nfs_serv.c revision 1.95.2.1
1 /* $NetBSD: nfs_serv.c,v 1.95.2.1 2005/07/07 13:03:20 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 1.95.2.1 2005/07/07 13:03:20 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 #include <ufs/ufs/dir.h> 74 75 #include <uvm/uvm.h> 76 77 #include <nfs/nfsproto.h> 78 #include <nfs/rpcv2.h> 79 #include <nfs/nfs.h> 80 #include <nfs/xdr_subs.h> 81 #include <nfs/nfsm_subs.h> 82 #include <nfs/nqnfs.h> 83 #include <nfs/nfs_var.h> 84 85 /* Global vars */ 86 extern u_int32_t nfs_xdrneg1; 87 extern u_int32_t nfs_false, nfs_true; 88 extern enum vtype nv3tov_type[8]; 89 extern struct nfsstats nfsstats; 90 extern const nfstype nfsv2_type[9]; 91 extern const nfstype nfsv3_type[9]; 92 int nfsrvw_procrastinate = NFS_GATHERDELAY * 1000; 93 int nfsd_use_loan = 1; /* use page-loan for READ OP */ 94 95 /* 96 * nfs v3 access service 97 */ 98 int 99 nfsrv3_access(nfsd, slp, procp, mrq) 100 struct nfsrv_descript *nfsd; 101 struct nfssvc_sock *slp; 102 struct proc *procp; 103 struct mbuf **mrq; 104 { 105 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 106 struct mbuf *nam = nfsd->nd_nam; 107 caddr_t dpos = nfsd->nd_dpos; 108 struct ucred *cred = &nfsd->nd_cr; 109 struct vnode *vp; 110 nfsfh_t nfh; 111 fhandle_t *fhp; 112 u_int32_t *tl; 113 int32_t t1; 114 caddr_t bpos; 115 int error = 0, rdonly, cache = 0, getret; 116 char *cp2; 117 struct mbuf *mb, *mreq; 118 struct vattr va; 119 u_long inmode, testmode, outmode; 120 u_quad_t frev; 121 122 fhp = &nfh.fh_generic; 123 nfsm_srvmtofh(fhp); 124 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 125 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, 126 (nfsd->nd_flag & ND_KERBAUTH), FALSE); 127 if (error) { 128 nfsm_reply(NFSX_UNSIGNED); 129 nfsm_srvpostop_attr(1, (struct vattr *)0); 130 return (0); 131 } 132 inmode = fxdr_unsigned(u_int32_t, *tl); 133 outmode = 0; 134 if ((inmode & NFSV3ACCESS_READ) && 135 nfsrv_access(vp, VREAD, cred, rdonly, procp, 0) == 0) 136 outmode |= NFSV3ACCESS_READ; 137 if (vp->v_type != VDIR) { 138 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND); 139 if (testmode && 140 nfsrv_access(vp, VWRITE, cred, rdonly, procp, 0) == 0) 141 outmode |= testmode; 142 if ((inmode & NFSV3ACCESS_EXECUTE) && 143 nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0) == 0) 144 outmode |= NFSV3ACCESS_EXECUTE; 145 } else { 146 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND | 147 NFSV3ACCESS_DELETE); 148 if (testmode && 149 nfsrv_access(vp, VWRITE, cred, rdonly, procp, 0) == 0) 150 outmode |= testmode; 151 if ((inmode & NFSV3ACCESS_LOOKUP) && 152 nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0) == 0) 153 outmode |= NFSV3ACCESS_LOOKUP; 154 } 155 getret = VOP_GETATTR(vp, &va, cred, procp); 156 vput(vp); 157 nfsm_reply(NFSX_POSTOPATTR(1) + NFSX_UNSIGNED); 158 nfsm_srvpostop_attr(getret, &va); 159 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 160 *tl = txdr_unsigned(outmode); 161 nfsm_srvdone; 162 } 163 164 /* 165 * nfs getattr service 166 */ 167 int 168 nfsrv_getattr(nfsd, slp, procp, mrq) 169 struct nfsrv_descript *nfsd; 170 struct nfssvc_sock *slp; 171 struct proc *procp; 172 struct mbuf **mrq; 173 { 174 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 175 struct mbuf *nam = nfsd->nd_nam; 176 caddr_t dpos = nfsd->nd_dpos; 177 struct ucred *cred = &nfsd->nd_cr; 178 struct nfs_fattr *fp; 179 struct vattr va; 180 struct vnode *vp; 181 nfsfh_t nfh; 182 fhandle_t *fhp; 183 u_int32_t *tl; 184 int32_t t1; 185 caddr_t bpos; 186 int error = 0, rdonly, cache; 187 char *cp2; 188 struct mbuf *mb, *mreq; 189 u_quad_t frev; 190 191 fhp = &nfh.fh_generic; 192 nfsm_srvmtofh(fhp); 193 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, 194 (nfsd->nd_flag & ND_KERBAUTH), FALSE); 195 if (error) { 196 nfsm_reply(0); 197 return (0); 198 } 199 nqsrv_getl(vp, ND_READ); 200 error = VOP_GETATTR(vp, &va, cred, procp); 201 vput(vp); 202 nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); 203 if (error) 204 return (0); 205 nfsm_build(fp, struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); 206 nfsm_srvfillattr(&va, fp); 207 nfsm_srvdone; 208 } 209 210 /* 211 * nfs setattr service 212 */ 213 int 214 nfsrv_setattr(nfsd, slp, procp, mrq) 215 struct nfsrv_descript *nfsd; 216 struct nfssvc_sock *slp; 217 struct proc *procp; 218 struct mbuf **mrq; 219 { 220 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 221 struct mbuf *nam = nfsd->nd_nam; 222 caddr_t dpos = nfsd->nd_dpos; 223 struct ucred *cred = &nfsd->nd_cr; 224 struct vattr va, preat; 225 struct nfsv2_sattr *sp; 226 struct nfs_fattr *fp; 227 struct vnode *vp; 228 nfsfh_t nfh; 229 fhandle_t *fhp; 230 u_int32_t *tl; 231 int32_t t1; 232 caddr_t bpos; 233 int error = 0, rdonly, cache, preat_ret = 1, postat_ret = 1; 234 int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0; 235 char *cp2; 236 struct mbuf *mb, *mreq; 237 u_quad_t frev; 238 struct timespec guard; 239 struct mount *mp = NULL; 240 241 fhp = &nfh.fh_generic; 242 nfsm_srvmtofh(fhp); 243 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 244 return (ESTALE); 245 vn_start_write(NULL, &mp, V_WAIT); 246 VATTR_NULL(&va); 247 if (v3) { 248 nfsm_srvsattr(&va); 249 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 250 gcheck = fxdr_unsigned(int, *tl); 251 if (gcheck) { 252 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 253 fxdr_nfsv3time(tl, &guard); 254 } 255 } else { 256 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 257 /* 258 * Nah nah nah nah na nah 259 * There is a bug in the Sun client that puts 0xffff in the mode 260 * field of sattr when it should put in 0xffffffff. The u_short 261 * doesn't sign extend. 262 * --> check the low order 2 bytes for 0xffff 263 */ 264 if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff) 265 va.va_mode = nfstov_mode(sp->sa_mode); 266 if (sp->sa_uid != nfs_xdrneg1) 267 va.va_uid = fxdr_unsigned(uid_t, sp->sa_uid); 268 if (sp->sa_gid != nfs_xdrneg1) 269 va.va_gid = fxdr_unsigned(gid_t, sp->sa_gid); 270 if (sp->sa_size != nfs_xdrneg1) 271 va.va_size = fxdr_unsigned(u_quad_t, sp->sa_size); 272 if (sp->sa_atime.nfsv2_sec != nfs_xdrneg1) { 273 #ifdef notyet 274 fxdr_nfsv2time(&sp->sa_atime, &va.va_atime); 275 #else 276 va.va_atime.tv_sec = 277 fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec); 278 va.va_atime.tv_nsec = 0; 279 #endif 280 } 281 if (sp->sa_mtime.nfsv2_sec != nfs_xdrneg1) 282 fxdr_nfsv2time(&sp->sa_mtime, &va.va_mtime); 283 284 } 285 286 /* 287 * Now that we have all the fields, lets do it. 288 */ 289 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, 290 (nfsd->nd_flag & ND_KERBAUTH), FALSE); 291 if (error) { 292 nfsm_reply(2 * NFSX_UNSIGNED); 293 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 294 vn_finished_write(mp, 0); 295 return (0); 296 } 297 nqsrv_getl(vp, ND_WRITE); 298 if (v3) { 299 error = preat_ret = VOP_GETATTR(vp, &preat, cred, procp); 300 if (!error && gcheck && 301 (preat.va_ctime.tv_sec != guard.tv_sec || 302 preat.va_ctime.tv_nsec != guard.tv_nsec)) 303 error = NFSERR_NOT_SYNC; 304 if (error) { 305 vput(vp); 306 nfsm_reply(NFSX_WCCDATA(v3)); 307 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 308 vn_finished_write(mp, 0); 309 return (0); 310 } 311 } 312 313 /* 314 * If the size is being changed write acces is required, otherwise 315 * just check for a read only file system. 316 */ 317 if (va.va_size == ((u_quad_t)((quad_t) -1))) { 318 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 319 error = EROFS; 320 goto out; 321 } 322 } else { 323 if (vp->v_type == VDIR) { 324 error = EISDIR; 325 goto out; 326 } else if ((error = nfsrv_access(vp, VWRITE, cred, rdonly, 327 procp, 0)) != 0) 328 goto out; 329 } 330 error = VOP_SETATTR(vp, &va, cred, procp); 331 postat_ret = VOP_GETATTR(vp, &va, cred, procp); 332 if (!error) 333 error = postat_ret; 334 out: 335 vput(vp); 336 vn_finished_write(mp, 0); 337 nfsm_reply(NFSX_WCCORFATTR(v3)); 338 if (v3) { 339 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 340 return (0); 341 } else { 342 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 343 nfsm_srvfillattr(&va, fp); 344 } 345 nfsm_srvdone; 346 } 347 348 /* 349 * nfs lookup rpc 350 */ 351 int 352 nfsrv_lookup(nfsd, slp, procp, mrq) 353 struct nfsrv_descript *nfsd; 354 struct nfssvc_sock *slp; 355 struct proc *procp; 356 struct mbuf **mrq; 357 { 358 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 359 struct mbuf *nam = nfsd->nd_nam; 360 caddr_t dpos = nfsd->nd_dpos; 361 struct ucred *cred = &nfsd->nd_cr; 362 struct nfs_fattr *fp; 363 struct nameidata nd, ind, *ndp = &nd; 364 struct vnode *vp, *dirp; 365 nfsfh_t nfh; 366 fhandle_t *fhp; 367 caddr_t cp; 368 u_int32_t *tl; 369 int32_t t1; 370 caddr_t bpos; 371 int error = 0, cache, dirattr_ret = 1; 372 uint32_t len; 373 int v3 = (nfsd->nd_flag & ND_NFSV3), pubflag; 374 char *cp2; 375 struct mbuf *mb, *mreq; 376 struct vattr va, dirattr; 377 u_quad_t frev; 378 379 fhp = &nfh.fh_generic; 380 nfsm_srvmtofh(fhp); 381 nfsm_srvnamesiz(len); 382 383 pubflag = nfs_ispublicfh(fhp); 384 385 nd.ni_cnd.cn_cred = cred; 386 nd.ni_cnd.cn_nameiop = LOOKUP; 387 nd.ni_cnd.cn_flags = LOCKLEAF | SAVESTART; 388 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 389 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), pubflag); 390 391 if (!error && pubflag) { 392 if (nd.ni_vp->v_type == VDIR && nfs_pub.np_index != NULL) { 393 /* 394 * Setup call to lookup() to see if we can find 395 * the index file. Arguably, this doesn't belong 396 * in a kernel.. Ugh. 397 */ 398 ind = nd; 399 VOP_UNLOCK(nd.ni_vp, 0); 400 ind.ni_pathlen = strlen(nfs_pub.np_index); 401 ind.ni_cnd.cn_nameptr = ind.ni_cnd.cn_pnbuf = 402 nfs_pub.np_index; 403 ind.ni_startdir = nd.ni_vp; 404 VREF(ind.ni_startdir); 405 error = lookup(&ind); 406 if (!error) { 407 /* 408 * Found an index file. Get rid of 409 * the old references. 410 */ 411 if (dirp) 412 vrele(dirp); 413 dirp = nd.ni_vp; 414 vrele(nd.ni_startdir); 415 ndp = &ind; 416 } else 417 error = 0; 418 } 419 /* 420 * If the public filehandle was used, check that this lookup 421 * didn't result in a filehandle outside the publicly exported 422 * filesystem. 423 */ 424 425 if (!error && ndp->ni_vp->v_mount != nfs_pub.np_mount) { 426 vput(nd.ni_vp); 427 error = EPERM; 428 } 429 } 430 431 if (dirp) { 432 if (v3) 433 dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred, 434 procp); 435 vrele(dirp); 436 } 437 438 if (error) { 439 nfsm_reply(NFSX_POSTOPATTR(v3)); 440 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 441 return (0); 442 } 443 444 nqsrv_getl(ndp->ni_startdir, ND_READ); 445 vrele(ndp->ni_startdir); 446 PNBUF_PUT(nd.ni_cnd.cn_pnbuf); 447 vp = ndp->ni_vp; 448 memset((caddr_t)fhp, 0, sizeof(nfh)); 449 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 450 error = VFS_VPTOFH(vp, &fhp->fh_fid); 451 if (!error) 452 error = VOP_GETATTR(vp, &va, cred, procp); 453 vput(vp); 454 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPORFATTR(v3) + NFSX_POSTOPATTR(v3)); 455 if (error) { 456 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 457 return (0); 458 } 459 nfsm_srvfhtom(fhp, v3); 460 if (v3) { 461 nfsm_srvpostop_attr(0, &va); 462 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 463 } else { 464 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 465 nfsm_srvfillattr(&va, fp); 466 } 467 nfsm_srvdone; 468 } 469 470 /* 471 * nfs readlink service 472 */ 473 int 474 nfsrv_readlink(nfsd, slp, procp, mrq) 475 struct nfsrv_descript *nfsd; 476 struct nfssvc_sock *slp; 477 struct proc *procp; 478 struct mbuf **mrq; 479 { 480 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 481 struct mbuf *nam = nfsd->nd_nam; 482 caddr_t dpos = nfsd->nd_dpos; 483 struct ucred *cred = &nfsd->nd_cr; 484 struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN]; 485 struct iovec *ivp = iv; 486 struct mbuf *mp; 487 u_int32_t *tl; 488 int32_t t1; 489 caddr_t bpos; 490 int error = 0, rdonly, cache, i, padlen, getret; 491 uint32_t len; 492 int v3 = (nfsd->nd_flag & ND_NFSV3); 493 char *cp2; 494 struct mbuf *mb, *mp2 = NULL, *mp3 = NULL, *mreq; 495 struct vnode *vp; 496 struct vattr attr; 497 nfsfh_t nfh; 498 fhandle_t *fhp; 499 struct uio io, *uiop = &io; 500 u_quad_t frev; 501 502 fhp = &nfh.fh_generic; 503 nfsm_srvmtofh(fhp); 504 len = 0; 505 i = 0; 506 while (len < NFS_MAXPATHLEN) { 507 mp = m_get(M_WAIT, MT_DATA); 508 MCLAIM(mp, &nfs_mowner); 509 m_clget(mp, M_WAIT); 510 mp->m_len = NFSMSIZ(mp); 511 if (len == 0) 512 mp3 = mp2 = mp; 513 else { 514 mp2->m_next = mp; 515 mp2 = mp; 516 } 517 if ((len+mp->m_len) > NFS_MAXPATHLEN) { 518 mp->m_len = NFS_MAXPATHLEN-len; 519 len = NFS_MAXPATHLEN; 520 } else 521 len += mp->m_len; 522 ivp->iov_base = mtod(mp, caddr_t); 523 ivp->iov_len = mp->m_len; 524 i++; 525 ivp++; 526 } 527 uiop->uio_iov = iv; 528 uiop->uio_iovcnt = i; 529 uiop->uio_offset = 0; 530 uiop->uio_resid = len; 531 uiop->uio_rw = UIO_READ; 532 uiop->uio_segflg = UIO_SYSSPACE; 533 uiop->uio_procp = NULL; 534 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 535 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 536 if (error) { 537 m_freem(mp3); 538 nfsm_reply(2 * NFSX_UNSIGNED); 539 nfsm_srvpostop_attr(1, (struct vattr *)0); 540 return (0); 541 } 542 if (vp->v_type != VLNK) { 543 if (v3) 544 error = EINVAL; 545 else 546 error = ENXIO; 547 goto out; 548 } 549 nqsrv_getl(vp, ND_READ); 550 error = VOP_READLINK(vp, uiop, cred); 551 out: 552 getret = VOP_GETATTR(vp, &attr, cred, procp); 553 vput(vp); 554 if (error) 555 m_freem(mp3); 556 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED); 557 if (v3) { 558 nfsm_srvpostop_attr(getret, &attr); 559 if (error) 560 return (0); 561 } 562 len -= uiop->uio_resid; 563 padlen = nfsm_padlen(len); 564 if (uiop->uio_resid || padlen) 565 nfs_zeropad(mp3, uiop->uio_resid, padlen); 566 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 567 *tl = txdr_unsigned(len); 568 mb->m_next = mp3; 569 nfsm_srvdone; 570 } 571 572 /* 573 * nfs read service 574 */ 575 int 576 nfsrv_read(nfsd, slp, procp, mrq) 577 struct nfsrv_descript *nfsd; 578 struct nfssvc_sock *slp; 579 struct proc *procp; 580 struct mbuf **mrq; 581 { 582 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 583 struct mbuf *nam = nfsd->nd_nam; 584 caddr_t dpos = nfsd->nd_dpos; 585 struct ucred *cred = &nfsd->nd_cr; 586 struct mbuf *m; 587 struct nfs_fattr *fp; 588 u_int32_t *tl; 589 int32_t t1; 590 int i; 591 caddr_t bpos; 592 int error = 0, rdonly, cache, getret; 593 int v3 = (nfsd->nd_flag & ND_NFSV3); 594 uint32_t reqlen, len, cnt, left; 595 int padlen; 596 char *cp2; 597 struct mbuf *mb, *mreq; 598 struct vnode *vp; 599 nfsfh_t nfh; 600 fhandle_t *fhp; 601 struct uio io, *uiop = &io; 602 struct vattr va; 603 off_t off; 604 u_quad_t frev; 605 606 fhp = &nfh.fh_generic; 607 nfsm_srvmtofh(fhp); 608 if (v3) { 609 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 610 off = fxdr_hyper(tl); 611 } else { 612 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 613 off = (off_t)fxdr_unsigned(u_int32_t, *tl); 614 } 615 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 616 reqlen = fxdr_unsigned(uint32_t, *tl); 617 reqlen = MIN(reqlen, NFS_SRVMAXDATA(nfsd)); 618 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 619 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 620 if (error) { 621 nfsm_reply(2 * NFSX_UNSIGNED); 622 nfsm_srvpostop_attr(1, (struct vattr *)0); 623 return (0); 624 } 625 if (vp->v_type != VREG) { 626 if (v3) 627 error = EINVAL; 628 else 629 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 630 } 631 if (!error) { 632 nqsrv_getl(vp, ND_READ); 633 if ((error = nfsrv_access(vp, VREAD, cred, rdonly, procp, 1)) != 0) 634 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 1); 635 } 636 getret = VOP_GETATTR(vp, &va, cred, procp); 637 if (!error) 638 error = getret; 639 if (error) { 640 vput(vp); 641 nfsm_reply(NFSX_POSTOPATTR(v3)); 642 nfsm_srvpostop_attr(getret, &va); 643 return (0); 644 } 645 if (off >= va.va_size) 646 cnt = 0; 647 else if ((off + reqlen) > va.va_size) 648 cnt = va.va_size - off; 649 else 650 cnt = reqlen; 651 nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt)); 652 if (v3) { 653 nfsm_build(tl, u_int32_t *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED); 654 *tl++ = nfs_true; 655 fp = (struct nfs_fattr *)tl; 656 tl += (NFSX_V3FATTR / sizeof (u_int32_t)); 657 } else { 658 nfsm_build(tl, u_int32_t *, NFSX_V2FATTR + NFSX_UNSIGNED); 659 fp = (struct nfs_fattr *)tl; 660 tl += (NFSX_V2FATTR / sizeof (u_int32_t)); 661 } 662 len = left = cnt; 663 if (cnt > 0) { 664 if (nfsd_use_loan) { 665 struct vm_page **pgpp; 666 voff_t pgoff = trunc_page(off); 667 int npages; 668 vaddr_t lva; 669 670 npages = (round_page(off + cnt) - pgoff) >> PAGE_SHIFT; 671 KASSERT(npages <= M_EXT_MAXPAGES); /* XXX */ 672 673 /* allocate kva for mbuf data */ 674 lva = sokvaalloc(npages << PAGE_SHIFT, slp->ns_so); 675 if (lva == 0) { 676 /* fall back to VOP_READ */ 677 goto loan_fail; 678 } 679 680 /* allocate mbuf */ 681 m = m_get(M_WAIT, MT_DATA); 682 MCLAIM(m, &nfs_mowner); 683 pgpp = m->m_ext.ext_pgs; 684 685 /* loan pages */ 686 error = uvm_loanuobjpages(&vp->v_uobj, pgoff, npages, 687 pgpp); 688 if (error) { 689 sokvafree(lva, npages << PAGE_SHIFT); 690 m_free(m); 691 if (error == EBUSY) 692 goto loan_fail; 693 goto read_error; 694 } 695 696 /* associate kva to mbuf */ 697 MEXTADD(m, (void *)(lva + ((vaddr_t)off & PAGE_MASK)), 698 cnt, M_MBUF, soloanfree, slp->ns_so); 699 m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP; 700 m->m_len = cnt; 701 702 #if defined(__HAVE_LAZY_MBUF) 703 m->m_flags |= M_EXT_LAZY; 704 m->m_ext.ext_flags |= M_EXT_LAZY; 705 #else /* defined(__HAVE_LAZY_MBUF) */ 706 /* map pages */ 707 for (i = 0; i < npages; i++) { 708 pmap_kenter_pa(lva, VM_PAGE_TO_PHYS(pgpp[i]), 709 VM_PROT_READ); 710 lva += PAGE_SIZE; 711 } 712 713 pmap_update(pmap_kernel()); 714 #endif /* defined(__HAVE_LAZY_MBUF) */ 715 716 mb->m_next = m; 717 mb = m; 718 error = 0; 719 uiop->uio_resid = 0; 720 } else { 721 struct iovec *iv; 722 struct iovec *iv2; 723 struct mbuf *m2; 724 int siz; 725 loan_fail: 726 /* 727 * Generate the mbuf list with the uio_iov ref. to it. 728 */ 729 i = 0; 730 m = m2 = mb; 731 while (left > 0) { 732 siz = min(M_TRAILINGSPACE(m), left); 733 if (siz > 0) { 734 left -= siz; 735 i++; 736 } 737 if (left > 0) { 738 m = m_get(M_WAIT, MT_DATA); 739 MCLAIM(m, &nfs_mowner); 740 m_clget(m, M_WAIT); 741 m->m_len = 0; 742 m2->m_next = m; 743 m2 = m; 744 } 745 } 746 iv = malloc(i * sizeof(struct iovec), M_TEMP, M_WAITOK); 747 uiop->uio_iov = iv2 = iv; 748 m = mb; 749 left = cnt; 750 i = 0; 751 while (left > 0) { 752 if (m == NULL) 753 panic("nfsrv_read iov"); 754 siz = min(M_TRAILINGSPACE(m), left); 755 if (siz > 0) { 756 iv->iov_base = mtod(m, caddr_t) + 757 m->m_len; 758 iv->iov_len = siz; 759 m->m_len += siz; 760 left -= siz; 761 iv++; 762 i++; 763 } 764 m = m->m_next; 765 } 766 uiop->uio_iovcnt = i; 767 uiop->uio_offset = off; 768 uiop->uio_resid = cnt; 769 uiop->uio_rw = UIO_READ; 770 uiop->uio_segflg = UIO_SYSSPACE; 771 error = VOP_READ(vp, uiop, IO_NODELOCKED, cred); 772 free((caddr_t)iv2, M_TEMP); 773 } 774 read_error: 775 if (error || (getret = VOP_GETATTR(vp, &va, cred, procp)) != 0){ 776 if (!error) 777 error = getret; 778 m_freem(mreq); 779 vput(vp); 780 nfsm_reply(NFSX_POSTOPATTR(v3)); 781 nfsm_srvpostop_attr(getret, &va); 782 return (0); 783 } 784 } else { 785 uiop->uio_resid = 0; 786 } 787 vput(vp); 788 nfsm_srvfillattr(&va, fp); 789 len -= uiop->uio_resid; 790 padlen = nfsm_padlen(len); 791 if (uiop->uio_resid || padlen) 792 nfs_zeropad(mb, uiop->uio_resid, padlen); 793 if (v3) { 794 /* count */ 795 *tl++ = txdr_unsigned(len); 796 /* eof */ 797 if (off + len >= va.va_size) 798 *tl++ = nfs_true; 799 else 800 *tl++ = nfs_false; 801 } 802 *tl = txdr_unsigned(len); 803 nfsm_srvdone; 804 } 805 806 /* 807 * nfs write service 808 */ 809 int 810 nfsrv_write(nfsd, slp, procp, mrq) 811 struct nfsrv_descript *nfsd; 812 struct nfssvc_sock *slp; 813 struct proc *procp; 814 struct mbuf **mrq; 815 { 816 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 817 struct mbuf *nam = nfsd->nd_nam; 818 caddr_t dpos = nfsd->nd_dpos; 819 struct ucred *cred = &nfsd->nd_cr; 820 struct iovec *ivp; 821 int i, cnt; 822 struct mbuf *mp; 823 struct nfs_fattr *fp; 824 struct iovec *iv; 825 struct vattr va, forat; 826 u_int32_t *tl; 827 int32_t t1; 828 caddr_t bpos; 829 int error = 0, rdonly, cache, len, forat_ret = 1; 830 int ioflags, aftat_ret = 1, retlen, zeroing, adjust; 831 int stable = NFSV3WRITE_FILESYNC; 832 int v3 = (nfsd->nd_flag & ND_NFSV3); 833 char *cp2; 834 struct mbuf *mb, *mreq; 835 struct vnode *vp; 836 nfsfh_t nfh; 837 fhandle_t *fhp; 838 struct uio io, *uiop = &io; 839 off_t off; 840 u_quad_t frev; 841 struct mount *mntp = NULL; 842 843 if (mrep == NULL) { 844 *mrq = NULL; 845 return (0); 846 } 847 fhp = &nfh.fh_generic; 848 nfsm_srvmtofh(fhp); 849 if ((mntp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 850 return (ESTALE); 851 vn_start_write(NULL, &mntp, V_WAIT); 852 if (v3) { 853 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 854 off = fxdr_hyper(tl); 855 tl += 3; 856 stable = fxdr_unsigned(int, *tl++); 857 } else { 858 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 859 off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 860 tl += 2; 861 } 862 retlen = len = fxdr_unsigned(int32_t, *tl); 863 cnt = i = 0; 864 865 /* 866 * For NFS Version 2, it is not obvious what a write of zero length 867 * should do, but I might as well be consistent with Version 3, 868 * which is to return ok so long as there are no permission problems. 869 */ 870 if (len > 0) { 871 zeroing = 1; 872 mp = mrep; 873 while (mp) { 874 if (mp == md) { 875 zeroing = 0; 876 adjust = dpos - mtod(mp, caddr_t); 877 mp->m_len -= adjust; 878 if (mp->m_len > 0 && adjust > 0) 879 NFSMADV(mp, adjust); 880 } 881 if (zeroing) 882 mp->m_len = 0; 883 else if (mp->m_len > 0) { 884 i += mp->m_len; 885 if (i > len) { 886 mp->m_len -= (i - len); 887 zeroing = 1; 888 } 889 if (mp->m_len > 0) 890 cnt++; 891 } 892 mp = mp->m_next; 893 } 894 } 895 if (len > NFS_MAXDATA || len < 0 || i < len) { 896 error = EIO; 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 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 903 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 904 if (error) { 905 nfsm_reply(2 * NFSX_UNSIGNED); 906 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 907 vn_finished_write(mntp, 0); 908 return (0); 909 } 910 if (v3) 911 forat_ret = VOP_GETATTR(vp, &forat, cred, procp); 912 if (vp->v_type != VREG) { 913 if (v3) 914 error = EINVAL; 915 else 916 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 917 } 918 if (!error) { 919 nqsrv_getl(vp, ND_WRITE); 920 error = nfsrv_access(vp, VWRITE, cred, rdonly, procp, 1); 921 } 922 if (error) { 923 vput(vp); 924 nfsm_reply(NFSX_WCCDATA(v3)); 925 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 926 vn_finished_write(mntp, 0); 927 return (0); 928 } 929 930 if (len > 0) { 931 ivp = malloc(cnt * sizeof (struct iovec), M_TEMP, M_WAITOK); 932 uiop->uio_iov = iv = ivp; 933 uiop->uio_iovcnt = cnt; 934 mp = mrep; 935 while (mp) { 936 if (mp->m_len > 0) { 937 ivp->iov_base = mtod(mp, caddr_t); 938 ivp->iov_len = mp->m_len; 939 ivp++; 940 } 941 mp = mp->m_next; 942 } 943 944 /* 945 * XXX 946 * The IO_METASYNC flag indicates that all metadata (and not 947 * just enough to ensure data integrity) must be written to 948 * stable storage synchronously. 949 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.) 950 */ 951 if (stable == NFSV3WRITE_UNSTABLE) 952 ioflags = IO_NODELOCKED; 953 else if (stable == NFSV3WRITE_DATASYNC) 954 ioflags = (IO_SYNC | IO_NODELOCKED); 955 else 956 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 957 uiop->uio_resid = len; 958 uiop->uio_rw = UIO_WRITE; 959 uiop->uio_segflg = UIO_SYSSPACE; 960 uiop->uio_procp = NULL; 961 uiop->uio_offset = off; 962 error = VOP_WRITE(vp, uiop, ioflags, cred); 963 nfsstats.srvvop_writes++; 964 free(iv, M_TEMP); 965 } 966 aftat_ret = VOP_GETATTR(vp, &va, cred, procp); 967 vput(vp); 968 vn_finished_write(mntp, 0); 969 if (!error) 970 error = aftat_ret; 971 nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) + 972 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3)); 973 if (v3) { 974 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 975 if (error) 976 return (0); 977 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 978 *tl++ = txdr_unsigned(retlen); 979 if (stable == NFSV3WRITE_UNSTABLE) 980 *tl++ = txdr_unsigned(stable); 981 else 982 *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC); 983 /* 984 * Actually, there is no need to txdr these fields, 985 * but it may make the values more human readable, 986 * for debugging purposes. 987 */ 988 *tl++ = txdr_unsigned(boottime.tv_sec); 989 *tl = txdr_unsigned(boottime.tv_usec); 990 } else { 991 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 992 nfsm_srvfillattr(&va, fp); 993 } 994 nfsm_srvdone; 995 } 996 997 /* 998 * NFS write service with write gathering support. Called when 999 * nfsrvw_procrastinate > 0. 1000 * See: Chet Juszczak, "Improving the Write Performance of an NFS Server", 1001 * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco, 1002 * Jan. 1994. 1003 */ 1004 int 1005 nfsrv_writegather(ndp, slp, procp, mrq) 1006 struct nfsrv_descript **ndp; 1007 struct nfssvc_sock *slp; 1008 struct proc *procp; 1009 struct mbuf **mrq; 1010 { 1011 struct iovec *ivp; 1012 struct mbuf *mp; 1013 struct nfsrv_descript *wp, *nfsd, *owp, *swp; 1014 struct nfs_fattr *fp; 1015 int i = 0; 1016 struct iovec *iov; 1017 struct nfsrvw_delayhash *wpp; 1018 struct ucred *cred; 1019 struct vattr va, forat; 1020 u_int32_t *tl; 1021 int32_t t1; 1022 caddr_t bpos, dpos; 1023 int error = 0, rdonly, cache, len = 0, forat_ret = 1; 1024 int ioflags, aftat_ret = 1, s, adjust, v3, zeroing; 1025 char *cp2; 1026 struct mbuf *mb, *mreq, *mrep, *md; 1027 struct vnode *vp; 1028 struct uio io, *uiop = &io; 1029 u_quad_t frev, cur_usec; 1030 struct mount *mntp = NULL; 1031 1032 *mrq = NULL; 1033 if (*ndp) { 1034 nfsd = *ndp; 1035 *ndp = NULL; 1036 mrep = nfsd->nd_mrep; 1037 md = nfsd->nd_md; 1038 dpos = nfsd->nd_dpos; 1039 cred = &nfsd->nd_cr; 1040 v3 = (nfsd->nd_flag & ND_NFSV3); 1041 LIST_INIT(&nfsd->nd_coalesce); 1042 nfsd->nd_mreq = NULL; 1043 nfsd->nd_stable = NFSV3WRITE_FILESYNC; 1044 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec; 1045 nfsd->nd_time = cur_usec + nfsrvw_procrastinate; 1046 1047 /* 1048 * Now, get the write header.. 1049 */ 1050 nfsm_srvmtofh(&nfsd->nd_fh); 1051 if (v3) { 1052 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 1053 nfsd->nd_off = fxdr_hyper(tl); 1054 tl += 3; 1055 nfsd->nd_stable = fxdr_unsigned(int, *tl++); 1056 } else { 1057 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1058 nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 1059 tl += 2; 1060 } 1061 len = fxdr_unsigned(int32_t, *tl); 1062 nfsd->nd_len = len; 1063 nfsd->nd_eoff = nfsd->nd_off + len; 1064 1065 /* 1066 * Trim the header out of the mbuf list and trim off any trailing 1067 * junk so that the mbuf list has only the write data. 1068 */ 1069 zeroing = 1; 1070 i = 0; 1071 mp = mrep; 1072 while (mp) { 1073 if (mp == md) { 1074 zeroing = 0; 1075 adjust = dpos - mtod(mp, caddr_t); 1076 mp->m_len -= adjust; 1077 if (mp->m_len > 0 && adjust > 0) 1078 NFSMADV(mp, adjust); 1079 } 1080 if (zeroing) 1081 mp->m_len = 0; 1082 else { 1083 i += mp->m_len; 1084 if (i > len) { 1085 mp->m_len -= (i - len); 1086 zeroing = 1; 1087 } 1088 } 1089 mp = mp->m_next; 1090 } 1091 if (len > NFS_MAXDATA || len < 0 || i < len) { 1092 nfsmout: 1093 m_freem(mrep); 1094 error = EIO; 1095 nfsm_writereply(2 * NFSX_UNSIGNED, v3); 1096 if (v3) 1097 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1098 nfsd->nd_mreq = mreq; 1099 nfsd->nd_mrep = NULL; 1100 nfsd->nd_time = 0; 1101 } 1102 1103 /* 1104 * Add this entry to the hash and time queues. 1105 */ 1106 s = splsoftclock(); 1107 owp = NULL; 1108 wp = LIST_FIRST(&slp->ns_tq); 1109 while (wp && wp->nd_time < nfsd->nd_time) { 1110 owp = wp; 1111 wp = LIST_NEXT(wp, nd_tq); 1112 } 1113 if (owp) { 1114 LIST_INSERT_AFTER(owp, nfsd, nd_tq); 1115 } else { 1116 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1117 } 1118 if (nfsd->nd_mrep) { 1119 wpp = NWDELAYHASH(slp, nfsd->nd_fh.fh_fid.fid_data); 1120 owp = NULL; 1121 wp = LIST_FIRST(wpp); 1122 while (wp && 1123 memcmp((caddr_t)&nfsd->nd_fh, (caddr_t)&wp->nd_fh, NFSX_V3FH)) { 1124 owp = wp; 1125 wp = LIST_NEXT(wp, nd_hash); 1126 } 1127 while (wp && wp->nd_off < nfsd->nd_off && 1128 !memcmp((caddr_t)&nfsd->nd_fh, (caddr_t)&wp->nd_fh, NFSX_V3FH)) { 1129 owp = wp; 1130 wp = LIST_NEXT(wp, nd_hash); 1131 } 1132 if (owp) { 1133 LIST_INSERT_AFTER(owp, nfsd, nd_hash); 1134 1135 /* 1136 * Search the hash list for overlapping entries and 1137 * coalesce. 1138 */ 1139 for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) { 1140 wp = LIST_NEXT(nfsd, nd_hash); 1141 if (NFSW_SAMECRED(owp, nfsd)) 1142 nfsrvw_coalesce(owp, nfsd); 1143 } 1144 } else { 1145 LIST_INSERT_HEAD(wpp, nfsd, nd_hash); 1146 } 1147 } 1148 splx(s); 1149 } 1150 1151 /* 1152 * Now, do VOP_WRITE()s for any one(s) that need to be done now 1153 * and generate the associated reply mbuf list(s). 1154 */ 1155 loop1: 1156 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec; 1157 s = splsoftclock(); 1158 for (nfsd = LIST_FIRST(&slp->ns_tq); nfsd; nfsd = owp) { 1159 owp = LIST_NEXT(nfsd, nd_tq); 1160 if (nfsd->nd_time > cur_usec) 1161 break; 1162 if (nfsd->nd_mreq) 1163 continue; 1164 LIST_REMOVE(nfsd, nd_tq); 1165 LIST_REMOVE(nfsd, nd_hash); 1166 splx(s); 1167 mrep = nfsd->nd_mrep; 1168 nfsd->nd_mrep = NULL; 1169 cred = &nfsd->nd_cr; 1170 v3 = (nfsd->nd_flag & ND_NFSV3); 1171 forat_ret = aftat_ret = 1; 1172 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp, 1173 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH), 1174 FALSE); 1175 if (!error) { 1176 if (v3) 1177 forat_ret = VOP_GETATTR(vp, &forat, cred, procp); 1178 if (vp->v_type != VREG) { 1179 if (v3) 1180 error = EINVAL; 1181 else 1182 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 1183 } 1184 } else 1185 vp = NULL; 1186 if (!error) { 1187 nqsrv_getl(vp, ND_WRITE); 1188 error = nfsrv_access(vp, VWRITE, cred, rdonly, procp, 1); 1189 } 1190 1191 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE) 1192 ioflags = IO_NODELOCKED; 1193 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC) 1194 ioflags = (IO_SYNC | IO_NODELOCKED); 1195 else 1196 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 1197 uiop->uio_rw = UIO_WRITE; 1198 uiop->uio_segflg = UIO_SYSSPACE; 1199 uiop->uio_procp = NULL; 1200 uiop->uio_offset = nfsd->nd_off; 1201 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off; 1202 if (uiop->uio_resid > 0) { 1203 mp = mrep; 1204 i = 0; 1205 while (mp) { 1206 if (mp->m_len > 0) 1207 i++; 1208 mp = mp->m_next; 1209 } 1210 uiop->uio_iovcnt = i; 1211 iov = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK); 1212 uiop->uio_iov = ivp = iov; 1213 mp = mrep; 1214 while (mp) { 1215 if (mp->m_len > 0) { 1216 ivp->iov_base = mtod(mp, caddr_t); 1217 ivp->iov_len = mp->m_len; 1218 ivp++; 1219 } 1220 mp = mp->m_next; 1221 } 1222 if (!error) { 1223 if (vn_start_write(vp, &mntp, V_NOWAIT) != 0) { 1224 VOP_UNLOCK(vp, 0); 1225 vn_start_write(NULL, &mntp, V_WAIT); 1226 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1227 } 1228 if (!error) { 1229 error = VOP_WRITE(vp, uiop, ioflags, cred); 1230 nfsstats.srvvop_writes++; 1231 vn_finished_write(mntp, 0); 1232 } 1233 } 1234 free((caddr_t)iov, M_TEMP); 1235 } 1236 m_freem(mrep); 1237 if (vp) { 1238 aftat_ret = VOP_GETATTR(vp, &va, cred, procp); 1239 vput(vp); 1240 } 1241 1242 /* 1243 * Loop around generating replies for all write rpcs that have 1244 * now been completed. 1245 */ 1246 swp = nfsd; 1247 do { 1248 if (error) { 1249 nfsm_writereply(NFSX_WCCDATA(v3), v3); 1250 if (v3) { 1251 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1252 } 1253 } else { 1254 nfsm_writereply(NFSX_PREOPATTR(v3) + 1255 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED + 1256 NFSX_WRITEVERF(v3), v3); 1257 if (v3) { 1258 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1259 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1260 *tl++ = txdr_unsigned(nfsd->nd_len); 1261 *tl++ = txdr_unsigned(swp->nd_stable); 1262 /* 1263 * Actually, there is no need to txdr these fields, 1264 * but it may make the values more human readable, 1265 * for debugging purposes. 1266 */ 1267 *tl++ = txdr_unsigned(boottime.tv_sec); 1268 *tl = txdr_unsigned(boottime.tv_usec); 1269 } else { 1270 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1271 nfsm_srvfillattr(&va, fp); 1272 } 1273 } 1274 nfsd->nd_mreq = mreq; 1275 if (nfsd->nd_mrep) 1276 panic("nfsrv_write: nd_mrep not free"); 1277 1278 /* 1279 * Done. Put it at the head of the timer queue so that 1280 * the final phase can return the reply. 1281 */ 1282 s = splsoftclock(); 1283 if (nfsd != swp) { 1284 nfsd->nd_time = 0; 1285 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1286 } 1287 nfsd = LIST_FIRST(&swp->nd_coalesce); 1288 if (nfsd) { 1289 LIST_REMOVE(nfsd, nd_tq); 1290 } 1291 splx(s); 1292 } while (nfsd); 1293 s = splsoftclock(); 1294 swp->nd_time = 0; 1295 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq); 1296 splx(s); 1297 goto loop1; 1298 } 1299 splx(s); 1300 1301 /* 1302 * Search for a reply to return. 1303 */ 1304 s = splsoftclock(); 1305 LIST_FOREACH(nfsd, &slp->ns_tq, nd_tq) { 1306 if (nfsd->nd_mreq) { 1307 LIST_REMOVE(nfsd, nd_tq); 1308 *mrq = nfsd->nd_mreq; 1309 *ndp = nfsd; 1310 break; 1311 } 1312 } 1313 splx(s); 1314 return (0); 1315 } 1316 1317 /* 1318 * Coalesce the write request nfsd into owp. To do this we must: 1319 * - remove nfsd from the queues 1320 * - merge nfsd->nd_mrep into owp->nd_mrep 1321 * - update the nd_eoff and nd_stable for owp 1322 * - put nfsd on owp's nd_coalesce list 1323 * NB: Must be called at splsoftclock(). 1324 */ 1325 void 1326 nfsrvw_coalesce(owp, nfsd) 1327 struct nfsrv_descript *owp; 1328 struct nfsrv_descript *nfsd; 1329 { 1330 int overlap; 1331 struct mbuf *mp; 1332 struct nfsrv_descript *m; 1333 1334 LIST_REMOVE(nfsd, nd_hash); 1335 LIST_REMOVE(nfsd, nd_tq); 1336 if (owp->nd_eoff < nfsd->nd_eoff) { 1337 overlap = owp->nd_eoff - nfsd->nd_off; 1338 if (overlap < 0) 1339 panic("nfsrv_coalesce: bad off"); 1340 if (overlap > 0) 1341 m_adj(nfsd->nd_mrep, overlap); 1342 mp = owp->nd_mrep; 1343 while (mp->m_next) 1344 mp = mp->m_next; 1345 mp->m_next = nfsd->nd_mrep; 1346 owp->nd_eoff = nfsd->nd_eoff; 1347 } else 1348 m_freem(nfsd->nd_mrep); 1349 nfsd->nd_mrep = NULL; 1350 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC) 1351 owp->nd_stable = NFSV3WRITE_FILESYNC; 1352 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC && 1353 owp->nd_stable == NFSV3WRITE_UNSTABLE) 1354 owp->nd_stable = NFSV3WRITE_DATASYNC; 1355 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq); 1356 /* 1357 * nfsd might hold coalesce elements! Move them to owp. 1358 * Otherwise, requests may be lost and clients will be stuck. 1359 */ 1360 while ((m = LIST_FIRST(&nfsd->nd_coalesce)) != NULL) { 1361 LIST_REMOVE(m, nd_tq); 1362 LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq); 1363 } 1364 } 1365 1366 /* 1367 * nfs create service 1368 * now does a truncate to 0 length via. setattr if it already exists 1369 */ 1370 int 1371 nfsrv_create(nfsd, slp, procp, mrq) 1372 struct nfsrv_descript *nfsd; 1373 struct nfssvc_sock *slp; 1374 struct proc *procp; 1375 struct mbuf **mrq; 1376 { 1377 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1378 struct mbuf *nam = nfsd->nd_nam; 1379 caddr_t dpos = nfsd->nd_dpos; 1380 struct ucred *cred = &nfsd->nd_cr; 1381 struct nfs_fattr *fp; 1382 struct vattr va, dirfor, diraft; 1383 struct nfsv2_sattr *sp; 1384 u_int32_t *tl; 1385 struct nameidata nd; 1386 caddr_t cp; 1387 int32_t t1; 1388 caddr_t bpos; 1389 int error = 0, cache, len, tsize, dirfor_ret = 1, diraft_ret = 1; 1390 int rdev = 0; 1391 int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0; 1392 char *cp2; 1393 struct mbuf *mb, *mreq; 1394 struct vnode *vp = NULL, *dirp = NULL; 1395 nfsfh_t nfh; 1396 fhandle_t *fhp; 1397 u_quad_t frev, tempsize; 1398 u_char cverf[NFSX_V3CREATEVERF]; 1399 struct mount *mp = NULL; 1400 1401 nd.ni_cnd.cn_nameiop = 0; 1402 fhp = &nfh.fh_generic; 1403 nfsm_srvmtofh(fhp); 1404 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1405 return (ESTALE); 1406 vn_start_write(NULL, &mp, V_WAIT); 1407 nfsm_srvnamesiz(len); 1408 nd.ni_cnd.cn_cred = cred; 1409 nd.ni_cnd.cn_nameiop = CREATE; 1410 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1411 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1412 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1413 if (dirp) { 1414 if (v3) 1415 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, 1416 procp); 1417 else { 1418 vrele(dirp); 1419 dirp = (struct vnode *)0; 1420 } 1421 } 1422 if (error) { 1423 nfsm_reply(NFSX_WCCDATA(v3)); 1424 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1425 if (dirp) 1426 vrele(dirp); 1427 vn_finished_write(mp, 0); 1428 return (0); 1429 } 1430 VATTR_NULL(&va); 1431 if (v3) { 1432 va.va_mode = 0; 1433 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1434 how = fxdr_unsigned(int, *tl); 1435 switch (how) { 1436 case NFSV3CREATE_GUARDED: 1437 if (nd.ni_vp) { 1438 error = EEXIST; 1439 break; 1440 } 1441 case NFSV3CREATE_UNCHECKED: 1442 nfsm_srvsattr(&va); 1443 break; 1444 case NFSV3CREATE_EXCLUSIVE: 1445 nfsm_dissect(cp, caddr_t, NFSX_V3CREATEVERF); 1446 memcpy(cverf, cp, NFSX_V3CREATEVERF); 1447 exclusive_flag = 1; 1448 break; 1449 }; 1450 va.va_type = VREG; 1451 } else { 1452 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 1453 va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode)); 1454 if (va.va_type == VNON) 1455 va.va_type = VREG; 1456 va.va_mode = nfstov_mode(sp->sa_mode); 1457 switch (va.va_type) { 1458 case VREG: 1459 tsize = fxdr_unsigned(int32_t, sp->sa_size); 1460 if (tsize != -1) 1461 va.va_size = (u_quad_t)tsize; 1462 break; 1463 case VCHR: 1464 case VBLK: 1465 case VFIFO: 1466 rdev = fxdr_unsigned(int32_t, sp->sa_size); 1467 break; 1468 default: 1469 break; 1470 }; 1471 } 1472 1473 /* 1474 * Iff doesn't exist, create it 1475 * otherwise just truncate to 0 length 1476 * should I set the mode too ?? 1477 */ 1478 if (nd.ni_vp == NULL) { 1479 if (va.va_type == VREG || va.va_type == VSOCK) { 1480 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1481 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1482 if (!error) { 1483 if (exclusive_flag) { 1484 exclusive_flag = 0; 1485 VATTR_NULL(&va); 1486 /* 1487 * XXX 1488 * assuming NFSX_V3CREATEVERF 1489 * == sizeof(nfstime3) 1490 */ 1491 fxdr_nfsv3time(cverf, &va.va_atime); 1492 error = VOP_SETATTR(nd.ni_vp, &va, cred, 1493 procp); 1494 } 1495 } 1496 } else if (va.va_type == VCHR || va.va_type == VBLK || 1497 va.va_type == VFIFO) { 1498 if (va.va_type == VCHR && rdev == 0xffffffff) 1499 va.va_type = VFIFO; 1500 if (va.va_type != VFIFO && 1501 (error = suser(cred, (u_short *)0))) { 1502 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1503 vput(nd.ni_dvp); 1504 nfsm_reply(0); 1505 vn_finished_write(mp, 0); 1506 return (error); 1507 } else 1508 va.va_rdev = (dev_t)rdev; 1509 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1510 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, 1511 &va); 1512 if (error) { 1513 nfsm_reply(0); 1514 } 1515 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1516 vrele(nd.ni_dvp); 1517 vput(nd.ni_vp); 1518 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1519 error = EINVAL; 1520 nfsm_reply(0); 1521 } 1522 } else { 1523 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1524 vput(nd.ni_dvp); 1525 error = ENXIO; 1526 } 1527 vp = nd.ni_vp; 1528 } else { 1529 vp = nd.ni_vp; 1530 if (nd.ni_dvp == vp) 1531 vrele(nd.ni_dvp); 1532 else 1533 vput(nd.ni_dvp); 1534 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1535 if (!error && va.va_size != -1) { 1536 error = nfsrv_access(vp, VWRITE, cred, 1537 (nd.ni_cnd.cn_flags & RDONLY), procp, 0); 1538 if (!error) { 1539 nqsrv_getl(vp, ND_WRITE); 1540 tempsize = va.va_size; 1541 VATTR_NULL(&va); 1542 va.va_size = tempsize; 1543 error = VOP_SETATTR(vp, &va, cred, 1544 procp); 1545 } 1546 } 1547 if (error) 1548 vput(vp); 1549 } 1550 if (!error) { 1551 memset((caddr_t)fhp, 0, sizeof(nfh)); 1552 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1553 error = VFS_VPTOFH(vp, &fhp->fh_fid); 1554 if (!error) 1555 error = VOP_GETATTR(vp, &va, cred, procp); 1556 vput(vp); 1557 } 1558 if (v3) { 1559 if (exclusive_flag && !error) { 1560 /* 1561 * XXX assuming NFSX_V3CREATEVERF == sizeof(nfstime3) 1562 */ 1563 char oldverf[NFSX_V3CREATEVERF]; 1564 1565 txdr_nfsv3time(&va.va_atime, oldverf); 1566 if (memcmp(cverf, oldverf, NFSX_V3CREATEVERF)) 1567 error = EEXIST; 1568 } 1569 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); 1570 vrele(dirp); 1571 } 1572 nfsm_reply(NFSX_SRVFH(v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3)); 1573 if (v3) { 1574 if (!error) { 1575 nfsm_srvpostop_fh(fhp); 1576 nfsm_srvpostop_attr(0, &va); 1577 } 1578 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1579 } else { 1580 nfsm_srvfhtom(fhp, v3); 1581 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1582 nfsm_srvfillattr(&va, fp); 1583 } 1584 vn_finished_write(mp, 0); 1585 return (0); 1586 nfsmout: 1587 if (dirp) 1588 vrele(dirp); 1589 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1590 if (nd.ni_dvp == nd.ni_vp) 1591 vrele(nd.ni_dvp); 1592 else 1593 vput(nd.ni_dvp); 1594 if (nd.ni_vp) 1595 vput(nd.ni_vp); 1596 vn_finished_write(mp, 0); 1597 return (error); 1598 } 1599 1600 /* 1601 * nfs v3 mknod service 1602 */ 1603 int 1604 nfsrv_mknod(nfsd, slp, procp, mrq) 1605 struct nfsrv_descript *nfsd; 1606 struct nfssvc_sock *slp; 1607 struct proc *procp; 1608 struct mbuf **mrq; 1609 { 1610 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1611 struct mbuf *nam = nfsd->nd_nam; 1612 caddr_t dpos = nfsd->nd_dpos; 1613 struct ucred *cred = &nfsd->nd_cr; 1614 struct vattr va, dirfor, diraft; 1615 u_int32_t *tl; 1616 struct nameidata nd; 1617 int32_t t1; 1618 caddr_t bpos; 1619 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 1620 u_int32_t major, minor; 1621 enum vtype vtyp; 1622 char *cp2; 1623 struct mbuf *mb, *mreq; 1624 struct vnode *vp, *dirp = (struct vnode *)0; 1625 nfsfh_t nfh; 1626 fhandle_t *fhp; 1627 u_quad_t frev; 1628 struct mount *mp = NULL; 1629 1630 nd.ni_cnd.cn_nameiop = 0; 1631 fhp = &nfh.fh_generic; 1632 nfsm_srvmtofh(fhp); 1633 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1634 return (ESTALE); 1635 vn_start_write(NULL, &mp, V_WAIT); 1636 nfsm_srvnamesiz(len); 1637 nd.ni_cnd.cn_cred = cred; 1638 nd.ni_cnd.cn_nameiop = CREATE; 1639 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1640 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1641 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1642 if (dirp) 1643 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, procp); 1644 if (error) { 1645 nfsm_reply(NFSX_WCCDATA(1)); 1646 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1647 if (dirp) 1648 vrele(dirp); 1649 vn_finished_write(mp, 0); 1650 return (0); 1651 } 1652 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1653 vtyp = nfsv3tov_type(*tl); 1654 if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) { 1655 error = NFSERR_BADTYPE; 1656 goto abort; 1657 } 1658 VATTR_NULL(&va); 1659 va.va_mode = 0; 1660 nfsm_srvsattr(&va); 1661 if (vtyp == VCHR || vtyp == VBLK) { 1662 dev_t rdev; 1663 1664 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1665 major = fxdr_unsigned(u_int32_t, *tl++); 1666 minor = fxdr_unsigned(u_int32_t, *tl); 1667 rdev = makedev(major, minor); 1668 if (major(rdev) != major || minor(rdev) != minor) { 1669 error = EINVAL; 1670 goto abort; 1671 } 1672 va.va_rdev = rdev; 1673 } 1674 1675 /* 1676 * Iff doesn't exist, create it. 1677 */ 1678 if (nd.ni_vp) { 1679 error = EEXIST; 1680 abort: 1681 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1682 if (nd.ni_dvp == nd.ni_vp) 1683 vrele(nd.ni_dvp); 1684 else 1685 vput(nd.ni_dvp); 1686 if (nd.ni_vp) 1687 vput(nd.ni_vp); 1688 goto out; 1689 } 1690 va.va_type = vtyp; 1691 if (vtyp == VSOCK) { 1692 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1693 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1694 } else { 1695 if (va.va_type != VFIFO && 1696 (error = suser(cred, (u_short *)0))) { 1697 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1698 vput(nd.ni_dvp); 1699 goto out; 1700 } 1701 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1702 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1703 if (error) { 1704 goto out; 1705 } 1706 if (error) 1707 goto out; 1708 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1709 vput(nd.ni_vp); 1710 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1711 error = EINVAL; 1712 } 1713 } 1714 out: 1715 vp = nd.ni_vp; 1716 if (!error) { 1717 memset((caddr_t)fhp, 0, sizeof(nfh)); 1718 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1719 error = VFS_VPTOFH(vp, &fhp->fh_fid); 1720 if (!error) 1721 error = VOP_GETATTR(vp, &va, cred, procp); 1722 vput(vp); 1723 } 1724 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); 1725 vrele(dirp); 1726 nfsm_reply(NFSX_SRVFH(1) + NFSX_POSTOPATTR(1) + NFSX_WCCDATA(1)); 1727 if (!error) { 1728 nfsm_srvpostop_fh(fhp); 1729 nfsm_srvpostop_attr(0, &va); 1730 } 1731 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1732 vn_finished_write(mp, 0); 1733 return (0); 1734 nfsmout: 1735 if (dirp) 1736 vrele(dirp); 1737 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1738 if (nd.ni_dvp == nd.ni_vp) 1739 vrele(nd.ni_dvp); 1740 else 1741 vput(nd.ni_dvp); 1742 if (nd.ni_vp) 1743 vput(nd.ni_vp); 1744 vn_finished_write(mp, 0); 1745 return (error); 1746 } 1747 1748 /* 1749 * nfs remove service 1750 */ 1751 int 1752 nfsrv_remove(nfsd, slp, procp, mrq) 1753 struct nfsrv_descript *nfsd; 1754 struct nfssvc_sock *slp; 1755 struct proc *procp; 1756 struct mbuf **mrq; 1757 { 1758 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1759 struct mbuf *nam = nfsd->nd_nam; 1760 caddr_t dpos = nfsd->nd_dpos; 1761 struct ucred *cred = &nfsd->nd_cr; 1762 struct nameidata nd; 1763 u_int32_t *tl; 1764 int32_t t1; 1765 caddr_t bpos; 1766 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 1767 int v3 = (nfsd->nd_flag & ND_NFSV3); 1768 char *cp2; 1769 struct mbuf *mb, *mreq; 1770 struct vnode *vp, *dirp; 1771 struct vattr dirfor, diraft; 1772 nfsfh_t nfh; 1773 fhandle_t *fhp; 1774 u_quad_t frev; 1775 struct mount *mp = NULL; 1776 1777 #ifndef nolint 1778 vp = (struct vnode *)0; 1779 #endif 1780 fhp = &nfh.fh_generic; 1781 nfsm_srvmtofh(fhp); 1782 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1783 return (ESTALE); 1784 vn_start_write(NULL, &mp, V_WAIT); 1785 nfsm_srvnamesiz(len); 1786 nd.ni_cnd.cn_cred = cred; 1787 nd.ni_cnd.cn_nameiop = DELETE; 1788 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1789 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1790 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1791 if (dirp) { 1792 if (v3) 1793 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, 1794 procp); 1795 else 1796 vrele(dirp); 1797 } 1798 if (!error) { 1799 vp = nd.ni_vp; 1800 if (vp->v_type == VDIR && 1801 (error = suser(cred, (u_short *)0)) != 0) 1802 goto out; 1803 /* 1804 * The root of a mounted filesystem cannot be deleted. 1805 */ 1806 if (vp->v_flag & VROOT) { 1807 error = EBUSY; 1808 goto out; 1809 } 1810 out: 1811 if (!error) { 1812 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1813 nqsrv_getl(vp, ND_WRITE); 1814 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1815 } else { 1816 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1817 if (nd.ni_dvp == vp) 1818 vrele(nd.ni_dvp); 1819 else 1820 vput(nd.ni_dvp); 1821 vput(vp); 1822 } 1823 } 1824 if (dirp && v3) { 1825 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); 1826 vrele(dirp); 1827 } 1828 nfsm_reply(NFSX_WCCDATA(v3)); 1829 if (v3) { 1830 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1831 vn_finished_write(mp, 0); 1832 return (0); 1833 } 1834 vn_finished_write(mp, 0); 1835 nfsm_srvdone; 1836 } 1837 1838 /* 1839 * nfs rename service 1840 */ 1841 int 1842 nfsrv_rename(nfsd, slp, procp, mrq) 1843 struct nfsrv_descript *nfsd; 1844 struct nfssvc_sock *slp; 1845 struct proc *procp; 1846 struct mbuf **mrq; 1847 { 1848 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1849 struct mbuf *nam = nfsd->nd_nam; 1850 caddr_t dpos = nfsd->nd_dpos; 1851 struct ucred *cred = &nfsd->nd_cr; 1852 u_int32_t *tl; 1853 int32_t t1; 1854 caddr_t bpos; 1855 int error = 0, cache, fdirfor_ret = 1, fdiraft_ret = 1; 1856 uint32_t len, len2; 1857 int tdirfor_ret = 1, tdiraft_ret = 1; 1858 int v3 = (nfsd->nd_flag & ND_NFSV3); 1859 char *cp2; 1860 struct mbuf *mb, *mreq; 1861 struct nameidata fromnd, tond; 1862 struct vnode *fvp, *tvp, *tdvp, *fdirp = (struct vnode *)0; 1863 struct vnode *tdirp = (struct vnode *)0; 1864 struct vattr fdirfor, fdiraft, tdirfor, tdiraft; 1865 nfsfh_t fnfh, tnfh; 1866 fhandle_t *ffhp, *tfhp; 1867 u_quad_t frev; 1868 uid_t saved_uid; 1869 struct mount *mp = NULL; 1870 1871 #ifndef nolint 1872 fvp = (struct vnode *)0; 1873 #endif 1874 ffhp = &fnfh.fh_generic; 1875 tfhp = &tnfh.fh_generic; 1876 fromnd.ni_cnd.cn_nameiop = 0; 1877 tond.ni_cnd.cn_nameiop = 0; 1878 nfsm_srvmtofh(ffhp); 1879 if ((mp = vfs_getvfs(&ffhp->fh_fsid)) == NULL) 1880 return (ESTALE); 1881 vn_start_write(NULL, &mp, V_WAIT); 1882 nfsm_srvnamesiz(len); 1883 /* 1884 * Remember our original uid so that we can reset cr_uid before 1885 * the second nfs_namei() call, in case it is remapped. 1886 */ 1887 saved_uid = cred->cr_uid; 1888 fromnd.ni_cnd.cn_cred = cred; 1889 fromnd.ni_cnd.cn_nameiop = DELETE; 1890 fromnd.ni_cnd.cn_flags = WANTPARENT | SAVESTART; 1891 error = nfs_namei(&fromnd, ffhp, len, slp, nam, &md, 1892 &dpos, &fdirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1893 if (fdirp) { 1894 if (v3) 1895 fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred, 1896 procp); 1897 else { 1898 vrele(fdirp); 1899 fdirp = (struct vnode *)0; 1900 } 1901 } 1902 if (error) { 1903 nfsm_reply(2 * NFSX_WCCDATA(v3)); 1904 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 1905 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 1906 if (fdirp) 1907 vrele(fdirp); 1908 vn_finished_write(mp, 0); 1909 return (0); 1910 } 1911 fvp = fromnd.ni_vp; 1912 nfsm_srvmtofh(tfhp); 1913 if (v3) { 1914 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 1915 len2 = fxdr_unsigned(uint32_t, *tl); 1916 /* len2 will be checked by nfs_namei */ 1917 } 1918 else { 1919 /* NFSv2 */ 1920 nfsm_strsiz(len2, NFS_MAXNAMLEN); 1921 } 1922 cred->cr_uid = saved_uid; 1923 tond.ni_cnd.cn_cred = cred; 1924 tond.ni_cnd.cn_nameiop = RENAME; 1925 tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART; 1926 error = nfs_namei(&tond, tfhp, len2, slp, nam, &md, 1927 &dpos, &tdirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1928 if (tdirp) { 1929 if (v3) 1930 tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred, 1931 procp); 1932 else { 1933 vrele(tdirp); 1934 tdirp = (struct vnode *)0; 1935 } 1936 } 1937 if (error) { 1938 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1939 vrele(fromnd.ni_dvp); 1940 vrele(fvp); 1941 goto out1; 1942 } 1943 tdvp = tond.ni_dvp; 1944 tvp = tond.ni_vp; 1945 if (tvp != NULL) { 1946 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 1947 if (v3) 1948 error = EEXIST; 1949 else 1950 error = EISDIR; 1951 goto out; 1952 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 1953 if (v3) 1954 error = EEXIST; 1955 else 1956 error = ENOTDIR; 1957 goto out; 1958 } 1959 if (tvp->v_type == VDIR && tvp->v_mountedhere) { 1960 if (v3) 1961 error = EXDEV; 1962 else 1963 error = ENOTEMPTY; 1964 goto out; 1965 } 1966 } 1967 if (fvp->v_type == VDIR && fvp->v_mountedhere) { 1968 if (v3) 1969 error = EXDEV; 1970 else 1971 error = ENOTEMPTY; 1972 goto out; 1973 } 1974 if (fvp->v_mount != tdvp->v_mount) { 1975 if (v3) 1976 error = EXDEV; 1977 else 1978 error = ENOTEMPTY; 1979 goto out; 1980 } 1981 if (fvp == tdvp) { 1982 if (v3) 1983 error = EINVAL; 1984 else 1985 error = ENOTEMPTY; 1986 } 1987 /* 1988 * If source is the same as the destination (that is the 1989 * same vnode with the same name in the same directory), 1990 * then there is nothing to do. 1991 */ 1992 if (fvp == tvp && fromnd.ni_dvp == tdvp && 1993 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 1994 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr, 1995 fromnd.ni_cnd.cn_namelen)) 1996 error = -1; 1997 out: 1998 if (!error) { 1999 nqsrv_getl(fromnd.ni_dvp, ND_WRITE); 2000 nqsrv_getl(tdvp, ND_WRITE); 2001 if (tvp) { 2002 nqsrv_getl(tvp, ND_WRITE); 2003 } 2004 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 2005 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 2006 } else { 2007 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 2008 if (tdvp == tvp) 2009 vrele(tdvp); 2010 else 2011 vput(tdvp); 2012 if (tvp) 2013 vput(tvp); 2014 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2015 vrele(fromnd.ni_dvp); 2016 vrele(fvp); 2017 if (error == -1) 2018 error = 0; 2019 } 2020 vrele(tond.ni_startdir); 2021 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 2022 out1: 2023 if (fdirp) { 2024 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred, procp); 2025 vrele(fdirp); 2026 } 2027 if (tdirp) { 2028 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred, procp); 2029 vrele(tdirp); 2030 } 2031 vrele(fromnd.ni_startdir); 2032 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2033 nfsm_reply(2 * NFSX_WCCDATA(v3)); 2034 if (v3) { 2035 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 2036 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 2037 } 2038 vn_finished_write(mp, 0); 2039 return (0); 2040 2041 nfsmout: 2042 if (fdirp) 2043 vrele(fdirp); 2044 if (tdirp) 2045 vrele(tdirp); 2046 if (tond.ni_cnd.cn_nameiop) { 2047 vrele(tond.ni_startdir); 2048 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 2049 } 2050 if (fromnd.ni_cnd.cn_nameiop) { 2051 vrele(fromnd.ni_startdir); 2052 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2053 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2054 vrele(fromnd.ni_dvp); 2055 vrele(fvp); 2056 } 2057 return (error); 2058 } 2059 2060 /* 2061 * nfs link service 2062 */ 2063 int 2064 nfsrv_link(nfsd, slp, procp, mrq) 2065 struct nfsrv_descript *nfsd; 2066 struct nfssvc_sock *slp; 2067 struct proc *procp; 2068 struct mbuf **mrq; 2069 { 2070 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2071 struct mbuf *nam = nfsd->nd_nam; 2072 caddr_t dpos = nfsd->nd_dpos; 2073 struct ucred *cred = &nfsd->nd_cr; 2074 struct nameidata nd; 2075 u_int32_t *tl; 2076 int32_t t1; 2077 caddr_t bpos; 2078 int error = 0, rdonly, cache, len, dirfor_ret = 1, diraft_ret = 1; 2079 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3); 2080 char *cp2; 2081 struct mbuf *mb, *mreq; 2082 struct vnode *vp, *xp, *dirp = (struct vnode *)0; 2083 struct vattr dirfor, diraft, at; 2084 nfsfh_t nfh, dnfh; 2085 fhandle_t *fhp, *dfhp; 2086 u_quad_t frev; 2087 struct mount *mp = NULL; 2088 2089 fhp = &nfh.fh_generic; 2090 dfhp = &dnfh.fh_generic; 2091 nfsm_srvmtofh(fhp); 2092 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2093 return (ESTALE); 2094 vn_start_write(NULL, &mp, V_WAIT); 2095 nfsm_srvmtofh(dfhp); 2096 nfsm_srvnamesiz(len); 2097 error = nfsrv_fhtovp(fhp, FALSE, &vp, cred, slp, nam, 2098 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2099 if (error) { 2100 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2101 nfsm_srvpostop_attr(getret, &at); 2102 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2103 vn_finished_write(mp, 0); 2104 return (0); 2105 } 2106 if (vp->v_type == VDIR && (error = suser(cred, (u_short *)0)) != 0) 2107 goto out1; 2108 nd.ni_cnd.cn_cred = cred; 2109 nd.ni_cnd.cn_nameiop = CREATE; 2110 nd.ni_cnd.cn_flags = LOCKPARENT; 2111 error = nfs_namei(&nd, dfhp, len, slp, nam, &md, &dpos, 2112 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2113 if (dirp) { 2114 if (v3) 2115 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, 2116 procp); 2117 else { 2118 vrele(dirp); 2119 dirp = (struct vnode *)0; 2120 } 2121 } 2122 if (error) 2123 goto out1; 2124 xp = nd.ni_vp; 2125 if (xp != NULL) { 2126 error = EEXIST; 2127 goto out; 2128 } 2129 xp = nd.ni_dvp; 2130 if (vp->v_mount != xp->v_mount) 2131 error = EXDEV; 2132 out: 2133 if (!error) { 2134 nqsrv_getl(vp, ND_WRITE); 2135 nqsrv_getl(xp, ND_WRITE); 2136 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2137 } else { 2138 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2139 if (nd.ni_dvp == nd.ni_vp) 2140 vrele(nd.ni_dvp); 2141 else 2142 vput(nd.ni_dvp); 2143 if (nd.ni_vp) 2144 vrele(nd.ni_vp); 2145 } 2146 out1: 2147 if (v3) 2148 getret = VOP_GETATTR(vp, &at, cred, procp); 2149 if (dirp) { 2150 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); 2151 vrele(dirp); 2152 } 2153 vrele(vp); 2154 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2155 if (v3) { 2156 nfsm_srvpostop_attr(getret, &at); 2157 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2158 vn_finished_write(mp, 0); 2159 return (0); 2160 } 2161 vn_finished_write(mp, 0); 2162 nfsm_srvdone; 2163 } 2164 2165 /* 2166 * nfs symbolic link service 2167 */ 2168 int 2169 nfsrv_symlink(nfsd, slp, procp, mrq) 2170 struct nfsrv_descript *nfsd; 2171 struct nfssvc_sock *slp; 2172 struct proc *procp; 2173 struct mbuf **mrq; 2174 { 2175 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2176 struct mbuf *nam = nfsd->nd_nam; 2177 caddr_t dpos = nfsd->nd_dpos; 2178 struct ucred *cred = &nfsd->nd_cr; 2179 struct vattr va, dirfor, diraft; 2180 struct nameidata nd; 2181 u_int32_t *tl; 2182 int32_t t1; 2183 struct nfsv2_sattr *sp; 2184 char *bpos, *pathcp = NULL, *cp2; 2185 struct uio io; 2186 struct iovec iv; 2187 int error = 0, cache, dirfor_ret = 1, diraft_ret = 1; 2188 uint32_t len, len2; 2189 int v3 = (nfsd->nd_flag & ND_NFSV3); 2190 struct mbuf *mb, *mreq; 2191 struct vnode *dirp = (struct vnode *)0; 2192 nfsfh_t nfh; 2193 fhandle_t *fhp; 2194 u_quad_t frev; 2195 struct mount *mp = NULL; 2196 2197 nd.ni_cnd.cn_nameiop = 0; 2198 fhp = &nfh.fh_generic; 2199 nfsm_srvmtofh(fhp); 2200 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2201 return (ESTALE); 2202 vn_start_write(NULL, &mp, V_WAIT); 2203 nfsm_srvnamesiz(len); 2204 nd.ni_cnd.cn_cred = cred; 2205 nd.ni_cnd.cn_nameiop = CREATE; 2206 nd.ni_cnd.cn_flags = LOCKPARENT; 2207 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2208 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2209 if (dirp) { 2210 if (v3) 2211 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, 2212 procp); 2213 else { 2214 vrele(dirp); 2215 dirp = (struct vnode *)0; 2216 } 2217 } 2218 if (error) 2219 goto out; 2220 VATTR_NULL(&va); 2221 if (v3) { 2222 va.va_mode = 0; 2223 nfsm_srvsattr(&va); 2224 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 2225 len2 = fxdr_unsigned(uint32_t, *tl); 2226 if (len2 > PATH_MAX) { 2227 /* XXX should check _PC_NO_TRUNC */ 2228 error = ENAMETOOLONG; 2229 goto abortop; 2230 } 2231 } 2232 else { 2233 /* NFSv2 */ 2234 nfsm_strsiz(len2, NFS_MAXPATHLEN); 2235 } 2236 pathcp = malloc(len2 + 1, M_TEMP, M_WAITOK); 2237 iv.iov_base = pathcp; 2238 iv.iov_len = len2; 2239 io.uio_resid = len2; 2240 io.uio_offset = 0; 2241 io.uio_iov = &iv; 2242 io.uio_iovcnt = 1; 2243 io.uio_segflg = UIO_SYSSPACE; 2244 io.uio_rw = UIO_READ; 2245 io.uio_procp = NULL; 2246 nfsm_mtouio(&io, len2); 2247 if (!v3) { 2248 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 2249 va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode); 2250 } 2251 *(pathcp + len2) = '\0'; 2252 if (nd.ni_vp) { 2253 error = EEXIST; 2254 abortop: 2255 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2256 if (nd.ni_dvp == nd.ni_vp) 2257 vrele(nd.ni_dvp); 2258 else 2259 vput(nd.ni_dvp); 2260 if (nd.ni_vp) 2261 vrele(nd.ni_vp); 2262 goto out; 2263 } 2264 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2265 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp); 2266 if (!error) { 2267 if (v3) { 2268 memset((caddr_t)fhp, 0, sizeof(nfh)); 2269 fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsidx; 2270 error = VFS_VPTOFH(nd.ni_vp, &fhp->fh_fid); 2271 if (!error) 2272 error = VOP_GETATTR(nd.ni_vp, &va, cred, 2273 procp); 2274 vput(nd.ni_vp); 2275 } else { 2276 vput(nd.ni_vp); 2277 } 2278 } 2279 out: 2280 if (pathcp) 2281 free(pathcp, M_TEMP); 2282 if (dirp) { 2283 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); 2284 vrele(dirp); 2285 } 2286 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2287 if (v3) { 2288 if (!error) { 2289 nfsm_srvpostop_fh(fhp); 2290 nfsm_srvpostop_attr(0, &va); 2291 } 2292 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2293 } 2294 vn_finished_write(mp, 0); 2295 return (0); 2296 nfsmout: 2297 if (dirp) 2298 vrele(dirp); 2299 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2300 if (nd.ni_dvp == nd.ni_vp) 2301 vrele(nd.ni_dvp); 2302 else 2303 vput(nd.ni_dvp); 2304 if (nd.ni_vp) 2305 vrele(nd.ni_vp); 2306 if (pathcp) 2307 free(pathcp, M_TEMP); 2308 vn_finished_write(mp, 0); 2309 return (error); 2310 } 2311 2312 /* 2313 * nfs mkdir service 2314 */ 2315 int 2316 nfsrv_mkdir(nfsd, slp, procp, mrq) 2317 struct nfsrv_descript *nfsd; 2318 struct nfssvc_sock *slp; 2319 struct proc *procp; 2320 struct mbuf **mrq; 2321 { 2322 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2323 struct mbuf *nam = nfsd->nd_nam; 2324 caddr_t dpos = nfsd->nd_dpos; 2325 struct ucred *cred = &nfsd->nd_cr; 2326 struct vattr va, dirfor, diraft; 2327 struct nfs_fattr *fp; 2328 struct nameidata nd; 2329 caddr_t cp; 2330 u_int32_t *tl; 2331 int32_t t1; 2332 caddr_t bpos; 2333 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 2334 int v3 = (nfsd->nd_flag & ND_NFSV3); 2335 char *cp2; 2336 struct mbuf *mb, *mreq; 2337 struct vnode *vp, *dirp = (struct vnode *)0; 2338 nfsfh_t nfh; 2339 fhandle_t *fhp; 2340 u_quad_t frev; 2341 struct mount *mp = NULL; 2342 2343 fhp = &nfh.fh_generic; 2344 nfsm_srvmtofh(fhp); 2345 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2346 return (ESTALE); 2347 vn_start_write(NULL, &mp, V_WAIT); 2348 nfsm_srvnamesiz(len); 2349 nd.ni_cnd.cn_cred = cred; 2350 nd.ni_cnd.cn_nameiop = CREATE; 2351 nd.ni_cnd.cn_flags = LOCKPARENT; 2352 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2353 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2354 if (dirp) { 2355 if (v3) 2356 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, 2357 procp); 2358 else { 2359 vrele(dirp); 2360 dirp = (struct vnode *)0; 2361 } 2362 } 2363 if (error) { 2364 nfsm_reply(NFSX_WCCDATA(v3)); 2365 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2366 if (dirp) 2367 vrele(dirp); 2368 vn_finished_write(mp, 0); 2369 return (0); 2370 } 2371 VATTR_NULL(&va); 2372 if (v3) { 2373 va.va_mode = 0; 2374 nfsm_srvsattr(&va); 2375 } else { 2376 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 2377 va.va_mode = nfstov_mode(*tl++); 2378 } 2379 va.va_type = VDIR; 2380 vp = nd.ni_vp; 2381 if (vp != NULL) { 2382 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2383 if (nd.ni_dvp == vp) 2384 vrele(nd.ni_dvp); 2385 else 2386 vput(nd.ni_dvp); 2387 vrele(vp); 2388 error = EEXIST; 2389 goto out; 2390 } 2391 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2392 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 2393 if (!error) { 2394 vp = nd.ni_vp; 2395 memset((caddr_t)fhp, 0, sizeof(nfh)); 2396 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 2397 error = VFS_VPTOFH(vp, &fhp->fh_fid); 2398 if (!error) 2399 error = VOP_GETATTR(vp, &va, cred, procp); 2400 vput(vp); 2401 } 2402 out: 2403 if (dirp) { 2404 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); 2405 vrele(dirp); 2406 } 2407 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2408 if (v3) { 2409 if (!error) { 2410 nfsm_srvpostop_fh(fhp); 2411 nfsm_srvpostop_attr(0, &va); 2412 } 2413 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2414 } else { 2415 nfsm_srvfhtom(fhp, v3); 2416 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 2417 nfsm_srvfillattr(&va, fp); 2418 } 2419 vn_finished_write(mp, 0); 2420 return (0); 2421 nfsmout: 2422 if (dirp) 2423 vrele(dirp); 2424 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2425 if (nd.ni_dvp == nd.ni_vp) 2426 vrele(nd.ni_dvp); 2427 else 2428 vput(nd.ni_dvp); 2429 if (nd.ni_vp) 2430 vrele(nd.ni_vp); 2431 vn_finished_write(mp, 0); 2432 return (error); 2433 } 2434 2435 /* 2436 * nfs rmdir service 2437 */ 2438 int 2439 nfsrv_rmdir(nfsd, slp, procp, mrq) 2440 struct nfsrv_descript *nfsd; 2441 struct nfssvc_sock *slp; 2442 struct proc *procp; 2443 struct mbuf **mrq; 2444 { 2445 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2446 struct mbuf *nam = nfsd->nd_nam; 2447 caddr_t dpos = nfsd->nd_dpos; 2448 struct ucred *cred = &nfsd->nd_cr; 2449 u_int32_t *tl; 2450 int32_t t1; 2451 caddr_t bpos; 2452 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 2453 int v3 = (nfsd->nd_flag & ND_NFSV3); 2454 char *cp2; 2455 struct mbuf *mb, *mreq; 2456 struct vnode *vp, *dirp = (struct vnode *)0; 2457 struct vattr dirfor, diraft; 2458 nfsfh_t nfh; 2459 fhandle_t *fhp; 2460 struct nameidata nd; 2461 u_quad_t frev; 2462 struct mount *mp = NULL; 2463 2464 fhp = &nfh.fh_generic; 2465 nfsm_srvmtofh(fhp); 2466 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2467 return (ESTALE); 2468 vn_start_write(NULL, &mp, V_WAIT); 2469 nfsm_srvnamesiz(len); 2470 nd.ni_cnd.cn_cred = cred; 2471 nd.ni_cnd.cn_nameiop = DELETE; 2472 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 2473 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2474 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2475 if (dirp) { 2476 if (v3) 2477 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, 2478 procp); 2479 else { 2480 vrele(dirp); 2481 dirp = (struct vnode *)0; 2482 } 2483 } 2484 if (error) { 2485 nfsm_reply(NFSX_WCCDATA(v3)); 2486 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2487 if (dirp) 2488 vrele(dirp); 2489 vn_finished_write(mp, 0); 2490 return (0); 2491 } 2492 vp = nd.ni_vp; 2493 if (vp->v_type != VDIR) { 2494 error = ENOTDIR; 2495 goto out; 2496 } 2497 /* 2498 * No rmdir "." please. 2499 */ 2500 if (nd.ni_dvp == vp) { 2501 error = EINVAL; 2502 goto out; 2503 } 2504 /* 2505 * The root of a mounted filesystem cannot be deleted. 2506 */ 2507 if (vp->v_flag & VROOT) 2508 error = EBUSY; 2509 out: 2510 if (!error) { 2511 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2512 nqsrv_getl(vp, ND_WRITE); 2513 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2514 } else { 2515 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2516 if (nd.ni_dvp == nd.ni_vp) 2517 vrele(nd.ni_dvp); 2518 else 2519 vput(nd.ni_dvp); 2520 vput(vp); 2521 } 2522 if (dirp) { 2523 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); 2524 vrele(dirp); 2525 } 2526 nfsm_reply(NFSX_WCCDATA(v3)); 2527 if (v3) { 2528 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2529 vn_finished_write(mp, 0); 2530 return (0); 2531 } 2532 vn_finished_write(mp, 0); 2533 nfsm_srvdone; 2534 } 2535 2536 /* 2537 * nfs readdir service 2538 * - mallocs what it thinks is enough to read 2539 * count rounded up to a multiple of NFS_DIRBLKSIZ <= NFS_MAXREADDIR 2540 * - calls VOP_READDIR() 2541 * - loops around building the reply 2542 * if the output generated exceeds count break out of loop 2543 * The nfsm_clget macro is used here so that the reply will be packed 2544 * tightly in mbuf clusters. 2545 * - it only knows that it has encountered eof when the VOP_READDIR() 2546 * reads nothing 2547 * - as such one readdir rpc will return eof false although you are there 2548 * and then the next will return eof 2549 * - it trims out records with d_fileno == 0 2550 * this doesn't matter for Unix clients, but they might confuse clients 2551 * for other os'. 2552 * - it trims out records with d_type == DT_WHT 2553 * these cannot be seen through NFS (unless we extend the protocol) 2554 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less 2555 * than requested, but this may not apply to all filesystems. For 2556 * example, client NFS does not { although it is never remote mounted 2557 * anyhow } 2558 * The alternate call nfsrv_readdirplus() does lookups as well. 2559 * PS: The NFS protocol spec. does not clarify what the "count" byte 2560 * argument is a count of.. just name strings and file id's or the 2561 * entire reply rpc or ... 2562 * I tried just file name and id sizes and it confused the Sun client, 2563 * so I am using the full rpc size now. The "paranoia.." comment refers 2564 * to including the status longwords that are not a part of the dir. 2565 * "entry" structures, but are in the rpc. 2566 */ 2567 struct flrep { 2568 nfsuint64 fl_off; 2569 u_int32_t fl_postopok; 2570 u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)]; 2571 u_int32_t fl_fhok; 2572 u_int32_t fl_fhsize; 2573 u_int32_t fl_nfh[NFSX_V3FH / sizeof (u_int32_t)]; 2574 }; 2575 2576 int 2577 nfsrv_readdir(nfsd, slp, procp, mrq) 2578 struct nfsrv_descript *nfsd; 2579 struct nfssvc_sock *slp; 2580 struct proc *procp; 2581 struct mbuf **mrq; 2582 { 2583 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2584 struct mbuf *nam = nfsd->nd_nam; 2585 caddr_t dpos = nfsd->nd_dpos; 2586 struct ucred *cred = &nfsd->nd_cr; 2587 char *bp, *be; 2588 struct mbuf *mp; 2589 struct dirent *dp; 2590 caddr_t cp; 2591 u_int32_t *tl; 2592 int32_t t1; 2593 caddr_t bpos; 2594 struct mbuf *mb, *mreq, *mp2; 2595 char *cpos, *cend, *cp2, *rbuf; 2596 struct vnode *vp; 2597 struct vattr at; 2598 nfsfh_t nfh; 2599 fhandle_t *fhp; 2600 struct uio io; 2601 struct iovec iv; 2602 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2603 int siz, cnt, fullsiz, eofflag, rdonly, cache, ncookies; 2604 int v3 = (nfsd->nd_flag & ND_NFSV3); 2605 u_quad_t frev, off, toff, verf; 2606 off_t *cookies = NULL, *cookiep; 2607 nfsuint64 jar; 2608 2609 fhp = &nfh.fh_generic; 2610 nfsm_srvmtofh(fhp); 2611 if (v3) { 2612 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 2613 toff = fxdr_hyper(tl); 2614 tl += 2; 2615 verf = fxdr_hyper(tl); 2616 tl += 2; 2617 } else { 2618 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2619 toff = fxdr_unsigned(u_quad_t, *tl++); 2620 } 2621 off = toff; 2622 cnt = fxdr_unsigned(int, *tl); 2623 siz = ((cnt + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1)); 2624 xfer = NFS_SRVMAXDATA(nfsd); 2625 if (siz > xfer) 2626 siz = xfer; 2627 fullsiz = siz; 2628 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 2629 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2630 if (!error && vp->v_type != VDIR) { 2631 error = ENOTDIR; 2632 vput(vp); 2633 } 2634 if (error) { 2635 nfsm_reply(NFSX_UNSIGNED); 2636 nfsm_srvpostop_attr(getret, &at); 2637 return (0); 2638 } 2639 nqsrv_getl(vp, ND_READ); 2640 if (v3) { 2641 error = getret = VOP_GETATTR(vp, &at, cred, procp); 2642 #ifdef NFS3_STRICTVERF 2643 /* 2644 * XXX This check is too strict for Solaris 2.5 clients. 2645 */ 2646 if (!error && toff && verf != at.va_filerev) 2647 error = NFSERR_BAD_COOKIE; 2648 #endif 2649 } 2650 if (!error) 2651 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0); 2652 if (error) { 2653 vput(vp); 2654 nfsm_reply(NFSX_POSTOPATTR(v3)); 2655 nfsm_srvpostop_attr(getret, &at); 2656 return (0); 2657 } 2658 VOP_UNLOCK(vp, 0); 2659 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2660 again: 2661 iv.iov_base = rbuf; 2662 iv.iov_len = fullsiz; 2663 io.uio_iov = &iv; 2664 io.uio_iovcnt = 1; 2665 io.uio_offset = (off_t)off; 2666 io.uio_resid = fullsiz; 2667 io.uio_segflg = UIO_SYSSPACE; 2668 io.uio_rw = UIO_READ; 2669 io.uio_procp = NULL; 2670 eofflag = 0; 2671 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2672 2673 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2674 2675 off = (off_t)io.uio_offset; 2676 if (!cookies && !error) 2677 error = NFSERR_PERM; 2678 if (v3) { 2679 getret = VOP_GETATTR(vp, &at, cred, procp); 2680 if (!error) 2681 error = getret; 2682 } 2683 2684 VOP_UNLOCK(vp, 0); 2685 if (error) { 2686 vrele(vp); 2687 free((caddr_t)rbuf, M_TEMP); 2688 if (cookies) 2689 free((caddr_t)cookies, M_TEMP); 2690 nfsm_reply(NFSX_POSTOPATTR(v3)); 2691 nfsm_srvpostop_attr(getret, &at); 2692 return (0); 2693 } 2694 if (io.uio_resid) { 2695 siz -= io.uio_resid; 2696 2697 /* 2698 * If nothing read, return eof 2699 * rpc reply 2700 */ 2701 if (siz == 0) { 2702 vrele(vp); 2703 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + 2704 2 * NFSX_UNSIGNED); 2705 if (v3) { 2706 nfsm_srvpostop_attr(getret, &at); 2707 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2708 txdr_hyper(at.va_filerev, tl); 2709 tl += 2; 2710 } else 2711 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2712 *tl++ = nfs_false; 2713 *tl = nfs_true; 2714 free((caddr_t)rbuf, M_TEMP); 2715 free((caddr_t)cookies, M_TEMP); 2716 return (0); 2717 } 2718 } 2719 2720 /* 2721 * Check for degenerate cases of nothing useful read. 2722 * If so go try again 2723 */ 2724 cpos = rbuf; 2725 cend = rbuf + siz; 2726 dp = (struct dirent *)cpos; 2727 cookiep = cookies; 2728 2729 while (cpos < cend && ncookies > 0 && 2730 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2731 cpos += dp->d_reclen; 2732 dp = (struct dirent *)cpos; 2733 cookiep++; 2734 ncookies--; 2735 } 2736 if (cpos >= cend || ncookies == 0) { 2737 toff = off; 2738 siz = fullsiz; 2739 goto again; 2740 } 2741 2742 len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */ 2743 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz); 2744 if (v3) { 2745 nfsm_srvpostop_attr(getret, &at); 2746 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2747 txdr_hyper(at.va_filerev, tl); 2748 } 2749 mp = mp2 = mb; 2750 bp = bpos; 2751 be = bp + M_TRAILINGSPACE(mp); 2752 2753 /* Loop through the records and build reply */ 2754 while (cpos < cend && ncookies > 0) { 2755 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 2756 nlen = dp->d_namlen; 2757 rem = nfsm_rndup(nlen)-nlen; 2758 len += (4 * NFSX_UNSIGNED + nlen + rem); 2759 if (v3) 2760 len += 2 * NFSX_UNSIGNED; 2761 if (len > cnt) { 2762 eofflag = 0; 2763 break; 2764 } 2765 /* 2766 * Build the directory record xdr from 2767 * the dirent entry. 2768 */ 2769 nfsm_clget; 2770 *tl = nfs_true; 2771 bp += NFSX_UNSIGNED; 2772 if (v3) { 2773 nfsm_clget; 2774 *tl = 0; 2775 bp += NFSX_UNSIGNED; 2776 } 2777 nfsm_clget; 2778 *tl = txdr_unsigned(dp->d_fileno); 2779 bp += NFSX_UNSIGNED; 2780 nfsm_clget; 2781 *tl = txdr_unsigned(nlen); 2782 bp += NFSX_UNSIGNED; 2783 2784 /* And loop around copying the name */ 2785 xfer = nlen; 2786 cp = dp->d_name; 2787 while (xfer > 0) { 2788 nfsm_clget; 2789 if ((bp+xfer) > be) 2790 tsiz = be-bp; 2791 else 2792 tsiz = xfer; 2793 memcpy(bp, cp, tsiz); 2794 bp += tsiz; 2795 xfer -= tsiz; 2796 if (xfer > 0) 2797 cp += tsiz; 2798 } 2799 /* And null pad to an int32_t boundary */ 2800 for (i = 0; i < rem; i++) 2801 *bp++ = '\0'; 2802 nfsm_clget; 2803 2804 /* Finish off the record */ 2805 txdr_hyper(*cookiep, &jar); 2806 if (v3) { 2807 *tl = jar.nfsuquad[0]; 2808 bp += NFSX_UNSIGNED; 2809 nfsm_clget; 2810 } 2811 *tl = jar.nfsuquad[1]; 2812 bp += NFSX_UNSIGNED; 2813 } 2814 cpos += dp->d_reclen; 2815 dp = (struct dirent *)cpos; 2816 cookiep++; 2817 ncookies--; 2818 } 2819 vrele(vp); 2820 nfsm_clget; 2821 *tl = nfs_false; 2822 bp += NFSX_UNSIGNED; 2823 nfsm_clget; 2824 if (eofflag) 2825 *tl = nfs_true; 2826 else 2827 *tl = nfs_false; 2828 bp += NFSX_UNSIGNED; 2829 if (mp != mb) { 2830 if (bp < be) 2831 mp->m_len = bp - mtod(mp, caddr_t); 2832 } else 2833 mp->m_len += bp - bpos; 2834 free((caddr_t)rbuf, M_TEMP); 2835 free((caddr_t)cookies, M_TEMP); 2836 nfsm_srvdone; 2837 } 2838 2839 int 2840 nfsrv_readdirplus(nfsd, slp, procp, mrq) 2841 struct nfsrv_descript *nfsd; 2842 struct nfssvc_sock *slp; 2843 struct proc *procp; 2844 struct mbuf **mrq; 2845 { 2846 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2847 struct mbuf *nam = nfsd->nd_nam; 2848 caddr_t dpos = nfsd->nd_dpos; 2849 struct ucred *cred = &nfsd->nd_cr; 2850 char *bp, *be; 2851 struct mbuf *mp; 2852 struct dirent *dp; 2853 caddr_t cp; 2854 u_int32_t *tl; 2855 int32_t t1; 2856 caddr_t bpos; 2857 struct mbuf *mb, *mreq, *mp2; 2858 char *cpos, *cend, *cp2, *rbuf; 2859 struct vnode *vp, *nvp; 2860 struct flrep fl; 2861 nfsfh_t nfh; 2862 fhandle_t *fhp, *nfhp = (fhandle_t *)fl.fl_nfh; 2863 struct uio io; 2864 struct iovec iv; 2865 struct vattr va, at, *vap = &va; 2866 struct nfs_fattr *fp; 2867 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2868 int siz, cnt, fullsiz, eofflag, rdonly, cache, dirlen, ncookies; 2869 u_quad_t frev, off, toff, verf; 2870 off_t *cookies = NULL, *cookiep; 2871 2872 fhp = &nfh.fh_generic; 2873 nfsm_srvmtofh(fhp); 2874 nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED); 2875 toff = fxdr_hyper(tl); 2876 tl += 2; 2877 verf = fxdr_hyper(tl); 2878 tl += 2; 2879 siz = fxdr_unsigned(int, *tl++); 2880 cnt = fxdr_unsigned(int, *tl); 2881 off = toff; 2882 siz = ((siz + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1)); 2883 xfer = NFS_SRVMAXDATA(nfsd); 2884 if (siz > xfer) 2885 siz = xfer; 2886 fullsiz = siz; 2887 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 2888 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2889 if (!error && vp->v_type != VDIR) { 2890 error = ENOTDIR; 2891 vput(vp); 2892 } 2893 if (error) { 2894 nfsm_reply(NFSX_UNSIGNED); 2895 nfsm_srvpostop_attr(getret, &at); 2896 return (0); 2897 } 2898 error = getret = VOP_GETATTR(vp, &at, cred, procp); 2899 #ifdef NFS3_STRICTVERF 2900 /* 2901 * XXX This check is too strict for Solaris 2.5 clients. 2902 */ 2903 if (!error && toff && verf != at.va_filerev) 2904 error = NFSERR_BAD_COOKIE; 2905 #endif 2906 if (!error) { 2907 nqsrv_getl(vp, ND_READ); 2908 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0); 2909 } 2910 if (error) { 2911 vput(vp); 2912 nfsm_reply(NFSX_V3POSTOPATTR); 2913 nfsm_srvpostop_attr(getret, &at); 2914 return (0); 2915 } 2916 VOP_UNLOCK(vp, 0); 2917 2918 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2919 again: 2920 iv.iov_base = rbuf; 2921 iv.iov_len = fullsiz; 2922 io.uio_iov = &iv; 2923 io.uio_iovcnt = 1; 2924 io.uio_offset = (off_t)off; 2925 io.uio_resid = fullsiz; 2926 io.uio_segflg = UIO_SYSSPACE; 2927 io.uio_rw = UIO_READ; 2928 io.uio_procp = NULL; 2929 eofflag = 0; 2930 2931 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2932 2933 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2934 2935 off = (u_quad_t)io.uio_offset; 2936 getret = VOP_GETATTR(vp, &at, cred, procp); 2937 2938 VOP_UNLOCK(vp, 0); 2939 2940 /* 2941 * If the VGET operation doesn't work for this filesystem, 2942 * we can't support readdirplus. Returning NOTSUPP should 2943 * make clients fall back to plain readdir. 2944 * There's no need to check for VPTOFH as well, we wouldn't 2945 * even be here otherwise. 2946 */ 2947 if (!getret) { 2948 if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp))) 2949 getret = (getret == EOPNOTSUPP) ? 2950 NFSERR_NOTSUPP : NFSERR_IO; 2951 else 2952 vput(nvp); 2953 } 2954 2955 if (!cookies && !error) 2956 error = NFSERR_PERM; 2957 if (!error) 2958 error = getret; 2959 if (error) { 2960 vrele(vp); 2961 if (cookies) 2962 free((caddr_t)cookies, M_TEMP); 2963 free((caddr_t)rbuf, M_TEMP); 2964 nfsm_reply(NFSX_V3POSTOPATTR); 2965 nfsm_srvpostop_attr(getret, &at); 2966 return (0); 2967 } 2968 if (io.uio_resid) { 2969 siz -= io.uio_resid; 2970 2971 /* 2972 * If nothing read, return eof 2973 * rpc reply 2974 */ 2975 if (siz == 0) { 2976 vrele(vp); 2977 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2978 2 * NFSX_UNSIGNED); 2979 nfsm_srvpostop_attr(getret, &at); 2980 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2981 txdr_hyper(at.va_filerev, tl); 2982 tl += 2; 2983 *tl++ = nfs_false; 2984 *tl = nfs_true; 2985 free((caddr_t)cookies, M_TEMP); 2986 free((caddr_t)rbuf, M_TEMP); 2987 return (0); 2988 } 2989 } 2990 2991 /* 2992 * Check for degenerate cases of nothing useful read. 2993 * If so go try again 2994 */ 2995 cpos = rbuf; 2996 cend = rbuf + siz; 2997 dp = (struct dirent *)cpos; 2998 cookiep = cookies; 2999 3000 while (cpos < cend && ncookies > 0 && 3001 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 3002 cpos += dp->d_reclen; 3003 dp = (struct dirent *)cpos; 3004 cookiep++; 3005 ncookies--; 3006 } 3007 if (cpos >= cend || ncookies == 0) { 3008 toff = off; 3009 siz = fullsiz; 3010 goto again; 3011 } 3012 3013 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED; 3014 nfsm_reply(cnt); 3015 nfsm_srvpostop_attr(getret, &at); 3016 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 3017 txdr_hyper(at.va_filerev, tl); 3018 mp = mp2 = mb; 3019 bp = bpos; 3020 be = bp + M_TRAILINGSPACE(mp); 3021 3022 /* Loop through the records and build reply */ 3023 while (cpos < cend && ncookies > 0) { 3024 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 3025 nlen = dp->d_namlen; 3026 rem = nfsm_rndup(nlen)-nlen; 3027 3028 /* 3029 * For readdir_and_lookup get the vnode using 3030 * the file number. 3031 */ 3032 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp)) 3033 goto invalid; 3034 memset((caddr_t)nfhp, 0, NFSX_V3FH); 3035 nfhp->fh_fsid = 3036 nvp->v_mount->mnt_stat.f_fsidx; 3037 if (VFS_VPTOFH(nvp, &nfhp->fh_fid)) { 3038 vput(nvp); 3039 goto invalid; 3040 } 3041 if (VOP_GETATTR(nvp, vap, cred, procp)) { 3042 vput(nvp); 3043 goto invalid; 3044 } 3045 vput(nvp); 3046 3047 /* 3048 * If either the dircount or maxcount will be 3049 * exceeded, get out now. Both of these lengths 3050 * are calculated conservatively, including all 3051 * XDR overheads. 3052 */ 3053 len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH + 3054 NFSX_V3POSTOPATTR); 3055 dirlen += (6 * NFSX_UNSIGNED + nlen + rem); 3056 if (len > cnt || dirlen > fullsiz) { 3057 eofflag = 0; 3058 break; 3059 } 3060 3061 /* 3062 * Build the directory record xdr from 3063 * the dirent entry. 3064 */ 3065 fp = (struct nfs_fattr *)&fl.fl_fattr; 3066 nfsm_srvfillattr(vap, fp); 3067 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH); 3068 fl.fl_fhok = nfs_true; 3069 fl.fl_postopok = nfs_true; 3070 txdr_hyper(*cookiep, fl.fl_off.nfsuquad); 3071 3072 nfsm_clget; 3073 *tl = nfs_true; 3074 bp += NFSX_UNSIGNED; 3075 nfsm_clget; 3076 *tl = 0; 3077 bp += NFSX_UNSIGNED; 3078 nfsm_clget; 3079 *tl = txdr_unsigned(dp->d_fileno); 3080 bp += NFSX_UNSIGNED; 3081 nfsm_clget; 3082 *tl = txdr_unsigned(nlen); 3083 bp += NFSX_UNSIGNED; 3084 3085 /* And loop around copying the name */ 3086 xfer = nlen; 3087 cp = dp->d_name; 3088 while (xfer > 0) { 3089 nfsm_clget; 3090 if ((bp + xfer) > be) 3091 tsiz = be - bp; 3092 else 3093 tsiz = xfer; 3094 memcpy(bp, cp, tsiz); 3095 bp += tsiz; 3096 xfer -= tsiz; 3097 if (xfer > 0) 3098 cp += tsiz; 3099 } 3100 /* And null pad to an int32_t boundary */ 3101 for (i = 0; i < rem; i++) 3102 *bp++ = '\0'; 3103 3104 /* 3105 * Now copy the flrep structure out. 3106 */ 3107 xfer = sizeof (struct flrep); 3108 cp = (caddr_t)&fl; 3109 while (xfer > 0) { 3110 nfsm_clget; 3111 if ((bp + xfer) > be) 3112 tsiz = be - bp; 3113 else 3114 tsiz = xfer; 3115 memcpy(bp, cp, tsiz); 3116 bp += tsiz; 3117 xfer -= tsiz; 3118 if (xfer > 0) 3119 cp += tsiz; 3120 } 3121 } 3122 invalid: 3123 cpos += dp->d_reclen; 3124 dp = (struct dirent *)cpos; 3125 cookiep++; 3126 ncookies--; 3127 } 3128 vrele(vp); 3129 nfsm_clget; 3130 *tl = nfs_false; 3131 bp += NFSX_UNSIGNED; 3132 nfsm_clget; 3133 if (eofflag) 3134 *tl = nfs_true; 3135 else 3136 *tl = nfs_false; 3137 bp += NFSX_UNSIGNED; 3138 if (mp != mb) { 3139 if (bp < be) 3140 mp->m_len = bp - mtod(mp, caddr_t); 3141 } else 3142 mp->m_len += bp - bpos; 3143 free((caddr_t)cookies, M_TEMP); 3144 free((caddr_t)rbuf, M_TEMP); 3145 nfsm_srvdone; 3146 } 3147 3148 /* 3149 * nfs commit service 3150 */ 3151 int 3152 nfsrv_commit(nfsd, slp, procp, mrq) 3153 struct nfsrv_descript *nfsd; 3154 struct nfssvc_sock *slp; 3155 struct proc *procp; 3156 struct mbuf **mrq; 3157 { 3158 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3159 struct mbuf *nam = nfsd->nd_nam; 3160 caddr_t dpos = nfsd->nd_dpos; 3161 struct ucred *cred = &nfsd->nd_cr; 3162 struct vattr bfor, aft; 3163 struct vnode *vp; 3164 nfsfh_t nfh; 3165 fhandle_t *fhp; 3166 u_int32_t *tl; 3167 int32_t t1; 3168 caddr_t bpos; 3169 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cache; 3170 uint32_t cnt; 3171 char *cp2; 3172 struct mbuf *mb, *mreq; 3173 u_quad_t frev, off, end; 3174 struct mount *mp = NULL; 3175 3176 #ifndef nolint 3177 cache = 0; 3178 #endif 3179 fhp = &nfh.fh_generic; 3180 nfsm_srvmtofh(fhp); 3181 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 3182 return (ESTALE); 3183 vn_start_write(NULL, &mp, V_WAIT); 3184 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED); 3185 3186 off = fxdr_hyper(tl); 3187 tl += 2; 3188 cnt = fxdr_unsigned(uint32_t, *tl); 3189 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3190 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3191 if (error) { 3192 nfsm_reply(2 * NFSX_UNSIGNED); 3193 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3194 vn_finished_write(mp, 0); 3195 return (0); 3196 } 3197 for_ret = VOP_GETATTR(vp, &bfor, cred, procp); 3198 end = (cnt > 0) ? off + cnt : vp->v_size; 3199 if (end < off || end > vp->v_size) 3200 end = vp->v_size; 3201 if (off < vp->v_size) 3202 error = VOP_FSYNC(vp, cred, FSYNC_WAIT, off, end, procp); 3203 /* else error == 0, from nfsrv_fhtovp() */ 3204 aft_ret = VOP_GETATTR(vp, &aft, cred, procp); 3205 vput(vp); 3206 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF); 3207 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3208 if (!error) { 3209 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF); 3210 *tl++ = txdr_unsigned(boottime.tv_sec); 3211 *tl = txdr_unsigned(boottime.tv_usec); 3212 } else { 3213 vn_finished_write(mp, 0); 3214 return (0); 3215 } 3216 vn_finished_write(mp, 0); 3217 nfsm_srvdone; 3218 } 3219 3220 /* 3221 * nfs statfs service 3222 */ 3223 int 3224 nfsrv_statfs(nfsd, slp, procp, mrq) 3225 struct nfsrv_descript *nfsd; 3226 struct nfssvc_sock *slp; 3227 struct proc *procp; 3228 struct mbuf **mrq; 3229 { 3230 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3231 struct mbuf *nam = nfsd->nd_nam; 3232 caddr_t dpos = nfsd->nd_dpos; 3233 struct ucred *cred = &nfsd->nd_cr; 3234 struct statvfs *sf; 3235 struct nfs_statfs *sfp; 3236 u_int32_t *tl; 3237 int32_t t1; 3238 caddr_t bpos; 3239 int error = 0, rdonly, cache, getret = 1; 3240 int v3 = (nfsd->nd_flag & ND_NFSV3); 3241 char *cp2; 3242 struct mbuf *mb, *mreq; 3243 struct vnode *vp; 3244 struct vattr at; 3245 nfsfh_t nfh; 3246 fhandle_t *fhp; 3247 struct statvfs statvfs; 3248 u_quad_t frev, tval; 3249 3250 #ifndef nolint 3251 cache = 0; 3252 #endif 3253 fhp = &nfh.fh_generic; 3254 nfsm_srvmtofh(fhp); 3255 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3256 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3257 if (error) { 3258 nfsm_reply(NFSX_UNSIGNED); 3259 nfsm_srvpostop_attr(getret, &at); 3260 return (0); 3261 } 3262 sf = &statvfs; 3263 error = VFS_STATVFS(vp->v_mount, sf, procp); 3264 getret = VOP_GETATTR(vp, &at, cred, procp); 3265 vput(vp); 3266 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3)); 3267 if (v3) 3268 nfsm_srvpostop_attr(getret, &at); 3269 if (error) 3270 return (0); 3271 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3)); 3272 if (v3) { 3273 tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_frsize); 3274 txdr_hyper(tval, &sfp->sf_tbytes); 3275 tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_frsize); 3276 txdr_hyper(tval, &sfp->sf_fbytes); 3277 tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_frsize); 3278 txdr_hyper(tval, &sfp->sf_abytes); 3279 tval = (u_quad_t)sf->f_files; 3280 txdr_hyper(tval, &sfp->sf_tfiles); 3281 tval = (u_quad_t)sf->f_ffree; 3282 txdr_hyper(tval, &sfp->sf_ffiles); 3283 txdr_hyper(tval, &sfp->sf_afiles); 3284 sfp->sf_invarsec = 0; 3285 } else { 3286 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA); 3287 sfp->sf_bsize = txdr_unsigned(sf->f_frsize); 3288 sfp->sf_blocks = txdr_unsigned(sf->f_blocks); 3289 sfp->sf_bfree = txdr_unsigned(sf->f_bfree); 3290 sfp->sf_bavail = txdr_unsigned(sf->f_bavail); 3291 } 3292 nfsm_srvdone; 3293 } 3294 3295 /* 3296 * nfs fsinfo service 3297 */ 3298 int 3299 nfsrv_fsinfo(nfsd, slp, procp, mrq) 3300 struct nfsrv_descript *nfsd; 3301 struct nfssvc_sock *slp; 3302 struct proc *procp; 3303 struct mbuf **mrq; 3304 { 3305 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3306 struct mbuf *nam = nfsd->nd_nam; 3307 caddr_t dpos = nfsd->nd_dpos; 3308 struct ucred *cred = &nfsd->nd_cr; 3309 u_int32_t *tl; 3310 struct nfsv3_fsinfo *sip; 3311 int32_t t1; 3312 caddr_t bpos; 3313 int error = 0, rdonly, cache, getret = 1; 3314 uint32_t maxdata; 3315 char *cp2; 3316 struct mbuf *mb, *mreq; 3317 struct vnode *vp; 3318 struct vattr at; 3319 nfsfh_t nfh; 3320 fhandle_t *fhp; 3321 u_quad_t frev, maxfsize; 3322 struct statvfs sb; 3323 3324 #ifndef nolint 3325 cache = 0; 3326 #endif 3327 fhp = &nfh.fh_generic; 3328 nfsm_srvmtofh(fhp); 3329 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3330 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3331 if (error) { 3332 nfsm_reply(NFSX_UNSIGNED); 3333 nfsm_srvpostop_attr(getret, &at); 3334 return (0); 3335 } 3336 3337 /* XXX Try to make a guess on the max file size. */ 3338 VFS_STATVFS(vp->v_mount, &sb, (struct proc *)0); 3339 maxfsize = (u_quad_t)0x80000000 * sb.f_frsize - 1; 3340 3341 getret = VOP_GETATTR(vp, &at, cred, procp); 3342 vput(vp); 3343 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO); 3344 nfsm_srvpostop_attr(getret, &at); 3345 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO); 3346 3347 /* 3348 * XXX 3349 * There should be file system VFS OP(s) to get this information. 3350 * For now, assume ufs. 3351 */ 3352 if (slp->ns_so->so_type == SOCK_DGRAM) 3353 maxdata = NFS_MAXDGRAMDATA; 3354 else 3355 maxdata = NFS_MAXDATA; 3356 sip->fs_rtmax = txdr_unsigned(maxdata); 3357 sip->fs_rtpref = txdr_unsigned(maxdata); 3358 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE); 3359 sip->fs_wtmax = txdr_unsigned(maxdata); 3360 sip->fs_wtpref = txdr_unsigned(maxdata); 3361 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE); 3362 sip->fs_dtpref = txdr_unsigned(maxdata); 3363 txdr_hyper(maxfsize, &sip->fs_maxfilesize); 3364 sip->fs_timedelta.nfsv3_sec = 0; 3365 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1); 3366 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK | 3367 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS | 3368 NFSV3FSINFO_CANSETTIME); 3369 nfsm_srvdone; 3370 } 3371 3372 /* 3373 * nfs pathconf service 3374 */ 3375 int 3376 nfsrv_pathconf(nfsd, slp, procp, mrq) 3377 struct nfsrv_descript *nfsd; 3378 struct nfssvc_sock *slp; 3379 struct proc *procp; 3380 struct mbuf **mrq; 3381 { 3382 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3383 struct mbuf *nam = nfsd->nd_nam; 3384 caddr_t dpos = nfsd->nd_dpos; 3385 struct ucred *cred = &nfsd->nd_cr; 3386 u_int32_t *tl; 3387 struct nfsv3_pathconf *pc; 3388 int32_t t1; 3389 caddr_t bpos; 3390 int error = 0, rdonly, cache, getret = 1; 3391 register_t linkmax, namemax, chownres, notrunc; 3392 char *cp2; 3393 struct mbuf *mb, *mreq; 3394 struct vnode *vp; 3395 struct vattr at; 3396 nfsfh_t nfh; 3397 fhandle_t *fhp; 3398 u_quad_t frev; 3399 3400 #ifndef nolint 3401 cache = 0; 3402 #endif 3403 fhp = &nfh.fh_generic; 3404 nfsm_srvmtofh(fhp); 3405 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3406 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3407 if (error) { 3408 nfsm_reply(NFSX_UNSIGNED); 3409 nfsm_srvpostop_attr(getret, &at); 3410 return (0); 3411 } 3412 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax); 3413 if (!error) 3414 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax); 3415 if (!error) 3416 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres); 3417 if (!error) 3418 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc); 3419 getret = VOP_GETATTR(vp, &at, cred, procp); 3420 vput(vp); 3421 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF); 3422 nfsm_srvpostop_attr(getret, &at); 3423 if (error) 3424 return (0); 3425 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF); 3426 3427 pc->pc_linkmax = txdr_unsigned(linkmax); 3428 pc->pc_namemax = txdr_unsigned(namemax); 3429 pc->pc_notrunc = txdr_unsigned(notrunc); 3430 pc->pc_chownrestricted = txdr_unsigned(chownres); 3431 3432 /* 3433 * These should probably be supported by VOP_PATHCONF(), but 3434 * until msdosfs is exportable (why would you want to?), the 3435 * Unix defaults should be ok. 3436 */ 3437 pc->pc_caseinsensitive = nfs_false; 3438 pc->pc_casepreserving = nfs_true; 3439 nfsm_srvdone; 3440 } 3441 3442 /* 3443 * Null operation, used by clients to ping server 3444 */ 3445 /* ARGSUSED */ 3446 int 3447 nfsrv_null(nfsd, slp, procp, mrq) 3448 struct nfsrv_descript *nfsd; 3449 struct nfssvc_sock *slp; 3450 struct proc *procp; 3451 struct mbuf **mrq; 3452 { 3453 struct mbuf *mrep = nfsd->nd_mrep; 3454 caddr_t bpos; 3455 int error = NFSERR_RETVOID, cache = 0; 3456 struct mbuf *mb, *mreq; 3457 u_quad_t frev; 3458 3459 nfsm_reply(0); 3460 return (0); 3461 } 3462 3463 /* 3464 * No operation, used for obsolete procedures 3465 */ 3466 /* ARGSUSED */ 3467 int 3468 nfsrv_noop(nfsd, slp, procp, mrq) 3469 struct nfsrv_descript *nfsd; 3470 struct nfssvc_sock *slp; 3471 struct proc *procp; 3472 struct mbuf **mrq; 3473 { 3474 struct mbuf *mrep = nfsd->nd_mrep; 3475 caddr_t bpos; 3476 int error, cache = 0; 3477 struct mbuf *mb, *mreq; 3478 u_quad_t frev; 3479 3480 if (nfsd->nd_repstat) 3481 error = nfsd->nd_repstat; 3482 else 3483 error = EPROCUNAVAIL; 3484 nfsm_reply(0); 3485 return (0); 3486 } 3487 3488 /* 3489 * Perform access checking for vnodes obtained from file handles that would 3490 * refer to files already opened by a Unix client. You cannot just use 3491 * vn_writechk() and VOP_ACCESS() for two reasons. 3492 * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case 3493 * 2 - The owner is to be given access irrespective of mode bits for some 3494 * operations, so that processes that chmod after opening a file don't 3495 * break. I don't like this because it opens a security hole, but since 3496 * the nfs server opens a security hole the size of a barn door anyhow, 3497 * what the heck. 3498 * 3499 * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS() 3500 * will return EPERM instead of EACCESS. EPERM is always an error. 3501 */ 3502 int 3503 nfsrv_access(vp, flags, cred, rdonly, p, override) 3504 struct vnode *vp; 3505 int flags; 3506 struct ucred *cred; 3507 int rdonly; 3508 struct proc *p; 3509 { 3510 struct vattr vattr; 3511 int error; 3512 if (flags & VWRITE) { 3513 /* Just vn_writechk() changed to check rdonly */ 3514 /* 3515 * Disallow write attempts on read-only file systems; 3516 * unless the file is a socket or a block or character 3517 * device resident on the file system. 3518 */ 3519 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 3520 switch (vp->v_type) { 3521 case VREG: 3522 case VDIR: 3523 case VLNK: 3524 return (EROFS); 3525 default: 3526 break; 3527 } 3528 } 3529 3530 /* 3531 * If the vnode is in use as a process's text, 3532 * we can't allow writing. 3533 */ 3534 if (vp->v_flag & VTEXT) 3535 return (ETXTBSY); 3536 } 3537 error = VOP_GETATTR(vp, &vattr, cred, p); 3538 if (error) 3539 return (error); 3540 error = VOP_ACCESS(vp, flags, cred, p); 3541 /* 3542 * Allow certain operations for the owner (reads and writes 3543 * on files that are already open). 3544 */ 3545 if (override && error == EACCES && cred->cr_uid == vattr.va_uid) 3546 error = 0; 3547 return error; 3548 } 3549
Indexes created Thu Oct 02 14:10:14 GMT 2025