Merge branch 'i2c/for-5.2' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux

Pull i2c updates from Wolfram Sang:

 - API for late atomic transfers (e.g. to shut down via PMIC). We have a
   seperate callback now which is called under clearly defined
   conditions. In-kernel users are converted, too.

 - new driver for the AMD PCIe MP2 I2C controller

 - large refactoring for at91 and bcm-iproc (both gain slave support due
   to this)

 - and a good share of various driver improvements anf fixes

* 'i2c/for-5.2' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux: (57 commits)
  dt-bindings: i2c: riic: document r7s9210 support
  i2c: imx-lpi2c: Use __maybe_unused instead of #if CONFIG_PM_SLEEP
  i2c-piix4: Add Hygon Dhyana SMBus support
  i2c: core: apply 'is_suspended' check for SMBus, too
  i2c: core: ratelimit 'transfer when suspended' errors
  i2c: iproc: Change driver to use 'BIT' macro
  i2c: riic: Add Runtime PM support
  i2c: mux: demux-pinctrl: use struct_size() in devm_kzalloc()
  i2c: mux: pca954x: allow management of device idle state via sysfs
  i2c: mux: pca9541: remove support for unused platform data
  i2c: mux: pca954x: remove support for unused platform data
  dt-bindings: i2c: i2c-mtk: add support for MT8516
  i2c: axxia: use auto cmd for last message
  i2c: gpio: flag atomic capability if possible
  i2c: algo: bit: add flag to whitelist atomic transfers
  i2c: stu300: use xfer_atomic callback to bail out early
  i2c: ocores: enable atomic xfers
  i2c: ocores: refactor setup for polling
  i2c: tegra-bpmp: convert to use new atomic callbacks
  i2c: omap: Add the master_xfer_atomic hook
  ...

1 files changed, 801 insertions, 0 deletions

diff --git a/drivers/i2c/busses/i2c-at91-master.c b/drivers/i2c/busses/i2c-at91-master.c
new file mode 100644
index 000000000000..e87232f2e708
--- /dev/null
+++ b/drivers/i2c/busses/i2c-at91-master.c
@@ -0,0 +1,801 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *  i2c Support for Atmel's AT91 Two-Wire Interface (TWI)
+ *
+ *  Copyright (C) 2011 Weinmann Medical GmbH
+ *  Author: Nikolaus Voss <n.voss@weinmann.de>
+ *
+ *  Evolved from original work by:
+ *  Copyright (C) 2004 Rick Bronson
+ *  Converted to 2.6 by Andrew Victor <andrew@sanpeople.com>
+ *
+ *  Borrowed heavily from original work by:
+ *  Copyright (C) 2000 Philip Edelbrock <phil@stimpy.netroedge.com>
+ */
+
+#include <linux/clk.h>
+#include <linux/completion.h>
+#include <linux/dma-mapping.h>
+#include <linux/dmaengine.h>
+#include <linux/err.h>
+#include <linux/i2c.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/of.h>
+#include <linux/of_device.h>
+#include <linux/platform_device.h>
+#include <linux/platform_data/dma-atmel.h>
+#include <linux/pm_runtime.h>
+
+#include "i2c-at91.h"
+
+void at91_init_twi_bus_master(struct at91_twi_dev *dev)
+{
+	/* FIFO should be enabled immediately after the software reset */
+	if (dev->fifo_size)
+		at91_twi_write(dev, AT91_TWI_CR, AT91_TWI_FIFOEN);
+	at91_twi_write(dev, AT91_TWI_CR, AT91_TWI_MSEN);
+	at91_twi_write(dev, AT91_TWI_CR, AT91_TWI_SVDIS);
+	at91_twi_write(dev, AT91_TWI_CWGR, dev->twi_cwgr_reg);
+}
+
+/*
+ * Calculate symmetric clock as stated in datasheet:
+ * twi_clk = F_MAIN / (2 * (cdiv * (1 << ckdiv) + offset))
+ */
+static void at91_calc_twi_clock(struct at91_twi_dev *dev)
+{
+	int ckdiv, cdiv, div, hold = 0;
+	struct at91_twi_pdata *pdata = dev->pdata;
+	int offset = pdata->clk_offset;
+	int max_ckdiv = pdata->clk_max_div;
+	struct i2c_timings timings, *t = &timings;
+
+	i2c_parse_fw_timings(dev->dev, t, true);
+
+	div = max(0, (int)DIV_ROUND_UP(clk_get_rate(dev->clk),
+				       2 * t->bus_freq_hz) - offset);
+	ckdiv = fls(div >> 8);
+	cdiv = div >> ckdiv;
+
+	if (ckdiv > max_ckdiv) {
+		dev_warn(dev->dev, "%d exceeds ckdiv max value which is %d.\n",
+			 ckdiv, max_ckdiv);
+		ckdiv = max_ckdiv;
+		cdiv = 255;
+	}
+
+	if (pdata->has_hold_field) {
+		/*
+		 * hold time = HOLD + 3 x T_peripheral_clock
+		 * Use clk rate in kHz to prevent overflows when computing
+		 * hold.
+		 */
+		hold = DIV_ROUND_UP(t->sda_hold_ns
+				    * (clk_get_rate(dev->clk) / 1000), 1000000);
+		hold -= 3;
+		if (hold < 0)
+			hold = 0;
+		if (hold > AT91_TWI_CWGR_HOLD_MAX) {
+			dev_warn(dev->dev,
+				 "HOLD field set to its maximum value (%d instead of %d)\n",
+				 AT91_TWI_CWGR_HOLD_MAX, hold);
+			hold = AT91_TWI_CWGR_HOLD_MAX;
+		}
+	}
+
+	dev->twi_cwgr_reg = (ckdiv << 16) | (cdiv << 8) | cdiv
+			    | AT91_TWI_CWGR_HOLD(hold);
+
+	dev_dbg(dev->dev, "cdiv %d ckdiv %d hold %d (%d ns)\n",
+		cdiv, ckdiv, hold, t->sda_hold_ns);
+}
+
+static void at91_twi_dma_cleanup(struct at91_twi_dev *dev)
+{
+	struct at91_twi_dma *dma = &dev->dma;
+
+	at91_twi_irq_save(dev);
+
+	if (dma->xfer_in_progress) {
+		if (dma->direction == DMA_FROM_DEVICE)
+			dmaengine_terminate_all(dma->chan_rx);
+		else
+			dmaengine_terminate_all(dma->chan_tx);
+		dma->xfer_in_progress = false;
+	}
+	if (dma->buf_mapped) {
+		dma_unmap_single(dev->dev, sg_dma_address(&dma->sg[0]),
+				 dev->buf_len, dma->direction);
+		dma->buf_mapped = false;
+	}
+
+	at91_twi_irq_restore(dev);
+}
+
+static void at91_twi_write_next_byte(struct at91_twi_dev *dev)
+{
+	if (!dev->buf_len)
+		return;
+
+	/* 8bit write works with and without FIFO */
+	writeb_relaxed(*dev->buf, dev->base + AT91_TWI_THR);
+
+	/* send stop when last byte has been written */
+	if (--dev->buf_len == 0)
+		if (!dev->use_alt_cmd)
+			at91_twi_write(dev, AT91_TWI_CR, AT91_TWI_STOP);
+
+	dev_dbg(dev->dev, "wrote 0x%x, to go %zu\n", *dev->buf, dev->buf_len);
+
+	++dev->buf;
+}
+
+static void at91_twi_write_data_dma_callback(void *data)
+{
+	struct at91_twi_dev *dev = (struct at91_twi_dev *)data;
+
+	dma_unmap_single(dev->dev, sg_dma_address(&dev->dma.sg[0]),
+			 dev->buf_len, DMA_TO_DEVICE);
+
+	/*
+	 * When this callback is called, THR/TX FIFO is likely not to be empty
+	 * yet. So we have to wait for TXCOMP or NACK bits to be set into the
+	 * Status Register to be sure that the STOP bit has been sent and the
+	 * transfer is completed. The NACK interrupt has already been enabled,
+	 * we just have to enable TXCOMP one.
+	 */
+	at91_twi_write(dev, AT91_TWI_IER, AT91_TWI_TXCOMP);
+	if (!dev->use_alt_cmd)
+		at91_twi_write(dev, AT91_TWI_CR, AT91_TWI_STOP);
+}
+
+static void at91_twi_write_data_dma(struct at91_twi_dev *dev)
+{
+	dma_addr_t dma_addr;
+	struct dma_async_tx_descriptor *txdesc;
+	struct at91_twi_dma *dma = &dev->dma;
+	struct dma_chan *chan_tx = dma->chan_tx;
+	unsigned int sg_len = 1;
+
+	if (!dev->buf_len)
+		return;
+
+	dma->direction = DMA_TO_DEVICE;
+
+	at91_twi_irq_save(dev);
+	dma_addr = dma_map_single(dev->dev, dev->buf, dev->buf_len,
+				  DMA_TO_DEVICE);
+	if (dma_mapping_error(dev->dev, dma_addr)) {
+		dev_err(dev->dev, "dma map failed\n");
+		return;
+	}
+	dma->buf_mapped = true;
+	at91_twi_irq_restore(dev);
+
+	if (dev->fifo_size) {
+		size_t part1_len, part2_len;
+		struct scatterlist *sg;
+		unsigned fifo_mr;
+
+		sg_len = 0;
+
+		part1_len = dev->buf_len & ~0x3;
+		if (part1_len) {
+			sg = &dma->sg[sg_len++];
+			sg_dma_len(sg) = part1_len;
+			sg_dma_address(sg) = dma_addr;
+		}
+
+		part2_len = dev->buf_len & 0x3;
+		if (part2_len) {
+			sg = &dma->sg[sg_len++];
+			sg_dma_len(sg) = part2_len;
+			sg_dma_address(sg) = dma_addr + part1_len;
+		}
+
+		/*
+		 * DMA controller is triggered when at least 4 data can be
+		 * written into the TX FIFO
+		 */
+		fifo_mr = at91_twi_read(dev, AT91_TWI_FMR);
+		fifo_mr &= ~AT91_TWI_FMR_TXRDYM_MASK;
+		fifo_mr |= AT91_TWI_FMR_TXRDYM(AT91_TWI_FOUR_DATA);
+		at91_twi_write(dev, AT91_TWI_FMR, fifo_mr);
+	} else {
+		sg_dma_len(&dma->sg[0]) = dev->buf_len;
+		sg_dma_address(&dma->sg[0]) = dma_addr;
+	}
+
+	txdesc = dmaengine_prep_slave_sg(chan_tx, dma->sg, sg_len,
+					 DMA_MEM_TO_DEV,
+					 DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
+	if (!txdesc) {
+		dev_err(dev->dev, "dma prep slave sg failed\n");
+		goto error;
+	}
+
+	txdesc->callback = at91_twi_write_data_dma_callback;
+	txdesc->callback_param = dev;
+
+	dma->xfer_in_progress = true;
+	dmaengine_submit(txdesc);
+	dma_async_issue_pending(chan_tx);
+
+	return;
+
+error:
+	at91_twi_dma_cleanup(dev);
+}
+
+static void at91_twi_read_next_byte(struct at91_twi_dev *dev)
+{
+	/*
+	 * If we are in this case, it means there is garbage data in RHR, so
+	 * delete them.
+	 */
+	if (!dev->buf_len) {
+		at91_twi_read(dev, AT91_TWI_RHR);
+		return;
+	}
+
+	/* 8bit read works with and without FIFO */
+	*dev->buf = readb_relaxed(dev->base + AT91_TWI_RHR);
+	--dev->buf_len;
+
+	/* return if aborting, we only needed to read RHR to clear RXRDY*/
+	if (dev->recv_len_abort)
+		return;
+
+	/* handle I2C_SMBUS_BLOCK_DATA */
+	if (unlikely(dev->msg->flags & I2C_M_RECV_LEN)) {
+		/* ensure length byte is a valid value */
+		if (*dev->buf <= I2C_SMBUS_BLOCK_MAX && *dev->buf > 0) {
+			dev->msg->flags &= ~I2C_M_RECV_LEN;
+			dev->buf_len += *dev->buf;
+			dev->msg->len = dev->buf_len + 1;
+			dev_dbg(dev->dev, "received block length %zu\n",
+					 dev->buf_len);
+		} else {
+			/* abort and send the stop by reading one more byte */
+			dev->recv_len_abort = true;
+			dev->buf_len = 1;
+		}
+	}
+
+	/* send stop if second but last byte has been read */
+	if (!dev->use_alt_cmd && dev->buf_len == 1)
+		at91_twi_write(dev, AT91_TWI_CR, AT91_TWI_STOP);
+
+	dev_dbg(dev->dev, "read 0x%x, to go %zu\n", *dev->buf, dev->buf_len);
+
+	++dev->buf;
+}
+
+static void at91_twi_read_data_dma_callback(void *data)
+{
+	struct at91_twi_dev *dev = (struct at91_twi_dev *)data;
+	unsigned ier = AT91_TWI_TXCOMP;
+
+	dma_unmap_single(dev->dev, sg_dma_address(&dev->dma.sg[0]),
+			 dev->buf_len, DMA_FROM_DEVICE);
+
+	if (!dev->use_alt_cmd) {
+		/* The last two bytes have to be read without using dma */
+		dev->buf += dev->buf_len - 2;
+		dev->buf_len = 2;
+		ier |= AT91_TWI_RXRDY;
+	}
+	at91_twi_write(dev, AT91_TWI_IER, ier);
+}
+
+static void at91_twi_read_data_dma(struct at91_twi_dev *dev)
+{
+	dma_addr_t dma_addr;
+	struct dma_async_tx_descriptor *rxdesc;
+	struct at91_twi_dma *dma = &dev->dma;
+	struct dma_chan *chan_rx = dma->chan_rx;
+	size_t buf_len;
+
+	buf_len = (dev->use_alt_cmd) ? dev->buf_len : dev->buf_len - 2;
+	dma->direction = DMA_FROM_DEVICE;
+
+	/* Keep in mind that we won't use dma to read the last two bytes */
+	at91_twi_irq_save(dev);
+	dma_addr = dma_map_single(dev->dev, dev->buf, buf_len, DMA_FROM_DEVICE);
+	if (dma_mapping_error(dev->dev, dma_addr)) {
+		dev_err(dev->dev, "dma map failed\n");
+		return;
+	}
+	dma->buf_mapped = true;
+	at91_twi_irq_restore(dev);
+
+	if (dev->fifo_size && IS_ALIGNED(buf_len, 4)) {
+		unsigned fifo_mr;
+
+		/*
+		 * DMA controller is triggered when at least 4 data can be
+		 * read from the RX FIFO
+		 */
+		fifo_mr = at91_twi_read(dev, AT91_TWI_FMR);
+		fifo_mr &= ~AT91_TWI_FMR_RXRDYM_MASK;
+		fifo_mr |= AT91_TWI_FMR_RXRDYM(AT91_TWI_FOUR_DATA);
+		at91_twi_write(dev, AT91_TWI_FMR, fifo_mr);
+	}
+
+	sg_dma_len(&dma->sg[0]) = buf_len;
+	sg_dma_address(&dma->sg[0]) = dma_addr;
+
+	rxdesc = dmaengine_prep_slave_sg(chan_rx, dma->sg, 1, DMA_DEV_TO_MEM,
+					 DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
+	if (!rxdesc) {
+		dev_err(dev->dev, "dma prep slave sg failed\n");
+		goto error;
+	}
+
+	rxdesc->callback = at91_twi_read_data_dma_callback;
+	rxdesc->callback_param = dev;
+
+	dma->xfer_in_progress = true;
+	dmaengine_submit(rxdesc);
+	dma_async_issue_pending(dma->chan_rx);
+
+	return;
+
+error:
+	at91_twi_dma_cleanup(dev);
+}
+
+static irqreturn_t atmel_twi_interrupt(int irq, void *dev_id)
+{
+	struct at91_twi_dev *dev = dev_id;
+	const unsigned status = at91_twi_read(dev, AT91_TWI_SR);
+	const unsigned irqstatus = status & at91_twi_read(dev, AT91_TWI_IMR);
+
+	if (!irqstatus)
+		return IRQ_NONE;
+	/*
+	 * In reception, the behavior of the twi device (before sama5d2) is
+	 * weird. There is some magic about RXRDY flag! When a data has been
+	 * almost received, the reception of a new one is anticipated if there
+	 * is no stop command to send. That is the reason why ask for sending
+	 * the stop command not on the last data but on the second last one.
+	 *
+	 * Unfortunately, we could still have the RXRDY flag set even if the
+	 * transfer is done and we have read the last data. It might happen
+	 * when the i2c slave device sends too quickly data after receiving the
+	 * ack from the master. The data has been almost received before having
+	 * the order to send stop. In this case, sending the stop command could
+	 * cause a RXRDY interrupt with a TXCOMP one. It is better to manage
+	 * the RXRDY interrupt first in order to not keep garbage data in the
+	 * Receive Holding Register for the next transfer.
+	 */
+	if (irqstatus & AT91_TWI_RXRDY) {
+		/*
+		 * Read all available bytes at once by polling RXRDY usable w/
+		 * and w/o FIFO. With FIFO enabled we could also read RXFL and
+		 * avoid polling RXRDY.
+		 */
+		do {
+			at91_twi_read_next_byte(dev);
+		} while (at91_twi_read(dev, AT91_TWI_SR) & AT91_TWI_RXRDY);
+	}
+
+	/*
+	 * When a NACK condition is detected, the I2C controller sets the NACK,
+	 * TXCOMP and TXRDY bits all together in the Status Register (SR).
+	 *
+	 * 1 - Handling NACK errors with CPU write transfer.
+	 *
+	 * In such case, we should not write the next byte into the Transmit
+	 * Holding Register (THR) otherwise the I2C controller would start a new
+	 * transfer and the I2C slave is likely to reply by another NACK.
+	 *
+	 * 2 - Handling NACK errors with DMA write transfer.
+	 *
+	 * By setting the TXRDY bit in the SR, the I2C controller also triggers
+	 * the DMA controller to write the next data into the THR. Then the
+	 * result depends on the hardware version of the I2C controller.
+	 *
+	 * 2a - Without support of the Alternative Command mode.
+	 *
+	 * This is the worst case: the DMA controller is triggered to write the
+	 * next data into the THR, hence starting a new transfer: the I2C slave
+	 * is likely to reply by another NACK.
+	 * Concurrently, this interrupt handler is likely to be called to manage
+	 * the first NACK before the I2C controller detects the second NACK and
+	 * sets once again the NACK bit into the SR.
+	 * When handling the first NACK, this interrupt handler disables the I2C
+	 * controller interruptions, especially the NACK interrupt.
+	 * Hence, the NACK bit is pending into the SR. This is why we should
+	 * read the SR to clear all pending interrupts at the beginning of
+	 * at91_do_twi_transfer() before actually starting a new transfer.
+	 *
+	 * 2b - With support of the Alternative Command mode.
+	 *
+	 * When a NACK condition is detected, the I2C controller also locks the
+	 * THR (and sets the LOCK bit in the SR): even though the DMA controller
+	 * is triggered by the TXRDY bit to write the next data into the THR,
+	 * this data actually won't go on the I2C bus hence a second NACK is not
+	 * generated.
+	 */
+	if (irqstatus & (AT91_TWI_TXCOMP | AT91_TWI_NACK)) {
+		at91_disable_twi_interrupts(dev);
+		complete(&dev->cmd_complete);
+	} else if (irqstatus & AT91_TWI_TXRDY) {
+		at91_twi_write_next_byte(dev);
+	}
+
+	/* catch error flags */
+	dev->transfer_status |= status;
+
+	return IRQ_HANDLED;
+}
+
+static int at91_do_twi_transfer(struct at91_twi_dev *dev)
+{
+	int ret;
+	unsigned long time_left;
+	bool has_unre_flag = dev->pdata->has_unre_flag;
+	bool has_alt_cmd = dev->pdata->has_alt_cmd;
+
+	/*
+	 * WARNING: the TXCOMP bit in the Status Register is NOT a clear on
+	 * read flag but shows the state of the transmission at the time the
+	 * Status Register is read. According to the programmer datasheet,
+	 * TXCOMP is set when both holding register and internal shifter are
+	 * empty and STOP condition has been sent.
+	 * Consequently, we should enable NACK interrupt rather than TXCOMP to
+	 * detect transmission failure.
+	 * Indeed let's take the case of an i2c write command using DMA.
+	 * Whenever the slave doesn't acknowledge a byte, the LOCK, NACK and
+	 * TXCOMP bits are set together into the Status Register.
+	 * LOCK is a clear on write bit, which is set to prevent the DMA
+	 * controller from sending new data on the i2c bus after a NACK
+	 * condition has happened. Once locked, this i2c peripheral stops
+	 * triggering the DMA controller for new data but it is more than
+	 * likely that a new DMA transaction is already in progress, writing
+	 * into the Transmit Holding Register. Since the peripheral is locked,
+	 * these new data won't be sent to the i2c bus but they will remain
+	 * into the Transmit Holding Register, so TXCOMP bit is cleared.
+	 * Then when the interrupt handler is called, the Status Register is
+	 * read: the TXCOMP bit is clear but NACK bit is still set. The driver
+	 * manage the error properly, without waiting for timeout.
+	 * This case can be reproduced easyly when writing into an at24 eeprom.
+	 *
+	 * Besides, the TXCOMP bit is already set before the i2c transaction
+	 * has been started. For read transactions, this bit is cleared when
+	 * writing the START bit into the Control Register. So the
+	 * corresponding interrupt can safely be enabled just after.
+	 * However for write transactions managed by the CPU, we first write
+	 * into THR, so TXCOMP is cleared. Then we can safely enable TXCOMP
+	 * interrupt. If TXCOMP interrupt were enabled before writing into THR,
+	 * the interrupt handler would be called immediately and the i2c command
+	 * would be reported as completed.
+	 * Also when a write transaction is managed by the DMA controller,
+	 * enabling the TXCOMP interrupt in this function may lead to a race
+	 * condition since we don't know whether the TXCOMP interrupt is enabled
+	 * before or after the DMA has started to write into THR. So the TXCOMP
+	 * interrupt is enabled later by at91_twi_write_data_dma_callback().
+	 * Immediately after in that DMA callback, if the alternative command
+	 * mode is not used, we still need to send the STOP condition manually
+	 * writing the corresponding bit into the Control Register.
+	 */
+
+	dev_dbg(dev->dev, "transfer: %s %zu bytes.\n",
+		(dev->msg->flags & I2C_M_RD) ? "read" : "write", dev->buf_len);
+
+	reinit_completion(&dev->cmd_complete);
+	dev->transfer_status = 0;
+
+	/* Clear pending interrupts, such as NACK. */
+	at91_twi_read(dev, AT91_TWI_SR);
+
+	if (dev->fifo_size) {
+		unsigned fifo_mr = at91_twi_read(dev, AT91_TWI_FMR);
+
+		/* Reset FIFO mode register */
+		fifo_mr &= ~(AT91_TWI_FMR_TXRDYM_MASK |
+			     AT91_TWI_FMR_RXRDYM_MASK);
+		fifo_mr |= AT91_TWI_FMR_TXRDYM(AT91_TWI_ONE_DATA);
+		fifo_mr |= AT91_TWI_FMR_RXRDYM(AT91_TWI_ONE_DATA);
+		at91_twi_write(dev, AT91_TWI_FMR, fifo_mr);
+
+		/* Flush FIFOs */
+		at91_twi_write(dev, AT91_TWI_CR,
+			       AT91_TWI_THRCLR | AT91_TWI_RHRCLR);
+	}
+
+	if (!dev->buf_len) {
+		at91_twi_write(dev, AT91_TWI_CR, AT91_TWI_QUICK);
+		at91_twi_write(dev, AT91_TWI_IER, AT91_TWI_TXCOMP);
+	} else if (dev->msg->flags & I2C_M_RD) {
+		unsigned start_flags = AT91_TWI_START;
+
+		/* if only one byte is to be read, immediately stop transfer */
+		if (!dev->use_alt_cmd && dev->buf_len <= 1 &&
+		    !(dev->msg->flags & I2C_M_RECV_LEN))
+			start_flags |= AT91_TWI_STOP;
+		at91_twi_write(dev, AT91_TWI_CR, start_flags);
+		/*
+		 * When using dma without alternative command mode, the last
+		 * byte has to be read manually in order to not send the stop
+		 * command too late and then to receive extra data.
+		 * In practice, there are some issues if you use the dma to
+		 * read n-1 bytes because of latency.
+		 * Reading n-2 bytes with dma and the two last ones manually
+		 * seems to be the best solution.
+		 */
+		if (dev->use_dma && (dev->buf_len > AT91_I2C_DMA_THRESHOLD)) {
+			at91_twi_write(dev, AT91_TWI_IER, AT91_TWI_NACK);
+			at91_twi_read_data_dma(dev);
+		} else {
+			at91_twi_write(dev, AT91_TWI_IER,
+				       AT91_TWI_TXCOMP |
+				       AT91_TWI_NACK |
+				       AT91_TWI_RXRDY);
+		}
+	} else {
+		if (dev->use_dma && (dev->buf_len > AT91_I2C_DMA_THRESHOLD)) {
+			at91_twi_write(dev, AT91_TWI_IER, AT91_TWI_NACK);
+			at91_twi_write_data_dma(dev);
+		} else {
+			at91_twi_write_next_byte(dev);
+			at91_twi_write(dev, AT91_TWI_IER,
+				       AT91_TWI_TXCOMP |
+				       AT91_TWI_NACK |
+				       AT91_TWI_TXRDY);
+		}
+	}
+
+	time_left = wait_for_completion_timeout(&dev->cmd_complete,
+					      dev->adapter.timeout);
+	if (time_left == 0) {
+		dev->transfer_status |= at91_twi_read(dev, AT91_TWI_SR);
+		dev_err(dev->dev, "controller timed out\n");
+		at91_init_twi_bus(dev);
+		ret = -ETIMEDOUT;
+		goto error;
+	}
+	if (dev->transfer_status & AT91_TWI_NACK) {
+		dev_dbg(dev->dev, "received nack\n");
+		ret = -EREMOTEIO;
+		goto error;
+	}
+	if (dev->transfer_status & AT91_TWI_OVRE) {
+		dev_err(dev->dev, "overrun while reading\n");
+		ret = -EIO;
+		goto error;
+	}
+	if (has_unre_flag && dev->transfer_status & AT91_TWI_UNRE) {
+		dev_err(dev->dev, "underrun while writing\n");
+		ret = -EIO;
+		goto error;
+	}
+	if ((has_alt_cmd || dev->fifo_size) &&
+	    (dev->transfer_status & AT91_TWI_LOCK)) {
+		dev_err(dev->dev, "tx locked\n");
+		ret = -EIO;
+		goto error;
+	}
+	if (dev->recv_len_abort) {
+		dev_err(dev->dev, "invalid smbus block length recvd\n");
+		ret = -EPROTO;
+		goto error;
+	}
+
+	dev_dbg(dev->dev, "transfer complete\n");
+
+	return 0;
+
+error:
+	/* first stop DMA transfer if still in progress */
+	at91_twi_dma_cleanup(dev);
+	/* then flush THR/FIFO and unlock TX if locked */
+	if ((has_alt_cmd || dev->fifo_size) &&
+	    (dev->transfer_status & AT91_TWI_LOCK)) {
+		dev_dbg(dev->dev, "unlock tx\n");
+		at91_twi_write(dev, AT91_TWI_CR,
+			       AT91_TWI_THRCLR | AT91_TWI_LOCKCLR);
+	}
+	return ret;
+}
+
+static int at91_twi_xfer(struct i2c_adapter *adap, struct i2c_msg *msg, int num)
+{
+	struct at91_twi_dev *dev = i2c_get_adapdata(adap);
+	int ret;
+	unsigned int_addr_flag = 0;
+	struct i2c_msg *m_start = msg;
+	bool is_read;
+
+	dev_dbg(&adap->dev, "at91_xfer: processing %d messages:\n", num);
+
+	ret = pm_runtime_get_sync(dev->dev);
+	if (ret < 0)
+		goto out;
+
+	if (num == 2) {
+		int internal_address = 0;
+		int i;
+
+		/* 1st msg is put into the internal address, start with 2nd */
+		m_start = &msg[1];
+		for (i = 0; i < msg->len; ++i) {
+			const unsigned addr = msg->buf[msg->len - 1 - i];
+
+			internal_address |= addr << (8 * i);
+			int_addr_flag += AT91_TWI_IADRSZ_1;
+		}
+		at91_twi_write(dev, AT91_TWI_IADR, internal_address);
+	}
+
+	dev->use_alt_cmd = false;
+	is_read = (m_start->flags & I2C_M_RD);
+	if (dev->pdata->has_alt_cmd) {
+		if (m_start->len > 0 &&
+		    m_start->len < AT91_I2C_MAX_ALT_CMD_DATA_SIZE) {
+			at91_twi_write(dev, AT91_TWI_CR, AT91_TWI_ACMEN);
+			at91_twi_write(dev, AT91_TWI_ACR,
+				       AT91_TWI_ACR_DATAL(m_start->len) |
+				       ((is_read) ? AT91_TWI_ACR_DIR : 0));
+			dev->use_alt_cmd = true;
+		} else {
+			at91_twi_write(dev, AT91_TWI_CR, AT91_TWI_ACMDIS);
+		}
+	}
+
+	at91_twi_write(dev, AT91_TWI_MMR,
+		       (m_start->addr << 16) |
+		       int_addr_flag |
+		       ((!dev->use_alt_cmd && is_read) ? AT91_TWI_MREAD : 0));
+
+	dev->buf_len = m_start->len;
+	dev->buf = m_start->buf;
+	dev->msg = m_start;
+	dev->recv_len_abort = false;
+
+	ret = at91_do_twi_transfer(dev);
+
+	ret = (ret < 0) ? ret : num;
+out:
+	pm_runtime_mark_last_busy(dev->dev);
+	pm_runtime_put_autosuspend(dev->dev);
+
+	return ret;
+}
+
+/*
+ * The hardware can handle at most two messages concatenated by a
+ * repeated start via it's internal address feature.
+ */
+static const struct i2c_adapter_quirks at91_twi_quirks = {
+	.flags = I2C_AQ_COMB | I2C_AQ_COMB_WRITE_FIRST | I2C_AQ_COMB_SAME_ADDR,
+	.max_comb_1st_msg_len = 3,
+};
+
+static u32 at91_twi_func(struct i2c_adapter *adapter)
+{
+	return I2C_FUNC_I2C | I2C_FUNC_SMBUS_EMUL
+		| I2C_FUNC_SMBUS_READ_BLOCK_DATA;
+}
+
+static const struct i2c_algorithm at91_twi_algorithm = {
+	.master_xfer	= at91_twi_xfer,
+	.functionality	= at91_twi_func,
+};
+
+static int at91_twi_configure_dma(struct at91_twi_dev *dev, u32 phy_addr)
+{
+	int ret = 0;
+	struct dma_slave_config slave_config;
+	struct at91_twi_dma *dma = &dev->dma;
+	enum dma_slave_buswidth addr_width = DMA_SLAVE_BUSWIDTH_1_BYTE;
+
+	/*
+	 * The actual width of the access will be chosen in
+	 * dmaengine_prep_slave_sg():
+	 * for each buffer in the scatter-gather list, if its size is aligned
+	 * to addr_width then addr_width accesses will be performed to transfer
+	 * the buffer. On the other hand, if the buffer size is not aligned to
+	 * addr_width then the buffer is transferred using single byte accesses.
+	 * Please refer to the Atmel eXtended DMA controller driver.
+	 * When FIFOs are used, the TXRDYM threshold can always be set to
+	 * trigger the XDMAC when at least 4 data can be written into the TX
+	 * FIFO, even if single byte accesses are performed.
+	 * However the RXRDYM threshold must be set to fit the access width,
+	 * deduced from buffer length, so the XDMAC is triggered properly to
+	 * read data from the RX FIFO.
+	 */
+	if (dev->fifo_size)
+		addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
+
+	memset(&slave_config, 0, sizeof(slave_config));
+	slave_config.src_addr = (dma_addr_t)phy_addr + AT91_TWI_RHR;
+	slave_config.src_addr_width = addr_width;
+	slave_config.src_maxburst = 1;
+	slave_config.dst_addr = (dma_addr_t)phy_addr + AT91_TWI_THR;
+	slave_config.dst_addr_width = addr_width;
+	slave_config.dst_maxburst = 1;
+	slave_config.device_fc = false;
+
+	dma->chan_tx = dma_request_slave_channel_reason(dev->dev, "tx");
+	if (IS_ERR(dma->chan_tx)) {
+		ret = PTR_ERR(dma->chan_tx);
+		dma->chan_tx = NULL;
+		goto error;
+	}
+
+	dma->chan_rx = dma_request_slave_channel_reason(dev->dev, "rx");
+	if (IS_ERR(dma->chan_rx)) {
+		ret = PTR_ERR(dma->chan_rx);
+		dma->chan_rx = NULL;
+		goto error;
+	}
+
+	slave_config.direction = DMA_MEM_TO_DEV;
+	if (dmaengine_slave_config(dma->chan_tx, &slave_config)) {
+		dev_err(dev->dev, "failed to configure tx channel\n");
+		ret = -EINVAL;
+		goto error;
+	}
+
+	slave_config.direction = DMA_DEV_TO_MEM;
+	if (dmaengine_slave_config(dma->chan_rx, &slave_config)) {
+		dev_err(dev->dev, "failed to configure rx channel\n");
+		ret = -EINVAL;
+		goto error;
+	}
+
+	sg_init_table(dma->sg, 2);
+	dma->buf_mapped = false;
+	dma->xfer_in_progress = false;
+	dev->use_dma = true;
+
+	dev_info(dev->dev, "using %s (tx) and %s (rx) for DMA transfers\n",
+		 dma_chan_name(dma->chan_tx), dma_chan_name(dma->chan_rx));
+
+	return ret;
+
+error:
+	if (ret != -EPROBE_DEFER)
+		dev_info(dev->dev, "can't get DMA channel, continue without DMA support\n");
+	if (dma->chan_rx)
+		dma_release_channel(dma->chan_rx);
+	if (dma->chan_tx)
+		dma_release_channel(dma->chan_tx);
+	return ret;
+}
+
+int at91_twi_probe_master(struct platform_device *pdev,
+			  u32 phy_addr, struct at91_twi_dev *dev)
+{
+	int rc;
+
+	init_completion(&dev->cmd_complete);
+
+	rc = devm_request_irq(&pdev->dev, dev->irq, atmel_twi_interrupt, 0,
+			      dev_name(dev->dev), dev);
+	if (rc) {
+		dev_err(dev->dev, "Cannot get irq %d: %d\n", dev->irq, rc);
+		return rc;
+	}
+
+	if (dev->dev->of_node) {
+		rc = at91_twi_configure_dma(dev, phy_addr);
+		if (rc == -EPROBE_DEFER)
+			return rc;
+	}
+
+	if (!of_property_read_u32(pdev->dev.of_node, "atmel,fifo-size",
+				  &dev->fifo_size)) {
+		dev_info(dev->dev, "Using FIFO (%u data)\n", dev->fifo_size);
+	}
+
+	at91_calc_twi_clock(dev);
+
+	dev->adapter.algo = &at91_twi_algorithm;
+	dev->adapter.quirks = &at91_twi_quirks;
+
+	return 0;
+}