[llvm] [hwasan] Add intrinsics for fixed shadow on Aarch64 (PR #89319)

[Vitaly Buka via llvm-commits]

================
@@ -625,14 +638,35 @@ void AArch64AsmPrinter::emitHwasanMemaccessSymbols(Module &M) {
                                      .addImm(4)
                                      .addImm(55),
                                  *STI);
-    OutStreamer->emitInstruction(
-        MCInstBuilder(AArch64::LDRBBroX)
-            .addReg(AArch64::W16)
-            .addReg(IsShort ? AArch64::X20 : AArch64::X9)
-            .addReg(AArch64::X16)
-            .addImm(0)
-            .addImm(0),
-        *STI);
+
+    if (IsFixedShadow) {
+      // Aarch64 makes it difficult to embed large constants in the code.
+      // Fortuitously, kShadowBaseAlignment == 32, so we use the 32-bit
+      // left-shift option in the MOV instruction. Combined with the 16-bit
+      // immediate, this is enough to represent any offset up to 2**48.
+      OutStreamer->emitInstruction(MCInstBuilder(AArch64::MOVZXi)
+                                       .addReg(AArch64::X17)
+                                       .addImm(FixedShadowOffset >> 32)
----------------
vitalybuka wrote:

or add `if (!aligned)` and insert longer less efficient but correct sequence

https://github.com/llvm/llvm-project/pull/89319