Skip to content

WIP: make rasdaemon more modular and add support to use other databases - #258

Draft
mchehab wants to merge 86 commits into
masterfrom
next
Draft

WIP: make rasdaemon more modular and add support to use other databases#258
mchehab wants to merge 86 commits into
masterfrom
next

Conversation

@mchehab

@mchehab mchehab commented Aug 8, 2026 •
edited
Loading

Copy link
Copy Markdown
Owner

In preparation for rasdaemon version 1.0.0, improve its internal architecture.

Goals:

  • make rasdaemon modular;
  • allow using more databases;
  • add unit tests

Updates:

  • mysql support added;
  • postgresql support added;
  • sqlite3, mysql and postgresql unit tests passed;
  • converted to use meson instead of auto-tools;
  • CI jobs adjusted to use meson and all off them passes;
  • Improved README.md file content.

I appreciate feedbacks and PRs against the next branch to help with the redesign.

mchehab added 10 commits August 8, 2026 13:36
@mchehab
ras-record: use a more generic way to describe database fields
0c9e159 
Right now, the DB logic was written to support only sqlite3.
As we're aiming to add support for other databases, use a more
generic way to describe DB fields.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
ras-record: add a generic binding routine
24c47c7 
Convert sqlite3-specific bindings into a generic one inside
ras-record, to allow adding support later for other databases.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
Get rid of sqlite3 specifics outside ras-record
4183ddf 
In preparation for adding support to other databases, don't
keep any sqlite3 stuff outside ras-record.

For now, we'll keep HAVE_SQLITE3 as an intermediate step,
as right now the only SQL supported database is sqlite3.

In the future, this will be broken on two separate fields,
and the checks inside each rasdaemon file will disapear, as
the SQL backend core will take care of it.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
ras-reri-handler: fix file permissions
64b0275 
Those should be 644, not 755.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
rasdaemon: add an interface to register its modules
472b67a 
Right now, rasdaemon is only partially modular: at build time,
each individual feature can be selected and added, but the code
requires lots of ifdefs.

This worked fine at the beginning when there weren't much options,
but now the code has lots of optional features. Use a more modular
design by adding a module register logic.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
configure.ac: bump its version to mark it as a draft
bed2908 
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
Reorganize source tree
6c1bc90 
As the project grew, there are now too much files at the
core directory, making harder to maintain.

Move them to subdirs.

No functional changes.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
core/modules: refactor module register logic
1c4c87a 
Make it more generic, aiming to allow automatic module load
via __attribute__((constructor)).

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
core/modules: keep modules entry sorted
55a21d1 
As we're aiming to let module registration to happen in any
order, make the logic deterministic by inserting modules in
alphabetic order.

This makes insert being O(n) instead of O(1), but performance
here is not a problem.

While here, fix a bug when checking if allocation succeded.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
core/modules: fix several bugs at the logic
3ea90df 
While the code compiles, it fails at the test cases.

Fix the issues.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
tests: add unit tests for modules
e449a7d 
Add unit tests to check if modules logic works.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
mchehab added 4 commits August 9, 2026 11:57
@mchehab
core/modules: better handle cleanup logic
e782f72 
Allow modules to optionally set a private data and ensure that
cleanup will receive enough context.

This is needed to check modules functionality via unit tests,
and could help to ensure that cleanup will do the right thing.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
tests/test-modules: add an unit test for module init/cleanup
b389650 
Ensure that init and cleanup logic will work as expected.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
tests/test-modules: add a logic to better check errors
8fca264 
Instead of relying on manual conference, check programaticly
if module cleanups will happen at the expected way.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
core/ras-logger: add support for mocking output
1f63e90 
When doing unit tests, it doesn't make any sense to write at
syslog. Also, on success, the normal behavior is to just mask
the error output.

Implement support for it and modify tests/test-modules to
use such feature.

After such change, a successful run will be cleaner:

	Testing modules functionality:
	01/06:   Running NULL entry
	02/06:   Running single module
	03/06:   Running modules in order
	04/06:   Running modules out of order
	05/06:   Running multiple levels
	06/06:   Running init/cleanup
	All module registration tests passed.

And if the logs are always shown, there will be a tab on
each new line, which will produce a better visual output
as well:

	Testing modules functionality:
	01/06:   Running NULL entry
        module entry is missing!
	02/06:   Running single module
	03/06:   Running modules in order
	04/06:   Running modules out of order
	05/06:   Running multiple levels
	06/06:   Running init/cleanup
	        module alpha enabled
	        module charlie enabled
	        module foxtrot enabled
	        module delta enabled
	        module echo enabled
	        module beta enabled
	All module registration tests passed.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
mchehab added 8 commits August 10, 2026 17:43
@mchehab
rasdaemon: abstract sqlite3 logic
82f5e27 
Move all sqlite3-specific logit to db/ras-db, preparing the code
to support other databases in the future.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
db/ras-db: rename database operations
47de667 
change the namespace for all database-related ops.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
db/ras-db: document the database interface
b3a8b5c 
Use kernel-doc style markups to document the database
generic interface.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
db/ras-db: Rename the C code to db/db-sqlite3.c
46fc46a 
The code there contains only sqlite3 specifics. Rename it to
better reflect its content.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
ras-db: add an database abstraction layer
287de7c 
In preparation to support other database types, add an
abstraction layer.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
db/db-sqlite3: use the modules register logic
df617a4 
As the idea is to make the logic more modular to give more
flexibility and remove ifdef code from rasdaemon, move
the DB backend initialization to be done via module register.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
db/ras-db: add missing code
8c16aef 
Forgot to add ras-db code on a previous commit.

While here, update .gitignore.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
modules: add a function to initialize a single module.
ab7faba 
This is useful for unit tests.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
mchehab force-pushed the next branch 2 times, most recently from 45e9484 to 38bf93a Compare August 11, 2026 12:01
@mchehab
ras-db: implement unit tests for db-sqlite3
350c595 
Add a way to check if db-sqlite3 implementation is working.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
db/ras-db: some fixes and cleanups
d9018b9 
The logic inside ras-db to pick the next backend is moving the
pointer twice, producing sigfaults.

While here, do a couple of cleanups.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
mchehab force-pushed the next branch 9 times, most recently from b15e0ba to 110ef8b Compare August 18, 2026 08:08
mchehab added 2 commits August 18, 2026 10:41
@mchehab
rasdaemon: clean more files with distclean
d8bf53f 
Ensure that the build and SRPMS/ directories will be removed.
While here, also remove m4/ directory.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
actions: use yamlfmt to format YAML files the same way
47b234c 
Use a more canonical identation format, as given by yamlfmt.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
mchehab added 2 commits August 18, 2026 11:49
@mchehab
rasdaemon: add support for PostgreSQL database backend
229fdbd 
Now that we have a generic interface, add PostgreSQL support.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
actions: expand unit test to also cover PostgreSQL backend
0379979 
let CI do unit tests for PostgreSQL as well.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
mchehab added 3 commits August 18, 2026 11:54
@mchehab
README.md: Add information about database backend and unit teststhe
8e3e9e5 
Database setup is somewhat complex. Add a set of instructions about
it.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
non-standard: add some missing ifdefs if no database
25661ca 
There are some code there which should be compiled only if
a database backend is selected.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
rasdaemon: use HAVE_DB instead of HAVE_SQLITE3
766459e 
As there are now multiple database backends, use the generic
HAVE_DB to ensure that the code will work with any backend.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
mchehab added 10 commits August 18, 2026 17:10
@mchehab
rasdaemon: auto-load rasdaemon configuration
485b912 
Until this change, rasdaemon had an env file stored under:
	/etc/sysconfig/rasdaemon.env

However, the content of this file was never read. Also, it
doesn't follow the way other files are stored there: without
a .env suffix.

Change the logic to make rasdaemon to automatically load its
configuration if the file is present.

If there's no /etc/sysconfig/rasdaemon, a warning will be
issued but the logic will proceed.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
misc/rasdaemon.env: add database connection parameters
811ef6a 
Add support to configure a database via env parameters.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
rasdaemon: add support for initializing backends
93e76af 
Allow rasdaemon to select the right SQL backend.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
rasdaemon: add an option to select the config file
ae52d2f 
Allow overriding the default using a --config (-c) option
to optionally select a different config file.

This can be very helpful when testing rasdaemon with different
DB backends.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
db/ras-db.h: avoid adding a duplicated db_backend_enable()
645d1d4 
Make the empty stub static inline, as otherwise some setups
may end declaring it multiple times.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
rasdaemon: cleanup some logs
21eb624 
Remove some logs that were useful while changing rasdaemon
architecture.

While here, replace the meaningless error message:

	rasdaemon: Failed to open directory

to one that actually makes sense:

	rasdaemon: Failed to open /sys/fs/pstore/erst directory

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
ras-logger: use ansi colors on tty terminals
64c3b77 
Rasdaemon produces lots of logs. Make it clearer to identify
potential problems by adding colors to the messages.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
@mchehab
FOO
4b29f86
@mchehab
Revert "FOO"
1cff62b 
This reverts commit 4b29f86.
@mchehab
meson.build: fix Fedora builds with mock
2208d3a 
We want to allow installing inside a directory with:

meson build --reconfigure  -Dprefix=`pwd`/instal

but, at the same time, we need to support "make mock"
target.

The first one works, but `make mock` is adding /usr
twice:

	./BUILDROOT/usr/usr/lib/systemd/system/ras-mc-ctl.service
	./BUILDROOT/usr/usr/lib/systemd/system/rasdaemon.service

Add an extra check to prevent that.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@mchehab