cmd/link, etc: store typelinks as offsets
This is the first in a series of CLs to replace the use of pointers
in binary read-only data with offsets.
In standard Go binaries these CLs have a small effect, shrinking
8-byte pointers to 4-bytes. In position-independent code, it also
saves the dynamic relocation for the pointer. This has a significant
effect on the binary size when building as PIE, c-archive, or
c-shared.
darwin/amd64:
cmd/go: -12KB (0.1%)
jujud: -82KB (0.1%)
linux/amd64 PIE:
cmd/go: -86KB (0.7%)
jujud: -569KB (0.7%)
For #6853.
Change-Id: Iad5625bbeba58dabfd4d334dbee3fcbfe04b2dcf
Reviewed-on: https://go-review.googlesource.com/21284
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
diff --git a/src/cmd/compile/internal/gc/go.go b/src/cmd/compile/internal/gc/go.go
index d9b28ff..5df49b5 100644
--- a/src/cmd/compile/internal/gc/go.go
+++ b/src/cmd/compile/internal/gc/go.go
@@ -171,8 +171,6 @@
var typepkg *Pkg // fake package for runtime type info (headers)
-var typelinkpkg *Pkg // fake package for runtime type info (data)
-
var unsafepkg *Pkg // package unsafe
var trackpkg *Pkg // fake package for field tracking
diff --git a/src/cmd/compile/internal/gc/main.go b/src/cmd/compile/internal/gc/main.go
index 26acf88..45a510d 100644
--- a/src/cmd/compile/internal/gc/main.go
+++ b/src/cmd/compile/internal/gc/main.go
@@ -126,10 +126,6 @@
itabpkg.Name = "go.itab"
itabpkg.Prefix = "go.itab" // not go%2eitab
- typelinkpkg = mkpkg("go.typelink")
- typelinkpkg.Name = "go.typelink"
- typelinkpkg.Prefix = "go.typelink" // not go%2etypelink
-
itablinkpkg = mkpkg("go.itablink")
itablinkpkg.Name = "go.itablink"
itablinkpkg.Prefix = "go.itablink" // not go%2eitablink
diff --git a/src/cmd/compile/internal/gc/obj.go b/src/cmd/compile/internal/gc/obj.go
index 23c8be6..eed0ed6 100644
--- a/src/cmd/compile/internal/gc/obj.go
+++ b/src/cmd/compile/internal/gc/obj.go
@@ -321,6 +321,12 @@
return off
}
+func dsymptrOffLSym(s *obj.LSym, off int, x *obj.LSym, xoff int) int {
+ s.WriteOff(Ctxt, int64(off), x, int64(xoff))
+ off += 4
+ return off
+}
+
func gdata(nam *Node, nr *Node, wid int) {
if nam.Op != ONAME {
Fatalf("gdata nam op %v", opnames[nam.Op])
diff --git a/src/cmd/compile/internal/gc/reflect.go b/src/cmd/compile/internal/gc/reflect.go
index df9ef27..ea67634 100644
--- a/src/cmd/compile/internal/gc/reflect.go
+++ b/src/cmd/compile/internal/gc/reflect.go
@@ -879,7 +879,7 @@
return Pkglookup(Tconv(t, FmtLeft)+"."+f.Sym.Name, trackpkg)
}
-func typelinksym(t *Type) *Sym {
+func typelinkLSym(t *Type) *obj.LSym {
// %-uT is what the generated Type's string field says.
// It uses (ambiguous) package names instead of import paths.
// %-T is the complete, unambiguous type name.
@@ -889,13 +889,8 @@
// ensure the types appear sorted by their string field. The
// names are a little long but they are discarded by the linker
// and do not end up in the symbol table of the final binary.
- p := Tconv(t, FmtLeft|FmtUnsigned) + "\t" + Tconv(t, FmtLeft)
-
- s := Pkglookup(p, typelinkpkg)
-
- //print("typelinksym: %s -> %+S\n", p, s);
-
- return s
+ name := "go.typelink." + Tconv(t, FmtLeft|FmtUnsigned) + "\t" + Tconv(t, FmtLeft)
+ return obj.Linklookup(Ctxt, name, 0)
}
func typesymprefix(prefix string, t *Type) *Sym {
@@ -1298,9 +1293,9 @@
if t.Sym == nil {
switch t.Etype {
case TPTR32, TPTR64, TARRAY, TCHAN, TFUNC, TMAP, TSTRUCT:
- slink := typelinksym(t)
- dsymptr(slink, 0, s, 0)
- ggloblsym(slink, int32(Widthptr), int16(dupok|obj.RODATA))
+ slink := typelinkLSym(t)
+ dsymptrOffLSym(slink, 0, Linksym(s), 0)
+ ggloblLSym(slink, 4, int16(dupok|obj.RODATA))
}
}
diff --git a/src/cmd/internal/obj/data.go b/src/cmd/internal/obj/data.go
index 37ab70b..546ff37 100644
--- a/src/cmd/internal/obj/data.go
+++ b/src/cmd/internal/obj/data.go
@@ -111,17 +111,36 @@
// rsym and roff specify the relocation for the address.
func (s *LSym) WriteAddr(ctxt *Link, off int64, siz int, rsym *LSym, roff int64) {
if siz != ctxt.Arch.PtrSize {
- ctxt.Diag("WriteAddr: bad address size: %d", siz)
+ ctxt.Diag("WriteAddr: bad address size %d in %s", siz, s.Name)
}
s.prepwrite(ctxt, off, siz)
r := Addrel(s)
r.Off = int32(off)
+ if int64(r.Off) != off {
+ ctxt.Diag("WriteAddr: off overflow %d in %s", off, s.Name)
+ }
r.Siz = uint8(siz)
r.Sym = rsym
r.Type = R_ADDR
r.Add = roff
}
+// 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 (s *LSym) WriteOff(ctxt *Link, off int64, rsym *LSym, roff int64) {
+ s.prepwrite(ctxt, off, 4)
+ r := Addrel(s)
+ r.Off = int32(off)
+ if int64(r.Off) != off {
+ ctxt.Diag("WriteOff: off overflow %d in %s", off, s.Name)
+ }
+ r.Siz = 4
+ r.Sym = rsym
+ r.Type = R_ADDROFF
+ r.Add = roff
+}
+
// WriteString writes a string of size siz into s at offset off.
func (s *LSym) WriteString(ctxt *Link, off int64, siz int, str string) {
if siz < len(str) {
diff --git a/src/cmd/internal/obj/link.go b/src/cmd/internal/obj/link.go
index 62175f9..d44d439 100644
--- a/src/cmd/internal/obj/link.go
+++ b/src/cmd/internal/obj/link.go
@@ -457,6 +457,9 @@
// R_ADDRMIPS (only used on mips64) resolves to a 32-bit external address,
// by loading the address into a register with two instructions (lui, ori).
R_ADDRMIPS
+ // R_ADDROFF resolves to an offset from the beginning of the section holding
+ // the data being relocated to the referenced symbol.
+ R_ADDROFF
R_SIZE
R_CALL
R_CALLARM
diff --git a/src/cmd/link/internal/ld/data.go b/src/cmd/link/internal/ld/data.go
index ae7c287..cf51b0a 100644
--- a/src/cmd/link/internal/ld/data.go
+++ b/src/cmd/link/internal/ld/data.go
@@ -525,6 +525,9 @@
}
o = Symaddr(r.Sym) + r.Add - int64(r.Sym.Sect.Vaddr)
+ case obj.R_ADDROFF:
+ o = Symaddr(r.Sym) - int64(r.Sym.Sect.Vaddr) + r.Add
+
// r->sym can be null when CALL $(constant) is transformed from absolute PC to relative PC call.
case obj.R_CALL, obj.R_GOTPCREL, obj.R_PCREL:
if Linkmode == LinkExternal && r.Sym != nil && r.Sym.Type != obj.SCONST && (r.Sym.Sect != Ctxt.Cursym.Sect || r.Type == obj.R_GOTPCREL) {
@@ -1599,6 +1602,10 @@
sect.Vaddr = 0
Linklookup(Ctxt, "runtime.rodata", 0).Sect = sect
Linklookup(Ctxt, "runtime.erodata", 0).Sect = sect
+ if !UseRelro() {
+ Linklookup(Ctxt, "runtime.types", 0).Sect = sect
+ Linklookup(Ctxt, "runtime.etypes", 0).Sect = sect
+ }
for ; s != nil && s.Type < obj.STYPERELRO; s = s.Next {
datsize = aligndatsize(datsize, s)
s.Sect = sect
@@ -1631,6 +1638,8 @@
sect.Align = maxalign(s, obj.STYPELINK-1)
datsize = Rnd(datsize, int64(sect.Align))
sect.Vaddr = 0
+ Linklookup(Ctxt, "runtime.types", 0).Sect = sect
+ Linklookup(Ctxt, "runtime.etypes", 0).Sect = sect
for ; s != nil && s.Type < obj.STYPELINK; s = s.Next {
datsize = aligndatsize(datsize, s)
if s.Outer != nil && s.Outer.Sect != nil && s.Outer.Sect != sect {
@@ -1970,10 +1979,12 @@
} else {
rodata = text.Next
}
+ var relrodata *Section
typelink := rodata.Next
if UseRelro() {
// There is another section (.data.rel.ro) when building a shared
// object on elf systems.
+ relrodata = typelink
typelink = typelink.Next
}
itablink := typelink.Next
@@ -2007,6 +2018,11 @@
s.Value = int64(sectSym.Sect.Vaddr + 16)
}
+ types := relrodata
+ if types == nil {
+ types = rodata
+ }
+
xdefine("runtime.text", obj.STEXT, int64(text.Vaddr))
xdefine("runtime.etext", obj.STEXT, int64(text.Vaddr+text.Length))
if HEADTYPE == obj.Hwindows {
@@ -2014,6 +2030,8 @@
}
xdefine("runtime.rodata", obj.SRODATA, int64(rodata.Vaddr))
xdefine("runtime.erodata", obj.SRODATA, int64(rodata.Vaddr+rodata.Length))
+ xdefine("runtime.types", obj.SRODATA, int64(types.Vaddr))
+ xdefine("runtime.etypes", obj.SRODATA, int64(types.Vaddr+types.Length))
xdefine("runtime.typelink", obj.SRODATA, int64(typelink.Vaddr))
xdefine("runtime.etypelink", obj.SRODATA, int64(typelink.Vaddr+typelink.Length))
xdefine("runtime.itablink", obj.SRODATA, int64(itablink.Vaddr))
diff --git a/src/cmd/link/internal/ld/symtab.go b/src/cmd/link/internal/ld/symtab.go
index ae0b17c..678ed38 100644
--- a/src/cmd/link/internal/ld/symtab.go
+++ b/src/cmd/link/internal/ld/symtab.go
@@ -329,6 +329,8 @@
xdefine("runtime.eitablink", obj.SRODATA, 0)
xdefine("runtime.rodata", obj.SRODATA, 0)
xdefine("runtime.erodata", obj.SRODATA, 0)
+ xdefine("runtime.types", obj.SRODATA, 0)
+ xdefine("runtime.etypes", obj.SRODATA, 0)
xdefine("runtime.noptrdata", obj.SNOPTRDATA, 0)
xdefine("runtime.enoptrdata", obj.SNOPTRDATA, 0)
xdefine("runtime.data", obj.SDATA, 0)
@@ -537,6 +539,8 @@
Addaddr(Ctxt, moduledata, Linklookup(Ctxt, "runtime.end", 0))
Addaddr(Ctxt, moduledata, Linklookup(Ctxt, "runtime.gcdata", 0))
Addaddr(Ctxt, moduledata, Linklookup(Ctxt, "runtime.gcbss", 0))
+ Addaddr(Ctxt, moduledata, Linklookup(Ctxt, "runtime.types", 0))
+ Addaddr(Ctxt, moduledata, Linklookup(Ctxt, "runtime.etypes", 0))
// The typelinks slice
Addaddr(Ctxt, moduledata, Linklookup(Ctxt, "runtime.typelink", 0))
adduint(Ctxt, moduledata, uint64(ntypelinks))
diff --git a/src/reflect/export_test.go b/src/reflect/export_test.go
index ddc64b4..037c953 100644
--- a/src/reflect/export_test.go
+++ b/src/reflect/export_test.go
@@ -46,9 +46,11 @@
func TypeLinks() []string {
var r []string
- for _, m := range typelinks() {
- for _, t := range m {
- r = append(r, t.string)
+ sections, offset := typelinks()
+ for i, offs := range offset {
+ rodata := sections[i]
+ for _, off := range offs {
+ r = append(r, rtypeOff(rodata, off).string)
}
}
return r
diff --git a/src/reflect/type.go b/src/reflect/type.go
index 8f13acf..7104fde 100644
--- a/src/reflect/type.go
+++ b/src/reflect/type.go
@@ -1558,30 +1558,48 @@
}
// typelinks is implemented in package runtime.
-// It returns a slice of all the 'typelink' information in the binary,
-// which is to say a slice of known types, sorted by string.
+// It returns a slice of the sections in each module,
+// and a slice of *rtype offsets in each module.
+//
+// The types in each module are sorted by string. That is, the first
+// two linked types of the first module are:
+//
+// d0 := sections[0]
+// t1 := (*rtype)(add(d0, offset[0][0]))
+// t2 := (*rtype)(add(d0, offset[0][1]))
+//
+// and
+//
+// t1.string < t2.string
+//
// Note that strings are not unique identifiers for types:
// there can be more than one with a given string.
// Only types we might want to look up are included:
// pointers, channels, maps, slices, and arrays.
-func typelinks() [][]*rtype
+func typelinks() (sections []unsafe.Pointer, offset [][]int32)
+
+func rtypeOff(section unsafe.Pointer, off int32) *rtype {
+ return (*rtype)(add(section, uintptr(off)))
+}
// typesByString returns the subslice of typelinks() whose elements have
// the given string representation.
// It may be empty (no known types with that string) or may have
// multiple elements (multiple types with that string).
func typesByString(s string) []*rtype {
- typs := typelinks()
+ sections, offset := typelinks()
var ret []*rtype
- for _, typ := range typs {
+ for offsI, offs := range offset {
+ section := sections[offsI]
+
// We are looking for the first index i where the string becomes >= s.
// This is a copy of sort.Search, with f(h) replaced by (*typ[h].string >= s).
- i, j := 0, len(typ)
+ i, j := 0, len(offs)
for i < j {
h := i + (j-i)/2 // avoid overflow when computing h
// i ≤ h < j
- if !(typ[h].string >= s) {
+ if !(rtypeOff(section, offs[h]).string >= s) {
i = h + 1 // preserves f(i-1) == false
} else {
j = h // preserves f(j) == true
@@ -1592,17 +1610,12 @@
// Having found the first, linear scan forward to find the last.
// We could do a second binary search, but the caller is going
// to do a linear scan anyway.
- j = i
- for j < len(typ) && typ[j].string == s {
- j++
- }
-
- if j > i {
- if ret == nil {
- ret = typ[i:j:j]
- } else {
- ret = append(ret, typ[i:j]...)
+ for j := i; j < len(offs); j++ {
+ typ := rtypeOff(section, offs[j])
+ if typ.string != s {
+ break
}
+ ret = append(ret, typ)
}
}
return ret
diff --git a/src/runtime/runtime1.go b/src/runtime/runtime1.go
index 95bebac..e195656 100644
--- a/src/runtime/runtime1.go
+++ b/src/runtime/runtime1.go
@@ -477,10 +477,12 @@
}
//go:linkname reflect_typelinks reflect.typelinks
-func reflect_typelinks() [][]*_type {
- ret := [][]*_type{firstmoduledata.typelinks}
+func reflect_typelinks() ([]unsafe.Pointer, [][]int32) {
+ sections := []unsafe.Pointer{unsafe.Pointer(firstmoduledata.types)}
+ ret := [][]int32{firstmoduledata.typelinks}
for datap := firstmoduledata.next; datap != nil; datap = datap.next {
+ sections = append(sections, unsafe.Pointer(datap.types))
ret = append(ret, datap.typelinks)
}
- return ret
+ return sections, ret
}
diff --git a/src/runtime/symtab.go b/src/runtime/symtab.go
index 158bdce..8c70f22 100644
--- a/src/runtime/symtab.go
+++ b/src/runtime/symtab.go
@@ -127,8 +127,9 @@
bss, ebss uintptr
noptrbss, enoptrbss uintptr
end, gcdata, gcbss uintptr
+ types, etypes uintptr
- typelinks []*_type
+ typelinks []int32 // offsets from types
itablinks []*itab
modulename string
