nfs_serv.c revision 1.79.2.7
1 /* $NetBSD: nfs_serv.c,v 1.79.2.7 2005/03/04 16:54:20 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 1.79.2.7 2005/03/04 16:54:20 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 = NULL; 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 if (error == EBUSY) 691 goto loan_fail; 692 goto read_error; 693 } 694 695 /* associate kva to mbuf */ 696 MEXTADD(m, (void *)(lva + ((vaddr_t)off & PAGE_MASK)), 697 cnt, M_MBUF, soloanfree, slp->ns_so); 698 m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP; 699 m->m_len = cnt; 700 701 /* map pages */ 702 for (i = 0; i < npages; i++) { 703 pmap_kenter_pa(lva, VM_PAGE_TO_PHYS(pgpp[i]), 704 VM_PROT_READ); 705 lva += PAGE_SIZE; 706 } 707 708 pmap_update(pmap_kernel()); 709 710 mb->m_next = m; 711 mb = m; 712 error = 0; 713 uiop->uio_resid = 0; 714 } else { 715 struct iovec *iv; 716 struct iovec *iv2; 717 struct mbuf *m2; 718 int siz; 719 loan_fail: 720 /* 721 * Generate the mbuf list with the uio_iov ref. to it. 722 */ 723 i = 0; 724 m = m2 = mb; 725 while (left > 0) { 726 siz = min(M_TRAILINGSPACE(m), left); 727 if (siz > 0) { 728 left -= siz; 729 i++; 730 } 731 if (left > 0) { 732 m = m_get(M_WAIT, MT_DATA); 733 MCLAIM(m, &nfs_mowner); 734 m_clget(m, M_WAIT); 735 m->m_len = 0; 736 m2->m_next = m; 737 m2 = m; 738 } 739 } 740 iv = malloc(i * sizeof(struct iovec), M_TEMP, M_WAITOK); 741 uiop->uio_iov = iv2 = iv; 742 m = mb; 743 left = cnt; 744 i = 0; 745 while (left > 0) { 746 if (m == NULL) 747 panic("nfsrv_read iov"); 748 siz = min(M_TRAILINGSPACE(m), left); 749 if (siz > 0) { 750 iv->iov_base = mtod(m, caddr_t) + 751 m->m_len; 752 iv->iov_len = siz; 753 m->m_len += siz; 754 left -= siz; 755 iv++; 756 i++; 757 } 758 m = m->m_next; 759 } 760 uiop->uio_iovcnt = i; 761 uiop->uio_offset = off; 762 uiop->uio_resid = cnt; 763 uiop->uio_rw = UIO_READ; 764 uiop->uio_segflg = UIO_SYSSPACE; 765 error = VOP_READ(vp, uiop, IO_NODELOCKED, cred); 766 free((caddr_t)iv2, M_TEMP); 767 } 768 read_error: 769 if (error || (getret = VOP_GETATTR(vp, &va, cred, lwp)) != 0){ 770 if (!error) 771 error = getret; 772 m_freem(mreq); 773 vput(vp); 774 nfsm_reply(NFSX_POSTOPATTR(v3)); 775 nfsm_srvpostop_attr(getret, &va); 776 return (0); 777 } 778 } else { 779 uiop->uio_resid = 0; 780 } 781 vput(vp); 782 nfsm_srvfillattr(&va, fp); 783 len -= uiop->uio_resid; 784 padlen = nfsm_padlen(len); 785 if (uiop->uio_resid || padlen) 786 nfs_zeropad(mb, uiop->uio_resid, padlen); 787 if (v3) { 788 /* count */ 789 *tl++ = txdr_unsigned(len); 790 /* eof */ 791 if (off + len >= va.va_size) 792 *tl++ = nfs_true; 793 else 794 *tl++ = nfs_false; 795 } 796 *tl = txdr_unsigned(len); 797 nfsm_srvdone; 798 } 799 800 /* 801 * nfs write service 802 */ 803 int 804 nfsrv_write(nfsd, slp, lwp, mrq) 805 struct nfsrv_descript *nfsd; 806 struct nfssvc_sock *slp; 807 struct lwp *lwp; 808 struct mbuf **mrq; 809 { 810 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 811 struct mbuf *nam = nfsd->nd_nam; 812 caddr_t dpos = nfsd->nd_dpos; 813 struct ucred *cred = &nfsd->nd_cr; 814 struct iovec *ivp; 815 int i, cnt; 816 struct mbuf *mp; 817 struct nfs_fattr *fp; 818 struct iovec *iv; 819 struct vattr va, forat; 820 u_int32_t *tl; 821 int32_t t1; 822 caddr_t bpos; 823 int error = 0, rdonly, cache, len, forat_ret = 1; 824 int ioflags, aftat_ret = 1, retlen, zeroing, adjust; 825 int stable = NFSV3WRITE_FILESYNC; 826 int v3 = (nfsd->nd_flag & ND_NFSV3); 827 char *cp2; 828 struct mbuf *mb, *mreq; 829 struct vnode *vp; 830 nfsfh_t nfh; 831 fhandle_t *fhp; 832 struct uio io, *uiop = &io; 833 off_t off; 834 u_quad_t frev; 835 struct mount *mntp = NULL; 836 837 if (mrep == NULL) { 838 *mrq = NULL; 839 return (0); 840 } 841 fhp = &nfh.fh_generic; 842 nfsm_srvmtofh(fhp); 843 if ((mntp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 844 return (ESTALE); 845 vn_start_write(NULL, &mntp, V_WAIT); 846 if (v3) { 847 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 848 off = fxdr_hyper(tl); 849 tl += 3; 850 stable = fxdr_unsigned(int, *tl++); 851 } else { 852 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 853 off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 854 tl += 2; 855 } 856 retlen = len = fxdr_unsigned(int32_t, *tl); 857 cnt = i = 0; 858 859 /* 860 * For NFS Version 2, it is not obvious what a write of zero length 861 * should do, but I might as well be consistent with Version 3, 862 * which is to return ok so long as there are no permission problems. 863 */ 864 if (len > 0) { 865 zeroing = 1; 866 mp = mrep; 867 while (mp) { 868 if (mp == md) { 869 zeroing = 0; 870 adjust = dpos - mtod(mp, caddr_t); 871 mp->m_len -= adjust; 872 if (mp->m_len > 0 && adjust > 0) 873 NFSMADV(mp, adjust); 874 } 875 if (zeroing) 876 mp->m_len = 0; 877 else if (mp->m_len > 0) { 878 i += mp->m_len; 879 if (i > len) { 880 mp->m_len -= (i - len); 881 zeroing = 1; 882 } 883 if (mp->m_len > 0) 884 cnt++; 885 } 886 mp = mp->m_next; 887 } 888 } 889 if (len > NFS_MAXDATA || len < 0 || i < len) { 890 error = EIO; 891 nfsm_reply(2 * NFSX_UNSIGNED); 892 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 893 vn_finished_write(mntp, 0); 894 return (0); 895 } 896 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 897 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 898 if (error) { 899 nfsm_reply(2 * NFSX_UNSIGNED); 900 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 901 vn_finished_write(mntp, 0); 902 return (0); 903 } 904 if (v3) 905 forat_ret = VOP_GETATTR(vp, &forat, cred, lwp); 906 if (vp->v_type != VREG) { 907 if (v3) 908 error = EINVAL; 909 else 910 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 911 } 912 if (!error) { 913 nqsrv_getl(vp, ND_WRITE); 914 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1); 915 } 916 if (error) { 917 vput(vp); 918 nfsm_reply(NFSX_WCCDATA(v3)); 919 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 920 vn_finished_write(mntp, 0); 921 return (0); 922 } 923 924 if (len > 0) { 925 ivp = malloc(cnt * sizeof (struct iovec), M_TEMP, M_WAITOK); 926 uiop->uio_iov = iv = ivp; 927 uiop->uio_iovcnt = cnt; 928 mp = mrep; 929 while (mp) { 930 if (mp->m_len > 0) { 931 ivp->iov_base = mtod(mp, caddr_t); 932 ivp->iov_len = mp->m_len; 933 ivp++; 934 } 935 mp = mp->m_next; 936 } 937 938 /* 939 * XXX 940 * The IO_METASYNC flag indicates that all metadata (and not 941 * just enough to ensure data integrity) must be written to 942 * stable storage synchronously. 943 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.) 944 */ 945 if (stable == NFSV3WRITE_UNSTABLE) 946 ioflags = IO_NODELOCKED; 947 else if (stable == NFSV3WRITE_DATASYNC) 948 ioflags = (IO_SYNC | IO_NODELOCKED); 949 else 950 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 951 uiop->uio_resid = len; 952 uiop->uio_rw = UIO_WRITE; 953 uiop->uio_segflg = UIO_SYSSPACE; 954 uiop->uio_lwp = NULL; 955 uiop->uio_offset = off; 956 error = VOP_WRITE(vp, uiop, ioflags, cred); 957 nfsstats.srvvop_writes++; 958 free(iv, M_TEMP); 959 } 960 aftat_ret = VOP_GETATTR(vp, &va, cred, lwp); 961 vput(vp); 962 vn_finished_write(mntp, 0); 963 if (!error) 964 error = aftat_ret; 965 nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) + 966 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3)); 967 if (v3) { 968 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 969 if (error) 970 return (0); 971 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 972 *tl++ = txdr_unsigned(retlen); 973 if (stable == NFSV3WRITE_UNSTABLE) 974 *tl++ = txdr_unsigned(stable); 975 else 976 *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC); 977 /* 978 * Actually, there is no need to txdr these fields, 979 * but it may make the values more human readable, 980 * for debugging purposes. 981 */ 982 *tl++ = txdr_unsigned(boottime.tv_sec); 983 *tl = txdr_unsigned(boottime.tv_usec); 984 } else { 985 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 986 nfsm_srvfillattr(&va, fp); 987 } 988 nfsm_srvdone; 989 } 990 991 /* 992 * NFS write service with write gathering support. Called when 993 * nfsrvw_procrastinate > 0. 994 * See: Chet Juszczak, "Improving the Write Performance of an NFS Server", 995 * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco, 996 * Jan. 1994. 997 */ 998 int 999 nfsrv_writegather(ndp, slp, lwp, mrq) 1000 struct nfsrv_descript **ndp; 1001 struct nfssvc_sock *slp; 1002 struct lwp *lwp; 1003 struct mbuf **mrq; 1004 { 1005 struct iovec *ivp; 1006 struct mbuf *mp; 1007 struct nfsrv_descript *wp, *nfsd, *owp, *swp; 1008 struct nfs_fattr *fp; 1009 int i = 0; 1010 struct iovec *iov; 1011 struct nfsrvw_delayhash *wpp; 1012 struct ucred *cred; 1013 struct vattr va, forat; 1014 u_int32_t *tl; 1015 int32_t t1; 1016 caddr_t bpos, dpos; 1017 int error = 0, rdonly, cache, len = 0, forat_ret = 1; 1018 int ioflags, aftat_ret = 1, s, adjust, v3, zeroing; 1019 char *cp2; 1020 struct mbuf *mb, *mreq, *mrep, *md; 1021 struct vnode *vp; 1022 struct uio io, *uiop = &io; 1023 u_quad_t frev, cur_usec; 1024 struct mount *mntp = NULL; 1025 1026 *mrq = NULL; 1027 if (*ndp) { 1028 nfsd = *ndp; 1029 *ndp = NULL; 1030 mrep = nfsd->nd_mrep; 1031 md = nfsd->nd_md; 1032 dpos = nfsd->nd_dpos; 1033 cred = &nfsd->nd_cr; 1034 v3 = (nfsd->nd_flag & ND_NFSV3); 1035 LIST_INIT(&nfsd->nd_coalesce); 1036 nfsd->nd_mreq = NULL; 1037 nfsd->nd_stable = NFSV3WRITE_FILESYNC; 1038 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec; 1039 nfsd->nd_time = cur_usec + nfsrvw_procrastinate; 1040 1041 /* 1042 * Now, get the write header.. 1043 */ 1044 nfsm_srvmtofh(&nfsd->nd_fh); 1045 if (v3) { 1046 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 1047 nfsd->nd_off = fxdr_hyper(tl); 1048 tl += 3; 1049 nfsd->nd_stable = fxdr_unsigned(int, *tl++); 1050 } else { 1051 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1052 nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 1053 tl += 2; 1054 } 1055 len = fxdr_unsigned(int32_t, *tl); 1056 nfsd->nd_len = len; 1057 nfsd->nd_eoff = nfsd->nd_off + len; 1058 1059 /* 1060 * Trim the header out of the mbuf list and trim off any trailing 1061 * junk so that the mbuf list has only the write data. 1062 */ 1063 zeroing = 1; 1064 i = 0; 1065 mp = mrep; 1066 while (mp) { 1067 if (mp == md) { 1068 zeroing = 0; 1069 adjust = dpos - mtod(mp, caddr_t); 1070 mp->m_len -= adjust; 1071 if (mp->m_len > 0 && adjust > 0) 1072 NFSMADV(mp, adjust); 1073 } 1074 if (zeroing) 1075 mp->m_len = 0; 1076 else { 1077 i += mp->m_len; 1078 if (i > len) { 1079 mp->m_len -= (i - len); 1080 zeroing = 1; 1081 } 1082 } 1083 mp = mp->m_next; 1084 } 1085 if (len > NFS_MAXDATA || len < 0 || i < len) { 1086 nfsmout: 1087 m_freem(mrep); 1088 error = EIO; 1089 nfsm_writereply(2 * NFSX_UNSIGNED, v3); 1090 if (v3) 1091 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1092 nfsd->nd_mreq = mreq; 1093 nfsd->nd_mrep = NULL; 1094 nfsd->nd_time = 0; 1095 } 1096 1097 /* 1098 * Add this entry to the hash and time queues. 1099 */ 1100 s = splsoftclock(); 1101 owp = NULL; 1102 wp = LIST_FIRST(&slp->ns_tq); 1103 while (wp && wp->nd_time < nfsd->nd_time) { 1104 owp = wp; 1105 wp = LIST_NEXT(wp, nd_tq); 1106 } 1107 if (owp) { 1108 LIST_INSERT_AFTER(owp, nfsd, nd_tq); 1109 } else { 1110 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1111 } 1112 if (nfsd->nd_mrep) { 1113 wpp = NWDELAYHASH(slp, nfsd->nd_fh.fh_fid.fid_data); 1114 owp = NULL; 1115 wp = LIST_FIRST(wpp); 1116 while (wp && 1117 memcmp((caddr_t)&nfsd->nd_fh, (caddr_t)&wp->nd_fh, NFSX_V3FH)) { 1118 owp = wp; 1119 wp = LIST_NEXT(wp, nd_hash); 1120 } 1121 while (wp && wp->nd_off < nfsd->nd_off && 1122 !memcmp((caddr_t)&nfsd->nd_fh, (caddr_t)&wp->nd_fh, NFSX_V3FH)) { 1123 owp = wp; 1124 wp = LIST_NEXT(wp, nd_hash); 1125 } 1126 if (owp) { 1127 LIST_INSERT_AFTER(owp, nfsd, nd_hash); 1128 1129 /* 1130 * Search the hash list for overlapping entries and 1131 * coalesce. 1132 */ 1133 for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) { 1134 wp = LIST_NEXT(nfsd, nd_hash); 1135 if (NFSW_SAMECRED(owp, nfsd)) 1136 nfsrvw_coalesce(owp, nfsd); 1137 } 1138 } else { 1139 LIST_INSERT_HEAD(wpp, nfsd, nd_hash); 1140 } 1141 } 1142 splx(s); 1143 } 1144 1145 /* 1146 * Now, do VOP_WRITE()s for any one(s) that need to be done now 1147 * and generate the associated reply mbuf list(s). 1148 */ 1149 loop1: 1150 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec; 1151 s = splsoftclock(); 1152 for (nfsd = LIST_FIRST(&slp->ns_tq); nfsd; nfsd = owp) { 1153 owp = LIST_NEXT(nfsd, nd_tq); 1154 if (nfsd->nd_time > cur_usec) 1155 break; 1156 if (nfsd->nd_mreq) 1157 continue; 1158 LIST_REMOVE(nfsd, nd_tq); 1159 LIST_REMOVE(nfsd, nd_hash); 1160 splx(s); 1161 mrep = nfsd->nd_mrep; 1162 nfsd->nd_mrep = NULL; 1163 cred = &nfsd->nd_cr; 1164 v3 = (nfsd->nd_flag & ND_NFSV3); 1165 forat_ret = aftat_ret = 1; 1166 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp, 1167 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH), 1168 FALSE); 1169 if (!error) { 1170 if (v3) 1171 forat_ret = VOP_GETATTR(vp, &forat, cred, lwp); 1172 if (vp->v_type != VREG) { 1173 if (v3) 1174 error = EINVAL; 1175 else 1176 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 1177 } 1178 } else 1179 vp = NULL; 1180 if (!error) { 1181 nqsrv_getl(vp, ND_WRITE); 1182 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1); 1183 } 1184 1185 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE) 1186 ioflags = IO_NODELOCKED; 1187 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC) 1188 ioflags = (IO_SYNC | IO_NODELOCKED); 1189 else 1190 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 1191 uiop->uio_rw = UIO_WRITE; 1192 uiop->uio_segflg = UIO_SYSSPACE; 1193 uiop->uio_lwp = NULL; 1194 uiop->uio_offset = nfsd->nd_off; 1195 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off; 1196 if (uiop->uio_resid > 0) { 1197 mp = mrep; 1198 i = 0; 1199 while (mp) { 1200 if (mp->m_len > 0) 1201 i++; 1202 mp = mp->m_next; 1203 } 1204 uiop->uio_iovcnt = i; 1205 iov = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK); 1206 uiop->uio_iov = ivp = iov; 1207 mp = mrep; 1208 while (mp) { 1209 if (mp->m_len > 0) { 1210 ivp->iov_base = mtod(mp, caddr_t); 1211 ivp->iov_len = mp->m_len; 1212 ivp++; 1213 } 1214 mp = mp->m_next; 1215 } 1216 if (!error) { 1217 if (vn_start_write(vp, &mntp, V_NOWAIT) != 0) { 1218 VOP_UNLOCK(vp, 0); 1219 vn_start_write(NULL, &mntp, V_WAIT); 1220 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1221 } 1222 if (!error) { 1223 error = VOP_WRITE(vp, uiop, ioflags, cred); 1224 nfsstats.srvvop_writes++; 1225 vn_finished_write(mntp, 0); 1226 } 1227 } 1228 free((caddr_t)iov, M_TEMP); 1229 } 1230 m_freem(mrep); 1231 if (vp) { 1232 aftat_ret = VOP_GETATTR(vp, &va, cred, lwp); 1233 vput(vp); 1234 } 1235 1236 /* 1237 * Loop around generating replies for all write rpcs that have 1238 * now been completed. 1239 */ 1240 swp = nfsd; 1241 do { 1242 if (error) { 1243 nfsm_writereply(NFSX_WCCDATA(v3), v3); 1244 if (v3) { 1245 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1246 } 1247 } else { 1248 nfsm_writereply(NFSX_PREOPATTR(v3) + 1249 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED + 1250 NFSX_WRITEVERF(v3), v3); 1251 if (v3) { 1252 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1253 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1254 *tl++ = txdr_unsigned(nfsd->nd_len); 1255 *tl++ = txdr_unsigned(swp->nd_stable); 1256 /* 1257 * Actually, there is no need to txdr these fields, 1258 * but it may make the values more human readable, 1259 * for debugging purposes. 1260 */ 1261 *tl++ = txdr_unsigned(boottime.tv_sec); 1262 *tl = txdr_unsigned(boottime.tv_usec); 1263 } else { 1264 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1265 nfsm_srvfillattr(&va, fp); 1266 } 1267 } 1268 nfsd->nd_mreq = mreq; 1269 if (nfsd->nd_mrep) 1270 panic("nfsrv_write: nd_mrep not free"); 1271 1272 /* 1273 * Done. Put it at the head of the timer queue so that 1274 * the final phase can return the reply. 1275 */ 1276 s = splsoftclock(); 1277 if (nfsd != swp) { 1278 nfsd->nd_time = 0; 1279 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1280 } 1281 nfsd = LIST_FIRST(&swp->nd_coalesce); 1282 if (nfsd) { 1283 LIST_REMOVE(nfsd, nd_tq); 1284 } 1285 splx(s); 1286 } while (nfsd); 1287 s = splsoftclock(); 1288 swp->nd_time = 0; 1289 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq); 1290 splx(s); 1291 goto loop1; 1292 } 1293 splx(s); 1294 1295 /* 1296 * Search for a reply to return. 1297 */ 1298 s = splsoftclock(); 1299 LIST_FOREACH(nfsd, &slp->ns_tq, nd_tq) { 1300 if (nfsd->nd_mreq) { 1301 LIST_REMOVE(nfsd, nd_tq); 1302 *mrq = nfsd->nd_mreq; 1303 *ndp = nfsd; 1304 break; 1305 } 1306 } 1307 splx(s); 1308 return (0); 1309 } 1310 1311 /* 1312 * Coalesce the write request nfsd into owp. To do this we must: 1313 * - remove nfsd from the queues 1314 * - merge nfsd->nd_mrep into owp->nd_mrep 1315 * - update the nd_eoff and nd_stable for owp 1316 * - put nfsd on owp's nd_coalesce list 1317 * NB: Must be called at splsoftclock(). 1318 */ 1319 void 1320 nfsrvw_coalesce(owp, nfsd) 1321 struct nfsrv_descript *owp; 1322 struct nfsrv_descript *nfsd; 1323 { 1324 int overlap; 1325 struct mbuf *mp; 1326 struct nfsrv_descript *m; 1327 1328 LIST_REMOVE(nfsd, nd_hash); 1329 LIST_REMOVE(nfsd, nd_tq); 1330 if (owp->nd_eoff < nfsd->nd_eoff) { 1331 overlap = owp->nd_eoff - nfsd->nd_off; 1332 if (overlap < 0) 1333 panic("nfsrv_coalesce: bad off"); 1334 if (overlap > 0) 1335 m_adj(nfsd->nd_mrep, overlap); 1336 mp = owp->nd_mrep; 1337 while (mp->m_next) 1338 mp = mp->m_next; 1339 mp->m_next = nfsd->nd_mrep; 1340 owp->nd_eoff = nfsd->nd_eoff; 1341 } else 1342 m_freem(nfsd->nd_mrep); 1343 nfsd->nd_mrep = NULL; 1344 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC) 1345 owp->nd_stable = NFSV3WRITE_FILESYNC; 1346 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC && 1347 owp->nd_stable == NFSV3WRITE_UNSTABLE) 1348 owp->nd_stable = NFSV3WRITE_DATASYNC; 1349 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq); 1350 /* 1351 * nfsd might hold coalesce elements! Move them to owp. 1352 * Otherwise, requests may be lost and clients will be stuck. 1353 */ 1354 while ((m = LIST_FIRST(&nfsd->nd_coalesce)) != NULL) { 1355 LIST_REMOVE(m, nd_tq); 1356 LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq); 1357 } 1358 } 1359 1360 /* 1361 * nfs create service 1362 * now does a truncate to 0 length via. setattr if it already exists 1363 */ 1364 int 1365 nfsrv_create(nfsd, slp, lwp, mrq) 1366 struct nfsrv_descript *nfsd; 1367 struct nfssvc_sock *slp; 1368 struct lwp *lwp; 1369 struct mbuf **mrq; 1370 { 1371 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1372 struct mbuf *nam = nfsd->nd_nam; 1373 caddr_t dpos = nfsd->nd_dpos; 1374 struct ucred *cred = &nfsd->nd_cr; 1375 struct nfs_fattr *fp; 1376 struct vattr va, dirfor, diraft; 1377 struct nfsv2_sattr *sp; 1378 u_int32_t *tl; 1379 struct nameidata nd; 1380 caddr_t cp; 1381 int32_t t1; 1382 caddr_t bpos; 1383 int error = 0, cache, len, tsize, dirfor_ret = 1, diraft_ret = 1; 1384 int rdev = 0; 1385 int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0; 1386 char *cp2; 1387 struct mbuf *mb, *mreq; 1388 struct vnode *vp = NULL, *dirp = NULL; 1389 nfsfh_t nfh; 1390 fhandle_t *fhp; 1391 u_quad_t frev, tempsize; 1392 u_char cverf[NFSX_V3CREATEVERF]; 1393 struct mount *mp = NULL; 1394 1395 nd.ni_cnd.cn_nameiop = 0; 1396 fhp = &nfh.fh_generic; 1397 nfsm_srvmtofh(fhp); 1398 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1399 return (ESTALE); 1400 vn_start_write(NULL, &mp, V_WAIT); 1401 nfsm_srvnamesiz(len); 1402 nd.ni_cnd.cn_cred = cred; 1403 nd.ni_cnd.cn_nameiop = CREATE; 1404 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1405 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1406 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1407 if (dirp) { 1408 if (v3) 1409 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1410 else { 1411 vrele(dirp); 1412 dirp = (struct vnode *)0; 1413 } 1414 } 1415 if (error) { 1416 nfsm_reply(NFSX_WCCDATA(v3)); 1417 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1418 if (dirp) 1419 vrele(dirp); 1420 vn_finished_write(mp, 0); 1421 return (0); 1422 } 1423 VATTR_NULL(&va); 1424 if (v3) { 1425 va.va_mode = 0; 1426 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1427 how = fxdr_unsigned(int, *tl); 1428 switch (how) { 1429 case NFSV3CREATE_GUARDED: 1430 if (nd.ni_vp) { 1431 error = EEXIST; 1432 break; 1433 } 1434 case NFSV3CREATE_UNCHECKED: 1435 nfsm_srvsattr(&va); 1436 break; 1437 case NFSV3CREATE_EXCLUSIVE: 1438 nfsm_dissect(cp, caddr_t, NFSX_V3CREATEVERF); 1439 memcpy(cverf, cp, NFSX_V3CREATEVERF); 1440 exclusive_flag = 1; 1441 break; 1442 }; 1443 va.va_type = VREG; 1444 } else { 1445 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 1446 va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode)); 1447 if (va.va_type == VNON) 1448 va.va_type = VREG; 1449 va.va_mode = nfstov_mode(sp->sa_mode); 1450 switch (va.va_type) { 1451 case VREG: 1452 tsize = fxdr_unsigned(int32_t, sp->sa_size); 1453 if (tsize != -1) 1454 va.va_size = (u_quad_t)tsize; 1455 break; 1456 case VCHR: 1457 case VBLK: 1458 case VFIFO: 1459 rdev = fxdr_unsigned(int32_t, sp->sa_size); 1460 break; 1461 default: 1462 break; 1463 }; 1464 } 1465 1466 /* 1467 * Iff doesn't exist, create it 1468 * otherwise just truncate to 0 length 1469 * should I set the mode too ?? 1470 */ 1471 if (nd.ni_vp == NULL) { 1472 if (va.va_type == VREG || va.va_type == VSOCK) { 1473 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1474 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1475 if (!error) { 1476 if (exclusive_flag) { 1477 exclusive_flag = 0; 1478 VATTR_NULL(&va); 1479 /* 1480 * XXX 1481 * assuming NFSX_V3CREATEVERF 1482 * == sizeof(nfstime3) 1483 */ 1484 fxdr_nfsv3time(cverf, &va.va_atime); 1485 error = VOP_SETATTR(nd.ni_vp, &va, cred, 1486 lwp); 1487 } 1488 } 1489 } else if (va.va_type == VCHR || va.va_type == VBLK || 1490 va.va_type == VFIFO) { 1491 if (va.va_type == VCHR && rdev == 0xffffffff) 1492 va.va_type = VFIFO; 1493 if (va.va_type != VFIFO && 1494 (error = suser(cred, (u_short *)0))) { 1495 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1496 vput(nd.ni_dvp); 1497 nfsm_reply(0); 1498 vn_finished_write(mp, 0); 1499 return (error); 1500 } else 1501 va.va_rdev = (dev_t)rdev; 1502 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1503 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, 1504 &va); 1505 if (error) { 1506 nfsm_reply(0); 1507 } 1508 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1509 vrele(nd.ni_dvp); 1510 vput(nd.ni_vp); 1511 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1512 error = EINVAL; 1513 nfsm_reply(0); 1514 } 1515 } else { 1516 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1517 vput(nd.ni_dvp); 1518 error = ENXIO; 1519 } 1520 vp = nd.ni_vp; 1521 } else { 1522 vp = nd.ni_vp; 1523 if (nd.ni_dvp == vp) 1524 vrele(nd.ni_dvp); 1525 else 1526 vput(nd.ni_dvp); 1527 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1528 if (!error && va.va_size != -1) { 1529 error = nfsrv_access(vp, VWRITE, cred, 1530 (nd.ni_cnd.cn_flags & RDONLY), lwp, 0); 1531 if (!error) { 1532 nqsrv_getl(vp, ND_WRITE); 1533 tempsize = va.va_size; 1534 VATTR_NULL(&va); 1535 va.va_size = tempsize; 1536 error = VOP_SETATTR(vp, &va, cred, lwp); 1537 } 1538 } 1539 if (error) 1540 vput(vp); 1541 } 1542 if (!error) { 1543 memset((caddr_t)fhp, 0, sizeof(nfh)); 1544 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1545 error = VFS_VPTOFH(vp, &fhp->fh_fid); 1546 if (!error) 1547 error = VOP_GETATTR(vp, &va, cred, lwp); 1548 vput(vp); 1549 } 1550 if (v3) { 1551 if (exclusive_flag && !error) { 1552 /* 1553 * XXX assuming NFSX_V3CREATEVERF == sizeof(nfstime3) 1554 */ 1555 char oldverf[NFSX_V3CREATEVERF]; 1556 1557 txdr_nfsv3time(&va.va_atime, oldverf); 1558 if (memcmp(cverf, oldverf, NFSX_V3CREATEVERF)) 1559 error = EEXIST; 1560 } 1561 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1562 vrele(dirp); 1563 } 1564 nfsm_reply(NFSX_SRVFH(v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3)); 1565 if (v3) { 1566 if (!error) { 1567 nfsm_srvpostop_fh(fhp); 1568 nfsm_srvpostop_attr(0, &va); 1569 } 1570 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1571 } else { 1572 nfsm_srvfhtom(fhp, v3); 1573 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1574 nfsm_srvfillattr(&va, fp); 1575 } 1576 vn_finished_write(mp, 0); 1577 return (0); 1578 nfsmout: 1579 if (dirp) 1580 vrele(dirp); 1581 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1582 if (nd.ni_dvp == nd.ni_vp) 1583 vrele(nd.ni_dvp); 1584 else 1585 vput(nd.ni_dvp); 1586 if (nd.ni_vp) 1587 vput(nd.ni_vp); 1588 vn_finished_write(mp, 0); 1589 return (error); 1590 } 1591 1592 /* 1593 * nfs v3 mknod service 1594 */ 1595 int 1596 nfsrv_mknod(nfsd, slp, lwp, mrq) 1597 struct nfsrv_descript *nfsd; 1598 struct nfssvc_sock *slp; 1599 struct lwp *lwp; 1600 struct mbuf **mrq; 1601 { 1602 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1603 struct mbuf *nam = nfsd->nd_nam; 1604 caddr_t dpos = nfsd->nd_dpos; 1605 struct ucred *cred = &nfsd->nd_cr; 1606 struct vattr va, dirfor, diraft; 1607 u_int32_t *tl; 1608 struct nameidata nd; 1609 int32_t t1; 1610 caddr_t bpos; 1611 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 1612 u_int32_t major, minor; 1613 enum vtype vtyp; 1614 char *cp2; 1615 struct mbuf *mb, *mreq; 1616 struct vnode *vp, *dirp = (struct vnode *)0; 1617 nfsfh_t nfh; 1618 fhandle_t *fhp; 1619 u_quad_t frev; 1620 struct mount *mp = NULL; 1621 1622 nd.ni_cnd.cn_nameiop = 0; 1623 fhp = &nfh.fh_generic; 1624 nfsm_srvmtofh(fhp); 1625 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1626 return (ESTALE); 1627 vn_start_write(NULL, &mp, V_WAIT); 1628 nfsm_srvnamesiz(len); 1629 nd.ni_cnd.cn_cred = cred; 1630 nd.ni_cnd.cn_nameiop = CREATE; 1631 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1632 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1633 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1634 if (dirp) 1635 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1636 if (error) { 1637 nfsm_reply(NFSX_WCCDATA(1)); 1638 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1639 if (dirp) 1640 vrele(dirp); 1641 vn_finished_write(mp, 0); 1642 return (0); 1643 } 1644 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1645 vtyp = nfsv3tov_type(*tl); 1646 if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) { 1647 error = NFSERR_BADTYPE; 1648 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1649 if (nd.ni_dvp == nd.ni_vp) 1650 vrele(nd.ni_dvp); 1651 else 1652 vput(nd.ni_dvp); 1653 if (nd.ni_vp) 1654 vput(nd.ni_vp); 1655 goto out; 1656 } 1657 VATTR_NULL(&va); 1658 va.va_mode = 0; 1659 nfsm_srvsattr(&va); 1660 if (vtyp == VCHR || vtyp == VBLK) { 1661 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1662 major = fxdr_unsigned(u_int32_t, *tl++); 1663 minor = fxdr_unsigned(u_int32_t, *tl); 1664 va.va_rdev = makedev(major, minor); 1665 } 1666 1667 /* 1668 * Iff doesn't exist, create it. 1669 */ 1670 if (nd.ni_vp) { 1671 error = EEXIST; 1672 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1673 if (nd.ni_dvp == nd.ni_vp) 1674 vrele(nd.ni_dvp); 1675 else 1676 vput(nd.ni_dvp); 1677 vput(nd.ni_vp); 1678 goto out; 1679 } 1680 va.va_type = vtyp; 1681 if (vtyp == VSOCK) { 1682 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1683 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1684 } else { 1685 if (va.va_type != VFIFO && 1686 (error = suser(cred, (u_short *)0))) { 1687 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1688 vput(nd.ni_dvp); 1689 goto out; 1690 } 1691 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1692 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1693 if (error) { 1694 goto out; 1695 } 1696 if (error) 1697 goto out; 1698 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1699 vput(nd.ni_vp); 1700 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1701 error = EINVAL; 1702 } 1703 } 1704 out: 1705 vp = nd.ni_vp; 1706 if (!error) { 1707 memset((caddr_t)fhp, 0, sizeof(nfh)); 1708 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1709 error = VFS_VPTOFH(vp, &fhp->fh_fid); 1710 if (!error) 1711 error = VOP_GETATTR(vp, &va, cred, lwp); 1712 vput(vp); 1713 } 1714 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1715 vrele(dirp); 1716 nfsm_reply(NFSX_SRVFH(1) + NFSX_POSTOPATTR(1) + NFSX_WCCDATA(1)); 1717 if (!error) { 1718 nfsm_srvpostop_fh(fhp); 1719 nfsm_srvpostop_attr(0, &va); 1720 } 1721 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1722 vn_finished_write(mp, 0); 1723 return (0); 1724 nfsmout: 1725 if (dirp) 1726 vrele(dirp); 1727 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1728 if (nd.ni_dvp == nd.ni_vp) 1729 vrele(nd.ni_dvp); 1730 else 1731 vput(nd.ni_dvp); 1732 if (nd.ni_vp) 1733 vput(nd.ni_vp); 1734 vn_finished_write(mp, 0); 1735 return (error); 1736 } 1737 1738 /* 1739 * nfs remove service 1740 */ 1741 int 1742 nfsrv_remove(nfsd, slp, lwp, mrq) 1743 struct nfsrv_descript *nfsd; 1744 struct nfssvc_sock *slp; 1745 struct lwp *lwp; 1746 struct mbuf **mrq; 1747 { 1748 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1749 struct mbuf *nam = nfsd->nd_nam; 1750 caddr_t dpos = nfsd->nd_dpos; 1751 struct ucred *cred = &nfsd->nd_cr; 1752 struct nameidata nd; 1753 u_int32_t *tl; 1754 int32_t t1; 1755 caddr_t bpos; 1756 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 1757 int v3 = (nfsd->nd_flag & ND_NFSV3); 1758 char *cp2; 1759 struct mbuf *mb, *mreq; 1760 struct vnode *vp, *dirp; 1761 struct vattr dirfor, diraft; 1762 nfsfh_t nfh; 1763 fhandle_t *fhp; 1764 u_quad_t frev; 1765 struct mount *mp = NULL; 1766 1767 #ifndef nolint 1768 vp = (struct vnode *)0; 1769 #endif 1770 fhp = &nfh.fh_generic; 1771 nfsm_srvmtofh(fhp); 1772 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1773 return (ESTALE); 1774 vn_start_write(NULL, &mp, V_WAIT); 1775 nfsm_srvnamesiz(len); 1776 nd.ni_cnd.cn_cred = cred; 1777 nd.ni_cnd.cn_nameiop = DELETE; 1778 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1779 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1780 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1781 if (dirp) { 1782 if (v3) 1783 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1784 else 1785 vrele(dirp); 1786 } 1787 if (!error) { 1788 vp = nd.ni_vp; 1789 if (vp->v_type == VDIR && 1790 (error = suser(cred, (u_short *)0)) != 0) 1791 goto out; 1792 /* 1793 * The root of a mounted filesystem cannot be deleted. 1794 */ 1795 if (vp->v_flag & VROOT) { 1796 error = EBUSY; 1797 goto out; 1798 } 1799 out: 1800 if (!error) { 1801 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1802 nqsrv_getl(vp, ND_WRITE); 1803 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1804 } else { 1805 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1806 if (nd.ni_dvp == vp) 1807 vrele(nd.ni_dvp); 1808 else 1809 vput(nd.ni_dvp); 1810 vput(vp); 1811 } 1812 } 1813 if (dirp && v3) { 1814 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1815 vrele(dirp); 1816 } 1817 nfsm_reply(NFSX_WCCDATA(v3)); 1818 if (v3) { 1819 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1820 vn_finished_write(mp, 0); 1821 return (0); 1822 } 1823 vn_finished_write(mp, 0); 1824 nfsm_srvdone; 1825 } 1826 1827 /* 1828 * nfs rename service 1829 */ 1830 int 1831 nfsrv_rename(nfsd, slp, lwp, mrq) 1832 struct nfsrv_descript *nfsd; 1833 struct nfssvc_sock *slp; 1834 struct lwp *lwp; 1835 struct mbuf **mrq; 1836 { 1837 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1838 struct mbuf *nam = nfsd->nd_nam; 1839 caddr_t dpos = nfsd->nd_dpos; 1840 struct ucred *cred = &nfsd->nd_cr; 1841 u_int32_t *tl; 1842 int32_t t1; 1843 caddr_t bpos; 1844 int error = 0, cache, fdirfor_ret = 1, fdiraft_ret = 1; 1845 uint32_t len, len2; 1846 int tdirfor_ret = 1, tdiraft_ret = 1; 1847 int v3 = (nfsd->nd_flag & ND_NFSV3); 1848 char *cp2; 1849 struct mbuf *mb, *mreq; 1850 struct nameidata fromnd, tond; 1851 struct vnode *fvp, *tvp, *tdvp, *fdirp = (struct vnode *)0; 1852 struct vnode *tdirp = (struct vnode *)0; 1853 struct vattr fdirfor, fdiraft, tdirfor, tdiraft; 1854 nfsfh_t fnfh, tnfh; 1855 fhandle_t *ffhp, *tfhp; 1856 u_quad_t frev; 1857 uid_t saved_uid; 1858 struct mount *mp = NULL; 1859 1860 #ifndef nolint 1861 fvp = (struct vnode *)0; 1862 #endif 1863 ffhp = &fnfh.fh_generic; 1864 tfhp = &tnfh.fh_generic; 1865 fromnd.ni_cnd.cn_nameiop = 0; 1866 tond.ni_cnd.cn_nameiop = 0; 1867 nfsm_srvmtofh(ffhp); 1868 if ((mp = vfs_getvfs(&ffhp->fh_fsid)) == NULL) 1869 return (ESTALE); 1870 vn_start_write(NULL, &mp, V_WAIT); 1871 nfsm_srvnamesiz(len); 1872 /* 1873 * Remember our original uid so that we can reset cr_uid before 1874 * the second nfs_namei() call, in case it is remapped. 1875 */ 1876 saved_uid = cred->cr_uid; 1877 fromnd.ni_cnd.cn_cred = cred; 1878 fromnd.ni_cnd.cn_nameiop = DELETE; 1879 fromnd.ni_cnd.cn_flags = WANTPARENT | SAVESTART; 1880 error = nfs_namei(&fromnd, ffhp, len, slp, nam, &md, 1881 &dpos, &fdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1882 if (fdirp) { 1883 if (v3) 1884 fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred, lwp); 1885 else { 1886 vrele(fdirp); 1887 fdirp = (struct vnode *)0; 1888 } 1889 } 1890 if (error) { 1891 nfsm_reply(2 * NFSX_WCCDATA(v3)); 1892 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 1893 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 1894 if (fdirp) 1895 vrele(fdirp); 1896 vn_finished_write(mp, 0); 1897 return (0); 1898 } 1899 fvp = fromnd.ni_vp; 1900 nfsm_srvmtofh(tfhp); 1901 if (v3) { 1902 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 1903 len2 = fxdr_unsigned(uint32_t, *tl); 1904 /* len2 will be checked by nfs_namei */ 1905 } 1906 else { 1907 /* NFSv2 */ 1908 nfsm_strsiz(len2, NFS_MAXNAMLEN); 1909 } 1910 cred->cr_uid = saved_uid; 1911 tond.ni_cnd.cn_cred = cred; 1912 tond.ni_cnd.cn_nameiop = RENAME; 1913 tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART; 1914 error = nfs_namei(&tond, tfhp, len2, slp, nam, &md, 1915 &dpos, &tdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1916 if (tdirp) { 1917 if (v3) 1918 tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred, lwp); 1919 else { 1920 vrele(tdirp); 1921 tdirp = (struct vnode *)0; 1922 } 1923 } 1924 if (error) { 1925 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1926 vrele(fromnd.ni_dvp); 1927 vrele(fvp); 1928 goto out1; 1929 } 1930 tdvp = tond.ni_dvp; 1931 tvp = tond.ni_vp; 1932 if (tvp != NULL) { 1933 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 1934 if (v3) 1935 error = EEXIST; 1936 else 1937 error = EISDIR; 1938 goto out; 1939 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 1940 if (v3) 1941 error = EEXIST; 1942 else 1943 error = ENOTDIR; 1944 goto out; 1945 } 1946 if (tvp->v_type == VDIR && tvp->v_mountedhere) { 1947 if (v3) 1948 error = EXDEV; 1949 else 1950 error = ENOTEMPTY; 1951 goto out; 1952 } 1953 } 1954 if (fvp->v_type == VDIR && fvp->v_mountedhere) { 1955 if (v3) 1956 error = EXDEV; 1957 else 1958 error = ENOTEMPTY; 1959 goto out; 1960 } 1961 if (fvp->v_mount != tdvp->v_mount) { 1962 if (v3) 1963 error = EXDEV; 1964 else 1965 error = ENOTEMPTY; 1966 goto out; 1967 } 1968 if (fvp == tdvp) { 1969 if (v3) 1970 error = EINVAL; 1971 else 1972 error = ENOTEMPTY; 1973 } 1974 /* 1975 * If source is the same as the destination (that is the 1976 * same vnode with the same name in the same directory), 1977 * then there is nothing to do. 1978 */ 1979 if (fvp == tvp && fromnd.ni_dvp == tdvp && 1980 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 1981 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr, 1982 fromnd.ni_cnd.cn_namelen)) 1983 error = -1; 1984 out: 1985 if (!error) { 1986 nqsrv_getl(fromnd.ni_dvp, ND_WRITE); 1987 nqsrv_getl(tdvp, ND_WRITE); 1988 if (tvp) { 1989 nqsrv_getl(tvp, ND_WRITE); 1990 } 1991 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 1992 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 1993 } else { 1994 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 1995 if (tdvp == tvp) 1996 vrele(tdvp); 1997 else 1998 vput(tdvp); 1999 if (tvp) 2000 vput(tvp); 2001 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2002 vrele(fromnd.ni_dvp); 2003 vrele(fvp); 2004 if (error == -1) 2005 error = 0; 2006 } 2007 vrele(tond.ni_startdir); 2008 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 2009 out1: 2010 if (fdirp) { 2011 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred, lwp); 2012 vrele(fdirp); 2013 } 2014 if (tdirp) { 2015 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred, lwp); 2016 vrele(tdirp); 2017 } 2018 vrele(fromnd.ni_startdir); 2019 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2020 nfsm_reply(2 * NFSX_WCCDATA(v3)); 2021 if (v3) { 2022 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 2023 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 2024 } 2025 vn_finished_write(mp, 0); 2026 return (0); 2027 2028 nfsmout: 2029 if (fdirp) 2030 vrele(fdirp); 2031 if (tdirp) 2032 vrele(tdirp); 2033 if (tond.ni_cnd.cn_nameiop) { 2034 vrele(tond.ni_startdir); 2035 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 2036 } 2037 if (fromnd.ni_cnd.cn_nameiop) { 2038 vrele(fromnd.ni_startdir); 2039 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2040 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2041 vrele(fromnd.ni_dvp); 2042 vrele(fvp); 2043 } 2044 return (error); 2045 } 2046 2047 /* 2048 * nfs link service 2049 */ 2050 int 2051 nfsrv_link(nfsd, slp, lwp, mrq) 2052 struct nfsrv_descript *nfsd; 2053 struct nfssvc_sock *slp; 2054 struct lwp *lwp; 2055 struct mbuf **mrq; 2056 { 2057 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2058 struct mbuf *nam = nfsd->nd_nam; 2059 caddr_t dpos = nfsd->nd_dpos; 2060 struct ucred *cred = &nfsd->nd_cr; 2061 struct nameidata nd; 2062 u_int32_t *tl; 2063 int32_t t1; 2064 caddr_t bpos; 2065 int error = 0, rdonly, cache, len, dirfor_ret = 1, diraft_ret = 1; 2066 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3); 2067 char *cp2; 2068 struct mbuf *mb, *mreq; 2069 struct vnode *vp, *xp, *dirp = (struct vnode *)0; 2070 struct vattr dirfor, diraft, at; 2071 nfsfh_t nfh, dnfh; 2072 fhandle_t *fhp, *dfhp; 2073 u_quad_t frev; 2074 struct mount *mp = NULL; 2075 2076 fhp = &nfh.fh_generic; 2077 dfhp = &dnfh.fh_generic; 2078 nfsm_srvmtofh(fhp); 2079 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2080 return (ESTALE); 2081 vn_start_write(NULL, &mp, V_WAIT); 2082 nfsm_srvmtofh(dfhp); 2083 nfsm_srvnamesiz(len); 2084 error = nfsrv_fhtovp(fhp, FALSE, &vp, cred, slp, nam, 2085 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2086 if (error) { 2087 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2088 nfsm_srvpostop_attr(getret, &at); 2089 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2090 vn_finished_write(mp, 0); 2091 return (0); 2092 } 2093 if (vp->v_type == VDIR && (error = suser(cred, (u_short *)0)) != 0) 2094 goto out1; 2095 nd.ni_cnd.cn_cred = cred; 2096 nd.ni_cnd.cn_nameiop = CREATE; 2097 nd.ni_cnd.cn_flags = LOCKPARENT; 2098 error = nfs_namei(&nd, dfhp, len, slp, nam, &md, &dpos, 2099 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2100 if (dirp) { 2101 if (v3) 2102 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2103 else { 2104 vrele(dirp); 2105 dirp = (struct vnode *)0; 2106 } 2107 } 2108 if (error) 2109 goto out1; 2110 xp = nd.ni_vp; 2111 if (xp != NULL) { 2112 error = EEXIST; 2113 goto out; 2114 } 2115 xp = nd.ni_dvp; 2116 if (vp->v_mount != xp->v_mount) 2117 error = EXDEV; 2118 out: 2119 if (!error) { 2120 nqsrv_getl(vp, ND_WRITE); 2121 nqsrv_getl(xp, ND_WRITE); 2122 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2123 } else { 2124 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2125 if (nd.ni_dvp == nd.ni_vp) 2126 vrele(nd.ni_dvp); 2127 else 2128 vput(nd.ni_dvp); 2129 if (nd.ni_vp) 2130 vrele(nd.ni_vp); 2131 } 2132 out1: 2133 if (v3) 2134 getret = VOP_GETATTR(vp, &at, cred, lwp); 2135 if (dirp) { 2136 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2137 vrele(dirp); 2138 } 2139 vrele(vp); 2140 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2141 if (v3) { 2142 nfsm_srvpostop_attr(getret, &at); 2143 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2144 vn_finished_write(mp, 0); 2145 return (0); 2146 } 2147 vn_finished_write(mp, 0); 2148 nfsm_srvdone; 2149 } 2150 2151 /* 2152 * nfs symbolic link service 2153 */ 2154 int 2155 nfsrv_symlink(nfsd, slp, lwp, mrq) 2156 struct nfsrv_descript *nfsd; 2157 struct nfssvc_sock *slp; 2158 struct lwp *lwp; 2159 struct mbuf **mrq; 2160 { 2161 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2162 struct mbuf *nam = nfsd->nd_nam; 2163 caddr_t dpos = nfsd->nd_dpos; 2164 struct ucred *cred = &nfsd->nd_cr; 2165 struct vattr va, dirfor, diraft; 2166 struct nameidata nd; 2167 u_int32_t *tl; 2168 int32_t t1; 2169 struct nfsv2_sattr *sp; 2170 char *bpos, *pathcp = NULL, *cp2; 2171 struct uio io; 2172 struct iovec iv; 2173 int error = 0, cache, dirfor_ret = 1, diraft_ret = 1; 2174 uint32_t len, len2; 2175 int v3 = (nfsd->nd_flag & ND_NFSV3); 2176 struct mbuf *mb, *mreq; 2177 struct vnode *dirp = (struct vnode *)0; 2178 nfsfh_t nfh; 2179 fhandle_t *fhp; 2180 u_quad_t frev; 2181 struct mount *mp = NULL; 2182 2183 nd.ni_cnd.cn_nameiop = 0; 2184 fhp = &nfh.fh_generic; 2185 nfsm_srvmtofh(fhp); 2186 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2187 return (ESTALE); 2188 vn_start_write(NULL, &mp, V_WAIT); 2189 nfsm_srvnamesiz(len); 2190 nd.ni_cnd.cn_cred = cred; 2191 nd.ni_cnd.cn_nameiop = CREATE; 2192 nd.ni_cnd.cn_flags = LOCKPARENT; 2193 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2194 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2195 if (dirp) { 2196 if (v3) 2197 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2198 else { 2199 vrele(dirp); 2200 dirp = (struct vnode *)0; 2201 } 2202 } 2203 if (error) 2204 goto out; 2205 VATTR_NULL(&va); 2206 if (v3) { 2207 va.va_mode = 0; 2208 nfsm_srvsattr(&va); 2209 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 2210 len2 = fxdr_unsigned(uint32_t, *tl); 2211 if (len2 > PATH_MAX) { 2212 /* XXX should check _PC_NO_TRUNC */ 2213 error = ENAMETOOLONG; 2214 goto abortop; 2215 } 2216 } 2217 else { 2218 /* NFSv2 */ 2219 nfsm_strsiz(len2, NFS_MAXPATHLEN); 2220 } 2221 pathcp = malloc(len2 + 1, M_TEMP, M_WAITOK); 2222 iv.iov_base = pathcp; 2223 iv.iov_len = len2; 2224 io.uio_resid = len2; 2225 io.uio_offset = 0; 2226 io.uio_iov = &iv; 2227 io.uio_iovcnt = 1; 2228 io.uio_segflg = UIO_SYSSPACE; 2229 io.uio_rw = UIO_READ; 2230 io.uio_lwp = NULL; 2231 nfsm_mtouio(&io, len2); 2232 if (!v3) { 2233 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 2234 va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode); 2235 } 2236 *(pathcp + len2) = '\0'; 2237 if (nd.ni_vp) { 2238 error = EEXIST; 2239 abortop: 2240 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2241 if (nd.ni_dvp == nd.ni_vp) 2242 vrele(nd.ni_dvp); 2243 else 2244 vput(nd.ni_dvp); 2245 if (nd.ni_vp) 2246 vrele(nd.ni_vp); 2247 goto out; 2248 } 2249 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2250 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp); 2251 if (!error) { 2252 if (v3) { 2253 memset((caddr_t)fhp, 0, sizeof(nfh)); 2254 fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsidx; 2255 error = VFS_VPTOFH(nd.ni_vp, &fhp->fh_fid); 2256 if (!error) 2257 error = VOP_GETATTR(nd.ni_vp, &va, cred, lwp); 2258 vput(nd.ni_vp); 2259 } else { 2260 vput(nd.ni_vp); 2261 } 2262 } 2263 out: 2264 if (pathcp) 2265 free(pathcp, M_TEMP); 2266 if (dirp) { 2267 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2268 vrele(dirp); 2269 } 2270 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2271 if (v3) { 2272 if (!error) { 2273 nfsm_srvpostop_fh(fhp); 2274 nfsm_srvpostop_attr(0, &va); 2275 } 2276 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2277 } 2278 vn_finished_write(mp, 0); 2279 return (0); 2280 nfsmout: 2281 if (dirp) 2282 vrele(dirp); 2283 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2284 if (nd.ni_dvp == nd.ni_vp) 2285 vrele(nd.ni_dvp); 2286 else 2287 vput(nd.ni_dvp); 2288 if (nd.ni_vp) 2289 vrele(nd.ni_vp); 2290 if (pathcp) 2291 free(pathcp, M_TEMP); 2292 vn_finished_write(mp, 0); 2293 return (error); 2294 } 2295 2296 /* 2297 * nfs mkdir service 2298 */ 2299 int 2300 nfsrv_mkdir(nfsd, slp, lwp, mrq) 2301 struct nfsrv_descript *nfsd; 2302 struct nfssvc_sock *slp; 2303 struct lwp *lwp; 2304 struct mbuf **mrq; 2305 { 2306 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2307 struct mbuf *nam = nfsd->nd_nam; 2308 caddr_t dpos = nfsd->nd_dpos; 2309 struct ucred *cred = &nfsd->nd_cr; 2310 struct vattr va, dirfor, diraft; 2311 struct nfs_fattr *fp; 2312 struct nameidata nd; 2313 caddr_t cp; 2314 u_int32_t *tl; 2315 int32_t t1; 2316 caddr_t bpos; 2317 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 2318 int v3 = (nfsd->nd_flag & ND_NFSV3); 2319 char *cp2; 2320 struct mbuf *mb, *mreq; 2321 struct vnode *vp, *dirp = (struct vnode *)0; 2322 nfsfh_t nfh; 2323 fhandle_t *fhp; 2324 u_quad_t frev; 2325 struct mount *mp = NULL; 2326 2327 fhp = &nfh.fh_generic; 2328 nfsm_srvmtofh(fhp); 2329 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2330 return (ESTALE); 2331 vn_start_write(NULL, &mp, V_WAIT); 2332 nfsm_srvnamesiz(len); 2333 nd.ni_cnd.cn_cred = cred; 2334 nd.ni_cnd.cn_nameiop = CREATE; 2335 nd.ni_cnd.cn_flags = LOCKPARENT; 2336 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2337 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2338 if (dirp) { 2339 if (v3) 2340 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2341 else { 2342 vrele(dirp); 2343 dirp = (struct vnode *)0; 2344 } 2345 } 2346 if (error) { 2347 nfsm_reply(NFSX_WCCDATA(v3)); 2348 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2349 if (dirp) 2350 vrele(dirp); 2351 vn_finished_write(mp, 0); 2352 return (0); 2353 } 2354 VATTR_NULL(&va); 2355 if (v3) { 2356 va.va_mode = 0; 2357 nfsm_srvsattr(&va); 2358 } else { 2359 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 2360 va.va_mode = nfstov_mode(*tl++); 2361 } 2362 va.va_type = VDIR; 2363 vp = nd.ni_vp; 2364 if (vp != NULL) { 2365 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2366 if (nd.ni_dvp == vp) 2367 vrele(nd.ni_dvp); 2368 else 2369 vput(nd.ni_dvp); 2370 vrele(vp); 2371 error = EEXIST; 2372 goto out; 2373 } 2374 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2375 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 2376 if (!error) { 2377 vp = nd.ni_vp; 2378 memset((caddr_t)fhp, 0, sizeof(nfh)); 2379 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 2380 error = VFS_VPTOFH(vp, &fhp->fh_fid); 2381 if (!error) 2382 error = VOP_GETATTR(vp, &va, cred, lwp); 2383 vput(vp); 2384 } 2385 out: 2386 if (dirp) { 2387 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2388 vrele(dirp); 2389 } 2390 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2391 if (v3) { 2392 if (!error) { 2393 nfsm_srvpostop_fh(fhp); 2394 nfsm_srvpostop_attr(0, &va); 2395 } 2396 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2397 } else { 2398 nfsm_srvfhtom(fhp, v3); 2399 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 2400 nfsm_srvfillattr(&va, fp); 2401 } 2402 vn_finished_write(mp, 0); 2403 return (0); 2404 nfsmout: 2405 if (dirp) 2406 vrele(dirp); 2407 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2408 if (nd.ni_dvp == nd.ni_vp) 2409 vrele(nd.ni_dvp); 2410 else 2411 vput(nd.ni_dvp); 2412 if (nd.ni_vp) 2413 vrele(nd.ni_vp); 2414 vn_finished_write(mp, 0); 2415 return (error); 2416 } 2417 2418 /* 2419 * nfs rmdir service 2420 */ 2421 int 2422 nfsrv_rmdir(nfsd, slp, lwp, mrq) 2423 struct nfsrv_descript *nfsd; 2424 struct nfssvc_sock *slp; 2425 struct lwp *lwp; 2426 struct mbuf **mrq; 2427 { 2428 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2429 struct mbuf *nam = nfsd->nd_nam; 2430 caddr_t dpos = nfsd->nd_dpos; 2431 struct ucred *cred = &nfsd->nd_cr; 2432 u_int32_t *tl; 2433 int32_t t1; 2434 caddr_t bpos; 2435 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 2436 int v3 = (nfsd->nd_flag & ND_NFSV3); 2437 char *cp2; 2438 struct mbuf *mb, *mreq; 2439 struct vnode *vp, *dirp = (struct vnode *)0; 2440 struct vattr dirfor, diraft; 2441 nfsfh_t nfh; 2442 fhandle_t *fhp; 2443 struct nameidata nd; 2444 u_quad_t frev; 2445 struct mount *mp = NULL; 2446 2447 fhp = &nfh.fh_generic; 2448 nfsm_srvmtofh(fhp); 2449 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2450 return (ESTALE); 2451 vn_start_write(NULL, &mp, V_WAIT); 2452 nfsm_srvnamesiz(len); 2453 nd.ni_cnd.cn_cred = cred; 2454 nd.ni_cnd.cn_nameiop = DELETE; 2455 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 2456 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2457 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2458 if (dirp) { 2459 if (v3) 2460 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2461 else { 2462 vrele(dirp); 2463 dirp = (struct vnode *)0; 2464 } 2465 } 2466 if (error) { 2467 nfsm_reply(NFSX_WCCDATA(v3)); 2468 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2469 if (dirp) 2470 vrele(dirp); 2471 vn_finished_write(mp, 0); 2472 return (0); 2473 } 2474 vp = nd.ni_vp; 2475 if (vp->v_type != VDIR) { 2476 error = ENOTDIR; 2477 goto out; 2478 } 2479 /* 2480 * No rmdir "." please. 2481 */ 2482 if (nd.ni_dvp == vp) { 2483 error = EINVAL; 2484 goto out; 2485 } 2486 /* 2487 * The root of a mounted filesystem cannot be deleted. 2488 */ 2489 if (vp->v_flag & VROOT) 2490 error = EBUSY; 2491 out: 2492 if (!error) { 2493 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2494 nqsrv_getl(vp, ND_WRITE); 2495 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2496 } else { 2497 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2498 if (nd.ni_dvp == nd.ni_vp) 2499 vrele(nd.ni_dvp); 2500 else 2501 vput(nd.ni_dvp); 2502 vput(vp); 2503 } 2504 if (dirp) { 2505 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2506 vrele(dirp); 2507 } 2508 nfsm_reply(NFSX_WCCDATA(v3)); 2509 if (v3) { 2510 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2511 vn_finished_write(mp, 0); 2512 return (0); 2513 } 2514 vn_finished_write(mp, 0); 2515 nfsm_srvdone; 2516 } 2517 2518 /* 2519 * nfs readdir service 2520 * - mallocs what it thinks is enough to read 2521 * count rounded up to a multiple of NFS_DIRBLKSIZ <= NFS_MAXREADDIR 2522 * - calls VOP_READDIR() 2523 * - loops around building the reply 2524 * if the output generated exceeds count break out of loop 2525 * The nfsm_clget macro is used here so that the reply will be packed 2526 * tightly in mbuf clusters. 2527 * - it only knows that it has encountered eof when the VOP_READDIR() 2528 * reads nothing 2529 * - as such one readdir rpc will return eof false although you are there 2530 * and then the next will return eof 2531 * - it trims out records with d_fileno == 0 2532 * this doesn't matter for Unix clients, but they might confuse clients 2533 * for other os'. 2534 * - it trims out records with d_type == DT_WHT 2535 * these cannot be seen through NFS (unless we extend the protocol) 2536 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less 2537 * than requested, but this may not apply to all filesystems. For 2538 * example, client NFS does not { although it is never remote mounted 2539 * anyhow } 2540 * The alternate call nfsrv_readdirplus() does lookups as well. 2541 * PS: The NFS protocol spec. does not clarify what the "count" byte 2542 * argument is a count of.. just name strings and file id's or the 2543 * entire reply rpc or ... 2544 * I tried just file name and id sizes and it confused the Sun client, 2545 * so I am using the full rpc size now. The "paranoia.." comment refers 2546 * to including the status longwords that are not a part of the dir. 2547 * "entry" structures, but are in the rpc. 2548 */ 2549 struct flrep { 2550 nfsuint64 fl_off; 2551 u_int32_t fl_postopok; 2552 u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)]; 2553 u_int32_t fl_fhok; 2554 u_int32_t fl_fhsize; 2555 u_int32_t fl_nfh[NFSX_V3FH / sizeof (u_int32_t)]; 2556 }; 2557 2558 int 2559 nfsrv_readdir(nfsd, slp, lwp, mrq) 2560 struct nfsrv_descript *nfsd; 2561 struct nfssvc_sock *slp; 2562 struct lwp *lwp; 2563 struct mbuf **mrq; 2564 { 2565 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2566 struct mbuf *nam = nfsd->nd_nam; 2567 caddr_t dpos = nfsd->nd_dpos; 2568 struct ucred *cred = &nfsd->nd_cr; 2569 char *bp, *be; 2570 struct mbuf *mp; 2571 struct dirent *dp; 2572 caddr_t cp; 2573 u_int32_t *tl; 2574 int32_t t1; 2575 caddr_t bpos; 2576 struct mbuf *mb, *mreq, *mp2; 2577 char *cpos, *cend, *cp2, *rbuf; 2578 struct vnode *vp; 2579 struct vattr at; 2580 nfsfh_t nfh; 2581 fhandle_t *fhp; 2582 struct uio io; 2583 struct iovec iv; 2584 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2585 int siz, cnt, fullsiz, eofflag, rdonly, cache, ncookies; 2586 int v3 = (nfsd->nd_flag & ND_NFSV3); 2587 u_quad_t frev, off, toff, verf; 2588 off_t *cookies = NULL, *cookiep; 2589 nfsuint64 jar; 2590 2591 fhp = &nfh.fh_generic; 2592 nfsm_srvmtofh(fhp); 2593 if (v3) { 2594 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 2595 toff = fxdr_hyper(tl); 2596 tl += 2; 2597 verf = fxdr_hyper(tl); 2598 tl += 2; 2599 } else { 2600 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2601 toff = fxdr_unsigned(u_quad_t, *tl++); 2602 } 2603 off = toff; 2604 cnt = fxdr_unsigned(int, *tl); 2605 siz = ((cnt + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1)); 2606 xfer = NFS_SRVMAXDATA(nfsd); 2607 if (siz > xfer) 2608 siz = xfer; 2609 fullsiz = siz; 2610 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 2611 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2612 if (!error && vp->v_type != VDIR) { 2613 error = ENOTDIR; 2614 vput(vp); 2615 } 2616 if (error) { 2617 nfsm_reply(NFSX_UNSIGNED); 2618 nfsm_srvpostop_attr(getret, &at); 2619 return (0); 2620 } 2621 nqsrv_getl(vp, ND_READ); 2622 if (v3) { 2623 error = getret = VOP_GETATTR(vp, &at, cred, lwp); 2624 #ifdef NFS3_STRICTVERF 2625 /* 2626 * XXX This check is too strict for Solaris 2.5 clients. 2627 */ 2628 if (!error && toff && verf != at.va_filerev) 2629 error = NFSERR_BAD_COOKIE; 2630 #endif 2631 } 2632 if (!error) 2633 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0); 2634 if (error) { 2635 vput(vp); 2636 nfsm_reply(NFSX_POSTOPATTR(v3)); 2637 nfsm_srvpostop_attr(getret, &at); 2638 return (0); 2639 } 2640 VOP_UNLOCK(vp, 0); 2641 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2642 again: 2643 iv.iov_base = rbuf; 2644 iv.iov_len = fullsiz; 2645 io.uio_iov = &iv; 2646 io.uio_iovcnt = 1; 2647 io.uio_offset = (off_t)off; 2648 io.uio_resid = fullsiz; 2649 io.uio_segflg = UIO_SYSSPACE; 2650 io.uio_rw = UIO_READ; 2651 io.uio_lwp = NULL; 2652 eofflag = 0; 2653 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2654 2655 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2656 2657 off = (off_t)io.uio_offset; 2658 if (!cookies && !error) 2659 error = NFSERR_PERM; 2660 if (v3) { 2661 getret = VOP_GETATTR(vp, &at, cred, lwp); 2662 if (!error) 2663 error = getret; 2664 } 2665 2666 VOP_UNLOCK(vp, 0); 2667 if (error) { 2668 vrele(vp); 2669 free((caddr_t)rbuf, M_TEMP); 2670 if (cookies) 2671 free((caddr_t)cookies, M_TEMP); 2672 nfsm_reply(NFSX_POSTOPATTR(v3)); 2673 nfsm_srvpostop_attr(getret, &at); 2674 return (0); 2675 } 2676 if (io.uio_resid) { 2677 siz -= io.uio_resid; 2678 2679 /* 2680 * If nothing read, return eof 2681 * rpc reply 2682 */ 2683 if (siz == 0) { 2684 vrele(vp); 2685 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + 2686 2 * NFSX_UNSIGNED); 2687 if (v3) { 2688 nfsm_srvpostop_attr(getret, &at); 2689 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2690 txdr_hyper(at.va_filerev, tl); 2691 tl += 2; 2692 } else 2693 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2694 *tl++ = nfs_false; 2695 *tl = nfs_true; 2696 free((caddr_t)rbuf, M_TEMP); 2697 free((caddr_t)cookies, M_TEMP); 2698 return (0); 2699 } 2700 } 2701 2702 /* 2703 * Check for degenerate cases of nothing useful read. 2704 * If so go try again 2705 */ 2706 cpos = rbuf; 2707 cend = rbuf + siz; 2708 dp = (struct dirent *)cpos; 2709 cookiep = cookies; 2710 2711 while (cpos < cend && ncookies > 0 && 2712 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2713 cpos += dp->d_reclen; 2714 dp = (struct dirent *)cpos; 2715 cookiep++; 2716 ncookies--; 2717 } 2718 if (cpos >= cend || ncookies == 0) { 2719 toff = off; 2720 siz = fullsiz; 2721 goto again; 2722 } 2723 2724 len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */ 2725 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz); 2726 if (v3) { 2727 nfsm_srvpostop_attr(getret, &at); 2728 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2729 txdr_hyper(at.va_filerev, tl); 2730 } 2731 mp = mp2 = mb; 2732 bp = bpos; 2733 be = bp + M_TRAILINGSPACE(mp); 2734 2735 /* Loop through the records and build reply */ 2736 while (cpos < cend && ncookies > 0) { 2737 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 2738 nlen = dp->d_namlen; 2739 rem = nfsm_rndup(nlen)-nlen; 2740 len += (4 * NFSX_UNSIGNED + nlen + rem); 2741 if (v3) 2742 len += 2 * NFSX_UNSIGNED; 2743 if (len > cnt) { 2744 eofflag = 0; 2745 break; 2746 } 2747 /* 2748 * Build the directory record xdr from 2749 * the dirent entry. 2750 */ 2751 nfsm_clget; 2752 *tl = nfs_true; 2753 bp += NFSX_UNSIGNED; 2754 if (v3) { 2755 nfsm_clget; 2756 *tl = 0; 2757 bp += NFSX_UNSIGNED; 2758 } 2759 nfsm_clget; 2760 *tl = txdr_unsigned(dp->d_fileno); 2761 bp += NFSX_UNSIGNED; 2762 nfsm_clget; 2763 *tl = txdr_unsigned(nlen); 2764 bp += NFSX_UNSIGNED; 2765 2766 /* And loop around copying the name */ 2767 xfer = nlen; 2768 cp = dp->d_name; 2769 while (xfer > 0) { 2770 nfsm_clget; 2771 if ((bp+xfer) > be) 2772 tsiz = be-bp; 2773 else 2774 tsiz = xfer; 2775 memcpy(bp, cp, tsiz); 2776 bp += tsiz; 2777 xfer -= tsiz; 2778 if (xfer > 0) 2779 cp += tsiz; 2780 } 2781 /* And null pad to an int32_t boundary */ 2782 for (i = 0; i < rem; i++) 2783 *bp++ = '\0'; 2784 nfsm_clget; 2785 2786 /* Finish off the record */ 2787 txdr_hyper(*cookiep, &jar); 2788 if (v3) { 2789 *tl = jar.nfsuquad[0]; 2790 bp += NFSX_UNSIGNED; 2791 nfsm_clget; 2792 } 2793 *tl = jar.nfsuquad[1]; 2794 bp += NFSX_UNSIGNED; 2795 } 2796 cpos += dp->d_reclen; 2797 dp = (struct dirent *)cpos; 2798 cookiep++; 2799 ncookies--; 2800 } 2801 vrele(vp); 2802 nfsm_clget; 2803 *tl = nfs_false; 2804 bp += NFSX_UNSIGNED; 2805 nfsm_clget; 2806 if (eofflag) 2807 *tl = nfs_true; 2808 else 2809 *tl = nfs_false; 2810 bp += NFSX_UNSIGNED; 2811 if (mp != mb) { 2812 if (bp < be) 2813 mp->m_len = bp - mtod(mp, caddr_t); 2814 } else 2815 mp->m_len += bp - bpos; 2816 free((caddr_t)rbuf, M_TEMP); 2817 free((caddr_t)cookies, M_TEMP); 2818 nfsm_srvdone; 2819 } 2820 2821 int 2822 nfsrv_readdirplus(nfsd, slp, lwp, mrq) 2823 struct nfsrv_descript *nfsd; 2824 struct nfssvc_sock *slp; 2825 struct lwp *lwp; 2826 struct mbuf **mrq; 2827 { 2828 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2829 struct mbuf *nam = nfsd->nd_nam; 2830 caddr_t dpos = nfsd->nd_dpos; 2831 struct ucred *cred = &nfsd->nd_cr; 2832 char *bp, *be; 2833 struct mbuf *mp; 2834 struct dirent *dp; 2835 caddr_t cp; 2836 u_int32_t *tl; 2837 int32_t t1; 2838 caddr_t bpos; 2839 struct mbuf *mb, *mreq, *mp2; 2840 char *cpos, *cend, *cp2, *rbuf; 2841 struct vnode *vp, *nvp; 2842 struct flrep fl; 2843 nfsfh_t nfh; 2844 fhandle_t *fhp, *nfhp = (fhandle_t *)fl.fl_nfh; 2845 struct uio io; 2846 struct iovec iv; 2847 struct vattr va, at, *vap = &va; 2848 struct nfs_fattr *fp; 2849 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2850 int siz, cnt, fullsiz, eofflag, rdonly, cache, dirlen, ncookies; 2851 u_quad_t frev, off, toff, verf; 2852 off_t *cookies = NULL, *cookiep; 2853 2854 fhp = &nfh.fh_generic; 2855 nfsm_srvmtofh(fhp); 2856 nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED); 2857 toff = fxdr_hyper(tl); 2858 tl += 2; 2859 verf = fxdr_hyper(tl); 2860 tl += 2; 2861 siz = fxdr_unsigned(int, *tl++); 2862 cnt = fxdr_unsigned(int, *tl); 2863 off = toff; 2864 siz = ((siz + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1)); 2865 xfer = NFS_SRVMAXDATA(nfsd); 2866 if (siz > xfer) 2867 siz = xfer; 2868 fullsiz = siz; 2869 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 2870 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2871 if (!error && vp->v_type != VDIR) { 2872 error = ENOTDIR; 2873 vput(vp); 2874 } 2875 if (error) { 2876 nfsm_reply(NFSX_UNSIGNED); 2877 nfsm_srvpostop_attr(getret, &at); 2878 return (0); 2879 } 2880 error = getret = VOP_GETATTR(vp, &at, cred, lwp); 2881 #ifdef NFS3_STRICTVERF 2882 /* 2883 * XXX This check is too strict for Solaris 2.5 clients. 2884 */ 2885 if (!error && toff && verf != at.va_filerev) 2886 error = NFSERR_BAD_COOKIE; 2887 #endif 2888 if (!error) { 2889 nqsrv_getl(vp, ND_READ); 2890 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0); 2891 } 2892 if (error) { 2893 vput(vp); 2894 nfsm_reply(NFSX_V3POSTOPATTR); 2895 nfsm_srvpostop_attr(getret, &at); 2896 return (0); 2897 } 2898 VOP_UNLOCK(vp, 0); 2899 2900 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2901 again: 2902 iv.iov_base = rbuf; 2903 iv.iov_len = fullsiz; 2904 io.uio_iov = &iv; 2905 io.uio_iovcnt = 1; 2906 io.uio_offset = (off_t)off; 2907 io.uio_resid = fullsiz; 2908 io.uio_segflg = UIO_SYSSPACE; 2909 io.uio_rw = UIO_READ; 2910 io.uio_lwp = NULL; 2911 eofflag = 0; 2912 2913 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2914 2915 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2916 2917 off = (u_quad_t)io.uio_offset; 2918 getret = VOP_GETATTR(vp, &at, cred, lwp); 2919 2920 VOP_UNLOCK(vp, 0); 2921 2922 /* 2923 * If the VGET operation doesn't work for this filesystem, 2924 * we can't support readdirplus. Returning NOTSUPP should 2925 * make clients fall back to plain readdir. 2926 * There's no need to check for VPTOFH as well, we wouldn't 2927 * even be here otherwise. 2928 */ 2929 if (!getret) { 2930 if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp))) 2931 getret = (getret == EOPNOTSUPP) ? 2932 NFSERR_NOTSUPP : NFSERR_IO; 2933 else 2934 vput(nvp); 2935 } 2936 2937 if (!cookies && !error) 2938 error = NFSERR_PERM; 2939 if (!error) 2940 error = getret; 2941 if (error) { 2942 vrele(vp); 2943 if (cookies) 2944 free((caddr_t)cookies, M_TEMP); 2945 free((caddr_t)rbuf, M_TEMP); 2946 nfsm_reply(NFSX_V3POSTOPATTR); 2947 nfsm_srvpostop_attr(getret, &at); 2948 return (0); 2949 } 2950 if (io.uio_resid) { 2951 siz -= io.uio_resid; 2952 2953 /* 2954 * If nothing read, return eof 2955 * rpc reply 2956 */ 2957 if (siz == 0) { 2958 vrele(vp); 2959 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2960 2 * NFSX_UNSIGNED); 2961 nfsm_srvpostop_attr(getret, &at); 2962 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2963 txdr_hyper(at.va_filerev, tl); 2964 tl += 2; 2965 *tl++ = nfs_false; 2966 *tl = nfs_true; 2967 free((caddr_t)cookies, M_TEMP); 2968 free((caddr_t)rbuf, M_TEMP); 2969 return (0); 2970 } 2971 } 2972 2973 /* 2974 * Check for degenerate cases of nothing useful read. 2975 * If so go try again 2976 */ 2977 cpos = rbuf; 2978 cend = rbuf + siz; 2979 dp = (struct dirent *)cpos; 2980 cookiep = cookies; 2981 2982 while (cpos < cend && ncookies > 0 && 2983 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2984 cpos += dp->d_reclen; 2985 dp = (struct dirent *)cpos; 2986 cookiep++; 2987 ncookies--; 2988 } 2989 if (cpos >= cend || ncookies == 0) { 2990 toff = off; 2991 siz = fullsiz; 2992 goto again; 2993 } 2994 2995 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED; 2996 nfsm_reply(cnt); 2997 nfsm_srvpostop_attr(getret, &at); 2998 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2999 txdr_hyper(at.va_filerev, tl); 3000 mp = mp2 = mb; 3001 bp = bpos; 3002 be = bp + M_TRAILINGSPACE(mp); 3003 3004 /* Loop through the records and build reply */ 3005 while (cpos < cend && ncookies > 0) { 3006 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 3007 nlen = dp->d_namlen; 3008 rem = nfsm_rndup(nlen)-nlen; 3009 3010 /* 3011 * For readdir_and_lookup get the vnode using 3012 * the file number. 3013 */ 3014 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp)) 3015 goto invalid; 3016 memset((caddr_t)nfhp, 0, NFSX_V3FH); 3017 nfhp->fh_fsid = 3018 nvp->v_mount->mnt_stat.f_fsidx; 3019 if (VFS_VPTOFH(nvp, &nfhp->fh_fid)) { 3020 vput(nvp); 3021 goto invalid; 3022 } 3023 if (VOP_GETATTR(nvp, vap, cred, lwp)) { 3024 vput(nvp); 3025 goto invalid; 3026 } 3027 vput(nvp); 3028 3029 /* 3030 * If either the dircount or maxcount will be 3031 * exceeded, get out now. Both of these lengths 3032 * are calculated conservatively, including all 3033 * XDR overheads. 3034 */ 3035 len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH + 3036 NFSX_V3POSTOPATTR); 3037 dirlen += (6 * NFSX_UNSIGNED + nlen + rem); 3038 if (len > cnt || dirlen > fullsiz) { 3039 eofflag = 0; 3040 break; 3041 } 3042 3043 /* 3044 * Build the directory record xdr from 3045 * the dirent entry. 3046 */ 3047 fp = (struct nfs_fattr *)&fl.fl_fattr; 3048 nfsm_srvfillattr(vap, fp); 3049 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH); 3050 fl.fl_fhok = nfs_true; 3051 fl.fl_postopok = nfs_true; 3052 txdr_hyper(*cookiep, fl.fl_off.nfsuquad); 3053 3054 nfsm_clget; 3055 *tl = nfs_true; 3056 bp += NFSX_UNSIGNED; 3057 nfsm_clget; 3058 *tl = 0; 3059 bp += NFSX_UNSIGNED; 3060 nfsm_clget; 3061 *tl = txdr_unsigned(dp->d_fileno); 3062 bp += NFSX_UNSIGNED; 3063 nfsm_clget; 3064 *tl = txdr_unsigned(nlen); 3065 bp += NFSX_UNSIGNED; 3066 3067 /* And loop around copying the name */ 3068 xfer = nlen; 3069 cp = dp->d_name; 3070 while (xfer > 0) { 3071 nfsm_clget; 3072 if ((bp + xfer) > be) 3073 tsiz = be - bp; 3074 else 3075 tsiz = xfer; 3076 memcpy(bp, cp, tsiz); 3077 bp += tsiz; 3078 xfer -= tsiz; 3079 if (xfer > 0) 3080 cp += tsiz; 3081 } 3082 /* And null pad to an int32_t boundary */ 3083 for (i = 0; i < rem; i++) 3084 *bp++ = '\0'; 3085 3086 /* 3087 * Now copy the flrep structure out. 3088 */ 3089 xfer = sizeof (struct flrep); 3090 cp = (caddr_t)&fl; 3091 while (xfer > 0) { 3092 nfsm_clget; 3093 if ((bp + xfer) > be) 3094 tsiz = be - bp; 3095 else 3096 tsiz = xfer; 3097 memcpy(bp, cp, tsiz); 3098 bp += tsiz; 3099 xfer -= tsiz; 3100 if (xfer > 0) 3101 cp += tsiz; 3102 } 3103 } 3104 invalid: 3105 cpos += dp->d_reclen; 3106 dp = (struct dirent *)cpos; 3107 cookiep++; 3108 ncookies--; 3109 } 3110 vrele(vp); 3111 nfsm_clget; 3112 *tl = nfs_false; 3113 bp += NFSX_UNSIGNED; 3114 nfsm_clget; 3115 if (eofflag) 3116 *tl = nfs_true; 3117 else 3118 *tl = nfs_false; 3119 bp += NFSX_UNSIGNED; 3120 if (mp != mb) { 3121 if (bp < be) 3122 mp->m_len = bp - mtod(mp, caddr_t); 3123 } else 3124 mp->m_len += bp - bpos; 3125 free((caddr_t)cookies, M_TEMP); 3126 free((caddr_t)rbuf, M_TEMP); 3127 nfsm_srvdone; 3128 } 3129 3130 /* 3131 * nfs commit service 3132 */ 3133 int 3134 nfsrv_commit(nfsd, slp, lwp, mrq) 3135 struct nfsrv_descript *nfsd; 3136 struct nfssvc_sock *slp; 3137 struct lwp *lwp; 3138 struct mbuf **mrq; 3139 { 3140 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3141 struct mbuf *nam = nfsd->nd_nam; 3142 caddr_t dpos = nfsd->nd_dpos; 3143 struct ucred *cred = &nfsd->nd_cr; 3144 struct vattr bfor, aft; 3145 struct vnode *vp; 3146 nfsfh_t nfh; 3147 fhandle_t *fhp; 3148 u_int32_t *tl; 3149 int32_t t1; 3150 caddr_t bpos; 3151 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cache; 3152 uint32_t cnt; 3153 char *cp2; 3154 struct mbuf *mb, *mreq; 3155 u_quad_t frev, off, end; 3156 struct mount *mp = NULL; 3157 3158 #ifndef nolint 3159 cache = 0; 3160 #endif 3161 fhp = &nfh.fh_generic; 3162 nfsm_srvmtofh(fhp); 3163 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 3164 return (ESTALE); 3165 vn_start_write(NULL, &mp, V_WAIT); 3166 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED); 3167 3168 off = fxdr_hyper(tl); 3169 tl += 2; 3170 cnt = fxdr_unsigned(uint32_t, *tl); 3171 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3172 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3173 if (error) { 3174 nfsm_reply(2 * NFSX_UNSIGNED); 3175 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3176 vn_finished_write(mp, 0); 3177 return (0); 3178 } 3179 for_ret = VOP_GETATTR(vp, &bfor, cred, lwp); 3180 end = (cnt > 0) ? off + cnt : vp->v_size; 3181 if (end < off || end > vp->v_size) 3182 end = vp->v_size; 3183 if (off < vp->v_size) 3184 error = VOP_FSYNC(vp, cred, FSYNC_WAIT, off, end, lwp); 3185 /* else error == 0, from nfsrv_fhtovp() */ 3186 aft_ret = VOP_GETATTR(vp, &aft, cred, lwp); 3187 vput(vp); 3188 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF); 3189 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3190 if (!error) { 3191 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF); 3192 *tl++ = txdr_unsigned(boottime.tv_sec); 3193 *tl = txdr_unsigned(boottime.tv_usec); 3194 } else { 3195 vn_finished_write(mp, 0); 3196 return (0); 3197 } 3198 vn_finished_write(mp, 0); 3199 nfsm_srvdone; 3200 } 3201 3202 /* 3203 * nfs statfs service 3204 */ 3205 int 3206 nfsrv_statfs(nfsd, slp, lwp, mrq) 3207 struct nfsrv_descript *nfsd; 3208 struct nfssvc_sock *slp; 3209 struct lwp *lwp; 3210 struct mbuf **mrq; 3211 { 3212 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3213 struct mbuf *nam = nfsd->nd_nam; 3214 caddr_t dpos = nfsd->nd_dpos; 3215 struct ucred *cred = &nfsd->nd_cr; 3216 struct statvfs *sf; 3217 struct nfs_statfs *sfp; 3218 u_int32_t *tl; 3219 int32_t t1; 3220 caddr_t bpos; 3221 int error = 0, rdonly, cache, getret = 1; 3222 int v3 = (nfsd->nd_flag & ND_NFSV3); 3223 char *cp2; 3224 struct mbuf *mb, *mreq; 3225 struct vnode *vp; 3226 struct vattr at; 3227 nfsfh_t nfh; 3228 fhandle_t *fhp; 3229 struct statvfs statvfs; 3230 u_quad_t frev, tval; 3231 3232 #ifndef nolint 3233 cache = 0; 3234 #endif 3235 fhp = &nfh.fh_generic; 3236 nfsm_srvmtofh(fhp); 3237 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3238 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3239 if (error) { 3240 nfsm_reply(NFSX_UNSIGNED); 3241 nfsm_srvpostop_attr(getret, &at); 3242 return (0); 3243 } 3244 sf = &statvfs; 3245 error = VFS_STATVFS(vp->v_mount, sf, lwp); 3246 getret = VOP_GETATTR(vp, &at, cred, lwp); 3247 vput(vp); 3248 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3)); 3249 if (v3) 3250 nfsm_srvpostop_attr(getret, &at); 3251 if (error) 3252 return (0); 3253 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3)); 3254 if (v3) { 3255 tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_frsize); 3256 txdr_hyper(tval, &sfp->sf_tbytes); 3257 tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_frsize); 3258 txdr_hyper(tval, &sfp->sf_fbytes); 3259 tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_frsize); 3260 txdr_hyper(tval, &sfp->sf_abytes); 3261 tval = (u_quad_t)sf->f_files; 3262 txdr_hyper(tval, &sfp->sf_tfiles); 3263 tval = (u_quad_t)sf->f_ffree; 3264 txdr_hyper(tval, &sfp->sf_ffiles); 3265 txdr_hyper(tval, &sfp->sf_afiles); 3266 sfp->sf_invarsec = 0; 3267 } else { 3268 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA); 3269 sfp->sf_bsize = txdr_unsigned(sf->f_frsize); 3270 sfp->sf_blocks = txdr_unsigned(sf->f_blocks); 3271 sfp->sf_bfree = txdr_unsigned(sf->f_bfree); 3272 sfp->sf_bavail = txdr_unsigned(sf->f_bavail); 3273 } 3274 nfsm_srvdone; 3275 } 3276 3277 /* 3278 * nfs fsinfo service 3279 */ 3280 int 3281 nfsrv_fsinfo(nfsd, slp, lwp, mrq) 3282 struct nfsrv_descript *nfsd; 3283 struct nfssvc_sock *slp; 3284 struct lwp *lwp; 3285 struct mbuf **mrq; 3286 { 3287 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3288 struct mbuf *nam = nfsd->nd_nam; 3289 caddr_t dpos = nfsd->nd_dpos; 3290 struct ucred *cred = &nfsd->nd_cr; 3291 u_int32_t *tl; 3292 struct nfsv3_fsinfo *sip; 3293 int32_t t1; 3294 caddr_t bpos; 3295 int error = 0, rdonly, cache, getret = 1; 3296 uint32_t maxdata; 3297 char *cp2; 3298 struct mbuf *mb, *mreq; 3299 struct vnode *vp; 3300 struct vattr at; 3301 nfsfh_t nfh; 3302 fhandle_t *fhp; 3303 u_quad_t frev, maxfsize; 3304 struct statvfs sb; 3305 3306 #ifndef nolint 3307 cache = 0; 3308 #endif 3309 fhp = &nfh.fh_generic; 3310 nfsm_srvmtofh(fhp); 3311 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3312 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3313 if (error) { 3314 nfsm_reply(NFSX_UNSIGNED); 3315 nfsm_srvpostop_attr(getret, &at); 3316 return (0); 3317 } 3318 3319 /* XXX Try to make a guess on the max file size. */ 3320 VFS_STATVFS(vp->v_mount, &sb, (struct lwp *)0); 3321 maxfsize = (u_quad_t)0x80000000 * sb.f_frsize - 1; 3322 3323 getret = VOP_GETATTR(vp, &at, cred, lwp); 3324 vput(vp); 3325 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO); 3326 nfsm_srvpostop_attr(getret, &at); 3327 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO); 3328 3329 /* 3330 * XXX 3331 * There should be file system VFS OP(s) to get this information. 3332 * For now, assume ufs. 3333 */ 3334 if (slp->ns_so->so_type == SOCK_DGRAM) 3335 maxdata = NFS_MAXDGRAMDATA; 3336 else 3337 maxdata = NFS_MAXDATA; 3338 sip->fs_rtmax = txdr_unsigned(maxdata); 3339 sip->fs_rtpref = txdr_unsigned(maxdata); 3340 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE); 3341 sip->fs_wtmax = txdr_unsigned(maxdata); 3342 sip->fs_wtpref = txdr_unsigned(maxdata); 3343 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE); 3344 sip->fs_dtpref = txdr_unsigned(maxdata); 3345 txdr_hyper(maxfsize, &sip->fs_maxfilesize); 3346 sip->fs_timedelta.nfsv3_sec = 0; 3347 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1); 3348 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK | 3349 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS | 3350 NFSV3FSINFO_CANSETTIME); 3351 nfsm_srvdone; 3352 } 3353 3354 /* 3355 * nfs pathconf service 3356 */ 3357 int 3358 nfsrv_pathconf(nfsd, slp, lwp, mrq) 3359 struct nfsrv_descript *nfsd; 3360 struct nfssvc_sock *slp; 3361 struct lwp *lwp; 3362 struct mbuf **mrq; 3363 { 3364 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3365 struct mbuf *nam = nfsd->nd_nam; 3366 caddr_t dpos = nfsd->nd_dpos; 3367 struct ucred *cred = &nfsd->nd_cr; 3368 u_int32_t *tl; 3369 struct nfsv3_pathconf *pc; 3370 int32_t t1; 3371 caddr_t bpos; 3372 int error = 0, rdonly, cache, getret = 1; 3373 register_t linkmax, namemax, chownres, notrunc; 3374 char *cp2; 3375 struct mbuf *mb, *mreq; 3376 struct vnode *vp; 3377 struct vattr at; 3378 nfsfh_t nfh; 3379 fhandle_t *fhp; 3380 u_quad_t frev; 3381 3382 #ifndef nolint 3383 cache = 0; 3384 #endif 3385 fhp = &nfh.fh_generic; 3386 nfsm_srvmtofh(fhp); 3387 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3388 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3389 if (error) { 3390 nfsm_reply(NFSX_UNSIGNED); 3391 nfsm_srvpostop_attr(getret, &at); 3392 return (0); 3393 } 3394 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax); 3395 if (!error) 3396 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax); 3397 if (!error) 3398 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres); 3399 if (!error) 3400 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc); 3401 getret = VOP_GETATTR(vp, &at, cred, lwp); 3402 vput(vp); 3403 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF); 3404 nfsm_srvpostop_attr(getret, &at); 3405 if (error) 3406 return (0); 3407 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF); 3408 3409 pc->pc_linkmax = txdr_unsigned(linkmax); 3410 pc->pc_namemax = txdr_unsigned(namemax); 3411 pc->pc_notrunc = txdr_unsigned(notrunc); 3412 pc->pc_chownrestricted = txdr_unsigned(chownres); 3413 3414 /* 3415 * These should probably be supported by VOP_PATHCONF(), but 3416 * until msdosfs is exportable (why would you want to?), the 3417 * Unix defaults should be ok. 3418 */ 3419 pc->pc_caseinsensitive = nfs_false; 3420 pc->pc_casepreserving = nfs_true; 3421 nfsm_srvdone; 3422 } 3423 3424 /* 3425 * Null operation, used by clients to ping server 3426 */ 3427 /* ARGSUSED */ 3428 int 3429 nfsrv_null(nfsd, slp, lwp, mrq) 3430 struct nfsrv_descript *nfsd; 3431 struct nfssvc_sock *slp; 3432 struct lwp *lwp; 3433 struct mbuf **mrq; 3434 { 3435 struct mbuf *mrep = nfsd->nd_mrep; 3436 caddr_t bpos; 3437 int error = NFSERR_RETVOID, cache = 0; 3438 struct mbuf *mb, *mreq; 3439 u_quad_t frev; 3440 3441 nfsm_reply(0); 3442 return (0); 3443 } 3444 3445 /* 3446 * No operation, used for obsolete procedures 3447 */ 3448 /* ARGSUSED */ 3449 int 3450 nfsrv_noop(nfsd, slp, lwp, mrq) 3451 struct nfsrv_descript *nfsd; 3452 struct nfssvc_sock *slp; 3453 struct lwp *lwp; 3454 struct mbuf **mrq; 3455 { 3456 struct mbuf *mrep = nfsd->nd_mrep; 3457 caddr_t bpos; 3458 int error, cache = 0; 3459 struct mbuf *mb, *mreq; 3460 u_quad_t frev; 3461 3462 if (nfsd->nd_repstat) 3463 error = nfsd->nd_repstat; 3464 else 3465 error = EPROCUNAVAIL; 3466 nfsm_reply(0); 3467 return (0); 3468 } 3469 3470 /* 3471 * Perform access checking for vnodes obtained from file handles that would 3472 * refer to files already opened by a Unix client. You cannot just use 3473 * vn_writechk() and VOP_ACCESS() for two reasons. 3474 * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case 3475 * 2 - The owner is to be given access irrespective of mode bits for some 3476 * operations, so that processes that chmod after opening a file don't 3477 * break. I don't like this because it opens a security hole, but since 3478 * the nfs server opens a security hole the size of a barn door anyhow, 3479 * what the heck. 3480 * 3481 * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS() 3482 * will return EPERM instead of EACCESS. EPERM is always an error. 3483 */ 3484 int 3485 nfsrv_access(vp, flags, cred, rdonly, lwp, override) 3486 struct vnode *vp; 3487 int flags; 3488 struct ucred *cred; 3489 int rdonly; 3490 struct lwp *lwp; 3491 { 3492 struct vattr vattr; 3493 int error; 3494 if (flags & VWRITE) { 3495 /* Just vn_writechk() changed to check rdonly */ 3496 /* 3497 * Disallow write attempts on read-only file systems; 3498 * unless the file is a socket or a block or character 3499 * device resident on the file system. 3500 */ 3501 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 3502 switch (vp->v_type) { 3503 case VREG: 3504 case VDIR: 3505 case VLNK: 3506 return (EROFS); 3507 default: 3508 break; 3509 } 3510 } 3511 3512 /* 3513 * If the vnode is in use as a process's text, 3514 * we can't allow writing. 3515 */ 3516 if (vp->v_flag & VTEXT) 3517 return (ETXTBSY); 3518 } 3519 error = VOP_GETATTR(vp, &vattr, cred, lwp); 3520 if (error) 3521 return (error); 3522 error = VOP_ACCESS(vp, flags, cred, lwp); 3523 /* 3524 * Allow certain operations for the owner (reads and writes 3525 * on files that are already open). 3526 */ 3527 if (override && error == EACCES && cred->cr_uid == vattr.va_uid) 3528 error = 0; 3529 return error; 3530 } 3531
Indexes created Thu Oct 02 14:10:14 GMT 2025