Re: Linux 5.0 regression: BUG: unable to handle kernel paging request at ffff888023e26778 RIP: e030:move_page_tables+0x7c1/0xae0

[Juergen Gross]

On 09/02/2019 19:51, Sander Eikelenboom wrote:
> On 09/02/2019 19:48, Juergen Gross wrote:
>> On 09/02/2019 19:45, Sander Eikelenboom wrote:
>>> On 09/02/2019 09:26, Sander Eikelenboom wrote:
>>>> L.S.,
>>>>
>>>>
>>>> While testing a Linux 5.0-rc5-ish kernel (pull of yesterday) with some additional patches for
>>>> already reported other issues i came across the issue below which i haven't seen with 4.20.x
>>>>
>>>> I haven't got a reproducer so i might be hard to hit it again, 
>>>> system is AMD and this is from the host kernel running under
>>>> the Xen hypervisor might it matter.
>>>
>>>> --
>>>>
>>>> Sander
>>>
>>> Hi Boris / Juergen,
>>>
>>> The commit causing this is:
>>> 2c91bd4a4e2e530582d6fd643ea7b86b27907151 mm: speed up mremap by 20x on large regions
>>>
>>> Since it seems there haven't been any other reports about this .. 
>>> could it be this doesn't specifically work well with a Xen PVH dom0 ?
>>
>> PVH? Not PV?
> 
> Ah sorry, indeed PV !

Okay, found the problem. set_pmd_at() is used in above commit for
writing a PMD entry with a page table address. In all other places
it is used only for huge pages, which are not possible in PV guests.

I'll send a patch on Monday.


Juergen