[PPC64] Add .toc section after .got section

[syzaara]

Summary

PPC64 maintains a compiler managed got in the .toc section. When accessing a global variable through got-indirect access, a .toc entry is created for the variable. The relocation for the got-indirect access will refer to the .toc section rather than the symbol that is actually accessed. The .toc entry contains the address of the global variable. We evaluate the offset from r2 (which is the TOC base) to the address of the toc entry for the global variable. Currently, the .toc is not near the .got. This causes errors because the offset from r2 to the toc section is too large. The linker needs to add all the .toc input sections to the .got output section, merging the compiler managed got with the linker got. This ensures that the offsets from the TOC base to the toc entries are not too large.

This patch puts the .toc section right after the .got section.

[ruiu]

I wonder if you really need to merge .toc into .got. If I understand your description correctly, it doesn't matter whether they are merged or not as long as they are close enough. So I'm wondering if you can just manipulate the section output order so that .toc and .got are output next to each other.

[syzaara]

In D45833#1072806, @ruiu wrote:

I wonder if you really need to merge .toc into .got. If I understand your description correctly, it doesn't matter whether they are merged or not as long as they are close enough. So I'm wondering if you can just manipulate the section output order so that .toc and .got are output next to each other.

Changing the output order to add .toc right after .got will be functionally correct. However, we also want to be similar to gold and bfd. The other linkers don't display toc as its own output section, but rather add it to the got output section.

[syzaara]

@ruiu Hi Rui, can you please finish reviewing this?

[sfertile]

In D45833#1072806, @ruiu wrote:

I wonder if you really need to merge .toc into .got. If I understand your description correctly, it doesn't matter whether they are merged or not as long as they are close enough. So I'm wondering if you can just manipulate the section output order so that .toc and .got are output next to each other.

This would work, but I think we eventually want to merge them together just to match the behaviors of bfd and gold anyway. If that's the case we may as well just perform that merging now.

[syzaara]

Change the sorting so that .toc is emitted right after .got rather than adding it into the got.

[syzaara]

@ruiu Hi Rui, I've changed the output section order now. Can you take a look at this patch again?

[ruiu]

I'm sorry that I missed this patch. LGTM

I'm fine if you eventually have to merge .toc and .got. As a reviewer, I believe that my role is to ask if something isn't clear. I asked a question because I was genuinely wondering, and it didn't mean that I was requesting you to make a change.

lld/trunk/ELF/Writer.cpp

if (Type == SHT_INIT_ARRAY || Type == SHT_FINI_ARRAY ||
     return true;
 
   // .got contains pointers to external symbols. They are resolved by
   // the dynamic linker when a module is loaded into memory, and after
   // that they are not expected to change. So, it can be in RELRO.
   if (InX::Got && Sec == InX::Got->getParent())
     return true;
 
+  if (Sec->Name.equals(".toc"))
+    return true;
+
   // .got.plt contains pointers to external function symbols. They are
   // by default resolved lazily, so we usually cannot put it into RELRO.
   // However, if "-z now" is given, the lazy symbol resolution is
   // disabled, which enables us to put it into RELRO.
   if (Sec == InX::GotPlt->getParent())
     return Config->ZNow;
 
   // .dynamic section contains data for the dynamic linker, and
enum RankFlags {
   RF_EXEC = 1 << 13,
   RF_NON_TLS_BSS = 1 << 12,
   RF_NON_TLS_BSS_RO = 1 << 11,
   RF_NOT_TLS = 1 << 10,
   RF_ALLOC_FIRST = 1 << 9,
   RF_BSS = 1 << 8,
   RF_NOTE = 1 << 7,
   RF_PPC_NOT_TOCBSS = 1 << 6,
-  RF_PPC_TOCL = 1 << 4,
-  RF_PPC_TOC = 1 << 3,
+  RF_PPC_TOCL = 1 << 5,
+  RF_PPC_TOC = 1 << 4,
+  RF_PPC_GOT = 1 << 3,
   RF_PPC_BRANCH_LT = 1 << 2,
   RF_MIPS_GPREL = 1 << 1,
   RF_MIPS_NOT_GOT = 1 << 0
 };
 
 static unsigned getSectionRank(const OutputSection *Sec) {
   unsigned Rank = 0;
 
if (Name != ".tocbss")
       Rank |= RF_PPC_NOT_TOCBSS;
 
     if (Name == ".toc1")
       Rank |= RF_PPC_TOCL;
 
     if (Name == ".toc")
       Rank |= RF_PPC_TOC;
 
+    if (Name == ".got")
+      Rank |= RF_PPC_GOT;
+
     if (Name == ".branch_lt")
       Rank |= RF_PPC_BRANCH_LT;
   }
 
   if (Config->EMachine == EM_MIPS) {
     // All sections with SHF_MIPS_GPREL flag should be grouped together
     // because data in these sections is addressable with a gp relative address.
     if (Sec->Flags & SHF_MIPS_GPREL)

lld/trunk/test/ELF/ppc64-func-entry-points.s

 # point and we branch to start of foo_external_same.
 
 // CHECK: _start:
 // CHECK: 10010020:       {{.*}}     bl .+144
 // CHECK: 10010034:       {{.*}}     bl .+84
 // CHECK: foo_external_diff:
 // CHECK-NEXT: 10010080:       {{.*}}     addis 2, 12, 2
 // CHECK-NEXT: 10010084:       {{.*}}     addi 2, 2, 32640
-// CHECK-NEXT: 10010088:       {{.*}}     addis 5, 2, -1
+// CHECK-NEXT: 10010088:       {{.*}}     addis 5, 2, 0
 // CHECK: foo_external_same:
 // CHECK-NEXT: 100100b0:       {{.*}}     add 3, 4, 3

lld/trunk/test/ELF/ppc64-got-indirect.s

+# REQUIRES: ppc
+
+# RUN: llvm-mc -filetype=obj -triple=powerpc64le-unknown-linux %s -o %t.o
+# RUN: llvm-readobj -relocations %t.o | FileCheck -check-prefix=RELOCS-LE %s
+# RUN: ld.lld %t.o -o %t2
+# RUN: llvm-objdump -D %t2 | FileCheck %s --check-prefix=CHECK-LE
+# RUN: llvm-objdump -D %t2 | FileCheck %s
+
+# RUN: llvm-mc -filetype=obj -triple=powerpc64-unknown-linux %s -o %t.o
+# RUN: llvm-readobj -relocations %t.o | FileCheck -check-prefix=RELOCS-BE %s
+# RUN: ld.lld %t.o -o %t2
+# RUN: llvm-objdump -D %t2 | FileCheck %s --check-prefix=CHECK-BE
+# RUN: llvm-objdump -D %t2 | FileCheck %s
+
+# Make sure we calculate the offset correctly for a got-indirect access to a
+# global variable as described by the PPC64 ELF V2 abi.
+  .text
+  .abiversion 2
+  .globl  _start                    # -- Begin function _start
+  .p2align  4
+  .type  _start,@function
+_start:                                   # @_start
+.Lfunc_begin0:
+.Lfunc_gep0:
+  addis 2, 12, .TOC.-.Lfunc_gep0@ha
+  addi 2, 2, .TOC.-.Lfunc_gep0@l
+.Lfunc_lep0:
+  .localentry  _start, .Lfunc_lep0-.Lfunc_gep0
+# %bb.0:                                # %entry
+  addis 3, 2, .LC0@toc@ha
+  ld 3, .LC0@toc@l(3)
+  li 4, 0
+  stw 4, -12(1)
+  li 0,1
+        lwa 3, 0(3)
+  sc
+  .long  0
+  .quad  0
+.Lfunc_end0:
+  .size  _start, .Lfunc_end0-.Lfunc_begin0
+                                        # -- End function
+  .section  .toc,"aw",@progbits
+.LC0:
+  .tc glob[TC],glob
+  .type  glob,@object            # @glob
+  .data
+  .globl  glob
+  .p2align  2
+glob:
+  .long  55                      # 0x37
+  .size  glob, 4
+
+# Verify the relocations emitted for glob are through the .toc
+
+# RELOCS-LE: Relocations [
+# RELOCS-LE:  .rela.text {
+# RELOCS-LE:     0x0 R_PPC64_REL16_HA .TOC. 0x0
+# RELOCS-LE:     0x4 R_PPC64_REL16_LO .TOC. 0x4
+# RELOCS-LE:     0x8 R_PPC64_TOC16_HA .toc 0x0
+# RELOCS-LE:     0xC R_PPC64_TOC16_LO_DS .toc 0x0
+# RELOCS-LE:   }
+# RELOCS-LE:   .rela.toc {
+# RELOCS-LE:     0x0 R_PPC64_ADDR64 glob 0x0
+# RELOCS-LE:   }
+
+# RELOCS-BE: Relocations [
+# RELOCS-BE:  .rela.text {
+# RELOCS-BE:    0x2 R_PPC64_REL16_HA .TOC. 0x2
+# RELOCS-BE:    0x6 R_PPC64_REL16_LO .TOC. 0x6
+# RELOCS-BE:    0xA R_PPC64_TOC16_HA .toc 0x0
+# RELOCS-BE:    0xE R_PPC64_TOC16_LO_DS .toc 0x0
+# RELOCS-BE:  }
+# RELOCS-BE:  .rela.toc {
+# RELOCS-BE:    0x0 R_PPC64_ADDR64 glob 0x0
+# RELOCS-BE:  }
+# RELOCS-BE:]
+
+# Verify that the global variable access is done through the correct
+# toc entry:
+# r2 = .TOC. = 0x10038000.
+# r3 = r2 - 32760 = 0x10030008 -> .toc entry for glob.
+
+# CHECK: _start:
+# CHECK-NEXT: 10010000:  {{.*}}   addis 2, 12, 3
+# CHECK-NEXT: 10010004:  {{.*}}   addi 2, 2, -32768
+# CHECK-NEXT: 10010008:  {{.*}}   addis 3, 2, 0
+# CHECK-NEXT: 1001000c:  {{.*}}   ld 3, -32760(3)
+# CHECK: 1001001c:  {{.*}}   lwa 3, 0(3)
+
+# CHECK-LE: Disassembly of section .data:
+# CHECK-LE-NEXT: glob:
+# CHECK-LE-NEXT: 10020000:       37 00 00 00
+
+# CHECK-LE: Disassembly of section .got:
+# CHECK-LE-NEXT: .got:
+# CHECK-LE-NEXT: 10030000:       00 80 03 10
+# CHECK-LE-NEXT: 10030004:       00 00 00 00
+
+# Verify that .toc comes right after .got
+# CHECK-LE: Disassembly of section .toc:
+# CHECK-LE: 10030008:       00 00 02 10
+
+# CHECK-BE: Disassembly of section .data:
+# CHECK-BE-NEXT: glob:
+# CHECK-BE-NEXT: 10020000:       00 00 00 37
+
+# CHECK-BE: Disassembly of section .got:
+# CHECK-BE-NEXT: .got:
+# CHECK-BE-NEXT: 10030000:       00 00 00 00
+# CHECK-BE-NEXT: 10030004:       10 03 80 00
+
+# Verify that .toc comes right after .got
+# CHECK-BE: Disassembly of section .toc:
+# CHECK-BE: 10030008:       00 00 00 00
+# CHECK-BE: 1003000c:       10 02 00 00

lld/trunk/test/ELF/ppc64-relocs.s

 
 .section .R_PPC64_TOC16_HI,"ax",@progbits
 .globl .FR_PPC64_TOC16_HI
 .FR_PPC64_TOC16_HI:
   addis 1, 2, .L1@toc@h
 
 # CHECK: Disassembly of section .R_PPC64_TOC16_HI:
 # CHECK: .FR_PPC64_TOC16_HI:
-# CHECK: 10010014: {{.*}} addis 1, 2, -2
+# CHECK: 10010014: {{.*}} addis 1, 2, -1
 
 .section .R_PPC64_TOC16_HA,"ax",@progbits
 .globl .FR_PPC64_TOC16_HA
 .FR_PPC64_TOC16_HA:
   addis 1, 2, .L1@toc@ha
 
 # CHECK: Disassembly of section .R_PPC64_TOC16_HA:
 # CHECK: .FR_PPC64_TOC16_HA:
-# CHECK: 10010018: {{.*}} addis 1, 2, -1
+# CHECK: 10010018: {{.*}} addis 1, 2, 0
 
 .section .R_PPC64_REL24,"ax",@progbits
 .globl .FR_PPC64_REL24
 .FR_PPC64_REL24:
   b .Lfoox
 .section .R_PPC64_REL24_2,"ax",@progbits
 .Lfoox:
 

 # DATABE: .rodata:
 # DATABE: 10000190: 00 00 fe b4
 
 # Address of rodata + value stored at rodata entry
 # should equal address of LBB0_2.
 # 0x10000190 + 0xfeb4 = 0x10010044
 # CHECK: Disassembly of section .R_PPC64_REL32:
 # CHECK: .FR_PPC64_REL32:
-# CHECK: 1001003c: {{.*}} addis 5, 2, -1
+# CHECK: 1001003c: {{.*}} addis 5, 2, 0
 # CHECK: 10010040: {{.*}} ld 5, -32736(5)
 # CHECK: 10010044: {{.*}} add 3, 3, 4
 
 .section .R_PPC64_REL64, "ax",@progbits
 .globl  .FR_PPC64_REL64
 .FR_PPC64_REL64:
         .cfi_startproc
         .cfi_personality 148, __foo