On Fri, Oct 26, 2018 at 04:48:43PM +0200, Ulf Hansson wrote:
> There's no point to register the cpuidle driver for the current CPU, when
> the initialization of the arch specific back-end data fails by returning
> -ENXIO.
>
> Instead, let's re-order the sequence to its original flow, by first trying
> to initialize the back-end part and then act accordingly on the returned
> error code. Additionally, let's print the error message, no matter of what
> error code that was returned.
>
> Signed-off-by: Ulf Hansson <ulf.hansson@xxxxxxxxxx>
> ---
>
> Note, as far as I can tell it's only qcom-spm that may return -ENXIO, which also
> was original reason to why this error path was invented.
>
> Unfurtunate I don't have a qcom spm platfrom at hand, however I manually tried
> this by hacking the PSCI driver to return -ENXIO for some CPUs. So I assume it
> should then also work for qcom spm.
>
> ---
> drivers/cpuidle/cpuidle-arm.c | 22 ++++++++++------------
> 1 file changed, 10 insertions(+), 12 deletions(-)
>
> diff --git a/drivers/cpuidle/cpuidle-arm.c b/drivers/cpuidle/cpuidle-arm.c
> index 073557f433eb..df564d783216 100644
> --- a/drivers/cpuidle/cpuidle-arm.c
> +++ b/drivers/cpuidle/cpuidle-arm.c
> @@ -103,13 +103,6 @@ static int __init arm_idle_init_cpu(int cpu)
> goto out_kfree_drv;
> }
>
> - ret = cpuidle_register_driver(drv);
> - if (ret) {
> - if (ret != -EBUSY)
> - pr_err("Failed to register cpuidle driver\n");
> - goto out_kfree_drv;
> - }
> -
> /*
> * Call arch CPU operations in order to initialize
> * idle states suspend back-end specific data
> @@ -117,15 +110,20 @@ static int __init arm_idle_init_cpu(int cpu)
> ret = arm_cpuidle_init(cpu);
>
> /*
> - * Skip the cpuidle device initialization if the reported
> + * Allow the initialization to continue for other CPUs, if the reported
> * failure is a HW misconfiguration/breakage (-ENXIO).
> */
> - if (ret == -ENXIO)
> - return 0;
I wonder whether moving the drv allocation and dt initialization after
arm_cpuidle_init(cpu) would simplify the error path; the patch makes sense
the error path is getting a bit hard to read - maybe we can simplify it.
I think, if you can single it out, it would be good to add a Fixes:
tag too.
Thanks,
Lorenzo
> -
> if (ret) {
> pr_err("CPU %d failed to init idle CPU ops\n", cpu);
> - goto out_unregister_drv;
> + ret = ret == -ENXIO ? 0 : ret;
> + goto out_kfree_drv;
> + }
> +
> + ret = cpuidle_register_driver(drv);
> + if (ret) {
> + if (ret != -EBUSY)
> + pr_err("Failed to register cpuidle driver\n");
> + goto out_kfree_drv;
> }
>
> dev = kzalloc(sizeof(*dev), GFP_KERNEL);
> --
> 2.17.1
>
