freebsd-nq/usr.sbin/gssd/gssd.c

1028 lines
28 KiB
C
Raw Normal View History

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
/*-
* Copyright (c) 2008 Isilon Inc http://www.isilon.com/
* Authors: Doug Rabson <dfr@rabson.org>
* Developed with Red Inc: Alfred Perlstein <alfred@freebsd.org>
*
* 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.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR 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 AUTHOR 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.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/stat.h>
#include <sys/linker.h>
#include <sys/module.h>
#include <sys/queue.h>
#include <sys/syslog.h>
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
#include <ctype.h>
#include <dirent.h>
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
#include <err.h>
#include <errno.h>
#ifndef WITHOUT_KERBEROS
#include <krb5.h>
#endif
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
#include <pwd.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <gssapi/gssapi.h>
#include <rpc/rpc.h>
#include <rpc/rpc_com.h>
#include "gssd.h"
#ifndef _PATH_GSS_MECH
#define _PATH_GSS_MECH "/etc/gss/mech"
#endif
#ifndef _PATH_GSSDSOCK
#define _PATH_GSSDSOCK "/var/run/gssd.sock"
#endif
struct gss_resource {
LIST_ENTRY(gss_resource) gr_link;
uint64_t gr_id; /* indentifier exported to kernel */
void* gr_res; /* GSS-API resource pointer */
};
LIST_HEAD(gss_resource_list, gss_resource) gss_resources;
int gss_resource_count;
uint32_t gss_next_id;
uint32_t gss_start_time;
int debug_level;
static char ccfile_dirlist[PATH_MAX + 1], ccfile_substring[NAME_MAX + 1];
static char pref_realm[1024];
static int verbose;
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
static void gssd_load_mech(void);
static int find_ccache_file(const char *, uid_t, char *);
static int is_a_valid_tgt_cache(const char *, uid_t, int *, time_t *);
static void gssd_verbose_out(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
extern void gssd_1(struct svc_req *rqstp, SVCXPRT *transp);
extern int gssd_syscall(char *path);
int
main(int argc, char **argv)
{
/*
* We provide an RPC service on a local-domain socket. The
* kernel's GSS-API code will pass what it can't handle
* directly to us.
*/
struct sockaddr_un sun;
int fd, oldmask, ch, debug;
SVCXPRT *xprt;
/*
* Initialize the credential cache file name substring and the
* search directory list.
*/
strlcpy(ccfile_substring, "krb5cc_", sizeof(ccfile_substring));
ccfile_dirlist[0] = '\0';
pref_realm[0] = '\0';
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
debug = 0;
verbose = 0;
while ((ch = getopt(argc, argv, "dvs:c:r:")) != -1) {
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
switch (ch) {
case 'd':
debug_level++;
break;
case 'v':
verbose = 1;
break;
case 's':
#ifndef WITHOUT_KERBEROS
/*
* Set the directory search list. This enables use of
* find_ccache_file() to search the directories for a
* suitable credentials cache file.
*/
strlcpy(ccfile_dirlist, optarg, sizeof(ccfile_dirlist));
#else
errx(1, "This option not available when built"
" without MK_KERBEROS\n");
#endif
break;
case 'c':
/*
* Specify a non-default credential cache file
* substring.
*/
strlcpy(ccfile_substring, optarg,
sizeof(ccfile_substring));
break;
case 'r':
/*
* Set the preferred realm for the credential cache tgt.
*/
strlcpy(pref_realm, optarg, sizeof(pref_realm));
break;
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
default:
fprintf(stderr,
"usage: %s [-d] [-s dir-list] [-c file-substring]"
" [-r preferred-realm]\n", argv[0]);
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
exit(1);
break;
}
}
gssd_load_mech();
if (!debug_level)
daemon(0, 0);
memset(&sun, 0, sizeof sun);
sun.sun_family = AF_LOCAL;
unlink(_PATH_GSSDSOCK);
strcpy(sun.sun_path, _PATH_GSSDSOCK);
sun.sun_len = SUN_LEN(&sun);
fd = socket(AF_LOCAL, SOCK_STREAM, 0);
if (!fd) {
if (debug_level == 0) {
syslog(LOG_ERR, "Can't create local gssd socket");
exit(1);
}
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
err(1, "Can't create local gssd socket");
}
oldmask = umask(S_IXUSR|S_IRWXG|S_IRWXO);
if (bind(fd, (struct sockaddr *) &sun, sun.sun_len) < 0) {
if (debug_level == 0) {
syslog(LOG_ERR, "Can't bind local gssd socket");
exit(1);
}
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
err(1, "Can't bind local gssd socket");
}
umask(oldmask);
if (listen(fd, SOMAXCONN) < 0) {
if (debug_level == 0) {
syslog(LOG_ERR, "Can't listen on local gssd socket");
exit(1);
}
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
err(1, "Can't listen on local gssd socket");
}
xprt = svc_vc_create(fd, RPC_MAXDATASIZE, RPC_MAXDATASIZE);
if (!xprt) {
if (debug_level == 0) {
syslog(LOG_ERR,
"Can't create transport for local gssd socket");
exit(1);
}
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
err(1, "Can't create transport for local gssd socket");
}
if (!svc_reg(xprt, GSSD, GSSDVERS, gssd_1, NULL)) {
if (debug_level == 0) {
syslog(LOG_ERR,
"Can't register service for local gssd socket");
exit(1);
}
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
err(1, "Can't register service for local gssd socket");
}
LIST_INIT(&gss_resources);
gss_next_id = 1;
gss_start_time = time(0);
gssd_syscall(_PATH_GSSDSOCK);
svc_run();
return (0);
}
static void
gssd_load_mech(void)
{
FILE *fp;
char buf[256];
char *p;
char *name, *oid, *lib, *kobj;
fp = fopen(_PATH_GSS_MECH, "r");
if (!fp)
return;
while (fgets(buf, sizeof(buf), fp)) {
if (*buf == '#')
continue;
p = buf;
name = strsep(&p, "\t\n ");
if (p) while (isspace(*p)) p++;
oid = strsep(&p, "\t\n ");
if (p) while (isspace(*p)) p++;
lib = strsep(&p, "\t\n ");
if (p) while (isspace(*p)) p++;
kobj = strsep(&p, "\t\n ");
if (!name || !oid || !lib || !kobj)
continue;
if (strcmp(kobj, "-")) {
/*
* Attempt to load the kernel module if its
* not already present.
*/
if (modfind(kobj) < 0) {
if (kldload(kobj) < 0) {
fprintf(stderr,
"%s: can't find or load kernel module %s for %s\n",
getprogname(), kobj, name);
}
}
}
}
fclose(fp);
}
static void *
gssd_find_resource(uint64_t id)
{
struct gss_resource *gr;
if (!id)
return (NULL);
LIST_FOREACH(gr, &gss_resources, gr_link)
if (gr->gr_id == id)
return (gr->gr_res);
return (NULL);
}
static uint64_t
gssd_make_resource(void *res)
{
struct gss_resource *gr;
if (!res)
return (0);
gr = malloc(sizeof(struct gss_resource));
if (!gr)
return (0);
gr->gr_id = (gss_next_id++) + ((uint64_t) gss_start_time << 32);
gr->gr_res = res;
LIST_INSERT_HEAD(&gss_resources, gr, gr_link);
gss_resource_count++;
if (debug_level > 1)
printf("%d resources allocated\n", gss_resource_count);
return (gr->gr_id);
}
static void
gssd_delete_resource(uint64_t id)
{
struct gss_resource *gr;
LIST_FOREACH(gr, &gss_resources, gr_link) {
if (gr->gr_id == id) {
LIST_REMOVE(gr, gr_link);
free(gr);
gss_resource_count--;
if (debug_level > 1)
printf("%d resources allocated\n",
gss_resource_count);
return;
}
}
}
static void
gssd_verbose_out(const char *fmt, ...)
{
va_list ap;
if (verbose != 0) {
va_start(ap, fmt);
if (debug_level == 0)
vsyslog(LOG_INFO | LOG_DAEMON, fmt, ap);
else
vfprintf(stderr, fmt, ap);
va_end(ap);
}
}
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
bool_t
gssd_null_1_svc(void *argp, void *result, struct svc_req *rqstp)
{
gssd_verbose_out("gssd_null: done\n");
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
return (TRUE);
}
bool_t
gssd_init_sec_context_1_svc(init_sec_context_args *argp, init_sec_context_res *result, struct svc_req *rqstp)
{
gss_cred_id_t cred = GSS_C_NO_CREDENTIAL;
gss_ctx_id_t ctx = GSS_C_NO_CONTEXT;
gss_name_t name = GSS_C_NO_NAME;
char ccname[PATH_MAX + 5 + 1], *cp, *cp2;
int gotone;
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
memset(result, 0, sizeof(*result));
if (ccfile_dirlist[0] != '\0' && argp->cred == 0) {
/*
* For the "-s" case and no credentials provided as an
* argument, search the directory list for an appropriate
* credential cache file. If the search fails, return failure.
*/
gotone = 0;
cp = ccfile_dirlist;
do {
cp2 = strchr(cp, ':');
if (cp2 != NULL)
*cp2 = '\0';
gotone = find_ccache_file(cp, argp->uid, ccname);
if (gotone != 0)
break;
if (cp2 != NULL)
*cp2++ = ':';
cp = cp2;
} while (cp != NULL && *cp != '\0');
if (gotone == 0) {
result->major_status = GSS_S_CREDENTIALS_EXPIRED;
gssd_verbose_out("gssd_init_sec_context: -s no"
" credential cache file found for uid=%d\n",
(int)argp->uid);
return (TRUE);
}
} else {
/*
* If there wasn't a "-s" option or the credentials have
* been provided as an argument, do it the old way.
* When credentials are provided, the uid should be root.
*/
if (argp->cred != 0 && argp->uid != 0) {
if (debug_level == 0)
syslog(LOG_ERR, "gss_init_sec_context:"
" cred for non-root");
else
fprintf(stderr, "gss_init_sec_context:"
" cred for non-root\n");
}
snprintf(ccname, sizeof(ccname), "FILE:/tmp/krb5cc_%d",
(int) argp->uid);
}
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
setenv("KRB5CCNAME", ccname, TRUE);
if (argp->cred) {
cred = gssd_find_resource(argp->cred);
if (!cred) {
result->major_status = GSS_S_CREDENTIALS_EXPIRED;
gssd_verbose_out("gssd_init_sec_context: cred"
" resource not found\n");
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
return (TRUE);
}
}
if (argp->ctx) {
ctx = gssd_find_resource(argp->ctx);
if (!ctx) {
result->major_status = GSS_S_CONTEXT_EXPIRED;
gssd_verbose_out("gssd_init_sec_context: context"
" resource not found\n");
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
return (TRUE);
}
}
if (argp->name) {
name = gssd_find_resource(argp->name);
if (!name) {
result->major_status = GSS_S_BAD_NAME;
gssd_verbose_out("gssd_init_sec_context: name"
" resource not found\n");
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
return (TRUE);
}
}
result->major_status = gss_init_sec_context(&result->minor_status,
cred, &ctx, name, argp->mech_type,
argp->req_flags, argp->time_req, argp->input_chan_bindings,
&argp->input_token, &result->actual_mech_type,
&result->output_token, &result->ret_flags, &result->time_rec);
gssd_verbose_out("gssd_init_sec_context: done major=0x%x minor=%d"
" uid=%d\n", (unsigned int)result->major_status,
(int)result->minor_status, (int)argp->uid);
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 (result->major_status == GSS_S_COMPLETE
|| result->major_status == GSS_S_CONTINUE_NEEDED) {
if (argp->ctx)
result->ctx = argp->ctx;
else
result->ctx = gssd_make_resource(ctx);
}
return (TRUE);
}
bool_t
gssd_accept_sec_context_1_svc(accept_sec_context_args *argp, accept_sec_context_res *result, struct svc_req *rqstp)
{
gss_ctx_id_t ctx = GSS_C_NO_CONTEXT;
gss_cred_id_t cred = GSS_C_NO_CREDENTIAL;
gss_name_t src_name;
gss_cred_id_t delegated_cred_handle;
memset(result, 0, sizeof(*result));
if (argp->ctx) {
ctx = gssd_find_resource(argp->ctx);
if (!ctx) {
result->major_status = GSS_S_CONTEXT_EXPIRED;
gssd_verbose_out("gssd_accept_sec_context: ctx"
" resource not found\n");
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
return (TRUE);
}
}
if (argp->cred) {
cred = gssd_find_resource(argp->cred);
if (!cred) {
result->major_status = GSS_S_CREDENTIALS_EXPIRED;
gssd_verbose_out("gssd_accept_sec_context: cred"
" resource not found\n");
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
return (TRUE);
}
}
memset(result, 0, sizeof(*result));
result->major_status = gss_accept_sec_context(&result->minor_status,
&ctx, cred, &argp->input_token, argp->input_chan_bindings,
&src_name, &result->mech_type, &result->output_token,
&result->ret_flags, &result->time_rec,
&delegated_cred_handle);
gssd_verbose_out("gssd_accept_sec_context: done major=0x%x minor=%d\n",
(unsigned int)result->major_status, (int)result->minor_status);
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 (result->major_status == GSS_S_COMPLETE
|| result->major_status == GSS_S_CONTINUE_NEEDED) {
if (argp->ctx)
result->ctx = argp->ctx;
else
result->ctx = gssd_make_resource(ctx);
result->src_name = gssd_make_resource(src_name);
result->delegated_cred_handle =
gssd_make_resource(delegated_cred_handle);
}
return (TRUE);
}
bool_t
gssd_delete_sec_context_1_svc(delete_sec_context_args *argp, delete_sec_context_res *result, struct svc_req *rqstp)
{
gss_ctx_id_t ctx = gssd_find_resource(argp->ctx);
if (ctx) {
result->major_status = gss_delete_sec_context(
&result->minor_status, &ctx, &result->output_token);
gssd_delete_resource(argp->ctx);
} else {
result->major_status = GSS_S_COMPLETE;
result->minor_status = 0;
}
gssd_verbose_out("gssd_delete_sec_context: done major=0x%x minor=%d\n",
(unsigned int)result->major_status, (int)result->minor_status);
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
return (TRUE);
}
bool_t
gssd_export_sec_context_1_svc(export_sec_context_args *argp, export_sec_context_res *result, struct svc_req *rqstp)
{
gss_ctx_id_t ctx = gssd_find_resource(argp->ctx);
if (ctx) {
result->major_status = gss_export_sec_context(
&result->minor_status, &ctx,
&result->interprocess_token);
result->format = KGSS_HEIMDAL_1_1;
gssd_delete_resource(argp->ctx);
} else {
result->major_status = GSS_S_FAILURE;
result->minor_status = 0;
result->interprocess_token.length = 0;
result->interprocess_token.value = NULL;
}
gssd_verbose_out("gssd_export_sec_context: done major=0x%x minor=%d\n",
(unsigned int)result->major_status, (int)result->minor_status);
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
return (TRUE);
}
bool_t
gssd_import_name_1_svc(import_name_args *argp, import_name_res *result, struct svc_req *rqstp)
{
gss_name_t name;
result->major_status = gss_import_name(&result->minor_status,
&argp->input_name_buffer, argp->input_name_type, &name);
gssd_verbose_out("gssd_import_name: done major=0x%x minor=%d\n",
(unsigned int)result->major_status, (int)result->minor_status);
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 (result->major_status == GSS_S_COMPLETE)
result->output_name = gssd_make_resource(name);
else
result->output_name = 0;
return (TRUE);
}
bool_t
gssd_canonicalize_name_1_svc(canonicalize_name_args *argp, canonicalize_name_res *result, struct svc_req *rqstp)
{
gss_name_t name = gssd_find_resource(argp->input_name);
gss_name_t output_name;
memset(result, 0, sizeof(*result));
if (!name) {
result->major_status = GSS_S_BAD_NAME;
return (TRUE);
}
result->major_status = gss_canonicalize_name(&result->minor_status,
name, argp->mech_type, &output_name);
gssd_verbose_out("gssd_canonicalize_name: done major=0x%x minor=%d\n",
(unsigned int)result->major_status, (int)result->minor_status);
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 (result->major_status == GSS_S_COMPLETE)
result->output_name = gssd_make_resource(output_name);
else
result->output_name = 0;
return (TRUE);
}
bool_t
gssd_export_name_1_svc(export_name_args *argp, export_name_res *result, struct svc_req *rqstp)
{
gss_name_t name = gssd_find_resource(argp->input_name);
memset(result, 0, sizeof(*result));
if (!name) {
result->major_status = GSS_S_BAD_NAME;
gssd_verbose_out("gssd_export_name: name resource not found\n");
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
return (TRUE);
}
result->major_status = gss_export_name(&result->minor_status,
name, &result->exported_name);
gssd_verbose_out("gssd_export_name: done major=0x%x minor=%d\n",
(unsigned int)result->major_status, (int)result->minor_status);
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
return (TRUE);
}
bool_t
gssd_release_name_1_svc(release_name_args *argp, release_name_res *result, struct svc_req *rqstp)
{
gss_name_t name = gssd_find_resource(argp->input_name);
if (name) {
result->major_status = gss_release_name(&result->minor_status,
&name);
gssd_delete_resource(argp->input_name);
} else {
result->major_status = GSS_S_COMPLETE;
result->minor_status = 0;
}
gssd_verbose_out("gssd_release_name: done major=0x%x minor=%d\n",
(unsigned int)result->major_status, (int)result->minor_status);
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
return (TRUE);
}
bool_t
gssd_pname_to_uid_1_svc(pname_to_uid_args *argp, pname_to_uid_res *result, struct svc_req *rqstp)
{
gss_name_t name = gssd_find_resource(argp->pname);
uid_t uid;
char buf[1024], *bufp;
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
struct passwd pwd, *pw;
size_t buflen;
int error;
static size_t buflen_hint = 1024;
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
memset(result, 0, sizeof(*result));
if (name) {
result->major_status =
gss_pname_to_uid(&result->minor_status,
name, argp->mech, &uid);
if (result->major_status == GSS_S_COMPLETE) {
result->uid = uid;
buflen = buflen_hint;
for (;;) {
pw = NULL;
bufp = buf;
if (buflen > sizeof(buf))
bufp = malloc(buflen);
if (bufp == NULL)
break;
error = getpwuid_r(uid, &pwd, bufp, buflen,
&pw);
if (error != ERANGE)
break;
if (buflen > sizeof(buf))
free(bufp);
buflen += 1024;
if (buflen > buflen_hint)
buflen_hint = buflen;
}
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 (pw) {
int len = NGRPS;
int groups[NGRPS];
result->gid = pw->pw_gid;
getgrouplist(pw->pw_name, pw->pw_gid,
groups, &len);
result->gidlist.gidlist_len = len;
result->gidlist.gidlist_val =
mem_alloc(len * sizeof(int));
memcpy(result->gidlist.gidlist_val, groups,
len * sizeof(int));
gssd_verbose_out("gssd_pname_to_uid: mapped"
" to uid=%d, gid=%d\n", (int)result->uid,
(int)result->gid);
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 {
result->gid = 65534;
result->gidlist.gidlist_len = 0;
result->gidlist.gidlist_val = NULL;
gssd_verbose_out("gssd_pname_to_uid: mapped"
" to uid=%d, but no groups\n",
(int)result->uid);
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 (bufp != NULL && buflen > sizeof(buf))
free(bufp);
} else
gssd_verbose_out("gssd_pname_to_uid: failed major=0x%x"
" minor=%d\n", (unsigned int)result->major_status,
(int)result->minor_status);
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 {
result->major_status = GSS_S_BAD_NAME;
result->minor_status = 0;
gssd_verbose_out("gssd_pname_to_uid: no name\n");
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
}
return (TRUE);
}
bool_t
gssd_acquire_cred_1_svc(acquire_cred_args *argp, acquire_cred_res *result, struct svc_req *rqstp)
{
gss_name_t desired_name = GSS_C_NO_NAME;
gss_cred_id_t cred;
char ccname[PATH_MAX + 5 + 1], *cp, *cp2;
int gotone;
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
memset(result, 0, sizeof(*result));
if (ccfile_dirlist[0] != '\0' && argp->desired_name == 0) {
/*
* For the "-s" case and no name provided as an
* argument, search the directory list for an appropriate
* credential cache file. If the search fails, return failure.
*/
gotone = 0;
cp = ccfile_dirlist;
do {
cp2 = strchr(cp, ':');
if (cp2 != NULL)
*cp2 = '\0';
gotone = find_ccache_file(cp, argp->uid, ccname);
if (gotone != 0)
break;
if (cp2 != NULL)
*cp2++ = ':';
cp = cp2;
} while (cp != NULL && *cp != '\0');
if (gotone == 0) {
result->major_status = GSS_S_CREDENTIALS_EXPIRED;
gssd_verbose_out("gssd_acquire_cred: no cred cache"
" file found\n");
return (TRUE);
}
} else {
/*
* If there wasn't a "-s" option or the name has
* been provided as an argument, do it the old way.
* When a name is provided, it will normally exist in the
* default keytab file and the uid will be root.
*/
if (argp->desired_name != 0 && argp->uid != 0) {
if (debug_level == 0)
syslog(LOG_ERR, "gss_acquire_cred:"
" principal_name for non-root");
else
fprintf(stderr, "gss_acquire_cred:"
" principal_name for non-root\n");
}
snprintf(ccname, sizeof(ccname), "FILE:/tmp/krb5cc_%d",
(int) argp->uid);
}
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
setenv("KRB5CCNAME", ccname, TRUE);
if (argp->desired_name) {
desired_name = gssd_find_resource(argp->desired_name);
if (!desired_name) {
result->major_status = GSS_S_BAD_NAME;
gssd_verbose_out("gssd_acquire_cred: no desired name"
" found\n");
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
return (TRUE);
}
}
result->major_status = gss_acquire_cred(&result->minor_status,
desired_name, argp->time_req, argp->desired_mechs,
argp->cred_usage, &cred, &result->actual_mechs, &result->time_rec);
gssd_verbose_out("gssd_acquire_cred: done major=0x%x minor=%d\n",
(unsigned int)result->major_status, (int)result->minor_status);
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 (result->major_status == GSS_S_COMPLETE)
result->output_cred = gssd_make_resource(cred);
else
result->output_cred = 0;
return (TRUE);
}
bool_t
gssd_set_cred_option_1_svc(set_cred_option_args *argp, set_cred_option_res *result, struct svc_req *rqstp)
{
gss_cred_id_t cred = gssd_find_resource(argp->cred);
memset(result, 0, sizeof(*result));
if (!cred) {
result->major_status = GSS_S_CREDENTIALS_EXPIRED;
gssd_verbose_out("gssd_set_cred: no credentials\n");
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
return (TRUE);
}
result->major_status = gss_set_cred_option(&result->minor_status,
&cred, argp->option_name, &argp->option_value);
gssd_verbose_out("gssd_set_cred: done major=0x%x minor=%d\n",
(unsigned int)result->major_status, (int)result->minor_status);
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
return (TRUE);
}
bool_t
gssd_release_cred_1_svc(release_cred_args *argp, release_cred_res *result, struct svc_req *rqstp)
{
gss_cred_id_t cred = gssd_find_resource(argp->cred);
if (cred) {
result->major_status = gss_release_cred(&result->minor_status,
&cred);
gssd_delete_resource(argp->cred);
} else {
result->major_status = GSS_S_COMPLETE;
result->minor_status = 0;
}
gssd_verbose_out("gssd_release_cred: done major=0x%x minor=%d\n",
(unsigned int)result->major_status, (int)result->minor_status);
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
return (TRUE);
}
bool_t
gssd_display_status_1_svc(display_status_args *argp, display_status_res *result, struct svc_req *rqstp)
{
result->message_context = argp->message_context;
result->major_status = gss_display_status(&result->minor_status,
argp->status_value, argp->status_type, argp->mech_type,
&result->message_context, &result->status_string);
gssd_verbose_out("gssd_display_status: done major=0x%x minor=%d\n",
(unsigned int)result->major_status, (int)result->minor_status);
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
return (TRUE);
}
int
gssd_1_freeresult(SVCXPRT *transp, xdrproc_t xdr_result, caddr_t result)
{
/*
* We don't use XDR to free the results - anything which was
* allocated came from GSS-API. We use xdr_result to figure
* out what to do.
*/
OM_uint32 junk;
if (xdr_result == (xdrproc_t) xdr_init_sec_context_res) {
init_sec_context_res *p = (init_sec_context_res *) result;
gss_release_buffer(&junk, &p->output_token);
} else if (xdr_result == (xdrproc_t) xdr_accept_sec_context_res) {
accept_sec_context_res *p = (accept_sec_context_res *) result;
gss_release_buffer(&junk, &p->output_token);
} else if (xdr_result == (xdrproc_t) xdr_delete_sec_context_res) {
delete_sec_context_res *p = (delete_sec_context_res *) result;
gss_release_buffer(&junk, &p->output_token);
} else if (xdr_result == (xdrproc_t) xdr_export_sec_context_res) {
export_sec_context_res *p = (export_sec_context_res *) result;
if (p->interprocess_token.length)
memset(p->interprocess_token.value, 0,
p->interprocess_token.length);
gss_release_buffer(&junk, &p->interprocess_token);
} else if (xdr_result == (xdrproc_t) xdr_export_name_res) {
export_name_res *p = (export_name_res *) result;
gss_release_buffer(&junk, &p->exported_name);
} else if (xdr_result == (xdrproc_t) xdr_acquire_cred_res) {
acquire_cred_res *p = (acquire_cred_res *) result;
gss_release_oid_set(&junk, &p->actual_mechs);
} else if (xdr_result == (xdrproc_t) xdr_pname_to_uid_res) {
pname_to_uid_res *p = (pname_to_uid_res *) result;
if (p->gidlist.gidlist_val)
free(p->gidlist.gidlist_val);
} else if (xdr_result == (xdrproc_t) xdr_display_status_res) {
display_status_res *p = (display_status_res *) result;
gss_release_buffer(&junk, &p->status_string);
}
return (TRUE);
}
/*
* Search a directory for the most likely candidate to be used as the
* credential cache for a uid. If successful, return 1 and fill the
* file's path id into "rpath". Otherwise, return 0.
*/
static int
find_ccache_file(const char *dirpath, uid_t uid, char *rpath)
{
DIR *dirp;
struct dirent *dp;
struct stat sb;
time_t exptime, oexptime;
int gotone, len, rating, orating;
char namepath[PATH_MAX + 5 + 1];
char retpath[PATH_MAX + 5 + 1];
dirp = opendir(dirpath);
if (dirp == NULL)
return (0);
gotone = 0;
orating = 0;
oexptime = 0;
while ((dp = readdir(dirp)) != NULL) {
len = snprintf(namepath, sizeof(namepath), "%s/%s", dirpath,
dp->d_name);
if (len < sizeof(namepath) &&
strstr(dp->d_name, ccfile_substring) != NULL &&
lstat(namepath, &sb) >= 0 &&
sb.st_uid == uid &&
S_ISREG(sb.st_mode)) {
len = snprintf(namepath, sizeof(namepath), "FILE:%s/%s",
dirpath, dp->d_name);
if (len < sizeof(namepath) &&
is_a_valid_tgt_cache(namepath, uid, &rating,
&exptime) != 0) {
if (gotone == 0 || rating > orating ||
(rating == orating && exptime > oexptime)) {
orating = rating;
oexptime = exptime;
strcpy(retpath, namepath);
gotone = 1;
}
}
}
}
closedir(dirp);
if (gotone != 0) {
strcpy(rpath, retpath);
return (1);
}
return (0);
}
/*
* Try to determine if the file is a valid tgt cache file.
* Check that the file has a valid tgt for a principal.
* If it does, return 1, otherwise return 0.
* It also returns a "rating" and the expiry time for the TGT, when found.
* This "rating" is higher based on heuristics that make it more
* likely to be the correct credential cache file to use. It can
* be used by the caller, along with expiry time, to select from
* multiple credential cache files.
*/
static int
is_a_valid_tgt_cache(const char *filepath, uid_t uid, int *retrating,
time_t *retexptime)
{
#ifndef WITHOUT_KERBEROS
krb5_context context;
krb5_principal princ;
krb5_ccache ccache;
krb5_error_code retval;
krb5_cc_cursor curse;
krb5_creds krbcred;
int gotone, orating, rating, ret;
struct passwd *pw;
char *cp, *cp2, *pname;
time_t exptime;
/* Find a likely name for the uid principal. */
pw = getpwuid(uid);
/*
* Do a bunch of krb5 library stuff to try and determine if
* this file is a credentials cache with an appropriate TGT
* in it.
*/
retval = krb5_init_context(&context);
if (retval != 0)
return (0);
retval = krb5_cc_resolve(context, filepath, &ccache);
if (retval != 0) {
krb5_free_context(context);
return (0);
}
ret = 0;
orating = 0;
exptime = 0;
retval = krb5_cc_start_seq_get(context, ccache, &curse);
if (retval == 0) {
while ((retval = krb5_cc_next_cred(context, ccache, &curse,
&krbcred)) == 0) {
gotone = 0;
rating = 0;
retval = krb5_unparse_name(context, krbcred.server,
&pname);
if (retval == 0) {
cp = strchr(pname, '/');
if (cp != NULL) {
*cp++ = '\0';
if (strcmp(pname, "krbtgt") == 0 &&
krbcred.times.endtime > time(NULL)
) {
gotone = 1;
/*
* Test to see if this is a
* tgt for cross-realm auth.
* Rate it higher, if it is not.
*/
cp2 = strchr(cp, '@');
if (cp2 != NULL) {
*cp2++ = '\0';
if (strcmp(cp, cp2) ==
0)
rating++;
}
}
}
free(pname);
}
if (gotone != 0) {
retval = krb5_unparse_name(context,
krbcred.client, &pname);
if (retval == 0) {
cp = strchr(pname, '@');
if (cp != NULL) {
*cp++ = '\0';
if (pw != NULL && strcmp(pname,
pw->pw_name) == 0)
rating++;
if (strchr(pname, '/') == NULL)
rating++;
if (pref_realm[0] != '\0' &&
strcmp(cp, pref_realm) == 0)
rating++;
}
}
free(pname);
if (rating > orating) {
orating = rating;
exptime = krbcred.times.endtime;
} else if (rating == orating &&
krbcred.times.endtime > exptime)
exptime = krbcred.times.endtime;
ret = 1;
}
krb5_free_cred_contents(context, &krbcred);
}
krb5_cc_end_seq_get(context, ccache, &curse);
}
krb5_cc_close(context, ccache);
krb5_free_context(context);
if (ret != 0) {
*retrating = orating;
*retexptime = exptime;
}
return (ret);
#else /* WITHOUT_KERBEROS */
return (0);
#endif /* !WITHOUT_KERBEROS */
}