Merge tag 'mmc-merge-for-3.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/cjb/mmc

Pull MMC changes from Chris Ball
 - at91-mci: This driver will be replaced by atmel-mci in 3.7.
 - atmel-mci: Add support for old at91-mci hardware.
 - dw_mmc: Allow multiple controllers; this previously caused
   corruption.
 - imxmmc: Remove this driver, replaced by mxcmmc.
 - mmci: Add device tree support.
 - omap: Allow multiple controllers.
 - omap_hsmmc: Auto CMD12, DDR support.
 - tegra: Support SD 3.0 spec.

Fix up the usual trivial conflicts in feature-removal-schedule.txt

* tag 'mmc-merge-for-3.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/cjb/mmc: (38 commits)
  mmc: at91-mci: this driver is now deprecated
  mmc: omap_hsmmc: pass IRQF_ONESHOT to request_threaded_irq
  mmc: block: Allow disabling 512B sector size emulation
  mmc: atmel-mci: add debug logs
  mmc: atmel-mci: add support for version lower than v2xx
  mmc: atmel-mci: change the state machine for compatibility with old IP
  mmc: atmel-mci: the r/w proof capability lack was not well managed
  mmc: dw_mmc: Fixed sdio interrupt mask bit setting bug
  mmc: omap: convert to module_platform_driver
  mmc: omap: make it behave well as a module
  mmc: omap: convert to per instance workqueue
  mmc: core: Remove dead code
  mmc: card: Avoid null pointer dereference
  mmc: core: Prevent eMMC VCC supply to be cut from late init
  mmc: dw_mmc: make multiple instances of dw_mci_card_workqueue
  mmc: queue: remove redundant memsets
  mmc: queue: rename mmc_request function
  mmc: core: skip card initialization if power class selection fails
  mmc: core: fix the signaling 1.8V for HS200
  mmc: core: fix the decision of HS200/DDR card-type
  ...

21 files changed, 630 insertions, 1601 deletions

diff --git a/drivers/mmc/card/block.c b/drivers/mmc/card/block.c
index dabec556ebb8..dd2d374dcc7a 100644
--- a/drivers/mmc/card/block.c
+++ b/drivers/mmc/card/block.c
@@ -384,7 +384,7 @@ static int mmc_blk_ioctl_cmd(struct block_device *bdev,
 	md = mmc_blk_get(bdev->bd_disk);
 	if (!md) {
 		err = -EINVAL;
-		goto cmd_done;
+		goto cmd_err;
 	}
 
 	card = md->queue.card;
@@ -483,6 +483,7 @@ cmd_rel_host:
 
 cmd_done:
 	mmc_blk_put(md);
+cmd_err:
 	kfree(idata->buf);
 	kfree(idata);
 	return err;
@@ -1283,7 +1284,7 @@ static int mmc_blk_issue_rw_rq(struct mmc_queue *mq, struct request *rqc)
 	int ret = 1, disable_multi = 0, retry = 0, type;
 	enum mmc_blk_status status;
 	struct mmc_queue_req *mq_rq;
-	struct request *req;
+	struct request *req = rqc;
 	struct mmc_async_req *areq;
 
 	if (!rqc && !mq->mqrq_prev->req)
@@ -1291,6 +1292,16 @@ static int mmc_blk_issue_rw_rq(struct mmc_queue *mq, struct request *rqc)
 
 	do {
 		if (rqc) {
+			/*
+			 * When 4KB native sector is enabled, only 8 blocks
+			 * multiple read or write is allowed
+			 */
+			if ((brq->data.blocks & 0x07) &&
+			    (card->ext_csd.data_sector_size == 4096)) {
+				pr_err("%s: Transfer size is not 4KB sector size aligned\n",
+					req->rq_disk->disk_name);
+				goto cmd_abort;
+			}
 			mmc_blk_rw_rq_prep(mq->mqrq_cur, card, 0, mq);
 			areq = &mq->mqrq_cur->mmc_active;
 		} else
@@ -1538,7 +1549,12 @@ static struct mmc_blk_data *mmc_blk_alloc_req(struct mmc_card *card,
 	snprintf(md->disk->disk_name, sizeof(md->disk->disk_name),
 		 "mmcblk%d%s", md->name_idx, subname ? subname : "");
 
-	blk_queue_logical_block_size(md->queue.queue, 512);
+	if (mmc_card_mmc(card))
+		blk_queue_logical_block_size(md->queue.queue,
+					     card->ext_csd.data_sector_size);
+	else
+		blk_queue_logical_block_size(md->queue.queue, 512);
+
 	set_capacity(md->disk, size);
 
 	if (mmc_host_cmd23(card->host)) {
diff --git a/drivers/mmc/card/queue.c b/drivers/mmc/card/queue.c
index 996f8e36e23d..e360a979857d 100644
--- a/drivers/mmc/card/queue.c
+++ b/drivers/mmc/card/queue.c
@@ -96,7 +96,7 @@ static int mmc_queue_thread(void *d)
  * on any queue on this host, and attempt to issue it.  This may
  * not be the queue we were asked to process.
  */
-static void mmc_request(struct request_queue *q)
+static void mmc_request_fn(struct request_queue *q)
 {
 	struct mmc_queue *mq = q->queuedata;
 	struct request *req;
@@ -171,12 +171,10 @@ int mmc_init_queue(struct mmc_queue *mq, struct mmc_card *card,
 		limit = *mmc_dev(host)->dma_mask;
 
 	mq->card = card;
-	mq->queue = blk_init_queue(mmc_request, lock);
+	mq->queue = blk_init_queue(mmc_request_fn, lock);
 	if (!mq->queue)
 		return -ENOMEM;
 
-	memset(&mq->mqrq_cur, 0, sizeof(mq->mqrq_cur));
-	memset(&mq->mqrq_prev, 0, sizeof(mq->mqrq_prev));
 	mq->mqrq_cur = mqrq_cur;
 	mq->mqrq_prev = mqrq_prev;
 	mq->queue->queuedata = mq;
diff --git a/drivers/mmc/core/bus.c b/drivers/mmc/core/bus.c
index c60cee92a2b2..9b68933f27e7 100644
--- a/drivers/mmc/core/bus.c
+++ b/drivers/mmc/core/bus.c
@@ -122,6 +122,7 @@ static int mmc_bus_remove(struct device *dev)
 	return 0;
 }
 
+#ifdef CONFIG_PM_SLEEP
 static int mmc_bus_suspend(struct device *dev)
 {
 	struct mmc_driver *drv = to_mmc_driver(dev->driver);
@@ -143,6 +144,7 @@ static int mmc_bus_resume(struct device *dev)
 		ret = drv->resume(card);
 	return ret;
 }
+#endif
 
 #ifdef CONFIG_PM_RUNTIME
 
diff --git a/drivers/mmc/core/cd-gpio.c b/drivers/mmc/core/cd-gpio.c
index 2c14be73254c..f13e38deceac 100644
--- a/drivers/mmc/core/cd-gpio.c
+++ b/drivers/mmc/core/cd-gpio.c
@@ -73,6 +73,9 @@ void mmc_cd_gpio_free(struct mmc_host *host)
 {
 	struct mmc_cd_gpio *cd = host->hotplug.handler_priv;
 
+	if (!cd)
+		return;
+
 	free_irq(host->hotplug.irq, host);
 	gpio_free(cd->gpio);
 	kfree(cd);
diff --git a/drivers/mmc/core/core.c b/drivers/mmc/core/core.c
index ba821fe70bca..0b6141d29dbd 100644
--- a/drivers/mmc/core/core.c
+++ b/drivers/mmc/core/core.c
@@ -42,6 +42,7 @@
 #include "sdio_ops.h"
 
 static struct workqueue_struct *workqueue;
+static const unsigned freqs[] = { 400000, 300000, 200000, 100000 };
 
 /*
  * Enabling software CRCs on the data blocks can be a significant (30%)
@@ -1157,6 +1158,9 @@ static void mmc_power_up(struct mmc_host *host)
 {
 	int bit;
 
+	if (host->ios.power_mode == MMC_POWER_ON)
+		return;
+
 	mmc_host_clk_hold(host);
 
 	/* If ocr is set, we use it */
@@ -1199,6 +1203,10 @@ static void mmc_power_up(struct mmc_host *host)
 void mmc_power_off(struct mmc_host *host)
 {
 	int err = 0;
+
+	if (host->ios.power_mode == MMC_POWER_OFF)
+		return;
+
 	mmc_host_clk_hold(host);
 
 	host->ios.clock = 0;
@@ -2005,7 +2013,6 @@ EXPORT_SYMBOL(mmc_detect_card_removed);
 
 void mmc_rescan(struct work_struct *work)
 {
-	static const unsigned freqs[] = { 400000, 300000, 200000, 100000 };
 	struct mmc_host *host =
 		container_of(work, struct mmc_host, detect.work);
 	int i;
@@ -2044,8 +2051,12 @@ void mmc_rescan(struct work_struct *work)
 	 */
 	mmc_bus_put(host);
 
-	if (host->ops->get_cd && host->ops->get_cd(host) == 0)
+	if (host->ops->get_cd && host->ops->get_cd(host) == 0) {
+		mmc_claim_host(host);
+		mmc_power_off(host);
+		mmc_release_host(host);
 		goto out;
+	}
 
 	mmc_claim_host(host);
 	for (i = 0; i < ARRAY_SIZE(freqs); i++) {
@@ -2063,7 +2074,8 @@ void mmc_rescan(struct work_struct *work)
 
 void mmc_start_host(struct mmc_host *host)
 {
-	mmc_power_off(host);
+	host->f_init = max(freqs[0], host->f_min);
+	mmc_power_up(host);
 	mmc_detect_change(host, 0);
 }
 
diff --git a/drivers/mmc/core/mmc.c b/drivers/mmc/core/mmc.c
index 54df5adc0413..2d4a4b746750 100644
--- a/drivers/mmc/core/mmc.c
+++ b/drivers/mmc/core/mmc.c
@@ -235,6 +235,36 @@ static int mmc_get_ext_csd(struct mmc_card *card, u8 **new_ext_csd)
 	return err;
 }
 
+static void mmc_select_card_type(struct mmc_card *card)
+{
+	struct mmc_host *host = card->host;
+	u8 card_type = card->ext_csd.raw_card_type & EXT_CSD_CARD_TYPE_MASK;
+	unsigned int caps = host->caps, caps2 = host->caps2;
+	unsigned int hs_max_dtr = 0;
+
+	if (card_type & EXT_CSD_CARD_TYPE_26)
+		hs_max_dtr = MMC_HIGH_26_MAX_DTR;
+
+	if (caps & MMC_CAP_MMC_HIGHSPEED &&
+			card_type & EXT_CSD_CARD_TYPE_52)
+		hs_max_dtr = MMC_HIGH_52_MAX_DTR;
+
+	if ((caps & MMC_CAP_1_8V_DDR &&
+			card_type & EXT_CSD_CARD_TYPE_DDR_1_8V) ||
+	    (caps & MMC_CAP_1_2V_DDR &&
+			card_type & EXT_CSD_CARD_TYPE_DDR_1_2V))
+		hs_max_dtr = MMC_HIGH_DDR_MAX_DTR;
+
+	if ((caps2 & MMC_CAP2_HS200_1_8V_SDR &&
+			card_type & EXT_CSD_CARD_TYPE_SDR_1_8V) ||
+	    (caps2 & MMC_CAP2_HS200_1_2V_SDR &&
+			card_type & EXT_CSD_CARD_TYPE_SDR_1_2V))
+		hs_max_dtr = MMC_HS200_MAX_DTR;
+
+	card->ext_csd.hs_max_dtr = hs_max_dtr;
+	card->ext_csd.card_type = card_type;
+}
+
 /*
  * Decode extended CSD.
  */
@@ -284,56 +314,9 @@ static int mmc_read_ext_csd(struct mmc_card *card, u8 *ext_csd)
 		if (card->ext_csd.sectors > (2u * 1024 * 1024 * 1024) / 512)
 			mmc_card_set_blockaddr(card);
 	}
+
 	card->ext_csd.raw_card_type = ext_csd[EXT_CSD_CARD_TYPE];
-	switch (ext_csd[EXT_CSD_CARD_TYPE] & EXT_CSD_CARD_TYPE_MASK) {
-	case EXT_CSD_CARD_TYPE_SDR_ALL:
-	case EXT_CSD_CARD_TYPE_SDR_ALL_DDR_1_8V:
-	case EXT_CSD_CARD_TYPE_SDR_ALL_DDR_1_2V:
-	case EXT_CSD_CARD_TYPE_SDR_ALL_DDR_52:
-		card->ext_csd.hs_max_dtr = 200000000;
-		card->ext_csd.card_type = EXT_CSD_CARD_TYPE_SDR_200;
-		break;
-	case EXT_CSD_CARD_TYPE_SDR_1_2V_ALL:
-	case EXT_CSD_CARD_TYPE_SDR_1_2V_DDR_1_8V:
-	case EXT_CSD_CARD_TYPE_SDR_1_2V_DDR_1_2V:
-	case EXT_CSD_CARD_TYPE_SDR_1_2V_DDR_52:
-		card->ext_csd.hs_max_dtr = 200000000;
-		card->ext_csd.card_type = EXT_CSD_CARD_TYPE_SDR_1_2V;
-		break;
-	case EXT_CSD_CARD_TYPE_SDR_1_8V_ALL:
-	case EXT_CSD_CARD_TYPE_SDR_1_8V_DDR_1_8V:
-	case EXT_CSD_CARD_TYPE_SDR_1_8V_DDR_1_2V:
-	case EXT_CSD_CARD_TYPE_SDR_1_8V_DDR_52:
-		card->ext_csd.hs_max_dtr = 200000000;
-		card->ext_csd.card_type = EXT_CSD_CARD_TYPE_SDR_1_8V;
-		break;
-	case EXT_CSD_CARD_TYPE_DDR_52 | EXT_CSD_CARD_TYPE_52 |
-	     EXT_CSD_CARD_TYPE_26:
-		card->ext_csd.hs_max_dtr = 52000000;
-		card->ext_csd.card_type = EXT_CSD_CARD_TYPE_DDR_52;
-		break;
-	case EXT_CSD_CARD_TYPE_DDR_1_2V | EXT_CSD_CARD_TYPE_52 |
-	     EXT_CSD_CARD_TYPE_26:
-		card->ext_csd.hs_max_dtr = 52000000;
-		card->ext_csd.card_type = EXT_CSD_CARD_TYPE_DDR_1_2V;
-		break;
-	case EXT_CSD_CARD_TYPE_DDR_1_8V | EXT_CSD_CARD_TYPE_52 |
-	     EXT_CSD_CARD_TYPE_26:
-		card->ext_csd.hs_max_dtr = 52000000;
-		card->ext_csd.card_type = EXT_CSD_CARD_TYPE_DDR_1_8V;
-		break;
-	case EXT_CSD_CARD_TYPE_52 | EXT_CSD_CARD_TYPE_26:
-		card->ext_csd.hs_max_dtr = 52000000;
-		break;
-	case EXT_CSD_CARD_TYPE_26:
-		card->ext_csd.hs_max_dtr = 26000000;
-		break;
-	default:
-		/* MMC v4 spec says this cannot happen */
-		pr_warning("%s: card is mmc v4 but doesn't "
-			"support any high-speed modes.\n",
-			mmc_hostname(card->host));
-	}
+	mmc_select_card_type(card);
 
 	card->ext_csd.raw_s_a_timeout = ext_csd[EXT_CSD_S_A_TIMEOUT];
 	card->ext_csd.raw_erase_timeout_mult =
@@ -533,6 +516,8 @@ static int mmc_read_ext_csd(struct mmc_card *card, u8 *ext_csd)
 		} else {
 			card->ext_csd.data_tag_unit_size = 0;
 		}
+	} else {
+		card->ext_csd.data_sector_size = 512;
 	}
 
 out:
@@ -556,14 +541,10 @@ static int mmc_compare_ext_csds(struct mmc_card *card, unsigned bus_width)
 	err = mmc_get_ext_csd(card, &bw_ext_csd);
 
 	if (err || bw_ext_csd == NULL) {
-		if (bus_width != MMC_BUS_WIDTH_1)
-			err = -EINVAL;
+		err = -EINVAL;
 		goto out;
 	}
 
-	if (bus_width == MMC_BUS_WIDTH_1)
-		goto out;
-
 	/* only compare read only fields */
 	err = !((card->ext_csd.raw_partition_support ==
 			bw_ext_csd[EXT_CSD_PARTITION_SUPPORT]) &&
@@ -736,6 +717,10 @@ static int mmc_select_powerclass(struct mmc_card *card,
 				 card->ext_csd.generic_cmd6_time);
 	}
 
+	if (err)
+		pr_err("%s: power class selection for ext_csd_bus_width %d"
+		       " failed\n", mmc_hostname(card->host), bus_width);
+
 	return err;
 }
 
@@ -745,7 +730,7 @@ static int mmc_select_powerclass(struct mmc_card *card,
  */
 static int mmc_select_hs200(struct mmc_card *card)
 {
-	int idx, err = 0;
+	int idx, err = -EINVAL;
 	struct mmc_host *host;
 	static unsigned ext_csd_bits[] = {
 		EXT_CSD_BUS_WIDTH_4,
@@ -761,10 +746,12 @@ static int mmc_select_hs200(struct mmc_card *card)
 	host = card->host;
 
 	if (card->ext_csd.card_type & EXT_CSD_CARD_TYPE_SDR_1_2V &&
-	    host->caps2 & MMC_CAP2_HS200_1_2V_SDR)
-		if (mmc_set_signal_voltage(host, MMC_SIGNAL_VOLTAGE_120, 0))
-			err = mmc_set_signal_voltage(host,
-						     MMC_SIGNAL_VOLTAGE_180, 0);
+			host->caps2 & MMC_CAP2_HS200_1_2V_SDR)
+		err = mmc_set_signal_voltage(host, MMC_SIGNAL_VOLTAGE_120, 0);
+
+	if (err && card->ext_csd.card_type & EXT_CSD_CARD_TYPE_SDR_1_8V &&
+			host->caps2 & MMC_CAP2_HS200_1_8V_SDR)
+		err = mmc_set_signal_voltage(host, MMC_SIGNAL_VOLTAGE_180, 0);
 
 	/* If fails try again during next card power cycle */
 	if (err)
@@ -1117,9 +1104,7 @@ static int mmc_init_card(struct mmc_host *host, u32 ocr,
 				EXT_CSD_BUS_WIDTH_8 : EXT_CSD_BUS_WIDTH_4;
 		err = mmc_select_powerclass(card, ext_csd_bits, ext_csd);
 		if (err)
-			pr_warning("%s: power class selection to bus width %d"
-				   " failed\n", mmc_hostname(card->host),
-				   1 << bus_width);
+			goto err;
 	}
 
 	/*
@@ -1151,10 +1136,7 @@ static int mmc_init_card(struct mmc_host *host, u32 ocr,
 			err = mmc_select_powerclass(card, ext_csd_bits[idx][0],
 						    ext_csd);
 			if (err)
-				pr_warning("%s: power class selection to "
-					   "bus width %d failed\n",
-					   mmc_hostname(card->host),
-					   1 << bus_width);
+				goto err;
 
 			err = mmc_switch(card, EXT_CSD_CMD_SET_NORMAL,
 					 EXT_CSD_BUS_WIDTH,
@@ -1182,10 +1164,7 @@ static int mmc_init_card(struct mmc_host *host, u32 ocr,
 			err = mmc_select_powerclass(card, ext_csd_bits[idx][1],
 						    ext_csd);
 			if (err)
-				pr_warning("%s: power class selection to "
-					   "bus width %d ddr %d failed\n",
-					   mmc_hostname(card->host),
-					   1 << bus_width, ddr);
+				goto err;
 
 			err = mmc_switch(card, EXT_CSD_CMD_SET_NORMAL,
 					 EXT_CSD_BUS_WIDTH,
diff --git a/drivers/mmc/core/sdio.c b/drivers/mmc/core/sdio.c
index 2c7c83f832d2..13d0e95380ab 100644
--- a/drivers/mmc/core/sdio.c
+++ b/drivers/mmc/core/sdio.c
@@ -947,7 +947,7 @@ static int mmc_sdio_resume(struct mmc_host *host)
 	}
 
 	if (!err && host->sdio_irqs)
-		mmc_signal_sdio_irq(host);
+		wake_up_process(host->sdio_irq_thread);
 	mmc_release_host(host);
 
 	/*
diff --git a/drivers/mmc/core/sdio_irq.c b/drivers/mmc/core/sdio_irq.c
index f573e7f9f740..3d8ceb4084de 100644
--- a/drivers/mmc/core/sdio_irq.c
+++ b/drivers/mmc/core/sdio_irq.c
@@ -28,18 +28,20 @@
 
 #include "sdio_ops.h"
 
-static int process_sdio_pending_irqs(struct mmc_card *card)
+static int process_sdio_pending_irqs(struct mmc_host *host)
 {
+	struct mmc_card *card = host->card;
 	int i, ret, count;
 	unsigned char pending;
 	struct sdio_func *func;
 
 	/*
 	 * Optimization, if there is only 1 function interrupt registered
-	 * call irq handler directly
+	 * and we know an IRQ was signaled then call irq handler directly.
+	 * Otherwise do the full probe.
 	 */
 	func = card->sdio_single_irq;
-	if (func) {
+	if (func && host->sdio_irq_pending) {
 		func->irq_handler(func);
 		return 1;
 	}
@@ -116,7 +118,8 @@ static int sdio_irq_thread(void *_host)
 		ret = __mmc_claim_host(host, &host->sdio_irq_thread_abort);
 		if (ret)
 			break;
-		ret = process_sdio_pending_irqs(host->card);
+		ret = process_sdio_pending_irqs(host);
+		host->sdio_irq_pending = false;
 		mmc_release_host(host);
 
 		/*
diff --git a/drivers/mmc/host/Kconfig b/drivers/mmc/host/Kconfig
index 2bc06e7344db..aa131b32e3b2 100644
--- a/drivers/mmc/host/Kconfig
+++ b/drivers/mmc/host/Kconfig
@@ -278,10 +278,13 @@ choice
 	  Choose which driver to use for the Atmel MCI Silicon
 
 config MMC_AT91
-	tristate "AT91 SD/MMC Card Interface support"
+	tristate "AT91 SD/MMC Card Interface support (DEPRECATED)"
 	depends on ARCH_AT91
 	help
-	  This selects the AT91 MCI controller.
+	  This selects the AT91 MCI controller. This driver will
+	  be removed soon (for more information have a look to
+	  Documentation/feature-removal-schedule.txt). Please use
+	  MMC_ATMEL_MCI.
 
 	  If unsure, say N.
 
@@ -307,16 +310,6 @@ config MMC_ATMELMCI_DMA
 
 	  If unsure, say N.
 
-config MMC_IMX
-	tristate "Motorola i.MX Multimedia Card Interface support"
-	depends on ARCH_MX1
-	help
-	  This selects the Motorola i.MX Multimedia card Interface.
-	  If you have a i.MX platform with a Multimedia Card slot,
-	  say Y or M here.
-
-	  If unsure, say N.
-
 config MMC_MSM
 	tristate "Qualcomm SDCC Controller Support"
 	depends on MMC && ARCH_MSM
diff --git a/drivers/mmc/host/Makefile b/drivers/mmc/host/Makefile
index 3e7e26d08073..8922b06be925 100644
--- a/drivers/mmc/host/Makefile
+++ b/drivers/mmc/host/Makefile
@@ -4,7 +4,6 @@
 
 obj-$(CONFIG_MMC_ARMMMCI)	+= mmci.o
 obj-$(CONFIG_MMC_PXA)		+= pxamci.o
-obj-$(CONFIG_MMC_IMX)		+= imxmmc.o
 obj-$(CONFIG_MMC_MXC)		+= mxcmmc.o
 obj-$(CONFIG_MMC_MXS)		+= mxs-mmc.o
 obj-$(CONFIG_MMC_SDHCI)		+= sdhci.o
diff --git a/drivers/mmc/host/atmel-mci.c b/drivers/mmc/host/atmel-mci.c
index e94476beca18..420aca642b14 100644
--- a/drivers/mmc/host/atmel-mci.c
+++ b/drivers/mmc/host/atmel-mci.c
@@ -45,19 +45,19 @@
 #define ATMCI_DMA_THRESHOLD	16
 
 enum {
-	EVENT_CMD_COMPLETE = 0,
+	EVENT_CMD_RDY = 0,
 	EVENT_XFER_COMPLETE,
-	EVENT_DATA_COMPLETE,
+	EVENT_NOTBUSY,
 	EVENT_DATA_ERROR,
 };
 
 enum atmel_mci_state {
 	STATE_IDLE = 0,
 	STATE_SENDING_CMD,
-	STATE_SENDING_DATA,
-	STATE_DATA_BUSY,
+	STATE_DATA_XFER,
+	STATE_WAITING_NOTBUSY,
 	STATE_SENDING_STOP,
-	STATE_DATA_ERROR,
+	STATE_END_REQUEST,
 };
 
 enum atmci_xfer_dir {
@@ -78,6 +78,9 @@ struct atmel_mci_caps {
 	bool    has_highspeed;
 	bool    has_rwproof;
 	bool	has_odd_clk_div;
+	bool	has_bad_data_ordering;
+	bool	need_reset_after_xfer;
+	bool	need_blksz_mul_4;
 };
 
 struct atmel_mci_dma {
@@ -91,6 +94,11 @@ struct atmel_mci_dma {
  * @regs: Pointer to MMIO registers.
  * @sg: Scatterlist entry currently being processed by PIO or PDC code.
  * @pio_offset: Offset into the current scatterlist entry.
+ * @buffer: Buffer used if we don't have the r/w proof capability. We
+ *      don't have the time to switch pdc buffers so we have to use only
+ *      one buffer for the full transaction.
+ * @buf_size: size of the buffer.
+ * @phys_buf_addr: buffer address needed for pdc.
  * @cur_slot: The slot which is currently using the controller.
  * @mrq: The request currently being processed on @cur_slot,
  *	or NULL if the controller is idle.
@@ -116,6 +124,7 @@ struct atmel_mci_dma {
  * @queue: List of slots waiting for access to the controller.
  * @need_clock_update: Update the clock rate before the next request.
  * @need_reset: Reset controller before next request.
+ * @timer: Timer to balance the data timeout error flag which cannot rise.
  * @mode_reg: Value of the MR register.
  * @cfg_reg: Value of the CFG register.
  * @bus_hz: The rate of @mck in Hz. This forms the basis for MMC bus
@@ -166,6 +175,9 @@ struct atmel_mci {
 
 	struct scatterlist	*sg;
 	unsigned int		pio_offset;
+	unsigned int		*buffer;
+	unsigned int		buf_size;
+	dma_addr_t		buf_phys_addr;
 
 	struct atmel_mci_slot	*cur_slot;
 	struct mmc_request	*mrq;
@@ -189,6 +201,7 @@ struct atmel_mci {
 
 	bool			need_clock_update;
 	bool			need_reset;
+	struct timer_list	timer;
 	u32			mode_reg;
 	u32			cfg_reg;
 	unsigned long		bus_hz;
@@ -480,6 +493,32 @@ err:
 	dev_err(&mmc->class_dev, "failed to initialize debugfs for slot\n");
 }
 
+static inline unsigned int atmci_get_version(struct atmel_mci *host)
+{
+	return atmci_readl(host, ATMCI_VERSION) & 0x00000fff;
+}
+
+static void atmci_timeout_timer(unsigned long data)
+{
+	struct atmel_mci *host;
+
+	host = (struct atmel_mci *)data;
+
+	dev_dbg(&host->pdev->dev, "software timeout\n");
+
+	if (host->mrq->cmd->data) {
+		host->mrq->cmd->data->error = -ETIMEDOUT;
+		host->data = NULL;
+	} else {
+		host->mrq->cmd->error = -ETIMEDOUT;
+		host->cmd = NULL;
+	}
+	host->need_reset = 1;
+	host->state = STATE_END_REQUEST;
+	smp_wmb();
+	tasklet_schedule(&host->tasklet);
+}
+
 static inline unsigned int atmci_ns_to_clocks(struct atmel_mci *host,
 					unsigned int ns)
 {
@@ -591,6 +630,7 @@ static void atmci_send_command(struct atmel_mci *host,
 
 static void atmci_send_stop_cmd(struct atmel_mci *host, struct mmc_data *data)
 {
+	dev_dbg(&host->pdev->dev, "send stop command\n");
 	atmci_send_command(host, data->stop, host->stop_cmdr);
 	atmci_writel(host, ATMCI_IER, ATMCI_CMDRDY);
 }
@@ -603,6 +643,7 @@ static void atmci_pdc_set_single_buf(struct atmel_mci *host,
 	enum atmci_xfer_dir dir, enum atmci_pdc_buf buf_nb)
 {
 	u32 pointer_reg, counter_reg;
+	unsigned int buf_size;
 
 	if (dir == XFER_RECEIVE) {
 		pointer_reg = ATMEL_PDC_RPR;
@@ -617,8 +658,15 @@ static void atmci_pdc_set_single_buf(struct atmel_mci *host,
 		counter_reg += ATMEL_PDC_SCND_BUF_OFF;
 	}
 
-	atmci_writel(host, pointer_reg, sg_dma_address(host->sg));
-	if (host->data_size <= sg_dma_len(host->sg)) {
+	if (!host->caps.has_rwproof) {
+		buf_size = host->buf_size;
+		atmci_writel(host, pointer_reg, host->buf_phys_addr);
+	} else {
+		buf_size = sg_dma_len(host->sg);
+		atmci_writel(host, pointer_reg, sg_dma_address(host->sg));
+	}
+
+	if (host->data_size <= buf_size) {
 		if (host->data_size & 0x3) {
 			/* If size is different from modulo 4, transfer bytes */
 			atmci_writel(host, counter_reg, host->data_size);
@@ -670,7 +718,20 @@ static void atmci_pdc_cleanup(struct atmel_mci *host)
  */
 static void atmci_pdc_complete(struct atmel_mci *host)
 {
+	int transfer_size = host->data->blocks * host->data->blksz;
+	int i;
+
 	atmci_writel(host, ATMEL_PDC_PTCR, ATMEL_PDC_RXTDIS | ATMEL_PDC_TXTDIS);
+
+	if ((!host->caps.has_rwproof)
+	    && (host->data->flags & MMC_DATA_READ)) {
+		if (host->caps.has_bad_data_ordering)
+			for (i = 0; i < transfer_size; i++)
+				host->buffer[i] = swab32(host->buffer[i]);
+		sg_copy_from_buffer(host->data->sg, host->data->sg_len,
+		                    host->buffer, transfer_size);
+	}
+
 	atmci_pdc_cleanup(host);
 
 	/*
@@ -678,9 +739,10 @@ static void atmci_pdc_complete(struct atmel_mci *host)
 	 * to send the stop command or waiting for NBUSY in this case.
 	 */
 	if (host->data) {
+		dev_dbg(&host->pdev->dev,
+		        "(%s) set pending xfer complete\n", __func__);
 		atmci_set_pending(host, EVENT_XFER_COMPLETE);
 		tasklet_schedule(&host->tasklet);
-		atmci_writel(host, ATMCI_IER, ATMCI_NOTBUSY);
 	}
 }
 
@@ -716,6 +778,8 @@ static void atmci_dma_complete(void *arg)
 	 * to send the stop command or waiting for NBUSY in this case.
 	 */
 	if (data) {
+		dev_dbg(&host->pdev->dev,
+		        "(%s) set pending xfer complete\n", __func__);
 		atmci_set_pending(host, EVENT_XFER_COMPLETE);
 		tasklet_schedule(&host->tasklet);
 
@@ -791,6 +855,7 @@ atmci_prepare_data_pdc(struct atmel_mci *host, struct mmc_data *data)
 	u32 iflags, tmp;
 	unsigned int sg_len;
 	enum dma_data_direction dir;
+	int i;
 
 	data->error = -EINPROGRESS;
 
@@ -806,7 +871,7 @@ atmci_prepare_data_pdc(struct atmel_mci *host, struct mmc_data *data)
 		iflags |= ATMCI_ENDRX | ATMCI_RXBUFF;
 	} else {
 		dir = DMA_TO_DEVICE;
-		iflags |= ATMCI_ENDTX | ATMCI_TXBUFE;
+		iflags |= ATMCI_ENDTX | ATMCI_TXBUFE | ATMCI_BLKE;
 	}
 
 	/* Set BLKLEN */
@@ -818,6 +883,16 @@ atmci_prepare_data_pdc(struct atmel_mci *host, struct mmc_data *data)
 	/* Configure PDC */
 	host->data_size = data->blocks * data->blksz;
 	sg_len = dma_map_sg(&host->pdev->dev, data->sg, data->sg_len, dir);
+
+	if ((!host->caps.has_rwproof)
+	    && (host->data->flags & MMC_DATA_WRITE)) {
+		sg_copy_to_buffer(host->data->sg, host->data->sg_len,
+		                  host->buffer, host->data_size);
+		if (host->caps.has_bad_data_ordering)
+			for (i = 0; i < host->data_size; i++)
+				host->buffer[i] = swab32(host->buffer[i]);
+	}
+
 	if (host->data_size)
 		atmci_pdc_set_both_buf(host,
 			((dir == DMA_FROM_DEVICE) ? XFER_RECEIVE : XFER_TRANSMIT));
@@ -931,6 +1006,8 @@ atmci_submit_data_dma(struct atmel_mci *host, struct mmc_data *data)
 
 static void atmci_stop_transfer(struct atmel_mci *host)
 {
+	dev_dbg(&host->pdev->dev,
+	        "(%s) set pending xfer complete\n", __func__);
 	atmci_set_pending(host, EVENT_XFER_COMPLETE);
 	atmci_writel(host, ATMCI_IER, ATMCI_NOTBUSY);
 }
@@ -940,8 +1017,7 @@ static void atmci_stop_transfer(struct atmel_mci *host)
  */
 static void atmci_stop_transfer_pdc(struct atmel_mci *host)
 {
-	atmci_set_pending(host, EVENT_XFER_COMPLETE);
-	atmci_writel(host, ATMCI_IER, ATMCI_NOTBUSY);
+	atmci_writel(host, ATMEL_PDC_PTCR, ATMEL_PDC_RXTDIS | ATMEL_PDC_TXTDIS);
 }
 
 static void atmci_stop_transfer_dma(struct atmel_mci *host)
@@ -953,6 +1029,8 @@ static void atmci_stop_transfer_dma(struct atmel_mci *host)
 		atmci_dma_cleanup(host);
 	} else {
 		/* Data transfer was stopped by the interrupt handler */
+		dev_dbg(&host->pdev->dev,
+		        "(%s) set pending xfer complete\n", __func__);
 		atmci_set_pending(host, EVENT_XFER_COMPLETE);
 		atmci_writel(host, ATMCI_IER, ATMCI_NOTBUSY);
 	}
@@ -977,9 +1055,12 @@ static void atmci_start_request(struct atmel_mci *host,
 
 	host->pending_events = 0;
 	host->completed_events = 0;
+	host->cmd_status = 0;
 	host->data_status = 0;
 
-	if (host->need_reset) {
+	dev_dbg(&host->pdev->dev, "start request: cmd %u\n", mrq->cmd->opcode);
+
+	if (host->need_reset || host->caps.need_reset_after_xfer) {
 		iflags = atmci_readl(host, ATMCI_IMR);
 		iflags &= (ATMCI_SDIOIRQA | ATMCI_SDIOIRQB);
 		atmci_writel(host, ATMCI_CR, ATMCI_CR_SWRST);
@@ -994,7 +1075,7 @@ static void atmci_start_request(struct atmel_mci *host,
 
 	iflags = atmci_readl(host, ATMCI_IMR);
 	if (iflags & ~(ATMCI_SDIOIRQA | ATMCI_SDIOIRQB))
-		dev_warn(&slot->mmc->class_dev, "WARNING: IMR=0x%08x\n",
+		dev_dbg(&slot->mmc->class_dev, "WARNING: IMR=0x%08x\n",
 				iflags);
 
 	if (unlikely(test_and_clear_bit(ATMCI_CARD_NEED_INIT, &slot->flags))) {
@@ -1043,6 +1124,8 @@ static void atmci_start_request(struct atmel_mci *host,
 	 * prepared yet.)
 	 */
 	atmci_writel(host, ATMCI_IER, iflags);
+
+	mod_timer(&host->timer, jiffies +  msecs_to_jiffies(2000));
 }
 
 static void atmci_queue_request(struct atmel_mci *host,
@@ -1057,6 +1140,7 @@ static void atmci_queue_request(struct atmel_mci *host,
 		host->state = STATE_SENDING_CMD;
 		atmci_start_request(host, slot);
 	} else {
+		dev_dbg(&host->pdev->dev, "queue request\n");
 		list_add_tail(&slot->queue_node, &host->queue);
 	}
 	spin_unlock_bh(&host->lock);
@@ -1069,6 +1153,7 @@ static void atmci_request(struct mmc_host *mmc, struct mmc_request *mrq)
 	struct mmc_data		*data;
 
 	WARN_ON(slot->mrq);
+	dev_dbg(&host->pdev->dev, "MRQ: cmd %u\n", mrq->cmd->opcode);
 
 	/*
 	 * We may "know" the card is gone even though there's still an
@@ -1308,6 +1393,8 @@ static void atmci_request_end(struct atmel_mci *host, struct mmc_request *mrq)
 		host->state = STATE_IDLE;
 	}
 
+	del_timer(&host->timer);
+
 	spin_unlock(&host->lock);
 	mmc_request_done(prev_mmc, mrq);
 	spin_lock(&host->lock);
@@ -1330,21 +1417,13 @@ static void atmci_command_complete(struct atmel_mci *host,
 		cmd->error = -EILSEQ;
 	else if (status & (ATMCI_RINDE | ATMCI_RDIRE | ATMCI_RENDE))
 		cmd->error = -EIO;
-	else
-		cmd->error = 0;
-
-	if (cmd->error) {
-		dev_dbg(&host->pdev->dev,
-			"command error: status=0x%08x\n", status);
-
-		if (cmd->data) {
-			host->stop_transfer(host);
-			host->data = NULL;
-			atmci_writel(host, ATMCI_IDR, ATMCI_NOTBUSY
-					| ATMCI_TXRDY | ATMCI_RXRDY
-					| ATMCI_DATA_ERROR_FLAGS);
+	else if (host->mrq->data && (host->mrq->data->blksz & 3)) {
+		if (host->caps.need_blksz_mul_4) {
+			cmd->error = -EINVAL;
+			host->need_reset = 1;
 		}
-	}
+	} else
+		cmd->error = 0;
 }
 
 static void atmci_detect_change(unsigned long data)
@@ -1407,23 +1486,21 @@ static void atmci_detect_change(unsigned long data)
 					break;
 				case STATE_SENDING_CMD:
 					mrq->cmd->error = -ENOMEDIUM;
-					if (!mrq->data)
-						break;
-					/* fall through */
-				case STATE_SENDING_DATA:
+					if (mrq->data)
+						host->stop_transfer(host);
+					break;
+				case STATE_DATA_XFER:
 					mrq->data->error = -ENOMEDIUM;
 					host->stop_transfer(host);
 					break;
-				case STATE_DATA_BUSY:
-				case STATE_DATA_ERROR:
-					if (mrq->data->error == -EINPROGRESS)
-						mrq->data->error = -ENOMEDIUM;
-					if (!mrq->stop)
-						break;
-					/* fall through */
+				case STATE_WAITING_NOTBUSY:
+					mrq->data->error = -ENOMEDIUM;
+					break;
 				case STATE_SENDING_STOP:
 					mrq->stop->error = -ENOMEDIUM;
 					break;
+				case STATE_END_REQUEST:
+					break;
 				}
 
 				atmci_request_end(host, mrq);
@@ -1451,7 +1528,6 @@ static void atmci_tasklet_func(unsigned long priv)
 	struct atmel_mci	*host = (struct atmel_mci *)priv;
 	struct mmc_request	*mrq = host->mrq;
 	struct mmc_data		*data = host->data;
-	struct mmc_command	*cmd = host->cmd;
 	enum atmel_mci_state	state = host->state;
 	enum atmel_mci_state	prev_state;
 	u32			status;
@@ -1467,107 +1543,186 @@ static void atmci_tasklet_func(unsigned long priv)
 
 	do {
 		prev_state = state;
+		dev_dbg(&host->pdev->dev, "FSM: state=%d\n", state);
 
 		switch (state) {
 		case STATE_IDLE:
 			break;
 
 		case STATE_SENDING_CMD:
+			/*
+			 * Command has been sent, we are waiting for command
+			 * ready. Then we have three next states possible:
+			 * END_REQUEST by default, WAITING_NOTBUSY if it's a
+			 * command needing it or DATA_XFER if there is data.
+			 */
+			dev_dbg(&host->pdev->dev, "FSM: cmd ready?\n");
 			if (!atmci_test_and_clear_pending(host,
-						EVENT_CMD_COMPLETE))
+						EVENT_CMD_RDY))
 				break;
 
+			dev_dbg(&host->pdev->dev, "set completed cmd ready\n");
 			host->cmd = NULL;
-			atmci_set_completed(host, EVENT_CMD_COMPLETE);
+			atmci_set_completed(host, EVENT_CMD_RDY);
 			atmci_command_complete(host, mrq->cmd);
-			if (!mrq->data || cmd->error) {
-				atmci_request_end(host, host->mrq);
-				goto unlock;
-			}
+			if (mrq->data) {
+				dev_dbg(&host->pdev->dev,
+				        "command with data transfer");
+				/*
+				 * If there is a command error don't start
+				 * data transfer.
+				 */
+				if (mrq->cmd->error) {
+					host->stop_transfer(host);
+					host->data = NULL;
+					atmci_writel(host, ATMCI_IDR,
+					             ATMCI_TXRDY | ATMCI_RXRDY
+					             | ATMCI_DATA_ERROR_FLAGS);
+					state = STATE_END_REQUEST;
+				} else
+					state = STATE_DATA_XFER;
+			} else if ((!mrq->data) && (mrq->cmd->flags & MMC_RSP_BUSY)) {
+				dev_dbg(&host->pdev->dev,
+				        "command response need waiting notbusy");
+				atmci_writel(host, ATMCI_IER, ATMCI_NOTBUSY);
+				state = STATE_WAITING_NOTBUSY;
+			} else
+				state = STATE_END_REQUEST;
 
-			prev_state = state = STATE_SENDING_DATA;
-			/* fall through */
+			break;
 
-		case STATE_SENDING_DATA:
+		case STATE_DATA_XFER:
 			if (atmci_test_and_clear_pending(host,
 						EVENT_DATA_ERROR)) {
-				host->stop_transfer(host);
-				if (data->stop)
-					atmci_send_stop_cmd(host, data);
-				state = STATE_DATA_ERROR;
+				dev_dbg(&host->pdev->dev, "set completed data error\n");
+				atmci_set_completed(host, EVENT_DATA_ERROR);
+				state = STATE_END_REQUEST;
 				break;
 			}
 
+			/*
+			 * A data transfer is in progress. The event expected
+			 * to move to the next state depends of data transfer
+			 * type (PDC or DMA). Once transfer done we can move
+			 * to the next step which is WAITING_NOTBUSY in write
+			 * case and directly SENDING_STOP in read case.
+			 */
+			dev_dbg(&host->pdev->dev, "FSM: xfer complete?\n");
 			if (!atmci_test_and_clear_pending(host,
 						EVENT_XFER_COMPLETE))
 				break;
 
+			dev_dbg(&host->pdev->dev,
+			        "(%s) set completed xfer complete\n",
+				__func__);
 			atmci_set_completed(host, EVENT_XFER_COMPLETE);
-			prev_state = state = STATE_DATA_BUSY;
-			/* fall through */
 
-		case STATE_DATA_BUSY:
-			if (!atmci_test_and_clear_pending(host,
-						EVENT_DATA_COMPLETE))
-				break;
-
-			host->data = NULL;
-			atmci_set_completed(host, EVENT_DATA_COMPLETE);
-			status = host->data_status;
-			if (unlikely(status & ATMCI_DATA_ERROR_FLAGS)) {
-				if (status & ATMCI_DTOE) {
-					dev_dbg(&host->pdev->dev,
-							"data timeout error\n");
-					data->error = -ETIMEDOUT;
-				} else if (status & ATMCI_DCRCE) {
-					dev_dbg(&host->pdev->dev,
-							"data CRC error\n");
-					data->error = -EILSEQ;
-				} else {
-					dev_dbg(&host->pdev->dev,
-						"data FIFO error (status=%08x)\n",
-						status);
-					data->error = -EIO;
-				}
+			if (host->data->flags & MMC_DATA_WRITE) {
+				atmci_writel(host, ATMCI_IER, ATMCI_NOTBUSY);
+				state = STATE_WAITING_NOTBUSY;
+			} else if (host->mrq->stop) {
+				atmci_writel(host, ATMCI_IER, ATMCI_CMDRDY);
+				atmci_send_stop_cmd(host, data);
+				state = STATE_SENDING_STOP;
 			} else {
+				host->data = NULL;
 				data->bytes_xfered = data->blocks * data->blksz;
 				data->error = 0;
-				atmci_writel(host, ATMCI_IDR, ATMCI_DATA_ERROR_FLAGS);
+				state = STATE_END_REQUEST;
 			}
+			break;
 
-			if (!data->stop) {
-				atmci_request_end(host, host->mrq);
-				goto unlock;
-			}
+		case STATE_WAITING_NOTBUSY:
+			/*
+			 * We can be in the state for two reasons: a command
+			 * requiring waiting not busy signal (stop command
+			 * included) or a write operation. In the latest case,
+			 * we need to send a stop command.
+			 */
+			dev_dbg(&host->pdev->dev, "FSM: not busy?\n");
+			if (!atmci_test_and_clear_pending(host,
+						EVENT_NOTBUSY))
+				break;
 
-			prev_state = state = STATE_SENDING_STOP;
-			if (!data->error)
-				atmci_send_stop_cmd(host, data);
-			/* fall through */
+			dev_dbg(&host->pdev->dev, "set completed not busy\n");
+			atmci_set_completed(host, EVENT_NOTBUSY);
+
+			if (host->data) {
+				/*
+				 * For some commands such as CMD53, even if
+				 * there is data transfer, there is no stop
+				 * command to send.
+				 */
+				if (host->mrq->stop) {
+					atmci_writel(host, ATMCI_IER,
+					             ATMCI_CMDRDY);
+					atmci_send_stop_cmd(host, data);
+					state = STATE_SENDING_STOP;
+				} else {
+					host->data = NULL;
+					data->bytes_xfered = data->blocks
+					                     * data->blksz;
+					data->error = 0;
+					state = STATE_END_REQUEST;
+				}
+			} else
+				state = STATE_END_REQUEST;
+			break;
 
 		case STATE_SENDING_STOP:
+			/*
+			 * In this state, it is important to set host->data to
+			 * NULL (which is tested in the waiting notbusy state)
+			 * in order to go to the end request state instead of
+			 * sending stop again.
+			 */
+			dev_dbg(&host->pdev->dev, "FSM: cmd ready?\n");
 			if (!atmci_test_and_clear_pending(host,
-						EVENT_CMD_COMPLETE))
+						EVENT_CMD_RDY))
 				break;
 
+			dev_dbg(&host->pdev->dev, "FSM: cmd ready\n");
 			host->cmd = NULL;
+			host->data = NULL;
+			data->bytes_xfered = data->blocks * data->blksz;
+			data->error = 0;
 			atmci_command_complete(host, mrq->stop);
-			atmci_request_end(host, host->mrq);
-			goto unlock;
+			if (mrq->stop->error) {
+				host->stop_transfer(host);
+				atmci_writel(host, ATMCI_IDR,
+				             ATMCI_TXRDY | ATMCI_RXRDY
+				             | ATMCI_DATA_ERROR_FLAGS);
+				state = STATE_END_REQUEST;
+			} else {
+				atmci_writel(host, ATMCI_IER, ATMCI_NOTBUSY);
+				state = STATE_WAITING_NOTBUSY;
+			}
+			break;
 
-		case STATE_DATA_ERROR:
-			if (!atmci_test_and_clear_pending(host,
-						EVENT_XFER_COMPLETE))
-				break;
+		case STATE_END_REQUEST:
+			atmci_writel(host, ATMCI_IDR, ATMCI_TXRDY | ATMCI_RXRDY
+			                   | ATMCI_DATA_ERROR_FLAGS);
+			status = host->data_status;
+			if (unlikely(status)) {
+				host->stop_transfer(host);
+				host->data = NULL;
+				if (status & ATMCI_DTOE) {
+					data->error = -ETIMEDOUT;
+				} else if (status & ATMCI_DCRCE) {
+					data->error = -EILSEQ;
+				} else {
+					data->error = -EIO;
+				}
+			}
 
-			state = STATE_DATA_BUSY;
+			atmci_request_end(host, host->mrq);
+			state = STATE_IDLE;
 			break;
 		}
 	} while (state != prev_state);
 
 	host->state = state;
 
-unlock:
 	spin_unlock(&host->lock);
 }
 
@@ -1620,9 +1775,6 @@ static void atmci_read_data_pio(struct atmel_mci *host)
 						| ATMCI_DATA_ERROR_FLAGS));
 			host->data_status = status;
 			data->bytes_xfered += nbytes;
-			smp_wmb();
-			atmci_set_pending(host, EVENT_DATA_ERROR);
-			tasklet_schedule(&host->tasklet);
 			return;
 		}
 	} while (status & ATMCI_RXRDY);
@@ -1691,9 +1843,6 @@ static void atmci_write_data_pio(struct atmel_mci *host)
 						| ATMCI_DATA_ERROR_FLAGS));
 			host->data_status = status;
 			data->bytes_xfered += nbytes;
-			smp_wmb();
-			atmci_set_pending(host, EVENT_DATA_ERROR);
-			tasklet_schedule(&host->tasklet);
 			return;
 		}
 	} while (status & ATMCI_TXRDY);
@@ -1711,16 +1860,6 @@ done:
 	atmci_set_pending(host, EVENT_XFER_COMPLETE);
 }
 
-static void atmci_cmd_interrupt(struct atmel_mci *host, u32 status)
-{
-	atmci_writel(host, ATMCI_IDR, ATMCI_CMDRDY);
-
-	host->cmd_status = status;
-	smp_wmb();
-	atmci_set_pending(host, EVENT_CMD_COMPLETE);
-	tasklet_schedule(&host->tasklet);
-}
-
 static void atmci_sdio_interrupt(struct atmel_mci *host, u32 status)
 {
 	int	i;
@@ -1748,17 +1887,21 @@ static irqreturn_t atmci_interrupt(int irq, void *dev_id)
 			break;
 
 		if (pending & ATMCI_DATA_ERROR_FLAGS) {
+			dev_dbg(&host->pdev->dev, "IRQ: data error\n");
 			atmci_writel(host, ATMCI_IDR, ATMCI_DATA_ERROR_FLAGS
-					| ATMCI_RXRDY | ATMCI_TXRDY);
-			pending &= atmci_readl(host, ATMCI_IMR);
+					| ATMCI_RXRDY | ATMCI_TXRDY
+					| ATMCI_ENDRX | ATMCI_ENDTX
+					| ATMCI_RXBUFF | ATMCI_TXBUFE);
 
 			host->data_status = status;
+			dev_dbg(&host->pdev->dev, "set pending data error\n");
 			smp_wmb();
 			atmci_set_pending(host, EVENT_DATA_ERROR);
 			tasklet_schedule(&host->tasklet);
 		}
 
 		if (pending & ATMCI_TXBUFE) {
+			dev_dbg(&host->pdev->dev, "IRQ: tx buffer empty\n");
 			atmci_writel(host, ATMCI_IDR, ATMCI_TXBUFE);
 			atmci_writel(host, ATMCI_IDR, ATMCI_ENDTX);
 			/*
@@ -1774,6 +1917,7 @@ static irqreturn_t atmci_interrupt(int irq, void *dev_id)
 				atmci_pdc_complete(host);
 			}
 		} else if (pending & ATMCI_ENDTX) {
+			dev_dbg(&host->pdev->dev, "IRQ: end of tx buffer\n");
 			atmci_writel(host, ATMCI_IDR, ATMCI_ENDTX);
 
 			if (host->data_size) {
@@ -1784,6 +1928,7 @@ static irqreturn_t atmci_interrupt(int irq, void *dev_id)
 		}
 
 		if (pending & ATMCI_RXBUFF) {
+			dev_dbg(&host->pdev->dev, "IRQ: rx buffer full\n");
 			atmci_writel(host, ATMCI_IDR, ATMCI_RXBUFF);
 			atmci_writel(host, ATMCI_IDR, ATMCI_ENDRX);
 			/*
@@ -1799,6 +1944,7 @@ static irqreturn_t atmci_interrupt(int irq, void *dev_id)
 				atmci_pdc_complete(host);
 			}
 		} else if (pending & ATMCI_ENDRX) {
+			dev_dbg(&host->pdev->dev, "IRQ: end of rx buffer\n");
 			atmci_writel(host, ATMCI_IDR, ATMCI_ENDRX);
 
 			if (host->data_size) {
@@ -1808,23 +1954,44 @@ static irqreturn_t atmci_interrupt(int irq, void *dev_id)
 			}
 		}
 
+		/*
+		 * First mci IPs, so mainly the ones having pdc, have some
+		 * issues with the notbusy signal. You can't get it after
+		 * data transmission if you have not sent a stop command.
+		 * The appropriate workaround is to use the BLKE signal.
+		 */
+		if (pending & ATMCI_BLKE) {
+			dev_dbg(&host->pdev->dev, "IRQ: blke\n");
+			atmci_writel(host, ATMCI_IDR, ATMCI_BLKE);
+			smp_wmb();
+			dev_dbg(&host->pdev->dev, "set pending notbusy\n");
+			atmci_set_pending(host, EVENT_NOTBUSY);
+			tasklet_schedule(&host->tasklet);
+		}
 
 		if (pending & ATMCI_NOTBUSY) {
-			atmci_writel(host, ATMCI_IDR,
-					ATMCI_DATA_ERROR_FLAGS | ATMCI_NOTBUSY);
-			if (!host->data_status)
-				host->data_status = status;
+			dev_dbg(&host->pdev->dev, "IRQ: not_busy\n");
+			atmci_writel(host, ATMCI_IDR, ATMCI_NOTBUSY);
 			smp_wmb();
-			atmci_set_pending(host, EVENT_DATA_COMPLETE);
+			dev_dbg(&host->pdev->dev, "set pending notbusy\n");
+			atmci_set_pending(host, EVENT_NOTBUSY);
 			tasklet_schedule(&host->tasklet);
 		}
+
 		if (pending & ATMCI_RXRDY)
 			atmci_read_data_pio(host);
 		if (pending & ATMCI_TXRDY)
 			atmci_write_data_pio(host);
 
-		if (pending & ATMCI_CMDRDY)
-			atmci_cmd_interrupt(host, status);
+		if (pending & ATMCI_CMDRDY) {
+			dev_dbg(&host->pdev->dev, "IRQ: cmd ready\n");
+			atmci_writel(host, ATMCI_IDR, ATMCI_CMDRDY);
+			host->cmd_status = status;
+			smp_wmb();
+			dev_dbg(&host->pdev->dev, "set pending cmd rdy\n");
+			atmci_set_pending(host, EVENT_CMD_RDY);
+			tasklet_schedule(&host->tasklet);
+		}
 
 		if (pending & (ATMCI_SDIOIRQA | ATMCI_SDIOIRQB))
 			atmci_sdio_interrupt(host, status);
@@ -1877,13 +2044,26 @@ static int __init atmci_init_slot(struct atmel_mci *host,
 		mmc->caps |= MMC_CAP_SDIO_IRQ;
 	if (host->caps.has_highspeed)
 		mmc->caps |= MMC_CAP_SD_HIGHSPEED;
-	if (slot_data->bus_width >= 4)
+	/*
+	 * Without the read/write proof capability, it is strongly suggested to
+	 * use only one bit for data to prevent fifo underruns and overruns
+	 * which will corrupt data.
+	 */
+	if ((slot_data->bus_width >= 4) && host->caps.has_rwproof)
 		mmc->caps |= MMC_CAP_4_BIT_DATA;
 
-	mmc->max_segs = 64;
-	mmc->max_req_size = 32768 * 512;
-	mmc->max_blk_size = 32768;
-	mmc->max_blk_count = 512;
+	if (atmci_get_version(host) < 0x200) {
+		mmc->max_segs = 256;
+		mmc->max_blk_size = 4095;
+		mmc->max_blk_count = 256;
+		mmc->max_req_size = mmc->max_blk_size * mmc->max_blk_count;
+		mmc->max_seg_size = mmc->max_blk_size * mmc->max_segs;
+	} else {
+		mmc->max_segs = 64;
+		mmc->max_req_size = 32768 * 512;
+		mmc->max_blk_size = 32768;
+		mmc->max_blk_count = 512;
+	}
 
 	/* Assume card is present initially */
 	set_bit(ATMCI_CARD_PRESENT, &slot->flags);
@@ -2007,11 +2187,6 @@ static bool atmci_configure_dma(struct atmel_mci *host)
 	}
 }
 
-static inline unsigned int atmci_get_version(struct atmel_mci *host)
-{
-	return atmci_readl(host, ATMCI_VERSION) & 0x00000fff;
-}
-
 /*
  * HSMCI (High Speed MCI) module is not fully compatible with MCI module.
  * HSMCI provides DMA support and a new config register but no more supports
@@ -2032,6 +2207,9 @@ static void __init atmci_get_cap(struct atmel_mci *host)
 	host->caps.has_highspeed = 0;
 	host->caps.has_rwproof = 0;
 	host->caps.has_odd_clk_div = 0;
+	host->caps.has_bad_data_ordering = 1;
+	host->caps.need_reset_after_xfer = 1;
+	host->caps.need_blksz_mul_4 = 1;
 
 	/* keep only major version number */
 	switch (version & 0xf00) {
@@ -2051,7 +2229,11 @@ static void __init atmci_get_cap(struct atmel_mci *host)
 		host->caps.has_highspeed = 1;
 	case 0x200:
 		host->caps.has_rwproof = 1;
+		host->caps.need_blksz_mul_4 = 0;
 	case 0x100:
+		host->caps.has_bad_data_ordering = 0;
+		host->caps.need_reset_after_xfer = 0;
+	case 0x0:
 		break;
 	default:
 		host->caps.has_pdc = 0;
@@ -2138,14 +2320,20 @@ static int __init atmci_probe(struct platform_device *pdev)
 	if (pdata->slot[0].bus_width) {
 		ret = atmci_init_slot(host, &pdata->slot[0],
 				0, ATMCI_SDCSEL_SLOT_A, ATMCI_SDIOIRQA);
-		if (!ret)
+		if (!ret) {
 			nr_slots++;
+			host->buf_size = host->slot[0]->mmc->max_req_size;
+		}
 	}
 	if (pdata->slot[1].bus_width) {
 		ret = atmci_init_slot(host, &pdata->slot[1],
 				1, ATMCI_SDCSEL_SLOT_B, ATMCI_SDIOIRQB);
-		if (!ret)
+		if (!ret) {
 			nr_slots++;
+			if (host->slot[1]->mmc->max_req_size > host->buf_size)
+				host->buf_size =
+					host->slot[1]->mmc->max_req_size;
+		}
 	}
 
 	if (!nr_slots) {
@@ -2153,6 +2341,19 @@ static int __init atmci_probe(struct platform_device *pdev)
 		goto err_init_slot;
 	}
 
+	if (!host->caps.has_rwproof) {
+		host->buffer = dma_alloc_coherent(&pdev->dev, host->buf_size,
+		                                  &host->buf_phys_addr,
+						  GFP_KERNEL);
+		if (!host->buffer) {
+			ret = -ENOMEM;
+			dev_err(&pdev->dev, "buffer allocation failed\n");
+			goto err_init_slot;
+		}
+	}
+
+	setup_timer(&host->timer, atmci_timeout_timer, (unsigned long)host);
+
 	dev_info(&pdev->dev,
 			"Atmel MCI controller at 0x%08lx irq %d, %u slots\n",
 			host->mapbase, irq, nr_slots);
@@ -2179,6 +2380,10 @@ static int __exit atmci_remove(struct platform_device *pdev)
 
 	platform_set_drvdata(pdev, NULL);
 
+	if (host->buffer)
+		dma_free_coherent(&pdev->dev, host->buf_size,
+		                  host->buffer, host->buf_phys_addr);
+
 	for (i = 0; i < ATMCI_MAX_NR_SLOTS; i++) {
 		if (host->slot[i])
 			atmci_cleanup_slot(host->slot[i], i);
diff --git a/drivers/mmc/host/davinci_mmc.c b/drivers/mmc/host/davinci_mmc.c
index c1f3673ae1ef..7cf6c624bf73 100644
--- a/drivers/mmc/host/davinci_mmc.c
+++ b/drivers/mmc/host/davinci_mmc.c
@@ -1533,4 +1533,5 @@ module_exit(davinci_mmcsd_exit);
 MODULE_AUTHOR("Texas Instruments India");
 MODULE_LICENSE("GPL");
 MODULE_DESCRIPTION("MMC/SD driver for Davinci MMC controller");
+MODULE_ALIAS("platform:davinci_mmc");
 
diff --git a/drivers/mmc/host/dw_mmc.c b/drivers/mmc/host/dw_mmc.c
index ab3fc4617107..9bbf45f8c538 100644
--- a/drivers/mmc/host/dw_mmc.c
+++ b/drivers/mmc/host/dw_mmc.c
@@ -100,8 +100,6 @@ struct dw_mci_slot {
 	int			last_detect_state;
 };
 
-static struct workqueue_struct *dw_mci_card_workqueue;
-
 #if defined(CONFIG_DEBUG_FS)
 static int dw_mci_req_show(struct seq_file *s, void *v)
 {
@@ -859,10 +857,10 @@ static void dw_mci_enable_sdio_irq(struct mmc_host *mmc, int enb)
 	int_mask = mci_readl(host, INTMASK);
 	if (enb) {
 		mci_writel(host, INTMASK,
-			   (int_mask | (1 << SDMMC_INT_SDIO(slot->id))));
+			   (int_mask | SDMMC_INT_SDIO(slot->id)));
 	} else {
 		mci_writel(host, INTMASK,
-			   (int_mask & ~(1 << SDMMC_INT_SDIO(slot->id))));
+			   (int_mask & ~SDMMC_INT_SDIO(slot->id)));
 	}
 }
 
@@ -1605,7 +1603,7 @@ static irqreturn_t dw_mci_interrupt(int irq, void *dev_id)
 
 		if (pending & SDMMC_INT_CD) {
 			mci_writel(host, RINTSTS, SDMMC_INT_CD);
-			queue_work(dw_mci_card_workqueue, &host->card_work);
+			queue_work(host->card_workqueue, &host->card_work);
 		}
 
 		/* Handle SDIO Interrupts */
@@ -1844,7 +1842,7 @@ static int __init dw_mci_init_slot(struct dw_mci *host, unsigned int id)
 	 * Card may have been plugged in prior to boot so we
 	 * need to run the detect tasklet
 	 */
-	queue_work(dw_mci_card_workqueue, &host->card_work);
+	queue_work(host->card_workqueue, &host->card_work);
 
 	return 0;
 }
@@ -2021,9 +2019,9 @@ int dw_mci_probe(struct dw_mci *host)
 	mci_writel(host, CLKSRC, 0);
 
 	tasklet_init(&host->tasklet, dw_mci_tasklet_func, (unsigned long)host);
-	dw_mci_card_workqueue = alloc_workqueue("dw-mci-card",
+	host->card_workqueue = alloc_workqueue("dw-mci-card",
 			WQ_MEM_RECLAIM | WQ_NON_REENTRANT, 1);
-	if (!dw_mci_card_workqueue)
+	if (!host->card_workqueue)
 		goto err_dmaunmap;
 	INIT_WORK(&host->card_work, dw_mci_work_routine_card);
 	ret = request_irq(host->irq, dw_mci_interrupt, host->irq_flags, "dw-mci", host);
@@ -2085,7 +2083,7 @@ err_init_slot:
 	free_irq(host->irq, host);
 
 err_workqueue:
-	destroy_workqueue(dw_mci_card_workqueue);
+	destroy_workqueue(host->card_workqueue);
 
 err_dmaunmap:
 	if (host->use_dma && host->dma_ops->exit)
@@ -2119,7 +2117,7 @@ void dw_mci_remove(struct dw_mci *host)
 	mci_writel(host, CLKSRC, 0);
 
 	free_irq(host->irq, host);
-	destroy_workqueue(dw_mci_card_workqueue);
+	destroy_workqueue(host->card_workqueue);
 	dma_free_coherent(&host->dev, PAGE_SIZE, host->sg_cpu, host->sg_dma);
 
 	if (host->use_dma && host->dma_ops->exit)
diff --git a/drivers/mmc/host/imxmmc.c b/drivers/mmc/host/imxmmc.c
deleted file mode 100644
index ea0f3cedef21..000000000000
--- a/drivers/mmc/host/imxmmc.c
+++ /dev/null
@@ -1,1169 +0,0 @@
-/*
- *  linux/drivers/mmc/host/imxmmc.c - Motorola i.MX MMCI driver
- *
- *  Copyright (C) 2004 Sascha Hauer, Pengutronix <sascha@saschahauer.de>
- *  Copyright (C) 2006 Pavel Pisa, PiKRON <ppisa@pikron.com>
- *
- *  derived from pxamci.c by Russell King
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of the GNU General Public License version 2 as
- * published by the Free Software Foundation.
- *
- */
-
-#include <linux/module.h>
-#include <linux/init.h>
-#include <linux/ioport.h>
-#include <linux/platform_device.h>
-#include <linux/interrupt.h>
-#include <linux/blkdev.h>
-#include <linux/dma-mapping.h>
-#include <linux/mmc/host.h>
-#include <linux/mmc/card.h>
-#include <linux/delay.h>
-#include <linux/clk.h>
-#include <linux/io.h>
-
-#include <asm/dma.h>
-#include <asm/irq.h>
-#include <asm/sizes.h>
-#include <mach/mmc.h>
-#include <mach/imx-dma.h>
-
-#include "imxmmc.h"
-
-#define DRIVER_NAME "imx-mmc"
-
-#define IMXMCI_INT_MASK_DEFAULT (INT_MASK_BUF_READY | INT_MASK_DATA_TRAN | \
-				 INT_MASK_WRITE_OP_DONE | INT_MASK_END_CMD_RES | \
-				 INT_MASK_AUTO_CARD_DETECT | INT_MASK_DAT0_EN | INT_MASK_SDIO)
-
-struct imxmci_host {
-	struct mmc_host		*mmc;
-	spinlock_t		lock;
-	struct resource		*res;
-	void __iomem		*base;
-	int			irq;
-	imx_dmach_t		dma;
-	volatile unsigned int	imask;
-	unsigned int		power_mode;
-	unsigned int		present;
-	struct imxmmc_platform_data *pdata;
-
-	struct mmc_request	*req;
-	struct mmc_command	*cmd;
-	struct mmc_data		*data;
-
-	struct timer_list	timer;
-	struct tasklet_struct	tasklet;
-	unsigned int		status_reg;
-	unsigned long		pending_events;
-	/* Next two fields are there for CPU driven transfers to overcome SDHC deficiencies */
-	u16			*data_ptr;
-	unsigned int		data_cnt;
-	atomic_t		stuck_timeout;
-
-	unsigned int		dma_nents;
-	unsigned int		dma_size;
-	unsigned int		dma_dir;
-	int			dma_allocated;
-
-	unsigned char		actual_bus_width;
-
-	int			prev_cmd_code;
-
-	struct clk		*clk;
-};
-
-#define IMXMCI_PEND_IRQ_b	0
-#define IMXMCI_PEND_DMA_END_b	1
-#define IMXMCI_PEND_DMA_ERR_b	2
-#define IMXMCI_PEND_WAIT_RESP_b	3
-#define IMXMCI_PEND_DMA_DATA_b	4
-#define IMXMCI_PEND_CPU_DATA_b	5
-#define IMXMCI_PEND_CARD_XCHG_b	6
-#define IMXMCI_PEND_SET_INIT_b	7
-#define IMXMCI_PEND_STARTED_b	8
-
-#define IMXMCI_PEND_IRQ_m	(1 << IMXMCI_PEND_IRQ_b)
-#define IMXMCI_PEND_DMA_END_m	(1 << IMXMCI_PEND_DMA_END_b)
-#define IMXMCI_PEND_DMA_ERR_m	(1 << IMXMCI_PEND_DMA_ERR_b)
-#define IMXMCI_PEND_WAIT_RESP_m	(1 << IMXMCI_PEND_WAIT_RESP_b)
-#define IMXMCI_PEND_DMA_DATA_m	(1 << IMXMCI_PEND_DMA_DATA_b)
-#define IMXMCI_PEND_CPU_DATA_m	(1 << IMXMCI_PEND_CPU_DATA_b)
-#define IMXMCI_PEND_CARD_XCHG_m	(1 << IMXMCI_PEND_CARD_XCHG_b)
-#define IMXMCI_PEND_SET_INIT_m	(1 << IMXMCI_PEND_SET_INIT_b)
-#define IMXMCI_PEND_STARTED_m	(1 << IMXMCI_PEND_STARTED_b)
-
-static void imxmci_stop_clock(struct imxmci_host *host)
-{
-	int i = 0;
-	u16 reg;
-
-	reg = readw(host->base + MMC_REG_STR_STP_CLK);
-	writew(reg & ~STR_STP_CLK_START_CLK, host->base + MMC_REG_STR_STP_CLK);
-	while (i < 0x1000) {
-		if (!(i & 0x7f)) {
-			reg = readw(host->base + MMC_REG_STR_STP_CLK);
-			writew(reg | STR_STP_CLK_STOP_CLK,
-					host->base + MMC_REG_STR_STP_CLK);
-		}
-
-		reg = readw(host->base + MMC_REG_STATUS);
-		if (!(reg & STATUS_CARD_BUS_CLK_RUN)) {
-			/* Check twice before cut */
-			reg = readw(host->base + MMC_REG_STATUS);
-			if (!(reg & STATUS_CARD_BUS_CLK_RUN))
-				return;
-		}
-
-		i++;
-	}
-	dev_dbg(mmc_dev(host->mmc), "imxmci_stop_clock blocked, no luck\n");
-}
-
-static int imxmci_start_clock(struct imxmci_host *host)
-{
-	unsigned int trials = 0;
-	unsigned int delay_limit = 128;
-	unsigned long flags;
-	u16 reg;
-
-	reg = readw(host->base + MMC_REG_STR_STP_CLK);
-	writew(reg & ~STR_STP_CLK_STOP_CLK, host->base + MMC_REG_STR_STP_CLK);
-
-	clear_bit(IMXMCI_PEND_STARTED_b, &host->pending_events);
-
-	/*
-	 * Command start of the clock, this usually succeeds in less
-	 * then 6 delay loops, but during card detection (low clockrate)
-	 * it takes up to 5000 delay loops and sometimes fails for the first time
-	 */
-	reg = readw(host->base + MMC_REG_STR_STP_CLK);
-	writew(reg | STR_STP_CLK_START_CLK, host->base + MMC_REG_STR_STP_CLK);
-
-	do {
-		unsigned int delay = delay_limit;
-
-		while (delay--) {
-			reg = readw(host->base + MMC_REG_STATUS);
-			if (reg & STATUS_CARD_BUS_CLK_RUN) {
-				/* Check twice before cut */
-				reg = readw(host->base + MMC_REG_STATUS);
-				if (reg & STATUS_CARD_BUS_CLK_RUN)
-					return 0;
-			}
-
-			if (test_bit(IMXMCI_PEND_STARTED_b, &host->pending_events))
-				return 0;
-		}
-
-		local_irq_save(flags);
-		/*
-		 * Ensure, that request is not doubled under all possible circumstances.
-		 * It is possible, that cock running state is missed, because some other
-		 * IRQ or schedule delays this function execution and the clocks has
-		 * been already stopped by other means (response processing, SDHC HW)
-		 */
-		if (!test_bit(IMXMCI_PEND_STARTED_b, &host->pending_events)) {
-			reg = readw(host->base + MMC_REG_STR_STP_CLK);
-			writew(reg | STR_STP_CLK_START_CLK,
-					host->base + MMC_REG_STR_STP_CLK);
-		}
-		local_irq_restore(flags);
-
-	} while (++trials < 256);
-
-	dev_err(mmc_dev(host->mmc), "imxmci_start_clock blocked, no luck\n");
-
-	return -1;
-}
-
-static void imxmci_softreset(struct imxmci_host *host)
-{
-	int i;
-
-	/* reset sequence */
-	writew(0x08, host->base + MMC_REG_STR_STP_CLK);
-	writew(0x0D, host->base + MMC_REG_STR_STP_CLK);
-
-	for (i = 0; i < 8; i++)
-		writew(0x05, host->base + MMC_REG_STR_STP_CLK);
-
-	writew(0xff, host->base + MMC_REG_RES_TO);
-	writew(512, host->base + MMC_REG_BLK_LEN);
-	writew(1, host->base + MMC_REG_NOB);
-}
-
-static int imxmci_busy_wait_for_status(struct imxmci_host *host,
-				       unsigned int *pstat, unsigned int stat_mask,
-				       int timeout, const char *where)
-{
-	int loops = 0;
-
-	while (!(*pstat & stat_mask)) {
-		loops += 2;
-		if (loops >= timeout) {
-			dev_dbg(mmc_dev(host->mmc), "busy wait timeout in %s, STATUS = 0x%x (0x%x)\n",
-				where, *pstat, stat_mask);
-			return -1;
-		}
-		udelay(2);
-		*pstat |= readw(host->base + MMC_REG_STATUS);
-	}
-	if (!loops)
-		return 0;
-
-	/* The busy-wait is expected there for clock <8MHz due to SDHC hardware flaws */
-	if (!(stat_mask & STATUS_END_CMD_RESP) || (host->mmc->ios.clock >= 8000000))
-		dev_info(mmc_dev(host->mmc), "busy wait for %d usec in %s, STATUS = 0x%x (0x%x)\n",
-			 loops, where, *pstat, stat_mask);
-	return loops;
-}
-
-static void imxmci_setup_data(struct imxmci_host *host, struct mmc_data *data)
-{
-	unsigned int nob = data->blocks;
-	unsigned int blksz = data->blksz;
-	unsigned int datasz = nob * blksz;
-	int i;
-
-	if (data->flags & MMC_DATA_STREAM)
-		nob = 0xffff;
-
-	host->data = data;
-	data->bytes_xfered = 0;
-
-	writew(nob, host->base + MMC_REG_NOB);
-	writew(blksz, host->base + MMC_REG_BLK_LEN);
-
-	/*
-	 * DMA cannot be used for small block sizes, we have to use CPU driven transfers otherwise.
-	 * We are in big troubles for non-512 byte transfers according to note in the paragraph
-	 * 20.6.7 of User Manual anyway, but we need to be able to transfer SCR at least.
-	 * The situation is even more complex in reality. The SDHC in not able to handle wll
-	 * partial FIFO fills and reads. The length has to be rounded up to burst size multiple.
-	 * This is required for SCR read at least.
-	 */
-	if (datasz < 512) {
-		host->dma_size = datasz;
-		if (data->flags & MMC_DATA_READ) {
-			host->dma_dir = DMA_FROM_DEVICE;
-
-			/* Hack to enable read SCR */
-			writew(1, host->base + MMC_REG_NOB);
-			writew(512, host->base + MMC_REG_BLK_LEN);
-		} else {
-			host->dma_dir = DMA_TO_DEVICE;
-		}
-
-		/* Convert back to virtual address */
-		host->data_ptr = (u16 *)sg_virt(data->sg);
-		host->data_cnt = 0;
-
-		clear_bit(IMXMCI_PEND_DMA_DATA_b, &host->pending_events);
-		set_bit(IMXMCI_PEND_CPU_DATA_b, &host->pending_events);
-
-		return;
-	}
-
-	if (data->flags & MMC_DATA_READ) {
-		host->dma_dir = DMA_FROM_DEVICE;
-		host->dma_nents = dma_map_sg(mmc_dev(host->mmc), data->sg,
-					     data->sg_len,  host->dma_dir);
-
-		imx_dma_setup_sg(host->dma, data->sg, data->sg_len, datasz,
-				 host->res->start + MMC_REG_BUFFER_ACCESS,
-				 DMA_MODE_READ);
-
-		/*imx_dma_setup_mem2dev_ccr(host->dma, DMA_MODE_READ, IMX_DMA_WIDTH_16, CCR_REN);*/
-		CCR(host->dma) = CCR_DMOD_LINEAR | CCR_DSIZ_32 | CCR_SMOD_FIFO | CCR_SSIZ_16 | CCR_REN;
-	} else {
-		host->dma_dir = DMA_TO_DEVICE;
-
-		host->dma_nents = dma_map_sg(mmc_dev(host->mmc), data->sg,
-					     data->sg_len,  host->dma_dir);
-
-		imx_dma_setup_sg(host->dma, data->sg, data->sg_len, datasz,
-				 host->res->start + MMC_REG_BUFFER_ACCESS,
-				 DMA_MODE_WRITE);
-
-		/*imx_dma_setup_mem2dev_ccr(host->dma, DMA_MODE_WRITE, IMX_DMA_WIDTH_16, CCR_REN);*/
-		CCR(host->dma) = CCR_SMOD_LINEAR | CCR_SSIZ_32 | CCR_DMOD_FIFO | CCR_DSIZ_16 | CCR_REN;
-	}
-
-#if 1	/* This code is there only for consistency checking and can be disabled in future */
-	host->dma_size = 0;
-	for (i = 0; i < host->dma_nents; i++)
-		host->dma_size += data->sg[i].length;
-
-	if (datasz > host->dma_size) {
-		dev_err(mmc_dev(host->mmc), "imxmci_setup_data datasz 0x%x > 0x%x dm_size\n",
-			datasz, host->dma_size);
-	}
-#endif
-
-	host->dma_size = datasz;
-
-	wmb();
-
-	set_bit(IMXMCI_PEND_DMA_DATA_b, &host->pending_events);
-	clear_bit(IMXMCI_PEND_CPU_DATA_b, &host->pending_events);
-
-	/* start DMA engine for read, write is delayed after initial response */
-	if (host->dma_dir == DMA_FROM_DEVICE)
-		imx_dma_enable(host->dma);
-}
-
-static void imxmci_start_cmd(struct imxmci_host *host, struct mmc_command *cmd, unsigned int cmdat)
-{
-	unsigned long flags;
-	u32 imask;
-
-	WARN_ON(host->cmd != NULL);
-	host->cmd = cmd;
-
-	/* Ensure, that clock are stopped else command programming and start fails */
-	imxmci_stop_clock(host);
-
-	if (cmd->flags & MMC_RSP_BUSY)
-		cmdat |= CMD_DAT_CONT_BUSY;
-
-	switch (mmc_resp_type(cmd)) {
-	case MMC_RSP_R1: /* short CRC, OPCODE */
-	case MMC_RSP_R1B:/* short CRC, OPCODE, BUSY */
-		cmdat |= CMD_DAT_CONT_RESPONSE_FORMAT_R1;
-		break;
-	case MMC_RSP_R2: /* long 136 bit + CRC */
-		cmdat |= CMD_DAT_CONT_RESPONSE_FORMAT_R2;
-		break;
-	case MMC_RSP_R3: /* short */
-		cmdat |= CMD_DAT_CONT_RESPONSE_FORMAT_R3;
-		break;
-	default:
-		break;
-	}
-
-	if (test_and_clear_bit(IMXMCI_PEND_SET_INIT_b, &host->pending_events))
-		cmdat |= CMD_DAT_CONT_INIT; /* This command needs init */
-
-	if (host->actual_bus_width == MMC_BUS_WIDTH_4)
-		cmdat |= CMD_DAT_CONT_BUS_WIDTH_4;
-
-	writew(cmd->opcode, host->base + MMC_REG_CMD);
-	writew(cmd->arg >> 16, host->base + MMC_REG_ARGH);
-	writew(cmd->arg & 0xffff, host->base + MMC_REG_ARGL);
-	writew(cmdat, host->base + MMC_REG_CMD_DAT_CONT);
-
-	atomic_set(&host->stuck_timeout, 0);
-	set_bit(IMXMCI_PEND_WAIT_RESP_b, &host->pending_events);
-
-
-	imask = IMXMCI_INT_MASK_DEFAULT;
-	imask &= ~INT_MASK_END_CMD_RES;
-	if (cmdat & CMD_DAT_CONT_DATA_ENABLE) {
-		/* imask &= ~INT_MASK_BUF_READY; */
-		imask &= ~INT_MASK_DATA_TRAN;
-		if (cmdat & CMD_DAT_CONT_WRITE)
-			imask &= ~INT_MASK_WRITE_OP_DONE;
-		if (test_bit(IMXMCI_PEND_CPU_DATA_b, &host->pending_events))
-			imask &= ~INT_MASK_BUF_READY;
-	}
-
-	spin_lock_irqsave(&host->lock, flags);
-	host->imask = imask;
-	writew(host->imask, host->base + MMC_REG_INT_MASK);
-	spin_unlock_irqrestore(&host->lock, flags);
-
-	dev_dbg(mmc_dev(host->mmc), "CMD%02d (0x%02x) mask set to 0x%04x\n",
-		cmd->opcode, cmd->opcode, imask);
-
-	imxmci_start_clock(host);
-}
-
-static void imxmci_finish_request(struct imxmci_host *host, struct mmc_request *req)
-{
-	unsigned long flags;
-
-	spin_lock_irqsave(&host->lock, flags);
-
-	host->pending_events &= ~(IMXMCI_PEND_WAIT_RESP_m | IMXMCI_PEND_DMA_END_m |
-				  IMXMCI_PEND_DMA_DATA_m | IMXMCI_PEND_CPU_DATA_m);
-
-	host->imask = IMXMCI_INT_MASK_DEFAULT;
-	writew(host->imask, host->base + MMC_REG_INT_MASK);
-
-	spin_unlock_irqrestore(&host->lock, flags);
-
-	if (req && req->cmd)
-		host->prev_cmd_code = req->cmd->opcode;
-
-	host->req = NULL;
-	host->cmd = NULL;
-	host->data = NULL;
-	mmc_request_done(host->mmc, req);
-}
-
-static int imxmci_finish_data(struct imxmci_host *host, unsigned int stat)
-{
-	struct mmc_data *data = host->data;
-	int data_error;
-
-	if (test_and_clear_bit(IMXMCI_PEND_DMA_DATA_b, &host->pending_events)) {
-		imx_dma_disable(host->dma);
-		dma_unmap_sg(mmc_dev(host->mmc), data->sg, host->dma_nents,
-			     host->dma_dir);
-	}
-
-	if (stat & STATUS_ERR_MASK) {
-		dev_dbg(mmc_dev(host->mmc), "request failed. status: 0x%08x\n", stat);
-		if (stat & (STATUS_CRC_READ_ERR | STATUS_CRC_WRITE_ERR))
-			data->error = -EILSEQ;
-		else if (stat & STATUS_TIME_OUT_READ)
-			data->error = -ETIMEDOUT;
-		else
-			data->error = -EIO;
-	} else {
-		data->bytes_xfered = host->dma_size;
-	}
-
-	data_error = data->error;
-
-	host->data = NULL;
-
-	return data_error;
-}
-
-static int imxmci_cmd_done(struct imxmci_host *host, unsigned int stat)
-{
-	struct mmc_command *cmd = host->cmd;
-	int i;
-	u32 a, b, c;
-	struct mmc_data *data = host->data;
-
-	if (!cmd)
-		return 0;
-
-	host->cmd = NULL;
-
-	if (stat & STATUS_TIME_OUT_RESP) {
-		dev_dbg(mmc_dev(host->mmc), "CMD TIMEOUT\n");
-		cmd->error = -ETIMEDOUT;
-	} else if (stat & STATUS_RESP_CRC_ERR && cmd->flags & MMC_RSP_CRC) {
-		dev_dbg(mmc_dev(host->mmc), "cmd crc error\n");
-		cmd->error = -EILSEQ;
-	}
-
-	if (cmd->flags & MMC_RSP_PRESENT) {
-		if (cmd->flags & MMC_RSP_136) {
-			for (i = 0; i < 4; i++) {
-				a = readw(host->base + MMC_REG_RES_FIFO);
-				b = readw(host->base + MMC_REG_RES_FIFO);
-				cmd->resp[i] = a << 16 | b;
-			}
-		} else {
-			a = readw(host->base + MMC_REG_RES_FIFO);
-			b = readw(host->base + MMC_REG_RES_FIFO);
-			c = readw(host->base + MMC_REG_RES_FIFO);
-			cmd->resp[0] = a << 24 | b << 8 | c >> 8;
-		}
-	}
-
-	dev_dbg(mmc_dev(host->mmc), "RESP 0x%08x, 0x%08x, 0x%08x, 0x%08x, error %d\n",
-		cmd->resp[0], cmd->resp[1], cmd->resp[2], cmd->resp[3], cmd->error);
-
-	if (data && !cmd->error && !(stat & STATUS_ERR_MASK)) {
-		if (host->req->data->flags & MMC_DATA_WRITE) {
-
-			/* Wait for FIFO to be empty before starting DMA write */
-
-			stat = readw(host->base + MMC_REG_STATUS);
-			if (imxmci_busy_wait_for_status(host, &stat,
-							STATUS_APPL_BUFF_FE,
-							40, "imxmci_cmd_done DMA WR") < 0) {
-				cmd->error = -EIO;
-				imxmci_finish_data(host, stat);
-				if (host->req)
-					imxmci_finish_request(host, host->req);
-				dev_warn(mmc_dev(host->mmc), "STATUS = 0x%04x\n",
-					 stat);
-				return 0;
-			}
-
-			if (test_bit(IMXMCI_PEND_DMA_DATA_b, &host->pending_events))
-				imx_dma_enable(host->dma);
-		}
-	} else {
-		struct mmc_request *req;
-		imxmci_stop_clock(host);
-		req = host->req;
-
-		if (data)
-			imxmci_finish_data(host, stat);
-
-		if (req)
-			imxmci_finish_request(host, req);
-		else
-			dev_warn(mmc_dev(host->mmc), "imxmci_cmd_done: no request to finish\n");
-	}
-
-	return 1;
-}
-
-static int imxmci_data_done(struct imxmci_host *host, unsigned int stat)
-{
-	struct mmc_data *data = host->data;
-	int data_error;
-
-	if (!data)
-		return 0;
-
-	data_error = imxmci_finish_data(host, stat);
-
-	if (host->req->stop) {
-		imxmci_stop_clock(host);
-		imxmci_start_cmd(host, host->req->stop, 0);
-	} else {
-		struct mmc_request *req;
-		req = host->req;
-		if (req)
-			imxmci_finish_request(host, req);
-		else
-			dev_warn(mmc_dev(host->mmc), "imxmci_data_done: no request to finish\n");
-	}
-
-	return 1;
-}
-
-static int imxmci_cpu_driven_data(struct imxmci_host *host, unsigned int *pstat)
-{
-	int i;
-	int burst_len;
-	int trans_done = 0;
-	unsigned int stat = *pstat;
-
-	if (host->actual_bus_width != MMC_BUS_WIDTH_4)
-		burst_len = 16;
-	else
-		burst_len = 64;
-
-	/* This is unfortunately required */
-	dev_dbg(mmc_dev(host->mmc), "imxmci_cpu_driven_data running STATUS = 0x%x\n",
-		stat);
-
-	udelay(20);	/* required for clocks < 8MHz*/
-
-	if (host->dma_dir == DMA_FROM_DEVICE) {
-		imxmci_busy_wait_for_status(host, &stat,
-					    STATUS_APPL_BUFF_FF | STATUS_DATA_TRANS_DONE |
-					    STATUS_TIME_OUT_READ,
-					    50, "imxmci_cpu_driven_data read");
-
-		while ((stat & (STATUS_APPL_BUFF_FF | STATUS_DATA_TRANS_DONE)) &&
-		       !(stat & STATUS_TIME_OUT_READ) &&
-		       (host->data_cnt < 512)) {
-
-			udelay(20);	/* required for clocks < 8MHz*/
-
-			for (i = burst_len; i >= 2 ; i -= 2) {
-				u16 data;
-				data = readw(host->base + MMC_REG_BUFFER_ACCESS);
-				udelay(10);	/* required for clocks < 8MHz*/
-				if (host->data_cnt+2 <= host->dma_size) {
-					*(host->data_ptr++) = data;
-				} else {
-					if (host->data_cnt < host->dma_size)
-						*(u8 *)(host->data_ptr) = data;
-				}
-				host->data_cnt += 2;
-			}
-
-			stat = readw(host->base + MMC_REG_STATUS);
-
-			dev_dbg(mmc_dev(host->mmc), "imxmci_cpu_driven_data read %d burst %d STATUS = 0x%x\n",
-				host->data_cnt, burst_len, stat);
-		}
-
-		if ((stat & STATUS_DATA_TRANS_DONE) && (host->data_cnt >= 512))
-			trans_done = 1;
-
-		if (host->dma_size & 0x1ff)
-			stat &= ~STATUS_CRC_READ_ERR;
-
-		if (stat & STATUS_TIME_OUT_READ) {
-			dev_dbg(mmc_dev(host->mmc), "imxmci_cpu_driven_data read timeout STATUS = 0x%x\n",
-				stat);
-			trans_done = -1;
-		}
-
-	} else {
-		imxmci_busy_wait_for_status(host, &stat,
-					    STATUS_APPL_BUFF_FE,
-					    20, "imxmci_cpu_driven_data write");
-
-		while ((stat & STATUS_APPL_BUFF_FE) &&
-		       (host->data_cnt < host->dma_size)) {
-			if (burst_len >= host->dma_size - host->data_cnt) {
-				burst_len = host->dma_size - host->data_cnt;
-				host->data_cnt = host->dma_size;
-				trans_done = 1;
-			} else {
-				host->data_cnt += burst_len;
-			}
-
-			for (i = burst_len; i > 0 ; i -= 2)
-				writew(*(host->data_ptr++), host->base + MMC_REG_BUFFER_ACCESS);
-
-			stat = readw(host->base + MMC_REG_STATUS);
-
-			dev_dbg(mmc_dev(host->mmc), "imxmci_cpu_driven_data write burst %d STATUS = 0x%x\n",
-				burst_len, stat);
-		}
-	}
-
-	*pstat = stat;
-
-	return trans_done;
-}
-
-static void imxmci_dma_irq(int dma, void *devid)
-{
-	struct imxmci_host *host = devid;
-	u32 stat = readw(host->base + MMC_REG_STATUS);
-
-	atomic_set(&host->stuck_timeout, 0);
-	host->status_reg = stat;
-	set_bit(IMXMCI_PEND_DMA_END_b, &host->pending_events);
-	tasklet_schedule(&host->tasklet);
-}
-
-static irqreturn_t imxmci_irq(int irq, void *devid)
-{
-	struct imxmci_host *host = devid;
-	u32 stat = readw(host->base + MMC_REG_STATUS);
-	int handled = 1;
-
-	writew(host->imask | INT_MASK_SDIO | INT_MASK_AUTO_CARD_DETECT,
-			host->base + MMC_REG_INT_MASK);
-
-	atomic_set(&host->stuck_timeout, 0);
-	host->status_reg = stat;
-	set_bit(IMXMCI_PEND_IRQ_b, &host->pending_events);
-	set_bit(IMXMCI_PEND_STARTED_b, &host->pending_events);
-	tasklet_schedule(&host->tasklet);
-
-	return IRQ_RETVAL(handled);
-}
-
-static void imxmci_tasklet_fnc(unsigned long data)
-{
-	struct imxmci_host *host = (struct imxmci_host *)data;
-	u32 stat;
-	unsigned int data_dir_mask = 0;	/* STATUS_WR_CRC_ERROR_CODE_MASK */
-	int timeout = 0;
-
-	if (atomic_read(&host->stuck_timeout) > 4) {
-		char *what;
-		timeout = 1;
-		stat = readw(host->base + MMC_REG_STATUS);
-		host->status_reg = stat;
-		if (test_bit(IMXMCI_PEND_WAIT_RESP_b, &host->pending_events))
-			if (test_bit(IMXMCI_PEND_DMA_DATA_b, &host->pending_events))
-				what = "RESP+DMA";
-			else
-				what = "RESP";
-		else
-			if (test_bit(IMXMCI_PEND_DMA_DATA_b, &host->pending_events))
-				if (test_bit(IMXMCI_PEND_DMA_END_b, &host->pending_events))
-					what = "DATA";
-				else
-					what = "DMA";
-			else
-				what = "???";
-
-		dev_err(mmc_dev(host->mmc),
-			"%s TIMEOUT, hardware stucked STATUS = 0x%04x IMASK = 0x%04x\n",
-			what, stat,
-			readw(host->base + MMC_REG_INT_MASK));
-		dev_err(mmc_dev(host->mmc),
-			"CMD_DAT_CONT = 0x%04x, MMC_BLK_LEN = 0x%04x, MMC_NOB = 0x%04x, DMA_CCR = 0x%08x\n",
-			readw(host->base + MMC_REG_CMD_DAT_CONT),
-			readw(host->base + MMC_REG_BLK_LEN),
-			readw(host->base + MMC_REG_NOB),
-			CCR(host->dma));
-		dev_err(mmc_dev(host->mmc), "CMD%d, prevCMD%d, bus %d-bit, dma_size = 0x%x\n",
-			host->cmd ? host->cmd->opcode : 0,
-			host->prev_cmd_code,
-			1 << host->actual_bus_width, host->dma_size);
-	}
-
-	if (!host->present || timeout)
-		host->status_reg = STATUS_TIME_OUT_RESP | STATUS_TIME_OUT_READ |
-			STATUS_CRC_READ_ERR | STATUS_CRC_WRITE_ERR;
-
-	if (test_bit(IMXMCI_PEND_IRQ_b, &host->pending_events) || timeout) {
-		clear_bit(IMXMCI_PEND_IRQ_b, &host->pending_events);
-
-		stat = readw(host->base + MMC_REG_STATUS);
-		/*
-		 * This is not required in theory, but there is chance to miss some flag
-		 * which clears automatically by mask write, FreeScale original code keeps
-		 * stat from IRQ time so do I
-		 */
-		stat |= host->status_reg;
-
-		if (test_bit(IMXMCI_PEND_CPU_DATA_b, &host->pending_events))
-			stat &= ~STATUS_CRC_READ_ERR;
-
-		if (test_bit(IMXMCI_PEND_WAIT_RESP_b, &host->pending_events)) {
-			imxmci_busy_wait_for_status(host, &stat,
-						    STATUS_END_CMD_RESP | STATUS_ERR_MASK,
-						    20, "imxmci_tasklet_fnc resp (ERRATUM #4)");
-		}
-
-		if (stat & (STATUS_END_CMD_RESP | STATUS_ERR_MASK)) {
-			if (test_and_clear_bit(IMXMCI_PEND_WAIT_RESP_b, &host->pending_events))
-				imxmci_cmd_done(host, stat);
-			if (host->data && (stat & STATUS_ERR_MASK))
-				imxmci_data_done(host, stat);
-		}
-
-		if (test_bit(IMXMCI_PEND_CPU_DATA_b, &host->pending_events)) {
-			stat |= readw(host->base + MMC_REG_STATUS);
-			if (imxmci_cpu_driven_data(host, &stat)) {
-				if (test_and_clear_bit(IMXMCI_PEND_WAIT_RESP_b, &host->pending_events))
-					imxmci_cmd_done(host, stat);
-				atomic_clear_mask(IMXMCI_PEND_IRQ_m|IMXMCI_PEND_CPU_DATA_m,
-						  &host->pending_events);
-				imxmci_data_done(host, stat);
-			}
-		}
-	}
-
-	if (test_bit(IMXMCI_PEND_DMA_END_b, &host->pending_events) &&
-	    !test_bit(IMXMCI_PEND_WAIT_RESP_b, &host->pending_events)) {
-
-		stat = readw(host->base + MMC_REG_STATUS);
-		/* Same as above */
-		stat |= host->status_reg;
-
-		if (host->dma_dir == DMA_TO_DEVICE)
-			data_dir_mask = STATUS_WRITE_OP_DONE;
-		else
-			data_dir_mask = STATUS_DATA_TRANS_DONE;
-
-		if (stat & data_dir_mask) {
-			clear_bit(IMXMCI_PEND_DMA_END_b, &host->pending_events);
-			imxmci_data_done(host, stat);
-		}
-	}
-
-	if (test_and_clear_bit(IMXMCI_PEND_CARD_XCHG_b, &host->pending_events)) {
-
-		if (host->cmd)
-			imxmci_cmd_done(host, STATUS_TIME_OUT_RESP);
-
-		if (host->data)
-			imxmci_data_done(host, STATUS_TIME_OUT_READ |
-					 STATUS_CRC_READ_ERR | STATUS_CRC_WRITE_ERR);
-
-		if (host->req)
-			imxmci_finish_request(host, host->req);
-
-		mmc_detect_change(host->mmc, msecs_to_jiffies(100));
-
-	}
-}
-
-static void imxmci_request(struct mmc_host *mmc, struct mmc_request *req)
-{
-	struct imxmci_host *host = mmc_priv(mmc);
-	unsigned int cmdat;
-
-	WARN_ON(host->req != NULL);
-
-	host->req = req;
-
-	cmdat = 0;
-
-	if (req->data) {
-		imxmci_setup_data(host, req->data);
-
-		cmdat |= CMD_DAT_CONT_DATA_ENABLE;
-
-		if (req->data->flags & MMC_DATA_WRITE)
-			cmdat |= CMD_DAT_CONT_WRITE;
-
-		if (req->data->flags & MMC_DATA_STREAM)
-			cmdat |= CMD_DAT_CONT_STREAM_BLOCK;
-	}
-
-	imxmci_start_cmd(host, req->cmd, cmdat);
-}
-
-#define CLK_RATE 19200000
-
-static void imxmci_set_ios(struct mmc_host *mmc, struct mmc_ios *ios)
-{
-	struct imxmci_host *host = mmc_priv(mmc);
-	int prescaler;
-
-	if (ios->bus_width == MMC_BUS_WIDTH_4) {
-		host->actual_bus_width = MMC_BUS_WIDTH_4;
-		imx_gpio_mode(PB11_PF_SD_DAT3);
-		BLR(host->dma) = 0;	/* burst 64 byte read/write */
-	} else {
-		host->actual_bus_width = MMC_BUS_WIDTH_1;
-		imx_gpio_mode(GPIO_PORTB | GPIO_IN | GPIO_PUEN | 11);
-		BLR(host->dma) = 16;	/* burst 16 byte read/write */
-	}
-
-	if (host->power_mode != ios->power_mode) {
-		switch (ios->power_mode) {
-		case MMC_POWER_OFF:
-			break;
-		case MMC_POWER_UP:
-			set_bit(IMXMCI_PEND_SET_INIT_b, &host->pending_events);
-			break;
-		case MMC_POWER_ON:
-			break;
-		}
-		host->power_mode = ios->power_mode;
-	}
-
-	if (ios->clock) {
-		unsigned int clk;
-		u16 reg;
-
-		/* The prescaler is 5 for PERCLK2 equal to 96MHz
-		 * then 96MHz / 5 = 19.2 MHz
-		 */
-		clk = clk_get_rate(host->clk);
-		prescaler = (clk + (CLK_RATE * 7) / 8) / CLK_RATE;
-		switch (prescaler) {
-		case 0:
-		case 1:	prescaler = 0;
-			break;
-		case 2:	prescaler = 1;
-			break;
-		case 3:	prescaler = 2;
-			break;
-		case 4:	prescaler = 4;
-			break;
-		default:
-		case 5:	prescaler = 5;
-			break;
-		}
-
-		dev_dbg(mmc_dev(host->mmc), "PERCLK2 %d MHz -> prescaler %d\n",
-			clk, prescaler);
-
-		for (clk = 0; clk < 8; clk++) {
-			int x;
-			x = CLK_RATE / (1 << clk);
-			if (x <= ios->clock)
-				break;
-		}
-
-		/* enable controller */
-		reg = readw(host->base + MMC_REG_STR_STP_CLK);
-		writew(reg | STR_STP_CLK_ENABLE,
-				host->base + MMC_REG_STR_STP_CLK);
-
-		imxmci_stop_clock(host);
-		writew((prescaler << 3) | clk, host->base + MMC_REG_CLK_RATE);
-		/*
-		 * Under my understanding, clock should not be started there, because it would
-		 * initiate SDHC sequencer and send last or random command into card
-		 */
-		/* imxmci_start_clock(host); */
-
-		dev_dbg(mmc_dev(host->mmc),
-			"MMC_CLK_RATE: 0x%08x\n",
-			readw(host->base + MMC_REG_CLK_RATE));
-	} else {
-		imxmci_stop_clock(host);
-	}
-}
-
-static int imxmci_get_ro(struct mmc_host *mmc)
-{
-	struct imxmci_host *host = mmc_priv(mmc);
-
-	if (host->pdata && host->pdata->get_ro)
-		return !!host->pdata->get_ro(mmc_dev(mmc));
-	/*
-	 * Board doesn't support read only detection; let the mmc core
-	 * decide what to do.
-	 */
-	return -ENOSYS;
-}
-
-
-static const struct mmc_host_ops imxmci_ops = {
-	.request	= imxmci_request,
-	.set_ios	= imxmci_set_ios,
-	.get_ro		= imxmci_get_ro,
-};
-
-static void imxmci_check_status(unsigned long data)
-{
-	struct imxmci_host *host = (struct imxmci_host *)data;
-
-	if (host->pdata && host->pdata->card_present &&
-	    host->pdata->card_present(mmc_dev(host->mmc)) != host->present) {
-		host->present ^= 1;
-		dev_info(mmc_dev(host->mmc), "card %s\n",
-		      host->present ? "inserted" : "removed");
-
-		set_bit(IMXMCI_PEND_CARD_XCHG_b, &host->pending_events);
-		tasklet_schedule(&host->tasklet);
-	}
-
-	if (test_bit(IMXMCI_PEND_WAIT_RESP_b, &host->pending_events) ||
-	    test_bit(IMXMCI_PEND_DMA_DATA_b, &host->pending_events)) {
-		atomic_inc(&host->stuck_timeout);
-		if (atomic_read(&host->stuck_timeout) > 4)
-			tasklet_schedule(&host->tasklet);
-	} else {
-		atomic_set(&host->stuck_timeout, 0);
-
-	}
-
-	mod_timer(&host->timer, jiffies + (HZ>>1));
-}
-
-static int __init imxmci_probe(struct platform_device *pdev)
-{
-	struct mmc_host *mmc;
-	struct imxmci_host *host = NULL;
-	struct resource *r;
-	int ret = 0, irq;
-	u16 rev_no;
-
-	pr_info("i.MX mmc driver\n");
-
-	r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
-	irq = platform_get_irq(pdev, 0);
-	if (!r || irq < 0)
-		return -ENXIO;
-
-	r = request_mem_region(r->start, resource_size(r), pdev->name);
-	if (!r)
-		return -EBUSY;
-
-	mmc = mmc_alloc_host(sizeof(struct imxmci_host), &pdev->dev);
-	if (!mmc) {
-		ret = -ENOMEM;
-		goto out;
-	}
-
-	mmc->ops = &imxmci_ops;
-	mmc->f_min = 150000;
-	mmc->f_max = CLK_RATE/2;
-	mmc->ocr_avail = MMC_VDD_32_33;
-	mmc->caps = MMC_CAP_4_BIT_DATA;
-
-	/* MMC core transfer sizes tunable parameters */
-	mmc->max_segs = 64;
-	mmc->max_seg_size = 64*512;	/* default PAGE_CACHE_SIZE */
-	mmc->max_req_size = 64*512;	/* default PAGE_CACHE_SIZE */
-	mmc->max_blk_size = 2048;
-	mmc->max_blk_count = 65535;
-
-	host = mmc_priv(mmc);
-	host->base = ioremap(r->start, resource_size(r));
-	if (!host->base) {
-		ret = -ENOMEM;
-		goto out;
-	}
-
-	host->mmc = mmc;
-	host->dma_allocated = 0;
-	host->pdata = pdev->dev.platform_data;
-	if (!host->pdata)
-		dev_warn(&pdev->dev, "No platform data provided!\n");
-
-	spin_lock_init(&host->lock);
-	host->res = r;
-	host->irq = irq;
-
-	host->clk = clk_get(&pdev->dev, "perclk2");
-	if (IS_ERR(host->clk)) {
-		ret = PTR_ERR(host->clk);
-		goto out;
-	}
-	clk_enable(host->clk);
-
-	imx_gpio_mode(PB8_PF_SD_DAT0);
-	imx_gpio_mode(PB9_PF_SD_DAT1);
-	imx_gpio_mode(PB10_PF_SD_DAT2);
-	/* Configured as GPIO with pull-up to ensure right MCC card mode */
-	/* Switched to PB11_PF_SD_DAT3 if 4 bit bus is configured */
-	imx_gpio_mode(GPIO_PORTB | GPIO_IN | GPIO_PUEN | 11);
-	/* imx_gpio_mode(PB11_PF_SD_DAT3); */
-	imx_gpio_mode(PB12_PF_SD_CLK);
-	imx_gpio_mode(PB13_PF_SD_CMD);
-
-	imxmci_softreset(host);
-
-	rev_no = readw(host->base + MMC_REG_REV_NO);
-	if (rev_no != 0x390) {
-		dev_err(mmc_dev(host->mmc), "wrong rev.no. 0x%08x. aborting.\n",
-			readw(host->base + MMC_REG_REV_NO));
-		goto out;
-	}
-
-	/* recommended in data sheet */
-	writew(0x2db4, host->base + MMC_REG_READ_TO);
-
-	host->imask = IMXMCI_INT_MASK_DEFAULT;
-	writew(host->imask, host->base + MMC_REG_INT_MASK);
-
-	host->dma = imx_dma_request_by_prio(DRIVER_NAME, DMA_PRIO_LOW);
-	if(host->dma < 0) {
-		dev_err(mmc_dev(host->mmc), "imx_dma_request_by_prio failed\n");
-		ret = -EBUSY;
-		goto out;
-	}
-	host->dma_allocated = 1;
-	imx_dma_setup_handlers(host->dma, imxmci_dma_irq, NULL, host);
-	RSSR(host->dma) = DMA_REQ_SDHC;
-
-	tasklet_init(&host->tasklet, imxmci_tasklet_fnc, (unsigned long)host);
-	host->status_reg=0;
-	host->pending_events=0;
-
-	ret = request_irq(host->irq, imxmci_irq, 0, DRIVER_NAME, host);
-	if (ret)
-		goto out;
-
-	if (host->pdata && host->pdata->card_present)
-		host->present = host->pdata->card_present(mmc_dev(mmc));
-	else	/* if there is no way to detect assume that card is present */
-		host->present = 1;
-
-	init_timer(&host->timer);
-	host->timer.data = (unsigned long)host;
-	host->timer.function = imxmci_check_status;
-	add_timer(&host->timer);
-	mod_timer(&host->timer, jiffies + (HZ >> 1));
-
-	platform_set_drvdata(pdev, mmc);
-
-	mmc_add_host(mmc);
-
-	return 0;
-
-out:
-	if (host) {
-		if (host->dma_allocated) {
-			imx_dma_free(host->dma);
-			host->dma_allocated = 0;
-		}
-		if (host->clk) {
-			clk_disable(host->clk);
-			clk_put(host->clk);
-		}
-		if (host->base)
-			iounmap(host->base);
-	}
-	if (mmc)
-		mmc_free_host(mmc);
-	release_mem_region(r->start, resource_size(r));
-	return ret;
-}
-
-static int __exit imxmci_remove(struct platform_device *pdev)
-{
-	struct mmc_host *mmc = platform_get_drvdata(pdev);
-
-	platform_set_drvdata(pdev, NULL);
-
-	if (mmc) {
-		struct imxmci_host *host = mmc_priv(mmc);
-
-		tasklet_disable(&host->tasklet);
-
-		del_timer_sync(&host->timer);
-		mmc_remove_host(mmc);
-
-		free_irq(host->irq, host);
-		iounmap(host->base);
-		if (host->dma_allocated) {
-			imx_dma_free(host->dma);
-			host->dma_allocated = 0;
-		}
-
-		tasklet_kill(&host->tasklet);
-
-		clk_disable(host->clk);
-		clk_put(host->clk);
-
-		release_mem_region(host->res->start, resource_size(host->res));
-
-		mmc_free_host(mmc);
-	}
-	return 0;
-}
-
-#ifdef CONFIG_PM
-static int imxmci_suspend(struct platform_device *dev, pm_message_t state)
-{
-	struct mmc_host *mmc = platform_get_drvdata(dev);
-	int ret = 0;
-
-	if (mmc)
-		ret = mmc_suspend_host(mmc);
-
-	return ret;
-}
-
-static int imxmci_resume(struct platform_device *dev)
-{
-	struct mmc_host *mmc = platform_get_drvdata(dev);
-	struct imxmci_host *host;
-	int ret = 0;
-
-	if (mmc) {
-		host = mmc_priv(mmc);
-		if (host)
-			set_bit(IMXMCI_PEND_SET_INIT_b, &host->pending_events);
-		ret = mmc_resume_host(mmc);
-	}
-
-	return ret;
-}
-#else
-#define imxmci_suspend  NULL
-#define imxmci_resume   NULL
-#endif /* CONFIG_PM */
-
-static struct platform_driver imxmci_driver = {
-	.remove		= __exit_p(imxmci_remove),
-	.suspend	= imxmci_suspend,
-	.resume		= imxmci_resume,
-	.driver		= {
-		.name		= DRIVER_NAME,
-		.owner		= THIS_MODULE,
-	}
-};
-
-static int __init imxmci_init(void)
-{
-	return platform_driver_probe(&imxmci_driver, imxmci_probe);
-}
-
-static void __exit imxmci_exit(void)
-{
-	platform_driver_unregister(&imxmci_driver);
-}
-
-module_init(imxmci_init);
-module_exit(imxmci_exit);
-
-MODULE_DESCRIPTION("i.MX Multimedia Card Interface Driver");
-MODULE_AUTHOR("Sascha Hauer, Pengutronix");
-MODULE_LICENSE("GPL");
-MODULE_ALIAS("platform:imx-mmc");
diff --git a/drivers/mmc/host/imxmmc.h b/drivers/mmc/host/imxmmc.h
deleted file mode 100644
index 09d5d4ee3a77..000000000000
--- a/drivers/mmc/host/imxmmc.h
+++ /dev/null
@@ -1,64 +0,0 @@
-#define MMC_REG_STR_STP_CLK		0x00
-#define MMC_REG_STATUS			0x04
-#define MMC_REG_CLK_RATE		0x08
-#define MMC_REG_CMD_DAT_CONT		0x0C
-#define MMC_REG_RES_TO			0x10
-#define MMC_REG_READ_TO			0x14
-#define MMC_REG_BLK_LEN			0x18
-#define MMC_REG_NOB			0x1C
-#define MMC_REG_REV_NO			0x20
-#define MMC_REG_INT_MASK		0x24
-#define MMC_REG_CMD			0x28
-#define MMC_REG_ARGH			0x2C
-#define MMC_REG_ARGL			0x30
-#define MMC_REG_RES_FIFO		0x34
-#define MMC_REG_BUFFER_ACCESS		0x38
-
-#define STR_STP_CLK_IPG_CLK_GATE_DIS    (1<<15)
-#define STR_STP_CLK_IPG_PERCLK_GATE_DIS (1<<14)
-#define STR_STP_CLK_ENDIAN              (1<<5)
-#define STR_STP_CLK_RESET               (1<<3)
-#define STR_STP_CLK_ENABLE              (1<<2)
-#define STR_STP_CLK_START_CLK           (1<<1)
-#define STR_STP_CLK_STOP_CLK            (1<<0)
-#define STATUS_CARD_PRESENCE            (1<<15)
-#define STATUS_SDIO_INT_ACTIVE          (1<<14)
-#define STATUS_END_CMD_RESP             (1<<13)
-#define STATUS_WRITE_OP_DONE            (1<<12)
-#define STATUS_DATA_TRANS_DONE          (1<<11)
-#define STATUS_WR_CRC_ERROR_CODE_MASK   (3<<10)
-#define STATUS_CARD_BUS_CLK_RUN         (1<<8)
-#define STATUS_APPL_BUFF_FF             (1<<7)
-#define STATUS_APPL_BUFF_FE             (1<<6)
-#define STATUS_RESP_CRC_ERR             (1<<5)
-#define STATUS_CRC_READ_ERR             (1<<3)
-#define STATUS_CRC_WRITE_ERR            (1<<2)
-#define STATUS_TIME_OUT_RESP            (1<<1)
-#define STATUS_TIME_OUT_READ            (1<<0)
-#define STATUS_ERR_MASK                 0x2f
-#define CLK_RATE_PRESCALER(x)           ((x) & 0x7)
-#define CLK_RATE_CLK_RATE(x)            (((x) & 0x7) << 3)
-#define CMD_DAT_CONT_CMD_RESP_LONG_OFF  (1<<12)
-#define CMD_DAT_CONT_STOP_READWAIT      (1<<11)
-#define CMD_DAT_CONT_START_READWAIT     (1<<10)
-#define CMD_DAT_CONT_BUS_WIDTH_1        (0<<8)
-#define CMD_DAT_CONT_BUS_WIDTH_4        (2<<8)
-#define CMD_DAT_CONT_INIT               (1<<7)
-#define CMD_DAT_CONT_BUSY               (1<<6)
-#define CMD_DAT_CONT_STREAM_BLOCK       (1<<5)
-#define CMD_DAT_CONT_WRITE              (1<<4)
-#define CMD_DAT_CONT_DATA_ENABLE        (1<<3)
-#define CMD_DAT_CONT_RESPONSE_FORMAT_R1 (1)
-#define CMD_DAT_CONT_RESPONSE_FORMAT_R2 (2)
-#define CMD_DAT_CONT_RESPONSE_FORMAT_R3 (3)
-#define CMD_DAT_CONT_RESPONSE_FORMAT_R4 (4)
-#define CMD_DAT_CONT_RESPONSE_FORMAT_R5 (5)
-#define CMD_DAT_CONT_RESPONSE_FORMAT_R6 (6)
-#define INT_MASK_AUTO_CARD_DETECT       (1<<6)
-#define INT_MASK_DAT0_EN                (1<<5)
-#define INT_MASK_SDIO                   (1<<4)
-#define INT_MASK_BUF_READY              (1<<3)
-#define INT_MASK_END_CMD_RES            (1<<2)
-#define INT_MASK_WRITE_OP_DONE          (1<<1)
-#define INT_MASK_DATA_TRAN              (1<<0)
-#define INT_ALL                         (0x7f)
diff --git a/drivers/mmc/host/mmci.c b/drivers/mmc/host/mmci.c
index b6f38421d541..f0fcce40cd8d 100644
--- a/drivers/mmc/host/mmci.c
+++ b/drivers/mmc/host/mmci.c
@@ -15,6 +15,7 @@
 #include <linux/device.h>
 #include <linux/interrupt.h>
 #include <linux/kernel.h>
+#include <linux/slab.h>
 #include <linux/delay.h>
 #include <linux/err.h>
 #include <linux/highmem.h>
@@ -25,6 +26,7 @@
 #include <linux/clk.h>
 #include <linux/scatterlist.h>
 #include <linux/gpio.h>
+#include <linux/of_gpio.h>
 #include <linux/regulator/consumer.h>
 #include <linux/dmaengine.h>
 #include <linux/dma-mapping.h>
@@ -1207,21 +1209,76 @@ static const struct mmc_host_ops mmci_ops = {
 	.get_cd		= mmci_get_cd,
 };
 
+#ifdef CONFIG_OF
+static void mmci_dt_populate_generic_pdata(struct device_node *np,
+					struct mmci_platform_data *pdata)
+{
+	int bus_width = 0;
+
+	pdata->gpio_wp = of_get_named_gpio(np, "wp-gpios", 0);
+	if (!pdata->gpio_wp)
+		pdata->gpio_wp = -1;
+
+	pdata->gpio_cd = of_get_named_gpio(np, "cd-gpios", 0);
+	if (!pdata->gpio_cd)
+		pdata->gpio_cd = -1;
+
+	if (of_get_property(np, "cd-inverted", NULL))
+		pdata->cd_invert = true;
+	else
+		pdata->cd_invert = false;
+
+	of_property_read_u32(np, "max-frequency", &pdata->f_max);
+	if (!pdata->f_max)
+		pr_warn("%s has no 'max-frequency' property\n", np->full_name);
+
+	if (of_get_property(np, "mmc-cap-mmc-highspeed", NULL))
+		pdata->capabilities |= MMC_CAP_MMC_HIGHSPEED;
+	if (of_get_property(np, "mmc-cap-sd-highspeed", NULL))
+		pdata->capabilities |= MMC_CAP_SD_HIGHSPEED;
+
+	of_property_read_u32(np, "bus-width", &bus_width);
+	switch (bus_width) {
+	case 0 :
+		/* No bus-width supplied. */
+		break;
+	case 4 :
+		pdata->capabilities |= MMC_CAP_4_BIT_DATA;
+		break;
+	case 8 :
+		pdata->capabilities |= MMC_CAP_8_BIT_DATA;
+		break;
+	default :
+		pr_warn("%s: Unsupported bus width\n", np->full_name);
+	}
+}
+#else
+static void mmci_dt_populate_generic_pdata(struct device_node *np,
+					struct mmci_platform_data *pdata)
+{
+	return;
+}
+#endif
+
 static int __devinit mmci_probe(struct amba_device *dev,
 	const struct amba_id *id)
 {
 	struct mmci_platform_data *plat = dev->dev.platform_data;
+	struct device_node *np = dev->dev.of_node;
 	struct variant_data *variant = id->data;
 	struct mmci_host *host;
 	struct mmc_host *mmc;
 	int ret;
 
-	/* must have platform data */
-	if (!plat) {
-		ret = -EINVAL;
-		goto out;
+	/* Must have platform data or Device Tree. */
+	if (!plat && !np) {
+		dev_err(&dev->dev, "No plat data or DT found\n");
+		return -EINVAL;
 	}
 
+	if (np)
+		mmci_dt_populate_generic_pdata(np, plat);
+
 	ret = amba_request_regions(dev, DRIVER_NAME);
 	if (ret)
 		goto out;
diff --git a/drivers/mmc/host/omap.c b/drivers/mmc/host/omap.c
index 887c0e598cf3..552196c764d4 100644
--- a/drivers/mmc/host/omap.c
+++ b/drivers/mmc/host/omap.c
@@ -169,11 +169,11 @@ struct mmc_omap_host {
 	struct timer_list       clk_timer;
 	spinlock_t		clk_lock;     /* for changing enabled state */
 	unsigned int            fclk_enabled:1;
+	struct workqueue_struct *mmc_omap_wq;
 
 	struct omap_mmc_platform_data *pdata;
 };
 
-static struct workqueue_struct *mmc_omap_wq;
 
 static void mmc_omap_fclk_offdelay(struct mmc_omap_slot *slot)
 {
@@ -291,7 +291,7 @@ static void mmc_omap_release_slot(struct mmc_omap_slot *slot, int clk_enabled)
 		host->next_slot = new_slot;
 		host->mmc = new_slot->mmc;
 		spin_unlock_irqrestore(&host->slot_lock, flags);
-		queue_work(mmc_omap_wq, &host->slot_release_work);
+		queue_work(host->mmc_omap_wq, &host->slot_release_work);
 		return;
 	}
 
@@ -459,7 +459,7 @@ mmc_omap_xfer_done(struct mmc_omap_host *host, struct mmc_data *data)
 	}
 
 	host->stop_data = data;
-	queue_work(mmc_omap_wq, &host->send_stop_work);
+	queue_work(host->mmc_omap_wq, &host->send_stop_work);
 }
 
 static void
@@ -639,7 +639,7 @@ mmc_omap_cmd_timer(unsigned long data)
 		OMAP_MMC_WRITE(host, IE, 0);
 		disable_irq(host->irq);
 		host->abort = 1;
-		queue_work(mmc_omap_wq, &host->cmd_abort_work);
+		queue_work(host->mmc_omap_wq, &host->cmd_abort_work);
 	}
 	spin_unlock_irqrestore(&host->slot_lock, flags);
 }
@@ -828,7 +828,7 @@ static irqreturn_t mmc_omap_irq(int irq, void *dev_id)
 		host->abort = 1;
 		OMAP_MMC_WRITE(host, IE, 0);
 		disable_irq_nosync(host->irq);
-		queue_work(mmc_omap_wq, &host->cmd_abort_work);
+		queue_work(host->mmc_omap_wq, &host->cmd_abort_work);
 		return IRQ_HANDLED;
 	}
 
@@ -1389,13 +1389,13 @@ static void mmc_omap_remove_slot(struct mmc_omap_slot *slot)
 
 	tasklet_kill(&slot->cover_tasklet);
 	del_timer_sync(&slot->cover_timer);
-	flush_workqueue(mmc_omap_wq);
+	flush_workqueue(slot->host->mmc_omap_wq);
 
 	mmc_remove_host(mmc);
 	mmc_free_host(mmc);
 }
 
-static int __init mmc_omap_probe(struct platform_device *pdev)
+static int __devinit mmc_omap_probe(struct platform_device *pdev)
 {
 	struct omap_mmc_platform_data *pdata = pdev->dev.platform_data;
 	struct mmc_omap_host *host = NULL;
@@ -1497,6 +1497,10 @@ static int __init mmc_omap_probe(struct platform_device *pdev)
 
 	host->reg_shift = (cpu_is_omap7xx() ? 1 : 2);
 
+	host->mmc_omap_wq = alloc_workqueue("mmc_omap", 0, 0);
+	if (!host->mmc_omap_wq)
+		goto err_plat_cleanup;
+
 	return 0;
 
 err_plat_cleanup:
@@ -1518,7 +1522,7 @@ err_free_mem_region:
 	return ret;
 }
 
-static int mmc_omap_remove(struct platform_device *pdev)
+static int __devexit mmc_omap_remove(struct platform_device *pdev)
 {
 	struct mmc_omap_host *host = platform_get_drvdata(pdev);
 	int i;
@@ -1542,6 +1546,7 @@ static int mmc_omap_remove(struct platform_device *pdev)
 	iounmap(host->virt_base);
 	release_mem_region(pdev->resource[0].start,
 			   pdev->resource[0].end - pdev->resource[0].start + 1);
+	destroy_workqueue(host->mmc_omap_wq);
 
 	kfree(host);
 
@@ -1599,7 +1604,8 @@ static int mmc_omap_resume(struct platform_device *pdev)
 #endif
 
 static struct platform_driver mmc_omap_driver = {
-	.remove		= mmc_omap_remove,
+	.probe		= mmc_omap_probe,
+	.remove		= __devexit_p(mmc_omap_remove),
 	.suspend	= mmc_omap_suspend,
 	.resume		= mmc_omap_resume,
 	.driver		= {
@@ -1608,29 +1614,7 @@ static struct platform_driver mmc_omap_driver = {
 	},
 };
 
-static int __init mmc_omap_init(void)
-{
-	int ret;
-
-	mmc_omap_wq = alloc_workqueue("mmc_omap", 0, 0);
-	if (!mmc_omap_wq)
-		return -ENOMEM;
-
-	ret = platform_driver_probe(&mmc_omap_driver, mmc_omap_probe);
-	if (ret)
-		destroy_workqueue(mmc_omap_wq);
-	return ret;
-}
-
-static void __exit mmc_omap_exit(void)
-{
-	platform_driver_unregister(&mmc_omap_driver);
-	destroy_workqueue(mmc_omap_wq);
-}
-
-module_init(mmc_omap_init);
-module_exit(mmc_omap_exit);
-
+module_platform_driver(mmc_omap_driver);
 MODULE_DESCRIPTION("OMAP Multimedia Card driver");
 MODULE_LICENSE("GPL");
 MODULE_ALIAS("platform:" DRIVER_NAME);
diff --git a/drivers/mmc/host/omap_hsmmc.c b/drivers/mmc/host/omap_hsmmc.c
index 56d4499d4388..a9fc714fb38d 100644
--- a/drivers/mmc/host/omap_hsmmc.c
+++ b/drivers/mmc/host/omap_hsmmc.c
@@ -85,12 +85,14 @@
 #define BRR_ENABLE		(1 << 5)
 #define DTO_ENABLE		(1 << 20)
 #define INIT_STREAM		(1 << 1)
+#define ACEN_ACMD12		(1 << 2)
 #define DP_SELECT		(1 << 21)
 #define DDIR			(1 << 4)
 #define DMA_EN			0x1
 #define MSBS			(1 << 5)
 #define BCE			(1 << 1)
 #define FOUR_BIT		(1 << 1)
+#define DDR			(1 << 19)
 #define DW8			(1 << 5)
 #define CC			0x1
 #define TC			0x02
@@ -115,6 +117,7 @@
 #define OMAP_MMC_MAX_CLOCK	52000000
 #define DRIVER_NAME		"omap_hsmmc"
 
+#define AUTO_CMD12		(1 << 0)	/* Auto CMD12 support */
 /*
  * One controller can have multiple slots, like on some omap boards using
  * omap.c controller driver. Luckily this is not currently done on any known
@@ -167,7 +170,6 @@ struct omap_hsmmc_host {
 	int			use_dma, dma_ch;
 	int			dma_line_tx, dma_line_rx;
 	int			slot_id;
-	int			got_dbclk;
 	int			response_busy;
 	int			context_loss;
 	int			vdd;
@@ -175,6 +177,7 @@ struct omap_hsmmc_host {
 	int			reqs_blocked;
 	int			use_reg;
 	int			req_in_progress;
+	unsigned int		flags;
 	struct omap_hsmmc_next	next_data;
 
 	struct	omap_mmc_platform_data	*pdata;
@@ -520,6 +523,10 @@ static void omap_hsmmc_set_bus_width(struct omap_hsmmc_host *host)
 	u32 con;
 
 	con = OMAP_HSMMC_READ(host->base, CON);
+	if (ios->timing == MMC_TIMING_UHS_DDR50)
+		con |= DDR;	/* configure in DDR mode */
+	else
+		con &= ~DDR;
 	switch (ios->bus_width) {
 	case MMC_BUS_WIDTH_8:
 		OMAP_HSMMC_WRITE(host->base, CON, con | DW8);
@@ -766,6 +773,8 @@ omap_hsmmc_start_command(struct omap_hsmmc_host *host, struct mmc_command *cmd,
 		cmdtype = 0x3;
 
 	cmdreg = (cmd->opcode << 24) | (resptype << 16) | (cmdtype << 22);
+	if ((host->flags & AUTO_CMD12) && mmc_op_multi(cmd->opcode))
+		cmdreg |= ACEN_ACMD12;
 
 	if (data) {
 		cmdreg |= DP_SELECT | MSBS | BCE;
@@ -796,11 +805,12 @@ omap_hsmmc_get_dma_dir(struct omap_hsmmc_host *host, struct mmc_data *data)
 static void omap_hsmmc_request_done(struct omap_hsmmc_host *host, struct mmc_request *mrq)
 {
 	int dma_ch;
+	unsigned long flags;
 
-	spin_lock(&host->irq_lock);
+	spin_lock_irqsave(&host->irq_lock, flags);
 	host->req_in_progress = 0;
 	dma_ch = host->dma_ch;
-	spin_unlock(&host->irq_lock);
+	spin_unlock_irqrestore(&host->irq_lock, flags);
 
 	omap_hsmmc_disable_irq(host);
 	/* Do not complete the request if DMA is still in progress */
@@ -837,11 +847,14 @@ omap_hsmmc_xfer_done(struct omap_hsmmc_host *host, struct mmc_data *data)
 	else
 		data->bytes_xfered = 0;
 
-	if (!data->stop) {
+	if (data->stop && ((!(host->flags & AUTO_CMD12)) || data->error)) {
+		omap_hsmmc_start_command(host, data->stop, NULL);
+	} else {
+		if (data->stop)
+			data->stop->resp[0] = OMAP_HSMMC_READ(host->base,
+							RSP76);
 		omap_hsmmc_request_done(host, data->mrq);
-		return;
 	}
-	omap_hsmmc_start_command(host, data->stop, NULL);
 }
 
 /*
@@ -874,13 +887,14 @@ omap_hsmmc_cmd_done(struct omap_hsmmc_host *host, struct mmc_command *cmd)
 static void omap_hsmmc_dma_cleanup(struct omap_hsmmc_host *host, int errno)
 {
 	int dma_ch;
+	unsigned long flags;
 
 	host->data->error = errno;
 
-	spin_lock(&host->irq_lock);
+	spin_lock_irqsave(&host->irq_lock, flags);
 	dma_ch = host->dma_ch;
 	host->dma_ch = -1;
-	spin_unlock(&host->irq_lock);
+	spin_unlock_irqrestore(&host->irq_lock, flags);
 
 	if (host->use_dma && dma_ch != -1) {
 		dma_unmap_sg(mmc_dev(host->mmc), host->data->sg,
@@ -1082,7 +1096,7 @@ static int omap_hsmmc_switch_opcond(struct omap_hsmmc_host *host, int vdd)
 
 	/* Disable the clocks */
 	pm_runtime_put_sync(host->dev);
-	if (host->got_dbclk)
+	if (host->dbclk)
 		clk_disable(host->dbclk);
 
 	/* Turn the power off */
@@ -1093,7 +1107,7 @@ static int omap_hsmmc_switch_opcond(struct omap_hsmmc_host *host, int vdd)
 		ret = mmc_slot(host).set_power(host->dev, host->slot_id, 1,
 					       vdd);
 	pm_runtime_get_sync(host->dev);
-	if (host->got_dbclk)
+	if (host->dbclk)
 		clk_enable(host->dbclk);
 
 	if (ret != 0)
@@ -1234,6 +1248,7 @@ static void omap_hsmmc_dma_cb(int lch, u16 ch_status, void *cb_data)
 	struct omap_hsmmc_host *host = cb_data;
 	struct mmc_data *data;
 	int dma_ch, req_in_progress;
+	unsigned long flags;
 
 	if (!(ch_status & OMAP_DMA_BLOCK_IRQ)) {
 		dev_warn(mmc_dev(host->mmc), "unexpected dma status %x\n",
@@ -1241,9 +1256,9 @@ static void omap_hsmmc_dma_cb(int lch, u16 ch_status, void *cb_data)
 		return;
 	}
 
-	spin_lock(&host->irq_lock);
+	spin_lock_irqsave(&host->irq_lock, flags);
 	if (host->dma_ch < 0) {
-		spin_unlock(&host->irq_lock);
+		spin_unlock_irqrestore(&host->irq_lock, flags);
 		return;
 	}
 
@@ -1253,7 +1268,7 @@ static void omap_hsmmc_dma_cb(int lch, u16 ch_status, void *cb_data)
 		/* Fire up the next transfer. */
 		omap_hsmmc_config_dma_params(host, data,
 					   data->sg + host->dma_sg_idx);
-		spin_unlock(&host->irq_lock);
+		spin_unlock_irqrestore(&host->irq_lock, flags);
 		return;
 	}
 
@@ -1264,7 +1279,7 @@ static void omap_hsmmc_dma_cb(int lch, u16 ch_status, void *cb_data)
 	req_in_progress = host->req_in_progress;
 	dma_ch = host->dma_ch;
 	host->dma_ch = -1;
-	spin_unlock(&host->irq_lock);
+	spin_unlock_irqrestore(&host->irq_lock, flags);
 
 	omap_free_dma(dma_ch);
 
@@ -1844,6 +1859,7 @@ static int __devinit omap_hsmmc_probe(struct platform_device *pdev)
 	host->mapbase	= res->start + pdata->reg_offset;
 	host->base	= ioremap(host->mapbase, SZ_4K);
 	host->power_mode = MMC_POWER_OFF;
+	host->flags	= AUTO_CMD12;
 	host->next_data.cookie = 1;
 
 	platform_set_drvdata(pdev, host);
@@ -1885,21 +1901,17 @@ static int __devinit omap_hsmmc_probe(struct platform_device *pdev)
 
 	omap_hsmmc_context_save(host);
 
-	if (cpu_is_omap2430()) {
-		host->dbclk = clk_get(&pdev->dev, "mmchsdb_fck");
-		/*
-		 * MMC can still work without debounce clock.
-		 */
-		if (IS_ERR(host->dbclk))
-			dev_warn(mmc_dev(host->mmc),
-				"Failed to get debounce clock\n");
-		else
-			host->got_dbclk = 1;
-
-		if (host->got_dbclk)
-			if (clk_enable(host->dbclk) != 0)
-				dev_dbg(mmc_dev(host->mmc), "Enabling debounce"
-							" clk failed\n");
+	host->dbclk = clk_get(&pdev->dev, "mmchsdb_fck");
+	/*
+	 * MMC can still work without debounce clock.
+	 */
+	if (IS_ERR(host->dbclk)) {
+		dev_warn(mmc_dev(host->mmc), "Failed to get debounce clk\n");
+		host->dbclk = NULL;
+	} else if (clk_enable(host->dbclk) != 0) {
+		dev_warn(mmc_dev(host->mmc), "Failed to enable debounce clk\n");
+		clk_put(host->dbclk);
+		host->dbclk = NULL;
 	}
 
 	/* Since we do only SG emulation, we can have as many segs
@@ -1969,7 +1981,7 @@ static int __devinit omap_hsmmc_probe(struct platform_device *pdev)
 		ret = request_threaded_irq(mmc_slot(host).card_detect_irq,
 					   NULL,
 					   omap_hsmmc_detect,
-					   IRQF_TRIGGER_RISING | IRQF_TRIGGER_FALLING,
+					   IRQF_TRIGGER_RISING | IRQF_TRIGGER_FALLING | IRQF_ONESHOT,
 					   mmc_hostname(mmc), host);
 		if (ret) {
 			dev_dbg(mmc_dev(host->mmc),
@@ -2019,7 +2031,7 @@ err_irq:
 	pm_runtime_put_sync(host->dev);
 	pm_runtime_disable(host->dev);
 	clk_put(host->fclk);
-	if (host->got_dbclk) {
+	if (host->dbclk) {
 		clk_disable(host->dbclk);
 		clk_put(host->dbclk);
 	}
@@ -2030,7 +2042,9 @@ err1:
 err_alloc:
 	omap_hsmmc_gpio_free(pdata);
 err:
-	release_mem_region(res->start, resource_size(res));
+	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	if (res)
+		release_mem_region(res->start, resource_size(res));
 	return ret;
 }
 
@@ -2052,7 +2066,7 @@ static int __devexit omap_hsmmc_remove(struct platform_device *pdev)
 	pm_runtime_put_sync(host->dev);
 	pm_runtime_disable(host->dev);
 	clk_put(host->fclk);
-	if (host->got_dbclk) {
+	if (host->dbclk) {
 		clk_disable(host->dbclk);
 		clk_put(host->dbclk);
 	}
@@ -2110,7 +2124,7 @@ static int omap_hsmmc_suspend(struct device *dev)
 				OMAP_HSMMC_READ(host->base, HCTL) & ~SDBP);
 	}
 
-	if (host->got_dbclk)
+	if (host->dbclk)
 		clk_disable(host->dbclk);
 err:
 	pm_runtime_put_sync(host->dev);
@@ -2131,7 +2145,7 @@ static int omap_hsmmc_resume(struct device *dev)
 
 	pm_runtime_get_sync(host->dev);
 
-	if (host->got_dbclk)
+	if (host->dbclk)
 		clk_enable(host->dbclk);
 
 	if (!(host->mmc->pm_flags & MMC_PM_KEEP_POWER))
diff --git a/drivers/mmc/host/sdhci-spear.c b/drivers/mmc/host/sdhci-spear.c
index 6dfa82e03c7e..1fe32dfa7cd4 100644
--- a/drivers/mmc/host/sdhci-spear.c
+++ b/drivers/mmc/host/sdhci-spear.c
@@ -75,8 +75,6 @@ static int __devinit sdhci_probe(struct platform_device *pdev)
 	struct spear_sdhci *sdhci;
 	int ret;
 
-	BUG_ON(pdev == NULL);
-
 	iomem = platform_get_resource(pdev, IORESOURCE_MEM, 0);
 	if (!iomem) {
 		ret = -ENOMEM;
@@ -84,18 +82,18 @@ static int __devinit sdhci_probe(struct platform_device *pdev)
 		goto err;
 	}
 
-	if (!request_mem_region(iomem->start, resource_size(iomem),
-				"spear-sdhci")) {
+	if (!devm_request_mem_region(&pdev->dev, iomem->start,
+				resource_size(iomem), "spear-sdhci")) {
 		ret = -EBUSY;
 		dev_dbg(&pdev->dev, "cannot request region\n");
 		goto err;
 	}
 
-	sdhci = kzalloc(sizeof(*sdhci), GFP_KERNEL);
+	sdhci = devm_kzalloc(&pdev->dev, sizeof(*sdhci), GFP_KERNEL);
 	if (!sdhci) {
 		ret = -ENOMEM;
 		dev_dbg(&pdev->dev, "cannot allocate memory for sdhci\n");
-		goto err_kzalloc;
+		goto err;
 	}
 
 	/* clk enable */
@@ -103,13 +101,13 @@ static int __devinit sdhci_probe(struct platform_device *pdev)
 	if (IS_ERR(sdhci->clk)) {
 		ret = PTR_ERR(sdhci->clk);
 		dev_dbg(&pdev->dev, "Error getting clock\n");
-		goto err_clk_get;
+		goto err;
 	}
 
 	ret = clk_enable(sdhci->clk);
 	if (ret) {
 		dev_dbg(&pdev->dev, "Error enabling clock\n");
-		goto err_clk_enb;
+		goto put_clk;
 	}
 
 	/* overwrite platform_data */
@@ -124,7 +122,7 @@ static int __devinit sdhci_probe(struct platform_device *pdev)
 	if (IS_ERR(host)) {
 		ret = PTR_ERR(host);
 		dev_dbg(&pdev->dev, "error allocating host\n");
-		goto err_alloc_host;
+		goto disable_clk;
 	}
 
 	host->hw_name = "sdhci";
@@ -132,17 +130,18 @@ static int __devinit sdhci_probe(struct platform_device *pdev)
 	host->irq = platform_get_irq(pdev, 0);
 	host->quirks = SDHCI_QUIRK_BROKEN_ADMA;
 
-	host->ioaddr = ioremap(iomem->start, resource_size(iomem));
+	host->ioaddr = devm_ioremap(&pdev->dev, iomem->start,
+			resource_size(iomem));
 	if (!host->ioaddr) {
 		ret = -ENOMEM;
 		dev_dbg(&pdev->dev, "failed to remap registers\n");
-		goto err_ioremap;
+		goto free_host;
 	}
 
 	ret = sdhci_add_host(host);
 	if (ret) {
 		dev_dbg(&pdev->dev, "error adding host\n");
-		goto err_add_host;
+		goto free_host;
 	}
 
 	platform_set_drvdata(pdev, host);
@@ -161,11 +160,12 @@ static int __devinit sdhci_probe(struct platform_device *pdev)
 	if (sdhci->data->card_power_gpio >= 0) {
 		int val = 0;
 
-		ret = gpio_request(sdhci->data->card_power_gpio, "sdhci");
+		ret = devm_gpio_request(&pdev->dev,
+				sdhci->data->card_power_gpio, "sdhci");
 		if (ret < 0) {
 			dev_dbg(&pdev->dev, "gpio request fail: %d\n",
 					sdhci->data->card_power_gpio);
-			goto err_pgpio_request;
+			goto set_drvdata;
 		}
 
 		if (sdhci->data->power_always_enb)
@@ -177,60 +177,48 @@ static int __devinit sdhci_probe(struct platform_device *pdev)
 		if (ret) {
 			dev_dbg(&pdev->dev, "gpio set direction fail: %d\n",
 					sdhci->data->card_power_gpio);
-			goto err_pgpio_direction;
+			goto set_drvdata;
 		}
 	}
 
 	if (sdhci->data->card_int_gpio >= 0) {
-		ret = gpio_request(sdhci->data->card_int_gpio, "sdhci");
+		ret = devm_gpio_request(&pdev->dev, sdhci->data->card_int_gpio,
+				"sdhci");
 		if (ret < 0) {
 			dev_dbg(&pdev->dev, "gpio request fail: %d\n",
 					sdhci->data->card_int_gpio);
-			goto err_igpio_request;
+			goto set_drvdata;
 		}
 
 		ret = gpio_direction_input(sdhci->data->card_int_gpio);
 		if (ret) {
 			dev_dbg(&pdev->dev, "gpio set direction fail: %d\n",
 					sdhci->data->card_int_gpio);
-			goto err_igpio_direction;
+			goto set_drvdata;
 		}
-		ret = request_irq(gpio_to_irq(sdhci->data->card_int_gpio),
+		ret = devm_request_irq(&pdev->dev,
+				gpio_to_irq(sdhci->data->card_int_gpio),
 				sdhci_gpio_irq, IRQF_TRIGGER_LOW,
 				mmc_hostname(host->mmc), pdev);
 		if (ret) {
 			dev_dbg(&pdev->dev, "gpio request irq fail: %d\n",
 					sdhci->data->card_int_gpio);
-			goto err_igpio_request_irq;
+			goto set_drvdata;
 		}
 
 	}
 
 	return 0;
 
-err_igpio_request_irq:
-err_igpio_direction:
-	if (sdhci->data->card_int_gpio >= 0)
-		gpio_free(sdhci->data->card_int_gpio);
-err_igpio_request:
-err_pgpio_direction:
-	if (sdhci->data->card_power_gpio >= 0)
-		gpio_free(sdhci->data->card_power_gpio);
-err_pgpio_request:
+set_drvdata:
 	platform_set_drvdata(pdev, NULL);
 	sdhci_remove_host(host, 1);
-err_add_host:
-	iounmap(host->ioaddr);
-err_ioremap:
+free_host:
 	sdhci_free_host(host);
-err_alloc_host:
+disable_clk:
 	clk_disable(sdhci->clk);
-err_clk_enb:
+put_clk:
 	clk_put(sdhci->clk);
-err_clk_get:
-	kfree(sdhci);
-err_kzalloc:
-	release_mem_region(iomem->start, resource_size(iomem));
 err:
 	dev_err(&pdev->dev, "spear-sdhci probe failed: %d\n", ret);
 	return ret;
@@ -239,35 +227,19 @@ err:
 static int __devexit sdhci_remove(struct platform_device *pdev)
 {
 	struct sdhci_host *host = platform_get_drvdata(pdev);
-	struct resource *iomem = platform_get_resource(pdev, IORESOURCE_MEM, 0);
 	struct spear_sdhci *sdhci = dev_get_platdata(&pdev->dev);
-	int dead;
+	int dead = 0;
 	u32 scratch;
 
-	if (sdhci->data) {
-		if (sdhci->data->card_int_gpio >= 0) {
-			free_irq(gpio_to_irq(sdhci->data->card_int_gpio), pdev);
-			gpio_free(sdhci->data->card_int_gpio);
-		}
-
-		if (sdhci->data->card_power_gpio >= 0)
-			gpio_free(sdhci->data->card_power_gpio);
-	}
-
 	platform_set_drvdata(pdev, NULL);
-	dead = 0;
 	scratch = readl(host->ioaddr + SDHCI_INT_STATUS);
 	if (scratch == (u32)-1)
 		dead = 1;
 
 	sdhci_remove_host(host, dead);
-	iounmap(host->ioaddr);
 	sdhci_free_host(host);
 	clk_disable(sdhci->clk);
 	clk_put(sdhci->clk);
-	kfree(sdhci);
-	if (iomem)
-		release_mem_region(iomem->start, resource_size(iomem));
 
 	return 0;
 }
diff --git a/drivers/mmc/host/sdhci-tegra.c b/drivers/mmc/host/sdhci-tegra.c
index ff5a16991939..b38d8a78f6a0 100644
--- a/drivers/mmc/host/sdhci-tegra.c
+++ b/drivers/mmc/host/sdhci-tegra.c
@@ -32,8 +32,13 @@
 
 #include "sdhci-pltfm.h"
 
+/* Tegra SDHOST controller vendor register definitions */
+#define SDHCI_TEGRA_VENDOR_MISC_CTRL		0x120
+#define SDHCI_MISC_CTRL_ENABLE_SDHCI_SPEC_300	0x20
+
 #define NVQUIRK_FORCE_SDHCI_SPEC_200	BIT(0)
 #define NVQUIRK_ENABLE_BLOCK_GAP_DET	BIT(1)
+#define NVQUIRK_ENABLE_SDHCI_SPEC_300	BIT(2)
 
 struct sdhci_tegra_soc_data {
 	struct sdhci_pltfm_data *pdata;
@@ -120,6 +125,25 @@ static irqreturn_t carddetect_irq(int irq, void *data)
 	return IRQ_HANDLED;
 };
 
+static void tegra_sdhci_reset_exit(struct sdhci_host *host, u8 mask)
+{
+	struct sdhci_pltfm_host *pltfm_host = sdhci_priv(host);
+	struct sdhci_tegra *tegra_host = pltfm_host->priv;
+	const struct sdhci_tegra_soc_data *soc_data = tegra_host->soc_data;
+
+	if (!(mask & SDHCI_RESET_ALL))
+		return;
+
+	/* Erratum: Enable SDHCI spec v3.00 support */
+	if (soc_data->nvquirks & NVQUIRK_ENABLE_SDHCI_SPEC_300) {
+		u32 misc_ctrl;
+
+		misc_ctrl = sdhci_readb(host, SDHCI_TEGRA_VENDOR_MISC_CTRL);
+		misc_ctrl |= SDHCI_MISC_CTRL_ENABLE_SDHCI_SPEC_300;
+		sdhci_writeb(host, misc_ctrl, SDHCI_TEGRA_VENDOR_MISC_CTRL);
+	}
+}
+
 static int tegra_sdhci_8bit(struct sdhci_host *host, int bus_width)
 {
 	struct sdhci_pltfm_host *pltfm_host = sdhci_priv(host);
@@ -148,6 +172,7 @@ static struct sdhci_ops tegra_sdhci_ops = {
 	.read_w     = tegra_sdhci_readw,
 	.write_l    = tegra_sdhci_writel,
 	.platform_8bit_width = tegra_sdhci_8bit,
+	.platform_reset_exit = tegra_sdhci_reset_exit,
 };
 
 #ifdef CONFIG_ARCH_TEGRA_2x_SOC
@@ -178,6 +203,7 @@ static struct sdhci_pltfm_data sdhci_tegra30_pdata = {
 
 static struct sdhci_tegra_soc_data soc_data_tegra30 = {
 	.pdata = &sdhci_tegra30_pdata,
+	.nvquirks = NVQUIRK_ENABLE_SDHCI_SPEC_300,
 };
 #endif
 
diff --git a/drivers/mmc/host/sdhci.c b/drivers/mmc/host/sdhci.c
index ccefdebeff14..e626732aff77 100644
--- a/drivers/mmc/host/sdhci.c
+++ b/drivers/mmc/host/sdhci.c
@@ -680,8 +680,8 @@ static u8 sdhci_calc_timeout(struct sdhci_host *host, struct mmc_command *cmd)
 	}
 
 	if (count >= 0xF) {
-		pr_warning("%s: Too large timeout requested for CMD%d!\n",
-		       mmc_hostname(host->mmc), cmd->opcode);
+		pr_warning("%s: Too large timeout 0x%x requested for CMD%d!\n",
+			   mmc_hostname(host->mmc), count, cmd->opcode);
 		count = 0xE;
 	}