lcmlog-server/lcmlog-server

#!/usr/bin/env python3

DIR         = "/var/local/log/lcmlog-data"
FROM_DOMAIN = "lcm.mi.infn.it"
TO_ADDRESS  = "working@lcm.mi.infn.it"
REPLY_TO    = True              # Does this make sense?

import os
import os.path
#from os import stat
#from pwd import getpwuid
import sys
import shlex
import pwd
import logging
import logging.handlers
import hashlib
import contextlib
import toml
import subprocess


# We log what happens every time someone connects
# Preparing the logger
logger = logging.getLogger(__name__)
file_formatter = logging.Formatter("%(asctime)s | %(levelname)8s | %(message)s")
logger.setLevel(logging.INFO)

# Logger handle to log all info
# We use a TimedRotatingFileHandler to rotate logs once a week
file_handler = logging.handlers.TimedRotatingFileHandler(filename = DIR + "/logs/logfile", when = "W6", backupCount = 10)
file_handler.setFormatter(file_formatter)
logger.addHandler(file_handler)

# Update logfile acl
#if pwd.getpwuid(os.stat(DIR + "/logs/logfile").st_uid).pw_name == pwd.getpwuid(os.geteuid()).pw_name:
#    subprocess.call(["touch", DIR + "/logs/logfile"])
#    #subprocess.call(["chmod", "444", DIR + "/logs/*"])
#    subprocess.call(["chmod", "666", DIR + "/logs/logfile"])


#------------------------------------------------------------------------------


def main():

	# The user is going to call us through ssh, so to know who he is we can simply get his effective uid
	user_id = os.geteuid()
	user_name = pwd.getpwuid(user_id).pw_name

	logger.info("Started by user " + user_name + " (id " + str(user_id) + ")")

	try:
		method = input() # Can be GET, POST or UPDATE
		logger.info("Method: " + method)
		if method == "UPDATE":	# We don't need more input lines for the UPDATE method
			auth(user_id, "UPDATE", "")	# Check if the user can update the database
			method_update()
		else:
			kind = input()	# This is the kind of the log, and it can be 150 or Admin
			logger.info("Kind: " + kind)
			# Now date check is implemented in the client side script
			# Maybe, for the future, date control can be implemented aslo here
			date = input()	# The date of the log
			logger.info("Date: " + date)
			tags = input()	# Tags are comma separated
			logger.info("Tags: " + tags)

			if kind != "150" and kind != "Admin":	# We only have this two log types
				raise KindError
			if method == "POST":
				auth(user_id, "POST", kind)     # Check if the user can post for the requested kind
				log = sys.stdin.read() # Read the log content
				method_post(kind, user_name, date, tags, log)
			elif method == "GET":
				auth(user_id, "GET", kind)	# Check if the user can get logs for the requested kind
				user_to_find = input() # Read the user name of the log writer to search
				logger.info("User to find: " + user_to_find)
				method_get(kind, user_to_find, date, tags)
			else:
				raise MethodError

	except EOFError as error:
		logger.critical("1 Not enough input lines")
		sys.exit(1)
	except FileNotFoundError as error:
		logger.critical("2 File not found")
		sys.exit(2)
	except FileExistsError as error:
		logger.critical("3 File already exists")
		sys.exit(3)
	except OSError as error:
		logger.critical("4 File error")
		sys.exit(4)
	except MethodError as error:
		logger.critical("5 Undefined method")
		sys.exit(5)
	except KindError as error:
		logger.critical("6 Undefined log kind")
		sys.exit(6)
	except AuthError as error:
		logger.critical("7 Authentication error")
		sys.exit(7)
	except Exception as error:
		logger.critical("8 Generic error: " + str(error))
		sys.exit(8)

	finally:
		logger.info("End\n")


#------------------------------------------------------------------------------


# Create new log file and adds it to the database
# kind, user_name, date and tags is the log metadata
# log is the log content
# The log metadata and content is hashed, and the hash is saved in the database and used as the filename for the log
# The return value of the function is the hash
def log_create(kind, user_name, date, tags, log):
	name = hashlib.sha512((kind + user_name + date + tags + log).encode("utf-8")).hexdigest()
	with open(DIR + "/data/" + name, "x") as f:
		f.write(name + "\n" + kind + "\n" + user_name + "\n" + date + "\n" + tags + "\n" + log)	# Write the file
	with open(DIR + "/data/.data", "a") as f:
		f.write(name + ":" + kind + ":" + user_name + ":" + date + ":" + tags + "\n")	# And add the entry to the .data file
	return name


# Search for the requested entry
# The functions returns a list containing the hash (saved in the database) of all the files that meet the specified criteria
# The kind parameter is mandatory (because different users have different privileges based on it).
# All the other arguments can be empty. Only the arguments that are not empty are taken into consideration for the search
def log_find(kind, user_name, date, tags):
	file_list = list()
	with open(DIR + "/data/.data", "r") as f:
		for line in f:
			found = True
			l = line.split(":")
			# The kind is different
			if l[1].find(kind) == -1:
				continue
			# The username is different, or we aren't searching by username
			if user_name and l[2].find(user_name) == -1:
				continue
			# The date is different, or we aren't searching by date
			if date and l[3].find(date) == -1:
				continue
			# Searh tags
			for t in tags.split(","):
				if t and l[4].find(t) == -1:
					found = False
					break
			# Save
			if found:
				file_list.append(l[0])
	return file_list


# TODO: the following functions work with the hash of the log files. The problem is that there are three different places where the hash is: the first line of the file,
# the database entry for the log and the filename of the log. I have to decide which function operates on which hash, because for example if the hash is changed in the file,
# it needs to be changed also in the other two locations.
# Add log file to .data
# This function reads an existing log file and adds it to the database
# Tha hash that is saved in the database is not calculated: the first line in the file is considered to be the hash. Use log_check to check if they are the same
def log_add(name):
	with open(DIR + "/data/" + name, "r") as f, open(DIR + "/data/.data", "a") as data:
		data.write(f.readline().rstrip("\n") + ":" +	# Hash
			   f.readline().rstrip("\n") + ":" +	# Kind
			   f.readline().rstrip("\n") + ":" +	# User name
			   f.readline().rstrip("\n") + ":" +	# Date
			   f.readline().rstrip("\n") + "\n") 	# Tags


# Check if the saved hash is correct, and if it is not, ask the user what to do
# This function calculates the hash of the file with filename name, and returns True if it is the same as the first line of the file, False otherwise
# If it doesn't correspond, it asks the user if he wants to keep it like it is or change it. Currently, it is pretty messed up: only the hash saved in the file is changed,
# not the one saved in the database or the file name. Also, the dialog to ask if the hash is to be changed probably should not be in this function.
def log_check(name):
	with open(DIR + "/data/" + name, "r") as f:
		saved_hash = f.readline().rstrip("\n")
		kind = f.readline().rstrip("\n")
		user = f.readline().rstrip("\n")
		date = f.readline().rstrip("\n")
		tags = f.readline().rstrip("\n")
		log = f.read()
	calc_hash = log_hash(name)
	result = saved_hash == calc_hash
	if not result:
		logger.warning(calc_hash + " hash does not correspond to saved one")
		while True:
			print("Warning: " + calc_hash + " sh does not correspond to saved one.\n" +
				"Do you want to: print the log (p), change the saved hash (c), or leave it as it is (l)?")
			c = input()
			if c == "p":
				sys.stdout.write("Hash: " + saved_hash + "\n")
				sys.stdout.write("Kind: " + kind + "\n")
				sys.stdout.write("User: " + user + "\n")
				sys.stdout.write("Date: " + date + "\n")
				sys.stdout.write("Tags: " + tags + "\n")
				sys.stdout.write("\n" + log + "\n")
			elif c == "l":
				logger.info("Hash unchanged")
				break
			elif c == "c":
				with open(DIR + "/data/" + name, "w") as f:
					f.write(calc_hash + "\n" + kind + "\n" + user + "\n" + date + "\n" + tags + "\n" + log)
				logger.info("Hash changed")
				break
	return result


# Calculates hash of file
# The first line of the file is the saved hash, therefore it is not considered in the calculation
def log_hash(name):
	with open(DIR + "/data/" + name, "r") as f:
		f.readline()	# The saved hash doesn't enter in the calculation
		kind = f.readline().rstrip("\n")
		user_name = f.readline().rstrip("\n")
		date = f.readline().rstrip("\n")
		tags = f.readline().rstrip("\n")
		log = f.read()
		return hashlib.sha512((kind + user_name + date + tags + log).encode("utf-8")).hexdigest()


# Email the log contents from USER_NAME@FROM_DOMAIN to TO_ADDRESS
# Emails the log contents using the 'mail(1)' program. If REPLY_TO is
# True add the 'Reply-to: ' header to the email
def log_mail(kind, user_name, date, tags, log):
	if REPLY_TO:
                subprocess.run("printf " + shlex.quote(log) + " | mail -s \"Log" + kind + " " + date + " " + tags +
                               "\" -r " + user + FROM_DOMAIN + " -a \"Reply-to:" TO_ADDRESS + " \" -- "
		               + TO_ADDRESS, shell=True)
	else:
		subprocess.run("printf \"%s\n\" " + shlex.quote(log) + " | mail -s \"Log" | kind + " " + date + " " + tags +
		               "\" -r " + user + FROM_DOMAIN + " -- " + TO_ADDRESS, shell=True)
	return


#------------------------------------------------------------------------------


# Print specified log on stdout
def method_get(kind, user_to_find, date, tags):
	file_list = log_find(kind, user_to_find, date, tags)
	for name in file_list:
		with open(DIR + "/data/" + name, "r") as f:
			sys.stdout.write("Hash: " + f.readline())
			sys.stdout.write("Kind: " + f.readline())
			sys.stdout.write("User: " + f.readline())
			sys.stdout.write("Date: " + f.readline())
			sys.stdout.write("Tags: " + f.readline())
			sys.stdout.write("\n" + f.read() + "------------------\n")
	logger.info("GET successful: got " + str(len(file_list)) + " files")


# Write log
def method_post(kind, user_name, date, tags, log):
	name = log_create(kind, user_name, date, tags, log)
	logger.info("POST successful: hash " + name)


# Generate .data file
def method_update():
	with contextlib.suppress(FileNotFoundError):
	    os.remove(DIR + "/data/.data")
	file_list = os.listdir(DIR + "/data/")
	open(DIR + "/data/.data", "x").close()
	for name in file_list:
            newname = log_hash(name)
            if not log_check(name):
                os.rename(DIR + "/data/" + name, DIR + "/data/" + newname)
            log_add(newname)
	logger.info("UPDATE successful: added " + str(len(file_list)) + " files")


#------------------------------------------------------------------------------


# Checks if the user has the permissions to use the requested method
def auth(user_id, method, kind):
	# Check if user id is in auth files
	# We suppose that every authorized user is ONLY IN A FILE!
	user_type = ""
	for file_name in ["150","Admin","Valhalla","Nirvana"]:
		with open(DIR + "/auth/" + file_name) as f:
			for line in f:
				line = line.split()[0]
				if int(line) == user_id:
					# If present, we consider only the user type
					user_type = file_name
					break
			if not user_type == "":
				break
	else:
		# If not in auth files, the user cannot do anything
		raise AuthError()
	# Now we check the user type permissions
	auth_list = toml.load(DIR + "/auth/auth.toml")[user_type]["auth"]
	if not method + " " + kind in auth_list:
		raise AuthError()
	return


#------------------------------------------------------------------------------


# Error definitions
class AuthError(Exception):
	pass
class KindError(Exception):
	pass
class MethodError(Exception):
	pass


#------------------------------------------------------------------------------


# Starting point
if __name__ == "__main__":
	main()
	# Change permissions to logfile just before leaving. Dirty fix to a not well understood problem
	subprocess.call(["chmod", "666", DIR + "/logs/logfile"])