From 6e86003651be7feb6da46360d6c411ff1c29b7f5 Mon Sep 17 00:00:00 2001
From: Austin Clements <austin@google.com>
Date: Fri, 31 Oct 2014 11:08:27 -0400
Subject: [PATCH] [dev.power64] 9g: fix under-zeroing in clearfat

All three cases of clearfat were wrong on power64x.

The cases that handle 1032 bytes and up and 32 bytes and up
both use MOVDU (one directly generated in a loop and the other
via duffzero), which leaves the pointer register pointing at
the *last written* address.  The generated code was not
accounting for this, so the byte fill loop was re-zeroing the
last zeroed dword, rather than the bytes following the last
zeroed dword.  Fix this by simply adding an additional 8 byte
offset to the byte zeroing loop.

The case that handled under 32 bytes was also wrong.  It
didn't update the pointer register at all, so the byte zeroing
loop was simply re-zeroing the beginning of region.  Again,
the fix is to add an offset to the byte zeroing loop to
account for this.

LGTM=dave, bradfitz
R=rsc, dave, bradfitz
CC=golang-codereviews
https://golang.org/cl/168870043
---
 src/cmd/9g/ggen.c          | 20 +++++++----
 src/runtime/asm_power64x.s |  2 +-
 test/clearfat.go           | 68 ++++++++++++++++++++++++++++++++++++++
 3 files changed, 82 insertions(+), 8 deletions(-)
 create mode 100644 test/clearfat.go

diff --git a/src/cmd/9g/ggen.c b/src/cmd/9g/ggen.c
index c41d8eb414..7d9cf5050d 100644
--- a/src/cmd/9g/ggen.c
+++ b/src/cmd/9g/ggen.c
@@ -900,7 +900,7 @@ ret:
 void
 clearfat(Node *nl)
 {
-	uint64 w, c, q, t;
+	uint64 w, c, q, t, boff;
 	Node dst, end, r0, *f;
 	Prog *p, *pl;
 
@@ -944,6 +944,8 @@ clearfat(Node *nl)
 		patch(gbranch(ABNE, T, 0), pl);
 
 		regfree(&end);
+		// The loop leaves R3 on the last zeroed dword
+		boff = 8;
 	} else if(q >= 4) {
 		p = gins(ASUB, N, &dst);
 		p->from.type = D_CONST;
@@ -953,17 +955,21 @@ clearfat(Node *nl)
 		afunclit(&p->to, f);
 		// 4 and 128 = magic constants: see ../../runtime/asm_power64x.s
 		p->to.offset = 4*(128-q);
-	} else
-	for(t = 0; t < q; t++) {
-		p = gins(AMOVD, &r0, &dst);
-		p->to.type = D_OREG;
-		p->to.offset = 8*t;
+		// duffzero leaves R3 on the last zeroed dword
+		boff = 8;
+	} else {
+		for(t = 0; t < q; t++) {
+			p = gins(AMOVD, &r0, &dst);
+			p->to.type = D_OREG;
+			p->to.offset = 8*t;
+		}
+		boff = 8*q;
 	}
 
 	for(t = 0; t < c; t++) {
 		p = gins(AMOVB, &r0, &dst);
 		p->to.type = D_OREG;
-		p->to.offset = t;
+		p->to.offset = t+boff;
 	}
 	reg[REGRT1]--;
 }
diff --git a/src/runtime/asm_power64x.s b/src/runtime/asm_power64x.s
index ab2db061c2..2ad3e56e94 100644
--- a/src/runtime/asm_power64x.s
+++ b/src/runtime/asm_power64x.s
@@ -829,7 +829,7 @@ notfound:
 // in ../../cmd/9g/ggen.c:/^clearfat.
 // R0: always zero
 // R3 (aka REGRT1): ptr to memory to be zeroed - 8
-// R3 is updated as a side effect.
+// On return, R3 points to the last zeroed dword.
 TEXT runtime·duffzero(SB), NOSPLIT, $-8-0
 	MOVDU	R0, 8(R3)
 	MOVDU	R0, 8(R3)
diff --git a/test/clearfat.go b/test/clearfat.go
new file mode 100644
index 0000000000..45d539306e
--- /dev/null
+++ b/test/clearfat.go
@@ -0,0 +1,68 @@
+// runoutput
+
+// Copyright 2014 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// Check that {5,6,8,9}g/ggen.c:clearfat is zeroing the entire object.
+
+package main
+
+import (
+	"bytes"
+	"fmt"
+	"strconv"
+	"strings"
+)
+
+const ntest = 1100
+
+func main() {
+	var decls, calls bytes.Buffer
+
+	for i := 1; i <= ntest; i++ {
+		s := strconv.Itoa(i)
+		decls.WriteString(strings.Replace(decl, "$", s, -1))
+		calls.WriteString(strings.Replace("poison$()\n\tclearfat$()\n\t", "$", s, -1))
+	}
+
+	program = strings.Replace(program, "$DECLS", decls.String(), 1)
+	program = strings.Replace(program, "$CALLS", calls.String(), 1)
+	fmt.Print(program)
+}
+
+var program = `package main
+
+var count int
+
+$DECLS
+
+func main() {
+	$CALLS
+	if count != 0 {
+		println("failed", count, "case(s)")
+	}
+}
+`
+
+const decl = `
+func poison$() {
+	// Grow and poison the stack space that will be used by clearfat$
+	var t [2*$]byte
+	for i := range t {
+		t[i] = 0xff
+	}
+}
+
+func clearfat$() {
+	var t [$]byte
+
+	for _, x := range t {
+		if x != 0 {
+//			println("clearfat$: index", i, "expected 0, got", x)
+			count++
+			break
+		}
+	}
+}
+`
-- 
2.50.0