obj

package
v0.0.0-...-a027466 Latest Latest
Warning

This package is not in the latest version of its module.

 Go to latest Published: Jun 7, 2017 License: BSD-3-Clause Imports: 13 Imported by: 0

Documentation

Index

Constants

View Source 
const (
	ABase386 = (1 + iota) << 10
	ABaseARM
	ABaseAMD64
	ABasePPC64
	ABaseARM64
	ABaseMIPS
	ABaseS390X

	AllowedOpCodes = 1 << 10            // The number of opcodes available for any given architecture.
	AMask          = AllowedOpCodes - 1 // AND with this to use the opcode as an array index.
)

Each architecture is allotted a distinct subspace of opcode values for declaring its arch-specific opcodes. Within this subspace, the first arch-specific opcode should be at offset A_ARCHSPECIFIC.

Subspaces are aligned to a power of two so opcodes can be masked with AMask and used as compact array indices.

View Source 
const (
	// Don't profile the marked routine.
	//
	// Deprecated: Not implemented, do not use.
	NOPROF = 1

	// It is ok for the linker to get multiple of these symbols. It will
	// pick one of the duplicates to use.
	DUPOK = 2

	// Don't insert stack check preamble.
	NOSPLIT = 4

	// Put this data in a read-only section.
	RODATA = 8

	// This data contains no pointers.
	NOPTR = 16

	// This is a wrapper function and should not count as disabling 'recover'.
	WRAPPER = 32

	// This function uses its incoming context register.
	NEEDCTXT = 64

	// When passed to ggloblsym, causes Local to be set to true on the LSym it creates.
	LOCAL = 128

	// Allocate a word of thread local storage and store the offset from the
	// thread local base to the thread local storage in this variable.
	TLSBSS = 256

	// Do not insert instructions to allocate a stack frame for this function.
	// Only valid on functions that declare a frame size of 0.
	// TODO(mwhudson): only implemented for ppc64x at present.
	NOFRAME = 512

	// Function can call reflect.Type.Method or reflect.Type.MethodByName.
	REFLECTMETHOD = 1024
)
View Source 
const (
	C_SCOND     = (1 << 4) - 1
	C_SBIT      = 1 << 4
	C_PBIT      = 1 << 5
	C_WBIT      = 1 << 6
	C_FBIT      = 1 << 7
	C_UBIT      = 1 << 7
	C_SCOND_XOR = 14
)

ARM scond byte

View Source 
const (
	// Because of masking operations in the encodings, each register
	// space should start at 0 modulo some power of 2.
	RBase386   = 1 * 1024
	RBaseAMD64 = 2 * 1024
	RBaseARM   = 3 * 1024
	RBasePPC64 = 4 * 1024  // range [4k, 8k)
	RBaseARM64 = 8 * 1024  // range [8k, 13k)
	RBaseMIPS  = 13 * 1024 // range [13k, 14k)
	RBaseS390X = 14 * 1024 // range [14k, 15k)
)
View Source 
const (
	LOG = 5
)
View Source 
const REG_NONE = 0

Variables

View Source 
var Anames = []string{
	"XXX",
	"CALL",
	"DUFFCOPY",
	"DUFFZERO",
	"END",
	"FUNCDATA",
	"JMP",
	"NOP",
	"PCDATA",
	"RET",
	"TEXT",
	"UNDEF",
}

Functions

func Bool2int 

func Bool2int(b bool) int

func CConv 

func CConv(s uint8) string

CConv formats ARM condition codes.

func Dconv 

func Dconv(p *Prog, a *Addr) string

func Flushplist 

func Flushplist(ctxt *Link, plist *Plist, newprog ProgAlloc)

func Mconv 

func Mconv(a *Addr) string

func Nopout 

func Nopout(p *Prog)

func Rconv 

func Rconv(reg int) string

func RegisterOpcode 

func RegisterOpcode(lo As, Anames []string)

RegisterOpcode binds a list of instruction names to a given instruction number range.

func RegisterRegister 

func RegisterRegister(lo, hi int, Rconv func(int) string)

RegisterRegister binds a pretty-printer (Rconv) for register numbers to a given register number range. Lo is inclusive, hi exclusive (valid registers are lo through hi-1).

func SortSlice 

func SortSlice(slice interface{}, less func(i, j int) bool)

func WriteObjFile 

func WriteObjFile(ctxt *Link, b *bufio.Writer)

Types

type Addr 

type Addr struct {
	Reg    int16
	Index  int16
	Scale  int16 // Sometimes holds a register.
	Type   AddrType
	Name   AddrName
	Class  int8
	Offset int64
	Sym    *LSym

	// argument value:
	//	for TYPE_SCONST, a string
	//	for TYPE_FCONST, a float64
	//	for TYPE_BRANCH, a *Prog (optional)
	//	for TYPE_TEXTSIZE, an int32 (optional)
	Val interface{}
}

An Addr is an argument to an instruction. The general forms and their encodings are: 

sym±offset(symkind)(reg)(index*scale)
	Memory reference at address &sym(symkind) + offset + reg + index*scale.
	Any of sym(symkind), ±offset, (reg), (index*scale), and *scale can be omitted.
	If (reg) and *scale are both omitted, the resulting expression (index) is parsed as (reg).
	To force a parsing as index*scale, write (index*1).
	Encoding:
		type = TYPE_MEM
		name = symkind (NAME_AUTO, ...) or 0 (NAME_NONE)
		sym = sym
		offset = ±offset
		reg = reg (REG_*)
		index = index (REG_*)
		scale = scale (1, 2, 4, 8)

$<mem>
	Effective address of memory reference <mem>, defined above.
	Encoding: same as memory reference, but type = TYPE_ADDR.

$<±integer value>
	This is a special case of $<mem>, in which only ±offset is present.
	It has a separate type for easy recognition.
	Encoding:
		type = TYPE_CONST
		offset = ±integer value

*<mem>
	Indirect reference through memory reference <mem>, defined above.
	Only used on x86 for CALL/JMP *sym(SB), which calls/jumps to a function
	pointer stored in the data word sym(SB), not a function named sym(SB).
	Encoding: same as above, but type = TYPE_INDIR.

$*$<mem>
	No longer used.
	On machines with actual SB registers, $*$<mem> forced the
	instruction encoding to use a full 32-bit constant, never a
	reference relative to SB.

$<floating point literal>
	Floating point constant value.
	Encoding:
		type = TYPE_FCONST
		val = floating point value

$<string literal, up to 8 chars>
	String literal value (raw bytes used for DATA instruction).
	Encoding:
		type = TYPE_SCONST
		val = string

<register name>
	Any register: integer, floating point, control, segment, and so on.
	If looking for specific register kind, must check type and reg value range.
	Encoding:
		type = TYPE_REG
		reg = reg (REG_*)

x(PC)
	Encoding:
		type = TYPE_BRANCH
		val = Prog* reference OR ELSE offset = target pc (branch takes priority)

$±x-±y
	Final argument to TEXT, specifying local frame size x and argument size y.
	In this form, x and y are integer literals only, not arbitrary expressions.
	This avoids parsing ambiguities due to the use of - as a separator.
	The ± are optional.
	If the final argument to TEXT omits the -±y, the encoding should still
	use TYPE_TEXTSIZE (not TYPE_CONST), with u.argsize = ArgsSizeUnknown.
	Encoding:
		type = TYPE_TEXTSIZE
		offset = x
		val = int32(y)

reg<<shift, reg>>shift, reg->shift, reg@>shift
	Shifted register value, for ARM and ARM64.
	In this form, reg must be a register and shift can be a register or an integer constant.
	Encoding:
		type = TYPE_SHIFT
	On ARM:
		offset = (reg&15) | shifttype<<5 | count
		shifttype = 0, 1, 2, 3 for <<, >>, ->, @>
		count = (reg&15)<<8 | 1<<4 for a register shift count, (n&31)<<7 for an integer constant.
	On ARM64:
		offset = (reg&31)<<16 | shifttype<<22 | (count&63)<<10
		shifttype = 0, 1, 2 for <<, >>, ->

(reg, reg)
	A destination register pair. When used as the last argument of an instruction,
	this form makes clear that both registers are destinations.
	Encoding:
		type = TYPE_REGREG
		reg = first register
		offset = second register

[reg, reg, reg-reg]
	Register list for ARM.
	Encoding:
		type = TYPE_REGLIST
		offset = bit mask of registers in list; R0 is low bit.

reg, reg
	Register pair for ARM.
	TYPE_REGREG2

(reg+reg)
	Register pair for PPC64.
	Encoding:
		type = TYPE_MEM
		reg = first register
		index = second register
		scale = 1

type AddrName 

type AddrName int8
const (
	NAME_NONE AddrName = iota
	NAME_EXTERN
	NAME_STATIC
	NAME_AUTO
	NAME_PARAM
	// A reference to name@GOT(SB) is a reference to the entry in the global offset
	// table for 'name'.
	NAME_GOTREF
)

type AddrType 

type AddrType uint8
const (
	TYPE_NONE AddrType = iota
	TYPE_BRANCH
	TYPE_TEXTSIZE
	TYPE_MEM
	TYPE_CONST
	TYPE_FCONST
	TYPE_SCONST
	TYPE_REG
	TYPE_ADDR
	TYPE_SHIFT
	TYPE_REGREG
	TYPE_REGREG2
	TYPE_INDIR
	TYPE_REGLIST
)

func (AddrType) String 

func (i AddrType) String() string

type As 

type As int16

An As denotes an assembler opcode. There are some portable opcodes, declared here in package obj, that are common to all architectures. However, the majority of opcodes are arch-specific and are declared in their respective architecture's subpackage. 

const (
	AXXX As = iota
	ACALL
	ADUFFCOPY
	ADUFFZERO
	AEND
	AFUNCDATA
	AJMP
	ANOP
	APCDATA
	ARET
	ATEXT
	AUNDEF
	A_ARCHSPECIFIC
)

These are the portable opcodes.

func (As) String 

func (a As) String() string

type Attribute 

type Attribute int16

Attribute is a set of symbol attributes. 

const (
	AttrDuplicateOK Attribute = 1 << iota
	AttrCFunc
	AttrNoSplit
	AttrLeaf
	AttrWrapper
	AttrNeedCtxt
	AttrNoFrame
	AttrSeenGlobl
	AttrOnList
	AttrStatic

	// MakeTypelink means that the type should have an entry in the typelink table.
	AttrMakeTypelink

	// ReflectMethod means the function may call reflect.Type.Method or
	// reflect.Type.MethodByName. Matching is imprecise (as reflect.Type
	// can be used through a custom interface), so ReflectMethod may be
	// set in some cases when the reflect package is not called.
	//
	// Used by the linker to determine what methods can be pruned.
	AttrReflectMethod

	// Local means make the symbol local even when compiling Go code to reference Go
	// symbols in other shared libraries, as in this mode symbols are global by
	// default. "local" here means in the sense of the dynamic linker, i.e. not
	// visible outside of the module (shared library or executable) that contains its
	// definition. (When not compiling to support Go shared libraries, all symbols are
	// local in this sense unless there is a cgo_export_* directive).
	AttrLocal
)

func (Attribute) CFunc 

func (a Attribute) CFunc() bool

func (Attribute) DuplicateOK 

func (a Attribute) DuplicateOK() bool

func (Attribute) Leaf 

func (a Attribute) Leaf() bool

func (Attribute) Local 

func (a Attribute) Local() bool
func (a Attribute) MakeTypelink() bool

func (Attribute) NeedCtxt 

func (a Attribute) NeedCtxt() bool

func (Attribute) NoFrame 

func (a Attribute) NoFrame() bool

func (Attribute) NoSplit 

func (a Attribute) NoSplit() bool

func (Attribute) OnList 

func (a Attribute) OnList() bool

func (Attribute) ReflectMethod 

func (a Attribute) ReflectMethod() bool

func (Attribute) SeenGlobl 

func (a Attribute) SeenGlobl() bool

func (*Attribute) Set 

func (a *Attribute) Set(flag Attribute, value bool)

func (Attribute) Static 

func (a Attribute) Static() bool

func (Attribute) TextAttrString 

func (a Attribute) TextAttrString() string

TextAttrString formats a for printing in as part of a TEXT prog.

func (Attribute) Wrapper 

func (a Attribute) Wrapper() bool

type Auto 

type Auto struct {
	Asym    *LSym
	Aoffset int32
	Name    AddrName
	Gotype  *LSym
}

type FuncInfo 

type FuncInfo struct {
	Args   int32
	Locals int32
	Text   *Prog
	Autom  []*Auto
	Pcln   Pcln

	GCArgs   LSym
	GCLocals LSym
	// contains filtered or unexported fields
}

A FuncInfo contains extra fields for STEXT symbols.

type InlTree 

type InlTree struct {
	// contains filtered or unexported fields
}

InlTree s a collection of inlined calls. The Parent field of an InlinedCall is the index of another InlinedCall in InlTree.

The compiler maintains a global inlining tree and adds a node to it every time a function is inlined. For example, suppose f() calls g() and g has two calls to h(), and that f, g, and h are inlineable: 

1 func main() {
2     f()
3 }
4 func f() {
5     g()
6 }
7 func g() {
8     h()
9     h()

10 }

Assuming the global tree starts empty, inlining will produce the following tree: 

[]InlinedCall{
  {Parent: -1, Func: "f", Pos: <line 2>},
  {Parent:  0, Func: "g", Pos: <line 5>},
  {Parent:  1, Func: "h", Pos: <line 8>},
  {Parent:  1, Func: "h", Pos: <line 9>},
}

The nodes of h inlined into main will have inlining indexes 2 and 3.

Eventually, the compiler extracts a per-function inlining tree from the global inlining tree (see pcln.go).

func (*InlTree) Add 

func (tree *InlTree) Add(parent int, pos src.XPos, func_ *LSym) int

Add adds a new call to the tree, returning its index.

type InlinedCall 

type InlinedCall struct {
	Parent int      // index of the parent in the InlTree or < 0 if outermost call
	Pos    src.XPos // position of the inlined call
	Func   *LSym    // function that was inlined
}

InlinedCall is a node in an InlTree.

type LSym 

type LSym struct {
	Name string
	Type objabi.SymKind
	Attribute

	RefIdx int // Index of this symbol in the symbol reference list.
	Size   int64
	Gotype *LSym
	P      []byte
	R      []Reloc

	Func *FuncInfo
}

An LSym is the sort of symbol that is written to an object file.

func (*LSym) Grow 

func (s *LSym) Grow(lsiz int64)

Grow increases the length of s.P to lsiz.

func (*LSym) GrowCap 

func (s *LSym) GrowCap(c int64)

GrowCap increases the capacity of s.P to c.

func (*LSym) Len 

func (s *LSym) Len() int64

func (*LSym) String 

func (s *LSym) String() string

The compiler needs LSym to satisfy fmt.Stringer, because it stores an LSym in ssa.ExternSymbol.

func (*LSym) WriteAddr 

func (s *LSym) WriteAddr(ctxt *Link, off int64, siz int, rsym *LSym, roff int64)

WriteAddr writes an address of size siz into s at offset off. rsym and roff specify the relocation for the address.

func (*LSym) WriteBytes 

func (s *LSym) WriteBytes(ctxt *Link, off int64, b []byte) int64

WriteBytes writes a slice of bytes into s at offset off.

func (*LSym) WriteFloat32 

func (s *LSym) WriteFloat32(ctxt *Link, off int64, f float32)

WriteFloat32 writes f into s at offset off.

func (*LSym) WriteFloat64 

func (s *LSym) WriteFloat64(ctxt *Link, off int64, f float64)

WriteFloat64 writes f into s at offset off.

func (*LSym) WriteInt 

func (s *LSym) WriteInt(ctxt *Link, off int64, siz int, i int64)

WriteInt writes an integer i of size siz into s at offset off.

func (*LSym) WriteOff 

func (s *LSym) WriteOff(ctxt *Link, off int64, rsym *LSym, roff int64)

WriteOff writes a 4 byte offset to rsym+roff into s at offset off. After linking the 4 bytes stored at s+off will be rsym+roff-(start of section that s is in).

func (*LSym) WriteString 

func (s *LSym) WriteString(ctxt *Link, off int64, siz int, str string)

WriteString writes a string of size siz into s at offset off.

func (*LSym) WriteWeakOff 

func (s *LSym) WriteWeakOff(ctxt *Link, off int64, rsym *LSym, roff int64)

WriteWeakOff writes a weak 4 byte offset to rsym+roff into s at offset off. After linking the 4 bytes stored at s+off will be rsym+roff-(start of section that s is in). 

type Link struct {
	Headtype      objabi.HeadType
	Arch          *LinkArch
	Debugasm      bool
	Debugvlog     bool
	Debugpcln     string
	Flag_shared   bool
	Flag_dynlink  bool
	Flag_optimize bool
	Bso           *bufio.Writer
	Pathname      string

	PosTable  src.PosTable
	InlTree   InlTree // global inlining tree used by gc/inl.go
	Imports   []string
	DiagFunc  func(string, ...interface{})
	DebugInfo func(fn *LSym, curfn interface{}) []dwarf.Scope // if non-nil, curfn is a *gc.Node
	Errors    int

	Framepointer_enabled bool

	// state for writing objects
	Text []*LSym
	Data []*LSym
	// contains filtered or unexported fields
}

Link holds the context for writing object code from a compiler to be linker input or for reading that input into the linker.

func Linknew 

func Linknew(arch *LinkArch) *Link

func (*Link) AddImport 

func (ctxt *Link) AddImport(pkg string)

AddImport adds a package to the list of imported packages.

func (*Link) CanReuseProgs 

func (ctxt *Link) CanReuseProgs() bool

func (*Link) Dconv 

func (ctxt *Link) Dconv(a *Addr) string

func (*Link) Diag 

func (ctxt *Link) Diag(format string, args ...interface{})

func (*Link) FixedFrameSize 

func (ctxt *Link) FixedFrameSize() int64

The smallest possible offset from the hardware stack pointer to a local variable on the stack. Architectures that use a link register save its value on the stack in the function prologue and so always have a pointer between the hardware stack pointer and the local variable area.

func (*Link) Float32Sym 

func (ctxt *Link) Float32Sym(f float32) *LSym

func (*Link) Float64Sym 

func (ctxt *Link) Float64Sym(f float64) *LSym

func (*Link) Globl 

func (ctxt *Link) Globl(s *LSym, size int64, flag int)

func (*Link) InitTextSym 

func (ctxt *Link) InitTextSym(s *LSym, flag int)

func (*Link) Int64Sym 

func (ctxt *Link) Int64Sym(i int64) *LSym

func (*Link) Logf 

func (ctxt *Link) Logf(format string, args ...interface{})

func (*Link) Lookup 

func (ctxt *Link) Lookup(name string) *LSym

Lookup looks up the symbol with name name. If it does not exist, it creates it.

func (*Link) LookupDerived 

func (ctxt *Link) LookupDerived(s *LSym, name string) *LSym

LookupDerived looks up or creates the symbol with name name derived from symbol s. The resulting symbol will be static iff s is.

func (*Link) LookupInit 

func (ctxt *Link) LookupInit(name string, init func(s *LSym)) *LSym

LookupInit looks up the symbol with name name. If it does not exist, it creates it and passes it to init for one-time initialization.

func (*Link) LookupStatic 

func (ctxt *Link) LookupStatic(name string) *LSym

LookupStatic looks up the static symbol with name name. If it does not exist, it creates it.

func (*Link) NewProg 

func (ctxt *Link) NewProg() *Prog

func (*Link) OutermostPos 

func (ctxt *Link) OutermostPos(xpos src.XPos) src.Pos

OutermostPos returns the outermost position corresponding to xpos, which is where xpos was ultimately inlined to. In the example for InlTree, main() contains inlined AST nodes from h(), but the outermost position for those nodes is line 2.

type LinkArch 

type LinkArch struct {
	*sys.Arch
	Init       func(*Link)
	Preprocess func(*Link, *LSym, ProgAlloc)
	Assemble   func(*Link, *LSym, ProgAlloc)
	Progedit   func(*Link, *Prog, ProgAlloc)
	UnaryDst   map[As]bool // Instruction takes one operand, a destination.
}

LinkArch is the definition of a single architecture.

type Pcdata 

type Pcdata struct {
	P []byte
}

type Pcln 

type Pcln struct {
	Pcsp        Pcdata
	Pcfile      Pcdata
	Pcline      Pcdata
	Pcinline    Pcdata
	Pcdata      []Pcdata
	Funcdata    []*LSym
	Funcdataoff []int64
	File        []string
	Lastfile    string
	Lastindex   int
	InlTree     InlTree // per-function inlining tree extracted from the global tree
}

type Plist 

type Plist struct {
	Firstpc *Prog
	Curfn   interface{} // holds a *gc.Node, if non-nil
}

type Prog 

type Prog struct {
	Ctxt   *Link    // linker context
	Link   *Prog    // next Prog in linked list
	From   Addr     // first source operand
	From3  *Addr    // third source operand (second is Reg below)
	To     Addr     // destination operand (second is RegTo2 below)
	Pcond  *Prog    // target of conditional jump
	Forwd  *Prog    // for x86 back end
	Rel    *Prog    // for x86, arm back ends
	Pc     int64    // for back ends or assembler: virtual or actual program counter, depending on phase
	Pos    src.XPos // source position of this instruction
	Spadj  int32    // effect of instruction on stack pointer (increment or decrement amount)
	As     As       // assembler opcode
	Reg    int16    // 2nd source operand
	RegTo2 int16    // 2nd destination operand
	Mark   uint16   // bitmask of arch-specific items
	Optab  uint16   // arch-specific opcode index
	Scond  uint8    // condition bits for conditional instruction (e.g., on ARM)
	Back   uint8    // for x86 back end: backwards branch state
	Ft     uint8    // for x86 back end: type index of Prog.From
	Tt     uint8    // for x86 back end: type index of Prog.To
	Isize  uint8    // for x86 back end: size of the instruction in bytes
}

Prog describes a single machine instruction.

The general instruction form is: 

As.Scond From, Reg, From3, To, RegTo2

where As is an opcode and the others are arguments: From, Reg, From3 are sources, and To, RegTo2 are destinations. Usually, not all arguments are present. For example, MOVL R1, R2 encodes using only As=MOVL, From=R1, To=R2. The Scond field holds additional condition bits for systems (like arm) that have generalized conditional execution.

Jump instructions use the Pcond field to point to the target instruction, which must be in the same linked list as the jump instruction.

The Progs for a given function are arranged in a list linked through the Link field.

Each Prog is charged to a specific source line in the debug information, specified by Pos.Line(). Every Prog has a Ctxt field that defines its context. For performance reasons, Progs usually are usually bulk allocated, cached, and reused; those bulk allocators should always be used, rather than new(Prog).

The other fields not yet mentioned are for use by the back ends and should be left zeroed by creators of Prog lists.

func Appendp 

func Appendp(q *Prog, newprog ProgAlloc) *Prog

func (*Prog) From3Type 

func (p *Prog) From3Type() AddrType

From3Type returns From3.Type, or TYPE_NONE when From3 is nil.

func (*Prog) Line 

func (p *Prog) Line() string

func (*Prog) String 

func (p *Prog) String() string

type ProgAlloc 

type ProgAlloc func() *Prog

ProgAlloc is a function that allocates Progs. It is used to provide access to cached/bulk-allocated Progs to the assemblers.

type Reloc 

type Reloc struct {
	Off  int32
	Siz  uint8
	Type objabi.RelocType
	Add  int64
	Sym  *LSym
}

func Addrel 

func Addrel(s *LSym) *Reloc

Source Files

View all Source files

Directories

Path Synopsis

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL
go.dev uses cookies from Google to deliver and enhance the quality of its services and to analyze traffic. Learn more.