ASoC: Intel: sst: fix sst_memcpy32 wrong with non-4x bytes issue

sst_memcpy32() only copied bytes/4 32bits, which means it dropped
the remaining bytes%4 bytes wrongly.

Here add copying those missing bytes, first to a 32bits tmp, and
then write the tmp to 32bits iomem.
Signed-off-by: Jie Yang <yang.jie@intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

sound/soc/intel/common/sst-firmware.c

@@ -51,8 +51,22 @@ struct sst_dma {
 
 static inline void sst_memcpy32(volatile void __iomem *dest, void *src, u32 bytes)
 {
+	u32 tmp = 0;
+	int i, m, n;
+	const u8 *src_byte = src;
+
+	m = bytes / 4;
+	n = bytes % 4;
+
 	/* __iowrite32_copy use 32bit size values so divide by 4 */
-	__iowrite32_copy((void *)dest, src, bytes/4);
+	__iowrite32_copy((void *)dest, src, m);
+
+	if (n) {
+		for (i = 0; i < n; i++)
+			tmp |= (u32)*(src_byte + m * 4 + i) << (i * 8);
+		__iowrite32_copy((void *)(dest + m * 4), &tmp, 1);
+	}
+
 }
 
 static void sst_dma_transfer_complete(void *arg)