nfs_serv.c revision 1.95.2.8
1 /* $NetBSD: nfs_serv.c,v 1.95.2.8 2008/01/21 09:47:33 yamt Exp $ */ 2 3 /* 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * This code is derived from software contributed to Berkeley by 8 * Rick Macklem at The University of Guelph. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * @(#)nfs_serv.c 8.8 (Berkeley) 7/31/95 35 */ 36 37 /* 38 * nfs version 2 and 3 server calls to vnode ops 39 * - these routines generally have 3 phases 40 * 1 - break down and validate rpc request in mbuf list 41 * 2 - do the vnode ops for the request 42 * (surprisingly ?? many are very similar to syscalls in vfs_syscalls.c) 43 * 3 - build the rpc reply in an mbuf list 44 * nb: 45 * - do not mix the phases, since the nfsm_?? macros can return failures 46 * on a bad rpc or similar and do not do any vrele() or vput()'s 47 * 48 * - the nfsm_reply() macro generates an nfs rpc reply with the nfs 49 * error number iff error != 0 whereas 50 * returning an error from the server function implies a fatal error 51 * such as a badly constructed rpc request that should be dropped without 52 * a reply. 53 * For Version 3, nfsm_reply() does not return for the error case, since 54 * most version 3 rpcs return more than the status for error cases. 55 */ 56 57 #include <sys/cdefs.h> 58 __KERNEL_RCSID(0, "$NetBSD: nfs_serv.c,v 1.95.2.8 2008/01/21 09:47:33 yamt Exp $"); 59 60 #include <sys/param.h> 61 #include <sys/systm.h> 62 #include <sys/proc.h> 63 #include <sys/file.h> 64 #include <sys/namei.h> 65 #include <sys/vnode.h> 66 #include <sys/mount.h> 67 #include <sys/socket.h> 68 #include <sys/socketvar.h> 69 #include <sys/mbuf.h> 70 #include <sys/dirent.h> 71 #include <sys/stat.h> 72 #include <sys/kernel.h> 73 #include <sys/hash.h> 74 #include <sys/kauth.h> 75 76 #include <uvm/uvm.h> 77 78 #include <nfs/nfsproto.h> 79 #include <nfs/rpcv2.h> 80 #include <nfs/nfs.h> 81 #include <nfs/xdr_subs.h> 82 #include <nfs/nfsm_subs.h> 83 #include <nfs/nfs_var.h> 84 85 /* Global vars */ 86 extern u_int32_t nfs_xdrneg1; 87 extern u_int32_t nfs_false, nfs_true; 88 extern enum vtype nv3tov_type[8]; 89 extern struct nfsstats nfsstats; 90 extern const nfstype nfsv2_type[9]; 91 extern const nfstype nfsv3_type[9]; 92 int nfsrvw_procrastinate = NFS_GATHERDELAY * 1000; 93 int nfsd_use_loan = 1; /* use page-loan for READ OP */ 94 95 #define nqsrv_getl(vp, rw) /* nothing */ 96 97 /* 98 * nfs v3 access service 99 */ 100 int 101 nfsrv3_access(nfsd, slp, lwp, mrq) 102 struct nfsrv_descript *nfsd; 103 struct nfssvc_sock *slp; 104 struct lwp *lwp; 105 struct mbuf **mrq; 106 { 107 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 108 struct mbuf *nam = nfsd->nd_nam; 109 char *dpos = nfsd->nd_dpos; 110 kauth_cred_t cred = nfsd->nd_cr; 111 struct vnode *vp; 112 nfsrvfh_t nsfh; 113 u_int32_t *tl; 114 int32_t t1; 115 char *bpos; 116 int error = 0, rdonly, cache = 0, getret; 117 char *cp2; 118 struct mbuf *mb, *mreq; 119 struct vattr va; 120 u_long inmode, testmode, outmode; 121 u_quad_t frev; 122 123 nfsm_srvmtofh(&nsfh); 124 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 125 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, &rdonly, 126 (nfsd->nd_flag & ND_KERBAUTH), false); 127 if (error) { 128 nfsm_reply(NFSX_UNSIGNED); 129 nfsm_srvpostop_attr(1, (struct vattr *)0); 130 return (0); 131 } 132 inmode = fxdr_unsigned(u_int32_t, *tl); 133 outmode = 0; 134 if ((inmode & NFSV3ACCESS_READ) && 135 nfsrv_access(vp, VREAD, cred, rdonly, lwp, 0) == 0) 136 outmode |= NFSV3ACCESS_READ; 137 if (vp->v_type != VDIR) { 138 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND); 139 if (testmode && 140 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0) 141 outmode |= testmode; 142 if ((inmode & NFSV3ACCESS_EXECUTE) && 143 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0) 144 outmode |= NFSV3ACCESS_EXECUTE; 145 } else { 146 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND | 147 NFSV3ACCESS_DELETE); 148 if (testmode && 149 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0) 150 outmode |= testmode; 151 if ((inmode & NFSV3ACCESS_LOOKUP) && 152 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0) 153 outmode |= NFSV3ACCESS_LOOKUP; 154 } 155 getret = VOP_GETATTR(vp, &va, cred); 156 vput(vp); 157 nfsm_reply(NFSX_POSTOPATTR(1) + NFSX_UNSIGNED); 158 nfsm_srvpostop_attr(getret, &va); 159 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 160 *tl = txdr_unsigned(outmode); 161 nfsm_srvdone; 162 } 163 164 /* 165 * nfs getattr service 166 */ 167 int 168 nfsrv_getattr(nfsd, slp, lwp, mrq) 169 struct nfsrv_descript *nfsd; 170 struct nfssvc_sock *slp; 171 struct lwp *lwp; 172 struct mbuf **mrq; 173 { 174 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 175 struct mbuf *nam = nfsd->nd_nam; 176 char *dpos = nfsd->nd_dpos; 177 kauth_cred_t cred = nfsd->nd_cr; 178 struct nfs_fattr *fp; 179 struct vattr va; 180 struct vnode *vp; 181 nfsrvfh_t nsfh; 182 u_int32_t *tl; 183 int32_t t1; 184 char *bpos; 185 int error = 0, rdonly, cache = 0; 186 char *cp2; 187 struct mbuf *mb, *mreq; 188 u_quad_t frev; 189 190 nfsm_srvmtofh(&nsfh); 191 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, &rdonly, 192 (nfsd->nd_flag & ND_KERBAUTH), false); 193 if (error) { 194 nfsm_reply(0); 195 return (0); 196 } 197 nqsrv_getl(vp, ND_READ); 198 error = VOP_GETATTR(vp, &va, cred); 199 vput(vp); 200 nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); 201 if (error) 202 return (0); 203 nfsm_build(fp, struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); 204 nfsm_srvfillattr(&va, fp); 205 nfsm_srvdone; 206 } 207 208 /* 209 * nfs setattr service 210 */ 211 int 212 nfsrv_setattr(nfsd, slp, lwp, mrq) 213 struct nfsrv_descript *nfsd; 214 struct nfssvc_sock *slp; 215 struct lwp *lwp; 216 struct mbuf **mrq; 217 { 218 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 219 struct mbuf *nam = nfsd->nd_nam; 220 char *dpos = nfsd->nd_dpos; 221 kauth_cred_t cred = nfsd->nd_cr; 222 struct vattr va, preat; 223 struct nfsv2_sattr *sp; 224 struct nfs_fattr *fp; 225 struct vnode *vp; 226 nfsrvfh_t nsfh; 227 u_int32_t *tl; 228 int32_t t1; 229 char *bpos; 230 int error = 0, rdonly, cache = 0, preat_ret = 1, postat_ret = 1; 231 int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0; 232 char *cp2; 233 struct mbuf *mb, *mreq; 234 u_quad_t frev; 235 struct timespec guard; 236 237 memset(&guard, 0, sizeof guard); /* XXX gcc */ 238 239 nfsm_srvmtofh(&nsfh); 240 VATTR_NULL(&va); 241 if (v3) { 242 nfsm_srvsattr(&va); 243 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 244 gcheck = fxdr_unsigned(int, *tl); 245 if (gcheck) { 246 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 247 fxdr_nfsv3time(tl, &guard); 248 } 249 } else { 250 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 251 /* 252 * Nah nah nah nah na nah 253 * There is a bug in the Sun client that puts 0xffff in the mode 254 * field of sattr when it should put in 0xffffffff. The u_short 255 * doesn't sign extend. 256 * --> check the low order 2 bytes for 0xffff 257 */ 258 if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff) 259 va.va_mode = nfstov_mode(sp->sa_mode); 260 if (sp->sa_uid != nfs_xdrneg1) 261 va.va_uid = fxdr_unsigned(uid_t, sp->sa_uid); 262 if (sp->sa_gid != nfs_xdrneg1) 263 va.va_gid = fxdr_unsigned(gid_t, sp->sa_gid); 264 if (sp->sa_size != nfs_xdrneg1) 265 va.va_size = fxdr_unsigned(u_quad_t, sp->sa_size); 266 if (sp->sa_atime.nfsv2_sec != nfs_xdrneg1) { 267 #ifdef notyet 268 fxdr_nfsv2time(&sp->sa_atime, &va.va_atime); 269 #else 270 va.va_atime.tv_sec = 271 fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec); 272 va.va_atime.tv_nsec = 0; 273 #endif 274 } 275 if (sp->sa_mtime.nfsv2_sec != nfs_xdrneg1) 276 fxdr_nfsv2time(&sp->sa_mtime, &va.va_mtime); 277 278 } 279 280 /* 281 * Now that we have all the fields, lets do it. 282 */ 283 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, &rdonly, 284 (nfsd->nd_flag & ND_KERBAUTH), false); 285 if (error) { 286 nfsm_reply(2 * NFSX_UNSIGNED); 287 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 288 return (0); 289 } 290 nqsrv_getl(vp, ND_WRITE); 291 if (v3) { 292 error = preat_ret = VOP_GETATTR(vp, &preat, cred); 293 if (!error && gcheck && 294 (preat.va_ctime.tv_sec != guard.tv_sec || 295 preat.va_ctime.tv_nsec != guard.tv_nsec)) 296 error = NFSERR_NOT_SYNC; 297 if (error) { 298 vput(vp); 299 nfsm_reply(NFSX_WCCDATA(v3)); 300 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 301 return (0); 302 } 303 } 304 305 /* 306 * If the size is being changed write acces is required, otherwise 307 * just check for a read only file system. 308 */ 309 if (va.va_size == ((u_quad_t)((quad_t) -1))) { 310 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 311 error = EROFS; 312 goto out; 313 } 314 } else { 315 if (vp->v_type == VDIR) { 316 error = EISDIR; 317 goto out; 318 } else if ((error = nfsrv_access(vp, VWRITE, cred, rdonly, 319 lwp, 0)) != 0) 320 goto out; 321 } 322 error = VOP_SETATTR(vp, &va, cred); 323 postat_ret = VOP_GETATTR(vp, &va, cred); 324 if (!error) 325 error = postat_ret; 326 out: 327 vput(vp); 328 nfsm_reply(NFSX_WCCORFATTR(v3)); 329 if (v3) { 330 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va); 331 return (0); 332 } else { 333 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 334 nfsm_srvfillattr(&va, fp); 335 } 336 nfsm_srvdone; 337 } 338 339 /* 340 * nfs lookup rpc 341 */ 342 int 343 nfsrv_lookup(nfsd, slp, lwp, mrq) 344 struct nfsrv_descript *nfsd; 345 struct nfssvc_sock *slp; 346 struct lwp *lwp; 347 struct mbuf **mrq; 348 { 349 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 350 struct mbuf *nam = nfsd->nd_nam; 351 char *dpos = nfsd->nd_dpos; 352 kauth_cred_t cred = nfsd->nd_cr; 353 struct nfs_fattr *fp; 354 struct nameidata nd, ind, *ndp = &nd; 355 struct vnode *vp, *dirp; 356 nfsrvfh_t nsfh; 357 char *cp; 358 u_int32_t *tl; 359 int32_t t1; 360 char *bpos; 361 int error = 0, cache = 0, dirattr_ret = 1; 362 uint32_t len; 363 int v3 = (nfsd->nd_flag & ND_NFSV3), pubflag; 364 char *cp2; 365 struct mbuf *mb, *mreq; 366 struct vattr va, dirattr; 367 u_quad_t frev; 368 369 nfsm_srvmtofh(&nsfh); 370 nfsm_srvnamesiz(len); 371 372 pubflag = nfs_ispublicfh(&nsfh); 373 374 nd.ni_cnd.cn_cred = cred; 375 nd.ni_cnd.cn_nameiop = LOOKUP; 376 nd.ni_cnd.cn_flags = LOCKLEAF | SAVESTART; 377 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 378 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), pubflag); 379 380 if (!error && pubflag) { 381 if (nd.ni_vp->v_type == VDIR && nfs_pub.np_index != NULL) { 382 /* 383 * Setup call to lookup() to see if we can find 384 * the index file. Arguably, this doesn't belong 385 * in a kernel.. Ugh. 386 */ 387 ind = nd; 388 VOP_UNLOCK(nd.ni_vp, 0); 389 ind.ni_pathlen = strlen(nfs_pub.np_index); 390 ind.ni_cnd.cn_nameptr = ind.ni_cnd.cn_pnbuf = 391 nfs_pub.np_index; 392 ind.ni_startdir = nd.ni_vp; 393 VREF(ind.ni_startdir); 394 error = lookup(&ind); 395 if (!error) { 396 /* 397 * Found an index file. Get rid of 398 * the old references. 399 */ 400 if (dirp) 401 vrele(dirp); 402 dirp = nd.ni_vp; 403 vrele(nd.ni_startdir); 404 ndp = &ind; 405 } else 406 error = 0; 407 } 408 /* 409 * If the public filehandle was used, check that this lookup 410 * didn't result in a filehandle outside the publicly exported 411 * filesystem. 412 */ 413 414 if (!error && ndp->ni_vp->v_mount != nfs_pub.np_mount) { 415 vput(nd.ni_vp); 416 error = EPERM; 417 } 418 } 419 420 if (dirp) { 421 if (v3) 422 dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred); 423 vrele(dirp); 424 } 425 426 if (error) { 427 nfsm_reply(NFSX_POSTOPATTR(v3)); 428 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 429 return (0); 430 } 431 432 nqsrv_getl(ndp->ni_startdir, ND_READ); 433 PNBUF_PUT(nd.ni_cnd.cn_pnbuf); 434 vp = ndp->ni_vp; 435 error = nfsrv_composefh(vp, &nsfh, v3); 436 if (!error) 437 error = VOP_GETATTR(vp, &va, cred); 438 vput(vp); 439 vrele(ndp->ni_startdir); 440 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_POSTOPORFATTR(v3) + 441 NFSX_POSTOPATTR(v3)); 442 if (error) { 443 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 444 return (0); 445 } 446 nfsm_srvfhtom(&nsfh, v3); 447 if (v3) { 448 nfsm_srvpostop_attr(0, &va); 449 nfsm_srvpostop_attr(dirattr_ret, &dirattr); 450 } else { 451 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 452 nfsm_srvfillattr(&va, fp); 453 } 454 nfsm_srvdone; 455 } 456 457 /* 458 * nfs readlink service 459 */ 460 int 461 nfsrv_readlink(nfsd, slp, lwp, mrq) 462 struct nfsrv_descript *nfsd; 463 struct nfssvc_sock *slp; 464 struct lwp *lwp; 465 struct mbuf **mrq; 466 { 467 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 468 struct mbuf *nam = nfsd->nd_nam; 469 char *dpos = nfsd->nd_dpos; 470 kauth_cred_t cred = nfsd->nd_cr; 471 struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN]; 472 struct iovec *ivp = iv; 473 struct mbuf *mp; 474 u_int32_t *tl; 475 int32_t t1; 476 char *bpos; 477 int error = 0, rdonly, cache = 0, i, padlen, getret; 478 uint32_t len; 479 int v3 = (nfsd->nd_flag & ND_NFSV3); 480 char *cp2; 481 struct mbuf *mb, *mp2 = NULL, *mp3 = NULL, *mreq; 482 struct vnode *vp; 483 struct vattr attr; 484 nfsrvfh_t nsfh; 485 struct uio io, *uiop = &io; 486 u_quad_t frev; 487 488 nfsm_srvmtofh(&nsfh); 489 len = 0; 490 i = 0; 491 while (len < NFS_MAXPATHLEN) { 492 mp = m_get(M_WAIT, MT_DATA); 493 MCLAIM(mp, &nfs_mowner); 494 m_clget(mp, M_WAIT); 495 mp->m_len = NFSMSIZ(mp); 496 if (len == 0) 497 mp3 = mp2 = mp; 498 else { 499 mp2->m_next = mp; 500 mp2 = mp; 501 } 502 if ((len+mp->m_len) > NFS_MAXPATHLEN) { 503 mp->m_len = NFS_MAXPATHLEN-len; 504 len = NFS_MAXPATHLEN; 505 } else 506 len += mp->m_len; 507 ivp->iov_base = mtod(mp, void *); 508 ivp->iov_len = mp->m_len; 509 i++; 510 ivp++; 511 } 512 uiop->uio_iov = iv; 513 uiop->uio_iovcnt = i; 514 uiop->uio_offset = 0; 515 uiop->uio_resid = len; 516 uiop->uio_rw = UIO_READ; 517 UIO_SETUP_SYSSPACE(uiop); 518 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 519 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 520 if (error) { 521 m_freem(mp3); 522 nfsm_reply(2 * NFSX_UNSIGNED); 523 nfsm_srvpostop_attr(1, (struct vattr *)0); 524 return (0); 525 } 526 if (vp->v_type != VLNK) { 527 if (v3) 528 error = EINVAL; 529 else 530 error = ENXIO; 531 goto out; 532 } 533 nqsrv_getl(vp, ND_READ); 534 error = VOP_READLINK(vp, uiop, cred); 535 out: 536 getret = VOP_GETATTR(vp, &attr, cred); 537 vput(vp); 538 if (error) 539 m_freem(mp3); 540 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED); 541 if (v3) { 542 nfsm_srvpostop_attr(getret, &attr); 543 if (error) 544 return (0); 545 } 546 len -= uiop->uio_resid; 547 padlen = nfsm_padlen(len); 548 if (uiop->uio_resid || padlen) 549 nfs_zeropad(mp3, uiop->uio_resid, padlen); 550 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 551 *tl = txdr_unsigned(len); 552 mb->m_next = mp3; 553 nfsm_srvdone; 554 } 555 556 /* 557 * nfs read service 558 */ 559 int 560 nfsrv_read(nfsd, slp, lwp, mrq) 561 struct nfsrv_descript *nfsd; 562 struct nfssvc_sock *slp; 563 struct lwp *lwp; 564 struct mbuf **mrq; 565 { 566 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 567 struct mbuf *nam = nfsd->nd_nam; 568 char *dpos = nfsd->nd_dpos; 569 kauth_cred_t cred = nfsd->nd_cr; 570 struct mbuf *m; 571 struct nfs_fattr *fp; 572 u_int32_t *tl; 573 int32_t t1; 574 int i; 575 char *bpos; 576 int error = 0, rdonly, cache = 0, getret; 577 int v3 = (nfsd->nd_flag & ND_NFSV3); 578 uint32_t reqlen, len, cnt, left; 579 int padlen; 580 char *cp2; 581 struct mbuf *mb, *mreq; 582 struct vnode *vp; 583 nfsrvfh_t nsfh; 584 struct uio io, *uiop = &io; 585 struct vattr va; 586 off_t off; 587 u_quad_t frev; 588 589 nfsm_srvmtofh(&nsfh); 590 if (v3) { 591 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 592 off = fxdr_hyper(tl); 593 } else { 594 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 595 off = (off_t)fxdr_unsigned(u_int32_t, *tl); 596 } 597 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 598 reqlen = fxdr_unsigned(uint32_t, *tl); 599 reqlen = MIN(reqlen, NFS_SRVMAXDATA(nfsd)); 600 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 601 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 602 if (error) { 603 nfsm_reply(2 * NFSX_UNSIGNED); 604 nfsm_srvpostop_attr(1, (struct vattr *)0); 605 return (0); 606 } 607 if (vp->v_type != VREG) { 608 if (v3) 609 error = EINVAL; 610 else 611 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 612 } 613 if (!error) { 614 nqsrv_getl(vp, ND_READ); 615 if ((error = nfsrv_access(vp, VREAD, cred, rdonly, lwp, 1)) != 0) 616 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 1); 617 } 618 getret = VOP_GETATTR(vp, &va, cred); 619 if (!error) 620 error = getret; 621 if (error) { 622 vput(vp); 623 nfsm_reply(NFSX_POSTOPATTR(v3)); 624 nfsm_srvpostop_attr(getret, &va); 625 return (0); 626 } 627 if (off >= va.va_size) 628 cnt = 0; 629 else if ((off + reqlen) > va.va_size) 630 cnt = va.va_size - off; 631 else 632 cnt = reqlen; 633 nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt)); 634 if (v3) { 635 nfsm_build(tl, u_int32_t *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED); 636 *tl++ = nfs_true; 637 fp = (struct nfs_fattr *)tl; 638 tl += (NFSX_V3FATTR / sizeof (u_int32_t)); 639 } else { 640 nfsm_build(tl, u_int32_t *, NFSX_V2FATTR + NFSX_UNSIGNED); 641 fp = (struct nfs_fattr *)tl; 642 tl += (NFSX_V2FATTR / sizeof (u_int32_t)); 643 } 644 len = left = cnt; 645 if (cnt > 0) { 646 if (nfsd_use_loan) { 647 struct vm_page **pgpp; 648 voff_t pgoff = trunc_page(off); 649 int npages; 650 vaddr_t lva; 651 652 npages = (round_page(off + cnt) - pgoff) >> PAGE_SHIFT; 653 KASSERT(npages <= M_EXT_MAXPAGES); /* XXX */ 654 655 /* allocate kva for mbuf data */ 656 lva = sokvaalloc(npages << PAGE_SHIFT, slp->ns_so); 657 if (lva == 0) { 658 /* fall back to VOP_READ */ 659 goto loan_fail; 660 } 661 662 /* allocate mbuf */ 663 m = m_get(M_WAIT, MT_DATA); 664 MCLAIM(m, &nfs_mowner); 665 pgpp = m->m_ext.ext_pgs; 666 667 /* loan pages */ 668 error = uvm_loanuobjpages(&vp->v_uobj, pgoff, npages, 669 pgpp); 670 if (error) { 671 sokvafree(lva, npages << PAGE_SHIFT); 672 m_free(m); 673 if (error == EBUSY) 674 goto loan_fail; 675 goto read_error; 676 } 677 678 /* associate kva to mbuf */ 679 MEXTADD(m, (void *)(lva + ((vaddr_t)off & PAGE_MASK)), 680 cnt, M_MBUF, soloanfree, slp->ns_so); 681 m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP; 682 m->m_len = cnt; 683 684 #if defined(__HAVE_LAZY_MBUF) 685 m->m_flags |= M_EXT_LAZY; 686 m->m_ext.ext_flags |= M_EXT_LAZY; 687 #else /* defined(__HAVE_LAZY_MBUF) */ 688 /* map pages */ 689 for (i = 0; i < npages; i++) { 690 pmap_kenter_pa(lva, VM_PAGE_TO_PHYS(pgpp[i]), 691 VM_PROT_READ); 692 lva += PAGE_SIZE; 693 } 694 695 pmap_update(pmap_kernel()); 696 #endif /* defined(__HAVE_LAZY_MBUF) */ 697 698 mb->m_next = m; 699 mb = m; 700 error = 0; 701 uiop->uio_resid = 0; 702 } else { 703 struct iovec *iv; 704 struct iovec *iv2; 705 struct mbuf *m2; 706 int siz; 707 loan_fail: 708 /* 709 * Generate the mbuf list with the uio_iov ref. to it. 710 */ 711 i = 0; 712 m = m2 = mb; 713 while (left > 0) { 714 siz = min(M_TRAILINGSPACE(m), left); 715 if (siz > 0) { 716 left -= siz; 717 i++; 718 } 719 if (left > 0) { 720 m = m_get(M_WAIT, MT_DATA); 721 MCLAIM(m, &nfs_mowner); 722 m_clget(m, M_WAIT); 723 m->m_len = 0; 724 m2->m_next = m; 725 m2 = m; 726 } 727 } 728 iv = malloc(i * sizeof(struct iovec), M_TEMP, M_WAITOK); 729 uiop->uio_iov = iv2 = iv; 730 m = mb; 731 left = cnt; 732 i = 0; 733 while (left > 0) { 734 if (m == NULL) 735 panic("nfsrv_read iov"); 736 siz = min(M_TRAILINGSPACE(m), left); 737 if (siz > 0) { 738 iv->iov_base = mtod(m, char *) + 739 m->m_len; 740 iv->iov_len = siz; 741 m->m_len += siz; 742 left -= siz; 743 iv++; 744 i++; 745 } 746 m = m->m_next; 747 } 748 uiop->uio_iovcnt = i; 749 uiop->uio_offset = off; 750 uiop->uio_resid = cnt; 751 uiop->uio_rw = UIO_READ; 752 UIO_SETUP_SYSSPACE(uiop); 753 error = VOP_READ(vp, uiop, IO_NODELOCKED, cred); 754 free((void *)iv2, M_TEMP); 755 } 756 read_error: 757 if (error || (getret = VOP_GETATTR(vp, &va, cred)) != 0){ 758 if (!error) 759 error = getret; 760 m_freem(mreq); 761 vput(vp); 762 nfsm_reply(NFSX_POSTOPATTR(v3)); 763 nfsm_srvpostop_attr(getret, &va); 764 return (0); 765 } 766 } else { 767 uiop->uio_resid = 0; 768 } 769 vput(vp); 770 nfsm_srvfillattr(&va, fp); 771 len -= uiop->uio_resid; 772 padlen = nfsm_padlen(len); 773 if (uiop->uio_resid || padlen) 774 nfs_zeropad(mb, uiop->uio_resid, padlen); 775 if (v3) { 776 /* count */ 777 *tl++ = txdr_unsigned(len); 778 /* eof */ 779 if (off + len >= va.va_size) 780 *tl++ = nfs_true; 781 else 782 *tl++ = nfs_false; 783 } 784 *tl = txdr_unsigned(len); 785 nfsm_srvdone; 786 } 787 788 /* 789 * nfs write service 790 */ 791 int 792 nfsrv_write(nfsd, slp, lwp, mrq) 793 struct nfsrv_descript *nfsd; 794 struct nfssvc_sock *slp; 795 struct lwp *lwp; 796 struct mbuf **mrq; 797 { 798 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 799 struct mbuf *nam = nfsd->nd_nam; 800 char *dpos = nfsd->nd_dpos; 801 kauth_cred_t cred = nfsd->nd_cr; 802 struct iovec *ivp; 803 int i, cnt; 804 struct mbuf *mp; 805 struct nfs_fattr *fp; 806 struct iovec *iv; 807 struct vattr va, forat; 808 u_int32_t *tl; 809 int32_t t1; 810 char *bpos; 811 int error = 0, rdonly, cache = 0, len, forat_ret = 1; 812 int ioflags, aftat_ret = 1, retlen, zeroing, adjust; 813 int stable = NFSV3WRITE_FILESYNC; 814 int v3 = (nfsd->nd_flag & ND_NFSV3); 815 char *cp2; 816 struct mbuf *mb, *mreq; 817 struct vnode *vp; 818 nfsrvfh_t nsfh; 819 struct uio io, *uiop = &io; 820 off_t off; 821 u_quad_t frev; 822 823 if (mrep == NULL) { 824 *mrq = NULL; 825 return (0); 826 } 827 nfsm_srvmtofh(&nsfh); 828 if (v3) { 829 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 830 off = fxdr_hyper(tl); 831 tl += 3; 832 stable = fxdr_unsigned(int, *tl++); 833 } else { 834 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 835 off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 836 tl += 2; 837 } 838 retlen = len = fxdr_unsigned(int32_t, *tl); 839 cnt = i = 0; 840 841 /* 842 * For NFS Version 2, it is not obvious what a write of zero length 843 * should do, but I might as well be consistent with Version 3, 844 * which is to return ok so long as there are no permission problems. 845 */ 846 if (len > 0) { 847 zeroing = 1; 848 mp = mrep; 849 while (mp) { 850 if (mp == md) { 851 zeroing = 0; 852 adjust = dpos - mtod(mp, char *); 853 mp->m_len -= adjust; 854 if (mp->m_len > 0 && adjust > 0) 855 NFSMADV(mp, adjust); 856 } 857 if (zeroing) 858 mp->m_len = 0; 859 else if (mp->m_len > 0) { 860 i += mp->m_len; 861 if (i > len) { 862 mp->m_len -= (i - len); 863 zeroing = 1; 864 } 865 if (mp->m_len > 0) 866 cnt++; 867 } 868 mp = mp->m_next; 869 } 870 } 871 if (len > NFS_MAXDATA || len < 0 || i < len) { 872 error = EIO; 873 nfsm_reply(2 * NFSX_UNSIGNED); 874 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 875 return (0); 876 } 877 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 878 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 879 if (error) { 880 nfsm_reply(2 * NFSX_UNSIGNED); 881 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 882 return (0); 883 } 884 if (v3) 885 forat_ret = VOP_GETATTR(vp, &forat, cred); 886 if (vp->v_type != VREG) { 887 if (v3) 888 error = EINVAL; 889 else 890 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 891 } 892 if (!error) { 893 nqsrv_getl(vp, ND_WRITE); 894 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1); 895 } 896 if (error) { 897 vput(vp); 898 nfsm_reply(NFSX_WCCDATA(v3)); 899 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 900 return (0); 901 } 902 903 if (len > 0) { 904 ivp = malloc(cnt * sizeof (struct iovec), M_TEMP, M_WAITOK); 905 uiop->uio_iov = iv = ivp; 906 uiop->uio_iovcnt = cnt; 907 mp = mrep; 908 while (mp) { 909 if (mp->m_len > 0) { 910 ivp->iov_base = mtod(mp, void *); 911 ivp->iov_len = mp->m_len; 912 ivp++; 913 } 914 mp = mp->m_next; 915 } 916 917 /* 918 * XXX 919 * The IO_METASYNC flag indicates that all metadata (and not 920 * just enough to ensure data integrity) must be written to 921 * stable storage synchronously. 922 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.) 923 */ 924 if (stable == NFSV3WRITE_UNSTABLE) 925 ioflags = IO_NODELOCKED; 926 else if (stable == NFSV3WRITE_DATASYNC) 927 ioflags = (IO_SYNC | IO_NODELOCKED); 928 else 929 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 930 uiop->uio_resid = len; 931 uiop->uio_rw = UIO_WRITE; 932 uiop->uio_offset = off; 933 UIO_SETUP_SYSSPACE(uiop); 934 error = VOP_WRITE(vp, uiop, ioflags, cred); 935 nfsstats.srvvop_writes++; 936 free(iv, M_TEMP); 937 } 938 aftat_ret = VOP_GETATTR(vp, &va, cred); 939 vput(vp); 940 if (!error) 941 error = aftat_ret; 942 nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) + 943 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3)); 944 if (v3) { 945 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 946 if (error) 947 return (0); 948 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 949 *tl++ = txdr_unsigned(retlen); 950 if (stable == NFSV3WRITE_UNSTABLE) 951 *tl++ = txdr_unsigned(stable); 952 else 953 *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC); 954 /* 955 * Actually, there is no need to txdr these fields, 956 * but it may make the values more human readable, 957 * for debugging purposes. 958 */ 959 *tl++ = txdr_unsigned(boottime.tv_sec); 960 *tl = txdr_unsigned(boottime.tv_usec); 961 } else { 962 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 963 nfsm_srvfillattr(&va, fp); 964 } 965 nfsm_srvdone; 966 } 967 968 /* 969 * XXX elad: the original NFSW_SAMECRED() macro also made sure the 970 * two nd_flag fields of the descriptors contained 971 * ND_KERBAUTH. 972 */ 973 static int 974 nfsrv_samecred(kauth_cred_t cred1, kauth_cred_t cred2) 975 { 976 int i, do_ngroups; 977 978 if (kauth_cred_geteuid(cred1) != kauth_cred_geteuid(cred2)) 979 return (0); 980 if (kauth_cred_ngroups(cred1) != kauth_cred_ngroups(cred2)) 981 return (0); 982 do_ngroups = kauth_cred_ngroups(cred1); 983 for (i = 0; i < do_ngroups; i++) 984 if (kauth_cred_group(cred1, i) != 985 kauth_cred_group(cred2, i)) 986 return (0); 987 988 return (1); 989 } 990 991 static struct nfsrvw_delayhash * 992 nfsrv_nwdelayhash(struct nfssvc_sock *slp, const nfsrvfh_t *nsfh) 993 { 994 uint32_t hash; 995 996 hash = hash32_buf(NFSRVFH_DATA(nsfh), NFSRVFH_SIZE(nsfh), 997 HASH32_BUF_INIT); 998 return &slp->ns_wdelayhashtbl[hash % NFS_WDELAYHASHSIZ]; 999 } 1000 1001 /* 1002 * NFS write service with write gathering support. Called when 1003 * nfsrvw_procrastinate > 0. 1004 * See: Chet Juszczak, "Improving the Write Performance of an NFS Server", 1005 * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco, 1006 * Jan. 1994. 1007 */ 1008 int 1009 nfsrv_writegather(ndp, slp, lwp, mrq) 1010 struct nfsrv_descript **ndp; 1011 struct nfssvc_sock *slp; 1012 struct lwp *lwp; 1013 struct mbuf **mrq; 1014 { 1015 struct timeval now; 1016 struct iovec *ivp; 1017 struct mbuf *mp; 1018 struct nfsrv_descript *wp, *nfsd, *owp, *swp; 1019 struct nfs_fattr *fp; 1020 int i = 0; 1021 struct iovec *iov; 1022 struct nfsrvw_delayhash *wpp; 1023 kauth_cred_t cred; 1024 struct vattr va, forat; 1025 u_int32_t *tl; 1026 int32_t t1; 1027 char *bpos, *dpos; 1028 int error = 0, rdonly, cache = 0, len = 0, forat_ret = 1; 1029 int ioflags, aftat_ret = 1, adjust, v3, zeroing; 1030 char *cp2; 1031 struct mbuf *mb, *mreq, *mrep, *md; 1032 struct vnode *vp; 1033 struct uio io, *uiop = &io; 1034 u_quad_t frev, cur_usec; 1035 1036 *mrq = NULL; 1037 if (*ndp) { 1038 nfsd = *ndp; 1039 *ndp = NULL; 1040 mrep = nfsd->nd_mrep; 1041 md = nfsd->nd_md; 1042 dpos = nfsd->nd_dpos; 1043 cred = nfsd->nd_cr; 1044 v3 = (nfsd->nd_flag & ND_NFSV3); 1045 LIST_INIT(&nfsd->nd_coalesce); 1046 nfsd->nd_mreq = NULL; 1047 nfsd->nd_stable = NFSV3WRITE_FILESYNC; 1048 getmicrotime(&now); 1049 cur_usec = (u_quad_t)now.tv_sec * 1000000 + (u_quad_t)now.tv_usec; 1050 nfsd->nd_time = cur_usec + nfsrvw_procrastinate; 1051 1052 /* 1053 * Now, get the write header.. 1054 */ 1055 nfsm_srvmtofh(&nfsd->nd_fh); 1056 if (v3) { 1057 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 1058 nfsd->nd_off = fxdr_hyper(tl); 1059 tl += 3; 1060 nfsd->nd_stable = fxdr_unsigned(int, *tl++); 1061 } else { 1062 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1063 nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl); 1064 tl += 2; 1065 } 1066 len = fxdr_unsigned(int32_t, *tl); 1067 nfsd->nd_len = len; 1068 nfsd->nd_eoff = nfsd->nd_off + len; 1069 1070 /* 1071 * Trim the header out of the mbuf list and trim off any trailing 1072 * junk so that the mbuf list has only the write data. 1073 */ 1074 zeroing = 1; 1075 i = 0; 1076 mp = mrep; 1077 while (mp) { 1078 if (mp == md) { 1079 zeroing = 0; 1080 adjust = dpos - mtod(mp, char *); 1081 mp->m_len -= adjust; 1082 if (mp->m_len > 0 && adjust > 0) 1083 NFSMADV(mp, adjust); 1084 } 1085 if (zeroing) 1086 mp->m_len = 0; 1087 else { 1088 i += mp->m_len; 1089 if (i > len) { 1090 mp->m_len -= (i - len); 1091 zeroing = 1; 1092 } 1093 } 1094 mp = mp->m_next; 1095 } 1096 if (len > NFS_MAXDATA || len < 0 || i < len) { 1097 nfsmout: 1098 m_freem(mrep); 1099 error = EIO; 1100 nfsm_writereply(2 * NFSX_UNSIGNED, v3); 1101 if (v3) 1102 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1103 nfsd->nd_mreq = mreq; 1104 nfsd->nd_mrep = NULL; 1105 nfsd->nd_time = 0; 1106 } 1107 1108 /* 1109 * Add this entry to the hash and time queues. 1110 */ 1111 owp = NULL; 1112 mutex_enter(&nfsd_lock); 1113 wp = LIST_FIRST(&slp->ns_tq); 1114 while (wp && wp->nd_time < nfsd->nd_time) { 1115 owp = wp; 1116 wp = LIST_NEXT(wp, nd_tq); 1117 } 1118 if (owp) { 1119 LIST_INSERT_AFTER(owp, nfsd, nd_tq); 1120 } else { 1121 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1122 } 1123 if (nfsd->nd_mrep) { 1124 wpp = nfsrv_nwdelayhash(slp, &nfsd->nd_fh); 1125 owp = NULL; 1126 wp = LIST_FIRST(wpp); 1127 while (wp && nfsrv_comparefh(&nfsd->nd_fh, &wp->nd_fh)) { 1128 owp = wp; 1129 wp = LIST_NEXT(wp, nd_hash); 1130 } 1131 while (wp && wp->nd_off < nfsd->nd_off && 1132 !nfsrv_comparefh(&nfsd->nd_fh, &wp->nd_fh)) { 1133 owp = wp; 1134 wp = LIST_NEXT(wp, nd_hash); 1135 } 1136 if (owp) { 1137 LIST_INSERT_AFTER(owp, nfsd, nd_hash); 1138 1139 /* 1140 * Search the hash list for overlapping entries and 1141 * coalesce. 1142 */ 1143 for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) { 1144 wp = LIST_NEXT(nfsd, nd_hash); 1145 if (nfsrv_samecred(owp->nd_cr, nfsd->nd_cr)) 1146 nfsrvw_coalesce(owp, nfsd); 1147 } 1148 } else { 1149 LIST_INSERT_HEAD(wpp, nfsd, nd_hash); 1150 } 1151 } 1152 mutex_exit(&nfsd_lock); 1153 } 1154 1155 /* 1156 * Now, do VOP_WRITE()s for any one(s) that need to be done now 1157 * and generate the associated reply mbuf list(s). 1158 */ 1159 loop1: 1160 getmicrotime(&now); 1161 cur_usec = (u_quad_t)now.tv_sec * 1000000 + (u_quad_t)now.tv_usec; 1162 mutex_enter(&nfsd_lock); 1163 for (nfsd = LIST_FIRST(&slp->ns_tq); nfsd; nfsd = owp) { 1164 owp = LIST_NEXT(nfsd, nd_tq); 1165 if (nfsd->nd_time > cur_usec) 1166 break; 1167 if (nfsd->nd_mreq) 1168 continue; 1169 LIST_REMOVE(nfsd, nd_tq); 1170 LIST_REMOVE(nfsd, nd_hash); 1171 mutex_exit(&nfsd_lock); 1172 1173 mrep = nfsd->nd_mrep; 1174 nfsd->nd_mrep = NULL; 1175 cred = nfsd->nd_cr; 1176 v3 = (nfsd->nd_flag & ND_NFSV3); 1177 forat_ret = aftat_ret = 1; 1178 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp, 1179 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH), 1180 false); 1181 if (!error) { 1182 if (v3) 1183 forat_ret = VOP_GETATTR(vp, &forat, cred); 1184 if (vp->v_type != VREG) { 1185 if (v3) 1186 error = EINVAL; 1187 else 1188 error = (vp->v_type == VDIR) ? EISDIR : EACCES; 1189 } 1190 } else 1191 vp = NULL; 1192 if (!error) { 1193 nqsrv_getl(vp, ND_WRITE); 1194 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1); 1195 } 1196 1197 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE) 1198 ioflags = IO_NODELOCKED; 1199 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC) 1200 ioflags = (IO_SYNC | IO_NODELOCKED); 1201 else 1202 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); 1203 uiop->uio_rw = UIO_WRITE; 1204 uiop->uio_offset = nfsd->nd_off; 1205 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off; 1206 UIO_SETUP_SYSSPACE(uiop); 1207 if (uiop->uio_resid > 0) { 1208 mp = mrep; 1209 i = 0; 1210 while (mp) { 1211 if (mp->m_len > 0) 1212 i++; 1213 mp = mp->m_next; 1214 } 1215 uiop->uio_iovcnt = i; 1216 iov = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK); 1217 uiop->uio_iov = ivp = iov; 1218 mp = mrep; 1219 while (mp) { 1220 if (mp->m_len > 0) { 1221 ivp->iov_base = mtod(mp, void *); 1222 ivp->iov_len = mp->m_len; 1223 ivp++; 1224 } 1225 mp = mp->m_next; 1226 } 1227 if (!error) { 1228 error = VOP_WRITE(vp, uiop, ioflags, cred); 1229 nfsstats.srvvop_writes++; 1230 } 1231 free((void *)iov, M_TEMP); 1232 } 1233 m_freem(mrep); 1234 if (vp) { 1235 aftat_ret = VOP_GETATTR(vp, &va, cred); 1236 vput(vp); 1237 } 1238 1239 /* 1240 * Loop around generating replies for all write rpcs that have 1241 * now been completed. 1242 */ 1243 swp = nfsd; 1244 do { 1245 if (error) { 1246 nfsm_writereply(NFSX_WCCDATA(v3), v3); 1247 if (v3) { 1248 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1249 } 1250 } else { 1251 nfsm_writereply(NFSX_PREOPATTR(v3) + 1252 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED + 1253 NFSX_WRITEVERF(v3), v3); 1254 if (v3) { 1255 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); 1256 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1257 *tl++ = txdr_unsigned(nfsd->nd_len); 1258 *tl++ = txdr_unsigned(swp->nd_stable); 1259 /* 1260 * Actually, there is no need to txdr these fields, 1261 * but it may make the values more human readable, 1262 * for debugging purposes. 1263 */ 1264 *tl++ = txdr_unsigned(boottime.tv_sec); 1265 *tl = txdr_unsigned(boottime.tv_usec); 1266 } else { 1267 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1268 nfsm_srvfillattr(&va, fp); 1269 } 1270 } 1271 nfsd->nd_mreq = mreq; 1272 if (nfsd->nd_mrep) 1273 panic("nfsrv_write: nd_mrep not free"); 1274 1275 /* 1276 * Done. Put it at the head of the timer queue so that 1277 * the final phase can return the reply. 1278 */ 1279 mutex_enter(&nfsd_lock); 1280 if (nfsd != swp) { 1281 nfsd->nd_time = 0; 1282 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); 1283 } 1284 nfsd = LIST_FIRST(&swp->nd_coalesce); 1285 if (nfsd) { 1286 LIST_REMOVE(nfsd, nd_tq); 1287 } 1288 mutex_exit(&nfsd_lock); 1289 } while (nfsd); 1290 swp->nd_time = 0; 1291 1292 mutex_enter(&nfsd_lock); 1293 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq); 1294 mutex_exit(&nfsd_lock); 1295 goto loop1; 1296 } 1297 mutex_exit(&nfsd_lock); 1298 nfs_timer_start(); 1299 1300 /* 1301 * Search for a reply to return. 1302 */ 1303 mutex_enter(&nfsd_lock); 1304 LIST_FOREACH(nfsd, &slp->ns_tq, nd_tq) { 1305 if (nfsd->nd_mreq) { 1306 LIST_REMOVE(nfsd, nd_tq); 1307 *mrq = nfsd->nd_mreq; 1308 *ndp = nfsd; 1309 break; 1310 } 1311 } 1312 mutex_exit(&nfsd_lock); 1313 return (0); 1314 } 1315 1316 /* 1317 * Coalesce the write request nfsd into owp. To do this we must: 1318 * - remove nfsd from the queues 1319 * - merge nfsd->nd_mrep into owp->nd_mrep 1320 * - update the nd_eoff and nd_stable for owp 1321 * - put nfsd on owp's nd_coalesce list 1322 * NB: Must be called at splsoftclock(). 1323 */ 1324 void 1325 nfsrvw_coalesce(owp, nfsd) 1326 struct nfsrv_descript *owp; 1327 struct nfsrv_descript *nfsd; 1328 { 1329 int overlap; 1330 struct mbuf *mp; 1331 struct nfsrv_descript *m; 1332 1333 KASSERT(mutex_owned(&nfsd_lock)); 1334 1335 LIST_REMOVE(nfsd, nd_hash); 1336 LIST_REMOVE(nfsd, nd_tq); 1337 if (owp->nd_eoff < nfsd->nd_eoff) { 1338 overlap = owp->nd_eoff - nfsd->nd_off; 1339 if (overlap < 0) 1340 panic("nfsrv_coalesce: bad off"); 1341 if (overlap > 0) 1342 m_adj(nfsd->nd_mrep, overlap); 1343 mp = owp->nd_mrep; 1344 while (mp->m_next) 1345 mp = mp->m_next; 1346 mp->m_next = nfsd->nd_mrep; 1347 owp->nd_eoff = nfsd->nd_eoff; 1348 } else 1349 m_freem(nfsd->nd_mrep); 1350 nfsd->nd_mrep = NULL; 1351 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC) 1352 owp->nd_stable = NFSV3WRITE_FILESYNC; 1353 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC && 1354 owp->nd_stable == NFSV3WRITE_UNSTABLE) 1355 owp->nd_stable = NFSV3WRITE_DATASYNC; 1356 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq); 1357 /* 1358 * nfsd might hold coalesce elements! Move them to owp. 1359 * Otherwise, requests may be lost and clients will be stuck. 1360 */ 1361 while ((m = LIST_FIRST(&nfsd->nd_coalesce)) != NULL) { 1362 LIST_REMOVE(m, nd_tq); 1363 LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq); 1364 } 1365 } 1366 1367 /* 1368 * nfs create service 1369 * now does a truncate to 0 length via. setattr if it already exists 1370 */ 1371 int 1372 nfsrv_create(nfsd, slp, lwp, mrq) 1373 struct nfsrv_descript *nfsd; 1374 struct nfssvc_sock *slp; 1375 struct lwp *lwp; 1376 struct mbuf **mrq; 1377 { 1378 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1379 struct mbuf *nam = nfsd->nd_nam; 1380 char *dpos = nfsd->nd_dpos; 1381 kauth_cred_t cred = nfsd->nd_cr; 1382 struct nfs_fattr *fp; 1383 struct vattr va, dirfor, diraft; 1384 struct nfsv2_sattr *sp; 1385 u_int32_t *tl; 1386 struct nameidata nd; 1387 char *cp; 1388 int32_t t1; 1389 char *bpos; 1390 int error = 0, cache = 0, len, tsize, dirfor_ret = 1, diraft_ret = 1; 1391 int rdev = 0; 1392 int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0; 1393 char *cp2; 1394 struct mbuf *mb, *mreq; 1395 struct vnode *vp = NULL, *dirp = NULL; 1396 nfsrvfh_t nsfh; 1397 u_quad_t frev, tempsize; 1398 u_char cverf[NFSX_V3CREATEVERF]; 1399 1400 nd.ni_cnd.cn_nameiop = 0; 1401 nfsm_srvmtofh(&nsfh); 1402 nfsm_srvnamesiz(len); 1403 nd.ni_cnd.cn_cred = cred; 1404 nd.ni_cnd.cn_nameiop = CREATE; 1405 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1406 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 1407 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1408 if (dirp && v3) { 1409 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred); 1410 } 1411 if (error) { 1412 nfsm_reply(NFSX_WCCDATA(v3)); 1413 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1414 if (dirp) 1415 vrele(dirp); 1416 return (0); 1417 } 1418 VATTR_NULL(&va); 1419 if (v3) { 1420 va.va_mode = 0; 1421 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1422 how = fxdr_unsigned(int, *tl); 1423 switch (how) { 1424 case NFSV3CREATE_GUARDED: 1425 if (nd.ni_vp) { 1426 error = EEXIST; 1427 break; 1428 } 1429 case NFSV3CREATE_UNCHECKED: 1430 nfsm_srvsattr(&va); 1431 break; 1432 case NFSV3CREATE_EXCLUSIVE: 1433 nfsm_dissect(cp, void *, NFSX_V3CREATEVERF); 1434 memcpy(cverf, cp, NFSX_V3CREATEVERF); 1435 exclusive_flag = 1; 1436 break; 1437 }; 1438 va.va_type = VREG; 1439 } else { 1440 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 1441 va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode)); 1442 if (va.va_type == VNON) 1443 va.va_type = VREG; 1444 va.va_mode = nfstov_mode(sp->sa_mode); 1445 switch (va.va_type) { 1446 case VREG: 1447 tsize = fxdr_unsigned(int32_t, sp->sa_size); 1448 if (tsize != -1) 1449 va.va_size = (u_quad_t)tsize; 1450 break; 1451 case VCHR: 1452 case VBLK: 1453 case VFIFO: 1454 rdev = fxdr_unsigned(int32_t, sp->sa_size); 1455 break; 1456 default: 1457 break; 1458 }; 1459 } 1460 1461 /* 1462 * Iff doesn't exist, create it 1463 * otherwise just truncate to 0 length 1464 * should I set the mode too ?? 1465 */ 1466 if (nd.ni_vp == NULL) { 1467 if (va.va_type == VREG || va.va_type == VSOCK) { 1468 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1469 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1470 if (!error) { 1471 if (exclusive_flag) { 1472 exclusive_flag = 0; 1473 VATTR_NULL(&va); 1474 /* 1475 * XXX 1476 * assuming NFSX_V3CREATEVERF 1477 * == sizeof(nfstime3) 1478 */ 1479 fxdr_nfsv3time(cverf, &va.va_atime); 1480 error = VOP_SETATTR(nd.ni_vp, &va, 1481 cred); 1482 } 1483 } 1484 } else if (va.va_type == VCHR || va.va_type == VBLK || 1485 va.va_type == VFIFO) { 1486 if (va.va_type == VCHR && rdev == 0xffffffff) 1487 va.va_type = VFIFO; 1488 if (va.va_type != VFIFO && 1489 (error = kauth_authorize_generic(cred, 1490 KAUTH_GENERIC_ISSUSER, NULL))) { 1491 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1492 vput(nd.ni_dvp); 1493 nfsm_reply(0); 1494 return (error); 1495 } else 1496 va.va_rdev = (dev_t)rdev; 1497 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1498 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, 1499 &va); 1500 if (error) { 1501 nfsm_reply(0); 1502 } 1503 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1504 vput(nd.ni_vp); 1505 vrele(nd.ni_dvp); 1506 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1507 error = EINVAL; 1508 nfsm_reply(0); 1509 } 1510 } else { 1511 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1512 vput(nd.ni_dvp); 1513 error = ENXIO; 1514 } 1515 vp = nd.ni_vp; 1516 } else { 1517 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1518 vp = nd.ni_vp; 1519 if (nd.ni_dvp == vp) 1520 vrele(nd.ni_dvp); 1521 else 1522 vput(nd.ni_dvp); 1523 if (!error && va.va_size != -1) { 1524 error = nfsrv_access(vp, VWRITE, cred, 1525 (nd.ni_cnd.cn_flags & RDONLY), lwp, 0); 1526 if (!error) { 1527 nqsrv_getl(vp, ND_WRITE); 1528 tempsize = va.va_size; 1529 VATTR_NULL(&va); 1530 va.va_size = tempsize; 1531 error = VOP_SETATTR(vp, &va, cred); 1532 } 1533 } 1534 if (error) 1535 vput(vp); 1536 } 1537 if (!error) { 1538 error = nfsrv_composefh(vp, &nsfh, v3); 1539 if (!error) 1540 error = VOP_GETATTR(vp, &va, cred); 1541 vput(vp); 1542 } 1543 if (v3) { 1544 if (exclusive_flag && !error) { 1545 /* 1546 * XXX assuming NFSX_V3CREATEVERF == sizeof(nfstime3) 1547 */ 1548 char oldverf[NFSX_V3CREATEVERF]; 1549 1550 txdr_nfsv3time(&va.va_atime, oldverf); 1551 if (memcmp(cverf, oldverf, NFSX_V3CREATEVERF)) 1552 error = EEXIST; 1553 } 1554 if (dirp) { 1555 diraft_ret = VOP_GETATTR(dirp, &diraft, cred); 1556 } 1557 } 1558 if (dirp) { 1559 vrele(dirp); 1560 } 1561 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3)); 1562 if (v3) { 1563 if (!error) { 1564 nfsm_srvpostop_fh(&nsfh); 1565 nfsm_srvpostop_attr(0, &va); 1566 } 1567 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1568 } else { 1569 nfsm_srvfhtom(&nsfh, v3); 1570 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 1571 nfsm_srvfillattr(&va, fp); 1572 } 1573 return (0); 1574 nfsmout: 1575 if (dirp) 1576 vrele(dirp); 1577 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1578 if (nd.ni_dvp == nd.ni_vp) 1579 vrele(nd.ni_dvp); 1580 else 1581 vput(nd.ni_dvp); 1582 if (nd.ni_vp) 1583 vput(nd.ni_vp); 1584 return (error); 1585 } 1586 1587 /* 1588 * nfs v3 mknod service 1589 */ 1590 int 1591 nfsrv_mknod(nfsd, slp, lwp, mrq) 1592 struct nfsrv_descript *nfsd; 1593 struct nfssvc_sock *slp; 1594 struct lwp *lwp; 1595 struct mbuf **mrq; 1596 { 1597 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1598 struct mbuf *nam = nfsd->nd_nam; 1599 char *dpos = nfsd->nd_dpos; 1600 kauth_cred_t cred = nfsd->nd_cr; 1601 struct vattr va, dirfor, diraft; 1602 u_int32_t *tl; 1603 struct nameidata nd; 1604 int32_t t1; 1605 char *bpos; 1606 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 1607 u_int32_t major, minor; 1608 enum vtype vtyp; 1609 char *cp2; 1610 struct mbuf *mb, *mreq; 1611 struct vnode *vp, *dirp = (struct vnode *)0; 1612 nfsrvfh_t nsfh; 1613 u_quad_t frev; 1614 1615 nd.ni_cnd.cn_nameiop = 0; 1616 nfsm_srvmtofh(&nsfh); 1617 nfsm_srvnamesiz(len); 1618 nd.ni_cnd.cn_cred = cred; 1619 nd.ni_cnd.cn_nameiop = CREATE; 1620 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1621 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 1622 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1623 if (dirp) 1624 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred); 1625 if (error) { 1626 nfsm_reply(NFSX_WCCDATA(1)); 1627 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1628 if (dirp) 1629 vrele(dirp); 1630 return (0); 1631 } 1632 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 1633 vtyp = nfsv3tov_type(*tl); 1634 if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) { 1635 error = NFSERR_BADTYPE; 1636 goto abort; 1637 } 1638 VATTR_NULL(&va); 1639 va.va_mode = 0; 1640 nfsm_srvsattr(&va); 1641 if (vtyp == VCHR || vtyp == VBLK) { 1642 dev_t rdev; 1643 1644 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1645 major = fxdr_unsigned(u_int32_t, *tl++); 1646 minor = fxdr_unsigned(u_int32_t, *tl); 1647 rdev = makedev(major, minor); 1648 if (major(rdev) != major || minor(rdev) != minor) { 1649 error = EINVAL; 1650 goto abort; 1651 } 1652 va.va_rdev = rdev; 1653 } 1654 1655 /* 1656 * Iff doesn't exist, create it. 1657 */ 1658 if (nd.ni_vp) { 1659 error = EEXIST; 1660 abort: 1661 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1662 if (nd.ni_dvp == nd.ni_vp) 1663 vrele(nd.ni_dvp); 1664 else 1665 vput(nd.ni_dvp); 1666 if (nd.ni_vp) 1667 vput(nd.ni_vp); 1668 goto out; 1669 } 1670 va.va_type = vtyp; 1671 if (vtyp == VSOCK) { 1672 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1673 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1674 } else { 1675 if (va.va_type != VFIFO && 1676 (error = kauth_authorize_generic(cred, 1677 KAUTH_GENERIC_ISSUSER, NULL))) { 1678 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1679 vput(nd.ni_dvp); 1680 goto out; 1681 } 1682 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1683 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 1684 if (error) 1685 goto out; 1686 if (nd.ni_cnd.cn_flags & ISSYMLINK) { 1687 vput(nd.ni_vp); 1688 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1689 error = EINVAL; 1690 } 1691 } 1692 out: 1693 vp = nd.ni_vp; 1694 if (!error) { 1695 error = nfsrv_composefh(vp, &nsfh, true); 1696 if (!error) 1697 error = VOP_GETATTR(vp, &va, cred); 1698 vput(vp); 1699 } 1700 if (dirp) { 1701 diraft_ret = VOP_GETATTR(dirp, &diraft, cred); 1702 vrele(dirp); 1703 } 1704 nfsm_reply(NFSX_SRVFH(&nsfh, true) + NFSX_POSTOPATTR(1) + 1705 NFSX_WCCDATA(1)); 1706 if (!error) { 1707 nfsm_srvpostop_fh(&nsfh); 1708 nfsm_srvpostop_attr(0, &va); 1709 } 1710 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1711 return (0); 1712 nfsmout: 1713 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1714 if (nd.ni_dvp == nd.ni_vp) 1715 vrele(nd.ni_dvp); 1716 else 1717 vput(nd.ni_dvp); 1718 if (nd.ni_vp) 1719 vput(nd.ni_vp); 1720 if (dirp) 1721 vrele(dirp); 1722 return (error); 1723 } 1724 1725 /* 1726 * nfs remove service 1727 */ 1728 int 1729 nfsrv_remove(nfsd, slp, lwp, mrq) 1730 struct nfsrv_descript *nfsd; 1731 struct nfssvc_sock *slp; 1732 struct lwp *lwp; 1733 struct mbuf **mrq; 1734 { 1735 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1736 struct mbuf *nam = nfsd->nd_nam; 1737 char *dpos = nfsd->nd_dpos; 1738 kauth_cred_t cred = nfsd->nd_cr; 1739 struct nameidata nd; 1740 u_int32_t *tl; 1741 int32_t t1; 1742 char *bpos; 1743 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 1744 int v3 = (nfsd->nd_flag & ND_NFSV3); 1745 char *cp2; 1746 struct mbuf *mb, *mreq; 1747 struct vnode *vp, *dirp; 1748 struct vattr dirfor, diraft; 1749 nfsrvfh_t nsfh; 1750 u_quad_t frev; 1751 1752 #ifndef nolint 1753 vp = (struct vnode *)0; 1754 #endif 1755 nfsm_srvmtofh(&nsfh); 1756 nfsm_srvnamesiz(len); 1757 nd.ni_cnd.cn_cred = cred; 1758 nd.ni_cnd.cn_nameiop = DELETE; 1759 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 1760 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 1761 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1762 if (dirp && v3) { 1763 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred); 1764 } 1765 if (!error) { 1766 vp = nd.ni_vp; 1767 if (vp->v_type == VDIR && 1768 (error = kauth_authorize_generic(cred, 1769 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 1770 goto out; 1771 /* 1772 * The root of a mounted filesystem cannot be deleted. 1773 */ 1774 if (vp->v_vflag & VV_ROOT) { 1775 error = EBUSY; 1776 goto out; 1777 } 1778 out: 1779 if (!error) { 1780 nqsrv_getl(nd.ni_dvp, ND_WRITE); 1781 nqsrv_getl(vp, ND_WRITE); 1782 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1783 } else { 1784 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1785 if (nd.ni_dvp == vp) 1786 vrele(nd.ni_dvp); 1787 else 1788 vput(nd.ni_dvp); 1789 vput(vp); 1790 } 1791 } 1792 if (dirp) { 1793 if (v3) { 1794 diraft_ret = VOP_GETATTR(dirp, &diraft, cred); 1795 } 1796 vrele(dirp); 1797 } 1798 nfsm_reply(NFSX_WCCDATA(v3)); 1799 if (v3) { 1800 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 1801 return (0); 1802 } 1803 nfsm_srvdone; 1804 } 1805 1806 /* 1807 * nfs rename service 1808 */ 1809 int 1810 nfsrv_rename(nfsd, slp, lwp, mrq) 1811 struct nfsrv_descript *nfsd; 1812 struct nfssvc_sock *slp; 1813 struct lwp *lwp; 1814 struct mbuf **mrq; 1815 { 1816 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 1817 struct mbuf *nam = nfsd->nd_nam; 1818 char *dpos = nfsd->nd_dpos; 1819 kauth_cred_t cred = nfsd->nd_cr; 1820 u_int32_t *tl; 1821 int32_t t1; 1822 char *bpos; 1823 int error = 0, cache = 0, fdirfor_ret = 1, fdiraft_ret = 1; 1824 uint32_t len, len2; 1825 int tdirfor_ret = 1, tdiraft_ret = 1; 1826 int v3 = (nfsd->nd_flag & ND_NFSV3); 1827 char *cp2; 1828 struct mbuf *mb, *mreq; 1829 struct nameidata fromnd, tond; 1830 struct vnode *fvp, *tvp, *tdvp; 1831 struct vnode *fdirp = NULL, *tdirp = NULL; 1832 struct vattr fdirfor, fdiraft, tdirfor, tdiraft; 1833 nfsrvfh_t fnsfh, tnsfh; 1834 u_quad_t frev; 1835 uid_t saved_uid; 1836 1837 #ifndef nolint 1838 fvp = (struct vnode *)0; 1839 #endif 1840 fromnd.ni_cnd.cn_nameiop = 0; 1841 tond.ni_cnd.cn_nameiop = 0; 1842 nfsm_srvmtofh(&fnsfh); 1843 nfsm_srvnamesiz(len); 1844 /* 1845 * Remember our original uid so that we can reset cr_uid before 1846 * the second nfs_namei() call, in case it is remapped. 1847 */ 1848 saved_uid = kauth_cred_geteuid(cred); 1849 fromnd.ni_cnd.cn_cred = cred; 1850 fromnd.ni_cnd.cn_nameiop = DELETE; 1851 fromnd.ni_cnd.cn_flags = LOCKPARENT | SAVESTART; 1852 error = nfs_namei(&fromnd, &fnsfh, len, slp, nam, &md, 1853 &dpos, &fdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1854 if (fdirp && v3) { 1855 fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred); 1856 } 1857 if (error) { 1858 nfsm_reply(2 * NFSX_WCCDATA(v3)); 1859 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 1860 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 1861 if (fdirp) 1862 vrele(fdirp); 1863 return (0); 1864 } 1865 if (fromnd.ni_dvp != fromnd.ni_vp) { 1866 VOP_UNLOCK(fromnd.ni_dvp, 0); 1867 } 1868 fvp = fromnd.ni_vp; 1869 nfsm_srvmtofh(&tnsfh); 1870 if (v3) { 1871 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 1872 len2 = fxdr_unsigned(uint32_t, *tl); 1873 /* len2 will be checked by nfs_namei */ 1874 } 1875 else { 1876 /* NFSv2 */ 1877 nfsm_strsiz(len2, NFS_MAXNAMLEN); 1878 } 1879 kauth_cred_seteuid(cred, saved_uid); 1880 tond.ni_cnd.cn_cred = cred; 1881 tond.ni_cnd.cn_nameiop = RENAME; 1882 tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART; 1883 error = nfs_namei(&tond, &tnsfh, len2, slp, nam, &md, 1884 &dpos, &tdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 1885 if (tdirp && v3) { 1886 tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred); 1887 } 1888 if (error) { 1889 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1890 vrele(fromnd.ni_dvp); 1891 vrele(fvp); 1892 goto out1; 1893 } 1894 tdvp = tond.ni_dvp; 1895 tvp = tond.ni_vp; 1896 if (tvp != NULL) { 1897 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 1898 if (v3) 1899 error = EEXIST; 1900 else 1901 error = EISDIR; 1902 goto out; 1903 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 1904 if (v3) 1905 error = EEXIST; 1906 else 1907 error = ENOTDIR; 1908 goto out; 1909 } 1910 if (tvp->v_type == VDIR && tvp->v_mountedhere) { 1911 if (v3) 1912 error = EXDEV; 1913 else 1914 error = ENOTEMPTY; 1915 goto out; 1916 } 1917 } 1918 if (fvp->v_type == VDIR && fvp->v_mountedhere) { 1919 if (v3) 1920 error = EXDEV; 1921 else 1922 error = ENOTEMPTY; 1923 goto out; 1924 } 1925 if (fvp->v_mount != tdvp->v_mount) { 1926 if (v3) 1927 error = EXDEV; 1928 else 1929 error = ENOTEMPTY; 1930 goto out; 1931 } 1932 if (fvp == tdvp) { 1933 if (v3) 1934 error = EINVAL; 1935 else 1936 error = ENOTEMPTY; 1937 } 1938 /* 1939 * If source is the same as the destination (that is the 1940 * same vnode with the same name in the same directory), 1941 * then there is nothing to do. 1942 */ 1943 if (fvp == tvp && fromnd.ni_dvp == tdvp && 1944 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 1945 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr, 1946 fromnd.ni_cnd.cn_namelen)) 1947 error = -1; 1948 out: 1949 if (!error) { 1950 nqsrv_getl(fromnd.ni_dvp, ND_WRITE); 1951 nqsrv_getl(tdvp, ND_WRITE); 1952 if (tvp) { 1953 nqsrv_getl(tvp, ND_WRITE); 1954 } 1955 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 1956 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 1957 } else { 1958 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 1959 if (tdvp == tvp) 1960 vrele(tdvp); 1961 else 1962 vput(tdvp); 1963 if (tvp) 1964 vput(tvp); 1965 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 1966 vrele(fromnd.ni_dvp); 1967 vrele(fvp); 1968 if (error == -1) 1969 error = 0; 1970 } 1971 vrele(tond.ni_startdir); 1972 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 1973 out1: 1974 if (fdirp) { 1975 if (v3) { 1976 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred); 1977 } 1978 vrele(fdirp); 1979 } 1980 if (tdirp) { 1981 if (v3) { 1982 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred); 1983 } 1984 vrele(tdirp); 1985 } 1986 vrele(fromnd.ni_startdir); 1987 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 1988 nfsm_reply(2 * NFSX_WCCDATA(v3)); 1989 if (v3) { 1990 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); 1991 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); 1992 } 1993 return (0); 1994 1995 nfsmout: 1996 if (fdirp) 1997 vrele(fdirp); 1998 #ifdef notdef 1999 if (tdirp) 2000 vrele(tdirp); 2001 #endif 2002 if (tond.ni_cnd.cn_nameiop) { 2003 vrele(tond.ni_startdir); 2004 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 2005 } 2006 if (fromnd.ni_cnd.cn_nameiop) { 2007 vrele(fromnd.ni_startdir); 2008 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 2009 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2010 vrele(fromnd.ni_dvp); 2011 vrele(fvp); 2012 } 2013 return (error); 2014 } 2015 2016 /* 2017 * nfs link service 2018 */ 2019 int 2020 nfsrv_link(nfsd, slp, lwp, mrq) 2021 struct nfsrv_descript *nfsd; 2022 struct nfssvc_sock *slp; 2023 struct lwp *lwp; 2024 struct mbuf **mrq; 2025 { 2026 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2027 struct mbuf *nam = nfsd->nd_nam; 2028 char *dpos = nfsd->nd_dpos; 2029 kauth_cred_t cred = nfsd->nd_cr; 2030 struct nameidata nd; 2031 u_int32_t *tl; 2032 int32_t t1; 2033 char *bpos; 2034 int error = 0, rdonly, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 2035 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3); 2036 char *cp2; 2037 struct mbuf *mb, *mreq; 2038 struct vnode *vp, *xp, *dirp = (struct vnode *)0; 2039 struct vattr dirfor, diraft, at; 2040 nfsrvfh_t nsfh, dnsfh; 2041 u_quad_t frev; 2042 2043 nfsm_srvmtofh(&nsfh); 2044 nfsm_srvmtofh(&dnsfh); 2045 nfsm_srvnamesiz(len); 2046 error = nfsrv_fhtovp(&nsfh, false, &vp, cred, slp, nam, 2047 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 2048 if (error) { 2049 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2050 nfsm_srvpostop_attr(getret, &at); 2051 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2052 return (0); 2053 } 2054 if (vp->v_type == VDIR && (error = kauth_authorize_generic(cred, 2055 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 2056 goto out1; 2057 nd.ni_cnd.cn_cred = cred; 2058 nd.ni_cnd.cn_nameiop = CREATE; 2059 nd.ni_cnd.cn_flags = LOCKPARENT; 2060 error = nfs_namei(&nd, &dnsfh, len, slp, nam, &md, &dpos, 2061 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2062 if (dirp && v3) { 2063 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred); 2064 } 2065 if (error) 2066 goto out1; 2067 xp = nd.ni_vp; 2068 if (xp != NULL) { 2069 error = EEXIST; 2070 goto out; 2071 } 2072 xp = nd.ni_dvp; 2073 if (vp->v_mount != xp->v_mount) 2074 error = EXDEV; 2075 out: 2076 if (!error) { 2077 nqsrv_getl(vp, ND_WRITE); 2078 nqsrv_getl(xp, ND_WRITE); 2079 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2080 } else { 2081 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2082 if (nd.ni_dvp == nd.ni_vp) 2083 vrele(nd.ni_dvp); 2084 else 2085 vput(nd.ni_dvp); 2086 if (nd.ni_vp) 2087 vrele(nd.ni_vp); 2088 } 2089 out1: 2090 if (v3) 2091 getret = VOP_GETATTR(vp, &at, cred); 2092 if (dirp) { 2093 if (v3) { 2094 diraft_ret = VOP_GETATTR(dirp, &diraft, cred); 2095 } 2096 vrele(dirp); 2097 } 2098 vrele(vp); 2099 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); 2100 if (v3) { 2101 nfsm_srvpostop_attr(getret, &at); 2102 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2103 return (0); 2104 } 2105 nfsm_srvdone; 2106 } 2107 2108 /* 2109 * nfs symbolic link service 2110 */ 2111 int 2112 nfsrv_symlink(nfsd, slp, lwp, mrq) 2113 struct nfsrv_descript *nfsd; 2114 struct nfssvc_sock *slp; 2115 struct lwp *lwp; 2116 struct mbuf **mrq; 2117 { 2118 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2119 struct mbuf *nam = nfsd->nd_nam; 2120 char *dpos = nfsd->nd_dpos; 2121 kauth_cred_t cred = nfsd->nd_cr; 2122 struct vattr va, dirfor, diraft; 2123 struct nameidata nd; 2124 u_int32_t *tl; 2125 int32_t t1; 2126 struct nfsv2_sattr *sp; 2127 char *bpos, *pathcp = NULL, *cp2; 2128 struct uio io; 2129 struct iovec iv; 2130 int error = 0, cache = 0, dirfor_ret = 1, diraft_ret = 1; 2131 uint32_t len, len2; 2132 int v3 = (nfsd->nd_flag & ND_NFSV3); 2133 struct mbuf *mb, *mreq; 2134 struct vnode *dirp = (struct vnode *)0; 2135 nfsrvfh_t nsfh; 2136 u_quad_t frev; 2137 2138 nd.ni_cnd.cn_nameiop = 0; 2139 nfsm_srvmtofh(&nsfh); 2140 nfsm_srvnamesiz(len); 2141 nd.ni_cnd.cn_cred = cred; 2142 nd.ni_cnd.cn_nameiop = CREATE; 2143 nd.ni_cnd.cn_flags = LOCKPARENT; 2144 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 2145 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2146 if (dirp && v3) { 2147 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred); 2148 } 2149 if (error) 2150 goto out; 2151 VATTR_NULL(&va); 2152 va.va_type = VLNK; 2153 if (v3) { 2154 va.va_mode = 0; 2155 nfsm_srvsattr(&va); 2156 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED); 2157 len2 = fxdr_unsigned(uint32_t, *tl); 2158 if (len2 > PATH_MAX) { 2159 /* XXX should check _PC_NO_TRUNC */ 2160 error = ENAMETOOLONG; 2161 goto abortop; 2162 } 2163 } 2164 else { 2165 /* NFSv2 */ 2166 nfsm_strsiz(len2, NFS_MAXPATHLEN); 2167 } 2168 pathcp = malloc(len2 + 1, M_TEMP, M_WAITOK); 2169 iv.iov_base = pathcp; 2170 iv.iov_len = len2; 2171 io.uio_resid = len2; 2172 io.uio_offset = 0; 2173 io.uio_iov = &iv; 2174 io.uio_iovcnt = 1; 2175 io.uio_rw = UIO_READ; 2176 UIO_SETUP_SYSSPACE(&io); 2177 nfsm_mtouio(&io, len2); 2178 if (!v3) { 2179 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 2180 va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode); 2181 } 2182 *(pathcp + len2) = '\0'; 2183 if (nd.ni_vp) { 2184 error = EEXIST; 2185 abortop: 2186 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2187 if (nd.ni_dvp == nd.ni_vp) 2188 vrele(nd.ni_dvp); 2189 else 2190 vput(nd.ni_dvp); 2191 if (nd.ni_vp) 2192 vrele(nd.ni_vp); 2193 goto out; 2194 } 2195 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2196 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp); 2197 if (!error) { 2198 if (v3) { 2199 error = nfsrv_composefh(nd.ni_vp, &nsfh, v3); 2200 if (!error) 2201 error = VOP_GETATTR(nd.ni_vp, &va, cred); 2202 vput(nd.ni_vp); 2203 } else { 2204 vput(nd.ni_vp); 2205 } 2206 } 2207 out: 2208 if (pathcp) 2209 free(pathcp, M_TEMP); 2210 if (dirp) { 2211 if (v3) { 2212 diraft_ret = VOP_GETATTR(dirp, &diraft, cred); 2213 } 2214 vrele(dirp); 2215 } 2216 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_POSTOPATTR(v3) + 2217 NFSX_WCCDATA(v3)); 2218 if (v3) { 2219 if (!error) { 2220 nfsm_srvpostop_fh(&nsfh); 2221 nfsm_srvpostop_attr(0, &va); 2222 } 2223 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2224 } 2225 return (0); 2226 nfsmout: 2227 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2228 if (nd.ni_dvp == nd.ni_vp) 2229 vrele(nd.ni_dvp); 2230 else 2231 vput(nd.ni_dvp); 2232 if (nd.ni_vp) 2233 vrele(nd.ni_vp); 2234 if (dirp) 2235 vrele(dirp); 2236 if (pathcp) 2237 free(pathcp, M_TEMP); 2238 return (error); 2239 } 2240 2241 /* 2242 * nfs mkdir service 2243 */ 2244 int 2245 nfsrv_mkdir(nfsd, slp, lwp, mrq) 2246 struct nfsrv_descript *nfsd; 2247 struct nfssvc_sock *slp; 2248 struct lwp *lwp; 2249 struct mbuf **mrq; 2250 { 2251 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2252 struct mbuf *nam = nfsd->nd_nam; 2253 char *dpos = nfsd->nd_dpos; 2254 kauth_cred_t cred = nfsd->nd_cr; 2255 struct vattr va, dirfor, diraft; 2256 struct nfs_fattr *fp; 2257 struct nameidata nd; 2258 char *cp; 2259 u_int32_t *tl; 2260 int32_t t1; 2261 char *bpos; 2262 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 2263 int v3 = (nfsd->nd_flag & ND_NFSV3); 2264 char *cp2; 2265 struct mbuf *mb, *mreq; 2266 struct vnode *vp, *dirp = (struct vnode *)0; 2267 nfsrvfh_t nsfh; 2268 u_quad_t frev; 2269 2270 nfsm_srvmtofh(&nsfh); 2271 nfsm_srvnamesiz(len); 2272 nd.ni_cnd.cn_cred = cred; 2273 nd.ni_cnd.cn_nameiop = CREATE; 2274 nd.ni_cnd.cn_flags = LOCKPARENT; 2275 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 2276 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2277 if (dirp && v3) { 2278 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred); 2279 } 2280 if (error) { 2281 nfsm_reply(NFSX_WCCDATA(v3)); 2282 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2283 if (dirp) 2284 vrele(dirp); 2285 return (0); 2286 } 2287 VATTR_NULL(&va); 2288 if (v3) { 2289 va.va_mode = 0; 2290 nfsm_srvsattr(&va); 2291 } else { 2292 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED); 2293 va.va_mode = nfstov_mode(*tl++); 2294 } 2295 va.va_type = VDIR; 2296 vp = nd.ni_vp; 2297 if (vp != NULL) { 2298 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2299 if (nd.ni_dvp == vp) 2300 vrele(nd.ni_dvp); 2301 else 2302 vput(nd.ni_dvp); 2303 vrele(vp); 2304 error = EEXIST; 2305 goto out; 2306 } 2307 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2308 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va); 2309 if (!error) { 2310 vp = nd.ni_vp; 2311 error = nfsrv_composefh(vp, &nsfh, v3); 2312 if (!error) 2313 error = VOP_GETATTR(vp, &va, cred); 2314 vput(vp); 2315 } 2316 out: 2317 if (dirp) { 2318 if (v3) { 2319 diraft_ret = VOP_GETATTR(dirp, &diraft, cred); 2320 } 2321 vrele(dirp); 2322 } 2323 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_POSTOPATTR(v3) + 2324 NFSX_WCCDATA(v3)); 2325 if (v3) { 2326 if (!error) { 2327 nfsm_srvpostop_fh(&nsfh); 2328 nfsm_srvpostop_attr(0, &va); 2329 } 2330 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2331 } else { 2332 nfsm_srvfhtom(&nsfh, v3); 2333 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); 2334 nfsm_srvfillattr(&va, fp); 2335 } 2336 return (0); 2337 nfsmout: 2338 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2339 if (nd.ni_dvp == nd.ni_vp) 2340 vrele(nd.ni_dvp); 2341 else 2342 vput(nd.ni_dvp); 2343 if (nd.ni_vp) 2344 vrele(nd.ni_vp); 2345 if (dirp) 2346 vrele(dirp); 2347 return (error); 2348 } 2349 2350 /* 2351 * nfs rmdir service 2352 */ 2353 int 2354 nfsrv_rmdir(nfsd, slp, lwp, mrq) 2355 struct nfsrv_descript *nfsd; 2356 struct nfssvc_sock *slp; 2357 struct lwp *lwp; 2358 struct mbuf **mrq; 2359 { 2360 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2361 struct mbuf *nam = nfsd->nd_nam; 2362 char *dpos = nfsd->nd_dpos; 2363 kauth_cred_t cred = nfsd->nd_cr; 2364 u_int32_t *tl; 2365 int32_t t1; 2366 char *bpos; 2367 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1; 2368 int v3 = (nfsd->nd_flag & ND_NFSV3); 2369 char *cp2; 2370 struct mbuf *mb, *mreq; 2371 struct vnode *vp, *dirp = (struct vnode *)0; 2372 struct vattr dirfor, diraft; 2373 nfsrvfh_t nsfh; 2374 struct nameidata nd; 2375 u_quad_t frev; 2376 2377 nfsm_srvmtofh(&nsfh); 2378 nfsm_srvnamesiz(len); 2379 nd.ni_cnd.cn_cred = cred; 2380 nd.ni_cnd.cn_nameiop = DELETE; 2381 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 2382 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos, 2383 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false); 2384 if (dirp && v3) { 2385 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred); 2386 } 2387 if (error) { 2388 nfsm_reply(NFSX_WCCDATA(v3)); 2389 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2390 if (dirp) 2391 vrele(dirp); 2392 return (0); 2393 } 2394 vp = nd.ni_vp; 2395 if (vp->v_type != VDIR) { 2396 error = ENOTDIR; 2397 goto out; 2398 } 2399 /* 2400 * No rmdir "." please. 2401 */ 2402 if (nd.ni_dvp == vp) { 2403 error = EINVAL; 2404 goto out; 2405 } 2406 /* 2407 * The root of a mounted filesystem cannot be deleted. 2408 */ 2409 if (vp->v_vflag & VV_ROOT) 2410 error = EBUSY; 2411 out: 2412 if (!error) { 2413 nqsrv_getl(nd.ni_dvp, ND_WRITE); 2414 nqsrv_getl(vp, ND_WRITE); 2415 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2416 } else { 2417 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2418 if (nd.ni_dvp == nd.ni_vp) 2419 vrele(nd.ni_dvp); 2420 else 2421 vput(nd.ni_dvp); 2422 vput(vp); 2423 } 2424 if (dirp) { 2425 if (v3) { 2426 diraft_ret = VOP_GETATTR(dirp, &diraft, cred); 2427 } 2428 vrele(dirp); 2429 } 2430 nfsm_reply(NFSX_WCCDATA(v3)); 2431 if (v3) { 2432 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); 2433 return (0); 2434 } 2435 nfsm_srvdone; 2436 } 2437 2438 /* 2439 * nfs readdir service 2440 * - mallocs what it thinks is enough to read 2441 * count rounded up to a multiple of NFS_SRVDIRBLKSIZ <= NFS_MAXREADDIR 2442 * - calls VOP_READDIR() 2443 * - loops around building the reply 2444 * if the output generated exceeds count break out of loop 2445 * The nfsm_clget macro is used here so that the reply will be packed 2446 * tightly in mbuf clusters. 2447 * - it only knows that it has encountered eof when the VOP_READDIR() 2448 * reads nothing 2449 * - as such one readdir rpc will return eof false although you are there 2450 * and then the next will return eof 2451 * - it trims out records with d_fileno == 0 2452 * this doesn't matter for Unix clients, but they might confuse clients 2453 * for other os'. 2454 * - it trims out records with d_type == DT_WHT 2455 * these cannot be seen through NFS (unless we extend the protocol) 2456 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less 2457 * than requested, but this may not apply to all filesystems. For 2458 * example, client NFS does not { although it is never remote mounted 2459 * anyhow } 2460 * The alternate call nfsrv_readdirplus() does lookups as well. 2461 * PS: The NFS protocol spec. does not clarify what the "count" byte 2462 * argument is a count of.. just name strings and file id's or the 2463 * entire reply rpc or ... 2464 * I tried just file name and id sizes and it confused the Sun client, 2465 * so I am using the full rpc size now. The "paranoia.." comment refers 2466 * to including the status longwords that are not a part of the dir. 2467 * "entry" structures, but are in the rpc. 2468 */ 2469 2470 #define NFS_SRVDIRBLKSIZ 1024 2471 2472 struct flrep { 2473 nfsuint64 fl_off; 2474 u_int32_t fl_postopok; 2475 u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)]; 2476 u_int32_t fl_fhok; 2477 u_int32_t fl_fhsize; 2478 }; 2479 2480 int 2481 nfsrv_readdir(nfsd, slp, lwp, mrq) 2482 struct nfsrv_descript *nfsd; 2483 struct nfssvc_sock *slp; 2484 struct lwp *lwp; 2485 struct mbuf **mrq; 2486 { 2487 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2488 struct mbuf *nam = nfsd->nd_nam; 2489 char *dpos = nfsd->nd_dpos; 2490 kauth_cred_t cred = nfsd->nd_cr; 2491 char *bp, *be; 2492 struct mbuf *mp; 2493 struct dirent *dp; 2494 char *cp; 2495 u_int32_t *tl; 2496 int32_t t1; 2497 char *bpos; 2498 struct mbuf *mb, *mreq, *mp2; 2499 char *cpos, *cend, *cp2, *rbuf; 2500 struct vnode *vp; 2501 struct vattr at; 2502 nfsrvfh_t nsfh; 2503 struct uio io; 2504 struct iovec iv; 2505 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2506 int siz, cnt, fullsiz, eofflag, rdonly, cache = 0, ncookies; 2507 int v3 = (nfsd->nd_flag & ND_NFSV3); 2508 u_quad_t frev, off, toff, verf; 2509 off_t *cookies = NULL, *cookiep; 2510 nfsuint64 jar; 2511 2512 nfsm_srvmtofh(&nsfh); 2513 if (v3) { 2514 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 2515 toff = fxdr_hyper(tl); 2516 tl += 2; 2517 verf = fxdr_hyper(tl); 2518 tl += 2; 2519 } else { 2520 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2521 toff = fxdr_unsigned(u_quad_t, *tl++); 2522 } 2523 off = toff; 2524 cnt = fxdr_unsigned(int, *tl); 2525 siz = ((cnt + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1)); 2526 xfer = NFS_SRVMAXDATA(nfsd); 2527 if (siz > xfer) 2528 siz = xfer; 2529 fullsiz = siz; 2530 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 2531 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 2532 if (!error && vp->v_type != VDIR) { 2533 error = ENOTDIR; 2534 vput(vp); 2535 } 2536 if (error) { 2537 nfsm_reply(NFSX_UNSIGNED); 2538 nfsm_srvpostop_attr(getret, &at); 2539 return (0); 2540 } 2541 nqsrv_getl(vp, ND_READ); 2542 if (v3) { 2543 error = getret = VOP_GETATTR(vp, &at, cred); 2544 #ifdef NFS3_STRICTVERF 2545 /* 2546 * XXX This check is too strict for Solaris 2.5 clients. 2547 */ 2548 if (!error && toff && verf != at.va_filerev) 2549 error = NFSERR_BAD_COOKIE; 2550 #endif 2551 } 2552 if (!error) 2553 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0); 2554 if (error) { 2555 vput(vp); 2556 nfsm_reply(NFSX_POSTOPATTR(v3)); 2557 nfsm_srvpostop_attr(getret, &at); 2558 return (0); 2559 } 2560 VOP_UNLOCK(vp, 0); 2561 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2562 again: 2563 iv.iov_base = rbuf; 2564 iv.iov_len = fullsiz; 2565 io.uio_iov = &iv; 2566 io.uio_iovcnt = 1; 2567 io.uio_offset = (off_t)off; 2568 io.uio_resid = fullsiz; 2569 io.uio_rw = UIO_READ; 2570 UIO_SETUP_SYSSPACE(&io); 2571 eofflag = 0; 2572 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2573 2574 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2575 2576 off = (off_t)io.uio_offset; 2577 if (!cookies && !error) 2578 error = NFSERR_PERM; 2579 if (v3) { 2580 getret = VOP_GETATTR(vp, &at, cred); 2581 if (!error) 2582 error = getret; 2583 } 2584 2585 VOP_UNLOCK(vp, 0); 2586 if (error) { 2587 vrele(vp); 2588 free((void *)rbuf, M_TEMP); 2589 if (cookies) 2590 free((void *)cookies, M_TEMP); 2591 nfsm_reply(NFSX_POSTOPATTR(v3)); 2592 nfsm_srvpostop_attr(getret, &at); 2593 return (0); 2594 } 2595 if (io.uio_resid) { 2596 siz -= io.uio_resid; 2597 2598 /* 2599 * If nothing read, return eof 2600 * rpc reply 2601 */ 2602 if (siz == 0) { 2603 vrele(vp); 2604 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + 2605 2 * NFSX_UNSIGNED); 2606 if (v3) { 2607 nfsm_srvpostop_attr(getret, &at); 2608 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2609 txdr_hyper(at.va_filerev, tl); 2610 tl += 2; 2611 } else 2612 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2613 *tl++ = nfs_false; 2614 *tl = nfs_true; 2615 free((void *)rbuf, M_TEMP); 2616 free((void *)cookies, M_TEMP); 2617 return (0); 2618 } 2619 } 2620 2621 /* 2622 * Check for degenerate cases of nothing useful read. 2623 * If so go try again 2624 */ 2625 cpos = rbuf; 2626 cend = rbuf + siz; 2627 dp = (struct dirent *)cpos; 2628 cookiep = cookies; 2629 2630 while (cpos < cend && ncookies > 0 && 2631 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2632 cpos += dp->d_reclen; 2633 dp = (struct dirent *)cpos; 2634 cookiep++; 2635 ncookies--; 2636 } 2637 if (cpos >= cend || ncookies == 0) { 2638 toff = off; 2639 siz = fullsiz; 2640 free(cookies, M_TEMP); 2641 cookies = NULL; 2642 goto again; 2643 } 2644 2645 len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */ 2646 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz); 2647 if (v3) { 2648 nfsm_srvpostop_attr(getret, &at); 2649 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2650 txdr_hyper(at.va_filerev, tl); 2651 } 2652 mp = mp2 = mb; 2653 bp = bpos; 2654 be = bp + M_TRAILINGSPACE(mp); 2655 2656 /* Loop through the records and build reply */ 2657 while (cpos < cend && ncookies > 0) { 2658 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 2659 nlen = dp->d_namlen; 2660 rem = nfsm_rndup(nlen)-nlen; 2661 len += (4 * NFSX_UNSIGNED + nlen + rem); 2662 if (v3) 2663 len += 2 * NFSX_UNSIGNED; 2664 if (len > cnt) { 2665 eofflag = 0; 2666 break; 2667 } 2668 /* 2669 * Build the directory record xdr from 2670 * the dirent entry. 2671 */ 2672 nfsm_clget; 2673 *tl = nfs_true; 2674 bp += NFSX_UNSIGNED; 2675 if (v3) { 2676 nfsm_clget; 2677 *tl = txdr_unsigned(dp->d_fileno >> 32); 2678 bp += NFSX_UNSIGNED; 2679 } 2680 nfsm_clget; 2681 *tl = txdr_unsigned(dp->d_fileno); 2682 bp += NFSX_UNSIGNED; 2683 nfsm_clget; 2684 *tl = txdr_unsigned(nlen); 2685 bp += NFSX_UNSIGNED; 2686 2687 /* And loop around copying the name */ 2688 xfer = nlen; 2689 cp = dp->d_name; 2690 while (xfer > 0) { 2691 nfsm_clget; 2692 if ((bp+xfer) > be) 2693 tsiz = be-bp; 2694 else 2695 tsiz = xfer; 2696 memcpy(bp, cp, tsiz); 2697 bp += tsiz; 2698 xfer -= tsiz; 2699 if (xfer > 0) 2700 cp += tsiz; 2701 } 2702 /* And null pad to an int32_t boundary */ 2703 for (i = 0; i < rem; i++) 2704 *bp++ = '\0'; 2705 nfsm_clget; 2706 2707 /* Finish off the record */ 2708 txdr_hyper(*cookiep, &jar); 2709 if (v3) { 2710 *tl = jar.nfsuquad[0]; 2711 bp += NFSX_UNSIGNED; 2712 nfsm_clget; 2713 } 2714 *tl = jar.nfsuquad[1]; 2715 bp += NFSX_UNSIGNED; 2716 } 2717 cpos += dp->d_reclen; 2718 dp = (struct dirent *)cpos; 2719 cookiep++; 2720 ncookies--; 2721 } 2722 vrele(vp); 2723 nfsm_clget; 2724 *tl = nfs_false; 2725 bp += NFSX_UNSIGNED; 2726 nfsm_clget; 2727 if (eofflag) 2728 *tl = nfs_true; 2729 else 2730 *tl = nfs_false; 2731 bp += NFSX_UNSIGNED; 2732 if (mp != mb) { 2733 if (bp < be) 2734 mp->m_len = bp - mtod(mp, char *); 2735 } else 2736 mp->m_len += bp - bpos; 2737 free((void *)rbuf, M_TEMP); 2738 free((void *)cookies, M_TEMP); 2739 nfsm_srvdone; 2740 } 2741 2742 int 2743 nfsrv_readdirplus(nfsd, slp, lwp, mrq) 2744 struct nfsrv_descript *nfsd; 2745 struct nfssvc_sock *slp; 2746 struct lwp *lwp; 2747 struct mbuf **mrq; 2748 { 2749 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 2750 struct mbuf *nam = nfsd->nd_nam; 2751 char *dpos = nfsd->nd_dpos; 2752 kauth_cred_t cred = nfsd->nd_cr; 2753 char *bp, *be; 2754 struct mbuf *mp; 2755 struct dirent *dp; 2756 char *cp; 2757 u_int32_t *tl; 2758 int32_t t1; 2759 char *bpos; 2760 struct mbuf *mb, *mreq, *mp2; 2761 char *cpos, *cend, *cp2, *rbuf; 2762 struct vnode *vp, *nvp; 2763 struct flrep fl; 2764 nfsrvfh_t nsfh; 2765 struct uio io; 2766 struct iovec iv; 2767 struct vattr va, at, *vap = &va; 2768 struct nfs_fattr *fp; 2769 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; 2770 int siz, cnt, fullsiz, eofflag, rdonly, cache = 0, dirlen, ncookies; 2771 u_quad_t frev, off, toff, verf; 2772 off_t *cookies = NULL, *cookiep; 2773 2774 nfsm_srvmtofh(&nsfh); 2775 nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED); 2776 toff = fxdr_hyper(tl); 2777 tl += 2; 2778 verf = fxdr_hyper(tl); 2779 tl += 2; 2780 siz = fxdr_unsigned(int, *tl++); 2781 cnt = fxdr_unsigned(int, *tl); 2782 off = toff; 2783 siz = ((siz + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1)); 2784 xfer = NFS_SRVMAXDATA(nfsd); 2785 if (siz > xfer) 2786 siz = xfer; 2787 fullsiz = siz; 2788 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 2789 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 2790 if (!error && vp->v_type != VDIR) { 2791 error = ENOTDIR; 2792 vput(vp); 2793 } 2794 if (error) { 2795 nfsm_reply(NFSX_UNSIGNED); 2796 nfsm_srvpostop_attr(getret, &at); 2797 return (0); 2798 } 2799 error = getret = VOP_GETATTR(vp, &at, cred); 2800 #ifdef NFS3_STRICTVERF 2801 /* 2802 * XXX This check is too strict for Solaris 2.5 clients. 2803 */ 2804 if (!error && toff && verf != at.va_filerev) 2805 error = NFSERR_BAD_COOKIE; 2806 #endif 2807 if (!error) { 2808 nqsrv_getl(vp, ND_READ); 2809 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0); 2810 } 2811 if (error) { 2812 vput(vp); 2813 nfsm_reply(NFSX_V3POSTOPATTR); 2814 nfsm_srvpostop_attr(getret, &at); 2815 return (0); 2816 } 2817 VOP_UNLOCK(vp, 0); 2818 2819 rbuf = malloc(siz, M_TEMP, M_WAITOK); 2820 again: 2821 iv.iov_base = rbuf; 2822 iv.iov_len = fullsiz; 2823 io.uio_iov = &iv; 2824 io.uio_iovcnt = 1; 2825 io.uio_offset = (off_t)off; 2826 io.uio_resid = fullsiz; 2827 io.uio_rw = UIO_READ; 2828 UIO_SETUP_SYSSPACE(&io); 2829 eofflag = 0; 2830 2831 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2832 2833 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies); 2834 2835 off = (u_quad_t)io.uio_offset; 2836 getret = VOP_GETATTR(vp, &at, cred); 2837 2838 VOP_UNLOCK(vp, 0); 2839 2840 /* 2841 * If the VGET operation doesn't work for this filesystem, 2842 * we can't support readdirplus. Returning NOTSUPP should 2843 * make clients fall back to plain readdir. 2844 * There's no need to check for VPTOFH as well, we wouldn't 2845 * even be here otherwise. 2846 */ 2847 if (!getret) { 2848 if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp))) 2849 getret = (getret == EOPNOTSUPP) ? 2850 NFSERR_NOTSUPP : NFSERR_IO; 2851 else 2852 vput(nvp); 2853 } 2854 2855 if (!cookies && !error) 2856 error = NFSERR_PERM; 2857 if (!error) 2858 error = getret; 2859 if (error) { 2860 vrele(vp); 2861 if (cookies) 2862 free((void *)cookies, M_TEMP); 2863 free((void *)rbuf, M_TEMP); 2864 nfsm_reply(NFSX_V3POSTOPATTR); 2865 nfsm_srvpostop_attr(getret, &at); 2866 return (0); 2867 } 2868 if (io.uio_resid) { 2869 siz -= io.uio_resid; 2870 2871 /* 2872 * If nothing read, return eof 2873 * rpc reply 2874 */ 2875 if (siz == 0) { 2876 vrele(vp); 2877 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2878 2 * NFSX_UNSIGNED); 2879 nfsm_srvpostop_attr(getret, &at); 2880 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 2881 txdr_hyper(at.va_filerev, tl); 2882 tl += 2; 2883 *tl++ = nfs_false; 2884 *tl = nfs_true; 2885 free((void *)cookies, M_TEMP); 2886 free((void *)rbuf, M_TEMP); 2887 return (0); 2888 } 2889 } 2890 2891 /* 2892 * Check for degenerate cases of nothing useful read. 2893 * If so go try again 2894 */ 2895 cpos = rbuf; 2896 cend = rbuf + siz; 2897 dp = (struct dirent *)cpos; 2898 cookiep = cookies; 2899 2900 while (cpos < cend && ncookies > 0 && 2901 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) { 2902 cpos += dp->d_reclen; 2903 dp = (struct dirent *)cpos; 2904 cookiep++; 2905 ncookies--; 2906 } 2907 if (cpos >= cend || ncookies == 0) { 2908 toff = off; 2909 siz = fullsiz; 2910 free(cookies, M_TEMP); 2911 cookies = NULL; 2912 goto again; 2913 } 2914 2915 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED; 2916 nfsm_reply(cnt); 2917 nfsm_srvpostop_attr(getret, &at); 2918 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2919 txdr_hyper(at.va_filerev, tl); 2920 mp = mp2 = mb; 2921 bp = bpos; 2922 be = bp + M_TRAILINGSPACE(mp); 2923 2924 /* Loop through the records and build reply */ 2925 while (cpos < cend && ncookies > 0) { 2926 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { 2927 nfsrvfh_t nnsfh; 2928 2929 nlen = dp->d_namlen; 2930 rem = nfsm_rndup(nlen)-nlen; 2931 2932 /* 2933 * For readdir_and_lookup get the vnode using 2934 * the file number. 2935 */ 2936 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp)) 2937 goto invalid; 2938 if (nfsrv_composefh(nvp, &nnsfh, true)) { 2939 vput(nvp); 2940 goto invalid; 2941 } 2942 if (VOP_GETATTR(nvp, vap, cred)) { 2943 vput(nvp); 2944 goto invalid; 2945 } 2946 vput(nvp); 2947 2948 /* 2949 * If either the dircount or maxcount will be 2950 * exceeded, get out now. Both of these lengths 2951 * are calculated conservatively, including all 2952 * XDR overheads. 2953 */ 2954 len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH + 2955 NFSX_V3POSTOPATTR); 2956 dirlen += (6 * NFSX_UNSIGNED + nlen + rem); 2957 if (len > cnt || dirlen > fullsiz) { 2958 eofflag = 0; 2959 break; 2960 } 2961 2962 /* 2963 * Build the directory record xdr from 2964 * the dirent entry. 2965 */ 2966 fp = (struct nfs_fattr *)&fl.fl_fattr; 2967 nfsm_srvfillattr(vap, fp); 2968 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH); 2969 fl.fl_fhok = nfs_true; 2970 fl.fl_postopok = nfs_true; 2971 txdr_hyper(*cookiep, fl.fl_off.nfsuquad); 2972 2973 nfsm_clget; 2974 *tl = nfs_true; 2975 bp += NFSX_UNSIGNED; 2976 nfsm_clget; 2977 *tl = txdr_unsigned(dp->d_fileno >> 32); 2978 bp += NFSX_UNSIGNED; 2979 nfsm_clget; 2980 *tl = txdr_unsigned(dp->d_fileno); 2981 bp += NFSX_UNSIGNED; 2982 nfsm_clget; 2983 *tl = txdr_unsigned(nlen); 2984 bp += NFSX_UNSIGNED; 2985 2986 /* And loop around copying the name */ 2987 xfer = nlen; 2988 cp = dp->d_name; 2989 while (xfer > 0) { 2990 nfsm_clget; 2991 if ((bp + xfer) > be) 2992 tsiz = be - bp; 2993 else 2994 tsiz = xfer; 2995 memcpy(bp, cp, tsiz); 2996 bp += tsiz; 2997 xfer -= tsiz; 2998 if (xfer > 0) 2999 cp += tsiz; 3000 } 3001 /* And null pad to an int32_t boundary */ 3002 for (i = 0; i < rem; i++) 3003 *bp++ = '\0'; 3004 3005 /* 3006 * Now copy the flrep structure out. 3007 */ 3008 xfer = sizeof(struct flrep); 3009 cp = (void *)&fl; 3010 while (xfer > 0) { 3011 nfsm_clget; 3012 if ((bp + xfer) > be) 3013 tsiz = be - bp; 3014 else 3015 tsiz = xfer; 3016 memcpy(bp, cp, tsiz); 3017 bp += tsiz; 3018 xfer -= tsiz; 3019 if (xfer > 0) 3020 cp += tsiz; 3021 } 3022 3023 /* 3024 * ... and filehandle. 3025 */ 3026 xfer = NFSRVFH_SIZE(&nnsfh); 3027 cp = NFSRVFH_DATA(&nnsfh); 3028 while (xfer > 0) { 3029 nfsm_clget; 3030 if ((bp + xfer) > be) 3031 tsiz = be - bp; 3032 else 3033 tsiz = xfer; 3034 memcpy(bp, cp, tsiz); 3035 bp += tsiz; 3036 xfer -= tsiz; 3037 if (xfer > 0) 3038 cp += tsiz; 3039 } 3040 } 3041 invalid: 3042 cpos += dp->d_reclen; 3043 dp = (struct dirent *)cpos; 3044 cookiep++; 3045 ncookies--; 3046 } 3047 vrele(vp); 3048 nfsm_clget; 3049 *tl = nfs_false; 3050 bp += NFSX_UNSIGNED; 3051 nfsm_clget; 3052 if (eofflag) 3053 *tl = nfs_true; 3054 else 3055 *tl = nfs_false; 3056 bp += NFSX_UNSIGNED; 3057 if (mp != mb) { 3058 if (bp < be) 3059 mp->m_len = bp - mtod(mp, char *); 3060 } else 3061 mp->m_len += bp - bpos; 3062 free((void *)cookies, M_TEMP); 3063 free((void *)rbuf, M_TEMP); 3064 nfsm_srvdone; 3065 } 3066 3067 /* 3068 * nfs commit service 3069 */ 3070 int 3071 nfsrv_commit(nfsd, slp, lwp, mrq) 3072 struct nfsrv_descript *nfsd; 3073 struct nfssvc_sock *slp; 3074 struct lwp *lwp; 3075 struct mbuf **mrq; 3076 { 3077 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3078 struct mbuf *nam = nfsd->nd_nam; 3079 char *dpos = nfsd->nd_dpos; 3080 kauth_cred_t cred = nfsd->nd_cr; 3081 struct vattr bfor, aft; 3082 struct vnode *vp; 3083 nfsrvfh_t nsfh; 3084 u_int32_t *tl; 3085 int32_t t1; 3086 char *bpos; 3087 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cache = 0; 3088 uint32_t cnt; 3089 char *cp2; 3090 struct mbuf *mb, *mreq; 3091 u_quad_t frev, off, end; 3092 3093 nfsm_srvmtofh(&nsfh); 3094 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED); 3095 3096 off = fxdr_hyper(tl); 3097 tl += 2; 3098 cnt = fxdr_unsigned(uint32_t, *tl); 3099 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3100 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3101 if (error) { 3102 nfsm_reply(2 * NFSX_UNSIGNED); 3103 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3104 return (0); 3105 } 3106 for_ret = VOP_GETATTR(vp, &bfor, cred); 3107 end = (cnt > 0) ? off + cnt : vp->v_size; 3108 if (end < off || end > vp->v_size) 3109 end = vp->v_size; 3110 if (off < vp->v_size) 3111 error = VOP_FSYNC(vp, cred, FSYNC_WAIT, off, end); 3112 /* else error == 0, from nfsrv_fhtovp() */ 3113 aft_ret = VOP_GETATTR(vp, &aft, cred); 3114 vput(vp); 3115 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF); 3116 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); 3117 if (!error) { 3118 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF); 3119 *tl++ = txdr_unsigned(boottime.tv_sec); 3120 *tl = txdr_unsigned(boottime.tv_usec); 3121 } else { 3122 return (0); 3123 } 3124 nfsm_srvdone; 3125 } 3126 3127 /* 3128 * nfs statfs service 3129 */ 3130 int 3131 nfsrv_statfs(nfsd, slp, lwp, mrq) 3132 struct nfsrv_descript *nfsd; 3133 struct nfssvc_sock *slp; 3134 struct lwp *lwp; 3135 struct mbuf **mrq; 3136 { 3137 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3138 struct mbuf *nam = nfsd->nd_nam; 3139 char *dpos = nfsd->nd_dpos; 3140 kauth_cred_t cred = nfsd->nd_cr; 3141 struct statvfs *sf = NULL; 3142 struct nfs_statfs *sfp; 3143 u_int32_t *tl; 3144 int32_t t1; 3145 char *bpos; 3146 int error = 0, rdonly, cache = 0, getret = 1; 3147 int v3 = (nfsd->nd_flag & ND_NFSV3); 3148 char *cp2; 3149 struct mbuf *mb, *mreq; 3150 struct vnode *vp; 3151 struct vattr at; 3152 nfsrvfh_t nsfh; 3153 u_quad_t frev, tval; 3154 3155 nfsm_srvmtofh(&nsfh); 3156 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3157 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3158 if (error) { 3159 nfsm_reply(NFSX_UNSIGNED); 3160 nfsm_srvpostop_attr(getret, &at); 3161 return (0); 3162 } 3163 sf = malloc(sizeof(*sf), M_TEMP, M_WAITOK); 3164 error = VFS_STATVFS(vp->v_mount, sf); 3165 getret = VOP_GETATTR(vp, &at, cred); 3166 vput(vp); 3167 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3)); 3168 if (v3) 3169 nfsm_srvpostop_attr(getret, &at); 3170 if (error) { 3171 free(sf, M_TEMP); 3172 return (0); 3173 } 3174 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3)); 3175 if (v3) { 3176 tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_frsize); 3177 txdr_hyper(tval, &sfp->sf_tbytes); 3178 tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_frsize); 3179 txdr_hyper(tval, &sfp->sf_fbytes); 3180 tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_frsize); 3181 txdr_hyper(tval, &sfp->sf_abytes); 3182 tval = (u_quad_t)sf->f_files; 3183 txdr_hyper(tval, &sfp->sf_tfiles); 3184 tval = (u_quad_t)sf->f_ffree; 3185 txdr_hyper(tval, &sfp->sf_ffiles); 3186 txdr_hyper(tval, &sfp->sf_afiles); 3187 sfp->sf_invarsec = 0; 3188 } else { 3189 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA); 3190 sfp->sf_bsize = txdr_unsigned(sf->f_frsize); 3191 sfp->sf_blocks = txdr_unsigned(sf->f_blocks); 3192 sfp->sf_bfree = txdr_unsigned(sf->f_bfree); 3193 sfp->sf_bavail = txdr_unsigned(sf->f_bavail); 3194 } 3195 nfsmout: 3196 if (sf) 3197 free(sf, M_TEMP); 3198 return error; 3199 } 3200 3201 /* 3202 * nfs fsinfo service 3203 */ 3204 int 3205 nfsrv_fsinfo(nfsd, slp, lwp, mrq) 3206 struct nfsrv_descript *nfsd; 3207 struct nfssvc_sock *slp; 3208 struct lwp *lwp; 3209 struct mbuf **mrq; 3210 { 3211 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3212 struct mbuf *nam = nfsd->nd_nam; 3213 char *dpos = nfsd->nd_dpos; 3214 kauth_cred_t cred = nfsd->nd_cr; 3215 u_int32_t *tl; 3216 struct nfsv3_fsinfo *sip; 3217 int32_t t1; 3218 char *bpos; 3219 int error = 0, rdonly, cache = 0, getret = 1; 3220 uint32_t maxdata; 3221 char *cp2; 3222 struct mbuf *mb, *mreq; 3223 struct vnode *vp; 3224 struct vattr at; 3225 nfsrvfh_t nsfh; 3226 u_quad_t frev, maxfsize; 3227 struct statvfs *sb; 3228 3229 nfsm_srvmtofh(&nsfh); 3230 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3231 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3232 if (error) { 3233 nfsm_reply(NFSX_UNSIGNED); 3234 nfsm_srvpostop_attr(getret, &at); 3235 return (0); 3236 } 3237 3238 /* XXX Try to make a guess on the max file size. */ 3239 sb = malloc(sizeof(*sb), M_TEMP, M_WAITOK); 3240 VFS_STATVFS(vp->v_mount, sb); 3241 maxfsize = (u_quad_t)0x80000000 * sb->f_frsize - 1; 3242 free(sb, M_TEMP); 3243 3244 getret = VOP_GETATTR(vp, &at, cred); 3245 vput(vp); 3246 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO); 3247 nfsm_srvpostop_attr(getret, &at); 3248 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO); 3249 3250 /* 3251 * XXX 3252 * There should be file system VFS OP(s) to get this information. 3253 * For now, assume ufs. 3254 */ 3255 if (slp->ns_so->so_type == SOCK_DGRAM) 3256 maxdata = NFS_MAXDGRAMDATA; 3257 else 3258 maxdata = NFS_MAXDATA; 3259 sip->fs_rtmax = txdr_unsigned(maxdata); 3260 sip->fs_rtpref = txdr_unsigned(maxdata); 3261 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE); 3262 sip->fs_wtmax = txdr_unsigned(maxdata); 3263 sip->fs_wtpref = txdr_unsigned(maxdata); 3264 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE); 3265 sip->fs_dtpref = txdr_unsigned(maxdata); 3266 txdr_hyper(maxfsize, &sip->fs_maxfilesize); 3267 sip->fs_timedelta.nfsv3_sec = 0; 3268 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1); 3269 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK | 3270 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS | 3271 NFSV3FSINFO_CANSETTIME); 3272 nfsm_srvdone; 3273 } 3274 3275 /* 3276 * nfs pathconf service 3277 */ 3278 int 3279 nfsrv_pathconf(nfsd, slp, lwp, mrq) 3280 struct nfsrv_descript *nfsd; 3281 struct nfssvc_sock *slp; 3282 struct lwp *lwp; 3283 struct mbuf **mrq; 3284 { 3285 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; 3286 struct mbuf *nam = nfsd->nd_nam; 3287 char *dpos = nfsd->nd_dpos; 3288 kauth_cred_t cred = nfsd->nd_cr; 3289 u_int32_t *tl; 3290 struct nfsv3_pathconf *pc; 3291 int32_t t1; 3292 char *bpos; 3293 int error = 0, rdonly, cache = 0, getret = 1; 3294 register_t linkmax, namemax, chownres, notrunc; 3295 char *cp2; 3296 struct mbuf *mb, *mreq; 3297 struct vnode *vp; 3298 struct vattr at; 3299 nfsrvfh_t nsfh; 3300 u_quad_t frev; 3301 3302 nfsm_srvmtofh(&nsfh); 3303 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, 3304 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false); 3305 if (error) { 3306 nfsm_reply(NFSX_UNSIGNED); 3307 nfsm_srvpostop_attr(getret, &at); 3308 return (0); 3309 } 3310 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax); 3311 if (!error) 3312 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax); 3313 if (!error) 3314 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres); 3315 if (!error) 3316 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc); 3317 getret = VOP_GETATTR(vp, &at, cred); 3318 vput(vp); 3319 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF); 3320 nfsm_srvpostop_attr(getret, &at); 3321 if (error) 3322 return (0); 3323 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF); 3324 3325 pc->pc_linkmax = txdr_unsigned(linkmax); 3326 pc->pc_namemax = txdr_unsigned(namemax); 3327 pc->pc_notrunc = txdr_unsigned(notrunc); 3328 pc->pc_chownrestricted = txdr_unsigned(chownres); 3329 3330 /* 3331 * These should probably be supported by VOP_PATHCONF(), but 3332 * until msdosfs is exportable (why would you want to?), the 3333 * Unix defaults should be ok. 3334 */ 3335 pc->pc_caseinsensitive = nfs_false; 3336 pc->pc_casepreserving = nfs_true; 3337 nfsm_srvdone; 3338 } 3339 3340 /* 3341 * Null operation, used by clients to ping server 3342 */ 3343 /* ARGSUSED */ 3344 int 3345 nfsrv_null(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, 3346 struct lwp *lwp, struct mbuf **mrq) 3347 { 3348 struct mbuf *mrep = nfsd->nd_mrep; 3349 char *bpos; 3350 int error = NFSERR_RETVOID, cache = 0; 3351 struct mbuf *mb, *mreq; 3352 u_quad_t frev; 3353 3354 nfsm_reply(0); 3355 return (0); 3356 } 3357 3358 /* 3359 * No operation, used for obsolete procedures 3360 */ 3361 /* ARGSUSED */ 3362 int 3363 nfsrv_noop(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, 3364 struct lwp *lwp, struct mbuf **mrq) 3365 { 3366 struct mbuf *mrep = nfsd->nd_mrep; 3367 char *bpos; 3368 int error, cache = 0; 3369 struct mbuf *mb, *mreq; 3370 u_quad_t frev; 3371 3372 if (nfsd->nd_repstat) 3373 error = nfsd->nd_repstat; 3374 else 3375 error = EPROCUNAVAIL; 3376 nfsm_reply(0); 3377 return (0); 3378 } 3379 3380 /* 3381 * Perform access checking for vnodes obtained from file handles that would 3382 * refer to files already opened by a Unix client. You cannot just use 3383 * vn_writechk() and VOP_ACCESS() for two reasons. 3384 * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case 3385 * 2 - The owner is to be given access irrespective of mode bits for some 3386 * operations, so that processes that chmod after opening a file don't 3387 * break. I don't like this because it opens a security hole, but since 3388 * the nfs server opens a security hole the size of a barn door anyhow, 3389 * what the heck. 3390 * 3391 * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS() 3392 * will return EPERM instead of EACCESS. EPERM is always an error. 3393 */ 3394 int 3395 nfsrv_access(vp, flags, cred, rdonly, lwp, override) 3396 struct vnode *vp; 3397 int flags; 3398 kauth_cred_t cred; 3399 int rdonly; 3400 struct lwp *lwp; 3401 int override; 3402 { 3403 struct vattr vattr; 3404 int error; 3405 if (flags & VWRITE) { 3406 /* Just vn_writechk() changed to check rdonly */ 3407 /* 3408 * Disallow write attempts on read-only file systems; 3409 * unless the file is a socket or a block or character 3410 * device resident on the file system. 3411 */ 3412 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { 3413 switch (vp->v_type) { 3414 case VREG: 3415 case VDIR: 3416 case VLNK: 3417 return (EROFS); 3418 default: 3419 break; 3420 } 3421 } 3422 3423 /* 3424 * If the vnode is in use as a process's text, 3425 * we can't allow writing. 3426 */ 3427 if (vp->v_iflag & VI_TEXT) 3428 return (ETXTBSY); 3429 } 3430 error = VOP_GETATTR(vp, &vattr, cred); 3431 if (error) 3432 return (error); 3433 error = VOP_ACCESS(vp, flags, cred); 3434 /* 3435 * Allow certain operations for the owner (reads and writes 3436 * on files that are already open). 3437 */ 3438 if (override && error == EACCES && kauth_cred_geteuid(cred) == vattr.va_uid) 3439 error = 0; 3440 return error; 3441 } 3442
Indexes created Thu Oct 02 14:10:14 GMT 2025