Split the vm_page flags PG_WRITEABLE and PG_REFERENCED into atomic

flags field. Updates to the atomic flags are performed using the atomic
ops on the containing word, do not require any vm lock to be held, and
are non-blocking. The vm_page_aflag_set(9) and vm_page_aflag_clear(9)
functions are provided to modify afalgs.

Document the changes to flags field to only require the page lock.

Introduce vm_page_reference(9) function to provide a stable KPI and
KBI for filesystems like tmpfs and zfs which need to mark a page as
referenced.

Reviewed by:    alc, attilio
Tested by:      marius, flo (sparc64); andreast (powerpc, powerpc64)
Approved by:	re (bz)

1 files changed, 22 insertions, 22 deletions

diff --git a/sys/mips/mips/pmap.c b/sys/mips/mips/pmap.c
index e6dcd77..7da76be 100644
--- a/sys/mips/mips/pmap.c
+++ b/sys/mips/mips/pmap.c
@@ -1432,7 +1432,7 @@ retry:
 			KASSERT(!pte_test(&oldpte, PTE_W),
 			    ("wired pte for unwired page"));
 			if (m->md.pv_flags & PV_TABLE_REF)
-				vm_page_flag_set(m, PG_REFERENCED);
+				vm_page_aflag_set(m, PGA_REFERENCED);
 			if (pte_test(&oldpte, PTE_D))
 				vm_page_dirty(m);
 			pmap_invalidate_page(pmap, va);
@@ -1448,7 +1448,7 @@ retry:
 				free_pv_entry(pv);
 		}
 		if (TAILQ_EMPTY(&m->md.pv_list)) {
-			vm_page_flag_clear(m, PG_WRITEABLE);
+			vm_page_aflag_clear(m, PGA_WRITEABLE);
 			m->md.pv_flags &= ~(PV_TABLE_REF | PV_TABLE_MOD);
 		}
 	}
@@ -1527,7 +1527,7 @@ pmap_remove_entry(pmap_t pmap, vm_page_t m, vm_offset_t va)
 	mtx_assert(&vm_page_queue_mtx, MA_OWNED);
 	pmap_pvh_free(&m->md, pmap, va);
 	if (TAILQ_EMPTY(&m->md.pv_list))
-		vm_page_flag_clear(m, PG_WRITEABLE);
+		vm_page_aflag_clear(m, PGA_WRITEABLE);
 }
 
 /*
@@ -1589,7 +1589,7 @@ pmap_remove_pte(struct pmap *pmap, pt_entry_t *ptq, vm_offset_t va)
 			vm_page_dirty(m);
 		}
 		if (m->md.pv_flags & PV_TABLE_REF)
-			vm_page_flag_set(m, PG_REFERENCED);
+			vm_page_aflag_set(m, PGA_REFERENCED);
 		m->md.pv_flags &= ~(PV_TABLE_REF | PV_TABLE_MOD);
 
 		pmap_remove_entry(pmap, m, va);
@@ -1713,7 +1713,7 @@ pmap_remove_all(vm_page_t m)
 	vm_page_lock_queues();
 
 	if (m->md.pv_flags & PV_TABLE_REF)
-		vm_page_flag_set(m, PG_REFERENCED);
+		vm_page_aflag_set(m, PGA_REFERENCED);
 
 	while ((pv = TAILQ_FIRST(&m->md.pv_list)) != NULL) {
 		PMAP_LOCK(pv->pv_pmap);
@@ -1757,7 +1757,7 @@ pmap_remove_all(vm_page_t m)
 		free_pv_entry(pv);
 	}
 
-	vm_page_flag_clear(m, PG_WRITEABLE);
+	vm_page_aflag_clear(m, PGA_WRITEABLE);
 	m->md.pv_flags &= ~(PV_TABLE_REF | PV_TABLE_MOD);
 	vm_page_unlock_queues();
 }
@@ -2004,7 +2004,7 @@ validate:
 			*pte = newpte;
 			if (page_is_managed(opa) && (opa != pa)) {
 				if (om->md.pv_flags & PV_TABLE_REF)
-					vm_page_flag_set(om, PG_REFERENCED);
+					vm_page_aflag_set(om, PGA_REFERENCED);
 				om->md.pv_flags &=
 				    ~(PV_TABLE_REF | PV_TABLE_MOD);
 			}
@@ -2017,7 +2017,7 @@ validate:
 			}
 			if (page_is_managed(opa) &&
 			    TAILQ_EMPTY(&om->md.pv_list))
-				vm_page_flag_clear(om, PG_WRITEABLE);
+				vm_page_aflag_clear(om, PGA_WRITEABLE);
 		} else {
 			*pte = newpte;
 		}
@@ -2535,7 +2535,7 @@ pmap_remove_pages(pmap_t pmap)
 		m->md.pv_list_count--;
 		TAILQ_REMOVE(&m->md.pv_list, pv, pv_list);
 		if (TAILQ_FIRST(&m->md.pv_list) == NULL) {
-			vm_page_flag_clear(m, PG_WRITEABLE);
+			vm_page_aflag_clear(m, PGA_WRITEABLE);
 		}
 		pmap_unuse_pt(pv->pv_pmap, pv->pv_va, pv->pv_ptem);
 		free_pv_entry(pv);
@@ -2615,7 +2615,7 @@ pmap_changebit(vm_page_t m, int bit, boolean_t setem)
 		PMAP_UNLOCK(pv->pv_pmap);
 	}
 	if (!setem && bit == PTE_D)
-		vm_page_flag_clear(m, PG_WRITEABLE);
+		vm_page_aflag_clear(m, PGA_WRITEABLE);
 }
 
 /*
@@ -2662,13 +2662,13 @@ pmap_remove_write(vm_page_t m)
 	    ("pmap_remove_write: page %p is not managed", m));
 
 	/*
-	 * If the page is not VPO_BUSY, then PG_WRITEABLE cannot be set by
-	 * another thread while the object is locked.  Thus, if PG_WRITEABLE
+	 * If the page is not VPO_BUSY, then PGA_WRITEABLE cannot be set by
+	 * another thread while the object is locked.  Thus, if PGA_WRITEABLE
 	 * is clear, no page table entries need updating.
 	 */
 	VM_OBJECT_LOCK_ASSERT(m->object, MA_OWNED);
 	if ((m->oflags & VPO_BUSY) == 0 &&
-	    (m->flags & PG_WRITEABLE) == 0)
+	    (m->aflags & PGA_WRITEABLE) == 0)
 		return;
 
 	/*
@@ -2685,7 +2685,7 @@ pmap_remove_write(vm_page_t m)
 		pmap_protect(pv->pv_pmap, va, va + PAGE_SIZE,
 		    VM_PROT_READ | VM_PROT_EXECUTE);
 	}
-	vm_page_flag_clear(m, PG_WRITEABLE);
+	vm_page_aflag_clear(m, PGA_WRITEABLE);
 	vm_page_unlock_queues();
 }
 
@@ -2724,13 +2724,13 @@ pmap_is_modified(vm_page_t m)
 	    ("pmap_is_modified: page %p is not managed", m));
 
 	/*
-	 * If the page is not VPO_BUSY, then PG_WRITEABLE cannot be
-	 * concurrently set while the object is locked.  Thus, if PG_WRITEABLE
+	 * If the page is not VPO_BUSY, then PGA_WRITEABLE cannot be
+	 * concurrently set while the object is locked.  Thus, if PGA_WRITEABLE
 	 * is clear, no PTEs can have PTE_D set.
 	 */
 	VM_OBJECT_LOCK_ASSERT(m->object, MA_OWNED);
 	if ((m->oflags & VPO_BUSY) == 0 &&
-	    (m->flags & PG_WRITEABLE) == 0)
+	    (m->aflags & PGA_WRITEABLE) == 0)
 		return (FALSE);
 	vm_page_lock_queues();
 	if (m->md.pv_flags & PV_TABLE_MOD)
@@ -2781,11 +2781,11 @@ pmap_clear_modify(vm_page_t m)
 	    ("pmap_clear_modify: page %p is busy", m));
 
 	/*
-	 * If the page is not PG_WRITEABLE, then no PTEs can have PTE_D set.
+	 * If the page is not PGA_WRITEABLE, then no PTEs can have PTE_D set.
 	 * If the object containing the page is locked and the page is not
-	 * VPO_BUSY, then PG_WRITEABLE cannot be concurrently set.
+	 * VPO_BUSY, then PGA_WRITEABLE cannot be concurrently set.
 	 */
-	if ((m->flags & PG_WRITEABLE) == 0)
+	if ((m->aflags & PGA_WRITEABLE) == 0)
 		return;
 	vm_page_lock_queues();
 	if (m->md.pv_flags & PV_TABLE_MOD) {
@@ -2929,7 +2929,7 @@ retry:
 		 * determine if the address is MINCORE_REFERENCED.  
 		 */
 		m = PHYS_TO_VM_PAGE(pa);
-		if ((m->flags & PG_REFERENCED) != 0)
+		if ((m->aflags & PGA_REFERENCED) != 0)
 			val |= MINCORE_REFERENCED | MINCORE_REFERENCED_OTHER;
 	}
 	if ((val & (MINCORE_MODIFIED_OTHER | MINCORE_REFERENCED_OTHER)) !=
@@ -3185,7 +3185,7 @@ init_pte_prot(vm_offset_t va, vm_page_t m, vm_prot_t prot)
 			rw =  PTE_V | PTE_D | PTE_C_CACHE;
 		else
 			rw = PTE_V | PTE_C_CACHE;
-		vm_page_flag_set(m, PG_WRITEABLE);
+		vm_page_aflag_set(m, PGA_WRITEABLE);
 	} else
 		/* Needn't emulate a modified bit for unmanaged pages. */
 		rw =  PTE_V | PTE_D | PTE_C_CACHE;