On Wed, Dec 21, 2016 at 11:19:34AM +0100, Alexander Stein wrote:
> Add ARM PMU removal function. This will be required by perf event drivers
> when option DEBUG_TEST_DRIVER_REMOVE is enabled.
>
> Signed-off-by: Alexander Stein <alexander.st...@systec-electronic.com>
> ---
> drivers/perf/arm_pmu.c | 14 ++++++++++++++
> include/linux/perf/arm_pmu.h | 2 ++
> 2 files changed, 16 insertions(+)
>
> diff --git a/drivers/perf/arm_pmu.c b/drivers/perf/arm_pmu.c
> index a9bbdbf..b7ddc4c 100644
> --- a/drivers/perf/arm_pmu.c
> +++ b/drivers/perf/arm_pmu.c
> @@ -1022,6 +1022,7 @@ int arm_pmu_device_probe(struct platform_device *pdev,
> armpmu_init(pmu);
>
> pmu->plat_device = pdev;
> + platform_set_drvdata(pdev, pmu);
>
> if (node && (of_id = of_match_node(of_table, pdev->dev.of_node))) {
> init_fn = of_id->data;
> @@ -1073,6 +1074,19 @@ int arm_pmu_device_probe(struct platform_device *pdev,
> return ret;
> }
>
> +int arm_pmu_device_remove(struct platform_device *pdev)
> +{
> + struct arm_pmu *pmu = platform_get_drvdata(pdev);
> +
> + __oprofile_cpu_pmu = NULL;
> +
> + perf_pmu_unregister(&pmu->pmu);
> +
> + cpu_pmu_destroy(pmu);
> +
> + return 0;
> +}
So normally, if there are events that use this pmu, we hold a reference
on its module, which avoids removal from happening.
How is that guarantee made by DEBUG_TEST_DRIVER_REMOVE ? Or will it
simply kill everything even though there's active events for the PMU?
