On s390x, we actually need a pte_mkyoung() / pte_mkdirty() instead of
going via the page and leaving the PTE unmodified. E.g., if we only
mark the page accessed via mark_page_accessed() when doing a FOLL_TOUCH,
we'll miss to clear the HW invalid bit in the pte and subsequent accesses
via the MMU would still require a pagefault.
Otherwise, buffered I/O will loop forever because it will keep stumling
over the set HW invalid bit, requiring a page fault.
Reported-by: Andreas Gruenbacher <agruenba@xxxxxxxxxx>
Signed-off-by: David Hildenbrand <david@xxxxxxxxxx>
---
mm/gup.c | 32 +++++++++++++++++++++++++-------
1 file changed, 25 insertions(+), 7 deletions(-)
diff --git a/mm/gup.c b/mm/gup.c
index a9d4d724aef7..de3311feb377 100644
--- a/mm/gup.c
+++ b/mm/gup.c
@@ -587,15 +587,33 @@ static struct page *follow_page_pte(struct vm_area_struct *vma,
}
}
if (flags & FOLL_TOUCH) {
- if ((flags & FOLL_WRITE) &&
- !pte_dirty(pte) && !PageDirty(page))
- set_page_dirty(page);
/*
- * pte_mkyoung() would be more correct here, but atomic care
- * is needed to avoid losing the dirty bit: it is easier to use
- * mark_page_accessed().
+ * We have to be careful with updating the PTE on architectures
+ * that have a HW dirty bit: while updating the PTE we might
+ * lose that bit again and we'd need an atomic update: it is
+ * easier to leave the PTE untouched for these architectures.
+ *
+ * s390x doesn't have a hw referenced / dirty bit and e.g., sets
+ * the hw invalid bit in pte_mkold(), to catch further
+ * references. We have to update the PTE here to e.g., clear the
+ * invalid bit; otherwise, callers that rely on not requiring
+ * an MMU fault once GUP(FOLL_TOUCH) succeeded will loop forever
+ * because the page won't actually be accessible via the MMU.
*/
- mark_page_accessed(page);
+ if (IS_ENABLED(CONFIG_S390)) {
+ pte = pte_mkyoung(pte);
+ if (flags & FOLL_WRITE)
+ pte = pte_mkdirty(pte);
+ if (!pte_same(pte, *ptep)) {
+ set_pte_at(vma->vm_mm, address, ptep, pte);
+ update_mmu_cache(vma, address, ptep);
+ }
+ } else {
+ if ((flags & FOLL_WRITE) &&
+ !pte_dirty(pte) && !PageDirty(page))
+ set_page_dirty(page);
+ mark_page_accessed(page);
+ }
}
if ((flags & FOLL_MLOCK) && (vma->vm_flags & VM_LOCKED)) {
/* Do not mlock pte-mapped THP */
--
2.35.1
--
Thanks,
David / dhildenb
