Reading internal flash triggers HardFault

[HaoboGu]

Hello, I'm using stm32h7b0 mcu and trying to emulate eeprom using mcu's flash and stm32h7xx-hal::flash. Here is my testing code for flash write/read:

        let (mut flash, _) = dp.FLASH.split();
        let mut bytes = [8u8;16];
        const FLASH_SECTOR_15_ADDR: u32 = 15 * 8192;
        let sector = flash.sector(FLASH_SECTOR_15_ADDR as usize).unwrap();
        info!("{}", sector.offset);
        if sector.contains(FLASH_SECTOR_15_ADDR as usize) {
            embedded_storage::nor_flash::ReadNorFlash::read(&mut flash, FLASH_SECTOR_15_ADDR, &mut bytes).unwrap();
        }

But the read function fails and the program jumps to HardFault

Here is the callstack:

[richardeoin]

Hi! Thanks for reporting. The stm32h7b0 is a device with only 128KB of flash, divided into 16 sectors of 8KB each. You're trying to read from the last of those sectors. In theory that should be possible.

Could you confirm:

• That you're using the --features=stm32h7b0 feature flag to build your test code.
• If possible, the value returned by stm32h7xx_hal::signature::FlashSize::bytes() on your stm32h7b0.

[HaoboGu]

@richardeoin

I checked my config, here is how i introduce stm32h7xx-hal in Cargo.toml:

stm32h7xx-hal = { version = "0.14.0", features = [
    "stm32h7b0",
    "rt",
    "log-rtt",
    "usb_hs",
] }

The value returned by stm32h7xx_hal::signature::FlashSize::bytes() is 131072, aka 128kb:

I've noticed there is a new version of stm32h7xx-hal published at crates.io, I also tried it, but my code breaks at:

note: required by a bound in `gpio::convert::<impl stm32h7xx_hal::gpio::Pin<P, N, MODE>>::into_alternate`
   --> /Users/haobogu/.cargo/registry/src/rsproxy.cn-0dccff568467c15b/stm32h7xx-hal-0.15.0/src/gpio/convert.rs:151:15
    |
147 |     pub fn into_alternate<const A: u8>(
    |            -------------- required by a bound in this associated function
...
151 |         Self: marker::IntoAf<A>,
    |               ^^^^^^^^^^^^^^^^^ required by this bound in `gpio::convert::<impl Pin<P, N, MODE>>::into_alternate`

error[E0277]: the trait bound `stm32h7xx_hal::gpio::Pin<'A', 12>: gpio::marker::IntoAf<10>` is not satisfied
   --> boards/stm32h7/src/main.rs:83:33
    |
83  |         let usb_dp = gpioa.pa12.into_alternate();
    |                                 ^^^^^^^^^^^^^^ the trait `gpio::marker::IntoAf<10>` is not implemented for `stm32h7xx_hal::gpio::Pin<'A', 12>`

[HaoboGu]

I tried to erase the sector using:

let (mut flash, _) = dp.FLASH.split();
let mut unlocked = flash.unlocked();
unlocked.erase_sector(15).unwrap();

The program reported the following error:

panicked at boards/stm32h7/src/main.rs:101:35:
called `Result::unwrap()` on an `Err` value: EccDoubleDetection

It seems that some of my write operations didn't complete, and then an ECC error occured.

@richardeoin Does stm32h7xx-hal crate provide any methods to fix flash ECC error?

[mlamoore]

Assuming the flash is fine and just got interrupted while writing or something, shouldn't a flash erase and write fix the ECC error?

…

On Thu, Oct 12, 2023, 8:20 AM HaoboGu ***@***.***> wrote: I tried to erase the sector using: let (mut flash, _) = dp.FLASH.split();let mut unlocked = flash.unlocked(); unlocked.erase_sector(15).unwrap(); The program reported the following error: panicked at boards/stm32h7/src/main.rs:101:35: called `Result::unwrap()` on an `Err` value: EccDoubleDetection It seems that some of my write operations didn't complete, and then an ECC error occured. @richardeoin <https://github.com/richardeoin> Does stm32h7xx-hal crate provide any methods to fix flash ECC error? — Reply to this email directly, view it on GitHub <#463 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAUS5FMRP2XU3WKH3HHWHGDX67VCRANCNFSM6AAAAAA534WUFU> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[HaoboGu]

I've reviewed my code. I use the following function to write the flash:

fn write_record(&mut self, record: EepromRecord) {
        match self
            .storage
            .write(STORAGE_START_ADDR + self.pos, &record.to_bytes())
        {
            Ok(_) => self.pos += 4,
            Err(_) => error!("Failed to write record to storage"),
        }
    }

after each write, the next write addr will be addr + 4. Is this right way to use the crate?

I believe I didn't write much data(about 300 bytes). So I don't know is the flash still fine. Erasing the sector with unlocked.erase_sector(15).unwrap();, reports EccDoubleDetection error. Does it mean the flash has been broken?

[richardeoin]

Your report about USB is a separate issue. I've posted a fix in #464

[HaoboGu]

Erasing the sector fixed the ECC error, closing this issue