Skip to contentSkip to conversation

arm: socfpga: implement proper peripheral reset handling

PATCH Superseded Simon Goldschmidt 3 participants 35 messages Jan 25, 2019 – Jan 29, 2019

Subsystems (7)

v1 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

cover arm: socfpga: implement proper peripheral reset handling

This series implements peripheral reset handling for socfpga gen5.

It moves from enabling all peripherals during SPL startup to using the socfpga reset driver from all peripherals and enabling peripherals when they are used.

As Linux cannot even handle this in 4.20, a Kconfig option is added to keep the old behaviour of just enabling all peripherals from SPL. This option is default 'y' for now but should be moved to default 'n' in the near future once a Linux kernel supports reset handling for all drivers.

Simon Goldschmidt (5):
arm: socfpga: gen5: add reset handles to devicetrees
arm: socfpga: move SDR reset handling to driver
mtd: rawnand: denali: add reset handling
spi: cadence_qspi: add reset handling
arm: socfpga: implement proper peripheral reset

 arch/arm/dts/socfpga_arria5_socdk-u-boot.dtsi |  6 ++
 arch/arm/dts/socfpga_cyclone5_dbm_soc1.dts    |  5 ++
 .../socfpga_cyclone5_de0_nano_soc-u-boot.dtsi |  6 ++
 arch/arm/dts/socfpga_cyclone5_de10_nano.dts   |  5 ++
 arch/arm/dts/socfpga_cyclone5_de1_soc.dts     |  5 ++
 arch/arm/dts/socfpga_cyclone5_is1.dts         |  5 ++
 .../dts/socfpga_cyclone5_socdk-u-boot.dtsi    |  6 ++
 .../dts/socfpga_cyclone5_sockit-u-boot.dtsi   |  6 ++
 .../dts/socfpga_cyclone5_socrates-u-boot.dtsi | 10 +++
 arch/arm/dts/socfpga_cyclone5_sr1500.dts      |  5 ++
 .../socfpga_cyclone5_vining_fpga-u-boot.dtsi  |  6 ++
 arch/arm/dts/socfpga_u-boot.dtsi              | 61 +++++++++++++++++++
 arch/arm/mach-socfpga/Kconfig                 | 10 +++
 arch/arm/mach-socfpga/misc_gen5.c             |  2 +
 arch/arm/mach-socfpga/spl_gen5.c              | 11 +++-
 drivers/ddr/altera/sdram_gen5.c               |  4 ++
 drivers/mtd/nand/raw/denali_dt.c              |  8 +++
 drivers/spi/cadence_qspi.c                    |  9 +++
 18 files changed, 169 insertions(+), 1 deletion(-)
 create mode 100644 arch/arm/dts/socfpga_u-boot.dtsi
Signature · 1 line
2.17.1
#1 lore raw

PATCH 1/5 arm: socfpga: gen5: add reset handles to devicetrees

The SPL for socfpga gen5 currently takes all peripherals out of reset unconditionally. To implement proper reset handling for peripherals,
some device tree nodes are missing reset handles.

Since the socfpga gen5 devicetrees are currently in sync with Linux 4.20, and these reset handles are not added to Linux, yet, let's add them to U-Boot specific dtsi files to ease future merging.

This adds a new file socfpga_u-boot.dtsi (note the '_' - if using a '-' this file gets pulled in for some a10/s10 devicetrees) that adds the missing reset handles.

---

 arch/arm/dts/socfpga_arria5_socdk-u-boot.dtsi |  6 ++
 arch/arm/dts/socfpga_cyclone5_dbm_soc1.dts    |  5 ++
 .../socfpga_cyclone5_de0_nano_soc-u-boot.dtsi |  6 ++
 arch/arm/dts/socfpga_cyclone5_de10_nano.dts   |  5 ++
 arch/arm/dts/socfpga_cyclone5_de1_soc.dts     |  5 ++
 arch/arm/dts/socfpga_cyclone5_is1.dts         |  5 ++
 .../dts/socfpga_cyclone5_socdk-u-boot.dtsi    |  6 ++
 .../dts/socfpga_cyclone5_sockit-u-boot.dtsi   |  6 ++
 .../dts/socfpga_cyclone5_socrates-u-boot.dtsi | 10 +++
 arch/arm/dts/socfpga_cyclone5_sr1500.dts      |  5 ++
 .../socfpga_cyclone5_vining_fpga-u-boot.dtsi  |  6 ++
 arch/arm/dts/socfpga_u-boot.dtsi              | 61 +++++++++++++++++++
 12 files changed, 126 insertions(+)
 create mode 100644 arch/arm/dts/socfpga_u-boot.dtsi
arch/arm/dts/socfpga_arria5_socdk-u-boot.dtsi +6 -0
index c44d1ee2fa..b4cd73d367 100644
@@ -6,6 +6,8 @@
  * Copyright (c) 2018 Simon Goldschmidt
  */
 
+#include "socfpga_u-boot.dtsi"
+
 /{
 	aliases {
 		spi0 = "/soc/spi at ff705000";
@@ -17,6 +19,10 @@
 	};
 };
 
+&rst {
+	u-boot,dm-pre-reloc;
+};
+
 &watchdog0 {
 	status = "disabled";
 };
arch/arm/dts/socfpga_cyclone5_dbm_soc1.dts +5 -0
index a387071674..f616740981 100644
@@ -4,6 +4,7 @@
  */
 
 #include "socfpga_cyclone5.dtsi"
+#include "socfpga_u-boot.dtsi"
 
 / {
 	model = "Devboards.de DBM-SoC1";
@@ -30,6 +31,10 @@
 	};
 };
 
+&rst {
+	u-boot,dm-pre-reloc;
+};
+
 &gmac1 {
 	status = "okay";
 	phy-mode = "rgmii";
arch/arm/dts/socfpga_cyclone5_de0_nano_soc-u-boot.dtsi +6 -0
index 08d81da169..63dd0de23b 100644
@@ -6,6 +6,8 @@
  * Copyright (c) 2018 Simon Goldschmidt
  */
 
+#include "socfpga_u-boot.dtsi"
+
 /{
 	aliases {
 		udc0 = &usb1;
@@ -16,6 +18,10 @@
 	};
 };
 
+&rst {
+	u-boot,dm-pre-reloc;
+};
+
 &watchdog0 {
 	status = "disabled";
 };
arch/arm/dts/socfpga_cyclone5_de10_nano.dts +5 -0
index e9105743ea..7106b0300a 100644
@@ -6,6 +6,7 @@
  */
 
 #include "socfpga_cyclone5.dtsi"
+#include "socfpga_u-boot.dtsi"
 
 / {
 	model = "Terasic DE10-Nano";
@@ -32,6 +33,10 @@
 	};
 };
 
+&rst {
+	u-boot,dm-pre-reloc;
+};
+
 &gmac1 {
 	status = "okay";
 	phy-mode = "rgmii";
arch/arm/dts/socfpga_cyclone5_de1_soc.dts +5 -0
index 4f076bce93..5d426dad68 100644
@@ -4,6 +4,7 @@
  */
 
 #include "socfpga_cyclone5.dtsi"
+#include "socfpga_u-boot.dtsi"
 
 / {
 	model = "Terasic DE1-SoC";
@@ -30,6 +31,10 @@
 	};
 };
 
+&rst {
+	u-boot,dm-pre-reloc;
+};
+
 &gmac1 {
 	status = "okay";
 	phy-mode = "rgmii";
arch/arm/dts/socfpga_cyclone5_is1.dts +5 -0
index b7054bfd5a..3383754cb4 100644
@@ -4,6 +4,7 @@
  */
 
 #include "socfpga_cyclone5.dtsi"
+#include "socfpga_u-boot.dtsi"
 
 / {
 	model = "SoCFPGA Cyclone V IS1";
@@ -37,6 +38,10 @@
 	};
 };
 
+&rst {
+	u-boot,dm-pre-reloc;
+};
+
 &gmac1 {
 	status = "okay";
 	phy-mode = "rgmii";
arch/arm/dts/socfpga_cyclone5_socdk-u-boot.dtsi +6 -0
index 9436e0fa8b..0c39017d55 100644
@@ -6,6 +6,8 @@
  * Copyright (c) 2018 Simon Goldschmidt
  */
 
+#include "socfpga_u-boot.dtsi"
+
 /{
 	aliases {
 		spi0 = "/soc/spi at ff705000";
@@ -17,6 +19,10 @@
 	};
 };
 
+&rst {
+	u-boot,dm-pre-reloc;
+};
+
 &can0 {
 	status = "okay";
 };
arch/arm/dts/socfpga_cyclone5_sockit-u-boot.dtsi +6 -0
index 648f1bd01d..284e19a66f 100644
@@ -6,6 +6,8 @@
  * Copyright (c) 2018 Simon Goldschmidt
  */
 
+#include "socfpga_u-boot.dtsi"
+
 /{
 	aliases {
 		spi0 = "/soc/spi at ff705000";
@@ -17,6 +19,10 @@
 	};
 };
 
+&rst {
+	u-boot,dm-pre-reloc;
+};
+
 &watchdog0 {
 	status = "disabled";
 };
arch/arm/dts/socfpga_cyclone5_socrates-u-boot.dtsi +10 -0
index 31bd1dba0f..4a0ca813cb 100644
@@ -6,6 +6,8 @@
  * Copyright (c) 2018 Simon Goldschmidt
  */
 
+#include "socfpga_u-boot.dtsi"
+
 /{
 	aliases {
 		spi0 = "/soc/spi at ff705000";
@@ -17,10 +19,18 @@
 	};
 };
 
+&rst {
+	u-boot,dm-pre-reloc;
+};
+
 &watchdog0 {
 	status = "disabled";
 };
 
+&rst {
+	u-boot,dm-pre-reloc;
+};
+
 &mmc {
 	u-boot,dm-pre-reloc;
 };
arch/arm/dts/socfpga_cyclone5_sr1500.dts +5 -0
index 6a6c29be79..0025baf29e 100644
@@ -4,6 +4,7 @@
  */
 
 #include "socfpga_cyclone5.dtsi"
+#include "socfpga_u-boot.dtsi"
 
 / {
 	model = "SoCFPGA Cyclone V SR1500";
@@ -33,6 +34,10 @@
 	};
 };
 
+&rst {
+	u-boot,dm-pre-reloc;
+};
+
 &gmac1 {
 	status = "okay";
 	phy-mode = "rgmii";
arch/arm/dts/socfpga_cyclone5_vining_fpga-u-boot.dtsi +6 -0
index 360b946ba2..4a24fcdacd 100644
@@ -6,6 +6,8 @@
  * Copyright (c) 2018 Simon Goldschmidt
  */
 
+#include "socfpga_u-boot.dtsi"
+
 /{
 	aliases {
 		spi0 = "/soc/spi at ff705000";
@@ -17,6 +19,10 @@
 	};
 };
 
+&rst {
+	u-boot,dm-pre-reloc;
+};
+
 &watchdog0 {
 	status = "disabled";
 };
arch/arm/dts/socfpga_u-boot.dtsi new file +61 -0

61 lines not on this page · raw message

Signature · 1 line
2.17.1
#2 lore raw

PATCH 1/5 arm: socfpga: gen5: add reset handles to devicetrees

Quoted 8 lines · Simon Goldschmidt

The SPL for socfpga gen5 currently takes all peripherals out of reset unconditionally. To implement proper reset handling for peripherals,
some device tree nodes are missing reset handles.

Since the socfpga gen5 devicetrees are currently in sync with Linux 4.20, and these reset handles are not added to Linux, yet, let's add them to U-Boot specific dtsi files to ease future merging.

Quoted 5 lines

This adds a new file socfpga_u-boot.dtsi (note the '_' - if using a '-' this file gets pulled in for some a10/s10 devicetrees) that adds the missing reset handles.

Shouldn't all this be in u-boot specific DTs, instead of the ones imported from Linux ?

Quoted 6 lines

---

 arch/arm/dts/socfpga_arria5_socdk-u-boot.dtsi |  6 ++
 arch/arm/dts/socfpga_cyclone5_dbm_soc1.dts    |  5 ++
 .../socfpga_cyclone5_de0_nano_soc-u-boot.dtsi |  6 ++
 arch/arm/dts/socfpga_cyclone5_de10_nano.dts   |  5 ++

[...]

Signature · 2 lines
Best regards,
Marek Vasut
#3 lore raw

PATCH 1/5 arm: socfpga: gen5: add reset handles to devicetrees

Am 26.01.2019 um 09:57 schrieb Marek Vasut:

Quoted 9 lines · Simon Goldschmidt
Quoted 8 lines

The SPL for socfpga gen5 currently takes all peripherals out of reset unconditionally. To implement proper reset handling for peripherals,
some device tree nodes are missing reset handles.

Since the socfpga gen5 devicetrees are currently in sync with Linux 4.20, and these reset handles are not added to Linux, yet, let's add them to U-Boot specific dtsi files to ease future merging.

Quoted 8 lines
Quoted 5 lines

This adds a new file socfpga_u-boot.dtsi (note the '_' - if using a '-' this file gets pulled in for some a10/s10 devicetrees) that adds the missing reset handles.

Shouldn't all this be in u-boot specific DTs, instead of the ones imported from Linux ?

The "u-boot,dm-pre-reloc" thing should of course stay in u-boot specific DTs. But I had to add many missing "resets = " references that are yet missing in upstream Linux DTs. Dinh is in the process of adding them and I'll make sure to sync that back to U-Boot once it's in Linux mainline git. I just did not want to wait until that happens...

Regards,
Simon

Quoted 9 lines
Quoted 7 lines

---

  arch/arm/dts/socfpga_arria5_socdk-u-boot.dtsi |  6 ++
  arch/arm/dts/socfpga_cyclone5_dbm_soc1.dts    |  5 ++
  .../socfpga_cyclone5_de0_nano_soc-u-boot.dtsi |  6 ++
  arch/arm/dts/socfpga_cyclone5_de10_nano.dts   |  5 ++

[...]

#4 lore raw

PATCH 1/5 arm: socfpga: gen5: add reset handles to devicetrees

Quoted 17 lines · Simon Goldschmidt

Am 26.01.2019 um 09:57 schrieb Marek Vasut:

Quoted 16 lines · Simon Goldschmidt
Quoted 14 lines

The SPL for socfpga gen5 currently takes all peripherals out of reset unconditionally. To implement proper reset handling for peripherals,   some device tree nodes are missing reset handles.

Since the socfpga gen5 devicetrees are currently in sync with Linux 4.20,
and these reset handles are not added to Linux, yet, let's add them to U-Boot specific dtsi files to ease future merging.

This adds a new file socfpga_u-boot.dtsi (note the '_' - if using a '-' this file gets pulled in for some a10/s10 devicetrees) that adds the missing reset handles.

Quoted 8 lines
Quoted 2 lines

Shouldn't all this be in u-boot specific DTs, instead of the ones imported from Linux ?

The "u-boot,dm-pre-reloc" thing should of course stay in u-boot specific DTs. But I had to add many missing "resets = " references that are yet missing in upstream Linux DTs. Dinh is in the process of adding them and I'll make sure to sync that back to U-Boot once it's in Linux mainline git. I just did not want to wait until that happens...

Maybe you can just submit this patch to Linux too ?

Signature · 2 lines
Best regards,
Marek Vasut
#5 lore raw

PATCH 1/5 arm: socfpga: gen5: add reset handles to devicetrees

Quoted 22 lines · Marek Vasut
Quoted 20 lines · Simon Goldschmidt

Am 26.01.2019 um 09:57 schrieb Marek Vasut:

Quoted 18 lines · Simon Goldschmidt
Quoted 9 lines

The SPL for socfpga gen5 currently takes all peripherals out of reset unconditionally. To implement proper reset handling for peripherals,
some device tree nodes are missing reset handles.

Since the socfpga gen5 devicetrees are currently in sync with Linux 4.20,
and these reset handles are not added to Linux, yet, let's add them to U-Boot specific dtsi files to ease future merging.

Quoted 5 lines

This adds a new file socfpga_u-boot.dtsi (note the '_' - if using a '-' this file gets pulled in for some a10/s10 devicetrees) that adds the missing reset handles.

Shouldn't all this be in u-boot specific DTs, instead of the ones imported from Linux ?

Quoted 7 lines
Quoted 5 lines

The "u-boot,dm-pre-reloc" thing should of course stay in u-boot specific DTs. But I had to add many missing "resets = " references that are yet missing in upstream Linux DTs. Dinh is in the process of adding them and I'll make sure to sync that back to U-Boot once it's in Linux mainline git. I just did not want to wait until that happens...

Maybe you can just submit this patch to Linux too ?

Dinh has already done that, that's why I have him in the loop here. But his patch was missing some reset pins, so this would need another update.

Regards,
Simon

#6 lore raw

PATCH 1/5 arm: socfpga: gen5: add reset handles to devicetrees

Quoted 29 lines · Simon Goldschmidt
Quoted 28 lines · Marek Vasut
Quoted 25 lines · Simon Goldschmidt

Am 26.01.2019 um 09:57 schrieb Marek Vasut:

Quoted 10 lines · Simon Goldschmidt
Quoted 9 lines

The SPL for socfpga gen5 currently takes all peripherals out of reset unconditionally. To implement proper reset handling for peripherals,
some device tree nodes are missing reset handles.

Since the socfpga gen5 devicetrees are currently in sync with Linux 4.20,
and these reset handles are not added to Linux, yet, let's add them to U-Boot specific dtsi files to ease future merging.

Quoted 8 lines
Quoted 5 lines

This adds a new file socfpga_u-boot.dtsi (note the '_' - if using a '-' this file gets pulled in for some a10/s10 devicetrees) that adds the missing reset handles.

Shouldn't all this be in u-boot specific DTs, instead of the ones imported from Linux ?

The "u-boot,dm-pre-reloc" thing should of course stay in u-boot specific DTs. But I had to add many missing "resets = " references that are yet missing in upstream Linux DTs. Dinh is in the process of adding them and I'll make sure to sync that back to U-Boot once it's in Linux mainline git. I just did not want to wait until that happens...

Quoted 4 lines
Quoted 1 line

Maybe you can just submit this patch to Linux too ?

Dinh has already done that, that's why I have him in the loop here. But his patch was missing some reset pins, so this would need another update.

OK, just make sure the DTs stay in sync.

Signature · 2 lines
Best regards,
Marek Vasut
#7 lore raw

PATCH 2/5 arm: socfpga: move SDR reset handling to driver

In reply to Simon Goldschmidt

To clean up reset handling for socfpga gen5, let's move the code snippet taking the DDR controller out of reset from SPL to the DDR driver.

---

 arch/arm/mach-socfpga/spl_gen5.c | 1 -
 drivers/ddr/altera/sdram_gen5.c  | 4 ++++
 2 files changed, 4 insertions(+), 1 deletion(-)
arch/arm/mach-socfpga/spl_gen5.c +0 -1
index ccdc661d05..f9bea892b1 100644
@@ -98,7 +98,6 @@ void board_init_f(ulong dummy)
 		socfpga_bridges_reset(1);
 	}
 
-	socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
 	socfpga_per_reset(SOCFPGA_RESET(UART0), 0);
 	socfpga_per_reset(SOCFPGA_RESET(OSC1TIMER0), 0);
 
drivers/ddr/altera/sdram_gen5.c +4 -0
index 821060459c..bd54c420f8 100644
@@ -7,6 +7,7 @@
 #include <div64.h>
 #include <watchdog.h>
 #include <asm/arch/fpga_manager.h>
+#include <asm/arch/reset_manager.h>
 #include <asm/arch/sdram.h>
 #include <asm/arch/system_manager.h>
 #include <asm/io.h>
@@ -434,6 +435,9 @@ int sdram_mmr_init_full(unsigned int sdr_phy_reg)
 			SDR_CTRLGRP_DRAMADDRW_ROWBITS_LSB;
 	int ret;
 
+	/* release DDR from reset */
+	socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
+
 	writel(rows, &sysmgr_regs->iswgrp_handoff[4]);
 
 	sdr_load_regs(cfg);
Signature · 1 line
2.17.1
#8 lore raw

PATCH 2/5 arm: socfpga: move SDR reset handling to driver

Quoted 35 lines · Simon Goldschmidt

To clean up reset handling for socfpga gen5, let's move the code snippet taking the DDR controller out of reset from SPL to the DDR driver.

---

 arch/arm/mach-socfpga/spl_gen5.c | 1 -
 drivers/ddr/altera/sdram_gen5.c  | 4 ++++
 2 files changed, 4 insertions(+), 1 deletion(-)
arch/arm/mach-socfpga/spl_gen5.c +0 -1
index ccdc661d05..f9bea892b1 100644
@@ -98,7 +98,6 @@ void board_init_f(ulong dummy)
 		socfpga_bridges_reset(1);
 	}
 
-	socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
 	socfpga_per_reset(SOCFPGA_RESET(UART0), 0);
 	socfpga_per_reset(SOCFPGA_RESET(OSC1TIMER0), 0);
 
drivers/ddr/altera/sdram_gen5.c +1 -0
index 821060459c..bd54c420f8 100644
@@ -7,6 +7,7 @@
 #include <div64.h>
 #include <watchdog.h>
 #include <asm/arch/fpga_manager.h>
+#include <asm/arch/reset_manager.h>
 #include <asm/arch/sdram.h>
 #include <asm/arch/system_manager.h>
 #include <asm/io.h>

drivers/ddr/altera/sdram_gen5.c int sdram_mmr_init_full(unsigned int sdr_phy_reg) · line 440

Quoted 6 lines
drivers/ddr/altera/sdram_gen5.c +3 -0
@@ -434,6 +435,9 @@ int sdram_mmr_init_full(unsigned int sdr_phy_reg)
 			SDR_CTRLGRP_DRAMADDRW_ROWBITS_LSB;
 	int ret;
 
+	/* release DDR from reset */
+	socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
+

Can the reset framework do this ?

Quoted 4 lines
drivers/ddr/altera/sdram_gen5.c +0 -0
 	writel(rows, &sysmgr_regs->iswgrp_handoff[4]);
 
 	sdr_load_regs(cfg);
Signature · 2 lines
Best regards,
Marek Vasut
#9 lore raw

PATCH 2/5 arm: socfpga: move SDR reset handling to driver

Am 26.01.2019 um 09:58 schrieb Marek Vasut:

Quoted 39 lines · Simon Goldschmidt
Quoted 38 lines

To clean up reset handling for socfpga gen5, let's move the code snippet taking the DDR controller out of reset from SPL to the DDR driver.

---

  arch/arm/mach-socfpga/spl_gen5.c | 1 -
  drivers/ddr/altera/sdram_gen5.c  | 4 ++++
  2 files changed, 4 insertions(+), 1 deletion(-)
arch/arm/mach-socfpga/spl_gen5.c +0 -1
index ccdc661d05..f9bea892b1 100644
@@ -98,7 +98,6 @@ void board_init_f(ulong dummy)
  		socfpga_bridges_reset(1);
  	}
 
-	socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
  	socfpga_per_reset(SOCFPGA_RESET(UART0), 0);
  	socfpga_per_reset(SOCFPGA_RESET(OSC1TIMER0), 0);
 
drivers/ddr/altera/sdram_gen5.c +1 -0
index 821060459c..bd54c420f8 100644
@@ -7,6 +7,7 @@
  #include <div64.h>
  #include <watchdog.h>
  #include <asm/arch/fpga_manager.h>
+#include <asm/arch/reset_manager.h>
  #include <asm/arch/sdram.h>
  #include <asm/arch/system_manager.h>
  #include <asm/io.h>
@@ -434,6 +435,9 @@ int sdram_mmr_init_full(unsigned int sdr_phy_reg)
  			SDR_CTRLGRP_DRAMADDRW_ROWBITS_LSB;
  	int ret;
 
Quoted 5 lines

drivers/ddr/altera/sdram_gen5.c int sdram_mmr_init_full(unsigned int sdr_phy_reg) · line 440

Quoted 3 lines
drivers/ddr/altera/sdram_gen5.c +3 -0
+	/* release DDR from reset */
+	socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
+

Can the reset framework do this ?

Hmm, it probably could, but I see that as a diferent patch. The altera DDR driver currently does not work with devicetree, so using the reset framework here would be a bigger patch, I think.

Can we do that later and clean up this by just moving the code?

Regards,
Simon

Quoted 7 lines
Quoted 5 lines
drivers/ddr/altera/sdram_gen5.c +0 -0
  	writel(rows, &sysmgr_regs->iswgrp_handoff[4]);
 
  	sdr_load_regs(cfg);
#10 lore raw

PATCH 2/5 arm: socfpga: move SDR reset handling to driver

Quoted 45 lines · Simon Goldschmidt

Am 26.01.2019 um 09:58 schrieb Marek Vasut:

Quoted 35 lines · Simon Goldschmidt
Quoted 34 lines

To clean up reset handling for socfpga gen5, let's move the code snippet taking the DDR controller out of reset from SPL to the DDR driver.

---

  arch/arm/mach-socfpga/spl_gen5.c | 1 -
  drivers/ddr/altera/sdram_gen5.c  | 4 ++++

  2 files changed, 4 insertions(+), 1 deletion(-)
arch/arm/mach-socfpga/spl_gen5.c +0 -0
index ccdc661d05..f9bea892b1 100644
@@ -98,7 +98,6 @@ void board_init_f(ulong dummy)
           socfpga_bridges_reset(1);
       }
   -    socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
       socfpga_per_reset(SOCFPGA_RESET(UART0), 0);
       socfpga_per_reset(SOCFPGA_RESET(OSC1TIMER0), 0);

  diff --git a/drivers/ddr/altera/sdram_gen5.c
b/drivers/ddr/altera/sdram_gen5.c
index 821060459c..bd54c420f8 100644

drivers/ddr/altera/sdram_gen5.c +1 -0
@@ -7,6 +7,7 @@
   #include <div64.h>
   #include <watchdog.h>
   #include <asm/arch/fpga_manager.h>
+#include <asm/arch/reset_manager.h>
   #include <asm/arch/sdram.h>
   #include <asm/arch/system_manager.h>
   #include <asm/io.h>
Quoted 8 lines
Quoted 6 lines
drivers/ddr/altera/sdram_gen5.c +0 -0
@@ -434,6 +435,9 @@ int sdram_mmr_init_full(unsigned int sdr_phy_reg)
               SDR_CTRLGRP_DRAMADDRW_ROWBITS_LSB;
       int ret;

  +    /* release DDR from reset */
+    socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
+

Can the reset framework do this ?

Quoted 5 lines

Hmm, it probably could, but I see that as a diferent patch. The altera DDR driver currently does not work with devicetree, so using the reset framework here would be a bigger patch, I think.

Can we do that later and clean up this by just moving the code?

How much effort is it to switch this driver over to DM ?

Signature · 2 lines
Best regards,
Marek Vasut
#11 lore raw

PATCH 2/5 arm: socfpga: move SDR reset handling to driver

Quoted 47 lines · Marek Vasut
Quoted 45 lines · Simon Goldschmidt

Am 26.01.2019 um 09:58 schrieb Marek Vasut:

Quoted 43 lines · Simon Goldschmidt
Quoted 35 lines

To clean up reset handling for socfpga gen5, let's move the code snippet taking the DDR controller out of reset from SPL to the DDR driver.

---

  arch/arm/mach-socfpga/spl_gen5.c | 1 -
  drivers/ddr/altera/sdram_gen5.c  | 4 ++++
  2 files changed, 4 insertions(+), 1 deletion(-)
arch/arm/mach-socfpga/spl_gen5.c +0 -0
index ccdc661d05..f9bea892b1 100644
@@ -98,7 +98,6 @@ void board_init_f(ulong dummy)
          socfpga_bridges_reset(1);
      }
  -    socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
      socfpga_per_reset(SOCFPGA_RESET(UART0), 0);
      socfpga_per_reset(SOCFPGA_RESET(OSC1TIMER0), 0);
  diff --git a/drivers/ddr/altera/sdram_gen5.c

b/drivers/ddr/altera/sdram_gen5.c
index 821060459c..bd54c420f8 100644

drivers/ddr/altera/sdram_gen5.c +1 -0
@@ -7,6 +7,7 @@
  #include <div64.h>
  #include <watchdog.h>
  #include <asm/arch/fpga_manager.h>
+#include <asm/arch/reset_manager.h>
  #include <asm/arch/sdram.h>
  #include <asm/arch/system_manager.h>
  #include <asm/io.h>

drivers/ddr/altera/sdram_gen5.c int sdram_mmr_init_full(unsigned int sdr_phy_reg) · line 439

Quoted 5 lines
drivers/ddr/altera/sdram_gen5.c +2 -0
@@ -434,6 +435,9 @@ int sdram_mmr_init_full(unsigned int sdr_phy_reg)
              SDR_CTRLGRP_DRAMADDRW_ROWBITS_LSB;
      int ret;
  +    /* release DDR from reset */
+    socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
+

Can the reset framework do this ?

Quoted 7 lines
Quoted 5 lines

Hmm, it probably could, but I see that as a diferent patch. The altera DDR driver currently does not work with devicetree, so using the reset framework here would be a bigger patch, I think.

Can we do that later and clean up this by just moving the code?

How much effort is it to switch this driver over to DM ?

I really don't know. Searching through drivers/ddr does not seem to give me an example of a DTS-enabled ddr driver. Given that, I'd rather just push this patch now. While it's true that it doesn't clean up everything, it's not as if it
would make things worse.

Regards,
Simon

#12 lore raw

PATCH 2/5 arm: socfpga: move SDR reset handling to driver

Quoted 54 lines · Simon Goldschmidt
Quoted 53 lines · Marek Vasut
Quoted 50 lines · Simon Goldschmidt

Am 26.01.2019 um 09:58 schrieb Marek Vasut:

Quoted 35 lines · Simon Goldschmidt
Quoted 34 lines

To clean up reset handling for socfpga gen5, let's move the code snippet taking the DDR controller out of reset from SPL to the DDR driver.

---

  arch/arm/mach-socfpga/spl_gen5.c | 1 -
  drivers/ddr/altera/sdram_gen5.c  | 4 ++++
  2 files changed, 4 insertions(+), 1 deletion(-)
arch/arm/mach-socfpga/spl_gen5.c +0 -0
index ccdc661d05..f9bea892b1 100644
@@ -98,7 +98,6 @@ void board_init_f(ulong dummy)
          socfpga_bridges_reset(1);
      }
  -    socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
      socfpga_per_reset(SOCFPGA_RESET(UART0), 0);
      socfpga_per_reset(SOCFPGA_RESET(OSC1TIMER0), 0);
  diff --git a/drivers/ddr/altera/sdram_gen5.c

b/drivers/ddr/altera/sdram_gen5.c
index 821060459c..bd54c420f8 100644

drivers/ddr/altera/sdram_gen5.c +1 -0
@@ -7,6 +7,7 @@
  #include <div64.h>
  #include <watchdog.h>
  #include <asm/arch/fpga_manager.h>
+#include <asm/arch/reset_manager.h>
  #include <asm/arch/sdram.h>
  #include <asm/arch/system_manager.h>
  #include <asm/io.h>
Quoted 8 lines

drivers/ddr/altera/sdram_gen5.c int sdram_mmr_init_full(unsigned int sdr_phy_reg) · line 439

Quoted 6 lines
drivers/ddr/altera/sdram_gen5.c +2 -0
@@ -434,6 +435,9 @@ int sdram_mmr_init_full(unsigned int sdr_phy_reg)
              SDR_CTRLGRP_DRAMADDRW_ROWBITS_LSB;
      int ret;
  +    /* release DDR from reset */
+    socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
+

Can the reset framework do this ?

Hmm, it probably could, but I see that as a diferent patch. The altera DDR driver currently does not work with devicetree, so using the reset framework here would be a bigger patch, I think.

Can we do that later and clean up this by just moving the code?

Quoted 7 lines
Quoted 1 line

How much effort is it to switch this driver over to DM ?

I really don't know. Searching through drivers/ddr does not seem to give me an example of a DTS-enabled ddr driver. Given that, I'd rather just push this patch now. While it's true that it doesn't clean up everything, it's not as if it
would make things worse.

That's a valid point.

I guess you can add DRAM uclass and just probe the driver, which should be all that's needed.

Signature · 2 lines
Best regards,
Marek Vasut
#13 lore raw

PATCH 2/5 arm: socfpga: move SDR reset handling to driver

Quoted 64 lines · Marek Vasut
Quoted 54 lines · Simon Goldschmidt
Quoted 53 lines · Marek Vasut
Quoted 50 lines · Simon Goldschmidt

Am 26.01.2019 um 09:58 schrieb Marek Vasut:

Quoted 35 lines · Simon Goldschmidt
Quoted 34 lines

To clean up reset handling for socfpga gen5, let's move the code snippet taking the DDR controller out of reset from SPL to the DDR driver.

---

  arch/arm/mach-socfpga/spl_gen5.c | 1 -
  drivers/ddr/altera/sdram_gen5.c  | 4 ++++
  2 files changed, 4 insertions(+), 1 deletion(-)
arch/arm/mach-socfpga/spl_gen5.c +0 -0
index ccdc661d05..f9bea892b1 100644
@@ -98,7 +98,6 @@ void board_init_f(ulong dummy)
          socfpga_bridges_reset(1);
      }
  -    socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
      socfpga_per_reset(SOCFPGA_RESET(UART0), 0);
      socfpga_per_reset(SOCFPGA_RESET(OSC1TIMER0), 0);
  diff --git a/drivers/ddr/altera/sdram_gen5.c

b/drivers/ddr/altera/sdram_gen5.c
index 821060459c..bd54c420f8 100644

drivers/ddr/altera/sdram_gen5.c +1 -0
@@ -7,6 +7,7 @@
  #include <div64.h>
  #include <watchdog.h>
  #include <asm/arch/fpga_manager.h>
+#include <asm/arch/reset_manager.h>
  #include <asm/arch/sdram.h>
  #include <asm/arch/system_manager.h>
  #include <asm/io.h>
Quoted 8 lines

drivers/ddr/altera/sdram_gen5.c int sdram_mmr_init_full(unsigned int sdr_phy_reg) · line 439

Quoted 6 lines
drivers/ddr/altera/sdram_gen5.c +2 -0
@@ -434,6 +435,9 @@ int sdram_mmr_init_full(unsigned int sdr_phy_reg)
              SDR_CTRLGRP_DRAMADDRW_ROWBITS_LSB;
      int ret;
  +    /* release DDR from reset */
+    socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
+

Can the reset framework do this ?

Hmm, it probably could, but I see that as a diferent patch. The altera DDR driver currently does not work with devicetree, so using the reset framework here would be a bigger patch, I think.

Can we do that later and clean up this by just moving the code?

Quoted 7 lines
Quoted 1 line

How much effort is it to switch this driver over to DM ?

I really don't know. Searching through drivers/ddr does not seem to give me an example of a DTS-enabled ddr driver. Given that, I'd rather just push this patch now. While it's true that it doesn't clean up everything, it's not as if it
would make things worse.

Quoted 4 lines

That's a valid point.

I guess you can add DRAM uclass and just probe the driver, which should be all that's needed.

Hmm, there *is* a UCLASS_RAM, but its drivers are in 'drivers/ram'. Is there any reason those are separated from 'drivers/ddr'?

Regards,
Simon

#14 lore raw

PATCH 2/5 arm: socfpga: move SDR reset handling to driver

Quoted 65 lines · Simon Goldschmidt
Quoted 64 lines · Marek Vasut
Quoted 61 lines · Simon Goldschmidt
Quoted 47 lines · Marek Vasut
Quoted 45 lines · Simon Goldschmidt

Am 26.01.2019 um 09:58 schrieb Marek Vasut:

Quoted 43 lines · Simon Goldschmidt
Quoted 35 lines

To clean up reset handling for socfpga gen5, let's move the code snippet taking the DDR controller out of reset from SPL to the DDR driver.

---

  arch/arm/mach-socfpga/spl_gen5.c | 1 -
  drivers/ddr/altera/sdram_gen5.c  | 4 ++++
  2 files changed, 4 insertions(+), 1 deletion(-)
arch/arm/mach-socfpga/spl_gen5.c +0 -0
index ccdc661d05..f9bea892b1 100644
@@ -98,7 +98,6 @@ void board_init_f(ulong dummy)
          socfpga_bridges_reset(1);
      }
  -    socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
      socfpga_per_reset(SOCFPGA_RESET(UART0), 0);
      socfpga_per_reset(SOCFPGA_RESET(OSC1TIMER0), 0);
  diff --git a/drivers/ddr/altera/sdram_gen5.c

b/drivers/ddr/altera/sdram_gen5.c
index 821060459c..bd54c420f8 100644

drivers/ddr/altera/sdram_gen5.c +1 -0
@@ -7,6 +7,7 @@
  #include <div64.h>
  #include <watchdog.h>
  #include <asm/arch/fpga_manager.h>
+#include <asm/arch/reset_manager.h>
  #include <asm/arch/sdram.h>
  #include <asm/arch/system_manager.h>
  #include <asm/io.h>

drivers/ddr/altera/sdram_gen5.c int sdram_mmr_init_full(unsigned int sdr_phy_reg) · line 439

Quoted 5 lines
drivers/ddr/altera/sdram_gen5.c +2 -0
@@ -434,6 +435,9 @@ int sdram_mmr_init_full(unsigned int sdr_phy_reg)
              SDR_CTRLGRP_DRAMADDRW_ROWBITS_LSB;
      int ret;
  +    /* release DDR from reset */
+    socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
+

Can the reset framework do this ?

Quoted 7 lines
Quoted 5 lines

Hmm, it probably could, but I see that as a diferent patch. The altera DDR driver currently does not work with devicetree, so using the reset framework here would be a bigger patch, I think.

Can we do that later and clean up this by just moving the code?

How much effort is it to switch this driver over to DM ?

I really don't know. Searching through drivers/ddr does not seem to give me an example of a DTS-enabled ddr driver. Given that, I'd rather just push this patch now. While it's true that it doesn't clean up everything, it's not as if it
would make things worse.

Quoted 7 lines
Quoted 4 lines

That's a valid point.

I guess you can add DRAM uclass and just probe the driver, which should be all that's needed.

Hmm, there *is* a UCLASS_RAM, but its drivers are in 'drivers/ram'. Is there any reason those are separated from 'drivers/ddr'?

I don't think so, seems like these two directories should be merged.

Signature · 2 lines
Best regards,
Marek Vasut
#15 lore raw

PATCH 2/5 arm: socfpga: move SDR reset handling to driver

Am 28.01.2019 um 20:02 schrieb Marek Vasut:

Quoted 68 lines · Simon Goldschmidt
Quoted 67 lines · Marek Vasut
Quoted 66 lines · Simon Goldschmidt
Quoted 54 lines · Marek Vasut
Quoted 53 lines · Simon Goldschmidt
Quoted 50 lines

Am 26.01.2019 um 09:58 schrieb Marek Vasut:

Quoted 35 lines · Simon Goldschmidt
Quoted 34 lines

To clean up reset handling for socfpga gen5, let's move the code snippet taking the DDR controller out of reset from SPL to the DDR driver.

---

   arch/arm/mach-socfpga/spl_gen5.c | 1 -
   drivers/ddr/altera/sdram_gen5.c  | 4 ++++
   2 files changed, 4 insertions(+), 1 deletion(-)
arch/arm/mach-socfpga/spl_gen5.c +0 -0
index ccdc661d05..f9bea892b1 100644
@@ -98,7 +98,6 @@ void board_init_f(ulong dummy)
           socfpga_bridges_reset(1);
       }
   -    socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
       socfpga_per_reset(SOCFPGA_RESET(UART0), 0);
       socfpga_per_reset(SOCFPGA_RESET(OSC1TIMER0), 0);
   diff --git a/drivers/ddr/altera/sdram_gen5.c

b/drivers/ddr/altera/sdram_gen5.c
index 821060459c..bd54c420f8 100644

drivers/ddr/altera/sdram_gen5.c +1 -0
@@ -7,6 +7,7 @@
   #include <div64.h>
   #include <watchdog.h>
   #include <asm/arch/fpga_manager.h>
+#include <asm/arch/reset_manager.h>
   #include <asm/arch/sdram.h>
   #include <asm/arch/system_manager.h>
   #include <asm/io.h>
Quoted 8 lines

drivers/ddr/altera/sdram_gen5.c int sdram_mmr_init_full(unsigned int sdr_phy_reg) · line 439

Quoted 6 lines
drivers/ddr/altera/sdram_gen5.c +2 -0
@@ -434,6 +435,9 @@ int sdram_mmr_init_full(unsigned int sdr_phy_reg)
               SDR_CTRLGRP_DRAMADDRW_ROWBITS_LSB;
       int ret;
   +    /* release DDR from reset */
+    socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
+

Can the reset framework do this ?

Hmm, it probably could, but I see that as a diferent patch. The altera DDR driver currently does not work with devicetree, so using the reset framework here would be a bigger patch, I think.

Can we do that later and clean up this by just moving the code?

Quoted 7 lines
Quoted 1 line

How much effort is it to switch this driver over to DM ?

I really don't know. Searching through drivers/ddr does not seem to give me an example of a DTS-enabled ddr driver. Given that, I'd rather just push this patch now. While it's true that it doesn't clean up everything, it's not as if it
would make things worse.

That's a valid point.

Quoted 7 lines
Quoted 5 lines
Quoted 2 lines

I guess you can add DRAM uclass and just probe the driver, which should be all that's needed.

Hmm, there *is* a UCLASS_RAM, but its drivers are in 'drivers/ram'. Is there any reason those are separated from 'drivers/ddr'?

I don't think so, seems like these two directories should be merged.

Yes, I think so too by now.

I'll see if I can change this patch to use UCLASS_RAM. A patch/series to merge drivers/ddr wih drivers/ram should be separate.

Regads,
Simon

#16 lore raw

PATCH 2/5 arm: socfpga: move SDR reset handling to driver

Quoted 82 lines · Simon Goldschmidt

Am 28.01.2019 um 20:02 schrieb Marek Vasut:

Quoted 81 lines · Simon Goldschmidt
Quoted 79 lines · Marek Vasut
Quoted 70 lines · Simon Goldschmidt
Quoted 60 lines · Marek Vasut
Quoted 55 lines · Simon Goldschmidt
Quoted 53 lines

Am 26.01.2019 um 09:58 schrieb Marek Vasut:

Quoted 37 lines · Simon Goldschmidt
Quoted 36 lines

To clean up reset handling for socfpga gen5, let's move the code snippet
taking the DDR controller out of reset from SPL to the DDR driver.

---

   arch/arm/mach-socfpga/spl_gen5.c | 1 -
   drivers/ddr/altera/sdram_gen5.c  | 4 ++++

   2 files changed, 4 insertions(+), 1 deletion(-)
arch/arm/mach-socfpga/spl_gen5.c +0 -0
index ccdc661d05..f9bea892b1 100644
@@ -98,7 +98,6 @@ void board_init_f(ulong dummy)
            socfpga_bridges_reset(1);
        }
    -    socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
        socfpga_per_reset(SOCFPGA_RESET(UART0), 0);
        socfpga_per_reset(SOCFPGA_RESET(OSC1TIMER0), 0);

   diff --git a/drivers/ddr/altera/sdram_gen5.c
b/drivers/ddr/altera/sdram_gen5.c
index 821060459c..bd54c420f8 100644

drivers/ddr/altera/sdram_gen5.c +1 -0
@@ -7,6 +7,7 @@
    #include <div64.h>
    #include <watchdog.h>
    #include <asm/arch/fpga_manager.h>
+#include <asm/arch/reset_manager.h>
    #include <asm/arch/sdram.h>
    #include <asm/arch/system_manager.h>
    #include <asm/io.h>
@@ -434,6 +435,9 @@ int sdram_mmr_init_full(unsigned int
Quoted 8 lines
Quoted 6 lines
drivers/ddr/altera/sdram_gen5.c +0 -0
 sdr_phy_reg)
                SDR_CTRLGRP_DRAMADDRW_ROWBITS_LSB;
        int ret;

   +    /* release DDR from reset */
+    socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
+

Can the reset framework do this ?

Hmm, it probably could, but I see that as a diferent patch. The altera
DDR driver currently does not work with devicetree, so using the reset
framework here would be a bigger patch, I think.

Quoted 3 lines
Quoted 1 line

Can we do that later and clean up this by just moving the code?

How much effort is it to switch this driver over to DM ?

Quoted 7 lines

I really don't know. Searching through drivers/ddr does not seem to give me
an example of a DTS-enabled ddr driver. Given that, I'd rather just push this
patch now. While it's true that it doesn't clean up everything, it's not as if it
would make things worse.

Quoted 4 lines

That's a valid point.

I guess you can add DRAM uclass and just probe the driver, which should be all that's needed.

Hmm, there *is* a UCLASS_RAM, but its drivers are in 'drivers/ram'. Is there
any reason those are separated from 'drivers/ddr'?

Quoted 6 lines
Quoted 1 line

I don't think so, seems like these two directories should be merged.

Yes, I think so too by now.

I'll see if I can change this patch to use UCLASS_RAM. A patch/series to merge drivers/ddr wih drivers/ram should be separate.

Sounds good.

Signature · 2 lines
Best regards,
Marek Vasut
#17 lore raw

PATCH 2/5 arm: socfpga: move SDR reset handling to driver

Quoted 84 lines · Marek Vasut
Quoted 82 lines · Simon Goldschmidt

Am 28.01.2019 um 20:02 schrieb Marek Vasut:

Quoted 81 lines · Simon Goldschmidt
Quoted 79 lines · Marek Vasut
Quoted 70 lines · Simon Goldschmidt
Quoted 60 lines · Marek Vasut
Quoted 55 lines · Simon Goldschmidt
Quoted 53 lines

Am 26.01.2019 um 09:58 schrieb Marek Vasut:

Quoted 37 lines · Simon Goldschmidt
Quoted 36 lines

To clean up reset handling for socfpga gen5, let's move the code snippet
taking the DDR controller out of reset from SPL to the DDR driver.

---

   arch/arm/mach-socfpga/spl_gen5.c | 1 -
   drivers/ddr/altera/sdram_gen5.c  | 4 ++++
   2 files changed, 4 insertions(+), 1 deletion(-)
arch/arm/mach-socfpga/spl_gen5.c +0 -0
index ccdc661d05..f9bea892b1 100644
@@ -98,7 +98,6 @@ void board_init_f(ulong dummy)
           socfpga_bridges_reset(1);
       }
   -    socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
       socfpga_per_reset(SOCFPGA_RESET(UART0), 0);
       socfpga_per_reset(SOCFPGA_RESET(OSC1TIMER0), 0);
   diff --git a/drivers/ddr/altera/sdram_gen5.c

b/drivers/ddr/altera/sdram_gen5.c
index 821060459c..bd54c420f8 100644

drivers/ddr/altera/sdram_gen5.c +1 -0
@@ -7,6 +7,7 @@
   #include <div64.h>
   #include <watchdog.h>
   #include <asm/arch/fpga_manager.h>
+#include <asm/arch/reset_manager.h>
   #include <asm/arch/sdram.h>
   #include <asm/arch/system_manager.h>
   #include <asm/io.h>
@@ -434,6 +435,9 @@ int sdram_mmr_init_full(unsigned int
Quoted 8 lines

drivers/ddr/altera/sdram_gen5.c int sdram_mmr_init_full(unsigned int · line 440

Quoted 6 lines
drivers/ddr/altera/sdram_gen5.c +2 -0
 sdr_phy_reg)
               SDR_CTRLGRP_DRAMADDRW_ROWBITS_LSB;
       int ret;
   +    /* release DDR from reset */
+    socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
+

Can the reset framework do this ?

Hmm, it probably could, but I see that as a diferent patch. The altera
DDR driver currently does not work with devicetree, so using the reset
framework here would be a bigger patch, I think.

Quoted 3 lines
Quoted 1 line

Can we do that later and clean up this by just moving the code?

How much effort is it to switch this driver over to DM ?

Quoted 7 lines

I really don't know. Searching through drivers/ddr does not seem to give me
an example of a DTS-enabled ddr driver. Given that, I'd rather just push this
patch now. While it's true that it doesn't clean up everything, it's not as if it
would make things worse.

Quoted 4 lines

That's a valid point.

I guess you can add DRAM uclass and just probe the driver, which should be all that's needed.

Hmm, there *is* a UCLASS_RAM, but its drivers are in 'drivers/ram'. Is there
any reason those are separated from 'drivers/ddr'?

Quoted 8 lines
Quoted 6 lines
Quoted 1 line

I don't think so, seems like these two directories should be merged.

Yes, I think so too by now.

I'll see if I can change this patch to use UCLASS_RAM. A patch/series to merge drivers/ddr wih drivers/ram should be separate.

Sounds good.

It kind of works with UCLASS_RAM, but there's a problem with the devicetree:
it describes the SDR controller starting at 0xffc25000 where the official memory map from Intel says it starts at 0xffc20000, but describes its registers starting
from offset 0x5000. However, in U-Boot, the file 'drivers/ddr/altera/sequencer.c'
uses those lower addresses.

So now I can either change the dts to let SDR registers start at 0xffc20000 instead of 0xffc25000 (and in consequence adapt Linux, too) or add a new driver for the sequencer that occupies this lower range (and make sdram_gen5.c use it somehow).

Before I spin another round, what would be the preferred way to take?

Anyway, I failed to find public documentation for this sequencer thing. Do you happen to know why it's done like this?

Regards,
Simon

#18 lore raw

PATCH 2/5 arm: socfpga: move SDR reset handling to driver

Quoted 102 lines · Simon Goldschmidt
Quoted 84 lines · Marek Vasut
Quoted 82 lines · Simon Goldschmidt

Am 28.01.2019 um 20:02 schrieb Marek Vasut:

Quoted 81 lines · Simon Goldschmidt
Quoted 79 lines · Marek Vasut
Quoted 70 lines · Simon Goldschmidt
Quoted 60 lines · Marek Vasut
Quoted 55 lines · Simon Goldschmidt
Quoted 53 lines

Am 26.01.2019 um 09:58 schrieb Marek Vasut:

Quoted 37 lines · Simon Goldschmidt
Quoted 36 lines

To clean up reset handling for socfpga gen5, let's move the code snippet
taking the DDR controller out of reset from SPL to the DDR driver.

---

   arch/arm/mach-socfpga/spl_gen5.c | 1 -
   drivers/ddr/altera/sdram_gen5.c  | 4 ++++
   2 files changed, 4 insertions(+), 1 deletion(-)
arch/arm/mach-socfpga/spl_gen5.c +0 -0
index ccdc661d05..f9bea892b1 100644
@@ -98,7 +98,6 @@ void board_init_f(ulong dummy)
           socfpga_bridges_reset(1);
       }
   -    socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
       socfpga_per_reset(SOCFPGA_RESET(UART0), 0);
       socfpga_per_reset(SOCFPGA_RESET(OSC1TIMER0), 0);
   diff --git a/drivers/ddr/altera/sdram_gen5.c

b/drivers/ddr/altera/sdram_gen5.c
index 821060459c..bd54c420f8 100644

drivers/ddr/altera/sdram_gen5.c +1 -0
@@ -7,6 +7,7 @@
   #include <div64.h>
   #include <watchdog.h>
   #include <asm/arch/fpga_manager.h>
+#include <asm/arch/reset_manager.h>
   #include <asm/arch/sdram.h>
   #include <asm/arch/system_manager.h>
   #include <asm/io.h>
@@ -434,6 +435,9 @@ int sdram_mmr_init_full(unsigned int
Quoted 8 lines

drivers/ddr/altera/sdram_gen5.c int sdram_mmr_init_full(unsigned int · line 440

Quoted 6 lines
drivers/ddr/altera/sdram_gen5.c +2 -0
 sdr_phy_reg)
               SDR_CTRLGRP_DRAMADDRW_ROWBITS_LSB;
       int ret;
   +    /* release DDR from reset */
+    socfpga_per_reset(SOCFPGA_RESET(SDR), 0);
+

Can the reset framework do this ?

Hmm, it probably could, but I see that as a diferent patch. The altera
DDR driver currently does not work with devicetree, so using the reset
framework here would be a bigger patch, I think.

Quoted 3 lines
Quoted 1 line

Can we do that later and clean up this by just moving the code?

How much effort is it to switch this driver over to DM ?

Quoted 7 lines

I really don't know. Searching through drivers/ddr does not seem to give me
an example of a DTS-enabled ddr driver. Given that, I'd rather just push this
patch now. While it's true that it doesn't clean up everything, it's not as if it
would make things worse.

Quoted 4 lines

That's a valid point.

I guess you can add DRAM uclass and just probe the driver, which should be all that's needed.

Hmm, there *is* a UCLASS_RAM, but its drivers are in 'drivers/ram'. Is there
any reason those are separated from 'drivers/ddr'?

Quoted 8 lines
Quoted 6 lines
Quoted 1 line

I don't think so, seems like these two directories should be merged.

Yes, I think so too by now.

I'll see if I can change this patch to use UCLASS_RAM. A patch/series to merge drivers/ddr wih drivers/ram should be separate.

Sounds good.

It kind of works with UCLASS_RAM, but there's a problem with the devicetree:
it describes the SDR controller starting at 0xffc25000 where the official memory map from Intel says it starts at 0xffc20000, but describes its registers starting
from offset 0x5000. However, in U-Boot, the file 'drivers/ddr/altera/sequencer.c'
uses those lower addresses.

Quoted 4 lines

So now I can either change the dts to let SDR registers start at 0xffc20000 instead of 0xffc25000 (and in consequence adapt Linux, too) or add a new driver for the sequencer that occupies this lower range (and make sdram_gen5.c use it somehow).

Fix the DT, I think the DT is buggy. The DRAM controller probably starts at 0xffc2_0000 and the various 4k chunks above 0xffc2_0000 are various subcomponents of the DDR controller.

Quoted 4 lines

Before I spin another round, what would be the preferred way to take?

Anyway, I failed to find public documentation for this sequencer thing. Do you happen to know why it's done like this?

I don't have one, but I _think_ it's a HW instance of the altera DDR3 controller which you can also put into the FPGA, and that one is in quartus.

Signature · 2 lines
Best regards,
Marek Vasut
#19 lore raw

PATCH 3/5 mtd: rawnand: denali: add reset handling

In reply to Simon Goldschmidt

This adds reset handling to the devicetree-enabled denalid nand driver.

For backwards compatibility, only a warning is printed when failing to get reset handles.

---

 drivers/mtd/nand/raw/denali_dt.c | 8 ++++++++
 1 file changed, 8 insertions(+)
drivers/mtd/nand/raw/denali_dt.c +8 -0
index d384b974df..21c4540637 100644
@@ -9,6 +9,7 @@
 #include <linux/io.h>
 #include <linux/ioport.h>
 #include <linux/printk.h>
+#include <reset.h>
 
 #include "denali.h"
 
@@ -64,6 +65,7 @@ static int denali_dt_probe(struct udevice *dev)
 	const struct denali_dt_data *data;
 	struct clk clk, clk_x, clk_ecc;
 	struct resource res;
+	struct reset_ctl_bulk reset_bulk;
 	int ret;
 
 	data = (void *)dev_get_driver_data(dev);
@@ -131,6 +133,12 @@ static int denali_dt_probe(struct udevice *dev)
 		denali->clk_x_rate = 200000000;
 	}
 
+	ret = reset_get_bulk(bus, &reset_bulk);
+	if (ret)
+		dev_warn(bus, "Cant' get reset: %d\n", ret);
+	else
+		reset_deassert_bulk(&reset_bulk);
+
 	return denali_init(denali);
 }
 
Signature · 1 line
2.17.1
#20 lore raw

PATCH 3/5 mtd: rawnand: denali: add reset handling

Quoted 32 lines · Simon Goldschmidt

This adds reset handling to the devicetree-enabled denalid nand driver.

For backwards compatibility, only a warning is printed when failing to get reset handles.

---

 drivers/mtd/nand/raw/denali_dt.c | 8 ++++++++
 1 file changed, 8 insertions(+)
drivers/mtd/nand/raw/denali_dt.c +2 -0
index d384b974df..21c4540637 100644
@@ -9,6 +9,7 @@
 #include <linux/io.h>
 #include <linux/ioport.h>
 #include <linux/printk.h>
+#include <reset.h>
 
 #include "denali.h"
 
@@ -64,6 +65,7 @@ static int denali_dt_probe(struct udevice *dev)
 	const struct denali_dt_data *data;
 	struct clk clk, clk_x, clk_ecc;
 	struct resource res;
+	struct reset_ctl_bulk reset_bulk;
 	int ret;
 
 	data = (void *)dev_get_driver_data(dev);

drivers/mtd/nand/raw/denali_dt.c static int denali_dt_probe(struct udevice *dev) · line 138

Quoted 6 lines
drivers/mtd/nand/raw/denali_dt.c +3 -0
@@ -131,6 +133,12 @@ static int denali_dt_probe(struct udevice *dev)
 		denali->clk_x_rate = 200000000;
 	}
 
+	ret = reset_get_bulk(bus, &reset_bulk);
+	if (ret)
+		dev_warn(bus, "Cant' get reset: %d\n", ret);

Can't , misplaced "'"

Quoted 7 lines
drivers/mtd/nand/raw/denali_dt.c +3 -0
+	else
+		reset_deassert_bulk(&reset_bulk);
+
 	return denali_init(denali);
 }
 
Signature · 2 lines
Best regards,
Marek Vasut
maintainer
#21 lore raw

PATCH 3/5 mtd: rawnand: denali: add reset handling

In reply to Simon Goldschmidt

Hi Simon,

Simon Goldschmidt <[email protected]> wrote on Fri, 25 Jan 2019 21:30:49 +0100:

Quoted 1 line

This adds reset handling to the devicetree-enabled denalid nand driver.

"Denali NAND driver"

Quoted 31 lines

For backwards compatibility, only a warning is printed when failing to get reset handles.

---

 drivers/mtd/nand/raw/denali_dt.c | 8 ++++++++
 1 file changed, 8 insertions(+)
drivers/mtd/nand/raw/denali_dt.c +2 -0
index d384b974df..21c4540637 100644
@@ -9,6 +9,7 @@
 #include <linux/io.h>
 #include <linux/ioport.h>
 #include <linux/printk.h>
+#include <reset.h>
 
 #include "denali.h"
 
@@ -64,6 +65,7 @@ static int denali_dt_probe(struct udevice *dev)
 	const struct denali_dt_data *data;
 	struct clk clk, clk_x, clk_ecc;
 	struct resource res;
+	struct reset_ctl_bulk reset_bulk;
 	int ret;
 
 	data = (void *)dev_get_driver_data(dev);

drivers/mtd/nand/raw/denali_dt.c static int denali_dt_probe(struct udevice *dev) · line 140

Quoted 8 lines
drivers/mtd/nand/raw/denali_dt.c +5 -0
@@ -131,6 +133,12 @@ static int denali_dt_probe(struct udevice *dev)
 		denali->clk_x_rate = 200000000;
 	}
 
+	ret = reset_get_bulk(bus, &reset_bulk);
+	if (ret)
+		dev_warn(bus, "Cant' get reset: %d\n", ret);
+	else
+		reset_deassert_bulk(&reset_bulk);

Where are reset_get_bulk/reset_deassert_bulk() defined?

drivers/mtd/nand/raw/denali_dt.c static int denali_dt_probe(struct udevice *dev) · line 144

Quoted 4 lines
drivers/mtd/nand/raw/denali_dt.c +1 -0
+
 	return denali_init(denali);
 }
 

Thanks,
Miquèl

#22 lore raw

PATCH 3/5 mtd: rawnand: denali: add reset handling

Hi Miquel,

Quoted 9 lines · Miquel Raynal

Hi Simon,

Simon Goldschmidt <[email protected]> wrote on Fri, 25 Jan 2019 21:30:49 +0100:

Quoted 1 line

This adds reset handling to the devicetree-enabled denalid nand driver.

"Denali NAND driver"

Right.

Quoted 35 lines
Quoted 35 lines

For backwards compatibility, only a warning is printed when failing to get reset handles.

---

 drivers/mtd/nand/raw/denali_dt.c | 8 ++++++++
 1 file changed, 8 insertions(+)
drivers/mtd/nand/raw/denali_dt.c +2 -0
index d384b974df..21c4540637 100644
@@ -9,6 +9,7 @@
 #include <linux/io.h>
 #include <linux/ioport.h>
 #include <linux/printk.h>
+#include <reset.h>
 
 #include "denali.h"
 
@@ -64,6 +65,7 @@ static int denali_dt_probe(struct udevice *dev)
      const struct denali_dt_data *data;
      struct clk clk, clk_x, clk_ecc;
      struct resource res;
+     struct reset_ctl_bulk reset_bulk;
      int ret;
 
      data = (void *)dev_get_driver_data(dev);
@@ -131,6 +133,12 @@ static int denali_dt_probe(struct udevice *dev)
              denali->clk_x_rate = 200000000;
      }
 
Quoted 7 lines

drivers/mtd/nand/raw/denali_dt.c static int denali_dt_probe(struct udevice *dev) · line 140

Quoted 5 lines
drivers/mtd/nand/raw/denali_dt.c +5 -0
+     ret = reset_get_bulk(bus, &reset_bulk);
+     if (ret)
+             dev_warn(bus, "Cant' get reset: %d\n", ret);
+     else
+             reset_deassert_bulk(&reset_bulk);

Where are reset_get_bulk/reset_deassert_bulk() defined?

In reset.h. That's why I added this include (see above)...

Regards,
Simon

Quoted 8 lines

drivers/mtd/nand/raw/denali_dt.c static int denali_dt_probe(struct udevice *dev) · line 144

Quoted 5 lines
drivers/mtd/nand/raw/denali_dt.c +1 -0
+
      return denali_init(denali);
 }
 

Thanks,
Miquèl

maintainer
#23 lore raw

PATCH 3/5 mtd: rawnand: denali: add reset handling

Hi Simon,

Simon Goldschmidt <[email protected]> wrote on Mon, 28 Jan 2019 10:16:28 +0100:

Quoted 59 lines

Hi Miquel,

Quoted 9 lines · Miquel Raynal

Hi Simon,

Simon Goldschmidt <[email protected]> wrote on Fri, 25 Jan 2019 21:30:49 +0100:

Quoted 1 line

This adds reset handling to the devicetree-enabled denalid nand driver.

"Denali NAND driver"

Right.

Quoted 35 lines
Quoted 35 lines

For backwards compatibility, only a warning is printed when failing to get reset handles.

---

 drivers/mtd/nand/raw/denali_dt.c | 8 ++++++++
 1 file changed, 8 insertions(+)
drivers/mtd/nand/raw/denali_dt.c +2 -0
index d384b974df..21c4540637 100644
@@ -9,6 +9,7 @@
 #include <linux/io.h>
 #include <linux/ioport.h>
 #include <linux/printk.h>
+#include <reset.h>
 
 #include "denali.h"
 
@@ -64,6 +65,7 @@ static int denali_dt_probe(struct udevice *dev)
      const struct denali_dt_data *data;
      struct clk clk, clk_x, clk_ecc;
      struct resource res;
+     struct reset_ctl_bulk reset_bulk;
      int ret;
 
      data = (void *)dev_get_driver_data(dev);
@@ -131,6 +133,12 @@ static int denali_dt_probe(struct udevice *dev)
              denali->clk_x_rate = 200000000;
      }
 
Quoted 7 lines

drivers/mtd/nand/raw/denali_dt.c static int denali_dt_probe(struct udevice *dev) · line 140

Quoted 5 lines
drivers/mtd/nand/raw/denali_dt.c +5 -0
+     ret = reset_get_bulk(bus, &reset_bulk);
+     if (ret)
+             dev_warn(bus, "Cant' get reset: %d\n", ret);
+     else
+             reset_deassert_bulk(&reset_bulk);  

Where are reset_get_bulk/reset_deassert_bulk() defined?

Quoted 1 line

In reset.h. That's why I added this include (see above)...

Yes but I guess it is something that will be merged in 5.1? Maybe I will wait for the next release to take this one.

Thanks,
Miquèl

#24 lore raw

PATCH 3/5 mtd: rawnand: denali: add reset handling

Hi Miquel,

Quoted 65 lines · Miquel Raynal

Hi Simon,

Simon Goldschmidt <[email protected]> wrote on Mon, 28 Jan 2019 10:16:28 +0100:

Quoted 59 lines

Hi Miquel,

Quoted 9 lines · Miquel Raynal

Hi Simon,

Simon Goldschmidt <[email protected]> wrote on Fri, 25 Jan 2019 21:30:49 +0100:

Quoted 1 line

This adds reset handling to the devicetree-enabled denalid nand driver.

"Denali NAND driver"

Right.

Quoted 42 lines
Quoted 32 lines

For backwards compatibility, only a warning is printed when failing to get reset handles.

---

 drivers/mtd/nand/raw/denali_dt.c | 8 ++++++++
 1 file changed, 8 insertions(+)
drivers/mtd/nand/raw/denali_dt.c +2 -0
index d384b974df..21c4540637 100644
@@ -9,6 +9,7 @@
 #include <linux/io.h>
 #include <linux/ioport.h>
 #include <linux/printk.h>
+#include <reset.h>
 
 #include "denali.h"
 
@@ -64,6 +65,7 @@ static int denali_dt_probe(struct udevice *dev)
      const struct denali_dt_data *data;
      struct clk clk, clk_x, clk_ecc;
      struct resource res;
+     struct reset_ctl_bulk reset_bulk;
      int ret;
 
      data = (void *)dev_get_driver_data(dev);

drivers/mtd/nand/raw/denali_dt.c static int denali_dt_probe(struct udevice *dev) · line 140

Quoted 8 lines
drivers/mtd/nand/raw/denali_dt.c +5 -0
@@ -131,6 +133,12 @@ static int denali_dt_probe(struct udevice *dev)
              denali->clk_x_rate = 200000000;
      }
 
+     ret = reset_get_bulk(bus, &reset_bulk);
+     if (ret)
+             dev_warn(bus, "Cant' get reset: %d\n", ret);
+     else
+             reset_deassert_bulk(&reset_bulk);

Where are reset_get_bulk/reset_deassert_bulk() defined?

Quoted 4 lines
Quoted 1 line

In reset.h. That's why I added this include (see above)...

Yes but I guess it is something that will be merged in 5.1? Maybe I will wait for the next release to take this one.

You mean Linux 5.1, right? This patch is for U-Boot, not Linux.

Regards,
Simon

maintainer
#25 lore raw

PATCH 3/5 mtd: rawnand: denali: add reset handling

Hi Simon,

Simon Goldschmidt <[email protected]> wrote on Mon, 28 Jan 2019 10:30:12 +0100:

Quoted 69 lines

Hi Miquel,

Quoted 65 lines · Miquel Raynal

Hi Simon,

Simon Goldschmidt <[email protected]> wrote on Mon, 28 Jan 2019 10:16:28 +0100:

Quoted 59 lines

Hi Miquel,

Quoted 9 lines · Miquel Raynal

Hi Simon,

Simon Goldschmidt <[email protected]> wrote on Fri, 25 Jan 2019 21:30:49 +0100:

Quoted 1 line

This adds reset handling to the devicetree-enabled denalid nand driver.

"Denali NAND driver"

Right.

Quoted 42 lines
Quoted 32 lines

For backwards compatibility, only a warning is printed when failing to get reset handles.

---

 drivers/mtd/nand/raw/denali_dt.c | 8 ++++++++
 1 file changed, 8 insertions(+)
drivers/mtd/nand/raw/denali_dt.c +2 -0
index d384b974df..21c4540637 100644
@@ -9,6 +9,7 @@
 #include <linux/io.h>
 #include <linux/ioport.h>
 #include <linux/printk.h>
+#include <reset.h>
 
 #include "denali.h"
 
@@ -64,6 +65,7 @@ static int denali_dt_probe(struct udevice *dev)
      const struct denali_dt_data *data;
      struct clk clk, clk_x, clk_ecc;
      struct resource res;
+     struct reset_ctl_bulk reset_bulk;
      int ret;
 
      data = (void *)dev_get_driver_data(dev);

drivers/mtd/nand/raw/denali_dt.c static int denali_dt_probe(struct udevice *dev) · line 140

Quoted 8 lines
drivers/mtd/nand/raw/denali_dt.c +5 -0
@@ -131,6 +133,12 @@ static int denali_dt_probe(struct udevice *dev)
              denali->clk_x_rate = 200000000;
      }
 
+     ret = reset_get_bulk(bus, &reset_bulk);
+     if (ret)
+             dev_warn(bus, "Cant' get reset: %d\n", ret);
+     else
+             reset_deassert_bulk(&reset_bulk);  

Where are reset_get_bulk/reset_deassert_bulk() defined?

Quoted 6 lines
Quoted 4 lines
Quoted 1 line

In reset.h. That's why I added this include (see above)...

Yes but I guess it is something that will be merged in 5.1? Maybe I will wait for the next release to take this one.

You mean Linux 5.1, right? This patch is for U-Boot, not Linux.

My bad, I had SIEVE rules issues, the patch did not got flagged for "U-Boot". Sorry for the misunderstanding.

Thanks,
Miquèl

#26 lore raw

PATCH 4/5 spi: cadence_qspi: add reset handling

In reply to Simon Goldschmidt

This adds reset handling to the devicetree-enabled denalid nand driver.

For backwards compatibility, only a warning is printed when failing to get reset handles.

---

 drivers/spi/cadence_qspi.c | 9 +++++++++
 1 file changed, 9 insertions(+)
drivers/spi/cadence_qspi.c +9 -0
index 11fce9c4fe..f68c827e6d 100644
@@ -8,6 +8,7 @@
 #include <dm.h>
 #include <fdtdec.h>
 #include <malloc.h>
+#include <reset.h>
 #include <spi.h>
 #include <linux/errno.h>
 #include "cadence_qspi.h"
@@ -154,10 +155,18 @@ static int cadence_spi_probe(struct udevice *bus)
 {
 	struct cadence_spi_platdata *plat = bus->platdata;
 	struct cadence_spi_priv *priv = dev_get_priv(bus);
+	struct reset_ctl_bulk reset_bulk;
+	int ret;
 
 	priv->regbase = plat->regbase;
 	priv->ahbbase = plat->ahbbase;
 
+	ret = reset_get_bulk(bus, &reset_bulk);
+	if (ret)
+		dev_warn(bus, "Cant' get reset: %d\n", ret);
+	else
+		reset_deassert_bulk(&reset_bulk);
+
 	if (!priv->qspi_is_init) {
 		cadence_qspi_apb_controller_init(plat);
 		priv->qspi_is_init = 1;
Signature · 1 line
2.17.1
#27 lore raw

PATCH 4/5 spi: cadence_qspi: add reset handling

Quoted 26 lines · Simon Goldschmidt

This adds reset handling to the devicetree-enabled denalid nand driver.

For backwards compatibility, only a warning is printed when failing to get reset handles.

---

 drivers/spi/cadence_qspi.c | 9 +++++++++
 1 file changed, 9 insertions(+)
drivers/spi/cadence_qspi.c +1 -0
index 11fce9c4fe..f68c827e6d 100644
@@ -8,6 +8,7 @@
 #include <dm.h>
 #include <fdtdec.h>
 #include <malloc.h>
+#include <reset.h>
 #include <spi.h>
 #include <linux/errno.h>
 #include "cadence_qspi.h"
@@ -154,10 +155,18 @@ static int cadence_spi_probe(struct udevice *bus)
 {
 	struct cadence_spi_platdata *plat = bus->platdata;

drivers/spi/cadence_qspi.c static int cadence_spi_probe(struct udevice *bus) · line 166

Quoted 10 lines
drivers/spi/cadence_qspi.c +5 -0
 	struct cadence_spi_priv *priv = dev_get_priv(bus);
+	struct reset_ctl_bulk reset_bulk;
+	int ret;
 
 	priv->regbase = plat->regbase;
 	priv->ahbbase = plat->ahbbase;
 
+	ret = reset_get_bulk(bus, &reset_bulk);
+	if (ret)
+		dev_warn(bus, "Cant' get reset: %d\n", ret);

Can't

Also, shouldn't there be some reset_release or reset_free somewhere ? What about .remove(), shouldn't that release the reset ?

Quoted 7 lines
drivers/spi/cadence_qspi.c +3 -0
+	else
+		reset_deassert_bulk(&reset_bulk);
+
 	if (!priv->qspi_is_init) {
 		cadence_qspi_apb_controller_init(plat);
 		priv->qspi_is_init = 1;
Signature · 2 lines
Best regards,
Marek Vasut
#28 lore raw

PATCH 4/5 spi: cadence_qspi: add reset handling

Am 26.01.2019 um 09:59 schrieb Marek Vasut:

Quoted 31 lines · Simon Goldschmidt
Quoted 30 lines

This adds reset handling to the devicetree-enabled denalid nand driver.

For backwards compatibility, only a warning is printed when failing to get reset handles.

---

  drivers/spi/cadence_qspi.c | 9 +++++++++
  1 file changed, 9 insertions(+)
drivers/spi/cadence_qspi.c +3 -0
index 11fce9c4fe..f68c827e6d 100644
@@ -8,6 +8,7 @@
  #include <dm.h>
  #include <fdtdec.h>
  #include <malloc.h>
+#include <reset.h>
  #include <spi.h>
  #include <linux/errno.h>
  #include "cadence_qspi.h"
@@ -154,10 +155,18 @@ static int cadence_spi_probe(struct udevice *bus)
  {
  	struct cadence_spi_platdata *plat = bus->platdata;
  	struct cadence_spi_priv *priv = dev_get_priv(bus);
+	struct reset_ctl_bulk reset_bulk;
+	int ret;
 
Quoted 8 lines

drivers/spi/cadence_qspi.c static int cadence_spi_probe(struct udevice *bus) · line 166

Quoted 6 lines
drivers/spi/cadence_qspi.c +3 -0
  	priv->regbase = plat->regbase;
  	priv->ahbbase = plat->ahbbase;
 
+	ret = reset_get_bulk(bus, &reset_bulk);
+	if (ret)
+		dev_warn(bus, "Cant' get reset: %d\n", ret);

Can't

Right. Copy & paste from 3/5...

Quoted 3 lines

Also, shouldn't there be some reset_release or reset_free somewhere ? What about .remove(), shouldn't that release the reset ?

Ehrm, well... this is my first driver using the reset framework. I haven't though of remove. I'll do a V2 for that soon.

Regards,
Simon

Quoted 10 lines
Quoted 8 lines
drivers/spi/cadence_qspi.c +3 -0
+	else
+		reset_deassert_bulk(&reset_bulk);
+
  	if (!priv->qspi_is_init) {
  		cadence_qspi_apb_controller_init(plat);
  		priv->qspi_is_init = 1;
#29 lore raw

PATCH 4/5 spi: cadence_qspi: add reset handling

Quoted 39 lines · Simon Goldschmidt

Am 26.01.2019 um 09:59 schrieb Marek Vasut:

Quoted 29 lines · Simon Goldschmidt
Quoted 28 lines

This adds reset handling to the devicetree-enabled denalid nand driver.

For backwards compatibility, only a warning is printed when failing to get reset handles.

---

  drivers/spi/cadence_qspi.c | 9 +++++++++

  1 file changed, 9 insertions(+)
drivers/spi/cadence_qspi.c +2 -0
index 11fce9c4fe..f68c827e6d 100644
@@ -8,6 +8,7 @@
   #include <dm.h>
   #include <fdtdec.h>
   #include <malloc.h>
+#include <reset.h>
   #include <spi.h>
   #include <linux/errno.h>
   #include "cadence_qspi.h"
@@ -154,10 +155,18 @@ static int cadence_spi_probe(struct udevice *bus)
   {
       struct cadence_spi_platdata *plat = bus->platdata;
       struct cadence_spi_priv *priv = dev_get_priv(bus);
+    struct reset_ctl_bulk reset_bulk;
Quoted 8 lines

drivers/spi/cadence_qspi.c static int cadence_spi_probe(struct udevice *bus) · line 166

Quoted 6 lines
drivers/spi/cadence_qspi.c +3 -0
+    int ret;
         priv->regbase = plat->regbase;
       priv->ahbbase = plat->ahbbase;
   +    ret = reset_get_bulk(bus, &reset_bulk);
+    if (ret)
+        dev_warn(bus, "Cant' get reset: %d\n", ret);

Can't

Quoted 8 lines

Right. Copy & paste from 3/5...

Quoted 3 lines

Also, shouldn't there be some reset_release or reset_free somewhere ? What about .remove(), shouldn't that release the reset ?

Ehrm, well... this is my first driver using the reset framework. I haven't though of remove. I'll do a V2 for that soon.

Thanks

Signature · 2 lines
Best regards,
Marek Vasut
#30 lore raw

PATCH 5/5 arm: socfpga: implement proper peripheral reset

In reply to Simon Goldschmidt

This commit removes ad-hoc reset handling for peripheral resets from SPL for socfpga gen5.

This is done because as U-Boot drivers support reset handling by now.

For kernels that don't support taking peripherals out of reset that are not enabled by U-Boot, a new Kconfig option "OLD_SOCFPGA_KERNEL_COMPAT" is added, which keeps the old behaviour of just enabling all peripherals.

This new option is enabled by default for now, as even Linux 4.20 does not support reset handling on all peripherals.

---

 arch/arm/mach-socfpga/Kconfig     | 10 ++++++++++
 arch/arm/mach-socfpga/misc_gen5.c |  2 ++
 arch/arm/mach-socfpga/spl_gen5.c  | 10 ++++++++++
 3 files changed, 22 insertions(+)
arch/arm/mach-socfpga/Kconfig +10 -0
index 5e87371f8c..89acced8d8 100644
@@ -142,4 +142,14 @@ config SYS_CONFIG_NAME
 	default "socfpga_stratix10_socdk" if TARGET_SOCFPGA_STRATIX10_SOCDK
 	default "socfpga_vining_fpga" if TARGET_SOCFPGA_SAMTEC_VINING_FPGA
 
+config OLD_SOCFPGA_KERNEL_COMPAT
+	bool "Enable workarounds for booting old kernels"
+	depends on TARGET_SOCFPGA_GEN5
+	default y
+	help
+	  Set this to enable various workarounds for old kernels (e.g. take all
+	  peripherals out of reset because old kernels cannot handle reset).
+	  This results in sub-optimal settings for newer kernels, only enable
+	  if needed.
+
 endif
arch/arm/mach-socfpga/misc_gen5.c +2 -0
index 04f237d100..168669923e 100644
@@ -243,6 +243,7 @@ int arch_early_init_r(void)
 	/* Add device descriptor to FPGA device table */
 	socfpga_fpga_add(&altera_fpga[0]);
 
+#ifdef CONFIG_OLD_SOCFPGA_KERNEL_COMPAT
 #ifdef CONFIG_DESIGNWARE_SPI
 	/* Get Designware SPI controller out of reset */
 	socfpga_per_reset(SOCFPGA_RESET(SPIM0), 0);
@@ -252,6 +253,7 @@ int arch_early_init_r(void)
 #ifdef CONFIG_NAND_DENALI
 	socfpga_per_reset(SOCFPGA_RESET(NAND), 0);
 #endif
+#endif /* CONFIG_OLD_SOCFPGA_KERNEL_COMPAT */
 
 	return 0;
 }
arch/arm/mach-socfpga/spl_gen5.c +10 -0
index f9bea892b1..19a211256d 100644
@@ -35,16 +35,22 @@ u32 spl_boot_device(void)
 		return BOOT_DEVICE_RAM;
 	case 0x2:	/* NAND Flash (1.8V) */
 	case 0x3:	/* NAND Flash (3.0V) */
+#ifdef CONFIG_OLD_SOCFPGA_KERNEL_COMPAT
 		socfpga_per_reset(SOCFPGA_RESET(NAND), 0);
+#endif
 		return BOOT_DEVICE_NAND;
 	case 0x4:	/* SD/MMC External Transceiver (1.8V) */
 	case 0x5:	/* SD/MMC Internal Transceiver (3.0V) */
+#ifdef CONFIG_OLD_SOCFPGA_KERNEL_COMPAT
 		socfpga_per_reset(SOCFPGA_RESET(SDMMC), 0);
 		socfpga_per_reset(SOCFPGA_RESET(DMA), 0);
+#endif
 		return BOOT_DEVICE_MMC1;
 	case 0x6:	/* QSPI Flash (1.8V) */
 	case 0x7:	/* QSPI Flash (3.0V) */
+#ifdef CONFIG_OLD_SOCFPGA_KERNEL_COMPAT
 		socfpga_per_reset(SOCFPGA_RESET(QSPI), 0);
+#endif
 		return BOOT_DEVICE_SPI;
 	default:
 		printf("Invalid boot device (bsel=%08x)!\n", bsel);
@@ -98,7 +104,9 @@ void board_init_f(ulong dummy)
 		socfpga_bridges_reset(1);
 	}
 
+#ifdef CONFIG_OLD_SOCFPGA_KERNEL_COMPAT
 	socfpga_per_reset(SOCFPGA_RESET(UART0), 0);
+#endif
 	socfpga_per_reset(SOCFPGA_RESET(OSC1TIMER0), 0);
 
 	timer_init();
@@ -122,9 +130,11 @@ void board_init_f(ulong dummy)
 	sysmgr_pinmux_init();
 	sysmgr_config_warmrstcfgio(0);
 
+#ifdef CONFIG_OLD_SOCFPGA_KERNEL_COMPAT
 	/* De-assert reset for peripherals and bridges based on handoff */
 	reset_deassert_peripherals_handoff();
 	socfpga_bridges_reset(0);
+#endif
 
 	debug("Unfreezing/Thaw all I/O banks\n");
 	/* unfreeze / thaw all IO banks */
Signature · 1 line
2.17.1
#31 lore raw

PATCH 5/5 arm: socfpga: implement proper peripheral reset

Quoted 5 lines · Simon Goldschmidt

This commit removes ad-hoc reset handling for peripheral resets from SPL for socfpga gen5.

This is done because as U-Boot drivers support reset handling by now.

Quoted 8 lines

For kernels that don't support taking peripherals out of reset that are not enabled by U-Boot, a new Kconfig option "OLD_SOCFPGA_KERNEL_COMPAT" is added, which keeps the old behaviour of just enabling all peripherals.

This new option is enabled by default for now, as even Linux 4.20 does not support reset handling on all peripherals.

Well ew, don't we have some pre-linux-boot hook which could just flip the reset register bits in one place ? And also, I'd prefer this to be runtime-configurable, not compile-time configurable.

Quoted 96 lines

---

 arch/arm/mach-socfpga/Kconfig     | 10 ++++++++++
 arch/arm/mach-socfpga/misc_gen5.c |  2 ++
 arch/arm/mach-socfpga/spl_gen5.c  | 10 ++++++++++
 3 files changed, 22 insertions(+)
arch/arm/mach-socfpga/Kconfig +10 -0
index 5e87371f8c..89acced8d8 100644
@@ -142,4 +142,14 @@ config SYS_CONFIG_NAME
 	default "socfpga_stratix10_socdk" if TARGET_SOCFPGA_STRATIX10_SOCDK
 	default "socfpga_vining_fpga" if TARGET_SOCFPGA_SAMTEC_VINING_FPGA
 
+config OLD_SOCFPGA_KERNEL_COMPAT
+	bool "Enable workarounds for booting old kernels"
+	depends on TARGET_SOCFPGA_GEN5
+	default y
+	help
+	  Set this to enable various workarounds for old kernels (e.g. take all
+	  peripherals out of reset because old kernels cannot handle reset).
+	  This results in sub-optimal settings for newer kernels, only enable
+	  if needed.
+
 endif
arch/arm/mach-socfpga/misc_gen5.c +2 -0
index 04f237d100..168669923e 100644
@@ -243,6 +243,7 @@ int arch_early_init_r(void)
 	/* Add device descriptor to FPGA device table */
 	socfpga_fpga_add(&altera_fpga[0]);
 
+#ifdef CONFIG_OLD_SOCFPGA_KERNEL_COMPAT
 #ifdef CONFIG_DESIGNWARE_SPI
 	/* Get Designware SPI controller out of reset */
 	socfpga_per_reset(SOCFPGA_RESET(SPIM0), 0);
@@ -252,6 +253,7 @@ int arch_early_init_r(void)
 #ifdef CONFIG_NAND_DENALI
 	socfpga_per_reset(SOCFPGA_RESET(NAND), 0);
 #endif
+#endif /* CONFIG_OLD_SOCFPGA_KERNEL_COMPAT */
 
 	return 0;
 }
arch/arm/mach-socfpga/spl_gen5.c +10 -0
index f9bea892b1..19a211256d 100644
@@ -35,16 +35,22 @@ u32 spl_boot_device(void)
 		return BOOT_DEVICE_RAM;
 	case 0x2:	/* NAND Flash (1.8V) */
 	case 0x3:	/* NAND Flash (3.0V) */
+#ifdef CONFIG_OLD_SOCFPGA_KERNEL_COMPAT
 		socfpga_per_reset(SOCFPGA_RESET(NAND), 0);
+#endif
 		return BOOT_DEVICE_NAND;
 	case 0x4:	/* SD/MMC External Transceiver (1.8V) */
 	case 0x5:	/* SD/MMC Internal Transceiver (3.0V) */
+#ifdef CONFIG_OLD_SOCFPGA_KERNEL_COMPAT
 		socfpga_per_reset(SOCFPGA_RESET(SDMMC), 0);
 		socfpga_per_reset(SOCFPGA_RESET(DMA), 0);
+#endif
 		return BOOT_DEVICE_MMC1;
 	case 0x6:	/* QSPI Flash (1.8V) */
 	case 0x7:	/* QSPI Flash (3.0V) */
+#ifdef CONFIG_OLD_SOCFPGA_KERNEL_COMPAT
 		socfpga_per_reset(SOCFPGA_RESET(QSPI), 0);
+#endif
 		return BOOT_DEVICE_SPI;
 	default:
 		printf("Invalid boot device (bsel=%08x)!\n", bsel);
@@ -98,7 +104,9 @@ void board_init_f(ulong dummy)
 		socfpga_bridges_reset(1);
 	}
 
+#ifdef CONFIG_OLD_SOCFPGA_KERNEL_COMPAT
 	socfpga_per_reset(SOCFPGA_RESET(UART0), 0);
+#endif
 	socfpga_per_reset(SOCFPGA_RESET(OSC1TIMER0), 0);
 
 	timer_init();
@@ -122,9 +130,11 @@ void board_init_f(ulong dummy)
 	sysmgr_pinmux_init();
 	sysmgr_config_warmrstcfgio(0);
 
+#ifdef CONFIG_OLD_SOCFPGA_KERNEL_COMPAT
 	/* De-assert reset for peripherals and bridges based on handoff */
 	reset_deassert_peripherals_handoff();
 	socfpga_bridges_reset(0);
+#endif
 
 	debug("Unfreezing/Thaw all I/O banks\n");
 	/* unfreeze / thaw all IO banks */
Signature · 2 lines
Best regards,
Marek Vasut
#32 lore raw

PATCH 5/5 arm: socfpga: implement proper peripheral reset

Am 26.01.2019 um 10:00 schrieb Marek Vasut:

Quoted 10 lines · Simon Goldschmidt
Quoted 9 lines

This commit removes ad-hoc reset handling for peripheral resets from SPL for socfpga gen5.

This is done because as U-Boot drivers support reset handling by now.

For kernels that don't support taking peripherals out of reset that are not enabled by U-Boot, a new Kconfig option "OLD_SOCFPGA_KERNEL_COMPAT" is added, which keeps the old behaviour of just enabling all peripherals.

Quoted 8 lines
Quoted 4 lines

This new option is enabled by default for now, as even Linux 4.20 does not support reset handling on all peripherals.

Well ew, don't we have some pre-linux-boot hook which could just flip the reset register bits in one place ? And also, I'd prefer this to be runtime-configurable, not compile-time configurable.

And this will be the biggest change for v2. It's also why I needed so long to send this series. What I sent in this version is how Tom said the same thing is being handled for mach-sunxi (Kconfig "OLD_SUNXI_KERNEL_COMPAT").

So if I remember your last mails about this topic, you'd prefer an environment variable with a callback that leads to enabling all peripherals just before jumping to Linux (yes, I'll find a hook for that).

I see two problems here:
- enabling all peripherals will need to be the default for some time, until Linux supports reset handling in all drivers. How do you want to deal with that?
- is this scheme used somewhere else for compatibility? I think it would make sense for multiple architectures to use the same way to handle such compatibility issues...

Regards,
Simon

Quoted 99 lines
Quoted 97 lines

---

  arch/arm/mach-socfpga/Kconfig     | 10 ++++++++++
  arch/arm/mach-socfpga/misc_gen5.c |  2 ++
  arch/arm/mach-socfpga/spl_gen5.c  | 10 ++++++++++
  3 files changed, 22 insertions(+)
arch/arm/mach-socfpga/Kconfig +10 -0
index 5e87371f8c..89acced8d8 100644
@@ -142,4 +142,14 @@ config SYS_CONFIG_NAME
  	default "socfpga_stratix10_socdk" if TARGET_SOCFPGA_STRATIX10_SOCDK
  	default "socfpga_vining_fpga" if TARGET_SOCFPGA_SAMTEC_VINING_FPGA
 
+config OLD_SOCFPGA_KERNEL_COMPAT
+	bool "Enable workarounds for booting old kernels"
+	depends on TARGET_SOCFPGA_GEN5
+	default y
+	help
+	  Set this to enable various workarounds for old kernels (e.g. take all
+	  peripherals out of reset because old kernels cannot handle reset).
+	  This results in sub-optimal settings for newer kernels, only enable
+	  if needed.
+
  endif
arch/arm/mach-socfpga/misc_gen5.c +2 -0
index 04f237d100..168669923e 100644
@@ -243,6 +243,7 @@ int arch_early_init_r(void)
  	/* Add device descriptor to FPGA device table */
  	socfpga_fpga_add(&altera_fpga[0]);
 
+#ifdef CONFIG_OLD_SOCFPGA_KERNEL_COMPAT
  #ifdef CONFIG_DESIGNWARE_SPI
  	/* Get Designware SPI controller out of reset */
  	socfpga_per_reset(SOCFPGA_RESET(SPIM0), 0);
@@ -252,6 +253,7 @@ int arch_early_init_r(void)
  #ifdef CONFIG_NAND_DENALI
  	socfpga_per_reset(SOCFPGA_RESET(NAND), 0);
  #endif
+#endif /* CONFIG_OLD_SOCFPGA_KERNEL_COMPAT */
 
  	return 0;
  }
arch/arm/mach-socfpga/spl_gen5.c +10 -0
index f9bea892b1..19a211256d 100644
@@ -35,16 +35,22 @@ u32 spl_boot_device(void)
  		return BOOT_DEVICE_RAM;
  	case 0x2:	/* NAND Flash (1.8V) */
  	case 0x3:	/* NAND Flash (3.0V) */
+#ifdef CONFIG_OLD_SOCFPGA_KERNEL_COMPAT
  		socfpga_per_reset(SOCFPGA_RESET(NAND), 0);
+#endif
  		return BOOT_DEVICE_NAND;
  	case 0x4:	/* SD/MMC External Transceiver (1.8V) */
  	case 0x5:	/* SD/MMC Internal Transceiver (3.0V) */
+#ifdef CONFIG_OLD_SOCFPGA_KERNEL_COMPAT
  		socfpga_per_reset(SOCFPGA_RESET(SDMMC), 0);
  		socfpga_per_reset(SOCFPGA_RESET(DMA), 0);
+#endif
  		return BOOT_DEVICE_MMC1;
  	case 0x6:	/* QSPI Flash (1.8V) */
  	case 0x7:	/* QSPI Flash (3.0V) */
+#ifdef CONFIG_OLD_SOCFPGA_KERNEL_COMPAT
  		socfpga_per_reset(SOCFPGA_RESET(QSPI), 0);
+#endif
  		return BOOT_DEVICE_SPI;
  	default:
  		printf("Invalid boot device (bsel=%08x)!\n", bsel);
@@ -98,7 +104,9 @@ void board_init_f(ulong dummy)
  		socfpga_bridges_reset(1);
  	}
 
+#ifdef CONFIG_OLD_SOCFPGA_KERNEL_COMPAT
  	socfpga_per_reset(SOCFPGA_RESET(UART0), 0);
+#endif
  	socfpga_per_reset(SOCFPGA_RESET(OSC1TIMER0), 0);
 
  	timer_init();
@@ -122,9 +130,11 @@ void board_init_f(ulong dummy)
  	sysmgr_pinmux_init();
  	sysmgr_config_warmrstcfgio(0);
 
+#ifdef CONFIG_OLD_SOCFPGA_KERNEL_COMPAT
  	/* De-assert reset for peripherals and bridges based on handoff */
  	reset_deassert_peripherals_handoff();
  	socfpga_bridges_reset(0);
+#endif
 
  	debug("Unfreezing/Thaw all I/O banks\n");
  	/* unfreeze / thaw all IO banks */
#33 lore raw

PATCH 5/5 arm: socfpga: implement proper peripheral reset

Quoted 22 lines · Simon Goldschmidt

Am 26.01.2019 um 10:00 schrieb Marek Vasut:

Quoted 11 lines · Simon Goldschmidt
Quoted 10 lines

This commit removes ad-hoc reset handling for peripheral resets from SPL for socfpga gen5.

This is done because as U-Boot drivers support reset handling by now.

For kernels that don't support taking peripherals out of reset that are not enabled by U-Boot, a new Kconfig option "OLD_SOCFPGA_KERNEL_COMPAT" is added, which keeps the old behaviour of just enabling all peripherals.

Quoted 8 lines
Quoted 4 lines

This new option is enabled by default for now, as even Linux 4.20 does not support reset handling on all peripherals.

Well ew, don't we have some pre-linux-boot hook which could just flip the reset register bits in one place ? And also, I'd prefer this to be runtime-configurable, not compile-time configurable.

Quoted 8 lines

And this will be the biggest change for v2. It's also why I needed so long to send this series. What I sent in this version is how Tom said the same thing is being handled for mach-sunxi (Kconfig "OLD_SUNXI_KERNEL_COMPAT").

So if I remember your last mails about this topic, you'd prefer an environment variable with a callback that leads to enabling all peripherals just before jumping to Linux (yes, I'll find a hook for that).

Rather, I'd prefer some .remove() code in the SoCFPGA reset driver which would check an environment variable to determine whether it should ungate the resets before booting Linux or not.

Quoted 4 lines

I see two problems here:
- enabling all peripherals will need to be the default for some time, until Linux supports reset handling in all drivers. How do you want to deal with that?

Set the env variable to default=1 for all socfpga boards ?

Quoted 3 lines

- is this scheme used somewhere else for compatibility? I think it would make sense for multiple architectures to use the same way to handle such compatibility issues...

I don't think so, however it's better than compile-time setting.

[...]

Signature · 2 lines
Best regards,
Marek Vasut
#34 lore raw

PATCH 5/5 arm: socfpga: implement proper peripheral reset

Quoted 29 lines · Marek Vasut
Quoted 27 lines · Simon Goldschmidt

Am 26.01.2019 um 10:00 schrieb Marek Vasut:

Quoted 11 lines · Simon Goldschmidt
Quoted 10 lines

This commit removes ad-hoc reset handling for peripheral resets from SPL for socfpga gen5.

This is done because as U-Boot drivers support reset handling by now.

For kernels that don't support taking peripherals out of reset that are not enabled by U-Boot, a new Kconfig option "OLD_SOCFPGA_KERNEL_COMPAT" is added, which keeps the old behaviour of just enabling all peripherals.

Quoted 8 lines
Quoted 4 lines

This new option is enabled by default for now, as even Linux 4.20 does not support reset handling on all peripherals.

Well ew, don't we have some pre-linux-boot hook which could just flip the reset register bits in one place ? And also, I'd prefer this to be runtime-configurable, not compile-time configurable.

And this will be the biggest change for v2. It's also why I needed so long to send this series. What I sent in this version is how Tom said the same thing is being handled for mach-sunxi (Kconfig "OLD_SUNXI_KERNEL_COMPAT").

Quoted 7 lines
Quoted 3 lines

So if I remember your last mails about this topic, you'd prefer an environment variable with a callback that leads to enabling all peripherals just before jumping to Linux (yes, I'll find a hook for that).

Rather, I'd prefer some .remove() code in the SoCFPGA reset driver which would check an environment variable to determine whether it should ungate the resets before booting Linux or not.

OK, that should work. I'll work on that for v2.

Regards,
Simon

Quoted 19 lines
Quoted 5 lines

I see two problems here:
- enabling all peripherals will need to be the default for some time, until Linux supports reset handling in all drivers. How do you want to deal with that?

Set the env variable to default=1 for all socfpga boards ?

Quoted 3 lines

- is this scheme used somewhere else for compatibility? I think it would make sense for multiple architectures to use the same way to handle such compatibility issues...

I don't think so, however it's better than compile-time setting.

[...]

--
Best regards,
Marek Vasut