Thread (107 messages) 107 messages, 7 authors, 2020-04-08

Re: [RFC PATCH v9 10/27] x86/mm: Update pte_modify, pmd_modify, and _PAGE_CHG_MASK for _PAGE_DIRTY_SW

From: Dave Hansen <hidden>
Date: 2020-02-26 22:02:10
Also in: linux-arch, linux-doc, linux-mm, lkml

The subject really needs work.  Could you think of a way to summarize
the changes here in english as opposed to just listing the symbols you
modified?

I think we could probably just auto-generate subjects for patches if the
existing one were sufficient.

On 2/5/20 10:19 AM, Yu-cheng Yu wrote:
After the introduction of _PAGE_DIRTY_SW, pte_modify and pmd_modify need to
set the Dirty bit accordingly: if Shadow Stack is enabled and _PAGE_RW is
cleared, use _PAGE_DIRTY_SW; otherwise _PAGE_DIRTY_HW.
You've basically gone and written the code's if() statement in english
here.  That doesn't really help me understand the patch.
Since the Dirty bit is modify by pte_modify(), remove _PAGE_DIRTY_HW from
PAGE_CHG_MASK.
			 ^ modified

This is a great example of a changelog that adds very little value.
It's following the comments and doing what they say, but it's pretty
obvious that the analysis stopped there.

What *kinds* of bits are in _PAGE_CHG_MASK or not?  What changed about
_PAGE_DIRTY_HW.  By this definition, shouldn't _PAGE_DIRTY_SW have
technically been in this mask before this patch?
quoted hunk ↗ jump to hunk
diff --git a/arch/x86/include/asm/pgtable.h b/arch/x86/include/asm/pgtable.h
index 62aeb118bc36..2733e7ec16b3 100644
--- a/arch/x86/include/asm/pgtable.h
+++ b/arch/x86/include/asm/pgtable.h
@@ -702,6 +702,14 @@ static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 	val &= _PAGE_CHG_MASK;
 	val |= check_pgprot(newprot) & ~_PAGE_CHG_MASK;
 	val = flip_protnone_guard(oldval, val, PTE_PFN_MASK);
+
+	if (pte_dirty(pte)) {
+		if (static_cpu_has(X86_FEATURE_SHSTK) && !(val & _PAGE_RW))
+			val |= _PAGE_DIRTY_SW;
+		else
+			val |= _PAGE_DIRTY_HW;
+	}
+
 	return __pte(val);
 }
OK, so this is a path we use for changing bunches of PTEs to 'newprot'.
 It doesn't use the pte_*() helpers that the previous patch fixed up, so
we need a new site.

Right?

Maybe that would make good changelog text.

Also, couldn't we just have a pte_fixup() function or something that did
this logic and could be shared?
quoted hunk ↗ jump to hunk
@@ -712,6 +720,14 @@ static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
 	val &= _HPAGE_CHG_MASK;
 	val |= check_pgprot(newprot) & ~_HPAGE_CHG_MASK;
 	val = flip_protnone_guard(oldval, val, PHYSICAL_PMD_PAGE_MASK);
+
+	if (pmd_dirty(pmd)) {
+		if (static_cpu_has(X86_FEATURE_SHSTK) && !(val & _PAGE_RW))
+			val |= _PAGE_DIRTY_SW;
+		else
+			val |= _PAGE_DIRTY_HW;
+	}
+
 	return __pmd(val);
 }
 
diff --git a/arch/x86/include/asm/pgtable_types.h b/arch/x86/include/asm/pgtable_types.h
index 826823df917f..e7e28bf7e919 100644
--- a/arch/x86/include/asm/pgtable_types.h
+++ b/arch/x86/include/asm/pgtable_types.h
@@ -150,8 +150,8 @@
  * instance, and is *not* included in this mask since
  * pte_modify() does modify it.
  */
-#define _PAGE_CHG_MASK	(PTE_PFN_MASK | _PAGE_PCD | _PAGE_PWT |		\
-			 _PAGE_SPECIAL | _PAGE_ACCESSED | _PAGE_DIRTY_HW |	\
+#define _PAGE_CHG_MASK	(PTE_PFN_MASK | _PAGE_PCD | _PAGE_PWT |	\
+			 _PAGE_SPECIAL | _PAGE_ACCESSED |	\
 			 _PAGE_SOFT_DIRTY | _PAGE_DEVMAP)
 #define _HPAGE_CHG_MASK (_PAGE_CHG_MASK | _PAGE_PSE)
  
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help