llvm-project/lld/COFF/Chunks.h

376 lines
12 KiB
C
Raw Normal View History

//===- Chunks.h -------------------------------------------------*- C++ -*-===//
//
// The LLVM Linker
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
#ifndef LLD_COFF_CHUNKS_H
#define LLD_COFF_CHUNKS_H
#include "Config.h"
[opt] Replace the recursive walk for GC with a worklist algorithm. This flattens the entire liveness walk from a recursive mark approach to a worklist approach. It also sinks the worklist management completely out of the SectionChunk and into the Writer by exposing the ability to iterato over children of a chunk and over the symbol bodies of relocated symbols. I'm not 100% happy with the API names, so suggestions welcome there. This allows us to use a single worklist for the entire recursive walk and would also be a natural place to take advantage of parallelism at some future point. With this, we completely inline away the GC walk into the Writer::markLive function and it makes it very easy to profile what is slow. Currently, time is being wasted checking whether a Chunk isa SectionChunk (it essentially always is), finding (or skipping) a replacement for a symbol, and chasing pointers between symbols and their chunks. There are a bunch of things we can do to fix this, and its easier to do them after this change IMO. This change alone saves 1-2% of the time for my self-link of lld.exe (which I'm running and benchmarking on Linux ironically). Perhaps more notably, we'll no longer blow out the stack for large links. =] Just as an FYI, at this point, I/O is starting to really dominate the profile. Well over 10% of the time appears to be inside the kernel doing page table silliness. I think a decent chunk of this can be nuked as well, but it's a little odd as cross-linking in this way isn't really the primary goal here. Differential Revision: http://reviews.llvm.org/D10790 llvm-svn: 240995
2015-06-30 05:12:49 +08:00
#include "InputFiles.h"
#include "lld/Core/LLVM.h"
#include "llvm/ADT/ArrayRef.h"
[opt] Replace the recursive walk for GC with a worklist algorithm. This flattens the entire liveness walk from a recursive mark approach to a worklist approach. It also sinks the worklist management completely out of the SectionChunk and into the Writer by exposing the ability to iterato over children of a chunk and over the symbol bodies of relocated symbols. I'm not 100% happy with the API names, so suggestions welcome there. This allows us to use a single worklist for the entire recursive walk and would also be a natural place to take advantage of parallelism at some future point. With this, we completely inline away the GC walk into the Writer::markLive function and it makes it very easy to profile what is slow. Currently, time is being wasted checking whether a Chunk isa SectionChunk (it essentially always is), finding (or skipping) a replacement for a symbol, and chasing pointers between symbols and their chunks. There are a bunch of things we can do to fix this, and its easier to do them after this change IMO. This change alone saves 1-2% of the time for my self-link of lld.exe (which I'm running and benchmarking on Linux ironically). Perhaps more notably, we'll no longer blow out the stack for large links. =] Just as an FYI, at this point, I/O is starting to really dominate the profile. Well over 10% of the time appears to be inside the kernel doing page table silliness. I think a decent chunk of this can be nuked as well, but it's a little odd as cross-linking in this way isn't really the primary goal here. Differential Revision: http://reviews.llvm.org/D10790 llvm-svn: 240995
2015-06-30 05:12:49 +08:00
#include "llvm/ADT/iterator.h"
#include "llvm/ADT/iterator_range.h"
#include "llvm/Object/COFF.h"
#include <utility>
#include <vector>
namespace lld {
namespace coff {
using llvm::COFF::ImportDirectoryTableEntry;
using llvm::object::COFFSymbolRef;
using llvm::object::SectionRef;
using llvm::object::coff_relocation;
using llvm::object::coff_section;
class Baserel;
class Defined;
class DefinedImportData;
class DefinedRegular;
class ObjectFile;
class OutputSection;
[opt] Replace the recursive walk for GC with a worklist algorithm. This flattens the entire liveness walk from a recursive mark approach to a worklist approach. It also sinks the worklist management completely out of the SectionChunk and into the Writer by exposing the ability to iterato over children of a chunk and over the symbol bodies of relocated symbols. I'm not 100% happy with the API names, so suggestions welcome there. This allows us to use a single worklist for the entire recursive walk and would also be a natural place to take advantage of parallelism at some future point. With this, we completely inline away the GC walk into the Writer::markLive function and it makes it very easy to profile what is slow. Currently, time is being wasted checking whether a Chunk isa SectionChunk (it essentially always is), finding (or skipping) a replacement for a symbol, and chasing pointers between symbols and their chunks. There are a bunch of things we can do to fix this, and its easier to do them after this change IMO. This change alone saves 1-2% of the time for my self-link of lld.exe (which I'm running and benchmarking on Linux ironically). Perhaps more notably, we'll no longer blow out the stack for large links. =] Just as an FYI, at this point, I/O is starting to really dominate the profile. Well over 10% of the time appears to be inside the kernel doing page table silliness. I think a decent chunk of this can be nuked as well, but it's a little odd as cross-linking in this way isn't really the primary goal here. Differential Revision: http://reviews.llvm.org/D10790 llvm-svn: 240995
2015-06-30 05:12:49 +08:00
class SymbolBody;
// Mask for section types (code, data, bss, disacardable, etc.)
// and permissions (writable, readable or executable).
const uint32_t PermMask = 0xFF0000F0;
// A Chunk represents a chunk of data that will occupy space in the
// output (if the resolver chose that). It may or may not be backed by
// a section of an input file. It could be linker-created data, or
// doesn't even have actual data (if common or bss).
class Chunk {
public:
enum Kind { SectionKind, OtherKind };
Kind kind() const { return ChunkKind; }
virtual ~Chunk() = default;
// Returns the size of this chunk (even if this is a common or BSS.)
virtual size_t getSize() const = 0;
// Write this chunk to a mmap'ed file, assuming Buf is pointing to
// beginning of the file. Because this function may use RVA values
// of other chunks for relocations, you need to set them properly
// before calling this function.
virtual void writeTo(uint8_t *Buf) const {}
// The writer sets and uses the addresses.
uint64_t getRVA() const { return RVA; }
uint32_t getAlign() const { return Align; }
void setRVA(uint64_t V) { RVA = V; }
// Returns true if this has non-zero data. BSS chunks return
// false. If false is returned, the space occupied by this chunk
// will be filled with zeros.
virtual bool hasData() const { return true; }
// Returns readable/writable/executable bits.
virtual uint32_t getPermissions() const { return 0; }
// Returns the section name if this is a section chunk.
// It is illegal to call this function on non-section chunks.
virtual StringRef getSectionName() const {
llvm_unreachable("unimplemented getSectionName");
}
// An output section has pointers to chunks in the section, and each
// chunk has a back pointer to an output section.
void setOutputSection(OutputSection *O) { Out = O; }
OutputSection *getOutputSection() { return Out; }
// Windows-specific.
// Collect all locations that contain absolute addresses for base relocations.
virtual void getBaserels(std::vector<Baserel> *Res) {}
// Returns a human-readable name of this chunk. Chunks are unnamed chunks of
// bytes, so this is used only for logging or debugging.
virtual StringRef getDebugName() { return ""; }
protected:
Chunk(Kind K = OtherKind) : ChunkKind(K) {}
const Kind ChunkKind;
// The alignment of this chunk. The writer uses the value.
uint32_t Align = 1;
// The RVA of this chunk in the output. The writer sets a value.
uint64_t RVA = 0;
public:
// The offset from beginning of the output section. The writer sets a value.
uint64_t OutputSectionOff = 0;
protected:
// The output section for this chunk.
OutputSection *Out = nullptr;
};
// A chunk corresponding a section of an input file.
class SectionChunk final : public Chunk {
// Identical COMDAT Folding feature accesses section internal data.
friend class ICF;
public:
[opt] Replace the recursive walk for GC with a worklist algorithm. This flattens the entire liveness walk from a recursive mark approach to a worklist approach. It also sinks the worklist management completely out of the SectionChunk and into the Writer by exposing the ability to iterato over children of a chunk and over the symbol bodies of relocated symbols. I'm not 100% happy with the API names, so suggestions welcome there. This allows us to use a single worklist for the entire recursive walk and would also be a natural place to take advantage of parallelism at some future point. With this, we completely inline away the GC walk into the Writer::markLive function and it makes it very easy to profile what is slow. Currently, time is being wasted checking whether a Chunk isa SectionChunk (it essentially always is), finding (or skipping) a replacement for a symbol, and chasing pointers between symbols and their chunks. There are a bunch of things we can do to fix this, and its easier to do them after this change IMO. This change alone saves 1-2% of the time for my self-link of lld.exe (which I'm running and benchmarking on Linux ironically). Perhaps more notably, we'll no longer blow out the stack for large links. =] Just as an FYI, at this point, I/O is starting to really dominate the profile. Well over 10% of the time appears to be inside the kernel doing page table silliness. I think a decent chunk of this can be nuked as well, but it's a little odd as cross-linking in this way isn't really the primary goal here. Differential Revision: http://reviews.llvm.org/D10790 llvm-svn: 240995
2015-06-30 05:12:49 +08:00
class symbol_iterator : public llvm::iterator_adaptor_base<
symbol_iterator, const coff_relocation *,
std::random_access_iterator_tag, SymbolBody *> {
friend SectionChunk;
ObjectFile *File;
symbol_iterator(ObjectFile *File, const coff_relocation *I)
: symbol_iterator::iterator_adaptor_base(I), File(File) {}
public:
symbol_iterator() = default;
SymbolBody *operator*() const {
return File->getSymbolBody(I->SymbolTableIndex);
}
};
SectionChunk(ObjectFile *File, const coff_section *Header);
static bool classof(const Chunk *C) { return C->kind() == SectionKind; }
size_t getSize() const override { return Header->SizeOfRawData; }
ArrayRef<uint8_t> getContents() const;
void writeTo(uint8_t *Buf) const override;
bool hasData() const override;
uint32_t getPermissions() const override;
StringRef getSectionName() const override { return SectionName; }
void getBaserels(std::vector<Baserel> *Res) override;
bool isCOMDAT() const;
void applyRelX64(uint8_t *Off, uint16_t Type, OutputSection *OS, uint64_t S,
uint64_t P) const;
void applyRelX86(uint8_t *Off, uint16_t Type, OutputSection *OS, uint64_t S,
uint64_t P) const;
void applyRelARM(uint8_t *Off, uint16_t Type, OutputSection *OS, uint64_t S,
uint64_t P) const;
void applyRelARM64(uint8_t *Off, uint16_t Type, OutputSection *OS, uint64_t S,
uint64_t P) const;
// Called if the garbage collector decides to not include this chunk
// in a final output. It's supposed to print out a log message to stdout.
void printDiscardedMessage() const;
// Adds COMDAT associative sections to this COMDAT section. A chunk
// and its children are treated as a group by the garbage collector.
void addAssociative(SectionChunk *Child);
StringRef getDebugName() override;
void setSymbol(DefinedRegular *S) { if (!Sym) Sym = S; }
// Returns true if the chunk was not dropped by GC or COMDAT deduplication.
bool isLive() { return Live && !Discarded; }
// Used by the garbage collector.
[opt] Replace the recursive walk for GC with a worklist algorithm. This flattens the entire liveness walk from a recursive mark approach to a worklist approach. It also sinks the worklist management completely out of the SectionChunk and into the Writer by exposing the ability to iterato over children of a chunk and over the symbol bodies of relocated symbols. I'm not 100% happy with the API names, so suggestions welcome there. This allows us to use a single worklist for the entire recursive walk and would also be a natural place to take advantage of parallelism at some future point. With this, we completely inline away the GC walk into the Writer::markLive function and it makes it very easy to profile what is slow. Currently, time is being wasted checking whether a Chunk isa SectionChunk (it essentially always is), finding (or skipping) a replacement for a symbol, and chasing pointers between symbols and their chunks. There are a bunch of things we can do to fix this, and its easier to do them after this change IMO. This change alone saves 1-2% of the time for my self-link of lld.exe (which I'm running and benchmarking on Linux ironically). Perhaps more notably, we'll no longer blow out the stack for large links. =] Just as an FYI, at this point, I/O is starting to really dominate the profile. Well over 10% of the time appears to be inside the kernel doing page table silliness. I think a decent chunk of this can be nuked as well, but it's a little odd as cross-linking in this way isn't really the primary goal here. Differential Revision: http://reviews.llvm.org/D10790 llvm-svn: 240995
2015-06-30 05:12:49 +08:00
void markLive() {
assert(Config->DoGC && "should only mark things live from GC");
assert(!isLive() && "Cannot mark an already live section!");
[opt] Replace the recursive walk for GC with a worklist algorithm. This flattens the entire liveness walk from a recursive mark approach to a worklist approach. It also sinks the worklist management completely out of the SectionChunk and into the Writer by exposing the ability to iterato over children of a chunk and over the symbol bodies of relocated symbols. I'm not 100% happy with the API names, so suggestions welcome there. This allows us to use a single worklist for the entire recursive walk and would also be a natural place to take advantage of parallelism at some future point. With this, we completely inline away the GC walk into the Writer::markLive function and it makes it very easy to profile what is slow. Currently, time is being wasted checking whether a Chunk isa SectionChunk (it essentially always is), finding (or skipping) a replacement for a symbol, and chasing pointers between symbols and their chunks. There are a bunch of things we can do to fix this, and its easier to do them after this change IMO. This change alone saves 1-2% of the time for my self-link of lld.exe (which I'm running and benchmarking on Linux ironically). Perhaps more notably, we'll no longer blow out the stack for large links. =] Just as an FYI, at this point, I/O is starting to really dominate the profile. Well over 10% of the time appears to be inside the kernel doing page table silliness. I think a decent chunk of this can be nuked as well, but it's a little odd as cross-linking in this way isn't really the primary goal here. Differential Revision: http://reviews.llvm.org/D10790 llvm-svn: 240995
2015-06-30 05:12:49 +08:00
Live = true;
}
// Returns true if this chunk was dropped by COMDAT deduplication.
bool isDiscarded() const { return Discarded; }
// Used by the SymbolTable when discarding unused comdat sections. This is
// redundant when GC is enabled, as all comdat sections will start out dead.
void markDiscarded() { Discarded = true; }
// True if this is a codeview debug info chunk. These will not be laid out in
// the image. Instead they will end up in the PDB, if one is requested.
bool isCodeView() const {
return SectionName == ".debug" || SectionName.startswith(".debug$");
}
// True if this is a DWARF debug info chunk.
bool isDWARF() const { return SectionName.startswith(".debug_"); }
[opt] Replace the recursive walk for GC with a worklist algorithm. This flattens the entire liveness walk from a recursive mark approach to a worklist approach. It also sinks the worklist management completely out of the SectionChunk and into the Writer by exposing the ability to iterato over children of a chunk and over the symbol bodies of relocated symbols. I'm not 100% happy with the API names, so suggestions welcome there. This allows us to use a single worklist for the entire recursive walk and would also be a natural place to take advantage of parallelism at some future point. With this, we completely inline away the GC walk into the Writer::markLive function and it makes it very easy to profile what is slow. Currently, time is being wasted checking whether a Chunk isa SectionChunk (it essentially always is), finding (or skipping) a replacement for a symbol, and chasing pointers between symbols and their chunks. There are a bunch of things we can do to fix this, and its easier to do them after this change IMO. This change alone saves 1-2% of the time for my self-link of lld.exe (which I'm running and benchmarking on Linux ironically). Perhaps more notably, we'll no longer blow out the stack for large links. =] Just as an FYI, at this point, I/O is starting to really dominate the profile. Well over 10% of the time appears to be inside the kernel doing page table silliness. I think a decent chunk of this can be nuked as well, but it's a little odd as cross-linking in this way isn't really the primary goal here. Differential Revision: http://reviews.llvm.org/D10790 llvm-svn: 240995
2015-06-30 05:12:49 +08:00
// Allow iteration over the bodies of this chunk's relocated symbols.
llvm::iterator_range<symbol_iterator> symbols() const {
return llvm::make_range(symbol_iterator(File, Relocs.begin()),
symbol_iterator(File, Relocs.end()));
}
// Allow iteration over the associated child chunks for this section.
ArrayRef<SectionChunk *> children() const { return AssocChildren; }
// A pointer pointing to a replacement for this chunk.
// Initially it points to "this" object. If this chunk is merged
// with other chunk by ICF, it points to another chunk,
// and this chunk is considrered as dead.
SectionChunk *Repl;
// The CRC of the contents as described in the COFF spec 4.5.5.
// Auxiliary Format 5: Section Definitions. Used for ICF.
uint32_t Checksum = 0;
const coff_section *Header;
// The file that this chunk was created from.
ObjectFile *File;
private:
StringRef SectionName;
std::vector<SectionChunk *> AssocChildren;
llvm::iterator_range<const coff_relocation *> Relocs;
size_t NumRelocs;
// True if this chunk was discarded because it was a duplicate comdat section.
bool Discarded;
// Used by the garbage collector.
bool Live;
// Used for ICF (Identical COMDAT Folding)
void replace(SectionChunk *Other);
uint32_t Class[2] = {0, 0};
// Sym points to a section symbol if this is a COMDAT chunk.
DefinedRegular *Sym = nullptr;
};
// A chunk for common symbols. Common chunks don't have actual data.
class CommonChunk : public Chunk {
public:
CommonChunk(const COFFSymbolRef Sym);
size_t getSize() const override { return Sym.getValue(); }
bool hasData() const override { return false; }
uint32_t getPermissions() const override;
StringRef getSectionName() const override { return ".bss"; }
private:
const COFFSymbolRef Sym;
};
// A chunk for linker-created strings.
class StringChunk : public Chunk {
public:
explicit StringChunk(StringRef S) : Str(S) {}
size_t getSize() const override { return Str.size() + 1; }
void writeTo(uint8_t *Buf) const override;
private:
StringRef Str;
};
static const uint8_t ImportThunkX86[] = {
0xff, 0x25, 0x00, 0x00, 0x00, 0x00, // JMP *0x0
};
static const uint8_t ImportThunkARM[] = {
0x40, 0xf2, 0x00, 0x0c, // mov.w ip, #0
0xc0, 0xf2, 0x00, 0x0c, // mov.t ip, #0
0xdc, 0xf8, 0x00, 0xf0, // ldr.w pc, [ip]
};
static const uint8_t ImportThunkARM64[] = {
0x10, 0x00, 0x00, 0x90, // adrp x16, #0
0x10, 0x02, 0x40, 0xf9, // ldr x16, [x16]
0x00, 0x02, 0x1f, 0xd6, // br x16
};
// Windows-specific.
// A chunk for DLL import jump table entry. In a final output, it's
// contents will be a JMP instruction to some __imp_ symbol.
class ImportThunkChunkX64 : public Chunk {
public:
explicit ImportThunkChunkX64(Defined *S);
size_t getSize() const override { return sizeof(ImportThunkX86); }
void writeTo(uint8_t *Buf) const override;
private:
Defined *ImpSymbol;
};
class ImportThunkChunkX86 : public Chunk {
public:
explicit ImportThunkChunkX86(Defined *S) : ImpSymbol(S) {}
size_t getSize() const override { return sizeof(ImportThunkX86); }
void getBaserels(std::vector<Baserel> *Res) override;
void writeTo(uint8_t *Buf) const override;
private:
Defined *ImpSymbol;
};
class ImportThunkChunkARM : public Chunk {
public:
explicit ImportThunkChunkARM(Defined *S) : ImpSymbol(S) {}
size_t getSize() const override { return sizeof(ImportThunkARM); }
void getBaserels(std::vector<Baserel> *Res) override;
void writeTo(uint8_t *Buf) const override;
private:
Defined *ImpSymbol;
};
class ImportThunkChunkARM64 : public Chunk {
public:
explicit ImportThunkChunkARM64(Defined *S) : ImpSymbol(S) {}
size_t getSize() const override { return sizeof(ImportThunkARM64); }
void writeTo(uint8_t *Buf) const override;
private:
Defined *ImpSymbol;
};
// Windows-specific.
// See comments for DefinedLocalImport class.
class LocalImportChunk : public Chunk {
public:
explicit LocalImportChunk(Defined *S) : Sym(S) {}
size_t getSize() const override;
void getBaserels(std::vector<Baserel> *Res) override;
void writeTo(uint8_t *Buf) const override;
private:
Defined *Sym;
};
// Windows-specific.
// A chunk for SEH table which contains RVAs of safe exception handler
// functions. x86-only.
class SEHTableChunk : public Chunk {
public:
explicit SEHTableChunk(std::set<Defined *> S) : Syms(std::move(S)) {}
size_t getSize() const override { return Syms.size() * 4; }
void writeTo(uint8_t *Buf) const override;
private:
std::set<Defined *> Syms;
};
// Windows-specific.
// This class represents a block in .reloc section.
// See the PE/COFF spec 5.6 for details.
class BaserelChunk : public Chunk {
public:
BaserelChunk(uint32_t Page, Baserel *Begin, Baserel *End);
size_t getSize() const override { return Data.size(); }
void writeTo(uint8_t *Buf) const override;
private:
std::vector<uint8_t> Data;
};
class Baserel {
public:
Baserel(uint32_t V, uint8_t Ty) : RVA(V), Type(Ty) {}
explicit Baserel(uint32_t V) : Baserel(V, getDefaultType()) {}
uint8_t getDefaultType();
uint32_t RVA;
uint8_t Type;
};
} // namespace coff
} // namespace lld
#endif