Re: [PATCH v2 1/3] arm64: hugetlb: Fix huge_pte_offset to return poisoned page table entries

3 messages, 3 authors, 2017-06-07 · open the first message on its own page

Re: [PATCH v2 1/3] arm64: hugetlb: Fix huge_pte_offset to return poisoned page table entries

From: Punit Agrawal <hidden>
Date: 2017-06-07 15:32:32

Will Deacon [off-list ref] writes:
On Wed, Jun 07, 2017 at 03:30:37PM +0100, Catalin Marinas wrote:
quoted
On Wed, Jun 07, 2017 at 02:47:32PM +0100, Will Deacon wrote:
quoted
On Wed, May 17, 2017 at 04:23:34PM +0100, Punit Agrawal wrote:
quoted
--- a/arch/arm64/mm/hugetlbpage.c
+++ b/arch/arm64/mm/hugetlbpage.c
@@ -136,36 +136,27 @@ pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr)
 {
 	pgd_t *pgd;
 	pud_t *pud;
-	pmd_t *pmd = NULL;
-	pte_t *pte = NULL;
+	pmd_t *pmd;
 
 	pgd = pgd_offset(mm, addr);
 	pr_debug("%s: addr:0x%lx pgd:%p\n", __func__, addr, pgd);
 	if (!pgd_present(*pgd))
 		return NULL;
+
 	pud = pud_offset(pgd, addr);
-	if (!pud_present(*pud))
+	if (pud_none(*pud))
 		return NULL;
Do you actually need this special case?
quoted
-
-	if (pud_huge(*pud))
+	/* swap or huge page */
+	if (!pud_present(*pud) || pud_huge(*pud))
... couldn't you just add a '|| pud_none(*pud)' in here?
I think an earlier version took this approach but...
quoted
quoted
quoted
 		return (pte_t *)pud;
But then you no longer return NULL if *pud == 0.
Does that actually matter? The bits of hugetlb code I looked at will
deferenced the returned pud and handle the huge_pte_none case correctly.
For hugetlb fault handling (hugetlb_fault()), returning NULL vs pointer
to the pud/pmd results in different behaviour. If we return the pud when
pud_none(), then we lose the resulting hugepage size check we get from
huge_pte_alloc().
Will

Re: [PATCH v2 1/3] arm64: hugetlb: Fix huge_pte_offset to return poisoned page table entries

From: Will Deacon <hidden>
Date: 2017-06-07 15:41:06

On Wed, Jun 07, 2017 at 04:32:28PM +0100, Punit Agrawal wrote:
Will Deacon [off-list ref] writes:
quoted
On Wed, Jun 07, 2017 at 03:30:37PM +0100, Catalin Marinas wrote:
quoted
On Wed, Jun 07, 2017 at 02:47:32PM +0100, Will Deacon wrote:
quoted
On Wed, May 17, 2017 at 04:23:34PM +0100, Punit Agrawal wrote:
quoted
--- a/arch/arm64/mm/hugetlbpage.c
+++ b/arch/arm64/mm/hugetlbpage.c
@@ -136,36 +136,27 @@ pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr)
 {
 	pgd_t *pgd;
 	pud_t *pud;
-	pmd_t *pmd = NULL;
-	pte_t *pte = NULL;
+	pmd_t *pmd;
 
 	pgd = pgd_offset(mm, addr);
 	pr_debug("%s: addr:0x%lx pgd:%p\n", __func__, addr, pgd);
 	if (!pgd_present(*pgd))
 		return NULL;
+
 	pud = pud_offset(pgd, addr);
-	if (!pud_present(*pud))
+	if (pud_none(*pud))
 		return NULL;
Do you actually need this special case?
quoted
-
-	if (pud_huge(*pud))
+	/* swap or huge page */
+	if (!pud_present(*pud) || pud_huge(*pud))
... couldn't you just add a '|| pud_none(*pud)' in here?
I think an earlier version took this approach but...
quoted
quoted
quoted
quoted
 		return (pte_t *)pud;
But then you no longer return NULL if *pud == 0.
Does that actually matter? The bits of hugetlb code I looked at will
deferenced the returned pud and handle the huge_pte_none case correctly.
For hugetlb fault handling (hugetlb_fault()), returning NULL vs pointer
to the pud/pmd results in different behaviour. If we return the pud when
pud_none(), then we lose the resulting hugepage size check we get from
huge_pte_alloc().
Ok, so does that mean that many of the huge_pte_none checks in mm/hugetlb.c
that operate on a huge_ptep_get of non-NULL output from huge_pte_offset are
actually redundant?

Will

Re: [PATCH v2 1/3] arm64: hugetlb: Fix huge_pte_offset to return poisoned page table entries

From: Catalin Marinas <catalin.marinas@arm.com>
Date: 2017-06-07 16:54:36

On Wed, Jun 07, 2017 at 04:32:28PM +0100, Punit Agrawal wrote:
Will Deacon [off-list ref] writes:
quoted
On Wed, Jun 07, 2017 at 03:30:37PM +0100, Catalin Marinas wrote:
quoted
On Wed, Jun 07, 2017 at 02:47:32PM +0100, Will Deacon wrote:
quoted
On Wed, May 17, 2017 at 04:23:34PM +0100, Punit Agrawal wrote:
quoted
--- a/arch/arm64/mm/hugetlbpage.c
+++ b/arch/arm64/mm/hugetlbpage.c
@@ -136,36 +136,27 @@ pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr)
 {
 	pgd_t *pgd;
 	pud_t *pud;
-	pmd_t *pmd = NULL;
-	pte_t *pte = NULL;
+	pmd_t *pmd;
 
 	pgd = pgd_offset(mm, addr);
 	pr_debug("%s: addr:0x%lx pgd:%p\n", __func__, addr, pgd);
 	if (!pgd_present(*pgd))
 		return NULL;
+
 	pud = pud_offset(pgd, addr);
-	if (!pud_present(*pud))
+	if (pud_none(*pud))
 		return NULL;
Do you actually need this special case?
quoted
-
-	if (pud_huge(*pud))
+	/* swap or huge page */
+	if (!pud_present(*pud) || pud_huge(*pud))
... couldn't you just add a '|| pud_none(*pud)' in here?
I think an earlier version took this approach but...
quoted
quoted
quoted
quoted
 		return (pte_t *)pud;
But then you no longer return NULL if *pud == 0.
Does that actually matter? The bits of hugetlb code I looked at will
deferenced the returned pud and handle the huge_pte_none case correctly.
For hugetlb fault handling (hugetlb_fault()), returning NULL vs pointer
to the pud/pmd results in different behaviour. If we return the pud when
pud_none(), then we lose the resulting hugepage size check we get from
huge_pte_alloc().
At a quick look, there are a few other places where not returning NULL
has some other effects (though I don't think any of them are fatal):

- copy_huge_tlb_page_range() - unnecessary allocation of a destination
  pud

- huge_pmd_share() - do we actually need the subsequent get_page()?

- page_vma_mapped_walk() - it even has a comment: "when pud is not
  present, pte will be NULL". Now, that's no longer true with swap
  entries but we'd never return NULL for a pud_none() case

Current code behaviour is to return NULL when !p*d_present(). We are
slightly relaxing this for swap entries while still returning NULL for
the p*d_none() case but I wouldn't go that far as to never return NULL
here.

-- 
Catalin
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help