runtime/cgo: take 2

This is a second attempt at submitting
https://golang.org/cl/3420043

A Mercurial problem lost the new files
in that submit.

TBR=r
CC=golang-dev
https://golang.org/cl/3511043

src/pkg/runtime/cgo/386.S

+// Copyright 2009 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.
+
+/*
+ * Apple still insists on underscore prefixes for C function names.
+ */
+#if defined(__APPLE__) || defined(_WIN32)
+#define EXT(s) _##s
+#else
+#define EXT(s) s
+#endif
+
+/*
+ * void crosscall_386(void (*fn)(void))
+ *
+ * Calling into the 8c tool chain, where all registers are caller save.
+ * Called from standard x86 ABI, where %ebp, %ebx, %esi,
+ * and %edi are callee-save, so they must be saved explicitly.
+ */
+.globl EXT(crosscall_386)
+EXT(crosscall_386):
+	pushl %ebp
+	movl %esp, %ebp
+	pushl %ebx
+	pushl %esi
+	pushl %edi
+
+	movl 8(%ebp), %eax	/* fn */
+	call *%eax
+
+	popl %edi
+	popl %esi
+	popl %ebx
+	popl %ebp
+	ret
+
+/*
+ * void crosscall2(void (*fn)(void*, int32), void*, int32)
+ *
+ * Save registers and call fn with two arguments.
+ */
+.globl EXT(crosscall2)
+EXT(crosscall2):
+	pushl %ebp
+	movl %esp, %ebp
+	pushl %ebx
+	pushl %esi
+	pushl %edi
+
+	pushl 16(%ebp)
+	pushl 12(%ebp)
+	mov 8(%ebp), %eax
+	call *%eax
+	addl $8,%esp
+
+	popl %edi
+	popl %esi
+	popl %ebx
+	popl %ebp
+	ret
+
+.globl EXT(__stack_chk_fail_local)
+EXT(__stack_chk_fail_local):
+1:
+	jmp 1b
+

src/pkg/runtime/cgo/Makefile

+# Copyright 2010 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.
+
+include ../../../Make.inc
+
+TARG=runtime/cgo
+
+GOFILES=\
+	cgo.go\
+
+# Unwarranted chumminess with Make.pkg's cgo rules.
+# Do not try this at home.
+GCC_OFILES=\
+	$(GOARCH).o\
+	$(GOOS)_$(GOARCH).o\
+	util.o\
+
+OFILES=\
+	iscgo.$O\
+	_cgo_import.$O\
+	$(GCC_OFILES)\
+
+CGO_LDFLAGS=-lpthread
+
+include ../../../Make.pkg
+
+$(GOARCH).o: $(GOARCH).S
+	$(HOST_CC) $(_CGO_CFLAGS_$(GOARCH)) -g -O2 -fPIC -o $@ -c $^
+
+$(GOOS)_$(GOARCH).o: $(GOOS)_$(GOARCH).c
+	$(HOST_CC) $(_CGO_CFLAGS_$(GOARCH)) -g -O2 -fPIC -o $@ -c $^
+
+%.o: %.c
+	$(HOST_CC) $(_CGO_CFLAGS_$(GOARCH)) -g -O2 -fPIC -o $@ -c $^

src/pkg/runtime/cgo/amd64.S

+// Copyright 2009 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.
+
+/*
+ * Apple still insists on underscore prefixes for C function names.
+ */
+#if defined(__APPLE__) || defined(_WIN32)
+#define EXT(s) _##s
+#else
+#define EXT(s) s
+#endif
+
+/*
+ * void crosscall_amd64(void (*fn)(void))
+ *
+ * Calling into the 6c tool chain, where all registers are caller save.
+ * Called from standard x86-64 ABI, where %rbx, %rbp, %r12-%r15
+ * are callee-save so they must be saved explicitly.
+ * The standard x86-64 ABI passes the three arguments m, g, fn
+ * in %rdi, %rsi, %rdx.
+ *
+ * Also need to set %r15 to g and %r14 to m (see ../pkg/runtime/mkasmh.sh)
+ * during the call.
+ */
+.globl EXT(crosscall_amd64)
+EXT(crosscall_amd64):
+	pushq %rbx
+	pushq %rbp
+	pushq %r12
+	pushq %r13
+	pushq %r14
+	pushq %r15
+
+	call *%rdi	/* fn */
+
+	popq %r15
+	popq %r14
+	popq %r13
+	popq %r12
+	popq %rbp
+	popq %rbx
+	ret
+
+/*
+ * void crosscall2(void (*fn)(void*, int32), void *arg, int32 argsize)
+ *
+ * Save registers and call fn with two arguments.  fn is a Go function
+ * which takes parameters on the stack rather than in registers.
+ */
+.globl EXT(crosscall2)
+EXT(crosscall2):
+	subq  $0x58, %rsp	/* keeps stack pointer 32-byte aligned */
+	movq  %rbx, 0x10(%rsp)
+	movq  %rbp, 0x18(%rsp)
+	movq  %r12, 0x20(%rsp)
+	movq  %r13, 0x28(%rsp)
+	movq  %r14, 0x30(%rsp)
+	movq  %r15, 0x38(%rsp)
+
+	movq  %rsi, 0(%rsp)	/* arg */
+	movq  %rdx, 8(%rsp)	/* argsize (includes padding) */
+
+	call *%rdi	/* fn */
+
+	movq  0x10(%rsp), %rbx
+	movq  0x18(%rsp), %rbp
+	movq  0x20(%rsp), %r12
+	movq  0x28(%rsp), %r13
+	movq  0x30(%rsp), %r14
+	movq  0x38(%rsp), %r15
+	addq  $0x58, %rsp
+	ret

src/pkg/runtime/cgo/cgo.go

+// Copyright 2010 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.
+
+/*
+Package cgo contains runtime support for code generated
+by the cgo tool.  See the documentation for the cgo command
+for details on using cgo.
+*/
+package cgo

src/pkg/runtime/cgo/darwin_386.c

+// Copyright 2009 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.
+
+#include <pthread.h>
+#include "libcgo.h"
+
+static void* threadentry(void*);
+static pthread_key_t k1, k2;
+
+static void
+inittls(void)
+{
+	uint32 x, y;
+	pthread_key_t tofree[16], k;
+	int i, ntofree;
+	int havek1, havek2;
+
+	/*
+	 * Allocate thread-local storage slots for m, g.
+	 * The key numbers start at 0x100, and we expect to be
+	 * one of the early calls to pthread_key_create, so we
+	 * should be able to get pretty low numbers.
+	 *
+	 * In Darwin/386 pthreads, %gs points at the thread
+	 * structure, and each key is an index into the thread-local
+	 * storage array that begins at offset 0x48 within in that structure.
+	 * It may happen that we are not quite the first function to try
+	 * to allocate thread-local storage keys, so instead of depending
+	 * on getting 0x100 and 0x101, we try for 0x108 and 0x109,
+	 * allocating keys until we get the ones we want and then freeing
+	 * the ones we didn't want.
+	 *
+	 * Thus the final offsets to use in %gs references are
+	 * 0x48+4*0x108 = 0x468 and 0x48+4*0x109 = 0x46c.
+	 *
+	 * The linker and runtime hard-code these constant offsets
+	 * from %gs where we expect to find m and g.  The code
+	 * below verifies that the constants are correct once it has
+	 * obtained the keys.  Known to ../cmd/8l/obj.c:/468
+	 * and to ../pkg/runtime/darwin/386/sys.s:/468
+	 *
+	 * This is truly disgusting and a bit fragile, but taking care
+	 * of it here protects the rest of the system from damage.
+	 * The alternative would be to use a global variable that
+	 * held the offset and refer to that variable each time we
+	 * need a %gs variable (m or g).  That approach would
+	 * require an extra instruction and memory reference in
+	 * every stack growth prolog and would also require
+	 * rewriting the code that 8c generates for extern registers.
+	 */
+	havek1 = 0;
+	havek2 = 0;
+	ntofree = 0;
+	while(!havek1 || !havek2) {
+		if(pthread_key_create(&k, nil) < 0) {
+			fprintf(stderr, "libcgo: pthread_key_create failed\n");
+			abort();
+		}
+		if(k == 0x108) {
+			havek1 = 1;
+			k1 = k;
+			continue;
+		}
+		if(k == 0x109) {
+			havek2 = 1;
+			k2 = k;
+			continue;
+		}
+		if(ntofree >= nelem(tofree)) {
+			fprintf(stderr, "libcgo: could not obtain pthread_keys\n");
+			fprintf(stderr, "\twanted 0x108 and 0x109\n");
+			fprintf(stderr, "\tgot");
+			for(i=0; i<ntofree; i++)
+				fprintf(stderr, " %#lx", tofree[i]);
+			fprintf(stderr, "\n");
+			abort();
+		}
+		tofree[ntofree++] = k;
+	}
+
+	for(i=0; i<ntofree; i++)
+		pthread_key_delete(tofree[i]);
+
+	/*
+	 * We got the keys we wanted.  Make sure that we observe
+	 * updates to k1 at 0x468, to verify that the TLS array
+	 * offset from %gs hasn't changed.
+	 */
+	pthread_setspecific(k1, (void*)0x12345678);
+	asm volatile("movl %%gs:0x468, %0" : "=r"(x));
+
+	pthread_setspecific(k1, (void*)0x87654321);
+	asm volatile("movl %%gs:0x468, %0" : "=r"(y));
+
+	if(x != 0x12345678 || y != 0x87654321) {
+		printf("libcgo: thread-local storage %#lx not at %%gs:0x468 - x=%#x y=%#x\n", k1, x, y);
+		abort();
+	}
+}
+
+static void
+xinitcgo(void)
+{
+	inittls();
+}
+
+void (*initcgo)(void) = xinitcgo;
+
+void
+libcgo_sys_thread_start(ThreadStart *ts)
+{
+	pthread_attr_t attr;
+	pthread_t p;
+	size_t size;
+
+	pthread_attr_init(&attr);
+	pthread_attr_getstacksize(&attr, &size);
+	ts->g->stackguard = size;
+	pthread_create(&p, &attr, threadentry, ts);
+}
+
+static void*
+threadentry(void *v)
+{
+	ThreadStart ts;
+
+	ts = *(ThreadStart*)v;
+	free(v);
+
+	ts.g->stackbase = (uintptr)&ts;
+
+	/*
+	 * libcgo_sys_thread_start set stackguard to stack size;
+	 * change to actual guard pointer.
+	 */
+	ts.g->stackguard = (uintptr)&ts - ts.g->stackguard + 4096;
+
+	pthread_setspecific(k1, (void*)ts.g);
+	pthread_setspecific(k2, (void*)ts.m);
+
+	crosscall_386(ts.fn);
+	return nil;
+}

src/pkg/runtime/cgo/darwin_amd64.c

+// Copyright 2009 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.
+
+#include <pthread.h>
+#include "libcgo.h"
+
+static void* threadentry(void*);
+static pthread_key_t k1, k2;
+
+static void
+inittls(void)
+{
+	uint64 x, y;
+	pthread_key_t tofree[16], k;
+	int i, ntofree;
+	int havek1, havek2;
+
+	/*
+	 * Same logic, code as darwin_386.c:/inittls, except that words
+	 * are 8 bytes long now, and the thread-local storage starts at 0x60.
+	 * So the offsets are
+	 * 0x60+8*0x108 = 0x8a0 and 0x60+8*0x109 = 0x8a8.
+	 *
+	 * The linker and runtime hard-code these constant offsets
+	 * from %gs where we expect to find m and g.  The code
+	 * below verifies that the constants are correct once it has
+	 * obtained the keys.  Known to ../cmd/6l/obj.c:/8a0
+	 * and to ../pkg/runtime/darwin/amd64/sys.s:/8a0
+	 *
+	 * As disgusting as on the 386; same justification.
+	 */
+	havek1 = 0;
+	havek2 = 0;
+	ntofree = 0;
+	while(!havek1 || !havek2) {
+		if(pthread_key_create(&k, nil) < 0) {
+			fprintf(stderr, "libcgo: pthread_key_create failed\n");
+			abort();
+		}
+		if(k == 0x108) {
+			havek1 = 1;
+			k1 = k;
+			continue;
+		}
+		if(k == 0x109) {
+			havek2 = 1;
+			k2 = k;
+			continue;
+		}
+		if(ntofree >= nelem(tofree)) {
+			fprintf(stderr, "libcgo: could not obtain pthread_keys\n");
+			fprintf(stderr, "\twanted 0x108 and 0x109\n");
+			fprintf(stderr, "\tgot");
+			for(i=0; i<ntofree; i++)
+				fprintf(stderr, " %#x", (unsigned)tofree[i]);
+			fprintf(stderr, "\n");
+			abort();
+		}
+		tofree[ntofree++] = k;
+	}
+
+	for(i=0; i<ntofree; i++)
+		pthread_key_delete(tofree[i]);
+
+	/*
+	 * We got the keys we wanted.  Make sure that we observe
+	 * updates to k1 at 0x8a0, to verify that the TLS array
+	 * offset from %gs hasn't changed.
+	 */
+	pthread_setspecific(k1, (void*)0x123456789abcdef0ULL);
+	asm volatile("movq %%gs:0x8a0, %0" : "=r"(x));
+
+	pthread_setspecific(k2, (void*)0x0fedcba987654321);
+	asm volatile("movq %%gs:0x8a8, %0" : "=r"(y));
+
+	if(x != 0x123456789abcdef0ULL || y != 0x0fedcba987654321) {
+		printf("libcgo: thread-local storage %#x not at %%gs:0x8a0 - x=%#llx y=%#llx\n", (unsigned)k1, x, y);
+		abort();
+	}
+}
+
+void
+xinitcgo(void)
+{
+	inittls();
+}
+
+void (*initcgo) = xinitcgo;
+
+void
+libcgo_sys_thread_start(ThreadStart *ts)
+{
+	pthread_attr_t attr;
+	pthread_t p;
+	size_t size;
+
+	pthread_attr_init(&attr);
+	pthread_attr_getstacksize(&attr, &size);
+	ts->g->stackguard = size;
+	pthread_create(&p, &attr, threadentry, ts);
+}
+
+static void*
+threadentry(void *v)
+{
+	ThreadStart ts;
+
+	ts = *(ThreadStart*)v;
+	free(v);
+
+	ts.g->stackbase = (uintptr)&ts;
+
+	/*
+	 * libcgo_sys_thread_start set stackguard to stack size;
+	 * change to actual guard pointer.
+	 */
+	ts.g->stackguard = (uintptr)&ts - ts.g->stackguard + 4096;
+
+	pthread_setspecific(k1, (void*)ts.g);
+	pthread_setspecific(k2, (void*)ts.m);
+
+	crosscall_amd64(ts.fn);
+	return nil;
+}

src/pkg/runtime/cgo/iscgo.c

+// Copyright 2010 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.
+
+// The runtime package contains an uninitialized definition
+// for runtime·iscgo.  Override it to tell the runtime we're here.
+// There are various function pointers that should be set too,
+// but those depend on dynamic linker magic to get initialized
+// correctly, and sometimes they break.  This variable is a
+// backup: it depends only on old C style static linking rules.
+
+#include "../runtime.h"
+
+bool runtime·iscgo = 1;

src/pkg/runtime/cgo/libcgo.h

+// Copyright 2009 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.
+
+#include <stdint.h>
+#include <stdlib.h>
+#include <stdio.h>
+
+#define nil ((void*)0)
+#define nelem(x) (sizeof(x)/sizeof((x)[0]))
+
+typedef uint32_t uint32;
+typedef uint64_t uint64;
+typedef uintptr_t uintptr;
+
+/*
+ * The beginning of the per-goroutine structure,
+ * as defined in ../pkg/runtime/runtime.h.
+ * Just enough to edit these two fields.
+ */
+typedef struct G G;
+struct G
+{
+	uintptr stackguard;
+	uintptr stackbase;
+};
+
+/*
+ * Arguments to the libcgo_thread_start call.
+ * Also known to ../pkg/runtime/runtime.h.
+ */
+typedef struct ThreadStart ThreadStart;
+struct ThreadStart
+{
+	uintptr m;
+	G *g;
+	void (*fn)(void);
+};
+
+/*
+ * Called by 5c/6c/8c world.
+ * Makes a local copy of the ThreadStart and
+ * calls libcgo_sys_thread_start(ts).
+ */
+void (*libcgo_thread_start)(ThreadStart *ts);
+
+/*
+ * Creates the new operating system thread (OS, arch dependent).
+ */
+void libcgo_sys_thread_start(ThreadStart *ts);
+
+/*
+ * Call fn in the 6c world.
+ */
+void crosscall_amd64(void (*fn)(void));
+
+/*
+ * Call fn in the 8c world.
+ */
+void crosscall_386(void (*fn)(void));

src/pkg/runtime/cgo/linux_386.c

+// Copyright 2009 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.
+
+#include <pthread.h>
+#include <string.h>
+#include "libcgo.h"
+
+static void *threadentry(void*);
+
+static void
+xinitcgo(void)
+{
+}
+
+void (*initcgo) = xinitcgo;
+
+void
+libcgo_sys_thread_start(ThreadStart *ts)
+{
+	pthread_attr_t attr;
+	pthread_t p;
+	size_t size;
+
+	// Not sure why the memset is necessary here,
+	// but without it, we get a bogus stack size
+	// out of pthread_attr_getstacksize.  C'est la Linux.
+	memset(&attr, 0, sizeof attr);
+	pthread_attr_init(&attr);
+	size = 0;
+	pthread_attr_getstacksize(&attr, &size);
+	ts->g->stackguard = size;
+	pthread_create(&p, &attr, threadentry, ts);
+}
+
+static void*
+threadentry(void *v)
+{
+	ThreadStart ts;
+
+	ts = *(ThreadStart*)v;
+	free(v);
+
+	ts.g->stackbase = (uintptr)&ts;
+
+	/*
+	 * libcgo_sys_thread_start set stackguard to stack size;
+	 * change to actual guard pointer.
+	 */
+	ts.g->stackguard = (uintptr)&ts - ts.g->stackguard + 4096;
+
+	/*
+	 * Set specific keys.  On Linux/ELF, the thread local storage
+	 * is just before %gs:0.  Our dynamic 8.out's reserve 8 bytes
+	 * for the two words g and m at %gs:-8 and %gs:-4.
+	 * Xen requires us to access those words indirect from %gs:0
+	 * which points at itself.
+	 */
+	asm volatile (
+		"movl %%gs:0, %%eax\n"		// MOVL 0(GS), tmp
+		"movl %0, -8(%%eax)\n"	// MOVL g, -8(GS)
+		"movl %1, -4(%%eax)\n"	// MOVL m, -4(GS)
+		:: "r"(ts.g), "r"(ts.m) : "%eax"
+	);
+
+	crosscall_386(ts.fn);
+	return nil;
+}

src/pkg/runtime/cgo/linux_amd64.c

+// Copyright 2009 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.
+
+#include <pthread.h>
+#include "libcgo.h"
+
+static void* threadentry(void*);
+
+void
+xinitcgo(void)
+{
+}
+
+void (*initcgo)(void) = xinitcgo;
+
+void
+libcgo_sys_thread_start(ThreadStart *ts)
+{
+	pthread_attr_t attr;
+	pthread_t p;
+	size_t size;
+
+	pthread_attr_init(&attr);
+	pthread_attr_getstacksize(&attr, &size);
+	ts->g->stackguard = size;
+	pthread_create(&p, &attr, threadentry, ts);
+}
+
+static void*
+threadentry(void *v)
+{
+	ThreadStart ts;
+
+	ts = *(ThreadStart*)v;
+	free(v);
+
+	ts.g->stackbase = (uintptr)&ts;
+
+	/*
+	 * libcgo_sys_thread_start set stackguard to stack size;
+	 * change to actual guard pointer.
+	 */
+	ts.g->stackguard = (uintptr)&ts - ts.g->stackguard + 4096;
+
+	/*
+	 * Set specific keys.  On Linux/ELF, the thread local storage
+	 * is just before %fs:0.  Our dynamic 6.out's reserve 16 bytes
+	 * for the two words g and m at %fs:-16 and %fs:-8.
+	 */
+	asm volatile (
+		"movq %0, %%fs:-16\n"	// MOVL g, -16(FS)
+		"movq %1, %%fs:-8\n"	// MOVL m, -8(FS)
+		:: "r"(ts.g), "r"(ts.m)
+	);
+	crosscall_amd64(ts.fn);
+	return nil;
+}

src/pkg/runtime/cgo/util.c

+// Copyright 2009 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.
+
+#include "libcgo.h"
+
+/* Stub for calling malloc from Go */
+static void
+x_cgo_malloc(void *p)
+{
+	struct a {
+		long long n;
+		void *ret;
+	} *a = p;
+
+	a->ret = malloc(a->n);
+}
+
+void (*_cgo_malloc)(void*) = x_cgo_malloc;
+
+/* Stub for calling from Go */
+static void
+x_cgo_free(void *p)
+{
+	struct a {
+		void *arg;
+	} *a = p;
+
+	free(a->arg);
+}
+
+void (*_cgo_free)(void*) = x_cgo_free;
+
+/* Stub for creating a new thread */
+static void
+xlibcgo_thread_start(ThreadStart *arg)
+{
+	ThreadStart *ts;
+
+	/* Make our own copy that can persist after we return. */
+	ts = malloc(sizeof *ts);
+	if(ts == nil) {
+		fprintf(stderr, "libcgo: out of memory in thread_start\n");
+		abort();
+	}
+	*ts = *arg;
+
+	libcgo_sys_thread_start(ts);	/* OS-dependent half */
+}
+
+void (*libcgo_thread_start)(ThreadStart*) = xlibcgo_thread_start;