sys/eepreg: use MTD by fabian18 · Pull Request #22158 · RIOT-OS/RIOT

fabian18 · 2026-03-30T18:47:26Z

Contribution description

Add an MTD to the EEPROM registry API, so that other MTD (EEPROMS) can also be used.

Extend EEPROM registry API
Create internal EEPROM MTD
Add mtd_eeprom to boards that support periph/eeprom
Adapt tests

Testing procedure

USEMODULE+=at24mac CFLAGS+=-DTEST_EEPREG_MTD="mtd_dev_get$1$" DEVELHELP=1 VERBOSE_ASSERT=1 make -C tests/sys/eepreg/ BOARD=same54-xpro flash term

2026-03-30 20:42:36,101 # START
2026-03-30 20:42:36,107 # main(): This is RIOT! (Version: 2026.04-devel-325-ge13ff-pr/eepreg_mtd)
2026-03-30 20:42:36,110 # EEPROM registry (eepreg) test routine
2026-03-30 20:42:36,121 # Testing new registry creation: reset check [SUCCESS]
2026-03-30 20:42:36,162 # Testing writing and reading entries: add write add read [SUCCESS]
2026-03-30 20:42:36,171 # Testing detection of conflicting size: add [SUCCESS]
2026-03-30 20:42:36,184 # Testing calculation of lengths: len len [SUCCESS]
2026-03-30 20:42:36,347 # Testing of successful data move after rm: rm read data [SUCCESS]
2026-03-30 20:42:36,373 # Testing of free space change after write: free add free [SUCCESS]
2026-03-30 20:42:36,384 # Testing of iteration over registry: iter bar foo [SUCCESS]
2026-03-30 20:42:36,385 # Tests complete!
2026-03-30 20:42:36,392 # { "threads": [{ "name": "main", "stack_size": 1536, "stack_used": 488 }]}

Issues/PRs references

crasbe

The static test has some comments too and why is mtd_eeprom.h sometimes conditionally and sometimes unconditionally included? 🤔

riot-ci · 2026-03-30T20:19:19Z

Murdock results

❌ FAILED

82dcd5b fixup! fixup! boards: expose default EEPROM MTD

Build failures (21)

Application	Target	Toolchain	Runtime (s)	Worker
tests/drivers/mtd_at24cxxx	native32	gnu	0.81	mobi6
tests/drivers/mtd_at24cxxx	native64	gnu	0.78	mobi6
tests/drivers/mtd_at24cxxx	native64	llvm	0.77	mobi7
tests/drivers/mtd_at24cxxx	native32	llvm	0.87	mobi6
tests/net/nanocoap_cli	native64	llvm	1.28	mobi7
tests/net/nanocoap_cli	native32	gnu	1.69	mobi6
tests/net/nanocoap_cli	native64	gnu	1.66	mobi6
tests/net/nanocoap_cli	native32	llvm	7.06	mobi7
tests/pkg/fatfs	native32	gnu	1.10	mobi6
tests/pkg/fatfs	native64	gnu	1.07	mobi6
tests/pkg/fatfs	native32	llvm	1.13	mobi6
tests/pkg/fatfs	native64	llvm	3.13	skyleaf
tests/pkg/spiffs	native64	gnu	0.78	mobi7
tests/pkg/spiffs	native32	gnu	1.18	mobi6
tests/pkg/spiffs	native64	llvm	1.23	mobi6
tests/pkg/spiffs	native32	llvm	3.06	skyleaf
tests/sys/shell_scripting	native32	gnu	0.91	mobi6
tests/sys/shell_scripting	native64	gnu	0.91	mobi6
tests/sys/shell_scripting	native64	llvm	0.75	mobi7
tests/sys/shell_scripting	native32	llvm	0.99	mobi6

and 74 more build failures...

Artifacts

crasbe · 2026-04-14T11:38:10Z

The build for arduino-leonardo currently fails:

buechse@skyleaf:~/RIOTstuff/riot-vanilla/RIOT/tests/drivers/mtd_eeprom$ BUILD_IN_DOCKER=1 BOARD=arduino-leonardo make
Launching build container using image "docker.io/riot/riotbuild@sha256:08fa7da2c702ac4db7cf57c23fc46c1971f3bffc4a6eff129793f853ec808736".
docker run --rm --tty --user $(id -u) --platform linux/amd64 -v '/home/buechse/RIOTstuff/riot-vanilla/RIOT:/data/riotbuild/riotbase:delegated' -v '/home/buechse/.cargo/registry:/data/riotbuild/.cargo/registry:delegated' -v '/home/buechse/.cargo/git:/data/riotbuild/.cargo/git:delegated' -e 'TZ=Europe/Berlin' -e 'RIOTBASE=/data/riotbuild/riotbase' -e 'CCACHE_BASEDIR=/data/riotbuild/riotbase' -e 'BUILD_DIR=/data/riotbuild/riotbase/build' -e 'BUILD_IN_DOCKER=0' -e 'RIOTPROJECT=/data/riotbuild/riotbase' -e 'RIOTCPU=/data/riotbuild/riotbase/cpu' -e 'RIOTBOARD=/data/riotbuild/riotbase/boards' -e 'RIOTMAKE=/data/riotbuild/riotbase/makefiles'      -e 'BOARD=arduino-leonardo' -e 'DISABLE_MODULE=' -e 'DEFAULT_MODULE=test_utils_interactive_sync test_utils_print_stack_usage' -e 'FEATURES_REQUIRED=' -e 'FEATURES_BLACKLIST=' -e 'FEATURES_OPTIONAL=' -e 'USEMODULE=embunit mtd_eeprom' -e 'USEPKG='  -w '/data/riotbuild/riotbase/tests/drivers/mtd_eeprom/' 'docker.io/riot/riotbuild@sha256:08fa7da2c702ac4db7cf57c23fc46c1971f3bffc4a6eff129793f853ec808736' make
Building application "tests_mtd_eeprom" for "arduino-leonardo" with CPU "atmega32u4".

"make" -C /data/riotbuild/riotbase/boards
...
"make" -C /data/riotbuild/riotbase/sys/tsrb
/data/riotbuild/riotbase/tests/drivers/mtd_eeprom/bin/arduino-leonardo/boards_common_atmega/board_common.o:(.roxfa.mtd_dev_xfa.5_0+0x0): multiple definition of `mtd0'
/data/riotbuild/riotbase/tests/drivers/mtd_eeprom/bin/arduino-leonardo/board/board.o:(.roxfa.mtd_dev_xfa.5_0+0x0): first defined here
collect2: error: ld returned 1 exit status
make: *** [/data/riotbuild/riotbase/Makefile.include:734: /data/riotbuild/riotbase/tests/drivers/mtd_eeprom/bin/arduino-leonardo/tests_mtd_eeprom.elf] Error 1
make: *** [/home/buechse/RIOTstuff/riot-vanilla/RIOT/makefiles/docker.inc.mk:395: ..in-docker-container] Error 2

crasbe · 2026-04-14T11:40:06Z

The tests/drivers/mtd_eeprom also needs a Makefile.ci which includes the atmega8.

crasbe · 2026-04-14T12:12:05Z

The tests/sys/eepreg/Makefile probably needs USEMODULE += mtd_eeprom, now that the MTD subsystem is used to access the EEPROM.

I'm unfamiliar with eepgreg, is that only for EEPROMs integrated into the microcontroller or also for external EEPROMs?

fabian18 · 2026-04-14T12:21:44Z

tests/sys/eepreg/Makefile --> eepreg --> mtd

The mtd pulls mtd_eeprom depending on the board.
Technically eepreg compiles with any mtd. I would just not guarantee that it works for anything else than EEPROM.
But mtd_at24cxxx works, which is an external EEPROM, hence it is a real alternative to mtd_eeprom.

crasbe · 2026-04-14T12:22:54Z

In general we might have to think about the name, mtd_eeprom does not really make it clear that it is only for internal EEPROMs.

I mean for external ones we have mtd_at24cxxx, so at the very least we should have a reference to that if someone is looking through the documentation and wonders where the documentation for external EEPROMs is.

Speaking of which... the documentation about the driver is a bit... slim.

You should add some information about how to use the driver (for new boards, CPUs at least), where it gets it's information from about the EEPROM size(s), how it's integrated, what it can be used for etc.

crasbe · 2026-04-14T12:42:18Z

@@ -1 +1 @@
-FEATURES_REQUIRED += periph_eeprom
+USEMODULE += mtd


Suggested change

USEMODULE += mtd

# eepreg works with other `mtd_%` subsystems than `mtd_eeprom` too as not all

# microcontrollers have built-in EEPROMs. However at least one `mtd_%`

# subsystem has to be configured, otherwise fall back to requiring the internal

# EEPROM. This makes sure the automatic feature resultion of the CI only

# tries to compile the test for boards that it would work on.

# Checking the USEMODULE before adding to it makes sure we don't test in the

# first round of dependency resolution to avoid false-negatives.

ifneq (,$(filter mtd,$(USEMODULE)))

ifeq (,$(filter mtd_%,$(USEMODULE)))

FEATURES_REQUIRED += periph_eeprom

endif

endif

USEMODULE += mtd

This should make sure that eepreg builds for all boards that expose an MTD_0 if they have an mtd_% configured or have periph_eeprom present.

buechse@skyleaf:~/RIOTstuff/riot-vanilla/RIOT/tests/sys/eepreg$ BUILD_IN_DOCKER=1 BOARD=nrf52dk make There are unsatisfied feature requirements: periph_eeprom /home/buechse/RIOTstuff/riot-vanilla/RIOT/tests/sys/eepreg/../../../Makefile.include:1033: *** You can let the build continue on expected errors by setting CONTINUE_ON_EXPECTED_ERRORS=1 to the command line. Stop. buechse@skyleaf:~/RIOTstuff/riot-vanilla/RIOT/tests/sys/eepreg$ BUILD_IN_DOCKER=1 BOARD=nrf52840dk make Launching build container using image "docker.io/riot/riotbuild@sha256:08fa7da2c702ac4db7cf57c23fc46c1971f3bffc4a6eff129793f853ec808736". docker run --rm --tty --user $(id -u) --platform linux/amd64 -v '/home/buechse/RIOTstuff/riot-vanilla/RIOT:/data/riotbuild/riotbase:delegated' -v '/home/buechse/.cargo/registry:/data/riotbuild/.cargo/registry:delegated' -v '/home/buechse/.cargo/git:/data/riotbuild/.cargo/git:delegated' -e 'TZ=Europe/Berlin' -e 'RIOTBASE=/data/riotbuild/riotbase' -e 'CCACHE_BASEDIR=/data/riotbuild/riotbase' -e 'BUILD_DIR=/data/riotbuild/riotbase/build' -e 'BUILD_IN_DOCKER=0' -e 'RIOTPROJECT=/data/riotbuild/riotbase' -e 'RIOTCPU=/data/riotbuild/riotbase/cpu' -e 'RIOTBOARD=/data/riotbuild/riotbase/boards' -e 'RIOTMAKE=/data/riotbuild/riotbase/makefiles' -e 'BOARD=nrf52840dk' -e 'DISABLE_MODULE=' -e 'DEFAULT_MODULE=test_utils_interactive_sync test_utils_print_stack_usage' -e 'FEATURES_REQUIRED=' -e 'FEATURES_BLACKLIST=' -e 'FEATURES_OPTIONAL=' -e 'USEMODULE=eepreg' -e 'USEPKG=' -w '/data/riotbuild/riotbase/tests/sys/eepreg/' 'docker.io/riot/riotbuild@sha256:08fa7da2c702ac4db7cf57c23fc46c1971f3bffc4a6eff129793f853ec808736' make Building application "tests_eepreg" for "nrf52840dk" with CPU "nrf52". "make" -C /data/riotbuild/riotbase/pkg/cmsis/ ... "make" -C /data/riotbuild/riotbase/sys/tsrb text data bss dec hex filename 16728 168 2548 19444 4bf4 /data/riotbuild/riotbase/tests/sys/eepreg/bin/nrf52840dk/tests_eepreg.elf

Given this patch I would see USEMODULE += mtd_eeprom there instead of FEATURES_REQUIRED += periph_eeprom. This in the end requires periph_eepom too but it is not explicitly mentioned.

So this will pull mtd_eeprom as a fallback if no particular MTD is used.

Sure. Then it depends on whether you want to open eepreg for other MTDs too.

You could also filter for mtd_24cxxxx only for EEPROM-like MTD devices instead of filtering for all mtd_%.

crasbe · 2026-04-14T13:07:42Z

+#ifndef TEST_EEPREG_MTD
+#  define TEST_EEPREG_MTD       MTD_0
+#endif
+


Suggested change

#ifndef MTD_0

# define MTD_0 mtd_dev_get(0)

#endif

This would be a workaround since not all boards provide MTD_0 (for example if mtd_sdmmc is used).

Or do you want to do a full treewide migration and add the appropriate MTD_0 defines? This would be relevant for devices that have more than one MTD 🤔

I think I added an MTD_<0|1> for all boards that support periph_eeprom.
Giving them an MTD is a benefit for boards that previously did not have any.

Why is mtd_sdmmc relevant? Because the CI will attempt to compile the eepreg test for a board that has only mtd_sdmmc but no MTD_0 defined? That's a a bit of a flaw. Maybe the workaround can be integrated in to mtd in general. I dont know yet

You removed the periph_eeprom feature requirement, so now CI will try to compile for all boards.

I assumed this was intentional.

Yes but I expected that every board supporting an MTD defines at least MTD_0.

I have not applied this yet. For native it is currently an issue that the eeprom would be MTD_1. The eepreg test for native MTD fails.

I would by now not use any MTD_ but rather introduce the naming convention that the internal EEPROM MTD must be calles "mtd_eeprom" for example.

crasbe · 2026-04-14T13:08:26Z


-#define DAT_START    (EEPROM_SIZE - EEPROM_RESERV_CPU_HI \
+#ifndef TEST_EEPREG_MTD


This configuration parameter should be documented in the tests/sys/eepreg/README.md, especially considering boards that might have multiple MTD devices.

crasbe · 2026-04-15T12:11:10Z

 WARNING: This will write to your EEPROM and probably corrupt any data that is
 there!
+
+Add `CFLAGS += -DTEST_EEPREG_MTD=MTD_<X>` if a different MTD should be used than MTD_0.


This has to be changed to MTD_EEPROM now I guess?

crasbe · 2026-04-15T12:11:26Z


 #ifndef TEST_EEPREG_MTD
-#  define TEST_EEPREG_MTD       MTD_0
+#  define TEST_EEPREG_MTD       (&mtd_eeprom.base)


Why not use MTD_EEPROM here?

crasbe · 2026-04-15T12:11:52Z

 #endif

+#if defined(MODULE_MTD_EEPROM)
+#include "mtd_eeprom.h"


Suggested change

#include "mtd_eeprom.h"

# include "mtd_eeprom.h"

crasbe · 2026-04-15T12:19:40Z

 #if MODULE_MTD_EEPROM
-static mtd_eeprom_t eeprom_mtd = MTD_EEPROM_INIT_VAL;
-MTD_XFA_ADD(eeprom_mtd, 0);
+mtd_eeprom_t mtd_eeprom = MTD_EEPROM_INIT_VAL;


Even though this is not a define, I'd propose to call it MTD_EEPROM to reduce the confusion with the module.

But still, there is confusion potential between the module and the MTD device when both have the same name 😅

But I don't have a better name for the MTD device unfortunately :(

fabian18 requested review from MichelRottleuthner, aabadie, kYc0o, leandrolanzieri and maribu as code owners March 30, 2026 18:47

crasbe added the CI: ready for build If set, CI server will compile all applications for all available boards for the labeled PR label Mar 30, 2026

crasbe reviewed Mar 30, 2026

View reviewed changes

crasbe added the Type: new feature The issue requests / The PR implemements a new feature for RIOT label Apr 3, 2026

fabian18 added 2 commits April 14, 2026 11:11

sys/eepreg: API change: use MTD

fb39f58

drivers: add mtd_eeprom for internal EEPROM

8b4a096

fabian18 force-pushed the pr/eepreg_mtd branch from d657b99 to 2115f82 Compare April 14, 2026 09:22

crasbe added the CI: no fast fail don't abort PR build after first error label Apr 14, 2026

fabian18 added 4 commits April 14, 2026 14:03

boards: add mtd_eeprom for atmega boards

39eec5f

boards: add mtd_eeprom for nucleo boards

5075ca2

boards: add mtd_eeprom for native

8839444

tests/drivers: add mtd_eeprom test

97c51cd

fabian18 force-pushed the pr/eepreg_mtd branch from 2115f82 to 97c51cd Compare April 14, 2026 12:03

crasbe reviewed Apr 14, 2026

View reviewed changes

fabian18 added 3 commits April 14, 2026 15:59

fixup! drivers: add mtd_eeprom for internal EEPROM

67237f0

fixup! tests/drivers: add mtd_eeprom test

02e2062

fixup! sys/eepreg: API change: use MTD

7ad6481

github-actions Bot added the Area: doc Area: Documentation label Apr 14, 2026

boards: expose default EEPROM MTD

1f0de5e

crasbe reviewed Apr 15, 2026

View reviewed changes

fabian18 added 3 commits April 16, 2026 09:58

fixup! boards: expose default EEPROM MTD

f0bce7d

fixup! fixup! drivers: add mtd_eeprom for internal EEPROM

28cf276

fixup! fixup! boards: expose default EEPROM MTD

82dcd5b

		@@ -1 +1 @@
		FEATURES_REQUIRED += periph_eeprom
		USEMODULE += mtd

-USEMODULE += mtd
+# eepreg works with other `mtd_%` subsystems than `mtd_eeprom` too as not all
+# microcontrollers have built-in EEPROMs. However at least one `mtd_%`
+# subsystem has to be configured, otherwise fall back to requiring the internal
+# EEPROM. This makes sure the automatic feature resultion of the CI only
+# tries to compile the test for boards that it would work on.
+# Checking the USEMODULE before adding to it makes sure we don't test in the
+# first round of dependency resolution to avoid false-negatives.
+ifneq (,$(filter mtd,$(USEMODULE)))
+  ifeq (,$(filter mtd_%,$(USEMODULE)))
+    FEATURES_REQUIRED += periph_eeprom
+  endif
+endif
+USEMODULE += mtd

+#ifndef MTD_0
+#  define MTD_0 mtd_dev_get(0)
+#endif


		#define DAT_START (EEPROM_SIZE - EEPROM_RESERV_CPU_HI \
		#ifndef TEST_EEPREG_MTD

Conversation

fabian18 commented Mar 30, 2026

Contribution description

Testing procedure

Issues/PRs references

Uh oh!

crasbe left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

riot-ci commented Mar 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Murdock results

Artifacts

Uh oh!

crasbe commented Apr 14, 2026

Uh oh!

crasbe commented Apr 14, 2026

Uh oh!

crasbe commented Apr 14, 2026

Uh oh!

fabian18 commented Apr 14, 2026

Uh oh!

crasbe commented Apr 14, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

fabian18 Apr 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

crasbe Apr 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

riot-ci commented Mar 30, 2026 •

edited

Loading

fabian18 Apr 14, 2026 •

edited

Loading

crasbe Apr 14, 2026 •

edited

Loading