On 08/02/2019 17:47, Roger Pau Monné wrote:
> On Fri, Feb 08, 2019 at 05:15:22PM +0100, Sander Eikelenboom wrote:
>> On 08/02/2019 16:10, Roger Pau Monné wrote:
>>> On Fri, Jan 25, 2019 at 07:44:40PM +0100, Sander Eikelenboom wrote:
>>>> On 25/01/2019 15:38, Roger Pau Monné wrote:
>>>>> On Thu, Jan 24, 2019 at 01:04:31PM +0100, Roger Pau Monné wrote:
>>>>> Sorry, fixing that error took longer than expected, and requires
>>>>> modifying quite a lot of code, so I'm not sure whether it's something
>>>>> to consider for 4.12, I have to think about it.
>>>>
>>>> I understand, especially since PVH dom0 is marked as experimental.
>>>>
>>>>> In the meantime, can you please test the following branch:
>>>>>
>>>>> git://xenbits.xen.org/people/royger/xen.git iommu-fixes-v2
>>>>>
>>>>> I've been able to successfully create a PVH guest from a PVH dom0 on
>>>>> AMD hardware using this branch.
>>>>
>>>> On the other hand, with a quick test I can confirm that booting a PVH guest from a PVH dom0 now works for me as well ! 
>>>> (and booting this build as PV dom0, with my normal PVH/HVM mix of guests still works, no regressions for me so far)
>>
>>
>>  
>>> Sorry for bothering you again, but could you give the following branch
>>> a test:
>>
>> No problem, happy to keep testing until it works and is in good enough shape to get committed.
>>
>> And fortunately you asked, because unfortunately it doesn't boot as pvhdom0, serial log is attached.
> 
> Thanks!
> 
> Can you try with the following debug patch on top? This should print a
> message before hitting the assert, hopefully giving us more
> information.
> 
> Roger.

Sure, I was also missing a sync_console on the Xen cmdline,
serial log attached.

--
Sander

> ---8<---
> diff --git a/xen/arch/x86/mm/p2m-pt.c b/xen/arch/x86/mm/p2m-pt.c
> index 5ad7a36269..bf647e7d26 100644
> --- a/xen/arch/x86/mm/p2m-pt.c
> +++ b/xen/arch/x86/mm/p2m-pt.c
> @@ -648,8 +648,13 @@ p2m_pt_set_entry(struct p2m_domain *p2m, gfn_t gfn_, mfn_t mfn,
>                     !rangeset_overlaps_range(mmio_ro_ranges, mfn_x(mfn),
>                                              mfn_x(mfn) + PFN_DOWN(MB(2))));
>          else
> +        {
> +            if ( !mfn_valid(mfn) && (!mfn_eq(mfn, INVALID_MFN) ||
> +                                     !p2m_allows_invalid_mfn(p2mt)) )
> +                printk("mfn: %#lx type: %d\n", mfn_x(mfn), p2mt);
>              ASSERT(mfn_valid(mfn) || (mfn_eq(mfn, INVALID_MFN) &&
>                                        p2m_allows_invalid_mfn(p2mt)));
> +        }
>          l2e_content = mfn_valid(mfn) || p2m_allows_invalid_mfn(p2mt)
>              ? p2m_l2e_from_pfn(mfn_x(mfn),
>                                 p2m_type_to_flags(p2m, p2mt, mfn, 1))
>