1 /* 2 * Copyright (c) 1989, 1993, 1995 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_vfsops.c 8.12 (Berkeley) 5/20/95 37 * $FreeBSD: src/sys/nfs/nfs_vfsops.c,v 1.91.2.7 2003/01/27 20:04:08 dillon Exp $ 38 */ 39 40 #include "opt_bootp.h" 41 42 #include <sys/param.h> 43 #include <sys/sockio.h> 44 #include <sys/proc.h> 45 #include <sys/vnode.h> 46 #include <sys/kernel.h> 47 #include <sys/sysctl.h> 48 #include <sys/malloc.h> 49 #include <sys/mount.h> 50 #include <sys/mbuf.h> 51 #include <sys/socket.h> 52 #include <sys/socketvar.h> 53 #include <sys/systm.h> 54 55 #include <vm/vm.h> 56 #include <vm/vm_extern.h> 57 #include <vm/vm_zone.h> 58 59 #include <net/if.h> 60 #include <net/route.h> 61 #include <netinet/in.h> 62 63 #include <nfs/rpcv2.h> 64 #include <nfs/nfsproto.h> 65 #include <nfs/nfs.h> 66 #include <nfs/nfsnode.h> 67 #include <nfs/nfsmount.h> 68 #include <nfs/xdr_subs.h> 69 #include <nfs/nfsm_subs.h> 70 #include <nfs/nfsdiskless.h> 71 #include <nfs/nqnfs.h> 72 73 extern int nfs_mountroot __P((struct mount *mp)); 74 75 extern int nfs_ticks; 76 77 MALLOC_DEFINE(M_NFSREQ, "NFS req", "NFS request header"); 78 MALLOC_DEFINE(M_NFSBIGFH, "NFSV3 bigfh", "NFS version 3 file handle"); 79 MALLOC_DEFINE(M_NFSD, "NFS daemon", "Nfs server daemon structure"); 80 MALLOC_DEFINE(M_NFSDIROFF, "NFSV3 diroff", "NFS directory offset data"); 81 MALLOC_DEFINE(M_NFSRVDESC, "NFSV3 srvdesc", "NFS server socket descriptor"); 82 MALLOC_DEFINE(M_NFSUID, "NFS uid", "Nfs uid mapping structure"); 83 MALLOC_DEFINE(M_NQLEASE, "NQNFS Lease", "Nqnfs lease"); 84 MALLOC_DEFINE(M_NFSHASH, "NFS hash", "NFS hash tables"); 85 86 vm_zone_t nfsmount_zone; 87 88 struct nfsstats nfsstats; 89 SYSCTL_NODE(_vfs, OID_AUTO, nfs, CTLFLAG_RW, 0, "NFS filesystem"); 90 SYSCTL_STRUCT(_vfs_nfs, NFS_NFSSTATS, nfsstats, CTLFLAG_RD, 91 &nfsstats, nfsstats, ""); 92 static int nfs_ip_paranoia = 1; 93 SYSCTL_INT(_vfs_nfs, OID_AUTO, nfs_ip_paranoia, CTLFLAG_RW, 94 &nfs_ip_paranoia, 0, ""); 95 #ifdef NFS_DEBUG 96 int nfs_debug; 97 SYSCTL_INT(_vfs_nfs, OID_AUTO, debug, CTLFLAG_RW, &nfs_debug, 0, ""); 98 #endif 99 100 static int nfs_iosize __P((struct nfsmount *nmp)); 101 static void nfs_decode_args __P((struct nfsmount *nmp, 102 struct nfs_args *argp)); 103 static int mountnfs __P((struct nfs_args *,struct mount *, 104 struct sockaddr *,char *,char *,struct vnode **)); 105 static int nfs_mount __P(( struct mount *mp, char *path, caddr_t data, 106 struct nameidata *ndp, struct proc *p)); 107 static int nfs_unmount __P(( struct mount *mp, int mntflags, 108 struct proc *p)); 109 static int nfs_root __P(( struct mount *mp, struct vnode **vpp)); 110 static int nfs_statfs __P(( struct mount *mp, struct statfs *sbp, 111 struct proc *p)); 112 static int nfs_sync __P(( struct mount *mp, int waitfor, 113 struct ucred *cred, struct proc *p)); 114 115 /* 116 * nfs vfs operations. 117 */ 118 static struct vfsops nfs_vfsops = { 119 nfs_mount, 120 vfs_stdstart, 121 nfs_unmount, 122 nfs_root, 123 vfs_stdquotactl, 124 nfs_statfs, 125 nfs_sync, 126 vfs_stdvget, 127 vfs_stdfhtovp, /* shouldn't happen */ 128 vfs_stdcheckexp, 129 vfs_stdvptofh, /* shouldn't happen */ 130 nfs_init, 131 nfs_uninit, 132 vfs_stdextattrctl, 133 }; 134 VFS_SET(nfs_vfsops, nfs, VFCF_NETWORK); 135 136 /* 137 * This structure must be filled in by a primary bootstrap or bootstrap 138 * server for a diskless/dataless machine. It is initialized below just 139 * to ensure that it is allocated to initialized data (.data not .bss). 140 */ 141 struct nfs_diskless nfs_diskless = { { { 0 } } }; 142 struct nfsv3_diskless nfsv3_diskless = { { { 0 } } }; 143 int nfs_diskless_valid = 0; 144 145 SYSCTL_INT(_vfs_nfs, OID_AUTO, diskless_valid, CTLFLAG_RD, 146 &nfs_diskless_valid, 0, ""); 147 148 SYSCTL_STRING(_vfs_nfs, OID_AUTO, diskless_rootpath, CTLFLAG_RD, 149 nfsv3_diskless.root_hostnam, 0, ""); 150 151 SYSCTL_OPAQUE(_vfs_nfs, OID_AUTO, diskless_rootaddr, CTLFLAG_RD, 152 &nfsv3_diskless.root_saddr, sizeof nfsv3_diskless.root_saddr, 153 "%Ssockaddr_in", ""); 154 155 SYSCTL_STRING(_vfs_nfs, OID_AUTO, diskless_swappath, CTLFLAG_RD, 156 nfsv3_diskless.swap_hostnam, 0, ""); 157 158 SYSCTL_OPAQUE(_vfs_nfs, OID_AUTO, diskless_swapaddr, CTLFLAG_RD, 159 &nfsv3_diskless.swap_saddr, sizeof nfsv3_diskless.swap_saddr, 160 "%Ssockaddr_in",""); 161 162 163 void nfsargs_ntoh __P((struct nfs_args *)); 164 static int nfs_mountdiskless __P((char *, char *, int, 165 struct sockaddr_in *, struct nfs_args *, 166 struct proc *, struct vnode **, 167 struct mount **)); 168 static void nfs_convert_diskless __P((void)); 169 static void nfs_convert_oargs __P((struct nfs_args *args, 170 struct onfs_args *oargs)); 171 172 static int 173 nfs_iosize(nmp) 174 struct nfsmount* nmp; 175 { 176 int iosize; 177 178 /* 179 * Calculate the size used for io buffers. Use the larger 180 * of the two sizes to minimise nfs requests but make sure 181 * that it is at least one VM page to avoid wasting buffer 182 * space. 183 */ 184 iosize = max(nmp->nm_rsize, nmp->nm_wsize); 185 if (iosize < PAGE_SIZE) iosize = PAGE_SIZE; 186 return iosize; 187 } 188 189 static void 190 nfs_convert_oargs(args, oargs) 191 struct nfs_args *args; 192 struct onfs_args *oargs; 193 { 194 args->version = NFS_ARGSVERSION; 195 args->addr = oargs->addr; 196 args->addrlen = oargs->addrlen; 197 args->sotype = oargs->sotype; 198 args->proto = oargs->proto; 199 args->fh = oargs->fh; 200 args->fhsize = oargs->fhsize; 201 args->flags = oargs->flags; 202 args->wsize = oargs->wsize; 203 args->rsize = oargs->rsize; 204 args->readdirsize = oargs->readdirsize; 205 args->timeo = oargs->timeo; 206 args->retrans = oargs->retrans; 207 args->maxgrouplist = oargs->maxgrouplist; 208 args->readahead = oargs->readahead; 209 args->leaseterm = oargs->leaseterm; 210 args->deadthresh = oargs->deadthresh; 211 args->hostname = oargs->hostname; 212 } 213 214 static void 215 nfs_convert_diskless() 216 { 217 bcopy(&nfs_diskless.myif, &nfsv3_diskless.myif, 218 sizeof(struct ifaliasreq)); 219 bcopy(&nfs_diskless.mygateway, &nfsv3_diskless.mygateway, 220 sizeof(struct sockaddr_in)); 221 nfs_convert_oargs(&nfsv3_diskless.swap_args,&nfs_diskless.swap_args); 222 nfsv3_diskless.swap_fhsize = NFSX_V2FH; 223 bcopy(nfs_diskless.swap_fh,nfsv3_diskless.swap_fh,NFSX_V2FH); 224 bcopy(&nfs_diskless.swap_saddr,&nfsv3_diskless.swap_saddr, 225 sizeof(struct sockaddr_in)); 226 bcopy(nfs_diskless.swap_hostnam,nfsv3_diskless.swap_hostnam, MNAMELEN); 227 nfsv3_diskless.swap_nblks = nfs_diskless.swap_nblks; 228 bcopy(&nfs_diskless.swap_ucred, &nfsv3_diskless.swap_ucred, 229 sizeof(struct ucred)); 230 nfs_convert_oargs(&nfsv3_diskless.root_args,&nfs_diskless.root_args); 231 nfsv3_diskless.root_fhsize = NFSX_V2FH; 232 bcopy(nfs_diskless.root_fh,nfsv3_diskless.root_fh,NFSX_V2FH); 233 bcopy(&nfs_diskless.root_saddr,&nfsv3_diskless.root_saddr, 234 sizeof(struct sockaddr_in)); 235 bcopy(nfs_diskless.root_hostnam,nfsv3_diskless.root_hostnam, MNAMELEN); 236 nfsv3_diskless.root_time = nfs_diskless.root_time; 237 bcopy(nfs_diskless.my_hostnam,nfsv3_diskless.my_hostnam, 238 MAXHOSTNAMELEN); 239 nfs_diskless_valid = 3; 240 } 241 242 /* 243 * nfs statfs call 244 */ 245 int 246 nfs_statfs(mp, sbp, p) 247 struct mount *mp; 248 register struct statfs *sbp; 249 struct proc *p; 250 { 251 register struct vnode *vp; 252 register struct nfs_statfs *sfp; 253 register caddr_t cp; 254 register u_int32_t *tl; 255 register int32_t t1, t2; 256 caddr_t bpos, dpos, cp2; 257 struct nfsmount *nmp = VFSTONFS(mp); 258 int error = 0, v3 = (nmp->nm_flag & NFSMNT_NFSV3), retattr; 259 struct mbuf *mreq, *mrep, *md, *mb, *mb2; 260 struct ucred *cred; 261 struct nfsnode *np; 262 u_quad_t tquad; 263 264 #ifndef nolint 265 sfp = (struct nfs_statfs *)0; 266 #endif 267 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np); 268 if (error) 269 return (error); 270 vp = NFSTOV(np); 271 cred = crget(); 272 cred->cr_ngroups = 1; 273 if (v3 && (nmp->nm_state & NFSSTA_GOTFSINFO) == 0) 274 (void)nfs_fsinfo(nmp, vp, cred, p); 275 nfsstats.rpccnt[NFSPROC_FSSTAT]++; 276 nfsm_reqhead(vp, NFSPROC_FSSTAT, NFSX_FH(v3)); 277 nfsm_fhtom(vp, v3); 278 nfsm_request(vp, NFSPROC_FSSTAT, p, cred); 279 if (v3) 280 nfsm_postop_attr(vp, retattr); 281 if (error) { 282 if (mrep != NULL) 283 m_freem(mrep); 284 goto nfsmout; 285 } 286 nfsm_dissect(sfp, struct nfs_statfs *, NFSX_STATFS(v3)); 287 sbp->f_flags = nmp->nm_flag; 288 sbp->f_iosize = nfs_iosize(nmp); 289 if (v3) { 290 sbp->f_bsize = NFS_FABLKSIZE; 291 tquad = fxdr_hyper(&sfp->sf_tbytes); 292 sbp->f_blocks = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE)); 293 tquad = fxdr_hyper(&sfp->sf_fbytes); 294 sbp->f_bfree = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE)); 295 tquad = fxdr_hyper(&sfp->sf_abytes); 296 sbp->f_bavail = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE)); 297 sbp->f_files = (fxdr_unsigned(int32_t, 298 sfp->sf_tfiles.nfsuquad[1]) & 0x7fffffff); 299 sbp->f_ffree = (fxdr_unsigned(int32_t, 300 sfp->sf_ffiles.nfsuquad[1]) & 0x7fffffff); 301 } else { 302 sbp->f_bsize = fxdr_unsigned(int32_t, sfp->sf_bsize); 303 sbp->f_blocks = fxdr_unsigned(int32_t, sfp->sf_blocks); 304 sbp->f_bfree = fxdr_unsigned(int32_t, sfp->sf_bfree); 305 sbp->f_bavail = fxdr_unsigned(int32_t, sfp->sf_bavail); 306 sbp->f_files = 0; 307 sbp->f_ffree = 0; 308 } 309 if (sbp != &mp->mnt_stat) { 310 sbp->f_type = mp->mnt_vfc->vfc_typenum; 311 bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN); 312 bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN); 313 } 314 nfsm_reqdone; 315 vput(vp); 316 crfree(cred); 317 return (error); 318 } 319 320 /* 321 * nfs version 3 fsinfo rpc call 322 */ 323 int 324 nfs_fsinfo(nmp, vp, cred, p) 325 register struct nfsmount *nmp; 326 register struct vnode *vp; 327 struct ucred *cred; 328 struct proc *p; 329 { 330 register struct nfsv3_fsinfo *fsp; 331 register caddr_t cp; 332 register int32_t t1, t2; 333 register u_int32_t *tl, pref, max; 334 caddr_t bpos, dpos, cp2; 335 int error = 0, retattr; 336 struct mbuf *mreq, *mrep, *md, *mb, *mb2; 337 u_int64_t maxfsize; 338 339 nfsstats.rpccnt[NFSPROC_FSINFO]++; 340 nfsm_reqhead(vp, NFSPROC_FSINFO, NFSX_FH(1)); 341 nfsm_fhtom(vp, 1); 342 nfsm_request(vp, NFSPROC_FSINFO, p, cred); 343 nfsm_postop_attr(vp, retattr); 344 if (!error) { 345 nfsm_dissect(fsp, struct nfsv3_fsinfo *, NFSX_V3FSINFO); 346 pref = fxdr_unsigned(u_int32_t, fsp->fs_wtpref); 347 if (pref < nmp->nm_wsize && pref >= NFS_FABLKSIZE) 348 nmp->nm_wsize = (pref + NFS_FABLKSIZE - 1) & 349 ~(NFS_FABLKSIZE - 1); 350 max = fxdr_unsigned(u_int32_t, fsp->fs_wtmax); 351 if (max < nmp->nm_wsize && max > 0) { 352 nmp->nm_wsize = max & ~(NFS_FABLKSIZE - 1); 353 if (nmp->nm_wsize == 0) 354 nmp->nm_wsize = max; 355 } 356 pref = fxdr_unsigned(u_int32_t, fsp->fs_rtpref); 357 if (pref < nmp->nm_rsize && pref >= NFS_FABLKSIZE) 358 nmp->nm_rsize = (pref + NFS_FABLKSIZE - 1) & 359 ~(NFS_FABLKSIZE - 1); 360 max = fxdr_unsigned(u_int32_t, fsp->fs_rtmax); 361 if (max < nmp->nm_rsize && max > 0) { 362 nmp->nm_rsize = max & ~(NFS_FABLKSIZE - 1); 363 if (nmp->nm_rsize == 0) 364 nmp->nm_rsize = max; 365 } 366 pref = fxdr_unsigned(u_int32_t, fsp->fs_dtpref); 367 if (pref < nmp->nm_readdirsize && pref >= NFS_DIRBLKSIZ) 368 nmp->nm_readdirsize = (pref + NFS_DIRBLKSIZ - 1) & 369 ~(NFS_DIRBLKSIZ - 1); 370 if (max < nmp->nm_readdirsize && max > 0) { 371 nmp->nm_readdirsize = max & ~(NFS_DIRBLKSIZ - 1); 372 if (nmp->nm_readdirsize == 0) 373 nmp->nm_readdirsize = max; 374 } 375 maxfsize = fxdr_hyper(&fsp->fs_maxfilesize); 376 if (maxfsize > 0 && maxfsize < nmp->nm_maxfilesize) 377 nmp->nm_maxfilesize = maxfsize; 378 nmp->nm_state |= NFSSTA_GOTFSINFO; 379 } 380 nfsm_reqdone; 381 return (error); 382 } 383 384 /* 385 * Mount a remote root fs via. nfs. This depends on the info in the 386 * nfs_diskless structure that has been filled in properly by some primary 387 * bootstrap. 388 * It goes something like this: 389 * - do enough of "ifconfig" by calling ifioctl() so that the system 390 * can talk to the server 391 * - If nfs_diskless.mygateway is filled in, use that address as 392 * a default gateway. 393 * - build the rootfs mount point and call mountnfs() to do the rest. 394 */ 395 int 396 nfs_mountroot(mp) 397 struct mount *mp; 398 { 399 struct mount *swap_mp; 400 struct nfsv3_diskless *nd = &nfsv3_diskless; 401 struct socket *so; 402 struct vnode *vp; 403 struct proc *p = curproc; /* XXX */ 404 int error, i; 405 u_long l; 406 char buf[128]; 407 408 #if defined(BOOTP_NFSROOT) && defined(BOOTP) 409 bootpc_init(); /* use bootp to get nfs_diskless filled in */ 410 #endif 411 412 /* 413 * XXX time must be non-zero when we init the interface or else 414 * the arp code will wedge... 415 */ 416 while (time_second == 0) 417 tsleep(&time_second, PZERO+8, "arpkludge", 10); 418 419 if (nfs_diskless_valid==1) 420 nfs_convert_diskless(); 421 422 /* 423 * XXX splnet, so networks will receive... 424 */ 425 splnet(); 426 427 #ifdef notyet 428 /* Set up swap credentials. */ 429 proc0.p_ucred->cr_uid = ntohl(nd->swap_ucred.cr_uid); 430 proc0.p_ucred->cr_gid = ntohl(nd->swap_ucred.cr_gid); 431 if ((proc0.p_ucred->cr_ngroups = ntohs(nd->swap_ucred.cr_ngroups)) > 432 NGROUPS) 433 proc0.p_ucred->cr_ngroups = NGROUPS; 434 for (i = 0; i < proc0.p_ucred->cr_ngroups; i++) 435 proc0.p_ucred->cr_groups[i] = ntohl(nd->swap_ucred.cr_groups[i]); 436 #endif 437 438 /* 439 * Do enough of ifconfig(8) so that the critical net interface can 440 * talk to the server. 441 */ 442 error = socreate(nd->myif.ifra_addr.sa_family, &so, SOCK_DGRAM, 0, p); 443 if (error) 444 panic("nfs_mountroot: socreate(%04x): %d", 445 nd->myif.ifra_addr.sa_family, error); 446 447 #if 0 /* XXX Bad idea */ 448 /* 449 * We might not have been told the right interface, so we pass 450 * over the first ten interfaces of the same kind, until we get 451 * one of them configured. 452 */ 453 454 for (i = strlen(nd->myif.ifra_name) - 1; 455 nd->myif.ifra_name[i] >= '0' && 456 nd->myif.ifra_name[i] <= '9'; 457 nd->myif.ifra_name[i] ++) { 458 error = ifioctl(so, SIOCAIFADDR, (caddr_t)&nd->myif, p); 459 if(!error) 460 break; 461 } 462 #endif 463 error = ifioctl(so, SIOCAIFADDR, (caddr_t)&nd->myif, p); 464 if (error) 465 panic("nfs_mountroot: SIOCAIFADDR: %d", error); 466 soclose(so); 467 468 /* 469 * If the gateway field is filled in, set it as the default route. 470 */ 471 if (nd->mygateway.sin_len != 0) { 472 struct sockaddr_in mask, sin; 473 474 bzero((caddr_t)&mask, sizeof(mask)); 475 sin = mask; 476 sin.sin_family = AF_INET; 477 sin.sin_len = sizeof(sin); 478 error = rtrequest(RTM_ADD, (struct sockaddr *)&sin, 479 (struct sockaddr *)&nd->mygateway, 480 (struct sockaddr *)&mask, 481 RTF_UP | RTF_GATEWAY, (struct rtentry **)0); 482 if (error) 483 panic("nfs_mountroot: RTM_ADD: %d", error); 484 } 485 486 /* 487 * Create the rootfs mount point. 488 */ 489 nd->root_args.fh = nd->root_fh; 490 nd->root_args.fhsize = nd->root_fhsize; 491 l = ntohl(nd->root_saddr.sin_addr.s_addr); 492 snprintf(buf, sizeof(buf), "%ld.%ld.%ld.%ld:%s", 493 (l >> 24) & 0xff, (l >> 16) & 0xff, 494 (l >> 8) & 0xff, (l >> 0) & 0xff,nd->root_hostnam); 495 printf("NFS ROOT: %s\n",buf); 496 if ((error = nfs_mountdiskless(buf, "/", MNT_RDONLY, 497 &nd->root_saddr, &nd->root_args, p, &vp, &mp)) != 0) { 498 if (swap_mp) { 499 mp->mnt_vfc->vfc_refcount--; 500 free(swap_mp, M_MOUNT); 501 } 502 return (error); 503 } 504 505 swap_mp = NULL; 506 if (nd->swap_nblks) { 507 508 /* Convert to DEV_BSIZE instead of Kilobyte */ 509 nd->swap_nblks *= 2; 510 511 /* 512 * Create a fake mount point just for the swap vnode so that the 513 * swap file can be on a different server from the rootfs. 514 */ 515 nd->swap_args.fh = nd->swap_fh; 516 nd->swap_args.fhsize = nd->swap_fhsize; 517 l = ntohl(nd->swap_saddr.sin_addr.s_addr); 518 snprintf(buf, sizeof(buf), "%ld.%ld.%ld.%ld:%s", 519 (l >> 24) & 0xff, (l >> 16) & 0xff, 520 (l >> 8) & 0xff, (l >> 0) & 0xff,nd->swap_hostnam); 521 printf("NFS SWAP: %s\n",buf); 522 if ((error = nfs_mountdiskless(buf, "/swap", 0, 523 &nd->swap_saddr, &nd->swap_args, p, &vp, &swap_mp)) != 0) 524 return (error); 525 vfs_unbusy(swap_mp, p); 526 527 VTONFS(vp)->n_size = VTONFS(vp)->n_vattr.va_size = 528 nd->swap_nblks * DEV_BSIZE ; 529 530 /* 531 * Since the swap file is not the root dir of a file system, 532 * hack it to a regular file. 533 */ 534 vp->v_type = VREG; 535 vp->v_flag = 0; 536 VREF(vp); 537 swaponvp(p, vp, NODEV, nd->swap_nblks); 538 } 539 540 mp->mnt_flag |= MNT_ROOTFS; 541 mp->mnt_vnodecovered = NULLVP; 542 rootvp = vp; 543 vfs_unbusy(mp, p); 544 545 /* 546 * This is not really an nfs issue, but it is much easier to 547 * set hostname here and then let the "/etc/rc.xxx" files 548 * mount the right /var based upon its preset value. 549 */ 550 bcopy(nd->my_hostnam, hostname, MAXHOSTNAMELEN); 551 hostname[MAXHOSTNAMELEN - 1] = '\0'; 552 for (i = 0; i < MAXHOSTNAMELEN; i++) 553 if (hostname[i] == '\0') 554 break; 555 inittodr(ntohl(nd->root_time)); 556 return (0); 557 } 558 559 /* 560 * Internal version of mount system call for diskless setup. 561 */ 562 static int 563 nfs_mountdiskless(path, which, mountflag, sin, args, p, vpp, mpp) 564 char *path; 565 char *which; 566 int mountflag; 567 struct sockaddr_in *sin; 568 struct nfs_args *args; 569 struct proc *p; 570 struct vnode **vpp; 571 struct mount **mpp; 572 { 573 struct mount *mp; 574 struct sockaddr *nam; 575 int error; 576 int didalloc = 0; 577 578 mp = *mpp; 579 580 if (mp == NULL) { 581 if ((error = vfs_rootmountalloc("nfs", path, &mp)) != 0) { 582 printf("nfs_mountroot: NFS not configured"); 583 return (error); 584 } 585 didalloc = 1; 586 } 587 588 mp->mnt_kern_flag = 0; 589 mp->mnt_flag = mountflag; 590 nam = dup_sockaddr((struct sockaddr *)sin, 1); 591 if ((error = mountnfs(args, mp, nam, which, path, vpp)) != 0) { 592 printf("nfs_mountroot: mount %s on %s: %d", path, which, error); 593 mp->mnt_vfc->vfc_refcount--; 594 vfs_unbusy(mp, p); 595 if (didalloc) 596 free(mp, M_MOUNT); 597 FREE(nam, M_SONAME); 598 return (error); 599 } 600 (void) copystr(which, mp->mnt_stat.f_mntonname, MNAMELEN - 1, 0); 601 *mpp = mp; 602 return (0); 603 } 604 605 static void 606 nfs_decode_args(nmp, argp) 607 struct nfsmount *nmp; 608 struct nfs_args *argp; 609 { 610 int s; 611 int adjsock; 612 int maxio; 613 614 s = splnet(); 615 /* 616 * Silently clear NFSMNT_NOCONN if it's a TCP mount, it makes 617 * no sense in that context. 618 */ 619 if (argp->sotype == SOCK_STREAM) 620 nmp->nm_flag &= ~NFSMNT_NOCONN; 621 622 /* Also clear RDIRPLUS if not NFSv3, it crashes some servers */ 623 if ((argp->flags & NFSMNT_NFSV3) == 0) 624 nmp->nm_flag &= ~NFSMNT_RDIRPLUS; 625 626 /* Re-bind if rsrvd port requested and wasn't on one */ 627 adjsock = !(nmp->nm_flag & NFSMNT_RESVPORT) 628 && (argp->flags & NFSMNT_RESVPORT); 629 /* Also re-bind if we're switching to/from a connected UDP socket */ 630 adjsock |= ((nmp->nm_flag & NFSMNT_NOCONN) != 631 (argp->flags & NFSMNT_NOCONN)); 632 633 /* Update flags atomically. Don't change the lock bits. */ 634 nmp->nm_flag = argp->flags | nmp->nm_flag; 635 splx(s); 636 637 if ((argp->flags & NFSMNT_TIMEO) && argp->timeo > 0) { 638 nmp->nm_timeo = (argp->timeo * NFS_HZ + 5) / 10; 639 if (nmp->nm_timeo < NFS_MINTIMEO) 640 nmp->nm_timeo = NFS_MINTIMEO; 641 else if (nmp->nm_timeo > NFS_MAXTIMEO) 642 nmp->nm_timeo = NFS_MAXTIMEO; 643 } 644 645 if ((argp->flags & NFSMNT_RETRANS) && argp->retrans > 1) { 646 nmp->nm_retry = argp->retrans; 647 if (nmp->nm_retry > NFS_MAXREXMIT) 648 nmp->nm_retry = NFS_MAXREXMIT; 649 } 650 651 if (argp->flags & NFSMNT_NFSV3) { 652 if (argp->sotype == SOCK_DGRAM) 653 maxio = NFS_MAXDGRAMDATA; 654 else 655 maxio = NFS_MAXDATA; 656 } else 657 maxio = NFS_V2MAXDATA; 658 659 if ((argp->flags & NFSMNT_WSIZE) && argp->wsize > 0) { 660 nmp->nm_wsize = argp->wsize; 661 /* Round down to multiple of blocksize */ 662 nmp->nm_wsize &= ~(NFS_FABLKSIZE - 1); 663 if (nmp->nm_wsize <= 0) 664 nmp->nm_wsize = NFS_FABLKSIZE; 665 } 666 if (nmp->nm_wsize > maxio) 667 nmp->nm_wsize = maxio; 668 if (nmp->nm_wsize > MAXBSIZE) 669 nmp->nm_wsize = MAXBSIZE; 670 671 if ((argp->flags & NFSMNT_RSIZE) && argp->rsize > 0) { 672 nmp->nm_rsize = argp->rsize; 673 /* Round down to multiple of blocksize */ 674 nmp->nm_rsize &= ~(NFS_FABLKSIZE - 1); 675 if (nmp->nm_rsize <= 0) 676 nmp->nm_rsize = NFS_FABLKSIZE; 677 } 678 if (nmp->nm_rsize > maxio) 679 nmp->nm_rsize = maxio; 680 if (nmp->nm_rsize > MAXBSIZE) 681 nmp->nm_rsize = MAXBSIZE; 682 683 if ((argp->flags & NFSMNT_READDIRSIZE) && argp->readdirsize > 0) { 684 nmp->nm_readdirsize = argp->readdirsize; 685 } 686 if (nmp->nm_readdirsize > maxio) 687 nmp->nm_readdirsize = maxio; 688 if (nmp->nm_readdirsize > nmp->nm_rsize) 689 nmp->nm_readdirsize = nmp->nm_rsize; 690 691 if ((argp->flags & NFSMNT_ACREGMIN) && argp->acregmin >= 0) 692 nmp->nm_acregmin = argp->acregmin; 693 else 694 nmp->nm_acregmin = NFS_MINATTRTIMO; 695 if ((argp->flags & NFSMNT_ACREGMAX) && argp->acregmax >= 0) 696 nmp->nm_acregmax = argp->acregmax; 697 else 698 nmp->nm_acregmax = NFS_MAXATTRTIMO; 699 if ((argp->flags & NFSMNT_ACDIRMIN) && argp->acdirmin >= 0) 700 nmp->nm_acdirmin = argp->acdirmin; 701 else 702 nmp->nm_acdirmin = NFS_MINDIRATTRTIMO; 703 if ((argp->flags & NFSMNT_ACDIRMAX) && argp->acdirmax >= 0) 704 nmp->nm_acdirmax = argp->acdirmax; 705 else 706 nmp->nm_acdirmax = NFS_MAXDIRATTRTIMO; 707 if (nmp->nm_acdirmin > nmp->nm_acdirmax) 708 nmp->nm_acdirmin = nmp->nm_acdirmax; 709 if (nmp->nm_acregmin > nmp->nm_acregmax) 710 nmp->nm_acregmin = nmp->nm_acregmax; 711 712 if ((argp->flags & NFSMNT_MAXGRPS) && argp->maxgrouplist >= 0) { 713 if (argp->maxgrouplist <= NFS_MAXGRPS) 714 nmp->nm_numgrps = argp->maxgrouplist; 715 else 716 nmp->nm_numgrps = NFS_MAXGRPS; 717 } 718 if ((argp->flags & NFSMNT_READAHEAD) && argp->readahead >= 0) { 719 if (argp->readahead <= NFS_MAXRAHEAD) 720 nmp->nm_readahead = argp->readahead; 721 else 722 nmp->nm_readahead = NFS_MAXRAHEAD; 723 } 724 if ((argp->flags & NFSMNT_LEASETERM) && argp->leaseterm >= 2) { 725 if (argp->leaseterm <= NQ_MAXLEASE) 726 nmp->nm_leaseterm = argp->leaseterm; 727 else 728 nmp->nm_leaseterm = NQ_MAXLEASE; 729 } 730 if ((argp->flags & NFSMNT_DEADTHRESH) && argp->deadthresh >= 1) { 731 if (argp->deadthresh <= NQ_NEVERDEAD) 732 nmp->nm_deadthresh = argp->deadthresh; 733 else 734 nmp->nm_deadthresh = NQ_NEVERDEAD; 735 } 736 737 adjsock |= ((nmp->nm_sotype != argp->sotype) || 738 (nmp->nm_soproto != argp->proto)); 739 nmp->nm_sotype = argp->sotype; 740 nmp->nm_soproto = argp->proto; 741 742 if (nmp->nm_so && adjsock) { 743 nfs_safedisconnect(nmp); 744 if (nmp->nm_sotype == SOCK_DGRAM) 745 while (nfs_connect(nmp, (struct nfsreq *)0)) { 746 printf("nfs_args: retrying connect\n"); 747 (void) tsleep((caddr_t)&lbolt, 748 PSOCK, "nfscon", 0); 749 } 750 } 751 } 752 753 /* 754 * VFS Operations. 755 * 756 * mount system call 757 * It seems a bit dumb to copyinstr() the host and path here and then 758 * bcopy() them in mountnfs(), but I wanted to detect errors before 759 * doing the sockargs() call because sockargs() allocates an mbuf and 760 * an error after that means that I have to release the mbuf. 761 */ 762 /* ARGSUSED */ 763 static int 764 nfs_mount(mp, path, data, ndp, p) 765 struct mount *mp; 766 char *path; 767 caddr_t data; 768 struct nameidata *ndp; 769 struct proc *p; 770 { 771 int error; 772 struct nfs_args args; 773 struct sockaddr *nam; 774 struct vnode *vp; 775 char pth[MNAMELEN], hst[MNAMELEN]; 776 size_t len; 777 u_char nfh[NFSX_V3FHMAX]; 778 779 if (path == NULL) { 780 nfs_mountroot(mp); 781 return (0); 782 } 783 error = copyin(data, (caddr_t)&args, sizeof (struct nfs_args)); 784 if (error) 785 return (error); 786 if (args.version != NFS_ARGSVERSION) { 787 #ifdef COMPAT_PRELITE2 788 /* 789 * If the argument version is unknown, then assume the 790 * caller is a pre-lite2 4.4BSD client and convert its 791 * arguments. 792 */ 793 struct onfs_args oargs; 794 error = copyin(data, (caddr_t)&oargs, sizeof (struct onfs_args)); 795 if (error) 796 return (error); 797 nfs_convert_oargs(&args,&oargs); 798 #else /* !COMPAT_PRELITE2 */ 799 return (EPROGMISMATCH); 800 #endif /* COMPAT_PRELITE2 */ 801 } 802 if (mp->mnt_flag & MNT_UPDATE) { 803 register struct nfsmount *nmp = VFSTONFS(mp); 804 805 if (nmp == NULL) 806 return (EIO); 807 /* 808 * When doing an update, we can't change from or to 809 * v3 and/or nqnfs, or change cookie translation 810 */ 811 args.flags = (args.flags & 812 ~(NFSMNT_NFSV3|NFSMNT_NQNFS /*|NFSMNT_XLATECOOKIE*/)) | 813 (nmp->nm_flag & 814 (NFSMNT_NFSV3|NFSMNT_NQNFS /*|NFSMNT_XLATECOOKIE*/)); 815 nfs_decode_args(nmp, &args); 816 return (0); 817 } 818 819 /* 820 * Make the nfs_ip_paranoia sysctl serve as the default connection 821 * or no-connection mode for those protocols that support 822 * no-connection mode (the flag will be cleared later for protocols 823 * that do not support no-connection mode). This will allow a client 824 * to receive replies from a different IP then the request was 825 * sent to. Note: default value for nfs_ip_paranoia is 1 (paranoid), 826 * not 0. 827 */ 828 if (nfs_ip_paranoia == 0) 829 args.flags |= NFSMNT_NOCONN; 830 if (args.fhsize < 0 || args.fhsize > NFSX_V3FHMAX) 831 return (EINVAL); 832 error = copyin((caddr_t)args.fh, (caddr_t)nfh, args.fhsize); 833 if (error) 834 return (error); 835 error = copyinstr(path, pth, MNAMELEN-1, &len); 836 if (error) 837 return (error); 838 bzero(&pth[len], MNAMELEN - len); 839 error = copyinstr(args.hostname, hst, MNAMELEN-1, &len); 840 if (error) 841 return (error); 842 bzero(&hst[len], MNAMELEN - len); 843 /* sockargs() call must be after above copyin() calls */ 844 error = getsockaddr(&nam, (caddr_t)args.addr, args.addrlen); 845 if (error) 846 return (error); 847 args.fh = nfh; 848 error = mountnfs(&args, mp, nam, pth, hst, &vp); 849 return (error); 850 } 851 852 /* 853 * Common code for mount and mountroot 854 */ 855 static int 856 mountnfs(argp, mp, nam, pth, hst, vpp) 857 register struct nfs_args *argp; 858 register struct mount *mp; 859 struct sockaddr *nam; 860 char *pth, *hst; 861 struct vnode **vpp; 862 { 863 register struct nfsmount *nmp; 864 struct nfsnode *np; 865 int error; 866 struct vattr attrs; 867 868 if (mp->mnt_flag & MNT_UPDATE) { 869 nmp = VFSTONFS(mp); 870 /* update paths, file handles, etc, here XXX */ 871 FREE(nam, M_SONAME); 872 return (0); 873 } else { 874 nmp = zalloc(nfsmount_zone); 875 bzero((caddr_t)nmp, sizeof (struct nfsmount)); 876 TAILQ_INIT(&nmp->nm_uidlruhead); 877 TAILQ_INIT(&nmp->nm_bufq); 878 mp->mnt_data = (qaddr_t)nmp; 879 } 880 vfs_getnewfsid(mp); 881 nmp->nm_mountp = mp; 882 if (argp->flags & NFSMNT_NQNFS) 883 /* 884 * We have to set mnt_maxsymlink to a non-zero value so 885 * that COMPAT_43 routines will know that we are setting 886 * the d_type field in directories (and can zero it for 887 * unsuspecting binaries). 888 */ 889 mp->mnt_maxsymlinklen = 1; 890 891 /* 892 * V2 can only handle 32 bit filesizes. A 4GB-1 limit may be too 893 * high, depending on whether we end up with negative offsets in 894 * the client or server somewhere. 2GB-1 may be safer. 895 * 896 * For V3, nfs_fsinfo will adjust this as necessary. Assume maximum 897 * that we can handle until we find out otherwise. 898 * XXX Our "safe" limit on the client is what we can store in our 899 * buffer cache using signed(!) block numbers. 900 */ 901 if ((argp->flags & NFSMNT_NFSV3) == 0) 902 nmp->nm_maxfilesize = 0xffffffffLL; 903 else 904 nmp->nm_maxfilesize = (u_int64_t)0x80000000 * DEV_BSIZE - 1; 905 906 nmp->nm_timeo = NFS_TIMEO; 907 nmp->nm_retry = NFS_RETRANS; 908 nmp->nm_wsize = NFS_WSIZE; 909 nmp->nm_rsize = NFS_RSIZE; 910 nmp->nm_readdirsize = NFS_READDIRSIZE; 911 nmp->nm_numgrps = NFS_MAXGRPS; 912 nmp->nm_readahead = NFS_DEFRAHEAD; 913 nmp->nm_leaseterm = NQ_DEFLEASE; 914 nmp->nm_deadthresh = NQ_DEADTHRESH; 915 CIRCLEQ_INIT(&nmp->nm_timerhead); 916 nmp->nm_inprog = NULLVP; 917 nmp->nm_fhsize = argp->fhsize; 918 bcopy((caddr_t)argp->fh, (caddr_t)nmp->nm_fh, argp->fhsize); 919 bcopy(hst, mp->mnt_stat.f_mntfromname, MNAMELEN); 920 bcopy(pth, mp->mnt_stat.f_mntonname, MNAMELEN); 921 nmp->nm_nam = nam; 922 /* Set up the sockets and per-host congestion */ 923 nmp->nm_sotype = argp->sotype; 924 nmp->nm_soproto = argp->proto; 925 926 nfs_decode_args(nmp, argp); 927 928 /* 929 * For Connection based sockets (TCP,...) defer the connect until 930 * the first request, in case the server is not responding. 931 */ 932 if (nmp->nm_sotype == SOCK_DGRAM && 933 (error = nfs_connect(nmp, (struct nfsreq *)0))) 934 goto bad; 935 936 /* 937 * This is silly, but it has to be set so that vinifod() works. 938 * We do not want to do an nfs_statfs() here since we can get 939 * stuck on a dead server and we are holding a lock on the mount 940 * point. 941 */ 942 mp->mnt_stat.f_iosize = nfs_iosize(nmp); 943 /* 944 * A reference count is needed on the nfsnode representing the 945 * remote root. If this object is not persistent, then backward 946 * traversals of the mount point (i.e. "..") will not work if 947 * the nfsnode gets flushed out of the cache. Ufs does not have 948 * this problem, because one can identify root inodes by their 949 * number == ROOTINO (2). 950 */ 951 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np); 952 if (error) 953 goto bad; 954 *vpp = NFSTOV(np); 955 956 /* 957 * Get file attributes for the mountpoint. This has the side 958 * effect of filling in (*vpp)->v_type with the correct value. 959 */ 960 VOP_GETATTR(*vpp, &attrs, curproc->p_ucred, curproc); 961 962 /* 963 * Lose the lock but keep the ref. 964 */ 965 VOP_UNLOCK(*vpp, 0, curproc); 966 967 return (0); 968 bad: 969 nfs_disconnect(nmp); 970 zfree(nfsmount_zone, nmp); 971 FREE(nam, M_SONAME); 972 return (error); 973 } 974 975 /* 976 * unmount system call 977 */ 978 static int 979 nfs_unmount(mp, mntflags, p) 980 struct mount *mp; 981 int mntflags; 982 struct proc *p; 983 { 984 register struct nfsmount *nmp; 985 int error, flags = 0; 986 987 if (mntflags & MNT_FORCE) 988 flags |= FORCECLOSE; 989 nmp = VFSTONFS(mp); 990 /* 991 * Goes something like this.. 992 * - Call vflush() to clear out vnodes for this file system 993 * - Close the socket 994 * - Free up the data structures 995 */ 996 /* In the forced case, cancel any outstanding requests. */ 997 if (flags & FORCECLOSE) { 998 error = nfs_nmcancelreqs(nmp); 999 if (error) 1000 return (error); 1001 } 1002 /* 1003 * Must handshake with nqnfs_clientd() if it is active. 1004 */ 1005 nmp->nm_state |= NFSSTA_DISMINPROG; 1006 while (nmp->nm_inprog != NULLVP) 1007 (void) tsleep((caddr_t)&lbolt, PSOCK, "nfsdism", 0); 1008 1009 /* We hold 1 extra ref on the root vnode; see comment in mountnfs(). */ 1010 error = vflush(mp, 1, flags); 1011 if (error) { 1012 nmp->nm_state &= ~NFSSTA_DISMINPROG; 1013 return (error); 1014 } 1015 1016 /* 1017 * We are now committed to the unmount. 1018 * For NQNFS, let the server daemon free the nfsmount structure. 1019 */ 1020 if (nmp->nm_flag & (NFSMNT_NQNFS | NFSMNT_KERB)) 1021 nmp->nm_state |= NFSSTA_DISMNT; 1022 1023 nfs_disconnect(nmp); 1024 FREE(nmp->nm_nam, M_SONAME); 1025 1026 if ((nmp->nm_flag & (NFSMNT_NQNFS | NFSMNT_KERB)) == 0) 1027 zfree(nfsmount_zone, nmp); 1028 return (0); 1029 } 1030 1031 /* 1032 * Return root of a filesystem 1033 */ 1034 static int 1035 nfs_root(mp, vpp) 1036 struct mount *mp; 1037 struct vnode **vpp; 1038 { 1039 register struct vnode *vp; 1040 struct nfsmount *nmp; 1041 struct nfsnode *np; 1042 int error; 1043 1044 nmp = VFSTONFS(mp); 1045 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np); 1046 if (error) 1047 return (error); 1048 vp = NFSTOV(np); 1049 if (vp->v_type == VNON) 1050 vp->v_type = VDIR; 1051 vp->v_flag = VROOT; 1052 *vpp = vp; 1053 return (0); 1054 } 1055 1056 extern int syncprt; 1057 1058 /* 1059 * Flush out the buffer cache 1060 */ 1061 /* ARGSUSED */ 1062 static int 1063 nfs_sync(mp, waitfor, cred, p) 1064 struct mount *mp; 1065 int waitfor; 1066 struct ucred *cred; 1067 struct proc *p; 1068 { 1069 register struct vnode *vp; 1070 int error, allerror = 0; 1071 1072 /* 1073 * Force stale buffer cache information to be flushed. 1074 */ 1075 loop: 1076 for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist); 1077 vp != NULL; 1078 vp = TAILQ_NEXT(vp, v_nmntvnodes)) { 1079 /* 1080 * If the vnode that we are about to sync is no longer 1081 * associated with this mount point, start over. 1082 */ 1083 if (vp->v_mount != mp) 1084 goto loop; 1085 if (VOP_ISLOCKED(vp, NULL) || TAILQ_EMPTY(&vp->v_dirtyblkhd) || 1086 waitfor == MNT_LAZY) 1087 continue; 1088 if (vget(vp, LK_EXCLUSIVE, p)) 1089 goto loop; 1090 error = VOP_FSYNC(vp, cred, waitfor, p); 1091 if (error) 1092 allerror = error; 1093 vput(vp); 1094 } 1095 return (allerror); 1096 } 1097 1098