Li RongQing <lirongqing@xxxxxxxxx> writes:
> From: Li Rongqing <lirongqing@xxxxxxxxx>
>
> move svm_hardware_unsetup before svm_hardware_setup, and call
> it to free all memory if fail to setup in svm_hardware_setup,
> otherwise memory will be leaked
>
> remove __exit attribute for svm_hardware_unsetup since it is
> called in __init function
>
> Signed-off-by: Li Rongqing <lirongqing@xxxxxxxxx>
> Signed-off-by: Zhang Yu <zhangyu31@xxxxxxxxx>
> ---
> arch/x86/kvm/svm.c | 31 +++++++++++++++----------------
> 1 file changed, 15 insertions(+), 16 deletions(-)
>
> diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
> index 1787a484d21c..276ab8ab6c95 100644
> --- a/arch/x86/kvm/svm.c
> +++ b/arch/x86/kvm/svm.c
> @@ -1290,6 +1290,20 @@ static void shrink_ple_window(struct kvm_vcpu *vcpu)
> control->pause_filter_count, old);
> }
>
> +static void svm_hardware_unsetup(void)
> +{
Unrelated to your patch but "unsetup" makes me feel sad. Why can't we
use e.g. "teardown" instead? (KVM is among just four users of 'unsetup'
word in kernel).
> + int cpu;
> +
> + if (svm_sev_enabled())
> + bitmap_free(sev_asid_bitmap);
> +
> + for_each_possible_cpu(cpu)
> + svm_cpu_uninit(cpu);
> +
> + __free_pages(pfn_to_page(iopm_base >> PAGE_SHIFT), IOPM_ALLOC_ORDER);
> + iopm_base = 0;
> +}
> +
> static __init int svm_hardware_setup(void)
> {
> int cpu;
> @@ -1396,25 +1410,10 @@ static __init int svm_hardware_setup(void)
> return 0;
>
> err:
> - __free_pages(iopm_pages, IOPM_ALLOC_ORDER);
> - iopm_base = 0;
> + svm_hardware_unsetup();
> return r;
> }
>
> -static __exit void svm_hardware_unsetup(void)
> -{
> - int cpu;
> -
> - if (svm_sev_enabled())
> - bitmap_free(sev_asid_bitmap);
> -
> - for_each_possible_cpu(cpu)
> - svm_cpu_uninit(cpu);
> -
> - __free_pages(pfn_to_page(iopm_base >> PAGE_SHIFT), IOPM_ALLOC_ORDER);
> - iopm_base = 0;
> -}
> -
> static void init_seg(struct vmcb_seg *seg)
> {
> seg->selector = 0;
Reviewed-by: Vitaly Kuznetsov <vkuznets@xxxxxxxxxx>
--
Vitaly
