Programming flash in RP2350

[highjeans]

In microzig v0.15.1, does programming flash on an RP2350 work? Currently I try to erase a sector from flash using a similar piece of code from the flash_program example, and it panics with "missing coprocessor usage = true". What can I do about this? Looking at the pico sdk, the functions for flash are the exact same for both the RP2040 and the RP2350, so I assumed the current code for erasing flash would work.

Edited: Turns out if all the bits of the flash are set to 1 already, it will crash when it tries to erase it. However, I can't seem to program it though as it crashes with "unaligned memory access = true" when writing anything from (seemingly) all 0xFFs.

[tact1m4n3]

I think rp2xxx.flash isn't yet ported to rp2350 (only works on rp2040). A clue for this is that the flash example is marked as rp2040 only still.

[highjeans]

The thing is that reading does work fine, and programming also did work yesterday morning, until I added some more code and then it stopped working. I'll see if I can produce a minimum reproducible example.

[highjeans]

Heres the minimum reproducible example I have so far (it includes uart debug logs since otherwise I won't really know whats happening):

const std = @import("std");
const microzig = @import("microzig");
const rp2xxx = microzig.hal;
const flash = rp2xxx.flash;
const time = rp2xxx.time;
const log = std.log;

pub fn panic(msg: []const u8, _: ?*std.builtin.StackTrace, _: ?usize) noreturn {
    while (true) {
        log.debug("panic occurred with message '{s}'", .{msg});
        time.sleep_ms(1000);
    }
}

const FLASH_TARGET_OFFSET: u32 = 256 * 1024;
const STORAGE_PTR: *const volatile u8 = @ptrFromInt(flash.XIP_BASE + FLASH_TARGET_OFFSET);

fn read() u3 {
    const raw = STORAGE_PTR.*;

    return @truncate(raw);
}

fn write(value: u3) void {
    var page_data: [flash.PAGE_SIZE]u8 = [_]u8{0xFF} ** flash.PAGE_SIZE;
    page_data[0] ^= 0b111;
    page_data[0] += value;

    // log.debug("page_data: {b64}", .{page_data});

    flash.range_program(FLASH_TARGET_OFFSET, &page_data);
}

const uart = rp2xxx.uart.instance.num(0);

const pin_config = rp2xxx.pins.GlobalConfiguration{
    .GPIO0 = .{
        .function = .UART0_TX,
        .name = "uart_tx",
    },
};

pub const microzig_options = microzig.Options{
    .log_level = .debug,
    .logFn = rp2xxx.uart.log,
};

pub fn main() !void {
    _ = pin_config.apply();
    uart.apply(.{ .clock_config = rp2xxx.clock_config });
    rp2xxx.uart.init_logger(uart);

    log.debug("Initialised logger", .{});

    log.debug("Starting loop", .{});

    while (true) {
        write(3);
        log.debug("wrote 3", .{});
        time.sleep_ms(5000);
        log.debug("read {d}", .{read()});
        time.sleep_ms(5000);
    }
}

This fails with the following:

[0.012867] error (cortex_m_debug): Bus Fault:
[0.022479] error (cortex_m_debug):   context                         =  r0:0x2007FB68  r1:0xFFFFFFF9  r2:0x00000000    r3:0x00000000
[0.037145] error (cortex_m_debug):                                     r12:0x5720A400  lr:0xFFFFFFF8  ra:0x78028400  xpsr:0x20000379
[0.050542] error (cortex_m_debug):   instruction bus error             = false
[0.058762] error (cortex_m_debug):   precice data bus error            = true
[0.067299] error (cortex_m_debug):   imprecice data bus error          = false
[0.075946] error (cortex_m_debug):   unstacking exception error        = false
[0.084891] error (cortex_m_debug):   exception stacking error          = false
[0.093590] error (cortex_m_debug):   fpu lazy state preservation fault = false
[0.102045] error (cortex_m_debug):   busfault address register valid   = true
[0.110179] error (cortex_m_debug):     busfault address register = 0x6850A217

I understand that you think rp2xxx.flash isn't ported for the rp2350, but I've manually looked through both the code that microzig has (at least for v0.15.1) and the pico sdk (https://github.com/raspberrypi/pico-sdk) and they both match, hence me asking here since I believe it wouldn't be an issue with the pico sdk. I will double check this though just in case.

[tact1m4n3]

After looking a bit into the problem and testing your example I found the issue. I made a patch for this (#934). Can you please test if this solves your problem?

[highjeans]

Absolute legend!!! Thanks!!!!