crypto-com / cosmos-sdk-codeql Goto Github PK

Hi,

Thanks for this great repo!

I was recently looking into the beginendblock-panic.ql and trying to understand it.
In particular, the query uses mayPanic() as an indicator that the function would panic.
The problem is, however, that mayPanic() will return true almost always, the only exception being a couple of built-in functions (as discussed in this codeQL issue).
This also means that the beginendblock-panic.ql will not be precise (because mayPanic() is not precise).

I believe there is a second problem in the query: namely, calling getParent*, which repeatedly calls the function getParent.
The problem is that getParent is not returning a function call parent, but an AST node parent (where the top ancestor is a file in which the call resides) .

Does this make sense?
In case I got things wrong, please let me know.

The beginendblock rule checks for panics in these ABCI calls. Is there a reason that this check is only performed in these 2 situations? For example, why not in the CheckTx ABCI call?

Thanks!

Hey,

As a part of my thesis I'm scanning live cosmos-sdk chains for possible vulnerabilities, and one way is using these CodeQl rules. However, I am unsure on why using variables called Bech32... is a vulnerability (like this rule suggests).

A more concrete example: this line in the osmosis repo is being flagged as a warning. I cannot seem to figure out why this would be insecure.

Would you mind explaining it a bit?

Thanks!