Re: [PATCH v2 1/2] sev/i386: Introduce sev_add_kernel_loader_hashes for measured linux boot

[Philippe Mathieu-Daudé]

On 6/22/21 12:26 PM, Dov Murik wrote:
> 
> 
> On 22/06/2021 12:49, Philippe Mathieu-Daudé wrote:
>> On 6/22/21 11:44 AM, Dov Murik wrote:
>>> On 21/06/2021 23:32, Philippe Mathieu-Daudé wrote:
>>
> 
> ...
> 
>>
>>>>> +    if (!pc_system_ovmf_table_find(SEV_HASH_TABLE_RV_GUID, &data, NULL)) 
>>>>> {
>>>>
>>>> If we never use the data_len argument, can we simplify the prototype?
>>>
>>> The current uses for the OVMF reset vector GUIDed table is for simple
>>> structs with known length (secret injection page address, SEV-ES reset
>>> address, SEV table of hashes address).  But keeping the length there
>>> allows adding variable-sized entries such as strings/blobs.
>>
>> OK. Good opportunity to document the prototype declaration ;)
> 
> Yep. I'll send as a separate standalone patch.

Sure.

> P.S.
> In a previous version you mentioned you ran into issues with a qemu
> build with SEV disabled.  I tried that by modifying
> default-configs/devices/i386-softmmu.mak and uncommenting CONFIG_SEV=n
> there.  Is there a friendlier way to create such a build?

Unfortunately not yet in mainstream (distributions tune their builds,
often disabling all recent features, and enable them on a case by case
basis once it is well tested).

Thankfully Alex posted a series to add the possibility:
https://www.mail-archive.com/qemu-devel@nongnu.org/msg817710.html

> 
> I'm currently building with:
> 
>     cd build
>     ../configure --target-list=x86_64-softmmu
>     make
> 
> 
> Thanks,
> -Dov
>