clk_disable_unprepare(mxs_dma->clk);
 }
 
+/*
+ * How to use the flags for ->device_prep_slave_sg() :
+ *    [1] If there is only one DMA command in the DMA chain, the code should be:
+ *            ......
+ *            ->device_prep_slave_sg(DMA_CTRL_ACK);
+ *            ......
+ *    [2] If there are two DMA commands in the DMA chain, the code should be
+ *            ......
+ *            ->device_prep_slave_sg(0);
+ *            ......
+ *            ->device_prep_slave_sg(DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
+ *            ......
+ *    [3] If there are more than two DMA commands in the DMA chain, the code
+ *        should be:
+ *            ......
+ *            ->device_prep_slave_sg(0);                                // First
+ *            ......
+ *            ->device_prep_slave_sg(DMA_PREP_INTERRUPT [| DMA_CTRL_ACK]);
+ *            ......
+ *            ->device_prep_slave_sg(DMA_PREP_INTERRUPT | DMA_CTRL_ACK); // Last
+ *            ......
+ */
 static struct dma_async_tx_descriptor *mxs_dma_prep_slave_sg(
                struct dma_chan *chan, struct scatterlist *sgl,
                unsigned int sg_len, enum dma_transfer_direction direction,
-               unsigned long append)
+               unsigned long flags)
 {
        struct mxs_dma_chan *mxs_chan = to_mxs_dma_chan(chan);
        struct mxs_dma_engine *mxs_dma = mxs_chan->mxs_dma;
        struct scatterlist *sg;
        int i, j;
        u32 *pio;
+       bool append = flags & DMA_PREP_INTERRUPT;
        int idx = append ? mxs_chan->desc_count : 0;
 
        if (mxs_chan->status == DMA_IN_PROGRESS && !append)
                ccw->bits |= CCW_CHAIN;
                ccw->bits &= ~CCW_IRQ;
                ccw->bits &= ~CCW_DEC_SEM;
-               ccw->bits &= ~CCW_WAIT4END;
        } else {
                idx = 0;
        }
                ccw->bits = 0;
                ccw->bits |= CCW_IRQ;
                ccw->bits |= CCW_DEC_SEM;
-               ccw->bits |= CCW_WAIT4END;
+               if (flags & DMA_CTRL_ACK)
+                       ccw->bits |= CCW_WAIT4END;
                ccw->bits |= CCW_HALT_ON_TERM;
                ccw->bits |= CCW_TERM_FLUSH;
                ccw->bits |= BF_CCW(sg_len, PIO_NUM);
                                ccw->bits &= ~CCW_CHAIN;
                                ccw->bits |= CCW_IRQ;
                                ccw->bits |= CCW_DEC_SEM;
-                               ccw->bits |= CCW_WAIT4END;
+                               if (flags & DMA_CTRL_ACK)
+                                       ccw->bits |= CCW_WAIT4END;
                        }
                }
        }
 
 }
 
 static struct dma_async_tx_descriptor *mxs_mmc_prep_dma(
-       struct mxs_mmc_host *host, unsigned int append)
+       struct mxs_mmc_host *host, unsigned long flags)
 {
        struct dma_async_tx_descriptor *desc;
        struct mmc_data *data = host->data;
        }
 
        desc = host->dmach->device->device_prep_slave_sg(host->dmach,
-                               sgl, sg_len, host->slave_dirn, append);
+                               sgl, sg_len, host->slave_dirn, flags);
        if (desc) {
                desc->callback = mxs_mmc_dma_irq_callback;
                desc->callback_param = host;
        host->ssp_pio_words[2] = cmd1;
        host->dma_dir = DMA_NONE;
        host->slave_dirn = DMA_TRANS_NONE;
-       desc = mxs_mmc_prep_dma(host, 0);
+       desc = mxs_mmc_prep_dma(host, DMA_CTRL_ACK);
        if (!desc)
                goto out;
 
        host->ssp_pio_words[2] = cmd1;
        host->dma_dir = DMA_NONE;
        host->slave_dirn = DMA_TRANS_NONE;
-       desc = mxs_mmc_prep_dma(host, 0);
+       desc = mxs_mmc_prep_dma(host, DMA_CTRL_ACK);
        if (!desc)
                goto out;
 
        host->data = data;
        host->dma_dir = dma_data_dir;
        host->slave_dirn = slave_dirn;
-       desc = mxs_mmc_prep_dma(host, 1);
+       desc = mxs_mmc_prep_dma(host, DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
        if (!desc)
                goto out;
 
 
        sg_init_one(sgl, this->cmd_buffer, this->command_length);
        dma_map_sg(this->dev, sgl, 1, DMA_TO_DEVICE);
        desc = channel->device->device_prep_slave_sg(channel,
-                                       sgl, 1, DMA_MEM_TO_DEV, 1);
+                               sgl, 1, DMA_MEM_TO_DEV,
+                               DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
+
        if (!desc) {
                pr_err("step 2 error\n");
                return -1;
        /* [2] send DMA request */
        prepare_data_dma(this, DMA_TO_DEVICE);
        desc = channel->device->device_prep_slave_sg(channel, &this->data_sgl,
-                                               1, DMA_MEM_TO_DEV, 1);
+                                       1, DMA_MEM_TO_DEV,
+                                       DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
        if (!desc) {
                pr_err("step 2 error\n");
                return -1;
        /* [2] : send DMA request */
        prepare_data_dma(this, DMA_FROM_DEVICE);
        desc = channel->device->device_prep_slave_sg(channel, &this->data_sgl,
-                                               1, DMA_DEV_TO_MEM, 1);
+                                       1, DMA_DEV_TO_MEM,
+                                       DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
        if (!desc) {
                pr_err("step 2 error\n");
                return -1;
 
        desc = channel->device->device_prep_slave_sg(channel,
                                        (struct scatterlist *)pio,
-                                       ARRAY_SIZE(pio), DMA_TRANS_NONE, 0);
+                                       ARRAY_SIZE(pio), DMA_TRANS_NONE,
+                                       DMA_CTRL_ACK);
        if (!desc) {
                pr_err("step 2 error\n");
                return -1;
        pio[5] = auxiliary;
        desc = channel->device->device_prep_slave_sg(channel,
                                        (struct scatterlist *)pio,
-                                       ARRAY_SIZE(pio), DMA_TRANS_NONE, 1);
+                                       ARRAY_SIZE(pio), DMA_TRANS_NONE,
+                                       DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
        if (!desc) {
                pr_err("step 2 error\n");
                return -1;
        pio[1] = 0;
        desc = channel->device->device_prep_slave_sg(channel,
                                (struct scatterlist *)pio, 2,
-                               DMA_TRANS_NONE, 1);
+                               DMA_TRANS_NONE,
+                               DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
        if (!desc) {
                pr_err("step 3 error\n");
                return -1;