Marek Vasut marex@denx.de schrieb am Fr., 3. Aug. 2018, 11:00:
On 08/03/2018 10:40 AM, Simon Goldschmidt wrote:
On 01.08.2018 09:34, Marek Vasut wrote:
On 08/01/2018 09:29 AM, Goldschmidt Simon wrote:
On 30.07.2018 16:04, Marek Vasut wrote:
On 07/30/2018 04:03 PM, Simon Goldschmidt wrote:
On 12.05.2018 22:28, Marek Vasut wrote: > Pull the serial port configuration from DT and use DM serial instead > of having the serial configuration in two places, DT and board > config. > > Signed-off-by: Marek Vasut marex@denx.de > Cc: Chin Liang See chin.liang.see@intel.com > Cc: Dinh Nguyen dinguyen@kernel.org > --- > arch/arm/Kconfig | 3 +++ > arch/arm/dts/socfpga.dtsi | 2 ++ > arch/arm/dts/socfpga_arria10_socdk_sdmmc.dts | 1 + > include/configs/socfpga_common.h | 8 -------- > 4 files changed, 6 insertions(+), 8 deletions(-) > > diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig > index 532aa41a87..2012ac6410 100644 > --- a/arch/arm/Kconfig > +++ b/arch/arm/Kconfig > @@ -737,6 +737,7 @@ config ARCH_SOCFPGA > select ARCH_MISC_INIT > select CPU_V7A > select DM > + select DM_SERIAL > select ENABLE_ARM_SOC_BOOT0_HOOK > select OF_CONTROL > select SPL_LIBCOMMON_SUPPORT > @@ -746,11 +747,13 @@ config ARCH_SOCFPGA > select SPL_NAND_SUPPORT if SPL_NAND_DENALI > select SPL_OF_CONTROL > select SPL_SERIAL_SUPPORT > + select SPL_DM_SERIAL > select SPL_SPI_FLASH_SUPPORT if SPL_SPI_SUPPORT > select SPL_SPI_SUPPORT if DM_SPI > select SPL_WATCHDOG_SUPPORT > select SUPPORT_SPL > select SYS_MMCSD_RAW_MODE_U_BOOT_USE_PARTITION_TYPE > + select SYS_NS16550 > select SYS_THUMB_BUILD > imply CMD_MTDPARTS > imply CRC32_VERIFY > diff --git a/arch/arm/dts/socfpga.dtsi b/arch/arm/dts/socfpga.dtsi > index e64127fcb2..314449478d 100644 > --- a/arch/arm/dts/socfpga.dtsi > +++ b/arch/arm/dts/socfpga.dtsi > @@ -737,6 +737,7 @@ > reg-shift = <2>; > reg-io-width = <4>; > clocks = <&l4_sp_clk>; > + clock-frequency = <100000000>; > }; > uart1: serial1@ffc03000 { > @@ -746,6 +747,7 @@ > reg-shift = <2>; > reg-io-width = <4>; > clocks = <&l4_sp_clk>; > + clock-frequency = <100000000>; > }; > rst: rstmgr@ffd05000 { > diff --git a/arch/arm/dts/socfpga_arria10_socdk_sdmmc.dts > b/arch/arm/dts/socfpga_arria10_socdk_sdmmc.dts > index b573d0e658..06b61cb0af 100644 > --- a/arch/arm/dts/socfpga_arria10_socdk_sdmmc.dts > +++ b/arch/arm/dts/socfpga_arria10_socdk_sdmmc.dts > @@ -24,6 +24,7 @@ > }; > &uart1 { > + clock-frequency = <50000000>; > u-boot,dm-pre-reloc; > status = "okay"; > }; > diff --git a/include/configs/socfpga_common.h > b/include/configs/socfpga_common.h > index 54b9edc97c..a60da85499 100644 > --- a/include/configs/socfpga_common.h > +++ b/include/configs/socfpga_common.h > @@ -173,14 +173,6 @@ unsigned int > cm_get_qspi_controller_clk_hz(void); > * Serial Driver > */ > #define CONFIG_SYS_NS16550_SERIAL > -#define CONFIG_SYS_NS16550_REG_SIZE -4 > -#if defined(CONFIG_TARGET_SOCFPGA_GEN5) > -#define CONFIG_SYS_NS16550_COM1 SOCFPGA_UART0_ADDRESS > -#define CONFIG_SYS_NS16550_CLK 100000000 > -#elif defined(CONFIG_TARGET_SOCFPGA_ARRIA10) > -#define CONFIG_SYS_NS16550_COM1 SOCFPGA_UART1_ADDRESS > -#define CONFIG_SYS_NS16550_CLK 50000000 > -#endif > /* > * USB > Unfortunately I saw this just now, but it seems this breaks GEN5 SPL? At least git-bisect told me that
73172753f4f3351ed1c9d2f6586fc599ce4e728c
is the first bad commit.
I tested socfpga_socrates_defconfig on my socrates board.
Any idea what's wrong there?
Nope, this should work fine. Can you investigate ?
Ok, so after adding "u-boot,dm-pre-reloc" to uart0 in socfpga_cyclone5_socrates.dts, U-Boot works (combined with an old SPL).
Good!
SPL still does not work. Any idea? How does SPL get the uart?
Thanks for any pointers.
Dig around the NS16550 driver .
OK, so my DS-5 debugger shows me it's not the driver that fails but "malloc_simple()" returns 0x0 when being called from uclass_get_device_by_seq(). Which is strange because gd->malloc_base is 0xffffe000.
malloc() returns NULL because it ran out of space ?
No, the malloc limit is not even set in gd. It just somehow fails to add malloc_base. I don't get why, as it should do so given the disassembly.
The next step is then zeroing the memory from calloc() and it doesn't return from there.
I guess next I could check if malloc_simple is called at all without serial DM. That would at least explain how this error is connected to your commit...
Also, gd->fdt_blob is not initialized at the time serial_init() is
called from SPL. Should it be? Because that way, we don't seem to have dts control over which uart is used as console?
Can someone shed some light on this? Should the device tree be available when allocating the SPL console? Because if not, it would just use the first driver available, or am I wrong?
Simon