#! /usr/bin/env python3
#
# This file is part of secnet.
# See README for full list of copyright holders.
#
# secnet is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 3 of the License, or
# (at your option) any later version.
# 
# secnet is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
# General Public License for more details.
# 
# You should have received a copy of the GNU General Public License
# version 3 along with secnet; if not, see
# https://www.gnu.org/licenses/gpl.html.

"""VPN sites file manipulation.

This program enables VPN site descriptions to be submitted for
inclusion in a central database, and allows the resulting database to
be turned into a secnet configuration file.

A database file can be turned into a secnet configuration file simply:
make-secnet-sites.py [infile [outfile]]

It would be wise to run secnet with the "--just-check-config" option
before installing the output on a live system.

The program expects to be invoked via userv to manage the database; it
relies on the USERV_USER and USERV_GROUP environment variables. The
command line arguments for this invocation are:

make-secnet-sites.py -u header-filename groupfiles-directory output-file \
  group

All but the last argument are expected to be set by userv; the 'group'
argument is provided by the user. A suitable userv configuration file
fragment is:

reset
no-disconnect-hup
no-suppress-args
cd ~/secnet/sites-test/
execute ~/secnet/make-secnet-sites.py -u vpnheader groupfiles sites

This program is part of secnet.

"""

from __future__ import print_function
from __future__ import unicode_literals
from builtins import int

import string
import time
import sys
import os
import getopt
import re
import argparse
import math

import ipaddress

# entry 0 is "near the executable", or maybe from PYTHONPATH=.,
# which we don't want to preempt
sys.path.insert(1,"/usr/local/share/secnet")
sys.path.insert(1,"/usr/share/secnet")
import ipaddrset

from argparseactionnoyes import ActionNoYes

VERSION="0.1.18"

from sys import version_info
if version_info.major == 2:  # for python2
    import codecs
    sys.stdin = codecs.getreader('utf-8')(sys.stdin)
    sys.stdout = codecs.getwriter('utf-8')(sys.stdout)
    import io
    open=lambda f,m='r': io.open(f,m,encoding='utf-8')

max={'rsa_bits':8200,'name':33,'dh_bits':8200}

class Tainted:
	def __init__(self,s,tline=None,tfile=None):
		self._s=s
		self._ok=None
		self._line=line if tline is None else tline
		self._file=file if tfile is None else tfile
	def __eq__(self,e):
		return self._s==e
	def __ne__(self,e):
		# for Python2
		return not self.__eq__(e)
	def __str__(self):
		raise RuntimeError('direct use of Tainted value')
	def __repr__(self):
		return 'Tainted(%s)' % repr(self._s)

	def _bad(self,what,why):
		assert(self._ok is not True)
		self._ok=False
		complain('bad parameter: %s: %s' % (what, why))
		return self

	def _max_ok(self,what,maxlen):
		if len(self._s) > maxlen:
			self._bad(what,'too long (max %d)' % maxlen)
		return self

	def _re_ok(self,bad,what,maxlen=None):
		if maxlen is None: maxlen=max[what]
		self._max_ok(what,maxlen)
		if self._ok is False: return self
		if bad.search(self._s): return self._bad(what,'bad syntax')
		return self

	def _rtnval(self, is_ok, ifgood, ifbad=''):
		if is_ok:
			assert(self._ok is not False)
			self._ok=True
			return ifgood
		else:
			assert(self._ok is not True)
			self._ok=False
			return ifbad

	def _rtn(self, is_ok, ifbad=''):
		return self._rtnval(is_ok, self._s, ifbad)

	def raw(self):
		return self._s
	def raw_mark_ok(self):
		# caller promises to throw if syntax was dangeorus
		return self._rtn(True)

	def output(self):
		if self._ok is False: return ''
		if self._ok is True: return self._s
		print('%s:%d: unchecked/unknown additional data "%s"' %
		      (self._file,self._line,self._s),
		      file=sys.stderr)
		sys.exit(1)

	bad_name=re.compile(r'^[^a-zA-Z]|[^-_0-9a-zA-Z]')
	# secnet accepts _ at start of names, but we reserve that
	bad_name_counter=0
	def name(self):
		ok=self._re_ok(Tainted.bad_name,'name')
		return self._rtn(ok,
				 '_line%d_%s' % (self._line, id(self)))

	def keyword(self):
		ok=self._s in keywords or self._s in levels
		if not ok:
			complain('unknown keyword %s' % self._s)
		return self._rtn(ok)

	bad_hex=re.compile(r'[^0-9a-fA-F]')
	def bignum_16(self,kind,what):
		maxlen=(max[kind+'_bits']+3)/4
		ok=self._re_ok(Tainted.bad_hex,what,maxlen)
		return self._rtn(ok)

	bad_num=re.compile(r'[^0-9]')
	def bignum_10(self,kind,what):
		maxlen=math.ceil(max[kind+'_bits'] / math.log10(2))
		ok=self._re_ok(Tainted.bad_num,what,maxlen)
		return self._rtn(ok)

	def number(self,minn,maxx,what='number'):
		# not for bignums
		ok=self._re_ok(Tainted.bad_num,what,10)
		if ok:
			v=int(self._s)
			if v<minn or v>maxx:
				ok=self._bad(what,'out of range %d..%d'
					     % (minn,maxx))
		return self._rtnval(ok,v,minn)

	bad_host=re.compile(r'[^-\][_.:0-9a-zA-Z]')
	# We permit _ so we can refer to special non-host domains
	# which have A and AAAA RRs.  This is a crude check and we may
	# still produce config files with syntactically invalid
	# domains or addresses, but that is OK.
	def host(self):
		ok=self._re_ok(Tainted.bad_host,'host/address',255)
		return self._rtn(ok)

	bad_email=re.compile(r'[^-._0-9a-z@!$%^&*=+~/]')
	# ^ This does not accept all valid email addresses.  That's
	# not really possible with this input syntax.  It accepts
	# all ones that don't require quoting anywhere in email
	# protocols (and also accepts some invalid ones).
	def email(self):
		ok=self._re_ok(Tainted.bad_email,'email address',1023)
		return self._rtn(ok)

	bad_groupname=re.compile(r'^[^_A-Za-z]|[^-+_0-9A-Za-z]')
	def groupname(self):
		ok=self._re_ok(Tainted.bad_groupname,'group name',64)
		return self._rtn(ok)

def parse_args():
	global service
	global inputfile
	global header
	global groupfiledir
	global sitesfile
	global group
	global user
	global of
	global prefix
	global key_prefix

	ap = argparse.ArgumentParser(description='process secnet sites files')
	ap.add_argument('--userv', '-u', action='store_true',
			help='userv service fragment update mode')
	ap.add_argument('--conf-key-prefix', action=ActionNoYes,
			default=True,
		 help='prefix conf file key names derived from sites data')
	ap.add_argument('--prefix', '-P', nargs=1,
			help='set prefix')
	ap.add_argument('arg',nargs=argparse.REMAINDER)
	av = ap.parse_args()
	#print(repr(av), file=sys.stderr)
	service = 1 if av.userv else 0
	prefix = '' if av.prefix is None else av.prefix[0]
	key_prefix = av.conf_key_prefix
	if service:
		if len(av.arg)!=4:
			print("Wrong number of arguments")
			sys.exit(1)
		(header, groupfiledir, sitesfile, group) = av.arg
		group = Tainted(group,0,'command line')
		# untrusted argument from caller
		if "USERV_USER" not in os.environ:
			print("Environment variable USERV_USER not found")
			sys.exit(1)
		user=os.environ["USERV_USER"]
		# Check that group is in USERV_GROUP
		if "USERV_GROUP" not in os.environ:
			print("Environment variable USERV_GROUP not found")
			sys.exit(1)
		ugs=os.environ["USERV_GROUP"]
		ok=0
		for i in ugs.split():
			if group==i: ok=1
		if not ok:
			print("caller not in group %s"%group)
			sys.exit(1)
	else:
		if len(av.arg)>3:
			print("Too many arguments")
			sys.exit(1)
		(inputfile, outputfile) = (av.arg + [None]*2)[0:2]
		if outputfile is None: of=sys.stdout
		else: of=open(outputfile,'w')

parse_args()

# Classes describing possible datatypes in the configuration file

class basetype:
	"Common protocol for configuration types."
	def add(self,obj,w):
		complain("%s %s already has property %s defined"%
			(obj.type,obj.name,w[0].raw()))

class conflist:
	"A list of some kind of configuration type."
	def __init__(self,subtype,w):
		self.subtype=subtype
		self.list=[subtype(w)]
	def add(self,obj,w):
		self.list.append(self.subtype(w))
	def __str__(self):
		return ', '.join(map(str, self.list))
def listof(subtype):
	return lambda w: conflist(subtype, w)

class single_ipaddr (basetype):
	"An IP address"
	def __init__(self,w):
		self.addr=ipaddress.ip_address(w[1].raw_mark_ok())
	def __str__(self):
		return '"%s"'%self.addr

class networks (basetype):
	"A set of IP addresses specified as a list of networks"
	def __init__(self,w):
		self.set=ipaddrset.IPAddressSet()
		for i in w[1:]:
			x=ipaddress.ip_network(i.raw_mark_ok(),strict=True)
			self.set.append([x])
	def __str__(self):
		return ",".join(map((lambda n: '"%s"'%n), self.set.networks()))

class dhgroup (basetype):
	"A Diffie-Hellman group"
	def __init__(self,w):
		self.mod=w[1].bignum_16('dh','dh mod')
		self.gen=w[2].bignum_16('dh','dh gen')
	def __str__(self):
		return 'diffie-hellman("%s","%s")'%(self.mod,self.gen)

class hash (basetype):
	"A choice of hash function"
	def __init__(self,w):
		hname=w[1]
		self.ht=hname.raw()
		if (self.ht!='md5' and self.ht!='sha1'):
			complain("unknown hash type %s"%(self.ht))
			self.ht=None
		else:
			hname.raw_mark_ok()
	def __str__(self):
		return '%s'%(self.ht)

class email (basetype):
	"An email address"
	def __init__(self,w):
		self.addr=w[1].email()
	def __str__(self):
		return '<%s>'%(self.addr)

class boolean (basetype):
	"A boolean"
	def __init__(self,w):
		v=w[1]
		if re.match('[TtYy1]',v.raw()):
			self.b=True
			v.raw_mark_ok()
		elif re.match('[FfNn0]',v.raw()):
			self.b=False
			v.raw_mark_ok()
		else:
			complain("invalid boolean value");
	def __str__(self):
		return ['False','True'][self.b]

class num (basetype):
	"A decimal number"
	def __init__(self,w):
		self.n=w[1].number(0,0x7fffffff)
	def __str__(self):
		return '%d'%(self.n)

class address (basetype):
	"A DNS name and UDP port number"
	def __init__(self,w):
		self.adr=w[1].host()
		self.port=w[2].number(1,65536,'port')
	def __str__(self):
		return '"%s"; port %d'%(self.adr,self.port)

class rsakey (basetype):
	"An RSA public key"
	def __init__(self,w):
		self.l=w[1].number(0,max['rsa_bits'],'rsa len')
		self.e=w[2].bignum_10('rsa','rsa e')
		self.n=w[3].bignum_10('rsa','rsa n')
		if len(w) >= 5: w[4].email()
	def __str__(self):
		return 'rsa-public("%s","%s")'%(self.e,self.n)

# Possible properties of configuration nodes
keywords={
 'contact':(email,"Contact address"),
 'dh':(dhgroup,"Diffie-Hellman group"),
 'hash':(hash,"Hash function"),
 'key-lifetime':(num,"Maximum key lifetime (ms)"),
 'setup-timeout':(num,"Key setup timeout (ms)"),
 'setup-retries':(num,"Maximum key setup packet retries"),
 'wait-time':(num,"Time to wait after unsuccessful key setup (ms)"),
 'renegotiate-time':(num,"Time after key setup to begin renegotiation (ms)"),
 'restrict-nets':(networks,"Allowable networks"),
 'networks':(networks,"Claimed networks"),
 'pubkey':(rsakey,"RSA public site key"),
 'peer':(single_ipaddr,"Tunnel peer IP address"),
 'address':(address,"External contact address and port"),
 'mobile':(boolean,"Site is mobile"),
}

def sp(name,value):
	"Simply output a property - the default case"
	return "%s %s;\n"%(name,value)

# All levels support these properties
global_properties={
	'contact':(lambda name,value:"# Contact email address: %s\n"%(value)),
	'dh':sp,
	'hash':sp,
	'key-lifetime':sp,
	'setup-timeout':sp,
	'setup-retries':sp,
	'wait-time':sp,
	'renegotiate-time':sp,
	'restrict-nets':(lambda name,value:"# restrict-nets %s\n"%value),
}

class level:
	"A level in the configuration hierarchy"
	depth=0
	leaf=0
	allow_properties={}
	require_properties={}
	def __init__(self,w):
		self.type=w[0].keyword()
		self.name=w[1].name()
		self.properties={}
		self.children={}
	def indent(self,w,t):
		w.write("                 "[:t])
	def prop_out(self,n):
		return self.allow_properties[n](n,str(self.properties[n]))
	def output_props(self,w,ind):
		for i in sorted(self.properties.keys()):
			if self.allow_properties[i]:
				self.indent(w,ind)
				w.write("%s"%self.prop_out(i))
	def kname(self):
		return ((self.type[0].upper() if key_prefix else '')
			+ self.name)
	def output_data(self,w,path):
		ind = 2*len(path)
		self.indent(w,ind)
		w.write("%s {\n"%(self.kname()))
		self.output_props(w,ind+2)
		if self.depth==1: w.write("\n");
		for k in sorted(self.children.keys()):
			c=self.children[k]
			c.output_data(w,path+(c,))
		self.indent(w,ind)
		w.write("};\n")

class vpnlevel(level):
	"VPN level in the configuration hierarchy"
	depth=1
	leaf=0
	type="vpn"
	allow_properties=global_properties.copy()
	require_properties={
	 'contact':"VPN admin contact address"
	}
	def __init__(self,w):
		level.__init__(self,w)
	def output_vpnflat(self,w,path):
		"Output flattened list of site names for this VPN"
		ind=2*(len(path)+1)
		self.indent(w,ind)
		w.write("%s {\n"%(self.kname()))
		for i in self.children.keys():
			self.children[i].output_vpnflat(w,path+(self,))
		w.write("\n")
		self.indent(w,ind+2)
		w.write("all-sites %s;\n"%
			','.join(map(lambda i: i.kname(),
				     self.children.values())))
		self.indent(w,ind)
		w.write("};\n")

class locationlevel(level):
	"Location level in the configuration hierarchy"
	depth=2
	leaf=0
	type="location"
	allow_properties=global_properties.copy()
	require_properties={
	 'contact':"Location admin contact address",
	}
	def __init__(self,w):
		level.__init__(self,w)
		self.group=w[2].groupname()
	def output_vpnflat(self,w,path):
		ind=2*(len(path)+1)
		self.indent(w,ind)
		# The "path=path,self=self" abomination below exists because
		# Python didn't support nested_scopes until version 2.1
		#
		#"/"+self.name+"/"+i
		w.write("%s %s;\n"%(self.kname(),','.join(
			map(lambda x,path=path,self=self:
			    '/'.join([prefix+"vpn-data"] + list(map(
				    lambda i: i.kname(),
				    path+(self,x)))),
			    self.children.values()))))

class sitelevel(level):
	"Site level (i.e. a leafnode) in the configuration hierarchy"
	depth=3
	leaf=1
	type="site"
	allow_properties=global_properties.copy()
	allow_properties.update({
	 'address':sp,
	 'networks':None,
	 'peer':None,
	 'pubkey':(lambda n,v:"key %s;\n"%v),
	 'mobile':sp,
	})
	require_properties={
	 'dh':"Diffie-Hellman group",
	 'contact':"Site admin contact address",
	 'networks':"Networks claimed by the site",
	 'hash':"hash function",
	 'peer':"Gateway address of the site",
	 'pubkey':"RSA public key of the site",
	}
	def __init__(self,w):
		level.__init__(self,w)
	def output_data(self,w,path):
		ind=2*len(path)
		np='/'.join(map(lambda i: i.name, path))
		self.indent(w,ind)
		w.write("%s {\n"%(self.kname()))
		self.indent(w,ind+2)
		w.write("name \"%s\";\n"%(np,))
		self.output_props(w,ind+2)
		self.indent(w,ind+2)
		w.write("link netlink {\n");
		self.indent(w,ind+4)
		w.write("routes %s;\n"%str(self.properties["networks"]))
		self.indent(w,ind+4)
		w.write("ptp-address %s;\n"%str(self.properties["peer"]))
		self.indent(w,ind+2)
		w.write("};\n")
		self.indent(w,ind)
		w.write("};\n")

# Levels in the configuration file
# (depth,properties)
levels={'vpn':vpnlevel, 'location':locationlevel, 'site':sitelevel}

def complain(msg):
	"Complain about a particular input line"
	global complaints
	print(("%s line %d: "%(file,line))+msg)
	complaints=complaints+1
def moan(msg):
	"Complain about something in general"
	global complaints
	print(msg);
	complaints=complaints+1

class UntaintedRoot():
	def __init__(self,s): self._s=s
	def name(self): return self._s
	def keyword(self): return self._s

root=level([UntaintedRoot(x) for x in ['root','root']])
# All vpns are children of this node
obstack=[root]
allow_defs=0   # Level above which new definitions are permitted

def set_property(obj,w):
	"Set a property on a configuration node"
	prop=w[0]
	if prop.raw() in obj.properties:
		obj.properties[prop.raw_mark_ok()].add(obj,w)
	else:
		obj.properties[prop.raw()]=keywords[prop.raw_mark_ok()][0](w)


def pline(il,allow_include=False):
	"Process a configuration file line"
	global allow_defs, obstack, root
	w=il.rstrip('\n').split()
	if len(w)==0: return ['']
	w=list([Tainted(x) for x in w])
	keyword=w[0]
	current=obstack[len(obstack)-1]
	copyout=lambda: ['    '*len(obstack) +
			' '.join([ww.output() for ww in w]) +
	                '\n']
	if keyword=='end-definitions':
		keyword.raw_mark_ok()
		allow_defs=sitelevel.depth
		obstack=[root]
		return copyout()
	if keyword=='include':
		if not allow_include:
			complain("include not permitted here")
			return []
		if len(w) != 2:
			complain("include requires one argument")
			return []
		newfile=os.path.join(os.path.dirname(file),w[1].raw_mark_ok())
		# ^ user of "include" is trusted so raw_mark_ok is good
		return pfilepath(newfile,allow_include=allow_include)
	if keyword.raw() in levels:
		# We may go up any number of levels, but only down by one
		newdepth=levels[keyword.raw_mark_ok()].depth
		currentdepth=len(obstack) # actually +1...
		if newdepth<=currentdepth:
			obstack=obstack[:newdepth]
		if newdepth>currentdepth:
			complain("May not go from level %d to level %d"%
				(currentdepth-1,newdepth))
		# See if it's a new one (and whether that's permitted)
		# or an existing one
		current=obstack[len(obstack)-1]
		tname=w[1].name()
		if tname in current.children:
			# Not new
			current=current.children[tname]
			if service and group and current.depth==2:
				if group!=current.group:
					complain("Incorrect group!")
				w[2].groupname()
		else:
			# New
			# Ignore depth check for now
			nl=levels[keyword.raw()](w)
			if nl.depth<allow_defs:
				complain("New definitions not allowed at "
					"level %d"%nl.depth)
				# we risk crashing if we continue
				sys.exit(1)
			current.children[tname]=nl
			current=nl
		obstack.append(current)
		return copyout()
	if keyword.raw() not in current.allow_properties:
		complain("Property %s not allowed at %s level"%
			(keyword.raw(),current.type))
		return []
	elif current.depth == vpnlevel.depth < allow_defs:
		complain("Not allowed to set VPN properties here")
		return []
	else:
		set_property(current,w)
		return copyout()

	complain("unknown keyword '%s'"%(keyword.raw()))

def pfilepath(pathname,allow_include=False):
	f=open(pathname)
	outlines=pfile(pathname,f.readlines(),allow_include=allow_include)
	f.close()
	return outlines

def pfile(name,lines,allow_include=False):
	"Process a file"
	global file,line
	file=name
	line=0
	outlines=[]
	for i in lines:
		line=line+1
		if (i[0]=='#'): continue
		outlines += pline(i,allow_include=allow_include)
	return outlines

def outputsites(w):
	"Output include file for secnet configuration"
	w.write("# secnet sites file autogenerated by make-secnet-sites "
		+"version %s\n"%VERSION)
	w.write("# %s\n"%time.asctime(time.localtime(time.time())))
	w.write("# Command line: %s\n\n"%' '.join(sys.argv))

	# Raw VPN data section of file
	w.write(prefix+"vpn-data {\n")
	for i in root.children.values():
		i.output_data(w,(i,))
	w.write("};\n")

	# Per-VPN flattened lists
	w.write(prefix+"vpn {\n")
	for i in root.children.values():
		i.output_vpnflat(w,())
	w.write("};\n")

	# Flattened list of sites
	w.write(prefix+"all-sites %s;\n"%",".join(
		map(lambda x:"%svpn/%s/all-sites"%(prefix,x.kname()),
			root.children.values())))

line=0
file=None
complaints=0

# Sanity check section
# Delete nodes where leaf=0 that have no children

def live(n):
	"Number of leafnodes below node n"
	if n.leaf: return 1
	for i in n.children.keys():
		if live(n.children[i]): return 1
	return 0
def delempty(n):
	"Delete nodes that have no leafnode children"
	for i in list(n.children.keys()):
		delempty(n.children[i])
		if not live(n.children[i]):
			del n.children[i]

# Check that all constraints are met (as far as I can tell
# restrict-nets/networks/peer are the only special cases)

def checkconstraints(n,p,ra):
	new_p=p.copy()
	new_p.update(n.properties)
	for i in n.require_properties.keys():
		if i not in new_p:
			moan("%s %s is missing property %s"%
				(n.type,n.name,i))
	for i in new_p.keys():
		if i not in n.allow_properties:
			moan("%s %s has forbidden property %s"%
				(n.type,n.name,i))
	# Check address range restrictions
	if "restrict-nets" in n.properties:
		new_ra=ra.intersection(n.properties["restrict-nets"].set)
	else:
		new_ra=ra
	if "networks" in n.properties:
		if not n.properties["networks"].set <= new_ra:
			moan("%s %s networks out of bounds"%(n.type,n.name))
		if "peer" in n.properties:
			if not n.properties["networks"].set.contains(
				n.properties["peer"].addr):
				moan("%s %s peer not in networks"%(n.type,n.name))
	for i in n.children.keys():
		checkconstraints(n.children[i],new_p,new_ra)

if service:
	headerinput=pfilepath(header,allow_include=True)
	userinput=sys.stdin.readlines()
	pfile("user input",userinput)
else:
	if inputfile is None:
		pfile("stdin",sys.stdin.readlines())
	else:
		pfilepath(inputfile)

delempty(root)
checkconstraints(root,{},ipaddrset.complete_set())

if complaints>0:
	if complaints==1: print("There was 1 problem.")
	else: print("There were %d problems."%(complaints))
	sys.exit(1)
complaints=None # arranges to crash if we complain later

if service:
	# Put the user's input into their group file, and rebuild the main
	# sites file
	f=open(groupfiledir+"/T"+group.groupname(),'w')
	f.write("# Section submitted by user %s, %s\n"%
		(user,time.asctime(time.localtime(time.time()))))
	f.write("# Checked by make-secnet-sites version %s\n\n"%VERSION)
	for i in userinput: f.write(i)
	f.write("\n")
	f.close()
	os.rename(groupfiledir+"/T"+group.groupname(),
		  groupfiledir+"/R"+group.groupname())
	f=open(sitesfile+"-tmp",'w')
	f.write("# sites file autogenerated by make-secnet-sites\n")
	f.write("# generated %s, invoked by %s\n"%
		(time.asctime(time.localtime(time.time())),user))
	f.write("# use make-secnet-sites to turn this file into a\n")
	f.write("# valid /etc/secnet/sites.conf file\n\n")
	for i in headerinput: f.write(i)
	files=os.listdir(groupfiledir)
	for i in files:
		if i[0]=='R':
			j=open(groupfiledir+"/"+i)
			f.write(j.read())
			j.close()
	f.write("# end of sites file\n")
	f.close()
	os.rename(sitesfile+"-tmp",sitesfile)
else:
	outputsites(of)