freebsd-dev/sbin/mount_nfs/mount_nfs.c

1097 lines
28 KiB
C
Raw Normal View History

/*-
* SPDX-License-Identifier: BSD-3-Clause
*
* Copyright (c) 1992, 1993, 1994
* The Regents of the University of California. All rights reserved.
*
* This code is derived from software contributed to Berkeley by
* Rick Macklem at The University of Guelph.
*
* 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.
*/
#if 0
#ifndef lint
static const char copyright[] =
"@(#) Copyright (c) 1992, 1993, 1994\n\
The Regents of the University of California. All rights reserved.\n";
#endif /* not lint */
#ifndef lint
static char sccsid[] = "@(#)mount_nfs.c 8.11 (Berkeley) 5/4/95";
#endif /* not lint */
#endif
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/linker.h>
#include <sys/module.h>
#include <sys/mount.h>
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
#include <sys/socket.h>
#include <sys/stat.h>
#include <sys/syslog.h>
#include <sys/uio.h>
#include <rpc/rpc.h>
#include <rpc/pmap_clnt.h>
#include <rpc/pmap_prot.h>
#include <rpcsvc/nfs_prot.h>
#include <rpcsvc/mount.h>
#include <fs/nfs/nfsproto.h>
#include <fs/nfs/nfsv4_errstr.h>
#include <arpa/inet.h>
#include <net/route.h>
#include <net/if.h>
#include <ctype.h>
#include <err.h>
#include <errno.h>
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
#include <fcntl.h>
#include <netdb.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <strings.h>
#include <sysexits.h>
#include <unistd.h>
#include "mntopts.h"
#include "mounttab.h"
/* Table for af,sotype -> netid conversions. */
static struct nc_protos {
const char *netid;
int af;
int sotype;
} nc_protos[] = {
{"udp", AF_INET, SOCK_DGRAM},
{"tcp", AF_INET, SOCK_STREAM},
{"udp6", AF_INET6, SOCK_DGRAM},
{"tcp6", AF_INET6, SOCK_STREAM},
{NULL, 0, 0}
};
struct nfhret {
u_long stat;
long vers;
long auth;
long fhsize;
u_char nfh[NFS3_FHSIZE];
};
#define BGRND 1
#define ISBGRND 2
#define OF_NOINET4 4
#define OF_NOINET6 8
static int retrycnt = -1;
static int opflags = 0;
static int nfsproto = IPPROTO_TCP;
static int mnttcp_ok = 1;
static int noconn = 0;
/* The 'portspec' is the server nfs port; NULL means look up via rpcbind. */
static const char *portspec = NULL;
static struct sockaddr *addr;
static int addrlen = 0;
static u_char *fh = NULL;
static int fhsize = 0;
static int secflavor = -1;
static int got_principal = 0;
static enum mountmode {
ANY,
V2,
V3,
V4
} mountmode = ANY;
/* Return codes for nfs_tryproto. */
enum tryret {
TRYRET_SUCCESS,
TRYRET_TIMEOUT, /* No response received. */
TRYRET_REMOTEERR, /* Error received from remote server. */
TRYRET_LOCALERR /* Local failure. */
};
static int sec_name_to_num(const char *sec);
static const char *sec_num_to_name(int num);
static int getnfsargs(char *, struct iovec **iov, int *iovlen);
2002-03-21 13:14:21 +00:00
/* void set_rpc_maxgrouplist(int); */
static struct netconfig *getnetconf_cached(const char *netid);
static const char *netidbytype(int af, int sotype);
static void usage(void) __dead2;
static int xdr_dir(XDR *, char *);
static int xdr_fh(XDR *, struct nfhret *);
static enum tryret nfs_tryproto(struct addrinfo *ai, char *hostp, char *spec,
char **errstr, struct iovec **iov, int *iovlen);
static enum tryret returncode(enum clnt_stat stat, struct rpc_err *rpcerr);
int
main(int argc, char *argv[])
{
int c;
struct iovec *iov;
int num, iovlen;
char *mntname, *p, *spec, *tmp;
char mntpath[MAXPATHLEN], errmsg[255];
char hostname[MAXHOSTNAMELEN + 1], gssn[MAXHOSTNAMELEN + 50];
const char *gssname, *nmount_errstr;
iov = NULL;
iovlen = 0;
memset(errmsg, 0, sizeof(errmsg));
gssname = NULL;
while ((c = getopt(argc, argv,
"23a:bcdD:g:I:iLlNo:PR:r:sTt:w:x:U")) != -1)
switch (c) {
case '2':
mountmode = V2;
break;
case '3':
mountmode = V3;
break;
case 'a':
printf("-a deprecated, use -o readahead=<value>\n");
build_iovec(&iov, &iovlen, "readahead", optarg, (size_t)-1);
break;
case 'b':
opflags |= BGRND;
break;
case 'c':
printf("-c deprecated, use -o noconn\n");
build_iovec(&iov, &iovlen, "noconn", NULL, 0);
noconn = 1;
break;
case 'D':
printf("-D deprecated, use -o deadthresh=<value>\n");
build_iovec(&iov, &iovlen, "deadthresh", optarg, (size_t)-1);
break;
case 'd':
printf("-d deprecated, use -o dumbtimer");
build_iovec(&iov, &iovlen, "dumbtimer", NULL, 0);
break;
case 'g':
printf("-g deprecated, use -o maxgroups");
num = strtol(optarg, &p, 10);
if (*p || num <= 0)
errx(1, "illegal -g value -- %s", optarg);
//set_rpc_maxgrouplist(num);
build_iovec(&iov, &iovlen, "maxgroups", optarg, (size_t)-1);
break;
case 'I':
printf("-I deprecated, use -o readdirsize=<value>\n");
build_iovec(&iov, &iovlen, "readdirsize", optarg, (size_t)-1);
break;
case 'i':
printf("-i deprecated, use -o intr\n");
build_iovec(&iov, &iovlen, "intr", NULL, 0);
break;
case 'L':
printf("-L deprecated, use -o nolockd\n");
build_iovec(&iov, &iovlen, "nolockd", NULL, 0);
break;
case 'l':
printf("-l deprecated, -o rdirplus\n");
build_iovec(&iov, &iovlen, "rdirplus", NULL, 0);
break;
case 'N':
printf("-N deprecated, do not specify -o resvport\n");
break;
case 'o': {
int pass_flag_to_nmount;
char *opt = optarg;
while (opt) {
char *pval = NULL;
char *pnextopt = NULL;
const char *val = "";
pass_flag_to_nmount = 1;
pnextopt = strchr(opt, ',');
if (pnextopt != NULL) {
*pnextopt = '\0';
pnextopt++;
}
pval = strchr(opt, '=');
if (pval != NULL) {
*pval = '\0';
val = pval + 1;
}
if (strcmp(opt, "bg") == 0) {
opflags |= BGRND;
pass_flag_to_nmount=0;
} else if (strcmp(opt, "fg") == 0) {
/* same as not specifying -o bg */
pass_flag_to_nmount=0;
} else if (strcmp(opt, "gssname") == 0) {
pass_flag_to_nmount = 0;
gssname = val;
} else if (strcmp(opt, "mntudp") == 0) {
mnttcp_ok = 0;
nfsproto = IPPROTO_UDP;
} else if (strcmp(opt, "udp") == 0) {
nfsproto = IPPROTO_UDP;
} else if (strcmp(opt, "tcp") == 0) {
nfsproto = IPPROTO_TCP;
} else if (strcmp(opt, "noinet4") == 0) {
pass_flag_to_nmount=0;
opflags |= OF_NOINET4;
} else if (strcmp(opt, "noinet6") == 0) {
pass_flag_to_nmount=0;
opflags |= OF_NOINET6;
} else if (strcmp(opt, "noconn") == 0) {
noconn = 1;
} else if (strcmp(opt, "nfsv2") == 0) {
pass_flag_to_nmount=0;
mountmode = V2;
} else if (strcmp(opt, "nfsv3") == 0) {
mountmode = V3;
} else if (strcmp(opt, "nfsv4") == 0) {
pass_flag_to_nmount=0;
mountmode = V4;
nfsproto = IPPROTO_TCP;
if (portspec == NULL)
portspec = "2049";
} else if (strcmp(opt, "port") == 0) {
pass_flag_to_nmount=0;
asprintf(&tmp, "%d", atoi(val));
if (tmp == NULL)
err(1, "asprintf");
portspec = tmp;
} else if (strcmp(opt, "principal") == 0) {
got_principal = 1;
} else if (strcmp(opt, "proto") == 0) {
pass_flag_to_nmount=0;
if (strcmp(val, "tcp") == 0) {
nfsproto = IPPROTO_TCP;
opflags |= OF_NOINET6;
build_iovec(&iov, &iovlen,
"tcp", NULL, 0);
} else if (strcmp(val, "udp") == 0) {
mnttcp_ok = 0;
nfsproto = IPPROTO_UDP;
opflags |= OF_NOINET6;
build_iovec(&iov, &iovlen,
"udp", NULL, 0);
} else if (strcmp(val, "tcp6") == 0) {
nfsproto = IPPROTO_TCP;
opflags |= OF_NOINET4;
build_iovec(&iov, &iovlen,
"tcp", NULL, 0);
} else if (strcmp(val, "udp6") == 0) {
mnttcp_ok = 0;
nfsproto = IPPROTO_UDP;
opflags |= OF_NOINET4;
build_iovec(&iov, &iovlen,
"udp", NULL, 0);
} else {
errx(1,
"illegal proto value -- %s",
val);
}
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
} else if (strcmp(opt, "sec") == 0) {
/*
* Don't add this option to
* the iovec yet - we will
* negotiate which sec flavor
* to use with the remote
* mountd.
*/
pass_flag_to_nmount=0;
secflavor = sec_name_to_num(val);
if (secflavor < 0) {
errx(1,
"illegal sec value -- %s",
val);
}
} else if (strcmp(opt, "retrycnt") == 0) {
pass_flag_to_nmount=0;
num = strtol(val, &p, 10);
if (*p || num < 0)
errx(1, "illegal retrycnt value -- %s", val);
retrycnt = num;
} else if (strcmp(opt, "maxgroups") == 0) {
num = strtol(val, &p, 10);
if (*p || num <= 0)
errx(1, "illegal maxgroups value -- %s", val);
//set_rpc_maxgrouplist(num);
} else if (strcmp(opt, "vers") == 0) {
num = strtol(val, &p, 10);
if (*p || num <= 0)
errx(1, "illegal vers value -- "
"%s", val);
switch (num) {
case 2:
mountmode = V2;
break;
case 3:
mountmode = V3;
build_iovec(&iov, &iovlen,
"nfsv3", NULL, 0);
break;
case 4:
mountmode = V4;
nfsproto = IPPROTO_TCP;
if (portspec == NULL)
portspec = "2049";
break;
default:
errx(1, "illegal nfs version "
"value -- %s", val);
}
pass_flag_to_nmount=0;
}
if (pass_flag_to_nmount) {
build_iovec(&iov, &iovlen, opt,
__DECONST(void *, val),
strlen(val) + 1);
}
opt = pnextopt;
}
}
break;
case 'P':
/* obsolete for -o noresvport now default */
printf("-P deprecated, use -o noresvport\n");
build_iovec(&iov, &iovlen, "noresvport", NULL, 0);
break;
case 'R':
printf("-R deprecated, use -o retrycnt=<retrycnt>\n");
num = strtol(optarg, &p, 10);
if (*p || num < 0)
errx(1, "illegal -R value -- %s", optarg);
retrycnt = num;
break;
case 'r':
printf("-r deprecated, use -o rsize=<rsize>\n");
build_iovec(&iov, &iovlen, "rsize", optarg, (size_t)-1);
break;
case 's':
printf("-s deprecated, use -o soft\n");
build_iovec(&iov, &iovlen, "soft", NULL, 0);
break;
case 'T':
nfsproto = IPPROTO_TCP;
printf("-T deprecated, use -o tcp\n");
break;
case 't':
printf("-t deprecated, use -o timeout=<value>\n");
build_iovec(&iov, &iovlen, "timeout", optarg, (size_t)-1);
break;
case 'w':
printf("-w deprecated, use -o wsize=<value>\n");
build_iovec(&iov, &iovlen, "wsize", optarg, (size_t)-1);
break;
case 'x':
printf("-x deprecated, use -o retrans=<value>\n");
build_iovec(&iov, &iovlen, "retrans", optarg, (size_t)-1);
break;
case 'U':
printf("-U deprecated, use -o mntudp\n");
mnttcp_ok = 0;
nfsproto = IPPROTO_UDP;
build_iovec(&iov, &iovlen, "mntudp", NULL, 0);
break;
default:
usage();
break;
}
argc -= optind;
argv += optind;
if (argc != 2) {
usage();
/* NOTREACHED */
}
spec = *argv++;
mntname = *argv;
if (retrycnt == -1)
/* The default is to keep retrying forever. */
retrycnt = 0;
if (modfind("nfscl") < 0) {
/* Not present in kernel, try loading it */
if (kldload("nfscl") < 0 ||
modfind("nfscl") < 0)
errx(1, "nfscl is not available");
}
/*
* Add the fqdn to the gssname, as required.
*/
if (gssname != NULL) {
if (strchr(gssname, '@') == NULL &&
gethostname(hostname, MAXHOSTNAMELEN) == 0) {
snprintf(gssn, sizeof (gssn), "%s@%s", gssname,
hostname);
gssname = gssn;
}
build_iovec(&iov, &iovlen, "gssname",
__DECONST(void *, gssname), strlen(gssname) + 1);
}
if (!getnfsargs(spec, &iov, &iovlen))
exit(1);
/* resolve the mountpoint with realpath(3) */
if (checkpath(mntname, mntpath) != 0)
err(1, "%s", mntpath);
build_iovec_argf(&iov, &iovlen, "fstype", "nfs");
build_iovec(&iov, &iovlen, "fspath", mntpath, (size_t)-1);
build_iovec(&iov, &iovlen, "errmsg", errmsg, sizeof(errmsg));
if (nmount(iov, iovlen, 0)) {
nmount_errstr = nfsv4_geterrstr(errno);
if (mountmode == V4 && nmount_errstr != NULL)
errx(1, "nmount: %s, %s", mntpath, nmount_errstr);
else
err(1, "nmount: %s%s%s", mntpath, errmsg[0] ? ", " : "",
errmsg);
}
exit(0);
}
static int
sec_name_to_num(const char *sec)
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
{
if (!strcmp(sec, "krb5"))
return (RPCSEC_GSS_KRB5);
if (!strcmp(sec, "krb5i"))
return (RPCSEC_GSS_KRB5I);
if (!strcmp(sec, "krb5p"))
return (RPCSEC_GSS_KRB5P);
if (!strcmp(sec, "sys"))
return (AUTH_SYS);
return (-1);
}
static const char *
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
sec_num_to_name(int flavor)
{
switch (flavor) {
case RPCSEC_GSS_KRB5:
return ("krb5");
case RPCSEC_GSS_KRB5I:
return ("krb5i");
case RPCSEC_GSS_KRB5P:
return ("krb5p");
case AUTH_SYS:
return ("sys");
}
return (NULL);
}
/*
* Wait for RTM_IFINFO message with interface that is IFF_UP and with
* link on, or until timeout expires. Returns seconds left.
*/
static time_t
rtm_ifinfo_sleep(time_t sec)
{
char buf[2048] __aligned(__alignof(struct if_msghdr));
fd_set rfds;
struct timeval tv, start;
ssize_t nread;
int n, s;
s = socket(PF_ROUTE, SOCK_RAW, 0);
if (s < 0)
err(EX_OSERR, "socket");
(void)gettimeofday(&start, NULL);
for (tv.tv_sec = sec, tv.tv_usec = 0;
tv.tv_sec > 0;
(void)gettimeofday(&tv, NULL),
tv.tv_sec = sec - (tv.tv_sec - start.tv_sec)) {
FD_ZERO(&rfds);
FD_SET(s, &rfds);
n = select(s + 1, &rfds, NULL, NULL, &tv);
if (n == 0)
continue;
if (n == -1) {
if (errno == EINTR)
continue;
else
err(EX_SOFTWARE, "select");
}
nread = read(s, buf, 2048);
if (nread < 0)
err(EX_OSERR, "read");
if ((size_t)nread >= sizeof(struct if_msghdr)) {
struct if_msghdr *ifm;
ifm = (struct if_msghdr *)buf;
if (ifm->ifm_version == RTM_VERSION &&
ifm->ifm_type == RTM_IFINFO &&
(ifm->ifm_flags & IFF_UP) &&
ifm->ifm_data.ifi_link_state != LINK_STATE_DOWN)
break;
}
}
close(s);
return (tv.tv_sec);
}
static int
getnfsargs(char *spec, struct iovec **iov, int *iovlen)
{
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
struct addrinfo hints, *ai_nfs, *ai;
enum tryret ret;
int ecode, speclen, remoteerr, offset, have_bracket = 0;
char *hostp, *delimp, *errstr;
size_t len;
static char nam[MNAMELEN + 1], pname[MAXHOSTNAMELEN + 5];
if (*spec == '[' && (delimp = strchr(spec + 1, ']')) != NULL &&
*(delimp + 1) == ':') {
hostp = spec + 1;
spec = delimp + 2;
have_bracket = 1;
} else if ((delimp = strrchr(spec, ':')) != NULL) {
hostp = spec;
spec = delimp + 1;
} else if ((delimp = strrchr(spec, '@')) != NULL) {
warnx("path@server syntax is deprecated, use server:path");
hostp = delimp + 1;
} else {
warnx("no <host>:<dirpath> nfs-name");
return (0);
}
*delimp = '\0';
/*
* If there has been a trailing slash at mounttime it seems
* that some mountd implementations fail to remove the mount
* entries from their mountlist while unmounting.
*/
for (speclen = strlen(spec);
speclen > 1 && spec[speclen - 1] == '/';
speclen--)
spec[speclen - 1] = '\0';
if (strlen(hostp) + strlen(spec) + 1 > MNAMELEN) {
warnx("%s:%s: %s", hostp, spec, strerror(ENAMETOOLONG));
return (0);
}
/* Make both '@' and ':' notations equal */
if (*hostp != '\0') {
len = strlen(hostp);
offset = 0;
if (have_bracket)
nam[offset++] = '[';
memmove(nam + offset, hostp, len);
if (have_bracket)
nam[len + offset++] = ']';
nam[len + offset++] = ':';
memmove(nam + len + offset, spec, speclen);
nam[len + speclen + offset] = '\0';
}
/*
* Handle an internet host address.
*/
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
memset(&hints, 0, sizeof hints);
hints.ai_flags = AI_NUMERICHOST;
if (nfsproto == IPPROTO_TCP)
hints.ai_socktype = SOCK_STREAM;
else if (nfsproto == IPPROTO_UDP)
hints.ai_socktype = SOCK_DGRAM;
if (getaddrinfo(hostp, portspec, &hints, &ai_nfs) != 0) {
hints.ai_flags = AI_CANONNAME;
if ((ecode = getaddrinfo(hostp, portspec, &hints, &ai_nfs))
!= 0) {
if (portspec == NULL)
errx(1, "%s: %s", hostp, gai_strerror(ecode));
else
errx(1, "%s:%s: %s", hostp, portspec,
gai_strerror(ecode));
return (0);
}
/*
* For a Kerberized nfs mount where the "principal"
* argument has not been set, add it here.
*/
if (got_principal == 0 && secflavor != AUTH_SYS &&
ai_nfs->ai_canonname != NULL) {
snprintf(pname, sizeof (pname), "nfs@%s",
ai_nfs->ai_canonname);
build_iovec(iov, iovlen, "principal", pname,
strlen(pname) + 1);
}
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
}
ret = TRYRET_LOCALERR;
for (;;) {
/*
* Try each entry returned by getaddrinfo(). Note the
2012-01-07 16:09:33 +00:00
* occurrence of remote errors by setting `remoteerr'.
*/
remoteerr = 0;
for (ai = ai_nfs; ai != NULL; ai = ai->ai_next) {
if ((ai->ai_family == AF_INET6) &&
(opflags & OF_NOINET6))
continue;
if ((ai->ai_family == AF_INET) &&
(opflags & OF_NOINET4))
continue;
ret = nfs_tryproto(ai, hostp, spec, &errstr, iov,
iovlen);
if (ret == TRYRET_SUCCESS)
break;
if (ret != TRYRET_LOCALERR)
remoteerr = 1;
if ((opflags & ISBGRND) == 0)
fprintf(stderr, "%s\n", errstr);
}
if (ret == TRYRET_SUCCESS)
break;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
/* Exit if all errors were local. */
if (!remoteerr)
exit(1);
/*
* If retrycnt == 0, we are to keep retrying forever.
* Otherwise decrement it, and exit if it hits zero.
*/
if (retrycnt != 0 && --retrycnt == 0)
exit(1);
if ((opflags & (BGRND | ISBGRND)) == BGRND) {
warnx("Cannot immediately mount %s:%s, backgrounding",
hostp, spec);
opflags |= ISBGRND;
if (daemon(0, 0) != 0)
err(1, "daemon");
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
}
/*
* If rtm_ifinfo_sleep() returns non-zero, don't count
* that as a retry attempt.
*/
if (rtm_ifinfo_sleep(60) && retrycnt != 0)
retrycnt++;
}
freeaddrinfo(ai_nfs);
build_iovec(iov, iovlen, "hostname", nam, (size_t)-1);
/* Add mounted file system to PATH_MOUNTTAB */
if (mountmode != V4 && !add_mtab(hostp, spec))
warnx("can't update %s for %s:%s", PATH_MOUNTTAB, hostp, spec);
return (1);
}
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
/*
* Try to set up the NFS arguments according to the address
* family, protocol (and possibly port) specified in `ai'.
*
* Returns TRYRET_SUCCESS if successful, or:
* TRYRET_TIMEOUT The server did not respond.
* TRYRET_REMOTEERR The server reported an error.
* TRYRET_LOCALERR Local failure.
*
* In all error cases, *errstr will be set to a statically-allocated string
* describing the error.
*/
static enum tryret
nfs_tryproto(struct addrinfo *ai, char *hostp, char *spec, char **errstr,
struct iovec **iov, int *iovlen)
{
static char errbuf[256];
struct sockaddr_storage nfs_ss;
struct netbuf nfs_nb;
struct nfhret nfhret;
struct timeval try;
struct rpc_err rpcerr;
CLIENT *clp;
struct netconfig *nconf, *nconf_mnt;
const char *netid, *netid_mnt, *secname;
int doconnect, nfsvers, mntvers, sotype;
enum clnt_stat clntstat;
enum mountmode trymntmode;
Bring in a hybrid of SunSoft's transport-independent RPC (TI-RPC) and associated changes that had to happen to make this possible as well as bugs fixed along the way. Bring in required TLI library routines to support this. Since we don't support TLI we've essentially copied what NetBSD has done, adding a thin layer to emulate direct the TLI calls into BSD socket calls. This is mostly from Sun's tirpc release that was made in 1994, however some fixes were backported from the 1999 release (supposedly only made available after this porting effort was underway). The submitter has agreed to continue on and bring us up to the 1999 release. Several key features are introduced with this update: Client calls are thread safe. (1999 code has server side thread safe) Updated, a more modern interface. Many userland updates were done to bring the code up to par with the recent RPC API. There is an update to the pthreads library, a function pthread_main_np() was added to emulate a function of Sun's threads library. While we're at it, bring in NetBSD's lockd, it's been far too long of a wait. New rpcbind(8) replaces portmap(8) (supporting communication over an authenticated Unix-domain socket, and by default only allowing set and unset requests over that channel). It's much more secure than the old portmapper. Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded to support TI-RPC and to support IPV6. Umount(8) is also fixed to unmount pathnames longer than 80 chars, which are currently truncated by the Kernel statfs structure. Submitted by: Martin Blapp <mb@imp.ch> Manpage review: ru Secure RPC implemented by: wpaul
2001-03-19 12:50:13 +00:00
sotype = 0;
trymntmode = mountmode;
errbuf[0] = '\0';
*errstr = errbuf;
if (nfsproto == IPPROTO_TCP)
sotype = SOCK_STREAM;
else if (nfsproto == IPPROTO_UDP)
sotype = SOCK_DGRAM;
if ((netid = netidbytype(ai->ai_family, sotype)) == NULL) {
snprintf(errbuf, sizeof errbuf,
"af %d sotype %d not supported", ai->ai_family, sotype);
return (TRYRET_LOCALERR);
}
if ((nconf = getnetconf_cached(netid)) == NULL) {
snprintf(errbuf, sizeof errbuf, "%s: %s", netid, nc_sperror());
return (TRYRET_LOCALERR);
}
/* The RPCPROG_MNT netid may be different. */
if (mnttcp_ok) {
netid_mnt = netid;
nconf_mnt = nconf;
} else {
if ((netid_mnt = netidbytype(ai->ai_family, SOCK_DGRAM))
== NULL) {
snprintf(errbuf, sizeof errbuf,
"af %d sotype SOCK_DGRAM not supported",
ai->ai_family);
return (TRYRET_LOCALERR);
}
if ((nconf_mnt = getnetconf_cached(netid_mnt)) == NULL) {
snprintf(errbuf, sizeof errbuf, "%s: %s", netid_mnt,
nc_sperror());
return (TRYRET_LOCALERR);
}
}
tryagain:
if (trymntmode == V4) {
nfsvers = 4;
mntvers = 3; /* Workaround for GCC. */
} else if (trymntmode == V2) {
nfsvers = 2;
mntvers = 1;
} else {
nfsvers = 3;
mntvers = 3;
}
if (portspec != NULL) {
/* `ai' contains the complete nfsd sockaddr. */
nfs_nb.buf = ai->ai_addr;
nfs_nb.len = nfs_nb.maxlen = ai->ai_addrlen;
} else {
/* Ask the remote rpcbind. */
nfs_nb.buf = &nfs_ss;
nfs_nb.len = nfs_nb.maxlen = sizeof nfs_ss;
if (!rpcb_getaddr(NFS_PROGRAM, nfsvers, nconf, &nfs_nb,
hostp)) {
if (rpc_createerr.cf_stat == RPC_PROGVERSMISMATCH &&
trymntmode == ANY) {
trymntmode = V2;
goto tryagain;
}
snprintf(errbuf, sizeof errbuf, "[%s] %s:%s: %s",
netid, hostp, spec,
clnt_spcreateerror("RPCPROG_NFS"));
return (returncode(rpc_createerr.cf_stat,
&rpc_createerr.cf_error));
}
}
/* Check that the server (nfsd) responds on the port we have chosen. */
clp = clnt_tli_create(RPC_ANYFD, nconf, &nfs_nb, NFS_PROGRAM, nfsvers,
0, 0);
if (clp == NULL) {
snprintf(errbuf, sizeof errbuf, "[%s] %s:%s: %s", netid,
hostp, spec, clnt_spcreateerror("nfsd: RPCPROG_NFS"));
return (returncode(rpc_createerr.cf_stat,
&rpc_createerr.cf_error));
}
if (sotype == SOCK_DGRAM && noconn == 0) {
/*
* Use connect(), to match what the kernel does. This
* catches cases where the server responds from the
* wrong source address.
*/
doconnect = 1;
if (!clnt_control(clp, CLSET_CONNECT, (char *)&doconnect)) {
clnt_destroy(clp);
snprintf(errbuf, sizeof errbuf,
"[%s] %s:%s: CLSET_CONNECT failed", netid, hostp,
spec);
return (TRYRET_LOCALERR);
}
}
try.tv_sec = 10;
try.tv_usec = 0;
clntstat = clnt_call(clp, NFSPROC_NULL, (xdrproc_t)xdr_void, NULL,
(xdrproc_t)xdr_void, NULL, try);
if (clntstat != RPC_SUCCESS) {
if (clntstat == RPC_PROGVERSMISMATCH && trymntmode == ANY) {
clnt_destroy(clp);
trymntmode = V2;
goto tryagain;
}
clnt_geterr(clp, &rpcerr);
snprintf(errbuf, sizeof errbuf, "[%s] %s:%s: %s", netid,
hostp, spec, clnt_sperror(clp, "NFSPROC_NULL"));
clnt_destroy(clp);
return (returncode(clntstat, &rpcerr));
}
clnt_destroy(clp);
/*
* For NFSv4, there is no mount protocol.
*/
if (trymntmode == V4) {
/*
* Store the server address in nfsargsp, making
* sure to copy any locally allocated structures.
*/
addrlen = nfs_nb.len;
addr = malloc(addrlen);
if (addr == NULL)
err(1, "malloc");
bcopy(nfs_nb.buf, addr, addrlen);
build_iovec(iov, iovlen, "addr", addr, addrlen);
secname = sec_num_to_name(secflavor);
if (secname != NULL) {
build_iovec(iov, iovlen, "sec",
__DECONST(void *, secname), (size_t)-1);
}
build_iovec(iov, iovlen, "nfsv4", NULL, 0);
build_iovec(iov, iovlen, "dirpath", spec, (size_t)-1);
return (TRYRET_SUCCESS);
}
/* Send the MOUNTPROC_MNT RPC to get the root filehandle. */
try.tv_sec = 10;
try.tv_usec = 0;
clp = clnt_tp_create(hostp, MOUNTPROG, mntvers, nconf_mnt);
if (clp == NULL) {
snprintf(errbuf, sizeof errbuf, "[%s] %s:%s: %s", netid_mnt,
hostp, spec, clnt_spcreateerror("RPCMNT: clnt_create"));
return (returncode(rpc_createerr.cf_stat,
&rpc_createerr.cf_error));
}
clp->cl_auth = authsys_create_default();
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
nfhret.auth = secflavor;
nfhret.vers = mntvers;
clntstat = clnt_call(clp, MOUNTPROC_MNT, (xdrproc_t)xdr_dir, spec,
(xdrproc_t)xdr_fh, &nfhret,
try);
auth_destroy(clp->cl_auth);
if (clntstat != RPC_SUCCESS) {
if (clntstat == RPC_PROGVERSMISMATCH && trymntmode == ANY) {
clnt_destroy(clp);
trymntmode = V2;
goto tryagain;
}
clnt_geterr(clp, &rpcerr);
snprintf(errbuf, sizeof errbuf, "[%s] %s:%s: %s", netid_mnt,
hostp, spec, clnt_sperror(clp, "RPCPROG_MNT"));
clnt_destroy(clp);
return (returncode(clntstat, &rpcerr));
}
clnt_destroy(clp);
if (nfhret.stat != 0) {
snprintf(errbuf, sizeof errbuf, "[%s] %s:%s: %s", netid_mnt,
hostp, spec, strerror(nfhret.stat));
return (TRYRET_REMOTEERR);
}
/*
* Store the filehandle and server address in nfsargsp, making
* sure to copy any locally allocated structures.
*/
addrlen = nfs_nb.len;
addr = malloc(addrlen);
fhsize = nfhret.fhsize;
fh = malloc(fhsize);
if (addr == NULL || fh == NULL)
err(1, "malloc");
bcopy(nfs_nb.buf, addr, addrlen);
bcopy(nfhret.nfh, fh, fhsize);
build_iovec(iov, iovlen, "addr", addr, addrlen);
build_iovec(iov, iovlen, "fh", fh, fhsize);
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
secname = sec_num_to_name(nfhret.auth);
if (secname) {
build_iovec(iov, iovlen, "sec",
__DECONST(void *, secname), (size_t)-1);
}
if (nfsvers == 3)
build_iovec(iov, iovlen, "nfsv3", NULL, 0);
return (TRYRET_SUCCESS);
}
/*
* Catagorise a RPC return status and error into an `enum tryret'
* return code.
*/
static enum tryret
returncode(enum clnt_stat clntstat, struct rpc_err *rpcerr)
{
switch (clntstat) {
case RPC_TIMEDOUT:
return (TRYRET_TIMEOUT);
case RPC_PMAPFAILURE:
case RPC_PROGNOTREGISTERED:
case RPC_PROGVERSMISMATCH:
/* XXX, these can be local or remote. */
case RPC_CANTSEND:
case RPC_CANTRECV:
return (TRYRET_REMOTEERR);
case RPC_SYSTEMERROR:
switch (rpcerr->re_errno) {
case ETIMEDOUT:
return (TRYRET_TIMEOUT);
case ENOMEM:
break;
default:
return (TRYRET_REMOTEERR);
}
/* FALLTHROUGH */
default:
break;
}
return (TRYRET_LOCALERR);
}
/*
* Look up a netid based on an address family and socket type.
* `af' is the address family, and `sotype' is SOCK_DGRAM or SOCK_STREAM.
*
* XXX there should be a library function for this.
*/
static const char *
netidbytype(int af, int sotype)
{
struct nc_protos *p;
for (p = nc_protos; p->netid != NULL; p++) {
if (af != p->af || sotype != p->sotype)
continue;
return (p->netid);
}
return (NULL);
}
/*
* Look up a netconfig entry based on a netid, and cache the result so
* that we don't need to remember to call freenetconfigent().
*
* Otherwise it behaves just like getnetconfigent(), so nc_*error()
* work on failure.
*/
static struct netconfig *
getnetconf_cached(const char *netid)
{
static struct nc_entry {
struct netconfig *nconf;
struct nc_entry *next;
} *head;
struct nc_entry *p;
struct netconfig *nconf;
for (p = head; p != NULL; p = p->next)
if (strcmp(netid, p->nconf->nc_netid) == 0)
return (p->nconf);
if ((nconf = getnetconfigent(netid)) == NULL)
return (NULL);
if ((p = malloc(sizeof(*p))) == NULL)
err(1, "malloc");
p->nconf = nconf;
p->next = head;
head = p;
return (p->nconf);
}
/*
* xdr routines for mount rpc's
*/
static int
xdr_dir(XDR *xdrsp, char *dirp)
{
return (xdr_string(xdrsp, &dirp, MNTPATHLEN));
}
static int
xdr_fh(XDR *xdrsp, struct nfhret *np)
{
int i;
long auth, authcnt, authfnd = 0;
if (!xdr_u_long(xdrsp, &np->stat))
return (0);
if (np->stat)
return (1);
switch (np->vers) {
case 1:
np->fhsize = NFS_FHSIZE;
return (xdr_opaque(xdrsp, (caddr_t)np->nfh, NFS_FHSIZE));
case 3:
if (!xdr_long(xdrsp, &np->fhsize))
return (0);
if (np->fhsize <= 0 || np->fhsize > NFS3_FHSIZE)
return (0);
if (!xdr_opaque(xdrsp, (caddr_t)np->nfh, np->fhsize))
return (0);
if (!xdr_long(xdrsp, &authcnt))
return (0);
for (i = 0; i < authcnt; i++) {
if (!xdr_long(xdrsp, &auth))
return (0);
if (np->auth == -1) {
np->auth = auth;
authfnd++;
} else if (auth == np->auth) {
authfnd++;
}
}
/*
* Some servers, such as DEC's OSF/1 return a nil authenticator
* list to indicate RPCAUTH_UNIX.
*/
if (authcnt == 0 && np->auth == -1)
np->auth = AUTH_SYS;
if (!authfnd && (authcnt > 0 || np->auth != AUTH_SYS))
np->stat = EAUTH;
return (1);
}
return (0);
}
static void
usage(void)
{
(void)fprintf(stderr, "%s\n%s\n%s\n%s\n",
"usage: mount_nfs [-23bcdiLlNPsTU] [-a maxreadahead] [-D deadthresh]",
" [-g maxgroups] [-I readdirsize] [-o options] [-R retrycnt]",
" [-r readsize] [-t timeout] [-w writesize] [-x retrans]",
" rhost:path node");
exit(1);
}