Skip to contentSkip to conversation

spi: spi-mem: Convert Aspeed SMC driver to spi-mem

PATCH In review Cédric Le Goater 4 participants 19 messages May 9, 2022 – May 17, 2022

Subsystems (6)
lore
Share & tools
lore

Cédric Le Goater

author maintainer
#0 lore raw

v7 cover spi: spi-mem: Convert Aspeed SMC driver to spi-mem

Hi,

This series adds a new SPI driver using the spi-mem interface for the Aspeed static memory controllers of the AST2600, AST2500 and AST2400 SoCs.

* AST2600 Firmware SPI Memory Controller (FMC)
* AST2600 SPI Flash Controller (SPI1 and SPI2)
* AST2500 Firmware SPI Memory Controller (FMC)
* AST2500 SPI Flash Controller (SPI1 and SPI2)
* AST2400 New Static Memory Controller (also referred as FMC)
* AST2400 SPI Flash Controller (SPI)

It is based on the current OpenBMC kernel driver [1], using directly the MTD SPI-NOR interface and on a patchset [2] previously proposed adding support for the AST2600 only. This driver takes a slightly different approach to cover all 6 controllers.

It does not make use of the controller register disabling Address and Data byte lanes because is not available on the AST2400 SoC. We could introduce a specific handler for new features available on recent SoCs if needed. As there is not much difference on performance, the driver chooses the common denominator: "User mode" which has been heavily tested in [1]. "User mode" is also used as a fall back method when flash device mapping window is too small.

Problems to address with spi-mem were the configuration of the mapping windows and the calibration of the read timings. The driver handles them in the direct mapping handler when some knowledge on the size of the flash device is know. It is not perfect but not incorrect either. The algorithm is one from [1] because it doesn't require the DMA registers which are not available on all controllers.

Direct mapping for writes is not supported (yet). I have seen some corruption with writes and I preferred to use the safer and proven method of the initial driver [1]. We can improve that later.

The driver supports Quad SPI RX transfers on the AST2600 SoC but it didn't have the expected results. Therefore it is not activated yet. There are some issues on the pinctrl to investigate first.

Tested on:

* OpenPOWER Palmetto (AST2400)
* Facebook Wedge 100 BMC (AST2400) by Tao Ren <[email protected]>
* Evaluation board (AST2500)
* Inspur FP5280G2 BMC (AST2500) by John Wang <[email protected]>
* Facebook Backpack CMM BMC (AST2500) by Tao Ren <[email protected]>
* OpenPOWER Witherspoon (AST2500)
* Evaluation board (AST2600 A0 and A3)
* Rainier board (AST2600)

[1] https://github.com/openbmc/linux/blob/dev-5.15/drivers/mtd/spi-nor/controllers/aspeed-smc.c
[2] https://patchwork.ozlabs.org/project/linux-aspeed/list/?series=212394

Thanks,

C.

Changes in v7:

- Moved removal of the old binding desccription in the binding patch

Changes in v6:

- Removed tabs for aspeed,ast2600-fmc.yaml

Changes in v5:

- Rebased on 5.18-rc5
- More AST2600 tests from Jae Hyun Yoo
- Cleanups of aspeed,ast2600-fmc.yaml
- Modified aspeed_spi_send_cmd_addr() routine to return an error
- Simplified conditions in exec_op() handler when computing the controller setting
- Dropped the use of memcpy_fromio for the SFDP address space

Changes in v4:

- Rebased on 5.18
- Removal of the SPI-NOR base driver (we had enough tests)
- Fix for small size flash devices on AST2600 (Potin)

Changes in v3:

- Fixed compile warning on aspeed_spi_dirmap_read() prototype reported by kernel test robot
- Removed unnecessary entry in ast2600-fmc.yaml
- New patch from Tao to set spi-max-frequency on all FMC devices

Changes in v2:

- Fixed dt_binding_check warnings (Rob)
- New entry in MAINTAINERS
- Addressed Lukas comments regarding the SPI controller registration and device removal. Checked with driver bind/unbind
- Introduced setup and cleanup handlers and removed routine looping on the DT children properties (Pratyush)
- Clarified in commit log requirements for training.
- Removed defconfig changes of patch 1 since they were reverted in the last patch (Joel)

C?dric Le Goater (9):
ARM: dts: aspeed: Adjust "reg" property of FMC/SPI controllers
dt-bindings: spi: Convert the Aspeed SMC controllers device tree
binding
spi: spi-mem: Convert Aspeed SMC driver to spi-mem
spi: aspeed: Add support for direct mapping
spi: aspeed: Adjust direct mapping to device size
spi: aspeed: Workaround AST2500 limitations
spi: aspeed: Add support for the AST2400 SPI controller
spi: aspeed: Calibrate read timings
ARM: dts: aspeed: Enable Dual SPI RX transfers

Potin Lai (1):
mtd: spi-nor: aspeed: set the decoding size to at least 2MB for

    AST2600

Tao Ren (1):
ARM: dts: aspeed-g4: Set spi-max-frequency for all flashes

 drivers/mtd/spi-nor/controllers/aspeed-smc.c  |  921 -------------
 drivers/spi/spi-aspeed-smc.c                  | 1210 +++++++++++++++++
 .../devicetree/bindings/mtd/aspeed-smc.txt    |   51 -
 .../bindings/spi/aspeed,ast2600-fmc.yaml      |   82 ++
 MAINTAINERS                                   |   10 +
 arch/arm/boot/dts/aspeed-g4.dtsi              |   16 +-
 arch/arm/boot/dts/aspeed-g5.dtsi              |   16 +-
 arch/arm/boot/dts/aspeed-g6.dtsi              |   17 +-
 drivers/mtd/spi-nor/controllers/Kconfig       |   10 -
 drivers/mtd/spi-nor/controllers/Makefile      |    1 -
 drivers/spi/Kconfig                           |   11 +
 drivers/spi/Makefile                          |    1 +
 12 files changed, 1347 insertions(+), 999 deletions(-)
 delete mode 100644 drivers/mtd/spi-nor/controllers/aspeed-smc.c
 create mode 100644 drivers/spi/spi-aspeed-smc.c
 delete mode 100644 Documentation/devicetree/bindings/mtd/aspeed-smc.txt
 create mode 100644 Documentation/devicetree/bindings/spi/aspeed,ast2600-fmc.yaml
Signature · 1 line
2.35.1
author maintainer
#1 lore raw

PATCH v7 1/11 ARM: dts: aspeed: Adjust "reg" property of FMC/SPI controllers

This is compatible with the current driver and addresses issues when running 'make dt_binding_check'.

---
 arch/arm/boot/dts/aspeed-g4.dtsi | 6 ++----
 arch/arm/boot/dts/aspeed-g5.dtsi | 9 +++------
 arch/arm/boot/dts/aspeed-g6.dtsi | 9 +++------
 3 files changed, 8 insertions(+), 16 deletions(-)
arch/arm/boot/dts/aspeed-g4.dtsi +2 -4
index fa8b581c3d6c..3b7b98e787e7 100644
@@ -54,8 +54,7 @@ ahb {
 		ranges;
 
 		fmc: spi at 1e620000 {
-			reg = < 0x1e620000 0x94
-				0x20000000 0x10000000 >;
+			reg = <0x1e620000 0x94>, <0x20000000 0x10000000>;
 			#address-cells = <1>;
 			#size-cells = <0>;
 			compatible = "aspeed,ast2400-fmc";
@@ -91,8 +90,7 @@ flash at 4 {
 		};
 
 		spi: spi at 1e630000 {
-			reg = < 0x1e630000 0x18
-				0x30000000 0x10000000 >;
+			reg = <0x1e630000 0x18>, <0x30000000 0x10000000>;
 			#address-cells = <1>;
 			#size-cells = <0>;
 			compatible = "aspeed,ast2400-spi";
arch/arm/boot/dts/aspeed-g5.dtsi +3 -6
index 4147b397c883..1e727cbaab14 100644
@@ -55,8 +55,7 @@ ahb {
 		ranges;
 
 		fmc: spi at 1e620000 {
-			reg = < 0x1e620000 0xc4
-				0x20000000 0x10000000 >;
+			reg = <0x1e620000 0xc4>, <0x20000000 0x10000000>;
 			#address-cells = <1>;
 			#size-cells = <0>;
 			compatible = "aspeed,ast2500-fmc";
@@ -84,8 +83,7 @@ flash at 2 {
 		};
 
 		spi1: spi at 1e630000 {
-			reg = < 0x1e630000 0xc4
-				0x30000000 0x08000000 >;
+			reg = <0x1e630000 0xc4>, <0x30000000 0x08000000>;
 			#address-cells = <1>;
 			#size-cells = <0>;
 			compatible = "aspeed,ast2500-spi";
@@ -106,8 +104,7 @@ flash at 1 {
 		};
 
 		spi2: spi at 1e631000 {
-			reg = < 0x1e631000 0xc4
-				0x38000000 0x08000000 >;
+			reg = <0x1e631000 0xc4>, <0x38000000 0x08000000>;
 			#address-cells = <1>;
 			#size-cells = <0>;
 			compatible = "aspeed,ast2500-spi";
arch/arm/boot/dts/aspeed-g6.dtsi +3 -6
index 3d5ce9da42c3..6ae5dfecbdf1 100644
@@ -95,8 +95,7 @@ gic: interrupt-controller at 40461000 {
 			};
 
 		fmc: spi at 1e620000 {
-			reg = < 0x1e620000 0xc4
-				0x20000000 0x10000000 >;
+			reg = <0x1e620000 0xc4>, <0x20000000 0x10000000>;
 			#address-cells = <1>;
 			#size-cells = <0>;
 			compatible = "aspeed,ast2600-fmc";
@@ -124,8 +123,7 @@ flash at 2 {
 		};
 
 		spi1: spi at 1e630000 {
-			reg = < 0x1e630000 0xc4
-				0x30000000 0x10000000 >;
+			reg = <0x1e630000 0xc4>, <0x30000000 0x10000000>;
 			#address-cells = <1>;
 			#size-cells = <0>;
 			compatible = "aspeed,ast2600-spi";
@@ -146,8 +144,7 @@ flash at 1 {
 		};
 
 		spi2: spi at 1e631000 {
-			reg = < 0x1e631000 0xc4
-				0x50000000 0x10000000 >;
+			reg = <0x1e631000 0xc4>, <0x50000000 0x10000000>;
 			#address-cells = <1>;
 			#size-cells = <0>;
 			compatible = "aspeed,ast2600-spi";
Signature · 1 line
2.35.1
author maintainer
#2 lore raw

PATCH v7 2/11 dt-bindings: spi: Convert the Aspeed SMC controllers device tree binding

In reply to Cédric Le Goater

The "interrupt" property is optional because it is only necessary for controllers supporting DMAs (Not implemented yet in the new driver).

---
 .../devicetree/bindings/mtd/aspeed-smc.txt    | 51 ------------
 .../bindings/spi/aspeed,ast2600-fmc.yaml      | 82 +++++++++++++++++++
 MAINTAINERS                                   |  9 ++
 3 files changed, 91 insertions(+), 51 deletions(-)
 delete mode 100644 Documentation/devicetree/bindings/mtd/aspeed-smc.txt
 create mode 100644 Documentation/devicetree/bindings/spi/aspeed,ast2600-fmc.yaml
Documentation/devicetree/bindings/mtd/aspeed-smc.txt deleted +0 -51
deleted file mode 100644
index 49f6528ef547..000000000000
@@ -1,51 +0,0 @@
-* Aspeed Firmware Memory controller
-* Aspeed SPI Flash Memory Controller
-
-The Firmware Memory Controller in the Aspeed AST2500 SoC supports
-three chip selects, two of which are always of SPI type and the third
-can be SPI or NOR type flash. These bindings only describe SPI.
-
-The two SPI flash memory controllers in the AST2500 each support two
-chip selects.
-
-Required properties:
-  - compatible : Should be one of
-	"aspeed,ast2400-fmc" for the AST2400 Firmware Memory Controller
-	"aspeed,ast2400-spi" for the AST2400 SPI Flash memory Controller
-	"aspeed,ast2500-fmc" for the AST2500 Firmware Memory Controller
-	"aspeed,ast2500-spi" for the AST2500 SPI flash memory controllers
-
-  - reg : the first contains the control register location and length,
-          the second contains the memory window mapping address and length
-  - #address-cells : must be 1 corresponding to chip select child binding
-  - #size-cells : must be 0 corresponding to chip select child binding
-
-Optional properties:
-  - interrupts : Should contain the interrupt for the dma device if an
-    FMC
-
-The child nodes are the SPI flash modules which must have a compatible
-property as specified in bindings/mtd/jedec,spi-nor.txt
-
-Optionally, the child node can contain properties for SPI mode (may be
-ignored):
-  - spi-max-frequency - max frequency of spi bus
-
-
-Example:
-fmc: fmc at 1e620000 {
-	compatible = "aspeed,ast2500-fmc";
-	reg = < 0x1e620000 0x94
-		0x20000000 0x02000000 >;
-	#address-cells = <1>;
-	#size-cells = <0>;
-	interrupts = <19>;
-	flash at 0 {
-		reg = < 0 >;
-		compatible = "jedec,spi-nor";
-		/* spi-max-frequency = <>; */
-		/* m25p,fast-read; */
-		#address-cells = <1>;
-		#size-cells = <1>;
-	};
-};
Documentation/devicetree/bindings/spi/aspeed,ast2600-fmc.yaml new file +82 -0

82 lines not on this page · raw message

MAINTAINERS +9 -0
index 9d47c5e7c6ae..149e84d32293 100644
@@ -3101,6 +3101,15 @@ S:	Maintained
 F:	Documentation/devicetree/bindings/mmc/aspeed,sdhci.yaml
 F:	drivers/mmc/host/sdhci-of-aspeed*
 
+ASPEED SMC SPI DRIVER
+M:	Chin-Ting Kuo <[email protected]>
+M:	C?dric Le Goater <[email protected]>
+L:	linux-aspeed at lists.ozlabs.org (moderated for non-subscribers)
+L:	openbmc at lists.ozlabs.org (moderated for non-subscribers)
+L:	linux-spi at vger.kernel.org
+S:	Maintained
+F:	Documentation/devicetree/bindings/spi/aspeed,ast2600-fmc.yaml
+
 ASPEED VIDEO ENGINE DRIVER
 M:	Eddie James <[email protected]>
 L:	linux-media at vger.kernel.org
Signature · 1 line
2.35.1
author maintainer
#3 lore raw

PATCH v7 3/11 spi: spi-mem: Convert Aspeed SMC driver to spi-mem

In reply to Cédric Le Goater

This SPI driver adds support for the Aspeed static memory controllers of the AST2600, AST2500 and AST2400 SoCs using the spi-mem interface.

* AST2600 Firmware SPI Memory Controller (FMC)
. BMC firmware
. 3 chip select pins (CE0 ~ CE2)
. Only supports SPI type flash memory
. different segment register interface
. single, dual and quad mode.

* AST2600 SPI Flash Controller (SPI1 and SPI2)
. host firmware
. 2 chip select pins (CE0 ~ CE1)
. different segment register interface
. single, dual and quad mode.

* AST2500 Firmware SPI Memory Controller (FMC)
. BMC firmware
. 3 chip select pins (CE0 ~ CE2)
. supports SPI type flash memory (CE0-CE1)
. CE2 can be of NOR type flash but this is not supported by the driver . single, dual mode.

* AST2500 SPI Flash Controller (SPI1 and SPI2)
. host firmware
. 2 chip select pins (CE0 ~ CE1)
. single, dual mode.

* AST2400 New Static Memory Controller (also referred as FMC) . BMC firmware
. New register set
. 5 chip select pins (CE0 ? CE4)
. supports NOR flash, NAND flash and SPI flash memory. . single, dual and quad mode.

Each controller has a memory range on which flash devices contents are mapped. Each device is assigned a window that can be changed at bootime with the Segment Address Registers.

Each SPI flash device can then be accessed in two modes: Command and User. When in User mode, SPI transfers are initiated with accesses to the memory segment of a device. When in Command mode, memory operations on the memory segment of a device generate SPI commands automatically using a Control Register for the settings.

This initial patch adds support for User mode. Command mode needs a little more work to check that the memory window on the AHB bus fits the device size. It will come later when support for direct mapping is added.

Single and dual mode RX transfers are supported. Other types than SPI are not supported.

---
 drivers/mtd/spi-nor/controllers/aspeed-smc.c | 921 -------------------
 drivers/spi/spi-aspeed-smc.c                 | 717 +++++++++++++++
 MAINTAINERS                                  |   1 +
 drivers/mtd/spi-nor/controllers/Kconfig      |  10 -
 drivers/mtd/spi-nor/controllers/Makefile     |   1 -
 drivers/spi/Kconfig                          |  11 +
 drivers/spi/Makefile                         |   1 +
 7 files changed, 730 insertions(+), 932 deletions(-)
 delete mode 100644 drivers/mtd/spi-nor/controllers/aspeed-smc.c
 create mode 100644 drivers/spi/spi-aspeed-smc.c
drivers/mtd/spi-nor/controllers/aspeed-smc.c deleted +0 -921
deleted file mode 100644
index acfe010f9dd7..000000000000
@@ -1,921 +0,0 @@
-// SPDX-License-Identifier: GPL-2.0-or-later
-/*
- * ASPEED Static Memory Controller driver
- *
- * Copyright (c) 2015-2016, IBM Corporation.
- */
-
-#include <linux/bug.h>
-#include <linux/device.h>
-#include <linux/io.h>
-#include <linux/module.h>
-#include <linux/mutex.h>
-#include <linux/mtd/mtd.h>
-#include <linux/mtd/partitions.h>
-#include <linux/mtd/spi-nor.h>
-#include <linux/of.h>
-#include <linux/of_platform.h>
-#include <linux/sizes.h>
-#include <linux/sysfs.h>
-
-#define DEVICE_NAME	"aspeed-smc"
-
-/*
- * The driver only support SPI flash
- */
-enum aspeed_smc_flash_type {
-	smc_type_nor  = 0,
-	smc_type_nand = 1,
-	smc_type_spi  = 2,
-};
-
-struct aspeed_smc_chip;
-
-struct aspeed_smc_info {
-	u32 maxsize;		/* maximum size of chip window */
-	u8 nce;			/* number of chip enables */
-	bool hastype;		/* flash type field exists in config reg */
-	u8 we0;			/* shift for write enable bit for CE0 */
-	u8 ctl0;		/* offset in regs of ctl for CE0 */
-
-	void (*set_4b)(struct aspeed_smc_chip *chip);
-};
-
-static void aspeed_smc_chip_set_4b_spi_2400(struct aspeed_smc_chip *chip);
-static void aspeed_smc_chip_set_4b(struct aspeed_smc_chip *chip);
-
-static const struct aspeed_smc_info fmc_2400_info = {
-	.maxsize = 64 * 1024 * 1024,
-	.nce = 5,
-	.hastype = true,
-	.we0 = 16,
-	.ctl0 = 0x10,
-	.set_4b = aspeed_smc_chip_set_4b,
-};
-
-static const struct aspeed_smc_info spi_2400_info = {
-	.maxsize = 64 * 1024 * 1024,
-	.nce = 1,
-	.hastype = false,
-	.we0 = 0,
-	.ctl0 = 0x04,
-	.set_4b = aspeed_smc_chip_set_4b_spi_2400,
-};
-
-static const struct aspeed_smc_info fmc_2500_info = {
-	.maxsize = 256 * 1024 * 1024,
-	.nce = 3,
-	.hastype = true,
-	.we0 = 16,
-	.ctl0 = 0x10,
-	.set_4b = aspeed_smc_chip_set_4b,
-};
-
-static const struct aspeed_smc_info spi_2500_info = {
-	.maxsize = 128 * 1024 * 1024,
-	.nce = 2,
-	.hastype = false,
-	.we0 = 16,
-	.ctl0 = 0x10,
-	.set_4b = aspeed_smc_chip_set_4b,
-};
-
-enum aspeed_smc_ctl_reg_value {
-	smc_base,		/* base value without mode for other commands */
-	smc_read,		/* command reg for (maybe fast) reads */
-	smc_write,		/* command reg for writes */
-	smc_max,
-};
-
-struct aspeed_smc_controller;
-
-struct aspeed_smc_chip {
-	int cs;
-	struct aspeed_smc_controller *controller;
-	void __iomem *ctl;			/* control register */
-	void __iomem *ahb_base;			/* base of chip window */
-	u32 ahb_window_size;			/* chip mapping window size */
-	u32 ctl_val[smc_max];			/* control settings */
-	enum aspeed_smc_flash_type type;	/* what type of flash */
-	struct spi_nor nor;
-};
-
-struct aspeed_smc_controller {
-	struct device *dev;
-
-	struct mutex mutex;			/* controller access mutex */
-	const struct aspeed_smc_info *info;	/* type info of controller */
-	void __iomem *regs;			/* controller registers */
-	void __iomem *ahb_base;			/* per-chip windows resource */
-	u32 ahb_window_size;			/* full mapping window size */
-
-	struct aspeed_smc_chip *chips[];	/* pointers to attached chips */
-};
-
-/*
- * SPI Flash Configuration Register (AST2500 SPI)
- *     or
- * Type setting Register (AST2500 FMC).
- * CE0 and CE1 can only be of type SPI. CE2 can be of type NOR but the
- * driver does not support it.
- */
-#define CONFIG_REG			0x0
-#define CONFIG_DISABLE_LEGACY		BIT(31) /* 1 */
-
-#define CONFIG_CE2_WRITE		BIT(18)
-#define CONFIG_CE1_WRITE		BIT(17)
-#define CONFIG_CE0_WRITE		BIT(16)
-
-#define CONFIG_CE2_TYPE			BIT(4) /* AST2500 FMC only */
-#define CONFIG_CE1_TYPE			BIT(2) /* AST2500 FMC only */
-#define CONFIG_CE0_TYPE			BIT(0) /* AST2500 FMC only */
-
-/*
- * CE Control Register
- */
-#define CE_CONTROL_REG			0x4
-
-/*
- * CEx Control Register
- */
-#define CONTROL_AAF_MODE		BIT(31)
-#define CONTROL_IO_MODE_MASK		GENMASK(30, 28)
-#define CONTROL_IO_DUAL_DATA		BIT(29)
-#define CONTROL_IO_DUAL_ADDR_DATA	(BIT(29) | BIT(28))
-#define CONTROL_IO_QUAD_DATA		BIT(30)
-#define CONTROL_IO_QUAD_ADDR_DATA	(BIT(30) | BIT(28))
-#define CONTROL_CE_INACTIVE_SHIFT	24
-#define CONTROL_CE_INACTIVE_MASK	GENMASK(27, \
-					CONTROL_CE_INACTIVE_SHIFT)
-/* 0 = 16T ... 15 = 1T   T=HCLK */
-#define CONTROL_COMMAND_SHIFT		16
-#define CONTROL_DUMMY_COMMAND_OUT	BIT(15)
-#define CONTROL_IO_DUMMY_HI		BIT(14)
-#define CONTROL_IO_DUMMY_HI_SHIFT	14
-#define CONTROL_CLK_DIV4		BIT(13) /* others */
-#define CONTROL_IO_ADDRESS_4B		BIT(13) /* AST2400 SPI */
-#define CONTROL_RW_MERGE		BIT(12)
-#define CONTROL_IO_DUMMY_LO_SHIFT	6
-#define CONTROL_IO_DUMMY_LO		GENMASK(7, \
-						CONTROL_IO_DUMMY_LO_SHIFT)
-#define CONTROL_IO_DUMMY_MASK		(CONTROL_IO_DUMMY_HI | \
-					 CONTROL_IO_DUMMY_LO)
-#define CONTROL_IO_DUMMY_SET(dummy)				 \
-	(((((dummy) >> 2) & 0x1) << CONTROL_IO_DUMMY_HI_SHIFT) | \
-	 (((dummy) & 0x3) << CONTROL_IO_DUMMY_LO_SHIFT))
-
-#define CONTROL_CLOCK_FREQ_SEL_SHIFT	8
-#define CONTROL_CLOCK_FREQ_SEL_MASK	GENMASK(11, \
-						CONTROL_CLOCK_FREQ_SEL_SHIFT)
-#define CONTROL_LSB_FIRST		BIT(5)
-#define CONTROL_CLOCK_MODE_3		BIT(4)
-#define CONTROL_IN_DUAL_DATA		BIT(3)
-#define CONTROL_CE_STOP_ACTIVE_CONTROL	BIT(2)
-#define CONTROL_COMMAND_MODE_MASK	GENMASK(1, 0)
-#define CONTROL_COMMAND_MODE_NORMAL	0
-#define CONTROL_COMMAND_MODE_FREAD	1
-#define CONTROL_COMMAND_MODE_WRITE	2
-#define CONTROL_COMMAND_MODE_USER	3
-
-#define CONTROL_KEEP_MASK						\
-	(CONTROL_AAF_MODE | CONTROL_CE_INACTIVE_MASK | CONTROL_CLK_DIV4 | \
-	 CONTROL_CLOCK_FREQ_SEL_MASK | CONTROL_LSB_FIRST | CONTROL_CLOCK_MODE_3)
-
-/*
- * The Segment Register uses a 8MB unit to encode the start address
- * and the end address of the mapping window of a flash SPI slave :
- *
- *        | byte 1 | byte 2 | byte 3 | byte 4 |
- *        +--------+--------+--------+--------+
- *        |  end   |  start |   0    |   0    |
- */
-#define SEGMENT_ADDR_REG0		0x30
-#define SEGMENT_ADDR_START(_r)		((((_r) >> 16) & 0xFF) << 23)
-#define SEGMENT_ADDR_END(_r)		((((_r) >> 24) & 0xFF) << 23)
-#define SEGMENT_ADDR_VALUE(start, end)					\
-	(((((start) >> 23) & 0xFF) << 16) | ((((end) >> 23) & 0xFF) << 24))
-#define SEGMENT_ADDR_REG(controller, cs)	\
-	((controller)->regs + SEGMENT_ADDR_REG0 + (cs) * 4)
-
-/*
- * In user mode all data bytes read or written to the chip decode address
- * range are transferred to or from the SPI bus. The range is treated as a
- * fifo of arbitratry 1, 2, or 4 byte width but each write has to be aligned
- * to its size. The address within the multiple 8kB range is ignored when
- * sending bytes to the SPI bus.
- *
- * On the arm architecture, as of Linux version 4.3, memcpy_fromio and
- * memcpy_toio on little endian targets use the optimized memcpy routines
- * that were designed for well behavied memory storage. These routines
- * have a stutter if the source and destination are not both word aligned,
- * once with a duplicate access to the source after aligning to the
- * destination to a word boundary, and again with a duplicate access to
- * the source when the final byte count is not word aligned.
- *
- * When writing or reading the fifo this stutter discards data or sends
- * too much data to the fifo and can not be used by this driver.
- *
- * While the low level io string routines that implement the insl family do
- * the desired accesses and memory increments, the cross architecture io
- * macros make them essentially impossible to use on a memory mapped address
- * instead of a a token from the call to iomap of an io port.
- *
- * These fifo routines use readl and friends to a constant io port and update
- * the memory buffer pointer and count via explicit code. The final updates
- * to len are optimistically suppressed.
- */
-static int aspeed_smc_read_from_ahb(void *buf, void __iomem *src, size_t len)
-{
-	size_t offset = 0;
-
-	if (IS_ALIGNED((uintptr_t)src, sizeof(uintptr_t)) &&
-	    IS_ALIGNED((uintptr_t)buf, sizeof(uintptr_t))) {
-		ioread32_rep(src, buf, len >> 2);
-		offset = len & ~0x3;
-		len -= offset;
-	}
-	ioread8_rep(src, (u8 *)buf + offset, len);
-	return 0;
-}
-
-static int aspeed_smc_write_to_ahb(void __iomem *dst, const void *buf,
-				   size_t len)
-{
-	size_t offset = 0;
-
-	if (IS_ALIGNED((uintptr_t)dst, sizeof(uintptr_t)) &&
-	    IS_ALIGNED((uintptr_t)buf, sizeof(uintptr_t))) {
-		iowrite32_rep(dst, buf, len >> 2);
-		offset = len & ~0x3;
-		len -= offset;
-	}
-	iowrite8_rep(dst, (const u8 *)buf + offset, len);
-	return 0;
-}
-
-static inline u32 aspeed_smc_chip_write_bit(struct aspeed_smc_chip *chip)
-{
-	return BIT(chip->controller->info->we0 + chip->cs);
-}
-
-static void aspeed_smc_chip_check_config(struct aspeed_smc_chip *chip)
-{
-	struct aspeed_smc_controller *controller = chip->controller;
-	u32 reg;
-
-	reg = readl(controller->regs + CONFIG_REG);
-
-	if (reg & aspeed_smc_chip_write_bit(chip))
-		return;
-
-	dev_dbg(controller->dev, "config write is not set ! @%p: 0x%08x\n",
-		controller->regs + CONFIG_REG, reg);
-	reg |= aspeed_smc_chip_write_bit(chip);
-	writel(reg, controller->regs + CONFIG_REG);
-}
-
-static void aspeed_smc_start_user(struct spi_nor *nor)
-{
-	struct aspeed_smc_chip *chip = nor->priv;
-	u32 ctl = chip->ctl_val[smc_base];
-
-	/*
-	 * When the chip is controlled in user mode, we need write
-	 * access to send the opcodes to it. So check the config.
-	 */
-	aspeed_smc_chip_check_config(chip);
-
-	ctl |= CONTROL_COMMAND_MODE_USER |
-		CONTROL_CE_STOP_ACTIVE_CONTROL;
-	writel(ctl, chip->ctl);
-
-	ctl &= ~CONTROL_CE_STOP_ACTIVE_CONTROL;
-	writel(ctl, chip->ctl);
-}
-
-static void aspeed_smc_stop_user(struct spi_nor *nor)
-{
-	struct aspeed_smc_chip *chip = nor->priv;
-
-	u32 ctl = chip->ctl_val[smc_read];
-	u32 ctl2 = ctl | CONTROL_COMMAND_MODE_USER |
-		CONTROL_CE_STOP_ACTIVE_CONTROL;
-
-	writel(ctl2, chip->ctl);	/* stop user CE control */
-	writel(ctl, chip->ctl);		/* default to fread or read mode */
-}
-
-static int aspeed_smc_prep(struct spi_nor *nor)
-{
-	struct aspeed_smc_chip *chip = nor->priv;
-
-	mutex_lock(&chip->controller->mutex);
-	return 0;
-}
-
-static void aspeed_smc_unprep(struct spi_nor *nor)
-{
-	struct aspeed_smc_chip *chip = nor->priv;
-
-	mutex_unlock(&chip->controller->mutex);
-}
-
-static int aspeed_smc_read_reg(struct spi_nor *nor, u8 opcode, u8 *buf,
-			       size_t len)
-{
-	struct aspeed_smc_chip *chip = nor->priv;
-
-	aspeed_smc_start_user(nor);
-	aspeed_smc_write_to_ahb(chip->ahb_base, &opcode, 1);
-	aspeed_smc_read_from_ahb(buf, chip->ahb_base, len);
-	aspeed_smc_stop_user(nor);
-	return 0;
-}
-
-static int aspeed_smc_write_reg(struct spi_nor *nor, u8 opcode, const u8 *buf,
-				size_t len)
-{
-	struct aspeed_smc_chip *chip = nor->priv;
-
-	aspeed_smc_start_user(nor);
-	aspeed_smc_write_to_ahb(chip->ahb_base, &opcode, 1);
-	aspeed_smc_write_to_ahb(chip->ahb_base, buf, len);
-	aspeed_smc_stop_user(nor);
-	return 0;
-}
-
-static void aspeed_smc_send_cmd_addr(struct spi_nor *nor, u8 cmd, u32 addr)
-{
-	struct aspeed_smc_chip *chip = nor->priv;
-	__be32 temp;
-	u32 cmdaddr;
-
-	switch (nor->addr_width) {
-	default:
-		WARN_ONCE(1, "Unexpected address width %u, defaulting to 3\n",
-			  nor->addr_width);
-		fallthrough;
-	case 3:
-		cmdaddr = addr & 0xFFFFFF;
-		cmdaddr |= cmd << 24;
-
-		temp = cpu_to_be32(cmdaddr);
-		aspeed_smc_write_to_ahb(chip->ahb_base, &temp, 4);
-		break;
-	case 4:
-		temp = cpu_to_be32(addr);
-		aspeed_smc_write_to_ahb(chip->ahb_base, &cmd, 1);
-		aspeed_smc_write_to_ahb(chip->ahb_base, &temp, 4);
-		break;
-	}
-}
-
-static ssize_t aspeed_smc_read_user(struct spi_nor *nor, loff_t from,
-				    size_t len, u_char *read_buf)
-{
-	struct aspeed_smc_chip *chip = nor->priv;
-	int i;
-	u8 dummy = 0xFF;
-
-	aspeed_smc_start_user(nor);
-	aspeed_smc_send_cmd_addr(nor, nor->read_opcode, from);
-	for (i = 0; i < chip->nor.read_dummy / 8; i++)
-		aspeed_smc_write_to_ahb(chip->ahb_base, &dummy, sizeof(dummy));
-
-	aspeed_smc_read_from_ahb(read_buf, chip->ahb_base, len);
-	aspeed_smc_stop_user(nor);
-	return len;
-}
-
-static ssize_t aspeed_smc_write_user(struct spi_nor *nor, loff_t to,
-				     size_t len, const u_char *write_buf)
-{
-	struct aspeed_smc_chip *chip = nor->priv;
-
-	aspeed_smc_start_user(nor);
-	aspeed_smc_send_cmd_addr(nor, nor->program_opcode, to);
-	aspeed_smc_write_to_ahb(chip->ahb_base, write_buf, len);
-	aspeed_smc_stop_user(nor);
-	return len;
-}
-
-static int aspeed_smc_unregister(struct aspeed_smc_controller *controller)
-{
-	struct aspeed_smc_chip *chip;
-	int n;
-
-	for (n = 0; n < controller->info->nce; n++) {
-		chip = controller->chips[n];
-		if (chip)
-			mtd_device_unregister(&chip->nor.mtd);
-	}
-
-	return 0;
-}
-
-static int aspeed_smc_remove(struct platform_device *dev)
-{
-	return aspeed_smc_unregister(platform_get_drvdata(dev));
-}
-
-static const struct of_device_id aspeed_smc_matches[] = {
-	{ .compatible = "aspeed,ast2400-fmc", .data = &fmc_2400_info },
-	{ .compatible = "aspeed,ast2400-spi", .data = &spi_2400_info },
-	{ .compatible = "aspeed,ast2500-fmc", .data = &fmc_2500_info },
-	{ .compatible = "aspeed,ast2500-spi", .data = &spi_2500_info },
-	{ }
-};
-MODULE_DEVICE_TABLE(of, aspeed_smc_matches);
-
-/*
- * Each chip has a mapping window defined by a segment address
- * register defining a start and an end address on the AHB bus. These
- * addresses can be configured to fit the chip size and offer a
- * contiguous memory region across chips. For the moment, we only
- * check that each chip segment is valid.
- */
-static void __iomem *aspeed_smc_chip_base(struct aspeed_smc_chip *chip,
-					  struct resource *res)
-{
-	struct aspeed_smc_controller *controller = chip->controller;
-	u32 offset = 0;
-	u32 reg;
-
-	if (controller->info->nce > 1) {
-		reg = readl(SEGMENT_ADDR_REG(controller, chip->cs));
-
-		if (SEGMENT_ADDR_START(reg) >= SEGMENT_ADDR_END(reg))
-			return NULL;
-
-		offset = SEGMENT_ADDR_START(reg) - res->start;
-	}
-
-	return controller->ahb_base + offset;
-}
-
-static u32 aspeed_smc_ahb_base_phy(struct aspeed_smc_controller *controller)
-{
-	u32 seg0_val = readl(SEGMENT_ADDR_REG(controller, 0));
-
-	return SEGMENT_ADDR_START(seg0_val);
-}
-
-static u32 chip_set_segment(struct aspeed_smc_chip *chip, u32 cs, u32 start,
-			    u32 size)
-{
-	struct aspeed_smc_controller *controller = chip->controller;
-	void __iomem *seg_reg;
-	u32 seg_oldval, seg_newval, ahb_base_phy, end;
-
-	ahb_base_phy = aspeed_smc_ahb_base_phy(controller);
-
-	seg_reg = SEGMENT_ADDR_REG(controller, cs);
-	seg_oldval = readl(seg_reg);
-
-	/*
-	 * If the chip size is not specified, use the default segment
-	 * size, but take into account the possible overlap with the
-	 * previous segment
-	 */
-	if (!size)
-		size = SEGMENT_ADDR_END(seg_oldval) - start;
-
-	/*
-	 * The segment cannot exceed the maximum window size of the
-	 * controller.
-	 */
-	if (start + size > ahb_base_phy + controller->ahb_window_size) {
-		size = ahb_base_phy + controller->ahb_window_size - start;
-		dev_warn(chip->nor.dev, "CE%d window resized to %dMB",
-			 cs, size >> 20);
-	}
-
-	end = start + size;
-	seg_newval = SEGMENT_ADDR_VALUE(start, end);
-	writel(seg_newval, seg_reg);
-
-	/*
-	 * Restore default value if something goes wrong. The chip
-	 * might have set some bogus value and we would loose access
-	 * to the chip.
-	 */
-	if (seg_newval != readl(seg_reg)) {
-		dev_err(chip->nor.dev, "CE%d window invalid", cs);
-		writel(seg_oldval, seg_reg);
-		start = SEGMENT_ADDR_START(seg_oldval);
-		end = SEGMENT_ADDR_END(seg_oldval);
-		size = end - start;
-	}
-
-	dev_info(chip->nor.dev, "CE%d window [ 0x%.8x - 0x%.8x ] %dMB",
-		 cs, start, end, size >> 20);
-
-	return size;
-}
-
-/*
- * The segment register defines the mapping window on the AHB bus and
- * it needs to be configured depending on the chip size. The segment
- * register of the following CE also needs to be tuned in order to
- * provide a contiguous window across multiple chips.
- *
- * This is expected to be called in increasing CE order
- */
-static u32 aspeed_smc_chip_set_segment(struct aspeed_smc_chip *chip)
-{
-	struct aspeed_smc_controller *controller = chip->controller;
-	u32 ahb_base_phy, start;
-	u32 size = chip->nor.mtd.size;
-
-	/*
-	 * Each controller has a chip size limit for direct memory
-	 * access
-	 */
-	if (size > controller->info->maxsize)
-		size = controller->info->maxsize;
-
-	/*
-	 * The AST2400 SPI controller only handles one chip and does
-	 * not have segment registers. Let's use the chip size for the
-	 * AHB window.
-	 */
-	if (controller->info == &spi_2400_info)
-		goto out;
-
-	/*
-	 * The AST2500 SPI controller has a HW bug when the CE0 chip
-	 * size reaches 128MB. Enforce a size limit of 120MB to
-	 * prevent the controller from using bogus settings in the
-	 * segment register.
-	 */
-	if (chip->cs == 0 && controller->info == &spi_2500_info &&
-	    size == SZ_128M) {
-		size = 120 << 20;
-		dev_info(chip->nor.dev,
-			 "CE%d window resized to %dMB (AST2500 HW quirk)",
-			 chip->cs, size >> 20);
-	}
-
-	ahb_base_phy = aspeed_smc_ahb_base_phy(controller);
-
-	/*
-	 * As a start address for the current segment, use the default
-	 * start address if we are handling CE0 or use the previous
-	 * segment ending address
-	 */
-	if (chip->cs) {
-		u32 prev = readl(SEGMENT_ADDR_REG(controller, chip->cs - 1));
-
-		start = SEGMENT_ADDR_END(prev);
-	} else {
-		start = ahb_base_phy;
-	}
-
-	size = chip_set_segment(chip, chip->cs, start, size);
-
-	/* Update chip base address on the AHB bus */
-	chip->ahb_base = controller->ahb_base + (start - ahb_base_phy);
-
-	/*
-	 * Now, make sure the next segment does not overlap with the
-	 * current one we just configured, even if there is no
-	 * available chip. That could break access in Command Mode.
-	 */
-	if (chip->cs < controller->info->nce - 1)
-		chip_set_segment(chip, chip->cs + 1, start + size, 0);
-
-out:
-	if (size < chip->nor.mtd.size)
-		dev_warn(chip->nor.dev,
-			 "CE%d window too small for chip %dMB",
-			 chip->cs, (u32)chip->nor.mtd.size >> 20);
-
-	return size;
-}
-
-static void aspeed_smc_chip_enable_write(struct aspeed_smc_chip *chip)
-{
-	struct aspeed_smc_controller *controller = chip->controller;
-	u32 reg;
-
-	reg = readl(controller->regs + CONFIG_REG);
-
-	reg |= aspeed_smc_chip_write_bit(chip);
-	writel(reg, controller->regs + CONFIG_REG);
-}
-
-static void aspeed_smc_chip_set_type(struct aspeed_smc_chip *chip, int type)
-{
-	struct aspeed_smc_controller *controller = chip->controller;
-	u32 reg;
-
-	chip->type = type;
-
-	reg = readl(controller->regs + CONFIG_REG);
-	reg &= ~(3 << (chip->cs * 2));
-	reg |= chip->type << (chip->cs * 2);
-	writel(reg, controller->regs + CONFIG_REG);
-}
-
-/*
- * The first chip of the AST2500 FMC flash controller is strapped by
- * hardware, or autodetected, but other chips need to be set. Enforce
- * the 4B setting for all chips.
- */
-static void aspeed_smc_chip_set_4b(struct aspeed_smc_chip *chip)
-{
-	struct aspeed_smc_controller *controller = chip->controller;
-	u32 reg;
-
-	reg = readl(controller->regs + CE_CONTROL_REG);
-	reg |= 1 << chip->cs;
-	writel(reg, controller->regs + CE_CONTROL_REG);
-}
-
-/*
- * The AST2400 SPI flash controller does not have a CE Control
- * register. It uses the CE0 control register to set 4Byte mode at the
- * controller level.
- */
-static void aspeed_smc_chip_set_4b_spi_2400(struct aspeed_smc_chip *chip)
-{
-	chip->ctl_val[smc_base] |= CONTROL_IO_ADDRESS_4B;
-	chip->ctl_val[smc_read] |= CONTROL_IO_ADDRESS_4B;
-}
-
-static int aspeed_smc_chip_setup_init(struct aspeed_smc_chip *chip,
-				      struct resource *res)
-{
-	struct aspeed_smc_controller *controller = chip->controller;
-	const struct aspeed_smc_info *info = controller->info;
-	u32 reg, base_reg;
-
-	/*
-	 * Always turn on the write enable bit to allow opcodes to be
-	 * sent in user mode.
-	 */
-	aspeed_smc_chip_enable_write(chip);
-
-	/* The driver only supports SPI type flash */
-	if (info->hastype)
-		aspeed_smc_chip_set_type(chip, smc_type_spi);
-
-	/*
-	 * Configure chip base address in memory
-	 */
-	chip->ahb_base = aspeed_smc_chip_base(chip, res);
-	if (!chip->ahb_base) {
-		dev_warn(chip->nor.dev, "CE%d window closed", chip->cs);
-		return -EINVAL;
-	}
-
-	/*
-	 * Get value of the inherited control register. U-Boot usually
-	 * does some timing calibration on the FMC chip, so it's good
-	 * to keep them. In the future, we should handle calibration
-	 * from Linux.
-	 */
-	reg = readl(chip->ctl);
-	dev_dbg(controller->dev, "control register: %08x\n", reg);
-
-	base_reg = reg & CONTROL_KEEP_MASK;
-	if (base_reg != reg) {
-		dev_dbg(controller->dev,
-			"control register changed to: %08x\n",
-			base_reg);
-	}
-	chip->ctl_val[smc_base] = base_reg;
-
-	/*
-	 * Retain the prior value of the control register as the
-	 * default if it was normal access mode. Otherwise start with
-	 * the sanitized base value set to read mode.
-	 */
-	if ((reg & CONTROL_COMMAND_MODE_MASK) ==
-	    CONTROL_COMMAND_MODE_NORMAL)
-		chip->ctl_val[smc_read] = reg;
-	else
-		chip->ctl_val[smc_read] = chip->ctl_val[smc_base] |
-			CONTROL_COMMAND_MODE_NORMAL;
-
-	dev_dbg(controller->dev, "default control register: %08x\n",
-		chip->ctl_val[smc_read]);
-	return 0;
-}
-
-static int aspeed_smc_chip_setup_finish(struct aspeed_smc_chip *chip)
-{
-	struct aspeed_smc_controller *controller = chip->controller;
-	const struct aspeed_smc_info *info = controller->info;
-	u32 cmd;
-
-	if (chip->nor.addr_width == 4 && info->set_4b)
-		info->set_4b(chip);
-
-	/* This is for direct AHB access when using Command Mode. */
-	chip->ahb_window_size = aspeed_smc_chip_set_segment(chip);
-
-	/*
-	 * base mode has not been optimized yet. use it for writes.
-	 */
-	chip->ctl_val[smc_write] = chip->ctl_val[smc_base] |
-		chip->nor.program_opcode << CONTROL_COMMAND_SHIFT |
-		CONTROL_COMMAND_MODE_WRITE;
-
-	dev_dbg(controller->dev, "write control register: %08x\n",
-		chip->ctl_val[smc_write]);
-
-	/*
-	 * TODO: Adjust clocks if fast read is supported and interpret
-	 * SPI NOR flags to adjust controller settings.
-	 */
-	if (chip->nor.read_proto == SNOR_PROTO_1_1_1) {
-		if (chip->nor.read_dummy == 0)
-			cmd = CONTROL_COMMAND_MODE_NORMAL;
-		else
-			cmd = CONTROL_COMMAND_MODE_FREAD;
-	} else {
-		dev_err(chip->nor.dev, "unsupported SPI read mode\n");
-		return -EINVAL;
-	}
-
-	chip->ctl_val[smc_read] |= cmd |
-		CONTROL_IO_DUMMY_SET(chip->nor.read_dummy / 8);
-
-	dev_dbg(controller->dev, "base control register: %08x\n",
-		chip->ctl_val[smc_read]);
-	return 0;
-}
-
-static const struct spi_nor_controller_ops aspeed_smc_controller_ops = {
-	.prepare = aspeed_smc_prep,
-	.unprepare = aspeed_smc_unprep,
-	.read_reg = aspeed_smc_read_reg,
-	.write_reg = aspeed_smc_write_reg,
-	.read = aspeed_smc_read_user,
-	.write = aspeed_smc_write_user,
-};
-
-static int aspeed_smc_setup_flash(struct aspeed_smc_controller *controller,
-				  struct device_node *np, struct resource *r)
-{
-	const struct spi_nor_hwcaps hwcaps = {
-		.mask = SNOR_HWCAPS_READ |
-			SNOR_HWCAPS_READ_FAST |
-			SNOR_HWCAPS_PP,
-	};
-	const struct aspeed_smc_info *info = controller->info;
-	struct device *dev = controller->dev;
-	struct device_node *child;
-	unsigned int cs;
-	int ret = -ENODEV;
-	bool found_one = false;
-
-	for_each_available_child_of_node(np, child) {
-		struct aspeed_smc_chip *chip;
-		struct spi_nor *nor;
-		struct mtd_info *mtd;
-
-		/* This driver does not support NAND or NOR flash devices. */
-		if (!of_device_is_compatible(child, "jedec,spi-nor"))
-			continue;
-
-		ret = of_property_read_u32(child, "reg", &cs);
-		if (ret) {
-			dev_err(dev, "Couldn't not read chip select.\n");
-			break;
-		}
-
-		if (cs >= info->nce) {
-			dev_err(dev, "Chip select %d out of range.\n",
-				cs);
-			ret = -ERANGE;
-			break;
-		}
-
-		if (controller->chips[cs]) {
-			dev_err(dev, "Chip select %d already in use by %s\n",
-				cs, dev_name(controller->chips[cs]->nor.dev));
-			ret = -EBUSY;
-			break;
-		}
-
-		chip = devm_kzalloc(controller->dev, sizeof(*chip), GFP_KERNEL);
-		if (!chip) {
-			ret = -ENOMEM;
-			break;
-		}
-
-		chip->controller = controller;
-		chip->ctl = controller->regs + info->ctl0 + cs * 4;
-		chip->cs = cs;
-
-		nor = &chip->nor;
-		mtd = &nor->mtd;
-
-		nor->dev = dev;
-		nor->priv = chip;
-		spi_nor_set_flash_node(nor, child);
-		nor->controller_ops = &aspeed_smc_controller_ops;
-
-		ret = aspeed_smc_chip_setup_init(chip, r);
-		if (ret)
-			break;
-
-		/*
-		 * TODO: Add support for Dual and Quad SPI protocols
-		 * attach when board support is present as determined
-		 * by of property.
-		 */
-		ret = spi_nor_scan(nor, NULL, &hwcaps);
-		/*
-		 * If we fail to scan the device it might not be present or
-		 * broken.  Don't fail the whole controller if others work.
-		 */
-		if (ret) {
-			if (found_one)
-				ret = 0;
-
-			devm_kfree(controller->dev, chip);
-			continue;
-		}
-
-		ret = aspeed_smc_chip_setup_finish(chip);
-		if (ret)
-			break;
-
-		ret = mtd_device_register(mtd, NULL, 0);
-		if (ret)
-			break;
-
-		controller->chips[cs] = chip;
-		found_one = true;
-	}
-
-	if (ret) {
-		of_node_put(child);
-		aspeed_smc_unregister(controller);
-	}
-
-	return ret;
-}
-
-static int aspeed_smc_probe(struct platform_device *pdev)
-{
-	struct device_node *np = pdev->dev.of_node;
-	struct device *dev = &pdev->dev;
-	struct aspeed_smc_controller *controller;
-	const struct of_device_id *match;
-	const struct aspeed_smc_info *info;
-	struct resource *res;
-	int ret;
-
-	match = of_match_device(aspeed_smc_matches, &pdev->dev);
-	if (!match || !match->data)
-		return -ENODEV;
-	info = match->data;
-
-	controller = devm_kzalloc(&pdev->dev,
-				  struct_size(controller, chips, info->nce),
-				  GFP_KERNEL);
-	if (!controller)
-		return -ENOMEM;
-	controller->info = info;
-	controller->dev = dev;
-
-	mutex_init(&controller->mutex);
-	platform_set_drvdata(pdev, controller);
-
-	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
-	controller->regs = devm_ioremap_resource(dev, res);
-	if (IS_ERR(controller->regs))
-		return PTR_ERR(controller->regs);
-
-	res = platform_get_resource(pdev, IORESOURCE_MEM, 1);
-	controller->ahb_base = devm_ioremap_resource(dev, res);
-	if (IS_ERR(controller->ahb_base))
-		return PTR_ERR(controller->ahb_base);
-
-	controller->ahb_window_size = resource_size(res);
-
-	ret = aspeed_smc_setup_flash(controller, np, res);
-	if (ret)
-		dev_err(dev, "Aspeed SMC probe failed %d\n", ret);
-
-	return ret;
-}
-
-static struct platform_driver aspeed_smc_driver = {
-	.probe = aspeed_smc_probe,
-	.remove = aspeed_smc_remove,
-	.driver = {
-		.name = DEVICE_NAME,
-		.of_match_table = aspeed_smc_matches,
-	}
-};
-
-module_platform_driver(aspeed_smc_driver);
-
-MODULE_DESCRIPTION("ASPEED Static Memory Controller Driver");
-MODULE_AUTHOR("Cedric Le Goater <[email protected]>");
-MODULE_LICENSE("GPL v2");
drivers/spi/spi-aspeed-smc.c new file +717 -0

717 lines not on this page · raw message

MAINTAINERS +1 -0
index 149e84d32293..fbad83094656 100644
@@ -3109,6 +3109,7 @@ L:	openbmc at lists.ozlabs.org (moderated for non-subscribers)
 L:	linux-spi at vger.kernel.org
 S:	Maintained
 F:	Documentation/devicetree/bindings/spi/aspeed,ast2600-fmc.yaml
+F:	drivers/spi/spi-aspeed-smc.c
 
 ASPEED VIDEO ENGINE DRIVER
 M:	Eddie James <[email protected]>
drivers/mtd/spi-nor/controllers/Kconfig +0 -10
index 50f4f3484d42..ca45dcd3ffe8 100644
@@ -1,14 +1,4 @@
 # SPDX-License-Identifier: GPL-2.0-only
-config SPI_ASPEED_SMC
-	tristate "Aspeed flash controllers in SPI mode"
-	depends on ARCH_ASPEED || COMPILE_TEST
-	depends on HAS_IOMEM && OF
-	help
-	  This enables support for the Firmware Memory controller (FMC)
-	  in the Aspeed AST2500/AST2400 SoCs when attached to SPI NOR chips,
-	  and support for the SPI flash memory controller (SPI) for
-	  the host firmware. The implementation only supports SPI NOR.
-
 config SPI_HISI_SFC
 	tristate "Hisilicon FMC SPI NOR Flash Controller(SFC)"
 	depends on ARCH_HISI || COMPILE_TEST
drivers/mtd/spi-nor/controllers/Makefile +0 -1
index 6e2a1dc68466..0b8e1d530913 100644
@@ -1,4 +1,3 @@
 # SPDX-License-Identifier: GPL-2.0
-obj-$(CONFIG_SPI_ASPEED_SMC)	+= aspeed-smc.o
 obj-$(CONFIG_SPI_HISI_SFC)	+= hisi-sfc.o
 obj-$(CONFIG_SPI_NXP_SPIFI)	+= nxp-spifi.o
drivers/spi/Kconfig +11 -0
index d2815eb361c0..4a3b02ec4ba9 100644
@@ -101,6 +101,17 @@ config SPI_ARMADA_3700
 	  This enables support for the SPI controller present on the
 	  Marvell Armada 3700 SoCs.
 
+config SPI_ASPEED_SMC
+	tristate "Aspeed flash controllers in SPI mode"
+	depends on ARCH_ASPEED || COMPILE_TEST
+	depends on OF
+	help
+	  This enables support for the Firmware Memory controller (FMC)
+	  in the Aspeed AST2600, AST2500 and AST2400 SoCs when attached
+	  to SPI NOR chips, and support for the SPI flash memory
+	  controller (SPI) for the host firmware. The implementation
+	  only supports SPI NOR.
+
 config SPI_ATMEL
 	tristate "Atmel SPI Controller"
 	depends on ARCH_AT91 || COMPILE_TEST
drivers/spi/Makefile +1 -0
index 3aa28ed3f761..1566388b96ae 100644
@@ -19,6 +19,7 @@ obj-$(CONFIG_SPI_ALTERA_CORE)		+= spi-altera-core.o
 obj-$(CONFIG_SPI_ALTERA_DFL)		+= spi-altera-dfl.o
 obj-$(CONFIG_SPI_AR934X)		+= spi-ar934x.o
 obj-$(CONFIG_SPI_ARMADA_3700)		+= spi-armada-3700.o
+obj-$(CONFIG_SPI_ASPEED_SMC)		+= spi-aspeed-smc.o
 obj-$(CONFIG_SPI_ATMEL)			+= spi-atmel.o
 obj-$(CONFIG_SPI_ATMEL_QUADSPI)		+= atmel-quadspi.o
 obj-$(CONFIG_SPI_AT91_USART)		+= spi-at91-usart.o
Signature · 1 line
2.35.1
author maintainer
#4 lore raw

PATCH v7 4/11 spi: aspeed: Add support for direct mapping

In reply to Cédric Le Goater

Use direct mapping to read the flash device contents. This operation mode is called "Command mode" on Aspeed SoC SMC controllers. It uses a Control Register for the settings to apply when a memory operation is performed on the flash device mapping window.

If the window is not big enough, fall back to the "User mode" to perform the read.

Direct mapping for writes will come later when validated.

---
 drivers/spi/spi-aspeed-smc.c | 68 ++++++++++++++++++++++++++++++++++++
 1 file changed, 68 insertions(+)
drivers/spi/spi-aspeed-smc.c +68 -0
index cb5a0652e5a4..50cc7bd7ba3e 100644
@@ -411,10 +411,78 @@ static int aspeed_spi_chip_set_default_window(struct aspeed_spi_chip *chip)
 	return chip->ahb_window_size ? 0 : -1;
 }
 
+static int aspeed_spi_dirmap_create(struct spi_mem_dirmap_desc *desc)
+{
+	struct aspeed_spi *aspi = spi_controller_get_devdata(desc->mem->spi->master);
+	struct aspeed_spi_chip *chip = &aspi->chips[desc->mem->spi->chip_select];
+	struct spi_mem_op *op = &desc->info.op_tmpl;
+	u32 ctl_val;
+	int ret = 0;
+
+	chip->clk_freq = desc->mem->spi->max_speed_hz;
+
+	/* Only for reads */
+	if (op->data.dir != SPI_MEM_DATA_IN)
+		return -EOPNOTSUPP;
+
+	if (desc->info.length > chip->ahb_window_size)
+		dev_warn(aspi->dev, "CE%d window (%dMB) too small for mapping",
+			 chip->cs, chip->ahb_window_size >> 20);
+
+	/* Define the default IO read settings */
+	ctl_val = readl(chip->ctl) & ~CTRL_IO_CMD_MASK;
+	ctl_val |= aspeed_spi_get_io_mode(op) |
+		op->cmd.opcode << CTRL_COMMAND_SHIFT |
+		CTRL_IO_DUMMY_SET(op->dummy.nbytes / op->dummy.buswidth) |
+		CTRL_IO_MODE_READ;
+
+	/* Tune 4BYTE address mode */
+	if (op->addr.nbytes) {
+		u32 addr_mode = readl(aspi->regs + CE_CTRL_REG);
+
+		if (op->addr.nbytes == 4)
+			addr_mode |= (0x11 << chip->cs);
+		else
+			addr_mode &= ~(0x11 << chip->cs);
+		writel(addr_mode, aspi->regs + CE_CTRL_REG);
+	}
+
+	/* READ mode is the controller default setting */
+	chip->ctl_val[ASPEED_SPI_READ] = ctl_val;
+	writel(chip->ctl_val[ASPEED_SPI_READ], chip->ctl);
+
+	dev_info(aspi->dev, "CE%d read buswidth:%d [0x%08x]\n",
+		 chip->cs, op->data.buswidth, chip->ctl_val[ASPEED_SPI_READ]);
+
+	return ret;
+}
+
+static ssize_t aspeed_spi_dirmap_read(struct spi_mem_dirmap_desc *desc,
+				      u64 offset, size_t len, void *buf)
+{
+	struct aspeed_spi *aspi = spi_controller_get_devdata(desc->mem->spi->master);
+	struct aspeed_spi_chip *chip = &aspi->chips[desc->mem->spi->chip_select];
+
+	/* Switch to USER command mode if mapping window is too small */
+	if (chip->ahb_window_size < offset + len) {
+		int ret;
+
+		ret = aspeed_spi_read_user(chip, &desc->info.op_tmpl, offset, len, buf);
+		if (ret < 0)
+			return ret;
+	} else {
+		memcpy_fromio(buf, chip->ahb_base + offset, len);
+	}
+
+	return len;
+}
+
 static const struct spi_controller_mem_ops aspeed_spi_mem_ops = {
 	.supports_op = aspeed_spi_supports_op,
 	.exec_op = aspeed_spi_exec_op,
 	.get_name = aspeed_spi_get_name,
+	.dirmap_create = aspeed_spi_dirmap_create,
+	.dirmap_read = aspeed_spi_dirmap_read,
 };
 
 static void aspeed_spi_chip_set_type(struct aspeed_spi *aspi, unsigned int cs, int type)
Signature · 1 line
2.35.1
author maintainer
#5 lore raw

PATCH v7 5/11 spi: aspeed: Adjust direct mapping to device size

In reply to Cédric Le Goater

The segment registers of the FMC/SPI controllers provide a way to configure the mapping window of the flash device contents on the AHB bus. Adjust this window to the size of the spi-mem mapping.

Things get more complex with multiple devices. The driver needs to also adjust the window of the next device to make sure that there is no overlap, even if there is no available device. The proposal below is not perfect but it is covering all the cases we have seen on different boards with one and two devices on the same bus.

---
 drivers/spi/spi-aspeed-smc.c | 88 ++++++++++++++++++++++++++++++++++++
 1 file changed, 88 insertions(+)
drivers/spi/spi-aspeed-smc.c +88 -0
index 50cc7bd7ba3e..0aff42e20b8d 100644
@@ -411,6 +411,92 @@ static int aspeed_spi_chip_set_default_window(struct aspeed_spi_chip *chip)
 	return chip->ahb_window_size ? 0 : -1;
 }
 
+static int aspeed_spi_set_window(struct aspeed_spi *aspi,
+				 const struct aspeed_spi_window *win)
+{
+	u32 start = aspi->ahb_base_phy + win->offset;
+	u32 end = start + win->size;
+	void __iomem *seg_reg = aspi->regs + CE0_SEGMENT_ADDR_REG + win->cs * 4;
+	u32 seg_val_backup = readl(seg_reg);
+	u32 seg_val = aspi->data->segment_reg(aspi, start, end);
+
+	if (seg_val == seg_val_backup)
+		return 0;
+
+	writel(seg_val, seg_reg);
+
+	/*
+	 * Restore initial value if something goes wrong else we could
+	 * loose access to the chip.
+	 */
+	if (seg_val != readl(seg_reg)) {
+		dev_err(aspi->dev, "CE%d invalid window [ 0x%.8x - 0x%.8x ] %dMB",
+			win->cs, start, end - 1, win->size >> 20);
+		writel(seg_val_backup, seg_reg);
+		return -EIO;
+	}
+
+	if (win->size)
+		dev_dbg(aspi->dev, "CE%d new window [ 0x%.8x - 0x%.8x ] %dMB",
+			win->cs, start, end - 1,  win->size >> 20);
+	else
+		dev_dbg(aspi->dev, "CE%d window closed", win->cs);
+
+	return 0;
+}
+
+/*
+ * Yet to be done when possible :
+ * - Align mappings on flash size (we don't have the info)
+ * - ioremap each window, not strictly necessary since the overall window
+ *   is correct.
+ */
+static int aspeed_spi_chip_adjust_window(struct aspeed_spi_chip *chip,
+					 u32 local_offset, u32 size)
+{
+	struct aspeed_spi *aspi = chip->aspi;
+	struct aspeed_spi_window windows[ASPEED_SPI_MAX_NUM_CS] = { 0 };
+	struct aspeed_spi_window *win = &windows[chip->cs];
+	int ret;
+
+	aspeed_spi_get_windows(aspi, windows);
+
+	/* Adjust this chip window */
+	win->offset += local_offset;
+	win->size = size;
+
+	if (win->offset + win->size > aspi->ahb_window_size) {
+		win->size = aspi->ahb_window_size - win->offset;
+		dev_warn(aspi->dev, "CE%d window resized to %dMB", chip->cs, win->size >> 20);
+	}
+
+	ret = aspeed_spi_set_window(aspi, win);
+	if (ret)
+		return ret;
+
+	/* Update chip mapping info */
+	chip->ahb_base = aspi->ahb_base + win->offset;
+	chip->ahb_window_size = win->size;
+
+	/*
+	 * Also adjust next chip window to make sure that it does not
+	 * overlap with the current window.
+	 */
+	if (chip->cs < aspi->data->max_cs - 1) {
+		struct aspeed_spi_window *next = &windows[chip->cs + 1];
+
+		/* Change offset and size to keep the same end address */
+		if ((next->offset + next->size) > (win->offset + win->size))
+			next->size = (next->offset + next->size) - (win->offset + win->size);
+		else
+			next->size = 0;
+		next->offset = win->offset + win->size;
+
+		aspeed_spi_set_window(aspi, next);
+	}
+	return 0;
+}
+
 static int aspeed_spi_dirmap_create(struct spi_mem_dirmap_desc *desc)
 {
 	struct aspeed_spi *aspi = spi_controller_get_devdata(desc->mem->spi->master);
@@ -425,6 +511,8 @@ static int aspeed_spi_dirmap_create(struct spi_mem_dirmap_desc *desc)
 	if (op->data.dir != SPI_MEM_DATA_IN)
 		return -EOPNOTSUPP;
 
+	aspeed_spi_chip_adjust_window(chip, desc->info.offset, desc->info.length);
+
 	if (desc->info.length > chip->ahb_window_size)
 		dev_warn(aspi->dev, "CE%d window (%dMB) too small for mapping",
 			 chip->cs, chip->ahb_window_size >> 20);
Signature · 1 line
2.35.1
author maintainer
#6 lore raw

PATCH v7 6/11 spi: aspeed: Workaround AST2500 limitations

In reply to Cédric Le Goater

It is not possible to configure a full 128MB window for a chip of the same size on the AST2500 SPI controller. For this case, the maximum window size is restricted to 120MB for CE0.

---
 drivers/spi/spi-aspeed-smc.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)
drivers/spi/spi-aspeed-smc.c +12 -0
index 0aff42e20b8d..d2b3342c133e 100644
@@ -451,6 +451,8 @@ static int aspeed_spi_set_window(struct aspeed_spi *aspi,
  * - ioremap each window, not strictly necessary since the overall window
  *   is correct.
  */
+static const struct aspeed_spi_data ast2500_spi_data;
+
 static int aspeed_spi_chip_adjust_window(struct aspeed_spi_chip *chip,
 					 u32 local_offset, u32 size)
 {
@@ -459,6 +461,16 @@ static int aspeed_spi_chip_adjust_window(struct aspeed_spi_chip *chip,
 	struct aspeed_spi_window *win = &windows[chip->cs];
 	int ret;
 
+	/*
+	 * Due to an HW issue on the AST2500 SPI controller, the CE0
+	 * window size should be smaller than the maximum 128MB.
+	 */
+	if (aspi->data == &ast2500_spi_data && chip->cs == 0 && size == SZ_128M) {
+		size = 120 << 20;
+		dev_info(aspi->dev, "CE%d window resized to %dMB (AST2500 HW quirk)",
+			 chip->cs, size >> 20);
+	}
+
 	aspeed_spi_get_windows(aspi, windows);
 
 	/* Adjust this chip window */
Signature · 1 line
2.35.1
author maintainer
#7 lore raw

PATCH v7 7/11 spi: aspeed: Add support for the AST2400 SPI controller

In reply to Cédric Le Goater

Extend the driver for the AST2400 SPI Flash Controller (SPI). This controller has a slightly different interface which requires adaptation of the 4B handling. Summary of features :

. host Firmware
. 1 chip select pin (CE0)
. slightly different register set, between AST2500 and the legacy
controller
. no segment registers
. single, dual mode.

---
 drivers/spi/spi-aspeed-smc.c | 33 ++++++++++++++++++++++++++++++++-
 1 file changed, 32 insertions(+), 1 deletion(-)
drivers/spi/spi-aspeed-smc.c +32 -1
index d2b3342c133e..d2b63c8bb88f 100644
@@ -30,6 +30,7 @@
 #define   CTRL_IO_DUAL_DATA		BIT(29)
 #define   CTRL_IO_QUAD_DATA		BIT(30)
 #define   CTRL_COMMAND_SHIFT		16
+#define   CTRL_IO_ADDRESS_4B		BIT(13)	/* AST2400 SPI only */
 #define   CTRL_IO_DUMMY_SET(dummy)					\
 	(((((dummy) >> 2) & 0x1) << 14) | (((dummy) & 0x3) << 6))
 #define   CTRL_CE_STOP_ACTIVE		BIT(2)
@@ -280,6 +281,8 @@ static bool aspeed_spi_supports_op(struct spi_mem *mem, const struct spi_mem_op
 	return spi_mem_default_supports_op(mem, op);
 }
 
+static const struct aspeed_spi_data ast2400_spi_data;
+
 static int do_aspeed_spi_exec_op(struct spi_mem *mem, const struct spi_mem_op *op)
 {
 	struct aspeed_spi *aspi = spi_controller_get_devdata(mem->spi->master);
@@ -309,6 +312,9 @@ static int do_aspeed_spi_exec_op(struct spi_mem *mem, const struct spi_mem_op *o
 			addr_mode |= (0x11 << chip->cs);
 		else
 			addr_mode &= ~(0x11 << chip->cs);
+
+		if (op->addr.nbytes == 4 && chip->aspi->data == &ast2400_spi_data)
+			ctl_val |= CTRL_IO_ADDRESS_4B;
 	}
 
 	if (op->dummy.nbytes)
@@ -398,7 +404,13 @@ static int aspeed_spi_chip_set_default_window(struct aspeed_spi_chip *chip)
 	struct aspeed_spi_window windows[ASPEED_SPI_MAX_NUM_CS] = { 0 };
 	struct aspeed_spi_window *win = &windows[chip->cs];
 
-	aspeed_spi_get_windows(aspi, windows);
+	/* No segment registers for the AST2400 SPI controller */
+	if (aspi->data == &ast2400_spi_data) {
+		win->offset = 0;
+		win->size = aspi->ahb_window_size;
+	} else {
+		aspeed_spi_get_windows(aspi, windows);
+	}
 
 	chip->ahb_base = aspi->ahb_base + win->offset;
 	chip->ahb_window_size = win->size;
@@ -461,6 +473,10 @@ static int aspeed_spi_chip_adjust_window(struct aspeed_spi_chip *chip,
 	struct aspeed_spi_window *win = &windows[chip->cs];
 	int ret;
 
+	/* No segment registers for the AST2400 SPI controller */
+	if (aspi->data == &ast2400_spi_data)
+		return 0;
+
 	/*
 	 * Due to an HW issue on the AST2500 SPI controller, the CE0
 	 * window size should be smaller than the maximum 128MB.
@@ -545,6 +561,12 @@ static int aspeed_spi_dirmap_create(struct spi_mem_dirmap_desc *desc)
 		else
 			addr_mode &= ~(0x11 << chip->cs);
 		writel(addr_mode, aspi->regs + CE_CTRL_REG);
+
+		/* AST2400 SPI controller sets 4BYTE address mode in
+		 * CE0 Control Register
+		 */
+		if (op->addr.nbytes == 4 && chip->aspi->data == &ast2400_spi_data)
+			ctl_val |= CTRL_IO_ADDRESS_4B;
 	}
 
 	/* READ mode is the controller default setting */
@@ -816,6 +838,14 @@ static const struct aspeed_spi_data ast2400_fmc_data = {
 	.segment_reg   = aspeed_spi_segment_reg,
 };
 
+static const struct aspeed_spi_data ast2400_spi_data = {
+	.max_cs	       = 1,
+	.hastype       = false,
+	.we0	       = 0,
+	.ctl0	       = 0x04,
+	/* No segment registers */
+};
+
 static const struct aspeed_spi_data ast2500_fmc_data = {
 	.max_cs	       = 3,
 	.hastype       = true,
@@ -860,6 +890,7 @@ static const struct aspeed_spi_data ast2600_spi_data = {
 
 static const struct of_device_id aspeed_spi_matches[] = {
 	{ .compatible = "aspeed,ast2400-fmc", .data = &ast2400_fmc_data },
+	{ .compatible = "aspeed,ast2400-spi", .data = &ast2400_spi_data },
 	{ .compatible = "aspeed,ast2500-fmc", .data = &ast2500_fmc_data },
 	{ .compatible = "aspeed,ast2500-spi", .data = &ast2500_spi_data },
 	{ .compatible = "aspeed,ast2600-fmc", .data = &ast2600_fmc_data },
Signature · 1 line
2.35.1
author maintainer
#8 lore raw

PATCH v7 8/11 spi: aspeed: Calibrate read timings

In reply to Cédric Le Goater

To accommodate the different response time of SPI transfers on different boards and different SPI NOR devices, the Aspeed controllers provide a set of Read Timing Compensation registers to tune the timing delays depending on the frequency being used. The AST2600 SoC has one of these registers per device. On the AST2500 and AST2400 SoCs, the timing register is shared by all devices which is problematic to get good results other than for one device.

The algorithm first reads a golden buffer at low speed and then performs reads with different clocks and delay cycle settings to find a breaking point. This selects a default good frequency for the CEx control register. The current settings are a bit optimistic as we pick the first delay giving good results. A safer approach would be to determine an interval and choose the middle value.

Calibration is performed when the direct mapping for reads is created. Since the underlying spi-nor object needs to be initialized to create the spi_mem operation for direct mapping, we should be fine. Having a specific API would clarify the requirements though.

---
 drivers/spi/spi-aspeed-smc.c | 281 +++++++++++++++++++++++++++++++++++
 1 file changed, 281 insertions(+)
drivers/spi/spi-aspeed-smc.c +281 -0
index d2b63c8bb88f..35f6934847b7 100644
@@ -33,6 +33,8 @@
 #define   CTRL_IO_ADDRESS_4B		BIT(13)	/* AST2400 SPI only */
 #define   CTRL_IO_DUMMY_SET(dummy)					\
 	(((((dummy) >> 2) & 0x1) << 14) | (((dummy) & 0x3) << 6))
+#define   CTRL_FREQ_SEL_SHIFT		8
+#define   CTRL_FREQ_SEL_MASK		GENMASK(11, CTRL_FREQ_SEL_SHIFT)
 #define   CTRL_CE_STOP_ACTIVE		BIT(2)
 #define   CTRL_IO_MODE_CMD_MASK		GENMASK(1, 0)
 #define   CTRL_IO_MODE_NORMAL		0x0
@@ -45,6 +47,9 @@
 /* CEx Address Decoding Range Register */
 #define CE0_SEGMENT_ADDR_REG		0x30
 
+/* CEx Read timing compensation register */
+#define CE0_TIMING_COMPENSATION_REG	0x94
+
 enum aspeed_spi_ctl_reg_value {
 	ASPEED_SPI_BASE,
 	ASPEED_SPI_READ,
@@ -70,10 +75,15 @@ struct aspeed_spi_data {
 	bool	hastype;
 	u32	mode_bits;
 	u32	we0;
+	u32	timing;
+	u32	hclk_mask;
+	u32	hdiv_max;
 
 	u32 (*segment_start)(struct aspeed_spi *aspi, u32 reg);
 	u32 (*segment_end)(struct aspeed_spi *aspi, u32 reg);
 	u32 (*segment_reg)(struct aspeed_spi *aspi, u32 start, u32 end);
+	int (*calibrate)(struct aspeed_spi_chip *chip, u32 hdiv,
+			 const u8 *golden_buf, u8 *test_buf);
 };
 
 #define ASPEED_SPI_MAX_NUM_CS	5
@@ -525,6 +535,8 @@ static int aspeed_spi_chip_adjust_window(struct aspeed_spi_chip *chip,
 	return 0;
 }
 
+static int aspeed_spi_do_calibration(struct aspeed_spi_chip *chip);
+
 static int aspeed_spi_dirmap_create(struct spi_mem_dirmap_desc *desc)
 {
 	struct aspeed_spi *aspi = spi_controller_get_devdata(desc->mem->spi->master);
@@ -573,6 +585,8 @@ static int aspeed_spi_dirmap_create(struct spi_mem_dirmap_desc *desc)
 	chip->ctl_val[ASPEED_SPI_READ] = ctl_val;
 	writel(chip->ctl_val[ASPEED_SPI_READ], chip->ctl);
 
+	ret = aspeed_spi_do_calibration(chip);
+
 	dev_info(aspi->dev, "CE%d read buswidth:%d [0x%08x]\n",
 		 chip->cs, op->data.buswidth, chip->ctl_val[ASPEED_SPI_READ]);
 
@@ -825,6 +839,249 @@ static u32 aspeed_spi_segment_ast2600_reg(struct aspeed_spi *aspi,
 		((end - 1) & AST2600_SEG_ADDR_MASK);
 }
 
+/*
+ * Read timing compensation sequences
+ */
+
+#define CALIBRATE_BUF_SIZE SZ_16K
+
+static bool aspeed_spi_check_reads(struct aspeed_spi_chip *chip,
+				   const u8 *golden_buf, u8 *test_buf)
+{
+	int i;
+
+	for (i = 0; i < 10; i++) {
+		memcpy_fromio(test_buf, chip->ahb_base, CALIBRATE_BUF_SIZE);
+		if (memcmp(test_buf, golden_buf, CALIBRATE_BUF_SIZE) != 0) {
+#if defined(VERBOSE_DEBUG)
+			print_hex_dump_bytes(DEVICE_NAME "  fail: ", DUMP_PREFIX_NONE,
+					     test_buf, 0x100);
+#endif
+			return false;
+		}
+	}
+	return true;
+}
+
+#define FREAD_TPASS(i)	(((i) / 2) | (((i) & 1) ? 0 : 8))
+
+/*
+ * The timing register is shared by all devices. Only update for CE0.
+ */
+static int aspeed_spi_calibrate(struct aspeed_spi_chip *chip, u32 hdiv,
+				const u8 *golden_buf, u8 *test_buf)
+{
+	struct aspeed_spi *aspi = chip->aspi;
+	const struct aspeed_spi_data *data = aspi->data;
+	int i;
+	int good_pass = -1, pass_count = 0;
+	u32 shift = (hdiv - 1) << 2;
+	u32 mask = ~(0xfu << shift);
+	u32 fread_timing_val = 0;
+
+	/* Try HCLK delay 0..5, each one with/without delay and look for a
+	 * good pair.
+	 */
+	for (i = 0; i < 12; i++) {
+		bool pass;
+
+		if (chip->cs == 0) {
+			fread_timing_val &= mask;
+			fread_timing_val |= FREAD_TPASS(i) << shift;
+			writel(fread_timing_val, aspi->regs + data->timing);
+		}
+		pass = aspeed_spi_check_reads(chip, golden_buf, test_buf);
+		dev_dbg(aspi->dev,
+			"  * [%08x] %d HCLK delay, %dns DI delay : %s",
+			fread_timing_val, i / 2, (i & 1) ? 0 : 4,
+			pass ? "PASS" : "FAIL");
+		if (pass) {
+			pass_count++;
+			if (pass_count == 3) {
+				good_pass = i - 1;
+				break;
+			}
+		} else {
+			pass_count = 0;
+		}
+	}
+
+	/* No good setting for this frequency */
+	if (good_pass < 0)
+		return -1;
+
+	/* We have@least one pass of margin, let's use first pass */
+	if (chip->cs == 0) {
+		fread_timing_val &= mask;
+		fread_timing_val |= FREAD_TPASS(good_pass) << shift;
+		writel(fread_timing_val, aspi->regs + data->timing);
+	}
+	dev_dbg(aspi->dev, " * -> good is pass %d [0x%08x]",
+		good_pass, fread_timing_val);
+	return 0;
+}
+
+static bool aspeed_spi_check_calib_data(const u8 *test_buf, u32 size)
+{
+	const u32 *tb32 = (const u32 *)test_buf;
+	u32 i, cnt = 0;
+
+	/* We check if we have enough words that are neither all 0
+	 * nor all 1's so the calibration can be considered valid.
+	 *
+	 * I use an arbitrary threshold for now of 64
+	 */
+	size >>= 2;
+	for (i = 0; i < size; i++) {
+		if (tb32[i] != 0 && tb32[i] != 0xffffffff)
+			cnt++;
+	}
+	return cnt >= 64;
+}
+
+static const u32 aspeed_spi_hclk_divs[] = {
+	0xf, /* HCLK */
+	0x7, /* HCLK/2 */
+	0xe, /* HCLK/3 */
+	0x6, /* HCLK/4 */
+	0xd, /* HCLK/5 */
+};
+
+#define ASPEED_SPI_HCLK_DIV(i) \
+	(aspeed_spi_hclk_divs[(i) - 1] << CTRL_FREQ_SEL_SHIFT)
+
+static int aspeed_spi_do_calibration(struct aspeed_spi_chip *chip)
+{
+	struct aspeed_spi *aspi = chip->aspi;
+	const struct aspeed_spi_data *data = aspi->data;
+	u32 ahb_freq = aspi->clk_freq;
+	u32 max_freq = chip->clk_freq;
+	u32 ctl_val;
+	u8 *golden_buf = NULL;
+	u8 *test_buf = NULL;
+	int i, rc, best_div = -1;
+
+	dev_dbg(aspi->dev, "calculate timing compensation - AHB freq: %d MHz",
+		ahb_freq / 1000000);
+
+	/*
+	 * use the related low frequency to get check calibration data
+	 * and get golden data.
+	 */
+	ctl_val = chip->ctl_val[ASPEED_SPI_READ] & data->hclk_mask;
+	writel(ctl_val, chip->ctl);
+
+	test_buf = kzalloc(CALIBRATE_BUF_SIZE * 2, GFP_KERNEL);
+	if (!test_buf)
+		return -ENOMEM;
+
+	golden_buf = test_buf + CALIBRATE_BUF_SIZE;
+
+	memcpy_fromio(golden_buf, chip->ahb_base, CALIBRATE_BUF_SIZE);
+	if (!aspeed_spi_check_calib_data(golden_buf, CALIBRATE_BUF_SIZE)) {
+		dev_info(aspi->dev, "Calibration area too uniform, using low speed");
+		goto no_calib;
+	}
+
+#if defined(VERBOSE_DEBUG)
+	print_hex_dump_bytes(DEVICE_NAME "  good: ", DUMP_PREFIX_NONE,
+			     golden_buf, 0x100);
+#endif
+
+	/* Now we iterate the HCLK dividers until we find our breaking point */
+	for (i = ARRAY_SIZE(aspeed_spi_hclk_divs); i > data->hdiv_max - 1; i--) {
+		u32 tv, freq;
+
+		freq = ahb_freq / i;
+		if (freq > max_freq)
+			continue;
+
+		/* Set the timing */
+		tv = chip->ctl_val[ASPEED_SPI_READ] | ASPEED_SPI_HCLK_DIV(i);
+		writel(tv, chip->ctl);
+		dev_dbg(aspi->dev, "Trying HCLK/%d [%08x] ...", i, tv);
+		rc = data->calibrate(chip, i, golden_buf, test_buf);
+		if (rc == 0)
+			best_div = i;
+	}
+
+	/* Nothing found ? */
+	if (best_div < 0) {
+		dev_warn(aspi->dev, "No good frequency, using dumb slow");
+	} else {
+		dev_dbg(aspi->dev, "Found good read timings at HCLK/%d", best_div);
+
+		/* Record the freq */
+		for (i = 0; i < ASPEED_SPI_MAX; i++)
+			chip->ctl_val[i] = (chip->ctl_val[i] & data->hclk_mask) |
+				ASPEED_SPI_HCLK_DIV(best_div);
+	}
+
+no_calib:
+	writel(chip->ctl_val[ASPEED_SPI_READ], chip->ctl);
+	kfree(test_buf);
+	return 0;
+}
+
+#define TIMING_DELAY_DI		BIT(3)
+#define TIMING_DELAY_HCYCLE_MAX	5
+#define TIMING_REG_AST2600(chip)				\
+	((chip)->aspi->regs + (chip)->aspi->data->timing +	\
+	 (chip)->cs * 4)
+
+static int aspeed_spi_ast2600_calibrate(struct aspeed_spi_chip *chip, u32 hdiv,
+					const u8 *golden_buf, u8 *test_buf)
+{
+	struct aspeed_spi *aspi = chip->aspi;
+	int hcycle;
+	u32 shift = (hdiv - 2) << 3;
+	u32 mask = ~(0xfu << shift);
+	u32 fread_timing_val = 0;
+
+	for (hcycle = 0; hcycle <= TIMING_DELAY_HCYCLE_MAX; hcycle++) {
+		int delay_ns;
+		bool pass = false;
+
+		fread_timing_val &= mask;
+		fread_timing_val |= hcycle << shift;
+
+		/* no DI input delay first  */
+		writel(fread_timing_val, TIMING_REG_AST2600(chip));
+		pass = aspeed_spi_check_reads(chip, golden_buf, test_buf);
+		dev_dbg(aspi->dev,
+			"  * [%08x] %d HCLK delay, DI delay none : %s",
+			fread_timing_val, hcycle, pass ? "PASS" : "FAIL");
+		if (pass)
+			return 0;
+
+		/* Add DI input delays  */
+		fread_timing_val &= mask;
+		fread_timing_val |= (TIMING_DELAY_DI | hcycle) << shift;
+
+		for (delay_ns = 0; delay_ns < 0x10; delay_ns++) {
+			fread_timing_val &= ~(0xf << (4 + shift));
+			fread_timing_val |= delay_ns << (4 + shift);
+
+			writel(fread_timing_val, TIMING_REG_AST2600(chip));
+			pass = aspeed_spi_check_reads(chip, golden_buf, test_buf);
+			dev_dbg(aspi->dev,
+				"  * [%08x] %d HCLK delay, DI delay %d.%dns : %s",
+				fread_timing_val, hcycle, (delay_ns + 1) / 2,
+				(delay_ns + 1) & 1 ? 5 : 5, pass ? "PASS" : "FAIL");
+			/*
+			 * TODO: This is optimistic. We should look
+			 * for a working interval and save the middle
+			 * value in the read timing register.
+			 */
+			if (pass)
+				return 0;
+		}
+	}
+
+	/* No good setting for this frequency */
+	return -1;
+}
+
 /*
  * Platform definitions
  */
@@ -833,6 +1090,10 @@ static const struct aspeed_spi_data ast2400_fmc_data = {
 	.hastype       = true,
 	.we0	       = 16,
 	.ctl0	       = CE0_CTRL_REG,
+	.timing	       = CE0_TIMING_COMPENSATION_REG,
+	.hclk_mask     = 0xfffff0ff,
+	.hdiv_max      = 1,
+	.calibrate     = aspeed_spi_calibrate,
 	.segment_start = aspeed_spi_segment_start,
 	.segment_end   = aspeed_spi_segment_end,
 	.segment_reg   = aspeed_spi_segment_reg,
@@ -843,6 +1104,10 @@ static const struct aspeed_spi_data ast2400_spi_data = {
 	.hastype       = false,
 	.we0	       = 0,
 	.ctl0	       = 0x04,
+	.timing	       = 0x14,
+	.hclk_mask     = 0xfffff0ff,
+	.hdiv_max      = 1,
+	.calibrate     = aspeed_spi_calibrate,
 	/* No segment registers */
 };
 
@@ -851,6 +1116,10 @@ static const struct aspeed_spi_data ast2500_fmc_data = {
 	.hastype       = true,
 	.we0	       = 16,
 	.ctl0	       = CE0_CTRL_REG,
+	.timing	       = CE0_TIMING_COMPENSATION_REG,
+	.hclk_mask     = 0xffffd0ff,
+	.hdiv_max      = 1,
+	.calibrate     = aspeed_spi_calibrate,
 	.segment_start = aspeed_spi_segment_start,
 	.segment_end   = aspeed_spi_segment_end,
 	.segment_reg   = aspeed_spi_segment_reg,
@@ -861,6 +1130,10 @@ static const struct aspeed_spi_data ast2500_spi_data = {
 	.hastype       = false,
 	.we0	       = 16,
 	.ctl0	       = CE0_CTRL_REG,
+	.timing	       = CE0_TIMING_COMPENSATION_REG,
+	.hclk_mask     = 0xffffd0ff,
+	.hdiv_max      = 1,
+	.calibrate     = aspeed_spi_calibrate,
 	.segment_start = aspeed_spi_segment_start,
 	.segment_end   = aspeed_spi_segment_end,
 	.segment_reg   = aspeed_spi_segment_reg,
@@ -872,6 +1145,10 @@ static const struct aspeed_spi_data ast2600_fmc_data = {
 	.mode_bits     = SPI_RX_QUAD | SPI_RX_QUAD,
 	.we0	       = 16,
 	.ctl0	       = CE0_CTRL_REG,
+	.timing	       = CE0_TIMING_COMPENSATION_REG,
+	.hclk_mask     = 0xf0fff0ff,
+	.hdiv_max      = 2,
+	.calibrate     = aspeed_spi_ast2600_calibrate,
 	.segment_start = aspeed_spi_segment_ast2600_start,
 	.segment_end   = aspeed_spi_segment_ast2600_end,
 	.segment_reg   = aspeed_spi_segment_ast2600_reg,
@@ -883,6 +1160,10 @@ static const struct aspeed_spi_data ast2600_spi_data = {
 	.mode_bits     = SPI_RX_QUAD | SPI_RX_QUAD,
 	.we0	       = 16,
 	.ctl0	       = CE0_CTRL_REG,
+	.timing	       = CE0_TIMING_COMPENSATION_REG,
+	.hclk_mask     = 0xf0fff0ff,
+	.hdiv_max      = 2,
+	.calibrate     = aspeed_spi_ast2600_calibrate,
 	.segment_start = aspeed_spi_segment_ast2600_start,
 	.segment_end   = aspeed_spi_segment_ast2600_end,
 	.segment_reg   = aspeed_spi_segment_ast2600_reg,
Signature · 1 line
2.35.1
author maintainer
#9 lore raw

PATCH v7 9/11 ARM: dts: aspeed: Enable Dual SPI RX transfers

In reply to Cédric Le Goater

All these controllers support at least Dual SPI. Update the DTs.

---
 arch/arm/boot/dts/aspeed-g4.dtsi | 6 ++++++
 arch/arm/boot/dts/aspeed-g5.dtsi | 7 +++++++
 arch/arm/boot/dts/aspeed-g6.dtsi | 8 ++++++++
 3 files changed, 21 insertions(+)
arch/arm/boot/dts/aspeed-g4.dtsi +6 -0
index 3b7b98e787e7..1c6354cec9f2 100644
@@ -64,27 +64,32 @@ fmc: spi at 1e620000 {
 			flash at 0 {
 				reg = < 0 >;
 				compatible = "jedec,spi-nor";
+				spi-rx-bus-width = <2>;
 				spi-max-frequency = <50000000>;
 				status = "disabled";
 			};
 			flash at 1 {
 				reg = < 1 >;
 				compatible = "jedec,spi-nor";
+				spi-rx-bus-width = <2>;
 				status = "disabled";
 			};
 			flash at 2 {
 				reg = < 2 >;
 				compatible = "jedec,spi-nor";
+				spi-rx-bus-width = <2>;
 				status = "disabled";
 			};
 			flash at 3 {
 				reg = < 3 >;
 				compatible = "jedec,spi-nor";
+				spi-rx-bus-width = <2>;
 				status = "disabled";
 			};
 			flash at 4 {
 				reg = < 4 >;
 				compatible = "jedec,spi-nor";
+				spi-rx-bus-width = <2>;
 				status = "disabled";
 			};
 		};
@@ -100,6 +105,7 @@ flash at 0 {
 				reg = < 0 >;
 				compatible = "jedec,spi-nor";
 				spi-max-frequency = <50000000>;
+				spi-rx-bus-width = <2>;
 				status = "disabled";
 			};
 		};
arch/arm/boot/dts/aspeed-g5.dtsi +7 -0
index 1e727cbaab14..c89092c3905b 100644
@@ -66,18 +66,21 @@ flash at 0 {
 				reg = < 0 >;
 				compatible = "jedec,spi-nor";
 				spi-max-frequency = <50000000>;
+				spi-rx-bus-width = <2>;
 				status = "disabled";
 			};
 			flash at 1 {
 				reg = < 1 >;
 				compatible = "jedec,spi-nor";
 				spi-max-frequency = <50000000>;
+				spi-rx-bus-width = <2>;
 				status = "disabled";
 			};
 			flash at 2 {
 				reg = < 2 >;
 				compatible = "jedec,spi-nor";
 				spi-max-frequency = <50000000>;
+				spi-rx-bus-width = <2>;
 				status = "disabled";
 			};
 		};
@@ -93,12 +96,14 @@ flash at 0 {
 				reg = < 0 >;
 				compatible = "jedec,spi-nor";
 				spi-max-frequency = <50000000>;
+				spi-rx-bus-width = <2>;
 				status = "disabled";
 			};
 			flash at 1 {
 				reg = < 1 >;
 				compatible = "jedec,spi-nor";
 				spi-max-frequency = <50000000>;
+				spi-rx-bus-width = <2>;
 				status = "disabled";
 			};
 		};
@@ -114,12 +119,14 @@ flash at 0 {
 				reg = < 0 >;
 				compatible = "jedec,spi-nor";
 				spi-max-frequency = <50000000>;
+				spi-rx-bus-width = <2>;
 				status = "disabled";
 			};
 			flash at 1 {
 				reg = < 1 >;
 				compatible = "jedec,spi-nor";
 				spi-max-frequency = <50000000>;
+				spi-rx-bus-width = <2>;
 				status = "disabled";
 			};
 		};
arch/arm/boot/dts/aspeed-g6.dtsi +8 -0
index 6ae5dfecbdf1..4264fc0a7c3f 100644
@@ -106,18 +106,21 @@ flash at 0 {
 				reg = < 0 >;
 				compatible = "jedec,spi-nor";
 				spi-max-frequency = <50000000>;
+				spi-rx-bus-width = <2>;
 				status = "disabled";
 			};
 			flash at 1 {
 				reg = < 1 >;
 				compatible = "jedec,spi-nor";
 				spi-max-frequency = <50000000>;
+				spi-rx-bus-width = <2>;
 				status = "disabled";
 			};
 			flash at 2 {
 				reg = < 2 >;
 				compatible = "jedec,spi-nor";
 				spi-max-frequency = <50000000>;
+				spi-rx-bus-width = <2>;
 				status = "disabled";
 			};
 		};
@@ -133,12 +136,14 @@ flash at 0 {
 				reg = < 0 >;
 				compatible = "jedec,spi-nor";
 				spi-max-frequency = <50000000>;
+				spi-rx-bus-width = <2>;
 				status = "disabled";
 			};
 			flash at 1 {
 				reg = < 1 >;
 				compatible = "jedec,spi-nor";
 				spi-max-frequency = <50000000>;
+				spi-rx-bus-width = <2>;
 				status = "disabled";
 			};
 		};
@@ -154,18 +159,21 @@ flash at 0 {
 				reg = < 0 >;
 				compatible = "jedec,spi-nor";
 				spi-max-frequency = <50000000>;
+				spi-rx-bus-width = <2>;
 				status = "disabled";
 			};
 			flash at 1 {
 				reg = < 1 >;
 				compatible = "jedec,spi-nor";
 				spi-max-frequency = <50000000>;
+				spi-rx-bus-width = <2>;
 				status = "disabled";
 			};
 			flash at 2 {
 				reg = < 2 >;
 				compatible = "jedec,spi-nor";
 				spi-max-frequency = <50000000>;
+				spi-rx-bus-width = <2>;
 				status = "disabled";
 			};
 		};
Signature · 1 line
2.35.1
author maintainer
#10 lore raw

PATCH v7 10/11 ARM: dts: aspeed-g4: Set spi-max-frequency for all flashes

In reply to Cédric Le Goater

From: Tao Ren <[email protected]>

Set "spi-max-frequency" to 50 MHz for all the flashes under the FMC controller to ensure the clock frequency is calculated correctly.

---
 arch/arm/boot/dts/aspeed-g4.dtsi | 4 ++++
 1 file changed, 4 insertions(+)
arch/arm/boot/dts/aspeed-g4.dtsi +4 -0
index 1c6354cec9f2..530491ae5eb2 100644
@@ -72,24 +72,28 @@ flash at 1 {
 				reg = < 1 >;
 				compatible = "jedec,spi-nor";
 				spi-rx-bus-width = <2>;
+				spi-max-frequency = <50000000>;
 				status = "disabled";
 			};
 			flash at 2 {
 				reg = < 2 >;
 				compatible = "jedec,spi-nor";
 				spi-rx-bus-width = <2>;
+				spi-max-frequency = <50000000>;
 				status = "disabled";
 			};
 			flash at 3 {
 				reg = < 3 >;
 				compatible = "jedec,spi-nor";
 				spi-rx-bus-width = <2>;
+				spi-max-frequency = <50000000>;
 				status = "disabled";
 			};
 			flash at 4 {
 				reg = < 4 >;
 				compatible = "jedec,spi-nor";
 				spi-rx-bus-width = <2>;
+				spi-max-frequency = <50000000>;
 				status = "disabled";
 			};
 		};
Signature · 1 line
2.35.1
author maintainer
#11 lore raw

PATCH v7 11/11 mtd: spi-nor: aspeed: set the decoding size to at least 2MB for AST2600

In reply to Cédric Le Goater

From: Potin Lai <[email protected]>

In AST2600, the unit of SPI CEx decoding range register is 1MB, and end address offset is set to the acctual offset - 1MB. If the flash only has 1MB, the end address will has same value as start address, which will causing unexpected errors.

This patch set the decoding size to at least 2MB to avoid decoding errors.

Tested:
root at bletchley:~# dmesg | grep "aspeed-smc 1e631000.spi: CE0 window"

[   59.328134] aspeed-smc 1e631000.spi: CE0 window resized to 2MB (AST2600 Decoding)
[   59.343001] aspeed-smc 1e631000.spi: CE0 window [ 0x50000000 - 0x50200000 ] 2MB

root at bletchley:~# devmem 0x1e631030
0x00100000

[ clg : Ported on new spi-mem driver ]
---
 drivers/spi/spi-aspeed-smc.c | 13 +++++++++++++
 1 file changed, 13 insertions(+)
drivers/spi/spi-aspeed-smc.c +13 -0
index 35f6934847b7..496f3e1e9079 100644
@@ -474,6 +474,8 @@ static int aspeed_spi_set_window(struct aspeed_spi *aspi,
  *   is correct.
  */
 static const struct aspeed_spi_data ast2500_spi_data;
+static const struct aspeed_spi_data ast2600_spi_data;
+static const struct aspeed_spi_data ast2600_fmc_data;
 
 static int aspeed_spi_chip_adjust_window(struct aspeed_spi_chip *chip,
 					 u32 local_offset, u32 size)
@@ -497,6 +499,17 @@ static int aspeed_spi_chip_adjust_window(struct aspeed_spi_chip *chip,
 			 chip->cs, size >> 20);
 	}
 
+	/*
+	 * The decoding size of AST2600 SPI controller should set at
+	 * least 2MB.
+	 */
+	if ((aspi->data == &ast2600_spi_data || aspi->data == &ast2600_fmc_data) &&
+	    size < SZ_2M) {
+		size = SZ_2M;
+		dev_info(aspi->dev, "CE%d window resized to %dMB (AST2600 Decoding)",
+			 chip->cs, size >> 20);
+	}
+
 	aspeed_spi_get_windows(aspi, windows);
 
 	/* Adjust this chip window */
Signature · 1 line
2.35.1
maintainer
#12 lore raw

Re: (subset) [PATCH v7 00/11] spi: spi-mem: Convert Aspeed SMC driver to spi-mem

In reply to Cédric Le Goater
Quoted 4 lines · C?dric Le Goater

This series adds a new SPI driver using the spi-mem interface for the Aspeed static memory controllers of the AST2600, AST2500 and AST2400 SoCs.

Quoted 8 lines

* AST2600 Firmware SPI Memory Controller (FMC)
* AST2600 SPI Flash Controller (SPI1 and SPI2)
* AST2500 Firmware SPI Memory Controller (FMC)
* AST2500 SPI Flash Controller (SPI1 and SPI2)
* AST2400 New Static Memory Controller (also referred as FMC)
* AST2400 SPI Flash Controller (SPI)

[...]

Applied to

https://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi.git for-next

Thanks!

[02/11] dt-bindings: spi: Convert the Aspeed SMC controllers device tree binding

        commit: ce9858ea499da025684a7a5f19823c2c3f14bdce
[03/11] spi: spi-mem: Convert Aspeed SMC driver to spi-mem
        commit: 9c63b846e6df43e5b3d31263f7db545f32deeda3
[04/11] spi: aspeed: Add support for direct mapping
        commit: 9da06d7bdec7dad8018c23b180e410ef2e7a4367
[05/11] spi: aspeed: Adjust direct mapping to device size
        commit: bb084f94e1bca4a5c4f689d7aa9b410220c1ed71
[06/11] spi: aspeed: Workaround AST2500 limitations
        commit: 5785eedee42c34cfec496199a80fa8ec9ddcf7fe
[07/11] spi: aspeed: Add support for the AST2400 SPI controller
        commit: 53526ab27d9c256504f267713aea60db7af18fb0
[08/11] spi: aspeed: Calibrate read timings
        commit: eeaec1ea05c0e0f08e04c6844f20cc24a2fcc0f4
[11/11] mtd: spi-nor: aspeed: set the decoding size to at least 2MB for AST2600
        commit: 73ae97e3cabb580639f02f12a192324a53c4bebb

All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.

You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.

If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.

Please add any relevant lists and maintainers to the CCs when replying to this mail.

Thanks,
Mark

#13 lore raw

Re: (subset) [PATCH v7 00/11] spi: spi-mem: Convert Aspeed SMC driver to spi-mem

Hi Cedric,

Quoted 21 lines · Mark Brown
Quoted 4 lines · C?dric Le Goater

This series adds a new SPI driver using the spi-mem interface for the Aspeed static memory controllers of the AST2600, AST2500 and AST2400 SoCs.

Quoted 8 lines

* AST2600 Firmware SPI Memory Controller (FMC)
* AST2600 SPI Flash Controller (SPI1 and SPI2)
* AST2500 Firmware SPI Memory Controller (FMC)
* AST2500 SPI Flash Controller (SPI1 and SPI2)
* AST2400 New Static Memory Controller (also referred as FMC)
* AST2400 SPI Flash Controller (SPI)

[...]

Applied to

https://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi.git for-next

Thanks!

[02/11] dt-bindings: spi: Convert the Aspeed SMC controllers device tree binding

Quoted 13 lines
        commit: ce9858ea499da025684a7a5f19823c2c3f14bdce
[03/11] spi: spi-mem: Convert Aspeed SMC driver to spi-mem
        commit: 9c63b846e6df43e5b3d31263f7db545f32deeda3
[04/11] spi: aspeed: Add support for direct mapping
        commit: 9da06d7bdec7dad8018c23b180e410ef2e7a4367
[05/11] spi: aspeed: Adjust direct mapping to device size
        commit: bb084f94e1bca4a5c4f689d7aa9b410220c1ed71
[06/11] spi: aspeed: Workaround AST2500 limitations
        commit: 5785eedee42c34cfec496199a80fa8ec9ddcf7fe
[07/11] spi: aspeed: Add support for the AST2400 SPI controller
        commit: 53526ab27d9c256504f267713aea60db7af18fb0
[08/11] spi: aspeed: Calibrate read timings
        commit: eeaec1ea05c0e0f08e04c6844f20cc24a2fcc0f4

I have repeatedly objected to this patch [0][1][2] and you have repeatedly decided to not address my objections. I won't spend any more time fighting it. But I will say that you should not expect any guarantees that SPI NOR or SPI NAND will not break your calibration in the future if they decide to move the dirmap_create() call around.

Quoted 3 lines

[11/11] mtd: spi-nor: aspeed: set the decoding size to at least 2MB for AST2600

        commit: 73ae97e3cabb580639f02f12a192324a53c4bebb

[0] https://patchwork.kernel.org/project/spi-devel-general/patch/20220325100849.2019209-9-clg at kaod.org/
[1] https://patchwork.kernel.org/project/spi-devel-general/patch/20220214094231.3753686-9-clg at kaod.org/
[2] https://lore.kernel.org/all/20220208190636.h6dubktkmuosvdxo at ti.com/

Signature · 3 lines
Regards,
Pratyush Yadav
Texas Instruments Inc.
author maintainer
#14 lore raw

Re: (subset) [PATCH v7 00/11] spi: spi-mem: Convert Aspeed SMC driver to spi-mem

Pratyush,

Quoted 38 lines · Pratyush Yadav

Hi Cedric,

Quoted 21 lines · Mark Brown
Quoted 4 lines · C?dric Le Goater

This series adds a new SPI driver using the spi-mem interface for the Aspeed static memory controllers of the AST2600, AST2500 and AST2400 SoCs.

Quoted 8 lines

* AST2600 Firmware SPI Memory Controller (FMC)
* AST2600 SPI Flash Controller (SPI1 and SPI2)
* AST2500 Firmware SPI Memory Controller (FMC)
* AST2500 SPI Flash Controller (SPI1 and SPI2)
* AST2400 New Static Memory Controller (also referred as FMC)
* AST2400 SPI Flash Controller (SPI)

[...]

Applied to

https://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi.git for-next

Thanks!

[02/11] dt-bindings: spi: Convert the Aspeed SMC controllers device tree binding

Quoted 13 lines
         commit: ce9858ea499da025684a7a5f19823c2c3f14bdce
[03/11] spi: spi-mem: Convert Aspeed SMC driver to spi-mem
         commit: 9c63b846e6df43e5b3d31263f7db545f32deeda3
[04/11] spi: aspeed: Add support for direct mapping
         commit: 9da06d7bdec7dad8018c23b180e410ef2e7a4367
[05/11] spi: aspeed: Adjust direct mapping to device size
         commit: bb084f94e1bca4a5c4f689d7aa9b410220c1ed71
[06/11] spi: aspeed: Workaround AST2500 limitations
         commit: 5785eedee42c34cfec496199a80fa8ec9ddcf7fe
[07/11] spi: aspeed: Add support for the AST2400 SPI controller
         commit: 53526ab27d9c256504f267713aea60db7af18fb0
[08/11] spi: aspeed: Calibrate read timings
         commit: eeaec1ea05c0e0f08e04c6844f20cc24a2fcc0f4
Quoted 2 lines

I have repeatedly objected to this patch [0][1][2] and you have repeatedly decided to not address my objections.

That's a very harsh way of saying things. I did not decide anything or ignore your comments. I answered your questions and acknowledged that indeed the read training was done under the dirmap handler but this was not a concern today since we had all the required information from spimem.

We waited _together_, 5 or 6 weeks, for more inputs on how to address the concerns you raised regarding the sustainability of this method.

Quoted 1 line

I won't spend any more time fighting it.

This is not a fight. I don't know why you interpret it that way.

Now, since you object so explicitly, and since this patchset has not reached the Linux kernel yet, we should consider dropping it. I rather do that than push crap in mainline. But then, please, provide solutions and not only objections !

Quoted 3 lines

But I will say that you should not expect any
guarantees that SPI NOR or SPI NAND will not break your calibration in the future if they decide to move the dirmap_create() call around.

If that's the case one day, we have multiple solutions :

- stop doing the training
- move the training to the appropriate handler if it exists
- use a default value

Quoted 7 lines
Quoted 3 lines

[11/11] mtd: spi-nor: aspeed: set the decoding size to at least 2MB for AST2600

         commit: 73ae97e3cabb580639f02f12a192324a53c4bebb

[0] https://patchwork.kernel.org/project/spi-devel-general/patch/20220325100849.2019209-9-clg at kaod.org/
[1] https://patchwork.kernel.org/project/spi-devel-general/patch/20220214094231.3753686-9-clg at kaod.org/
[2] https://lore.kernel.org/all/20220208190636.h6dubktkmuosvdxo at ti.com/

Regards,

C?dric.

#15 lore raw

Re: (subset) [PATCH v7 00/11] spi: spi-mem: Convert Aspeed SMC driver to spi-mem

Hi Cedric,

Quoted 41 lines · C?dric Le Goater

Pratyush,

Quoted 38 lines · Pratyush Yadav

Hi Cedric,

Quoted 34 lines · Mark Brown
Quoted 4 lines · C?dric Le Goater

This series adds a new SPI driver using the spi-mem interface for the Aspeed static memory controllers of the AST2600, AST2500 and AST2400 SoCs.

Quoted 8 lines

* AST2600 Firmware SPI Memory Controller (FMC)
* AST2600 SPI Flash Controller (SPI1 and SPI2)
* AST2500 Firmware SPI Memory Controller (FMC)
* AST2500 SPI Flash Controller (SPI1 and SPI2)
* AST2400 New Static Memory Controller (also referred as FMC)
* AST2400 SPI Flash Controller (SPI)

[...]

Applied to

https://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi.git for-next

Thanks!

[02/11] dt-bindings: spi: Convert the Aspeed SMC controllers device tree binding

         commit: ce9858ea499da025684a7a5f19823c2c3f14bdce
[03/11] spi: spi-mem: Convert Aspeed SMC driver to spi-mem
         commit: 9c63b846e6df43e5b3d31263f7db545f32deeda3
[04/11] spi: aspeed: Add support for direct mapping
         commit: 9da06d7bdec7dad8018c23b180e410ef2e7a4367
[05/11] spi: aspeed: Adjust direct mapping to device size
         commit: bb084f94e1bca4a5c4f689d7aa9b410220c1ed71
[06/11] spi: aspeed: Workaround AST2500 limitations
         commit: 5785eedee42c34cfec496199a80fa8ec9ddcf7fe
[07/11] spi: aspeed: Add support for the AST2400 SPI controller
         commit: 53526ab27d9c256504f267713aea60db7af18fb0
[08/11] spi: aspeed: Calibrate read timings
         commit: eeaec1ea05c0e0f08e04c6844f20cc24a2fcc0f4
Quoted 8 lines
Quoted 2 lines

I have repeatedly objected to this patch [0][1][2] and you have repeatedly decided to not address my objections.

That's a very harsh way of saying things. I did not decide anything or ignore your comments. I answered your questions and acknowledged that indeed the read training was done under the dirmap handler but this was not a concern today since we had all the required information from spimem.

I think the correct thing to do here would be to move this patch out of your main series. Since the old driver did not have this feature anyway, it is not necessary for the conversion to achieve feature parity. Then have a separate series to drive the conversation on how to best implement this contentious feature.

Quoted 7 lines

We waited _together_, 5 or 6 weeks, for more inputs on how to address the concerns you raised regarding the sustainability of this method.

Quoted 1 line

I won't spend any more time fighting it.

This is not a fight. I don't know why you interpret it that way.

I interpret it this way because I feel you have ignored my comments and have kept this patch in regardless. But maybe I _am_ being too harsh. So apologies if I have interpreted things wrong.

Quoted 5 lines

Now, since you object so explicitly, and since this patchset has not reached the Linux kernel yet, we should consider dropping it. I rather do that than push crap in mainline. But then, please, provide solutions and not only objections !

Yes please. I would like to have this patch dropped, and then reopen this discussion in a separate thread.

I have provided solutions before but maybe you missed them. Please see [3]. Specifically patches 1 and 3. These are still not very refined and might not be accepted in their current state (especially patch 3 since it does the calibration after registering the device with MTD), but they should be a good starting point for discussion.

And we might end up deciding to just use the dirmap_create() hook to do the calibration, but it is important to have this conversation separately and clearly instead of it being missed as part of a large patch series.

Quoted 9 lines
Quoted 4 lines

But I will say that you should not expect any
guarantees that SPI NOR or SPI NAND will not break your calibration in the future if they decide to move the dirmap_create() call around.

If that's the case one day, we have multiple solutions :

- stop doing the training
- move the training to the appropriate handler if it exists

This is my preferred option, [3] is a good starting point for proposing that handler.

Quoted 8 lines

- use a default value

Quoted 7 lines
Quoted 3 lines

[11/11] mtd: spi-nor: aspeed: set the decoding size to at least 2MB for AST2600

         commit: 73ae97e3cabb580639f02f12a192324a53c4bebb

[0] https://patchwork.kernel.org/project/spi-devel-general/patch/20220325100849.2019209-9-clg at kaod.org/
[1] https://patchwork.kernel.org/project/spi-devel-general/patch/20220214094231.3753686-9-clg at kaod.org/
[2] https://lore.kernel.org/all/20220208190636.h6dubktkmuosvdxo at ti.com/

[3] https://patchwork.ozlabs.org/project/linux-mtd/list/?series=233504&state=%2A&archive=both

Signature · 3 lines
Regards,
Pratyush Yadav
Texas Instruments Inc.
maintainer
#16 lore raw

Re: (subset) [PATCH v7 00/11] spi: spi-mem: Convert Aspeed SMC driver to spi-mem

In reply to Cédric Le Goater

Hi C?dric,

clg at kaod.org wrote on Tue, 17 May 2022 14:03:02 +0200:

Quoted 60 lines

Pratyush,

Quoted 38 lines · Pratyush Yadav

Hi Cedric,

Quoted 21 lines · Mark Brown
Quoted 4 lines · C?dric Le Goater

This series adds a new SPI driver using the spi-mem interface for the Aspeed static memory controllers of the AST2600, AST2500 and AST2400 SoCs.

Quoted 8 lines

* AST2600 Firmware SPI Memory Controller (FMC)
* AST2600 SPI Flash Controller (SPI1 and SPI2)
* AST2500 Firmware SPI Memory Controller (FMC)
* AST2500 SPI Flash Controller (SPI1 and SPI2)
* AST2400 New Static Memory Controller (also referred as FMC)
* AST2400 SPI Flash Controller (SPI)

[...]

Applied to

https://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi.git for-next

Thanks!

[02/11] dt-bindings: spi: Convert the Aspeed SMC controllers device tree binding

Quoted 13 lines
         commit: ce9858ea499da025684a7a5f19823c2c3f14bdce
[03/11] spi: spi-mem: Convert Aspeed SMC driver to spi-mem
         commit: 9c63b846e6df43e5b3d31263f7db545f32deeda3
[04/11] spi: aspeed: Add support for direct mapping
         commit: 9da06d7bdec7dad8018c23b180e410ef2e7a4367
[05/11] spi: aspeed: Adjust direct mapping to device size
         commit: bb084f94e1bca4a5c4f689d7aa9b410220c1ed71
[06/11] spi: aspeed: Workaround AST2500 limitations
         commit: 5785eedee42c34cfec496199a80fa8ec9ddcf7fe
[07/11] spi: aspeed: Add support for the AST2400 SPI controller
         commit: 53526ab27d9c256504f267713aea60db7af18fb0
[08/11] spi: aspeed: Calibrate read timings
         commit: eeaec1ea05c0e0f08e04c6844f20cc24a2fcc0f4  
Quoted 2 lines

I have repeatedly objected to this patch [0][1][2] and you have repeatedly decided to not address my objections.

That's a very harsh way of saying things. I did not decide anything or ignore your comments. I answered your questions and acknowledged that indeed the read training was done under the dirmap handler but this was not a concern today since we had all the required information from spimem.

We waited _together_, 5 or 6 weeks, for more inputs on how to address the concerns you raised regarding the sustainability of this method.

Quoted 1 line

I won't spend any more time fighting it.

This is not a fight. I don't know why you interpret it that way.

Now, since you object so explicitly, and since this patchset has not reached the Linux kernel yet, we should consider dropping it. I rather do that than push crap in mainline. But then, please, provide solutions and not only objections !

Quoted 7 lines
Quoted 3 lines

But I will say that you should not expect any
guarantees that SPI NOR or SPI NAND will not break your calibration in the future if they decide to move the dirmap_create() call around.

If that's the case one day, we have multiple solutions :

- stop doing the training

Not really an option, isn't it?

Quoted 1 line

- move the training to the appropriate handler if it exists

Sorry for jumping in so late, but why not using this solution already if there are concerns about the current location? It does not look like a complex solution to implement, but I might be overlooking something.

This can also be done for the next kernel release.

Quoted 5 lines

- use a default value

Quoted 4 lines
Quoted 3 lines

[11/11] mtd: spi-nor: aspeed: set the decoding size to at least 2MB for AST2600

         commit: 73ae97e3cabb580639f02f12a192324a53c4bebb
Quoted 7 lines
Quoted 3 lines

[0] https://patchwork.kernel.org/project/spi-devel-general/patch/20220325100849.2019209-9-clg at kaod.org/
[1] https://patchwork.kernel.org/project/spi-devel-general/patch/20220214094231.3753686-9-clg at kaod.org/
[2] https://lore.kernel.org/all/20220208190636.h6dubktkmuosvdxo at ti.com/

Regards,

C?dric.

Thanks,
Miqu?l

author maintainer
#17 lore raw

Re: (subset) [PATCH v7 00/11] spi: spi-mem: Convert Aspeed SMC driver to spi-mem

Hello Miquel,

Quoted 68 lines · Miquel Raynal

Hi C?dric,

clg at kaod.org wrote on Tue, 17 May 2022 14:03:02 +0200:

Quoted 64 lines

Pratyush,

Quoted 38 lines · Pratyush Yadav

Hi Cedric,

Quoted 21 lines · Mark Brown
Quoted 4 lines · C?dric Le Goater

This series adds a new SPI driver using the spi-mem interface for the Aspeed static memory controllers of the AST2600, AST2500 and AST2400 SoCs.

Quoted 8 lines

* AST2600 Firmware SPI Memory Controller (FMC)
* AST2600 SPI Flash Controller (SPI1 and SPI2)
* AST2500 Firmware SPI Memory Controller (FMC)
* AST2500 SPI Flash Controller (SPI1 and SPI2)
* AST2400 New Static Memory Controller (also referred as FMC)
* AST2400 SPI Flash Controller (SPI)

[...]

Applied to

https://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi.git for-next

Thanks!

[02/11] dt-bindings: spi: Convert the Aspeed SMC controllers device tree binding

Quoted 13 lines
          commit: ce9858ea499da025684a7a5f19823c2c3f14bdce
[03/11] spi: spi-mem: Convert Aspeed SMC driver to spi-mem
          commit: 9c63b846e6df43e5b3d31263f7db545f32deeda3
[04/11] spi: aspeed: Add support for direct mapping
          commit: 9da06d7bdec7dad8018c23b180e410ef2e7a4367
[05/11] spi: aspeed: Adjust direct mapping to device size
          commit: bb084f94e1bca4a5c4f689d7aa9b410220c1ed71
[06/11] spi: aspeed: Workaround AST2500 limitations
          commit: 5785eedee42c34cfec496199a80fa8ec9ddcf7fe
[07/11] spi: aspeed: Add support for the AST2400 SPI controller
          commit: 53526ab27d9c256504f267713aea60db7af18fb0
[08/11] spi: aspeed: Calibrate read timings
          commit: eeaec1ea05c0e0f08e04c6844f20cc24a2fcc0f4
Quoted 2 lines

I have repeatedly objected to this patch [0][1][2] and you have repeatedly decided to not address my objections.

That's a very harsh way of saying things. I did not decide anything or ignore your comments. I answered your questions and acknowledged that indeed the read training was done under the dirmap handler but this was not a concern today since we had all the required information from spimem.

We waited _together_, 5 or 6 weeks, for more inputs on how to address the concerns you raised regarding the sustainability of this method.

Quoted 1 line

I won't spend any more time fighting it.

This is not a fight. I don't know why you interpret it that way.

Now, since you object so explicitly, and since this patchset has not reached the Linux kernel yet, we should consider dropping it. I rather do that than push crap in mainline. But then, please, provide solutions and not only objections !

Quoted 3 lines

But I will say that you should not expect any
guarantees that SPI NOR or SPI NAND will not break your calibration in the future if they decide to move the dirmap_create() call around.

Quoted 5 lines
Quoted 3 lines

If that's the case one day, we have multiple solutions :

- stop doing the training

Not really an option, isn't it?

well, I would really rather keep the training. Else the device would be crippled and run at low speed. 1/16 HCLK is the default :/

Quoted 6 lines
Quoted 2 lines

- move the training to the appropriate handler if it exists

Sorry for jumping in so late, but why not using this solution already if there are concerns about the current location? It does not look like a complex solution to implement, but I might be overlooking something.

I think we need some more feedback on this proposal [0].

If that's not acceptable, then let's discuss on dirmap_create() and see if it would be acceptable to do calibration under this hook. Which is what this patchset is proposing since the spimem interface offers all the required information to do so.

However, Pratyush would prefer a stable interface, hence [0].

That said, if one day we merge a do_calibration() hook like proposed in [0], we can change the current Aspeed patchset to use it. That's why I didn't see the lack of specific calibration hook as a blocker.

Quoted 1 line

This can also be done for the next kernel release.

I was hoping this one. We would need a comment on the use of the dirmap_create() hook to clarify.

Thanks !

C?dric

[0] https://patchwork.ozlabs.org/project/linux-mtd/list/?series=233504&state=%2A&archive=both

maintainer
#18 lore raw

Re: (subset) [PATCH v7 00/11] spi: spi-mem: Convert Aspeed SMC driver to spi-mem

In reply to Pratyush Yadav
Quoted 11 lines · Pratyush Yadav
Quoted 4 lines · Mark Brown

On Mon, 9 May 2022 19:56:05 +0200, C?dric Le Goater wrote:

[08/11] spi: aspeed: Calibrate read timings

        commit: eeaec1ea05c0e0f08e04c6844f20cc24a2fcc0f4

I have repeatedly objected to this patch [0][1][2] and you have repeatedly decided to not address my objections. I won't spend any more time fighting it. But I will say that you should not expect any guarantees that SPI NOR or SPI NAND will not break your calibration in the future if they decide to move the dirmap_create() call around.

You == Cedric here. Sorry, I thought that had been addressed as there
hadn't been anything I'd noticed on the thread for a few versions. It
does look like there was agreement that there was a problem too :/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: not available
URL: <http://lists.ozlabs.org/pipermail/linux-aspeed/attachments/20220517/deb16e3a/attachment.sig>