Wireshark · Wireshark-dev: Re: [Wireshark-dev] Babel: prevent an infinite loop while parsing sub-TLV

[Pascal Quantin <pascal@xxxxxxxxxxxxx>]

Hi Juliusz,

Le ven. 18 oct. 2019 à 20:51, Juliusz Chroboczek <jch@xxxxxxx> a écrit :

Dear Pascal,

I've just seen your commit dd15b2, which I believe is incorrect.

  https://code.wireshark.org/review/gitweb?p=wireshark.git;a=commit;h=dd15b203c7ec8a8ec5c930cf018c838991ee3182

In dissect_babel_subtlvs, sublen is the length of the TLV body; the full
TLV is of length sublen+2.  At the end of the loop, the code is correct:

        beg += (sublen+2);

Hence, it is perfectly fine to have sublen=0.  Unless I'm wrong, your
commit is incorrect.

Maybe, honestly I do not remember the details as it was done a few months ago. It would require me to dive in once I have the time (so not now).

Could you please either explain why I'm wrong, or else revert commit dd15b2?

I will not revert this commit as it solved/workarounded an infinite loop, as seen in https://bugs.wireshark.org/bugzilla/show_bug.cgi?id=15856

But is you can come up with a better solution that ensures that we do not have the infinite loop while still being correct, you are welcome and I will try to review it

Best regards,

Pascal.