[ELF][PPC64] Synthesize _savegpr[01]_{14..31} and _restgpr[01]_{14..31}

In the 64-bit ELF V2 API Specification: Power Architecture, 2.3.3.1. GPR
Save and Restore Functions defines some special functions which may be
referenced by GCC produced assembly (LLVM does not reference them).

With GCC -Os, when the number of call-saved registers exceeds a certain
threshold, GCC generates `_savegpr0_* _restgpr0_*` calls and expects the
linker to define them. See
https://sourceware.org/pipermail/binutils/2002-February/017444.html and
https://sourceware.org/pipermail/binutils/2004-August/036765.html . This
is weird because libgcc.a would be the natural place. However, the linker
generation approach has the advantage that the linker can generate
multiple copies to avoid long branch thunks. We don't consider the
advantage significant enough to complicate our trunk implementation, so
we take a simple approach.

* Check whether `_savegpr0_{14..31}` are used
* If yes, define needed symbols and add an InputSection with the code sequence.

`_savegpr1_*` `_restgpr0_*` and `_restgpr1_*` are similar.

Reviewed By: sfertile

Differential Revision: https://reviews.llvm.org/D79977
This commit is contained in:
Fangrui Song 2020-05-14 13:55:20 -07:00
parent d6c8736287
commit bae7cf6746
8 changed files with 256 additions and 0 deletions

View File

@ -6,11 +6,13 @@
//
//===----------------------------------------------------------------------===//
#include "SymbolTable.h"
#include "Symbols.h"
#include "SyntheticSections.h"
#include "Target.h"
#include "Thunks.h"
#include "lld/Common/ErrorHandler.h"
#include "lld/Common/Memory.h"
#include "llvm/Support/Endian.h"
using namespace llvm;
@ -104,6 +106,84 @@ bool elf::isPPC64SmallCodeModelTocReloc(RelType type) {
return type == R_PPC64_TOC16 || type == R_PPC64_TOC16_DS;
}
static bool addOptional(StringRef name, uint64_t value,
std::vector<Defined *> &defined) {
Symbol *sym = symtab->find(name);
if (!sym || sym->isDefined())
return false;
sym->resolve(Defined{/*file=*/nullptr, saver.save(name), STB_GLOBAL,
STV_HIDDEN, STT_FUNC, value,
/*size=*/0, /*section=*/nullptr});
defined.push_back(cast<Defined>(sym));
return true;
}
// If from is 14, write ${prefix}14: firstInsn; ${prefix}15:
// firstInsn+0x200008; ...; ${prefix}31: firstInsn+(31-14)*0x200008; $tail
// The labels are defined only if they exist in the symbol table.
static void writeSequence(MutableArrayRef<uint32_t> buf, const char *prefix,
int from, uint32_t firstInsn,
ArrayRef<uint32_t> tail) {
std::vector<Defined *> defined;
char name[16];
int first;
uint32_t *ptr = buf.data();
for (int r = from; r < 32; ++r) {
format("%s%d", prefix, r).snprint(name, sizeof(name));
if (addOptional(name, 4 * (r - from), defined) && defined.size() == 1)
first = r - from;
write32(ptr++, firstInsn + 0x200008 * (r - from));
}
for (uint32_t insn : tail)
write32(ptr++, insn);
assert(ptr == &*buf.end());
if (defined.empty())
return;
// The full section content has the extent of [begin, end). We drop unused
// instructions and write [first,end).
auto *sec = make<InputSection>(
nullptr, SHF_ALLOC, SHT_PROGBITS, 4,
makeArrayRef(reinterpret_cast<uint8_t *>(buf.data() + first),
4 * (buf.size() - first)),
".text");
inputSections.push_back(sec);
for (Defined *sym : defined) {
sym->section = sec;
sym->value -= 4 * first;
}
}
// Implements some save and restore functions as described by ELF V2 ABI to be
// compatible with GCC. With GCC -Os, when the number of call-saved registers
// exceeds a certain threshold, GCC generates _savegpr0_* _restgpr0_* calls and
// expects the linker to define them. See
// https://sourceware.org/pipermail/binutils/2002-February/017444.html and
// https://sourceware.org/pipermail/binutils/2004-August/036765.html . This is
// weird because libgcc.a would be the natural place. The linker generation
// approach has the advantage that the linker can generate multiple copies to
// avoid long branch thunks. However, we don't consider the advantage
// significant enough to complicate our trunk implementation, so we take the
// simple approach and synthesize .text sections providing the implementation.
void elf::addPPC64SaveRestore() {
static uint32_t savegpr0[20], restgpr0[21], savegpr1[19], restgpr1[19];
constexpr uint32_t blr = 0x4e800020, mtlr_0 = 0x7c0803a6;
// _restgpr0_14: ld 14, -144(1); _restgpr0_15: ld 15, -136(1); ...
// Tail: ld 0, 16(1); mtlr 0; blr
writeSequence(restgpr0, "_restgpr0_", 14, 0xe9c1ff70,
{0xe8010010, mtlr_0, blr});
// _restgpr1_14: ld 14, -144(12); _restgpr1_15: ld 15, -136(12); ...
// Tail: blr
writeSequence(restgpr1, "_restgpr1_", 14, 0xe9ccff70, {blr});
// _savegpr0_14: std 14, -144(1); _savegpr0_15: std 15, -136(1); ...
// Tail: std 0, 16(1); blr
writeSequence(savegpr0, "_savegpr0_", 14, 0xf9c1ff70, {0xf8010010, blr});
// _savegpr1_14: std 14, -144(12); _savegpr1_15: std 15, -136(12); ...
// Tail: blr
writeSequence(savegpr1, "_savegpr1_", 14, 0xf9ccff70, {blr});
}
// Find the R_PPC64_ADDR64 in .rela.toc with matching offset.
template <typename ELFT>
static std::pair<Defined *, int64_t>

View File

@ -213,6 +213,7 @@ unsigned getPPC64GlobalEntryToLocalEntryOffset(uint8_t stOther);
// the .toc section.
bool isPPC64SmallCodeModelTocReloc(RelType type);
void addPPC64SaveRestore();
uint64_t getPPC64TocBase();
uint64_t getAArch64Page(uint64_t expr);

View File

@ -264,6 +264,8 @@ void elf::addReservedSymbols() {
// glibc *crt1.o has a undefined reference to _SDA_BASE_. Since we don't
// support Small Data Area, define it arbitrarily as 0.
addOptionalRegular("_SDA_BASE_", nullptr, 0, STV_HIDDEN);
} else if (config->emachine == EM_PPC64) {
addPPC64SaveRestore();
}
// The Power Architecture 64-bit v2 ABI defines a TableOfContents (TOC) which

View File

@ -0,0 +1,38 @@
# REQUIRES: ppc
## Test code sequences of synthesized _restgpr0_{14..31}
# RUN: llvm-mc -filetype=obj -triple=ppc64le %s -o %t14.o
# RUN: ld.lld %t14.o -o %t14
# RUN: llvm-objdump -d %t14 | FileCheck --check-prefix=R14 %s
# R14-LABEL: <_restgpr0_14>:
# R14-NEXT: ld 14, -144(1)
# R14-NEXT: ld 15, -136(1)
# R14-EMPTY:
# R14-NEXT: <_restgpr0_16>:
# R14-NEXT: ld 16, -128(1)
# R14: ld 31, -8(1)
# R14-NEXT: ld 0, 16(1)
# R14-NEXT: mtlr 0
# R14-NEXT: blr
## Don't synthesize _restgpr0_{14..30} because they are unused.
# RUN: echo 'bl _restgpr0_31' | llvm-mc -filetype=obj -triple=ppc64 - -o %t31.o
# RUN: ld.lld %t31.o -o %t31
# RUN: llvm-objdump -d %t31 | FileCheck --check-prefix=R31 %s
# R31-LABEL: Disassembly of section .text:
# R31-EMPTY:
# R31-NEXT: <_restgpr0_31>:
# R31-NEXT: ld 31, -8(1)
# R31-NEXT: ld 0, 16(1)
# R31-NEXT: mtlr 0
# R31-NEXT: blr
# RUN: echo 'bl _restgpr0_32' | llvm-mc -filetype=obj -triple=ppc64 - -o %t32.o
# RUN: not ld.lld %t32.o -o /dev/null
.globl _start
_start:
bl _restgpr0_14
bl _restgpr0_16

View File

@ -0,0 +1,34 @@
# REQUIRES: ppc
## Test code sequences of synthesized _restgpr1_{14..31}
# RUN: llvm-mc -filetype=obj -triple=ppc64le %s -o %t14.o
# RUN: ld.lld %t14.o -o %t14
# RUN: llvm-objdump -d %t14 | FileCheck --check-prefix=R14 %s
# R14: <_restgpr1_14>:
# R14-NEXT: ld 14, -144(12)
# R14-NEXT: ld 15, -136(12)
# R14-EMPTY:
# R14-NEXT: <_restgpr1_16>:
# R14-NEXT: ld 16, -128(12)
# R14: ld 31, -8(12)
# R14-NEXT: blr
## Don't synthesize _restgpr1_{14..30} because they are unused.
# RUN: echo 'bl _restgpr1_31' | llvm-mc -filetype=obj -triple=ppc64 - -o %t31.o
# RUN: ld.lld %t31.o -o %t31
# RUN: llvm-objdump -d %t31 | FileCheck --check-prefix=R31 %s
# R31-LABEL: Disassembly of section .text:
# R31-EMPTY:
# R31-NEXT: <_restgpr1_31>:
# R31-NEXT: ld 31, -8(12)
# R31-NEXT: blr
# RUN: echo 'bl _restgpr1_32' | llvm-mc -filetype=obj -triple=ppc64le - -o %t32.o
# RUN: not ld.lld %t32.o -o /dev/null
.globl _start
_start:
bl _restgpr1_14
bl _restgpr1_16

View File

@ -0,0 +1,36 @@
# REQUIRES: ppc
## Test code sequences of synthesized _savegpr0_{14..31}
# RUN: llvm-mc -filetype=obj -triple=ppc64le %s -o %t14.o
# RUN: ld.lld %t14.o -o %t14
# RUN: llvm-objdump -d %t14 | FileCheck --check-prefix=R14 %s
# R14-LABEL: <_savegpr0_14>:
# R14-NEXT: std 14, -144(1)
# R14-NEXT: std 15, -136(1)
# R14-EMPTY:
# R14-NEXT: <_savegpr0_16>:
# R14-NEXT: std 16, -128(1)
# R14: std 31, -8(1)
# R14-NEXT: std 0, 16(1)
# R14-NEXT: blr
## Don't synthesize _savegpr0_{14..30} because they are unused.
# RUN: echo 'bl _savegpr0_31' | llvm-mc -filetype=obj -triple=ppc64 - -o %t31.o
# RUN: ld.lld %t31.o -o %t31
# RUN: llvm-objdump -d %t31 | FileCheck --check-prefix=R31 %s
# R31-LABEL: Disassembly of section .text:
# R31-EMPTY:
# R31-NEXT: <_savegpr0_31>:
# R31-NEXT: std 31, -8(1)
# R31-NEXT: std 0, 16(1)
# R31-NEXT: blr
# RUN: echo 'bl _savegpr0_32' | llvm-mc -filetype=obj -triple=ppc64 - -o %t32.o
# RUN: not ld.lld %t32.o -o /dev/null
.globl _start
_start:
bl _savegpr0_14
bl _savegpr0_16

View File

@ -0,0 +1,34 @@
# REQUIRES: ppc
## Test code sequences of synthesized _savegpr1_{14..31}
# RUN: llvm-mc -filetype=obj -triple=ppc64le %s -o %t14.o
# RUN: ld.lld %t14.o -o %t14
# RUN: llvm-objdump -d %t14 | FileCheck --check-prefix=R14 %s
# R14-LABEL: <_savegpr1_14>:
# R14-NEXT: std 14, -144(12)
# R14-NEXT: std 15, -136(12)
# R14-EMPTY:
# R14-NEXT: <_savegpr1_16>:
# R14-NEXT: std 16, -128(12)
# R14: std 31, -8(12)
# R14-NEXT: blr
## Don't synthesize _savegpr1_{14..30} because they are unused.
# RUN: echo 'bl _savegpr1_31' | llvm-mc -filetype=obj -triple=ppc64 - -o %t31.o
# RUN: ld.lld %t31.o -o %t31
# RUN: llvm-objdump -d %t31 | FileCheck --check-prefix=R31 %s
# R31-LABEL: Disassembly of section .text:
# R31-EMPTY:
# R31-NEXT: <_savegpr1_31>:
# R31-NEXT: std 31, -8(12)
# R31-NEXT: blr
# RUN: echo 'bl _savegpr1_32' | llvm-mc -filetype=obj -triple=ppc64le - -o %t32.o
# RUN: not ld.lld %t32.o -o /dev/null
.globl _start
_start:
bl _savegpr1_14
bl _savegpr1_16

View File

@ -0,0 +1,31 @@
# REQUIRES: ppc
## Test that some save and restore functions can be synthesized.
## The code sequences are tested by ppc64-restgpr*.s and ppc64-savegpr*.s
# RUN: llvm-mc -filetype=obj -triple=ppc64le %s -o %t.o
# RUN: ld.lld -shared %t.o -o %t.so
# RUN: llvm-readelf -s %t.so | FileCheck --check-prefix=NM %s
# RUN: llvm-objdump -d %t.so | FileCheck %s
## The synthesized symbols are not exported.
# NM: FUNC LOCAL HIDDEN {{.*}} _restgpr0_30
# NM-NEXT: FUNC LOCAL HIDDEN {{.*}} _restgpr1_30
# NM-NEXT: FUNC LOCAL HIDDEN {{.*}} _savegpr0_30
# NM-NEXT: FUNC LOCAL HIDDEN {{.*}} _savegpr1_30
# CHECK: 00000000000[[#%x,RESTGPR0:]] <_restgpr0_30>:
# CHECK: 00000000000[[#%x,RESTGPR1:]] <_restgpr1_30>:
# CHECK: 00000000000[[#%x,SAVEGPR0:]] <_savegpr0_30>:
# CHECK: 00000000000[[#%x,SAVEGPR1:]] <_savegpr1_30>:
# CHECK-LABEL: <_start>:
# CHECK-NEXT: bl 0x[[#RESTGPR0]]
# CHECK-NEXT: bl 0x[[#RESTGPR1]]
# CHECK-NEXT: bl 0x[[#SAVEGPR0]]
# CHECK-NEXT: bl 0x[[#SAVEGPR1]]
.globl _start
_start:
bl _restgpr0_30
bl _restgpr1_30
bl _savegpr0_30
bl _savegpr1_30