问题
In this page, http://www.x86-64.org/pipermail/discuss/2004-August/005020.html He said that there is a way to mix 32-bit code and 64-bit code in a application. He assumed the application is 32-bit (in compatibility mode) and then switch to 64-bit mode to execute 64-bit code and vice versa.
Assume my OS is 64-bit linux and my application is 64-bit. I do a far jump to switch to compatibility mode and execute 32-bit code. Does it can work correctly when I do a system call or function call ?
Is there any overhead of mode switching between compatibility mode and 64-bit mode ? I think one of the overhead is I need separate stack for 32-bit and 64-bit.
Could I integrate this idea into JVM, maybe I can dynamic generate 32-bit code in 64-bit JVM, and execute it by mode switching ?
回答1:
Open-coded syscalls should be fine, since your 32-bit code will use the 32-bit kernel entry point.
Function calls can only be made to other 32-bit code, of course. This includes libc - so your 32-bit code will either have to be self-contained, or you will have to provide thunks for the library functions that it needs. Remember that usually syscalls are not called directly - you normally go via a libc wrapper that will be unavailable to your 32-bit code.
There is certainly an overhead for switching between modes. You should consult your processor documentation to find out what it is.
来源：https://stackoverflow.com/questions/5081203/is-it-possible-to-execute-32-bit-code-in-64-bit-process-by-doing-mode-switching