Re: [RFC PATCH 9/9] kvm_main.c: handle atomic memslot update

[Emanuele Giuseppe Esposito]

Am 13/09/2022 um 04:30 schrieb Yang, Weijiang:
> 
> On 9/9/2022 6:45 PM, Emanuele Giuseppe Esposito wrote:
>> When kvm_vm_ioctl_set_memory_region_list() is invoked, we need
>> to make sure that all memslots are updated in the inactive list
>> and then swap (preferreably only once) the lists, so that all
>> changes are visible immediately.
> [...]
>> +static int kvm_vm_ioctl_set_memory_region_list(struct kvm *kvm,
>> +               struct kvm_userspace_memory_region_list *list,
>> +               struct kvm_userspace_memory_region_entry __user *mem_arg)
>> +{
>> +    struct kvm_userspace_memory_region_entry *mem, *m_iter;
>> +    struct kvm_userspace_memory_region *mem_region;
>> +    struct kvm_internal_memory_region_list *batch, *b_iter;
>> +    int i, r = 0;
>> +    bool *as_to_swap;
>> +
>> +    /* TODO: limit the number of mem to a max? */
>> +
>> +    if (!list->nent)
>> +        return r;
>> +
>> +    mem = vmemdup_user(mem_arg, array_size(sizeof(*mem), list->nent));
>> +    if (IS_ERR(mem)) {
>> +        r = PTR_ERR(mem);
>> +        goto out;
>> +    }
> 
> IMO, it's more natural to dup the user memory at the first place, i.e.,
> kvm_vm_ioctl,
> 
> it also makes the outlets shorter.
> 

I followed the same pattern as kvm_vcpu_ioctl_set_cpuid2, which performs
the user memory dup inside the call :)

I see your point but I guess it's better to keep all ioctl
implementations similar.

Thank you,
Emanuele