boards/stm32l4: add riotboot support

[aabadie]

Contribution description

Now that #11643 is merged, this PR adds riotboot to all stm32l4 based boards. Changes provided in #11643 are compatible for all L4.

Not tested yet.

Testing procedure

• Flash riotboot and check application is working:

$ make BOARD=nucleo-l476rg -C tests/riotboot riotboot/flash term
> help
> curslotnr

• Flash an image on slot0:
  $ make BOARD=nucleo-l476rg -C tests/riotboot riotboot/flash-slot0 term

curslotnr # Should return slot 0

- Flash an image on slot1:
$ make BOARD=nucleo-l476rg -C tests/riotboot riotboot/flash-slot1 term
> curslotnr  # Should return slot 1

Issues/PRs references

Follow-up of #11643

[aabadie]

I just tested this PR on related L4 boards (except stm32l476-disco but it's the same CPU as nucleo-l476rg). I works except for nucleo-l432kc and nucleo-l433rc.
I need to find out what's different on them. Maybe I can remove them from this PR and open a follow-p once I can make them work ? @fjmolinas any thoughts?

[aabadie]

I works except for nucleo-l432kc and nucleo-l433rc.

I tested this again today but couldn't make it work on nucleo-l433rc. It looks like there's an issue when erasing/writing on the flash with these boards.
OpenOCD probably has the same issue because flashing the extended slot0 fails all the time.

I propose to dropped those 2 boards from this PR and try to fix this in follow-up PRs. @fjmolinas, is it ok for you ?

[aabadie]

I dropped the changes of nucleo-l432kc and nucleo-l433rc from this PR as it's still unclear what is the problem. Boards that work fine: nucleo-f496zg, b-l475-iot01a and stm32l476g-disco. nucleo-l452re is untested (I can't test this one)

[fjmolinas]

I'm not able to reproduce your issue for nucleo-l433rc, after flashing slot0/1:

2019-07-03 09:59:55,157 - INFO # Hello riotboot!
2019-07-03 09:59:55,160 - INFO # You are running RIOT on a(n) nucleo-l433rc board.
2019-07-03 09:59:55,163 - INFO # This board features a(n) stm32l4 MCU.
2019-07-03 09:59:55,166 - INFO # riotboot_test: running from slot 0
2019-07-03 09:59:55,170 - INFO # Image magic_number: 0x544f4952
2019-07-03 09:59:55,173 - INFO # Image Version: 0x5d1c6078
2019-07-03 09:59:55,174 - INFO # Image start address: 0x08001200
2019-07-03 09:59:55,176 - INFO # Header chksum: 0x22b17536
2019-07-03 09:59:55,177 - INFO # 
2019-07-03 10:00:06,192 - INFO #  main(): This is RIOT! (Version: 2019.07-devel-759-ge7aed-pr-11701)
2019-07-03 10:00:06,195 - INFO # Hello riotboot!
2019-07-03 10:00:06,197 - INFO # You are running RIOT on a(n) nucleo-l433rc board.
2019-07-03 10:00:06,199 - INFO # This board features a(n) stm32l4 MCU.
2019-07-03 10:00:06,202 - INFO # riotboot_test: running from slot 1
2019-07-03 10:00:06,205 - INFO # Image magic_number: 0x544f4952
2019-07-03 10:00:06,207 - INFO # Image Version: 0x5d1c6084
2019-07-03 10:00:06,210 - INFO # Image start address: 0x08020a00
2019-07-03 10:00:06,213 - INFO # Header chksum: 0x12e36d44
2019-07-03 10:00:06,213 - INFO # 

After applying #11776, tests/riotboot_flashwrite was also working for me.#11776.

I don't have nucleo-l432kc ATM to test.

[fjmolinas]

Ran successfully tests/riotboot_flashwrite on:

• nucleo-l433rc
• stm32l476g-disco

I don't have b-l475-iot01a, nucleo-l432kc, nucleo-f496zg or nucleo-l452re. @aabadie could you lend me the first 3 tomorrow?

@rdagher I think you have a nucleo-l452re, do you think you could test?

[aabadie]

After applying #11776, tests/riotboot_flashwrite was also working for me.

Are you sure no commits from #11776 were built in the firmware ? because for me it never worked.
I think nucleo-l432kc is in your office somewhere (I saw it yesterday ;) ), it's a nucleo-32 version.

[fjmolinas]

I think nucleo-l432kc is in your office somewhere (I saw it yesterday ;) ), it's a nucleo-32 version.

@aabadie You are right I though it was an f103 because it actually has one on the other side of the board (the stlink).

Anyway testing with your branch:

2019-07-03 16:04:07,823 - INFO # Hello riotboot!
2019-07-03 16:04:07,828 - INFO # You are running RIOT on a(n) nucleo-l432kc board.
2019-07-03 16:04:07,831 - INFO # This board features a(n) stm32l4 MCU.
2019-07-03 16:04:07,834 - INFO # riotboot_test: running from slot 0
2019-07-03 16:04:07,837 - INFO # Image magic_number: 0x544f4952
2019-07-03 16:04:07,840 - INFO # Image Version: 0x5d1cb5d2
2019-07-03 16:04:07,842 - INFO # Image start address: 0x08001200
2019-07-03 16:04:07,844 - INFO # Header chksum: 0x781aca90
2019-07-03 16:04:07,844 - INFO # 
> 2019-07-03 16:04:23,887 - INFO #  main(): This is RIOT! (Version: 2019.07-devel-759-ge7aed-pr-11701)
2019-07-03 16:04:23,889 - INFO # Hello riotboot!
2019-07-03 16:04:23,892 - INFO # You are running RIOT on a(n) nucleo-l432kc board.
2019-07-03 16:04:23,896 - INFO # This board features a(n) stm32l4 MCU.
2019-07-03 16:04:23,899 - INFO # riotboot_test: running from slot 1
2019-07-03 16:04:23,901 - INFO # Image magic_number: 0x544f4952
2019-07-03 16:04:23,904 - INFO # Image Version: 0x5d1cb5e6
2019-07-03 16:04:23,907 - INFO # Image start address: 0x08020a00
2019-07-03 16:04:23,909 - INFO # Header chksum: 0x686cc2a6
2019-07-03 16:04:23,910 - INFO # 

Could it be an openocd version?

Open On-Chip Debugger 0.10.0+dev-00704-gdb429c34 (2019-03-08-12:14)
Licensed under GNU GPL v2
For bug reports, read
	http://openocd.org/doc/doxygen/bugs.html

[fjmolinas]

Are you sure no commits from #11776 were built in the firmware ? because for me it never worked.

Yes, since if I remember well you where having flashing issues, and the changes in #11776 target flashpage code.

[aabadie]

Could it be an openocd version?

That could be but I rebuilt it very recently.

Open On-Chip Debugger 0.10.0+dev-00629-g46325790 (2019-06-05-14:34)
Licensed under GNU GPL v2
For bug reports, read
	http://openocd.org/doc/doxygen/bugs.html

[fjmolinas]

Ran successfully tests/riotboot_flashwrite on:

• b-l475-iot01a
• nucleo-l496zg

[aabadie]

After more investigation, I seems that my OpenOCD version was older than I thought and was missing a commit fixing support for single bank stm32l4. The commit in question is openocd-org/openocd@a4d5054.

After updating OpenOCD to latest master and checking that it contains the fix commit, I was able to flash a nucleo-l432kc.

So I'll re-add support for this board and nucleo-l433rc and add a comment regarding OpenOCD version for this 2 boards.

[aabadie]

@fjmolinas, I updated the PR following my previous comment. There's a comment in nucleo-l43x boards mentioning the potential problem with OpenOCD.

[fjmolinas]

Tested for all platforms, please squash!

[aabadie]

All green, go!