runtime/internal/atomic: add mips operators for And/Or

These primitives will be used by the new And/Or sync/atomic apis. Implemented for mips/mipsle and mips64/mips64le. For golang#61395 Change-Id: Icc604a2b5cdfe72646d47d3c6a0bb49a0fd0d353 GitHub-Last-Rev: 95dca2a GitHub-Pull-Request: golang#63297 Reviewed-on: https://go-review.googlesource.com/c/go/+/531835 Reviewed-by: Cherry Mui <cherryyz@google.com> Reviewed-by: Keith Randall <khr@google.com> Reviewed-by: Keith Randall <khr@golang.org> LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
lifubang · May 20, 2024 · 49cb8ac · 49cb8ac
1 parent 7f9edb4
commit 49cb8ac
Show file tree

Hide file tree

Showing 6 changed files with 158 additions and 4 deletions.
diff --git a/src/internal/runtime/atomic/atomic_andor_generic.go b/src/internal/runtime/atomic/atomic_andor_generic.go
@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-//go:build arm || mips || mipsle || mips64 || mips64le || wasm
+//go:build arm || wasm
 
 package atomic
 

diff --git a/src/internal/runtime/atomic/atomic_mips64x.go b/src/internal/runtime/atomic/atomic_mips64x.go
@@ -61,6 +61,24 @@ func And(ptr *uint32, val uint32)
 //go:noescape
 func Or(ptr *uint32, val uint32)
 
+//go:noescape
+func And32(ptr *uint32, val uint32) uint32
+
+//go:noescape
+func Or32(ptr *uint32, val uint32) uint32
+
+//go:noescape
+func And64(ptr *uint64, val uint64) uint64
+
+//go:noescape
+func Or64(ptr *uint64, val uint64) uint64
+
+//go:noescape
+func Anduintptr(ptr *uintptr, val uintptr) uintptr
+
+//go:noescape
+func Oruintptr(ptr *uintptr, val uintptr) uintptr
+
 //go:noescape
 func Cas64(ptr *uint64, old, new uint64) bool
 

diff --git a/src/internal/runtime/atomic/atomic_mips64x.s b/src/internal/runtime/atomic/atomic_mips64x.s
@@ -310,6 +310,70 @@ TEXT ·And(SB), NOSPLIT, $0-12
 	SYNC
 	RET
 
+// func Or32(addr *uint32, v uint32) old uint32
+TEXT ·Or32(SB), NOSPLIT, $0-20
+	MOVV	ptr+0(FP), R1
+	MOVW	val+8(FP), R2
+
+	SYNC
+	LL	(R1), R3
+	OR	R2, R3, R4
+	SC	R4, (R1)
+	BEQ	R4, -3(PC)
+	SYNC
+	MOVW	R3, ret+16(FP)
+	RET
+
+// func And32(addr *uint32, v uint32) old uint32
+TEXT ·And32(SB), NOSPLIT, $0-20
+	MOVV	ptr+0(FP), R1
+	MOVW	val+8(FP), R2
+
+	SYNC
+	LL	(R1), R3
+	AND	R2, R3, R4
+	SC	R4, (R1)
+	BEQ	R4, -3(PC)
+	SYNC
+	MOVW	R3, ret+16(FP)
+	RET
+
+// func Or64(addr *uint64, v uint64) old uint64
+TEXT ·Or64(SB), NOSPLIT, $0-24
+	MOVV	ptr+0(FP), R1
+	MOVV	val+8(FP), R2
+
+	SYNC
+	LLV	(R1), R3
+	OR	R2, R3, R4
+	SCV	R4, (R1)
+	BEQ	R4, -3(PC)
+	SYNC
+	MOVV	R3, ret+16(FP)
+	RET
+
+// func And64(addr *uint64, v uint64) old uint64
+TEXT ·And64(SB), NOSPLIT, $0-24
+	MOVV	ptr+0(FP), R1
+	MOVV	val+8(FP), R2
+
+	SYNC
+	LLV	(R1), R3
+	AND	R2, R3, R4
+	SCV	R4, (R1)
+	BEQ	R4, -3(PC)
+	SYNC
+	MOVV	R3, ret+16(FP)
+	RET
+
+// func Anduintptr(addr *uintptr, v uintptr) old uintptr
+TEXT ·Anduintptr(SB), NOSPLIT, $0-24
+	JMP	·And64(SB)
+
+// func Oruintptr(addr *uintptr, v uintptr) old uintptr
+TEXT ·Oruintptr(SB), NOSPLIT, $0-24
+	JMP	·Or64(SB)
+
 // uint32 ·Load(uint32 volatile* ptr)
 TEXT ·Load(SB),NOSPLIT|NOFRAME,$0-12
 	MOVV	ptr+0(FP), R1

diff --git a/src/internal/runtime/atomic/atomic_mipsx.go b/src/internal/runtime/atomic/atomic_mipsx.go
@@ -104,6 +104,26 @@ func Store64(addr *uint64, val uint64) {
 	return
 }
 
+//go:nosplit
+func Or64(addr *uint64, val uint64) (old uint64) {
+	for {
+		old = *addr
+		if Cas64(addr, old, old|val) {
+			return old
+		}
+	}
+}
+
+//go:nosplit
+func And64(addr *uint64, val uint64) (old uint64) {
+	for {
+		old = *addr
+		if Cas64(addr, old, old&val) {
+			return old
+		}
+	}
+}
+
 //go:noescape
 func Xadd(ptr *uint32, delta int32) uint32
 
@@ -143,6 +163,18 @@ func And(ptr *uint32, val uint32)
 //go:noescape
 func Or(ptr *uint32, val uint32)
 
+//go:noescape
+func And32(ptr *uint32, val uint32) uint32
+
+//go:noescape
+func Or32(ptr *uint32, val uint32) uint32
+
+//go:noescape
+func Anduintptr(ptr *uintptr, val uintptr) uintptr
+
+//go:noescape
+func Oruintptr(ptr *uintptr, val uintptr) uintptr
+
 //go:noescape
 func Store(ptr *uint32, val uint32)
 

diff --git a/src/internal/runtime/atomic/atomic_mipsx.s b/src/internal/runtime/atomic/atomic_mipsx.s
@@ -240,6 +240,42 @@ TEXT ·And(SB), NOSPLIT, $0-8
 	SYNC
 	RET
 
+// func Or32(addr *uint32, v uint32) old uint32
+TEXT ·Or32(SB), NOSPLIT, $0-12
+	MOVW	ptr+0(FP), R1
+	MOVW	val+4(FP), R2
+
+	SYNC
+	LL	(R1), R3
+	OR	R2, R3, R4
+	SC	R4, (R1)
+	BEQ	R4, -4(PC)
+	SYNC
+	MOVW	R3, ret+8(FP)
+	RET
+
+// func And32(addr *uint32, v uint32) old uint32
+TEXT ·And32(SB), NOSPLIT, $0-12
+	MOVW	ptr+0(FP), R1
+	MOVW	val+4(FP), R2
+
+	SYNC
+	LL	(R1), R3
+	AND	R2, R3, R4
+	SC	R4, (R1)
+	BEQ	R4, -4(PC)
+	SYNC
+	MOVW	R3, ret+8(FP)
+	RET
+
+// func Anduintptr(addr *uintptr, v uintptr) old uintptr
+TEXT ·Anduintptr(SB), NOSPLIT, $0-12
+	JMP	·And32(SB)
+
+// func Oruintptr(addr *uintptr, v uintptr) old uintptr
+TEXT ·Oruintptr(SB), NOSPLIT, $0-12
+	JMP	·Or32(SB)
+
 TEXT ·spinLock(SB),NOSPLIT,$0-4
 	MOVW	state+0(FP), R1
 	MOVW	$1, R2

diff --git a/src/runtime/cgo/gcc_stack_unix.c b/src/runtime/cgo/gcc_stack_unix.c
@@ -17,20 +17,24 @@ x_cgo_getstackbound(uintptr bounds[2])
 	pthread_attr_t attr;
 	void *addr;
 	size_t size;
+	int err;
 
 #if defined(__GLIBC__) || (defined(__sun) && !defined(__illumos__))
 	// pthread_getattr_np is a GNU extension supported in glibc.
 	// Solaris is not glibc but does support pthread_getattr_np
 	// (and the fallback doesn't work...). Illumos does not.
-	pthread_getattr_np(pthread_self(), &attr);  // GNU extension
+	err = pthread_getattr_np(pthread_self(), &attr);  // GNU extension
+	if (err != 0) {
+		// This is to have a backward compatibility with glibc < 2.32
+		pthread_attr_init(&attr);
+		pthread_getattr_np(pthread_self(), &attr);
+	}
 	pthread_attr_getstack(&attr, &addr, &size); // low address
 #elif defined(__illumos__)
 	pthread_attr_init(&attr);
 	pthread_attr_get_np(pthread_self(), &attr);
 	pthread_attr_getstack(&attr, &addr, &size); // low address
 #else
-	// We don't know how to get the current stacks, so assume they are the
-	// same as the default stack bounds.
 	pthread_attr_init(&attr);
 	pthread_attr_getstacksize(&attr, &size);
 	addr = __builtin_frame_address(0) + 4096 - size;