nfs_serv.c revision 1.95.2.10
1 /* $NetBSD: nfs_serv.c,v 1.95.2.10 2008/02/27 08:37:04 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 1.95.2.10 2008/02/27 08:37:04 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 const 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); 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); 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); 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); 323 postat_ret = VOP_GETATTR(vp, &va, cred); 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); 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); 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); 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); 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)) != 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); 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); 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, 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 owp = NULL; 1112 mutex_enter(&nfsd_lock); 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 mutex_exit(&nfsd_lock); 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 mutex_enter(&nfsd_lock); 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 mutex_exit(&nfsd_lock); 1172 1173 mrep = nfsd->nd_mrep; 1174 nfsd->nd_mrep = NULL; 1175 cred = nfsd->nd_cr; 1176 v3 = (nfsd->nd_flag & ND_NFSV3); 1177 forat_ret = aftat_ret = 1; 1178 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp, 1179 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH), 1180 false); 1181 if (!error) { 1182 if (v3) 1183 forat_ret = VOP_GETATTR(vp, &forat, cred); 1184 if (vp->v_type != VREG) { 1185 if (v3) 1186 error = EINVAL; 1187 else 1188 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 1189 } 1190 } else 1191 vp = NULL; 1192 if (!error) { 1193 nqsrv_getl(vp, ND_WRITE); 1194 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1); 1195 } 1196 1197 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE) 1198 ioflags = IO_NODELOCKED; 1199 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC) 1200 ioflags = (IO_SYNC | IO_NODELOCKED); 1201 else 1202 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 1203 uiop->uio_rw = UIO_WRITE; 1204 uiop->uio_offset = nfsd->nd_off; 1205 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off; 1206 UIO_SETUP_SYSSPACE(uiop); 1207 if (uiop->uio_resid > 0) { 1208 mp = mrep; 1209 i = 0; 1210 while (mp) { 1211 if (mp->m_len > 0) 1212 i++; 1213 mp = mp->m_next; 1214 } 1215 uiop->uio_iovcnt = i; 1216 iov = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK); 1217 uiop->uio_iov = ivp = iov; 1218 mp = mrep; 1219 while (mp) { 1220 if (mp->m_len > 0) { 1221 ivp->iov_base = mtod(mp, void *); 1222 ivp->iov_len = mp->m_len; 1223 ivp++; 1224 } 1225 mp = mp->m_next; 1226 } 1227 if (!error) { 1228 error = VOP_WRITE(vp, uiop, ioflags, cred); 1229 nfsstats.srvvop_writes++; 1230 } 1231 free((void *)iov, M_TEMP); 1232 } 1233 m_freem(mrep); 1234 if (vp) { 1235 aftat_ret = VOP_GETATTR(vp, &va, cred); 1236 vput(vp); 1237 } 1238 1239 /* 1240 * Loop around generating replies for all write rpcs that have 1241 * now been completed. 1242 */ 1243 swp = nfsd; 1244 do { 1245 if (error) { 1246 nfsm_writereply(NFSX_WCCDATA(v3), v3); 1247 if (v3) { 1248 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1249 } 1250 } else { 1251 nfsm_writereply(NFSX_PREOPATTR(v3) + 1252 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED + 1253 NFSX_WRITEVERF(v3), v3); 1254 if (v3) { 1255 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1256 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1257 *tl++ = txdr_unsigned(nfsd->nd_len); 1258 *tl++ = txdr_unsigned(swp->nd_stable); 1259 /* 1260 * Actually, there is no need to txdr these fields, 1261 * but it may make the values more human readable, 1262 * for debugging purposes. 1263 */ 1264 *tl++ = txdr_unsigned(boottime.tv_sec); 1265 *tl = txdr_unsigned(boottime.tv_usec); 1266 } else { 1267 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1268 nfsm_srvfillattr(&va, fp); 1269 } 1270 } 1271 nfsd->nd_mreq = mreq; 1272 if (nfsd->nd_mrep) 1273 panic("nfsrv_write: nd_mrep not free"); 1274 1275 /* 1276 * Done. Put it at the head of the timer queue so that 1277 * the final phase can return the reply. 1278 */ 1279 mutex_enter(&nfsd_lock); 1280 if (nfsd != swp) { 1281 nfsd->nd_time = 0; 1282 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1283 } 1284 nfsd = LIST_FIRST(&swp->nd_coalesce); 1285 if (nfsd) { 1286 LIST_REMOVE(nfsd, nd_tq); 1287 } 1288 mutex_exit(&nfsd_lock); 1289 } while (nfsd); 1290 swp->nd_time = 0; 1291 1292 mutex_enter(&nfsd_lock); 1293 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq); 1294 mutex_exit(&nfsd_lock); 1295 goto loop1; 1296 } 1297 mutex_exit(&nfsd_lock); 1298 nfs_timer_start(); 1299 1300 /* 1301 * Search for a reply to return. 1302 */ 1303 mutex_enter(&nfsd_lock); 1304 LIST_FOREACH(nfsd, &slp->ns_tq, nd_tq) { 1305 if (nfsd->nd_mreq) { 1306 LIST_REMOVE(nfsd, nd_tq); 1307 *mrq = nfsd->nd_mreq; 1308 *ndp = nfsd; 1309 break; 1310 } 1311 } 1312 mutex_exit(&nfsd_lock); 1313 return (0); 1314 } 1315 1316 /* 1317 * Coalesce the write request nfsd into owp. To do this we must: 1318 * - remove nfsd from the queues 1319 * - merge nfsd->nd_mrep into owp->nd_mrep 1320 * - update the nd_eoff and nd_stable for owp 1321 * - put nfsd on owp's nd_coalesce list 1322 * NB: Must be called at splsoftclock(). 1323 */ 1324 void 1325 nfsrvw_coalesce(owp, nfsd) 1326 struct nfsrv_descript *owp; 1327 struct nfsrv_descript *nfsd; 1328 { 1329 int overlap; 1330 struct mbuf *mp; 1331 struct nfsrv_descript *m; 1332 1333 KASSERT(mutex_owned(&nfsd_lock)); 1334 1335 LIST_REMOVE(nfsd, nd_hash); 1336 LIST_REMOVE(nfsd, nd_tq); 1337 if (owp->nd_eoff < nfsd->nd_eoff) { 1338 overlap = owp->nd_eoff - nfsd->nd_off; 1339 if (overlap < 0) 1340 panic("nfsrv_coalesce: bad off"); 1341 if (overlap > 0) 1342 m_adj(nfsd->nd_mrep, overlap); 1343 mp = owp->nd_mrep; 1344 while (mp->m_next) 1345 mp = mp->m_next; 1346 mp->m_next = nfsd->nd_mrep; 1347 owp->nd_eoff = nfsd->nd_eoff; 1348 } else 1349 m_freem(nfsd->nd_mrep); 1350 nfsd->nd_mrep = NULL; 1351 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC) 1352 owp->nd_stable = NFSV3WRITE_FILESYNC; 1353 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC && 1354 owp->nd_stable == NFSV3WRITE_UNSTABLE) 1355 owp->nd_stable = NFSV3WRITE_DATASYNC; 1356 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq); 1357 /* 1358 * nfsd might hold coalesce elements! Move them to owp. 1359 * Otherwise, requests may be lost and clients will be stuck. 1360 */ 1361 while ((m = LIST_FIRST(&nfsd->nd_coalesce)) != NULL) { 1362 LIST_REMOVE(m, nd_tq); 1363 LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq); 1364 } 1365 } 1366 1367 /* 1368 * nfs create service 1369 * now does a truncate to 0 length via. setattr if it already exists 1370 */ 1371 int 1372 nfsrv_create(nfsd, slp, lwp, mrq) 1373 struct nfsrv_descript *nfsd; 1374 struct nfssvc_sock *slp; 1375 struct lwp *lwp; 1376 struct mbuf **mrq; 1377 { 1378 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1379 struct mbuf *nam = nfsd->nd_nam; 1380 char *dpos = nfsd->nd_dpos; 1381 kauth_cred_t cred = nfsd->nd_cr; 1382 struct nfs_fattr *fp; 1383 struct vattr va, dirfor, diraft; 1384 struct nfsv2_sattr *sp; 1385 u_int32_t *tl; 1386 struct nameidata nd; 1387 char *cp; 1388 int32_t t1; 1389 char *bpos; 1390 int error = 0, cache = 0, len, tsize, dirfor_ret = 1, diraft_ret = 1; 1391 int rdev = 0; 1392 int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0; 1393 char *cp2; 1394 struct mbuf *mb, *mreq; 1395 struct vnode *vp = NULL, *dirp = NULL; 1396 nfsrvfh_t nsfh; 1397 u_quad_t frev, tempsize; 1398 u_char cverf[NFSX_V3CREATEVERF]; 1399 1400 nd.ni_cnd.cn_nameiop = 0; 1401 nfsm_srvmtofh(&nsfh); 1402 nfsm_srvnamesiz(len); 1403 nd.ni_cnd.cn_cred = cred; 1404 nd.ni_cnd.cn_nameiop = CREATE; 1405 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1406 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 1407 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1408 if (dirp && v3) { 1409 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred); 1410 } 1411 if (error) { 1412 nfsm_reply(NFSX_WCCDATA(v3)); 1413 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1414 if (dirp) 1415 vrele(dirp); 1416 return (0); 1417 } 1418 VATTR_NULL(&va); 1419 if (v3) { 1420 va.va_mode = 0; 1421 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1422 how = fxdr_unsigned(int, *tl); 1423 switch (how) { 1424 case NFSV3CREATE_GUARDED: 1425 if (nd.ni_vp) { 1426 error = EEXIST; 1427 break; 1428 } 1429 case NFSV3CREATE_UNCHECKED: 1430 nfsm_srvsattr(&va); 1431 break; 1432 case NFSV3CREATE_EXCLUSIVE: 1433 nfsm_dissect(cp, void *, NFSX_V3CREATEVERF); 1434 memcpy(cverf, cp, NFSX_V3CREATEVERF); 1435 exclusive_flag = 1; 1436 break; 1437 }; 1438 va.va_type = VREG; 1439 } else { 1440 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 1441 va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode)); 1442 if (va.va_type == VNON) 1443 va.va_type = VREG; 1444 va.va_mode = nfstov_mode(sp->sa_mode); 1445 switch (va.va_type) { 1446 case VREG: 1447 tsize = fxdr_unsigned(int32_t, sp->sa_size); 1448 if (tsize != -1) 1449 va.va_size = (u_quad_t)tsize; 1450 break; 1451 case VCHR: 1452 case VBLK: 1453 case VFIFO: 1454 rdev = fxdr_unsigned(int32_t, sp->sa_size); 1455 break; 1456 default: 1457 break; 1458 }; 1459 } 1460 1461 /* 1462 * Iff doesn't exist, create it 1463 * otherwise just truncate to 0 length 1464 * should I set the mode too ?? 1465 */ 1466 if (nd.ni_vp == NULL) { 1467 if (va.va_type == VREG || va.va_type == VSOCK) { 1468 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1469 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1470 if (!error) { 1471 if (exclusive_flag) { 1472 exclusive_flag = 0; 1473 VATTR_NULL(&va); 1474 /* 1475 * XXX 1476 * assuming NFSX_V3CREATEVERF 1477 * == sizeof(nfstime3) 1478 */ 1479 fxdr_nfsv3time(cverf, &va.va_atime); 1480 error = VOP_SETATTR(nd.ni_vp, &va, 1481 cred); 1482 } 1483 } 1484 } else if (va.va_type == VCHR || va.va_type == VBLK || 1485 va.va_type == VFIFO) { 1486 if (va.va_type == VCHR && rdev == 0xffffffff) 1487 va.va_type = VFIFO; 1488 if (va.va_type != VFIFO && 1489 (error = kauth_authorize_generic(cred, 1490 KAUTH_GENERIC_ISSUSER, NULL))) { 1491 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1492 vput(nd.ni_dvp); 1493 nfsm_reply(0); 1494 return (error); 1495 } else 1496 va.va_rdev = (dev_t)rdev; 1497 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1498 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, 1499 &va); 1500 if (error) { 1501 nfsm_reply(0); 1502 } 1503 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1504 vput(nd.ni_vp); 1505 vrele(nd.ni_dvp); 1506 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1507 error = EINVAL; 1508 nfsm_reply(0); 1509 } 1510 } else { 1511 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1512 vput(nd.ni_dvp); 1513 error = ENXIO; 1514 } 1515 vp = nd.ni_vp; 1516 } else { 1517 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1518 vp = nd.ni_vp; 1519 if (nd.ni_dvp == vp) 1520 vrele(nd.ni_dvp); 1521 else 1522 vput(nd.ni_dvp); 1523 if (!error && va.va_size != -1) { 1524 error = nfsrv_access(vp, VWRITE, cred, 1525 (nd.ni_cnd.cn_flags & RDONLY), lwp, 0); 1526 if (!error) { 1527 nqsrv_getl(vp, ND_WRITE); 1528 tempsize = va.va_size; 1529 VATTR_NULL(&va); 1530 va.va_size = tempsize; 1531 error = VOP_SETATTR(vp, &va, cred); 1532 } 1533 } 1534 if (error) 1535 vput(vp); 1536 } 1537 if (!error) { 1538 error = nfsrv_composefh(vp, &nsfh, v3); 1539 if (!error) 1540 error = VOP_GETATTR(vp, &va, cred); 1541 vput(vp); 1542 } 1543 if (v3) { 1544 if (exclusive_flag && !error) { 1545 /* 1546 * XXX assuming NFSX_V3CREATEVERF == sizeof(nfstime3) 1547 */ 1548 char oldverf[NFSX_V3CREATEVERF]; 1549 1550 txdr_nfsv3time(&va.va_atime, oldverf); 1551 if (memcmp(cverf, oldverf, NFSX_V3CREATEVERF)) 1552 error = EEXIST; 1553 } 1554 if (dirp) { 1555 diraft_ret = VOP_GETATTR(dirp, &diraft, cred); 1556 } 1557 } 1558 if (dirp) { 1559 vrele(dirp); 1560 } 1561 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3)); 1562 if (v3) { 1563 if (!error) { 1564 nfsm_srvpostop_fh(&nsfh); 1565 nfsm_srvpostop_attr(0, &va); 1566 } 1567 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1568 } else { 1569 nfsm_srvfhtom(&nsfh, v3); 1570 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1571 nfsm_srvfillattr(&va, fp); 1572 } 1573 return (0); 1574 nfsmout: 1575 if (dirp) 1576 vrele(dirp); 1577 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1578 if (nd.ni_dvp == nd.ni_vp) 1579 vrele(nd.ni_dvp); 1580 else 1581 vput(nd.ni_dvp); 1582 if (nd.ni_vp) 1583 vput(nd.ni_vp); 1584 return (error); 1585 } 1586 1587 /* 1588 * nfs v3 mknod service 1589 */ 1590 int 1591 nfsrv_mknod(nfsd, slp, lwp, mrq) 1592 struct nfsrv_descript *nfsd; 1593 struct nfssvc_sock *slp; 1594 struct lwp *lwp; 1595 struct mbuf **mrq; 1596 { 1597 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1598 struct mbuf *nam = nfsd->nd_nam; 1599 char *dpos = nfsd->nd_dpos; 1600 kauth_cred_t cred = nfsd->nd_cr; 1601 struct vattr va, dirfor, diraft; 1602 u_int32_t *tl; 1603 struct nameidata nd; 1604 int32_t t1; 1605 char *bpos; 1606 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 1607 u_int32_t major, minor; 1608 enum vtype vtyp; 1609 char *cp2; 1610 struct mbuf *mb, *mreq; 1611 struct vnode *vp, *dirp = (struct vnode *)0; 1612 nfsrvfh_t nsfh; 1613 u_quad_t frev; 1614 1615 nd.ni_cnd.cn_nameiop = 0; 1616 nfsm_srvmtofh(&nsfh); 1617 nfsm_srvnamesiz(len); 1618 nd.ni_cnd.cn_cred = cred; 1619 nd.ni_cnd.cn_nameiop = CREATE; 1620 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1621 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 1622 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1623 if (dirp) 1624 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred); 1625 if (error) { 1626 nfsm_reply(NFSX_WCCDATA(1)); 1627 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1628 if (dirp) 1629 vrele(dirp); 1630 return (0); 1631 } 1632 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1633 vtyp = nfsv3tov_type(*tl); 1634 if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) { 1635 error = NFSERR_BADTYPE; 1636 goto abort; 1637 } 1638 VATTR_NULL(&va); 1639 va.va_mode = 0; 1640 nfsm_srvsattr(&va); 1641 if (vtyp == VCHR || vtyp == VBLK) { 1642 dev_t rdev; 1643 1644 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1645 major = fxdr_unsigned(u_int32_t, *tl++); 1646 minor = fxdr_unsigned(u_int32_t, *tl); 1647 rdev = makedev(major, minor); 1648 if (major(rdev) != major || minor(rdev) != minor) { 1649 error = EINVAL; 1650 goto abort; 1651 } 1652 va.va_rdev = rdev; 1653 } 1654 1655 /* 1656 * Iff doesn't exist, create it. 1657 */ 1658 if (nd.ni_vp) { 1659 error = EEXIST; 1660 abort: 1661 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1662 if (nd.ni_dvp == nd.ni_vp) 1663 vrele(nd.ni_dvp); 1664 else 1665 vput(nd.ni_dvp); 1666 if (nd.ni_vp) 1667 vput(nd.ni_vp); 1668 goto out; 1669 } 1670 va.va_type = vtyp; 1671 if (vtyp == VSOCK) { 1672 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1673 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1674 } else { 1675 if (va.va_type != VFIFO && 1676 (error = kauth_authorize_generic(cred, 1677 KAUTH_GENERIC_ISSUSER, NULL))) { 1678 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1679 vput(nd.ni_dvp); 1680 goto out; 1681 } 1682 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1683 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1684 if (error) 1685 goto out; 1686 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1687 vput(nd.ni_vp); 1688 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1689 error = EINVAL; 1690 } 1691 } 1692 out: 1693 vp = nd.ni_vp; 1694 if (!error) { 1695 error = nfsrv_composefh(vp, &nsfh, true); 1696 if (!error) 1697 error = VOP_GETATTR(vp, &va, cred); 1698 vput(vp); 1699 } 1700 if (dirp) { 1701 diraft_ret = VOP_GETATTR(dirp, &diraft, cred); 1702 vrele(dirp); 1703 } 1704 nfsm_reply(NFSX_SRVFH(&nsfh, true) + NFSX_POSTOPATTR(1) + 1705 NFSX_WCCDATA(1)); 1706 if (!error) { 1707 nfsm_srvpostop_fh(&nsfh); 1708 nfsm_srvpostop_attr(0, &va); 1709 } 1710 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1711 return (0); 1712 nfsmout: 1713 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1714 if (nd.ni_dvp == nd.ni_vp) 1715 vrele(nd.ni_dvp); 1716 else 1717 vput(nd.ni_dvp); 1718 if (nd.ni_vp) 1719 vput(nd.ni_vp); 1720 if (dirp) 1721 vrele(dirp); 1722 return (error); 1723 } 1724 1725 /* 1726 * nfs remove service 1727 */ 1728 int 1729 nfsrv_remove(nfsd, slp, lwp, mrq) 1730 struct nfsrv_descript *nfsd; 1731 struct nfssvc_sock *slp; 1732 struct lwp *lwp; 1733 struct mbuf **mrq; 1734 { 1735 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1736 struct mbuf *nam = nfsd->nd_nam; 1737 char *dpos = nfsd->nd_dpos; 1738 kauth_cred_t cred = nfsd->nd_cr; 1739 struct nameidata nd; 1740 u_int32_t *tl; 1741 int32_t t1; 1742 char *bpos; 1743 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 1744 int v3 = (nfsd->nd_flag & ND_NFSV3); 1745 char *cp2; 1746 struct mbuf *mb, *mreq; 1747 struct vnode *vp, *dirp; 1748 struct vattr dirfor, diraft; 1749 nfsrvfh_t nsfh; 1750 u_quad_t frev; 1751 1752 #ifndef nolint 1753 vp = (struct vnode *)0; 1754 #endif 1755 nfsm_srvmtofh(&nsfh); 1756 nfsm_srvnamesiz(len); 1757 nd.ni_cnd.cn_cred = cred; 1758 nd.ni_cnd.cn_nameiop = DELETE; 1759 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1760 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 1761 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1762 if (dirp && v3) { 1763 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred); 1764 } 1765 if (!error) { 1766 vp = nd.ni_vp; 1767 if (vp->v_type == VDIR && 1768 (error = kauth_authorize_generic(cred, 1769 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 1770 goto out; 1771 /* 1772 * The root of a mounted filesystem cannot be deleted. 1773 */ 1774 if (vp->v_vflag & VV_ROOT) { 1775 error = EBUSY; 1776 goto out; 1777 } 1778 out: 1779 if (!error) { 1780 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1781 nqsrv_getl(vp, ND_WRITE); 1782 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1783 } else { 1784 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1785 if (nd.ni_dvp == vp) 1786 vrele(nd.ni_dvp); 1787 else 1788 vput(nd.ni_dvp); 1789 vput(vp); 1790 } 1791 } 1792 if (dirp) { 1793 if (v3) { 1794 diraft_ret = VOP_GETATTR(dirp, &diraft, cred); 1795 } 1796 vrele(dirp); 1797 } 1798 nfsm_reply(NFSX_WCCDATA(v3)); 1799 if (v3) { 1800 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1801 return (0); 1802 } 1803 nfsm_srvdone; 1804 } 1805 1806 /* 1807 * nfs rename service 1808 */ 1809 int 1810 nfsrv_rename(nfsd, slp, lwp, mrq) 1811 struct nfsrv_descript *nfsd; 1812 struct nfssvc_sock *slp; 1813 struct lwp *lwp; 1814 struct mbuf **mrq; 1815 { 1816 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1817 struct mbuf *nam = nfsd->nd_nam; 1818 char *dpos = nfsd->nd_dpos; 1819 kauth_cred_t cred = nfsd->nd_cr; 1820 u_int32_t *tl; 1821 int32_t t1; 1822 char *bpos; 1823 int error = 0, cache = 0, fdirfor_ret = 1, fdiraft_ret = 1; 1824 uint32_t len, len2; 1825 int tdirfor_ret = 1, tdiraft_ret = 1; 1826 int v3 = (nfsd->nd_flag & ND_NFSV3); 1827 char *cp2; 1828 struct mbuf *mb, *mreq; 1829 struct nameidata fromnd, tond; 1830 struct vnode *fvp, *tvp, *tdvp; 1831 struct vnode *fdirp = NULL, *tdirp = NULL; 1832 struct mount *localfs = NULL; 1833 struct vattr fdirfor, fdiraft, tdirfor, tdiraft; 1834 nfsrvfh_t fnsfh, tnsfh; 1835 u_quad_t frev; 1836 uid_t saved_uid; 1837 uint32_t saveflag; 1838 1839 #ifndef nolint 1840 fvp = (struct vnode *)0; 1841 #endif 1842 fromnd.ni_cnd.cn_nameiop = 0; 1843 tond.ni_cnd.cn_nameiop = 0; 1844 nfsm_srvmtofh(&fnsfh); 1845 nfsm_srvnamesiz(len); 1846 /* 1847 * Remember our original uid so that we can reset cr_uid before 1848 * the second nfs_namei() call, in case it is remapped. 1849 */ 1850 saved_uid = kauth_cred_geteuid(cred); 1851 fromnd.ni_cnd.cn_cred = cred; 1852 fromnd.ni_cnd.cn_nameiop = DELETE; 1853 fromnd.ni_cnd.cn_flags = LOCKPARENT | SAVESTART; 1854 error = nfs_namei(&fromnd, &fnsfh, len, slp, nam, &md, 1855 &dpos, &fdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1856 if (fdirp && v3) { 1857 fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred); 1858 } 1859 if (error) { 1860 nfsm_reply(2 * NFSX_WCCDATA(v3)); 1861 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 1862 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 1863 if (fdirp) 1864 vrele(fdirp); 1865 return (0); 1866 } 1867 if (fromnd.ni_dvp != fromnd.ni_vp) { 1868 VOP_UNLOCK(fromnd.ni_dvp, 0); 1869 } 1870 fvp = fromnd.ni_vp; 1871 1872 localfs = fvp->v_mount; 1873 error = VFS_RENAMELOCK_ENTER(localfs); 1874 if (error) { 1875 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1876 vrele(fromnd.ni_dvp); 1877 vrele(fvp); 1878 goto out1; 1879 } 1880 1881 /* Copied, regrettably, from vfs_syscalls.c (q.v.) */ 1882 vrele(fvp); 1883 saveflag = fromnd.ni_cnd.cn_flags & SAVESTART; 1884 fromnd.ni_cnd.cn_flags &= ~SAVESTART; 1885 vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY); 1886 error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd); 1887 fromnd.ni_cnd.cn_flags |= saveflag; 1888 if (error) { 1889 VOP_UNLOCK(fromnd.ni_dvp, 0); 1890 VFS_RENAMELOCK_EXIT(localfs); 1891 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1892 vrele(fromnd.ni_dvp); 1893 goto out1; 1894 } 1895 VOP_UNLOCK(fromnd.ni_vp, 0); 1896 if (fromnd.ni_dvp != fromnd.ni_vp) 1897 VOP_UNLOCK(fromnd.ni_dvp, 0); 1898 fvp = fromnd.ni_vp; 1899 1900 nfsm_srvmtofh(&tnsfh); 1901 if (v3) { 1902 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 1903 len2 = fxdr_unsigned(uint32_t, *tl); 1904 /* len2 will be checked by nfs_namei */ 1905 } 1906 else { 1907 /* NFSv2 */ 1908 nfsm_strsiz(len2, NFS_MAXNAMLEN); 1909 } 1910 kauth_cred_seteuid(cred, saved_uid); 1911 tond.ni_cnd.cn_cred = cred; 1912 tond.ni_cnd.cn_nameiop = RENAME; 1913 tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART; 1914 error = nfs_namei(&tond, &tnsfh, len2, slp, nam, &md, 1915 &dpos, &tdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1916 if (tdirp && v3) { 1917 tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred); 1918 } 1919 if (error) { 1920 VFS_RENAMELOCK_EXIT(localfs); 1921 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1922 vrele(fromnd.ni_dvp); 1923 vrele(fvp); 1924 goto out1; 1925 } 1926 tdvp = tond.ni_dvp; 1927 tvp = tond.ni_vp; 1928 if (tvp != NULL) { 1929 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 1930 if (v3) 1931 error = EEXIST; 1932 else 1933 error = EISDIR; 1934 goto out; 1935 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 1936 if (v3) 1937 error = EEXIST; 1938 else 1939 error = ENOTDIR; 1940 goto out; 1941 } 1942 if (tvp->v_type == VDIR && tvp->v_mountedhere) { 1943 if (v3) 1944 error = EXDEV; 1945 else 1946 error = ENOTEMPTY; 1947 goto out; 1948 } 1949 } 1950 if (fvp->v_type == VDIR && fvp->v_mountedhere) { 1951 if (v3) 1952 error = EXDEV; 1953 else 1954 error = ENOTEMPTY; 1955 goto out; 1956 } 1957 if (fvp->v_mount != tdvp->v_mount) { 1958 if (v3) 1959 error = EXDEV; 1960 else 1961 error = ENOTEMPTY; 1962 goto out; 1963 } 1964 if (fvp == tdvp) { 1965 if (v3) 1966 error = EINVAL; 1967 else 1968 error = ENOTEMPTY; 1969 } 1970 /* 1971 * If source is the same as the destination (that is the 1972 * same vnode with the same name in the same directory), 1973 * then there is nothing to do. 1974 */ 1975 if (fvp == tvp && fromnd.ni_dvp == tdvp && 1976 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 1977 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr, 1978 fromnd.ni_cnd.cn_namelen)) 1979 error = -1; 1980 out: 1981 if (!error) { 1982 nqsrv_getl(fromnd.ni_dvp, ND_WRITE); 1983 nqsrv_getl(tdvp, ND_WRITE); 1984 if (tvp) { 1985 nqsrv_getl(tvp, ND_WRITE); 1986 } 1987 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 1988 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 1989 VFS_RENAMELOCK_EXIT(localfs); 1990 } else { 1991 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 1992 if (tdvp == tvp) 1993 vrele(tdvp); 1994 else 1995 vput(tdvp); 1996 if (tvp) 1997 vput(tvp); 1998 VFS_RENAMELOCK_EXIT(localfs); 1999 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2000 vrele(fromnd.ni_dvp); 2001 vrele(fvp); 2002 if (error == -1) 2003 error = 0; 2004 } 2005 vrele(tond.ni_startdir); 2006 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 2007 out1: 2008 if (fdirp) { 2009 if (v3) { 2010 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred); 2011 } 2012 vrele(fdirp); 2013 } 2014 if (tdirp) { 2015 if (v3) { 2016 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred); 2017 } 2018 vrele(tdirp); 2019 } 2020 vrele(fromnd.ni_startdir); 2021 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2022 nfsm_reply(2 * NFSX_WCCDATA(v3)); 2023 if (v3) { 2024 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 2025 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 2026 } 2027 return (0); 2028 2029 nfsmout: 2030 if (fdirp) 2031 vrele(fdirp); 2032 #ifdef notdef 2033 if (tdirp) 2034 vrele(tdirp); 2035 #endif 2036 if (tond.ni_cnd.cn_nameiop) { 2037 vrele(tond.ni_startdir); 2038 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 2039 } 2040 if (localfs) { 2041 VFS_RENAMELOCK_EXIT(localfs); 2042 } 2043 if (fromnd.ni_cnd.cn_nameiop) { 2044 vrele(fromnd.ni_startdir); 2045 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2046 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2047 vrele(fromnd.ni_dvp); 2048 vrele(fvp); 2049 } 2050 return (error); 2051 } 2052 2053 /* 2054 * nfs link service 2055 */ 2056 int 2057 nfsrv_link(nfsd, slp, lwp, mrq) 2058 struct nfsrv_descript *nfsd; 2059 struct nfssvc_sock *slp; 2060 struct lwp *lwp; 2061 struct mbuf **mrq; 2062 { 2063 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2064 struct mbuf *nam = nfsd->nd_nam; 2065 char *dpos = nfsd->nd_dpos; 2066 kauth_cred_t cred = nfsd->nd_cr; 2067 struct nameidata nd; 2068 u_int32_t *tl; 2069 int32_t t1; 2070 char *bpos; 2071 int error = 0, rdonly, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 2072 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3); 2073 char *cp2; 2074 struct mbuf *mb, *mreq; 2075 struct vnode *vp, *xp, *dirp = (struct vnode *)0; 2076 struct vattr dirfor, diraft, at; 2077 nfsrvfh_t nsfh, dnsfh; 2078 u_quad_t frev; 2079 2080 nfsm_srvmtofh(&nsfh); 2081 nfsm_srvmtofh(&dnsfh); 2082 nfsm_srvnamesiz(len); 2083 error = nfsrv_fhtovp(&nsfh, false, &vp, cred, slp, nam, 2084 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 2085 if (error) { 2086 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2087 nfsm_srvpostop_attr(getret, &at); 2088 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2089 return (0); 2090 } 2091 if (vp->v_type == VDIR && (error = kauth_authorize_generic(cred, 2092 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 2093 goto out1; 2094 nd.ni_cnd.cn_cred = cred; 2095 nd.ni_cnd.cn_nameiop = CREATE; 2096 nd.ni_cnd.cn_flags = LOCKPARENT; 2097 error = nfs_namei(&nd, &dnsfh, len, slp, nam, &md, &dpos, 2098 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2099 if (dirp && v3) { 2100 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred); 2101 } 2102 if (error) 2103 goto out1; 2104 xp = nd.ni_vp; 2105 if (xp != NULL) { 2106 error = EEXIST; 2107 goto out; 2108 } 2109 xp = nd.ni_dvp; 2110 if (vp->v_mount != xp->v_mount) 2111 error = EXDEV; 2112 out: 2113 if (!error) { 2114 nqsrv_getl(vp, ND_WRITE); 2115 nqsrv_getl(xp, ND_WRITE); 2116 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2117 } else { 2118 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2119 if (nd.ni_dvp == nd.ni_vp) 2120 vrele(nd.ni_dvp); 2121 else 2122 vput(nd.ni_dvp); 2123 if (nd.ni_vp) 2124 vrele(nd.ni_vp); 2125 } 2126 out1: 2127 if (v3) 2128 getret = VOP_GETATTR(vp, &at, cred); 2129 if (dirp) { 2130 if (v3) { 2131 diraft_ret = VOP_GETATTR(dirp, &diraft, cred); 2132 } 2133 vrele(dirp); 2134 } 2135 vrele(vp); 2136 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2137 if (v3) { 2138 nfsm_srvpostop_attr(getret, &at); 2139 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2140 return (0); 2141 } 2142 nfsm_srvdone; 2143 } 2144 2145 /* 2146 * nfs symbolic link service 2147 */ 2148 int 2149 nfsrv_symlink(nfsd, slp, lwp, mrq) 2150 struct nfsrv_descript *nfsd; 2151 struct nfssvc_sock *slp; 2152 struct lwp *lwp; 2153 struct mbuf **mrq; 2154 { 2155 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2156 struct mbuf *nam = nfsd->nd_nam; 2157 char *dpos = nfsd->nd_dpos; 2158 kauth_cred_t cred = nfsd->nd_cr; 2159 struct vattr va, dirfor, diraft; 2160 struct nameidata nd; 2161 u_int32_t *tl; 2162 int32_t t1; 2163 struct nfsv2_sattr *sp; 2164 char *bpos, *pathcp = NULL, *cp2; 2165 struct uio io; 2166 struct iovec iv; 2167 int error = 0, cache = 0, dirfor_ret = 1, diraft_ret = 1; 2168 uint32_t len, len2; 2169 int v3 = (nfsd->nd_flag & ND_NFSV3); 2170 struct mbuf *mb, *mreq; 2171 struct vnode *dirp = (struct vnode *)0; 2172 nfsrvfh_t nsfh; 2173 u_quad_t frev; 2174 2175 nd.ni_cnd.cn_nameiop = 0; 2176 nfsm_srvmtofh(&nsfh); 2177 nfsm_srvnamesiz(len); 2178 nd.ni_cnd.cn_cred = cred; 2179 nd.ni_cnd.cn_nameiop = CREATE; 2180 nd.ni_cnd.cn_flags = LOCKPARENT; 2181 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 2182 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2183 if (dirp && v3) { 2184 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred); 2185 } 2186 if (error) 2187 goto out; 2188 VATTR_NULL(&va); 2189 va.va_type = VLNK; 2190 if (v3) { 2191 va.va_mode = 0; 2192 nfsm_srvsattr(&va); 2193 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 2194 len2 = fxdr_unsigned(uint32_t, *tl); 2195 if (len2 > PATH_MAX) { 2196 /* XXX should check _PC_NO_TRUNC */ 2197 error = ENAMETOOLONG; 2198 goto abortop; 2199 } 2200 } 2201 else { 2202 /* NFSv2 */ 2203 nfsm_strsiz(len2, NFS_MAXPATHLEN); 2204 } 2205 pathcp = malloc(len2 + 1, M_TEMP, M_WAITOK); 2206 iv.iov_base = pathcp; 2207 iv.iov_len = len2; 2208 io.uio_resid = len2; 2209 io.uio_offset = 0; 2210 io.uio_iov = &iv; 2211 io.uio_iovcnt = 1; 2212 io.uio_rw = UIO_READ; 2213 UIO_SETUP_SYSSPACE(&io); 2214 nfsm_mtouio(&io, len2); 2215 if (!v3) { 2216 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 2217 va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode); 2218 } 2219 *(pathcp + len2) = '\0'; 2220 if (nd.ni_vp) { 2221 error = EEXIST; 2222 abortop: 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 goto out; 2231 } 2232 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2233 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp); 2234 if (!error) { 2235 if (v3) { 2236 error = nfsrv_composefh(nd.ni_vp, &nsfh, v3); 2237 if (!error) 2238 error = VOP_GETATTR(nd.ni_vp, &va, cred); 2239 vput(nd.ni_vp); 2240 } else { 2241 vput(nd.ni_vp); 2242 } 2243 } 2244 out: 2245 if (pathcp) 2246 free(pathcp, M_TEMP); 2247 if (dirp) { 2248 if (v3) { 2249 diraft_ret = VOP_GETATTR(dirp, &diraft, cred); 2250 } 2251 vrele(dirp); 2252 } 2253 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_POSTOPATTR(v3) + 2254 NFSX_WCCDATA(v3)); 2255 if (v3) { 2256 if (!error) { 2257 nfsm_srvpostop_fh(&nsfh); 2258 nfsm_srvpostop_attr(0, &va); 2259 } 2260 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2261 } 2262 return (0); 2263 nfsmout: 2264 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2265 if (nd.ni_dvp == nd.ni_vp) 2266 vrele(nd.ni_dvp); 2267 else 2268 vput(nd.ni_dvp); 2269 if (nd.ni_vp) 2270 vrele(nd.ni_vp); 2271 if (dirp) 2272 vrele(dirp); 2273 if (pathcp) 2274 free(pathcp, M_TEMP); 2275 return (error); 2276 } 2277 2278 /* 2279 * nfs mkdir service 2280 */ 2281 int 2282 nfsrv_mkdir(nfsd, slp, lwp, mrq) 2283 struct nfsrv_descript *nfsd; 2284 struct nfssvc_sock *slp; 2285 struct lwp *lwp; 2286 struct mbuf **mrq; 2287 { 2288 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2289 struct mbuf *nam = nfsd->nd_nam; 2290 char *dpos = nfsd->nd_dpos; 2291 kauth_cred_t cred = nfsd->nd_cr; 2292 struct vattr va, dirfor, diraft; 2293 struct nfs_fattr *fp; 2294 struct nameidata nd; 2295 char *cp; 2296 u_int32_t *tl; 2297 int32_t t1; 2298 char *bpos; 2299 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 2300 int v3 = (nfsd->nd_flag & ND_NFSV3); 2301 char *cp2; 2302 struct mbuf *mb, *mreq; 2303 struct vnode *vp, *dirp = (struct vnode *)0; 2304 nfsrvfh_t nsfh; 2305 u_quad_t frev; 2306 2307 nfsm_srvmtofh(&nsfh); 2308 nfsm_srvnamesiz(len); 2309 nd.ni_cnd.cn_cred = cred; 2310 nd.ni_cnd.cn_nameiop = CREATE; 2311 nd.ni_cnd.cn_flags = LOCKPARENT; 2312 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 2313 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2314 if (dirp && v3) { 2315 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred); 2316 } 2317 if (error) { 2318 nfsm_reply(NFSX_WCCDATA(v3)); 2319 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2320 if (dirp) 2321 vrele(dirp); 2322 return (0); 2323 } 2324 VATTR_NULL(&va); 2325 if (v3) { 2326 va.va_mode = 0; 2327 nfsm_srvsattr(&va); 2328 } else { 2329 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 2330 va.va_mode = nfstov_mode(*tl++); 2331 } 2332 va.va_type = VDIR; 2333 vp = nd.ni_vp; 2334 if (vp != NULL) { 2335 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2336 if (nd.ni_dvp == vp) 2337 vrele(nd.ni_dvp); 2338 else 2339 vput(nd.ni_dvp); 2340 vrele(vp); 2341 error = EEXIST; 2342 goto out; 2343 } 2344 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2345 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 2346 if (!error) { 2347 vp = nd.ni_vp; 2348 error = nfsrv_composefh(vp, &nsfh, v3); 2349 if (!error) 2350 error = VOP_GETATTR(vp, &va, cred); 2351 vput(vp); 2352 } 2353 out: 2354 if (dirp) { 2355 if (v3) { 2356 diraft_ret = VOP_GETATTR(dirp, &diraft, cred); 2357 } 2358 vrele(dirp); 2359 } 2360 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_POSTOPATTR(v3) + 2361 NFSX_WCCDATA(v3)); 2362 if (v3) { 2363 if (!error) { 2364 nfsm_srvpostop_fh(&nsfh); 2365 nfsm_srvpostop_attr(0, &va); 2366 } 2367 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2368 } else { 2369 nfsm_srvfhtom(&nsfh, v3); 2370 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 2371 nfsm_srvfillattr(&va, fp); 2372 } 2373 return (0); 2374 nfsmout: 2375 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2376 if (nd.ni_dvp == nd.ni_vp) 2377 vrele(nd.ni_dvp); 2378 else 2379 vput(nd.ni_dvp); 2380 if (nd.ni_vp) 2381 vrele(nd.ni_vp); 2382 if (dirp) 2383 vrele(dirp); 2384 return (error); 2385 } 2386 2387 /* 2388 * nfs rmdir service 2389 */ 2390 int 2391 nfsrv_rmdir(nfsd, slp, lwp, mrq) 2392 struct nfsrv_descript *nfsd; 2393 struct nfssvc_sock *slp; 2394 struct lwp *lwp; 2395 struct mbuf **mrq; 2396 { 2397 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2398 struct mbuf *nam = nfsd->nd_nam; 2399 char *dpos = nfsd->nd_dpos; 2400 kauth_cred_t cred = nfsd->nd_cr; 2401 u_int32_t *tl; 2402 int32_t t1; 2403 char *bpos; 2404 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 2405 int v3 = (nfsd->nd_flag & ND_NFSV3); 2406 char *cp2; 2407 struct mbuf *mb, *mreq; 2408 struct vnode *vp, *dirp = (struct vnode *)0; 2409 struct vattr dirfor, diraft; 2410 nfsrvfh_t nsfh; 2411 struct nameidata nd; 2412 u_quad_t frev; 2413 2414 nfsm_srvmtofh(&nsfh); 2415 nfsm_srvnamesiz(len); 2416 nd.ni_cnd.cn_cred = cred; 2417 nd.ni_cnd.cn_nameiop = DELETE; 2418 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 2419 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 2420 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2421 if (dirp && v3) { 2422 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred); 2423 } 2424 if (error) { 2425 nfsm_reply(NFSX_WCCDATA(v3)); 2426 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2427 if (dirp) 2428 vrele(dirp); 2429 return (0); 2430 } 2431 vp = nd.ni_vp; 2432 if (vp->v_type != VDIR) { 2433 error = ENOTDIR; 2434 goto out; 2435 } 2436 /* 2437 * No rmdir "." please. 2438 */ 2439 if (nd.ni_dvp == vp) { 2440 error = EINVAL; 2441 goto out; 2442 } 2443 /* 2444 * The root of a mounted filesystem cannot be deleted. 2445 */ 2446 if (vp->v_vflag & VV_ROOT) 2447 error = EBUSY; 2448 out: 2449 if (!error) { 2450 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2451 nqsrv_getl(vp, ND_WRITE); 2452 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2453 } else { 2454 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2455 if (nd.ni_dvp == nd.ni_vp) 2456 vrele(nd.ni_dvp); 2457 else 2458 vput(nd.ni_dvp); 2459 vput(vp); 2460 } 2461 if (dirp) { 2462 if (v3) { 2463 diraft_ret = VOP_GETATTR(dirp, &diraft, cred); 2464 } 2465 vrele(dirp); 2466 } 2467 nfsm_reply(NFSX_WCCDATA(v3)); 2468 if (v3) { 2469 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2470 return (0); 2471 } 2472 nfsm_srvdone; 2473 } 2474 2475 /* 2476 * nfs readdir service 2477 * - mallocs what it thinks is enough to read 2478 * count rounded up to a multiple of NFS_SRVDIRBLKSIZ <= NFS_MAXREADDIR 2479 * - calls VOP_READDIR() 2480 * - loops around building the reply 2481 * if the output generated exceeds count break out of loop 2482 * The nfsm_clget macro is used here so that the reply will be packed 2483 * tightly in mbuf clusters. 2484 * - it only knows that it has encountered eof when the VOP_READDIR() 2485 * reads nothing 2486 * - as such one readdir rpc will return eof false although you are there 2487 * and then the next will return eof 2488 * - it trims out records with d_fileno == 0 2489 * this doesn't matter for Unix clients, but they might confuse clients 2490 * for other os'. 2491 * - it trims out records with d_type == DT_WHT 2492 * these cannot be seen through NFS (unless we extend the protocol) 2493 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less 2494 * than requested, but this may not apply to all filesystems. For 2495 * example, client NFS does not { although it is never remote mounted 2496 * anyhow } 2497 * The alternate call nfsrv_readdirplus() does lookups as well. 2498 * PS: The NFS protocol spec. does not clarify what the "count" byte 2499 * argument is a count of.. just name strings and file id's or the 2500 * entire reply rpc or ... 2501 * I tried just file name and id sizes and it confused the Sun client, 2502 * so I am using the full rpc size now. The "paranoia.." comment refers 2503 * to including the status longwords that are not a part of the dir. 2504 * "entry" structures, but are in the rpc. 2505 */ 2506 2507 #define NFS_SRVDIRBLKSIZ 1024 2508 2509 struct flrep { 2510 nfsuint64 fl_off; 2511 u_int32_t fl_postopok; 2512 u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)]; 2513 u_int32_t fl_fhok; 2514 u_int32_t fl_fhsize; 2515 }; 2516 2517 int 2518 nfsrv_readdir(nfsd, slp, lwp, mrq) 2519 struct nfsrv_descript *nfsd; 2520 struct nfssvc_sock *slp; 2521 struct lwp *lwp; 2522 struct mbuf **mrq; 2523 { 2524 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2525 struct mbuf *nam = nfsd->nd_nam; 2526 char *dpos = nfsd->nd_dpos; 2527 kauth_cred_t cred = nfsd->nd_cr; 2528 char *bp, *be; 2529 struct mbuf *mp; 2530 struct dirent *dp; 2531 char *cp; 2532 u_int32_t *tl; 2533 int32_t t1; 2534 char *bpos; 2535 struct mbuf *mb, *mreq, *mp2; 2536 char *cpos, *cend, *cp2, *rbuf; 2537 struct vnode *vp; 2538 struct vattr at; 2539 nfsrvfh_t nsfh; 2540 struct uio io; 2541 struct iovec iv; 2542 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2543 int siz, cnt, fullsiz, eofflag, rdonly, cache = 0, ncookies; 2544 int v3 = (nfsd->nd_flag & ND_NFSV3); 2545 u_quad_t frev, off, toff, verf; 2546 off_t *cookies = NULL, *cookiep; 2547 nfsuint64 jar; 2548 2549 nfsm_srvmtofh(&nsfh); 2550 if (v3) { 2551 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 2552 toff = fxdr_hyper(tl); 2553 tl += 2; 2554 verf = fxdr_hyper(tl); 2555 tl += 2; 2556 } else { 2557 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2558 toff = fxdr_unsigned(u_quad_t, *tl++); 2559 } 2560 off = toff; 2561 cnt = fxdr_unsigned(int, *tl); 2562 siz = ((cnt + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1)); 2563 xfer = NFS_SRVMAXDATA(nfsd); 2564 if (siz > xfer) 2565 siz = xfer; 2566 fullsiz = siz; 2567 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 2568 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 2569 if (!error && vp->v_type != VDIR) { 2570 error = ENOTDIR; 2571 vput(vp); 2572 } 2573 if (error) { 2574 nfsm_reply(NFSX_UNSIGNED); 2575 nfsm_srvpostop_attr(getret, &at); 2576 return (0); 2577 } 2578 nqsrv_getl(vp, ND_READ); 2579 if (v3) { 2580 error = getret = VOP_GETATTR(vp, &at, cred); 2581 #ifdef NFS3_STRICTVERF 2582 /* 2583 * XXX This check is too strict for Solaris 2.5 clients. 2584 */ 2585 if (!error && toff && verf != at.va_filerev) 2586 error = NFSERR_BAD_COOKIE; 2587 #endif 2588 } 2589 if (!error) 2590 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0); 2591 if (error) { 2592 vput(vp); 2593 nfsm_reply(NFSX_POSTOPATTR(v3)); 2594 nfsm_srvpostop_attr(getret, &at); 2595 return (0); 2596 } 2597 VOP_UNLOCK(vp, 0); 2598 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2599 again: 2600 iv.iov_base = rbuf; 2601 iv.iov_len = fullsiz; 2602 io.uio_iov = &iv; 2603 io.uio_iovcnt = 1; 2604 io.uio_offset = (off_t)off; 2605 io.uio_resid = fullsiz; 2606 io.uio_rw = UIO_READ; 2607 UIO_SETUP_SYSSPACE(&io); 2608 eofflag = 0; 2609 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2610 2611 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2612 2613 off = (off_t)io.uio_offset; 2614 if (!cookies && !error) 2615 error = NFSERR_PERM; 2616 if (v3) { 2617 getret = VOP_GETATTR(vp, &at, cred); 2618 if (!error) 2619 error = getret; 2620 } 2621 2622 VOP_UNLOCK(vp, 0); 2623 if (error) { 2624 vrele(vp); 2625 free((void *)rbuf, M_TEMP); 2626 if (cookies) 2627 free((void *)cookies, M_TEMP); 2628 nfsm_reply(NFSX_POSTOPATTR(v3)); 2629 nfsm_srvpostop_attr(getret, &at); 2630 return (0); 2631 } 2632 if (io.uio_resid) { 2633 siz -= io.uio_resid; 2634 2635 /* 2636 * If nothing read, return eof 2637 * rpc reply 2638 */ 2639 if (siz == 0) { 2640 vrele(vp); 2641 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + 2642 2 * NFSX_UNSIGNED); 2643 if (v3) { 2644 nfsm_srvpostop_attr(getret, &at); 2645 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2646 txdr_hyper(at.va_filerev, tl); 2647 tl += 2; 2648 } else 2649 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2650 *tl++ = nfs_false; 2651 *tl = nfs_true; 2652 free((void *)rbuf, M_TEMP); 2653 free((void *)cookies, M_TEMP); 2654 return (0); 2655 } 2656 } 2657 2658 /* 2659 * Check for degenerate cases of nothing useful read. 2660 * If so go try again 2661 */ 2662 cpos = rbuf; 2663 cend = rbuf + siz; 2664 dp = (struct dirent *)cpos; 2665 cookiep = cookies; 2666 2667 while (cpos < cend && ncookies > 0 && 2668 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2669 cpos += dp->d_reclen; 2670 dp = (struct dirent *)cpos; 2671 cookiep++; 2672 ncookies--; 2673 } 2674 if (cpos >= cend || ncookies == 0) { 2675 toff = off; 2676 siz = fullsiz; 2677 free(cookies, M_TEMP); 2678 cookies = NULL; 2679 goto again; 2680 } 2681 2682 len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */ 2683 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz); 2684 if (v3) { 2685 nfsm_srvpostop_attr(getret, &at); 2686 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2687 txdr_hyper(at.va_filerev, tl); 2688 } 2689 mp = mp2 = mb; 2690 bp = bpos; 2691 be = bp + M_TRAILINGSPACE(mp); 2692 2693 /* Loop through the records and build reply */ 2694 while (cpos < cend && ncookies > 0) { 2695 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 2696 nlen = dp->d_namlen; 2697 rem = nfsm_rndup(nlen)-nlen; 2698 len += (4 * NFSX_UNSIGNED + nlen + rem); 2699 if (v3) 2700 len += 2 * NFSX_UNSIGNED; 2701 if (len > cnt) { 2702 eofflag = 0; 2703 break; 2704 } 2705 /* 2706 * Build the directory record xdr from 2707 * the dirent entry. 2708 */ 2709 nfsm_clget; 2710 *tl = nfs_true; 2711 bp += NFSX_UNSIGNED; 2712 if (v3) { 2713 nfsm_clget; 2714 *tl = txdr_unsigned(dp->d_fileno >> 32); 2715 bp += NFSX_UNSIGNED; 2716 } 2717 nfsm_clget; 2718 *tl = txdr_unsigned(dp->d_fileno); 2719 bp += NFSX_UNSIGNED; 2720 nfsm_clget; 2721 *tl = txdr_unsigned(nlen); 2722 bp += NFSX_UNSIGNED; 2723 2724 /* And loop around copying the name */ 2725 xfer = nlen; 2726 cp = dp->d_name; 2727 while (xfer > 0) { 2728 nfsm_clget; 2729 if ((bp+xfer) > be) 2730 tsiz = be-bp; 2731 else 2732 tsiz = xfer; 2733 memcpy(bp, cp, tsiz); 2734 bp += tsiz; 2735 xfer -= tsiz; 2736 if (xfer > 0) 2737 cp += tsiz; 2738 } 2739 /* And null pad to an int32_t boundary */ 2740 for (i = 0; i < rem; i++) 2741 *bp++ = '\0'; 2742 nfsm_clget; 2743 2744 /* Finish off the record */ 2745 txdr_hyper(*cookiep, &jar); 2746 if (v3) { 2747 *tl = jar.nfsuquad[0]; 2748 bp += NFSX_UNSIGNED; 2749 nfsm_clget; 2750 } 2751 *tl = jar.nfsuquad[1]; 2752 bp += NFSX_UNSIGNED; 2753 } 2754 cpos += dp->d_reclen; 2755 dp = (struct dirent *)cpos; 2756 cookiep++; 2757 ncookies--; 2758 } 2759 vrele(vp); 2760 nfsm_clget; 2761 *tl = nfs_false; 2762 bp += NFSX_UNSIGNED; 2763 nfsm_clget; 2764 if (eofflag) 2765 *tl = nfs_true; 2766 else 2767 *tl = nfs_false; 2768 bp += NFSX_UNSIGNED; 2769 if (mp != mb) { 2770 if (bp < be) 2771 mp->m_len = bp - mtod(mp, char *); 2772 } else 2773 mp->m_len += bp - bpos; 2774 free((void *)rbuf, M_TEMP); 2775 free((void *)cookies, M_TEMP); 2776 nfsm_srvdone; 2777 } 2778 2779 int 2780 nfsrv_readdirplus(nfsd, slp, lwp, mrq) 2781 struct nfsrv_descript *nfsd; 2782 struct nfssvc_sock *slp; 2783 struct lwp *lwp; 2784 struct mbuf **mrq; 2785 { 2786 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2787 struct mbuf *nam = nfsd->nd_nam; 2788 char *dpos = nfsd->nd_dpos; 2789 kauth_cred_t cred = nfsd->nd_cr; 2790 char *bp, *be; 2791 struct mbuf *mp; 2792 struct dirent *dp; 2793 char *cp; 2794 u_int32_t *tl; 2795 int32_t t1; 2796 char *bpos; 2797 struct mbuf *mb, *mreq, *mp2; 2798 char *cpos, *cend, *cp2, *rbuf; 2799 struct vnode *vp, *nvp; 2800 struct flrep fl; 2801 nfsrvfh_t nsfh; 2802 struct uio io; 2803 struct iovec iv; 2804 struct vattr va, at, *vap = &va; 2805 struct nfs_fattr *fp; 2806 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2807 int siz, cnt, fullsiz, eofflag, rdonly, cache = 0, dirlen, ncookies; 2808 u_quad_t frev, off, toff, verf; 2809 off_t *cookies = NULL, *cookiep; 2810 2811 nfsm_srvmtofh(&nsfh); 2812 nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED); 2813 toff = fxdr_hyper(tl); 2814 tl += 2; 2815 verf = fxdr_hyper(tl); 2816 tl += 2; 2817 siz = fxdr_unsigned(int, *tl++); 2818 cnt = fxdr_unsigned(int, *tl); 2819 off = toff; 2820 siz = ((siz + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1)); 2821 xfer = NFS_SRVMAXDATA(nfsd); 2822 if (siz > xfer) 2823 siz = xfer; 2824 fullsiz = siz; 2825 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 2826 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 2827 if (!error && vp->v_type != VDIR) { 2828 error = ENOTDIR; 2829 vput(vp); 2830 } 2831 if (error) { 2832 nfsm_reply(NFSX_UNSIGNED); 2833 nfsm_srvpostop_attr(getret, &at); 2834 return (0); 2835 } 2836 error = getret = VOP_GETATTR(vp, &at, cred); 2837 #ifdef NFS3_STRICTVERF 2838 /* 2839 * XXX This check is too strict for Solaris 2.5 clients. 2840 */ 2841 if (!error && toff && verf != at.va_filerev) 2842 error = NFSERR_BAD_COOKIE; 2843 #endif 2844 if (!error) { 2845 nqsrv_getl(vp, ND_READ); 2846 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0); 2847 } 2848 if (error) { 2849 vput(vp); 2850 nfsm_reply(NFSX_V3POSTOPATTR); 2851 nfsm_srvpostop_attr(getret, &at); 2852 return (0); 2853 } 2854 VOP_UNLOCK(vp, 0); 2855 2856 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2857 again: 2858 iv.iov_base = rbuf; 2859 iv.iov_len = fullsiz; 2860 io.uio_iov = &iv; 2861 io.uio_iovcnt = 1; 2862 io.uio_offset = (off_t)off; 2863 io.uio_resid = fullsiz; 2864 io.uio_rw = UIO_READ; 2865 UIO_SETUP_SYSSPACE(&io); 2866 eofflag = 0; 2867 2868 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2869 2870 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2871 2872 off = (u_quad_t)io.uio_offset; 2873 getret = VOP_GETATTR(vp, &at, cred); 2874 2875 VOP_UNLOCK(vp, 0); 2876 2877 /* 2878 * If the VGET operation doesn't work for this filesystem, 2879 * we can't support readdirplus. Returning NOTSUPP should 2880 * make clients fall back to plain readdir. 2881 * There's no need to check for VPTOFH as well, we wouldn't 2882 * even be here otherwise. 2883 */ 2884 if (!getret) { 2885 if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp))) 2886 getret = (getret == EOPNOTSUPP) ? 2887 NFSERR_NOTSUPP : NFSERR_IO; 2888 else 2889 vput(nvp); 2890 } 2891 2892 if (!cookies && !error) 2893 error = NFSERR_PERM; 2894 if (!error) 2895 error = getret; 2896 if (error) { 2897 vrele(vp); 2898 if (cookies) 2899 free((void *)cookies, M_TEMP); 2900 free((void *)rbuf, M_TEMP); 2901 nfsm_reply(NFSX_V3POSTOPATTR); 2902 nfsm_srvpostop_attr(getret, &at); 2903 return (0); 2904 } 2905 if (io.uio_resid) { 2906 siz -= io.uio_resid; 2907 2908 /* 2909 * If nothing read, return eof 2910 * rpc reply 2911 */ 2912 if (siz == 0) { 2913 vrele(vp); 2914 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2915 2 * NFSX_UNSIGNED); 2916 nfsm_srvpostop_attr(getret, &at); 2917 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2918 txdr_hyper(at.va_filerev, tl); 2919 tl += 2; 2920 *tl++ = nfs_false; 2921 *tl = nfs_true; 2922 free((void *)cookies, M_TEMP); 2923 free((void *)rbuf, M_TEMP); 2924 return (0); 2925 } 2926 } 2927 2928 /* 2929 * Check for degenerate cases of nothing useful read. 2930 * If so go try again 2931 */ 2932 cpos = rbuf; 2933 cend = rbuf + siz; 2934 dp = (struct dirent *)cpos; 2935 cookiep = cookies; 2936 2937 while (cpos < cend && ncookies > 0 && 2938 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2939 cpos += dp->d_reclen; 2940 dp = (struct dirent *)cpos; 2941 cookiep++; 2942 ncookies--; 2943 } 2944 if (cpos >= cend || ncookies == 0) { 2945 toff = off; 2946 siz = fullsiz; 2947 free(cookies, M_TEMP); 2948 cookies = NULL; 2949 goto again; 2950 } 2951 2952 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED; 2953 nfsm_reply(cnt); 2954 nfsm_srvpostop_attr(getret, &at); 2955 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2956 txdr_hyper(at.va_filerev, tl); 2957 mp = mp2 = mb; 2958 bp = bpos; 2959 be = bp + M_TRAILINGSPACE(mp); 2960 2961 /* Loop through the records and build reply */ 2962 while (cpos < cend && ncookies > 0) { 2963 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 2964 nfsrvfh_t nnsfh; 2965 2966 nlen = dp->d_namlen; 2967 rem = nfsm_rndup(nlen)-nlen; 2968 2969 /* 2970 * For readdir_and_lookup get the vnode using 2971 * the file number. 2972 */ 2973 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp)) 2974 goto invalid; 2975 if (nfsrv_composefh(nvp, &nnsfh, true)) { 2976 vput(nvp); 2977 goto invalid; 2978 } 2979 if (VOP_GETATTR(nvp, vap, cred)) { 2980 vput(nvp); 2981 goto invalid; 2982 } 2983 vput(nvp); 2984 2985 /* 2986 * If either the dircount or maxcount will be 2987 * exceeded, get out now. Both of these lengths 2988 * are calculated conservatively, including all 2989 * XDR overheads. 2990 */ 2991 len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH + 2992 NFSX_V3POSTOPATTR); 2993 dirlen += (6 * NFSX_UNSIGNED + nlen + rem); 2994 if (len > cnt || dirlen > fullsiz) { 2995 eofflag = 0; 2996 break; 2997 } 2998 2999 /* 3000 * Build the directory record xdr from 3001 * the dirent entry. 3002 */ 3003 fp = (struct nfs_fattr *)&fl.fl_fattr; 3004 nfsm_srvfillattr(vap, fp); 3005 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH); 3006 fl.fl_fhok = nfs_true; 3007 fl.fl_postopok = nfs_true; 3008 txdr_hyper(*cookiep, fl.fl_off.nfsuquad); 3009 3010 nfsm_clget; 3011 *tl = nfs_true; 3012 bp += NFSX_UNSIGNED; 3013 nfsm_clget; 3014 *tl = txdr_unsigned(dp->d_fileno >> 32); 3015 bp += NFSX_UNSIGNED; 3016 nfsm_clget; 3017 *tl = txdr_unsigned(dp->d_fileno); 3018 bp += NFSX_UNSIGNED; 3019 nfsm_clget; 3020 *tl = txdr_unsigned(nlen); 3021 bp += NFSX_UNSIGNED; 3022 3023 /* And loop around copying the name */ 3024 xfer = nlen; 3025 cp = dp->d_name; 3026 while (xfer > 0) { 3027 nfsm_clget; 3028 if ((bp + xfer) > be) 3029 tsiz = be - bp; 3030 else 3031 tsiz = xfer; 3032 memcpy(bp, cp, tsiz); 3033 bp += tsiz; 3034 xfer -= tsiz; 3035 if (xfer > 0) 3036 cp += tsiz; 3037 } 3038 /* And null pad to an int32_t boundary */ 3039 for (i = 0; i < rem; i++) 3040 *bp++ = '\0'; 3041 3042 /* 3043 * Now copy the flrep structure out. 3044 */ 3045 xfer = sizeof(struct flrep); 3046 cp = (void *)&fl; 3047 while (xfer > 0) { 3048 nfsm_clget; 3049 if ((bp + xfer) > be) 3050 tsiz = be - bp; 3051 else 3052 tsiz = xfer; 3053 memcpy(bp, cp, tsiz); 3054 bp += tsiz; 3055 xfer -= tsiz; 3056 if (xfer > 0) 3057 cp += tsiz; 3058 } 3059 3060 /* 3061 * ... and filehandle. 3062 */ 3063 xfer = NFSRVFH_SIZE(&nnsfh); 3064 cp = NFSRVFH_DATA(&nnsfh); 3065 while (xfer > 0) { 3066 nfsm_clget; 3067 if ((bp + xfer) > be) 3068 tsiz = be - bp; 3069 else 3070 tsiz = xfer; 3071 memcpy(bp, cp, tsiz); 3072 bp += tsiz; 3073 xfer -= tsiz; 3074 if (xfer > 0) 3075 cp += tsiz; 3076 } 3077 } 3078 invalid: 3079 cpos += dp->d_reclen; 3080 dp = (struct dirent *)cpos; 3081 cookiep++; 3082 ncookies--; 3083 } 3084 vrele(vp); 3085 nfsm_clget; 3086 *tl = nfs_false; 3087 bp += NFSX_UNSIGNED; 3088 nfsm_clget; 3089 if (eofflag) 3090 *tl = nfs_true; 3091 else 3092 *tl = nfs_false; 3093 bp += NFSX_UNSIGNED; 3094 if (mp != mb) { 3095 if (bp < be) 3096 mp->m_len = bp - mtod(mp, char *); 3097 } else 3098 mp->m_len += bp - bpos; 3099 free((void *)cookies, M_TEMP); 3100 free((void *)rbuf, M_TEMP); 3101 nfsm_srvdone; 3102 } 3103 3104 /* 3105 * nfs commit service 3106 */ 3107 int 3108 nfsrv_commit(nfsd, slp, lwp, mrq) 3109 struct nfsrv_descript *nfsd; 3110 struct nfssvc_sock *slp; 3111 struct lwp *lwp; 3112 struct mbuf **mrq; 3113 { 3114 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3115 struct mbuf *nam = nfsd->nd_nam; 3116 char *dpos = nfsd->nd_dpos; 3117 kauth_cred_t cred = nfsd->nd_cr; 3118 struct vattr bfor, aft; 3119 struct vnode *vp; 3120 nfsrvfh_t nsfh; 3121 u_int32_t *tl; 3122 int32_t t1; 3123 char *bpos; 3124 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cache = 0; 3125 uint32_t cnt; 3126 char *cp2; 3127 struct mbuf *mb, *mreq; 3128 u_quad_t frev, off, end; 3129 3130 nfsm_srvmtofh(&nsfh); 3131 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED); 3132 3133 off = fxdr_hyper(tl); 3134 tl += 2; 3135 cnt = fxdr_unsigned(uint32_t, *tl); 3136 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3137 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3138 if (error) { 3139 nfsm_reply(2 * NFSX_UNSIGNED); 3140 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3141 return (0); 3142 } 3143 for_ret = VOP_GETATTR(vp, &bfor, cred); 3144 end = (cnt > 0) ? off + cnt : vp->v_size; 3145 if (end < off || end > vp->v_size) 3146 end = vp->v_size; 3147 if (off < vp->v_size) 3148 error = VOP_FSYNC(vp, cred, FSYNC_WAIT, off, end); 3149 /* else error == 0, from nfsrv_fhtovp() */ 3150 aft_ret = VOP_GETATTR(vp, &aft, cred); 3151 vput(vp); 3152 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF); 3153 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3154 if (!error) { 3155 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF); 3156 *tl++ = txdr_unsigned(boottime.tv_sec); 3157 *tl = txdr_unsigned(boottime.tv_usec); 3158 } else { 3159 return (0); 3160 } 3161 nfsm_srvdone; 3162 } 3163 3164 /* 3165 * nfs statfs service 3166 */ 3167 int 3168 nfsrv_statfs(nfsd, slp, lwp, mrq) 3169 struct nfsrv_descript *nfsd; 3170 struct nfssvc_sock *slp; 3171 struct lwp *lwp; 3172 struct mbuf **mrq; 3173 { 3174 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3175 struct mbuf *nam = nfsd->nd_nam; 3176 char *dpos = nfsd->nd_dpos; 3177 kauth_cred_t cred = nfsd->nd_cr; 3178 struct statvfs *sf = NULL; 3179 struct nfs_statfs *sfp; 3180 u_int32_t *tl; 3181 int32_t t1; 3182 char *bpos; 3183 int error = 0, rdonly, cache = 0, getret = 1; 3184 int v3 = (nfsd->nd_flag & ND_NFSV3); 3185 char *cp2; 3186 struct mbuf *mb, *mreq; 3187 struct vnode *vp; 3188 struct vattr at; 3189 nfsrvfh_t nsfh; 3190 u_quad_t frev, tval; 3191 3192 nfsm_srvmtofh(&nsfh); 3193 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3194 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3195 if (error) { 3196 nfsm_reply(NFSX_UNSIGNED); 3197 nfsm_srvpostop_attr(getret, &at); 3198 return (0); 3199 } 3200 sf = malloc(sizeof(*sf), M_TEMP, M_WAITOK); 3201 error = VFS_STATVFS(vp->v_mount, sf); 3202 getret = VOP_GETATTR(vp, &at, cred); 3203 vput(vp); 3204 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3)); 3205 if (v3) 3206 nfsm_srvpostop_attr(getret, &at); 3207 if (error) { 3208 free(sf, M_TEMP); 3209 return (0); 3210 } 3211 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3)); 3212 if (v3) { 3213 tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_frsize); 3214 txdr_hyper(tval, &sfp->sf_tbytes); 3215 tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_frsize); 3216 txdr_hyper(tval, &sfp->sf_fbytes); 3217 tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_frsize); 3218 txdr_hyper(tval, &sfp->sf_abytes); 3219 tval = (u_quad_t)sf->f_files; 3220 txdr_hyper(tval, &sfp->sf_tfiles); 3221 tval = (u_quad_t)sf->f_ffree; 3222 txdr_hyper(tval, &sfp->sf_ffiles); 3223 txdr_hyper(tval, &sfp->sf_afiles); 3224 sfp->sf_invarsec = 0; 3225 } else { 3226 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA); 3227 sfp->sf_bsize = txdr_unsigned(sf->f_frsize); 3228 sfp->sf_blocks = txdr_unsigned(sf->f_blocks); 3229 sfp->sf_bfree = txdr_unsigned(sf->f_bfree); 3230 sfp->sf_bavail = txdr_unsigned(sf->f_bavail); 3231 } 3232 nfsmout: 3233 if (sf) 3234 free(sf, M_TEMP); 3235 return error; 3236 } 3237 3238 /* 3239 * nfs fsinfo service 3240 */ 3241 int 3242 nfsrv_fsinfo(nfsd, slp, lwp, mrq) 3243 struct nfsrv_descript *nfsd; 3244 struct nfssvc_sock *slp; 3245 struct lwp *lwp; 3246 struct mbuf **mrq; 3247 { 3248 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3249 struct mbuf *nam = nfsd->nd_nam; 3250 char *dpos = nfsd->nd_dpos; 3251 kauth_cred_t cred = nfsd->nd_cr; 3252 u_int32_t *tl; 3253 struct nfsv3_fsinfo *sip; 3254 int32_t t1; 3255 char *bpos; 3256 int error = 0, rdonly, cache = 0, getret = 1; 3257 uint32_t maxdata; 3258 char *cp2; 3259 struct mbuf *mb, *mreq; 3260 struct vnode *vp; 3261 struct vattr at; 3262 nfsrvfh_t nsfh; 3263 u_quad_t frev, maxfsize; 3264 struct statvfs *sb; 3265 3266 nfsm_srvmtofh(&nsfh); 3267 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3268 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3269 if (error) { 3270 nfsm_reply(NFSX_UNSIGNED); 3271 nfsm_srvpostop_attr(getret, &at); 3272 return (0); 3273 } 3274 3275 /* XXX Try to make a guess on the max file size. */ 3276 sb = malloc(sizeof(*sb), M_TEMP, M_WAITOK); 3277 VFS_STATVFS(vp->v_mount, sb); 3278 maxfsize = (u_quad_t)0x80000000 * sb->f_frsize - 1; 3279 free(sb, M_TEMP); 3280 3281 getret = VOP_GETATTR(vp, &at, cred); 3282 vput(vp); 3283 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO); 3284 nfsm_srvpostop_attr(getret, &at); 3285 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO); 3286 3287 /* 3288 * XXX 3289 * There should be file system VFS OP(s) to get this information. 3290 * For now, assume ufs. 3291 */ 3292 if (slp->ns_so->so_type == SOCK_DGRAM) 3293 maxdata = NFS_MAXDGRAMDATA; 3294 else 3295 maxdata = NFS_MAXDATA; 3296 sip->fs_rtmax = txdr_unsigned(maxdata); 3297 sip->fs_rtpref = txdr_unsigned(maxdata); 3298 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE); 3299 sip->fs_wtmax = txdr_unsigned(maxdata); 3300 sip->fs_wtpref = txdr_unsigned(maxdata); 3301 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE); 3302 sip->fs_dtpref = txdr_unsigned(maxdata); 3303 txdr_hyper(maxfsize, &sip->fs_maxfilesize); 3304 sip->fs_timedelta.nfsv3_sec = 0; 3305 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1); 3306 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK | 3307 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS | 3308 NFSV3FSINFO_CANSETTIME); 3309 nfsm_srvdone; 3310 } 3311 3312 /* 3313 * nfs pathconf service 3314 */ 3315 int 3316 nfsrv_pathconf(nfsd, slp, lwp, mrq) 3317 struct nfsrv_descript *nfsd; 3318 struct nfssvc_sock *slp; 3319 struct lwp *lwp; 3320 struct mbuf **mrq; 3321 { 3322 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3323 struct mbuf *nam = nfsd->nd_nam; 3324 char *dpos = nfsd->nd_dpos; 3325 kauth_cred_t cred = nfsd->nd_cr; 3326 u_int32_t *tl; 3327 struct nfsv3_pathconf *pc; 3328 int32_t t1; 3329 char *bpos; 3330 int error = 0, rdonly, cache = 0, getret = 1; 3331 register_t linkmax, namemax, chownres, notrunc; 3332 char *cp2; 3333 struct mbuf *mb, *mreq; 3334 struct vnode *vp; 3335 struct vattr at; 3336 nfsrvfh_t nsfh; 3337 u_quad_t frev; 3338 3339 nfsm_srvmtofh(&nsfh); 3340 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3341 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3342 if (error) { 3343 nfsm_reply(NFSX_UNSIGNED); 3344 nfsm_srvpostop_attr(getret, &at); 3345 return (0); 3346 } 3347 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax); 3348 if (!error) 3349 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax); 3350 if (!error) 3351 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres); 3352 if (!error) 3353 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc); 3354 getret = VOP_GETATTR(vp, &at, cred); 3355 vput(vp); 3356 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF); 3357 nfsm_srvpostop_attr(getret, &at); 3358 if (error) 3359 return (0); 3360 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF); 3361 3362 pc->pc_linkmax = txdr_unsigned(linkmax); 3363 pc->pc_namemax = txdr_unsigned(namemax); 3364 pc->pc_notrunc = txdr_unsigned(notrunc); 3365 pc->pc_chownrestricted = txdr_unsigned(chownres); 3366 3367 /* 3368 * These should probably be supported by VOP_PATHCONF(), but 3369 * until msdosfs is exportable (why would you want to?), the 3370 * Unix defaults should be ok. 3371 */ 3372 pc->pc_caseinsensitive = nfs_false; 3373 pc->pc_casepreserving = nfs_true; 3374 nfsm_srvdone; 3375 } 3376 3377 /* 3378 * Null operation, used by clients to ping server 3379 */ 3380 /* ARGSUSED */ 3381 int 3382 nfsrv_null(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, 3383 struct lwp *lwp, struct mbuf **mrq) 3384 { 3385 struct mbuf *mrep = nfsd->nd_mrep; 3386 char *bpos; 3387 int error = NFSERR_RETVOID, cache = 0; 3388 struct mbuf *mb, *mreq; 3389 u_quad_t frev; 3390 3391 nfsm_reply(0); 3392 return (0); 3393 } 3394 3395 /* 3396 * No operation, used for obsolete procedures 3397 */ 3398 /* ARGSUSED */ 3399 int 3400 nfsrv_noop(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, 3401 struct lwp *lwp, struct mbuf **mrq) 3402 { 3403 struct mbuf *mrep = nfsd->nd_mrep; 3404 char *bpos; 3405 int error, cache = 0; 3406 struct mbuf *mb, *mreq; 3407 u_quad_t frev; 3408 3409 if (nfsd->nd_repstat) 3410 error = nfsd->nd_repstat; 3411 else 3412 error = EPROCUNAVAIL; 3413 nfsm_reply(0); 3414 return (0); 3415 } 3416 3417 /* 3418 * Perform access checking for vnodes obtained from file handles that would 3419 * refer to files already opened by a Unix client. You cannot just use 3420 * vn_writechk() and VOP_ACCESS() for two reasons. 3421 * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case 3422 * 2 - The owner is to be given access irrespective of mode bits for some 3423 * operations, so that processes that chmod after opening a file don't 3424 * break. I don't like this because it opens a security hole, but since 3425 * the nfs server opens a security hole the size of a barn door anyhow, 3426 * what the heck. 3427 * 3428 * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS() 3429 * will return EPERM instead of EACCESS. EPERM is always an error. 3430 */ 3431 int 3432 nfsrv_access(vp, flags, cred, rdonly, lwp, override) 3433 struct vnode *vp; 3434 int flags; 3435 kauth_cred_t cred; 3436 int rdonly; 3437 struct lwp *lwp; 3438 int override; 3439 { 3440 struct vattr vattr; 3441 int error; 3442 if (flags & VWRITE) { 3443 /* Just vn_writechk() changed to check rdonly */ 3444 /* 3445 * Disallow write attempts on read-only file systems; 3446 * unless the file is a socket or a block or character 3447 * device resident on the file system. 3448 */ 3449 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 3450 switch (vp->v_type) { 3451 case VREG: 3452 case VDIR: 3453 case VLNK: 3454 return (EROFS); 3455 default: 3456 break; 3457 } 3458 } 3459 3460 /* 3461 * If the vnode is in use as a process's text, 3462 * we can't allow writing. 3463 */ 3464 if (vp->v_iflag & VI_TEXT) 3465 return (ETXTBSY); 3466 } 3467 error = VOP_GETATTR(vp, &vattr, cred); 3468 if (error) 3469 return (error); 3470 error = VOP_ACCESS(vp, flags, cred); 3471 /* 3472 * Allow certain operations for the owner (reads and writes 3473 * on files that are already open). 3474 */ 3475 if (override && error == EACCES && kauth_cred_geteuid(cred) == vattr.va_uid) 3476 error = 0; 3477 return error; 3478 } 3479
Indexes created Thu Oct 02 14:10:14 GMT 2025