Skip to contentSkip to conversation

ARM: am335x: Add phyCORE AM335x R2 support

PATCH Superseded Niel Fourie 3 participants 9 messages May 6, 2019 – May 9, 2019

Subsystems (3)

v5 of 6 Superseded Latest is v6 →

lore
Share & tools
lore
#0 lore raw

PATCH v5 1/1 ARM: am335x: Add phyCORE AM335x R2 support

Support for Phytech phyCORE AM335x R2 SOM (PCL060) on the Phytec phyBOARD-Wega AM335x.

CPU  : AM335X-GP rev 2.1
Model: Phytec AM335x phyBOARD-WEGA
DRAM:  256 MiB
NAND:  256 MiB
MMC:   OMAP SD/MMC: 0
eth0: ethernet at 4a100000

Working:
- Eth0
- i2C
- MMC/SD
- NAND
- UART
- USB (host)

Device trees were taken from Linux mainline:
commit 37624b58542f ("Linux 5.1-rc7")

---
Changes for v2:
- Remove formatting changes to upstream Linux dtsi files
- Remove incorrectly added MACH_TYPE
- Rename board from phycore_pcl060 to phycore_am335x_r2
- Implement selecting memory size from device tree
- Remove non-DM Ethernet board code
- General clean-up

Changes for v3:
- Added kernel revision of upstream Linux dtsi files
- Place TARGET_PHYCORE_AM335X_R2 alphabetically in Kconfig
- Rework TPS65910 I2C init code
- Remove non-DM USB board macros
- Minor tweaks

Changes for v4:
- Propose abstracting common sections to am33xx-u-boot.dtsi
- Remove dead Falcon mode code
- Remove non-DM I2C support for TPS65910 I2C

Changes for v5:
- Revert proposed moving of sections to am33xx-u-boot.dtsi
- Remove redundant/incorrect lines from am33xx-u-boot.dtsi
- Add missing changelog

 arch/arm/dts/Makefile                      |   3 +-
 arch/arm/dts/am335x-phycore-som.dtsi       | 322 +++++++++++++++++++++
 arch/arm/dts/am335x-wega-rdk-u-boot.dtsi   |  31 ++
 arch/arm/dts/am335x-wega-rdk.dts           |  23 ++
 arch/arm/dts/am335x-wega.dtsi              | 230 +++++++++++++++
 arch/arm/mach-omap2/Kconfig                |   1 +
 arch/arm/mach-omap2/am33xx/Kconfig         |   7 +
 board/phytec/phycore_am335x_r2/Kconfig     |  15 +
 board/phytec/phycore_am335x_r2/MAINTAINERS |   7 +
 board/phytec/phycore_am335x_r2/Makefile    |  11 +
 board/phytec/phycore_am335x_r2/board.c     | 252 ++++++++++++++++
 board/phytec/phycore_am335x_r2/board.h     |  24 ++
 board/phytec/phycore_am335x_r2/mux.c       | 117 ++++++++
 configs/phycore-am335x-r2-wega_defconfig   |  79 +++++
 include/configs/phycore_am335x_r2.h        | 130 +++++++++
 15 files changed, 1251 insertions(+), 1 deletion(-)
 create mode 100644 arch/arm/dts/am335x-phycore-som.dtsi
 create mode 100644 arch/arm/dts/am335x-wega-rdk-u-boot.dtsi
 create mode 100644 arch/arm/dts/am335x-wega-rdk.dts
 create mode 100644 arch/arm/dts/am335x-wega.dtsi
 create mode 100644 board/phytec/phycore_am335x_r2/Kconfig
 create mode 100644 board/phytec/phycore_am335x_r2/MAINTAINERS
 create mode 100644 board/phytec/phycore_am335x_r2/Makefile
 create mode 100644 board/phytec/phycore_am335x_r2/board.c
 create mode 100644 board/phytec/phycore_am335x_r2/board.h
 create mode 100644 board/phytec/phycore_am335x_r2/mux.c
 create mode 100644 configs/phycore-am335x-r2-wega_defconfig
 create mode 100644 include/configs/phycore_am335x_r2.h
arch/arm/dts/Makefile +2 -1
index 8e082f2840..381a64ae13 100644
@@ -264,7 +264,8 @@ dtb-$(CONFIG_AM33XX) += \
 	am335x-chiliboard.dtb \
 	am335x-sl50.dtb \
 	am335x-base0033.dtb \
-	am335x-guardian.dtb
+	am335x-guardian.dtb \
+	am335x-wega-rdk.dtb
 dtb-$(CONFIG_AM43XX) += am437x-gp-evm.dtb am437x-sk-evm.dtb	\
 	am43x-epos-evm.dtb \
 	am437x-idk-evm.dtb \
arch/arm/dts/am335x-phycore-som.dtsi new file +322 -0

322 lines not on this page · raw message

arch/arm/dts/am335x-wega-rdk-u-boot.dtsi new file +31 -0
new file mode 100644
index 0000000000..36b761b7bf
@@ -0,0 +1,31 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2019 DENX Software Engineering GmbH
+ */
+
+/ {
+	chosen {
+		#address-cells = <1>;
+		#size-cells = <1>;
+
+		bootargs = "console=ttyO0,115200 earlyprintk";
+		stdout-path = &uart0;
+	};
+
+	ocp {
+		u-boot,dm-pre-reloc;
+	};
+
+	memory at 80000000 {
+		u-boot,dm-pre-reloc;
+	};
+};
+
+&uart0 {
+	u-boot,dm-pre-reloc;
+};
+
+&mmc1 {
+	u-boot,dm-pre-reloc;
+	cd-gpios = <&gpio0 6 GPIO_ACTIVE_LOW>;
+};
arch/arm/dts/am335x-wega-rdk.dts new file +23 -0
new file mode 100644
index 0000000000..fe50f3041a
@@ -0,0 +1,23 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2015 Phytec Messtechnik GmbH
+ * Author: Teresa Remmet <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+/dts-v1/;
+
+#include "am335x-phycore-som.dtsi"
+#include "am335x-wega.dtsi"
+
+/* SoM */
+&i2c_eeprom {
+	status = "okay";
+};
+
+&i2c_rtc {
+	status = "okay";
+};
arch/arm/dts/am335x-wega.dtsi new file +230 -0

230 lines not on this page · raw message

arch/arm/mach-omap2/Kconfig +1 -0
index d29f1ca0b5..3fa9dc89b8 100644
@@ -186,6 +186,7 @@ source "board/ti/am43xx/Kconfig"
 source "board/ti/am335x/Kconfig"
 source "board/compulab/cm_t335/Kconfig"
 source "board/compulab/cm_t43/Kconfig"
+source "board/phytec/phycore_am335x_r2/Kconfig"
 
 config SPL_LDSCRIPT
         default "arch/arm/mach-omap2/u-boot-spl.lds"
arch/arm/mach-omap2/am33xx/Kconfig +7 -0
index 500df1aa11..9da2a16a99 100644
@@ -177,6 +177,13 @@ config TARGET_PEPPER
 	select DM_SERIAL
 	imply CMD_DM
 
+config TARGET_PHYCORE_AM335X_R2
+	bool "Support phyCORE AM335X R2"
+	select DM
+	select DM_GPIO
+	select DM_SERIAL
+	imply CMD_DM
+
 config TARGET_PXM2
 	bool "Support pxm2"
 	select BOARD_LATE_INIT
board/phytec/phycore_am335x_r2/Kconfig new file +15 -0
new file mode 100644
index 0000000000..77055e043c
@@ -0,0 +1,15 @@
+if TARGET_PHYCORE_AM335X_R2
+
+config SYS_BOARD
+	default "phycore_am335x_r2"
+
+config SYS_VENDOR
+	default "phytec"
+
+config SYS_SOC
+	default "am33xx"
+
+config SYS_CONFIG_NAME
+	default "phycore_am335x_r2"
+
+endif
board/phytec/phycore_am335x_r2/MAINTAINERS new file +7 -0
new file mode 100644
index 0000000000..e56f30fdc0
@@ -0,0 +1,7 @@
+phyCORE AM335x R2 WEGA BOARD
+M:	Niel Fourie <[email protected]>
+M:	Parthiban Nallathambi <[email protected]>
+S:	Maintained
+F:	board/phytec/phycore_am335x_r2
+F:	include/configs/phycore_am335x_r2.h
+F:	configs/phycore-am335x-r2-wega_defconfig
board/phytec/phycore_am335x_r2/Makefile new file +11 -0
new file mode 100644
index 0000000000..ff6f8b4221
@@ -0,0 +1,11 @@
+# SPDX-License-Identifier: GPL-2.0+
+#
+# Makefile
+#
+# Copyright (C) 2011 Texas Instruments Incorporated - http://www.ti.com/
+
+ifdef CONFIG_SPL_BUILD
+obj-y	+= mux.o
+endif
+
+obj-y	+= board.o
board/phytec/phycore_am335x_r2/board.c new file +252 -0

252 lines not on this page · raw message

board/phytec/phycore_am335x_r2/board.h new file +24 -0
new file mode 100644
index 0000000000..68c3d56aa5
@@ -0,0 +1,24 @@
+/* SPDX-License-Identifier: GPL-2.0+ */
+/*
+ * board.h
+ *
+ * Phytec phyCORE-AM335x (pcl060) boards information header
+ *
+ * Copyright (C) 2013 Lars Poeschel, Lemonage Software GmbH
+ * Copyright (C) 2019 DENX Software Engineering GmbH
+ */
+
+#ifndef _BOARD_H_
+#define _BOARD_H_
+
+/*
+ * We have three pin mux functions that must exist.  We must be able to enable
+ * uart0, for initial output and i2c0 to access the PMIC. We then have a main
+ * pinmux function that can be overridden to enable all other pinmux that
+ * is required on the board.
+ */
+void enable_uart0_pin_mux(void);
+void enable_i2c0_pin_mux(void);
+void enable_board_pin_mux(void);
+void enable_cbmux_pin_mux(void);
+#endif
board/phytec/phycore_am335x_r2/mux.c new file +117 -0

117 lines not on this page · raw message

configs/phycore-am335x-r2-wega_defconfig new file +79 -0

79 lines not on this page · raw message

include/configs/phycore_am335x_r2.h new file +130 -0

130 lines not on this page · raw message

Signature · 1 line
2.21.0
#1 lore raw

PATCH v5 1/1 ARM: am335x: Add phyCORE AM335x R2 support

Quoted 22 lines · Niel Fourie

Support for Phytech phyCORE AM335x R2 SOM (PCL060) on the Phytec phyBOARD-Wega AM335x.

CPU  : AM335X-GP rev 2.1
Model: Phytec AM335x phyBOARD-WEGA
DRAM:  256 MiB
NAND:  256 MiB
MMC:   OMAP SD/MMC: 0
eth0: ethernet at 4a100000

Working:
- Eth0
- i2C
- MMC/SD
- NAND
- UART
- USB (host)

Device trees were taken from Linux mainline:
commit 37624b58542f ("Linux 5.1-rc7")

[snip]

arch/arm/dts/am335x-wega-rdk-u-boot.dtsi line 21

Quoted 3 lines
arch/arm/dts/am335x-wega-rdk-u-boot.dtsi new file +3 -0
+	memory at 80000000 {
+		u-boot,dm-pre-reloc;
+	};

You don't need this because:

[snip]

Quoted 15 lines
board/phytec/phycore_am335x_r2/board.c new file +15 -0
+void sdram_init(void)
+{
+	int ram_type_index = PHYCORE_R2_MT41K128M16JT_256MB;
+
+	if (fdtdec_setup_mem_size_base())
+		gd->ram_size = SZ_256M;
+
+	switch (gd->ram_size) {
+	case SZ_1G:
+		ram_type_index = PHYCORE_R2_MT41K512M16HA125IT_1024MB;
+		break;
+	case SZ_512M:
+		ram_type_index = PHYCORE_R2_MT41K256M16TW107IT_512MB;
+		break;
+	case SZ_256M:

board/phytec/phycore_am335x_r2/board.c line 155

Quoted 10 lines
board/phytec/phycore_am335x_r2/board.c new file +10 -0
+	default:
+		ram_type_index = PHYCORE_R2_MT41K128M16JT_256MB;
+		break;
+	}
+
+	config_ddr(DDR_CLK_MHZ, &ioregs,
+		   &physom_timings[ram_type_index].ddr3_data,
+		   &ddr3_cmd_ctrl_data,
+		   &physom_timings[ram_type_index].ddr3_emif_reg_data, 0);
+}

This is wrong. sdram_init() is called by arch/arm/mach-omap2/am33xx/board.c::dram_init() which then sets gd->ram_size based on what get_ram_size() determines. So this is all just a wrapper around how the various parts of the am33xx generations call some form of config_ddr(). And what you have here is a lot of unused code about which module provides how much memory. I assume there's some run-time method to determine which module you're on and thus determine that correct parameters to pass in for the chip that's in use. If you're not there yet then just make sdram_init() call config_ddr(...) with the correct enum for the 256M chip and then update this when you have real detection.

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190506/ec53b7d7/attachment.sig>
#2 lore raw

PATCH v5 1/1 ARM: am335x: Add phyCORE AM335x R2 support

Hi Tom,

Quoted 53 lines · Tom Rini
Quoted 37 lines · Niel Fourie

Support for Phytech phyCORE AM335x R2 SOM (PCL060) on the Phytec phyBOARD-Wega AM335x.

CPU  : AM335X-GP rev 2.1
Model: Phytec AM335x phyBOARD-WEGA
DRAM:  256 MiB
NAND:  256 MiB
MMC:   OMAP SD/MMC: 0
eth0: ethernet at 4a100000

Working:
- Eth0
- i2C
- MMC/SD
- NAND
- UART
- USB (host)

Device trees were taken from Linux mainline:
commit 37624b58542f ("Linux 5.1-rc7")

board/phytec/phycore_am335x_r2/board.c new file +15 -0
+void sdram_init(void)
+{
+	int ram_type_index = PHYCORE_R2_MT41K128M16JT_256MB;
+
+	if (fdtdec_setup_mem_size_base())
+		gd->ram_size = SZ_256M;
+
+	switch (gd->ram_size) {
+	case SZ_1G:
+		ram_type_index = PHYCORE_R2_MT41K512M16HA125IT_1024MB;
+		break;
+	case SZ_512M:
+		ram_type_index = PHYCORE_R2_MT41K256M16TW107IT_512MB;
+		break;
+	case SZ_256M:

board/phytec/phycore_am335x_r2/board.c line 155

Quoted 10 lines
board/phytec/phycore_am335x_r2/board.c new file +10 -0
+	default:
+		ram_type_index = PHYCORE_R2_MT41K128M16JT_256MB;
+		break;
+	}
+
+	config_ddr(DDR_CLK_MHZ, &ioregs,
+		   &physom_timings[ram_type_index].ddr3_data,
+		   &ddr3_cmd_ctrl_data,
+		   &physom_timings[ram_type_index].ddr3_emif_reg_data, 0);
+}

This is wrong. sdram_init() is called by arch/arm/mach-omap2/am33xx/board.c::dram_init() which then sets gd->ram_size based on what get_ram_size() determines. So this is all

Quoted 8 lines

just a wrapper around how the various parts of the am33xx generations call some form of config_ddr(). And what you have here is a lot of unused code about which module provides how much memory. I assume there's some run-time method to determine which module you're on and thus determine that correct parameters to pass in for the chip that's in use. If you're not there yet then just make sdram_init() call config_ddr(...) with the correct enum for the 256M chip and then update this when you have real detection.

Thanks for that input, you are right. I could not find any documented way to detect the exact module we are running on, but as you pointed out we can use get_ram_size() to find the size of the installed RAM. This is in fact exactly what barebox did, I just missed it. How is this for a replacement of the above?

void sdram_init(void)

{
	/* Configure memory to maximum supported size for detection */
	int ram_type_index = PHYCORE_R2_MT41K512M16HA125IT_1024MB;
	config_ddr(DDR_CLK_MHZ, &ioregs,
		   &physom_timings[ram_type_index].ddr3_data,
		   &ddr3_cmd_ctrl_data,
		   &physom_timings[ram_type_index].ddr3_emif_reg_data,
		   0);

	/* Detect memory physically present */
	gd->ram_size = get_ram_size((void *)CONFIG_SYS_SDRAM_BASE,
				    CONFIG_MAX_RAM_BANK_SIZE);

	/* Reconfigure memory for actual detected size */
	switch (gd->ram_size) {
	case SZ_1G:
		ram_type_index = PHYCORE_R2_MT41K512M16HA125IT_1024MB;
		break;
	case SZ_512M:
		ram_type_index = PHYCORE_R2_MT41K256M16TW107IT_512MB;
		break;
	case SZ_256M:
	default:
		ram_type_index = PHYCORE_R2_MT41K128M16JT_256MB;
		break;
	}
	config_ddr(DDR_CLK_MHZ, &ioregs,
		   &physom_timings[ram_type_index].ddr3_data,
		   &ddr3_cmd_ctrl_data,
		   &physom_timings[ram_type_index].ddr3_emif_reg_data,
		   0);
}

The ugliest part of this is, as you pointed out, that directly after this is called, get_ram_size() will be called again from sdram_init(). But it at least noninvasive, and no longer requires the device tree.

Best regards,
Niel Fourie

Signature · 3 lines
DENX Software Engineering GmbH,      Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: +49-8142-66989-21 Fax: +49-8142-66989-80  Email: lusus at denx.de
#3 lore raw

PATCH v5 1/1 ARM: am335x: Add phyCORE AM335x R2 support

Quoted 105 lines · Niel Fourie

Hi Tom,

Quoted 53 lines · Tom Rini
Quoted 37 lines · Niel Fourie

Support for Phytech phyCORE AM335x R2 SOM (PCL060) on the Phytec phyBOARD-Wega AM335x.

CPU  : AM335X-GP rev 2.1
Model: Phytec AM335x phyBOARD-WEGA
DRAM:  256 MiB
NAND:  256 MiB
MMC:   OMAP SD/MMC: 0
eth0: ethernet at 4a100000

Working:
- Eth0
- i2C
- MMC/SD
- NAND
- UART
- USB (host)

Device trees were taken from Linux mainline:
commit 37624b58542f ("Linux 5.1-rc7")

board/phytec/phycore_am335x_r2/board.c new file +15 -0
+void sdram_init(void)
+{
+	int ram_type_index = PHYCORE_R2_MT41K128M16JT_256MB;
+
+	if (fdtdec_setup_mem_size_base())
+		gd->ram_size = SZ_256M;
+
+	switch (gd->ram_size) {
+	case SZ_1G:
+		ram_type_index = PHYCORE_R2_MT41K512M16HA125IT_1024MB;
+		break;
+	case SZ_512M:
+		ram_type_index = PHYCORE_R2_MT41K256M16TW107IT_512MB;
+		break;
+	case SZ_256M:

board/phytec/phycore_am335x_r2/board.c line 155

Quoted 10 lines
board/phytec/phycore_am335x_r2/board.c new file +10 -0
+	default:
+		ram_type_index = PHYCORE_R2_MT41K128M16JT_256MB;
+		break;
+	}
+
+	config_ddr(DDR_CLK_MHZ, &ioregs,
+		   &physom_timings[ram_type_index].ddr3_data,
+		   &ddr3_cmd_ctrl_data,
+		   &physom_timings[ram_type_index].ddr3_emif_reg_data, 0);
+}

This is wrong. sdram_init() is called by arch/arm/mach-omap2/am33xx/board.c::dram_init() which then sets gd->ram_size based on what get_ram_size() determines. So this is all

Quoted 8 lines

just a wrapper around how the various parts of the am33xx generations call some form of config_ddr(). And what you have here is a lot of unused code about which module provides how much memory. I assume there's some run-time method to determine which module you're on and thus determine that correct parameters to pass in for the chip that's in use. If you're not there yet then just make sdram_init() call config_ddr(...) with the correct enum for the 256M chip and then update this when you have real detection.

Thanks for that input, you are right. I could not find any documented way to detect the exact module we are running on, but as you pointed out we can use get_ram_size() to find the size of the installed RAM. This is in fact exactly what barebox did, I just missed it. How is this for a replacement of the above?

void sdram_init(void)

{
	/* Configure memory to maximum supported size for detection */
	int ram_type_index = PHYCORE_R2_MT41K512M16HA125IT_1024MB;
	config_ddr(DDR_CLK_MHZ, &ioregs,
		   &physom_timings[ram_type_index].ddr3_data,
		   &ddr3_cmd_ctrl_data,
		   &physom_timings[ram_type_index].ddr3_emif_reg_data,
		   0);

	/* Detect memory physically present */
	gd->ram_size = get_ram_size((void *)CONFIG_SYS_SDRAM_BASE,
				    CONFIG_MAX_RAM_BANK_SIZE);

	/* Reconfigure memory for actual detected size */
	switch (gd->ram_size) {
	case SZ_1G:
		ram_type_index = PHYCORE_R2_MT41K512M16HA125IT_1024MB;
		break;
	case SZ_512M:
		ram_type_index = PHYCORE_R2_MT41K256M16TW107IT_512MB;
		break;
	case SZ_256M:
	default:
		ram_type_index = PHYCORE_R2_MT41K128M16JT_256MB;
		break;
	}
	config_ddr(DDR_CLK_MHZ, &ioregs,
		   &physom_timings[ram_type_index].ddr3_data,
		   &ddr3_cmd_ctrl_data,
		   &physom_timings[ram_type_index].ddr3_emif_reg_data,
		   0);
}
Quoted 3 lines

The ugliest part of this is, as you pointed out, that directly after this is called, get_ram_size() will be called again from sdram_init(). But it at least noninvasive, and no longer requires the device tree.

I don't think it's safe to call config_ddr twice, especially with the possibly wrong parameters. What's barebox doing in this case, being told the presumably correct DDR size in the device tree?

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190506/bf66167d/attachment.sig>
#4 lore raw

PATCH v5 1/1 ARM: am335x: Add phyCORE AM335x R2 support

Hi Tom,

Quoted 28 lines · Tom Rini
Quoted 27 lines · Niel Fourie

Hi Tom,

Quoted 24 lines · Tom Rini
Quoted 22 lines · Niel Fourie

Support for Phytech phyCORE AM335x R2 SOM (PCL060) on the Phytec phyBOARD-Wega AM335x.

CPU  : AM335X-GP rev 2.1
Model: Phytec AM335x phyBOARD-WEGA
DRAM:  256 MiB
NAND:  256 MiB
MMC:   OMAP SD/MMC: 0
eth0: ethernet at 4a100000

Working:
- Eth0
- i2C
- MMC/SD
- NAND
- UART
- USB (host)

Device trees were taken from Linux mainline:
commit 37624b58542f ("Linux 5.1-rc7")

[snip]

Quoted 35 lines
Quoted 35 lines

void sdram_init(void)

{
	/* Configure memory to maximum supported size for detection */
	int ram_type_index = PHYCORE_R2_MT41K512M16HA125IT_1024MB;
	config_ddr(DDR_CLK_MHZ, &ioregs,
		   &physom_timings[ram_type_index].ddr3_data,
		   &ddr3_cmd_ctrl_data,
		   &physom_timings[ram_type_index].ddr3_emif_reg_data,
		   0);

	/* Detect memory physically present */
	gd->ram_size = get_ram_size((void *)CONFIG_SYS_SDRAM_BASE,
				    CONFIG_MAX_RAM_BANK_SIZE);

	/* Reconfigure memory for actual detected size */
	switch (gd->ram_size) {
	case SZ_1G:
		ram_type_index = PHYCORE_R2_MT41K512M16HA125IT_1024MB;
		break;
	case SZ_512M:
		ram_type_index = PHYCORE_R2_MT41K256M16TW107IT_512MB;
		break;
	case SZ_256M:
	default:
		ram_type_index = PHYCORE_R2_MT41K128M16JT_256MB;
		break;
	}
	config_ddr(DDR_CLK_MHZ, &ioregs,
		   &physom_timings[ram_type_index].ddr3_data,
		   &ddr3_cmd_ctrl_data,
		   &physom_timings[ram_type_index].ddr3_emif_reg_data,
		   0);
}
Quoted 7 lines
Quoted 3 lines

The ugliest part of this is, as you pointed out, that directly after this is called, get_ram_size() will be called again from sdram_init(). But it at least noninvasive, and no longer requires the device tree.

I don't think it's safe to call config_ddr twice, especially with the possibly wrong parameters. What's barebox doing in this case, being told the presumably correct DDR size in the device tree?

Good point. Barebox uses the above mechanism to detect the memory size, and I could find no equivalent memory size specified in its internal device tree.

Marek originally proposed using the memory size specified in the device tree as an improvement over specifying the size in the defconfig (as in v1 of the patch).

Best regards,
Niel Fourie

Signature · 3 lines
DENX Software Engineering GmbH,      Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: +49-8142-66989-21 Fax: +49-8142-66989-80  Email: lusus at denx.de
#5 lore raw

PATCH v5 1/1 ARM: am335x: Add phyCORE AM335x R2 support

Quoted 71 lines · Niel Fourie

Hi Tom,

Quoted 28 lines · Tom Rini
Quoted 27 lines · Niel Fourie

Hi Tom,

Quoted 24 lines · Tom Rini
Quoted 22 lines · Niel Fourie

Support for Phytech phyCORE AM335x R2 SOM (PCL060) on the Phytec phyBOARD-Wega AM335x.

CPU  : AM335X-GP rev 2.1
Model: Phytec AM335x phyBOARD-WEGA
DRAM:  256 MiB
NAND:  256 MiB
MMC:   OMAP SD/MMC: 0
eth0: ethernet at 4a100000

Working:
- Eth0
- i2C
- MMC/SD
- NAND
- UART
- USB (host)

Device trees were taken from Linux mainline:
commit 37624b58542f ("Linux 5.1-rc7")

[snip]

Quoted 39 lines
Quoted 38 lines

void sdram_init(void)

{
	/* Configure memory to maximum supported size for detection */
	int ram_type_index = PHYCORE_R2_MT41K512M16HA125IT_1024MB;
	config_ddr(DDR_CLK_MHZ, &ioregs,
		   &physom_timings[ram_type_index].ddr3_data,
		   &ddr3_cmd_ctrl_data,
		   &physom_timings[ram_type_index].ddr3_emif_reg_data,
		   0);

	/* Detect memory physically present */
	gd->ram_size = get_ram_size((void *)CONFIG_SYS_SDRAM_BASE,
				    CONFIG_MAX_RAM_BANK_SIZE);

	/* Reconfigure memory for actual detected size */
	switch (gd->ram_size) {
	case SZ_1G:
		ram_type_index = PHYCORE_R2_MT41K512M16HA125IT_1024MB;
		break;
	case SZ_512M:
		ram_type_index = PHYCORE_R2_MT41K256M16TW107IT_512MB;
		break;
	case SZ_256M:
	default:
		ram_type_index = PHYCORE_R2_MT41K128M16JT_256MB;
		break;
	}
	config_ddr(DDR_CLK_MHZ, &ioregs,
		   &physom_timings[ram_type_index].ddr3_data,
		   &ddr3_cmd_ctrl_data,
		   &physom_timings[ram_type_index].ddr3_emif_reg_data,
		   0);
}

The ugliest part of this is, as you pointed out, that directly after this is called, get_ram_size() will be called again from sdram_init(). But it at least noninvasive, and no longer requires the device tree.

Quoted 7 lines
Quoted 3 lines

I don't think it's safe to call config_ddr twice, especially with the possibly wrong parameters. What's barebox doing in this case, being told the presumably correct DDR size in the device tree?

Good point. Barebox uses the above mechanism to detect the memory size, and I could find no equivalent memory size specified in its internal device tree.

Configure for 1GB and then see how much we can actually talk to?

Quoted 3 lines

Marek originally proposed using the memory size specified in the device tree as an improvement over specifying the size in the defconfig (as in v1 of the patch).

But then you aren't populating 3 device trees nor making it clear / easy to say which module you're on, and then still need to change the config for which DT you're picking up. These SOMs really don't provide any run-time method to see which one you're on? There's no GPIOs to poke?

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190507/30419d0d/attachment.sig>
#6 lore raw

PATCH v5 1/1 ARM: am335x: Add phyCORE AM335x R2 support

Hi Tom,

Quoted 76 lines · Tom Rini
Quoted 75 lines · Niel Fourie

Hi Tom,

Quoted 28 lines · Tom Rini
Quoted 27 lines · Niel Fourie

Hi Tom,

Quoted 24 lines · Tom Rini
Quoted 22 lines · Niel Fourie

Support for Phytech phyCORE AM335x R2 SOM (PCL060) on the Phytec phyBOARD-Wega AM335x.

CPU  : AM335X-GP rev 2.1
Model: Phytec AM335x phyBOARD-WEGA
DRAM:  256 MiB
NAND:  256 MiB
MMC:   OMAP SD/MMC: 0
eth0: ethernet at 4a100000

Working:
- Eth0
- i2C
- MMC/SD
- NAND
- UART
- USB (host)

Device trees were taken from Linux mainline:
commit 37624b58542f ("Linux 5.1-rc7")

[snip]

Quoted 42 lines
Quoted 35 lines

void sdram_init(void)

{
	/* Configure memory to maximum supported size for detection */
	int ram_type_index = PHYCORE_R2_MT41K512M16HA125IT_1024MB;
	config_ddr(DDR_CLK_MHZ, &ioregs,
		   &physom_timings[ram_type_index].ddr3_data,
		   &ddr3_cmd_ctrl_data,
		   &physom_timings[ram_type_index].ddr3_emif_reg_data,
		   0);

	/* Detect memory physically present */
	gd->ram_size = get_ram_size((void *)CONFIG_SYS_SDRAM_BASE,
				    CONFIG_MAX_RAM_BANK_SIZE);

	/* Reconfigure memory for actual detected size */
	switch (gd->ram_size) {
	case SZ_1G:
		ram_type_index = PHYCORE_R2_MT41K512M16HA125IT_1024MB;
		break;
	case SZ_512M:
		ram_type_index = PHYCORE_R2_MT41K256M16TW107IT_512MB;
		break;
	case SZ_256M:
	default:
		ram_type_index = PHYCORE_R2_MT41K128M16JT_256MB;
		break;
	}
	config_ddr(DDR_CLK_MHZ, &ioregs,
		   &physom_timings[ram_type_index].ddr3_data,
		   &ddr3_cmd_ctrl_data,
		   &physom_timings[ram_type_index].ddr3_emif_reg_data,
		   0);
}
Quoted 3 lines

The ugliest part of this is, as you pointed out, that directly after this is called, get_ram_size() will be called again from sdram_init(). But it at least noninvasive, and no longer requires the device tree.

I don't think it's safe to call config_ddr twice, especially with the possibly wrong parameters. What's barebox doing in this case, being told the presumably correct DDR size in the device tree?

Quoted 5 lines
Quoted 3 lines

Good point. Barebox uses the above mechanism to detect the memory size, and I could find no equivalent memory size specified in its internal device tree.

Configure for 1GB and then see how much we can actually talk to?

Yes. If you are interested, you can see their implementation here [1], where get_minimal_timings() returns the configuration for 1GiB; everything is in that file. (I did credit the author.)

[1]
https://git.pengutronix.de/cgit/barebox/tree/arch/arm/boards/phytec-som-am335x/lowlevel.c#n167

Quoted 8 lines
Quoted 3 lines

Marek originally proposed using the memory size specified in the device tree as an improvement over specifying the size in the defconfig (as in v1 of the patch).

But then you aren't populating 3 device trees nor making it clear / easy to say which module you're on, and then still need to change the config for which DT you're picking up. These SOMs really don't provide any run-time method to see which one you're on? There's no GPIOs to poke?

Agreed, the device tree solution is inferior to autodetection. The SOMs manual makes no mention of how different variants can be distinguished/detected, and the board specific code in Barebox (written by Phytec) does not contain any other detection code (except for the RAM), like checking GPIOs. Unfortunately there is no publicly available schematic, so I can't be completely sure. So I am going to assume it, there is no other way of detection.

Best regards,
Niel Fourie

Signature · 3 lines
DENX Software Engineering GmbH,      Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: +49-8142-66989-21 Fax: +49-8142-66989-80  Email: lusus at denx.de
#7 lore raw

PATCH v5 1/1 ARM: am335x: Add phyCORE AM335x R2 support

Quoted 83 lines · Niel Fourie

Hi Tom,

Quoted 80 lines · Tom Rini
Quoted 78 lines · Niel Fourie

Hi Tom,

Quoted 28 lines · Tom Rini
Quoted 27 lines · Niel Fourie

Hi Tom,

Quoted 24 lines · Tom Rini
Quoted 22 lines · Niel Fourie

Support for Phytech phyCORE AM335x R2 SOM (PCL060) on the Phytec phyBOARD-Wega AM335x.

CPU  : AM335X-GP rev 2.1
Model: Phytec AM335x phyBOARD-WEGA
DRAM:  256 MiB
NAND:  256 MiB
MMC:   OMAP SD/MMC: 0
eth0: ethernet at 4a100000

Working:
- Eth0
- i2C
- MMC/SD
- NAND
- UART
- USB (host)

Device trees were taken from Linux mainline:
commit 37624b58542f ("Linux 5.1-rc7")

[snip]

Quoted 35 lines
Quoted 35 lines

void sdram_init(void)

{
	/* Configure memory to maximum supported size for detection */
	int ram_type_index = PHYCORE_R2_MT41K512M16HA125IT_1024MB;
	config_ddr(DDR_CLK_MHZ, &ioregs,
		   &physom_timings[ram_type_index].ddr3_data,
		   &ddr3_cmd_ctrl_data,
		   &physom_timings[ram_type_index].ddr3_emif_reg_data,
		   0);

	/* Detect memory physically present */
	gd->ram_size = get_ram_size((void *)CONFIG_SYS_SDRAM_BASE,
				    CONFIG_MAX_RAM_BANK_SIZE);

	/* Reconfigure memory for actual detected size */
	switch (gd->ram_size) {
	case SZ_1G:
		ram_type_index = PHYCORE_R2_MT41K512M16HA125IT_1024MB;
		break;
	case SZ_512M:
		ram_type_index = PHYCORE_R2_MT41K256M16TW107IT_512MB;
		break;
	case SZ_256M:
	default:
		ram_type_index = PHYCORE_R2_MT41K128M16JT_256MB;
		break;
	}
	config_ddr(DDR_CLK_MHZ, &ioregs,
		   &physom_timings[ram_type_index].ddr3_data,
		   &ddr3_cmd_ctrl_data,
		   &physom_timings[ram_type_index].ddr3_emif_reg_data,
		   0);
}
Quoted 7 lines
Quoted 3 lines

The ugliest part of this is, as you pointed out, that directly after this is called, get_ram_size() will be called again from sdram_init(). But it at least noninvasive, and no longer requires the device tree.

I don't think it's safe to call config_ddr twice, especially with the possibly wrong parameters. What's barebox doing in this case, being told the presumably correct DDR size in the device tree?

Good point. Barebox uses the above mechanism to detect the memory size, and I could find no equivalent memory size specified in its internal device tree.

Quoted 7 lines
Quoted 1 line

Configure for 1GB and then see how much we can actually talk to?

Yes. If you are interested, you can see their implementation here [1], where get_minimal_timings() returns the configuration for 1GiB; everything is in that file. (I did credit the author.)

[1] https://git.pengutronix.de/cgit/barebox/tree/arch/arm/boards/phytec-som-am335x/lowlevel.c#n167

Thinking back to my time at TI, that leaves me a little worried, but if the vendor is OK with that approach, I'll leave it be, and yes, we can mirror that approach here. Thanks for the links!

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190507/1c400ee3/attachment.sig>
#8 lore raw

PATCH v5 1/1 ARM: am335x: Add phyCORE AM335x R2 support

In reply to Niel Fourie
Quoted 106 lines · Niel Fourie

Hi Tom,

Quoted 78 lines · Tom Rini
Quoted 77 lines · Niel Fourie

Hi Tom,

Quoted 28 lines · Tom Rini
Quoted 27 lines · Niel Fourie

Hi Tom,

Quoted 24 lines · Tom Rini
Quoted 22 lines · Niel Fourie

Support for Phytech phyCORE AM335x R2 SOM (PCL060) on the Phytec phyBOARD-Wega AM335x.

CPU  : AM335X-GP rev 2.1
Model: Phytec AM335x phyBOARD-WEGA
DRAM:  256 MiB
NAND:  256 MiB
MMC:   OMAP SD/MMC: 0
eth0: ethernet at 4a100000

Working:
  - Eth0
  - i2C
  - MMC/SD
  - NAND
  - UART
  - USB (host)

Device trees were taken from Linux mainline:
commit 37624b58542f ("Linux 5.1-rc7")

[snip]

Quoted 44 lines
Quoted 35 lines

void sdram_init(void)
{
    /* Configure memory to maximum supported size for detection */ int ram_type_index = PHYCORE_R2_MT41K512M16HA125IT_1024MB; config_ddr(DDR_CLK_MHZ, &ioregs, &physom_timings[ram_type_index].ddr3_data, &ddr3_cmd_ctrl_data, &physom_timings[ram_type_index].ddr3_emif_reg_data, 0);

    /* Detect memory physically present */ gd->ram_size = get_ram_size((void *)CONFIG_SYS_SDRAM_BASE, CONFIG_MAX_RAM_BANK_SIZE);

    /* Reconfigure memory for actual detected size */ switch (gd->ram_size) {
    case SZ_1G:
        ram_type_index = PHYCORE_R2_MT41K512M16HA125IT_1024MB; break;
    case SZ_512M:
        ram_type_index = PHYCORE_R2_MT41K256M16TW107IT_512MB; break;
    case SZ_256M:
    default:
        ram_type_index = PHYCORE_R2_MT41K128M16JT_256MB; break;
    }
    config_ddr(DDR_CLK_MHZ, &ioregs, &physom_timings[ram_type_index].ddr3_data, &ddr3_cmd_ctrl_data, &physom_timings[ram_type_index].ddr3_emif_reg_data, 0);
}

Quoted 5 lines

The ugliest part of this is, as you pointed out, that directly after this is
called, get_ram_size() will be called again from sdram_init(). But it at
least noninvasive, and no longer requires the device tree.

I don't think it's safe to call config_ddr twice, especially with the possibly wrong parameters.  What's barebox doing in this case, being told the presumably correct DDR size in the device tree?

Quoted 6 lines
Quoted 4 lines

Good point. Barebox uses the above mechanism to detect the memory size, and
I could find no equivalent memory size specified in its internal device tree.

Configure for 1GB and then see how much we can actually talk to?

Yes. If you are interested, you can see their implementation here [1], where get_minimal_timings() returns the configuration for 1GiB; everything is in that file. (I did credit the author.)

[1]
https://git.pengutronix.de/cgit/barebox/tree/arch/arm/boards/phytec-som-am335x/lowlevel.c#n167

Quoted 10 lines
Quoted 5 lines

Marek originally proposed using the memory size specified in the device tree
as an improvement over specifying the size in the defconfig (as in v1 of the
patch).

But then you aren't populating 3 device trees nor making it clear / easy to say which module you're on, and then still need to change the config for which DT you're picking up.  These SOMs really don't provide any run-time method to see which one you're on?  There's no GPIOs to poke?

Quoted 7 lines

Agreed, the device tree solution is inferior to autodetection. The SOMs manual makes no mention of how different variants can be distinguished/detected, and the board specific code in Barebox (written by Phytec) does not contain any other detection code (except for the RAM), like checking GPIOs. Unfortunately there is no publicly available schematic, so I can't be completely sure. So I am going to assume it, there is no other way of detection.

You are right, for our AM335 based SOMs there is no way to detect the variant. The barebox generates for every possible RAM configuration an own MLO. The nice thing is that barebox generates all variants with one build by using its multi image mechanism.

Besides the RAM, the phyCORE can be populated with different AM335x variants, NANDs, RTC/EEPROM/Eth-PHY(yes/no) and maybe other things I don't remember now.

I think it would be a good Idea to maintain a list of article numbers (that fully describes the SOM variant) which are supported by the current u-boot version.

Regards,
Wadim

Quoted 4 lines

Best regards,
Niel Fourie