1 /* $NetBSD: nfs_serv.c,v 1.134 2008/01/28 14:31:19 dholland 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 1.134 2008/01/28 14:31:19 dholland Exp $"); 59 60 #include <sys/param.h> 61 #include <sys/systm.h> 62 #include <sys/proc.h> 63 #include <sys/file.h> 64 #include <sys/namei.h> 65 #include <sys/vnode.h> 66 #include <sys/mount.h> 67 #include <sys/socket.h> 68 #include <sys/socketvar.h> 69 #include <sys/mbuf.h> 70 #include <sys/dirent.h> 71 #include <sys/stat.h> 72 #include <sys/kernel.h> 73 #include <sys/hash.h> 74 #include <sys/kauth.h> 75 76 #include <uvm/uvm.h> 77 78 #include <nfs/nfsproto.h> 79 #include <nfs/rpcv2.h> 80 #include <nfs/nfs.h> 81 #include <nfs/xdr_subs.h> 82 #include <nfs/nfsm_subs.h> 83 #include <nfs/nfs_var.h> 84 85 /* Global vars */ 86 extern u_int32_t nfs_xdrneg1; 87 extern u_int32_t nfs_false, nfs_true; 88 extern enum vtype nv3tov_type[8]; 89 extern struct nfsstats nfsstats; 90 extern const nfstype nfsv2_type[9]; 91 extern const nfstype nfsv3_type[9]; 92 int nfsrvw_procrastinate = NFS_GATHERDELAY * 1000; 93 int nfsd_use_loan = 1; /* use page-loan for READ OP */ 94 95 #define nqsrv_getl(vp, rw) /* nothing */ 96 97 /* 98 * nfs v3 access service 99 */ 100 int 101 nfsrv3_access(nfsd, slp, lwp, mrq) 102 struct nfsrv_descript *nfsd; 103 struct nfssvc_sock *slp; 104 struct lwp *lwp; 105 struct mbuf **mrq; 106 { 107 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 108 struct mbuf *nam = nfsd->nd_nam; 109 char *dpos = nfsd->nd_dpos; 110 kauth_cred_t cred = nfsd->nd_cr; 111 struct vnode *vp; 112 nfsrvfh_t nsfh; 113 u_int32_t *tl; 114 int32_t t1; 115 char *bpos; 116 int error = 0, rdonly, cache = 0, getret; 117 char *cp2; 118 struct mbuf *mb, *mreq; 119 struct vattr va; 120 u_long inmode, testmode, outmode; 121 u_quad_t frev; 122 123 nfsm_srvmtofh(&nsfh); 124 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 125 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, &rdonly, 126 (nfsd->nd_flag & ND_KERBAUTH), false); 127 if (error) { 128 nfsm_reply(NFSX_UNSIGNED); 129 nfsm_srvpostop_attr(1, (struct vattr *)0); 130 return (0); 131 } 132 inmode = fxdr_unsigned(u_int32_t, *tl); 133 outmode = 0; 134 if ((inmode & NFSV3ACCESS_READ) && 135 nfsrv_access(vp, VREAD, cred, rdonly, lwp, 0) == 0) 136 outmode |= NFSV3ACCESS_READ; 137 if (vp->v_type != VDIR) { 138 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND); 139 if (testmode && 140 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0) 141 outmode |= testmode; 142 if ((inmode & NFSV3ACCESS_EXECUTE) && 143 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0) 144 outmode |= NFSV3ACCESS_EXECUTE; 145 } else { 146 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND | 147 NFSV3ACCESS_DELETE); 148 if (testmode && 149 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0) 150 outmode |= testmode; 151 if ((inmode & NFSV3ACCESS_LOOKUP) && 152 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0) 153 outmode |= NFSV3ACCESS_LOOKUP; 154 } 155 getret = VOP_GETATTR(vp, &va, cred); 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 /* map pages */ 685 for (i = 0; i < npages; i++) { 686 pmap_kenter_pa(lva, VM_PAGE_TO_PHYS(pgpp[i]), 687 VM_PROT_READ); 688 lva += PAGE_SIZE; 689 } 690 691 pmap_update(pmap_kernel()); 692 693 mb->m_next = m; 694 mb = m; 695 error = 0; 696 uiop->uio_resid = 0; 697 } else { 698 struct iovec *iv; 699 struct iovec *iv2; 700 struct mbuf *m2; 701 int siz; 702 loan_fail: 703 /* 704 * Generate the mbuf list with the uio_iov ref. to it. 705 */ 706 i = 0; 707 m = m2 = mb; 708 while (left > 0) { 709 siz = min(M_TRAILINGSPACE(m), left); 710 if (siz > 0) { 711 left -= siz; 712 i++; 713 } 714 if (left > 0) { 715 m = m_get(M_WAIT, MT_DATA); 716 MCLAIM(m, &nfs_mowner); 717 m_clget(m, M_WAIT); 718 m->m_len = 0; 719 m2->m_next = m; 720 m2 = m; 721 } 722 } 723 iv = malloc(i * sizeof(struct iovec), M_TEMP, M_WAITOK); 724 uiop->uio_iov = iv2 = iv; 725 m = mb; 726 left = cnt; 727 i = 0; 728 while (left > 0) { 729 if (m == NULL) 730 panic("nfsrv_read iov"); 731 siz = min(M_TRAILINGSPACE(m), left); 732 if (siz > 0) { 733 iv->iov_base = mtod(m, char *) + 734 m->m_len; 735 iv->iov_len = siz; 736 m->m_len += siz; 737 left -= siz; 738 iv++; 739 i++; 740 } 741 m = m->m_next; 742 } 743 uiop->uio_iovcnt = i; 744 uiop->uio_offset = off; 745 uiop->uio_resid = cnt; 746 uiop->uio_rw = UIO_READ; 747 UIO_SETUP_SYSSPACE(uiop); 748 error = VOP_READ(vp, uiop, IO_NODELOCKED, cred); 749 free((void *)iv2, M_TEMP); 750 } 751 read_error: 752 if (error || (getret = VOP_GETATTR(vp, &va, cred)) != 0){ 753 if (!error) 754 error = getret; 755 m_freem(mreq); 756 vput(vp); 757 nfsm_reply(NFSX_POSTOPATTR(v3)); 758 nfsm_srvpostop_attr(getret, &va); 759 return (0); 760 } 761 } else { 762 uiop->uio_resid = 0; 763 } 764 vput(vp); 765 nfsm_srvfillattr(&va, fp); 766 len -= uiop->uio_resid; 767 padlen = nfsm_padlen(len); 768 if (uiop->uio_resid || padlen) 769 nfs_zeropad(mb, uiop->uio_resid, padlen); 770 if (v3) { 771 /* count */ 772 *tl++ = txdr_unsigned(len); 773 /* eof */ 774 if (off + len >= va.va_size) 775 *tl++ = nfs_true; 776 else 777 *tl++ = nfs_false; 778 } 779 *tl = txdr_unsigned(len); 780 nfsm_srvdone; 781 } 782 783 /* 784 * nfs write service 785 */ 786 int 787 nfsrv_write(nfsd, slp, lwp, mrq) 788 struct nfsrv_descript *nfsd; 789 struct nfssvc_sock *slp; 790 struct lwp *lwp; 791 struct mbuf **mrq; 792 { 793 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 794 struct mbuf *nam = nfsd->nd_nam; 795 char *dpos = nfsd->nd_dpos; 796 kauth_cred_t cred = nfsd->nd_cr; 797 struct iovec *ivp; 798 int i, cnt; 799 struct mbuf *mp; 800 struct nfs_fattr *fp; 801 struct iovec *iv; 802 struct vattr va, forat; 803 u_int32_t *tl; 804 int32_t t1; 805 char *bpos; 806 int error = 0, rdonly, cache = 0, len, forat_ret = 1; 807 int ioflags, aftat_ret = 1, retlen, zeroing, adjust; 808 int stable = NFSV3WRITE_FILESYNC; 809 int v3 = (nfsd->nd_flag & ND_NFSV3); 810 char *cp2; 811 struct mbuf *mb, *mreq; 812 struct vnode *vp; 813 nfsrvfh_t nsfh; 814 struct uio io, *uiop = &io; 815 off_t off; 816 u_quad_t frev; 817 818 if (mrep == NULL) { 819 *mrq = NULL; 820 return (0); 821 } 822 nfsm_srvmtofh(&nsfh); 823 if (v3) { 824 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 825 off = fxdr_hyper(tl); 826 tl += 3; 827 stable = fxdr_unsigned(int, *tl++); 828 } else { 829 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 830 off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 831 tl += 2; 832 } 833 retlen = len = fxdr_unsigned(int32_t, *tl); 834 cnt = i = 0; 835 836 /* 837 * For NFS Version 2, it is not obvious what a write of zero length 838 * should do, but I might as well be consistent with Version 3, 839 * which is to return ok so long as there are no permission problems. 840 */ 841 if (len > 0) { 842 zeroing = 1; 843 mp = mrep; 844 while (mp) { 845 if (mp == md) { 846 zeroing = 0; 847 adjust = dpos - mtod(mp, char *); 848 mp->m_len -= adjust; 849 if (mp->m_len > 0 && adjust > 0) 850 NFSMADV(mp, adjust); 851 } 852 if (zeroing) 853 mp->m_len = 0; 854 else if (mp->m_len > 0) { 855 i += mp->m_len; 856 if (i > len) { 857 mp->m_len -= (i - len); 858 zeroing = 1; 859 } 860 if (mp->m_len > 0) 861 cnt++; 862 } 863 mp = mp->m_next; 864 } 865 } 866 if (len > NFS_MAXDATA || len < 0 || i < len) { 867 error = EIO; 868 nfsm_reply(2 * NFSX_UNSIGNED); 869 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 870 return (0); 871 } 872 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 873 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 874 if (error) { 875 nfsm_reply(2 * NFSX_UNSIGNED); 876 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 877 return (0); 878 } 879 if (v3) 880 forat_ret = VOP_GETATTR(vp, &forat, cred); 881 if (vp->v_type != VREG) { 882 if (v3) 883 error = EINVAL; 884 else 885 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 886 } 887 if (!error) { 888 nqsrv_getl(vp, ND_WRITE); 889 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1); 890 } 891 if (error) { 892 vput(vp); 893 nfsm_reply(NFSX_WCCDATA(v3)); 894 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 895 return (0); 896 } 897 898 if (len > 0) { 899 ivp = malloc(cnt * sizeof (struct iovec), M_TEMP, M_WAITOK); 900 uiop->uio_iov = iv = ivp; 901 uiop->uio_iovcnt = cnt; 902 mp = mrep; 903 while (mp) { 904 if (mp->m_len > 0) { 905 ivp->iov_base = mtod(mp, void *); 906 ivp->iov_len = mp->m_len; 907 ivp++; 908 } 909 mp = mp->m_next; 910 } 911 912 /* 913 * XXX 914 * The IO_METASYNC flag indicates that all metadata (and not 915 * just enough to ensure data integrity) must be written to 916 * stable storage synchronously. 917 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.) 918 */ 919 if (stable == NFSV3WRITE_UNSTABLE) 920 ioflags = IO_NODELOCKED; 921 else if (stable == NFSV3WRITE_DATASYNC) 922 ioflags = (IO_SYNC | IO_NODELOCKED); 923 else 924 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 925 uiop->uio_resid = len; 926 uiop->uio_rw = UIO_WRITE; 927 uiop->uio_offset = off; 928 UIO_SETUP_SYSSPACE(uiop); 929 error = VOP_WRITE(vp, uiop, ioflags, cred); 930 nfsstats.srvvop_writes++; 931 free(iv, M_TEMP); 932 } 933 aftat_ret = VOP_GETATTR(vp, &va, cred); 934 vput(vp); 935 if (!error) 936 error = aftat_ret; 937 nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) + 938 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3)); 939 if (v3) { 940 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 941 if (error) 942 return (0); 943 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 944 *tl++ = txdr_unsigned(retlen); 945 if (stable == NFSV3WRITE_UNSTABLE) 946 *tl++ = txdr_unsigned(stable); 947 else 948 *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC); 949 /* 950 * Actually, there is no need to txdr these fields, 951 * but it may make the values more human readable, 952 * for debugging purposes. 953 */ 954 *tl++ = txdr_unsigned(boottime.tv_sec); 955 *tl = txdr_unsigned(boottime.tv_usec); 956 } else { 957 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 958 nfsm_srvfillattr(&va, fp); 959 } 960 nfsm_srvdone; 961 } 962 963 /* 964 * XXX elad: the original NFSW_SAMECRED() macro also made sure the 965 * two nd_flag fields of the descriptors contained 966 * ND_KERBAUTH. 967 */ 968 static int 969 nfsrv_samecred(kauth_cred_t cred1, kauth_cred_t cred2) 970 { 971 int i, do_ngroups; 972 973 if (kauth_cred_geteuid(cred1) != kauth_cred_geteuid(cred2)) 974 return (0); 975 if (kauth_cred_ngroups(cred1) != kauth_cred_ngroups(cred2)) 976 return (0); 977 do_ngroups = kauth_cred_ngroups(cred1); 978 for (i = 0; i < do_ngroups; i++) 979 if (kauth_cred_group(cred1, i) != 980 kauth_cred_group(cred2, i)) 981 return (0); 982 983 return (1); 984 } 985 986 static struct nfsrvw_delayhash * 987 nfsrv_nwdelayhash(struct nfssvc_sock *slp, const nfsrvfh_t *nsfh) 988 { 989 uint32_t hash; 990 991 hash = hash32_buf(NFSRVFH_DATA(nsfh), NFSRVFH_SIZE(nsfh), 992 HASH32_BUF_INIT); 993 return &slp->ns_wdelayhashtbl[hash % NFS_WDELAYHASHSIZ]; 994 } 995 996 /* 997 * NFS write service with write gathering support. Called when 998 * nfsrvw_procrastinate > 0. 999 * See: Chet Juszczak, "Improving the Write Performance of an NFS Server", 1000 * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco, 1001 * Jan. 1994. 1002 */ 1003 int 1004 nfsrv_writegather(ndp, slp, lwp, mrq) 1005 struct nfsrv_descript **ndp; 1006 struct nfssvc_sock *slp; 1007 struct lwp *lwp; 1008 struct mbuf **mrq; 1009 { 1010 struct timeval now; 1011 struct iovec *ivp; 1012 struct mbuf *mp; 1013 struct nfsrv_descript *wp, *nfsd, *owp, *swp; 1014 struct nfs_fattr *fp; 1015 int i = 0; 1016 struct iovec *iov; 1017 struct nfsrvw_delayhash *wpp; 1018 kauth_cred_t cred; 1019 struct vattr va, forat; 1020 u_int32_t *tl; 1021 int32_t t1; 1022 char *bpos, *dpos; 1023 int error = 0, rdonly, cache = 0, len = 0, forat_ret = 1; 1024 int ioflags, aftat_ret = 1, adjust, v3, zeroing; 1025 char *cp2; 1026 struct mbuf *mb, *mreq, *mrep, *md; 1027 struct vnode *vp; 1028 struct uio io, *uiop = &io; 1029 u_quad_t frev, cur_usec; 1030 1031 *mrq = NULL; 1032 if (*ndp) { 1033 nfsd = *ndp; 1034 *ndp = NULL; 1035 mrep = nfsd->nd_mrep; 1036 md = nfsd->nd_md; 1037 dpos = nfsd->nd_dpos; 1038 cred = nfsd->nd_cr; 1039 v3 = (nfsd->nd_flag & ND_NFSV3); 1040 LIST_INIT(&nfsd->nd_coalesce); 1041 nfsd->nd_mreq = NULL; 1042 nfsd->nd_stable = NFSV3WRITE_FILESYNC; 1043 getmicrotime(&now); 1044 cur_usec = (u_quad_t)now.tv_sec * 1000000 + (u_quad_t)now.tv_usec; 1045 nfsd->nd_time = cur_usec + nfsrvw_procrastinate; 1046 1047 /* 1048 * Now, get the write header.. 1049 */ 1050 nfsm_srvmtofh(&nfsd->nd_fh); 1051 if (v3) { 1052 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 1053 nfsd->nd_off = fxdr_hyper(tl); 1054 tl += 3; 1055 nfsd->nd_stable = fxdr_unsigned(int, *tl++); 1056 } else { 1057 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1058 nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 1059 tl += 2; 1060 } 1061 len = fxdr_unsigned(int32_t, *tl); 1062 nfsd->nd_len = len; 1063 nfsd->nd_eoff = nfsd->nd_off + len; 1064 1065 /* 1066 * Trim the header out of the mbuf list and trim off any trailing 1067 * junk so that the mbuf list has only the write data. 1068 */ 1069 zeroing = 1; 1070 i = 0; 1071 mp = mrep; 1072 while (mp) { 1073 if (mp == md) { 1074 zeroing = 0; 1075 adjust = dpos - mtod(mp, char *); 1076 mp->m_len -= adjust; 1077 if (mp->m_len > 0 && adjust > 0) 1078 NFSMADV(mp, adjust); 1079 } 1080 if (zeroing) 1081 mp->m_len = 0; 1082 else { 1083 i += mp->m_len; 1084 if (i > len) { 1085 mp->m_len -= (i - len); 1086 zeroing = 1; 1087 } 1088 } 1089 mp = mp->m_next; 1090 } 1091 if (len > NFS_MAXDATA || len < 0 || i < len) { 1092 nfsmout: 1093 m_freem(mrep); 1094 error = EIO; 1095 nfsm_writereply(2 * NFSX_UNSIGNED, v3); 1096 if (v3) 1097 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1098 nfsd->nd_mreq = mreq; 1099 nfsd->nd_mrep = NULL; 1100 nfsd->nd_time = 0; 1101 } 1102 1103 /* 1104 * Add this entry to the hash and time queues. 1105 */ 1106 owp = NULL; 1107 mutex_enter(&nfsd_lock); 1108 wp = LIST_FIRST(&slp->ns_tq); 1109 while (wp && wp->nd_time < nfsd->nd_time) { 1110 owp = wp; 1111 wp = LIST_NEXT(wp, nd_tq); 1112 } 1113 if (owp) { 1114 LIST_INSERT_AFTER(owp, nfsd, nd_tq); 1115 } else { 1116 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1117 } 1118 if (nfsd->nd_mrep) { 1119 wpp = nfsrv_nwdelayhash(slp, &nfsd->nd_fh); 1120 owp = NULL; 1121 wp = LIST_FIRST(wpp); 1122 while (wp && nfsrv_comparefh(&nfsd->nd_fh, &wp->nd_fh)) { 1123 owp = wp; 1124 wp = LIST_NEXT(wp, nd_hash); 1125 } 1126 while (wp && wp->nd_off < nfsd->nd_off && 1127 !nfsrv_comparefh(&nfsd->nd_fh, &wp->nd_fh)) { 1128 owp = wp; 1129 wp = LIST_NEXT(wp, nd_hash); 1130 } 1131 if (owp) { 1132 LIST_INSERT_AFTER(owp, nfsd, nd_hash); 1133 1134 /* 1135 * Search the hash list for overlapping entries and 1136 * coalesce. 1137 */ 1138 for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) { 1139 wp = LIST_NEXT(nfsd, nd_hash); 1140 if (nfsrv_samecred(owp->nd_cr, nfsd->nd_cr)) 1141 nfsrvw_coalesce(owp, nfsd); 1142 } 1143 } else { 1144 LIST_INSERT_HEAD(wpp, nfsd, nd_hash); 1145 } 1146 } 1147 mutex_exit(&nfsd_lock); 1148 } 1149 1150 /* 1151 * Now, do VOP_WRITE()s for any one(s) that need to be done now 1152 * and generate the associated reply mbuf list(s). 1153 */ 1154 loop1: 1155 getmicrotime(&now); 1156 cur_usec = (u_quad_t)now.tv_sec * 1000000 + (u_quad_t)now.tv_usec; 1157 mutex_enter(&nfsd_lock); 1158 for (nfsd = LIST_FIRST(&slp->ns_tq); nfsd; nfsd = owp) { 1159 owp = LIST_NEXT(nfsd, nd_tq); 1160 if (nfsd->nd_time > cur_usec) 1161 break; 1162 if (nfsd->nd_mreq) 1163 continue; 1164 LIST_REMOVE(nfsd, nd_tq); 1165 LIST_REMOVE(nfsd, nd_hash); 1166 mutex_exit(&nfsd_lock); 1167 1168 mrep = nfsd->nd_mrep; 1169 nfsd->nd_mrep = NULL; 1170 cred = nfsd->nd_cr; 1171 v3 = (nfsd->nd_flag & ND_NFSV3); 1172 forat_ret = aftat_ret = 1; 1173 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp, 1174 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH), 1175 false); 1176 if (!error) { 1177 if (v3) 1178 forat_ret = VOP_GETATTR(vp, &forat, cred); 1179 if (vp->v_type != VREG) { 1180 if (v3) 1181 error = EINVAL; 1182 else 1183 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 1184 } 1185 } else 1186 vp = NULL; 1187 if (!error) { 1188 nqsrv_getl(vp, ND_WRITE); 1189 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1); 1190 } 1191 1192 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE) 1193 ioflags = IO_NODELOCKED; 1194 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC) 1195 ioflags = (IO_SYNC | IO_NODELOCKED); 1196 else 1197 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 1198 uiop->uio_rw = UIO_WRITE; 1199 uiop->uio_offset = nfsd->nd_off; 1200 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off; 1201 UIO_SETUP_SYSSPACE(uiop); 1202 if (uiop->uio_resid > 0) { 1203 mp = mrep; 1204 i = 0; 1205 while (mp) { 1206 if (mp->m_len > 0) 1207 i++; 1208 mp = mp->m_next; 1209 } 1210 uiop->uio_iovcnt = i; 1211 iov = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK); 1212 uiop->uio_iov = ivp = iov; 1213 mp = mrep; 1214 while (mp) { 1215 if (mp->m_len > 0) { 1216 ivp->iov_base = mtod(mp, void *); 1217 ivp->iov_len = mp->m_len; 1218 ivp++; 1219 } 1220 mp = mp->m_next; 1221 } 1222 if (!error) { 1223 error = VOP_WRITE(vp, uiop, ioflags, cred); 1224 nfsstats.srvvop_writes++; 1225 } 1226 free((void *)iov, M_TEMP); 1227 } 1228 m_freem(mrep); 1229 if (vp) { 1230 aftat_ret = VOP_GETATTR(vp, &va, cred); 1231 vput(vp); 1232 } 1233 1234 /* 1235 * Loop around generating replies for all write rpcs that have 1236 * now been completed. 1237 */ 1238 swp = nfsd; 1239 do { 1240 if (error) { 1241 nfsm_writereply(NFSX_WCCDATA(v3), v3); 1242 if (v3) { 1243 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1244 } 1245 } else { 1246 nfsm_writereply(NFSX_PREOPATTR(v3) + 1247 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED + 1248 NFSX_WRITEVERF(v3), v3); 1249 if (v3) { 1250 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1251 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1252 *tl++ = txdr_unsigned(nfsd->nd_len); 1253 *tl++ = txdr_unsigned(swp->nd_stable); 1254 /* 1255 * Actually, there is no need to txdr these fields, 1256 * but it may make the values more human readable, 1257 * for debugging purposes. 1258 */ 1259 *tl++ = txdr_unsigned(boottime.tv_sec); 1260 *tl = txdr_unsigned(boottime.tv_usec); 1261 } else { 1262 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1263 nfsm_srvfillattr(&va, fp); 1264 } 1265 } 1266 nfsd->nd_mreq = mreq; 1267 if (nfsd->nd_mrep) 1268 panic("nfsrv_write: nd_mrep not free"); 1269 1270 /* 1271 * Done. Put it at the head of the timer queue so that 1272 * the final phase can return the reply. 1273 */ 1274 mutex_enter(&nfsd_lock); 1275 if (nfsd != swp) { 1276 nfsd->nd_time = 0; 1277 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1278 } 1279 nfsd = LIST_FIRST(&swp->nd_coalesce); 1280 if (nfsd) { 1281 LIST_REMOVE(nfsd, nd_tq); 1282 } 1283 mutex_exit(&nfsd_lock); 1284 } while (nfsd); 1285 swp->nd_time = 0; 1286 1287 mutex_enter(&nfsd_lock); 1288 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq); 1289 mutex_exit(&nfsd_lock); 1290 goto loop1; 1291 } 1292 mutex_exit(&nfsd_lock); 1293 nfs_timer_start(); 1294 1295 /* 1296 * Search for a reply to return. 1297 */ 1298 mutex_enter(&nfsd_lock); 1299 LIST_FOREACH(nfsd, &slp->ns_tq, nd_tq) { 1300 if (nfsd->nd_mreq) { 1301 LIST_REMOVE(nfsd, nd_tq); 1302 *mrq = nfsd->nd_mreq; 1303 *ndp = nfsd; 1304 break; 1305 } 1306 } 1307 mutex_exit(&nfsd_lock); 1308 return (0); 1309 } 1310 1311 /* 1312 * Coalesce the write request nfsd into owp. To do this we must: 1313 * - remove nfsd from the queues 1314 * - merge nfsd->nd_mrep into owp->nd_mrep 1315 * - update the nd_eoff and nd_stable for owp 1316 * - put nfsd on owp's nd_coalesce list 1317 * NB: Must be called at splsoftclock(). 1318 */ 1319 void 1320 nfsrvw_coalesce(owp, nfsd) 1321 struct nfsrv_descript *owp; 1322 struct nfsrv_descript *nfsd; 1323 { 1324 int overlap; 1325 struct mbuf *mp; 1326 struct nfsrv_descript *m; 1327 1328 KASSERT(mutex_owned(&nfsd_lock)); 1329 1330 LIST_REMOVE(nfsd, nd_hash); 1331 LIST_REMOVE(nfsd, nd_tq); 1332 if (owp->nd_eoff < nfsd->nd_eoff) { 1333 overlap = owp->nd_eoff - nfsd->nd_off; 1334 if (overlap < 0) 1335 panic("nfsrv_coalesce: bad off"); 1336 if (overlap > 0) 1337 m_adj(nfsd->nd_mrep, overlap); 1338 mp = owp->nd_mrep; 1339 while (mp->m_next) 1340 mp = mp->m_next; 1341 mp->m_next = nfsd->nd_mrep; 1342 owp->nd_eoff = nfsd->nd_eoff; 1343 } else 1344 m_freem(nfsd->nd_mrep); 1345 nfsd->nd_mrep = NULL; 1346 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC) 1347 owp->nd_stable = NFSV3WRITE_FILESYNC; 1348 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC && 1349 owp->nd_stable == NFSV3WRITE_UNSTABLE) 1350 owp->nd_stable = NFSV3WRITE_DATASYNC; 1351 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq); 1352 /* 1353 * nfsd might hold coalesce elements! Move them to owp. 1354 * Otherwise, requests may be lost and clients will be stuck. 1355 */ 1356 while ((m = LIST_FIRST(&nfsd->nd_coalesce)) != NULL) { 1357 LIST_REMOVE(m, nd_tq); 1358 LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq); 1359 } 1360 } 1361 1362 /* 1363 * nfs create service 1364 * now does a truncate to 0 length via. setattr if it already exists 1365 */ 1366 int 1367 nfsrv_create(nfsd, slp, lwp, mrq) 1368 struct nfsrv_descript *nfsd; 1369 struct nfssvc_sock *slp; 1370 struct lwp *lwp; 1371 struct mbuf **mrq; 1372 { 1373 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1374 struct mbuf *nam = nfsd->nd_nam; 1375 char *dpos = nfsd->nd_dpos; 1376 kauth_cred_t cred = nfsd->nd_cr; 1377 struct nfs_fattr *fp; 1378 struct vattr va, dirfor, diraft; 1379 struct nfsv2_sattr *sp; 1380 u_int32_t *tl; 1381 struct nameidata nd; 1382 char *cp; 1383 int32_t t1; 1384 char *bpos; 1385 int error = 0, cache = 0, len, tsize, dirfor_ret = 1, diraft_ret = 1; 1386 int rdev = 0; 1387 int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0; 1388 char *cp2; 1389 struct mbuf *mb, *mreq; 1390 struct vnode *vp = NULL, *dirp = NULL; 1391 nfsrvfh_t nsfh; 1392 u_quad_t frev, tempsize; 1393 u_char cverf[NFSX_V3CREATEVERF]; 1394 1395 nd.ni_cnd.cn_nameiop = 0; 1396 nfsm_srvmtofh(&nsfh); 1397 nfsm_srvnamesiz(len); 1398 nd.ni_cnd.cn_cred = cred; 1399 nd.ni_cnd.cn_nameiop = CREATE; 1400 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1401 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 1402 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1403 if (dirp && v3) { 1404 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred); 1405 } 1406 if (error) { 1407 nfsm_reply(NFSX_WCCDATA(v3)); 1408 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1409 if (dirp) 1410 vrele(dirp); 1411 return (0); 1412 } 1413 VATTR_NULL(&va); 1414 if (v3) { 1415 va.va_mode = 0; 1416 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1417 how = fxdr_unsigned(int, *tl); 1418 switch (how) { 1419 case NFSV3CREATE_GUARDED: 1420 if (nd.ni_vp) { 1421 error = EEXIST; 1422 break; 1423 } 1424 case NFSV3CREATE_UNCHECKED: 1425 nfsm_srvsattr(&va); 1426 break; 1427 case NFSV3CREATE_EXCLUSIVE: 1428 nfsm_dissect(cp, void *, NFSX_V3CREATEVERF); 1429 memcpy(cverf, cp, NFSX_V3CREATEVERF); 1430 exclusive_flag = 1; 1431 break; 1432 }; 1433 va.va_type = VREG; 1434 } else { 1435 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 1436 va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode)); 1437 if (va.va_type == VNON) 1438 va.va_type = VREG; 1439 va.va_mode = nfstov_mode(sp->sa_mode); 1440 switch (va.va_type) { 1441 case VREG: 1442 tsize = fxdr_unsigned(int32_t, sp->sa_size); 1443 if (tsize != -1) 1444 va.va_size = (u_quad_t)tsize; 1445 break; 1446 case VCHR: 1447 case VBLK: 1448 case VFIFO: 1449 rdev = fxdr_unsigned(int32_t, sp->sa_size); 1450 break; 1451 default: 1452 break; 1453 }; 1454 } 1455 1456 /* 1457 * Iff doesn't exist, create it 1458 * otherwise just truncate to 0 length 1459 * should I set the mode too ?? 1460 */ 1461 if (nd.ni_vp == NULL) { 1462 if (va.va_type == VREG || va.va_type == VSOCK) { 1463 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1464 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1465 if (!error) { 1466 if (exclusive_flag) { 1467 exclusive_flag = 0; 1468 VATTR_NULL(&va); 1469 /* 1470 * XXX 1471 * assuming NFSX_V3CREATEVERF 1472 * == sizeof(nfstime3) 1473 */ 1474 fxdr_nfsv3time(cverf, &va.va_atime); 1475 error = VOP_SETATTR(nd.ni_vp, &va, 1476 cred); 1477 } 1478 } 1479 } else if (va.va_type == VCHR || va.va_type == VBLK || 1480 va.va_type == VFIFO) { 1481 if (va.va_type == VCHR && rdev == 0xffffffff) 1482 va.va_type = VFIFO; 1483 if (va.va_type != VFIFO && 1484 (error = kauth_authorize_generic(cred, 1485 KAUTH_GENERIC_ISSUSER, NULL))) { 1486 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1487 vput(nd.ni_dvp); 1488 nfsm_reply(0); 1489 return (error); 1490 } else 1491 va.va_rdev = (dev_t)rdev; 1492 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1493 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, 1494 &va); 1495 if (error) { 1496 nfsm_reply(0); 1497 } 1498 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1499 vput(nd.ni_vp); 1500 vrele(nd.ni_dvp); 1501 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1502 error = EINVAL; 1503 nfsm_reply(0); 1504 } 1505 } else { 1506 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1507 vput(nd.ni_dvp); 1508 error = ENXIO; 1509 } 1510 vp = nd.ni_vp; 1511 } else { 1512 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1513 vp = nd.ni_vp; 1514 if (nd.ni_dvp == vp) 1515 vrele(nd.ni_dvp); 1516 else 1517 vput(nd.ni_dvp); 1518 if (!error && va.va_size != -1) { 1519 error = nfsrv_access(vp, VWRITE, cred, 1520 (nd.ni_cnd.cn_flags & RDONLY), lwp, 0); 1521 if (!error) { 1522 nqsrv_getl(vp, ND_WRITE); 1523 tempsize = va.va_size; 1524 VATTR_NULL(&va); 1525 va.va_size = tempsize; 1526 error = VOP_SETATTR(vp, &va, cred); 1527 } 1528 } 1529 if (error) 1530 vput(vp); 1531 } 1532 if (!error) { 1533 error = nfsrv_composefh(vp, &nsfh, v3); 1534 if (!error) 1535 error = VOP_GETATTR(vp, &va, cred); 1536 vput(vp); 1537 } 1538 if (v3) { 1539 if (exclusive_flag && !error) { 1540 /* 1541 * XXX assuming NFSX_V3CREATEVERF == sizeof(nfstime3) 1542 */ 1543 char oldverf[NFSX_V3CREATEVERF]; 1544 1545 txdr_nfsv3time(&va.va_atime, oldverf); 1546 if (memcmp(cverf, oldverf, NFSX_V3CREATEVERF)) 1547 error = EEXIST; 1548 } 1549 if (dirp) { 1550 diraft_ret = VOP_GETATTR(dirp, &diraft, cred); 1551 } 1552 } 1553 if (dirp) { 1554 vrele(dirp); 1555 } 1556 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3)); 1557 if (v3) { 1558 if (!error) { 1559 nfsm_srvpostop_fh(&nsfh); 1560 nfsm_srvpostop_attr(0, &va); 1561 } 1562 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1563 } else { 1564 nfsm_srvfhtom(&nsfh, v3); 1565 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1566 nfsm_srvfillattr(&va, fp); 1567 } 1568 return (0); 1569 nfsmout: 1570 if (dirp) 1571 vrele(dirp); 1572 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1573 if (nd.ni_dvp == nd.ni_vp) 1574 vrele(nd.ni_dvp); 1575 else 1576 vput(nd.ni_dvp); 1577 if (nd.ni_vp) 1578 vput(nd.ni_vp); 1579 return (error); 1580 } 1581 1582 /* 1583 * nfs v3 mknod service 1584 */ 1585 int 1586 nfsrv_mknod(nfsd, slp, lwp, mrq) 1587 struct nfsrv_descript *nfsd; 1588 struct nfssvc_sock *slp; 1589 struct lwp *lwp; 1590 struct mbuf **mrq; 1591 { 1592 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1593 struct mbuf *nam = nfsd->nd_nam; 1594 char *dpos = nfsd->nd_dpos; 1595 kauth_cred_t cred = nfsd->nd_cr; 1596 struct vattr va, dirfor, diraft; 1597 u_int32_t *tl; 1598 struct nameidata nd; 1599 int32_t t1; 1600 char *bpos; 1601 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 1602 u_int32_t major, minor; 1603 enum vtype vtyp; 1604 char *cp2; 1605 struct mbuf *mb, *mreq; 1606 struct vnode *vp, *dirp = (struct vnode *)0; 1607 nfsrvfh_t nsfh; 1608 u_quad_t frev; 1609 1610 nd.ni_cnd.cn_nameiop = 0; 1611 nfsm_srvmtofh(&nsfh); 1612 nfsm_srvnamesiz(len); 1613 nd.ni_cnd.cn_cred = cred; 1614 nd.ni_cnd.cn_nameiop = CREATE; 1615 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1616 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 1617 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1618 if (dirp) 1619 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred); 1620 if (error) { 1621 nfsm_reply(NFSX_WCCDATA(1)); 1622 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1623 if (dirp) 1624 vrele(dirp); 1625 return (0); 1626 } 1627 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1628 vtyp = nfsv3tov_type(*tl); 1629 if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) { 1630 error = NFSERR_BADTYPE; 1631 goto abort; 1632 } 1633 VATTR_NULL(&va); 1634 va.va_mode = 0; 1635 nfsm_srvsattr(&va); 1636 if (vtyp == VCHR || vtyp == VBLK) { 1637 dev_t rdev; 1638 1639 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1640 major = fxdr_unsigned(u_int32_t, *tl++); 1641 minor = fxdr_unsigned(u_int32_t, *tl); 1642 rdev = makedev(major, minor); 1643 if (major(rdev) != major || minor(rdev) != minor) { 1644 error = EINVAL; 1645 goto abort; 1646 } 1647 va.va_rdev = rdev; 1648 } 1649 1650 /* 1651 * Iff doesn't exist, create it. 1652 */ 1653 if (nd.ni_vp) { 1654 error = EEXIST; 1655 abort: 1656 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1657 if (nd.ni_dvp == nd.ni_vp) 1658 vrele(nd.ni_dvp); 1659 else 1660 vput(nd.ni_dvp); 1661 if (nd.ni_vp) 1662 vput(nd.ni_vp); 1663 goto out; 1664 } 1665 va.va_type = vtyp; 1666 if (vtyp == VSOCK) { 1667 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1668 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1669 } else { 1670 if (va.va_type != VFIFO && 1671 (error = kauth_authorize_generic(cred, 1672 KAUTH_GENERIC_ISSUSER, NULL))) { 1673 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1674 vput(nd.ni_dvp); 1675 goto out; 1676 } 1677 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1678 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1679 if (error) 1680 goto out; 1681 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1682 vput(nd.ni_vp); 1683 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1684 error = EINVAL; 1685 } 1686 } 1687 out: 1688 vp = nd.ni_vp; 1689 if (!error) { 1690 error = nfsrv_composefh(vp, &nsfh, true); 1691 if (!error) 1692 error = VOP_GETATTR(vp, &va, cred); 1693 vput(vp); 1694 } 1695 if (dirp) { 1696 diraft_ret = VOP_GETATTR(dirp, &diraft, cred); 1697 vrele(dirp); 1698 } 1699 nfsm_reply(NFSX_SRVFH(&nsfh, true) + NFSX_POSTOPATTR(1) + 1700 NFSX_WCCDATA(1)); 1701 if (!error) { 1702 nfsm_srvpostop_fh(&nsfh); 1703 nfsm_srvpostop_attr(0, &va); 1704 } 1705 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1706 return (0); 1707 nfsmout: 1708 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1709 if (nd.ni_dvp == nd.ni_vp) 1710 vrele(nd.ni_dvp); 1711 else 1712 vput(nd.ni_dvp); 1713 if (nd.ni_vp) 1714 vput(nd.ni_vp); 1715 if (dirp) 1716 vrele(dirp); 1717 return (error); 1718 } 1719 1720 /* 1721 * nfs remove service 1722 */ 1723 int 1724 nfsrv_remove(nfsd, slp, lwp, mrq) 1725 struct nfsrv_descript *nfsd; 1726 struct nfssvc_sock *slp; 1727 struct lwp *lwp; 1728 struct mbuf **mrq; 1729 { 1730 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1731 struct mbuf *nam = nfsd->nd_nam; 1732 char *dpos = nfsd->nd_dpos; 1733 kauth_cred_t cred = nfsd->nd_cr; 1734 struct nameidata nd; 1735 u_int32_t *tl; 1736 int32_t t1; 1737 char *bpos; 1738 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 1739 int v3 = (nfsd->nd_flag & ND_NFSV3); 1740 char *cp2; 1741 struct mbuf *mb, *mreq; 1742 struct vnode *vp, *dirp; 1743 struct vattr dirfor, diraft; 1744 nfsrvfh_t nsfh; 1745 u_quad_t frev; 1746 1747 #ifndef nolint 1748 vp = (struct vnode *)0; 1749 #endif 1750 nfsm_srvmtofh(&nsfh); 1751 nfsm_srvnamesiz(len); 1752 nd.ni_cnd.cn_cred = cred; 1753 nd.ni_cnd.cn_nameiop = DELETE; 1754 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1755 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 1756 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1757 if (dirp && v3) { 1758 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred); 1759 } 1760 if (!error) { 1761 vp = nd.ni_vp; 1762 if (vp->v_type == VDIR && 1763 (error = kauth_authorize_generic(cred, 1764 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 1765 goto out; 1766 /* 1767 * The root of a mounted filesystem cannot be deleted. 1768 */ 1769 if (vp->v_vflag & VV_ROOT) { 1770 error = EBUSY; 1771 goto out; 1772 } 1773 out: 1774 if (!error) { 1775 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1776 nqsrv_getl(vp, ND_WRITE); 1777 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1778 } else { 1779 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1780 if (nd.ni_dvp == vp) 1781 vrele(nd.ni_dvp); 1782 else 1783 vput(nd.ni_dvp); 1784 vput(vp); 1785 } 1786 } 1787 if (dirp) { 1788 if (v3) { 1789 diraft_ret = VOP_GETATTR(dirp, &diraft, cred); 1790 } 1791 vrele(dirp); 1792 } 1793 nfsm_reply(NFSX_WCCDATA(v3)); 1794 if (v3) { 1795 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1796 return (0); 1797 } 1798 nfsm_srvdone; 1799 } 1800 1801 /* 1802 * nfs rename service 1803 */ 1804 int 1805 nfsrv_rename(nfsd, slp, lwp, mrq) 1806 struct nfsrv_descript *nfsd; 1807 struct nfssvc_sock *slp; 1808 struct lwp *lwp; 1809 struct mbuf **mrq; 1810 { 1811 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1812 struct mbuf *nam = nfsd->nd_nam; 1813 char *dpos = nfsd->nd_dpos; 1814 kauth_cred_t cred = nfsd->nd_cr; 1815 u_int32_t *tl; 1816 int32_t t1; 1817 char *bpos; 1818 int error = 0, cache = 0, fdirfor_ret = 1, fdiraft_ret = 1; 1819 uint32_t len, len2; 1820 int tdirfor_ret = 1, tdiraft_ret = 1; 1821 int v3 = (nfsd->nd_flag & ND_NFSV3); 1822 char *cp2; 1823 struct mbuf *mb, *mreq; 1824 struct nameidata fromnd, tond; 1825 struct vnode *fvp, *tvp, *tdvp; 1826 struct vnode *fdirp = NULL, *tdirp = NULL; 1827 struct mount *localfs = NULL; 1828 struct vattr fdirfor, fdiraft, tdirfor, tdiraft; 1829 nfsrvfh_t fnsfh, tnsfh; 1830 u_quad_t frev; 1831 uid_t saved_uid; 1832 uint32_t saveflag; 1833 1834 #ifndef nolint 1835 fvp = (struct vnode *)0; 1836 #endif 1837 fromnd.ni_cnd.cn_nameiop = 0; 1838 tond.ni_cnd.cn_nameiop = 0; 1839 nfsm_srvmtofh(&fnsfh); 1840 nfsm_srvnamesiz(len); 1841 /* 1842 * Remember our original uid so that we can reset cr_uid before 1843 * the second nfs_namei() call, in case it is remapped. 1844 */ 1845 saved_uid = kauth_cred_geteuid(cred); 1846 fromnd.ni_cnd.cn_cred = cred; 1847 fromnd.ni_cnd.cn_nameiop = DELETE; 1848 fromnd.ni_cnd.cn_flags = LOCKPARENT | SAVESTART; 1849 error = nfs_namei(&fromnd, &fnsfh, len, slp, nam, &md, 1850 &dpos, &fdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1851 if (fdirp && v3) { 1852 fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred); 1853 } 1854 if (error) { 1855 nfsm_reply(2 * NFSX_WCCDATA(v3)); 1856 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 1857 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 1858 if (fdirp) 1859 vrele(fdirp); 1860 return (0); 1861 } 1862 if (fromnd.ni_dvp != fromnd.ni_vp) { 1863 VOP_UNLOCK(fromnd.ni_dvp, 0); 1864 } 1865 fvp = fromnd.ni_vp; 1866 1867 localfs = fvp->v_mount; 1868 error = VFS_RENAMELOCK_ENTER(localfs); 1869 if (error) { 1870 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1871 vrele(fromnd.ni_dvp); 1872 vrele(fvp); 1873 goto out1; 1874 } 1875 1876 /* Copied, regrettably, from vfs_syscalls.c (q.v.) */ 1877 vrele(fvp); 1878 saveflag = fromnd.ni_cnd.cn_flags & SAVESTART; 1879 fromnd.ni_cnd.cn_flags &= ~SAVESTART; 1880 vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY); 1881 error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd); 1882 fromnd.ni_cnd.cn_flags |= saveflag; 1883 if (error) { 1884 VOP_UNLOCK(fromnd.ni_dvp, 0); 1885 VFS_RENAMELOCK_EXIT(localfs); 1886 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1887 vrele(fromnd.ni_dvp); 1888 goto out1; 1889 } 1890 VOP_UNLOCK(fromnd.ni_vp, 0); 1891 if (fromnd.ni_dvp != fromnd.ni_vp) 1892 VOP_UNLOCK(fromnd.ni_dvp, 0); 1893 fvp = fromnd.ni_vp; 1894 1895 nfsm_srvmtofh(&tnsfh); 1896 if (v3) { 1897 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 1898 len2 = fxdr_unsigned(uint32_t, *tl); 1899 /* len2 will be checked by nfs_namei */ 1900 } 1901 else { 1902 /* NFSv2 */ 1903 nfsm_strsiz(len2, NFS_MAXNAMLEN); 1904 } 1905 kauth_cred_seteuid(cred, saved_uid); 1906 tond.ni_cnd.cn_cred = cred; 1907 tond.ni_cnd.cn_nameiop = RENAME; 1908 tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART; 1909 error = nfs_namei(&tond, &tnsfh, len2, slp, nam, &md, 1910 &dpos, &tdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1911 if (tdirp && v3) { 1912 tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred); 1913 } 1914 if (error) { 1915 VFS_RENAMELOCK_EXIT(localfs); 1916 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1917 vrele(fromnd.ni_dvp); 1918 vrele(fvp); 1919 goto out1; 1920 } 1921 tdvp = tond.ni_dvp; 1922 tvp = tond.ni_vp; 1923 if (tvp != NULL) { 1924 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 1925 if (v3) 1926 error = EEXIST; 1927 else 1928 error = EISDIR; 1929 goto out; 1930 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 1931 if (v3) 1932 error = EEXIST; 1933 else 1934 error = ENOTDIR; 1935 goto out; 1936 } 1937 if (tvp->v_type == VDIR && tvp->v_mountedhere) { 1938 if (v3) 1939 error = EXDEV; 1940 else 1941 error = ENOTEMPTY; 1942 goto out; 1943 } 1944 } 1945 if (fvp->v_type == VDIR && fvp->v_mountedhere) { 1946 if (v3) 1947 error = EXDEV; 1948 else 1949 error = ENOTEMPTY; 1950 goto out; 1951 } 1952 if (fvp->v_mount != tdvp->v_mount) { 1953 if (v3) 1954 error = EXDEV; 1955 else 1956 error = ENOTEMPTY; 1957 goto out; 1958 } 1959 if (fvp == tdvp) { 1960 if (v3) 1961 error = EINVAL; 1962 else 1963 error = ENOTEMPTY; 1964 } 1965 /* 1966 * If source is the same as the destination (that is the 1967 * same vnode with the same name in the same directory), 1968 * then there is nothing to do. 1969 */ 1970 if (fvp == tvp && fromnd.ni_dvp == tdvp && 1971 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 1972 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr, 1973 fromnd.ni_cnd.cn_namelen)) 1974 error = -1; 1975 out: 1976 if (!error) { 1977 nqsrv_getl(fromnd.ni_dvp, ND_WRITE); 1978 nqsrv_getl(tdvp, ND_WRITE); 1979 if (tvp) { 1980 nqsrv_getl(tvp, ND_WRITE); 1981 } 1982 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 1983 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 1984 VFS_RENAMELOCK_EXIT(localfs); 1985 } else { 1986 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 1987 if (tdvp == tvp) 1988 vrele(tdvp); 1989 else 1990 vput(tdvp); 1991 if (tvp) 1992 vput(tvp); 1993 VFS_RENAMELOCK_EXIT(localfs); 1994 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1995 vrele(fromnd.ni_dvp); 1996 vrele(fvp); 1997 if (error == -1) 1998 error = 0; 1999 } 2000 vrele(tond.ni_startdir); 2001 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 2002 out1: 2003 if (fdirp) { 2004 if (v3) { 2005 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred); 2006 } 2007 vrele(fdirp); 2008 } 2009 if (tdirp) { 2010 if (v3) { 2011 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred); 2012 } 2013 vrele(tdirp); 2014 } 2015 vrele(fromnd.ni_startdir); 2016 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2017 nfsm_reply(2 * NFSX_WCCDATA(v3)); 2018 if (v3) { 2019 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 2020 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 2021 } 2022 return (0); 2023 2024 nfsmout: 2025 if (fdirp) 2026 vrele(fdirp); 2027 #ifdef notdef 2028 if (tdirp) 2029 vrele(tdirp); 2030 #endif 2031 if (tond.ni_cnd.cn_nameiop) { 2032 vrele(tond.ni_startdir); 2033 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 2034 } 2035 if (localfs) { 2036 VFS_RENAMELOCK_EXIT(localfs); 2037 } 2038 if (fromnd.ni_cnd.cn_nameiop) { 2039 vrele(fromnd.ni_startdir); 2040 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2041 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2042 vrele(fromnd.ni_dvp); 2043 vrele(fvp); 2044 } 2045 return (error); 2046 } 2047 2048 /* 2049 * nfs link service 2050 */ 2051 int 2052 nfsrv_link(nfsd, slp, lwp, mrq) 2053 struct nfsrv_descript *nfsd; 2054 struct nfssvc_sock *slp; 2055 struct lwp *lwp; 2056 struct mbuf **mrq; 2057 { 2058 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2059 struct mbuf *nam = nfsd->nd_nam; 2060 char *dpos = nfsd->nd_dpos; 2061 kauth_cred_t cred = nfsd->nd_cr; 2062 struct nameidata nd; 2063 u_int32_t *tl; 2064 int32_t t1; 2065 char *bpos; 2066 int error = 0, rdonly, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 2067 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3); 2068 char *cp2; 2069 struct mbuf *mb, *mreq; 2070 struct vnode *vp, *xp, *dirp = (struct vnode *)0; 2071 struct vattr dirfor, diraft, at; 2072 nfsrvfh_t nsfh, dnsfh; 2073 u_quad_t frev; 2074 2075 nfsm_srvmtofh(&nsfh); 2076 nfsm_srvmtofh(&dnsfh); 2077 nfsm_srvnamesiz(len); 2078 error = nfsrv_fhtovp(&nsfh, false, &vp, cred, slp, nam, 2079 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 2080 if (error) { 2081 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2082 nfsm_srvpostop_attr(getret, &at); 2083 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2084 return (0); 2085 } 2086 if (vp->v_type == VDIR && (error = kauth_authorize_generic(cred, 2087 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 2088 goto out1; 2089 nd.ni_cnd.cn_cred = cred; 2090 nd.ni_cnd.cn_nameiop = CREATE; 2091 nd.ni_cnd.cn_flags = LOCKPARENT; 2092 error = nfs_namei(&nd, &dnsfh, len, slp, nam, &md, &dpos, 2093 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2094 if (dirp && v3) { 2095 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred); 2096 } 2097 if (error) 2098 goto out1; 2099 xp = nd.ni_vp; 2100 if (xp != NULL) { 2101 error = EEXIST; 2102 goto out; 2103 } 2104 xp = nd.ni_dvp; 2105 if (vp->v_mount != xp->v_mount) 2106 error = EXDEV; 2107 out: 2108 if (!error) { 2109 nqsrv_getl(vp, ND_WRITE); 2110 nqsrv_getl(xp, ND_WRITE); 2111 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2112 } else { 2113 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2114 if (nd.ni_dvp == nd.ni_vp) 2115 vrele(nd.ni_dvp); 2116 else 2117 vput(nd.ni_dvp); 2118 if (nd.ni_vp) 2119 vrele(nd.ni_vp); 2120 } 2121 out1: 2122 if (v3) 2123 getret = VOP_GETATTR(vp, &at, cred); 2124 if (dirp) { 2125 if (v3) { 2126 diraft_ret = VOP_GETATTR(dirp, &diraft, cred); 2127 } 2128 vrele(dirp); 2129 } 2130 vrele(vp); 2131 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2132 if (v3) { 2133 nfsm_srvpostop_attr(getret, &at); 2134 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2135 return (0); 2136 } 2137 nfsm_srvdone; 2138 } 2139 2140 /* 2141 * nfs symbolic link service 2142 */ 2143 int 2144 nfsrv_symlink(nfsd, slp, lwp, mrq) 2145 struct nfsrv_descript *nfsd; 2146 struct nfssvc_sock *slp; 2147 struct lwp *lwp; 2148 struct mbuf **mrq; 2149 { 2150 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2151 struct mbuf *nam = nfsd->nd_nam; 2152 char *dpos = nfsd->nd_dpos; 2153 kauth_cred_t cred = nfsd->nd_cr; 2154 struct vattr va, dirfor, diraft; 2155 struct nameidata nd; 2156 u_int32_t *tl; 2157 int32_t t1; 2158 struct nfsv2_sattr *sp; 2159 char *bpos, *pathcp = NULL, *cp2; 2160 struct uio io; 2161 struct iovec iv; 2162 int error = 0, cache = 0, dirfor_ret = 1, diraft_ret = 1; 2163 uint32_t len, len2; 2164 int v3 = (nfsd->nd_flag & ND_NFSV3); 2165 struct mbuf *mb, *mreq; 2166 struct vnode *dirp = (struct vnode *)0; 2167 nfsrvfh_t nsfh; 2168 u_quad_t frev; 2169 2170 nd.ni_cnd.cn_nameiop = 0; 2171 nfsm_srvmtofh(&nsfh); 2172 nfsm_srvnamesiz(len); 2173 nd.ni_cnd.cn_cred = cred; 2174 nd.ni_cnd.cn_nameiop = CREATE; 2175 nd.ni_cnd.cn_flags = LOCKPARENT; 2176 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 2177 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2178 if (dirp && v3) { 2179 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred); 2180 } 2181 if (error) 2182 goto out; 2183 VATTR_NULL(&va); 2184 va.va_type = VLNK; 2185 if (v3) { 2186 va.va_mode = 0; 2187 nfsm_srvsattr(&va); 2188 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 2189 len2 = fxdr_unsigned(uint32_t, *tl); 2190 if (len2 > PATH_MAX) { 2191 /* XXX should check _PC_NO_TRUNC */ 2192 error = ENAMETOOLONG; 2193 goto abortop; 2194 } 2195 } 2196 else { 2197 /* NFSv2 */ 2198 nfsm_strsiz(len2, NFS_MAXPATHLEN); 2199 } 2200 pathcp = malloc(len2 + 1, M_TEMP, M_WAITOK); 2201 iv.iov_base = pathcp; 2202 iv.iov_len = len2; 2203 io.uio_resid = len2; 2204 io.uio_offset = 0; 2205 io.uio_iov = &iv; 2206 io.uio_iovcnt = 1; 2207 io.uio_rw = UIO_READ; 2208 UIO_SETUP_SYSSPACE(&io); 2209 nfsm_mtouio(&io, len2); 2210 if (!v3) { 2211 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 2212 va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode); 2213 } 2214 *(pathcp + len2) = '\0'; 2215 if (nd.ni_vp) { 2216 error = EEXIST; 2217 abortop: 2218 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2219 if (nd.ni_dvp == nd.ni_vp) 2220 vrele(nd.ni_dvp); 2221 else 2222 vput(nd.ni_dvp); 2223 if (nd.ni_vp) 2224 vrele(nd.ni_vp); 2225 goto out; 2226 } 2227 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2228 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp); 2229 if (!error) { 2230 if (v3) { 2231 error = nfsrv_composefh(nd.ni_vp, &nsfh, v3); 2232 if (!error) 2233 error = VOP_GETATTR(nd.ni_vp, &va, cred); 2234 vput(nd.ni_vp); 2235 } else { 2236 vput(nd.ni_vp); 2237 } 2238 } 2239 out: 2240 if (pathcp) 2241 free(pathcp, M_TEMP); 2242 if (dirp) { 2243 if (v3) { 2244 diraft_ret = VOP_GETATTR(dirp, &diraft, cred); 2245 } 2246 vrele(dirp); 2247 } 2248 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_POSTOPATTR(v3) + 2249 NFSX_WCCDATA(v3)); 2250 if (v3) { 2251 if (!error) { 2252 nfsm_srvpostop_fh(&nsfh); 2253 nfsm_srvpostop_attr(0, &va); 2254 } 2255 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2256 } 2257 return (0); 2258 nfsmout: 2259 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2260 if (nd.ni_dvp == nd.ni_vp) 2261 vrele(nd.ni_dvp); 2262 else 2263 vput(nd.ni_dvp); 2264 if (nd.ni_vp) 2265 vrele(nd.ni_vp); 2266 if (dirp) 2267 vrele(dirp); 2268 if (pathcp) 2269 free(pathcp, M_TEMP); 2270 return (error); 2271 } 2272 2273 /* 2274 * nfs mkdir service 2275 */ 2276 int 2277 nfsrv_mkdir(nfsd, slp, lwp, mrq) 2278 struct nfsrv_descript *nfsd; 2279 struct nfssvc_sock *slp; 2280 struct lwp *lwp; 2281 struct mbuf **mrq; 2282 { 2283 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2284 struct mbuf *nam = nfsd->nd_nam; 2285 char *dpos = nfsd->nd_dpos; 2286 kauth_cred_t cred = nfsd->nd_cr; 2287 struct vattr va, dirfor, diraft; 2288 struct nfs_fattr *fp; 2289 struct nameidata nd; 2290 char *cp; 2291 u_int32_t *tl; 2292 int32_t t1; 2293 char *bpos; 2294 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 2295 int v3 = (nfsd->nd_flag & ND_NFSV3); 2296 char *cp2; 2297 struct mbuf *mb, *mreq; 2298 struct vnode *vp, *dirp = (struct vnode *)0; 2299 nfsrvfh_t nsfh; 2300 u_quad_t frev; 2301 2302 nfsm_srvmtofh(&nsfh); 2303 nfsm_srvnamesiz(len); 2304 nd.ni_cnd.cn_cred = cred; 2305 nd.ni_cnd.cn_nameiop = CREATE; 2306 nd.ni_cnd.cn_flags = LOCKPARENT; 2307 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 2308 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2309 if (dirp && v3) { 2310 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred); 2311 } 2312 if (error) { 2313 nfsm_reply(NFSX_WCCDATA(v3)); 2314 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2315 if (dirp) 2316 vrele(dirp); 2317 return (0); 2318 } 2319 VATTR_NULL(&va); 2320 if (v3) { 2321 va.va_mode = 0; 2322 nfsm_srvsattr(&va); 2323 } else { 2324 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 2325 va.va_mode = nfstov_mode(*tl++); 2326 } 2327 va.va_type = VDIR; 2328 vp = nd.ni_vp; 2329 if (vp != NULL) { 2330 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2331 if (nd.ni_dvp == vp) 2332 vrele(nd.ni_dvp); 2333 else 2334 vput(nd.ni_dvp); 2335 vrele(vp); 2336 error = EEXIST; 2337 goto out; 2338 } 2339 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2340 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 2341 if (!error) { 2342 vp = nd.ni_vp; 2343 error = nfsrv_composefh(vp, &nsfh, v3); 2344 if (!error) 2345 error = VOP_GETATTR(vp, &va, cred); 2346 vput(vp); 2347 } 2348 out: 2349 if (dirp) { 2350 if (v3) { 2351 diraft_ret = VOP_GETATTR(dirp, &diraft, cred); 2352 } 2353 vrele(dirp); 2354 } 2355 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_POSTOPATTR(v3) + 2356 NFSX_WCCDATA(v3)); 2357 if (v3) { 2358 if (!error) { 2359 nfsm_srvpostop_fh(&nsfh); 2360 nfsm_srvpostop_attr(0, &va); 2361 } 2362 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2363 } else { 2364 nfsm_srvfhtom(&nsfh, v3); 2365 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 2366 nfsm_srvfillattr(&va, fp); 2367 } 2368 return (0); 2369 nfsmout: 2370 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2371 if (nd.ni_dvp == nd.ni_vp) 2372 vrele(nd.ni_dvp); 2373 else 2374 vput(nd.ni_dvp); 2375 if (nd.ni_vp) 2376 vrele(nd.ni_vp); 2377 if (dirp) 2378 vrele(dirp); 2379 return (error); 2380 } 2381 2382 /* 2383 * nfs rmdir service 2384 */ 2385 int 2386 nfsrv_rmdir(nfsd, slp, lwp, mrq) 2387 struct nfsrv_descript *nfsd; 2388 struct nfssvc_sock *slp; 2389 struct lwp *lwp; 2390 struct mbuf **mrq; 2391 { 2392 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2393 struct mbuf *nam = nfsd->nd_nam; 2394 char *dpos = nfsd->nd_dpos; 2395 kauth_cred_t cred = nfsd->nd_cr; 2396 u_int32_t *tl; 2397 int32_t t1; 2398 char *bpos; 2399 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 2400 int v3 = (nfsd->nd_flag & ND_NFSV3); 2401 char *cp2; 2402 struct mbuf *mb, *mreq; 2403 struct vnode *vp, *dirp = (struct vnode *)0; 2404 struct vattr dirfor, diraft; 2405 nfsrvfh_t nsfh; 2406 struct nameidata nd; 2407 u_quad_t frev; 2408 2409 nfsm_srvmtofh(&nsfh); 2410 nfsm_srvnamesiz(len); 2411 nd.ni_cnd.cn_cred = cred; 2412 nd.ni_cnd.cn_nameiop = DELETE; 2413 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 2414 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 2415 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2416 if (dirp && v3) { 2417 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred); 2418 } 2419 if (error) { 2420 nfsm_reply(NFSX_WCCDATA(v3)); 2421 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2422 if (dirp) 2423 vrele(dirp); 2424 return (0); 2425 } 2426 vp = nd.ni_vp; 2427 if (vp->v_type != VDIR) { 2428 error = ENOTDIR; 2429 goto out; 2430 } 2431 /* 2432 * No rmdir "." please. 2433 */ 2434 if (nd.ni_dvp == vp) { 2435 error = EINVAL; 2436 goto out; 2437 } 2438 /* 2439 * The root of a mounted filesystem cannot be deleted. 2440 */ 2441 if (vp->v_vflag & VV_ROOT) 2442 error = EBUSY; 2443 out: 2444 if (!error) { 2445 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2446 nqsrv_getl(vp, ND_WRITE); 2447 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2448 } else { 2449 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2450 if (nd.ni_dvp == nd.ni_vp) 2451 vrele(nd.ni_dvp); 2452 else 2453 vput(nd.ni_dvp); 2454 vput(vp); 2455 } 2456 if (dirp) { 2457 if (v3) { 2458 diraft_ret = VOP_GETATTR(dirp, &diraft, cred); 2459 } 2460 vrele(dirp); 2461 } 2462 nfsm_reply(NFSX_WCCDATA(v3)); 2463 if (v3) { 2464 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2465 return (0); 2466 } 2467 nfsm_srvdone; 2468 } 2469 2470 /* 2471 * nfs readdir service 2472 * - mallocs what it thinks is enough to read 2473 * count rounded up to a multiple of NFS_SRVDIRBLKSIZ <= NFS_MAXREADDIR 2474 * - calls VOP_READDIR() 2475 * - loops around building the reply 2476 * if the output generated exceeds count break out of loop 2477 * The nfsm_clget macro is used here so that the reply will be packed 2478 * tightly in mbuf clusters. 2479 * - it only knows that it has encountered eof when the VOP_READDIR() 2480 * reads nothing 2481 * - as such one readdir rpc will return eof false although you are there 2482 * and then the next will return eof 2483 * - it trims out records with d_fileno == 0 2484 * this doesn't matter for Unix clients, but they might confuse clients 2485 * for other os'. 2486 * - it trims out records with d_type == DT_WHT 2487 * these cannot be seen through NFS (unless we extend the protocol) 2488 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less 2489 * than requested, but this may not apply to all filesystems. For 2490 * example, client NFS does not { although it is never remote mounted 2491 * anyhow } 2492 * The alternate call nfsrv_readdirplus() does lookups as well. 2493 * PS: The NFS protocol spec. does not clarify what the "count" byte 2494 * argument is a count of.. just name strings and file id's or the 2495 * entire reply rpc or ... 2496 * I tried just file name and id sizes and it confused the Sun client, 2497 * so I am using the full rpc size now. The "paranoia.." comment refers 2498 * to including the status longwords that are not a part of the dir. 2499 * "entry" structures, but are in the rpc. 2500 */ 2501 2502 #define NFS_SRVDIRBLKSIZ 1024 2503 2504 struct flrep { 2505 nfsuint64 fl_off; 2506 u_int32_t fl_postopok; 2507 u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)]; 2508 u_int32_t fl_fhok; 2509 u_int32_t fl_fhsize; 2510 }; 2511 2512 int 2513 nfsrv_readdir(nfsd, slp, lwp, mrq) 2514 struct nfsrv_descript *nfsd; 2515 struct nfssvc_sock *slp; 2516 struct lwp *lwp; 2517 struct mbuf **mrq; 2518 { 2519 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2520 struct mbuf *nam = nfsd->nd_nam; 2521 char *dpos = nfsd->nd_dpos; 2522 kauth_cred_t cred = nfsd->nd_cr; 2523 char *bp, *be; 2524 struct mbuf *mp; 2525 struct dirent *dp; 2526 char *cp; 2527 u_int32_t *tl; 2528 int32_t t1; 2529 char *bpos; 2530 struct mbuf *mb, *mreq, *mp2; 2531 char *cpos, *cend, *cp2, *rbuf; 2532 struct vnode *vp; 2533 struct vattr at; 2534 nfsrvfh_t nsfh; 2535 struct uio io; 2536 struct iovec iv; 2537 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2538 int siz, cnt, fullsiz, eofflag, rdonly, cache = 0, ncookies; 2539 int v3 = (nfsd->nd_flag & ND_NFSV3); 2540 u_quad_t frev, off, toff, verf; 2541 off_t *cookies = NULL, *cookiep; 2542 nfsuint64 jar; 2543 2544 nfsm_srvmtofh(&nsfh); 2545 if (v3) { 2546 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 2547 toff = fxdr_hyper(tl); 2548 tl += 2; 2549 verf = fxdr_hyper(tl); 2550 tl += 2; 2551 } else { 2552 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2553 toff = fxdr_unsigned(u_quad_t, *tl++); 2554 } 2555 off = toff; 2556 cnt = fxdr_unsigned(int, *tl); 2557 siz = ((cnt + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1)); 2558 xfer = NFS_SRVMAXDATA(nfsd); 2559 if (siz > xfer) 2560 siz = xfer; 2561 fullsiz = siz; 2562 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 2563 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 2564 if (!error && vp->v_type != VDIR) { 2565 error = ENOTDIR; 2566 vput(vp); 2567 } 2568 if (error) { 2569 nfsm_reply(NFSX_UNSIGNED); 2570 nfsm_srvpostop_attr(getret, &at); 2571 return (0); 2572 } 2573 nqsrv_getl(vp, ND_READ); 2574 if (v3) { 2575 error = getret = VOP_GETATTR(vp, &at, cred); 2576 #ifdef NFS3_STRICTVERF 2577 /* 2578 * XXX This check is too strict for Solaris 2.5 clients. 2579 */ 2580 if (!error && toff && verf != at.va_filerev) 2581 error = NFSERR_BAD_COOKIE; 2582 #endif 2583 } 2584 if (!error) 2585 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0); 2586 if (error) { 2587 vput(vp); 2588 nfsm_reply(NFSX_POSTOPATTR(v3)); 2589 nfsm_srvpostop_attr(getret, &at); 2590 return (0); 2591 } 2592 VOP_UNLOCK(vp, 0); 2593 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2594 again: 2595 iv.iov_base = rbuf; 2596 iv.iov_len = fullsiz; 2597 io.uio_iov = &iv; 2598 io.uio_iovcnt = 1; 2599 io.uio_offset = (off_t)off; 2600 io.uio_resid = fullsiz; 2601 io.uio_rw = UIO_READ; 2602 UIO_SETUP_SYSSPACE(&io); 2603 eofflag = 0; 2604 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2605 2606 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2607 2608 off = (off_t)io.uio_offset; 2609 if (!cookies && !error) 2610 error = NFSERR_PERM; 2611 if (v3) { 2612 getret = VOP_GETATTR(vp, &at, cred); 2613 if (!error) 2614 error = getret; 2615 } 2616 2617 VOP_UNLOCK(vp, 0); 2618 if (error) { 2619 vrele(vp); 2620 free((void *)rbuf, M_TEMP); 2621 if (cookies) 2622 free((void *)cookies, M_TEMP); 2623 nfsm_reply(NFSX_POSTOPATTR(v3)); 2624 nfsm_srvpostop_attr(getret, &at); 2625 return (0); 2626 } 2627 if (io.uio_resid) { 2628 siz -= io.uio_resid; 2629 2630 /* 2631 * If nothing read, return eof 2632 * rpc reply 2633 */ 2634 if (siz == 0) { 2635 vrele(vp); 2636 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + 2637 2 * NFSX_UNSIGNED); 2638 if (v3) { 2639 nfsm_srvpostop_attr(getret, &at); 2640 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2641 txdr_hyper(at.va_filerev, tl); 2642 tl += 2; 2643 } else 2644 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2645 *tl++ = nfs_false; 2646 *tl = nfs_true; 2647 free((void *)rbuf, M_TEMP); 2648 free((void *)cookies, M_TEMP); 2649 return (0); 2650 } 2651 } 2652 2653 /* 2654 * Check for degenerate cases of nothing useful read. 2655 * If so go try again 2656 */ 2657 cpos = rbuf; 2658 cend = rbuf + siz; 2659 dp = (struct dirent *)cpos; 2660 cookiep = cookies; 2661 2662 while (cpos < cend && ncookies > 0 && 2663 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2664 cpos += dp->d_reclen; 2665 dp = (struct dirent *)cpos; 2666 cookiep++; 2667 ncookies--; 2668 } 2669 if (cpos >= cend || ncookies == 0) { 2670 toff = off; 2671 siz = fullsiz; 2672 free(cookies, M_TEMP); 2673 cookies = NULL; 2674 goto again; 2675 } 2676 2677 len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */ 2678 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz); 2679 if (v3) { 2680 nfsm_srvpostop_attr(getret, &at); 2681 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2682 txdr_hyper(at.va_filerev, tl); 2683 } 2684 mp = mp2 = mb; 2685 bp = bpos; 2686 be = bp + M_TRAILINGSPACE(mp); 2687 2688 /* Loop through the records and build reply */ 2689 while (cpos < cend && ncookies > 0) { 2690 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 2691 nlen = dp->d_namlen; 2692 rem = nfsm_rndup(nlen)-nlen; 2693 len += (4 * NFSX_UNSIGNED + nlen + rem); 2694 if (v3) 2695 len += 2 * NFSX_UNSIGNED; 2696 if (len > cnt) { 2697 eofflag = 0; 2698 break; 2699 } 2700 /* 2701 * Build the directory record xdr from 2702 * the dirent entry. 2703 */ 2704 nfsm_clget; 2705 *tl = nfs_true; 2706 bp += NFSX_UNSIGNED; 2707 if (v3) { 2708 nfsm_clget; 2709 *tl = txdr_unsigned(dp->d_fileno >> 32); 2710 bp += NFSX_UNSIGNED; 2711 } 2712 nfsm_clget; 2713 *tl = txdr_unsigned(dp->d_fileno); 2714 bp += NFSX_UNSIGNED; 2715 nfsm_clget; 2716 *tl = txdr_unsigned(nlen); 2717 bp += NFSX_UNSIGNED; 2718 2719 /* And loop around copying the name */ 2720 xfer = nlen; 2721 cp = dp->d_name; 2722 while (xfer > 0) { 2723 nfsm_clget; 2724 if ((bp+xfer) > be) 2725 tsiz = be-bp; 2726 else 2727 tsiz = xfer; 2728 memcpy(bp, cp, tsiz); 2729 bp += tsiz; 2730 xfer -= tsiz; 2731 if (xfer > 0) 2732 cp += tsiz; 2733 } 2734 /* And null pad to an int32_t boundary */ 2735 for (i = 0; i < rem; i++) 2736 *bp++ = '\0'; 2737 nfsm_clget; 2738 2739 /* Finish off the record */ 2740 txdr_hyper(*cookiep, &jar); 2741 if (v3) { 2742 *tl = jar.nfsuquad[0]; 2743 bp += NFSX_UNSIGNED; 2744 nfsm_clget; 2745 } 2746 *tl = jar.nfsuquad[1]; 2747 bp += NFSX_UNSIGNED; 2748 } 2749 cpos += dp->d_reclen; 2750 dp = (struct dirent *)cpos; 2751 cookiep++; 2752 ncookies--; 2753 } 2754 vrele(vp); 2755 nfsm_clget; 2756 *tl = nfs_false; 2757 bp += NFSX_UNSIGNED; 2758 nfsm_clget; 2759 if (eofflag) 2760 *tl = nfs_true; 2761 else 2762 *tl = nfs_false; 2763 bp += NFSX_UNSIGNED; 2764 if (mp != mb) { 2765 if (bp < be) 2766 mp->m_len = bp - mtod(mp, char *); 2767 } else 2768 mp->m_len += bp - bpos; 2769 free((void *)rbuf, M_TEMP); 2770 free((void *)cookies, M_TEMP); 2771 nfsm_srvdone; 2772 } 2773 2774 int 2775 nfsrv_readdirplus(nfsd, slp, lwp, mrq) 2776 struct nfsrv_descript *nfsd; 2777 struct nfssvc_sock *slp; 2778 struct lwp *lwp; 2779 struct mbuf **mrq; 2780 { 2781 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2782 struct mbuf *nam = nfsd->nd_nam; 2783 char *dpos = nfsd->nd_dpos; 2784 kauth_cred_t cred = nfsd->nd_cr; 2785 char *bp, *be; 2786 struct mbuf *mp; 2787 struct dirent *dp; 2788 char *cp; 2789 u_int32_t *tl; 2790 int32_t t1; 2791 char *bpos; 2792 struct mbuf *mb, *mreq, *mp2; 2793 char *cpos, *cend, *cp2, *rbuf; 2794 struct vnode *vp, *nvp; 2795 struct flrep fl; 2796 nfsrvfh_t nsfh; 2797 struct uio io; 2798 struct iovec iv; 2799 struct vattr va, at, *vap = &va; 2800 struct nfs_fattr *fp; 2801 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2802 int siz, cnt, fullsiz, eofflag, rdonly, cache = 0, dirlen, ncookies; 2803 u_quad_t frev, off, toff, verf; 2804 off_t *cookies = NULL, *cookiep; 2805 2806 nfsm_srvmtofh(&nsfh); 2807 nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED); 2808 toff = fxdr_hyper(tl); 2809 tl += 2; 2810 verf = fxdr_hyper(tl); 2811 tl += 2; 2812 siz = fxdr_unsigned(int, *tl++); 2813 cnt = fxdr_unsigned(int, *tl); 2814 off = toff; 2815 siz = ((siz + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1)); 2816 xfer = NFS_SRVMAXDATA(nfsd); 2817 if (siz > xfer) 2818 siz = xfer; 2819 fullsiz = siz; 2820 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 2821 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 2822 if (!error && vp->v_type != VDIR) { 2823 error = ENOTDIR; 2824 vput(vp); 2825 } 2826 if (error) { 2827 nfsm_reply(NFSX_UNSIGNED); 2828 nfsm_srvpostop_attr(getret, &at); 2829 return (0); 2830 } 2831 error = getret = VOP_GETATTR(vp, &at, cred); 2832 #ifdef NFS3_STRICTVERF 2833 /* 2834 * XXX This check is too strict for Solaris 2.5 clients. 2835 */ 2836 if (!error && toff && verf != at.va_filerev) 2837 error = NFSERR_BAD_COOKIE; 2838 #endif 2839 if (!error) { 2840 nqsrv_getl(vp, ND_READ); 2841 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0); 2842 } 2843 if (error) { 2844 vput(vp); 2845 nfsm_reply(NFSX_V3POSTOPATTR); 2846 nfsm_srvpostop_attr(getret, &at); 2847 return (0); 2848 } 2849 VOP_UNLOCK(vp, 0); 2850 2851 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2852 again: 2853 iv.iov_base = rbuf; 2854 iv.iov_len = fullsiz; 2855 io.uio_iov = &iv; 2856 io.uio_iovcnt = 1; 2857 io.uio_offset = (off_t)off; 2858 io.uio_resid = fullsiz; 2859 io.uio_rw = UIO_READ; 2860 UIO_SETUP_SYSSPACE(&io); 2861 eofflag = 0; 2862 2863 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2864 2865 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2866 2867 off = (u_quad_t)io.uio_offset; 2868 getret = VOP_GETATTR(vp, &at, cred); 2869 2870 VOP_UNLOCK(vp, 0); 2871 2872 /* 2873 * If the VGET operation doesn't work for this filesystem, 2874 * we can't support readdirplus. Returning NOTSUPP should 2875 * make clients fall back to plain readdir. 2876 * There's no need to check for VPTOFH as well, we wouldn't 2877 * even be here otherwise. 2878 */ 2879 if (!getret) { 2880 if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp))) 2881 getret = (getret == EOPNOTSUPP) ? 2882 NFSERR_NOTSUPP : NFSERR_IO; 2883 else 2884 vput(nvp); 2885 } 2886 2887 if (!cookies && !error) 2888 error = NFSERR_PERM; 2889 if (!error) 2890 error = getret; 2891 if (error) { 2892 vrele(vp); 2893 if (cookies) 2894 free((void *)cookies, M_TEMP); 2895 free((void *)rbuf, M_TEMP); 2896 nfsm_reply(NFSX_V3POSTOPATTR); 2897 nfsm_srvpostop_attr(getret, &at); 2898 return (0); 2899 } 2900 if (io.uio_resid) { 2901 siz -= io.uio_resid; 2902 2903 /* 2904 * If nothing read, return eof 2905 * rpc reply 2906 */ 2907 if (siz == 0) { 2908 vrele(vp); 2909 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2910 2 * NFSX_UNSIGNED); 2911 nfsm_srvpostop_attr(getret, &at); 2912 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2913 txdr_hyper(at.va_filerev, tl); 2914 tl += 2; 2915 *tl++ = nfs_false; 2916 *tl = nfs_true; 2917 free((void *)cookies, M_TEMP); 2918 free((void *)rbuf, M_TEMP); 2919 return (0); 2920 } 2921 } 2922 2923 /* 2924 * Check for degenerate cases of nothing useful read. 2925 * If so go try again 2926 */ 2927 cpos = rbuf; 2928 cend = rbuf + siz; 2929 dp = (struct dirent *)cpos; 2930 cookiep = cookies; 2931 2932 while (cpos < cend && ncookies > 0 && 2933 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2934 cpos += dp->d_reclen; 2935 dp = (struct dirent *)cpos; 2936 cookiep++; 2937 ncookies--; 2938 } 2939 if (cpos >= cend || ncookies == 0) { 2940 toff = off; 2941 siz = fullsiz; 2942 free(cookies, M_TEMP); 2943 cookies = NULL; 2944 goto again; 2945 } 2946 2947 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED; 2948 nfsm_reply(cnt); 2949 nfsm_srvpostop_attr(getret, &at); 2950 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2951 txdr_hyper(at.va_filerev, tl); 2952 mp = mp2 = mb; 2953 bp = bpos; 2954 be = bp + M_TRAILINGSPACE(mp); 2955 2956 /* Loop through the records and build reply */ 2957 while (cpos < cend && ncookies > 0) { 2958 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 2959 nfsrvfh_t nnsfh; 2960 2961 nlen = dp->d_namlen; 2962 rem = nfsm_rndup(nlen)-nlen; 2963 2964 /* 2965 * For readdir_and_lookup get the vnode using 2966 * the file number. 2967 */ 2968 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp)) 2969 goto invalid; 2970 if (nfsrv_composefh(nvp, &nnsfh, true)) { 2971 vput(nvp); 2972 goto invalid; 2973 } 2974 if (VOP_GETATTR(nvp, vap, cred)) { 2975 vput(nvp); 2976 goto invalid; 2977 } 2978 vput(nvp); 2979 2980 /* 2981 * If either the dircount or maxcount will be 2982 * exceeded, get out now. Both of these lengths 2983 * are calculated conservatively, including all 2984 * XDR overheads. 2985 */ 2986 len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH + 2987 NFSX_V3POSTOPATTR); 2988 dirlen += (6 * NFSX_UNSIGNED + nlen + rem); 2989 if (len > cnt || dirlen > fullsiz) { 2990 eofflag = 0; 2991 break; 2992 } 2993 2994 /* 2995 * Build the directory record xdr from 2996 * the dirent entry. 2997 */ 2998 fp = (struct nfs_fattr *)&fl.fl_fattr; 2999 nfsm_srvfillattr(vap, fp); 3000 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH); 3001 fl.fl_fhok = nfs_true; 3002 fl.fl_postopok = nfs_true; 3003 txdr_hyper(*cookiep, fl.fl_off.nfsuquad); 3004 3005 nfsm_clget; 3006 *tl = nfs_true; 3007 bp += NFSX_UNSIGNED; 3008 nfsm_clget; 3009 *tl = txdr_unsigned(dp->d_fileno >> 32); 3010 bp += NFSX_UNSIGNED; 3011 nfsm_clget; 3012 *tl = txdr_unsigned(dp->d_fileno); 3013 bp += NFSX_UNSIGNED; 3014 nfsm_clget; 3015 *tl = txdr_unsigned(nlen); 3016 bp += NFSX_UNSIGNED; 3017 3018 /* And loop around copying the name */ 3019 xfer = nlen; 3020 cp = dp->d_name; 3021 while (xfer > 0) { 3022 nfsm_clget; 3023 if ((bp + xfer) > be) 3024 tsiz = be - bp; 3025 else 3026 tsiz = xfer; 3027 memcpy(bp, cp, tsiz); 3028 bp += tsiz; 3029 xfer -= tsiz; 3030 if (xfer > 0) 3031 cp += tsiz; 3032 } 3033 /* And null pad to an int32_t boundary */ 3034 for (i = 0; i < rem; i++) 3035 *bp++ = '\0'; 3036 3037 /* 3038 * Now copy the flrep structure out. 3039 */ 3040 xfer = sizeof(struct flrep); 3041 cp = (void *)&fl; 3042 while (xfer > 0) { 3043 nfsm_clget; 3044 if ((bp + xfer) > be) 3045 tsiz = be - bp; 3046 else 3047 tsiz = xfer; 3048 memcpy(bp, cp, tsiz); 3049 bp += tsiz; 3050 xfer -= tsiz; 3051 if (xfer > 0) 3052 cp += tsiz; 3053 } 3054 3055 /* 3056 * ... and filehandle. 3057 */ 3058 xfer = NFSRVFH_SIZE(&nnsfh); 3059 cp = NFSRVFH_DATA(&nnsfh); 3060 while (xfer > 0) { 3061 nfsm_clget; 3062 if ((bp + xfer) > be) 3063 tsiz = be - bp; 3064 else 3065 tsiz = xfer; 3066 memcpy(bp, cp, tsiz); 3067 bp += tsiz; 3068 xfer -= tsiz; 3069 if (xfer > 0) 3070 cp += tsiz; 3071 } 3072 } 3073 invalid: 3074 cpos += dp->d_reclen; 3075 dp = (struct dirent *)cpos; 3076 cookiep++; 3077 ncookies--; 3078 } 3079 vrele(vp); 3080 nfsm_clget; 3081 *tl = nfs_false; 3082 bp += NFSX_UNSIGNED; 3083 nfsm_clget; 3084 if (eofflag) 3085 *tl = nfs_true; 3086 else 3087 *tl = nfs_false; 3088 bp += NFSX_UNSIGNED; 3089 if (mp != mb) { 3090 if (bp < be) 3091 mp->m_len = bp - mtod(mp, char *); 3092 } else 3093 mp->m_len += bp - bpos; 3094 free((void *)cookies, M_TEMP); 3095 free((void *)rbuf, M_TEMP); 3096 nfsm_srvdone; 3097 } 3098 3099 /* 3100 * nfs commit service 3101 */ 3102 int 3103 nfsrv_commit(nfsd, slp, lwp, mrq) 3104 struct nfsrv_descript *nfsd; 3105 struct nfssvc_sock *slp; 3106 struct lwp *lwp; 3107 struct mbuf **mrq; 3108 { 3109 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3110 struct mbuf *nam = nfsd->nd_nam; 3111 char *dpos = nfsd->nd_dpos; 3112 kauth_cred_t cred = nfsd->nd_cr; 3113 struct vattr bfor, aft; 3114 struct vnode *vp; 3115 nfsrvfh_t nsfh; 3116 u_int32_t *tl; 3117 int32_t t1; 3118 char *bpos; 3119 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cache = 0; 3120 uint32_t cnt; 3121 char *cp2; 3122 struct mbuf *mb, *mreq; 3123 u_quad_t frev, off, end; 3124 3125 nfsm_srvmtofh(&nsfh); 3126 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED); 3127 3128 off = fxdr_hyper(tl); 3129 tl += 2; 3130 cnt = fxdr_unsigned(uint32_t, *tl); 3131 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3132 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3133 if (error) { 3134 nfsm_reply(2 * NFSX_UNSIGNED); 3135 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3136 return (0); 3137 } 3138 for_ret = VOP_GETATTR(vp, &bfor, cred); 3139 end = (cnt > 0) ? off + cnt : vp->v_size; 3140 if (end < off || end > vp->v_size) 3141 end = vp->v_size; 3142 if (off < vp->v_size) 3143 error = VOP_FSYNC(vp, cred, FSYNC_WAIT, off, end); 3144 /* else error == 0, from nfsrv_fhtovp() */ 3145 aft_ret = VOP_GETATTR(vp, &aft, cred); 3146 vput(vp); 3147 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF); 3148 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3149 if (!error) { 3150 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF); 3151 *tl++ = txdr_unsigned(boottime.tv_sec); 3152 *tl = txdr_unsigned(boottime.tv_usec); 3153 } else { 3154 return (0); 3155 } 3156 nfsm_srvdone; 3157 } 3158 3159 /* 3160 * nfs statfs service 3161 */ 3162 int 3163 nfsrv_statfs(nfsd, slp, lwp, mrq) 3164 struct nfsrv_descript *nfsd; 3165 struct nfssvc_sock *slp; 3166 struct lwp *lwp; 3167 struct mbuf **mrq; 3168 { 3169 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3170 struct mbuf *nam = nfsd->nd_nam; 3171 char *dpos = nfsd->nd_dpos; 3172 kauth_cred_t cred = nfsd->nd_cr; 3173 struct statvfs *sf = NULL; 3174 struct nfs_statfs *sfp; 3175 u_int32_t *tl; 3176 int32_t t1; 3177 char *bpos; 3178 int error = 0, rdonly, cache = 0, getret = 1; 3179 int v3 = (nfsd->nd_flag & ND_NFSV3); 3180 char *cp2; 3181 struct mbuf *mb, *mreq; 3182 struct vnode *vp; 3183 struct vattr at; 3184 nfsrvfh_t nsfh; 3185 u_quad_t frev, tval; 3186 3187 nfsm_srvmtofh(&nsfh); 3188 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3189 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3190 if (error) { 3191 nfsm_reply(NFSX_UNSIGNED); 3192 nfsm_srvpostop_attr(getret, &at); 3193 return (0); 3194 } 3195 sf = malloc(sizeof(*sf), M_TEMP, M_WAITOK); 3196 error = VFS_STATVFS(vp->v_mount, sf); 3197 getret = VOP_GETATTR(vp, &at, cred); 3198 vput(vp); 3199 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3)); 3200 if (v3) 3201 nfsm_srvpostop_attr(getret, &at); 3202 if (error) { 3203 free(sf, M_TEMP); 3204 return (0); 3205 } 3206 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3)); 3207 if (v3) { 3208 tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_frsize); 3209 txdr_hyper(tval, &sfp->sf_tbytes); 3210 tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_frsize); 3211 txdr_hyper(tval, &sfp->sf_fbytes); 3212 tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_frsize); 3213 txdr_hyper(tval, &sfp->sf_abytes); 3214 tval = (u_quad_t)sf->f_files; 3215 txdr_hyper(tval, &sfp->sf_tfiles); 3216 tval = (u_quad_t)sf->f_ffree; 3217 txdr_hyper(tval, &sfp->sf_ffiles); 3218 txdr_hyper(tval, &sfp->sf_afiles); 3219 sfp->sf_invarsec = 0; 3220 } else { 3221 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA); 3222 sfp->sf_bsize = txdr_unsigned(sf->f_frsize); 3223 sfp->sf_blocks = txdr_unsigned(sf->f_blocks); 3224 sfp->sf_bfree = txdr_unsigned(sf->f_bfree); 3225 sfp->sf_bavail = txdr_unsigned(sf->f_bavail); 3226 } 3227 nfsmout: 3228 if (sf) 3229 free(sf, M_TEMP); 3230 return error; 3231 } 3232 3233 /* 3234 * nfs fsinfo service 3235 */ 3236 int 3237 nfsrv_fsinfo(nfsd, slp, lwp, mrq) 3238 struct nfsrv_descript *nfsd; 3239 struct nfssvc_sock *slp; 3240 struct lwp *lwp; 3241 struct mbuf **mrq; 3242 { 3243 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3244 struct mbuf *nam = nfsd->nd_nam; 3245 char *dpos = nfsd->nd_dpos; 3246 kauth_cred_t cred = nfsd->nd_cr; 3247 u_int32_t *tl; 3248 struct nfsv3_fsinfo *sip; 3249 int32_t t1; 3250 char *bpos; 3251 int error = 0, rdonly, cache = 0, getret = 1; 3252 uint32_t maxdata; 3253 char *cp2; 3254 struct mbuf *mb, *mreq; 3255 struct vnode *vp; 3256 struct vattr at; 3257 nfsrvfh_t nsfh; 3258 u_quad_t frev, maxfsize; 3259 struct statvfs *sb; 3260 3261 nfsm_srvmtofh(&nsfh); 3262 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3263 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3264 if (error) { 3265 nfsm_reply(NFSX_UNSIGNED); 3266 nfsm_srvpostop_attr(getret, &at); 3267 return (0); 3268 } 3269 3270 /* XXX Try to make a guess on the max file size. */ 3271 sb = malloc(sizeof(*sb), M_TEMP, M_WAITOK); 3272 VFS_STATVFS(vp->v_mount, sb); 3273 maxfsize = (u_quad_t)0x80000000 * sb->f_frsize - 1; 3274 free(sb, M_TEMP); 3275 3276 getret = VOP_GETATTR(vp, &at, cred); 3277 vput(vp); 3278 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO); 3279 nfsm_srvpostop_attr(getret, &at); 3280 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO); 3281 3282 /* 3283 * XXX 3284 * There should be file system VFS OP(s) to get this information. 3285 * For now, assume ufs. 3286 */ 3287 if (slp->ns_so->so_type == SOCK_DGRAM) 3288 maxdata = NFS_MAXDGRAMDATA; 3289 else 3290 maxdata = NFS_MAXDATA; 3291 sip->fs_rtmax = txdr_unsigned(maxdata); 3292 sip->fs_rtpref = txdr_unsigned(maxdata); 3293 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE); 3294 sip->fs_wtmax = txdr_unsigned(maxdata); 3295 sip->fs_wtpref = txdr_unsigned(maxdata); 3296 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE); 3297 sip->fs_dtpref = txdr_unsigned(maxdata); 3298 txdr_hyper(maxfsize, &sip->fs_maxfilesize); 3299 sip->fs_timedelta.nfsv3_sec = 0; 3300 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1); 3301 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK | 3302 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS | 3303 NFSV3FSINFO_CANSETTIME); 3304 nfsm_srvdone; 3305 } 3306 3307 /* 3308 * nfs pathconf service 3309 */ 3310 int 3311 nfsrv_pathconf(nfsd, slp, lwp, mrq) 3312 struct nfsrv_descript *nfsd; 3313 struct nfssvc_sock *slp; 3314 struct lwp *lwp; 3315 struct mbuf **mrq; 3316 { 3317 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3318 struct mbuf *nam = nfsd->nd_nam; 3319 char *dpos = nfsd->nd_dpos; 3320 kauth_cred_t cred = nfsd->nd_cr; 3321 u_int32_t *tl; 3322 struct nfsv3_pathconf *pc; 3323 int32_t t1; 3324 char *bpos; 3325 int error = 0, rdonly, cache = 0, getret = 1; 3326 register_t linkmax, namemax, chownres, notrunc; 3327 char *cp2; 3328 struct mbuf *mb, *mreq; 3329 struct vnode *vp; 3330 struct vattr at; 3331 nfsrvfh_t nsfh; 3332 u_quad_t frev; 3333 3334 nfsm_srvmtofh(&nsfh); 3335 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3336 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3337 if (error) { 3338 nfsm_reply(NFSX_UNSIGNED); 3339 nfsm_srvpostop_attr(getret, &at); 3340 return (0); 3341 } 3342 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax); 3343 if (!error) 3344 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax); 3345 if (!error) 3346 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres); 3347 if (!error) 3348 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc); 3349 getret = VOP_GETATTR(vp, &at, cred); 3350 vput(vp); 3351 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF); 3352 nfsm_srvpostop_attr(getret, &at); 3353 if (error) 3354 return (0); 3355 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF); 3356 3357 pc->pc_linkmax = txdr_unsigned(linkmax); 3358 pc->pc_namemax = txdr_unsigned(namemax); 3359 pc->pc_notrunc = txdr_unsigned(notrunc); 3360 pc->pc_chownrestricted = txdr_unsigned(chownres); 3361 3362 /* 3363 * These should probably be supported by VOP_PATHCONF(), but 3364 * until msdosfs is exportable (why would you want to?), the 3365 * Unix defaults should be ok. 3366 */ 3367 pc->pc_caseinsensitive = nfs_false; 3368 pc->pc_casepreserving = nfs_true; 3369 nfsm_srvdone; 3370 } 3371 3372 /* 3373 * Null operation, used by clients to ping server 3374 */ 3375 /* ARGSUSED */ 3376 int 3377 nfsrv_null(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, 3378 struct lwp *lwp, struct mbuf **mrq) 3379 { 3380 struct mbuf *mrep = nfsd->nd_mrep; 3381 char *bpos; 3382 int error = NFSERR_RETVOID, cache = 0; 3383 struct mbuf *mb, *mreq; 3384 u_quad_t frev; 3385 3386 nfsm_reply(0); 3387 return (0); 3388 } 3389 3390 /* 3391 * No operation, used for obsolete procedures 3392 */ 3393 /* ARGSUSED */ 3394 int 3395 nfsrv_noop(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, 3396 struct lwp *lwp, struct mbuf **mrq) 3397 { 3398 struct mbuf *mrep = nfsd->nd_mrep; 3399 char *bpos; 3400 int error, cache = 0; 3401 struct mbuf *mb, *mreq; 3402 u_quad_t frev; 3403 3404 if (nfsd->nd_repstat) 3405 error = nfsd->nd_repstat; 3406 else 3407 error = EPROCUNAVAIL; 3408 nfsm_reply(0); 3409 return (0); 3410 } 3411 3412 /* 3413 * Perform access checking for vnodes obtained from file handles that would 3414 * refer to files already opened by a Unix client. You cannot just use 3415 * vn_writechk() and VOP_ACCESS() for two reasons. 3416 * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case 3417 * 2 - The owner is to be given access irrespective of mode bits for some 3418 * operations, so that processes that chmod after opening a file don't 3419 * break. I don't like this because it opens a security hole, but since 3420 * the nfs server opens a security hole the size of a barn door anyhow, 3421 * what the heck. 3422 * 3423 * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS() 3424 * will return EPERM instead of EACCESS. EPERM is always an error. 3425 */ 3426 int 3427 nfsrv_access(vp, flags, cred, rdonly, lwp, override) 3428 struct vnode *vp; 3429 int flags; 3430 kauth_cred_t cred; 3431 int rdonly; 3432 struct lwp *lwp; 3433 int override; 3434 { 3435 struct vattr vattr; 3436 int error; 3437 if (flags & VWRITE) { 3438 /* Just vn_writechk() changed to check rdonly */ 3439 /* 3440 * Disallow write attempts on read-only file systems; 3441 * unless the file is a socket or a block or character 3442 * device resident on the file system. 3443 */ 3444 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 3445 switch (vp->v_type) { 3446 case VREG: 3447 case VDIR: 3448 case VLNK: 3449 return (EROFS); 3450 default: 3451 break; 3452 } 3453 } 3454 3455 /* 3456 * If the vnode is in use as a process's text, 3457 * we can't allow writing. 3458 */ 3459 if (vp->v_iflag & VI_TEXT) 3460 return (ETXTBSY); 3461 } 3462 error = VOP_GETATTR(vp, &vattr, cred); 3463 if (error) 3464 return (error); 3465 error = VOP_ACCESS(vp, flags, cred); 3466 /* 3467 * Allow certain operations for the owner (reads and writes 3468 * on files that are already open). 3469 */ 3470 if (override && error == EACCES && kauth_cred_geteuid(cred) == vattr.va_uid) 3471 error = 0; 3472 return error; 3473 } 3474