cpu/stm32f4: add riotboot requirements

[fjmolinas]

Contribution description

This PR adds riotboot for stm32f4. Since F4 works with sectors for riotboot to work each slot must start at the beginning of a slotsector. For stm32f4 the smallest slotsector is of length 16kB, therefore that must be the length of the bootloader and the same amount must be taken from the second half of the flash to optimize flash usage.

This PR is part of 3, since just adding riotboot doesn't make much sense if it can't perform ota. Although it doesn't depend on them it is highly related to:

• Support flashpage: cpu/stm32f7/4/2: add flashpage and flashpage_raw #11681
• Don't erase fist page at un update finish: riotboot/flashwrite: use flashpage_raw when available #11683

Testing procedure

It has been tested on the supported boards: nucleo-f446re and stm32f429i-disc1.

1.- Test riotboot:

BOARD=nucleo-f446r FEATURES_REQUIRED+=riotboot make -C tests/xtimer_usleep riotboot/flash-extended-slot0 term

BOARD=nucleo-f446rFEATURES_REQUIRED+=riotboot make -C tests/xtimer_usleep riotboot/flash-slot1 term

In both cases it should run without issues.

2.- Since the sector division is important when updating the slots by writing to flash directly it would be ideal to reba se on top off flashpage: #11681 and #11683

and run tests/riotboot_flashwrite

The following patch can be applied to the test Makefile

diff --git a/tests/riotboot_flashwrite/Makefile b/tests/riotboot_flashwrite/Makefile
index 32f127d70..d3c9f20b4 100644
--- a/tests/riotboot_flashwrite/Makefile
+++ b/tests/riotboot_flashwrite/Makefile
@@ -39,11 +39,17 @@ LOW_MEMORY_BOARDS := nucleo-f334r8
 GNRC_NETIF_NUMOF := 2
 
 # uncomment these to use ethos
-#USEMODULE += ethos gnrc_uhcpc
+USEMODULE += ethos gnrc_uhcpc  stdio_uart_rx
 #
 ## ethos baudrate can be configured from make command
-#ETHOS_BAUDRATE ?= 115200
-#CFLAGS += -DETHOS_BAUDRATE=$(ETHOS_BAUDRATE) -DUSE_ETHOS_FOR_STDIO
+ETHOS_BAUDRATE ?= 115200
+CFLAGS += -DETHOS_BAUDRATE=$(ETHOS_BAUDRATE) -DUSE_ETHOS_FOR_STDIO
+
+TAP ?= tap0
+IPV6_PREFIX ?= 2001:db8::/64
+
+TERMPROG ?= sudo sh $(RIOTTOOLS)/ethos/start_network.sh
+TERMFLAGS ?= $(PORT) $(TAP) $(IPV6_PREFIX)
 
 ifneq (,$(filter $(BOARD),$(LOW_MEMORY_BOARDS)))
   $(info Using low-memory configuration for microcoap_server.)

Then to test:

• provided the node make -C tests/riotboot_flashwrite/ BOARD=stm32f429i-disc1 riotboot/flash term
• compile new firmware: make -C tests/riotboot_flashwrite/ BOARD=stm32f429i-disc1 riotboot
• launch an updated: coap-client -m post coap://[fe80::2%tap0]/flashwrite -f tests/riotboot_flashwrite/bin/stm32f429i-disc1/tests_riotboot_flashwrite-slot1.riot.bin -b 64

When the update finished reboot the node manually it should have started from a different slot.

Issues/PRs references

Related to #11681 and #11683

[aabadie]

Since F4 works with sectors for riotboot to work each slot must start at the beginning of a slot. For stm32f4 the smallest slot is of length 16kB

It seems there is a confusion between slots and sectors in this sentence. From my understanding, it should be: Since F4 works with sectors, for riotboot to work, each slot must start at the beginning of a sector. For stm32f4 the smallest sector is of length 16kB...

This PR also needs a rebase.

[aabadie]

I have comments on comments.

[fjmolinas]

@aabadie comments addressed.

[aabadie]

Code changes are good and well explained. I tested on nucleo-f446re and on nucleo-f410rb (even if the feature is not added on this one) and it worked on both.

ACK

[aabadie]

Please squash and trigger Murdock after.

[fjmolinas]

@aabadie I squashed and modified the commit message, there where some typos. I also removed:

SLOT0_OFFSET ?= $(RIOTBOOT_LEN)	
SLOT1_OFFSET ?= $(shell echo $$(($(SLOT0_OFFSET)+ $(RIOTBOOT_LEN) + $(SLOT0_LEN))))

Which was already defined for riotboot.

[aabadie]

Ok, I'll have a second look tomorrow, if I find time, just in case.

[aabadie]

I tested a second time this PR and it works. Let's go!

[fjmolinas]

@aabadie Thanks for the review!