From: Alexey Kardashevskiy <aik@amd.com>
To: <kvm@vger.kernel.org>
Cc: <iommu@lists.linux.dev>, <linux-coco@lists.linux.dev>,
<linux-pci@vger.kernel.org>,
Suravee Suthikulpanit <suravee.suthikulpanit@amd.com>,
Alex Williamson <alex.williamson@redhat.com>,
Dan Williams <dan.j.williams@intel.com>,
<pratikrajesh.sampat@amd.com>, <michael.day@amd.com>,
<david.kaplan@amd.com>, <dhaval.giani@amd.com>,
Santosh Shukla <santosh.shukla@amd.com>,
Tom Lendacky <thomas.lendacky@amd.com>,
Michael Roth <michael.roth@amd.com>,
"Alexander Graf" <agraf@suse.de>,
Nikunj A Dadhania <nikunj@amd.com>,
Vasant Hegde <vasant.hegde@amd.com>,
Lukas Wunner <lukas@wunner.de>,
Alexey Kardashevskiy <aik@amd.com>
Subject: [RFC PATCH 14/21] RFC: iommu/iommufd/amd: Add IOMMU_HWPT_TRUSTED flag, tweak DTE's DomainID, IOTLB
Date: Fri, 23 Aug 2024 23:21:28 +1000 [thread overview]
Message-ID: <20240823132137.336874-15-aik@amd.com> (raw)
In-Reply-To: <20240823132137.336874-1-aik@amd.com>
AMD IOMMUs use a device table where one entry (DTE) describes IOMMU
setup per a PCI BDFn. DMA accesses via these DTEs are always
unencrypted.
In order to allow DMA to/from private memory, AMD IOMMUs use another
memory structure called "secure device table" which entries (sDTEs)
are similar to DTE and contain configuration for private DMA operations.
The sDTE table is in the private memory and is managed by the PSP on
behalf of a SNP VM. So the host OS does not have access to it and
does not need to manage it.
However if sDTE is enabled, some fields of a DTE are now marked as
reserved in a DTE and managed by an sDTE instead (such as DomainID),
other fields need to stay in sync (IR/IW).
Mark IOMMU HW page table with a flag saying that the memory is
backed by KVM (effectively MEMFD).
Skip setting the DomainID in DTE. Enable IOTLB enable (bit 96) to
match what the PSP writes to sDTE.
Signed-off-by: Alexey Kardashevskiy <aik@amd.com>
---
drivers/iommu/amd/amd_iommu_types.h | 2 ++
include/uapi/linux/iommufd.h | 1 +
drivers/iommu/amd/iommu.c | 20 ++++++++++++++++++--
drivers/iommu/iommufd/hw_pagetable.c | 4 ++++
4 files changed, 25 insertions(+), 2 deletions(-)
diff --git a/drivers/iommu/amd/amd_iommu_types.h b/drivers/iommu/amd/amd_iommu_types.h
index 2b76b5dedc1d..cf435c1f2839 100644
--- a/drivers/iommu/amd/amd_iommu_types.h
+++ b/drivers/iommu/amd/amd_iommu_types.h
@@ -588,6 +588,8 @@ struct protection_domain {
struct mmu_notifier mn; /* mmu notifier for the SVA domain */
struct list_head dev_data_list; /* List of pdom_dev_data */
+
+ u32 flags;
};
/*
diff --git a/include/uapi/linux/iommufd.h b/include/uapi/linux/iommufd.h
index 4dde745cfb7e..c5536686b0b1 100644
--- a/include/uapi/linux/iommufd.h
+++ b/include/uapi/linux/iommufd.h
@@ -364,6 +364,7 @@ enum iommufd_hwpt_alloc_flags {
IOMMU_HWPT_ALLOC_NEST_PARENT = 1 << 0,
IOMMU_HWPT_ALLOC_DIRTY_TRACKING = 1 << 1,
IOMMU_HWPT_FAULT_ID_VALID = 1 << 2,
+ IOMMU_HWPT_TRUSTED = 1 << 3,
};
/**
diff --git a/drivers/iommu/amd/iommu.c b/drivers/iommu/amd/iommu.c
index b19e8c0f48fa..e2f8fb79ee53 100644
--- a/drivers/iommu/amd/iommu.c
+++ b/drivers/iommu/amd/iommu.c
@@ -1930,7 +1930,20 @@ static void set_dte_entry(struct amd_iommu *iommu,
}
flags &= ~DEV_DOMID_MASK;
- flags |= domid;
+
+ if (dev_data->dev->tdi_enabled && (domain->flags & IOMMU_HWPT_TRUSTED)) {
+ /*
+ * Do hack for VFIO with TSM enabled.
+ * This runs when VFIO is being bound to a device and before TDI is bound.
+ * Ideally TSM should change DTE only when TDI is bound.
+ * Probably better test for (domain->domain.type & __IOMMU_DOMAIN_DMA_API)
+ */
+ dev_info(dev_data->dev, "Skip DomainID=%x and set bit96\n", domid);
+ flags |= 1ULL << (96 - 64);
+ } else {
+ //dev_info(dev_data->dev, "Not skip DomainID=%x and not set bit96\n", domid);
+ flags |= domid;
+ }
old_domid = dev_table[devid].data[1] & DEV_DOMID_MASK;
dev_table[devid].data[1] = flags;
@@ -2413,6 +2426,8 @@ static struct iommu_domain *do_iommu_domain_alloc(unsigned int type,
if (dirty_tracking)
domain->domain.dirty_ops = &amd_dirty_ops;
+
+ domain->flags = flags;
}
return &domain->domain;
@@ -2437,7 +2452,8 @@ amd_iommu_domain_alloc_user(struct device *dev, u32 flags,
{
unsigned int type = IOMMU_DOMAIN_UNMANAGED;
- if ((flags & ~IOMMU_HWPT_ALLOC_DIRTY_TRACKING) || parent || user_data)
+ if ((flags & ~(IOMMU_HWPT_ALLOC_DIRTY_TRACKING | IOMMU_HWPT_TRUSTED)) ||
+ parent || user_data)
return ERR_PTR(-EOPNOTSUPP);
return do_iommu_domain_alloc(type, dev, flags);
diff --git a/drivers/iommu/iommufd/hw_pagetable.c b/drivers/iommu/iommufd/hw_pagetable.c
index aefde4443671..23ae95fc95ee 100644
--- a/drivers/iommu/iommufd/hw_pagetable.c
+++ b/drivers/iommu/iommufd/hw_pagetable.c
@@ -136,6 +136,10 @@ iommufd_hwpt_paging_alloc(struct iommufd_ctx *ictx, struct iommufd_ioas *ioas,
hwpt_paging->nest_parent = flags & IOMMU_HWPT_ALLOC_NEST_PARENT;
if (ops->domain_alloc_user) {
+ if (ictx->kvm) {
+ pr_info("Trusted domain");
+ flags |= IOMMU_HWPT_TRUSTED;
+ }
hwpt->domain = ops->domain_alloc_user(idev->dev, flags, NULL,
user_data);
if (IS_ERR(hwpt->domain)) {
--
2.45.2
next prev parent reply other threads:[~2024-08-23 13:31 UTC|newest]
Thread overview: 128+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-08-23 13:21 [RFC PATCH 00/21] Secure VFIO, TDISP, SEV TIO Alexey Kardashevskiy
2024-08-23 13:21 ` [RFC PATCH 01/21] tsm-report: Rename module to reflect what it does Alexey Kardashevskiy
2024-08-23 22:17 ` Bjorn Helgaas
2024-08-28 13:49 ` Jonathan Cameron
2024-08-30 0:13 ` Dan Williams
2024-09-02 1:29 ` Alexey Kardashevskiy
2024-08-23 13:21 ` [RFC PATCH 02/21] pci/doe: Define protocol types and make those public Alexey Kardashevskiy
2024-08-23 22:18 ` Bjorn Helgaas
2024-08-30 2:15 ` Dan Williams
2024-08-23 13:21 ` [RFC PATCH 03/21] pci: Define TEE-IO bit in PCIe device capabilities Alexey Kardashevskiy
2024-08-23 22:19 ` Bjorn Helgaas
2024-08-28 13:54 ` Jonathan Cameron
2024-08-30 2:21 ` Dan Williams
2024-08-30 4:04 ` Alexey Kardashevskiy
2024-08-30 21:37 ` Dan Williams
2024-08-23 13:21 ` [RFC PATCH 04/21] PCI/IDE: Define Integrity and Data Encryption (IDE) extended capability Alexey Kardashevskiy
2024-08-23 22:28 ` Bjorn Helgaas
2024-08-28 14:24 ` Jonathan Cameron
2024-08-30 2:41 ` Dan Williams
2024-08-23 13:21 ` [RFC PATCH 05/21] crypto/ccp: Make some SEV helpers public Alexey Kardashevskiy
2024-08-30 2:45 ` Dan Williams
2024-08-23 13:21 ` [RFC PATCH 06/21] crypto: ccp: Enable SEV-TIO feature in the PSP when supported Alexey Kardashevskiy
2024-08-28 14:32 ` Jonathan Cameron
2024-09-03 21:27 ` Dan Williams
2024-09-05 2:29 ` Alexey Kardashevskiy
2024-09-05 17:40 ` Dan Williams
2024-08-23 13:21 ` [RFC PATCH 07/21] pci/tdisp: Introduce tsm module Alexey Kardashevskiy
2024-08-27 12:32 ` Jason Gunthorpe
2024-08-28 3:00 ` Alexey Kardashevskiy
2024-08-28 23:42 ` Jason Gunthorpe
2024-08-29 0:00 ` Dan Williams
2024-08-29 0:09 ` Jason Gunthorpe
2024-08-29 0:20 ` Dan Williams
2024-08-29 12:03 ` Jason Gunthorpe
2024-08-29 4:57 ` Alexey Kardashevskiy
2024-08-29 12:07 ` Jason Gunthorpe
2024-09-02 0:52 ` Alexey Kardashevskiy
2024-08-28 15:04 ` Jonathan Cameron
2024-09-02 6:50 ` Aneesh Kumar K.V
2024-09-02 7:26 ` Alexey Kardashevskiy
2024-09-03 23:51 ` Dan Williams
2024-09-04 11:13 ` Alexey Kardashevskiy
2024-09-04 23:28 ` Dan Williams
2024-08-23 13:21 ` [RFC PATCH 08/21] crypto/ccp: Implement SEV TIO firmware interface Alexey Kardashevskiy
2024-08-28 15:39 ` Jonathan Cameron
2024-08-23 13:21 ` [RFC PATCH 09/21] kvm: Export kvm_vm_set_mem_attributes Alexey Kardashevskiy
2024-08-23 13:21 ` [RFC PATCH 10/21] vfio: Export helper to get vfio_device from fd Alexey Kardashevskiy
2024-08-23 13:21 ` [RFC PATCH 11/21] KVM: SEV: Add TIO VMGEXIT and bind TDI Alexey Kardashevskiy
2024-08-29 10:08 ` Xu Yilun
2024-08-30 4:00 ` Alexey Kardashevskiy
2024-08-30 7:02 ` Xu Yilun
2024-09-02 1:24 ` Alexey Kardashevskiy
2024-09-13 13:50 ` Zhi Wang
2024-09-13 22:08 ` Dan Williams
2024-09-14 2:47 ` Tian, Kevin
2024-09-14 5:19 ` Zhi Wang
2024-09-18 10:45 ` Xu Yilun
2024-09-20 3:41 ` Tian, Kevin
2024-08-23 13:21 ` [RFC PATCH 12/21] KVM: IOMMUFD: MEMFD: Map private pages Alexey Kardashevskiy
2024-08-26 8:39 ` Tian, Kevin
2024-08-26 12:30 ` Jason Gunthorpe
2024-08-29 9:34 ` Xu Yilun
2024-08-29 12:15 ` Jason Gunthorpe
2024-08-30 3:47 ` Alexey Kardashevskiy
2024-08-30 12:35 ` Jason Gunthorpe
2024-09-02 1:09 ` Alexey Kardashevskiy
2024-09-02 23:52 ` Jason Gunthorpe
2024-09-03 0:03 ` Alexey Kardashevskiy
2024-09-03 0:37 ` Jason Gunthorpe
2024-08-30 5:20 ` Xu Yilun
2024-08-30 12:36 ` Jason Gunthorpe
2024-09-03 20:34 ` Dan Williams
2024-09-04 0:02 ` Jason Gunthorpe
2024-09-04 0:59 ` Dan Williams
2024-09-05 8:29 ` Tian, Kevin
2024-09-05 12:02 ` Jason Gunthorpe
2024-09-05 12:07 ` Tian, Kevin
2024-09-05 12:00 ` Jason Gunthorpe
2024-09-05 12:17 ` Tian, Kevin
2024-09-05 12:23 ` Jason Gunthorpe
2024-09-05 20:53 ` Dan Williams
2024-09-05 23:06 ` Jason Gunthorpe
2024-09-06 2:46 ` Tian, Kevin
2024-09-06 13:54 ` Jason Gunthorpe
2024-09-06 2:41 ` Tian, Kevin
2024-08-27 2:27 ` Alexey Kardashevskiy
2024-08-27 2:31 ` Tian, Kevin
2024-09-15 21:07 ` Jason Gunthorpe
2024-09-20 21:10 ` Vishal Annapurve
2024-09-23 5:35 ` Tian, Kevin
2024-09-23 6:34 ` Vishal Annapurve
2024-09-23 8:24 ` Tian, Kevin
2024-09-23 16:02 ` Jason Gunthorpe
2024-09-23 23:52 ` Tian, Kevin
2024-09-24 12:07 ` Jason Gunthorpe
2024-09-25 8:44 ` Vishal Annapurve
2024-09-25 15:41 ` Jason Gunthorpe
2024-09-23 20:53 ` Vishal Annapurve
2024-09-23 23:55 ` Tian, Kevin
2024-08-23 13:21 ` [RFC PATCH 13/21] KVM: X86: Handle private MMIO as shared Alexey Kardashevskiy
2024-08-30 16:57 ` Xu Yilun
2024-09-02 2:22 ` Alexey Kardashevskiy
2024-09-03 5:13 ` Xu Yilun
2024-09-06 3:31 ` Alexey Kardashevskiy
2024-09-09 10:07 ` Xu Yilun
2024-09-10 1:28 ` Alexey Kardashevskiy
2024-08-23 13:21 ` Alexey Kardashevskiy [this message]
2024-08-27 12:17 ` [RFC PATCH 14/21] RFC: iommu/iommufd/amd: Add IOMMU_HWPT_TRUSTED flag, tweak DTE's DomainID, IOTLB Jason Gunthorpe
2024-08-23 13:21 ` [RFC PATCH 15/21] coco/sev-guest: Allow multiple source files in the driver Alexey Kardashevskiy
2024-08-23 13:21 ` [RFC PATCH 16/21] coco/sev-guest: Make SEV-to-PSP request helpers public Alexey Kardashevskiy
2024-08-23 13:21 ` [RFC PATCH 17/21] coco/sev-guest: Implement the guest side of things Alexey Kardashevskiy
2024-08-28 15:54 ` Jonathan Cameron
2024-09-14 7:19 ` Zhi Wang
2024-09-16 1:18 ` Alexey Kardashevskiy
2024-08-23 13:21 ` [RFC PATCH 18/21] RFC: pci: Add BUS_NOTIFY_PCI_BUS_MASTER event Alexey Kardashevskiy
2024-08-23 13:21 ` [RFC PATCH 19/21] sev-guest: Stop changing encrypted page state for TDISP devices Alexey Kardashevskiy
2024-08-23 13:21 ` [RFC PATCH 20/21] pci: Allow encrypted MMIO mapping via sysfs Alexey Kardashevskiy
2024-08-23 22:37 ` Bjorn Helgaas
2024-09-02 8:22 ` Alexey Kardashevskiy
2024-09-03 21:46 ` Bjorn Helgaas
2024-08-23 13:21 ` [RFC PATCH 21/21] pci: Define pci_iomap_range_encrypted Alexey Kardashevskiy
2024-08-28 20:43 ` [RFC PATCH 00/21] Secure VFIO, TDISP, SEV TIO Dan Williams
2024-08-29 14:13 ` Alexey Kardashevskiy
2024-08-29 23:41 ` Dan Williams
2024-08-30 4:38 ` Alexey Kardashevskiy
2024-08-30 21:57 ` Dan Williams
2024-09-05 8:21 ` Tian, Kevin
2024-09-03 15:56 ` Sean Christopherson
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=20240823132137.336874-15-aik@amd.com \
--to=aik@amd.com \
--cc=agraf@suse.de \
--cc=alex.williamson@redhat.com \
--cc=dan.j.williams@intel.com \
--cc=david.kaplan@amd.com \
--cc=dhaval.giani@amd.com \
--cc=iommu@lists.linux.dev \
--cc=kvm@vger.kernel.org \
--cc=linux-coco@lists.linux.dev \
--cc=linux-pci@vger.kernel.org \
--cc=lukas@wunner.de \
--cc=michael.day@amd.com \
--cc=michael.roth@amd.com \
--cc=nikunj@amd.com \
--cc=pratikrajesh.sampat@amd.com \
--cc=santosh.shukla@amd.com \
--cc=suravee.suthikulpanit@amd.com \
--cc=thomas.lendacky@amd.com \
--cc=vasant.hegde@amd.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).