Linux MM tree latest commits
 help / color / mirror / Atom feed
* [PATCH] mm/hmm: account for subpage offsets in device-private PMDs
@ 2026-10-31  4:07 Aveline Noir
  2026-11-22  0:00 ` Andrew Morton
  2026-11-22  0:01 ` + mm-hmm-account-for-subpage-offsets-in-device-private-pmds.patch added to mm-hotfixes-unstable branch Andrew Morton
  0 siblings, 2 replies; 10+ messages in thread
From: Aveline Noir @ 2026-10-31  4:07 UTC (permalink / raw)
  To: jgg, leonro, akpm
  Cc: airlied, apopple, balbirs, christian.koenig, dri-devel,
	Francois Dugast, intel-xe, jgg, leon, linux-kernel, linux-mm,
	Lorenzo Stoakes, Matthew Brost, mm-commits, mpenttil, Patchwork,
	SeongJae Park, thomas.hellstrom, Vlastimil Babka, ziy

When a query starts inside a device-private PMD owned by
range->dev_private_owner, hmm_vma_handle_absent_pmd() still fills the PFN
array from the folio's first page. The returned PFNs therefore refer to
earlier pages rather than the requested subrange.

Add the query's PMD-relative page offset to the initial PFN, as the
present PMD path already does.

Fixes: 10b9feee2d0d ("mm/hmm: populate PFNs from PMD swap entry")
Signed-off-by: Aveline Noir <jm5905938@gmail.com>

---
Found during a synthetic test_hmm check of device-private THP handling.
Tested PMD-aligned and non-aligned subranges, PMD boundaries, split
PTEs and fake-device data accesses with test_hmm in QEMU TCG; all
executed tests passed with the fix. No real GPU testing was performed.
---
 mm/hmm.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/mm/hmm.c b/mm/hmm.c
index 2f1e98c6b644..e89f140d3e5c 100644
--- a/mm/hmm.c
+++ b/mm/hmm.c
@@ -355,7 +355,8 @@ static int hmm_vma_handle_absent_pmd(struct mm_walk *walk, unsigned long start,
 	    range->dev_private_owner) {
 		unsigned long cpu_flags = HMM_PFN_VALID |
 			hmm_pfn_flags_order(PMD_SHIFT - PAGE_SHIFT);
-		unsigned long pfn = softleaf_to_pfn(entry);
+		unsigned long pfn = softleaf_to_pfn(entry) +
+			((start & ~PMD_MASK) >> PAGE_SHIFT);
 		unsigned long i;
 
 		if (softleaf_is_device_private_write(entry))

base-commit: d89e1d186ec9271bf78e62771779e9fc03bf0fc3
-- 
2.56.0


^ permalink raw reply related	[flat|nested] 10+ messages in thread

* Re: [PATCH] mm/hmm: account for subpage offsets in device-private PMDs
  2026-10-31  4:07 [PATCH] mm/hmm: account for subpage offsets in device-private PMDs Aveline Noir
@ 2026-11-22  0:00 ` Andrew Morton
  2026-11-22  5:55   ` Mika Penttilä
  2026-11-22  0:01 ` + mm-hmm-account-for-subpage-offsets-in-device-private-pmds.patch added to mm-hotfixes-unstable branch Andrew Morton
  1 sibling, 1 reply; 10+ messages in thread
From: Andrew Morton @ 2026-11-22  0:00 UTC (permalink / raw)
  To: Aveline Noir
  Cc: jgg, leonro, airlied, apopple, balbirs, christian.koenig,
	dri-devel, Francois Dugast, intel-xe, jgg, leon, linux-kernel,
	linux-mm, Lorenzo Stoakes, Matthew Brost, mm-commits, mpenttil,
	Patchwork, SeongJae Park, thomas.hellstrom, Vlastimil Babka, ziy

On Sat, 31 Oct 2026 15:07:58 +1100 Aveline Noir <jm5905938@gmail.com> wrote:

> When a query starts inside a device-private PMD owned by
> range->dev_private_owner, hmm_vma_handle_absent_pmd() still fills the PFN
> array from the folio's first page. The returned PFNs therefore refer to
> earlier pages rather than the requested subrange.
> 
> Add the query's PMD-relative page offset to the initial PFN, as the
> present PMD path already does.
> 
> Fixes: 10b9feee2d0d ("mm/hmm: populate PFNs from PMD swap entry")
> Signed-off-by: Aveline Noir <jm5905938@gmail.com>
> 
> ---
> Found during a synthetic test_hmm check of device-private THP handling.
> Tested PMD-aligned and non-aligned subranges, PMD boundaries, split
> PTEs and fake-device data accesses with test_hmm in QEMU TCG; all
> executed tests passed with the fix. No real GPU testing was performed.

Thanks.  It appears that the changelog somewhat understates the
seriousness of this so I came up with a more assertive one, below. 
Please check that.

I fed this through LLM-driven QEMU-based runtime testing and it looks
like it pretty much replicated your testing:

: I tested "mm/hmm: account for subpage offsets in device-private
: PMDs" on matched x86-64 guest kernels. An unprivileged process
: (UID 65534) migrated an anonymous 2 MiB THP to the test_hmm
: device. A second open of the device read pages inside the PMD
: while the first remained open. A snapshot confirmed a
: device-private PMD, and temporary kernel log messages confirmed
: that the reads reached hmm_vma_handle_absent_pmd().
: 
: Without the patch, requests for pages 1, 7, 255 and 511 returned
: page 0. A request for pages 31-33 returned pages 0-2. With the
: patch, all seven pages returned the requested contents. Page 0
: read correctly on both kernels. Thus an unprivileged device user
: can receive data from the wrong pages.
:
: The fake device needed an identical temporary backing-page
: correction on both kernels to make this read path work. That
: change was removed after testing. This exercises the HMM path
: through test_hmm; I did not test GPU hardware.

So it accidentally found a bug in lib/test_hmm.c on the way.  "The QEMU
run exposed a real test_hmm bug: a second open of /dev/hmm_dmirror0 can
put a device-private page into its page table, then dmirror_do_read()
tries to copy from that non-CPU-accessible page and oopses".  I'll prep
a fix for that separately.



I'll queue the below in mm.git's mm-hotfixes queue and shall await
maintainer input (please).


From: Aveline Noir <jm5905938@gmail.com>
Subject: mm/hmm: account for subpage offsets in device-private PMDs
Date: Sat, 31 Oct 2026 15:07:58 +1100

When a query starts inside a device-private PMD owned by
range->dev_private_owner, hmm_vma_handle_absent_pmd() still fills the PFN
array from the folio's first page.  The returned PFNs therefore refer to
earlier pages rather than the requested subrange.

Nouveau's GPU fault path can query one page of a device-private PMD at a
time.  Xe's SVM fault, prefetch and userptr mapping paths can also query
subranges of their own device-private PMDs.  An unprivileged process with
access to the GPU render node can reach these paths on hardware supporting
device-private THP migration.  Xe can read the wrong application data or
silently overwrite earlier pages of the same process's device-private
folio.

Nouveau aligns a one-page GPU fault to the PMD head by subtracting the
fault offset from the returned page.  The faulty HMM PFN already refers to
the head, so this subtraction can select a page before the folio.  If the
preceding page belongs to another allocation, GPU access can expose or
overwrite unrelated userspace data; an invalid page can also fault the
kernel during its address conversion.  This worst case follows from the
source, but has not been demonstrated on a real GPU.

Add the query's PMD-relative page offset to the initial PFN, as the
present PMD path already does.

The device-private PMD migration machinery first shipped in 6.19. 
Backport this fix to the maintained 7.3.y stable series; the HMM path
exists in 6.18.y but cannot be reached through an in-tree PMD-producing
device-private migration path there.

Link: https://patch.msgid.link/179342234355.4.16705453209886955762@gmail.com
Fixes: 10b9feee2d0d ("mm/hmm: populate PFNs from PMD swap entry")
Signed-off-by: Aveline Noir <jm5905938@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Leon Romanovsky <leon@kernel.org>
Cc: Francois Dugast <francois.dugast@intel.com>
Cc: Lorenzo Stoakes <ljs@kernel.org>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Patchwork <patchwork@emeril.freedesktop.org>
Cc: SJ Park <sj@kernel.org>
Cc: Vlastimil Babka <vbabka@kernel.org>
Cc: David Airlie <airlied@gmail.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Balbir Singh <balbirs@nvidia.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Mika Penttilä <mpenttil@redhat.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: <stable@vger.kernel.org>

 mm/hmm.c |    3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

--- a/mm/hmm.c~mm-hmm-account-for-subpage-offsets-in-device-private-pmds
+++ a/mm/hmm.c
@@ -355,7 +355,8 @@ static int hmm_vma_handle_absent_pmd(str
 	    range->dev_private_owner) {
 		unsigned long cpu_flags = HMM_PFN_VALID |
 			hmm_pfn_flags_order(PMD_SHIFT - PAGE_SHIFT);
-		unsigned long pfn = softleaf_to_pfn(entry);
+		unsigned long pfn = softleaf_to_pfn(entry) +
+			((start & ~PMD_MASK) >> PAGE_SHIFT);
 		unsigned long i;
 
 		if (softleaf_is_device_private_write(entry))
_


^ permalink raw reply	[flat|nested] 10+ messages in thread

* + mm-hmm-account-for-subpage-offsets-in-device-private-pmds.patch added to mm-hotfixes-unstable branch
  2026-10-31  4:07 [PATCH] mm/hmm: account for subpage offsets in device-private PMDs Aveline Noir
  2026-11-22  0:00 ` Andrew Morton
@ 2026-11-22  0:01 ` Andrew Morton
  1 sibling, 0 replies; 10+ messages in thread
From: Andrew Morton @ 2026-11-22  0:01 UTC (permalink / raw)
  To: mm-commits, ziy, vbabka, thomas.hellstrom, stable, sj, patchwork,
	mpenttil, matthew.brost, ljs, leon, jgg, francois.dugast,
	christian.koenig, balbirs, apopple, akpm, airlied, jm5905938

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 5304 bytes --]


The patch titled
     Subject: mm/hmm: account for subpage offsets in device-private PMDs
has been added to the -mm mm-hotfixes-unstable branch.  Its filename is
     mm-hmm-account-for-subpage-offsets-in-device-private-pmds.patch

This patch will shortly appear at
     https://git.kernel.org/pub/scm/linux/kernel/git/akpm/25-new.git/tree/patches/mm-hmm-account-for-subpage-offsets-in-device-private-pmds.patch

This patch will later appear in the mm-hotfixes-unstable branch at
    git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm

Before you just go and hit "reply", please:
   a) Consider who else should be cc'ed
   b) Prefer to cc a suitable mailing list as well
   c) Ideally: find the original patch on the mailing list and do a
      reply-to-all to that, adding suitable additional cc's

*** Remember to use Documentation/process/submit-checklist.rst when testing your code ***

The -mm tree is included into linux-next via various
branches at git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
and is updated there most days

------------------------------------------------------
From: Aveline Noir <jm5905938@gmail.com>
Subject: mm/hmm: account for subpage offsets in device-private PMDs
Date: Sat, 31 Oct 2026 15:07:58 +1100

When a query starts inside a device-private PMD owned by
range->dev_private_owner, hmm_vma_handle_absent_pmd() still fills the PFN
array from the folio's first page.  The returned PFNs therefore refer to
earlier pages rather than the requested subrange.

Nouveau's GPU fault path can query one page of a device-private PMD at a
time.  Xe's SVM fault, prefetch and userptr mapping paths can also query
subranges of their own device-private PMDs.  An unprivileged process with
access to the GPU render node can reach these paths on hardware supporting
device-private THP migration.  Xe can read the wrong application data or
silently overwrite earlier pages of the same process's device-private
folio.

Nouveau aligns a one-page GPU fault to the PMD head by subtracting the
fault offset from the returned page.  The faulty HMM PFN already refers to
the head, so this subtraction can select a page before the folio.  If the
preceding page belongs to another allocation, GPU access can expose or
overwrite unrelated userspace data; an invalid page can also fault the
kernel during its address conversion.  This worst case follows from the
source, but has not been demonstrated on a real GPU.

Add the query's PMD-relative page offset to the initial PFN, as the
present PMD path already does.

The device-private PMD migration machinery first shipped in 6.19. 
Backport this fix to the maintained 7.3.y stable series; the HMM path
exists in 6.18.y but cannot be reached through an in-tree PMD-producing
device-private migration path there.

Link: https://patch.msgid.link/179342234355.4.16705453209886955762@gmail.com
Fixes: 10b9feee2d0d ("mm/hmm: populate PFNs from PMD swap entry")
Signed-off-by: Aveline Noir <jm5905938@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Tested-by: Andrew Morton <akpm@linux-foundation.org>
  Link: https://lore.kernel.org/20261121160037.1a479f3c8b4e860bc23c106d@linux-foundation.org
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Leon Romanovsky <leon@kernel.org>
Cc: Francois Dugast <francois.dugast@intel.com>
Cc: Lorenzo Stoakes <ljs@kernel.org>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Patchwork <patchwork@emeril.freedesktop.org>
Cc: SJ Park <sj@kernel.org>
Cc: Vlastimil Babka <vbabka@kernel.org>
Cc: David Airlie <airlied@gmail.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Balbir Singh <balbirs@nvidia.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Mika Penttilä <mpenttil@redhat.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: <stable@vger.kernel.org>

 mm/hmm.c |    3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

--- a/mm/hmm.c~mm-hmm-account-for-subpage-offsets-in-device-private-pmds
+++ a/mm/hmm.c
@@ -355,7 +355,8 @@ static int hmm_vma_handle_absent_pmd(str
 	    range->dev_private_owner) {
 		unsigned long cpu_flags = HMM_PFN_VALID |
 			hmm_pfn_flags_order(PMD_SHIFT - PAGE_SHIFT);
-		unsigned long pfn = softleaf_to_pfn(entry);
+		unsigned long pfn = softleaf_to_pfn(entry) +
+			((start & ~PMD_MASK) >> PAGE_SHIFT);
 		unsigned long i;
 
 		if (softleaf_is_device_private_write(entry))
_

Patches currently in -mm which might be from jm5905938@gmail.com are

mm-hmm-account-for-subpage-offsets-in-device-private-pmds.patch
mm-damon-ops-common-align-mmu-notifier-range-when-clearing-pmd-young-bit.patch
mm-damon-stat-avoid-32-bit-overflow-in-idle-percentiles.patch
mm-damon-pvaddr-skip-folios-already-on-the-migration-target-node.patch
docs-admin-guide-mm-damon-start-fix-grammar.patch
docs-admin-guide-mm-damon-stat-fix-pronoun-agreement.patch
docs-admin-guide-mm-damon-usage-fix-grammar.patch
docs-admin-guide-mm-damon-lru_sort-fix-remaining-grammar.patch
docs-mm-damon-design-fix-grammar-and-typos.patch
docs-mm-damon-api-fix-verb-form.patch
docs-mm-damon-maintainer-profile-fix-subject-verb-agreement.patch
docs-mm-damon-monitoring_intervals_tuning_example-fix-grammar-and-typos.patch
mm-damon-paddr-add-kunit-tests-for-addr_unit-conversion.patch
mm-damon-paddr-add-kunit-tests-for-addr_unit-conversion-fix.patch


^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH] mm/hmm: account for subpage offsets in device-private PMDs
  2026-11-22  0:00 ` Andrew Morton
@ 2026-11-22  5:55   ` Mika Penttilä
  2026-11-22  9:52     ` Andrew Morton
  0 siblings, 1 reply; 10+ messages in thread
From: Mika Penttilä @ 2026-11-22  5:55 UTC (permalink / raw)
  To: Andrew Morton, Aveline Noir
  Cc: jgg, leonro, airlied, apopple, balbirs, christian.koenig,
	dri-devel, Francois Dugast, intel-xe, jgg, leon, linux-kernel,
	linux-mm, Lorenzo Stoakes, Matthew Brost, mm-commits, Patchwork,
	SeongJae Park, thomas.hellstrom, Vlastimil Babka, ziy


On 11/22/26 02:00, Andrew Morton wrote:
> On Sat, 31 Oct 2026 15:07:58 +1100 Aveline Noir <jm5905938@gmail.com> wrote:
>
>> When a query starts inside a device-private PMD owned by
>> range->dev_private_owner, hmm_vma_handle_absent_pmd() still fills the PFN
>> array from the folio's first page. The returned PFNs therefore refer to
>> earlier pages rather than the requested subrange.
>>
>> Add the query's PMD-relative page offset to the initial PFN, as the
>> present PMD path already does.
>>
>> Fixes: 10b9feee2d0d ("mm/hmm: populate PFNs from PMD swap entry")
>> Signed-off-by: Aveline Noir <jm5905938@gmail.com>
>>
>> ---
>> Found during a synthetic test_hmm check of device-private THP handling.
>> Tested PMD-aligned and non-aligned subranges, PMD boundaries, split
>> PTEs and fake-device data accesses with test_hmm in QEMU TCG; all
>> executed tests passed with the fix. No real GPU testing was performed.
> Thanks.  It appears that the changelog somewhat understates the
> seriousness of this so I came up with a more assertive one, below. 
> Please check that.
>
> I fed this through LLM-driven QEMU-based runtime testing and it looks
> like it pretty much replicated your testing:
>
> : I tested "mm/hmm: account for subpage offsets in device-private
> : PMDs" on matched x86-64 guest kernels. An unprivileged process
> : (UID 65534) migrated an anonymous 2 MiB THP to the test_hmm
> : device. A second open of the device read pages inside the PMD
> : while the first remained open. A snapshot confirmed a
> : device-private PMD, and temporary kernel log messages confirmed
> : that the reads reached hmm_vma_handle_absent_pmd().
> : 
> : Without the patch, requests for pages 1, 7, 255 and 511 returned
> : page 0. A request for pages 31-33 returned pages 0-2. With the
> : patch, all seven pages returned the requested contents. Page 0
> : read correctly on both kernels. Thus an unprivileged device user
> : can receive data from the wrong pages.
> :
> : The fake device needed an identical temporary backing-page
> : correction on both kernels to make this read path work. That
> : change was removed after testing. This exercises the HMM path
> : through test_hmm; I did not test GPU hardware.
>
> So it accidentally found a bug in lib/test_hmm.c on the way.  "The QEMU
> run exposed a real test_hmm bug: a second open of /dev/hmm_dmirror0 can
> put a device-private page into its page table, then dmirror_do_read()
> tries to copy from that non-CPU-accessible page and oopses".  I'll prep
> a fix for that separately.

I noticed that also and will make changes around it in my series.

>
>
>
> I'll queue the below in mm.git's mm-hotfixes queue and shall await
> maintainer input (please).
>
>
> From: Aveline Noir <jm5905938@gmail.com>
> Subject: mm/hmm: account for subpage offsets in device-private PMDs
> Date: Sat, 31 Oct 2026 15:07:58 +1100
>
> When a query starts inside a device-private PMD owned by
> range->dev_private_owner, hmm_vma_handle_absent_pmd() still fills the PFN
> array from the folio's first page.  The returned PFNs therefore refer to
> earlier pages rather than the requested subrange.
>
> Nouveau's GPU fault path can query one page of a device-private PMD at a
> time.  Xe's SVM fault, prefetch and userptr mapping paths can also query
> subranges of their own device-private PMDs.  An unprivileged process with
> access to the GPU render node can reach these paths on hardware supporting
> device-private THP migration.  Xe can read the wrong application data or
> silently overwrite earlier pages of the same process's device-private
> folio.
>
> Nouveau aligns a one-page GPU fault to the PMD head by subtracting the
> fault offset from the returned page.  The faulty HMM PFN already refers to
> the head, so this subtraction can select a page before the folio.  If the
> preceding page belongs to another allocation, GPU access can expose or
> overwrite unrelated userspace data; an invalid page can also fault the
> kernel during its address conversion.  This worst case follows from the
> source, but has not been demonstrated on a real GPU.
>
> Add the query's PMD-relative page offset to the initial PFN, as the
> present PMD path already does.
>
> The device-private PMD migration machinery first shipped in 6.19. 
> Backport this fix to the maintained 7.3.y stable series; the HMM path
> exists in 6.18.y but cannot be reached through an in-tree PMD-producing
> device-private migration path there.
>
> Link: https://patch.msgid.link/179342234355.4.16705453209886955762@gmail.com
> Fixes: 10b9feee2d0d ("mm/hmm: populate PFNs from PMD swap entry")
> Signed-off-by: Aveline Noir <jm5905938@gmail.com>
> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
> Cc: Jason Gunthorpe <jgg@ziepe.ca>
> Cc: Leon Romanovsky <leon@kernel.org>
> Cc: Francois Dugast <francois.dugast@intel.com>
> Cc: Lorenzo Stoakes <ljs@kernel.org>
> Cc: Matthew Brost <matthew.brost@intel.com>
> Cc: Patchwork <patchwork@emeril.freedesktop.org>
> Cc: SJ Park <sj@kernel.org>
> Cc: Vlastimil Babka <vbabka@kernel.org>
> Cc: David Airlie <airlied@gmail.com>
> Cc: Alistair Popple <apopple@nvidia.com>
> Cc: Balbir Singh <balbirs@nvidia.com>
> Cc: Christian König <christian.koenig@amd.com>
> Cc: Mika Penttilä <mpenttil@redhat.com>
> Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
> Cc: Zi Yan <ziy@nvidia.com>
> Cc: <stable@vger.kernel.org>
>
>  mm/hmm.c |    3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
> --- a/mm/hmm.c~mm-hmm-account-for-subpage-offsets-in-device-private-pmds
> +++ a/mm/hmm.c
> @@ -355,7 +355,8 @@ static int hmm_vma_handle_absent_pmd(str
>  	    range->dev_private_owner) {
>  		unsigned long cpu_flags = HMM_PFN_VALID |
>  			hmm_pfn_flags_order(PMD_SHIFT - PAGE_SHIFT);
> -		unsigned long pfn = softleaf_to_pfn(entry);
> +		unsigned long pfn = softleaf_to_pfn(entry) +
> +			((start & ~PMD_MASK) >> PAGE_SHIFT);
>  		unsigned long i;
>  
>  		if (softleaf_is_device_private_write(entry))
> _

For this change:

Reviewed-by: Mika Penttilä <mpenttil@redhat.com>



^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH] mm/hmm: account for subpage offsets in device-private PMDs
  2026-11-22  5:55   ` Mika Penttilä
@ 2026-11-22  9:52     ` Andrew Morton
  2026-11-22 12:04       ` Mika Penttilä
  2026-11-22 14:45       ` Aveline Noir
  0 siblings, 2 replies; 10+ messages in thread
From: Andrew Morton @ 2026-11-22  9:52 UTC (permalink / raw)
  To: Mika Penttilä
  Cc: Aveline Noir, jgg, leonro, airlied, apopple, balbirs,
	christian.koenig, dri-devel, Francois Dugast, intel-xe, jgg, leon,
	linux-kernel, linux-mm, Lorenzo Stoakes, Matthew Brost,
	mm-commits, Patchwork, SeongJae Park, thomas.hellstrom,
	Vlastimil Babka, ziy

On Sun, 22 Nov 2026 07:55:37 +0200 Mika Penttilä <mpenttil@redhat.com> wrote:

> > So it accidentally found a bug in lib/test_hmm.c on the way.  "The QEMU
> > run exposed a real test_hmm bug: a second open of /dev/hmm_dmirror0 can
> > put a device-private page into its page table, then dmirror_do_read()
> > tries to copy from that non-CPU-accessible page and oopses".  I'll prep
> > a fix for that separately.
> 
> I noticed that also and will make changes around it in my series.

I came up with the below.  It all tested out OK.  wdyt?  I guess it
should have cc:stable>



From: Andrew Morton <akpm@linux-foundation.org>
Subject: lib/test_hmm: use backing pages for device-private mirror faults
Date: Sat, 21 Nov 2026 17:06:45 -0800

When one hmm_dmirror descriptor migrates memory to device-private and a
second descriptor reads it, dmirror_do_fault() stores the device page in
the second descriptor's page table.  dmirror_do_read() then passes that
page to memcpy_from_page(), which can oops the kernel.  This was
reproduced with an unprivileged process using the fake device in a QEMU
guest.

Store the corresponding RAM backing page instead.  For a large device
folio, retain the subpage offset when selecting the backing page; using
the backing folio's head for every entry would return the wrong data. 
Extend the existing PMD migration selftest to read the last subpage
through a second open before faulting the memory back to RAM.

Fixes: b2ef9f5a5cb3 ("mm/hmm/test: add selftest driver for HMM")
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Leon Romanovsky <leon@kernel.org>
Cc: Shuah Khan <shuah@kernel.org>
---
tl;dr: The unpatched kernel oopsed when an unprivileged process opened the
fake HMM device twice, migrated an anonymous 2 MB PMD to device-private
memory through one descriptor, then accessed its last page through the
other.  The patched kernel returned the expected bytes on all three ioctl
paths.

A was reproduced and saved before B was built.  As uid 65534, a static
userspace program verified that all 512 pages were device-private PMD
entries.  A READ of the last page through the second descriptor produced a
KASAN-enabled kernel Oops in dmirror_do_read().  Independent A boots also
oopsed on WRITE and READ_UNLOCKED at the same last-page access.

With only this patch added, READ and READ_UNLOCKED returned the expected
0x0a byte, and WRITE returned the expected 0xa5 byte after migration back
to RAM.  The modified in-tree hmm_device_private.migrate_anon_huge_fault
selftest also passed (1/1).

This reproduces a userspace-triggered kernel Oops in the fake HMM device
used by test_hmm; it does not establish reachability through physical GPU
hardware.  QEMU guests used matching kernel configurations with
CONFIG_TEST_HMM=y and KASAN.  The guest device permissions were set for
uid 65534.

 lib/test_hmm.c                         |    7 +++++++
 tools/testing/selftests/mm/hmm-tests.c |   14 ++++++++++++++
 2 files changed, 21 insertions(+)

--- a/lib/test_hmm.c~lib-test_hmm-use-backing-pages-for-device-private-mirror-faults
+++ a/lib/test_hmm.c
@@ -292,6 +292,13 @@ static int dmirror_do_fault(struct dmirr
 		page = hmm_pfn_to_page(*pfns);
 		WARN_ON(!page);
 
+		if (is_device_private_page(page)) {
+			struct folio *folio = page_folio(page);
+			struct page *backing = BACKING_PAGE(&folio->page);
+
+			page = folio_page(page_folio(backing),
+					  folio_page_idx(folio, page));
+		}
 		entry = page;
 		if (*pfns & HMM_PFN_WRITE)
 			entry = xa_tag_pointer(entry, DPT_XA_TAG_WRITE);
--- a/tools/testing/selftests/mm/hmm-tests.c~lib-test_hmm-use-backing-pages-for-device-private-mirror-faults
+++ a/tools/testing/selftests/mm/hmm-tests.c
@@ -2087,6 +2087,7 @@ TEST_F(hmm, migrate_anon_huge_fault)
 	void *old_ptr;
 	void *map;
 	int pagemap_fd;
+	int fd2 = -1;
 	int *ptr;
 	int ret;
 
@@ -2143,12 +2144,25 @@ TEST_F(hmm, migrate_anon_huge_fault)
 		}
 
 		close(pagemap_fd);
+
+		/* A second open must read the last device-private subpage. */
+		fd2 = hmm_open(variant->device_number);
+		ASSERT_GE(fd2, 0);
+		buffer->ptr = (char *)map + (npages - 1) * self->page_size;
+		ret = hmm_dmirror_cmd(fd2, HMM_DMIRROR_READ, buffer, 1);
+		buffer->ptr = map;
+		ASSERT_EQ(ret, 0);
+		ASSERT_EQ(buffer->cpages, 1);
+		ASSERT_EQ(*(int *)buffer->mirror,
+			  (npages - 1) * self->page_size / sizeof(int));
 	}
 
 	/* Fault pages back to system memory and check them. */
 	for (i = 0, ptr = buffer->ptr; i < size / sizeof(*ptr); ++i)
 		ASSERT_EQ(ptr[i], i);
 
+	if (fd2 >= 0)
+		ASSERT_EQ(close(fd2), 0);
 	buffer->ptr = old_ptr;
 	hmm_buffer_free(buffer);
 }
_


^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH] mm/hmm: account for subpage offsets in device-private PMDs
  2026-11-22  9:52     ` Andrew Morton
@ 2026-11-22 12:04       ` Mika Penttilä
  2026-11-22 14:45       ` Aveline Noir
  1 sibling, 0 replies; 10+ messages in thread
From: Mika Penttilä @ 2026-11-22 12:04 UTC (permalink / raw)
  To: Andrew Morton
  Cc: Aveline Noir, jgg, leonro, airlied, apopple, balbirs,
	christian.koenig, dri-devel, Francois Dugast, intel-xe, jgg, leon,
	linux-kernel, linux-mm, Lorenzo Stoakes, Matthew Brost,
	mm-commits, Patchwork, SeongJae Park, thomas.hellstrom,
	Vlastimil Babka, ziy

On 11/22/26 11:52, Andrew Morton wrote:

> On Sun, 22 Nov 2026 07:55:37 +0200 Mika Penttilä <mpenttil@redhat.com> wrote:
>
>>> So it accidentally found a bug in lib/test_hmm.c on the way.  "The QEMU
>>> run exposed a real test_hmm bug: a second open of /dev/hmm_dmirror0 can
>>> put a device-private page into its page table, then dmirror_do_read()
>>> tries to copy from that non-CPU-accessible page and oopses".  I'll prep
>>> a fix for that separately.
>> I noticed that also and will make changes around it in my series.
> I came up with the below.  It all tested out OK.  wdyt?  I guess it
> should have cc:stable>
>
>
>
> From: Andrew Morton <akpm@linux-foundation.org>
> Subject: lib/test_hmm: use backing pages for device-private mirror faults
> Date: Sat, 21 Nov 2026 17:06:45 -0800
>
> When one hmm_dmirror descriptor migrates memory to device-private and a
> second descriptor reads it, dmirror_do_fault() stores the device page in
> the second descriptor's page table.  dmirror_do_read() then passes that
> page to memcpy_from_page(), which can oops the kernel.  This was
> reproduced with an unprivileged process using the fake device in a QEMU
> guest.
>
> Store the corresponding RAM backing page instead.  For a large device
> folio, retain the subpage offset when selecting the backing page; using
> the backing folio's head for every entry would return the wrong data. 
> Extend the existing PMD migration selftest to read the last subpage
> through a second open before faulting the memory back to RAM.
>
> Fixes: b2ef9f5a5cb3 ("mm/hmm/test: add selftest driver for HMM")
> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
> Cc: Jason Gunthorpe <jgg@ziepe.ca>
> Cc: Leon Romanovsky <leon@kernel.org>
> Cc: Shuah Khan <shuah@kernel.org>
> ---
> tl;dr: The unpatched kernel oopsed when an unprivileged process opened the
> fake HMM device twice, migrated an anonymous 2 MB PMD to device-private
> memory through one descriptor, then accessed its last page through the
> other.  The patched kernel returned the expected bytes on all three ioctl
> paths.
>
> A was reproduced and saved before B was built.  As uid 65534, a static
> userspace program verified that all 512 pages were device-private PMD
> entries.  A READ of the last page through the second descriptor produced a
> KASAN-enabled kernel Oops in dmirror_do_read().  Independent A boots also
> oopsed on WRITE and READ_UNLOCKED at the same last-page access.
>
> With only this patch added, READ and READ_UNLOCKED returned the expected
> 0x0a byte, and WRITE returned the expected 0xa5 byte after migration back
> to RAM.  The modified in-tree hmm_device_private.migrate_anon_huge_fault
> selftest also passed (1/1).
>
> This reproduces a userspace-triggered kernel Oops in the fake HMM device
> used by test_hmm; it does not establish reachability through physical GPU
> hardware.  QEMU guests used matching kernel configurations with
> CONFIG_TEST_HMM=y and KASAN.  The guest device permissions were set for
> uid 65534.
>
>  lib/test_hmm.c                         |    7 +++++++
>  tools/testing/selftests/mm/hmm-tests.c |   14 ++++++++++++++
>  2 files changed, 21 insertions(+)
>
> --- a/lib/test_hmm.c~lib-test_hmm-use-backing-pages-for-device-private-mirror-faults
> +++ a/lib/test_hmm.c
> @@ -292,6 +292,13 @@ static int dmirror_do_fault(struct dmirr
>  		page = hmm_pfn_to_page(*pfns);
>  		WARN_ON(!page);
>  
> +		if (is_device_private_page(page)) {
> +			struct folio *folio = page_folio(page);
> +			struct page *backing = BACKING_PAGE(&folio->page);
> +
> +			page = folio_page(page_folio(backing),
> +					  folio_page_idx(folio, page));
> +		}
>  		entry = page;
>  		if (*pfns & HMM_PFN_WRITE)

I have this same chunk. I think it's needed, all dmirror readers expect to
copy_to_user() from pt.


>  			entry = xa_tag_pointer(entry, DPT_XA_TAG_WRITE);
> --- a/tools/testing/selftests/mm/hmm-tests.c~lib-test_hmm-use-backing-pages-for-device-private-mirror-faults
> +++ a/tools/testing/selftests/mm/hmm-tests.c
> @@ -2087,6 +2087,7 @@ TEST_F(hmm, migrate_anon_huge_fault)
>  	void *old_ptr;
>  	void *map;
>  	int pagemap_fd;
> +	int fd2 = -1;
>  	int *ptr;
>  	int ret;
>  
> @@ -2143,12 +2144,25 @@ TEST_F(hmm, migrate_anon_huge_fault)
>  		}
>  
>  		close(pagemap_fd);
> +
> +		/* A second open must read the last device-private subpage. */
> +		fd2 = hmm_open(variant->device_number);
> +		ASSERT_GE(fd2, 0);
> +		buffer->ptr = (char *)map + (npages - 1) * self->page_size;
> +		ret = hmm_dmirror_cmd(fd2, HMM_DMIRROR_READ, buffer, 1);
> +		buffer->ptr = map;
> +		ASSERT_EQ(ret, 0);
> +		ASSERT_EQ(buffer->cpages, 1);
> +		ASSERT_EQ(*(int *)buffer->mirror,
> +			  (npages - 1) * self->page_size / sizeof(int));
>  	}
>  
>  	/* Fault pages back to system memory and check them. */
>  	for (i = 0, ptr = buffer->ptr; i < size / sizeof(*ptr); ++i)
>  		ASSERT_EQ(ptr[i], i);
>  
> +	if (fd2 >= 0)
> +		ASSERT_EQ(close(fd2), 0);
>  	buffer->ptr = old_ptr;
>  	hmm_buffer_free(buffer);
>  }
> _

Looks good to me also!

--Mika


^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH] mm/hmm: account for subpage offsets in device-private PMDs
  2026-11-22  9:52     ` Andrew Morton
  2026-11-22 12:04       ` Mika Penttilä
@ 2026-11-22 14:45       ` Aveline Noir
  2026-11-22 16:41         ` Mika Penttilä
  1 sibling, 1 reply; 10+ messages in thread
From: Aveline Noir @ 2026-11-22 14:45 UTC (permalink / raw)
  To: Andrew Morton
  Cc: airlied, apopple, balbirs, christian.koenig, dri-devel,
	Francois Dugast, intel-xe, jgg, jgg, leon, leonro, linux-kernel,
	linux-mm, Lorenzo Stoakes, Matthew Brost, mm-commits,
	Mika Penttilä, Patchwork, Shuah Khan, SeongJae Park,
	thomas.hellstrom, Vlastimil Babka, ziy

Hi Andrew,

On Sun, 22 Nov 2026 01:52:46 -0800 Andrew Morton wrote:

> Store the corresponding RAM backing page instead.

This avoids the oops on the first read through the second descriptor,
but leaves a stale backing page pointer after CPU faultback.

Since the mirror XArray entries don't take a page reference, they rely
on invalidation.  When the CPU later faults the page back to RAM,
dmirror_interval_invalidate() skips MMU_NOTIFY_MIGRATE for the same
mdevice, and dmirror_devmem_fault_alloc_and_copy() only clears the
migrating descriptor's entry before the backing folio is freed.  The
second descriptor still points to that freed backing page.

Reading the same page again through the second descriptor in QEMU
triggers a KASAN use-after-free in dmirror_do_read(), and the read
returns the stale data instead of what the CPU wrote.

The second descriptor's entry probably needs to be invalidated before
the backing page is freed, and it'd be nice to cover a second read
after faultback in the selftest too.

Thanks,
Aveline Noir

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH] mm/hmm: account for subpage offsets in device-private PMDs
  2026-11-22 14:45       ` Aveline Noir
@ 2026-11-22 16:41         ` Mika Penttilä
  2026-11-22 17:14           ` Andrew Morton
  0 siblings, 1 reply; 10+ messages in thread
From: Mika Penttilä @ 2026-11-22 16:41 UTC (permalink / raw)
  To: Aveline Noir, Andrew Morton
  Cc: airlied, apopple, balbirs, christian.koenig, dri-devel,
	Francois Dugast, intel-xe, jgg, jgg, leon, leonro, linux-kernel,
	linux-mm, Lorenzo Stoakes, Matthew Brost, mm-commits, Patchwork,
	Shuah Khan, SeongJae Park, thomas.hellstrom, Vlastimil Babka, ziy


On 11/22/26 16:45, Aveline Noir wrote:

> Hi Andrew,
>
> On Sun, 22 Nov 2026 01:52:46 -0800 Andrew Morton wrote:
>
>> Store the corresponding RAM backing page instead.
> This avoids the oops on the first read through the second descriptor,
> but leaves a stale backing page pointer after CPU faultback.
>
> Since the mirror XArray entries don't take a page reference, they rely
> on invalidation.  When the CPU later faults the page back to RAM,
> dmirror_interval_invalidate() skips MMU_NOTIFY_MIGRATE for the same
> mdevice, and dmirror_devmem_fault_alloc_and_copy() only clears the
> migrating descriptor's entry before the backing folio is freed.  The
> second descriptor still points to that freed backing page.
>
> Reading the same page again through the second descriptor in QEMU
> triggers a KASAN use-after-free in dmirror_do_read(), and the read
> returns the stale data instead of what the CPU wrote.

I think this is right, but a fairly artificial scenario: it needs two
descriptors open on the same test device in one test, a migration through one,
a read through the other, a CPU fault back, and then another read through
the second descriptor..

The fix lets the device private page content be read without an oops
in the context of one fd which is an improvement. 


>
> The second descriptor's entry probably needs to be invalidated before
> the backing page is freed, and it'd be nice to cover a second read
> after faultback in the selftest too.
>
> Thanks,
> Aveline Noir

--Mika


^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH] mm/hmm: account for subpage offsets in device-private PMDs
  2026-11-22 16:41         ` Mika Penttilä
@ 2026-11-22 17:14           ` Andrew Morton
  2026-11-22 17:37             ` Mika Penttilä
  0 siblings, 1 reply; 10+ messages in thread
From: Andrew Morton @ 2026-11-22 17:14 UTC (permalink / raw)
  To: Mika Penttilä
  Cc: Aveline Noir, airlied, apopple, balbirs, christian.koenig,
	dri-devel, Francois Dugast, intel-xe, jgg, jgg, leon, leonro,
	linux-kernel, linux-mm, Lorenzo Stoakes, Matthew Brost,
	mm-commits, Patchwork, Shuah Khan, SeongJae Park,
	thomas.hellstrom, Vlastimil Babka, ziy

On Sun, 22 Nov 2026 18:41:39 +0200 Mika Penttilä <mpenttil@redhat.com> wrote:

> 
> On 11/22/26 16:45, Aveline Noir wrote:
> 
> > Hi Andrew,
> >
> > On Sun, 22 Nov 2026 01:52:46 -0800 Andrew Morton wrote:
> >
> >> Store the corresponding RAM backing page instead.
> > This avoids the oops on the first read through the second descriptor,
> > but leaves a stale backing page pointer after CPU faultback.
> >
> > Since the mirror XArray entries don't take a page reference, they rely
> > on invalidation.  When the CPU later faults the page back to RAM,
> > dmirror_interval_invalidate() skips MMU_NOTIFY_MIGRATE for the same
> > mdevice, and dmirror_devmem_fault_alloc_and_copy() only clears the
> > migrating descriptor's entry before the backing folio is freed.  The
> > second descriptor still points to that freed backing page.
> >
> > Reading the same page again through the second descriptor in QEMU
> > triggers a KASAN use-after-free in dmirror_do_read(), and the read
> > returns the stale data instead of what the CPU wrote.
> 
> I think this is right, but a fairly artificial scenario: it needs two
> descriptors open on the same test device in one test, a migration through one,
> a read through the other, a CPU fault back, and then another read through
> the second descriptor..
> 
> The fix lets the device private page content be read without an oops
> in the context of one fd which is an improvement. 

May as well fix it.  I'll test this updated version.

Mika, please feel free to take this over, perhaps integrate it (or
something like it) into whatever you've got going on?  Or not, let me
know ;)


From: Andrew Morton <akpm@linux-foundation.org>
Subject: lib/test_hmm: invalidate all mirrors on device-private migration
Date: Sun, 22 Nov 2026 08:53:12 -0800

When one open of the fake HMM device migrates a folio to device-private
memory, a second open can fault the same address into its mirror page
table. The second mirror then stores the inaccessible device page, which
can oops the kernel when dmirror_do_read() copies from it. Store the
corresponding RAM backing subpage instead.

That conversion also makes the backing page's lifetime significant to
the second mirror. migrate_vma_setup() notifies all mirrors before CPU
faultback, but dmirror_interval_invalidate() skips the notification for
every open on the same device. The faulting open clears only its own
XArray entry. The other open can read or write the backing page after
migrate_vma_finalize() frees it. Aveline reported a KASAN use-after-free
and stale data through the second open.

Handle migration invalidation for every mirror under its mutex. This
clears pointers before faultback can free the backing pages and waits
for an active mirror read or write. During system-to-device migration,
the migrating open installs new entries after migration; other opens
refault on their next access. Remove the now redundant XArray erase
from the faultback copy loop.

Extend the PMD migration selftest to read its last device-private
subpage through a second open, then fault it back to RAM, change it,
and read the new value through that open. The trigger is available to
a process permitted to open /dev/hmm_dmirror* when CONFIG_TEST_HMM is
enabled; it does not depend on physical GPU hardware.

Link: https://lore.kernel.org/r/20261122015246.d8b6b98ff5a8157168156339@linux-foundation.org
Link: https://lore.kernel.org/r/179535872126.34.4789080961796247940.b4-reply@b4
Fixes: b2ef9f5a5cb3 ("mm/hmm/test: add selftest driver for HMM")
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reported-by: Aveline Noir <jm5905938@gmail.com>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Leon Romanovsky <leon@kernel.org>
Cc: Mika Penttilä <mpenttil@redhat.com>
Cc: Shuah Khan <shuah@kernel.org>
Cc: David Hildenbrand <david@kernel.org>
Cc: Liam R. Howlett <liam@infradead.org>
Cc: Vlastimil Babka <vbabka@kernel.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: <stable@vger.kernel.org>

 lib/test_hmm.c                         |   20 +++++++-----------
 tools/testing/selftests/mm/hmm-tests.c |   25 +++++++++++++++++++++++
 2 files changed, 33 insertions(+), 12 deletions(-)

--- a/lib/test_hmm.c~lib-test_hmm-invalidate-all-mirrors-on-device-private-migration
+++ a/lib/test_hmm.c
@@ -292,6 +292,12 @@ static int dmirror_do_fault(struct dmirr
 		page = hmm_pfn_to_page(*pfns);
 		WARN_ON(!page);
 
+		if (is_device_private_page(page)) {
+			struct folio *folio = page_folio(page);
+
+			page = BACKING_PAGE(&folio->page) +
+				folio_page_idx(folio, page);
+		}
 		entry = page;
 		if (*pfns & HMM_PFN_WRITE)
 			entry = xa_tag_pointer(entry, DPT_XA_TAG_WRITE);
@@ -327,14 +333,6 @@ static bool dmirror_interval_invalidate(
 {
 	struct dmirror *dmirror = container_of(mni, struct dmirror, notifier);
 
-	/*
-	 * Ignore invalidation callbacks for device private pages since
-	 * the invalidation is handled as part of the migration process.
-	 */
-	if (range->event == MMU_NOTIFY_MIGRATE &&
-	    range->owner == dmirror->mdevice)
-		return true;
-
 	if (mmu_notifier_range_blockable(range))
 		mutex_lock(&dmirror->mutex);
 	else if (!mutex_trylock(&dmirror->mutex))
@@ -1198,7 +1196,6 @@ static vm_fault_t dmirror_devmem_fault_a
 
 			src_page = pfn_to_page(page_to_pfn(spage) + i);
 
-			xa_erase(&dmirror->pt, addr >> PAGE_SHIFT);
 			addr += PAGE_SIZE;
 			copy_highpage(dst_page, src_page);
 		}
@@ -1822,9 +1819,8 @@ static vm_fault_t dmirror_devmem_fault(s
 	}
 	migrate_vma_pages(&args);
 	/*
-	 * No device finalize step is needed since
-	 * dmirror_devmem_fault_alloc_and_copy() will have already
-	 * invalidated the device page table.
+	 * The migration notifier has invalidated all mirror page tables
+	 * before the backing pages can be freed.
 	 */
 	migrate_vma_finalize(&args);
 err:
--- a/tools/testing/selftests/mm/hmm-tests.c~lib-test_hmm-invalidate-all-mirrors-on-device-private-migration
+++ a/tools/testing/selftests/mm/hmm-tests.c
@@ -2087,6 +2087,7 @@ TEST_F(hmm, migrate_anon_huge_fault)
 	void *old_ptr;
 	void *map;
 	int pagemap_fd;
+	int fd2 = -1;
 	int *ptr;
 	int ret;
 
@@ -2143,12 +2144,36 @@ TEST_F(hmm, migrate_anon_huge_fault)
 		}
 
 		close(pagemap_fd);
+
+		/* Read a device-private subpage through another open. */
+		fd2 = hmm_open(variant->device_number);
+		ASSERT_GE(fd2, 0);
+		buffer->ptr = (char *)map + (npages - 1) * self->page_size;
+		ret = hmm_dmirror_cmd(fd2, HMM_DMIRROR_READ, buffer, 1);
+		buffer->ptr = map;
+		ASSERT_EQ(ret, 0);
+		ASSERT_EQ(buffer->cpages, 1);
+		ASSERT_EQ(*(int *)buffer->mirror,
+			  (npages - 1) * self->page_size / sizeof(int));
 	}
 
 	/* Fault pages back to system memory and check them. */
 	for (i = 0, ptr = buffer->ptr; i < size / sizeof(*ptr); ++i)
 		ASSERT_EQ(ptr[i], i);
 
+	if (fd2 >= 0) {
+		/* This open must not retain the freed device backing page. */
+		ptr[(npages - 1) * self->page_size / sizeof(*ptr)] = -1;
+		buffer->ptr = (char *)map + (npages - 1) * self->page_size;
+		ret = hmm_dmirror_cmd(fd2, HMM_DMIRROR_READ, buffer, 1);
+		buffer->ptr = map;
+		ASSERT_EQ(ret, 0);
+		ASSERT_EQ(buffer->cpages, 1);
+		ASSERT_EQ(buffer->faults, 1);
+		ASSERT_EQ(*(int *)buffer->mirror, -1);
+		ASSERT_EQ(close(fd2), 0);
+	}
+
 	buffer->ptr = old_ptr;
 	hmm_buffer_free(buffer);
 }
_


^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH] mm/hmm: account for subpage offsets in device-private PMDs
  2026-11-22 17:14           ` Andrew Morton
@ 2026-11-22 17:37             ` Mika Penttilä
  0 siblings, 0 replies; 10+ messages in thread
From: Mika Penttilä @ 2026-11-22 17:37 UTC (permalink / raw)
  To: Andrew Morton
  Cc: Aveline Noir, airlied, apopple, balbirs, christian.koenig,
	dri-devel, Francois Dugast, intel-xe, jgg, jgg, leon, leonro,
	linux-kernel, linux-mm, Lorenzo Stoakes, Matthew Brost,
	mm-commits, Patchwork, Shuah Khan, SeongJae Park,
	thomas.hellstrom, Vlastimil Babka, ziy


On 11/22/26 19:14, Andrew Morton wrote:
> On Sun, 22 Nov 2026 18:41:39 +0200 Mika Penttilä <mpenttil@redhat.com> wrote:
>
>> On 11/22/26 16:45, Aveline Noir wrote:
>>
>>> Hi Andrew,
>>>
>>> On Sun, 22 Nov 2026 01:52:46 -0800 Andrew Morton wrote:
>>>
>>>> Store the corresponding RAM backing page instead.
>>> This avoids the oops on the first read through the second descriptor,
>>> but leaves a stale backing page pointer after CPU faultback.
>>>
>>> Since the mirror XArray entries don't take a page reference, they rely
>>> on invalidation.  When the CPU later faults the page back to RAM,
>>> dmirror_interval_invalidate() skips MMU_NOTIFY_MIGRATE for the same
>>> mdevice, and dmirror_devmem_fault_alloc_and_copy() only clears the
>>> migrating descriptor's entry before the backing folio is freed.  The
>>> second descriptor still points to that freed backing page.
>>>
>>> Reading the same page again through the second descriptor in QEMU
>>> triggers a KASAN use-after-free in dmirror_do_read(), and the read
>>> returns the stale data instead of what the CPU wrote.
>> I think this is right, but a fairly artificial scenario: it needs two
>> descriptors open on the same test device in one test, a migration through one,
>> a read through the other, a CPU fault back, and then another read through
>> the second descriptor..
>>
>> The fix lets the device private page content be read without an oops
>> in the context of one fd which is an improvement. 
> May as well fix it.  I'll test this updated version.
>
> Mika, please feel free to take this over, perhaps integrate it (or
> something like it) into whatever you've got going on?  Or not, let me
> know ;)

I can handle it in the context where it made difference for me.

>
> From: Andrew Morton <akpm@linux-foundation.org>
> Subject: lib/test_hmm: invalidate all mirrors on device-private migration
> Date: Sun, 22 Nov 2026 08:53:12 -0800
>
> When one open of the fake HMM device migrates a folio to device-private
> memory, a second open can fault the same address into its mirror page
> table. The second mirror then stores the inaccessible device page, which
> can oops the kernel when dmirror_do_read() copies from it. Store the
> corresponding RAM backing subpage instead.
>
> That conversion also makes the backing page's lifetime significant to
> the second mirror. migrate_vma_setup() notifies all mirrors before CPU
> faultback, but dmirror_interval_invalidate() skips the notification for
> every open on the same device. The faulting open clears only its own
> XArray entry. The other open can read or write the backing page after
> migrate_vma_finalize() frees it. Aveline reported a KASAN use-after-free
> and stale data through the second open.
>
> Handle migration invalidation for every mirror under its mutex. This
> clears pointers before faultback can free the backing pages and waits
> for an active mirror read or write. During system-to-device migration,
> the migrating open installs new entries after migration; other opens
> refault on their next access. Remove the now redundant XArray erase
> from the faultback copy loop.
>
> Extend the PMD migration selftest to read its last device-private
> subpage through a second open, then fault it back to RAM, change it,
> and read the new value through that open. The trigger is available to
> a process permitted to open /dev/hmm_dmirror* when CONFIG_TEST_HMM is
> enabled; it does not depend on physical GPU hardware.
>
> Link: https://lore.kernel.org/r/20261122015246.d8b6b98ff5a8157168156339@linux-foundation.org
> Link: https://lore.kernel.org/r/179535872126.34.4789080961796247940.b4-reply@b4
> Fixes: b2ef9f5a5cb3 ("mm/hmm/test: add selftest driver for HMM")
> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
> Reported-by: Aveline Noir <jm5905938@gmail.com>
> Cc: Ralph Campbell <rcampbell@nvidia.com>
> Cc: Jason Gunthorpe <jgg@ziepe.ca>
> Cc: Leon Romanovsky <leon@kernel.org>
> Cc: Mika Penttilä <mpenttil@redhat.com>
> Cc: Shuah Khan <shuah@kernel.org>
> Cc: David Hildenbrand <david@kernel.org>
> Cc: Liam R. Howlett <liam@infradead.org>
> Cc: Vlastimil Babka <vbabka@kernel.org>
> Cc: Mike Rapoport <rppt@kernel.org>
> Cc: Michal Hocko <mhocko@suse.com>
> Cc: <stable@vger.kernel.org>
>
>  lib/test_hmm.c                         |   20 +++++++-----------
>  tools/testing/selftests/mm/hmm-tests.c |   25 +++++++++++++++++++++++
>  2 files changed, 33 insertions(+), 12 deletions(-)
>
> --- a/lib/test_hmm.c~lib-test_hmm-invalidate-all-mirrors-on-device-private-migration
> +++ a/lib/test_hmm.c
> @@ -292,6 +292,12 @@ static int dmirror_do_fault(struct dmirr
>  		page = hmm_pfn_to_page(*pfns);
>  		WARN_ON(!page);
>  
> +		if (is_device_private_page(page)) {
> +			struct folio *folio = page_folio(page);
> +
> +			page = BACKING_PAGE(&folio->page) +
> +				folio_page_idx(folio, page);
> +		}
>  		entry = page;
>  		if (*pfns & HMM_PFN_WRITE)
>  			entry = xa_tag_pointer(entry, DPT_XA_TAG_WRITE);
> @@ -327,14 +333,6 @@ static bool dmirror_interval_invalidate(
>  {
>  	struct dmirror *dmirror = container_of(mni, struct dmirror, notifier);
>  
> -	/*
> -	 * Ignore invalidation callbacks for device private pages since
> -	 * the invalidation is handled as part of the migration process.
> -	 */
> -	if (range->event == MMU_NOTIFY_MIGRATE &&
> -	    range->owner == dmirror->mdevice)
> -		return true;
> -
>  	if (mmu_notifier_range_blockable(range))
>  		mutex_lock(&dmirror->mutex);
>  	else if (!mutex_trylock(&dmirror->mutex))
> @@ -1198,7 +1196,6 @@ static vm_fault_t dmirror_devmem_fault_a
>  
>  			src_page = pfn_to_page(page_to_pfn(spage) + i);
>  
> -			xa_erase(&dmirror->pt, addr >> PAGE_SHIFT);
>  			addr += PAGE_SIZE;
>  			copy_highpage(dst_page, src_page);
>  		}
> @@ -1822,9 +1819,8 @@ static vm_fault_t dmirror_devmem_fault(s
>  	}
>  	migrate_vma_pages(&args);
>  	/*
> -	 * No device finalize step is needed since
> -	 * dmirror_devmem_fault_alloc_and_copy() will have already
> -	 * invalidated the device page table.
> +	 * The migration notifier has invalidated all mirror page tables
> +	 * before the backing pages can be freed.
>  	 */
>  	migrate_vma_finalize(&args);
>  err:
> --- a/tools/testing/selftests/mm/hmm-tests.c~lib-test_hmm-invalidate-all-mirrors-on-device-private-migration
> +++ a/tools/testing/selftests/mm/hmm-tests.c
> @@ -2087,6 +2087,7 @@ TEST_F(hmm, migrate_anon_huge_fault)
>  	void *old_ptr;
>  	void *map;
>  	int pagemap_fd;
> +	int fd2 = -1;
>  	int *ptr;
>  	int ret;
>  
> @@ -2143,12 +2144,36 @@ TEST_F(hmm, migrate_anon_huge_fault)
>  		}
>  
>  		close(pagemap_fd);
> +
> +		/* Read a device-private subpage through another open. */
> +		fd2 = hmm_open(variant->device_number);
> +		ASSERT_GE(fd2, 0);
> +		buffer->ptr = (char *)map + (npages - 1) * self->page_size;
> +		ret = hmm_dmirror_cmd(fd2, HMM_DMIRROR_READ, buffer, 1);
> +		buffer->ptr = map;
> +		ASSERT_EQ(ret, 0);
> +		ASSERT_EQ(buffer->cpages, 1);
> +		ASSERT_EQ(*(int *)buffer->mirror,
> +			  (npages - 1) * self->page_size / sizeof(int));
>  	}
>  
>  	/* Fault pages back to system memory and check them. */
>  	for (i = 0, ptr = buffer->ptr; i < size / sizeof(*ptr); ++i)
>  		ASSERT_EQ(ptr[i], i);
>  
> +	if (fd2 >= 0) {
> +		/* This open must not retain the freed device backing page. */
> +		ptr[(npages - 1) * self->page_size / sizeof(*ptr)] = -1;
> +		buffer->ptr = (char *)map + (npages - 1) * self->page_size;
> +		ret = hmm_dmirror_cmd(fd2, HMM_DMIRROR_READ, buffer, 1);
> +		buffer->ptr = map;
> +		ASSERT_EQ(ret, 0);
> +		ASSERT_EQ(buffer->cpages, 1);
> +		ASSERT_EQ(buffer->faults, 1);
> +		ASSERT_EQ(*(int *)buffer->mirror, -1);
> +		ASSERT_EQ(close(fd2), 0);
> +	}
> +
>  	buffer->ptr = old_ptr;
>  	hmm_buffer_free(buffer);
>  }
> _

--Mika


^ permalink raw reply	[flat|nested] 10+ messages in thread

end of thread, other threads:[~2026-11-22 17:37 UTC | newest]

Thread overview: 10+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-10-31  4:07 [PATCH] mm/hmm: account for subpage offsets in device-private PMDs Aveline Noir
2026-11-22  0:00 ` Andrew Morton
2026-11-22  5:55   ` Mika Penttilä
2026-11-22  9:52     ` Andrew Morton
2026-11-22 12:04       ` Mika Penttilä
2026-11-22 14:45       ` Aveline Noir
2026-11-22 16:41         ` Mika Penttilä
2026-11-22 17:14           ` Andrew Morton
2026-11-22 17:37             ` Mika Penttilä
2026-11-22  0:01 ` + mm-hmm-account-for-subpage-offsets-in-device-private-pmds.patch added to mm-hotfixes-unstable branch Andrew Morton

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox