Re: [PATCH v2 12/14] arm64: realm: Support nonsecure ITS emulation shared

[Steven Price]

On 15/05/2024 12:01, Catalin Marinas wrote:
> On Fri, Apr 12, 2024 at 09:42:11AM +0100, Steven Price wrote:
>> @@ -198,6 +201,33 @@ static DEFINE_IDA(its_vpeid_ida);
>>  #define gic_data_rdist_rd_base()	(gic_data_rdist()->rd_base)
>>  #define gic_data_rdist_vlpi_base()	(gic_data_rdist_rd_base() + SZ_128K)
>>  
>> +static struct page *its_alloc_shared_pages_node(int node, gfp_t gfp,
>> +						unsigned int order)
>> +{
>> +	struct page *page;
>> +
>> +	if (node == NUMA_NO_NODE)
>> +		page = alloc_pages(gfp, order);
>> +	else
>> +		page = alloc_pages_node(node, gfp, order);
> 
> I think you can just call alloc_pages_node() in both cases. This
> function takes care of the NUMA_NO_NODE case itself.
> 
>> +
>> +	if (page)
>> +		set_memory_decrypted((unsigned long)page_address(page),
>> +				     1 << order);
>> +	return page;
>> +}
>> +
>> +static struct page *its_alloc_shared_pages(gfp_t gfp, unsigned int order)
>> +{
>> +	return its_alloc_shared_pages_node(NUMA_NO_NODE, gfp, order);
>> +}
>> +
>> +static void its_free_shared_pages(void *addr, unsigned int order)
>> +{
>> +	set_memory_encrypted((unsigned long)addr, 1 << order);
>> +	free_pages((unsigned long)addr, order);
>> +}
> 
> More of a nitpick on the naming: Are these functions used by the host as
> well? The 'shared' part of the name does not make much sense, so maybe
> just call them its_alloc_page() etc.

Yes, the host is emulating this so the pages need to be in the shared
region. However this is only for realms, for a normal guest this
functions obviously aren't "sharing" with anything - so perhaps dropping
the 'shared' part makes sense.

>> @@ -3432,7 +3468,16 @@ static struct its_device *its_create_device(struct its_node *its, u32 dev_id,
>>  	nr_ites = max(2, nvecs);
>>  	sz = nr_ites * (FIELD_GET(GITS_TYPER_ITT_ENTRY_SIZE, its->typer) + 1);
>>  	sz = max(sz, ITS_ITT_ALIGN) + ITS_ITT_ALIGN - 1;
>> -	itt = kzalloc_node(sz, GFP_KERNEL, its->numa_node);
>> +	itt_order = get_order(sz);
>> +	page = its_alloc_shared_pages_node(its->numa_node,
>> +					   GFP_KERNEL | __GFP_ZERO,
>> +					   itt_order);
> 
> How much do we waste by going for a full page always if this is going to
> be used on the host?

sz is a minimum of ITS_ITT_ALIGN*2-1 - which is 511 bytes. So
potentially PAGE_SIZE-512 bytes could be wasted here (minus kmalloc
overhead).

>> +	if (!page)
>> +		return NULL;
>> +	itt = (void *)page_address(page);
> 
> page_address() has the void * type already.
> 

Indeed, the cast is pointless. I'll remove.

Thanks,

Steve