[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <df9c87c7-b7d8-4de7-99a8-ad5ae133bf36@amd.com>
Date: Thu, 2 Jun 2016 15:38:43 -0500
From: Suravee Suthikulanit <suravee.suthikulpanit@....com>
To: Joerg Roedel <joro@...tes.org>
CC: <pbonzini@...hat.com>, <rkrcmar@...hat.com>, <bp@...en8.de>,
<gleb@...nel.org>, <alex.williamson@...hat.com>,
<kvm@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
<wei@...hat.com>, <sherry.hurwitz@....com>
Subject: Re: [PART2 RFC v1 1/9] iommu/amd: Detect and enable guest vAPIC
support
On 5/9/2016 6:49 AM, Joerg Roedel wrote:
> On Fri, Apr 08, 2016 at 07:49:22AM -0500, Suthikulpanit, Suravee wrote:
>> From: Suravee Suthikulpanit <suravee.suthikulpanit@....com>
>>
>> This patch introduces a new IOMMU driver parameter, amd_iommu_guest_ir,
>> which can be used to specify different interrupt remapping mode for
>> passthrough devices to VM guest:
>> * legacy: Legacy interrupt remapping mode (w/ 32-bit IRTE)
>> * ga : Guest vAPIC interrupt remapping mode (w/ 128-bit IRTE)
>>
>> Note that the GA mode also supports legacy interrupt remapping
>> for non-passthrough devices with the 128-bit IRTE.
>
> Does this need to be under user control? The code can just check what
> the hardware supports and use the 128bit IRTEs if supported, no?
>
> Joerg
It does not need to be signified by user.
Currently, if the MMIO Offset 30h[GASup] bit (of IOMMU Extended Feature
Register) is set, the driver should default to using the 128bit IRTE by
setting MMIO Offset 0018h[GAEn] bit (of IOMMU Control Register). The
default is also enabling GA mode (by setting MMIO 0018h[GAEn] if MMIO
0030h[GASup] is set).
However, if SVM AVIC is not enabled, or if the AVIC HW cannot support
the type of interrupt (e.g. multicast/broadcast), it falls back to use
legacy interrupt remapping mode w/ 128-bit IRTE.
This option is intended for the case when we want to force IOMMU to use
legacy interrupt remapping (hence no need for 128-bit IRTE).
I will improve on the documentation in the next patch series.
Thanks,
Suravee
Powered by blists - more mailing lists