Re: [Xen-devel] [PATCH v3 07/16] x86/p2m/pod: make it build with !CONFIG_HVM

[George Dunlap <george.dunlap@xxxxxxxxxx>]

On 09/06/2018 04:05 PM, Jan Beulich wrote:
>>>> On 04.09.18 at 18:15, <wei.liu2@xxxxxxxxxx> wrote:
>> --- a/xen/arch/x86/mm/p2m.c
>> +++ b/xen/arch/x86/mm/p2m.c
>> @@ -52,15 +52,20 @@ DEFINE_PERCPU_RWLOCK_GLOBAL(p2m_percpu_rwlock);
>>  /* Init the datastructures for later use by the p2m code */
>>  static int p2m_initialise(struct domain *d, struct p2m_domain *p2m)
>>  {
>> -    unsigned int i;
>>      int ret = 0;
>> +#ifdef CONFIG_HVM
>> +    unsigned int i;
>> +#endif
>>  
>>      mm_rwlock_init(&p2m->lock);
>> -    mm_lock_init(&p2m->pod.lock);
>>      INIT_LIST_HEAD(&p2m->np2m_list);
>>      INIT_PAGE_LIST_HEAD(&p2m->pages);
>> +
>> +#ifdef CONFIG_HVM
>> +    mm_lock_init(&p2m->pod.lock);
>>      INIT_PAGE_LIST_HEAD(&p2m->pod.super);
>>      INIT_PAGE_LIST_HEAD(&p2m->pod.single);
>> +#endif
>>  
>>      p2m->domain = d;
>>      p2m->default_access = p2m_access_rwx;
>> @@ -69,8 +74,10 @@ static int p2m_initialise(struct domain *d, struct 
>> p2m_domain *p2m)
>>      p2m->np2m_base = P2M_BASE_EADDR;
>>      p2m->np2m_generation = 0;
>>  
>> +#ifdef CONFIG_HVM
>>      for ( i = 0; i < ARRAY_SIZE(p2m->pod.mrp.list); ++i )
>>          p2m->pod.mrp.list[i] = gfn_x(INVALID_GFN);
>> +#endif
> 
> I wonder whether all of this wouldn't better go into p2m_pod_init()
> (or alike), to limit/avoid the #ifdef-ary here.

The cover letter for this series did say the initial goal was just to
make it build; and cleaning up could be done later.  Adding
p2m_pod_init() would be better long term, I think, but I'm also happy
with leaving if #ifdef's for now and cleaning them up later.

> 
>> @@ -2560,7 +2573,10 @@ void audit_p2m(struct domain *d,
>>      P2M_PRINTK("p2m audit starts\n");
>>  
>>      p2m_lock(p2m);
>> +
>> +#ifdef CONFIG_HVM
>>      pod_lock(p2m);
>> +#endif
>>  
>>      if (p2m->audit_p2m)
>>          pmbad = p2m->audit_p2m(p2m);
>> @@ -2621,7 +2637,9 @@ void audit_p2m(struct domain *d,
>>      }
>>      spin_unlock(&d->page_alloc_lock);
>>  
>> +#ifdef CONFIG_HVM
>>      pod_unlock(p2m);
>> +#endif
> 
> Perhaps better make them inline stubs? Otoh - isn't the whole
> function useful for HVM only?

Oh yes -- now that we've gotten rid of full translate guests for PV, I
guess so.

> 
>> --- a/xen/common/vm_event.c
>> +++ b/xen/common/vm_event.c
>> @@ -630,7 +630,9 @@ int vm_event_domctl(struct domain *d, struct 
>> xen_domctl_vm_event_op *vec,
>>          {
>>          case XEN_VM_EVENT_ENABLE:
>>          {
>> +#ifdef CONFIG_HVM
>>              struct p2m_domain *p2m = p2m_get_hostp2m(d);
>> +#endif
>>  
>>              rc = -EOPNOTSUPP;
>>              /* hvm fixme: p2m_is_foreign types need addressing */
>> @@ -647,10 +649,12 @@ int vm_event_domctl(struct domain *d, struct 
>> xen_domctl_vm_event_op *vec,
>>              if ( unlikely(need_iommu(d)) )
>>                  break;
>>  
>> +#ifdef CONFIG_HVM
>>              rc = -EXDEV;
>>              /* Disallow paging in a PoD guest */
>>              if ( p2m->pod.entry_count )
>>                  break;
>> +#endif
> 
> Perhaps simply ditch the local variable?

That seems like a good idea.

Everything else looks good to me.

 -George

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel