Does #Meltdown bypass segment limit? To be specific, is e.g. Linux 2.0 on 32-bit x86 vulnerable? If not, maybe consider http://wiki.osdev.org/X86-64#Segmentation_in_Long_Mode …
-
-
Replying to @solardiz
Evil X Retweeted grsecurity
Did you see https://twitter.com/grsecurity/status/949273306603147266 … https://twitter.com/grsecurity/status/949433055764275200 … and https://twitter.com/grsecurity/status/949499167700865024 … ?
Evil X added,
1 reply 0 retweets 0 likes -
Replying to @Evil_X_
Thanks! So segment limit works against Meltdown, but conflicts with old? TLS ABI. Should we trap & emulate the problematic TLS accesses in the kernel for no or lower typical (even if much higher worst case) performance impact than KPTI? Both i386 & x86_64.
@RichFelker@grsecurity3 replies 0 retweets 0 likes -
It's not old; it's current/only ABI. Trap&emulate would be very very slow. GCC(+ld due to relaxation) could be trained not to use the local exec model & then maybe trap would be ok for old code not recompiled...?
1 reply 0 retweets 0 likes -
Replying to @RichFelker @solardiz and
Actually the option already exists. -mno-tls-direct-seg-refs https://gcc.gnu.org/onlinedocs/gcc/x86-Options.html …
2 replies 0 retweets 0 likes
See also https://www.akkadia.org/drepper/tls.pdf page 47 - it looks like Drepper overlooked a very good reason why Sun did things the way they did when he NIH'd the GNU variant of it.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.