[U-Boot] [PATCH 00/11] Enable edma support for ti-qspi

[U-Boot] [PATCH 00/11] Enable edma support for ti-qspi

[Vignesh R]

This patch series enables DMA for QSPI on dra7xx and am43xx.
Patch 1 and 2 make sure the buffers are mem aligned for dma operation.
Patch 3-6 add support for enabling and disabling of edma3 clocks on
am43xx and dra7xx. Patch 7-9 add dma functionality to ti_qspi driver.
Patch 10-11 enable edma3 support on dra7xx and am43xx in respective
config files.

Tested on DRA74 EVM: http://pastebin.ubuntu.com/11813977/
Tested on AM437X EVM SK: http://pastebin.ubuntu.com/11813994/



Kishon Vijay Abraham I (2):
  ARM: AM43xx: Add support for disabling clocks in uboot
  ARM: OMAP5: Add support for disabling clocks in uboot

Ravi Babu (2):
  sf: allocate cache aligned buffers to copy from flash
  env: use cache line aligned memory for flash read

Tom Rini (1):
  sf: ops: Add spi_flash_copy_mmap function

Vignesh R (6):
  ARM: OMAP5: Add functions to enable and disable EDMA3 clocks
  ARM: AM43XX: Add functions to enable and disable EDMA3 clocks
  spi: ti_qspi: Use DMA to read from qspi flash
  dma: ti-edma3: Add BIT(x) macro definition
  ARM: dra7xx_evm: Enable EDMA3 in SPL to support DMA on qspi
  ARM: am43xx_evm: Enable EDMA3 support DMA on qspi

 arch/arm/cpu/armv7/am33xx/clock.c                  | 52 ++++++++++++
 arch/arm/cpu/armv7/am33xx/clock_am43xx.c           | 36 +++++++++
 arch/arm/cpu/armv7/omap-common/clocks-common.c     | 53 ++++++++++++
 arch/arm/cpu/armv7/omap5/hw_data.c                 | 41 ++++++++++
 arch/arm/cpu/armv7/omap5/prcm-regs.c               |  4 +
 arch/arm/include/asm/arch-am33xx/clock.h           |  1 +
 arch/arm/include/asm/arch-am33xx/hardware_am43xx.h |  3 +
 arch/arm/include/asm/arch-omap5/hardware.h         |  5 ++
 arch/arm/include/asm/omap_common.h                 | 13 +++
 common/cmd_sf.c                                    |  6 +-
 common/env_sf.c                                    | 11 +--
 drivers/dma/ti-edma3.c                             |  2 +
 drivers/mtd/spi/sf_ops.c                           |  8 +-
 drivers/spi/ti_qspi.c                              | 93 ++++++++++++++++++++++
 include/configs/am43xx_evm.h                       |  2 +
 include/configs/dra7xx_evm.h                       |  2 +
 include/spi.h                                      |  3 +
 17 files changed, 326 insertions(+), 9 deletions(-)

-- 
2.4.5

[U-Boot] [PATCH 01/11] sf: allocate cache aligned buffers to copy from flash

[Vignesh R]

From: Ravi Babu <ravibabu@ti.com>

Use memalign() with ARCH_DMA_MINALIGN to allocate read buffers.
This is required because, flash drivers may use DMA for read operations
and may have to invalidate the buffer before read.

Signed-off-by: Ravi Babu <ravibabu@ti.com>
Signed-off-by: Vignesh R <vigneshr@ti.com>
---
 common/cmd_sf.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/common/cmd_sf.c b/common/cmd_sf.c
index aef8c2a5ea84..db84a78e1698 100644
--- a/common/cmd_sf.c
+++ b/common/cmd_sf.c
@@ -223,7 +223,7 @@ static int spi_flash_update(struct spi_flash *flash, u32 offset,
 
 	if (end - buf >= 200)
 		scale = (end - buf) / 100;
-	cmp_buf = malloc(flash->sector_size);
+	cmp_buf = memalign(ARCH_DMA_MINALIGN, flash->sector_size);
 	if (cmp_buf) {
 		ulong last_update = get_timer(0);
 
@@ -480,12 +480,12 @@ static int do_spi_flash_test(int argc, char * const argv[])
 	if (*argv[2] == 0 || *endp != 0)
 		return -1;
 
-	vbuf = malloc(len);
+	vbuf = memalign(ARCH_DMA_MINALIGN, len);
 	if (!vbuf) {
 		printf("Cannot allocate memory (%lu bytes)\n", len);
 		return 1;
 	}
-	buf = malloc(len);
+	buf = memalign(ARCH_DMA_MINALIGN, len);
 	if (!buf) {
 		free(vbuf);
 		printf("Cannot allocate memory (%lu bytes)\n", len);
-- 
2.4.5

[U-Boot] [PATCH 02/11] env: use cache line aligned memory for flash read

[Vignesh R]

From: Ravi Babu <ravibabu@ti.com>

Use memalign() with ARCH_DMA_MINALIGN to allocate read buffers.
This is required because, flash drivers may use DMA for read operations
and may have to invalidate the buffer before read.

Signed-off-by: Ravi Babu <ravibabu@ti.com>
Signed-off-by: Vignesh R <vigneshr@ti.com>
---
 common/env_sf.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/common/env_sf.c b/common/env_sf.c
index e928f5752cc7..940983124fbb 100644
--- a/common/env_sf.c
+++ b/common/env_sf.c
@@ -79,7 +79,7 @@ int saveenv(void)
 	if (CONFIG_ENV_SECT_SIZE > CONFIG_ENV_SIZE) {
 		saved_size = CONFIG_ENV_SECT_SIZE - CONFIG_ENV_SIZE;
 		saved_offset = env_new_offset + CONFIG_ENV_SIZE;
-		saved_buffer = malloc(saved_size);
+		saved_buffer = memalign(ARCH_DMA_MINALIGN, saved_size);
 		if (!saved_buffer) {
 			ret = 1;
 			goto done;
@@ -142,9 +142,10 @@ void env_relocate_spec(void)
 	env_t *tmp_env2 = NULL;
 	env_t *ep = NULL;
 
-	tmp_env1 = (env_t *)malloc(CONFIG_ENV_SIZE);
-	tmp_env2 = (env_t *)malloc(CONFIG_ENV_SIZE);
-
+	tmp_env1 = (env_t *)memalign(ARCH_DMA_MINALIGN,
+			CONFIG_ENV_SIZE);
+	tmp_env2 = (env_t *)memalign(ARCH_DMA_MINALIGN,
+			CONFIG_ENV_SIZE);
 	if (!tmp_env1 || !tmp_env2) {
 		set_default_env("!malloc() failed");
 		goto out;
@@ -295,7 +296,7 @@ void env_relocate_spec(void)
 	int ret;
 	char *buf = NULL;
 
-	buf = (char *)malloc(CONFIG_ENV_SIZE);
+	buf = (char *)memalign(ARCH_DMA_MINALIGN, CONFIG_ENV_SIZE);
 	env_flash = spi_flash_probe(CONFIG_ENV_SPI_BUS, CONFIG_ENV_SPI_CS,
 			CONFIG_ENV_SPI_MAX_HZ, CONFIG_ENV_SPI_MODE);
 	if (!env_flash) {
-- 
2.4.5

[U-Boot] [PATCH 03/11] ARM: AM43xx: Add support for disabling clocks in uboot

[Vignesh R]

From: Kishon Vijay Abraham I <kishon@ti.com>

Add do_disable_clocks() to disable clock domains and module clocks.
These clocks are enabled using do_enable_clocks().

Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
Signed-off-by: Vignesh R <vigneshr@ti.com>
---
 arch/arm/cpu/armv7/am33xx/clock.c        | 52 ++++++++++++++++++++++++++++++++
 arch/arm/include/asm/arch-am33xx/clock.h |  1 +
 2 files changed, 53 insertions(+)

diff --git a/arch/arm/cpu/armv7/am33xx/clock.c b/arch/arm/cpu/armv7/am33xx/clock.c
index ec7d46838b74..595c951ed245 100644
--- a/arch/arm/cpu/armv7/am33xx/clock.c
+++ b/arch/arm/cpu/armv7/am33xx/clock.c
@@ -144,6 +144,33 @@ static inline void enable_clock_module(u32 *const clkctrl_addr, u32 enable_mode,
 		wait_for_clk_enable(clkctrl_addr);
 }
 
+static inline void wait_for_clk_disable(u32 *clkctrl_addr)
+{
+	u32 clkctrl, idlest = MODULE_CLKCTRL_IDLEST_FULLY_FUNCTIONAL;
+	u32 bound = LDELAY;
+
+	while ((idlest != MODULE_CLKCTRL_IDLEST_DISABLED)) {
+		clkctrl = readl(clkctrl_addr);
+		idlest = (clkctrl & MODULE_CLKCTRL_IDLEST_MASK) >>
+			  MODULE_CLKCTRL_IDLEST_SHIFT;
+		if (--bound == 0) {
+			printf("Clock disable failed for 0x%p idlest 0x%x\n",
+			       clkctrl_addr, clkctrl);
+			 return;
+		}
+	}
+}
+static inline void disable_clock_module(u32 *const clkctrl_addr,
+					u32 wait_for_disable)
+{
+	clrsetbits_le32(clkctrl_addr, MODULE_CLKCTRL_MODULEMODE_MASK,
+			MODULE_CLKCTRL_MODULEMODE_SW_DISABLE <<
+			MODULE_CLKCTRL_MODULEMODE_SHIFT);
+	debug("Disable clock module - %p\n", clkctrl_addr);
+	if (wait_for_disable)
+		wait_for_clk_disable(clkctrl_addr);
+}
+
 static inline void enable_clock_domain(u32 *const clkctrl_reg, u32 enable_mode)
 {
 	clrsetbits_le32(clkctrl_reg, CD_CLKCTRL_CLKTRCTRL_MASK,
@@ -151,6 +178,14 @@ static inline void enable_clock_domain(u32 *const clkctrl_reg, u32 enable_mode)
 	debug("Enable clock domain - %p\n", clkctrl_reg);
 }
 
+static inline void disable_clock_domain(u32 *const clkctrl_reg)
+{
+	clrsetbits_le32(clkctrl_reg, CD_CLKCTRL_CLKTRCTRL_MASK,
+			CD_CLKCTRL_CLKTRCTRL_SW_SLEEP <<
+			CD_CLKCTRL_CLKTRCTRL_SHIFT);
+	debug("Disable clock domain - %p\n", clkctrl_reg);
+}
+
 void do_enable_clocks(u32 *const *clk_domains,
 		      u32 *const *clk_modules_explicit_en, u8 wait_for_enable)
 {
@@ -170,6 +205,23 @@ void do_enable_clocks(u32 *const *clk_domains,
 	};
 }
 
+void do_disable_clocks(u32 *const *clk_domains,
+			u32 *const *clk_modules_disable,
+			u8 wait_for_disable)
+{
+	u32 i, max = 100;
+
+
+	/* Clock modules that need to be put in SW_DISABLE */
+	for (i = 0; (i < max) && clk_modules_disable[i]; i++)
+		disable_clock_module(clk_modules_disable[i],
+				     wait_for_disable);
+
+	/* Put the clock domains in SW_SLEEP mode */
+	for (i = 0; (i < max) && clk_domains[i]; i++)
+		disable_clock_domain(clk_domains[i]);
+}
+
 /*
  * Before scaling up the clocks we need to have the PMIC scale up the
  * voltages first.  This will be dependent on which PMIC is in use
diff --git a/arch/arm/include/asm/arch-am33xx/clock.h b/arch/arm/include/asm/arch-am33xx/clock.h
index 4af6b57e42f5..a6d2419fb843 100644
--- a/arch/arm/include/asm/arch-am33xx/clock.h
+++ b/arch/arm/include/asm/arch-am33xx/clock.h
@@ -112,5 +112,6 @@ void do_setup_dpll(const struct dpll_regs *, const struct dpll_params *);
 void prcm_init(void);
 void enable_basic_clocks(void);
 void do_enable_clocks(u32 *const *, u32 *const *, u8);
+void do_disable_clocks(u32 *const *, u32 *const *, u8);
 
 #endif
-- 
2.4.5

[U-Boot] [PATCH 04/11] ARM: OMAP5: Add support for disabling clocks in uboot

[Vignesh R]

From: Kishon Vijay Abraham I <kishon@ti.com>

Add do_disable_clocks() to disable clock domains and module clocks.
These clocks are enabled using do_enable_clocks().

Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
Signed-off-by: Vignesh R <vigneshr@ti.com>
---
 arch/arm/cpu/armv7/omap-common/clocks-common.c | 53 ++++++++++++++++++++++++++
 arch/arm/include/asm/omap_common.h             |  4 ++
 2 files changed, 57 insertions(+)

diff --git a/arch/arm/cpu/armv7/omap-common/clocks-common.c b/arch/arm/cpu/armv7/omap-common/clocks-common.c
index c94a80781931..e28b79568d1d 100644
--- a/arch/arm/cpu/armv7/omap-common/clocks-common.c
+++ b/arch/arm/cpu/armv7/omap-common/clocks-common.c
@@ -648,6 +648,14 @@ static inline void enable_clock_domain(u32 const clkctrl_reg, u32 enable_mode)
 	debug("Enable clock domain - %x\n", clkctrl_reg);
 }
 
+static inline void disable_clock_domain(u32 const clkctrl_reg)
+{
+	clrsetbits_le32(clkctrl_reg, CD_CLKCTRL_CLKTRCTRL_MASK,
+			CD_CLKCTRL_CLKTRCTRL_SW_SLEEP <<
+			CD_CLKCTRL_CLKTRCTRL_SHIFT);
+	debug("Disable clock domain - %x\n", clkctrl_reg);
+}
+
 static inline void wait_for_clk_enable(u32 clkctrl_addr)
 {
 	u32 clkctrl, idlest = MODULE_CLKCTRL_IDLEST_DISABLED;
@@ -677,6 +685,34 @@ static inline void enable_clock_module(u32 const clkctrl_addr, u32 enable_mode,
 		wait_for_clk_enable(clkctrl_addr);
 }
 
+static inline void wait_for_clk_disable(u32 clkctrl_addr)
+{
+	u32 clkctrl, idlest = MODULE_CLKCTRL_IDLEST_FULLY_FUNCTIONAL;
+	u32 bound = LDELAY;
+
+	while ((idlest != MODULE_CLKCTRL_IDLEST_DISABLED)) {
+		clkctrl = readl(clkctrl_addr);
+		idlest = (clkctrl & MODULE_CLKCTRL_IDLEST_MASK) >>
+			 MODULE_CLKCTRL_IDLEST_SHIFT;
+		if (--bound == 0) {
+			printf("Clock disable failed for 0x%x idlest 0x%x\n",
+			       clkctrl_addr, clkctrl);
+			return;
+		}
+	}
+}
+
+static inline void disable_clock_module(u32 const clkctrl_addr,
+					u32 wait_for_disable)
+{
+	clrsetbits_le32(clkctrl_addr, MODULE_CLKCTRL_MODULEMODE_MASK,
+			MODULE_CLKCTRL_MODULEMODE_SW_DISABLE <<
+			MODULE_CLKCTRL_MODULEMODE_SHIFT);
+	debug("Disable clock module - %x\n", clkctrl_addr);
+	if (wait_for_disable)
+		wait_for_clk_disable(clkctrl_addr);
+}
+
 void freq_update_core(void)
 {
 	u32 freq_config1 = 0;
@@ -800,6 +836,23 @@ void do_enable_clocks(u32 const *clk_domains,
 	}
 }
 
+void do_disable_clocks(u32 const *clk_domains,
+			    u32 const *clk_modules_disable,
+			    u8 wait_for_disable)
+{
+	u32 i, max = 100;
+
+
+	/* Clock modules that need to be put in SW_DISABLE */
+	for (i = 0; (i < max) && clk_modules_disable[i]; i++)
+		disable_clock_module(clk_modules_disable[i],
+				     wait_for_disable);
+
+	/* Put the clock domains in SW_SLEEP mode */
+	for (i = 0; (i < max) && clk_domains[i]; i++)
+		disable_clock_domain(clk_domains[i]);
+}
+
 void prcm_init(void)
 {
 	switch (omap_hw_init_context()) {
diff --git a/arch/arm/include/asm/omap_common.h b/arch/arm/include/asm/omap_common.h
index 5469435cc756..30e899ad46e3 100644
--- a/arch/arm/include/asm/omap_common.h
+++ b/arch/arm/include/asm/omap_common.h
@@ -575,6 +575,10 @@ void do_enable_clocks(u32 const *clk_domains,
 		      u32 const *clk_modules_explicit_en,
 		      u8 wait_for_enable);
 
+void do_disable_clocks(u32 const *clk_domains,
+		       u32 const *clk_modules_disable,
+		       u8 wait_for_disable);
+
 void setup_post_dividers(u32 const base,
 			const struct dpll_params *params);
 u32 omap_ddr_clk(void);
-- 
2.4.5

[U-Boot] [PATCH 05/11] ARM: OMAP5: Add functions to enable and disable EDMA3 clocks

[Vignesh R]

Adds functions to enable and disable edma3 clocks which can be invoked
by drivers using edma3 to control the clocks.

Signed-off-by: Vignesh R <vigneshr@ti.com>
---
 arch/arm/cpu/armv7/omap5/hw_data.c   | 41 ++++++++++++++++++++++++++++++++++++
 arch/arm/cpu/armv7/omap5/prcm-regs.c |  4 ++++
 arch/arm/include/asm/omap_common.h   |  9 ++++++++
 3 files changed, 54 insertions(+)

diff --git a/arch/arm/cpu/armv7/omap5/hw_data.c b/arch/arm/cpu/armv7/omap5/hw_data.c
index 3a723cace71a..33f92b7e225d 100644
--- a/arch/arm/cpu/armv7/omap5/hw_data.c
+++ b/arch/arm/cpu/armv7/omap5/hw_data.c
@@ -565,6 +565,47 @@ void enable_basic_uboot_clocks(void)
 			 1);
 }
 
+#ifdef CONFIG_TI_EDMA3
+void enable_edma3_clocks(void)
+{
+	u32 const clk_domains_edma3[] = {
+		0
+	};
+
+	u32 const clk_modules_hw_auto_edma3[] = {
+		(*prcm)->cm_l3main1_tptc1_clkctrl,
+		(*prcm)->cm_l3main1_tptc2_clkctrl,
+		0
+	};
+
+	u32 const clk_modules_explicit_en_edma3[] = {
+		0
+	};
+
+	do_enable_clocks(clk_domains_edma3,
+			 clk_modules_hw_auto_edma3,
+			 clk_modules_explicit_en_edma3,
+			 1);
+}
+
+void disable_edma3_clocks(void)
+{
+	u32 const clk_domains_edma3[] = {
+		0
+	};
+
+	u32 const clk_modules_disable_edma3[] = {
+		(*prcm)->cm_l3main1_tptc1_clkctrl,
+		(*prcm)->cm_l3main1_tptc2_clkctrl,
+		0
+	};
+
+	do_disable_clocks(clk_domains_edma3,
+			  clk_modules_disable_edma3,
+			  1);
+}
+#endif
+
 const struct ctrl_ioregs ioregs_omap5430 = {
 	.ctrl_ddrch = DDR_IO_I_34OHM_SR_FASTEST_WD_DQ_NO_PULL_DQS_PULL_DOWN,
 	.ctrl_lpddr2ch = DDR_IO_I_34OHM_SR_FASTEST_WD_CK_CKE_NCS_CA_PULL_DOWN,
diff --git a/arch/arm/cpu/armv7/omap5/prcm-regs.c b/arch/arm/cpu/armv7/omap5/prcm-regs.c
index cd51fe7678be..d01ce88306ee 100644
--- a/arch/arm/cpu/armv7/omap5/prcm-regs.c
+++ b/arch/arm/cpu/armv7/omap5/prcm-regs.c
@@ -989,4 +989,8 @@ struct prcm_regs const dra7xx_prcm = {
 
 	.prm_abbldo_mpu_setup			= 0x4AE07DDC,
 	.prm_abbldo_mpu_ctrl			= 0x4AE07DE0,
+
+	/*l3main1 edma*/
+	.cm_l3main1_tptc1_clkctrl               = 0x4a008778,
+	.cm_l3main1_tptc2_clkctrl               = 0x4a008780,
 };
diff --git a/arch/arm/include/asm/omap_common.h b/arch/arm/include/asm/omap_common.h
index 30e899ad46e3..1d758fc9050c 100644
--- a/arch/arm/include/asm/omap_common.h
+++ b/arch/arm/include/asm/omap_common.h
@@ -349,6 +349,10 @@ struct prcm_regs {
 	/* IPU */
 	u32 cm_ipu_clkstctrl;
 	u32 cm_ipu_i2c5_clkctrl;
+
+	/*l3main1 edma*/
+	u32 cm_l3main1_tptc1_clkctrl;
+	u32 cm_l3main1_tptc2_clkctrl;
 };
 
 struct omap_sys_ctrl_regs {
@@ -598,6 +602,11 @@ void recalibrate_iodelay(void);
 
 void omap_smc1(u32 service, u32 val);
 
+#ifdef CONFIG_TI_EDMA3
+void enable_edma3_clocks(void);
+void disable_edma3_clocks(void);
+#endif
+
 /* ABB */
 #define OMAP_ABB_NOMINAL_OPP		0
 #define OMAP_ABB_FAST_OPP		1
-- 
2.4.5

[U-Boot] [PATCH 06/11] ARM: AM43XX: Add functions to enable and disable EDMA3 clocks

[Vignesh R]

Adds functions to enable and disable edma3 clocks which can be invoked
by drivers using edma3 to control the clocks.

Signed-off-by: Vignesh R <vigneshr@ti.com>
---
 arch/arm/cpu/armv7/am33xx/clock_am43xx.c | 36 ++++++++++++++++++++++++++++++++
 1 file changed, 36 insertions(+)

diff --git a/arch/arm/cpu/armv7/am33xx/clock_am43xx.c b/arch/arm/cpu/armv7/am33xx/clock_am43xx.c
index b1c0025eebe8..4a410a21952e 100644
--- a/arch/arm/cpu/armv7/am33xx/clock_am43xx.c
+++ b/arch/arm/cpu/armv7/am33xx/clock_am43xx.c
@@ -134,3 +134,39 @@ void enable_basic_clocks(void)
 	/* For OPP100 the mac clock should be /5. */
 	writel(0x4, &cmdpll->clkselmacclk);
 }
+
+#ifdef CONFIG_TI_EDMA3
+void enable_edma3_clocks(void)
+{
+	u32 *const clk_domains_edma3[] = {
+		0
+	};
+
+	u32 *const clk_modules_explicit_en_edma3[] = {
+		&cmper->tpccclkctrl,
+		&cmper->tptc0clkctrl,
+		0
+	};
+
+	do_enable_clocks(clk_domains_edma3,
+			 clk_modules_explicit_en_edma3,
+			 1);
+}
+
+void disable_edma3_clocks(void)
+{
+	u32 *const clk_domains_edma3[] = {
+		0
+	};
+
+	u32 *const clk_modules_disable_edma3[] = {
+		&cmper->tpccclkctrl,
+		&cmper->tptc0clkctrl,
+		0
+	};
+
+	do_disable_clocks(clk_domains_edma3,
+			  clk_modules_disable_edma3,
+			  1);
+}
+#endif
-- 
2.4.5

[U-Boot] [PATCH 07/11] sf: ops: Add spi_flash_copy_mmap function

[Vignesh R]

From: Tom Rini <trini@ti.com>

When doing a memory mapped copy we may have DMA available and thus need
to have this copy abstracted so that the driver can do it, rather than a
simple memcpy.

Signed-off-by: Vignesh R <vigneshr@ti.com>
---
 drivers/mtd/spi/sf_ops.c | 8 +++++++-
 include/spi.h            | 3 +++
 2 files changed, 10 insertions(+), 1 deletion(-)

diff --git a/drivers/mtd/spi/sf_ops.c b/drivers/mtd/spi/sf_ops.c
index 38592f518b72..900ec1f2a9ce 100644
--- a/drivers/mtd/spi/sf_ops.c
+++ b/drivers/mtd/spi/sf_ops.c
@@ -14,6 +14,7 @@
 #include <spi.h>
 #include <spi_flash.h>
 #include <watchdog.h>
+#include <linux/compiler.h>
 
 #include "sf_internal.h"
 
@@ -378,6 +379,11 @@ int spi_flash_read_common(struct spi_flash *flash, const u8 *cmd,
 	return ret;
 }
 
+void __weak spi_flash_copy_mmap(void *data, void *offset, size_t len)
+{
+	memcpy(data, offset, len);
+}
+
 int spi_flash_cmd_read_ops(struct spi_flash *flash, u32 offset,
 		size_t len, void *data)
 {
@@ -394,7 +400,7 @@ int spi_flash_cmd_read_ops(struct spi_flash *flash, u32 offset,
 			return ret;
 		}
 		spi_xfer(flash->spi, 0, NULL, NULL, SPI_XFER_MMAP);
-		memcpy(data, flash->memory_map + offset, len);
+		spi_flash_copy_mmap(data, flash->memory_map + offset, len);
 		spi_xfer(flash->spi, 0, NULL, NULL, SPI_XFER_MMAP_END);
 		spi_release_bus(flash->spi);
 		return 0;
diff --git a/include/spi.h b/include/spi.h
index 18362364cf0e..51fdfd6d7360 100644
--- a/include/spi.h
+++ b/include/spi.h
@@ -272,6 +272,9 @@ int spi_set_wordlen(struct spi_slave *slave, unsigned int wordlen);
 int  spi_xfer(struct spi_slave *slave, unsigned int bitlen, const void *dout,
 		void *din, unsigned long flags);
 
+/* Copy memory mapped data */
+void spi_flash_copy_mmap(void *data, void *offset, size_t len);
+
 /**
  * Determine if a SPI chipselect is valid.
  * This function is provided by the board if the low-level SPI driver
-- 
2.4.5

[U-Boot] [PATCH 08/11] spi: ti_qspi: Use DMA to read from qspi flash

[Vignesh R]

ti_qspi uses memory map mode for faster read. Enabling DMA will increase
read speed by 3x @48MHz on DRA74 EVM.

Signed-off-by: Vignesh R <vigneshr@ti.com>
---
 drivers/spi/ti_qspi.c | 93 +++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 93 insertions(+)

diff --git a/drivers/spi/ti_qspi.c b/drivers/spi/ti_qspi.c
index 3356c0f072e5..0c533e295962 100644
--- a/drivers/spi/ti_qspi.c
+++ b/drivers/spi/ti_qspi.c
@@ -13,6 +13,8 @@
 #include <spi.h>
 #include <asm/gpio.h>
 #include <asm/omap_gpio.h>
+#include <asm/omap_common.h>
+#include <asm/ti-common/ti-edma3.h>
 
 /* ti qpsi register bit masks */
 #define QSPI_TIMEOUT                    2000000
@@ -347,3 +349,94 @@ int spi_xfer(struct spi_slave *slave, unsigned int bitlen, const void *dout,
 
 	return 0;
 }
+#ifdef CONFIG_TI_EDMA3
+void spi_flash_copy_mmap(void *data, void *offset, size_t len)
+{
+	struct edma3_slot_config	slot;
+	struct edma3_channel_config	edma_channel;
+	int				b_cnt_value = 1;
+	int				rem_bytes  = 0;
+	int				a_cnt_value = len;
+	unsigned int			addr = (unsigned int) (data);
+	unsigned int			max_acnt  = 0x7FFFU;
+	unsigned int			edma_slot_num = 1;
+
+	if (len > max_acnt) {
+		b_cnt_value = (len / max_acnt);
+		rem_bytes  = (len % max_acnt);
+		a_cnt_value = max_acnt;
+	}
+
+	/* Invalidate the area, so no writeback into the RAM races with DMA */
+	invalidate_dcache_range(addr, addr + roundup(len, ARCH_DMA_MINALIGN));
+
+	/* enable edma3 clocks */
+	enable_edma3_clocks();
+	/* Compute QSPI address and size */
+	slot.opt	= 0;
+	slot.src	= ((unsigned int) offset);
+	slot.acnt	= a_cnt_value;
+	slot.bcnt	= b_cnt_value;
+	slot.ccnt	= 1;
+	slot.src_bidx	= a_cnt_value;
+	slot.dst_bidx	= a_cnt_value;
+	slot.src_cidx	= 0;
+	slot.dst_cidx	= 0;
+	slot.link	= EDMA3_PARSET_NULL_LINK;
+	slot.bcntrld	= 0;
+	slot.opt	= EDMA3_SLOPT_TRANS_COMP_INT_ENB |
+			  EDMA3_SLOPT_COMP_CODE(0) |
+			  EDMA3_SLOPT_STATIC | EDMA3_SLOPT_AB_SYNC;
+
+	edma3_slot_configure(EDMA3_BASE, edma_slot_num, &slot);
+	edma_channel.slot = edma_slot_num;
+	edma_channel.chnum = 0;
+	edma_channel.complete_code = 0;
+	 /* set event trigger to dst update */
+	edma_channel.trigger_slot_word = EDMA3_TWORD(dst);
+
+	qedma3_start(EDMA3_BASE, &edma_channel);
+	edma3_set_dest_addr(EDMA3_BASE, edma_channel.slot, addr);
+
+	while (edma3_check_for_transfer(EDMA3_BASE, &edma_channel))
+		;
+	qedma3_stop(EDMA3_BASE, &edma_channel);
+
+	if (rem_bytes != 0) {
+		/* Compute QSPI address and size */
+		slot.opt	= 0;
+		slot.src	=
+			(b_cnt_value * max_acnt) + ((unsigned int) offset);
+		slot.acnt	= rem_bytes;
+		slot.bcnt	= 1;
+		slot.ccnt	= 1;
+		slot.src_bidx	= rem_bytes;
+		slot.dst_bidx	= rem_bytes;
+		slot.src_cidx	= 0;
+		slot.dst_cidx	= 0;
+		slot.link	= EDMA3_PARSET_NULL_LINK;
+		slot.bcntrld	= 0;
+		slot.opt	= EDMA3_SLOPT_TRANS_COMP_INT_ENB |
+				  EDMA3_SLOPT_COMP_CODE(0) |
+				  EDMA3_SLOPT_STATIC | EDMA3_SLOPT_AB_SYNC;
+		edma3_slot_configure(EDMA3_BASE, edma_slot_num, &slot);
+		edma_channel.slot = edma_slot_num;
+		edma_channel.chnum = 0;
+		edma_channel.complete_code = 0;
+		/* set event trigger to dst update */
+		edma_channel.trigger_slot_word = EDMA3_TWORD(dst);
+
+		qedma3_start(EDMA3_BASE, &edma_channel);
+		edma3_set_dest_addr(EDMA3_BASE, edma_channel.slot, addr +
+				    (max_acnt * b_cnt_value));
+
+		while (edma3_check_for_transfer(EDMA3_BASE, &edma_channel))
+			;
+		qedma3_stop(EDMA3_BASE, &edma_channel);
+	}
+	*((unsigned int *)offset) += len;
+
+	/* disable edma3 clocks */
+	disable_edma3_clocks();
+}
+#endif
-- 
2.4.5

[U-Boot] [PATCH 09/11] dma: ti-edma3: Add BIT(x) macro definition

[Vignesh R]

Add BIT(x) macro definition to edma3 driver file. Fixes following
compiler warning when CONFIG_TI_EDMA3 is enabled for am437x and dra7x:

drivers/dma/ti-edma3.c: In function ?edma3_set_dest?:
drivers/dma/ti-edma3.c:92:10: warning: implicit declaration of function
?BIT? [-Wimplicit-function-declaration]

Signed-off-by: Vignesh R <vigneshr@ti.com>
---
 drivers/dma/ti-edma3.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/dma/ti-edma3.c b/drivers/dma/ti-edma3.c
index 8184ded9fa81..638cf525a278 100644
--- a/drivers/dma/ti-edma3.c
+++ b/drivers/dma/ti-edma3.c
@@ -13,6 +13,8 @@
 #include <common.h>
 #include <asm/ti-common/ti-edma3.h>
 
+#define BIT(x)					(1 << (x))
+
 #define EDMA3_SL_BASE(slot)			(0x4000 + ((slot) << 5))
 #define EDMA3_SL_MAX_NUM			512
 #define EDMA3_SLOPT_FIFO_WIDTH_MASK		(0x7 << 8)
-- 
2.4.5

[U-Boot] [PATCH 10/11] ARM: dra7xx_evm: Enable EDMA3 in SPL to support DMA on qspi

[Vignesh R]

Enable TI_EDMA3 and SPL_DMA support, so as to reduce boot time. With
DMA enabled there is almost 3x improvement in read performance. This
helps in reducing boot time in qspiboot mode

Also add EDMA3 base address for DRA7XX and AM57XX.

Signed-off-by: Vignesh R <vigneshr@ti.com>
---
 arch/arm/include/asm/arch-omap5/hardware.h | 5 +++++
 include/configs/dra7xx_evm.h               | 2 ++
 2 files changed, 7 insertions(+)

diff --git a/arch/arm/include/asm/arch-omap5/hardware.h b/arch/arm/include/asm/arch-omap5/hardware.h
index f7011b4e904a..a5bd60009289 100644
--- a/arch/arm/include/asm/arch-omap5/hardware.h
+++ b/arch/arm/include/asm/arch-omap5/hardware.h
@@ -23,4 +23,9 @@
 /* GPMC Base address */
 #define GPMC_BASE			0x50000000
 
+/* EDMA3 Base address for DRA7XX and AM57XX */
+#if defined(CONFIG_DRA7XX) || defined(CONFIG_AM57XX)
+#define EDMA3_BASE			0x43300000
+#endif
+
 #endif
diff --git a/include/configs/dra7xx_evm.h b/include/configs/dra7xx_evm.h
index d84427d1d8bc..fe09aac27f71 100644
--- a/include/configs/dra7xx_evm.h
+++ b/include/configs/dra7xx_evm.h
@@ -165,6 +165,8 @@
 
 /* SPI SPL */
 #define CONFIG_SPL_SPI_SUPPORT
+#define CONFIG_SPL_DMA_SUPPORT
+#define CONFIG_TI_EDMA3
 #define CONFIG_SPL_SPI_LOAD
 #define CONFIG_SPL_SPI_FLASH_SUPPORT
 #define CONFIG_SYS_SPI_U_BOOT_OFFS     0x40000
-- 
2.4.5

[U-Boot] [PATCH 11/11] ARM: am43xx_evm: Enable EDMA3 to support DMA on qspi

[Vignesh R]

Enable CONFIG_TI_EDMA3 for am43xx, this increases read performance by 1.5x.

Also add EDMA3 base address for am43xx.

Signed-off-by: Vignesh R <vigneshr@ti.com>
---
 arch/arm/include/asm/arch-am33xx/hardware_am43xx.h | 3 +++
 include/configs/am43xx_evm.h                       | 2 ++
 2 files changed, 5 insertions(+)

diff --git a/arch/arm/include/asm/arch-am33xx/hardware_am43xx.h b/arch/arm/include/asm/arch-am33xx/hardware_am43xx.h
index 479893e47ea1..a7da6b5cfde6 100644
--- a/arch/arm/include/asm/arch-am33xx/hardware_am43xx.h
+++ b/arch/arm/include/asm/arch-am33xx/hardware_am43xx.h
@@ -98,4 +98,7 @@
 
 #define NUM_CRYSTAL_FREQ			0x4
 
+/* EDMA3 Base Address */
+#define EDMA3_BASE				0x49000000
+
 #endif /* __AM43XX_HARDWARE_AM43XX_H */
diff --git a/include/configs/am43xx_evm.h b/include/configs/am43xx_evm.h
index 33e534a76583..5f733ea9905b 100644
--- a/include/configs/am43xx_evm.h
+++ b/include/configs/am43xx_evm.h
@@ -208,6 +208,8 @@
 #define CONFIG_SF_DEFAULT_SPEED                48000000
 #define CONFIG_DEFAULT_SPI_MODE                SPI_MODE_3
 
+#define CONFIG_TI_EDMA3
+
 /* Enhance our eMMC support / experience. */
 #define CONFIG_CMD_GPT
 #define CONFIG_EFI_PARTITION
-- 
2.4.5

[U-Boot] [PATCH 08/11] spi: ti_qspi: Use DMA to read from qspi flash

[Tom Rini]

On Fri, Jul 03, 2015 at 04:46:10PM +0530, Vignesh R wrote:

> ti_qspi uses memory map mode for faster read. Enabling DMA will increase
> read speed by 3x @48MHz on DRA74 EVM.
> 
> Signed-off-by: Vignesh R <vigneshr@ti.com>

This ignores the feedback from
http://lists.denx.de/pipermail/u-boot/2014-July/183715.html where we
need to model the DMA changes on how it's done for mxs_spi.c

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20150703/aba6a985/attachment.sig>

[U-Boot] [PATCH 09/11] dma: ti-edma3: Add BIT(x) macro definition

[Andy Pont]

Vignesh wrote...

[snip]

> +#define BIT(x)					(1 << (x))
> +

Is this not something that would be better in a global header file somewhere rather than it starting a trend of a per-driver, per-arch, etc. definitions?

Andy.

[U-Boot] [PATCH 08/11] spi: ti_qspi: Use DMA to read from qspi flash

[R, Vignesh]

On 7/3/2015 5:12 PM, Tom Rini wrote:
> On Fri, Jul 03, 2015 at 04:46:10PM +0530, Vignesh R wrote:
> 
>> ti_qspi uses memory map mode for faster read. Enabling DMA will increase
>> read speed by 3x @48MHz on DRA74 EVM.
>>
>> Signed-off-by: Vignesh R <vigneshr@ti.com>
> 
> This ignores the feedback from
> http://lists.denx.de/pipermail/u-boot/2014-July/183715.html where we
> need to model the DMA changes on how it's done for mxs_spi.c
> 

Sorry.. I didn't look into that before.
mxs_spi uses peripheral DMA to read/write flash. But ti_qspi can use DMA
to read from flash in mmap mode only. In current u-boot, defining
CONFIG_TI_SPI_MMAP will make memory map address available
(spi_flash->memory_map) to sf layer and spi_flash_cmd_read_ops() (in
sf_ops.c) directly calls memcpy() to read data from flash into buffer.
There is no spi_xfer() call to the ti_qspi driver at all.

In order to implement mxs_spi like approach for ti_qspi.c, I can delete
mmap handling in sf_ops.c( I don't think any other spi driver uses this
part of code), so that spi_xfer() is always called. And then, in
spi_xfer() implementation of ti_qspi, I can do DMA transfer similar to
mxs_spi.c. Is this approach ok?

And are you ok with patch 1 and 2 of this series?

Regards
Vignesh

[U-Boot] [PATCH 09/11] dma: ti-edma3: Add BIT(x) macro definition

[R, Vignesh]

On 7/3/2015 7:27 PM, Andy Pont wrote:
> Vignesh wrote...
> 
> [snip]
> 
>> +#define BIT(x)					(1 << (x))
>> +
> 
> Is this not something that would be better in a global header file somewhere rather than it starting a trend of a per-driver, per-arch, etc. definitions?

I agree there are few per arch defintions.
How about adding it to include/linux/bitops.h as in linux kernel?

Regards
Vignesh

[U-Boot] [PATCH 09/11] dma: ti-edma3: Add BIT(x) macro definition

[Jagan Teki]

On 4 July 2015 at 18:23, R, Vignesh <vigneshr@ti.com> wrote:
>
>
> On 7/3/2015 7:27 PM, Andy Pont wrote:
>> Vignesh wrote...
>>
>> [snip]
>>
>>> +#define BIT(x)                                      (1 << (x))
>>> +
>>
>> Is this not something that would be better in a global header file somewhere rather than it starting a trend of a per-driver, per-arch, etc. definitions?
>
> I agree there are few per arch defintions.
> How about adding it to include/linux/bitops.h as in linux kernel?

Patch is under review [1] in fact you may give comment on the same.

[1] https://patchwork.ozlabs.org/patch/470478/

thanks!
-- 
Jagan | openedev.

[U-Boot] [PATCH 08/11] spi: ti_qspi: Use DMA to read from qspi flash

[Vignesh R]

On Saturday 04 July 2015 06:23 PM, R, Vignesh wrote:
> 
> 
> On 7/3/2015 5:12 PM, Tom Rini wrote:
>> On Fri, Jul 03, 2015 at 04:46:10PM +0530, Vignesh R wrote:
>>
>>> ti_qspi uses memory map mode for faster read. Enabling DMA will increase
>>> read speed by 3x @48MHz on DRA74 EVM.
>>>
>>> Signed-off-by: Vignesh R <vigneshr@ti.com>
>>
>> This ignores the feedback from
>> http://lists.denx.de/pipermail/u-boot/2014-July/183715.html where we
>> need to model the DMA changes on how it's done for mxs_spi.c
>>
> 
> Sorry.. I didn't look into that before.
> mxs_spi uses peripheral DMA to read/write flash. But ti_qspi can use DMA
> to read from flash in mmap mode only. In current u-boot, defining
> CONFIG_TI_SPI_MMAP will make memory map address available
> (spi_flash->memory_map) to sf layer and spi_flash_cmd_read_ops() (in
> sf_ops.c) directly calls memcpy() to read data from flash into buffer.
> There is no spi_xfer() call to the ti_qspi driver at all.
> 
> In order to implement mxs_spi like approach for ti_qspi.c, I can delete
> mmap handling in sf_ops.c( I don't think any other spi driver uses this
> part of code), so that spi_xfer() is always called. And then, in
> spi_xfer() implementation of ti_qspi, I can do DMA transfer similar to
> mxs_spi.c. Is this approach ok?
> 

I think I misinterpreted the thread previously. The suggestion is to
move DMA initialization related code from to ti-edma3.c and use
spi_flash_copy_mmap() just to pass addresses to ti-edma3 apis. Am I correct?

-- 
Regards
Vignesh

[U-Boot] [PATCH 08/11] spi: ti_qspi: Use DMA to read from qspi flash

[Vignesh R]

On 07/03/2015 05:12 PM, Tom Rini wrote:
> On Fri, Jul 03, 2015 at 04:46:10PM +0530, Vignesh R wrote:
> 
>> ti_qspi uses memory map mode for faster read. Enabling DMA will increase
>> read speed by 3x @48MHz on DRA74 EVM.
>>
>> Signed-off-by: Vignesh R <vigneshr@ti.com>
> 
> This ignores the feedback from
> http://lists.denx.de/pipermail/u-boot/2014-July/183715.html where we
> need to model the DMA changes on how it's done for mxs_spi.c
> 
Is the following patch an acceptable solution?

8<-------------------------------------------------------------------

Move DMA related initialization code to helper function in ti-edma3
driver. Use this function for scheduling DMA transfer from ti_qspi driver.

diff --git a/arch/arm/include/asm/ti-common/ti-edma3.h
b/arch/arm/include/asm/ti-common/ti-edma3.h
index 5adc1dac0e65..6a7a321c1bdf 100644
--- a/arch/arm/include/asm/ti-common/ti-edma3.h
+++ b/arch/arm/include/asm/ti-common/ti-edma3.h
@@ -117,5 +117,7 @@ void edma3_set_src_addr(u32 base, int slot, u32 src);
 void edma3_set_transfer_params(u32 base, int slot, int acnt,
 			       int bcnt, int ccnt, u16 bcnt_rld,
 			       enum edma3_sync_dimension sync_mode);
+void edma3_transfer(unsigned long edma3_base_addr, unsigned int
+		edma_slot_num, void *dst, void *src, size_t len);

 #endif
diff --git a/drivers/dma/ti-edma3.c b/drivers/dma/ti-edma3.c
index 8184ded9fa81..d6a427f2e21d 100644
--- a/drivers/dma/ti-edma3.c
+++ b/drivers/dma/ti-edma3.c
@@ -382,3 +382,81 @@ void qedma3_stop(u32 base, struct
edma3_channel_config *cfg)
 	/* Clear the channel map */
 	__raw_writel(0, base + EDMA3_QCHMAP(cfg->chnum));
 }
+
+void edma3_transfer(unsigned long edma3_base_addr, unsigned int
+		    edma_slot_num, void *dst, void *src, size_t len)
+{
+	struct edma3_slot_config        slot;
+	struct edma3_channel_config     edma_channel;
+	int                             b_cnt_value = 1;
+	int                             rem_bytes  = 0;
+	int                             a_cnt_value = len;
+	unsigned int                    addr = (unsigned int) (dst);
+	unsigned int                    max_acnt  = 0x7FFFU;
+
+	if (len > max_acnt) {
+		b_cnt_value = (len / max_acnt);
+		rem_bytes  = (len % max_acnt);
+		a_cnt_value = max_acnt;
+	}
+
+	slot.opt        = 0;
+	slot.src        = ((unsigned int) src);
+	slot.acnt       = a_cnt_value;
+	slot.bcnt       = b_cnt_value;
+	slot.ccnt       = 1;
+	slot.src_bidx   = a_cnt_value;
+	slot.dst_bidx   = a_cnt_value;
+	slot.src_cidx   = 0;
+	slot.dst_cidx   = 0;
+	slot.link       = EDMA3_PARSET_NULL_LINK;
+	slot.bcntrld    = 0;
+	slot.opt        = EDMA3_SLOPT_TRANS_COMP_INT_ENB |
+			  EDMA3_SLOPT_COMP_CODE(0) |
+			  EDMA3_SLOPT_STATIC | EDMA3_SLOPT_AB_SYNC;
+
+	edma3_slot_configure(edma3_base_addr, edma_slot_num, &slot);
+	edma_channel.slot = edma_slot_num;
+	edma_channel.chnum = 0;
+	edma_channel.complete_code = 0;
+	 /* set event trigger to dst update */
+	edma_channel.trigger_slot_word = EDMA3_TWORD(dst);
+
+	qedma3_start(edma3_base_addr, &edma_channel);
+	edma3_set_dest_addr(edma3_base_addr, edma_channel.slot, addr);
+
+	while (edma3_check_for_transfer(edma3_base_addr, &edma_channel))
+		;
+	qedma3_stop(edma3_base_addr, &edma_channel);
+
+	if (rem_bytes != 0) {
+		slot.opt        = 0;
+		slot.src        =
+			(b_cnt_value * max_acnt) + ((unsigned int) src);
+		slot.acnt       = rem_bytes;
+		slot.bcnt       = 1;
+		slot.ccnt       = 1;
+		slot.src_bidx   = rem_bytes;
+		slot.dst_bidx   = rem_bytes;
+		slot.src_cidx   = 0;
+		slot.dst_cidx   = 0;
+		slot.link       = EDMA3_PARSET_NULL_LINK;
+		slot.bcntrld    = 0;
+		slot.opt        = EDMA3_SLOPT_TRANS_COMP_INT_ENB |
+				  EDMA3_SLOPT_COMP_CODE(0) |
+				  EDMA3_SLOPT_STATIC | EDMA3_SLOPT_AB_SYNC;
+		edma3_slot_configure(edma3_base_addr, edma_slot_num, &slot);
+		edma_channel.slot = edma_slot_num;
+		edma_channel.chnum = 0;
+		edma_channel.complete_code = 0;
+		/* set event trigger to dst update */
+		edma_channel.trigger_slot_word = EDMA3_TWORD(dst);
+
+		qedma3_start(edma3_base_addr, &edma_channel);
+		edma3_set_dest_addr(edma3_base_addr, edma_channel.slot, addr +
+				    (max_acnt * b_cnt_value));
+		while (edma3_check_for_transfer(edma3_base_addr, &edma_channel))
+			;
+		qedma3_stop(edma3_base_addr, &edma_channel);
+	}
+}
diff --git a/drivers/spi/ti_qspi.c b/drivers/spi/ti_qspi.c
index 3356c0f072e5..753d68980bd6 100644
--- a/drivers/spi/ti_qspi.c
+++ b/drivers/spi/ti_qspi.c
@@ -13,6 +13,8 @@
 #include <spi.h>
 #include <asm/gpio.h>
 #include <asm/omap_gpio.h>
+#include <asm/omap_common.h>
+#include <asm/ti-common/ti-edma3.h>

 /* ti qpsi register bit masks */
 #define QSPI_TIMEOUT                    2000000
@@ -347,3 +349,24 @@ int spi_xfer(struct spi_slave *slave, unsigned int
bitlen, const void *dout,

 	return 0;
 }
+#ifdef CONFIG_TI_EDMA3
+void spi_flash_copy_mmap(void *data, void *offset, size_t len)
+{
+	unsigned int			addr = (unsigned int) (data);
+	unsigned int			edma_slot_num = 1;
+
+	/* Invalidate the area, so no writeback into the RAM races with DMA */
+	invalidate_dcache_range(addr, addr + roundup(len, ARCH_DMA_MINALIGN));
+
+	/* enable edma3 clocks */
+	enable_edma3_clocks();
+
+	/* Call edma3 api to do actual DMA transfer	*/
+	edma3_transfer(EDMA3_BASE, edma_slot_num, data, offset, len);
+
+	/* disable edma3 clocks */
+	disable_edma3_clocks();
+
+	*((unsigned int *)offset) += len;
+}
+#endif


-- 
Regards
Vignesh

[U-Boot] [PATCH 08/11] spi: ti_qspi: Use DMA to read from qspi flash

[Tom Rini]

On Thu, Jul 09, 2015 at 12:10:03PM +0530, Vignesh R wrote:
> 
> 
> On 07/03/2015 05:12 PM, Tom Rini wrote:
> > On Fri, Jul 03, 2015 at 04:46:10PM +0530, Vignesh R wrote:
> > 
> >> ti_qspi uses memory map mode for faster read. Enabling DMA will increase
> >> read speed by 3x @48MHz on DRA74 EVM.
> >>
> >> Signed-off-by: Vignesh R <vigneshr@ti.com>
> > 
> > This ignores the feedback from
> > http://lists.denx.de/pipermail/u-boot/2014-July/183715.html where we
> > need to model the DMA changes on how it's done for mxs_spi.c
> > 
> Is the following patch an acceptable solution?
> 

Jagan, are you OK with the SPI side of this?  Thanks!

> 8<-------------------------------------------------------------------
> 
> Move DMA related initialization code to helper function in ti-edma3
> driver. Use this function for scheduling DMA transfer from ti_qspi driver.
> 
> diff --git a/arch/arm/include/asm/ti-common/ti-edma3.h
> b/arch/arm/include/asm/ti-common/ti-edma3.h
> index 5adc1dac0e65..6a7a321c1bdf 100644
> --- a/arch/arm/include/asm/ti-common/ti-edma3.h
> +++ b/arch/arm/include/asm/ti-common/ti-edma3.h
> @@ -117,5 +117,7 @@ void edma3_set_src_addr(u32 base, int slot, u32 src);
>  void edma3_set_transfer_params(u32 base, int slot, int acnt,
>  			       int bcnt, int ccnt, u16 bcnt_rld,
>  			       enum edma3_sync_dimension sync_mode);
> +void edma3_transfer(unsigned long edma3_base_addr, unsigned int
> +		edma_slot_num, void *dst, void *src, size_t len);
> 
>  #endif
> diff --git a/drivers/dma/ti-edma3.c b/drivers/dma/ti-edma3.c
> index 8184ded9fa81..d6a427f2e21d 100644
> --- a/drivers/dma/ti-edma3.c
> +++ b/drivers/dma/ti-edma3.c
> @@ -382,3 +382,81 @@ void qedma3_stop(u32 base, struct
> edma3_channel_config *cfg)
>  	/* Clear the channel map */
>  	__raw_writel(0, base + EDMA3_QCHMAP(cfg->chnum));
>  }
> +
> +void edma3_transfer(unsigned long edma3_base_addr, unsigned int
> +		    edma_slot_num, void *dst, void *src, size_t len)
> +{
> +	struct edma3_slot_config        slot;
> +	struct edma3_channel_config     edma_channel;
> +	int                             b_cnt_value = 1;
> +	int                             rem_bytes  = 0;
> +	int                             a_cnt_value = len;
> +	unsigned int                    addr = (unsigned int) (dst);
> +	unsigned int                    max_acnt  = 0x7FFFU;
> +
> +	if (len > max_acnt) {
> +		b_cnt_value = (len / max_acnt);
> +		rem_bytes  = (len % max_acnt);
> +		a_cnt_value = max_acnt;
> +	}
> +
> +	slot.opt        = 0;
> +	slot.src        = ((unsigned int) src);
> +	slot.acnt       = a_cnt_value;
> +	slot.bcnt       = b_cnt_value;
> +	slot.ccnt       = 1;
> +	slot.src_bidx   = a_cnt_value;
> +	slot.dst_bidx   = a_cnt_value;
> +	slot.src_cidx   = 0;
> +	slot.dst_cidx   = 0;
> +	slot.link       = EDMA3_PARSET_NULL_LINK;
> +	slot.bcntrld    = 0;
> +	slot.opt        = EDMA3_SLOPT_TRANS_COMP_INT_ENB |
> +			  EDMA3_SLOPT_COMP_CODE(0) |
> +			  EDMA3_SLOPT_STATIC | EDMA3_SLOPT_AB_SYNC;
> +
> +	edma3_slot_configure(edma3_base_addr, edma_slot_num, &slot);
> +	edma_channel.slot = edma_slot_num;
> +	edma_channel.chnum = 0;
> +	edma_channel.complete_code = 0;
> +	 /* set event trigger to dst update */
> +	edma_channel.trigger_slot_word = EDMA3_TWORD(dst);
> +
> +	qedma3_start(edma3_base_addr, &edma_channel);
> +	edma3_set_dest_addr(edma3_base_addr, edma_channel.slot, addr);
> +
> +	while (edma3_check_for_transfer(edma3_base_addr, &edma_channel))
> +		;
> +	qedma3_stop(edma3_base_addr, &edma_channel);
> +
> +	if (rem_bytes != 0) {
> +		slot.opt        = 0;
> +		slot.src        =
> +			(b_cnt_value * max_acnt) + ((unsigned int) src);
> +		slot.acnt       = rem_bytes;
> +		slot.bcnt       = 1;
> +		slot.ccnt       = 1;
> +		slot.src_bidx   = rem_bytes;
> +		slot.dst_bidx   = rem_bytes;
> +		slot.src_cidx   = 0;
> +		slot.dst_cidx   = 0;
> +		slot.link       = EDMA3_PARSET_NULL_LINK;
> +		slot.bcntrld    = 0;
> +		slot.opt        = EDMA3_SLOPT_TRANS_COMP_INT_ENB |
> +				  EDMA3_SLOPT_COMP_CODE(0) |
> +				  EDMA3_SLOPT_STATIC | EDMA3_SLOPT_AB_SYNC;
> +		edma3_slot_configure(edma3_base_addr, edma_slot_num, &slot);
> +		edma_channel.slot = edma_slot_num;
> +		edma_channel.chnum = 0;
> +		edma_channel.complete_code = 0;
> +		/* set event trigger to dst update */
> +		edma_channel.trigger_slot_word = EDMA3_TWORD(dst);
> +
> +		qedma3_start(edma3_base_addr, &edma_channel);
> +		edma3_set_dest_addr(edma3_base_addr, edma_channel.slot, addr +
> +				    (max_acnt * b_cnt_value));
> +		while (edma3_check_for_transfer(edma3_base_addr, &edma_channel))
> +			;
> +		qedma3_stop(edma3_base_addr, &edma_channel);
> +	}
> +}
> diff --git a/drivers/spi/ti_qspi.c b/drivers/spi/ti_qspi.c
> index 3356c0f072e5..753d68980bd6 100644
> --- a/drivers/spi/ti_qspi.c
> +++ b/drivers/spi/ti_qspi.c
> @@ -13,6 +13,8 @@
>  #include <spi.h>
>  #include <asm/gpio.h>
>  #include <asm/omap_gpio.h>
> +#include <asm/omap_common.h>
> +#include <asm/ti-common/ti-edma3.h>
> 
>  /* ti qpsi register bit masks */
>  #define QSPI_TIMEOUT                    2000000
> @@ -347,3 +349,24 @@ int spi_xfer(struct spi_slave *slave, unsigned int
> bitlen, const void *dout,
> 
>  	return 0;
>  }
> +#ifdef CONFIG_TI_EDMA3
> +void spi_flash_copy_mmap(void *data, void *offset, size_t len)
> +{
> +	unsigned int			addr = (unsigned int) (data);
> +	unsigned int			edma_slot_num = 1;
> +
> +	/* Invalidate the area, so no writeback into the RAM races with DMA */
> +	invalidate_dcache_range(addr, addr + roundup(len, ARCH_DMA_MINALIGN));
> +
> +	/* enable edma3 clocks */
> +	enable_edma3_clocks();
> +
> +	/* Call edma3 api to do actual DMA transfer	*/
> +	edma3_transfer(EDMA3_BASE, edma_slot_num, data, offset, len);
> +
> +	/* disable edma3 clocks */
> +	disable_edma3_clocks();
> +
> +	*((unsigned int *)offset) += len;
> +}
> +#endif
> 
> 
> -- 
> Regards
> Vignesh

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20150714/3cabfe00/attachment.sig>

[U-Boot] [PATCH 08/11] spi: ti_qspi: Use DMA to read from qspi flash

[R, Vignesh]

On 7/15/2015 12:32 AM, Tom Rini wrote:
> On Thu, Jul 09, 2015 at 12:10:03PM +0530, Vignesh R wrote:
>>
>>
>> On 07/03/2015 05:12 PM, Tom Rini wrote:
>>> On Fri, Jul 03, 2015 at 04:46:10PM +0530, Vignesh R wrote:
>>>
>>>> ti_qspi uses memory map mode for faster read. Enabling DMA will increase
>>>> read speed by 3x @48MHz on DRA74 EVM.
>>>>
>>>> Signed-off-by: Vignesh R <vigneshr@ti.com>
>>>
>>> This ignores the feedback from
>>> http://lists.denx.de/pipermail/u-boot/2014-July/183715.html where we
>>> need to model the DMA changes on how it's done for mxs_spi.c
>>>
>> Is the following patch an acceptable solution?
>>
> 
> Jagan, are you OK with the SPI side of this?  Thanks!

Gentle ping... Any comments? I will send a v2 for this series if the
below patch is acceptable.

> 
>> 8<-------------------------------------------------------------------
>>
>> Move DMA related initialization code to helper function in ti-edma3
>> driver. Use this function for scheduling DMA transfer from ti_qspi driver.
>>
>> diff --git a/arch/arm/include/asm/ti-common/ti-edma3.h
>> b/arch/arm/include/asm/ti-common/ti-edma3.h
>> index 5adc1dac0e65..6a7a321c1bdf 100644
>> --- a/arch/arm/include/asm/ti-common/ti-edma3.h
>> +++ b/arch/arm/include/asm/ti-common/ti-edma3.h
>> @@ -117,5 +117,7 @@ void edma3_set_src_addr(u32 base, int slot, u32 src);
>>  void edma3_set_transfer_params(u32 base, int slot, int acnt,
>>  			       int bcnt, int ccnt, u16 bcnt_rld,
>>  			       enum edma3_sync_dimension sync_mode);
>> +void edma3_transfer(unsigned long edma3_base_addr, unsigned int
>> +		edma_slot_num, void *dst, void *src, size_t len);
>>
>>  #endif
>> diff --git a/drivers/dma/ti-edma3.c b/drivers/dma/ti-edma3.c
>> index 8184ded9fa81..d6a427f2e21d 100644
>> --- a/drivers/dma/ti-edma3.c
>> +++ b/drivers/dma/ti-edma3.c
>> @@ -382,3 +382,81 @@ void qedma3_stop(u32 base, struct
>> edma3_channel_config *cfg)
>>  	/* Clear the channel map */
>>  	__raw_writel(0, base + EDMA3_QCHMAP(cfg->chnum));
>>  }
>> +
>> +void edma3_transfer(unsigned long edma3_base_addr, unsigned int
>> +		    edma_slot_num, void *dst, void *src, size_t len)
>> +{
>> +	struct edma3_slot_config        slot;
>> +	struct edma3_channel_config     edma_channel;
>> +	int                             b_cnt_value = 1;
>> +	int                             rem_bytes  = 0;
>> +	int                             a_cnt_value = len;
>> +	unsigned int                    addr = (unsigned int) (dst);
>> +	unsigned int                    max_acnt  = 0x7FFFU;
>> +
>> +	if (len > max_acnt) {
>> +		b_cnt_value = (len / max_acnt);
>> +		rem_bytes  = (len % max_acnt);
>> +		a_cnt_value = max_acnt;
>> +	}
>> +
>> +	slot.opt        = 0;
>> +	slot.src        = ((unsigned int) src);
>> +	slot.acnt       = a_cnt_value;
>> +	slot.bcnt       = b_cnt_value;
>> +	slot.ccnt       = 1;
>> +	slot.src_bidx   = a_cnt_value;
>> +	slot.dst_bidx   = a_cnt_value;
>> +	slot.src_cidx   = 0;
>> +	slot.dst_cidx   = 0;
>> +	slot.link       = EDMA3_PARSET_NULL_LINK;
>> +	slot.bcntrld    = 0;
>> +	slot.opt        = EDMA3_SLOPT_TRANS_COMP_INT_ENB |
>> +			  EDMA3_SLOPT_COMP_CODE(0) |
>> +			  EDMA3_SLOPT_STATIC | EDMA3_SLOPT_AB_SYNC;
>> +
>> +	edma3_slot_configure(edma3_base_addr, edma_slot_num, &slot);
>> +	edma_channel.slot = edma_slot_num;
>> +	edma_channel.chnum = 0;
>> +	edma_channel.complete_code = 0;
>> +	 /* set event trigger to dst update */
>> +	edma_channel.trigger_slot_word = EDMA3_TWORD(dst);
>> +
>> +	qedma3_start(edma3_base_addr, &edma_channel);
>> +	edma3_set_dest_addr(edma3_base_addr, edma_channel.slot, addr);
>> +
>> +	while (edma3_check_for_transfer(edma3_base_addr, &edma_channel))
>> +		;
>> +	qedma3_stop(edma3_base_addr, &edma_channel);
>> +
>> +	if (rem_bytes != 0) {
>> +		slot.opt        = 0;
>> +		slot.src        =
>> +			(b_cnt_value * max_acnt) + ((unsigned int) src);
>> +		slot.acnt       = rem_bytes;
>> +		slot.bcnt       = 1;
>> +		slot.ccnt       = 1;
>> +		slot.src_bidx   = rem_bytes;
>> +		slot.dst_bidx   = rem_bytes;
>> +		slot.src_cidx   = 0;
>> +		slot.dst_cidx   = 0;
>> +		slot.link       = EDMA3_PARSET_NULL_LINK;
>> +		slot.bcntrld    = 0;
>> +		slot.opt        = EDMA3_SLOPT_TRANS_COMP_INT_ENB |
>> +				  EDMA3_SLOPT_COMP_CODE(0) |
>> +				  EDMA3_SLOPT_STATIC | EDMA3_SLOPT_AB_SYNC;
>> +		edma3_slot_configure(edma3_base_addr, edma_slot_num, &slot);
>> +		edma_channel.slot = edma_slot_num;
>> +		edma_channel.chnum = 0;
>> +		edma_channel.complete_code = 0;
>> +		/* set event trigger to dst update */
>> +		edma_channel.trigger_slot_word = EDMA3_TWORD(dst);
>> +
>> +		qedma3_start(edma3_base_addr, &edma_channel);
>> +		edma3_set_dest_addr(edma3_base_addr, edma_channel.slot, addr +
>> +				    (max_acnt * b_cnt_value));
>> +		while (edma3_check_for_transfer(edma3_base_addr, &edma_channel))
>> +			;
>> +		qedma3_stop(edma3_base_addr, &edma_channel);
>> +	}
>> +}
>> diff --git a/drivers/spi/ti_qspi.c b/drivers/spi/ti_qspi.c
>> index 3356c0f072e5..753d68980bd6 100644
>> --- a/drivers/spi/ti_qspi.c
>> +++ b/drivers/spi/ti_qspi.c
>> @@ -13,6 +13,8 @@
>>  #include <spi.h>
>>  #include <asm/gpio.h>
>>  #include <asm/omap_gpio.h>
>> +#include <asm/omap_common.h>
>> +#include <asm/ti-common/ti-edma3.h>
>>
>>  /* ti qpsi register bit masks */
>>  #define QSPI_TIMEOUT                    2000000
>> @@ -347,3 +349,24 @@ int spi_xfer(struct spi_slave *slave, unsigned int
>> bitlen, const void *dout,
>>
>>  	return 0;
>>  }
>> +#ifdef CONFIG_TI_EDMA3
>> +void spi_flash_copy_mmap(void *data, void *offset, size_t len)
>> +{
>> +	unsigned int			addr = (unsigned int) (data);
>> +	unsigned int			edma_slot_num = 1;
>> +
>> +	/* Invalidate the area, so no writeback into the RAM races with DMA */
>> +	invalidate_dcache_range(addr, addr + roundup(len, ARCH_DMA_MINALIGN));
>> +
>> +	/* enable edma3 clocks */
>> +	enable_edma3_clocks();
>> +
>> +	/* Call edma3 api to do actual DMA transfer	*/
>> +	edma3_transfer(EDMA3_BASE, edma_slot_num, data, offset, len);
>> +
>> +	/* disable edma3 clocks */
>> +	disable_edma3_clocks();
>> +
>> +	*((unsigned int *)offset) += len;
>> +}
>> +#endif
>>
>>
>> -- 
>> Regards
>> Vignesh
>