darwin-xnu/bsd/nfs/nfs_vfsops.c
<<
>>
Prefs
   1/*
   2 * Copyright (c) 2000-2005 Apple Computer, Inc. All rights reserved.
   3 *
   4 * @APPLE_LICENSE_HEADER_START@
   5 * 
   6 * The contents of this file constitute Original Code as defined in and
   7 * are subject to the Apple Public Source License Version 1.1 (the
   8 * "License").  You may not use this file except in compliance with the
   9 * License.  Please obtain a copy of the License at
  10 * http://www.apple.com/publicsource and read it before using this file.
  11 * 
  12 * This Original Code and all software distributed under the License are
  13 * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  14 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  15 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  16 * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT.  Please see the
  17 * License for the specific language governing rights and limitations
  18 * under the License.
  19 * 
  20 * @APPLE_LICENSE_HEADER_END@
  21 */
  22/* Copyright (c) 1995 NeXT Computer, Inc. All Rights Reserved */
  23/*
  24 * Copyright (c) 1989, 1993, 1995
  25 *      The Regents of the University of California.  All rights reserved.
  26 *
  27 * This code is derived from software contributed to Berkeley by
  28 * Rick Macklem at The University of Guelph.
  29 *
  30 * Redistribution and use in source and binary forms, with or without
  31 * modification, are permitted provided that the following conditions
  32 * are met:
  33 * 1. Redistributions of source code must retain the above copyright
  34 *    notice, this list of conditions and the following disclaimer.
  35 * 2. Redistributions in binary form must reproduce the above copyright
  36 *    notice, this list of conditions and the following disclaimer in the
  37 *    documentation and/or other materials provided with the distribution.
  38 * 3. All advertising materials mentioning features or use of this software
  39 *    must display the following acknowledgement:
  40 *      This product includes software developed by the University of
  41 *      California, Berkeley and its contributors.
  42 * 4. Neither the name of the University nor the names of its contributors
  43 *    may be used to endorse or promote products derived from this software
  44 *    without specific prior written permission.
  45 *
  46 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  47 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  48 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  49 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  50 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  51 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  52 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  53 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  54 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  55 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  56 * SUCH DAMAGE.
  57 *
  58 *      @(#)nfs_vfsops.c        8.12 (Berkeley) 5/20/95
  59 * FreeBSD-Id: nfs_vfsops.c,v 1.52 1997/11/12 05:42:21 julian Exp $
  60 */
  61
  62#include <sys/param.h>
  63#include <sys/systm.h>
  64#include <sys/conf.h>
  65#include <sys/ioctl.h>
  66#include <sys/signal.h>
  67#include <sys/proc_internal.h> /* for fs rooting to update rootdir in fdp */
  68#include <sys/kauth.h>
  69#include <sys/vnode_internal.h>
  70#include <sys/malloc.h>
  71#include <sys/kernel.h>
  72#include <sys/sysctl.h>
  73#include <sys/mount_internal.h>
  74#include <sys/kpi_mbuf.h>
  75#include <sys/socket.h>
  76#include <sys/socketvar.h>
  77#include <sys/fcntl.h>
  78#include <libkern/OSAtomic.h>
  79
  80#include <sys/vm.h>
  81#include <sys/vmparam.h>
  82
  83#if !defined(NO_MOUNT_PRIVATE)
  84#include <sys/filedesc.h>
  85#endif /* NO_MOUNT_PRIVATE */
  86
  87#include <net/if.h>
  88#include <net/route.h>
  89#include <netinet/in.h>
  90
  91#include <nfs/rpcv2.h>
  92#include <nfs/nfsproto.h>
  93#include <nfs/nfs.h>
  94#include <nfs/nfsnode.h>
  95#include <nfs/nfsmount.h>
  96#include <nfs/xdr_subs.h>
  97#include <nfs/nfsm_subs.h>
  98#include <nfs/nfsdiskless.h>
  99#include <nfs/nfs_lock.h>
 100
 101extern int      nfs_mountroot(void);
 102
 103extern int      nfs_ticks;
 104extern int      nfs_mount_type;
 105extern int      nfs_resv_mounts;
 106
 107struct nfsstats nfsstats;
 108static int nfs_sysctl(int *, u_int, user_addr_t, size_t *, user_addr_t, size_t, vfs_context_t);
 109/* XXX CSM 11/25/97 Upgrade sysctl.h someday */
 110#ifdef notyet
 111SYSCTL_NODE(_vfs, MOUNT_NFS, nfs, CTLFLAG_RW, 0, "NFS filesystem");
 112SYSCTL_STRUCT(_vfs_nfs, NFS_NFSSTATS, nfsstats, CTLFLAG_RD,
 113        &nfsstats, nfsstats, "");
 114#endif
 115
 116SYSCTL_DECL(_vfs_generic_nfs);
 117SYSCTL_NODE(_vfs_generic_nfs, OID_AUTO, client, CTLFLAG_RW, 0,
 118    "nfs client hinge");
 119/* how long NFS will wait before signalling vfs that it's down. */
 120static int nfs_tprintf_initial_delay = NFS_TPRINTF_INITIAL_DELAY;
 121SYSCTL_INT(_vfs_generic_nfs_client, NFS_TPRINTF_INITIAL_DELAY,
 122    initialdowndelay, CTLFLAG_RW, &nfs_tprintf_initial_delay, 0, "");
 123/* how long between console messages "nfs server foo not responding" */
 124static int nfs_tprintf_delay = NFS_TPRINTF_DELAY;
 125SYSCTL_INT(_vfs_generic_nfs_client, NFS_TPRINTF_DELAY,
 126    nextdowndelay, CTLFLAG_RW, &nfs_tprintf_delay, 0, "");
 127
 128static int      nfs_iosize(struct nfsmount *nmp);
 129static int      mountnfs(struct user_nfs_args *,mount_t,mbuf_t,proc_t,vnode_t *);
 130static int      nfs_mount(mount_t mp, vnode_t vp, user_addr_t data, vfs_context_t context);
 131static int      nfs_start(mount_t mp, int flags, vfs_context_t context);
 132static int      nfs_unmount(mount_t mp, int mntflags, vfs_context_t context);
 133static int      nfs_root(mount_t mp, vnode_t *vpp, vfs_context_t context);
 134static int      nfs_statfs(mount_t mp, struct vfsstatfs *sbp, vfs_context_t context);
 135static int      nfs_vfs_getattr(mount_t mp, struct vfs_attr *fsap, vfs_context_t context);
 136static int      nfs_sync( mount_t mp, int waitfor, vfs_context_t context);
 137static int      nfs_vptofh(vnode_t vp, int *fhlenp, unsigned char *fhp, vfs_context_t context);
 138static int      nfs_fhtovp(mount_t mp, int fhlen, unsigned char *fhp, vnode_t *vpp, vfs_context_t context);
 139static int      nfs_vget(mount_t , ino64_t, vnode_t *, vfs_context_t context);
 140
 141
 142/*
 143 * nfs vfs operations.
 144 */
 145struct vfsops nfs_vfsops = {
 146        nfs_mount,
 147        nfs_start,
 148        nfs_unmount,
 149        nfs_root,
 150        NULL,           /* quotactl */
 151        nfs_vfs_getattr,
 152        nfs_sync,
 153        nfs_vget,
 154        nfs_fhtovp,
 155        nfs_vptofh,
 156        nfs_init,
 157        nfs_sysctl,
 158        NULL            /* setattr */
 159};
 160
 161
 162static int
 163nfs_mount_diskless(struct nfs_dlmount *, const char *, int, vnode_t *, mount_t *);
 164#if !defined(NO_MOUNT_PRIVATE)
 165static int
 166nfs_mount_diskless_private(struct nfs_dlmount *, const char *, int, vnode_t *, mount_t *);
 167#endif /* NO_MOUNT_PRIVATE */
 168
 169static int nfs_iosize(nmp)
 170        struct nfsmount* nmp;
 171{
 172        int iosize;
 173
 174        /*
 175         * Calculate the size used for io buffers.  Use the larger
 176         * of the two sizes to minimise nfs requests but make sure
 177         * that it is at least one VM page to avoid wasting buffer
 178         * space and to allow easy mmapping of I/O buffers.
 179         * The read/write rpc calls handle the splitting up of
 180         * buffers into multiple requests if the buffer size is
 181         * larger than the I/O size.
 182         */
 183        iosize = max(nmp->nm_rsize, nmp->nm_wsize);
 184        if (iosize < PAGE_SIZE)
 185                iosize = PAGE_SIZE;
 186        return (trunc_page_32(iosize));
 187}
 188
 189/*
 190 * nfs statfs call
 191 */
 192int
 193nfs_statfs(mount_t mp, struct vfsstatfs *sbp, vfs_context_t context)
 194{
 195        proc_t p = vfs_context_proc(context);
 196        vnode_t vp;
 197        struct nfs_statfs *sfp;
 198        caddr_t cp;
 199        u_long *tl;
 200        long t1, t2;
 201        caddr_t bpos, dpos, cp2;
 202        struct nfsmount *nmp = VFSTONFS(mp);
 203        int error = 0, v3 = (nmp->nm_flag & NFSMNT_NFSV3), retattr;
 204        mbuf_t mreq, mrep, md, mb, mb2;
 205        u_int64_t xid;
 206        kauth_cred_t cred;
 207        struct ucred temp_cred;
 208
 209#ifndef nolint
 210        sfp = (struct nfs_statfs *)0;
 211#endif
 212        vp = nmp->nm_dvp;
 213        if ((error = vnode_get(vp)))
 214                return(error);
 215
 216        bzero(&temp_cred, sizeof(temp_cred));
 217        temp_cred.cr_ngroups = 1;
 218        cred = kauth_cred_create(&temp_cred);
 219
 220        if (v3 && (nmp->nm_state & NFSSTA_GOTFSINFO) == 0)
 221                nfs_fsinfo(nmp, vp, cred, p);
 222        nfsm_reqhead(NFSX_FH(v3));
 223        if (error) {
 224                kauth_cred_rele(cred);
 225                vnode_put(vp);
 226                return (error);
 227        }
 228        OSAddAtomic(1, (SInt32*)&nfsstats.rpccnt[NFSPROC_FSSTAT]);
 229        nfsm_fhtom(vp, v3);
 230        nfsm_request(vp, NFSPROC_FSSTAT, p, cred, &xid);
 231        if (v3 && mrep)
 232                nfsm_postop_attr_update(vp, v3, retattr, &xid);
 233        nfsm_dissect(sfp, struct nfs_statfs *, NFSX_STATFS(v3));
 234
 235        sbp->f_flags = nmp->nm_flag;
 236        sbp->f_iosize = nfs_iosize(nmp);
 237        if (v3) {
 238                /*
 239                 * Adjust block size to get total block count to fit in a long.
 240                 * If we can't increase block size enough, clamp to max long.
 241                 */
 242                u_quad_t tquad, tquad2, bsize;
 243                bsize = NFS_FABLKSIZE;
 244
 245                fxdr_hyper(&sfp->sf_tbytes, &tquad);
 246                tquad /= bsize;
 247                while ((tquad & ~0x7fffffff) && (bsize < 0x40000000)) {
 248                        bsize <<= 1;
 249                        tquad >>= 1;
 250                }
 251                sbp->f_blocks = (tquad & ~0x7fffffff) ? 0x7fffffff : (long)tquad;
 252
 253                fxdr_hyper(&sfp->sf_fbytes, &tquad);
 254                tquad /= bsize;
 255                sbp->f_bfree = (tquad & ~0x7fffffff) ? 0x7fffffff : (long)tquad;
 256
 257                fxdr_hyper(&sfp->sf_abytes, &tquad);
 258                tquad /= bsize;
 259                sbp->f_bavail = (tquad & ~0x7fffffff) ? 0x7fffffff : (long)tquad;
 260
 261                sbp->f_bsize = (long)bsize;
 262
 263                /* adjust file slots too... */
 264                fxdr_hyper(&sfp->sf_tfiles, &tquad);
 265                fxdr_hyper(&sfp->sf_ffiles, &tquad2);
 266                while (tquad & ~0x7fffffff) {
 267                        tquad >>= 1;
 268                        tquad2 >>= 1;
 269                }
 270                sbp->f_files = tquad;
 271                sbp->f_ffree = tquad2;
 272        } else {
 273                sbp->f_bsize = fxdr_unsigned(long, sfp->sf_bsize);
 274                sbp->f_blocks = fxdr_unsigned(long, sfp->sf_blocks);
 275                sbp->f_bfree = fxdr_unsigned(long, sfp->sf_bfree);
 276                sbp->f_bavail = fxdr_unsigned(long, sfp->sf_bavail);
 277                sbp->f_files = 0;
 278                sbp->f_ffree = 0;
 279        }
 280        nfsm_reqdone;
 281        kauth_cred_rele(cred);
 282        vnode_put(vp);
 283        return (error);
 284}
 285
 286/*
 287 * The nfs_statfs code is complicated, and used by mountnfs(), so leave it as-is
 288 * and handle VFS_GETATTR by calling nfs_statfs and copying fields.
 289 */
 290static int
 291nfs_vfs_getattr(mount_t mp, struct vfs_attr *fsap, vfs_context_t context)
 292{
 293        int error = 0;
 294        
 295        if (VFSATTR_IS_ACTIVE(fsap, f_bsize)  ||
 296            VFSATTR_IS_ACTIVE(fsap, f_iosize) ||
 297            VFSATTR_IS_ACTIVE(fsap, f_blocks) ||
 298            VFSATTR_IS_ACTIVE(fsap, f_bfree)  ||
 299            VFSATTR_IS_ACTIVE(fsap, f_bavail) ||
 300            VFSATTR_IS_ACTIVE(fsap, f_bused)  ||
 301            VFSATTR_IS_ACTIVE(fsap, f_files)  ||
 302            VFSATTR_IS_ACTIVE(fsap, f_ffree)) {
 303                struct vfsstatfs sb;
 304
 305                error = nfs_statfs(mp, &sb, context);
 306                if (!error) {
 307                        VFSATTR_RETURN(fsap, f_bsize, sb.f_bsize);
 308                        VFSATTR_RETURN(fsap, f_iosize, sb.f_iosize);
 309                        VFSATTR_RETURN(fsap, f_blocks, sb.f_blocks);
 310                        VFSATTR_RETURN(fsap, f_bfree, sb.f_bfree);
 311                        VFSATTR_RETURN(fsap, f_bavail, sb.f_bavail);
 312                        VFSATTR_RETURN(fsap, f_bused, sb.f_blocks - sb.f_bfree);
 313                        VFSATTR_RETURN(fsap, f_files, sb.f_files);
 314                        VFSATTR_RETURN(fsap, f_ffree, sb.f_ffree);
 315                }
 316        }
 317
 318        if (VFSATTR_IS_ACTIVE(fsap, f_capabilities)) {
 319                struct nfsmount *nmp;
 320                struct nfsv3_pathconf pc;
 321                u_int32_t caps, valid;
 322                vnode_t vp;
 323                int v3;
 324
 325                if (!(nmp = VFSTONFS(mp)))
 326                        return (ENXIO);
 327                vp = nmp->nm_dvp;
 328                v3 = (nmp->nm_flag & NFSMNT_NFSV3);
 329
 330                /*
 331                 * The capabilities[] array defines what this volume supports.
 332                 *
 333                 * The valid[] array defines which bits this code understands
 334                 * the meaning of (whether the volume has that capability or not).
 335                 * Any zero bits here means "I don't know what you're asking about"
 336                 * and the caller cannot tell whether that capability is
 337                 * present or not.
 338                 */
 339                caps = valid = 0;
 340                if (v3) {
 341                        /* try to get fsinfo if we haven't already */
 342                        if (!(nmp->nm_state & NFSSTA_GOTFSINFO)) {
 343                                nfs_fsinfo(nmp, vp, vfs_context_ucred(context),
 344                                        vfs_context_proc(context));
 345                                if (!(nmp = VFSTONFS(vnode_mount(vp))))
 346                                        return (ENXIO);
 347                        }
 348                        if (nmp->nm_state & NFSSTA_GOTFSINFO) {
 349                                /* fsinfo indicates (non)support of links and symlinks */
 350                                valid |= VOL_CAP_FMT_SYMBOLICLINKS |
 351                                         VOL_CAP_FMT_HARDLINKS;
 352                                if (nmp->nm_fsinfo.fsproperties & NFSV3FSINFO_SYMLINK)
 353                                        caps |= VOL_CAP_FMT_SYMBOLICLINKS;
 354                                if (nmp->nm_fsinfo.fsproperties & NFSV3FSINFO_LINK)
 355                                        caps |= VOL_CAP_FMT_HARDLINKS;
 356                                /* if fsinfo indicates all pathconf info is the same, */
 357                                /* we can use it to report case attributes */
 358                                if ((nmp->nm_fsinfo.fsproperties & NFSV3FSINFO_HOMOGENEOUS) &&
 359                                    !(nmp->nm_state & NFSSTA_GOTPATHCONF)) {
 360                                        /* no cached pathconf info, try to get now */
 361                                        error = nfs_pathconfrpc(vp, &pc,
 362                                                        vfs_context_ucred(context),
 363                                                        vfs_context_proc(context));
 364                                        if (!(nmp = VFSTONFS(vnode_mount(vp))))
 365                                                return (ENXIO);
 366                                        if (!error) {
 367                                                /* all files have the same pathconf info, */
 368                                                /* so cache a copy of the results */
 369                                                nfs_pathconf_cache(nmp, &pc);
 370                                        }
 371                                }
 372                                if (nmp->nm_state & NFSSTA_GOTPATHCONF) {
 373                                        valid |= VOL_CAP_FMT_CASE_SENSITIVE |
 374                                                 VOL_CAP_FMT_CASE_PRESERVING;
 375                                        if (!(nmp->nm_fsinfo.pcflags &
 376                                                NFSPCINFO_CASE_INSENSITIVE))
 377                                                caps |= VOL_CAP_FMT_CASE_SENSITIVE;
 378                                        if (nmp->nm_fsinfo.pcflags &
 379                                                NFSPCINFO_CASE_PRESERVING)
 380                                                caps |= VOL_CAP_FMT_CASE_PRESERVING;
 381                                }
 382                                /* Is server's max file size at least 2TB? */
 383                                if (nmp->nm_fsinfo.maxfilesize >= 0x20000000000ULL)
 384                                        caps |= VOL_CAP_FMT_2TB_FILESIZE;
 385                        } else {
 386                                /*
 387                                 * NFSv3 supports 64 bits of file size.
 388                                 * Without FSINFO from the server, we'll
 389                                 * just assume maxfilesize >= 2TB
 390                                 */
 391                                caps |= VOL_CAP_FMT_2TB_FILESIZE;
 392                        }
 393                }
 394                fsap->f_capabilities.capabilities[VOL_CAPABILITIES_FORMAT] =
 395                        // VOL_CAP_FMT_PERSISTENTOBJECTIDS |
 396                        // VOL_CAP_FMT_SYMBOLICLINKS |
 397                        // VOL_CAP_FMT_HARDLINKS |
 398                        // VOL_CAP_FMT_JOURNAL |
 399                        // VOL_CAP_FMT_JOURNAL_ACTIVE |
 400                        // VOL_CAP_FMT_NO_ROOT_TIMES |
 401                        // VOL_CAP_FMT_SPARSE_FILES |
 402                        // VOL_CAP_FMT_ZERO_RUNS |
 403                        // VOL_CAP_FMT_CASE_SENSITIVE |
 404                        // VOL_CAP_FMT_CASE_PRESERVING |
 405                        // VOL_CAP_FMT_FAST_STATFS |
 406                        // VOL_CAP_FMT_2TB_FILESIZE |
 407                        caps;
 408                fsap->f_capabilities.valid[VOL_CAPABILITIES_FORMAT] =
 409                        VOL_CAP_FMT_PERSISTENTOBJECTIDS |
 410                        // VOL_CAP_FMT_SYMBOLICLINKS |
 411                        // VOL_CAP_FMT_HARDLINKS |
 412                        // VOL_CAP_FMT_JOURNAL |
 413                        // VOL_CAP_FMT_JOURNAL_ACTIVE |
 414                        // VOL_CAP_FMT_NO_ROOT_TIMES |
 415                        // VOL_CAP_FMT_SPARSE_FILES |
 416                        // VOL_CAP_FMT_ZERO_RUNS |
 417                        // VOL_CAP_FMT_CASE_SENSITIVE |
 418                        // VOL_CAP_FMT_CASE_PRESERVING |
 419                        VOL_CAP_FMT_FAST_STATFS |
 420                        VOL_CAP_FMT_2TB_FILESIZE |
 421                        valid;
 422
 423                /*
 424                 * We don't support most of the interfaces.
 425                 *
 426                 * We MAY support locking, but we don't have any easy way of probing.
 427                 * We can tell if there's no lockd running or if locks have been
 428                 * disabled for a mount, so we can definitely answer NO in that case.
 429                 * Any attempt to send a request to lockd to test for locking support
 430                 * may cause the lazily-launched locking daemons to be started
 431                 * unnecessarily.  So we avoid that.  However, we do record if we ever
 432                 * successfully perform a lock operation on a mount point, so if it
 433                 * looks like lock ops have worked, we do report that we support them.
 434                 */
 435                caps = valid = 0;
 436                if ((!nfslockdvnode && !nfslockdwaiting) ||
 437                    (nmp->nm_flag & NFSMNT_NOLOCKS)) {
 438                        /* locks disabled on this mount, so they definitely won't work */
 439                        valid = VOL_CAP_INT_ADVLOCK | VOL_CAP_INT_FLOCK;
 440                } else if (nmp->nm_state & NFSSTA_LOCKSWORK) {
 441                        caps = VOL_CAP_INT_ADVLOCK | VOL_CAP_INT_FLOCK;
 442                        valid = VOL_CAP_INT_ADVLOCK | VOL_CAP_INT_FLOCK;
 443                }
 444                fsap->f_capabilities.capabilities[VOL_CAPABILITIES_INTERFACES] =
 445                        // VOL_CAP_INT_SEARCHFS |
 446                        // VOL_CAP_INT_ATTRLIST |
 447                        // VOL_CAP_INT_NFSEXPORT |
 448                        // VOL_CAP_INT_READDIRATTR |
 449                        // VOL_CAP_INT_EXCHANGEDATA |
 450                        // VOL_CAP_INT_COPYFILE |
 451                        // VOL_CAP_INT_ALLOCATE |
 452                        // VOL_CAP_INT_VOL_RENAME |
 453                        // VOL_CAP_INT_ADVLOCK |
 454                        // VOL_CAP_INT_FLOCK |
 455                        // VOL_CAP_INT_EXTENDED_SECURITY |
 456                        // VOL_CAP_INT_USERACCESS |
 457                        caps;
 458                fsap->f_capabilities.valid[VOL_CAPABILITIES_INTERFACES] =
 459                        VOL_CAP_INT_SEARCHFS |
 460                        VOL_CAP_INT_ATTRLIST |
 461                        VOL_CAP_INT_NFSEXPORT |
 462                        VOL_CAP_INT_READDIRATTR |
 463                        VOL_CAP_INT_EXCHANGEDATA |
 464                        VOL_CAP_INT_COPYFILE |
 465                        VOL_CAP_INT_ALLOCATE |
 466                        VOL_CAP_INT_VOL_RENAME |
 467                        // VOL_CAP_INT_ADVLOCK |
 468                        // VOL_CAP_INT_FLOCK |
 469                        // VOL_CAP_INT_EXTENDED_SECURITY |
 470                        // VOL_CAP_INT_USERACCESS |
 471                        valid;
 472
 473                fsap->f_capabilities.capabilities[VOL_CAPABILITIES_RESERVED1] = 0;
 474                fsap->f_capabilities.valid[VOL_CAPABILITIES_RESERVED1] = 0;
 475
 476                fsap->f_capabilities.capabilities[VOL_CAPABILITIES_RESERVED2] = 0;
 477                fsap->f_capabilities.valid[VOL_CAPABILITIES_RESERVED2] = 0;
 478
 479                VFSATTR_SET_SUPPORTED(fsap, f_capabilities);
 480        }
 481
 482        if (VFSATTR_IS_ACTIVE(fsap, f_attributes)) {
 483                fsap->f_attributes.validattr.commonattr = 0;
 484                fsap->f_attributes.validattr.volattr =
 485                        ATTR_VOL_CAPABILITIES | ATTR_VOL_ATTRIBUTES;
 486                fsap->f_attributes.validattr.dirattr = 0;
 487                fsap->f_attributes.validattr.fileattr = 0;
 488                fsap->f_attributes.validattr.forkattr = 0;
 489
 490                fsap->f_attributes.nativeattr.commonattr = 0;
 491                fsap->f_attributes.nativeattr.volattr =
 492                        ATTR_VOL_CAPABILITIES | ATTR_VOL_ATTRIBUTES;
 493                fsap->f_attributes.nativeattr.dirattr = 0;
 494                fsap->f_attributes.nativeattr.fileattr = 0;
 495                fsap->f_attributes.nativeattr.forkattr = 0;
 496
 497                VFSATTR_SET_SUPPORTED(fsap, f_attributes);
 498        }
 499
 500        return (error);
 501}
 502
 503/*
 504 * nfs version 3 fsinfo rpc call
 505 */
 506int
 507nfs_fsinfo(nmp, vp, cred, p)
 508        struct nfsmount *nmp;
 509        vnode_t vp;
 510        kauth_cred_t cred;
 511        proc_t p;
 512{
 513        struct nfsv3_fsinfo *fsp;
 514        caddr_t cp;
 515        long t1, t2;
 516        u_long *tl;
 517        int prefsize, maxsize;
 518        caddr_t bpos, dpos, cp2;
 519        int error = 0, retattr;
 520        mbuf_t mreq, mrep, md, mb, mb2;
 521        u_int64_t xid;
 522
 523        nfsm_reqhead(NFSX_FH(1));
 524        if (error)
 525                return (error);
 526        OSAddAtomic(1, (SInt32*)&nfsstats.rpccnt[NFSPROC_FSINFO]);
 527        nfsm_fhtom(vp, 1);
 528        nfsm_request(vp, NFSPROC_FSINFO, p, cred, &xid);
 529        if (mrep) {
 530                nfsm_postop_attr_update(vp, 1, retattr, &xid);
 531        }
 532        if (!error) {
 533                nfsm_dissect(fsp, struct nfsv3_fsinfo *, NFSX_V3FSINFO);
 534                prefsize = fxdr_unsigned(u_long, fsp->fs_wtpref);
 535                if (prefsize < nmp->nm_wsize)
 536                        nmp->nm_wsize = (prefsize + NFS_FABLKSIZE - 1) &
 537                                ~(NFS_FABLKSIZE - 1);
 538                maxsize = fxdr_unsigned(u_long, fsp->fs_wtmax);
 539                if (maxsize < nmp->nm_wsize) {
 540                        nmp->nm_wsize = maxsize & ~(NFS_FABLKSIZE - 1);
 541                        if (nmp->nm_wsize == 0)
 542                                nmp->nm_wsize = maxsize;
 543                }
 544                prefsize = fxdr_unsigned(u_long, fsp->fs_rtpref);
 545                if (prefsize < nmp->nm_rsize)
 546                        nmp->nm_rsize = (prefsize + NFS_FABLKSIZE - 1) &
 547                                ~(NFS_FABLKSIZE - 1);
 548                maxsize = fxdr_unsigned(u_long, fsp->fs_rtmax);
 549                if (maxsize < nmp->nm_rsize) {
 550                        nmp->nm_rsize = maxsize & ~(NFS_FABLKSIZE - 1);
 551                        if (nmp->nm_rsize == 0)
 552                                nmp->nm_rsize = maxsize;
 553                }
 554                prefsize = fxdr_unsigned(u_long, fsp->fs_dtpref);
 555                if (prefsize < nmp->nm_readdirsize)
 556                        nmp->nm_readdirsize = prefsize;
 557                if (maxsize < nmp->nm_readdirsize) {
 558                        nmp->nm_readdirsize = maxsize;
 559                }
 560                fxdr_hyper(&fsp->fs_maxfilesize, &nmp->nm_fsinfo.maxfilesize);
 561                nmp->nm_fsinfo.fsproperties = fxdr_unsigned(u_long, fsp->fs_properties);
 562                nmp->nm_state |= NFSSTA_GOTFSINFO;
 563        }
 564        nfsm_reqdone;
 565        return (error);
 566}
 567
 568/*
 569 * Mount a remote root fs via. nfs. This depends on the info in the
 570 * nfs_diskless structure that has been filled in properly by some primary
 571 * bootstrap.
 572 * It goes something like this:
 573 * - do enough of "ifconfig" by calling ifioctl() so that the system
 574 *   can talk to the server
 575 * - If nfs_diskless.mygateway is filled in, use that address as
 576 *   a default gateway.
 577 * - hand craft the swap nfs vnode hanging off a fake mount point
 578 *      if swdevt[0].sw_dev == NODEV
 579 * - build the rootfs mount point and call mountnfs() to do the rest.
 580 */
 581int
 582nfs_mountroot()
 583{
 584        struct nfs_diskless nd;
 585        struct nfs_vattr nvattr;
 586        mount_t mp;
 587        vnode_t vp;
 588        proc_t procp;
 589        int error;
 590#if !defined(NO_MOUNT_PRIVATE)
 591        mount_t mppriv;
 592        vnode_t vppriv;
 593#endif /* NO_MOUNT_PRIVATE */
 594        int v3, sotype;
 595
 596        procp = current_proc(); /* XXX */
 597
 598        /*
 599         * Call nfs_boot_init() to fill in the nfs_diskless struct.
 600         * Note: networking must already have been configured before
 601         * we're called.
 602         */
 603        bzero((caddr_t) &nd, sizeof(nd));
 604        error = nfs_boot_init(&nd, procp);
 605        if (error) {
 606                panic("nfs_boot_init failed with %d\n", error);
 607        }
 608
 609        /*
 610         * Try NFSv3 first, then fallback to NFSv2.
 611         * Likewise, try TCP first, then fall back to UDP.
 612         */
 613        v3 = 1;
 614        sotype = SOCK_STREAM;
 615
 616tryagain:
 617        error = nfs_boot_getfh(&nd, procp, v3, sotype);
 618        if (error) {
 619                if (error == EHOSTDOWN || error == EHOSTUNREACH) {
 620                        if (nd.nd_root.ndm_path)
 621                                FREE_ZONE(nd.nd_root.ndm_path, 
 622                                          MAXPATHLEN, M_NAMEI);
 623                        if (nd.nd_private.ndm_path)
 624                                FREE_ZONE(nd.nd_private.ndm_path, 
 625                                          MAXPATHLEN, M_NAMEI);
 626                        return (error);
 627                }
 628                if (v3) {
 629                        if (sotype == SOCK_STREAM) {
 630                                printf("nfs_boot_getfh(v3,TCP) failed with %d, trying UDP...\n", error);
 631                                sotype = SOCK_DGRAM;
 632                                goto tryagain;
 633                        }
 634                        printf("nfs_boot_getfh(v3,UDP) failed with %d, trying v2...\n", error);
 635                        v3 = 0;
 636                        sotype = SOCK_STREAM;
 637                        goto tryagain;
 638                } else if (sotype == SOCK_STREAM) {
 639                        printf("nfs_boot_getfh(v2,TCP) failed with %d, trying UDP...\n", error);
 640                        sotype = SOCK_DGRAM;
 641                        goto tryagain;
 642                }
 643                panic("nfs_boot_getfh(v2,UDP) failed with %d\n", error);
 644        }
 645
 646        /*
 647         * Create the root mount point.
 648         */
 649#if !defined(NO_MOUNT_PRIVATE)
 650        if ((error = nfs_mount_diskless(&nd.nd_root, "/", MNT_RDONLY|MNT_ROOTFS, &vp, &mp)))
 651#else
 652        if ((error = nfs_mount_diskless(&nd.nd_root, "/", MNT_ROOTFS, &vp, &mp)))
 653#endif /* NO_MOUNT_PRIVATE */
 654        {
 655                if (v3) {
 656                        if (sotype == SOCK_STREAM) {
 657                                printf("nfs_mount_diskless(v3,TCP) failed with %d, trying UDP...\n", error);
 658                                sotype = SOCK_DGRAM;
 659                                goto tryagain;
 660                        }
 661                        printf("nfs_mount_diskless(v3,UDP) failed with %d, trying v2...\n", error);
 662                        v3 = 0;
 663                        sotype = SOCK_STREAM;
 664                        goto tryagain;
 665                } else if (sotype == SOCK_STREAM) {
 666                        printf("nfs_mount_diskless(v2,TCP) failed with %d, trying UDP...\n", error);
 667                        sotype = SOCK_DGRAM;
 668                        goto tryagain;
 669                }
 670                panic("nfs_mount_diskless(v2,UDP) root failed with %d\n", error);
 671        }
 672        printf("root on %s\n", (char *)&nd.nd_root.ndm_host);
 673
 674        vfs_unbusy(mp);
 675        mount_list_add(mp);
 676        rootvp = vp;
 677        
 678#if !defined(NO_MOUNT_PRIVATE)
 679        if (nd.nd_private.ndm_saddr.sin_addr.s_addr) {
 680            error = nfs_mount_diskless_private(&nd.nd_private, "/private",
 681                                               0, &vppriv, &mppriv);
 682            if (error) {
 683                panic("nfs_mount_diskless private failed with %d\n", error);
 684            }
 685            printf("private on %s\n", (char *)&nd.nd_private.ndm_host);
 686            
 687            vfs_unbusy(mppriv);
 688            mount_list_add(mppriv);
 689        }
 690
 691#endif /* NO_MOUNT_PRIVATE */
 692
 693        if (nd.nd_root.ndm_path)
 694                FREE_ZONE(nd.nd_root.ndm_path, MAXPATHLEN, M_NAMEI);
 695        if (nd.nd_private.ndm_path)
 696                FREE_ZONE(nd.nd_private.ndm_path, MAXPATHLEN, M_NAMEI);
 697
 698        /* Get root attributes (for the time). */
 699        error = nfs_getattr(vp, &nvattr, kauth_cred_get(), procp);
 700        if (error) panic("nfs_mountroot: getattr for root");
 701        return (0);
 702}
 703
 704/*
 705 * Internal version of mount system call for diskless setup.
 706 */
 707static int
 708nfs_mount_diskless(
 709        struct nfs_dlmount *ndmntp,
 710        const char *mntname,
 711        int mntflag,
 712        vnode_t *vpp,
 713        mount_t *mpp)
 714{
 715        struct user_nfs_args args;
 716        mount_t mp;
 717        mbuf_t m;
 718        int error;
 719        proc_t procp;
 720
 721        procp = current_proc(); /* XXX */
 722
 723        if ((error = vfs_rootmountalloc("nfs", ndmntp->ndm_host, &mp))) {
 724                printf("nfs_mount_diskless: NFS not configured");
 725                return (error);
 726        }
 727
 728        mp->mnt_flag |= mntflag;
 729        if (!(mntflag & MNT_RDONLY))
 730                mp->mnt_flag &= ~MNT_RDONLY;
 731
 732        /* Initialize mount args. */
 733        bzero((caddr_t) &args, sizeof(args));
 734        args.addr     = CAST_USER_ADDR_T(&ndmntp->ndm_saddr);
 735        args.addrlen  = ndmntp->ndm_saddr.sin_len;
 736        args.sotype   = ndmntp->ndm_sotype;
 737        args.fh       = CAST_USER_ADDR_T(&ndmntp->ndm_fh[0]);
 738        args.fhsize   = ndmntp->ndm_fhlen;
 739        args.hostname = CAST_USER_ADDR_T(ndmntp->ndm_host);
 740        args.flags    = NFSMNT_RESVPORT;
 741        if (ndmntp->ndm_nfsv3)
 742                args.flags |= NFSMNT_NFSV3;
 743
 744        error = mbuf_get(MBUF_WAITOK, MBUF_TYPE_SONAME, &m);
 745        if (error) {
 746                printf("nfs_mount_diskless: mbuf_get(soname) failed");
 747                return (error);
 748        }
 749        mbuf_setlen(m, ndmntp->ndm_saddr.sin_len);
 750        bcopy((caddr_t)args.addr, mbuf_data(m), ndmntp->ndm_saddr.sin_len);
 751        if ((error = mountnfs(&args, mp, m, procp, vpp))) {
 752                printf("nfs_mountroot: mount %s failed: %d\n", mntname, error);
 753                // XXX vfs_rootmountfailed(mp);
 754                mount_list_lock();
 755                mp->mnt_vtable->vfc_refcount--;
 756                mount_list_unlock();
 757                vfs_unbusy(mp);
 758                mount_lock_destroy(mp);
 759                FREE_ZONE(mp, sizeof(struct mount), M_MOUNT);
 760                return (error);
 761        }
 762        *mpp = mp;
 763        return (0);
 764}
 765
 766#if !defined(NO_MOUNT_PRIVATE)
 767/*
 768 * Internal version of mount system call to mount "/private"
 769 * separately in diskless setup
 770 */
 771static int
 772nfs_mount_diskless_private(
 773        struct nfs_dlmount *ndmntp,
 774        const char *mntname,
 775        int mntflag,
 776        vnode_t *vpp,
 777        mount_t *mpp)
 778{
 779        struct user_nfs_args args;
 780        mount_t mp;
 781        mbuf_t m;
 782        int error;
 783        proc_t procp;
 784        struct vfstable *vfsp;
 785        struct nameidata nd;
 786        vnode_t vp;
 787        struct vfs_context context;
 788
 789        procp = current_proc(); /* XXX */
 790        context.vc_proc = procp;
 791        context.vc_ucred = kauth_cred_get();
 792
 793        {
 794        /*
 795         * mimic main()!. Temporarily set up rootvnode and other stuff so
 796         * that namei works. Need to undo this because main() does it, too
 797         */
 798                struct filedesc *fdp;   /* pointer to file descriptor state */
 799                fdp = procp->p_fd;
 800                mountlist.tqh_first->mnt_flag |= MNT_ROOTFS;
 801
 802                /* Get the vnode for '/'. Set fdp->fd_cdir to reference it. */
 803                if (VFS_ROOT(mountlist.tqh_first, &rootvnode, NULL))
 804                        panic("cannot find root vnode");
 805                error = vnode_ref(rootvnode);
 806                if (error) {
 807                        printf("nfs_mountroot: vnode_ref() failed on root vnode!\n");
 808                        return (error);
 809                }
 810                fdp->fd_cdir = rootvnode;
 811                fdp->fd_rdir = NULL;
 812        }
 813
 814        /*
 815         * Get vnode to be covered
 816         */
 817        NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE32,
 818            mntname, &context);
 819        if ((error = namei(&nd))) {
 820                printf("nfs_mountroot: private namei failed!\n");
 821                return (error);
 822        }
 823        {
 824                /* undo vnode_ref() in mimic main()! */
 825                vnode_rele(rootvnode);
 826        }
 827        nameidone(&nd);
 828        vp = nd.ni_vp;
 829
 830        if ((error = VNOP_FSYNC(vp, MNT_WAIT, &context)) ||
 831            (error = buf_invalidateblks(vp, BUF_WRITE_DATA, 0, 0))) {
 832                vnode_put(vp);
 833                return (error);
 834        }
 835        if (vnode_vtype(vp) != VDIR) {
 836                vnode_put(vp);
 837                return (ENOTDIR);
 838        }
 839        for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next)
 840                if (!strcmp(vfsp->vfc_name, "nfs"))
 841                        break;
 842        if (vfsp == NULL) {
 843                printf("nfs_mountroot: private NFS not configured\n");
 844                vnode_put(vp);
 845                return (ENODEV);
 846        }
 847        if (vnode_mountedhere(vp) != NULL) {
 848                vnode_put(vp);
 849                return (EBUSY);
 850        }
 851
 852        /*
 853         * Allocate and initialize the filesystem.
 854         */
 855        mp = _MALLOC_ZONE((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK);
 856        if (!mp) {
 857                printf("nfs_mountroot: unable to allocate mount structure\n");
 858                vnode_put(vp);
 859                return (ENOMEM);
 860        }
 861        bzero((char *)mp, (u_long)sizeof(struct mount));
 862
 863        /* Initialize the default IO constraints */
 864        mp->mnt_maxreadcnt = mp->mnt_maxwritecnt = MAXPHYS;
 865        mp->mnt_segreadcnt = mp->mnt_segwritecnt = 32;
 866
 867        mount_lock_init(mp);
 868        TAILQ_INIT(&mp->mnt_vnodelist);
 869        TAILQ_INIT(&mp->mnt_workerqueue);
 870        TAILQ_INIT(&mp->mnt_newvnodes);
 871        (void)vfs_busy(mp, LK_NOWAIT);
 872        TAILQ_INIT(&mp->mnt_vnodelist);
 873        mount_list_lock();
 874        vfsp->vfc_refcount++;
 875        mount_list_unlock();
 876        mp->mnt_vtable = vfsp;
 877        mp->mnt_op = vfsp->vfc_vfsops;
 878        // mp->mnt_stat.f_type = vfsp->vfc_typenum;
 879        mp->mnt_flag = mntflag;
 880        mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
 881        strncpy(mp->mnt_vfsstat.f_fstypename, vfsp->vfc_name, MFSNAMELEN-1);
 882        vp->v_mountedhere = mp;
 883        mp->mnt_vnodecovered = vp;
 884        mp->mnt_vfsstat.f_owner = kauth_cred_getuid(kauth_cred_get());
 885        (void) copystr(mntname, mp->mnt_vfsstat.f_mntonname, MNAMELEN - 1, 0);
 886        (void) copystr(ndmntp->ndm_host, mp->mnt_vfsstat.f_mntfromname, MNAMELEN - 1, 0);
 887
 888        /* Initialize mount args. */
 889        bzero((caddr_t) &args, sizeof(args));
 890        args.addr     = CAST_USER_ADDR_T(&ndmntp->ndm_saddr);
 891        args.addrlen  = ndmntp->ndm_saddr.sin_len;
 892        args.sotype   = ndmntp->ndm_sotype;
 893        args.fh       = CAST_USER_ADDR_T(ndmntp->ndm_fh);
 894        args.fhsize   = ndmntp->ndm_fhlen;
 895        args.hostname = CAST_USER_ADDR_T(ndmntp->ndm_host);
 896        args.flags    = NFSMNT_RESVPORT;
 897        if (ndmntp->ndm_nfsv3)
 898                args.flags |= NFSMNT_NFSV3;
 899
 900        error = mbuf_get(MBUF_WAITOK, MBUF_TYPE_SONAME, &m);
 901        if (error) {
 902                printf("nfs_mount_diskless_private: mbuf_get(soname) failed");
 903                return (error);
 904        }
 905        mbuf_setlen(m, ndmntp->ndm_saddr.sin_len);
 906        bcopy((caddr_t)args.addr, mbuf_data(m), ndmntp->ndm_saddr.sin_len);
 907        if ((error = mountnfs(&args, mp, m, procp, &vp))) {
 908                printf("nfs_mountroot: mount %s failed: %d\n", mntname, error);
 909                mount_list_lock();
 910                vfsp->vfc_refcount--;
 911                mount_list_unlock();
 912                vfs_unbusy(mp);
 913                mount_lock_destroy(mp);
 914                FREE_ZONE(mp, sizeof (struct mount), M_MOUNT);
 915                return (error);
 916        }
 917
 918        *mpp = mp;
 919        *vpp = vp;
 920        return (0);
 921}
 922#endif /* NO_MOUNT_PRIVATE */
 923
 924/*
 925 * VFS Operations.
 926 *
 927 * mount system call
 928 */
 929static int
 930nfs_mount(mount_t mp, vnode_t vp, user_addr_t data, vfs_context_t context)
 931{
 932        proc_t p = vfs_context_proc(context);
 933        int error, argsvers;
 934        struct user_nfs_args args;
 935        struct nfs_args tempargs;
 936        mbuf_t nam;
 937        size_t len;
 938        u_char nfh[NFSX_V3FHMAX];
 939        char *mntfrom;
 940
 941        error = copyin(data, (caddr_t)&argsvers, sizeof (argsvers));
 942        if (error)
 943                return (error);
 944
 945        switch (argsvers) {
 946        case 3:
 947                if (vfs_context_is64bit(context))
 948                        error = copyin(data, (caddr_t)&args, sizeof (struct user_nfs_args3));
 949                else
 950                        error = copyin(data, (caddr_t)&tempargs, sizeof (struct nfs_args3));
 951                break;
 952        case 4:
 953                if (vfs_context_is64bit(context))
 954                        error = copyin(data, (caddr_t)&args, sizeof (args));
 955                else
 956                        error = copyin(data, (caddr_t)&tempargs, sizeof (tempargs));
 957                break;
 958        default:
 959                return (EPROGMISMATCH);
 960        }
 961        if (error)
 962                return (error);
 963
 964        if (!vfs_context_is64bit(context)) {
 965                args.version = tempargs.version;
 966                args.addrlen = tempargs.addrlen;
 967                args.sotype = tempargs.sotype;
 968                args.proto = tempargs.proto;
 969                args.fhsize = tempargs.fhsize;
 970                args.flags = tempargs.flags;
 971                args.wsize = tempargs.wsize;
 972                args.rsize = tempargs.rsize;
 973                args.readdirsize = tempargs.readdirsize;
 974                args.timeo = tempargs.timeo;
 975                args.retrans = tempargs.retrans;
 976                args.maxgrouplist = tempargs.maxgrouplist;
 977                args.readahead = tempargs.readahead;
 978                args.leaseterm = tempargs.leaseterm;
 979                args.deadthresh = tempargs.deadthresh;
 980                args.addr = CAST_USER_ADDR_T(tempargs.addr);
 981                args.fh = CAST_USER_ADDR_T(tempargs.fh);
 982                args.hostname = CAST_USER_ADDR_T(tempargs.hostname);
 983                if (argsvers >= 4) {
 984                        args.acregmin = tempargs.acregmin;
 985                        args.acregmax = tempargs.acregmax;
 986                        args.acdirmin = tempargs.acdirmin;
 987                        args.acdirmax = tempargs.acdirmax;
 988                }
 989        }
 990
 991        if (args.fhsize < 0 || args.fhsize > NFSX_V3FHMAX)
 992                return (EINVAL);
 993        error = copyin(args.fh, (caddr_t)nfh, args.fhsize);
 994        if (error)
 995                return (error);
 996
 997        mntfrom = &vfs_statfs(mp)->f_mntfromname[0];
 998        error = copyinstr(args.hostname, mntfrom, MAXPATHLEN-1, &len);
 999        if (error)
1000                return (error);
1001        bzero(&mntfrom[len], MAXPATHLEN - len);
1002
1003        /* sockargs() call must be after above copyin() calls */
1004        error = sockargs(&nam, args.addr, args.addrlen, MBUF_TYPE_SONAME);
1005        if (error)
1006                return (error);
1007
1008        args.fh = CAST_USER_ADDR_T(&nfh[0]);
1009        error = mountnfs(&args, mp, nam, p, &vp);
1010        return (error);
1011}
1012
1013/*
1014 * Common code for mount and mountroot
1015 */
1016static int
1017mountnfs(
1018        struct user_nfs_args *argp,
1019        mount_t mp,
1020        mbuf_t nam,
1021        proc_t p,
1022        vnode_t *vpp)
1023{
1024        struct nfsmount *nmp;
1025        struct nfsnode *np;
1026        int error, maxio;
1027        struct nfs_vattr nvattrs;
1028        struct vfs_context context; /* XXX get from caller? */
1029        u_int64_t xid;
1030
1031        /*
1032         * Silently clear NFSMNT_NOCONN if it's a TCP mount, it makes
1033         * no sense in that context.
1034         */
1035        if (argp->sotype == SOCK_STREAM)
1036                argp->flags &= ~NFSMNT_NOCONN;
1037        
1038        if (vfs_flags(mp) & MNT_UPDATE) {
1039                nmp = VFSTONFS(mp);
1040                /* update paths, file handles, etc, here        XXX */
1041                mbuf_freem(nam);
1042                return (0);
1043        } else {
1044                MALLOC_ZONE(nmp, struct nfsmount *,
1045                                sizeof (struct nfsmount), M_NFSMNT, M_WAITOK);
1046                if (!nmp) {
1047                        mbuf_freem(nam);
1048                        return (ENOMEM);
1049                }
1050                bzero((caddr_t)nmp, sizeof (struct nfsmount));
1051                TAILQ_INIT(&nmp->nm_uidlruhead);
1052                TAILQ_INIT(&nmp->nm_bufq);
1053                vfs_setfsprivate(mp, nmp);
1054        }
1055
1056        /* setup defaults */
1057        nmp->nm_timeo = NFS_TIMEO;
1058        nmp->nm_retry = NFS_RETRANS;
1059        if (argp->sotype == SOCK_DGRAM) {
1060                nmp->nm_wsize = NFS_DGRAM_WSIZE;
1061                nmp->nm_rsize = NFS_DGRAM_RSIZE;
1062        } else {
1063                nmp->nm_wsize = NFS_WSIZE;
1064                nmp->nm_rsize = NFS_RSIZE;
1065        }
1066        nmp->nm_readdirsize = NFS_READDIRSIZE;
1067        nmp->nm_numgrps = NFS_MAXGRPS;
1068        nmp->nm_readahead = NFS_DEFRAHEAD;
1069        nmp->nm_tprintf_delay = nfs_tprintf_delay;
1070        if (nmp->nm_tprintf_delay < 0)
1071                nmp->nm_tprintf_delay = 0;
1072        nmp->nm_tprintf_initial_delay = nfs_tprintf_initial_delay;
1073        if (nmp->nm_tprintf_initial_delay < 0)
1074                nmp->nm_tprintf_initial_delay = 0;
1075        nmp->nm_acregmin = NFS_MINATTRTIMO;
1076        nmp->nm_acregmax = NFS_MAXATTRTIMO;
1077        nmp->nm_acdirmin = NFS_MINDIRATTRTIMO;
1078        nmp->nm_acdirmax = NFS_MAXDIRATTRTIMO;
1079
1080        vfs_getnewfsid(mp);
1081        nmp->nm_mountp = mp;
1082        vfs_setauthopaque(mp);
1083        nmp->nm_flag = argp->flags;
1084        nmp->nm_nam = nam;
1085
1086        if ((argp->flags & NFSMNT_TIMEO) && argp->timeo > 0) {
1087                nmp->nm_timeo = (argp->timeo * NFS_HZ + 5) / 10;
1088                if (nmp->nm_timeo < NFS_MINTIMEO)
1089                        nmp->nm_timeo = NFS_MINTIMEO;
1090                else if (nmp->nm_timeo > NFS_MAXTIMEO)
1091                        nmp->nm_timeo = NFS_MAXTIMEO;
1092        }
1093
1094        if ((argp->flags & NFSMNT_RETRANS) && argp->retrans > 1) {
1095                nmp->nm_retry = argp->retrans;
1096                if (nmp->nm_retry > NFS_MAXREXMIT)
1097                        nmp->nm_retry = NFS_MAXREXMIT;
1098        }
1099
1100        if (argp->flags & NFSMNT_NFSV3) {
1101                if (argp->sotype == SOCK_DGRAM)
1102                        maxio = NFS_MAXDGRAMDATA;
1103                else
1104                        maxio = NFS_MAXDATA;
1105        } else
1106                maxio = NFS_V2MAXDATA;
1107
1108        if ((argp->flags & NFSMNT_WSIZE) && argp->wsize > 0) {
1109                nmp->nm_wsize = argp->wsize;
1110                /* Round down to multiple of blocksize */
1111                nmp->nm_wsize &= ~(NFS_FABLKSIZE - 1);
1112                if (nmp->nm_wsize <= 0)
1113                        nmp->nm_wsize = NFS_FABLKSIZE;
1114        }
1115        if (nmp->nm_wsize > maxio)
1116                nmp->nm_wsize = maxio;
1117        if (nmp->nm_wsize > MAXBSIZE)
1118                nmp->nm_wsize = MAXBSIZE;
1119
1120        if ((argp->flags & NFSMNT_RSIZE) && argp->rsize > 0) {
1121                nmp->nm_rsize = argp->rsize;
1122                /* Round down to multiple of blocksize */
1123                nmp->nm_rsize &= ~(NFS_FABLKSIZE - 1);
1124                if (nmp->nm_rsize <= 0)
1125                        nmp->nm_rsize = NFS_FABLKSIZE;
1126        }
1127        if (nmp->nm_rsize > maxio)
1128                nmp->nm_rsize = maxio;
1129        if (nmp->nm_rsize > MAXBSIZE)
1130                nmp->nm_rsize = MAXBSIZE;
1131
1132        if ((argp->flags & NFSMNT_READDIRSIZE) && argp->readdirsize > 0) {
1133                nmp->nm_readdirsize = argp->readdirsize;
1134        }
1135        if (nmp->nm_readdirsize > maxio)
1136                nmp->nm_readdirsize = maxio;
1137        if (nmp->nm_readdirsize > nmp->nm_rsize)
1138                nmp->nm_readdirsize = nmp->nm_rsize;
1139
1140        if ((argp->flags & NFSMNT_MAXGRPS) && argp->maxgrouplist >= 0 &&
1141                argp->maxgrouplist <= NFS_MAXGRPS)
1142                nmp->nm_numgrps = argp->maxgrouplist;
1143        if ((argp->flags & NFSMNT_READAHEAD) && argp->readahead >= 0 &&
1144                argp->readahead <= NFS_MAXRAHEAD)
1145                nmp->nm_readahead = argp->readahead;
1146
1147        if (argp->version >= 4) {
1148                if ((argp->flags & NFSMNT_ACREGMIN) && argp->acregmin >= 0)
1149                        nmp->nm_acregmin = argp->acregmin;
1150                if ((argp->flags & NFSMNT_ACREGMAX) && argp->acregmax >= 0)
1151                        nmp->nm_acregmax = argp->acregmax;
1152                if ((argp->flags & NFSMNT_ACDIRMIN) && argp->acdirmin >= 0)
1153                        nmp->nm_acdirmin = argp->acdirmin;
1154                if ((argp->flags & NFSMNT_ACDIRMAX) && argp->acdirmax >= 0)
1155                        nmp->nm_acdirmax = argp->acdirmax;
1156                if (nmp->nm_acregmin > nmp->nm_acregmax)
1157                        nmp->nm_acregmin = nmp->nm_acregmax;
1158                if (nmp->nm_acdirmin > nmp->nm_acdirmax)
1159                        nmp->nm_acdirmin = nmp->nm_acdirmax;
1160        }
1161
1162        /* Set up the sockets and per-host congestion */
1163        nmp->nm_sotype = argp->sotype;
1164        nmp->nm_soproto = argp->proto;
1165
1166        /* make sure mbuf constants are set up */
1167        if (!nfs_mbuf_mlen)
1168                nfs_mbuf_init();
1169
1170        /*
1171         * For Connection based sockets (TCP,...) defer the connect until
1172         * the first request, in case the server is not responding.
1173         */
1174        if (nmp->nm_sotype == SOCK_DGRAM &&
1175                (error = nfs_connect(nmp, (struct nfsreq *)0)))
1176                goto bad;
1177
1178        /*
1179         * Get file attributes for the mountpoint.  These are needed
1180         * in order to properly create the root vnode.
1181         */
1182        // LP64todo - fix CAST_DOWN of argp->fh
1183        error = nfs_getattr_no_vnode(mp, CAST_DOWN(caddr_t, argp->fh), argp->fhsize,
1184                        proc_ucred(p), p, &nvattrs, &xid);
1185        if (error) {
1186                /*
1187                 * we got problems... we couldn't get the attributes
1188                 * from the NFS server... so the mount fails.
1189                 */
1190                goto bad;
1191        }
1192
1193        /*
1194         * A reference count is needed on the nfsnode representing the
1195         * remote root.  If this object is not persistent, then backward
1196         * traversals of the mount point (i.e. "..") will not work if
1197         * the nfsnode gets flushed out of the cache. UFS does not have
1198         * this problem, because one can identify root inodes by their
1199         * number == ROOTINO (2).
1200         */
1201        error = nfs_nget(mp, NULL, NULL, CAST_DOWN(caddr_t, argp->fh), argp->fhsize,
1202                        &nvattrs, &xid, NG_MARKROOT, &np);
1203        if (error)
1204                goto bad;
1205
1206        /*
1207         * save this vnode pointer. That way nfs_unmount()
1208         * does not need to call nfs_nget() just get it to drop
1209         * this vnode reference.
1210         */
1211        nmp->nm_dvp = *vpp = NFSTOV(np);
1212        /* get usecount and drop iocount */
1213        error = vnode_ref(*vpp);
1214        if (error) {
1215                vnode_put(*vpp);
1216                goto bad;
1217        }
1218        vnode_put(*vpp);
1219
1220        /*
1221         * Set the mount point's block I/O size.
1222         * We really need to do this after we get info back from
1223         * the server about what its preferred I/O sizes are.
1224         */
1225        if (nmp->nm_flag & NFSMNT_NFSV3)
1226                nfs_fsinfo(nmp, *vpp, proc_ucred(p), p);
1227        vfs_statfs(mp)->f_iosize = nfs_iosize(nmp);
1228
1229        /*
1230         * V3 mounts give us a (relatively) reliable remote access(2)
1231         * call, so advertise the fact.
1232         *
1233         * XXX this may not be the best way to go, as the granularity
1234         *     offered isn't a good match to our needs.
1235         */
1236        if (nmp->nm_flag & NFSMNT_NFSV3)
1237                vfs_setauthopaqueaccess(mp);
1238
1239        /*
1240         * Do statfs to ensure static info gets set to reasonable values.
1241         */
1242        context.vc_proc = p;
1243        context.vc_ucred = proc_ucred(p);
1244        nfs_statfs(mp, vfs_statfs(mp), &context);
1245
1246        if (nmp->nm_flag & NFSMNT_RESVPORT)
1247                nfs_resv_mounts++;
1248        nmp->nm_state |= NFSSTA_MOUNTED;
1249        return (0);
1250bad:
1251        nfs_disconnect(nmp);
1252        FREE_ZONE((caddr_t)nmp, sizeof (struct nfsmount), M_NFSMNT);
1253        mbuf_freem(nam);
1254        return (error);
1255}
1256
1257
1258/*
1259 * unmount system call
1260 */
1261static int
1262nfs_unmount(
1263        mount_t mp,
1264        int mntflags,
1265        __unused vfs_context_t context)
1266{
1267        register struct nfsmount *nmp;
1268        vnode_t vp;
1269        int error, flags = 0;
1270
1271        nmp = VFSTONFS(mp);
1272        /*
1273         * During a force unmount we want to...
1274         *   Mark that we are doing a force unmount.
1275         *   Make the mountpoint soft.
1276         */
1277        if (mntflags & MNT_FORCE) {
1278                flags |= FORCECLOSE;
1279                nmp->nm_state |= NFSSTA_FORCE;
1280                nmp->nm_flag |= NFSMNT_SOFT;
1281        }
1282        /*
1283         * Goes something like this..
1284         * - Call vflush() to clear out vnodes for this file system,
1285         *   except for the swap files. Deal with them in 2nd pass.
1286         * - Decrement reference on the vnode representing remote root.
1287         * - Close the socket
1288         * - Free up the data structures
1289         */
1290        vp = nmp->nm_dvp;
1291        
1292        /*
1293         * vflush will check for busy vnodes on mountpoint. 
1294         * Will do the right thing for MNT_FORCE. That is, we should
1295         * not get EBUSY back.
1296         */
1297        error = vflush(mp, vp, SKIPSWAP | flags);
1298        if (mntflags & MNT_FORCE) {
1299                error = vflush(mp, NULLVP, flags); /* locks vp in the process */
1300        } else {
1301                if (vnode_isinuse(vp, 1))
1302                        return (EBUSY);
1303                error = vflush(mp, vp, flags);
1304        }
1305        if (error)
1306                return (error);
1307
1308        nmp->nm_state &= ~NFSSTA_MOUNTED;
1309        if (nmp->nm_flag & NFSMNT_RESVPORT) {
1310                if (--nfs_resv_mounts == 0)
1311                        nfs_bind_resv_thread_wake();
1312        }
1313
1314        /*
1315         * Release the root vnode reference held by mountnfs()
1316         */
1317        vnode_rele(vp);
1318
1319        (void)vflush(mp, NULLVP, FORCECLOSE);
1320        vfs_setfsprivate(mp, 0); /* don't want to end up using stale vp */
1321
1322        nfs_disconnect(nmp);
1323        mbuf_freem(nmp->nm_nam);
1324
1325        if ((nmp->nm_flag & NFSMNT_KERB) == 0) {
1326                struct nfsreq *rp;
1327                /*
1328                 * Loop through outstanding request list and remove dangling
1329                 * references to defunct nfsmount struct
1330                 */
1331                for (rp = nfs_reqq.tqh_first; rp; rp = rp->r_chain.tqe_next)
1332                        if (rp->r_nmp == nmp)
1333                                rp->r_nmp = (struct nfsmount *)0;
1334                /* Need to wake up any rcvlock waiters so they notice the unmount. */
1335                if (nmp->nm_state & NFSSTA_WANTRCV) {
1336                        nmp->nm_state &= ~NFSSTA_WANTRCV;
1337                        wakeup(&nmp->nm_state);
1338                }
1339                FREE_ZONE((caddr_t)nmp, sizeof (struct nfsmount), M_NFSMNT);
1340        }
1341        return (0);
1342}
1343
1344/*
1345 * Return root of a filesystem
1346 */
1347static int
1348nfs_root(mount_t mp, vnode_t *vpp, __unused vfs_context_t context)
1349{
1350        vnode_t vp;
1351        struct nfsmount *nmp;
1352        int error;
1353        u_long vpid;
1354
1355        nmp = VFSTONFS(mp);
1356        vp = nmp->nm_dvp;
1357        vpid = vnode_vid(vp);
1358        while ((error = vnode_getwithvid(vp, vpid))) {
1359                /* vnode_get() may return ENOENT if the dir changes. */
1360                /* If that happens, just try it again, else return the error. */
1361                if ((error != ENOENT) || (vnode_vid(vp) == vpid))
1362                        return (error);
1363                vpid = vnode_vid(vp);
1364        }
1365        *vpp = vp;
1366        return (0);
1367}
1368
1369/*
1370 * Flush out the buffer cache
1371 */
1372
1373struct nfs_sync_cargs {
1374        vfs_context_t context;
1375        int    waitfor;
1376        int    error;
1377};
1378
1379static int
1380nfs_sync_callout(vnode_t vp, void *arg)
1381{
1382        struct nfs_sync_cargs *cargs = (struct nfs_sync_cargs*)arg;
1383        int error;
1384
1385        if (LIST_EMPTY(&VTONFS(vp)->n_dirtyblkhd))
1386                return (VNODE_RETURNED);
1387        if (VTONFS(vp)->n_flag & NWRBUSY)
1388                return (VNODE_RETURNED);
1389
1390        error = nfs_flush(vp, cargs->waitfor,
1391                        vfs_context_ucred(cargs->context),
1392                        vfs_context_proc(cargs->context), 0);
1393        if (error)
1394                cargs->error = error;
1395
1396        return (VNODE_RETURNED);
1397}
1398
1399static int
1400nfs_sync(mount_t mp, int waitfor, vfs_context_t context)
1401{
1402        struct nfs_sync_cargs cargs;
1403
1404        cargs.waitfor = waitfor;
1405        cargs.context = context;
1406        cargs.error = 0;
1407
1408        vnode_iterate(mp, 0, nfs_sync_callout, &cargs);
1409
1410        return (cargs.error);
1411}
1412
1413/*
1414 * NFS flat namespace lookup.
1415 * Currently unsupported.
1416 */
1417/*ARGSUSED*/
1418static int
1419nfs_vget(
1420        __unused mount_t mp,
1421        __unused ino64_t ino,
1422        __unused vnode_t *vpp,
1423        __unused vfs_context_t context)
1424{
1425
1426        return (ENOTSUP);
1427}
1428
1429/*
1430 * At this point, this should never happen
1431 */
1432/*ARGSUSED*/
1433static int
1434nfs_fhtovp(
1435        __unused mount_t mp,
1436        __unused int fhlen,
1437        __unused unsigned char *fhp,
1438        __unused vnode_t *vpp,
1439        __unused vfs_context_t context)
1440{
1441
1442        return (ENOTSUP);
1443}
1444
1445/*
1446 * Vnode pointer to File handle, should never happen either
1447 */
1448/*ARGSUSED*/
1449static int
1450nfs_vptofh(
1451        __unused vnode_t vp,
1452        __unused int *fhlenp,
1453        __unused unsigned char *fhp,
1454        __unused vfs_context_t context)
1455{
1456
1457        return (ENOTSUP);
1458}
1459
1460/*
1461 * Vfs start routine, a no-op.
1462 */
1463/*ARGSUSED*/
1464static int
1465nfs_start(
1466        __unused mount_t mp,
1467        __unused int flags,
1468        __unused vfs_context_t context)
1469{
1470
1471        return (0);
1472}
1473
1474/*
1475 * Do that sysctl thang...
1476 */
1477static int
1478nfs_sysctl(int *name, u_int namelen, user_addr_t oldp, size_t *oldlenp, 
1479           user_addr_t newp, size_t newlen, vfs_context_t context)
1480{
1481        int error = 0, val;
1482        struct sysctl_req *req = NULL;
1483        struct vfsidctl vc;
1484        struct user_vfsidctl user_vc;
1485        mount_t mp;
1486        struct nfsmount *nmp = NULL;
1487        struct vfsquery vq;
1488        boolean_t is_64_bit;
1489
1490        /*
1491         * All names at this level are terminal.
1492         */
1493        if(namelen > 1)
1494                return ENOTDIR; /* overloaded */
1495
1496        is_64_bit = vfs_context_is64bit(context);
1497
1498        /* common code for "new style" VFS_CTL sysctl, get the mount. */
1499        switch (name[0]) {
1500        case VFS_CTL_TIMEO:
1501        case VFS_CTL_QUERY:
1502        case VFS_CTL_NOLOCKS:
1503                req = CAST_DOWN(struct sysctl_req *, oldp);
1504                if (is_64_bit) {
1505                        error = SYSCTL_IN(req, &user_vc, sizeof(user_vc));
1506                        if (error)
1507                                 return (error);
1508                        mp = vfs_getvfs(&user_vc.vc_fsid);
1509                } 
1510                else {
1511                        error = SYSCTL_IN(req, &vc, sizeof(vc));
1512                        if (error)
1513                                return (error);
1514                        mp = vfs_getvfs(&vc.vc_fsid);
1515                }
1516                if (mp == NULL)
1517                        return (ENOENT);
1518                nmp = VFSTONFS(mp);
1519                if (nmp == NULL)
1520                        return (ENOENT);
1521                bzero(&vq, sizeof(vq));
1522                req->newidx = 0;
1523                if (is_64_bit) {
1524                        req->newptr = user_vc.vc_ptr;
1525                        req->newlen = (size_t)user_vc.vc_len;
1526                }
1527                else {
1528                        req->newptr = CAST_USER_ADDR_T(vc.vc_ptr);
1529                        req->newlen = vc.vc_len;
1530                }
1531        }
1532
1533        switch(name[0]) {
1534        case NFS_NFSSTATS:
1535                if(!oldp) {
1536                        *oldlenp = sizeof nfsstats;
1537                        return 0;
1538                }
1539
1540                if(*oldlenp < sizeof nfsstats) {
1541                        *oldlenp = sizeof nfsstats;
1542                        return ENOMEM;
1543                }
1544
1545                error = copyout(&nfsstats, oldp, sizeof nfsstats);
1546                if (error)
1547                        return (error);
1548
1549                if(newp && newlen != sizeof nfsstats)
1550                        return EINVAL;
1551
1552                if(newp) {
1553                        return copyin(newp, &nfsstats, sizeof nfsstats);
1554                }
1555                return 0;
1556        case VFS_CTL_NOLOCKS:
1557                val = (nmp->nm_flag & NFSMNT_NOLOCKS) ? 1 : 0;
1558                if (req->oldptr != USER_ADDR_NULL) {
1559                        error = SYSCTL_OUT(req, &val, sizeof(val));
1560                        if (error)
1561                                return (error);
1562                }
1563                if (req->newptr != USER_ADDR_NULL) {
1564                        error = SYSCTL_IN(req, &val, sizeof(val));
1565                        if (error)
1566                                return (error);
1567                        if (val)
1568                                nmp->nm_flag |= NFSMNT_NOLOCKS;
1569                        else
1570                                nmp->nm_flag &= ~NFSMNT_NOLOCKS;
1571                }
1572                break;
1573        case VFS_CTL_QUERY:
1574                if (nmp->nm_state & NFSSTA_TIMEO)
1575                        vq.vq_flags |= VQ_NOTRESP;
1576                if (!(nmp->nm_flag & NFSMNT_NOLOCKS) &&
1577                    (nmp->nm_state & NFSSTA_LOCKTIMEO))
1578                        vq.vq_flags |= VQ_NOTRESPLOCK;
1579                error = SYSCTL_OUT(req, &vq, sizeof(vq));
1580                break;
1581        case VFS_CTL_TIMEO:
1582                if (req->oldptr != USER_ADDR_NULL) {
1583                        error = SYSCTL_OUT(req, &nmp->nm_tprintf_initial_delay,
1584                            sizeof(nmp->nm_tprintf_initial_delay));
1585                        if (error)
1586                                return (error);
1587                }
1588                if (req->newptr != USER_ADDR_NULL) {
1589                        error = SYSCTL_IN(req, &nmp->nm_tprintf_initial_delay,
1590                            sizeof(nmp->nm_tprintf_initial_delay));
1591                        if (error)
1592                                return (error);
1593                        if (nmp->nm_tprintf_initial_delay < 0)
1594                                nmp->nm_tprintf_initial_delay = 0;
1595                }
1596                break;
1597        default:
1598                return (ENOTSUP);
1599        }
1600        return (error);
1601}
1602
1603
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.