[libvirt] [PATCH v2 02/14] Introduce NVDIMM memory model

[John Ferlan]

[...]


>>      virBufferAdjustIndent(buf, -2);
>>      virBufferAddLit(buf, "</source>\n");
>>  
> 
> So no need

Looks like this lost thought got somehow cut-n-paste below ...


>> diff --git a/src/conf/domain_conf.h b/src/conf/domain_conf.h
>> index 1e53cc328..dc949d3c9 100644
>> --- a/src/conf/domain_conf.h
>> +++ b/src/conf/domain_conf.h
>> @@ -1996,6 +1996,7 @@ struct _virDomainRNGDef {
>>  typedef enum {
>>      VIR_DOMAIN_MEMORY_MODEL_NONE,
>>      VIR_DOMAIN_MEMORY_MODEL_DIMM, /* dimm hotpluggable memory device */
>> +    VIR_DOMAIN_MEMORY_MODEL_NVDIMM, /* nvdimm memory device */
>>  
>>      VIR_DOMAIN_MEMORY_MODEL_LAST
>>  } virDomainMemoryModel;
>> @@ -2004,6 +2005,7 @@ struct _virDomainMemoryDef {
>>      /* source */
>>      virBitmapPtr sourceNodes;
>>      unsigned long long pagesize; /* kibibytes */
>> +    char *path;
> 
> Since it's "hard" to find path in sources in a simple (hah) search, can
> this be nvdimm_path or something more specific?
> 

As in 'here'...

> I would also think there'd need to be some Mem ABI Stability check added
> in virDomainMemoryDefCheckABIStability that the src/dst path's are the same.
> 
> Searching on 'nmems':
> 
>  1. Will virDomainDefPostParseMemory need any adjustment to not account
> for this type of memory or should it be included?
> 
>  2. Similar for virDomainDefGetMemoryInitial
> 
>  3. Will alignment be needed? qemuDomainAlignMemorySizes
> 
> 

[...]

John