Re: Turning on DOITM at last

[Christian Ludloff]

Continuing this summer's thread...

  https://lore.kernel.org/lkml/14962fdf-66e9-4b64-a37e-0e697c875cc0@xxxxxxxxxx/

The DOITM U/S spec arrived, as mentioned on the gcc-patches list.

  https://gcc.gnu.org/pipermail/gcc-patches/2026-November/738803.html

  https://www.intel.com.br/content/www/br/pt/content-details/932747/user-doitm-and-sup-doitm-extensions-architecture-specification.html

In section 3.1.3 it proposes two changes to the published DOIT list.

  https://www.intel.com/content/www/us/en/developer/articles/technical/software-security-guidance/resources/data-operand-independent-timing-instructions.html

Namely, for shift/rotate and for rep stosb.

How is this not problematic? Will Intel do the required SW work? How?

And what good is a DOIT list... when it then changes under one's feet?

(Other insanity: the naming prefix/suffix mess and the cpl1/cpl2 mess.)

--
C.