[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CAPDyKFoFBoeA0aqnFTNh83QGvb9j6QRwnacXcYFYzKMa_Y9iBg@mail.gmail.com>
Date: Fri, 22 Jan 2021 10:26:25 +0100
From: Ulf Hansson <ulf.hansson@...aro.org>
To: Andrew Jeffery <andrew@...id.au>
Cc: "linux-mmc@...r.kernel.org" <linux-mmc@...r.kernel.org>,
Adrian Hunter <adrian.hunter@...el.com>,
Joel Stanley <joel@....id.au>,
linux-aspeed <linux-aspeed@...ts.ozlabs.org>,
Linux ARM <linux-arm-kernel@...ts.infradead.org>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
brendanhiggins@...gle.com, Randy Dunlap <rdunlap@...radead.org>
Subject: Re: [PATCH] mmc: sdhci-of-aspeed: Fix kunit-related build error
On Fri, 22 Jan 2021 at 04:23, Andrew Jeffery <andrew@...id.au> wrote:
>
> Randy found that with the following Kconfig settings we have duplicate
> definitions (e.g. __inittest()) in sdhci-of-aspeed due to competing
> module_init()/module_exit() calls from kunit and driver the itself.
>
> ```
> CONFIG_MMC_SDHCI_OF_ASPEED=m
> CONFIG_MMC_SDHCI_OF_ASPEED_TEST=y
> ```
>
> Conditionally open-code the kunit initialisation to avoid the error.
>
> Fixes: 7efa02a981d6 ("mmc: sdhci-of-aspeed: Add KUnit tests for phase calculations")
> Reported-by: Randy Dunlap <rdunlap@...radead.org>
> Signed-off-by: Andrew Jeffery <andrew@...id.au>
> ---
> drivers/mmc/host/sdhci-of-aspeed-test.c | 9 ++++++++-
> drivers/mmc/host/sdhci-of-aspeed.c | 27 ++++++++++++++++++++-----
> 2 files changed, 30 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/mmc/host/sdhci-of-aspeed-test.c b/drivers/mmc/host/sdhci-of-aspeed-test.c
> index 34070605b28b..e5edda6c6e60 100644
> --- a/drivers/mmc/host/sdhci-of-aspeed-test.c
> +++ b/drivers/mmc/host/sdhci-of-aspeed-test.c
> @@ -95,4 +95,11 @@ static struct kunit_suite aspeed_sdhci_test_suite = {
> .name = "sdhci-of-aspeed",
> .test_cases = aspeed_sdhci_test_cases,
> };
> -kunit_test_suite(aspeed_sdhci_test_suite);
> +
> +static struct kunit_suite *aspeed_sdhci_test_suite_array[] = {
> + &aspeed_sdhci_test_suite,
> + NULL,
> +};
> +
> +static struct kunit_suite **aspeed_sdhci_test_suites
> + __used __section(".kunit_test_suites") = aspeed_sdhci_test_suite_array;
> diff --git a/drivers/mmc/host/sdhci-of-aspeed.c b/drivers/mmc/host/sdhci-of-aspeed.c
> index 3b0d381e1215..1446f23a2bfb 100644
> --- a/drivers/mmc/host/sdhci-of-aspeed.c
> +++ b/drivers/mmc/host/sdhci-of-aspeed.c
> @@ -556,6 +556,10 @@ static struct platform_driver aspeed_sdc_driver = {
> .remove = aspeed_sdc_remove,
> };
>
> +#if defined(CONFIG_MMC_SDHCI_OF_ASPEED_TEST)
> +#include "sdhci-of-aspeed-test.c"
> +#endif
> +
> static int __init aspeed_sdc_init(void)
> {
> int rc;
> @@ -566,7 +570,20 @@ static int __init aspeed_sdc_init(void)
>
> rc = platform_driver_register(&aspeed_sdc_driver);
> if (rc < 0)
> - platform_driver_unregister(&aspeed_sdhci_driver);
> + goto cleanup_sdhci;
> +
> +#if defined(CONFIG_MMC_SDHCI_OF_ASPEED_TEST)
I would like us to avoid #if defined" in code like this.
Can you instead declare a function that you can stub in case
CONFIG_MMC_SDHCI_OF_ASPEED_TEST is unset?
> + rc = __kunit_test_suites_init(aspeed_sdhci_test_suites);
> + if (rc < 0) {
> + platform_driver_unregister(&aspeed_sdc_driver);
> + goto cleanup_sdhci;
> + }
> +#endif
> +
> + return 0;
> +
> +cleanup_sdhci:
> + platform_driver_unregister(&aspeed_sdhci_driver);
>
> return rc;
> }
> @@ -574,15 +591,15 @@ module_init(aspeed_sdc_init);
>
> static void __exit aspeed_sdc_exit(void)
> {
> +#if defined(CONFIG_MMC_SDHCI_OF_ASPEED_TEST)
> + __kunit_test_suites_exit(aspeed_sdhci_test_suites);
> +#endif
> +
> platform_driver_unregister(&aspeed_sdc_driver);
> platform_driver_unregister(&aspeed_sdhci_driver);
> }
> module_exit(aspeed_sdc_exit);
>
> -#if defined(CONFIG_MMC_SDHCI_OF_ASPEED_TEST)
> -#include "sdhci-of-aspeed-test.c"
> -#endif
> -
> MODULE_DESCRIPTION("Driver for the ASPEED SD/SDIO/SDHCI Controllers");
> MODULE_AUTHOR("Ryan Chen <ryan_chen@...eedtech.com>");
> MODULE_AUTHOR("Andrew Jeffery <andrew@...id.au>");
> --
> 2.27.0
>
Kind regards
Uffe
Powered by blists - more mailing lists