[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20260114190715.GC961588@nvidia.com>
Date: Wed, 14 Jan 2026 15:07:15 -0400
From: Jason Gunthorpe <jgg@...dia.com>
To: Samiullah Khawaja <skhawaja@...gle.com>
Cc: "Tian, Kevin" <kevin.tian@...el.com>,
Baolu Lu <baolu.lu@...ux.intel.com>, Joerg Roedel <joro@...tes.org>,
Will Deacon <will@...nel.org>, Robin Murphy <robin.murphy@....com>,
Dmytro Maluka <dmaluka@...omium.org>,
"iommu@...ts.linux.dev" <iommu@...ts.linux.dev>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 3/3] iommu/vt-d: Rework hitless PASID entry replacement
On Wed, Jan 14, 2026 at 10:51:02AM -0800, Samiullah Khawaja wrote:
> There is no immediate use case for first level, as I understand we use
> second level for all our VM use cases. But looking at the code and
> specs, the first level is used for pasid compatible domains (using
> IOMMU_HWPT_ALLOC_PASID flag) and it will not be replaceable
> hitlessly..
If the VMM allocates using the NEST_PARENT flag it will always get a
S2, and that is a reasonable thing for the VMM to do.
Jason
Powered by blists - more mailing lists