Ubuntu 16.10 (Yakkety) 上部署 ocserv 并且解决掉几个坑的记录

Ubuntu 16.10 (Yakkety)正式发布已经有1天了,ocserv 这种主流的包也有直接提供,所以简易安装一下就好:

$ sudo apt-get install ocserv

因为我自己有证书,所以直接把相关证书链合并之后随意扔到某个目录就行,以下3个位置需要自行提供文件:

/etc/.passwd
/etc/server-cert.pem
/etc/server-key.pem
/etc/dh.pem
/etc/ca.pem

Generating ocpasswd:

sudo ocpasswd -c /etc/.passwd xxx

我的配置文件:

# User authentication method. Could be set multiple times and in
# that case all should succeed. To enable multiple methods use
# multiple auth directives. Available options: certificate,
# plain, pam, radius, gssapi.
#
# Note that authentication methods cannot be changed with reload.

# certificate:
#  This indicates that all connecting users must present a certificate.
#
# pam[gid-min=1000]:
#  This enabled PAM authentication of the user. The gid-min option is used
# by auto-select-group option, in order to select the minimum valid group ID.
#
# plain[passwd=/etc/ocserv/ocpasswd,otp=/etc/ocserv/users.otp]
#  The plain option requires specifying a password file which contains
# entries of the following format.
# "username:groupname1,groupname2:encoded-password"
# One entry must be listed per line, and 'ocpasswd' should be used
# to generate password entries. The 'otp' suboption allows to specify
# an oath password file to be used for one time passwords; the format of
# the file is described in https://code.google.com/p/mod-authn-otp/wiki/UsersFile
#
# radius[config=/etc/radiusclient/radiusclient.conf,groupconfig=true,nas-identifier=name,override-interim-updates=false]:
#  The radius option requires specifying freeradius-client configuration
# file. If the groupconfig option is set, then config-per-user/group will be overriden,
# and all configuration will be read from radius. The 'override-interim-updates' if set to
# true will ignore Acct-Interim-Interval from the server and 'stats-report-time' will be considered.
#
# gssapi[keytab=/etc/key.tab,require-local-user-map=true,tgt-freshness-time=900]
#  The gssapi option allows to use authentication methods supported by GSSAPI,
# such as Kerberos tickets with ocserv. It should be best used as an alternative
# to PAM (i.e., have pam in auth and gssapi in enable-auth), to allow users with
# tickets and without tickets to login. The default value for require-local-user-map
# is true. The 'tgt-freshness-time' if set, it would require the TGT tickets presented
# to have been issued within the provided number of seconds. That option is used to
# restrict logins even if the KDC provides long time TGT tickets.

#auth = "pam"
#auth = "pam[gid-min=1000]"
#auth = "plain[passwd=./sample.passwd,otp=./sample.otp]"
auth = "plain[passwd=/etc/.passwd]"
#auth = "certificate"
#auth = "radius[config=/etc/radiusclient/radiusclient.conf,groupconfig=true]"

# Specify alternative authentication methods that are sufficient
# for authentication. That is, if set, any of the methods enabled
# will be sufficient to login.
#enable-auth = "certificate"
#enable-auth = "gssapi"
#enable-auth = "gssapi[keytab=/etc/key.tab,require-local-user-map=true,tgt-freshness-time=900]"

# Accounting methods available:
# radius: can be combined with any authentication method, it provides
#      radius accounting to available users (see also stats-report-time).
#
# pam: can be combined with any authentication method, it provides
#      a validation of the connecting user's name using PAM. It is
#      superfluous to use this method when authentication is already
#      PAM.
#
# Only one accounting method can be specified.
#acct = "radius[config=/etc/radiusclient/radiusclient.conf]"

# Use listen-host to limit to specific IPs or to the IPs of a provided
# hostname.
listen-host = i.example.com

# When the server has a dynamic DNS address (that may change),
# should set that to true to ask the client to resolve again on
# reconnects.
#listen-host-is-dyndns = true

# TCP and UDP port number
tcp-port = 443
udp-port = 443

# Accept connections using a socket file. It accepts HTTP
# connections (i.e., without SSL/TLS unlike its TCP counterpart),
# and uses it as the primary channel. That option cannot be
# combined with certificate authentication.
#listen-clear-file = /var/run/ocserv-conn.socket

# The user the worker processes will be run as. It should be
# unique (no other services run as this user).
run-as-user = nobody
run-as-group = daemon

# socket file used for IPC with occtl. You only need to set that,
# if you use more than a single servers.
#occtl-socket-file = /var/run/occtl.socket

# socket file used for server IPC (worker-main), will be appended with .PID
# It must be accessible within the chroot environment (if any), so it is best
# specified relatively to the chroot directory.
socket-file = /var/run/ocserv-socket

# The default server directory. Does not require any devices present.
#chroot-dir = /path/to/chroot

# The key and the certificates of the server
# The key may be a file, or any URL supported by GnuTLS (e.g.,
# tpmkey:uuid=xxxxxxx-xxxx-xxxx-xxxx-xxxxxxxx;storage=user
# or pkcs11:object=my-vpn-key;object-type=private)
#
# The server-cert file may contain a single certificate, or
# a sorted certificate chain.
#
# There may be multiple server-cert and server-key directives,
# but each key should correspond to the preceding certificate.
server-cert = /etc/server-cert.pem
server-key = /etc/server-key.pem

# Diffie-Hellman parameters. Only needed if you require support
# for the DHE ciphersuites (by default this server supports ECDHE).
# Can be generated using:
# certtool --generate-dh-params --outfile /path/to/dh.pem
dh-params = /etc/dh.pem

# In case PKCS #11, TPM or encrypted keys are used the PINs should be available
# in files. The srk-pin-file is applicable to TPM keys only, and is the
# storage root key.
#pin-file = /path/to/pin.txt
#srk-pin-file = /path/to/srkpin.txt

# The password or PIN needed to unlock the key in server-key file.
# Only needed if the file is encrypted or a PKCS #11 object. This
# is an alternative method to pin-file.
#key-pin = 1234

# The SRK PIN for TPM.
# This is an alternative method to srk-pin-file.
#srk-pin = 1234

# The Certificate Authority that will be used to verify
# client certificates (public keys) if certificate authentication
# is set.
ca-cert = /etc/ca.pem


### All configuration options below this line are reloaded on a SIGHUP.
### The options above, will remain unchanged. Note however, that the
### server-cert, server-key, dh-params and ca-cert options will be reloaded
### if the provided file changes, on server reload. That allows certificate
### rotation, but requires the server key to remain the same for seamless
### operation. If the server key changes on reload, there may be connection
### failures during the reloading time.


# Whether to enable seccomp/Linux namespaces worker isolation. That restricts the number of
# system calls allowed to a worker process, in order to reduce damage from a
# bug in the worker process. It is available on Linux systems at a performance cost.
# The performance cost is roughly 2% overhead at transfer time (tested on a Linux 3.17.8).
# Note however, that process isolation is restricted to the specific libc versions
# the isolation was tested at. If you get random failures on worker processes, try
# disabling that option and report the failures you, along with system and debugging
# information at: https://gitlab.com/ocserv/ocserv/issues
isolate-workers = true

# A banner to be displayed on clients
#banner = "Welcome"

# Limit the number of clients. Unset or set to zero for unlimited.
#max-clients = 1024
max-clients = 16

# Limit the number of identical clients (i.e., users connecting
# multiple times). Unset or set to zero for unlimited.
max-same-clients = 2

# When the server receives connections from a proxy, like haproxy
# which supports the proxy protocol, set this to obtain the correct
# client addresses. The proxy protocol (v2) would then be expected in
# the TCP or UNIX socket (not the UDP one).
#listen-proxy-proto = true

# Limit the number of client connections to one every X milliseconds
# (X is the provided value). Set to zero for no limit.
#rate-limit-ms = 100

# Stats report time. The number of seconds after which each
# worker process will report its usage statistics (number of
# bytes transferred etc). This is useful when accounting like
# radius is in use.
#stats-report-time = 360

# Keepalive in seconds
keepalive = 32400

# Dead peer detection in seconds.
# Note that when the client is behind a NAT this value
# needs to be short enough to prevent the NAT disassociating
# his UDP session from the port number. Otherwise the client
# could have his UDP connection stalled, for several minutes.
dpd = 90

# Dead peer detection for mobile clients. That needs to
# be higher to prevent such clients being awaken too
# often by the DPD messages, and save battery.
# The mobile clients are distinguished from the header
# 'X-AnyConnect-Identifier-DeviceType'.
mobile-dpd = 1800

# MTU discovery (DPD must be enabled)
try-mtu-discovery = false

# If you have a certificate from a CA that provides an OCSP
# service you may provide a fresh OCSP status response within
# the TLS handshake. That will prevent the client from connecting
# independently on the OCSP server.
# You can update this response periodically using:
# ocsptool --ask --load-cert=your_cert --load-issuer=your_ca --outfile response
# Make sure that you replace the following file in an atomic way.
#ocsp-response = /path/to/ocsp.der

# The object identifier that will be used to read the user ID in the client
# certificate. The object identifier should be part of the certificate's DN
# Useful OIDs are:
#  CN = 2.5.4.3, UID = 0.9.2342.19200300.100.1.1
cert-user-oid = 0.9.2342.19200300.100.1.1

# The object identifier that will be used to read the user group in the
# client certificate. The object identifier should be part of the certificate's
# DN. If the user may belong to multiple groups, then use multiple such fields
# in the certificate's DN. Useful OIDs are:
#  OU (organizational unit) = 2.5.4.11
#cert-group-oid = 2.5.4.11

# The revocation list of the certificates issued by the 'ca-cert' above.
# See the manual to generate an empty CRL initially. The CRL will be reloaded
# periodically when ocserv detects a change in the file. To force a reload use
# SIGHUP.
#crl = /path/to/crl.pem

# Uncomment this to enable compression negotiation (LZS, LZ4).
#compression = true

# Set the minimum size under which a packet will not be compressed.
# That is to allow low-latency for VoIP packets. The default size
# is 256 bytes. Modify it if the clients typically use compression
# as well of VoIP with codecs that exceed the default value.
#no-compress-limit = 256

# GnuTLS priority string; note that SSL 3.0 is disabled by default
# as there are no openconnect (and possibly anyconnect clients) using
# that protocol. The string below does not enforce perfect forward
# secrecy, in order to be compatible with legacy clients.
#
# Note that the most performant ciphersuites are the moment are the ones
# involving AES-GCM. These are very fast in x86 and x86-64 hardware, and
# in addition require no padding, thus taking full advantage of the MTU.
# For that to be taken advantage of, the openconnect client must be
# used, and the server must be compiled against GnuTLS 3.2.7 or later.
# Use "gnutls-cli --benchmark-tls-ciphers", to see the performance
# difference with AES_128_CBC_SHA1 (the default for anyconnect clients)
# in your system.

tls-priorities = "NORMAL:%SERVER_PRECEDENCE:%COMPAT:-VERS-SSL3.0"

# More combinations in priority strings are available, check
# http://gnutls.org/manual/html_node/Priority-Strings.html
# E.g., the string below enforces perfect forward secrecy (PFS)
# on the main channel.
#tls-priorities = "NORMAL:%SERVER_PRECEDENCE:%COMPAT:-RSA:-VERS-SSL3.0:-ARCFOUR-128"

# The time (in seconds) that a client is allowed to stay connected prior
# to authentication
auth-timeout = 240

# The time (in seconds) that a client is allowed to stay idle (no traffic)
# before being disconnected. Unset to disable.
#idle-timeout = 1200

# The time (in seconds) that a mobile client is allowed to stay idle (no
# traffic) before being disconnected. Unset to disable.
#mobile-idle-timeout = 2400

# The time (in seconds) that a client is not allowed to reconnect after
# a failed authentication attempt.
min-reauth-time = 300

# Banning clients in ocserv works with a point system. IP addresses
# that get a score over that configured number are banned for
# min-reauth-time seconds. By default a wrong password attempt is 10 points,
# a KKDCP POST is 1 point, and a connection is 1 point. Note that
# due to difference processes being involved the count of points
# will not be real-time precise.
#
# Score banning cannot be reliably used when receiving proxied connections
# locally from an HTTP server (i.e., when listen-clear-file is used).
#
# Set to zero to disable.
max-ban-score = 50

# The time (in seconds) that all score kept for a client is reset.
ban-reset-time = 300

# In case you'd like to change the default points.
#ban-points-wrong-password = 10
#ban-points-connection = 1
#ban-points-kkdcp = 1

# Cookie timeout (in seconds)
# Once a client is authenticated he's provided a cookie with
# which he can reconnect. That cookie will be invalidated if not
# used within this timeout value. This cookie remains valid, during
# the user's connected time, and after user disconnection it
# remains active for this amount of time. That setting should allow a
# reasonable amount of time for roaming between different networks.
cookie-timeout = 300

# If this is enabled (not recommended) the cookies will stay
# valid even after a user manually disconnects, and until they
# expire. This may improve roaming with some broken clients.
#persistent-cookies = true

# Whether roaming is allowed, i.e., if true a cookie is
# restricted to a single IP address and cannot be re-used
# from a different IP.
deny-roaming = false

# ReKey time (in seconds)
# ocserv will ask the client to refresh keys periodically once
# this amount of seconds is elapsed. Set to zero to disable (note
# that, some clients fail if rekey is disabled).
rekey-time = 172800

# ReKey method
# Valid options: ssl, new-tunnel
#  ssl: Will perform an efficient rehandshake on the channel allowing
#       a seamless connection during rekey.
#  new-tunnel: Will instruct the client to discard and re-establish the channel.
#       Use this option only if the connecting clients have issues with the ssl
#       option.
rekey-method = ssl

# Script to call when a client connects and obtains an IP.
# The following parameters are passed on the environment.
# REASON, USERNAME, GROUPNAME, DEVICE, IP_REAL (the real IP of the client),
# IP_REAL_LOCAL (the local interface IP the client connected), IP_LOCAL
# (the local IP in the P-t-P connection), IP_REMOTE (the VPN IP of the client),
# IPV6_LOCAL (the IPv6 local address if there are both IPv4 and IPv6
# assigned), IPV6_REMOTE (the IPv6 remote address), IPV6_PREFIX, and
# ID (a unique numeric ID); REASON may be "connect" or "disconnect".
# In addition the following variables OCSERV_ROUTES (the applied routes for this
# client), OCSERV_NO_ROUTES, OCSERV_DNS (the DNS servers for this client),
# will contain a space separated list of routes or DNS servers. A version
# of these variables with the 4 or 6 suffix will contain only the IPv4 or
# IPv6 values.

# The disconnect script will receive the additional values: STATS_BYTES_IN,
# STATS_BYTES_OUT, STATS_DURATION that contain a 64-bit counter of the bytes
# output from the tun device, and the duration of the session in seconds.

#connect-script = /usr/bin/myscript
#disconnect-script = /usr/bin/myscript

# UTMP
# Register the connected clients to utmp. This will allow viewing
# the connected clients using the command 'who'.
#use-utmp = true

# Whether to enable support for the occtl tool (i.e., either through D-BUS,
# or via a unix socket).
use-occtl = true

# PID file. It can be overriden in the command line.
pid-file = /var/run/ocserv.pid

# Set the protocol-defined priority (SO_PRIORITY) for packets to
# be sent. That is a number from 0 to 6 with 0 being the lowest
# priority. Alternatively this can be used to set the IP Type-
# Of-Service, by setting it to a hexadecimal number (e.g., 0x20).
# This can be set per user/group or globally.
#net-priority = 3

# Set the VPN worker process into a specific cgroup. This is Linux
# specific and can be set per user/group or globally.
#cgroup = "cpuset,cpu:test"

#
# Network settings
#

# The name to use for the tun device
device = vpns

# Whether the generated IPs will be predictable, i.e., IP stays the
# same for the same user when possible.
predictable-ips = true

# The default domain to be advertised
default-domain = i.gnchen.com

# The pool of addresses that leases will be given from. If the leases
# are given via Radius, or via the explicit-ip? per-user config option then
# these network values should contain a network with at least a single
# address that will remain under the full control of ocserv (that is
# to be able to assign the local part of the tun device address).
# Note that, you could use addresses from a subnet of your LAN network if you
# enable proxy arp in the LAN interface (see http://infradead.org/ocserv/recipes-ocserv-pseudo-bridge.html);
# in that case it is recommended to set ping-leases to true.
ipv4-network = 192.168.1.0
ipv4-netmask = 255.255.255.0

# An alternative way of specifying the network:
#ipv4-network = 192.168.1.0/24

# The IPv6 subnet that leases will be given from.
#ipv6-network = fda9:4efe:7e3b:03ea::/48

# Specify the size of the network to provide to clients. It is
# generally recommended to provide clients with a /64 network in
# IPv6, but any subnet may be specified. To provide clients only
# with a single IP use the prefix 128.
#ipv6-subnet-prefix = 128
#ipv6-subnet-prefix = 64

# Whether to tunnel all DNS queries via the VPN. This is the default
# when a default route is set.
#tunnel-all-dns = true

# The advertized DNS server. Use multiple lines for
# multiple servers.
# dns = fc00::4be0
dns = 8.8.8.8
dns = 8.8.4.4

# The NBNS server (if any)
#nbns = 192.168.1.3

# The domains over which the provided DNS should be used. Use
# multiple lines for multiple domains.
#split-dns = example.com

# Prior to leasing any IP from the pool ping it to verify that
# it is not in use by another (unrelated to this server) host.
# Only set to true, if there can be occupied addresses in the
# IP range for leases.
ping-leases = false

# Use this option to set a link MTU value to the incoming
# connections. Unset to use the default MTU of the TUN device.
# Note that the MTU is negotiated using the value set and the
# value sent by the peer.
#mtu = 1420

# Unset to enable bandwidth restrictions (in bytes/sec). The
# setting here is global, but can also be set per user or per group.
#rx-data-per-sec = 40000
#tx-data-per-sec = 40000

# The number of packets (of MTU size) that are available in
# the output buffer. The default is low to improve latency.
# Setting it higher will improve throughput.
#output-buffer = 10

# Routes to be forwarded to the client. If you need the
# client to forward routes to the server, you may use the
# config-per-user/group or even connect and disconnect scripts.
#
# To set the server as the default gateway for the client just
# comment out all routes from the server, or use the special keyword
# 'default'.

#route = 10.10.10.0/255.255.255.0
#route = 192.168.1.0/255.255.0.0


#route = fef4:db8:1000:1001::/64
#route = default

# Subsets of the routes above that will not be routed by
# the server.

#no-route = 192.168.5.0/255.255.255.0
no-route = 1.0.0.0/255.192.0.0
no-route = 1.64.0.0/255.224.0.0
no-route = 1.112.0.0/255.248.0.0
no-route = 1.176.0.0/255.240.0.0
no-route = 1.192.0.0/255.240.0.0
no-route = 14.0.0.0/255.224.0.0
no-route = 14.96.0.0/255.224.0.0
no-route = 14.128.0.0/255.224.0.0
no-route = 14.192.0.0/255.224.0.0
no-route = 27.0.0.0/255.192.0.0
no-route = 27.96.0.0/255.224.0.0
no-route = 27.128.0.0/255.224.0.0
no-route = 27.176.0.0/255.240.0.0
no-route = 27.192.0.0/255.224.0.0
no-route = 27.224.0.0/255.252.0.0
no-route = 36.0.0.0/255.192.0.0
no-route = 36.96.0.0/255.224.0.0
no-route = 36.128.0.0/255.192.0.0
no-route = 36.192.0.0/255.224.0.0
no-route = 36.240.0.0/255.240.0.0
no-route = 39.0.0.0/255.255.0.0
no-route = 39.64.0.0/255.224.0.0
no-route = 39.96.0.0/255.240.0.0
no-route = 39.128.0.0/255.192.0.0
no-route = 40.72.0.0/255.254.0.0
no-route = 40.124.0.0/255.252.0.0
no-route = 42.0.0.0/255.248.0.0
no-route = 42.48.0.0/255.240.0.0
no-route = 42.80.0.0/255.240.0.0
no-route = 42.96.0.0/255.224.0.0
no-route = 42.128.0.0/255.128.0.0
no-route = 43.224.0.0/255.224.0.0
no-route = 45.3.32.0/255.255.224.0
no-route = 45.65.16.0/255.255.240.0
no-route = 45.78.80.0/255.255.240.0
no-route = 45.112.0.0/255.240.0.0
no-route = 45.248.0.0/255.248.0.0
no-route = 47.92.0.0/255.252.0.0
no-route = 47.96.0.0/255.224.0.0
no-route = 49.0.0.0/255.128.0.0
no-route = 49.128.0.0/255.224.0.0
no-route = 49.192.0.0/255.192.0.0
no-route = 52.80.0.0/255.252.0.0
no-route = 54.222.0.0/255.254.0.0
no-route = 58.0.0.0/255.128.0.0
no-route = 58.128.0.0/255.224.0.0
no-route = 58.192.0.0/255.224.0.0
no-route = 58.240.0.0/255.240.0.0
no-route = 59.32.0.0/255.224.0.0
no-route = 59.64.0.0/255.224.0.0
no-route = 59.96.0.0/255.240.0.0
no-route = 59.144.0.0/255.240.0.0
no-route = 59.160.0.0/255.224.0.0
no-route = 59.192.0.0/255.192.0.0
no-route = 60.0.0.0/255.224.0.0
no-route = 60.48.0.0/255.240.0.0
no-route = 60.160.0.0/255.224.0.0
no-route = 60.192.0.0/255.192.0.0
no-route = 61.0.0.0/255.192.0.0
no-route = 61.80.0.0/255.248.0.0
no-route = 61.128.0.0/255.192.0.0
no-route = 61.224.0.0/255.224.0.0
no-route = 91.234.36.0/255.255.255.0
no-route = 101.0.0.0/255.128.0.0
no-route = 101.128.0.0/255.224.0.0
no-route = 101.192.0.0/255.240.0.0
no-route = 101.224.0.0/255.224.0.0
no-route = 103.0.0.0/255.0.0.0
no-route = 104.167.16.0/255.255.240.0
no-route = 104.207.32.0/255.255.224.0
no-route = 106.0.0.0/255.128.0.0
no-route = 106.224.0.0/255.240.0.0
no-route = 110.0.0.0/255.128.0.0
no-route = 110.144.0.0/255.240.0.0
no-route = 110.160.0.0/255.224.0.0
no-route = 110.192.0.0/255.192.0.0
no-route = 111.0.0.0/255.192.0.0
no-route = 111.64.0.0/255.224.0.0
no-route = 111.112.0.0/255.240.0.0
no-route = 111.128.0.0/255.192.0.0
no-route = 111.192.0.0/255.224.0.0
no-route = 111.224.0.0/255.240.0.0
no-route = 112.0.0.0/255.128.0.0
no-route = 112.128.0.0/255.240.0.0
no-route = 112.192.0.0/255.252.0.0
no-route = 112.224.0.0/255.224.0.0
no-route = 113.0.0.0/255.128.0.0
no-route = 113.128.0.0/255.240.0.0
no-route = 113.192.0.0/255.192.0.0
no-route = 114.16.0.0/255.240.0.0
no-route = 114.48.0.0/255.240.0.0
no-route = 114.64.0.0/255.192.0.0
no-route = 114.128.0.0/255.240.0.0
no-route = 114.192.0.0/255.192.0.0
no-route = 115.0.0.0/255.0.0.0
no-route = 116.0.0.0/255.0.0.0
no-route = 117.0.0.0/255.128.0.0
no-route = 117.128.0.0/255.192.0.0
no-route = 118.16.0.0/255.240.0.0
no-route = 118.64.0.0/255.192.0.0
no-route = 118.128.0.0/255.128.0.0
no-route = 119.0.0.0/255.128.0.0
no-route = 119.128.0.0/255.192.0.0
no-route = 119.224.0.0/255.224.0.0
no-route = 120.0.0.0/255.192.0.0
no-route = 120.64.0.0/255.224.0.0
no-route = 120.128.0.0/255.240.0.0
no-route = 120.192.0.0/255.192.0.0
no-route = 121.0.0.0/255.128.0.0
no-route = 121.192.0.0/255.192.0.0
no-route = 122.0.0.0/254.0.0.0
no-route = 124.0.0.0/255.0.0.0
no-route = 125.0.0.0/255.128.0.0
no-route = 125.160.0.0/255.224.0.0
no-route = 125.192.0.0/255.192.0.0
no-route = 137.59.59.0/255.255.255.0
no-route = 137.59.88.0/255.255.252.0
no-route = 139.0.0.0/255.224.0.0
no-route = 139.128.0.0/255.128.0.0
no-route = 140.64.0.0/255.240.0.0
no-route = 140.128.0.0/255.240.0.0
no-route = 140.192.0.0/255.192.0.0
no-route = 144.0.0.0/255.248.0.0
no-route = 144.12.0.0/255.255.0.0
no-route = 144.48.0.0/255.248.0.0
no-route = 144.123.0.0/255.255.0.0
no-route = 144.255.0.0/255.255.0.0
no-route = 146.196.0.0/255.255.128.0
no-route = 150.0.0.0/255.255.0.0
no-route = 150.96.0.0/255.224.0.0
no-route = 150.128.0.0/255.240.0.0
no-route = 150.192.0.0/255.192.0.0
no-route = 152.104.128.0/255.255.128.0
no-route = 153.0.0.0/255.192.0.0
no-route = 153.96.0.0/255.224.0.0
no-route = 157.0.0.0/255.255.0.0
no-route = 157.18.0.0/255.255.0.0
no-route = 157.61.0.0/255.255.0.0
no-route = 157.112.0.0/255.240.0.0
no-route = 157.144.0.0/255.240.0.0
no-route = 157.255.0.0/255.255.0.0
no-route = 159.226.0.0/255.255.0.0
no-route = 160.19.208.0/255.255.240.0
no-route = 160.20.48.0/255.255.252.0
no-route = 160.202.0.0/255.255.0.0
no-route = 160.238.64.0/255.255.252.0
no-route = 161.207.0.0/255.255.0.0
no-route = 162.105.0.0/255.255.0.0
no-route = 163.0.0.0/255.192.0.0
no-route = 163.96.0.0/255.224.0.0
no-route = 163.128.0.0/255.192.0.0
no-route = 163.192.0.0/255.224.0.0
no-route = 166.111.0.0/255.255.0.0
no-route = 167.139.0.0/255.255.0.0
no-route = 167.189.0.0/255.255.0.0
no-route = 167.220.244.0/255.255.252.0
no-route = 168.160.0.0/255.255.0.0
no-route = 170.179.0.0/255.255.0.0
no-route = 171.0.0.0/255.128.0.0
no-route = 171.192.0.0/255.224.0.0
no-route = 175.0.0.0/255.128.0.0
no-route = 175.128.0.0/255.192.0.0
no-route = 180.64.0.0/255.192.0.0
no-route = 180.128.0.0/255.128.0.0
no-route = 182.0.0.0/255.0.0.0
no-route = 183.0.0.0/255.192.0.0
no-route = 183.64.0.0/255.224.0.0
no-route = 183.128.0.0/255.128.0.0
no-route = 192.124.154.0/255.255.255.0
no-route = 192.140.128.0/255.255.128.0
no-route = 202.0.0.0/255.128.0.0
no-route = 202.128.0.0/255.192.0.0
no-route = 202.192.0.0/255.224.0.0
no-route = 203.0.0.0/255.0.0.0
no-route = 210.0.0.0/255.192.0.0
no-route = 210.64.0.0/255.224.0.0
no-route = 210.160.0.0/255.224.0.0
no-route = 210.192.0.0/255.224.0.0
no-route = 211.64.0.0/255.248.0.0
no-route = 211.80.0.0/255.240.0.0
no-route = 211.96.0.0/255.248.0.0
no-route = 211.136.0.0/255.248.0.0
no-route = 211.144.0.0/255.240.0.0
no-route = 211.160.0.0/255.248.0.0
no-route = 218.0.0.0/255.128.0.0
no-route = 218.160.0.0/255.224.0.0
no-route = 218.192.0.0/255.192.0.0
no-route = 219.64.0.0/255.224.0.0
no-route = 219.128.0.0/255.224.0.0
no-route = 219.192.0.0/255.192.0.0
no-route = 220.96.0.0/255.224.0.0
no-route = 220.128.0.0/255.128.0.0
no-route = 221.0.0.0/255.224.0.0
no-route = 221.96.0.0/255.224.0.0
no-route = 221.128.0.0/255.128.0.0
no-route = 222.0.0.0/255.0.0.0
no-route = 223.0.0.0/255.224.0.0
no-route = 223.64.0.0/255.192.0.0
no-route = 223.128.0.0/255.128.0.0
# Note the that following two firewalling options currently are available
# in Linux systems with iptables software.

# If set, the script /usr/bin/ocserv-fw will be called to restrict
# the user to its allowed routes and prevent him from accessing
# any other routes. In case of defaultroute, the no-routes are restricted.
# All the routes applied by ocserv can be reverted using /usr/bin/ocserv-fw
# --removeall. This option can be set globally or in the per-user configuration.
#restrict-user-to-routes = true

# This option implies restrict-user-to-routes set to true. If set, the
# script /usr/bin/ocserv-fw will be called to restrict the user to
# access specific ports in the network. This option can be set globally
# or in the per-user configuration.
#restrict-user-to-ports = "tcp(443), tcp(80), udp(443), sctp(99), tcp(583), icmp(), icmpv6()"

# You could also use negation, i.e., block the user from accessing these ports only.
#restrict-user-to-ports = "!(tcp(443), tcp(80))"

# When set to true, all client's iroutes are made visible to all
# connecting clients except for the ones offering them. This option
# only makes sense if config-per-user is set.
#expose-iroutes = true

# Groups that a client is allowed to select from.
# A client may belong in multiple groups, and in certain use-cases
# it is needed to switch between them. For these cases the client can
# select prior to authentication. Add multiple entries for multiple groups.
# The group may be followed by a user-friendly name in brackets.
#select-group = group1
#select-group = group2[My special group]

# The name of the (virtual) group that if selected it would assign the user
# to its default group.
#default-select-group = DEFAULT

# Instead of specifying manually all the allowed groups, you may instruct
# ocserv to scan all available groups and include the full list.
#auto-select-group = true

# Configuration files that will be applied per user connection or
# per group. Each file name on these directories must match the username
# or the groupname.
# The options allowed in the configuration files are dns, nbns,
#  ipv?-network, ipv4-netmask, rx/tx-per-sec, iroute, route, no-route,
#  explicit-ipv4, explicit-ipv6, net-priority, deny-roaming, no-udp,
#  keepalive, dpd, mobile-dpd, max-same-clients, tunnel-all-dns,
#  restrict-user-to-routes, user-profile, cgroup, stats-report-time,
#  mtu, idle-timeout, mobile-idle-timeout, restrict-user-to-ports,
#  and session-timeout.
#
# Note that the 'iroute' option allows to add routes on the server
# based on a user or group. The syntax depends on the input accepted
# by the commands route-add-cmd and route-del-cmd (see below). The no-udp
# is a boolean option (e.g., no-udp = true), and will prevent a UDP session
# for that specific user or group. The hostname option will set a
# hostname to override any proposed by the user. Note also, that, any
# routes, no-routes, DNS or NBNS servers present will overwrite the global ones.

#config-per-user = /etc/ocserv/config-per-user/
#config-per-group = /etc/ocserv/config-per-group/

# When config-per-xxx is specified and there is no group or user that
# matches, then utilize the following configuration.
#default-user-config = /etc/ocserv/defaults/user.conf
#default-group-config = /etc/ocserv/defaults/group.conf

# The system command to use to setup a route. %{R} will be replaced with the
# route/mask, %{RI} with the route in CIDR format, and %{D} with the (tun) device.
#
# The following example is from linux systems. %{R} should be something
# like 192.168.2.0/255.255.255.0 and %{RI} 192.168.2.0/24 (the argument of iroute).

route-add-cmd = "ip route add 192.168.1.0 dev tun0"
route-del-cmd = "ip route delete 192.168.1.0 dev tun0"

# This option allows to forward a proxy. The special keywords '%{U}'
# and '%{G}', if present will be replaced by the username and group name.
#proxy-url = http://example.com/
#proxy-url = http://example.com/%{U}/

# This option allows you to specify a URL location where a client can
# post using MS-KKDCP, and the message will be forwarded to the provided
# KDC server. That is a translation URL between HTTP and Kerberos.
# In MIT kerberos you'll need to add in realms:
#   EXAMPLE.COM = {
#     kdc = https://ocserv.example.com/KdcProxy
#     http_anchors = FILE:/etc/ocserv-ca.pem
#   }
# In some distributions the krb5-k5tls plugin of kinit is required.
#
# The following option is available in ocserv, when compiled with GSSAPI support.

#kkdcp = "SERVER-PATH KERBEROS-REALM PROTOCOL@SERVER:PORT"
#kkdcp = "/KdcProxy KERBEROS.REALM udp@127.0.0.1:88"
#kkdcp = "/KdcProxy KERBEROS.REALM tcp@127.0.0.1:88"
#kkdcp = "/KdcProxy KERBEROS.REALM tcp@[::1]:88"

#
# The following options are for (experimental) AnyConnect client
# compatibility.

# This option must be set to true to support legacy CISCO clients.
# A side effect of this option is that it will no longer be required
# for clients to present their certificate on every connection.
# That is they may resume a cookie without presenting a certificate
# (when certificate authentication is used).
cisco-client-compat = true

# Client profile xml. A sample file exists in doc/profile.xml.
# It is required by some of the CISCO clients.
# This file must be accessible from inside the worker's chroot.
# Note that enabling this option is not recommended as it will allow
# the worker processes to open arbitrary files (when isolate-workers is
# set to true).
#user-profile = /path/to/file.xml

#Advanced options

# Option to allow sending arbitrary custom headers to the client after
# authentication and prior to VPN tunnel establishment. You shouldn't
# need to use this option normally; if you do and you think that
# this may help others, please send your settings and reason to
# the openconnect mailing list. The special keywords '%{U}'
# and '%{G}', if present will be replaced by the username and group name.
#custom-header = "X-My-Header: hi there"

弄完之后发现了一个bug,用

systemctl status ocserv

查看状态时发现了如下的提示:

GnuTLS error (at worker-vpn.c:585): Error in the system's randomness device.

Google了一下发现Debian的Mailing List已经有人解决了这个问题(http://www.mail-archive.com/debian-bugs-dist@lists.debian.org/msg1459291.html):

Hi

I can confirm that setting "isolate-workers=false" the connection can be established.
Still I think/feel this should be fixed one way or the other.

regards,
Gergely


On 2016-10-06 16:39, Nikos Mavrogiannopoulos wrote:
Hi,
  You can work-around the issue by setting isolate-workers=false. The
problem is that the getrandom() system call is not included in the
whitelisted seccomp filter. The "right" solution is to either apply
patch [0] or move to 0.11.5.

regards,
Nikos

[0]. 
http://pkgs.fedoraproject.org/cgit/rpms/ocserv.git/commit/?id=d0dbbc1a1988c995771c0bbb85894e723049b5ef

On Thu, Oct 6, 2016 at 4:19 PM, Gergely Katona <n...@tfw.hu> wrote:
Subject: ocserv: GnuTLS error (at worker-vpn.c:585): Error in the system's
randomness device.
Package: ocserv
Version: 0.11.4-1+b1
Severity: important

Dear Maintainer,


I've started the ocserv service and tried to connect with an android phone
and later on with a linux machine.
Both times I recived:
GnuTLS error (at worker-vpn.c:585): Error in the system's randomness device.
On the client's side:
LIB: SSL negotiation with srv3.unnamedhost.somewhere
LIB: SSL connection failure: The TLS connection was non-properly terminated


Oct 06 15:19:19 srv3 systemd[1]: Started OpenConnect SSL VPN server.
Oct 06 15:19:19 srv3 ocserv[8425]: Setting 'radius' as primary
authentication method
Oct 06 15:19:19 srv3 ocserv[8425]: Setting 'radius' as accounting method
Oct 06 15:19:19 srv3 ocserv[8425]: Setting 'file' as supplemental config
option
Oct 06 15:19:19 srv3 ocserv[8425]: listening on 2 systemd sockets...
Oct 06 15:19:19 srv3 ocserv[8425]: main: initialized ocserv 0.11.4
Oct 06 15:19:19 srv3 ocserv[8438]: sec-mod: reading supplemental config from
files
Oct 06 15:19:19 srv3 ocserv[8438]: sec-mod: sec-mod initialized (socket:
/var/run/ocserv-socket.8425)
Oct 06 15:19:48 srv3 ocserv[8445]: GnuTLS error (at worker-vpn.c:585): Error
in the system's randomness device.
Oct 06 15:19:48 srv3 ocserv[8425]: main: [::ffff:192.168.31.230]:36872 user
disconnected (reason: unspecified, rx: 0, tx: 0)

也就是在 /etc/ocserv/ocserv.conf 里把 isolate-workers 改成false就行了(默认是true)

Q2: 为何改/etc/ocserv/ocserv.conf里的端口不行?
A2: 其实是因为忘记了一个东西,ocserv在安装时乃还能记得它会添加两个 systemd service 吗?对,另一个socket service改一下就好了,其实改不改也无所谓,不过我是有强迫症的,直接用网上某些方法,比如 sudo ocserv -c conf 启动显然不专业,这种方式启动也非常不优雅!不过看个人的吧,我把具体路径贴出来就好:
2016-10-15_191922

/lib/systemd/system/ocserv.socket

把端口改一下就好。

Q3: UFW启用之后的问题
A3:

ufw allow 443
ufw allow 443/udp

echo "*nat" >> /etc/ufw/before.rules
echo ":POSTROUTING ACCEPT [0:0]" >> /etc/ufw/before.rules
echo "-A POSTROUTING -s 10.88.0.0/24 -o eth0 -j MASQUERADE" >> /etc/ufw/before.rules
echo "COMMIT" >> /etc/ufw/before.rules
ufw disable && sudo ufw enable
anyShare分享到:

0 条评论

昵称

沙发空缺中,还不快抢~