arm: atmel: sama5d3: enable spl boot from SD card
Subsystems (2)
v3 cover arm: atmel: sama5d3: enable spl boot from SD card
This patch series enable spl boot from SD card, it only can boot u-boot itself.
Changes in v3:
- Correct the clock enable code, the ID can not OR
- Move to at91 common folder
- Move plla and mck configure to spl.c file
Changes in v2:
- Move spl related code to at91-common folder
Bo Shen (6):
arm: atmel: sama5d3: correct the ID for DBGU and PIT
arm: atmel: sama5d3: correct the error define of DIV
arm: atmel: sama5d3: the offset of MULA is 18
arm: atmel: sama5d3: early enable PIO peripherals
arm: atmel: add ddr2 initialization function
arm: atmel: sama5d3: spl boot from fat fs SD card
arch/arm/cpu/armv7/Makefile | 2 +- arch/arm/cpu/armv7/at91/sama5d3_devices.c | 2 +- arch/arm/cpu/armv7/at91/timer.c | 2 +- arch/arm/cpu/at91-common/Makefile | 33 +++++++ arch/arm/cpu/at91-common/mpddrc.c | 123 +++++++++++++++++++++++++ arch/arm/cpu/at91-common/spl.c | 90 ++++++++++++++++++ arch/arm/cpu/at91-common/u-boot-spl.lds | 50 ++++++++++ arch/arm/include/asm/arch-at91/at91_common.h | 4 + arch/arm/include/asm/arch-at91/at91_pmc.h | 8 +- arch/arm/include/asm/arch-at91/atmel_mpddrc.h | 111 ++++++++++++++++++++++ arch/arm/include/asm/arch-at91/spl.h | 20 ++++ board/atmel/sama5d3xek/sama5d3xek.c | 88 ++++++++++++++++++ include/configs/sama5d3xek.h | 34 +++++++ spl/Makefile | 4 + 14 files changed, 566 insertions(+), 5 deletions(-) create mode 100644 arch/arm/cpu/at91-common/Makefile create mode 100644 arch/arm/cpu/at91-common/mpddrc.c create mode 100644 arch/arm/cpu/at91-common/spl.c create mode 100644 arch/arm/cpu/at91-common/u-boot-spl.lds create mode 100644 arch/arm/include/asm/arch-at91/atmel_mpddrc.h create mode 100644 arch/arm/include/asm/arch-at91/spl.h
Signature · 1 line
1.7.9.5
PATCH v3 1/6 arm: atmel: sama5d3: correct the ID for DBGU and PIT
As the DBGU and PIT has its own ID on sama5d3 SoC, while not share with SYS ID. So, correct them.
---
Changes in v3:
- None
Changes in v2:
- None
arch/arm/cpu/armv7/at91/sama5d3_devices.c | 2 +- arch/arm/cpu/armv7/at91/timer.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
arch/arm/cpu/armv7/at91/sama5d3_devices.c +1 -1
@@ -82,7 +82,7 @@ void at91_seriald_hw_init(void)at91_set_a_periph (AT91_PIO_PORTB ,30 ,0 );/* DRXD */ /* Enable clock */ -at91_periph_clk_enable (ATMEL_ID_SYS ); +at91_periph_clk_enable (ATMEL_ID_DBGU ); }#if defined(CONFIG_ATMEL_SPI )
arch/arm/cpu/armv7/at91/timer.c +1 -1
@@ -60,7 +60,7 @@ int timer_init(void)at91_pit_t *pit = (at91_pit_t *)ATMEL_BASE_PIT ;/* Enable PITC Clock */ -at91_periph_clk_enable (ATMEL_ID_SYS ); +at91_periph_clk_enable (ATMEL_ID_PIT );/* Enable PITC */ writel (TIMER_LOAD_VAL |AT91_PIT_MR_EN , &pit->mr);
Signature · 1 line
1.7.9.5
PATCH v3 1/6 arm: atmel: sama5d3: correct the ID for DBGU and PIT
Hi Bo,
Quoted 6 lines · Bo Shen
As the DBGU and PIT has its own ID on sama5d3 SoC, while not share with SYS ID. So, correct them.
Signed-off-by Bo Shen <[email protected]>---
no complaints
Best regards
Andreas Bie?mann
PATCH v3 2/6 arm: atmel: sama5d3: correct the error define of DIV
In reply to Bo ShenCorrect the error define of DIV.
---
Changes in v3:
- None
Changes in v2:
- None
arch/arm/include/asm/arch-at91/at91_pmc.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
arch/arm/include/asm/arch-at91/at91_pmc.h +2 -2
@@ -124,8 +124,8 @@ typedef struct at91_pmc {#define AT91_PMC_MCKR_MDIV_MASK 0x00000300#endif -#define AT91_PMC_MCKR_PLLADIV_1 0x00001000 -#define AT91_PMC_MCKR_PLLADIV_2 0x00002000 +#define AT91_PMC_MCKR_PLLADIV_1 0x00000000 +#define AT91_PMC_MCKR_PLLADIV_2 0x00001000#define AT91_PMC_IXR_MOSCS 0x00000001#define AT91_PMC_IXR_LOCKA 0x00000002
Signature · 1 line
1.7.9.5
PATCH v3 2/6 arm: atmel: sama5d3: correct the error define of DIV
Hi Bo,
Quoted 19 lines · Bo Shen
Correct the error define of DIV.
Signed-off-by Bo Shen <[email protected]>---
Changes in v3:
- NoneChanges in v2:
- Nonearch/arm/include/asm/arch-at91/at91_pmc.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)arch/arm/include/asm/arch-at91/at91_pmc.h +0
-0
arch/arm/include/asm/arch-at91/at91_pmc.h typedef struct at91_pmc { · line 128
Quoted 7 lines
arch/arm/include/asm/arch-at91/at91_pmc.h +2
-2@@ -124,8 +124,8 @@ typedef struct at91_pmc {#define AT91_PMC_MCKR_MDIV_MASK 0x00000300#endif -#define AT91_PMC_MCKR_PLLADIV_1 0x00001000 -#define AT91_PMC_MCKR_PLLADIV_2 0x00002000 +#define AT91_PMC_MCKR_PLLADIV_1 0x00000000 +#define AT91_PMC_MCKR_PLLADIV_2 0x00001000
this change will touch pm9261 board. I dunno if your change is correct for that board. I also wonder why one should set bit position 13 in MCKR ... I can't find any source what that bit could be. However mature documentation for sam9261 (which is the SoC on pm9261) doesn't name bit position 12 either.
So I'm fine with your change, but please fix the pm9261 board also (use the PLLADIV_2 define).
arch/arm/include/asm/arch-at91/at91_pmc.h typedef struct at91_pmc {
Quoted 4 lines
arch/arm/include/asm/arch-at91/at91_pmc.h +0
-0#define AT91_PMC_IXR_MOSCS 0x00000001#define AT91_PMC_IXR_LOCKA 0x00000002
Best regards
Andreas Bie?mann
PATCH v3 2/6 arm: atmel: sama5d3: correct the error define of DIV
Hi Andreas,
Quoted 36 lines · Andreas Bie?mann
Hi Bo,
Quoted 19 lines · Bo Shen
Correct the error define of DIV.
Signed-off-by Bo Shen <[email protected]>---
Changes in v3:
- NoneChanges in v2:
- Nonearch/arm/include/asm/arch-at91/at91_pmc.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)arch/arm/include/asm/arch-at91/at91_pmc.h +0
-0arch/arm/include/asm/arch-at91/at91_pmc.h typedef struct at91_pmc { · line 128
Quoted 7 lines
arch/arm/include/asm/arch-at91/at91_pmc.h +2
-2@@ -124,8 +124,8 @@ typedef struct at91_pmc {#define AT91_PMC_MCKR_MDIV_MASK 0x00000300#endif -#define AT91_PMC_MCKR_PLLADIV_1 0x00001000 -#define AT91_PMC_MCKR_PLLADIV_2 0x00002000 +#define AT91_PMC_MCKR_PLLADIV_1 0x00000000 +#define AT91_PMC_MCKR_PLLADIV_2 0x00001000this change will touch pm9261 board. I dunno if your change is correct for that board. I also wonder why one should set bit position 13 in MCKR ... I can't find any source what that bit could be. However mature documentation for sam9261 (which is the SoC on pm9261) doesn't name bit position 12 either.
Quoted 2 lines
So I'm fine with your change, but please fix the pm9261 board also (use the
PLLADIV_2define).
Ok, thanks.
Quoted 9 lines
arch/arm/include/asm/arch-at91/at91_pmc.h typedef struct at91_pmc {
Quoted 4 lines
arch/arm/include/asm/arch-at91/at91_pmc.h +0
-0#define AT91_PMC_IXR_MOSCS 0x00000001#define AT91_PMC_IXR_LOCKA 0x00000002Best regards
Andreas Bie?mann
Best Regards,
Bo Shen
PATCH v3 3/6 arm: atmel: sama5d3: the offset of MULA is 18
In reply to Bo ShenThe offset of MULA field in PLLA register in sama5d3 is 18, and the length only 7 bits.
---
Changes in v3:
- None
Changes in v2:
- None
arch/arm/include/asm/arch-at91/at91_pmc.h | 4 ++++ 1 file changed, 4 insertions(+)
arch/arm/include/asm/arch-at91/at91_pmc.h +4 -0
@@ -73,7 +73,11 @@ typedef struct at91_pmc {#define AT91_PMC_PLLXR_DIV (x) (x & 0xFF)#define AT91_PMC_PLLXR_PLLCOUNT (x) ((x & 0x3F) << 8)#define AT91_PMC_PLLXR_OUT (x) ((x & 0x03) << 14) +#ifdef CONFIG_SAMA5D3 +#define AT91_PMC_PLLXR_MUL (x) ((x & 0x7F) << 18) +#else #define AT91_PMC_PLLXR_MUL (x) ((x & 0x7FF) << 16) +#endif #define AT91_PMC_PLLAR_29 0x20000000#define AT91_PMC_PLLBR_USBDIV_1 0x00000000#define AT91_PMC_PLLBR_USBDIV_2 0x10000000
Signature · 1 line
1.7.9.5
PATCH v3 3/6 arm: atmel: sama5d3: the offset of MULA is 18
Hi Bo,
Quoted 6 lines · Bo Shen
The offset of MULA field in PLLA register in sama5d3 is 18, and the length only 7 bits.
Signed-off-by Bo Shen <[email protected]>---
no complaints
Best regards
Andreas Bie?mann
PATCH v3 4/6 arm: atmel: sama5d3: early enable PIO peripherals
In reply to Bo ShenEnable the PIO peripherals early than other peripherals.
---
Changes in v3:
- Correct the clock enable code, the ID can not OR
Changes in v2:
- None
board/atmel/sama5d3xek/sama5d3xek.c | 6 ++++++ 1 file changed, 6 insertions(+)
board/atmel/sama5d3xek/sama5d3xek.c +6 -0
@@ -158,6 +158,12 @@ void lcd_show_board_info(void)int board_early_init_f (void ) { +at91_periph_clk_enable (ATMEL_ID_PIOA ); +at91_periph_clk_enable (ATMEL_ID_PIOB ); +at91_periph_clk_enable (ATMEL_ID_PIOC ); +at91_periph_clk_enable (ATMEL_ID_PIOD ); +at91_periph_clk_enable (ATMEL_ID_PIOE ); +at91_seriald_hw_init ();return 0 ;
Signature · 1 line
1.7.9.5
PATCH v3 4/6 arm: atmel: sama5d3: early enable PIO peripherals
Hi Bo,
Quoted 19 lines · Bo Shen
Enable the PIO peripherals early than other peripherals.
Signed-off-by Bo Shen <[email protected]>---
Changes in v3:
- Correct the clock enable code, the ID can not ORChanges in v2:
- Noneboard/atmel/sama5d3xek/sama5d3xek.c | 6 ++++++ 1 file changed, 6 insertions(+)board/atmel/sama5d3xek/sama5d3xek.c +0
-0
board/atmel/sama5d3xek/sama5d3xek.c void lcd_show_board_info(void) · line 159
Quoted 2 lines
board/atmel/sama5d3xek/sama5d3xek.c +0
-0@@ -158,6 +158,12 @@ void lcd_show_board_info(void)int board_early_init_f (void )
I'm still not really sure if _board_ init is the right place ... I feel more like some _arch_ init.
But I would accept it here for getting the first SPL in at91. We need to rework that whole arch vs. board related init anyways, so we could change it then (and insisting on that _now_ would definitely break the release date).
board/atmel/sama5d3xek/sama5d3xek.c void lcd_show_board_info(void) · line 169
Quoted 11 lines
board/atmel/sama5d3xek/sama5d3xek.c +6
-0{ +at91_periph_clk_enable (ATMEL_ID_PIOA ); +at91_periph_clk_enable (ATMEL_ID_PIOB ); +at91_periph_clk_enable (ATMEL_ID_PIOC ); +at91_periph_clk_enable (ATMEL_ID_PIOD ); +at91_periph_clk_enable (ATMEL_ID_PIOE ); +at91_seriald_hw_init ();return 0 ;
To say it clear, for this patch no change requested.
Best regards
Andreas Bie?mann
PATCH v3 5/6 arm: atmel: add ddr2 initialization function
In reply to Bo ShenThe MPDDRC supports different type of SDRAM
This patch add ddr2 initialization function
---
Changes in v3:
- Move to at91 common folder
Changes in v2:
- None
arch/arm/cpu/at91-common/Makefile | 32 +++++++ arch/arm/cpu/at91-common/mpddrc.c | 123 +++++++++++++++++++++++++ arch/arm/include/asm/arch-at91/atmel_mpddrc.h | 111 ++++++++++++++++++++++ spl/Makefile | 4 + 4 files changed, 270 insertions(+) create mode 100644 arch/arm/cpu/at91-common/Makefile create mode 100644 arch/arm/cpu/at91-common/mpddrc.c create mode 100644 arch/arm/include/asm/arch-at91/atmel_mpddrc.h
arch/arm/cpu/at91-common/Makefile new file +32 -0
@@ -0,0 +1,32 @@ +# +# (C) Copyright 2000-2008 +# Wolfgang Denk, DENX Software Engineering, wd at denx.de. +# +# (C) Copyright 2013 Atmel Corporation +# Bo Shen <[email protected]> +# +# SPDX-License-Identifier: GPL-2.0+ +# + +include $(TOPDIR)/config.mk + +LIB = $(obj)libat91-common.o + +COBJS-$(CONFIG_SPL_BUILD) += mpddrc.o + +SRCS := $(SOBJS-y:.o=.S) $(COBJS-y:.o=.c) +OBJS := $(addprefix $(obj),$(SOBJS-y) $(COBJS-y)) + +all: $(obj).depend $(LIB) + +$(LIB): $(OBJS) + $(call cmd_link_o_target, $(OBJS)) + +######################################################################### + +# defines $(obj).depend target +include $(SRCTREE)/rules.mk + +sinclude $(obj).depend + +#########################################################################
arch/arm/cpu/at91-common/mpddrc.c new file +123 -0
123 lines not on this page · raw message
arch/arm/include/asm/arch-at91/atmel_mpddrc.h new file +111 -0
111 lines not on this page · raw message
spl/Makefile +4 -0
@@ -123,6 +123,10 @@ ifeq ($(SOC),exynos) LIBS-y += $(CPUDIR)/s5p-common/libs5p-common.o endif +ifeq ($(SOC),at91) +LIBS-y += arch/$(ARCH)/cpu/at91-common/libat91-common.o +endif + # Add GCC lib ifeq ("$(USE_PRIVATE_LIBGCC)", "yes") PLATFORM_LIBGCC = $(SPLTREE)/arch/$(ARCH)/lib/libgcc.o
Signature · 1 line
1.7.9.5
PATCH v3 5/6 arm: atmel: add ddr2 initialization function
Hi Bo,
Quoted 49 lines · Bo Shen
The MPDDRC supports different type of SDRAM
This patch add ddr2 initialization functionSigned-off-by Bo Shen <[email protected]>---
Changes in v3:
- Move to at91 common folderChanges in v2:
- Nonearch/arm/cpu/at91-common/Makefile | 32 +++++++ arch/arm/cpu/at91-common/mpddrc.c | 123 +++++++++++++++++++++++++ arch/arm/include/asm/arch-at91/atmel_mpddrc.h | 111 ++++++++++++++++++++++ spl/Makefile | 4 + 4 files changed, 270 insertions(+) create mode 100644 arch/arm/cpu/at91-common/Makefile create mode 100644 arch/arm/cpu/at91-common/mpddrc.c create mode 100644 arch/arm/include/asm/arch-at91/atmel_mpddrc.harch/arm/cpu/at91-common/Makefile new file +22
-0@@ -0,0 +1,32 @@ +# +# (C) Copyright 2000-2008 +# Wolfgang Denk, DENX Software Engineering, wd at denx.de. +# +# (C) Copyright 2013 Atmel Corporation +# Bo Shen <[email protected]> +# +# SPDX-License-Identifier: GPL-2.0+ +# + +include $(TOPDIR)/config.mk + +LIB = $(obj)libat91-common.o + +COBJS-$(CONFIG_SPL_BUILD) += mpddrc.o + +SRCS := $(SOBJS-y:.o=.S) $(COBJS-y:.o=.c) +OBJS := $(addprefix $(obj),$(SOBJS-y) $(COBJS-y)) + +all: $(obj).depend $(LIB) + +$(LIB): $(OBJS)
arch/arm/cpu/at91-common/Makefile line 32
Quoted 10 lines
arch/arm/cpu/at91-common/Makefile new file +10
-0+ $(call cmd_link_o_target, $(OBJS)) + +######################################################################### + +# defines $(obj).depend target +include $(SRCTREE)/rules.mk + +sinclude $(obj).depend + +#########################################################################
please rewrite in KBuild style.
Quoted 8 lines
arch/arm/cpu/at91-common/mpddrc.c new file +2
-0@@ -0,0 +1,123 @@ +/* +* Copyright (C) 2013 Atmel Corporation
arch/arm/cpu/at91-common/mpddrc.c line 12
Quoted 10 lines
arch/arm/cpu/at91-common/mpddrc.c new file +10
-0+* Bo Shen <[email protected]> +* +* SPDX-License-Identifier: GPL-2.0+ +*/ + +#include <common.h> +#include <asm/io.h> +#include <asm/arch/atmel_mpddrc.h> + +static void atmel_mpddr_op (int mode,u32 ram_address)
static inline, could give the compiler a hint to optimize here.
arch/arm/cpu/at91-common/mpddrc.c line 20
Quoted 8 lines
arch/arm/cpu/at91-common/mpddrc.c new file +8
-0+{ +struct atmel_mpddr *mpddr = (struct atmel_mpddr *)ATMEL_BASE_MPDDRC ; + +writel (mode, &mpddr->mr); +writel (0 , ram_address); +} + +int ddr2_init (u32 ram_address,struct atmel_mpddr *mpddr_value)
could you please constify mpddr_value for the very same reason?
Quoted 61 lines
arch/arm/cpu/at91-common/mpddrc.c new file +61
-0+{ +struct atmel_mpddr *mpddr = (struct atmel_mpddr *)ATMEL_BASE_MPDDRC ; +u32 ba_off, cr; + +/* Compute bank offset according to NC in configuration register */ + ba_off = (mpddr_value->cr &ATMEL_MPDDRC_CR_NC_MASK ) +9 ; +if (!(mpddr_value->cr &ATMEL_MPDDRC_CR_DECOD_INTERLEAVED )) + ba_off += ((mpddr->cr &ATMEL_MPDDRC_CR_NR_MASK ) >>2 ) +11 ; + + ba_off += (mpddr_value->mdr &ATMEL_MPDDRC_MDR_DBW_MASK ) ?1 :2 ; + +/* Program the memory device type into the memory device register */ +writel (mpddr_value->mdr, &mpddr->mdr); + +/* Program the configuration register */ +writel (mpddr_value->cr, &mpddr->cr); + +/* Program the timing register */ +writel (mpddr_value->tp0r, &mpddr->tp0r); +writel (mpddr_value->tp1r, &mpddr->tp1r); +writel (mpddr_value->tp2r, &mpddr->tp2r); + +/* Issue a NOP command */ +atmel_mpddr_op (ATMEL_MPDDRC_MR_NOP_CMD , ram_address); + +/* A 200 us is provided to precede any signal toggle */ +udelay (200 ); + +/* Issue a NOP command */ +atmel_mpddr_op (ATMEL_MPDDRC_MR_NOP_CMD , ram_address); + +/* Issue an all banks precharge command */ +atmel_mpddr_op (ATMEL_MPDDRC_MR_PRCGALL_CMD , ram_address); + +/* Issue an extended mode register set(EMRS2) to choose operation */ +atmel_mpddr_op (ATMEL_MPDDRC_MR_EXT_LMR_CMD , + ram_address + (0x2 << ba_off)); + +/* Issue an extended mode register set(EMRS3) to set EMSR to 0 */ +atmel_mpddr_op (ATMEL_MPDDRC_MR_EXT_LMR_CMD , + ram_address + (0x3 << ba_off)); + +/* +* Issue an extended mode register set(EMRS1) to enable DLL and +* program D.I.C (output driver impedance control) +*/ +atmel_mpddr_op (ATMEL_MPDDRC_MR_EXT_LMR_CMD , + ram_address + (0x1 << ba_off)); + +/* Enable DLL reset */ + cr =readl (&mpddr->cr); +writel (cr |ATMEL_MPDDRC_CR_EN_RESET_DLL , &mpddr->cr); + +/* A mode register set(MRS) cycle is issued to reset DLL */ +atmel_mpddr_op (ATMEL_MPDDRC_MR_LMR_CMD , ram_address); + +/* Issue an all banks precharge command */ +atmel_mpddr_op (ATMEL_MPDDRC_MR_PRCGALL_CMD , ram_address); + +/* Two auto-refresh (CBR) cycles are provided */ +atmel_mpddr_op (ATMEL_MPDDRC_MR_RFSH_CMD , ram_address);
arch/arm/cpu/at91-common/mpddrc.c line 91
Quoted 10 lines
arch/arm/cpu/at91-common/mpddrc.c new file +10
-0+atmel_mpddr_op (ATMEL_MPDDRC_MR_RFSH_CMD , ram_address); + +/* Disable DLL reset */ + cr =readl (&mpddr->cr); +writel (cr & (~ATMEL_MPDDRC_CR_EN_RESET_DLL ), &mpddr->cr); + +/* A mode register set (MRS) cycle is issued to disable DLL reset */ +atmel_mpddr_op (ATMEL_MPDDRC_MR_LMR_CMD , ram_address); + +/* Set OCD calibration in defautl state */
typo default
Quoted 20 lines
arch/arm/cpu/at91-common/mpddrc.c new file +20
-0+ cr =readl (&mpddr->cr); +writel (cr |ATMEL_MPDDRC_CR_OCD_DEFAULT , &mpddr->cr); + +/* +* An extended mode register set (EMRS1) cycle is issued +* to OCD default value +*/ +atmel_mpddr_op (ATMEL_MPDDRC_MR_EXT_LMR_CMD , + ram_address + (0x1 << ba_off)); + +/* OCD calibration mode exit */ + cr =readl (&mpddr->cr); +writel (cr & (~ATMEL_MPDDRC_CR_OCD_DEFAULT ), &mpddr->cr); + +/* +* An extended mode register set (EMRS1) cycle is issued +* to enable OCD exit +*/ +atmel_mpddr_op (ATMEL_MPDDRC_MR_EXT_LMR_CMD , + ram_address + (0x1 << ba_off));
arch/arm/cpu/at91-common/mpddrc.c line 121
Quoted 10 lines
arch/arm/cpu/at91-common/mpddrc.c new file +10
-0+ +/* A nornal mode command is provided */ +atmel_mpddr_op (ATMEL_MPDDRC_MR_NORMAL_CMD , ram_address); + +/* Perform a write access to any DDR2-SDRAM address */ +writel (0 , ram_address); + +/* Write the refresh rate */ +writel (mpddr_value->rtr, &mpddr->rtr); +
I haven't checked that sequence deeply, I trust in you that it is correct ;)
Quoted 13 lines
arch/arm/cpu/at91-common/mpddrc.c new file +2
-0+return 0 ; +}arch/arm/include/asm/arch-at91/atmel_mpddrc.h new file +5
-0@@ -0,0 +1,111 @@ +/* +* Copyright (C) 2013 Atmel Corporation +* Bo Shen <[email protected]> +* +* SPDX-License-Identifier: GPL-2.0+
arch/arm/include/asm/arch-at91/atmel_mpddrc.h line 15
Quoted 10 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h new file +10
-0+*/ + +#ifndef __ATMEL_MPDDRC_H__ +#define __ATMEL_MPDDRC_H__ + +struct atmel_mpddr { + u32 mr; + u32 rtr; + u32 cr; + u32 tp0r;
Datasheet names them tprX
arch/arm/include/asm/arch-at91/atmel_mpddrc.h line 19
Quoted 4 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h new file +4
-0+u32 tp1r; +u32 tp2r; +u32 reserved[2 ]; +u32 mdr;
Datasheet names it just md.
I think it would be good to also add the other register positions or at least mention that these the only one needed currently (in a comment right here in the struct).
Quoted 4 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h new file +4
-0+}; + +int ddr2_init (unsigned int ram_address, +struct atmel_mpddr *mpddr);
arch/arm/include/asm/arch-at91/atmel_mpddrc.h line 33
Quoted 10 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h new file +10
-0+ +/* bit field in mode register */ +#define ATMEL_MPDDRC_MR_NORMAL_CMD 0x0 +#define ATMEL_MPDDRC_MR_NOP_CMD 0x1 +#define ATMEL_MPDDRC_MR_PRCGALL_CMD 0x2 +#define ATMEL_MPDDRC_MR_LMR_CMD 0x3 +#define ATMEL_MPDDRC_MR_RFSH_CMD 0x4 +#define ATMEL_MPDDRC_MR_EXT_LMR_CMD 0x5 +#define ATMEL_MPDDRC_MR_DEEP_CMD 0x6 +#define ATMEL_MPDDRC_MR_LPDDR2_CMD 0x7
Could you please drop the tabs between 'define' and the definition.
Quoted 17 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h new file +17
-0+ +/* bit field in configuration register */ +#define ATMEL_MPDDRC_CR_NC_MASK 0x3 +#define ATMEL_MPDDRC_CR_NC_COL_9 0x0 +#define ATMEL_MPDDRC_CR_NC_COL_10 0x1 +#define ATMEL_MPDDRC_CR_NC_COL_11 0x2 +#define ATMEL_MPDDRC_CR_NC_COL_12 0x3 +#define ATMEL_MPDDRC_CR_NR_MASK (0x3 << 2) +#define ATMEL_MPDDRC_CR_NR_ROW_11 (0x0 << 2) +#define ATMEL_MPDDRC_CR_NR_ROW_12 (0x1 << 2) +#define ATMEL_MPDDRC_CR_NR_ROW_13 (0x2 << 2) +#define ATMEL_MPDDRC_CR_NR_ROW_14 (0x3 << 2) +#define ATMEL_MPDDRC_CR_CAS (0x7 << 4) +#define ATMEL_MPDDRC_CR_CAS_2 (0x2 << 4) +#define ATMEL_MPDDRC_CR_CAS_3 (0x3 << 4) +#define ATMEL_MPDDRC_CR_CAS_4 (0x4 << 4) +#define ATMEL_MPDDRC_CR_CAS_5 (0x5 << 4)
arch/arm/include/asm/arch-at91/atmel_mpddrc.h line 60
Quoted 10 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h new file +10
-0+#define ATMEL_MPDDRC_CR_CAS_6 (0x6 << 4) +#define ATMEL_MPDDRC_CR_EN_RESET_DLL (0x1 << 7) +#define ATMEL_MPDDRC_CR_DIC_DS (0x1 << 8) +#define ATMEL_MPDDRC_CR_DIS_DLL (0x1 << 9) +#define ATMEL_MPDDRC_CR_OCD_DEFAULT (0x7 << 12) +#define ATMEL_MPDDRC_CR_EN_ENRDM (0x1 << 17) +#define ATMEL_MPDDRC_CR_NB_8BANKS (0x1 << 20) +#define ATMEL_MPDDRC_CR_DIS_NDQS (0x1 << 21) +#define ATMEL_MPDDRC_CR_DECOD_INTERLEAVED (0x1 << 22) +#define ATMEL_MPDDRC_CR_UNAL_SUPPORTED (0x1 << 23)
Some of the defines have the strict scheme <offset><register name><field name> with <offset> being ATMEL_MPDDRC, <register name> CR here and the respective field names from datasheet. Some however have some more information like UNAL_SUPPORTED or DECOD_INTERLEAVED (those two are Ok I think, but we could discuss if we like to have the strict scheme or leave some space). But EN_ENRDM is definitely too much ;) Has anyone a opinion about the strict scheme?
Regarding EN_ENRDM I think using ENRDM_ON would be better.
Quoted 56 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h new file +51
-0+ +/* bit field in timing parameter 0 register */ +#define ATMEL_MPDDRC_TP0R_TRAS_OFFSET 0 +#define ATMEL_MPDDRC_TP0R_TRAS_MASK 0xf +#define ATMEL_MPDDRC_TP0R_TRCD_OFFSET 4 +#define ATMEL_MPDDRC_TP0R_TRCD_MASK 0xf +#define ATMEL_MPDDRC_TP0R_TWR_OFFSET 8 +#define ATMEL_MPDDRC_TP0R_TWR_MASK 0xf +#define ATMEL_MPDDRC_TP0R_TRC_OFFSET 12 +#define ATMEL_MPDDRC_TP0R_TRC_MASK 0xf +#define ATMEL_MPDDRC_TP0R_TRP_OFFSET 16 +#define ATMEL_MPDDRC_TP0R_TRP_MASK 0xf +#define ATMEL_MPDDRC_TP0R_TRRD_OFFSET 20 +#define ATMEL_MPDDRC_TP0R_TRRD_MASK 0xf +#define ATMEL_MPDDRC_TP0R_TWTR_OFFSET 24 +#define ATMEL_MPDDRC_TP0R_TWTR_MASK 0x7 +#define ATMEL_MPDDRC_TP0R_RDC_WRRD_OFFSET 27 +#define ATMEL_MPDDRC_TP0R_RDC_WRRD_MASK 0x1 +#define ATMEL_MPDDRC_TP0R_TMRD_OFFSET 28 +#define ATMEL_MPDDRC_TP0R_TMRD_MASK 0xf + +/* bit field in timing parameter 1 register */ +#define ATMEL_MPDDRC_TP1R_TRFC_OFFSET 0 +#define ATMEL_MPDDRC_TP1R_TRFC_MASK 0x7f +#define ATMEL_MPDDRC_TP1R_TXSNR_OFFSET 8 +#define ATMEL_MPDDRC_TP1R_TXSNR_MASK 0xff +#define ATMEL_MPDDRC_TP1R_TXSRD_OFFSET 16 +#define ATMEL_MPDDRC_TP1R_TXSRD_MASK 0xff +#define ATMEL_MPDDRC_TP1R_TXP_OFFSET 24 +#define ATMEL_MPDDRC_TP1R_TXP_MASK 0xf + +/* bit field in timing parameter 2 register */ +#define ATMEL_MPDDRC_TP2R_TXARD_OFFSET 0 +#define ATMEL_MPDDRC_TP2R_TXARD_MASK 0xf +#define ATMEL_MPDDRC_TP2R_TXARDS_OFFSET 4 +#define ATMEL_MPDDRC_TP2R_TXARDS_MASK 0xf +#define ATMEL_MPDDRC_TP2R_TRPA_OFFSET 8 +#define ATMEL_MPDDRC_TP2R_TRPA_MASK 0xf +#define ATMEL_MPDDRC_TP2R_TRTP_OFFSET 12 +#define ATMEL_MPDDRC_TP2R_TRTP_MASK 0x7 +#define ATMEL_MPDDRC_TP2R_TFAW_OFFSET 16 +#define ATMEL_MPDDRC_TP2R_TFAW_MASK 0xf + +/* bit field in Memory Device Register */ +#define ATMEL_MPDDRC_MDR_LPDDR_SDRAM 0x3 +#define ATMEL_MPDDRC_MDR_DDR2_SDRAM 0x6 +#define ATMEL_MPDDRC_MDR_DBW_MASK (0x1 << 4) +#define ATMEL_MPDDRC_MDR_DBW_32BITS (0x0 << 4) +#define ATMEL_MPDDRC_MDR_DBW_16BITS (0x1 << 4) + +#endif spl/Makefile +0
-0
spl/Makefile ifeq ($(SOC),exynos) · line 128
Quoted 6 lines
spl/Makefile +3
-0@@ -123,6 +123,10 @@ ifeq ($(SOC),exynos) LIBS-y += $(CPUDIR)/s5p-common/libs5p-common.o endif +ifeq ($(SOC),at91) +LIBS-y += arch/$(ARCH)/cpu/at91-common/libat91-common.o +endif
The libat91-common.o should be built in any case. The mpddrc.o should only be included for SPL build (that mentioned Heiko in another mail).
spl/Makefile ifeq ($(SOC),exynos) · line 132
Quoted 5 lines
spl/Makefile +1
-0+ # Add GCC lib ifeq ("$(USE_PRIVATE_LIBGCC)", "yes") PLATFORM_LIBGCC = $(SPLTREE)/arch/$(ARCH)/lib/libgcc.o
Best regards
Andreas Bie?mann
PATCH v3 5/6 arm: atmel: add ddr2 initialization function
Hi Andreas,
Quoted 56 lines · Andreas Bie?mann
Hi Bo,
Quoted 53 lines · Bo Shen
The MPDDRC supports different type of SDRAM
This patch add ddr2 initialization functionSigned-off-by Bo Shen <[email protected]>---
Changes in v3:
- Move to at91 common folderChanges in v2:
- Nonearch/arm/cpu/at91-common/Makefile | 32 +++++++ arch/arm/cpu/at91-common/mpddrc.c | 123 +++++++++++++++++++++++++ arch/arm/include/asm/arch-at91/atmel_mpddrc.h | 111 ++++++++++++++++++++++ spl/Makefile | 4 + 4 files changed, 270 insertions(+)create mode 100644
arch/arm/cpu/at91-common/Makefilecreate mode 100644arch/arm/cpu/at91-common/mpddrc.ccreate mode 100644arch/arm/include/asm/arch-at91/atmel_mpddrc.harch/arm/cpu/at91-common/Makefile new file +26
-0@@ -0,0 +1,32 @@ +# +# (C) Copyright 2000-2008 +# Wolfgang Denk, DENX Software Engineering, wd at denx.de. +# +# (C) Copyright 2013 Atmel Corporation +# Bo Shen <[email protected]> +# +# SPDX-License-Identifier: GPL-2.0+ +# + +include $(TOPDIR)/config.mk + +LIB = $(obj)libat91-common.o + +COBJS-$(CONFIG_SPL_BUILD) += mpddrc.o + +SRCS := $(SOBJS-y:.o=.S) $(COBJS-y:.o=.c) +OBJS := $(addprefix $(obj),$(SOBJS-y) $(COBJS-y)) + +all: $(obj).depend $(LIB) + +$(LIB): $(OBJS) + $(call cmd_link_o_target, $(OBJS)) + +######################################################################### +
Quoted 8 lines
arch/arm/cpu/at91-common/Makefile line 32
Quoted 6 lines
arch/arm/cpu/at91-common/Makefile new file +6
-0+# defines $(obj).depend target +include $(SRCTREE)/rules.mk + +sinclude $(obj).depend + +#########################################################################please rewrite in KBuild style.
OK, I will use KBuild style in next version.
Quoted 12 lines
Quoted 12 lines
arch/arm/cpu/at91-common/mpddrc.c new file +6
-0@@ -0,0 +1,123 @@ +/* +* Copyright (C) 2013 Atmel Corporation +* Bo Shen <[email protected]> +* +* SPDX-License-Identifier: GPL-2.0+ +*/
Quoted 8 lines
arch/arm/cpu/at91-common/mpddrc.c line 12
Quoted 6 lines
arch/arm/cpu/at91-common/mpddrc.c new file +6
-0+ +#include <common.h> +#include <asm/io.h> +#include <asm/arch/atmel_mpddrc.h> + +static void atmel_mpddr_op (int mode,u32 ram_address)static inline, could give the compiler a hint to optimize here.
I am not sure whether the write(0, ram_address) will be optimized. Because this must excute later than write mode to let it work.
Quoted 10 lines
arch/arm/cpu/at91-common/mpddrc.c line 20
Quoted 8 lines
arch/arm/cpu/at91-common/mpddrc.c new file +8
-0+{ +struct atmel_mpddr *mpddr = (struct atmel_mpddr *)ATMEL_BASE_MPDDRC ; + +writel (mode, &mpddr->mr); +writel (0 , ram_address); +} + +int ddr2_init (u32 ram_address,struct atmel_mpddr *mpddr_value)could you please constify
mpddr_valuefor the very same reason?
OK.
Quoted 65 lines
Quoted 65 lines
arch/arm/cpu/at91-common/mpddrc.c new file +65
-0+{ +struct atmel_mpddr *mpddr = (struct atmel_mpddr *)ATMEL_BASE_MPDDRC ; +u32 ba_off, cr; + +/* Compute bank offset according to NC in configuration register */ + ba_off = (mpddr_value->cr &ATMEL_MPDDRC_CR_NC_MASK ) +9 ; +if (!(mpddr_value->cr &ATMEL_MPDDRC_CR_DECOD_INTERLEAVED )) + ba_off += ((mpddr->cr &ATMEL_MPDDRC_CR_NR_MASK ) >>2 ) +11 ; + + ba_off += (mpddr_value->mdr &ATMEL_MPDDRC_MDR_DBW_MASK ) ?1 :2 ; + +/* Program the memory device type into the memory device register */ +writel (mpddr_value->mdr, &mpddr->mdr); + +/* Program the configuration register */ +writel (mpddr_value->cr, &mpddr->cr); + +/* Program the timing register */ +writel (mpddr_value->tp0r, &mpddr->tp0r); +writel (mpddr_value->tp1r, &mpddr->tp1r); +writel (mpddr_value->tp2r, &mpddr->tp2r); + +/* Issue a NOP command */ +atmel_mpddr_op (ATMEL_MPDDRC_MR_NOP_CMD , ram_address); + +/* A 200 us is provided to precede any signal toggle */ +udelay (200 ); + +/* Issue a NOP command */ +atmel_mpddr_op (ATMEL_MPDDRC_MR_NOP_CMD , ram_address); + +/* Issue an all banks precharge command */ +atmel_mpddr_op (ATMEL_MPDDRC_MR_PRCGALL_CMD , ram_address); + +/* Issue an extended mode register set(EMRS2) to choose operation */ +atmel_mpddr_op (ATMEL_MPDDRC_MR_EXT_LMR_CMD , + ram_address + (0x2 << ba_off)); + +/* Issue an extended mode register set(EMRS3) to set EMSR to 0 */ +atmel_mpddr_op (ATMEL_MPDDRC_MR_EXT_LMR_CMD , + ram_address + (0x3 << ba_off)); + +/* +* Issue an extended mode register set(EMRS1) to enable DLL and +* program D.I.C (output driver impedance control) +*/ +atmel_mpddr_op (ATMEL_MPDDRC_MR_EXT_LMR_CMD , + ram_address + (0x1 << ba_off)); + +/* Enable DLL reset */ + cr =readl (&mpddr->cr); +writel (cr |ATMEL_MPDDRC_CR_EN_RESET_DLL , &mpddr->cr); + +/* A mode register set(MRS) cycle is issued to reset DLL */ +atmel_mpddr_op (ATMEL_MPDDRC_MR_LMR_CMD , ram_address); + +/* Issue an all banks precharge command */ +atmel_mpddr_op (ATMEL_MPDDRC_MR_PRCGALL_CMD , ram_address); + +/* Two auto-refresh (CBR) cycles are provided */ +atmel_mpddr_op (ATMEL_MPDDRC_MR_RFSH_CMD , ram_address); +atmel_mpddr_op (ATMEL_MPDDRC_MR_RFSH_CMD , ram_address); + +/* Disable DLL reset */ + cr =readl (&mpddr->cr);
Quoted 8 lines
arch/arm/cpu/at91-common/mpddrc.c line 91
Quoted 6 lines
arch/arm/cpu/at91-common/mpddrc.c new file +6
-0+writel (cr & (~ATMEL_MPDDRC_CR_EN_RESET_DLL ), &mpddr->cr); + +/* A mode register set (MRS) cycle is issued to disable DLL reset */ +atmel_mpddr_op (ATMEL_MPDDRC_MR_LMR_CMD , ram_address); + +/* Set OCD calibration in defautl state */ typo default
thanks.
Quoted 50 lines
Quoted 20 lines
arch/arm/cpu/at91-common/mpddrc.c new file +20
-0+ cr =readl (&mpddr->cr); +writel (cr |ATMEL_MPDDRC_CR_OCD_DEFAULT , &mpddr->cr); + +/* +* An extended mode register set (EMRS1) cycle is issued +* to OCD default value +*/ +atmel_mpddr_op (ATMEL_MPDDRC_MR_EXT_LMR_CMD , + ram_address + (0x1 << ba_off)); + +/* OCD calibration mode exit */ + cr =readl (&mpddr->cr); +writel (cr & (~ATMEL_MPDDRC_CR_OCD_DEFAULT ), &mpddr->cr); + +/* +* An extended mode register set (EMRS1) cycle is issued +* to enable OCD exit +*/ +atmel_mpddr_op (ATMEL_MPDDRC_MR_EXT_LMR_CMD , + ram_address + (0x1 << ba_off));arch/arm/cpu/at91-common/mpddrc.c line 121
Quoted 10 lines
arch/arm/cpu/at91-common/mpddrc.c new file +10
-0+ +/* A nornal mode command is provided */ +atmel_mpddr_op (ATMEL_MPDDRC_MR_NORMAL_CMD , ram_address); + +/* Perform a write access to any DDR2-SDRAM address */ +writel (0 , ram_address); + +/* Write the refresh rate */ +writel (mpddr_value->rtr, &mpddr->rtr); +I haven't checked that sequence deeply, I trust in you that it is correct ;)
Quoted 17 lines
arch/arm/cpu/at91-common/mpddrc.c new file +2
-0+return 0 ; +}arch/arm/include/asm/arch-at91/atmel_mpddrc.h new file +9
-0@@ -0,0 +1,111 @@ +/* +* Copyright (C) 2013 Atmel Corporation +* Bo Shen <[email protected]> +* +* SPDX-License-Identifier: GPL-2.0+ +*/ + +#ifndef __ATMEL_MPDDRC_H__ +#define __ATMEL_MPDDRC_H__
Quoted 8 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h line 15
Quoted 6 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h new file +6
-0+ +struct atmel_mpddr { +u32 mr; +u32 rtr; +u32 cr; +u32 tp0r;Datasheet names them tprX
Actually, this name is Timing Parameter 0 Register, Timing Parameter 1 Register.
Quoted 6 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h line 19
Quoted 4 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h new file +4
-0+u32 tp1r; +u32 tp2r; +u32 reserved[2 ]; +u32 mdr;Datasheet names it just md.
All other registers with "r" suffix, so, add r to this register name too.
Quoted 3 lines
I think it would be good to also add the other register positions or at least mention that these the only one needed currently (in a comment right here in the struct).
OK, I will add comments here.
Quoted 8 lines
Quoted 8 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h new file +8
-0+}; + +int ddr2_init (unsigned int ram_address, +struct atmel_mpddr *mpddr); + +/* bit field in mode register */ +#define ATMEL_MPDDRC_MR_NORMAL_CMD 0x0 +#define ATMEL_MPDDRC_MR_NOP_CMD 0x1
Quoted 8 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h line 33
Quoted 6 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h new file +6
-0+#define ATMEL_MPDDRC_MR_PRCGALL_CMD 0x2 +#define ATMEL_MPDDRC_MR_LMR_CMD 0x3 +#define ATMEL_MPDDRC_MR_RFSH_CMD 0x4 +#define ATMEL_MPDDRC_MR_EXT_LMR_CMD 0x5 +#define ATMEL_MPDDRC_MR_DEEP_CMD 0x6 +#define ATMEL_MPDDRC_MR_LPDDR2_CMD 0x7Could you please drop the tabs between 'define' and the definition.
OK
Quoted 36 lines
Quoted 17 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h new file +17
-0+ +/* bit field in configuration register */ +#define ATMEL_MPDDRC_CR_NC_MASK 0x3 +#define ATMEL_MPDDRC_CR_NC_COL_9 0x0 +#define ATMEL_MPDDRC_CR_NC_COL_10 0x1 +#define ATMEL_MPDDRC_CR_NC_COL_11 0x2 +#define ATMEL_MPDDRC_CR_NC_COL_12 0x3 +#define ATMEL_MPDDRC_CR_NR_MASK (0x3 << 2) +#define ATMEL_MPDDRC_CR_NR_ROW_11 (0x0 << 2) +#define ATMEL_MPDDRC_CR_NR_ROW_12 (0x1 << 2) +#define ATMEL_MPDDRC_CR_NR_ROW_13 (0x2 << 2) +#define ATMEL_MPDDRC_CR_NR_ROW_14 (0x3 << 2) +#define ATMEL_MPDDRC_CR_CAS (0x7 << 4) +#define ATMEL_MPDDRC_CR_CAS_2 (0x2 << 4) +#define ATMEL_MPDDRC_CR_CAS_3 (0x3 << 4) +#define ATMEL_MPDDRC_CR_CAS_4 (0x4 << 4) +#define ATMEL_MPDDRC_CR_CAS_5 (0x5 << 4)arch/arm/include/asm/arch-at91/atmel_mpddrc.h line 60
Quoted 10 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h new file +10
-0+#define ATMEL_MPDDRC_CR_CAS_6 (0x6 << 4) +#define ATMEL_MPDDRC_CR_EN_RESET_DLL (0x1 << 7) +#define ATMEL_MPDDRC_CR_DIC_DS (0x1 << 8) +#define ATMEL_MPDDRC_CR_DIS_DLL (0x1 << 9) +#define ATMEL_MPDDRC_CR_OCD_DEFAULT (0x7 << 12) +#define ATMEL_MPDDRC_CR_EN_ENRDM (0x1 << 17) +#define ATMEL_MPDDRC_CR_NB_8BANKS (0x1 << 20) +#define ATMEL_MPDDRC_CR_DIS_NDQS (0x1 << 21) +#define ATMEL_MPDDRC_CR_DECOD_INTERLEAVED (0x1 << 22) +#define ATMEL_MPDDRC_CR_UNAL_SUPPORTED (0x1 << 23)Some of the defines have the strict scheme <offset><register name><field name> with <offset> being
ATMEL_MPDDRC, <register name> CR here and the respective field names from datasheet. Some however have some more information likeUNAL_SUPPORTEDorDECOD_INTERLEAVED(those two are Ok I think, but we could discuss if we like to have the strict scheme or leave some space). ButEN_ENRDMis definitely too much ;) Has anyone a opinion about the strict scheme?
Quoted 1 line
Regarding
EN_ENRDMI think usingENRDM_ONwould be better
Agree.
Quoted 60 lines
Quoted 60 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h new file +51
-0+ +/* bit field in timing parameter 0 register */ +#define ATMEL_MPDDRC_TP0R_TRAS_OFFSET 0 +#define ATMEL_MPDDRC_TP0R_TRAS_MASK 0xf +#define ATMEL_MPDDRC_TP0R_TRCD_OFFSET 4 +#define ATMEL_MPDDRC_TP0R_TRCD_MASK 0xf +#define ATMEL_MPDDRC_TP0R_TWR_OFFSET 8 +#define ATMEL_MPDDRC_TP0R_TWR_MASK 0xf +#define ATMEL_MPDDRC_TP0R_TRC_OFFSET 12 +#define ATMEL_MPDDRC_TP0R_TRC_MASK 0xf +#define ATMEL_MPDDRC_TP0R_TRP_OFFSET 16 +#define ATMEL_MPDDRC_TP0R_TRP_MASK 0xf +#define ATMEL_MPDDRC_TP0R_TRRD_OFFSET 20 +#define ATMEL_MPDDRC_TP0R_TRRD_MASK 0xf +#define ATMEL_MPDDRC_TP0R_TWTR_OFFSET 24 +#define ATMEL_MPDDRC_TP0R_TWTR_MASK 0x7 +#define ATMEL_MPDDRC_TP0R_RDC_WRRD_OFFSET 27 +#define ATMEL_MPDDRC_TP0R_RDC_WRRD_MASK 0x1 +#define ATMEL_MPDDRC_TP0R_TMRD_OFFSET 28 +#define ATMEL_MPDDRC_TP0R_TMRD_MASK 0xf + +/* bit field in timing parameter 1 register */ +#define ATMEL_MPDDRC_TP1R_TRFC_OFFSET 0 +#define ATMEL_MPDDRC_TP1R_TRFC_MASK 0x7f +#define ATMEL_MPDDRC_TP1R_TXSNR_OFFSET 8 +#define ATMEL_MPDDRC_TP1R_TXSNR_MASK 0xff +#define ATMEL_MPDDRC_TP1R_TXSRD_OFFSET 16 +#define ATMEL_MPDDRC_TP1R_TXSRD_MASK 0xff +#define ATMEL_MPDDRC_TP1R_TXP_OFFSET 24 +#define ATMEL_MPDDRC_TP1R_TXP_MASK 0xf + +/* bit field in timing parameter 2 register */ +#define ATMEL_MPDDRC_TP2R_TXARD_OFFSET 0 +#define ATMEL_MPDDRC_TP2R_TXARD_MASK 0xf +#define ATMEL_MPDDRC_TP2R_TXARDS_OFFSET 4 +#define ATMEL_MPDDRC_TP2R_TXARDS_MASK 0xf +#define ATMEL_MPDDRC_TP2R_TRPA_OFFSET 8 +#define ATMEL_MPDDRC_TP2R_TRPA_MASK 0xf +#define ATMEL_MPDDRC_TP2R_TRTP_OFFSET 12 +#define ATMEL_MPDDRC_TP2R_TRTP_MASK 0x7 +#define ATMEL_MPDDRC_TP2R_TFAW_OFFSET 16 +#define ATMEL_MPDDRC_TP2R_TFAW_MASK 0xf + +/* bit field in Memory Device Register */ +#define ATMEL_MPDDRC_MDR_LPDDR_SDRAM 0x3 +#define ATMEL_MPDDRC_MDR_DDR2_SDRAM 0x6 +#define ATMEL_MPDDRC_MDR_DBW_MASK (0x1 << 4) +#define ATMEL_MPDDRC_MDR_DBW_32BITS (0x0 << 4) +#define ATMEL_MPDDRC_MDR_DBW_16BITS (0x1 << 4) + +#endif spl/Makefile +0
-0@@ -123,6 +123,10 @@ ifeq ($(SOC),exynos) LIBS-y += $(CPUDIR)/s5p-common/libs5p-common.o endif
Quoted 6 lines
spl/Makefile ifeq ($(SOC),exynos) · line 128
Quoted 3 lines
spl/Makefile +3
-0+ifeq ($(SOC),at91) +LIBS-y += arch/$(ARCH)/cpu/at91-common/libat91-common.o +endifThe libat91-common.o should be built in any case. The mpddrc.o should only be included for SPL build (that mentioned Heiko in another mail).
OK.
Quoted 10 lines
spl/Makefile ifeq ($(SOC),exynos) · line 132
Quoted 5 lines
spl/Makefile +1
-0+ # Add GCC lib ifeq ("$(USE_PRIVATE_LIBGCC)", "yes") PLATFORM_LIBGCC = $(SPLTREE)/arch/$(ARCH)/lib/libgcc.oBest regards
Andreas Bie?mann
Best Regards,
Bo Shen
PATCH v3 5/6 arm: atmel: add ddr2 initialization function
Hi Bo,
Quoted 2 lines · Bo Shen
Quoted 1 line · Andreas Bie?mann
On 11/06/2013 06:29 AM, Bo Shen wrote:
<snip>
Quoted 6 lines
Quoted 3 lines
Quoted 1 line
+static void
atmel_mpddr_op(int mode, u32 ram_address)static inline, could give the compiler a hint to optimize here.
I am not sure whether the
write(0, ram_address)will be optimized. Because this must excute later than write mode to let it work.
well, I don't mean reordering for optimization here.
The point is the atmel_mpddr_op will get two mostly static values (it is not fully true cause ram_address will be calculated, but didn't knew when writing). So when using the function we need to setup two registers, branch (and do all the stuff involved: saving to the stack a.s.o.) and write the register content to two different addresses.
If we instruct the compiler to inline atmel_mpddr_op() he could decide to just take the const value given into ddr2_init to be written to the relevant address. I think that is optimization too. It could however increase the text segment by some bytes also, that is to be checked (cause it would be getting worse and that would be no optimization).
Quoted 12 lines
Quoted 11 lines
arch/arm/cpu/at91-common/mpddrc.c line 20
Quoted 9 lines
(quoted diff) +2
-0+{ + struct atmel_mpddr *mpddr = (struct atmel_mpddr*)ATMEL_BASE_MPDDRC;
arch/arm/cpu/at91-common/mpddrc.c new file +6
-0+ +writel (mode, &mpddr->mr); +writel (0 , ram_address); +} + +int ddr2_init (u32 ram_address,struct atmel_mpddr *mpddr_value)could you please constify
mpddr_valuefor the very same reason?
Quoted 1 line
OK.
I think ram_address should also be declared const.
<snip>
Quoted 11 lines
Quoted 10 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h line 15
Quoted 8 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h new file +8
-0+#ifndef __ATMEL_MPDDRC_H__ +#define __ATMEL_MPDDRC_H__ + +struct atmel_mpddr { + u32 mr; + u32 rtr; + u32 cr; + u32 tp0r;Datasheet names them tprX
Quoted 2 lines
Actually, this name is Timing Parameter 0 Register, Timing Parameter 1 Register.
Well, I know. But my data sheet names it MPDDRC_TPR0 (the abbreviation in table 28-13). I think we should use the names in data sheet here. I often search the data sheet for the given name. mpddr will point to the correct subsystem in SoC, that is good. But tp0r will find nothing. This just consumes time when looking up the spec and is IMHO annoying.
Quoted 8 lines
Quoted 6 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h line 19
Quoted 4 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h new file +4
-0+u32 tp1r; +u32 tp2r; +u32 reserved[2 ]; +u32 mdr;Datasheet names it just md.
All other registers with "r" suffix, so, add r to this register name too.
But the name in data sheet is MPDDRC_MD ...
Best regards
Andreas Bie?mann
PATCH v3 5/6 arm: atmel: add ddr2 initialization function
Hi Andreas,
Quoted 22 lines · Andreas Bie?mann
Hi Bo,
Quoted 2 lines · Bo Shen
Quoted 1 line · Andreas Bie?mann
On 11/06/2013 06:29 AM, Bo Shen wrote:
<snip>
Quoted 6 lines
Quoted 3 lines
Quoted 1 line
+static void
atmel_mpddr_op(int mode, u32 ram_address)static inline, could give the compiler a hint to optimize here.
I am not sure whether the
write(0, ram_address)will be optimized. Because this must excute later than write mode to let it work.well, I don't mean reordering for optimization here.
The point is theatmel_mpddr_opwill get two mostly static values (it is not fully true causeram_addresswill be calculated, but didn't knew when writing). So when using the function we need to setup two registers, branch (and do all the stuff involved: saving to the stack a.s.o.) and write the register content to two different addresses.
Quoted 5 lines
If we instruct the compiler to inline
atmel_mpddr_op()he could decide to just take the const value given intoddr2_initto be written to the relevant address. I think that is optimization too. It could however increase the text segment by some bytes also, that is to be checked (cause it would be getting worse and that would be no optimization).
Got you means, ok, I will test it.
Quoted 43 lines
Quoted 12 lines
Quoted 11 lines
arch/arm/cpu/at91-common/mpddrc.c line 20
Quoted 9 lines
(quoted diff) +2
-0+{ + struct atmel_mpddr *mpddr = (struct atmel_mpddr*)ATMEL_BASE_MPDDRC;
arch/arm/cpu/at91-common/mpddrc.c new file +6
-0+ +writel (mode, &mpddr->mr); +writel (0 , ram_address); +} + +int ddr2_init (u32 ram_address,struct atmel_mpddr *mpddr_value)could you please constify
mpddr_valuefor the very same reason?Quoted 1 line
OK.
I think
ram_addressshould also be declared const.<snip>
Quoted 11 lines
Quoted 10 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h line 15
Quoted 8 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h new file +8
-0+#ifndef __ATMEL_MPDDRC_H__ +#define __ATMEL_MPDDRC_H__ + +struct atmel_mpddr { + u32 mr; + u32 rtr; + u32 cr; + u32 tp0r;Datasheet names them tprX
Quoted 2 lines
Actually, this name is Timing Parameter 0 Register, Timing Parameter 1 Register.
Well, I know. But my data sheet names it
MPDDRC_TPR0(the abbreviation in table 28-13). I think we should use the names in data sheet here. I often search the data sheet for the given name. mpddr will point to the correct subsystem in SoC, that is good. But tp0r will find nothing. This just consumes time when looking up the spec and is IMHO annoying.Quoted 5 lines
Quoted 5 lines
Quoted 4 lines
arch/arm/include/asm/arch-at91/atmel_mpddrc.h new file +4
-0+u32 tp1r; +u32 tp2r; +u32 reserved[2 ]; +u32 mdr;
Quoted 5 lines
Quoted 3 lines
Quoted 1 line
Datasheet names it just md.
All other registers with "r" suffix, so, add r to this register name too.
But the name in data sheet is
MPDDRC_MD...
OK, for register name, I will following up the table. Thanks.
Quoted 4 lines
Best regards
Andreas Bie?mann
Best Regards,
Bo Shen
PATCH v3 6/6 arm: atmel: sama5d3: spl boot from fat fs SD card
In reply to Bo ShenEnable Atmel sama5d3xek boart spl boot support, which can load u-boot from SD card with FAT file system.
---
Changes in v3:
- Move plla and mck configure to spl.c file
Changes in v2:
- Move spl related code to at91-common folder
arch/arm/cpu/armv7/Makefile | 2 +- arch/arm/cpu/at91-common/Makefile | 1 + arch/arm/cpu/at91-common/spl.c | 90 ++++++++++++++++++++++++++ arch/arm/cpu/at91-common/u-boot-spl.lds | 50 ++++++++++++++ arch/arm/include/asm/arch-at91/at91_common.h | 4 ++ arch/arm/include/asm/arch-at91/spl.h | 20 ++++++ board/atmel/sama5d3xek/sama5d3xek.c | 82 +++++++++++++++++++++++ include/configs/sama5d3xek.h | 34 ++++++++++ 8 files changed, 282 insertions(+), 1 deletion(-) create mode 100644 arch/arm/cpu/at91-common/spl.c create mode 100644 arch/arm/cpu/at91-common/u-boot-spl.lds create mode 100644 arch/arm/include/asm/arch-at91/spl.h
arch/arm/cpu/armv7/Makefile +1 -1
@@ -16,7 +16,7 @@ COBJS += cache_v7.o COBJS += cpu.o COBJS += syslib.o -ifneq ($(CONFIG_AM43XX)$(CONFIG_AM33XX)$(CONFIG_OMAP44XX)$(CONFIG_OMAP54XX)$(CONFIG_TEGRA)$(CONFIG_MX6)$(CONFIG_TI81XX),) +ifneq ($(CONFIG_AM43XX)$(CONFIG_AM33XX)$(CONFIG_OMAP44XX)$(CONFIG_OMAP54XX)$(CONFIG_TEGRA)$(CONFIG_MX6)$(CONFIG_TI81XX)$(CONFIG_AT91FAMILY),) SOBJS += lowlevel_init.o endif
arch/arm/cpu/at91-common/Makefile +1 -0
@@ -13,6 +13,7 @@ include $(TOPDIR)/config.mk LIB = $(obj)libat91-common.o COBJS-$(CONFIG_SPL_BUILD) += mpddrc.o +COBJS-$(CONFIG_SPL_BUILD) += spl.o SRCS := $(SOBJS-y:.o=.S) $(COBJS-y:.o=.c) OBJS := $(addprefix $(obj),$(SOBJS-y) $(COBJS-y))
arch/arm/cpu/at91-common/spl.c new file +90 -0
90 lines not on this page · raw message
arch/arm/cpu/at91-common/u-boot-spl.lds new file +50 -0
50 lines not on this page · raw message
arch/arm/include/asm/arch-at91/at91_common.h +4 -0
@@ -22,5 +22,9 @@ void at91_spi1_hw_init(unsigned long cs_mask);void at91_udp_hw_init (void );void at91_uhp_hw_init (void );void at91_lcd_hw_init (void ); +void at91_plla_init (u32 pllar); +void at91_mck_init (u32 mckr); +void at91_pmc_init (void ); +void mem_init (void );#endif /* AT91_COMMON_H */
arch/arm/include/asm/arch-at91/spl.h new file +20 -0
@@ -0,0 +1,20 @@ +/* +* Copyright (C) 2013 Atmel Corporation +* Bo Shen <[email protected]> +* +* SPDX-License-Identifier: GPL-2.0+ +*/ + +#ifndef _ASM_ARCH_SPL_H_ +#define _ASM_ARCH_SPL_H_ + +enum { +BOOT_DEVICE_NONE , +#ifdef CONFIG_SYS_USE_MMC +BOOT_DEVICE_MMC1 , +BOOT_DEVICE_MMC2 , +BOOT_DEVICE_MMC2_2 , +#endif +}; + +#endif
board/atmel/sama5d3xek/sama5d3xek.c +82 -0
91 lines not on this page · raw message
include/configs/sama5d3xek.h +34 -0
@@ -25,7 +25,10 @@#define CONFIG_AT91FAMILY #define CONFIG_ARCH_CPU_INIT +#ifndef CONFIG_SPL_BUILD #define CONFIG_SKIP_LOWLEVEL_INIT +#endif +#define CONFIG_BOARD_EARLY_INIT_F #define CONFIG_DISPLAY_CPUINFO
@@ -94,8 +97,12 @@#define CONFIG_SYS_SDRAM_BASE ATMEL_BASE_DDRCS#define CONFIG_SYS_SDRAM_SIZE 0x20000000 +#ifdef CONFIG_SPL_BUILD +#define CONFIG_SYS_INIT_SP_ADDR 0x310000 +#else #define CONFIG_SYS_INIT_SP_ADDR \ (CONFIG_SYS_SDRAM_BASE + 4 * 1024 - GENERATED_GBL_DATA_SIZE) +#endif /* SerialFlash */ #define CONFIG_CMD_SF
@@ -235,4 +242,31 @@/* Size of malloc() pool */ #define CONFIG_SYS_MALLOC_LEN (1024 * 1024) +/* SPL */ +#define CONFIG_SPL +#define CONFIG_SPL_FRAMEWORK +#define CONFIG_SPL_TEXT_BASE 0x300000 +#define CONFIG_SPL_MAX_SIZE 0x10000 +#define CONFIG_SPL_BSS_START_ADDR 0x20000000 +#define CONFIG_SPL_BSS_MAX_SIZE 0x80000 +#define CONFIG_SYS_SPL_MALLOC_START 0x20080000 +#define CONFIG_SYS_SPL_MALLOC_SIZE 0x80000 + +#define CONFIG_SPL_LIBCOMMON_SUPPORT +#define CONFIG_SPL_LIBGENERIC_SUPPORT +#define CONFIG_SPL_GPIO_SUPPORT +#define CONFIG_SPL_SERIAL_SUPPORT + +#define CONFIG_SPL_BOARD_INIT +#ifdef CONFIG_SYS_USE_MMC +#define CONFIG_SPL_LDSCRIPT arch/arm/cpu/at91-common/u-boot-spl.lds +#define CONFIG_SPL_MMC_SUPPORT +#define CONFIG_SYS_U_BOOT_MAX_SIZE_SECTORS 0x400 +#define CONFIG_SYS_MMCSD_RAW_MODE_U_BOOT_SECTOR 0x200 +#define CONFIG_SYS_MMC_SD_FAT_BOOT_PARTITION 1 +#define CONFIG_SPL_FAT_LOAD_PAYLOAD_NAME "u-boot.img" +#define CONFIG_SPL_FAT_SUPPORT +#define CONFIG_SPL_LIBDISK_SUPPORT +#endif +#endif
Signature · 1 line
1.7.9.5
Andreas Bießmann
NAKPATCH v3 6/6 arm: atmel: sama5d3: spl boot from fat fs SD card
Hi Bo,
Quoted 62 lines · Bo Shen
Enable Atmel sama5d3xek boart spl boot support, which can load u-boot from SD card with FAT file system.
Signed-off-by Bo Shen <[email protected]>---
Changes in v3:
- Move plla and mck configure tospl.cfileChanges in v2:
- Move spl related code to at91-common folderarch/arm/cpu/armv7/Makefile | 2 +- arch/arm/cpu/at91-common/Makefile | 1 + arch/arm/cpu/at91-common/spl.c | 90 ++++++++++++++++++++++++++ arch/arm/cpu/at91-common/u-boot-spl.lds | 50 ++++++++++++++ arch/arm/include/asm/arch-at91/at91_common.h | 4 ++ arch/arm/include/asm/arch-at91/spl.h | 20 ++++++ board/atmel/sama5d3xek/sama5d3xek.c | 82 +++++++++++++++++++++++ include/configs/sama5d3xek.h | 34 ++++++++++ 8 files changed, 282 insertions(+), 1 deletion(-) create mode 100644 arch/arm/cpu/at91-common/spl.c create mode 100644 arch/arm/cpu/at91-common/u-boot-spl.lds create mode 100644 arch/arm/include/asm/arch-at91/spl.harch/arm/cpu/armv7/Makefile +1
-1@@ -16,7 +16,7 @@ COBJS += cache_v7.o COBJS += cpu.o COBJS += syslib.o -ifneq ($(CONFIG_AM43XX)$(CONFIG_AM33XX)$(CONFIG_OMAP44XX)$(CONFIG_OMAP54XX)$(CONFIG_TEGRA)$(CONFIG_MX6)$(CONFIG_TI81XX),) +ifneq ($(CONFIG_AM43XX)$(CONFIG_AM33XX)$(CONFIG_OMAP44XX)$(CONFIG_OMAP54XX)$(CONFIG_TEGRA)$(CONFIG_MX6)$(CONFIG_TI81XX)$(CONFIG_AT91FAMILY),) SOBJS += lowlevel_init.o endifarch/arm/cpu/at91-common/Makefile +1
-0@@ -13,6 +13,7 @@ include $(TOPDIR)/config.mk LIB = $(obj)libat91-common.o COBJS-$(CONFIG_SPL_BUILD) += mpddrc.o +COBJS-$(CONFIG_SPL_BUILD) += spl.o SRCS := $(SOBJS-y:.o=.S) $(COBJS-y:.o=.c) OBJS := $(addprefix $(obj),$(SOBJS-y) $(COBJS-y))arch/arm/cpu/at91-common/spl.c new file +6
-0@@ -0,0 +1,90 @@ +/* +* Copyright (C) 2013 Atmel Corporation +* Bo Shen <[email protected]> +* +* SPDX-License-Identifier: GPL-2.0+ +*/
arch/arm/cpu/at91-common/spl.c line 16
Quoted 10 lines
arch/arm/cpu/at91-common/spl.c new file +10
-0+ +#include <common.h> +#include <asm/io.h> +#include <asm/arch/at91_common.h> +#include <asm/arch/at91_pmc.h> +#include <asm/arch/at91_wdt.h> +#include <asm/arch/clk.h> +#include <spl.h> + +static void at91_disable_wdt (void )
Why should we disable the WDT in SPL? I think it would be better to configure a working timer value than just disable it.
Well it's easy and works, but for the future I think it would be good to let it run while in SPL and u-boot.
arch/arm/cpu/at91-common/spl.c line 26
Quoted 10 lines
arch/arm/cpu/at91-common/spl.c new file +10
-0+{ +struct at91_wdt *wdt = (struct at91_wdt *)ATMEL_BASE_WDT ; + +writel (AT91_WDT_MR_WDDIS , &wdt->mr); +} + +void at91_plla_init (u32 pllar) +{ +struct at91_pmc *pmc = (struct at91_pmc *)ATMEL_BASE_PMC ; +
We should ensure bit 29 to be '1' here.
arch/arm/cpu/at91-common/spl.c line 29
Quoted 3 lines
arch/arm/cpu/at91-common/spl.c new file +3
-0+writel (pllar, &pmc->pllar); +while (!(readl (&pmc->sr) & (AT91_PMC_LOCKA |AT91_PMC_MCKRDY ))) + ;
Especially for doing such things it would be best handled by the WDT on error.
Quoted 4 lines
arch/arm/cpu/at91-common/spl.c new file +4
-0+} + +void at91_mck_init (u32 mckr) +{
arch/arm/cpu/at91-common/spl.c line 43
Quoted 10 lines
arch/arm/cpu/at91-common/spl.c new file +10
-0+struct at91_pmc *pmc = (struct at91_pmc *)ATMEL_BASE_PMC ; +u32 tmp; + + tmp =readl (&pmc->mckr); + tmp &= ~(AT91_PMC_MCKR_PRES_MASK | +AT91_PMC_MCKR_MDIV_MASK | +AT91_PMC_MCKR_PLLADIV_2 ); + tmp |= mckr & (AT91_PMC_MCKR_PRES_MASK | +AT91_PMC_MCKR_MDIV_MASK | +AT91_PMC_MCKR_PLLADIV_2 );
Why gets the at91_mck_init() just some parts of the MCK register (some fields are preserved here) while the at91_plla_init() just rewrites the PLLA register?
I think it is not much more than hiding the writel() register access. I think a better API would be to request some specific frequency and we calculate the register values with that input.
Please let us discuss this.
arch/arm/cpu/at91-common/spl.c line 55
Quoted 12 lines
arch/arm/cpu/at91-common/spl.c new file +12
-0+writel (tmp, &pmc->mckr); + +while (!(readl (&pmc->sr) &AT91_PMC_MCKRDY )) + ; +} + + +u32 spl_boot_device (void ) +{ +#ifdef CONFIG_SYS_USE_MMC +return BOOT_DEVICE_MMC1 ; +#endif
Isn't there some way to detect the boot source by asking for example the ROM code? Is there some register that holds that information?
Quoted 163 lines
arch/arm/cpu/at91-common/spl.c new file +35
-0+return BOOT_DEVICE_NONE ; +} + +u32 spl_boot_mode (void ) +{ +switch (spl_boot_device ()) { +#ifdef CONFIG_SYS_USE_MMC +case BOOT_DEVICE_MMC1 : +return MMCSD_MODE_FAT ; +break ; +#endif +case BOOT_DEVICE_NONE : +default : +hang (); + } +} + +void s_init (void ) +{ +/* disable watchdog */ +at91_disable_wdt (); + +/* PMC configuration */ +at91_pmc_init (); + +at91_clock_init (CONFIG_SYS_AT91_MAIN_CLOCK ); + +timer_init (); + +board_early_init_f (); + +preloader_console_init (); + +mem_init (); +}arch/arm/cpu/at91-common/u-boot-spl.lds new file +50
-0@@ -0,0 +1,50 @@ +/* + * (C) Copyright 2002 + * Gary Jennejohn, DENX Software Engineering, <[email protected]> + * + * (C) Copyright 2010 + * Texas Instruments, <www.ti.com> + * Aneesh V <[email protected]> + * + * (C) 2013 Atmel Corporation + * Bo Shen <[email protected]> + * + * SPDX-License-Identifier: GPL-2.0+ + */ + +MEMORY { .sram : ORIGIN = CONFIG_SPL_TEXT_BASE, \ + LENGTH = CONFIG_SPL_MAX_SIZE } +MEMORY { .sdram : ORIGIN = CONFIG_SPL_BSS_START_ADDR, \ + LENGTH = CONFIG_SPL_BSS_MAX_SIZE } + +OUTPUT_FORMAT("elf32-littlearm", "elf32-littlearm", "elf32-littlearm") +OUTPUT_ARCH(arm) +ENTRY(_start) +SECTIONS +{ + .text : + { + __start = .; + arch/arm/cpu/armv7/start.o (.text*) + *(.text*) + } >.sram + + . = ALIGN(4); + .rodata : { *(SORT_BY_ALIGNMENT(.rodata*)) } >.sram + + . = ALIGN(4); + .data : { *(SORT_BY_ALIGNMENT(.data*)) } >.sram + + . = ALIGN(4); + __image_copy_end = .; + _end = .; + + .bss : + { + . = ALIGN(4); + __bss_start = .; + *(.bss*) + . = ALIGN(4); + __bss_end = .; + } >.sdram +}arch/arm/include/asm/arch-at91/at91_common.h +4
-0@@ -22,5 +22,9 @@ void at91_spi1_hw_init(unsigned long cs_mask);void at91_udp_hw_init (void );void at91_uhp_hw_init (void );void at91_lcd_hw_init (void ); +void at91_plla_init (u32 pllar); +void at91_mck_init (u32 mckr); +void at91_pmc_init (void ); +void mem_init (void );#endif /* AT91_COMMON_H */ arch/arm/include/asm/arch-at91/spl.h new file +20
-0@@ -0,0 +1,20 @@ +/* +* Copyright (C) 2013 Atmel Corporation +* Bo Shen <[email protected]> +* +* SPDX-License-Identifier: GPL-2.0+ +*/ + +#ifndef _ASM_ARCH_SPL_H_ +#define _ASM_ARCH_SPL_H_ + +enum { +BOOT_DEVICE_NONE , +#ifdef CONFIG_SYS_USE_MMC +BOOT_DEVICE_MMC1 , +BOOT_DEVICE_MMC2 , +BOOT_DEVICE_MMC2_2 , +#endif +}; + +#endif board/atmel/sama5d3xek/sama5d3xek.c +17
-0@@ -20,6 +20,9 @@#include <micrel.h> #include <net.h> #include <netdev.h> +#include <spl.h> +#include <asm/arch/atmel_mpddrc.h> +#include <asm/arch/at91_wdt.h> #ifdef CONFIG_USB_GADGET_ATMEL_USBA #include <asm/arch/atmel_usba_udc.h> @@ -296,3 +299,82 @@ void spi_cs_deactivate(struct spi_slave *slave) } }#endif /* CONFIG_ATMEL_SPI */ + +/* SPL */ +#ifdef CONFIG_SPL_BUILD +void spl_board_init (void ) +{ +#ifdef CONFIG_SYS_USE_MMC +sama5d3xek_mci_hw_init (); +#endif +} + +void ddr2_conf (struct atmel_mpddr *ddr2) +{ + ddr2->mdr = (ATMEL_MPDDRC_MDR_DBW_32BITS |ATMEL_MPDDRC_MDR_DDR2_SDRAM ); +
board/atmel/sama5d3xek/sama5d3xek.c void spi_cs_deactivate(struct spi_slave *slave) · line 325
Quoted 10 lines
board/atmel/sama5d3xek/sama5d3xek.c +10
-0+ ddr2->cr = (ATMEL_MPDDRC_CR_NC_COL_10 | +ATMEL_MPDDRC_CR_NR_ROW_14 | +ATMEL_MPDDRC_CR_CAS_3 | +ATMEL_MPDDRC_CR_EN_ENRDM | +ATMEL_MPDDRC_CR_NB_8BANKS | +ATMEL_MPDDRC_CR_DIS_NDQS | +ATMEL_MPDDRC_CR_DECOD_INTERLEAVED | +ATMEL_MPDDRC_CR_UNAL_SUPPORTED ); + + ddr2->rtr =0x411 ;
Please document magic 0x411.
Quoted 11 lines
board/atmel/sama5d3xek/sama5d3xek.c +11
-0+ + ddr2->tp0r = (6 <<ATMEL_MPDDRC_TP0R_TRAS_OFFSET | +2 <<ATMEL_MPDDRC_TP0R_TRCD_OFFSET | +2 <<ATMEL_MPDDRC_TP0R_TWR_OFFSET | +8 <<ATMEL_MPDDRC_TP0R_TRC_OFFSET | +2 <<ATMEL_MPDDRC_TP0R_TRP_OFFSET | +2 <<ATMEL_MPDDRC_TP0R_TRRD_OFFSET | +2 <<ATMEL_MPDDRC_TP0R_TWTR_OFFSET | +2 <<ATMEL_MPDDRC_TP0R_TMRD_OFFSET ); + + ddr2->tp1r = (2 << ATMEL_MPDDRC_TP1R_TXP_OFFSET |
board/atmel/sama5d3xek/sama5d3xek.c void spi_cs_deactivate(struct spi_slave *slave) · line 346
Quoted 10 lines
board/atmel/sama5d3xek/sama5d3xek.c +10
-0+200 <<ATMEL_MPDDRC_TP1R_TXSRD_OFFSET | +28 <<ATMEL_MPDDRC_TP1R_TXSNR_OFFSET | + 26 << ATMEL_MPDDRC_TP1R_TRFC_OFFSET); + + ddr2->tp2r = (7 <<ATMEL_MPDDRC_TP2R_TFAW_OFFSET | +2 <<ATMEL_MPDDRC_TP2R_TRTP_OFFSET | +2 <<ATMEL_MPDDRC_TP2R_TRPA_OFFSET | +7 <<ATMEL_MPDDRC_TP2R_TXARDS_OFFSET | +8 <<ATMEL_MPDDRC_TP2R_TXARD_OFFSET ); +}
NAK, that is just copying data from ro section to bss section, please setup a static struct atmel_mpddr with static initialization. I can't see use cases where this won't work. You could have just more than one register setup and choose the correct one. Or modify a single instance before feeding it into ddr2_init().
Quoted 16 lines
board/atmel/sama5d3xek/sama5d3xek.c +16
-0+ +void mem_init (void ) +{ +struct at91_pmc *pmc = (struct at91_pmc *)ATMEL_BASE_PMC ; +struct atmel_mpddr ddr2; + +ddr2_conf (&ddr2); + +/* enable MPDDR clock */ +at91_periph_clk_enable (ATMEL_ID_MPDDRC ); +writel (0x4 , &pmc->scer); + +/* DDRAM2 Controller initialize */ +ddr2_init (ATMEL_BASE_DDRCS , &ddr2); +} +
board/atmel/sama5d3xek/sama5d3xek.c void spi_cs_deactivate(struct spi_slave *slave) · line 372
Quoted 10 lines
board/atmel/sama5d3xek/sama5d3xek.c +10
-0+void at91_pmc_init (void ) +{ +struct at91_pmc *pmc = (struct at91_pmc *)ATMEL_BASE_PMC ; +u32 tmp; + + tmp =AT91_PMC_PLLAR_29 | +AT91_PMC_PLLXR_PLLCOUNT (0x3f ) | +AT91_PMC_PLLXR_MUL (43 ) | +AT91_PMC_PLLXR_DIV (1 ); +at91_plla_init (tmp);
That could also be written with
writel(tmp, &pmc->pllar);
Ok, and wait for lock ... You know what I mean? As mentioned above it is just writing the register, not really much abstraction for that at91_plla_init() function.
But please do not remove that function. I think it worth to have it in the current setup.
But I also think we could do it better, for example the board just calls some function to setup 400MHz (and will then possibly use another frequency cause the given oscillators could not reach exactely that frequency).
board/atmel/sama5d3xek/sama5d3xek.c void spi_cs_deactivate(struct spi_slave *slave) · line 374
Quoted 2 lines
board/atmel/sama5d3xek/sama5d3xek.c +2
-0+ +writel (0x3 <<8 , &pmc->pllicpr);
Datasheet says PLLICPR->IPLL_PLLA should be written to '0'. Why do you write it to three? Is there something missing in the spec?
Quoted 57 lines
board/atmel/sama5d3xek/sama5d3xek.c +6
-0+ + tmp =AT91_PMC_MCKR_MDIV_4 | +AT91_PMC_MCKR_CSS_PLLA ; + at91_mck_init(tmp); +} +#endif include/configs/sama5d3xek.h +25
-0@@ -25,7 +25,10 @@#define CONFIG_AT91FAMILY #define CONFIG_ARCH_CPU_INIT +#ifndef CONFIG_SPL_BUILD #define CONFIG_SKIP_LOWLEVEL_INIT +#endif +#define CONFIG_BOARD_EARLY_INIT_F #define CONFIG_DISPLAY_CPUINFO @@ -94,8 +97,12 @@#define CONFIG_SYS_SDRAM_BASE ATMEL_BASE_DDRCS#define CONFIG_SYS_SDRAM_SIZE 0x20000000 +#ifdef CONFIG_SPL_BUILD +#define CONFIG_SYS_INIT_SP_ADDR 0x310000 +#else #define CONFIG_SYS_INIT_SP_ADDR \ (CONFIG_SYS_SDRAM_BASE + 4 * 1024 - GENERATED_GBL_DATA_SIZE) +#endif /* SerialFlash */ #define CONFIG_CMD_SF @@ -235,4 +242,31 @@/* Size of malloc() pool */ #define CONFIG_SYS_MALLOC_LEN (1024 * 1024) +/* SPL */ +#define CONFIG_SPL +#define CONFIG_SPL_FRAMEWORK +#define CONFIG_SPL_TEXT_BASE 0x300000 +#define CONFIG_SPL_MAX_SIZE 0x10000 +#define CONFIG_SPL_BSS_START_ADDR 0x20000000 +#define CONFIG_SPL_BSS_MAX_SIZE 0x80000 +#define CONFIG_SYS_SPL_MALLOC_START 0x20080000 +#define CONFIG_SYS_SPL_MALLOC_SIZE 0x80000 + +#define CONFIG_SPL_LIBCOMMON_SUPPORT +#define CONFIG_SPL_LIBGENERIC_SUPPORT +#define CONFIG_SPL_GPIO_SUPPORT +#define CONFIG_SPL_SERIAL_SUPPORT + +#define CONFIG_SPL_BOARD_INIT +#ifdef CONFIG_SYS_USE_MMC +#define CONFIG_SPL_LDSCRIPT arch/arm/cpu/at91-common/u-boot-spl.lds
include/configs/sama5d3xek.h line 272
Quoted 10 lines
include/configs/sama5d3xek.h +9
-0+#define CONFIG_SPL_MMC_SUPPORT +#define CONFIG_SYS_U_BOOT_MAX_SIZE_SECTORS 0x400 +#define CONFIG_SYS_MMCSD_RAW_MODE_U_BOOT_SECTOR 0x200 +#define CONFIG_SYS_MMC_SD_FAT_BOOT_PARTITION 1 +#define CONFIG_SPL_FAT_LOAD_PAYLOAD_NAME "u-boot.img" +#define CONFIG_SPL_FAT_SUPPORT +#define CONFIG_SPL_LIBDISK_SUPPORT +#endif +#endif
Looks good to me. The only thing I'd really like to change now is the ddr setup with a function, the rest should be discussed.
Best regards
Andreas Bie?mann
PATCH v3 6/6 arm: atmel: sama5d3: spl boot from fat fs SD card
Hello Andreas,
Am 13.11.2013 14:34, schrieb Andreas Bie?mann:
Quoted 27 lines
Hi Bo,
Quoted 24 lines · Bo Shen
Enable Atmel sama5d3xek boart spl boot support, which can load u-boot from SD card with FAT file system.
Signed-off-by Bo Shen<[email protected]>---
Changes in v3:
- Move plla and mck configure tospl.cfileChanges in v2:
- Move spl related code to at91-common folderarch/arm/cpu/armv7/Makefile | 2 +- arch/arm/cpu/at91-common/Makefile | 1 + arch/arm/cpu/at91-common/spl.c | 90 ++++++++++++++++++++++++++ arch/arm/cpu/at91-common/u-boot-spl.lds | 50 ++++++++++++++ arch/arm/include/asm/arch-at91/at91_common.h | 4 ++ arch/arm/include/asm/arch-at91/spl.h | 20 ++++++ board/atmel/sama5d3xek/sama5d3xek.c | 82 +++++++++++++++++++++++ include/configs/sama5d3xek.h | 34 ++++++++++ 8 files changed, 282 insertions(+), 1 deletion(-)create mode 100644
arch/arm/cpu/at91-common/spl.ccreate mode 100644arch/arm/cpu/at91-common/u-boot-spl.ldscreate mode 100644arch/arm/include/asm/arch-at91/spl.h
[...]
Quoted 17 lines
Quoted 17 lines
arch/arm/cpu/at91-common/spl.c new file +11
-0@@ -0,0 +1,90 @@ +/* +* Copyright (C) 2013 Atmel Corporation +* Bo Shen<[email protected]> +* +* SPDX-License-Identifier: GPL-2.0+ +*/ + +#include <common.h> +#include <asm/io.h> +#include <asm/arch/at91_common.h> +#include <asm/arch/at91_pmc.h>
Quoted 8 lines
arch/arm/cpu/at91-common/spl.c line 16
Quoted 5 lines
arch/arm/cpu/at91-common/spl.c new file +5
-0+#include <asm/arch/at91_wdt.h> +#include <asm/arch/clk.h> +#include <spl.h> + +static void at91_disable_wdt (void )Why should we disable the WDT in SPL? I think it would be better to configure a working timer value than just disable it.
This is currently done in the at91bootstrap code too...
Quoted 2 lines
Well it's easy and works, but for the future I think it would be good to let it run while in SPL and u-boot.
We should have the option to enable/disable it ...
Quoted 4 lines
Quoted 4 lines
arch/arm/cpu/at91-common/spl.c new file +4
-0+{ +struct at91_wdt *wdt = (struct at91_wdt *)ATMEL_BASE_WDT ; + +writel (AT91_WDT_MR_WDDIS ,&wdt->mr);
Quoted 8 lines
arch/arm/cpu/at91-common/spl.c line 26
Quoted 6 lines
arch/arm/cpu/at91-common/spl.c new file +6
-0+} + +void at91_plla_init (u32 pllar) +{ +struct at91_pmc *pmc = (struct at91_pmc *)ATMEL_BASE_PMC ; +We should ensure bit 29 to be '1' here.
What is bit 29 ?
Quoted 26 lines
arch/arm/cpu/at91-common/spl.c line 29
Quoted 3 lines
arch/arm/cpu/at91-common/spl.c new file +3
-0+writel (pllar,&pmc->pllar); +while (!(readl (&pmc->sr)& (AT91_PMC_LOCKA |AT91_PMC_MCKRDY ))) + ;Especially for doing such things it would be best handled by the WDT on error.
Quoted 4 lines
arch/arm/cpu/at91-common/spl.c new file +4
-0+} + +void at91_mck_init (u32 mckr) +{arch/arm/cpu/at91-common/spl.c line 43
Quoted 10 lines
arch/arm/cpu/at91-common/spl.c new file +10
-0+struct at91_pmc *pmc = (struct at91_pmc *)ATMEL_BASE_PMC ; +u32 tmp; + + tmp =readl (&pmc->mckr); + tmp&= ~(AT91_PMC_MCKR_PRES_MASK | +AT91_PMC_MCKR_MDIV_MASK | +AT91_PMC_MCKR_PLLADIV_2 ); + tmp |= mckr& (AT91_PMC_MCKR_PRES_MASK | +AT91_PMC_MCKR_MDIV_MASK | +AT91_PMC_MCKR_PLLADIV_2 );Why gets the
at91_mck_init()just some parts of the MCK register (some fields are preserved here) while theat91_plla_init()just rewrites the PLLA register?
Quoted 4 lines
I think it is not much more than hiding the
writel()register access. I think a better API would be to request some specific frequency and we calculate the register values with that input.
Please let us discuss this.
Such a function would be nice, indeed. But I would accept this function, to get SPL code into mainline... (I have the same function ;-)
Quoted 7 lines
Quoted 7 lines
arch/arm/cpu/at91-common/spl.c new file +7
-0+writel (tmp,&pmc->mckr); + +while (!(readl (&pmc->sr)&AT91_PMC_MCKRDY )) + ; +} + +
Quoted 8 lines
arch/arm/cpu/at91-common/spl.c line 55
Quoted 5 lines
arch/arm/cpu/at91-common/spl.c new file +5
-0+u32 spl_boot_device (void ) +{ +#ifdef CONFIG_SYS_USE_MMC +return BOOT_DEVICE_MMC1 ; +#endif Isn't there some way to detect the boot source by asking for example the ROM code? Is there some register that holds that information?
Yeah, that would be nice if we have such an information. Just got the information, that on the taurus board is also a SPI dataflash, from which it can boot. So I have the problem, that I have two boot sources, and if I want only one spl image, I have to detect from where SPL code got loaded ...
[...]
bye,
Heiko
Signature · 2 lines
DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
PATCH v3 6/6 arm: atmel: sama5d3: spl boot from fat fs SD card
Hello Heiko,
Quoted 10 lines · Heiko Schocher
Am 13.11.2013 14:34, schrieb Andreas Bie?mann:
Quoted 9 lines
Hi Bo,
Quoted 6 lines · Bo Shen
Enable Atmel sama5d3xek boart spl boot support, which can load u-boot from SD card with FAT file system.
Signed-off-by Bo Shen<[email protected]>---
<snip>
Quoted 6 lines
Quoted 4 lines
Quoted 1 line
+static void
at91_disable_wdt(void)Why should we disable the WDT in SPL? I think it would be better to configure a working timer value than just disable it.
This is currently done in the at91bootstrap code too...
I know ...
Quoted 5 lines
Quoted 3 lines
Well it's easy and works, but for the future I think it would be good to let it run while in SPL and u-boot.
We should have the option to enable/disable it ...
I just like to mention that this should be changed in near future.
Quoted 13 lines
Quoted 4 lines
Quoted 4 lines
arch/arm/cpu/at91-common/spl.c new file +4
-0+{ +struct at91_wdt *wdt = (struct at91_wdt *)ATMEL_BASE_WDT ; + +writel (AT91_WDT_MR_WDDIS ,&wdt->mr);Quoted 8 lines
arch/arm/cpu/at91-common/spl.c line 26
Quoted 6 lines
arch/arm/cpu/at91-common/spl.c new file +6
-0+} + +void at91_plla_init (u32 pllar) +{ +struct at91_pmc *pmc = (struct at91_pmc *)ATMEL_BASE_PMC ; +We should ensure bit 29 to be '1' here.
Quoted 1 line
What is bit 29 ?
'ONE' ;) Spec says it must be written to '1'
Quoted 17 lines
Quoted 17 lines
Review
Quoted 3 lines
(quoted diff) +3
-0+ writel(pllar,&pmc->pllar); + while (!(readl(&pmc->sr)& (AT91_PMC_LOCKA | AT91_PMC_MCKRDY))) + ;Especially for doing such things it would be best handled by the WDT on error.
Quoted 10 lines
arch/arm/cpu/at91-common/spl.c new file +10
-0+} + +void at91_mck_init (u32 mckr) +{ +struct at91_pmc *pmc = (struct at91_pmc *)ATMEL_BASE_PMC ; +u32 tmp; + + tmp =readl (&pmc->mckr); + tmp&= ~(AT91_PMC_MCKR_PRES_MASK | + AT91_PMC_MCKR_MDIV_MASK |
Quoted 8 lines
Quoted 8 lines
arch/arm/cpu/at91-common/spl.c line 43
Quoted 4 lines
arch/arm/cpu/at91-common/spl.c new file +4
-0+ AT91_PMC_MCKR_PLLADIV_2); + tmp |= mckr& (AT91_PMC_MCKR_PRES_MASK | +AT91_PMC_MCKR_MDIV_MASK | +AT91_PMC_MCKR_PLLADIV_2 );Why gets the
at91_mck_init()just some parts of the MCK register (some fields are preserved here) while theat91_plla_init()just rewrites the PLLA register?
Don't you see the error in my sentence here? ;) I thought some parts of the register content where preserved. The other way round is true, we need to clean up relevant parts (PRES, MDIV, and PLLADIV Bit(s)). Sorry, my fault here.
Quoted 7 lines
Quoted 4 lines
I think it is not much more than hiding the
writel()register access. I think a better API would be to request some specific frequency and we calculate the register values with that input.
Please let us discuss this.Such a function would be nice, indeed. But I would accept this function, to get SPL code into mainline... (I have the same function ;-)
Yea, no problem with this function. I just wonder why the at91_plla_init() is a plain writel() (plus waiting for lock) and this one does a bit more (preserve CSS... We could also declare the whole register content when calling at91_mckr_init(), don't we?
This is no change request in general, but please lets discuss why we not just write the given register content.
Best regards
Andreas Bie?mann
PATCH v3 6/6 arm: atmel: sama5d3: spl boot from fat fs SD card
Hi Andreas,
Quoted 16 lines · Andreas Bie?mann
Hello Heiko,
Quoted 10 lines · Heiko Schocher
Am 13.11.2013 14:34, schrieb Andreas Bie?mann:
Quoted 9 lines
Hi Bo,
Quoted 6 lines · Bo Shen
Enable Atmel sama5d3xek boart spl boot support, which can load u-boot from SD card with FAT file system.
Signed-off-by Bo Shen<[email protected]>---
<snip>
Quoted 13 lines
Quoted 6 lines
Quoted 4 lines
Quoted 1 line
+static void
at91_disable_wdt(void)Why should we disable the WDT in SPL? I think it would be better to configure a working timer value than just disable it.
This is currently done in the at91bootstrap code too...
I know ...
Quoted 5 lines
Quoted 3 lines
Well it's easy and works, but for the future I think it would be good to let it run while in SPL and u-boot.
We should have the option to enable/disable it ...
The watchdog is one time configurable. If configurabled, can not change anymore, so we disable it by default.
Quoted 15 lines
I just like to mention that this should be changed in near future.
Quoted 13 lines
Quoted 12 lines
arch/arm/cpu/at91-common/spl.c line 26
Quoted 10 lines
arch/arm/cpu/at91-common/spl.c new file +10
-0+{ +struct at91_wdt *wdt = (struct at91_wdt *)ATMEL_BASE_WDT ; + +writel (AT91_WDT_MR_WDDIS ,&wdt->mr); +} + +void at91_plla_init (u32 pllar) +{ +struct at91_pmc *pmc = (struct at91_pmc *)ATMEL_BASE_PMC ; +We should ensure bit 29 to be '1' here.
Quoted 3 lines
Quoted 1 line
What is bit 29 ?
'ONE' ;) Spec says it must be written to '1'
OK, I will add a check here.
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Review
Quoted 3 lines
(quoted diff) +3
-0+ writel(pllar,&pmc->pllar); + while (!(readl(&pmc->sr)& (AT91_PMC_LOCKA | AT91_PMC_MCKRDY))) + ;Especially for doing such things it would be best handled by the WDT on error.
As watchdog is disabled, we can not use it here.
Quoted 19 lines
Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
arch/arm/cpu/at91-common/spl.c new file +10
-0+} + +void at91_mck_init (u32 mckr) +{ +struct at91_pmc *pmc = (struct at91_pmc *)ATMEL_BASE_PMC ; +u32 tmp; + + tmp =readl (&pmc->mckr); + tmp&= ~(AT91_PMC_MCKR_PRES_MASK | + AT91_PMC_MCKR_MDIV_MASK |Quoted 8 lines
Quoted 8 lines
arch/arm/cpu/at91-common/spl.c line 43
Quoted 4 lines
arch/arm/cpu/at91-common/spl.c new file +4
-0+ AT91_PMC_MCKR_PLLADIV_2); + tmp |= mckr& (AT91_PMC_MCKR_PRES_MASK | +AT91_PMC_MCKR_MDIV_MASK | +AT91_PMC_MCKR_PLLADIV_2 );Why gets the
at91_mck_init()just some parts of the MCK register (some fields are preserved here) while theat91_plla_init()just rewrites the PLLA register?
Quoted 4 lines
Don't you see the error in my sentence here? ;) I thought some parts of the register content where preserved. The other way round is true, we need to clean up relevant parts (PRES, MDIV, and PLLADIV Bit(s)). Sorry, my fault here.
So, keep it as is(?)
Quoted 5 lines
Quoted 5 lines
Quoted 4 lines
I think it is not much more than hiding the
writel()register access. I think a better API would be to request some specific frequency and we calculate the register values with that input.
Please let us discuss this.
Quoted 7 lines
Quoted 2 lines
Such a function would be nice, indeed. But I would accept this function, to get SPL code into mainline... (I have the same function ;-)
Yea, no problem with this function. I just wonder why the
at91_plla_init()is a plainwritel()(plus waiting for lock) and this one does a bit more (preserve CSS... We could also declare the whole register content when callingat91_mckr_init(), don't we?
at91_plla_init()'s main purpose is to wait the lock.
Quoted 2 lines
This is no change request in general, but please lets discuss why we not just write the given register content.
This make we focus on what we need to configure.
Best Regards,
Bo Shen
PATCH v3 6/6 arm: atmel: sama5d3: spl boot from fat fs SD card
Hi Bo,
Quoted 24 lines · Bo Shen
Quoted 23 lines · Andreas Bie?mann
Quoted 10 lines · Heiko Schocher
Am 13.11.2013 14:34, schrieb Andreas Bie?mann:
Quoted 9 lines
Hi Bo,
Quoted 6 lines · Bo Shen
Enable Atmel sama5d3xek boart spl boot support, which can load u-boot from SD card with FAT file system.
Signed-off-by Bo Shen<[email protected]>---
<snip>
Quoted 6 lines
Quoted 4 lines
Quoted 1 line
+static void
at91_disable_wdt(void)Why should we disable the WDT in SPL? I think it would be better to configure a working timer value than just disable it.
This is currently done in the at91bootstrap code too...
I know ...
Quoted 1 line
Quoted 8 lines
Quoted 5 lines
Quoted 5 lines
Quoted 3 lines
Well it's easy and works, but for the future I think it would be good to
let it run while in SPL and u-boot.We should have the option to enable/disable it ...
The watchdog is one time configurable. If configurabled, can not change anymore, so we disable it by default.
But we should give an easy way to enable it. I swear industrial users require this, so we should at least add a possibility to enable it easily. Take it as a comment, no need to change it for that patch.
Quoted 17 lines
Quoted 17 lines
Quoted 16 lines
Quoted 16 lines
arch/arm/cpu/at91-common/spl.c line 43
Quoted 12 lines
arch/arm/cpu/at91-common/spl.c new file +12
-0+void at91_mck_init (u32 mckr) +{ +struct at91_pmc *pmc = (struct at91_pmc *)ATMEL_BASE_PMC ; +u32 tmp; + + tmp =readl (&pmc->mckr); + tmp&= ~(AT91_PMC_MCKR_PRES_MASK | +AT91_PMC_MCKR_MDIV_MASK | +AT91_PMC_MCKR_PLLADIV_2 ); + tmp |= mckr& (AT91_PMC_MCKR_PRES_MASK | +AT91_PMC_MCKR_MDIV_MASK | +AT91_PMC_MCKR_PLLADIV_2 );Why gets the
at91_mck_init()just some parts of the MCK register (some fields are preserved here) while theat91_plla_init()just rewrites the PLLA register?
Quoted 6 lines
Quoted 4 lines
Don't you see the error in my sentence here? ;) I thought some parts of the register content where preserved. The other way round is true, we need to clean up relevant parts (PRES, MDIV, and PLLADIV Bit(s)). Sorry, my fault here.
So, keep it as is(?)
For now yes ... we need to think about a better solution in the future.
Best regards
Andreas Bie?mann