[Network] fix oom bug

[J-N-K]

Fixes #3145

Large networks (like /8) caused an OOM exception because 16 million IP addresses were cconstructed during network discovery. This reduces the network size to the number of requested addresses (usually 255 which results in /24).

Verified locally.

[davidgraeff]

Hm. Why did that happen in the first place. maximumPerInterface should have limited the number of IP addresses per interface.

[J-N-K]

Old l. 139 constructs an array of all addresses which is later truncated to the number of requested addresses in l. 144.

This fails on systems with low memory (on my development machine I wasn‘t able to reproduce it until I reduced the vm heap size).

[davidgraeff]

Ah ok. The problem is the SubnetUtils class, which does not provide means to limit the returned addresses. So this workaround is to modify the subnet, I see.

But there is also an IPv4 matcher included. That should be removed, because the interface IPs are provided by getInterfaceIPs which should only return IPv4 addresses. (I checked the code and it does not so atm, the correct fix is to check if the address is an instance of IPv4Address in

https://github.com/openhab/openhab2-addons/blob/fe8cbb0e79707064049a7984f2e41e84b4e53730/addons/binding/org.openhab.binding.network/src/main/java/org/openhab/binding/network/internal/utils/NetworkUtils.java#L91

[J-N-K]

I can add that there, too. But I need the matcher anyway to separate address and CIDR suffix, so it will result in a double check.

[davidgraeff]

Took some time but I have found it:

https://github.com/openhab/openhab-core/blob/master/bundles/org.openhab.core/src/main/java/org/eclipse/smarthome/core/net/CidrAddress.java

There is a class in the core bundle (also in eclipse smarthome core) that wraps an InetAddress and a prefix length. We can just return that in getInterfaceIPs and the prefixLength check can be much shorter in the method in question.

See also: https://github.com/openhab/openhab-core/blob/master/bundles/org.openhab.core/src/main/java/org/eclipse/smarthome/core/net/NetUtil.java#L373

That is a static method that makes the network binding method superfluous.

[J-N-K]

I‘ll have a look.

[J-N-K]

@davidgraeff, I also removed the exception handling whcih was only necessary if an invalid address was passed to new SubnetUtils(address).

[davidgraeff]

LGTM

[openhab-bot]

This pull request has been mentioned on openHAB Community. There might be relevant details there:

https://community.openhab.org/t/network-binding-problem-heap-space/72243/2