nfs_serv.c revision 1.98
1 /* $NetBSD: nfs_serv.c,v 1.98 2005/10/06 10:23:01 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 1.98 2005/10/06 10:23:01 yamt Exp $"); 59 60 #include <sys/param.h> 61 #include <sys/systm.h> 62 #include <sys/proc.h> 63 #include <sys/file.h> 64 #include <sys/namei.h> 65 #include <sys/vnode.h> 66 #include <sys/mount.h> 67 #include <sys/socket.h> 68 #include <sys/socketvar.h> 69 #include <sys/mbuf.h> 70 #include <sys/dirent.h> 71 #include <sys/stat.h> 72 #include <sys/kernel.h> 73 74 #include <uvm/uvm.h> 75 76 #include <nfs/nfsproto.h> 77 #include <nfs/rpcv2.h> 78 #include <nfs/nfs.h> 79 #include <nfs/xdr_subs.h> 80 #include <nfs/nfsm_subs.h> 81 #include <nfs/nqnfs.h> 82 #include <nfs/nfs_var.h> 83 84 /* Global vars */ 85 extern u_int32_t nfs_xdrneg1; 86 extern u_int32_t nfs_false, nfs_true; 87 extern enum vtype nv3tov_type[8]; 88 extern struct nfsstats nfsstats; 89 extern const nfstype nfsv2_type[9]; 90 extern const nfstype nfsv3_type[9]; 91 int nfsrvw_procrastinate = NFS_GATHERDELAY * 1000; 92 int nfsd_use_loan = 1; /* use page-loan for READ OP */ 93 94 /* 95 * nfs v3 access service 96 */ 97 int 98 nfsrv3_access(nfsd, slp, procp, mrq) 99 struct nfsrv_descript *nfsd; 100 struct nfssvc_sock *slp; 101 struct proc *procp; 102 struct mbuf **mrq; 103 { 104 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 105 struct mbuf *nam = nfsd->nd_nam; 106 caddr_t dpos = nfsd->nd_dpos; 107 struct ucred *cred = &nfsd->nd_cr; 108 struct vnode *vp; 109 nfsfh_t nfh; 110 fhandle_t *fhp; 111 u_int32_t *tl; 112 int32_t t1; 113 caddr_t bpos; 114 int error = 0, rdonly, cache = 0, getret; 115 char *cp2; 116 struct mbuf *mb, *mreq; 117 struct vattr va; 118 u_long inmode, testmode, outmode; 119 u_quad_t frev; 120 121 fhp = &nfh.fh_generic; 122 nfsm_srvmtofh(fhp); 123 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 124 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, 125 (nfsd->nd_flag & ND_KERBAUTH), FALSE); 126 if (error) { 127 nfsm_reply(NFSX_UNSIGNED); 128 nfsm_srvpostop_attr(1, (struct vattr *)0); 129 return (0); 130 } 131 inmode = fxdr_unsigned(u_int32_t, *tl); 132 outmode = 0; 133 if ((inmode & NFSV3ACCESS_READ) && 134 nfsrv_access(vp, VREAD, cred, rdonly, procp, 0) == 0) 135 outmode |= NFSV3ACCESS_READ; 136 if (vp->v_type != VDIR) { 137 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND); 138 if (testmode && 139 nfsrv_access(vp, VWRITE, cred, rdonly, procp, 0) == 0) 140 outmode |= testmode; 141 if ((inmode & NFSV3ACCESS_EXECUTE) && 142 nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0) == 0) 143 outmode |= NFSV3ACCESS_EXECUTE; 144 } else { 145 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND | 146 NFSV3ACCESS_DELETE); 147 if (testmode && 148 nfsrv_access(vp, VWRITE, cred, rdonly, procp, 0) == 0) 149 outmode |= testmode; 150 if ((inmode & NFSV3ACCESS_LOOKUP) && 151 nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0) == 0) 152 outmode |= NFSV3ACCESS_LOOKUP; 153 } 154 getret = VOP_GETATTR(vp, &va, cred, procp); 155 vput(vp); 156 nfsm_reply(NFSX_POSTOPATTR(1) + NFSX_UNSIGNED); 157 nfsm_srvpostop_attr(getret, &va); 158 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 159 *tl = txdr_unsigned(outmode); 160 nfsm_srvdone; 161 } 162 163 /* 164 * nfs getattr service 165 */ 166 int 167 nfsrv_getattr(nfsd, slp, procp, mrq) 168 struct nfsrv_descript *nfsd; 169 struct nfssvc_sock *slp; 170 struct proc *procp; 171 struct mbuf **mrq; 172 { 173 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 174 struct mbuf *nam = nfsd->nd_nam; 175 caddr_t dpos = nfsd->nd_dpos; 176 struct ucred *cred = &nfsd->nd_cr; 177 struct nfs_fattr *fp; 178 struct vattr va; 179 struct vnode *vp; 180 nfsfh_t nfh; 181 fhandle_t *fhp; 182 u_int32_t *tl; 183 int32_t t1; 184 caddr_t bpos; 185 int error = 0, rdonly, cache; 186 char *cp2; 187 struct mbuf *mb, *mreq; 188 u_quad_t frev; 189 190 fhp = &nfh.fh_generic; 191 nfsm_srvmtofh(fhp); 192 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, 193 (nfsd->nd_flag & ND_KERBAUTH), FALSE); 194 if (error) { 195 nfsm_reply(0); 196 return (0); 197 } 198 nqsrv_getl(vp, ND_READ); 199 error = VOP_GETATTR(vp, &va, cred, procp); 200 vput(vp); 201 nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); 202 if (error) 203 return (0); 204 nfsm_build(fp, struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); 205 nfsm_srvfillattr(&va, fp); 206 nfsm_srvdone; 207 } 208 209 /* 210 * nfs setattr service 211 */ 212 int 213 nfsrv_setattr(nfsd, slp, procp, mrq) 214 struct nfsrv_descript *nfsd; 215 struct nfssvc_sock *slp; 216 struct proc *procp; 217 struct mbuf **mrq; 218 { 219 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 220 struct mbuf *nam = nfsd->nd_nam; 221 caddr_t dpos = nfsd->nd_dpos; 222 struct ucred *cred = &nfsd->nd_cr; 223 struct vattr va, preat; 224 struct nfsv2_sattr *sp; 225 struct nfs_fattr *fp; 226 struct vnode *vp; 227 nfsfh_t nfh; 228 fhandle_t *fhp; 229 u_int32_t *tl; 230 int32_t t1; 231 caddr_t bpos; 232 int error = 0, rdonly, cache, preat_ret = 1, postat_ret = 1; 233 int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0; 234 char *cp2; 235 struct mbuf *mb, *mreq; 236 u_quad_t frev; 237 struct timespec guard; 238 struct mount *mp = NULL; 239 240 fhp = &nfh.fh_generic; 241 nfsm_srvmtofh(fhp); 242 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 243 return (ESTALE); 244 vn_start_write(NULL, &mp, V_WAIT); 245 VATTR_NULL(&va); 246 if (v3) { 247 nfsm_srvsattr(&va); 248 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 249 gcheck = fxdr_unsigned(int, *tl); 250 if (gcheck) { 251 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 252 fxdr_nfsv3time(tl, &guard); 253 } 254 } else { 255 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 256 /* 257 * Nah nah nah nah na nah 258 * There is a bug in the Sun client that puts 0xffff in the mode 259 * field of sattr when it should put in 0xffffffff. The u_short 260 * doesn't sign extend. 261 * --> check the low order 2 bytes for 0xffff 262 */ 263 if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff) 264 va.va_mode = nfstov_mode(sp->sa_mode); 265 if (sp->sa_uid != nfs_xdrneg1) 266 va.va_uid = fxdr_unsigned(uid_t, sp->sa_uid); 267 if (sp->sa_gid != nfs_xdrneg1) 268 va.va_gid = fxdr_unsigned(gid_t, sp->sa_gid); 269 if (sp->sa_size != nfs_xdrneg1) 270 va.va_size = fxdr_unsigned(u_quad_t, sp->sa_size); 271 if (sp->sa_atime.nfsv2_sec != nfs_xdrneg1) { 272 #ifdef notyet 273 fxdr_nfsv2time(&sp->sa_atime, &va.va_atime); 274 #else 275 va.va_atime.tv_sec = 276 fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec); 277 va.va_atime.tv_nsec = 0; 278 #endif 279 } 280 if (sp->sa_mtime.nfsv2_sec != nfs_xdrneg1) 281 fxdr_nfsv2time(&sp->sa_mtime, &va.va_mtime); 282 283 } 284 285 /* 286 * Now that we have all the fields, lets do it. 287 */ 288 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, 289 (nfsd->nd_flag & ND_KERBAUTH), FALSE); 290 if (error) { 291 nfsm_reply(2 * NFSX_UNSIGNED); 292 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 293 vn_finished_write(mp, 0); 294 return (0); 295 } 296 nqsrv_getl(vp, ND_WRITE); 297 if (v3) { 298 error = preat_ret = VOP_GETATTR(vp, &preat, cred, procp); 299 if (!error && gcheck && 300 (preat.va_ctime.tv_sec != guard.tv_sec || 301 preat.va_ctime.tv_nsec != guard.tv_nsec)) 302 error = NFSERR_NOT_SYNC; 303 if (error) { 304 vput(vp); 305 nfsm_reply(NFSX_WCCDATA(v3)); 306 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 307 vn_finished_write(mp, 0); 308 return (0); 309 } 310 } 311 312 /* 313 * If the size is being changed write acces is required, otherwise 314 * just check for a read only file system. 315 */ 316 if (va.va_size == ((u_quad_t)((quad_t) -1))) { 317 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 318 error = EROFS; 319 goto out; 320 } 321 } else { 322 if (vp->v_type == VDIR) { 323 error = EISDIR; 324 goto out; 325 } else if ((error = nfsrv_access(vp, VWRITE, cred, rdonly, 326 procp, 0)) != 0) 327 goto out; 328 } 329 error = VOP_SETATTR(vp, &va, cred, procp); 330 postat_ret = VOP_GETATTR(vp, &va, cred, procp); 331 if (!error) 332 error = postat_ret; 333 out: 334 vput(vp); 335 vn_finished_write(mp, 0); 336 nfsm_reply(NFSX_WCCORFATTR(v3)); 337 if (v3) { 338 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 339 return (0); 340 } else { 341 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 342 nfsm_srvfillattr(&va, fp); 343 } 344 nfsm_srvdone; 345 } 346 347 /* 348 * nfs lookup rpc 349 */ 350 int 351 nfsrv_lookup(nfsd, slp, procp, mrq) 352 struct nfsrv_descript *nfsd; 353 struct nfssvc_sock *slp; 354 struct proc *procp; 355 struct mbuf **mrq; 356 { 357 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 358 struct mbuf *nam = nfsd->nd_nam; 359 caddr_t dpos = nfsd->nd_dpos; 360 struct ucred *cred = &nfsd->nd_cr; 361 struct nfs_fattr *fp; 362 struct nameidata nd, ind, *ndp = &nd; 363 struct vnode *vp, *dirp; 364 nfsfh_t nfh; 365 fhandle_t *fhp; 366 caddr_t cp; 367 u_int32_t *tl; 368 int32_t t1; 369 caddr_t bpos; 370 int error = 0, cache, dirattr_ret = 1; 371 uint32_t len; 372 int v3 = (nfsd->nd_flag & ND_NFSV3), pubflag; 373 char *cp2; 374 struct mbuf *mb, *mreq; 375 struct vattr va, dirattr; 376 u_quad_t frev; 377 378 fhp = &nfh.fh_generic; 379 nfsm_srvmtofh(fhp); 380 nfsm_srvnamesiz(len); 381 382 pubflag = nfs_ispublicfh(fhp); 383 384 nd.ni_cnd.cn_cred = cred; 385 nd.ni_cnd.cn_nameiop = LOOKUP; 386 nd.ni_cnd.cn_flags = LOCKLEAF | SAVESTART; 387 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 388 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), pubflag); 389 390 if (!error && pubflag) { 391 if (nd.ni_vp->v_type == VDIR && nfs_pub.np_index != NULL) { 392 /* 393 * Setup call to lookup() to see if we can find 394 * the index file. Arguably, this doesn't belong 395 * in a kernel.. Ugh. 396 */ 397 ind = nd; 398 VOP_UNLOCK(nd.ni_vp, 0); 399 ind.ni_pathlen = strlen(nfs_pub.np_index); 400 ind.ni_cnd.cn_nameptr = ind.ni_cnd.cn_pnbuf = 401 nfs_pub.np_index; 402 ind.ni_startdir = nd.ni_vp; 403 VREF(ind.ni_startdir); 404 error = lookup(&ind); 405 if (!error) { 406 /* 407 * Found an index file. Get rid of 408 * the old references. 409 */ 410 if (dirp) 411 vrele(dirp); 412 dirp = nd.ni_vp; 413 vrele(nd.ni_startdir); 414 ndp = &ind; 415 } else 416 error = 0; 417 } 418 /* 419 * If the public filehandle was used, check that this lookup 420 * didn't result in a filehandle outside the publicly exported 421 * filesystem. 422 */ 423 424 if (!error && ndp->ni_vp->v_mount != nfs_pub.np_mount) { 425 vput(nd.ni_vp); 426 error = EPERM; 427 } 428 } 429 430 if (dirp) { 431 if (v3) 432 dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred, 433 procp); 434 vrele(dirp); 435 } 436 437 if (error) { 438 nfsm_reply(NFSX_POSTOPATTR(v3)); 439 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 440 return (0); 441 } 442 443 nqsrv_getl(ndp->ni_startdir, ND_READ); 444 vrele(ndp->ni_startdir); 445 PNBUF_PUT(nd.ni_cnd.cn_pnbuf); 446 vp = ndp->ni_vp; 447 memset((caddr_t)fhp, 0, sizeof(nfh)); 448 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 449 error = VFS_VPTOFH(vp, &fhp->fh_fid); 450 if (!error) 451 error = VOP_GETATTR(vp, &va, cred, procp); 452 vput(vp); 453 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPORFATTR(v3) + NFSX_POSTOPATTR(v3)); 454 if (error) { 455 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 456 return (0); 457 } 458 nfsm_srvfhtom(fhp, v3); 459 if (v3) { 460 nfsm_srvpostop_attr(0, &va); 461 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 462 } else { 463 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 464 nfsm_srvfillattr(&va, fp); 465 } 466 nfsm_srvdone; 467 } 468 469 /* 470 * nfs readlink service 471 */ 472 int 473 nfsrv_readlink(nfsd, slp, procp, mrq) 474 struct nfsrv_descript *nfsd; 475 struct nfssvc_sock *slp; 476 struct proc *procp; 477 struct mbuf **mrq; 478 { 479 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 480 struct mbuf *nam = nfsd->nd_nam; 481 caddr_t dpos = nfsd->nd_dpos; 482 struct ucred *cred = &nfsd->nd_cr; 483 struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN]; 484 struct iovec *ivp = iv; 485 struct mbuf *mp; 486 u_int32_t *tl; 487 int32_t t1; 488 caddr_t bpos; 489 int error = 0, rdonly, cache, i, padlen, getret; 490 uint32_t len; 491 int v3 = (nfsd->nd_flag & ND_NFSV3); 492 char *cp2; 493 struct mbuf *mb, *mp2 = NULL, *mp3 = NULL, *mreq; 494 struct vnode *vp; 495 struct vattr attr; 496 nfsfh_t nfh; 497 fhandle_t *fhp; 498 struct uio io, *uiop = &io; 499 u_quad_t frev; 500 501 fhp = &nfh.fh_generic; 502 nfsm_srvmtofh(fhp); 503 len = 0; 504 i = 0; 505 while (len < NFS_MAXPATHLEN) { 506 mp = m_get(M_WAIT, MT_DATA); 507 MCLAIM(mp, &nfs_mowner); 508 m_clget(mp, M_WAIT); 509 mp->m_len = NFSMSIZ(mp); 510 if (len == 0) 511 mp3 = mp2 = mp; 512 else { 513 mp2->m_next = mp; 514 mp2 = mp; 515 } 516 if ((len+mp->m_len) > NFS_MAXPATHLEN) { 517 mp->m_len = NFS_MAXPATHLEN-len; 518 len = NFS_MAXPATHLEN; 519 } else 520 len += mp->m_len; 521 ivp->iov_base = mtod(mp, caddr_t); 522 ivp->iov_len = mp->m_len; 523 i++; 524 ivp++; 525 } 526 uiop->uio_iov = iv; 527 uiop->uio_iovcnt = i; 528 uiop->uio_offset = 0; 529 uiop->uio_resid = len; 530 uiop->uio_rw = UIO_READ; 531 uiop->uio_segflg = UIO_SYSSPACE; 532 uiop->uio_procp = NULL; 533 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 534 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 535 if (error) { 536 m_freem(mp3); 537 nfsm_reply(2 * NFSX_UNSIGNED); 538 nfsm_srvpostop_attr(1, (struct vattr *)0); 539 return (0); 540 } 541 if (vp->v_type != VLNK) { 542 if (v3) 543 error = EINVAL; 544 else 545 error = ENXIO; 546 goto out; 547 } 548 nqsrv_getl(vp, ND_READ); 549 error = VOP_READLINK(vp, uiop, cred); 550 out: 551 getret = VOP_GETATTR(vp, &attr, cred, procp); 552 vput(vp); 553 if (error) 554 m_freem(mp3); 555 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED); 556 if (v3) { 557 nfsm_srvpostop_attr(getret, &attr); 558 if (error) 559 return (0); 560 } 561 len -= uiop->uio_resid; 562 padlen = nfsm_padlen(len); 563 if (uiop->uio_resid || padlen) 564 nfs_zeropad(mp3, uiop->uio_resid, padlen); 565 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 566 *tl = txdr_unsigned(len); 567 mb->m_next = mp3; 568 nfsm_srvdone; 569 } 570 571 /* 572 * nfs read service 573 */ 574 int 575 nfsrv_read(nfsd, slp, procp, mrq) 576 struct nfsrv_descript *nfsd; 577 struct nfssvc_sock *slp; 578 struct proc *procp; 579 struct mbuf **mrq; 580 { 581 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 582 struct mbuf *nam = nfsd->nd_nam; 583 caddr_t dpos = nfsd->nd_dpos; 584 struct ucred *cred = &nfsd->nd_cr; 585 struct mbuf *m; 586 struct nfs_fattr *fp; 587 u_int32_t *tl; 588 int32_t t1; 589 int i; 590 caddr_t bpos; 591 int error = 0, rdonly, cache, getret; 592 int v3 = (nfsd->nd_flag & ND_NFSV3); 593 uint32_t reqlen, len, cnt, left; 594 int padlen; 595 char *cp2; 596 struct mbuf *mb, *mreq; 597 struct vnode *vp; 598 nfsfh_t nfh; 599 fhandle_t *fhp; 600 struct uio io, *uiop = &io; 601 struct vattr va; 602 off_t off; 603 u_quad_t frev; 604 605 fhp = &nfh.fh_generic; 606 nfsm_srvmtofh(fhp); 607 if (v3) { 608 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 609 off = fxdr_hyper(tl); 610 } else { 611 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 612 off = (off_t)fxdr_unsigned(u_int32_t, *tl); 613 } 614 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 615 reqlen = fxdr_unsigned(uint32_t, *tl); 616 reqlen = MIN(reqlen, NFS_SRVMAXDATA(nfsd)); 617 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 618 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 619 if (error) { 620 nfsm_reply(2 * NFSX_UNSIGNED); 621 nfsm_srvpostop_attr(1, (struct vattr *)0); 622 return (0); 623 } 624 if (vp->v_type != VREG) { 625 if (v3) 626 error = EINVAL; 627 else 628 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 629 } 630 if (!error) { 631 nqsrv_getl(vp, ND_READ); 632 if ((error = nfsrv_access(vp, VREAD, cred, rdonly, procp, 1)) != 0) 633 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 1); 634 } 635 getret = VOP_GETATTR(vp, &va, cred, procp); 636 if (!error) 637 error = getret; 638 if (error) { 639 vput(vp); 640 nfsm_reply(NFSX_POSTOPATTR(v3)); 641 nfsm_srvpostop_attr(getret, &va); 642 return (0); 643 } 644 if (off >= va.va_size) 645 cnt = 0; 646 else if ((off + reqlen) > va.va_size) 647 cnt = va.va_size - off; 648 else 649 cnt = reqlen; 650 nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt)); 651 if (v3) { 652 nfsm_build(tl, u_int32_t *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED); 653 *tl++ = nfs_true; 654 fp = (struct nfs_fattr *)tl; 655 tl += (NFSX_V3FATTR / sizeof (u_int32_t)); 656 } else { 657 nfsm_build(tl, u_int32_t *, NFSX_V2FATTR + NFSX_UNSIGNED); 658 fp = (struct nfs_fattr *)tl; 659 tl += (NFSX_V2FATTR / sizeof (u_int32_t)); 660 } 661 len = left = cnt; 662 if (cnt > 0) { 663 if (nfsd_use_loan) { 664 struct vm_page **pgpp; 665 voff_t pgoff = trunc_page(off); 666 int npages; 667 vaddr_t lva; 668 669 npages = (round_page(off + cnt) - pgoff) >> PAGE_SHIFT; 670 KASSERT(npages <= M_EXT_MAXPAGES); /* XXX */ 671 672 /* allocate kva for mbuf data */ 673 lva = sokvaalloc(npages << PAGE_SHIFT, slp->ns_so); 674 if (lva == 0) { 675 /* fall back to VOP_READ */ 676 goto loan_fail; 677 } 678 679 /* allocate mbuf */ 680 m = m_get(M_WAIT, MT_DATA); 681 MCLAIM(m, &nfs_mowner); 682 pgpp = m->m_ext.ext_pgs; 683 684 /* loan pages */ 685 error = uvm_loanuobjpages(&vp->v_uobj, pgoff, npages, 686 pgpp); 687 if (error) { 688 sokvafree(lva, npages << PAGE_SHIFT); 689 m_free(m); 690 if (error == EBUSY) 691 goto loan_fail; 692 goto read_error; 693 } 694 695 /* associate kva to mbuf */ 696 MEXTADD(m, (void *)(lva + ((vaddr_t)off & PAGE_MASK)), 697 cnt, M_MBUF, soloanfree, slp->ns_so); 698 m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP; 699 m->m_len = cnt; 700 701 /* map pages */ 702 for (i = 0; i < npages; i++) { 703 pmap_kenter_pa(lva, VM_PAGE_TO_PHYS(pgpp[i]), 704 VM_PROT_READ); 705 lva += PAGE_SIZE; 706 } 707 708 pmap_update(pmap_kernel()); 709 710 mb->m_next = m; 711 mb = m; 712 error = 0; 713 uiop->uio_resid = 0; 714 } else { 715 struct iovec *iv; 716 struct iovec *iv2; 717 struct mbuf *m2; 718 int siz; 719 loan_fail: 720 /* 721 * Generate the mbuf list with the uio_iov ref. to it. 722 */ 723 i = 0; 724 m = m2 = mb; 725 while (left > 0) { 726 siz = min(M_TRAILINGSPACE(m), left); 727 if (siz > 0) { 728 left -= siz; 729 i++; 730 } 731 if (left > 0) { 732 m = m_get(M_WAIT, MT_DATA); 733 MCLAIM(m, &nfs_mowner); 734 m_clget(m, M_WAIT); 735 m->m_len = 0; 736 m2->m_next = m; 737 m2 = m; 738 } 739 } 740 iv = malloc(i * sizeof(struct iovec), M_TEMP, M_WAITOK); 741 uiop->uio_iov = iv2 = iv; 742 m = mb; 743 left = cnt; 744 i = 0; 745 while (left > 0) { 746 if (m == NULL) 747 panic("nfsrv_read iov"); 748 siz = min(M_TRAILINGSPACE(m), left); 749 if (siz > 0) { 750 iv->iov_base = mtod(m, caddr_t) + 751 m->m_len; 752 iv->iov_len = siz; 753 m->m_len += siz; 754 left -= siz; 755 iv++; 756 i++; 757 } 758 m = m->m_next; 759 } 760 uiop->uio_iovcnt = i; 761 uiop->uio_offset = off; 762 uiop->uio_resid = cnt; 763 uiop->uio_rw = UIO_READ; 764 uiop->uio_segflg = UIO_SYSSPACE; 765 error = VOP_READ(vp, uiop, IO_NODELOCKED, cred); 766 free((caddr_t)iv2, M_TEMP); 767 } 768 read_error: 769 if (error || (getret = VOP_GETATTR(vp, &va, cred, procp)) != 0){ 770 if (!error) 771 error = getret; 772 m_freem(mreq); 773 vput(vp); 774 nfsm_reply(NFSX_POSTOPATTR(v3)); 775 nfsm_srvpostop_attr(getret, &va); 776 return (0); 777 } 778 } else { 779 uiop->uio_resid = 0; 780 } 781 vput(vp); 782 nfsm_srvfillattr(&va, fp); 783 len -= uiop->uio_resid; 784 padlen = nfsm_padlen(len); 785 if (uiop->uio_resid || padlen) 786 nfs_zeropad(mb, uiop->uio_resid, padlen); 787 if (v3) { 788 /* count */ 789 *tl++ = txdr_unsigned(len); 790 /* eof */ 791 if (off + len >= va.va_size) 792 *tl++ = nfs_true; 793 else 794 *tl++ = nfs_false; 795 } 796 *tl = txdr_unsigned(len); 797 nfsm_srvdone; 798 } 799 800 /* 801 * nfs write service 802 */ 803 int 804 nfsrv_write(nfsd, slp, procp, mrq) 805 struct nfsrv_descript *nfsd; 806 struct nfssvc_sock *slp; 807 struct proc *procp; 808 struct mbuf **mrq; 809 { 810 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 811 struct mbuf *nam = nfsd->nd_nam; 812 caddr_t dpos = nfsd->nd_dpos; 813 struct ucred *cred = &nfsd->nd_cr; 814 struct iovec *ivp; 815 int i, cnt; 816 struct mbuf *mp; 817 struct nfs_fattr *fp; 818 struct iovec *iv; 819 struct vattr va, forat; 820 u_int32_t *tl; 821 int32_t t1; 822 caddr_t bpos; 823 int error = 0, rdonly, cache, len, forat_ret = 1; 824 int ioflags, aftat_ret = 1, retlen, zeroing, adjust; 825 int stable = NFSV3WRITE_FILESYNC; 826 int v3 = (nfsd->nd_flag & ND_NFSV3); 827 char *cp2; 828 struct mbuf *mb, *mreq; 829 struct vnode *vp; 830 nfsfh_t nfh; 831 fhandle_t *fhp; 832 struct uio io, *uiop = &io; 833 off_t off; 834 u_quad_t frev; 835 struct mount *mntp = NULL; 836 837 if (mrep == NULL) { 838 *mrq = NULL; 839 return (0); 840 } 841 fhp = &nfh.fh_generic; 842 nfsm_srvmtofh(fhp); 843 if ((mntp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 844 return (ESTALE); 845 vn_start_write(NULL, &mntp, V_WAIT); 846 if (v3) { 847 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 848 off = fxdr_hyper(tl); 849 tl += 3; 850 stable = fxdr_unsigned(int, *tl++); 851 } else { 852 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 853 off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 854 tl += 2; 855 } 856 retlen = len = fxdr_unsigned(int32_t, *tl); 857 cnt = i = 0; 858 859 /* 860 * For NFS Version 2, it is not obvious what a write of zero length 861 * should do, but I might as well be consistent with Version 3, 862 * which is to return ok so long as there are no permission problems. 863 */ 864 if (len > 0) { 865 zeroing = 1; 866 mp = mrep; 867 while (mp) { 868 if (mp == md) { 869 zeroing = 0; 870 adjust = dpos - mtod(mp, caddr_t); 871 mp->m_len -= adjust; 872 if (mp->m_len > 0 && adjust > 0) 873 NFSMADV(mp, adjust); 874 } 875 if (zeroing) 876 mp->m_len = 0; 877 else if (mp->m_len > 0) { 878 i += mp->m_len; 879 if (i > len) { 880 mp->m_len -= (i - len); 881 zeroing = 1; 882 } 883 if (mp->m_len > 0) 884 cnt++; 885 } 886 mp = mp->m_next; 887 } 888 } 889 if (len > NFS_MAXDATA || len < 0 || i < len) { 890 error = EIO; 891 nfsm_reply(2 * NFSX_UNSIGNED); 892 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 893 vn_finished_write(mntp, 0); 894 return (0); 895 } 896 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 897 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 898 if (error) { 899 nfsm_reply(2 * NFSX_UNSIGNED); 900 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 901 vn_finished_write(mntp, 0); 902 return (0); 903 } 904 if (v3) 905 forat_ret = VOP_GETATTR(vp, &forat, cred, procp); 906 if (vp->v_type != VREG) { 907 if (v3) 908 error = EINVAL; 909 else 910 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 911 } 912 if (!error) { 913 nqsrv_getl(vp, ND_WRITE); 914 error = nfsrv_access(vp, VWRITE, cred, rdonly, procp, 1); 915 } 916 if (error) { 917 vput(vp); 918 nfsm_reply(NFSX_WCCDATA(v3)); 919 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 920 vn_finished_write(mntp, 0); 921 return (0); 922 } 923 924 if (len > 0) { 925 ivp = malloc(cnt * sizeof (struct iovec), M_TEMP, M_WAITOK); 926 uiop->uio_iov = iv = ivp; 927 uiop->uio_iovcnt = cnt; 928 mp = mrep; 929 while (mp) { 930 if (mp->m_len > 0) { 931 ivp->iov_base = mtod(mp, caddr_t); 932 ivp->iov_len = mp->m_len; 933 ivp++; 934 } 935 mp = mp->m_next; 936 } 937 938 /* 939 * XXX 940 * The IO_METASYNC flag indicates that all metadata (and not 941 * just enough to ensure data integrity) must be written to 942 * stable storage synchronously. 943 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.) 944 */ 945 if (stable == NFSV3WRITE_UNSTABLE) 946 ioflags = IO_NODELOCKED; 947 else if (stable == NFSV3WRITE_DATASYNC) 948 ioflags = (IO_SYNC | IO_NODELOCKED); 949 else 950 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 951 uiop->uio_resid = len; 952 uiop->uio_rw = UIO_WRITE; 953 uiop->uio_segflg = UIO_SYSSPACE; 954 uiop->uio_procp = NULL; 955 uiop->uio_offset = off; 956 error = VOP_WRITE(vp, uiop, ioflags, cred); 957 nfsstats.srvvop_writes++; 958 free(iv, M_TEMP); 959 } 960 aftat_ret = VOP_GETATTR(vp, &va, cred, procp); 961 vput(vp); 962 vn_finished_write(mntp, 0); 963 if (!error) 964 error = aftat_ret; 965 nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) + 966 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3)); 967 if (v3) { 968 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 969 if (error) 970 return (0); 971 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 972 *tl++ = txdr_unsigned(retlen); 973 if (stable == NFSV3WRITE_UNSTABLE) 974 *tl++ = txdr_unsigned(stable); 975 else 976 *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC); 977 /* 978 * Actually, there is no need to txdr these fields, 979 * but it may make the values more human readable, 980 * for debugging purposes. 981 */ 982 *tl++ = txdr_unsigned(boottime.tv_sec); 983 *tl = txdr_unsigned(boottime.tv_usec); 984 } else { 985 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 986 nfsm_srvfillattr(&va, fp); 987 } 988 nfsm_srvdone; 989 } 990 991 /* 992 * NFS write service with write gathering support. Called when 993 * nfsrvw_procrastinate > 0. 994 * See: Chet Juszczak, "Improving the Write Performance of an NFS Server", 995 * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco, 996 * Jan. 1994. 997 */ 998 int 999 nfsrv_writegather(ndp, slp, procp, mrq) 1000 struct nfsrv_descript **ndp; 1001 struct nfssvc_sock *slp; 1002 struct proc *procp; 1003 struct mbuf **mrq; 1004 { 1005 struct iovec *ivp; 1006 struct mbuf *mp; 1007 struct nfsrv_descript *wp, *nfsd, *owp, *swp; 1008 struct nfs_fattr *fp; 1009 int i = 0; 1010 struct iovec *iov; 1011 struct nfsrvw_delayhash *wpp; 1012 struct ucred *cred; 1013 struct vattr va, forat; 1014 u_int32_t *tl; 1015 int32_t t1; 1016 caddr_t bpos, dpos; 1017 int error = 0, rdonly, cache, len = 0, forat_ret = 1; 1018 int ioflags, aftat_ret = 1, s, adjust, v3, zeroing; 1019 char *cp2; 1020 struct mbuf *mb, *mreq, *mrep, *md; 1021 struct vnode *vp; 1022 struct uio io, *uiop = &io; 1023 u_quad_t frev, cur_usec; 1024 struct mount *mntp = NULL; 1025 1026 *mrq = NULL; 1027 if (*ndp) { 1028 nfsd = *ndp; 1029 *ndp = NULL; 1030 mrep = nfsd->nd_mrep; 1031 md = nfsd->nd_md; 1032 dpos = nfsd->nd_dpos; 1033 cred = &nfsd->nd_cr; 1034 v3 = (nfsd->nd_flag & ND_NFSV3); 1035 LIST_INIT(&nfsd->nd_coalesce); 1036 nfsd->nd_mreq = NULL; 1037 nfsd->nd_stable = NFSV3WRITE_FILESYNC; 1038 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec; 1039 nfsd->nd_time = cur_usec + nfsrvw_procrastinate; 1040 1041 /* 1042 * Now, get the write header.. 1043 */ 1044 nfsm_srvmtofh(&nfsd->nd_fh); 1045 if (v3) { 1046 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 1047 nfsd->nd_off = fxdr_hyper(tl); 1048 tl += 3; 1049 nfsd->nd_stable = fxdr_unsigned(int, *tl++); 1050 } else { 1051 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1052 nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 1053 tl += 2; 1054 } 1055 len = fxdr_unsigned(int32_t, *tl); 1056 nfsd->nd_len = len; 1057 nfsd->nd_eoff = nfsd->nd_off + len; 1058 1059 /* 1060 * Trim the header out of the mbuf list and trim off any trailing 1061 * junk so that the mbuf list has only the write data. 1062 */ 1063 zeroing = 1; 1064 i = 0; 1065 mp = mrep; 1066 while (mp) { 1067 if (mp == md) { 1068 zeroing = 0; 1069 adjust = dpos - mtod(mp, caddr_t); 1070 mp->m_len -= adjust; 1071 if (mp->m_len > 0 && adjust > 0) 1072 NFSMADV(mp, adjust); 1073 } 1074 if (zeroing) 1075 mp->m_len = 0; 1076 else { 1077 i += mp->m_len; 1078 if (i > len) { 1079 mp->m_len -= (i - len); 1080 zeroing = 1; 1081 } 1082 } 1083 mp = mp->m_next; 1084 } 1085 if (len > NFS_MAXDATA || len < 0 || i < len) { 1086 nfsmout: 1087 m_freem(mrep); 1088 error = EIO; 1089 nfsm_writereply(2 * NFSX_UNSIGNED, v3); 1090 if (v3) 1091 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1092 nfsd->nd_mreq = mreq; 1093 nfsd->nd_mrep = NULL; 1094 nfsd->nd_time = 0; 1095 } 1096 1097 /* 1098 * Add this entry to the hash and time queues. 1099 */ 1100 s = splsoftclock(); 1101 owp = NULL; 1102 wp = LIST_FIRST(&slp->ns_tq); 1103 while (wp && wp->nd_time < nfsd->nd_time) { 1104 owp = wp; 1105 wp = LIST_NEXT(wp, nd_tq); 1106 } 1107 if (owp) { 1108 LIST_INSERT_AFTER(owp, nfsd, nd_tq); 1109 } else { 1110 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1111 } 1112 if (nfsd->nd_mrep) { 1113 wpp = NWDELAYHASH(slp, nfsd->nd_fh.fh_fid.fid_data); 1114 owp = NULL; 1115 wp = LIST_FIRST(wpp); 1116 while (wp && 1117 memcmp((caddr_t)&nfsd->nd_fh, (caddr_t)&wp->nd_fh, NFSX_V3FH)) { 1118 owp = wp; 1119 wp = LIST_NEXT(wp, nd_hash); 1120 } 1121 while (wp && wp->nd_off < nfsd->nd_off && 1122 !memcmp((caddr_t)&nfsd->nd_fh, (caddr_t)&wp->nd_fh, NFSX_V3FH)) { 1123 owp = wp; 1124 wp = LIST_NEXT(wp, nd_hash); 1125 } 1126 if (owp) { 1127 LIST_INSERT_AFTER(owp, nfsd, nd_hash); 1128 1129 /* 1130 * Search the hash list for overlapping entries and 1131 * coalesce. 1132 */ 1133 for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) { 1134 wp = LIST_NEXT(nfsd, nd_hash); 1135 if (NFSW_SAMECRED(owp, nfsd)) 1136 nfsrvw_coalesce(owp, nfsd); 1137 } 1138 } else { 1139 LIST_INSERT_HEAD(wpp, nfsd, nd_hash); 1140 } 1141 } 1142 splx(s); 1143 } 1144 1145 /* 1146 * Now, do VOP_WRITE()s for any one(s) that need to be done now 1147 * and generate the associated reply mbuf list(s). 1148 */ 1149 loop1: 1150 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec; 1151 s = splsoftclock(); 1152 for (nfsd = LIST_FIRST(&slp->ns_tq); nfsd; nfsd = owp) { 1153 owp = LIST_NEXT(nfsd, nd_tq); 1154 if (nfsd->nd_time > cur_usec) 1155 break; 1156 if (nfsd->nd_mreq) 1157 continue; 1158 LIST_REMOVE(nfsd, nd_tq); 1159 LIST_REMOVE(nfsd, nd_hash); 1160 splx(s); 1161 mrep = nfsd->nd_mrep; 1162 nfsd->nd_mrep = NULL; 1163 cred = &nfsd->nd_cr; 1164 v3 = (nfsd->nd_flag & ND_NFSV3); 1165 forat_ret = aftat_ret = 1; 1166 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp, 1167 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH), 1168 FALSE); 1169 if (!error) { 1170 if (v3) 1171 forat_ret = VOP_GETATTR(vp, &forat, cred, procp); 1172 if (vp->v_type != VREG) { 1173 if (v3) 1174 error = EINVAL; 1175 else 1176 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 1177 } 1178 } else 1179 vp = NULL; 1180 if (!error) { 1181 nqsrv_getl(vp, ND_WRITE); 1182 error = nfsrv_access(vp, VWRITE, cred, rdonly, procp, 1); 1183 } 1184 1185 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE) 1186 ioflags = IO_NODELOCKED; 1187 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC) 1188 ioflags = (IO_SYNC | IO_NODELOCKED); 1189 else 1190 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 1191 uiop->uio_rw = UIO_WRITE; 1192 uiop->uio_segflg = UIO_SYSSPACE; 1193 uiop->uio_procp = NULL; 1194 uiop->uio_offset = nfsd->nd_off; 1195 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off; 1196 if (uiop->uio_resid > 0) { 1197 mp = mrep; 1198 i = 0; 1199 while (mp) { 1200 if (mp->m_len > 0) 1201 i++; 1202 mp = mp->m_next; 1203 } 1204 uiop->uio_iovcnt = i; 1205 iov = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK); 1206 uiop->uio_iov = ivp = iov; 1207 mp = mrep; 1208 while (mp) { 1209 if (mp->m_len > 0) { 1210 ivp->iov_base = mtod(mp, caddr_t); 1211 ivp->iov_len = mp->m_len; 1212 ivp++; 1213 } 1214 mp = mp->m_next; 1215 } 1216 if (!error) { 1217 if (vn_start_write(vp, &mntp, V_NOWAIT) != 0) { 1218 VOP_UNLOCK(vp, 0); 1219 vn_start_write(NULL, &mntp, V_WAIT); 1220 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1221 } 1222 if (!error) { 1223 error = VOP_WRITE(vp, uiop, ioflags, cred); 1224 nfsstats.srvvop_writes++; 1225 vn_finished_write(mntp, 0); 1226 } 1227 } 1228 free((caddr_t)iov, M_TEMP); 1229 } 1230 m_freem(mrep); 1231 if (vp) { 1232 aftat_ret = VOP_GETATTR(vp, &va, cred, procp); 1233 vput(vp); 1234 } 1235 1236 /* 1237 * Loop around generating replies for all write rpcs that have 1238 * now been completed. 1239 */ 1240 swp = nfsd; 1241 do { 1242 if (error) { 1243 nfsm_writereply(NFSX_WCCDATA(v3), v3); 1244 if (v3) { 1245 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1246 } 1247 } else { 1248 nfsm_writereply(NFSX_PREOPATTR(v3) + 1249 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED + 1250 NFSX_WRITEVERF(v3), v3); 1251 if (v3) { 1252 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1253 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1254 *tl++ = txdr_unsigned(nfsd->nd_len); 1255 *tl++ = txdr_unsigned(swp->nd_stable); 1256 /* 1257 * Actually, there is no need to txdr these fields, 1258 * but it may make the values more human readable, 1259 * for debugging purposes. 1260 */ 1261 *tl++ = txdr_unsigned(boottime.tv_sec); 1262 *tl = txdr_unsigned(boottime.tv_usec); 1263 } else { 1264 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1265 nfsm_srvfillattr(&va, fp); 1266 } 1267 } 1268 nfsd->nd_mreq = mreq; 1269 if (nfsd->nd_mrep) 1270 panic("nfsrv_write: nd_mrep not free"); 1271 1272 /* 1273 * Done. Put it at the head of the timer queue so that 1274 * the final phase can return the reply. 1275 */ 1276 s = splsoftclock(); 1277 if (nfsd != swp) { 1278 nfsd->nd_time = 0; 1279 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1280 } 1281 nfsd = LIST_FIRST(&swp->nd_coalesce); 1282 if (nfsd) { 1283 LIST_REMOVE(nfsd, nd_tq); 1284 } 1285 splx(s); 1286 } while (nfsd); 1287 s = splsoftclock(); 1288 swp->nd_time = 0; 1289 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq); 1290 splx(s); 1291 goto loop1; 1292 } 1293 splx(s); 1294 1295 /* 1296 * Search for a reply to return. 1297 */ 1298 s = splsoftclock(); 1299 LIST_FOREACH(nfsd, &slp->ns_tq, nd_tq) { 1300 if (nfsd->nd_mreq) { 1301 LIST_REMOVE(nfsd, nd_tq); 1302 *mrq = nfsd->nd_mreq; 1303 *ndp = nfsd; 1304 break; 1305 } 1306 } 1307 splx(s); 1308 return (0); 1309 } 1310 1311 /* 1312 * Coalesce the write request nfsd into owp. To do this we must: 1313 * - remove nfsd from the queues 1314 * - merge nfsd->nd_mrep into owp->nd_mrep 1315 * - update the nd_eoff and nd_stable for owp 1316 * - put nfsd on owp's nd_coalesce list 1317 * NB: Must be called at splsoftclock(). 1318 */ 1319 void 1320 nfsrvw_coalesce(owp, nfsd) 1321 struct nfsrv_descript *owp; 1322 struct nfsrv_descript *nfsd; 1323 { 1324 int overlap; 1325 struct mbuf *mp; 1326 struct nfsrv_descript *m; 1327 1328 LIST_REMOVE(nfsd, nd_hash); 1329 LIST_REMOVE(nfsd, nd_tq); 1330 if (owp->nd_eoff < nfsd->nd_eoff) { 1331 overlap = owp->nd_eoff - nfsd->nd_off; 1332 if (overlap < 0) 1333 panic("nfsrv_coalesce: bad off"); 1334 if (overlap > 0) 1335 m_adj(nfsd->nd_mrep, overlap); 1336 mp = owp->nd_mrep; 1337 while (mp->m_next) 1338 mp = mp->m_next; 1339 mp->m_next = nfsd->nd_mrep; 1340 owp->nd_eoff = nfsd->nd_eoff; 1341 } else 1342 m_freem(nfsd->nd_mrep); 1343 nfsd->nd_mrep = NULL; 1344 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC) 1345 owp->nd_stable = NFSV3WRITE_FILESYNC; 1346 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC && 1347 owp->nd_stable == NFSV3WRITE_UNSTABLE) 1348 owp->nd_stable = NFSV3WRITE_DATASYNC; 1349 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq); 1350 /* 1351 * nfsd might hold coalesce elements! Move them to owp. 1352 * Otherwise, requests may be lost and clients will be stuck. 1353 */ 1354 while ((m = LIST_FIRST(&nfsd->nd_coalesce)) != NULL) { 1355 LIST_REMOVE(m, nd_tq); 1356 LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq); 1357 } 1358 } 1359 1360 /* 1361 * nfs create service 1362 * now does a truncate to 0 length via. setattr if it already exists 1363 */ 1364 int 1365 nfsrv_create(nfsd, slp, procp, mrq) 1366 struct nfsrv_descript *nfsd; 1367 struct nfssvc_sock *slp; 1368 struct proc *procp; 1369 struct mbuf **mrq; 1370 { 1371 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1372 struct mbuf *nam = nfsd->nd_nam; 1373 caddr_t dpos = nfsd->nd_dpos; 1374 struct ucred *cred = &nfsd->nd_cr; 1375 struct nfs_fattr *fp; 1376 struct vattr va, dirfor, diraft; 1377 struct nfsv2_sattr *sp; 1378 u_int32_t *tl; 1379 struct nameidata nd; 1380 caddr_t cp; 1381 int32_t t1; 1382 caddr_t bpos; 1383 int error = 0, cache, len, tsize, dirfor_ret = 1, diraft_ret = 1; 1384 int rdev = 0; 1385 int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0; 1386 char *cp2; 1387 struct mbuf *mb, *mreq; 1388 struct vnode *vp = NULL, *dirp = NULL; 1389 nfsfh_t nfh; 1390 fhandle_t *fhp; 1391 u_quad_t frev, tempsize; 1392 u_char cverf[NFSX_V3CREATEVERF]; 1393 struct mount *mp = NULL; 1394 1395 nd.ni_cnd.cn_nameiop = 0; 1396 fhp = &nfh.fh_generic; 1397 nfsm_srvmtofh(fhp); 1398 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1399 return (ESTALE); 1400 vn_start_write(NULL, &mp, V_WAIT); 1401 nfsm_srvnamesiz(len); 1402 nd.ni_cnd.cn_cred = cred; 1403 nd.ni_cnd.cn_nameiop = CREATE; 1404 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1405 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1406 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1407 if (dirp) { 1408 if (v3) 1409 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, 1410 procp); 1411 else { 1412 vrele(dirp); 1413 dirp = (struct vnode *)0; 1414 } 1415 } 1416 if (error) { 1417 nfsm_reply(NFSX_WCCDATA(v3)); 1418 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1419 if (dirp) 1420 vrele(dirp); 1421 vn_finished_write(mp, 0); 1422 return (0); 1423 } 1424 VATTR_NULL(&va); 1425 if (v3) { 1426 va.va_mode = 0; 1427 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1428 how = fxdr_unsigned(int, *tl); 1429 switch (how) { 1430 case NFSV3CREATE_GUARDED: 1431 if (nd.ni_vp) { 1432 error = EEXIST; 1433 break; 1434 } 1435 case NFSV3CREATE_UNCHECKED: 1436 nfsm_srvsattr(&va); 1437 break; 1438 case NFSV3CREATE_EXCLUSIVE: 1439 nfsm_dissect(cp, caddr_t, NFSX_V3CREATEVERF); 1440 memcpy(cverf, cp, NFSX_V3CREATEVERF); 1441 exclusive_flag = 1; 1442 break; 1443 }; 1444 va.va_type = VREG; 1445 } else { 1446 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 1447 va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode)); 1448 if (va.va_type == VNON) 1449 va.va_type = VREG; 1450 va.va_mode = nfstov_mode(sp->sa_mode); 1451 switch (va.va_type) { 1452 case VREG: 1453 tsize = fxdr_unsigned(int32_t, sp->sa_size); 1454 if (tsize != -1) 1455 va.va_size = (u_quad_t)tsize; 1456 break; 1457 case VCHR: 1458 case VBLK: 1459 case VFIFO: 1460 rdev = fxdr_unsigned(int32_t, sp->sa_size); 1461 break; 1462 default: 1463 break; 1464 }; 1465 } 1466 1467 /* 1468 * Iff doesn't exist, create it 1469 * otherwise just truncate to 0 length 1470 * should I set the mode too ?? 1471 */ 1472 if (nd.ni_vp == NULL) { 1473 if (va.va_type == VREG || va.va_type == VSOCK) { 1474 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1475 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1476 if (!error) { 1477 if (exclusive_flag) { 1478 exclusive_flag = 0; 1479 VATTR_NULL(&va); 1480 /* 1481 * XXX 1482 * assuming NFSX_V3CREATEVERF 1483 * == sizeof(nfstime3) 1484 */ 1485 fxdr_nfsv3time(cverf, &va.va_atime); 1486 error = VOP_SETATTR(nd.ni_vp, &va, cred, 1487 procp); 1488 } 1489 } 1490 } else if (va.va_type == VCHR || va.va_type == VBLK || 1491 va.va_type == VFIFO) { 1492 if (va.va_type == VCHR && rdev == 0xffffffff) 1493 va.va_type = VFIFO; 1494 if (va.va_type != VFIFO && 1495 (error = suser(cred, (u_short *)0))) { 1496 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1497 vput(nd.ni_dvp); 1498 nfsm_reply(0); 1499 vn_finished_write(mp, 0); 1500 return (error); 1501 } else 1502 va.va_rdev = (dev_t)rdev; 1503 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1504 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, 1505 &va); 1506 if (error) { 1507 nfsm_reply(0); 1508 } 1509 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1510 vrele(nd.ni_dvp); 1511 vput(nd.ni_vp); 1512 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1513 error = EINVAL; 1514 nfsm_reply(0); 1515 } 1516 } else { 1517 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1518 vput(nd.ni_dvp); 1519 error = ENXIO; 1520 } 1521 vp = nd.ni_vp; 1522 } else { 1523 vp = nd.ni_vp; 1524 if (nd.ni_dvp == vp) 1525 vrele(nd.ni_dvp); 1526 else 1527 vput(nd.ni_dvp); 1528 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1529 if (!error && va.va_size != -1) { 1530 error = nfsrv_access(vp, VWRITE, cred, 1531 (nd.ni_cnd.cn_flags & RDONLY), procp, 0); 1532 if (!error) { 1533 nqsrv_getl(vp, ND_WRITE); 1534 tempsize = va.va_size; 1535 VATTR_NULL(&va); 1536 va.va_size = tempsize; 1537 error = VOP_SETATTR(vp, &va, cred, 1538 procp); 1539 } 1540 } 1541 if (error) 1542 vput(vp); 1543 } 1544 if (!error) { 1545 memset((caddr_t)fhp, 0, sizeof(nfh)); 1546 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1547 error = VFS_VPTOFH(vp, &fhp->fh_fid); 1548 if (!error) 1549 error = VOP_GETATTR(vp, &va, cred, procp); 1550 vput(vp); 1551 } 1552 if (v3) { 1553 if (exclusive_flag && !error) { 1554 /* 1555 * XXX assuming NFSX_V3CREATEVERF == sizeof(nfstime3) 1556 */ 1557 char oldverf[NFSX_V3CREATEVERF]; 1558 1559 txdr_nfsv3time(&va.va_atime, oldverf); 1560 if (memcmp(cverf, oldverf, NFSX_V3CREATEVERF)) 1561 error = EEXIST; 1562 } 1563 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); 1564 vrele(dirp); 1565 } 1566 nfsm_reply(NFSX_SRVFH(v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3)); 1567 if (v3) { 1568 if (!error) { 1569 nfsm_srvpostop_fh(fhp); 1570 nfsm_srvpostop_attr(0, &va); 1571 } 1572 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1573 } else { 1574 nfsm_srvfhtom(fhp, v3); 1575 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1576 nfsm_srvfillattr(&va, fp); 1577 } 1578 vn_finished_write(mp, 0); 1579 return (0); 1580 nfsmout: 1581 if (dirp) 1582 vrele(dirp); 1583 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1584 if (nd.ni_dvp == nd.ni_vp) 1585 vrele(nd.ni_dvp); 1586 else 1587 vput(nd.ni_dvp); 1588 if (nd.ni_vp) 1589 vput(nd.ni_vp); 1590 vn_finished_write(mp, 0); 1591 return (error); 1592 } 1593 1594 /* 1595 * nfs v3 mknod service 1596 */ 1597 int 1598 nfsrv_mknod(nfsd, slp, procp, mrq) 1599 struct nfsrv_descript *nfsd; 1600 struct nfssvc_sock *slp; 1601 struct proc *procp; 1602 struct mbuf **mrq; 1603 { 1604 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1605 struct mbuf *nam = nfsd->nd_nam; 1606 caddr_t dpos = nfsd->nd_dpos; 1607 struct ucred *cred = &nfsd->nd_cr; 1608 struct vattr va, dirfor, diraft; 1609 u_int32_t *tl; 1610 struct nameidata nd; 1611 int32_t t1; 1612 caddr_t bpos; 1613 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 1614 u_int32_t major, minor; 1615 enum vtype vtyp; 1616 char *cp2; 1617 struct mbuf *mb, *mreq; 1618 struct vnode *vp, *dirp = (struct vnode *)0; 1619 nfsfh_t nfh; 1620 fhandle_t *fhp; 1621 u_quad_t frev; 1622 struct mount *mp = NULL; 1623 1624 nd.ni_cnd.cn_nameiop = 0; 1625 fhp = &nfh.fh_generic; 1626 nfsm_srvmtofh(fhp); 1627 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1628 return (ESTALE); 1629 vn_start_write(NULL, &mp, V_WAIT); 1630 nfsm_srvnamesiz(len); 1631 nd.ni_cnd.cn_cred = cred; 1632 nd.ni_cnd.cn_nameiop = CREATE; 1633 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1634 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1635 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1636 if (dirp) 1637 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, procp); 1638 if (error) { 1639 nfsm_reply(NFSX_WCCDATA(1)); 1640 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1641 if (dirp) 1642 vrele(dirp); 1643 vn_finished_write(mp, 0); 1644 return (0); 1645 } 1646 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1647 vtyp = nfsv3tov_type(*tl); 1648 if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) { 1649 error = NFSERR_BADTYPE; 1650 goto abort; 1651 } 1652 VATTR_NULL(&va); 1653 va.va_mode = 0; 1654 nfsm_srvsattr(&va); 1655 if (vtyp == VCHR || vtyp == VBLK) { 1656 dev_t rdev; 1657 1658 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1659 major = fxdr_unsigned(u_int32_t, *tl++); 1660 minor = fxdr_unsigned(u_int32_t, *tl); 1661 rdev = makedev(major, minor); 1662 if (major(rdev) != major || minor(rdev) != minor) { 1663 error = EINVAL; 1664 goto abort; 1665 } 1666 va.va_rdev = rdev; 1667 } 1668 1669 /* 1670 * Iff doesn't exist, create it. 1671 */ 1672 if (nd.ni_vp) { 1673 error = EEXIST; 1674 abort: 1675 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1676 if (nd.ni_dvp == nd.ni_vp) 1677 vrele(nd.ni_dvp); 1678 else 1679 vput(nd.ni_dvp); 1680 if (nd.ni_vp) 1681 vput(nd.ni_vp); 1682 goto out; 1683 } 1684 va.va_type = vtyp; 1685 if (vtyp == VSOCK) { 1686 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1687 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1688 } else { 1689 if (va.va_type != VFIFO && 1690 (error = suser(cred, (u_short *)0))) { 1691 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1692 vput(nd.ni_dvp); 1693 goto out; 1694 } 1695 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1696 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1697 if (error) { 1698 goto out; 1699 } 1700 if (error) 1701 goto out; 1702 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1703 vput(nd.ni_vp); 1704 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1705 error = EINVAL; 1706 } 1707 } 1708 out: 1709 vp = nd.ni_vp; 1710 if (!error) { 1711 memset((caddr_t)fhp, 0, sizeof(nfh)); 1712 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1713 error = VFS_VPTOFH(vp, &fhp->fh_fid); 1714 if (!error) 1715 error = VOP_GETATTR(vp, &va, cred, procp); 1716 vput(vp); 1717 } 1718 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); 1719 vrele(dirp); 1720 nfsm_reply(NFSX_SRVFH(1) + NFSX_POSTOPATTR(1) + NFSX_WCCDATA(1)); 1721 if (!error) { 1722 nfsm_srvpostop_fh(fhp); 1723 nfsm_srvpostop_attr(0, &va); 1724 } 1725 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1726 vn_finished_write(mp, 0); 1727 return (0); 1728 nfsmout: 1729 if (dirp) 1730 vrele(dirp); 1731 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1732 if (nd.ni_dvp == nd.ni_vp) 1733 vrele(nd.ni_dvp); 1734 else 1735 vput(nd.ni_dvp); 1736 if (nd.ni_vp) 1737 vput(nd.ni_vp); 1738 vn_finished_write(mp, 0); 1739 return (error); 1740 } 1741 1742 /* 1743 * nfs remove service 1744 */ 1745 int 1746 nfsrv_remove(nfsd, slp, procp, mrq) 1747 struct nfsrv_descript *nfsd; 1748 struct nfssvc_sock *slp; 1749 struct proc *procp; 1750 struct mbuf **mrq; 1751 { 1752 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1753 struct mbuf *nam = nfsd->nd_nam; 1754 caddr_t dpos = nfsd->nd_dpos; 1755 struct ucred *cred = &nfsd->nd_cr; 1756 struct nameidata nd; 1757 u_int32_t *tl; 1758 int32_t t1; 1759 caddr_t bpos; 1760 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 1761 int v3 = (nfsd->nd_flag & ND_NFSV3); 1762 char *cp2; 1763 struct mbuf *mb, *mreq; 1764 struct vnode *vp, *dirp; 1765 struct vattr dirfor, diraft; 1766 nfsfh_t nfh; 1767 fhandle_t *fhp; 1768 u_quad_t frev; 1769 struct mount *mp = NULL; 1770 1771 #ifndef nolint 1772 vp = (struct vnode *)0; 1773 #endif 1774 fhp = &nfh.fh_generic; 1775 nfsm_srvmtofh(fhp); 1776 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1777 return (ESTALE); 1778 vn_start_write(NULL, &mp, V_WAIT); 1779 nfsm_srvnamesiz(len); 1780 nd.ni_cnd.cn_cred = cred; 1781 nd.ni_cnd.cn_nameiop = DELETE; 1782 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1783 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1784 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1785 if (dirp) { 1786 if (v3) 1787 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, 1788 procp); 1789 else 1790 vrele(dirp); 1791 } 1792 if (!error) { 1793 vp = nd.ni_vp; 1794 if (vp->v_type == VDIR && 1795 (error = suser(cred, (u_short *)0)) != 0) 1796 goto out; 1797 /* 1798 * The root of a mounted filesystem cannot be deleted. 1799 */ 1800 if (vp->v_flag & VROOT) { 1801 error = EBUSY; 1802 goto out; 1803 } 1804 out: 1805 if (!error) { 1806 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1807 nqsrv_getl(vp, ND_WRITE); 1808 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1809 } else { 1810 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1811 if (nd.ni_dvp == vp) 1812 vrele(nd.ni_dvp); 1813 else 1814 vput(nd.ni_dvp); 1815 vput(vp); 1816 } 1817 } 1818 if (dirp && v3) { 1819 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); 1820 vrele(dirp); 1821 } 1822 nfsm_reply(NFSX_WCCDATA(v3)); 1823 if (v3) { 1824 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1825 vn_finished_write(mp, 0); 1826 return (0); 1827 } 1828 vn_finished_write(mp, 0); 1829 nfsm_srvdone; 1830 } 1831 1832 /* 1833 * nfs rename service 1834 */ 1835 int 1836 nfsrv_rename(nfsd, slp, procp, mrq) 1837 struct nfsrv_descript *nfsd; 1838 struct nfssvc_sock *slp; 1839 struct proc *procp; 1840 struct mbuf **mrq; 1841 { 1842 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1843 struct mbuf *nam = nfsd->nd_nam; 1844 caddr_t dpos = nfsd->nd_dpos; 1845 struct ucred *cred = &nfsd->nd_cr; 1846 u_int32_t *tl; 1847 int32_t t1; 1848 caddr_t bpos; 1849 int error = 0, cache, fdirfor_ret = 1, fdiraft_ret = 1; 1850 uint32_t len, len2; 1851 int tdirfor_ret = 1, tdiraft_ret = 1; 1852 int v3 = (nfsd->nd_flag & ND_NFSV3); 1853 char *cp2; 1854 struct mbuf *mb, *mreq; 1855 struct nameidata fromnd, tond; 1856 struct vnode *fvp, *tvp, *tdvp, *fdirp = (struct vnode *)0; 1857 struct vnode *tdirp = (struct vnode *)0; 1858 struct vattr fdirfor, fdiraft, tdirfor, tdiraft; 1859 nfsfh_t fnfh, tnfh; 1860 fhandle_t *ffhp, *tfhp; 1861 u_quad_t frev; 1862 uid_t saved_uid; 1863 struct mount *mp = NULL; 1864 1865 #ifndef nolint 1866 fvp = (struct vnode *)0; 1867 #endif 1868 ffhp = &fnfh.fh_generic; 1869 tfhp = &tnfh.fh_generic; 1870 fromnd.ni_cnd.cn_nameiop = 0; 1871 tond.ni_cnd.cn_nameiop = 0; 1872 nfsm_srvmtofh(ffhp); 1873 if ((mp = vfs_getvfs(&ffhp->fh_fsid)) == NULL) 1874 return (ESTALE); 1875 vn_start_write(NULL, &mp, V_WAIT); 1876 nfsm_srvnamesiz(len); 1877 /* 1878 * Remember our original uid so that we can reset cr_uid before 1879 * the second nfs_namei() call, in case it is remapped. 1880 */ 1881 saved_uid = cred->cr_uid; 1882 fromnd.ni_cnd.cn_cred = cred; 1883 fromnd.ni_cnd.cn_nameiop = DELETE; 1884 fromnd.ni_cnd.cn_flags = WANTPARENT | SAVESTART; 1885 error = nfs_namei(&fromnd, ffhp, len, slp, nam, &md, 1886 &dpos, &fdirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1887 if (fdirp) { 1888 if (v3) 1889 fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred, 1890 procp); 1891 else { 1892 vrele(fdirp); 1893 fdirp = (struct vnode *)0; 1894 } 1895 } 1896 if (error) { 1897 nfsm_reply(2 * NFSX_WCCDATA(v3)); 1898 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 1899 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 1900 if (fdirp) 1901 vrele(fdirp); 1902 vn_finished_write(mp, 0); 1903 return (0); 1904 } 1905 fvp = fromnd.ni_vp; 1906 nfsm_srvmtofh(tfhp); 1907 if (v3) { 1908 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 1909 len2 = fxdr_unsigned(uint32_t, *tl); 1910 /* len2 will be checked by nfs_namei */ 1911 } 1912 else { 1913 /* NFSv2 */ 1914 nfsm_strsiz(len2, NFS_MAXNAMLEN); 1915 } 1916 cred->cr_uid = saved_uid; 1917 tond.ni_cnd.cn_cred = cred; 1918 tond.ni_cnd.cn_nameiop = RENAME; 1919 tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART; 1920 error = nfs_namei(&tond, tfhp, len2, slp, nam, &md, 1921 &dpos, &tdirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1922 if (tdirp) { 1923 if (v3) 1924 tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred, 1925 procp); 1926 else { 1927 vrele(tdirp); 1928 tdirp = (struct vnode *)0; 1929 } 1930 } 1931 if (error) { 1932 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1933 vrele(fromnd.ni_dvp); 1934 vrele(fvp); 1935 goto out1; 1936 } 1937 tdvp = tond.ni_dvp; 1938 tvp = tond.ni_vp; 1939 if (tvp != NULL) { 1940 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 1941 if (v3) 1942 error = EEXIST; 1943 else 1944 error = EISDIR; 1945 goto out; 1946 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 1947 if (v3) 1948 error = EEXIST; 1949 else 1950 error = ENOTDIR; 1951 goto out; 1952 } 1953 if (tvp->v_type == VDIR && tvp->v_mountedhere) { 1954 if (v3) 1955 error = EXDEV; 1956 else 1957 error = ENOTEMPTY; 1958 goto out; 1959 } 1960 } 1961 if (fvp->v_type == VDIR && fvp->v_mountedhere) { 1962 if (v3) 1963 error = EXDEV; 1964 else 1965 error = ENOTEMPTY; 1966 goto out; 1967 } 1968 if (fvp->v_mount != tdvp->v_mount) { 1969 if (v3) 1970 error = EXDEV; 1971 else 1972 error = ENOTEMPTY; 1973 goto out; 1974 } 1975 if (fvp == tdvp) { 1976 if (v3) 1977 error = EINVAL; 1978 else 1979 error = ENOTEMPTY; 1980 } 1981 /* 1982 * If source is the same as the destination (that is the 1983 * same vnode with the same name in the same directory), 1984 * then there is nothing to do. 1985 */ 1986 if (fvp == tvp && fromnd.ni_dvp == tdvp && 1987 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 1988 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr, 1989 fromnd.ni_cnd.cn_namelen)) 1990 error = -1; 1991 out: 1992 if (!error) { 1993 nqsrv_getl(fromnd.ni_dvp, ND_WRITE); 1994 nqsrv_getl(tdvp, ND_WRITE); 1995 if (tvp) { 1996 nqsrv_getl(tvp, ND_WRITE); 1997 } 1998 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 1999 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 2000 } else { 2001 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 2002 if (tdvp == tvp) 2003 vrele(tdvp); 2004 else 2005 vput(tdvp); 2006 if (tvp) 2007 vput(tvp); 2008 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2009 vrele(fromnd.ni_dvp); 2010 vrele(fvp); 2011 if (error == -1) 2012 error = 0; 2013 } 2014 vrele(tond.ni_startdir); 2015 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 2016 out1: 2017 if (fdirp) { 2018 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred, procp); 2019 vrele(fdirp); 2020 } 2021 if (tdirp) { 2022 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred, procp); 2023 vrele(tdirp); 2024 } 2025 vrele(fromnd.ni_startdir); 2026 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2027 nfsm_reply(2 * NFSX_WCCDATA(v3)); 2028 if (v3) { 2029 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 2030 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 2031 } 2032 vn_finished_write(mp, 0); 2033 return (0); 2034 2035 nfsmout: 2036 if (fdirp) 2037 vrele(fdirp); 2038 if (tdirp) 2039 vrele(tdirp); 2040 if (tond.ni_cnd.cn_nameiop) { 2041 vrele(tond.ni_startdir); 2042 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 2043 } 2044 if (fromnd.ni_cnd.cn_nameiop) { 2045 vrele(fromnd.ni_startdir); 2046 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2047 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2048 vrele(fromnd.ni_dvp); 2049 vrele(fvp); 2050 } 2051 return (error); 2052 } 2053 2054 /* 2055 * nfs link service 2056 */ 2057 int 2058 nfsrv_link(nfsd, slp, procp, mrq) 2059 struct nfsrv_descript *nfsd; 2060 struct nfssvc_sock *slp; 2061 struct proc *procp; 2062 struct mbuf **mrq; 2063 { 2064 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2065 struct mbuf *nam = nfsd->nd_nam; 2066 caddr_t dpos = nfsd->nd_dpos; 2067 struct ucred *cred = &nfsd->nd_cr; 2068 struct nameidata nd; 2069 u_int32_t *tl; 2070 int32_t t1; 2071 caddr_t bpos; 2072 int error = 0, rdonly, cache, len, dirfor_ret = 1, diraft_ret = 1; 2073 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3); 2074 char *cp2; 2075 struct mbuf *mb, *mreq; 2076 struct vnode *vp, *xp, *dirp = (struct vnode *)0; 2077 struct vattr dirfor, diraft, at; 2078 nfsfh_t nfh, dnfh; 2079 fhandle_t *fhp, *dfhp; 2080 u_quad_t frev; 2081 struct mount *mp = NULL; 2082 2083 fhp = &nfh.fh_generic; 2084 dfhp = &dnfh.fh_generic; 2085 nfsm_srvmtofh(fhp); 2086 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2087 return (ESTALE); 2088 vn_start_write(NULL, &mp, V_WAIT); 2089 nfsm_srvmtofh(dfhp); 2090 nfsm_srvnamesiz(len); 2091 error = nfsrv_fhtovp(fhp, FALSE, &vp, cred, slp, nam, 2092 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2093 if (error) { 2094 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2095 nfsm_srvpostop_attr(getret, &at); 2096 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2097 vn_finished_write(mp, 0); 2098 return (0); 2099 } 2100 if (vp->v_type == VDIR && (error = suser(cred, (u_short *)0)) != 0) 2101 goto out1; 2102 nd.ni_cnd.cn_cred = cred; 2103 nd.ni_cnd.cn_nameiop = CREATE; 2104 nd.ni_cnd.cn_flags = LOCKPARENT; 2105 error = nfs_namei(&nd, dfhp, len, slp, nam, &md, &dpos, 2106 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2107 if (dirp) { 2108 if (v3) 2109 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, 2110 procp); 2111 else { 2112 vrele(dirp); 2113 dirp = (struct vnode *)0; 2114 } 2115 } 2116 if (error) 2117 goto out1; 2118 xp = nd.ni_vp; 2119 if (xp != NULL) { 2120 error = EEXIST; 2121 goto out; 2122 } 2123 xp = nd.ni_dvp; 2124 if (vp->v_mount != xp->v_mount) 2125 error = EXDEV; 2126 out: 2127 if (!error) { 2128 nqsrv_getl(vp, ND_WRITE); 2129 nqsrv_getl(xp, ND_WRITE); 2130 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2131 } else { 2132 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2133 if (nd.ni_dvp == nd.ni_vp) 2134 vrele(nd.ni_dvp); 2135 else 2136 vput(nd.ni_dvp); 2137 if (nd.ni_vp) 2138 vrele(nd.ni_vp); 2139 } 2140 out1: 2141 if (v3) 2142 getret = VOP_GETATTR(vp, &at, cred, procp); 2143 if (dirp) { 2144 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); 2145 vrele(dirp); 2146 } 2147 vrele(vp); 2148 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2149 if (v3) { 2150 nfsm_srvpostop_attr(getret, &at); 2151 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2152 vn_finished_write(mp, 0); 2153 return (0); 2154 } 2155 vn_finished_write(mp, 0); 2156 nfsm_srvdone; 2157 } 2158 2159 /* 2160 * nfs symbolic link service 2161 */ 2162 int 2163 nfsrv_symlink(nfsd, slp, procp, mrq) 2164 struct nfsrv_descript *nfsd; 2165 struct nfssvc_sock *slp; 2166 struct proc *procp; 2167 struct mbuf **mrq; 2168 { 2169 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2170 struct mbuf *nam = nfsd->nd_nam; 2171 caddr_t dpos = nfsd->nd_dpos; 2172 struct ucred *cred = &nfsd->nd_cr; 2173 struct vattr va, dirfor, diraft; 2174 struct nameidata nd; 2175 u_int32_t *tl; 2176 int32_t t1; 2177 struct nfsv2_sattr *sp; 2178 char *bpos, *pathcp = NULL, *cp2; 2179 struct uio io; 2180 struct iovec iv; 2181 int error = 0, cache, dirfor_ret = 1, diraft_ret = 1; 2182 uint32_t len, len2; 2183 int v3 = (nfsd->nd_flag & ND_NFSV3); 2184 struct mbuf *mb, *mreq; 2185 struct vnode *dirp = (struct vnode *)0; 2186 nfsfh_t nfh; 2187 fhandle_t *fhp; 2188 u_quad_t frev; 2189 struct mount *mp = NULL; 2190 2191 nd.ni_cnd.cn_nameiop = 0; 2192 fhp = &nfh.fh_generic; 2193 nfsm_srvmtofh(fhp); 2194 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2195 return (ESTALE); 2196 vn_start_write(NULL, &mp, V_WAIT); 2197 nfsm_srvnamesiz(len); 2198 nd.ni_cnd.cn_cred = cred; 2199 nd.ni_cnd.cn_nameiop = CREATE; 2200 nd.ni_cnd.cn_flags = LOCKPARENT; 2201 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2202 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2203 if (dirp) { 2204 if (v3) 2205 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, 2206 procp); 2207 else { 2208 vrele(dirp); 2209 dirp = (struct vnode *)0; 2210 } 2211 } 2212 if (error) 2213 goto out; 2214 VATTR_NULL(&va); 2215 va.va_type = VLNK; 2216 if (v3) { 2217 va.va_mode = 0; 2218 nfsm_srvsattr(&va); 2219 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 2220 len2 = fxdr_unsigned(uint32_t, *tl); 2221 if (len2 > PATH_MAX) { 2222 /* XXX should check _PC_NO_TRUNC */ 2223 error = ENAMETOOLONG; 2224 goto abortop; 2225 } 2226 } 2227 else { 2228 /* NFSv2 */ 2229 nfsm_strsiz(len2, NFS_MAXPATHLEN); 2230 } 2231 pathcp = malloc(len2 + 1, M_TEMP, M_WAITOK); 2232 iv.iov_base = pathcp; 2233 iv.iov_len = len2; 2234 io.uio_resid = len2; 2235 io.uio_offset = 0; 2236 io.uio_iov = &iv; 2237 io.uio_iovcnt = 1; 2238 io.uio_segflg = UIO_SYSSPACE; 2239 io.uio_rw = UIO_READ; 2240 io.uio_procp = NULL; 2241 nfsm_mtouio(&io, len2); 2242 if (!v3) { 2243 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 2244 va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode); 2245 } 2246 *(pathcp + len2) = '\0'; 2247 if (nd.ni_vp) { 2248 error = EEXIST; 2249 abortop: 2250 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2251 if (nd.ni_dvp == nd.ni_vp) 2252 vrele(nd.ni_dvp); 2253 else 2254 vput(nd.ni_dvp); 2255 if (nd.ni_vp) 2256 vrele(nd.ni_vp); 2257 goto out; 2258 } 2259 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2260 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp); 2261 if (!error) { 2262 if (v3) { 2263 memset((caddr_t)fhp, 0, sizeof(nfh)); 2264 fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsidx; 2265 error = VFS_VPTOFH(nd.ni_vp, &fhp->fh_fid); 2266 if (!error) 2267 error = VOP_GETATTR(nd.ni_vp, &va, cred, 2268 procp); 2269 vput(nd.ni_vp); 2270 } else { 2271 vput(nd.ni_vp); 2272 } 2273 } 2274 out: 2275 if (pathcp) 2276 free(pathcp, M_TEMP); 2277 if (dirp) { 2278 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); 2279 vrele(dirp); 2280 } 2281 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2282 if (v3) { 2283 if (!error) { 2284 nfsm_srvpostop_fh(fhp); 2285 nfsm_srvpostop_attr(0, &va); 2286 } 2287 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2288 } 2289 vn_finished_write(mp, 0); 2290 return (0); 2291 nfsmout: 2292 if (dirp) 2293 vrele(dirp); 2294 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2295 if (nd.ni_dvp == nd.ni_vp) 2296 vrele(nd.ni_dvp); 2297 else 2298 vput(nd.ni_dvp); 2299 if (nd.ni_vp) 2300 vrele(nd.ni_vp); 2301 if (pathcp) 2302 free(pathcp, M_TEMP); 2303 vn_finished_write(mp, 0); 2304 return (error); 2305 } 2306 2307 /* 2308 * nfs mkdir service 2309 */ 2310 int 2311 nfsrv_mkdir(nfsd, slp, procp, mrq) 2312 struct nfsrv_descript *nfsd; 2313 struct nfssvc_sock *slp; 2314 struct proc *procp; 2315 struct mbuf **mrq; 2316 { 2317 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2318 struct mbuf *nam = nfsd->nd_nam; 2319 caddr_t dpos = nfsd->nd_dpos; 2320 struct ucred *cred = &nfsd->nd_cr; 2321 struct vattr va, dirfor, diraft; 2322 struct nfs_fattr *fp; 2323 struct nameidata nd; 2324 caddr_t cp; 2325 u_int32_t *tl; 2326 int32_t t1; 2327 caddr_t bpos; 2328 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 2329 int v3 = (nfsd->nd_flag & ND_NFSV3); 2330 char *cp2; 2331 struct mbuf *mb, *mreq; 2332 struct vnode *vp, *dirp = (struct vnode *)0; 2333 nfsfh_t nfh; 2334 fhandle_t *fhp; 2335 u_quad_t frev; 2336 struct mount *mp = NULL; 2337 2338 fhp = &nfh.fh_generic; 2339 nfsm_srvmtofh(fhp); 2340 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2341 return (ESTALE); 2342 vn_start_write(NULL, &mp, V_WAIT); 2343 nfsm_srvnamesiz(len); 2344 nd.ni_cnd.cn_cred = cred; 2345 nd.ni_cnd.cn_nameiop = CREATE; 2346 nd.ni_cnd.cn_flags = LOCKPARENT; 2347 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2348 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2349 if (dirp) { 2350 if (v3) 2351 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, 2352 procp); 2353 else { 2354 vrele(dirp); 2355 dirp = (struct vnode *)0; 2356 } 2357 } 2358 if (error) { 2359 nfsm_reply(NFSX_WCCDATA(v3)); 2360 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2361 if (dirp) 2362 vrele(dirp); 2363 vn_finished_write(mp, 0); 2364 return (0); 2365 } 2366 VATTR_NULL(&va); 2367 if (v3) { 2368 va.va_mode = 0; 2369 nfsm_srvsattr(&va); 2370 } else { 2371 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 2372 va.va_mode = nfstov_mode(*tl++); 2373 } 2374 va.va_type = VDIR; 2375 vp = nd.ni_vp; 2376 if (vp != NULL) { 2377 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2378 if (nd.ni_dvp == vp) 2379 vrele(nd.ni_dvp); 2380 else 2381 vput(nd.ni_dvp); 2382 vrele(vp); 2383 error = EEXIST; 2384 goto out; 2385 } 2386 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2387 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 2388 if (!error) { 2389 vp = nd.ni_vp; 2390 memset((caddr_t)fhp, 0, sizeof(nfh)); 2391 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 2392 error = VFS_VPTOFH(vp, &fhp->fh_fid); 2393 if (!error) 2394 error = VOP_GETATTR(vp, &va, cred, procp); 2395 vput(vp); 2396 } 2397 out: 2398 if (dirp) { 2399 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); 2400 vrele(dirp); 2401 } 2402 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2403 if (v3) { 2404 if (!error) { 2405 nfsm_srvpostop_fh(fhp); 2406 nfsm_srvpostop_attr(0, &va); 2407 } 2408 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2409 } else { 2410 nfsm_srvfhtom(fhp, v3); 2411 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 2412 nfsm_srvfillattr(&va, fp); 2413 } 2414 vn_finished_write(mp, 0); 2415 return (0); 2416 nfsmout: 2417 if (dirp) 2418 vrele(dirp); 2419 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2420 if (nd.ni_dvp == nd.ni_vp) 2421 vrele(nd.ni_dvp); 2422 else 2423 vput(nd.ni_dvp); 2424 if (nd.ni_vp) 2425 vrele(nd.ni_vp); 2426 vn_finished_write(mp, 0); 2427 return (error); 2428 } 2429 2430 /* 2431 * nfs rmdir service 2432 */ 2433 int 2434 nfsrv_rmdir(nfsd, slp, procp, mrq) 2435 struct nfsrv_descript *nfsd; 2436 struct nfssvc_sock *slp; 2437 struct proc *procp; 2438 struct mbuf **mrq; 2439 { 2440 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2441 struct mbuf *nam = nfsd->nd_nam; 2442 caddr_t dpos = nfsd->nd_dpos; 2443 struct ucred *cred = &nfsd->nd_cr; 2444 u_int32_t *tl; 2445 int32_t t1; 2446 caddr_t bpos; 2447 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 2448 int v3 = (nfsd->nd_flag & ND_NFSV3); 2449 char *cp2; 2450 struct mbuf *mb, *mreq; 2451 struct vnode *vp, *dirp = (struct vnode *)0; 2452 struct vattr dirfor, diraft; 2453 nfsfh_t nfh; 2454 fhandle_t *fhp; 2455 struct nameidata nd; 2456 u_quad_t frev; 2457 struct mount *mp = NULL; 2458 2459 fhp = &nfh.fh_generic; 2460 nfsm_srvmtofh(fhp); 2461 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2462 return (ESTALE); 2463 vn_start_write(NULL, &mp, V_WAIT); 2464 nfsm_srvnamesiz(len); 2465 nd.ni_cnd.cn_cred = cred; 2466 nd.ni_cnd.cn_nameiop = DELETE; 2467 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 2468 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2469 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2470 if (dirp) { 2471 if (v3) 2472 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, 2473 procp); 2474 else { 2475 vrele(dirp); 2476 dirp = (struct vnode *)0; 2477 } 2478 } 2479 if (error) { 2480 nfsm_reply(NFSX_WCCDATA(v3)); 2481 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2482 if (dirp) 2483 vrele(dirp); 2484 vn_finished_write(mp, 0); 2485 return (0); 2486 } 2487 vp = nd.ni_vp; 2488 if (vp->v_type != VDIR) { 2489 error = ENOTDIR; 2490 goto out; 2491 } 2492 /* 2493 * No rmdir "." please. 2494 */ 2495 if (nd.ni_dvp == vp) { 2496 error = EINVAL; 2497 goto out; 2498 } 2499 /* 2500 * The root of a mounted filesystem cannot be deleted. 2501 */ 2502 if (vp->v_flag & VROOT) 2503 error = EBUSY; 2504 out: 2505 if (!error) { 2506 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2507 nqsrv_getl(vp, ND_WRITE); 2508 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2509 } else { 2510 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2511 if (nd.ni_dvp == nd.ni_vp) 2512 vrele(nd.ni_dvp); 2513 else 2514 vput(nd.ni_dvp); 2515 vput(vp); 2516 } 2517 if (dirp) { 2518 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); 2519 vrele(dirp); 2520 } 2521 nfsm_reply(NFSX_WCCDATA(v3)); 2522 if (v3) { 2523 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2524 vn_finished_write(mp, 0); 2525 return (0); 2526 } 2527 vn_finished_write(mp, 0); 2528 nfsm_srvdone; 2529 } 2530 2531 /* 2532 * nfs readdir service 2533 * - mallocs what it thinks is enough to read 2534 * count rounded up to a multiple of NFS_SRVDIRBLKSIZ <= NFS_MAXREADDIR 2535 * - calls VOP_READDIR() 2536 * - loops around building the reply 2537 * if the output generated exceeds count break out of loop 2538 * The nfsm_clget macro is used here so that the reply will be packed 2539 * tightly in mbuf clusters. 2540 * - it only knows that it has encountered eof when the VOP_READDIR() 2541 * reads nothing 2542 * - as such one readdir rpc will return eof false although you are there 2543 * and then the next will return eof 2544 * - it trims out records with d_fileno == 0 2545 * this doesn't matter for Unix clients, but they might confuse clients 2546 * for other os'. 2547 * - it trims out records with d_type == DT_WHT 2548 * these cannot be seen through NFS (unless we extend the protocol) 2549 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less 2550 * than requested, but this may not apply to all filesystems. For 2551 * example, client NFS does not { although it is never remote mounted 2552 * anyhow } 2553 * The alternate call nfsrv_readdirplus() does lookups as well. 2554 * PS: The NFS protocol spec. does not clarify what the "count" byte 2555 * argument is a count of.. just name strings and file id's or the 2556 * entire reply rpc or ... 2557 * I tried just file name and id sizes and it confused the Sun client, 2558 * so I am using the full rpc size now. The "paranoia.." comment refers 2559 * to including the status longwords that are not a part of the dir. 2560 * "entry" structures, but are in the rpc. 2561 */ 2562 2563 #define NFS_SRVDIRBLKSIZ 1024 2564 2565 struct flrep { 2566 nfsuint64 fl_off; 2567 u_int32_t fl_postopok; 2568 u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)]; 2569 u_int32_t fl_fhok; 2570 u_int32_t fl_fhsize; 2571 u_int32_t fl_nfh[NFSX_V3FH / sizeof (u_int32_t)]; 2572 }; 2573 2574 int 2575 nfsrv_readdir(nfsd, slp, procp, mrq) 2576 struct nfsrv_descript *nfsd; 2577 struct nfssvc_sock *slp; 2578 struct proc *procp; 2579 struct mbuf **mrq; 2580 { 2581 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2582 struct mbuf *nam = nfsd->nd_nam; 2583 caddr_t dpos = nfsd->nd_dpos; 2584 struct ucred *cred = &nfsd->nd_cr; 2585 char *bp, *be; 2586 struct mbuf *mp; 2587 struct dirent *dp; 2588 caddr_t cp; 2589 u_int32_t *tl; 2590 int32_t t1; 2591 caddr_t bpos; 2592 struct mbuf *mb, *mreq, *mp2; 2593 char *cpos, *cend, *cp2, *rbuf; 2594 struct vnode *vp; 2595 struct vattr at; 2596 nfsfh_t nfh; 2597 fhandle_t *fhp; 2598 struct uio io; 2599 struct iovec iv; 2600 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2601 int siz, cnt, fullsiz, eofflag, rdonly, cache, ncookies; 2602 int v3 = (nfsd->nd_flag & ND_NFSV3); 2603 u_quad_t frev, off, toff, verf; 2604 off_t *cookies = NULL, *cookiep; 2605 nfsuint64 jar; 2606 2607 fhp = &nfh.fh_generic; 2608 nfsm_srvmtofh(fhp); 2609 if (v3) { 2610 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 2611 toff = fxdr_hyper(tl); 2612 tl += 2; 2613 verf = fxdr_hyper(tl); 2614 tl += 2; 2615 } else { 2616 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2617 toff = fxdr_unsigned(u_quad_t, *tl++); 2618 } 2619 off = toff; 2620 cnt = fxdr_unsigned(int, *tl); 2621 siz = ((cnt + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1)); 2622 xfer = NFS_SRVMAXDATA(nfsd); 2623 if (siz > xfer) 2624 siz = xfer; 2625 fullsiz = siz; 2626 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 2627 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2628 if (!error && vp->v_type != VDIR) { 2629 error = ENOTDIR; 2630 vput(vp); 2631 } 2632 if (error) { 2633 nfsm_reply(NFSX_UNSIGNED); 2634 nfsm_srvpostop_attr(getret, &at); 2635 return (0); 2636 } 2637 nqsrv_getl(vp, ND_READ); 2638 if (v3) { 2639 error = getret = VOP_GETATTR(vp, &at, cred, procp); 2640 #ifdef NFS3_STRICTVERF 2641 /* 2642 * XXX This check is too strict for Solaris 2.5 clients. 2643 */ 2644 if (!error && toff && verf != at.va_filerev) 2645 error = NFSERR_BAD_COOKIE; 2646 #endif 2647 } 2648 if (!error) 2649 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0); 2650 if (error) { 2651 vput(vp); 2652 nfsm_reply(NFSX_POSTOPATTR(v3)); 2653 nfsm_srvpostop_attr(getret, &at); 2654 return (0); 2655 } 2656 VOP_UNLOCK(vp, 0); 2657 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2658 again: 2659 iv.iov_base = rbuf; 2660 iv.iov_len = fullsiz; 2661 io.uio_iov = &iv; 2662 io.uio_iovcnt = 1; 2663 io.uio_offset = (off_t)off; 2664 io.uio_resid = fullsiz; 2665 io.uio_segflg = UIO_SYSSPACE; 2666 io.uio_rw = UIO_READ; 2667 io.uio_procp = NULL; 2668 eofflag = 0; 2669 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2670 2671 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2672 2673 off = (off_t)io.uio_offset; 2674 if (!cookies && !error) 2675 error = NFSERR_PERM; 2676 if (v3) { 2677 getret = VOP_GETATTR(vp, &at, cred, procp); 2678 if (!error) 2679 error = getret; 2680 } 2681 2682 VOP_UNLOCK(vp, 0); 2683 if (error) { 2684 vrele(vp); 2685 free((caddr_t)rbuf, M_TEMP); 2686 if (cookies) 2687 free((caddr_t)cookies, M_TEMP); 2688 nfsm_reply(NFSX_POSTOPATTR(v3)); 2689 nfsm_srvpostop_attr(getret, &at); 2690 return (0); 2691 } 2692 if (io.uio_resid) { 2693 siz -= io.uio_resid; 2694 2695 /* 2696 * If nothing read, return eof 2697 * rpc reply 2698 */ 2699 if (siz == 0) { 2700 vrele(vp); 2701 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + 2702 2 * NFSX_UNSIGNED); 2703 if (v3) { 2704 nfsm_srvpostop_attr(getret, &at); 2705 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2706 txdr_hyper(at.va_filerev, tl); 2707 tl += 2; 2708 } else 2709 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2710 *tl++ = nfs_false; 2711 *tl = nfs_true; 2712 free((caddr_t)rbuf, M_TEMP); 2713 free((caddr_t)cookies, M_TEMP); 2714 return (0); 2715 } 2716 } 2717 2718 /* 2719 * Check for degenerate cases of nothing useful read. 2720 * If so go try again 2721 */ 2722 cpos = rbuf; 2723 cend = rbuf + siz; 2724 dp = (struct dirent *)cpos; 2725 cookiep = cookies; 2726 2727 while (cpos < cend && ncookies > 0 && 2728 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2729 cpos += dp->d_reclen; 2730 dp = (struct dirent *)cpos; 2731 cookiep++; 2732 ncookies--; 2733 } 2734 if (cpos >= cend || ncookies == 0) { 2735 toff = off; 2736 siz = fullsiz; 2737 goto again; 2738 } 2739 2740 len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */ 2741 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz); 2742 if (v3) { 2743 nfsm_srvpostop_attr(getret, &at); 2744 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2745 txdr_hyper(at.va_filerev, tl); 2746 } 2747 mp = mp2 = mb; 2748 bp = bpos; 2749 be = bp + M_TRAILINGSPACE(mp); 2750 2751 /* Loop through the records and build reply */ 2752 while (cpos < cend && ncookies > 0) { 2753 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 2754 nlen = dp->d_namlen; 2755 rem = nfsm_rndup(nlen)-nlen; 2756 len += (4 * NFSX_UNSIGNED + nlen + rem); 2757 if (v3) 2758 len += 2 * NFSX_UNSIGNED; 2759 if (len > cnt) { 2760 eofflag = 0; 2761 break; 2762 } 2763 /* 2764 * Build the directory record xdr from 2765 * the dirent entry. 2766 */ 2767 nfsm_clget; 2768 *tl = nfs_true; 2769 bp += NFSX_UNSIGNED; 2770 if (v3) { 2771 nfsm_clget; 2772 *tl = txdr_unsigned(dp->d_fileno >> 32); 2773 bp += NFSX_UNSIGNED; 2774 } 2775 nfsm_clget; 2776 *tl = txdr_unsigned(dp->d_fileno); 2777 bp += NFSX_UNSIGNED; 2778 nfsm_clget; 2779 *tl = txdr_unsigned(nlen); 2780 bp += NFSX_UNSIGNED; 2781 2782 /* And loop around copying the name */ 2783 xfer = nlen; 2784 cp = dp->d_name; 2785 while (xfer > 0) { 2786 nfsm_clget; 2787 if ((bp+xfer) > be) 2788 tsiz = be-bp; 2789 else 2790 tsiz = xfer; 2791 memcpy(bp, cp, tsiz); 2792 bp += tsiz; 2793 xfer -= tsiz; 2794 if (xfer > 0) 2795 cp += tsiz; 2796 } 2797 /* And null pad to an int32_t boundary */ 2798 for (i = 0; i < rem; i++) 2799 *bp++ = '\0'; 2800 nfsm_clget; 2801 2802 /* Finish off the record */ 2803 txdr_hyper(*cookiep, &jar); 2804 if (v3) { 2805 *tl = jar.nfsuquad[0]; 2806 bp += NFSX_UNSIGNED; 2807 nfsm_clget; 2808 } 2809 *tl = jar.nfsuquad[1]; 2810 bp += NFSX_UNSIGNED; 2811 } 2812 cpos += dp->d_reclen; 2813 dp = (struct dirent *)cpos; 2814 cookiep++; 2815 ncookies--; 2816 } 2817 vrele(vp); 2818 nfsm_clget; 2819 *tl = nfs_false; 2820 bp += NFSX_UNSIGNED; 2821 nfsm_clget; 2822 if (eofflag) 2823 *tl = nfs_true; 2824 else 2825 *tl = nfs_false; 2826 bp += NFSX_UNSIGNED; 2827 if (mp != mb) { 2828 if (bp < be) 2829 mp->m_len = bp - mtod(mp, caddr_t); 2830 } else 2831 mp->m_len += bp - bpos; 2832 free((caddr_t)rbuf, M_TEMP); 2833 free((caddr_t)cookies, M_TEMP); 2834 nfsm_srvdone; 2835 } 2836 2837 int 2838 nfsrv_readdirplus(nfsd, slp, procp, mrq) 2839 struct nfsrv_descript *nfsd; 2840 struct nfssvc_sock *slp; 2841 struct proc *procp; 2842 struct mbuf **mrq; 2843 { 2844 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2845 struct mbuf *nam = nfsd->nd_nam; 2846 caddr_t dpos = nfsd->nd_dpos; 2847 struct ucred *cred = &nfsd->nd_cr; 2848 char *bp, *be; 2849 struct mbuf *mp; 2850 struct dirent *dp; 2851 caddr_t cp; 2852 u_int32_t *tl; 2853 int32_t t1; 2854 caddr_t bpos; 2855 struct mbuf *mb, *mreq, *mp2; 2856 char *cpos, *cend, *cp2, *rbuf; 2857 struct vnode *vp, *nvp; 2858 struct flrep fl; 2859 nfsfh_t nfh; 2860 fhandle_t *fhp, *nfhp = (fhandle_t *)fl.fl_nfh; 2861 struct uio io; 2862 struct iovec iv; 2863 struct vattr va, at, *vap = &va; 2864 struct nfs_fattr *fp; 2865 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2866 int siz, cnt, fullsiz, eofflag, rdonly, cache, dirlen, ncookies; 2867 u_quad_t frev, off, toff, verf; 2868 off_t *cookies = NULL, *cookiep; 2869 2870 fhp = &nfh.fh_generic; 2871 nfsm_srvmtofh(fhp); 2872 nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED); 2873 toff = fxdr_hyper(tl); 2874 tl += 2; 2875 verf = fxdr_hyper(tl); 2876 tl += 2; 2877 siz = fxdr_unsigned(int, *tl++); 2878 cnt = fxdr_unsigned(int, *tl); 2879 off = toff; 2880 siz = ((siz + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1)); 2881 xfer = NFS_SRVMAXDATA(nfsd); 2882 if (siz > xfer) 2883 siz = xfer; 2884 fullsiz = siz; 2885 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 2886 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2887 if (!error && vp->v_type != VDIR) { 2888 error = ENOTDIR; 2889 vput(vp); 2890 } 2891 if (error) { 2892 nfsm_reply(NFSX_UNSIGNED); 2893 nfsm_srvpostop_attr(getret, &at); 2894 return (0); 2895 } 2896 error = getret = VOP_GETATTR(vp, &at, cred, procp); 2897 #ifdef NFS3_STRICTVERF 2898 /* 2899 * XXX This check is too strict for Solaris 2.5 clients. 2900 */ 2901 if (!error && toff && verf != at.va_filerev) 2902 error = NFSERR_BAD_COOKIE; 2903 #endif 2904 if (!error) { 2905 nqsrv_getl(vp, ND_READ); 2906 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0); 2907 } 2908 if (error) { 2909 vput(vp); 2910 nfsm_reply(NFSX_V3POSTOPATTR); 2911 nfsm_srvpostop_attr(getret, &at); 2912 return (0); 2913 } 2914 VOP_UNLOCK(vp, 0); 2915 2916 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2917 again: 2918 iv.iov_base = rbuf; 2919 iv.iov_len = fullsiz; 2920 io.uio_iov = &iv; 2921 io.uio_iovcnt = 1; 2922 io.uio_offset = (off_t)off; 2923 io.uio_resid = fullsiz; 2924 io.uio_segflg = UIO_SYSSPACE; 2925 io.uio_rw = UIO_READ; 2926 io.uio_procp = NULL; 2927 eofflag = 0; 2928 2929 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2930 2931 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2932 2933 off = (u_quad_t)io.uio_offset; 2934 getret = VOP_GETATTR(vp, &at, cred, procp); 2935 2936 VOP_UNLOCK(vp, 0); 2937 2938 /* 2939 * If the VGET operation doesn't work for this filesystem, 2940 * we can't support readdirplus. Returning NOTSUPP should 2941 * make clients fall back to plain readdir. 2942 * There's no need to check for VPTOFH as well, we wouldn't 2943 * even be here otherwise. 2944 */ 2945 if (!getret) { 2946 if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp))) 2947 getret = (getret == EOPNOTSUPP) ? 2948 NFSERR_NOTSUPP : NFSERR_IO; 2949 else 2950 vput(nvp); 2951 } 2952 2953 if (!cookies && !error) 2954 error = NFSERR_PERM; 2955 if (!error) 2956 error = getret; 2957 if (error) { 2958 vrele(vp); 2959 if (cookies) 2960 free((caddr_t)cookies, M_TEMP); 2961 free((caddr_t)rbuf, M_TEMP); 2962 nfsm_reply(NFSX_V3POSTOPATTR); 2963 nfsm_srvpostop_attr(getret, &at); 2964 return (0); 2965 } 2966 if (io.uio_resid) { 2967 siz -= io.uio_resid; 2968 2969 /* 2970 * If nothing read, return eof 2971 * rpc reply 2972 */ 2973 if (siz == 0) { 2974 vrele(vp); 2975 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2976 2 * NFSX_UNSIGNED); 2977 nfsm_srvpostop_attr(getret, &at); 2978 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2979 txdr_hyper(at.va_filerev, tl); 2980 tl += 2; 2981 *tl++ = nfs_false; 2982 *tl = nfs_true; 2983 free((caddr_t)cookies, M_TEMP); 2984 free((caddr_t)rbuf, M_TEMP); 2985 return (0); 2986 } 2987 } 2988 2989 /* 2990 * Check for degenerate cases of nothing useful read. 2991 * If so go try again 2992 */ 2993 cpos = rbuf; 2994 cend = rbuf + siz; 2995 dp = (struct dirent *)cpos; 2996 cookiep = cookies; 2997 2998 while (cpos < cend && ncookies > 0 && 2999 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 3000 cpos += dp->d_reclen; 3001 dp = (struct dirent *)cpos; 3002 cookiep++; 3003 ncookies--; 3004 } 3005 if (cpos >= cend || ncookies == 0) { 3006 toff = off; 3007 siz = fullsiz; 3008 goto again; 3009 } 3010 3011 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED; 3012 nfsm_reply(cnt); 3013 nfsm_srvpostop_attr(getret, &at); 3014 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 3015 txdr_hyper(at.va_filerev, tl); 3016 mp = mp2 = mb; 3017 bp = bpos; 3018 be = bp + M_TRAILINGSPACE(mp); 3019 3020 /* Loop through the records and build reply */ 3021 while (cpos < cend && ncookies > 0) { 3022 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 3023 nlen = dp->d_namlen; 3024 rem = nfsm_rndup(nlen)-nlen; 3025 3026 /* 3027 * For readdir_and_lookup get the vnode using 3028 * the file number. 3029 */ 3030 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp)) 3031 goto invalid; 3032 memset((caddr_t)nfhp, 0, NFSX_V3FH); 3033 nfhp->fh_fsid = 3034 nvp->v_mount->mnt_stat.f_fsidx; 3035 if (VFS_VPTOFH(nvp, &nfhp->fh_fid)) { 3036 vput(nvp); 3037 goto invalid; 3038 } 3039 if (VOP_GETATTR(nvp, vap, cred, procp)) { 3040 vput(nvp); 3041 goto invalid; 3042 } 3043 vput(nvp); 3044 3045 /* 3046 * If either the dircount or maxcount will be 3047 * exceeded, get out now. Both of these lengths 3048 * are calculated conservatively, including all 3049 * XDR overheads. 3050 */ 3051 len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH + 3052 NFSX_V3POSTOPATTR); 3053 dirlen += (6 * NFSX_UNSIGNED + nlen + rem); 3054 if (len > cnt || dirlen > fullsiz) { 3055 eofflag = 0; 3056 break; 3057 } 3058 3059 /* 3060 * Build the directory record xdr from 3061 * the dirent entry. 3062 */ 3063 fp = (struct nfs_fattr *)&fl.fl_fattr; 3064 nfsm_srvfillattr(vap, fp); 3065 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH); 3066 fl.fl_fhok = nfs_true; 3067 fl.fl_postopok = nfs_true; 3068 txdr_hyper(*cookiep, fl.fl_off.nfsuquad); 3069 3070 nfsm_clget; 3071 *tl = nfs_true; 3072 bp += NFSX_UNSIGNED; 3073 nfsm_clget; 3074 *tl = txdr_unsigned(dp->d_fileno >> 32); 3075 bp += NFSX_UNSIGNED; 3076 nfsm_clget; 3077 *tl = txdr_unsigned(dp->d_fileno); 3078 bp += NFSX_UNSIGNED; 3079 nfsm_clget; 3080 *tl = txdr_unsigned(nlen); 3081 bp += NFSX_UNSIGNED; 3082 3083 /* And loop around copying the name */ 3084 xfer = nlen; 3085 cp = dp->d_name; 3086 while (xfer > 0) { 3087 nfsm_clget; 3088 if ((bp + xfer) > be) 3089 tsiz = be - bp; 3090 else 3091 tsiz = xfer; 3092 memcpy(bp, cp, tsiz); 3093 bp += tsiz; 3094 xfer -= tsiz; 3095 if (xfer > 0) 3096 cp += tsiz; 3097 } 3098 /* And null pad to an int32_t boundary */ 3099 for (i = 0; i < rem; i++) 3100 *bp++ = '\0'; 3101 3102 /* 3103 * Now copy the flrep structure out. 3104 */ 3105 xfer = sizeof (struct flrep); 3106 cp = (caddr_t)&fl; 3107 while (xfer > 0) { 3108 nfsm_clget; 3109 if ((bp + xfer) > be) 3110 tsiz = be - bp; 3111 else 3112 tsiz = xfer; 3113 memcpy(bp, cp, tsiz); 3114 bp += tsiz; 3115 xfer -= tsiz; 3116 if (xfer > 0) 3117 cp += tsiz; 3118 } 3119 } 3120 invalid: 3121 cpos += dp->d_reclen; 3122 dp = (struct dirent *)cpos; 3123 cookiep++; 3124 ncookies--; 3125 } 3126 vrele(vp); 3127 nfsm_clget; 3128 *tl = nfs_false; 3129 bp += NFSX_UNSIGNED; 3130 nfsm_clget; 3131 if (eofflag) 3132 *tl = nfs_true; 3133 else 3134 *tl = nfs_false; 3135 bp += NFSX_UNSIGNED; 3136 if (mp != mb) { 3137 if (bp < be) 3138 mp->m_len = bp - mtod(mp, caddr_t); 3139 } else 3140 mp->m_len += bp - bpos; 3141 free((caddr_t)cookies, M_TEMP); 3142 free((caddr_t)rbuf, M_TEMP); 3143 nfsm_srvdone; 3144 } 3145 3146 /* 3147 * nfs commit service 3148 */ 3149 int 3150 nfsrv_commit(nfsd, slp, procp, mrq) 3151 struct nfsrv_descript *nfsd; 3152 struct nfssvc_sock *slp; 3153 struct proc *procp; 3154 struct mbuf **mrq; 3155 { 3156 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3157 struct mbuf *nam = nfsd->nd_nam; 3158 caddr_t dpos = nfsd->nd_dpos; 3159 struct ucred *cred = &nfsd->nd_cr; 3160 struct vattr bfor, aft; 3161 struct vnode *vp; 3162 nfsfh_t nfh; 3163 fhandle_t *fhp; 3164 u_int32_t *tl; 3165 int32_t t1; 3166 caddr_t bpos; 3167 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cache; 3168 uint32_t cnt; 3169 char *cp2; 3170 struct mbuf *mb, *mreq; 3171 u_quad_t frev, off, end; 3172 struct mount *mp = NULL; 3173 3174 #ifndef nolint 3175 cache = 0; 3176 #endif 3177 fhp = &nfh.fh_generic; 3178 nfsm_srvmtofh(fhp); 3179 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 3180 return (ESTALE); 3181 vn_start_write(NULL, &mp, V_WAIT); 3182 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED); 3183 3184 off = fxdr_hyper(tl); 3185 tl += 2; 3186 cnt = fxdr_unsigned(uint32_t, *tl); 3187 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3188 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3189 if (error) { 3190 nfsm_reply(2 * NFSX_UNSIGNED); 3191 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3192 vn_finished_write(mp, 0); 3193 return (0); 3194 } 3195 for_ret = VOP_GETATTR(vp, &bfor, cred, procp); 3196 end = (cnt > 0) ? off + cnt : vp->v_size; 3197 if (end < off || end > vp->v_size) 3198 end = vp->v_size; 3199 if (off < vp->v_size) 3200 error = VOP_FSYNC(vp, cred, FSYNC_WAIT, off, end, procp); 3201 /* else error == 0, from nfsrv_fhtovp() */ 3202 aft_ret = VOP_GETATTR(vp, &aft, cred, procp); 3203 vput(vp); 3204 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF); 3205 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3206 if (!error) { 3207 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF); 3208 *tl++ = txdr_unsigned(boottime.tv_sec); 3209 *tl = txdr_unsigned(boottime.tv_usec); 3210 } else { 3211 vn_finished_write(mp, 0); 3212 return (0); 3213 } 3214 vn_finished_write(mp, 0); 3215 nfsm_srvdone; 3216 } 3217 3218 /* 3219 * nfs statfs service 3220 */ 3221 int 3222 nfsrv_statfs(nfsd, slp, procp, mrq) 3223 struct nfsrv_descript *nfsd; 3224 struct nfssvc_sock *slp; 3225 struct proc *procp; 3226 struct mbuf **mrq; 3227 { 3228 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3229 struct mbuf *nam = nfsd->nd_nam; 3230 caddr_t dpos = nfsd->nd_dpos; 3231 struct ucred *cred = &nfsd->nd_cr; 3232 struct statvfs *sf; 3233 struct nfs_statfs *sfp; 3234 u_int32_t *tl; 3235 int32_t t1; 3236 caddr_t bpos; 3237 int error = 0, rdonly, cache, getret = 1; 3238 int v3 = (nfsd->nd_flag & ND_NFSV3); 3239 char *cp2; 3240 struct mbuf *mb, *mreq; 3241 struct vnode *vp; 3242 struct vattr at; 3243 nfsfh_t nfh; 3244 fhandle_t *fhp; 3245 struct statvfs statvfs; 3246 u_quad_t frev, tval; 3247 3248 #ifndef nolint 3249 cache = 0; 3250 #endif 3251 fhp = &nfh.fh_generic; 3252 nfsm_srvmtofh(fhp); 3253 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3254 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3255 if (error) { 3256 nfsm_reply(NFSX_UNSIGNED); 3257 nfsm_srvpostop_attr(getret, &at); 3258 return (0); 3259 } 3260 sf = &statvfs; 3261 error = VFS_STATVFS(vp->v_mount, sf, procp); 3262 getret = VOP_GETATTR(vp, &at, cred, procp); 3263 vput(vp); 3264 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3)); 3265 if (v3) 3266 nfsm_srvpostop_attr(getret, &at); 3267 if (error) 3268 return (0); 3269 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3)); 3270 if (v3) { 3271 tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_frsize); 3272 txdr_hyper(tval, &sfp->sf_tbytes); 3273 tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_frsize); 3274 txdr_hyper(tval, &sfp->sf_fbytes); 3275 tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_frsize); 3276 txdr_hyper(tval, &sfp->sf_abytes); 3277 tval = (u_quad_t)sf->f_files; 3278 txdr_hyper(tval, &sfp->sf_tfiles); 3279 tval = (u_quad_t)sf->f_ffree; 3280 txdr_hyper(tval, &sfp->sf_ffiles); 3281 txdr_hyper(tval, &sfp->sf_afiles); 3282 sfp->sf_invarsec = 0; 3283 } else { 3284 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA); 3285 sfp->sf_bsize = txdr_unsigned(sf->f_frsize); 3286 sfp->sf_blocks = txdr_unsigned(sf->f_blocks); 3287 sfp->sf_bfree = txdr_unsigned(sf->f_bfree); 3288 sfp->sf_bavail = txdr_unsigned(sf->f_bavail); 3289 } 3290 nfsm_srvdone; 3291 } 3292 3293 /* 3294 * nfs fsinfo service 3295 */ 3296 int 3297 nfsrv_fsinfo(nfsd, slp, procp, mrq) 3298 struct nfsrv_descript *nfsd; 3299 struct nfssvc_sock *slp; 3300 struct proc *procp; 3301 struct mbuf **mrq; 3302 { 3303 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3304 struct mbuf *nam = nfsd->nd_nam; 3305 caddr_t dpos = nfsd->nd_dpos; 3306 struct ucred *cred = &nfsd->nd_cr; 3307 u_int32_t *tl; 3308 struct nfsv3_fsinfo *sip; 3309 int32_t t1; 3310 caddr_t bpos; 3311 int error = 0, rdonly, cache, getret = 1; 3312 uint32_t maxdata; 3313 char *cp2; 3314 struct mbuf *mb, *mreq; 3315 struct vnode *vp; 3316 struct vattr at; 3317 nfsfh_t nfh; 3318 fhandle_t *fhp; 3319 u_quad_t frev, maxfsize; 3320 struct statvfs sb; 3321 3322 #ifndef nolint 3323 cache = 0; 3324 #endif 3325 fhp = &nfh.fh_generic; 3326 nfsm_srvmtofh(fhp); 3327 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3328 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3329 if (error) { 3330 nfsm_reply(NFSX_UNSIGNED); 3331 nfsm_srvpostop_attr(getret, &at); 3332 return (0); 3333 } 3334 3335 /* XXX Try to make a guess on the max file size. */ 3336 VFS_STATVFS(vp->v_mount, &sb, (struct proc *)0); 3337 maxfsize = (u_quad_t)0x80000000 * sb.f_frsize - 1; 3338 3339 getret = VOP_GETATTR(vp, &at, cred, procp); 3340 vput(vp); 3341 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO); 3342 nfsm_srvpostop_attr(getret, &at); 3343 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO); 3344 3345 /* 3346 * XXX 3347 * There should be file system VFS OP(s) to get this information. 3348 * For now, assume ufs. 3349 */ 3350 if (slp->ns_so->so_type == SOCK_DGRAM) 3351 maxdata = NFS_MAXDGRAMDATA; 3352 else 3353 maxdata = NFS_MAXDATA; 3354 sip->fs_rtmax = txdr_unsigned(maxdata); 3355 sip->fs_rtpref = txdr_unsigned(maxdata); 3356 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE); 3357 sip->fs_wtmax = txdr_unsigned(maxdata); 3358 sip->fs_wtpref = txdr_unsigned(maxdata); 3359 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE); 3360 sip->fs_dtpref = txdr_unsigned(maxdata); 3361 txdr_hyper(maxfsize, &sip->fs_maxfilesize); 3362 sip->fs_timedelta.nfsv3_sec = 0; 3363 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1); 3364 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK | 3365 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS | 3366 NFSV3FSINFO_CANSETTIME); 3367 nfsm_srvdone; 3368 } 3369 3370 /* 3371 * nfs pathconf service 3372 */ 3373 int 3374 nfsrv_pathconf(nfsd, slp, procp, mrq) 3375 struct nfsrv_descript *nfsd; 3376 struct nfssvc_sock *slp; 3377 struct proc *procp; 3378 struct mbuf **mrq; 3379 { 3380 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3381 struct mbuf *nam = nfsd->nd_nam; 3382 caddr_t dpos = nfsd->nd_dpos; 3383 struct ucred *cred = &nfsd->nd_cr; 3384 u_int32_t *tl; 3385 struct nfsv3_pathconf *pc; 3386 int32_t t1; 3387 caddr_t bpos; 3388 int error = 0, rdonly, cache, getret = 1; 3389 register_t linkmax, namemax, chownres, notrunc; 3390 char *cp2; 3391 struct mbuf *mb, *mreq; 3392 struct vnode *vp; 3393 struct vattr at; 3394 nfsfh_t nfh; 3395 fhandle_t *fhp; 3396 u_quad_t frev; 3397 3398 #ifndef nolint 3399 cache = 0; 3400 #endif 3401 fhp = &nfh.fh_generic; 3402 nfsm_srvmtofh(fhp); 3403 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3404 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3405 if (error) { 3406 nfsm_reply(NFSX_UNSIGNED); 3407 nfsm_srvpostop_attr(getret, &at); 3408 return (0); 3409 } 3410 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax); 3411 if (!error) 3412 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax); 3413 if (!error) 3414 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres); 3415 if (!error) 3416 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc); 3417 getret = VOP_GETATTR(vp, &at, cred, procp); 3418 vput(vp); 3419 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF); 3420 nfsm_srvpostop_attr(getret, &at); 3421 if (error) 3422 return (0); 3423 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF); 3424 3425 pc->pc_linkmax = txdr_unsigned(linkmax); 3426 pc->pc_namemax = txdr_unsigned(namemax); 3427 pc->pc_notrunc = txdr_unsigned(notrunc); 3428 pc->pc_chownrestricted = txdr_unsigned(chownres); 3429 3430 /* 3431 * These should probably be supported by VOP_PATHCONF(), but 3432 * until msdosfs is exportable (why would you want to?), the 3433 * Unix defaults should be ok. 3434 */ 3435 pc->pc_caseinsensitive = nfs_false; 3436 pc->pc_casepreserving = nfs_true; 3437 nfsm_srvdone; 3438 } 3439 3440 /* 3441 * Null operation, used by clients to ping server 3442 */ 3443 /* ARGSUSED */ 3444 int 3445 nfsrv_null(nfsd, slp, procp, mrq) 3446 struct nfsrv_descript *nfsd; 3447 struct nfssvc_sock *slp; 3448 struct proc *procp; 3449 struct mbuf **mrq; 3450 { 3451 struct mbuf *mrep = nfsd->nd_mrep; 3452 caddr_t bpos; 3453 int error = NFSERR_RETVOID, cache = 0; 3454 struct mbuf *mb, *mreq; 3455 u_quad_t frev; 3456 3457 nfsm_reply(0); 3458 return (0); 3459 } 3460 3461 /* 3462 * No operation, used for obsolete procedures 3463 */ 3464 /* ARGSUSED */ 3465 int 3466 nfsrv_noop(nfsd, slp, procp, mrq) 3467 struct nfsrv_descript *nfsd; 3468 struct nfssvc_sock *slp; 3469 struct proc *procp; 3470 struct mbuf **mrq; 3471 { 3472 struct mbuf *mrep = nfsd->nd_mrep; 3473 caddr_t bpos; 3474 int error, cache = 0; 3475 struct mbuf *mb, *mreq; 3476 u_quad_t frev; 3477 3478 if (nfsd->nd_repstat) 3479 error = nfsd->nd_repstat; 3480 else 3481 error = EPROCUNAVAIL; 3482 nfsm_reply(0); 3483 return (0); 3484 } 3485 3486 /* 3487 * Perform access checking for vnodes obtained from file handles that would 3488 * refer to files already opened by a Unix client. You cannot just use 3489 * vn_writechk() and VOP_ACCESS() for two reasons. 3490 * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case 3491 * 2 - The owner is to be given access irrespective of mode bits for some 3492 * operations, so that processes that chmod after opening a file don't 3493 * break. I don't like this because it opens a security hole, but since 3494 * the nfs server opens a security hole the size of a barn door anyhow, 3495 * what the heck. 3496 * 3497 * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS() 3498 * will return EPERM instead of EACCESS. EPERM is always an error. 3499 */ 3500 int 3501 nfsrv_access(vp, flags, cred, rdonly, p, override) 3502 struct vnode *vp; 3503 int flags; 3504 struct ucred *cred; 3505 int rdonly; 3506 struct proc *p; 3507 { 3508 struct vattr vattr; 3509 int error; 3510 if (flags & VWRITE) { 3511 /* Just vn_writechk() changed to check rdonly */ 3512 /* 3513 * Disallow write attempts on read-only file systems; 3514 * unless the file is a socket or a block or character 3515 * device resident on the file system. 3516 */ 3517 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 3518 switch (vp->v_type) { 3519 case VREG: 3520 case VDIR: 3521 case VLNK: 3522 return (EROFS); 3523 default: 3524 break; 3525 } 3526 } 3527 3528 /* 3529 * If the vnode is in use as a process's text, 3530 * we can't allow writing. 3531 */ 3532 if (vp->v_flag & VTEXT) 3533 return (ETXTBSY); 3534 } 3535 error = VOP_GETATTR(vp, &vattr, cred, p); 3536 if (error) 3537 return (error); 3538 error = VOP_ACCESS(vp, flags, cred, p); 3539 /* 3540 * Allow certain operations for the owner (reads and writes 3541 * on files that are already open). 3542 */ 3543 if (override && error == EACCES && cred->cr_uid == vattr.va_uid) 3544 error = 0; 3545 return error; 3546 } 3547
Indexes created Wed Sep 24 02:09:48 GMT 2025