On 30/10/2024 07:55, Aneesh Kumar K.V wrote:
> Steven Price <steven.price@xxxxxxx> writes:
>
>> +
>> +out_undelegate_tables:
>> + while (--i >= 0) {
>> + phys_addr_t pgd_phys = kvm->arch.mmu.pgd_phys + i * PAGE_SIZE;
>> +
>> + WARN_ON(rmi_granule_undelegate(pgd_phys));
>> + }
>> + WARN_ON(rmi_granule_undelegate(rd_phys));
>> +free_rd:
>> + free_page((unsigned long)rd);
>> + return r;
>> +}
>> +
>
> we should avoid that free_page on an undelegate failure? rd_phys we can
> handle here. Not sure how to handle the pgd_phys.
Good point. I think for pgd_phys setting kvm->arch.mmu.pgt=NULL should
be sufficient to prevent the pages being freed.
Thanks,
Steve
