Conversation

Here, I am just trying to summarize the differences (compare to Scudo), please correct me, if I misunderstood: 1. Scudo can't reliably detect invalid free. 2. It lacks fine grained randomization. Is there any differences if we try to compare all 3 above mentioned allocators?🤔
3
1
The comparison in that paper is very incomplete and misleading. It oversimplifies a lot of the differences and boils down complex things to a single point, which is wrong. It's not where you should your information about this. hardened_malloc is also not OpenBSD malloc.
1
All I can do is strongly suggest reading through github.com/GrapheneOS/har, testing it out (github.com/GrapheneOS/har shows the baseline size class partitioning and there are tests demonstrating some other things in github.com/GrapheneOS/har) and looking at the implementation.
1
Please don't try making a comparison like this before doing the research: twitter.com/_neerajpal/sta Someone else at Mozilla recently did the same and attacked it as nearly completely useless and easily replicated elsewhere based on false claims, and I don't appreciate it at all.
Quote Tweet
Here, I am just trying to summarize the differences (compare to Scudo), please correct me, if I misunderstood: 1. Scudo can't reliably detect invalid free. 2. It lacks fine grained randomization. Is there any differences if we try to compare all 3 above mentioned allocators?🤔
Show this thread