nfs_serv.c revision 1.101.4.2
1 /* $NetBSD: nfs_serv.c,v 1.101.4.2 2006/03/10 13:37:46 elad Exp $ */ 2 3 /* 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * This code is derived from software contributed to Berkeley by 8 * Rick Macklem at The University of Guelph. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * @(#)nfs_serv.c 8.8 (Berkeley) 7/31/95 35 */ 36 37 /* 38 * nfs version 2 and 3 server calls to vnode ops 39 * - these routines generally have 3 phases 40 * 1 - break down and validate rpc request in mbuf list 41 * 2 - do the vnode ops for the request 42 * (surprisingly ?? many are very similar to syscalls in vfs_syscalls.c) 43 * 3 - build the rpc reply in an mbuf list 44 * nb: 45 * - do not mix the phases, since the nfsm_?? macros can return failures 46 * on a bad rpc or similar and do not do any vrele() or vput()'s 47 * 48 * - the nfsm_reply() macro generates an nfs rpc reply with the nfs 49 * error number iff error != 0 whereas 50 * returning an error from the server function implies a fatal error 51 * such as a badly constructed rpc request that should be dropped without 52 * a reply. 53 * For Version 3, nfsm_reply() does not return for the error case, since 54 * most version 3 rpcs return more than the status for error cases. 55 */ 56 57 #include <sys/cdefs.h> 58 __KERNEL_RCSID(0, "$NetBSD: nfs_serv.c,v 1.101.4.2 2006/03/10 13:37:46 elad Exp $"); 59 60 #include <sys/param.h> 61 #include <sys/systm.h> 62 #include <sys/proc.h> 63 #include <sys/file.h> 64 #include <sys/namei.h> 65 #include <sys/vnode.h> 66 #include <sys/mount.h> 67 #include <sys/socket.h> 68 #include <sys/socketvar.h> 69 #include <sys/mbuf.h> 70 #include <sys/dirent.h> 71 #include <sys/stat.h> 72 #include <sys/kernel.h> 73 74 #include <uvm/uvm.h> 75 76 #include <nfs/nfsproto.h> 77 #include <nfs/rpcv2.h> 78 #include <nfs/nfs.h> 79 #include <nfs/xdr_subs.h> 80 #include <nfs/nfsm_subs.h> 81 #include <nfs/nqnfs.h> 82 #include <nfs/nfs_var.h> 83 84 /* Global vars */ 85 extern u_int32_t nfs_xdrneg1; 86 extern u_int32_t nfs_false, nfs_true; 87 extern enum vtype nv3tov_type[8]; 88 extern struct nfsstats nfsstats; 89 extern const nfstype nfsv2_type[9]; 90 extern const nfstype nfsv3_type[9]; 91 int nfsrvw_procrastinate = NFS_GATHERDELAY * 1000; 92 int nfsd_use_loan = 1; /* use page-loan for READ OP */ 93 94 /* 95 * nfs v3 access service 96 */ 97 int 98 nfsrv3_access(nfsd, slp, lwp, mrq) 99 struct nfsrv_descript *nfsd; 100 struct nfssvc_sock *slp; 101 struct lwp *lwp; 102 struct mbuf **mrq; 103 { 104 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 105 struct mbuf *nam = nfsd->nd_nam; 106 caddr_t dpos = nfsd->nd_dpos; 107 kauth_cred_t cred = nfsd->nd_cr; 108 struct vnode *vp; 109 nfsfh_t nfh; 110 fhandle_t *fhp; 111 u_int32_t *tl; 112 int32_t t1; 113 caddr_t bpos; 114 int error = 0, rdonly, cache = 0, getret; 115 char *cp2; 116 struct mbuf *mb, *mreq; 117 struct vattr va; 118 u_long inmode, testmode, outmode; 119 u_quad_t frev; 120 121 fhp = &nfh.fh_generic; 122 nfsm_srvmtofh(fhp); 123 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 124 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, 125 (nfsd->nd_flag & ND_KERBAUTH), FALSE); 126 if (error) { 127 nfsm_reply(NFSX_UNSIGNED); 128 nfsm_srvpostop_attr(1, (struct vattr *)0); 129 return (0); 130 } 131 inmode = fxdr_unsigned(u_int32_t, *tl); 132 outmode = 0; 133 if ((inmode & NFSV3ACCESS_READ) && 134 nfsrv_access(vp, VREAD, cred, rdonly, lwp, 0) == 0) 135 outmode |= NFSV3ACCESS_READ; 136 if (vp->v_type != VDIR) { 137 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND); 138 if (testmode && 139 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0) 140 outmode |= testmode; 141 if ((inmode & NFSV3ACCESS_EXECUTE) && 142 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0) 143 outmode |= NFSV3ACCESS_EXECUTE; 144 } else { 145 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND | 146 NFSV3ACCESS_DELETE); 147 if (testmode && 148 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0) 149 outmode |= testmode; 150 if ((inmode & NFSV3ACCESS_LOOKUP) && 151 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0) 152 outmode |= NFSV3ACCESS_LOOKUP; 153 } 154 getret = VOP_GETATTR(vp, &va, cred, lwp); 155 vput(vp); 156 nfsm_reply(NFSX_POSTOPATTR(1) + NFSX_UNSIGNED); 157 nfsm_srvpostop_attr(getret, &va); 158 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 159 *tl = txdr_unsigned(outmode); 160 nfsm_srvdone; 161 } 162 163 /* 164 * nfs getattr service 165 */ 166 int 167 nfsrv_getattr(nfsd, slp, lwp, mrq) 168 struct nfsrv_descript *nfsd; 169 struct nfssvc_sock *slp; 170 struct lwp *lwp; 171 struct mbuf **mrq; 172 { 173 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 174 struct mbuf *nam = nfsd->nd_nam; 175 caddr_t dpos = nfsd->nd_dpos; 176 kauth_cred_t cred = nfsd->nd_cr; 177 struct nfs_fattr *fp; 178 struct vattr va; 179 struct vnode *vp; 180 nfsfh_t nfh; 181 fhandle_t *fhp; 182 u_int32_t *tl; 183 int32_t t1; 184 caddr_t bpos; 185 int error = 0, rdonly, cache; 186 char *cp2; 187 struct mbuf *mb, *mreq; 188 u_quad_t frev; 189 190 fhp = &nfh.fh_generic; 191 nfsm_srvmtofh(fhp); 192 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, 193 (nfsd->nd_flag & ND_KERBAUTH), FALSE); 194 if (error) { 195 nfsm_reply(0); 196 return (0); 197 } 198 nqsrv_getl(vp, ND_READ); 199 error = VOP_GETATTR(vp, &va, cred, lwp); 200 vput(vp); 201 nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); 202 if (error) 203 return (0); 204 nfsm_build(fp, struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); 205 nfsm_srvfillattr(&va, fp); 206 nfsm_srvdone; 207 } 208 209 /* 210 * nfs setattr service 211 */ 212 int 213 nfsrv_setattr(nfsd, slp, lwp, mrq) 214 struct nfsrv_descript *nfsd; 215 struct nfssvc_sock *slp; 216 struct lwp *lwp; 217 struct mbuf **mrq; 218 { 219 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 220 struct mbuf *nam = nfsd->nd_nam; 221 caddr_t dpos = nfsd->nd_dpos; 222 kauth_cred_t cred = nfsd->nd_cr; 223 struct vattr va, preat; 224 struct nfsv2_sattr *sp; 225 struct nfs_fattr *fp; 226 struct vnode *vp; 227 nfsfh_t nfh; 228 fhandle_t *fhp; 229 u_int32_t *tl; 230 int32_t t1; 231 caddr_t bpos; 232 int error = 0, rdonly, cache, preat_ret = 1, postat_ret = 1; 233 int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0; 234 char *cp2; 235 struct mbuf *mb, *mreq; 236 u_quad_t frev; 237 struct timespec guard; 238 struct mount *mp = NULL; 239 240 fhp = &nfh.fh_generic; 241 nfsm_srvmtofh(fhp); 242 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 243 return (ESTALE); 244 vn_start_write(NULL, &mp, V_WAIT); 245 VATTR_NULL(&va); 246 if (v3) { 247 nfsm_srvsattr(&va); 248 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 249 gcheck = fxdr_unsigned(int, *tl); 250 if (gcheck) { 251 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 252 fxdr_nfsv3time(tl, &guard); 253 } 254 } else { 255 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 256 /* 257 * Nah nah nah nah na nah 258 * There is a bug in the Sun client that puts 0xffff in the mode 259 * field of sattr when it should put in 0xffffffff. The u_short 260 * doesn't sign extend. 261 * --> check the low order 2 bytes for 0xffff 262 */ 263 if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff) 264 va.va_mode = nfstov_mode(sp->sa_mode); 265 if (sp->sa_uid != nfs_xdrneg1) 266 va.va_uid = fxdr_unsigned(uid_t, sp->sa_uid); 267 if (sp->sa_gid != nfs_xdrneg1) 268 va.va_gid = fxdr_unsigned(gid_t, sp->sa_gid); 269 if (sp->sa_size != nfs_xdrneg1) 270 va.va_size = fxdr_unsigned(u_quad_t, sp->sa_size); 271 if (sp->sa_atime.nfsv2_sec != nfs_xdrneg1) { 272 #ifdef notyet 273 fxdr_nfsv2time(&sp->sa_atime, &va.va_atime); 274 #else 275 va.va_atime.tv_sec = 276 fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec); 277 va.va_atime.tv_nsec = 0; 278 #endif 279 } 280 if (sp->sa_mtime.nfsv2_sec != nfs_xdrneg1) 281 fxdr_nfsv2time(&sp->sa_mtime, &va.va_mtime); 282 283 } 284 285 /* 286 * Now that we have all the fields, lets do it. 287 */ 288 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, 289 (nfsd->nd_flag & ND_KERBAUTH), FALSE); 290 if (error) { 291 nfsm_reply(2 * NFSX_UNSIGNED); 292 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 293 vn_finished_write(mp, 0); 294 return (0); 295 } 296 nqsrv_getl(vp, ND_WRITE); 297 if (v3) { 298 error = preat_ret = VOP_GETATTR(vp, &preat, cred, lwp); 299 if (!error && gcheck && 300 (preat.va_ctime.tv_sec != guard.tv_sec || 301 preat.va_ctime.tv_nsec != guard.tv_nsec)) 302 error = NFSERR_NOT_SYNC; 303 if (error) { 304 vput(vp); 305 nfsm_reply(NFSX_WCCDATA(v3)); 306 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 307 vn_finished_write(mp, 0); 308 return (0); 309 } 310 } 311 312 /* 313 * If the size is being changed write acces is required, otherwise 314 * just check for a read only file system. 315 */ 316 if (va.va_size == ((u_quad_t)((quad_t) -1))) { 317 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 318 error = EROFS; 319 goto out; 320 } 321 } else { 322 if (vp->v_type == VDIR) { 323 error = EISDIR; 324 goto out; 325 } else if ((error = nfsrv_access(vp, VWRITE, cred, rdonly, 326 lwp, 0)) != 0) 327 goto out; 328 } 329 error = VOP_SETATTR(vp, &va, cred, lwp); 330 postat_ret = VOP_GETATTR(vp, &va, cred, lwp); 331 if (!error) 332 error = postat_ret; 333 out: 334 vput(vp); 335 vn_finished_write(mp, 0); 336 nfsm_reply(NFSX_WCCORFATTR(v3)); 337 if (v3) { 338 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 339 return (0); 340 } else { 341 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 342 nfsm_srvfillattr(&va, fp); 343 } 344 nfsm_srvdone; 345 } 346 347 /* 348 * nfs lookup rpc 349 */ 350 int 351 nfsrv_lookup(nfsd, slp, lwp, mrq) 352 struct nfsrv_descript *nfsd; 353 struct nfssvc_sock *slp; 354 struct lwp *lwp; 355 struct mbuf **mrq; 356 { 357 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 358 struct mbuf *nam = nfsd->nd_nam; 359 caddr_t dpos = nfsd->nd_dpos; 360 kauth_cred_t cred = nfsd->nd_cr; 361 struct nfs_fattr *fp; 362 struct nameidata nd, ind, *ndp = &nd; 363 struct vnode *vp, *dirp; 364 nfsfh_t nfh; 365 fhandle_t *fhp; 366 caddr_t cp; 367 u_int32_t *tl; 368 int32_t t1; 369 caddr_t bpos; 370 int error = 0, cache, dirattr_ret = 1; 371 uint32_t len; 372 int v3 = (nfsd->nd_flag & ND_NFSV3), pubflag; 373 char *cp2; 374 struct mbuf *mb, *mreq; 375 struct vattr va, dirattr; 376 u_quad_t frev; 377 378 fhp = &nfh.fh_generic; 379 nfsm_srvmtofh(fhp); 380 nfsm_srvnamesiz(len); 381 382 pubflag = nfs_ispublicfh(fhp); 383 384 nd.ni_cnd.cn_cred = cred; 385 nd.ni_cnd.cn_nameiop = LOOKUP; 386 nd.ni_cnd.cn_flags = LOCKLEAF | SAVESTART; 387 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 388 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), pubflag); 389 390 if (!error && pubflag) { 391 if (nd.ni_vp->v_type == VDIR && nfs_pub.np_index != NULL) { 392 /* 393 * Setup call to lookup() to see if we can find 394 * the index file. Arguably, this doesn't belong 395 * in a kernel.. Ugh. 396 */ 397 ind = nd; 398 VOP_UNLOCK(nd.ni_vp, 0); 399 ind.ni_pathlen = strlen(nfs_pub.np_index); 400 ind.ni_cnd.cn_nameptr = ind.ni_cnd.cn_pnbuf = 401 nfs_pub.np_index; 402 ind.ni_startdir = nd.ni_vp; 403 VREF(ind.ni_startdir); 404 error = lookup(&ind); 405 if (!error) { 406 /* 407 * Found an index file. Get rid of 408 * the old references. 409 */ 410 if (dirp) 411 vrele(dirp); 412 dirp = nd.ni_vp; 413 vrele(nd.ni_startdir); 414 ndp = &ind; 415 } else 416 error = 0; 417 } 418 /* 419 * If the public filehandle was used, check that this lookup 420 * didn't result in a filehandle outside the publicly exported 421 * filesystem. 422 */ 423 424 if (!error && ndp->ni_vp->v_mount != nfs_pub.np_mount) { 425 vput(nd.ni_vp); 426 error = EPERM; 427 } 428 } 429 430 if (dirp) { 431 if (v3) 432 dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred, lwp); 433 vrele(dirp); 434 } 435 436 if (error) { 437 nfsm_reply(NFSX_POSTOPATTR(v3)); 438 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 439 return (0); 440 } 441 442 nqsrv_getl(ndp->ni_startdir, ND_READ); 443 vrele(ndp->ni_startdir); 444 PNBUF_PUT(nd.ni_cnd.cn_pnbuf); 445 vp = ndp->ni_vp; 446 memset((caddr_t)fhp, 0, sizeof(nfh)); 447 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 448 error = VFS_VPTOFH(vp, &fhp->fh_fid); 449 if (!error) 450 error = VOP_GETATTR(vp, &va, cred, lwp); 451 vput(vp); 452 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPORFATTR(v3) + NFSX_POSTOPATTR(v3)); 453 if (error) { 454 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 455 return (0); 456 } 457 nfsm_srvfhtom(fhp, v3); 458 if (v3) { 459 nfsm_srvpostop_attr(0, &va); 460 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 461 } else { 462 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 463 nfsm_srvfillattr(&va, fp); 464 } 465 nfsm_srvdone; 466 } 467 468 /* 469 * nfs readlink service 470 */ 471 int 472 nfsrv_readlink(nfsd, slp, lwp, mrq) 473 struct nfsrv_descript *nfsd; 474 struct nfssvc_sock *slp; 475 struct lwp *lwp; 476 struct mbuf **mrq; 477 { 478 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 479 struct mbuf *nam = nfsd->nd_nam; 480 caddr_t dpos = nfsd->nd_dpos; 481 kauth_cred_t cred = nfsd->nd_cr; 482 struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN]; 483 struct iovec *ivp = iv; 484 struct mbuf *mp; 485 u_int32_t *tl; 486 int32_t t1; 487 caddr_t bpos; 488 int error = 0, rdonly, cache, i, padlen, getret; 489 uint32_t len; 490 int v3 = (nfsd->nd_flag & ND_NFSV3); 491 char *cp2; 492 struct mbuf *mb, *mp2 = NULL, *mp3 = NULL, *mreq; 493 struct vnode *vp; 494 struct vattr attr; 495 nfsfh_t nfh; 496 fhandle_t *fhp; 497 struct uio io, *uiop = &io; 498 u_quad_t frev; 499 500 fhp = &nfh.fh_generic; 501 nfsm_srvmtofh(fhp); 502 len = 0; 503 i = 0; 504 while (len < NFS_MAXPATHLEN) { 505 mp = m_get(M_WAIT, MT_DATA); 506 MCLAIM(mp, &nfs_mowner); 507 m_clget(mp, M_WAIT); 508 mp->m_len = NFSMSIZ(mp); 509 if (len == 0) 510 mp3 = mp2 = mp; 511 else { 512 mp2->m_next = mp; 513 mp2 = mp; 514 } 515 if ((len+mp->m_len) > NFS_MAXPATHLEN) { 516 mp->m_len = NFS_MAXPATHLEN-len; 517 len = NFS_MAXPATHLEN; 518 } else 519 len += mp->m_len; 520 ivp->iov_base = mtod(mp, caddr_t); 521 ivp->iov_len = mp->m_len; 522 i++; 523 ivp++; 524 } 525 uiop->uio_iov = iv; 526 uiop->uio_iovcnt = i; 527 uiop->uio_offset = 0; 528 uiop->uio_resid = len; 529 uiop->uio_rw = UIO_READ; 530 UIO_SETUP_SYSSPACE(uiop); 531 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 532 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 533 if (error) { 534 m_freem(mp3); 535 nfsm_reply(2 * NFSX_UNSIGNED); 536 nfsm_srvpostop_attr(1, (struct vattr *)0); 537 return (0); 538 } 539 if (vp->v_type != VLNK) { 540 if (v3) 541 error = EINVAL; 542 else 543 error = ENXIO; 544 goto out; 545 } 546 nqsrv_getl(vp, ND_READ); 547 error = VOP_READLINK(vp, uiop, cred); 548 out: 549 getret = VOP_GETATTR(vp, &attr, cred, lwp); 550 vput(vp); 551 if (error) 552 m_freem(mp3); 553 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED); 554 if (v3) { 555 nfsm_srvpostop_attr(getret, &attr); 556 if (error) 557 return (0); 558 } 559 len -= uiop->uio_resid; 560 padlen = nfsm_padlen(len); 561 if (uiop->uio_resid || padlen) 562 nfs_zeropad(mp3, uiop->uio_resid, padlen); 563 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 564 *tl = txdr_unsigned(len); 565 mb->m_next = mp3; 566 nfsm_srvdone; 567 } 568 569 /* 570 * nfs read service 571 */ 572 int 573 nfsrv_read(nfsd, slp, lwp, mrq) 574 struct nfsrv_descript *nfsd; 575 struct nfssvc_sock *slp; 576 struct lwp *lwp; 577 struct mbuf **mrq; 578 { 579 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 580 struct mbuf *nam = nfsd->nd_nam; 581 caddr_t dpos = nfsd->nd_dpos; 582 kauth_cred_t cred = nfsd->nd_cr; 583 struct mbuf *m; 584 struct nfs_fattr *fp; 585 u_int32_t *tl; 586 int32_t t1; 587 int i; 588 caddr_t bpos; 589 int error = 0, rdonly, cache, getret; 590 int v3 = (nfsd->nd_flag & ND_NFSV3); 591 uint32_t reqlen, len, cnt, left; 592 int padlen; 593 char *cp2; 594 struct mbuf *mb, *mreq; 595 struct vnode *vp; 596 nfsfh_t nfh; 597 fhandle_t *fhp; 598 struct uio io, *uiop = &io; 599 struct vattr va; 600 off_t off; 601 u_quad_t frev; 602 603 fhp = &nfh.fh_generic; 604 nfsm_srvmtofh(fhp); 605 if (v3) { 606 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 607 off = fxdr_hyper(tl); 608 } else { 609 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 610 off = (off_t)fxdr_unsigned(u_int32_t, *tl); 611 } 612 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 613 reqlen = fxdr_unsigned(uint32_t, *tl); 614 reqlen = MIN(reqlen, NFS_SRVMAXDATA(nfsd)); 615 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 616 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 617 if (error) { 618 nfsm_reply(2 * NFSX_UNSIGNED); 619 nfsm_srvpostop_attr(1, (struct vattr *)0); 620 return (0); 621 } 622 if (vp->v_type != VREG) { 623 if (v3) 624 error = EINVAL; 625 else 626 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 627 } 628 if (!error) { 629 nqsrv_getl(vp, ND_READ); 630 if ((error = nfsrv_access(vp, VREAD, cred, rdonly, lwp, 1)) != 0) 631 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 1); 632 } 633 getret = VOP_GETATTR(vp, &va, cred, lwp); 634 if (!error) 635 error = getret; 636 if (error) { 637 vput(vp); 638 nfsm_reply(NFSX_POSTOPATTR(v3)); 639 nfsm_srvpostop_attr(getret, &va); 640 return (0); 641 } 642 if (off >= va.va_size) 643 cnt = 0; 644 else if ((off + reqlen) > va.va_size) 645 cnt = va.va_size - off; 646 else 647 cnt = reqlen; 648 nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt)); 649 if (v3) { 650 nfsm_build(tl, u_int32_t *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED); 651 *tl++ = nfs_true; 652 fp = (struct nfs_fattr *)tl; 653 tl += (NFSX_V3FATTR / sizeof (u_int32_t)); 654 } else { 655 nfsm_build(tl, u_int32_t *, NFSX_V2FATTR + NFSX_UNSIGNED); 656 fp = (struct nfs_fattr *)tl; 657 tl += (NFSX_V2FATTR / sizeof (u_int32_t)); 658 } 659 len = left = cnt; 660 if (cnt > 0) { 661 if (nfsd_use_loan) { 662 struct vm_page **pgpp; 663 voff_t pgoff = trunc_page(off); 664 int npages; 665 vaddr_t lva; 666 667 npages = (round_page(off + cnt) - pgoff) >> PAGE_SHIFT; 668 KASSERT(npages <= M_EXT_MAXPAGES); /* XXX */ 669 670 /* allocate kva for mbuf data */ 671 lva = sokvaalloc(npages << PAGE_SHIFT, slp->ns_so); 672 if (lva == 0) { 673 /* fall back to VOP_READ */ 674 goto loan_fail; 675 } 676 677 /* allocate mbuf */ 678 m = m_get(M_WAIT, MT_DATA); 679 MCLAIM(m, &nfs_mowner); 680 pgpp = m->m_ext.ext_pgs; 681 682 /* loan pages */ 683 error = uvm_loanuobjpages(&vp->v_uobj, pgoff, npages, 684 pgpp); 685 if (error) { 686 sokvafree(lva, npages << PAGE_SHIFT); 687 m_free(m); 688 if (error == EBUSY) 689 goto loan_fail; 690 goto read_error; 691 } 692 693 /* associate kva to mbuf */ 694 MEXTADD(m, (void *)(lva + ((vaddr_t)off & PAGE_MASK)), 695 cnt, M_MBUF, soloanfree, slp->ns_so); 696 m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP; 697 m->m_len = cnt; 698 699 /* map pages */ 700 for (i = 0; i < npages; i++) { 701 pmap_kenter_pa(lva, VM_PAGE_TO_PHYS(pgpp[i]), 702 VM_PROT_READ); 703 lva += PAGE_SIZE; 704 } 705 706 pmap_update(pmap_kernel()); 707 708 mb->m_next = m; 709 mb = m; 710 error = 0; 711 uiop->uio_resid = 0; 712 } else { 713 struct iovec *iv; 714 struct iovec *iv2; 715 struct mbuf *m2; 716 int siz; 717 loan_fail: 718 /* 719 * Generate the mbuf list with the uio_iov ref. to it. 720 */ 721 i = 0; 722 m = m2 = mb; 723 while (left > 0) { 724 siz = min(M_TRAILINGSPACE(m), left); 725 if (siz > 0) { 726 left -= siz; 727 i++; 728 } 729 if (left > 0) { 730 m = m_get(M_WAIT, MT_DATA); 731 MCLAIM(m, &nfs_mowner); 732 m_clget(m, M_WAIT); 733 m->m_len = 0; 734 m2->m_next = m; 735 m2 = m; 736 } 737 } 738 iv = malloc(i * sizeof(struct iovec), M_TEMP, M_WAITOK); 739 uiop->uio_iov = iv2 = iv; 740 m = mb; 741 left = cnt; 742 i = 0; 743 while (left > 0) { 744 if (m == NULL) 745 panic("nfsrv_read iov"); 746 siz = min(M_TRAILINGSPACE(m), left); 747 if (siz > 0) { 748 iv->iov_base = mtod(m, caddr_t) + 749 m->m_len; 750 iv->iov_len = siz; 751 m->m_len += siz; 752 left -= siz; 753 iv++; 754 i++; 755 } 756 m = m->m_next; 757 } 758 uiop->uio_iovcnt = i; 759 uiop->uio_offset = off; 760 uiop->uio_resid = cnt; 761 uiop->uio_rw = UIO_READ; 762 UIO_SETUP_SYSSPACE(uiop); 763 error = VOP_READ(vp, uiop, IO_NODELOCKED, cred); 764 free((caddr_t)iv2, M_TEMP); 765 } 766 read_error: 767 if (error || (getret = VOP_GETATTR(vp, &va, cred, lwp)) != 0){ 768 if (!error) 769 error = getret; 770 m_freem(mreq); 771 vput(vp); 772 nfsm_reply(NFSX_POSTOPATTR(v3)); 773 nfsm_srvpostop_attr(getret, &va); 774 return (0); 775 } 776 } else { 777 uiop->uio_resid = 0; 778 } 779 vput(vp); 780 nfsm_srvfillattr(&va, fp); 781 len -= uiop->uio_resid; 782 padlen = nfsm_padlen(len); 783 if (uiop->uio_resid || padlen) 784 nfs_zeropad(mb, uiop->uio_resid, padlen); 785 if (v3) { 786 /* count */ 787 *tl++ = txdr_unsigned(len); 788 /* eof */ 789 if (off + len >= va.va_size) 790 *tl++ = nfs_true; 791 else 792 *tl++ = nfs_false; 793 } 794 *tl = txdr_unsigned(len); 795 nfsm_srvdone; 796 } 797 798 /* 799 * nfs write service 800 */ 801 int 802 nfsrv_write(nfsd, slp, lwp, mrq) 803 struct nfsrv_descript *nfsd; 804 struct nfssvc_sock *slp; 805 struct lwp *lwp; 806 struct mbuf **mrq; 807 { 808 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 809 struct mbuf *nam = nfsd->nd_nam; 810 caddr_t dpos = nfsd->nd_dpos; 811 kauth_cred_t cred = nfsd->nd_cr; 812 struct iovec *ivp; 813 int i, cnt; 814 struct mbuf *mp; 815 struct nfs_fattr *fp; 816 struct iovec *iv; 817 struct vattr va, forat; 818 u_int32_t *tl; 819 int32_t t1; 820 caddr_t bpos; 821 int error = 0, rdonly, cache, len, forat_ret = 1; 822 int ioflags, aftat_ret = 1, retlen, zeroing, adjust; 823 int stable = NFSV3WRITE_FILESYNC; 824 int v3 = (nfsd->nd_flag & ND_NFSV3); 825 char *cp2; 826 struct mbuf *mb, *mreq; 827 struct vnode *vp; 828 nfsfh_t nfh; 829 fhandle_t *fhp; 830 struct uio io, *uiop = &io; 831 off_t off; 832 u_quad_t frev; 833 struct mount *mntp = NULL; 834 835 if (mrep == NULL) { 836 *mrq = NULL; 837 return (0); 838 } 839 fhp = &nfh.fh_generic; 840 nfsm_srvmtofh(fhp); 841 if ((mntp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 842 return (ESTALE); 843 vn_start_write(NULL, &mntp, V_WAIT); 844 if (v3) { 845 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 846 off = fxdr_hyper(tl); 847 tl += 3; 848 stable = fxdr_unsigned(int, *tl++); 849 } else { 850 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 851 off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 852 tl += 2; 853 } 854 retlen = len = fxdr_unsigned(int32_t, *tl); 855 cnt = i = 0; 856 857 /* 858 * For NFS Version 2, it is not obvious what a write of zero length 859 * should do, but I might as well be consistent with Version 3, 860 * which is to return ok so long as there are no permission problems. 861 */ 862 if (len > 0) { 863 zeroing = 1; 864 mp = mrep; 865 while (mp) { 866 if (mp == md) { 867 zeroing = 0; 868 adjust = dpos - mtod(mp, caddr_t); 869 mp->m_len -= adjust; 870 if (mp->m_len > 0 && adjust > 0) 871 NFSMADV(mp, adjust); 872 } 873 if (zeroing) 874 mp->m_len = 0; 875 else if (mp->m_len > 0) { 876 i += mp->m_len; 877 if (i > len) { 878 mp->m_len -= (i - len); 879 zeroing = 1; 880 } 881 if (mp->m_len > 0) 882 cnt++; 883 } 884 mp = mp->m_next; 885 } 886 } 887 if (len > NFS_MAXDATA || len < 0 || i < len) { 888 error = EIO; 889 nfsm_reply(2 * NFSX_UNSIGNED); 890 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 891 vn_finished_write(mntp, 0); 892 return (0); 893 } 894 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 895 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 896 if (error) { 897 nfsm_reply(2 * NFSX_UNSIGNED); 898 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 899 vn_finished_write(mntp, 0); 900 return (0); 901 } 902 if (v3) 903 forat_ret = VOP_GETATTR(vp, &forat, cred, lwp); 904 if (vp->v_type != VREG) { 905 if (v3) 906 error = EINVAL; 907 else 908 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 909 } 910 if (!error) { 911 nqsrv_getl(vp, ND_WRITE); 912 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1); 913 } 914 if (error) { 915 vput(vp); 916 nfsm_reply(NFSX_WCCDATA(v3)); 917 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 918 vn_finished_write(mntp, 0); 919 return (0); 920 } 921 922 if (len > 0) { 923 ivp = malloc(cnt * sizeof (struct iovec), M_TEMP, M_WAITOK); 924 uiop->uio_iov = iv = ivp; 925 uiop->uio_iovcnt = cnt; 926 mp = mrep; 927 while (mp) { 928 if (mp->m_len > 0) { 929 ivp->iov_base = mtod(mp, caddr_t); 930 ivp->iov_len = mp->m_len; 931 ivp++; 932 } 933 mp = mp->m_next; 934 } 935 936 /* 937 * XXX 938 * The IO_METASYNC flag indicates that all metadata (and not 939 * just enough to ensure data integrity) must be written to 940 * stable storage synchronously. 941 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.) 942 */ 943 if (stable == NFSV3WRITE_UNSTABLE) 944 ioflags = IO_NODELOCKED; 945 else if (stable == NFSV3WRITE_DATASYNC) 946 ioflags = (IO_SYNC | IO_NODELOCKED); 947 else 948 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 949 uiop->uio_resid = len; 950 uiop->uio_rw = UIO_WRITE; 951 uiop->uio_offset = off; 952 UIO_SETUP_SYSSPACE(uiop); 953 error = VOP_WRITE(vp, uiop, ioflags, cred); 954 nfsstats.srvvop_writes++; 955 free(iv, M_TEMP); 956 } 957 aftat_ret = VOP_GETATTR(vp, &va, cred, lwp); 958 vput(vp); 959 vn_finished_write(mntp, 0); 960 if (!error) 961 error = aftat_ret; 962 nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) + 963 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3)); 964 if (v3) { 965 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 966 if (error) 967 return (0); 968 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 969 *tl++ = txdr_unsigned(retlen); 970 if (stable == NFSV3WRITE_UNSTABLE) 971 *tl++ = txdr_unsigned(stable); 972 else 973 *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC); 974 /* 975 * Actually, there is no need to txdr these fields, 976 * but it may make the values more human readable, 977 * for debugging purposes. 978 */ 979 *tl++ = txdr_unsigned(boottime.tv_sec); 980 *tl = txdr_unsigned(boottime.tv_usec); 981 } else { 982 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 983 nfsm_srvfillattr(&va, fp); 984 } 985 nfsm_srvdone; 986 } 987 988 /* 989 * NFS write service with write gathering support. Called when 990 * nfsrvw_procrastinate > 0. 991 * See: Chet Juszczak, "Improving the Write Performance of an NFS Server", 992 * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco, 993 * Jan. 1994. 994 */ 995 int 996 nfsrv_writegather(ndp, slp, lwp, mrq) 997 struct nfsrv_descript **ndp; 998 struct nfssvc_sock *slp; 999 struct lwp *lwp; 1000 struct mbuf **mrq; 1001 { 1002 struct iovec *ivp; 1003 struct mbuf *mp; 1004 struct nfsrv_descript *wp, *nfsd, *owp, *swp; 1005 struct nfs_fattr *fp; 1006 int i = 0; 1007 struct iovec *iov; 1008 struct nfsrvw_delayhash *wpp; 1009 kauth_cred_t cred; 1010 struct vattr va, forat; 1011 u_int32_t *tl; 1012 int32_t t1; 1013 caddr_t bpos, dpos; 1014 int error = 0, rdonly, cache, len = 0, forat_ret = 1; 1015 int ioflags, aftat_ret = 1, s, adjust, v3, zeroing; 1016 char *cp2; 1017 struct mbuf *mb, *mreq, *mrep, *md; 1018 struct vnode *vp; 1019 struct uio io, *uiop = &io; 1020 u_quad_t frev, cur_usec; 1021 struct mount *mntp = NULL; 1022 1023 *mrq = NULL; 1024 if (*ndp) { 1025 nfsd = *ndp; 1026 *ndp = NULL; 1027 mrep = nfsd->nd_mrep; 1028 md = nfsd->nd_md; 1029 dpos = nfsd->nd_dpos; 1030 cred = nfsd->nd_cr; 1031 v3 = (nfsd->nd_flag & ND_NFSV3); 1032 LIST_INIT(&nfsd->nd_coalesce); 1033 nfsd->nd_mreq = NULL; 1034 nfsd->nd_stable = NFSV3WRITE_FILESYNC; 1035 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec; 1036 nfsd->nd_time = cur_usec + nfsrvw_procrastinate; 1037 1038 /* 1039 * Now, get the write header.. 1040 */ 1041 nfsm_srvmtofh(&nfsd->nd_fh); 1042 if (v3) { 1043 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 1044 nfsd->nd_off = fxdr_hyper(tl); 1045 tl += 3; 1046 nfsd->nd_stable = fxdr_unsigned(int, *tl++); 1047 } else { 1048 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1049 nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 1050 tl += 2; 1051 } 1052 len = fxdr_unsigned(int32_t, *tl); 1053 nfsd->nd_len = len; 1054 nfsd->nd_eoff = nfsd->nd_off + len; 1055 1056 /* 1057 * Trim the header out of the mbuf list and trim off any trailing 1058 * junk so that the mbuf list has only the write data. 1059 */ 1060 zeroing = 1; 1061 i = 0; 1062 mp = mrep; 1063 while (mp) { 1064 if (mp == md) { 1065 zeroing = 0; 1066 adjust = dpos - mtod(mp, caddr_t); 1067 mp->m_len -= adjust; 1068 if (mp->m_len > 0 && adjust > 0) 1069 NFSMADV(mp, adjust); 1070 } 1071 if (zeroing) 1072 mp->m_len = 0; 1073 else { 1074 i += mp->m_len; 1075 if (i > len) { 1076 mp->m_len -= (i - len); 1077 zeroing = 1; 1078 } 1079 } 1080 mp = mp->m_next; 1081 } 1082 if (len > NFS_MAXDATA || len < 0 || i < len) { 1083 nfsmout: 1084 m_freem(mrep); 1085 error = EIO; 1086 nfsm_writereply(2 * NFSX_UNSIGNED, v3); 1087 if (v3) 1088 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1089 nfsd->nd_mreq = mreq; 1090 nfsd->nd_mrep = NULL; 1091 nfsd->nd_time = 0; 1092 } 1093 1094 /* 1095 * Add this entry to the hash and time queues. 1096 */ 1097 s = splsoftclock(); 1098 owp = NULL; 1099 wp = LIST_FIRST(&slp->ns_tq); 1100 while (wp && wp->nd_time < nfsd->nd_time) { 1101 owp = wp; 1102 wp = LIST_NEXT(wp, nd_tq); 1103 } 1104 if (owp) { 1105 LIST_INSERT_AFTER(owp, nfsd, nd_tq); 1106 } else { 1107 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1108 } 1109 if (nfsd->nd_mrep) { 1110 wpp = NWDELAYHASH(slp, nfsd->nd_fh.fh_fid.fid_data); 1111 owp = NULL; 1112 wp = LIST_FIRST(wpp); 1113 while (wp && 1114 memcmp(&nfsd->nd_fh, &wp->nd_fh, NFSX_V3FH)) { 1115 owp = wp; 1116 wp = LIST_NEXT(wp, nd_hash); 1117 } 1118 while (wp && wp->nd_off < nfsd->nd_off && 1119 !memcmp(&nfsd->nd_fh, &wp->nd_fh, NFSX_V3FH)) { 1120 owp = wp; 1121 wp = LIST_NEXT(wp, nd_hash); 1122 } 1123 if (owp) { 1124 LIST_INSERT_AFTER(owp, nfsd, nd_hash); 1125 1126 /* 1127 * Search the hash list for overlapping entries and 1128 * coalesce. 1129 */ 1130 for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) { 1131 wp = LIST_NEXT(nfsd, nd_hash); 1132 if (NFSW_SAMECRED(owp, nfsd)) 1133 nfsrvw_coalesce(owp, nfsd); 1134 } 1135 } else { 1136 LIST_INSERT_HEAD(wpp, nfsd, nd_hash); 1137 } 1138 } 1139 splx(s); 1140 } 1141 1142 /* 1143 * Now, do VOP_WRITE()s for any one(s) that need to be done now 1144 * and generate the associated reply mbuf list(s). 1145 */ 1146 loop1: 1147 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec; 1148 s = splsoftclock(); 1149 for (nfsd = LIST_FIRST(&slp->ns_tq); nfsd; nfsd = owp) { 1150 owp = LIST_NEXT(nfsd, nd_tq); 1151 if (nfsd->nd_time > cur_usec) 1152 break; 1153 if (nfsd->nd_mreq) 1154 continue; 1155 LIST_REMOVE(nfsd, nd_tq); 1156 LIST_REMOVE(nfsd, nd_hash); 1157 splx(s); 1158 mrep = nfsd->nd_mrep; 1159 nfsd->nd_mrep = NULL; 1160 cred = nfsd->nd_cr; 1161 v3 = (nfsd->nd_flag & ND_NFSV3); 1162 forat_ret = aftat_ret = 1; 1163 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp, 1164 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH), 1165 FALSE); 1166 if (!error) { 1167 if (v3) 1168 forat_ret = VOP_GETATTR(vp, &forat, cred, lwp); 1169 if (vp->v_type != VREG) { 1170 if (v3) 1171 error = EINVAL; 1172 else 1173 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 1174 } 1175 } else 1176 vp = NULL; 1177 if (!error) { 1178 nqsrv_getl(vp, ND_WRITE); 1179 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1); 1180 } 1181 1182 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE) 1183 ioflags = IO_NODELOCKED; 1184 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC) 1185 ioflags = (IO_SYNC | IO_NODELOCKED); 1186 else 1187 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 1188 uiop->uio_rw = UIO_WRITE; 1189 uiop->uio_offset = nfsd->nd_off; 1190 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off; 1191 UIO_SETUP_SYSSPACE(uiop); 1192 if (uiop->uio_resid > 0) { 1193 mp = mrep; 1194 i = 0; 1195 while (mp) { 1196 if (mp->m_len > 0) 1197 i++; 1198 mp = mp->m_next; 1199 } 1200 uiop->uio_iovcnt = i; 1201 iov = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK); 1202 uiop->uio_iov = ivp = iov; 1203 mp = mrep; 1204 while (mp) { 1205 if (mp->m_len > 0) { 1206 ivp->iov_base = mtod(mp, caddr_t); 1207 ivp->iov_len = mp->m_len; 1208 ivp++; 1209 } 1210 mp = mp->m_next; 1211 } 1212 if (!error) { 1213 if (vn_start_write(vp, &mntp, V_NOWAIT) != 0) { 1214 VOP_UNLOCK(vp, 0); 1215 vn_start_write(NULL, &mntp, V_WAIT); 1216 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1217 } 1218 if (!error) { 1219 error = VOP_WRITE(vp, uiop, ioflags, cred); 1220 nfsstats.srvvop_writes++; 1221 vn_finished_write(mntp, 0); 1222 } 1223 } 1224 free((caddr_t)iov, M_TEMP); 1225 } 1226 m_freem(mrep); 1227 if (vp) { 1228 aftat_ret = VOP_GETATTR(vp, &va, cred, lwp); 1229 vput(vp); 1230 } 1231 1232 /* 1233 * Loop around generating replies for all write rpcs that have 1234 * now been completed. 1235 */ 1236 swp = nfsd; 1237 do { 1238 if (error) { 1239 nfsm_writereply(NFSX_WCCDATA(v3), v3); 1240 if (v3) { 1241 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1242 } 1243 } else { 1244 nfsm_writereply(NFSX_PREOPATTR(v3) + 1245 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED + 1246 NFSX_WRITEVERF(v3), v3); 1247 if (v3) { 1248 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1249 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1250 *tl++ = txdr_unsigned(nfsd->nd_len); 1251 *tl++ = txdr_unsigned(swp->nd_stable); 1252 /* 1253 * Actually, there is no need to txdr these fields, 1254 * but it may make the values more human readable, 1255 * for debugging purposes. 1256 */ 1257 *tl++ = txdr_unsigned(boottime.tv_sec); 1258 *tl = txdr_unsigned(boottime.tv_usec); 1259 } else { 1260 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1261 nfsm_srvfillattr(&va, fp); 1262 } 1263 } 1264 nfsd->nd_mreq = mreq; 1265 if (nfsd->nd_mrep) 1266 panic("nfsrv_write: nd_mrep not free"); 1267 1268 /* 1269 * Done. Put it at the head of the timer queue so that 1270 * the final phase can return the reply. 1271 */ 1272 s = splsoftclock(); 1273 if (nfsd != swp) { 1274 nfsd->nd_time = 0; 1275 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1276 } 1277 nfsd = LIST_FIRST(&swp->nd_coalesce); 1278 if (nfsd) { 1279 LIST_REMOVE(nfsd, nd_tq); 1280 } 1281 splx(s); 1282 } while (nfsd); 1283 s = splsoftclock(); 1284 swp->nd_time = 0; 1285 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq); 1286 splx(s); 1287 goto loop1; 1288 } 1289 splx(s); 1290 1291 /* 1292 * Search for a reply to return. 1293 */ 1294 s = splsoftclock(); 1295 LIST_FOREACH(nfsd, &slp->ns_tq, nd_tq) { 1296 if (nfsd->nd_mreq) { 1297 LIST_REMOVE(nfsd, nd_tq); 1298 *mrq = nfsd->nd_mreq; 1299 *ndp = nfsd; 1300 break; 1301 } 1302 } 1303 splx(s); 1304 return (0); 1305 } 1306 1307 /* 1308 * Coalesce the write request nfsd into owp. To do this we must: 1309 * - remove nfsd from the queues 1310 * - merge nfsd->nd_mrep into owp->nd_mrep 1311 * - update the nd_eoff and nd_stable for owp 1312 * - put nfsd on owp's nd_coalesce list 1313 * NB: Must be called at splsoftclock(). 1314 */ 1315 void 1316 nfsrvw_coalesce(owp, nfsd) 1317 struct nfsrv_descript *owp; 1318 struct nfsrv_descript *nfsd; 1319 { 1320 int overlap; 1321 struct mbuf *mp; 1322 struct nfsrv_descript *m; 1323 1324 LIST_REMOVE(nfsd, nd_hash); 1325 LIST_REMOVE(nfsd, nd_tq); 1326 if (owp->nd_eoff < nfsd->nd_eoff) { 1327 overlap = owp->nd_eoff - nfsd->nd_off; 1328 if (overlap < 0) 1329 panic("nfsrv_coalesce: bad off"); 1330 if (overlap > 0) 1331 m_adj(nfsd->nd_mrep, overlap); 1332 mp = owp->nd_mrep; 1333 while (mp->m_next) 1334 mp = mp->m_next; 1335 mp->m_next = nfsd->nd_mrep; 1336 owp->nd_eoff = nfsd->nd_eoff; 1337 } else 1338 m_freem(nfsd->nd_mrep); 1339 nfsd->nd_mrep = NULL; 1340 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC) 1341 owp->nd_stable = NFSV3WRITE_FILESYNC; 1342 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC && 1343 owp->nd_stable == NFSV3WRITE_UNSTABLE) 1344 owp->nd_stable = NFSV3WRITE_DATASYNC; 1345 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq); 1346 /* 1347 * nfsd might hold coalesce elements! Move them to owp. 1348 * Otherwise, requests may be lost and clients will be stuck. 1349 */ 1350 while ((m = LIST_FIRST(&nfsd->nd_coalesce)) != NULL) { 1351 LIST_REMOVE(m, nd_tq); 1352 LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq); 1353 } 1354 } 1355 1356 /* 1357 * nfs create service 1358 * now does a truncate to 0 length via. setattr if it already exists 1359 */ 1360 int 1361 nfsrv_create(nfsd, slp, lwp, mrq) 1362 struct nfsrv_descript *nfsd; 1363 struct nfssvc_sock *slp; 1364 struct lwp *lwp; 1365 struct mbuf **mrq; 1366 { 1367 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1368 struct mbuf *nam = nfsd->nd_nam; 1369 caddr_t dpos = nfsd->nd_dpos; 1370 kauth_cred_t cred = nfsd->nd_cr; 1371 struct nfs_fattr *fp; 1372 struct vattr va, dirfor, diraft; 1373 struct nfsv2_sattr *sp; 1374 u_int32_t *tl; 1375 struct nameidata nd; 1376 caddr_t cp; 1377 int32_t t1; 1378 caddr_t bpos; 1379 int error = 0, cache, len, tsize, dirfor_ret = 1, diraft_ret = 1; 1380 int rdev = 0; 1381 int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0; 1382 char *cp2; 1383 struct mbuf *mb, *mreq; 1384 struct vnode *vp = NULL, *dirp = NULL; 1385 nfsfh_t nfh; 1386 fhandle_t *fhp; 1387 u_quad_t frev, tempsize; 1388 u_char cverf[NFSX_V3CREATEVERF]; 1389 struct mount *mp = NULL; 1390 1391 nd.ni_cnd.cn_nameiop = 0; 1392 fhp = &nfh.fh_generic; 1393 nfsm_srvmtofh(fhp); 1394 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1395 return (ESTALE); 1396 vn_start_write(NULL, &mp, V_WAIT); 1397 nfsm_srvnamesiz(len); 1398 nd.ni_cnd.cn_cred = cred; 1399 nd.ni_cnd.cn_nameiop = CREATE; 1400 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1401 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1402 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1403 if (dirp) { 1404 if (v3) 1405 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1406 else { 1407 vrele(dirp); 1408 dirp = (struct vnode *)0; 1409 } 1410 } 1411 if (error) { 1412 nfsm_reply(NFSX_WCCDATA(v3)); 1413 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1414 if (dirp) 1415 vrele(dirp); 1416 vn_finished_write(mp, 0); 1417 return (0); 1418 } 1419 VATTR_NULL(&va); 1420 if (v3) { 1421 va.va_mode = 0; 1422 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1423 how = fxdr_unsigned(int, *tl); 1424 switch (how) { 1425 case NFSV3CREATE_GUARDED: 1426 if (nd.ni_vp) { 1427 error = EEXIST; 1428 break; 1429 } 1430 case NFSV3CREATE_UNCHECKED: 1431 nfsm_srvsattr(&va); 1432 break; 1433 case NFSV3CREATE_EXCLUSIVE: 1434 nfsm_dissect(cp, caddr_t, NFSX_V3CREATEVERF); 1435 memcpy(cverf, cp, NFSX_V3CREATEVERF); 1436 exclusive_flag = 1; 1437 break; 1438 }; 1439 va.va_type = VREG; 1440 } else { 1441 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 1442 va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode)); 1443 if (va.va_type == VNON) 1444 va.va_type = VREG; 1445 va.va_mode = nfstov_mode(sp->sa_mode); 1446 switch (va.va_type) { 1447 case VREG: 1448 tsize = fxdr_unsigned(int32_t, sp->sa_size); 1449 if (tsize != -1) 1450 va.va_size = (u_quad_t)tsize; 1451 break; 1452 case VCHR: 1453 case VBLK: 1454 case VFIFO: 1455 rdev = fxdr_unsigned(int32_t, sp->sa_size); 1456 break; 1457 default: 1458 break; 1459 }; 1460 } 1461 1462 /* 1463 * Iff doesn't exist, create it 1464 * otherwise just truncate to 0 length 1465 * should I set the mode too ?? 1466 */ 1467 if (nd.ni_vp == NULL) { 1468 if (va.va_type == VREG || va.va_type == VSOCK) { 1469 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1470 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1471 if (!error) { 1472 if (exclusive_flag) { 1473 exclusive_flag = 0; 1474 VATTR_NULL(&va); 1475 /* 1476 * XXX 1477 * assuming NFSX_V3CREATEVERF 1478 * == sizeof(nfstime3) 1479 */ 1480 fxdr_nfsv3time(cverf, &va.va_atime); 1481 error = VOP_SETATTR(nd.ni_vp, &va, cred, 1482 lwp); 1483 } 1484 } 1485 } else if (va.va_type == VCHR || va.va_type == VBLK || 1486 va.va_type == VFIFO) { 1487 if (va.va_type == VCHR && rdev == 0xffffffff) 1488 va.va_type = VFIFO; 1489 if (va.va_type != VFIFO && 1490 (error = kauth_authorize_generic(cred, 1491 KAUTH_GENERIC_ISSUSER, (u_short *)0))) { 1492 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1493 vput(nd.ni_dvp); 1494 nfsm_reply(0); 1495 vn_finished_write(mp, 0); 1496 return (error); 1497 } else 1498 va.va_rdev = (dev_t)rdev; 1499 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1500 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, 1501 &va); 1502 if (error) { 1503 nfsm_reply(0); 1504 } 1505 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1506 vrele(nd.ni_dvp); 1507 vput(nd.ni_vp); 1508 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1509 error = EINVAL; 1510 nfsm_reply(0); 1511 } 1512 } else { 1513 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1514 vput(nd.ni_dvp); 1515 error = ENXIO; 1516 } 1517 vp = nd.ni_vp; 1518 } else { 1519 vp = nd.ni_vp; 1520 if (nd.ni_dvp == vp) 1521 vrele(nd.ni_dvp); 1522 else 1523 vput(nd.ni_dvp); 1524 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1525 if (!error && va.va_size != -1) { 1526 error = nfsrv_access(vp, VWRITE, cred, 1527 (nd.ni_cnd.cn_flags & RDONLY), lwp, 0); 1528 if (!error) { 1529 nqsrv_getl(vp, ND_WRITE); 1530 tempsize = va.va_size; 1531 VATTR_NULL(&va); 1532 va.va_size = tempsize; 1533 error = VOP_SETATTR(vp, &va, cred, lwp); 1534 } 1535 } 1536 if (error) 1537 vput(vp); 1538 } 1539 if (!error) { 1540 memset((caddr_t)fhp, 0, sizeof(nfh)); 1541 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1542 error = VFS_VPTOFH(vp, &fhp->fh_fid); 1543 if (!error) 1544 error = VOP_GETATTR(vp, &va, cred, lwp); 1545 vput(vp); 1546 } 1547 if (v3) { 1548 if (exclusive_flag && !error) { 1549 /* 1550 * XXX assuming NFSX_V3CREATEVERF == sizeof(nfstime3) 1551 */ 1552 char oldverf[NFSX_V3CREATEVERF]; 1553 1554 txdr_nfsv3time(&va.va_atime, oldverf); 1555 if (memcmp(cverf, oldverf, NFSX_V3CREATEVERF)) 1556 error = EEXIST; 1557 } 1558 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1559 vrele(dirp); 1560 } 1561 nfsm_reply(NFSX_SRVFH(v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3)); 1562 if (v3) { 1563 if (!error) { 1564 nfsm_srvpostop_fh(fhp); 1565 nfsm_srvpostop_attr(0, &va); 1566 } 1567 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1568 } else { 1569 nfsm_srvfhtom(fhp, v3); 1570 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1571 nfsm_srvfillattr(&va, fp); 1572 } 1573 vn_finished_write(mp, 0); 1574 return (0); 1575 nfsmout: 1576 if (dirp) 1577 vrele(dirp); 1578 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1579 if (nd.ni_dvp == nd.ni_vp) 1580 vrele(nd.ni_dvp); 1581 else 1582 vput(nd.ni_dvp); 1583 if (nd.ni_vp) 1584 vput(nd.ni_vp); 1585 vn_finished_write(mp, 0); 1586 return (error); 1587 } 1588 1589 /* 1590 * nfs v3 mknod service 1591 */ 1592 int 1593 nfsrv_mknod(nfsd, slp, lwp, mrq) 1594 struct nfsrv_descript *nfsd; 1595 struct nfssvc_sock *slp; 1596 struct lwp *lwp; 1597 struct mbuf **mrq; 1598 { 1599 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1600 struct mbuf *nam = nfsd->nd_nam; 1601 caddr_t dpos = nfsd->nd_dpos; 1602 kauth_cred_t cred = nfsd->nd_cr; 1603 struct vattr va, dirfor, diraft; 1604 u_int32_t *tl; 1605 struct nameidata nd; 1606 int32_t t1; 1607 caddr_t bpos; 1608 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 1609 u_int32_t major, minor; 1610 enum vtype vtyp; 1611 char *cp2; 1612 struct mbuf *mb, *mreq; 1613 struct vnode *vp, *dirp = (struct vnode *)0; 1614 nfsfh_t nfh; 1615 fhandle_t *fhp; 1616 u_quad_t frev; 1617 struct mount *mp = NULL; 1618 1619 nd.ni_cnd.cn_nameiop = 0; 1620 fhp = &nfh.fh_generic; 1621 nfsm_srvmtofh(fhp); 1622 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1623 return (ESTALE); 1624 vn_start_write(NULL, &mp, V_WAIT); 1625 nfsm_srvnamesiz(len); 1626 nd.ni_cnd.cn_cred = cred; 1627 nd.ni_cnd.cn_nameiop = CREATE; 1628 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1629 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1630 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1631 if (dirp) 1632 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1633 if (error) { 1634 nfsm_reply(NFSX_WCCDATA(1)); 1635 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1636 if (dirp) 1637 vrele(dirp); 1638 vn_finished_write(mp, 0); 1639 return (0); 1640 } 1641 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1642 vtyp = nfsv3tov_type(*tl); 1643 if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) { 1644 error = NFSERR_BADTYPE; 1645 goto abort; 1646 } 1647 VATTR_NULL(&va); 1648 va.va_mode = 0; 1649 nfsm_srvsattr(&va); 1650 if (vtyp == VCHR || vtyp == VBLK) { 1651 dev_t rdev; 1652 1653 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1654 major = fxdr_unsigned(u_int32_t, *tl++); 1655 minor = fxdr_unsigned(u_int32_t, *tl); 1656 rdev = makedev(major, minor); 1657 if (major(rdev) != major || minor(rdev) != minor) { 1658 error = EINVAL; 1659 goto abort; 1660 } 1661 va.va_rdev = rdev; 1662 } 1663 1664 /* 1665 * Iff doesn't exist, create it. 1666 */ 1667 if (nd.ni_vp) { 1668 error = EEXIST; 1669 abort: 1670 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1671 if (nd.ni_dvp == nd.ni_vp) 1672 vrele(nd.ni_dvp); 1673 else 1674 vput(nd.ni_dvp); 1675 if (nd.ni_vp) 1676 vput(nd.ni_vp); 1677 goto out; 1678 } 1679 va.va_type = vtyp; 1680 if (vtyp == VSOCK) { 1681 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1682 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1683 } else { 1684 if (va.va_type != VFIFO && 1685 (error = kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER, 1686 (u_short *)0))) { 1687 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1688 vput(nd.ni_dvp); 1689 goto out; 1690 } 1691 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1692 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1693 if (error) { 1694 goto out; 1695 } 1696 if (error) 1697 goto out; 1698 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1699 vput(nd.ni_vp); 1700 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1701 error = EINVAL; 1702 } 1703 } 1704 out: 1705 vp = nd.ni_vp; 1706 if (!error) { 1707 memset((caddr_t)fhp, 0, sizeof(nfh)); 1708 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1709 error = VFS_VPTOFH(vp, &fhp->fh_fid); 1710 if (!error) 1711 error = VOP_GETATTR(vp, &va, cred, lwp); 1712 vput(vp); 1713 } 1714 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1715 vrele(dirp); 1716 nfsm_reply(NFSX_SRVFH(1) + NFSX_POSTOPATTR(1) + NFSX_WCCDATA(1)); 1717 if (!error) { 1718 nfsm_srvpostop_fh(fhp); 1719 nfsm_srvpostop_attr(0, &va); 1720 } 1721 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1722 vn_finished_write(mp, 0); 1723 return (0); 1724 nfsmout: 1725 if (dirp) 1726 vrele(dirp); 1727 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1728 if (nd.ni_dvp == nd.ni_vp) 1729 vrele(nd.ni_dvp); 1730 else 1731 vput(nd.ni_dvp); 1732 if (nd.ni_vp) 1733 vput(nd.ni_vp); 1734 vn_finished_write(mp, 0); 1735 return (error); 1736 } 1737 1738 /* 1739 * nfs remove service 1740 */ 1741 int 1742 nfsrv_remove(nfsd, slp, lwp, mrq) 1743 struct nfsrv_descript *nfsd; 1744 struct nfssvc_sock *slp; 1745 struct lwp *lwp; 1746 struct mbuf **mrq; 1747 { 1748 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1749 struct mbuf *nam = nfsd->nd_nam; 1750 caddr_t dpos = nfsd->nd_dpos; 1751 kauth_cred_t cred = nfsd->nd_cr; 1752 struct nameidata nd; 1753 u_int32_t *tl; 1754 int32_t t1; 1755 caddr_t bpos; 1756 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 1757 int v3 = (nfsd->nd_flag & ND_NFSV3); 1758 char *cp2; 1759 struct mbuf *mb, *mreq; 1760 struct vnode *vp, *dirp; 1761 struct vattr dirfor, diraft; 1762 nfsfh_t nfh; 1763 fhandle_t *fhp; 1764 u_quad_t frev; 1765 struct mount *mp = NULL; 1766 1767 #ifndef nolint 1768 vp = (struct vnode *)0; 1769 #endif 1770 fhp = &nfh.fh_generic; 1771 nfsm_srvmtofh(fhp); 1772 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 1773 return (ESTALE); 1774 vn_start_write(NULL, &mp, V_WAIT); 1775 nfsm_srvnamesiz(len); 1776 nd.ni_cnd.cn_cred = cred; 1777 nd.ni_cnd.cn_nameiop = DELETE; 1778 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1779 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 1780 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1781 if (dirp) { 1782 if (v3) 1783 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1784 else 1785 vrele(dirp); 1786 } 1787 if (!error) { 1788 vp = nd.ni_vp; 1789 if (vp->v_type == VDIR && 1790 (error = kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER, 1791 (u_short *)0)) != 0) 1792 goto out; 1793 /* 1794 * The root of a mounted filesystem cannot be deleted. 1795 */ 1796 if (vp->v_flag & VROOT) { 1797 error = EBUSY; 1798 goto out; 1799 } 1800 out: 1801 if (!error) { 1802 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1803 nqsrv_getl(vp, ND_WRITE); 1804 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1805 } else { 1806 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1807 if (nd.ni_dvp == vp) 1808 vrele(nd.ni_dvp); 1809 else 1810 vput(nd.ni_dvp); 1811 vput(vp); 1812 } 1813 } 1814 if (dirp && v3) { 1815 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1816 vrele(dirp); 1817 } 1818 nfsm_reply(NFSX_WCCDATA(v3)); 1819 if (v3) { 1820 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1821 vn_finished_write(mp, 0); 1822 return (0); 1823 } 1824 vn_finished_write(mp, 0); 1825 nfsm_srvdone; 1826 } 1827 1828 /* 1829 * nfs rename service 1830 */ 1831 int 1832 nfsrv_rename(nfsd, slp, lwp, mrq) 1833 struct nfsrv_descript *nfsd; 1834 struct nfssvc_sock *slp; 1835 struct lwp *lwp; 1836 struct mbuf **mrq; 1837 { 1838 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1839 struct mbuf *nam = nfsd->nd_nam; 1840 caddr_t dpos = nfsd->nd_dpos; 1841 kauth_cred_t cred = nfsd->nd_cr; 1842 u_int32_t *tl; 1843 int32_t t1; 1844 caddr_t bpos; 1845 int error = 0, cache, fdirfor_ret = 1, fdiraft_ret = 1; 1846 uint32_t len, len2; 1847 int tdirfor_ret = 1, tdiraft_ret = 1; 1848 int v3 = (nfsd->nd_flag & ND_NFSV3); 1849 char *cp2; 1850 struct mbuf *mb, *mreq; 1851 struct nameidata fromnd, tond; 1852 struct vnode *fvp, *tvp, *tdvp, *fdirp = (struct vnode *)0; 1853 struct vnode *tdirp = (struct vnode *)0; 1854 struct vattr fdirfor, fdiraft, tdirfor, tdiraft; 1855 nfsfh_t fnfh, tnfh; 1856 fhandle_t *ffhp, *tfhp; 1857 u_quad_t frev; 1858 uid_t saved_uid; 1859 struct mount *mp = NULL; 1860 1861 #ifndef nolint 1862 fvp = (struct vnode *)0; 1863 #endif 1864 ffhp = &fnfh.fh_generic; 1865 tfhp = &tnfh.fh_generic; 1866 fromnd.ni_cnd.cn_nameiop = 0; 1867 tond.ni_cnd.cn_nameiop = 0; 1868 nfsm_srvmtofh(ffhp); 1869 if ((mp = vfs_getvfs(&ffhp->fh_fsid)) == NULL) 1870 return (ESTALE); 1871 vn_start_write(NULL, &mp, V_WAIT); 1872 nfsm_srvnamesiz(len); 1873 /* 1874 * Remember our original uid so that we can reset cr_uid before 1875 * the second nfs_namei() call, in case it is remapped. 1876 */ 1877 saved_uid = kauth_cred_geteuid(cred); 1878 fromnd.ni_cnd.cn_cred = cred; 1879 fromnd.ni_cnd.cn_nameiop = DELETE; 1880 fromnd.ni_cnd.cn_flags = WANTPARENT | SAVESTART; 1881 error = nfs_namei(&fromnd, ffhp, len, slp, nam, &md, 1882 &dpos, &fdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1883 if (fdirp) { 1884 if (v3) 1885 fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred, lwp); 1886 else { 1887 vrele(fdirp); 1888 fdirp = (struct vnode *)0; 1889 } 1890 } 1891 if (error) { 1892 nfsm_reply(2 * NFSX_WCCDATA(v3)); 1893 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 1894 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 1895 if (fdirp) 1896 vrele(fdirp); 1897 vn_finished_write(mp, 0); 1898 return (0); 1899 } 1900 fvp = fromnd.ni_vp; 1901 nfsm_srvmtofh(tfhp); 1902 if (v3) { 1903 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 1904 len2 = fxdr_unsigned(uint32_t, *tl); 1905 /* len2 will be checked by nfs_namei */ 1906 } 1907 else { 1908 /* NFSv2 */ 1909 nfsm_strsiz(len2, NFS_MAXNAMLEN); 1910 } 1911 kauth_cred_seteuid(cred, saved_uid); 1912 tond.ni_cnd.cn_cred = cred; 1913 tond.ni_cnd.cn_nameiop = RENAME; 1914 tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART; 1915 error = nfs_namei(&tond, tfhp, len2, slp, nam, &md, 1916 &dpos, &tdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 1917 if (tdirp) { 1918 if (v3) 1919 tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred, lwp); 1920 else { 1921 vrele(tdirp); 1922 tdirp = (struct vnode *)0; 1923 } 1924 } 1925 if (error) { 1926 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1927 vrele(fromnd.ni_dvp); 1928 vrele(fvp); 1929 goto out1; 1930 } 1931 tdvp = tond.ni_dvp; 1932 tvp = tond.ni_vp; 1933 if (tvp != NULL) { 1934 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 1935 if (v3) 1936 error = EEXIST; 1937 else 1938 error = EISDIR; 1939 goto out; 1940 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 1941 if (v3) 1942 error = EEXIST; 1943 else 1944 error = ENOTDIR; 1945 goto out; 1946 } 1947 if (tvp->v_type == VDIR && tvp->v_mountedhere) { 1948 if (v3) 1949 error = EXDEV; 1950 else 1951 error = ENOTEMPTY; 1952 goto out; 1953 } 1954 } 1955 if (fvp->v_type == VDIR && fvp->v_mountedhere) { 1956 if (v3) 1957 error = EXDEV; 1958 else 1959 error = ENOTEMPTY; 1960 goto out; 1961 } 1962 if (fvp->v_mount != tdvp->v_mount) { 1963 if (v3) 1964 error = EXDEV; 1965 else 1966 error = ENOTEMPTY; 1967 goto out; 1968 } 1969 if (fvp == tdvp) { 1970 if (v3) 1971 error = EINVAL; 1972 else 1973 error = ENOTEMPTY; 1974 } 1975 /* 1976 * If source is the same as the destination (that is the 1977 * same vnode with the same name in the same directory), 1978 * then there is nothing to do. 1979 */ 1980 if (fvp == tvp && fromnd.ni_dvp == tdvp && 1981 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 1982 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr, 1983 fromnd.ni_cnd.cn_namelen)) 1984 error = -1; 1985 out: 1986 if (!error) { 1987 nqsrv_getl(fromnd.ni_dvp, ND_WRITE); 1988 nqsrv_getl(tdvp, ND_WRITE); 1989 if (tvp) { 1990 nqsrv_getl(tvp, ND_WRITE); 1991 } 1992 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 1993 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 1994 } else { 1995 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 1996 if (tdvp == tvp) 1997 vrele(tdvp); 1998 else 1999 vput(tdvp); 2000 if (tvp) 2001 vput(tvp); 2002 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2003 vrele(fromnd.ni_dvp); 2004 vrele(fvp); 2005 if (error == -1) 2006 error = 0; 2007 } 2008 vrele(tond.ni_startdir); 2009 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 2010 out1: 2011 if (fdirp) { 2012 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred, lwp); 2013 vrele(fdirp); 2014 } 2015 if (tdirp) { 2016 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred, lwp); 2017 vrele(tdirp); 2018 } 2019 vrele(fromnd.ni_startdir); 2020 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2021 nfsm_reply(2 * NFSX_WCCDATA(v3)); 2022 if (v3) { 2023 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 2024 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 2025 } 2026 vn_finished_write(mp, 0); 2027 return (0); 2028 2029 nfsmout: 2030 if (fdirp) 2031 vrele(fdirp); 2032 if (tdirp) 2033 vrele(tdirp); 2034 if (tond.ni_cnd.cn_nameiop) { 2035 vrele(tond.ni_startdir); 2036 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 2037 } 2038 if (fromnd.ni_cnd.cn_nameiop) { 2039 vrele(fromnd.ni_startdir); 2040 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2041 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2042 vrele(fromnd.ni_dvp); 2043 vrele(fvp); 2044 } 2045 return (error); 2046 } 2047 2048 /* 2049 * nfs link service 2050 */ 2051 int 2052 nfsrv_link(nfsd, slp, lwp, mrq) 2053 struct nfsrv_descript *nfsd; 2054 struct nfssvc_sock *slp; 2055 struct lwp *lwp; 2056 struct mbuf **mrq; 2057 { 2058 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2059 struct mbuf *nam = nfsd->nd_nam; 2060 caddr_t dpos = nfsd->nd_dpos; 2061 kauth_cred_t cred = nfsd->nd_cr; 2062 struct nameidata nd; 2063 u_int32_t *tl; 2064 int32_t t1; 2065 caddr_t bpos; 2066 int error = 0, rdonly, cache, len, dirfor_ret = 1, diraft_ret = 1; 2067 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3); 2068 char *cp2; 2069 struct mbuf *mb, *mreq; 2070 struct vnode *vp, *xp, *dirp = (struct vnode *)0; 2071 struct vattr dirfor, diraft, at; 2072 nfsfh_t nfh, dnfh; 2073 fhandle_t *fhp, *dfhp; 2074 u_quad_t frev; 2075 struct mount *mp = NULL; 2076 2077 fhp = &nfh.fh_generic; 2078 dfhp = &dnfh.fh_generic; 2079 nfsm_srvmtofh(fhp); 2080 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2081 return (ESTALE); 2082 vn_start_write(NULL, &mp, V_WAIT); 2083 nfsm_srvmtofh(dfhp); 2084 nfsm_srvnamesiz(len); 2085 error = nfsrv_fhtovp(fhp, FALSE, &vp, cred, slp, nam, 2086 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2087 if (error) { 2088 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2089 nfsm_srvpostop_attr(getret, &at); 2090 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2091 vn_finished_write(mp, 0); 2092 return (0); 2093 } 2094 if (vp->v_type == VDIR && (error = kauth_authorize_generic(cred, 2095 KAUTH_GENERIC_ISSUSER, (u_short *)0)) != 0) 2096 goto out1; 2097 nd.ni_cnd.cn_cred = cred; 2098 nd.ni_cnd.cn_nameiop = CREATE; 2099 nd.ni_cnd.cn_flags = LOCKPARENT; 2100 error = nfs_namei(&nd, dfhp, len, slp, nam, &md, &dpos, 2101 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2102 if (dirp) { 2103 if (v3) 2104 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2105 else { 2106 vrele(dirp); 2107 dirp = (struct vnode *)0; 2108 } 2109 } 2110 if (error) 2111 goto out1; 2112 xp = nd.ni_vp; 2113 if (xp != NULL) { 2114 error = EEXIST; 2115 goto out; 2116 } 2117 xp = nd.ni_dvp; 2118 if (vp->v_mount != xp->v_mount) 2119 error = EXDEV; 2120 out: 2121 if (!error) { 2122 nqsrv_getl(vp, ND_WRITE); 2123 nqsrv_getl(xp, ND_WRITE); 2124 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2125 } else { 2126 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2127 if (nd.ni_dvp == nd.ni_vp) 2128 vrele(nd.ni_dvp); 2129 else 2130 vput(nd.ni_dvp); 2131 if (nd.ni_vp) 2132 vrele(nd.ni_vp); 2133 } 2134 out1: 2135 if (v3) 2136 getret = VOP_GETATTR(vp, &at, cred, lwp); 2137 if (dirp) { 2138 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2139 vrele(dirp); 2140 } 2141 vrele(vp); 2142 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2143 if (v3) { 2144 nfsm_srvpostop_attr(getret, &at); 2145 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2146 vn_finished_write(mp, 0); 2147 return (0); 2148 } 2149 vn_finished_write(mp, 0); 2150 nfsm_srvdone; 2151 } 2152 2153 /* 2154 * nfs symbolic link service 2155 */ 2156 int 2157 nfsrv_symlink(nfsd, slp, lwp, mrq) 2158 struct nfsrv_descript *nfsd; 2159 struct nfssvc_sock *slp; 2160 struct lwp *lwp; 2161 struct mbuf **mrq; 2162 { 2163 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2164 struct mbuf *nam = nfsd->nd_nam; 2165 caddr_t dpos = nfsd->nd_dpos; 2166 kauth_cred_t cred = nfsd->nd_cr; 2167 struct vattr va, dirfor, diraft; 2168 struct nameidata nd; 2169 u_int32_t *tl; 2170 int32_t t1; 2171 struct nfsv2_sattr *sp; 2172 char *bpos, *pathcp = NULL, *cp2; 2173 struct uio io; 2174 struct iovec iv; 2175 int error = 0, cache, dirfor_ret = 1, diraft_ret = 1; 2176 uint32_t len, len2; 2177 int v3 = (nfsd->nd_flag & ND_NFSV3); 2178 struct mbuf *mb, *mreq; 2179 struct vnode *dirp = (struct vnode *)0; 2180 nfsfh_t nfh; 2181 fhandle_t *fhp; 2182 u_quad_t frev; 2183 struct mount *mp = NULL; 2184 2185 nd.ni_cnd.cn_nameiop = 0; 2186 fhp = &nfh.fh_generic; 2187 nfsm_srvmtofh(fhp); 2188 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2189 return (ESTALE); 2190 vn_start_write(NULL, &mp, V_WAIT); 2191 nfsm_srvnamesiz(len); 2192 nd.ni_cnd.cn_cred = cred; 2193 nd.ni_cnd.cn_nameiop = CREATE; 2194 nd.ni_cnd.cn_flags = LOCKPARENT; 2195 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2196 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2197 if (dirp) { 2198 if (v3) 2199 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2200 else { 2201 vrele(dirp); 2202 dirp = (struct vnode *)0; 2203 } 2204 } 2205 if (error) 2206 goto out; 2207 VATTR_NULL(&va); 2208 va.va_type = VLNK; 2209 if (v3) { 2210 va.va_mode = 0; 2211 nfsm_srvsattr(&va); 2212 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 2213 len2 = fxdr_unsigned(uint32_t, *tl); 2214 if (len2 > PATH_MAX) { 2215 /* XXX should check _PC_NO_TRUNC */ 2216 error = ENAMETOOLONG; 2217 goto abortop; 2218 } 2219 } 2220 else { 2221 /* NFSv2 */ 2222 nfsm_strsiz(len2, NFS_MAXPATHLEN); 2223 } 2224 pathcp = malloc(len2 + 1, M_TEMP, M_WAITOK); 2225 iv.iov_base = pathcp; 2226 iv.iov_len = len2; 2227 io.uio_resid = len2; 2228 io.uio_offset = 0; 2229 io.uio_iov = &iv; 2230 io.uio_iovcnt = 1; 2231 io.uio_rw = UIO_READ; 2232 UIO_SETUP_SYSSPACE(&io); 2233 nfsm_mtouio(&io, len2); 2234 if (!v3) { 2235 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 2236 va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode); 2237 } 2238 *(pathcp + len2) = '\0'; 2239 if (nd.ni_vp) { 2240 error = EEXIST; 2241 abortop: 2242 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2243 if (nd.ni_dvp == nd.ni_vp) 2244 vrele(nd.ni_dvp); 2245 else 2246 vput(nd.ni_dvp); 2247 if (nd.ni_vp) 2248 vrele(nd.ni_vp); 2249 goto out; 2250 } 2251 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2252 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp); 2253 if (!error) { 2254 if (v3) { 2255 memset((caddr_t)fhp, 0, sizeof(nfh)); 2256 fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsidx; 2257 error = VFS_VPTOFH(nd.ni_vp, &fhp->fh_fid); 2258 if (!error) 2259 error = VOP_GETATTR(nd.ni_vp, &va, cred, lwp); 2260 vput(nd.ni_vp); 2261 } else { 2262 vput(nd.ni_vp); 2263 } 2264 } 2265 out: 2266 if (pathcp) 2267 free(pathcp, M_TEMP); 2268 if (dirp) { 2269 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2270 vrele(dirp); 2271 } 2272 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2273 if (v3) { 2274 if (!error) { 2275 nfsm_srvpostop_fh(fhp); 2276 nfsm_srvpostop_attr(0, &va); 2277 } 2278 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2279 } 2280 vn_finished_write(mp, 0); 2281 return (0); 2282 nfsmout: 2283 if (dirp) 2284 vrele(dirp); 2285 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2286 if (nd.ni_dvp == nd.ni_vp) 2287 vrele(nd.ni_dvp); 2288 else 2289 vput(nd.ni_dvp); 2290 if (nd.ni_vp) 2291 vrele(nd.ni_vp); 2292 if (pathcp) 2293 free(pathcp, M_TEMP); 2294 vn_finished_write(mp, 0); 2295 return (error); 2296 } 2297 2298 /* 2299 * nfs mkdir service 2300 */ 2301 int 2302 nfsrv_mkdir(nfsd, slp, lwp, mrq) 2303 struct nfsrv_descript *nfsd; 2304 struct nfssvc_sock *slp; 2305 struct lwp *lwp; 2306 struct mbuf **mrq; 2307 { 2308 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2309 struct mbuf *nam = nfsd->nd_nam; 2310 caddr_t dpos = nfsd->nd_dpos; 2311 kauth_cred_t cred = nfsd->nd_cr; 2312 struct vattr va, dirfor, diraft; 2313 struct nfs_fattr *fp; 2314 struct nameidata nd; 2315 caddr_t cp; 2316 u_int32_t *tl; 2317 int32_t t1; 2318 caddr_t bpos; 2319 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 2320 int v3 = (nfsd->nd_flag & ND_NFSV3); 2321 char *cp2; 2322 struct mbuf *mb, *mreq; 2323 struct vnode *vp, *dirp = (struct vnode *)0; 2324 nfsfh_t nfh; 2325 fhandle_t *fhp; 2326 u_quad_t frev; 2327 struct mount *mp = NULL; 2328 2329 fhp = &nfh.fh_generic; 2330 nfsm_srvmtofh(fhp); 2331 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2332 return (ESTALE); 2333 vn_start_write(NULL, &mp, V_WAIT); 2334 nfsm_srvnamesiz(len); 2335 nd.ni_cnd.cn_cred = cred; 2336 nd.ni_cnd.cn_nameiop = CREATE; 2337 nd.ni_cnd.cn_flags = LOCKPARENT; 2338 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2339 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2340 if (dirp) { 2341 if (v3) 2342 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2343 else { 2344 vrele(dirp); 2345 dirp = (struct vnode *)0; 2346 } 2347 } 2348 if (error) { 2349 nfsm_reply(NFSX_WCCDATA(v3)); 2350 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2351 if (dirp) 2352 vrele(dirp); 2353 vn_finished_write(mp, 0); 2354 return (0); 2355 } 2356 VATTR_NULL(&va); 2357 if (v3) { 2358 va.va_mode = 0; 2359 nfsm_srvsattr(&va); 2360 } else { 2361 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 2362 va.va_mode = nfstov_mode(*tl++); 2363 } 2364 va.va_type = VDIR; 2365 vp = nd.ni_vp; 2366 if (vp != NULL) { 2367 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2368 if (nd.ni_dvp == vp) 2369 vrele(nd.ni_dvp); 2370 else 2371 vput(nd.ni_dvp); 2372 vrele(vp); 2373 error = EEXIST; 2374 goto out; 2375 } 2376 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2377 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 2378 if (!error) { 2379 vp = nd.ni_vp; 2380 memset((caddr_t)fhp, 0, sizeof(nfh)); 2381 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 2382 error = VFS_VPTOFH(vp, &fhp->fh_fid); 2383 if (!error) 2384 error = VOP_GETATTR(vp, &va, cred, lwp); 2385 vput(vp); 2386 } 2387 out: 2388 if (dirp) { 2389 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2390 vrele(dirp); 2391 } 2392 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2393 if (v3) { 2394 if (!error) { 2395 nfsm_srvpostop_fh(fhp); 2396 nfsm_srvpostop_attr(0, &va); 2397 } 2398 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2399 } else { 2400 nfsm_srvfhtom(fhp, v3); 2401 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 2402 nfsm_srvfillattr(&va, fp); 2403 } 2404 vn_finished_write(mp, 0); 2405 return (0); 2406 nfsmout: 2407 if (dirp) 2408 vrele(dirp); 2409 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2410 if (nd.ni_dvp == nd.ni_vp) 2411 vrele(nd.ni_dvp); 2412 else 2413 vput(nd.ni_dvp); 2414 if (nd.ni_vp) 2415 vrele(nd.ni_vp); 2416 vn_finished_write(mp, 0); 2417 return (error); 2418 } 2419 2420 /* 2421 * nfs rmdir service 2422 */ 2423 int 2424 nfsrv_rmdir(nfsd, slp, lwp, mrq) 2425 struct nfsrv_descript *nfsd; 2426 struct nfssvc_sock *slp; 2427 struct lwp *lwp; 2428 struct mbuf **mrq; 2429 { 2430 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2431 struct mbuf *nam = nfsd->nd_nam; 2432 caddr_t dpos = nfsd->nd_dpos; 2433 kauth_cred_t cred = nfsd->nd_cr; 2434 u_int32_t *tl; 2435 int32_t t1; 2436 caddr_t bpos; 2437 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; 2438 int v3 = (nfsd->nd_flag & ND_NFSV3); 2439 char *cp2; 2440 struct mbuf *mb, *mreq; 2441 struct vnode *vp, *dirp = (struct vnode *)0; 2442 struct vattr dirfor, diraft; 2443 nfsfh_t nfh; 2444 fhandle_t *fhp; 2445 struct nameidata nd; 2446 u_quad_t frev; 2447 struct mount *mp = NULL; 2448 2449 fhp = &nfh.fh_generic; 2450 nfsm_srvmtofh(fhp); 2451 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 2452 return (ESTALE); 2453 vn_start_write(NULL, &mp, V_WAIT); 2454 nfsm_srvnamesiz(len); 2455 nd.ni_cnd.cn_cred = cred; 2456 nd.ni_cnd.cn_nameiop = DELETE; 2457 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 2458 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, 2459 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2460 if (dirp) { 2461 if (v3) 2462 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2463 else { 2464 vrele(dirp); 2465 dirp = (struct vnode *)0; 2466 } 2467 } 2468 if (error) { 2469 nfsm_reply(NFSX_WCCDATA(v3)); 2470 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2471 if (dirp) 2472 vrele(dirp); 2473 vn_finished_write(mp, 0); 2474 return (0); 2475 } 2476 vp = nd.ni_vp; 2477 if (vp->v_type != VDIR) { 2478 error = ENOTDIR; 2479 goto out; 2480 } 2481 /* 2482 * No rmdir "." please. 2483 */ 2484 if (nd.ni_dvp == vp) { 2485 error = EINVAL; 2486 goto out; 2487 } 2488 /* 2489 * The root of a mounted filesystem cannot be deleted. 2490 */ 2491 if (vp->v_flag & VROOT) 2492 error = EBUSY; 2493 out: 2494 if (!error) { 2495 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2496 nqsrv_getl(vp, ND_WRITE); 2497 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2498 } else { 2499 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2500 if (nd.ni_dvp == nd.ni_vp) 2501 vrele(nd.ni_dvp); 2502 else 2503 vput(nd.ni_dvp); 2504 vput(vp); 2505 } 2506 if (dirp) { 2507 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2508 vrele(dirp); 2509 } 2510 nfsm_reply(NFSX_WCCDATA(v3)); 2511 if (v3) { 2512 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2513 vn_finished_write(mp, 0); 2514 return (0); 2515 } 2516 vn_finished_write(mp, 0); 2517 nfsm_srvdone; 2518 } 2519 2520 /* 2521 * nfs readdir service 2522 * - mallocs what it thinks is enough to read 2523 * count rounded up to a multiple of NFS_SRVDIRBLKSIZ <= NFS_MAXREADDIR 2524 * - calls VOP_READDIR() 2525 * - loops around building the reply 2526 * if the output generated exceeds count break out of loop 2527 * The nfsm_clget macro is used here so that the reply will be packed 2528 * tightly in mbuf clusters. 2529 * - it only knows that it has encountered eof when the VOP_READDIR() 2530 * reads nothing 2531 * - as such one readdir rpc will return eof false although you are there 2532 * and then the next will return eof 2533 * - it trims out records with d_fileno == 0 2534 * this doesn't matter for Unix clients, but they might confuse clients 2535 * for other os'. 2536 * - it trims out records with d_type == DT_WHT 2537 * these cannot be seen through NFS (unless we extend the protocol) 2538 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less 2539 * than requested, but this may not apply to all filesystems. For 2540 * example, client NFS does not { although it is never remote mounted 2541 * anyhow } 2542 * The alternate call nfsrv_readdirplus() does lookups as well. 2543 * PS: The NFS protocol spec. does not clarify what the "count" byte 2544 * argument is a count of.. just name strings and file id's or the 2545 * entire reply rpc or ... 2546 * I tried just file name and id sizes and it confused the Sun client, 2547 * so I am using the full rpc size now. The "paranoia.." comment refers 2548 * to including the status longwords that are not a part of the dir. 2549 * "entry" structures, but are in the rpc. 2550 */ 2551 2552 #define NFS_SRVDIRBLKSIZ 1024 2553 2554 struct flrep { 2555 nfsuint64 fl_off; 2556 u_int32_t fl_postopok; 2557 u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)]; 2558 u_int32_t fl_fhok; 2559 u_int32_t fl_fhsize; 2560 u_int32_t fl_nfh[NFSX_V3FH / sizeof (u_int32_t)]; 2561 }; 2562 2563 int 2564 nfsrv_readdir(nfsd, slp, lwp, mrq) 2565 struct nfsrv_descript *nfsd; 2566 struct nfssvc_sock *slp; 2567 struct lwp *lwp; 2568 struct mbuf **mrq; 2569 { 2570 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2571 struct mbuf *nam = nfsd->nd_nam; 2572 caddr_t dpos = nfsd->nd_dpos; 2573 kauth_cred_t cred = nfsd->nd_cr; 2574 char *bp, *be; 2575 struct mbuf *mp; 2576 struct dirent *dp; 2577 caddr_t cp; 2578 u_int32_t *tl; 2579 int32_t t1; 2580 caddr_t bpos; 2581 struct mbuf *mb, *mreq, *mp2; 2582 char *cpos, *cend, *cp2, *rbuf; 2583 struct vnode *vp; 2584 struct vattr at; 2585 nfsfh_t nfh; 2586 fhandle_t *fhp; 2587 struct uio io; 2588 struct iovec iv; 2589 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2590 int siz, cnt, fullsiz, eofflag, rdonly, cache, ncookies; 2591 int v3 = (nfsd->nd_flag & ND_NFSV3); 2592 u_quad_t frev, off, toff, verf; 2593 off_t *cookies = NULL, *cookiep; 2594 nfsuint64 jar; 2595 2596 fhp = &nfh.fh_generic; 2597 nfsm_srvmtofh(fhp); 2598 if (v3) { 2599 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 2600 toff = fxdr_hyper(tl); 2601 tl += 2; 2602 verf = fxdr_hyper(tl); 2603 tl += 2; 2604 } else { 2605 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2606 toff = fxdr_unsigned(u_quad_t, *tl++); 2607 } 2608 off = toff; 2609 cnt = fxdr_unsigned(int, *tl); 2610 siz = ((cnt + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1)); 2611 xfer = NFS_SRVMAXDATA(nfsd); 2612 if (siz > xfer) 2613 siz = xfer; 2614 fullsiz = siz; 2615 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 2616 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2617 if (!error && vp->v_type != VDIR) { 2618 error = ENOTDIR; 2619 vput(vp); 2620 } 2621 if (error) { 2622 nfsm_reply(NFSX_UNSIGNED); 2623 nfsm_srvpostop_attr(getret, &at); 2624 return (0); 2625 } 2626 nqsrv_getl(vp, ND_READ); 2627 if (v3) { 2628 error = getret = VOP_GETATTR(vp, &at, cred, lwp); 2629 #ifdef NFS3_STRICTVERF 2630 /* 2631 * XXX This check is too strict for Solaris 2.5 clients. 2632 */ 2633 if (!error && toff && verf != at.va_filerev) 2634 error = NFSERR_BAD_COOKIE; 2635 #endif 2636 } 2637 if (!error) 2638 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0); 2639 if (error) { 2640 vput(vp); 2641 nfsm_reply(NFSX_POSTOPATTR(v3)); 2642 nfsm_srvpostop_attr(getret, &at); 2643 return (0); 2644 } 2645 VOP_UNLOCK(vp, 0); 2646 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2647 again: 2648 iv.iov_base = rbuf; 2649 iv.iov_len = fullsiz; 2650 io.uio_iov = &iv; 2651 io.uio_iovcnt = 1; 2652 io.uio_offset = (off_t)off; 2653 io.uio_resid = fullsiz; 2654 io.uio_rw = UIO_READ; 2655 UIO_SETUP_SYSSPACE(&io); 2656 eofflag = 0; 2657 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2658 2659 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2660 2661 off = (off_t)io.uio_offset; 2662 if (!cookies && !error) 2663 error = NFSERR_PERM; 2664 if (v3) { 2665 getret = VOP_GETATTR(vp, &at, cred, lwp); 2666 if (!error) 2667 error = getret; 2668 } 2669 2670 VOP_UNLOCK(vp, 0); 2671 if (error) { 2672 vrele(vp); 2673 free((caddr_t)rbuf, M_TEMP); 2674 if (cookies) 2675 free((caddr_t)cookies, M_TEMP); 2676 nfsm_reply(NFSX_POSTOPATTR(v3)); 2677 nfsm_srvpostop_attr(getret, &at); 2678 return (0); 2679 } 2680 if (io.uio_resid) { 2681 siz -= io.uio_resid; 2682 2683 /* 2684 * If nothing read, return eof 2685 * rpc reply 2686 */ 2687 if (siz == 0) { 2688 vrele(vp); 2689 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + 2690 2 * NFSX_UNSIGNED); 2691 if (v3) { 2692 nfsm_srvpostop_attr(getret, &at); 2693 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2694 txdr_hyper(at.va_filerev, tl); 2695 tl += 2; 2696 } else 2697 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2698 *tl++ = nfs_false; 2699 *tl = nfs_true; 2700 free((caddr_t)rbuf, M_TEMP); 2701 free((caddr_t)cookies, M_TEMP); 2702 return (0); 2703 } 2704 } 2705 2706 /* 2707 * Check for degenerate cases of nothing useful read. 2708 * If so go try again 2709 */ 2710 cpos = rbuf; 2711 cend = rbuf + siz; 2712 dp = (struct dirent *)cpos; 2713 cookiep = cookies; 2714 2715 while (cpos < cend && ncookies > 0 && 2716 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2717 cpos += dp->d_reclen; 2718 dp = (struct dirent *)cpos; 2719 cookiep++; 2720 ncookies--; 2721 } 2722 if (cpos >= cend || ncookies == 0) { 2723 toff = off; 2724 siz = fullsiz; 2725 goto again; 2726 } 2727 2728 len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */ 2729 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz); 2730 if (v3) { 2731 nfsm_srvpostop_attr(getret, &at); 2732 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2733 txdr_hyper(at.va_filerev, tl); 2734 } 2735 mp = mp2 = mb; 2736 bp = bpos; 2737 be = bp + M_TRAILINGSPACE(mp); 2738 2739 /* Loop through the records and build reply */ 2740 while (cpos < cend && ncookies > 0) { 2741 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 2742 nlen = dp->d_namlen; 2743 rem = nfsm_rndup(nlen)-nlen; 2744 len += (4 * NFSX_UNSIGNED + nlen + rem); 2745 if (v3) 2746 len += 2 * NFSX_UNSIGNED; 2747 if (len > cnt) { 2748 eofflag = 0; 2749 break; 2750 } 2751 /* 2752 * Build the directory record xdr from 2753 * the dirent entry. 2754 */ 2755 nfsm_clget; 2756 *tl = nfs_true; 2757 bp += NFSX_UNSIGNED; 2758 if (v3) { 2759 nfsm_clget; 2760 *tl = txdr_unsigned(dp->d_fileno >> 32); 2761 bp += NFSX_UNSIGNED; 2762 } 2763 nfsm_clget; 2764 *tl = txdr_unsigned(dp->d_fileno); 2765 bp += NFSX_UNSIGNED; 2766 nfsm_clget; 2767 *tl = txdr_unsigned(nlen); 2768 bp += NFSX_UNSIGNED; 2769 2770 /* And loop around copying the name */ 2771 xfer = nlen; 2772 cp = dp->d_name; 2773 while (xfer > 0) { 2774 nfsm_clget; 2775 if ((bp+xfer) > be) 2776 tsiz = be-bp; 2777 else 2778 tsiz = xfer; 2779 memcpy(bp, cp, tsiz); 2780 bp += tsiz; 2781 xfer -= tsiz; 2782 if (xfer > 0) 2783 cp += tsiz; 2784 } 2785 /* And null pad to an int32_t boundary */ 2786 for (i = 0; i < rem; i++) 2787 *bp++ = '\0'; 2788 nfsm_clget; 2789 2790 /* Finish off the record */ 2791 txdr_hyper(*cookiep, &jar); 2792 if (v3) { 2793 *tl = jar.nfsuquad[0]; 2794 bp += NFSX_UNSIGNED; 2795 nfsm_clget; 2796 } 2797 *tl = jar.nfsuquad[1]; 2798 bp += NFSX_UNSIGNED; 2799 } 2800 cpos += dp->d_reclen; 2801 dp = (struct dirent *)cpos; 2802 cookiep++; 2803 ncookies--; 2804 } 2805 vrele(vp); 2806 nfsm_clget; 2807 *tl = nfs_false; 2808 bp += NFSX_UNSIGNED; 2809 nfsm_clget; 2810 if (eofflag) 2811 *tl = nfs_true; 2812 else 2813 *tl = nfs_false; 2814 bp += NFSX_UNSIGNED; 2815 if (mp != mb) { 2816 if (bp < be) 2817 mp->m_len = bp - mtod(mp, caddr_t); 2818 } else 2819 mp->m_len += bp - bpos; 2820 free((caddr_t)rbuf, M_TEMP); 2821 free((caddr_t)cookies, M_TEMP); 2822 nfsm_srvdone; 2823 } 2824 2825 int 2826 nfsrv_readdirplus(nfsd, slp, lwp, mrq) 2827 struct nfsrv_descript *nfsd; 2828 struct nfssvc_sock *slp; 2829 struct lwp *lwp; 2830 struct mbuf **mrq; 2831 { 2832 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2833 struct mbuf *nam = nfsd->nd_nam; 2834 caddr_t dpos = nfsd->nd_dpos; 2835 kauth_cred_t cred = nfsd->nd_cr; 2836 char *bp, *be; 2837 struct mbuf *mp; 2838 struct dirent *dp; 2839 caddr_t cp; 2840 u_int32_t *tl; 2841 int32_t t1; 2842 caddr_t bpos; 2843 struct mbuf *mb, *mreq, *mp2; 2844 char *cpos, *cend, *cp2, *rbuf; 2845 struct vnode *vp, *nvp; 2846 struct flrep fl; 2847 nfsfh_t nfh; 2848 fhandle_t *fhp, *nfhp = (fhandle_t *)fl.fl_nfh; 2849 struct uio io; 2850 struct iovec iv; 2851 struct vattr va, at, *vap = &va; 2852 struct nfs_fattr *fp; 2853 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2854 int siz, cnt, fullsiz, eofflag, rdonly, cache, dirlen, ncookies; 2855 u_quad_t frev, off, toff, verf; 2856 off_t *cookies = NULL, *cookiep; 2857 2858 fhp = &nfh.fh_generic; 2859 nfsm_srvmtofh(fhp); 2860 nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED); 2861 toff = fxdr_hyper(tl); 2862 tl += 2; 2863 verf = fxdr_hyper(tl); 2864 tl += 2; 2865 siz = fxdr_unsigned(int, *tl++); 2866 cnt = fxdr_unsigned(int, *tl); 2867 off = toff; 2868 siz = ((siz + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1)); 2869 xfer = NFS_SRVMAXDATA(nfsd); 2870 if (siz > xfer) 2871 siz = xfer; 2872 fullsiz = siz; 2873 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 2874 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 2875 if (!error && vp->v_type != VDIR) { 2876 error = ENOTDIR; 2877 vput(vp); 2878 } 2879 if (error) { 2880 nfsm_reply(NFSX_UNSIGNED); 2881 nfsm_srvpostop_attr(getret, &at); 2882 return (0); 2883 } 2884 error = getret = VOP_GETATTR(vp, &at, cred, lwp); 2885 #ifdef NFS3_STRICTVERF 2886 /* 2887 * XXX This check is too strict for Solaris 2.5 clients. 2888 */ 2889 if (!error && toff && verf != at.va_filerev) 2890 error = NFSERR_BAD_COOKIE; 2891 #endif 2892 if (!error) { 2893 nqsrv_getl(vp, ND_READ); 2894 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0); 2895 } 2896 if (error) { 2897 vput(vp); 2898 nfsm_reply(NFSX_V3POSTOPATTR); 2899 nfsm_srvpostop_attr(getret, &at); 2900 return (0); 2901 } 2902 VOP_UNLOCK(vp, 0); 2903 2904 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2905 again: 2906 iv.iov_base = rbuf; 2907 iv.iov_len = fullsiz; 2908 io.uio_iov = &iv; 2909 io.uio_iovcnt = 1; 2910 io.uio_offset = (off_t)off; 2911 io.uio_resid = fullsiz; 2912 io.uio_rw = UIO_READ; 2913 UIO_SETUP_SYSSPACE(&io); 2914 eofflag = 0; 2915 2916 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2917 2918 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2919 2920 off = (u_quad_t)io.uio_offset; 2921 getret = VOP_GETATTR(vp, &at, cred, lwp); 2922 2923 VOP_UNLOCK(vp, 0); 2924 2925 /* 2926 * If the VGET operation doesn't work for this filesystem, 2927 * we can't support readdirplus. Returning NOTSUPP should 2928 * make clients fall back to plain readdir. 2929 * There's no need to check for VPTOFH as well, we wouldn't 2930 * even be here otherwise. 2931 */ 2932 if (!getret) { 2933 if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp))) 2934 getret = (getret == EOPNOTSUPP) ? 2935 NFSERR_NOTSUPP : NFSERR_IO; 2936 else 2937 vput(nvp); 2938 } 2939 2940 if (!cookies && !error) 2941 error = NFSERR_PERM; 2942 if (!error) 2943 error = getret; 2944 if (error) { 2945 vrele(vp); 2946 if (cookies) 2947 free((caddr_t)cookies, M_TEMP); 2948 free((caddr_t)rbuf, M_TEMP); 2949 nfsm_reply(NFSX_V3POSTOPATTR); 2950 nfsm_srvpostop_attr(getret, &at); 2951 return (0); 2952 } 2953 if (io.uio_resid) { 2954 siz -= io.uio_resid; 2955 2956 /* 2957 * If nothing read, return eof 2958 * rpc reply 2959 */ 2960 if (siz == 0) { 2961 vrele(vp); 2962 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2963 2 * NFSX_UNSIGNED); 2964 nfsm_srvpostop_attr(getret, &at); 2965 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2966 txdr_hyper(at.va_filerev, tl); 2967 tl += 2; 2968 *tl++ = nfs_false; 2969 *tl = nfs_true; 2970 free((caddr_t)cookies, M_TEMP); 2971 free((caddr_t)rbuf, M_TEMP); 2972 return (0); 2973 } 2974 } 2975 2976 /* 2977 * Check for degenerate cases of nothing useful read. 2978 * If so go try again 2979 */ 2980 cpos = rbuf; 2981 cend = rbuf + siz; 2982 dp = (struct dirent *)cpos; 2983 cookiep = cookies; 2984 2985 while (cpos < cend && ncookies > 0 && 2986 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2987 cpos += dp->d_reclen; 2988 dp = (struct dirent *)cpos; 2989 cookiep++; 2990 ncookies--; 2991 } 2992 if (cpos >= cend || ncookies == 0) { 2993 toff = off; 2994 siz = fullsiz; 2995 goto again; 2996 } 2997 2998 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED; 2999 nfsm_reply(cnt); 3000 nfsm_srvpostop_attr(getret, &at); 3001 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 3002 txdr_hyper(at.va_filerev, tl); 3003 mp = mp2 = mb; 3004 bp = bpos; 3005 be = bp + M_TRAILINGSPACE(mp); 3006 3007 /* Loop through the records and build reply */ 3008 while (cpos < cend && ncookies > 0) { 3009 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 3010 nlen = dp->d_namlen; 3011 rem = nfsm_rndup(nlen)-nlen; 3012 3013 /* 3014 * For readdir_and_lookup get the vnode using 3015 * the file number. 3016 */ 3017 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp)) 3018 goto invalid; 3019 memset((caddr_t)nfhp, 0, NFSX_V3FH); 3020 nfhp->fh_fsid = 3021 nvp->v_mount->mnt_stat.f_fsidx; 3022 if (VFS_VPTOFH(nvp, &nfhp->fh_fid)) { 3023 vput(nvp); 3024 goto invalid; 3025 } 3026 if (VOP_GETATTR(nvp, vap, cred, lwp)) { 3027 vput(nvp); 3028 goto invalid; 3029 } 3030 vput(nvp); 3031 3032 /* 3033 * If either the dircount or maxcount will be 3034 * exceeded, get out now. Both of these lengths 3035 * are calculated conservatively, including all 3036 * XDR overheads. 3037 */ 3038 len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH + 3039 NFSX_V3POSTOPATTR); 3040 dirlen += (6 * NFSX_UNSIGNED + nlen + rem); 3041 if (len > cnt || dirlen > fullsiz) { 3042 eofflag = 0; 3043 break; 3044 } 3045 3046 /* 3047 * Build the directory record xdr from 3048 * the dirent entry. 3049 */ 3050 fp = (struct nfs_fattr *)&fl.fl_fattr; 3051 nfsm_srvfillattr(vap, fp); 3052 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH); 3053 fl.fl_fhok = nfs_true; 3054 fl.fl_postopok = nfs_true; 3055 txdr_hyper(*cookiep, fl.fl_off.nfsuquad); 3056 3057 nfsm_clget; 3058 *tl = nfs_true; 3059 bp += NFSX_UNSIGNED; 3060 nfsm_clget; 3061 *tl = txdr_unsigned(dp->d_fileno >> 32); 3062 bp += NFSX_UNSIGNED; 3063 nfsm_clget; 3064 *tl = txdr_unsigned(dp->d_fileno); 3065 bp += NFSX_UNSIGNED; 3066 nfsm_clget; 3067 *tl = txdr_unsigned(nlen); 3068 bp += NFSX_UNSIGNED; 3069 3070 /* And loop around copying the name */ 3071 xfer = nlen; 3072 cp = dp->d_name; 3073 while (xfer > 0) { 3074 nfsm_clget; 3075 if ((bp + xfer) > be) 3076 tsiz = be - bp; 3077 else 3078 tsiz = xfer; 3079 memcpy(bp, cp, tsiz); 3080 bp += tsiz; 3081 xfer -= tsiz; 3082 if (xfer > 0) 3083 cp += tsiz; 3084 } 3085 /* And null pad to an int32_t boundary */ 3086 for (i = 0; i < rem; i++) 3087 *bp++ = '\0'; 3088 3089 /* 3090 * Now copy the flrep structure out. 3091 */ 3092 xfer = sizeof (struct flrep); 3093 cp = (caddr_t)&fl; 3094 while (xfer > 0) { 3095 nfsm_clget; 3096 if ((bp + xfer) > be) 3097 tsiz = be - bp; 3098 else 3099 tsiz = xfer; 3100 memcpy(bp, cp, tsiz); 3101 bp += tsiz; 3102 xfer -= tsiz; 3103 if (xfer > 0) 3104 cp += tsiz; 3105 } 3106 } 3107 invalid: 3108 cpos += dp->d_reclen; 3109 dp = (struct dirent *)cpos; 3110 cookiep++; 3111 ncookies--; 3112 } 3113 vrele(vp); 3114 nfsm_clget; 3115 *tl = nfs_false; 3116 bp += NFSX_UNSIGNED; 3117 nfsm_clget; 3118 if (eofflag) 3119 *tl = nfs_true; 3120 else 3121 *tl = nfs_false; 3122 bp += NFSX_UNSIGNED; 3123 if (mp != mb) { 3124 if (bp < be) 3125 mp->m_len = bp - mtod(mp, caddr_t); 3126 } else 3127 mp->m_len += bp - bpos; 3128 free((caddr_t)cookies, M_TEMP); 3129 free((caddr_t)rbuf, M_TEMP); 3130 nfsm_srvdone; 3131 } 3132 3133 /* 3134 * nfs commit service 3135 */ 3136 int 3137 nfsrv_commit(nfsd, slp, lwp, mrq) 3138 struct nfsrv_descript *nfsd; 3139 struct nfssvc_sock *slp; 3140 struct lwp *lwp; 3141 struct mbuf **mrq; 3142 { 3143 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3144 struct mbuf *nam = nfsd->nd_nam; 3145 caddr_t dpos = nfsd->nd_dpos; 3146 kauth_cred_t cred = nfsd->nd_cr; 3147 struct vattr bfor, aft; 3148 struct vnode *vp; 3149 nfsfh_t nfh; 3150 fhandle_t *fhp; 3151 u_int32_t *tl; 3152 int32_t t1; 3153 caddr_t bpos; 3154 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cache; 3155 uint32_t cnt; 3156 char *cp2; 3157 struct mbuf *mb, *mreq; 3158 u_quad_t frev, off, end; 3159 struct mount *mp = NULL; 3160 3161 #ifndef nolint 3162 cache = 0; 3163 #endif 3164 fhp = &nfh.fh_generic; 3165 nfsm_srvmtofh(fhp); 3166 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) 3167 return (ESTALE); 3168 vn_start_write(NULL, &mp, V_WAIT); 3169 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED); 3170 3171 off = fxdr_hyper(tl); 3172 tl += 2; 3173 cnt = fxdr_unsigned(uint32_t, *tl); 3174 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3175 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3176 if (error) { 3177 nfsm_reply(2 * NFSX_UNSIGNED); 3178 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3179 vn_finished_write(mp, 0); 3180 return (0); 3181 } 3182 for_ret = VOP_GETATTR(vp, &bfor, cred, lwp); 3183 end = (cnt > 0) ? off + cnt : vp->v_size; 3184 if (end < off || end > vp->v_size) 3185 end = vp->v_size; 3186 if (off < vp->v_size) 3187 error = VOP_FSYNC(vp, cred, FSYNC_WAIT, off, end, lwp); 3188 /* else error == 0, from nfsrv_fhtovp() */ 3189 aft_ret = VOP_GETATTR(vp, &aft, cred, lwp); 3190 vput(vp); 3191 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF); 3192 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3193 if (!error) { 3194 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF); 3195 *tl++ = txdr_unsigned(boottime.tv_sec); 3196 *tl = txdr_unsigned(boottime.tv_usec); 3197 } else { 3198 vn_finished_write(mp, 0); 3199 return (0); 3200 } 3201 vn_finished_write(mp, 0); 3202 nfsm_srvdone; 3203 } 3204 3205 /* 3206 * nfs statfs service 3207 */ 3208 int 3209 nfsrv_statfs(nfsd, slp, lwp, mrq) 3210 struct nfsrv_descript *nfsd; 3211 struct nfssvc_sock *slp; 3212 struct lwp *lwp; 3213 struct mbuf **mrq; 3214 { 3215 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3216 struct mbuf *nam = nfsd->nd_nam; 3217 caddr_t dpos = nfsd->nd_dpos; 3218 kauth_cred_t cred = nfsd->nd_cr; 3219 struct statvfs *sf; 3220 struct nfs_statfs *sfp; 3221 u_int32_t *tl; 3222 int32_t t1; 3223 caddr_t bpos; 3224 int error = 0, rdonly, cache, getret = 1; 3225 int v3 = (nfsd->nd_flag & ND_NFSV3); 3226 char *cp2; 3227 struct mbuf *mb, *mreq; 3228 struct vnode *vp; 3229 struct vattr at; 3230 nfsfh_t nfh; 3231 fhandle_t *fhp; 3232 struct statvfs statvfs; 3233 u_quad_t frev, tval; 3234 3235 #ifndef nolint 3236 cache = 0; 3237 #endif 3238 fhp = &nfh.fh_generic; 3239 nfsm_srvmtofh(fhp); 3240 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3241 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3242 if (error) { 3243 nfsm_reply(NFSX_UNSIGNED); 3244 nfsm_srvpostop_attr(getret, &at); 3245 return (0); 3246 } 3247 sf = &statvfs; 3248 error = VFS_STATVFS(vp->v_mount, sf, lwp); 3249 getret = VOP_GETATTR(vp, &at, cred, lwp); 3250 vput(vp); 3251 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3)); 3252 if (v3) 3253 nfsm_srvpostop_attr(getret, &at); 3254 if (error) 3255 return (0); 3256 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3)); 3257 if (v3) { 3258 tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_frsize); 3259 txdr_hyper(tval, &sfp->sf_tbytes); 3260 tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_frsize); 3261 txdr_hyper(tval, &sfp->sf_fbytes); 3262 tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_frsize); 3263 txdr_hyper(tval, &sfp->sf_abytes); 3264 tval = (u_quad_t)sf->f_files; 3265 txdr_hyper(tval, &sfp->sf_tfiles); 3266 tval = (u_quad_t)sf->f_ffree; 3267 txdr_hyper(tval, &sfp->sf_ffiles); 3268 txdr_hyper(tval, &sfp->sf_afiles); 3269 sfp->sf_invarsec = 0; 3270 } else { 3271 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA); 3272 sfp->sf_bsize = txdr_unsigned(sf->f_frsize); 3273 sfp->sf_blocks = txdr_unsigned(sf->f_blocks); 3274 sfp->sf_bfree = txdr_unsigned(sf->f_bfree); 3275 sfp->sf_bavail = txdr_unsigned(sf->f_bavail); 3276 } 3277 nfsm_srvdone; 3278 } 3279 3280 /* 3281 * nfs fsinfo service 3282 */ 3283 int 3284 nfsrv_fsinfo(nfsd, slp, lwp, mrq) 3285 struct nfsrv_descript *nfsd; 3286 struct nfssvc_sock *slp; 3287 struct lwp *lwp; 3288 struct mbuf **mrq; 3289 { 3290 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3291 struct mbuf *nam = nfsd->nd_nam; 3292 caddr_t dpos = nfsd->nd_dpos; 3293 kauth_cred_t cred = nfsd->nd_cr; 3294 u_int32_t *tl; 3295 struct nfsv3_fsinfo *sip; 3296 int32_t t1; 3297 caddr_t bpos; 3298 int error = 0, rdonly, cache, getret = 1; 3299 uint32_t maxdata; 3300 char *cp2; 3301 struct mbuf *mb, *mreq; 3302 struct vnode *vp; 3303 struct vattr at; 3304 nfsfh_t nfh; 3305 fhandle_t *fhp; 3306 u_quad_t frev, maxfsize; 3307 struct statvfs sb; 3308 3309 #ifndef nolint 3310 cache = 0; 3311 #endif 3312 fhp = &nfh.fh_generic; 3313 nfsm_srvmtofh(fhp); 3314 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3315 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3316 if (error) { 3317 nfsm_reply(NFSX_UNSIGNED); 3318 nfsm_srvpostop_attr(getret, &at); 3319 return (0); 3320 } 3321 3322 /* XXX Try to make a guess on the max file size. */ 3323 VFS_STATVFS(vp->v_mount, &sb, (struct lwp *)0); 3324 maxfsize = (u_quad_t)0x80000000 * sb.f_frsize - 1; 3325 3326 getret = VOP_GETATTR(vp, &at, cred, lwp); 3327 vput(vp); 3328 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO); 3329 nfsm_srvpostop_attr(getret, &at); 3330 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO); 3331 3332 /* 3333 * XXX 3334 * There should be file system VFS OP(s) to get this information. 3335 * For now, assume ufs. 3336 */ 3337 if (slp->ns_so->so_type == SOCK_DGRAM) 3338 maxdata = NFS_MAXDGRAMDATA; 3339 else 3340 maxdata = NFS_MAXDATA; 3341 sip->fs_rtmax = txdr_unsigned(maxdata); 3342 sip->fs_rtpref = txdr_unsigned(maxdata); 3343 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE); 3344 sip->fs_wtmax = txdr_unsigned(maxdata); 3345 sip->fs_wtpref = txdr_unsigned(maxdata); 3346 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE); 3347 sip->fs_dtpref = txdr_unsigned(maxdata); 3348 txdr_hyper(maxfsize, &sip->fs_maxfilesize); 3349 sip->fs_timedelta.nfsv3_sec = 0; 3350 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1); 3351 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK | 3352 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS | 3353 NFSV3FSINFO_CANSETTIME); 3354 nfsm_srvdone; 3355 } 3356 3357 /* 3358 * nfs pathconf service 3359 */ 3360 int 3361 nfsrv_pathconf(nfsd, slp, lwp, mrq) 3362 struct nfsrv_descript *nfsd; 3363 struct nfssvc_sock *slp; 3364 struct lwp *lwp; 3365 struct mbuf **mrq; 3366 { 3367 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3368 struct mbuf *nam = nfsd->nd_nam; 3369 caddr_t dpos = nfsd->nd_dpos; 3370 kauth_cred_t cred = nfsd->nd_cr; 3371 u_int32_t *tl; 3372 struct nfsv3_pathconf *pc; 3373 int32_t t1; 3374 caddr_t bpos; 3375 int error = 0, rdonly, cache, getret = 1; 3376 register_t linkmax, namemax, chownres, notrunc; 3377 char *cp2; 3378 struct mbuf *mb, *mreq; 3379 struct vnode *vp; 3380 struct vattr at; 3381 nfsfh_t nfh; 3382 fhandle_t *fhp; 3383 u_quad_t frev; 3384 3385 #ifndef nolint 3386 cache = 0; 3387 #endif 3388 fhp = &nfh.fh_generic; 3389 nfsm_srvmtofh(fhp); 3390 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, 3391 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE); 3392 if (error) { 3393 nfsm_reply(NFSX_UNSIGNED); 3394 nfsm_srvpostop_attr(getret, &at); 3395 return (0); 3396 } 3397 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax); 3398 if (!error) 3399 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax); 3400 if (!error) 3401 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres); 3402 if (!error) 3403 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc); 3404 getret = VOP_GETATTR(vp, &at, cred, lwp); 3405 vput(vp); 3406 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF); 3407 nfsm_srvpostop_attr(getret, &at); 3408 if (error) 3409 return (0); 3410 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF); 3411 3412 pc->pc_linkmax = txdr_unsigned(linkmax); 3413 pc->pc_namemax = txdr_unsigned(namemax); 3414 pc->pc_notrunc = txdr_unsigned(notrunc); 3415 pc->pc_chownrestricted = txdr_unsigned(chownres); 3416 3417 /* 3418 * These should probably be supported by VOP_PATHCONF(), but 3419 * until msdosfs is exportable (why would you want to?), the 3420 * Unix defaults should be ok. 3421 */ 3422 pc->pc_caseinsensitive = nfs_false; 3423 pc->pc_casepreserving = nfs_true; 3424 nfsm_srvdone; 3425 } 3426 3427 /* 3428 * Null operation, used by clients to ping server 3429 */ 3430 /* ARGSUSED */ 3431 int 3432 nfsrv_null(nfsd, slp, lwp, mrq) 3433 struct nfsrv_descript *nfsd; 3434 struct nfssvc_sock *slp; 3435 struct lwp *lwp; 3436 struct mbuf **mrq; 3437 { 3438 struct mbuf *mrep = nfsd->nd_mrep; 3439 caddr_t bpos; 3440 int error = NFSERR_RETVOID, cache = 0; 3441 struct mbuf *mb, *mreq; 3442 u_quad_t frev; 3443 3444 nfsm_reply(0); 3445 return (0); 3446 } 3447 3448 /* 3449 * No operation, used for obsolete procedures 3450 */ 3451 /* ARGSUSED */ 3452 int 3453 nfsrv_noop(nfsd, slp, lwp, mrq) 3454 struct nfsrv_descript *nfsd; 3455 struct nfssvc_sock *slp; 3456 struct lwp *lwp; 3457 struct mbuf **mrq; 3458 { 3459 struct mbuf *mrep = nfsd->nd_mrep; 3460 caddr_t bpos; 3461 int error, cache = 0; 3462 struct mbuf *mb, *mreq; 3463 u_quad_t frev; 3464 3465 if (nfsd->nd_repstat) 3466 error = nfsd->nd_repstat; 3467 else 3468 error = EPROCUNAVAIL; 3469 nfsm_reply(0); 3470 return (0); 3471 } 3472 3473 /* 3474 * Perform access checking for vnodes obtained from file handles that would 3475 * refer to files already opened by a Unix client. You cannot just use 3476 * vn_writechk() and VOP_ACCESS() for two reasons. 3477 * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case 3478 * 2 - The owner is to be given access irrespective of mode bits for some 3479 * operations, so that processes that chmod after opening a file don't 3480 * break. I don't like this because it opens a security hole, but since 3481 * the nfs server opens a security hole the size of a barn door anyhow, 3482 * what the heck. 3483 * 3484 * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS() 3485 * will return EPERM instead of EACCESS. EPERM is always an error. 3486 */ 3487 int 3488 nfsrv_access(vp, flags, cred, rdonly, lwp, override) 3489 struct vnode *vp; 3490 int flags; 3491 kauth_cred_t cred; 3492 int rdonly; 3493 struct lwp *lwp; 3494 { 3495 struct vattr vattr; 3496 int error; 3497 if (flags & VWRITE) { 3498 /* Just vn_writechk() changed to check rdonly */ 3499 /* 3500 * Disallow write attempts on read-only file systems; 3501 * unless the file is a socket or a block or character 3502 * device resident on the file system. 3503 */ 3504 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 3505 switch (vp->v_type) { 3506 case VREG: 3507 case VDIR: 3508 case VLNK: 3509 return (EROFS); 3510 default: 3511 break; 3512 } 3513 } 3514 3515 /* 3516 * If the vnode is in use as a process's text, 3517 * we can't allow writing. 3518 */ 3519 if (vp->v_flag & VTEXT) 3520 return (ETXTBSY); 3521 } 3522 error = VOP_GETATTR(vp, &vattr, cred, lwp); 3523 if (error) 3524 return (error); 3525 error = VOP_ACCESS(vp, flags, cred, lwp); 3526 /* 3527 * Allow certain operations for the owner (reads and writes 3528 * on files that are already open). 3529 */ 3530 if (override && error == EACCES && kauth_cred_geteuid(cred) == vattr.va_uid) 3531 error = 0; 3532 return error; 3533 } 3534
Indexes created Thu Oct 02 14:10:14 GMT 2025