ar71xx: boost SPI flash read performance

mtd_speedtest results:
                           page read speed
                    old         new        delta
DB120            929 KiB/s   2597 KiB/s   +179.55%
TL-WR1043ND v1   754 KiB/s   2166 KiB/s   +187.27%
TL-WR703N v1     745 KiB/s   2176 KiB/s   +192.08%
TL-MR3220 v1     752 KiB/s   2154 KiB/s   +186.44%
TL-WR2543ND v1   564 KiB/s   2130 KiB/s   +277.66%
TL-WR741ND v2    525 KiB/s   1767 KiB/s   +236.57%
ALFA-AP96        702 KiB/s   1977 KiB/s   +181.62%
WNDR3700         697 KiB/s   1965 KiB/s   +181.92%

git-svn-id: svn://svn.openwrt.org/openwrt/trunk@31118 3c298f89-4303-0410-b956-a3cf2f4a3e73

target/linux/ar71xx/files/arch/mips/ath79/dev-m25p80.c

@@ -51,6 +51,7 @@ void __init ath79_register_m25p80(struct flash_platform_data *pdata)
 {
 	ath79_spi_data.bus_num = 0;
 	ath79_spi_data.num_chipselect = 1;
+	ath79_spi0_cdata.is_flash = true;
 	ath79_spi_info[0].platform_data = pdata;
 	ath79_register_spi(&ath79_spi_data, ath79_spi_info, 1);
 }

target/linux/ar71xx/files/arch/mips/ath79/mach-alfa-ap96.c

@@ -62,6 +62,7 @@ static struct mmc_spi_platform_data alfa_ap96_mmc_data = {
 static struct ath79_spi_controller_data ap96_spi0_cdata = {
 	.cs_type = ATH79_SPI_CS_TYPE_INTERNAL,
 	.cs_line = 0,
+	.is_flash = true,
 };
 
 static struct ath79_spi_controller_data ap96_spi1_cdata = {

target/linux/ar71xx/files/arch/mips/ath79/mach-tl-wr2543n.c

@@ -40,7 +40,6 @@ static const char *tl_wr2543n_part_probes[] = {
 
 static struct flash_platform_data tl_wr2543n_flash_data = {
 	.part_probes	= tl_wr2543n_part_probes,
-	.max_read_len	= 64,
 };
 
 static struct gpio_led tl_wr2543n_leds_gpio[] __initdata = {

target/linux/ar71xx/patches-3.2/460-spi-bitbang-export-spi_bitbang_bufs.patch

@@ -0,0 +1,28 @@
+--- a/drivers/spi/spi-bitbang.c
++++ b/drivers/spi/spi-bitbang.c
+@@ -234,13 +234,14 @@ void spi_bitbang_cleanup(struct spi_devi
+ }
+ EXPORT_SYMBOL_GPL(spi_bitbang_cleanup);
+ 
+-static int spi_bitbang_bufs(struct spi_device *spi, struct spi_transfer *t)
++int spi_bitbang_bufs(struct spi_device *spi, struct spi_transfer *t)
+ {
+ 	struct spi_bitbang_cs	*cs = spi->controller_state;
+ 	unsigned		nsecs = cs->nsecs;
+ 
+ 	return cs->txrx_bufs(spi, cs->txrx_word, nsecs, t);
+ }
++EXPORT_SYMBOL_GPL(spi_bitbang_bufs);
+ 
+ /*----------------------------------------------------------------------*/
+ 
+--- a/include/linux/spi/spi_bitbang.h
++++ b/include/linux/spi/spi_bitbang.h
+@@ -44,6 +44,7 @@ extern void spi_bitbang_cleanup(struct s
+ extern int spi_bitbang_transfer(struct spi_device *spi, struct spi_message *m);
+ extern int spi_bitbang_setup_transfer(struct spi_device *spi,
+ 				      struct spi_transfer *t);
++extern int spi_bitbang_bufs(struct spi_device *spi, struct spi_transfer *t);
+ 
+ /* start or stop queue processing */
+ extern int spi_bitbang_start(struct spi_bitbang *spi);

target/linux/ar71xx/patches-3.2/461-spi-add-type-field-to-spi_transfer.patch

@@ -0,0 +1,23 @@
+--- a/include/linux/spi/spi.h
++++ b/include/linux/spi/spi.h
+@@ -344,6 +344,12 @@ extern struct spi_master *spi_busnum_to_
+ 
+ /*---------------------------------------------------------------------------*/
+ 
++enum spi_transfer_type {
++	SPI_TRANSFER_GENERIC = 0,
++	SPI_TRANSFER_FLASH_READ_CMD,
++	SPI_TRANSFER_FLASH_READ_DATA,
++};
++
+ /*
+  * I/O INTERFACE between SPI controller and protocol drivers
+  *
+@@ -446,6 +452,7 @@ struct spi_transfer {
+ 	u8		bits_per_word;
+ 	u16		delay_usecs;
+ 	u32		speed_hz;
++	enum spi_transfer_type type;
+ 
+ 	struct list_head transfer_list;
+ };

target/linux/ar71xx/patches-3.2/462-mtd-m25p80-set-spi-transfer-type.patch

@@ -0,0 +1,15 @@
+--- a/drivers/mtd/devices/m25p80.c
++++ b/drivers/mtd/devices/m25p80.c
+@@ -372,10 +372,12 @@ static int m25p80_read(struct mtd_info *
+ 	 * OPCODE_FAST_READ (if available) is faster.
+ 	 * Should add 1 byte DUMMY_BYTE.
+ 	 */
++	t[0].type = SPI_TRANSFER_FLASH_READ_CMD;
+ 	t[0].tx_buf = flash->command;
+ 	t[0].len = m25p_cmdsz(flash) + FAST_READ_DUMMY_BYTE;
+ 	spi_message_add_tail(&t[0], &m);
+ 
++	t[1].type = SPI_TRANSFER_FLASH_READ_DATA;
+ 	spi_message_add_tail(&t[1], &m);
+ 
+ 	/* Byte count starts at zero. */

target/linux/ar71xx/patches-3.2/463-spi-ath79-add-fast-flash-read.patch

@@ -0,0 +1,185 @@
+--- a/drivers/spi/spi-ath79.c
++++ b/drivers/spi/spi-ath79.c
+@@ -37,6 +37,11 @@
+ 
+ #define ATH79_SPI_CS_LINE_MAX		2
+ 
++enum ath79_spi_state {
++	ATH79_SPI_STATE_WAIT_CMD = 0,
++	ATH79_SPI_STATE_WAIT_READ,
++};
++
+ struct ath79_spi {
+ 	struct spi_bitbang	bitbang;
+ 	u32			ioc_base;
+@@ -44,6 +49,11 @@ struct ath79_spi {
+ 	void __iomem		*base;
+ 	struct clk		*clk;
+ 	unsigned		rrw_delay;
++
++	enum ath79_spi_state	state;
++	u32			clk_div;
++	unsigned long 		read_addr;
++	unsigned long		ahb_rate;
+ };
+ 
+ static inline u32 ath79_spi_rr(struct ath79_spi *sp, unsigned reg)
+@@ -108,9 +118,6 @@ static void ath79_spi_enable(struct ath7
+ 	/* save CTRL register */
+ 	sp->reg_ctrl = ath79_spi_rr(sp, AR71XX_SPI_REG_CTRL);
+ 	sp->ioc_base = ath79_spi_rr(sp, AR71XX_SPI_REG_IOC);
+-
+-	/* TODO: setup speed? */
+-	ath79_spi_wr(sp, AR71XX_SPI_REG_CTRL, 0x43);
+ }
+ 
+ static void ath79_spi_disable(struct ath79_spi *sp)
+@@ -222,6 +229,110 @@ static u32 ath79_spi_txrx_mode0(struct s
+ 	return ath79_spi_rr(sp, AR71XX_SPI_REG_RDS);
+ }
+ 
++static int ath79_spi_do_read_flash_data(struct spi_device *spi,
++					struct spi_transfer *t)
++{
++	struct ath79_spi *sp = ath79_spidev_to_sp(spi);
++
++	/* disable GPIO mode */
++	ath79_spi_wr(sp, AR71XX_SPI_REG_FS, 0);
++
++	memcpy_fromio(t->rx_buf, sp->base + sp->read_addr, t->len);
++
++	/* enable GPIO mode */
++	ath79_spi_wr(sp, AR71XX_SPI_REG_FS, AR71XX_SPI_FS_GPIO);
++
++	/* restore IOC register */
++	ath79_spi_wr(sp, AR71XX_SPI_REG_IOC, sp->ioc_base);
++
++	return t->len;
++}
++
++static int ath79_spi_do_read_flash_cmd(struct spi_device *spi,
++				       struct spi_transfer *t)
++{
++	struct ath79_spi *sp = ath79_spidev_to_sp(spi);
++	int len;
++	const u8 *p;
++
++	sp->read_addr = 0;
++
++	len = t->len - 1;
++	p = t->tx_buf;
++
++	while (len--) {
++		p++;
++		sp->read_addr <<= 8;
++		sp->read_addr |= *p;
++	}
++
++	return t->len;
++}
++
++static bool ath79_spi_is_read_cmd(struct spi_device *spi,
++				 struct spi_transfer *t)
++{
++	return t->type == SPI_TRANSFER_FLASH_READ_CMD;
++}
++
++static bool ath79_spi_is_data_read(struct spi_device *spi,
++				  struct spi_transfer *t)
++{
++	return t->type == SPI_TRANSFER_FLASH_READ_DATA;
++}
++
++static int ath79_spi_txrx_bufs(struct spi_device *spi, struct spi_transfer *t)
++{
++	struct ath79_spi *sp = ath79_spidev_to_sp(spi);
++	int ret;
++
++	switch (sp->state) {
++	case ATH79_SPI_STATE_WAIT_CMD:
++		if (ath79_spi_is_read_cmd(spi, t)) {
++			ret = ath79_spi_do_read_flash_cmd(spi, t);
++			sp->state = ATH79_SPI_STATE_WAIT_READ;
++		} else {
++			ret = spi_bitbang_bufs(spi, t);
++		}
++		break;
++
++	case ATH79_SPI_STATE_WAIT_READ:
++		if (ath79_spi_is_data_read(spi, t)) {
++			ret = ath79_spi_do_read_flash_data(spi, t);
++		} else {
++			dev_warn(&spi->dev, "flash data read expected\n");
++			ret = -EIO;
++		}
++		sp->state = ATH79_SPI_STATE_WAIT_CMD;
++		break;
++
++	default:
++		BUG();
++	}
++
++	return ret;
++}
++
++static int ath79_spi_setup_transfer(struct spi_device *spi,
++				    struct spi_transfer *t)
++{
++	struct ath79_spi *sp = ath79_spidev_to_sp(spi);
++	struct ath79_spi_controller_data *cdata;
++	int ret;
++
++	ret = spi_bitbang_setup_transfer(spi, t);
++	if (ret)
++		return ret;
++
++	cdata = spi->controller_data;
++	if (cdata->is_flash)
++		sp->bitbang.txrx_bufs = ath79_spi_txrx_bufs;
++	else
++		sp->bitbang.txrx_bufs = spi_bitbang_bufs;
++
++	return ret;
++}
++
+ static __devinit int ath79_spi_probe(struct platform_device *pdev)
+ {
+ 	struct spi_master *master;
+@@ -244,6 +355,8 @@ static __devinit int ath79_spi_probe(str
+ 	sp = spi_master_get_devdata(master);
+ 	platform_set_drvdata(pdev, sp);
+ 
++	sp->state = ATH79_SPI_STATE_WAIT_CMD;
++
+ 	master->setup = ath79_spi_setup;
+ 	master->cleanup = ath79_spi_cleanup;
+ 	master->bus_num = pdata->bus_num;
+@@ -252,7 +365,7 @@ static __devinit int ath79_spi_probe(str
+ 	sp->bitbang.master = spi_master_get(master);
+ 	sp->bitbang.chipselect = ath79_spi_chipselect;
+ 	sp->bitbang.txrx_word[SPI_MODE_0] = ath79_spi_txrx_mode0;
+-	sp->bitbang.setup_transfer = spi_bitbang_setup_transfer;
++	sp->bitbang.setup_transfer = ath79_spi_setup_transfer;
+ 	sp->bitbang.flags = SPI_CS_HIGH;
+ 
+ 	r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+@@ -277,7 +390,8 @@ static __devinit int ath79_spi_probe(str
+ 	if (ret)
+ 		goto err_clk_put;
+ 
+-	rate = DIV_ROUND_UP(clk_get_rate(sp->clk), MHZ);
++	sp->ahb_rate = clk_get_rate(sp->clk);
++	rate = DIV_ROUND_UP(sp->ahb_rate, MHZ);
+ 	if (!rate) {
+ 		ret = -EINVAL;
+ 		goto err_clk_disable;
+--- a/arch/mips/include/asm/mach-ath79/ath79_spi_platform.h
++++ b/arch/mips/include/asm/mach-ath79/ath79_spi_platform.h
+@@ -24,6 +24,7 @@ enum ath79_spi_cs_type {
+ struct ath79_spi_controller_data {
+ 	enum ath79_spi_cs_type cs_type;
+ 	unsigned cs_line;
++	bool is_flash;
+ };
+ 
+ #endif /* _ATH79_SPI_PLATFORM_H */

target/linux/ar71xx/patches-3.3/460-spi-bitbang-export-spi_bitbang_bufs.patch

@@ -0,0 +1,28 @@
+--- a/drivers/spi/spi-bitbang.c
++++ b/drivers/spi/spi-bitbang.c
+@@ -234,13 +234,14 @@ void spi_bitbang_cleanup(struct spi_devi
+ }
+ EXPORT_SYMBOL_GPL(spi_bitbang_cleanup);
+ 
+-static int spi_bitbang_bufs(struct spi_device *spi, struct spi_transfer *t)
++int spi_bitbang_bufs(struct spi_device *spi, struct spi_transfer *t)
+ {
+ 	struct spi_bitbang_cs	*cs = spi->controller_state;
+ 	unsigned		nsecs = cs->nsecs;
+ 
+ 	return cs->txrx_bufs(spi, cs->txrx_word, nsecs, t);
+ }
++EXPORT_SYMBOL_GPL(spi_bitbang_bufs);
+ 
+ /*----------------------------------------------------------------------*/
+ 
+--- a/include/linux/spi/spi_bitbang.h
++++ b/include/linux/spi/spi_bitbang.h
+@@ -44,6 +44,7 @@ extern void spi_bitbang_cleanup(struct s
+ extern int spi_bitbang_transfer(struct spi_device *spi, struct spi_message *m);
+ extern int spi_bitbang_setup_transfer(struct spi_device *spi,
+ 				      struct spi_transfer *t);
++extern int spi_bitbang_bufs(struct spi_device *spi, struct spi_transfer *t);
+ 
+ /* start or stop queue processing */
+ extern int spi_bitbang_start(struct spi_bitbang *spi);

target/linux/ar71xx/patches-3.3/461-spi-add-type-field-to-spi_transfer.patch

@@ -0,0 +1,23 @@
+--- a/include/linux/spi/spi.h
++++ b/include/linux/spi/spi.h
+@@ -355,6 +355,12 @@ extern struct spi_master *spi_busnum_to_
+ 
+ /*---------------------------------------------------------------------------*/
+ 
++enum spi_transfer_type {
++	SPI_TRANSFER_GENERIC = 0,
++	SPI_TRANSFER_FLASH_READ_CMD,
++	SPI_TRANSFER_FLASH_READ_DATA,
++};
++
+ /*
+  * I/O INTERFACE between SPI controller and protocol drivers
+  *
+@@ -457,6 +463,7 @@ struct spi_transfer {
+ 	u8		bits_per_word;
+ 	u16		delay_usecs;
+ 	u32		speed_hz;
++	enum spi_transfer_type type;
+ 
+ 	struct list_head transfer_list;
+ };

target/linux/ar71xx/patches-3.3/462-mtd-m25p80-set-spi-transfer-type.patch

@@ -0,0 +1,15 @@
+--- a/drivers/mtd/devices/m25p80.c
++++ b/drivers/mtd/devices/m25p80.c
+@@ -372,10 +372,12 @@ static int m25p80_read(struct mtd_info *
+ 	 * OPCODE_FAST_READ (if available) is faster.
+ 	 * Should add 1 byte DUMMY_BYTE.
+ 	 */
++	t[0].type = SPI_TRANSFER_FLASH_READ_CMD;
+ 	t[0].tx_buf = flash->command;
+ 	t[0].len = m25p_cmdsz(flash) + FAST_READ_DUMMY_BYTE;
+ 	spi_message_add_tail(&t[0], &m);
+ 
++	t[1].type = SPI_TRANSFER_FLASH_READ_DATA;
+ 	spi_message_add_tail(&t[1], &m);
+ 
+ 	/* Byte count starts at zero. */

target/linux/ar71xx/patches-3.3/463-spi-ath79-add-fast-flash-read.patch

@@ -0,0 +1,185 @@
+--- a/drivers/spi/spi-ath79.c
++++ b/drivers/spi/spi-ath79.c
+@@ -37,6 +37,11 @@
+ 
+ #define ATH79_SPI_CS_LINE_MAX		2
+ 
++enum ath79_spi_state {
++	ATH79_SPI_STATE_WAIT_CMD = 0,
++	ATH79_SPI_STATE_WAIT_READ,
++};
++
+ struct ath79_spi {
+ 	struct spi_bitbang	bitbang;
+ 	u32			ioc_base;
+@@ -44,6 +49,11 @@ struct ath79_spi {
+ 	void __iomem		*base;
+ 	struct clk		*clk;
+ 	unsigned		rrw_delay;
++
++	enum ath79_spi_state	state;
++	u32			clk_div;
++	unsigned long 		read_addr;
++	unsigned long		ahb_rate;
+ };
+ 
+ static inline u32 ath79_spi_rr(struct ath79_spi *sp, unsigned reg)
+@@ -108,9 +118,6 @@ static void ath79_spi_enable(struct ath7
+ 	/* save CTRL register */
+ 	sp->reg_ctrl = ath79_spi_rr(sp, AR71XX_SPI_REG_CTRL);
+ 	sp->ioc_base = ath79_spi_rr(sp, AR71XX_SPI_REG_IOC);
+-
+-	/* TODO: setup speed? */
+-	ath79_spi_wr(sp, AR71XX_SPI_REG_CTRL, 0x43);
+ }
+ 
+ static void ath79_spi_disable(struct ath79_spi *sp)
+@@ -222,6 +229,110 @@ static u32 ath79_spi_txrx_mode0(struct s
+ 	return ath79_spi_rr(sp, AR71XX_SPI_REG_RDS);
+ }
+ 
++static int ath79_spi_do_read_flash_data(struct spi_device *spi,
++					struct spi_transfer *t)
++{
++	struct ath79_spi *sp = ath79_spidev_to_sp(spi);
++
++	/* disable GPIO mode */
++	ath79_spi_wr(sp, AR71XX_SPI_REG_FS, 0);
++
++	memcpy_fromio(t->rx_buf, sp->base + sp->read_addr, t->len);
++
++	/* enable GPIO mode */
++	ath79_spi_wr(sp, AR71XX_SPI_REG_FS, AR71XX_SPI_FS_GPIO);
++
++	/* restore IOC register */
++	ath79_spi_wr(sp, AR71XX_SPI_REG_IOC, sp->ioc_base);
++
++	return t->len;
++}
++
++static int ath79_spi_do_read_flash_cmd(struct spi_device *spi,
++				       struct spi_transfer *t)
++{
++	struct ath79_spi *sp = ath79_spidev_to_sp(spi);
++	int len;
++	const u8 *p;
++
++	sp->read_addr = 0;
++
++	len = t->len - 1;
++	p = t->tx_buf;
++
++	while (len--) {
++		p++;
++		sp->read_addr <<= 8;
++		sp->read_addr |= *p;
++	}
++
++	return t->len;
++}
++
++static bool ath79_spi_is_read_cmd(struct spi_device *spi,
++				 struct spi_transfer *t)
++{
++	return t->type == SPI_TRANSFER_FLASH_READ_CMD;
++}
++
++static bool ath79_spi_is_data_read(struct spi_device *spi,
++				  struct spi_transfer *t)
++{
++	return t->type == SPI_TRANSFER_FLASH_READ_DATA;
++}
++
++static int ath79_spi_txrx_bufs(struct spi_device *spi, struct spi_transfer *t)
++{
++	struct ath79_spi *sp = ath79_spidev_to_sp(spi);
++	int ret;
++
++	switch (sp->state) {
++	case ATH79_SPI_STATE_WAIT_CMD:
++		if (ath79_spi_is_read_cmd(spi, t)) {
++			ret = ath79_spi_do_read_flash_cmd(spi, t);
++			sp->state = ATH79_SPI_STATE_WAIT_READ;
++		} else {
++			ret = spi_bitbang_bufs(spi, t);
++		}
++		break;
++
++	case ATH79_SPI_STATE_WAIT_READ:
++		if (ath79_spi_is_data_read(spi, t)) {
++			ret = ath79_spi_do_read_flash_data(spi, t);
++		} else {
++			dev_warn(&spi->dev, "flash data read expected\n");
++			ret = -EIO;
++		}
++		sp->state = ATH79_SPI_STATE_WAIT_CMD;
++		break;
++
++	default:
++		BUG();
++	}
++
++	return ret;
++}
++
++static int ath79_spi_setup_transfer(struct spi_device *spi,
++				    struct spi_transfer *t)
++{
++	struct ath79_spi *sp = ath79_spidev_to_sp(spi);
++	struct ath79_spi_controller_data *cdata;
++	int ret;
++
++	ret = spi_bitbang_setup_transfer(spi, t);
++	if (ret)
++		return ret;
++
++	cdata = spi->controller_data;
++	if (cdata->is_flash)
++		sp->bitbang.txrx_bufs = ath79_spi_txrx_bufs;
++	else
++		sp->bitbang.txrx_bufs = spi_bitbang_bufs;
++
++	return ret;
++}
++
+ static __devinit int ath79_spi_probe(struct platform_device *pdev)
+ {
+ 	struct spi_master *master;
+@@ -244,6 +355,8 @@ static __devinit int ath79_spi_probe(str
+ 	sp = spi_master_get_devdata(master);
+ 	platform_set_drvdata(pdev, sp);
+ 
++	sp->state = ATH79_SPI_STATE_WAIT_CMD;
++
+ 	master->setup = ath79_spi_setup;
+ 	master->cleanup = ath79_spi_cleanup;
+ 	master->bus_num = pdata->bus_num;
+@@ -252,7 +365,7 @@ static __devinit int ath79_spi_probe(str
+ 	sp->bitbang.master = spi_master_get(master);
+ 	sp->bitbang.chipselect = ath79_spi_chipselect;
+ 	sp->bitbang.txrx_word[SPI_MODE_0] = ath79_spi_txrx_mode0;
+-	sp->bitbang.setup_transfer = spi_bitbang_setup_transfer;
++	sp->bitbang.setup_transfer = ath79_spi_setup_transfer;
+ 	sp->bitbang.flags = SPI_CS_HIGH;
+ 
+ 	r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+@@ -277,7 +390,8 @@ static __devinit int ath79_spi_probe(str
+ 	if (ret)
+ 		goto err_clk_put;
+ 
+-	rate = DIV_ROUND_UP(clk_get_rate(sp->clk), MHZ);
++	sp->ahb_rate = clk_get_rate(sp->clk);
++	rate = DIV_ROUND_UP(sp->ahb_rate, MHZ);
+ 	if (!rate) {
+ 		ret = -EINVAL;
+ 		goto err_clk_disable;
+--- a/arch/mips/include/asm/mach-ath79/ath79_spi_platform.h
++++ b/arch/mips/include/asm/mach-ath79/ath79_spi_platform.h
+@@ -24,6 +24,7 @@ enum ath79_spi_cs_type {
+ struct ath79_spi_controller_data {
+ 	enum ath79_spi_cs_type cs_type;
+ 	unsigned cs_line;
++	bool is_flash;
+ };
+ 
+ #endif /* _ATH79_SPI_PLATFORM_H */