cmd/cgo, runtime/cgo: support ppc64

This implements support for calls to and from C in the ppc64 C ABI, as
well as supporting functionality such as an entry point from the
dynamic linker.

Change-Id: I68da6df50d5638cb1a3d3fef773fb412d7bf631a
Reviewed-on: https://go-review.googlesource.com/2009
Reviewed-by: Russ Cox <rsc@golang.org>

11 files changed, 619 insertions, 15 deletions

diff --git a/src/runtime/asm_ppc64x.s b/src/runtime/asm_ppc64x.s
index a2aba632e9..a708aa9377 100644
--- a/src/runtime/asm_ppc64x.s
+++ b/src/runtime/asm_ppc64x.s
@@ -29,8 +29,29 @@ TEXT runtime·rt0_go(SB),NOSPLIT,$0
 	MOVD	R3, (g_stack+stack_lo)(g)
 	MOVD	R1, (g_stack+stack_hi)(g)
 
-	// TODO: if there is a _cgo_init, call it.
-	// TODO: add TLS
+	// if there is a _cgo_init, call it using the gcc ABI.
+	MOVD	_cgo_init(SB), R12
+	CMP	R0, R12
+	BEQ	nocgo
+	MOVD	R12, CTR		// r12 = "global function entry point"
+	MOVD	R13, R5			// arg 2: TLS base pointer
+	MOVD	$setg_gcc<>(SB), R4 	// arg 1: setg
+	MOVD	g, R3			// arg 0: G
+	// C functions expect 32 bytes of space on caller stack frame
+	// and a 16-byte aligned R1
+	MOVD	R1, R14			// save current stack
+	SUB	$32, R1			// reserve 32 bytes
+	RLDCR	$0, R1, $~15, R1	// 16-byte align
+	BL	(CTR)			// may clobber R0, R3-R12
+	MOVD	R14, R1			// restore stack
+	XOR	R0, R0			// fix R0
+
+nocgo:
+	// update stackguard after _cgo_init
+	MOVD	(g_stack+stack_lo)(g), R3
+	ADD	$const__StackGuard, R3
+	MOVD	R3, g_stackguard0(g)
+	MOVD	R3, g_stackguard1(g)
 
 	// set the per-goroutine and per-mach "registers"
 	MOVD	$runtime·m0(SB), R3
@@ -71,6 +92,11 @@ TEXT runtime·breakpoint(SB),NOSPLIT,$-8-0
 TEXT runtime·asminit(SB),NOSPLIT,$-8-0
 	RETURN
 
+TEXT _cgo_reginit(SB),NOSPLIT,$-8-0
+	// crosscall_ppc64 and crosscall2 need to reginit, but can't
+	// get at the 'runtime.reginit' symbol.
+	BR	runtime·reginit(SB)
+
 TEXT runtime·reginit(SB),NOSPLIT,$-8-0
 	// set R0 to zero, it's expected by the toolchain
 	XOR R0, R0
@@ -625,26 +651,207 @@ TEXT gosave<>(SB),NOSPLIT,$-8
 // aligned appropriately for the gcc ABI.
 // See cgocall.c for more details.
 TEXT ·asmcgocall(SB),NOSPLIT,$0-16
-	MOVD	R0, 21(R0)
+	MOVD	fn+0(FP), R3
+	MOVD	arg+8(FP), R4
+	BL	asmcgocall<>(SB)
+	RET
+
+TEXT ·asmcgocall_errno(SB),NOSPLIT,$0-24
+	MOVD	fn+0(FP), R3
+	MOVD	arg+8(FP), R4
+	BL	asmcgocall<>(SB)
+	MOVD	R3, ret+16(FP)
+	RET
+
+// asmcgocall common code. fn in R3, arg in R4. returns errno in R3.
+TEXT asmcgocall<>(SB),NOSPLIT,$0-0
+	MOVD	R1, R2		// save original stack pointer
+	MOVD	g, R5
+
+	// Figure out if we need to switch to m->g0 stack.
+	// We get called to create new OS threads too, and those
+	// come in on the m->g0 stack already.
+	MOVD	g_m(g), R6
+	MOVD	m_g0(R6), R6
+	CMP	R6, g
+	BEQ	g0
+	BL	gosave<>(SB)
+	MOVD	R6, g
+	BL	runtime·save_g(SB)
+	MOVD	(g_sched+gobuf_sp)(g), R1
+
+	// Now on a scheduling stack (a pthread-created stack).
+g0:
+	// Save room for two of our pointers, plus 32 bytes of callee
+	// save area that lives on the caller stack.
+	SUB	$48, R1
+	RLDCR	$0, R1, $~15, R1	// 16-byte alignment for gcc ABI
+	MOVD	R5, 40(R1)	// save old g on stack
+	MOVD	(g_stack+stack_hi)(R5), R5
+	SUB	R2, R5
+	MOVD	R5, 32(R1)	// save depth in old g stack (can't just save SP, as stack might be copied during a callback)
+	MOVD	R0, 0(R1)	// clear back chain pointer (TODO can we give it real back trace information?)
+	// This is a "global call", so put the global entry point in r12
+	MOVD	R3, R12
+	MOVD	R12, CTR
+	MOVD	R4, R3		// arg in r3
+	BL	(CTR)
+
+	// C code can clobber R0, so set it back to 0.  F27-F31 are
+	// callee save, so we don't need to recover those.
+	XOR	R0, R0
+	// Restore g, stack pointer.  R3 is errno, so don't touch it
+	MOVD	40(R1), g
+	BL	runtime·save_g(SB)
+	MOVD	(g_stack+stack_hi)(g), R5
+	MOVD	32(R1), R6
+	SUB	R6, R5
+	MOVD	R5, R1
+	RET
 
 // cgocallback(void (*fn)(void*), void *frame, uintptr framesize)
 // Turn the fn into a Go func (by taking its address) and call
 // cgocallback_gofunc.
 TEXT runtime·cgocallback(SB),NOSPLIT,$24-24
-	MOVD	R0, 22(R0)
+	MOVD	$fn+0(FP), R3
+	MOVD	R3, 8(R1)
+	MOVD	frame+8(FP), R3
+	MOVD	R3, 16(R1)
+	MOVD	framesize+16(FP), R3
+	MOVD	R3, 24(R1)
+	MOVD	$runtime·cgocallback_gofunc(SB), R3
+	MOVD	R3, CTR
+	BL	(CTR)
+	RET
 
 // cgocallback_gofunc(FuncVal*, void *frame, uintptr framesize)
 // See cgocall.c for more details.
-TEXT ·cgocallback_gofunc(SB),NOSPLIT,$8-24
-	MOVD	R0, 23(R0)
+TEXT ·cgocallback_gofunc(SB),NOSPLIT,$16-24
+	NO_LOCAL_POINTERS
+
+	// Load m and g from thread-local storage.
+	MOVB	runtime·iscgo(SB), R3
+	CMP	R3, $0
+	BEQ	nocgo
+	BL	runtime·load_g(SB)
+nocgo:
+
+	// If g is nil, Go did not create the current thread.
+	// Call needm to obtain one for temporary use.
+	// In this case, we're running on the thread stack, so there's
+	// lots of space, but the linker doesn't know. Hide the call from
+	// the linker analysis by using an indirect call.
+	CMP	g, $0
+	BNE	havem
+	MOVD	g, savedm-8(SP) // g is zero, so is m.
+	MOVD	$runtime·needm(SB), R3
+	MOVD	R3, CTR
+	BL	(CTR)
+
+	// Set m->sched.sp = SP, so that if a panic happens
+	// during the function we are about to execute, it will
+	// have a valid SP to run on the g0 stack.
+	// The next few lines (after the havem label)
+	// will save this SP onto the stack and then write
+	// the same SP back to m->sched.sp. That seems redundant,
+	// but if an unrecovered panic happens, unwindm will
+	// restore the g->sched.sp from the stack location
+	// and then systemstack will try to use it. If we don't set it here,
+	// that restored SP will be uninitialized (typically 0) and
+	// will not be usable.
+	MOVD	g_m(g), R3
+	MOVD	m_g0(R3), R3
+	MOVD	R1, (g_sched+gobuf_sp)(R3)
+
+havem:
+	MOVD	g_m(g), R8
+	MOVD	R8, savedm-8(SP)
+	// Now there's a valid m, and we're running on its m->g0.
+	// Save current m->g0->sched.sp on stack and then set it to SP.
+	// Save current sp in m->g0->sched.sp in preparation for
+	// switch back to m->curg stack.
+	// NOTE: unwindm knows that the saved g->sched.sp is at 8(R1) aka savedsp-16(SP).
+	MOVD	m_g0(R8), R3
+	MOVD	(g_sched+gobuf_sp)(R3), R4
+	MOVD	R4, savedsp-16(SP)
+	MOVD	R1, (g_sched+gobuf_sp)(R3)
+
+	// Switch to m->curg stack and call runtime.cgocallbackg.
+	// Because we are taking over the execution of m->curg
+	// but *not* resuming what had been running, we need to
+	// save that information (m->curg->sched) so we can restore it.
+	// We can restore m->curg->sched.sp easily, because calling
+	// runtime.cgocallbackg leaves SP unchanged upon return.
+	// To save m->curg->sched.pc, we push it onto the stack.
+	// This has the added benefit that it looks to the traceback
+	// routine like cgocallbackg is going to return to that
+	// PC (because the frame we allocate below has the same
+	// size as cgocallback_gofunc's frame declared above)
+	// so that the traceback will seamlessly trace back into
+	// the earlier calls.
+	//
+	// In the new goroutine, -16(SP) and -8(SP) are unused.
+	MOVD	m_curg(R8), g
+	BL	runtime·save_g(SB)
+	MOVD	(g_sched+gobuf_sp)(g), R4 // prepare stack as R4
+	MOVD	(g_sched+gobuf_pc)(g), R5
+	MOVD	R5, -24(R4)
+	MOVD	$-24(R4), R1
+	BL	runtime·cgocallbackg(SB)
+
+	// Restore g->sched (== m->curg->sched) from saved values.
+	MOVD	0(R1), R5
+	MOVD	R5, (g_sched+gobuf_pc)(g)
+	MOVD	$24(R1), R4
+	MOVD	R4, (g_sched+gobuf_sp)(g)
+
+	// Switch back to m->g0's stack and restore m->g0->sched.sp.
+	// (Unlike m->curg, the g0 goroutine never uses sched.pc,
+	// so we do not have to restore it.)
+	MOVD	g_m(g), R8
+	MOVD	m_g0(R8), g
+	BL	runtime·save_g(SB)
+	MOVD	(g_sched+gobuf_sp)(g), R1
+	MOVD	savedsp-16(SP), R4
+	MOVD	R4, (g_sched+gobuf_sp)(g)
+
+	// If the m on entry was nil, we called needm above to borrow an m
+	// for the duration of the call. Since the call is over, return it with dropm.
+	MOVD	savedm-8(SP), R6
+	CMP	R6, $0
+	BNE	droppedm
+	MOVD	$runtime·dropm(SB), R3
+	MOVD	R3, CTR
+	BL	(CTR)
+droppedm:
+
+	// Done!
+	RET
 
 // void setg(G*); set g. for use by needm.
 TEXT runtime·setg(SB), NOSPLIT, $0-8
-	MOVD	R0, 24(R0)
+	MOVD	gg+0(FP), g
+	// This only happens if iscgo, so jump straight to save_g
+	BL	runtime·save_g(SB)
+	RET
 
-// void setg_gcc(G*); set g called from gcc.
-TEXT setg_gcc<>(SB),NOSPLIT,$0
-	MOVD	R0, 25(R0)
+// void setg_gcc(G*); set g in C TLS.
+// Must obey the gcc calling convention.
+TEXT setg_gcc<>(SB),NOSPLIT,$-8-0
+	// The standard prologue clobbers R31, which is callee-save in
+	// the C ABI, so we have to use $-8-0 and save LR ourselves.
+	MOVD	LR, R4
+	// Also save g and R31, since they're callee-save in C ABI
+	MOVD	R31, R5
+	MOVD	g, R6
+
+	MOVD	R3, g
+	BL	runtime·save_g(SB)
+
+	MOVD	R6, g
+	MOVD	R5, R31
+	MOVD	R4, LR
+	RET
 
 TEXT runtime·getcallerpc(SB),NOSPLIT,$-8-16
 	MOVD	0(R1), R3
@@ -989,8 +1196,21 @@ TEXT runtime·return0(SB), NOSPLIT, $0
 
 // Called from cgo wrappers, this function returns g->m->curg.stack.hi.
 // Must obey the gcc calling convention.
-TEXT _cgo_topofstack(SB),NOSPLIT,$0
-	MOVD	R0, 26(R0)
+TEXT _cgo_topofstack(SB),NOSPLIT,$-8
+	// g (R30) and R31 are callee-save in the C ABI, so save them
+	MOVD	g, R4
+	MOVD	R31, R5
+	MOVD	LR, R6
+
+	BL	runtime·load_g(SB)	// clobbers g (R30), R31
+	MOVD	g_m(g), R3
+	MOVD	m_curg(R3), R3
+	MOVD	(g_stack+stack_hi)(R3), R3
+
+	MOVD	R4, g
+	MOVD	R5, R31
+	MOVD	R6, LR
+	RET
 
 // The top-most function running on a goroutine
 // returns to goexit+PCQuantum.
diff --git a/src/runtime/cgo/asm_ppc64x.s b/src/runtime/cgo/asm_ppc64x.s
new file mode 100644
index 0000000000..0c08a1d6b5
--- /dev/null
+++ b/src/runtime/cgo/asm_ppc64x.s
@@ -0,0 +1,124 @@
+// Copyright 2014 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.
+
+// +build ppc64 ppc64le
+
+#include "textflag.h"
+
+/*
+ * void crosscall2(void (*fn)(void*, int32), void*, int32)
+ * Save registers and call fn with two arguments.
+ * crosscall2 obeys the C ABI; fn obeys the Go ABI.
+ */
+TEXT crosscall2(SB),NOSPLIT,$-8
+	// TODO(austin): ABI v1 (fn is probably a function descriptor)
+
+	// Start with standard C stack frame layout and linkage
+	MOVD	LR, R0
+	MOVD	R0, 16(R1)	// Save LR in caller's frame
+	MOVD	R2, 24(R1)	// Save TOC in caller's frame
+
+	BL	saveregs2<>(SB)
+
+	MOVDU	R1, (-288-3*8)(R1)
+
+	// Initialize Go ABI environment
+	BL	runtime·reginit(SB)
+	BL	runtime·load_g(SB)
+
+	MOVD	R3, CTR
+	MOVD	R4, 8(R1)
+	MOVD	R5, 16(R1)
+	BL	(CTR)
+
+	ADD	$(288+3*8), R1
+
+	BL	restoreregs2<>(SB)
+
+	MOVD	24(R1), R2
+	MOVD	16(R1), R0
+	MOVD	R0, LR
+	RET
+
+TEXT saveregs2<>(SB),NOSPLIT,$-8
+	// O=-288; for R in R{14..31}; do echo "\tMOVD\t$R, $O(R1)"|sed s/R30/g/; ((O+=8)); done; for F in F{14..31}; do echo "\tFMOVD\t$F, $O(R1)"; ((O+=8)); done
+	MOVD	R14, -288(R1)
+	MOVD	R15, -280(R1)
+	MOVD	R16, -272(R1)
+	MOVD	R17, -264(R1)
+	MOVD	R18, -256(R1)
+	MOVD	R19, -248(R1)
+	MOVD	R20, -240(R1)
+	MOVD	R21, -232(R1)
+	MOVD	R22, -224(R1)
+	MOVD	R23, -216(R1)
+	MOVD	R24, -208(R1)
+	MOVD	R25, -200(R1)
+	MOVD	R26, -192(R1)
+	MOVD	R27, -184(R1)
+	MOVD	R28, -176(R1)
+	MOVD	R29, -168(R1)
+	MOVD	g, -160(R1)
+	MOVD	R31, -152(R1)
+	FMOVD	F14, -144(R1)
+	FMOVD	F15, -136(R1)
+	FMOVD	F16, -128(R1)
+	FMOVD	F17, -120(R1)
+	FMOVD	F18, -112(R1)
+	FMOVD	F19, -104(R1)
+	FMOVD	F20, -96(R1)
+	FMOVD	F21, -88(R1)
+	FMOVD	F22, -80(R1)
+	FMOVD	F23, -72(R1)
+	FMOVD	F24, -64(R1)
+	FMOVD	F25, -56(R1)
+	FMOVD	F26, -48(R1)
+	FMOVD	F27, -40(R1)
+	FMOVD	F28, -32(R1)
+	FMOVD	F29, -24(R1)
+	FMOVD	F30, -16(R1)
+	FMOVD	F31, -8(R1)
+
+	RET
+
+TEXT restoreregs2<>(SB),NOSPLIT,$-8
+	// O=-288; for R in R{14..31}; do echo "\tMOVD\t$O(R1), $R"|sed s/R30/g/; ((O+=8)); done; for F in F{14..31}; do echo "\tFMOVD\t$O(R1), $F"; ((O+=8)); done
+	MOVD	-288(R1), R14
+	MOVD	-280(R1), R15
+	MOVD	-272(R1), R16
+	MOVD	-264(R1), R17
+	MOVD	-256(R1), R18
+	MOVD	-248(R1), R19
+	MOVD	-240(R1), R20
+	MOVD	-232(R1), R21
+	MOVD	-224(R1), R22
+	MOVD	-216(R1), R23
+	MOVD	-208(R1), R24
+	MOVD	-200(R1), R25
+	MOVD	-192(R1), R26
+	MOVD	-184(R1), R27
+	MOVD	-176(R1), R28
+	MOVD	-168(R1), R29
+	MOVD	-160(R1), g
+	MOVD	-152(R1), R31
+	FMOVD	-144(R1), F14
+	FMOVD	-136(R1), F15
+	FMOVD	-128(R1), F16
+	FMOVD	-120(R1), F17
+	FMOVD	-112(R1), F18
+	FMOVD	-104(R1), F19
+	FMOVD	-96(R1), F20
+	FMOVD	-88(R1), F21
+	FMOVD	-80(R1), F22
+	FMOVD	-72(R1), F23
+	FMOVD	-64(R1), F24
+	FMOVD	-56(R1), F25
+	FMOVD	-48(R1), F26
+	FMOVD	-40(R1), F27
+	FMOVD	-32(R1), F28
+	FMOVD	-24(R1), F29
+	FMOVD	-16(R1), F30
+	FMOVD	-8(R1), F31
+
+	RET
diff --git a/src/runtime/cgo/gcc_linux_ppc64x.c b/src/runtime/cgo/gcc_linux_ppc64x.c
new file mode 100644
index 0000000000..b1762957a2
--- /dev/null
+++ b/src/runtime/cgo/gcc_linux_ppc64x.c
@@ -0,0 +1,70 @@
+// Copyright 2014 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.
+
+// +build ppc64 ppc64le
+
+#include <pthread.h>
+#include <string.h>
+#include <signal.h>
+#include "libcgo.h"
+
+static void *threadentry(void*);
+
+void (*x_cgo_inittls)(void **tlsg, void **tlsbase);
+static void (*setg_gcc)(void*);
+
+void
+x_cgo_init(G *g, void (*setg)(void*), void **tlsbase)
+{
+	pthread_attr_t attr;
+	size_t size;
+
+	setg_gcc = setg;
+	pthread_attr_init(&attr);
+	pthread_attr_getstacksize(&attr, &size);
+	g->stacklo = (uintptr)&attr - size + 4096;
+	pthread_attr_destroy(&attr);
+}
+
+void
+_cgo_sys_thread_start(ThreadStart *ts)
+{
+	pthread_attr_t attr;
+	sigset_t ign, oset;
+	pthread_t p;
+	size_t size;
+	int err;
+
+	sigfillset(&ign);
+	pthread_sigmask(SIG_SETMASK, &ign, &oset);
+
+	pthread_attr_init(&attr);
+	pthread_attr_getstacksize(&attr, &size);
+	// Leave stacklo=0 and set stackhi=size; mstack will do the rest.
+	ts->g->stackhi = size;
+	err = pthread_create(&p, &attr, threadentry, ts);
+
+	pthread_sigmask(SIG_SETMASK, &oset, nil);
+
+	if (err != 0) {
+		fatalf("pthread_create failed: %s", strerror(err));
+	}
+}
+
+extern void crosscall_ppc64(void (*fn)(void), void *g);
+
+static void*
+threadentry(void *v)
+{
+	ThreadStart ts;
+
+	ts = *(ThreadStart*)v;
+	free(v);
+
+	// Save g for this thread in C TLS
+	setg_gcc((void*)ts.g);
+
+	crosscall_ppc64(ts.fn, (void*)ts.g);
+	return nil;
+}
diff --git a/src/runtime/cgo/gcc_ppc64x.S b/src/runtime/cgo/gcc_ppc64x.S
new file mode 100644
index 0000000000..fc202778c7
--- /dev/null
+++ b/src/runtime/cgo/gcc_ppc64x.S
@@ -0,0 +1,140 @@
+// Copyright 2014 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.
+
+// +build ppc64 ppc64le
+
+/*
+ * Apple still insists on underscore prefixes for C function names.
+ */
+#if defined(__APPLE__)
+#define EXT(s) _##s
+#else
+#define EXT(s) s
+#endif
+
+/*
+ * void crosscall_ppc64(void (*fn)(void), void *g)
+ *
+ * Calling into the 9g tool chain, where all registers are caller save.
+ * Called from standard ppc64 C ABI, where r2, r14-r31, f14-f31 are
+ * callee-save, so they must be saved explicitly.
+ */
+.globl EXT(crosscall_ppc64)
+EXT(crosscall_ppc64):
+	// Start with standard C stack frame layout and linkage
+	mflr	%r0
+	std	%r0, 16(%r1)	// Save LR in caller's frame
+	std	%r2, 24(%r1)	// Save TOC in caller's frame
+	bl	saveregs
+	stdu	%r1, -296(%r1)
+
+	// Set up Go ABI constant registers
+	bl	_cgo_reginit
+
+	// Restore g pointer (r30 in Go ABI, which may have been clobbered by C)
+	mr	%r30, %r4
+
+	// Call fn
+	mtctr	%r3
+	bctrl
+
+	addi	%r1, %r1, 296
+	bl	restoreregs
+	ld	%r2, 24(%r1)
+	ld	%r0, 16(%r1)
+	mtlr	%r0
+	blr
+
+saveregs:
+	// Save callee-save registers
+	// O=-288; for R in %r{14..31}; do echo "\tstd\t$R, $O(%r1)"; ((O+=8)); done; for F in f{14..31}; do echo "\tstfd\t$F, $O(%r1)"; ((O+=8)); done
+	std	%r14, -288(%r1)
+	std	%r15, -280(%r1)
+	std	%r16, -272(%r1)
+	std	%r17, -264(%r1)
+	std	%r18, -256(%r1)
+	std	%r19, -248(%r1)
+	std	%r20, -240(%r1)
+	std	%r21, -232(%r1)
+	std	%r22, -224(%r1)
+	std	%r23, -216(%r1)
+	std	%r24, -208(%r1)
+	std	%r25, -200(%r1)
+	std	%r26, -192(%r1)
+	std	%r27, -184(%r1)
+	std	%r28, -176(%r1)
+	std	%r29, -168(%r1)
+	std	%r30, -160(%r1)
+	std	%r31, -152(%r1)
+	stfd	%f14, -144(%r1)
+	stfd	%f15, -136(%r1)
+	stfd	%f16, -128(%r1)
+	stfd	%f17, -120(%r1)
+	stfd	%f18, -112(%r1)
+	stfd	%f19, -104(%r1)
+	stfd	%f20, -96(%r1)
+	stfd	%f21, -88(%r1)
+	stfd	%f22, -80(%r1)
+	stfd	%f23, -72(%r1)
+	stfd	%f24, -64(%r1)
+	stfd	%f25, -56(%r1)
+	stfd	%f26, -48(%r1)
+	stfd	%f27, -40(%r1)
+	stfd	%f28, -32(%r1)
+	stfd	%f29, -24(%r1)
+	stfd	%f30, -16(%r1)
+	stfd	%f31, -8(%r1)
+
+	blr
+
+restoreregs:
+	// O=-288; for R in %r{14..31}; do echo "\tld\t$R, $O(%r1)"; ((O+=8)); done; for F in %f{14..31}; do echo "\tlfd\t$F, $O(%r1)"; ((O+=8)); done
+	ld	%r14, -288(%r1)
+	ld	%r15, -280(%r1)
+	ld	%r16, -272(%r1)
+	ld	%r17, -264(%r1)
+	ld	%r18, -256(%r1)
+	ld	%r19, -248(%r1)
+	ld	%r20, -240(%r1)
+	ld	%r21, -232(%r1)
+	ld	%r22, -224(%r1)
+	ld	%r23, -216(%r1)
+	ld	%r24, -208(%r1)
+	ld	%r25, -200(%r1)
+	ld	%r26, -192(%r1)
+	ld	%r27, -184(%r1)
+	ld	%r28, -176(%r1)
+	ld	%r29, -168(%r1)
+	ld	%r30, -160(%r1)
+	ld	%r31, -152(%r1)
+	lfd	%f14, -144(%r1)
+	lfd	%f15, -136(%r1)
+	lfd	%f16, -128(%r1)
+	lfd	%f17, -120(%r1)
+	lfd	%f18, -112(%r1)
+	lfd	%f19, -104(%r1)
+	lfd	%f20, -96(%r1)
+	lfd	%f21, -88(%r1)
+	lfd	%f22, -80(%r1)
+	lfd	%f23, -72(%r1)
+	lfd	%f24, -64(%r1)
+	lfd	%f25, -56(%r1)
+	lfd	%f26, -48(%r1)
+	lfd	%f27, -40(%r1)
+	lfd	%f28, -32(%r1)
+	lfd	%f29, -24(%r1)
+	lfd	%f30, -16(%r1)
+	lfd	%f31, -8(%r1)
+
+	blr
+
+.globl EXT(__stack_chk_fail_local)
+EXT(__stack_chk_fail_local):
+1:
+	// TODO(austin)
+	b 1b
+
+#ifdef __ELF__
+.section .note.GNU-stack,"",%progbits
+#endif
diff --git a/src/runtime/cgocall.go b/src/runtime/cgocall.go
index dbeea200d4..96873cc2da 100644
--- a/src/runtime/cgocall.go
+++ b/src/runtime/cgocall.go
@@ -229,6 +229,11 @@ func cgocallbackg1() {
 	case "386":
 		// On 386, stack frame is three words, plus caller PC.
 		cb = (*args)(unsafe.Pointer(sp + 4*ptrSize))
+	case "ppc64", "ppc64le":
+		// On ppc64, stack frame is two words and there's a
+		// saved LR between SP and the stack frame and between
+		// the stack frame and the arguments.
+		cb = (*args)(unsafe.Pointer(sp + 4*ptrSize))
 	}
 
 	// Invoke callback.
@@ -263,6 +268,8 @@ func unwindm(restore *bool) {
 		sched.sp = *(*uintptr)(unsafe.Pointer(sched.sp))
 	case "arm":
 		sched.sp = *(*uintptr)(unsafe.Pointer(sched.sp + 4))
+	case "ppc64", "ppc64le":
+		sched.sp = *(*uintptr)(unsafe.Pointer(sched.sp + 8))
 	}
 	releasem(mp)
 }
diff --git a/src/runtime/crash_cgo_test.go b/src/runtime/crash_cgo_test.go
index 29f90fa36d..7152b93195 100644
--- a/src/runtime/crash_cgo_test.go
+++ b/src/runtime/crash_cgo_test.go
@@ -68,6 +68,11 @@ func TestCgoExternalThreadSIGPROF(t *testing.T) {
 			t.Skipf("no external linking on OS X 10.6")
 		}
 	}
+	if runtime.GOARCH == "ppc64" || runtime.GOARCH == "ppc64le" {
+		// TODO(austin) External linking not implemented on
+		// ppc64 (issue #8912)
+		t.Skipf("no external linking on ppc64")
+	}
 	got := executeTest(t, cgoExternalThreadSIGPROFSource, nil)
 	want := "OK\n"
 	if got != want {
diff --git a/src/runtime/rt0_linux_ppc64.s b/src/runtime/rt0_linux_ppc64.s
index 33bbbbd1bf..33e973db96 100644
--- a/src/runtime/rt0_linux_ppc64.s
+++ b/src/runtime/rt0_linux_ppc64.s
@@ -7,6 +7,12 @@ TEXT _rt0_ppc64_linux(SB),NOSPLIT,$0
 	DWORD $0
 
 TEXT _main<>(SB),NOSPLIT,$-8
+	// In a statically linked binary, the stack contains argc,
+	// argv as argc string pointers followed by a NULL, envv as a
+	// sequence of string pointers followed by a NULL, and auxv.
+	// There is no TLS base pointer.
+	//
+	// TODO(austin): Support ABI v1 dynamic linking entry point
 	MOVD 0(R1), R3 // argc
 	ADD $8, R1, R4 // argv
 	BR main(SB)
diff --git a/src/runtime/rt0_linux_ppc64le.s b/src/runtime/rt0_linux_ppc64le.s
index dda29ab3a0..f5c0af5c71 100644
--- a/src/runtime/rt0_linux_ppc64le.s
+++ b/src/runtime/rt0_linux_ppc64le.s
@@ -4,11 +4,29 @@ TEXT _rt0_ppc64le_linux(SB),NOSPLIT,$0
 	BR _main<>(SB)
 
 TEXT _main<>(SB),NOSPLIT,$-8
-	MOVD 0(R1), R3 // argc
-	ADD $8, R1, R4 // argv
+	// In a statically linked binary, the stack contains argc,
+	// argv as argc string pointers followed by a NULL, envv as a
+	// sequence of string pointers followed by a NULL, and auxv.
+	// There is no TLS base pointer.
+	//
+	// In a dynamically linked binary, r3 contains argc, r4
+	// contains argv, r5 contains envp, r6 contains auxv, and r13
+	// contains the TLS pointer.
+	//
+	// Figure out which case this is by looking at r4: if it's 0,
+	// we're statically linked; otherwise we're dynamically
+	// linked.
+	CMP	R0, R4
+	BNE	dlink
+
+	// Statically linked
+	MOVD	0(R1), R3 // argc
+	ADD	$8, R1, R4 // argv
 	MOVD	$runtime·tls0(SB), R13 // TLS
 	ADD	$0x7000, R13
-	BR main(SB)
+
+dlink:
+	BR	main(SB)
 
 TEXT main(SB),NOSPLIT,$-8
 	MOVD	$runtime·rt0_go(SB), R31
diff --git a/src/runtime/runtime1.go b/src/runtime/runtime1.go
index 495b5f915a..e6510a8aa3 100644
--- a/src/runtime/runtime1.go
+++ b/src/runtime/runtime1.go
@@ -77,6 +77,9 @@ func goargs() {
 }
 
 func goenvs_unix() {
+	// TODO(austin): ppc64 in dynamic linking mode doesn't
+	// guarantee env[] will immediately follow argv.  Might cause
+	// problems.
 	n := int32(0)
 	for argv_index(argv, argc+1+n) != nil {
 		n++
diff --git a/src/runtime/sys_linux_ppc64x.s b/src/runtime/sys_linux_ppc64x.s
index 4a4f440c53..b9d8be11c5 100644
--- a/src/runtime/sys_linux_ppc64x.s
+++ b/src/runtime/sys_linux_ppc64x.s
@@ -193,6 +193,13 @@ TEXT runtime·_sigtramp(SB),NOSPLIT,$64
 	// initialize essential registers (just in case)
 	BL	runtime·reginit(SB)
 
+	// this might be called in external code context,
+	// where g is not set.
+	MOVB	runtime·iscgo(SB), R6
+	CMP 	R6, $0
+	BEQ	2(PC)
+	BL	runtime·load_g(SB)
+
 	// check that g exists
 	CMP	g, $0
 	BNE	6(PC)
diff --git a/src/runtime/tls_ppc64x.s b/src/runtime/tls_ppc64x.s
index fa1f9ac6f0..fc1718f508 100644
--- a/src/runtime/tls_ppc64x.s
+++ b/src/runtime/tls_ppc64x.s
@@ -20,6 +20,8 @@
 // ppc64 code that will overwrite this register.
 //
 // If !iscgo, this is a no-op.
+//
+// NOTE: setg_gcc<> assume this clobbers only R31.
 TEXT runtime·save_g(SB),NOSPLIT,$-8-0
 	MOVB	runtime·iscgo(SB), R31
 	CMP	R31, $0
@@ -46,6 +48,8 @@ nocgo:
 // This is never called directly from C code (it doesn't have to
 // follow the C ABI), but it may be called from a C context, where the
 // usual Go registers aren't set up.
+//
+// NOTE: _cgo_topofstack assumes this only clobbers g (R30), and R31.
 TEXT runtime·load_g(SB),NOSPLIT,$-8-0
 	MOVD	$runtime·tlsg(SB), R31
 	// R13 is the C ABI TLS base pointer + 0x7000