ethtool ioctl crashes if number of strings exceeds hardcoded value

[pspacek]

Version affected: c7da937

Problematic code:
https://github.com/svinota/pyroute2/blob/c7da937999dc59167f81f9466c80e891fb7be736/pyroute2/ethtool/ioctl.py#L284C14-L284C14

It turns out that comment

        # If you have more than 256 features on your NIC
        # they will not be seen by it

... is not correct. The code outright segfaults if number of strings returned exceeds hardcoded value. Easiest way to test is to manually drop the number to a small value, but it is practical problem e.g. in AWS VM type c5n.18xlarge which can return 1105 strings.

We already know the correct values from ioctl ETHTOOL_GSSET_INFO but it is not used in structure definition or checked anywhere. I might try to fix it but it will require redefining array sizes dynamically, I think.

[svinota]

Oh. Thanks for the report. If you try to provide a PR it would be nice, otherwise I'm to address this bug.

[pspacek]

Turns out I did not have time for it - I can get to it in couple weeks. If you are faster so be it ...

[svinota]

Reproduced.

I will try to fix the issue this week.

[svinota]

It took some more time than expected.

Could you please check if it works for you?

[pspacek]

This fix partially works in a sense that it does not crash if you ask for the same stringset every time. It can potentially crash when you change set_id on the fly. I've attempted to generalize this in #1126 . It makes the structures a bit less dynamic but correctly takes into account different string set sizes for distinct IDs.