Re: [PATCH] dma-direct: Force unencrypted DMA under SME for certain DMA masks

[Lendacky, Thomas]

On 7/24/19 10:55 AM, Kirill A. Shutemov wrote:
> On Wed, Jul 10, 2019 at 07:01:19PM +0000, Lendacky, Thomas wrote:
>> @@ -351,6 +355,32 @@ bool sev_active(void)
>>  }
>>  EXPORT_SYMBOL(sev_active);
>>  
>> +/* Override for DMA direct allocation check - ARCH_HAS_FORCE_DMA_UNENCRYPTED */
>> +bool force_dma_unencrypted(struct device *dev)
>> +{
>> +	/*
>> +	 * For SEV, all DMA must be to unencrypted addresses.
>> +	 */
>> +	if (sev_active())
>> +		return true;
>> +
>> +	/*
>> +	 * For SME, all DMA must be to unencrypted addresses if the
>> +	 * device does not support DMA to addresses that include the
>> +	 * encryption mask.
>> +	 */
>> +	if (sme_active()) {
>> +		u64 dma_enc_mask = DMA_BIT_MASK(__ffs64(sme_me_mask));
>> +		u64 dma_dev_mask = min_not_zero(dev->coherent_dma_mask,
>> +						dev->bus_dma_mask);
>> +
>> +		if (dma_dev_mask <= dma_enc_mask)
>> +			return true;
> 
> Hm. What is wrong with the dev mask being equal to enc mask? IIUC, it
> means that device mask is wide enough to cover encryption bit, doesn't it?

Not really...  it's the way DMA_BIT_MASK works vs bit numbering. Let's say
that sme_me_mask has bit 47 set. __ffs64 returns 47 and DMA_BIT_MASK(47)
will generate a mask without bit 47 set (0x7fffffffffff). So the check
will catch anything that does not support at least 48-bit DMA.

Thanks,
Tom

> 
>> +	}
>> +
>> +	return false;
>> +}
>