821 if (uiop->uio_resid || padlen) 822 nfs_zeropad(mb, uiop->uio_resid, padlen); 823 if (v3) { 824 *tl++ = txdr_unsigned(len); 825 if (len < reqlen) 826 *tl++ = nfs_true; 827 else 828 *tl++ = nfs_false; 829 } 830 *tl = txdr_unsigned(len); 831 nfsm_srvdone; 832 } 833 834 /* 835 * nfs write service 836 */ 837 int 838 nfsrv_write(nfsd, slp, lwp, mrq) 839 struct nfsrv_descript *nfsd; 840 struct nfssvc_sock *slp; 841 struct lwp *lwp; 842 struct mbuf **mrq; 843 { 844 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 845 struct mbuf *nam = nfsd->nd_nam; 846 caddr_t dpos = nfsd->nd_dpos; 847 struct ucred *cred = &nfsd->nd_cr; 848 struct iovec *ivp; 849 int i, cnt; 850 struct mbuf *mp; 851 struct nfs_fattr *fp; 852 struct iovec *iv; 853 struct vattr va, forat; 854 u_int32_t *tl; 855 int32_t t1; 856 caddr_t bpos; 857 int error = 0, rdonly, cache, len, forat_ret = 1; 858 int ioflags, aftat_ret = 1, retlen, zeroing, adjust; 859 int stable = NFSV3WRITE_FILESYNC; 860 int v3 = (nfsd->nd_flag & ND_NFSV3); 861 char *cp2; 862 struct mbuf *mb, *mreq; 863 struct vnode *vp; 864 nfsfh_t nfh; 865 fhandle_t *fhp; 866 struct uio io, *uiop = &io; 867 off_t off; 868 u_quad_t frev; 869 870 if (mrep == NULL) { 871 *mrq = NULL; 872 return (0); 873 } 874 fhp = &nfh.fh_generic; 875 nfsm_srvmtofh(fhp); 876 if (v3) { 877 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 878 off = fxdr_hyper(tl); 879 tl += 3; 880 stable = fxdr_unsigned(int, *tl++); 881 } else { 882 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 883 off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 884 tl += 2; 885 } 886 retlen = len = fxdr_unsigned(int32_t, *tl); 887 cnt = i = 0; 888 889 /* 890 * For NFS Version 2, it is not obvious what a write of zero length 891 * should do, but I might as well be consistent with Version 3, 892 * which is to return ok so long as there are no permission problems. 893 */ 894 if (len > 0) { 895 zeroing = 1; 896 mp = mrep; 897 while (mp) { 898 if (mp == md) { 899 zeroing = 0; 900 adjust = dpos - mtod(mp, caddr_t); 901 mp->m_len -= adjust; 902 if (mp->m_len > 0 && adjust > 0) 903 NFSMADV(mp, adjust); 904 } 905 if (zeroing) 906 mp->m_len = 0; 907 else if (mp->m_len > 0) { 908 i += mp->m_len; 909 if (i > len) { 910 mp->m_len -= (i - len); 911 zeroing = 1; 912 } 913 if (mp->m_len > 0) 914 cnt++; 915 } 916 mp = mp->m_next; 917 } 918 } 919 if (len > NFS_MAXDATA || len < 0 || i < len) { 920 error = EIO; 921 nfsm_reply(2 * NFSX_UNSIGNED); 922 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 923 return (0); 924 } 925 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 926 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 927 if (error) { 928 nfsm_reply(2 * NFSX_UNSIGNED); 929 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 930 return (0); 931 } 932 if (v3) 933 forat_ret = VOP_GETATTR(vp, &forat, cred, lwp); 934 if (vp->v_type != VREG) { 935 if (v3) 936 error = EINVAL; 937 else 938 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 939 } 940 if (!error) { 941 nqsrv_getl(vp, ND_WRITE); 942 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1); 943 } 944 if (error) { 945 vput(vp); 946 nfsm_reply(NFSX_WCCDATA(v3)); 947 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 948 return (0); 949 } 950 951 if (len > 0) { 952 ivp = malloc(cnt * sizeof (struct iovec), M_TEMP, M_WAITOK); 953 uiop->uio_iov = iv = ivp; 954 uiop->uio_iovcnt = cnt; 955 mp = mrep; 956 while (mp) { 957 if (mp->m_len > 0) { 958 ivp->iov_base = mtod(mp, caddr_t); 959 ivp->iov_len = mp->m_len; 960 ivp++; 961 } 962 mp = mp->m_next; 963 } 964 965 /* 966 * XXX 967 * The IO_METASYNC flag indicates that all metadata (and not 968 * just enough to ensure data integrity) must be written to 969 * stable storage synchronously. 970 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.) 971 */ 972 if (stable == NFSV3WRITE_UNSTABLE) 973 ioflags = IO_NODELOCKED; 974 else if (stable == NFSV3WRITE_DATASYNC) 975 ioflags = (IO_SYNC | IO_NODELOCKED); 976 else 977 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 978 uiop->uio_resid = len; 979 uiop->uio_rw = UIO_WRITE; 980 uiop->uio_segflg = UIO_SYSSPACE; 981 uiop->uio_lwp = (struct lwp *)0; 982 uiop->uio_offset = off; 983 error = VOP_WRITE(vp, uiop, ioflags, cred); 984 nfsstats.srvvop_writes++; 985 free(iv, M_TEMP); 986 } 987 aftat_ret = VOP_GETATTR(vp, &va, cred, lwp); 988 vput(vp); 989 if (!error) 990 error = aftat_ret; 991 nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) + 992 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3)); 993 if (v3) { 994 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 995 if (error) 996 return (0); 997 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 998 *tl++ = txdr_unsigned(retlen); 999 if (stable == NFSV3WRITE_UNSTABLE) 1000 *tl++ = txdr_unsigned(stable); 1001 else 1002 *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC); 1003 /* 1004 * Actually, there is no need to txdr these fields, 1005 * but it may make the values more human readable, 1006 * for debugging purposes. 1007 */ 1008 *tl++ = txdr_unsigned(boottime.tv_sec); 1009 *tl = txdr_unsigned(boottime.tv_usec); 1010 } else { 1011 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1012 nfsm_srvfillattr(&va, fp); 1013 } 1014 nfsm_srvdone; 1015 } 1016 1017 /* 1018 * NFS write service with write gathering support. Called when 1019 * nfsrvw_procrastinate > 0. 1020 * See: Chet Juszczak, "Improving the Write Performance of an NFS Server", 1021 * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco, 1022 * Jan. 1994. 1023 */ 1024 int 1025 nfsrv_writegather(ndp, slp, lwp, mrq) 1026 struct nfsrv_descript **ndp; 1027 struct nfssvc_sock *slp; 1028 struct lwp *lwp; 1029 struct mbuf **mrq; 1030 { 1031 struct iovec *ivp; 1032 struct mbuf *mp; 1033 struct nfsrv_descript *wp, *nfsd, *owp, *swp; 1034 struct nfs_fattr *fp; 1035 int i = 0; 1036 struct iovec *iov; 1037 struct nfsrvw_delayhash *wpp; 1038 struct ucred *cred; 1039 struct vattr va, forat; 1040 u_int32_t *tl; 1041 int32_t t1; 1042 caddr_t bpos, dpos; 1043 int error = 0, rdonly, cache, len = 0, forat_ret = 1; 1044 int ioflags, aftat_ret = 1, s, adjust, v3, zeroing; 1045 char *cp2; 1046 struct mbuf *mb, *mreq, *mrep, *md; 1047 struct vnode *vp; 1048 struct uio io, *uiop = &io; 1049 u_quad_t frev, cur_usec; 1050 1051 *mrq = NULL; 1052 if (*ndp) { 1053 nfsd = *ndp; 1054 *ndp = NULL; 1055 mrep = nfsd->nd_mrep; 1056 md = nfsd->nd_md; 1057 dpos = nfsd->nd_dpos; 1058 cred = &nfsd->nd_cr; 1059 v3 = (nfsd->nd_flag & ND_NFSV3); 1060 LIST_INIT(&nfsd->nd_coalesce); 1061 nfsd->nd_mreq = NULL; 1062 nfsd->nd_stable = NFSV3WRITE_FILESYNC; 1063 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec; 1064 nfsd->nd_time = cur_usec + nfsrvw_procrastinate; 1065 1066 /* 1067 * Now, get the write header.. 1068 */ 1069 nfsm_srvmtofh(&nfsd->nd_fh); 1070 if (v3) { 1071 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 1072 nfsd->nd_off = fxdr_hyper(tl); 1073 tl += 3; 1074 nfsd->nd_stable = fxdr_unsigned(int, *tl++); 1075 } else { 1076 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1077 nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 1078 tl += 2; 1079 } 1080 len = fxdr_unsigned(int32_t, *tl); 1081 nfsd->nd_len = len; 1082 nfsd->nd_eoff = nfsd->nd_off + len; 1083 1084 /* 1085 * Trim the header out of the mbuf list and trim off any trailing 1086 * junk so that the mbuf list has only the write data. 1087 */ 1088 zeroing = 1; 1089 i = 0; 1090 mp = mrep; 1091 while (mp) { 1092 if (mp == md) { 1093 zeroing = 0; 1094 adjust = dpos - mtod(mp, caddr_t); 1095 mp->m_len -= adjust; 1096 if (mp->m_len > 0 && adjust > 0) 1097 NFSMADV(mp, adjust); 1098 } 1099 if (zeroing) 1100 mp->m_len = 0; 1101 else { 1102 i += mp->m_len; 1103 if (i > len) { 1104 mp->m_len -= (i - len); 1105 zeroing = 1; 1106 } 1107 } 1108 mp = mp->m_next; 1109 } 1110 if (len > NFS_MAXDATA || len < 0 || i < len) { 1111 nfsmout: 1112 m_freem(mrep); 1113 error = EIO; 1114 nfsm_writereply(2 * NFSX_UNSIGNED, v3); 1115 if (v3) 1116 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1117 nfsd->nd_mreq = mreq; 1118 nfsd->nd_mrep = NULL; 1119 nfsd->nd_time = 0; 1120 } 1121 1122 /* 1123 * Add this entry to the hash and time queues. 1124 */ 1125 s = splsoftclock(); 1126 owp = NULL; 1127 wp = LIST_FIRST(&slp->ns_tq); 1128 while (wp && wp->nd_time < nfsd->nd_time) { 1129 owp = wp; 1130 wp = LIST_NEXT(wp, nd_tq); 1131 } 1132 if (owp) { 1133 LIST_INSERT_AFTER(owp, nfsd, nd_tq); 1134 } else { 1135 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1136 } 1137 if (nfsd->nd_mrep) { 1138 wpp = NWDELAYHASH(slp, nfsd->nd_fh.fh_fid.fid_data); 1139 owp = NULL; 1140 wp = LIST_FIRST(wpp); 1141 while (wp && 1142 memcmp((caddr_t)&nfsd->nd_fh, (caddr_t)&wp->nd_fh, NFSX_V3FH)) { 1143 owp = wp; 1144 wp = LIST_NEXT(wp, nd_hash); 1145 } 1146 while (wp && wp->nd_off < nfsd->nd_off && 1147 !memcmp((caddr_t)&nfsd->nd_fh, (caddr_t)&wp->nd_fh, NFSX_V3FH)) { 1148 owp = wp; 1149 wp = LIST_NEXT(wp, nd_hash); 1150 } 1151 if (owp) { 1152 LIST_INSERT_AFTER(owp, nfsd, nd_hash); 1153 1154 /* 1155 * Search the hash list for overlapping entries and 1156 * coalesce. 1157 */ 1158 for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) { 1159 wp = LIST_NEXT(nfsd, nd_hash); 1160 if (NFSW_SAMECRED(owp, nfsd)) 1161 nfsrvw_coalesce(owp, nfsd); 1162 } 1163 } else { 1164 LIST_INSERT_HEAD(wpp, nfsd, nd_hash); 1165 } 1166 } 1167 splx(s); 1168 } 1169 1170 /* 1171 * Now, do VOP_WRITE()s for any one(s) that need to be done now 1172 * and generate the associated reply mbuf list(s). 1173 */ 1174 loop1: 1175 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec; 1176 s = splsoftclock(); 1177 for (nfsd = LIST_FIRST(&slp->ns_tq); nfsd; nfsd = owp) { 1178 owp = LIST_NEXT(nfsd, nd_tq); 1179 if (nfsd->nd_time > cur_usec) 1180 break; 1181 if (nfsd->nd_mreq) 1182 continue; 1183 LIST_REMOVE(nfsd, nd_tq); 1184 LIST_REMOVE(nfsd, nd_hash); 1185 splx(s); 1186 mrep = nfsd->nd_mrep; 1187 nfsd->nd_mrep = NULL; 1188 cred = &nfsd->nd_cr; 1189 v3 = (nfsd->nd_flag & ND_NFSV3); 1190 forat_ret = aftat_ret = 1; 1191 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp, 1192 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH), 1193 FALSE); 1194 if (!error) { 1195 if (v3) 1196 forat_ret = VOP_GETATTR(vp, &forat, cred, lwp); 1197 if (vp->v_type != VREG) { 1198 if (v3) 1199 error = EINVAL; 1200 else 1201 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 1202 } 1203 } else 1204 vp = NULL; 1205 if (!error) { 1206 nqsrv_getl(vp, ND_WRITE); 1207 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1); 1208 } 1209 1210 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE) 1211 ioflags = IO_NODELOCKED; 1212 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC) 1213 ioflags = (IO_SYNC | IO_NODELOCKED); 1214 else 1215 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 1216 uiop->uio_rw = UIO_WRITE; 1217 uiop->uio_segflg = UIO_SYSSPACE; 1218 uiop->uio_lwp = (struct lwp *)0; 1219 uiop->uio_offset = nfsd->nd_off; 1220 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off; 1221 if (uiop->uio_resid > 0) { 1222 mp = mrep; 1223 i = 0; 1224 while (mp) { 1225 if (mp->m_len > 0) 1226 i++; 1227 mp = mp->m_next; 1228 } 1229 uiop->uio_iovcnt = i; 1230 iov = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK); 1231 uiop->uio_iov = ivp = iov; 1232 mp = mrep; 1233 while (mp) { 1234 if (mp->m_len > 0) { 1235 ivp->iov_base = mtod(mp, caddr_t); 1236 ivp->iov_len = mp->m_len; 1237 ivp++; 1238 } 1239 mp = mp->m_next; 1240 } 1241 if (!error) { 1242 error = VOP_WRITE(vp, uiop, ioflags, cred); 1243 nfsstats.srvvop_writes++; 1244 } 1245 free((caddr_t)iov, M_TEMP); 1246 } 1247 m_freem(mrep); 1248 if (vp) { 1249 aftat_ret = VOP_GETATTR(vp, &va, cred, lwp); 1250 vput(vp); 1251 } 1252 1253 /* 1254 * Loop around generating replies for all write rpcs that have 1255 * now been completed. 1256 */ 1257 swp = nfsd; 1258 do { 1259 if (error) { 1260 nfsm_writereply(NFSX_WCCDATA(v3), v3); 1261 if (v3) { 1262 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1263 } 1264 } else { 1265 nfsm_writereply(NFSX_PREOPATTR(v3) + 1266 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED + 1267 NFSX_WRITEVERF(v3), v3); 1268 if (v3) { 1269 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1270 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1271 *tl++ = txdr_unsigned(nfsd->nd_len); 1272 *tl++ = txdr_unsigned(swp->nd_stable); 1273 /* 1274 * Actually, there is no need to txdr these fields, 1275 * but it may make the values more human readable, 1276 * for debugging purposes. 1277 */ 1278 *tl++ = txdr_unsigned(boottime.tv_sec); 1279 *tl = txdr_unsigned(boottime.tv_usec); 1280 } else { 1281 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1282 nfsm_srvfillattr(&va, fp); 1283 } 1284 } 1285 nfsd->nd_mreq = mreq; 1286 if (nfsd->nd_mrep) 1287 panic("nfsrv_write: nd_mrep not free"); 1288 1289 /* 1290 * Done. Put it at the head of the timer queue so that 1291 * the final phase can return the reply. 1292 */ 1293 s = splsoftclock(); 1294 if (nfsd != swp) { 1295 nfsd->nd_time = 0; 1296 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1297 } 1298 nfsd = LIST_FIRST(&swp->nd_coalesce); 1299 if (nfsd) { 1300 LIST_REMOVE(nfsd, nd_tq); 1301 } 1302 splx(s); 1303 } while (nfsd); 1304 s = splsoftclock(); 1305 swp->nd_time = 0; 1306 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq); 1307 splx(s); 1308 goto loop1; 1309 } 1310 splx(s); 1311 1312 /* 1313 * Search for a reply to return. 1314 */ 1315 s = splsoftclock(); 1316 LIST_FOREACH(nfsd, &slp->ns_tq, nd_tq) { 1317 if (nfsd->nd_mreq) { 1318 LIST_REMOVE(nfsd, nd_tq); 1319 *mrq = nfsd->nd_mreq; 1320 *ndp = nfsd; 1321 break; 1322 } 1323 } 1324 splx(s); 1325 return (0); 1326 } 1327 1328 /* 1329 * Coalesce the write request nfsd into owp. To do this we must: 1330 * - remove nfsd from the queues 1331 * - merge nfsd->nd_mrep into owp->nd_mrep 1332 * - update the nd_eoff and nd_stable for owp 1333 * - put nfsd on owp's nd_coalesce list 1334 * NB: Must be called at splsoftclock(). 1335 */ 1336 void 1337 nfsrvw_coalesce(owp, nfsd) 1338 struct nfsrv_descript *owp; 1339 struct nfsrv_descript *nfsd; 1340 { 1341 int overlap; 1342 struct mbuf *mp; 1343 struct nfsrv_descript *m; 1344 1345 LIST_REMOVE(nfsd, nd_hash); 1346 LIST_REMOVE(nfsd, nd_tq); 1347 if (owp->nd_eoff < nfsd->nd_eoff) { 1348 overlap = owp->nd_eoff - nfsd->nd_off; 1349 if (overlap < 0) 1350 panic("nfsrv_coalesce: bad off"); 1351 if (overlap > 0) 1352 m_adj(nfsd->nd_mrep, overlap); 1353 mp = owp->nd_mrep; 1354 while (mp->m_next) 1355 mp = mp->m_next; 1356 mp->m_next = nfsd->nd_mrep; 1357 owp->nd_eoff = nfsd->nd_eoff; 1358 } else 1359 m_freem(nfsd->nd_mrep); 1360 nfsd->nd_mrep = NULL; 1361 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC) 1362 owp->nd_stable = NFSV3WRITE_FILESYNC; 1363 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC && 1364 owp->nd_stable == NFSV3WRITE_UNSTABLE) 1365 owp->nd_stable = NFSV3WRITE_DATASYNC; 1366 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq); 1367 /* 1368 * nfsd might hold coalesce elements! Move them to owp. 1369 * Otherwise, requests may be lost and clients will be stuck. 1370 */ 1371 while ((m = LIST_FIRST(&nfsd->nd_coalesce)) != NULL) { 1372 LIST_REMOVE(m, nd_tq); 1373 LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq); 1374 } 1375 } 1376 1377 /* 1378 * nfs create service 1379 * now does a truncate to 0 length via. setattr if it already exists 1380 */ 1381 int 1382 nfsrv_create(nfsd, slp, lwp, mrq) 1383 struct nfsrv_descript *nfsd; 1384 struct nfssvc_sock *slp; 1385 struct lwp *lwp; 1386 struct mbuf **mrq; 1387 { 1388 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1389 struct mbuf *nam = nfsd->nd_nam; 1390 caddr_t dpos = nfsd->nd_dpos; 1391 struct ucred *cred = &nfsd->nd_cr; 1392 struct nfs_fattr *fp; 1393 struct vattr va, dirfor, diraft; 1394 struct nfsv2_sattr *sp; 1395 u_int32_t *tl; 1396 struct nameidata nd; 1397 caddr_t cp; 1398 int32_t t1; 1399 caddr_t bpos; 1400 int error = 0, cache, len, tsize, dirfor_ret = 1, diraft_ret = 1; 1401 int rdev = 0; 1402 int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0; 1403 char *cp2; 1404 struct mbuf *mb, *mreq; 1405 struct vnode *vp = NULL, *dirp = NULL; 1406 nfsfh_t nfh; 1407 fhandle_t *fhp; 1408 u_quad_t frev, tempsize; 1409 u_char cverf[NFSX_V3CREATEVERF]; 1410 1411 nd.ni_cnd.cn_nameiop = 0; 1412 fhp = &nfh.fh_generic; 1413 nfsm_srvmtofh(fhp); 1414 nfsm_srvnamesiz(len); 1415 nd.ni_cnd.cn_cred = cred; 1416 nd.ni_cnd.cn_nameiop = CREATE; 1417 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1418 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1419 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1420 if (dirp) { 1421 if (v3) 1422 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1423 else { 1424 vrele(dirp); 1425 dirp = (struct vnode *)0; 1426 } 1427 } 1428 if (error) { 1429 nfsm_reply(NFSX_WCCDATA(v3)); 1430 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1431 if (dirp) 1432 vrele(dirp); 1433 return (0); 1434 } 1435 VATTR_NULL(&va); 1436 if (v3) { 1437 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1438 how = fxdr_unsigned(int, *tl); 1439 switch (how) { 1440 case NFSV3CREATE_GUARDED: 1441 if (nd.ni_vp) { 1442 error = EEXIST; 1443 break; 1444 } 1445 case NFSV3CREATE_UNCHECKED: 1446 nfsm_srvsattr(&va); 1447 break; 1448 case NFSV3CREATE_EXCLUSIVE: 1449 nfsm_dissect(cp, caddr_t, NFSX_V3CREATEVERF); 1450 memcpy(cverf, cp, NFSX_V3CREATEVERF); 1451 exclusive_flag = 1; 1452 if (nd.ni_vp == NULL) 1453 va.va_mode = 0; 1454 break; 1455 }; 1456 va.va_type = VREG; 1457 } else { 1458 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 1459 va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode)); 1460 if (va.va_type == VNON) 1461 va.va_type = VREG; 1462 va.va_mode = nfstov_mode(sp->sa_mode); 1463 switch (va.va_type) { 1464 case VREG: 1465 tsize = fxdr_unsigned(int32_t, sp->sa_size); 1466 if (tsize != -1) 1467 va.va_size = (u_quad_t)tsize; 1468 break; 1469 case VCHR: 1470 case VBLK: 1471 case VFIFO: 1472 rdev = fxdr_unsigned(int32_t, sp->sa_size); 1473 break; 1474 default: 1475 break; 1476 }; 1477 } 1478 1479 /* 1480 * Iff doesn't exist, create it 1481 * otherwise just truncate to 0 length 1482 * should I set the mode too ?? 1483 */ 1484 if (nd.ni_vp == NULL) { 1485 if (va.va_type == VREG || va.va_type == VSOCK) { 1486 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1487 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1488 if (!error) { 1489 if (exclusive_flag) { 1490 exclusive_flag = 0; 1491 VATTR_NULL(&va); 1492 memcpy((caddr_t)&va.va_atime, cverf, 1493 NFSX_V3CREATEVERF); 1494 error = VOP_SETATTR(nd.ni_vp, &va, cred, 1495 lwp); 1496 } 1497 } 1498 } else if (va.va_type == VCHR || va.va_type == VBLK || 1499 va.va_type == VFIFO) { 1500 if (va.va_type == VCHR && rdev == 0xffffffff) 1501 va.va_type = VFIFO; 1502 if (va.va_type != VFIFO && 1503 (error = suser(cred, (u_short *)0))) { 1504 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1505 vput(nd.ni_dvp); 1506 nfsm_reply(0); 1507 return (error); 1508 } else 1509 va.va_rdev = (dev_t)rdev; 1510 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1511 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, 1512 &va); 1513 if (error) { 1514 nfsm_reply(0); 1515 } 1516 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1517 vrele(nd.ni_dvp); 1518 vput(nd.ni_vp); 1519 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1520 error = EINVAL; 1521 nfsm_reply(0); 1522 } 1523 } else { 1524 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1525 vput(nd.ni_dvp); 1526 error = ENXIO; 1527 } 1528 vp = nd.ni_vp; 1529 } else { 1530 vp = nd.ni_vp; 1531 if (nd.ni_dvp == vp) 1532 vrele(nd.ni_dvp); 1533 else 1534 vput(nd.ni_dvp); 1535 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1536 if (!error && va.va_size != -1) { 1537 error = nfsrv_access(vp, VWRITE, cred, 1538 (nd.ni_cnd.cn_flags & RDONLY), lwp, 0); 1539 if (!error) { 1540 nqsrv_getl(vp, ND_WRITE); 1541 tempsize = va.va_size; 1542 VATTR_NULL(&va); 1543 va.va_size = tempsize; 1544 error = VOP_SETATTR(vp, &va, cred, lwp); 1545 } 1546 } 1547 if (error) 1548 vput(vp); 1549 } 1550 if (!error) { 1551 memset((caddr_t)fhp, 0, sizeof(nfh)); 1552 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; 1553 error = VFS_VPTOFH(vp, &fhp->fh_fid); 1554 if (!error) 1555 error = VOP_GETATTR(vp, &va, cred, lwp); 1556 vput(vp); 1557 } 1558 if (v3) { 1559 if (exclusive_flag && !error && 1560 memcmp(cverf, (caddr_t)&va.va_atime, NFSX_V3CREATEVERF)) 1561 error = EEXIST; 1562 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1563 vrele(dirp); 1564 } 1565 nfsm_reply(NFSX_SRVFH(v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3)); 1566 if (v3) { 1567 if (!error) { 1568 nfsm_srvpostop_fh(fhp); 1569 nfsm_srvpostop_attr(0, &va); 1570 } 1571 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1572 } else { 1573 nfsm_srvfhtom(fhp, v3); 1574 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1575 nfsm_srvfillattr(&va, fp); 1576 } 1577 return (0); 1578 nfsmout: 1579 if (dirp) 1580 vrele(dirp); 1581 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1582 if (nd.ni_dvp == nd.ni_vp) 1583 vrele(nd.ni_dvp); 1584 else 1585 vput(nd.ni_dvp); 1586 if (nd.ni_vp) 1587 vput(nd.ni_vp); 1588 return (error); 1589 } 1590 1591 /* 1592 * nfs v3 mknod service 1593 */ 1594 int 1595 nfsrv_mknod(nfsd, slp, lwp, mrq) 1596 struct nfsrv_descript *nfsd; 1597 struct nfssvc_sock *slp; 1598 struct lwp *lwp; 1599 struct mbuf **mrq; 1600 { 1601 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1602 struct mbuf *nam = nfsd->nd_nam; 1603 caddr_t dpos = nfsd->nd_dpos; 1604 struct ucred *cred = &nfsd->nd_cr; 1605 struct vattr va, dirfor, diraft; 1606 u_int32_t *tl; 1607 struct nameidata nd; 1608 int32_t t1; 1609 caddr_t bpos; 1610 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 1611 u_int32_t major, minor; 1612 enum vtype vtyp; 1613 char *cp2; 1614 struct mbuf *mb, *mreq; 1615 struct vnode *vp, *dirp = (struct vnode *)0; 1616 nfsfh_t nfh; 1617 fhandle_t *fhp; 1618 u_quad_t frev; 1619 1620 nd.ni_cnd.cn_nameiop = 0; 1621 fhp = &nfh.fh_generic; 1622 nfsm_srvmtofh(fhp); 1623 nfsm_srvnamesiz(len); 1624 nd.ni_cnd.cn_cred = cred; 1625 nd.ni_cnd.cn_nameiop = CREATE; 1626 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1627 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1628 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1629 if (dirp) 1630 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1631 if (error) { 1632 nfsm_reply(NFSX_WCCDATA(1)); 1633 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1634 if (dirp) 1635 vrele(dirp); 1636 return (0); 1637 } 1638 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1639 vtyp = nfsv3tov_type(*tl); 1640 if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) { 1641 error = NFSERR_BADTYPE; 1642 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1643 if (nd.ni_dvp == nd.ni_vp) 1644 vrele(nd.ni_dvp); 1645 else 1646 vput(nd.ni_dvp); 1647 if (nd.ni_vp) 1648 vput(nd.ni_vp); 1649 goto out; 1650 } 1651 VATTR_NULL(&va); 1652 nfsm_srvsattr(&va); 1653 if (vtyp == VCHR || vtyp == VBLK) { 1654 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1655 major = fxdr_unsigned(u_int32_t, *tl++); 1656 minor = fxdr_unsigned(u_int32_t, *tl); 1657 va.va_rdev = makedev(major, minor); 1658 } 1659 1660 /* 1661 * Iff doesn't exist, create it. 1662 */ 1663 if (nd.ni_vp) { 1664 error = EEXIST; 1665 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1666 if (nd.ni_dvp == nd.ni_vp) 1667 vrele(nd.ni_dvp); 1668 else 1669 vput(nd.ni_dvp); 1670 vput(nd.ni_vp); 1671 goto out; 1672 } 1673 va.va_type = vtyp; 1674 if (vtyp == VSOCK) { 1675 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1676 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1677 } else { 1678 if (va.va_type != VFIFO && 1679 (error = suser(cred, (u_short *)0))) { 1680 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1681 vput(nd.ni_dvp); 1682 goto out; 1683 } 1684 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1685 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1686 if (error) { 1687 goto out; 1688 } 1689 if (error) 1690 goto out; 1691 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1692 vput(nd.ni_vp); 1693 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1694 error = EINVAL; 1695 } 1696 } 1697 out: 1698 vp = nd.ni_vp; 1699 if (!error) { 1700 memset((caddr_t)fhp, 0, sizeof(nfh)); 1701 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; 1702 error = VFS_VPTOFH(vp, &fhp->fh_fid); 1703 if (!error) 1704 error = VOP_GETATTR(vp, &va, cred, lwp); 1705 vput(vp); 1706 } 1707 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1708 vrele(dirp); 1709 nfsm_reply(NFSX_SRVFH(1) + NFSX_POSTOPATTR(1) + NFSX_WCCDATA(1)); 1710 if (!error) { 1711 nfsm_srvpostop_fh(fhp); 1712 nfsm_srvpostop_attr(0, &va); 1713 } 1714 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1715 return (0); 1716 nfsmout: 1717 if (dirp) 1718 vrele(dirp); 1719 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1720 if (nd.ni_dvp == nd.ni_vp) 1721 vrele(nd.ni_dvp); 1722 else 1723 vput(nd.ni_dvp); 1724 if (nd.ni_vp) 1725 vput(nd.ni_vp); 1726 return (error); 1727 } 1728 1729 /* 1730 * nfs remove service 1731 */ 1732 int 1733 nfsrv_remove(nfsd, slp, lwp, mrq) 1734 struct nfsrv_descript *nfsd; 1735 struct nfssvc_sock *slp; 1736 struct lwp *lwp; 1737 struct mbuf **mrq; 1738 { 1739 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1740 struct mbuf *nam = nfsd->nd_nam; 1741 caddr_t dpos = nfsd->nd_dpos; 1742 struct ucred *cred = &nfsd->nd_cr; 1743 struct nameidata nd; 1744 u_int32_t *tl; 1745 int32_t t1; 1746 caddr_t bpos; 1747 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 1748 int v3 = (nfsd->nd_flag & ND_NFSV3); 1749 char *cp2; 1750 struct mbuf *mb, *mreq; 1751 struct vnode *vp, *dirp; 1752 struct vattr dirfor, diraft; 1753 nfsfh_t nfh; 1754 fhandle_t *fhp; 1755 u_quad_t frev; 1756 1757 #ifndef nolint 1758 vp = (struct vnode *)0; 1759 #endif 1760 fhp = &nfh.fh_generic; 1761 nfsm_srvmtofh(fhp); 1762 nfsm_srvnamesiz(len); 1763 nd.ni_cnd.cn_cred = cred; 1764 nd.ni_cnd.cn_nameiop = DELETE; 1765 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1766 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1767 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1768 if (dirp) { 1769 if (v3) 1770 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1771 else 1772 vrele(dirp); 1773 } 1774 if (!error) { 1775 vp = nd.ni_vp; 1776 if (vp->v_type == VDIR && 1777 (error = suser(cred, (u_short *)0)) != 0) 1778 goto out; 1779 /* 1780 * The root of a mounted filesystem cannot be deleted. 1781 */ 1782 if (vp->v_flag & VROOT) { 1783 error = EBUSY; 1784 goto out; 1785 } 1786 out: 1787 if (!error) { 1788 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1789 nqsrv_getl(vp, ND_WRITE); 1790 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1791 } else { 1792 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1793 if (nd.ni_dvp == vp) 1794 vrele(nd.ni_dvp); 1795 else 1796 vput(nd.ni_dvp); 1797 vput(vp); 1798 } 1799 } 1800 if (dirp && v3) { 1801 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1802 vrele(dirp); 1803 } 1804 nfsm_reply(NFSX_WCCDATA(v3)); 1805 if (v3) { 1806 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1807 return (0); 1808 } 1809 nfsm_srvdone; 1810 } 1811 1812 /* 1813 * nfs rename service 1814 */ 1815 int 1816 nfsrv_rename(nfsd, slp, lwp, mrq) 1817 struct nfsrv_descript *nfsd; 1818 struct nfssvc_sock *slp; 1819 struct lwp *lwp; 1820 struct mbuf **mrq; 1821 { 1822 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1823 struct mbuf *nam = nfsd->nd_nam; 1824 caddr_t dpos = nfsd->nd_dpos; 1825 struct ucred *cred = &nfsd->nd_cr; 1826 u_int32_t *tl; 1827 int32_t t1; 1828 caddr_t bpos; 1829 int error = 0, cache, fdirfor_ret = 1, fdiraft_ret = 1; 1830 uint32_t len, len2; 1831 int tdirfor_ret = 1, tdiraft_ret = 1; 1832 int v3 = (nfsd->nd_flag & ND_NFSV3); 1833 char *cp2; 1834 struct mbuf *mb, *mreq; 1835 struct nameidata fromnd, tond; 1836 struct vnode *fvp, *tvp, *tdvp, *fdirp = (struct vnode *)0; 1837 struct vnode *tdirp = (struct vnode *)0; 1838 struct vattr fdirfor, fdiraft, tdirfor, tdiraft; 1839 nfsfh_t fnfh, tnfh; 1840 fhandle_t *ffhp, *tfhp; 1841 u_quad_t frev; 1842 uid_t saved_uid; 1843 1844 #ifndef nolint 1845 fvp = (struct vnode *)0; 1846 #endif 1847 ffhp = &fnfh.fh_generic; 1848 tfhp = &tnfh.fh_generic; 1849 fromnd.ni_cnd.cn_nameiop = 0; 1850 tond.ni_cnd.cn_nameiop = 0; 1851 nfsm_srvmtofh(ffhp); 1852 nfsm_srvnamesiz(len); 1853 /* 1854 * Remember our original uid so that we can reset cr_uid before 1855 * the second nfs_namei() call, in case it is remapped. 1856 */ 1857 saved_uid = cred->cr_uid; 1858 fromnd.ni_cnd.cn_cred = cred; 1859 fromnd.ni_cnd.cn_nameiop = DELETE; 1860 fromnd.ni_cnd.cn_flags = WANTPARENT | SAVESTART; 1861 error = nfs_namei(&fromnd, ffhp, len, slp, nam, &md, 1862 &dpos, &fdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1863 if (fdirp) { 1864 if (v3) 1865 fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred, lwp); 1866 else { 1867 vrele(fdirp); 1868 fdirp = (struct vnode *)0; 1869 } 1870 } 1871 if (error) { 1872 nfsm_reply(2 * NFSX_WCCDATA(v3)); 1873 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 1874 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 1875 if (fdirp) 1876 vrele(fdirp); 1877 return (0); 1878 } 1879 fvp = fromnd.ni_vp; 1880 nfsm_srvmtofh(tfhp); 1881 if (v3) { 1882 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 1883 len2 = fxdr_unsigned(uint32_t, *tl); 1884 /* len2 will be checked by nfs_namei */ 1885 } 1886 else { 1887 /* NFSv2 */ 1888 nfsm_strsiz(len2, NFS_MAXNAMLEN); 1889 } 1890 cred->cr_uid = saved_uid; 1891 tond.ni_cnd.cn_cred = cred; 1892 tond.ni_cnd.cn_nameiop = RENAME; 1893 tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART; 1894 error = nfs_namei(&tond, tfhp, len2, slp, nam, &md, 1895 &dpos, &tdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1896 if (tdirp) { 1897 if (v3) 1898 tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred, lwp); 1899 else { 1900 vrele(tdirp); 1901 tdirp = (struct vnode *)0; 1902 } 1903 } 1904 if (error) { 1905 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1906 vrele(fromnd.ni_dvp); 1907 vrele(fvp); 1908 goto out1; 1909 } 1910 tdvp = tond.ni_dvp; 1911 tvp = tond.ni_vp; 1912 if (tvp != NULL) { 1913 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 1914 if (v3) 1915 error = EEXIST; 1916 else 1917 error = EISDIR; 1918 goto out; 1919 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 1920 if (v3) 1921 error = EEXIST; 1922 else 1923 error = ENOTDIR; 1924 goto out; 1925 } 1926 if (tvp->v_type == VDIR && tvp->v_mountedhere) { 1927 if (v3) 1928 error = EXDEV; 1929 else 1930 error = ENOTEMPTY; 1931 goto out; 1932 } 1933 } 1934 if (fvp->v_type == VDIR && fvp->v_mountedhere) { 1935 if (v3) 1936 error = EXDEV; 1937 else 1938 error = ENOTEMPTY; 1939 goto out; 1940 } 1941 if (fvp->v_mount != tdvp->v_mount) { 1942 if (v3) 1943 error = EXDEV; 1944 else 1945 error = ENOTEMPTY; 1946 goto out; 1947 } 1948 if (fvp == tdvp) { 1949 if (v3) 1950 error = EINVAL; 1951 else 1952 error = ENOTEMPTY; 1953 } 1954 /* 1955 * If source is the same as the destination (that is the 1956 * same vnode with the same name in the same directory), 1957 * then there is nothing to do. 1958 */ 1959 if (fvp == tvp && fromnd.ni_dvp == tdvp && 1960 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 1961 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr, 1962 fromnd.ni_cnd.cn_namelen)) 1963 error = -1; 1964 out: 1965 if (!error) { 1966 nqsrv_getl(fromnd.ni_dvp, ND_WRITE); 1967 nqsrv_getl(tdvp, ND_WRITE); 1968 if (tvp) { 1969 nqsrv_getl(tvp, ND_WRITE); 1970 } 1971 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 1972 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 1973 } else { 1974 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 1975 if (tdvp == tvp) 1976 vrele(tdvp); 1977 else 1978 vput(tdvp); 1979 if (tvp) 1980 vput(tvp); 1981 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1982 vrele(fromnd.ni_dvp); 1983 vrele(fvp); 1984 if (error == -1) 1985 error = 0; 1986 } 1987 vrele(tond.ni_startdir); 1988 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 1989 out1: 1990 if (fdirp) { 1991 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred, lwp); 1992 vrele(fdirp); 1993 } 1994 if (tdirp) { 1995 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred, lwp); 1996 vrele(tdirp); 1997 } 1998 vrele(fromnd.ni_startdir); 1999 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2000 nfsm_reply(2 * NFSX_WCCDATA(v3)); 2001 if (v3) { 2002 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 2003 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 2004 } 2005 return (0); 2006 2007 nfsmout: 2008 if (fdirp) 2009 vrele(fdirp); 2010 if (tdirp) 2011 vrele(tdirp); 2012 if (tond.ni_cnd.cn_nameiop) { 2013 vrele(tond.ni_startdir); 2014 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 2015 } 2016 if (fromnd.ni_cnd.cn_nameiop) { 2017 vrele(fromnd.ni_startdir); 2018 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2019 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2020 vrele(fromnd.ni_dvp); 2021 vrele(fvp); 2022 } 2023 return (error); 2024 } 2025 2026 /* 2027 * nfs link service 2028 */ 2029 int 2030 nfsrv_link(nfsd, slp, lwp, mrq) 2031 struct nfsrv_descript *nfsd; 2032 struct nfssvc_sock *slp; 2033 struct lwp *lwp; 2034 struct mbuf **mrq; 2035 { 2036 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2037 struct mbuf *nam = nfsd->nd_nam; 2038 caddr_t dpos = nfsd->nd_dpos; 2039 struct ucred *cred = &nfsd->nd_cr; 2040 struct nameidata nd; 2041 u_int32_t *tl; 2042 int32_t t1; 2043 caddr_t bpos; 2044 int error = 0, rdonly, cache, len, dirfor_ret = 1, diraft_ret = 1; 2045 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3); 2046 char *cp2; 2047 struct mbuf *mb, *mreq; 2048 struct vnode *vp, *xp, *dirp = (struct vnode *)0; 2049 struct vattr dirfor, diraft, at; 2050 nfsfh_t nfh, dnfh; 2051 fhandle_t *fhp, *dfhp; 2052 u_quad_t frev; 2053 2054 fhp = &nfh.fh_generic; 2055 dfhp = &dnfh.fh_generic; 2056 nfsm_srvmtofh(fhp); 2057 nfsm_srvmtofh(dfhp); 2058 nfsm_srvnamesiz(len); 2059 error = nfsrv_fhtovp(fhp, FALSE, &vp, cred, slp, nam, 2060 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2061 if (error) { 2062 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2063 nfsm_srvpostop_attr(getret, &at); 2064 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2065 return (0); 2066 } 2067 if (vp->v_type == VDIR && (error = suser(cred, (u_short *)0)) != 0) 2068 goto out1; 2069 nd.ni_cnd.cn_cred = cred; 2070 nd.ni_cnd.cn_nameiop = CREATE; 2071 nd.ni_cnd.cn_flags = LOCKPARENT; 2072 error = nfs_namei(&nd, dfhp, len, slp, nam, &md, &dpos, 2073 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2074 if (dirp) { 2075 if (v3) 2076 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2077 else { 2078 vrele(dirp); 2079 dirp = (struct vnode *)0; 2080 } 2081 } 2082 if (error) 2083 goto out1; 2084 xp = nd.ni_vp; 2085 if (xp != NULL) { 2086 error = EEXIST; 2087 goto out; 2088 } 2089 xp = nd.ni_dvp; 2090 if (vp->v_mount != xp->v_mount) 2091 error = EXDEV; 2092 out: 2093 if (!error) { 2094 nqsrv_getl(vp, ND_WRITE); 2095 nqsrv_getl(xp, ND_WRITE); 2096 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2097 } else { 2098 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2099 if (nd.ni_dvp == nd.ni_vp) 2100 vrele(nd.ni_dvp); 2101 else 2102 vput(nd.ni_dvp); 2103 if (nd.ni_vp) 2104 vrele(nd.ni_vp); 2105 } 2106 out1: 2107 if (v3) 2108 getret = VOP_GETATTR(vp, &at, cred, lwp); 2109 if (dirp) { 2110 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2111 vrele(dirp); 2112 } 2113 vrele(vp); 2114 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2115 if (v3) { 2116 nfsm_srvpostop_attr(getret, &at); 2117 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2118 return (0); 2119 } 2120 nfsm_srvdone; 2121 } 2122 2123 /* 2124 * nfs symbolic link service 2125 */ 2126 int 2127 nfsrv_symlink(nfsd, slp, lwp, mrq) 2128 struct nfsrv_descript *nfsd; 2129 struct nfssvc_sock *slp; 2130 struct lwp *lwp; 2131 struct mbuf **mrq; 2132 { 2133 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2134 struct mbuf *nam = nfsd->nd_nam; 2135 caddr_t dpos = nfsd->nd_dpos; 2136 struct ucred *cred = &nfsd->nd_cr; 2137 struct vattr va, dirfor, diraft; 2138 struct nameidata nd; 2139 u_int32_t *tl; 2140 int32_t t1; 2141 struct nfsv2_sattr *sp; 2142 char *bpos, *pathcp = NULL, *cp2; 2143 struct uio io; 2144 struct iovec iv; 2145 int error = 0, cache, dirfor_ret = 1, diraft_ret = 1; 2146 uint32_t len, len2; 2147 int v3 = (nfsd->nd_flag & ND_NFSV3); 2148 struct mbuf *mb, *mreq; 2149 struct vnode *dirp = (struct vnode *)0; 2150 nfsfh_t nfh; 2151 fhandle_t *fhp; 2152 u_quad_t frev; 2153 2154 nd.ni_cnd.cn_nameiop = 0; 2155 fhp = &nfh.fh_generic; 2156 nfsm_srvmtofh(fhp); 2157 nfsm_srvnamesiz(len); 2158 nd.ni_cnd.cn_cred = cred; 2159 nd.ni_cnd.cn_nameiop = CREATE; 2160 nd.ni_cnd.cn_flags = LOCKPARENT; 2161 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2162 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2163 if (dirp) { 2164 if (v3) 2165 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2166 else { 2167 vrele(dirp); 2168 dirp = (struct vnode *)0; 2169 } 2170 } 2171 if (error) 2172 goto out; 2173 VATTR_NULL(&va); 2174 if (v3) { 2175 nfsm_srvsattr(&va); 2176 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 2177 len2 = fxdr_unsigned(uint32_t, *tl); 2178 if (len2 > PATH_MAX) { 2179 /* XXX should check _PC_NO_TRUNC */ 2180 error = ENAMETOOLONG; 2181 goto abortop; 2182 } 2183 } 2184 else { 2185 /* NFSv2 */ 2186 nfsm_strsiz(len2, NFS_MAXPATHLEN); 2187 } 2188 pathcp = malloc(len2 + 1, M_TEMP, M_WAITOK); 2189 iv.iov_base = pathcp; 2190 iv.iov_len = len2; 2191 io.uio_resid = len2; 2192 io.uio_offset = 0; 2193 io.uio_iov = &iv; 2194 io.uio_iovcnt = 1; 2195 io.uio_segflg = UIO_SYSSPACE; 2196 io.uio_rw = UIO_READ; 2197 io.uio_lwp = (struct lwp *)0; 2198 nfsm_mtouio(&io, len2); 2199 if (!v3) { 2200 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 2201 va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode); 2202 } 2203 *(pathcp + len2) = '\0'; 2204 if (nd.ni_vp) { 2205 error = EEXIST; 2206 abortop: 2207 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2208 if (nd.ni_dvp == nd.ni_vp) 2209 vrele(nd.ni_dvp); 2210 else 2211 vput(nd.ni_dvp); 2212 if (nd.ni_vp) 2213 vrele(nd.ni_vp); 2214 goto out; 2215 } 2216 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2217 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp); 2218 if (!error) { 2219 if (v3) { 2220 memset((caddr_t)fhp, 0, sizeof(nfh)); 2221 fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsid; 2222 error = VFS_VPTOFH(nd.ni_vp, &fhp->fh_fid); 2223 if (!error) 2224 error = VOP_GETATTR(nd.ni_vp, &va, cred, lwp); 2225 vput(nd.ni_vp); 2226 } else { 2227 vput(nd.ni_vp); 2228 } 2229 } 2230 out: 2231 if (pathcp) 2232 free(pathcp, M_TEMP); 2233 if (dirp) { 2234 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2235 vrele(dirp); 2236 } 2237 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2238 if (v3) { 2239 if (!error) { 2240 nfsm_srvpostop_fh(fhp); 2241 nfsm_srvpostop_attr(0, &va); 2242 } 2243 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2244 } 2245 return (0); 2246 nfsmout: 2247 if (dirp) 2248 vrele(dirp); 2249 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2250 if (nd.ni_dvp == nd.ni_vp) 2251 vrele(nd.ni_dvp); 2252 else 2253 vput(nd.ni_dvp); 2254 if (nd.ni_vp) 2255 vrele(nd.ni_vp); 2256 if (pathcp) 2257 free(pathcp, M_TEMP); 2258 return (error); 2259 } 2260 2261 /* 2262 * nfs mkdir service 2263 */ 2264 int 2265 nfsrv_mkdir(nfsd, slp, lwp, mrq) 2266 struct nfsrv_descript *nfsd; 2267 struct nfssvc_sock *slp; 2268 struct lwp *lwp; 2269 struct mbuf **mrq; 2270 { 2271 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2272 struct mbuf *nam = nfsd->nd_nam; 2273 caddr_t dpos = nfsd->nd_dpos; 2274 struct ucred *cred = &nfsd->nd_cr; 2275 struct vattr va, dirfor, diraft; 2276 struct nfs_fattr *fp; 2277 struct nameidata nd; 2278 caddr_t cp; 2279 u_int32_t *tl; 2280 int32_t t1; 2281 caddr_t bpos; 2282 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 2283 int v3 = (nfsd->nd_flag & ND_NFSV3); 2284 char *cp2; 2285 struct mbuf *mb, *mreq; 2286 struct vnode *vp, *dirp = (struct vnode *)0; 2287 nfsfh_t nfh; 2288 fhandle_t *fhp; 2289 u_quad_t frev; 2290 2291 fhp = &nfh.fh_generic; 2292 nfsm_srvmtofh(fhp); 2293 nfsm_srvnamesiz(len); 2294 nd.ni_cnd.cn_cred = cred; 2295 nd.ni_cnd.cn_nameiop = CREATE; 2296 nd.ni_cnd.cn_flags = LOCKPARENT; 2297 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2298 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2299 if (dirp) { 2300 if (v3) 2301 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2302 else { 2303 vrele(dirp); 2304 dirp = (struct vnode *)0; 2305 } 2306 } 2307 if (error) { 2308 nfsm_reply(NFSX_WCCDATA(v3)); 2309 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2310 if (dirp) 2311 vrele(dirp); 2312 return (0); 2313 } 2314 VATTR_NULL(&va); 2315 if (v3) { 2316 nfsm_srvsattr(&va); 2317 } else { 2318 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 2319 va.va_mode = nfstov_mode(*tl++); 2320 } 2321 va.va_type = VDIR; 2322 vp = nd.ni_vp; 2323 if (vp != NULL) { 2324 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2325 if (nd.ni_dvp == vp) 2326 vrele(nd.ni_dvp); 2327 else 2328 vput(nd.ni_dvp); 2329 vrele(vp); 2330 error = EEXIST; 2331 goto out; 2332 } 2333 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2334 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 2335 if (!error) { 2336 vp = nd.ni_vp; 2337 memset((caddr_t)fhp, 0, sizeof(nfh)); 2338 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; 2339 error = VFS_VPTOFH(vp, &fhp->fh_fid); 2340 if (!error) 2341 error = VOP_GETATTR(vp, &va, cred, lwp); 2342 vput(vp); 2343 } 2344 out: 2345 if (dirp) { 2346 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2347 vrele(dirp); 2348 } 2349 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2350 if (v3) { 2351 if (!error) { 2352 nfsm_srvpostop_fh(fhp); 2353 nfsm_srvpostop_attr(0, &va); 2354 } 2355 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2356 } else { 2357 nfsm_srvfhtom(fhp, v3); 2358 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 2359 nfsm_srvfillattr(&va, fp); 2360 } 2361 return (0); 2362 nfsmout: 2363 if (dirp) 2364 vrele(dirp); 2365 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2366 if (nd.ni_dvp == nd.ni_vp) 2367 vrele(nd.ni_dvp); 2368 else 2369 vput(nd.ni_dvp); 2370 if (nd.ni_vp) 2371 vrele(nd.ni_vp); 2372 return (error); 2373 } 2374 2375 /* 2376 * nfs rmdir service 2377 */ 2378 int 2379 nfsrv_rmdir(nfsd, slp, lwp, mrq) 2380 struct nfsrv_descript *nfsd; 2381 struct nfssvc_sock *slp; 2382 struct lwp *lwp; 2383 struct mbuf **mrq; 2384 { 2385 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2386 struct mbuf *nam = nfsd->nd_nam; 2387 caddr_t dpos = nfsd->nd_dpos; 2388 struct ucred *cred = &nfsd->nd_cr; 2389 u_int32_t *tl; 2390 int32_t t1; 2391 caddr_t bpos; 2392 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 2393 int v3 = (nfsd->nd_flag & ND_NFSV3); 2394 char *cp2; 2395 struct mbuf *mb, *mreq; 2396 struct vnode *vp, *dirp = (struct vnode *)0; 2397 struct vattr dirfor, diraft; 2398 nfsfh_t nfh; 2399 fhandle_t *fhp; 2400 struct nameidata nd; 2401 u_quad_t frev; 2402 2403 fhp = &nfh.fh_generic; 2404 nfsm_srvmtofh(fhp); 2405 nfsm_srvnamesiz(len); 2406 nd.ni_cnd.cn_cred = cred; 2407 nd.ni_cnd.cn_nameiop = DELETE; 2408 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 2409 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2410 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2411 if (dirp) { 2412 if (v3) 2413 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2414 else { 2415 vrele(dirp); 2416 dirp = (struct vnode *)0; 2417 } 2418 } 2419 if (error) { 2420 nfsm_reply(NFSX_WCCDATA(v3)); 2421 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2422 if (dirp) 2423 vrele(dirp); 2424 return (0); 2425 } 2426 vp = nd.ni_vp; 2427 if (vp->v_type != VDIR) { 2428 error = ENOTDIR; 2429 goto out; 2430 } 2431 /* 2432 * No rmdir "." please. 2433 */ 2434 if (nd.ni_dvp == vp) { 2435 error = EINVAL; 2436 goto out; 2437 } 2438 /* 2439 * The root of a mounted filesystem cannot be deleted. 2440 */ 2441 if (vp->v_flag & VROOT) 2442 error = EBUSY; 2443 out: 2444 if (!error) { 2445 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2446 nqsrv_getl(vp, ND_WRITE); 2447 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2448 } else { 2449 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2450 if (nd.ni_dvp == nd.ni_vp) 2451 vrele(nd.ni_dvp); 2452 else 2453 vput(nd.ni_dvp); 2454 vput(vp); 2455 } 2456 if (dirp) { 2457 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2458 vrele(dirp); 2459 } 2460 nfsm_reply(NFSX_WCCDATA(v3)); 2461 if (v3) { 2462 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2463 return (0); 2464 } 2465 nfsm_srvdone; 2466 } 2467 2468 /* 2469 * nfs readdir service 2470 * - mallocs what it thinks is enough to read 2471 * count rounded up to a multiple of NFS_DIRBLKSIZ <= NFS_MAXREADDIR 2472 * - calls VOP_READDIR() 2473 * - loops around building the reply 2474 * if the output generated exceeds count break out of loop 2475 * The nfsm_clget macro is used here so that the reply will be packed 2476 * tightly in mbuf clusters. 2477 * - it only knows that it has encountered eof when the VOP_READDIR() 2478 * reads nothing 2479 * - as such one readdir rpc will return eof false although you are there 2480 * and then the next will return eof 2481 * - it trims out records with d_fileno == 0 2482 * this doesn't matter for Unix clients, but they might confuse clients 2483 * for other os'. 2484 * - it trims out records with d_type == DT_WHT 2485 * these cannot be seen through NFS (unless we extend the protocol) 2486 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less 2487 * than requested, but this may not apply to all filesystems. For 2488 * example, client NFS does not { although it is never remote mounted 2489 * anyhow } 2490 * The alternate call nfsrv_readdirplus() does lookups as well. 2491 * PS: The NFS protocol spec. does not clarify what the "count" byte 2492 * argument is a count of.. just name strings and file id's or the 2493 * entire reply rpc or ... 2494 * I tried just file name and id sizes and it confused the Sun client, 2495 * so I am using the full rpc size now. The "paranoia.." comment refers 2496 * to including the status longwords that are not a part of the dir. 2497 * "entry" structures, but are in the rpc. 2498 */ 2499 struct flrep { 2500 nfsuint64 fl_off; 2501 u_int32_t fl_postopok; 2502 u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)]; 2503 u_int32_t fl_fhok; 2504 u_int32_t fl_fhsize; 2505 u_int32_t fl_nfh[NFSX_V3FH / sizeof (u_int32_t)]; 2506 }; 2507 2508 int 2509 nfsrv_readdir(nfsd, slp, lwp, mrq) 2510 struct nfsrv_descript *nfsd; 2511 struct nfssvc_sock *slp; 2512 struct lwp *lwp; 2513 struct mbuf **mrq; 2514 { 2515 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2516 struct mbuf *nam = nfsd->nd_nam; 2517 caddr_t dpos = nfsd->nd_dpos; 2518 struct ucred *cred = &nfsd->nd_cr; 2519 char *bp, *be; 2520 struct mbuf *mp; 2521 struct dirent *dp; 2522 caddr_t cp; 2523 u_int32_t *tl; 2524 int32_t t1; 2525 caddr_t bpos; 2526 struct mbuf *mb, *mreq, *mp2; 2527 char *cpos, *cend, *cp2, *rbuf; 2528 struct vnode *vp; 2529 struct vattr at; 2530 nfsfh_t nfh; 2531 fhandle_t *fhp; 2532 struct uio io; 2533 struct iovec iv; 2534 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2535 int siz, cnt, fullsiz, eofflag, rdonly, cache, ncookies; 2536 int v3 = (nfsd->nd_flag & ND_NFSV3); 2537 u_quad_t frev, off, toff, verf; 2538 off_t *cookies = NULL, *cookiep; 2539 nfsuint64 jar; 2540 2541 fhp = &nfh.fh_generic; 2542 nfsm_srvmtofh(fhp); 2543 if (v3) { 2544 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 2545 toff = fxdr_hyper(tl); 2546 tl += 2; 2547 verf = fxdr_hyper(tl); 2548 tl += 2; 2549 } else { 2550 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2551 toff = fxdr_unsigned(u_quad_t, *tl++); 2552 } 2553 off = toff; 2554 cnt = fxdr_unsigned(int, *tl); 2555 siz = ((cnt + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1)); 2556 xfer = NFS_SRVMAXDATA(nfsd); 2557 if (siz > xfer) 2558 siz = xfer; 2559 fullsiz = siz; 2560 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 2561 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2562 if (!error && vp->v_type != VDIR) { 2563 error = ENOTDIR; 2564 vput(vp); 2565 } 2566 if (error) { 2567 nfsm_reply(NFSX_UNSIGNED); 2568 nfsm_srvpostop_attr(getret, &at); 2569 return (0); 2570 } 2571 nqsrv_getl(vp, ND_READ); 2572 if (v3) { 2573 error = getret = VOP_GETATTR(vp, &at, cred, lwp); 2574 #ifdef NFS3_STRICTVERF 2575 /* 2576 * XXX This check is too strict for Solaris 2.5 clients. 2577 */ 2578 if (!error && toff && verf != at.va_filerev) 2579 error = NFSERR_BAD_COOKIE; 2580 #endif 2581 } 2582 if (!error) 2583 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0); 2584 if (error) { 2585 vput(vp); 2586 nfsm_reply(NFSX_POSTOPATTR(v3)); 2587 nfsm_srvpostop_attr(getret, &at); 2588 return (0); 2589 } 2590 VOP_UNLOCK(vp, 0); 2591 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2592 again: 2593 iv.iov_base = rbuf; 2594 iv.iov_len = fullsiz; 2595 io.uio_iov = &iv; 2596 io.uio_iovcnt = 1; 2597 io.uio_offset = (off_t)off; 2598 io.uio_resid = fullsiz; 2599 io.uio_segflg = UIO_SYSSPACE; 2600 io.uio_rw = UIO_READ; 2601 io.uio_lwp = (struct lwp *)0; 2602 eofflag = 0; 2603 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2604 2605 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2606 2607 off = (off_t)io.uio_offset; 2608 if (!cookies && !error) 2609 error = NFSERR_PERM; 2610 if (v3) { 2611 getret = VOP_GETATTR(vp, &at, cred, lwp); 2612 if (!error) 2613 error = getret; 2614 } 2615 2616 VOP_UNLOCK(vp, 0); 2617 if (error) { 2618 vrele(vp); 2619 free((caddr_t)rbuf, M_TEMP); 2620 if (cookies) 2621 free((caddr_t)cookies, M_TEMP); 2622 nfsm_reply(NFSX_POSTOPATTR(v3)); 2623 nfsm_srvpostop_attr(getret, &at); 2624 return (0); 2625 } 2626 if (io.uio_resid) { 2627 siz -= io.uio_resid; 2628 2629 /* 2630 * If nothing read, return eof 2631 * rpc reply 2632 */ 2633 if (siz == 0) { 2634 vrele(vp); 2635 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + 2636 2 * NFSX_UNSIGNED); 2637 if (v3) { 2638 nfsm_srvpostop_attr(getret, &at); 2639 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2640 txdr_hyper(at.va_filerev, tl); 2641 tl += 2; 2642 } else 2643 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2644 *tl++ = nfs_false; 2645 *tl = nfs_true; 2646 free((caddr_t)rbuf, M_TEMP); 2647 free((caddr_t)cookies, M_TEMP); 2648 return (0); 2649 } 2650 } 2651 2652 /* 2653 * Check for degenerate cases of nothing useful read. 2654 * If so go try again 2655 */ 2656 cpos = rbuf; 2657 cend = rbuf + siz; 2658 dp = (struct dirent *)cpos; 2659 cookiep = cookies; 2660 2661 while (cpos < cend && ncookies > 0 && 2662 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2663 cpos += dp->d_reclen; 2664 dp = (struct dirent *)cpos; 2665 cookiep++; 2666 ncookies--; 2667 } 2668 if (cpos >= cend || ncookies == 0) { 2669 toff = off; 2670 siz = fullsiz; 2671 goto again; 2672 } 2673 2674 len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */ 2675 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz); 2676 if (v3) { 2677 nfsm_srvpostop_attr(getret, &at); 2678 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2679 txdr_hyper(at.va_filerev, tl); 2680 } 2681 mp = mp2 = mb; 2682 bp = bpos; 2683 be = bp + M_TRAILINGSPACE(mp); 2684 2685 /* Loop through the records and build reply */ 2686 while (cpos < cend && ncookies > 0) { 2687 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 2688 nlen = dp->d_namlen; 2689 rem = nfsm_rndup(nlen)-nlen; 2690 len += (4 * NFSX_UNSIGNED + nlen + rem); 2691 if (v3) 2692 len += 2 * NFSX_UNSIGNED; 2693 if (len > cnt) { 2694 eofflag = 0; 2695 break; 2696 } 2697 /* 2698 * Build the directory record xdr from 2699 * the dirent entry. 2700 */ 2701 nfsm_clget; 2702 *tl = nfs_true; 2703 bp += NFSX_UNSIGNED; 2704 if (v3) { 2705 nfsm_clget; 2706 *tl = 0; 2707 bp += NFSX_UNSIGNED; 2708 } 2709 nfsm_clget; 2710 *tl = txdr_unsigned(dp->d_fileno); 2711 bp += NFSX_UNSIGNED; 2712 nfsm_clget; 2713 *tl = txdr_unsigned(nlen); 2714 bp += NFSX_UNSIGNED; 2715 2716 /* And loop around copying the name */ 2717 xfer = nlen; 2718 cp = dp->d_name; 2719 while (xfer > 0) { 2720 nfsm_clget; 2721 if ((bp+xfer) > be) 2722 tsiz = be-bp; 2723 else 2724 tsiz = xfer; 2725 memcpy(bp, cp, tsiz); 2726 bp += tsiz; 2727 xfer -= tsiz; 2728 if (xfer > 0) 2729 cp += tsiz; 2730 } 2731 /* And null pad to an int32_t boundary */ 2732 for (i = 0; i < rem; i++) 2733 *bp++ = '\0'; 2734 nfsm_clget; 2735 2736 /* Finish off the record */ 2737 txdr_hyper(*cookiep, &jar); 2738 if (v3) { 2739 *tl = jar.nfsuquad[0]; 2740 bp += NFSX_UNSIGNED; 2741 nfsm_clget; 2742 } 2743 *tl = jar.nfsuquad[1]; 2744 bp += NFSX_UNSIGNED; 2745 } 2746 cpos += dp->d_reclen; 2747 dp = (struct dirent *)cpos; 2748 cookiep++; 2749 ncookies--; 2750 } 2751 vrele(vp); 2752 nfsm_clget; 2753 *tl = nfs_false; 2754 bp += NFSX_UNSIGNED; 2755 nfsm_clget; 2756 if (eofflag) 2757 *tl = nfs_true; 2758 else 2759 *tl = nfs_false; 2760 bp += NFSX_UNSIGNED; 2761 if (mp != mb) { 2762 if (bp < be) 2763 mp->m_len = bp - mtod(mp, caddr_t); 2764 } else 2765 mp->m_len += bp - bpos; 2766 free((caddr_t)rbuf, M_TEMP); 2767 free((caddr_t)cookies, M_TEMP); 2768 nfsm_srvdone; 2769 } 2770 2771 int 2772 nfsrv_readdirplus(nfsd, slp, lwp, mrq) 2773 struct nfsrv_descript *nfsd; 2774 struct nfssvc_sock *slp; 2775 struct lwp *lwp; 2776 struct mbuf **mrq; 2777 { 2778 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2779 struct mbuf *nam = nfsd->nd_nam; 2780 caddr_t dpos = nfsd->nd_dpos; 2781 struct ucred *cred = &nfsd->nd_cr; 2782 char *bp, *be; 2783 struct mbuf *mp; 2784 struct dirent *dp; 2785 caddr_t cp; 2786 u_int32_t *tl; 2787 int32_t t1; 2788 caddr_t bpos; 2789 struct mbuf *mb, *mreq, *mp2; 2790 char *cpos, *cend, *cp2, *rbuf; 2791 struct vnode *vp, *nvp; 2792 struct flrep fl; 2793 nfsfh_t nfh; 2794 fhandle_t *fhp, *nfhp = (fhandle_t *)fl.fl_nfh; 2795 struct uio io; 2796 struct iovec iv; 2797 struct vattr va, at, *vap = &va; 2798 struct nfs_fattr *fp; 2799 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2800 int siz, cnt, fullsiz, eofflag, rdonly, cache, dirlen, ncookies; 2801 u_quad_t frev, off, toff, verf; 2802 off_t *cookies = NULL, *cookiep; 2803 2804 fhp = &nfh.fh_generic; 2805 nfsm_srvmtofh(fhp); 2806 nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED); 2807 toff = fxdr_hyper(tl); 2808 tl += 2; 2809 verf = fxdr_hyper(tl); 2810 tl += 2; 2811 siz = fxdr_unsigned(int, *tl++); 2812 cnt = fxdr_unsigned(int, *tl); 2813 off = toff; 2814 siz = ((siz + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1)); 2815 xfer = NFS_SRVMAXDATA(nfsd); 2816 if (siz > xfer) 2817 siz = xfer; 2818 fullsiz = siz; 2819 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 2820 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2821 if (!error && vp->v_type != VDIR) { 2822 error = ENOTDIR; 2823 vput(vp); 2824 } 2825 if (error) { 2826 nfsm_reply(NFSX_UNSIGNED); 2827 nfsm_srvpostop_attr(getret, &at); 2828 return (0); 2829 } 2830 error = getret = VOP_GETATTR(vp, &at, cred, lwp); 2831 #ifdef NFS3_STRICTVERF 2832 /* 2833 * XXX This check is too strict for Solaris 2.5 clients. 2834 */ 2835 if (!error && toff && verf != at.va_filerev) 2836 error = NFSERR_BAD_COOKIE; 2837 #endif 2838 if (!error) { 2839 nqsrv_getl(vp, ND_READ); 2840 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0); 2841 } 2842 if (error) { 2843 vput(vp); 2844 nfsm_reply(NFSX_V3POSTOPATTR); 2845 nfsm_srvpostop_attr(getret, &at); 2846 return (0); 2847 } 2848 VOP_UNLOCK(vp, 0); 2849 2850 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2851 again: 2852 iv.iov_base = rbuf; 2853 iv.iov_len = fullsiz; 2854 io.uio_iov = &iv; 2855 io.uio_iovcnt = 1; 2856 io.uio_offset = (off_t)off; 2857 io.uio_resid = fullsiz; 2858 io.uio_segflg = UIO_SYSSPACE; 2859 io.uio_rw = UIO_READ; 2860 io.uio_lwp = (struct lwp *)0; 2861 eofflag = 0; 2862 2863 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2864 2865 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2866 2867 off = (u_quad_t)io.uio_offset; 2868 getret = VOP_GETATTR(vp, &at, cred, lwp); 2869 2870 VOP_UNLOCK(vp, 0); 2871 2872 /* 2873 * If the VGET operation doesn't work for this filesystem, 2874 * we can't support readdirplus. Returning NOTSUPP should 2875 * make clients fall back to plain readdir. 2876 * There's no need to check for VPTOFH as well, we wouldn't 2877 * even be here otherwise. 2878 */ 2879 if (!getret) { 2880 if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp))) 2881 getret = (getret == EOPNOTSUPP) ? 2882 NFSERR_NOTSUPP : NFSERR_IO; 2883 else 2884 vput(nvp); 2885 } 2886 2887 if (!cookies && !error) 2888 error = NFSERR_PERM; 2889 if (!error) 2890 error = getret; 2891 if (error) { 2892 vrele(vp); 2893 if (cookies) 2894 free((caddr_t)cookies, M_TEMP); 2895 free((caddr_t)rbuf, M_TEMP); 2896 nfsm_reply(NFSX_V3POSTOPATTR); 2897 nfsm_srvpostop_attr(getret, &at); 2898 return (0); 2899 } 2900 if (io.uio_resid) { 2901 siz -= io.uio_resid; 2902 2903 /* 2904 * If nothing read, return eof 2905 * rpc reply 2906 */ 2907 if (siz == 0) { 2908 vrele(vp); 2909 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2910 2 * NFSX_UNSIGNED); 2911 nfsm_srvpostop_attr(getret, &at); 2912 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2913 txdr_hyper(at.va_filerev, tl); 2914 tl += 2; 2915 *tl++ = nfs_false; 2916 *tl = nfs_true; 2917 free((caddr_t)cookies, M_TEMP); 2918 free((caddr_t)rbuf, M_TEMP); 2919 return (0); 2920 } 2921 } 2922 2923 /* 2924 * Check for degenerate cases of nothing useful read. 2925 * If so go try again 2926 */ 2927 cpos = rbuf; 2928 cend = rbuf + siz; 2929 dp = (struct dirent *)cpos; 2930 cookiep = cookies; 2931 2932 while (cpos < cend && ncookies > 0 && 2933 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2934 cpos += dp->d_reclen; 2935 dp = (struct dirent *)cpos; 2936 cookiep++; 2937 ncookies--; 2938 } 2939 if (cpos >= cend || ncookies == 0) { 2940 toff = off; 2941 siz = fullsiz; 2942 goto again; 2943 } 2944 2945 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED; 2946 nfsm_reply(cnt); 2947 nfsm_srvpostop_attr(getret, &at); 2948 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2949 txdr_hyper(at.va_filerev, tl); 2950 mp = mp2 = mb; 2951 bp = bpos; 2952 be = bp + M_TRAILINGSPACE(mp); 2953 2954 /* Loop through the records and build reply */ 2955 while (cpos < cend && ncookies > 0) { 2956 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 2957 nlen = dp->d_namlen; 2958 rem = nfsm_rndup(nlen)-nlen; 2959 2960 /* 2961 * For readdir_and_lookup get the vnode using 2962 * the file number. 2963 */ 2964 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp)) 2965 goto invalid; 2966 memset((caddr_t)nfhp, 0, NFSX_V3FH); 2967 nfhp->fh_fsid = 2968 nvp->v_mount->mnt_stat.f_fsid; 2969 if (VFS_VPTOFH(nvp, &nfhp->fh_fid)) { 2970 vput(nvp); 2971 goto invalid; 2972 } 2973 if (VOP_GETATTR(nvp, vap, cred, lwp)) { 2974 vput(nvp); 2975 goto invalid; 2976 } 2977 vput(nvp); 2978 2979 /* 2980 * If either the dircount or maxcount will be 2981 * exceeded, get out now. Both of these lengths 2982 * are calculated conservatively, including all 2983 * XDR overheads. 2984 */ 2985 len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH + 2986 NFSX_V3POSTOPATTR); 2987 dirlen += (6 * NFSX_UNSIGNED + nlen + rem); 2988 if (len > cnt || dirlen > fullsiz) { 2989 eofflag = 0; 2990 break; 2991 } 2992 2993 /* 2994 * Build the directory record xdr from 2995 * the dirent entry. 2996 */ 2997 fp = (struct nfs_fattr *)&fl.fl_fattr; 2998 nfsm_srvfillattr(vap, fp); 2999 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH); 3000 fl.fl_fhok = nfs_true; 3001 fl.fl_postopok = nfs_true; 3002 txdr_hyper(*cookiep, fl.fl_off.nfsuquad); 3003 3004 nfsm_clget; 3005 *tl = nfs_true; 3006 bp += NFSX_UNSIGNED; 3007 nfsm_clget; 3008 *tl = 0; 3009 bp += NFSX_UNSIGNED; 3010 nfsm_clget; 3011 *tl = txdr_unsigned(dp->d_fileno); 3012 bp += NFSX_UNSIGNED; 3013 nfsm_clget; 3014 *tl = txdr_unsigned(nlen); 3015 bp += NFSX_UNSIGNED; 3016 3017 /* And loop around copying the name */ 3018 xfer = nlen; 3019 cp = dp->d_name; 3020 while (xfer > 0) { 3021 nfsm_clget; 3022 if ((bp + xfer) > be) 3023 tsiz = be - bp; 3024 else 3025 tsiz = xfer; 3026 memcpy(bp, cp, tsiz); 3027 bp += tsiz; 3028 xfer -= tsiz; 3029 if (xfer > 0) 3030 cp += tsiz; 3031 } 3032 /* And null pad to an int32_t boundary */ 3033 for (i = 0; i < rem; i++) 3034 *bp++ = '\0'; 3035 3036 /* 3037 * Now copy the flrep structure out. 3038 */ 3039 xfer = sizeof (struct flrep); 3040 cp = (caddr_t)&fl; 3041 while (xfer > 0) { 3042 nfsm_clget; 3043 if ((bp + xfer) > be) 3044 tsiz = be - bp; 3045 else 3046 tsiz = xfer; 3047 memcpy(bp, cp, tsiz); 3048 bp += tsiz; 3049 xfer -= tsiz; 3050 if (xfer > 0) 3051 cp += tsiz; 3052 } 3053 } 3054 invalid: 3055 cpos += dp->d_reclen; 3056 dp = (struct dirent *)cpos; 3057 cookiep++; 3058 ncookies--; 3059 } 3060 vrele(vp); 3061 nfsm_clget; 3062 *tl = nfs_false; 3063 bp += NFSX_UNSIGNED; 3064 nfsm_clget; 3065 if (eofflag) 3066 *tl = nfs_true; 3067 else 3068 *tl = nfs_false; 3069 bp += NFSX_UNSIGNED; 3070 if (mp != mb) { 3071 if (bp < be) 3072 mp->m_len = bp - mtod(mp, caddr_t); 3073 } else 3074 mp->m_len += bp - bpos; 3075 free((caddr_t)cookies, M_TEMP); 3076 free((caddr_t)rbuf, M_TEMP); 3077 nfsm_srvdone; 3078 } 3079 3080 /* 3081 * nfs commit service 3082 */ 3083 int 3084 nfsrv_commit(nfsd, slp, lwp, mrq) 3085 struct nfsrv_descript *nfsd; 3086 struct nfssvc_sock *slp; 3087 struct lwp *lwp; 3088 struct mbuf **mrq; 3089 { 3090 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3091 struct mbuf *nam = nfsd->nd_nam; 3092 caddr_t dpos = nfsd->nd_dpos; 3093 struct ucred *cred = &nfsd->nd_cr; 3094 struct vattr bfor, aft; 3095 struct vnode *vp; 3096 nfsfh_t nfh; 3097 fhandle_t *fhp; 3098 u_int32_t *tl; 3099 int32_t t1; 3100 caddr_t bpos; 3101 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cnt, cache; 3102 char *cp2; 3103 struct mbuf *mb, *mreq; 3104 u_quad_t frev, off, end; 3105 3106 #ifndef nolint 3107 cache = 0; 3108 #endif 3109 fhp = &nfh.fh_generic; 3110 nfsm_srvmtofh(fhp); 3111 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED); 3112 3113 off = fxdr_hyper(tl); 3114 tl += 2; 3115 cnt = fxdr_unsigned(int, *tl); 3116 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3117 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3118 if (error) { 3119 nfsm_reply(2 * NFSX_UNSIGNED); 3120 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3121 return (0); 3122 } 3123 for_ret = VOP_GETATTR(vp, &bfor, cred, lwp); 3124 end = (cnt > 0) ? off + cnt : vp->v_size; 3125 if (end < off || end > vp->v_size) 3126 end = vp->v_size; 3127 error = VOP_FSYNC(vp, cred, FSYNC_WAIT, off, end, lwp); 3128 aft_ret = VOP_GETATTR(vp, &aft, cred, lwp); 3129 vput(vp); 3130 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF); 3131 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3132 if (!error) { 3133 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF); 3134 *tl++ = txdr_unsigned(boottime.tv_sec); 3135 *tl = txdr_unsigned(boottime.tv_usec); 3136 } else 3137 return (0); 3138 nfsm_srvdone; 3139 } 3140 3141 /* 3142 * nfs statfs service 3143 */ 3144 int 3145 nfsrv_statfs(nfsd, slp, lwp, mrq) 3146 struct nfsrv_descript *nfsd; 3147 struct nfssvc_sock *slp; 3148 struct lwp *lwp; 3149 struct mbuf **mrq; 3150 { 3151 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3152 struct mbuf *nam = nfsd->nd_nam; 3153 caddr_t dpos = nfsd->nd_dpos; 3154 struct ucred *cred = &nfsd->nd_cr; 3155 struct statfs *sf; 3156 struct nfs_statfs *sfp; 3157 u_int32_t *tl; 3158 int32_t t1; 3159 caddr_t bpos; 3160 int error = 0, rdonly, cache, getret = 1; 3161 int v3 = (nfsd->nd_flag & ND_NFSV3); 3162 char *cp2; 3163 struct mbuf *mb, *mreq; 3164 struct vnode *vp; 3165 struct vattr at; 3166 nfsfh_t nfh; 3167 fhandle_t *fhp; 3168 struct statfs statfs; 3169 u_quad_t frev, tval; 3170 3171 #ifndef nolint 3172 cache = 0; 3173 #endif 3174 fhp = &nfh.fh_generic; 3175 nfsm_srvmtofh(fhp); 3176 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3177 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3178 if (error) { 3179 nfsm_reply(NFSX_UNSIGNED); 3180 nfsm_srvpostop_attr(getret, &at); 3181 return (0); 3182 } 3183 sf = &statfs; 3184 error = VFS_STATFS(vp->v_mount, sf, lwp); 3185 getret = VOP_GETATTR(vp, &at, cred, lwp); 3186 vput(vp); 3187 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3)); 3188 if (v3) 3189 nfsm_srvpostop_attr(getret, &at); 3190 if (error) 3191 return (0); 3192 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3)); 3193 if (v3) { 3194 tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_bsize); 3195 txdr_hyper(tval, &sfp->sf_tbytes); 3196 tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_bsize); 3197 txdr_hyper(tval, &sfp->sf_fbytes); 3198 tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_bsize); 3199 txdr_hyper(tval, &sfp->sf_abytes); 3200 tval = (u_quad_t)sf->f_files; 3201 txdr_hyper(tval, &sfp->sf_tfiles); 3202 tval = (u_quad_t)sf->f_ffree; 3203 txdr_hyper(tval, &sfp->sf_ffiles); 3204 txdr_hyper(tval, &sfp->sf_afiles); 3205 sfp->sf_invarsec = 0; 3206 } else { 3207 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA); 3208 sfp->sf_bsize = txdr_unsigned(sf->f_bsize); 3209 sfp->sf_blocks = txdr_unsigned(sf->f_blocks); 3210 sfp->sf_bfree = txdr_unsigned(sf->f_bfree); 3211 sfp->sf_bavail = txdr_unsigned(sf->f_bavail); 3212 } 3213 nfsm_srvdone; 3214 } 3215 3216 /* 3217 * nfs fsinfo service 3218 */ 3219 int 3220 nfsrv_fsinfo(nfsd, slp, lwp, mrq) 3221 struct nfsrv_descript *nfsd; 3222 struct nfssvc_sock *slp; 3223 struct lwp *lwp; 3224 struct mbuf **mrq; 3225 { 3226 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3227 struct mbuf *nam = nfsd->nd_nam; 3228 caddr_t dpos = nfsd->nd_dpos; 3229 struct ucred *cred = &nfsd->nd_cr; 3230 u_int32_t *tl; 3231 struct nfsv3_fsinfo *sip; 3232 int32_t t1; 3233 caddr_t bpos; 3234 int error = 0, rdonly, cache, getret = 1; 3235 uint32_t maxdata; 3236 char *cp2; 3237 struct mbuf *mb, *mreq; 3238 struct vnode *vp; 3239 struct vattr at; 3240 nfsfh_t nfh; 3241 fhandle_t *fhp; 3242 u_quad_t frev, maxfsize; 3243 struct statfs sb; 3244 3245 #ifndef nolint 3246 cache = 0; 3247 #endif 3248 fhp = &nfh.fh_generic; 3249 nfsm_srvmtofh(fhp); 3250 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3251 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3252 if (error) { 3253 nfsm_reply(NFSX_UNSIGNED); 3254 nfsm_srvpostop_attr(getret, &at); 3255 return (0); 3256 } 3257 3258 /* XXX Try to make a guess on the max file size. */ 3259 VFS_STATFS(vp->v_mount, &sb, (struct lwp *)0); 3260 maxfsize = (u_quad_t)0x80000000 * sb.f_bsize - 1; 3261 3262 getret = VOP_GETATTR(vp, &at, cred, lwp); 3263 vput(vp); 3264 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO); 3265 nfsm_srvpostop_attr(getret, &at); 3266 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO); 3267 3268 /* 3269 * XXX 3270 * There should be file system VFS OP(s) to get this information. 3271 * For now, assume ufs. 3272 */ 3273 if (slp->ns_so->so_type == SOCK_DGRAM) 3274 maxdata = NFS_MAXDGRAMDATA; 3275 else 3276 maxdata = NFS_MAXDATA; 3277 sip->fs_rtmax = txdr_unsigned(maxdata); 3278 sip->fs_rtpref = txdr_unsigned(maxdata); 3279 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE); 3280 sip->fs_wtmax = txdr_unsigned(maxdata); 3281 sip->fs_wtpref = txdr_unsigned(maxdata); 3282 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE); 3283 sip->fs_dtpref = txdr_unsigned(maxdata); 3284 txdr_hyper(maxfsize, &sip->fs_maxfilesize); 3285 sip->fs_timedelta.nfsv3_sec = 0; 3286 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1); 3287 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK | 3288 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS | 3289 NFSV3FSINFO_CANSETTIME); 3290 nfsm_srvdone; 3291 } 3292 3293 /* 3294 * nfs pathconf service 3295 */ 3296 int 3297 nfsrv_pathconf(nfsd, slp, lwp, mrq) 3298 struct nfsrv_descript *nfsd; 3299 struct nfssvc_sock *slp; 3300 struct lwp *lwp; 3301 struct mbuf **mrq; 3302 { 3303 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3304 struct mbuf *nam = nfsd->nd_nam; 3305 caddr_t dpos = nfsd->nd_dpos; 3306 struct ucred *cred = &nfsd->nd_cr; 3307 u_int32_t *tl; 3308 struct nfsv3_pathconf *pc; 3309 int32_t t1; 3310 caddr_t bpos; 3311 int error = 0, rdonly, cache, getret = 1; 3312 register_t linkmax, namemax, chownres, notrunc; 3313 char *cp2; 3314 struct mbuf *mb, *mreq; 3315 struct vnode *vp; 3316 struct vattr at; 3317 nfsfh_t nfh; 3318 fhandle_t *fhp; 3319 u_quad_t frev; 3320 3321 #ifndef nolint 3322 cache = 0; 3323 #endif 3324 fhp = &nfh.fh_generic; 3325 nfsm_srvmtofh(fhp); 3326 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3327 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3328 if (error) { 3329 nfsm_reply(NFSX_UNSIGNED); 3330 nfsm_srvpostop_attr(getret, &at); 3331 return (0); 3332 } 3333 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax); 3334 if (!error) 3335 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax); 3336 if (!error) 3337 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres); 3338 if (!error) 3339 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc); 3340 getret = VOP_GETATTR(vp, &at, cred, lwp); 3341 vput(vp); 3342 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF); 3343 nfsm_srvpostop_attr(getret, &at); 3344 if (error) 3345 return (0); 3346 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF); 3347 3348 pc->pc_linkmax = txdr_unsigned(linkmax); 3349 pc->pc_namemax = txdr_unsigned(namemax); 3350 pc->pc_notrunc = txdr_unsigned(notrunc); 3351 pc->pc_chownrestricted = txdr_unsigned(chownres); 3352 3353 /* 3354 * These should probably be supported by VOP_PATHCONF(), but 3355 * until msdosfs is exportable (why would you want to?), the 3356 * Unix defaults should be ok. 3357 */ 3358 pc->pc_caseinsensitive = nfs_false; 3359 pc->pc_casepreserving = nfs_true; 3360 nfsm_srvdone; 3361 } 3362 3363 /* 3364 * Null operation, used by clients to ping server 3365 */ 3366 /* ARGSUSED */ 3367 int 3368 nfsrv_null(nfsd, slp, lwp, mrq) 3369 struct nfsrv_descript *nfsd; 3370 struct nfssvc_sock *slp; 3371 struct lwp *lwp; 3372 struct mbuf **mrq; 3373 { 3374 struct mbuf *mrep = nfsd->nd_mrep; 3375 caddr_t bpos; 3376 int error = NFSERR_RETVOID, cache = 0; 3377 struct mbuf *mb, *mreq; 3378 u_quad_t frev; 3379 3380 nfsm_reply(0); 3381 return (0); 3382 } 3383 3384 /* 3385 * No operation, used for obsolete procedures 3386 */ 3387 /* ARGSUSED */ 3388 int 3389 nfsrv_noop(nfsd, slp, lwp, mrq) 3390 struct nfsrv_descript *nfsd; 3391 struct nfssvc_sock *slp; 3392 struct lwp *lwp; 3393 struct mbuf **mrq; 3394 { 3395 struct mbuf *mrep = nfsd->nd_mrep; 3396 caddr_t bpos; 3397 int error, cache = 0; 3398 struct mbuf *mb, *mreq; 3399 u_quad_t frev; 3400 3401 if (nfsd->nd_repstat) 3402 error = nfsd->nd_repstat; 3403 else 3404 error = EPROCUNAVAIL; 3405 nfsm_reply(0); 3406 return (0); 3407 } 3408 3409 /* 3410 * Perform access checking for vnodes obtained from file handles that would 3411 * refer to files already opened by a Unix client. You cannot just use 3412 * vn_writechk() and VOP_ACCESS() for two reasons. 3413 * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case 3414 * 2 - The owner is to be given access irrespective of mode bits for some 3415 * operations, so that processes that chmod after opening a file don't 3416 * break. I don't like this because it opens a security hole, but since 3417 * the nfs server opens a security hole the size of a barn door anyhow, 3418 * what the heck. 3419 * 3420 * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS() 3421 * will return EPERM instead of EACCESS. EPERM is always an error. 3422 */ 3423 int 3424 nfsrv_access(vp, flags, cred, rdonly, lwp, override) 3425 struct vnode *vp; 3426 int flags; 3427 struct ucred *cred; 3428 int rdonly; 3429 struct lwp *lwp; 3430 { 3431 struct vattr vattr; 3432 int error; 3433 if (flags & VWRITE) { 3434 /* Just vn_writechk() changed to check rdonly */ 3435 /* 3436 * Disallow write attempts on read-only file systems; 3437 * unless the file is a socket or a block or character 3438 * device resident on the file system. 3439 */ 3440 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 3441 switch (vp->v_type) { 3442 case VREG: 3443 case VDIR: 3444 case VLNK: 3445 return (EROFS); 3446 default: 3447 break; 3448 } 3449 } 3450 3451 /* 3452 * If the vnode is in use as a process's text, 3453 * we can't allow writing. 3454 */ 3455 if (vp->v_flag & VTEXT) 3456 return (ETXTBSY); 3457 } 3458 error = VOP_GETATTR(vp, &vattr, cred, lwp); 3459 if (error) 3460 return (error); 3461 error = VOP_ACCESS(vp, flags, cred, lwp); 3462 /* 3463 * Allow certain operations for the owner (reads and writes 3464 * on files that are already open). 3465 */ 3466 if (override && error == EACCES && cred->cr_uid == vattr.va_uid) 3467 error = 0; 3468 return error; 3469 } 3470
Indexes created Thu Sep 25 16:09:42 GMT 2025