[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [RFC] arm/p2m: call iommu iotlb flush if iommu exists and enabled

> -----Original Message-----
> From: Andrii Anisov [mailto:andrii.anisov@xxxxxxxxx]
> Sent: 23 January 2019 12:40
> To: Paul Durrant <Paul.Durrant@xxxxxxxxxx>; 'Julien Grall'
> <julien.grall@xxxxxxx>; xen-devel@xxxxxxxxxxxxxxxxxxxx
> Cc: Stefano Stabellini <sstabellini@xxxxxxxxxx>; Andrii Anisov
> <andrii_anisov@xxxxxxxx>
> Subject: Re: [RFC] arm/p2m: call iommu iotlb flush if iommu exists and
> enabled
> Hello Paul,
> On 23.01.19 13:45, Paul Durrant wrote:
> > Yes, this was a mistake when moving from the old macros and need_iommu.
> Andrii is correct that need_iommu_pt_sync() is supposed to gate whether an
> explicit map/unmap is needed. The need for flush should only depend on
> has_iommu_pt().
> So I fix the commit message and send V2 for XEN 4.12.


> > There is also iommu_use_hap_pt() which is actually defined as
> has_iommu_pt(), but I wonder whether that should just go away for ARM
> since the page tables are always shared.
> Actually, the mainline currently supports only SMMU, which does share the
> page table.
> But it is not always true, we are using HW, where PT is not shared. And
> going to upstream that support one day.

Oh, ok. Leave it in then :-)


> --
> Sincerely,
> Andrii Anisov.
Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.