nfs_serv.c revision 1.79.2.3
1 /* $NetBSD: nfs_serv.c,v 1.79.2.3 2004/08/24 17:57:41 skrll Exp $ */ 2 3 /* 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * This code is derived from software contributed to Berkeley by 8 * Rick Macklem at The University of Guelph. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * @(#)nfs_serv.c 8.8 (Berkeley) 7/31/95 35 */ 36 37 /* 38 * nfs version 2 and 3 server calls to vnode ops 39 * - these routines generally have 3 phases 40 * 1 - break down and validate rpc request in mbuf list 41 * 2 - do the vnode ops for the request 42 * (surprisingly ?? many are very similar to syscalls in vfs_syscalls.c) 43 * 3 - build the rpc reply in an mbuf list 44 * nb: 45 * - do not mix the phases, since the nfsm_?? macros can return failures 46 * on a bad rpc or similar and do not do any vrele() or vput()'s 47 * 48 * - the nfsm_reply() macro generates an nfs rpc reply with the nfs 49 * error number iff error != 0 whereas 50 * returning an error from the server function implies a fatal error 51 * such as a badly constructed rpc request that should be dropped without 52 * a reply. 53 * For Version 3, nfsm_reply() does not return for the error case, since 54 * most version 3 rpcs return more than the status for error cases. 55 */ 56 57 #include <sys/cdefs.h> 58 __KERNEL_RCSID(0, "$NetBSD: nfs_serv.c,v 1.79.2.3 2004/08/24 17:57:41 skrll Exp $"); 59 60 #include <sys/param.h> 61 #include <sys/systm.h> 62 #include <sys/proc.h> 63 #include <sys/file.h> 64 #include <sys/namei.h> 65 #include <sys/vnode.h> 66 #include <sys/mount.h> 67 #include <sys/socket.h> 68 #include <sys/socketvar.h> 69 #include <sys/mbuf.h> 70 #include <sys/dirent.h> 71 #include <sys/stat.h> 72 #include <sys/kernel.h> 73 #include <ufs/ufs/dir.h> 74 75 #include <uvm/uvm.h> 76 77 #include <nfs/nfsproto.h> 78 #include <nfs/rpcv2.h> 79 #include <nfs/nfs.h> 80 #include <nfs/xdr_subs.h> 81 #include <nfs/nfsm_subs.h> 82 #include <nfs/nqnfs.h> 83 #include <nfs/nfs_var.h> 84 85 /* Global vars */ 86 extern u_int32_t nfs_xdrneg1; 87 extern u_int32_t nfs_false, nfs_true; 88 extern enum vtype nv3tov_type[8]; 89 extern struct nfsstats nfsstats; 90 extern const nfstype nfsv2_type[9]; 91 extern const nfstype nfsv3_type[9]; 92 int nfsrvw_procrastinate = NFS_GATHERDELAY * 1000; 93 int nfsd_use_loan = 1; /* use page-loan for READ OP */ 94 95 /* 96 * nfs v3 access service 97 */ 98 int 99 nfsrv3_access(nfsd, slp, lwp, mrq) 100 struct nfsrv_descript *nfsd; 101 struct nfssvc_sock *slp; 102 struct lwp *lwp; 103 struct mbuf **mrq; 104 { 105 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 106 struct mbuf *nam = nfsd->nd_nam; 107 caddr_t dpos = nfsd->nd_dpos; 108 struct ucred *cred = &nfsd->nd_cr; 109 struct vnode *vp; 110 nfsfh_t nfh; 111 fhandle_t *fhp; 112 u_int32_t *tl; 113 int32_t t1; 114 caddr_t bpos; 115 int error = 0, rdonly, cache = 0, getret; 116 char *cp2; 117 struct mbuf *mb, *mreq; 118 struct vattr va; 119 u_long inmode, testmode, outmode; 120 u_quad_t frev; 121 122 fhp = &nfh.fh_generic; 123 nfsm_srvmtofh(fhp); 124 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 125 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, 126 (nfsd->nd_flag & ND_KERBAUTH), FALSE); 127 if (error) { 128 nfsm_reply(NFSX_UNSIGNED); 129 nfsm_srvpostop_attr(1, (struct vattr *)0); 130 return (0); 131 } 132 inmode = fxdr_unsigned(u_int32_t, *tl); 133 outmode = 0; 134 if ((inmode & NFSV3ACCESS_READ) && 135 nfsrv_access(vp, VREAD, cred, rdonly, lwp, 0) == 0) 136 outmode |= NFSV3ACCESS_READ; 137 if (vp->v_type != VDIR) { 138 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND); 139 if (testmode && 140 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0) 141 outmode |= testmode; 142 if ((inmode & NFSV3ACCESS_EXECUTE) && 143 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0) 144 outmode |= NFSV3ACCESS_EXECUTE; 145 } else { 146 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND | 147 NFSV3ACCESS_DELETE); 148 if (testmode && 149 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0) 150 outmode |= testmode; 151 if ((inmode & NFSV3ACCESS_LOOKUP) && 152 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0) 153 outmode |= NFSV3ACCESS_LOOKUP; 154 } 155 getret = VOP_GETATTR(vp, &va, cred, lwp); 156 vput(vp); 157 nfsm_reply(NFSX_POSTOPATTR(1) + NFSX_UNSIGNED); 158 nfsm_srvpostop_attr(getret, &va); 159 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 160 *tl = txdr_unsigned(outmode); 161 nfsm_srvdone; 162 } 163 164 /* 165 * nfs getattr service 166 */ 167 int 168 nfsrv_getattr(nfsd, slp, lwp, mrq) 169 struct nfsrv_descript *nfsd; 170 struct nfssvc_sock *slp; 171 struct lwp *lwp; 172 struct mbuf **mrq; 173 { 174 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 175 struct mbuf *nam = nfsd->nd_nam; 176 caddr_t dpos = nfsd->nd_dpos; 177 struct ucred *cred = &nfsd->nd_cr; 178 struct nfs_fattr *fp; 179 struct vattr va; 180 struct vnode *vp; 181 nfsfh_t nfh; 182 fhandle_t *fhp; 183 u_int32_t *tl; 184 int32_t t1; 185 caddr_t bpos; 186 int error = 0, rdonly, cache; 187 char *cp2; 188 struct mbuf *mb, *mreq; 189 u_quad_t frev; 190 191 fhp = &nfh.fh_generic; 192 nfsm_srvmtofh(fhp); 193 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, 194 (nfsd->nd_flag & ND_KERBAUTH), FALSE); 195 if (error) { 196 nfsm_reply(0); 197 return (0); 198 } 199 nqsrv_getl(vp, ND_READ); 200 error = VOP_GETATTR(vp, &va, cred, lwp); 201 vput(vp); 202 nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); 203 if (error) 204 return (0); 205 nfsm_build(fp, struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); 206 nfsm_srvfillattr(&va, fp); 207 nfsm_srvdone; 208 } 209 210 /* 211 * nfs setattr service 212 */ 213 int 214 nfsrv_setattr(nfsd, slp, lwp, mrq) 215 struct nfsrv_descript *nfsd; 216 struct nfssvc_sock *slp; 217 struct lwp *lwp; 218 struct mbuf **mrq; 219 { 220 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 221 struct mbuf *nam = nfsd->nd_nam; 222 caddr_t dpos = nfsd->nd_dpos; 223 struct ucred *cred = &nfsd->nd_cr; 224 struct vattr va, preat; 225 struct nfsv2_sattr *sp; 226 struct nfs_fattr *fp; 227 struct vnode *vp; 228 nfsfh_t nfh; 229 fhandle_t *fhp; 230 u_int32_t *tl; 231 int32_t t1; 232 caddr_t bpos; 233 int error = 0, rdonly, cache, preat_ret = 1, postat_ret = 1; 234 int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0; 235 char *cp2; 236 struct mbuf *mb, *mreq; 237 u_quad_t frev; 238 struct timespec guard; 239 struct mount *mp = NULL; 240 241 fhp = &nfh.fh_generic; 242 nfsm_srvmtofh(fhp); 243 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 244 return (ESTALE); 245 vn_start_write(NULL, &mp, V_WAIT); 246 VATTR_NULL(&va); 247 if (v3) { 248 nfsm_srvsattr(&va); 249 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 250 gcheck = fxdr_unsigned(int, *tl); 251 if (gcheck) { 252 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 253 fxdr_nfsv3time(tl, &guard); 254 } 255 } else { 256 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 257 /* 258 * Nah nah nah nah na nah 259 * There is a bug in the Sun client that puts 0xffff in the mode 260 * field of sattr when it should put in 0xffffffff. The u_short 261 * doesn't sign extend. 262 * --> check the low order 2 bytes for 0xffff 263 */ 264 if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff) 265 va.va_mode = nfstov_mode(sp->sa_mode); 266 if (sp->sa_uid != nfs_xdrneg1) 267 va.va_uid = fxdr_unsigned(uid_t, sp->sa_uid); 268 if (sp->sa_gid != nfs_xdrneg1) 269 va.va_gid = fxdr_unsigned(gid_t, sp->sa_gid); 270 if (sp->sa_size != nfs_xdrneg1) 271 va.va_size = fxdr_unsigned(u_quad_t, sp->sa_size); 272 if (sp->sa_atime.nfsv2_sec != nfs_xdrneg1) { 273 #ifdef notyet 274 fxdr_nfsv2time(&sp->sa_atime, &va.va_atime); 275 #else 276 va.va_atime.tv_sec = 277 fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec); 278 va.va_atime.tv_nsec = 0; 279 #endif 280 } 281 if (sp->sa_mtime.nfsv2_sec != nfs_xdrneg1) 282 fxdr_nfsv2time(&sp->sa_mtime, &va.va_mtime); 283 284 } 285 286 /* 287 * Now that we have all the fields, lets do it. 288 */ 289 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, 290 (nfsd->nd_flag & ND_KERBAUTH), FALSE); 291 if (error) { 292 nfsm_reply(2 * NFSX_UNSIGNED); 293 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 294 vn_finished_write(mp, 0); 295 return (0); 296 } 297 nqsrv_getl(vp, ND_WRITE); 298 if (v3) { 299 error = preat_ret = VOP_GETATTR(vp, &preat, cred, lwp); 300 if (!error && gcheck && 301 (preat.va_ctime.tv_sec != guard.tv_sec || 302 preat.va_ctime.tv_nsec != guard.tv_nsec)) 303 error = NFSERR_NOT_SYNC; 304 if (error) { 305 vput(vp); 306 nfsm_reply(NFSX_WCCDATA(v3)); 307 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 308 vn_finished_write(mp, 0); 309 return (0); 310 } 311 } 312 313 /* 314 * If the size is being changed write acces is required, otherwise 315 * just check for a read only file system. 316 */ 317 if (va.va_size == ((u_quad_t)((quad_t) -1))) { 318 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 319 error = EROFS; 320 goto out; 321 } 322 } else { 323 if (vp->v_type == VDIR) { 324 error = EISDIR; 325 goto out; 326 } else if ((error = nfsrv_access(vp, VWRITE, cred, rdonly, 327 lwp, 0)) != 0) 328 goto out; 329 } 330 error = VOP_SETATTR(vp, &va, cred, lwp); 331 postat_ret = VOP_GETATTR(vp, &va, cred, lwp); 332 if (!error) 333 error = postat_ret; 334 out: 335 vput(vp); 336 vn_finished_write(mp, 0); 337 nfsm_reply(NFSX_WCCORFATTR(v3)); 338 if (v3) { 339 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 340 return (0); 341 } else { 342 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 343 nfsm_srvfillattr(&va, fp); 344 } 345 nfsm_srvdone; 346 } 347 348 /* 349 * nfs lookup rpc 350 */ 351 int 352 nfsrv_lookup(nfsd, slp, lwp, mrq) 353 struct nfsrv_descript *nfsd; 354 struct nfssvc_sock *slp; 355 struct lwp *lwp; 356 struct mbuf **mrq; 357 { 358 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 359 struct mbuf *nam = nfsd->nd_nam; 360 caddr_t dpos = nfsd->nd_dpos; 361 struct ucred *cred = &nfsd->nd_cr; 362 struct nfs_fattr *fp; 363 struct nameidata nd, ind, *ndp = &nd; 364 struct vnode *vp, *dirp; 365 nfsfh_t nfh; 366 fhandle_t *fhp; 367 caddr_t cp; 368 u_int32_t *tl; 369 int32_t t1; 370 caddr_t bpos; 371 int error = 0, cache, dirattr_ret = 1; 372 uint32_t len; 373 int v3 = (nfsd->nd_flag & ND_NFSV3), pubflag; 374 char *cp2; 375 struct mbuf *mb, *mreq; 376 struct vattr va, dirattr; 377 u_quad_t frev; 378 379 fhp = &nfh.fh_generic; 380 nfsm_srvmtofh(fhp); 381 nfsm_srvnamesiz(len); 382 383 pubflag = nfs_ispublicfh(fhp); 384 385 nd.ni_cnd.cn_cred = cred; 386 nd.ni_cnd.cn_nameiop = LOOKUP; 387 nd.ni_cnd.cn_flags = LOCKLEAF | SAVESTART; 388 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 389 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), pubflag); 390 391 if (!error && pubflag) { 392 if (nd.ni_vp->v_type == VDIR && nfs_pub.np_index != NULL) { 393 /* 394 * Setup call to lookup() to see if we can find 395 * the index file. Arguably, this doesn't belong 396 * in a kernel.. Ugh. 397 */ 398 ind = nd; 399 VOP_UNLOCK(nd.ni_vp, 0); 400 ind.ni_pathlen = strlen(nfs_pub.np_index); 401 ind.ni_cnd.cn_nameptr = ind.ni_cnd.cn_pnbuf = 402 nfs_pub.np_index; 403 ind.ni_startdir = nd.ni_vp; 404 VREF(ind.ni_startdir); 405 error = lookup(&ind); 406 if (!error) { 407 /* 408 * Found an index file. Get rid of 409 * the old references. 410 */ 411 if (dirp) 412 vrele(dirp); 413 dirp = nd.ni_vp; 414 vrele(nd.ni_startdir); 415 ndp = &ind; 416 } else 417 error = 0; 418 } 419 /* 420 * If the public filehandle was used, check that this lookup 421 * didn't result in a filehandle outside the publicly exported 422 * filesystem. 423 */ 424 425 if (!error && ndp->ni_vp->v_mount != nfs_pub.np_mount) { 426 vput(nd.ni_vp); 427 error = EPERM; 428 } 429 } 430 431 if (dirp) { 432 if (v3) 433 dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred, lwp); 434 vrele(dirp); 435 } 436 437 if (error) { 438 nfsm_reply(NFSX_POSTOPATTR(v3)); 439 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 440 return (0); 441 } 442 443 nqsrv_getl(ndp->ni_startdir, ND_READ); 444 vrele(ndp->ni_startdir); 445 PNBUF_PUT(nd.ni_cnd.cn_pnbuf); 446 vp = ndp->ni_vp; 447 memset((caddr_t)fhp, 0, sizeof(nfh)); 448 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 449 error = VFS_VPTOFH(vp, &fhp->fh_fid); 450 if (!error) 451 error = VOP_GETATTR(vp, &va, cred, lwp); 452 vput(vp); 453 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPORFATTR(v3) + NFSX_POSTOPATTR(v3)); 454 if (error) { 455 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 456 return (0); 457 } 458 nfsm_srvfhtom(fhp, v3); 459 if (v3) { 460 nfsm_srvpostop_attr(0, &va); 461 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 462 } else { 463 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 464 nfsm_srvfillattr(&va, fp); 465 } 466 nfsm_srvdone; 467 } 468 469 /* 470 * nfs readlink service 471 */ 472 int 473 nfsrv_readlink(nfsd, slp, lwp, mrq) 474 struct nfsrv_descript *nfsd; 475 struct nfssvc_sock *slp; 476 struct lwp *lwp; 477 struct mbuf **mrq; 478 { 479 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 480 struct mbuf *nam = nfsd->nd_nam; 481 caddr_t dpos = nfsd->nd_dpos; 482 struct ucred *cred = &nfsd->nd_cr; 483 struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN]; 484 struct iovec *ivp = iv; 485 struct mbuf *mp; 486 u_int32_t *tl; 487 int32_t t1; 488 caddr_t bpos; 489 int error = 0, rdonly, cache, i, padlen, getret; 490 uint32_t len; 491 int v3 = (nfsd->nd_flag & ND_NFSV3); 492 char *cp2; 493 struct mbuf *mb, *mp2 = NULL, *mp3 = NULL, *mreq; 494 struct vnode *vp; 495 struct vattr attr; 496 nfsfh_t nfh; 497 fhandle_t *fhp; 498 struct uio io, *uiop = &io; 499 u_quad_t frev; 500 501 fhp = &nfh.fh_generic; 502 nfsm_srvmtofh(fhp); 503 len = 0; 504 i = 0; 505 while (len < NFS_MAXPATHLEN) { 506 mp = m_get(M_WAIT, MT_DATA); 507 MCLAIM(mp, &nfs_mowner); 508 m_clget(mp, M_WAIT); 509 mp->m_len = NFSMSIZ(mp); 510 if (len == 0) 511 mp3 = mp2 = mp; 512 else { 513 mp2->m_next = mp; 514 mp2 = mp; 515 } 516 if ((len+mp->m_len) > NFS_MAXPATHLEN) { 517 mp->m_len = NFS_MAXPATHLEN-len; 518 len = NFS_MAXPATHLEN; 519 } else 520 len += mp->m_len; 521 ivp->iov_base = mtod(mp, caddr_t); 522 ivp->iov_len = mp->m_len; 523 i++; 524 ivp++; 525 } 526 uiop->uio_iov = iv; 527 uiop->uio_iovcnt = i; 528 uiop->uio_offset = 0; 529 uiop->uio_resid = len; 530 uiop->uio_rw = UIO_READ; 531 uiop->uio_segflg = UIO_SYSSPACE; 532 uiop->uio_lwp = (struct lwp *)0; 533 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 534 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 535 if (error) { 536 m_freem(mp3); 537 nfsm_reply(2 * NFSX_UNSIGNED); 538 nfsm_srvpostop_attr(1, (struct vattr *)0); 539 return (0); 540 } 541 if (vp->v_type != VLNK) { 542 if (v3) 543 error = EINVAL; 544 else 545 error = ENXIO; 546 goto out; 547 } 548 nqsrv_getl(vp, ND_READ); 549 error = VOP_READLINK(vp, uiop, cred); 550 out: 551 getret = VOP_GETATTR(vp, &attr, cred, lwp); 552 vput(vp); 553 if (error) 554 m_freem(mp3); 555 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED); 556 if (v3) { 557 nfsm_srvpostop_attr(getret, &attr); 558 if (error) 559 return (0); 560 } 561 len -= uiop->uio_resid; 562 padlen = nfsm_padlen(len); 563 if (uiop->uio_resid || padlen) 564 nfs_zeropad(mp3, uiop->uio_resid, padlen); 565 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 566 *tl = txdr_unsigned(len); 567 mb->m_next = mp3; 568 nfsm_srvdone; 569 } 570 571 /* 572 * nfs read service 573 */ 574 int 575 nfsrv_read(nfsd, slp, lwp, mrq) 576 struct nfsrv_descript *nfsd; 577 struct nfssvc_sock *slp; 578 struct lwp *lwp; 579 struct mbuf **mrq; 580 { 581 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 582 struct mbuf *nam = nfsd->nd_nam; 583 caddr_t dpos = nfsd->nd_dpos; 584 struct ucred *cred = &nfsd->nd_cr; 585 struct mbuf *m; 586 struct nfs_fattr *fp; 587 u_int32_t *tl; 588 int32_t t1; 589 int i; 590 caddr_t bpos; 591 int error = 0, rdonly, cache, getret; 592 int v3 = (nfsd->nd_flag & ND_NFSV3); 593 uint32_t reqlen, len, cnt, left; 594 int padlen; 595 char *cp2; 596 struct mbuf *mb, *mreq; 597 struct vnode *vp; 598 nfsfh_t nfh; 599 fhandle_t *fhp; 600 struct uio io, *uiop = &io; 601 struct vattr va; 602 off_t off; 603 u_quad_t frev; 604 605 fhp = &nfh.fh_generic; 606 nfsm_srvmtofh(fhp); 607 if (v3) { 608 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 609 off = fxdr_hyper(tl); 610 } else { 611 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 612 off = (off_t)fxdr_unsigned(u_int32_t, *tl); 613 } 614 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 615 reqlen = fxdr_unsigned(uint32_t, *tl); 616 reqlen = MIN(reqlen, NFS_SRVMAXDATA(nfsd)); 617 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 618 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 619 if (error) { 620 nfsm_reply(2 * NFSX_UNSIGNED); 621 nfsm_srvpostop_attr(1, (struct vattr *)0); 622 return (0); 623 } 624 if (vp->v_type != VREG) { 625 if (v3) 626 error = EINVAL; 627 else 628 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 629 } 630 if (!error) { 631 nqsrv_getl(vp, ND_READ); 632 if ((error = nfsrv_access(vp, VREAD, cred, rdonly, lwp, 1)) != 0) 633 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 1); 634 } 635 getret = VOP_GETATTR(vp, &va, cred, lwp); 636 if (!error) 637 error = getret; 638 if (error) { 639 vput(vp); 640 nfsm_reply(NFSX_POSTOPATTR(v3)); 641 nfsm_srvpostop_attr(getret, &va); 642 return (0); 643 } 644 if (off >= va.va_size) 645 cnt = 0; 646 else if ((off + reqlen) > va.va_size) 647 cnt = va.va_size - off; 648 else 649 cnt = reqlen; 650 nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt)); 651 if (v3) { 652 nfsm_build(tl, u_int32_t *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED); 653 *tl++ = nfs_true; 654 fp = (struct nfs_fattr *)tl; 655 tl += (NFSX_V3FATTR / sizeof (u_int32_t)); 656 } else { 657 nfsm_build(tl, u_int32_t *, NFSX_V2FATTR + NFSX_UNSIGNED); 658 fp = (struct nfs_fattr *)tl; 659 tl += (NFSX_V2FATTR / sizeof (u_int32_t)); 660 } 661 len = left = cnt; 662 if (cnt > 0) { 663 if (nfsd_use_loan) { 664 struct vm_page **pgpp; 665 voff_t pgoff = trunc_page(off); 666 int npages; 667 vaddr_t lva; 668 669 npages = (round_page(off + cnt) - pgoff) >> PAGE_SHIFT; 670 KASSERT(npages <= M_EXT_MAXPAGES); /* XXX */ 671 672 /* allocate kva for mbuf data */ 673 lva = sokvaalloc(npages << PAGE_SHIFT, slp->ns_so); 674 if (lva == 0) { 675 /* fall back to VOP_READ */ 676 goto loan_fail; 677 } 678 679 /* allocate mbuf */ 680 m = m_get(M_WAIT, MT_DATA); 681 MCLAIM(m, &nfs_mowner); 682 pgpp = m->m_ext.ext_pgs; 683 684 /* loan pages */ 685 error = uvm_loanuobjpages(&vp->v_uobj, pgoff, npages, 686 pgpp); 687 if (error) { 688 sokvafree(lva, npages << PAGE_SHIFT); 689 m_free(m); 690 goto read_error; 691 } 692 693 /* associate kva to mbuf */ 694 MEXTADD(m, (void *)(lva + ((vaddr_t)off & PAGE_MASK)), 695 cnt, M_MBUF, soloanfree, slp->ns_so); 696 m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP; 697 m->m_len = cnt; 698 699 /* map pages */ 700 for (i = 0; i < npages; i++) { 701 pmap_kenter_pa(lva, VM_PAGE_TO_PHYS(pgpp[i]), 702 VM_PROT_READ); 703 lva += PAGE_SIZE; 704 } 705 706 pmap_update(pmap_kernel()); 707 708 mb->m_next = m; 709 mb = m; 710 error = 0; 711 uiop->uio_resid = 0; 712 } else { 713 struct iovec *iv; 714 struct iovec *iv2; 715 struct mbuf *m2; 716 int siz; 717 loan_fail: 718 /* 719 * Generate the mbuf list with the uio_iov ref. to it. 720 */ 721 i = 0; 722 m = m2 = mb; 723 while (left > 0) { 724 siz = min(M_TRAILINGSPACE(m), left); 725 if (siz > 0) { 726 left -= siz; 727 i++; 728 } 729 if (left > 0) { 730 m = m_get(M_WAIT, MT_DATA); 731 MCLAIM(m, &nfs_mowner); 732 m_clget(m, M_WAIT); 733 m->m_len = 0; 734 m2->m_next = m; 735 m2 = m; 736 } 737 } 738 iv = malloc(i * sizeof(struct iovec), M_TEMP, M_WAITOK); 739 uiop->uio_iov = iv2 = iv; 740 m = mb; 741 left = cnt; 742 i = 0; 743 while (left > 0) { 744 if (m == NULL) 745 panic("nfsrv_read iov"); 746 siz = min(M_TRAILINGSPACE(m), left); 747 if (siz > 0) { 748 iv->iov_base = mtod(m, caddr_t) + 749 m->m_len; 750 iv->iov_len = siz; 751 m->m_len += siz; 752 left -= siz; 753 iv++; 754 i++; 755 } 756 m = m->m_next; 757 } 758 uiop->uio_iovcnt = i; 759 uiop->uio_offset = off; 760 uiop->uio_resid = cnt; 761 uiop->uio_rw = UIO_READ; 762 uiop->uio_segflg = UIO_SYSSPACE; 763 error = VOP_READ(vp, uiop, IO_NODELOCKED, cred); 764 free((caddr_t)iv2, M_TEMP); 765 } 766 read_error: 767 if (error || (getret = VOP_GETATTR(vp, &va, cred, lwp)) != 0){ 768 if (!error) 769 error = getret; 770 m_freem(mreq); 771 vput(vp); 772 nfsm_reply(NFSX_POSTOPATTR(v3)); 773 nfsm_srvpostop_attr(getret, &va); 774 return (0); 775 } 776 } else { 777 uiop->uio_resid = 0; 778 } 779 vput(vp); 780 nfsm_srvfillattr(&va, fp); 781 len -= uiop->uio_resid; 782 padlen = nfsm_padlen(len); 783 if (uiop->uio_resid || padlen) 784 nfs_zeropad(mb, uiop->uio_resid, padlen); 785 if (v3) { 786 /* count */ 787 *tl++ = txdr_unsigned(len); 788 /* eof */ 789 if (off + len >= va.va_size) 790 *tl++ = nfs_true; 791 else 792 *tl++ = nfs_false; 793 } 794 *tl = txdr_unsigned(len); 795 nfsm_srvdone; 796 } 797 798 /* 799 * nfs write service 800 */ 801 int 802 nfsrv_write(nfsd, slp, lwp, mrq) 803 struct nfsrv_descript *nfsd; 804 struct nfssvc_sock *slp; 805 struct lwp *lwp; 806 struct mbuf **mrq; 807 { 808 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 809 struct mbuf *nam = nfsd->nd_nam; 810 caddr_t dpos = nfsd->nd_dpos; 811 struct ucred *cred = &nfsd->nd_cr; 812 struct iovec *ivp; 813 int i, cnt; 814 struct mbuf *mp; 815 struct nfs_fattr *fp; 816 struct iovec *iv; 817 struct vattr va, forat; 818 u_int32_t *tl; 819 int32_t t1; 820 caddr_t bpos; 821 int error = 0, rdonly, cache, len, forat_ret = 1; 822 int ioflags, aftat_ret = 1, retlen, zeroing, adjust; 823 int stable = NFSV3WRITE_FILESYNC; 824 int v3 = (nfsd->nd_flag & ND_NFSV3); 825 char *cp2; 826 struct mbuf *mb, *mreq; 827 struct vnode *vp; 828 nfsfh_t nfh; 829 fhandle_t *fhp; 830 struct uio io, *uiop = &io; 831 off_t off; 832 u_quad_t frev; 833 struct mount *mntp = NULL; 834 835 if (mrep == NULL) { 836 *mrq = NULL; 837 return (0); 838 } 839 fhp = &nfh.fh_generic; 840 nfsm_srvmtofh(fhp); 841 if ((mntp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 842 return (ESTALE); 843 vn_start_write(NULL, &mntp, V_WAIT); 844 if (v3) { 845 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 846 off = fxdr_hyper(tl); 847 tl += 3; 848 stable = fxdr_unsigned(int, *tl++); 849 } else { 850 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 851 off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 852 tl += 2; 853 } 854 retlen = len = fxdr_unsigned(int32_t, *tl); 855 cnt = i = 0; 856 857 /* 858 * For NFS Version 2, it is not obvious what a write of zero length 859 * should do, but I might as well be consistent with Version 3, 860 * which is to return ok so long as there are no permission problems. 861 */ 862 if (len > 0) { 863 zeroing = 1; 864 mp = mrep; 865 while (mp) { 866 if (mp == md) { 867 zeroing = 0; 868 adjust = dpos - mtod(mp, caddr_t); 869 mp->m_len -= adjust; 870 if (mp->m_len > 0 && adjust > 0) 871 NFSMADV(mp, adjust); 872 } 873 if (zeroing) 874 mp->m_len = 0; 875 else if (mp->m_len > 0) { 876 i += mp->m_len; 877 if (i > len) { 878 mp->m_len -= (i - len); 879 zeroing = 1; 880 } 881 if (mp->m_len > 0) 882 cnt++; 883 } 884 mp = mp->m_next; 885 } 886 } 887 if (len > NFS_MAXDATA || len < 0 || i < len) { 888 error = EIO; 889 nfsm_reply(2 * NFSX_UNSIGNED); 890 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 891 vn_finished_write(mntp, 0); 892 return (0); 893 } 894 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 895 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 896 if (error) { 897 nfsm_reply(2 * NFSX_UNSIGNED); 898 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 899 vn_finished_write(mntp, 0); 900 return (0); 901 } 902 if (v3) 903 forat_ret = VOP_GETATTR(vp, &forat, cred, lwp); 904 if (vp->v_type != VREG) { 905 if (v3) 906 error = EINVAL; 907 else 908 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 909 } 910 if (!error) { 911 nqsrv_getl(vp, ND_WRITE); 912 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1); 913 } 914 if (error) { 915 vput(vp); 916 nfsm_reply(NFSX_WCCDATA(v3)); 917 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 918 vn_finished_write(mntp, 0); 919 return (0); 920 } 921 922 if (len > 0) { 923 ivp = malloc(cnt * sizeof (struct iovec), M_TEMP, M_WAITOK); 924 uiop->uio_iov = iv = ivp; 925 uiop->uio_iovcnt = cnt; 926 mp = mrep; 927 while (mp) { 928 if (mp->m_len > 0) { 929 ivp->iov_base = mtod(mp, caddr_t); 930 ivp->iov_len = mp->m_len; 931 ivp++; 932 } 933 mp = mp->m_next; 934 } 935 936 /* 937 * XXX 938 * The IO_METASYNC flag indicates that all metadata (and not 939 * just enough to ensure data integrity) must be written to 940 * stable storage synchronously. 941 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.) 942 */ 943 if (stable == NFSV3WRITE_UNSTABLE) 944 ioflags = IO_NODELOCKED; 945 else if (stable == NFSV3WRITE_DATASYNC) 946 ioflags = (IO_SYNC | IO_NODELOCKED); 947 else 948 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 949 uiop->uio_resid = len; 950 uiop->uio_rw = UIO_WRITE; 951 uiop->uio_segflg = UIO_SYSSPACE; 952 uiop->uio_lwp = (struct lwp *)0; 953 uiop->uio_offset = off; 954 error = VOP_WRITE(vp, uiop, ioflags, cred); 955 nfsstats.srvvop_writes++; 956 free(iv, M_TEMP); 957 } 958 aftat_ret = VOP_GETATTR(vp, &va, cred, lwp); 959 vput(vp); 960 vn_finished_write(mntp, 0); 961 if (!error) 962 error = aftat_ret; 963 nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) + 964 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3)); 965 if (v3) { 966 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 967 if (error) 968 return (0); 969 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 970 *tl++ = txdr_unsigned(retlen); 971 if (stable == NFSV3WRITE_UNSTABLE) 972 *tl++ = txdr_unsigned(stable); 973 else 974 *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC); 975 /* 976 * Actually, there is no need to txdr these fields, 977 * but it may make the values more human readable, 978 * for debugging purposes. 979 */ 980 *tl++ = txdr_unsigned(boottime.tv_sec); 981 *tl = txdr_unsigned(boottime.tv_usec); 982 } else { 983 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 984 nfsm_srvfillattr(&va, fp); 985 } 986 nfsm_srvdone; 987 } 988 989 /* 990 * NFS write service with write gathering support. Called when 991 * nfsrvw_procrastinate > 0. 992 * See: Chet Juszczak, "Improving the Write Performance of an NFS Server", 993 * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco, 994 * Jan. 1994. 995 */ 996 int 997 nfsrv_writegather(ndp, slp, lwp, mrq) 998 struct nfsrv_descript **ndp; 999 struct nfssvc_sock *slp; 1000 struct lwp *lwp; 1001 struct mbuf **mrq; 1002 { 1003 struct iovec *ivp; 1004 struct mbuf *mp; 1005 struct nfsrv_descript *wp, *nfsd, *owp, *swp; 1006 struct nfs_fattr *fp; 1007 int i = 0; 1008 struct iovec *iov; 1009 struct nfsrvw_delayhash *wpp; 1010 struct ucred *cred; 1011 struct vattr va, forat; 1012 u_int32_t *tl; 1013 int32_t t1; 1014 caddr_t bpos, dpos; 1015 int error = 0, rdonly, cache, len = 0, forat_ret = 1; 1016 int ioflags, aftat_ret = 1, s, adjust, v3, zeroing; 1017 char *cp2; 1018 struct mbuf *mb, *mreq, *mrep, *md; 1019 struct vnode *vp; 1020 struct uio io, *uiop = &io; 1021 u_quad_t frev, cur_usec; 1022 struct mount *mntp = NULL; 1023 1024 *mrq = NULL; 1025 if (*ndp) { 1026 nfsd = *ndp; 1027 *ndp = NULL; 1028 mrep = nfsd->nd_mrep; 1029 md = nfsd->nd_md; 1030 dpos = nfsd->nd_dpos; 1031 cred = &nfsd->nd_cr; 1032 v3 = (nfsd->nd_flag & ND_NFSV3); 1033 LIST_INIT(&nfsd->nd_coalesce); 1034 nfsd->nd_mreq = NULL; 1035 nfsd->nd_stable = NFSV3WRITE_FILESYNC; 1036 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec; 1037 nfsd->nd_time = cur_usec + nfsrvw_procrastinate; 1038 1039 /* 1040 * Now, get the write header.. 1041 */ 1042 nfsm_srvmtofh(&nfsd->nd_fh); 1043 if (v3) { 1044 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 1045 nfsd->nd_off = fxdr_hyper(tl); 1046 tl += 3; 1047 nfsd->nd_stable = fxdr_unsigned(int, *tl++); 1048 } else { 1049 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1050 nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 1051 tl += 2; 1052 } 1053 len = fxdr_unsigned(int32_t, *tl); 1054 nfsd->nd_len = len; 1055 nfsd->nd_eoff = nfsd->nd_off + len; 1056 1057 /* 1058 * Trim the header out of the mbuf list and trim off any trailing 1059 * junk so that the mbuf list has only the write data. 1060 */ 1061 zeroing = 1; 1062 i = 0; 1063 mp = mrep; 1064 while (mp) { 1065 if (mp == md) { 1066 zeroing = 0; 1067 adjust = dpos - mtod(mp, caddr_t); 1068 mp->m_len -= adjust; 1069 if (mp->m_len > 0 && adjust > 0) 1070 NFSMADV(mp, adjust); 1071 } 1072 if (zeroing) 1073 mp->m_len = 0; 1074 else { 1075 i += mp->m_len; 1076 if (i > len) { 1077 mp->m_len -= (i - len); 1078 zeroing = 1; 1079 } 1080 } 1081 mp = mp->m_next; 1082 } 1083 if (len > NFS_MAXDATA || len < 0 || i < len) { 1084 nfsmout: 1085 m_freem(mrep); 1086 error = EIO; 1087 nfsm_writereply(2 * NFSX_UNSIGNED, v3); 1088 if (v3) 1089 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1090 nfsd->nd_mreq = mreq; 1091 nfsd->nd_mrep = NULL; 1092 nfsd->nd_time = 0; 1093 } 1094 1095 /* 1096 * Add this entry to the hash and time queues. 1097 */ 1098 s = splsoftclock(); 1099 owp = NULL; 1100 wp = LIST_FIRST(&slp->ns_tq); 1101 while (wp && wp->nd_time < nfsd->nd_time) { 1102 owp = wp; 1103 wp = LIST_NEXT(wp, nd_tq); 1104 } 1105 if (owp) { 1106 LIST_INSERT_AFTER(owp, nfsd, nd_tq); 1107 } else { 1108 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1109 } 1110 if (nfsd->nd_mrep) { 1111 wpp = NWDELAYHASH(slp, nfsd->nd_fh.fh_fid.fid_data); 1112 owp = NULL; 1113 wp = LIST_FIRST(wpp); 1114 while (wp && 1115 memcmp((caddr_t)&nfsd->nd_fh, (caddr_t)&wp->nd_fh, NFSX_V3FH)) { 1116 owp = wp; 1117 wp = LIST_NEXT(wp, nd_hash); 1118 } 1119 while (wp && wp->nd_off < nfsd->nd_off && 1120 !memcmp((caddr_t)&nfsd->nd_fh, (caddr_t)&wp->nd_fh, NFSX_V3FH)) { 1121 owp = wp; 1122 wp = LIST_NEXT(wp, nd_hash); 1123 } 1124 if (owp) { 1125 LIST_INSERT_AFTER(owp, nfsd, nd_hash); 1126 1127 /* 1128 * Search the hash list for overlapping entries and 1129 * coalesce. 1130 */ 1131 for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) { 1132 wp = LIST_NEXT(nfsd, nd_hash); 1133 if (NFSW_SAMECRED(owp, nfsd)) 1134 nfsrvw_coalesce(owp, nfsd); 1135 } 1136 } else { 1137 LIST_INSERT_HEAD(wpp, nfsd, nd_hash); 1138 } 1139 } 1140 splx(s); 1141 } 1142 1143 /* 1144 * Now, do VOP_WRITE()s for any one(s) that need to be done now 1145 * and generate the associated reply mbuf list(s). 1146 */ 1147 loop1: 1148 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec; 1149 s = splsoftclock(); 1150 for (nfsd = LIST_FIRST(&slp->ns_tq); nfsd; nfsd = owp) { 1151 owp = LIST_NEXT(nfsd, nd_tq); 1152 if (nfsd->nd_time > cur_usec) 1153 break; 1154 if (nfsd->nd_mreq) 1155 continue; 1156 LIST_REMOVE(nfsd, nd_tq); 1157 LIST_REMOVE(nfsd, nd_hash); 1158 splx(s); 1159 mrep = nfsd->nd_mrep; 1160 nfsd->nd_mrep = NULL; 1161 cred = &nfsd->nd_cr; 1162 v3 = (nfsd->nd_flag & ND_NFSV3); 1163 forat_ret = aftat_ret = 1; 1164 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp, 1165 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH), 1166 FALSE); 1167 if (!error) { 1168 if (v3) 1169 forat_ret = VOP_GETATTR(vp, &forat, cred, lwp); 1170 if (vp->v_type != VREG) { 1171 if (v3) 1172 error = EINVAL; 1173 else 1174 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 1175 } 1176 } else 1177 vp = NULL; 1178 if (!error) { 1179 nqsrv_getl(vp, ND_WRITE); 1180 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1); 1181 } 1182 1183 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE) 1184 ioflags = IO_NODELOCKED; 1185 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC) 1186 ioflags = (IO_SYNC | IO_NODELOCKED); 1187 else 1188 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 1189 uiop->uio_rw = UIO_WRITE; 1190 uiop->uio_segflg = UIO_SYSSPACE; 1191 uiop->uio_lwp = (struct lwp *)0; 1192 uiop->uio_offset = nfsd->nd_off; 1193 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off; 1194 if (uiop->uio_resid > 0) { 1195 mp = mrep; 1196 i = 0; 1197 while (mp) { 1198 if (mp->m_len > 0) 1199 i++; 1200 mp = mp->m_next; 1201 } 1202 uiop->uio_iovcnt = i; 1203 iov = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK); 1204 uiop->uio_iov = ivp = iov; 1205 mp = mrep; 1206 while (mp) { 1207 if (mp->m_len > 0) { 1208 ivp->iov_base = mtod(mp, caddr_t); 1209 ivp->iov_len = mp->m_len; 1210 ivp++; 1211 } 1212 mp = mp->m_next; 1213 } 1214 if (!error) { 1215 if (vn_start_write(vp, &mntp, V_NOWAIT) != 0) { 1216 VOP_UNLOCK(vp, 0); 1217 vn_start_write(NULL, &mntp, V_WAIT); 1218 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1219 } 1220 if (!error) { 1221 error = VOP_WRITE(vp, uiop, ioflags, cred); 1222 nfsstats.srvvop_writes++; 1223 vn_finished_write(mntp, 0); 1224 } 1225 } 1226 free((caddr_t)iov, M_TEMP); 1227 } 1228 m_freem(mrep); 1229 if (vp) { 1230 aftat_ret = VOP_GETATTR(vp, &va, cred, lwp); 1231 vput(vp); 1232 } 1233 1234 /* 1235 * Loop around generating replies for all write rpcs that have 1236 * now been completed. 1237 */ 1238 swp = nfsd; 1239 do { 1240 if (error) { 1241 nfsm_writereply(NFSX_WCCDATA(v3), v3); 1242 if (v3) { 1243 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1244 } 1245 } else { 1246 nfsm_writereply(NFSX_PREOPATTR(v3) + 1247 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED + 1248 NFSX_WRITEVERF(v3), v3); 1249 if (v3) { 1250 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1251 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1252 *tl++ = txdr_unsigned(nfsd->nd_len); 1253 *tl++ = txdr_unsigned(swp->nd_stable); 1254 /* 1255 * Actually, there is no need to txdr these fields, 1256 * but it may make the values more human readable, 1257 * for debugging purposes. 1258 */ 1259 *tl++ = txdr_unsigned(boottime.tv_sec); 1260 *tl = txdr_unsigned(boottime.tv_usec); 1261 } else { 1262 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1263 nfsm_srvfillattr(&va, fp); 1264 } 1265 } 1266 nfsd->nd_mreq = mreq; 1267 if (nfsd->nd_mrep) 1268 panic("nfsrv_write: nd_mrep not free"); 1269 1270 /* 1271 * Done. Put it at the head of the timer queue so that 1272 * the final phase can return the reply. 1273 */ 1274 s = splsoftclock(); 1275 if (nfsd != swp) { 1276 nfsd->nd_time = 0; 1277 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1278 } 1279 nfsd = LIST_FIRST(&swp->nd_coalesce); 1280 if (nfsd) { 1281 LIST_REMOVE(nfsd, nd_tq); 1282 } 1283 splx(s); 1284 } while (nfsd); 1285 s = splsoftclock(); 1286 swp->nd_time = 0; 1287 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq); 1288 splx(s); 1289 goto loop1; 1290 } 1291 splx(s); 1292 1293 /* 1294 * Search for a reply to return. 1295 */ 1296 s = splsoftclock(); 1297 LIST_FOREACH(nfsd, &slp->ns_tq, nd_tq) { 1298 if (nfsd->nd_mreq) { 1299 LIST_REMOVE(nfsd, nd_tq); 1300 *mrq = nfsd->nd_mreq; 1301 *ndp = nfsd; 1302 break; 1303 } 1304 } 1305 splx(s); 1306 return (0); 1307 } 1308 1309 /* 1310 * Coalesce the write request nfsd into owp. To do this we must: 1311 * - remove nfsd from the queues 1312 * - merge nfsd->nd_mrep into owp->nd_mrep 1313 * - update the nd_eoff and nd_stable for owp 1314 * - put nfsd on owp's nd_coalesce list 1315 * NB: Must be called at splsoftclock(). 1316 */ 1317 void 1318 nfsrvw_coalesce(owp, nfsd) 1319 struct nfsrv_descript *owp; 1320 struct nfsrv_descript *nfsd; 1321 { 1322 int overlap; 1323 struct mbuf *mp; 1324 struct nfsrv_descript *m; 1325 1326 LIST_REMOVE(nfsd, nd_hash); 1327 LIST_REMOVE(nfsd, nd_tq); 1328 if (owp->nd_eoff < nfsd->nd_eoff) { 1329 overlap = owp->nd_eoff - nfsd->nd_off; 1330 if (overlap < 0) 1331 panic("nfsrv_coalesce: bad off"); 1332 if (overlap > 0) 1333 m_adj(nfsd->nd_mrep, overlap); 1334 mp = owp->nd_mrep; 1335 while (mp->m_next) 1336 mp = mp->m_next; 1337 mp->m_next = nfsd->nd_mrep; 1338 owp->nd_eoff = nfsd->nd_eoff; 1339 } else 1340 m_freem(nfsd->nd_mrep); 1341 nfsd->nd_mrep = NULL; 1342 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC) 1343 owp->nd_stable = NFSV3WRITE_FILESYNC; 1344 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC && 1345 owp->nd_stable == NFSV3WRITE_UNSTABLE) 1346 owp->nd_stable = NFSV3WRITE_DATASYNC; 1347 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq); 1348 /* 1349 * nfsd might hold coalesce elements! Move them to owp. 1350 * Otherwise, requests may be lost and clients will be stuck. 1351 */ 1352 while ((m = LIST_FIRST(&nfsd->nd_coalesce)) != NULL) { 1353 LIST_REMOVE(m, nd_tq); 1354 LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq); 1355 } 1356 } 1357 1358 /* 1359 * nfs create service 1360 * now does a truncate to 0 length via. setattr if it already exists 1361 */ 1362 int 1363 nfsrv_create(nfsd, slp, lwp, mrq) 1364 struct nfsrv_descript *nfsd; 1365 struct nfssvc_sock *slp; 1366 struct lwp *lwp; 1367 struct mbuf **mrq; 1368 { 1369 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1370 struct mbuf *nam = nfsd->nd_nam; 1371 caddr_t dpos = nfsd->nd_dpos; 1372 struct ucred *cred = &nfsd->nd_cr; 1373 struct nfs_fattr *fp; 1374 struct vattr va, dirfor, diraft; 1375 struct nfsv2_sattr *sp; 1376 u_int32_t *tl; 1377 struct nameidata nd; 1378 caddr_t cp; 1379 int32_t t1; 1380 caddr_t bpos; 1381 int error = 0, cache, len, tsize, dirfor_ret = 1, diraft_ret = 1; 1382 int rdev = 0; 1383 int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0; 1384 char *cp2; 1385 struct mbuf *mb, *mreq; 1386 struct vnode *vp = NULL, *dirp = NULL; 1387 nfsfh_t nfh; 1388 fhandle_t *fhp; 1389 u_quad_t frev, tempsize; 1390 u_char cverf[NFSX_V3CREATEVERF]; 1391 struct mount *mp = NULL; 1392 1393 nd.ni_cnd.cn_nameiop = 0; 1394 fhp = &nfh.fh_generic; 1395 nfsm_srvmtofh(fhp); 1396 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1397 return (ESTALE); 1398 vn_start_write(NULL, &mp, V_WAIT); 1399 nfsm_srvnamesiz(len); 1400 nd.ni_cnd.cn_cred = cred; 1401 nd.ni_cnd.cn_nameiop = CREATE; 1402 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1403 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1404 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1405 if (dirp) { 1406 if (v3) 1407 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1408 else { 1409 vrele(dirp); 1410 dirp = (struct vnode *)0; 1411 } 1412 } 1413 if (error) { 1414 nfsm_reply(NFSX_WCCDATA(v3)); 1415 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1416 if (dirp) 1417 vrele(dirp); 1418 vn_finished_write(mp, 0); 1419 return (0); 1420 } 1421 VATTR_NULL(&va); 1422 if (v3) { 1423 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1424 how = fxdr_unsigned(int, *tl); 1425 switch (how) { 1426 case NFSV3CREATE_GUARDED: 1427 if (nd.ni_vp) { 1428 error = EEXIST; 1429 break; 1430 } 1431 case NFSV3CREATE_UNCHECKED: 1432 nfsm_srvsattr(&va); 1433 break; 1434 case NFSV3CREATE_EXCLUSIVE: 1435 nfsm_dissect(cp, caddr_t, NFSX_V3CREATEVERF); 1436 memcpy(cverf, cp, NFSX_V3CREATEVERF); 1437 exclusive_flag = 1; 1438 if (nd.ni_vp == NULL) 1439 va.va_mode = 0; 1440 break; 1441 }; 1442 va.va_type = VREG; 1443 } else { 1444 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 1445 va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode)); 1446 if (va.va_type == VNON) 1447 va.va_type = VREG; 1448 va.va_mode = nfstov_mode(sp->sa_mode); 1449 switch (va.va_type) { 1450 case VREG: 1451 tsize = fxdr_unsigned(int32_t, sp->sa_size); 1452 if (tsize != -1) 1453 va.va_size = (u_quad_t)tsize; 1454 break; 1455 case VCHR: 1456 case VBLK: 1457 case VFIFO: 1458 rdev = fxdr_unsigned(int32_t, sp->sa_size); 1459 break; 1460 default: 1461 break; 1462 }; 1463 } 1464 1465 /* 1466 * Iff doesn't exist, create it 1467 * otherwise just truncate to 0 length 1468 * should I set the mode too ?? 1469 */ 1470 if (nd.ni_vp == NULL) { 1471 if (va.va_type == VREG || va.va_type == VSOCK) { 1472 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1473 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1474 if (!error) { 1475 if (exclusive_flag) { 1476 exclusive_flag = 0; 1477 VATTR_NULL(&va); 1478 /* 1479 * XXX 1480 * assuming NFSX_V3CREATEVERF 1481 * == sizeof(nfstime3) 1482 */ 1483 fxdr_nfsv3time(cverf, &va.va_atime); 1484 error = VOP_SETATTR(nd.ni_vp, &va, cred, 1485 lwp); 1486 } 1487 } 1488 } else if (va.va_type == VCHR || va.va_type == VBLK || 1489 va.va_type == VFIFO) { 1490 if (va.va_type == VCHR && rdev == 0xffffffff) 1491 va.va_type = VFIFO; 1492 if (va.va_type != VFIFO && 1493 (error = suser(cred, (u_short *)0))) { 1494 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1495 vput(nd.ni_dvp); 1496 nfsm_reply(0); 1497 vn_finished_write(mp, 0); 1498 return (error); 1499 } else 1500 va.va_rdev = (dev_t)rdev; 1501 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1502 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, 1503 &va); 1504 if (error) { 1505 nfsm_reply(0); 1506 } 1507 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1508 vrele(nd.ni_dvp); 1509 vput(nd.ni_vp); 1510 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1511 error = EINVAL; 1512 nfsm_reply(0); 1513 } 1514 } else { 1515 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1516 vput(nd.ni_dvp); 1517 error = ENXIO; 1518 } 1519 vp = nd.ni_vp; 1520 } else { 1521 vp = nd.ni_vp; 1522 if (nd.ni_dvp == vp) 1523 vrele(nd.ni_dvp); 1524 else 1525 vput(nd.ni_dvp); 1526 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1527 if (!error && va.va_size != -1) { 1528 error = nfsrv_access(vp, VWRITE, cred, 1529 (nd.ni_cnd.cn_flags & RDONLY), lwp, 0); 1530 if (!error) { 1531 nqsrv_getl(vp, ND_WRITE); 1532 tempsize = va.va_size; 1533 VATTR_NULL(&va); 1534 va.va_size = tempsize; 1535 error = VOP_SETATTR(vp, &va, cred, lwp); 1536 } 1537 } 1538 if (error) 1539 vput(vp); 1540 } 1541 if (!error) { 1542 memset((caddr_t)fhp, 0, sizeof(nfh)); 1543 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1544 error = VFS_VPTOFH(vp, &fhp->fh_fid); 1545 if (!error) 1546 error = VOP_GETATTR(vp, &va, cred, lwp); 1547 vput(vp); 1548 } 1549 if (v3) { 1550 if (exclusive_flag && !error) { 1551 /* 1552 * XXX assuming NFSX_V3CREATEVERF == sizeof(nfstime3) 1553 */ 1554 char oldverf[NFSX_V3CREATEVERF]; 1555 1556 txdr_nfsv3time(&va.va_atime, oldverf); 1557 if (memcmp(cverf, oldverf, NFSX_V3CREATEVERF)) 1558 error = EEXIST; 1559 } 1560 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1561 vrele(dirp); 1562 } 1563 nfsm_reply(NFSX_SRVFH(v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3)); 1564 if (v3) { 1565 if (!error) { 1566 nfsm_srvpostop_fh(fhp); 1567 nfsm_srvpostop_attr(0, &va); 1568 } 1569 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1570 } else { 1571 nfsm_srvfhtom(fhp, v3); 1572 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1573 nfsm_srvfillattr(&va, fp); 1574 } 1575 vn_finished_write(mp, 0); 1576 return (0); 1577 nfsmout: 1578 if (dirp) 1579 vrele(dirp); 1580 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1581 if (nd.ni_dvp == nd.ni_vp) 1582 vrele(nd.ni_dvp); 1583 else 1584 vput(nd.ni_dvp); 1585 if (nd.ni_vp) 1586 vput(nd.ni_vp); 1587 vn_finished_write(mp, 0); 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 struct mount *mp = NULL; 1620 1621 nd.ni_cnd.cn_nameiop = 0; 1622 fhp = &nfh.fh_generic; 1623 nfsm_srvmtofh(fhp); 1624 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1625 return (ESTALE); 1626 vn_start_write(NULL, &mp, V_WAIT); 1627 nfsm_srvnamesiz(len); 1628 nd.ni_cnd.cn_cred = cred; 1629 nd.ni_cnd.cn_nameiop = CREATE; 1630 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1631 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1632 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1633 if (dirp) 1634 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1635 if (error) { 1636 nfsm_reply(NFSX_WCCDATA(1)); 1637 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1638 if (dirp) 1639 vrele(dirp); 1640 vn_finished_write(mp, 0); 1641 return (0); 1642 } 1643 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1644 vtyp = nfsv3tov_type(*tl); 1645 if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) { 1646 error = NFSERR_BADTYPE; 1647 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1648 if (nd.ni_dvp == nd.ni_vp) 1649 vrele(nd.ni_dvp); 1650 else 1651 vput(nd.ni_dvp); 1652 if (nd.ni_vp) 1653 vput(nd.ni_vp); 1654 goto out; 1655 } 1656 VATTR_NULL(&va); 1657 nfsm_srvsattr(&va); 1658 if (vtyp == VCHR || vtyp == VBLK) { 1659 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1660 major = fxdr_unsigned(u_int32_t, *tl++); 1661 minor = fxdr_unsigned(u_int32_t, *tl); 1662 va.va_rdev = makedev(major, minor); 1663 } 1664 1665 /* 1666 * Iff doesn't exist, create it. 1667 */ 1668 if (nd.ni_vp) { 1669 error = EEXIST; 1670 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1671 if (nd.ni_dvp == nd.ni_vp) 1672 vrele(nd.ni_dvp); 1673 else 1674 vput(nd.ni_dvp); 1675 vput(nd.ni_vp); 1676 goto out; 1677 } 1678 va.va_type = vtyp; 1679 if (vtyp == VSOCK) { 1680 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1681 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1682 } else { 1683 if (va.va_type != VFIFO && 1684 (error = suser(cred, (u_short *)0))) { 1685 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1686 vput(nd.ni_dvp); 1687 goto out; 1688 } 1689 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1690 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1691 if (error) { 1692 goto out; 1693 } 1694 if (error) 1695 goto out; 1696 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1697 vput(nd.ni_vp); 1698 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1699 error = EINVAL; 1700 } 1701 } 1702 out: 1703 vp = nd.ni_vp; 1704 if (!error) { 1705 memset((caddr_t)fhp, 0, sizeof(nfh)); 1706 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1707 error = VFS_VPTOFH(vp, &fhp->fh_fid); 1708 if (!error) 1709 error = VOP_GETATTR(vp, &va, cred, lwp); 1710 vput(vp); 1711 } 1712 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1713 vrele(dirp); 1714 nfsm_reply(NFSX_SRVFH(1) + NFSX_POSTOPATTR(1) + NFSX_WCCDATA(1)); 1715 if (!error) { 1716 nfsm_srvpostop_fh(fhp); 1717 nfsm_srvpostop_attr(0, &va); 1718 } 1719 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1720 vn_finished_write(mp, 0); 1721 return (0); 1722 nfsmout: 1723 if (dirp) 1724 vrele(dirp); 1725 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1726 if (nd.ni_dvp == nd.ni_vp) 1727 vrele(nd.ni_dvp); 1728 else 1729 vput(nd.ni_dvp); 1730 if (nd.ni_vp) 1731 vput(nd.ni_vp); 1732 vn_finished_write(mp, 0); 1733 return (error); 1734 } 1735 1736 /* 1737 * nfs remove service 1738 */ 1739 int 1740 nfsrv_remove(nfsd, slp, lwp, mrq) 1741 struct nfsrv_descript *nfsd; 1742 struct nfssvc_sock *slp; 1743 struct lwp *lwp; 1744 struct mbuf **mrq; 1745 { 1746 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1747 struct mbuf *nam = nfsd->nd_nam; 1748 caddr_t dpos = nfsd->nd_dpos; 1749 struct ucred *cred = &nfsd->nd_cr; 1750 struct nameidata nd; 1751 u_int32_t *tl; 1752 int32_t t1; 1753 caddr_t bpos; 1754 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 1755 int v3 = (nfsd->nd_flag & ND_NFSV3); 1756 char *cp2; 1757 struct mbuf *mb, *mreq; 1758 struct vnode *vp, *dirp; 1759 struct vattr dirfor, diraft; 1760 nfsfh_t nfh; 1761 fhandle_t *fhp; 1762 u_quad_t frev; 1763 struct mount *mp = NULL; 1764 1765 #ifndef nolint 1766 vp = (struct vnode *)0; 1767 #endif 1768 fhp = &nfh.fh_generic; 1769 nfsm_srvmtofh(fhp); 1770 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1771 return (ESTALE); 1772 vn_start_write(NULL, &mp, V_WAIT); 1773 nfsm_srvnamesiz(len); 1774 nd.ni_cnd.cn_cred = cred; 1775 nd.ni_cnd.cn_nameiop = DELETE; 1776 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1777 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1778 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1779 if (dirp) { 1780 if (v3) 1781 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1782 else 1783 vrele(dirp); 1784 } 1785 if (!error) { 1786 vp = nd.ni_vp; 1787 if (vp->v_type == VDIR && 1788 (error = suser(cred, (u_short *)0)) != 0) 1789 goto out; 1790 /* 1791 * The root of a mounted filesystem cannot be deleted. 1792 */ 1793 if (vp->v_flag & VROOT) { 1794 error = EBUSY; 1795 goto out; 1796 } 1797 out: 1798 if (!error) { 1799 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1800 nqsrv_getl(vp, ND_WRITE); 1801 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1802 } else { 1803 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1804 if (nd.ni_dvp == vp) 1805 vrele(nd.ni_dvp); 1806 else 1807 vput(nd.ni_dvp); 1808 vput(vp); 1809 } 1810 } 1811 if (dirp && v3) { 1812 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1813 vrele(dirp); 1814 } 1815 nfsm_reply(NFSX_WCCDATA(v3)); 1816 if (v3) { 1817 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1818 vn_finished_write(mp, 0); 1819 return (0); 1820 } 1821 vn_finished_write(mp, 0); 1822 nfsm_srvdone; 1823 } 1824 1825 /* 1826 * nfs rename service 1827 */ 1828 int 1829 nfsrv_rename(nfsd, slp, lwp, mrq) 1830 struct nfsrv_descript *nfsd; 1831 struct nfssvc_sock *slp; 1832 struct lwp *lwp; 1833 struct mbuf **mrq; 1834 { 1835 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1836 struct mbuf *nam = nfsd->nd_nam; 1837 caddr_t dpos = nfsd->nd_dpos; 1838 struct ucred *cred = &nfsd->nd_cr; 1839 u_int32_t *tl; 1840 int32_t t1; 1841 caddr_t bpos; 1842 int error = 0, cache, fdirfor_ret = 1, fdiraft_ret = 1; 1843 uint32_t len, len2; 1844 int tdirfor_ret = 1, tdiraft_ret = 1; 1845 int v3 = (nfsd->nd_flag & ND_NFSV3); 1846 char *cp2; 1847 struct mbuf *mb, *mreq; 1848 struct nameidata fromnd, tond; 1849 struct vnode *fvp, *tvp, *tdvp, *fdirp = (struct vnode *)0; 1850 struct vnode *tdirp = (struct vnode *)0; 1851 struct vattr fdirfor, fdiraft, tdirfor, tdiraft; 1852 nfsfh_t fnfh, tnfh; 1853 fhandle_t *ffhp, *tfhp; 1854 u_quad_t frev; 1855 uid_t saved_uid; 1856 struct mount *mp = NULL; 1857 1858 #ifndef nolint 1859 fvp = (struct vnode *)0; 1860 #endif 1861 ffhp = &fnfh.fh_generic; 1862 tfhp = &tnfh.fh_generic; 1863 fromnd.ni_cnd.cn_nameiop = 0; 1864 tond.ni_cnd.cn_nameiop = 0; 1865 nfsm_srvmtofh(ffhp); 1866 if ((mp = vfs_getvfs(&ffhp->fh_fsid)) == NULL) 1867 return (ESTALE); 1868 vn_start_write(NULL, &mp, V_WAIT); 1869 nfsm_srvnamesiz(len); 1870 /* 1871 * Remember our original uid so that we can reset cr_uid before 1872 * the second nfs_namei() call, in case it is remapped. 1873 */ 1874 saved_uid = cred->cr_uid; 1875 fromnd.ni_cnd.cn_cred = cred; 1876 fromnd.ni_cnd.cn_nameiop = DELETE; 1877 fromnd.ni_cnd.cn_flags = WANTPARENT | SAVESTART; 1878 error = nfs_namei(&fromnd, ffhp, len, slp, nam, &md, 1879 &dpos, &fdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1880 if (fdirp) { 1881 if (v3) 1882 fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred, lwp); 1883 else { 1884 vrele(fdirp); 1885 fdirp = (struct vnode *)0; 1886 } 1887 } 1888 if (error) { 1889 nfsm_reply(2 * NFSX_WCCDATA(v3)); 1890 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 1891 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 1892 if (fdirp) 1893 vrele(fdirp); 1894 vn_finished_write(mp, 0); 1895 return (0); 1896 } 1897 fvp = fromnd.ni_vp; 1898 nfsm_srvmtofh(tfhp); 1899 if (v3) { 1900 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 1901 len2 = fxdr_unsigned(uint32_t, *tl); 1902 /* len2 will be checked by nfs_namei */ 1903 } 1904 else { 1905 /* NFSv2 */ 1906 nfsm_strsiz(len2, NFS_MAXNAMLEN); 1907 } 1908 cred->cr_uid = saved_uid; 1909 tond.ni_cnd.cn_cred = cred; 1910 tond.ni_cnd.cn_nameiop = RENAME; 1911 tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART; 1912 error = nfs_namei(&tond, tfhp, len2, slp, nam, &md, 1913 &dpos, &tdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1914 if (tdirp) { 1915 if (v3) 1916 tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred, lwp); 1917 else { 1918 vrele(tdirp); 1919 tdirp = (struct vnode *)0; 1920 } 1921 } 1922 if (error) { 1923 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1924 vrele(fromnd.ni_dvp); 1925 vrele(fvp); 1926 goto out1; 1927 } 1928 tdvp = tond.ni_dvp; 1929 tvp = tond.ni_vp; 1930 if (tvp != NULL) { 1931 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 1932 if (v3) 1933 error = EEXIST; 1934 else 1935 error = EISDIR; 1936 goto out; 1937 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 1938 if (v3) 1939 error = EEXIST; 1940 else 1941 error = ENOTDIR; 1942 goto out; 1943 } 1944 if (tvp->v_type == VDIR && tvp->v_mountedhere) { 1945 if (v3) 1946 error = EXDEV; 1947 else 1948 error = ENOTEMPTY; 1949 goto out; 1950 } 1951 } 1952 if (fvp->v_type == VDIR && fvp->v_mountedhere) { 1953 if (v3) 1954 error = EXDEV; 1955 else 1956 error = ENOTEMPTY; 1957 goto out; 1958 } 1959 if (fvp->v_mount != tdvp->v_mount) { 1960 if (v3) 1961 error = EXDEV; 1962 else 1963 error = ENOTEMPTY; 1964 goto out; 1965 } 1966 if (fvp == tdvp) { 1967 if (v3) 1968 error = EINVAL; 1969 else 1970 error = ENOTEMPTY; 1971 } 1972 /* 1973 * If source is the same as the destination (that is the 1974 * same vnode with the same name in the same directory), 1975 * then there is nothing to do. 1976 */ 1977 if (fvp == tvp && fromnd.ni_dvp == tdvp && 1978 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 1979 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr, 1980 fromnd.ni_cnd.cn_namelen)) 1981 error = -1; 1982 out: 1983 if (!error) { 1984 nqsrv_getl(fromnd.ni_dvp, ND_WRITE); 1985 nqsrv_getl(tdvp, ND_WRITE); 1986 if (tvp) { 1987 nqsrv_getl(tvp, ND_WRITE); 1988 } 1989 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 1990 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 1991 } else { 1992 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 1993 if (tdvp == tvp) 1994 vrele(tdvp); 1995 else 1996 vput(tdvp); 1997 if (tvp) 1998 vput(tvp); 1999 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2000 vrele(fromnd.ni_dvp); 2001 vrele(fvp); 2002 if (error == -1) 2003 error = 0; 2004 } 2005 vrele(tond.ni_startdir); 2006 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 2007 out1: 2008 if (fdirp) { 2009 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred, lwp); 2010 vrele(fdirp); 2011 } 2012 if (tdirp) { 2013 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred, lwp); 2014 vrele(tdirp); 2015 } 2016 vrele(fromnd.ni_startdir); 2017 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2018 nfsm_reply(2 * NFSX_WCCDATA(v3)); 2019 if (v3) { 2020 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 2021 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 2022 } 2023 vn_finished_write(mp, 0); 2024 return (0); 2025 2026 nfsmout: 2027 if (fdirp) 2028 vrele(fdirp); 2029 if (tdirp) 2030 vrele(tdirp); 2031 if (tond.ni_cnd.cn_nameiop) { 2032 vrele(tond.ni_startdir); 2033 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 2034 } 2035 if (fromnd.ni_cnd.cn_nameiop) { 2036 vrele(fromnd.ni_startdir); 2037 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2038 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2039 vrele(fromnd.ni_dvp); 2040 vrele(fvp); 2041 } 2042 return (error); 2043 } 2044 2045 /* 2046 * nfs link service 2047 */ 2048 int 2049 nfsrv_link(nfsd, slp, lwp, mrq) 2050 struct nfsrv_descript *nfsd; 2051 struct nfssvc_sock *slp; 2052 struct lwp *lwp; 2053 struct mbuf **mrq; 2054 { 2055 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2056 struct mbuf *nam = nfsd->nd_nam; 2057 caddr_t dpos = nfsd->nd_dpos; 2058 struct ucred *cred = &nfsd->nd_cr; 2059 struct nameidata nd; 2060 u_int32_t *tl; 2061 int32_t t1; 2062 caddr_t bpos; 2063 int error = 0, rdonly, cache, len, dirfor_ret = 1, diraft_ret = 1; 2064 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3); 2065 char *cp2; 2066 struct mbuf *mb, *mreq; 2067 struct vnode *vp, *xp, *dirp = (struct vnode *)0; 2068 struct vattr dirfor, diraft, at; 2069 nfsfh_t nfh, dnfh; 2070 fhandle_t *fhp, *dfhp; 2071 u_quad_t frev; 2072 struct mount *mp = NULL; 2073 2074 fhp = &nfh.fh_generic; 2075 dfhp = &dnfh.fh_generic; 2076 nfsm_srvmtofh(fhp); 2077 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2078 return (ESTALE); 2079 vn_start_write(NULL, &mp, V_WAIT); 2080 nfsm_srvmtofh(dfhp); 2081 nfsm_srvnamesiz(len); 2082 error = nfsrv_fhtovp(fhp, FALSE, &vp, cred, slp, nam, 2083 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2084 if (error) { 2085 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2086 nfsm_srvpostop_attr(getret, &at); 2087 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2088 vn_finished_write(mp, 0); 2089 return (0); 2090 } 2091 if (vp->v_type == VDIR && (error = suser(cred, (u_short *)0)) != 0) 2092 goto out1; 2093 nd.ni_cnd.cn_cred = cred; 2094 nd.ni_cnd.cn_nameiop = CREATE; 2095 nd.ni_cnd.cn_flags = LOCKPARENT; 2096 error = nfs_namei(&nd, dfhp, len, slp, nam, &md, &dpos, 2097 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2098 if (dirp) { 2099 if (v3) 2100 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2101 else { 2102 vrele(dirp); 2103 dirp = (struct vnode *)0; 2104 } 2105 } 2106 if (error) 2107 goto out1; 2108 xp = nd.ni_vp; 2109 if (xp != NULL) { 2110 error = EEXIST; 2111 goto out; 2112 } 2113 xp = nd.ni_dvp; 2114 if (vp->v_mount != xp->v_mount) 2115 error = EXDEV; 2116 out: 2117 if (!error) { 2118 nqsrv_getl(vp, ND_WRITE); 2119 nqsrv_getl(xp, ND_WRITE); 2120 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2121 } else { 2122 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2123 if (nd.ni_dvp == nd.ni_vp) 2124 vrele(nd.ni_dvp); 2125 else 2126 vput(nd.ni_dvp); 2127 if (nd.ni_vp) 2128 vrele(nd.ni_vp); 2129 } 2130 out1: 2131 if (v3) 2132 getret = VOP_GETATTR(vp, &at, cred, lwp); 2133 if (dirp) { 2134 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2135 vrele(dirp); 2136 } 2137 vrele(vp); 2138 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2139 if (v3) { 2140 nfsm_srvpostop_attr(getret, &at); 2141 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2142 vn_finished_write(mp, 0); 2143 return (0); 2144 } 2145 vn_finished_write(mp, 0); 2146 nfsm_srvdone; 2147 } 2148 2149 /* 2150 * nfs symbolic link service 2151 */ 2152 int 2153 nfsrv_symlink(nfsd, slp, lwp, mrq) 2154 struct nfsrv_descript *nfsd; 2155 struct nfssvc_sock *slp; 2156 struct lwp *lwp; 2157 struct mbuf **mrq; 2158 { 2159 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2160 struct mbuf *nam = nfsd->nd_nam; 2161 caddr_t dpos = nfsd->nd_dpos; 2162 struct ucred *cred = &nfsd->nd_cr; 2163 struct vattr va, dirfor, diraft; 2164 struct nameidata nd; 2165 u_int32_t *tl; 2166 int32_t t1; 2167 struct nfsv2_sattr *sp; 2168 char *bpos, *pathcp = NULL, *cp2; 2169 struct uio io; 2170 struct iovec iv; 2171 int error = 0, cache, dirfor_ret = 1, diraft_ret = 1; 2172 uint32_t len, len2; 2173 int v3 = (nfsd->nd_flag & ND_NFSV3); 2174 struct mbuf *mb, *mreq; 2175 struct vnode *dirp = (struct vnode *)0; 2176 nfsfh_t nfh; 2177 fhandle_t *fhp; 2178 u_quad_t frev; 2179 struct mount *mp = NULL; 2180 2181 nd.ni_cnd.cn_nameiop = 0; 2182 fhp = &nfh.fh_generic; 2183 nfsm_srvmtofh(fhp); 2184 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2185 return (ESTALE); 2186 vn_start_write(NULL, &mp, V_WAIT); 2187 nfsm_srvnamesiz(len); 2188 nd.ni_cnd.cn_cred = cred; 2189 nd.ni_cnd.cn_nameiop = CREATE; 2190 nd.ni_cnd.cn_flags = LOCKPARENT; 2191 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2192 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2193 if (dirp) { 2194 if (v3) 2195 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2196 else { 2197 vrele(dirp); 2198 dirp = (struct vnode *)0; 2199 } 2200 } 2201 if (error) 2202 goto out; 2203 VATTR_NULL(&va); 2204 if (v3) { 2205 nfsm_srvsattr(&va); 2206 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 2207 len2 = fxdr_unsigned(uint32_t, *tl); 2208 if (len2 > PATH_MAX) { 2209 /* XXX should check _PC_NO_TRUNC */ 2210 error = ENAMETOOLONG; 2211 goto abortop; 2212 } 2213 } 2214 else { 2215 /* NFSv2 */ 2216 nfsm_strsiz(len2, NFS_MAXPATHLEN); 2217 } 2218 pathcp = malloc(len2 + 1, M_TEMP, M_WAITOK); 2219 iv.iov_base = pathcp; 2220 iv.iov_len = len2; 2221 io.uio_resid = len2; 2222 io.uio_offset = 0; 2223 io.uio_iov = &iv; 2224 io.uio_iovcnt = 1; 2225 io.uio_segflg = UIO_SYSSPACE; 2226 io.uio_rw = UIO_READ; 2227 io.uio_lwp = (struct lwp *)0; 2228 nfsm_mtouio(&io, len2); 2229 if (!v3) { 2230 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 2231 va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode); 2232 } 2233 *(pathcp + len2) = '\0'; 2234 if (nd.ni_vp) { 2235 error = EEXIST; 2236 abortop: 2237 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2238 if (nd.ni_dvp == nd.ni_vp) 2239 vrele(nd.ni_dvp); 2240 else 2241 vput(nd.ni_dvp); 2242 if (nd.ni_vp) 2243 vrele(nd.ni_vp); 2244 goto out; 2245 } 2246 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2247 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp); 2248 if (!error) { 2249 if (v3) { 2250 memset((caddr_t)fhp, 0, sizeof(nfh)); 2251 fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsidx; 2252 error = VFS_VPTOFH(nd.ni_vp, &fhp->fh_fid); 2253 if (!error) 2254 error = VOP_GETATTR(nd.ni_vp, &va, cred, lwp); 2255 vput(nd.ni_vp); 2256 } else { 2257 vput(nd.ni_vp); 2258 } 2259 } 2260 out: 2261 if (pathcp) 2262 free(pathcp, M_TEMP); 2263 if (dirp) { 2264 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2265 vrele(dirp); 2266 } 2267 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2268 if (v3) { 2269 if (!error) { 2270 nfsm_srvpostop_fh(fhp); 2271 nfsm_srvpostop_attr(0, &va); 2272 } 2273 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2274 } 2275 vn_finished_write(mp, 0); 2276 return (0); 2277 nfsmout: 2278 if (dirp) 2279 vrele(dirp); 2280 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2281 if (nd.ni_dvp == nd.ni_vp) 2282 vrele(nd.ni_dvp); 2283 else 2284 vput(nd.ni_dvp); 2285 if (nd.ni_vp) 2286 vrele(nd.ni_vp); 2287 if (pathcp) 2288 free(pathcp, M_TEMP); 2289 vn_finished_write(mp, 0); 2290 return (error); 2291 } 2292 2293 /* 2294 * nfs mkdir service 2295 */ 2296 int 2297 nfsrv_mkdir(nfsd, slp, lwp, mrq) 2298 struct nfsrv_descript *nfsd; 2299 struct nfssvc_sock *slp; 2300 struct lwp *lwp; 2301 struct mbuf **mrq; 2302 { 2303 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2304 struct mbuf *nam = nfsd->nd_nam; 2305 caddr_t dpos = nfsd->nd_dpos; 2306 struct ucred *cred = &nfsd->nd_cr; 2307 struct vattr va, dirfor, diraft; 2308 struct nfs_fattr *fp; 2309 struct nameidata nd; 2310 caddr_t cp; 2311 u_int32_t *tl; 2312 int32_t t1; 2313 caddr_t bpos; 2314 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 2315 int v3 = (nfsd->nd_flag & ND_NFSV3); 2316 char *cp2; 2317 struct mbuf *mb, *mreq; 2318 struct vnode *vp, *dirp = (struct vnode *)0; 2319 nfsfh_t nfh; 2320 fhandle_t *fhp; 2321 u_quad_t frev; 2322 struct mount *mp = NULL; 2323 2324 fhp = &nfh.fh_generic; 2325 nfsm_srvmtofh(fhp); 2326 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2327 return (ESTALE); 2328 vn_start_write(NULL, &mp, V_WAIT); 2329 nfsm_srvnamesiz(len); 2330 nd.ni_cnd.cn_cred = cred; 2331 nd.ni_cnd.cn_nameiop = CREATE; 2332 nd.ni_cnd.cn_flags = LOCKPARENT; 2333 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2334 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2335 if (dirp) { 2336 if (v3) 2337 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2338 else { 2339 vrele(dirp); 2340 dirp = (struct vnode *)0; 2341 } 2342 } 2343 if (error) { 2344 nfsm_reply(NFSX_WCCDATA(v3)); 2345 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2346 if (dirp) 2347 vrele(dirp); 2348 vn_finished_write(mp, 0); 2349 return (0); 2350 } 2351 VATTR_NULL(&va); 2352 if (v3) { 2353 nfsm_srvsattr(&va); 2354 } else { 2355 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 2356 va.va_mode = nfstov_mode(*tl++); 2357 } 2358 va.va_type = VDIR; 2359 vp = nd.ni_vp; 2360 if (vp != NULL) { 2361 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2362 if (nd.ni_dvp == vp) 2363 vrele(nd.ni_dvp); 2364 else 2365 vput(nd.ni_dvp); 2366 vrele(vp); 2367 error = EEXIST; 2368 goto out; 2369 } 2370 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2371 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 2372 if (!error) { 2373 vp = nd.ni_vp; 2374 memset((caddr_t)fhp, 0, sizeof(nfh)); 2375 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 2376 error = VFS_VPTOFH(vp, &fhp->fh_fid); 2377 if (!error) 2378 error = VOP_GETATTR(vp, &va, cred, lwp); 2379 vput(vp); 2380 } 2381 out: 2382 if (dirp) { 2383 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2384 vrele(dirp); 2385 } 2386 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2387 if (v3) { 2388 if (!error) { 2389 nfsm_srvpostop_fh(fhp); 2390 nfsm_srvpostop_attr(0, &va); 2391 } 2392 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2393 } else { 2394 nfsm_srvfhtom(fhp, v3); 2395 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 2396 nfsm_srvfillattr(&va, fp); 2397 } 2398 vn_finished_write(mp, 0); 2399 return (0); 2400 nfsmout: 2401 if (dirp) 2402 vrele(dirp); 2403 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2404 if (nd.ni_dvp == nd.ni_vp) 2405 vrele(nd.ni_dvp); 2406 else 2407 vput(nd.ni_dvp); 2408 if (nd.ni_vp) 2409 vrele(nd.ni_vp); 2410 vn_finished_write(mp, 0); 2411 return (error); 2412 } 2413 2414 /* 2415 * nfs rmdir service 2416 */ 2417 int 2418 nfsrv_rmdir(nfsd, slp, lwp, mrq) 2419 struct nfsrv_descript *nfsd; 2420 struct nfssvc_sock *slp; 2421 struct lwp *lwp; 2422 struct mbuf **mrq; 2423 { 2424 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2425 struct mbuf *nam = nfsd->nd_nam; 2426 caddr_t dpos = nfsd->nd_dpos; 2427 struct ucred *cred = &nfsd->nd_cr; 2428 u_int32_t *tl; 2429 int32_t t1; 2430 caddr_t bpos; 2431 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 2432 int v3 = (nfsd->nd_flag & ND_NFSV3); 2433 char *cp2; 2434 struct mbuf *mb, *mreq; 2435 struct vnode *vp, *dirp = (struct vnode *)0; 2436 struct vattr dirfor, diraft; 2437 nfsfh_t nfh; 2438 fhandle_t *fhp; 2439 struct nameidata nd; 2440 u_quad_t frev; 2441 struct mount *mp = NULL; 2442 2443 fhp = &nfh.fh_generic; 2444 nfsm_srvmtofh(fhp); 2445 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2446 return (ESTALE); 2447 vn_start_write(NULL, &mp, V_WAIT); 2448 nfsm_srvnamesiz(len); 2449 nd.ni_cnd.cn_cred = cred; 2450 nd.ni_cnd.cn_nameiop = DELETE; 2451 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 2452 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2453 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2454 if (dirp) { 2455 if (v3) 2456 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2457 else { 2458 vrele(dirp); 2459 dirp = (struct vnode *)0; 2460 } 2461 } 2462 if (error) { 2463 nfsm_reply(NFSX_WCCDATA(v3)); 2464 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2465 if (dirp) 2466 vrele(dirp); 2467 vn_finished_write(mp, 0); 2468 return (0); 2469 } 2470 vp = nd.ni_vp; 2471 if (vp->v_type != VDIR) { 2472 error = ENOTDIR; 2473 goto out; 2474 } 2475 /* 2476 * No rmdir "." please. 2477 */ 2478 if (nd.ni_dvp == vp) { 2479 error = EINVAL; 2480 goto out; 2481 } 2482 /* 2483 * The root of a mounted filesystem cannot be deleted. 2484 */ 2485 if (vp->v_flag & VROOT) 2486 error = EBUSY; 2487 out: 2488 if (!error) { 2489 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2490 nqsrv_getl(vp, ND_WRITE); 2491 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2492 } else { 2493 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2494 if (nd.ni_dvp == nd.ni_vp) 2495 vrele(nd.ni_dvp); 2496 else 2497 vput(nd.ni_dvp); 2498 vput(vp); 2499 } 2500 if (dirp) { 2501 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2502 vrele(dirp); 2503 } 2504 nfsm_reply(NFSX_WCCDATA(v3)); 2505 if (v3) { 2506 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2507 vn_finished_write(mp, 0); 2508 return (0); 2509 } 2510 vn_finished_write(mp, 0); 2511 nfsm_srvdone; 2512 } 2513 2514 /* 2515 * nfs readdir service 2516 * - mallocs what it thinks is enough to read 2517 * count rounded up to a multiple of NFS_DIRBLKSIZ <= NFS_MAXREADDIR 2518 * - calls VOP_READDIR() 2519 * - loops around building the reply 2520 * if the output generated exceeds count break out of loop 2521 * The nfsm_clget macro is used here so that the reply will be packed 2522 * tightly in mbuf clusters. 2523 * - it only knows that it has encountered eof when the VOP_READDIR() 2524 * reads nothing 2525 * - as such one readdir rpc will return eof false although you are there 2526 * and then the next will return eof 2527 * - it trims out records with d_fileno == 0 2528 * this doesn't matter for Unix clients, but they might confuse clients 2529 * for other os'. 2530 * - it trims out records with d_type == DT_WHT 2531 * these cannot be seen through NFS (unless we extend the protocol) 2532 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less 2533 * than requested, but this may not apply to all filesystems. For 2534 * example, client NFS does not { although it is never remote mounted 2535 * anyhow } 2536 * The alternate call nfsrv_readdirplus() does lookups as well. 2537 * PS: The NFS protocol spec. does not clarify what the "count" byte 2538 * argument is a count of.. just name strings and file id's or the 2539 * entire reply rpc or ... 2540 * I tried just file name and id sizes and it confused the Sun client, 2541 * so I am using the full rpc size now. The "paranoia.." comment refers 2542 * to including the status longwords that are not a part of the dir. 2543 * "entry" structures, but are in the rpc. 2544 */ 2545 struct flrep { 2546 nfsuint64 fl_off; 2547 u_int32_t fl_postopok; 2548 u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)]; 2549 u_int32_t fl_fhok; 2550 u_int32_t fl_fhsize; 2551 u_int32_t fl_nfh[NFSX_V3FH / sizeof (u_int32_t)]; 2552 }; 2553 2554 int 2555 nfsrv_readdir(nfsd, slp, lwp, mrq) 2556 struct nfsrv_descript *nfsd; 2557 struct nfssvc_sock *slp; 2558 struct lwp *lwp; 2559 struct mbuf **mrq; 2560 { 2561 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2562 struct mbuf *nam = nfsd->nd_nam; 2563 caddr_t dpos = nfsd->nd_dpos; 2564 struct ucred *cred = &nfsd->nd_cr; 2565 char *bp, *be; 2566 struct mbuf *mp; 2567 struct dirent *dp; 2568 caddr_t cp; 2569 u_int32_t *tl; 2570 int32_t t1; 2571 caddr_t bpos; 2572 struct mbuf *mb, *mreq, *mp2; 2573 char *cpos, *cend, *cp2, *rbuf; 2574 struct vnode *vp; 2575 struct vattr at; 2576 nfsfh_t nfh; 2577 fhandle_t *fhp; 2578 struct uio io; 2579 struct iovec iv; 2580 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2581 int siz, cnt, fullsiz, eofflag, rdonly, cache, ncookies; 2582 int v3 = (nfsd->nd_flag & ND_NFSV3); 2583 u_quad_t frev, off, toff, verf; 2584 off_t *cookies = NULL, *cookiep; 2585 nfsuint64 jar; 2586 2587 fhp = &nfh.fh_generic; 2588 nfsm_srvmtofh(fhp); 2589 if (v3) { 2590 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 2591 toff = fxdr_hyper(tl); 2592 tl += 2; 2593 verf = fxdr_hyper(tl); 2594 tl += 2; 2595 } else { 2596 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2597 toff = fxdr_unsigned(u_quad_t, *tl++); 2598 } 2599 off = toff; 2600 cnt = fxdr_unsigned(int, *tl); 2601 siz = ((cnt + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1)); 2602 xfer = NFS_SRVMAXDATA(nfsd); 2603 if (siz > xfer) 2604 siz = xfer; 2605 fullsiz = siz; 2606 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 2607 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2608 if (!error && vp->v_type != VDIR) { 2609 error = ENOTDIR; 2610 vput(vp); 2611 } 2612 if (error) { 2613 nfsm_reply(NFSX_UNSIGNED); 2614 nfsm_srvpostop_attr(getret, &at); 2615 return (0); 2616 } 2617 nqsrv_getl(vp, ND_READ); 2618 if (v3) { 2619 error = getret = VOP_GETATTR(vp, &at, cred, lwp); 2620 #ifdef NFS3_STRICTVERF 2621 /* 2622 * XXX This check is too strict for Solaris 2.5 clients. 2623 */ 2624 if (!error && toff && verf != at.va_filerev) 2625 error = NFSERR_BAD_COOKIE; 2626 #endif 2627 } 2628 if (!error) 2629 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0); 2630 if (error) { 2631 vput(vp); 2632 nfsm_reply(NFSX_POSTOPATTR(v3)); 2633 nfsm_srvpostop_attr(getret, &at); 2634 return (0); 2635 } 2636 VOP_UNLOCK(vp, 0); 2637 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2638 again: 2639 iv.iov_base = rbuf; 2640 iv.iov_len = fullsiz; 2641 io.uio_iov = &iv; 2642 io.uio_iovcnt = 1; 2643 io.uio_offset = (off_t)off; 2644 io.uio_resid = fullsiz; 2645 io.uio_segflg = UIO_SYSSPACE; 2646 io.uio_rw = UIO_READ; 2647 io.uio_lwp = (struct lwp *)0; 2648 eofflag = 0; 2649 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2650 2651 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2652 2653 off = (off_t)io.uio_offset; 2654 if (!cookies && !error) 2655 error = NFSERR_PERM; 2656 if (v3) { 2657 getret = VOP_GETATTR(vp, &at, cred, lwp); 2658 if (!error) 2659 error = getret; 2660 } 2661 2662 VOP_UNLOCK(vp, 0); 2663 if (error) { 2664 vrele(vp); 2665 free((caddr_t)rbuf, M_TEMP); 2666 if (cookies) 2667 free((caddr_t)cookies, M_TEMP); 2668 nfsm_reply(NFSX_POSTOPATTR(v3)); 2669 nfsm_srvpostop_attr(getret, &at); 2670 return (0); 2671 } 2672 if (io.uio_resid) { 2673 siz -= io.uio_resid; 2674 2675 /* 2676 * If nothing read, return eof 2677 * rpc reply 2678 */ 2679 if (siz == 0) { 2680 vrele(vp); 2681 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + 2682 2 * NFSX_UNSIGNED); 2683 if (v3) { 2684 nfsm_srvpostop_attr(getret, &at); 2685 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2686 txdr_hyper(at.va_filerev, tl); 2687 tl += 2; 2688 } else 2689 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2690 *tl++ = nfs_false; 2691 *tl = nfs_true; 2692 free((caddr_t)rbuf, M_TEMP); 2693 free((caddr_t)cookies, M_TEMP); 2694 return (0); 2695 } 2696 } 2697 2698 /* 2699 * Check for degenerate cases of nothing useful read. 2700 * If so go try again 2701 */ 2702 cpos = rbuf; 2703 cend = rbuf + siz; 2704 dp = (struct dirent *)cpos; 2705 cookiep = cookies; 2706 2707 while (cpos < cend && ncookies > 0 && 2708 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2709 cpos += dp->d_reclen; 2710 dp = (struct dirent *)cpos; 2711 cookiep++; 2712 ncookies--; 2713 } 2714 if (cpos >= cend || ncookies == 0) { 2715 toff = off; 2716 siz = fullsiz; 2717 goto again; 2718 } 2719 2720 len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */ 2721 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz); 2722 if (v3) { 2723 nfsm_srvpostop_attr(getret, &at); 2724 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2725 txdr_hyper(at.va_filerev, tl); 2726 } 2727 mp = mp2 = mb; 2728 bp = bpos; 2729 be = bp + M_TRAILINGSPACE(mp); 2730 2731 /* Loop through the records and build reply */ 2732 while (cpos < cend && ncookies > 0) { 2733 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 2734 nlen = dp->d_namlen; 2735 rem = nfsm_rndup(nlen)-nlen; 2736 len += (4 * NFSX_UNSIGNED + nlen + rem); 2737 if (v3) 2738 len += 2 * NFSX_UNSIGNED; 2739 if (len > cnt) { 2740 eofflag = 0; 2741 break; 2742 } 2743 /* 2744 * Build the directory record xdr from 2745 * the dirent entry. 2746 */ 2747 nfsm_clget; 2748 *tl = nfs_true; 2749 bp += NFSX_UNSIGNED; 2750 if (v3) { 2751 nfsm_clget; 2752 *tl = 0; 2753 bp += NFSX_UNSIGNED; 2754 } 2755 nfsm_clget; 2756 *tl = txdr_unsigned(dp->d_fileno); 2757 bp += NFSX_UNSIGNED; 2758 nfsm_clget; 2759 *tl = txdr_unsigned(nlen); 2760 bp += NFSX_UNSIGNED; 2761 2762 /* And loop around copying the name */ 2763 xfer = nlen; 2764 cp = dp->d_name; 2765 while (xfer > 0) { 2766 nfsm_clget; 2767 if ((bp+xfer) > be) 2768 tsiz = be-bp; 2769 else 2770 tsiz = xfer; 2771 memcpy(bp, cp, tsiz); 2772 bp += tsiz; 2773 xfer -= tsiz; 2774 if (xfer > 0) 2775 cp += tsiz; 2776 } 2777 /* And null pad to an int32_t boundary */ 2778 for (i = 0; i < rem; i++) 2779 *bp++ = '\0'; 2780 nfsm_clget; 2781 2782 /* Finish off the record */ 2783 txdr_hyper(*cookiep, &jar); 2784 if (v3) { 2785 *tl = jar.nfsuquad[0]; 2786 bp += NFSX_UNSIGNED; 2787 nfsm_clget; 2788 } 2789 *tl = jar.nfsuquad[1]; 2790 bp += NFSX_UNSIGNED; 2791 } 2792 cpos += dp->d_reclen; 2793 dp = (struct dirent *)cpos; 2794 cookiep++; 2795 ncookies--; 2796 } 2797 vrele(vp); 2798 nfsm_clget; 2799 *tl = nfs_false; 2800 bp += NFSX_UNSIGNED; 2801 nfsm_clget; 2802 if (eofflag) 2803 *tl = nfs_true; 2804 else 2805 *tl = nfs_false; 2806 bp += NFSX_UNSIGNED; 2807 if (mp != mb) { 2808 if (bp < be) 2809 mp->m_len = bp - mtod(mp, caddr_t); 2810 } else 2811 mp->m_len += bp - bpos; 2812 free((caddr_t)rbuf, M_TEMP); 2813 free((caddr_t)cookies, M_TEMP); 2814 nfsm_srvdone; 2815 } 2816 2817 int 2818 nfsrv_readdirplus(nfsd, slp, lwp, mrq) 2819 struct nfsrv_descript *nfsd; 2820 struct nfssvc_sock *slp; 2821 struct lwp *lwp; 2822 struct mbuf **mrq; 2823 { 2824 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2825 struct mbuf *nam = nfsd->nd_nam; 2826 caddr_t dpos = nfsd->nd_dpos; 2827 struct ucred *cred = &nfsd->nd_cr; 2828 char *bp, *be; 2829 struct mbuf *mp; 2830 struct dirent *dp; 2831 caddr_t cp; 2832 u_int32_t *tl; 2833 int32_t t1; 2834 caddr_t bpos; 2835 struct mbuf *mb, *mreq, *mp2; 2836 char *cpos, *cend, *cp2, *rbuf; 2837 struct vnode *vp, *nvp; 2838 struct flrep fl; 2839 nfsfh_t nfh; 2840 fhandle_t *fhp, *nfhp = (fhandle_t *)fl.fl_nfh; 2841 struct uio io; 2842 struct iovec iv; 2843 struct vattr va, at, *vap = &va; 2844 struct nfs_fattr *fp; 2845 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2846 int siz, cnt, fullsiz, eofflag, rdonly, cache, dirlen, ncookies; 2847 u_quad_t frev, off, toff, verf; 2848 off_t *cookies = NULL, *cookiep; 2849 2850 fhp = &nfh.fh_generic; 2851 nfsm_srvmtofh(fhp); 2852 nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED); 2853 toff = fxdr_hyper(tl); 2854 tl += 2; 2855 verf = fxdr_hyper(tl); 2856 tl += 2; 2857 siz = fxdr_unsigned(int, *tl++); 2858 cnt = fxdr_unsigned(int, *tl); 2859 off = toff; 2860 siz = ((siz + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1)); 2861 xfer = NFS_SRVMAXDATA(nfsd); 2862 if (siz > xfer) 2863 siz = xfer; 2864 fullsiz = siz; 2865 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 2866 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2867 if (!error && vp->v_type != VDIR) { 2868 error = ENOTDIR; 2869 vput(vp); 2870 } 2871 if (error) { 2872 nfsm_reply(NFSX_UNSIGNED); 2873 nfsm_srvpostop_attr(getret, &at); 2874 return (0); 2875 } 2876 error = getret = VOP_GETATTR(vp, &at, cred, lwp); 2877 #ifdef NFS3_STRICTVERF 2878 /* 2879 * XXX This check is too strict for Solaris 2.5 clients. 2880 */ 2881 if (!error && toff && verf != at.va_filerev) 2882 error = NFSERR_BAD_COOKIE; 2883 #endif 2884 if (!error) { 2885 nqsrv_getl(vp, ND_READ); 2886 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0); 2887 } 2888 if (error) { 2889 vput(vp); 2890 nfsm_reply(NFSX_V3POSTOPATTR); 2891 nfsm_srvpostop_attr(getret, &at); 2892 return (0); 2893 } 2894 VOP_UNLOCK(vp, 0); 2895 2896 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2897 again: 2898 iv.iov_base = rbuf; 2899 iv.iov_len = fullsiz; 2900 io.uio_iov = &iv; 2901 io.uio_iovcnt = 1; 2902 io.uio_offset = (off_t)off; 2903 io.uio_resid = fullsiz; 2904 io.uio_segflg = UIO_SYSSPACE; 2905 io.uio_rw = UIO_READ; 2906 io.uio_lwp = (struct lwp *)0; 2907 eofflag = 0; 2908 2909 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2910 2911 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2912 2913 off = (u_quad_t)io.uio_offset; 2914 getret = VOP_GETATTR(vp, &at, cred, lwp); 2915 2916 VOP_UNLOCK(vp, 0); 2917 2918 /* 2919 * If the VGET operation doesn't work for this filesystem, 2920 * we can't support readdirplus. Returning NOTSUPP should 2921 * make clients fall back to plain readdir. 2922 * There's no need to check for VPTOFH as well, we wouldn't 2923 * even be here otherwise. 2924 */ 2925 if (!getret) { 2926 if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp))) 2927 getret = (getret == EOPNOTSUPP) ? 2928 NFSERR_NOTSUPP : NFSERR_IO; 2929 else 2930 vput(nvp); 2931 } 2932 2933 if (!cookies && !error) 2934 error = NFSERR_PERM; 2935 if (!error) 2936 error = getret; 2937 if (error) { 2938 vrele(vp); 2939 if (cookies) 2940 free((caddr_t)cookies, M_TEMP); 2941 free((caddr_t)rbuf, M_TEMP); 2942 nfsm_reply(NFSX_V3POSTOPATTR); 2943 nfsm_srvpostop_attr(getret, &at); 2944 return (0); 2945 } 2946 if (io.uio_resid) { 2947 siz -= io.uio_resid; 2948 2949 /* 2950 * If nothing read, return eof 2951 * rpc reply 2952 */ 2953 if (siz == 0) { 2954 vrele(vp); 2955 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2956 2 * NFSX_UNSIGNED); 2957 nfsm_srvpostop_attr(getret, &at); 2958 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2959 txdr_hyper(at.va_filerev, tl); 2960 tl += 2; 2961 *tl++ = nfs_false; 2962 *tl = nfs_true; 2963 free((caddr_t)cookies, M_TEMP); 2964 free((caddr_t)rbuf, M_TEMP); 2965 return (0); 2966 } 2967 } 2968 2969 /* 2970 * Check for degenerate cases of nothing useful read. 2971 * If so go try again 2972 */ 2973 cpos = rbuf; 2974 cend = rbuf + siz; 2975 dp = (struct dirent *)cpos; 2976 cookiep = cookies; 2977 2978 while (cpos < cend && ncookies > 0 && 2979 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2980 cpos += dp->d_reclen; 2981 dp = (struct dirent *)cpos; 2982 cookiep++; 2983 ncookies--; 2984 } 2985 if (cpos >= cend || ncookies == 0) { 2986 toff = off; 2987 siz = fullsiz; 2988 goto again; 2989 } 2990 2991 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED; 2992 nfsm_reply(cnt); 2993 nfsm_srvpostop_attr(getret, &at); 2994 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2995 txdr_hyper(at.va_filerev, tl); 2996 mp = mp2 = mb; 2997 bp = bpos; 2998 be = bp + M_TRAILINGSPACE(mp); 2999 3000 /* Loop through the records and build reply */ 3001 while (cpos < cend && ncookies > 0) { 3002 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 3003 nlen = dp->d_namlen; 3004 rem = nfsm_rndup(nlen)-nlen; 3005 3006 /* 3007 * For readdir_and_lookup get the vnode using 3008 * the file number. 3009 */ 3010 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp)) 3011 goto invalid; 3012 memset((caddr_t)nfhp, 0, NFSX_V3FH); 3013 nfhp->fh_fsid = 3014 nvp->v_mount->mnt_stat.f_fsidx; 3015 if (VFS_VPTOFH(nvp, &nfhp->fh_fid)) { 3016 vput(nvp); 3017 goto invalid; 3018 } 3019 if (VOP_GETATTR(nvp, vap, cred, lwp)) { 3020 vput(nvp); 3021 goto invalid; 3022 } 3023 vput(nvp); 3024 3025 /* 3026 * If either the dircount or maxcount will be 3027 * exceeded, get out now. Both of these lengths 3028 * are calculated conservatively, including all 3029 * XDR overheads. 3030 */ 3031 len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH + 3032 NFSX_V3POSTOPATTR); 3033 dirlen += (6 * NFSX_UNSIGNED + nlen + rem); 3034 if (len > cnt || dirlen > fullsiz) { 3035 eofflag = 0; 3036 break; 3037 } 3038 3039 /* 3040 * Build the directory record xdr from 3041 * the dirent entry. 3042 */ 3043 fp = (struct nfs_fattr *)&fl.fl_fattr; 3044 nfsm_srvfillattr(vap, fp); 3045 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH); 3046 fl.fl_fhok = nfs_true; 3047 fl.fl_postopok = nfs_true; 3048 txdr_hyper(*cookiep, fl.fl_off.nfsuquad); 3049 3050 nfsm_clget; 3051 *tl = nfs_true; 3052 bp += NFSX_UNSIGNED; 3053 nfsm_clget; 3054 *tl = 0; 3055 bp += NFSX_UNSIGNED; 3056 nfsm_clget; 3057 *tl = txdr_unsigned(dp->d_fileno); 3058 bp += NFSX_UNSIGNED; 3059 nfsm_clget; 3060 *tl = txdr_unsigned(nlen); 3061 bp += NFSX_UNSIGNED; 3062 3063 /* And loop around copying the name */ 3064 xfer = nlen; 3065 cp = dp->d_name; 3066 while (xfer > 0) { 3067 nfsm_clget; 3068 if ((bp + xfer) > be) 3069 tsiz = be - bp; 3070 else 3071 tsiz = xfer; 3072 memcpy(bp, cp, tsiz); 3073 bp += tsiz; 3074 xfer -= tsiz; 3075 if (xfer > 0) 3076 cp += tsiz; 3077 } 3078 /* And null pad to an int32_t boundary */ 3079 for (i = 0; i < rem; i++) 3080 *bp++ = '\0'; 3081 3082 /* 3083 * Now copy the flrep structure out. 3084 */ 3085 xfer = sizeof (struct flrep); 3086 cp = (caddr_t)&fl; 3087 while (xfer > 0) { 3088 nfsm_clget; 3089 if ((bp + xfer) > be) 3090 tsiz = be - bp; 3091 else 3092 tsiz = xfer; 3093 memcpy(bp, cp, tsiz); 3094 bp += tsiz; 3095 xfer -= tsiz; 3096 if (xfer > 0) 3097 cp += tsiz; 3098 } 3099 } 3100 invalid: 3101 cpos += dp->d_reclen; 3102 dp = (struct dirent *)cpos; 3103 cookiep++; 3104 ncookies--; 3105 } 3106 vrele(vp); 3107 nfsm_clget; 3108 *tl = nfs_false; 3109 bp += NFSX_UNSIGNED; 3110 nfsm_clget; 3111 if (eofflag) 3112 *tl = nfs_true; 3113 else 3114 *tl = nfs_false; 3115 bp += NFSX_UNSIGNED; 3116 if (mp != mb) { 3117 if (bp < be) 3118 mp->m_len = bp - mtod(mp, caddr_t); 3119 } else 3120 mp->m_len += bp - bpos; 3121 free((caddr_t)cookies, M_TEMP); 3122 free((caddr_t)rbuf, M_TEMP); 3123 nfsm_srvdone; 3124 } 3125 3126 /* 3127 * nfs commit service 3128 */ 3129 int 3130 nfsrv_commit(nfsd, slp, lwp, mrq) 3131 struct nfsrv_descript *nfsd; 3132 struct nfssvc_sock *slp; 3133 struct lwp *lwp; 3134 struct mbuf **mrq; 3135 { 3136 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3137 struct mbuf *nam = nfsd->nd_nam; 3138 caddr_t dpos = nfsd->nd_dpos; 3139 struct ucred *cred = &nfsd->nd_cr; 3140 struct vattr bfor, aft; 3141 struct vnode *vp; 3142 nfsfh_t nfh; 3143 fhandle_t *fhp; 3144 u_int32_t *tl; 3145 int32_t t1; 3146 caddr_t bpos; 3147 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cnt, cache; 3148 char *cp2; 3149 struct mbuf *mb, *mreq; 3150 u_quad_t frev, off, end; 3151 struct mount *mp = NULL; 3152 3153 #ifndef nolint 3154 cache = 0; 3155 #endif 3156 fhp = &nfh.fh_generic; 3157 nfsm_srvmtofh(fhp); 3158 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 3159 return (ESTALE); 3160 vn_start_write(NULL, &mp, V_WAIT); 3161 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED); 3162 3163 off = fxdr_hyper(tl); 3164 tl += 2; 3165 cnt = fxdr_unsigned(int, *tl); 3166 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3167 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3168 if (error) { 3169 nfsm_reply(2 * NFSX_UNSIGNED); 3170 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3171 vn_finished_write(mp, 0); 3172 return (0); 3173 } 3174 for_ret = VOP_GETATTR(vp, &bfor, cred, lwp); 3175 end = (cnt > 0) ? off + cnt : vp->v_size; 3176 if (end < off || end > vp->v_size) 3177 end = vp->v_size; 3178 if (off < vp->v_size) 3179 error = VOP_FSYNC(vp, cred, FSYNC_WAIT, off, end, lwp); 3180 /* else error == 0, from nfsrv_fhtovp() */ 3181 aft_ret = VOP_GETATTR(vp, &aft, cred, lwp); 3182 vput(vp); 3183 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF); 3184 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3185 if (!error) { 3186 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF); 3187 *tl++ = txdr_unsigned(boottime.tv_sec); 3188 *tl = txdr_unsigned(boottime.tv_usec); 3189 } else { 3190 vn_finished_write(mp, 0); 3191 return (0); 3192 } 3193 vn_finished_write(mp, 0); 3194 nfsm_srvdone; 3195 } 3196 3197 /* 3198 * nfs statfs service 3199 */ 3200 int 3201 nfsrv_statfs(nfsd, slp, lwp, mrq) 3202 struct nfsrv_descript *nfsd; 3203 struct nfssvc_sock *slp; 3204 struct lwp *lwp; 3205 struct mbuf **mrq; 3206 { 3207 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3208 struct mbuf *nam = nfsd->nd_nam; 3209 caddr_t dpos = nfsd->nd_dpos; 3210 struct ucred *cred = &nfsd->nd_cr; 3211 struct statvfs *sf; 3212 struct nfs_statfs *sfp; 3213 u_int32_t *tl; 3214 int32_t t1; 3215 caddr_t bpos; 3216 int error = 0, rdonly, cache, getret = 1; 3217 int v3 = (nfsd->nd_flag & ND_NFSV3); 3218 char *cp2; 3219 struct mbuf *mb, *mreq; 3220 struct vnode *vp; 3221 struct vattr at; 3222 nfsfh_t nfh; 3223 fhandle_t *fhp; 3224 struct statvfs statvfs; 3225 u_quad_t frev, tval; 3226 3227 #ifndef nolint 3228 cache = 0; 3229 #endif 3230 fhp = &nfh.fh_generic; 3231 nfsm_srvmtofh(fhp); 3232 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3233 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3234 if (error) { 3235 nfsm_reply(NFSX_UNSIGNED); 3236 nfsm_srvpostop_attr(getret, &at); 3237 return (0); 3238 } 3239 sf = &statvfs; 3240 error = VFS_STATVFS(vp->v_mount, sf, lwp); 3241 getret = VOP_GETATTR(vp, &at, cred, lwp); 3242 vput(vp); 3243 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3)); 3244 if (v3) 3245 nfsm_srvpostop_attr(getret, &at); 3246 if (error) 3247 return (0); 3248 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3)); 3249 if (v3) { 3250 tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_frsize); 3251 txdr_hyper(tval, &sfp->sf_tbytes); 3252 tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_frsize); 3253 txdr_hyper(tval, &sfp->sf_fbytes); 3254 tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_frsize); 3255 txdr_hyper(tval, &sfp->sf_abytes); 3256 tval = (u_quad_t)sf->f_files; 3257 txdr_hyper(tval, &sfp->sf_tfiles); 3258 tval = (u_quad_t)sf->f_ffree; 3259 txdr_hyper(tval, &sfp->sf_ffiles); 3260 txdr_hyper(tval, &sfp->sf_afiles); 3261 sfp->sf_invarsec = 0; 3262 } else { 3263 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA); 3264 sfp->sf_bsize = txdr_unsigned(sf->f_frsize); 3265 sfp->sf_blocks = txdr_unsigned(sf->f_blocks); 3266 sfp->sf_bfree = txdr_unsigned(sf->f_bfree); 3267 sfp->sf_bavail = txdr_unsigned(sf->f_bavail); 3268 } 3269 nfsm_srvdone; 3270 } 3271 3272 /* 3273 * nfs fsinfo service 3274 */ 3275 int 3276 nfsrv_fsinfo(nfsd, slp, lwp, mrq) 3277 struct nfsrv_descript *nfsd; 3278 struct nfssvc_sock *slp; 3279 struct lwp *lwp; 3280 struct mbuf **mrq; 3281 { 3282 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3283 struct mbuf *nam = nfsd->nd_nam; 3284 caddr_t dpos = nfsd->nd_dpos; 3285 struct ucred *cred = &nfsd->nd_cr; 3286 u_int32_t *tl; 3287 struct nfsv3_fsinfo *sip; 3288 int32_t t1; 3289 caddr_t bpos; 3290 int error = 0, rdonly, cache, getret = 1; 3291 uint32_t maxdata; 3292 char *cp2; 3293 struct mbuf *mb, *mreq; 3294 struct vnode *vp; 3295 struct vattr at; 3296 nfsfh_t nfh; 3297 fhandle_t *fhp; 3298 u_quad_t frev, maxfsize; 3299 struct statvfs sb; 3300 3301 #ifndef nolint 3302 cache = 0; 3303 #endif 3304 fhp = &nfh.fh_generic; 3305 nfsm_srvmtofh(fhp); 3306 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3307 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3308 if (error) { 3309 nfsm_reply(NFSX_UNSIGNED); 3310 nfsm_srvpostop_attr(getret, &at); 3311 return (0); 3312 } 3313 3314 /* XXX Try to make a guess on the max file size. */ 3315 VFS_STATVFS(vp->v_mount, &sb, (struct lwp *)0); 3316 maxfsize = (u_quad_t)0x80000000 * sb.f_frsize - 1; 3317 3318 getret = VOP_GETATTR(vp, &at, cred, lwp); 3319 vput(vp); 3320 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO); 3321 nfsm_srvpostop_attr(getret, &at); 3322 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO); 3323 3324 /* 3325 * XXX 3326 * There should be file system VFS OP(s) to get this information. 3327 * For now, assume ufs. 3328 */ 3329 if (slp->ns_so->so_type == SOCK_DGRAM) 3330 maxdata = NFS_MAXDGRAMDATA; 3331 else 3332 maxdata = NFS_MAXDATA; 3333 sip->fs_rtmax = txdr_unsigned(maxdata); 3334 sip->fs_rtpref = txdr_unsigned(maxdata); 3335 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE); 3336 sip->fs_wtmax = txdr_unsigned(maxdata); 3337 sip->fs_wtpref = txdr_unsigned(maxdata); 3338 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE); 3339 sip->fs_dtpref = txdr_unsigned(maxdata); 3340 txdr_hyper(maxfsize, &sip->fs_maxfilesize); 3341 sip->fs_timedelta.nfsv3_sec = 0; 3342 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1); 3343 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK | 3344 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS | 3345 NFSV3FSINFO_CANSETTIME); 3346 nfsm_srvdone; 3347 } 3348 3349 /* 3350 * nfs pathconf service 3351 */ 3352 int 3353 nfsrv_pathconf(nfsd, slp, lwp, mrq) 3354 struct nfsrv_descript *nfsd; 3355 struct nfssvc_sock *slp; 3356 struct lwp *lwp; 3357 struct mbuf **mrq; 3358 { 3359 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3360 struct mbuf *nam = nfsd->nd_nam; 3361 caddr_t dpos = nfsd->nd_dpos; 3362 struct ucred *cred = &nfsd->nd_cr; 3363 u_int32_t *tl; 3364 struct nfsv3_pathconf *pc; 3365 int32_t t1; 3366 caddr_t bpos; 3367 int error = 0, rdonly, cache, getret = 1; 3368 register_t linkmax, namemax, chownres, notrunc; 3369 char *cp2; 3370 struct mbuf *mb, *mreq; 3371 struct vnode *vp; 3372 struct vattr at; 3373 nfsfh_t nfh; 3374 fhandle_t *fhp; 3375 u_quad_t frev; 3376 3377 #ifndef nolint 3378 cache = 0; 3379 #endif 3380 fhp = &nfh.fh_generic; 3381 nfsm_srvmtofh(fhp); 3382 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3383 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3384 if (error) { 3385 nfsm_reply(NFSX_UNSIGNED); 3386 nfsm_srvpostop_attr(getret, &at); 3387 return (0); 3388 } 3389 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax); 3390 if (!error) 3391 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax); 3392 if (!error) 3393 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres); 3394 if (!error) 3395 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc); 3396 getret = VOP_GETATTR(vp, &at, cred, lwp); 3397 vput(vp); 3398 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF); 3399 nfsm_srvpostop_attr(getret, &at); 3400 if (error) 3401 return (0); 3402 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF); 3403 3404 pc->pc_linkmax = txdr_unsigned(linkmax); 3405 pc->pc_namemax = txdr_unsigned(namemax); 3406 pc->pc_notrunc = txdr_unsigned(notrunc); 3407 pc->pc_chownrestricted = txdr_unsigned(chownres); 3408 3409 /* 3410 * These should probably be supported by VOP_PATHCONF(), but 3411 * until msdosfs is exportable (why would you want to?), the 3412 * Unix defaults should be ok. 3413 */ 3414 pc->pc_caseinsensitive = nfs_false; 3415 pc->pc_casepreserving = nfs_true; 3416 nfsm_srvdone; 3417 } 3418 3419 /* 3420 * Null operation, used by clients to ping server 3421 */ 3422 /* ARGSUSED */ 3423 int 3424 nfsrv_null(nfsd, slp, lwp, mrq) 3425 struct nfsrv_descript *nfsd; 3426 struct nfssvc_sock *slp; 3427 struct lwp *lwp; 3428 struct mbuf **mrq; 3429 { 3430 struct mbuf *mrep = nfsd->nd_mrep; 3431 caddr_t bpos; 3432 int error = NFSERR_RETVOID, cache = 0; 3433 struct mbuf *mb, *mreq; 3434 u_quad_t frev; 3435 3436 nfsm_reply(0); 3437 return (0); 3438 } 3439 3440 /* 3441 * No operation, used for obsolete procedures 3442 */ 3443 /* ARGSUSED */ 3444 int 3445 nfsrv_noop(nfsd, slp, lwp, mrq) 3446 struct nfsrv_descript *nfsd; 3447 struct nfssvc_sock *slp; 3448 struct lwp *lwp; 3449 struct mbuf **mrq; 3450 { 3451 struct mbuf *mrep = nfsd->nd_mrep; 3452 caddr_t bpos; 3453 int error, cache = 0; 3454 struct mbuf *mb, *mreq; 3455 u_quad_t frev; 3456 3457 if (nfsd->nd_repstat) 3458 error = nfsd->nd_repstat; 3459 else 3460 error = EPROCUNAVAIL; 3461 nfsm_reply(0); 3462 return (0); 3463 } 3464 3465 /* 3466 * Perform access checking for vnodes obtained from file handles that would 3467 * refer to files already opened by a Unix client. You cannot just use 3468 * vn_writechk() and VOP_ACCESS() for two reasons. 3469 * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case 3470 * 2 - The owner is to be given access irrespective of mode bits for some 3471 * operations, so that processes that chmod after opening a file don't 3472 * break. I don't like this because it opens a security hole, but since 3473 * the nfs server opens a security hole the size of a barn door anyhow, 3474 * what the heck. 3475 * 3476 * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS() 3477 * will return EPERM instead of EACCESS. EPERM is always an error. 3478 */ 3479 int 3480 nfsrv_access(vp, flags, cred, rdonly, lwp, override) 3481 struct vnode *vp; 3482 int flags; 3483 struct ucred *cred; 3484 int rdonly; 3485 struct lwp *lwp; 3486 { 3487 struct vattr vattr; 3488 int error; 3489 if (flags & VWRITE) { 3490 /* Just vn_writechk() changed to check rdonly */ 3491 /* 3492 * Disallow write attempts on read-only file systems; 3493 * unless the file is a socket or a block or character 3494 * device resident on the file system. 3495 */ 3496 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 3497 switch (vp->v_type) { 3498 case VREG: 3499 case VDIR: 3500 case VLNK: 3501 return (EROFS); 3502 default: 3503 break; 3504 } 3505 } 3506 3507 /* 3508 * If the vnode is in use as a process's text, 3509 * we can't allow writing. 3510 */ 3511 if (vp->v_flag & VTEXT) 3512 return (ETXTBSY); 3513 } 3514 error = VOP_GETATTR(vp, &vattr, cred, lwp); 3515 if (error) 3516 return (error); 3517 error = VOP_ACCESS(vp, flags, cred, lwp); 3518 /* 3519 * Allow certain operations for the owner (reads and writes 3520 * on files that are already open). 3521 */ 3522 if (override && error == EACCES && cred->cr_uid == vattr.va_uid) 3523 error = 0; 3524 return error; 3525 } 3526
Indexes created Thu Oct 02 14:10:14 GMT 2025