cmd/internal/obj, cmd/link, runtime: handle TLS more like a platform linker on ppc64

On ppc64x, the thread pointer, held in R13, points 0x7000 bytes past where
thread-local storage begins (presumably to maximize the amount of storage that
can be accessed with a 16-bit signed displacement). The relocations used to
indicate thread-local storage to the platform linker account for this, so to be
able to support external linking we need to change things so the linker applies
this offset instead of the runtime assembly.

Change-Id: I2556c249ab2d802cae62c44b2b4c5b44787d7059
Reviewed-on: https://go-review.googlesource.com/14233
Reviewed-by: Russ Cox <rsc@golang.org>
Reviewed-by: Austin Clements <austin@google.com>

src/cmd/internal/obj/link.go

@@ -451,6 +451,14 @@ const (
 	// the thread local base and the thread local variable defined by the
 	// referenced (thread local) symbol from the GOT.
 	R_ARM64_TLS_IE
+
+	// PPC64.
+
+	// R_POWER_TLS_LE is used to implement the "local exec" model for tls
+	// access. It resolves to the offset of the thread-local symbol from the
+	// thread pointer (R13) and inserts this value into the low 16 bits of an
+	// instruction word.
+	R_POWER_TLS_LE
 )
 
 type Auto struct {

src/cmd/internal/obj/ppc64/a.out.go

@@ -220,6 +220,7 @@ const (
 	C_ANY
 	C_GOK
 	C_ADDR
+	C_TLS_LE
 	C_TEXTSIZE
 
 	C_NCLASS /* must be the last */

src/cmd/internal/obj/ppc64/anames9.go

@@ -35,6 +35,7 @@ var cnames9 = []string{
 	"ANY",
 	"GOK",
 	"ADDR",
+	"TLS_LE",
 	"TEXTSIZE",
 	"NCLASS",
 }

src/cmd/internal/obj/ppc64/asm9.go

@@ -245,6 +245,8 @@ var optab = []Optab{
 	{AMOVBZ, C_ADDR, C_NONE, C_NONE, C_REG, 75, 8, 0},
 	{AMOVB, C_ADDR, C_NONE, C_NONE, C_REG, 76, 12, 0},
 
+	{AMOVD, C_TLS_LE, C_NONE, C_NONE, C_REG, 79, 4, 0},
+
 	/* load constant */
 	{AMOVD, C_SECON, C_NONE, C_NONE, C_REG, 3, 4, REGSB},
 	{AMOVD, C_SACON, C_NONE, C_NONE, C_REG, 3, 4, REGSP},
@@ -583,6 +585,9 @@ func aclass(ctxt *obj.Link, a *obj.Addr) int {
 			}
 			ctxt.Instoffset = a.Offset
 			if a.Sym != nil { // use relocation
+				if a.Sym.Type == obj.STLSBSS {
+					return C_TLS_LE
+				}
 				return C_ADDR
 			}
 			return C_LEXT
@@ -2396,6 +2401,17 @@ func asmout(ctxt *obj.Link, p *obj.Prog, o *Optab, out []uint32) {
 
 		//if(dlm) reloc(&p->from, p->pc, 1);
 
+	case 79:
+		if p.From.Offset != 0 {
+			ctxt.Diag("invalid offset against tls var %v", p)
+		}
+		o1 = AOP_IRR(OP_ADDI, uint32(p.To.Reg), REGZERO, 0)
+		rel := obj.Addrel(ctxt.Cursym)
+		rel.Off = int32(ctxt.Pc)
+		rel.Siz = 4
+		rel.Sym = p.From.Sym
+		rel.Type = obj.R_POWER_TLS_LE
+
 	}
 
 	out[0] = o1

src/cmd/link/internal/ppc64/asm.go

@@ -408,6 +408,18 @@ func archreloc(r *ld.Reloc, s *ld.LSym, val *int64) int {
 		*val = ld.Symaddr(r.Sym) + r.Add - symtoc(s)
 
 		return 0
+
+	case obj.R_POWER_TLS_LE:
+		// The thread pointer points 0x7000 bytes after the start of the the
+		// thread local storage area as documented in section "3.7.2 TLS
+		// Runtime Handling" of "Power Architecture 64-Bit ELF V2 ABI
+		// Specification".
+		v := r.Sym.Value - 0x7000
+		if int64(int16(v)) != v {
+			ld.Diag("TLS offset out of range %d", v)
+		}
+		*val = (*val &^ 0xffff) | (v & 0xffff)
+		return 0
 	}
 
 	return -1

src/runtime/tls_ppc64x.s

@@ -26,16 +26,8 @@ TEXT runtime·save_g(SB),NOSPLIT|NOFRAME,$0-0
 	MOVB	runtime·iscgo(SB), R31
 	CMP	R31, $0
 	BEQ	nocgo
-
-	// $runtime.tlsg(SB) is a special linker symbol.
-	// It is the offset from the start of TLS to our
-	// thread-local storage for g.
-	MOVD	$runtime·tls_g(SB), R31
+	MOVD	runtime·tls_g(SB), R31
 	ADD	R13, R31
-	// The actual TLS base is 0x7000 below R13
-	SUB	$0x7000, R31
-
-	// Store g in TLS
 	MOVD	g, 0(R31)
 
 nocgo:
@@ -51,11 +43,8 @@ nocgo:
 //
 // NOTE: _cgo_topofstack assumes this only clobbers g (R30), and R31.
 TEXT runtime·load_g(SB),NOSPLIT|NOFRAME,$0-0
-	MOVD	$runtime·tls_g(SB), R31
-	// R13 is the C ABI TLS base pointer + 0x7000
+	MOVD	runtime·tls_g(SB), R31
 	ADD	R13, R31
-	SUB	$0x7000, R31
-
 	MOVD	0(R31), g
 	RET