Wrong memory pages permissions restoration

[xtsop]

I see that after decrypting .rodata section you are reverting the decrypted memory pages using mprotect() to Read-Execute permissions. While this is necessary when decrypting the .text, it is not so for data sections, essentially making it a hot target for shellcode placement. If I'm missing anything, let me know.

[tweksteen]

By default, the .rodata section is already executable (see http://xn--thibaud-dya.fr/rodata.html) and yes, it is an heresy. For now, the code is simply returning that page into its previous state.

Feel free to push upstream (glibc, gold) to try to get that change implemented/widely used. Additionally, I could also modify that part of the code to verify the previous permissions and restore them exactly (feel free to send a patch).