[PATCH net-next v3] net: mctp: Make flow KUnit coverage an explicit opt-in

[PATCH net-next v3] net: mctp: Make flow KUnit coverage an explicit opt-in

[Karl Mehltretter]

MCTP_TEST selects MCTP_FLOWS, so KUNIT_ALL_TESTS builds flow support and
SKB_EXTENSIONS even when no transport needs them. KUnit's style guide
says tests should not select otherwise unused functionality. The CRC
and crypto library tests use explicit ENABLE_ALL_FOR_KUNIT options for
full coverage.

Remove the selection from MCTP_TEST. The core suite remains available
without flows; its two flow cases use their existing skip paths. Add
MCTP_ENABLE_ALL_FOR_KUNIT to opt into full coverage, and enable it in
all_tests.config and a dedicated net/mctp/.kunitconfig. Both profiles
run the flow cases without a hardware transport.

A plain MCTP_TEST=y configuration without the opt-in now skips those
two cases. This is the coverage cost of leaving otherwise unused flow
support disabled.

Assisted-by: LLM
Signed-off-by: Karl Mehltretter <kmehltretter@gmail.com>
---
Changes in v3:
- Keep the core suite available without MCTP_FLOWS; a plain MCTP_TEST=y
  configuration now skips only the two flow cases.
- Add MCTP_ENABLE_ALL_FOR_KUNIT to retain full coverage in
  all_tests.config and a focused net/mctp/.kunitconfig.
- Drop the Fixes tag because this changes test configuration rather
  than fixing a runtime MCTP bug.

v2: https://lore.kernel.org/r/20260928205352.30502-1-kmehltretter@gmail.com/
v1: https://lore.kernel.org/r/20260925192510.93840-1-kmehltretter@gmail.com/

Validation: The MCTP KUnit suites were built with GCC 15.2 and booted
under QEMU 10.2.1 on an isolated source export from net-next
014d795c7383, with the changed Kconfig files copied from 071876fd5048.
Without the opt-in, mctp-sock passed 5/5 tests and mctp-route passed
21/23 tests with two expected flow-case skips. With the opt-in, both
suites passed all 5 and 23 tests, respectively. On the full
071876fd5048 tree, the new net/mctp/.kunitconfig resolves all requested
symbols and selects MCTP_FLOWS and SKB_EXTENSIONS.

 net/mctp/.kunitconfig                        |  5 +++++
 net/mctp/Kconfig                             | 16 +++++++++++++++-
 tools/testing/kunit/configs/all_tests.config |  2 +-
 3 files changed, 21 insertions(+), 2 deletions(-)
 create mode 100644 net/mctp/.kunitconfig

diff --git a/net/mctp/.kunitconfig b/net/mctp/.kunitconfig
new file mode 100644
index 000000000..4b2969e42
--- /dev/null
+++ b/net/mctp/.kunitconfig
@@ -0,0 +1,5 @@
+CONFIG_KUNIT=y
+CONFIG_NET=y
+CONFIG_MCTP=y
+CONFIG_MCTP_TEST=y
+CONFIG_MCTP_ENABLE_ALL_FOR_KUNIT=y
diff --git a/net/mctp/Kconfig b/net/mctp/Kconfig
index d8d3413a3..a89678428 100644
--- a/net/mctp/Kconfig
+++ b/net/mctp/Kconfig
@@ -12,9 +12,23 @@ menuconfig MCTP
 	  devices, you'll want to enable a driver for a specific hardware
 	  channel.
 
+config MCTP_ENABLE_ALL_FOR_KUNIT
+	bool "Enable all optional MCTP features for KUnit tests"
+	depends on MCTP=y && KUNIT=y
+	select MCTP_FLOWS
+	help
+	  Enable all optional MCTP features covered by KUnit tests, even
+	  when no transport needs them. Currently this selects MCTP_FLOWS
+	  so the two flow cases can run.
+
+	  Without this option, the core suite remains available; the flow
+	  cases skip unless a transport enables flow support.
+
+	  You still need to enable MCTP_TEST, either individually or through
+	  KUNIT_ALL_TESTS.
+
 config MCTP_TEST
         bool "MCTP core tests" if !KUNIT_ALL_TESTS
-        select MCTP_FLOWS
         depends on MCTP=y && KUNIT=y
         default KUNIT_ALL_TESTS
 
diff --git a/tools/testing/kunit/configs/all_tests.config b/tools/testing/kunit/configs/all_tests.config
index 31e930a0f..b3ea56f4e 100644
--- a/tools/testing/kunit/configs/all_tests.config
+++ b/tools/testing/kunit/configs/all_tests.config
@@ -26,7 +26,7 @@ CONFIG_GPIOLIB=y
 
 CONFIG_NET=y
 CONFIG_MCTP=y
-CONFIG_MCTP_FLOWS=y
+CONFIG_MCTP_ENABLE_ALL_FOR_KUNIT=y
 
 CONFIG_INET=y
 CONFIG_MPTCP=y

base-commit: 071876fd50482a68603a9460d80dd6dd58827ee1
-- 
2.39.5 (Apple Git-154)

Re: [PATCH net-next v3] net: mctp: Make flow KUnit coverage an explicit opt-in

[Jakub Kicinski]

On Fri,  2 Oct 2026 23:20:58 +0200 Karl Mehltretter wrote:
> MCTP_TEST selects MCTP_FLOWS, so KUNIT_ALL_TESTS builds flow support and
> SKB_EXTENSIONS even when no transport needs them. KUnit's style guide
> says tests should not select otherwise unused functionality. The CRC
> and crypto library tests use explicit ENABLE_ALL_FOR_KUNIT options for
> full coverage.
> 
> Remove the selection from MCTP_TEST. The core suite remains available
> without flows; its two flow cases use their existing skip paths. Add
> MCTP_ENABLE_ALL_FOR_KUNIT to opt into full coverage, and enable it in
> all_tests.config and a dedicated net/mctp/.kunitconfig. Both profiles
> run the flow cases without a hardware transport.
> 
> A plain MCTP_TEST=y configuration without the opt-in now skips those
> two cases. This is the coverage cost of leaving otherwise unused flow
> support disabled.
> 
> Assisted-by: LLM
> Signed-off-by: Karl Mehltretter <kmehltretter@gmail.com>
> ---
> Changes in v3:
> - Keep the core suite available without MCTP_FLOWS; a plain MCTP_TEST=y
>   configuration now skips only the two flow cases.
> - Add MCTP_ENABLE_ALL_FOR_KUNIT to retain full coverage in
>   all_tests.config and a focused net/mctp/.kunitconfig.
> - Drop the Fixes tag because this changes test configuration rather
>   than fixing a runtime MCTP bug.
> 
> v2: https://lore.kernel.org/r/20260928205352.30502-1-kmehltretter@gmail.com/
> v1: https://lore.kernel.org/r/20260925192510.93840-1-kmehltretter@gmail.com/
> 
> Validation: The MCTP KUnit suites were built with GCC 15.2 and booted
> under QEMU 10.2.1 on an isolated source export from net-next
> 014d795c7383, with the changed Kconfig files copied from 071876fd5048.
> Without the opt-in, mctp-sock passed 5/5 tests and mctp-route passed
> 21/23 tests with two expected flow-case skips. With the opt-in, both
> suites passed all 5 and 23 tests, respectively. On the full
> 071876fd5048 tree, the new net/mctp/.kunitconfig resolves all requested
> symbols and selects MCTP_FLOWS and SKB_EXTENSIONS.

I don't think this is worth applying
-- 
pw-bot: reject