mm/userfaultfd: Allow hugetlb change protection upon poison entry
Subsystems (3)
PATCH 1/1 mm/userfaultfd: Allow hugetlb change protection upon poison entry
From: Peter Xu <[email protected]>
After UFFDIO_POISON, there can be two kinds of hugetlb pte markers, either the POISON one or UFFD_WP one.
Allow change protection to run on a poisoned marker just like !hugetlb cases, ignoring the marker irrelevant of the permission.
Here the two bits are mutual exclusive. For example, when install a poisoned entry it must not be UFFD_WP already (by checking pte_none() before such install). And it also means if UFFD_WP is set there must have no POISON bit set. It makes sense because UFFD_WP is a bit to reflect permission, and permissions do not apply if the pte is poisoned and destined to sigbus.
So here we simply check uffd_wp bit set first, do nothing otherwise.
Attach the Fixes to UFFDIO_POISON work, as before that it should not be possible to have poison entry for hugetlb (e.g., hugetlb doesn't do swap, so no chance of swapin errors).
--- mm/hugetlb.c | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-)
mm/hugetlb.c +7 -3
@@ -6960,9 +6960,13 @@ long hugetlb_change_protection(struct vm_area_struct *vma,if (!pte_same (pte, newpte))set_huge_pte_at (mm, address, ptep, newpte, psize); }else if (unlikely (is_pte_marker (pte))) { -/* No other markers apply for now. */ -WARN_ON_ONCE (!pte_marker_uffd_wp (pte)); -if (uffd_wp_resolve) +/* +* Do nothing on a poison marker; page is +* corrupted, permissons do not apply. Here +* pte_marker_uffd_wp()==true implies !poison +* because they're mutual exclusive. +*/ +if (pte_marker_uffd_wp (pte) && uffd_wp_resolve)/* Safe to modify directly (non-present->none). */ huge_pte_clear (mm, address, ptep, psize); }else if (!huge_pte_none (pte)) {
Signature · 1 line
2.44.0
Quoted 43 lines · On 06.04.24 01:19, [email protected] wrote:
From: Peter Xu <[email protected]>
After
UFFDIO_POISON, there can be two kinds of hugetlb pte markers, either the POISON one orUFFD_WPone.Allow change protection to run on a poisoned marker just like !hugetlb cases, ignoring the marker irrelevant of the permission.
Here the two bits are mutual exclusive. For example, when install a poisoned entry it must not be
UFFD_WPalready (by checkingpte_none()before such install). And it also means ifUFFD_WPis set there must have no POISON bit set. It makes sense becauseUFFD_WPis a bit to reflect permission, and permissions do not apply if the pte is poisoned and destined to sigbus.So here we simply check
uffd_wpbit set first, do nothing otherwise.Attach the Fixes to
UFFDIO_POISONwork, as before that it should not be possible to have poison entry for hugetlb (e.g., hugetlb doesn't do swap, so no chance of swapin errors).Cc Axel Rasmussen <[email protected]>Cc David Hildenbrand <[email protected]>Cc linux-stable <[email protected]> # 6.6+Reported-by [email protected]Fixes fc71884a5f59 ("mm: userfaultfd: add new UFFDIO_POISON ioctl")Signed-off-by Peter Xu <[email protected]>--- mm/hugetlb.c | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-)mm/hugetlb.c +0
-3@@ -6960,9 +6960,13 @@ long hugetlb_change_protection(struct vm_area_struct *vma,if (!pte_same (pte, newpte))set_huge_pte_at (mm, address, ptep, newpte, psize); }else if (unlikely (is_pte_marker (pte))) { -/* No other markers apply for now. */ -WARN_ON_ONCE (!pte_marker_uffd_wp (pte)); -if (uffd_wp_resolve)
mm/hugetlb.c long hugetlb_change_protection(struct vm_area_struct *vma, · line 6972
Quoted 10 lines
mm/hugetlb.c +7
-0+/* +* Do nothing on a poison marker; page is +* corrupted, permissons do not apply. Here +* pte_marker_uffd_wp()==true implies !poison +* because they're mutual exclusive. +*/ +if (pte_marker_uffd_wp (pte) && uffd_wp_resolve)/* Safe to modify directly (non-present->none). */ huge_pte_clear (mm, address, ptep, psize); }else if (!huge_pte_none (pte)) {
Signature · 3 lines
Cheers, David / dhildenb
Axel Rasmussen
Reviewed-byThanks for the fix, Peter!
Quoted 63 lines · David Hildenbrand
Quoted 43 lines · On 06.04.24 01:19, [email protected] wrote:
From: Peter Xu <[email protected]>
After
UFFDIO_POISON, there can be two kinds of hugetlb pte markers, either the POISON one orUFFD_WPone.Allow change protection to run on a poisoned marker just like !hugetlb cases, ignoring the marker irrelevant of the permission.
Here the two bits are mutual exclusive. For example, when install a poisoned entry it must not be
UFFD_WPalready (by checkingpte_none()before such install). And it also means ifUFFD_WPis set there must have no POISON bit set. It makes sense becauseUFFD_WPis a bit to reflect permission, and permissions do not apply if the pte is poisoned and destined to sigbus.So here we simply check
uffd_wpbit set first, do nothing otherwise.Attach the Fixes to
UFFDIO_POISONwork, as before that it should not be possible to have poison entry for hugetlb (e.g., hugetlb doesn't do swap, so no chance of swapin errors).Cc Axel Rasmussen <[email protected]>Cc David Hildenbrand <[email protected]>Cc linux-stable <[email protected]> # 6.6+Reported-by [email protected]Fixes fc71884a5f59 ("mm: userfaultfd: add new UFFDIO_POISON ioctl")Signed-off-by Peter Xu <[email protected]>--- mm/hugetlb.c | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-)mm/hugetlb.c +0
-3@@ -6960,9 +6960,13 @@ long hugetlb_change_protection(struct vm_area_struct *vma,if (!pte_same (pte, newpte))set_huge_pte_at (mm, address, ptep, newpte, psize); }else if (unlikely (is_pte_marker (pte))) { -/* No other markers apply for now. */ -WARN_ON_ONCE (!pte_marker_uffd_wp (pte)); -if (uffd_wp_resolve)mm/hugetlb.c long hugetlb_change_protection(struct vm_area_struct *vma, · line 6972
Quoted 10 lines
mm/hugetlb.c +7
-0+/* +* Do nothing on a poison marker; page is +* corrupted, permissons do not apply. Here +* pte_marker_uffd_wp()==true implies !poison +* because they're mutual exclusive. +*/ +if (pte_marker_uffd_wp (pte) && uffd_wp_resolve)/* Safe to modify directly (non-present->none). */ huge_pte_clear (mm, address, ptep, psize); }else if (!huge_pte_none (pte)) {Reviewed-by David Hildenbrand <[email protected]>--
Cheers,David / dhildenb