From: Jason Gunthorpe <jgg@ziepe.ca>
To: Vasant Hegde <vasant.hegde@amd.com>
Cc: Baolu Lu <baolu.lu@linux.intel.com>,
iommu@lists.linux.dev, joro@8bytes.org, will@kernel.org,
robin.murphy@arm.com, suravee.suthikulpanit@amd.com,
yi.l.liu@intel.com, kevin.tian@intel.com,
jacob.pan@linux.microsoft.com
Subject: Re: [PATCH v2 0/8] iommu: Domain allocation enhancements
Date: Wed, 9 Oct 2024 09:15:48 -0300 [thread overview]
Message-ID: <20241009121548.GE762027@ziepe.ca> (raw)
In-Reply-To: <9d0101b4-af13-4bf6-94a5-43a79f4d9989@amd.com>
On Wed, Oct 09, 2024 at 03:23:16PM +0530, Vasant Hegde wrote:
> > This change might cause a functional regression when it comes to nested
> > translation. In nested translation mode, the user page table (e.g.,
> > created and managed by a guest VM for guest kernel DMA) must be in the
> > first-stage page table format. Then, it can be nested on a second-stage
> > page table managed by the host kernel.
> >
> > Currently, the kernel automatically selects the page table formats. For
> > example, the Intel IOMMU driver always uses the first-stage page table
> > for guest kernel DMA. After this change, this assumption no longer holds
> > true. This means the kernel might use a second-stage page table for
> > guest kernel DMA, breaking nested translation.
>
> Hmmm. I assumed after discussion in v1 series you are fine. Looks like I misread it?
It is a bug in the implementation in the intel driver.
intel_iommu_domain_alloc_user() should always allocate a page table
that works, if you are in a guest context it must allocate a first
level/guest page table otherwise nested VFIO will be broken.
For this reason Intel driver should always allocate the guest
compatible page table unless IOMMU_HWPT_ALLOC_NEST_PARENT is
specified.
Something like this:
diff --git a/drivers/iommu/intel/iommu.c b/drivers/iommu/intel/iommu.c
index 9f6b0780f2ef5e..cfae6c2973e0ee 100644
--- a/drivers/iommu/intel/iommu.c
+++ b/drivers/iommu/intel/iommu.c
@@ -1440,7 +1440,7 @@ static void free_dmar_iommu(struct intel_iommu *iommu)
* Check and return whether first level is used by default for
* DMA translation.
*/
-static bool first_level_by_default(unsigned int type)
+static bool first_level_by_default(void)
{
/* Only SL is available in legacy mode */
if (!scalable_mode_support())
@@ -1450,8 +1450,8 @@ static bool first_level_by_default(unsigned int type)
if (intel_cap_flts_sanity() ^ intel_cap_slts_sanity())
return intel_cap_flts_sanity();
- /* Both levels are available, decide it based on domain type */
- return type != IOMMU_DOMAIN_UNMANAGED;
+ /* Both levels are available, use FL */
+ return true;
}
static struct dmar_domain *alloc_domain(unsigned int type)
@@ -1463,7 +1463,7 @@ static struct dmar_domain *alloc_domain(unsigned int type)
return NULL;
domain->nid = NUMA_NO_NODE;
- if (first_level_by_default(type))
+ if (first_level_by_default())
domain->use_first_level = true;
INIT_LIST_HEAD(&domain->devices);
INIT_LIST_HEAD(&domain->dev_pasids);
@@ -3287,8 +3287,7 @@ int __init intel_iommu_init(void)
* is likely to be much lower than the overhead of synchronizing
* the virtual and physical IOMMU page-tables.
*/
- if (cap_caching_mode(iommu->cap) &&
- !first_level_by_default(IOMMU_DOMAIN_DMA)) {
+ if (cap_caching_mode(iommu->cap) && !first_level_by_default()) {
pr_info_once("IOMMU batching disallowed due to virtualization\n");
iommu_set_dma_strict();
}
@@ -3530,6 +3529,7 @@ intel_iommu_domain_alloc_user(struct device *dev, u32 flags,
struct intel_iommu *iommu = info->iommu;
struct dmar_domain *dmar_domain;
struct iommu_domain *domain;
+ bool first_level;
/* Must be NESTING domain */
if (parent) {
@@ -3541,13 +3541,18 @@ intel_iommu_domain_alloc_user(struct device *dev, u32 flags,
if (flags &
(~(IOMMU_HWPT_ALLOC_NEST_PARENT | IOMMU_HWPT_ALLOC_DIRTY_TRACKING)))
return ERR_PTR(-EOPNOTSUPP);
+ if ((flags & IOMMU_HWPT_ALLOC_NEST_PARENT) && !intel_cap_slts_sanity())
+ return ERR_PTR(-EOPNOTSUPP);
if (nested_parent && !nested_supported(iommu))
return ERR_PTR(-EOPNOTSUPP);
if (user_data || (dirty_tracking && !ssads_supported(iommu)))
return ERR_PTR(-EOPNOTSUPP);
- /* Do not use first stage for user domain translation. */
- dmar_domain = paging_domain_alloc(dev, false);
+ if (flags & IOMMU_HWPT_ALLOC_NEST_PARENT)
+ first_level = false;
+ else
+ first_level = first_level_by_default();
+ dmar_domain = paging_domain_alloc(dev, first_level);
if (IS_ERR(dmar_domain))
return ERR_CAST(dmar_domain);
domain = &dmar_domain->domain;
next prev parent reply other threads:[~2024-10-09 12:15 UTC|newest]
Thread overview: 75+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-09-11 10:19 [PATCH v2 0/8] iommu: Domain allocation enhancements Vasant Hegde
2024-09-11 10:19 ` [PATCH v2 1/8] iommu: Refactor __iommu_domain_alloc() Vasant Hegde
2024-09-12 1:50 ` Baolu Lu
2024-09-13 4:02 ` Jacob Pan
2024-09-26 10:17 ` Vasant Hegde
2024-09-30 17:55 ` Jacob Pan
2024-10-01 4:31 ` Vasant Hegde
2024-10-02 5:11 ` Jacob Pan
[not found] ` <66fae60d.170a0220.280357.3d11SMTPIN_ADDED_BROKEN@mx.google.com>
2024-10-02 14:19 ` Jason Gunthorpe
2024-10-02 16:16 ` Jacob Pan
2024-10-02 19:09 ` Jason Gunthorpe
2024-10-15 8:12 ` Tian, Kevin
2024-09-11 10:19 ` [PATCH v2 2/8] iommu: Introduce iommu_paging_domain_alloc_flags() Vasant Hegde
2024-09-12 4:04 ` Baolu Lu
2024-09-26 10:43 ` Vasant Hegde
2024-10-15 8:24 ` Tian, Kevin
2024-10-15 12:31 ` Jason Gunthorpe
2024-10-16 2:44 ` Tian, Kevin
2024-10-02 19:12 ` Jason Gunthorpe
2024-10-09 21:14 ` Jacob Pan
2024-10-16 10:14 ` Vasant Hegde
2024-09-11 10:19 ` [PATCH v2 3/8] iommu: Add new flag to explictly request PASID capable domain Vasant Hegde
2024-09-12 4:14 ` Baolu Lu
2024-09-26 10:29 ` Vasant Hegde
2024-09-26 11:01 ` Vasant Hegde
2024-10-02 14:23 ` Jason Gunthorpe
2024-10-02 19:02 ` Jacob Pan
[not found] ` <66fd98e3.170a0220.23d7ae.c2a9SMTPIN_ADDED_BROKEN@mx.google.com>
2024-10-02 19:07 ` Jason Gunthorpe
2024-10-03 16:00 ` Jacob Pan
2024-10-02 19:23 ` Jason Gunthorpe
2024-10-04 8:12 ` Vasant Hegde
2024-10-04 12:46 ` Jason Gunthorpe
2024-10-15 8:31 ` Tian, Kevin
2024-09-11 10:19 ` [PATCH v2 4/8] iommu/amd: Separate page table setup from domain allocation Vasant Hegde
2024-09-13 17:08 ` Jacob Pan
2024-10-02 19:24 ` Jason Gunthorpe
2024-09-11 10:19 ` [PATCH v2 5/8] iommu/amd: Pass page table type as param to pdom_setup_pgtable() Vasant Hegde
2024-09-13 21:39 ` Jacob Pan
2024-09-26 10:25 ` Vasant Hegde
2024-09-30 17:57 ` Jacob Pan
[not found] ` <66e4b125.170a0220.2fa213.1e2cSMTPIN_ADDED_BROKEN@mx.google.com>
2024-09-20 13:02 ` Jason Gunthorpe
2024-09-11 10:19 ` [PATCH v2 6/8] iommu/amd: Enhance domain_alloc_user() to allocate PASID capable domain Vasant Hegde
2024-10-02 19:31 ` Jason Gunthorpe
2024-10-04 8:18 ` Vasant Hegde
2024-10-04 12:48 ` Jason Gunthorpe
2024-10-04 14:32 ` Vasant Hegde
2024-10-15 8:41 ` Tian, Kevin
2024-10-15 12:40 ` Jason Gunthorpe
2024-10-16 2:48 ` Tian, Kevin
2024-10-16 15:28 ` Jason Gunthorpe
2024-10-17 6:11 ` Tian, Kevin
2024-10-17 11:03 ` Vasant Hegde
2024-09-11 10:19 ` [PATCH v2 7/8] iommu/amd: Add iommu_ops->domain_alloc_paging support Vasant Hegde
2024-10-02 19:33 ` Jason Gunthorpe
2024-10-04 11:55 ` Vasant Hegde
2024-10-04 12:56 ` Jason Gunthorpe
2024-10-04 14:30 ` Vasant Hegde
2024-10-04 15:31 ` Jason Gunthorpe
2024-10-08 10:08 ` Vasant Hegde
2024-09-11 10:19 ` [PATCH v2 8/8] iommu/amd: Implement global identity domain Vasant Hegde
2024-10-02 19:36 ` Jason Gunthorpe
2024-10-04 11:42 ` Vasant Hegde
2024-10-02 5:30 ` [PATCH v2 0/8] iommu: Domain allocation enhancements Vasant Hegde
2024-10-02 14:24 ` Jason Gunthorpe
2024-10-04 6:11 ` Vasant Hegde
2024-10-09 2:47 ` Baolu Lu
2024-10-09 9:53 ` Vasant Hegde
2024-10-09 12:15 ` Jason Gunthorpe [this message]
2024-10-10 6:40 ` Baolu Lu
2024-10-10 6:48 ` Baolu Lu
2024-10-10 11:38 ` Jason Gunthorpe
2024-10-10 14:06 ` Baolu Lu
2024-10-11 5:06 ` Tian, Kevin
2024-10-11 11:39 ` Jason Gunthorpe
2024-10-15 8:10 ` Tian, Kevin
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=20241009121548.GE762027@ziepe.ca \
--to=jgg@ziepe.ca \
--cc=baolu.lu@linux.intel.com \
--cc=iommu@lists.linux.dev \
--cc=jacob.pan@linux.microsoft.com \
--cc=joro@8bytes.org \
--cc=kevin.tian@intel.com \
--cc=robin.murphy@arm.com \
--cc=suravee.suthikulpanit@amd.com \
--cc=vasant.hegde@amd.com \
--cc=will@kernel.org \
--cc=yi.l.liu@intel.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