Re: [PATCH] x86: add 'runtime constant' infrastructure

[H. Peter Anvin]

On June 9, 2024 4:22:40 AM PDT, Borislav Petkov <bp@xxxxxxxxx> wrote:
>On Sat, Jun 08, 2024 at 12:35:05PM -0700, Linus Torvalds wrote:
>> Ingo / Peter / Borislav - I enabled this for 32-bit x86 too, because it
>> was literally trivial (had to remove a "q" from "movq").  I did a
>> test-build and it looks find, but I didn't actually try to boot it. 
>
>Will do once you have your final version. I still have an Atom, 32-bit
>only laptop lying around here.
>
>> +#define runtime_const_ptr(sym) ({				\
>> +	typeof(sym) __ret;					\
>> +	asm("mov %1,%0\n1:\n"					\
>> +		".pushsection runtime_ptr_" #sym ",\"a\"\n\t"	\
>> +		".long 1b - %c2 - .\n\t"			\
>> +		".popsection"					\
>> +		:"=r" (__ret)					\
>> +		:"i" ((unsigned long)0x0123456789abcdefull),	\
>> +		 "i" (sizeof(long)));				\
>> +	__ret; })
>
>You might wanna use asm symbolic names for the operands so that it is
>more readable:
>
>#define runtime_const_ptr(sym) ({                                               \
>        typeof(sym) __ret;                                                      \
>        asm("mov %[constant] ,%[__ret]\n1:\n"                                   \
>                ".pushsection runtime_ptr_" #sym ",\"a\"\n\t"                   \
>                ".long 1b - %c[sizeoflong] - .\n\t"                             \
>                ".popsection"                                                   \
>                : [__ret] "=r" (__ret)                                          \
>                : [constant] "i" ((unsigned long)0x0123456789abcdefull),        \
>                  [sizeoflong] "i" (sizeof(long)));                             \
>        __ret; })
>
>For example.
>
>> +// The 'typeof' will create at _least_ a 32-bit type, but
>> +// will happily also take a bigger type and the 'shrl' will
>> +// clear the upper bits
>
>Can we pls use the multiline comments, like you do below in the same
>file.
>
>Otherwise, it looks ok to me and it boots in a guest.
>
>I'll take the final version for a spin on real hw in a couple of days,
>once the review dust settles.
>
>Thx.
>

So the biggest difference versus what I had in progress was that I had the idea of basically doing "ro_after_init" behavior by doing memory references until alternatives are run.

I don't know if that was overthinking the problem...