9

I'm searching for a php algorithm that efficiently test if one cidr notated network overlaps another.

Basically I have the following situation:

Array of cidr adresses:

$cidrNetworks = array(
    '192.168.10.0/24',
    '10.10.0.30/20',
    etc.
);

I have a method that adds networks to the array, but this method should throw an exception when a network is added that overlaps with a network allready in the array.

So ie. if 192.168.10.0/25 is added an exception should be thrown.

Does anyone have/know/"can think of" an method to test this efficiently?

Damien Overeem
  • 4,487
  • 4
  • 36
  • 55

4 Answers4

18

Here is an updated version of the class previously discussed in chat. It can do what you require, as well as many other useful things.

<?php

    class IPv4Subnet implements ArrayAccess, Iterator {

        /*
         * Address format constants
         */
        const ADDRESS_BINARY = 0x01;
        const ADDRESS_INT = 0x02;
        const ADDRESS_DOTDEC = 0x04;
        const ADDRESS_SUBNET = 0x08;

        /*
         * Constants to control whether getHosts() returns the network/broadcast addresses
         */
        const HOSTS_WITH_NETWORK = 0x10;
        const HOSTS_WITH_BROADCAST = 0x20;
        const HOSTS_ALL = 0x30;

        /*
         * Properties to store base address and subnet mask as binary strings
         */
        protected $address;
        protected $mask;

        /*
         * Counter to track the current iteration offset
         */
        private $iteratorOffset = 0;

        /*
         * Array to hold values retrieved via ArrayAccess
         */
        private $arrayAccessObjects = array();

        /*
         * Helper methods
         */
        private function longToBinary ($long) {
            return pack('N', $long);
        }
        private function longToDottedDecimal ($long) {
            return ($long >> 24 & 0xFF).'.'.($long >> 16 & 0xFF).'.'.($long >> 8 & 0xFF).'.'.($long & 0xFF);
        }
        private function longToByteArray ($long) {
            return array(
                $long >> 24 & 0xFF,
                $long >> 16 & 0xFF,
                $long >> 8 & 0xFF,
                $long & 0xFF
            );
        }
        private function longToSubnet ($long) {
            if (!isset($this->arrayAccessObjects[$long])) {
                $this->arrayAccessObjects[$long] = new self($long);
            }
            return $this->arrayAccessObjects[$long];
        }
        private function binaryToLong ($binary) {
            return current(unpack('N', $binary));
        }
        private function binaryToDottedDecimal ($binary) {
            return implode('.', unpack('C*', $binary));
        }
        private function binaryToX ($binary, $mode) {
            if ($mode & self::ADDRESS_BINARY) {
                $result = $binary;
            } else if ($mode & self::ADDRESS_INT) {
                $result = $this->binaryToLong($binary);
            } else if ($mode & self::ADDRESS_DOTDEC) {
                $result = $this->binaryToDottedDecimal($binary);
            } else {
                $result = $this->longToSubnet($this->binaryToLong($binary));
            }
            return $result;
        }
        private function byteArrayToLong($bytes) {
            return ($bytes[0] << 24) | ($bytes[1] << 16) | ($bytes[2] << 8) | $bytes[3];
        }
        private function byteArrayToBinary($bytes) {
            return pack('C*', $bytes[0], $bytes[1], $bytes[2], $bytes[3]);
        }

        private function normaliseComparisonSubject (&$subject) {
            if (!is_object($subject)) {
                $subject = new self($subject);
            }
            if (!($subject instanceof self)) {
                throw new InvalidArgumentException('Subject must be an instance of IPv4Subnet');
            }
        }

        private function validateOctetArray (&$octets) {
            foreach ($octets as &$octet) {
                $octet = (int) $octet;
                if ($octet < 0 || $octet > 255) {
                    return FALSE;
                }
            }
            return TRUE;
        }

        /*
         * Constructor
         */
        public function __construct ($address = NULL, $mask = NULL) {
            if ($address === NULL || (is_string($address) && trim($address) === '')) {
                $address = array(0, 0, 0, 0);
            } else if (is_int($address)) {
                $address = $this->longToByteArray($address);
            } else if (is_string($address)) {
                $parts = preg_split('#\s*/\s*#', trim($address), -1, PREG_SPLIT_NO_EMPTY);
                if (count($parts) > 2) {
                    throw new InvalidArgumentException('No usable IP address supplied: Syntax error');
                } else if ($parts[0] === '') {
                    throw new InvalidArgumentException('No usable IP address supplied: IP address empty');
                }
                if (!empty($parts[1]) && !isset($mask)) {
                    $mask = $parts[1];
                }
                $address = preg_split('#\s*\.\s*#', $parts[0], -1, PREG_SPLIT_NO_EMPTY);
            } else if (is_array($address)) {
                $address = array_values($address);
            } else {
                throw new InvalidArgumentException('No usable IP address supplied: Value must be a string or an integer');
            }

            $suppliedAddressOctets = count($address);
            $address += array(0, 0, 0, 0);
            if ($suppliedAddressOctets > 4) {
                throw new InvalidArgumentException('No usable IP address supplied: IP address has more than 4 octets');
            } else if (!$this->validateOctetArray($address)) {
                throw new InvalidArgumentException('No usable IP address supplied: At least one octet value outside acceptable range 0 - 255');
            }

            if ($mask === NULL) {
                $mask = array_pad(array(), $suppliedAddressOctets, 255) + array(0, 0, 0, 0);
            } else if (is_int($mask)) {
                $mask = $this->longToByteArray($mask);
            } else if (is_string($mask)) {
                $mask = preg_split('#\s*\.\s*#', trim($mask), -1, PREG_SPLIT_NO_EMPTY);

                switch (count($mask)) {
                    case 1: // CIDR
                        $cidr = (int) $mask[0];
                        if ($cidr === 0) {
                            // Shifting 32 bits on a 32 bit system doesn't work, so treat this as a special case
                            $mask = array(0, 0, 0, 0);
                        } else if ($cidr <= 32) {
                            // This looks odd, but it's the nicest way I have found to get the 32 least significant bits set in a
                            // way that works on both 32 and 64 bit platforms
                            $base = ~((~0 << 16) << 16);
                            $mask = $this->longToByteArray($base << (32 - $cidr));
                        } else {
                            throw new InvalidArgumentException('Supplied mask invalid: CIDR outside acceptable range 0 - 32');
                        }
                        break;
                    case 4: break; // Dotted decimal
                    default: throw new InvalidArgumentException('Supplied mask invalid: Must be either a full dotted-decimal or a CIDR');
                }
            } else if (is_array($mask)) {
                $mask = array_values($mask);
            } else {
                throw new InvalidArgumentException('Supplied mask invalid: Type invalid');
            }

            if (!$this->validateOctetArray($mask)) {
                throw new InvalidArgumentException('Supplied mask invalid: At least one octet value outside acceptable range 0 - 255');
            }
            // Check bits are contiguous from left
            // TODO: Improve this mechanism
            $asciiBits = sprintf('%032b', $this->byteArrayToLong($mask));
            if (strpos(rtrim($asciiBits, '0'), '0') !== FALSE) {
                throw new InvalidArgumentException('Supplied mask invalid: Set bits are not contiguous from the most significant bit');
            }

            $this->mask = $this->byteArrayToBinary($mask);
            $this->address = $this->byteArrayToBinary($address) & $this->mask;
        }

        /*
         * ArrayAccess interface methods (read only)
         */
        public function offsetExists ($offset) {
            if ($offset === 'network' || $offset === 'broadcast') {
                return TRUE;
            }

            $offset = filter_var($offset, FILTER_VALIDATE_INT);
            if ($offset === FALSE || $offset < 0) {
                return FALSE;
            }

            return $offset < $this->getHostsCount();
        }
        public function offsetGet ($offset) {
            if (!$this->offsetExists($offset)) {
                return NULL;
            }

            if ($offset === 'network') {
                $address = $this->getNetworkAddress(self::ADDRESS_INT);
            } else if ($offset === 'broadcast') {
                $address = $this->getBroadcastAddress(self::ADDRESS_INT);
            } else {
                // How much the address needs to be adjusted by to account for network address
                $adjustment = (int) ($this->getHostsCount() > 2);
                $address = $this->binaryToLong($this->address) + $offset + $adjustment;
            }

            return $this->longToSubnet($address);
        }
        public function offsetSet ($offset, $value) {}
        public function offsetUnset ($offset) {}

        /*
         * Iterator interface methods
         */
        public function current () {
            return $this->offsetGet($this->iteratorOffset);
        }
        public function key () {
            return $this->iteratorOffset;
        }
        public function next () {
            $this->iteratorOffset++;
        }
        public function rewind () {
            $this->iteratorOffset = 0;
        }
        public function valid () {
            return $this->iteratorOffset < $this->getHostsCount();
        }

        /*
         * Data access methods
         */
        public function getHosts ($mode = self::ADDRESS_SUBNET) {
            // Parse flags and initialise vars
            $bin = (bool) ($mode & self::ADDRESS_BINARY);
            $int = (bool) ($mode & self::ADDRESS_INT);
            $dd = (bool) ($mode & self::ADDRESS_DOTDEC);
            $base = $this->binaryToLong($this->address);
            $mask = $this->binaryToLong($this->mask);
            $hasNwBc = !($mask & 0x03);
            $result = array();

            // Get network address if requested
            if (($mode & self::HOSTS_WITH_NETWORK) && $hasNwBc) {
                $result[] = $base;
            }

            // Get hosts
            for ($current = $hasNwBc ? $base + 1 : $base; ($current & $mask) === $base; $current++) {
                $result[] = $current;
            }

            // Remove broadcast address if present and not requested
            if ($hasNwBc && !($mode & self::HOSTS_WITH_BROADCAST)) {
                array_pop($result);
            }

            // Convert to the correct type
            if ($bin) {
                $result = array_map(array($this, 'longToBinary'), $result);
            } else if ($dd) {
                $result = array_map(array($this, 'longToDottedDecimal'), $result);
            } else if (!$int) {
                $result = array_map(array($this, 'longToSubnet'), $result);
            }

            return $result;
        }
        public function getHostsCount () {
            $count = $this->getBroadcastAddress(self::ADDRESS_INT) - $this->getNetworkAddress(self::ADDRESS_INT);
            return $count > 2 ? $count - 1 : $count + 1; // Adjust return value to exclude network/broadcast addresses
        }
        public function getNetworkAddress ($mode = self::ADDRESS_SUBNET) {
            return $this->binaryToX($this->address, $mode);
        }
        public function getBroadcastAddress ($mode = self::ADDRESS_SUBNET) {
            return $this->binaryToX($this->address | ~$this->mask, $mode);
        }
        public function getMask ($mode = self::ADDRESS_DOTDEC) {
            return $this->binaryToX($this->mask, $mode);
        }

        /*
         * Stringify methods
         */
        public function __toString () {
            if ($this->getHostsCount() === 1) {
                $result = $this->toDottedDecimal();
            } else {
                $result = $this->toCIDR();
            }
            return $result;
        }
        public function toDottedDecimal () {
            $result = $this->getNetworkAddress(self::ADDRESS_DOTDEC);
            if ($this->mask !== "\xFF\xFF\xFF\xFF") {
                $result .= '/'.$this->getMask(self::ADDRESS_DOTDEC);
            }
            return $result;
        }
        public function toCIDR () {
            $address = $this->getNetworkAddress(self::ADDRESS_DOTDEC);
            $cidr = strlen(trim(sprintf('%b', $this->getMask(self::ADDRESS_INT)), '0')); // TODO: Improve this mechanism
            return $address.'/'.$cidr;
        }

        /*
         * Comparison methods
         */
        public function contains ($subject) {
            $this->normaliseComparisonSubject($subject);

            $subjectAddress = $subject->getNetworkAddress(self::ADDRESS_BINARY);
            $subjectMask = $subject->getMask(self::ADDRESS_BINARY);

            return $this->mask !== $subjectMask && ($this->mask | ($this->mask ^ $subjectMask)) !== $this->mask && ($subjectAddress & $this->mask) === $this->address;
        }

        public function within ($subject) {
            $this->normaliseComparisonSubject($subject);

            $subjectAddress = $subject->getNetworkAddress(self::ADDRESS_BINARY);
            $subjectMask = $subject->getMask(self::ADDRESS_BINARY);

            return $this->mask !== $subjectMask && ($this->mask | ($this->mask ^ $subjectMask)) === $this->mask && ($this->address & $subjectMask) === $subjectAddress;
        }
        public function equalTo ($subject) {
            $this->normaliseComparisonSubject($subject);

            return $this->address === $subject->getNetworkAddress(self::ADDRESS_BINARY) && $this->mask === $subject->getMask(self::ADDRESS_BINARY);
        }
        public function intersect ($subject) {
            $this->normaliseComparisonSubject($subject);

            return $this->equalTo($subject) || $this->contains($subject) || $this->within($subject);
        }

    }

In order to do what you desire, the class provides 4 methods:

contains()
within()
equalTo()
intersect()

Example usage of these:

// Also accepts dotted decimal mask. The mask may also be passed to the second
// argument. Any valid combination of dotted decimal, CIDR and integers will be
// accepted
$subnet = new IPv4Subnet('192.168.0.0/24');

// These methods will accept a string or another instance
var_dump($subnet->contains('192.168.0.1')); //TRUE
var_dump($subnet->contains('192.168.1.1')); //FALSE
var_dump($subnet->contains('192.168.0.0/16')); //FALSE
var_dump($subnet->within('192.168.0.0/16')); //TRUE
// ...hopefully you get the picture. intersect() returns TRUE if any of the
// other three match.

The class also implements the Iterator interface, allowing you to iterate over all the addresses in a subnet. The iterator excludes the network and broadcast addresses, which can be retrieved separately.

Example:

$subnet = new IPv4Subnet('192.168.0.0/28');
echo "Network: ", $subnet->getNetworkAddress(),
     "; Broadcast: ", $subnet->getBroadcastAddress(),
     "\nHosts:\n";
foreach ($subnet as $host) {
    echo $host, "\n";
}

The class also implements ArrayAccess, allowing you to treat it as an array:

$subnet = new IPv4Subnet('192.168.0.0/28');
echo $subnet['network'], "\n"; // 192.168.0.0
echo $subnet[0], "\n"; // 192.168.0.1
// ...
echo $subnet[13], "\n"; // 192.168.0.14
echo $subnet['broadcast'], "\n"; // 192.168.0.15

NB: The iterator/array methods of accessing the subnet's host addresses will return another IPv4Subnet object. The class implements __toString(), which will return the IP address as a dotted decimal if it represents a single address, or the CIDR if it represents more than one. The data can be accessed directly as a string or an integer by calling the relevant get*() method and passing the desired flag(s) (see constants defined at the top of the class).

All operations are 32- and 64-bit safe. Compatibility should be (although not thoroughly tested) 5.2+

See it working


For completeness, I imagine your use case would be implemented something along these lines:

public function addSubnet ($newSubnet) {
    $newSubnet = new IPv4Subnet($newSubnet);
    foreach ($this->subnets as &$existingSubnet) {
        if ($existingSubnet->contains($newSubnet)) {
            throw new Exception('Subnet already added');
        } else if ($existingSubnet->within($newSubnet)) {
            $existingSubnet = $newSubnet;
            return;
        }
    }
    $this->subnets[] = $newSubnet;
}

See it working

DaveRandom
  • 87,921
  • 11
  • 154
  • 174
  • +1 for going beyond the call of duty ;) I'll try and understand all the bitwise magic in `contains` after some caffeine. – Leigh Nov 29 '12 at 06:47
  • Thanks again Dave. I yet have to test it, but i'm quite sure the magic I need is in here :) And if not I'll just bug you in chat :) – Damien Overeem Nov 29 '12 at 08:00
  • Just a small thing though, your class does not override ArrayAccess' offsetUnset method. Easily fixed though. – Damien Overeem Nov 29 '12 at 09:11
  • Oh actually it does.. why the hell was my netbeans bugging me about it :) – Damien Overeem Nov 29 '12 at 09:23
  • @DamienOvereem The class is designed to be immutable, as allowing the address/masked to be changed opens up a whole can of worms. Also it would make no sense to allow altering a single item in the middle of a subnet, because then it would no longer be one complete subnet. If your IDE is bugging you about the empty methods, adding a `return;` statement to them will most likely make it happy again. – DaveRandom Nov 29 '12 at 09:56
  • Busy implementing it now, seems to be working like a charm. Good work :) – Damien Overeem Nov 29 '12 at 10:02
  • @Leigh There are three operations: make sure the masks are not equal, make sure the subject mask is narrower than the current mask (xor them to get the difference, then the or of the current mask and the difference should equal the current mask if it is narrower), make sure that the subject base address equals my base address when my mask is applied to it. `within()` just inverses the last two operations. – DaveRandom Nov 29 '12 at 10:07
  • @DaveRandom +1, Very nice! When is you IPv6Subnet version coming out? =oP – kittycat Jan 02 '13 at 17:18
  • @cryptic As soon as I need it :-P I actually started writing it once then I stopped because I came to a decision that I couldn't make regarding the fact that IPv6 addresses are effectively 128-bit ints so they cannot be represented as ints. I will dig it out and have a look at it later and try and get it finished. – DaveRandom Jan 02 '13 at 17:22
  • @DaveRandom, wouldn't BCMath functions allow using it since they use numeric strings and have afaik no limit to their size? – kittycat Jan 02 '13 at 17:25
  • @cryptic Indeed they would, but I'd rather not couple to an extension that's not necessarily available if I can avoid it. I'm not sure there's that much use for an IPv6 address as an int though, every use case I can think of for one you can do bitwise ops to get the info you need or simply store it as a packed binary string (for storage in DB etc). Thoughts? (possibly better for /rooms/11/php if you have any, lest we get told off by SO for having an extended discussion in comments) – DaveRandom Jan 02 '13 at 17:30
3

As discussed briefly in PHP chat, here's how I would implement it, to compare any two addresses.

  1. Convert the IP addresses to their binary form
  2. Extract the masks from the CIDR format
  3. Take the minimum mask of the two (least specific = contains more addresses)
  4. Use the mask on both binary representations.
  5. Compare the two.

If there is a match, then one is contained within the other.

Here's some example code, it's not very pretty and you'll want to adapt it to cater for your array.

function bin_pad($num)
{
    return str_pad(decbin($num), 8, '0', STR_PAD_LEFT);
}

$ip1 = '192.168.0.0/23';
$ip2 = '192.168.1.0/24';

$regex = '~(\d+)\.(\d+)\.(\d+)\.(\d+)/(\d+)~';

preg_match($regex, $ip1, $ip1);
preg_match($regex, $ip2, $ip2);

$mask = min($ip1[5], $ip2[5]);

$ip1 = substr(
    bin_pad($ip1[1]) . bin_pad($ip1[2]) .
    bin_pad($ip1[3]) . bin_pad($ip1[4]),
    0, $mask
);

$ip2 = substr(
    bin_pad($ip2[1]) . bin_pad($ip2[2]) .
    bin_pad($ip2[3]) . bin_pad($ip2[4]),
    0, $mask
);

var_dump($ip1, $ip2, $ip1 === $ip2);

I had trouble making it 32 bit compatible, which is why I eventually opted for converting each octet of the IP address into binary individually, and then using substr.

I started off using pack('C4', $ip[1] .. $ip[4]) but when it came to using a full 32 bit mask I ran into problems converting it into binary (since PHP integers are signed). Thought for a future implementation though!

Leigh
  • 12,859
  • 3
  • 39
  • 60
0

Intuitively I would suggest you'd want to do something like:

  1. Let the new entry be X
  2. Convert X to single integer form, let that integer be Y
  3. Let mask length of any entry A be mask(A)
  4. Compare any existing entries where mask(entry) = mask(Y)
  5. Mask off existing entries where mask(entry) > mask(Y) and compare with Y
  6. Mask off Y for each existing entry where mask(entry) < mask(X), such that mask(Y) = mask(entry) and compare

Provided you encounter no collisions, all is well.

Of course this does not check if the proposed subnet is valid.

My proposition of correctness here is that I can't think of a counter-example, but there may well be one so I offer this as a basis for further thought - hope this helps.

jstephenson
  • 2,170
  • 14
  • 14
0
<?php
    function checkOverlap ($net1, $net2) {

        $mask1 = explode("/", $net1)[1];
        $net1 = explode("/", $net1)[0];
        $netArr1 = explode(".",$net1);

        $mask2 = explode("/", $net2)[1];
        $net2 = explode("/", $net2)[0];
        $netArr2 = explode(".",$net2);

        $newnet1 = $newnet2 = "";

        foreach($netArr1 as $num) {
            $binnum = decbin($num);
            $length = strlen($binnum);
            for ($i = 0; $i < 8-$length; $i++) {
                $binnum = '0'.$binnum;
            }
            $newnet1 .= $binnum;
        }

        foreach($netArr2 as $num) {
            $binnum = decbin($num);
            $length = strlen($binnum);
            for ($i = 0; $i < 8-$length; $i++) {
                $binnum = '0'.$binnum;
            }
            $newnet2 .= $binnum;
        }

        $length = min($mask1, $mask2);

        $newnet1 = substr($newnet1,0,$length);
        $newnet2 = substr($newnet2,0,$length);

        $overlap = 0;
        if ($newnet1 == $newnet2) $overlap = 1;

        return $overlap;
    }

    function networksOverlap ($networks, $newnet) {

        $overlap = false;

        foreach ($networks as $network) {
            $overlap = checkOverlap($network, $newnet);
            if ($overlap) return 1;
        }

        return $overlap;        

    }

    $cidrNetworks = array(
        '192.168.10.0/24',
        '10.10.0.30/20'
    );

    $newnet = "192.168.10.0/25";

    $overlap = networksOverlap($cidrNetworks, $newnet);
?>

Not sure if this is 100% correct but try it out see if it works.

GTCrais
  • 2,039
  • 2
  • 26
  • 32
  • 1
    Couple of things, you're doing the same `explode` operation a couple of times. You could use `list($mask1, $net1)` to only perform them once. Check out `str_pad`, those padding loops are unnecessary. The ternary comparison to return the maximum value is unnecessary, use `max` (although I'm sure you actually want `min`). Other than that, it essentially does the same thing as mine. – Leigh Nov 28 '12 at 16:48