mm/hugetlb: don't map folios writable without VM_WRITE when copying during fork()
Subsystems (2)
PATCH 1/1 mm/hugetlb: don't map folios writable without VM_WRITE when copying during fork()
If we have to trigger a hugetlb folio copy during fork() because the anon folio might be pinned, we currently unconditionally create a writable PTE.
However, the VMA might not have write permissions (VM_WRITE) at that point.
Fix it by checking the VMA for VM_WRITE. Make the code less error prone by moving checking for VM_WRITE into make_huge_pte(), and letting callers only specify whether we should try making it writable.
A simple reproducer that longter-pins the folios using liburing to then mprotect(PROT_READ) the folios befor fork() [1] results in:
Before:
[FAIL] access should not have worked
After:
[PASS] access did not work as expected
[1] https://gitlab.com/davidhildenbrand/scratchspace/-/raw/main/reproducers/hugetlb-mkwrite-fork.c
This is rather a corner case, so stable might not be warranted.
--- mm/hugetlb.c | 18 ++++++------------ 1 file changed, 6 insertions(+), 12 deletions(-)
mm/hugetlb.c +6 -12
@@ -5141,12 +5141,12 @@ const struct vm_operations_struct hugetlb_vm_ops = { };static pte_t make_huge_pte (struct vm_area_struct *vma,struct page *page, -int writable) +bool try_mkwrite) {pte_t entry;unsigned int shift =huge_page_shift (hstate_vma (vma)); -if (writable) { +if (try_mkwrite && (vma->vm_flags &VM_WRITE )) { entry =huge_pte_mkwrite (huge_pte_mkdirty (mk_huge_pte (page, vma->vm_page_prot))); }else {
@@ -5199,7 +5199,7 @@ static void hugetlb_install_folio(struct vm_area_struct *vma, pte_t *ptep, unsigned long addr, struct folio *new_folio, pte_t old, unsigned long sz) { - pte_t newpte = make_huge_pte(vma, &new_folio->page, 1); + pte_t newpte = make_huge_pte(vma, &new_folio->page, true);__folio_mark_uptodate (new_folio);hugetlb_add_new_anon_rmap (new_folio, vma, addr);
@@ -6223,8 +6223,7 @@ static vm_fault_t hugetlb_no_page(struct address_space *mapping,hugetlb_add_new_anon_rmap (folio, vma, vmf->address); elsehugetlb_add_file_rmap (folio ); - new_pte =make_huge_pte (vma, &folio->page, ((vma->vm_flags &VM_WRITE ) - && (vma->vm_flags &VM_SHARED ))); + new_pte =make_huge_pte (vma, &folio->page, vma->vm_flags &VM_SHARED );/* * If this pte was previously wr-protected, keep it wr-protected even * if populated.
@@ -6556,7 +6555,6 @@ int hugetlb_mfill_atomic_pte(pte_t *dst_pte,spinlock_t *ptl;int ret = -ENOMEM ;struct folio *folio; -int writable;bool folio_in_pagecache = false;if (uffd_flags_mode_is (flags,MFILL_ATOMIC_POISON )) {
@@ -6710,12 +6708,8 @@ int hugetlb_mfill_atomic_pte(pte_t *dst_pte,* For either: (1) CONTINUE on a non-shared VMA, or (2) UFFDIO_COPY * with wp flag set, don't set pte write bit. */ -if (wp_enabled || (is_continue && !vm_shared)) - writable =0 ; -else - writable = dst_vma->vm_flags &VM_WRITE ; - - _dst_pte =make_huge_pte (dst_vma, &folio->page, writable); + _dst_pte =make_huge_pte (dst_vma, &folio->page, + !wp_enabled && !(is_continue && !vm_shared));/* * Always mark UFFDIO_COPY page dirty; note that this may not be * extremely important for hugetlbfs for now since swapping is not
Signature · 1 line
2.47.1
Quoted 22 lines · David Hildenbrand
If we have to trigger a hugetlb folio copy during
fork()because the anon folio might be pinned, we currently unconditionally create a writable PTE.However, the VMA might not have write permissions (
VM_WRITE) at that point.Fix it by checking the VMA for
VM_WRITE. Make the code less error prone by moving checking forVM_WRITEintomake_huge_pte(), and letting callers only specify whether we should try making it writable.A simple reproducer that longter-pins the folios using liburing to then
mprotect(PROT_READ)the folios beforfork()[1] results in:Before:
[FAIL] access should not have workedAfter:
[PASS] access did not work as expected[1] https://gitlab.com/davidhildenbrand/scratchspace/-/raw/main/reproducers/hugetlb-mkwrite-fork.c
Quoted 8 lines
This is rather a corner case, so stable might not be warranted.
Fixes 4eae4efa2c29 ("hugetlb: do early cow when page pinned on src mm")Cc Muchun Song <[email protected]>Cc Andrew Morton <[email protected]>Cc Peter Xu <[email protected]>Cc Guillaume Morin <[email protected]>Signed-off-by David Hildenbrand <[email protected]>
Thanks,
Signature · 1 line
Peter Xu