Re: [PATCH net-next] x86: bpf_jit_comp: secure bpf jit againstspraying attacks

From: Eric Dumazet
Date: Mon May 20 2013 - 09:52:20 EST

On Mon, 2013-05-20 at 11:50 +0200, Daniel Borkmann wrote:

> Here seems also to be another approach ...
> via:

Well, there are many approaches, and I have another one as well provided
by H. Peter Anvin.

Idea was to allow the code being relocated outside of the 2GB space that
we use for kernel code (including module_alloc())

So every call helper, coded in "e8 xx xx xx xx" was replaced by

"48 c7 c0 yy yy yy yy mov $foo,%rax"
"ff d0 call *%rax"

The RO protection + random holes idea was a solution with no performance

Another idea is to limit BPF JIT to root users. I do not think BPF JIT
is mandatory at all, as tcpdump is already restricted.

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at
Please read the FAQ at