[lld] 80caa1e - [lld/mac] Add support for segment$start$ and segment$end$ symbols
Nico Weber via llvm-commits
llvm-commits at lists.llvm.org
Sun Jul 25 15:27:46 PDT 2021
Author: Nico Weber
Date: 2021-07-25T18:25:13-04:00
New Revision: 80caa1eb4a0eab61debdcda515d00461a20520a6
URL: https://github.com/llvm/llvm-project/commit/80caa1eb4a0eab61debdcda515d00461a20520a6
DIFF: https://github.com/llvm/llvm-project/commit/80caa1eb4a0eab61debdcda515d00461a20520a6.diff
LOG: [lld/mac] Add support for segment$start$ and segment$end$ symbols
These symbols are somewhat interesting in that they create non-existing
segments, which as far as I know is the only way to create segments
that don't contain any sections.
Final part of part of PR50760. Like D106629, but for segments instead
of sections. I'm not aware of anything that needs this in practice.
Differential Revision: https://reviews.llvm.org/D106767
Added:
Modified:
lld/MachO/OutputSegment.cpp
lld/MachO/OutputSegment.h
lld/MachO/SymbolTable.cpp
lld/MachO/Writer.cpp
lld/test/MachO/start-end.s
Removed:
################################################################################
diff --git a/lld/MachO/OutputSegment.cpp b/lld/MachO/OutputSegment.cpp
index 7d9544d06ea1..3bbaf7f0304e 100644
--- a/lld/MachO/OutputSegment.cpp
+++ b/lld/MachO/OutputSegment.cpp
@@ -9,6 +9,7 @@
#include "OutputSegment.h"
#include "ConcatOutputSection.h"
#include "InputSection.h"
+#include "Symbols.h"
#include "SyntheticSections.h"
#include "lld/Common/ErrorHandler.h"
@@ -145,6 +146,13 @@ void OutputSegment::sortOutputSections() {
llvm::stable_sort(sections, compareByOrder<OutputSection *>(sectionOrder));
}
+void OutputSegment::assignAddressesToStartEndSymbols() {
+ for (Defined *d : segmentStartSymbols)
+ d->value = addr;
+ for (Defined *d : segmentEndSymbols)
+ d->value = addr + vmSize;
+}
+
void macho::sortOutputSegments() {
llvm::stable_sort(outputSegments,
compareByOrder<OutputSegment *>(segmentOrder));
diff --git a/lld/MachO/OutputSegment.h b/lld/MachO/OutputSegment.h
index 10d60b989a9b..b3863f4148d9 100644
--- a/lld/MachO/OutputSegment.h
+++ b/lld/MachO/OutputSegment.h
@@ -10,7 +10,9 @@
#define LLD_MACHO_OUTPUT_SEGMENT_H
#include "OutputSection.h"
+#include "Symbols.h"
#include "lld/Common/LLVM.h"
+#include "llvm/ADT/TinyPtrVector.h"
#include <limits>
#include <vector>
@@ -41,6 +43,7 @@ class OutputSegment {
public:
void addOutputSection(OutputSection *os);
void sortOutputSections();
+ void assignAddressesToStartEndSymbols();
const std::vector<OutputSection *> &getSections() const { return sections; }
size_t numNonHiddenSections() const;
@@ -55,6 +58,9 @@ class OutputSegment {
uint32_t initProt = 0;
uint8_t index;
+ llvm::TinyPtrVector<Defined *> segmentStartSymbols;
+ llvm::TinyPtrVector<Defined *> segmentEndSymbols;
+
private:
std::vector<OutputSection *> sections;
};
diff --git a/lld/MachO/SymbolTable.cpp b/lld/MachO/SymbolTable.cpp
index a09ff94c8873..c5808a89bf57 100644
--- a/lld/MachO/SymbolTable.cpp
+++ b/lld/MachO/SymbolTable.cpp
@@ -203,6 +203,12 @@ enum class Boundary {
End,
};
+static Defined *createBoundarySymbol(const Undefined &sym) {
+ return symtab->addSynthetic(
+ sym.getName(), /*isec=*/nullptr, /*value=*/-1, /*isPrivateExtern=*/true,
+ /*includeInSymtab=*/false, /*referencedDynamically=*/false);
+}
+
static void handleSectionBoundarySymbol(const Undefined &sym, StringRef segSect,
Boundary which) {
StringRef segName, sectName;
@@ -238,19 +244,19 @@ static void handleSectionBoundarySymbol(const Undefined &sym, StringRef segSect,
inputSections.push_back(isec);
}
- Defined *boundarySym = symtab->addSynthetic(
- sym.getName(), /*isec=*/nullptr, /*value=*/-1, /*isPrivateExtern=*/true,
- /*includeInSymtab=*/false, /*referencedDynamically=*/false);
if (which == Boundary::Start)
- osec->sectionStartSymbols.push_back(boundarySym);
+ osec->sectionStartSymbols.push_back(createBoundarySymbol(sym));
else
- osec->sectionEndSymbols.push_back(boundarySym);
+ osec->sectionEndSymbols.push_back(createBoundarySymbol(sym));
}
static void handleSegmentBoundarySymbol(const Undefined &sym, StringRef segName,
Boundary which) {
- // FIXME
- error("segment$start$ and segment$end$ symbols are not yet implemented");
+ OutputSegment *seg = getOrCreateOutputSegment(segName);
+ if (which == Boundary::Start)
+ seg->segmentStartSymbols.push_back(createBoundarySymbol(sym));
+ else
+ seg->segmentEndSymbols.push_back(createBoundarySymbol(sym));
}
void lld::macho::treatUndefinedSymbol(const Undefined &sym, StringRef source) {
diff --git a/lld/MachO/Writer.cpp b/lld/MachO/Writer.cpp
index 1f66cb1d06e0..d9c9cf570054 100644
--- a/lld/MachO/Writer.cpp
+++ b/lld/MachO/Writer.cpp
@@ -992,6 +992,7 @@ void Writer::finalizeAddresses() {
addr = alignTo(addr, pageSize);
seg->vmSize = addr - seg->addr;
seg->fileSize = fileOff - seg->fileOff;
+ seg->assignAddressesToStartEndSymbols();
}
}
diff --git a/lld/test/MachO/start-end.s b/lld/test/MachO/start-end.s
index a55ec6eecc80..dbe4f88f412b 100644
--- a/lld/test/MachO/start-end.s
+++ b/lld/test/MachO/start-end.s
@@ -1,11 +1,10 @@
# REQUIRES: x86
-## FIXME: Add tests for segment$start$foo, segment$end$foo once implemented.
-
# RUN: rm -rf %t; split-file %s %t
# RUN: llvm-mc -filetype=obj -triple=x86_64-apple-darwin %t/main.s -o %t/main.o
# RUN: llvm-mc -filetype=obj -triple=x86_64-apple-darwin %t/foo.s -o %t/foo.o
+# RUN: llvm-mc -filetype=obj -triple=x86_64-apple-darwin %t/seg.s -o %t/seg.o
# RUN: %lld -lSystem %t/main.o %t/foo.o -o %t.out \
# RUN: -rename_section __FOO __bar __BAZ __quux \
@@ -171,6 +170,101 @@
# CHECK: [[#%.16x, TEXTSTART]]
# CHECK-NOT: [[#%.16x, TEXTEND]]
+###############################################################################
+## Test segment$start and segment$end.
+
+# RUN: %lld -lSystem %t/seg.o -o %t.seg.out \
+# RUN: -rename_segment __FOO __BAZ \
+# RUN: -rename_segment __WHAT __FOO \
+# RUN: -u 'segment$start$__UFLAG_SEG' \
+# RUN: -e 'segment$start$__TEXT'
+# RUN: llvm-objdump --macho --syms %t.seg.out > %t-seg-dump.txt
+## llvm-objdump can't dump segment names; use lld-otool for this.
+# RUN: llvm-otool -l %t.seg.out | grep -A6 LC_SEGMENT >> %t-seg-dump.txt
+# RUN: llvm-objdump --macho -d --no-symbolic-operands --no-show-raw-insn %t.seg.out >> %t-seg-dump.txt
+# RUN: llvm-objdump --macho --function-starts %t.out >> %t-seg-dump.txt
+# RUN: FileCheck %s --check-prefix=SEG < %t-seg-dump.txt
+
+# SEG-LABEL: SYMBOL TABLE:
+# SEG-NOT: segment$start$__TEXT
+# SEG-NOT: segment$end$__TEXT
+# SEG-NOT: segment$start$__FOO
+# SEG-NOT: segment$end$__FOO
+# SEG-NOT: segment$start$__BAZ
+# SEG-NOT: segment$end$__BAZ
+# SEG-NOT: segment$start$__WHAT
+# SEG-NOT: segment$end$__WHAT
+# SEG-NOT: segment$start$__UFLAG_SEG
+# SEG-NOT: segment$start$__UFLAG_SEG
+# SEG: segment$start$REGULAR
+# SEG: segment$end$REGULAR
+
+# SEG: cmd LC_SEGMENT_64
+# SEG-NEXT: cmdsize
+# SEG-NEXT: segname __PAGEZERO
+
+# SEG: cmd LC_SEGMENT_64
+# SEG-NEXT: cmdsize
+# SEG-NEXT: segname __TEXT
+# SEG-NEXT: vmaddr 0x[[#%x, TEXTSTART:]]
+# SEG-NEXT: vmsize 0x[[#%x, TEXTSIZE:]]
+
+# SEG: cmd LC_SEGMENT_64
+# SEG-NEXT: cmdsize
+# SEG-NEXT: segname __BAZ
+# SEG-NEXT: vmaddr 0x[[#%x, BAZSTART:]]
+# SEG-NEXT: vmsize 0x[[#%x, BAZSIZE:]]
+
+# SEG: cmd LC_SEGMENT_64
+# SEG-NEXT: cmdsize
+# SEG-NEXT: segname __FOO
+# SEG-NEXT: vmaddr 0x[[#%x, FOOSTART:]]
+# SEG-NEXT: vmsize 0x[[#%x, FOOSIZE:]]
+
+# SEG: cmd LC_SEGMENT_64
+# SEG-NEXT: cmdsize
+# SEG-NEXT: segname __UFLAG_SEG
+# SEG-NEXT: vmaddr 0x[[#%x, UFLAGSTART:]]
+# SEG-NEXT: vmsize 0x0000000000000000
+
+# SEG: cmd LC_SEGMENT_64
+# SEG-NEXT: cmdsize
+# SEG-NEXT: segname ASDF
+# SEG-NEXT: vmaddr 0x[[#%x, ASDFSTART:]]
+# SEG-NEXT: vmsize 0x0000000000000000
+
+# SEG: _main
+
+## segment$start$__TEXT / segment$end$__TEXT
+# SEG: [[#%x, PC1:]]:
+# SEG-SAME: leaq [[#%d, TEXTSTART - PC1 - 7]](%rip), %rax
+# SEG-NEXT: [[#%x, PC2:]]:
+# SEG-SAME: leaq [[#%d, TEXTSTART + TEXTSIZE - PC2 - 7]](%rip), %rbx
+
+## segment$start$__FOO / segment$end$__FOO, which is renamed to __BAZ
+# SEG: [[#%x, PC3:]]:
+# SEG-SAME: leaq [[#%d, BAZSTART - PC3 - 7]](%rip), %rax
+# SEG-NEXT: [[#%x, PC4:]]:
+# SEG-SAME: leaq [[#%d, BAZSTART + BAZSIZE - PC4 - 7]](%rip), %rbx
+
+## segment$start$__BAZ / segment$end$__BAZ
+# SEG: [[#%x, PC5:]]:
+# SEG-SAME: leaq [[#%d, BAZSTART - PC5 - 7]](%rip), %rax
+# SEG-NEXT: [[#%x, PC6:]]:
+# SEG-SAME: leaq [[#%d, BAZSTART + BAZSIZE - PC6 - 7]](%rip), %rbx
+
+## segment$start$__WHAT / segment$end$__WHAT, which is renamed to __FOO
+# SEG: [[#%x, PC7:]]:
+# SEG-SAME: leaq [[#%d, FOOSTART - PC7 - 7]](%rip), %rax
+# SEG-NEXT: [[#%x, PC8:]]:
+# SEG-SAME: leaq [[#%d, FOOSTART + FOOSIZE - PC8 - 7]](%rip), %rbx
+
+## segment$start$ASDF / segment$end$ASDF
+# SEG: [[#%x, PC9:]]:
+# SEG-SAME: leaq [[#%d, ASDFSTART - PC9 - 7]](%rip), %rax
+# SEG-NEXT: [[#%x, PC10:]]:
+# SEG-SAME: leaq [[#%d, ASDFSTART - PC10 - 7]](%rip), %rbx
+
#--- order.txt
_otherfun
_main
@@ -272,3 +366,53 @@ section$end$ACTUAL$symbol:
.fill 1
.subsections_via_symbols
+
+#--- seg.s
+## Renamed to __BAZ,__bar by -rename_segment
+.section __FOO,__bar
+.space 1
+
+## Renamed to __FOO,__ever by -rename_segment
+.section __WHAT,__ever
+.space 1
+
+.text
+.globl segment$start$REGULAR
+segment$start$REGULAR:
+ retq
+
+.globl segment$end$REGULAR
+segment$end$REGULAR:
+ retq
+
+.globl _main
+_main:
+ movq segment$start$__TEXT at GOTPCREL(%rip), %rax
+ movq segment$end$__TEXT at GOTPCREL(%rip), %rbx
+
+ movq segment$start$__FOO at GOTPCREL(%rip), %rax
+ movq segment$end$__FOO at GOTPCREL(%rip), %rbx
+
+ movq segment$start$__BAZ at GOTPCREL(%rip), %rax
+ movq segment$end$__BAZ at GOTPCREL(%rip), %rbx
+
+ # (These two lines make ld64 crash linking the .o file.
+ # The rest of the test works with ld64 too.)
+ movq segment$start$__WHAT at GOTPCREL(%rip), %rax
+ movq segment$end$__WHAT at GOTPCREL(%rip), %rbx
+
+ # References to non-existing segments create that segment,
+ # without any sections in it.
+ movq segment$start$ASDF at GOTPCREL(%rip), %rax
+ movq segment$end$ASDF at GOTPCREL(%rip), %rbx
+
+ ## Non-undefined symbols don't create segments.
+ callq segment$start$REGULAR
+ callq segment$end$REGULAR
+
+ ret
+
+.section __TEXT,__fill
+.space 578
+
+.subsections_via_symbols
More information about the llvm-commits
mailing list