As Xu mentioned, that dealing with multiple verifier behaviors in production is painful. Recently, I hit a pointer-arithmetic related BPF verifier failures with an XDP program on Ubuntu 24.04.4 LTS running on 6.8.0-124-generic kernel, while the same code loaded fine on Debian 13's 6.12.63+deb13-amd64 and also on Ubuntu 24.04.4 LTS with 6.17.0-35-generic.
In practice, I've found these kinds of verifier differences more troublesome than missing helpers or kfuncs, since the latter are generally easier to detect and work around.
Would a modularized verifier actually help reduce this kind of drift, or are these differences typically rooted in interactions outside verifier.c?
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。