Bare metal, Assembly language • Re: pi 4B assembly UART

Hi rgz,

I didn't dive too deep into your issue, but instead put something together to achieve what you were trying to do. Perhaps by comparing mine vs yours, you'll work out what was wrong.

I created the following uart.s file:

Code:

.global _start.align 2_start:  mrs     x0, mpidr_el1                           // x0 = Multiprocessor Affinity Register value.  and     x0, x0, #0x3                            // x0 = core number.  cbz     x0, 2f                                  // Put all cores except core 0 to sleep.1:  wfi  b 1b2:# ------------------------------------------------------------------------------# Initialise the Mini UART interface for logging over serial port.# Note, this is Broadcomm's own UART, not the ARM licenced UART interface.# ------------------------------------------------------------------------------  ldr     x1, =0xfe215000                         // x1 = aux_base  ldr     w2, [x1, #0x4]                          // w2 = [AUX_ENABLES] (Auxiliary enables)  orr     w2, w2, #1  str     w2, [x1, #0x4]                          //   [AUX_ENABLES] |= 0x00000001 => Enable Mini UART.  str     wzr, [x1, #0x44]                        //   [AUX_MU_IER] = 0x00000000 => Disable Mini UART interrupts.  str     wzr, [x1, #0x60]                        //   [AUX_MU_CNTL] = 0x00000000 => Disable Mini UART Tx/Rx  mov     w2, #0x6                                // w2 = 6  str     w2, [x1, #0x48]                         //   [AUX_MU_IIR] = 0x00000006 => Mini UART clear Tx, Rx FIFOs  mov     w3, #0x3                                // w3 = 3  str     w3, [x1, #0x4c]                         //   [AUX_MU_LCR] = 0x00000003 => Mini UART in 8-bit mode.  str     wzr, [x1, #0x50]                        //   [AUX_MU_MCR] = 0x00000000 => Set UART1_RTS line high.  mov     w2, 0x0000021d  str     w2, [x1, #0x68]                         //   [AUX_MU_BAUD] = 0x0000021d                                                  //         => baudrate = system_clock_freq/(8*([AUX_MU_BAUD]+1))                                                  //                       (as close to 115200 as possible)  ldr     x4, =0xfe200000                         // x4 = [gpio_base] = 0xfe200000  ldr     w2, [x4, #0x4]                          // w2 = [GPFSEL1]  and     w2, w2, #0xfffc0fff                     // Unset bits 12, 13, 14 (FSEL14 => GPIO Pin 14 is an input).                                                  // Unset bits 15, 16, 17 (FSEL15 => GPIO Pin 15 is an input).  orr     w2, w2, #0x00002000                     // Set bit 13 (FSEL14 => GPIO Pin 14 takes alternative function 5).  orr     w2, w2, #0x00010000                     // Set bit 16 (FSEL15 => GPIO Pin 15 takes alternative function 5).  str     w2, [x4, #0x4]                          //   [GPFSEL1] = updated value => Enable UART 1.  str     wzr, [x4, #0x94]                        //   [GPPUD] = 0x00000000 => GPIO Pull up/down = OFF  mov     x5, #0x96                               // Wait 150 instruction cycles (as stipulated by datasheet).  3:    subs    x5, x5, #0x1                          // x0 -= 1    b.ne    3b                                    // Repeat until x0 == 0.  mov     w2, #0xc000                             // w2 = 2^14 + 2^15  str     w2, [x4, #0x98]                         //   [GPPUDCLK0] = 0x0000c000 => Control signal to lines 14, 15.  mov     x0, #0x96                               // Wait 150 instruction cycles (as stipulated by datasheet).  4:    subs    x0, x0, #0x1                          // x0 -= 1    b.ne    4b                                    // Repeat until x0 == 0.  str     wzr, [x4, #0x98]                        //   [GPPUDCLK0] = 0x00000000 => Remove control signal to lines 14, 15.  str     w3, [x1, #0x60]                         //   [AUX_MU_CNTL] = 0x00000003 => Enable Mini UART Tx/Rx# ------------------------------------------------------------------------------# Send 'ABC' over Mini UART# ------------------------------------------------------------------------------  mov     w0, #0x41                               // start char 'A'  mov     w3, #0x43                               // end char 'C'  5:    6:      ldr     w2, [x1, #0x54]                     // w2 = [AUX_MU_LSR]      tbz     x2, #5, 6b                          // Repeat last statement until bit 5 is set.    strb    w0, [x1, #0x40]                       //   [AUX_MU_IO] = w0    add      w0, w0, #1    cmp      w0, w3    b.le       5b  b 1b

Then I built it as follows:

Code:

$ aarch64-none-elf-as -o uart.o uart.s$ aarch64-none-elf-ld --no-warn-rwx-segments -N -Ttext=0x80000 -o uart.elf uart.o$ aarch64-none-elf-objcopy --set-start=0x80000 uart.elf -O binary uart.img

Note, the kernel image I created is called uart.img. Then I created the following config.txt:

Code:

# Stop start.elf from filling in ATAGS (memory from 0x100) before launching the# kernel.disable_commandline_tags=1# Disable initial turbo since it affects Mini UART timing.# See https://github.com/raspberrypi/firmware/issues/1945initial_turbo=0# Dynamic frequency clocking minimum value needed for Mini UART to function correctly.core_freq_min=500# Ensure the bootloader loads a 64-bit kernel (not entirely necessary but a# good safety check).arm_64bit=1# Kernel image to loadkernel=uart.img

I deployed this to my rasbperry pi 400, and it worked. Hopefully it should for you too!

One note: in your code, I see you are setting x29 and x30 to peripheral addresses. This may work, but is somewhat unconventional. Normally x29 is used as the frame pointer, and x30 is the link register. When you call a function with a bl instruction, x30 will be set to the return address, so that ret instruction functions correctly. So as soon as you execute a bl instruction, the x30 register will be overwritten. For this reason, you might prefer to use different registers for storing those peripheral addresses.

Statistics: Posted by rgz — Fri Mar 28, 2025 12:03 pm