LLVM has built-in heuristics for adding stack canaries to functions. These
heuristics can be selected with LLVM function attributes. This patch adds a
rustc option `-Z stack-protector={none,basic,strong,all}` which controls the use
of these attributes. This gives rustc the same stack smash protection support as
clang offers through options `-fno-stack-protector`, `-fstack-protector`,
`-fstack-protector-strong`, and `-fstack-protector-all`. The protection this can
offer is demonstrated in test/ui/abi/stack-protector.rs. This fills a gap in the
current list of rustc exploit
mitigations (https://doc.rust-lang.org/rustc/exploit-mitigations.html),
originally discussed in #15179.
Stack smash protection adds runtime overhead and is therefore still off by
default, but now users have the option to trade performance for security as they
see fit. An example use case is adding Rust code in an existing C/C++ code base
compiled with stack smash protection. Without the ability to add stack smash
protection to the Rust code, the code base artifacts could be exploitable in
ways not possible if the code base remained pure C/C++.
Stack smash protection support is present in LLVM for almost all the current
tier 1/tier 2 targets: see
test/assembly/stack-protector/stack-protector-target-support.rs. The one
exception is nvptx64-nvidia-cuda. This patch follows clang's example, and adds a
warning message printed if stack smash protection is used with this target (see
test/ui/stack-protector/warn-stack-protector-unsupported.rs). Support for tier 3
targets has not been checked.
Since the heuristics are applied at the LLVM level, the heuristics are expected
to add stack smash protection to a fraction of functions comparable to C/C++.
Some experiments demonstrating how Rust code is affected by the different
heuristics can be found in
test/assembly/stack-protector/stack-protector-heuristics-effect.rs. There is
potential for better heuristics using Rust-specific safety information. For
example it might be reasonable to skip stack smash protection in functions which
transitively only use safe Rust code, or which uses only a subset of functions
the user declares safe (such as anything under `std.*`). Such alternative
heuristics could be added at a later point.
LLVM also offers a "safestack" sanitizer as an alternative way to guard against
stack smashing (see #26612). This could possibly also be included as a
stack-protection heuristic. An alternative is to add it as a sanitizer (#39699).
This is what clang does: safestack is exposed with option
`-fsanitize=safe-stack`.
The options are only supported by the LLVM backend, but as with other codegen
options it is visible in the main codegen option help menu. The heuristic names
"basic", "strong", and "all" are hopefully sufficiently generic to be usable in
other backends as well.
Reviewed-by: Nikita Popov <nikic@php.net>
Extra commits during review:
- [address-review] make the stack-protector option unstable
- [address-review] reduce detail level of stack-protector option help text
- [address-review] correct grammar in comment
- [address-review] use compiler flag to avoid merging functions in test
- [address-review] specify min LLVM version in fortanix stack-protector test
Only for Fortanix test, since this target specifically requests the
`--x86-experimental-lvi-inline-asm-hardening` flag.
- [address-review] specify required LLVM components in stack-protector tests
- move stack protector option enum closer to other similar option enums
- rustc_interface/tests: sort debug option list in tracking hash test
- add an explicit `none` stack-protector option
Revert "set LLVM requirements for all stack protector support test revisions"
This reverts commit a49b74f92a4e7d701d6f6cf63d207a8aff2e0f68.
Methods that were only blocked on `const_panic` have been stabilized.
The remaining methods of `duration_consts_2` are all related to floats,
and as such have been placed behind the `duration_consts_float` feature
gate.
Mention std::io::Error::from(ErrorKind) in Error::new() docs
This conversion is not very discoverable for the cases
where an error is required without extra payload.
Previously, we set the default color for <a> tags to black, and then
had an override with a bunch of not() clauses to set anchors in
docblocks to blue.
Instead, we should set the default color for <a> to blue (or equivalent
in other themes), and override it for places like the sidebar or search
results, where we don't want them to be styled as links.
Check for duplicate attributes.
This adds some checks for duplicate attributes. In many cases, the duplicates were being ignored without error or warning. This adds several kinds of checks (see `AttributeDuplicates` enum).
The motivation here is to issue unused warnings with similar reasoning for any unused lint, and to error for cases where there are conflicts.
This also adds a check for empty attribute lists in a few attributes where this causes the attribute to be ignored.
Closes#55112.
Simplify `for` loop desugar
Basically two intermediate bindings are inlined. I could have left one intermediate binding in place as this would simplify some diagnostic logic, but I think the difference in that regard would be negligible, so it is better to have a minimal HIR.
For checking that the pattern is irrefutable, I added a special case when the `match` is found to be non-exhaustive.
The reordering of the arms is purely stylistic. I don't *think* there are any perf implications.
```diff
match IntoIterator::into_iter($head) {
mut iter => {
$label: loop {
- let mut __next;
match Iterator::next(&mut iter) {
- Some(val) => __next = val,
None => break,
+ Some($pat) => $block,
}
- let $pat = __next;
- $block
}
}
}
```
libcore: assume the input of `next_code_point` and `next_code_point_reverse` is UTF-8-like
The functions are now `unsafe` and they use `Option::unwrap_unchecked` instead of `unwrap_or_0`
`unwrap_or_0` was added in 42357d772b. I guess `unwrap_unchecked` was not available back then.
Given this example:
```rust
pub fn first_char(s: &str) -> Option<char> {
s.chars().next()
}
```
Previously, the following assembly was produced:
```asm
_ZN7example10first_char17ha056ddea6bafad1cE:
.cfi_startproc
test rsi, rsi
je .LBB0_1
movzx edx, byte ptr [rdi]
test dl, dl
js .LBB0_3
mov eax, edx
ret
.LBB0_1:
mov eax, 1114112
ret
.LBB0_3:
lea r8, [rdi + rsi]
xor eax, eax
mov r9, r8
cmp rsi, 1
je .LBB0_5
movzx eax, byte ptr [rdi + 1]
add rdi, 2
and eax, 63
mov r9, rdi
.LBB0_5:
mov ecx, edx
and ecx, 31
cmp dl, -33
jbe .LBB0_6
cmp r9, r8
je .LBB0_9
movzx esi, byte ptr [r9]
add r9, 1
and esi, 63
shl eax, 6
or eax, esi
cmp dl, -16
jb .LBB0_12
.LBB0_13:
cmp r9, r8
je .LBB0_14
movzx edx, byte ptr [r9]
and edx, 63
jmp .LBB0_16
.LBB0_6:
shl ecx, 6
or eax, ecx
ret
.LBB0_9:
xor esi, esi
mov r9, r8
shl eax, 6
or eax, esi
cmp dl, -16
jae .LBB0_13
.LBB0_12:
shl ecx, 12
or eax, ecx
ret
.LBB0_14:
xor edx, edx
.LBB0_16:
and ecx, 7
shl ecx, 18
shl eax, 6
or eax, ecx
or eax, edx
ret
```
After this change, the assembly is reduced to:
```asm
_ZN7example10first_char17h4318683472f884ccE:
.cfi_startproc
test rsi, rsi
je .LBB0_1
movzx ecx, byte ptr [rdi]
test cl, cl
js .LBB0_3
mov eax, ecx
ret
.LBB0_1:
mov eax, 1114112
ret
.LBB0_3:
mov eax, ecx
and eax, 31
movzx esi, byte ptr [rdi + 1]
and esi, 63
cmp cl, -33
jbe .LBB0_4
movzx edx, byte ptr [rdi + 2]
shl esi, 6
and edx, 63
or edx, esi
cmp cl, -16
jb .LBB0_7
movzx ecx, byte ptr [rdi + 3]
and eax, 7
shl eax, 18
shl edx, 6
and ecx, 63
or ecx, edx
or eax, ecx
ret
.LBB0_4:
shl eax, 6
or eax, esi
ret
.LBB0_7:
shl eax, 12
or eax, edx
ret
```
The functions are now `unsafe` and they use `Option::unwrap_unchecked` instead of `unwrap_or_0`
`unwrap_or_0` was added in 42357d772b. I guess `unwrap_unchecked` was not available back then.
Given this example:
```rust
pub fn first_char(s: &str) -> Option<char> {
s.chars().next()
}
```
Previously, the following assembly was produced:
```asm
_ZN7example10first_char17ha056ddea6bafad1cE:
.cfi_startproc
test rsi, rsi
je .LBB0_1
movzx edx, byte ptr [rdi]
test dl, dl
js .LBB0_3
mov eax, edx
ret
.LBB0_1:
mov eax, 1114112
ret
.LBB0_3:
lea r8, [rdi + rsi]
xor eax, eax
mov r9, r8
cmp rsi, 1
je .LBB0_5
movzx eax, byte ptr [rdi + 1]
add rdi, 2
and eax, 63
mov r9, rdi
.LBB0_5:
mov ecx, edx
and ecx, 31
cmp dl, -33
jbe .LBB0_6
cmp r9, r8
je .LBB0_9
movzx esi, byte ptr [r9]
add r9, 1
and esi, 63
shl eax, 6
or eax, esi
cmp dl, -16
jb .LBB0_12
.LBB0_13:
cmp r9, r8
je .LBB0_14
movzx edx, byte ptr [r9]
and edx, 63
jmp .LBB0_16
.LBB0_6:
shl ecx, 6
or eax, ecx
ret
.LBB0_9:
xor esi, esi
mov r9, r8
shl eax, 6
or eax, esi
cmp dl, -16
jae .LBB0_13
.LBB0_12:
shl ecx, 12
or eax, ecx
ret
.LBB0_14:
xor edx, edx
.LBB0_16:
and ecx, 7
shl ecx, 18
shl eax, 6
or eax, ecx
or eax, edx
ret
```
After this change, the assembly is reduced to:
```asm
_ZN7example10first_char17h4318683472f884ccE:
.cfi_startproc
test rsi, rsi
je .LBB0_1
movzx ecx, byte ptr [rdi]
test cl, cl
js .LBB0_3
mov eax, ecx
ret
.LBB0_1:
mov eax, 1114112
ret
.LBB0_3:
mov eax, ecx
and eax, 31
movzx esi, byte ptr [rdi + 1]
and esi, 63
cmp cl, -33
jbe .LBB0_4
movzx edx, byte ptr [rdi + 2]
shl esi, 6
and edx, 63
or edx, esi
cmp cl, -16
jb .LBB0_7
movzx ecx, byte ptr [rdi + 3]
and eax, 7
shl eax, 18
shl edx, 6
and ecx, 63
or ecx, edx
or eax, ecx
ret
.LBB0_4:
shl eax, 6
or eax, esi
ret
.LBB0_7:
shl eax, 12
or eax, edx
ret
```
Rollup of 4 pull requests
Successful merges:
- #91008 (Adds IEEE 754-2019 minimun and maximum functions for f32/f64)
- #91070 (Make `LLVMRustGetOrInsertGlobal` always return a `GlobalVariable`)
- #91097 (Add spaces in opaque `impl Trait` with more than one trait)
- #91098 (Don't suggest certain fixups (`.field`, `.await`, etc) when reporting errors while matching on arrays )
Failed merges:
r? `@ghost`
`@rustbot` modify labels: rollup
Don't suggest certain fixups (`.field`, `.await`, etc) when reporting errors while matching on arrays
When we have a type mismatch with a `cause.code` that is an `ObligationCauseCode::Pattern`, skip suggesting fixes like adding `.await` or accessing a struct's `.field` if the pattern's `root_ty` differs from the `expected` ty. This occurs in situations like this:
```rust
struct S(());
fn main() {
let array = [S(())];
match array {
[()] => {}
_ => {}
}
}
```
I think what's happening here is a layer of `[_; N]` is peeled off of both types and we end up seeing the mismatch between just `S` and `()`, but when we suggest a fixup, that applies to the expression with type `root_ty`.
---
Questions:
1. Should this check live here, above all of the suggestions, or should I push this down into every suggestion when we match `ObligationCauseCode`?
2. Any other `ObligationCauseCode`s to check here?
3. Am I overlooking an easier way to get to this same conclusion without pattern matching on `ObligationCauseCode` and comparing `root_ty`?
Fixes#91058
Make `LLVMRustGetOrInsertGlobal` always return a `GlobalVariable`
`Module::getOrInsertGlobal` returns a `Constant*`, which is a super
class of `GlobalVariable`, but if the given type doesn't match an
existing declaration, it returns a bitcast of that global instead.
This causes UB when we pass that to `LLVMGetVisibility` which
unconditionally casts the opaque argument to a `GlobalValue*`.
Instead, we can do our own get-or-insert without worrying whether
existing types match exactly. It's not relevant when we're just trying
to get/set the linkage and visibility, and if types are needed we can
bitcast or error nicely from `rustc_codegen_llvm` instead.
Fixes#91050, fixes#87933, fixes#87813.
Adds IEEE 754-2019 minimun and maximum functions for f32/f64
IEEE 754-2019 removed the `minNum` (`min` in Rust) and `maxNum` (`max` in Rust) operations in favor of the newly created `minimum` and `maximum` operations due to their [non-associativity](https://grouper.ieee.org/groups/msc/ANSI_IEEE-Std-754-2019/background/minNum_maxNum_Removal_Demotion_v3.pdf) that cannot be fix in a backwards compatible manner. This PR adds `fN::{minimun,maximum}` functions following the new rules.
### IEEE 754-2019 Rules
> **minimum(x, y)** is x if x < y, y if y < x, and a quiet NaN if either operand is a NaN, according to 6.2.
For this operation, −0 compares less than +0. Otherwise (i.e., when x = y and signs are the same)
it is either x or y.
> **maximum(x, y)** is x if x > y, y if y > x, and a quiet NaN if either operand is a NaN, according to 6.2.
For this operation, +0 compares greater than −0. Otherwise (i.e., when x = y and signs are the
same) it is either x or y.
"IEEE Standard for Floating-Point Arithmetic," in IEEE Std 754-2019 (Revision of IEEE 754-2008) , vol., no., pp.1-84, 22 July 2019, doi: 10.1109/IEEESTD.2019.8766229.
### Implementation
This implementation is inspired by the one in [`glibc` ](90f0ac10a7/math/s_fminimum_template.c) (it self derived from the C2X draft) expect that:
- it doesn't use `copysign` because it's not available in `core` and also because `copysign` is unnecessary (we only want to check the sign, no need to create a new float)
- it also prefer `other > self` instead of `self < other` like IEEE 754-2019 does
I originally tried to implement them [using intrinsics](1d8aa13bc3) but LLVM [error out](https://godbolt.org/z/7sMrxW49a) when trying to lower them to machine intructions, GCC doesn't yet have built-ins for them, only cranelift support them nativelly (as it doesn't support the nativelly the old sementics).
Helps with https://github.com/rust-lang/rust/issues/83984
A byproduct of using `<details>` and `<summary>` to show/hide detailed
documentation was that clicking any part of a method heading (or impl
heading) would show or hide the documentation. This was not super
noticeable because clicking a link inside the method heading would
navigate to that link. But clicking any unlinked black text in a method
heading would trigger the behavior.
That behavior was somewhat unexpected, and means that if you try to click
a type name in a method heading, but miss by a few pixels, you get a
confusing surprise.
This change inhibits that behavior by putting an event listener on most
summaries that cancels the event unless the event target was the summary
itself. In practice, that means it cancels the event unless the target
was the "[+]" / "[-]", because the rest of the heading is wrapped inside
a `<div>`, which is the target for anything that doesn't have a more
specific target.
Point at source of trait bound obligations in more places
Be more thorough in using `ItemObligation` and `BindingObligation` when
evaluating obligations so that we can point at trait bounds that
introduced unfulfilled obligations. We no longer incorrectly point at
unrelated trait bounds (`substs-ppaux.verbose.stderr`).
In particular, we now point at trait bounds on method calls.
We no longer point at "obvious" obligation sources (we no longer have a
note pointing at `Trait` saying "required by a bound in `Trait`", like
in `associated-types-no-suitable-supertrait*`).
We no longer point at associated items (`ImplObligation`), as they didn't
add any user actionable information, they just added noise.
Address part of #89418.
Rollup of 6 pull requests
Successful merges:
- #89741 (Mark `Arc::from_inner` / `Rc::from_inner` as unsafe)
- #90927 (Fix float ICE)
- #90994 (Fix ICE `#90993`: add missing call to cancel)
- #91018 (Adopt let_else in more places in rustc_mir_build)
- #91022 (Suggest `await` in more situations where infer types are involved)
- #91088 (Revert "require full validity when determining the discriminant of a value")
Failed merges:
r? `@ghost`
`@rustbot` modify labels: rollup
Suggest `await` in more situations where infer types are involved
Currently we use `TyS::same_type` in diagnostics that suggest adding `.await` to opaque future types.
This change makes the suggestion slightly more general, when we're comparing types like `Result<T, E>` and `Result<_, _>` which happens sometimes in places like `match` patterns or `let` statements with partially-elaborated types.
----
Question:
1. Is this change worthwhile? Totally fine if it doesn't make sense adding.
2. Should `same_type_modulo_infer` live in `rustc_infer::infer::error_reporting` or alongside the other method in `rustc_middle::ty::util`?
3. Should we generalize this change? I wanted to change all usages, but I don't want erroneous suggestions when adding `.field_name`...
Mark `Arc::from_inner` / `Rc::from_inner` as unsafe
While it's an internal function, it is easy to create invalid Arc/Rcs to
a dangling pointer with it.
Fixes https://github.com/rust-lang/rust/issues/89740
