diff --git a/src/cmd/gc/plive.c b/src/cmd/gc/plive.c
index d2f384b2dd4..0feb2c710ac 100644
--- a/src/cmd/gc/plive.c
+++ b/src/cmd/gc/plive.c
@@ -1936,7 +1936,11 @@ twobitwritesymbol(Array *arr, Sym *sym)
 			break;
 		for(j = 0; j < bv->n; j += 32) {
 			word = bv->b[j/32];
-			off = duint32(sym, off, word);
+			// Runtime reads the bitmaps as byte arrays. Oblige.
+			off = duint8(sym, off, word);
+			off = duint8(sym, off, word>>8);
+			off = duint8(sym, off, word>>16);
+			off = duint8(sym, off, word>>24);
 		}
 	}
 	duint32(sym, 0, i); // number of bitmaps
diff --git a/src/runtime/heapdump.c b/src/runtime/heapdump.c
index db2c552b72c..3d7960aab0a 100644
--- a/src/runtime/heapdump.c
+++ b/src/runtime/heapdump.c
@@ -250,7 +250,7 @@ dumpbv(BitVector *bv, uintptr offset)
 	uintptr i;
 
 	for(i = 0; i < bv->n; i += BitsPerPointer) {
-		switch(bv->data[i/32] >> i%32 & 3) {
+		switch(bv->bytedata[i/8] >> i%8 & 3) {
 		case BitsDead:
 		case BitsScalar:
 			break;
@@ -259,7 +259,7 @@ dumpbv(BitVector *bv, uintptr offset)
 			dumpint(offset + i / BitsPerPointer * PtrSize);
 			break;
 		case BitsMultiWord:
-			switch(bv->data[(i+BitsPerPointer)/32] >> (i+BitsPerPointer)%32 & 3) {
+			switch(bv->bytedata[(i+BitsPerPointer)/8] >> (i+BitsPerPointer)%8 & 3) {
 			default:
 				runtime·throw("unexpected garbage collection bits");
 			case BitsIface:
@@ -813,9 +813,9 @@ dumpbvtypes(BitVector *bv, byte *base)
 	uintptr i;
 
 	for(i = 0; i < bv->n; i += BitsPerPointer) {
-		if((bv->data[i/32] >> i%32 & 3) != BitsMultiWord)
+		if((bv->bytedata[i/8] >> i%8 & 3) != BitsMultiWord)
 			continue;
-		switch(bv->data[(i+BitsPerPointer)/32] >> (i+BitsPerPointer)%32 & 3) {
+		switch(bv->bytedata[(i+BitsPerPointer)/8] >> (i+BitsPerPointer)%8 & 3) {
 		default:
 			runtime·throw("unexpected garbage collection bits");
 		case BitsIface:
@@ -860,5 +860,5 @@ makeheapobjbv(byte *p, uintptr size)
 		tmpbuf[i*BitsPerPointer/8] &= ~(BitsMask<<((i*BitsPerPointer)%8));
 		tmpbuf[i*BitsPerPointer/8] |= bits<<((i*BitsPerPointer)%8);
 	}
-	return (BitVector){i*BitsPerPointer, (uint32*)tmpbuf};
+	return (BitVector){i*BitsPerPointer, tmpbuf};
 }
diff --git a/src/runtime/malloc.h b/src/runtime/malloc.h
index b6856768db1..60b87da78c1 100644
--- a/src/runtime/malloc.h
+++ b/src/runtime/malloc.h
@@ -589,14 +589,14 @@ bool	runtime·freespecial(Special *s, void *p, uintptr size, bool freed);
 struct BitVector
 {
 	int32 n; // # of bits
-	uint32 *data;
+	uint8 *bytedata;
 };
 typedef struct StackMap StackMap;
 struct StackMap
 {
 	int32 n; // number of bitmaps
 	int32 nbit; // number of bits in each bitmap
-	uint32 data[];
+	uint8 bytedata[]; // bitmaps, each starting on a 32-bit boundary
 };
 // Returns pointer map data for the given stackmap index
 // (the index is encoded in PCDATA_StackMapIndex).
diff --git a/src/runtime/mgc0.c b/src/runtime/mgc0.c
index af0b6285a30..4221aaab2fc 100644
--- a/src/runtime/mgc0.c
+++ b/src/runtime/mgc0.c
@@ -438,11 +438,11 @@ markroot(ParFor *desc, uint32 i)
 	// Note: if you add a case here, please also update heapdump.c:dumproots.
 	switch(i) {
 	case RootData:
-		scanblock(runtime·data, runtime·edata - runtime·data, (byte*)runtime·gcdatamask.data);
+		scanblock(runtime·data, runtime·edata - runtime·data, runtime·gcdatamask.bytedata);
 		break;
 
 	case RootBss:
-		scanblock(runtime·bss, runtime·ebss - runtime·bss, (byte*)runtime·gcbssmask.data);
+		scanblock(runtime·bss, runtime·ebss - runtime·bss, runtime·gcbssmask.bytedata);
 		break;
 
 	case RootFinalizers:
@@ -610,7 +610,7 @@ runtime·stackmapdata(StackMap *stackmap, int32 n)
 {
 	if(n < 0 || n >= stackmap->n)
 		runtime·throw("stackmapdata: index out of range");
-	return (BitVector){stackmap->nbit, stackmap->data + n*((stackmap->nbit+31)/32)};
+	return (BitVector){stackmap->nbit, stackmap->bytedata + n*((stackmap->nbit+31)/32*4)};
 }
 
 // Scan a stack frame: local variables and function arguments/results.
@@ -668,17 +668,17 @@ scanframe(Stkframe *frame, void *unused)
 		}
 		bv = runtime·stackmapdata(stackmap, pcdata);
 		size = (bv.n * PtrSize) / BitsPerPointer;
-		scanblock((byte*)(frame->varp - size), bv.n/BitsPerPointer*PtrSize, (byte*)bv.data);
+		scanblock((byte*)(frame->varp - size), bv.n/BitsPerPointer*PtrSize, bv.bytedata);
 	}
 
 	// Scan arguments.
 	// Use pointer information if known.
 	if(frame->argmap != nil) {
 		bv = *frame->argmap;
-		scanblock((byte*)frame->argp, bv.n/BitsPerPointer*PtrSize, (byte*)bv.data);
+		scanblock((byte*)frame->argp, bv.n/BitsPerPointer*PtrSize, bv.bytedata);
 	} else if((stackmap = runtime·funcdata(f, FUNCDATA_ArgsPointerMaps)) != nil) {
 		bv = runtime·stackmapdata(stackmap, pcdata);
-		scanblock((byte*)frame->argp, bv.n/BitsPerPointer*PtrSize, (byte*)bv.data);
+		scanblock((byte*)frame->argp, bv.n/BitsPerPointer*PtrSize, bv.bytedata);
 	} else {
 		if(Debug > 2)
 			runtime·printf("frame %s conservative args %p+%p\n", runtime·funcname(f), frame->argp, (uintptr)frame->arglen);
@@ -1642,7 +1642,7 @@ unrollglobgcprog(byte *prog, uintptr size)
 		runtime·throw("unrollglobgcprog: program does not end with insEnd");
 	if(mask[masksize] != 0xa1)
 		runtime·throw("unrollglobgcprog: overflow");
-	return (BitVector){masksize*8, (uint32*)mask};
+	return (BitVector){masksize*8, mask};
 }
 
 void
@@ -1833,7 +1833,7 @@ runtime·getgcmask(byte *p, Type *t, byte **mask, uintptr *len)
 		*mask = runtime·mallocgc(*len, nil, FlagNoScan);
 		for(i = 0; i < n; i += PtrSize) {
 			off = (p+i-runtime·data)/PtrSize;
-			bits = (((byte*)runtime·gcdatamask.data)[off/PointersPerByte] >> ((off%PointersPerByte)*BitsPerPointer))&BitsMask;
+			bits = (runtime·gcdatamask.bytedata[off/PointersPerByte] >> ((off%PointersPerByte)*BitsPerPointer))&BitsMask;
 			(*mask)[i/PtrSize] = bits;
 		}
 		return;
@@ -1845,7 +1845,7 @@ runtime·getgcmask(byte *p, Type *t, byte **mask, uintptr *len)
 		*mask = runtime·mallocgc(*len, nil, FlagNoScan);
 		for(i = 0; i < n; i += PtrSize) {
 			off = (p+i-runtime·bss)/PtrSize;
-			bits = (((byte*)runtime·gcbssmask.data)[off/PointersPerByte] >> ((off%PointersPerByte)*BitsPerPointer))&BitsMask;
+			bits = (runtime·gcbssmask.bytedata[off/PointersPerByte] >> ((off%PointersPerByte)*BitsPerPointer))&BitsMask;
 			(*mask)[i/PtrSize] = bits;
 		}
 		return;
@@ -1895,7 +1895,7 @@ runtime·getgcmask(byte *p, Type *t, byte **mask, uintptr *len)
 		*mask = runtime·mallocgc(*len, nil, FlagNoScan);
 		for(i = 0; i < n; i += PtrSize) {
 			off = (p+i-(byte*)frame.varp+size)/PtrSize;
-			bits = (bv.data[off*BitsPerPointer/32] >> ((off*BitsPerPointer)%32))&BitsMask;
+			bits = (bv.bytedata[off*BitsPerPointer/8] >> ((off*BitsPerPointer)%8))&BitsMask;
 			(*mask)[i/PtrSize] = bits;
 		}
 	}
diff --git a/src/runtime/stack.c b/src/runtime/stack.c
index 53ad90a5deb..f29266eb6be 100644
--- a/src/runtime/stack.c
+++ b/src/runtime/stack.c
@@ -370,8 +370,8 @@ adjustpointers(byte **scanp, BitVector *bv, AdjustInfo *adjinfo, Func *f)
 	num = bv->n / BitsPerPointer;
 	for(i = 0; i < num; i++) {
 		if(StackDebug >= 4)
-			runtime·printf("        %p:%s:%p\n", &scanp[i], mapnames[bv->data[i / (32 / BitsPerPointer)] >> (i * BitsPerPointer & 31) & 3], scanp[i]);
-		switch(bv->data[i / (32 / BitsPerPointer)] >> (i * BitsPerPointer & 31) & 3) {
+			runtime·printf("        %p:%s:%p\n", &scanp[i], mapnames[bv->bytedata[i / (8 / BitsPerPointer)] >> (i * BitsPerPointer & 7) & 3], scanp[i]);
+		switch(bv->bytedata[i / (8 / BitsPerPointer)] >> (i * BitsPerPointer & 7) & 3) {
 		case BitsDead:
 			if(runtime·debug.gcdead)
 				scanp[i] = (byte*)PoisonStack;
@@ -394,7 +394,7 @@ adjustpointers(byte **scanp, BitVector *bv, AdjustInfo *adjinfo, Func *f)
 			}
 			break;
 		case BitsMultiWord:
-			switch(bv->data[(i+1) / (32 / BitsPerPointer)] >> ((i+1) * BitsPerPointer & 31) & 3) {
+			switch(bv->bytedata[(i+1) / (8 / BitsPerPointer)] >> ((i+1) * BitsPerPointer & 7) & 3) {
 			default:
 				runtime·throw("unexpected garbage collection bits");
 			case BitsEface: