From: Jacob Pan <jacob.pan@linux.microsoft.com>
To: Jason Gunthorpe <jgg@nvidia.com>
Cc: iommu@lists.linux.dev,
Jean-Philippe Brucker <jean-philippe@linaro.org>,
Joerg Roedel <joro@8bytes.org>,
Robin Murphy <robin.murphy@arm.com>,
virtualization@lists.linux.dev, Will Deacon <will@kernel.org>,
Eric Auger <eric.auger@redhat.com>,
patches@lists.linux.dev, jacob.pan@linux.microsoft.com
Subject: Re: [PATCH 0/5] Convert virtio-iommu to domain_alloc_paging()
Date: Tue, 11 Feb 2025 16:41:51 -0800 [thread overview]
Message-ID: <20250211164151.723cfce0@DESKTOP-0403QTC.> (raw)
In-Reply-To: <0-v1-91eed9c8014a+53a37-iommu_virtio_domains_jgg@nvidia.com>
Hi Jason,
On Fri, 7 Feb 2025 10:46:00 -0400
Jason Gunthorpe <jgg@nvidia.com> wrote:
> virtio-iommu and fsl_pamu are the only two drivers left that still
> rely on ops.domain_alloc()
>
> Update virtio-iommu to have a global static identity domain, implement
> domain_alloc_paging(), and finalize its domain during allocation
> instead of on first attach.
>
> As virtio-iommu was the last real iommu driver using domain_alloc()
> update the core code to isolate the remains to fsl_pamu.
>
> I tested this using qemu x86 7.0.0 and 9.2.0 with a mlx5 VFIO PCI
> device. The kernel behaves the same after this series.
>
> However, there seem to be some unrelated qemu bugs. 7.0.0 does not
> have working VIRTIO_IOMMU_F_BYPASS_CONFIG and neither qemu works with
> an IDENTITY domain using !VIRTIO_IOMMU_F_BYPASS_CONFIG. It prints:
>
> qemu-system-x86_64: iommu has granularity incompatible with target AS
> qemu-system-x86_64: iommu map to non memory area 80000000
> qemu-system-x86_64: iommu map to non memory area c0000000
> qemu-system-x86_64: iommu map to non memory area e0000000
> qemu-system-x86_64: iommu map to non memory area f0000000
> qemu-system-x86_64: iommu map to non memory area f8000000
> qemu-system-x86_64: iommu map to non memory area fc000000
> qemu-system-x86_64: iommu map to non memory area fe000000
> qemu-system-x86_64: iommu map to non memory area fe800000
> qemu-system-x86_64: iommu map to non memory area 0
> qemu-system-x86_64: iommu map to non memory area fef00000
> qemu-system-x86_64: iommu map to non memory area ff000000
> qemu-system-x86_64: iommu has granularity incompatible with target AS
> qemu-system-x86_64: iommu map to non memory area 200000000
> qemu-system-x86_64: iommu map to non memory area 400000000
> qemu-system-x86_64: iommu map to non memory area 800000000
> qemu-system-x86_64: iommu map to non memory area 1000000000
> qemu-system-x86_64: iommu map to non memory area 2000000000
> qemu-system-x86_64: iommu map to non memory area 4000000000
I see the same on arm64 with v9.0, assigned an ixgbe nic via VFIO.
qemu-system-aarch64: iommu map to non memory area 0
qemu-system-aarch64: iommu map to non memory area 8100000
qemu-system-aarch64: iommu map to non memory area 8200000
qemu-system-aarch64: iommu map to non memory area 8400000
qemu-system-aarch64: iommu map to non memory area 8800000
qemu-system-aarch64: iommu map to non memory area 0
qemu-system-aarch64: iommu map to non memory area 0
qemu-system-aarch64: iommu map to non memory area c000000
qemu-system-aarch64: iommu map to non memory area 10000000
qemu-system-aarch64: iommu map to non memory area 20000000
qemu-system-aarch64: iommu has granularity incompatible with target AS
qemu-system-aarch64: iommu map to non memory area 200000000
qemu-system-aarch64: iommu map to non memory area 400000000
qemu-system-aarch64: iommu map to non memory area 800000000
qemu-system-aarch64: iommu map to non memory area 1000000000
qemu-system-aarch64: iommu map to non memory area 2000000000
qemu-system-aarch64: iommu map to non memory area 4000000000
qemu-system-aarch64: iommu map to non memory area 8000000000
qemu-system-aarch64: iommu map to non memory area 10000000000
qemu-system-aarch64: iommu map to non memory area 20000000000
qemu-system-aarch64: iommu map to non memory area 40000000000
qemu-system-aarch64: iommu map to non memory area 80000000000
qemu-system-aarch64: iommu map to non memory area 100000000000
qemu-system-aarch64: iommu map to non memory area 200000000000
qemu-system-aarch64: iommu map to non memory area 400000000000
qemu-system-aarch64: iommu map to non memory area 800000000000
> Jason Gunthorpe (5):
> iommu/virtio: Break out bypass identity support into a global static
> iommu: Add domain_alloc_identity()
> iommu/virtio: Move to domain_alloc_paging()
> iommu: Do not call domain_alloc() in iommu_sva_domain_alloc()
> iommu: Hide ops.domain_alloc behind CONFIG_FSL_PAMU
>
> drivers/iommu/iommu-sva.c | 15 ++-
> drivers/iommu/iommu.c | 16 +--
> drivers/iommu/virtio-iommu.c | 200
> +++++++++++++++++++++-------------- include/linux/iommu.h |
> 10 +- 4 files changed, 141 insertions(+), 100 deletions(-)
>
>
> base-commit: 2014c95afecee3e76ca4a56956a936e23283f05b
next prev parent reply other threads:[~2025-02-12 0:41 UTC|newest]
Thread overview: 37+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-02-07 14:46 [PATCH 0/5] Convert virtio-iommu to domain_alloc_paging() Jason Gunthorpe
2025-02-07 14:46 ` [PATCH 1/5] iommu/virtio: Break out bypass identity support into a global static Jason Gunthorpe
2025-02-12 0:43 ` Jacob Pan
2025-02-18 18:21 ` Jason Gunthorpe
2025-02-21 11:35 ` Jean-Philippe Brucker
2025-02-24 19:37 ` Jason Gunthorpe
2025-02-07 14:46 ` [PATCH 2/5] iommu: Add domain_alloc_identity() Jason Gunthorpe
2025-02-12 13:56 ` Robin Murphy
2025-02-12 14:03 ` Jason Gunthorpe
2025-02-12 14:16 ` Robin Murphy
2025-02-12 14:45 ` Jason Gunthorpe
2025-02-07 14:46 ` [PATCH 3/5] iommu/virtio: Move to domain_alloc_paging() Jason Gunthorpe
2025-02-12 19:22 ` Jacob Pan
2025-02-12 23:30 ` Jason Gunthorpe
2025-02-13 5:47 ` Jacob Pan
2025-02-18 20:01 ` Jason Gunthorpe
[not found] ` <67ad876d.170a0220.3c21dc.85ceSMTPIN_ADDED_BROKEN@mx.google.com>
2025-02-13 9:46 ` Jean-Philippe Brucker
2025-02-13 17:03 ` Yu Zhang
2025-02-13 18:09 ` Jean-Philippe Brucker
2025-02-19 9:39 ` Yu Zhang
2025-02-19 10:35 ` Jean-Philippe Brucker
2025-02-19 11:11 ` Yu Zhang
2025-02-19 11:57 ` Jean-Philippe Brucker
2025-02-19 13:10 ` Yi Liu
2025-02-20 2:58 ` Baolu Lu
2025-02-20 3:44 ` Yu Zhang
2025-02-07 14:46 ` [PATCH 4/5] iommu: Do not call domain_alloc() in iommu_sva_domain_alloc() Jason Gunthorpe
2025-02-07 14:46 ` [PATCH 5/5] iommu: Hide ops.domain_alloc behind CONFIG_FSL_PAMU Jason Gunthorpe
2025-02-12 0:41 ` Jacob Pan [this message]
2025-02-12 12:50 ` [PATCH 0/5] Convert virtio-iommu to domain_alloc_paging() Jason Gunthorpe
2025-02-12 18:50 ` Jacob Pan
2025-02-12 20:10 ` Robin Murphy
2025-02-21 11:42 ` Jean-Philippe Brucker
2025-02-24 19:39 ` Jason Gunthorpe
[not found] ` <67abee53.170a0220.154671.ae28SMTPIN_ADDED_BROKEN@mx.google.com>
2025-02-12 11:58 ` Jean-Philippe Brucker
2025-02-12 17:05 ` Jacob Pan
[not found] ` <67acd4e2.630a0220.365aab.e098SMTPIN_ADDED_BROKEN@mx.google.com>
2025-02-12 19:16 ` Jean-Philippe Brucker
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20250211164151.723cfce0@DESKTOP-0403QTC. \
--to=jacob.pan@linux.microsoft.com \
--cc=eric.auger@redhat.com \
--cc=iommu@lists.linux.dev \
--cc=jean-philippe@linaro.org \
--cc=jgg@nvidia.com \
--cc=joro@8bytes.org \
--cc=patches@lists.linux.dev \
--cc=robin.murphy@arm.com \
--cc=virtualization@lists.linux.dev \
--cc=will@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox