Add MMC/SD support for S700/S900
Subsystems (5)
v3 cover Add MMC/SD support for S700/S900
From: Amit Singh Tomar <[email protected]>
This series(v3) addresses review comments provided by Andre and Jaehoon. Earlier in MMC driver, we were *not* checking the command response errors(for instance, the CRC error) based on STATU register. This is now addressed in patch 5/6.
Also, MMC driver(pacth 5/6) now has replaced the switch/case with if/else to correlate MMC_RSP_Rx bitmask values with register bits as suggested by Andre.
Apart from it, there are changes in patch 2/6 to fix weird assignments to div variable.
----------------------------------------------------------------------------------
Series(v2) has few important updates, while loading large files we found that MMC framework puts "0x1fffe00" into DMA Frame Length(DMA_FRAME_LEN 0x000C) but register itself is limited to 24 bits and hence it was failing. It is due to the wrong Block count(cfg->b_max) used in driver, that should be just 512. This is now fixed in patch 5/6.
Apart from it, as Andre pointed that we might be just overclocking the MMC/SD clock, and to confirm this we run following test:
$ md5sum clang 349eac46cbbe28f8e44da2dce07fa7b7 clang
U-Boot => ext4load mmc 0:2 0x0 clang
503316480 bytes read in 19516 ms (24.6 MiB/s)
U-Boot => md5sum 0x0 0x1e000000
md5 for 00000000 ... 1dffffff ==> d793bb51c4a1cf83c96d1980927461ff
Even though file gets loaded but md5sum doesn't match. This is now fixed in patch 2/6
U-Boot => ext4load mmc 0:2 0x0 clang
503316480 bytes read in 41524 ms (11.6 MiB/s)
U-Boot => md5sum 0x0 0x1e000000
md5 for 00000000 ... 1dffffff ==> 349eac46cbbe28f8e44da2dce07fa7b7
-----------------------------------------------------------------------------
At the moment on S700 based platforms, only way to load/boot the Kernel is from Ethernet, and with these patches one can now load/boot the Kernel from uSD card.
Patches(1/6 and 2/6) adds changes needed for MMC/SD clock. It introduces set/get callback routine and get/set MMC/SD clock rate based on device id.
Patch 4/6 adds MMC/SD node in U-boot specific dtsi file, which is used by MMC/SD driver to read controller base address later on.
Patch 5/6 adds driver for MMC/SD controller present on S700 SoC, and its based on Mainline Linux driver and DMA related bits is picked and simpilified from vendor source.
Final patch 6/6 enables the driver support along with MMC commands in Cubieboard7 config file.
Also, while at it just took the opportunity to synchronize the S700 SoC DT with Linux in patch 3/6.
This patch-set is tested on Cubieboard7-lite board with following results:
U-Boot 2021.01-rc1-04434-g6589149-dirty (Dec 13 2020 - 13:51:07 +0530) cubieboard7
DRAM: 1 GiB PSCI: v0.2 MMC: mmc at e0210000: 0 In: serial at e0126000 Out: serial at e0126000 Err: serial at e0126000 Net: eth0: ethernet at e0220000 Hit any key to stop autoboot: 0
U-Boot =>
U-Boot =>
U-Boot =>
U-Boot => mmc info
Device: mmc at e0210000
Manufacturer ID: 3
OEM: 5344
Name: SC16G
Bus Speed: 50000000
Mode: SD High Speed (50MHz)
Rd Block Len: 512
SD version 3.0
High Capacity: Yes
Capacity: 14.8 GiB
Bus Width: 4-bit
Erase Group Size: 512 Bytes
U-Boot => setenv bootargs console=ttyOWL3,115200n8 earlycon=owl,0xe0126000 init=/sbin/init root=/dev/mmcblk0p2 rw rootwait U-Boot => setenv kernel_addr_r 0x80000;setenv fdt_addr_r 0x10000000;
U-Boot => fatload mmc 0:1 ${kernel_addr_r} image ;fatload mmc 0:1 ${fdt_addr_r} s700-cubieboard7.dtb 27480576 bytes read in 1041 ms (25.2 MiB/s)
7056 bytes read in 2 ms (3.4 MiB/s)
U-Boot => booti $kernel_addr_r - $fdt_addr_r
## Flattened Device Tree blob at 10000000
Booting using the fdt blob at 0x10000000
Loading Device Tree to 000000003df56000, end 000000003df5ab8f ... OK
Starting kernel ...
[ 0.000000] Booting Linux on physical CPU 0x0000000000 [0x410fd034] [ 0.000000] Linux version 5.7.0-rc6-00022-g99f1c330923b-dirty (amit at amit-ThinkPad-X230) (gcc version 7.3.1 20180425 [linaro-7.3-2018.05 revision d29120a424ecfbc167ef90065c0eeb7f91977701] (Linaro GCC 7.3-2018.05), GNU ld (GNU Binutils for Ubuntu) 2.26.1) #474 SMP PREEMPT Thu Aug 13 15:27:10 IST 2020 [ 0.000000] Machine model: CubieBoard7 [ 0.000000] earlycon: owl0 at MMIO 0x00000000e0126000 (options '') [ 0.000000] printk: bootconsole [owl0] enabled [ 0.000000] efi: UEFI not found. [ 0.000000] cma: Reserved 32 MiB at 0x000000003e000000
Amit Singh Tomar (6):
clk: actions: Introduce dummy get/set_rate callbacks
clk: actions: Add SD/MMC clocks
ARM: dts: sync Actions Semi S700 DT from Linux 5.10-rc7
ARM: dts: s700: add MMC/SD controller node
mmc: actions: add MMC driver for Actions OWL S700/S900
configs: Enable mmc support
arch/arm/dts/s700-u-boot.dtsi | 10 + arch/arm/dts/s700.dtsi | 17 +- configs/cubieboard7_defconfig | 3 + drivers/clk/owl/clk_owl.c | 99 ++++++ drivers/mmc/Kconfig | 7 + drivers/mmc/Makefile | 1 + drivers/mmc/owl_mmc.c | 401 +++++++++++++++++++++++++ include/dt-bindings/power/owl-s700-powergate.h | 19 ++ 8 files changed, 556 insertions(+), 1 deletion(-) create mode 100644 drivers/mmc/owl_mmc.c create mode 100644 include/dt-bindings/power/owl-s700-powergate.h
Signature · 1 line
2.7.4
PATCH v3 1/6 clk: actions: Introduce dummy get/set_rate callbacks
From: Amit Singh Tomar <[email protected]>
This commit introduces get/set_rate callbacks, these are dummy at the moment, and can be used to get/set clock for various devices based on the clk id.
---
Changes since v2:
* No changes.
Changes since previous version:
* Removed premature initialization to avoid compiler warnings.
--- drivers/clk/owl/clk_owl.c | 26 ++++++++++++++++++++++++++ 1 file changed, 26 insertions(+)
drivers/clk/owl/clk_owl.c +26 -0
@@ -128,6 +128,30 @@ int owl_clk_disable(struct clk *clk)return 0 ; } +static ulong owl_clk_get_rate (struct clk *clk) +{ +ulong rate; + +switch (clk->id) { +default : +return -ENOENT ; + } + +return rate; +} + +static ulong owl_clk_set_rate (struct clk *clk,ulong rate) +{ +ulong new_rate; + +switch (clk->id) { +default : +return -ENOENT ; + } + +return new_rate; +} +static int owl_clk_probe (struct udevice *dev) {struct owl_clk_priv *priv =dev_get_priv (dev);
@@ -145,6 +169,8 @@ static int owl_clk_probe(struct udevice *dev)static const struct clk_ops owl_clk_ops = { .enable = owl_clk_enable, .disable = owl_clk_disable, + .get_rate = owl_clk_get_rate, + .set_rate = owl_clk_set_rate, };static const struct udevice_id owl_clk_ids[] = {
Signature · 1 line
2.7.4
PATCH v3 2/6 clk: actions: Add SD/MMC clocks
In reply to Amit Singh TomarFrom: Amit Singh Tomar <[email protected]>
This commit adds SD/MMC clocks, and provides .set/get_rate callbacks for SD/MMC device present on Actions OWL S700 SoCs.
---
Changes since v2:
* Fixed the not weird div assignment.
* Moved the clock bit for SD from header file to driver file.
* Removed "< 0" part while comparing unsigned.
Changes since previous version:
* Removed rate *= 2 as this just overclocks.
* Separated the divide by 128 bit from divider value.
* Provided the separate routine to get sd parent rate based on bit 9.
* Removed unnecessary initialization.
--- drivers/clk/owl/clk_owl.c | 73 +++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 73 insertions(+)
drivers/clk/owl/clk_owl.c +73 -0
@@ -20,6 +20,8 @@#include <linux/bitops.h> #include <linux/delay.h> +#define CMU_DEVCLKEN0_SD0 BIT(22) +void owl_clk_init (struct owl_clk_priv *priv) {u32 bus_clk =0 , core_pll, dev_pll;
@@ -92,6 +94,9 @@ int owl_clk_enable(struct clk *clk)setbits_le32 (priv->base +CMU_DEVCLKEN1 ,CMU_DEVCLKEN1_ETH );setbits_le32 (priv->base +CMU_ETHERNETPLL ,5 );break ; +case CLK_SD0 : +setbits_le32 (priv->base +CMU_DEVCLKEN0 ,CMU_DEVCLKEN0_SD0 ); +break ;default :return -EINVAL ; }
@@ -121,6 +126,9 @@ int owl_clk_disable(struct clk *clk)case CLK_ETHERNET :clrbits_le32 (priv->base +CMU_DEVCLKEN1 ,CMU_DEVCLKEN1_ETH );break ; +case CLK_SD0 : +clrbits_le32 (priv->base +CMU_DEVCLKEN0 ,CMU_DEVCLKEN0_SD0 ); +break ;default :return -EINVAL ; }
@@ -128,11 +136,72 @@ int owl_clk_disable(struct clk *clk)return 0 ; } +static ulong get_sd_parent_rate (struct owl_clk_priv *priv,u32 dev_index) +{ +ulong rate; +u32 reg; + + reg =readl (priv->base + (CMU_SD0CLK + dev_index *0x4 )); +/* Clock output of DEV/NAND_PLL +* Range: 48M ~ 756M +* Frequency= PLLCLK * 6 +*/ +if (reg &0x200 ) + rate =readl (priv->base +CMU_NANDPLL ) &0x7f ; +else + rate =readl (priv->base +CMU_DEVPLL ) &0x7f ; + + rate *=6000000 ; + +return rate; +} + +static ulong owl_get_sd_clk_rate (struct owl_clk_priv *priv,int sd_index) +{ +uint div, val; +ulong parent_rate =get_sd_parent_rate (priv, sd_index); + + val =readl (priv->base + (CMU_SD0CLK + sd_index *0x4 )); + div = (val &0x1f ) +1 ; + +return (parent_rate / div); +} + +static ulong owl_set_sd_clk_rate (struct owl_clk_priv *priv,ulong rate, +int sd_index) +{ +uint div, val; +ulong parent_rate =get_sd_parent_rate (priv, sd_index); + +if (rate ==0 ) +return rate; + + div = (parent_rate / rate); + + val =readl (priv->base + (CMU_SD0CLK + sd_index *0x4 )); +/* Bits 4..0 is used to program div value and bit 8 to enable +* divide by 128 circuit +*/ + val &= ~0x11f ; +if (div >=128 ) { + div = div /128 ; + val |=0x100 ;/* enable divide by 128 circuit */ + } + val |= ((div -1 ) &0x1f ); +writel (val, priv->base + (CMU_SD0CLK + sd_index *0x4 )); + +return owl_get_sd_clk_rate (priv,0 ); +} +static ulong owl_clk_get_rate (struct clk *clk) { +struct owl_clk_priv *priv =dev_get_priv (clk->dev);ulong rate;switch (clk->id) { +case CLK_SD0 : + rate =owl_get_sd_clk_rate (priv,0 ); +break ;default :return -ENOENT ; }
@@ -142,9 +211,13 @@ static ulong owl_clk_get_rate(struct clk *clk)static ulong owl_clk_set_rate (struct clk *clk,ulong rate) { +struct owl_clk_priv *priv =dev_get_priv (clk->dev);ulong new_rate;switch (clk->id) { +case CLK_SD0 : + new_rate =owl_set_sd_clk_rate (priv, rate,0 ); +break ;default :return -ENOENT ; }
Signature · 1 line
2.7.4
PATCH v3 3/6 ARM: dts: sync Actions Semi S700 DT from Linux 5.10-rc7
In reply to Amit Singh TomarFrom: Amit Singh Tomar <[email protected]>
This Synchronizes the Actions Semi S700 SoC DT changes from commit "0477e9288185" ("Linux 5.10-rc7").
---
Changes since previous versions
* No change.
--- arch/arm/dts/s700.dtsi | 17 ++++++++++++++++- include/dt-bindings/power/owl-s700-powergate.h | 19 +++++++++++++++++++ 2 files changed, 35 insertions(+), 1 deletion(-) create mode 100644 include/dt-bindings/power/owl-s700-powergate.h
arch/arm/dts/s700.dtsi +16 -1
@@ -5,6 +5,7 @@ #include <dt-bindings/clock/actions,s700-cmu.h> #include <dt-bindings/interrupt-controller/arm-gic.h> +#include <dt-bindings/power/owl-s700-powergate.h> #include <dt-bindings/reset/actions,s700-reset.h> / {
@@ -231,7 +232,7 @@ pinctrl: pinctrl at e01b0000 { compatible = "actions,s700-pinctrl"; - reg = <0x0 0xe01b0000 0x0 0x1000>; + reg = <0x0 0xe01b0000 0x0 0x100>; clocks = <&cmu CLK_GPIO>; gpio-controller; gpio-ranges = <&pinctrl 0 0 136>;
@@ -244,5 +245,19 @@ <GIC_SPI 39 IRQ_TYPE_LEVEL_HIGH>, <GIC_SPI 40 IRQ_TYPE_LEVEL_HIGH>; }; + + dma: dma-controller at e0230000 { + compatible = "actions,s700-dma"; + reg = <0x0 0xe0230000 0x0 0x1000>; + interrupts = <GIC_SPI 57 IRQ_TYPE_LEVEL_HIGH>, + <GIC_SPI 58 IRQ_TYPE_LEVEL_HIGH>, + <GIC_SPI 59 IRQ_TYPE_LEVEL_HIGH>, + <GIC_SPI 60 IRQ_TYPE_LEVEL_HIGH>; + #dma-cells = <1>; + dma-channels = <10>; + dma-requests = <44>; + clocks = <&cmu CLK_DMAC>; + power-domains = <&sps S700_PD_DMA>; + }; }; };
include/dt-bindings/power/owl-s700-powergate.h new file +19 -0
@@ -0,0 +1,19 @@ +// SPDX-License-Identifier: (GPL-2.0+ OR MIT) +/* +* Actions Semi S700 SPS +* +* Copyright (c) 2017 Andreas F?rber +*/ +#ifndef DT_BINDINGS_POWER_OWL_S700_POWERGATE_H +#define DT_BINDINGS_POWER_OWL_S700_POWERGATE_H + +#define S700_PD_VDE 0 +#define S700_PD_VCE_SI 1 +#define S700_PD_USB2_1 2 +#define S700_PD_HDE 3 +#define S700_PD_DMA 4 +#define S700_PD_DS 5 +#define S700_PD_USB3 6 +#define S700_PD_USB2_0 7 + +#endif
Signature · 1 line
2.7.4
PATCH v3 4/6 ARM: dts: s700: add MMC/SD controller node
In reply to Amit Singh TomarFrom: Amit Singh Tomar <[email protected]>
This patch adds node for mmc/sd controller found on Action Semi OWL S700 SoC.
Since, upstream Linux binding has not been merged for S700 MMC/SD controller, Changes are put in u-boot specific dtsi file.
---
Changes since previous versions
* No change.
--- arch/arm/dts/s700-u-boot.dtsi | 10 ++++++++++ 1 file changed, 10 insertions(+)
arch/arm/dts/s700-u-boot.dtsi +10 -0
@@ -19,6 +19,16 @@ status = "okay"; }; + mmc0: mmc at e0210000 { + compatible = "actions,s700-mmc", "actions,owl-mmc"; + reg = <0x0 0xe0210000 0x0 0x4000>; + interrupts = <GIC_SPI 42 IRQ_TYPE_LEVEL_HIGH>; + clocks = <&cmu CLK_SD0>; + dmas = <&dma 2>; + dma-names = "mmc"; + bus-width = <4>; + status = "okay"; + }; }; };
Signature · 1 line
2.7.4
PATCH v3 5/6 mmc: actions: add MMC driver for Actions OWL S700/S900
In reply to Amit Singh TomarFrom: Amit Singh Tomar <[email protected]>
This commit adds support for MMC controllers found on Actions OWL SoC platform(S700/S900).
---
Changes since v2:
* Progammed read/write delays as suggested by
Jaehoon, its more readable now.
* Updated commit message and source file to reflect
it supports S900 as well.
* Updated SD_DMA_CHANNEL macro to include channel offset.
* Checked the command response Error based on SDC STATU register bits.
* Replaced switch/case with if/else to correlate with MMC_RSP_Rx bitmask values, as suggested by Andre.
* Introduced clk_disable based on mmc->clk_disable flag.
* Guarded the clk_set_rate under if statment to avoid reprogramming the rate.
* Fetched the DRQ number from using args.args[0].
Changes since previous version
* Corrected block count to 512.
* Changed the command timeout value to 30ms.
* Used readl_poll_timeout.
* Read DMA parameters from DT instead of hardcoding it.
* Reduced number of arguments passed to own_dma_cofig.
* Removed debug leftover.
* Used mmc_of_parse().
--- drivers/mmc/Kconfig | 7 + drivers/mmc/Makefile | 1 + drivers/mmc/owl_mmc.c | 401 ++++++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 409 insertions(+) create mode 100644 drivers/mmc/owl_mmc.c
drivers/mmc/Kconfig +7 -0
@@ -289,6 +289,13 @@ config MMC_MXC If unsure, say N. +config MMC_OWL + bool "Actions OWL Multimedia Card Interface support" + depends on ARCH_OWL && DM_MMC && BLK + help + This selects the OWL SD/MMC host controller found on board + based on Actions S700/S900 SoC. + config MMC_MXS bool "Freescale MXS Multimedia Card Interface support" depends on MX23 || MX28 || MX6 || MX7
drivers/mmc/Makefile +1 -0
@@ -38,6 +38,7 @@ obj-$(CONFIG_MMC_OMAP_HS) += omap_hsmmc.o obj-$(CONFIG_MMC_MXC) += mxcmmc.o obj-$(CONFIG_MMC_MXS) += mxsmmc.o obj-$(CONFIG_MMC_OCTEONTX) += octeontx_hsmmc.o +obj-$(CONFIG_MMC_OWL) += owl_mmc.o obj-$(CONFIG_MMC_PCI) += pci_mmc.o obj-$(CONFIG_PXA_MMC_GENERIC) += pxa_mmc_gen.o obj-$(CONFIG_$(SPL_TPL_)SUPPORT_EMMC_RPMB) += rpmb.o
drivers/mmc/owl_mmc.c new file +401 -0
401 lines not on this page · raw message
Signature · 1 line
2.7.4
PATCH v3 6/6 configs: Enable mmc support
In reply to Amit Singh TomarFrom: Amit Singh Tomar <[email protected]>
This commits enables mmc on the Actions Cubieboard7 board.
---
Changes since previous versions
* No change.
--- configs/cubieboard7_defconfig | 3 +++ 1 file changed, 3 insertions(+)
configs/cubieboard7_defconfig +3 -0
@@ -14,3 +14,6 @@ CONFIG_PHY_REALTEK=y CONFIG_RTL8201F_PHY_S700_RMII_TIMINGS=y CONFIG_ETH_DESIGNWARE=y CONFIG_ETH_DESIGNWARE_S700=y +CONFIG_DM_MMC=y +CONFIG_MMC_OWL=y +CONFIG_CMD_MMC=y
Signature · 1 line
2.7.4
PATCH v3 5/6 mmc: actions: add MMC driver for Actions OWL S700/S900
In reply to Amit Singh TomarQuoted 1 line · Amit Singh Tomar
From: Amit Singh Tomar <[email protected]>
Hi,
Quoted 149 lines
This commit adds support for MMC controllers found on Actions OWL SoC platform(S700/S900).
Signed-off-by Amit Singh Tomar <[email protected]>---
Changes since v2:
* Progammed read/write delays as suggested by
Jaehoon, its more readable now.
* Updated commit message and source file to reflect
it supports S900 as well.
* UpdatedSD_DMA_CHANNELmacro to include channel offset.
* Checked the command response Error based on SDC STATU register bits.
* Replaced switch/case with if/else to correlate withMMC_RSP_Rxbitmask values, as suggested by Andre.
* Introducedclk_disablebased onmmc->clk_disableflag.
* Guarded theclk_set_rateunder if statment to avoid reprogramming the rate.
* Fetched the DRQ number from using args.args[0].
Changes since previous version
* Corrected block count to 512.
* Changed the command timeout value to 30ms.
* Usedreadl_poll_timeout.
* Read DMA parameters from DT instead of hardcoding it.
* Reduced number of arguments passed toown_dma_cofig.
* Removed debug leftover.
* Usedmmc_of_parse().--- drivers/mmc/Kconfig | 7 + drivers/mmc/Makefile | 1 + drivers/mmc/owl_mmc.c | 401 ++++++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 409 insertions(+) create mode 100644 drivers/mmc/owl_mmc.cdrivers/mmc/Kconfig +7
-0@@ -289,6 +289,13 @@ config MMC_MXC If unsure, say N. +config MMC_OWL + bool "Actions OWL Multimedia Card Interface support" + depends on ARCH_OWL && DM_MMC && BLK + help + This selects the OWL SD/MMC host controller found on board + based on Actions S700/S900 SoC. + config MMC_MXS bool "Freescale MXS Multimedia Card Interface support" depends on MX23 || MX28 || MX6 || MX7drivers/mmc/Makefile +1
-0@@ -38,6 +38,7 @@ obj-$(CONFIG_MMC_OMAP_HS) += omap_hsmmc.o obj-$(CONFIG_MMC_MXC) += mxcmmc.o obj-$(CONFIG_MMC_MXS) += mxsmmc.o obj-$(CONFIG_MMC_OCTEONTX) += octeontx_hsmmc.o +obj-$(CONFIG_MMC_OWL) += owl_mmc.o obj-$(CONFIG_MMC_PCI) += pci_mmc.o obj-$(CONFIG_PXA_MMC_GENERIC) += pxa_mmc_gen.o obj-$(CONFIG_$(SPL_TPL_)SUPPORT_EMMC_RPMB) += rpmb.odrivers/mmc/owl_mmc.c new file +79
-0@@ -0,0 +1,401 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* +* Copyright (C) 2020 Amit Singh Tomar <[email protected]> +* +* Driver for SD/MMC controller present on Actions Semi S700/S900 SoC, based +* on Linux Driver "drivers/mmc/host/owl-mmc.c". +* +* Though, there is a bit (BSEL, BUS or DMA Special Channel Selection) that +* controls the data transfer from SDx_DAT register either using CPU AHB Bus +* or DMA channel, but seems like, it only works correctly using external DMA +* channel, and those special bits used in this driver is picked from vendor +* source exclusively for MMC/SD. +*/ +#include <common.h> +#include <clk.h> +#include <cpu_func.h> +#include <dm.h> +#include <errno.h> +#include <log.h> +#include <mmc.h> +#include <asm/io.h> +#include <linux/bitops.h> +#include <linux/delay.h> +#include <linux/err.h> +#include <linux/iopoll.h> + +/* +* SDC registers +*/ +#define OWL_REG_SD_EN 0x0000 +#define OWL_REG_SD_CTL 0x0004 +#define OWL_REG_SD_STATE 0x0008 +#define OWL_REG_SD_CMD 0x000c +#define OWL_REG_SD_ARG 0x0010 +#define OWL_REG_SD_RSPBUF0 0x0014 +#define OWL_REG_SD_RSPBUF1 0x0018 +#define OWL_REG_SD_RSPBUF2 0x001c +#define OWL_REG_SD_RSPBUF3 0x0020 +#define OWL_REG_SD_RSPBUF4 0x0024 +#define OWL_REG_SD_DAT 0x0028 +#define OWL_REG_SD_BLK_SIZE 0x002c +#define OWL_REG_SD_BLK_NUM 0x0030 +#define OWL_REG_SD_BUF_SIZE 0x0034 + +/* SD_EN Bits */ +#define OWL_SD_EN_RANE BIT(31) +#define OWL_SD_EN_RESE BIT(10) +#define OWL_SD_ENABLE BIT(7) +#define OWL_SD_EN_BSEL BIT(6) +#define OWL_SD_EN_DATAWID (x) (((x) & 0x3) << 0) +#define OWL_SD_EN_DATAWID_MASK 0x03 + +/* SD_CTL Bits */ +#define OWL_SD_CTL_TOUTEN BIT(31) +#define OWL_SD_CTL_DELAY_MSK GENMASK(23, 16) +#define OWL_SD_CTL_RDELAY (x) (((x) & 0xf) << 20) +#define OWL_SD_CTL_WDELAY (x) (((x) & 0xf) << 16) +#define OWL_SD_CTL_TS BIT(7) +#define OWL_SD_CTL_LBE BIT(6) +#define OWL_SD_CTL_TM (x) (((x) & 0xf) << 0) + +#define OWL_SD_DELAY_LOW_CLK 0x0f +#define OWL_SD_DELAY_MID_CLK 0x0a +#define OWL_SD_RDELAY_HIGH 0x08 +#define OWL_SD_WDELAY_HIGH 0x09 + +/* SD_STATE Bits */ +#define OWL_SD_STATE_DAT0S BIT(7) +#define OWL_SD_STATE_CLNR BIT(4) +#define OWL_SD_STATE_CRC7ER BIT(0) + +#define OWL_MMC_OCR (MMC_VDD_32_33 | MMC_VDD_33_34 | \ + MMC_VDD_165_195) + +#define DATA_TRANSFER_TIMEOUT 30000 + +/* +* Simple DMA transfer operations defines for MMC/SD card +*/
drivers/mmc/owl_mmc.c line 89
Quoted 10 lines
drivers/mmc/owl_mmc.c new file +10
-0+#define SD_DMA_CHANNEL (base, channel) ((base) + 0x100 + 0x100 * (channel)) + +#define DMA_MODE 0x0000 +#define DMA_SOURCE 0x0004 +#define DMA_DESTINATION 0x0008 +#define DMA_FRAME_LEN 0x000C +#define DMA_FRAME_CNT 0x0010 +#define DMA_START 0x0024 + +/* DMAx_MODE */
There are still spaces in the below section.
Quoted 18 lines
drivers/mmc/owl_mmc.c new file +18
-0+#define DMA_MODE_ST (x) (((x) & 0x3) << 8) +#define DMA_MODE_ST_DEV DMA_MODE_ST(0) +#define DMA_MODE_DT (x) (((x) & 0x3) << 10) +#define DMA_MODE_DT_DCU DMA_MODE_DT(2) +#define DMA_MODE_SAM (x) (((x) & 0x3) << 16) +#define DMA_MODE_SAM_CONST DMA_MODE_SAM(0) +#define DMA_MODE_DAM (x) (((x) & 0x3) << 18) +#define DMA_MODE_DAM_INC DMA_MODE_DAM(1) + +struct owl_mmc_plat { +struct mmc_config cfg; +struct mmc mmc; +}; + +struct owl_mmc_priv { +void *reg_base; +void *dma_channel; +struct clk clk;
drivers/mmc/owl_mmc.c line 117
Quoted 10 lines
drivers/mmc/owl_mmc.c new file +10
-0+unsigned int clock;/* Current clock */ +unsigned int dma_drq;/* Trigger Source */ +}; + +static void owl_dma_config (struct owl_mmc_priv *priv,unsigned int src, +unsigned int dst,unsigned int len) +{ +unsigned int mode = priv->dma_drq; + +/* Set Source and Destination adderess mode */
address
Quoted 9 lines
drivers/mmc/owl_mmc.c new file +9
-0+ mode |= (DMA_MODE_ST_DEV |DMA_MODE_SAM_CONST |DMA_MODE_DT_DCU | +DMA_MODE_DAM_INC ); + +writel (mode,SD_DMA_CHANNEL (priv->dma_channel,0 ) +DMA_MODE ); +writel (src,SD_DMA_CHANNEL (priv->dma_channel,0 ) +DMA_SOURCE ); +writel (dst,SD_DMA_CHANNEL (priv->dma_channel,0 ) +DMA_DESTINATION ); +writel (len,SD_DMA_CHANNEL (priv->dma_channel,0 ) +DMA_FRAME_LEN ); +writel (0x1 ,SD_DMA_CHANNEL (priv->dma_channel,0 ) +DMA_FRAME_CNT ); +}
drivers/mmc/owl_mmc.c line 136
Quoted 10 lines
drivers/mmc/owl_mmc.c new file +10
-0+ +static void owl_mmc_prepare_data (struct owl_mmc_priv *priv, +struct mmc_data *data) +{ +unsigned int reg, total; +u32 buf =0 ; + + reg =readl (priv->reg_base +OWL_REG_SD_EN ); + reg |=OWL_SD_EN_BSEL ; +writel (reg, priv->reg_base +OWL_REG_SD_EN );
Can't you use setbits_le32() here? The same to all the other occasions where you just set or clear some constant bits in a register.
drivers/mmc/owl_mmc.c line 143
Quoted 7 lines
drivers/mmc/owl_mmc.c new file +7
-0+ +writel (data->blocks, priv->reg_base +OWL_REG_SD_BLK_NUM ); +writel (data->blocksize, priv->reg_base +OWL_REG_SD_BLK_SIZE ); + total = data->blocksize * data->blocks; + +if (data->blocksize <512 ) +writel (total, priv->reg_base +OWL_REG_SD_BUF_SIZE );
Is that correct? I would think you either check for total being smaller than 512, or you write data->blocksize into the register.
Quoted 27 lines
drivers/mmc/owl_mmc.c new file +27
-0+else +writel (512 , priv->reg_base +OWL_REG_SD_BUF_SIZE ); + +/* DMA STOP */ +writel (0x0 ,SD_DMA_CHANNEL (priv->dma_channel,0 ) +DMA_START ); + +if (data) { +if (data->flags ==MMC_DATA_READ ) { + buf = (ulong) (data->dest); +owl_dma_config (priv, (ulong ) priv->reg_base + +OWL_REG_SD_DAT , buf, total); +invalidate_dcache_range (buf, buf + total); + }else { + buf = (ulong) (data->src); +owl_dma_config (priv, buf, (ulong ) priv->reg_base + +OWL_REG_SD_DAT , total); +flush_dcache_range (buf, buf + total); + } +/* DMA START */ +writel (0x1 ,SD_DMA_CHANNEL (priv->dma_channel,0 ) +DMA_START ); + } +} + +static int owl_mmc_send_cmd (struct udevice *dev,struct mmc_cmd *cmd, +struct mmc_data *data) +{ +struct owl_mmc_priv *priv =dev_get_priv (dev);
drivers/mmc/owl_mmc.c line 180
Quoted 10 lines
drivers/mmc/owl_mmc.c new file +10
-0+unsigned int cmd_rsp_mask, mode, reg; +int ret; + + reg =readl (priv->reg_base +OWL_REG_SD_EN ); + reg |=OWL_SD_ENABLE ; +writel (reg, priv->reg_base +OWL_REG_SD_EN ); + +/* setup response */ +if (cmd->resp_type ==MMC_RSP_NONE ) + mode =OWL_SD_CTL_TM (0 );
So this whole mode handling here looks dodgy. Below you mix "assignments to mode" with "ORing in values", without actually ever initialising mode explicitly. I wonder why the compiler doesn't warn about this, I can see paths were you OR into an uninitialised value.
Also, the above two lines are actually: if (resp_type == 0) mode = 0; So I would suggest:
mode = 0;
if (cmd->resp_type != MMC_RSP_NONE)
cmd_rsp_mask = OWL_SD_STATE_CLNR | OWL_SD_STATE_CRC7ER;
if (md->resp_type == MMC_RSP_R1) {
.... (always assign to "mode")(I haven't checked yet whether the actual combinations are correct.)
Quoted 40 lines
drivers/mmc/owl_mmc.c new file +40
-0+else { + cmd_rsp_mask =OWL_SD_STATE_CLNR |OWL_SD_STATE_CRC7ER ; + +if (cmd->resp_type ==MMC_RSP_R1 ) { +if (data) { +if (data->flags ==MMC_DATA_READ ) + mode |=OWL_SD_CTL_TM (4 ); +else + mode |=OWL_SD_CTL_TM (5 ); + }else + mode |=OWL_SD_CTL_TM (1 ); + }else if (cmd->resp_type ==MMC_RSP_R2 ) { + mode =OWL_SD_CTL_TM (2 ); + }else if (cmd->resp_type == MMC_RSP_R1b) { + mode =OWL_SD_CTL_TM (3 ); + }else if (cmd->resp_type ==MMC_RSP_R3 ) { + cmd_rsp_mask =OWL_SD_STATE_CLNR ; + mode =OWL_SD_CTL_TM (1 ); + } + } + + mode |= (readl (priv->reg_base +OWL_REG_SD_CTL ) & (0xff <<16 )); + +/* setup command */ +writel (cmd->cmdidx, priv->reg_base +OWL_REG_SD_CMD ); +writel (cmd->cmdarg, priv->reg_base +OWL_REG_SD_ARG ); + +/* Set LBE to send clk at the end of last read block */ +if (data) + mode |= (OWL_SD_CTL_TS |OWL_SD_CTL_LBE |0xE4000000 ); +else + mode |=OWL_SD_CTL_TS ; + +if (data) +owl_mmc_prepare_data (priv, data); + +/* Start transfer */ +writel (mode, priv->reg_base +OWL_REG_SD_CTL ); + + ret = readl_poll_timeout(priv->reg_base + OWL_REG_SD_CTL, reg,
drivers/mmc/owl_mmc.c line 230
Quoted 10 lines
drivers/mmc/owl_mmc.c new file +10
-0+ !(reg & OWL_SD_CTL_TS), DATA_TRANSFER_TIMEOUT); + +if (ret == -ETIMEDOUT ) { +debug ("error: transferred data timeout\n" ); +return ret; + } + + reg =readl (priv->reg_base +OWL_REG_SD_STATE ); +if (cmd->resp_type &MMC_RSP_PRESENT ) { +if (cmd_rsp_mask & reg) {
Can't you mask the bits in reg, then save the outer if?
Quoted 24 lines
drivers/mmc/owl_mmc.c new file +24
-0+if (reg &OWL_SD_STATE_CLNR ) { +printf ("Error CMD_NO_RSP\n" ); +return -1 ; + } + +if (reg &OWL_SD_STATE_CRC7ER ) { +printf ("Error CMD_RSP_CRC\n" ); +return -1 ; + } + + } + +if (cmd->resp_type &MMC_RSP_136 ) { + cmd->response[3 ] =readl (priv->reg_base +OWL_REG_SD_RSPBUF0 ); + cmd->response[2 ] =readl (priv->reg_base +OWL_REG_SD_RSPBUF1 ); + cmd->response[1 ] =readl (priv->reg_base +OWL_REG_SD_RSPBUF2 ); + cmd->response[0 ] =readl (priv->reg_base +OWL_REG_SD_RSPBUF3 ); + }else { +u32 rsp[2 ]; + + rsp[0 ] =readl (priv->reg_base +OWL_REG_SD_RSPBUF0 ); + rsp[1 ] =readl (priv->reg_base +OWL_REG_SD_RSPBUF1 ); + cmd->response[0 ] = rsp[1 ] <<24 | rsp[0 ] >>8 ; + cmd->response[1 ] = rsp[1 ] >>8 ;
drivers/mmc/owl_mmc.c line 264
Quoted 10 lines
drivers/mmc/owl_mmc.c new file +10
-0+ } + } + +if (data) { +/* DMA STOP */ +writel (0x0 ,SD_DMA_CHANNEL (priv->dma_channel,0 ) +DMA_START ); +/* Transmission STOP */ +while (readl (priv->reg_base +OWL_REG_SD_CTL ) &OWL_SD_CTL_TS ) +clrbits_le32 (priv->reg_base +OWL_REG_SD_CTL , +OWL_SD_CTL_TS );
Should there be a timeout here?
Quoted 12 lines
drivers/mmc/owl_mmc.c new file +12
-0+ } + +return 0 ; +} + +static void owl_mmc_clk_set (struct owl_mmc_priv *priv,int rate) +{ +u32 reg, wdelay, rdelay; + + reg =readl (priv->reg_base +OWL_REG_SD_CTL ); + reg &= ~OWL_SD_CTL_DELAY_MSK ; +
drivers/mmc/owl_mmc.c line 286
Quoted 10 lines
drivers/mmc/owl_mmc.c new file +10
-0+/* Set RDELAY and WDELAY based on the clock */ +if (rate <=1000000 ) + rdelay = wdelay =OWL_SD_DELAY_LOW_CLK ; +else if ((rate >1000000 ) && (rate <=26000000 )) + rdelay = wdelay =OWL_SD_DELAY_MID_CLK ; +else if ((rate >26000000 ) && (rate <=52000000 )) { + rdelay =OWL_SD_RDELAY_HIGH ; + wdelay =OWL_SD_WDELAY_HIGH ; + }else +debug ("SD clock rate not supported\n" );
Shouldn't this be more fatal? You still continue to set the register below. Why not make this function return an int, and check for an error in the caller?
Cheers,
Andre
Quoted 116 lines
drivers/mmc/owl_mmc.c new file +115
-0+ +writel (reg |OWL_SD_CTL_RDELAY (rdelay) |OWL_SD_CTL_WDELAY (wdelay), + priv->reg_base +OWL_REG_SD_CTL ); +} + +static int owl_mmc_set_ios (struct udevice *dev) +{ +struct owl_mmc_priv *priv =dev_get_priv (dev); +struct owl_mmc_plat *plat =dev_get_platdata (dev); +struct mmc *mmc = &plat->mmc; +u32 reg, ret; + +if (mmc->clock != priv->clock) { + priv->clock = mmc->clock; +owl_mmc_clk_set (priv, mmc->clock); + + ret =clk_set_rate (&priv->clk, mmc->clock); +if (IS_ERR_VALUE (ret)) +return ret; + } + +if (mmc->clk_disable) + ret =clk_disable (&priv->clk); +else + ret =clk_enable (&priv->clk); +if (ret) +return ret; + +/* Set the Bus width */ + reg =readl (priv->reg_base +OWL_REG_SD_EN ); + reg &= ~OWL_SD_EN_DATAWID_MASK ; +if (mmc->bus_width ==8 ) + reg |=OWL_SD_EN_DATAWID (2 ); +else if (mmc->bus_width ==4 ) + reg |=OWL_SD_EN_DATAWID (1 ); + +writel (reg, priv->reg_base +OWL_REG_SD_EN ); + +return 0 ; +} + +static const struct dm_mmc_ops owl_mmc_ops = { + .send_cmd = owl_mmc_send_cmd, + .set_ios = owl_mmc_set_ios, +}; + +static int owl_mmc_probe (struct udevice *dev) +{ +struct mmc_uclass_priv *upriv =dev_get_uclass_priv (dev); +struct owl_mmc_plat *plat =dev_get_platdata (dev); +struct owl_mmc_priv *priv =dev_get_priv (dev); +struct mmc_config *cfg = &plat->cfg; +struct ofnode_phandle_args args; +int ret; +fdt_addr_t addr; + + cfg->name = dev->name; + cfg->voltages =OWL_MMC_OCR ; + cfg->f_min =400000 ; + cfg->f_max =52000000 ; + cfg->b_max =512 ; + cfg->host_caps =MMC_MODE_HS | MMC_MODE_HS_52MHz; + + ret =mmc_of_parse (dev, cfg); +if (ret) +return ret; + + addr =dev_read_addr (dev); +if (addr ==FDT_ADDR_T_NONE ) +return -EINVAL ; + + priv->reg_base = (void *)addr; + + ret =dev_read_phandle_with_args (dev,"dmas" ,"#dma-cells" ,0 ,0 , + &args); +if (ret) +return ret; + + priv->dma_channel = (void *)ofnode_get_addr (args.node); + priv->dma_drq = args.args[0 ]; + + ret =clk_get_by_index (dev,0 , &priv->clk); +if (ret) { +debug ("clk_get_by_index() failed: %d\n" , ret); +return ret; + } + + upriv->mmc = &plat->mmc; + +return 0 ; +} + +static int owl_mmc_bind (struct udevice *dev) +{ +struct owl_mmc_plat *plat =dev_get_platdata (dev); + +return mmc_bind (dev, &plat->mmc, &plat->cfg); +} + +static const struct udevice_id owl_mmc_ids[] = { + { .compatible ="actions,s700-mmc" }, + { .compatible ="actions,owl-mmc" }, + { } +}; + +U_BOOT_DRIVER (owl_mmc_drv) = { + .name ="owl_mmc" , + .id =UCLASS_MMC , + .of_match = owl_mmc_ids, + .bind = owl_mmc_bind, + .probe = owl_mmc_probe, + .ops = &owl_mmc_ops, + .platdata_auto_alloc_size =sizeof (struct owl_mmc_plat ), + .priv_auto_alloc_size =sizeof (struct owl_mmc_priv ), +};
PATCH v3 5/6 mmc: actions: add MMC driver for Actions OWL S700/S900
Hi,
Quoted 6 lines
So this whole mode handling here looks dodgy. Below you mix "assignments to mode" with "ORing in values", without actually ever initialising mode explicitly. I wonder why the compiler doesn't warn about this, I can see paths were you OR into an uninitialised value.
But the compiler already has initialized mode to 0, that is why there is
no warning.
In order to test , printed out mode value which suggests this variable is initialized.
Thanks
-Amit
PATCH v3 5/6 mmc: actions: add MMC driver for Actions OWL S700/S900
On Mon, 1 Mar 2021 18:47:26 +0530
Amit Tomar <[email protected]> wrote:
Hi,
Quoted 6 lines
Quoted 6 lines
So this whole mode handling here looks dodgy. Below you mix "assignments to mode" with "ORing in values", without actually ever initialising mode explicitly. I wonder why the compiler doesn't warn about this, I can see paths were you OR into an uninitialised value.
But the compiler already has initialized mode to 0,
Why? Where? I just see a local, non-static definition of mode, meaning it won't be initialised.
Quoted 3 lines
that is why there is no warning.
In order to test , printed out mode value which suggests this variable is initialized.
To what? Just because you printed 0(?) in your test doesn't mean that this will always be the case.
Cheers,
Andre
PATCH v3 5/6 mmc: actions: add MMC driver for Actions OWL S700/S900
Quoted 10 lines · Andre Przywara
On Mon, 1 Mar 2021 18:47:26 +0530
Amit Tomar <[email protected]> wrote:Hi,
Quoted 5 lines
Quoted 5 lines
So this whole mode handling here looks dodgy. Below you mix "assignments to mode" with "ORing in values", without actually ever initialising mode explicitly. I wonder why the compiler doesn't warn about this, I can see paths were you OR into an uninitialised value.
Quoted 4 lines
Quoted 1 line
Quoted 1 line
But the compiler already has initialized mode to 0,
Why? Where? I just see a local, non-static definition of mode, meaning it won't be initialised.
It seems that it doesn't initialize in this function. I think that resp_type may be always matched one of condition. I think that's why the compiler doesn't warn.
Best Regards,
Jaehoon Chung
Quoted 11 lines
Quoted 4 lines
that is why there is no warning.
In order to test , printed out mode value which suggests this variable is initialized.To what? Just because you printed 0(?) in your test doesn't mean that this will always be the case.
Cheers,
Andre
PATCH v3 5/6 mmc: actions: add MMC driver for Actions OWL S700/S900
Quoted 11 lines · On Tue, 2 Mar 2021 07:11:54 +0900 Jaehoon Chung <[email protected]> wrote:
Quoted 10 lines · Andre Przywara
On Mon, 1 Mar 2021 18:47:26 +0530
Amit Tomar <[email protected]> wrote:Hi,
Quoted 5 lines
Quoted 5 lines
So this whole mode handling here looks dodgy. Below you mix "assignments to mode" with "ORing in values", without actually ever initialising mode explicitly. I wonder why the compiler doesn't warn about this, I can see paths were you OR into an uninitialised value.
Quoted 7 lines
Quoted 4 lines
Quoted 1 line
Quoted 1 line
But the compiler already has initialized mode to 0,
Why? Where? I just see a local, non-static definition of mode, meaning it won't be initialised.
It seems that it doesn't initialize in this function. I think that
resp_typemay be always matched one of condition.
Possibly, but the problem is that the MMC_RSP_R1 clause also uses "|=", so it's the same issue here.
One a first glance into the assembly it looks like the compiler optimises this whole clause away? Undefined behaviour?
Anyway, I am not really sure we need to argue here, when the fix is dead easy ...
Cheers,
Andre
Quoted 17 lines
I think that's why the compiler doesn't warn.
Best Regards,
Jaehoon ChungQuoted 11 lines
Quoted 4 lines
that is why there is no warning.
In order to test , printed out mode value which suggests this variable is initialized.To what? Just because you printed 0(?) in your test doesn't mean that this will always be the case.
Cheers,
Andre
PATCH v3 5/6 mmc: actions: add MMC driver for Actions OWL S700/S900
Quoted 22 lines · Andre Przywara
Quoted 11 lines · On Tue, 2 Mar 2021 07:11:54 +0900 Jaehoon Chung <[email protected]> wrote:
Quoted 10 lines · Andre Przywara
On Mon, 1 Mar 2021 18:47:26 +0530
Amit Tomar <[email protected]> wrote:Hi,
Quoted 5 lines
Quoted 5 lines
So this whole mode handling here looks dodgy. Below you mix "assignments to mode" with "ORing in values", without actually ever initialising mode explicitly. I wonder why the compiler doesn't warn about this, I can see paths were you OR into an uninitialised value.
Quoted 7 lines
Quoted 4 lines
Quoted 1 line
Quoted 1 line
But the compiler already has initialized mode to 0,
Why? Where? I just see a local, non-static definition of mode, meaning it won't be initialised.
It seems that it doesn't initialize in this function. I think that
resp_typemay be always matched one of condition.
Quoted 7 lines
Possibly, but the problem is that the
MMC_RSP_R1clause also uses "|=", so it's the same issue here.
One a first glance into the assembly it looks like the compiler optimises this whole clause away? Undefined behaviour?Anyway, I am not really sure we need to argue here, when the fix is dead easy ...
Agreed, It doesn't need to discuss about this. If there is a potential problem, it needs to fix it.
Best Regards,
Jaehoon Chung
Quoted 23 lines
Cheers,
AndreQuoted 17 lines
I think that's why the compiler doesn't warn.
Best Regards,
Jaehoon ChungQuoted 11 lines
Quoted 4 lines
that is why there is no warning.
In order to test , printed out mode value which suggests this variable is initialized.To what? Just because you printed 0(?) in your test doesn't mean that this will always be the case.
Cheers,
Andre