)
const (
- // TODO: the Microsoft doco says IMAGE_SYM_DTYPE_ARRAY is 3 (same with IMAGE_SYM_DTYPE_POINTER and IMAGE_SYM_DTYPE_FUNCTION)
IMAGE_SYM_UNDEFINED = 0
IMAGE_SYM_ABSOLUTE = -1
IMAGE_SYM_DEBUG = -2
IMAGE_SYM_TYPE_DWORD = 15
IMAGE_SYM_TYPE_PCODE = 32768
IMAGE_SYM_DTYPE_NULL = 0
- IMAGE_SYM_DTYPE_POINTER = 0x10
- IMAGE_SYM_DTYPE_FUNCTION = 0x20
- IMAGE_SYM_DTYPE_ARRAY = 0x30
+ IMAGE_SYM_DTYPE_POINTER = 1
+ IMAGE_SYM_DTYPE_FUNCTION = 2
+ IMAGE_SYM_DTYPE_ARRAY = 3
IMAGE_SYM_CLASS_END_OF_FUNCTION = -1
IMAGE_SYM_CLASS_NULL = 0
IMAGE_SYM_CLASS_AUTOMATIC = 1
IMAGE_REL_THUMB_BRANCH24 = 0x0014
IMAGE_REL_THUMB_BLX23 = 0x0015
IMAGE_REL_ARM_PAIR = 0x0016
+ IMAGE_REL_ARM64_ABSOLUTE = 0x0000
+ IMAGE_REL_ARM64_ADDR32 = 0x0001
+ IMAGE_REL_ARM64_ADDR32NB = 0x0002
+ IMAGE_REL_ARM64_BRANCH26 = 0x0003
+ IMAGE_REL_ARM64_PAGEBASE_REL21 = 0x0004
+ IMAGE_REL_ARM64_REL21 = 0x0005
+ IMAGE_REL_ARM64_PAGEOFFSET_12A = 0x0006
+ IMAGE_REL_ARM64_PAGEOFFSET_12L = 0x0007
+ IMAGE_REL_ARM64_SECREL = 0x0008
+ IMAGE_REL_ARM64_SECREL_LOW12A = 0x0009
+ IMAGE_REL_ARM64_SECREL_HIGH12A = 0x000A
+ IMAGE_REL_ARM64_SECREL_LOW12L = 0x000B
+ IMAGE_REL_ARM64_TOKEN = 0x000C
+ IMAGE_REL_ARM64_SECTION = 0x000D
+ IMAGE_REL_ARM64_ADDR64 = 0x000E
+ IMAGE_REL_ARM64_BRANCH19 = 0x000F
+ IMAGE_REL_ARM64_BRANCH14 = 0x0010
+ IMAGE_REL_ARM64_REL32 = 0x0011
)
-// TODO(crawshaw): de-duplicate these symbols with cmd/internal/ld, ideally in debug/pe.
const (
- IMAGE_SCN_CNT_CODE = 0x00000020
- IMAGE_SCN_CNT_INITIALIZED_DATA = 0x00000040
- IMAGE_SCN_CNT_UNINITIALIZED_DATA = 0x00000080
- IMAGE_SCN_MEM_DISCARDABLE = 0x02000000
- IMAGE_SCN_MEM_EXECUTE = 0x20000000
- IMAGE_SCN_MEM_READ = 0x40000000
- IMAGE_SCN_MEM_WRITE = 0x80000000
+ // When stored into the PLT value for a symbol, this token tells
+ // windynrelocsym to redirect direct references to this symbol to a stub
+ // that loads from the corresponding import symbol and then does
+ // a jump to the loaded value.
+ CreateImportStubPltToken = -2
+
+ // When stored into the GOT value for an import symbol __imp_X this
+ // token tells windynrelocsym to redirect references to the
+ // underlying DYNIMPORT symbol X.
+ RedirectToDynImportGotToken = -2
)
// TODO(brainman): maybe just add ReadAt method to bio.Reader instead of creating peBiobuf
return bld
}
+// peImportSymsState tracks the set of DLL import symbols we've seen
+// while reading host objects. We create a singleton instance of this
+// type, which will persist across multiple host objects.
+type peImportSymsState struct {
+
+ // Text and non-text sections read in by the host object loader.
+ secSyms []loader.Sym
+
+ // Loader and arch, for use in postprocessing.
+ l *loader.Loader
+ arch *sys.Arch
+}
+
+var importSymsState *peImportSymsState
+
+func createImportSymsState(l *loader.Loader, arch *sys.Arch) {
+ if importSymsState != nil {
+ return
+ }
+ importSymsState = &peImportSymsState{
+ l: l,
+ arch: arch,
+ }
+}
+
+// peLoaderState holds various bits of useful state information needed
+// while loading a single PE object file.
+type peLoaderState struct {
+ l *loader.Loader
+ arch *sys.Arch
+ f *pe.File
+ pn string
+ sectsyms map[*pe.Section]loader.Sym
+ comdats map[uint16]int64 // key is section index, val is size
+ sectdata map[*pe.Section][]byte
+ localSymVersion int
+}
+
+// comdatDefinitions records the names of symbols for which we've
+// previously seen a definition in COMDAT. Key is symbol name, value
+// is symbol size (or -1 if we're using the "any" strategy).
+var comdatDefinitions map[string]int64
+
+// Symbols contains the symbols that can be loaded from a PE file.
+type Symbols struct {
+ Textp []loader.Sym // text symbols
+ Resources []loader.Sym // .rsrc section or set of .rsrc$xx sections
+ PData loader.Sym
+ XData loader.Sym
+}
+
// Load loads the PE file pn from input.
-// Symbols are written into syms, and a slice of the text symbols is returned.
-// If an .rsrc section or set of .rsrc$xx sections is found, its symbols are
-// returned as rsrc.
-func Load(l *loader.Loader, arch *sys.Arch, localSymVersion int, input *bio.Reader, pkg string, length int64, pn string) (textp []loader.Sym, rsrc []loader.Sym, err error) {
- lookup := func(name string, version int) (*loader.SymbolBuilder, loader.Sym) {
- s := l.LookupOrCreateSym(name, version)
- sb := l.MakeSymbolUpdater(s)
- return sb, s
+// Symbols from the object file are created via the loader 'l'.
+func Load(l *loader.Loader, arch *sys.Arch, localSymVersion int, input *bio.Reader, pkg string, length int64, pn string) (*Symbols, error) {
+ state := &peLoaderState{
+ l: l,
+ arch: arch,
+ sectsyms: make(map[*pe.Section]loader.Sym),
+ sectdata: make(map[*pe.Section][]byte),
+ localSymVersion: localSymVersion,
+ pn: pn,
+ }
+ createImportSymsState(state.l, state.arch)
+ if comdatDefinitions == nil {
+ comdatDefinitions = make(map[string]int64)
}
- sectsyms := make(map[*pe.Section]loader.Sym)
- sectdata := make(map[*pe.Section][]byte)
// Some input files are archives containing multiple of
// object files, and pe.NewFile seeks to the start of
// TODO: replace pe.NewFile with pe.Load (grep for "add Load function" in debug/pe for details)
f, err := pe.NewFile(sr)
if err != nil {
- return nil, nil, err
+ return nil, err
}
defer f.Close()
+ state.f = f
+
+ var ls Symbols
// TODO return error if found .cormeta
// create symbols for mapped sections
for _, sect := range f.Sections {
- if sect.Characteristics&IMAGE_SCN_MEM_DISCARDABLE != 0 {
+ if sect.Characteristics&pe.IMAGE_SCN_MEM_DISCARDABLE != 0 {
continue
}
- if sect.Characteristics&(IMAGE_SCN_CNT_CODE|IMAGE_SCN_CNT_INITIALIZED_DATA|IMAGE_SCN_CNT_UNINITIALIZED_DATA) == 0 {
+ if sect.Characteristics&(pe.IMAGE_SCN_CNT_CODE|pe.IMAGE_SCN_CNT_INITIALIZED_DATA|pe.IMAGE_SCN_CNT_UNINITIALIZED_DATA) == 0 {
// This has been seen for .idata sections, which we
// want to ignore. See issues 5106 and 5273.
continue
}
name := fmt.Sprintf("%s(%s)", pkg, sect.Name)
- bld, s := lookup(name, localSymVersion)
-
- switch sect.Characteristics & (IMAGE_SCN_CNT_UNINITIALIZED_DATA | IMAGE_SCN_CNT_INITIALIZED_DATA | IMAGE_SCN_MEM_READ | IMAGE_SCN_MEM_WRITE | IMAGE_SCN_CNT_CODE | IMAGE_SCN_MEM_EXECUTE) {
- case IMAGE_SCN_CNT_INITIALIZED_DATA | IMAGE_SCN_MEM_READ: //.rdata
- bld.SetType(sym.SRODATA)
+ s := state.l.LookupOrCreateCgoExport(name, localSymVersion)
+ bld := l.MakeSymbolUpdater(s)
+
+ switch sect.Characteristics & (pe.IMAGE_SCN_CNT_UNINITIALIZED_DATA | pe.IMAGE_SCN_CNT_INITIALIZED_DATA | pe.IMAGE_SCN_MEM_READ | pe.IMAGE_SCN_MEM_WRITE | pe.IMAGE_SCN_CNT_CODE | pe.IMAGE_SCN_MEM_EXECUTE) {
+ case pe.IMAGE_SCN_CNT_INITIALIZED_DATA | pe.IMAGE_SCN_MEM_READ: //.rdata
+ if issehsect(arch, sect) {
+ bld.SetType(sym.SSEHSECT)
+ bld.SetAlign(4)
+ } else {
+ bld.SetType(sym.SRODATA)
+ }
- case IMAGE_SCN_CNT_UNINITIALIZED_DATA | IMAGE_SCN_MEM_READ | IMAGE_SCN_MEM_WRITE: //.bss
+ case pe.IMAGE_SCN_CNT_UNINITIALIZED_DATA | pe.IMAGE_SCN_MEM_READ | pe.IMAGE_SCN_MEM_WRITE: //.bss
bld.SetType(sym.SNOPTRBSS)
- case IMAGE_SCN_CNT_INITIALIZED_DATA | IMAGE_SCN_MEM_READ | IMAGE_SCN_MEM_WRITE: //.data
+ case pe.IMAGE_SCN_CNT_INITIALIZED_DATA | pe.IMAGE_SCN_MEM_READ | pe.IMAGE_SCN_MEM_WRITE: //.data
bld.SetType(sym.SNOPTRDATA)
- case IMAGE_SCN_CNT_CODE | IMAGE_SCN_MEM_EXECUTE | IMAGE_SCN_MEM_READ: //.text
+ case pe.IMAGE_SCN_CNT_CODE | pe.IMAGE_SCN_MEM_EXECUTE | pe.IMAGE_SCN_MEM_READ: //.text
bld.SetType(sym.STEXT)
default:
- return nil, nil, fmt.Errorf("unexpected flags %#06x for PE section %s", sect.Characteristics, sect.Name)
+ return nil, fmt.Errorf("unexpected flags %#06x for PE section %s", sect.Characteristics, sect.Name)
}
if bld.Type() != sym.SNOPTRBSS {
data, err := sect.Data()
if err != nil {
- return nil, nil, err
+ return nil, err
}
- sectdata[sect] = data
+ state.sectdata[sect] = data
bld.SetData(data)
}
bld.SetSize(int64(sect.Size))
- sectsyms[sect] = s
+ state.sectsyms[sect] = s
if sect.Name == ".rsrc" || strings.HasPrefix(sect.Name, ".rsrc$") {
- rsrc = append(rsrc, s)
+ ls.Resources = append(ls.Resources, s)
+ } else if bld.Type() == sym.SSEHSECT {
+ if sect.Name == ".pdata" {
+ ls.PData = s
+ } else if sect.Name == ".xdata" {
+ ls.XData = s
+ }
}
}
+ // Make a prepass over the symbols to collect info about COMDAT symbols.
+ if err := state.preprocessSymbols(); err != nil {
+ return nil, err
+ }
+
// load relocations
for _, rsect := range f.Sections {
- if _, found := sectsyms[rsect]; !found {
+ if _, found := state.sectsyms[rsect]; !found {
continue
}
if rsect.NumberOfRelocations == 0 {
continue
}
- if rsect.Characteristics&IMAGE_SCN_MEM_DISCARDABLE != 0 {
+ if rsect.Characteristics&pe.IMAGE_SCN_MEM_DISCARDABLE != 0 {
continue
}
- if rsect.Characteristics&(IMAGE_SCN_CNT_CODE|IMAGE_SCN_CNT_INITIALIZED_DATA|IMAGE_SCN_CNT_UNINITIALIZED_DATA) == 0 {
+ if rsect.Characteristics&(pe.IMAGE_SCN_CNT_CODE|pe.IMAGE_SCN_CNT_INITIALIZED_DATA|pe.IMAGE_SCN_CNT_UNINITIALIZED_DATA) == 0 {
// This has been seen for .idata sections, which we
// want to ignore. See issues 5106 and 5273.
continue
}
splitResources := strings.HasPrefix(rsect.Name, ".rsrc$")
- sb := l.MakeSymbolUpdater(sectsyms[rsect])
+ issehsect := issehsect(arch, rsect)
+ sb := l.MakeSymbolUpdater(state.sectsyms[rsect])
for j, r := range rsect.Relocs {
if int(r.SymbolTableIndex) >= len(f.COFFSymbols) {
- return nil, nil, fmt.Errorf("relocation number %d symbol index idx=%d cannot be large then number of symbols %d", j, r.SymbolTableIndex, len(f.COFFSymbols))
+ return nil, fmt.Errorf("relocation number %d symbol index idx=%d cannot be large then number of symbols %d", j, r.SymbolTableIndex, len(f.COFFSymbols))
}
pesym := &f.COFFSymbols[r.SymbolTableIndex]
- _, gosym, err := readpesym(l, arch, l.LookupOrCreateSym, f, pesym, sectsyms, localSymVersion)
+ _, gosym, err := state.readpesym(pesym)
if err != nil {
- return nil, nil, err
+ return nil, err
}
if gosym == 0 {
name, err := pesym.FullName(f.StringTable)
if err != nil {
name = string(pesym.Name[:])
}
- return nil, nil, fmt.Errorf("reloc of invalid sym %s idx=%d type=%d", name, r.SymbolTableIndex, pesym.Type)
+ return nil, fmt.Errorf("reloc of invalid sym %s idx=%d type=%d", name, r.SymbolTableIndex, pesym.Type)
}
rSym := gosym
var rType objabi.RelocType
switch arch.Family {
default:
- return nil, nil, fmt.Errorf("%s: unsupported arch %v", pn, arch.Family)
+ return nil, fmt.Errorf("%s: unsupported arch %v", pn, arch.Family)
case sys.I386, sys.AMD64:
switch r.Type {
default:
- return nil, nil, fmt.Errorf("%s: %v: unknown relocation type %v", pn, sectsyms[rsect], r.Type)
+ return nil, fmt.Errorf("%s: %v: unknown relocation type %v", pn, state.sectsyms[rsect], r.Type)
case IMAGE_REL_I386_REL32, IMAGE_REL_AMD64_REL32,
IMAGE_REL_AMD64_ADDR32, // R_X86_64_PC32
IMAGE_REL_AMD64_ADDR32NB:
- rType = objabi.R_PCREL
+ if r.Type == IMAGE_REL_AMD64_ADDR32NB {
+ rType = objabi.R_PEIMAGEOFF
+ } else {
+ rType = objabi.R_PCREL
+ }
- rAdd = int64(int32(binary.LittleEndian.Uint32(sectdata[rsect][rOff:])))
+ rAdd = int64(int32(binary.LittleEndian.Uint32(state.sectdata[rsect][rOff:])))
case IMAGE_REL_I386_DIR32NB, IMAGE_REL_I386_DIR32:
- rType = objabi.R_ADDR
+ if r.Type == IMAGE_REL_I386_DIR32NB {
+ rType = objabi.R_PEIMAGEOFF
+ } else {
+ rType = objabi.R_ADDR
+ }
// load addend from image
- rAdd = int64(int32(binary.LittleEndian.Uint32(sectdata[rsect][rOff:])))
+ rAdd = int64(int32(binary.LittleEndian.Uint32(state.sectdata[rsect][rOff:])))
case IMAGE_REL_AMD64_ADDR64: // R_X86_64_64
rSize = 8
rType = objabi.R_ADDR
// load addend from image
- rAdd = int64(binary.LittleEndian.Uint64(sectdata[rsect][rOff:]))
+ rAdd = int64(binary.LittleEndian.Uint64(state.sectdata[rsect][rOff:]))
}
case sys.ARM:
switch r.Type {
default:
- return nil, nil, fmt.Errorf("%s: %v: unknown ARM relocation type %v", pn, sectsyms[rsect], r.Type)
+ return nil, fmt.Errorf("%s: %v: unknown ARM relocation type %v", pn, state.sectsyms[rsect], r.Type)
case IMAGE_REL_ARM_SECREL:
rType = objabi.R_PCREL
- rAdd = int64(int32(binary.LittleEndian.Uint32(sectdata[rsect][rOff:])))
+ rAdd = int64(int32(binary.LittleEndian.Uint32(state.sectdata[rsect][rOff:])))
case IMAGE_REL_ARM_ADDR32, IMAGE_REL_ARM_ADDR32NB:
- rType = objabi.R_ADDR
+ if r.Type == IMAGE_REL_ARM_ADDR32NB {
+ rType = objabi.R_PEIMAGEOFF
+ } else {
+ rType = objabi.R_ADDR
+ }
- rAdd = int64(int32(binary.LittleEndian.Uint32(sectdata[rsect][rOff:])))
+ rAdd = int64(int32(binary.LittleEndian.Uint32(state.sectdata[rsect][rOff:])))
case IMAGE_REL_ARM_BRANCH24:
rType = objabi.R_CALLARM
- rAdd = int64(int32(binary.LittleEndian.Uint32(sectdata[rsect][rOff:])))
+ rAdd = int64(int32(binary.LittleEndian.Uint32(state.sectdata[rsect][rOff:])))
+ }
+
+ case sys.ARM64:
+ switch r.Type {
+ default:
+ return nil, fmt.Errorf("%s: %v: unknown ARM64 relocation type %v", pn, state.sectsyms[rsect], r.Type)
+
+ case IMAGE_REL_ARM64_ADDR32, IMAGE_REL_ARM64_ADDR32NB:
+ if r.Type == IMAGE_REL_ARM64_ADDR32NB {
+ rType = objabi.R_PEIMAGEOFF
+ } else {
+ rType = objabi.R_ADDR
+ }
+
+ rAdd = int64(int32(binary.LittleEndian.Uint32(state.sectdata[rsect][rOff:])))
}
}
if issect(pesym) || splitResources {
rAdd += int64(pesym.Value)
}
+ if issehsect {
+ // .pdata and .xdata sections can contain records
+ // associated to functions that won't be used in
+ // the final binary, in which case the relocation
+ // target symbol won't be reachable.
+ rType |= objabi.R_WEAK
+ }
rel, _ := sb.AddRel(rType)
rel.SetOff(rOff)
rel.SetSiz(rSize)
rel.SetSym(rSym)
rel.SetAdd(rAdd)
+
}
sb.SortRelocs()
name, err := pesym.FullName(f.StringTable)
if err != nil {
- return nil, nil, err
+ return nil, err
}
if name == "" {
continue
var sect *pe.Section
if pesym.SectionNumber > 0 {
sect = f.Sections[pesym.SectionNumber-1]
- if _, found := sectsyms[sect]; !found {
+ if _, found := state.sectsyms[sect]; !found {
continue
}
}
- bld, s, err := readpesym(l, arch, l.LookupOrCreateSym, f, pesym, sectsyms, localSymVersion)
+ bld, s, err := state.readpesym(pesym)
if err != nil {
- return nil, nil, err
+ return nil, err
}
if pesym.SectionNumber == 0 { // extern
- if l.SymType(s) == sym.SDYNIMPORT {
- bld = makeUpdater(l, bld, s)
- bld.SetPlt(-2) // flag for dynimport in PE object files.
- }
if l.SymType(s) == sym.SXREF && pesym.Value > 0 { // global data
bld = makeUpdater(l, bld, s)
bld.SetType(sym.SNOPTRDATA)
continue
} else if pesym.SectionNumber > 0 && int(pesym.SectionNumber) <= len(f.Sections) {
sect = f.Sections[pesym.SectionNumber-1]
- if _, found := sectsyms[sect]; !found {
- return nil, nil, fmt.Errorf("%s: %v: missing sect.sym", pn, s)
+ if _, found := state.sectsyms[sect]; !found {
+ return nil, fmt.Errorf("%s: %v: missing sect.sym", pn, s)
}
} else {
- return nil, nil, fmt.Errorf("%s: %v: sectnum < 0!", pn, s)
+ return nil, fmt.Errorf("%s: %v: sectnum < 0!", pn, s)
}
if sect == nil {
- return nil, nil, nil
+ return nil, nil
}
+ // Check for COMDAT symbol.
+ if sz, ok1 := state.comdats[uint16(pesym.SectionNumber-1)]; ok1 {
+ if psz, ok2 := comdatDefinitions[l.SymName(s)]; ok2 {
+ if sz == psz {
+ // OK to discard, we've seen an instance
+ // already.
+ continue
+ }
+ }
+ }
if l.OuterSym(s) != 0 {
if l.AttrDuplicateOK(s) {
continue
}
outerName := l.SymName(l.OuterSym(s))
- sectName := l.SymName(sectsyms[sect])
- return nil, nil, fmt.Errorf("%s: duplicate symbol reference: %s in both %s and %s", pn, l.SymName(s), outerName, sectName)
+ sectName := l.SymName(state.sectsyms[sect])
+ return nil, fmt.Errorf("%s: duplicate symbol reference: %s in both %s and %s", pn, l.SymName(s), outerName, sectName)
}
bld = makeUpdater(l, bld, s)
- sectsym := sectsyms[sect]
+ sectsym := state.sectsyms[sect]
bld.SetType(l.SymType(sectsym))
l.AddInteriorSym(sectsym, s)
bld.SetValue(int64(pesym.Value))
bld.SetSize(4)
if l.SymType(sectsym) == sym.STEXT {
if bld.External() && !bld.DuplicateOK() {
- return nil, nil, fmt.Errorf("%s: duplicate symbol definition", l.SymName(s))
+ return nil, fmt.Errorf("%s: duplicate symbol definition", l.SymName(s))
}
bld.SetExternal(true)
}
+ if sz, ok := state.comdats[uint16(pesym.SectionNumber-1)]; ok {
+ // This is a COMDAT definition. Record that we're picking
+ // this instance so that we can ignore future defs.
+ if _, ok := comdatDefinitions[l.SymName(s)]; ok {
+ return nil, fmt.Errorf("internal error: preexisting COMDAT definition for %q", name)
+ }
+ comdatDefinitions[l.SymName(s)] = sz
+ }
}
// Sort outer lists by address, adding to textp.
// This keeps textp in increasing address order.
for _, sect := range f.Sections {
- s := sectsyms[sect]
+ s := state.sectsyms[sect]
if s == 0 {
continue
}
l.SortSub(s)
+ importSymsState.secSyms = append(importSymsState.secSyms, s)
if l.SymType(s) == sym.STEXT {
for ; s != 0; s = l.SubSym(s) {
if l.AttrOnList(s) {
- return nil, nil, fmt.Errorf("symbol %s listed multiple times", l.SymName(s))
+ return nil, fmt.Errorf("symbol %s listed multiple times", l.SymName(s))
}
l.SetAttrOnList(s, true)
- textp = append(textp, s)
+ ls.Textp = append(ls.Textp, s)
+ }
+ }
+ }
+
+ if ls.PData != 0 {
+ processSEH(l, arch, ls.PData, ls.XData)
+ }
+
+ return &ls, nil
+}
+
+// PostProcessImports works to resolve inconsistencies with DLL import
+// symbols; it is needed when building with more "modern" C compilers
+// with internal linkage.
+//
+// Background: DLL import symbols are data (SNOPTRDATA) symbols whose
+// name is of the form "__imp_XXX", which contain a pointer/reference
+// to symbol XXX. It's possible to have import symbols for both data
+// symbols ("__imp__fmode") and text symbols ("__imp_CreateEventA").
+// In some case import symbols are just references to some external
+// thing, and in other cases we see actual definitions of import
+// symbols when reading host objects.
+//
+// Previous versions of the linker would in most cases immediately
+// "forward" import symbol references, e.g. treat a references to
+// "__imp_XXX" a references to "XXX", however this doesn't work well
+// with more modern compilers, where you can sometimes see import
+// symbols that are defs (as opposed to external refs).
+//
+// The main actions taken below are to search for references to
+// SDYNIMPORT symbols in host object text/data sections and flag the
+// symbols for later fixup. When we see a reference to an import
+// symbol __imp_XYZ where XYZ corresponds to some SDYNIMPORT symbol,
+// we flag the symbol (via GOT setting) so that it can be redirected
+// to XYZ later in windynrelocsym. When we see a direct reference to
+// an SDYNIMPORT symbol XYZ, we also flag the symbol (via PLT setting)
+// to indicated that the reference will need to be redirected to a
+// stub.
+func PostProcessImports() error {
+ ldr := importSymsState.l
+ arch := importSymsState.arch
+ keeprelocneeded := make(map[loader.Sym]loader.Sym)
+ for _, s := range importSymsState.secSyms {
+ isText := ldr.SymType(s) == sym.STEXT
+ relocs := ldr.Relocs(s)
+ for i := 0; i < relocs.Count(); i++ {
+ r := relocs.At(i)
+ rs := r.Sym()
+ if ldr.SymType(rs) == sym.SDYNIMPORT {
+ // Tag the symbol for later stub generation.
+ ldr.SetPlt(rs, CreateImportStubPltToken)
+ continue
+ }
+ isym, err := LookupBaseFromImport(rs, ldr, arch)
+ if err != nil {
+ return err
+ }
+ if isym == 0 {
+ continue
}
+ if ldr.SymType(isym) != sym.SDYNIMPORT {
+ continue
+ }
+ // For non-text symbols, forward the reference from __imp_X to
+ // X immediately.
+ if !isText {
+ r.SetSym(isym)
+ continue
+ }
+ // Flag this imp symbol to be processed later in windynrelocsym.
+ ldr.SetGot(rs, RedirectToDynImportGotToken)
+ // Consistency check: should be no PLT token here.
+ splt := ldr.SymPlt(rs)
+ if splt != -1 {
+ return fmt.Errorf("internal error: import symbol %q has invalid PLT setting %d", ldr.SymName(rs), splt)
+ }
+ // Flag for dummy relocation.
+ keeprelocneeded[rs] = isym
}
}
+ for k, v := range keeprelocneeded {
+ sb := ldr.MakeSymbolUpdater(k)
+ r, _ := sb.AddRel(objabi.R_KEEP)
+ r.SetSym(v)
+ }
+ importSymsState = nil
+ return nil
+}
- return textp, rsrc, nil
+func issehsect(arch *sys.Arch, s *pe.Section) bool {
+ return arch.Family == sys.AMD64 && (s.Name == ".pdata" || s.Name == ".xdata")
}
func issect(s *pe.COFFSymbol) bool {
return s.StorageClass == IMAGE_SYM_CLASS_STATIC && s.Type == 0 && s.Name[0] == '.'
}
-func readpesym(l *loader.Loader, arch *sys.Arch, lookup func(string, int) loader.Sym, f *pe.File, pesym *pe.COFFSymbol, sectsyms map[*pe.Section]loader.Sym, localSymVersion int) (*loader.SymbolBuilder, loader.Sym, error) {
- symname, err := pesym.FullName(f.StringTable)
+func (state *peLoaderState) readpesym(pesym *pe.COFFSymbol) (*loader.SymbolBuilder, loader.Sym, error) {
+ symname, err := pesym.FullName(state.f.StringTable)
if err != nil {
return nil, 0, err
}
var name string
if issect(pesym) {
- name = l.SymName(sectsyms[f.Sections[pesym.SectionNumber-1]])
+ name = state.l.SymName(state.sectsyms[state.f.Sections[pesym.SectionNumber-1]])
} else {
name = symname
- switch arch.Family {
- case sys.AMD64:
- if name == "__imp___acrt_iob_func" {
- // Do not rename __imp___acrt_iob_func into __acrt_iob_func,
- // because __imp___acrt_iob_func symbol is real
- // (see commit b295099 from git://git.code.sf.net/p/mingw-w64/mingw-w64 for details).
- } else {
- name = strings.TrimPrefix(name, "__imp_") // __imp_Name => Name
- }
- case sys.I386:
- if name == "__imp____acrt_iob_func" {
- // Do not rename __imp____acrt_iob_func into ___acrt_iob_func,
- // because __imp____acrt_iob_func symbol is real
- // (see commit b295099 from git://git.code.sf.net/p/mingw-w64/mingw-w64 for details).
- } else {
- name = strings.TrimPrefix(name, "__imp_") // __imp_Name => Name
- }
- if name[0] == '_' {
- name = name[1:] // _Name => Name
- }
+ // A note on the "_main" exclusion below: the main routine
+ // defined by the Go runtime is named "_main", not "main", so
+ // when reading references to _main from a host object we want
+ // to avoid rewriting "_main" to "main" in this specific
+ // instance. See #issuecomment-1143698749 on #35006 for more
+ // details on this problem.
+ if state.arch.Family == sys.I386 && name[0] == '_' && name != "_main" && !strings.HasPrefix(name, "__imp_") {
+ name = name[1:] // _Name => Name
}
}
var s loader.Sym
var bld *loader.SymbolBuilder
- switch pesym.Type {
+ // Microsoft's PE documentation is contradictory. It says that the symbol's complex type
+ // is stored in the pesym.Type most significant byte, but MSVC, LLVM, and mingw store it
+ // in the 4 high bits of the less significant byte.
+ switch uint8(pesym.Type&0xf0) >> 4 {
default:
return nil, 0, fmt.Errorf("%s: invalid symbol type %d", symname, pesym.Type)
case IMAGE_SYM_DTYPE_FUNCTION, IMAGE_SYM_DTYPE_NULL:
switch pesym.StorageClass {
case IMAGE_SYM_CLASS_EXTERNAL: //global
- s = lookup(name, 0)
+ s = state.l.LookupOrCreateCgoExport(name, 0)
case IMAGE_SYM_CLASS_NULL, IMAGE_SYM_CLASS_STATIC, IMAGE_SYM_CLASS_LABEL:
- s = lookup(name, localSymVersion)
- bld = makeUpdater(l, bld, s)
+ s = state.l.LookupOrCreateCgoExport(name, state.localSymVersion)
+ bld = makeUpdater(state.l, bld, s)
bld.SetDuplicateOK(true)
default:
}
}
- if s != 0 && l.SymType(s) == 0 && (pesym.StorageClass != IMAGE_SYM_CLASS_STATIC || pesym.Value != 0) {
- bld = makeUpdater(l, bld, s)
+ if s != 0 && state.l.SymType(s) == 0 && (pesym.StorageClass != IMAGE_SYM_CLASS_STATIC || pesym.Value != 0) {
+ bld = makeUpdater(state.l, bld, s)
bld.SetType(sym.SXREF)
}
- if strings.HasPrefix(symname, "__imp_") {
- bld = makeUpdater(l, bld, s)
- bld.SetGot(-2) // flag for __imp_
- }
return bld, s, nil
}
+
+// preprocessSymbols walks the COFF symbols for the PE file we're
+// reading and looks for cases where we have both a symbol definition
+// for "XXX" and an "__imp_XXX" symbol, recording these cases in a map
+// in the state struct. This information will be used in readpesym()
+// above to give such symbols special treatment. This function also
+// gathers information about COMDAT sections/symbols for later use
+// in readpesym().
+func (state *peLoaderState) preprocessSymbols() error {
+
+ // Locate comdat sections.
+ state.comdats = make(map[uint16]int64)
+ for i, s := range state.f.Sections {
+ if s.Characteristics&uint32(pe.IMAGE_SCN_LNK_COMDAT) != 0 {
+ state.comdats[uint16(i)] = int64(s.Size)
+ }
+ }
+
+ // Examine symbol defs.
+ for i, numaux := 0, 0; i < len(state.f.COFFSymbols); i += numaux + 1 {
+ pesym := &state.f.COFFSymbols[i]
+ numaux = int(pesym.NumberOfAuxSymbols)
+ if pesym.SectionNumber == 0 { // extern
+ continue
+ }
+ symname, err := pesym.FullName(state.f.StringTable)
+ if err != nil {
+ return err
+ }
+ if _, isc := state.comdats[uint16(pesym.SectionNumber-1)]; !isc {
+ continue
+ }
+ if pesym.StorageClass != uint8(IMAGE_SYM_CLASS_STATIC) {
+ continue
+ }
+ // This symbol corresponds to a COMDAT section. Read the
+ // aux data for it.
+ auxsymp, err := state.f.COFFSymbolReadSectionDefAux(i)
+ if err != nil {
+ return fmt.Errorf("unable to read aux info for section def symbol %d %s: pe.COFFSymbolReadComdatInfo returns %v", i, symname, err)
+ }
+ if auxsymp.Selection == pe.IMAGE_COMDAT_SELECT_SAME_SIZE {
+ // This is supported.
+ } else if auxsymp.Selection == pe.IMAGE_COMDAT_SELECT_ANY {
+ // Also supported.
+ state.comdats[uint16(pesym.SectionNumber-1)] = int64(-1)
+ } else {
+ // We don't support any of the other strategies at the
+ // moment. I suspect that we may need to also support
+ // "associative", we'll see.
+ return fmt.Errorf("internal error: unsupported COMDAT selection strategy found in path=%s sec=%d strategy=%d idx=%d, please file a bug", state.pn, auxsymp.SecNum, auxsymp.Selection, i)
+ }
+ }
+ return nil
+}
+
+// LookupBaseFromImport examines the symbol "s" to see if it
+// corresponds to an import symbol (name of the form "__imp_XYZ") and
+// if so, it looks up the underlying target of the import symbol and
+// returns it. An error is returned if the symbol is of the form
+// "__imp_XYZ" but no XYZ can be found.
+func LookupBaseFromImport(s loader.Sym, ldr *loader.Loader, arch *sys.Arch) (loader.Sym, error) {
+ sname := ldr.SymName(s)
+ if !strings.HasPrefix(sname, "__imp_") {
+ return 0, nil
+ }
+ basename := sname[len("__imp_"):]
+ if arch.Family == sys.I386 && basename[0] == '_' {
+ basename = basename[1:] // _Name => Name
+ }
+ isym := ldr.Lookup(basename, 0)
+ if isym == 0 {
+ return 0, fmt.Errorf("internal error: import symbol %q with no underlying sym", sname)
+ }
+ return isym, nil
+}