On Tue, Dec 03 2024 at 18:24, Frank Li wrote:
> On Tue, Dec 03, 2024 at 11:15:27PM +0100, Thomas Gleixner wrote:
>> The fact that a MSI parent domain supports DOMAIN_BUS_DEVICE_MSI does
>> not guarantee that the parent is translation table based.
>>
>> As this is intended to be a generic library for all sorts of EP
>> implementations, there needs to be
>>
>> - either a mechanism to prevent the initialization if the underlying
>> MSI parent domain does not provide immutable messages
>
> How to know such information?
It obviously needs to be flagged somehow in the domain and the EP magic
needs to check that flag.
>>
>> - or support for endpoint specific msi_write_msg() implementations
>
> Even provide specific msi_write_msg(), write to address/data to shared
> memory.
>
> host driver:
> 1. read address/data from shared memory
> 2. write data to address.
>
> 1 and 2 is not atomic. So it can't avoid above raise conditon.
Correct. So you cannot support that case, which in turn requires to have
a mechanism to check for the immutable property.
Thanks,
tglx
