[llvm-branch-commits] [lld] [PAC][lld][AArch64][ELF] Support signed TLSDESC (PR #113817)

Daniil Kovalev via llvm-branch-commits llvm-branch-commits at lists.llvm.org
Fri Nov 1 03:30:51 PDT 2024


================
@@ -1355,6 +1355,36 @@ unsigned RelocationScanner::handleTlsRelocation(RelExpr expr, RelType type,
     return 1;
   }
 
+  auto fatalBothAuthAndNonAuth = [&sym]() {
+    fatal("both AUTH and non-AUTH TLSDESC entries for '" + sym.getName() +
+          "' requested, but only one type of TLSDESC entry per symbol is "
+          "supported");
+  };
+
+  // Do not optimize signed TLSDESC as described in pauthabielf64 to LE/IE.
+  // https://github.com/ARM-software/abi-aa/blob/main/pauthabielf64/pauthabielf64.rst#general-restrictions
+  // > PAUTHELF64 only supports the descriptor based TLS (TLSDESC).
+  if (oneof<R_AARCH64_AUTH_TLSDESC_PAGE, RelExpr::R_AARCH64_AUTH_TLSDESC>(
+          expr)) {
+    assert(ctx.arg.emachine == EM_AARCH64);
+    if (!sym.hasFlag(NEEDS_TLSDESC))
+      sym.setFlags(NEEDS_TLSDESC | NEEDS_TLSDESC_AUTH);
+    else if (!sym.hasFlag(NEEDS_TLSDESC_AUTH))
+      fatalBothAuthAndNonAuth();
+    sec->addReloc({expr, type, offset, addend, &sym});
+    return 1;
+  }
+
+  if (sym.hasFlag(NEEDS_TLSDESC_AUTH)) {
+    assert(ctx.arg.emachine == EM_AARCH64);
+    // TLSDESC_CALL hint relocation probably should not be emitted by compiler
+    // with signed TLSDESC enabled since it does not give any value, but leave a
+    // check against that just in case someone uses it.
+    if (expr != R_TLSDESC_CALL)
+      fatalBothAuthAndNonAuth();
----------------
kovdan01 wrote:

The logic of this code and code above is the following. We check rel expr against AUTH variants.

1. If yes, the symbol should either already have both `NEEDS_TLSDESC` and `NEEDS_TLSDESC_AUTH` or have none of them. The symbol having only `NEEDS_TLSDESC` means that a non-auth entry was requested previously, and now we are requesting an auth one - which is currently not supported.

2. If no, but `NEEDS_TLSDESC_AUTH` flag was already set previously, it means that auth entry was requested previously, and non-auth one is requested now, which is currently not supported. The only case when we don't emit an error is expr equal to `R_TLSDESC_CALL` - this is a hint reloc which does not result in a non-auth entry by itself. See comment right above this if statement.

https://github.com/llvm/llvm-project/pull/113817


More information about the llvm-branch-commits mailing list