nfs_subs.c revision 104908
1/* 2 * Copyright (c) 1989, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * This code is derived from software contributed to Berkeley by 6 * Rick Macklem at The University of Guelph. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. All advertising materials mentioning features or use of this software 17 * must display the following acknowledgement: 18 * This product includes software developed by the University of 19 * California, Berkeley and its contributors. 20 * 4. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * @(#)nfs_subs.c 8.8 (Berkeley) 5/22/95 37 */ 38 39#include <sys/cdefs.h> 40__FBSDID("$FreeBSD: head/sys/nfsclient/nfs_subs.c 104908 2002-10-11 14:58:34Z mike $"); 41 42/* 43 * These functions support the macros and help fiddle mbuf chains for 44 * the nfs op functions. They do things like create the rpc header and 45 * copy data between mbuf chains and uio lists. 46 */ 47 48#include <sys/param.h> 49#include <sys/systm.h> 50#include <sys/kernel.h> 51#include <sys/bio.h> 52#include <sys/buf.h> 53#include <sys/proc.h> 54#include <sys/mount.h> 55#include <sys/vnode.h> 56#include <sys/namei.h> 57#include <sys/mbuf.h> 58#include <sys/socket.h> 59#include <sys/stat.h> 60#include <sys/malloc.h> 61#include <sys/sysent.h> 62#include <sys/syscall.h> 63#include <sys/sysproto.h> 64 65#include <vm/vm.h> 66#include <vm/vm_object.h> 67#include <vm/vm_extern.h> 68#include <vm/uma.h> 69 70#include <nfs/rpcv2.h> 71#include <nfs/nfsproto.h> 72#include <nfsclient/nfs.h> 73#include <nfsclient/nfsnode.h> 74#include <nfs/xdr_subs.h> 75#include <nfsclient/nfsm_subs.h> 76#include <nfsclient/nfsmount.h> 77 78#include <netinet/in.h> 79 80/* 81 * Data items converted to xdr at startup, since they are constant 82 * This is kinda hokey, but may save a little time doing byte swaps 83 */ 84u_int32_t nfs_xdrneg1; 85u_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr, 86 rpc_mismatch, rpc_auth_unix, rpc_msgaccepted; 87u_int32_t nfs_true, nfs_false; 88 89/* And other global data */ 90static u_int32_t nfs_xid = 0; 91static enum vtype nv2tov_type[8]= { 92 VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON 93}; 94 95int nfs_ticks; 96int nfs_pbuf_freecnt = -1; /* start out unlimited */ 97 98struct nfs_reqq nfs_reqq; 99struct nfs_bufq nfs_bufq; 100 101static int nfs_prev_nfsclnt_sy_narg; 102static sy_call_t *nfs_prev_nfsclnt_sy_call; 103 104/* 105 * and the reverse mapping from generic to Version 2 procedure numbers 106 */ 107int nfsv2_procid[NFS_NPROCS] = { 108 NFSV2PROC_NULL, 109 NFSV2PROC_GETATTR, 110 NFSV2PROC_SETATTR, 111 NFSV2PROC_LOOKUP, 112 NFSV2PROC_NOOP, 113 NFSV2PROC_READLINK, 114 NFSV2PROC_READ, 115 NFSV2PROC_WRITE, 116 NFSV2PROC_CREATE, 117 NFSV2PROC_MKDIR, 118 NFSV2PROC_SYMLINK, 119 NFSV2PROC_CREATE, 120 NFSV2PROC_REMOVE, 121 NFSV2PROC_RMDIR, 122 NFSV2PROC_RENAME, 123 NFSV2PROC_LINK, 124 NFSV2PROC_READDIR, 125 NFSV2PROC_NOOP, 126 NFSV2PROC_STATFS, 127 NFSV2PROC_NOOP, 128 NFSV2PROC_NOOP, 129 NFSV2PROC_NOOP, 130 NFSV2PROC_NOOP, 131}; 132 133LIST_HEAD(nfsnodehashhead, nfsnode); 134 135/* 136 * Create the header for an rpc request packet 137 * The hsiz is the size of the rest of the nfs request header. 138 * (just used to decide if a cluster is a good idea) 139 */ 140struct mbuf * 141nfsm_reqhead(struct vnode *vp, u_long procid, int hsiz) 142{ 143 struct mbuf *mb; 144 145 MGET(mb, M_TRYWAIT, MT_DATA); 146 if (hsiz >= MINCLSIZE) 147 MCLGET(mb, M_TRYWAIT); 148 mb->m_len = 0; 149 return (mb); 150} 151 152/* 153 * Build the RPC header and fill in the authorization info. 154 * The authorization string argument is only used when the credentials 155 * come from outside of the kernel. 156 * Returns the head of the mbuf list. 157 */ 158struct mbuf * 159nfsm_rpchead(struct ucred *cr, int nmflag, int procid, int auth_type, 160 int auth_len, struct mbuf *mrest, int mrest_len, struct mbuf **mbp, 161 u_int32_t *xidp) 162{ 163 struct mbuf *mb; 164 u_int32_t *tl; 165 caddr_t bpos; 166 int i; 167 struct mbuf *mreq; 168 int grpsiz, authsiz; 169 170 authsiz = nfsm_rndup(auth_len); 171 MGETHDR(mb, M_TRYWAIT, MT_DATA); 172 if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) { 173 MCLGET(mb, M_TRYWAIT); 174 } else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) { 175 MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED); 176 } else { 177 MH_ALIGN(mb, 8 * NFSX_UNSIGNED); 178 } 179 mb->m_len = 0; 180 mreq = mb; 181 bpos = mtod(mb, caddr_t); 182 183 /* 184 * First the RPC header. 185 */ 186 tl = nfsm_build(u_int32_t *, 8 * NFSX_UNSIGNED); 187 188 /* Get a pretty random xid to start with */ 189 if (!nfs_xid) 190 nfs_xid = random(); 191 /* 192 * Skip zero xid if it should ever happen. 193 */ 194 if (++nfs_xid == 0) 195 nfs_xid++; 196 197 *tl++ = *xidp = txdr_unsigned(nfs_xid); 198 *tl++ = rpc_call; 199 *tl++ = rpc_vers; 200 *tl++ = txdr_unsigned(NFS_PROG); 201 if (nmflag & NFSMNT_NFSV3) { 202 *tl++ = txdr_unsigned(NFS_VER3); 203 *tl++ = txdr_unsigned(procid); 204 } else { 205 *tl++ = txdr_unsigned(NFS_VER2); 206 *tl++ = txdr_unsigned(nfsv2_procid[procid]); 207 } 208 209 /* 210 * And then the authorization cred. 211 */ 212 *tl++ = txdr_unsigned(auth_type); 213 *tl = txdr_unsigned(authsiz); 214 switch (auth_type) { 215 case RPCAUTH_UNIX: 216 tl = nfsm_build(u_int32_t *, auth_len); 217 *tl++ = 0; /* stamp ?? */ 218 *tl++ = 0; /* NULL hostname */ 219 *tl++ = txdr_unsigned(cr->cr_uid); 220 *tl++ = txdr_unsigned(cr->cr_groups[0]); 221 grpsiz = (auth_len >> 2) - 5; 222 *tl++ = txdr_unsigned(grpsiz); 223 for (i = 1; i <= grpsiz; i++) 224 *tl++ = txdr_unsigned(cr->cr_groups[i]); 225 break; 226 } 227 228 /* 229 * And the verifier... 230 */ 231 tl = nfsm_build(u_int32_t *, 2 * NFSX_UNSIGNED); 232 *tl++ = txdr_unsigned(RPCAUTH_NULL); 233 *tl = 0; 234 mb->m_next = mrest; 235 mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len; 236 mreq->m_pkthdr.rcvif = NULL; 237 *mbp = mb; 238 return (mreq); 239} 240 241/* 242 * copies a uio scatter/gather list to an mbuf chain. 243 * NOTE: can ony handle iovcnt == 1 244 */ 245int 246nfsm_uiotombuf(struct uio *uiop, struct mbuf **mq, int siz, caddr_t *bpos) 247{ 248 char *uiocp; 249 struct mbuf *mp, *mp2; 250 int xfer, left, mlen; 251 int uiosiz, clflg, rem; 252 char *cp; 253 254#ifdef DIAGNOSTIC 255 if (uiop->uio_iovcnt != 1) 256 panic("nfsm_uiotombuf: iovcnt != 1"); 257#endif 258 259 if (siz > MLEN) /* or should it >= MCLBYTES ?? */ 260 clflg = 1; 261 else 262 clflg = 0; 263 rem = nfsm_rndup(siz)-siz; 264 mp = mp2 = *mq; 265 while (siz > 0) { 266 left = uiop->uio_iov->iov_len; 267 uiocp = uiop->uio_iov->iov_base; 268 if (left > siz) 269 left = siz; 270 uiosiz = left; 271 while (left > 0) { 272 mlen = M_TRAILINGSPACE(mp); 273 if (mlen == 0) { 274 MGET(mp, M_TRYWAIT, MT_DATA); 275 if (clflg) 276 MCLGET(mp, M_TRYWAIT); 277 mp->m_len = 0; 278 mp2->m_next = mp; 279 mp2 = mp; 280 mlen = M_TRAILINGSPACE(mp); 281 } 282 xfer = (left > mlen) ? mlen : left; 283#ifdef notdef 284 /* Not Yet.. */ 285 if (uiop->uio_iov->iov_op != NULL) 286 (*(uiop->uio_iov->iov_op)) 287 (uiocp, mtod(mp, caddr_t)+mp->m_len, xfer); 288 else 289#endif 290 if (uiop->uio_segflg == UIO_SYSSPACE) 291 bcopy(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer); 292 else 293 copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer); 294 mp->m_len += xfer; 295 left -= xfer; 296 uiocp += xfer; 297 uiop->uio_offset += xfer; 298 uiop->uio_resid -= xfer; 299 } 300 uiop->uio_iov->iov_base = 301 (char *)uiop->uio_iov->iov_base + uiosiz; 302 uiop->uio_iov->iov_len -= uiosiz; 303 siz -= uiosiz; 304 } 305 if (rem > 0) { 306 if (rem > M_TRAILINGSPACE(mp)) { 307 MGET(mp, M_TRYWAIT, MT_DATA); 308 mp->m_len = 0; 309 mp2->m_next = mp; 310 } 311 cp = mtod(mp, caddr_t)+mp->m_len; 312 for (left = 0; left < rem; left++) 313 *cp++ = '\0'; 314 mp->m_len += rem; 315 *bpos = cp; 316 } else 317 *bpos = mtod(mp, caddr_t)+mp->m_len; 318 *mq = mp; 319 return (0); 320} 321 322/* 323 * Copy a string into mbufs for the hard cases... 324 */ 325int 326nfsm_strtmbuf(struct mbuf **mb, char **bpos, const char *cp, long siz) 327{ 328 struct mbuf *m1 = NULL, *m2; 329 long left, xfer, len, tlen; 330 u_int32_t *tl; 331 int putsize; 332 333 putsize = 1; 334 m2 = *mb; 335 left = M_TRAILINGSPACE(m2); 336 if (left > 0) { 337 tl = ((u_int32_t *)(*bpos)); 338 *tl++ = txdr_unsigned(siz); 339 putsize = 0; 340 left -= NFSX_UNSIGNED; 341 m2->m_len += NFSX_UNSIGNED; 342 if (left > 0) { 343 bcopy(cp, (caddr_t) tl, left); 344 siz -= left; 345 cp += left; 346 m2->m_len += left; 347 left = 0; 348 } 349 } 350 /* Loop around adding mbufs */ 351 while (siz > 0) { 352 MGET(m1, M_TRYWAIT, MT_DATA); 353 if (siz > MLEN) 354 MCLGET(m1, M_TRYWAIT); 355 m1->m_len = NFSMSIZ(m1); 356 m2->m_next = m1; 357 m2 = m1; 358 tl = mtod(m1, u_int32_t *); 359 tlen = 0; 360 if (putsize) { 361 *tl++ = txdr_unsigned(siz); 362 m1->m_len -= NFSX_UNSIGNED; 363 tlen = NFSX_UNSIGNED; 364 putsize = 0; 365 } 366 if (siz < m1->m_len) { 367 len = nfsm_rndup(siz); 368 xfer = siz; 369 if (xfer < len) 370 *(tl+(xfer>>2)) = 0; 371 } else { 372 xfer = len = m1->m_len; 373 } 374 bcopy(cp, (caddr_t) tl, xfer); 375 m1->m_len = len+tlen; 376 siz -= xfer; 377 cp += xfer; 378 } 379 *mb = m1; 380 *bpos = mtod(m1, caddr_t)+m1->m_len; 381 return (0); 382} 383 384/* 385 * Called once to initialize data structures... 386 */ 387int 388nfs_init(struct vfsconf *vfsp) 389{ 390 int i; 391 392 nfsmount_zone = uma_zcreate("NFSMOUNT", sizeof(struct nfsmount), 393 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); 394 rpc_vers = txdr_unsigned(RPC_VER2); 395 rpc_call = txdr_unsigned(RPC_CALL); 396 rpc_reply = txdr_unsigned(RPC_REPLY); 397 rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED); 398 rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED); 399 rpc_mismatch = txdr_unsigned(RPC_MISMATCH); 400 rpc_autherr = txdr_unsigned(RPC_AUTHERR); 401 rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX); 402 nfs_true = txdr_unsigned(TRUE); 403 nfs_false = txdr_unsigned(FALSE); 404 nfs_xdrneg1 = txdr_unsigned(-1); 405 nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000; 406 if (nfs_ticks < 1) 407 nfs_ticks = 1; 408 /* Ensure async daemons disabled */ 409 for (i = 0; i < NFS_MAXASYNCDAEMON; i++) { 410 nfs_iodwant[i] = NULL; 411 nfs_iodmount[i] = NULL; 412 } 413 nfs_nhinit(); /* Init the nfsnode table */ 414 415 /* 416 * Initialize reply list and start timer 417 */ 418 TAILQ_INIT(&nfs_reqq); 419 420 nfs_timer(0); 421 422 nfs_prev_nfsclnt_sy_narg = sysent[SYS_nfsclnt].sy_narg; 423 sysent[SYS_nfsclnt].sy_narg = 2; 424 nfs_prev_nfsclnt_sy_call = sysent[SYS_nfsclnt].sy_call; 425 sysent[SYS_nfsclnt].sy_call = (sy_call_t *)nfsclnt; 426 427 nfs_pbuf_freecnt = nswbuf / 2 + 1; 428 429 return (0); 430} 431 432int 433nfs_uninit(struct vfsconf *vfsp) 434{ 435 436 untimeout(nfs_timer, (void *)NULL, nfs_timer_handle); 437 sysent[SYS_nfsclnt].sy_narg = nfs_prev_nfsclnt_sy_narg; 438 sysent[SYS_nfsclnt].sy_call = nfs_prev_nfsclnt_sy_call; 439 return (0); 440} 441 442/* 443 * Attribute cache routines. 444 * nfs_loadattrcache() - loads or updates the cache contents from attributes 445 * that are on the mbuf list 446 * nfs_getattrcache() - returns valid attributes if found in cache, returns 447 * error otherwise 448 */ 449 450/* 451 * Load the attribute cache (that lives in the nfsnode entry) with 452 * the values on the mbuf list and 453 * Iff vap not NULL 454 * copy the attributes to *vaper 455 */ 456int 457nfs_loadattrcache(struct vnode **vpp, struct mbuf **mdp, caddr_t *dposp, 458 struct vattr *vaper, int dontshrink) 459{ 460 struct vnode *vp = *vpp; 461 struct vattr *vap; 462 struct nfs_fattr *fp; 463 struct nfsnode *np; 464 int32_t t1; 465 caddr_t cp2; 466 int rdev; 467 struct mbuf *md; 468 enum vtype vtyp; 469 u_short vmode; 470 struct timespec mtime; 471 int v3 = NFS_ISV3(vp); 472 473 md = *mdp; 474 t1 = (mtod(md, caddr_t) + md->m_len) - *dposp; 475 cp2 = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1); 476 if (cp2 == NULL) 477 return EBADRPC; 478 fp = (struct nfs_fattr *)cp2; 479 if (v3) { 480 vtyp = nfsv3tov_type(fp->fa_type); 481 vmode = fxdr_unsigned(u_short, fp->fa_mode); 482 rdev = makeudev(fxdr_unsigned(int, fp->fa3_rdev.specdata1), 483 fxdr_unsigned(int, fp->fa3_rdev.specdata2)); 484 fxdr_nfsv3time(&fp->fa3_mtime, &mtime); 485 } else { 486 vtyp = nfsv2tov_type(fp->fa_type); 487 vmode = fxdr_unsigned(u_short, fp->fa_mode); 488 /* 489 * XXX 490 * 491 * The duplicate information returned in fa_type and fa_mode 492 * is an ambiguity in the NFS version 2 protocol. 493 * 494 * VREG should be taken literally as a regular file. If a 495 * server intents to return some type information differently 496 * in the upper bits of the mode field (e.g. for sockets, or 497 * FIFOs), NFSv2 mandates fa_type to be VNON. Anyway, we 498 * leave the examination of the mode bits even in the VREG 499 * case to avoid breakage for bogus servers, but we make sure 500 * that there are actually type bits set in the upper part of 501 * fa_mode (and failing that, trust the va_type field). 502 * 503 * NFSv3 cleared the issue, and requires fa_mode to not 504 * contain any type information (while also introduing sockets 505 * and FIFOs for fa_type). 506 */ 507 if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0)) 508 vtyp = IFTOVT(vmode); 509 rdev = fxdr_unsigned(int32_t, fp->fa2_rdev); 510 fxdr_nfsv2time(&fp->fa2_mtime, &mtime); 511 512 /* 513 * Really ugly NFSv2 kludge. 514 */ 515 if (vtyp == VCHR && rdev == 0xffffffff) 516 vtyp = VFIFO; 517 } 518 519 /* 520 * If v_type == VNON it is a new node, so fill in the v_type, 521 * n_mtime fields. Check to see if it represents a special 522 * device, and if so, check for a possible alias. Once the 523 * correct vnode has been obtained, fill in the rest of the 524 * information. 525 */ 526 np = VTONFS(vp); 527 if (vp->v_type != vtyp) { 528 vp->v_type = vtyp; 529 if (vp->v_type == VFIFO) { 530 vp->v_op = fifo_nfsv2nodeop_p; 531 } 532 if (vp->v_type == VCHR || vp->v_type == VBLK) { 533 vp->v_op = spec_nfsv2nodeop_p; 534 vp = addaliasu(vp, rdev); 535 np->n_vnode = vp; 536 } 537 np->n_mtime = mtime.tv_sec; 538 } 539 vap = &np->n_vattr; 540 vap->va_type = vtyp; 541 vap->va_mode = (vmode & 07777); 542 vap->va_rdev = rdev; 543 vap->va_mtime = mtime; 544 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0]; 545 if (v3) { 546 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); 547 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); 548 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); 549 vap->va_size = fxdr_hyper(&fp->fa3_size); 550 vap->va_blocksize = NFS_FABLKSIZE; 551 vap->va_bytes = fxdr_hyper(&fp->fa3_used); 552 vap->va_fileid = fxdr_unsigned(int32_t, 553 fp->fa3_fileid.nfsuquad[1]); 554 fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime); 555 fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime); 556 vap->va_flags = 0; 557 vap->va_filerev = 0; 558 } else { 559 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); 560 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); 561 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); 562 vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size); 563 vap->va_blocksize = fxdr_unsigned(int32_t, fp->fa2_blocksize); 564 vap->va_bytes = (u_quad_t)fxdr_unsigned(int32_t, fp->fa2_blocks) 565 * NFS_FABLKSIZE; 566 vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid); 567 fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime); 568 vap->va_flags = 0; 569 vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t, 570 fp->fa2_ctime.nfsv2_sec); 571 vap->va_ctime.tv_nsec = 0; 572 vap->va_gen = fxdr_unsigned(u_int32_t, fp->fa2_ctime.nfsv2_usec); 573 vap->va_filerev = 0; 574 } 575 np->n_attrstamp = time_second; 576 if (vap->va_size != np->n_size) { 577 if (vap->va_type == VREG) { 578 if (dontshrink && vap->va_size < np->n_size) { 579 /* 580 * We've been told not to shrink the file; 581 * zero np->n_attrstamp to indicate that 582 * the attributes are stale. 583 */ 584 vap->va_size = np->n_size; 585 np->n_attrstamp = 0; 586 } else if (np->n_flag & NMODIFIED) { 587 if (vap->va_size < np->n_size) 588 vap->va_size = np->n_size; 589 else 590 np->n_size = vap->va_size; 591 } else { 592 np->n_size = vap->va_size; 593 } 594 vnode_pager_setsize(vp, np->n_size); 595 } else { 596 np->n_size = vap->va_size; 597 } 598 } 599 if (vaper != NULL) { 600 bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(*vap)); 601 if (np->n_flag & NCHG) { 602 if (np->n_flag & NACC) 603 vaper->va_atime = np->n_atim; 604 if (np->n_flag & NUPD) 605 vaper->va_mtime = np->n_mtim; 606 } 607 } 608 return (0); 609} 610 611#ifdef NFS_ACDEBUG 612#include <sys/sysctl.h> 613SYSCTL_DECL(_vfs_nfs); 614static int nfs_acdebug; 615SYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0, ""); 616#endif 617 618/* 619 * Check the time stamp 620 * If the cache is valid, copy contents to *vap and return 0 621 * otherwise return an error 622 */ 623int 624nfs_getattrcache(struct vnode *vp, struct vattr *vaper) 625{ 626 struct nfsnode *np; 627 struct vattr *vap; 628 struct nfsmount *nmp; 629 int timeo; 630 631 np = VTONFS(vp); 632 vap = &np->n_vattr; 633 nmp = VFSTONFS(vp->v_mount); 634 /* XXX n_mtime doesn't seem to be updated on a miss-and-reload */ 635 timeo = (time_second - np->n_mtime) / 10; 636 637#ifdef NFS_ACDEBUG 638 if (nfs_acdebug>1) 639 printf("nfs_getattrcache: initial timeo = %d\n", timeo); 640#endif 641 642 if (vap->va_type == VDIR) { 643 if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin) 644 timeo = nmp->nm_acdirmin; 645 else if (timeo > nmp->nm_acdirmax) 646 timeo = nmp->nm_acdirmax; 647 } else { 648 if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin) 649 timeo = nmp->nm_acregmin; 650 else if (timeo > nmp->nm_acregmax) 651 timeo = nmp->nm_acregmax; 652 } 653 654#ifdef NFS_ACDEBUG 655 if (nfs_acdebug > 2) 656 printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n", 657 nmp->nm_acregmin, nmp->nm_acregmax, 658 nmp->nm_acdirmin, nmp->nm_acdirmax); 659 660 if (nfs_acdebug) 661 printf("nfs_getattrcache: age = %d; final timeo = %d\n", 662 (time_second - np->n_attrstamp), timeo); 663#endif 664 665 if ((time_second - np->n_attrstamp) >= timeo) { 666 nfsstats.attrcache_misses++; 667 return (ENOENT); 668 } 669 nfsstats.attrcache_hits++; 670 if (vap->va_size != np->n_size) { 671 if (vap->va_type == VREG) { 672 if (np->n_flag & NMODIFIED) { 673 if (vap->va_size < np->n_size) 674 vap->va_size = np->n_size; 675 else 676 np->n_size = vap->va_size; 677 } else { 678 np->n_size = vap->va_size; 679 } 680 vnode_pager_setsize(vp, np->n_size); 681 } else { 682 np->n_size = vap->va_size; 683 } 684 } 685 bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr)); 686 if (np->n_flag & NCHG) { 687 if (np->n_flag & NACC) 688 vaper->va_atime = np->n_atim; 689 if (np->n_flag & NUPD) 690 vaper->va_mtime = np->n_mtim; 691 } 692 return (0); 693} 694 695static nfsuint64 nfs_nullcookie = { { 0, 0 } }; 696/* 697 * This function finds the directory cookie that corresponds to the 698 * logical byte offset given. 699 */ 700nfsuint64 * 701nfs_getcookie(struct nfsnode *np, off_t off, int add) 702{ 703 struct nfsdmap *dp, *dp2; 704 int pos; 705 706 pos = (uoff_t)off / NFS_DIRBLKSIZ; 707 if (pos == 0 || off < 0) { 708#ifdef DIAGNOSTIC 709 if (add) 710 panic("nfs getcookie add at <= 0"); 711#endif 712 return (&nfs_nullcookie); 713 } 714 pos--; 715 dp = LIST_FIRST(&np->n_cookies); 716 if (!dp) { 717 if (add) { 718 MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap), 719 M_NFSDIROFF, M_WAITOK); 720 dp->ndm_eocookie = 0; 721 LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list); 722 } else 723 return (NULL); 724 } 725 while (pos >= NFSNUMCOOKIES) { 726 pos -= NFSNUMCOOKIES; 727 if (LIST_NEXT(dp, ndm_list)) { 728 if (!add && dp->ndm_eocookie < NFSNUMCOOKIES && 729 pos >= dp->ndm_eocookie) 730 return (NULL); 731 dp = LIST_NEXT(dp, ndm_list); 732 } else if (add) { 733 MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap), 734 M_NFSDIROFF, M_WAITOK); 735 dp2->ndm_eocookie = 0; 736 LIST_INSERT_AFTER(dp, dp2, ndm_list); 737 dp = dp2; 738 } else 739 return (NULL); 740 } 741 if (pos >= dp->ndm_eocookie) { 742 if (add) 743 dp->ndm_eocookie = pos + 1; 744 else 745 return (NULL); 746 } 747 return (&dp->ndm_cookies[pos]); 748} 749 750/* 751 * Invalidate cached directory information, except for the actual directory 752 * blocks (which are invalidated separately). 753 * Done mainly to avoid the use of stale offset cookies. 754 */ 755void 756nfs_invaldir(struct vnode *vp) 757{ 758 struct nfsnode *np = VTONFS(vp); 759 760#ifdef DIAGNOSTIC 761 if (vp->v_type != VDIR) 762 panic("nfs: invaldir not dir"); 763#endif 764 np->n_direofoffset = 0; 765 np->n_cookieverf.nfsuquad[0] = 0; 766 np->n_cookieverf.nfsuquad[1] = 0; 767 if (LIST_FIRST(&np->n_cookies)) 768 LIST_FIRST(&np->n_cookies)->ndm_eocookie = 0; 769} 770 771/* 772 * The write verifier has changed (probably due to a server reboot), so all 773 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the 774 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT 775 * and B_CLUSTEROK flags. Once done the new write verifier can be set for the 776 * mount point. 777 * 778 * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data 779 * writes are not clusterable. 780 */ 781void 782nfs_clearcommit(struct mount *mp) 783{ 784 struct vnode *vp, *nvp; 785 struct buf *bp, *nbp; 786 int s; 787 788 GIANT_REQUIRED; 789 790 s = splbio(); 791 mtx_lock(&mntvnode_mtx); 792loop: 793 for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist); vp; vp = nvp) { 794 if (vp->v_mount != mp) /* Paranoia */ 795 goto loop; 796 nvp = TAILQ_NEXT(vp, v_nmntvnodes); 797 VI_LOCK(vp); 798 mtx_unlock(&mntvnode_mtx); 799 for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) { 800 nbp = TAILQ_NEXT(bp, b_vnbufs); 801 if (BUF_REFCNT(bp) == 0 && 802 (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT)) 803 == (B_DELWRI | B_NEEDCOMMIT)) 804 bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK); 805 } 806 VI_UNLOCK(vp); 807 mtx_lock(&mntvnode_mtx); 808 } 809 mtx_unlock(&mntvnode_mtx); 810 splx(s); 811} 812 813/* 814 * Helper functions for former macros. Some of these should be 815 * moved to their callers. 816 */ 817 818int 819nfsm_mtofh_xx(struct vnode *d, struct vnode **v, int v3, int *f, 820 struct mbuf **md, caddr_t *dpos) 821{ 822 struct nfsnode *ttnp; 823 struct vnode *ttvp; 824 nfsfh_t *ttfhp; 825 u_int32_t *tl; 826 int ttfhsize; 827 int t1; 828 829 if (v3) { 830 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos); 831 if (tl == NULL) 832 return EBADRPC; 833 *f = fxdr_unsigned(int, *tl); 834 } else 835 *f = 1; 836 if (*f) { 837 t1 = nfsm_getfh_xx(&ttfhp, &ttfhsize, (v3), md, dpos); 838 if (t1 != 0) 839 return t1; 840 t1 = nfs_nget(d->v_mount, ttfhp, ttfhsize, &ttnp); 841 if (t1 != 0) 842 return t1; 843 *v = NFSTOV(ttnp); 844 } 845 if (v3) { 846 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos); 847 if (tl == NULL) 848 return EBADRPC; 849 if (*f) 850 *f = fxdr_unsigned(int, *tl); 851 else if (fxdr_unsigned(int, *tl)) 852 nfsm_adv_xx(NFSX_V3FATTR, md, dpos); 853 } 854 if (*f) { 855 ttvp = *v; 856 t1 = nfs_loadattrcache(&ttvp, md, dpos, NULL, 0); 857 if (t1) 858 return t1; 859 *v = ttvp; 860 } 861 return 0; 862} 863 864int 865nfsm_getfh_xx(nfsfh_t **f, int *s, int v3, struct mbuf **md, caddr_t *dpos) 866{ 867 u_int32_t *tl; 868 869 if (v3) { 870 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos); 871 if (tl == NULL) 872 return EBADRPC; 873 *s = fxdr_unsigned(int, *tl); 874 if (*s <= 0 || *s > NFSX_V3FHMAX) 875 return EBADRPC; 876 } else 877 *s = NFSX_V2FH; 878 *f = nfsm_dissect_xx(nfsm_rndup(*s), md, dpos); 879 if (*f == NULL) 880 return EBADRPC; 881 else 882 return 0; 883} 884 885 886int 887nfsm_loadattr_xx(struct vnode **v, struct vattr *va, struct mbuf **md, 888 caddr_t *dpos) 889{ 890 int t1; 891 892 struct vnode *ttvp = *v; 893 t1 = nfs_loadattrcache(&ttvp, md, dpos, va, 0); 894 if (t1 != 0) 895 return t1; 896 *v = ttvp; 897 return 0; 898} 899 900int 901nfsm_postop_attr_xx(struct vnode **v, int *f, struct mbuf **md, 902 caddr_t *dpos) 903{ 904 u_int32_t *tl; 905 int t1; 906 907 struct vnode *ttvp = *v; 908 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos); 909 if (tl == NULL) 910 return EBADRPC; 911 *f = fxdr_unsigned(int, *tl); 912 if (*f != 0) { 913 t1 = nfs_loadattrcache(&ttvp, md, dpos, NULL, 1); 914 if (t1 != 0) { 915 *f = 0; 916 return t1; 917 } 918 *v = ttvp; 919 } 920 return 0; 921} 922 923int 924nfsm_wcc_data_xx(struct vnode **v, int *f, struct mbuf **md, caddr_t *dpos) 925{ 926 u_int32_t *tl; 927 int ttattrf, ttretf = 0; 928 int t1; 929 930 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos); 931 if (tl == NULL) 932 return EBADRPC; 933 if (*tl == nfs_true) { 934 tl = nfsm_dissect_xx(6 * NFSX_UNSIGNED, md, dpos); 935 if (tl == NULL) 936 return EBADRPC; 937 if (*f) 938 ttretf = (VTONFS(*v)->n_mtime == 939 fxdr_unsigned(u_int32_t, *(tl + 2))); 940 } 941 t1 = nfsm_postop_attr_xx(v, &ttattrf, md, dpos); 942 if (t1) 943 return t1; 944 if (*f) 945 *f = ttretf; 946 else 947 *f = ttattrf; 948 return 0; 949} 950 951int 952nfsm_strtom_xx(const char *a, int s, int m, struct mbuf **mb, caddr_t *bpos) 953{ 954 u_int32_t *tl; 955 int t1; 956 957 if (s > m) 958 return ENAMETOOLONG; 959 t1 = nfsm_rndup(s) + NFSX_UNSIGNED; 960 if (t1 <= M_TRAILINGSPACE(*mb)) { 961 tl = nfsm_build_xx(t1, mb, bpos); 962 *tl++ = txdr_unsigned(s); 963 *(tl + ((t1 >> 2) - 2)) = 0; 964 bcopy(a, tl, s); 965 } else { 966 t1 = nfsm_strtmbuf(mb, bpos, a, s); 967 if (t1 != 0) 968 return t1; 969 } 970 return 0; 971} 972 973int 974nfsm_fhtom_xx(struct vnode *v, int v3, struct mbuf **mb, caddr_t *bpos) 975{ 976 u_int32_t *tl; 977 int t1; 978 caddr_t cp; 979 980 if (v3) { 981 t1 = nfsm_rndup(VTONFS(v)->n_fhsize) + NFSX_UNSIGNED; 982 if (t1 < M_TRAILINGSPACE(*mb)) { 983 tl = nfsm_build_xx(t1, mb, bpos); 984 *tl++ = txdr_unsigned(VTONFS(v)->n_fhsize); 985 *(tl + ((t1 >> 2) - 2)) = 0; 986 bcopy(VTONFS(v)->n_fhp, tl, VTONFS(v)->n_fhsize); 987 } else { 988 t1 = nfsm_strtmbuf(mb, bpos, 989 (const char *)VTONFS(v)->n_fhp, 990 VTONFS(v)->n_fhsize); 991 if (t1 != 0) 992 return t1; 993 } 994 } else { 995 cp = nfsm_build_xx(NFSX_V2FH, mb, bpos); 996 bcopy(VTONFS(v)->n_fhp, cp, NFSX_V2FH); 997 } 998 return 0; 999} 1000 1001void 1002nfsm_v3attrbuild_xx(struct vattr *va, int full, struct mbuf **mb, 1003 caddr_t *bpos) 1004{ 1005 u_int32_t *tl; 1006 1007 if (va->va_mode != (mode_t)VNOVAL) { 1008 tl = nfsm_build_xx(2 * NFSX_UNSIGNED, mb, bpos); 1009 *tl++ = nfs_true; 1010 *tl = txdr_unsigned(va->va_mode); 1011 } else { 1012 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos); 1013 *tl = nfs_false; 1014 } 1015 if (full && va->va_uid != (uid_t)VNOVAL) { 1016 tl = nfsm_build_xx(2 * NFSX_UNSIGNED, mb, bpos); 1017 *tl++ = nfs_true; 1018 *tl = txdr_unsigned(va->va_uid); 1019 } else { 1020 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos); 1021 *tl = nfs_false; 1022 } 1023 if (full && va->va_gid != (gid_t)VNOVAL) { 1024 tl = nfsm_build_xx(2 * NFSX_UNSIGNED, mb, bpos); 1025 *tl++ = nfs_true; 1026 *tl = txdr_unsigned(va->va_gid); 1027 } else { 1028 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos); 1029 *tl = nfs_false; 1030 } 1031 if (full && va->va_size != VNOVAL) { 1032 tl = nfsm_build_xx(3 * NFSX_UNSIGNED, mb, bpos); 1033 *tl++ = nfs_true; 1034 txdr_hyper(va->va_size, tl); 1035 } else { 1036 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos); 1037 *tl = nfs_false; 1038 } 1039 if (va->va_atime.tv_sec != VNOVAL) { 1040 if (va->va_atime.tv_sec != time_second) { 1041 tl = nfsm_build_xx(3 * NFSX_UNSIGNED, mb, bpos); 1042 *tl++ = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT); 1043 txdr_nfsv3time(&va->va_atime, tl); 1044 } else { 1045 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos); 1046 *tl = txdr_unsigned(NFSV3SATTRTIME_TOSERVER); 1047 } 1048 } else { 1049 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos); 1050 *tl = txdr_unsigned(NFSV3SATTRTIME_DONTCHANGE); 1051 } 1052 if (va->va_mtime.tv_sec != VNOVAL) { 1053 if (va->va_mtime.tv_sec != time_second) { 1054 tl = nfsm_build_xx(3 * NFSX_UNSIGNED, mb, bpos); 1055 *tl++ = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT); 1056 txdr_nfsv3time(&va->va_mtime, tl); 1057 } else { 1058 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos); 1059 *tl = txdr_unsigned(NFSV3SATTRTIME_TOSERVER); 1060 } 1061 } else { 1062 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos); 1063 *tl = txdr_unsigned(NFSV3SATTRTIME_DONTCHANGE); 1064 } 1065} 1066