nfs_serv.c revision 1.95.2.6
1 /* $NetBSD: nfs_serv.c,v 1.95.2.6 2007/10/27 11:36:16 yamt Exp $ */ 2 3 /* 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * This code is derived from software contributed to Berkeley by 8 * Rick Macklem at The University of Guelph. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * @(#)nfs_serv.c 8.8 (Berkeley) 7/31/95 35 */ 36 37 /* 38 * nfs version 2 and 3 server calls to vnode ops 39 * - these routines generally have 3 phases 40 * 1 - break down and validate rpc request in mbuf list 41 * 2 - do the vnode ops for the request 42 * (surprisingly ?? many are very similar to syscalls in vfs_syscalls.c) 43 * 3 - build the rpc reply in an mbuf list 44 * nb: 45 * - do not mix the phases, since the nfsm_?? macros can return failures 46 * on a bad rpc or similar and do not do any vrele() or vput()'s 47 * 48 * - the nfsm_reply() macro generates an nfs rpc reply with the nfs 49 * error number iff error != 0 whereas 50 * returning an error from the server function implies a fatal error 51 * such as a badly constructed rpc request that should be dropped without 52 * a reply. 53 * For Version 3, nfsm_reply() does not return for the error case, since 54 * most version 3 rpcs return more than the status for error cases. 55 */ 56 57 #include <sys/cdefs.h> 58 __KERNEL_RCSID(0, "$NetBSD: nfs_serv.c,v 1.95.2.6 2007/10/27 11:36:16 yamt Exp $"); 59 60 #include <sys/param.h> 61 #include <sys/systm.h> 62 #include <sys/proc.h> 63 #include <sys/file.h> 64 #include <sys/namei.h> 65 #include <sys/vnode.h> 66 #include <sys/mount.h> 67 #include <sys/socket.h> 68 #include <sys/socketvar.h> 69 #include <sys/mbuf.h> 70 #include <sys/dirent.h> 71 #include <sys/stat.h> 72 #include <sys/kernel.h> 73 #include <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 #if defined(__HAVE_LAZY_MBUF) 685 m->m_flags |= M_EXT_LAZY; 686 m->m_ext.ext_flags |= M_EXT_LAZY; 687 #else /* defined(__HAVE_LAZY_MBUF) */ 688 /* map pages */ 689 for (i = 0; i < npages; i++) { 690 pmap_kenter_pa(lva, VM_PAGE_TO_PHYS(pgpp[i]), 691 VM_PROT_READ); 692 lva += PAGE_SIZE; 693 } 694 695 pmap_update(pmap_kernel()); 696 #endif /* defined(__HAVE_LAZY_MBUF) */ 697 698 mb->m_next = m; 699 mb = m; 700 error = 0; 701 uiop->uio_resid = 0; 702 } else { 703 struct iovec *iv; 704 struct iovec *iv2; 705 struct mbuf *m2; 706 int siz; 707 loan_fail: 708 /* 709 * Generate the mbuf list with the uio_iov ref. to it. 710 */ 711 i = 0; 712 m = m2 = mb; 713 while (left > 0) { 714 siz = min(M_TRAILINGSPACE(m), left); 715 if (siz > 0) { 716 left -= siz; 717 i++; 718 } 719 if (left > 0) { 720 m = m_get(M_WAIT, MT_DATA); 721 MCLAIM(m, &nfs_mowner); 722 m_clget(m, M_WAIT); 723 m->m_len = 0; 724 m2->m_next = m; 725 m2 = m; 726 } 727 } 728 iv = malloc(i * sizeof(struct iovec), M_TEMP, M_WAITOK); 729 uiop->uio_iov = iv2 = iv; 730 m = mb; 731 left = cnt; 732 i = 0; 733 while (left > 0) { 734 if (m == NULL) 735 panic("nfsrv_read iov"); 736 siz = min(M_TRAILINGSPACE(m), left); 737 if (siz > 0) { 738 iv->iov_base = mtod(m, char *) + 739 m->m_len; 740 iv->iov_len = siz; 741 m->m_len += siz; 742 left -= siz; 743 iv++; 744 i++; 745 } 746 m = m->m_next; 747 } 748 uiop->uio_iovcnt = i; 749 uiop->uio_offset = off; 750 uiop->uio_resid = cnt; 751 uiop->uio_rw = UIO_READ; 752 UIO_SETUP_SYSSPACE(uiop); 753 error = VOP_READ(vp, uiop, IO_NODELOCKED, cred); 754 free((void *)iv2, M_TEMP); 755 } 756 read_error: 757 if (error || (getret = VOP_GETATTR(vp, &va, cred, lwp)) != 0){ 758 if (!error) 759 error = getret; 760 m_freem(mreq); 761 vput(vp); 762 nfsm_reply(NFSX_POSTOPATTR(v3)); 763 nfsm_srvpostop_attr(getret, &va); 764 return (0); 765 } 766 } else { 767 uiop->uio_resid = 0; 768 } 769 vput(vp); 770 nfsm_srvfillattr(&va, fp); 771 len -= uiop->uio_resid; 772 padlen = nfsm_padlen(len); 773 if (uiop->uio_resid || padlen) 774 nfs_zeropad(mb, uiop->uio_resid, padlen); 775 if (v3) { 776 /* count */ 777 *tl++ = txdr_unsigned(len); 778 /* eof */ 779 if (off + len >= va.va_size) 780 *tl++ = nfs_true; 781 else 782 *tl++ = nfs_false; 783 } 784 *tl = txdr_unsigned(len); 785 nfsm_srvdone; 786 } 787 788 /* 789 * nfs write service 790 */ 791 int 792 nfsrv_write(nfsd, slp, lwp, mrq) 793 struct nfsrv_descript *nfsd; 794 struct nfssvc_sock *slp; 795 struct lwp *lwp; 796 struct mbuf **mrq; 797 { 798 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 799 struct mbuf *nam = nfsd->nd_nam; 800 char *dpos = nfsd->nd_dpos; 801 kauth_cred_t cred = nfsd->nd_cr; 802 struct iovec *ivp; 803 int i, cnt; 804 struct mbuf *mp; 805 struct nfs_fattr *fp; 806 struct iovec *iv; 807 struct vattr va, forat; 808 u_int32_t *tl; 809 int32_t t1; 810 char *bpos; 811 int error = 0, rdonly, cache = 0, len, forat_ret = 1; 812 int ioflags, aftat_ret = 1, retlen, zeroing, adjust; 813 int stable = NFSV3WRITE_FILESYNC; 814 int v3 = (nfsd->nd_flag & ND_NFSV3); 815 char *cp2; 816 struct mbuf *mb, *mreq; 817 struct vnode *vp; 818 nfsrvfh_t nsfh; 819 struct uio io, *uiop = &io; 820 off_t off; 821 u_quad_t frev; 822 823 if (mrep == NULL) { 824 *mrq = NULL; 825 return (0); 826 } 827 nfsm_srvmtofh(&nsfh); 828 if (v3) { 829 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 830 off = fxdr_hyper(tl); 831 tl += 3; 832 stable = fxdr_unsigned(int, *tl++); 833 } else { 834 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 835 off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 836 tl += 2; 837 } 838 retlen = len = fxdr_unsigned(int32_t, *tl); 839 cnt = i = 0; 840 841 /* 842 * For NFS Version 2, it is not obvious what a write of zero length 843 * should do, but I might as well be consistent with Version 3, 844 * which is to return ok so long as there are no permission problems. 845 */ 846 if (len > 0) { 847 zeroing = 1; 848 mp = mrep; 849 while (mp) { 850 if (mp == md) { 851 zeroing = 0; 852 adjust = dpos - mtod(mp, char *); 853 mp->m_len -= adjust; 854 if (mp->m_len > 0 && adjust > 0) 855 NFSMADV(mp, adjust); 856 } 857 if (zeroing) 858 mp->m_len = 0; 859 else if (mp->m_len > 0) { 860 i += mp->m_len; 861 if (i > len) { 862 mp->m_len -= (i - len); 863 zeroing = 1; 864 } 865 if (mp->m_len > 0) 866 cnt++; 867 } 868 mp = mp->m_next; 869 } 870 } 871 if (len > NFS_MAXDATA || len < 0 || i < len) { 872 error = EIO; 873 nfsm_reply(2 * NFSX_UNSIGNED); 874 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 875 return (0); 876 } 877 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 878 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 879 if (error) { 880 nfsm_reply(2 * NFSX_UNSIGNED); 881 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 882 return (0); 883 } 884 if (v3) 885 forat_ret = VOP_GETATTR(vp, &forat, cred, lwp); 886 if (vp->v_type != VREG) { 887 if (v3) 888 error = EINVAL; 889 else 890 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 891 } 892 if (!error) { 893 nqsrv_getl(vp, ND_WRITE); 894 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1); 895 } 896 if (error) { 897 vput(vp); 898 nfsm_reply(NFSX_WCCDATA(v3)); 899 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 900 return (0); 901 } 902 903 if (len > 0) { 904 ivp = malloc(cnt * sizeof (struct iovec), M_TEMP, M_WAITOK); 905 uiop->uio_iov = iv = ivp; 906 uiop->uio_iovcnt = cnt; 907 mp = mrep; 908 while (mp) { 909 if (mp->m_len > 0) { 910 ivp->iov_base = mtod(mp, void *); 911 ivp->iov_len = mp->m_len; 912 ivp++; 913 } 914 mp = mp->m_next; 915 } 916 917 /* 918 * XXX 919 * The IO_METASYNC flag indicates that all metadata (and not 920 * just enough to ensure data integrity) must be written to 921 * stable storage synchronously. 922 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.) 923 */ 924 if (stable == NFSV3WRITE_UNSTABLE) 925 ioflags = IO_NODELOCKED; 926 else if (stable == NFSV3WRITE_DATASYNC) 927 ioflags = (IO_SYNC | IO_NODELOCKED); 928 else 929 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 930 uiop->uio_resid = len; 931 uiop->uio_rw = UIO_WRITE; 932 uiop->uio_offset = off; 933 UIO_SETUP_SYSSPACE(uiop); 934 error = VOP_WRITE(vp, uiop, ioflags, cred); 935 nfsstats.srvvop_writes++; 936 free(iv, M_TEMP); 937 } 938 aftat_ret = VOP_GETATTR(vp, &va, cred, lwp); 939 vput(vp); 940 if (!error) 941 error = aftat_ret; 942 nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) + 943 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3)); 944 if (v3) { 945 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 946 if (error) 947 return (0); 948 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 949 *tl++ = txdr_unsigned(retlen); 950 if (stable == NFSV3WRITE_UNSTABLE) 951 *tl++ = txdr_unsigned(stable); 952 else 953 *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC); 954 /* 955 * Actually, there is no need to txdr these fields, 956 * but it may make the values more human readable, 957 * for debugging purposes. 958 */ 959 *tl++ = txdr_unsigned(boottime.tv_sec); 960 *tl = txdr_unsigned(boottime.tv_usec); 961 } else { 962 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 963 nfsm_srvfillattr(&va, fp); 964 } 965 nfsm_srvdone; 966 } 967 968 /* 969 * XXX elad: the original NFSW_SAMECRED() macro also made sure the 970 * two nd_flag fields of the descriptors contained 971 * ND_KERBAUTH. 972 */ 973 static int 974 nfsrv_samecred(kauth_cred_t cred1, kauth_cred_t cred2) 975 { 976 int i, do_ngroups; 977 978 if (kauth_cred_geteuid(cred1) != kauth_cred_geteuid(cred2)) 979 return (0); 980 if (kauth_cred_ngroups(cred1) != kauth_cred_ngroups(cred2)) 981 return (0); 982 do_ngroups = kauth_cred_ngroups(cred1); 983 for (i = 0; i < do_ngroups; i++) 984 if (kauth_cred_group(cred1, i) != 985 kauth_cred_group(cred2, i)) 986 return (0); 987 988 return (1); 989 } 990 991 static struct nfsrvw_delayhash * 992 nfsrv_nwdelayhash(struct nfssvc_sock *slp, const nfsrvfh_t *nsfh) 993 { 994 uint32_t hash; 995 996 hash = hash32_buf(NFSRVFH_DATA(nsfh), NFSRVFH_SIZE(nsfh), 997 HASH32_BUF_INIT); 998 return &slp->ns_wdelayhashtbl[hash % NFS_WDELAYHASHSIZ]; 999 } 1000 1001 /* 1002 * NFS write service with write gathering support. Called when 1003 * nfsrvw_procrastinate > 0. 1004 * See: Chet Juszczak, "Improving the Write Performance of an NFS Server", 1005 * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco, 1006 * Jan. 1994. 1007 */ 1008 int 1009 nfsrv_writegather(ndp, slp, lwp, mrq) 1010 struct nfsrv_descript **ndp; 1011 struct nfssvc_sock *slp; 1012 struct lwp *lwp; 1013 struct mbuf **mrq; 1014 { 1015 struct timeval now; 1016 struct iovec *ivp; 1017 struct mbuf *mp; 1018 struct nfsrv_descript *wp, *nfsd, *owp, *swp; 1019 struct nfs_fattr *fp; 1020 int i = 0; 1021 struct iovec *iov; 1022 struct nfsrvw_delayhash *wpp; 1023 kauth_cred_t cred; 1024 struct vattr va, forat; 1025 u_int32_t *tl; 1026 int32_t t1; 1027 char *bpos, *dpos; 1028 int error = 0, rdonly, cache = 0, len = 0, forat_ret = 1; 1029 int ioflags, aftat_ret = 1, s, adjust, v3, zeroing; 1030 char *cp2; 1031 struct mbuf *mb, *mreq, *mrep, *md; 1032 struct vnode *vp; 1033 struct uio io, *uiop = &io; 1034 u_quad_t frev, cur_usec; 1035 1036 *mrq = NULL; 1037 if (*ndp) { 1038 nfsd = *ndp; 1039 *ndp = NULL; 1040 mrep = nfsd->nd_mrep; 1041 md = nfsd->nd_md; 1042 dpos = nfsd->nd_dpos; 1043 cred = nfsd->nd_cr; 1044 v3 = (nfsd->nd_flag & ND_NFSV3); 1045 LIST_INIT(&nfsd->nd_coalesce); 1046 nfsd->nd_mreq = NULL; 1047 nfsd->nd_stable = NFSV3WRITE_FILESYNC; 1048 getmicrotime(&now); 1049 cur_usec = (u_quad_t)now.tv_sec * 1000000 + (u_quad_t)now.tv_usec; 1050 nfsd->nd_time = cur_usec + nfsrvw_procrastinate; 1051 1052 /* 1053 * Now, get the write header.. 1054 */ 1055 nfsm_srvmtofh(&nfsd->nd_fh); 1056 if (v3) { 1057 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 1058 nfsd->nd_off = fxdr_hyper(tl); 1059 tl += 3; 1060 nfsd->nd_stable = fxdr_unsigned(int, *tl++); 1061 } else { 1062 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1063 nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 1064 tl += 2; 1065 } 1066 len = fxdr_unsigned(int32_t, *tl); 1067 nfsd->nd_len = len; 1068 nfsd->nd_eoff = nfsd->nd_off + len; 1069 1070 /* 1071 * Trim the header out of the mbuf list and trim off any trailing 1072 * junk so that the mbuf list has only the write data. 1073 */ 1074 zeroing = 1; 1075 i = 0; 1076 mp = mrep; 1077 while (mp) { 1078 if (mp == md) { 1079 zeroing = 0; 1080 adjust = dpos - mtod(mp, char *); 1081 mp->m_len -= adjust; 1082 if (mp->m_len > 0 && adjust > 0) 1083 NFSMADV(mp, adjust); 1084 } 1085 if (zeroing) 1086 mp->m_len = 0; 1087 else { 1088 i += mp->m_len; 1089 if (i > len) { 1090 mp->m_len -= (i - len); 1091 zeroing = 1; 1092 } 1093 } 1094 mp = mp->m_next; 1095 } 1096 if (len > NFS_MAXDATA || len < 0 || i < len) { 1097 nfsmout: 1098 m_freem(mrep); 1099 error = EIO; 1100 nfsm_writereply(2 * NFSX_UNSIGNED, v3); 1101 if (v3) 1102 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1103 nfsd->nd_mreq = mreq; 1104 nfsd->nd_mrep = NULL; 1105 nfsd->nd_time = 0; 1106 } 1107 1108 /* 1109 * Add this entry to the hash and time queues. 1110 */ 1111 s = splsoftclock(); 1112 owp = NULL; 1113 wp = LIST_FIRST(&slp->ns_tq); 1114 while (wp && wp->nd_time < nfsd->nd_time) { 1115 owp = wp; 1116 wp = LIST_NEXT(wp, nd_tq); 1117 } 1118 if (owp) { 1119 LIST_INSERT_AFTER(owp, nfsd, nd_tq); 1120 } else { 1121 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1122 } 1123 if (nfsd->nd_mrep) { 1124 wpp = nfsrv_nwdelayhash(slp, &nfsd->nd_fh); 1125 owp = NULL; 1126 wp = LIST_FIRST(wpp); 1127 while (wp && nfsrv_comparefh(&nfsd->nd_fh, &wp->nd_fh)) { 1128 owp = wp; 1129 wp = LIST_NEXT(wp, nd_hash); 1130 } 1131 while (wp && wp->nd_off < nfsd->nd_off && 1132 !nfsrv_comparefh(&nfsd->nd_fh, &wp->nd_fh)) { 1133 owp = wp; 1134 wp = LIST_NEXT(wp, nd_hash); 1135 } 1136 if (owp) { 1137 LIST_INSERT_AFTER(owp, nfsd, nd_hash); 1138 1139 /* 1140 * Search the hash list for overlapping entries and 1141 * coalesce. 1142 */ 1143 for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) { 1144 wp = LIST_NEXT(nfsd, nd_hash); 1145 if (nfsrv_samecred(owp->nd_cr, nfsd->nd_cr)) 1146 nfsrvw_coalesce(owp, nfsd); 1147 } 1148 } else { 1149 LIST_INSERT_HEAD(wpp, nfsd, nd_hash); 1150 } 1151 } 1152 splx(s); 1153 } 1154 1155 /* 1156 * Now, do VOP_WRITE()s for any one(s) that need to be done now 1157 * and generate the associated reply mbuf list(s). 1158 */ 1159 loop1: 1160 getmicrotime(&now); 1161 cur_usec = (u_quad_t)now.tv_sec * 1000000 + (u_quad_t)now.tv_usec; 1162 s = splsoftclock(); 1163 for (nfsd = LIST_FIRST(&slp->ns_tq); nfsd; nfsd = owp) { 1164 owp = LIST_NEXT(nfsd, nd_tq); 1165 if (nfsd->nd_time > cur_usec) 1166 break; 1167 if (nfsd->nd_mreq) 1168 continue; 1169 LIST_REMOVE(nfsd, nd_tq); 1170 LIST_REMOVE(nfsd, nd_hash); 1171 splx(s); 1172 mrep = nfsd->nd_mrep; 1173 nfsd->nd_mrep = NULL; 1174 cred = nfsd->nd_cr; 1175 v3 = (nfsd->nd_flag & ND_NFSV3); 1176 forat_ret = aftat_ret = 1; 1177 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp, 1178 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH), 1179 false); 1180 if (!error) { 1181 if (v3) 1182 forat_ret = VOP_GETATTR(vp, &forat, cred, lwp); 1183 if (vp->v_type != VREG) { 1184 if (v3) 1185 error = EINVAL; 1186 else 1187 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 1188 } 1189 } else 1190 vp = NULL; 1191 if (!error) { 1192 nqsrv_getl(vp, ND_WRITE); 1193 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1); 1194 } 1195 1196 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE) 1197 ioflags = IO_NODELOCKED; 1198 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC) 1199 ioflags = (IO_SYNC | IO_NODELOCKED); 1200 else 1201 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 1202 uiop->uio_rw = UIO_WRITE; 1203 uiop->uio_offset = nfsd->nd_off; 1204 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off; 1205 UIO_SETUP_SYSSPACE(uiop); 1206 if (uiop->uio_resid > 0) { 1207 mp = mrep; 1208 i = 0; 1209 while (mp) { 1210 if (mp->m_len > 0) 1211 i++; 1212 mp = mp->m_next; 1213 } 1214 uiop->uio_iovcnt = i; 1215 iov = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK); 1216 uiop->uio_iov = ivp = iov; 1217 mp = mrep; 1218 while (mp) { 1219 if (mp->m_len > 0) { 1220 ivp->iov_base = mtod(mp, void *); 1221 ivp->iov_len = mp->m_len; 1222 ivp++; 1223 } 1224 mp = mp->m_next; 1225 } 1226 if (!error) { 1227 error = VOP_WRITE(vp, uiop, ioflags, cred); 1228 nfsstats.srvvop_writes++; 1229 } 1230 free((void *)iov, M_TEMP); 1231 } 1232 m_freem(mrep); 1233 if (vp) { 1234 aftat_ret = VOP_GETATTR(vp, &va, cred, lwp); 1235 vput(vp); 1236 } 1237 1238 /* 1239 * Loop around generating replies for all write rpcs that have 1240 * now been completed. 1241 */ 1242 swp = nfsd; 1243 do { 1244 if (error) { 1245 nfsm_writereply(NFSX_WCCDATA(v3), v3); 1246 if (v3) { 1247 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1248 } 1249 } else { 1250 nfsm_writereply(NFSX_PREOPATTR(v3) + 1251 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED + 1252 NFSX_WRITEVERF(v3), v3); 1253 if (v3) { 1254 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1255 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1256 *tl++ = txdr_unsigned(nfsd->nd_len); 1257 *tl++ = txdr_unsigned(swp->nd_stable); 1258 /* 1259 * Actually, there is no need to txdr these fields, 1260 * but it may make the values more human readable, 1261 * for debugging purposes. 1262 */ 1263 *tl++ = txdr_unsigned(boottime.tv_sec); 1264 *tl = txdr_unsigned(boottime.tv_usec); 1265 } else { 1266 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1267 nfsm_srvfillattr(&va, fp); 1268 } 1269 } 1270 nfsd->nd_mreq = mreq; 1271 if (nfsd->nd_mrep) 1272 panic("nfsrv_write: nd_mrep not free"); 1273 1274 /* 1275 * Done. Put it at the head of the timer queue so that 1276 * the final phase can return the reply. 1277 */ 1278 s = splsoftclock(); 1279 if (nfsd != swp) { 1280 nfsd->nd_time = 0; 1281 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1282 } 1283 nfsd = LIST_FIRST(&swp->nd_coalesce); 1284 if (nfsd) { 1285 LIST_REMOVE(nfsd, nd_tq); 1286 } 1287 splx(s); 1288 } while (nfsd); 1289 s = splsoftclock(); 1290 swp->nd_time = 0; 1291 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq); 1292 splx(s); 1293 goto loop1; 1294 } 1295 nfs_timer_start(); 1296 splx(s); 1297 1298 /* 1299 * Search for a reply to return. 1300 */ 1301 s = splsoftclock(); 1302 LIST_FOREACH(nfsd, &slp->ns_tq, nd_tq) { 1303 if (nfsd->nd_mreq) { 1304 LIST_REMOVE(nfsd, nd_tq); 1305 *mrq = nfsd->nd_mreq; 1306 *ndp = nfsd; 1307 break; 1308 } 1309 } 1310 splx(s); 1311 return (0); 1312 } 1313 1314 /* 1315 * Coalesce the write request nfsd into owp. To do this we must: 1316 * - remove nfsd from the queues 1317 * - merge nfsd->nd_mrep into owp->nd_mrep 1318 * - update the nd_eoff and nd_stable for owp 1319 * - put nfsd on owp's nd_coalesce list 1320 * NB: Must be called at splsoftclock(). 1321 */ 1322 void 1323 nfsrvw_coalesce(owp, nfsd) 1324 struct nfsrv_descript *owp; 1325 struct nfsrv_descript *nfsd; 1326 { 1327 int overlap; 1328 struct mbuf *mp; 1329 struct nfsrv_descript *m; 1330 1331 LIST_REMOVE(nfsd, nd_hash); 1332 LIST_REMOVE(nfsd, nd_tq); 1333 if (owp->nd_eoff < nfsd->nd_eoff) { 1334 overlap = owp->nd_eoff - nfsd->nd_off; 1335 if (overlap < 0) 1336 panic("nfsrv_coalesce: bad off"); 1337 if (overlap > 0) 1338 m_adj(nfsd->nd_mrep, overlap); 1339 mp = owp->nd_mrep; 1340 while (mp->m_next) 1341 mp = mp->m_next; 1342 mp->m_next = nfsd->nd_mrep; 1343 owp->nd_eoff = nfsd->nd_eoff; 1344 } else 1345 m_freem(nfsd->nd_mrep); 1346 nfsd->nd_mrep = NULL; 1347 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC) 1348 owp->nd_stable = NFSV3WRITE_FILESYNC; 1349 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC && 1350 owp->nd_stable == NFSV3WRITE_UNSTABLE) 1351 owp->nd_stable = NFSV3WRITE_DATASYNC; 1352 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq); 1353 /* 1354 * nfsd might hold coalesce elements! Move them to owp. 1355 * Otherwise, requests may be lost and clients will be stuck. 1356 */ 1357 while ((m = LIST_FIRST(&nfsd->nd_coalesce)) != NULL) { 1358 LIST_REMOVE(m, nd_tq); 1359 LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq); 1360 } 1361 } 1362 1363 /* 1364 * nfs create service 1365 * now does a truncate to 0 length via. setattr if it already exists 1366 */ 1367 int 1368 nfsrv_create(nfsd, slp, lwp, mrq) 1369 struct nfsrv_descript *nfsd; 1370 struct nfssvc_sock *slp; 1371 struct lwp *lwp; 1372 struct mbuf **mrq; 1373 { 1374 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1375 struct mbuf *nam = nfsd->nd_nam; 1376 char *dpos = nfsd->nd_dpos; 1377 kauth_cred_t cred = nfsd->nd_cr; 1378 struct nfs_fattr *fp; 1379 struct vattr va, dirfor, diraft; 1380 struct nfsv2_sattr *sp; 1381 u_int32_t *tl; 1382 struct nameidata nd; 1383 char *cp; 1384 int32_t t1; 1385 char *bpos; 1386 int error = 0, cache = 0, len, tsize, dirfor_ret = 1, diraft_ret = 1; 1387 int rdev = 0; 1388 int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0; 1389 char *cp2; 1390 struct mbuf *mb, *mreq; 1391 struct vnode *vp = NULL, *dirp = NULL; 1392 nfsrvfh_t nsfh; 1393 u_quad_t frev, tempsize; 1394 u_char cverf[NFSX_V3CREATEVERF]; 1395 1396 nd.ni_cnd.cn_nameiop = 0; 1397 nfsm_srvmtofh(&nsfh); 1398 nfsm_srvnamesiz(len); 1399 nd.ni_cnd.cn_cred = cred; 1400 nd.ni_cnd.cn_nameiop = CREATE; 1401 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1402 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 1403 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1404 if (dirp && v3) { 1405 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1406 } 1407 if (error) { 1408 nfsm_reply(NFSX_WCCDATA(v3)); 1409 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1410 if (dirp) 1411 vrele(dirp); 1412 return (0); 1413 } 1414 VATTR_NULL(&va); 1415 if (v3) { 1416 va.va_mode = 0; 1417 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1418 how = fxdr_unsigned(int, *tl); 1419 switch (how) { 1420 case NFSV3CREATE_GUARDED: 1421 if (nd.ni_vp) { 1422 error = EEXIST; 1423 break; 1424 } 1425 case NFSV3CREATE_UNCHECKED: 1426 nfsm_srvsattr(&va); 1427 break; 1428 case NFSV3CREATE_EXCLUSIVE: 1429 nfsm_dissect(cp, void *, NFSX_V3CREATEVERF); 1430 memcpy(cverf, cp, NFSX_V3CREATEVERF); 1431 exclusive_flag = 1; 1432 break; 1433 }; 1434 va.va_type = VREG; 1435 } else { 1436 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 1437 va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode)); 1438 if (va.va_type == VNON) 1439 va.va_type = VREG; 1440 va.va_mode = nfstov_mode(sp->sa_mode); 1441 switch (va.va_type) { 1442 case VREG: 1443 tsize = fxdr_unsigned(int32_t, sp->sa_size); 1444 if (tsize != -1) 1445 va.va_size = (u_quad_t)tsize; 1446 break; 1447 case VCHR: 1448 case VBLK: 1449 case VFIFO: 1450 rdev = fxdr_unsigned(int32_t, sp->sa_size); 1451 break; 1452 default: 1453 break; 1454 }; 1455 } 1456 1457 /* 1458 * Iff doesn't exist, create it 1459 * otherwise just truncate to 0 length 1460 * should I set the mode too ?? 1461 */ 1462 if (nd.ni_vp == NULL) { 1463 if (va.va_type == VREG || va.va_type == VSOCK) { 1464 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1465 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1466 if (!error) { 1467 if (exclusive_flag) { 1468 exclusive_flag = 0; 1469 VATTR_NULL(&va); 1470 /* 1471 * XXX 1472 * assuming NFSX_V3CREATEVERF 1473 * == sizeof(nfstime3) 1474 */ 1475 fxdr_nfsv3time(cverf, &va.va_atime); 1476 error = VOP_SETATTR(nd.ni_vp, &va, cred, 1477 lwp); 1478 } 1479 } 1480 } else if (va.va_type == VCHR || va.va_type == VBLK || 1481 va.va_type == VFIFO) { 1482 if (va.va_type == VCHR && rdev == 0xffffffff) 1483 va.va_type = VFIFO; 1484 if (va.va_type != VFIFO && 1485 (error = kauth_authorize_generic(cred, 1486 KAUTH_GENERIC_ISSUSER, NULL))) { 1487 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1488 vput(nd.ni_dvp); 1489 nfsm_reply(0); 1490 return (error); 1491 } else 1492 va.va_rdev = (dev_t)rdev; 1493 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1494 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, 1495 &va); 1496 if (error) { 1497 nfsm_reply(0); 1498 } 1499 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1500 vput(nd.ni_vp); 1501 vrele(nd.ni_dvp); 1502 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1503 error = EINVAL; 1504 nfsm_reply(0); 1505 } 1506 } else { 1507 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1508 vput(nd.ni_dvp); 1509 error = ENXIO; 1510 } 1511 vp = nd.ni_vp; 1512 } else { 1513 vp = nd.ni_vp; 1514 if (nd.ni_dvp == vp) 1515 vrele(nd.ni_dvp); 1516 else 1517 vput(nd.ni_dvp); 1518 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1519 if (!error && va.va_size != -1) { 1520 error = nfsrv_access(vp, VWRITE, cred, 1521 (nd.ni_cnd.cn_flags & RDONLY), lwp, 0); 1522 if (!error) { 1523 nqsrv_getl(vp, ND_WRITE); 1524 tempsize = va.va_size; 1525 VATTR_NULL(&va); 1526 va.va_size = tempsize; 1527 error = VOP_SETATTR(vp, &va, cred, lwp); 1528 } 1529 } 1530 if (error) 1531 vput(vp); 1532 } 1533 if (!error) { 1534 error = nfsrv_composefh(vp, &nsfh, v3); 1535 if (!error) 1536 error = VOP_GETATTR(vp, &va, cred, lwp); 1537 vput(vp); 1538 } 1539 if (v3) { 1540 if (exclusive_flag && !error) { 1541 /* 1542 * XXX assuming NFSX_V3CREATEVERF == sizeof(nfstime3) 1543 */ 1544 char oldverf[NFSX_V3CREATEVERF]; 1545 1546 txdr_nfsv3time(&va.va_atime, oldverf); 1547 if (memcmp(cverf, oldverf, NFSX_V3CREATEVERF)) 1548 error = EEXIST; 1549 } 1550 if (dirp) { 1551 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1552 } 1553 } 1554 if (dirp) { 1555 vrele(dirp); 1556 } 1557 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3)); 1558 if (v3) { 1559 if (!error) { 1560 nfsm_srvpostop_fh(&nsfh); 1561 nfsm_srvpostop_attr(0, &va); 1562 } 1563 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1564 } else { 1565 nfsm_srvfhtom(&nsfh, v3); 1566 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1567 nfsm_srvfillattr(&va, fp); 1568 } 1569 return (0); 1570 nfsmout: 1571 if (dirp) 1572 vrele(dirp); 1573 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1574 if (nd.ni_dvp == nd.ni_vp) 1575 vrele(nd.ni_dvp); 1576 else 1577 vput(nd.ni_dvp); 1578 if (nd.ni_vp) 1579 vput(nd.ni_vp); 1580 return (error); 1581 } 1582 1583 /* 1584 * nfs v3 mknod service 1585 */ 1586 int 1587 nfsrv_mknod(nfsd, slp, lwp, mrq) 1588 struct nfsrv_descript *nfsd; 1589 struct nfssvc_sock *slp; 1590 struct lwp *lwp; 1591 struct mbuf **mrq; 1592 { 1593 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1594 struct mbuf *nam = nfsd->nd_nam; 1595 char *dpos = nfsd->nd_dpos; 1596 kauth_cred_t cred = nfsd->nd_cr; 1597 struct vattr va, dirfor, diraft; 1598 u_int32_t *tl; 1599 struct nameidata nd; 1600 int32_t t1; 1601 char *bpos; 1602 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 1603 u_int32_t major, minor; 1604 enum vtype vtyp; 1605 char *cp2; 1606 struct mbuf *mb, *mreq; 1607 struct vnode *vp, *dirp = (struct vnode *)0; 1608 nfsrvfh_t nsfh; 1609 u_quad_t frev; 1610 1611 nd.ni_cnd.cn_nameiop = 0; 1612 nfsm_srvmtofh(&nsfh); 1613 nfsm_srvnamesiz(len); 1614 nd.ni_cnd.cn_cred = cred; 1615 nd.ni_cnd.cn_nameiop = CREATE; 1616 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1617 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 1618 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1619 if (dirp) 1620 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1621 if (error) { 1622 nfsm_reply(NFSX_WCCDATA(1)); 1623 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1624 if (dirp) 1625 vrele(dirp); 1626 return (0); 1627 } 1628 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1629 vtyp = nfsv3tov_type(*tl); 1630 if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) { 1631 error = NFSERR_BADTYPE; 1632 goto abort; 1633 } 1634 VATTR_NULL(&va); 1635 va.va_mode = 0; 1636 nfsm_srvsattr(&va); 1637 if (vtyp == VCHR || vtyp == VBLK) { 1638 dev_t rdev; 1639 1640 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1641 major = fxdr_unsigned(u_int32_t, *tl++); 1642 minor = fxdr_unsigned(u_int32_t, *tl); 1643 rdev = makedev(major, minor); 1644 if (major(rdev) != major || minor(rdev) != minor) { 1645 error = EINVAL; 1646 goto abort; 1647 } 1648 va.va_rdev = rdev; 1649 } 1650 1651 /* 1652 * Iff doesn't exist, create it. 1653 */ 1654 if (nd.ni_vp) { 1655 error = EEXIST; 1656 abort: 1657 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1658 if (nd.ni_dvp == nd.ni_vp) 1659 vrele(nd.ni_dvp); 1660 else 1661 vput(nd.ni_dvp); 1662 if (nd.ni_vp) 1663 vput(nd.ni_vp); 1664 goto out; 1665 } 1666 va.va_type = vtyp; 1667 if (vtyp == VSOCK) { 1668 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1669 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1670 } else { 1671 if (va.va_type != VFIFO && 1672 (error = kauth_authorize_generic(cred, 1673 KAUTH_GENERIC_ISSUSER, NULL))) { 1674 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1675 vput(nd.ni_dvp); 1676 goto out; 1677 } 1678 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1679 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1680 if (error) 1681 goto out; 1682 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1683 vput(nd.ni_vp); 1684 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1685 error = EINVAL; 1686 } 1687 } 1688 out: 1689 vp = nd.ni_vp; 1690 if (!error) { 1691 error = nfsrv_composefh(vp, &nsfh, true); 1692 if (!error) 1693 error = VOP_GETATTR(vp, &va, cred, lwp); 1694 vput(vp); 1695 } 1696 if (dirp) { 1697 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1698 vrele(dirp); 1699 } 1700 nfsm_reply(NFSX_SRVFH(&nsfh, true) + NFSX_POSTOPATTR(1) + 1701 NFSX_WCCDATA(1)); 1702 if (!error) { 1703 nfsm_srvpostop_fh(&nsfh); 1704 nfsm_srvpostop_attr(0, &va); 1705 } 1706 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1707 return (0); 1708 nfsmout: 1709 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1710 if (nd.ni_dvp == nd.ni_vp) 1711 vrele(nd.ni_dvp); 1712 else 1713 vput(nd.ni_dvp); 1714 if (nd.ni_vp) 1715 vput(nd.ni_vp); 1716 if (dirp) 1717 vrele(dirp); 1718 return (error); 1719 } 1720 1721 /* 1722 * nfs remove service 1723 */ 1724 int 1725 nfsrv_remove(nfsd, slp, lwp, mrq) 1726 struct nfsrv_descript *nfsd; 1727 struct nfssvc_sock *slp; 1728 struct lwp *lwp; 1729 struct mbuf **mrq; 1730 { 1731 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1732 struct mbuf *nam = nfsd->nd_nam; 1733 char *dpos = nfsd->nd_dpos; 1734 kauth_cred_t cred = nfsd->nd_cr; 1735 struct nameidata nd; 1736 u_int32_t *tl; 1737 int32_t t1; 1738 char *bpos; 1739 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 1740 int v3 = (nfsd->nd_flag & ND_NFSV3); 1741 char *cp2; 1742 struct mbuf *mb, *mreq; 1743 struct vnode *vp, *dirp; 1744 struct vattr dirfor, diraft; 1745 nfsrvfh_t nsfh; 1746 u_quad_t frev; 1747 1748 #ifndef nolint 1749 vp = (struct vnode *)0; 1750 #endif 1751 nfsm_srvmtofh(&nsfh); 1752 nfsm_srvnamesiz(len); 1753 nd.ni_cnd.cn_cred = cred; 1754 nd.ni_cnd.cn_nameiop = DELETE; 1755 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1756 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 1757 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1758 if (dirp && v3) { 1759 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 1760 } 1761 if (!error) { 1762 vp = nd.ni_vp; 1763 if (vp->v_type == VDIR && 1764 (error = kauth_authorize_generic(cred, 1765 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 1766 goto out; 1767 /* 1768 * The root of a mounted filesystem cannot be deleted. 1769 */ 1770 if (vp->v_vflag & VV_ROOT) { 1771 error = EBUSY; 1772 goto out; 1773 } 1774 out: 1775 if (!error) { 1776 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1777 nqsrv_getl(vp, ND_WRITE); 1778 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1779 } else { 1780 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1781 if (nd.ni_dvp == vp) 1782 vrele(nd.ni_dvp); 1783 else 1784 vput(nd.ni_dvp); 1785 vput(vp); 1786 } 1787 } 1788 if (dirp) { 1789 if (v3) { 1790 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 1791 } 1792 vrele(dirp); 1793 } 1794 nfsm_reply(NFSX_WCCDATA(v3)); 1795 if (v3) { 1796 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1797 return (0); 1798 } 1799 nfsm_srvdone; 1800 } 1801 1802 /* 1803 * nfs rename service 1804 */ 1805 int 1806 nfsrv_rename(nfsd, slp, lwp, mrq) 1807 struct nfsrv_descript *nfsd; 1808 struct nfssvc_sock *slp; 1809 struct lwp *lwp; 1810 struct mbuf **mrq; 1811 { 1812 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1813 struct mbuf *nam = nfsd->nd_nam; 1814 char *dpos = nfsd->nd_dpos; 1815 kauth_cred_t cred = nfsd->nd_cr; 1816 u_int32_t *tl; 1817 int32_t t1; 1818 char *bpos; 1819 int error = 0, cache = 0, fdirfor_ret = 1, fdiraft_ret = 1; 1820 uint32_t len, len2; 1821 int tdirfor_ret = 1, tdiraft_ret = 1; 1822 int v3 = (nfsd->nd_flag & ND_NFSV3); 1823 char *cp2; 1824 struct mbuf *mb, *mreq; 1825 struct nameidata fromnd, tond; 1826 struct vnode *fvp, *tvp, *tdvp; 1827 struct vnode *fdirp = NULL, *tdirp = NULL; 1828 struct vattr fdirfor, fdiraft, tdirfor, tdiraft; 1829 nfsrvfh_t fnsfh, tnsfh; 1830 u_quad_t frev; 1831 uid_t saved_uid; 1832 1833 #ifndef nolint 1834 fvp = (struct vnode *)0; 1835 #endif 1836 fromnd.ni_cnd.cn_nameiop = 0; 1837 tond.ni_cnd.cn_nameiop = 0; 1838 nfsm_srvmtofh(&fnsfh); 1839 nfsm_srvnamesiz(len); 1840 /* 1841 * Remember our original uid so that we can reset cr_uid before 1842 * the second nfs_namei() call, in case it is remapped. 1843 */ 1844 saved_uid = kauth_cred_geteuid(cred); 1845 fromnd.ni_cnd.cn_cred = cred; 1846 fromnd.ni_cnd.cn_nameiop = DELETE; 1847 fromnd.ni_cnd.cn_flags = LOCKPARENT | SAVESTART; 1848 error = nfs_namei(&fromnd, &fnsfh, len, slp, nam, &md, 1849 &dpos, &fdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1850 if (fdirp && v3) { 1851 fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred, lwp); 1852 } 1853 if (error) { 1854 nfsm_reply(2 * NFSX_WCCDATA(v3)); 1855 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 1856 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 1857 if (fdirp) 1858 vrele(fdirp); 1859 return (0); 1860 } 1861 if (fromnd.ni_dvp != fromnd.ni_vp) { 1862 VOP_UNLOCK(fromnd.ni_dvp, 0); 1863 } 1864 fvp = fromnd.ni_vp; 1865 nfsm_srvmtofh(&tnsfh); 1866 if (v3) { 1867 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 1868 len2 = fxdr_unsigned(uint32_t, *tl); 1869 /* len2 will be checked by nfs_namei */ 1870 } 1871 else { 1872 /* NFSv2 */ 1873 nfsm_strsiz(len2, NFS_MAXNAMLEN); 1874 } 1875 kauth_cred_seteuid(cred, saved_uid); 1876 tond.ni_cnd.cn_cred = cred; 1877 tond.ni_cnd.cn_nameiop = RENAME; 1878 tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART; 1879 error = nfs_namei(&tond, &tnsfh, len2, slp, nam, &md, 1880 &dpos, &tdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1881 if (tdirp && v3) { 1882 tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred, lwp); 1883 } 1884 if (error) { 1885 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1886 vrele(fromnd.ni_dvp); 1887 vrele(fvp); 1888 goto out1; 1889 } 1890 tdvp = tond.ni_dvp; 1891 tvp = tond.ni_vp; 1892 if (tvp != NULL) { 1893 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 1894 if (v3) 1895 error = EEXIST; 1896 else 1897 error = EISDIR; 1898 goto out; 1899 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 1900 if (v3) 1901 error = EEXIST; 1902 else 1903 error = ENOTDIR; 1904 goto out; 1905 } 1906 if (tvp->v_type == VDIR && tvp->v_mountedhere) { 1907 if (v3) 1908 error = EXDEV; 1909 else 1910 error = ENOTEMPTY; 1911 goto out; 1912 } 1913 } 1914 if (fvp->v_type == VDIR && fvp->v_mountedhere) { 1915 if (v3) 1916 error = EXDEV; 1917 else 1918 error = ENOTEMPTY; 1919 goto out; 1920 } 1921 if (fvp->v_mount != tdvp->v_mount) { 1922 if (v3) 1923 error = EXDEV; 1924 else 1925 error = ENOTEMPTY; 1926 goto out; 1927 } 1928 if (fvp == tdvp) { 1929 if (v3) 1930 error = EINVAL; 1931 else 1932 error = ENOTEMPTY; 1933 } 1934 /* 1935 * If source is the same as the destination (that is the 1936 * same vnode with the same name in the same directory), 1937 * then there is nothing to do. 1938 */ 1939 if (fvp == tvp && fromnd.ni_dvp == tdvp && 1940 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 1941 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr, 1942 fromnd.ni_cnd.cn_namelen)) 1943 error = -1; 1944 out: 1945 if (!error) { 1946 nqsrv_getl(fromnd.ni_dvp, ND_WRITE); 1947 nqsrv_getl(tdvp, ND_WRITE); 1948 if (tvp) { 1949 nqsrv_getl(tvp, ND_WRITE); 1950 } 1951 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 1952 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 1953 } else { 1954 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 1955 if (tdvp == tvp) 1956 vrele(tdvp); 1957 else 1958 vput(tdvp); 1959 if (tvp) 1960 vput(tvp); 1961 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1962 vrele(fromnd.ni_dvp); 1963 vrele(fvp); 1964 if (error == -1) 1965 error = 0; 1966 } 1967 vrele(tond.ni_startdir); 1968 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 1969 out1: 1970 if (fdirp) { 1971 if (v3) { 1972 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred, lwp); 1973 } 1974 vrele(fdirp); 1975 } 1976 if (tdirp) { 1977 if (v3) { 1978 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred, lwp); 1979 } 1980 vrele(tdirp); 1981 } 1982 vrele(fromnd.ni_startdir); 1983 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 1984 nfsm_reply(2 * NFSX_WCCDATA(v3)); 1985 if (v3) { 1986 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 1987 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 1988 } 1989 return (0); 1990 1991 nfsmout: 1992 if (fdirp) 1993 vrele(fdirp); 1994 #ifdef notdef 1995 if (tdirp) 1996 vrele(tdirp); 1997 #endif 1998 if (tond.ni_cnd.cn_nameiop) { 1999 vrele(tond.ni_startdir); 2000 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 2001 } 2002 if (fromnd.ni_cnd.cn_nameiop) { 2003 vrele(fromnd.ni_startdir); 2004 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2005 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2006 vrele(fromnd.ni_dvp); 2007 vrele(fvp); 2008 } 2009 return (error); 2010 } 2011 2012 /* 2013 * nfs link service 2014 */ 2015 int 2016 nfsrv_link(nfsd, slp, lwp, mrq) 2017 struct nfsrv_descript *nfsd; 2018 struct nfssvc_sock *slp; 2019 struct lwp *lwp; 2020 struct mbuf **mrq; 2021 { 2022 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2023 struct mbuf *nam = nfsd->nd_nam; 2024 char *dpos = nfsd->nd_dpos; 2025 kauth_cred_t cred = nfsd->nd_cr; 2026 struct nameidata nd; 2027 u_int32_t *tl; 2028 int32_t t1; 2029 char *bpos; 2030 int error = 0, rdonly, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 2031 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3); 2032 char *cp2; 2033 struct mbuf *mb, *mreq; 2034 struct vnode *vp, *xp, *dirp = (struct vnode *)0; 2035 struct vattr dirfor, diraft, at; 2036 nfsrvfh_t nsfh, dnsfh; 2037 u_quad_t frev; 2038 2039 nfsm_srvmtofh(&nsfh); 2040 nfsm_srvmtofh(&dnsfh); 2041 nfsm_srvnamesiz(len); 2042 error = nfsrv_fhtovp(&nsfh, false, &vp, cred, slp, nam, 2043 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 2044 if (error) { 2045 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2046 nfsm_srvpostop_attr(getret, &at); 2047 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2048 return (0); 2049 } 2050 if (vp->v_type == VDIR && (error = kauth_authorize_generic(cred, 2051 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 2052 goto out1; 2053 nd.ni_cnd.cn_cred = cred; 2054 nd.ni_cnd.cn_nameiop = CREATE; 2055 nd.ni_cnd.cn_flags = LOCKPARENT; 2056 error = nfs_namei(&nd, &dnsfh, len, slp, nam, &md, &dpos, 2057 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2058 if (dirp && v3) { 2059 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2060 } 2061 if (error) 2062 goto out1; 2063 xp = nd.ni_vp; 2064 if (xp != NULL) { 2065 error = EEXIST; 2066 goto out; 2067 } 2068 xp = nd.ni_dvp; 2069 if (vp->v_mount != xp->v_mount) 2070 error = EXDEV; 2071 out: 2072 if (!error) { 2073 nqsrv_getl(vp, ND_WRITE); 2074 nqsrv_getl(xp, ND_WRITE); 2075 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2076 } else { 2077 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2078 if (nd.ni_dvp == nd.ni_vp) 2079 vrele(nd.ni_dvp); 2080 else 2081 vput(nd.ni_dvp); 2082 if (nd.ni_vp) 2083 vrele(nd.ni_vp); 2084 } 2085 out1: 2086 if (v3) 2087 getret = VOP_GETATTR(vp, &at, cred, lwp); 2088 if (dirp) { 2089 if (v3) { 2090 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2091 } 2092 vrele(dirp); 2093 } 2094 vrele(vp); 2095 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2096 if (v3) { 2097 nfsm_srvpostop_attr(getret, &at); 2098 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2099 return (0); 2100 } 2101 nfsm_srvdone; 2102 } 2103 2104 /* 2105 * nfs symbolic link service 2106 */ 2107 int 2108 nfsrv_symlink(nfsd, slp, lwp, mrq) 2109 struct nfsrv_descript *nfsd; 2110 struct nfssvc_sock *slp; 2111 struct lwp *lwp; 2112 struct mbuf **mrq; 2113 { 2114 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2115 struct mbuf *nam = nfsd->nd_nam; 2116 char *dpos = nfsd->nd_dpos; 2117 kauth_cred_t cred = nfsd->nd_cr; 2118 struct vattr va, dirfor, diraft; 2119 struct nameidata nd; 2120 u_int32_t *tl; 2121 int32_t t1; 2122 struct nfsv2_sattr *sp; 2123 char *bpos, *pathcp = NULL, *cp2; 2124 struct uio io; 2125 struct iovec iv; 2126 int error = 0, cache = 0, dirfor_ret = 1, diraft_ret = 1; 2127 uint32_t len, len2; 2128 int v3 = (nfsd->nd_flag & ND_NFSV3); 2129 struct mbuf *mb, *mreq; 2130 struct vnode *dirp = (struct vnode *)0; 2131 nfsrvfh_t nsfh; 2132 u_quad_t frev; 2133 2134 nd.ni_cnd.cn_nameiop = 0; 2135 nfsm_srvmtofh(&nsfh); 2136 nfsm_srvnamesiz(len); 2137 nd.ni_cnd.cn_cred = cred; 2138 nd.ni_cnd.cn_nameiop = CREATE; 2139 nd.ni_cnd.cn_flags = LOCKPARENT; 2140 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 2141 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2142 if (dirp && v3) { 2143 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2144 } 2145 if (error) 2146 goto out; 2147 VATTR_NULL(&va); 2148 va.va_type = VLNK; 2149 if (v3) { 2150 va.va_mode = 0; 2151 nfsm_srvsattr(&va); 2152 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 2153 len2 = fxdr_unsigned(uint32_t, *tl); 2154 if (len2 > PATH_MAX) { 2155 /* XXX should check _PC_NO_TRUNC */ 2156 error = ENAMETOOLONG; 2157 goto abortop; 2158 } 2159 } 2160 else { 2161 /* NFSv2 */ 2162 nfsm_strsiz(len2, NFS_MAXPATHLEN); 2163 } 2164 pathcp = malloc(len2 + 1, M_TEMP, M_WAITOK); 2165 iv.iov_base = pathcp; 2166 iv.iov_len = len2; 2167 io.uio_resid = len2; 2168 io.uio_offset = 0; 2169 io.uio_iov = &iv; 2170 io.uio_iovcnt = 1; 2171 io.uio_rw = UIO_READ; 2172 UIO_SETUP_SYSSPACE(&io); 2173 nfsm_mtouio(&io, len2); 2174 if (!v3) { 2175 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 2176 va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode); 2177 } 2178 *(pathcp + len2) = '\0'; 2179 if (nd.ni_vp) { 2180 error = EEXIST; 2181 abortop: 2182 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2183 if (nd.ni_dvp == nd.ni_vp) 2184 vrele(nd.ni_dvp); 2185 else 2186 vput(nd.ni_dvp); 2187 if (nd.ni_vp) 2188 vrele(nd.ni_vp); 2189 goto out; 2190 } 2191 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2192 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp); 2193 if (!error) { 2194 if (v3) { 2195 error = nfsrv_composefh(nd.ni_vp, &nsfh, v3); 2196 if (!error) 2197 error = VOP_GETATTR(nd.ni_vp, &va, cred, lwp); 2198 vput(nd.ni_vp); 2199 } else { 2200 vput(nd.ni_vp); 2201 } 2202 } 2203 out: 2204 if (pathcp) 2205 free(pathcp, M_TEMP); 2206 if (dirp) { 2207 if (v3) { 2208 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2209 } 2210 vrele(dirp); 2211 } 2212 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_POSTOPATTR(v3) + 2213 NFSX_WCCDATA(v3)); 2214 if (v3) { 2215 if (!error) { 2216 nfsm_srvpostop_fh(&nsfh); 2217 nfsm_srvpostop_attr(0, &va); 2218 } 2219 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2220 } 2221 return (0); 2222 nfsmout: 2223 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2224 if (nd.ni_dvp == nd.ni_vp) 2225 vrele(nd.ni_dvp); 2226 else 2227 vput(nd.ni_dvp); 2228 if (nd.ni_vp) 2229 vrele(nd.ni_vp); 2230 if (dirp) 2231 vrele(dirp); 2232 if (pathcp) 2233 free(pathcp, M_TEMP); 2234 return (error); 2235 } 2236 2237 /* 2238 * nfs mkdir service 2239 */ 2240 int 2241 nfsrv_mkdir(nfsd, slp, lwp, mrq) 2242 struct nfsrv_descript *nfsd; 2243 struct nfssvc_sock *slp; 2244 struct lwp *lwp; 2245 struct mbuf **mrq; 2246 { 2247 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2248 struct mbuf *nam = nfsd->nd_nam; 2249 char *dpos = nfsd->nd_dpos; 2250 kauth_cred_t cred = nfsd->nd_cr; 2251 struct vattr va, dirfor, diraft; 2252 struct nfs_fattr *fp; 2253 struct nameidata nd; 2254 char *cp; 2255 u_int32_t *tl; 2256 int32_t t1; 2257 char *bpos; 2258 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 2259 int v3 = (nfsd->nd_flag & ND_NFSV3); 2260 char *cp2; 2261 struct mbuf *mb, *mreq; 2262 struct vnode *vp, *dirp = (struct vnode *)0; 2263 nfsrvfh_t nsfh; 2264 u_quad_t frev; 2265 2266 nfsm_srvmtofh(&nsfh); 2267 nfsm_srvnamesiz(len); 2268 nd.ni_cnd.cn_cred = cred; 2269 nd.ni_cnd.cn_nameiop = CREATE; 2270 nd.ni_cnd.cn_flags = LOCKPARENT; 2271 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 2272 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2273 if (dirp && v3) { 2274 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2275 } 2276 if (error) { 2277 nfsm_reply(NFSX_WCCDATA(v3)); 2278 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2279 if (dirp) 2280 vrele(dirp); 2281 return (0); 2282 } 2283 VATTR_NULL(&va); 2284 if (v3) { 2285 va.va_mode = 0; 2286 nfsm_srvsattr(&va); 2287 } else { 2288 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 2289 va.va_mode = nfstov_mode(*tl++); 2290 } 2291 va.va_type = VDIR; 2292 vp = nd.ni_vp; 2293 if (vp != NULL) { 2294 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2295 if (nd.ni_dvp == vp) 2296 vrele(nd.ni_dvp); 2297 else 2298 vput(nd.ni_dvp); 2299 vrele(vp); 2300 error = EEXIST; 2301 goto out; 2302 } 2303 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2304 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 2305 if (!error) { 2306 vp = nd.ni_vp; 2307 error = nfsrv_composefh(vp, &nsfh, v3); 2308 if (!error) 2309 error = VOP_GETATTR(vp, &va, cred, lwp); 2310 vput(vp); 2311 } 2312 out: 2313 if (dirp) { 2314 if (v3) { 2315 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2316 } 2317 vrele(dirp); 2318 } 2319 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_POSTOPATTR(v3) + 2320 NFSX_WCCDATA(v3)); 2321 if (v3) { 2322 if (!error) { 2323 nfsm_srvpostop_fh(&nsfh); 2324 nfsm_srvpostop_attr(0, &va); 2325 } 2326 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2327 } else { 2328 nfsm_srvfhtom(&nsfh, v3); 2329 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 2330 nfsm_srvfillattr(&va, fp); 2331 } 2332 return (0); 2333 nfsmout: 2334 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2335 if (nd.ni_dvp == nd.ni_vp) 2336 vrele(nd.ni_dvp); 2337 else 2338 vput(nd.ni_dvp); 2339 if (nd.ni_vp) 2340 vrele(nd.ni_vp); 2341 if (dirp) 2342 vrele(dirp); 2343 return (error); 2344 } 2345 2346 /* 2347 * nfs rmdir service 2348 */ 2349 int 2350 nfsrv_rmdir(nfsd, slp, lwp, mrq) 2351 struct nfsrv_descript *nfsd; 2352 struct nfssvc_sock *slp; 2353 struct lwp *lwp; 2354 struct mbuf **mrq; 2355 { 2356 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2357 struct mbuf *nam = nfsd->nd_nam; 2358 char *dpos = nfsd->nd_dpos; 2359 kauth_cred_t cred = nfsd->nd_cr; 2360 u_int32_t *tl; 2361 int32_t t1; 2362 char *bpos; 2363 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 2364 int v3 = (nfsd->nd_flag & ND_NFSV3); 2365 char *cp2; 2366 struct mbuf *mb, *mreq; 2367 struct vnode *vp, *dirp = (struct vnode *)0; 2368 struct vattr dirfor, diraft; 2369 nfsrvfh_t nsfh; 2370 struct nameidata nd; 2371 u_quad_t frev; 2372 2373 nfsm_srvmtofh(&nsfh); 2374 nfsm_srvnamesiz(len); 2375 nd.ni_cnd.cn_cred = cred; 2376 nd.ni_cnd.cn_nameiop = DELETE; 2377 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 2378 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 2379 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2380 if (dirp && v3) { 2381 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp); 2382 } 2383 if (error) { 2384 nfsm_reply(NFSX_WCCDATA(v3)); 2385 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2386 if (dirp) 2387 vrele(dirp); 2388 return (0); 2389 } 2390 vp = nd.ni_vp; 2391 if (vp->v_type != VDIR) { 2392 error = ENOTDIR; 2393 goto out; 2394 } 2395 /* 2396 * No rmdir "." please. 2397 */ 2398 if (nd.ni_dvp == vp) { 2399 error = EINVAL; 2400 goto out; 2401 } 2402 /* 2403 * The root of a mounted filesystem cannot be deleted. 2404 */ 2405 if (vp->v_vflag & VV_ROOT) 2406 error = EBUSY; 2407 out: 2408 if (!error) { 2409 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2410 nqsrv_getl(vp, ND_WRITE); 2411 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2412 } else { 2413 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2414 if (nd.ni_dvp == nd.ni_vp) 2415 vrele(nd.ni_dvp); 2416 else 2417 vput(nd.ni_dvp); 2418 vput(vp); 2419 } 2420 if (dirp) { 2421 if (v3) { 2422 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp); 2423 } 2424 vrele(dirp); 2425 } 2426 nfsm_reply(NFSX_WCCDATA(v3)); 2427 if (v3) { 2428 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2429 return (0); 2430 } 2431 nfsm_srvdone; 2432 } 2433 2434 /* 2435 * nfs readdir service 2436 * - mallocs what it thinks is enough to read 2437 * count rounded up to a multiple of NFS_SRVDIRBLKSIZ <= NFS_MAXREADDIR 2438 * - calls VOP_READDIR() 2439 * - loops around building the reply 2440 * if the output generated exceeds count break out of loop 2441 * The nfsm_clget macro is used here so that the reply will be packed 2442 * tightly in mbuf clusters. 2443 * - it only knows that it has encountered eof when the VOP_READDIR() 2444 * reads nothing 2445 * - as such one readdir rpc will return eof false although you are there 2446 * and then the next will return eof 2447 * - it trims out records with d_fileno == 0 2448 * this doesn't matter for Unix clients, but they might confuse clients 2449 * for other os'. 2450 * - it trims out records with d_type == DT_WHT 2451 * these cannot be seen through NFS (unless we extend the protocol) 2452 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less 2453 * than requested, but this may not apply to all filesystems. For 2454 * example, client NFS does not { although it is never remote mounted 2455 * anyhow } 2456 * The alternate call nfsrv_readdirplus() does lookups as well. 2457 * PS: The NFS protocol spec. does not clarify what the "count" byte 2458 * argument is a count of.. just name strings and file id's or the 2459 * entire reply rpc or ... 2460 * I tried just file name and id sizes and it confused the Sun client, 2461 * so I am using the full rpc size now. The "paranoia.." comment refers 2462 * to including the status longwords that are not a part of the dir. 2463 * "entry" structures, but are in the rpc. 2464 */ 2465 2466 #define NFS_SRVDIRBLKSIZ 1024 2467 2468 struct flrep { 2469 nfsuint64 fl_off; 2470 u_int32_t fl_postopok; 2471 u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)]; 2472 u_int32_t fl_fhok; 2473 u_int32_t fl_fhsize; 2474 }; 2475 2476 int 2477 nfsrv_readdir(nfsd, slp, lwp, mrq) 2478 struct nfsrv_descript *nfsd; 2479 struct nfssvc_sock *slp; 2480 struct lwp *lwp; 2481 struct mbuf **mrq; 2482 { 2483 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2484 struct mbuf *nam = nfsd->nd_nam; 2485 char *dpos = nfsd->nd_dpos; 2486 kauth_cred_t cred = nfsd->nd_cr; 2487 char *bp, *be; 2488 struct mbuf *mp; 2489 struct dirent *dp; 2490 char *cp; 2491 u_int32_t *tl; 2492 int32_t t1; 2493 char *bpos; 2494 struct mbuf *mb, *mreq, *mp2; 2495 char *cpos, *cend, *cp2, *rbuf; 2496 struct vnode *vp; 2497 struct vattr at; 2498 nfsrvfh_t nsfh; 2499 struct uio io; 2500 struct iovec iv; 2501 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2502 int siz, cnt, fullsiz, eofflag, rdonly, cache = 0, ncookies; 2503 int v3 = (nfsd->nd_flag & ND_NFSV3); 2504 u_quad_t frev, off, toff, verf; 2505 off_t *cookies = NULL, *cookiep; 2506 nfsuint64 jar; 2507 2508 nfsm_srvmtofh(&nsfh); 2509 if (v3) { 2510 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 2511 toff = fxdr_hyper(tl); 2512 tl += 2; 2513 verf = fxdr_hyper(tl); 2514 tl += 2; 2515 } else { 2516 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2517 toff = fxdr_unsigned(u_quad_t, *tl++); 2518 } 2519 off = toff; 2520 cnt = fxdr_unsigned(int, *tl); 2521 siz = ((cnt + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1)); 2522 xfer = NFS_SRVMAXDATA(nfsd); 2523 if (siz > xfer) 2524 siz = xfer; 2525 fullsiz = siz; 2526 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 2527 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 2528 if (!error && vp->v_type != VDIR) { 2529 error = ENOTDIR; 2530 vput(vp); 2531 } 2532 if (error) { 2533 nfsm_reply(NFSX_UNSIGNED); 2534 nfsm_srvpostop_attr(getret, &at); 2535 return (0); 2536 } 2537 nqsrv_getl(vp, ND_READ); 2538 if (v3) { 2539 error = getret = VOP_GETATTR(vp, &at, cred, lwp); 2540 #ifdef NFS3_STRICTVERF 2541 /* 2542 * XXX This check is too strict for Solaris 2.5 clients. 2543 */ 2544 if (!error && toff && verf != at.va_filerev) 2545 error = NFSERR_BAD_COOKIE; 2546 #endif 2547 } 2548 if (!error) 2549 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0); 2550 if (error) { 2551 vput(vp); 2552 nfsm_reply(NFSX_POSTOPATTR(v3)); 2553 nfsm_srvpostop_attr(getret, &at); 2554 return (0); 2555 } 2556 VOP_UNLOCK(vp, 0); 2557 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2558 again: 2559 iv.iov_base = rbuf; 2560 iv.iov_len = fullsiz; 2561 io.uio_iov = &iv; 2562 io.uio_iovcnt = 1; 2563 io.uio_offset = (off_t)off; 2564 io.uio_resid = fullsiz; 2565 io.uio_rw = UIO_READ; 2566 UIO_SETUP_SYSSPACE(&io); 2567 eofflag = 0; 2568 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2569 2570 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2571 2572 off = (off_t)io.uio_offset; 2573 if (!cookies && !error) 2574 error = NFSERR_PERM; 2575 if (v3) { 2576 getret = VOP_GETATTR(vp, &at, cred, lwp); 2577 if (!error) 2578 error = getret; 2579 } 2580 2581 VOP_UNLOCK(vp, 0); 2582 if (error) { 2583 vrele(vp); 2584 free((void *)rbuf, M_TEMP); 2585 if (cookies) 2586 free((void *)cookies, M_TEMP); 2587 nfsm_reply(NFSX_POSTOPATTR(v3)); 2588 nfsm_srvpostop_attr(getret, &at); 2589 return (0); 2590 } 2591 if (io.uio_resid) { 2592 siz -= io.uio_resid; 2593 2594 /* 2595 * If nothing read, return eof 2596 * rpc reply 2597 */ 2598 if (siz == 0) { 2599 vrele(vp); 2600 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + 2601 2 * NFSX_UNSIGNED); 2602 if (v3) { 2603 nfsm_srvpostop_attr(getret, &at); 2604 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2605 txdr_hyper(at.va_filerev, tl); 2606 tl += 2; 2607 } else 2608 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2609 *tl++ = nfs_false; 2610 *tl = nfs_true; 2611 free((void *)rbuf, M_TEMP); 2612 free((void *)cookies, M_TEMP); 2613 return (0); 2614 } 2615 } 2616 2617 /* 2618 * Check for degenerate cases of nothing useful read. 2619 * If so go try again 2620 */ 2621 cpos = rbuf; 2622 cend = rbuf + siz; 2623 dp = (struct dirent *)cpos; 2624 cookiep = cookies; 2625 2626 while (cpos < cend && ncookies > 0 && 2627 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2628 cpos += dp->d_reclen; 2629 dp = (struct dirent *)cpos; 2630 cookiep++; 2631 ncookies--; 2632 } 2633 if (cpos >= cend || ncookies == 0) { 2634 toff = off; 2635 siz = fullsiz; 2636 free(cookies, M_TEMP); 2637 cookies = NULL; 2638 goto again; 2639 } 2640 2641 len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */ 2642 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz); 2643 if (v3) { 2644 nfsm_srvpostop_attr(getret, &at); 2645 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2646 txdr_hyper(at.va_filerev, tl); 2647 } 2648 mp = mp2 = mb; 2649 bp = bpos; 2650 be = bp + M_TRAILINGSPACE(mp); 2651 2652 /* Loop through the records and build reply */ 2653 while (cpos < cend && ncookies > 0) { 2654 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 2655 nlen = dp->d_namlen; 2656 rem = nfsm_rndup(nlen)-nlen; 2657 len += (4 * NFSX_UNSIGNED + nlen + rem); 2658 if (v3) 2659 len += 2 * NFSX_UNSIGNED; 2660 if (len > cnt) { 2661 eofflag = 0; 2662 break; 2663 } 2664 /* 2665 * Build the directory record xdr from 2666 * the dirent entry. 2667 */ 2668 nfsm_clget; 2669 *tl = nfs_true; 2670 bp += NFSX_UNSIGNED; 2671 if (v3) { 2672 nfsm_clget; 2673 *tl = txdr_unsigned(dp->d_fileno >> 32); 2674 bp += NFSX_UNSIGNED; 2675 } 2676 nfsm_clget; 2677 *tl = txdr_unsigned(dp->d_fileno); 2678 bp += NFSX_UNSIGNED; 2679 nfsm_clget; 2680 *tl = txdr_unsigned(nlen); 2681 bp += NFSX_UNSIGNED; 2682 2683 /* And loop around copying the name */ 2684 xfer = nlen; 2685 cp = dp->d_name; 2686 while (xfer > 0) { 2687 nfsm_clget; 2688 if ((bp+xfer) > be) 2689 tsiz = be-bp; 2690 else 2691 tsiz = xfer; 2692 memcpy(bp, cp, tsiz); 2693 bp += tsiz; 2694 xfer -= tsiz; 2695 if (xfer > 0) 2696 cp += tsiz; 2697 } 2698 /* And null pad to an int32_t boundary */ 2699 for (i = 0; i < rem; i++) 2700 *bp++ = '\0'; 2701 nfsm_clget; 2702 2703 /* Finish off the record */ 2704 txdr_hyper(*cookiep, &jar); 2705 if (v3) { 2706 *tl = jar.nfsuquad[0]; 2707 bp += NFSX_UNSIGNED; 2708 nfsm_clget; 2709 } 2710 *tl = jar.nfsuquad[1]; 2711 bp += NFSX_UNSIGNED; 2712 } 2713 cpos += dp->d_reclen; 2714 dp = (struct dirent *)cpos; 2715 cookiep++; 2716 ncookies--; 2717 } 2718 vrele(vp); 2719 nfsm_clget; 2720 *tl = nfs_false; 2721 bp += NFSX_UNSIGNED; 2722 nfsm_clget; 2723 if (eofflag) 2724 *tl = nfs_true; 2725 else 2726 *tl = nfs_false; 2727 bp += NFSX_UNSIGNED; 2728 if (mp != mb) { 2729 if (bp < be) 2730 mp->m_len = bp - mtod(mp, char *); 2731 } else 2732 mp->m_len += bp - bpos; 2733 free((void *)rbuf, M_TEMP); 2734 free((void *)cookies, M_TEMP); 2735 nfsm_srvdone; 2736 } 2737 2738 int 2739 nfsrv_readdirplus(nfsd, slp, lwp, mrq) 2740 struct nfsrv_descript *nfsd; 2741 struct nfssvc_sock *slp; 2742 struct lwp *lwp; 2743 struct mbuf **mrq; 2744 { 2745 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2746 struct mbuf *nam = nfsd->nd_nam; 2747 char *dpos = nfsd->nd_dpos; 2748 kauth_cred_t cred = nfsd->nd_cr; 2749 char *bp, *be; 2750 struct mbuf *mp; 2751 struct dirent *dp; 2752 char *cp; 2753 u_int32_t *tl; 2754 int32_t t1; 2755 char *bpos; 2756 struct mbuf *mb, *mreq, *mp2; 2757 char *cpos, *cend, *cp2, *rbuf; 2758 struct vnode *vp, *nvp; 2759 struct flrep fl; 2760 nfsrvfh_t nsfh; 2761 struct uio io; 2762 struct iovec iv; 2763 struct vattr va, at, *vap = &va; 2764 struct nfs_fattr *fp; 2765 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2766 int siz, cnt, fullsiz, eofflag, rdonly, cache = 0, dirlen, ncookies; 2767 u_quad_t frev, off, toff, verf; 2768 off_t *cookies = NULL, *cookiep; 2769 2770 nfsm_srvmtofh(&nsfh); 2771 nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED); 2772 toff = fxdr_hyper(tl); 2773 tl += 2; 2774 verf = fxdr_hyper(tl); 2775 tl += 2; 2776 siz = fxdr_unsigned(int, *tl++); 2777 cnt = fxdr_unsigned(int, *tl); 2778 off = toff; 2779 siz = ((siz + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1)); 2780 xfer = NFS_SRVMAXDATA(nfsd); 2781 if (siz > xfer) 2782 siz = xfer; 2783 fullsiz = siz; 2784 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 2785 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 2786 if (!error && vp->v_type != VDIR) { 2787 error = ENOTDIR; 2788 vput(vp); 2789 } 2790 if (error) { 2791 nfsm_reply(NFSX_UNSIGNED); 2792 nfsm_srvpostop_attr(getret, &at); 2793 return (0); 2794 } 2795 error = getret = VOP_GETATTR(vp, &at, cred, lwp); 2796 #ifdef NFS3_STRICTVERF 2797 /* 2798 * XXX This check is too strict for Solaris 2.5 clients. 2799 */ 2800 if (!error && toff && verf != at.va_filerev) 2801 error = NFSERR_BAD_COOKIE; 2802 #endif 2803 if (!error) { 2804 nqsrv_getl(vp, ND_READ); 2805 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0); 2806 } 2807 if (error) { 2808 vput(vp); 2809 nfsm_reply(NFSX_V3POSTOPATTR); 2810 nfsm_srvpostop_attr(getret, &at); 2811 return (0); 2812 } 2813 VOP_UNLOCK(vp, 0); 2814 2815 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2816 again: 2817 iv.iov_base = rbuf; 2818 iv.iov_len = fullsiz; 2819 io.uio_iov = &iv; 2820 io.uio_iovcnt = 1; 2821 io.uio_offset = (off_t)off; 2822 io.uio_resid = fullsiz; 2823 io.uio_rw = UIO_READ; 2824 UIO_SETUP_SYSSPACE(&io); 2825 eofflag = 0; 2826 2827 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2828 2829 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2830 2831 off = (u_quad_t)io.uio_offset; 2832 getret = VOP_GETATTR(vp, &at, cred, lwp); 2833 2834 VOP_UNLOCK(vp, 0); 2835 2836 /* 2837 * If the VGET operation doesn't work for this filesystem, 2838 * we can't support readdirplus. Returning NOTSUPP should 2839 * make clients fall back to plain readdir. 2840 * There's no need to check for VPTOFH as well, we wouldn't 2841 * even be here otherwise. 2842 */ 2843 if (!getret) { 2844 if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp))) 2845 getret = (getret == EOPNOTSUPP) ? 2846 NFSERR_NOTSUPP : NFSERR_IO; 2847 else 2848 vput(nvp); 2849 } 2850 2851 if (!cookies && !error) 2852 error = NFSERR_PERM; 2853 if (!error) 2854 error = getret; 2855 if (error) { 2856 vrele(vp); 2857 if (cookies) 2858 free((void *)cookies, M_TEMP); 2859 free((void *)rbuf, M_TEMP); 2860 nfsm_reply(NFSX_V3POSTOPATTR); 2861 nfsm_srvpostop_attr(getret, &at); 2862 return (0); 2863 } 2864 if (io.uio_resid) { 2865 siz -= io.uio_resid; 2866 2867 /* 2868 * If nothing read, return eof 2869 * rpc reply 2870 */ 2871 if (siz == 0) { 2872 vrele(vp); 2873 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2874 2 * NFSX_UNSIGNED); 2875 nfsm_srvpostop_attr(getret, &at); 2876 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2877 txdr_hyper(at.va_filerev, tl); 2878 tl += 2; 2879 *tl++ = nfs_false; 2880 *tl = nfs_true; 2881 free((void *)cookies, M_TEMP); 2882 free((void *)rbuf, M_TEMP); 2883 return (0); 2884 } 2885 } 2886 2887 /* 2888 * Check for degenerate cases of nothing useful read. 2889 * If so go try again 2890 */ 2891 cpos = rbuf; 2892 cend = rbuf + siz; 2893 dp = (struct dirent *)cpos; 2894 cookiep = cookies; 2895 2896 while (cpos < cend && ncookies > 0 && 2897 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2898 cpos += dp->d_reclen; 2899 dp = (struct dirent *)cpos; 2900 cookiep++; 2901 ncookies--; 2902 } 2903 if (cpos >= cend || ncookies == 0) { 2904 toff = off; 2905 siz = fullsiz; 2906 free(cookies, M_TEMP); 2907 cookies = NULL; 2908 goto again; 2909 } 2910 2911 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED; 2912 nfsm_reply(cnt); 2913 nfsm_srvpostop_attr(getret, &at); 2914 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2915 txdr_hyper(at.va_filerev, tl); 2916 mp = mp2 = mb; 2917 bp = bpos; 2918 be = bp + M_TRAILINGSPACE(mp); 2919 2920 /* Loop through the records and build reply */ 2921 while (cpos < cend && ncookies > 0) { 2922 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 2923 nfsrvfh_t nnsfh; 2924 2925 nlen = dp->d_namlen; 2926 rem = nfsm_rndup(nlen)-nlen; 2927 2928 /* 2929 * For readdir_and_lookup get the vnode using 2930 * the file number. 2931 */ 2932 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp)) 2933 goto invalid; 2934 if (nfsrv_composefh(nvp, &nnsfh, true)) { 2935 vput(nvp); 2936 goto invalid; 2937 } 2938 if (VOP_GETATTR(nvp, vap, cred, lwp)) { 2939 vput(nvp); 2940 goto invalid; 2941 } 2942 vput(nvp); 2943 2944 /* 2945 * If either the dircount or maxcount will be 2946 * exceeded, get out now. Both of these lengths 2947 * are calculated conservatively, including all 2948 * XDR overheads. 2949 */ 2950 len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH + 2951 NFSX_V3POSTOPATTR); 2952 dirlen += (6 * NFSX_UNSIGNED + nlen + rem); 2953 if (len > cnt || dirlen > fullsiz) { 2954 eofflag = 0; 2955 break; 2956 } 2957 2958 /* 2959 * Build the directory record xdr from 2960 * the dirent entry. 2961 */ 2962 fp = (struct nfs_fattr *)&fl.fl_fattr; 2963 nfsm_srvfillattr(vap, fp); 2964 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH); 2965 fl.fl_fhok = nfs_true; 2966 fl.fl_postopok = nfs_true; 2967 txdr_hyper(*cookiep, fl.fl_off.nfsuquad); 2968 2969 nfsm_clget; 2970 *tl = nfs_true; 2971 bp += NFSX_UNSIGNED; 2972 nfsm_clget; 2973 *tl = txdr_unsigned(dp->d_fileno >> 32); 2974 bp += NFSX_UNSIGNED; 2975 nfsm_clget; 2976 *tl = txdr_unsigned(dp->d_fileno); 2977 bp += NFSX_UNSIGNED; 2978 nfsm_clget; 2979 *tl = txdr_unsigned(nlen); 2980 bp += NFSX_UNSIGNED; 2981 2982 /* And loop around copying the name */ 2983 xfer = nlen; 2984 cp = dp->d_name; 2985 while (xfer > 0) { 2986 nfsm_clget; 2987 if ((bp + xfer) > be) 2988 tsiz = be - bp; 2989 else 2990 tsiz = xfer; 2991 memcpy(bp, cp, tsiz); 2992 bp += tsiz; 2993 xfer -= tsiz; 2994 if (xfer > 0) 2995 cp += tsiz; 2996 } 2997 /* And null pad to an int32_t boundary */ 2998 for (i = 0; i < rem; i++) 2999 *bp++ = '\0'; 3000 3001 /* 3002 * Now copy the flrep structure out. 3003 */ 3004 xfer = sizeof(struct flrep); 3005 cp = (void *)&fl; 3006 while (xfer > 0) { 3007 nfsm_clget; 3008 if ((bp + xfer) > be) 3009 tsiz = be - bp; 3010 else 3011 tsiz = xfer; 3012 memcpy(bp, cp, tsiz); 3013 bp += tsiz; 3014 xfer -= tsiz; 3015 if (xfer > 0) 3016 cp += tsiz; 3017 } 3018 3019 /* 3020 * ... and filehandle. 3021 */ 3022 xfer = NFSRVFH_SIZE(&nnsfh); 3023 cp = NFSRVFH_DATA(&nnsfh); 3024 while (xfer > 0) { 3025 nfsm_clget; 3026 if ((bp + xfer) > be) 3027 tsiz = be - bp; 3028 else 3029 tsiz = xfer; 3030 memcpy(bp, cp, tsiz); 3031 bp += tsiz; 3032 xfer -= tsiz; 3033 if (xfer > 0) 3034 cp += tsiz; 3035 } 3036 } 3037 invalid: 3038 cpos += dp->d_reclen; 3039 dp = (struct dirent *)cpos; 3040 cookiep++; 3041 ncookies--; 3042 } 3043 vrele(vp); 3044 nfsm_clget; 3045 *tl = nfs_false; 3046 bp += NFSX_UNSIGNED; 3047 nfsm_clget; 3048 if (eofflag) 3049 *tl = nfs_true; 3050 else 3051 *tl = nfs_false; 3052 bp += NFSX_UNSIGNED; 3053 if (mp != mb) { 3054 if (bp < be) 3055 mp->m_len = bp - mtod(mp, char *); 3056 } else 3057 mp->m_len += bp - bpos; 3058 free((void *)cookies, M_TEMP); 3059 free((void *)rbuf, M_TEMP); 3060 nfsm_srvdone; 3061 } 3062 3063 /* 3064 * nfs commit service 3065 */ 3066 int 3067 nfsrv_commit(nfsd, slp, lwp, mrq) 3068 struct nfsrv_descript *nfsd; 3069 struct nfssvc_sock *slp; 3070 struct lwp *lwp; 3071 struct mbuf **mrq; 3072 { 3073 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3074 struct mbuf *nam = nfsd->nd_nam; 3075 char *dpos = nfsd->nd_dpos; 3076 kauth_cred_t cred = nfsd->nd_cr; 3077 struct vattr bfor, aft; 3078 struct vnode *vp; 3079 nfsrvfh_t nsfh; 3080 u_int32_t *tl; 3081 int32_t t1; 3082 char *bpos; 3083 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cache = 0; 3084 uint32_t cnt; 3085 char *cp2; 3086 struct mbuf *mb, *mreq; 3087 u_quad_t frev, off, end; 3088 3089 nfsm_srvmtofh(&nsfh); 3090 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED); 3091 3092 off = fxdr_hyper(tl); 3093 tl += 2; 3094 cnt = fxdr_unsigned(uint32_t, *tl); 3095 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3096 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3097 if (error) { 3098 nfsm_reply(2 * NFSX_UNSIGNED); 3099 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3100 return (0); 3101 } 3102 for_ret = VOP_GETATTR(vp, &bfor, cred, lwp); 3103 end = (cnt > 0) ? off + cnt : vp->v_size; 3104 if (end < off || end > vp->v_size) 3105 end = vp->v_size; 3106 if (off < vp->v_size) 3107 error = VOP_FSYNC(vp, cred, FSYNC_WAIT, off, end, lwp); 3108 /* else error == 0, from nfsrv_fhtovp() */ 3109 aft_ret = VOP_GETATTR(vp, &aft, cred, lwp); 3110 vput(vp); 3111 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF); 3112 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3113 if (!error) { 3114 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF); 3115 *tl++ = txdr_unsigned(boottime.tv_sec); 3116 *tl = txdr_unsigned(boottime.tv_usec); 3117 } else { 3118 return (0); 3119 } 3120 nfsm_srvdone; 3121 } 3122 3123 /* 3124 * nfs statfs service 3125 */ 3126 int 3127 nfsrv_statfs(nfsd, slp, lwp, mrq) 3128 struct nfsrv_descript *nfsd; 3129 struct nfssvc_sock *slp; 3130 struct lwp *lwp; 3131 struct mbuf **mrq; 3132 { 3133 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3134 struct mbuf *nam = nfsd->nd_nam; 3135 char *dpos = nfsd->nd_dpos; 3136 kauth_cred_t cred = nfsd->nd_cr; 3137 struct statvfs *sf = NULL; 3138 struct nfs_statfs *sfp; 3139 u_int32_t *tl; 3140 int32_t t1; 3141 char *bpos; 3142 int error = 0, rdonly, cache = 0, getret = 1; 3143 int v3 = (nfsd->nd_flag & ND_NFSV3); 3144 char *cp2; 3145 struct mbuf *mb, *mreq; 3146 struct vnode *vp; 3147 struct vattr at; 3148 nfsrvfh_t nsfh; 3149 u_quad_t frev, tval; 3150 3151 nfsm_srvmtofh(&nsfh); 3152 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3153 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3154 if (error) { 3155 nfsm_reply(NFSX_UNSIGNED); 3156 nfsm_srvpostop_attr(getret, &at); 3157 return (0); 3158 } 3159 sf = malloc(sizeof(*sf), M_TEMP, M_WAITOK); 3160 error = VFS_STATVFS(vp->v_mount, sf, lwp); 3161 getret = VOP_GETATTR(vp, &at, cred, lwp); 3162 vput(vp); 3163 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3)); 3164 if (v3) 3165 nfsm_srvpostop_attr(getret, &at); 3166 if (error) { 3167 free(sf, M_TEMP); 3168 return (0); 3169 } 3170 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3)); 3171 if (v3) { 3172 tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_frsize); 3173 txdr_hyper(tval, &sfp->sf_tbytes); 3174 tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_frsize); 3175 txdr_hyper(tval, &sfp->sf_fbytes); 3176 tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_frsize); 3177 txdr_hyper(tval, &sfp->sf_abytes); 3178 tval = (u_quad_t)sf->f_files; 3179 txdr_hyper(tval, &sfp->sf_tfiles); 3180 tval = (u_quad_t)sf->f_ffree; 3181 txdr_hyper(tval, &sfp->sf_ffiles); 3182 txdr_hyper(tval, &sfp->sf_afiles); 3183 sfp->sf_invarsec = 0; 3184 } else { 3185 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA); 3186 sfp->sf_bsize = txdr_unsigned(sf->f_frsize); 3187 sfp->sf_blocks = txdr_unsigned(sf->f_blocks); 3188 sfp->sf_bfree = txdr_unsigned(sf->f_bfree); 3189 sfp->sf_bavail = txdr_unsigned(sf->f_bavail); 3190 } 3191 nfsmout: 3192 if (sf) 3193 free(sf, M_TEMP); 3194 return error; 3195 } 3196 3197 /* 3198 * nfs fsinfo service 3199 */ 3200 int 3201 nfsrv_fsinfo(nfsd, slp, lwp, mrq) 3202 struct nfsrv_descript *nfsd; 3203 struct nfssvc_sock *slp; 3204 struct lwp *lwp; 3205 struct mbuf **mrq; 3206 { 3207 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3208 struct mbuf *nam = nfsd->nd_nam; 3209 char *dpos = nfsd->nd_dpos; 3210 kauth_cred_t cred = nfsd->nd_cr; 3211 u_int32_t *tl; 3212 struct nfsv3_fsinfo *sip; 3213 int32_t t1; 3214 char *bpos; 3215 int error = 0, rdonly, cache = 0, getret = 1; 3216 uint32_t maxdata; 3217 char *cp2; 3218 struct mbuf *mb, *mreq; 3219 struct vnode *vp; 3220 struct vattr at; 3221 nfsrvfh_t nsfh; 3222 u_quad_t frev, maxfsize; 3223 struct statvfs *sb; 3224 3225 nfsm_srvmtofh(&nsfh); 3226 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3227 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3228 if (error) { 3229 nfsm_reply(NFSX_UNSIGNED); 3230 nfsm_srvpostop_attr(getret, &at); 3231 return (0); 3232 } 3233 3234 /* XXX Try to make a guess on the max file size. */ 3235 sb = malloc(sizeof(*sb), M_TEMP, M_WAITOK); 3236 VFS_STATVFS(vp->v_mount, sb, (struct lwp *)0); 3237 maxfsize = (u_quad_t)0x80000000 * sb->f_frsize - 1; 3238 free(sb, M_TEMP); 3239 3240 getret = VOP_GETATTR(vp, &at, cred, lwp); 3241 vput(vp); 3242 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO); 3243 nfsm_srvpostop_attr(getret, &at); 3244 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO); 3245 3246 /* 3247 * XXX 3248 * There should be file system VFS OP(s) to get this information. 3249 * For now, assume ufs. 3250 */ 3251 if (slp->ns_so->so_type == SOCK_DGRAM) 3252 maxdata = NFS_MAXDGRAMDATA; 3253 else 3254 maxdata = NFS_MAXDATA; 3255 sip->fs_rtmax = txdr_unsigned(maxdata); 3256 sip->fs_rtpref = txdr_unsigned(maxdata); 3257 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE); 3258 sip->fs_wtmax = txdr_unsigned(maxdata); 3259 sip->fs_wtpref = txdr_unsigned(maxdata); 3260 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE); 3261 sip->fs_dtpref = txdr_unsigned(maxdata); 3262 txdr_hyper(maxfsize, &sip->fs_maxfilesize); 3263 sip->fs_timedelta.nfsv3_sec = 0; 3264 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1); 3265 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK | 3266 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS | 3267 NFSV3FSINFO_CANSETTIME); 3268 nfsm_srvdone; 3269 } 3270 3271 /* 3272 * nfs pathconf service 3273 */ 3274 int 3275 nfsrv_pathconf(nfsd, slp, lwp, mrq) 3276 struct nfsrv_descript *nfsd; 3277 struct nfssvc_sock *slp; 3278 struct lwp *lwp; 3279 struct mbuf **mrq; 3280 { 3281 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3282 struct mbuf *nam = nfsd->nd_nam; 3283 char *dpos = nfsd->nd_dpos; 3284 kauth_cred_t cred = nfsd->nd_cr; 3285 u_int32_t *tl; 3286 struct nfsv3_pathconf *pc; 3287 int32_t t1; 3288 char *bpos; 3289 int error = 0, rdonly, cache = 0, getret = 1; 3290 register_t linkmax, namemax, chownres, notrunc; 3291 char *cp2; 3292 struct mbuf *mb, *mreq; 3293 struct vnode *vp; 3294 struct vattr at; 3295 nfsrvfh_t nsfh; 3296 u_quad_t frev; 3297 3298 nfsm_srvmtofh(&nsfh); 3299 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3300 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3301 if (error) { 3302 nfsm_reply(NFSX_UNSIGNED); 3303 nfsm_srvpostop_attr(getret, &at); 3304 return (0); 3305 } 3306 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax); 3307 if (!error) 3308 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax); 3309 if (!error) 3310 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres); 3311 if (!error) 3312 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc); 3313 getret = VOP_GETATTR(vp, &at, cred, lwp); 3314 vput(vp); 3315 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF); 3316 nfsm_srvpostop_attr(getret, &at); 3317 if (error) 3318 return (0); 3319 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF); 3320 3321 pc->pc_linkmax = txdr_unsigned(linkmax); 3322 pc->pc_namemax = txdr_unsigned(namemax); 3323 pc->pc_notrunc = txdr_unsigned(notrunc); 3324 pc->pc_chownrestricted = txdr_unsigned(chownres); 3325 3326 /* 3327 * These should probably be supported by VOP_PATHCONF(), but 3328 * until msdosfs is exportable (why would you want to?), the 3329 * Unix defaults should be ok. 3330 */ 3331 pc->pc_caseinsensitive = nfs_false; 3332 pc->pc_casepreserving = nfs_true; 3333 nfsm_srvdone; 3334 } 3335 3336 /* 3337 * Null operation, used by clients to ping server 3338 */ 3339 /* ARGSUSED */ 3340 int 3341 nfsrv_null(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, 3342 struct lwp *lwp, struct mbuf **mrq) 3343 { 3344 struct mbuf *mrep = nfsd->nd_mrep; 3345 char *bpos; 3346 int error = NFSERR_RETVOID, cache = 0; 3347 struct mbuf *mb, *mreq; 3348 u_quad_t frev; 3349 3350 nfsm_reply(0); 3351 return (0); 3352 } 3353 3354 /* 3355 * No operation, used for obsolete procedures 3356 */ 3357 /* ARGSUSED */ 3358 int 3359 nfsrv_noop(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, 3360 struct lwp *lwp, struct mbuf **mrq) 3361 { 3362 struct mbuf *mrep = nfsd->nd_mrep; 3363 char *bpos; 3364 int error, cache = 0; 3365 struct mbuf *mb, *mreq; 3366 u_quad_t frev; 3367 3368 if (nfsd->nd_repstat) 3369 error = nfsd->nd_repstat; 3370 else 3371 error = EPROCUNAVAIL; 3372 nfsm_reply(0); 3373 return (0); 3374 } 3375 3376 /* 3377 * Perform access checking for vnodes obtained from file handles that would 3378 * refer to files already opened by a Unix client. You cannot just use 3379 * vn_writechk() and VOP_ACCESS() for two reasons. 3380 * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case 3381 * 2 - The owner is to be given access irrespective of mode bits for some 3382 * operations, so that processes that chmod after opening a file don't 3383 * break. I don't like this because it opens a security hole, but since 3384 * the nfs server opens a security hole the size of a barn door anyhow, 3385 * what the heck. 3386 * 3387 * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS() 3388 * will return EPERM instead of EACCESS. EPERM is always an error. 3389 */ 3390 int 3391 nfsrv_access(vp, flags, cred, rdonly, lwp, override) 3392 struct vnode *vp; 3393 int flags; 3394 kauth_cred_t cred; 3395 int rdonly; 3396 struct lwp *lwp; 3397 int override; 3398 { 3399 struct vattr vattr; 3400 int error; 3401 if (flags & VWRITE) { 3402 /* Just vn_writechk() changed to check rdonly */ 3403 /* 3404 * Disallow write attempts on read-only file systems; 3405 * unless the file is a socket or a block or character 3406 * device resident on the file system. 3407 */ 3408 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 3409 switch (vp->v_type) { 3410 case VREG: 3411 case VDIR: 3412 case VLNK: 3413 return (EROFS); 3414 default: 3415 break; 3416 } 3417 } 3418 3419 /* 3420 * If the vnode is in use as a process's text, 3421 * we can't allow writing. 3422 */ 3423 if (vp->v_iflag & VI_TEXT) 3424 return (ETXTBSY); 3425 } 3426 error = VOP_GETATTR(vp, &vattr, cred, lwp); 3427 if (error) 3428 return (error); 3429 error = VOP_ACCESS(vp, flags, cred, lwp); 3430 /* 3431 * Allow certain operations for the owner (reads and writes 3432 * on files that are already open). 3433 */ 3434 if (override && error == EACCES && kauth_cred_geteuid(cred) == vattr.va_uid) 3435 error = 0; 3436 return error; 3437 } 3438
Indexes created Thu Oct 02 14:10:14 GMT 2025