Re: [PATCH 6/9] test: Probe RTC early in dm_test_host()

Hi Heinrich,

On 2026-05-18T05:57:19, Heinrich Schuchardt
<[email protected]> wrote:
> test: Probe RTC early in dm_test_host()
>
> The ext4 driver probes and reads the RTC which allocates memory.
>
> Ensure that the device is already probed and read once in dm_test_hook()
> to avoid false positives.
>
> Signed-off-by: Heinrich Schuchardt <[email protected]>
>
> test/dm/host.c | 13 +++++++++++++
>  1 file changed, 13 insertions(+)

> diff --git a/test/dm/host.c b/test/dm/host.c
> @@ -26,6 +27,18 @@ static int dm_test_host(struct unit_test_state *uts)
>       ulong mem_start;
>       loff_t actwrite;
>
> +     /*
> +      * Probing and first read from the RTC allocates memory.
> +      * Do it before the measurement.
> +      */

Please mention here (and in the commit message) which call inside this
test ends up touching the RTC. It is non-obvious that fs_write()
further down now reaches into the RTC via patch 5, and a future reader
will not understand why an RTC probe belongs in a host test.

> diff --git a/test/dm/host.c b/test/dm/host.c
> @@ -26,6 +27,18 @@ static int dm_test_host(struct unit_test_state *uts)
> +     if (CONFIG_IS_ENABLED(DM_RTC)) {
> +             struct rtc_time tm;
> +
> +             uclass_first_device(UCLASS_RTC, &dev);
> +             if (dev)
> +                     dm_rtc_get(dev, &tm);
> +     }

Since this is a DM test you should assert each of these calls.

Also, would it be simpler to take a second mem_start sample after the
warm-up, so the RTC allocation is naturally outside the measured
window without needing to know it exists?

> Ensure that the device is already probed and read once in dm_test_hook()
> to avoid false positives.

Should be dm_test_host(). Also spell out that it is the
ut_check_delta() leak check at the end of the test that gets fooled by
the RTC's one-shot allocation.

Regards,
Simon