[SPARC64] Preserve rd of TLS *_ADD relocations during optimization

GCC might decide to put the result of the operation in another register,
so remove hardcodings of %o0 as the destination register (`rd` field).

Author: koachan

src/arch-sparc64.cc

@@ -390,7 +390,8 @@ void InputSection<E>::apply_reloc_alloc(Context<E> &ctx, u8 *base) {
         }
       } else {
         u32 rs2 = bits(*(ub32 *)loc, 4, 0);
-        *(ub32 *)loc = 0x9001'c000 | rs2; // add %g7, %reg, %o0
+        u32 rd = bits(*(ub32 *)loc, 29, 25);
+        *(ub32 *)loc = 0x9001'c000 | (rd << 25) | rs2; // add %g7, %reg, %rd
       }
       break;
     case R_SPARC_TLS_GD_CALL:
@@ -420,7 +421,8 @@ void InputSection<E>::apply_reloc_alloc(Context<E> &ctx, u8 *base) {
         // do nothing
       } else {
         u32 rs2 = bits(*(ub32 *)loc, 4, 0);
-        *(ub32 *)loc = 0x9021'c000 | rs2; // sub %g7, %reg, %o0
+        u32 rd = bits(*(ub32 *)loc, 29, 25);
+        *(ub32 *)loc = 0x9021'c000 | (rd << 25) | rs2; // sub %g7, %reg, %rd
       }
       break;
     case R_SPARC_TLS_LDM_CALL:
@@ -429,6 +431,13 @@ void InputSection<E>::apply_reloc_alloc(Context<E> &ctx, u8 *base) {
         *(ub32 *)loc |= bits(addr + A - P, 31, 2);
       } else {
         *(ub32 *)loc = 0x0100'0000; // nop
+
+        // If the argument to __tls_get_addr is set in the delay slot of the call,
+        // turn the delay slot instruction into a nop too.
+        u32 ds_rd = bits(*(ub32 *)(loc + 4), 29, 25);
+        if (ds_rd == 8) { // rd is %o0
+          *(ub32 *)(loc + 4) = 0x0100'0000; // nop
+        }
       }
       break;
     case R_SPARC_TLS_LDO_HIX22: