board: atmel: sama5d2_wlsom1_ek: Fix spurious serial output
PATCH v2 1/2 board: atmel: sama5d2_wlsom1_ek: Fix spurious serial output
From: Fabio Estevam <[email protected]>
Currently, there are always some spurious characters showing up right before SPL banner is shown:
RomBOOT
���=������
U-Boot SPL 2025.04-rc5-00023-g9ed4e2c45f25 (Apr 03 2025 - 23:23:17 -0300) Trying to boot from MMC1
<debug_uart>
...
The reason for the spurious characters is that the UART0 clock is being enabled too early, prior to adjusting its correct frequency.
Fix the problem by removing the early UART0 clock enablement in C code.
SPL DM code will adjust the correct UART0 clock frequency and then enable it at the appropriate time.
---
Changes since v1:
- Make the minimal change to fix the spurious characters. Improve the explanation in the commit log.
board/atmel/sama5d27_wlsom1_ek/sama5d27_wlsom1_ek.c | 2 -- 1 file changed, 2 deletions(-)
board/atmel/sama5d27_wlsom1_ek/sama5d27_wlsom1_ek.c +0 -2
@@ -45,8 +45,6 @@ static void board_uart0_hw_init(void) {atmel_pio4_set_c_periph (AT91_PIO_PORTB ,26 ,ATMEL_PIO_PUEN_MASK );/* URXD0 */ atmel_pio4_set_c_periph (AT91_PIO_PORTB ,27 ,0 );/* UTXD0 */ - -at91_periph_clk_enable (ATMEL_ID_UART0 ); }void board_debug_uart_init (void )
Signature · 1 line
2.34.1
PATCH v2 2/2 board: atmel: sama5d2_wlsom1_ek: Remove UART pinctrl initialization
From: Fabio Estevam <[email protected]>
The UART0 pinctrl initialization is already done by DM via CONFIG_PINCTRL_AT91PIO4=y.
Remove the unnecessary UART0 pinctrl initialization done in board code.
---
Changes since v1:
- Newly introduced.
board/atmel/sama5d27_wlsom1_ek/sama5d27_wlsom1_ek.c | 7 ------- 1 file changed, 7 deletions(-)
board/atmel/sama5d27_wlsom1_ek/sama5d27_wlsom1_ek.c +0 -7
@@ -41,15 +41,8 @@ int board_late_init(void)#endif #ifdef CONFIG_DEBUG_UART_BOARD_INIT -static void board_uart0_hw_init (void ) -{ -atmel_pio4_set_c_periph (AT91_PIO_PORTB ,26 ,ATMEL_PIO_PUEN_MASK );/* URXD0 */ -atmel_pio4_set_c_periph (AT91_PIO_PORTB ,27 ,0 );/* UTXD0 */ -} -void board_debug_uart_init (void ) { -board_uart0_hw_init (); }#endif
Signature · 1 line
2.34.1
Quoted 10 lines · Fabio Estevam
From: Fabio Estevam <[email protected]>
Currently, there are always some spurious characters showing up right before SPL banner is shown:
RomBOOT
���=������
U-Boot SPL 2025.04-rc5-00023-g9ed4e2c45f25 (Apr 03 2025 - 23:23:17 -0300) Trying to boot from MMC1
Quoted 7 lines
<
debug_uart>
...The reason for the spurious characters is that the UART0 clock is being enabled too early, prior to adjusting its correct frequency.
Fix the problem by removing the early UART0 clock enablement in C code.
I am not convinced.
The purpose of this early UART would be to get output *before* the DM code would enable clocks, pins, etc, for exactly that purpose, early debug output.
So to remove all this and make the initialization rely on the SPL DM code would defeat the purpose.
Quoted 26 lines
SPL DM code will adjust the correct UART0 clock frequency and then enable it at the appropriate time.
Signed-off-by Fabio Estevam <[email protected]>---
Changes since v1:
- Make the minimal change to fix the spurious characters. Improve the explanation in the commit log.board/atmel/sama5d27_wlsom1_ek/sama5d27_wlsom1_ek.c | 2 -- 1 file changed, 2 deletions(-)board/atmel/sama5d27_wlsom1_ek/sama5d27_wlsom1_ek.c +0
-2@@ -45,8 +45,6 @@ static void board_uart0_hw_init(void) {atmel_pio4_set_c_periph (AT91_PIO_PORTB ,26 ,ATMEL_PIO_PUEN_MASK );/* URXD0 */ atmel_pio4_set_c_periph (AT91_PIO_PORTB ,27 ,0 );/* UTXD0 */ - -at91_periph_clk_enable (ATMEL_ID_UART0 ); }void board_debug_uart_init (void )
Quoted 6 lines · Eugen Hristev
I am not convinced.
The purpose of this early UART would be to get output *before* the DM code would enable clocks, pins, etc, for exactly that purpose, early debug output.
So to remove all this and make the initialization rely on the SPL DM code would defeat the purpose.
Ok, but if we have CONFIG_DEBUG_UART enabled, shouldn't we then disable CONFIG_SPL_SERIAL?
The change below fixes the serial garbage shown in SPL:
configs/sama5d27_wlsom1_ek_mmc_defconfig +0 -1
@@ -16,7 +16,6 @@ CONFIG_OF_LIBFDT_OVERLAY=y CONFIG_DM_RESET=y CONFIG_SYS_MONITOR_LEN=524288 CONFIG_SPL_MMC=y -CONFIG_SPL_SERIAL=y CONFIG_SPL_DRIVERS_MISC=y CONFIG_SPL_STACK=0x218000 CONFIG_SPL_TEXT_BASE=0x200000
I suspect CONFIG_DEBUG_UART and CONFIG_SPL_SERIAL conflict with each other.
Is the change above acceptable? If not, what would you suggest?
Quoted 25 lines · Fabio Estevam
Quoted 6 lines · Eugen Hristev
I am not convinced.
The purpose of this early UART would be to get output *before* the DM code would enable clocks, pins, etc, for exactly that purpose, early debug output.
So to remove all this and make the initialization rely on the SPL DM code would defeat the purpose.Ok, but if we have
CONFIG_DEBUG_UARTenabled, shouldn't we then disableCONFIG_SPL_SERIAL?The change below fixes the serial garbage shown in SPL:
configs/sama5d27_wlsom1_ek_mmc_defconfig +0
-1@@ -16,7 +16,6 @@ CONFIG_OF_LIBFDT_OVERLAY=y CONFIG_DM_RESET=y CONFIG_SYS_MONITOR_LEN=524288 CONFIG_SPL_MMC=y -CONFIG_SPL_SERIAL=y CONFIG_SPL_DRIVERS_MISC=y CONFIG_SPL_STACK=0x218000 CONFIG_SPL_TEXT_BASE=0x200000
Quoted 1 line
I suspect
CONFIG_DEBUG_UARTandCONFIG_SPL_SERIALconflict with each other.
If they conflict, it means one of them is wrongly configured, since it's the same hardware serial used for both.
So my feeling is that there is a wrong clock config in one of the serials that causes the bad characters. Wrong clock config because it appears to have a wrong computed divider to get to the desired baud. So , disabling one of them is not really fixing the problem.
Quoted 2 lines
Is the change above acceptable? If not, what would you suggest?