nrf52: Add riotboot support by bergzand · Pull Request #11126 · RIOT-OS/RIOT

bergzand · 2019-03-07T13:33:38Z

Contribution description

This PR attempts to add riotboot support to the nrf52 class mcu's. Unfortunately currently not compatible with the nordic_softdevice_ble package.

Testing procedure

the default example (with and without bootloader) and the tests/riotboot test should both work with this PR

Issues/PRs references

Depends on

nrf52: use cortexm.ld script when applicable nrf52: use cortexm.ld script when applicable #11127
makefiles/tools/jlink.inc.mk: use FLASHFILE makefiles/tools/jlink.inc.mk: use FLASHFILE #11130

jlink: handle flashing at IMAGE_OFFSET jlink: handle flashing at IMAGE_OFFSET #11200
Not needed ~~riotboot.mk: get variable as hex rather than dec riotboot.mk: get variable as hex rather than dec #11201~~

kaspar030 · 2019-03-07T13:38:56Z

#9095 would automate some of the testing.

kYc0o · 2019-03-07T13:53:04Z

Nice one! However, it might be better to split b15625a in another PR, so we can show how easy is to add riotboot to a new board using the corexm.ld script 😄

cladmi · 2019-03-07T14:09:41Z

Nice :)

I think we could also go for cortexm.ld with nordick_softdevice_ble, if I am correct, it would just need to define ROM_OFFSET ?= 0x1f000.

I will check this.

bergzand · 2019-03-07T14:10:04Z

@kYc0o So 1 PR refactoring to the common cortexm.ld and one PR (this one) adding the riotboot feature?

kYc0o · 2019-03-07T14:11:04Z

Yep! Basically just the commit I pointed out in another PR. Thanks @bergzand !

cladmi · 2019-03-07T14:13:27Z

If we handle ROM_OFFSET in riotboot.mk and cortexm_common/Makefile.include we could even work with nordick_softdevice_ble I think… But it is another PR :)

cladmi · 2019-03-07T14:19:28Z

It should also either add some information about the hack for flashing with JLink as now it is said that HEXFILE is here to work for edbg.

Or if we can get this one in #11112 I can update jlink to use FLASHFILE directly.

bergzand · 2019-03-07T14:21:13Z

@kYc0o See #11127

bergzand · 2019-03-07T14:27:51Z

It should also either add some information about the hack for flashing with JLink as now it is said that HEXFILE is here to work for edbg.

I could use a bit more context here. Is this still in scope of this PR or is this a note for your follow up PR?

cladmi · 2019-03-07T14:32:19Z

Sorry, I meant that in the following files there are infos about the fact that currently doing make flash or make riotboot/flash-slot0 needs hack to work.

RIOT/tests/riotboot/Makefile

Line 31 in 87d0757

# edbg

RIOT/makefiles/boot/riotboot.mk

Line 114 in 87d0757

# Flashing rule for edbg to flash combined/extended binaries

And they do not mention JLink, and work because currently also jlink.inc.mk does HEXFILE = $(BINFILE) and flash with HEXFILE.

A fix for this is to either update the mentions to also say "it is also a hack for jlink" or rush to have FLASHFILE used in Jlink. I can do the PR for this one and see.

cladmi · 2019-03-07T14:42:23Z

I created the FLASHFILE pull request. I will rush to add tests for it.
Better to fix it properly if it is working than updating the work around :)

The riotboot feature is not compatible with the nordic_softdevice_ble package at the moment

bergzand · 2019-03-14T11:04:35Z

Rebased

cladmi · 2019-03-14T11:35:55Z

I will re-test this :)

bergzand · 2019-03-14T11:37:13Z

I will re-test this :)

Thanks!

cladmi · 2019-03-14T12:35:12Z

riotboot/flash-slot1 is not working due to the way flashing offset is handled.

JLink.sh only relies on FLASH_ADDR for the address to flash, which is described as offset to flash code into ROM memory in makefiles/vars.inc.mk, so maybe not really an address.

For flashing with an offset, I am using IMAGE_OFFSET in makefiles/boot/riotboot.mk as was done by openocd and now edbg, so I think this should now be harmonized.

I can put a hack to also take IMAGE_OFFSET into account for JLink.

cladmi · 2019-03-14T12:48:48Z

With this patch, jlink.sh correctly gets the address for the riotboot/flash-slot-* but it looks like it is not being flashed...

diff --git a/makefiles/boot/riotboot.mk b/makefiles/boot/riotboot.mk
index 0abe2c1d7..2f7c7e1fc 100644
--- a/makefiles/boot/riotboot.mk
+++ b/makefiles/boot/riotboot.mk
@@ -128,6 +128,7 @@ riotboot/flash-extended-slot0: $(RIOTBOOT_EXTENDED_BIN) $(FLASHDEPS)

 # Flashing rule for slot 0
 riotboot/flash-slot0: export IMAGE_OFFSET=$(SLOT0_OFFSET)
+riotboot/flash-slot0: export FLASH_ADDR:=$(shell echo $$(($${FLASH_ADDR:-0}+$(SLOT0_OFFSET))))
 # Flashing rule for edbg to flash only slot 0
 riotboot/flash-slot0: HEXFILE=$(SLOT0_RIOT_BIN)
 # openocd
@@ -138,6 +139,7 @@ riotboot/flash-slot0: $(SLOT0_RIOT_BIN) $(FLASHDEPS)

 # Flashing rule for slot 1
 riotboot/flash-slot1: export IMAGE_OFFSET=$(SLOT1_OFFSET)
+riotboot/flash-slot1: export FLASH_ADDR:=$(shell echo $$(($${FLASH_ADDR:-0}+$(SLOT1_OFFSET))))
 # Flashing rule for edbg to flash only slot 1
 riotboot/flash-slot1: HEXFILE=$(SLOT1_RIOT_BIN)
 # openocd

The commands I ran:

# Flash using `flash` target
# EDITED because APP_VER is build date...
git clean -xdf tests/riotboot ; APP_VER=0 RIOT_CI_BUILD=1 BOARD=nrf52dk make -C tests/riotboot flash test 
# Correctly tests and uses slot0 with "version == 0"

# flash using `flash-slot0` target and a bigger image version
git clean -xdf tests/riotboot ; APP_VER=1 RIOT_CI_BUILD=1 BOARD=nrf52dk make -C tests/riotboot riotboot/flash-slot0 test
# ERROR still running slot0 version 0

# flash using `flash-slot1` target and an even bigger image version
git clean -xdf tests/riotboot; APP_VER=3 BOARD=nrf52dk make -C tests/riotboot clean riotboot/flash-slot1 test
# ERROR still running slot0 version 0

cladmi · 2019-03-14T12:57:03Z

When flashing I get this error:

Downloading file [/home/harter/work/git/RIOT/tests/riotboot/bin/nrf52dk/tests_riotboot-slot1.riot.bin]...
Unspecified error -1

But flash does not exit with an error. I already noticed this but should prepare a pull request for it:
cedba94

If I flash using PROGRAMMER=openocd it correctly updates the slots.
So it looks like completely jlink related.

bergzand · 2019-03-14T14:53:14Z

Ah, that explains. I've only tested this with PROGRAMMER=openocd so far. Better block this until this also works with the jlink software, as that is the default PROGRAMMER with these boards at the moment.

danpetry · 2019-03-18T10:59:14Z

There were a couple of things going on here. #11200 implements handling of IMAGE_OFFSET in jlink. #11201 implements correct calculation (in hex) of the slot 1 image offset.

They're somewhat interrelated but suggest merging #11200 first, and only testing slot 0. #11201 is needed for slot 1 to work correctly, so testing of slot 1 can be done there.

cladmi · 2019-03-19T17:51:58Z

The updated procedure from #11126 (comment) works when #11200 is applied alone. I will finish the review over there.

cladmi · 2019-03-20T18:29:54Z

I rebased on master now that #11200 is merged and ran the tests from #11126 (comment) which work with both PROGRAMMER=jlink and PROGRAMMER=openocd.

cladmi

Reviewed and tested with the different riotboot/flash- targets with both jlink and openocd.

You can merge when CI gives the greenlight.

bergzand · 2019-03-20T18:34:23Z

I rebased on master now that #11200 is merged and ran the tests from #11126 (comment) which work with both PROGRAMMER=jlink and PROGRAMMER=openocd.

Awesome, thanks for the effort you guys put into this @danpetry @cladmi!

danpetry · 2019-03-21T12:33:13Z

Np!

Bluetooth OTA upgrade here we come :)

brent7984 · 2019-05-23T13:32:20Z

Hey guys, does the cc2538dk have riotboot support. When i try to compile i receive this error, "/bin/sh: 1: Illegal number: 0xfffffffffffff800
make: *** [/home/brenton/Downloads/RIOT-pr-nrf52-riotboot/makefiles/boot/riotboot.mk:39: /home/brenton/Downloads/RIOT-pr-nrf52-riotboot/tests/riotboot/bin/cc2538dk/tests_riotboot-slot0.elf] Error 2"
Does anyone have any idea on how i can fix this issue?

emmanuelsearch · 2019-05-23T14:09:04Z

Hi @brent7984 no the cc2538 is not supported yet by riotboot.
Since it is a Cortex-M3 there is no conceptual hurdle to support it, though.
Are you interested in a port?

brent7984 · 2019-05-24T06:10:05Z

Hi there, I am interested in a port, could you please point me towards what needs to be done for a port. I really want to attempt this..

fjmolinas · 2019-07-24T07:27:03Z

@cladmi are you still able to flash the second slot with Jlink? I'm currently unable to do it (still able with openocd)

ROMTbl[0] @ E00FF000
ROMTbl[0][0]: E000E000, CID: B105E00D, PID: 000BB00C SCS-M7
ROMTbl[0][1]: E0001000, CID: B105E00D, PID: 003BB002 DWT
ROMTbl[0][2]: E0002000, CID: B105E00D, PID: 002BB003 FPB
ROMTbl[0][3]: E0000000, CID: B105E00D, PID: 003BB001 ITM
ROMTbl[0][4]: E0040000, CID: B105900D, PID: 000BB9A1 TPIU
ROMTbl[0][5]: E0041000, CID: B105900D, PID: 000BB925 ETM
Cortex-M4 identified.
Halting CPU for downloading file.
Downloading file [/home/francisco/workspace/RIOT/tests/riotboot/bin/nrf52840dk/tests_riotboot-slot1.1563953199.riot.bin]...
Unspecified error -1

Script processing completed.

cladmi · 2019-07-24T09:52:51Z

The command still works for me without error:

(filesize 12560)

BOARD=nrf52dk make -C tests/riotboot riotboot/flash-slot1


### Flashing Target ###
### Flashing at base address 0x0 with offset 266240 ###
...

Cortex-M4 identified.
Halting CPU for downloading file.
Downloading file [/home/harter/work/git/RIOT/tests/riotboot/bin/nrf52dk/tests_riotboot-slot1.0.riot.bin]...
Comparing flash   [100%] Done.
Erasing flash     [100%] Done.
Programming flash [100%] Done.
Verifying flash   [100%] Done.
J-Link: Flash download: Bank 0 @ 0x00000000: 1 range affected (16384 bytes)
J-Link: Flash download: Total time needed: 0.290s (Prepare: 0.042s, Compare: 0.004s, Erase: 0.000s, Program: 0.234s, Verify: 0.000s, Restore: 0.008s)
O.K.

Reset delay: 0 ms
Reset type NORMAL: Resets core & peripherals via SYSRESETREQ & VECTRESET bit.
Reset: Halt core after reset via DEMCR.VC_CORERESET.
Reset: Reset device via AIRCR.SYSRESETREQ.



Script processing completed.

make: Leaving directory '/home/harter/work/git/RIOT/tests/riotboot'

However it already happened to me that reset with JLink was doing nothing, and that it was working with openocd.
On my test machine I am using openocd for the board with #11470 too.

fjmolinas · 2019-07-24T10:03:53Z

@cladmi do you have an nrf52840dk? It is the one I'm using and which is failing, the exact command I'm running:

make -C tests/riotboot BOARD=nrf52840dk -j3 riotboot/flash-slot1

However it already happened to me that reset with JLink was doing nothing, and that it was working with openocd.

Flashing slot0 does succeed, but not flashing slot0-extended, so it doesn't seem to be a reset problem to me.

cladmi · 2019-07-24T11:09:51Z

Oh no I don't have that one.

bergzand added Type: enhancement The issue suggests enhanceable parts / The PR enhances parts of the codebase / documentation Area: cpu Area: CPU/MCU ports labels Mar 7, 2019

bergzand requested review from cladmi, haukepetersen and kYc0o March 7, 2019 13:33

bergzand mentioned this pull request Mar 7, 2019

nrf52: use cortexm.ld script when applicable #11127

Merged

cladmi mentioned this pull request Mar 7, 2019

makefiles/tools/jlink.inc.mk: use FLASHFILE #11130

Merged

3 tasks

bergzand added the State: waiting for other PR State: The PR requires another PR to be merged first label Mar 7, 2019

cladmi added this to the Release 2019.04 milestone Mar 11, 2019

nrf52: include riotboot feature when possible

005710e

The riotboot feature is not compatible with the nordic_softdevice_ble package at the moment

bergzand force-pushed the pr/nrf52/riotboot branch from a937ca5 to 005710e Compare March 14, 2019 11:03

bergzand added CI: ready for build If set, CI server will compile all applications for all available boards for the labeled PR and removed State: waiting for other PR State: The PR requires another PR to be merged first labels Mar 14, 2019

cladmi added the CI: run tests If set, CI server will run tests on hardware for the labeled PR label Mar 14, 2019

kaspar030 added CI: ready for build If set, CI server will compile all applications for all available boards for the labeled PR and removed CI: ready for build If set, CI server will compile all applications for all available boards for the labeled PR labels Mar 14, 2019

This was referenced Mar 18, 2019

jlink: handle flashing at IMAGE_OFFSET #11200

Merged

riotboot.mk: get variable as hex rather than dec #11201

Closed

danpetry added the State: waiting for other PR State: The PR requires another PR to be merged first label Mar 18, 2019

cladmi approved these changes Mar 20, 2019

View reviewed changes

cladmi merged commit 2523c1b into RIOT-OS:master Mar 20, 2019

Conversation

bergzand commented Mar 7, 2019 • edited by cladmi Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Contribution description

Testing procedure

Issues/PRs references

Uh oh!

kaspar030 commented Mar 7, 2019

Uh oh!

kYc0o commented Mar 7, 2019

Uh oh!

cladmi commented Mar 7, 2019 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

bergzand commented Mar 7, 2019

Uh oh!

kYc0o commented Mar 7, 2019

Uh oh!

cladmi commented Mar 7, 2019

Uh oh!

cladmi commented Mar 7, 2019

Uh oh!

bergzand commented Mar 7, 2019

Uh oh!

bergzand commented Mar 7, 2019

Uh oh!

cladmi commented Mar 7, 2019 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

cladmi commented Mar 7, 2019

Uh oh!

bergzand commented Mar 14, 2019

Uh oh!

cladmi commented Mar 14, 2019

Uh oh!

bergzand commented Mar 14, 2019

Uh oh!

cladmi commented Mar 14, 2019

Uh oh!

cladmi commented Mar 14, 2019 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

cladmi commented Mar 14, 2019

Uh oh!

bergzand commented Mar 14, 2019

Uh oh!

danpetry commented Mar 18, 2019

Uh oh!

cladmi commented Mar 19, 2019

Uh oh!

cladmi commented Mar 20, 2019

Uh oh!

cladmi left a comment

Choose a reason for hiding this comment

Uh oh!

bergzand commented Mar 20, 2019

Uh oh!

danpetry commented Mar 21, 2019

Uh oh!

brent7984 commented May 23, 2019

Uh oh!

emmanuelsearch commented May 23, 2019

Uh oh!

brent7984 commented May 24, 2019

Uh oh!

fjmolinas commented Jul 24, 2019

Uh oh!

cladmi commented Jul 24, 2019

Uh oh!

fjmolinas commented Jul 24, 2019

Uh oh!

cladmi commented Jul 24, 2019

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

bergzand commented Mar 7, 2019 •

edited by cladmi

Loading

cladmi commented Mar 7, 2019 •

edited

Loading

cladmi commented Mar 7, 2019 •

edited

Loading

cladmi commented Mar 14, 2019 •

edited

Loading