add rustc option for using LLVM stack smash protection

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 <[email protected]>

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.

Author: bbjornse

compiler/rustc_codegen_llvm/src/attributes.rs

@@ -12,7 +12,7 @@ use rustc_middle::ty::{self, TyCtxt};
 use rustc_session::config::OptLevel;
 use rustc_session::Session;
 use rustc_target::spec::abi::Abi;
-use rustc_target::spec::{FramePointer, SanitizerSet, StackProbeType};
+use rustc_target::spec::{FramePointer, SanitizerSet, StackProbeType, StackProtector};
 
 use crate::attributes;
 use crate::llvm::AttributePlace::Function;
@@ -161,6 +161,17 @@ fn set_probestack(cx: &CodegenCx<'ll, '_>, llfn: &'ll Value) {
     }
 }
 
+fn set_stackprotector(cx: &CodegenCx<'ll, '_>, llfn: &'ll Value) {
+    let sspattr = match cx.sess().stack_protector() {
+        StackProtector::None => return,
+        StackProtector::All => Attribute::StackProtectReq,
+        StackProtector::Strong => Attribute::StackProtectStrong,
+        StackProtector::Basic => Attribute::StackProtect,
+    };
+
+    sspattr.apply_llfn(Function, llfn)
+}
+
 pub fn apply_target_cpu_attr(cx: &CodegenCx<'ll, '_>, llfn: &'ll Value) {
     let target_cpu = SmallCStr::new(llvm_util::target_cpu(cx.tcx.sess));
     llvm::AddFunctionAttrStringValue(
@@ -271,6 +282,7 @@ pub fn from_fn_attrs(cx: &CodegenCx<'ll, 'tcx>, llfn: &'ll Value, instance: ty::
     set_frame_pointer_type(cx, llfn);
     set_instrument_function(cx, llfn);
     set_probestack(cx, llfn);
+    set_stackprotector(cx, llfn);
 
     if codegen_fn_attrs.flags.contains(CodegenFnAttrFlags::COLD) {
         Attribute::Cold.apply_llfn(Function, llfn);

compiler/rustc_codegen_llvm/src/lib.rs

@@ -288,6 +288,31 @@ impl CodegenBackend for LlvmCodegenBackend {
                 }
                 println!();
             }
+            PrintRequest::StackProtectorStrategies => {
+                println!(
+                    r#"Available stack protector strategies:
+    all
+        Generate stack canaries in all functions.
+
+    strong
+        Generate stack canaries in a function if it either:
+        - has a local variable of `[T; N]` type, regardless of `T` and `N`
+        - takes the address of a local variable.
+
+          (Note that a local variable being borrowed is not equivalent to its
+          address being taken: e.g. some borrows may be removed by optimization,
+          while by-value argument passing may be implemented with reference to a
+          local stack variable in the ABI.)
+
+    basic
+        Generate stack canaries in functions with:
+        - local variables of `[T; N]` type, where `T` is byte-sized and `N` > 8.
+
+    none
+        Do not generate stack canaries.
+"#
+                );
+            }
             req => llvm_util::print(req, sess),
         }
     }

compiler/rustc_codegen_llvm/src/llvm/ffi.rs

@@ -166,6 +166,9 @@ pub enum Attribute {
     InaccessibleMemOnly = 27,
     SanitizeHWAddress = 28,
     WillReturn = 29,
+    StackProtectReq = 30,
+    StackProtectStrong = 31,
+    StackProtect = 32,
 }
 
 /// LLVMIntPredicate

compiler/rustc_driver/src/lib.rs

@@ -736,7 +736,12 @@ impl RustcDefaultCalls {
                         println!("{}", cfg);
                     }
                 }
-                RelocationModels | CodeModels | TlsModels | TargetCPUs | TargetFeatures => {
+                RelocationModels
+                | CodeModels
+                | TlsModels
+                | TargetCPUs
+                | StackProtectorStrategies
+                | TargetFeatures => {
                     codegen_backend.print(*req, sess);
                 }
                 // Any output here interferes with Cargo's parsing of other printed output

compiler/rustc_interface/src/tests.rs

@@ -20,7 +20,9 @@ use rustc_span::edition::{Edition, DEFAULT_EDITION};
 use rustc_span::symbol::sym;
 use rustc_span::SourceFileHashAlgorithm;
 use rustc_target::spec::{CodeModel, LinkerFlavor, MergeFunctions, PanicStrategy};
-use rustc_target::spec::{RelocModel, RelroLevel, SanitizerSet, SplitDebuginfo, TlsModel};
+use rustc_target::spec::{
+    RelocModel, RelroLevel, SanitizerSet, SplitDebuginfo, StackProtector, TlsModel,
+};
 
 use std::collections::{BTreeMap, BTreeSet};
 use std::iter::FromIterator;
@@ -713,8 +715,8 @@ fn test_debugging_options_tracking_hash() {
     // This list is in alphabetical order.
     tracked!(allow_features, Some(vec![String::from("lang_items")]));
     tracked!(always_encode_mir, true);
-    tracked!(assume_incomplete_release, true);
     tracked!(asm_comments, true);
+    tracked!(assume_incomplete_release, true);
     tracked!(binary_dep_depinfo, true);
     tracked!(chalk, true);
     tracked!(codegen_backend, Some("abc".to_string()));
@@ -731,8 +733,8 @@ fn test_debugging_options_tracking_hash() {
     tracked!(human_readable_cgu_names, true);
     tracked!(inline_in_all_cgus, Some(true));
     tracked!(inline_mir, Some(true));
-    tracked!(inline_mir_threshold, Some(123));
     tracked!(inline_mir_hint_threshold, Some(123));
+    tracked!(inline_mir_threshold, Some(123));
     tracked!(instrument_coverage, Some(InstrumentCoverage::All));
     tracked!(instrument_mcount, true);
     tracked!(link_only, true);
@@ -764,23 +766,24 @@ fn test_debugging_options_tracking_hash() {
     tracked!(relax_elf_relocations, Some(true));
     tracked!(relro_level, Some(RelroLevel::Full));
     tracked!(remap_cwd_prefix, Some(PathBuf::from("abc")));
-    tracked!(simulate_remapped_rust_src_base, Some(PathBuf::from("/rustc/abc")));
     tracked!(report_delayed_bugs, true);
     tracked!(sanitizer, SanitizerSet::ADDRESS);
     tracked!(sanitizer_memory_track_origins, 2);
     tracked!(sanitizer_recover, SanitizerSet::ADDRESS);
     tracked!(saturating_float_casts, Some(true));
     tracked!(share_generics, Some(true));
     tracked!(show_span, Some(String::from("abc")));
+    tracked!(simulate_remapped_rust_src_base, Some(PathBuf::from("/rustc/abc")));
     tracked!(src_hash_algorithm, Some(SourceFileHashAlgorithm::Sha1));
+    tracked!(stack_protector, StackProtector::All);
     tracked!(symbol_mangling_version, Some(SymbolManglingVersion::V0));
     tracked!(teach, true);
     tracked!(thinlto, Some(true));
     tracked!(thir_unsafeck, true);
-    tracked!(tune_cpu, Some(String::from("abc")));
     tracked!(tls_model, Some(TlsModel::GeneralDynamic));
     tracked!(trap_unreachable, Some(false));
     tracked!(treat_err_as_bug, NonZeroUsize::new(1));
+    tracked!(tune_cpu, Some(String::from("abc")));
     tracked!(unleash_the_miri_inside_of_you, true);
     tracked!(use_ctors_section, Some(true));
     tracked!(verify_llvm_ir, true);

compiler/rustc_llvm/llvm-wrapper/LLVMWrapper.h

@@ -79,6 +79,9 @@ enum LLVMRustAttribute {
   InaccessibleMemOnly = 27,
   SanitizeHWAddress = 28,
   WillReturn = 29,
+  StackProtectReq = 30,
+  StackProtectStrong = 31,
+  StackProtect = 32,
 };
 
 typedef struct OpaqueRustString *RustStringRef;

compiler/rustc_llvm/llvm-wrapper/RustWrapper.cpp

@@ -213,6 +213,12 @@ static Attribute::AttrKind fromRust(LLVMRustAttribute Kind) {
     return Attribute::SanitizeHWAddress;
   case WillReturn:
     return Attribute::WillReturn;
+  case StackProtectReq:
+    return Attribute::StackProtectReq;
+  case StackProtectStrong:
+    return Attribute::StackProtectStrong;
+  case StackProtect:
+    return Attribute::StackProtect;
   }
   report_fatal_error("bad AttributeKind");
 }

compiler/rustc_session/src/config.rs

@@ -538,6 +538,7 @@ pub enum PrintRequest {
     TlsModels,
     TargetSpec,
     NativeStaticLibs,
+    StackProtectorStrategies,
 }
 
 #[derive(Copy, Clone)]
@@ -1110,8 +1111,8 @@ pub fn rustc_short_optgroups() -> Vec<RustcOptGroup> {
             "print",
             "Compiler information to print on stdout",
             "[crate-name|file-names|sysroot|target-libdir|cfg|target-list|\
-             target-cpus|target-features|relocation-models|\
-             code-models|tls-models|target-spec-json|native-static-libs]",
+             target-cpus|target-features|relocation-models|code-models|\
+             tls-models|target-spec-json|native-static-libs|stack-protector-strategies]",
         ),
         opt::flagmulti_s("g", "", "Equivalent to -C debuginfo=2"),
         opt::flagmulti_s("O", "", "Equivalent to -C opt-level=2"),
@@ -1527,6 +1528,7 @@ fn collect_print_requests(
         "code-models" => PrintRequest::CodeModels,
         "tls-models" => PrintRequest::TlsModels,
         "native-static-libs" => PrintRequest::NativeStaticLibs,
+        "stack-protector-strategies" => PrintRequest::StackProtectorStrategies,
         "target-spec-json" => {
             if dopts.unstable_options {
                 PrintRequest::TargetSpec
@@ -2494,7 +2496,9 @@ crate mod dep_tracking {
     use rustc_span::edition::Edition;
     use rustc_span::RealFileName;
     use rustc_target::spec::{CodeModel, MergeFunctions, PanicStrategy, RelocModel};
-    use rustc_target::spec::{RelroLevel, SanitizerSet, SplitDebuginfo, TargetTriple, TlsModel};
+    use rustc_target::spec::{
+        RelroLevel, SanitizerSet, SplitDebuginfo, StackProtector, TargetTriple, TlsModel,
+    };
     use std::collections::hash_map::DefaultHasher;
     use std::collections::BTreeMap;
     use std::hash::Hash;
@@ -2568,6 +2572,7 @@ crate mod dep_tracking {
         Edition,
         LinkerPluginLto,
         SplitDebuginfo,
+        StackProtector,
         SwitchWithOptPath,
         SymbolManglingVersion,
         SourceFileHashAlgorithm,

compiler/rustc_session/src/options.rs

@@ -5,7 +5,9 @@ use crate::lint;
 use crate::search_paths::SearchPath;
 use crate::utils::NativeLib;
 use rustc_target::spec::{CodeModel, LinkerFlavor, MergeFunctions, PanicStrategy, SanitizerSet};
-use rustc_target::spec::{RelocModel, RelroLevel, SplitDebuginfo, TargetTriple, TlsModel};
+use rustc_target::spec::{
+    RelocModel, RelroLevel, SplitDebuginfo, StackProtector, TargetTriple, TlsModel,
+};
 
 use rustc_feature::UnstableFeatures;
 use rustc_span::edition::Edition;
@@ -385,6 +387,8 @@ mod desc {
     pub const parse_split_debuginfo: &str =
         "one of supported split-debuginfo modes (`off`, `packed`, or `unpacked`)";
     pub const parse_gcc_ld: &str = "one of: no value, `lld`";
+    pub const parse_stack_protector: &str =
+        "one of (`none` (default), `basic`, `strong`, or `all`)";
 }
 
 mod parse {
@@ -917,6 +921,14 @@ mod parse {
         }
         true
     }
+
+    crate fn parse_stack_protector(slot: &mut StackProtector, v: Option<&str>) -> bool {
+        match v.and_then(|s| StackProtector::from_str(s).ok()) {
+            Some(ssp) => *slot = ssp,
+            _ => return false,
+        }
+        true
+    }
 }
 
 options! {
@@ -1330,6 +1342,8 @@ options! {
         "exclude spans when debug-printing compiler state (default: no)"),
     src_hash_algorithm: Option<SourceFileHashAlgorithm> = (None, parse_src_file_hash, [TRACKED],
         "hash algorithm of source files in debug info (`md5`, `sha1`, or `sha256`)"),
+    stack_protector: StackProtector = (StackProtector::None, parse_stack_protector, [TRACKED],
+        "control stack smash protection strategy (`rustc --print stack-protector-strategies` for details)"),
     strip: Strip = (Strip::None, parse_strip, [UNTRACKED],
         "tell the linker which information to strip (`none` (default), `debuginfo` or `symbols`)"),
     split_dwarf_inlining: bool = (true, parse_bool, [UNTRACKED],

compiler/rustc_session/src/session.rs

@@ -27,7 +27,9 @@ use rustc_span::source_map::{FileLoader, MultiSpan, RealFileLoader, SourceMap, S
 use rustc_span::{sym, SourceFileHashAlgorithm, Symbol};
 use rustc_target::asm::InlineAsmArch;
 use rustc_target::spec::{CodeModel, PanicStrategy, RelocModel, RelroLevel};
-use rustc_target::spec::{SanitizerSet, SplitDebuginfo, Target, TargetTriple, TlsModel};
+use rustc_target::spec::{
+    SanitizerSet, SplitDebuginfo, StackProtector, Target, TargetTriple, TlsModel,
+};
 
 use std::cell::{self, RefCell};
 use std::env;
@@ -732,6 +734,14 @@ impl Session {
         self.opts.cg.split_debuginfo.unwrap_or(self.target.split_debuginfo)
     }
 
+    pub fn stack_protector(&self) -> StackProtector {
+        if self.target.options.supports_stack_protector {
+            self.opts.debugging_opts.stack_protector
+        } else {
+            StackProtector::None
+        }
+    }
+
     pub fn target_can_use_split_dwarf(&self) -> bool {
         !self.target.is_like_windows && !self.target.is_like_osx
     }
@@ -1411,6 +1421,15 @@ fn validate_commandline_args_with_session_available(sess: &Session) {
             sess.err("`-Zsanitizer=cfi` requires `-Clto`");
         }
     }
+
+    if sess.opts.debugging_opts.stack_protector != StackProtector::None {
+        if !sess.target.options.supports_stack_protector {
+            sess.warn(&format!(
+                "`-Z stack-protector={}` is not supported for target {} and will be ignored",
+                sess.opts.debugging_opts.stack_protector, sess.opts.target_triple
+            ))
+        }
+    }
 }
 
 /// Holds data on the current incremental compilation session, if there is one.

compiler/rustc_target/src/spec/mod.rs

@@ -712,6 +712,59 @@ impl ToJson for FramePointer {
     }
 }
 
+/// Controls use of stack canaries.
+#[derive(Clone, Copy, Debug, PartialEq, Hash, Eq)]
+pub enum StackProtector {
+    /// Disable stack canary generation.
+    None,
+
+    /// On LLVM, mark all generated LLVM functions with the `ssp` attribute (see
+    /// llvm/docs/LangRef.rst). This triggers stack canary generation in
+    /// functions which contain an array of a byte-sized type with more than
+    /// eight elements.
+    Basic,
+
+    /// On LLVM, mark all generated LLVM functions with the `sspstrong`
+    /// attribute (see llvm/docs/LangRef.rst). This triggers stack canary
+    /// generation in functions which either contain an array, or which take
+    /// the address of a local variable.
+    Strong,
+
+    /// Generate stack canaries in all functions.
+    All,
+}
+
+impl StackProtector {
+    fn as_str(&self) -> &'static str {
+        match self {
+            StackProtector::None => "none",
+            StackProtector::Basic => "basic",
+            StackProtector::Strong => "strong",
+            StackProtector::All => "all",
+        }
+    }
+}
+
+impl FromStr for StackProtector {
+    type Err = ();
+
+    fn from_str(s: &str) -> Result<StackProtector, ()> {
+        Ok(match s {
+            "none" => StackProtector::None,
+            "basic" => StackProtector::Basic,
+            "strong" => StackProtector::Strong,
+            "all" => StackProtector::All,
+            _ => return Err(()),
+        })
+    }
+}
+
+impl fmt::Display for StackProtector {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        f.write_str(self.as_str())
+    }
+}
+
 macro_rules! supported_targets {
     ( $(($( $triple:literal, )+ $module:ident ),)+ ) => {
         $(mod $module;)+
@@ -1360,6 +1413,10 @@ pub struct TargetOptions {
 
     /// Whether or not the DWARF `.debug_aranges` section should be generated.
     pub generate_arange_section: bool,
+
+    /// Whether the target supports stack canary checks. `true` by default,
+    /// since this is most common among tier 1 and tier 2 targets.
+    pub supports_stack_protector: bool,
 }
 
 impl Default for TargetOptions {
@@ -1466,6 +1523,7 @@ impl Default for TargetOptions {
             default_adjusted_cabi: None,
             c_enum_min_bits: 32,
             generate_arange_section: true,
+            supports_stack_protector: true,
         }
     }
 }
@@ -2052,6 +2110,7 @@ impl Target {
         key!(default_adjusted_cabi, Option<Abi>)?;
         key!(c_enum_min_bits, u64);
         key!(generate_arange_section, bool);
+        key!(supports_stack_protector, bool);
 
         if base.is_builtin {
             // This can cause unfortunate ICEs later down the line.
@@ -2292,6 +2351,7 @@ impl ToJson for Target {
         target_option_val!(supported_sanitizers);
         target_option_val!(c_enum_min_bits);
         target_option_val!(generate_arange_section);
+        target_option_val!(supports_stack_protector);
 
         if let Some(abi) = self.default_adjusted_cabi {
             d.insert("default-adjusted-cabi".to_string(), Abi::name(abi).to_json());