Re: [PATCH] s390x: Properly fetch and test the short psw on diag308 subc 0/1

[David Hildenbrand]

> Am 11.11.2019 um 14:52 schrieb Janosch Frank <address@hidden>:
> 
> On 11/5/19 9:07 PM, David Hildenbrand wrote:
>>> On 05.11.19 20:34, Janosch Frank wrote:
>>> On 11/5/19 8:29 PM, David Hildenbrand wrote:
>>>> On 05.11.19 19:44, Janosch Frank wrote:
> [...]
>> Note: For TCG we miss many of these validity checks. For KVM, most 
>> should be triggered when running the VCPU AFAIK (that means, we don't 
>> have to check for any other scenarios here). Checking for the special 
>> case as given in the programming note should be sufficient.
>> 
>> 
>> I'll have to think about how to best handle that for TCG (mazbe what I 
>> proposed works). We could ignore TCG for now and add a TODO. Then, just 
>> wrap the exception in a "if (kvm_enabled())". You could also document 
>> why we only have to check for this very specific bit and not the other 
>> bits (handled by HW later).
>> 
> 
> After some discussion I got an interesting answer:
> For KVM we need to expand the short psw and invert bit 12.
> The next SIE entry will automatically report a PIC 6.

He, that‘s a nice trick.

> 
> I'd propose that I'll fix this patch and send a v2 and you can fix TCG :)

Yes please. Perform the conversion unconditionally and add a comment - we 
should add the same checks when running the TCG main loop after modifying the 
PSW mask.

>