On Sun, Jan 14, 2018 at 11:40 AM, Jan Hubicka <hubi...@ucw.cz> wrote: >> Hi HJ, >> >> > -----Original Message----- >> > From: gcc-patches-ow...@gcc.gnu.org [mailto:gcc-patches- >> > ow...@gcc.gnu.org] On Behalf Of H.J. Lu >> > Sent: Sunday, January 14, 2018 9:07 AM >> > To: gcc-patches@gcc.gnu.org >> > Subject: [PATCH 0/5] x86: CVE-2017-5715, aka Spectre >> > >> > This set of patches for GCC 8 mitigates variant #2 of the speculative >> > execution vulnerabilities on x86 processors identified by CVE-2017-5715, >> > aka >> > Spectre. They convert indirect branches and function returns to call and >> > return thunks to avoid speculative execution via indirect call, jmp and >> > ret. >> > >> > H.J. Lu (5): >> > x86: Add -mindirect-branch= >> > x86: Add -mfunction-return= >> > x86: Add -mindirect-branch-register >> > x86: Add 'V' register operand modifier >> > x86: Disallow -mindirect-branch=/-mfunction-return= with >> > -mcmodel=large >> >> Current set of patches don't seem to have any option to generate "lfence" as >> the loop filler in "retpoline", which is required by AMD. >> Can you please clarify the plan. We would like to get this checked-in GCC 8. > > Since thunks are output as strings, it is easy to add the option > on the top of patch #1 of the series. I do not fully understand > the reason for choosing pause over lfence for Intel, but if we need > to do both, we need to have command line option (and possibly attribute). > What would be reasonable name for it?
Looking at the kernel patch [1], the loop filler should be "pause;lfence" sequence, and should be universally accepted for Intel and AMD targets. [1] https://www.spinics.net/lists/kernel/msg2697507.html Uros.
