serial: ns16550: Support run-time configuration
PATCH v2 1/4 serial: ns16550: Support run-time configuration
At present this driver uses an assortment of CONFIG options to control how it accesses the hardware. This is painful for platforms that are supposed to be controlled by a device tree or a previous-stage bootloader.
Add a new CONFIG option to enable fully dynamic configuration. This controls register spacing, size, offset and endianness.
---
Changes in v2:
- runtime -> run-time
- Enable run-time config for slimbootloader too
- Improve Kconfig help based on Bin's comments
- Use ns16550 in patch subject
drivers/serial/Kconfig | 21 +++++++++++++++ drivers/serial/ns16550.c | 57 ++++++++++++++++++++++++++++++++++------ include/ns16550.h | 13 +++++++++ 3 files changed, 83 insertions(+), 8 deletions(-)
drivers/serial/Kconfig +21 -0
@@ -598,6 +598,27 @@ config SYS_NS16550 be used. It can be a constant or a function to get clock, eg, get_serial_clock(). +config NS16550_DYNAMIC + bool "Allow NS16550 to be configured at runtime" + default y if SYS_COREBOOT || SYS_SLIMBOOTLOADER + help + Enable this option to allow device-tree control of the driver. + + Normally this driver is controlled by the following options: + + CONFIG_SYS_NS16550_PORT_MAPPED - indicates that port I/O is used for + access. If not enabled, then the UART is memory-mapped. + CONFIG_SYS_NS16550_MEM32 - if memory-mapped, indicates that 32-bit + access should be used (instead of 8-bit) + CONFIG_SYS_NS16550_REG_SIZE - indicates register width and also + endianness. If positive, big-endian access is used. If negative, + little-endian is used. + + It is not a good practice for a driver to be statically configured, + since it prevents the same driver being used for different types of + UARTs in a system. This option avoids this problem at the cost of a + slightly increased code size. + config INTEL_MID_SERIAL bool "Intel MID platform UART support" depends on DM_SERIAL && OF_CONTROL
drivers/serial/ns16550.c +49 -8
77 lines not on this page · raw message
include/ns16550.h +13 -0
@@ -31,6 +31,9 @@#define CONFIG_SYS_NS16550_REG_SIZE (-1)#endif +#ifdef CONFIG_NS16550_DYNAMIC +#define UART_REG (x) unsigned char x +#else #if !defined(CONFIG_SYS_NS16550_REG_SIZE ) || (CONFIG_SYS_NS16550_REG_SIZE ==0 )#error "Please define NS16550 registers size."#elif defined(CONFIG_SYS_NS16550_MEM32 ) && !defined(CONFIG_DM_SERIAL )
@@ -44,6 +47,12 @@unsigned char x; \ unsigned char postpad_##x[-CONFIG_SYS_NS16550_REG_SIZE - 1];#endif +#endif /* CONFIG_NS16550_DYNAMIC */ + +enum ns16550_flags { +NS16550_FLAG_IO =1 <<0 ,/* Use I/O access (else mem-mapped) */ +NS16550_FLAG_BE =1 <<1 ,/* Use big-endian access (else little) */ +};/** * struct ns16550_platdata - information about a NS16550 port
@@ -51,7 +60,10 @@* @base: Base register address * @reg_width: IO accesses size of registers (in bytes) * @reg_shift: Shift size of registers (0=byte, 1=16bit, 2=32bit...) +* @reg_offset: Offset to start of registers * @clock: UART base clock speed in Hz +* @fcr: Offset of FCR register (normally UART_FCR_DEFVAL) +* @flags: A few flags (enum ns16550_flags) * @bdf: PCI slot/function (pci_dev_t) */ struct ns16550_platdata {
@@ -61,6 +73,7 @@ struct ns16550_platdata {int reg_offset;int clock;u32 fcr; +int flags;#if defined(CONFIG_PCI ) && defined(CONFIG_SPL )int bdf;#endif
Signature · 1 line
2.24.0.393.g34dc348eaf-goog
PATCH v2 2/4 x86: Update coreboot serial table struct
Since mid 2016, coreboot has additional fields in the serial struct that it passes down to U-Boot. Add these so we are in sync.
---
Changes in v2: None
arch/x86/include/asm/coreboot_tables.h | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+)
arch/x86/include/asm/coreboot_tables.h +19 -0
@@ -97,6 +97,25 @@ struct cb_serial {u32 type;u32 baseaddr;u32 baud; +u32 regwidth; + +/* +* Crystal or input frequency to the chip containing the UART. +* Provide the board specific details to allow the payload to +* initialize the chip containing the UART and make independent +* decisions as to which dividers to select and their values +* to eventually arrive at the desired console baud-rate. +*/ +u32 input_hertz; + +/* +* UART PCI address: bus, device, function +* 1 << 31 - Valid bit, PCI UART in use +* Bus << 20 +* Device << 15 +* Function << 12 +*/ +u32 uart_pci_addr; };#define CB_TAG_CONSOLE 0x0010
Signature · 1 line
2.24.0.393.g34dc348eaf-goog
PATCH v2 3/4 x86: serial: Add a coreboot serial driver
In reply to Simon GlassCoreboot can provide information about the serial device in use on a platform. Add a driver that uses this information to produce a working UART.
---
Changes in v2: None
drivers/serial/Kconfig | 11 ++++++++ drivers/serial/Makefile | 1 + drivers/serial/serial_coreboot.c | 46 ++++++++++++++++++++++++++++++++ 3 files changed, 58 insertions(+) create mode 100644 drivers/serial/serial_coreboot.c
drivers/serial/Kconfig +11 -0
@@ -539,6 +539,17 @@ config BCM6345_SERIAL help Select this to enable UART on BCM6345 SoCs. +config COREBOOT_SERIAL + bool "Coreboot UART support" + depends on DM_SERIAL + default y if SYS_COREBOOT + select SYS_NS16550 + help + Select this to enable a ns16550-style UART where the platform data + comes from the coreboot 'sysinfo' tables. This allows U-Boot to have + a serial console on any platform without needing to change the + device tree, etc. + config FSL_LINFLEXUART bool "Freescale Linflex UART support" depends on DM_SERIAL
drivers/serial/Makefile +1 -0
@@ -35,6 +35,7 @@ obj-$(CONFIG_AR933X_UART) += serial_ar933x.o obj-$(CONFIG_ARM_DCC) += arm_dcc.o obj-$(CONFIG_ATMEL_USART) += atmel_usart.o obj-$(CONFIG_BCM6345_SERIAL) += serial_bcm6345.o +obj-$(CONFIG_COREBOOT_SERIAL) += serial_coreboot.o obj-$(CONFIG_EFI_APP) += serial_efi.o obj-$(CONFIG_LPC32XX_HSUART) += lpc32xx_hsuart.o obj-$(CONFIG_MCFUART) += mcfuart.o
drivers/serial/serial_coreboot.c new file +46 -0
46 lines not on this page · raw message
Signature · 1 line
2.24.0.393.g34dc348eaf-goog
PATCH v2 4/4 x86: Move coreboot over to use the coreboot UART
In reply to Simon GlassUse this UART to improve the compatibility of U-Boot when used as a coreboot payload.
---
Changes in v2: None
arch/x86/dts/coreboot.dts | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-)
arch/x86/dts/coreboot.dts +5 -1
@@ -8,7 +8,6 @@ /dts-v1/; /include/ "skeleton.dtsi" -/include/ "serial.dtsi" /include/ "keyboard.dtsi" /include/ "pcspkr.dtsi" /include/ "reset.dtsi"
@@ -40,6 +39,11 @@ u-boot,dm-pre-reloc; }; + serial: serial { + u-boot,dm-pre-reloc; + compatible = "coreboot-serial"; + }; + coreboot-fb { compatible = "coreboot-fb"; };
Signature · 1 line
2.24.0.393.g34dc348eaf-goog
PATCH v2 1/4 serial: ns16550: Support run-time configuration
In reply to Simon GlassAm 09.12.2019 um 17:59 schrieb Simon Glass:
Quoted 175 lines
At present this driver uses an assortment of CONFIG options to control how it accesses the hardware. This is painful for platforms that are supposed to be controlled by a device tree or a previous-stage bootloader.
Add a new CONFIG option to enable fully dynamic configuration. This controls register spacing, size, offset and endianness.
Signed-off-by Simon Glass <[email protected]>---
Changes in v2:
- runtime -> run-time
- Enable run-time config for slimbootloader too
- ImproveKconfighelp based on Bin's comments
- Use ns16550 in patch subjectdrivers/serial/Kconfig | 21 +++++++++++++++ drivers/serial/ns16550.c | 57 ++++++++++++++++++++++++++++++++++------ include/ns16550.h | 13 +++++++++ 3 files changed, 83 insertions(+), 8 deletions(-)drivers/serial/Kconfig +21
-0@@ -598,6 +598,27 @@ config SYS_NS16550 be used. It can be a constant or a function to get clock, eg, get_serial_clock(). +config NS16550_DYNAMIC + bool "Allow NS16550 to be configured at runtime" + default y if SYS_COREBOOT || SYS_SLIMBOOTLOADER + help + Enable this option to allow device-tree control of the driver. + + Normally this driver is controlled by the following options: + + CONFIG_SYS_NS16550_PORT_MAPPED - indicates that port I/O is used for + access. If not enabled, then the UART is memory-mapped. + CONFIG_SYS_NS16550_MEM32 - if memory-mapped, indicates that 32-bit + access should be used (instead of 8-bit) + CONFIG_SYS_NS16550_REG_SIZE - indicates register width and also + endianness. If positive, big-endian access is used. If negative, + little-endian is used. + + It is not a good practice for a driver to be statically configured, + since it prevents the same driver being used for different types of + UARTs in a system. This option avoids this problem at the cost of a + slightly increased code size. + config INTEL_MID_SERIAL bool "Intel MID platform UART support" depends on DM_SERIAL && OF_CONTROLdrivers/serial/ns16550.c +49
-8@@ -92,19 +92,57 @@ static inline int serial_in_shift(void *addr, int shift)#define CONFIG_SYS_NS16550_CLK 0#endif +static void serial_out_dynamic (struct ns16550_platdata *plat,u8 *addr, +int value) +{ +if (plat->flags &NS16550_FLAG_BE ) { +if (plat->reg_width ==1 ) +writeb (value, addr + (1 << plat->reg_shift) -1 ); +else if (plat->flags &NS16550_FLAG_IO ) +out_be32 (addr, value); +else +writel (value, addr); + }else { +if (plat->reg_width ==1 ) +writeb (value, addr); +else if (plat->flags &NS16550_FLAG_IO ) +out_le32 (addr, value); +else +writel (value, addr); + } +} + +static int serial_in_dynamic (struct ns16550_platdata *plat,u8 *addr) +{ +if (plat->flags &NS16550_FLAG_BE ) { +if (plat->reg_width ==1 ) +return readb (addr + (1 << plat->reg_shift) -1 ); +else if (plat->flags &NS16550_FLAG_IO ) +return in_be32 (addr); +else +return readl (addr); + }else { +if (plat->reg_width ==1 ) +return readb (addr); +else if (plat->flags &NS16550_FLAG_IO ) +return in_le32 (addr); +else +return readl (addr); + } +} +static void ns16550_writeb (NS16550_t port,int offset,int value) {struct ns16550_platdata *plat = port->plat;unsigned char *addr; offset *=1 << plat->reg_shift; - addr = (unsigned char *)plat->base + offset; + addr = (unsigned char *)plat->base + offset + plat->reg_offset; -/* -* As far as we know it doesn't make sense to support selection of -* these options at run-time, so use the existing CONFIG options. -*/ -serial_out_shift (addr + plat->reg_offset, plat->reg_shift, value); +if (IS_ENABLED (CONFIG_NS16550_DYNAMIC )) +serial_out_dynamic (plat, addr, value); +else +serial_out_shift (addr, plat->reg_shift, value); }static int ns16550_readb (NS16550_t port,int offset)@@ -113,9 +151,12 @@ static int ns16550_readb(NS16550_t port, int offset)unsigned char *addr; offset *=1 << plat->reg_shift; - addr = (unsigned char *)plat->base + offset; + addr = (unsigned char *)plat->base + offset + plat->reg_offset; -return serial_in_shift (addr + plat->reg_offset, plat->reg_shift); +if (IS_ENABLED (CONFIG_NS16550_DYNAMIC )) +return serial_in_dynamic (plat, addr); +else +return serial_in_shift (addr, plat->reg_shift); }static u32 ns16550_getfcr (NS16550_t port)include/ns16550.h +12
-0@@ -31,6 +31,9 @@#define CONFIG_SYS_NS16550_REG_SIZE (-1)#endif +#ifdef CONFIG_NS16550_DYNAMIC +#define UART_REG (x) unsigned char x +#else #if !defined(CONFIG_SYS_NS16550_REG_SIZE ) || (CONFIG_SYS_NS16550_REG_SIZE ==0 )#error "Please define NS16550 registers size."#elif defined(CONFIG_SYS_NS16550_MEM32 ) && !defined(CONFIG_DM_SERIAL )@@ -44,6 +47,12 @@unsigned char x; \ unsigned char postpad_##x[-CONFIG_SYS_NS16550_REG_SIZE - 1];#endif +#endif /* CONFIG_NS16550_DYNAMIC */ + +enum ns16550_flags { +NS16550_FLAG_IO =1 <<0 ,/* Use I/O access (else mem-mapped) */ +NS16550_FLAG_BE =1 <<1 ,/* Use big-endian access (else little) */ +};/** * struct ns16550_platdata - information about a NS16550 port @@ -51,7 +60,10 @@* @base: Base register address * @reg_width: IO accesses size of registers (in bytes) * @reg_shift: Shift size of registers (0=byte, 1=16bit, 2=32bit...) +* @reg_offset: Offset to start of registers * @clock: UART base clock speed in Hz +* @fcr: Offset of FCR register (normally UART_FCR_DEFVAL) +* @flags: A few flags (enum ns16550_flags) * @bdf: PCI slot/function (pci_dev_t) */ struct ns16550_platdata {
include/ns16550.h struct ns16550_platdata { · line 76
Quoted 4 lines
include/ns16550.h +1
-0@@ -61,6 +73,7 @@ struct ns16550_platdata {int reg_offset;int clock;u32 fcr; +int flags;
I'm not too fond of adding 4 bytes here for nothing on most platforms, but unless you come up with a better idea than cluttering this file with ifdefs:
Quoted 4 lines
include/ns16550.h +0
-0#if defined(CONFIG_PCI ) && defined(CONFIG_SPL )int bdf;#endif
PATCH v2 1/4 serial: ns16550: Support run-time configuration
In reply to Simon GlassQuoted 17 lines · Simon Glass
At present this driver uses an assortment of CONFIG options to control how it accesses the hardware. This is painful for platforms that are supposed to be controlled by a device tree or a previous-stage bootloader.
Add a new CONFIG option to enable fully dynamic configuration. This controls register spacing, size, offset and endianness.
Signed-off-by Simon Glass <[email protected]>---
Changes in v2:
- runtime -> run-time
- Enable run-time config for slimbootloader too
- ImproveKconfighelp based on Bin's comments
- Use ns16550 in patch subject
Quoted 5 lines
drivers/serial/Kconfig | 21 +++++++++++++++ drivers/serial/ns16550.c | 57 ++++++++++++++++++++++++++++++++++------ include/ns16550.h | 13 +++++++++ 3 files changed, 83 insertions(+), 8 deletions(-)
Park
PATCH v2 1/4 serial: ns16550: Support run-time configuration
In reply to Simon GlassHi Simon,
Quoted 79 lines
-----Original Message-----
From: U-Boot <[email protected]> On Behalf Of Simon Glass
Sent: Monday, December 9, 2019 8:59 AM
To: U-Boot Mailing List <[email protected]>
Cc: Stefan Roese <[email protected]>; Angelo Dureghello <[email protected]>
Subject: [PATCH v2 1/4] serial: ns16550: Support run-time configurationAt present this driver uses an assortment of CONFIG options to control how it accesses the hardware. This is painful for platforms that are supposed to be controlled by a device tree or a previous-stage bootloader.
Add a new CONFIG option to enable fully dynamic configuration. This controls register spacing, size, offset and endianness.
Signed-off-by Simon Glass <[email protected]>---
Changes in v2:
- runtime -> run-time
- Enable run-time config for slimbootloader too
- ImproveKconfighelp based on Bin's comments
- Use ns16550 in patch subjectdrivers/serial/Kconfig | 21 +++++++++++++++ drivers/serial/ns16550.c | 57 ++++++++++++++++++++++++++++++++++---- -- include/ns16550.h | 13 +++++++++ 3 files changed, 83 insertions(+), 8 deletions(-)drivers/serial/Kconfig index renamed from drivers/serial/Kconfig +0
-0
ece7d87d4c..472a9f0929100644drivers/serial/Kconfig +9
-0@@ -598,6 +598,27 @@ config SYS_NS16550 be used. It can be a constant or a function to get clock, eg, get_serial_clock(). +config NS16550_DYNAMIC + bool "Allow NS16550 to be configured at runtime" + default y if SYS_COREBOOT || SYS_SLIMBOOTLOADER + help + Enable this option to allow device-tree control of the driver. + + Normally this driver is controlled by the following options: + + CONFIG_SYS_NS16550_PORT_MAPPED - indicates that port I/O isused for
drivers/serial/Kconfig +2
-0+ access. If not enabled, then the UART is memory-mapped. + CONFIG_SYS_NS16550_MEM32 - if memory-mapped, indicates that32-bit
drivers/serial/Kconfig +10
-0+ access should be used (instead of 8-bit) + CONFIG_SYS_NS16550_REG_SIZE - indicates register width and also + endianness. If positive, big-endian access is used. If negative, + little-endian is used. + + It is not a good practice for a driver to be statically configured, + since it prevents the same driver being used for different types of + UARTs in a system. This option avoids this problem at the cost of a + slightly increased code size. + config INTEL_MID_SERIAL bool "Intel MID platform UART support" depends on DM_SERIAL && OF_CONTROLdrivers/serial/ns16550.c index renamed from drivers/serial/ns16550.c +0
-0
754b6e9921..96c4471efd100644drivers/serial/ns16550.c +9
-0@@ -92,19 +92,57 @@ static inline int serial_in_shift(void *addr, int shift)#define CONFIG_SYS_NS16550_CLK 0 #endif +static void serial_out_dynamic (struct ns16550_platdata *plat,u8 *addr, +int value) +{ +if (plat->flags &NS16550_FLAG_BE ) { +if (plat->reg_width ==1 ) +writeb (value, addr + (1 << plat->reg_shift) -1 ); +else if (plat->flags &NS16550_FLAG_IO ) +out_be32 (addr, value); +else
drivers/serial/ns16550.c static inline int serial_in_shift(void *addr, int shift) · line 112
Quoted 10 lines
drivers/serial/ns16550.c +10
-0+writel (value, addr); + }else { +if (plat->reg_width ==1 ) +writeb (value, addr); +else if (plat->flags &NS16550_FLAG_IO ) +out_le32 (addr, value); +else +writel (value, addr); + } +}
IO needs to use outb(). It breaks QEMU 0x3f8 IO (reg_width = 1, flags=IO). I have verified 0x3f8 IO on QEMU and MMIO32 on APL with below code.
if (plat->flags & NS16550_FLAG_IO) {
outb(value, addr);
} else {
if (plat->flags & NS16550_FLAG_BE) {
if (plat->reg_width == 1)
writeb(value, addr + (1 << plat->reg_shift) - 1);
else
out_be32(addr, value);
} else {
if (plat->reg_width == 1)
writeb(value, addr);
else
out_le32(addr, value);
}
}Quoted 8 lines
drivers/serial/ns16550.c +8
-0+ +static int serial_in_dynamic (struct ns16550_platdata *plat,u8 *addr) { +if (plat->flags &NS16550_FLAG_BE ) { +if (plat->reg_width ==1 ) +return readb (addr + (1 << plat->reg_shift) -1 ); +else if (plat->flags &NS16550_FLAG_IO ) +return in_be32 (addr); +else
drivers/serial/ns16550.c static inline int serial_in_shift(void *addr, int shift) · line 132
Quoted 10 lines
drivers/serial/ns16550.c +10
-0+return readl (addr); + }else { +if (plat->reg_width ==1 ) +return readb (addr); +else if (plat->flags &NS16550_FLAG_IO ) +return in_le32 (addr); +else +return readl (addr); + } +}
Similar as above.
if (plat->flags & NS16550_FLAG_IO) {
return inb((ulong)addr);
} else {
if (plat->flags & NS16550_FLAG_BE) {
if (plat->reg_width == 1)
return readb(addr + (1 << plat->reg_shift) - 1);
else
return in_be32(addr);
} else {
if (plat->reg_width == 1)
return readb(addr);
else
return in_le32(addr);
}
}Quoted 84 lines
drivers/serial/ns16550.c +11
-8+static void ns16550_writeb (NS16550_t port,int offset,int value) {struct ns16550_platdata *plat = port->plat;unsigned char *addr; offset *=1 << plat->reg_shift; - addr = (unsigned char *)plat->base + offset; + addr = (unsigned char *)plat->base + offset + plat->reg_offset; -/* -* As far as we know it doesn't make sense to support selection of -* these options at run-time, so use the existing CONFIG options. -*/ -serial_out_shift (addr + plat->reg_offset, plat->reg_shift, value); +if (IS_ENABLED (CONFIG_NS16550_DYNAMIC )) +serial_out_dynamic (plat, addr, value); +else +serial_out_shift (addr, plat->reg_shift, value); } static int ns16550_readb(NS16550_t port, int offset) @@ -113,9 +151,12 @@ static int ns16550_readb(NS16550_t port, int offset) unsigned char *addr; offset *=1 << plat->reg_shift; - addr = (unsigned char *)plat->base + offset; + addr = (unsigned char *)plat->base + offset + plat->reg_offset; -return serial_in_shift (addr + plat->reg_offset, plat->reg_shift); +if (IS_ENABLED (CONFIG_NS16550_DYNAMIC )) +return serial_in_dynamic (plat, addr); +else +return serial_in_shift (addr, plat->reg_shift); } static u32 ns16550_getfcr(NS16550_t port) diff --git a/include/ns16550.h
b/include/ns16550.hindex701efeea85..ba9b71962d100644include/ns16550.h +3
-0@@ -31,6 +31,9 @@#define CONFIG_SYS_NS16550_REG_SIZE (-1) #endif +#ifdef CONFIG_NS16550_DYNAMIC +#define UART_REG (x) unsigned char x +#else #if !defined(CONFIG_SYS_NS16550_REG_SIZE) ||(
CONFIG_SYS_NS16550_REG_SIZE== 0) #error "Please define NS16550 registers size."
#elifdefined(CONFIG_SYS_NS16550_MEM32)
&& !defined(CONFIG_DM_SERIAL) @@ -44,6 +47,12 @@include/ns16550.h +0
-0unsigned char x; \ unsigned char postpad_##x[-CONFIG_SYS_NS16550_REG_SIZE - 1];#endif
include/ns16550.h +4
-0+#endif /* CONFIG_NS16550_DYNAMIC */ + +enum ns16550_flags { +NS16550_FLAG_IO =1 <<0 ,/* Use I/O access (else mem- mapped) */
include/ns16550.h +1
-0+ NS16550_FLAG_BE = 1 << 1, /* Use big-endian access (else little)*/
include/ns16550.h +6
-1+};/** * struct ns16550_platdata - information about a NS16550 port @@ -51,7 +60,10 @@ * @base: Base register address * @reg_width: IO accesses size of registers (in bytes) * @reg_shift: Shift size of registers (0=byte, 1=16bit, 2=32bit...) +* @reg_offset: Offset to start of registers * @clock: UART base clock speed in Hz +* @fcr: Offset of FCR register (normally UART_FCR_DEFVAL) +* @flags: A few flags (enum ns16550_flags) * @bdf: PCI slot/function (pci_dev_t) */ struct ns16550_platdata {@@ -61,6 +73,7 @@ struct ns16550_platdata {int reg_offset;int clock;u32 fcr; +int flags;#if defined(CONFIG_PCI ) && defined(CONFIG_SPL )int bdf;#endif --
Quoted 1 line
2.24.0.393.g34dc348eaf-goog
Best Regards,
Aiden
PATCH v2 1/4 serial: ns16550: Support run-time configuration
Hi Simon,
Quoted 94 lines · Aiden
Hi Simon,
Quoted 79 lines
-----Original Message-----
From: U-Boot <[email protected]> On Behalf Of Simon Glass
Sent: Monday, December 9, 2019 8:59 AM
To: U-Boot Mailing List <[email protected]>
Cc: Stefan Roese <[email protected]>; Angelo Dureghello <[email protected]>
Subject: [PATCH v2 1/4] serial: ns16550: Support run-time configurationAt present this driver uses an assortment of CONFIG options to control how it accesses the hardware. This is painful for platforms that are supposed to be controlled by a device tree or a previous-stage bootloader.
Add a new CONFIG option to enable fully dynamic configuration. This controls register spacing, size, offset and endianness.
Signed-off-by Simon Glass <[email protected]>---
Changes in v2:
- runtime -> run-time
- Enable run-time config for slimbootloader too
- ImproveKconfighelp based on Bin's comments
- Use ns16550 in patch subjectdrivers/serial/Kconfig | 21 +++++++++++++++ drivers/serial/ns16550.c | 57 ++++++++++++++++++++++++++++++++++---- -- include/ns16550.h | 13 +++++++++ 3 files changed, 83 insertions(+), 8 deletions(-)drivers/serial/Kconfig index renamed from drivers/serial/Kconfig +0
-0
ece7d87d4c..472a9f0929100644drivers/serial/Kconfig +9
-0@@ -598,6 +598,27 @@ config SYS_NS16550 be used. It can be a constant or a function to get clock, eg, get_serial_clock(). +config NS16550_DYNAMIC + bool "Allow NS16550 to be configured at runtime" + default y if SYS_COREBOOT || SYS_SLIMBOOTLOADER + help + Enable this option to allow device-tree control of the driver. + + Normally this driver is controlled by the following options: + + CONFIG_SYS_NS16550_PORT_MAPPED - indicates that port I/O isused for
drivers/serial/Kconfig +2
-0+ access. If not enabled, then the UART is memory-mapped. + CONFIG_SYS_NS16550_MEM32 - if memory-mapped, indicates that32-bit
drivers/serial/Kconfig +10
-0+ access should be used (instead of 8-bit) + CONFIG_SYS_NS16550_REG_SIZE - indicates register width and also + endianness. If positive, big-endian access is used. If negative, + little-endian is used. + + It is not a good practice for a driver to be statically configured, + since it prevents the same driver being used for different types of + UARTs in a system. This option avoids this problem at the cost of a + slightly increased code size. + config INTEL_MID_SERIAL bool "Intel MID platform UART support" depends on DM_SERIAL && OF_CONTROLdrivers/serial/ns16550.c index renamed from drivers/serial/ns16550.c +0
-0
754b6e9921..96c4471efd100644drivers/serial/ns16550.c +9
-0@@ -92,19 +92,57 @@ static inline int serial_in_shift(void *addr, int shift)#define CONFIG_SYS_NS16550_CLK 0 #endif +static void serial_out_dynamic (struct ns16550_platdata *plat,u8 *addr, +int value) +{ +if (plat->flags &NS16550_FLAG_BE ) { +if (plat->reg_width ==1 ) +writeb (value, addr + (1 << plat->reg_shift) -1 ); +else if (plat->flags &NS16550_FLAG_IO ) +out_be32 (addr, value); +else drivers/serial/ns16550.c static inline int serial_in_shift(void *addr, int shift) · line 112
Quoted 10 lines
drivers/serial/ns16550.c +10
-0+writel (value, addr); + }else { +if (plat->reg_width ==1 ) +writeb (value, addr); +else if (plat->flags &NS16550_FLAG_IO ) +out_le32 (addr, value); +else +writel (value, addr); + } +}IO needs to use
outb(). It breaks QEMU0x3f8IO (reg_width= 1, flags=IO). I have verified0x3f8IO on QEMU and MMIO32 on APL with below code.
Quoted 16 lines
if (plat->flags & NS16550_FLAG_IO) {
outb(value, addr);
} else {
if (plat->flags & NS16550_FLAG_BE) {
if (plat->reg_width == 1)
writeb(value, addr + (1 << plat->reg_shift) - 1);
else
out_be32(addr, value);
} else {
if (plat->reg_width == 1)
writeb(value, addr);
else
out_le32(addr, value);
}
}Would you post a v3 that fixes the issue that Aiden pointed out?
Regards,
Bin
PATCH v2 1/4 serial: ns16550: Support run-time configuration
Hi Bin,
Quoted 115 lines · Bin Meng
Hi Simon,
Quoted 94 lines · Aiden
Hi Simon,
Quoted 79 lines
-----Original Message-----
From: U-Boot <[email protected]> On Behalf Of Simon Glass
Sent: Monday, December 9, 2019 8:59 AM
To: U-Boot Mailing List <[email protected]>
Cc: Stefan Roese <[email protected]>; Angelo Dureghello <[email protected]>
Subject: [PATCH v2 1/4] serial: ns16550: Support run-time configurationAt present this driver uses an assortment of CONFIG options to control how it accesses the hardware. This is painful for platforms that are supposed to be controlled by a device tree or a previous-stage bootloader.
Add a new CONFIG option to enable fully dynamic configuration. This controls register spacing, size, offset and endianness.
Signed-off-by Simon Glass <[email protected]>---
Changes in v2:
- runtime -> run-time
- Enable run-time config for slimbootloader too
- ImproveKconfighelp based on Bin's comments
- Use ns16550 in patch subjectdrivers/serial/Kconfig | 21 +++++++++++++++ drivers/serial/ns16550.c | 57 ++++++++++++++++++++++++++++++++++---- -- include/ns16550.h | 13 +++++++++ 3 files changed, 83 insertions(+), 8 deletions(-)drivers/serial/Kconfig index renamed from drivers/serial/Kconfig +0
-0
ece7d87d4c..472a9f0929100644drivers/serial/Kconfig +9
-0@@ -598,6 +598,27 @@ config SYS_NS16550 be used. It can be a constant or a function to get clock, eg, get_serial_clock(). +config NS16550_DYNAMIC + bool "Allow NS16550 to be configured at runtime" + default y if SYS_COREBOOT || SYS_SLIMBOOTLOADER + help + Enable this option to allow device-tree control of the driver. + + Normally this driver is controlled by the following options: + + CONFIG_SYS_NS16550_PORT_MAPPED - indicates that port I/O isused for
drivers/serial/Kconfig +2
-0+ access. If not enabled, then the UART is memory-mapped. + CONFIG_SYS_NS16550_MEM32 - if memory-mapped, indicates that32-bit
drivers/serial/Kconfig +10
-0+ access should be used (instead of 8-bit) + CONFIG_SYS_NS16550_REG_SIZE - indicates register width and also + endianness. If positive, big-endian access is used. If negative, + little-endian is used. + + It is not a good practice for a driver to be statically configured, + since it prevents the same driver being used for different types of + UARTs in a system. This option avoids this problem at the cost of a + slightly increased code size. + config INTEL_MID_SERIAL bool "Intel MID platform UART support" depends on DM_SERIAL && OF_CONTROLdrivers/serial/ns16550.c index renamed from drivers/serial/ns16550.c +0
-0
754b6e9921..96c4471efd100644drivers/serial/ns16550.c +9
-0@@ -92,19 +92,57 @@ static inline int serial_in_shift(void *addr, int shift)#define CONFIG_SYS_NS16550_CLK 0 #endif +static void serial_out_dynamic (struct ns16550_platdata *plat,u8 *addr, +int value) +{ +if (plat->flags &NS16550_FLAG_BE ) { +if (plat->reg_width ==1 ) +writeb (value, addr + (1 << plat->reg_shift) -1 ); +else if (plat->flags &NS16550_FLAG_IO ) +out_be32 (addr, value); +else drivers/serial/ns16550.c static inline int serial_in_shift(void *addr, int shift) · line 112
Quoted 10 lines
drivers/serial/ns16550.c +10
-0+writel (value, addr); + }else { +if (plat->reg_width ==1 ) +writeb (value, addr); +else if (plat->flags &NS16550_FLAG_IO ) +out_le32 (addr, value); +else +writel (value, addr); + } +}IO needs to use
outb(). It breaks QEMU0x3f8IO (reg_width= 1, flags=IO). I have verified0x3f8IO on QEMU and MMIO32 on APL with below code.Quoted 16 lines
if (plat->flags & NS16550_FLAG_IO) { outb(value, addr); } else { if (plat->flags & NS16550_FLAG_BE) { if (plat->reg_width == 1) writeb(value, addr + (1 << plat->reg_shift) - 1); else out_be32(addr, value); } else { if (plat->reg_width == 1) writeb(value, addr); else out_le32(addr, value); } }
Quoted 1 line
Would you post a v3 that fixes the issue that Aiden pointed out?
Yes, but actually I have found that we need one more case, so it needs a bit more effort. Will post an update series by Monday.
Regards,
SImon