STM32F1 Blue Pill stuck in DFU mode after upload

pfeerick · March 15, 2019, 7:20am

The core and generic stuff is basically fine as that’s just a configuration value.

The ‘hack’ to the board is something I’ll chase up as an issue → PR on the relevant github repo (I’ll link the issue later tonight when I make it)… then hopefully I can also find out why the way I thought I could do it properly and do a PR to include it doesn’t work…

So it does actually work (i.e. the board is running code fine)? I don’t have an STM32F103 handy so haven’t gone further than a compile/concept test. That was really all I was waiting on before opening an issue/PR.

ifconfig · March 15, 2019, 7:29am

Yes, as far as I can tell, all of the functionality is present. I haven’t been able to test it with a program above the 64K size limit, but that shouldn’t be an issue, right?

pfeerick · March 15, 2019, 8:55am

Fingers crossed, no… but I reserve the right to be proven wrong!

The issue is here, let’s see where this rabbit hole leads…

github.com/platformio/platform-ststm32

128k flash variants of STM32F103 (Arduino framework)

opened 08:52AM - 15 Mar 19 UTC

pfeerick

When looking at the STM32F103, I realised the 128k board variant doesn't appear …to work as the linker script is missing. Since I wanted to try to get it working, I checked how the stm32dunio core handled the 64/128k variants, and it seems that it is merely changing the FLASH line in the linker script to 128k ... https://github.com/platformio/platform-ststm32/blob/a1c78e09e4a966a4240d9ad5ab1ff6c8103e47fe/ldscripts/stm32f103x8.ld#L4 ... creating a linker file as `stm32f103x8_128k.ld` and updating the path to point to it in ... https://github.com/platformio/platform-ststm32/blob/a1c78e09e4a966a4240d9ad5ab1ff6c8103e47fe/boards/bluepill_f103c8_128k.json#L17 ... should have been enough, but instead, I get a `Warning! Cannot find linker script for the current target!` and the build fails. What am I missing? Changing the `bluepill_f103c8.json` or `genericSTM32F103CB.json` and `stm32f103x8.ld` in place seems to work just fine though, and the generated code does appear to be functioning on a test board. It's just not the appropriate way to fix it! And on a semi-related note, I see that https://github.com/platformio/platform-ststm32/blob/develop/boards/genericSTM32F103CB.json points to the version of the linker file that specified 64k of flash, whereas both the chip and the `maximum_size` are for 128k, so that is completely wrong. If I can get the bluepill 128k working, I can do PRs for both it and the STM32F103CB.

pfeerick · March 15, 2019, 11:15am

This seems to be another way to do a hacky linker script redirect, but it is still a band-aid fix as this particular case is a bug in my view.

ifconfig · March 15, 2019, 3:10pm

Sure, but as Max said, that thread is a bit hard to comprehend. At first glance it looks like that fix doesn’t make changes to the same files that we did, but I’m not sure which it is.

pfeerick · March 16, 2019, 4:31am

It’s just a different way to tackle the linker script change… instead of altering the file used by platformio by default, you specify the path to a custom version of the file via the pre_extra_script, and in this instance, it was included in the platformio project directory. That particular issue was about having a custom linker script for a different board, hence the difference in file name, but the theory of operation should still work. And would survive updates to the platform, as your linker script is actually local to your project.