[AArch64,ELF] Allow implicit $d/$x at section beginning
The start state of a new section is `EMS_None`, often leading to a $d/$x at offset 0. Introduce a MCTargetOption/cl::opt "implicit-mapsyms" to allow an alternative behavior (https://github.com/ARM-software/abi-aa/issues/274): * Set the start state to `EMS_Data` or `EMS_A64`. * For text sections, add an ending $x only if the final data is not instructions. * For non-text sections, add an ending $d only if the final data is not data commands. ``` .section .text.1,"ax" nop // emit $d .long 42 // emit $x .section .text.2,"ax" nop ``` This new behavior decreases the .symtab size significantly: ``` % bloaty a64-2/bin/clang -- a64-0/bin/clang FILE SIZE VM SIZE -------------- -------------- -5.4% -1.13Mi [ = ] 0 .strtab -50.9% -4.09Mi [ = ] 0 .symtab -4.0% -5.22Mi [ = ] 0 TOTAL ``` --- This scheme works as long as the user can rule out some error scenarios: * .text.1 assembled using the traditional behavior is combined with .text.2 using the new behavior * A linker script combining non-text sections and text sections. The lack of mapping symbols in the non-text sections could make them treated as code, unless the linker inserts extra mapping symbols. The above mix-and-match scenarios aren't an issue at all for a significant portion of users. A text section may start with data commands in rare cases (e.g. -fsanitize=function) that many users don't care about. When combing `(.text.0; .word 0)` and `(.text.1; .word 0)`, the ending $x of .text.0 and the initial $d of .text.1 may have the same address. If both sections reside in the same file, ensure the ending symbol comes before the initial $d of .text.1, so that a dumb linker respecting the symbol order will place the ending $x before the initial $d. Disassemblers using stable sort will see both symbols at the same address, and the second will win. When section ordering mechanisms (e.g. --symbol-ordering-file, --call-graph-profile-sort, `.text : { second.o(.text) first.o(.text) }`) are involved, the initial data in a text section following a text section with trailing data could be misidentified as code, but the issue is local and the risk could be acceptable. Pull Request: https://github.com/llvm/llvm-project/pull/99718
This commit is contained in:
Fangrui Song
GitHub
parent
0bd90ec421
commit
46707b0a83
45
lld/test/ELF/aarch64-mapsyms-implicit.s
Normal file
45
lld/test/ELF/aarch64-mapsyms-implicit.s
Normal file
@ -0,0 +1,45 @@
|
||||
# REQUIRES: aarch64
|
||||
# RUN: llvm-mc -filetype=obj -triple=aarch64 -implicit-mapsyms %s -o %t.o
|
||||
# RUN: ld.lld %t.o -z keep-text-section-prefix -o %t
|
||||
# RUN: llvm-objdump -d --no-print-imm-hex --show-all-symbols %t | FileCheck %s
|
||||
|
||||
# CHECK: <_start>:
|
||||
# CHECK-NEXT: nop
|
||||
# CHECK-EMPTY:
|
||||
# CHECK-NEXT: <$d>:
|
||||
# CHECK-NEXT: .word 0x0000002a
|
||||
# CHECK-EMPTY:
|
||||
# CHECK-NEXT: <$x>:
|
||||
# CHECK-NEXT: nop
|
||||
# CHECK-EMPTY:
|
||||
# CHECK-NEXT: Disassembly of section .text.hot:
|
||||
# CHECK-EMPTY:
|
||||
# CHECK-NEXT: <.text.hot>:
|
||||
# CHECK-NEXT: nop
|
||||
# CHECK-EMPTY:
|
||||
# CHECK-NEXT: <$d>:
|
||||
# CHECK-NEXT: .word 0x0000002a
|
||||
# CHECK-EMPTY:
|
||||
# CHECK-NEXT: <$d>:
|
||||
# CHECK-NEXT: <$x>:
|
||||
# CHECK-NEXT: udf #42
|
||||
# CHECK-EMPTY:
|
||||
# CHECK-NEXT: <$x>:
|
||||
# CHECK-NEXT: nop
|
||||
|
||||
## Trailing data followed by a section starting with an instruction.
|
||||
.section .text.1,"ax"
|
||||
.globl _start
|
||||
_start:
|
||||
nop
|
||||
.long 42
|
||||
.section .text.2,"ax"
|
||||
nop
|
||||
|
||||
## Trailing data followed by a section starting with a data directive.
|
||||
.section .text.hot.1,"ax"
|
||||
nop
|
||||
.long 42
|
||||
.section .text.hot.2,"ax"
|
||||
.long 42
|
||||
nop
|
||||
@ -218,6 +218,7 @@ public:
|
||||
const_iterator begin() const { return Sections.begin(); }
|
||||
const_iterator end() const { return Sections.end(); }
|
||||
|
||||
SmallVectorImpl<const MCSymbol *> &getSymbols() { return Symbols; }
|
||||
iterator_range<pointee_iterator<
|
||||
typename SmallVector<const MCSymbol *, 0>::const_iterator>>
|
||||
symbols() const {
|
||||
|
||||
@ -64,6 +64,8 @@ public:
|
||||
// Use CREL relocation format for ELF.
|
||||
bool Crel = false;
|
||||
|
||||
bool ImplicitMapSyms = false;
|
||||
|
||||
// If true, prefer R_X86_64_[REX_]GOTPCRELX to R_X86_64_GOTPCREL on x86-64
|
||||
// ELF.
|
||||
bool X86RelaxRelocations = true;
|
||||
|
||||
@ -53,6 +53,8 @@ bool getSaveTempLabels();
|
||||
|
||||
bool getCrel();
|
||||
|
||||
bool getImplicitMapSyms();
|
||||
|
||||
bool getX86RelaxRelocations();
|
||||
|
||||
bool getX86Sse2Avx();
|
||||
|
||||
@ -48,6 +48,7 @@ MCOPT(bool, NoDeprecatedWarn)
|
||||
MCOPT(bool, NoTypeCheck)
|
||||
MCOPT(bool, SaveTempLabels)
|
||||
MCOPT(bool, Crel)
|
||||
MCOPT(bool, ImplicitMapSyms)
|
||||
MCOPT(bool, X86RelaxRelocations)
|
||||
MCOPT(bool, X86Sse2Avx)
|
||||
MCOPT(std::string, ABIName)
|
||||
@ -134,6 +135,14 @@ llvm::mc::RegisterMCTargetOptionsFlags::RegisterMCTargetOptionsFlags() {
|
||||
cl::desc("Use CREL relocation format for ELF"));
|
||||
MCBINDOPT(Crel);
|
||||
|
||||
static cl::opt<bool> ImplicitMapSyms(
|
||||
"implicit-mapsyms",
|
||||
cl::desc("Allow mapping symbol at section beginning to be implicit, "
|
||||
"lowering number of mapping symbols at the expense of some "
|
||||
"portability. Recommended for projects that can build all their "
|
||||
"object files using this option"));
|
||||
MCBINDOPT(ImplicitMapSyms);
|
||||
|
||||
static cl::opt<bool> X86RelaxRelocations(
|
||||
"x86-relax-relocations",
|
||||
cl::desc(
|
||||
@ -174,6 +183,7 @@ MCTargetOptions llvm::mc::InitMCTargetOptionsFromFlags() {
|
||||
Options.MCNoTypeCheck = getNoTypeCheck();
|
||||
Options.MCSaveTempLabels = getSaveTempLabels();
|
||||
Options.Crel = getCrel();
|
||||
Options.ImplicitMapSyms = getImplicitMapSyms();
|
||||
Options.X86RelaxRelocations = getX86RelaxRelocations();
|
||||
Options.X86Sse2Avx = getX86Sse2Avx();
|
||||
Options.EmitDwarfUnwind = getEmitDwarfUnwind();
|
||||
|
||||
@ -24,14 +24,15 @@
|
||||
#include "llvm/MC/MCAssembler.h"
|
||||
#include "llvm/MC/MCCodeEmitter.h"
|
||||
#include "llvm/MC/MCContext.h"
|
||||
#include "llvm/MC/MCELFObjectWriter.h"
|
||||
#include "llvm/MC/MCELFStreamer.h"
|
||||
#include "llvm/MC/MCExpr.h"
|
||||
#include "llvm/MC/MCInst.h"
|
||||
#include "llvm/MC/MCObjectWriter.h"
|
||||
#include "llvm/MC/MCSectionELF.h"
|
||||
#include "llvm/MC/MCStreamer.h"
|
||||
#include "llvm/MC/MCSubtargetInfo.h"
|
||||
#include "llvm/MC/MCSymbolELF.h"
|
||||
#include "llvm/MC/MCTargetOptions.h"
|
||||
#include "llvm/MC/MCWinCOFFStreamer.h"
|
||||
#include "llvm/Support/Casting.h"
|
||||
#include "llvm/Support/FormattedStream.h"
|
||||
@ -176,19 +177,29 @@ void AArch64TargetAsmStreamer::emitInst(uint32_t Inst) {
|
||||
/// by MachO. Beware!
|
||||
class AArch64ELFStreamer : public MCELFStreamer {
|
||||
public:
|
||||
friend AArch64TargetELFStreamer;
|
||||
AArch64ELFStreamer(MCContext &Context, std::unique_ptr<MCAsmBackend> TAB,
|
||||
std::unique_ptr<MCObjectWriter> OW,
|
||||
std::unique_ptr<MCCodeEmitter> Emitter)
|
||||
: MCELFStreamer(Context, std::move(TAB), std::move(OW),
|
||||
std::move(Emitter)),
|
||||
LastEMS(EMS_None) {}
|
||||
LastEMS(EMS_None) {
|
||||
auto *TO = getContext().getTargetOptions();
|
||||
ImplicitMapSyms = TO && TO->ImplicitMapSyms;
|
||||
}
|
||||
|
||||
void changeSection(MCSection *Section, uint32_t Subsection = 0) override {
|
||||
// We have to keep track of the mapping symbol state of any sections we
|
||||
// use. Each one should start off as EMS_None, which is provided as the
|
||||
// default constructor by DenseMap::lookup.
|
||||
// Save the mapping symbol state for potential reuse when revisiting the
|
||||
// section. When ImplicitMapSyms is true, the initial state is
|
||||
// EMS_A64 for text sections and EMS_Data for the others.
|
||||
LastMappingSymbols[getCurrentSection().first] = LastEMS;
|
||||
LastEMS = LastMappingSymbols.lookup(Section);
|
||||
auto It = LastMappingSymbols.find(Section);
|
||||
if (It != LastMappingSymbols.end())
|
||||
LastEMS = It->second;
|
||||
else if (ImplicitMapSyms)
|
||||
LastEMS = Section->isText() ? EMS_A64 : EMS_Data;
|
||||
else
|
||||
LastEMS = EMS_None;
|
||||
|
||||
MCELFStreamer::changeSection(Section, Subsection);
|
||||
}
|
||||
@ -269,13 +280,15 @@ private:
|
||||
LastEMS = EMS_A64;
|
||||
}
|
||||
|
||||
void emitMappingSymbol(StringRef Name) {
|
||||
MCSymbol *emitMappingSymbol(StringRef Name) {
|
||||
auto *Symbol = cast<MCSymbolELF>(getContext().createLocalSymbol(Name));
|
||||
emitLabel(Symbol);
|
||||
return Symbol;
|
||||
}
|
||||
|
||||
DenseMap<const MCSection *, ElfMappingSymbol> LastMappingSymbols;
|
||||
ElfMappingSymbol LastEMS;
|
||||
bool ImplicitMapSyms;
|
||||
};
|
||||
} // end anonymous namespace
|
||||
|
||||
@ -297,6 +310,58 @@ void AArch64TargetELFStreamer::finish() {
|
||||
AArch64ELFStreamer &S = getStreamer();
|
||||
MCContext &Ctx = S.getContext();
|
||||
auto &Asm = S.getAssembler();
|
||||
|
||||
// If ImplicitMapSyms is specified, ensure that text sections end with
|
||||
// the A64 state while non-text sections end with the data state. When
|
||||
// sections are combined by the linker, the subsequent section will start with
|
||||
// the right state. The ending mapping symbol is added right after the last
|
||||
// symbol relative to the section. When a dumb linker combines (.text.0; .word
|
||||
// 0) and (.text.1; .word 0), the ending $x of .text.0 precedes the $d of
|
||||
// .text.1, even if they have the same address.
|
||||
if (S.ImplicitMapSyms) {
|
||||
auto &Syms = Asm.getSymbols();
|
||||
const size_t NumSyms = Syms.size();
|
||||
DenseMap<MCSection *, std::pair<size_t, MCSymbol *>> EndMapSym;
|
||||
for (MCSection &Sec : Asm) {
|
||||
S.switchSection(&Sec);
|
||||
if (S.LastEMS == (Sec.isText() ? AArch64ELFStreamer::EMS_Data
|
||||
: AArch64ELFStreamer::EMS_A64))
|
||||
EndMapSym.insert(
|
||||
{&Sec, {NumSyms, S.emitMappingSymbol(Sec.isText() ? "$x" : "$d")}});
|
||||
}
|
||||
if (Syms.size() != NumSyms) {
|
||||
SmallVector<const MCSymbol *, 0> NewSyms;
|
||||
DenseMap<MCSection *, size_t> Cnt;
|
||||
Syms.truncate(NumSyms);
|
||||
// Find the last symbol index for each candidate section.
|
||||
for (auto [I, Sym] : llvm::enumerate(Syms)) {
|
||||
if (!Sym->isInSection())
|
||||
continue;
|
||||
auto It = EndMapSym.find(&Sym->getSection());
|
||||
if (It != EndMapSym.end())
|
||||
It->second.first = I;
|
||||
}
|
||||
SmallVector<size_t, 0> Idx;
|
||||
for (auto [I, Sym] : llvm::enumerate(Syms)) {
|
||||
NewSyms.push_back(Sym);
|
||||
if (!Sym->isInSection())
|
||||
continue;
|
||||
auto It = EndMapSym.find(&Sym->getSection());
|
||||
// If `Sym` is the last symbol relative to the section, add the ending
|
||||
// mapping symbol after `Sym`.
|
||||
if (It != EndMapSym.end() && I == It->second.first) {
|
||||
NewSyms.push_back(It->second.second);
|
||||
Idx.push_back(I);
|
||||
}
|
||||
}
|
||||
Syms = std::move(NewSyms);
|
||||
// F.second holds the number of symbols added before the FILE symbol.
|
||||
// Take into account the inserted mapping symbols.
|
||||
for (auto &F : S.getWriter().getFileNames())
|
||||
F.second += llvm::lower_bound(Idx, F.second) - Idx.begin();
|
||||
}
|
||||
}
|
||||
|
||||
MCSectionELF *MemtagSec = nullptr;
|
||||
for (const MCSymbol &Symbol : Asm.symbols()) {
|
||||
const auto &Sym = cast<MCSymbolELF>(Symbol);
|
||||
|
||||
@ -1,5 +1,10 @@
|
||||
// RUN: llvm-mc -triple=aarch64 -filetype=obj %s | llvm-objdump -t - | FileCheck %s --match-full-lines
|
||||
// RUN: llvm-mc -triple=aarch64 -filetype=obj -implicit-mapsyms %s | llvm-objdump -t - | FileCheck %s --check-prefix=CHECK1 --match-full-lines
|
||||
|
||||
/// The test covers many state transitions. Let's use the first state and the last state to describe a section.
|
||||
/// .text goes through cd -> dd -> cc -> dd.
|
||||
/// .data goes through dd -> dc -> cd.
|
||||
.file "0.s"
|
||||
.section .text1,"ax"
|
||||
add w0, w0, w0
|
||||
|
||||
@ -12,29 +17,61 @@ add w0, w0, w0
|
||||
.popsection
|
||||
|
||||
.text
|
||||
add w1, w1, w1
|
||||
.word 42
|
||||
|
||||
.section .text1,"ax"
|
||||
add w1, w1, w1
|
||||
|
||||
.text
|
||||
add w1, w1, w1
|
||||
|
||||
.section .data,"aw"
|
||||
.word 42
|
||||
add w0, w0, w0
|
||||
|
||||
.text
|
||||
.word 42
|
||||
|
||||
## .rodata and subsequent symbols should be after the FILE symbol of "1.s".
|
||||
.file "1.s"
|
||||
.section .rodata,"a"
|
||||
.word 42
|
||||
add w0, w0, w0
|
||||
|
||||
.section .data,"aw"
|
||||
add w0, w0, w0
|
||||
.word 42
|
||||
|
||||
.text
|
||||
|
||||
.ident "clang"
|
||||
.section ".note.GNU-stack","",@progbits
|
||||
|
||||
// CHECK: SYMBOL TABLE:
|
||||
// CHECK-NEXT: 0000000000000000 l .text1 0000000000000000 $x
|
||||
// CHECK-NEXT: 0000000000000000 l .text 0000000000000000 $x
|
||||
// CHECK-NEXT: 0000000000000004 l .text 0000000000000000 $d
|
||||
// CHECK-NEXT: 0000000000000000 l .data 0000000000000000 $d
|
||||
// CHECK-NEXT: 0000000000000008 l .text 0000000000000000 $x
|
||||
// CHECK-NEXT: 000000000000000c l .text 0000000000000000 $d
|
||||
// CHECK-NEXT: 0000000000000000 l .rodata 0000000000000000 $d
|
||||
// CHECK-NEXT: 0000000000000004 l .rodata 0000000000000000 $x
|
||||
// CHECK-NEXT: 0000000000000000 l .comment 0000000000000000 $d
|
||||
// CHECK-NEXT: 0000000000000000 l df *ABS* 0000000000000000 0.s
|
||||
// CHECK-NEXT: 0000000000000000 l .text1 0000000000000000 $x
|
||||
// CHECK-NEXT: 0000000000000000 l .text 0000000000000000 $x
|
||||
// CHECK-NEXT: 0000000000000004 l .text 0000000000000000 $d
|
||||
// CHECK-NEXT: 0000000000000000 l .data 0000000000000000 $d
|
||||
// CHECK-NEXT: 000000000000000c l .text 0000000000000000 $x
|
||||
// CHECK-NEXT: 0000000000000008 l .data 0000000000000000 $x
|
||||
// CHECK-NEXT: 0000000000000010 l .text 0000000000000000 $d
|
||||
// CHECK-NEXT: 0000000000000000 l df *ABS* 0000000000000000 1.s
|
||||
// CHECK-NEXT: 0000000000000000 l .rodata 0000000000000000 $d
|
||||
// CHECK-NEXT: 0000000000000004 l .rodata 0000000000000000 $x
|
||||
// CHECK-NEXT: 0000000000000010 l .data 0000000000000000 $d
|
||||
// CHECK-NEXT: 0000000000000000 l .comment 0000000000000000 $d
|
||||
// CHECK-NOT: {{.}}
|
||||
|
||||
// CHECK1: SYMBOL TABLE:
|
||||
// CHECK1-NEXT: 0000000000000000 l df *ABS* 0000000000000000 0.s
|
||||
// CHECK1-NEXT: 0000000000000004 l .text 0000000000000000 $d
|
||||
// CHECK1-NEXT: 000000000000000c l .text 0000000000000000 $x
|
||||
// CHECK1-NEXT: 0000000000000008 l .data 0000000000000000 $x
|
||||
// CHECK1-NEXT: 0000000000000010 l .text 0000000000000000 $d
|
||||
// CHECK1-NEXT: 0000000000000014 l .text 0000000000000000 $x
|
||||
// CHECK1-NEXT: 0000000000000000 l df *ABS* 0000000000000000 1.s
|
||||
// CHECK1-NEXT: 0000000000000004 l .rodata 0000000000000000 $x
|
||||
// CHECK1-NEXT: 0000000000000008 l .rodata 0000000000000000 $d
|
||||
// CHECK1-NEXT: 0000000000000010 l .data 0000000000000000 $d
|
||||
// CHECK1-NOT: {{.}}
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user