rapid7/metasploit-framework

View on GitHub
docs/metasploit-framework.wiki/Common-Metasploit-Module-Coding-Mistakes.md

Summary

Maintainability
Test Coverage
# Deprecation notice!

Please see [CONTRIBUTING.md](https://github.com/rapid7/metasploit-framework/blob/master/CONTRIBUTING.md) for an authoritative coding guide. This document has fallen out of date. We don't write bad code any more! Hooray!

This is a collection of all the bad code we often see in Metasploit modules.  You should avoid them, too.

Note: Some of these examples use puts() for demo purposes, but you should always use print_status / print_error when writing a module.

### Bad Examples You Should NOT Follow:

1. Not checking the return value of a Metasploit API
2. Ruby 1.9.3 vs 1.8.7... gotcha!
3. Not checking the return value when using match()
4. Not checking nil before accessing a method
5. Using exception handling to shut an error up
6. Not taking advantage of the 'ensure' block
7. Adding the 'VERBOSE' option
8. Neglecting to use 'vars_post' for send_request_cgi() when crafting a POST request
9. Bad variable naming style
10. Using global variables
11. Modifying the datastore during execution

**1. Not checking the return value of a Metasploit API**

```ruby
res = send_request_cgi({
  'method' => 'GET',
  'uri' => '/app/index.php'
})

# There's a bug here, because res can return nil (due to a timeout or other reasons)
# If that happens, you will hit a "undefined method `code' for nil:NilClass" error.
# The correct way should be: if res && res.code == 200
if res.code == 200
  print_status("Response looks good")
else
  print_error("Unexpected response")
end
```

**2. Ruby 1.9.3 vs 1.8.7... gotcha!**

```ruby
some_string = "ABC"

# This can cause unexpected results to your module.
# Better to always do: char = some_string[1, 1]
char = some_string[1]

if char == 'B'
  puts "You will see this message in Ruby 1.9.3"
elsif char == 66
  puts "You will see this message in Ruby 1.8.7"
end
```

```ruby
# 1.9 allows a comma after the last argument when calling
# a method while 1.8 does not.  The most common place to
# see this error is in the update_info() section in a
# module's constructor.
some_method(
  "arg1",
  "arg2",  # <-- This comma is a syntax error on 1.8.x
)
```

**3. Not checking the return value when using match()**

```ruby
str = "dragon! drag on! Not lizard, I don't do that tongue thing"

# This tries to print "Not snake", but it's not in the string,
# so you'll get this error: "undefined method `[]' for nil:NilClass"
puts str.match(/(Not snake)/)[0]
```

```ruby
# The above is better written as:
if (str =~ /(Not snake)/)
  puts $1
end
```

**4. Not checking nil first before accessing a method**

```ruby
str = "These things are round and tasty, let's call them... tastycles!"

food = str.scan(/donut holes/)[0]

# food is nil, and nil has no method called "empty".
# This will throw an error: "undefined method `empty?' for nil:NilClass"
if food.empty? or food.nil?
  puts "I don't know what it's called"
end
```

**5. Using exception handling to shut an error up**

```ruby
begin
  # This block has 2 issues:
  # Issue #1: sample() is not a method in 1.8.7
  # Issue #2: Divided by 0 (race condition)
  n = [0, 1, 2, 3, 4, 5].sample
  1/n
rescue
  # If the user reports a bug saying this code isn't
  # working, it can be hard to debug exactly what went
  # wrong for the user without a backtrace.
  # When you do this, the error also won't be logged in
  # framework.log, either.
  # Note that rescuing ::Exception is especially harmful
  # because it can even hide syntax errors.
end
```

**6. Not taking advantage of the 'ensure' block**

```ruby
# You should use the ensure block to make sure x always has a value,
# which also avoids repeating code
begin
  n = [0, 1, 2].sample
  x = 1/n
rescue ZeroDivisionError => e
  puts "Are you smarter than a 5th grader? #{e.message}"
  x = 0  # Can put this in the ensure block
rescue NoMethodError
  puts "You must be using an older Ruby"
  x = 0 # Can put this in the ensure block
end

puts "Value is #{x.to_s}"
```

**7. Adding the 'VERBOSE' option**

```ruby
register_options(
  [
    # You already have this. Just type 'show advanced' and you'll see it.
    # So no need to register again
    OptBool.new("VERBOSE", [false, 'Enable detailed status messages', false])
  ], self.class)
```

**8. Neglecting to use send_request_cgi()'s vars_get or vars_get when crafting a POST/GET request**

```ruby
data_post = 'user=jsmith&pass=hello123'

# You should use the 'vars_post' key instead of 'data',
# unless you're trying to avoid the API escaping your
# parameter names
send_request_cgi({
  'method' => 'POST',
  'uri'    => '/',
  'data'   => data_post
})
```

**9. Bad variable naming style**

```ruby
# What's this, Java?
# The proper naming style in this case should be: my_string
myString = "hello, world"
```

**10. Using global variables**

```ruby
# $msg is a global variable that can be accessed anywhere within the program.
# This can induce bugs to other modules or mixins that are hard to debug.
# Use @instance variables instead.
# This is also mentioned in your HACKING file :-)

class Opinion
  def initialize
    # This variable shouldn't be shared with other classes
    $msg = "It's called the Freedom of Information Act. The Hippies finally got something right."
  end
end

class Metasploit3
  def initialize
    puts $msg
  end
end

Opinion.new
Metasploit3.new
```

**11. Modifying the datastore during execution**

```ruby
# https://github.com/rapid7/metasploit-framework/issues/3853
datastore['BAD'] = 'This is bad.'
```