runtime: always pass type to mallocgc when allocating scannable memory

We allocate scannable memory w/o type only in few places in runtime.
All these cases are not-performance critical (e.g. G or finq args buffer),
and in long term they all need to go away.
It's not worth it to have special code for this case in mallocgc.
So use special fake "notype" type for such allocations.

LGTM=khr
R=golang-codereviews, khr
CC=golang-codereviews, rlh, rsc
https://golang.org/cl/127450044

3 files changed, 57 insertions, 49 deletions

diff --git a/src/pkg/runtime/malloc.c b/src/pkg/runtime/malloc.c
index f4143669e7..ca8ebf5dd6 100644
--- a/src/pkg/runtime/malloc.c
+++ b/src/pkg/runtime/malloc.c
@@ -21,7 +21,10 @@ MHeap runtime·mheap;
 #pragma dataflag NOPTR
 MStats runtime·memstats;
 
+static Type* notype;
+
 void runtime·cmallocgc(uintptr size, Type *typ, uint32 flag, void **ret);
+void runtime·gc_notype_ptr(Eface*);
 
 void*
 runtime·mallocgc(uintptr size, Type *typ, uint32 flag)
@@ -31,6 +34,8 @@ runtime·mallocgc(uintptr size, Type *typ, uint32 flag)
 	// Call into the Go version of mallocgc.
 	// TODO: maybe someday we can get rid of this.  It is
 	// probably the only location where we run Go code on the M stack.
+	if((flag&FlagNoScan) == 0 && typ == nil)
+		typ = notype;
 	runtime·cmallocgc(size, typ, flag, &ret);
 	return ret;
 }
@@ -124,6 +129,7 @@ runtime·mallocinit(void)
 	uintptr limit;
 	uint64 i;
 	bool reserved;
+	Eface notype_eface;
 
 	p = nil;
 	p_size = 0;
@@ -251,6 +257,9 @@ runtime·mallocinit(void)
 	// Initialize the rest of the allocator.	
 	runtime·MHeap_Init(&runtime·mheap);
 	g->m->mcache = runtime·allocmcache();
+
+	runtime·gc_notype_ptr(&notype_eface);
+	notype = notype_eface.type;
 }
 
 void*
diff --git a/src/pkg/runtime/malloc.go b/src/pkg/runtime/malloc.go
index 84c69abde7..152b3b6b68 100644
--- a/src/pkg/runtime/malloc.go
+++ b/src/pkg/runtime/malloc.go
@@ -225,66 +225,56 @@ func gomallocgc(size uintptr, typ *_type, flags int) unsafe.Pointer {
 			*xbits |= (bitsPointer << 2) << shift
 			goto marked
 		}
-		if typ != nil && (uintptr(typ.gc[0])|uintptr(typ.gc[1])) != 0 && uintptr(typ.size) > ptrSize {
-			if typ.kind&kindGCProg != 0 {
-				nptr := (uintptr(typ.size) + ptrSize - 1) / ptrSize
-				masksize := nptr
-				if masksize%2 != 0 {
-					masksize *= 2 // repeated
-				}
-				masksize = masksize * pointersPerByte / 8 // 4 bits per word
-				masksize++                                // unroll flag in the beginning
-				if masksize > maxGCMask && typ.gc[1] != 0 {
-					// If the mask is too large, unroll the program directly
-					// into the GC bitmap. It's 7 times slower than copying
-					// from the pre-unrolled mask, but saves 1/16 of type size
-					// memory for the mask.
-					mp := acquirem()
-					mp.ptrarg[0] = x
-					mp.ptrarg[1] = unsafe.Pointer(typ)
-					mp.scalararg[0] = uint(size)
-					mp.scalararg[1] = uint(size0)
-					onM(&unrollgcproginplace_m)
-					releasem(mp)
-					goto marked
-				}
-				ptrmask = (*uint8)(unsafe.Pointer(uintptr(typ.gc[0])))
-				// Check whether the program is already unrolled.
-				if uintptr(goatomicloadp(unsafe.Pointer(ptrmask)))&0xff == 0 {
-					mp := acquirem()
-					mp.ptrarg[0] = unsafe.Pointer(typ)
-					onM(&unrollgcprog_m)
-					releasem(mp)
-				}
-				ptrmask = (*uint8)(add(unsafe.Pointer(ptrmask), 1)) // skip the unroll flag byte
-			} else {
-				ptrmask = (*uint8)(unsafe.Pointer(&typ.gc[0])) // embed mask
+		if typ.kind&kindGCProg != 0 {
+			nptr := (uintptr(typ.size) + ptrSize - 1) / ptrSize
+			masksize := nptr
+			if masksize%2 != 0 {
+				masksize *= 2 // repeated
 			}
-			if size == 2*ptrSize {
-				xbitsb := (*uint8)(add(unsafe.Pointer(xbits), shift/8))
-				*xbitsb = *ptrmask | bitBoundary
+			masksize = masksize * pointersPerByte / 8 // 4 bits per word
+			masksize++                                // unroll flag in the beginning
+			if masksize > maxGCMask && typ.gc[1] != 0 {
+				// If the mask is too large, unroll the program directly
+				// into the GC bitmap. It's 7 times slower than copying
+				// from the pre-unrolled mask, but saves 1/16 of type size
+				// memory for the mask.
+				mp := acquirem()
+				mp.ptrarg[0] = x
+				mp.ptrarg[1] = unsafe.Pointer(typ)
+				mp.scalararg[0] = uint(size)
+				mp.scalararg[1] = uint(size0)
+				onM(&unrollgcproginplace_m)
+				releasem(mp)
 				goto marked
 			}
-			te = uintptr(typ.size) / ptrSize
-			// If the type occupies odd number of words, its mask is repeated.
-			if te%2 == 0 {
-				te /= 2
+			ptrmask = (*uint8)(unsafe.Pointer(uintptr(typ.gc[0])))
+			// Check whether the program is already unrolled.
+			if uintptr(goatomicloadp(unsafe.Pointer(ptrmask)))&0xff == 0 {
+				mp := acquirem()
+				mp.ptrarg[0] = unsafe.Pointer(typ)
+				onM(&unrollgcprog_m)
+				releasem(mp)
 			}
+			ptrmask = (*uint8)(add(unsafe.Pointer(ptrmask), 1)) // skip the unroll flag byte
+		} else {
+			ptrmask = (*uint8)(unsafe.Pointer(&typ.gc[0])) // embed mask
 		}
 		if size == 2*ptrSize {
 			xbitsb := (*uint8)(add(unsafe.Pointer(xbits), shift/8))
-			*xbitsb = (bitsPointer << 2) | (bitsPointer << 6) | bitBoundary
+			*xbitsb = *ptrmask | bitBoundary
 			goto marked
 		}
+		te = uintptr(typ.size) / ptrSize
+		// If the type occupies odd number of words, its mask is repeated.
+		if te%2 == 0 {
+			te /= 2
+		}
 		// Copy pointer bitmask into the bitmap.
 		for i := uintptr(0); i < size0; i += 2 * ptrSize {
-			v := uint8((bitsPointer << 2) | (bitsPointer << 6))
-			if ptrmask != nil {
-				v = *(*uint8)(add(unsafe.Pointer(ptrmask), ti))
-				ti++
-				if ti == te {
-					ti = 0
-				}
+			v := *(*uint8)(add(unsafe.Pointer(ptrmask), ti))
+			ti++
+			if ti == te {
+				ti = 0
 			}
 			if i == 0 {
 				v |= bitBoundary
diff --git a/src/pkg/runtime/mgc0.go b/src/pkg/runtime/mgc0.go
index 624485d18b..496725f6a7 100644
--- a/src/pkg/runtime/mgc0.go
+++ b/src/pkg/runtime/mgc0.go
@@ -19,6 +19,15 @@ func gc_itab_ptr(ret *interface{}) {
 	*ret = (*itab)(nil)
 }
 
+// Type used for "conservative" allocations in C code.
+type notype [8]*byte
+
+// Called from C. Returns the Go type used for C allocations w/o type.
+func gc_notype_ptr(ret *interface{}) {
+	var x notype
+	*ret = x
+}
+
 func timenow() (sec int64, nsec int32)
 
 func gc_unixnanotime(now *int64) {