nfs_serv.c revision 1.101
1 /* $NetBSD: nfs_serv.c,v 1.101 2006/03/01 12:38:32 yamt Exp $ */ 2 3 /* 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * This code is derived from software contributed to Berkeley by 8 * Rick Macklem at The University of Guelph. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * @(#)nfs_serv.c 8.8 (Berkeley) 7/31/95 35 */ 36 37 /* 38 * nfs version 2 and 3 server calls to vnode ops 39 * - these routines generally have 3 phases 40 * 1 - break down and validate rpc request in mbuf list 41 * 2 - do the vnode ops for the request 42 * (surprisingly ?? many are very similar to syscalls in vfs_syscalls.c) 43 * 3 - build the rpc reply in an mbuf list 44 * nb: 45 * - do not mix the phases, since the nfsm_?? macros can return failures 46 * on a bad rpc or similar and do not do any vrele() or vput()'s 47 * 48 * - the nfsm_reply() macro generates an nfs rpc reply with the nfs 49 * error number iff error != 0 whereas 50 * returning an error from the server function implies a fatal error 51 * such as a badly constructed rpc request that should be dropped without 52 * a reply. 53 * For Version 3, nfsm_reply() does not return for the error case, since 54 * most version 3 rpcs return more than the status for error cases. 55 */ 56 57 #include <sys/cdefs.h> 58 __KERNEL_RCSID(0, "$NetBSD: nfs_serv.c,v 1.101 2006/03/01 12:38:32 yamt Exp $"); 59 60 #include <sys/param.h> 61 #include <sys/systm.h> 62 #include <sys/proc.h> 63 #include <sys/file.h> 64 #include <sys/namei.h> 65 #include <sys/vnode.h> 66 #include <sys/mount.h> 67 #include <sys/socket.h> 68 #include <sys/socketvar.h> 69 #include <sys/mbuf.h> 70 #include <sys/dirent.h> 71 #include <sys/stat.h> 72 #include <sys/kernel.h> 73 74 #include <uvm/uvm.h> 75 76 #include <nfs/nfsproto.h> 77 #include <nfs/rpcv2.h> 78 #include <nfs/nfs.h> 79 #include <nfs/xdr_subs.h> 80 #include <nfs/nfsm_subs.h> 81 #include <nfs/nqnfs.h> 82 #include <nfs/nfs_var.h> 83 84 /* Global vars */ 85 extern u_int32_t nfs_xdrneg1; 86 extern u_int32_t nfs_false, nfs_true; 87 extern enum vtype nv3tov_type[8]; 88 extern struct nfsstats nfsstats; 89 extern const nfstype nfsv2_type[9]; 90 extern const nfstype nfsv3_type[9]; 91 int nfsrvw_procrastinate = NFS_GATHERDELAY * 1000; 92 int nfsd_use_loan = 1; /* use page-loan for READ OP */ 93 94 /* 95 * nfs v3 access service 96 */ 97 int 98 nfsrv3_access(nfsd, slp, lwp, mrq) 99 struct nfsrv_descript *nfsd; 100 struct nfssvc_sock *slp; 101 struct lwp *lwp; 102 struct mbuf **mrq; 103 { 104 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 105 struct mbuf *nam = nfsd->nd_nam; 106 caddr_t dpos = nfsd->nd_dpos; 107 struct ucred *cred = &nfsd->nd_cr; 108 struct vnode *vp; 109 nfsfh_t nfh; 110 fhandle_t *fhp; 111 u_int32_t *tl; 112 int32_t t1; 113 caddr_t bpos; 114 int error = 0, rdonly, cache = 0, getret; 115 char *cp2; 116 struct mbuf *mb, *mreq; 117 struct vattr va; 118 u_long inmode, testmode, outmode; 119 u_quad_t frev; 120 121 fhp = &nfh.fh_generic; 122 nfsm_srvmtofh(fhp); 123 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 124 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, 125 (nfsd->nd_flag & ND_KERBAUTH), FALSE); 126 if (error) { 127 nfsm_reply(NFSX_UNSIGNED); 128 nfsm_srvpostop_attr(1, (struct vattr *)0); 129 return (0); 130 } 131 inmode = fxdr_unsigned(u_int32_t, *tl); 132 outmode = 0; 133 if ((inmode & NFSV3ACCESS_READ) && 134 nfsrv_access(vp, VREAD, cred, rdonly, lwp, 0) == 0) 135 outmode |= NFSV3ACCESS_READ; 136 if (vp->v_type != VDIR) { 137 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND); 138 if (testmode && 139 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0) 140 outmode |= testmode; 141 if ((inmode & NFSV3ACCESS_EXECUTE) && 142 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0) 143 outmode |= NFSV3ACCESS_EXECUTE; 144 } else { 145 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND | 146 NFSV3ACCESS_DELETE); 147 if (testmode && 148 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0) 149 outmode |= testmode; 150 if ((inmode & NFSV3ACCESS_LOOKUP) && 151 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0) 152 outmode |= NFSV3ACCESS_LOOKUP; 153 } 154 getret = VOP_GETATTR(vp, &va, cred, lwp); 155 vput(vp); 156 nfsm_reply(NFSX_POSTOPATTR(1) + NFSX_UNSIGNED); 157 nfsm_srvpostop_attr(getret, &va); 158 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 159 *tl = txdr_unsigned(outmode); 160 nfsm_srvdone; 161 } 162 163 /* 164 * nfs getattr service 165 */ 166 int 167 nfsrv_getattr(nfsd, slp, lwp, mrq) 168 struct nfsrv_descript *nfsd; 169 struct nfssvc_sock *slp; 170 struct lwp *lwp; 171 struct mbuf **mrq; 172 { 173 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 174 struct mbuf *nam = nfsd->nd_nam; 175 caddr_t dpos = nfsd->nd_dpos; 176 struct ucred *cred = &nfsd->nd_cr; 177 struct nfs_fattr *fp; 178 struct vattr va; 179 struct vnode *vp; 180 nfsfh_t nfh; 181 fhandle_t *fhp; 182 u_int32_t *tl; 183 int32_t t1; 184 caddr_t bpos; 185 int error = 0, rdonly, cache; 186 char *cp2; 187 struct mbuf *mb, *mreq; 188 u_quad_t frev; 189 190 fhp = &nfh.fh_generic; 191 nfsm_srvmtofh(fhp); 192 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, 193 (nfsd->nd_flag & ND_KERBAUTH), FALSE); 194 if (error) { 195 nfsm_reply(0); 196 return (0); 197 } 198 nqsrv_getl(vp, ND_READ); 199 error = VOP_GETATTR(vp, &va, cred, lwp); 200 vput(vp); 201 nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); 202 if (error) 203 return (0); 204 nfsm_build(fp, struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); 205 nfsm_srvfillattr(&va, fp); 206 nfsm_srvdone; 207 } 208 209 /* 210 * nfs setattr service 211 */ 212 int 213 nfsrv_setattr(nfsd, slp, lwp, mrq) 214 struct nfsrv_descript *nfsd; 215 struct nfssvc_sock *slp; 216 struct lwp *lwp; 217 struct mbuf **mrq; 218 { 219 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 220 struct mbuf *nam = nfsd->nd_nam; 221 caddr_t dpos = nfsd->nd_dpos; 222 struct ucred *cred = &nfsd->nd_cr; 223 struct vattr va, preat; 224 struct nfsv2_sattr *sp; 225 struct nfs_fattr *fp; 226 struct vnode *vp; 227 nfsfh_t nfh; 228 fhandle_t *fhp; 229 u_int32_t *tl; 230 int32_t t1; 231 caddr_t bpos; 232 int error = 0, rdonly, cache, preat_ret = 1, postat_ret = 1; 233 int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0; 234 char *cp2; 235 struct mbuf *mb, *mreq; 236 u_quad_t frev; 237 struct timespec guard; 238 struct mount *mp = NULL; 239 240 fhp = &nfh.fh_generic; 241 nfsm_srvmtofh(fhp); 242 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 243 return (ESTALE); 244 vn_start_write(NULL, &mp, V_WAIT); 245 VATTR_NULL(&va); 246 if (v3) { 247 nfsm_srvsattr(&va); 248 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 249 gcheck = fxdr_unsigned(int, *tl); 250 if (gcheck) { 251 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 252 fxdr_nfsv3time(tl, &guard); 253 } 254 } else { 255 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 256 /* 257 * Nah nah nah nah na nah 258 * There is a bug in the Sun client that puts 0xffff in the mode 259 * field of sattr when it should put in 0xffffffff. The u_short 260 * doesn't sign extend. 261 * --> check the low order 2 bytes for 0xffff 262 */ 263 if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff) 264 va.va_mode = nfstov_mode(sp->sa_mode); 265 if (sp->sa_uid != nfs_xdrneg1) 266 va.va_uid = fxdr_unsigned(uid_t, sp->sa_uid); 267 if (sp->sa_gid != nfs_xdrneg1) 268 va.va_gid = fxdr_unsigned(gid_t, sp->sa_gid); 269 if (sp->sa_size != nfs_xdrneg1) 270 va.va_size = fxdr_unsigned(u_quad_t, sp->sa_size); 271 if (sp->sa_atime.nfsv2_sec != nfs_xdrneg1) { 272 #ifdef notyet 273 fxdr_nfsv2time(&sp->sa_atime, &va.va_atime); 274 #else 275 va.va_atime.tv_sec = 276 fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec); 277 va.va_atime.tv_nsec = 0; 278 #endif 279 } 280 if (sp->sa_mtime.nfsv2_sec != nfs_xdrneg1) 281 fxdr_nfsv2time(&sp->sa_mtime, &va.va_mtime); 282 283 } 284 285 /* 286 * Now that we have all the fields, lets do it. 287 */ 288 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, 289 (nfsd->nd_flag & ND_KERBAUTH), FALSE); 290 if (error) { 291 nfsm_reply(2 * NFSX_UNSIGNED); 292 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 293 vn_finished_write(mp, 0); 294 return (0); 295 } 296 nqsrv_getl(vp, ND_WRITE); 297 if (v3) { 298 error = preat_ret = VOP_GETATTR(vp, &preat, cred, lwp); 299 if (!error && gcheck && 300 (preat.va_ctime.tv_sec != guard.tv_sec || 301 preat.va_ctime.tv_nsec != guard.tv_nsec)) 302 error = NFSERR_NOT_SYNC; 303 if (error) { 304 vput(vp); 305 nfsm_reply(NFSX_WCCDATA(v3)); 306 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 307 vn_finished_write(mp, 0); 308 return (0); 309 } 310 } 311 312 /* 313 * If the size is being changed write acces is required, otherwise 314 * just check for a read only file system. 315 */ 316 if (va.va_size == ((u_quad_t)((quad_t) -1))) { 317 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 318 error = EROFS; 319 goto out; 320 } 321 } else { 322 if (vp->v_type == VDIR) { 323 error = EISDIR; 324 goto out; 325 } else if ((error = nfsrv_access(vp, VWRITE, cred, rdonly, 326 lwp, 0)) != 0) 327 goto out; 328 } 329 error = VOP_SETATTR(vp, &va, cred, lwp); 330 postat_ret = VOP_GETATTR(vp, &va, cred, lwp); 331 if (!error) 332 error = postat_ret; 333 out: 334 vput(vp); 335 vn_finished_write(mp, 0); 336 nfsm_reply(NFSX_WCCORFATTR(v3)); 337 if (v3) { 338 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 339 return (0); 340 } else { 341 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 342 nfsm_srvfillattr(&va, fp); 343 } 344 nfsm_srvdone; 345 } 346 347 /* 348 * nfs lookup rpc 349 */ 350 int 351 nfsrv_lookup(nfsd, slp, lwp, mrq) 352 struct nfsrv_descript *nfsd; 353 struct nfssvc_sock *slp; 354 struct lwp *lwp; 355 struct mbuf **mrq; 356 { 357 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 358 struct mbuf *nam = nfsd->nd_nam; 359 caddr_t dpos = nfsd->nd_dpos; 360 struct ucred *cred = &nfsd->nd_cr; 361 struct nfs_fattr *fp; 362 struct nameidata nd, ind, *ndp = &nd; 363 struct vnode *vp, *dirp; 364 nfsfh_t nfh; 365 fhandle_t *fhp; 366 caddr_t cp; 367 u_int32_t *tl; 368 int32_t t1; 369 caddr_t bpos; 370 int error = 0, cache, dirattr_ret = 1; 371 uint32_t len; 372 int v3 = (nfsd->nd_flag & ND_NFSV3), pubflag; 373 char *cp2; 374 struct mbuf *mb, *mreq; 375 struct vattr va, dirattr; 376 u_quad_t frev; 377 378 fhp = &nfh.fh_generic; 379 nfsm_srvmtofh(fhp); 380 nfsm_srvnamesiz(len); 381 382 pubflag = nfs_ispublicfh(fhp); 383 384 nd.ni_cnd.cn_cred = cred; 385 nd.ni_cnd.cn_nameiop = LOOKUP; 386 nd.ni_cnd.cn_flags = LOCKLEAF | SAVESTART; 387 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 388 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), pubflag); 389 390 if (!error && pubflag) { 391 if (nd.ni_vp->v_type == VDIR && nfs_pub.np_index != NULL) { 392 /* 393 * Setup call to lookup() to see if we can find 394 * the index file. Arguably, this doesn't belong 395 * in a kernel.. Ugh. 396 */ 397 ind = nd; 398 VOP_UNLOCK(nd.ni_vp, 0); 399 ind.ni_pathlen = strlen(nfs_pub.np_index); 400 ind.ni_cnd.cn_nameptr = ind.ni_cnd.cn_pnbuf = 401 nfs_pub.np_index; 402 ind.ni_startdir = nd.ni_vp; 403 VREF(ind.ni_startdir); 404 error = lookup(&ind); 405 if (!error) { 406 /* 407 * Found an index file. Get rid of 408 * the old references. 409 */ 410 if (dirp) 411 vrele(dirp); 412 dirp = nd.ni_vp; 413 vrele(nd.ni_startdir); 414 ndp = &ind; 415 } else 416 error = 0; 417 } 418 /* 419 * If the public filehandle was used, check that this lookup 420 * didn't result in a filehandle outside the publicly exported 421 * filesystem. 422 */ 423 424 if (!error && ndp->ni_vp->v_mount != nfs_pub.np_mount) { 425 vput(nd.ni_vp); 426 error = EPERM; 427 } 428 } 429 430 if (dirp) { 431 if (v3) 432 dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred, lwp); 433 vrele(dirp); 434 } 435 436 if (error) { 437 nfsm_reply(NFSX_POSTOPATTR(v3)); 438 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 439 return (0); 440 } 441 442 nqsrv_getl(ndp->ni_startdir, ND_READ); 443 vrele(ndp->ni_startdir); 444 PNBUF_PUT(nd.ni_cnd.cn_pnbuf); 445 vp = ndp->ni_vp; 446 memset((caddr_t)fhp, 0, sizeof(nfh)); 447 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 448 error = VFS_VPTOFH(vp, &fhp->fh_fid); 449 if (!error) 450 error = VOP_GETATTR(vp, &va, cred, lwp); 451 vput(vp); 452 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPORFATTR(v3) + NFSX_POSTOPATTR(v3)); 453 if (error) { 454 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 455 return (0); 456 } 457 nfsm_srvfhtom(fhp, v3); 458 if (v3) { 459 nfsm_srvpostop_attr(0, &va); 460 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 461 } else { 462 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 463 nfsm_srvfillattr(&va, fp); 464 } 465 nfsm_srvdone; 466 } 467 468 /* 469 * nfs readlink service 470 */ 471 int 472 nfsrv_readlink(nfsd, slp, lwp, mrq) 473 struct nfsrv_descript *nfsd; 474 struct nfssvc_sock *slp; 475 struct lwp *lwp; 476 struct mbuf **mrq; 477 { 478 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 479 struct mbuf *nam = nfsd->nd_nam; 480 caddr_t dpos = nfsd->nd_dpos; 481 struct ucred *cred = &nfsd->nd_cr; 482 struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN]; 483 struct iovec *ivp = iv; 484 struct mbuf *mp; 485 u_int32_t *tl; 486 int32_t t1; 487 caddr_t bpos; 488 int error = 0, rdonly, cache, i, padlen, getret; 489 uint32_t len; 490 int v3 = (nfsd->nd_flag & ND_NFSV3); 491 char *cp2; 492 struct mbuf *mb, *mp2 = NULL, *mp3 = NULL, *mreq; 493 struct vnode *vp; 494 struct vattr attr; 495 nfsfh_t nfh; 496 fhandle_t *fhp; 497 struct uio io, *uiop = &io; 498 u_quad_t frev; 499 500 fhp = &nfh.fh_generic; 501 nfsm_srvmtofh(fhp); 502 len = 0; 503 i = 0; 504 while (len < NFS_MAXPATHLEN) { 505 mp = m_get(M_WAIT, MT_DATA); 506 MCLAIM(mp, &nfs_mowner); 507 m_clget(mp, M_WAIT); 508 mp->m_len = NFSMSIZ(mp); 509 if (len == 0) 510 mp3 = mp2 = mp; 511 else { 512 mp2->m_next = mp; 513 mp2 = mp; 514 } 515 if ((len+mp->m_len) > NFS_MAXPATHLEN) { 516 mp->m_len = NFS_MAXPATHLEN-len; 517 len = NFS_MAXPATHLEN; 518 } else 519 len += mp->m_len; 520 ivp->iov_base = mtod(mp, caddr_t); 521 ivp->iov_len = mp->m_len; 522 i++; 523 ivp++; 524 } 525 uiop->uio_iov = iv; 526 uiop->uio_iovcnt = i; 527 uiop->uio_offset = 0; 528 uiop->uio_resid = len; 529 uiop->uio_rw = UIO_READ; 530 UIO_SETUP_SYSSPACE(uiop); 531 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 532 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 533 if (error) { 534 m_freem(mp3); 535 nfsm_reply(2 * NFSX_UNSIGNED); 536 nfsm_srvpostop_attr(1, (struct vattr *)0); 537 return (0); 538 } 539 if (vp->v_type != VLNK) { 540 if (v3) 541 error = EINVAL; 542 else 543 error = ENXIO; 544 goto out; 545 } 546 nqsrv_getl(vp, ND_READ); 547 error = VOP_READLINK(vp, uiop, cred); 548 out: 549 getret = VOP_GETATTR(vp, &attr, cred, lwp); 550 vput(vp); 551 if (error) 552 m_freem(mp3); 553 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED); 554 if (v3) { 555 nfsm_srvpostop_attr(getret, &attr); 556 if (error) 557 return (0); 558 } 559 len -= uiop->uio_resid; 560 padlen = nfsm_padlen(len); 561 if (uiop->uio_resid || padlen) 562 nfs_zeropad(mp3, uiop->uio_resid, padlen); 563 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 564 *tl = txdr_unsigned(len); 565 mb->m_next = mp3; 566 nfsm_srvdone; 567 } 568 569 /* 570 * nfs read service 571 */ 572 int 573 nfsrv_read(nfsd, slp, lwp, mrq) 574 struct nfsrv_descript *nfsd; 575 struct nfssvc_sock *slp; 576 struct lwp *lwp; 577 struct mbuf **mrq; 578 { 579 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 580 struct mbuf *nam = nfsd->nd_nam; 581 caddr_t dpos = nfsd->nd_dpos; 582 struct ucred *cred = &nfsd->nd_cr; 583 struct mbuf *m; 584 struct nfs_fattr *fp; 585 u_int32_t *tl; 586 int32_t t1; 587 int i; 588 caddr_t bpos; 589 int error = 0, rdonly, cache, getret; 590 int v3 = (nfsd->nd_flag & ND_NFSV3); 591 uint32_t reqlen, len, cnt, left; 592 int padlen; 593 char *cp2; 594 struct mbuf *mb, *mreq; 595 struct vnode *vp; 596 nfsfh_t nfh; 597 fhandle_t *fhp; 598 struct uio io, *uiop = &io; 599 struct vattr va; 600 off_t off; 601 u_quad_t frev; 602 603 fhp = &nfh.fh_generic; 604 nfsm_srvmtofh(fhp); 605 if (v3) { 606 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 607 off = fxdr_hyper(tl); 608 } else { 609 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 610 off = (off_t)fxdr_unsigned(u_int32_t, *tl); 611 } 612 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 613 reqlen = fxdr_unsigned(uint32_t, *tl); 614 reqlen = MIN(reqlen, NFS_SRVMAXDATA(nfsd)); 615 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 616 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 617 if (error) { 618 nfsm_reply(2 * NFSX_UNSIGNED); 619 nfsm_srvpostop_attr(1, (struct vattr *)0); 620 return (0); 621 } 622 if (vp->v_type != VREG) { 623 if (v3) 624 error = EINVAL; 625 else 626 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 627 } 628 if (!error) { 629 nqsrv_getl(vp, ND_READ); 630 if ((error = nfsrv_access(vp, VREAD, cred, rdonly, lwp, 1)) != 0) 631 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 1); 632 } 633 getret = VOP_GETATTR(vp, &va, cred, lwp); 634 if (!error) 635 error = getret; 636 if (error) { 637 vput(vp); 638 nfsm_reply(NFSX_POSTOPATTR(v3)); 639 nfsm_srvpostop_attr(getret, &va); 640 return (0); 641 } 642 if (off >= va.va_size) 643 cnt = 0; 644 else if ((off + reqlen) > va.va_size) 645 cnt = va.va_size - off; 646 else 647 cnt = reqlen; 648 nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt)); 649 if (v3) { 650 nfsm_build(tl, u_int32_t *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED); 651 *tl++ = nfs_true; 652 fp = (struct nfs_fattr *)tl; 653 tl += (NFSX_V3FATTR / sizeof (u_int32_t)); 654 } else { 655 nfsm_build(tl, u_int32_t *, NFSX_V2FATTR + NFSX_UNSIGNED); 656 fp = (struct nfs_fattr *)tl; 657 tl += (NFSX_V2FATTR / sizeof (u_int32_t)); 658 } 659 len = left = cnt; 660 if (cnt > 0) { 661 if (nfsd_use_loan) { 662 struct vm_page **pgpp; 663 voff_t pgoff = trunc_page(off); 664 int npages; 665 vaddr_t lva; 666 667 npages = (round_page(off + cnt) - pgoff) >> PAGE_SHIFT; 668 KASSERT(npages <= M_EXT_MAXPAGES); /* XXX */ 669 670 /* allocate kva for mbuf data */ 671 lva = sokvaalloc(npages << PAGE_SHIFT, slp->ns_so); 672 if (lva == 0) { 673 /* fall back to VOP_READ */ 674 goto loan_fail; 675 } 676 677 /* allocate mbuf */ 678 m = m_get(M_WAIT, MT_DATA); 679 MCLAIM(m, &nfs_mowner); 680 pgpp = m->m_ext.ext_pgs; 681 682 /* loan pages */ 683 error = uvm_loanuobjpages(&vp->v_uobj, pgoff, npages, 684 pgpp); 685 if (error) { 686 sokvafree(lva, npages << PAGE_SHIFT); 687 m_free(m); 688 if (error == EBUSY) 689 goto loan_fail; 690 goto read_error; 691 } 692 693 /* associate kva to mbuf */ 694 MEXTADD(m, (void *)(lva + ((vaddr_t)off & PAGE_MASK)), 695 cnt, M_MBUF, soloanfree, slp->ns_so); 696 m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP; 697 m->m_len = cnt; 698 699 /* map pages */ 700 for (i = 0; i < npages; i++) { 701 pmap_kenter_pa(lva, VM_PAGE_TO_PHYS(pgpp[i]), 702 VM_PROT_READ); 703 lva += PAGE_SIZE; 704 } 705 706 pmap_update(pmap_kernel()); 707 708 mb->m_next = m; 709 mb = m; 710 error = 0; 711 uiop->uio_resid = 0; 712 } else { 713 struct iovec *iv; 714 struct iovec *iv2; 715 struct mbuf *m2; 716 int siz; 717 loan_fail: 718 /* 719 * Generate the mbuf list with the uio_iov ref. to it. 720 */ 721 i = 0; 722 m = m2 = mb; 723 while (left > 0) { 724 siz = min(M_TRAILINGSPACE(m), left); 725 if (siz > 0) { 726 left -= siz; 727 i++; 728 } 729 if (left > 0) { 730 m = m_get(M_WAIT, MT_DATA); 731 MCLAIM(m, &nfs_mowner); 732 m_clget(m, M_WAIT); 733 m->m_len = 0; 734 m2->m_next = m; 735 m2 = m; 736 } 737 } 738 iv = malloc(i * sizeof(struct iovec), M_TEMP, M_WAITOK); 739 uiop->uio_iov = iv2 = iv; 740 m = mb; 741 left = cnt; 742 i = 0; 743 while (left > 0) { 744 if (m == NULL) 745 panic("nfsrv_read iov"); 746 siz = min(M_TRAILINGSPACE(m), left); 747 if (siz > 0) { 748 iv->iov_base = mtod(m, caddr_t) + 749 m->m_len; 750 iv->iov_len = siz; 751 m->m_len += siz; 752 left -= siz; 753 iv++; 754 i++; 755 } 756 m = m->m_next; 757 } 758 uiop->uio_iovcnt = i; 759 uiop->uio_offset = off; 760 uiop->uio_resid = cnt; 761 uiop->uio_rw = UIO_READ; 762 UIO_SETUP_SYSSPACE(uiop); 763 error = VOP_READ(vp, uiop, IO_NODELOCKED, cred); 764 free((caddr_t)iv2, M_TEMP); 765 } 766 read_error: 767 if (error || (getret = VOP_GETATTR(vp, &va, cred, lwp)) != 0){ 768 if (!error) 769 error = getret; 770 m_freem(mreq); 771 vput(vp); 772 nfsm_reply(NFSX_POSTOPATTR(v3)); 773 nfsm_srvpostop_attr(getret, &va); 774 return (0); 775 } 776 } else { 777 uiop->uio_resid = 0; 778 } 779 vput(vp); 780 nfsm_srvfillattr(&va, fp); 781 len -= uiop->uio_resid; 782 padlen = nfsm_padlen(len); 783 if (uiop->uio_resid || padlen) 784 nfs_zeropad(mb, uiop->uio_resid, padlen); 785 if (v3) { 786 /* count */ 787 *tl++ = txdr_unsigned(len); 788 /* eof */ 789 if (off + len >= va.va_size) 790 *tl++ = nfs_true; 791 else 792 *tl++ = nfs_false; 793 } 794 *tl = txdr_unsigned(len); 795 nfsm_srvdone; 796 } 797 798 /* 799 * nfs write service 800 */ 801 int 802 nfsrv_write(nfsd, slp, lwp, mrq) 803 struct nfsrv_descript *nfsd; 804 struct nfssvc_sock *slp; 805 struct lwp *lwp; 806 struct mbuf **mrq; 807 { 808 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 809 struct mbuf *nam = nfsd->nd_nam; 810 caddr_t dpos = nfsd->nd_dpos; 811 struct ucred *cred = &nfsd->nd_cr; 812 struct iovec *ivp; 813 int i, cnt; 814 struct mbuf *mp; 815 struct nfs_fattr *fp; 816 struct iovec *iv; 817 struct vattr va, forat; 818 u_int32_t *tl; 819 int32_t t1; 820 caddr_t bpos; 821 int error = 0, rdonly, cache, len, forat_ret = 1; 822 int ioflags, aftat_ret = 1, retlen, zeroing, adjust; 823 int stable = NFSV3WRITE_FILESYNC; 824 int v3 = (nfsd->nd_flag & ND_NFSV3); 825 char *cp2; 826 struct mbuf *mb, *mreq; 827 struct vnode *vp; 828 nfsfh_t nfh; 829 fhandle_t *fhp; 830 struct uio io, *uiop = &io; 831 off_t off; 832 u_quad_t frev; 833 struct mount *mntp = NULL; 834 835 if (mrep == NULL) { 836 *mrq = NULL; 837 return (0); 838 } 839 fhp = &nfh.fh_generic; 840 nfsm_srvmtofh(fhp); 841 if ((mntp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 842 return (ESTALE); 843 vn_start_write(NULL, &mntp, V_WAIT); 844 if (v3) { 845 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 846 off = fxdr_hyper(tl); 847 tl += 3; 848 stable = fxdr_unsigned(int, *tl++); 849 } else { 850 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 851 off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 852 tl += 2; 853 } 854 retlen = len = fxdr_unsigned(int32_t, *tl); 855 cnt = i = 0; 856 857 /* 858 * For NFS Version 2, it is not obvious what a write of zero length 859 * should do, but I might as well be consistent with Version 3, 860 * which is to return ok so long as there are no permission problems. 861 */ 862 if (len > 0) { 863 zeroing = 1; 864 mp = mrep; 865 while (mp) { 866 if (mp == md) { 867 zeroing = 0; 868 adjust = dpos - mtod(mp, caddr_t); 869 mp->m_len -= adjust; 870 if (mp->m_len > 0 && adjust > 0) 871 NFSMADV(mp, adjust); 872 } 873 if (zeroing) 874 mp->m_len = 0; 875 else if (mp->m_len > 0) { 876 i += mp->m_len; 877 if (i > len) { 878 mp->m_len -= (i - len); 879 zeroing = 1; 880 } 881 if (mp->m_len > 0) 882 cnt++; 883 } 884 mp = mp->m_next; 885 } 886 } 887 if (len > NFS_MAXDATA || len < 0 || i < len) { 888 error = EIO; 889 nfsm_reply(2 * NFSX_UNSIGNED); 890 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 891 vn_finished_write(mntp, 0); 892 return (0); 893 } 894 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 895 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 896 if (error) { 897 nfsm_reply(2 * NFSX_UNSIGNED); 898 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 899 vn_finished_write(mntp, 0); 900 return (0); 901 } 902 if (v3) 903 forat_ret = VOP_GETATTR(vp, &forat, cred, lwp); 904 if (vp->v_type != VREG) { 905 if (v3) 906 error = EINVAL; 907 else 908 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 909 } 910 if (!error) { 911 nqsrv_getl(vp, ND_WRITE); 912 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1); 913 } 914 if (error) { 915 vput(vp); 916 nfsm_reply(NFSX_WCCDATA(v3)); 917 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 918 vn_finished_write(mntp, 0); 919 return (0); 920 } 921 922 if (len > 0) { 923 ivp = malloc(cnt * sizeof (struct iovec), M_TEMP, M_WAITOK); 924 uiop->uio_iov = iv = ivp; 925 uiop->uio_iovcnt = cnt; 926 mp = mrep; 927 while (mp) { 928 if (mp->m_len > 0) { 929 ivp->iov_base = mtod(mp, caddr_t); 930 ivp->iov_len = mp->m_len; 931 ivp++; 932 } 933 mp = mp->m_next; 934 } 935 936 /* 937 * XXX 938 * The IO_METASYNC flag indicates that all metadata (and not 939 * just enough to ensure data integrity) must be written to 940 * stable storage synchronously. 941 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.) 942 */ 943 if (stable == NFSV3WRITE_UNSTABLE) 944 ioflags = IO_NODELOCKED; 945 else if (stable == NFSV3WRITE_DATASYNC) 946 ioflags = (IO_SYNC | IO_NODELOCKED); 947 else 948 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 949 uiop->uio_resid = len; 950 uiop->uio_rw = UIO_WRITE; 951 uiop->uio_offset = off; 952 UIO_SETUP_SYSSPACE(uiop); 953 error = VOP_WRITE(vp, uiop, ioflags, cred); 954 nfsstats.srvvop_writes++; 955 free(iv, M_TEMP); 956 } 957 aftat_ret = VOP_GETATTR(vp, &va, cred, lwp); 958 vput(vp); 959 vn_finished_write(mntp, 0); 960 if (!error) 961 error = aftat_ret; 962 nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) + 963 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3)); 964 if (v3) { 965 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 966 if (error) 967 return (0); 968 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 969 *tl++ = txdr_unsigned(retlen); 970 if (stable == NFSV3WRITE_UNSTABLE) 971 *tl++ = txdr_unsigned(stable); 972 else 973 *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC); 974 /* 975 * Actually, there is no need to txdr these fields, 976 * but it may make the values more human readable, 977 * for debugging purposes. 978 */ 979 *tl++ = txdr_unsigned(boottime.tv_sec); 980 *tl = txdr_unsigned(boottime.tv_usec); 981 } else { 982 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 983 nfsm_srvfillattr(&va, fp); 984 } 985 nfsm_srvdone; 986 } 987 988 /* 989 * NFS write service with write gathering support. Called when 990 * nfsrvw_procrastinate > 0. 991 * See: Chet Juszczak, "Improving the Write Performance of an NFS Server", 992 * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco, 993 * Jan. 1994. 994 */ 995 int 996 nfsrv_writegather(ndp, slp, lwp, mrq) 997 struct nfsrv_descript **ndp; 998 struct nfssvc_sock *slp; 999 struct lwp *lwp; 1000 struct mbuf **mrq; 1001 { 1002 struct iovec *ivp; 1003 struct mbuf *mp; 1004 struct nfsrv_descript *wp, *nfsd, *owp, *swp; 1005 struct nfs_fattr *fp; 1006 int i = 0; 1007 struct iovec *iov; 1008 struct nfsrvw_delayhash *wpp; 1009 struct ucred *cred; 1010 struct vattr va, forat; 1011 u_int32_t *tl; 1012 int32_t t1; 1013 caddr_t bpos, dpos; 1014 int error = 0, rdonly, cache, len = 0, forat_ret = 1; 1015 int ioflags, aftat_ret = 1, s, adjust, v3, zeroing; 1016 char *cp2; 1017 struct mbuf *mb, *mreq, *mrep, *md; 1018 struct vnode *vp; 1019 struct uio io, *uiop = &io; 1020 u_quad_t frev, cur_usec; 1021 struct mount *mntp = NULL; 1022 1023 *mrq = NULL; 1024 if (*ndp) { 1025 nfsd = *ndp; 1026 *ndp = NULL; 1027 mrep = nfsd->nd_mrep; 1028 md = nfsd->nd_md; 1029 dpos = nfsd->nd_dpos; 1030 cred = &nfsd->nd_cr; 1031 v3 = (nfsd->nd_flag & ND_NFSV3); 1032 LIST_INIT(&nfsd->nd_coalesce); 1033 nfsd->nd_mreq = NULL; 1034 nfsd->nd_stable = NFSV3WRITE_FILESYNC; 1035 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec; 1036 nfsd->nd_time = cur_usec + nfsrvw_procrastinate; 1037 1038 /* 1039 * Now, get the write header.. 1040 */ 1041 nfsm_srvmtofh(&nfsd->nd_fh); 1042 if (v3) { 1043 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 1044 nfsd->nd_off = fxdr_hyper(tl); 1045 tl += 3; 1046 nfsd->nd_stable = fxdr_unsigned(int, *tl++); 1047 } else { 1048 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1049 nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 1050 tl += 2; 1051 } 1052 len = fxdr_unsigned(int32_t, *tl); 1053 nfsd->nd_len = len; 1054 nfsd->nd_eoff = nfsd->nd_off + len; 1055 1056 /* 1057 * Trim the header out of the mbuf list and trim off any trailing 1058 * junk so that the mbuf list has only the write data. 1059 */ 1060 zeroing = 1; 1061 i = 0; 1062 mp = mrep; 1063 while (mp) { 1064 if (mp == md) { 1065 zeroing = 0; 1066 adjust = dpos - mtod(mp, caddr_t); 1067 mp->m_len -= adjust; 1068 if (mp->m_len > 0 && adjust > 0) 1069 NFSMADV(mp, adjust); 1070 } 1071 if (zeroing) 1072 mp->m_len = 0; 1073 else { 1074 i += mp->m_len; 1075 if (i > len) { 1076 mp->m_len -= (i - len); 1077 zeroing = 1; 1078 } 1079 } 1080 mp = mp->m_next; 1081 } 1082 if (len > NFS_MAXDATA || len < 0 || i < len) { 1083 nfsmout: 1084 m_freem(mrep); 1085 error = EIO; 1086 nfsm_writereply(2 * NFSX_UNSIGNED, v3); 1087 if (v3) 1088 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1089 nfsd->nd_mreq = mreq; 1090 nfsd->nd_mrep = NULL; 1091 nfsd->nd_time = 0; 1092 } 1093 1094 /* 1095 * Add this entry to the hash and time queues. 1096 */ 1097 s = splsoftclock(); 1098 owp = NULL; 1099 wp = LIST_FIRST(&slp->ns_tq); 1100 while (wp && wp->nd_time < nfsd->nd_time) { 1101 owp = wp; 1102 wp = LIST_NEXT(wp, nd_tq); 1103 } 1104 if (owp) { 1105 LIST_INSERT_AFTER(owp, nfsd, nd_tq); 1106 } else { 1107 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1108 } 1109 if (nfsd->nd_mrep) { 1110 wpp = NWDELAYHASH(slp, nfsd->nd_fh.fh_fid.fid_data); 1111 owp = NULL; 1112 wp = LIST_FIRST(wpp); 1113 while (wp && 1114 memcmp(&nfsd->nd_fh, &wp->nd_fh, NFSX_V3FH)) { 1115 owp = wp; 1116 wp = LIST_NEXT(wp, nd_hash); 1117 } 1118 while (wp && wp->nd_off < nfsd->nd_off && 1119 !memcmp(&nfsd->nd_fh, &wp->nd_fh, NFSX_V3FH)) { 1120 owp = wp; 1121 wp = LIST_NEXT(wp, nd_hash); 1122 } 1123 if (owp) { 1124 LIST_INSERT_AFTER(owp, nfsd, nd_hash); 1125 1126 /* 1127 * Search the hash list for overlapping entries and 1128 * coalesce. 1129 */ 1130 for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) { 1131 wp = LIST_NEXT(nfsd, nd_hash); 1132 if (NFSW_SAMECRED(owp, nfsd)) 1133 nfsrvw_coalesce(owp, nfsd); 1134 } 1135 } else { 1136 LIST_INSERT_HEAD(wpp, nfsd, nd_hash); 1137 } 1138 } 1139 splx(s); 1140 } 1141 1142 /* 1143 * Now, do VOP_WRITE()s for any one(s) that need to be done now 1144 * and generate the associated reply mbuf list(s). 1145 */ 1146 loop1: 1147 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec; 1148 s = splsoftclock(); 1149 for (nfsd = LIST_FIRST(&slp->ns_tq); nfsd; nfsd = owp) { 1150 owp = LIST_NEXT(nfsd, nd_tq); 1151 if (nfsd->nd_time > cur_usec) 1152 break; 1153 if (nfsd->nd_mreq) 1154 continue; 1155 LIST_REMOVE(nfsd, nd_tq); 1156 LIST_REMOVE(nfsd, nd_hash); 1157 splx(s); 1158 mrep = nfsd->nd_mrep; 1159 nfsd->nd_mrep = NULL; 1160 cred = &nfsd->nd_cr; 1161 v3 = (nfsd->nd_flag & ND_NFSV3); 1162 forat_ret = aftat_ret = 1; 1163 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp, 1164 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH), 1165 FALSE); 1166 if (!error) { 1167 if (v3) 1168 forat_ret = VOP_GETATTR(vp, &forat, cred, lwp); 1169 if (vp->v_type != VREG) { 1170 if (v3) 1171 error = EINVAL; 1172 else 1173 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 1174 } 1175 } else 1176 vp = NULL; 1177 if (!error) { 1178 nqsrv_getl(vp, ND_WRITE); 1179 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1); 1180 } 1181 1182 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE) 1183 ioflags = IO_NODELOCKED; 1184 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC) 1185 ioflags = (IO_SYNC | IO_NODELOCKED); 1186 else 1187 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 1188 uiop->uio_rw = UIO_WRITE; 1189 uiop->uio_offset = nfsd->nd_off; 1190 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off; 1191 UIO_SETUP_SYSSPACE(uiop); 1192 if (uiop->uio_resid > 0) { 1193 mp = mrep; 1194 i = 0; 1195 while (mp) { 1196 if (mp->m_len > 0) 1197 i++; 1198 mp = mp->m_next; 1199 } 1200 uiop->uio_iovcnt = i; 1201 iov = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK); 1202 uiop->uio_iov = ivp = iov; 1203 mp = mrep; 1204 while (mp) { 1205 if (mp->m_len > 0) { 1206 ivp->iov_base = mtod(mp, caddr_t); 1207 ivp->iov_len = mp->m_len; 1208 ivp++; 1209 } 1210 mp = mp->m_next; 1211 } 1212 if (!error) { 1213 if (vn_start_write(vp, &mntp, V_NOWAIT) != 0) { 1214 VOP_UNLOCK(vp, 0); 1215 vn_start_write(NULL, &mntp, V_WAIT); 1216 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1217 } 1218 if (!error) { 1219 error = VOP_WRITE(vp, uiop, ioflags, cred); 1220 nfsstats.srvvop_writes++; 1221 vn_finished_write(mntp, 0); 1222 } 1223 } 1224 free((caddr_t)iov, M_TEMP); 1225 } 1226 m_freem(mrep); 1227 if (vp) { 1228 aftat_ret = VOP_GETATTR(vp, &va, cred, lwp); 1229 vput(vp); 1230 } 1231 1232 /* 1233 * Loop around generating replies for all write rpcs that have 1234 * now been completed. 1235 */ 1236 swp = nfsd; 1237 do { 1238 if (error) { 1239 nfsm_writereply(NFSX_WCCDATA(v3), v3); 1240 if (v3) { 1241 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1242 } 1243 } else { 1244 nfsm_writereply(NFSX_PREOPATTR(v3) + 1245 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED + 1246 NFSX_WRITEVERF(v3), v3); 1247 if (v3) { 1248 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1249 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1250 *tl++ = txdr_unsigned(nfsd->nd_len); 1251 *tl++ = txdr_unsigned(swp->nd_stable); 1252 /* 1253 * Actually, there is no need to txdr these fields, 1254 * but it may make the values more human readable, 1255 * for debugging purposes. 1256 */ 1257 *tl++ = txdr_unsigned(boottime.tv_sec); 1258 *tl = txdr_unsigned(boottime.tv_usec); 1259 } else { 1260 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1261 nfsm_srvfillattr(&va, fp); 1262 } 1263 } 1264 nfsd->nd_mreq = mreq; 1265 if (nfsd->nd_mrep) 1266 panic("nfsrv_write: nd_mrep not free"); 1267 1268 /* 1269 * Done. Put it at the head of the timer queue so that 1270 * the final phase can return the reply. 1271 */ 1272 s = splsoftclock(); 1273 if (nfsd != swp) { 1274 nfsd->nd_time = 0; 1275 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1276 } 1277 nfsd = LIST_FIRST(&swp->nd_coalesce); 1278 if (nfsd) { 1279 LIST_REMOVE(nfsd, nd_tq); 1280 } 1281 splx(s); 1282 } while (nfsd); 1283 s = splsoftclock(); 1284 swp->nd_time = 0; 1285 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq); 1286 splx(s); 1287 goto loop1; 1288 } 1289 splx(s); 1290 1291 /* 1292 * Search for a reply to return. 1293 */ 1294 s = splsoftclock(); 1295 LIST_FOREACH(nfsd, &slp->ns_tq, nd_tq) { 1296 if (nfsd->nd_mreq) { 1297 LIST_REMOVE(nfsd, nd_tq); 1298 *mrq = nfsd->nd_mreq; 1299 *ndp = nfsd; 1300 break; 1301 } 1302 } 1303 splx(s); 1304 return (0); 1305 } 1306 1307 /* 1308 * Coalesce the write request nfsd into owp. To do this we must: 1309 * - remove nfsd from the queues 1310 * - merge nfsd->nd_mrep into owp->nd_mrep 1311 * - update the nd_eoff and nd_stable for owp 1312 * - put nfsd on owp's nd_coalesce list 1313 * NB: Must be called at splsoftclock(). 1314 */ 1315 void 1316 nfsrvw_coalesce(owp, nfsd) 1317 struct nfsrv_descript *owp; 1318 struct nfsrv_descript *nfsd; 1319 { 1320 int overlap; 1321 struct mbuf *mp; 1322 struct nfsrv_descript *m; 1323 1324 LIST_REMOVE(nfsd, nd_hash); 1325 LIST_REMOVE(nfsd, nd_tq); 1326 if (owp->nd_eoff < nfsd->nd_eoff) { 1327 overlap = owp->nd_eoff - nfsd->nd_off; 1328 if (overlap < 0) 1329 panic("nfsrv_coalesce: bad off"); 1330 if (overlap > 0) 1331 m_adj(nfsd->nd_mrep, overlap); 1332 mp = owp->nd_mrep; 1333 while (mp->m_next) 1334 mp = mp->m_next; 1335 mp->m_next = nfsd->nd_mrep; 1336 owp->nd_eoff = nfsd->nd_eoff; 1337 } else 1338 m_freem(nfsd->nd_mrep); 1339 nfsd->nd_mrep = NULL; 1340 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC) 1341 owp->nd_stable = NFSV3WRITE_FILESYNC; 1342 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC && 1343 owp->nd_stable == NFSV3WRITE_UNSTABLE) 1344 owp->nd_stable = NFSV3WRITE_DATASYNC; 1345 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq); 1346 /* 1347 * nfsd might hold coalesce elements! Move them to owp. 1348 * Otherwise, requests may be lost and clients will be stuck. 1349 */ 1350 while ((m = LIST_FIRST(&nfsd->nd_coalesce)) != NULL) { 1351 LIST_REMOVE(m, nd_tq); 1352 LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq); 1353 } 1354 } 1355 1356 /* 1357 * nfs create service 1358 * now does a truncate to 0 length via. setattr if it already exists 1359 */ 1360 int 1361 nfsrv_create(nfsd, slp, lwp, mrq) 1362 struct nfsrv_descript *nfsd; 1363 struct nfssvc_sock *slp; 1364 struct lwp *lwp; 1365 struct mbuf **mrq; 1366 { 1367 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1368 struct mbuf *nam = nfsd->nd_nam; 1369 caddr_t dpos = nfsd->nd_dpos; 1370 struct ucred *cred = &nfsd->nd_cr; 1371 struct nfs_fattr *fp; 1372 struct vattr va, dirfor, diraft; 1373 struct nfsv2_sattr *sp; 1374 u_int32_t *tl; 1375 struct nameidata nd; 1376 caddr_t cp; 1377 int32_t t1; 1378 caddr_t bpos; 1379 int error = 0, cache, len, tsize, dirfor_ret = 1, diraft_ret = 1; 1380 int rdev = 0; 1381 int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0; 1382 char *cp2; 1383 struct mbuf *mb, *mreq; 1384 struct vnode *vp = NULL, *dirp = NULL; 1385 nfsfh_t nfh; 1386 fhandle_t *fhp; 1387 u_quad_t frev, tempsize; 1388 u_char cverf[NFSX_V3CREATEVERF]; 1389 struct mount *mp = NULL; 1390 1391 nd.ni_cnd.cn_nameiop = 0; 1392 fhp = &nfh.fh_generic; 1393 nfsm_srvmtofh(fhp); 1394 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1395 return (ESTALE); 1396 vn_start_write(NULL, &mp, V_WAIT); 1397 nfsm_srvnamesiz(len); 1398 nd.ni_cnd.cn_cred = cred; 1399 nd.ni_cnd.cn_nameiop = CREATE; 1400 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1401 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1402 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1403 if (dirp) { 1404 if (v3) 1405 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1406 else { 1407 vrele(dirp); 1408 dirp = (struct vnode *)0; 1409 } 1410 } 1411 if (error) { 1412 nfsm_reply(NFSX_WCCDATA(v3)); 1413 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1414 if (dirp) 1415 vrele(dirp); 1416 vn_finished_write(mp, 0); 1417 return (0); 1418 } 1419 VATTR_NULL(&va); 1420 if (v3) { 1421 va.va_mode = 0; 1422 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1423 how = fxdr_unsigned(int, *tl); 1424 switch (how) { 1425 case NFSV3CREATE_GUARDED: 1426 if (nd.ni_vp) { 1427 error = EEXIST; 1428 break; 1429 } 1430 case NFSV3CREATE_UNCHECKED: 1431 nfsm_srvsattr(&va); 1432 break; 1433 case NFSV3CREATE_EXCLUSIVE: 1434 nfsm_dissect(cp, caddr_t, NFSX_V3CREATEVERF); 1435 memcpy(cverf, cp, NFSX_V3CREATEVERF); 1436 exclusive_flag = 1; 1437 break; 1438 }; 1439 va.va_type = VREG; 1440 } else { 1441 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 1442 va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode)); 1443 if (va.va_type == VNON) 1444 va.va_type = VREG; 1445 va.va_mode = nfstov_mode(sp->sa_mode); 1446 switch (va.va_type) { 1447 case VREG: 1448 tsize = fxdr_unsigned(int32_t, sp->sa_size); 1449 if (tsize != -1) 1450 va.va_size = (u_quad_t)tsize; 1451 break; 1452 case VCHR: 1453 case VBLK: 1454 case VFIFO: 1455 rdev = fxdr_unsigned(int32_t, sp->sa_size); 1456 break; 1457 default: 1458 break; 1459 }; 1460 } 1461 1462 /* 1463 * Iff doesn't exist, create it 1464 * otherwise just truncate to 0 length 1465 * should I set the mode too ?? 1466 */ 1467 if (nd.ni_vp == NULL) { 1468 if (va.va_type == VREG || va.va_type == VSOCK) { 1469 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1470 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1471 if (!error) { 1472 if (exclusive_flag) { 1473 exclusive_flag = 0; 1474 VATTR_NULL(&va); 1475 /* 1476 * XXX 1477 * assuming NFSX_V3CREATEVERF 1478 * == sizeof(nfstime3) 1479 */ 1480 fxdr_nfsv3time(cverf, &va.va_atime); 1481 error = VOP_SETATTR(nd.ni_vp, &va, cred, 1482 lwp); 1483 } 1484 } 1485 } else if (va.va_type == VCHR || va.va_type == VBLK || 1486 va.va_type == VFIFO) { 1487 if (va.va_type == VCHR && rdev == 0xffffffff) 1488 va.va_type = VFIFO; 1489 if (va.va_type != VFIFO && 1490 (error = suser(cred, (u_short *)0))) { 1491 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1492 vput(nd.ni_dvp); 1493 nfsm_reply(0); 1494 vn_finished_write(mp, 0); 1495 return (error); 1496 } else 1497 va.va_rdev = (dev_t)rdev; 1498 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1499 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, 1500 &va); 1501 if (error) { 1502 nfsm_reply(0); 1503 } 1504 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1505 vrele(nd.ni_dvp); 1506 vput(nd.ni_vp); 1507 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1508 error = EINVAL; 1509 nfsm_reply(0); 1510 } 1511 } else { 1512 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1513 vput(nd.ni_dvp); 1514 error = ENXIO; 1515 } 1516 vp = nd.ni_vp; 1517 } else { 1518 vp = nd.ni_vp; 1519 if (nd.ni_dvp == vp) 1520 vrele(nd.ni_dvp); 1521 else 1522 vput(nd.ni_dvp); 1523 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1524 if (!error && va.va_size != -1) { 1525 error = nfsrv_access(vp, VWRITE, cred, 1526 (nd.ni_cnd.cn_flags & RDONLY), lwp, 0); 1527 if (!error) { 1528 nqsrv_getl(vp, ND_WRITE); 1529 tempsize = va.va_size; 1530 VATTR_NULL(&va); 1531 va.va_size = tempsize; 1532 error = VOP_SETATTR(vp, &va, cred, lwp); 1533 } 1534 } 1535 if (error) 1536 vput(vp); 1537 } 1538 if (!error) { 1539 memset((caddr_t)fhp, 0, sizeof(nfh)); 1540 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1541 error = VFS_VPTOFH(vp, &fhp->fh_fid); 1542 if (!error) 1543 error = VOP_GETATTR(vp, &va, cred, lwp); 1544 vput(vp); 1545 } 1546 if (v3) { 1547 if (exclusive_flag && !error) { 1548 /* 1549 * XXX assuming NFSX_V3CREATEVERF == sizeof(nfstime3) 1550 */ 1551 char oldverf[NFSX_V3CREATEVERF]; 1552 1553 txdr_nfsv3time(&va.va_atime, oldverf); 1554 if (memcmp(cverf, oldverf, NFSX_V3CREATEVERF)) 1555 error = EEXIST; 1556 } 1557 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1558 vrele(dirp); 1559 } 1560 nfsm_reply(NFSX_SRVFH(v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3)); 1561 if (v3) { 1562 if (!error) { 1563 nfsm_srvpostop_fh(fhp); 1564 nfsm_srvpostop_attr(0, &va); 1565 } 1566 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1567 } else { 1568 nfsm_srvfhtom(fhp, v3); 1569 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1570 nfsm_srvfillattr(&va, fp); 1571 } 1572 vn_finished_write(mp, 0); 1573 return (0); 1574 nfsmout: 1575 if (dirp) 1576 vrele(dirp); 1577 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1578 if (nd.ni_dvp == nd.ni_vp) 1579 vrele(nd.ni_dvp); 1580 else 1581 vput(nd.ni_dvp); 1582 if (nd.ni_vp) 1583 vput(nd.ni_vp); 1584 vn_finished_write(mp, 0); 1585 return (error); 1586 } 1587 1588 /* 1589 * nfs v3 mknod service 1590 */ 1591 int 1592 nfsrv_mknod(nfsd, slp, lwp, mrq) 1593 struct nfsrv_descript *nfsd; 1594 struct nfssvc_sock *slp; 1595 struct lwp *lwp; 1596 struct mbuf **mrq; 1597 { 1598 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1599 struct mbuf *nam = nfsd->nd_nam; 1600 caddr_t dpos = nfsd->nd_dpos; 1601 struct ucred *cred = &nfsd->nd_cr; 1602 struct vattr va, dirfor, diraft; 1603 u_int32_t *tl; 1604 struct nameidata nd; 1605 int32_t t1; 1606 caddr_t bpos; 1607 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 1608 u_int32_t major, minor; 1609 enum vtype vtyp; 1610 char *cp2; 1611 struct mbuf *mb, *mreq; 1612 struct vnode *vp, *dirp = (struct vnode *)0; 1613 nfsfh_t nfh; 1614 fhandle_t *fhp; 1615 u_quad_t frev; 1616 struct mount *mp = NULL; 1617 1618 nd.ni_cnd.cn_nameiop = 0; 1619 fhp = &nfh.fh_generic; 1620 nfsm_srvmtofh(fhp); 1621 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1622 return (ESTALE); 1623 vn_start_write(NULL, &mp, V_WAIT); 1624 nfsm_srvnamesiz(len); 1625 nd.ni_cnd.cn_cred = cred; 1626 nd.ni_cnd.cn_nameiop = CREATE; 1627 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1628 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1629 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1630 if (dirp) 1631 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1632 if (error) { 1633 nfsm_reply(NFSX_WCCDATA(1)); 1634 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1635 if (dirp) 1636 vrele(dirp); 1637 vn_finished_write(mp, 0); 1638 return (0); 1639 } 1640 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1641 vtyp = nfsv3tov_type(*tl); 1642 if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) { 1643 error = NFSERR_BADTYPE; 1644 goto abort; 1645 } 1646 VATTR_NULL(&va); 1647 va.va_mode = 0; 1648 nfsm_srvsattr(&va); 1649 if (vtyp == VCHR || vtyp == VBLK) { 1650 dev_t rdev; 1651 1652 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1653 major = fxdr_unsigned(u_int32_t, *tl++); 1654 minor = fxdr_unsigned(u_int32_t, *tl); 1655 rdev = makedev(major, minor); 1656 if (major(rdev) != major || minor(rdev) != minor) { 1657 error = EINVAL; 1658 goto abort; 1659 } 1660 va.va_rdev = rdev; 1661 } 1662 1663 /* 1664 * Iff doesn't exist, create it. 1665 */ 1666 if (nd.ni_vp) { 1667 error = EEXIST; 1668 abort: 1669 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1670 if (nd.ni_dvp == nd.ni_vp) 1671 vrele(nd.ni_dvp); 1672 else 1673 vput(nd.ni_dvp); 1674 if (nd.ni_vp) 1675 vput(nd.ni_vp); 1676 goto out; 1677 } 1678 va.va_type = vtyp; 1679 if (vtyp == VSOCK) { 1680 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1681 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1682 } else { 1683 if (va.va_type != VFIFO && 1684 (error = suser(cred, (u_short *)0))) { 1685 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1686 vput(nd.ni_dvp); 1687 goto out; 1688 } 1689 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1690 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1691 if (error) { 1692 goto out; 1693 } 1694 if (error) 1695 goto out; 1696 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1697 vput(nd.ni_vp); 1698 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1699 error = EINVAL; 1700 } 1701 } 1702 out: 1703 vp = nd.ni_vp; 1704 if (!error) { 1705 memset((caddr_t)fhp, 0, sizeof(nfh)); 1706 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1707 error = VFS_VPTOFH(vp, &fhp->fh_fid); 1708 if (!error) 1709 error = VOP_GETATTR(vp, &va, cred, lwp); 1710 vput(vp); 1711 } 1712 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1713 vrele(dirp); 1714 nfsm_reply(NFSX_SRVFH(1) + NFSX_POSTOPATTR(1) + NFSX_WCCDATA(1)); 1715 if (!error) { 1716 nfsm_srvpostop_fh(fhp); 1717 nfsm_srvpostop_attr(0, &va); 1718 } 1719 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1720 vn_finished_write(mp, 0); 1721 return (0); 1722 nfsmout: 1723 if (dirp) 1724 vrele(dirp); 1725 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1726 if (nd.ni_dvp == nd.ni_vp) 1727 vrele(nd.ni_dvp); 1728 else 1729 vput(nd.ni_dvp); 1730 if (nd.ni_vp) 1731 vput(nd.ni_vp); 1732 vn_finished_write(mp, 0); 1733 return (error); 1734 } 1735 1736 /* 1737 * nfs remove service 1738 */ 1739 int 1740 nfsrv_remove(nfsd, slp, lwp, mrq) 1741 struct nfsrv_descript *nfsd; 1742 struct nfssvc_sock *slp; 1743 struct lwp *lwp; 1744 struct mbuf **mrq; 1745 { 1746 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1747 struct mbuf *nam = nfsd->nd_nam; 1748 caddr_t dpos = nfsd->nd_dpos; 1749 struct ucred *cred = &nfsd->nd_cr; 1750 struct nameidata nd; 1751 u_int32_t *tl; 1752 int32_t t1; 1753 caddr_t bpos; 1754 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 1755 int v3 = (nfsd->nd_flag & ND_NFSV3); 1756 char *cp2; 1757 struct mbuf *mb, *mreq; 1758 struct vnode *vp, *dirp; 1759 struct vattr dirfor, diraft; 1760 nfsfh_t nfh; 1761 fhandle_t *fhp; 1762 u_quad_t frev; 1763 struct mount *mp = NULL; 1764 1765 #ifndef nolint 1766 vp = (struct vnode *)0; 1767 #endif 1768 fhp = &nfh.fh_generic; 1769 nfsm_srvmtofh(fhp); 1770 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1771 return (ESTALE); 1772 vn_start_write(NULL, &mp, V_WAIT); 1773 nfsm_srvnamesiz(len); 1774 nd.ni_cnd.cn_cred = cred; 1775 nd.ni_cnd.cn_nameiop = DELETE; 1776 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1777 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1778 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1779 if (dirp) { 1780 if (v3) 1781 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1782 else 1783 vrele(dirp); 1784 } 1785 if (!error) { 1786 vp = nd.ni_vp; 1787 if (vp->v_type == VDIR && 1788 (error = suser(cred, (u_short *)0)) != 0) 1789 goto out; 1790 /* 1791 * The root of a mounted filesystem cannot be deleted. 1792 */ 1793 if (vp->v_flag & VROOT) { 1794 error = EBUSY; 1795 goto out; 1796 } 1797 out: 1798 if (!error) { 1799 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1800 nqsrv_getl(vp, ND_WRITE); 1801 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1802 } else { 1803 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1804 if (nd.ni_dvp == vp) 1805 vrele(nd.ni_dvp); 1806 else 1807 vput(nd.ni_dvp); 1808 vput(vp); 1809 } 1810 } 1811 if (dirp && v3) { 1812 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1813 vrele(dirp); 1814 } 1815 nfsm_reply(NFSX_WCCDATA(v3)); 1816 if (v3) { 1817 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1818 vn_finished_write(mp, 0); 1819 return (0); 1820 } 1821 vn_finished_write(mp, 0); 1822 nfsm_srvdone; 1823 } 1824 1825 /* 1826 * nfs rename service 1827 */ 1828 int 1829 nfsrv_rename(nfsd, slp, lwp, mrq) 1830 struct nfsrv_descript *nfsd; 1831 struct nfssvc_sock *slp; 1832 struct lwp *lwp; 1833 struct mbuf **mrq; 1834 { 1835 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1836 struct mbuf *nam = nfsd->nd_nam; 1837 caddr_t dpos = nfsd->nd_dpos; 1838 struct ucred *cred = &nfsd->nd_cr; 1839 u_int32_t *tl; 1840 int32_t t1; 1841 caddr_t bpos; 1842 int error = 0, cache, fdirfor_ret = 1, fdiraft_ret = 1; 1843 uint32_t len, len2; 1844 int tdirfor_ret = 1, tdiraft_ret = 1; 1845 int v3 = (nfsd->nd_flag & ND_NFSV3); 1846 char *cp2; 1847 struct mbuf *mb, *mreq; 1848 struct nameidata fromnd, tond; 1849 struct vnode *fvp, *tvp, *tdvp, *fdirp = (struct vnode *)0; 1850 struct vnode *tdirp = (struct vnode *)0; 1851 struct vattr fdirfor, fdiraft, tdirfor, tdiraft; 1852 nfsfh_t fnfh, tnfh; 1853 fhandle_t *ffhp, *tfhp; 1854 u_quad_t frev; 1855 uid_t saved_uid; 1856 struct mount *mp = NULL; 1857 1858 #ifndef nolint 1859 fvp = (struct vnode *)0; 1860 #endif 1861 ffhp = &fnfh.fh_generic; 1862 tfhp = &tnfh.fh_generic; 1863 fromnd.ni_cnd.cn_nameiop = 0; 1864 tond.ni_cnd.cn_nameiop = 0; 1865 nfsm_srvmtofh(ffhp); 1866 if ((mp = vfs_getvfs(&ffhp->fh_fsid)) == NULL) 1867 return (ESTALE); 1868 vn_start_write(NULL, &mp, V_WAIT); 1869 nfsm_srvnamesiz(len); 1870 /* 1871 * Remember our original uid so that we can reset cr_uid before 1872 * the second nfs_namei() call, in case it is remapped. 1873 */ 1874 saved_uid = cred->cr_uid; 1875 fromnd.ni_cnd.cn_cred = cred; 1876 fromnd.ni_cnd.cn_nameiop = DELETE; 1877 fromnd.ni_cnd.cn_flags = WANTPARENT | SAVESTART; 1878 error = nfs_namei(&fromnd, ffhp, len, slp, nam, &md, 1879 &dpos, &fdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1880 if (fdirp) { 1881 if (v3) 1882 fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred, lwp); 1883 else { 1884 vrele(fdirp); 1885 fdirp = (struct vnode *)0; 1886 } 1887 } 1888 if (error) { 1889 nfsm_reply(2 * NFSX_WCCDATA(v3)); 1890 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 1891 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 1892 if (fdirp) 1893 vrele(fdirp); 1894 vn_finished_write(mp, 0); 1895 return (0); 1896 } 1897 fvp = fromnd.ni_vp; 1898 nfsm_srvmtofh(tfhp); 1899 if (v3) { 1900 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 1901 len2 = fxdr_unsigned(uint32_t, *tl); 1902 /* len2 will be checked by nfs_namei */ 1903 } 1904 else { 1905 /* NFSv2 */ 1906 nfsm_strsiz(len2, NFS_MAXNAMLEN); 1907 } 1908 cred->cr_uid = saved_uid; 1909 tond.ni_cnd.cn_cred = cred; 1910 tond.ni_cnd.cn_nameiop = RENAME; 1911 tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART; 1912 error = nfs_namei(&tond, tfhp, len2, slp, nam, &md, 1913 &dpos, &tdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1914 if (tdirp) { 1915 if (v3) 1916 tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred, lwp); 1917 else { 1918 vrele(tdirp); 1919 tdirp = (struct vnode *)0; 1920 } 1921 } 1922 if (error) { 1923 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1924 vrele(fromnd.ni_dvp); 1925 vrele(fvp); 1926 goto out1; 1927 } 1928 tdvp = tond.ni_dvp; 1929 tvp = tond.ni_vp; 1930 if (tvp != NULL) { 1931 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 1932 if (v3) 1933 error = EEXIST; 1934 else 1935 error = EISDIR; 1936 goto out; 1937 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 1938 if (v3) 1939 error = EEXIST; 1940 else 1941 error = ENOTDIR; 1942 goto out; 1943 } 1944 if (tvp->v_type == VDIR && tvp->v_mountedhere) { 1945 if (v3) 1946 error = EXDEV; 1947 else 1948 error = ENOTEMPTY; 1949 goto out; 1950 } 1951 } 1952 if (fvp->v_type == VDIR && fvp->v_mountedhere) { 1953 if (v3) 1954 error = EXDEV; 1955 else 1956 error = ENOTEMPTY; 1957 goto out; 1958 } 1959 if (fvp->v_mount != tdvp->v_mount) { 1960 if (v3) 1961 error = EXDEV; 1962 else 1963 error = ENOTEMPTY; 1964 goto out; 1965 } 1966 if (fvp == tdvp) { 1967 if (v3) 1968 error = EINVAL; 1969 else 1970 error = ENOTEMPTY; 1971 } 1972 /* 1973 * If source is the same as the destination (that is the 1974 * same vnode with the same name in the same directory), 1975 * then there is nothing to do. 1976 */ 1977 if (fvp == tvp && fromnd.ni_dvp == tdvp && 1978 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 1979 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr, 1980 fromnd.ni_cnd.cn_namelen)) 1981 error = -1; 1982 out: 1983 if (!error) { 1984 nqsrv_getl(fromnd.ni_dvp, ND_WRITE); 1985 nqsrv_getl(tdvp, ND_WRITE); 1986 if (tvp) { 1987 nqsrv_getl(tvp, ND_WRITE); 1988 } 1989 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 1990 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 1991 } else { 1992 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 1993 if (tdvp == tvp) 1994 vrele(tdvp); 1995 else 1996 vput(tdvp); 1997 if (tvp) 1998 vput(tvp); 1999 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2000 vrele(fromnd.ni_dvp); 2001 vrele(fvp); 2002 if (error == -1) 2003 error = 0; 2004 } 2005 vrele(tond.ni_startdir); 2006 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 2007 out1: 2008 if (fdirp) { 2009 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred, lwp); 2010 vrele(fdirp); 2011 } 2012 if (tdirp) { 2013 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred, lwp); 2014 vrele(tdirp); 2015 } 2016 vrele(fromnd.ni_startdir); 2017 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2018 nfsm_reply(2 * NFSX_WCCDATA(v3)); 2019 if (v3) { 2020 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 2021 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 2022 } 2023 vn_finished_write(mp, 0); 2024 return (0); 2025 2026 nfsmout: 2027 if (fdirp) 2028 vrele(fdirp); 2029 if (tdirp) 2030 vrele(tdirp); 2031 if (tond.ni_cnd.cn_nameiop) { 2032 vrele(tond.ni_startdir); 2033 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 2034 } 2035 if (fromnd.ni_cnd.cn_nameiop) { 2036 vrele(fromnd.ni_startdir); 2037 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2038 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2039 vrele(fromnd.ni_dvp); 2040 vrele(fvp); 2041 } 2042 return (error); 2043 } 2044 2045 /* 2046 * nfs link service 2047 */ 2048 int 2049 nfsrv_link(nfsd, slp, lwp, mrq) 2050 struct nfsrv_descript *nfsd; 2051 struct nfssvc_sock *slp; 2052 struct lwp *lwp; 2053 struct mbuf **mrq; 2054 { 2055 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2056 struct mbuf *nam = nfsd->nd_nam; 2057 caddr_t dpos = nfsd->nd_dpos; 2058 struct ucred *cred = &nfsd->nd_cr; 2059 struct nameidata nd; 2060 u_int32_t *tl; 2061 int32_t t1; 2062 caddr_t bpos; 2063 int error = 0, rdonly, cache, len, dirfor_ret = 1, diraft_ret = 1; 2064 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3); 2065 char *cp2; 2066 struct mbuf *mb, *mreq; 2067 struct vnode *vp, *xp, *dirp = (struct vnode *)0; 2068 struct vattr dirfor, diraft, at; 2069 nfsfh_t nfh, dnfh; 2070 fhandle_t *fhp, *dfhp; 2071 u_quad_t frev; 2072 struct mount *mp = NULL; 2073 2074 fhp = &nfh.fh_generic; 2075 dfhp = &dnfh.fh_generic; 2076 nfsm_srvmtofh(fhp); 2077 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2078 return (ESTALE); 2079 vn_start_write(NULL, &mp, V_WAIT); 2080 nfsm_srvmtofh(dfhp); 2081 nfsm_srvnamesiz(len); 2082 error = nfsrv_fhtovp(fhp, FALSE, &vp, cred, slp, nam, 2083 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2084 if (error) { 2085 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2086 nfsm_srvpostop_attr(getret, &at); 2087 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2088 vn_finished_write(mp, 0); 2089 return (0); 2090 } 2091 if (vp->v_type == VDIR && (error = suser(cred, (u_short *)0)) != 0) 2092 goto out1; 2093 nd.ni_cnd.cn_cred = cred; 2094 nd.ni_cnd.cn_nameiop = CREATE; 2095 nd.ni_cnd.cn_flags = LOCKPARENT; 2096 error = nfs_namei(&nd, dfhp, len, slp, nam, &md, &dpos, 2097 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2098 if (dirp) { 2099 if (v3) 2100 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2101 else { 2102 vrele(dirp); 2103 dirp = (struct vnode *)0; 2104 } 2105 } 2106 if (error) 2107 goto out1; 2108 xp = nd.ni_vp; 2109 if (xp != NULL) { 2110 error = EEXIST; 2111 goto out; 2112 } 2113 xp = nd.ni_dvp; 2114 if (vp->v_mount != xp->v_mount) 2115 error = EXDEV; 2116 out: 2117 if (!error) { 2118 nqsrv_getl(vp, ND_WRITE); 2119 nqsrv_getl(xp, ND_WRITE); 2120 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2121 } else { 2122 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2123 if (nd.ni_dvp == nd.ni_vp) 2124 vrele(nd.ni_dvp); 2125 else 2126 vput(nd.ni_dvp); 2127 if (nd.ni_vp) 2128 vrele(nd.ni_vp); 2129 } 2130 out1: 2131 if (v3) 2132 getret = VOP_GETATTR(vp, &at, cred, lwp); 2133 if (dirp) { 2134 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2135 vrele(dirp); 2136 } 2137 vrele(vp); 2138 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2139 if (v3) { 2140 nfsm_srvpostop_attr(getret, &at); 2141 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2142 vn_finished_write(mp, 0); 2143 return (0); 2144 } 2145 vn_finished_write(mp, 0); 2146 nfsm_srvdone; 2147 } 2148 2149 /* 2150 * nfs symbolic link service 2151 */ 2152 int 2153 nfsrv_symlink(nfsd, slp, lwp, mrq) 2154 struct nfsrv_descript *nfsd; 2155 struct nfssvc_sock *slp; 2156 struct lwp *lwp; 2157 struct mbuf **mrq; 2158 { 2159 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2160 struct mbuf *nam = nfsd->nd_nam; 2161 caddr_t dpos = nfsd->nd_dpos; 2162 struct ucred *cred = &nfsd->nd_cr; 2163 struct vattr va, dirfor, diraft; 2164 struct nameidata nd; 2165 u_int32_t *tl; 2166 int32_t t1; 2167 struct nfsv2_sattr *sp; 2168 char *bpos, *pathcp = NULL, *cp2; 2169 struct uio io; 2170 struct iovec iv; 2171 int error = 0, cache, dirfor_ret = 1, diraft_ret = 1; 2172 uint32_t len, len2; 2173 int v3 = (nfsd->nd_flag & ND_NFSV3); 2174 struct mbuf *mb, *mreq; 2175 struct vnode *dirp = (struct vnode *)0; 2176 nfsfh_t nfh; 2177 fhandle_t *fhp; 2178 u_quad_t frev; 2179 struct mount *mp = NULL; 2180 2181 nd.ni_cnd.cn_nameiop = 0; 2182 fhp = &nfh.fh_generic; 2183 nfsm_srvmtofh(fhp); 2184 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2185 return (ESTALE); 2186 vn_start_write(NULL, &mp, V_WAIT); 2187 nfsm_srvnamesiz(len); 2188 nd.ni_cnd.cn_cred = cred; 2189 nd.ni_cnd.cn_nameiop = CREATE; 2190 nd.ni_cnd.cn_flags = LOCKPARENT; 2191 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2192 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2193 if (dirp) { 2194 if (v3) 2195 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2196 else { 2197 vrele(dirp); 2198 dirp = (struct vnode *)0; 2199 } 2200 } 2201 if (error) 2202 goto out; 2203 VATTR_NULL(&va); 2204 va.va_type = VLNK; 2205 if (v3) { 2206 va.va_mode = 0; 2207 nfsm_srvsattr(&va); 2208 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 2209 len2 = fxdr_unsigned(uint32_t, *tl); 2210 if (len2 > PATH_MAX) { 2211 /* XXX should check _PC_NO_TRUNC */ 2212 error = ENAMETOOLONG; 2213 goto abortop; 2214 } 2215 } 2216 else { 2217 /* NFSv2 */ 2218 nfsm_strsiz(len2, NFS_MAXPATHLEN); 2219 } 2220 pathcp = malloc(len2 + 1, M_TEMP, M_WAITOK); 2221 iv.iov_base = pathcp; 2222 iv.iov_len = len2; 2223 io.uio_resid = len2; 2224 io.uio_offset = 0; 2225 io.uio_iov = &iv; 2226 io.uio_iovcnt = 1; 2227 io.uio_rw = UIO_READ; 2228 UIO_SETUP_SYSSPACE(&io); 2229 nfsm_mtouio(&io, len2); 2230 if (!v3) { 2231 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 2232 va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode); 2233 } 2234 *(pathcp + len2) = '\0'; 2235 if (nd.ni_vp) { 2236 error = EEXIST; 2237 abortop: 2238 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2239 if (nd.ni_dvp == nd.ni_vp) 2240 vrele(nd.ni_dvp); 2241 else 2242 vput(nd.ni_dvp); 2243 if (nd.ni_vp) 2244 vrele(nd.ni_vp); 2245 goto out; 2246 } 2247 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2248 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp); 2249 if (!error) { 2250 if (v3) { 2251 memset((caddr_t)fhp, 0, sizeof(nfh)); 2252 fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsidx; 2253 error = VFS_VPTOFH(nd.ni_vp, &fhp->fh_fid); 2254 if (!error) 2255 error = VOP_GETATTR(nd.ni_vp, &va, cred, lwp); 2256 vput(nd.ni_vp); 2257 } else { 2258 vput(nd.ni_vp); 2259 } 2260 } 2261 out: 2262 if (pathcp) 2263 free(pathcp, M_TEMP); 2264 if (dirp) { 2265 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2266 vrele(dirp); 2267 } 2268 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2269 if (v3) { 2270 if (!error) { 2271 nfsm_srvpostop_fh(fhp); 2272 nfsm_srvpostop_attr(0, &va); 2273 } 2274 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2275 } 2276 vn_finished_write(mp, 0); 2277 return (0); 2278 nfsmout: 2279 if (dirp) 2280 vrele(dirp); 2281 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2282 if (nd.ni_dvp == nd.ni_vp) 2283 vrele(nd.ni_dvp); 2284 else 2285 vput(nd.ni_dvp); 2286 if (nd.ni_vp) 2287 vrele(nd.ni_vp); 2288 if (pathcp) 2289 free(pathcp, M_TEMP); 2290 vn_finished_write(mp, 0); 2291 return (error); 2292 } 2293 2294 /* 2295 * nfs mkdir service 2296 */ 2297 int 2298 nfsrv_mkdir(nfsd, slp, lwp, mrq) 2299 struct nfsrv_descript *nfsd; 2300 struct nfssvc_sock *slp; 2301 struct lwp *lwp; 2302 struct mbuf **mrq; 2303 { 2304 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2305 struct mbuf *nam = nfsd->nd_nam; 2306 caddr_t dpos = nfsd->nd_dpos; 2307 struct ucred *cred = &nfsd->nd_cr; 2308 struct vattr va, dirfor, diraft; 2309 struct nfs_fattr *fp; 2310 struct nameidata nd; 2311 caddr_t cp; 2312 u_int32_t *tl; 2313 int32_t t1; 2314 caddr_t bpos; 2315 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 2316 int v3 = (nfsd->nd_flag & ND_NFSV3); 2317 char *cp2; 2318 struct mbuf *mb, *mreq; 2319 struct vnode *vp, *dirp = (struct vnode *)0; 2320 nfsfh_t nfh; 2321 fhandle_t *fhp; 2322 u_quad_t frev; 2323 struct mount *mp = NULL; 2324 2325 fhp = &nfh.fh_generic; 2326 nfsm_srvmtofh(fhp); 2327 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2328 return (ESTALE); 2329 vn_start_write(NULL, &mp, V_WAIT); 2330 nfsm_srvnamesiz(len); 2331 nd.ni_cnd.cn_cred = cred; 2332 nd.ni_cnd.cn_nameiop = CREATE; 2333 nd.ni_cnd.cn_flags = LOCKPARENT; 2334 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2335 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2336 if (dirp) { 2337 if (v3) 2338 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2339 else { 2340 vrele(dirp); 2341 dirp = (struct vnode *)0; 2342 } 2343 } 2344 if (error) { 2345 nfsm_reply(NFSX_WCCDATA(v3)); 2346 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2347 if (dirp) 2348 vrele(dirp); 2349 vn_finished_write(mp, 0); 2350 return (0); 2351 } 2352 VATTR_NULL(&va); 2353 if (v3) { 2354 va.va_mode = 0; 2355 nfsm_srvsattr(&va); 2356 } else { 2357 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 2358 va.va_mode = nfstov_mode(*tl++); 2359 } 2360 va.va_type = VDIR; 2361 vp = nd.ni_vp; 2362 if (vp != NULL) { 2363 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2364 if (nd.ni_dvp == vp) 2365 vrele(nd.ni_dvp); 2366 else 2367 vput(nd.ni_dvp); 2368 vrele(vp); 2369 error = EEXIST; 2370 goto out; 2371 } 2372 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2373 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 2374 if (!error) { 2375 vp = nd.ni_vp; 2376 memset((caddr_t)fhp, 0, sizeof(nfh)); 2377 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 2378 error = VFS_VPTOFH(vp, &fhp->fh_fid); 2379 if (!error) 2380 error = VOP_GETATTR(vp, &va, cred, lwp); 2381 vput(vp); 2382 } 2383 out: 2384 if (dirp) { 2385 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2386 vrele(dirp); 2387 } 2388 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2389 if (v3) { 2390 if (!error) { 2391 nfsm_srvpostop_fh(fhp); 2392 nfsm_srvpostop_attr(0, &va); 2393 } 2394 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2395 } else { 2396 nfsm_srvfhtom(fhp, v3); 2397 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 2398 nfsm_srvfillattr(&va, fp); 2399 } 2400 vn_finished_write(mp, 0); 2401 return (0); 2402 nfsmout: 2403 if (dirp) 2404 vrele(dirp); 2405 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2406 if (nd.ni_dvp == nd.ni_vp) 2407 vrele(nd.ni_dvp); 2408 else 2409 vput(nd.ni_dvp); 2410 if (nd.ni_vp) 2411 vrele(nd.ni_vp); 2412 vn_finished_write(mp, 0); 2413 return (error); 2414 } 2415 2416 /* 2417 * nfs rmdir service 2418 */ 2419 int 2420 nfsrv_rmdir(nfsd, slp, lwp, mrq) 2421 struct nfsrv_descript *nfsd; 2422 struct nfssvc_sock *slp; 2423 struct lwp *lwp; 2424 struct mbuf **mrq; 2425 { 2426 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2427 struct mbuf *nam = nfsd->nd_nam; 2428 caddr_t dpos = nfsd->nd_dpos; 2429 struct ucred *cred = &nfsd->nd_cr; 2430 u_int32_t *tl; 2431 int32_t t1; 2432 caddr_t bpos; 2433 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 2434 int v3 = (nfsd->nd_flag & ND_NFSV3); 2435 char *cp2; 2436 struct mbuf *mb, *mreq; 2437 struct vnode *vp, *dirp = (struct vnode *)0; 2438 struct vattr dirfor, diraft; 2439 nfsfh_t nfh; 2440 fhandle_t *fhp; 2441 struct nameidata nd; 2442 u_quad_t frev; 2443 struct mount *mp = NULL; 2444 2445 fhp = &nfh.fh_generic; 2446 nfsm_srvmtofh(fhp); 2447 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2448 return (ESTALE); 2449 vn_start_write(NULL, &mp, V_WAIT); 2450 nfsm_srvnamesiz(len); 2451 nd.ni_cnd.cn_cred = cred; 2452 nd.ni_cnd.cn_nameiop = DELETE; 2453 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 2454 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2455 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2456 if (dirp) { 2457 if (v3) 2458 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2459 else { 2460 vrele(dirp); 2461 dirp = (struct vnode *)0; 2462 } 2463 } 2464 if (error) { 2465 nfsm_reply(NFSX_WCCDATA(v3)); 2466 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2467 if (dirp) 2468 vrele(dirp); 2469 vn_finished_write(mp, 0); 2470 return (0); 2471 } 2472 vp = nd.ni_vp; 2473 if (vp->v_type != VDIR) { 2474 error = ENOTDIR; 2475 goto out; 2476 } 2477 /* 2478 * No rmdir "." please. 2479 */ 2480 if (nd.ni_dvp == vp) { 2481 error = EINVAL; 2482 goto out; 2483 } 2484 /* 2485 * The root of a mounted filesystem cannot be deleted. 2486 */ 2487 if (vp->v_flag & VROOT) 2488 error = EBUSY; 2489 out: 2490 if (!error) { 2491 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2492 nqsrv_getl(vp, ND_WRITE); 2493 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2494 } else { 2495 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2496 if (nd.ni_dvp == nd.ni_vp) 2497 vrele(nd.ni_dvp); 2498 else 2499 vput(nd.ni_dvp); 2500 vput(vp); 2501 } 2502 if (dirp) { 2503 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2504 vrele(dirp); 2505 } 2506 nfsm_reply(NFSX_WCCDATA(v3)); 2507 if (v3) { 2508 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2509 vn_finished_write(mp, 0); 2510 return (0); 2511 } 2512 vn_finished_write(mp, 0); 2513 nfsm_srvdone; 2514 } 2515 2516 /* 2517 * nfs readdir service 2518 * - mallocs what it thinks is enough to read 2519 * count rounded up to a multiple of NFS_SRVDIRBLKSIZ <= NFS_MAXREADDIR 2520 * - calls VOP_READDIR() 2521 * - loops around building the reply 2522 * if the output generated exceeds count break out of loop 2523 * The nfsm_clget macro is used here so that the reply will be packed 2524 * tightly in mbuf clusters. 2525 * - it only knows that it has encountered eof when the VOP_READDIR() 2526 * reads nothing 2527 * - as such one readdir rpc will return eof false although you are there 2528 * and then the next will return eof 2529 * - it trims out records with d_fileno == 0 2530 * this doesn't matter for Unix clients, but they might confuse clients 2531 * for other os'. 2532 * - it trims out records with d_type == DT_WHT 2533 * these cannot be seen through NFS (unless we extend the protocol) 2534 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less 2535 * than requested, but this may not apply to all filesystems. For 2536 * example, client NFS does not { although it is never remote mounted 2537 * anyhow } 2538 * The alternate call nfsrv_readdirplus() does lookups as well. 2539 * PS: The NFS protocol spec. does not clarify what the "count" byte 2540 * argument is a count of.. just name strings and file id's or the 2541 * entire reply rpc or ... 2542 * I tried just file name and id sizes and it confused the Sun client, 2543 * so I am using the full rpc size now. The "paranoia.." comment refers 2544 * to including the status longwords that are not a part of the dir. 2545 * "entry" structures, but are in the rpc. 2546 */ 2547 2548 #define NFS_SRVDIRBLKSIZ 1024 2549 2550 struct flrep { 2551 nfsuint64 fl_off; 2552 u_int32_t fl_postopok; 2553 u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)]; 2554 u_int32_t fl_fhok; 2555 u_int32_t fl_fhsize; 2556 u_int32_t fl_nfh[NFSX_V3FH / sizeof (u_int32_t)]; 2557 }; 2558 2559 int 2560 nfsrv_readdir(nfsd, slp, lwp, mrq) 2561 struct nfsrv_descript *nfsd; 2562 struct nfssvc_sock *slp; 2563 struct lwp *lwp; 2564 struct mbuf **mrq; 2565 { 2566 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2567 struct mbuf *nam = nfsd->nd_nam; 2568 caddr_t dpos = nfsd->nd_dpos; 2569 struct ucred *cred = &nfsd->nd_cr; 2570 char *bp, *be; 2571 struct mbuf *mp; 2572 struct dirent *dp; 2573 caddr_t cp; 2574 u_int32_t *tl; 2575 int32_t t1; 2576 caddr_t bpos; 2577 struct mbuf *mb, *mreq, *mp2; 2578 char *cpos, *cend, *cp2, *rbuf; 2579 struct vnode *vp; 2580 struct vattr at; 2581 nfsfh_t nfh; 2582 fhandle_t *fhp; 2583 struct uio io; 2584 struct iovec iv; 2585 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2586 int siz, cnt, fullsiz, eofflag, rdonly, cache, ncookies; 2587 int v3 = (nfsd->nd_flag & ND_NFSV3); 2588 u_quad_t frev, off, toff, verf; 2589 off_t *cookies = NULL, *cookiep; 2590 nfsuint64 jar; 2591 2592 fhp = &nfh.fh_generic; 2593 nfsm_srvmtofh(fhp); 2594 if (v3) { 2595 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 2596 toff = fxdr_hyper(tl); 2597 tl += 2; 2598 verf = fxdr_hyper(tl); 2599 tl += 2; 2600 } else { 2601 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2602 toff = fxdr_unsigned(u_quad_t, *tl++); 2603 } 2604 off = toff; 2605 cnt = fxdr_unsigned(int, *tl); 2606 siz = ((cnt + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1)); 2607 xfer = NFS_SRVMAXDATA(nfsd); 2608 if (siz > xfer) 2609 siz = xfer; 2610 fullsiz = siz; 2611 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 2612 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2613 if (!error && vp->v_type != VDIR) { 2614 error = ENOTDIR; 2615 vput(vp); 2616 } 2617 if (error) { 2618 nfsm_reply(NFSX_UNSIGNED); 2619 nfsm_srvpostop_attr(getret, &at); 2620 return (0); 2621 } 2622 nqsrv_getl(vp, ND_READ); 2623 if (v3) { 2624 error = getret = VOP_GETATTR(vp, &at, cred, lwp); 2625 #ifdef NFS3_STRICTVERF 2626 /* 2627 * XXX This check is too strict for Solaris 2.5 clients. 2628 */ 2629 if (!error && toff && verf != at.va_filerev) 2630 error = NFSERR_BAD_COOKIE; 2631 #endif 2632 } 2633 if (!error) 2634 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0); 2635 if (error) { 2636 vput(vp); 2637 nfsm_reply(NFSX_POSTOPATTR(v3)); 2638 nfsm_srvpostop_attr(getret, &at); 2639 return (0); 2640 } 2641 VOP_UNLOCK(vp, 0); 2642 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2643 again: 2644 iv.iov_base = rbuf; 2645 iv.iov_len = fullsiz; 2646 io.uio_iov = &iv; 2647 io.uio_iovcnt = 1; 2648 io.uio_offset = (off_t)off; 2649 io.uio_resid = fullsiz; 2650 io.uio_rw = UIO_READ; 2651 UIO_SETUP_SYSSPACE(&io); 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 = txdr_unsigned(dp->d_fileno >> 32); 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 + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 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_rw = UIO_READ; 2909 UIO_SETUP_SYSSPACE(&io); 2910 eofflag = 0; 2911 2912 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2913 2914 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2915 2916 off = (u_quad_t)io.uio_offset; 2917 getret = VOP_GETATTR(vp, &at, cred, lwp); 2918 2919 VOP_UNLOCK(vp, 0); 2920 2921 /* 2922 * If the VGET operation doesn't work for this filesystem, 2923 * we can't support readdirplus. Returning NOTSUPP should 2924 * make clients fall back to plain readdir. 2925 * There's no need to check for VPTOFH as well, we wouldn't 2926 * even be here otherwise. 2927 */ 2928 if (!getret) { 2929 if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp))) 2930 getret = (getret == EOPNOTSUPP) ? 2931 NFSERR_NOTSUPP : NFSERR_IO; 2932 else 2933 vput(nvp); 2934 } 2935 2936 if (!cookies && !error) 2937 error = NFSERR_PERM; 2938 if (!error) 2939 error = getret; 2940 if (error) { 2941 vrele(vp); 2942 if (cookies) 2943 free((caddr_t)cookies, M_TEMP); 2944 free((caddr_t)rbuf, M_TEMP); 2945 nfsm_reply(NFSX_V3POSTOPATTR); 2946 nfsm_srvpostop_attr(getret, &at); 2947 return (0); 2948 } 2949 if (io.uio_resid) { 2950 siz -= io.uio_resid; 2951 2952 /* 2953 * If nothing read, return eof 2954 * rpc reply 2955 */ 2956 if (siz == 0) { 2957 vrele(vp); 2958 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2959 2 * NFSX_UNSIGNED); 2960 nfsm_srvpostop_attr(getret, &at); 2961 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2962 txdr_hyper(at.va_filerev, tl); 2963 tl += 2; 2964 *tl++ = nfs_false; 2965 *tl = nfs_true; 2966 free((caddr_t)cookies, M_TEMP); 2967 free((caddr_t)rbuf, M_TEMP); 2968 return (0); 2969 } 2970 } 2971 2972 /* 2973 * Check for degenerate cases of nothing useful read. 2974 * If so go try again 2975 */ 2976 cpos = rbuf; 2977 cend = rbuf + siz; 2978 dp = (struct dirent *)cpos; 2979 cookiep = cookies; 2980 2981 while (cpos < cend && ncookies > 0 && 2982 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2983 cpos += dp->d_reclen; 2984 dp = (struct dirent *)cpos; 2985 cookiep++; 2986 ncookies--; 2987 } 2988 if (cpos >= cend || ncookies == 0) { 2989 toff = off; 2990 siz = fullsiz; 2991 goto again; 2992 } 2993 2994 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED; 2995 nfsm_reply(cnt); 2996 nfsm_srvpostop_attr(getret, &at); 2997 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2998 txdr_hyper(at.va_filerev, tl); 2999 mp = mp2 = mb; 3000 bp = bpos; 3001 be = bp + M_TRAILINGSPACE(mp); 3002 3003 /* Loop through the records and build reply */ 3004 while (cpos < cend && ncookies > 0) { 3005 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 3006 nlen = dp->d_namlen; 3007 rem = nfsm_rndup(nlen)-nlen; 3008 3009 /* 3010 * For readdir_and_lookup get the vnode using 3011 * the file number. 3012 */ 3013 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp)) 3014 goto invalid; 3015 memset((caddr_t)nfhp, 0, NFSX_V3FH); 3016 nfhp->fh_fsid = 3017 nvp->v_mount->mnt_stat.f_fsidx; 3018 if (VFS_VPTOFH(nvp, &nfhp->fh_fid)) { 3019 vput(nvp); 3020 goto invalid; 3021 } 3022 if (VOP_GETATTR(nvp, vap, cred, lwp)) { 3023 vput(nvp); 3024 goto invalid; 3025 } 3026 vput(nvp); 3027 3028 /* 3029 * If either the dircount or maxcount will be 3030 * exceeded, get out now. Both of these lengths 3031 * are calculated conservatively, including all 3032 * XDR overheads. 3033 */ 3034 len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH + 3035 NFSX_V3POSTOPATTR); 3036 dirlen += (6 * NFSX_UNSIGNED + nlen + rem); 3037 if (len > cnt || dirlen > fullsiz) { 3038 eofflag = 0; 3039 break; 3040 } 3041 3042 /* 3043 * Build the directory record xdr from 3044 * the dirent entry. 3045 */ 3046 fp = (struct nfs_fattr *)&fl.fl_fattr; 3047 nfsm_srvfillattr(vap, fp); 3048 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH); 3049 fl.fl_fhok = nfs_true; 3050 fl.fl_postopok = nfs_true; 3051 txdr_hyper(*cookiep, fl.fl_off.nfsuquad); 3052 3053 nfsm_clget; 3054 *tl = nfs_true; 3055 bp += NFSX_UNSIGNED; 3056 nfsm_clget; 3057 *tl = txdr_unsigned(dp->d_fileno >> 32); 3058 bp += NFSX_UNSIGNED; 3059 nfsm_clget; 3060 *tl = txdr_unsigned(dp->d_fileno); 3061 bp += NFSX_UNSIGNED; 3062 nfsm_clget; 3063 *tl = txdr_unsigned(nlen); 3064 bp += NFSX_UNSIGNED; 3065 3066 /* And loop around copying the name */ 3067 xfer = nlen; 3068 cp = dp->d_name; 3069 while (xfer > 0) { 3070 nfsm_clget; 3071 if ((bp + xfer) > be) 3072 tsiz = be - bp; 3073 else 3074 tsiz = xfer; 3075 memcpy(bp, cp, tsiz); 3076 bp += tsiz; 3077 xfer -= tsiz; 3078 if (xfer > 0) 3079 cp += tsiz; 3080 } 3081 /* And null pad to an int32_t boundary */ 3082 for (i = 0; i < rem; i++) 3083 *bp++ = '\0'; 3084 3085 /* 3086 * Now copy the flrep structure out. 3087 */ 3088 xfer = sizeof (struct flrep); 3089 cp = (caddr_t)&fl; 3090 while (xfer > 0) { 3091 nfsm_clget; 3092 if ((bp + xfer) > be) 3093 tsiz = be - bp; 3094 else 3095 tsiz = xfer; 3096 memcpy(bp, cp, tsiz); 3097 bp += tsiz; 3098 xfer -= tsiz; 3099 if (xfer > 0) 3100 cp += tsiz; 3101 } 3102 } 3103 invalid: 3104 cpos += dp->d_reclen; 3105 dp = (struct dirent *)cpos; 3106 cookiep++; 3107 ncookies--; 3108 } 3109 vrele(vp); 3110 nfsm_clget; 3111 *tl = nfs_false; 3112 bp += NFSX_UNSIGNED; 3113 nfsm_clget; 3114 if (eofflag) 3115 *tl = nfs_true; 3116 else 3117 *tl = nfs_false; 3118 bp += NFSX_UNSIGNED; 3119 if (mp != mb) { 3120 if (bp < be) 3121 mp->m_len = bp - mtod(mp, caddr_t); 3122 } else 3123 mp->m_len += bp - bpos; 3124 free((caddr_t)cookies, M_TEMP); 3125 free((caddr_t)rbuf, M_TEMP); 3126 nfsm_srvdone; 3127 } 3128 3129 /* 3130 * nfs commit service 3131 */ 3132 int 3133 nfsrv_commit(nfsd, slp, lwp, mrq) 3134 struct nfsrv_descript *nfsd; 3135 struct nfssvc_sock *slp; 3136 struct lwp *lwp; 3137 struct mbuf **mrq; 3138 { 3139 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3140 struct mbuf *nam = nfsd->nd_nam; 3141 caddr_t dpos = nfsd->nd_dpos; 3142 struct ucred *cred = &nfsd->nd_cr; 3143 struct vattr bfor, aft; 3144 struct vnode *vp; 3145 nfsfh_t nfh; 3146 fhandle_t *fhp; 3147 u_int32_t *tl; 3148 int32_t t1; 3149 caddr_t bpos; 3150 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cache; 3151 uint32_t cnt; 3152 char *cp2; 3153 struct mbuf *mb, *mreq; 3154 u_quad_t frev, off, end; 3155 struct mount *mp = NULL; 3156 3157 #ifndef nolint 3158 cache = 0; 3159 #endif 3160 fhp = &nfh.fh_generic; 3161 nfsm_srvmtofh(fhp); 3162 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 3163 return (ESTALE); 3164 vn_start_write(NULL, &mp, V_WAIT); 3165 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED); 3166 3167 off = fxdr_hyper(tl); 3168 tl += 2; 3169 cnt = fxdr_unsigned(uint32_t, *tl); 3170 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3171 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3172 if (error) { 3173 nfsm_reply(2 * NFSX_UNSIGNED); 3174 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3175 vn_finished_write(mp, 0); 3176 return (0); 3177 } 3178 for_ret = VOP_GETATTR(vp, &bfor, cred, lwp); 3179 end = (cnt > 0) ? off + cnt : vp->v_size; 3180 if (end < off || end > vp->v_size) 3181 end = vp->v_size; 3182 if (off < vp->v_size) 3183 error = VOP_FSYNC(vp, cred, FSYNC_WAIT, off, end, lwp); 3184 /* else error == 0, from nfsrv_fhtovp() */ 3185 aft_ret = VOP_GETATTR(vp, &aft, cred, lwp); 3186 vput(vp); 3187 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF); 3188 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3189 if (!error) { 3190 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF); 3191 *tl++ = txdr_unsigned(boottime.tv_sec); 3192 *tl = txdr_unsigned(boottime.tv_usec); 3193 } else { 3194 vn_finished_write(mp, 0); 3195 return (0); 3196 } 3197 vn_finished_write(mp, 0); 3198 nfsm_srvdone; 3199 } 3200 3201 /* 3202 * nfs statfs service 3203 */ 3204 int 3205 nfsrv_statfs(nfsd, slp, lwp, mrq) 3206 struct nfsrv_descript *nfsd; 3207 struct nfssvc_sock *slp; 3208 struct lwp *lwp; 3209 struct mbuf **mrq; 3210 { 3211 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3212 struct mbuf *nam = nfsd->nd_nam; 3213 caddr_t dpos = nfsd->nd_dpos; 3214 struct ucred *cred = &nfsd->nd_cr; 3215 struct statvfs *sf; 3216 struct nfs_statfs *sfp; 3217 u_int32_t *tl; 3218 int32_t t1; 3219 caddr_t bpos; 3220 int error = 0, rdonly, cache, getret = 1; 3221 int v3 = (nfsd->nd_flag & ND_NFSV3); 3222 char *cp2; 3223 struct mbuf *mb, *mreq; 3224 struct vnode *vp; 3225 struct vattr at; 3226 nfsfh_t nfh; 3227 fhandle_t *fhp; 3228 struct statvfs statvfs; 3229 u_quad_t frev, tval; 3230 3231 #ifndef nolint 3232 cache = 0; 3233 #endif 3234 fhp = &nfh.fh_generic; 3235 nfsm_srvmtofh(fhp); 3236 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3237 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3238 if (error) { 3239 nfsm_reply(NFSX_UNSIGNED); 3240 nfsm_srvpostop_attr(getret, &at); 3241 return (0); 3242 } 3243 sf = &statvfs; 3244 error = VFS_STATVFS(vp->v_mount, sf, lwp); 3245 getret = VOP_GETATTR(vp, &at, cred, lwp); 3246 vput(vp); 3247 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3)); 3248 if (v3) 3249 nfsm_srvpostop_attr(getret, &at); 3250 if (error) 3251 return (0); 3252 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3)); 3253 if (v3) { 3254 tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_frsize); 3255 txdr_hyper(tval, &sfp->sf_tbytes); 3256 tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_frsize); 3257 txdr_hyper(tval, &sfp->sf_fbytes); 3258 tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_frsize); 3259 txdr_hyper(tval, &sfp->sf_abytes); 3260 tval = (u_quad_t)sf->f_files; 3261 txdr_hyper(tval, &sfp->sf_tfiles); 3262 tval = (u_quad_t)sf->f_ffree; 3263 txdr_hyper(tval, &sfp->sf_ffiles); 3264 txdr_hyper(tval, &sfp->sf_afiles); 3265 sfp->sf_invarsec = 0; 3266 } else { 3267 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA); 3268 sfp->sf_bsize = txdr_unsigned(sf->f_frsize); 3269 sfp->sf_blocks = txdr_unsigned(sf->f_blocks); 3270 sfp->sf_bfree = txdr_unsigned(sf->f_bfree); 3271 sfp->sf_bavail = txdr_unsigned(sf->f_bavail); 3272 } 3273 nfsm_srvdone; 3274 } 3275 3276 /* 3277 * nfs fsinfo service 3278 */ 3279 int 3280 nfsrv_fsinfo(nfsd, slp, lwp, mrq) 3281 struct nfsrv_descript *nfsd; 3282 struct nfssvc_sock *slp; 3283 struct lwp *lwp; 3284 struct mbuf **mrq; 3285 { 3286 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3287 struct mbuf *nam = nfsd->nd_nam; 3288 caddr_t dpos = nfsd->nd_dpos; 3289 struct ucred *cred = &nfsd->nd_cr; 3290 u_int32_t *tl; 3291 struct nfsv3_fsinfo *sip; 3292 int32_t t1; 3293 caddr_t bpos; 3294 int error = 0, rdonly, cache, getret = 1; 3295 uint32_t maxdata; 3296 char *cp2; 3297 struct mbuf *mb, *mreq; 3298 struct vnode *vp; 3299 struct vattr at; 3300 nfsfh_t nfh; 3301 fhandle_t *fhp; 3302 u_quad_t frev, maxfsize; 3303 struct statvfs sb; 3304 3305 #ifndef nolint 3306 cache = 0; 3307 #endif 3308 fhp = &nfh.fh_generic; 3309 nfsm_srvmtofh(fhp); 3310 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3311 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3312 if (error) { 3313 nfsm_reply(NFSX_UNSIGNED); 3314 nfsm_srvpostop_attr(getret, &at); 3315 return (0); 3316 } 3317 3318 /* XXX Try to make a guess on the max file size. */ 3319 VFS_STATVFS(vp->v_mount, &sb, (struct lwp *)0); 3320 maxfsize = (u_quad_t)0x80000000 * sb.f_frsize - 1; 3321 3322 getret = VOP_GETATTR(vp, &at, cred, lwp); 3323 vput(vp); 3324 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO); 3325 nfsm_srvpostop_attr(getret, &at); 3326 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO); 3327 3328 /* 3329 * XXX 3330 * There should be file system VFS OP(s) to get this information. 3331 * For now, assume ufs. 3332 */ 3333 if (slp->ns_so->so_type == SOCK_DGRAM) 3334 maxdata = NFS_MAXDGRAMDATA; 3335 else 3336 maxdata = NFS_MAXDATA; 3337 sip->fs_rtmax = txdr_unsigned(maxdata); 3338 sip->fs_rtpref = txdr_unsigned(maxdata); 3339 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE); 3340 sip->fs_wtmax = txdr_unsigned(maxdata); 3341 sip->fs_wtpref = txdr_unsigned(maxdata); 3342 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE); 3343 sip->fs_dtpref = txdr_unsigned(maxdata); 3344 txdr_hyper(maxfsize, &sip->fs_maxfilesize); 3345 sip->fs_timedelta.nfsv3_sec = 0; 3346 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1); 3347 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK | 3348 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS | 3349 NFSV3FSINFO_CANSETTIME); 3350 nfsm_srvdone; 3351 } 3352 3353 /* 3354 * nfs pathconf service 3355 */ 3356 int 3357 nfsrv_pathconf(nfsd, slp, lwp, mrq) 3358 struct nfsrv_descript *nfsd; 3359 struct nfssvc_sock *slp; 3360 struct lwp *lwp; 3361 struct mbuf **mrq; 3362 { 3363 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3364 struct mbuf *nam = nfsd->nd_nam; 3365 caddr_t dpos = nfsd->nd_dpos; 3366 struct ucred *cred = &nfsd->nd_cr; 3367 u_int32_t *tl; 3368 struct nfsv3_pathconf *pc; 3369 int32_t t1; 3370 caddr_t bpos; 3371 int error = 0, rdonly, cache, getret = 1; 3372 register_t linkmax, namemax, chownres, notrunc; 3373 char *cp2; 3374 struct mbuf *mb, *mreq; 3375 struct vnode *vp; 3376 struct vattr at; 3377 nfsfh_t nfh; 3378 fhandle_t *fhp; 3379 u_quad_t frev; 3380 3381 #ifndef nolint 3382 cache = 0; 3383 #endif 3384 fhp = &nfh.fh_generic; 3385 nfsm_srvmtofh(fhp); 3386 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3387 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3388 if (error) { 3389 nfsm_reply(NFSX_UNSIGNED); 3390 nfsm_srvpostop_attr(getret, &at); 3391 return (0); 3392 } 3393 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax); 3394 if (!error) 3395 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax); 3396 if (!error) 3397 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres); 3398 if (!error) 3399 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc); 3400 getret = VOP_GETATTR(vp, &at, cred, lwp); 3401 vput(vp); 3402 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF); 3403 nfsm_srvpostop_attr(getret, &at); 3404 if (error) 3405 return (0); 3406 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF); 3407 3408 pc->pc_linkmax = txdr_unsigned(linkmax); 3409 pc->pc_namemax = txdr_unsigned(namemax); 3410 pc->pc_notrunc = txdr_unsigned(notrunc); 3411 pc->pc_chownrestricted = txdr_unsigned(chownres); 3412 3413 /* 3414 * These should probably be supported by VOP_PATHCONF(), but 3415 * until msdosfs is exportable (why would you want to?), the 3416 * Unix defaults should be ok. 3417 */ 3418 pc->pc_caseinsensitive = nfs_false; 3419 pc->pc_casepreserving = nfs_true; 3420 nfsm_srvdone; 3421 } 3422 3423 /* 3424 * Null operation, used by clients to ping server 3425 */ 3426 /* ARGSUSED */ 3427 int 3428 nfsrv_null(nfsd, slp, lwp, mrq) 3429 struct nfsrv_descript *nfsd; 3430 struct nfssvc_sock *slp; 3431 struct lwp *lwp; 3432 struct mbuf **mrq; 3433 { 3434 struct mbuf *mrep = nfsd->nd_mrep; 3435 caddr_t bpos; 3436 int error = NFSERR_RETVOID, cache = 0; 3437 struct mbuf *mb, *mreq; 3438 u_quad_t frev; 3439 3440 nfsm_reply(0); 3441 return (0); 3442 } 3443 3444 /* 3445 * No operation, used for obsolete procedures 3446 */ 3447 /* ARGSUSED */ 3448 int 3449 nfsrv_noop(nfsd, slp, lwp, mrq) 3450 struct nfsrv_descript *nfsd; 3451 struct nfssvc_sock *slp; 3452 struct lwp *lwp; 3453 struct mbuf **mrq; 3454 { 3455 struct mbuf *mrep = nfsd->nd_mrep; 3456 caddr_t bpos; 3457 int error, cache = 0; 3458 struct mbuf *mb, *mreq; 3459 u_quad_t frev; 3460 3461 if (nfsd->nd_repstat) 3462 error = nfsd->nd_repstat; 3463 else 3464 error = EPROCUNAVAIL; 3465 nfsm_reply(0); 3466 return (0); 3467 } 3468 3469 /* 3470 * Perform access checking for vnodes obtained from file handles that would 3471 * refer to files already opened by a Unix client. You cannot just use 3472 * vn_writechk() and VOP_ACCESS() for two reasons. 3473 * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case 3474 * 2 - The owner is to be given access irrespective of mode bits for some 3475 * operations, so that processes that chmod after opening a file don't 3476 * break. I don't like this because it opens a security hole, but since 3477 * the nfs server opens a security hole the size of a barn door anyhow, 3478 * what the heck. 3479 * 3480 * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS() 3481 * will return EPERM instead of EACCESS. EPERM is always an error. 3482 */ 3483 int 3484 nfsrv_access(vp, flags, cred, rdonly, lwp, override) 3485 struct vnode *vp; 3486 int flags; 3487 struct ucred *cred; 3488 int rdonly; 3489 struct lwp *lwp; 3490 { 3491 struct vattr vattr; 3492 int error; 3493 if (flags & VWRITE) { 3494 /* Just vn_writechk() changed to check rdonly */ 3495 /* 3496 * Disallow write attempts on read-only file systems; 3497 * unless the file is a socket or a block or character 3498 * device resident on the file system. 3499 */ 3500 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 3501 switch (vp->v_type) { 3502 case VREG: 3503 case VDIR: 3504 case VLNK: 3505 return (EROFS); 3506 default: 3507 break; 3508 } 3509 } 3510 3511 /* 3512 * If the vnode is in use as a process's text, 3513 * we can't allow writing. 3514 */ 3515 if (vp->v_flag & VTEXT) 3516 return (ETXTBSY); 3517 } 3518 error = VOP_GETATTR(vp, &vattr, cred, lwp); 3519 if (error) 3520 return (error); 3521 error = VOP_ACCESS(vp, flags, cred, lwp); 3522 /* 3523 * Allow certain operations for the owner (reads and writes 3524 * on files that are already open). 3525 */ 3526 if (override && error == EACCES && cred->cr_uid == vattr.va_uid) 3527 error = 0; 3528 return error; 3529 } 3530
Indexes created Mon Sep 22 21:09:42 GMT 2025