From: Fangrui Song <i@maskray.me>
Date: Sat, 1 May 2021 00:19:45 +0000 (-0700)
Subject: [ELF] Simplify the condition adding .got header
X-Git-Tag: llvmorg-14-init~7897
X-Git-Url: http://review.tizen.org/git/?a=commitdiff_plain;h=818b508953c7684a6c104b89c6f6ce84e961d82e;p=platform%2Fupstream%2Fllvm.git

[ELF] Simplify the condition adding .got header

Adopt my suggestion in https://reviews.llvm.org/D91426#2653926 ,
generalizing the ppc64 specific code.

GNU ld and glibc ld.so has a contract about the first few entries of .got .
There are somewhat complex conditions when the header is needed. This patch
switches to a simpler approach: add a header unconditionally if
_GLOBAL_OFFSET_TABLE_ is used or the number of entries is more than just the
header.
---

diff --git a/lld/ELF/SyntheticSections.cpp b/lld/ELF/SyntheticSections.cpp
index 76f6d37..9c80edf 100644
--- a/lld/ELF/SyntheticSections.cpp
+++ b/lld/ELF/SyntheticSections.cpp
@@ -646,13 +646,7 @@ void EhFrameSection::writeTo(uint8_t *buf) {
 GotSection::GotSection()
     : SyntheticSection(SHF_ALLOC | SHF_WRITE, SHT_PROGBITS, config->wordsize,
                        ".got") {
-  // If ElfSym::globalOffsetTable is relative to .got and is referenced,
-  // increase numEntries by the number of entries used to emit
-  // ElfSym::globalOffsetTable.
-  // On PP64 we always add the header at the start.
-  if ((ElfSym::globalOffsetTable && !target->gotBaseSymInGotPlt) ||
-      config->emachine == EM_PPC64)
-    numEntries += target->gotHeaderEntriesNum;
+  numEntries = target->gotHeaderEntriesNum;
 }
 
 void GotSection::addEntry(Symbol &sym) {
@@ -696,16 +690,9 @@ void GotSection::finalizeContents() {
 }
 
 bool GotSection::isNeeded() const {
-  // We need to emit a GOT even if it's empty if there's a relocation that is
-  // relative to GOT(such as GOTOFFREL).
-
-  // On PPC64 we need to check that the number of entries is more than just the
-  // size of the header since the header is always added. A GOT with just the
-  // header may not actually be needed.
-  if (config->emachine == EM_PPC64)
-    return numEntries > target->gotHeaderEntriesNum || hasGotOffRel;
-
-  return numEntries || hasGotOffRel;
+  // Needed if the GOT symbol is used or the number of entries is more than just
+  // the header. A GOT with just the header may not be needed.
+  return hasGotOffRel || numEntries > target->gotHeaderEntriesNum;
 }
 
 void GotSection::writeTo(uint8_t *buf) {
diff --git a/lld/test/ELF/ppc32-tls-gd.s b/lld/test/ELF/ppc32-tls-gd.s
index c3a1ffd..dfff837 100644
--- a/lld/test/ELF/ppc32-tls-gd.s
+++ b/lld/test/ELF/ppc32-tls-gd.s
@@ -21,26 +21,26 @@
 # GD-DYN: PPC_GOT 0x20358
 
 # GD-REL:      .rela.dyn {
-# GD-REL-NEXT:   0x20358 R_PPC_DTPMOD32 a 0x0
-# GD-REL-NEXT:   0x2035C R_PPC_DTPREL32 a 0x0
-# GD-REL-NEXT:   0x20360 R_PPC_DTPMOD32 b 0x0
-# GD-REL-NEXT:   0x20364 R_PPC_DTPREL32 b 0x0
-# GD-REL-NEXT:   0x20368 R_PPC_DTPMOD32 c 0x0
-# GD-REL-NEXT:   0x2036C R_PPC_DTPREL32 c 0x0
+# GD-REL-NEXT:   0x20364 R_PPC_DTPMOD32 a 0x0
+# GD-REL-NEXT:   0x20368 R_PPC_DTPREL32 a 0x0
+# GD-REL-NEXT:   0x2036C R_PPC_DTPMOD32 b 0x0
+# GD-REL-NEXT:   0x20370 R_PPC_DTPREL32 b 0x0
+# GD-REL-NEXT:   0x20374 R_PPC_DTPMOD32 c 0x0
+# GD-REL-NEXT:   0x20378 R_PPC_DTPREL32 c 0x0
 # GD-REL-NEXT: }
 
-## &DTPMOD(a) - _GLOBAL_OFFSET_TABLE_ = 0x20078 - 0x20078 = 0
-# GD:      addi 3, 31, 0
+## &DTPMOD(a) - _GLOBAL_OFFSET_TABLE_ = 12
+# GD:      addi 3, 31, 12
 # GD-NEXT: bl 0x1028c
 # GD-NEXT: lwz 3, 0(3)
 
-## &DTPMOD(b) - _GLOBAL_OFFSET_TABLE_ = 0x20080 - 0x20078 = 8
-# GD-NEXT: addi 3, 31, 8
+## &DTPMOD(b) - _GLOBAL_OFFSET_TABLE_ = 20
+# GD-NEXT: addi 3, 31, 20
 # GD-NEXT: bl 0x1028c
 # GD-NEXT: lwz 3, 0(3)
 
-## &DTPMOD(c) - _GLOBAL_OFFSET_TABLE_ = 0x20088 - 0x20078 = 16
-# GD-NEXT: addi 3, 9, 16
+## &DTPMOD(c) - _GLOBAL_OFFSET_TABLE_ = 28
+# GD-NEXT: addi 3, 9, 28
 # GD-NEXT: bl 0x1028c
 # GD-NEXT: lwz 3, 0(3)
 
@@ -60,8 +60,8 @@
 # LE-NEXT: lwz 3, 0(3)
 
 # IE-REL:      .rela.dyn {
-# IE-REL-NEXT:   0x10020274 R_PPC_TPREL32 b 0x0
-# IE-REL-NEXT:   0x10020278 R_PPC_TPREL32 c 0x0
+# IE-REL-NEXT:   0x10020280 R_PPC_TPREL32 b 0x0
+# IE-REL-NEXT:   0x10020284 R_PPC_TPREL32 c 0x0
 # IE-REL-NEXT: }
 
 ## a is relaxed to use LE.
@@ -69,12 +69,12 @@
 # IE:      addis 3, 2, 0
 # IE-NEXT: addi 3, 3, -28664
 # IE-NEXT: lwz 3, 0(3)
-## &.got[0] - _GLOBAL_OFFSET_TABLE_ = 0
-# IE-NEXT: lwz 3, 0(31)
+## &.got[3] - _GLOBAL_OFFSET_TABLE_ = 12
+# IE-NEXT: lwz 3, 12(31)
 # IE-NEXT: add 3, 3, 2
 # IE-NEXT: lwz 3, 0(3)
-## &.got[1] - _GLOBAL_OFFSET_TABLE_ = 4
-# IE-NEXT: lwz 3, 4(9)
+## &.got[4] - _GLOBAL_OFFSET_TABLE_ = 16
+# IE-NEXT: lwz 3, 16(9)
 # IE-NEXT: add 3, 3, 2
 # IE-NEXT: lwz 3, 0(3)
 
diff --git a/lld/test/ELF/ppc32-tls-ie.s b/lld/test/ELF/ppc32-tls-ie.s
index 9bc8ffc..9157129 100644
--- a/lld/test/ELF/ppc32-tls-ie.s
+++ b/lld/test/ELF/ppc32-tls-ie.s
@@ -11,15 +11,15 @@
 
 ## A non-preemptable symbol (b) has 0 st_shndx.
 # IE-REL:      .rela.dyn {
-# IE-REL-NEXT:   0x20224 R_PPC_TPREL32 - 0xC
-# IE-REL-NEXT:   0x20220 R_PPC_TPREL32 a 0x0
+# IE-REL-NEXT:   0x20230 R_PPC_TPREL32 - 0xC
+# IE-REL-NEXT:   0x2022C R_PPC_TPREL32 a 0x0
 # IE-REL-NEXT: }
 
-## &.got[0] - _GLOBAL_OFFSET_TABLE_ = 0
-# IE:      lwz 10, 0(9)
+## &.got[3] - _GLOBAL_OFFSET_TABLE_ = 12
+# IE:      lwz 10, 12(9)
 # IE-NEXT: add 10, 10, 2
-## &.got[1] - _GLOBAL_OFFSET_TABLE_ = 4
-# IE-NEXT: lwz 8, 4(7)
+## &.got[4] - _GLOBAL_OFFSET_TABLE_ = 16
+# IE-NEXT: lwz 8, 16(7)
 # IE-NEXT: lbzx 10, 8, 2
 
 # NOREL: no relocations
diff --git a/lld/test/ELF/ppc32-tls-ld.s b/lld/test/ELF/ppc32-tls-ld.s
index 0eda8e9..d70a52a 100644
--- a/lld/test/ELF/ppc32-tls-ld.s
+++ b/lld/test/ELF/ppc32-tls-ld.s
@@ -11,11 +11,11 @@
 # RUN: llvm-objdump -d --no-show-raw-insn %t | FileCheck --check-prefix=LE %s
 
 # LD-REL:      .rela.dyn {
-# LD-REL-NEXT:   0x202C4 R_PPC_DTPMOD32 - 0x0
+# LD-REL-NEXT:   0x202D0 R_PPC_DTPMOD32 - 0x0
 # LD-REL-NEXT: }
 
-## .got - _GLOBAL_OFFSET_TABLE_ = 0
-# LD:      addi 3, 30, 0
+## &.got[3] - _GLOBAL_OFFSET_TABLE_ = 12
+# LD:      addi 3, 30, 12
 # LD-NEXT: bl 0x101f8
 ## a@dtprel = st_value(a)-0x8000 = 65540-0x8000 = 65536*1-32764
 ## b@dtprel = st_value(a)-0x8000 = 131080-0x8000 = 65536*2-32760
@@ -28,8 +28,8 @@
 
 ## Check that b@got@tlsld does not allocate another GOT entry.
 ## It shares In.Got->TlsIndexOff allocated when processing a@got@tlsld.
-## .got - _GLOBAL_OFFSET_TABLE_ = 0
-# LD-NEXT: addi 3, 9, 0
+## &.got[3] - _GLOBAL_OFFSET_TABLE_ = 12
+# LD-NEXT: addi 3, 9, 12
 # LD-NEXT: bl 0x101f8
 ## b@dtprel = st_value(a)-0x8000 = 131080-0x8000 = 65536*2-32760
 # LD-NEXT: addis 29, 3, 2
diff --git a/lld/test/ELF/riscv-tls-gd.s b/lld/test/ELF/riscv-tls-gd.s
index 0781994..21a9d42 100644
--- a/lld/test/ELF/riscv-tls-gd.s
+++ b/lld/test/ELF/riscv-tls-gd.s
@@ -46,40 +46,40 @@
 # RUN: llvm-readelf -x .got %t.64 | FileCheck --check-prefix=IE64-GOT %s
 
 # GD32-REL:      .rela.dyn {
-# GD32-REL-NEXT:   0x2310 R_RISCV_TLS_DTPMOD32 a 0x0
-# GD32-REL-NEXT:   0x2314 R_RISCV_TLS_DTPREL32 a 0x0
-# GD32-REL-NEXT:   0x2318 R_RISCV_TLS_DTPMOD32 b 0x0
-# GD32-REL-NEXT:   0x231C R_RISCV_TLS_DTPREL32 b 0x0
+# GD32-REL-NEXT:   0x2314 R_RISCV_TLS_DTPMOD32 a 0x0
+# GD32-REL-NEXT:   0x2318 R_RISCV_TLS_DTPREL32 a 0x0
+# GD32-REL-NEXT:   0x231C R_RISCV_TLS_DTPMOD32 b 0x0
+# GD32-REL-NEXT:   0x2320 R_RISCV_TLS_DTPREL32 b 0x0
 # GD32-REL-NEXT: }
 
-## &DTPMOD(a) - . = 0x2310 - 0x1250 = 4096*1+192
+## &DTPMOD(a) - . = 0x2314 - 0x1250 = 4096*1+196
 # GD32:      1250: auipc a0, 1
-# GD32-NEXT:       addi a0, a0, 192
+# GD32-NEXT:       addi a0, a0, 196
 # GD32-NEXT:       auipc ra, 0
 # GD32-NEXT:       jalr 56(ra)
 
-## &DTPMOD(b) - . = 0x2318 - 0x1260 = 4096*1+184
+## &DTPMOD(b) - . = 0x231C - 0x1260 = 4096*1+188
 # GD32:      1260: auipc a0, 1
-# GD32-NEXT:       addi a0, a0, 184
+# GD32-NEXT:       addi a0, a0, 188
 # GD32-NEXT:       auipc ra, 0
 # GD32-NEXT:       jalr 40(ra)
 
 # GD64-REL:      .rela.dyn {
-# GD64-REL-NEXT:   0x24D0 R_RISCV_TLS_DTPMOD64 a 0x0
-# GD64-REL-NEXT:   0x24D8 R_RISCV_TLS_DTPREL64 a 0x0
-# GD64-REL-NEXT:   0x24E0 R_RISCV_TLS_DTPMOD64 b 0x0
-# GD64-REL-NEXT:   0x24E8 R_RISCV_TLS_DTPREL64 b 0x0
+# GD64-REL-NEXT:   0x24D8 R_RISCV_TLS_DTPMOD64 a 0x0
+# GD64-REL-NEXT:   0x24E0 R_RISCV_TLS_DTPREL64 a 0x0
+# GD64-REL-NEXT:   0x24E8 R_RISCV_TLS_DTPMOD64 b 0x0
+# GD64-REL-NEXT:   0x24F0 R_RISCV_TLS_DTPREL64 b 0x0
 # GD64-REL-NEXT: }
 
-## &DTPMOD(a) - . = 0x24d0 - 0x1398 = 4096*1+312
+## &DTPMOD(a) - . = 0x24d8 - 0x1398 = 4096*1+320
 # GD64:      1398: auipc a0, 1
-# GD64-NEXT:       addi a0, a0, 312
+# GD64-NEXT:       addi a0, a0, 320
 # GD64-NEXT:       auipc ra, 0
 # GD64-NEXT:       jalr 64(ra)
 
-## &DTPMOD(b) - . = 0x24e0 - 0x13a8 = 4096*1+312
+## &DTPMOD(b) - . = 0x24e8 - 0x13a8 = 4096*1+320
 # GD64:      13a8: auipc a0, 1
-# GD64-NEXT:       addi a0, a0, 312
+# GD64-NEXT:       addi a0, a0, 320
 # GD64-NEXT:       auipc ra, 0
 # GD64-NEXT:       jalr 48(ra)
 
@@ -89,27 +89,31 @@
 ## a@dtprel = st_value(a)-0x800 = 0xfffff808
 ## b@dtprel = st_value(b)-0x800 = 0xfffff80c
 # LE32-GOT: section '.got':
-# LE32-GOT-NEXT: 0x{{[0-9a-f]+}} 01000000 08f8ffff 01000000 0cf8ffff
+# LE32-GOT-NEXT: 0x[[#%x,A:]] [[#%x,GOT:]] 01000000 08f8ffff 01000000
+# LE32-GOT-NEXT: 0x[[#%x,A:]] 0cf8ffff
 # LE64-GOT: section '.got':
-# LE64-GOT-NEXT: 0x{{[0-9a-f]+}} 01000000 00000000 08f8ffff ffffffff
-# LE64-GOT-NEXT: 0x{{[0-9a-f]+}} 01000000 00000000 0cf8ffff ffffffff
+# LE64-GOT-NEXT: 0x[[#%x,A:]] [[#%x,GOT:]] [[#%x,GOT:]] 01000000 00000000
+# LE64-GOT-NEXT: 0x[[#%x,A:]] 08f8ffff ffffffff 01000000 00000000
+# LE64-GOT-NEXT: 0x[[#%x,A:]] 0cf8ffff ffffffff
 
 ## a is local - relaxed to LE - its DTPMOD/DTPREL slots are link-time constants.
 ## b is external - DTPMOD/DTPREL dynamic relocations are required.
 # IE32-REL:      .rela.dyn {
-# IE32-REL-NEXT:   0x12230 R_RISCV_TLS_DTPMOD32 b 0x0
-# IE32-REL-NEXT:   0x12234 R_RISCV_TLS_DTPREL32 b 0x0
+# IE32-REL-NEXT:   0x12234 R_RISCV_TLS_DTPMOD32 b 0x0
+# IE32-REL-NEXT:   0x12238 R_RISCV_TLS_DTPREL32 b 0x0
 # IE32-REL-NEXT: }
 # IE32-GOT:      section '.got':
-# IE32-GOT-NEXT: 0x00012228 01000000 08f8ffff 00000000 00000000
+# IE32-GOT-NEXT: 0x00012228 c8210100 01000000 08f8ffff 00000000
+# IE32-GOT-NEXT: 0x00012238 00000000
 
 # IE64-REL:      .rela.dyn {
-# IE64-REL-NEXT:   0x12390 R_RISCV_TLS_DTPMOD64 b 0x0
-# IE64-REL-NEXT:   0x12398 R_RISCV_TLS_DTPREL64 b 0x0
+# IE64-REL-NEXT:   0x12398 R_RISCV_TLS_DTPMOD64 b 0x0
+# IE64-REL-NEXT:   0x123A0 R_RISCV_TLS_DTPREL64 b 0x0
 # IE64-REL-NEXT: }
 # IE64-GOT:      section '.got':
-# IE64-GOT-NEXT: 0x00012380 01000000 00000000 08f8ffff ffffffff
-# IE64-GOT-NEXT: 0x00012390 00000000 00000000 00000000 00000000
+# IE64-GOT-NEXT: 0x00012380 c0220100 00000000 01000000 00000000
+# IE64-GOT-NEXT: 0x00012390 08f8ffff ffffffff 00000000 00000000
+# IE64-GOT-NEXT: 0x000123a0 00000000 00000000
 
 la.tls.gd a0,a
 call __tls_get_addr@plt
diff --git a/lld/test/ELF/riscv-tls-ie.s b/lld/test/ELF/riscv-tls-ie.s
index 424edb9..6d58c99 100644
--- a/lld/test/ELF/riscv-tls-ie.s
+++ b/lld/test/ELF/riscv-tls-ie.s
@@ -24,27 +24,27 @@
 
 # IE32-REL:      FLAGS STATIC_TLS
 # IE32-REL:      .rela.dyn {
-# IE32-REL-NEXT:   0x2218 R_RISCV_TLS_TPREL32 - 0xC
-# IE32-REL-NEXT:   0x2214 R_RISCV_TLS_TPREL32 a 0x0
+# IE32-REL-NEXT:   0x221C R_RISCV_TLS_TPREL32 - 0xC
+# IE32-REL-NEXT:   0x2218 R_RISCV_TLS_TPREL32 a 0x0
 # IE32-REL-NEXT: }
 
 # IE64-REL:      FLAGS STATIC_TLS
 # IE64-REL:      .rela.dyn {
-# IE64-REL-NEXT:   0x2370 R_RISCV_TLS_TPREL64 - 0xC
-# IE64-REL-NEXT:   0x2368 R_RISCV_TLS_TPREL64 a 0x0
+# IE64-REL-NEXT:   0x2378 R_RISCV_TLS_TPREL64 - 0xC
+# IE64-REL-NEXT:   0x2370 R_RISCV_TLS_TPREL64 a 0x0
 # IE64-REL-NEXT: }
 
-## rv32: &.got[0] - . = 0x2214 - . = 4096*1+112
-## rv64: &.got[0] - . = 0x2368 - . = 4096*1+200
+## rv32: &.got[0] - . = 0x2218 - . = 4096*1+116
+## rv64: &.got[0] - . = 0x2378 - . = 4096*1+208
 # IE:              auipc a4, 1
-# IE32-NEXT:       lw a4, 112(a4)
-# IE64-NEXT:       ld a4, 200(a4)
+# IE32-NEXT:       lw a4, 116(a4)
+# IE64-NEXT:       ld a4, 208(a4)
 # IE-NEXT:         add a4, a4, tp
-## rv32: &.got[1] - . = 0x2218 - . = 4096*1+104
-## rv64: &.got[1] - . = 0x2370 - . = 4096*1+196
+## rv32: &.got[1] - . = 0x221c - . = 4096*1+108
+## rv64: &.got[1] - . = 0x2378 - . = 4096*1+204
 # IE:              auipc a5, 1
-# IE32-NEXT:       lw a5, 104(a5)
-# IE64-NEXT:       ld a5, 196(a5)
+# IE32-NEXT:       lw a5, 108(a5)
+# IE64-NEXT:       ld a5, 204(a5)
 # IE-NEXT:         add a5, a5, tp
 
 # NOREL: no relocations
@@ -52,23 +52,24 @@
 # a@tprel = st_value(a) = 0x8
 # b@tprel = st_value(a) = 0xc
 # LE32-GOT: section '.got':
-# LE32-GOT-NEXT: 0x0001212c 08000000 0c000000
+# LE32-GOT-NEXT: 0x0001212c 00000000 08000000 0c000000
 # LE64-GOT: section '.got':
-# LE64-GOT-NEXT: 0x000121e0 08000000 00000000 0c000000 00000000
+# LE64-GOT-NEXT: 0x000121e0 00000000 00000000 08000000 00000000
+# LE64-GOT-NEXT: 0x000121f0 0c000000 00000000
 
-## rv32: &.got[0] - . = 0x1212c - 0x11114 = 4096*1+24
-## rv64: &.got[0] - . = 0x121e0 - 0x111c8 = 4096*1+24
+## rv32: &.got[0] - . = 0x12130 - 0x11114 = 4096*1+28
+## rv64: &.got[0] - . = 0x121e8 - 0x111c8 = 4096*1+32
 # LE32:      11114: auipc a4, 1
-# LE32-NEXT:        lw a4, 24(a4)
+# LE32-NEXT:        lw a4, 28(a4)
 # LE64:      111c8: auipc a4, 1
-# LE64-NEXT:        ld a4, 24(a4)
+# LE64-NEXT:        ld a4, 32(a4)
 # LE-NEXT:          add a4, a4, tp
-## rv32: &.got[1] - . = 0x12130 - 0x11120 = 4096*1+16
-## rv64: &.got[1] - . = 0x121e8 - 0x111d4 = 4096*1+20
+## rv32: &.got[1] - . = 0x12134 - 0x11120 = 4096*1+20
+## rv64: &.got[1] - . = 0x121f0 - 0x111d4 = 4096*1+28
 # LE32:      11120: auipc a5, 1
-# LE32-NEXT:        lw a5, 16(a5)
+# LE32-NEXT:        lw a5, 20(a5)
 # LE64:      111d4: auipc a5, 1
-# LE64-NEXT:        ld a5, 20(a5)
+# LE64-NEXT:        ld a5, 28(a5)
 # LE-NEXT:          add a5, a5, tp
 
 la.tls.ie a4,a