nfs_serv.c revision 1.128
1 /* $NetBSD: nfs_serv.c,v 1.128 2007/04/06 14:28:28 hannken 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 1.128 2007/04/06 14:28:28 hannken Exp $"); 59 60 #include <sys/param.h> 61 #include <sys/systm.h> 62 #include <sys/proc.h> 63 #include <sys/file.h> 64 #include <sys/namei.h> 65 #include <sys/vnode.h> 66 #include <sys/mount.h> 67 #include <sys/socket.h> 68 #include <sys/socketvar.h> 69 #include <sys/mbuf.h> 70 #include <sys/dirent.h> 71 #include <sys/stat.h> 72 #include <sys/kernel.h> 73 #include <sys/hash.h> 74 #include <sys/kauth.h> 75 76 #include <uvm/uvm.h> 77 78 #include <nfs/nfsproto.h> 79 #include <nfs/rpcv2.h> 80 #include <nfs/nfs.h> 81 #include <nfs/xdr_subs.h> 82 #include <nfs/nfsm_subs.h> 83 #include <nfs/nfs_var.h> 84 85 /* Global vars */ 86 extern u_int32_t nfs_xdrneg1; 87 extern u_int32_t nfs_false, nfs_true; 88 extern enum vtype nv3tov_type[8]; 89 extern struct nfsstats nfsstats; 90 extern const nfstype nfsv2_type[9]; 91 extern const nfstype nfsv3_type[9]; 92 int nfsrvw_procrastinate = NFS_GATHERDELAY * 1000; 93 int nfsd_use_loan = 1; /* use page-loan for READ OP */ 94 95 #define nqsrv_getl(vp, rw) /* nothing */ 96 97 /* 98 * nfs v3 access service 99 */ 100 int 101 nfsrv3_access(nfsd, slp, lwp, mrq) 102 struct nfsrv_descript *nfsd; 103 struct nfssvc_sock *slp; 104 struct lwp *lwp; 105 struct mbuf **mrq; 106 { 107 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 108 struct mbuf *nam = nfsd->nd_nam; 109 char *dpos = nfsd->nd_dpos; 110 kauth_cred_t cred = nfsd->nd_cr; 111 struct vnode *vp; 112 nfsrvfh_t nsfh; 113 u_int32_t *tl; 114 int32_t t1; 115 char *bpos; 116 int error = 0, rdonly, cache = 0, getret; 117 char *cp2; 118 struct mbuf *mb, *mreq; 119 struct vattr va; 120 u_long inmode, testmode, outmode; 121 u_quad_t frev; 122 123 nfsm_srvmtofh(&nsfh); 124 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 125 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, &rdonly, 126 (nfsd->nd_flag & ND_KERBAUTH), false); 127 if (error) { 128 nfsm_reply(NFSX_UNSIGNED); 129 nfsm_srvpostop_attr(1, (struct vattr *)0); 130 return (0); 131 } 132 inmode = fxdr_unsigned(u_int32_t, *tl); 133 outmode = 0; 134 if ((inmode & NFSV3ACCESS_READ) && 135 nfsrv_access(vp, VREAD, cred, rdonly, lwp, 0) == 0) 136 outmode |= NFSV3ACCESS_READ; 137 if (vp->v_type != VDIR) { 138 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND); 139 if (testmode && 140 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0) 141 outmode |= testmode; 142 if ((inmode & NFSV3ACCESS_EXECUTE) && 143 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0) 144 outmode |= NFSV3ACCESS_EXECUTE; 145 } else { 146 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND | 147 NFSV3ACCESS_DELETE); 148 if (testmode && 149 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0) 150 outmode |= testmode; 151 if ((inmode & NFSV3ACCESS_LOOKUP) && 152 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0) 153 outmode |= NFSV3ACCESS_LOOKUP; 154 } 155 getret = VOP_GETATTR(vp, &va, cred, lwp); 156 vput(vp); 157 nfsm_reply(NFSX_POSTOPATTR(1) + NFSX_UNSIGNED); 158 nfsm_srvpostop_attr(getret, &va); 159 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 160 *tl = txdr_unsigned(outmode); 161 nfsm_srvdone; 162 } 163 164 /* 165 * nfs getattr service 166 */ 167 int 168 nfsrv_getattr(nfsd, slp, lwp, mrq) 169 struct nfsrv_descript *nfsd; 170 struct nfssvc_sock *slp; 171 struct lwp *lwp; 172 struct mbuf **mrq; 173 { 174 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 175 struct mbuf *nam = nfsd->nd_nam; 176 char *dpos = nfsd->nd_dpos; 177 kauth_cred_t cred = nfsd->nd_cr; 178 struct nfs_fattr *fp; 179 struct vattr va; 180 struct vnode *vp; 181 nfsrvfh_t nsfh; 182 u_int32_t *tl; 183 int32_t t1; 184 char *bpos; 185 int error = 0, rdonly, cache = 0; 186 char *cp2; 187 struct mbuf *mb, *mreq; 188 u_quad_t frev; 189 190 nfsm_srvmtofh(&nsfh); 191 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, &rdonly, 192 (nfsd->nd_flag & ND_KERBAUTH), false); 193 if (error) { 194 nfsm_reply(0); 195 return (0); 196 } 197 nqsrv_getl(vp, ND_READ); 198 error = VOP_GETATTR(vp, &va, cred, lwp); 199 vput(vp); 200 nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); 201 if (error) 202 return (0); 203 nfsm_build(fp, struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); 204 nfsm_srvfillattr(&va, fp); 205 nfsm_srvdone; 206 } 207 208 /* 209 * nfs setattr service 210 */ 211 int 212 nfsrv_setattr(nfsd, slp, lwp, mrq) 213 struct nfsrv_descript *nfsd; 214 struct nfssvc_sock *slp; 215 struct lwp *lwp; 216 struct mbuf **mrq; 217 { 218 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 219 struct mbuf *nam = nfsd->nd_nam; 220 char *dpos = nfsd->nd_dpos; 221 kauth_cred_t cred = nfsd->nd_cr; 222 struct vattr va, preat; 223 struct nfsv2_sattr *sp; 224 struct nfs_fattr *fp; 225 struct vnode *vp; 226 nfsrvfh_t nsfh; 227 u_int32_t *tl; 228 int32_t t1; 229 char *bpos; 230 int error = 0, rdonly, cache = 0, preat_ret = 1, postat_ret = 1; 231 int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0; 232 char *cp2; 233 struct mbuf *mb, *mreq; 234 u_quad_t frev; 235 struct timespec guard; 236 237 memset(&guard, 0, sizeof guard); /* XXX gcc */ 238 239 nfsm_srvmtofh(&nsfh); 240 VATTR_NULL(&va); 241 if (v3) { 242 nfsm_srvsattr(&va); 243 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 244 gcheck = fxdr_unsigned(int, *tl); 245 if (gcheck) { 246 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 247 fxdr_nfsv3time(tl, &guard); 248 } 249 } else { 250 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 251 /* 252 * Nah nah nah nah na nah 253 * There is a bug in the Sun client that puts 0xffff in the mode 254 * field of sattr when it should put in 0xffffffff. The u_short 255 * doesn't sign extend. 256 * --> check the low order 2 bytes for 0xffff 257 */ 258 if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff) 259 va.va_mode = nfstov_mode(sp->sa_mode); 260 if (sp->sa_uid != nfs_xdrneg1) 261 va.va_uid = fxdr_unsigned(uid_t, sp->sa_uid); 262 if (sp->sa_gid != nfs_xdrneg1) 263 va.va_gid = fxdr_unsigned(gid_t, sp->sa_gid); 264 if (sp->sa_size != nfs_xdrneg1) 265 va.va_size = fxdr_unsigned(u_quad_t, sp->sa_size); 266 if (sp->sa_atime.nfsv2_sec != nfs_xdrneg1) { 267 #ifdef notyet 268 fxdr_nfsv2time(&sp->sa_atime, &va.va_atime); 269 #else 270 va.va_atime.tv_sec = 271 fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec); 272 va.va_atime.tv_nsec = 0; 273 #endif 274 } 275 if (sp->sa_mtime.nfsv2_sec != nfs_xdrneg1) 276 fxdr_nfsv2time(&sp->sa_mtime, &va.va_mtime); 277 278 } 279 280 /* 281 * Now that we have all the fields, lets do it. 282 */ 283 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, &rdonly, 284 (nfsd->nd_flag & ND_KERBAUTH), false); 285 if (error) { 286 nfsm_reply(2 * NFSX_UNSIGNED); 287 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 288 return (0); 289 } 290 nqsrv_getl(vp, ND_WRITE); 291 if (v3) { 292 error = preat_ret = VOP_GETATTR(vp, &preat, cred, lwp); 293 if (!error && gcheck && 294 (preat.va_ctime.tv_sec != guard.tv_sec || 295 preat.va_ctime.tv_nsec != guard.tv_nsec)) 296 error = NFSERR_NOT_SYNC; 297 if (error) { 298 vput(vp); 299 nfsm_reply(NFSX_WCCDATA(v3)); 300 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 301 return (0); 302 } 303 } 304 305 /* 306 * If the size is being changed write acces is required, otherwise 307 * just check for a read only file system. 308 */ 309 if (va.va_size == ((u_quad_t)((quad_t) -1))) { 310 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 311 error = EROFS; 312 goto out; 313 } 314 } else { 315 if (vp->v_type == VDIR) { 316 error = EISDIR; 317 goto out; 318 } else if ((error = nfsrv_access(vp, VWRITE, cred, rdonly, 319 lwp, 0)) != 0) 320 goto out; 321 } 322 error = VOP_SETATTR(vp, &va, cred, lwp); 323 postat_ret = VOP_GETATTR(vp, &va, cred, lwp); 324 if (!error) 325 error = postat_ret; 326 out: 327 vput(vp); 328 nfsm_reply(NFSX_WCCORFATTR(v3)); 329 if (v3) { 330 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 331 return (0); 332 } else { 333 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 334 nfsm_srvfillattr(&va, fp); 335 } 336 nfsm_srvdone; 337 } 338 339 /* 340 * nfs lookup rpc 341 */ 342 int 343 nfsrv_lookup(nfsd, slp, lwp, mrq) 344 struct nfsrv_descript *nfsd; 345 struct nfssvc_sock *slp; 346 struct lwp *lwp; 347 struct mbuf **mrq; 348 { 349 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 350 struct mbuf *nam = nfsd->nd_nam; 351 char *dpos = nfsd->nd_dpos; 352 kauth_cred_t cred = nfsd->nd_cr; 353 struct nfs_fattr *fp; 354 struct nameidata nd, ind, *ndp = &nd; 355 struct vnode *vp, *dirp; 356 nfsrvfh_t nsfh; 357 char *cp; 358 u_int32_t *tl; 359 int32_t t1; 360 char *bpos; 361 int error = 0, cache = 0, dirattr_ret = 1; 362 uint32_t len; 363 int v3 = (nfsd->nd_flag & ND_NFSV3), pubflag; 364 char *cp2; 365 struct mbuf *mb, *mreq; 366 struct vattr va, dirattr; 367 u_quad_t frev; 368 369 nfsm_srvmtofh(&nsfh); 370 nfsm_srvnamesiz(len); 371 372 pubflag = nfs_ispublicfh(&nsfh); 373 374 nd.ni_cnd.cn_cred = cred; 375 nd.ni_cnd.cn_nameiop = LOOKUP; 376 nd.ni_cnd.cn_flags = LOCKLEAF | SAVESTART; 377 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 378 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), pubflag); 379 380 if (!error && pubflag) { 381 if (nd.ni_vp->v_type == VDIR && nfs_pub.np_index != NULL) { 382 /* 383 * Setup call to lookup() to see if we can find 384 * the index file. Arguably, this doesn't belong 385 * in a kernel.. Ugh. 386 */ 387 ind = nd; 388 VOP_UNLOCK(nd.ni_vp, 0); 389 ind.ni_pathlen = strlen(nfs_pub.np_index); 390 ind.ni_cnd.cn_nameptr = ind.ni_cnd.cn_pnbuf = 391 nfs_pub.np_index; 392 ind.ni_startdir = nd.ni_vp; 393 VREF(ind.ni_startdir); 394 error = lookup(&ind); 395 if (!error) { 396 /* 397 * Found an index file. Get rid of 398 * the old references. 399 */ 400 if (dirp) 401 vrele(dirp); 402 dirp = nd.ni_vp; 403 vrele(nd.ni_startdir); 404 ndp = &ind; 405 } else 406 error = 0; 407 } 408 /* 409 * If the public filehandle was used, check that this lookup 410 * didn't result in a filehandle outside the publicly exported 411 * filesystem. 412 */ 413 414 if (!error && ndp->ni_vp->v_mount != nfs_pub.np_mount) { 415 vput(nd.ni_vp); 416 error = EPERM; 417 } 418 } 419 420 if (dirp) { 421 if (v3) 422 dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred, lwp); 423 vrele(dirp); 424 } 425 426 if (error) { 427 nfsm_reply(NFSX_POSTOPATTR(v3)); 428 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 429 return (0); 430 } 431 432 nqsrv_getl(ndp->ni_startdir, ND_READ); 433 PNBUF_PUT(nd.ni_cnd.cn_pnbuf); 434 vp = ndp->ni_vp; 435 error = nfsrv_composefh(vp, &nsfh, v3); 436 if (!error) 437 error = VOP_GETATTR(vp, &va, cred, lwp); 438 vput(vp); 439 vrele(ndp->ni_startdir); 440 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_POSTOPORFATTR(v3) + 441 NFSX_POSTOPATTR(v3)); 442 if (error) { 443 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 444 return (0); 445 } 446 nfsm_srvfhtom(&nsfh, v3); 447 if (v3) { 448 nfsm_srvpostop_attr(0, &va); 449 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 450 } else { 451 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 452 nfsm_srvfillattr(&va, fp); 453 } 454 nfsm_srvdone; 455 } 456 457 /* 458 * nfs readlink service 459 */ 460 int 461 nfsrv_readlink(nfsd, slp, lwp, mrq) 462 struct nfsrv_descript *nfsd; 463 struct nfssvc_sock *slp; 464 struct lwp *lwp; 465 struct mbuf **mrq; 466 { 467 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 468 struct mbuf *nam = nfsd->nd_nam; 469 char *dpos = nfsd->nd_dpos; 470 kauth_cred_t cred = nfsd->nd_cr; 471 struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN]; 472 struct iovec *ivp = iv; 473 struct mbuf *mp; 474 u_int32_t *tl; 475 int32_t t1; 476 char *bpos; 477 int error = 0, rdonly, cache = 0, i, padlen, getret; 478 uint32_t len; 479 int v3 = (nfsd->nd_flag & ND_NFSV3); 480 char *cp2; 481 struct mbuf *mb, *mp2 = NULL, *mp3 = NULL, *mreq; 482 struct vnode *vp; 483 struct vattr attr; 484 nfsrvfh_t nsfh; 485 struct uio io, *uiop = &io; 486 u_quad_t frev; 487 488 nfsm_srvmtofh(&nsfh); 489 len = 0; 490 i = 0; 491 while (len < NFS_MAXPATHLEN) { 492 mp = m_get(M_WAIT, MT_DATA); 493 MCLAIM(mp, &nfs_mowner); 494 m_clget(mp, M_WAIT); 495 mp->m_len = NFSMSIZ(mp); 496 if (len == 0) 497 mp3 = mp2 = mp; 498 else { 499 mp2->m_next = mp; 500 mp2 = mp; 501 } 502 if ((len+mp->m_len) > NFS_MAXPATHLEN) { 503 mp->m_len = NFS_MAXPATHLEN-len; 504 len = NFS_MAXPATHLEN; 505 } else 506 len += mp->m_len; 507 ivp->iov_base = mtod(mp, void *); 508 ivp->iov_len = mp->m_len; 509 i++; 510 ivp++; 511 } 512 uiop->uio_iov = iv; 513 uiop->uio_iovcnt = i; 514 uiop->uio_offset = 0; 515 uiop->uio_resid = len; 516 uiop->uio_rw = UIO_READ; 517 UIO_SETUP_SYSSPACE(uiop); 518 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 519 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 520 if (error) { 521 m_freem(mp3); 522 nfsm_reply(2 * NFSX_UNSIGNED); 523 nfsm_srvpostop_attr(1, (struct vattr *)0); 524 return (0); 525 } 526 if (vp->v_type != VLNK) { 527 if (v3) 528 error = EINVAL; 529 else 530 error = ENXIO; 531 goto out; 532 } 533 nqsrv_getl(vp, ND_READ); 534 error = VOP_READLINK(vp, uiop, cred); 535 out: 536 getret = VOP_GETATTR(vp, &attr, cred, lwp); 537 vput(vp); 538 if (error) 539 m_freem(mp3); 540 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED); 541 if (v3) { 542 nfsm_srvpostop_attr(getret, &attr); 543 if (error) 544 return (0); 545 } 546 len -= uiop->uio_resid; 547 padlen = nfsm_padlen(len); 548 if (uiop->uio_resid || padlen) 549 nfs_zeropad(mp3, uiop->uio_resid, padlen); 550 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 551 *tl = txdr_unsigned(len); 552 mb->m_next = mp3; 553 nfsm_srvdone; 554 } 555 556 /* 557 * nfs read service 558 */ 559 int 560 nfsrv_read(nfsd, slp, lwp, mrq) 561 struct nfsrv_descript *nfsd; 562 struct nfssvc_sock *slp; 563 struct lwp *lwp; 564 struct mbuf **mrq; 565 { 566 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 567 struct mbuf *nam = nfsd->nd_nam; 568 char *dpos = nfsd->nd_dpos; 569 kauth_cred_t cred = nfsd->nd_cr; 570 struct mbuf *m; 571 struct nfs_fattr *fp; 572 u_int32_t *tl; 573 int32_t t1; 574 int i; 575 char *bpos; 576 int error = 0, rdonly, cache = 0, getret; 577 int v3 = (nfsd->nd_flag & ND_NFSV3); 578 uint32_t reqlen, len, cnt, left; 579 int padlen; 580 char *cp2; 581 struct mbuf *mb, *mreq; 582 struct vnode *vp; 583 nfsrvfh_t nsfh; 584 struct uio io, *uiop = &io; 585 struct vattr va; 586 off_t off; 587 u_quad_t frev; 588 589 nfsm_srvmtofh(&nsfh); 590 if (v3) { 591 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 592 off = fxdr_hyper(tl); 593 } else { 594 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 595 off = (off_t)fxdr_unsigned(u_int32_t, *tl); 596 } 597 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 598 reqlen = fxdr_unsigned(uint32_t, *tl); 599 reqlen = MIN(reqlen, NFS_SRVMAXDATA(nfsd)); 600 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 601 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 602 if (error) { 603 nfsm_reply(2 * NFSX_UNSIGNED); 604 nfsm_srvpostop_attr(1, (struct vattr *)0); 605 return (0); 606 } 607 if (vp->v_type != VREG) { 608 if (v3) 609 error = EINVAL; 610 else 611 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 612 } 613 if (!error) { 614 nqsrv_getl(vp, ND_READ); 615 if ((error = nfsrv_access(vp, VREAD, cred, rdonly, lwp, 1)) != 0) 616 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 1); 617 } 618 getret = VOP_GETATTR(vp, &va, cred, lwp); 619 if (!error) 620 error = getret; 621 if (error) { 622 vput(vp); 623 nfsm_reply(NFSX_POSTOPATTR(v3)); 624 nfsm_srvpostop_attr(getret, &va); 625 return (0); 626 } 627 if (off >= va.va_size) 628 cnt = 0; 629 else if ((off + reqlen) > va.va_size) 630 cnt = va.va_size - off; 631 else 632 cnt = reqlen; 633 nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt)); 634 if (v3) { 635 nfsm_build(tl, u_int32_t *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED); 636 *tl++ = nfs_true; 637 fp = (struct nfs_fattr *)tl; 638 tl += (NFSX_V3FATTR / sizeof (u_int32_t)); 639 } else { 640 nfsm_build(tl, u_int32_t *, NFSX_V2FATTR + NFSX_UNSIGNED); 641 fp = (struct nfs_fattr *)tl; 642 tl += (NFSX_V2FATTR / sizeof (u_int32_t)); 643 } 644 len = left = cnt; 645 if (cnt > 0) { 646 if (nfsd_use_loan) { 647 struct vm_page **pgpp; 648 voff_t pgoff = trunc_page(off); 649 int npages; 650 vaddr_t lva; 651 652 npages = (round_page(off + cnt) - pgoff) >> PAGE_SHIFT; 653 KASSERT(npages <= M_EXT_MAXPAGES); /* XXX */ 654 655 /* allocate kva for mbuf data */ 656 lva = sokvaalloc(npages << PAGE_SHIFT, slp->ns_so); 657 if (lva == 0) { 658 /* fall back to VOP_READ */ 659 goto loan_fail; 660 } 661 662 /* allocate mbuf */ 663 m = m_get(M_WAIT, MT_DATA); 664 MCLAIM(m, &nfs_mowner); 665 pgpp = m->m_ext.ext_pgs; 666 667 /* loan pages */ 668 error = uvm_loanuobjpages(&vp->v_uobj, pgoff, npages, 669 pgpp); 670 if (error) { 671 sokvafree(lva, npages << PAGE_SHIFT); 672 m_free(m); 673 if (error == EBUSY) 674 goto loan_fail; 675 goto read_error; 676 } 677 678 /* associate kva to mbuf */ 679 MEXTADD(m, (void *)(lva + ((vaddr_t)off & PAGE_MASK)), 680 cnt, M_MBUF, soloanfree, slp->ns_so); 681 m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP; 682 m->m_len = cnt; 683 684 /* map pages */ 685 for (i = 0; i < npages; i++) { 686 pmap_kenter_pa(lva, VM_PAGE_TO_PHYS(pgpp[i]), 687 VM_PROT_READ); 688 lva += PAGE_SIZE; 689 } 690 691 pmap_update(pmap_kernel()); 692 693 mb->m_next = m; 694 mb = m; 695 error = 0; 696 uiop->uio_resid = 0; 697 } else { 698 struct iovec *iv; 699 struct iovec *iv2; 700 struct mbuf *m2; 701 int siz; 702 loan_fail: 703 /* 704 * Generate the mbuf list with the uio_iov ref. to it. 705 */ 706 i = 0; 707 m = m2 = mb; 708 while (left > 0) { 709 siz = min(M_TRAILINGSPACE(m), left); 710 if (siz > 0) { 711 left -= siz; 712 i++; 713 } 714 if (left > 0) { 715 m = m_get(M_WAIT, MT_DATA); 716 MCLAIM(m, &nfs_mowner); 717 m_clget(m, M_WAIT); 718 m->m_len = 0; 719 m2->m_next = m; 720 m2 = m; 721 } 722 } 723 iv = malloc(i * sizeof(struct iovec), M_TEMP, M_WAITOK); 724 uiop->uio_iov = iv2 = iv; 725 m = mb; 726 left = cnt; 727 i = 0; 728 while (left > 0) { 729 if (m == NULL) 730 panic("nfsrv_read iov"); 731 siz = min(M_TRAILINGSPACE(m), left); 732 if (siz > 0) { 733 iv->iov_base = mtod(m, char *) + 734 m->m_len; 735 iv->iov_len = siz; 736 m->m_len += siz; 737 left -= siz; 738 iv++; 739 i++; 740 } 741 m = m->m_next; 742 } 743 uiop->uio_iovcnt = i; 744 uiop->uio_offset = off; 745 uiop->uio_resid = cnt; 746 uiop->uio_rw = UIO_READ; 747 UIO_SETUP_SYSSPACE(uiop); 748 error = VOP_READ(vp, uiop, IO_NODELOCKED, cred); 749 free((void *)iv2, M_TEMP); 750 } 751 read_error: 752 if (error || (getret = VOP_GETATTR(vp, &va, cred, lwp)) != 0){ 753 if (!error) 754 error = getret; 755 m_freem(mreq); 756 vput(vp); 757 nfsm_reply(NFSX_POSTOPATTR(v3)); 758 nfsm_srvpostop_attr(getret, &va); 759 return (0); 760 } 761 } else { 762 uiop->uio_resid = 0; 763 } 764 vput(vp); 765 nfsm_srvfillattr(&va, fp); 766 len -= uiop->uio_resid; 767 padlen = nfsm_padlen(len); 768 if (uiop->uio_resid || padlen) 769 nfs_zeropad(mb, uiop->uio_resid, padlen); 770 if (v3) { 771 /* count */ 772 *tl++ = txdr_unsigned(len); 773 /* eof */ 774 if (off + len >= va.va_size) 775 *tl++ = nfs_true; 776 else 777 *tl++ = nfs_false; 778 } 779 *tl = txdr_unsigned(len); 780 nfsm_srvdone; 781 } 782 783 /* 784 * nfs write service 785 */ 786 int 787 nfsrv_write(nfsd, slp, lwp, mrq) 788 struct nfsrv_descript *nfsd; 789 struct nfssvc_sock *slp; 790 struct lwp *lwp; 791 struct mbuf **mrq; 792 { 793 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 794 struct mbuf *nam = nfsd->nd_nam; 795 char *dpos = nfsd->nd_dpos; 796 kauth_cred_t cred = nfsd->nd_cr; 797 struct iovec *ivp; 798 int i, cnt; 799 struct mbuf *mp; 800 struct nfs_fattr *fp; 801 struct iovec *iv; 802 struct vattr va, forat; 803 u_int32_t *tl; 804 int32_t t1; 805 char *bpos; 806 int error = 0, rdonly, cache = 0, len, forat_ret = 1; 807 int ioflags, aftat_ret = 1, retlen, zeroing, adjust; 808 int stable = NFSV3WRITE_FILESYNC; 809 int v3 = (nfsd->nd_flag & ND_NFSV3); 810 char *cp2; 811 struct mbuf *mb, *mreq; 812 struct vnode *vp; 813 nfsrvfh_t nsfh; 814 struct uio io, *uiop = &io; 815 off_t off; 816 u_quad_t frev; 817 818 if (mrep == NULL) { 819 *mrq = NULL; 820 return (0); 821 } 822 nfsm_srvmtofh(&nsfh); 823 if (v3) { 824 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 825 off = fxdr_hyper(tl); 826 tl += 3; 827 stable = fxdr_unsigned(int, *tl++); 828 } else { 829 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 830 off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 831 tl += 2; 832 } 833 retlen = len = fxdr_unsigned(int32_t, *tl); 834 cnt = i = 0; 835 836 /* 837 * For NFS Version 2, it is not obvious what a write of zero length 838 * should do, but I might as well be consistent with Version 3, 839 * which is to return ok so long as there are no permission problems. 840 */ 841 if (len > 0) { 842 zeroing = 1; 843 mp = mrep; 844 while (mp) { 845 if (mp == md) { 846 zeroing = 0; 847 adjust = dpos - mtod(mp, char *); 848 mp->m_len -= adjust; 849 if (mp->m_len > 0 && adjust > 0) 850 NFSMADV(mp, adjust); 851 } 852 if (zeroing) 853 mp->m_len = 0; 854 else if (mp->m_len > 0) { 855 i += mp->m_len; 856 if (i > len) { 857 mp->m_len -= (i - len); 858 zeroing = 1; 859 } 860 if (mp->m_len > 0) 861 cnt++; 862 } 863 mp = mp->m_next; 864 } 865 } 866 if (len > NFS_MAXDATA || len < 0 || i < len) { 867 error = EIO; 868 nfsm_reply(2 * NFSX_UNSIGNED); 869 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 870 return (0); 871 } 872 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 873 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 874 if (error) { 875 nfsm_reply(2 * NFSX_UNSIGNED); 876 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 877 return (0); 878 } 879 if (v3) 880 forat_ret = VOP_GETATTR(vp, &forat, cred, lwp); 881 if (vp->v_type != VREG) { 882 if (v3) 883 error = EINVAL; 884 else 885 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 886 } 887 if (!error) { 888 nqsrv_getl(vp, ND_WRITE); 889 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1); 890 } 891 if (error) { 892 vput(vp); 893 nfsm_reply(NFSX_WCCDATA(v3)); 894 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 895 return (0); 896 } 897 898 if (len > 0) { 899 ivp = malloc(cnt * sizeof (struct iovec), M_TEMP, M_WAITOK); 900 uiop->uio_iov = iv = ivp; 901 uiop->uio_iovcnt = cnt; 902 mp = mrep; 903 while (mp) { 904 if (mp->m_len > 0) { 905 ivp->iov_base = mtod(mp, void *); 906 ivp->iov_len = mp->m_len; 907 ivp++; 908 } 909 mp = mp->m_next; 910 } 911 912 /* 913 * XXX 914 * The IO_METASYNC flag indicates that all metadata (and not 915 * just enough to ensure data integrity) must be written to 916 * stable storage synchronously. 917 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.) 918 */ 919 if (stable == NFSV3WRITE_UNSTABLE) 920 ioflags = IO_NODELOCKED; 921 else if (stable == NFSV3WRITE_DATASYNC) 922 ioflags = (IO_SYNC | IO_NODELOCKED); 923 else 924 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 925 uiop->uio_resid = len; 926 uiop->uio_rw = UIO_WRITE; 927 uiop->uio_offset = off; 928 UIO_SETUP_SYSSPACE(uiop); 929 error = VOP_WRITE(vp, uiop, ioflags, cred); 930 nfsstats.srvvop_writes++; 931 free(iv, M_TEMP); 932 } 933 aftat_ret = VOP_GETATTR(vp, &va, cred, lwp); 934 vput(vp); 935 if (!error) 936 error = aftat_ret; 937 nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) + 938 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3)); 939 if (v3) { 940 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 941 if (error) 942 return (0); 943 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 944 *tl++ = txdr_unsigned(retlen); 945 if (stable == NFSV3WRITE_UNSTABLE) 946 *tl++ = txdr_unsigned(stable); 947 else 948 *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC); 949 /* 950 * Actually, there is no need to txdr these fields, 951 * but it may make the values more human readable, 952 * for debugging purposes. 953 */ 954 *tl++ = txdr_unsigned(boottime.tv_sec); 955 *tl = txdr_unsigned(boottime.tv_usec); 956 } else { 957 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 958 nfsm_srvfillattr(&va, fp); 959 } 960 nfsm_srvdone; 961 } 962 963 /* 964 * XXX elad: the original NFSW_SAMECRED() macro also made sure the 965 * two nd_flag fields of the descriptors contained 966 * ND_KERBAUTH. 967 */ 968 static int 969 nfsrv_samecred(kauth_cred_t cred1, kauth_cred_t cred2) 970 { 971 int i, do_ngroups; 972 973 if (kauth_cred_geteuid(cred1) != kauth_cred_geteuid(cred2)) 974 return (0); 975 if (kauth_cred_ngroups(cred1) != kauth_cred_ngroups(cred2)) 976 return (0); 977 do_ngroups = kauth_cred_ngroups(cred1); 978 for (i = 0; i < do_ngroups; i++) 979 if (kauth_cred_group(cred1, i) != 980 kauth_cred_group(cred2, i)) 981 return (0); 982 983 return (1); 984 } 985 986 static struct nfsrvw_delayhash * 987 nfsrv_nwdelayhash(struct nfssvc_sock *slp, const nfsrvfh_t *nsfh) 988 { 989 uint32_t hash; 990 991 hash = hash32_buf(NFSRVFH_DATA(nsfh), NFSRVFH_SIZE(nsfh), 992 HASH32_BUF_INIT); 993 return &slp->ns_wdelayhashtbl[hash % NFS_WDELAYHASHSIZ]; 994 } 995 996 /* 997 * NFS write service with write gathering support. Called when 998 * nfsrvw_procrastinate > 0. 999 * See: Chet Juszczak, "Improving the Write Performance of an NFS Server", 1000 * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco, 1001 * Jan. 1994. 1002 */ 1003 int 1004 nfsrv_writegather(ndp, slp, lwp, mrq) 1005 struct nfsrv_descript **ndp; 1006 struct nfssvc_sock *slp; 1007 struct lwp *lwp; 1008 struct mbuf **mrq; 1009 { 1010 struct timeval now; 1011 struct iovec *ivp; 1012 struct mbuf *mp; 1013 struct nfsrv_descript *wp, *nfsd, *owp, *swp; 1014 struct nfs_fattr *fp; 1015 int i = 0; 1016 struct iovec *iov; 1017 struct nfsrvw_delayhash *wpp; 1018 kauth_cred_t cred; 1019 struct vattr va, forat; 1020 u_int32_t *tl; 1021 int32_t t1; 1022 char *bpos, *dpos; 1023 int error = 0, rdonly, cache = 0, len = 0, forat_ret = 1; 1024 int ioflags, aftat_ret = 1, s, adjust, v3, zeroing; 1025 char *cp2; 1026 struct mbuf *mb, *mreq, *mrep, *md; 1027 struct vnode *vp; 1028 struct uio io, *uiop = &io; 1029 u_quad_t frev, cur_usec; 1030 1031 *mrq = NULL; 1032 if (*ndp) { 1033 nfsd = *ndp; 1034 *ndp = NULL; 1035 mrep = nfsd->nd_mrep; 1036 md = nfsd->nd_md; 1037 dpos = nfsd->nd_dpos; 1038 cred = nfsd->nd_cr; 1039 v3 = (nfsd->nd_flag & ND_NFSV3); 1040 LIST_INIT(&nfsd->nd_coalesce); 1041 nfsd->nd_mreq = NULL; 1042 nfsd->nd_stable = NFSV3WRITE_FILESYNC; 1043 getmicrotime(&now); 1044 cur_usec = (u_quad_t)now.tv_sec * 1000000 + (u_quad_t)now.tv_usec; 1045 nfsd->nd_time = cur_usec + nfsrvw_procrastinate; 1046 1047 /* 1048 * Now, get the write header.. 1049 */ 1050 nfsm_srvmtofh(&nfsd->nd_fh); 1051 if (v3) { 1052 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 1053 nfsd->nd_off = fxdr_hyper(tl); 1054 tl += 3; 1055 nfsd->nd_stable = fxdr_unsigned(int, *tl++); 1056 } else { 1057 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1058 nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 1059 tl += 2; 1060 } 1061 len = fxdr_unsigned(int32_t, *tl); 1062 nfsd->nd_len = len; 1063 nfsd->nd_eoff = nfsd->nd_off + len; 1064 1065 /* 1066 * Trim the header out of the mbuf list and trim off any trailing 1067 * junk so that the mbuf list has only the write data. 1068 */ 1069 zeroing = 1; 1070 i = 0; 1071 mp = mrep; 1072 while (mp) { 1073 if (mp == md) { 1074 zeroing = 0; 1075 adjust = dpos - mtod(mp, char *); 1076 mp->m_len -= adjust; 1077 if (mp->m_len > 0 && adjust > 0) 1078 NFSMADV(mp, adjust); 1079 } 1080 if (zeroing) 1081 mp->m_len = 0; 1082 else { 1083 i += mp->m_len; 1084 if (i > len) { 1085 mp->m_len -= (i - len); 1086 zeroing = 1; 1087 } 1088 } 1089 mp = mp->m_next; 1090 } 1091 if (len > NFS_MAXDATA || len < 0 || i < len) { 1092 nfsmout: 1093 m_freem(mrep); 1094 error = EIO; 1095 nfsm_writereply(2 * NFSX_UNSIGNED, v3); 1096 if (v3) 1097 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1098 nfsd->nd_mreq = mreq; 1099 nfsd->nd_mrep = NULL; 1100 nfsd->nd_time = 0; 1101 } 1102 1103 /* 1104 * Add this entry to the hash and time queues. 1105 */ 1106 s = splsoftclock(); 1107 owp = NULL; 1108 wp = LIST_FIRST(&slp->ns_tq); 1109 while (wp && wp->nd_time < nfsd->nd_time) { 1110 owp = wp; 1111 wp = LIST_NEXT(wp, nd_tq); 1112 } 1113 if (owp) { 1114 LIST_INSERT_AFTER(owp, nfsd, nd_tq); 1115 } else { 1116 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1117 } 1118 if (nfsd->nd_mrep) { 1119 wpp = nfsrv_nwdelayhash(slp, &nfsd->nd_fh); 1120 owp = NULL; 1121 wp = LIST_FIRST(wpp); 1122 while (wp && nfsrv_comparefh(&nfsd->nd_fh, &wp->nd_fh)) { 1123 owp = wp; 1124 wp = LIST_NEXT(wp, nd_hash); 1125 } 1126 while (wp && wp->nd_off < nfsd->nd_off && 1127 !nfsrv_comparefh(&nfsd->nd_fh, &wp->nd_fh)) { 1128 owp = wp; 1129 wp = LIST_NEXT(wp, nd_hash); 1130 } 1131 if (owp) { 1132 LIST_INSERT_AFTER(owp, nfsd, nd_hash); 1133 1134 /* 1135 * Search the hash list for overlapping entries and 1136 * coalesce. 1137 */ 1138 for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) { 1139 wp = LIST_NEXT(nfsd, nd_hash); 1140 if (nfsrv_samecred(owp->nd_cr, nfsd->nd_cr)) 1141 nfsrvw_coalesce(owp, nfsd); 1142 } 1143 } else { 1144 LIST_INSERT_HEAD(wpp, nfsd, nd_hash); 1145 } 1146 } 1147 splx(s); 1148 } 1149 1150 /* 1151 * Now, do VOP_WRITE()s for any one(s) that need to be done now 1152 * and generate the associated reply mbuf list(s). 1153 */ 1154 loop1: 1155 getmicrotime(&now); 1156 cur_usec = (u_quad_t)now.tv_sec * 1000000 + (u_quad_t)now.tv_usec; 1157 s = splsoftclock(); 1158 for (nfsd = LIST_FIRST(&slp->ns_tq); nfsd; nfsd = owp) { 1159 owp = LIST_NEXT(nfsd, nd_tq); 1160 if (nfsd->nd_time > cur_usec) 1161 break; 1162 if (nfsd->nd_mreq) 1163 continue; 1164 LIST_REMOVE(nfsd, nd_tq); 1165 LIST_REMOVE(nfsd, nd_hash); 1166 splx(s); 1167 mrep = nfsd->nd_mrep; 1168 nfsd->nd_mrep = NULL; 1169 cred = nfsd->nd_cr; 1170 v3 = (nfsd->nd_flag & ND_NFSV3); 1171 forat_ret = aftat_ret = 1; 1172 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp, 1173 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH), 1174 false); 1175 if (!error) { 1176 if (v3) 1177 forat_ret = VOP_GETATTR(vp, &forat, cred, lwp); 1178 if (vp->v_type != VREG) { 1179 if (v3) 1180 error = EINVAL; 1181 else 1182 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 1183 } 1184 } else 1185 vp = NULL; 1186 if (!error) { 1187 nqsrv_getl(vp, ND_WRITE); 1188 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1); 1189 } 1190 1191 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE) 1192 ioflags = IO_NODELOCKED; 1193 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC) 1194 ioflags = (IO_SYNC | IO_NODELOCKED); 1195 else 1196 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 1197 uiop->uio_rw = UIO_WRITE; 1198 uiop->uio_offset = nfsd->nd_off; 1199 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off; 1200 UIO_SETUP_SYSSPACE(uiop); 1201 if (uiop->uio_resid > 0) { 1202 mp = mrep; 1203 i = 0; 1204 while (mp) { 1205 if (mp->m_len > 0) 1206 i++; 1207 mp = mp->m_next; 1208 } 1209 uiop->uio_iovcnt = i; 1210 iov = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK); 1211 uiop->uio_iov = ivp = iov; 1212 mp = mrep; 1213 while (mp) { 1214 if (mp->m_len > 0) { 1215 ivp->iov_base = mtod(mp, void *); 1216 ivp->iov_len = mp->m_len; 1217 ivp++; 1218 } 1219 mp = mp->m_next; 1220 } 1221 if (!error) { 1222 error = VOP_WRITE(vp, uiop, ioflags, cred); 1223 nfsstats.srvvop_writes++; 1224 } 1225 free((void *)iov, M_TEMP); 1226 } 1227 m_freem(mrep); 1228 if (vp) { 1229 aftat_ret = VOP_GETATTR(vp, &va, cred, lwp); 1230 vput(vp); 1231 } 1232 1233 /* 1234 * Loop around generating replies for all write rpcs that have 1235 * now been completed. 1236 */ 1237 swp = nfsd; 1238 do { 1239 if (error) { 1240 nfsm_writereply(NFSX_WCCDATA(v3), v3); 1241 if (v3) { 1242 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1243 } 1244 } else { 1245 nfsm_writereply(NFSX_PREOPATTR(v3) + 1246 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED + 1247 NFSX_WRITEVERF(v3), v3); 1248 if (v3) { 1249 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1250 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1251 *tl++ = txdr_unsigned(nfsd->nd_len); 1252 *tl++ = txdr_unsigned(swp->nd_stable); 1253 /* 1254 * Actually, there is no need to txdr these fields, 1255 * but it may make the values more human readable, 1256 * for debugging purposes. 1257 */ 1258 *tl++ = txdr_unsigned(boottime.tv_sec); 1259 *tl = txdr_unsigned(boottime.tv_usec); 1260 } else { 1261 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1262 nfsm_srvfillattr(&va, fp); 1263 } 1264 } 1265 nfsd->nd_mreq = mreq; 1266 if (nfsd->nd_mrep) 1267 panic("nfsrv_write: nd_mrep not free"); 1268 1269 /* 1270 * Done. Put it at the head of the timer queue so that 1271 * the final phase can return the reply. 1272 */ 1273 s = splsoftclock(); 1274 if (nfsd != swp) { 1275 nfsd->nd_time = 0; 1276 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1277 } 1278 nfsd = LIST_FIRST(&swp->nd_coalesce); 1279 if (nfsd) { 1280 LIST_REMOVE(nfsd, nd_tq); 1281 } 1282 splx(s); 1283 } while (nfsd); 1284 s = splsoftclock(); 1285 swp->nd_time = 0; 1286 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq); 1287 splx(s); 1288 goto loop1; 1289 } 1290 splx(s); 1291 1292 /* 1293 * Search for a reply to return. 1294 */ 1295 s = splsoftclock(); 1296 LIST_FOREACH(nfsd, &slp->ns_tq, nd_tq) { 1297 if (nfsd->nd_mreq) { 1298 LIST_REMOVE(nfsd, nd_tq); 1299 *mrq = nfsd->nd_mreq; 1300 *ndp = nfsd; 1301 break; 1302 } 1303 } 1304 splx(s); 1305 return (0); 1306 } 1307 1308 /* 1309 * Coalesce the write request nfsd into owp. To do this we must: 1310 * - remove nfsd from the queues 1311 * - merge nfsd->nd_mrep into owp->nd_mrep 1312 * - update the nd_eoff and nd_stable for owp 1313 * - put nfsd on owp's nd_coalesce list 1314 * NB: Must be called at splsoftclock(). 1315 */ 1316 void 1317 nfsrvw_coalesce(owp, nfsd) 1318 struct nfsrv_descript *owp; 1319 struct nfsrv_descript *nfsd; 1320 { 1321 int overlap; 1322 struct mbuf *mp; 1323 struct nfsrv_descript *m; 1324 1325 LIST_REMOVE(nfsd, nd_hash); 1326 LIST_REMOVE(nfsd, nd_tq); 1327 if (owp->nd_eoff < nfsd->nd_eoff) { 1328 overlap = owp->nd_eoff - nfsd->nd_off; 1329 if (overlap < 0) 1330 panic("nfsrv_coalesce: bad off"); 1331 if (overlap > 0) 1332 m_adj(nfsd->nd_mrep, overlap); 1333 mp = owp->nd_mrep; 1334 while (mp->m_next) 1335 mp = mp->m_next; 1336 mp->m_next = nfsd->nd_mrep; 1337 owp->nd_eoff = nfsd->nd_eoff; 1338 } else 1339 m_freem(nfsd->nd_mrep); 1340 nfsd->nd_mrep = NULL; 1341 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC) 1342 owp->nd_stable = NFSV3WRITE_FILESYNC; 1343 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC && 1344 owp->nd_stable == NFSV3WRITE_UNSTABLE) 1345 owp->nd_stable = NFSV3WRITE_DATASYNC; 1346 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq); 1347 /* 1348 * nfsd might hold coalesce elements! Move them to owp. 1349 * Otherwise, requests may be lost and clients will be stuck. 1350 */ 1351 while ((m = LIST_FIRST(&nfsd->nd_coalesce)) != NULL) { 1352 LIST_REMOVE(m, nd_tq); 1353 LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq); 1354 } 1355 } 1356 1357 /* 1358 * nfs create service 1359 * now does a truncate to 0 length via. setattr if it already exists 1360 */ 1361 int 1362 nfsrv_create(nfsd, slp, lwp, mrq) 1363 struct nfsrv_descript *nfsd; 1364 struct nfssvc_sock *slp; 1365 struct lwp *lwp; 1366 struct mbuf **mrq; 1367 { 1368 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1369 struct mbuf *nam = nfsd->nd_nam; 1370 char *dpos = nfsd->nd_dpos; 1371 kauth_cred_t cred = nfsd->nd_cr; 1372 struct nfs_fattr *fp; 1373 struct vattr va, dirfor, diraft; 1374 struct nfsv2_sattr *sp; 1375 u_int32_t *tl; 1376 struct nameidata nd; 1377 char *cp; 1378 int32_t t1; 1379 char *bpos; 1380 int error = 0, cache = 0, len, tsize, dirfor_ret = 1, diraft_ret = 1; 1381 int rdev = 0; 1382 int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0; 1383 char *cp2; 1384 struct mbuf *mb, *mreq; 1385 struct vnode *vp = NULL, *dirp = NULL; 1386 nfsrvfh_t nsfh; 1387 u_quad_t frev, tempsize; 1388 u_char cverf[NFSX_V3CREATEVERF]; 1389 1390 nd.ni_cnd.cn_nameiop = 0; 1391 nfsm_srvmtofh(&nsfh); 1392 nfsm_srvnamesiz(len); 1393 nd.ni_cnd.cn_cred = cred; 1394 nd.ni_cnd.cn_nameiop = CREATE; 1395 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1396 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 1397 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1398 if (dirp && v3) { 1399 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1400 } 1401 if (error) { 1402 nfsm_reply(NFSX_WCCDATA(v3)); 1403 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1404 if (dirp) 1405 vrele(dirp); 1406 return (0); 1407 } 1408 VATTR_NULL(&va); 1409 if (v3) { 1410 va.va_mode = 0; 1411 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1412 how = fxdr_unsigned(int, *tl); 1413 switch (how) { 1414 case NFSV3CREATE_GUARDED: 1415 if (nd.ni_vp) { 1416 error = EEXIST; 1417 break; 1418 } 1419 case NFSV3CREATE_UNCHECKED: 1420 nfsm_srvsattr(&va); 1421 break; 1422 case NFSV3CREATE_EXCLUSIVE: 1423 nfsm_dissect(cp, void *, NFSX_V3CREATEVERF); 1424 memcpy(cverf, cp, NFSX_V3CREATEVERF); 1425 exclusive_flag = 1; 1426 break; 1427 }; 1428 va.va_type = VREG; 1429 } else { 1430 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 1431 va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode)); 1432 if (va.va_type == VNON) 1433 va.va_type = VREG; 1434 va.va_mode = nfstov_mode(sp->sa_mode); 1435 switch (va.va_type) { 1436 case VREG: 1437 tsize = fxdr_unsigned(int32_t, sp->sa_size); 1438 if (tsize != -1) 1439 va.va_size = (u_quad_t)tsize; 1440 break; 1441 case VCHR: 1442 case VBLK: 1443 case VFIFO: 1444 rdev = fxdr_unsigned(int32_t, sp->sa_size); 1445 break; 1446 default: 1447 break; 1448 }; 1449 } 1450 1451 /* 1452 * Iff doesn't exist, create it 1453 * otherwise just truncate to 0 length 1454 * should I set the mode too ?? 1455 */ 1456 if (nd.ni_vp == NULL) { 1457 if (va.va_type == VREG || va.va_type == VSOCK) { 1458 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1459 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1460 if (!error) { 1461 if (exclusive_flag) { 1462 exclusive_flag = 0; 1463 VATTR_NULL(&va); 1464 /* 1465 * XXX 1466 * assuming NFSX_V3CREATEVERF 1467 * == sizeof(nfstime3) 1468 */ 1469 fxdr_nfsv3time(cverf, &va.va_atime); 1470 error = VOP_SETATTR(nd.ni_vp, &va, cred, 1471 lwp); 1472 } 1473 } 1474 } else if (va.va_type == VCHR || va.va_type == VBLK || 1475 va.va_type == VFIFO) { 1476 if (va.va_type == VCHR && rdev == 0xffffffff) 1477 va.va_type = VFIFO; 1478 if (va.va_type != VFIFO && 1479 (error = kauth_authorize_generic(cred, 1480 KAUTH_GENERIC_ISSUSER, NULL))) { 1481 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1482 vput(nd.ni_dvp); 1483 nfsm_reply(0); 1484 return (error); 1485 } else 1486 va.va_rdev = (dev_t)rdev; 1487 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1488 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, 1489 &va); 1490 if (error) { 1491 nfsm_reply(0); 1492 } 1493 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1494 vput(nd.ni_vp); 1495 vrele(nd.ni_dvp); 1496 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1497 error = EINVAL; 1498 nfsm_reply(0); 1499 } 1500 } else { 1501 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1502 vput(nd.ni_dvp); 1503 error = ENXIO; 1504 } 1505 vp = nd.ni_vp; 1506 } else { 1507 vp = nd.ni_vp; 1508 if (nd.ni_dvp == vp) 1509 vrele(nd.ni_dvp); 1510 else 1511 vput(nd.ni_dvp); 1512 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1513 if (!error && va.va_size != -1) { 1514 error = nfsrv_access(vp, VWRITE, cred, 1515 (nd.ni_cnd.cn_flags & RDONLY), lwp, 0); 1516 if (!error) { 1517 nqsrv_getl(vp, ND_WRITE); 1518 tempsize = va.va_size; 1519 VATTR_NULL(&va); 1520 va.va_size = tempsize; 1521 error = VOP_SETATTR(vp, &va, cred, lwp); 1522 } 1523 } 1524 if (error) 1525 vput(vp); 1526 } 1527 if (!error) { 1528 error = nfsrv_composefh(vp, &nsfh, v3); 1529 if (!error) 1530 error = VOP_GETATTR(vp, &va, cred, lwp); 1531 vput(vp); 1532 } 1533 if (v3) { 1534 if (exclusive_flag && !error) { 1535 /* 1536 * XXX assuming NFSX_V3CREATEVERF == sizeof(nfstime3) 1537 */ 1538 char oldverf[NFSX_V3CREATEVERF]; 1539 1540 txdr_nfsv3time(&va.va_atime, oldverf); 1541 if (memcmp(cverf, oldverf, NFSX_V3CREATEVERF)) 1542 error = EEXIST; 1543 } 1544 if (dirp) { 1545 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1546 } 1547 } 1548 if (dirp) { 1549 vrele(dirp); 1550 } 1551 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3)); 1552 if (v3) { 1553 if (!error) { 1554 nfsm_srvpostop_fh(&nsfh); 1555 nfsm_srvpostop_attr(0, &va); 1556 } 1557 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1558 } else { 1559 nfsm_srvfhtom(&nsfh, v3); 1560 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1561 nfsm_srvfillattr(&va, fp); 1562 } 1563 return (0); 1564 nfsmout: 1565 if (dirp) 1566 vrele(dirp); 1567 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1568 if (nd.ni_dvp == nd.ni_vp) 1569 vrele(nd.ni_dvp); 1570 else 1571 vput(nd.ni_dvp); 1572 if (nd.ni_vp) 1573 vput(nd.ni_vp); 1574 return (error); 1575 } 1576 1577 /* 1578 * nfs v3 mknod service 1579 */ 1580 int 1581 nfsrv_mknod(nfsd, slp, lwp, mrq) 1582 struct nfsrv_descript *nfsd; 1583 struct nfssvc_sock *slp; 1584 struct lwp *lwp; 1585 struct mbuf **mrq; 1586 { 1587 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1588 struct mbuf *nam = nfsd->nd_nam; 1589 char *dpos = nfsd->nd_dpos; 1590 kauth_cred_t cred = nfsd->nd_cr; 1591 struct vattr va, dirfor, diraft; 1592 u_int32_t *tl; 1593 struct nameidata nd; 1594 int32_t t1; 1595 char *bpos; 1596 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 1597 u_int32_t major, minor; 1598 enum vtype vtyp; 1599 char *cp2; 1600 struct mbuf *mb, *mreq; 1601 struct vnode *vp, *dirp = (struct vnode *)0; 1602 nfsrvfh_t nsfh; 1603 u_quad_t frev; 1604 1605 nd.ni_cnd.cn_nameiop = 0; 1606 nfsm_srvmtofh(&nsfh); 1607 nfsm_srvnamesiz(len); 1608 nd.ni_cnd.cn_cred = cred; 1609 nd.ni_cnd.cn_nameiop = CREATE; 1610 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1611 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 1612 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1613 if (dirp) 1614 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1615 if (error) { 1616 nfsm_reply(NFSX_WCCDATA(1)); 1617 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1618 if (dirp) 1619 vrele(dirp); 1620 return (0); 1621 } 1622 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1623 vtyp = nfsv3tov_type(*tl); 1624 if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) { 1625 error = NFSERR_BADTYPE; 1626 goto abort; 1627 } 1628 VATTR_NULL(&va); 1629 va.va_mode = 0; 1630 nfsm_srvsattr(&va); 1631 if (vtyp == VCHR || vtyp == VBLK) { 1632 dev_t rdev; 1633 1634 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1635 major = fxdr_unsigned(u_int32_t, *tl++); 1636 minor = fxdr_unsigned(u_int32_t, *tl); 1637 rdev = makedev(major, minor); 1638 if (major(rdev) != major || minor(rdev) != minor) { 1639 error = EINVAL; 1640 goto abort; 1641 } 1642 va.va_rdev = rdev; 1643 } 1644 1645 /* 1646 * Iff doesn't exist, create it. 1647 */ 1648 if (nd.ni_vp) { 1649 error = EEXIST; 1650 abort: 1651 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1652 if (nd.ni_dvp == nd.ni_vp) 1653 vrele(nd.ni_dvp); 1654 else 1655 vput(nd.ni_dvp); 1656 if (nd.ni_vp) 1657 vput(nd.ni_vp); 1658 goto out; 1659 } 1660 va.va_type = vtyp; 1661 if (vtyp == VSOCK) { 1662 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1663 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1664 } else { 1665 if (va.va_type != VFIFO && 1666 (error = kauth_authorize_generic(cred, 1667 KAUTH_GENERIC_ISSUSER, NULL))) { 1668 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1669 vput(nd.ni_dvp); 1670 goto out; 1671 } 1672 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1673 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1674 if (error) 1675 goto out; 1676 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1677 vput(nd.ni_vp); 1678 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1679 error = EINVAL; 1680 } 1681 } 1682 out: 1683 vp = nd.ni_vp; 1684 if (!error) { 1685 error = nfsrv_composefh(vp, &nsfh, true); 1686 if (!error) 1687 error = VOP_GETATTR(vp, &va, cred, lwp); 1688 vput(vp); 1689 } 1690 if (dirp) { 1691 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1692 vrele(dirp); 1693 } 1694 nfsm_reply(NFSX_SRVFH(&nsfh, true) + NFSX_POSTOPATTR(1) + 1695 NFSX_WCCDATA(1)); 1696 if (!error) { 1697 nfsm_srvpostop_fh(&nsfh); 1698 nfsm_srvpostop_attr(0, &va); 1699 } 1700 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1701 return (0); 1702 nfsmout: 1703 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1704 if (nd.ni_dvp == nd.ni_vp) 1705 vrele(nd.ni_dvp); 1706 else 1707 vput(nd.ni_dvp); 1708 if (nd.ni_vp) 1709 vput(nd.ni_vp); 1710 if (dirp) 1711 vrele(dirp); 1712 return (error); 1713 } 1714 1715 /* 1716 * nfs remove service 1717 */ 1718 int 1719 nfsrv_remove(nfsd, slp, lwp, mrq) 1720 struct nfsrv_descript *nfsd; 1721 struct nfssvc_sock *slp; 1722 struct lwp *lwp; 1723 struct mbuf **mrq; 1724 { 1725 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1726 struct mbuf *nam = nfsd->nd_nam; 1727 char *dpos = nfsd->nd_dpos; 1728 kauth_cred_t cred = nfsd->nd_cr; 1729 struct nameidata nd; 1730 u_int32_t *tl; 1731 int32_t t1; 1732 char *bpos; 1733 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 1734 int v3 = (nfsd->nd_flag & ND_NFSV3); 1735 char *cp2; 1736 struct mbuf *mb, *mreq; 1737 struct vnode *vp, *dirp; 1738 struct vattr dirfor, diraft; 1739 nfsrvfh_t nsfh; 1740 u_quad_t frev; 1741 1742 #ifndef nolint 1743 vp = (struct vnode *)0; 1744 #endif 1745 nfsm_srvmtofh(&nsfh); 1746 nfsm_srvnamesiz(len); 1747 nd.ni_cnd.cn_cred = cred; 1748 nd.ni_cnd.cn_nameiop = DELETE; 1749 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1750 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 1751 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1752 if (dirp && v3) { 1753 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1754 } 1755 if (!error) { 1756 vp = nd.ni_vp; 1757 if (vp->v_type == VDIR && 1758 (error = kauth_authorize_generic(cred, 1759 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 1760 goto out; 1761 /* 1762 * The root of a mounted filesystem cannot be deleted. 1763 */ 1764 if (vp->v_flag & VROOT) { 1765 error = EBUSY; 1766 goto out; 1767 } 1768 out: 1769 if (!error) { 1770 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1771 nqsrv_getl(vp, ND_WRITE); 1772 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1773 } else { 1774 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1775 if (nd.ni_dvp == vp) 1776 vrele(nd.ni_dvp); 1777 else 1778 vput(nd.ni_dvp); 1779 vput(vp); 1780 } 1781 } 1782 if (dirp) { 1783 if (v3) { 1784 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1785 } 1786 vrele(dirp); 1787 } 1788 nfsm_reply(NFSX_WCCDATA(v3)); 1789 if (v3) { 1790 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1791 return (0); 1792 } 1793 nfsm_srvdone; 1794 } 1795 1796 /* 1797 * nfs rename service 1798 */ 1799 int 1800 nfsrv_rename(nfsd, slp, lwp, mrq) 1801 struct nfsrv_descript *nfsd; 1802 struct nfssvc_sock *slp; 1803 struct lwp *lwp; 1804 struct mbuf **mrq; 1805 { 1806 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1807 struct mbuf *nam = nfsd->nd_nam; 1808 char *dpos = nfsd->nd_dpos; 1809 kauth_cred_t cred = nfsd->nd_cr; 1810 u_int32_t *tl; 1811 int32_t t1; 1812 char *bpos; 1813 int error = 0, cache = 0, fdirfor_ret = 1, fdiraft_ret = 1; 1814 uint32_t len, len2; 1815 int tdirfor_ret = 1, tdiraft_ret = 1; 1816 int v3 = (nfsd->nd_flag & ND_NFSV3); 1817 char *cp2; 1818 struct mbuf *mb, *mreq; 1819 struct nameidata fromnd, tond; 1820 struct vnode *fvp, *tvp, *tdvp; 1821 struct vnode *fdirp = NULL, *tdirp = NULL; 1822 struct vattr fdirfor, fdiraft, tdirfor, tdiraft; 1823 nfsrvfh_t fnsfh, tnsfh; 1824 u_quad_t frev; 1825 uid_t saved_uid; 1826 1827 #ifndef nolint 1828 fvp = (struct vnode *)0; 1829 #endif 1830 fromnd.ni_cnd.cn_nameiop = 0; 1831 tond.ni_cnd.cn_nameiop = 0; 1832 nfsm_srvmtofh(&fnsfh); 1833 nfsm_srvnamesiz(len); 1834 /* 1835 * Remember our original uid so that we can reset cr_uid before 1836 * the second nfs_namei() call, in case it is remapped. 1837 */ 1838 saved_uid = kauth_cred_geteuid(cred); 1839 fromnd.ni_cnd.cn_cred = cred; 1840 fromnd.ni_cnd.cn_nameiop = DELETE; 1841 fromnd.ni_cnd.cn_flags = LOCKPARENT | SAVESTART; 1842 error = nfs_namei(&fromnd, &fnsfh, len, slp, nam, &md, 1843 &dpos, &fdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1844 if (fdirp && v3) { 1845 fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred, lwp); 1846 } 1847 if (error) { 1848 nfsm_reply(2 * NFSX_WCCDATA(v3)); 1849 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 1850 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 1851 if (fdirp) 1852 vrele(fdirp); 1853 return (0); 1854 } 1855 if (fromnd.ni_dvp != fromnd.ni_vp) { 1856 VOP_UNLOCK(fromnd.ni_dvp, 0); 1857 } 1858 fvp = fromnd.ni_vp; 1859 nfsm_srvmtofh(&tnsfh); 1860 if (v3) { 1861 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 1862 len2 = fxdr_unsigned(uint32_t, *tl); 1863 /* len2 will be checked by nfs_namei */ 1864 } 1865 else { 1866 /* NFSv2 */ 1867 nfsm_strsiz(len2, NFS_MAXNAMLEN); 1868 } 1869 kauth_cred_seteuid(cred, saved_uid); 1870 tond.ni_cnd.cn_cred = cred; 1871 tond.ni_cnd.cn_nameiop = RENAME; 1872 tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART; 1873 error = nfs_namei(&tond, &tnsfh, len2, slp, nam, &md, 1874 &dpos, &tdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1875 if (tdirp && v3) { 1876 tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred, lwp); 1877 } 1878 if (error) { 1879 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1880 vrele(fromnd.ni_dvp); 1881 vrele(fvp); 1882 goto out1; 1883 } 1884 tdvp = tond.ni_dvp; 1885 tvp = tond.ni_vp; 1886 if (tvp != NULL) { 1887 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 1888 if (v3) 1889 error = EEXIST; 1890 else 1891 error = EISDIR; 1892 goto out; 1893 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 1894 if (v3) 1895 error = EEXIST; 1896 else 1897 error = ENOTDIR; 1898 goto out; 1899 } 1900 if (tvp->v_type == VDIR && tvp->v_mountedhere) { 1901 if (v3) 1902 error = EXDEV; 1903 else 1904 error = ENOTEMPTY; 1905 goto out; 1906 } 1907 } 1908 if (fvp->v_type == VDIR && fvp->v_mountedhere) { 1909 if (v3) 1910 error = EXDEV; 1911 else 1912 error = ENOTEMPTY; 1913 goto out; 1914 } 1915 if (fvp->v_mount != tdvp->v_mount) { 1916 if (v3) 1917 error = EXDEV; 1918 else 1919 error = ENOTEMPTY; 1920 goto out; 1921 } 1922 if (fvp == tdvp) { 1923 if (v3) 1924 error = EINVAL; 1925 else 1926 error = ENOTEMPTY; 1927 } 1928 /* 1929 * If source is the same as the destination (that is the 1930 * same vnode with the same name in the same directory), 1931 * then there is nothing to do. 1932 */ 1933 if (fvp == tvp && fromnd.ni_dvp == tdvp && 1934 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 1935 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr, 1936 fromnd.ni_cnd.cn_namelen)) 1937 error = -1; 1938 out: 1939 if (!error) { 1940 nqsrv_getl(fromnd.ni_dvp, ND_WRITE); 1941 nqsrv_getl(tdvp, ND_WRITE); 1942 if (tvp) { 1943 nqsrv_getl(tvp, ND_WRITE); 1944 } 1945 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 1946 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 1947 } else { 1948 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 1949 if (tdvp == tvp) 1950 vrele(tdvp); 1951 else 1952 vput(tdvp); 1953 if (tvp) 1954 vput(tvp); 1955 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1956 vrele(fromnd.ni_dvp); 1957 vrele(fvp); 1958 if (error == -1) 1959 error = 0; 1960 } 1961 vrele(tond.ni_startdir); 1962 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 1963 out1: 1964 if (fdirp) { 1965 if (v3) { 1966 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred, lwp); 1967 } 1968 vrele(fdirp); 1969 } 1970 if (tdirp) { 1971 if (v3) { 1972 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred, lwp); 1973 } 1974 vrele(tdirp); 1975 } 1976 vrele(fromnd.ni_startdir); 1977 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 1978 nfsm_reply(2 * NFSX_WCCDATA(v3)); 1979 if (v3) { 1980 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 1981 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 1982 } 1983 return (0); 1984 1985 nfsmout: 1986 if (fdirp) 1987 vrele(fdirp); 1988 #ifdef notdef 1989 if (tdirp) 1990 vrele(tdirp); 1991 #endif 1992 if (tond.ni_cnd.cn_nameiop) { 1993 vrele(tond.ni_startdir); 1994 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 1995 } 1996 if (fromnd.ni_cnd.cn_nameiop) { 1997 vrele(fromnd.ni_startdir); 1998 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 1999 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2000 vrele(fromnd.ni_dvp); 2001 vrele(fvp); 2002 } 2003 return (error); 2004 } 2005 2006 /* 2007 * nfs link service 2008 */ 2009 int 2010 nfsrv_link(nfsd, slp, lwp, mrq) 2011 struct nfsrv_descript *nfsd; 2012 struct nfssvc_sock *slp; 2013 struct lwp *lwp; 2014 struct mbuf **mrq; 2015 { 2016 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2017 struct mbuf *nam = nfsd->nd_nam; 2018 char *dpos = nfsd->nd_dpos; 2019 kauth_cred_t cred = nfsd->nd_cr; 2020 struct nameidata nd; 2021 u_int32_t *tl; 2022 int32_t t1; 2023 char *bpos; 2024 int error = 0, rdonly, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 2025 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3); 2026 char *cp2; 2027 struct mbuf *mb, *mreq; 2028 struct vnode *vp, *xp, *dirp = (struct vnode *)0; 2029 struct vattr dirfor, diraft, at; 2030 nfsrvfh_t nsfh, dnsfh; 2031 u_quad_t frev; 2032 2033 nfsm_srvmtofh(&nsfh); 2034 nfsm_srvmtofh(&dnsfh); 2035 nfsm_srvnamesiz(len); 2036 error = nfsrv_fhtovp(&nsfh, false, &vp, cred, slp, nam, 2037 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 2038 if (error) { 2039 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2040 nfsm_srvpostop_attr(getret, &at); 2041 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2042 return (0); 2043 } 2044 if (vp->v_type == VDIR && (error = kauth_authorize_generic(cred, 2045 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 2046 goto out1; 2047 nd.ni_cnd.cn_cred = cred; 2048 nd.ni_cnd.cn_nameiop = CREATE; 2049 nd.ni_cnd.cn_flags = LOCKPARENT; 2050 error = nfs_namei(&nd, &dnsfh, len, slp, nam, &md, &dpos, 2051 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2052 if (dirp && v3) { 2053 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2054 } 2055 if (error) 2056 goto out1; 2057 xp = nd.ni_vp; 2058 if (xp != NULL) { 2059 error = EEXIST; 2060 goto out; 2061 } 2062 xp = nd.ni_dvp; 2063 if (vp->v_mount != xp->v_mount) 2064 error = EXDEV; 2065 out: 2066 if (!error) { 2067 nqsrv_getl(vp, ND_WRITE); 2068 nqsrv_getl(xp, ND_WRITE); 2069 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2070 } else { 2071 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2072 if (nd.ni_dvp == nd.ni_vp) 2073 vrele(nd.ni_dvp); 2074 else 2075 vput(nd.ni_dvp); 2076 if (nd.ni_vp) 2077 vrele(nd.ni_vp); 2078 } 2079 out1: 2080 if (v3) 2081 getret = VOP_GETATTR(vp, &at, cred, lwp); 2082 if (dirp) { 2083 if (v3) { 2084 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2085 } 2086 vrele(dirp); 2087 } 2088 vrele(vp); 2089 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2090 if (v3) { 2091 nfsm_srvpostop_attr(getret, &at); 2092 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2093 return (0); 2094 } 2095 nfsm_srvdone; 2096 } 2097 2098 /* 2099 * nfs symbolic link service 2100 */ 2101 int 2102 nfsrv_symlink(nfsd, slp, lwp, mrq) 2103 struct nfsrv_descript *nfsd; 2104 struct nfssvc_sock *slp; 2105 struct lwp *lwp; 2106 struct mbuf **mrq; 2107 { 2108 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2109 struct mbuf *nam = nfsd->nd_nam; 2110 char *dpos = nfsd->nd_dpos; 2111 kauth_cred_t cred = nfsd->nd_cr; 2112 struct vattr va, dirfor, diraft; 2113 struct nameidata nd; 2114 u_int32_t *tl; 2115 int32_t t1; 2116 struct nfsv2_sattr *sp; 2117 char *bpos, *pathcp = NULL, *cp2; 2118 struct uio io; 2119 struct iovec iv; 2120 int error = 0, cache = 0, dirfor_ret = 1, diraft_ret = 1; 2121 uint32_t len, len2; 2122 int v3 = (nfsd->nd_flag & ND_NFSV3); 2123 struct mbuf *mb, *mreq; 2124 struct vnode *dirp = (struct vnode *)0; 2125 nfsrvfh_t nsfh; 2126 u_quad_t frev; 2127 2128 nd.ni_cnd.cn_nameiop = 0; 2129 nfsm_srvmtofh(&nsfh); 2130 nfsm_srvnamesiz(len); 2131 nd.ni_cnd.cn_cred = cred; 2132 nd.ni_cnd.cn_nameiop = CREATE; 2133 nd.ni_cnd.cn_flags = LOCKPARENT; 2134 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 2135 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2136 if (dirp && v3) { 2137 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2138 } 2139 if (error) 2140 goto out; 2141 VATTR_NULL(&va); 2142 va.va_type = VLNK; 2143 if (v3) { 2144 va.va_mode = 0; 2145 nfsm_srvsattr(&va); 2146 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 2147 len2 = fxdr_unsigned(uint32_t, *tl); 2148 if (len2 > PATH_MAX) { 2149 /* XXX should check _PC_NO_TRUNC */ 2150 error = ENAMETOOLONG; 2151 goto abortop; 2152 } 2153 } 2154 else { 2155 /* NFSv2 */ 2156 nfsm_strsiz(len2, NFS_MAXPATHLEN); 2157 } 2158 pathcp = malloc(len2 + 1, M_TEMP, M_WAITOK); 2159 iv.iov_base = pathcp; 2160 iv.iov_len = len2; 2161 io.uio_resid = len2; 2162 io.uio_offset = 0; 2163 io.uio_iov = &iv; 2164 io.uio_iovcnt = 1; 2165 io.uio_rw = UIO_READ; 2166 UIO_SETUP_SYSSPACE(&io); 2167 nfsm_mtouio(&io, len2); 2168 if (!v3) { 2169 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 2170 va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode); 2171 } 2172 *(pathcp + len2) = '\0'; 2173 if (nd.ni_vp) { 2174 error = EEXIST; 2175 abortop: 2176 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2177 if (nd.ni_dvp == nd.ni_vp) 2178 vrele(nd.ni_dvp); 2179 else 2180 vput(nd.ni_dvp); 2181 if (nd.ni_vp) 2182 vrele(nd.ni_vp); 2183 goto out; 2184 } 2185 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2186 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp); 2187 if (!error) { 2188 if (v3) { 2189 error = nfsrv_composefh(nd.ni_vp, &nsfh, v3); 2190 if (!error) 2191 error = VOP_GETATTR(nd.ni_vp, &va, cred, lwp); 2192 vput(nd.ni_vp); 2193 } else { 2194 vput(nd.ni_vp); 2195 } 2196 } 2197 out: 2198 if (pathcp) 2199 free(pathcp, M_TEMP); 2200 if (dirp) { 2201 if (v3) { 2202 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2203 } 2204 vrele(dirp); 2205 } 2206 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_POSTOPATTR(v3) + 2207 NFSX_WCCDATA(v3)); 2208 if (v3) { 2209 if (!error) { 2210 nfsm_srvpostop_fh(&nsfh); 2211 nfsm_srvpostop_attr(0, &va); 2212 } 2213 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2214 } 2215 return (0); 2216 nfsmout: 2217 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2218 if (nd.ni_dvp == nd.ni_vp) 2219 vrele(nd.ni_dvp); 2220 else 2221 vput(nd.ni_dvp); 2222 if (nd.ni_vp) 2223 vrele(nd.ni_vp); 2224 if (dirp) 2225 vrele(dirp); 2226 if (pathcp) 2227 free(pathcp, M_TEMP); 2228 return (error); 2229 } 2230 2231 /* 2232 * nfs mkdir service 2233 */ 2234 int 2235 nfsrv_mkdir(nfsd, slp, lwp, mrq) 2236 struct nfsrv_descript *nfsd; 2237 struct nfssvc_sock *slp; 2238 struct lwp *lwp; 2239 struct mbuf **mrq; 2240 { 2241 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2242 struct mbuf *nam = nfsd->nd_nam; 2243 char *dpos = nfsd->nd_dpos; 2244 kauth_cred_t cred = nfsd->nd_cr; 2245 struct vattr va, dirfor, diraft; 2246 struct nfs_fattr *fp; 2247 struct nameidata nd; 2248 char *cp; 2249 u_int32_t *tl; 2250 int32_t t1; 2251 char *bpos; 2252 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 2253 int v3 = (nfsd->nd_flag & ND_NFSV3); 2254 char *cp2; 2255 struct mbuf *mb, *mreq; 2256 struct vnode *vp, *dirp = (struct vnode *)0; 2257 nfsrvfh_t nsfh; 2258 u_quad_t frev; 2259 2260 nfsm_srvmtofh(&nsfh); 2261 nfsm_srvnamesiz(len); 2262 nd.ni_cnd.cn_cred = cred; 2263 nd.ni_cnd.cn_nameiop = CREATE; 2264 nd.ni_cnd.cn_flags = LOCKPARENT; 2265 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 2266 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2267 if (dirp && v3) { 2268 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2269 } 2270 if (error) { 2271 nfsm_reply(NFSX_WCCDATA(v3)); 2272 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2273 if (dirp) 2274 vrele(dirp); 2275 return (0); 2276 } 2277 VATTR_NULL(&va); 2278 if (v3) { 2279 va.va_mode = 0; 2280 nfsm_srvsattr(&va); 2281 } else { 2282 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 2283 va.va_mode = nfstov_mode(*tl++); 2284 } 2285 va.va_type = VDIR; 2286 vp = nd.ni_vp; 2287 if (vp != NULL) { 2288 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2289 if (nd.ni_dvp == vp) 2290 vrele(nd.ni_dvp); 2291 else 2292 vput(nd.ni_dvp); 2293 vrele(vp); 2294 error = EEXIST; 2295 goto out; 2296 } 2297 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2298 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 2299 if (!error) { 2300 vp = nd.ni_vp; 2301 error = nfsrv_composefh(vp, &nsfh, v3); 2302 if (!error) 2303 error = VOP_GETATTR(vp, &va, cred, lwp); 2304 vput(vp); 2305 } 2306 out: 2307 if (dirp) { 2308 if (v3) { 2309 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2310 } 2311 vrele(dirp); 2312 } 2313 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_POSTOPATTR(v3) + 2314 NFSX_WCCDATA(v3)); 2315 if (v3) { 2316 if (!error) { 2317 nfsm_srvpostop_fh(&nsfh); 2318 nfsm_srvpostop_attr(0, &va); 2319 } 2320 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2321 } else { 2322 nfsm_srvfhtom(&nsfh, v3); 2323 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 2324 nfsm_srvfillattr(&va, fp); 2325 } 2326 return (0); 2327 nfsmout: 2328 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2329 if (nd.ni_dvp == nd.ni_vp) 2330 vrele(nd.ni_dvp); 2331 else 2332 vput(nd.ni_dvp); 2333 if (nd.ni_vp) 2334 vrele(nd.ni_vp); 2335 if (dirp) 2336 vrele(dirp); 2337 return (error); 2338 } 2339 2340 /* 2341 * nfs rmdir service 2342 */ 2343 int 2344 nfsrv_rmdir(nfsd, slp, lwp, mrq) 2345 struct nfsrv_descript *nfsd; 2346 struct nfssvc_sock *slp; 2347 struct lwp *lwp; 2348 struct mbuf **mrq; 2349 { 2350 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2351 struct mbuf *nam = nfsd->nd_nam; 2352 char *dpos = nfsd->nd_dpos; 2353 kauth_cred_t cred = nfsd->nd_cr; 2354 u_int32_t *tl; 2355 int32_t t1; 2356 char *bpos; 2357 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 2358 int v3 = (nfsd->nd_flag & ND_NFSV3); 2359 char *cp2; 2360 struct mbuf *mb, *mreq; 2361 struct vnode *vp, *dirp = (struct vnode *)0; 2362 struct vattr dirfor, diraft; 2363 nfsrvfh_t nsfh; 2364 struct nameidata nd; 2365 u_quad_t frev; 2366 2367 nfsm_srvmtofh(&nsfh); 2368 nfsm_srvnamesiz(len); 2369 nd.ni_cnd.cn_cred = cred; 2370 nd.ni_cnd.cn_nameiop = DELETE; 2371 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 2372 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 2373 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2374 if (dirp && v3) { 2375 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2376 } 2377 if (error) { 2378 nfsm_reply(NFSX_WCCDATA(v3)); 2379 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2380 if (dirp) 2381 vrele(dirp); 2382 return (0); 2383 } 2384 vp = nd.ni_vp; 2385 if (vp->v_type != VDIR) { 2386 error = ENOTDIR; 2387 goto out; 2388 } 2389 /* 2390 * No rmdir "." please. 2391 */ 2392 if (nd.ni_dvp == vp) { 2393 error = EINVAL; 2394 goto out; 2395 } 2396 /* 2397 * The root of a mounted filesystem cannot be deleted. 2398 */ 2399 if (vp->v_flag & VROOT) 2400 error = EBUSY; 2401 out: 2402 if (!error) { 2403 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2404 nqsrv_getl(vp, ND_WRITE); 2405 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2406 } else { 2407 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2408 if (nd.ni_dvp == nd.ni_vp) 2409 vrele(nd.ni_dvp); 2410 else 2411 vput(nd.ni_dvp); 2412 vput(vp); 2413 } 2414 if (dirp) { 2415 if (v3) { 2416 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2417 } 2418 vrele(dirp); 2419 } 2420 nfsm_reply(NFSX_WCCDATA(v3)); 2421 if (v3) { 2422 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2423 return (0); 2424 } 2425 nfsm_srvdone; 2426 } 2427 2428 /* 2429 * nfs readdir service 2430 * - mallocs what it thinks is enough to read 2431 * count rounded up to a multiple of NFS_SRVDIRBLKSIZ <= NFS_MAXREADDIR 2432 * - calls VOP_READDIR() 2433 * - loops around building the reply 2434 * if the output generated exceeds count break out of loop 2435 * The nfsm_clget macro is used here so that the reply will be packed 2436 * tightly in mbuf clusters. 2437 * - it only knows that it has encountered eof when the VOP_READDIR() 2438 * reads nothing 2439 * - as such one readdir rpc will return eof false although you are there 2440 * and then the next will return eof 2441 * - it trims out records with d_fileno == 0 2442 * this doesn't matter for Unix clients, but they might confuse clients 2443 * for other os'. 2444 * - it trims out records with d_type == DT_WHT 2445 * these cannot be seen through NFS (unless we extend the protocol) 2446 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less 2447 * than requested, but this may not apply to all filesystems. For 2448 * example, client NFS does not { although it is never remote mounted 2449 * anyhow } 2450 * The alternate call nfsrv_readdirplus() does lookups as well. 2451 * PS: The NFS protocol spec. does not clarify what the "count" byte 2452 * argument is a count of.. just name strings and file id's or the 2453 * entire reply rpc or ... 2454 * I tried just file name and id sizes and it confused the Sun client, 2455 * so I am using the full rpc size now. The "paranoia.." comment refers 2456 * to including the status longwords that are not a part of the dir. 2457 * "entry" structures, but are in the rpc. 2458 */ 2459 2460 #define NFS_SRVDIRBLKSIZ 1024 2461 2462 struct flrep { 2463 nfsuint64 fl_off; 2464 u_int32_t fl_postopok; 2465 u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)]; 2466 u_int32_t fl_fhok; 2467 u_int32_t fl_fhsize; 2468 }; 2469 2470 int 2471 nfsrv_readdir(nfsd, slp, lwp, mrq) 2472 struct nfsrv_descript *nfsd; 2473 struct nfssvc_sock *slp; 2474 struct lwp *lwp; 2475 struct mbuf **mrq; 2476 { 2477 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2478 struct mbuf *nam = nfsd->nd_nam; 2479 char *dpos = nfsd->nd_dpos; 2480 kauth_cred_t cred = nfsd->nd_cr; 2481 char *bp, *be; 2482 struct mbuf *mp; 2483 struct dirent *dp; 2484 char *cp; 2485 u_int32_t *tl; 2486 int32_t t1; 2487 char *bpos; 2488 struct mbuf *mb, *mreq, *mp2; 2489 char *cpos, *cend, *cp2, *rbuf; 2490 struct vnode *vp; 2491 struct vattr at; 2492 nfsrvfh_t nsfh; 2493 struct uio io; 2494 struct iovec iv; 2495 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2496 int siz, cnt, fullsiz, eofflag, rdonly, cache = 0, ncookies; 2497 int v3 = (nfsd->nd_flag & ND_NFSV3); 2498 u_quad_t frev, off, toff, verf; 2499 off_t *cookies = NULL, *cookiep; 2500 nfsuint64 jar; 2501 2502 nfsm_srvmtofh(&nsfh); 2503 if (v3) { 2504 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 2505 toff = fxdr_hyper(tl); 2506 tl += 2; 2507 verf = fxdr_hyper(tl); 2508 tl += 2; 2509 } else { 2510 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2511 toff = fxdr_unsigned(u_quad_t, *tl++); 2512 } 2513 off = toff; 2514 cnt = fxdr_unsigned(int, *tl); 2515 siz = ((cnt + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1)); 2516 xfer = NFS_SRVMAXDATA(nfsd); 2517 if (siz > xfer) 2518 siz = xfer; 2519 fullsiz = siz; 2520 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 2521 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 2522 if (!error && vp->v_type != VDIR) { 2523 error = ENOTDIR; 2524 vput(vp); 2525 } 2526 if (error) { 2527 nfsm_reply(NFSX_UNSIGNED); 2528 nfsm_srvpostop_attr(getret, &at); 2529 return (0); 2530 } 2531 nqsrv_getl(vp, ND_READ); 2532 if (v3) { 2533 error = getret = VOP_GETATTR(vp, &at, cred, lwp); 2534 #ifdef NFS3_STRICTVERF 2535 /* 2536 * XXX This check is too strict for Solaris 2.5 clients. 2537 */ 2538 if (!error && toff && verf != at.va_filerev) 2539 error = NFSERR_BAD_COOKIE; 2540 #endif 2541 } 2542 if (!error) 2543 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0); 2544 if (error) { 2545 vput(vp); 2546 nfsm_reply(NFSX_POSTOPATTR(v3)); 2547 nfsm_srvpostop_attr(getret, &at); 2548 return (0); 2549 } 2550 VOP_UNLOCK(vp, 0); 2551 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2552 again: 2553 iv.iov_base = rbuf; 2554 iv.iov_len = fullsiz; 2555 io.uio_iov = &iv; 2556 io.uio_iovcnt = 1; 2557 io.uio_offset = (off_t)off; 2558 io.uio_resid = fullsiz; 2559 io.uio_rw = UIO_READ; 2560 UIO_SETUP_SYSSPACE(&io); 2561 eofflag = 0; 2562 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2563 2564 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2565 2566 off = (off_t)io.uio_offset; 2567 if (!cookies && !error) 2568 error = NFSERR_PERM; 2569 if (v3) { 2570 getret = VOP_GETATTR(vp, &at, cred, lwp); 2571 if (!error) 2572 error = getret; 2573 } 2574 2575 VOP_UNLOCK(vp, 0); 2576 if (error) { 2577 vrele(vp); 2578 free((void *)rbuf, M_TEMP); 2579 if (cookies) 2580 free((void *)cookies, M_TEMP); 2581 nfsm_reply(NFSX_POSTOPATTR(v3)); 2582 nfsm_srvpostop_attr(getret, &at); 2583 return (0); 2584 } 2585 if (io.uio_resid) { 2586 siz -= io.uio_resid; 2587 2588 /* 2589 * If nothing read, return eof 2590 * rpc reply 2591 */ 2592 if (siz == 0) { 2593 vrele(vp); 2594 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + 2595 2 * NFSX_UNSIGNED); 2596 if (v3) { 2597 nfsm_srvpostop_attr(getret, &at); 2598 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2599 txdr_hyper(at.va_filerev, tl); 2600 tl += 2; 2601 } else 2602 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2603 *tl++ = nfs_false; 2604 *tl = nfs_true; 2605 free((void *)rbuf, M_TEMP); 2606 free((void *)cookies, M_TEMP); 2607 return (0); 2608 } 2609 } 2610 2611 /* 2612 * Check for degenerate cases of nothing useful read. 2613 * If so go try again 2614 */ 2615 cpos = rbuf; 2616 cend = rbuf + siz; 2617 dp = (struct dirent *)cpos; 2618 cookiep = cookies; 2619 2620 while (cpos < cend && ncookies > 0 && 2621 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2622 cpos += dp->d_reclen; 2623 dp = (struct dirent *)cpos; 2624 cookiep++; 2625 ncookies--; 2626 } 2627 if (cpos >= cend || ncookies == 0) { 2628 toff = off; 2629 siz = fullsiz; 2630 free(cookies, M_TEMP); 2631 cookies = NULL; 2632 goto again; 2633 } 2634 2635 len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */ 2636 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz); 2637 if (v3) { 2638 nfsm_srvpostop_attr(getret, &at); 2639 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2640 txdr_hyper(at.va_filerev, tl); 2641 } 2642 mp = mp2 = mb; 2643 bp = bpos; 2644 be = bp + M_TRAILINGSPACE(mp); 2645 2646 /* Loop through the records and build reply */ 2647 while (cpos < cend && ncookies > 0) { 2648 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 2649 nlen = dp->d_namlen; 2650 rem = nfsm_rndup(nlen)-nlen; 2651 len += (4 * NFSX_UNSIGNED + nlen + rem); 2652 if (v3) 2653 len += 2 * NFSX_UNSIGNED; 2654 if (len > cnt) { 2655 eofflag = 0; 2656 break; 2657 } 2658 /* 2659 * Build the directory record xdr from 2660 * the dirent entry. 2661 */ 2662 nfsm_clget; 2663 *tl = nfs_true; 2664 bp += NFSX_UNSIGNED; 2665 if (v3) { 2666 nfsm_clget; 2667 *tl = txdr_unsigned(dp->d_fileno >> 32); 2668 bp += NFSX_UNSIGNED; 2669 } 2670 nfsm_clget; 2671 *tl = txdr_unsigned(dp->d_fileno); 2672 bp += NFSX_UNSIGNED; 2673 nfsm_clget; 2674 *tl = txdr_unsigned(nlen); 2675 bp += NFSX_UNSIGNED; 2676 2677 /* And loop around copying the name */ 2678 xfer = nlen; 2679 cp = dp->d_name; 2680 while (xfer > 0) { 2681 nfsm_clget; 2682 if ((bp+xfer) > be) 2683 tsiz = be-bp; 2684 else 2685 tsiz = xfer; 2686 memcpy(bp, cp, tsiz); 2687 bp += tsiz; 2688 xfer -= tsiz; 2689 if (xfer > 0) 2690 cp += tsiz; 2691 } 2692 /* And null pad to an int32_t boundary */ 2693 for (i = 0; i < rem; i++) 2694 *bp++ = '\0'; 2695 nfsm_clget; 2696 2697 /* Finish off the record */ 2698 txdr_hyper(*cookiep, &jar); 2699 if (v3) { 2700 *tl = jar.nfsuquad[0]; 2701 bp += NFSX_UNSIGNED; 2702 nfsm_clget; 2703 } 2704 *tl = jar.nfsuquad[1]; 2705 bp += NFSX_UNSIGNED; 2706 } 2707 cpos += dp->d_reclen; 2708 dp = (struct dirent *)cpos; 2709 cookiep++; 2710 ncookies--; 2711 } 2712 vrele(vp); 2713 nfsm_clget; 2714 *tl = nfs_false; 2715 bp += NFSX_UNSIGNED; 2716 nfsm_clget; 2717 if (eofflag) 2718 *tl = nfs_true; 2719 else 2720 *tl = nfs_false; 2721 bp += NFSX_UNSIGNED; 2722 if (mp != mb) { 2723 if (bp < be) 2724 mp->m_len = bp - mtod(mp, char *); 2725 } else 2726 mp->m_len += bp - bpos; 2727 free((void *)rbuf, M_TEMP); 2728 free((void *)cookies, M_TEMP); 2729 nfsm_srvdone; 2730 } 2731 2732 int 2733 nfsrv_readdirplus(nfsd, slp, lwp, mrq) 2734 struct nfsrv_descript *nfsd; 2735 struct nfssvc_sock *slp; 2736 struct lwp *lwp; 2737 struct mbuf **mrq; 2738 { 2739 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2740 struct mbuf *nam = nfsd->nd_nam; 2741 char *dpos = nfsd->nd_dpos; 2742 kauth_cred_t cred = nfsd->nd_cr; 2743 char *bp, *be; 2744 struct mbuf *mp; 2745 struct dirent *dp; 2746 char *cp; 2747 u_int32_t *tl; 2748 int32_t t1; 2749 char *bpos; 2750 struct mbuf *mb, *mreq, *mp2; 2751 char *cpos, *cend, *cp2, *rbuf; 2752 struct vnode *vp, *nvp; 2753 struct flrep fl; 2754 nfsrvfh_t nsfh; 2755 struct uio io; 2756 struct iovec iv; 2757 struct vattr va, at, *vap = &va; 2758 struct nfs_fattr *fp; 2759 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2760 int siz, cnt, fullsiz, eofflag, rdonly, cache = 0, dirlen, ncookies; 2761 u_quad_t frev, off, toff, verf; 2762 off_t *cookies = NULL, *cookiep; 2763 2764 nfsm_srvmtofh(&nsfh); 2765 nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED); 2766 toff = fxdr_hyper(tl); 2767 tl += 2; 2768 verf = fxdr_hyper(tl); 2769 tl += 2; 2770 siz = fxdr_unsigned(int, *tl++); 2771 cnt = fxdr_unsigned(int, *tl); 2772 off = toff; 2773 siz = ((siz + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1)); 2774 xfer = NFS_SRVMAXDATA(nfsd); 2775 if (siz > xfer) 2776 siz = xfer; 2777 fullsiz = siz; 2778 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 2779 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 2780 if (!error && vp->v_type != VDIR) { 2781 error = ENOTDIR; 2782 vput(vp); 2783 } 2784 if (error) { 2785 nfsm_reply(NFSX_UNSIGNED); 2786 nfsm_srvpostop_attr(getret, &at); 2787 return (0); 2788 } 2789 error = getret = VOP_GETATTR(vp, &at, cred, lwp); 2790 #ifdef NFS3_STRICTVERF 2791 /* 2792 * XXX This check is too strict for Solaris 2.5 clients. 2793 */ 2794 if (!error && toff && verf != at.va_filerev) 2795 error = NFSERR_BAD_COOKIE; 2796 #endif 2797 if (!error) { 2798 nqsrv_getl(vp, ND_READ); 2799 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0); 2800 } 2801 if (error) { 2802 vput(vp); 2803 nfsm_reply(NFSX_V3POSTOPATTR); 2804 nfsm_srvpostop_attr(getret, &at); 2805 return (0); 2806 } 2807 VOP_UNLOCK(vp, 0); 2808 2809 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2810 again: 2811 iv.iov_base = rbuf; 2812 iv.iov_len = fullsiz; 2813 io.uio_iov = &iv; 2814 io.uio_iovcnt = 1; 2815 io.uio_offset = (off_t)off; 2816 io.uio_resid = fullsiz; 2817 io.uio_rw = UIO_READ; 2818 UIO_SETUP_SYSSPACE(&io); 2819 eofflag = 0; 2820 2821 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2822 2823 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2824 2825 off = (u_quad_t)io.uio_offset; 2826 getret = VOP_GETATTR(vp, &at, cred, lwp); 2827 2828 VOP_UNLOCK(vp, 0); 2829 2830 /* 2831 * If the VGET operation doesn't work for this filesystem, 2832 * we can't support readdirplus. Returning NOTSUPP should 2833 * make clients fall back to plain readdir. 2834 * There's no need to check for VPTOFH as well, we wouldn't 2835 * even be here otherwise. 2836 */ 2837 if (!getret) { 2838 if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp))) 2839 getret = (getret == EOPNOTSUPP) ? 2840 NFSERR_NOTSUPP : NFSERR_IO; 2841 else 2842 vput(nvp); 2843 } 2844 2845 if (!cookies && !error) 2846 error = NFSERR_PERM; 2847 if (!error) 2848 error = getret; 2849 if (error) { 2850 vrele(vp); 2851 if (cookies) 2852 free((void *)cookies, M_TEMP); 2853 free((void *)rbuf, M_TEMP); 2854 nfsm_reply(NFSX_V3POSTOPATTR); 2855 nfsm_srvpostop_attr(getret, &at); 2856 return (0); 2857 } 2858 if (io.uio_resid) { 2859 siz -= io.uio_resid; 2860 2861 /* 2862 * If nothing read, return eof 2863 * rpc reply 2864 */ 2865 if (siz == 0) { 2866 vrele(vp); 2867 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2868 2 * NFSX_UNSIGNED); 2869 nfsm_srvpostop_attr(getret, &at); 2870 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2871 txdr_hyper(at.va_filerev, tl); 2872 tl += 2; 2873 *tl++ = nfs_false; 2874 *tl = nfs_true; 2875 free((void *)cookies, M_TEMP); 2876 free((void *)rbuf, M_TEMP); 2877 return (0); 2878 } 2879 } 2880 2881 /* 2882 * Check for degenerate cases of nothing useful read. 2883 * If so go try again 2884 */ 2885 cpos = rbuf; 2886 cend = rbuf + siz; 2887 dp = (struct dirent *)cpos; 2888 cookiep = cookies; 2889 2890 while (cpos < cend && ncookies > 0 && 2891 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2892 cpos += dp->d_reclen; 2893 dp = (struct dirent *)cpos; 2894 cookiep++; 2895 ncookies--; 2896 } 2897 if (cpos >= cend || ncookies == 0) { 2898 toff = off; 2899 siz = fullsiz; 2900 free(cookies, M_TEMP); 2901 cookies = NULL; 2902 goto again; 2903 } 2904 2905 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED; 2906 nfsm_reply(cnt); 2907 nfsm_srvpostop_attr(getret, &at); 2908 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2909 txdr_hyper(at.va_filerev, tl); 2910 mp = mp2 = mb; 2911 bp = bpos; 2912 be = bp + M_TRAILINGSPACE(mp); 2913 2914 /* Loop through the records and build reply */ 2915 while (cpos < cend && ncookies > 0) { 2916 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 2917 nfsrvfh_t nnsfh; 2918 2919 nlen = dp->d_namlen; 2920 rem = nfsm_rndup(nlen)-nlen; 2921 2922 /* 2923 * For readdir_and_lookup get the vnode using 2924 * the file number. 2925 */ 2926 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp)) 2927 goto invalid; 2928 if (nfsrv_composefh(nvp, &nnsfh, true)) { 2929 vput(nvp); 2930 goto invalid; 2931 } 2932 if (VOP_GETATTR(nvp, vap, cred, lwp)) { 2933 vput(nvp); 2934 goto invalid; 2935 } 2936 vput(nvp); 2937 2938 /* 2939 * If either the dircount or maxcount will be 2940 * exceeded, get out now. Both of these lengths 2941 * are calculated conservatively, including all 2942 * XDR overheads. 2943 */ 2944 len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH + 2945 NFSX_V3POSTOPATTR); 2946 dirlen += (6 * NFSX_UNSIGNED + nlen + rem); 2947 if (len > cnt || dirlen > fullsiz) { 2948 eofflag = 0; 2949 break; 2950 } 2951 2952 /* 2953 * Build the directory record xdr from 2954 * the dirent entry. 2955 */ 2956 fp = (struct nfs_fattr *)&fl.fl_fattr; 2957 nfsm_srvfillattr(vap, fp); 2958 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH); 2959 fl.fl_fhok = nfs_true; 2960 fl.fl_postopok = nfs_true; 2961 txdr_hyper(*cookiep, fl.fl_off.nfsuquad); 2962 2963 nfsm_clget; 2964 *tl = nfs_true; 2965 bp += NFSX_UNSIGNED; 2966 nfsm_clget; 2967 *tl = txdr_unsigned(dp->d_fileno >> 32); 2968 bp += NFSX_UNSIGNED; 2969 nfsm_clget; 2970 *tl = txdr_unsigned(dp->d_fileno); 2971 bp += NFSX_UNSIGNED; 2972 nfsm_clget; 2973 *tl = txdr_unsigned(nlen); 2974 bp += NFSX_UNSIGNED; 2975 2976 /* And loop around copying the name */ 2977 xfer = nlen; 2978 cp = dp->d_name; 2979 while (xfer > 0) { 2980 nfsm_clget; 2981 if ((bp + xfer) > be) 2982 tsiz = be - bp; 2983 else 2984 tsiz = xfer; 2985 memcpy(bp, cp, tsiz); 2986 bp += tsiz; 2987 xfer -= tsiz; 2988 if (xfer > 0) 2989 cp += tsiz; 2990 } 2991 /* And null pad to an int32_t boundary */ 2992 for (i = 0; i < rem; i++) 2993 *bp++ = '\0'; 2994 2995 /* 2996 * Now copy the flrep structure out. 2997 */ 2998 xfer = sizeof(struct flrep); 2999 cp = (void *)&fl; 3000 while (xfer > 0) { 3001 nfsm_clget; 3002 if ((bp + xfer) > be) 3003 tsiz = be - bp; 3004 else 3005 tsiz = xfer; 3006 memcpy(bp, cp, tsiz); 3007 bp += tsiz; 3008 xfer -= tsiz; 3009 if (xfer > 0) 3010 cp += tsiz; 3011 } 3012 3013 /* 3014 * ... and filehandle. 3015 */ 3016 xfer = NFSRVFH_SIZE(&nnsfh); 3017 cp = NFSRVFH_DATA(&nnsfh); 3018 while (xfer > 0) { 3019 nfsm_clget; 3020 if ((bp + xfer) > be) 3021 tsiz = be - bp; 3022 else 3023 tsiz = xfer; 3024 memcpy(bp, cp, tsiz); 3025 bp += tsiz; 3026 xfer -= tsiz; 3027 if (xfer > 0) 3028 cp += tsiz; 3029 } 3030 } 3031 invalid: 3032 cpos += dp->d_reclen; 3033 dp = (struct dirent *)cpos; 3034 cookiep++; 3035 ncookies--; 3036 } 3037 vrele(vp); 3038 nfsm_clget; 3039 *tl = nfs_false; 3040 bp += NFSX_UNSIGNED; 3041 nfsm_clget; 3042 if (eofflag) 3043 *tl = nfs_true; 3044 else 3045 *tl = nfs_false; 3046 bp += NFSX_UNSIGNED; 3047 if (mp != mb) { 3048 if (bp < be) 3049 mp->m_len = bp - mtod(mp, char *); 3050 } else 3051 mp->m_len += bp - bpos; 3052 free((void *)cookies, M_TEMP); 3053 free((void *)rbuf, M_TEMP); 3054 nfsm_srvdone; 3055 } 3056 3057 /* 3058 * nfs commit service 3059 */ 3060 int 3061 nfsrv_commit(nfsd, slp, lwp, mrq) 3062 struct nfsrv_descript *nfsd; 3063 struct nfssvc_sock *slp; 3064 struct lwp *lwp; 3065 struct mbuf **mrq; 3066 { 3067 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3068 struct mbuf *nam = nfsd->nd_nam; 3069 char *dpos = nfsd->nd_dpos; 3070 kauth_cred_t cred = nfsd->nd_cr; 3071 struct vattr bfor, aft; 3072 struct vnode *vp; 3073 nfsrvfh_t nsfh; 3074 u_int32_t *tl; 3075 int32_t t1; 3076 char *bpos; 3077 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cache = 0; 3078 uint32_t cnt; 3079 char *cp2; 3080 struct mbuf *mb, *mreq; 3081 u_quad_t frev, off, end; 3082 3083 nfsm_srvmtofh(&nsfh); 3084 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED); 3085 3086 off = fxdr_hyper(tl); 3087 tl += 2; 3088 cnt = fxdr_unsigned(uint32_t, *tl); 3089 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3090 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3091 if (error) { 3092 nfsm_reply(2 * NFSX_UNSIGNED); 3093 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3094 return (0); 3095 } 3096 for_ret = VOP_GETATTR(vp, &bfor, cred, lwp); 3097 end = (cnt > 0) ? off + cnt : vp->v_size; 3098 if (end < off || end > vp->v_size) 3099 end = vp->v_size; 3100 if (off < vp->v_size) 3101 error = VOP_FSYNC(vp, cred, FSYNC_WAIT, off, end, lwp); 3102 /* else error == 0, from nfsrv_fhtovp() */ 3103 aft_ret = VOP_GETATTR(vp, &aft, cred, lwp); 3104 vput(vp); 3105 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF); 3106 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3107 if (!error) { 3108 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF); 3109 *tl++ = txdr_unsigned(boottime.tv_sec); 3110 *tl = txdr_unsigned(boottime.tv_usec); 3111 } else { 3112 return (0); 3113 } 3114 nfsm_srvdone; 3115 } 3116 3117 /* 3118 * nfs statfs service 3119 */ 3120 int 3121 nfsrv_statfs(nfsd, slp, lwp, mrq) 3122 struct nfsrv_descript *nfsd; 3123 struct nfssvc_sock *slp; 3124 struct lwp *lwp; 3125 struct mbuf **mrq; 3126 { 3127 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3128 struct mbuf *nam = nfsd->nd_nam; 3129 char *dpos = nfsd->nd_dpos; 3130 kauth_cred_t cred = nfsd->nd_cr; 3131 struct statvfs *sf = NULL; 3132 struct nfs_statfs *sfp; 3133 u_int32_t *tl; 3134 int32_t t1; 3135 char *bpos; 3136 int error = 0, rdonly, cache = 0, getret = 1; 3137 int v3 = (nfsd->nd_flag & ND_NFSV3); 3138 char *cp2; 3139 struct mbuf *mb, *mreq; 3140 struct vnode *vp; 3141 struct vattr at; 3142 nfsrvfh_t nsfh; 3143 u_quad_t frev, tval; 3144 3145 nfsm_srvmtofh(&nsfh); 3146 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3147 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3148 if (error) { 3149 nfsm_reply(NFSX_UNSIGNED); 3150 nfsm_srvpostop_attr(getret, &at); 3151 return (0); 3152 } 3153 sf = malloc(sizeof(*sf), M_TEMP, M_WAITOK); 3154 error = VFS_STATVFS(vp->v_mount, sf, lwp); 3155 getret = VOP_GETATTR(vp, &at, cred, lwp); 3156 vput(vp); 3157 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3)); 3158 if (v3) 3159 nfsm_srvpostop_attr(getret, &at); 3160 if (error) { 3161 free(sf, M_TEMP); 3162 return (0); 3163 } 3164 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3)); 3165 if (v3) { 3166 tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_frsize); 3167 txdr_hyper(tval, &sfp->sf_tbytes); 3168 tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_frsize); 3169 txdr_hyper(tval, &sfp->sf_fbytes); 3170 tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_frsize); 3171 txdr_hyper(tval, &sfp->sf_abytes); 3172 tval = (u_quad_t)sf->f_files; 3173 txdr_hyper(tval, &sfp->sf_tfiles); 3174 tval = (u_quad_t)sf->f_ffree; 3175 txdr_hyper(tval, &sfp->sf_ffiles); 3176 txdr_hyper(tval, &sfp->sf_afiles); 3177 sfp->sf_invarsec = 0; 3178 } else { 3179 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA); 3180 sfp->sf_bsize = txdr_unsigned(sf->f_frsize); 3181 sfp->sf_blocks = txdr_unsigned(sf->f_blocks); 3182 sfp->sf_bfree = txdr_unsigned(sf->f_bfree); 3183 sfp->sf_bavail = txdr_unsigned(sf->f_bavail); 3184 } 3185 nfsmout: 3186 if (sf) 3187 free(sf, M_TEMP); 3188 return error; 3189 } 3190 3191 /* 3192 * nfs fsinfo service 3193 */ 3194 int 3195 nfsrv_fsinfo(nfsd, slp, lwp, mrq) 3196 struct nfsrv_descript *nfsd; 3197 struct nfssvc_sock *slp; 3198 struct lwp *lwp; 3199 struct mbuf **mrq; 3200 { 3201 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3202 struct mbuf *nam = nfsd->nd_nam; 3203 char *dpos = nfsd->nd_dpos; 3204 kauth_cred_t cred = nfsd->nd_cr; 3205 u_int32_t *tl; 3206 struct nfsv3_fsinfo *sip; 3207 int32_t t1; 3208 char *bpos; 3209 int error = 0, rdonly, cache = 0, getret = 1; 3210 uint32_t maxdata; 3211 char *cp2; 3212 struct mbuf *mb, *mreq; 3213 struct vnode *vp; 3214 struct vattr at; 3215 nfsrvfh_t nsfh; 3216 u_quad_t frev, maxfsize; 3217 struct statvfs *sb; 3218 3219 nfsm_srvmtofh(&nsfh); 3220 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3221 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3222 if (error) { 3223 nfsm_reply(NFSX_UNSIGNED); 3224 nfsm_srvpostop_attr(getret, &at); 3225 return (0); 3226 } 3227 3228 /* XXX Try to make a guess on the max file size. */ 3229 sb = malloc(sizeof(*sb), M_TEMP, M_WAITOK); 3230 VFS_STATVFS(vp->v_mount, sb, (struct lwp *)0); 3231 maxfsize = (u_quad_t)0x80000000 * sb->f_frsize - 1; 3232 free(sb, M_TEMP); 3233 3234 getret = VOP_GETATTR(vp, &at, cred, lwp); 3235 vput(vp); 3236 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO); 3237 nfsm_srvpostop_attr(getret, &at); 3238 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO); 3239 3240 /* 3241 * XXX 3242 * There should be file system VFS OP(s) to get this information. 3243 * For now, assume ufs. 3244 */ 3245 if (slp->ns_so->so_type == SOCK_DGRAM) 3246 maxdata = NFS_MAXDGRAMDATA; 3247 else 3248 maxdata = NFS_MAXDATA; 3249 sip->fs_rtmax = txdr_unsigned(maxdata); 3250 sip->fs_rtpref = txdr_unsigned(maxdata); 3251 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE); 3252 sip->fs_wtmax = txdr_unsigned(maxdata); 3253 sip->fs_wtpref = txdr_unsigned(maxdata); 3254 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE); 3255 sip->fs_dtpref = txdr_unsigned(maxdata); 3256 txdr_hyper(maxfsize, &sip->fs_maxfilesize); 3257 sip->fs_timedelta.nfsv3_sec = 0; 3258 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1); 3259 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK | 3260 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS | 3261 NFSV3FSINFO_CANSETTIME); 3262 nfsm_srvdone; 3263 } 3264 3265 /* 3266 * nfs pathconf service 3267 */ 3268 int 3269 nfsrv_pathconf(nfsd, slp, lwp, mrq) 3270 struct nfsrv_descript *nfsd; 3271 struct nfssvc_sock *slp; 3272 struct lwp *lwp; 3273 struct mbuf **mrq; 3274 { 3275 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3276 struct mbuf *nam = nfsd->nd_nam; 3277 char *dpos = nfsd->nd_dpos; 3278 kauth_cred_t cred = nfsd->nd_cr; 3279 u_int32_t *tl; 3280 struct nfsv3_pathconf *pc; 3281 int32_t t1; 3282 char *bpos; 3283 int error = 0, rdonly, cache = 0, getret = 1; 3284 register_t linkmax, namemax, chownres, notrunc; 3285 char *cp2; 3286 struct mbuf *mb, *mreq; 3287 struct vnode *vp; 3288 struct vattr at; 3289 nfsrvfh_t nsfh; 3290 u_quad_t frev; 3291 3292 nfsm_srvmtofh(&nsfh); 3293 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3294 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3295 if (error) { 3296 nfsm_reply(NFSX_UNSIGNED); 3297 nfsm_srvpostop_attr(getret, &at); 3298 return (0); 3299 } 3300 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax); 3301 if (!error) 3302 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax); 3303 if (!error) 3304 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres); 3305 if (!error) 3306 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc); 3307 getret = VOP_GETATTR(vp, &at, cred, lwp); 3308 vput(vp); 3309 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF); 3310 nfsm_srvpostop_attr(getret, &at); 3311 if (error) 3312 return (0); 3313 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF); 3314 3315 pc->pc_linkmax = txdr_unsigned(linkmax); 3316 pc->pc_namemax = txdr_unsigned(namemax); 3317 pc->pc_notrunc = txdr_unsigned(notrunc); 3318 pc->pc_chownrestricted = txdr_unsigned(chownres); 3319 3320 /* 3321 * These should probably be supported by VOP_PATHCONF(), but 3322 * until msdosfs is exportable (why would you want to?), the 3323 * Unix defaults should be ok. 3324 */ 3325 pc->pc_caseinsensitive = nfs_false; 3326 pc->pc_casepreserving = nfs_true; 3327 nfsm_srvdone; 3328 } 3329 3330 /* 3331 * Null operation, used by clients to ping server 3332 */ 3333 /* ARGSUSED */ 3334 int 3335 nfsrv_null(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, 3336 struct lwp *lwp, struct mbuf **mrq) 3337 { 3338 struct mbuf *mrep = nfsd->nd_mrep; 3339 char *bpos; 3340 int error = NFSERR_RETVOID, cache = 0; 3341 struct mbuf *mb, *mreq; 3342 u_quad_t frev; 3343 3344 nfsm_reply(0); 3345 return (0); 3346 } 3347 3348 /* 3349 * No operation, used for obsolete procedures 3350 */ 3351 /* ARGSUSED */ 3352 int 3353 nfsrv_noop(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, 3354 struct lwp *lwp, struct mbuf **mrq) 3355 { 3356 struct mbuf *mrep = nfsd->nd_mrep; 3357 char *bpos; 3358 int error, cache = 0; 3359 struct mbuf *mb, *mreq; 3360 u_quad_t frev; 3361 3362 if (nfsd->nd_repstat) 3363 error = nfsd->nd_repstat; 3364 else 3365 error = EPROCUNAVAIL; 3366 nfsm_reply(0); 3367 return (0); 3368 } 3369 3370 /* 3371 * Perform access checking for vnodes obtained from file handles that would 3372 * refer to files already opened by a Unix client. You cannot just use 3373 * vn_writechk() and VOP_ACCESS() for two reasons. 3374 * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case 3375 * 2 - The owner is to be given access irrespective of mode bits for some 3376 * operations, so that processes that chmod after opening a file don't 3377 * break. I don't like this because it opens a security hole, but since 3378 * the nfs server opens a security hole the size of a barn door anyhow, 3379 * what the heck. 3380 * 3381 * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS() 3382 * will return EPERM instead of EACCESS. EPERM is always an error. 3383 */ 3384 int 3385 nfsrv_access(vp, flags, cred, rdonly, lwp, override) 3386 struct vnode *vp; 3387 int flags; 3388 kauth_cred_t cred; 3389 int rdonly; 3390 struct lwp *lwp; 3391 int override; 3392 { 3393 struct vattr vattr; 3394 int error; 3395 if (flags & VWRITE) { 3396 /* Just vn_writechk() changed to check rdonly */ 3397 /* 3398 * Disallow write attempts on read-only file systems; 3399 * unless the file is a socket or a block or character 3400 * device resident on the file system. 3401 */ 3402 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 3403 switch (vp->v_type) { 3404 case VREG: 3405 case VDIR: 3406 case VLNK: 3407 return (EROFS); 3408 default: 3409 break; 3410 } 3411 } 3412 3413 /* 3414 * If the vnode is in use as a process's text, 3415 * we can't allow writing. 3416 */ 3417 if (vp->v_flag & VTEXT) 3418 return (ETXTBSY); 3419 } 3420 error = VOP_GETATTR(vp, &vattr, cred, lwp); 3421 if (error) 3422 return (error); 3423 error = VOP_ACCESS(vp, flags, cred, lwp); 3424 /* 3425 * Allow certain operations for the owner (reads and writes 3426 * on files that are already open). 3427 */ 3428 if (override && error == EACCES && kauth_cred_geteuid(cred) == vattr.va_uid) 3429 error = 0; 3430 return error; 3431 } 3432
Indexes created Wed Sep 24 05:09:52 GMT 2025