-Zharden-sls flag (target modifier) added to enable mitigation against straight line speculation (SLS) #136597
Conversation
rustbot
added
S-waiting-on-review
This comment has been minimized.
This comment has been minimized.
|
Some changes occurred in compiler/rustc_codegen_gcc |
|
First commit adds retpoline feature, but PR descriptions says about sls. |
Yes, first commit is in another PR #135927. |
|
Marking this as blocked on #135927 |
Dylan-DPC
added
S-blocked
|
I filed an MCP for this flag: rust-lang/compiler-team#869 |
|
☔ The latest upstream changes (presumably #141232) made this pull request unmergeable. Please resolve the merge conflicts. |
| #![feature(no_core, lang_items)] | ||
| #![no_std] | ||
| #![no_core] | ||
|
|
||
| #[lang = "sized"] | ||
| pub trait Sized {} |
There was a problem hiding this comment.
Use add-core-stubs, don't handwrite the minicore.
| warning: target feature `retpoline-external-thunk` cannot be enabled with `-Ctarget-feature`: use `x86-retpoline` target modifier flag instead | ||
| | | ||
| = note: this was previously accepted by the compiler but is being phased out; it will become a hard error in a future release! | ||
| = note: for more information, see issue #116344 <https://github.com/rust-lang/rust/issues/116344> | ||
|
|
||
| warning: 1 warning emitted |
There was a problem hiding this comment.
Why is this possible at all?
| //@ compile-flags: --target=x86_64-unknown-linux-gnu --crate-type=lib | ||
| //@ needs-llvm-components: x86 | ||
| //@ [by_flag]compile-flags: -Zharden-sls=all | ||
| //@ [by_feature]compile-flags: -Ctarget-feature=+harden-sls-ijmp,+harden-sls-ret |
There was a problem hiding this comment.
These should not be set by -Ctarget-feature.
Darksonn
added
S-waiting-on-author
rustbot
added
the
A-LLVM
rustbot
added
the
S-waiting-on-review
|
Some changes occurred in compiler/rustc_codegen_ssa |
rustbot
removed
the
S-waiting-on-author
|
Converted to "Ready for review" because retpoline PR was merged. |
|
Adding myself to review as part of the RfL project goal I signed up for 🙂 r? wesleywiser |
|
Thanks Wesley! Much appreciated. |
|
By the way, @azhogin, I think we should have an assembly test for this one too, even if small/trivial, to double-check LLVM is actually doing something, like in the other flags. (Another rebase is needed, given the recent renaming of the folders) Thanks! |
|
This PR was rebased onto a different master commit. Here's a range-diff highlighting what actually changed. Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers. |
Added |
tests/assembly-llvm/x86_64-sls.rs
Outdated
| 4 => unsafe { return std::intrinsics::unchecked_div(b, a) + 4 }, | ||
| 5 => unsafe { return std::intrinsics::unchecked_div(b, a) + 5 }, | ||
| 6 => unsafe { return std::intrinsics::unchecked_div(b, a) + 6 }, |
There was a problem hiding this comment.
I assume all these are here to ensure the compiler is likely to generate a jump table?
By the way, you can remove all the returns, and it will be a bit more readable since the formatting will be redone.
tests/assembly-llvm/x86_64-sls.rs
Outdated
| if a > 0 { | ||
| return unsafe { std::intrinsics::unchecked_div(a, b) }; | ||
| } else { | ||
| return unsafe { std::intrinsics::unchecked_div(b, a) }; | ||
| } |
There was a problem hiding this comment.
| if a > 0 { | |
| return unsafe { std::intrinsics::unchecked_div(a, b) }; | |
| } else { | |
| return unsafe { std::intrinsics::unchecked_div(b, a) }; | |
| } | |
| if a > 0 { | |
| unsafe { std::intrinsics::unchecked_div(a, b) } | |
| } else { | |
| unsafe { std::intrinsics::unchecked_div(b, a) } | |
| } |
Thanks! |
…t straight line speculation (SLS)
Flag (target modifier) to mitigate against straight line speculation (SLS).
-Zharden-sls=[none|all|return|indirect-jmp].The flag enables the related features:
+harden-sls-ijmp,+harden-sls-ret.The flag is tracked as a target modifier to be equal between linked crates.
Tracking issue: #116851