blob: 03dc9fb88bc769157b2cd4414e843f0a607052b9 [file] [log] [blame]
// Copyright 2013 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// NOTE: There are *three* independent implementations of this object
// file format in the Go source tree:
//
// - cmd/internal/goobj/read.go (used by cmd/addr2line, cmd/nm, cmd/objdump, cmd/pprof)
// - cmd/internal/obj/objfile.go (used by cmd/asm and cmd/compile)
// - cmd/link/internal/objfile.go (used by cmd/link)
//
// When changing the object file format, remember to change all three.
// Originally, Go object files were Plan 9 object files, but no longer.
// Now they are more like standard object files, in that each symbol is defined
// by an associated memory image (bytes) and a list of relocations to apply
// during linking. We do not (yet?) use a standard file format, however.
// For now, the format is chosen to be as simple as possible to read and write.
// It may change for reasons of efficiency, or we may even switch to a
// standard file format if there are compelling benefits to doing so.
// See golang.org/s/go13linker for more background.
//
// The file format is:
//
// - magic header: "\x00go112ld"
// - byte 1 - version number
// - sequence of strings giving dependencies (imported packages)
// - empty string (marks end of sequence)
// - sequence of symbol references used by the defined symbols
// - byte 0xff (marks end of sequence)
// - sequence of integer lengths:
// - total data length
// - total number of relocations
// - total number of pcdata
// - total number of automatics
// - total number of funcdata
// - total number of files
// - data, the content of the defined symbols
// - sequence of defined symbols
// - byte 0xff (marks end of sequence)
// - magic footer: "\xffgo112ld"
//
// All integers are stored in a zigzag varint format.
// See golang.org/s/go12symtab for a definition.
//
// Data blocks and strings are both stored as an integer
// followed by that many bytes.
//
// A symbol reference is a string name followed by an ABI or -1 for static.
//
// A symbol points to other symbols using an index into the symbol
// reference sequence. Index 0 corresponds to a nil symbol pointer.
// In the symbol layout described below "symref index" stands for this
// index.
//
// Each symbol is laid out as the following fields:
//
// - byte 0xfe (sanity check for synchronization)
// - type [byte]
// - name & ABI [symref index]
// - flags [int]
// 1<<0 dupok
// 1<<1 local
// 1<<2 add to typelink table
// - size [int]
// - gotype [symref index]
// - p [data block]
// - nr [int]
// - r [nr relocations, sorted by off]
//
// If type == STEXT, there are a few more fields:
//
// - args [int]
// - locals [int]
// - nosplit [int]
// - flags [int]
// 1<<0 leaf
// 1<<1 C function
// 1<<2 function may call reflect.Type.Method
// 1<<3 function compiled with -shared
// - nlocal [int]
// - local [nlocal automatics]
// - pcln [pcln table]
//
// Each relocation has the encoding:
//
// - off [int]
// - siz [int]
// - type [int]
// - add [int]
// - sym [symref index]
//
// Each local has the encoding:
//
// - asym [symref index]
// - offset [int]
// - type [int]
// - gotype [symref index]
//
// The pcln table has the encoding:
//
// - pcsp [data block]
// - pcfile [data block]
// - pcline [data block]
// - pcinline [data block]
// - npcdata [int]
// - pcdata [npcdata data blocks]
// - nfuncdata [int]
// - funcdata [nfuncdata symref index]
// - funcdatasym [nfuncdata ints]
// - nfile [int]
// - file [nfile symref index]
// - ninlinedcall [int]
// - inlinedcall [ninlinedcall int symref int symref]
//
// The file layout and meaning of type integers are architecture-independent.
//
// TODO(rsc): The file format is good for a first pass but needs work.
// - There are SymID in the object file that should really just be strings.
package objabi