Re: [Xen-devel] [PATCH v2 4/4] x86/mm/p2m: stop checking for IOMMU shared page tables in mmio_order()

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 04.12.18 at 16:22, <Paul.Durrant@xxxxxxxxxx> wrote:
>>  -----Original Message-----
>> From: Jan Beulich [mailto:JBeulich@xxxxxxxx]
>> Sent: 04 December 2018 15:21
>> To: Paul Durrant <Paul.Durrant@xxxxxxxxxx>
>> Cc: Andrew Cooper <Andrew.Cooper3@xxxxxxxxxx>; Roger Pau Monne
>> <roger.pau@xxxxxxxxxx>; Wei Liu <wei.liu2@xxxxxxxxxx>; George Dunlap
>> <George.Dunlap@xxxxxxxxxx>; xen-devel <xen-devel@xxxxxxxxxxxxxxxxxxxx>
>> Subject: Re: [PATCH v2 4/4] x86/mm/p2m: stop checking for IOMMU shared
>> page tables in mmio_order()
>> 
>> >>> On 03.12.18 at 18:40, <paul.durrant@xxxxxxxxxx> wrote:
>> > Now that the iommu_map() and iommu_unmap() operations take an order
>> > parameter and elide flushing there's no strong reason why modifying MMIO
>> > ranges in the p2m should be restricted to a 4k granularity simply
>> because
>> > the IOMMU is enabled but shared page tables are not in operation.
>> 
>> I'm afraid the two improvements are not enough for this restriction
>> to be lifted: There's still no preemption in the processing of the
>> higher order values.
> 
> Why? 1G orders are already ruled out and testing shows that 2M orders cause 
> no problems on EPYC systems.

Hmm, yes, agreed.

Jan



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel