Potential security issue by disabling W^X in the original function and the jit memory

For functions in the memory, we usually have W^X protection to avoid buffer overflow attacking.

In the patch_function, it disabled the W^X first to be able to inject code to the original function address:

pub(crate) unsafe fn patch_function(func: *mut u8, patch: &[u8]) {
    make_memory_writable_and_executable(func); // <--------- HERE

    inject_asm_code(patch, func);
}

Should we consider doing it like this:

• Change the memory page to Writable only (or Writable and Executable)
• Inject the code.
• Change the memory page to Executable only.

This idea also applies to the allocate_jit_memory.

Start by locating patch_function and allocate_jit_memory, then inspect how make_memory_writable_and_executable and inject_asm_code manage page permissions. Done means the injection flow is preserved while writable and executable permissions are not left enabled together longer than necessary, with the behavior verified by the relevant existing checks.