+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/i2c.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/sched.h>
+#include <linux/delay.h>
+#include <linux/errno.h>
+#include <linux/err.h>
+#include <linux/device.h>
+#include <linux/platform_device.h>
+#include <linux/mm.h>
+#include <linux/dma-mapping.h>
+#include <linux/scatterlist.h>
+#include <linux/io.h>
+#include <linux/of_address.h>
+#include <linux/of_irq.h>
+#include <linux/clk.h>
+#include <linux/completion.h>

Please sort the includes to avoid duplicates.

+struct mtk_i2c_compatible {
+	const struct i2c_adapter_quirks *quirks;
+	unsigned char pmic_i2c;
+	unsigned char dcm;
+};

I wonder if the unsigned char options can be bits in a flags variable?

+static const struct i2c_adapter_quirks mt6577_i2c_quirks = {
+	.flags = I2C_AQ_COMB_WRITE_THEN_READ,
+	.max_num_msgs = MAX_MSG_NUM_MT6577,
+	.max_write_len = MAX_DMA_TRANS_SIZE_MT6577,
+	.max_read_len = MAX_DMA_TRANS_SIZE_MT6577,
+	.max_comb_1st_msg_len = MAX_DMA_TRANS_SIZE_MT6577,
+	.max_comb_2nd_msg_len = MAX_WRRD_TRANS_SIZE_MT6577,
+};

I would think plain numbers are much more readable than defines here.
They are used only once, too.

+static const struct of_device_id mtk_i2c_of_match[] = {
+	{ .compatible = "mediatek,mt6577-i2c", .data = (void *)&mt6577_compat },
+	{ .compatible = "mediatek,mt6589-i2c", .data = (void *)&mt6589_compat },
+	{}
+};
+MODULE_DEVICE_TABLE(of, mtk_i2c_of_match);

No need for casts.

+static inline void mtk_i2c_writew(u16 value, struct mtk_i2c *i2c, u8 offset)
+{
+	writew(value, i2c->base + offset);
+}
+
+static inline u16 mtk_i2c_readw(struct mtk_i2c *i2c, u8 offset)
+{
+	return readw(i2c->base + offset);
+}

I am not a big fan of such extremly thin wrappers, but if you like them...

+		rpaddr = dma_map_single(i2c->adap.dev.parent, msgs->buf,
+						msgs->len, DMA_FROM_DEVICE);

I think you shouldn't use the adapter device here and later, but the dma
channel device.

+	/* flush before sending start */
+	mb();
+	mtk_i2c_writel_dma(I2C_DMA_START_EN, i2c, OFFSET_EN);

Is mb() really needed when you use writel after that?

+	if (i2c->irq_stat & (I2C_HS_NACKERR | I2C_ACKERR)) {
+		dev_dbg(i2c->dev, "addr: %x, transfer ACK error\n", msgs->addr);
+		mtk_i2c_init_hw(i2c);
+		return -EREMOTEIO;

-ENXIO. Please check Documentation/i2c/fault-codes for a reference and
check all your error values.

+	ret = of_property_read_u32(np, "clock-div", clk_src_div);
+	if (ret < 0)
+		return ret;

Do we need a property for that in the i2c-block? Can't the clock
provider driver deliver a properly divided clock already?


And cppcheck says:

drivers/i2c/busses/i2c-mt65xx.c:133: style: struct or union member 'mtk_i2c_data::clk_frequency' is never used.
drivers/i2c/busses/i2c-mt65xx.c:135: style: struct or union member 'mtk_i2c_data::clk_src_div' is never used.

Rest looks good, thanks!

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20150512/6e08758c/attachment.sig>