Roughly half of a typical pmap_mincore() implementation is machine-

independent code.  Move this code into mincore(), and eliminate the
page queues lock from pmap_mincore().

Push down the page queues lock into pmap_clear_modify(),
pmap_clear_reference(), and pmap_is_modified().  Assert that these
functions are never passed an unmanaged page.

Eliminate an inaccurate comment from powerpc/powerpc/mmu_if.m:
Contrary to what the comment says, pmap_mincore() is not simply an
optimization.  Without a complete pmap_mincore() implementation,
mincore() cannot return either MINCORE_MODIFIED or MINCORE_REFERENCED
because only the pmap can provide this information.

Eliminate the page queues lock from vfs_setdirty_locked_object(),
vm_pageout_clean(), vm_object_page_collect_flush(), and
vm_object_page_clean().  Generally speaking, these are all accesses
to the page's dirty field, which are synchronized by the containing
vm object's lock.

Reduce the scope of the page queues lock in vm_object_madvise() and
vm_page_dontneed().

Reviewed by:	kib (an earlier version)

2 files changed, 68 insertions, 12 deletions

diff --git a/sys/powerpc/aim/mmu_oea.c b/sys/powerpc/aim/mmu_oea.c
index a2f120b..b5d0ae7 100644
--- a/sys/powerpc/aim/mmu_oea.c
+++ b/sys/powerpc/aim/mmu_oea.c
@@ -1290,29 +1290,57 @@ moea_is_referenced(mmu_t mmu, vm_page_t m)
 boolean_t
 moea_is_modified(mmu_t mmu, vm_page_t m)
 {
+	boolean_t rv;
 
-	if ((m->flags & (PG_FICTITIOUS |PG_UNMANAGED)) != 0)
-		return (FALSE);
+	KASSERT((m->flags & (PG_FICTITIOUS | PG_UNMANAGED)) == 0,
+	    ("moea_is_modified: page %p is not managed", m));
 
-	return (moea_query_bit(m, PTE_CHG));
+	/*
+	 * If the page is not VPO_BUSY, then PG_WRITEABLE cannot be
+	 * concurrently set while the object is locked.  Thus, if PG_WRITEABLE
+	 * is clear, no PTEs can have PTE_CHG set.
+	 */
+	VM_OBJECT_LOCK_ASSERT(m->object, MA_OWNED);
+	if ((m->oflags & VPO_BUSY) == 0 &&
+	    (m->flags & PG_WRITEABLE) == 0)
+		return (FALSE);
+	vm_page_lock_queues();
+	rv = moea_query_bit(m, PTE_CHG);
+	vm_page_unlock_queues();
+	return (rv);
 }
 
 void
 moea_clear_reference(mmu_t mmu, vm_page_t m)
 {
 
-	if ((m->flags & (PG_FICTITIOUS | PG_UNMANAGED)) != 0)
-		return;
+	KASSERT((m->flags & (PG_FICTITIOUS | PG_UNMANAGED)) == 0,
+	    ("moea_clear_reference: page %p is not managed", m));
+	vm_page_lock_queues();
 	moea_clear_bit(m, PTE_REF, NULL);
+	vm_page_unlock_queues();
 }
 
 void
 moea_clear_modify(mmu_t mmu, vm_page_t m)
 {
 
-	if ((m->flags & (PG_FICTITIOUS | PG_UNMANAGED)) != 0)
+	KASSERT((m->flags & (PG_FICTITIOUS | PG_UNMANAGED)) == 0,
+	    ("moea_clear_modify: page %p is not managed", m));
+	VM_OBJECT_LOCK_ASSERT(m->object, MA_OWNED);
+	KASSERT((m->oflags & VPO_BUSY) == 0,
+	    ("moea_clear_modify: page %p is busy", m));
+
+	/*
+	 * If the page is not PG_WRITEABLE, then no PTEs can have PTE_CHG
+	 * set.  If the object containing the page is locked and the page is
+	 * not VPO_BUSY, then PG_WRITEABLE cannot be concurrently set.
+	 */
+	if ((m->flags & PG_WRITEABLE) == 0)
 		return;
+	vm_page_lock_queues();
 	moea_clear_bit(m, PTE_CHG, NULL);
+	vm_page_unlock_queues();
 }
 
 /*
diff --git a/sys/powerpc/aim/mmu_oea64.c b/sys/powerpc/aim/mmu_oea64.c
index 8b6453a..140e949 100644
--- a/sys/powerpc/aim/mmu_oea64.c
+++ b/sys/powerpc/aim/mmu_oea64.c
@@ -1485,29 +1485,57 @@ moea64_is_referenced(mmu_t mmu, vm_page_t m)
 boolean_t
 moea64_is_modified(mmu_t mmu, vm_page_t m)
 {
+	boolean_t rv;
 
-	if ((m->flags & (PG_FICTITIOUS | PG_UNMANAGED)) != 0)
-		return (FALSE);
+	KASSERT((m->flags & (PG_FICTITIOUS | PG_UNMANAGED)) == 0,
+	    ("moea64_is_modified: page %p is not managed", m));
 
-	return (moea64_query_bit(m, LPTE_CHG));
+	/*
+	 * If the page is not VPO_BUSY, then PG_WRITEABLE cannot be
+	 * concurrently set while the object is locked.  Thus, if PG_WRITEABLE
+	 * is clear, no PTEs can have LPTE_CHG set.
+	 */
+	VM_OBJECT_LOCK_ASSERT(m->object, MA_OWNED);
+	if ((m->oflags & VPO_BUSY) == 0 &&
+	    (m->flags & PG_WRITEABLE) == 0)
+		return (FALSE);
+	vm_page_lock_queues();
+	rv = moea64_query_bit(m, LPTE_CHG);
+	vm_page_unlock_queues();
+	return (rv);
 }
 
 void
 moea64_clear_reference(mmu_t mmu, vm_page_t m)
 {
 
-	if ((m->flags & (PG_FICTITIOUS | PG_UNMANAGED)) != 0)
-		return;
+	KASSERT((m->flags & (PG_FICTITIOUS | PG_UNMANAGED)) == 0,
+	    ("moea64_clear_reference: page %p is not managed", m));
+	vm_page_lock_queues();
 	moea64_clear_bit(m, LPTE_REF, NULL);
+	vm_page_unlock_queues();
 }
 
 void
 moea64_clear_modify(mmu_t mmu, vm_page_t m)
 {
 
-	if ((m->flags & (PG_FICTITIOUS | PG_UNMANAGED)) != 0)
+	KASSERT((m->flags & (PG_FICTITIOUS | PG_UNMANAGED)) == 0,
+	    ("moea64_clear_modify: page %p is not managed", m));
+	VM_OBJECT_LOCK_ASSERT(m->object, MA_OWNED);
+	KASSERT((m->oflags & VPO_BUSY) == 0,
+	    ("moea64_clear_modify: page %p is busy", m));
+
+	/*
+	 * If the page is not PG_WRITEABLE, then no PTEs can have LPTE_CHG
+	 * set.  If the object containing the page is locked and the page is
+	 * not VPO_BUSY, then PG_WRITEABLE cannot be concurrently set.
+	 */
+	if ((m->flags & PG_WRITEABLE) == 0)
 		return;
+	vm_page_lock_queues();
 	moea64_clear_bit(m, LPTE_CHG, NULL);
+	vm_page_unlock_queues();
 }
 
 /*