Re: [PATCH 1/5] x86/pv: Fix the determiniation of whether to inject #DB

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 14/09/2023 3:40 pm, Jan Beulich wrote:
> On 13.09.2023 01:21, Andrew Cooper wrote:
>> We long ago fixed the emulator to not inject exceptions behind our back.
>> Therefore, assert that that a PV event (including interrupts, because that
>> would be buggy too) isn't pending, rather than skipping the #DB injection if
>> one is.
>>
>> On the other hand, the io_emul() stubs which use X86EMUL_DONE rather than
>> X86EMUL_OKAY may have pending breakpoints to inject after the IO access is
>> complete, not to mention a pending singlestep.
> If you look at the uses of X86EMUL_DONE you'll see that this error code is
> not intended to ever come back from the emulator. It's solely used to
> communicate between hooks and the core emulator. Therefore I think this
> part of the description and the added case label are wrong here. With them
> dropped again ...

Oh.  I see that now you've pointed it out, but it's far from clear.

I'd suggest that we extend the the debug wrapper for x86_emulate() with
an assertion to this effect.  It also has a knock-on effect in later
patches.

With the DONE part dropped, this probably wants merging into patch 4. 
Thoughts?

~Andrew