Auto merge of #99679 - repnop:kernel-address-sanitizer, r=cuviper

Add `kernel-address` sanitizer support for freestanding targets

This PR adds support for KASan (kernel address sanitizer) instrumentation in freestanding targets. I included the minimal set of `x86_64-unknown-none`, `riscv64{imac, gc}-unknown-none-elf`, and `aarch64-unknown-none` but there's likely other targets it can be added to. (`linux_kernel_base.rs`?) KASan uses the address sanitizer attributes but has the `CompileKernel` parameter set to `true` in the pass creation.

18 files changed, 142 insertions, 12 deletions

diff --git a/compiler/rustc_codegen_llvm/src/attributes.rs b/compiler/rustc_codegen_llvm/src/attributes.rs
index 7a4ec494c8e..651d644ebb6 100644
--- a/compiler/rustc_codegen_llvm/src/attributes.rs
+++ b/compiler/rustc_codegen_llvm/src/attributes.rs
@@ -62,7 +62,7 @@ pub fn sanitize_attrs<'ll>(
 ) -> SmallVec<[&'ll Attribute; 4]> {
     let mut attrs = SmallVec::new();
     let enabled = cx.tcx.sess.opts.unstable_opts.sanitizer - no_sanitize;
-    if enabled.contains(SanitizerSet::ADDRESS) {
+    if enabled.contains(SanitizerSet::ADDRESS) || enabled.contains(SanitizerSet::KERNELADDRESS) {
         attrs.push(llvm::AttributeKind::SanitizeAddress.create_attr(cx.llcx));
     }
     if enabled.contains(SanitizerSet::MEMORY) {
diff --git a/compiler/rustc_codegen_llvm/src/back/write.rs b/compiler/rustc_codegen_llvm/src/back/write.rs
index fb160669436..40f0594b40d 100644
--- a/compiler/rustc_codegen_llvm/src/back/write.rs
+++ b/compiler/rustc_codegen_llvm/src/back/write.rs
@@ -442,6 +442,10 @@ pub(crate) unsafe fn llvm_optimize(
             sanitize_thread: config.sanitizer.contains(SanitizerSet::THREAD),
             sanitize_hwaddress: config.sanitizer.contains(SanitizerSet::HWADDRESS),
             sanitize_hwaddress_recover: config.sanitizer_recover.contains(SanitizerSet::HWADDRESS),
+            sanitize_kernel_address: config.sanitizer.contains(SanitizerSet::KERNELADDRESS),
+            sanitize_kernel_address_recover: config
+                .sanitizer_recover
+                .contains(SanitizerSet::KERNELADDRESS),
         })
     } else {
         None
diff --git a/compiler/rustc_codegen_llvm/src/llvm/ffi.rs b/compiler/rustc_codegen_llvm/src/llvm/ffi.rs
index 8b4861962b2..7aab666fc5e 100644
--- a/compiler/rustc_codegen_llvm/src/llvm/ffi.rs
+++ b/compiler/rustc_codegen_llvm/src/llvm/ffi.rs
@@ -482,6 +482,8 @@ pub struct SanitizerOptions {
     pub sanitize_thread: bool,
     pub sanitize_hwaddress: bool,
     pub sanitize_hwaddress_recover: bool,
+    pub sanitize_kernel_address: bool,
+    pub sanitize_kernel_address_recover: bool,
 }
 
 /// LLVMRelocMode
diff --git a/compiler/rustc_codegen_ssa/src/codegen_attrs.rs b/compiler/rustc_codegen_ssa/src/codegen_attrs.rs
index 3ebbb2cbdfb..7d5c0048626 100644
--- a/compiler/rustc_codegen_ssa/src/codegen_attrs.rs
+++ b/compiler/rustc_codegen_ssa/src/codegen_attrs.rs
@@ -295,7 +295,8 @@ fn codegen_fn_attrs(tcx: TyCtxt<'_>, did: DefId) -> CodegenFnAttrs {
             if let Some(list) = attr.meta_item_list() {
                 for item in list.iter() {
                     if item.has_name(sym::address) {
-                        codegen_fn_attrs.no_sanitize |= SanitizerSet::ADDRESS;
+                        codegen_fn_attrs.no_sanitize |=
+                            SanitizerSet::ADDRESS | SanitizerSet::KERNELADDRESS;
                     } else if item.has_name(sym::cfi) {
                         codegen_fn_attrs.no_sanitize |= SanitizerSet::CFI;
                     } else if item.has_name(sym::kcfi) {
diff --git a/compiler/rustc_llvm/llvm-wrapper/PassWrapper.cpp b/compiler/rustc_llvm/llvm-wrapper/PassWrapper.cpp
index fbc1d8ef310..0a42265a6ba 100644
--- a/compiler/rustc_llvm/llvm-wrapper/PassWrapper.cpp
+++ b/compiler/rustc_llvm/llvm-wrapper/PassWrapper.cpp
@@ -594,6 +594,8 @@ struct LLVMRustSanitizerOptions {
   bool SanitizeThread;
   bool SanitizeHWAddress;
   bool SanitizeHWAddressRecover;
+  bool SanitizeKernelAddress;
+  bool SanitizeKernelAddressRecover;
 };
 
 extern "C" LLVMRustResult
@@ -765,15 +767,17 @@ LLVMRustOptimize(
       );
     }
 
-    if (SanitizerOptions->SanitizeAddress) {
+    if (SanitizerOptions->SanitizeAddress || SanitizerOptions->SanitizeKernelAddress) {
       OptimizerLastEPCallbacks.push_back(
         [SanitizerOptions](ModulePassManager &MPM, OptimizationLevel Level) {
+          auto CompileKernel = SanitizerOptions->SanitizeKernelAddress;
 #if LLVM_VERSION_LT(15, 0)
           MPM.addPass(RequireAnalysisPass<ASanGlobalsMetadataAnalysis, Module>());
 #endif
           AddressSanitizerOptions opts = AddressSanitizerOptions{
-            /*CompileKernel=*/false,
-            SanitizerOptions->SanitizeAddressRecover,
+            CompileKernel,
+            SanitizerOptions->SanitizeAddressRecover
+              || SanitizerOptions->SanitizeKernelAddressRecover,
             /*UseAfterScope=*/true,
             AsanDetectStackUseAfterReturnMode::Runtime,
           };
diff --git a/compiler/rustc_session/src/config.rs b/compiler/rustc_session/src/config.rs
index 4da6acad2c0..295e93f6103 100644
--- a/compiler/rustc_session/src/config.rs
+++ b/compiler/rustc_session/src/config.rs
@@ -1022,7 +1022,13 @@ fn default_configuration(sess: &Session) -> CrateConfig {
     let panic_strategy = sess.panic_strategy();
     ret.insert((sym::panic, Some(panic_strategy.desc_symbol())));
 
-    for s in sess.opts.unstable_opts.sanitizer {
+    for mut s in sess.opts.unstable_opts.sanitizer {
+        // KASAN should use the same attribute name as ASAN, as it's still ASAN
+        // under the hood
+        if s == SanitizerSet::KERNELADDRESS {
+            s = SanitizerSet::ADDRESS;
+        }
+
         let symbol = Symbol::intern(&s.to_string());
         ret.insert((sym::sanitize, Some(symbol)));
     }
diff --git a/compiler/rustc_session/src/options.rs b/compiler/rustc_session/src/options.rs
index c784582012a..0d5818bd39c 100644
--- a/compiler/rustc_session/src/options.rs
+++ b/compiler/rustc_session/src/options.rs
@@ -370,7 +370,7 @@ mod desc {
     pub const parse_opt_panic_strategy: &str = parse_panic_strategy;
     pub const parse_oom_strategy: &str = "either `panic` or `abort`";
     pub const parse_relro_level: &str = "one of: `full`, `partial`, or `off`";
-    pub const parse_sanitizers: &str = "comma separated list of sanitizers: `address`, `cfi`, `hwaddress`, `kcfi`, `leak`, `memory`, `memtag`, `shadow-call-stack`, or `thread`";
+    pub const parse_sanitizers: &str = "comma separated list of sanitizers: `address`, `cfi`, `hwaddress`, `kcfi`, `kernel-address`, `leak`, `memory`, `memtag`, `shadow-call-stack`, or `thread`";
     pub const parse_sanitizer_memory_track_origins: &str = "0, 1, or 2";
     pub const parse_cfguard: &str =
         "either a boolean (`yes`, `no`, `on`, `off`, etc), `checks`, or `nochecks`";
@@ -684,6 +684,7 @@ mod parse {
                     "address" => SanitizerSet::ADDRESS,
                     "cfi" => SanitizerSet::CFI,
                     "kcfi" => SanitizerSet::KCFI,
+                    "kernel-address" => SanitizerSet::KERNELADDRESS,
                     "leak" => SanitizerSet::LEAK,
                     "memory" => SanitizerSet::MEMORY,
                     "memtag" => SanitizerSet::MEMTAG,
diff --git a/compiler/rustc_session/src/session.rs b/compiler/rustc_session/src/session.rs
index e608b9fe0b3..3dc09854b3c 100644
--- a/compiler/rustc_session/src/session.rs
+++ b/compiler/rustc_session/src/session.rs
@@ -954,10 +954,10 @@ impl Session {
     /// Checks if LLVM lifetime markers should be emitted.
     pub fn emit_lifetime_markers(&self) -> bool {
         self.opts.optimize != config::OptLevel::No
-        // AddressSanitizer uses lifetimes to detect use after scope bugs.
+        // AddressSanitizer and KernelAddressSanitizer uses lifetimes to detect use after scope bugs.
         // MemorySanitizer uses lifetimes to detect use of uninitialized stack variables.
         // HWAddressSanitizer will use lifetimes to detect use after scope bugs in the future.
-        || self.opts.unstable_opts.sanitizer.intersects(SanitizerSet::ADDRESS | SanitizerSet::MEMORY | SanitizerSet::HWADDRESS)
+        || self.opts.unstable_opts.sanitizer.intersects(SanitizerSet::ADDRESS | SanitizerSet::KERNELADDRESS | SanitizerSet::MEMORY | SanitizerSet::HWADDRESS)
     }
 
     pub fn is_proc_macro_attr(&self, attr: &Attribute) -> bool {
diff --git a/compiler/rustc_target/src/spec/aarch64_unknown_none.rs b/compiler/rustc_target/src/spec/aarch64_unknown_none.rs
index 30fbe6f3c15..be27302f7f5 100644
--- a/compiler/rustc_target/src/spec/aarch64_unknown_none.rs
+++ b/compiler/rustc_target/src/spec/aarch64_unknown_none.rs
@@ -15,7 +15,7 @@ pub fn target() -> Target {
         linker_flavor: LinkerFlavor::Gnu(Cc::No, Lld::Yes),
         linker: Some("rust-lld".into()),
         features: "+v8a,+strict-align,+neon,+fp-armv8".into(),
-        supported_sanitizers: SanitizerSet::KCFI,
+        supported_sanitizers: SanitizerSet::KCFI | SanitizerSet::KERNELADDRESS,
         relocation_model: RelocModel::Static,
         disable_redzone: true,
         max_atomic_width: Some(128),
diff --git a/compiler/rustc_target/src/spec/mod.rs b/compiler/rustc_target/src/spec/mod.rs
index ef60956a617..0d86a3032a6 100644
--- a/compiler/rustc_target/src/spec/mod.rs
+++ b/compiler/rustc_target/src/spec/mod.rs
@@ -812,6 +812,7 @@ bitflags::bitflags! {
         const MEMTAG  = 1 << 6;
         const SHADOWCALLSTACK = 1 << 7;
         const KCFI    = 1 << 8;
+        const KERNELADDRESS = 1 << 9;
     }
 }
 
@@ -824,6 +825,7 @@ impl SanitizerSet {
             SanitizerSet::ADDRESS => "address",
             SanitizerSet::CFI => "cfi",
             SanitizerSet::KCFI => "kcfi",
+            SanitizerSet::KERNELADDRESS => "kernel-address",
             SanitizerSet::LEAK => "leak",
             SanitizerSet::MEMORY => "memory",
             SanitizerSet::MEMTAG => "memtag",
@@ -866,6 +868,7 @@ impl IntoIterator for SanitizerSet {
             SanitizerSet::SHADOWCALLSTACK,
             SanitizerSet::THREAD,
             SanitizerSet::HWADDRESS,
+            SanitizerSet::KERNELADDRESS,
         ]
         .iter()
         .copied()
@@ -2341,6 +2344,7 @@ impl Target {
                                 Some("address") => SanitizerSet::ADDRESS,
                                 Some("cfi") => SanitizerSet::CFI,
                                 Some("kcfi") => SanitizerSet::KCFI,
+                                Some("kernel-address") => SanitizerSet::KERNELADDRESS,
                                 Some("leak") => SanitizerSet::LEAK,
                                 Some("memory") => SanitizerSet::MEMORY,
                                 Some("memtag") => SanitizerSet::MEMTAG,
diff --git a/compiler/rustc_target/src/spec/riscv64gc_unknown_none_elf.rs b/compiler/rustc_target/src/spec/riscv64gc_unknown_none_elf.rs
index 409b0b26961..ab3c14e3fe7 100644
--- a/compiler/rustc_target/src/spec/riscv64gc_unknown_none_elf.rs
+++ b/compiler/rustc_target/src/spec/riscv64gc_unknown_none_elf.rs
@@ -1,6 +1,8 @@
 use crate::spec::{Cc, CodeModel, LinkerFlavor, Lld, PanicStrategy};
 use crate::spec::{RelocModel, Target, TargetOptions};
 
+use super::SanitizerSet;
+
 pub fn target() -> Target {
     Target {
         data_layout: "e-m:e-p:64:64-i64:64-i128:128-n32:64-S128".into(),
@@ -20,6 +22,7 @@ pub fn target() -> Target {
             code_model: Some(CodeModel::Medium),
             emit_debug_gdb_scripts: false,
             eh_frame_header: false,
+            supported_sanitizers: SanitizerSet::KERNELADDRESS,
             ..Default::default()
         },
     }
diff --git a/compiler/rustc_target/src/spec/riscv64imac_unknown_none_elf.rs b/compiler/rustc_target/src/spec/riscv64imac_unknown_none_elf.rs
index 87aba9171b4..0f1821c9985 100644
--- a/compiler/rustc_target/src/spec/riscv64imac_unknown_none_elf.rs
+++ b/compiler/rustc_target/src/spec/riscv64imac_unknown_none_elf.rs
@@ -1,5 +1,5 @@
 use crate::spec::{Cc, CodeModel, LinkerFlavor, Lld, PanicStrategy};
-use crate::spec::{RelocModel, Target, TargetOptions};
+use crate::spec::{RelocModel, SanitizerSet, Target, TargetOptions};
 
 pub fn target() -> Target {
     Target {
@@ -19,6 +19,7 @@ pub fn target() -> Target {
             code_model: Some(CodeModel::Medium),
             emit_debug_gdb_scripts: false,
             eh_frame_header: false,
+            supported_sanitizers: SanitizerSet::KERNELADDRESS,
             ..Default::default()
         },
     }
diff --git a/compiler/rustc_target/src/spec/x86_64_unknown_none.rs b/compiler/rustc_target/src/spec/x86_64_unknown_none.rs
index 32060c35c11..43c5ce78ce3 100644
--- a/compiler/rustc_target/src/spec/x86_64_unknown_none.rs
+++ b/compiler/rustc_target/src/spec/x86_64_unknown_none.rs
@@ -20,7 +20,7 @@ pub fn target() -> Target {
         features:
             "-mmx,-sse,-sse2,-sse3,-ssse3,-sse4.1,-sse4.2,-3dnow,-3dnowa,-avx,-avx2,+soft-float"
                 .into(),
-        supported_sanitizers: SanitizerSet::KCFI,
+        supported_sanitizers: SanitizerSet::KCFI | SanitizerSet::KERNELADDRESS,
         disable_redzone: true,
         panic_strategy: PanicStrategy::Abort,
         code_model: Some(CodeModel::Kernel),
diff --git a/src/doc/unstable-book/src/compiler-flags/sanitizer.md b/src/doc/unstable-book/src/compiler-flags/sanitizer.md
index 70c3a445b86..262cef3454a 100644
--- a/src/doc/unstable-book/src/compiler-flags/sanitizer.md
+++ b/src/doc/unstable-book/src/compiler-flags/sanitizer.md
@@ -531,6 +531,24 @@ LLVM KCFI is supported on the following targets:
 See the [Clang KernelControlFlowIntegrity documentation][clang-kcfi] for more
 details.
 
+# KernelAddressSanitizer
+
+KernelAddressSanitizer (KASAN) is a freestanding version of AddressSanitizer
+which is suitable for detecting memory errors in programs which do not have a
+runtime environment, such as operating system kernels. KernelAddressSanitizer
+requires manual implementation of the underlying functions used for tracking
+KernelAddressSanitizer state.
+
+KernelAddressSanitizer is supported on the following targets:
+
+* `aarch64-unknown-none`
+* `riscv64gc-unknown-none-elf`
+* `riscv64imac-unknown-none-elf`
+* `x86_64-unknown-none`
+
+See the [Linux Kernel's KernelAddressSanitizer documentation][linux-kasan] for
+more details.
+
 # LeakSanitizer
 
 LeakSanitizer is run-time memory leak detector.
@@ -714,6 +732,7 @@ Sanitizers produce symbolized stacktraces when llvm-symbolizer binary is in `PAT
 * [AddressSanitizer in Clang][clang-asan]
 * [ControlFlowIntegrity in Clang][clang-cfi]
 * [HWAddressSanitizer in Clang][clang-hwasan]
+* [Linux Kernel's KernelAddressSanitizer documentation][linux-kasan]
 * [LeakSanitizer in Clang][clang-lsan]
 * [MemorySanitizer in Clang][clang-msan]
 * [MemTagSanitizer in LLVM][llvm-memtag]
@@ -727,4 +746,5 @@ Sanitizers produce symbolized stacktraces when llvm-symbolizer binary is in `PAT
 [clang-msan]: https://clang.llvm.org/docs/MemorySanitizer.html
 [clang-scs]: https://clang.llvm.org/docs/ShadowCallStack.html
 [clang-tsan]: https://clang.llvm.org/docs/ThreadSanitizer.html
+[linux-kasan]: https://www.kernel.org/doc/html/latest/dev-tools/kasan.html
 [llvm-memtag]: https://llvm.org/docs/MemTagSanitizer.html
diff --git a/src/tools/compiletest/src/header.rs b/src/tools/compiletest/src/header.rs
index e11ebca6ea9..d9b39927ca4 100644
--- a/src/tools/compiletest/src/header.rs
+++ b/src/tools/compiletest/src/header.rs
@@ -935,6 +935,7 @@ pub fn make_test_description<R: Read>(
     let has_asan = util::ASAN_SUPPORTED_TARGETS.contains(&&*config.target);
     let has_cfi = util::CFI_SUPPORTED_TARGETS.contains(&&*config.target);
     let has_kcfi = util::KCFI_SUPPORTED_TARGETS.contains(&&*config.target);
+    let has_kasan = util::KASAN_SUPPORTED_TARGETS.contains(&&*config.target);
     let has_lsan = util::LSAN_SUPPORTED_TARGETS.contains(&&*config.target);
     let has_msan = util::MSAN_SUPPORTED_TARGETS.contains(&&*config.target);
     let has_tsan = util::TSAN_SUPPORTED_TARGETS.contains(&&*config.target);
@@ -1010,6 +1011,7 @@ pub fn make_test_description<R: Read>(
         reason!(!has_asan && config.parse_name_directive(ln, "needs-sanitizer-address"));
         reason!(!has_cfi && config.parse_name_directive(ln, "needs-sanitizer-cfi"));
         reason!(!has_kcfi && config.parse_name_directive(ln, "needs-sanitizer-kcfi"));
+        reason!(!has_kasan && config.parse_name_directive(ln, "needs-sanitizer-kasan"));
         reason!(!has_lsan && config.parse_name_directive(ln, "needs-sanitizer-leak"));
         reason!(!has_msan && config.parse_name_directive(ln, "needs-sanitizer-memory"));
         reason!(!has_tsan && config.parse_name_directive(ln, "needs-sanitizer-thread"));
diff --git a/src/tools/compiletest/src/util.rs b/src/tools/compiletest/src/util.rs
index 67f49bb6397..0ce9f2622e1 100644
--- a/src/tools/compiletest/src/util.rs
+++ b/src/tools/compiletest/src/util.rs
@@ -45,6 +45,13 @@ pub const CFI_SUPPORTED_TARGETS: &[&str] = &[
 
 pub const KCFI_SUPPORTED_TARGETS: &[&str] = &["aarch64-linux-none", "x86_64-linux-none"];
 
+pub const KASAN_SUPPORTED_TARGETS: &[&str] = &[
+    "aarch64-unknown-none",
+    "riscv64gc-unknown-none-elf",
+    "riscv64imac-unknown-none-elf",
+    "x86_64-unknown-none",
+];
+
 pub const LSAN_SUPPORTED_TARGETS: &[&str] = &[
     // FIXME: currently broken, see #88132
     // "aarch64-apple-darwin",
diff --git a/tests/codegen/sanitizer-kasan-emits-instrumentation.rs b/tests/codegen/sanitizer-kasan-emits-instrumentation.rs
new file mode 100644
index 00000000000..d6e3f2719df
--- /dev/null
+++ b/tests/codegen/sanitizer-kasan-emits-instrumentation.rs
@@ -0,0 +1,47 @@
+// Verifies that `-Zsanitizer=kernel-address` emits sanitizer instrumentation.
+
+// compile-flags: -Zsanitizer=kernel-address
+// revisions: aarch64 riscv64imac riscv64gc x86_64
+//[aarch64] compile-flags: --target aarch64-unknown-none
+//[aarch64] needs-llvm-components: aarch64
+//[riscv64imac] compile-flags: --target riscv64imac-unknown-none-elf
+//[riscv64imac] needs-llvm-components: riscv
+//[riscv64imac] min-llvm-version: 16
+//[riscv64gc] compile-flags: --target riscv64gc-unknown-none-elf
+//[riscv64gc] needs-llvm-components: riscv
+//[riscv64gc] min-llvm-version: 16
+//[x86_64] compile-flags: --target x86_64-unknown-none
+//[x86_64] needs-llvm-components: x86
+
+#![crate_type = "rlib"]
+#![feature(no_core, no_sanitize, lang_items)]
+#![no_core]
+
+#[lang = "sized"]
+trait Sized {}
+
+#[lang = "copy"]
+trait Copy {}
+
+impl Copy for u8 {}
+
+// CHECK-LABEL: ; sanitizer_kasan_emits_instrumentation::unsanitized
+// CHECK-NEXT:  ; Function Attrs:
+// CHECK-NOT:   sanitize_address
+// CHECK:       start:
+// CHECK-NOT:   call void @__asan_report_load
+// CHECK:       }
+#[no_sanitize(address)]
+pub fn unsanitized(b: &mut u8) -> u8 {
+    *b
+}
+
+// CHECK-LABEL: ; sanitizer_kasan_emits_instrumentation::sanitized
+// CHECK-NEXT:  ; Function Attrs:
+// CHECK:       sanitize_address
+// CHECK:       start:
+// CHECK:       call void @__asan_report_load
+// CHECK:       }
+pub fn sanitized(b: &mut u8) -> u8 {
+    *b
+}
diff --git a/tests/ui/sanitize/cfg-kasan.rs b/tests/ui/sanitize/cfg-kasan.rs
new file mode 100644
index 00000000000..d721011f3ee
--- /dev/null
+++ b/tests/ui/sanitize/cfg-kasan.rs
@@ -0,0 +1,28 @@
+// Verifies that when compiling with -Zsanitizer=kernel-address,
+// the `#[cfg(sanitize = "address")]` attribute is configured.
+
+// check-pass
+// compile-flags: -Zsanitizer=kernel-address --cfg kernel_address
+// revisions: aarch64 riscv64imac riscv64gc x86_64
+//[aarch64] compile-flags: --target aarch64-unknown-none
+//[aarch64] needs-llvm-components: aarch64
+//[riscv64imac] compile-flags: --target riscv64imac-unknown-none-elf
+//[riscv64imac] needs-llvm-components: riscv
+//[riscv64imac] min-llvm-version: 16
+//[riscv64gc] compile-flags: --target riscv64gc-unknown-none-elf
+//[riscv64gc] needs-llvm-components: riscv
+//[riscv64gc] min-llvm-version: 16
+//[x86_64] compile-flags: --target x86_64-unknown-none
+//[x86_64] needs-llvm-components: x86
+
+#![crate_type = "rlib"]
+#![feature(cfg_sanitize, no_core, lang_items)]
+#![no_core]
+
+#[lang = "sized"]
+trait Sized {}
+
+const _: fn() -> () = main;
+
+#[cfg(all(sanitize = "address", kernel_address))]
+fn main() {}