/*      $NetBSD: ops_nfs.c,v 1.1.1.3 2015/01/17 16:34:15 christos Exp $ */

/*
* Copyright (c) 1997-2014 Erez Zadok
* Copyright (c) 1990 Jan-Simon Pendry
* Copyright (c) 1990 Imperial College of Science, Technology & Medicine
* Copyright (c) 1990 The Regents of the University of California.
* All rights reserved.
*
* This code is derived from software contributed to Berkeley by
* Jan-Simon Pendry at Imperial College, London.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
*    notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
*    notice, this list of conditions and the following disclaimer in the
*    documentation and/or other materials provided with the distribution.
* 3. Neither the name of the University nor the names of its contributors
*    may be used to endorse or promote products derived from this software
*    without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
*
* File: am-utils/amd/ops_nfs.c
*
*/

/*
* Network file system
*/

#ifdef HAVE_CONFIG_H
# include <config.h>
#endif /* HAVE_CONFIG_H */
#include <am_defs.h>
#include <amd.h>

/*
* Convert from nfsstat to UN*X error code
*/
#define unx_error(e)    ((int)(e))

/*
* FH_TTL is the time a file handle will remain in the cache since
* last being used.  If the file handle becomes invalid, then it
* will be flushed anyway.
*/
#define FH_TTL                  (5 * 60) /* five minutes */
#define FH_TTL_ERROR            (30) /* 30 seconds */
#define FHID_ALLOC()            (++fh_id)

/*
* The NFS layer maintains a cache of file handles.
* This is *fundamental* to the implementation and
* also allows quick remounting when a filesystem
* is accessed soon after timing out.
*
* The NFS server layer knows to flush this cache
* when a server goes down so avoiding stale handles.
*
* Each cache entry keeps a hard reference to
* the corresponding server.  This ensures that
* the server keepalive information is maintained.
*
* The copy of the sockaddr_in here is taken so
* that the port can be twiddled to talk to mountd
* instead of portmap or the NFS server as used
* elsewhere.
* The port# is flushed if a server goes down.
* The IP address is never flushed - we assume
* that the address of a mounted machine never
* changes.  If it does, then you have other
* problems...
*/
typedef struct fh_cache fh_cache;
struct fh_cache {
 qelem                 fh_q;           /* List header */
 wchan_t               fh_wchan;       /* Wait channel */
 int                   fh_error;       /* Valid data? */
 int                   fh_id;          /* Unique id */
 int                   fh_cid;         /* Callout id */
 u_long                fh_nfs_version; /* highest NFS version on host */
 am_nfs_handle_t       fh_nfs_handle;  /* Handle on filesystem */
 int                   fh_status;      /* Status of last rpc */
 struct sockaddr_in    fh_sin;         /* Address of mountd */
 fserver               *fh_fs;         /* Server holding filesystem */
 char                  *fh_path;       /* Filesystem on host */
};

/* forward definitions */
static int nfs_init(mntfs *mf);
static char *nfs_match(am_opts *fo);
static int nfs_mount(am_node *am, mntfs *mf);
static int nfs_umount(am_node *am, mntfs *mf);
static void nfs_umounted(mntfs *mf);
static int call_mountd(fh_cache *fp, u_long proc, fwd_fun f, wchan_t wchan);
static int webnfs_lookup(fh_cache *fp, fwd_fun f, wchan_t wchan);
static int fh_id = 0;

/*
* clamp the filehandle version to 3, so that we can fail back to nfsv3
* since nfsv4 does not have file handles
*/
#define SET_FH_VERSION(fs) \
   (fs)->fs_version > NFS_VERSION3 ? NFS_VERSION3 : (fs)->fs_version;

/* globals */
AUTH *nfs_auth;
qelem fh_head = {&fh_head, &fh_head};

/*
* Network file system operations
*/
am_ops nfs_ops =
{
 "nfs",
 nfs_match,
 nfs_init,
 nfs_mount,
 nfs_umount,
 amfs_error_lookup_child,
 amfs_error_mount_child,
 amfs_error_readdir,
 0,                            /* nfs_readlink */
 0,                            /* nfs_mounted */
 nfs_umounted,
 find_nfs_srvr,
 0,                            /* nfs_get_wchan */
 FS_MKMNT | FS_BACKGROUND | FS_AMQINFO,        /* nfs_fs_flags */
#ifdef HAVE_FS_AUTOFS
 AUTOFS_NFS_FS_FLAGS,
#endif /* HAVE_FS_AUTOFS */
};


static fh_cache *
find_nfs_fhandle_cache(opaque_t arg, int done)
{
 fh_cache *fp, *fp2 = NULL;
 int id = (long) arg;          /* for 64-bit archs */

 ITER(fp, fh_cache, &fh_head) {
   if (fp->fh_id == id) {
     fp2 = fp;
     break;
   }
 }

 if (fp2) {
   dlog("fh cache gives fp %#lx, fs %s", (unsigned long) fp2, fp2->fh_path);
 } else {
   dlog("fh cache search failed");
 }

 if (fp2 && !done) {
   fp2->fh_error = ETIMEDOUT;
   return 0;
 }

 return fp2;
}


/*
* Called when a filehandle appears via the mount protocol
*/
static void
got_nfs_fh_mount(voidp pkt, int len, struct sockaddr_in *sa, struct sockaddr_in *ia, opaque_t arg, int done)
{
 fh_cache *fp;
 struct fhstatus res;
#ifdef HAVE_FS_NFS3
 struct am_mountres3 res3;
#endif /* HAVE_FS_NFS3 */

 fp = find_nfs_fhandle_cache(arg, done);
 if (!fp)
   return;

 /*
  * retrieve the correct RPC reply for the file handle, based on the
  * NFS protocol version.
  */
#ifdef HAVE_FS_NFS3
 if (fp->fh_nfs_version == NFS_VERSION3) {
   memset(&res3, 0, sizeof(res3));
   fp->fh_error = pickup_rpc_reply(pkt, len, (voidp) &res3,
                                   (XDRPROC_T_TYPE) xdr_am_mountres3);
   fp->fh_status = unx_error(res3.fhs_status);
   memset(&fp->fh_nfs_handle.v3, 0, sizeof(am_nfs_fh3));
   fp->fh_nfs_handle.v3.am_fh3_length = res3.mountres3_u.mountinfo.fhandle.fhandle3_len;
   memmove(fp->fh_nfs_handle.v3.am_fh3_data,
           res3.mountres3_u.mountinfo.fhandle.fhandle3_val,
           fp->fh_nfs_handle.v3.am_fh3_length);

   XFREE(res3.mountres3_u.mountinfo.fhandle.fhandle3_val);
   if (res3.mountres3_u.mountinfo.auth_flavors.auth_flavors_val)
     XFREE(res3.mountres3_u.mountinfo.auth_flavors.auth_flavors_val);
 } else {
#endif /* HAVE_FS_NFS3 */
   memset(&res, 0, sizeof(res));
   fp->fh_error = pickup_rpc_reply(pkt, len, (voidp) &res,
                                   (XDRPROC_T_TYPE) xdr_fhstatus);
   fp->fh_status = unx_error(res.fhs_status);
   memmove(&fp->fh_nfs_handle.v2, &res.fhs_fh, NFS_FHSIZE);
#ifdef HAVE_FS_NFS3
 }
#endif /* HAVE_FS_NFS3 */

 if (!fp->fh_error) {
   dlog("got filehandle for %s:%s", fp->fh_fs->fs_host, fp->fh_path);
 } else {
   plog(XLOG_USER, "filehandle denied for %s:%s", fp->fh_fs->fs_host, fp->fh_path);
   /*
    * Force the error to be EACCES. It's debatable whether it should be
    * ENOENT instead, but the server really doesn't give us any clues, and
    * EACCES is more in line with the "filehandle denied" message.
    */
   fp->fh_error = EACCES;
 }

 /*
  * Wakeup anything sleeping on this filehandle
  */
 if (fp->fh_wchan) {
   dlog("Calling wakeup on %#lx", (unsigned long) fp->fh_wchan);
   wakeup(fp->fh_wchan);
 }
}


/*
* Called when a filehandle appears via WebNFS
*/
static void
got_nfs_fh_webnfs(voidp pkt, int len, struct sockaddr_in *sa, struct sockaddr_in *ia, opaque_t arg, int done)
{
 fh_cache *fp;
 nfsdiropres res;
#ifdef HAVE_FS_NFS3
 am_LOOKUP3res res3;
#endif /* HAVE_FS_NFS3 */

 fp = find_nfs_fhandle_cache(arg, done);
 if (!fp)
   return;

 /*
  * retrieve the correct RPC reply for the file handle, based on the
  * NFS protocol version.
  */
#ifdef HAVE_FS_NFS3
 if (fp->fh_nfs_version == NFS_VERSION3) {
   memset(&res3, 0, sizeof(res3));
   fp->fh_error = pickup_rpc_reply(pkt, len, (voidp) &res3,
                                   (XDRPROC_T_TYPE) xdr_am_LOOKUP3res);
   fp->fh_status = unx_error(res3.status);
   memset(&fp->fh_nfs_handle.v3, 0, sizeof(am_nfs_fh3));
   fp->fh_nfs_handle.v3.am_fh3_length = res3.res_u.ok.object.am_fh3_length;
   memmove(fp->fh_nfs_handle.v3.am_fh3_data,
           res3.res_u.ok.object.am_fh3_data,
           fp->fh_nfs_handle.v3.am_fh3_length);
 } else {
#endif /* HAVE_FS_NFS3 */
   memset(&res, 0, sizeof(res));
   fp->fh_error = pickup_rpc_reply(pkt, len, (voidp) &res,
                                   (XDRPROC_T_TYPE) xdr_diropres);
   fp->fh_status = unx_error(res.dr_status);
   memmove(&fp->fh_nfs_handle.v2, &res.dr_u.dr_drok_u.drok_fhandle, NFS_FHSIZE);
#ifdef HAVE_FS_NFS3
 }
#endif /* HAVE_FS_NFS3 */

 if (!fp->fh_error) {
   dlog("got filehandle for %s:%s", fp->fh_fs->fs_host, fp->fh_path);
 } else {
   plog(XLOG_USER, "filehandle denied for %s:%s", fp->fh_fs->fs_host, fp->fh_path);
   /*
    * Force the error to be EACCES. It's debatable whether it should be
    * ENOENT instead, but the server really doesn't give us any clues, and
    * EACCES is more in line with the "filehandle denied" message.
    */
   fp->fh_error = EACCES;
 }

 /*
  * Wakeup anything sleeping on this filehandle
  */
 if (fp->fh_wchan) {
   dlog("Calling wakeup on %#lx", (unsigned long) fp->fh_wchan);
   wakeup(fp->fh_wchan);
 }
}


void
flush_nfs_fhandle_cache(fserver *fs)
{
 fh_cache *fp;

 ITER(fp, fh_cache, &fh_head) {
   if (fp->fh_fs == fs || fs == NULL) {
     /*
      * Only invalidate port info for non-WebNFS servers
      */
     if (!(fp->fh_fs->fs_flags & FSF_WEBNFS))
       fp->fh_sin.sin_port = (u_short) 0;
     fp->fh_error = -1;
   }
 }
}


static void
discard_fh(opaque_t arg)
{
 fh_cache *fp = (fh_cache *) arg;

 rem_que(&fp->fh_q);
 if (fp->fh_fs) {
   dlog("Discarding filehandle for %s:%s", fp->fh_fs->fs_host, fp->fh_path);
   free_srvr(fp->fh_fs);
 }
 XFREE(fp->fh_path);
 XFREE(fp);
}


/*
* Determine the file handle for a node
*/
static int
prime_nfs_fhandle_cache(char *path, fserver *fs, am_nfs_handle_t *fhbuf, mntfs *mf)
{
 fh_cache *fp, *fp_save = NULL;
 int error;
 int reuse_id = FALSE;

 dlog("Searching cache for %s:%s", fs->fs_host, path);

 /*
  * First search the cache
  */
 ITER(fp, fh_cache, &fh_head) {
   if (fs != fp->fh_fs  ||  !STREQ(path, fp->fh_path))
     continue;                 /* skip to next ITER item */
   /* else we got a match */
   switch (fp->fh_error) {
   case 0:
     plog(XLOG_INFO, "prime_nfs_fhandle_cache: NFS version %d", (int) fp->fh_nfs_version);

     error = fp->fh_error = fp->fh_status;

     if (error == 0) {
       if (mf->mf_flags & MFF_NFS_SCALEDOWN) {
         fp_save = fp;
         /* XXX: why reuse the ID? */
         reuse_id = TRUE;
         break;
       }

       if (fhbuf) {
#ifdef HAVE_FS_NFS3
         if (fp->fh_nfs_version == NFS_VERSION3) {
           memmove((voidp) &(fhbuf->v3), (voidp) &(fp->fh_nfs_handle.v3),
                   sizeof(fp->fh_nfs_handle.v3));
         } else
#endif /* HAVE_FS_NFS3 */
           {
             memmove((voidp) &(fhbuf->v2), (voidp) &(fp->fh_nfs_handle.v2),
                     sizeof(fp->fh_nfs_handle.v2));
           }
       }
       if (fp->fh_cid)
         untimeout(fp->fh_cid);
       fp->fh_cid = timeout(FH_TTL, discard_fh, (opaque_t) fp);
     } else if (error == EACCES) {
       /*
        * Now decode the file handle return code.
        */
       plog(XLOG_INFO, "Filehandle denied for \"%s:%s\"",
            fs->fs_host, path);
     } else {
       errno = error;  /* XXX */
       plog(XLOG_INFO, "Filehandle error for \"%s:%s\": %m",
            fs->fs_host, path);
     }

     /*
      * The error was returned from the remote mount daemon.
      * Policy: this error will be cached for now...
      */
     return error;

   case -1:
     /*
      * Still thinking about it, but we can re-use.
      */
     fp_save = fp;
     reuse_id = TRUE;
     break;

   default:
     /*
      * Return the error.
      * Policy: make sure we recompute if required again
      * in case this was caused by a network failure.
      * This can thrash mountd's though...  If you find
      * your mountd going slowly then:
      * 1.  Add a fork() loop to main.
      * 2.  Remove the call to innetgr() and don't use
      *     netgroups, especially if you don't use YP.
      */
     error = fp->fh_error;
     fp->fh_error = -1;
     return error;
   }   /* end of switch statement */
 } /* end of ITER loop */

 /*
  * Not in cache
  */
 if (fp_save) {
   fp = fp_save;
   /*
    * Re-use existing slot
    */
   untimeout(fp->fh_cid);
   free_srvr(fp->fh_fs);
   XFREE(fp->fh_path);
 } else {
   fp = ALLOC(struct fh_cache);
   memset((voidp) fp, 0, sizeof(struct fh_cache));
   ins_que(&fp->fh_q, &fh_head);
 }
 if (!reuse_id)
   fp->fh_id = FHID_ALLOC();
 fp->fh_wchan = get_mntfs_wchan(mf);
 fp->fh_error = -1;
 fp->fh_cid = timeout(FH_TTL, discard_fh, (opaque_t) fp);

 /*
  * If fs->fs_ip is null, remote server is probably down.
  */
 if (!fs->fs_ip) {
   /* Mark the fileserver down and invalid again */
   fs->fs_flags &= ~FSF_VALID;
   fs->fs_flags |= FSF_DOWN;
   error = AM_ERRNO_HOST_DOWN;
   return error;
 }

 /*
  * Either fp has been freshly allocated or the address has changed.
  * Initialize address and nfs version.  Don't try to re-use the port
  * information unless using WebNFS where the port is fixed either by
  * the spec or the "port" mount option.
  */
 if (fp->fh_sin.sin_addr.s_addr != fs->fs_ip->sin_addr.s_addr) {
   fp->fh_sin = *fs->fs_ip;
   if (!(mf->mf_flags & MFF_WEBNFS))
       fp->fh_sin.sin_port = 0;
   fp->fh_nfs_version = SET_FH_VERSION(fs);
 }

 fp->fh_fs = dup_srvr(fs);
 fp->fh_path = xstrdup(path);

 if (mf->mf_flags & MFF_WEBNFS)
   error = webnfs_lookup(fp, got_nfs_fh_webnfs, get_mntfs_wchan(mf));
 else
   error = call_mountd(fp, MOUNTPROC_MNT, got_nfs_fh_mount, get_mntfs_wchan(mf));
 if (error) {
   /*
    * Local error - cache for a short period
    * just to prevent thrashing.
    */
   untimeout(fp->fh_cid);
   fp->fh_cid = timeout(error < 0 ? 2 * ALLOWED_MOUNT_TIME : FH_TTL_ERROR,
                        discard_fh, (opaque_t) fp);
   fp->fh_error = error;
 } else {
   error = fp->fh_error;
 }

 return error;
}


int
make_nfs_auth(void)
{
 AUTH_CREATE_GIDLIST_TYPE group_wheel = 0;

 /* Some NFS mounts (particularly cross-domain) require FQDNs to succeed */

#ifdef HAVE_TRANSPORT_TYPE_TLI
 if (gopt.flags & CFM_FULLY_QUALIFIED_HOSTS) {
   plog(XLOG_INFO, "Using NFS auth for FQHN \"%s\"", hostd);
   nfs_auth = authsys_create(hostd, 0, 0, 1, &group_wheel);
 } else {
   nfs_auth = authsys_create_default();
 }
#else /* not HAVE_TRANSPORT_TYPE_TLI */
 if (gopt.flags & CFM_FULLY_QUALIFIED_HOSTS) {
   plog(XLOG_INFO, "Using NFS auth for FQHN \"%s\"", hostd);
   nfs_auth = authunix_create(hostd, 0, 0, 1, &group_wheel);
 } else {
   nfs_auth = authunix_create_default();
 }
#endif /* not HAVE_TRANSPORT_TYPE_TLI */

 if (!nfs_auth)
   return ENOBUFS;

 return 0;
}


static int
call_mountd(fh_cache *fp, u_long proc, fwd_fun fun, wchan_t wchan)
{
 struct rpc_msg mnt_msg;
 int len;
 char iobuf[UDPMSGSIZE];
 int error;
 u_long mnt_version;

 if (!nfs_auth) {
   error = make_nfs_auth();
   if (error)
     return error;
 }

 if (fp->fh_sin.sin_port == 0) {
   u_short mountd_port;
   error = get_mountd_port(fp->fh_fs, &mountd_port, wchan);
   if (error)
     return error;
   fp->fh_sin.sin_port = mountd_port;
   dlog("%s: New %d mountd port", __func__, fp->fh_sin.sin_port);
 } else
   dlog("%s: Already had %d mountd port", __func__, fp->fh_sin.sin_port);

 /* find the right version of the mount protocol */
#ifdef HAVE_FS_NFS3
 if (fp->fh_nfs_version == NFS_VERSION3)
   mnt_version = AM_MOUNTVERS3;
 else
#endif /* HAVE_FS_NFS3 */
   mnt_version = MOUNTVERS;
 plog(XLOG_INFO, "call_mountd: NFS version %d, mount version %d",
      (int) fp->fh_nfs_version, (int) mnt_version);

 rpc_msg_init(&mnt_msg, MOUNTPROG, mnt_version, MOUNTPROC_NULL);
 len = make_rpc_packet(iobuf,
                       sizeof(iobuf),
                       proc,
                       &mnt_msg,
                       (voidp) &fp->fh_path,
                       (XDRPROC_T_TYPE) xdr_nfspath,
                       nfs_auth);

 if (len > 0) {
   error = fwd_packet(MK_RPC_XID(RPC_XID_MOUNTD, fp->fh_id),
                      iobuf,
                      len,
                      &fp->fh_sin,
                      &fp->fh_sin,
                      (opaque_t) ((long) fp->fh_id), /* cast to long needed for 64-bit archs */
                      fun);
 } else {
   error = -len;
 }

 /*
  * It may be the case that we're sending to the wrong MOUNTD port.  This
  * occurs if mountd is restarted on the server after the port has been
  * looked up and stored in the filehandle cache somewhere.  The correct
  * solution, if we're going to cache port numbers is to catch the ICMP
  * port unreachable reply from the server and cause the portmap request
  * to be redone.  The quick solution here is to invalidate the MOUNTD
  * port.
  */
 fp->fh_sin.sin_port = 0;

 return error;
}


static int
webnfs_lookup(fh_cache *fp, fwd_fun fun, wchan_t wchan)
{
 struct rpc_msg wnfs_msg;
 int len;
 char iobuf[UDPMSGSIZE];
 int error;
 u_long proc;
 XDRPROC_T_TYPE xdr_fn;
 voidp argp;
 nfsdiropargs args;
#ifdef HAVE_FS_NFS3
 am_LOOKUP3args args3;
#endif /* HAVE_FS_NFS3 */
 char *wnfs_path;
 size_t l;

 if (!nfs_auth) {
   error = make_nfs_auth();
   if (error)
     return error;
 }

 if (fp->fh_sin.sin_port == 0) {
   /* FIXME: wrong, don't discard sin_port in the first place for WebNFS. */
   plog(XLOG_WARNING, "webnfs_lookup: port == 0 for nfs on %s, fixed",
        fp->fh_fs->fs_host);
   fp->fh_sin.sin_port = htons(NFS_PORT);
 }

 /*
  * Use native path like the rest of amd (cf. RFC 2054, 6.1).
  */
 l = strlen(fp->fh_path) + 2;
 wnfs_path = (char *) xmalloc(l);
 wnfs_path[0] = 0x80;
 xstrlcpy(wnfs_path + 1, fp->fh_path, l - 1);

 /* find the right program and lookup procedure */
#ifdef HAVE_FS_NFS3
 if (fp->fh_nfs_version == NFS_VERSION3) {
   proc = AM_NFSPROC3_LOOKUP;
   xdr_fn = (XDRPROC_T_TYPE) xdr_am_LOOKUP3args;
   argp = &args3;
   /* WebNFS public file handle */
   args3.what.dir.am_fh3_length = 0;
   args3.what.name = wnfs_path;
 } else {
#endif /* HAVE_FS_NFS3 */
   proc = NFSPROC_LOOKUP;
   xdr_fn = (XDRPROC_T_TYPE) xdr_diropargs;
   argp = &args;
   /* WebNFS public file handle */
   memset(&args.da_fhandle, 0, NFS_FHSIZE);
   args.da_name = wnfs_path;
#ifdef HAVE_FS_NFS3
 }
#endif /* HAVE_FS_NFS3 */

 plog(XLOG_INFO, "webnfs_lookup: NFS version %d", (int) fp->fh_nfs_version);

 rpc_msg_init(&wnfs_msg, NFS_PROGRAM, fp->fh_nfs_version, proc);
 len = make_rpc_packet(iobuf,
                       sizeof(iobuf),
                       proc,
                       &wnfs_msg,
                       argp,
                       (XDRPROC_T_TYPE) xdr_fn,
                       nfs_auth);

 if (len > 0) {
   error = fwd_packet(MK_RPC_XID(RPC_XID_WEBNFS, fp->fh_id),
                      iobuf,
                      len,
                      &fp->fh_sin,
                      &fp->fh_sin,
                      (opaque_t) ((long) fp->fh_id), /* cast to long needed for 64-bit archs */
                      fun);
 } else {
   error = -len;
 }

 XFREE(wnfs_path);
 return error;
}


/*
* NFS needs the local filesystem, remote filesystem
* remote hostname.
* Local filesystem defaults to remote and vice-versa.
*/
static char *
nfs_match(am_opts *fo)
{
 char *xmtab;
 size_t l;

 if (fo->opt_fs && !fo->opt_rfs)
   fo->opt_rfs = fo->opt_fs;
 if (!fo->opt_rfs) {
   plog(XLOG_USER, "nfs: no remote filesystem specified");
   return NULL;
 }
 if (!fo->opt_rhost) {
   plog(XLOG_USER, "nfs: no remote host specified");
   return NULL;
 }

 /*
  * Determine magic cookie to put in mtab
  */
 l = strlen(fo->opt_rhost) + strlen(fo->opt_rfs) + 2;
 xmtab = (char *) xmalloc(l);
 xsnprintf(xmtab, l, "%s:%s", fo->opt_rhost, fo->opt_rfs);
 dlog("NFS: mounting remote server \"%s\", remote fs \"%s\" on \"%s\"",
      fo->opt_rhost, fo->opt_rfs, fo->opt_fs);

 return xmtab;
}


/*
* Initialize am structure for nfs
*/
static int
nfs_init(mntfs *mf)
{
 int error;
 am_nfs_handle_t fhs;
 char *colon;

#ifdef NO_FALLBACK
 /*
  * We don't need file handles for NFS version 4, but we can fall back to
  * version 3, so we allocate anyway
  */
#ifdef HAVE_FS_NFS4
 if (mf->mf_server->fs_version == NFS_VERSION4)
   return 0;
#endif /* HAVE_FS_NFS4 */
#endif /* NO_FALLBACK */

 if (mf->mf_private) {
   if (mf->mf_flags & MFF_NFS_SCALEDOWN) {
     fserver *fs;

     /* tell remote mountd that we're done with this filehandle */
     mf->mf_ops->umounted(mf);

     mf->mf_prfree(mf->mf_private);
     mf->mf_private = NULL;
     mf->mf_prfree = NULL;

     fs = mf->mf_ops->ffserver(mf);
     free_srvr(mf->mf_server);
     mf->mf_server = fs;
   } else
     return 0;
 }

 colon = strchr(mf->mf_info, ':');
 if (colon == 0)
   return ENOENT;

 error = prime_nfs_fhandle_cache(colon + 1, mf->mf_server, &fhs, mf);
 if (!error) {
   mf->mf_private = (opaque_t) ALLOC(am_nfs_handle_t);
   mf->mf_prfree = (void (*)(opaque_t)) free;
   memmove(mf->mf_private, (voidp) &fhs, sizeof(fhs));
 }
 return error;
}


int
mount_nfs_fh(am_nfs_handle_t *fhp, char *mntdir, char *fs_name, mntfs *mf)
{
 MTYPE_TYPE type;
 char *colon;
 char *xopts=NULL, transp_timeo_opts[40], transp_retrans_opts[40];
 char host[MAXHOSTNAMELEN + MAXPATHLEN + 2];
 fserver *fs = mf->mf_server;
 u_long nfs_version = fs->fs_version;
 char *nfs_proto = fs->fs_proto; /* "tcp" or "udp" */
 int on_autofs = mf->mf_flags & MFF_ON_AUTOFS;
 int error;
 int genflags;
 int retry;
 int proto = AMU_TYPE_NONE;
 mntent_t mnt;
 void *argsp;
 nfs_args_t nfs_args;
#ifdef HAVE_FS_NFS4
 nfs4_args_t nfs4_args;
#endif /* HAVE_FS_NFS4 */

 /*
  * Extract HOST name to give to kernel.
  * Some systems like osf1/aix3/bsd44 variants may need old code
  * for NFS_ARGS_NEEDS_PATH.
  */
 if (!(colon = strchr(fs_name, ':')))
   return ENOENT;
#ifdef MOUNT_TABLE_ON_FILE
 *colon = '\0';
#endif /* MOUNT_TABLE_ON_FILE */
 xstrlcpy(host, fs_name, sizeof(host));
#ifdef MOUNT_TABLE_ON_FILE
 *colon = ':';
#endif /* MOUNT_TABLE_ON_FILE */
#ifdef MAXHOSTNAMELEN
 /* most kernels have a name length restriction */
 if (strlen(host) >= MAXHOSTNAMELEN)
   xstrlcpy(host + MAXHOSTNAMELEN - 3, "..",
            sizeof(host) - MAXHOSTNAMELEN + 3);
#endif /* MAXHOSTNAMELEN */

 /*
  * Create option=VAL for udp/tcp specific timeouts and retrans values, but
  * only if these options were specified.
  */

 transp_timeo_opts[0] = transp_retrans_opts[0] = '\0'; /* initialize */
 if (STREQ(nfs_proto, "udp"))
   proto = AMU_TYPE_UDP;
 else if (STREQ(nfs_proto, "tcp"))
   proto = AMU_TYPE_TCP;
 if (proto != AMU_TYPE_NONE) {
   if (gopt.amfs_auto_timeo[proto] > 0)
     xsnprintf(transp_timeo_opts, sizeof(transp_timeo_opts), "%s=%d,",
               MNTTAB_OPT_TIMEO, gopt.amfs_auto_timeo[proto]);
   if (gopt.amfs_auto_retrans[proto] > 0)
     xsnprintf(transp_retrans_opts, sizeof(transp_retrans_opts), "%s=%d,",
               MNTTAB_OPT_RETRANS, gopt.amfs_auto_retrans[proto]);
 }

 if (mf->mf_remopts && *mf->mf_remopts &&
     !islocalnet(fs->fs_ip->sin_addr.s_addr)) {
   plog(XLOG_INFO, "Using remopts=\"%s\"", mf->mf_remopts);
   /* use transp_opts first, so map-specific opts will override */
   xopts = str3cat(xopts, transp_timeo_opts, transp_retrans_opts, mf->mf_remopts);
 } else {
   /* use transp_opts first, so map-specific opts will override */
   xopts = str3cat(xopts, transp_timeo_opts, transp_retrans_opts, mf->mf_mopts);
 }

 memset((voidp) &mnt, 0, sizeof(mnt));
 mnt.mnt_dir = mntdir;
 mnt.mnt_fsname = fs_name;
 mnt.mnt_opts = xopts;

 /*
  * Set mount types accordingly
  */
#ifdef HAVE_FS_NFS3
 if (nfs_version == NFS_VERSION3) {
   type = MOUNT_TYPE_NFS3;
   /*
    * Systems that include the mount table "vers" option generally do not
    * set the mnttab entry to "nfs3", but to "nfs" and then they set
    * "vers=3".  Setting it to "nfs3" works, but it may break some things
    * like "df -t nfs" and the "quota" program (esp. on Solaris and Irix).
    * So on those systems, set it to "nfs".
    * Note: MNTTAB_OPT_VERS is always set for NFS3 (see am_compat.h).
    */
   argsp = &nfs_args;
# if defined(MNTTAB_OPT_VERS) && defined(MOUNT_TABLE_ON_FILE)
   mnt.mnt_type = MNTTAB_TYPE_NFS;
# else /* defined(MNTTAB_OPT_VERS) && defined(MOUNT_TABLE_ON_FILE) */
   mnt.mnt_type = MNTTAB_TYPE_NFS3;
# endif /* defined(MNTTAB_OPT_VERS) && defined(MOUNT_TABLE_ON_FILE) */
# ifdef HAVE_FS_NFS4
 } else if (nfs_version == NFS_VERSION4) {
   argsp = &nfs4_args;
   type = MOUNT_TYPE_NFS4;
   mnt.mnt_type = MNTTAB_TYPE_NFS4;
# endif /* HAVE_FS_NFS4 */
 } else
#endif /* HAVE_FS_NFS3 */
 {
   argsp = &nfs_args;
   type = MOUNT_TYPE_NFS;
   mnt.mnt_type = MNTTAB_TYPE_NFS;
 }
 plog(XLOG_INFO, "mount_nfs_fh: NFS version %d", (int) nfs_version);
 plog(XLOG_INFO, "mount_nfs_fh: using NFS transport %s", nfs_proto);

 retry = hasmntval(&mnt, MNTTAB_OPT_RETRY);
 if (retry <= 0)
   retry = 1;                  /* XXX */

 genflags = compute_mount_flags(&mnt);
#ifdef HAVE_FS_AUTOFS
 if (on_autofs)
   genflags |= autofs_compute_mount_flags(&mnt);
#endif /* HAVE_FS_AUTOFS */

  /* setup the many fields and flags within nfs_args */
  compute_nfs_args(argsp,
                   &mnt,
                   genflags,
                   NULL,       /* struct netconfig *nfsncp */
                   fs->fs_ip,
                   nfs_version,
                   nfs_proto,
                   fhp,
                   host,
                   fs_name);

 /* finally call the mounting function */
 if (amuDebug(D_TRACE)) {
   print_nfs_args(argsp, nfs_version);
   plog(XLOG_DEBUG, "Generic mount flags 0x%x used for NFS mount", genflags);
 }
 error = mount_fs(&mnt, genflags, argsp, retry, type,
                  nfs_version, nfs_proto, mnttab_file_name, on_autofs);
 XFREE(mnt.mnt_opts);
 discard_nfs_args(argsp, nfs_version);

#ifdef HAVE_FS_NFS4
# ifndef NO_FALLBACK
 /*
  * If we are using a v4 file handle, we try a v3 if we get back:
  *    ENOENT: NFS v4 has a different export list than v3
  *    EPERM: Kernels <= 2.6.18 return that, instead of ENOENT
  */
 if ((error == ENOENT || error == EPERM) && nfs_version == NFS_VERSION4) {
   plog(XLOG_DEBUG, "Could not find NFS 4 mount, trying again with NFS 3");
   fs->fs_version = NFS_VERSION3;
   error = mount_nfs_fh(fhp, mntdir, fs_name, mf);
   if (error)
     fs->fs_version = NFS_VERSION4;
 }
# endif /* NO_FALLBACK */
#endif /* HAVE_FS_NFS4 */

 return error;
}


static int
nfs_mount(am_node *am, mntfs *mf)
{
 int error = 0;
 mntent_t mnt;

 if (!mf->mf_private && mf->mf_server->fs_version != 4) {
   plog(XLOG_ERROR, "Missing filehandle for %s", mf->mf_info);
   return EINVAL;
 }

 if (mf->mf_mopts == NULL) {
   plog(XLOG_ERROR, "Missing mount options for %s", mf->mf_info);
   return EINVAL;
 }

 mnt.mnt_opts = mf->mf_mopts;
 if (amu_hasmntopt(&mnt, "softlookup") ||
     (amu_hasmntopt(&mnt, "soft") && !amu_hasmntopt(&mnt, "nosoftlookup")))
   am->am_flags |= AMF_SOFTLOOKUP;

 error = mount_nfs_fh((am_nfs_handle_t *) mf->mf_private,
                      mf->mf_mount,
                      mf->mf_info,
                      mf);

 if (error) {
   errno = error;
   dlog("mount_nfs: %m");
 }

 return error;
}


static int
nfs_umount(am_node *am, mntfs *mf)
{
 int unmount_flags, new_unmount_flags, error;

 dlog("attempting nfs umount");
 unmount_flags = (mf->mf_flags & MFF_ON_AUTOFS) ? AMU_UMOUNT_AUTOFS : 0;
 error = UMOUNT_FS(mf->mf_mount, mnttab_file_name, unmount_flags);

#if defined(HAVE_UMOUNT2) && (defined(MNT2_GEN_OPT_FORCE) || defined(MNT2_GEN_OPT_DETACH))
 /*
  * If the attempt to unmount failed with EBUSY, and this fserver was
  * marked for forced unmounts, then use forced/lazy unmounts.
  */
 if (error == EBUSY &&
     gopt.flags & CFM_FORCED_UNMOUNTS &&
     mf->mf_server->fs_flags & FSF_FORCE_UNMOUNT) {
   plog(XLOG_INFO, "EZK: nfs_umount: trying forced/lazy unmounts");
   /*
    * XXX: turning off the FSF_FORCE_UNMOUNT may not be perfectly
    * incorrect.  Multiple nodes may need to be timed out and restarted for
    * a single hung fserver.
    */
   mf->mf_server->fs_flags &= ~FSF_FORCE_UNMOUNT;
   new_unmount_flags = unmount_flags | AMU_UMOUNT_FORCE | AMU_UMOUNT_DETACH;
   error = UMOUNT_FS(mf->mf_mount, mnttab_file_name, new_unmount_flags);
 }
#endif /* HAVE_UMOUNT2 && (MNT2_GEN_OPT_FORCE || MNT2_GEN_OPT_DETACH) */

 /*
  * Here is some code to unmount 'restarted' file systems.
  * The restarted file systems are marked as 'nfs', not
  * 'host', so we only have the map information for the
  * the top-level mount.  The unmount will fail (EBUSY)
  * if there are anything else from the NFS server mounted
  * below the mount-point.  This code checks to see if there
  * is anything mounted with the same prefix as the
  * file system to be unmounted ("/a/b/c" when unmounting "/a/b").
  * If there is, and it is a 'restarted' file system, we unmount
  * it.
  * Added by Mike Mitchell, [email protected], 09/08/93
  */
 if (error == EBUSY) {
   mntfs *new_mf;
   int len = strlen(mf->mf_mount);
   int didsome = 0;

   ITER(new_mf, mntfs, &mfhead) {
     if (new_mf->mf_ops != mf->mf_ops ||
         new_mf->mf_refc > 1 ||
         mf == new_mf ||
         ((new_mf->mf_flags & (MFF_MOUNTED | MFF_UNMOUNTING | MFF_RESTART)) == (MFF_MOUNTED | MFF_RESTART)))
       continue;

     if (NSTREQ(mf->mf_mount, new_mf->mf_mount, len) &&
         new_mf->mf_mount[len] == '/') {
       new_unmount_flags =
         (new_mf->mf_flags & MFF_ON_AUTOFS) ? AMU_UMOUNT_AUTOFS : 0;
       UMOUNT_FS(new_mf->mf_mount, mnttab_file_name, new_unmount_flags);
       didsome = 1;
     }
   }
   if (didsome)
     error = UMOUNT_FS(mf->mf_mount, mnttab_file_name, unmount_flags);
 }
 if (error)
   return error;

 return 0;
}


static void
nfs_umounted(mntfs *mf)
{
 fserver *fs;
 char *colon, *path;

 if (mf->mf_error || mf->mf_refc > 1)
   return;

 /*
  * No need to inform mountd when WebNFS is in use.
  */
 if (mf->mf_flags & MFF_WEBNFS)
   return;

 /*
  * Call the mount daemon on the server to announce that we are not using
  * the fs any more.
  *
  * XXX: This is *wrong*.  The mountd should be called when the fhandle is
  * flushed from the cache, and a reference held to the cached entry while
  * the fs is mounted...
  */
 fs = mf->mf_server;
 colon = path = strchr(mf->mf_info, ':');
 if (fs && colon) {
   fh_cache f;

   dlog("calling mountd for %s", mf->mf_info);
   *path++ = '\0';
   f.fh_path = path;
   f.fh_sin = *fs->fs_ip;
   f.fh_sin.sin_port = (u_short) 0;
   f.fh_nfs_version = SET_FH_VERSION(fs);
   f.fh_fs = fs;
   f.fh_id = 0;
   f.fh_error = 0;
   prime_nfs_fhandle_cache(colon + 1, mf->mf_server, (am_nfs_handle_t *) NULL, mf);
   call_mountd(&f, MOUNTPROC_UMNT, (fwd_fun *) NULL, (wchan_t) NULL);
   *colon = ':';
 }
}