From: Jason Gunthorpe <jgg@nvidia.com>
To: Alexey Kardashevskiy <aik@amd.com>
Cc: Xu Yilun <yilun.xu@linux.intel.com>,
"Aneesh Kumar K.V" <aneesh.kumar@kernel.org>,
Dan Williams <dan.j.williams@intel.com>,
linux-coco@lists.linux.dev, linux-pci@vger.kernel.org,
gregkh@linuxfoundation.org, lukas@wunner.de,
suzuki.poulose@arm.com, sameo@rivosinc.com, zhiw@nvidia.com
Subject: Re: [RFC PATCH 3/3] iommufd/tsm: Add tsm_bind/unbind iommufd ioctls
Date: Tue, 10 Jun 2025 15:19:54 -0300 [thread overview]
Message-ID: <20250610181954.GK543171@nvidia.com> (raw)
In-Reply-To: <d56b85e7-3081-4c99-b3db-7adf6604951c@amd.com>
On Tue, Jun 10, 2025 at 05:05:21PM +1000, Alexey Kardashevskiy wrote:
>
>
> On 4/6/25 22:36, Jason Gunthorpe wrote:
> > On Wed, Jun 04, 2025 at 01:58:55PM +0800, Xu Yilun wrote:
> >
> > > But the p2p case may impact AMD, AMD have legacy IOMMUPT on its secure
> > > DMA path. And if you invalidate MMIO (in turn unmaps IOMMUPT) when
> > > bound, may trigger HW protection mechanism against DMA silent drop.
> >
> > As I understand AMD it sort of has a single translation and relies on
> > its RMP for security.
>
> Well, two levels with the first page table living in the guest
> memory + RMP for the second page table in the host memory.
Yes, 'level' is a bit unclear here.. Intel and ARM have four levels.
T=0 S2 lives in hypervisor memory
T=1 S2 lives in TSM memory (S-EPT)
T=0 S1 lives in guest shared memory
T=1 S1 lives in guest private memory
So compared to AMD both the RMP equivalent and the secure S2 live in
TSM memory and provide protection..
I've forgotten exactly how AMD manages to secure the IOMMU S2 in
hypervisor memory, but it seemed unique to AMD.
> > > "If a bound TDI sends a request to the root complex, and the IOMMU detects a fault caused by host
> > > configuration, the root complex fences the ASID from all further I/O to or from that guest. A host
> > > fault is either a host page table fault or an RMP check violation. ASID fencing means that the
> > > IOMMU blocks all further I/O from the root complex to the guest that the TDI was bound, and the
> > > root complex blocks all MMIO accesses by the guest. When a guest writes to MMIO, the write is
> > > silently dropped. When a guest reads from MMIO, the guest reads 1s."
> >
> > Sounds to me like the guest has to do things properly or the guest
> > gets itself killed. I wonder how feasible this really is..
>
> What does look especially worrying? So far the process has been
> pretty straightforward.
I think it makes debugging guest bugs hard if the guest explodes on an
errant DMA.
Jason
next prev parent reply other threads:[~2025-06-10 18:19 UTC|newest]
Thread overview: 173+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-05-16 5:47 [PATCH v3 00/13] PCI/TSM: Core infrastructure for PCI device security (TDISP) Dan Williams
2025-05-16 5:47 ` [PATCH v3 01/13] coco/tsm: Introduce a core device for TEE Security Managers Dan Williams
2025-06-02 13:18 ` Jason Gunthorpe
2025-06-04 0:42 ` Dan Williams
2025-06-04 1:15 ` Dan Williams
2025-06-04 12:15 ` Jason Gunthorpe
2025-06-04 12:14 ` Jason Gunthorpe
2025-06-06 3:33 ` Alexey Kardashevskiy
2025-06-06 2:09 ` Alexey Kardashevskiy
2025-05-16 5:47 ` [PATCH v3 02/13] PCI/IDE: Enumerate Selective Stream IDE capabilities Dan Williams
2025-06-17 12:16 ` Jonathan Cameron
2025-07-12 22:31 ` dan.j.williams
2025-05-16 5:47 ` [PATCH v3 03/13] PCI/TSM: Authenticate devices via platform TSM Dan Williams
2025-05-21 15:32 ` Aneesh Kumar K.V
2025-06-03 19:53 ` Dan Williams
2025-06-04 8:04 ` Aneesh Kumar K.V
2025-06-17 12:51 ` Jonathan Cameron
2025-07-12 22:07 ` dan.j.williams
2025-08-26 1:31 ` Alexey Kardashevskiy
2025-08-26 23:54 ` dan.j.williams
2025-08-27 4:44 ` Alexey Kardashevskiy
2025-08-28 19:27 ` dan.j.williams
2025-08-26 3:08 ` Alexey Kardashevskiy
2025-08-26 23:58 ` dan.j.williams
2025-08-27 5:06 ` Alexey Kardashevskiy
2025-08-26 10:22 ` Alexey Kardashevskiy
2025-08-27 0:15 ` dan.j.williams
2025-08-27 5:02 ` Alexey Kardashevskiy
2025-08-28 19:32 ` dan.j.williams
2025-05-16 5:47 ` [PATCH v3 04/13] PCI: Enable host-bridge emulation for PCI_DOMAINS_GENERIC platforms Dan Williams
2025-05-16 5:47 ` [PATCH v3 05/13] PCI: vmd: Switch to pci_bus_find_emul_domain_nr() Dan Williams
2025-05-16 5:47 ` [PATCH v3 06/13] samples/devsec: Introduce a PCI device-security bus + endpoint sample Dan Williams
2025-06-17 13:30 ` Jonathan Cameron
2025-07-13 1:58 ` dan.j.williams
2025-05-16 5:47 ` [PATCH v3 07/13] PCI: Add PCIe Device 3 Extended Capability enumeration Dan Williams
2025-06-17 13:36 ` Jonathan Cameron
2025-05-16 5:47 ` [PATCH v3 08/13] PCI/IDE: Add IDE establishment helpers Dan Williams
2025-06-17 14:04 ` Jonathan Cameron
2025-07-14 18:25 ` dan.j.williams
2025-07-03 2:59 ` Alexey Kardashevskiy
2025-05-16 5:47 ` [PATCH v3 09/13] PCI/IDE: Report available IDE streams Dan Williams
2025-05-18 12:48 ` kernel test robot
2025-06-17 14:16 ` Jonathan Cameron
2025-07-14 20:16 ` dan.j.williams
2025-05-16 5:47 ` [PATCH v3 10/13] PCI/TSM: Report active " Dan Williams
2025-06-17 14:21 ` Jonathan Cameron
2025-07-14 20:49 ` dan.j.williams
2025-05-16 5:47 ` [PATCH v3 11/13] samples/devsec: Add sample IDE establishment Dan Williams
2025-06-17 14:26 ` Jonathan Cameron
2025-07-14 20:59 ` dan.j.williams
2025-05-16 5:47 ` [PATCH v3 12/13] PCI/TSM: support TDI related operations for host TSM driver Dan Williams
2025-05-16 6:52 ` Xu Yilun
2025-05-20 7:17 ` Aneesh Kumar K.V
2025-05-21 9:35 ` Xu Yilun
2025-05-26 5:05 ` Aneesh Kumar K.V
2025-05-26 7:52 ` Alexey Kardashevskiy
2025-05-26 15:44 ` Aneesh Kumar K.V
2025-05-27 1:01 ` Alexey Kardashevskiy
2025-05-27 11:48 ` Aneesh Kumar K.V
2025-05-27 13:06 ` Jason Gunthorpe
2025-05-27 14:26 ` Aneesh Kumar K.V
2025-05-27 14:45 ` Jason Gunthorpe
2025-05-28 12:17 ` Aneesh Kumar K.V
2025-05-28 16:42 ` Jason Gunthorpe
2025-05-28 16:52 ` Jason Gunthorpe
2025-05-29 9:30 ` Xu Yilun
2025-05-29 13:43 ` Aneesh Kumar K.V
2025-05-29 14:09 ` Jason Gunthorpe
2025-05-30 3:00 ` Alexey Kardashevskiy
2025-05-30 13:21 ` Jason Gunthorpe
2025-05-29 13:49 ` Xu Yilun
2025-05-29 14:05 ` Jason Gunthorpe
2025-05-29 3:03 ` Alexey Kardashevskiy
2025-05-29 13:34 ` Aneesh Kumar K.V
2025-05-29 13:37 ` [RFC PATCH 1/3] coco: tsm: Add tsm_bind/unbind helpers Aneesh Kumar K.V (Arm)
2025-05-29 13:37 ` [RFC PATCH 2/3] iommufd/viommu: Add support to associate viommu with kvm instance Aneesh Kumar K.V (Arm)
2025-05-29 14:13 ` Jason Gunthorpe
2025-05-29 13:37 ` [RFC PATCH 3/3] iommufd/tsm: Add tsm_bind/unbind iommufd ioctls Aneesh Kumar K.V (Arm)
2025-05-29 14:32 ` Jason Gunthorpe
2025-05-30 8:33 ` Aneesh Kumar K.V
2025-05-30 18:18 ` Jason Gunthorpe
2025-05-31 16:25 ` Xu Yilun
2025-06-02 4:52 ` Alexey Kardashevskiy
2025-06-02 17:17 ` Xu Yilun
2025-06-04 1:47 ` Alexey Kardashevskiy
2025-06-04 5:02 ` Xu Yilun
2025-06-04 12:37 ` Jason Gunthorpe
2025-06-06 15:40 ` Xu Yilun
2025-06-06 16:34 ` Jason Gunthorpe
2025-06-09 4:47 ` Xu Yilun
2025-06-02 11:08 ` Aneesh Kumar K.V
2025-06-02 16:25 ` Xu Yilun
2025-06-02 16:48 ` Jason Gunthorpe
2025-06-03 4:05 ` Xu Yilun
2025-06-03 12:11 ` Jason Gunthorpe
2025-06-04 5:58 ` Xu Yilun
2025-06-04 12:36 ` Jason Gunthorpe
2025-06-05 3:05 ` Xu Yilun
2025-06-10 7:05 ` Alexey Kardashevskiy
2025-06-10 18:19 ` Jason Gunthorpe [this message]
2025-06-11 1:26 ` Alexey Kardashevskiy
2025-06-10 4:47 ` Alexey Kardashevskiy
2025-06-10 18:21 ` Jason Gunthorpe
2025-06-12 4:15 ` Xu Yilun
2025-06-03 5:00 ` Aneesh Kumar K.V
2025-06-03 10:50 ` Xu Yilun
2025-06-03 12:14 ` Jason Gunthorpe
2025-06-04 5:31 ` Xu Yilun
2025-06-04 12:31 ` Jason Gunthorpe
2025-06-05 3:25 ` Xu Yilun
2025-06-05 14:54 ` Jason Gunthorpe
2025-06-09 6:10 ` Xu Yilun
2025-06-09 16:42 ` Suzuki K Poulose
2025-06-09 18:07 ` Jason Gunthorpe
2025-06-10 7:31 ` Alexey Kardashevskiy
2025-06-12 5:44 ` Xu Yilun
2025-06-03 12:18 ` Jason Gunthorpe
2025-06-04 1:06 ` Dan Williams
2025-06-04 12:18 ` Jason Gunthorpe
2025-06-02 12:47 ` Jason Gunthorpe
2025-06-03 3:47 ` Xu Yilun
2025-06-03 12:08 ` Jason Gunthorpe
2025-06-04 6:39 ` Xu Yilun
2025-06-04 12:39 ` Jason Gunthorpe
2025-06-05 1:56 ` Xu Yilun
2025-07-15 10:29 ` Xu Yilun
2025-07-15 13:09 ` Jason Gunthorpe
2025-07-16 15:41 ` Xu Yilun
2025-07-16 16:31 ` Jason Gunthorpe
2025-07-17 8:28 ` Xu Yilun
2025-07-17 12:43 ` Jason Gunthorpe
2025-07-18 9:15 ` Xu Yilun
2025-07-18 12:26 ` Jason Gunthorpe
2025-07-20 2:37 ` Xu Yilun
2025-05-30 2:44 ` [PATCH v3 12/13] PCI/TSM: support TDI related operations for host TSM driver Alexey Kardashevskiy
2025-05-27 10:25 ` Suzuki K Poulose
2025-06-03 22:47 ` Dan Williams
2025-06-04 1:35 ` Alexey Kardashevskiy
2025-06-04 1:52 ` Dan Williams
2025-06-04 1:54 ` Dan Williams
2025-06-05 10:56 ` Alexey Kardashevskiy
2025-06-07 1:56 ` Dan Williams
2025-06-11 4:40 ` Alexey Kardashevskiy
2025-06-13 3:06 ` Dan Williams
2025-06-03 22:40 ` Dan Williams
2025-05-19 10:20 ` Alexey Kardashevskiy
2025-05-20 20:12 ` Dan Williams
2025-05-21 9:28 ` Xu Yilun
2025-05-26 8:08 ` Alexey Kardashevskiy
2025-05-29 14:20 ` Xu Yilun
2025-05-30 2:54 ` Alexey Kardashevskiy
2025-05-31 15:26 ` Xu Yilun
2025-06-02 4:51 ` Alexey Kardashevskiy
2025-06-02 18:51 ` Xu Yilun
2025-06-03 19:12 ` Dan Williams
2025-07-07 7:17 ` Aneesh Kumar K.V
2025-05-20 5:20 ` Aneesh Kumar K.V
2025-05-20 21:12 ` Dan Williams
2025-05-16 5:47 ` [PATCH v3 13/13] PCI/TSM: Add Guest TSM Support Dan Williams
2025-05-19 10:20 ` Alexey Kardashevskiy
2025-05-20 21:11 ` Dan Williams
2025-05-22 4:07 ` Alexey Kardashevskiy
2025-06-03 22:26 ` Dan Williams
2025-06-03 22:33 ` Jason Gunthorpe
2025-06-10 8:31 ` Alexey Kardashevskiy
2025-07-11 23:04 ` dan.j.williams
2025-05-20 9:25 ` Aneesh Kumar K.V
2025-05-20 21:27 ` Dan Williams
2025-05-20 11:00 ` Aneesh Kumar K.V
2025-05-20 21:31 ` Dan Williams
2025-06-03 19:07 ` Dan Williams
2025-05-21 15:03 ` Xu Yilun
2025-06-03 19:20 ` Dan Williams
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=20250610181954.GK543171@nvidia.com \
--to=jgg@nvidia.com \
--cc=aik@amd.com \
--cc=aneesh.kumar@kernel.org \
--cc=dan.j.williams@intel.com \
--cc=gregkh@linuxfoundation.org \
--cc=linux-coco@lists.linux.dev \
--cc=linux-pci@vger.kernel.org \
--cc=lukas@wunner.de \
--cc=sameo@rivosinc.com \
--cc=suzuki.poulose@arm.com \
--cc=yilun.xu@linux.intel.com \
--cc=zhiw@nvidia.com \
/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;
as well as URLs for NNTP newsgroup(s).