News for hackers. Stuff that matters.
Re:Towards Linux Kernel Memory Safety by pjmlp on Thursday November 16, 2017 @03:45AM (#15711133)
Languages like Ada just forbid many of the errors that lead to memory corruption in C, by default.

So it would only happen with a binary Ada library in two ways:

- the binary was manipulated and thus does not map to the original source code

- the code was compiled with permission for Uncheked packages enabled

In C, 100% of the source code is unsafe, specially when compiled with -O3.

Re:Towards Linux Kernel Memory Safety by joosters on Thursday November 16, 2017 @04:13AM (#15711250)
I was referring to your original comment: 'they only work properly if 100% of the source code is available.'

This project is adding protection to the linux kernel. Now, this bounds checking could be added to a kernel build with 100% of the source code available, in which case all is good. Or, it could be added to a kernel build where some of the drivers are available only in a 'binary blob' form. In which case, rewriting the kernel in Ada (or whatever) would not help you with the binary blobs, since they are out of your control. So, as I understand it, your point of 'only working properly if 100% of the source code is available' would seem to be just as true for a code rewrite as it is for the paper's approach to the problem.