[lld] Add target support for SystemZ (s390x)

This patch adds full support for linking SystemZ (ELF s390x) object
files.  Support should be generally complete:
- All relocation types are supported.
- Full shared library support (DYNAMIC, GOT, PLT, ifunc).
- Relaxation of TLS and GOT relocations where appropriate.
- Platform-specific test cases.

In addition to new platform code and the obvious changes, there were
a few additional changes to common code:

- Add three new RelExpr members (R_GOTPLT_OFF, R_GOTPLT_PC, and
  R_PLT_GOTREL) needed to support certain s390x relocations.
  I chose not to use a platform-specific name since nothing in
  the definition of these relocs is actually platform-specific;
  it is well possible that other platforms will need the same.

- A couple of tweaks to TLS relocation handling, as the particular
  semantics of the s390x versions differ slightly.  See comments
  in the code.

This was tested by building and testing >1500 Fedora packages,
with only a handful of failures; as these also have issues when
building with LLD on other architectures, they seem unrelated.

Co-authored-by: Tulio Magno Quites Machado Filho <[email protected]>

Author: uweigand

lld/ELF/Arch/SystemZ.cpp

@@ -33,6 +33,7 @@ add_lld_library(lldELF
   Arch/PPC64.cpp
   Arch/RISCV.cpp
   Arch/SPARCV9.cpp
+  Arch/SystemZ.cpp
   Arch/X86.cpp
   Arch/X86_64.cpp
   ARMErrataFix.cpp

lld/ELF/CMakeLists.txt

@@ -200,6 +200,7 @@ static std::tuple<ELFKind, uint16_t, uint8_t> parseEmulation(StringRef emul) {
           .Case("msp430elf", {ELF32LEKind, EM_MSP430})
           .Case("elf64_amdgpu", {ELF64LEKind, EM_AMDGPU})
           .Case("elf64loongarch", {ELF64LEKind, EM_LOONGARCH})
+          .Case("elf64_s390", {ELF64BEKind, EM_S390})
           .Default({ELFNoneKind, EM_NONE});
 
   if (ret.first == ELFNoneKind)
@@ -1136,7 +1137,7 @@ static SmallVector<StringRef, 0> getSymbolOrderingFile(MemoryBufferRef mb) {
 static bool getIsRela(opt::InputArgList &args) {
   // The psABI specifies the default relocation entry format.
   bool rela = is_contained({EM_AARCH64, EM_AMDGPU, EM_HEXAGON, EM_LOONGARCH,
-                            EM_PPC, EM_PPC64, EM_RISCV, EM_X86_64},
+                            EM_PPC, EM_PPC64, EM_RISCV, EM_S390, EM_X86_64},
                            config->emachine);
   // If -z rel or -z rela is specified, use the last option.
   for (auto *arg : args.filtered(OPT_z)) {

lld/ELF/Driver.cpp

@@ -1614,6 +1614,8 @@ static uint16_t getBitcodeMachineKind(StringRef path, const Triple &t) {
     return EM_RISCV;
   case Triple::sparcv9:
     return EM_SPARCV9;
+  case Triple::systemz:
+    return EM_S390;
   case Triple::x86:
     return t.isOSIAMCU() ? EM_IAMCU : EM_386;
   case Triple::x86_64:

lld/ELF/InputFiles.cpp

@@ -654,6 +654,7 @@ static int64_t getTlsTpOffset(const Symbol &s) {
 
     // Variant 2.
   case EM_HEXAGON:
+  case EM_S390:
   case EM_SPARCV9:
   case EM_386:
   case EM_X86_64:
@@ -708,6 +709,8 @@ uint64_t InputSectionBase::getRelocTargetVA(const InputFile *file, RelType type,
   case R_GOT_OFF:
   case R_RELAX_TLS_GD_TO_IE_GOT_OFF:
     return sym.getGotOffset() + a;
+  case R_GOTPLT_OFF:
+    return sym.getGotPltOffset() + a;
   case R_AARCH64_GOT_PAGE_PC:
   case R_AARCH64_RELAX_TLS_GD_TO_IE_PAGE_PC:
     return getAArch64Page(sym.getGotVA() + a) - getAArch64Page(p);
@@ -716,6 +719,8 @@ uint64_t InputSectionBase::getRelocTargetVA(const InputFile *file, RelType type,
   case R_GOT_PC:
   case R_RELAX_TLS_GD_TO_IE:
     return sym.getGotVA() + a - p;
+  case R_GOTPLT_PC:
+    return sym.getGotPltVA() + a - p;
   case R_LOONGARCH_GOT_PAGE_PC:
     if (sym.hasFlag(NEEDS_TLSGD))
       return getLoongArchPageDelta(in.got->getGlobalDynAddr(sym) + a, p, type);
@@ -807,6 +812,8 @@ uint64_t InputSectionBase::getRelocTargetVA(const InputFile *file, RelType type,
     return getLoongArchPageDelta(sym.getPltVA() + a, p, type);
   case R_PLT_GOTPLT:
     return sym.getPltVA() + a - in.gotPlt->getVA();
+  case R_PLT_GOTREL:
+    return sym.getPltVA() + a - in.got->getVA();
   case R_PPC32_PLTREL:
     // R_PPC_PLTREL24 uses the addend (usually 0 or 0x8000) to indicate r30
     // stores _GLOBAL_OFFSET_TABLE_ or .got2+0x8000. The addend is ignored for

lld/ELF/InputSection.cpp

@@ -1364,8 +1364,8 @@ static unsigned handleTlsRelocation(RelType type, Symbol &sym,
             R_LOONGARCH_GOT_PAGE_PC, R_GOT_OFF, R_TLSIE_HINT>(expr)) {
     ctx.hasTlsIe.store(true, std::memory_order_relaxed);
     // Initial-Exec relocs can be relaxed to Local-Exec if the symbol is locally
-    // defined.
-    if (toExecRelax && isLocalInExecutable) {
+    // defined.  This is not supported on SystemZ.
+    if (toExecRelax && isLocalInExecutable && config->emachine != EM_S390) {
       c.addReloc({R_RELAX_TLS_IE_TO_LE, type, offset, addend, &sym});
     } else if (expr != R_TLSIE_HINT) {
       sym.setFlags(NEEDS_TLSIE);
@@ -1411,6 +1411,26 @@ template <class ELFT, class RelTy> void RelocationScanner::scanOne(RelTy *&i) {
   if (expr == R_NONE)
     return;
 
+  // Like other platforms, calls to the TLS helper routine on SystemZ carry
+  // two relocations, one for the helper routine itself, and a TLS marker
+  // relocation.  When relaxing the TLS model, the helper routine is no longer
+  // needed, and its relocation should be removed.  Unlike other platforms,
+  // on SystemZ the TLS marker routine typically comes *after* the helper
+  // routine relocation, so the getTlsGdRelaxSkip mechanism used by
+  // handleTlsRelocation does not work on this platform.
+  //
+  // Instead, check for this case here: if we are building a main executable
+  // (i.e. TLS relaxation applies), and the relocation *after* the current one
+  // is a TLS call marker instruction matching the current instruction, then
+  // skip this relocation.
+  if (config->emachine == EM_S390 && !config->shared) {
+    if (i < end && getter.get(i->r_offset) == offset - 2) {
+      RelType nextType = i->getType(/*isMips64EL=*/false);
+      if (nextType == R_390_TLS_GDCALL || nextType == R_390_TLS_LDCALL)
+        return;
+    }
+  }
+
   // Error if the target symbol is undefined. Symbol index 0 may be used by
   // marker relocations, e.g. R_*_NONE and R_ARM_V4BX. Don't error on them.
   if (sym.isUndefined() && symIndex != 0 &&

lld/ELF/Relocations.cpp

@@ -40,11 +40,14 @@ enum RelExpr {
   R_GOTPLT,
   R_GOTPLTREL,
   R_GOTREL,
+  R_GOTPLT_OFF,
+  R_GOTPLT_PC,
   R_NONE,
   R_PC,
   R_PLT,
   R_PLT_PC,
   R_PLT_GOTPLT,
+  R_PLT_GOTREL,
   R_RELAX_HINT,
   R_RELAX_GOT_PC,
   R_RELAX_GOT_PC_NOPIC,

lld/ELF/Relocations.h

@@ -445,6 +445,7 @@ static std::pair<ELFKind, uint16_t> parseBfdName(StringRef s) {
       .Case("elf32-msp430", {ELF32LEKind, EM_MSP430})
       .Case("elf32-loongarch", {ELF32LEKind, EM_LOONGARCH})
       .Case("elf64-loongarch", {ELF64LEKind, EM_LOONGARCH})
+      .Case("elf64-s390", {ELF64BEKind, EM_S390})
       .Default({ELFNoneKind, EM_NONE});
 }
 

lld/ELF/ScriptParser.cpp

@@ -1419,6 +1419,9 @@ DynamicSection<ELFT>::computeContents() {
     case EM_MIPS:
       addInSec(DT_MIPS_PLTGOT, *in.gotPlt);
       break;
+    case EM_S390:
+      addInSec(DT_PLTGOT, *in.got);
+      break;
     case EM_SPARCV9:
       addInSec(DT_PLTGOT, *in.plt);
       break;

lld/ELF/SyntheticSections.cpp

@@ -87,6 +87,8 @@ TargetInfo *elf::getTarget() {
     return getRISCVTargetInfo();
   case EM_SPARCV9:
     return getSPARCV9TargetInfo();
+  case EM_S390:
+    return getSystemZTargetInfo();
   case EM_X86_64:
     return getX86_64TargetInfo();
   }

lld/ELF/Target.cpp

@@ -186,6 +186,7 @@ TargetInfo *getPPC64TargetInfo();
 TargetInfo *getPPCTargetInfo();
 TargetInfo *getRISCVTargetInfo();
 TargetInfo *getSPARCV9TargetInfo();
+TargetInfo *getSystemZTargetInfo();
 TargetInfo *getX86TargetInfo();
 TargetInfo *getX86_64TargetInfo();
 template <class ELFT> TargetInfo *getMipsTargetInfo();

lld/ELF/Target.h

@@ -0,0 +1,5 @@
+// glibc < 2.39 used to align .init and .fini code at a 4-byte boundary.
+// This file aims to recreate that behavior.
+	.section        .init,"ax",@progbits
+	.align	4
+	lg %r4, 272(%r15)

lld/test/ELF/Inputs/systemz-init.s

@@ -0,0 +1,63 @@
+# REQUIRES: systemz
+# RUN: llvm-mc -filetype=obj -triple=s390x-unknown-linux %s -o %t.o
+# RUN: ld.lld --hash-style=sysv -discard-all -shared %t.o -o %t.so
+# RUN: llvm-readelf --file-header --program-headers --section-headers --dynamic-table %t.so | FileCheck %s
+
+# Exits with return code 55 on linux.
+.text
+  lghi 2,55
+  svc 1
+
+// CHECK:       ELF Header:
+// CHECK-NEXT:  Magic:   7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00
+// CHECK-NEXT:  Class:                             ELF64
+// CHECK-NEXT:  Data:                              2's complement, big endian
+// CHECK-NEXT:  Version:                           1 (current)
+// CHECK-NEXT:  OS/ABI:                            UNIX - System V
+// CHECK-NEXT:  ABI Version:                       0
+// CHECK-NEXT:  Type:                              DYN (Shared object file)
+// CHECK-NEXT:  Machine:                           IBM S/390
+// CHECK-NEXT:  Version:                           0x1
+// CHECK-NEXT:  Entry point address:               0x0
+// CHECK-NEXT:  Start of program headers:          64 (bytes into file)
+// CHECK-NEXT:  Start of section headers:          768 (bytes into file)
+// CHECK-NEXT:  Flags:                             0x0
+// CHECK-NEXT:  Size of this header:               64 (bytes)
+// CHECK-NEXT:  Size of program headers:           56 (bytes)
+// CHECK-NEXT:  Number of program headers:         7
+// CHECK-NEXT:  Size of section headers:           64 (bytes)
+// CHECK-NEXT:  Number of section headers:         11
+// CHECK-NEXT:  Section header string table index: 9
+
+// CHECK:       Section Headers:
+// CHECK-NEXT:  [Nr] Name              Type            Address          Off    Size   ES Flg Lk Inf Al
+// CHECK-NEXT:  [ 0]                   NULL            0000000000000000 000000 000000 00      0   0  0
+// CHECK-NEXT:  [ 1] .dynsym           DYNSYM          00000000000001c8 0001c8 000018 18   A  3   1  8
+// CHECK-NEXT:  [ 2] .hash             HASH            00000000000001e0 0001e0 000010 04   A  1   0  4
+// CHECK-NEXT:  [ 3] .dynstr           STRTAB          00000000000001f0 0001f0 000001 00   A  0   0  1
+// CHECK-NEXT:  [ 4] .text             PROGBITS        00000000000011f4 0001f4 000006 00  AX  0   0  4
+// CHECK-NEXT:  [ 5] .dynamic          DYNAMIC         0000000000002200 000200 000060 10  WA  3   0  8
+// CHECK-NEXT:  [ 6] .relro_padding    NOBITS          0000000000002260 000260 000da0 00  WA  0   0  1
+// CHECK-NEXT:  [ 7] .comment          PROGBITS        0000000000000000 000260 000008 01  MS  0   0  1
+// CHECK-NEXT:  [ 8] .symtab           SYMTAB          0000000000000000 000268 000030 18     10   2  8
+// CHECK-NEXT:  [ 9] .shstrtab         STRTAB          0000000000000000 000298 000058 00      0   0  1
+// CHECK-NEXT:  [10] .strtab           STRTAB          0000000000000000 0002f0 00000a 00      0   0  1
+
+// CHECK:       Program Headers:
+// CHECK-NEXT:  Type           Offset   VirtAddr           PhysAddr           FileSiz  MemSiz   Flg Align
+// CHECK-NEXT:  PHDR           0x000040 0x0000000000000040 0x0000000000000040 0x000188 0x000188 R   0x8
+// CHECK-NEXT:  LOAD           0x000000 0x0000000000000000 0x0000000000000000 0x0001f1 0x0001f1 R   0x1000
+// CHECK-NEXT:  LOAD           0x0001f4 0x00000000000011f4 0x00000000000011f4 0x000006 0x000006 R E 0x1000
+// CHECK-NEXT:  LOAD           0x000200 0x0000000000002200 0x0000000000002200 0x000060 0x000e00 RW  0x1000
+// CHECK-NEXT:  DYNAMIC        0x000200 0x0000000000002200 0x0000000000002200 0x000060 0x000060 RW  0x8
+// CHECK-NEXT:  GNU_RELRO      0x000200 0x0000000000002200 0x0000000000002200 0x000060 0x000e00 R   0x1
+// CHECK-NEXT:  GNU_STACK      0x000000 0x0000000000000000 0x0000000000000000 0x000000 0x000000 RW  0x0
+
+// CHECK:       Dynamic section at offset 0x200 contains 6 entries:
+// CHECK-NEXT:  Tag                Type     Name/Value
+// CHECK-NEXT:  0x0000000000000006 (SYMTAB) 0x1c8
+// CHECK-NEXT:  0x000000000000000b (SYMENT) 24 (bytes)
+// CHECK-NEXT:  0x0000000000000005 (STRTAB) 0x1f0
+// CHECK-NEXT:  0x000000000000000a (STRSZ)  1 (bytes)
+// CHECK-NEXT:  0x0000000000000004 (HASH)   0x1e0
+// CHECK-NEXT:  0x0000000000000000 (NULL)   0x0

lld/test/ELF/basic-systemz.s

@@ -0,0 +1,29 @@
+# REQUIRES: systemz
+# RUN: llvm-mc -filetype=obj -triple=s390x-unknown-linux %s -o %t
+# RUN: ld.lld -m elf64_s390 %t -o %t2
+# RUN: llvm-readelf --file-header %t2 | FileCheck %s
+# RUN: ld.lld %t -o %t3
+# RUN: llvm-readelf --file-header %t3 | FileCheck %s
+# RUN: echo 'OUTPUT_FORMAT(elf64-s390)' > %t.script
+# RUN: ld.lld %t.script  %t -o %t4
+# RUN: llvm-readelf --file-header %t4 | FileCheck %s
+
+// CHECK:       ELF Header:
+// CHECK-NEXT:  Magic:   7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00
+// CHECK-NEXT:  Class:                             ELF64
+// CHECK-NEXT:  Data:                              2's complement, big endian
+// CHECK-NEXT:  Version:                           1 (current)
+// CHECK-NEXT:  OS/ABI:                            UNIX - System V
+// CHECK-NEXT:  ABI Version:                       0
+// CHECK-NEXT:  Type:                              EXEC (Executable file)
+// CHECK-NEXT:  Machine:                           IBM S/390
+// CHECK-NEXT:  Version:                           0x1
+// CHECK-NEXT:  Entry point address:
+// CHECK-NEXT:  Start of program headers:          64 (bytes into file)
+// CHECK-NEXT:  Start of section headers:
+// CHECK-NEXT:  Flags:                             0x0
+// CHECK-NEXT:  Size of this header:               64 (bytes)
+// CHECK-NEXT:  Size of program headers:           56 (bytes)
+
+.globl _start
+_start:

lld/test/ELF/emulation-systemz.s

@@ -0,0 +1,18 @@
+; REQUIRES: systemz
+;; Test we can infer the e_machine value EM_S390 from a bitcode file.
+
+; RUN: llvm-as %s -o %t.o
+; RUN: ld.lld %t.o -o %t
+; RUN: llvm-readobj -h %t | FileCheck %s
+
+; CHECK: Class: 64-bit
+; CHECK: DataEncoding: BigEndian
+; CHECK: Machine: EM_S390
+
+target datalayout = "E-m:e-i1:8:16-i8:8:16-i64:64-f128:64-v128:64-a:8:16-n32:64"
+target triple = "s390x-unknown-linux-gnu"
+
+define void @_start() {
+entry:
+  ret void
+}

lld/test/ELF/lto/systemz.ll

@@ -0,0 +1,72 @@
+// REQUIRES: systemz
+// RUN: llvm-mc -filetype=obj -triple=s390x-none-linux-gnu %s -o %t.o
+// RUN: ld.lld -static %t.o -o %tout
+// RUN: llvm-objdump --no-print-imm-hex -d --no-show-raw-insn %tout | FileCheck %s --check-prefix=DISASM
+// RUN: llvm-readelf --section-headers --relocations --symbols %tout | FileCheck %s
+
+// CHECK:      There are 9 section headers
+// CHECK:      Section Headers:
+// CHECK-NEXT:  [Nr] Name              Type            Address          Off    Size   ES Flg Lk Inf Al
+// CHECK-NEXT:  [ 0]                   NULL            0000000000000000 000000 000000 00      0   0  0
+// CHECK-NEXT:  [ 1] .rela.dyn         RELA            0000000001000158 000158 000030 18  AI  0   4  8
+// CHECK-NEXT:  [ 2] .text             PROGBITS        0000000001001188 000188 00001c 00  AX  0   0  4
+// CHECK-NEXT:  [ 3] .iplt             PROGBITS        00000000010011b0 0001b0 000040 00  AX  0   0 16
+// CHECK-NEXT:  [ 4] .got.plt          PROGBITS        00000000010021f0 0001f0 000010 00  WA  0   0  8
+// CHECK-NEXT:  [ 5] .comment          PROGBITS        0000000000000000 000200 000008 01  MS  0   0  1
+// CHECK-NEXT:  [ 6] .symtab           SYMTAB          0000000000000000 000208 000090 18      8   3  8
+// CHECK-NEXT:  [ 7] .shstrtab         STRTAB          0000000000000000 000298 000043 00      0   0  1
+// CHECK-NEXT:  [ 8] .strtab           STRTAB          0000000000000000 0002db 000032 00      0   0  1
+
+// CHECK:      Relocation section '.rela.dyn' at offset 0x158 contains 2 entries:
+// CHECK-NEXT:     Offset             Info             Type               Symbol's Value  Symbol's Name + Addend
+// CHECK-NEXT: 00000000010021f0  000000000000003d R_390_IRELATIVE                   1001188
+// CHECK-NEXT: 00000000010021f8  000000000000003d R_390_IRELATIVE                   100118a
+
+// CHECK:      Symbol table '.symtab' contains 6 entries:
+// CHECK-NEXT:   Num:    Value          Size Type    Bind   Vis       Ndx Name
+// CHECK-NEXT:     0: 0000000000000000     0 NOTYPE  LOCAL  DEFAULT   UND
+// CHECK-NEXT:     1: 0000000001000158     0 NOTYPE  LOCAL  HIDDEN      1 __rela_iplt_start
+// CHECK-NEXT:     2: 0000000001000188     0 NOTYPE  LOCAL  HIDDEN      1 __rela_iplt_end
+// CHECK-NEXT:     3: 0000000001001188     0 IFUNC   GLOBAL DEFAULT     2 foo
+// CHECK-NEXT:     4: 000000000100118a     0 IFUNC   GLOBAL DEFAULT     2 bar
+// CHECK-NEXT:     5: 000000000100118c     0 NOTYPE  GLOBAL DEFAULT     2 _start
+
+// DISASM: Disassembly of section .text:
+// DISASM-EMPTY:
+// DISASM-NEXT: <foo>:
+// DISASM-NEXT:  1001188: br      %r14
+// DISASM: <bar>:
+// DISASM-NEXT:  100118a: br      %r14
+// DISASM:      <_start>:
+// DISASM-NEXT:  100118c: brasl   %r14, 0x10011b0
+// DISASM-NEXT:  1001192: brasl   %r14, 0x10011d0
+// DISASM-NEXT:  1001198: larl    %r2, 0x1000158
+// DISASM-NEXT:  100119e: larl    %r2, 0x1000188
+// DISASM-EMPTY:
+// DISASM-NEXT: Disassembly of section .iplt:
+// DISASM-EMPTY:
+// DISASM-NEXT: <.iplt>:
+// DISASM:        10011b0:       larl    %r1, 0x10021f0
+// DISASM-NEXT:   10011b6:       lg      %r1, 0(%r1)
+// DISASM-NEXT:   10011bc:       br      %r1
+// DISASM:        10011d0:       larl    %r1, 0x10021f8
+// DISASM-NEXT:   10011d6:       lg      %r1, 0(%r1)
+// DISASM-NEXT:   10011dc:       br      %r1
+
+.text
+.type foo STT_GNU_IFUNC
+.globl foo
+foo:
+ br %r14
+
+.type bar STT_GNU_IFUNC
+.globl bar
+bar:
+ br %r14
+
+.globl _start
+_start:
+ brasl %r14, foo@plt
+ brasl %r14, bar@plt
+ larl %r2, __rela_iplt_start
+ larl %r2, __rela_iplt_end

lld/test/ELF/systemz-gnu-ifunc.s

@@ -0,0 +1,16 @@
+// REQUIRES: systemz
+// RUN: llvm-mc -filetype=obj -triple=s390x-unknown-linux %s -o %t.o
+// RUN: llvm-mc -filetype=obj -triple=s390x-unknown-linux %p/Inputs/shared.s -o %t2.o
+// RUN: ld.lld -shared %t2.o -soname=%t2.so -o %t2.so
+
+// RUN: ld.lld -dynamic-linker /lib/ld64.so.1 %t.o %t2.so -o %t
+// RUN: llvm-readelf -S -r  %t | FileCheck %s
+
+// CHECK: .got              PROGBITS        {{.*}} {{.*}} 000020 00  WA  0   0  8
+
+// CHECK: Relocation section '.rela.dyn' at offset {{.*}} contains 1 entries:
+// CHECK: {{.*}}  000000010000000a R_390_GLOB_DAT         0000000000000000 bar + 0
+
+.global _start
+_start:
+	lgrl  %r1,bar@GOT