Change the busy page mgmt, so that when pages are freed, they

MUST be PG_BUSY.  It is bogus to free a page that isn't busy,
because it is in a state of being "unavailable" when being
freed.  The additional advantage is that the page_remove code
has a better cross-check that the page should be busy and
unavailable for other use.  There were some minor problems
with the collapse code, and this plugs those subtile "holes."

Also, the vfs_bio code wasn't checking correctly for PG_BUSY
pages.  I am going to develop a more consistant scheme for
grabbing pages, busy or otherwise.  For now, we are stuck
with the current morass.

16 files changed, 253 insertions, 125 deletions

diff --git a/sys/amd64/amd64/pmap.c b/sys/amd64/amd64/pmap.c
index 0004625..22281dd 100644
--- a/sys/amd64/amd64/pmap.c
+++ b/sys/amd64/amd64/pmap.c
@@ -39,7 +39,7 @@
  * SUCH DAMAGE.
  *
  *	from:	@(#)pmap.c	7.7 (Berkeley)	5/12/91
- *	$Id: pmap.c,v 1.177 1998/01/17 09:16:18 dyson Exp $
+ *	$Id: pmap.c,v 1.178 1998/01/22 17:29:30 dyson Exp $
  */
 
 /*
@@ -855,11 +855,11 @@ void
 pmap_new_proc(p)
 	struct proc *p;
 {
-	int i;
+	int i, updateneeded;
 	vm_object_t upobj;
 	vm_page_t m;
 	struct user *up;
-	unsigned *ptek;
+	unsigned *ptek, oldpte;
 
 	/*
 	 * allocate object for the upages
@@ -880,6 +880,7 @@ pmap_new_proc(p)
 
 	ptek = (unsigned *) vtopte((vm_offset_t) up);
 
+	updateneeded = 0;
 	for(i=0;i<UPAGES;i++) {
 		/*
 		 * Get a kernel stack page
@@ -895,15 +896,25 @@ pmap_new_proc(p)
 		m->wire_count++;
 		cnt.v_wire_count++;
 
+		oldpte = *(ptek + i);
 		/*
 		 * Enter the page into the kernel address space.
 		 */
 		*(ptek + i) = VM_PAGE_TO_PHYS(m) | PG_RW | PG_V | pgeflag;
+		if (oldpte) {
+			if ((oldpte & PG_G) || (cpu_class > CPUCLASS_386)) {
+				invlpg((vm_offset_t) up + i * PAGE_SIZE);
+			} else {
+				updateneeded = 1;
+			}
+		}
 
 		m->flags &= ~(PG_ZERO|PG_BUSY);
 		m->flags |= PG_MAPPED|PG_WRITEABLE;
 		m->valid = VM_PAGE_BITS_ALL;
 	}
+	if (updateneeded)
+		invltlb();
 }
 
 /*
@@ -917,18 +928,20 @@ pmap_dispose_proc(p)
 	int i;
 	vm_object_t upobj;
 	vm_page_t m;
-	unsigned *ptek;
-
+	unsigned *ptek, oldpte;
 
 	upobj = p->p_upages_obj;
 
 	ptek = (unsigned *) vtopte((vm_offset_t) p->p_addr);
 	for(i=0;i<UPAGES;i++) {
+
 		if ((m = vm_page_lookup(upobj, i)) == NULL)
 			panic("pmap_dispose_proc: upage already missing???");
+		m->flags |= PG_BUSY;
 
+		oldpte = *(ptek + i);
 		*(ptek + i) = 0;
-		if (cpu_class >= CPUCLASS_586)
+		if ((oldpte & PG_G) || (cpu_class > CPUCLASS_386))
 			invlpg((vm_offset_t) p->p_addr + i * PAGE_SIZE);
 		vm_page_unwire(m);
 		vm_page_free(m);
@@ -1062,6 +1075,7 @@ _pmap_unwire_pte_hold(pmap_t pmap, vm_page_t m) {
 				wakeup(m);
 			}
 
+			m->flags |= PG_BUSY;
 			vm_page_free_zero(m);
 			--cnt.v_wire_count;
 		}
@@ -1237,6 +1251,7 @@ pmap_release_free_page(pmap, p)
 	if (pmap->pm_ptphint && (pmap->pm_ptphint->pindex == p->pindex))
 		pmap->pm_ptphint = NULL;
 
+	p->flags |= PG_BUSY;
 	vm_page_free_zero(p);
 	splx(s);
 	return 1;
@@ -2335,7 +2350,6 @@ retry:
 			m[0] = p;
 
 			if (vm_pager_get_pages(object, m, 1, 0) != VM_PAGER_OK) {
-				PAGE_WAKEUP(p);
 				vm_page_free(p);
 				return;
 			}
diff --git a/sys/i386/i386/pmap.c b/sys/i386/i386/pmap.c
index 0004625..22281dd 100644
--- a/sys/i386/i386/pmap.c
+++ b/sys/i386/i386/pmap.c
@@ -39,7 +39,7 @@
  * SUCH DAMAGE.
  *
  *	from:	@(#)pmap.c	7.7 (Berkeley)	5/12/91
- *	$Id: pmap.c,v 1.177 1998/01/17 09:16:18 dyson Exp $
+ *	$Id: pmap.c,v 1.178 1998/01/22 17:29:30 dyson Exp $
  */
 
 /*
@@ -855,11 +855,11 @@ void
 pmap_new_proc(p)
 	struct proc *p;
 {
-	int i;
+	int i, updateneeded;
 	vm_object_t upobj;
 	vm_page_t m;
 	struct user *up;
-	unsigned *ptek;
+	unsigned *ptek, oldpte;
 
 	/*
 	 * allocate object for the upages
@@ -880,6 +880,7 @@ pmap_new_proc(p)
 
 	ptek = (unsigned *) vtopte((vm_offset_t) up);
 
+	updateneeded = 0;
 	for(i=0;i<UPAGES;i++) {
 		/*
 		 * Get a kernel stack page
@@ -895,15 +896,25 @@ pmap_new_proc(p)
 		m->wire_count++;
 		cnt.v_wire_count++;
 
+		oldpte = *(ptek + i);
 		/*
 		 * Enter the page into the kernel address space.
 		 */
 		*(ptek + i) = VM_PAGE_TO_PHYS(m) | PG_RW | PG_V | pgeflag;
+		if (oldpte) {
+			if ((oldpte & PG_G) || (cpu_class > CPUCLASS_386)) {
+				invlpg((vm_offset_t) up + i * PAGE_SIZE);
+			} else {
+				updateneeded = 1;
+			}
+		}
 
 		m->flags &= ~(PG_ZERO|PG_BUSY);
 		m->flags |= PG_MAPPED|PG_WRITEABLE;
 		m->valid = VM_PAGE_BITS_ALL;
 	}
+	if (updateneeded)
+		invltlb();
 }
 
 /*
@@ -917,18 +928,20 @@ pmap_dispose_proc(p)
 	int i;
 	vm_object_t upobj;
 	vm_page_t m;
-	unsigned *ptek;
-
+	unsigned *ptek, oldpte;
 
 	upobj = p->p_upages_obj;
 
 	ptek = (unsigned *) vtopte((vm_offset_t) p->p_addr);
 	for(i=0;i<UPAGES;i++) {
+
 		if ((m = vm_page_lookup(upobj, i)) == NULL)
 			panic("pmap_dispose_proc: upage already missing???");
+		m->flags |= PG_BUSY;
 
+		oldpte = *(ptek + i);
 		*(ptek + i) = 0;
-		if (cpu_class >= CPUCLASS_586)
+		if ((oldpte & PG_G) || (cpu_class > CPUCLASS_386))
 			invlpg((vm_offset_t) p->p_addr + i * PAGE_SIZE);
 		vm_page_unwire(m);
 		vm_page_free(m);
@@ -1062,6 +1075,7 @@ _pmap_unwire_pte_hold(pmap_t pmap, vm_page_t m) {
 				wakeup(m);
 			}
 
+			m->flags |= PG_BUSY;
 			vm_page_free_zero(m);
 			--cnt.v_wire_count;
 		}
@@ -1237,6 +1251,7 @@ pmap_release_free_page(pmap, p)
 	if (pmap->pm_ptphint && (pmap->pm_ptphint->pindex == p->pindex))
 		pmap->pm_ptphint = NULL;
 
+	p->flags |= PG_BUSY;
 	vm_page_free_zero(p);
 	splx(s);
 	return 1;
@@ -2335,7 +2350,6 @@ retry:
 			m[0] = p;
 
 			if (vm_pager_get_pages(object, m, 1, 0) != VM_PAGER_OK) {
-				PAGE_WAKEUP(p);
 				vm_page_free(p);
 				return;
 			}
diff --git a/sys/kern/vfs_bio.c b/sys/kern/vfs_bio.c
index 39fcd54..953dfae 100644
--- a/sys/kern/vfs_bio.c
+++ b/sys/kern/vfs_bio.c
@@ -11,7 +11,7 @@
  * 2. Absolutely no warranty of function or purpose is made by the author
  *		John S. Dyson.
  *
- * $Id: vfs_bio.c,v 1.145 1998/01/24 02:01:20 dyson Exp $
+ * $Id: vfs_bio.c,v 1.146 1998/01/25 06:24:01 dyson Exp $
  */
 
 /*
@@ -796,8 +796,7 @@ vfs_vmio_release(bp)
 					else
 						vm_page_deactivate(m);
 				} else if (m->hold_count == 0) {
-					struct vnode *vp;
-					vp = bp->b_vp;
+					m->flags |= PG_BUSY;
 					vm_page_protect(m, VM_PROT_NONE);
 					vm_page_free(m);
 				}
@@ -2145,7 +2144,7 @@ vfs_page_set_valid(struct buf *bp, vm_ooffset_t off, int pageno, vm_page_t m)
 void
 vfs_busy_pages(struct buf * bp, int clear_modify)
 {
-	int i;
+	int i,s;
 
 	if (bp->b_flags & B_VMIO) {
 		struct vnode *vp = bp->b_vp;
@@ -2156,7 +2155,24 @@ vfs_busy_pages(struct buf * bp, int clear_modify)
 			foff = (vm_ooffset_t) DEV_BSIZE * bp->b_lblkno;
 		else
 			foff = (vm_ooffset_t) vp->v_mount->mnt_stat.f_iosize * bp->b_lblkno;
+
 		vfs_setdirty(bp);
+
+retry:
+		for (i = 0; i < bp->b_npages; i++) {
+			vm_page_t m = bp->b_pages[i];
+
+			if (m && (m->flags & PG_BUSY)) {
+				s = splvm();
+				while (m->flags & PG_BUSY) {
+					m->flags |= PG_WANTED;
+					tsleep(m, PVM, "vbpage", 0);
+				}
+				splx(s);
+				goto retry;
+			}
+		}
+
 		for (i = 0; i < bp->b_npages; i++, foff += PAGE_SIZE) {
 			vm_page_t m = bp->b_pages[i];
 
@@ -2164,6 +2180,7 @@ vfs_busy_pages(struct buf * bp, int clear_modify)
 				obj->paging_in_progress++;
 				m->busy++;
 			}
+
 			vm_page_protect(m, VM_PROT_NONE);
 			if (clear_modify)
 				vfs_page_set_valid(bp, foff, i, m);
@@ -2299,6 +2316,7 @@ vm_hold_free_pages(struct buf * bp, vm_offset_t from, vm_offset_t to)
 #endif
 			bp->b_pages[index] = NULL;
 			pmap_kremove(pg);
+			p->flags |= PG_BUSY;
 			vm_page_unwire(p);
 			vm_page_free(p);
 		}
diff --git a/sys/kern/vfs_cluster.c b/sys/kern/vfs_cluster.c
index 70f1765..880ca1f 100644
--- a/sys/kern/vfs_cluster.c
+++ b/sys/kern/vfs_cluster.c
@@ -33,7 +33,7 @@
  * SUCH DAMAGE.
  *
  *	@(#)vfs_cluster.c	8.7 (Berkeley) 2/13/94
- * $Id: vfs_cluster.c,v 1.51 1998/01/24 02:01:21 dyson Exp $
+ * $Id: vfs_cluster.c,v 1.52 1998/01/31 07:23:11 eivind Exp $
  */
 
 #include "opt_debug_cluster.h"
@@ -720,9 +720,19 @@ cluster_wbuild(vp, size, start_lbn, len)
 				tbp->b_flags &= ~B_DONE;
 				splx(s);
 			}
+
 			if (tbp->b_flags & B_VMIO) {
+				vm_page_t m;
+
+				if (i != 0) {
+					for (j = 0; j < tbp->b_npages; j += 1) {
+						m = tbp->b_pages[j];
+						if (m->flags & PG_BUSY)
+							goto finishcluster;
+					}
+				}
+					
 				for (j = 0; j < tbp->b_npages; j += 1) {
-					vm_page_t m;
 					m = tbp->b_pages[j];
 					++m->busy;
 					++m->object->paging_in_progress;
@@ -746,6 +756,7 @@ cluster_wbuild(vp, size, start_lbn, len)
 			TAILQ_INSERT_TAIL(&bp->b_cluster.cluster_head,
 				tbp, b_cluster.cluster_entry);
 		}
+	finishcluster:
 		pmap_qenter(trunc_page((vm_offset_t) bp->b_data),
 			(vm_page_t *) bp->b_pages, bp->b_npages);
 		if (bp->b_bufsize > bp->b_kvasize)
diff --git a/sys/nfs/nfs_serv.c b/sys/nfs/nfs_serv.c
index 029897b..4bcca7d 100644
--- a/sys/nfs/nfs_serv.c
+++ b/sys/nfs/nfs_serv.c
@@ -34,7 +34,7 @@
  * SUCH DAMAGE.
  *
  *	@(#)nfs_serv.c	8.3 (Berkeley) 1/12/94
- * $Id: nfs_serv.c,v 1.53 1997/12/27 02:56:34 bde Exp $
+ * $Id: nfs_serv.c,v 1.54 1997/12/29 00:23:34 dyson Exp $
  */
 
 /*
@@ -76,6 +76,7 @@
 #include <vm/vm.h>
 #include <vm/vm_extern.h>
 #include <vm/vm_zone.h>
+#include <vm/vm_object.h>
 
 #include <nfs/nfsproto.h>
 #include <nfs/rpcv2.h>
@@ -3123,6 +3124,10 @@ nfsrv_commit(nfsd, slp, procp, mrq)
 		return (0);
 	}
 	for_ret = VOP_GETATTR(vp, &bfor, cred, procp);
+	if (vp->v_object &&
+	   (vp->v_object->flags & OBJ_MIGHTBEDIRTY)) {
+			vm_object_page_clean(vp->v_object, 0, 0, TRUE);
+	}
 	error = VOP_FSYNC(vp, cred, MNT_WAIT, procp);
 	aft_ret = VOP_GETATTR(vp, &aft, cred, procp);
 	vput(vp);
diff --git a/sys/nfsserver/nfs_serv.c b/sys/nfsserver/nfs_serv.c
index 029897b..4bcca7d 100644
--- a/sys/nfsserver/nfs_serv.c
+++ b/sys/nfsserver/nfs_serv.c
@@ -34,7 +34,7 @@
  * SUCH DAMAGE.
  *
  *	@(#)nfs_serv.c	8.3 (Berkeley) 1/12/94
- * $Id: nfs_serv.c,v 1.53 1997/12/27 02:56:34 bde Exp $
+ * $Id: nfs_serv.c,v 1.54 1997/12/29 00:23:34 dyson Exp $
  */
 
 /*
@@ -76,6 +76,7 @@
 #include <vm/vm.h>
 #include <vm/vm_extern.h>
 #include <vm/vm_zone.h>
+#include <vm/vm_object.h>
 
 #include <nfs/nfsproto.h>
 #include <nfs/rpcv2.h>
@@ -3123,6 +3124,10 @@ nfsrv_commit(nfsd, slp, procp, mrq)
 		return (0);
 	}
 	for_ret = VOP_GETATTR(vp, &bfor, cred, procp);
+	if (vp->v_object &&
+	   (vp->v_object->flags & OBJ_MIGHTBEDIRTY)) {
+			vm_object_page_clean(vp->v_object, 0, 0, TRUE);
+	}
 	error = VOP_FSYNC(vp, cred, MNT_WAIT, procp);
 	aft_ret = VOP_GETATTR(vp, &aft, cred, procp);
 	vput(vp);
diff --git a/sys/vm/device_pager.c b/sys/vm/device_pager.c
index 23f1440..e274095 100644
--- a/sys/vm/device_pager.c
+++ b/sys/vm/device_pager.c
@@ -36,7 +36,7 @@
  * SUCH DAMAGE.
  *
  *	@(#)device_pager.c	8.1 (Berkeley) 6/11/93
- * $Id: device_pager.c,v 1.26 1997/08/25 22:15:11 bde Exp $
+ * $Id: device_pager.c,v 1.27 1997/09/01 03:17:12 bde Exp $
  */
 
 #include <sys/param.h>
@@ -214,7 +214,6 @@ dev_pager_getpages(object, m, count, reqpage)
 	page = dev_pager_getfake(paddr);
 	TAILQ_INSERT_TAIL(&object->un_pager.devp.devp_pglist, page, pageq);
 	for (i = 0; i < count; i++) {
-		PAGE_WAKEUP(m[i]);
 		vm_page_free(m[i]);
 	}
 	s = splhigh();
diff --git a/sys/vm/swap_pager.c b/sys/vm/swap_pager.c
index f5fd447..f62819b 100644
--- a/sys/vm/swap_pager.c
+++ b/sys/vm/swap_pager.c
@@ -39,7 +39,7 @@
  * from: Utah $Hdr: swap_pager.c 1.4 91/04/30$
  *
  *	@(#)swap_pager.c	8.9 (Berkeley) 3/21/94
- * $Id: swap_pager.c,v 1.81 1998/01/17 09:16:47 dyson Exp $
+ * $Id: swap_pager.c,v 1.82 1998/01/22 17:30:32 dyson Exp $
  */
 
 /*
@@ -849,7 +849,6 @@ static void
 swap_pager_freepage(m)
 	vm_page_t m;
 {
-	PAGE_WAKEUP(m);
 	vm_page_free(m);
 }
 
@@ -1489,8 +1488,7 @@ retryfree:
 				 * optimization, if a page has been read
 				 * during the pageout process, we activate it.
 				 */
-				if ((m[i]->queue != PQ_ACTIVE) &&
-				    ((m[i]->flags & (PG_WANTED|PG_REFERENCED)) ||
+				if (((m[i]->flags & (PG_WANTED|PG_REFERENCED)) ||
 				    pmap_ts_referenced(VM_PAGE_TO_PHYS(m[i])))) {
 					vm_page_activate(m[i]);
 				}
@@ -1597,8 +1595,7 @@ swap_pager_finish(spc)
 		for (i = 0; i < spc->spc_count; i++) {
 			pmap_clear_modify(VM_PAGE_TO_PHYS(spc->spc_m[i]));
 			spc->spc_m[i]->dirty = 0;
-			if ((spc->spc_m[i]->queue != PQ_ACTIVE) &&
-			    ((spc->spc_m[i]->flags & PG_WANTED) || pmap_ts_referenced(VM_PAGE_TO_PHYS(spc->spc_m[i]))))
+			if (((spc->spc_m[i]->flags & PG_WANTED) || pmap_ts_referenced(VM_PAGE_TO_PHYS(spc->spc_m[i]))))
 				vm_page_activate(spc->spc_m[i]);
 		}
 	}
diff --git a/sys/vm/vm_fault.c b/sys/vm/vm_fault.c
index 62cda5a..d00e232 100644
--- a/sys/vm/vm_fault.c
+++ b/sys/vm/vm_fault.c
@@ -66,7 +66,7 @@
  * any improvements or extensions that they make and grant Carnegie the
  * rights to redistribute these changes.
  *
- * $Id: vm_fault.c,v 1.75 1998/01/17 09:16:49 dyson Exp $
+ * $Id: vm_fault.c,v 1.76 1998/01/22 17:30:33 dyson Exp $
  */
 
 /*
@@ -146,16 +146,12 @@ vm_fault(vm_map_t map, vm_offset_t vaddr, vm_prot_t fault_type, int fault_flags)
  *	Recovery actions
  */
 #define	FREE_PAGE(m)	{				\
-	PAGE_WAKEUP(m);					\
 	vm_page_free(m);				\
 }
 
 #define	RELEASE_PAGE(m)	{				\
 	PAGE_WAKEUP(m);					\
-	if (m->queue != PQ_ACTIVE) { \
-		vm_page_activate(m);		\
-		m->act_count = 0; \
-	} \
+	vm_page_activate(m);		\
 }
 
 #define	UNLOCK_MAP	{				\
@@ -613,8 +609,7 @@ readrest:
 				 * get rid of the unnecessary page
 				 */
 				vm_page_protect(first_m, VM_PROT_NONE);
-				PAGE_WAKEUP(first_m);
-				vm_page_free(first_m);
+				FREE_PAGE(first_m);
 				/*
 				 * grab the page and put it into the process'es object
 				 */
@@ -630,10 +625,7 @@ readrest:
 			}
 
 			if (m) {
-				if (m->queue != PQ_ACTIVE) {
-					vm_page_activate(m);
-					m->act_count = 0;
-				}
+				vm_page_activate(m);
 
 			/*
 			 * We no longer need the old page or object.
@@ -757,8 +749,7 @@ readrest:
 		else
 			vm_page_unwire(m);
 	} else {
-		if (m->queue != PQ_ACTIVE)
-			vm_page_activate(m);
+		vm_page_activate(m);
 	}
 
 	if (curproc && (curproc->p_flag & P_INMEM) && curproc->p_stats) {
diff --git a/sys/vm/vm_kern.c b/sys/vm/vm_kern.c
index a20cda4..0077ca2 100644
--- a/sys/vm/vm_kern.c
+++ b/sys/vm/vm_kern.c
@@ -61,7 +61,7 @@
  * any improvements or extensions that they make and grant Carnegie the
  * rights to redistribute these changes.
  *
- * $Id: vm_kern.c,v 1.39 1997/08/05 00:01:52 dyson Exp $
+ * $Id: vm_kern.c,v 1.40 1998/01/22 17:30:35 dyson Exp $
  */
 
 /*
@@ -328,7 +328,6 @@ retry:
 				i -= PAGE_SIZE;
 				m = vm_page_lookup(kmem_object,
 					OFF_TO_IDX(offset + i));
-				PAGE_WAKEUP(m);
 				vm_page_free(m);
 			}
 			vm_map_delete(map, addr, addr + size);
diff --git a/sys/vm/vm_map.c b/sys/vm/vm_map.c
index 7afc84c..5fd8d30 100644
--- a/sys/vm/vm_map.c
+++ b/sys/vm/vm_map.c
@@ -61,7 +61,7 @@
  * any improvements or extensions that they make and grant Carnegie the
  * rights to redistribute these changes.
  *
- * $Id: vm_map.c,v 1.107 1998/01/21 12:18:00 dyson Exp $
+ * $Id: vm_map.c,v 1.108 1998/01/22 17:30:37 dyson Exp $
  */
 
 /*
@@ -2678,7 +2678,8 @@ m_outretry:
 				while (m_out && (m_out->flags & PG_BUSY)) {
 					m_out->flags |= PG_WANTED;
 					tsleep(m_out, PVM, "pwtfrz", 0);
-					m_out = vm_page_lookup(robject, idx);
+					splx(s);
+					goto m_outretry;
 				}
 				splx(s);
 			}
@@ -2708,7 +2709,8 @@ m_inretry:
 					while (m_in && (m_in->busy || (m_in->flags & PG_BUSY))) {
 						m_in->flags |= PG_WANTED;
 						tsleep(m_in, PVM, "pwtfrz", 0);
-						m_in = vm_page_lookup(object, bo_pindex + idx);
+						splx(s);
+						goto m_inretry;
 					}
 					splx(s);
 					if (m_in == NULL) {
diff --git a/sys/vm/vm_object.c b/sys/vm/vm_object.c
index a833fab..cd7dc0f 100644
--- a/sys/vm/vm_object.c
+++ b/sys/vm/vm_object.c
@@ -61,7 +61,7 @@
  * any improvements or extensions that they make and grant Carnegie the
  * rights to redistribute these changes.
  *
- * $Id: vm_object.c,v 1.107 1998/01/17 09:16:55 dyson Exp $
+ * $Id: vm_object.c,v 1.108 1998/01/22 17:30:39 dyson Exp $
  */
 
 /*
@@ -155,6 +155,8 @@ _vm_object_allocate(type, size, object)
 	object->behavior = OBJ_NORMAL;
 	object->paging_in_progress = 0;
 	object->resident_page_count = 0;
+	object->cache_count = 0;
+	object->wire_count = 0;
 	object->shadow_count = 0;
 	object->pg_color = next_index;
 	if ( size > (PQ_L2_SIZE / 3 + PQ_PRIME1))
@@ -376,6 +378,7 @@ doterm:
 			temp->shadow_count--;
 			if (temp->ref_count == 0)
 				temp->flags &= ~OBJ_OPT;
+			temp->generation++;
 		}
 		vm_object_terminate(object);
 		/* unlocks and deallocates object */
@@ -445,7 +448,7 @@ vm_object_terminate(object)
 		while ((p = TAILQ_FIRST(&object->memq)) != NULL) {
 			if (p->busy || (p->flags & PG_BUSY))
 				printf("vm_object_terminate: freeing busy page\n");
-			PAGE_WAKEUP(p);
+			p->flags |= PG_BUSY;
 			vm_page_free(p);
 			cnt.v_pfree++;
 		}
@@ -529,11 +532,15 @@ vm_object_page_clean(object, start, end, syncio)
 	} else {
 		tend = end;
 	}
+
+	for(p = TAILQ_FIRST(&object->memq); p; p = TAILQ_NEXT(p, listq)) {
+		p->flags |= PG_CLEANCHK;
+		vm_page_protect(p, VM_PROT_READ);
+	}
+
 	if ((tstart == 0) && (tend == object->size)) {
 		object->flags &= ~(OBJ_WRITEABLE|OBJ_MIGHTBEDIRTY);
 	}
-	for(p = TAILQ_FIRST(&object->memq); p; p = TAILQ_NEXT(p, listq))
-		p->flags |= PG_CLEANCHK;
 
 rescan:
 	curgeneration = object->generation;
@@ -565,9 +572,7 @@ rescan:
 				goto rescan;
 			}
 		}
-		splx(s);
-			
-		s = splvm();
+
 		maxf = 0;
 		for(i=1;i<vm_pageout_page_count;i++) {
 			if (tp = vm_page_lookup(object, pi + i)) {
@@ -812,8 +817,7 @@ shadowlookup:
 		}
 
 		if (advise == MADV_WILLNEED) {
-			if (m->queue != PQ_ACTIVE)
-				vm_page_activate(m);
+			vm_page_activate(m);
 		} else if (advise == MADV_DONTNEED) {
 			vm_page_deactivate(m);
 		} else if (advise == MADV_FREE) {
@@ -868,7 +872,8 @@ vm_object_shadow(object, offset, length)
 	result->backing_object = source;
 	if (source) {
 		TAILQ_INSERT_TAIL(&source->shadow_head, result, shadow_list);
-		++source->shadow_count;
+		source->shadow_count++;
+		source->generation++;
 	}
 
 	/*
@@ -924,6 +929,8 @@ vm_object_qcollapse(object)
 			p = next;
 			continue;
 		}
+		p->flags |= PG_BUSY;
+
 		new_pindex = p->pindex - backing_offset_index;
 		if (p->pindex < backing_offset_index ||
 		    new_pindex >= size) {
@@ -935,7 +942,8 @@ vm_object_qcollapse(object)
 			vm_page_free(p);
 		} else {
 			pp = vm_page_lookup(object, new_pindex);
-			if (pp != NULL || (object->type == OBJT_SWAP && vm_pager_has_page(object,
+			if (pp != NULL ||
+				(object->type == OBJT_SWAP && vm_pager_has_page(object,
 				    paging_offset_index + new_pindex, NULL, NULL))) {
 				if (backing_object->type == OBJT_SWAP)
 					swap_pager_freespace(backing_object,
@@ -946,6 +954,7 @@ vm_object_qcollapse(object)
 				if (backing_object->type == OBJT_SWAP)
 					swap_pager_freespace(backing_object,
 					    backing_object_paging_offset_index + p->pindex, 1);
+
 				vm_page_rename(p, object, new_pindex);
 				vm_page_protect(p, VM_PROT_NONE);
 				p->dirty = VM_PAGE_BITS_ALL;
@@ -1041,6 +1050,7 @@ vm_object_collapse(object)
 			while ((p = TAILQ_FIRST(&backing_object->memq)) != 0) {
 
 				new_pindex = p->pindex - backing_offset_index;
+				p->flags |= PG_BUSY;
 
 				/*
 				 * If the parent has a page here, or if this
@@ -1053,14 +1063,12 @@ vm_object_collapse(object)
 				if (p->pindex < backing_offset_index ||
 				    new_pindex >= size) {
 					vm_page_protect(p, VM_PROT_NONE);
-					PAGE_WAKEUP(p);
 					vm_page_free(p);
 				} else {
 					pp = vm_page_lookup(object, new_pindex);
 					if (pp != NULL || (object->type == OBJT_SWAP && vm_pager_has_page(object,
 					    OFF_TO_IDX(object->paging_offset) + new_pindex, NULL, NULL))) {
 						vm_page_protect(p, VM_PROT_NONE);
-						PAGE_WAKEUP(p);
 						vm_page_free(p);
 					} else {
 						vm_page_protect(p, VM_PROT_NONE);
@@ -1133,17 +1141,20 @@ vm_object_collapse(object)
 
 			TAILQ_REMOVE(&object->backing_object->shadow_head, object,
 			    shadow_list);
-			--object->backing_object->shadow_count;
+			object->backing_object->shadow_count--;
+			object->backing_object->generation++;
 			if (backing_object->backing_object) {
 				TAILQ_REMOVE(&backing_object->backing_object->shadow_head,
 				    backing_object, shadow_list);
-				--backing_object->backing_object->shadow_count;
+				backing_object->backing_object->shadow_count--;
+				backing_object->backing_object->generation++;
 			}
 			object->backing_object = backing_object->backing_object;
 			if (object->backing_object) {
 				TAILQ_INSERT_TAIL(&object->backing_object->shadow_head,
 				    object, shadow_list);
-				++object->backing_object->shadow_count;
+				object->backing_object->shadow_count++;
+				object->backing_object->generation++;
 			}
 
 			object->backing_object_offset += backing_object->backing_object_offset;
@@ -1182,7 +1193,11 @@ vm_object_collapse(object)
 			 * here.
 			 */
 
-			for (p = TAILQ_FIRST(&backing_object->memq); p; p = TAILQ_NEXT(p, listq)) {
+			for (p = TAILQ_FIRST(&backing_object->memq); p;
+					p = TAILQ_NEXT(p, listq)) {
+
+				p->flags |= PG_BUSY;
+
 				new_pindex = p->pindex - backing_offset_index;
 
 				/*
@@ -1198,15 +1213,25 @@ vm_object_collapse(object)
 
 					pp = vm_page_lookup(object, new_pindex);
 
-					if ((pp == NULL || pp->valid == 0) &&
+					if ((pp == NULL) || (pp->flags & PG_BUSY) || pp->busy) {
+						PAGE_WAKEUP(p);
+						return;
+					}
+
+					pp->flags |= PG_BUSY;
+					if ((pp->valid == 0) &&
 				   	    !vm_pager_has_page(object, OFF_TO_IDX(object->paging_offset) + new_pindex, NULL, NULL)) {
 						/*
 						 * Page still needed. Can't go any
 						 * further.
 						 */
+						PAGE_WAKEUP(pp);
+						PAGE_WAKEUP(p);
 						return;
 					}
+					PAGE_WAKEUP(pp);
 				}
+				PAGE_WAKEUP(p);
 			}
 
 			/*
@@ -1217,14 +1242,16 @@ vm_object_collapse(object)
 
 			TAILQ_REMOVE(&backing_object->shadow_head,
 			    object, shadow_list);
-			--backing_object->shadow_count;
+			backing_object->shadow_count--;
+			backing_object->generation++;
 
 			new_backing_object = backing_object->backing_object;
 			if (object->backing_object = new_backing_object) {
 				vm_object_reference(new_backing_object);
 				TAILQ_INSERT_TAIL(&new_backing_object->shadow_head,
 				    object, shadow_list);
-				++new_backing_object->shadow_count;
+				new_backing_object->shadow_count++;
+				new_backing_object->generation++;
 				object->backing_object_offset +=
 					backing_object->backing_object_offset;
 			}
@@ -1232,12 +1259,11 @@ vm_object_collapse(object)
 			/*
 			 * Drop the reference count on backing_object. Since
 			 * its ref_count was at least 2, it will not vanish;
-			 * so we don't need to call vm_object_deallocate.
+			 * so we don't need to call vm_object_deallocate, but
+			 * we do anyway.
 			 */
 			vm_object_deallocate(backing_object);
-
 			object_bypasses++;
-
 		}
 
 		/*
@@ -1303,14 +1329,16 @@ again:
 					if (p->valid & p->dirty)
 						continue;
 				}
+
+				p->flags |= PG_BUSY;
 				vm_page_protect(p, VM_PROT_NONE);
-				PAGE_WAKEUP(p);
 				vm_page_free(p);
 			}
 		}
 	} else {
 		while (size > 0) {
 			if ((p = vm_page_lookup(object, start)) != 0) {
+
 				if (p->wire_count != 0) {
 					p->valid = 0;
 					vm_page_protect(p, VM_PROT_NONE);
@@ -1318,6 +1346,7 @@ again:
 					size -= 1;
 					continue;
 				}
+
 				/*
 				 * The busy flags are only cleared at
 				 * interrupt -- minimize the spl transitions
@@ -1332,6 +1361,7 @@ again:
 					}
 					splx(s);
 				}
+
 				if (clean_only) {
 					vm_page_test_dirty(p);
 					if (p->valid & p->dirty) {
@@ -1340,8 +1370,9 @@ again:
 						continue;
 					}
 				}
+
+				p->flags |= PG_BUSY;
 				vm_page_protect(p, VM_PROT_NONE);
-				PAGE_WAKEUP(p);
 				vm_page_free(p);
 			}
 			start += 1;
diff --git a/sys/vm/vm_object.h b/sys/vm/vm_object.h
index 36bdb87..e67460e 100644
--- a/sys/vm/vm_object.h
+++ b/sys/vm/vm_object.h
@@ -61,7 +61,7 @@
  * any improvements or extensions that they make and grant Carnegie the
  * rights to redistribute these changes.
  *
- * $Id: vm_object.h,v 1.42 1998/01/17 09:16:56 dyson Exp $
+ * $Id: vm_object.h,v 1.43 1998/01/22 17:30:40 dyson Exp $
  */
 
 /*
@@ -97,6 +97,8 @@ struct vm_object {
 	u_short paging_in_progress;	/* Paging (in or out) so don't collapse or destroy */
 	u_short	behavior;		/* see below */
 	int resident_page_count;	/* number of resident pages */
+	int cache_count;			/* number of cached pages */
+	int	wire_count;			/* number of wired pages */
 	vm_ooffset_t paging_offset;	/* Offset into paging space */
 	struct vm_object *backing_object; /* object that I'm a shadow of */
 	vm_ooffset_t backing_object_offset;/* Offset in backing object */
diff --git a/sys/vm/vm_page.c b/sys/vm/vm_page.c
index 2072bf3..6837ecf 100644
--- a/sys/vm/vm_page.c
+++ b/sys/vm/vm_page.c
@@ -34,7 +34,7 @@
  * SUCH DAMAGE.
  *
  *	from: @(#)vm_page.c	7.4 (Berkeley) 5/7/91
- *	$Id: vm_page.c,v 1.86 1998/01/17 09:16:59 dyson Exp $
+ *	$Id: vm_page.c,v 1.87 1998/01/22 17:30:41 dyson Exp $
  */
 
 /*
@@ -416,6 +416,12 @@ vm_page_insert(m, object, pindex)
 	m->object->page_hint = m;
 	m->object->generation++;
 
+	if (m->wire_count)
+		object->wire_count++;
+
+	if ((m->queue - m->pc) == PQ_CACHE)
+		object->cache_count++;
+
 	/*
 	 * And show that the object has one more resident page.
 	 */
@@ -442,9 +448,25 @@ vm_page_remove(m)
 	if (!(m->flags & PG_TABLED))
 		return;
 
+	if ((m->flags & PG_BUSY) == 0) {
+		panic("vm_page_remove: page not busy");
+	}
+	
+	m->flags &= ~PG_BUSY;
+	if (m->flags & PG_WANTED) {
+		m->flags &= ~PG_WANTED;
+		wakeup(m);
+	}
+
 	if (m->object->page_hint == m)
 		m->object->page_hint = NULL;
 
+	if (m->wire_count)
+		m->object->wire_count--;
+
+	if ((m->queue - m->pc) == PQ_CACHE) 
+		m->object->cache_count--;
+
 	/*
 	 * Remove from the object_object/offset hash table
 	 */
@@ -542,8 +564,11 @@ vm_page_unqueue_nowakeup(m)
 		pq = &vm_page_queues[queue];
 		m->queue = PQ_NONE;
 		TAILQ_REMOVE(pq->pl, m, pageq);
-		--(*pq->cnt);
-		--(*pq->lcnt);
+		(*pq->cnt)--;
+		(*pq->lcnt)--;
+		if ((queue - m->pc) == PQ_CACHE) {
+			m->object->cache_count--;
+		}
 	}
 }
 
@@ -560,12 +585,13 @@ vm_page_unqueue(m)
 		m->queue = PQ_NONE;
 		pq = &vm_page_queues[queue];
 		TAILQ_REMOVE(pq->pl, m, pageq);
-		--(*pq->cnt);
-		--(*pq->lcnt);
+		(*pq->cnt)--;
+		(*pq->lcnt)--;
 		if ((queue - m->pc) == PQ_CACHE) {
 			if ((cnt.v_cache_count + cnt.v_free_count) <
 				(cnt.v_free_reserved + cnt.v_cache_min))
 				pagedaemon_wakeup();
+			m->object->cache_count--;
 		}
 	}
 }
@@ -868,16 +894,17 @@ vm_page_alloc(object, pindex, page_req)
 	queue = m->queue;
 	qtype = queue - m->pc;
 	if (qtype == PQ_ZERO)
-		--vm_page_zero_count;
+		vm_page_zero_count--;
 	pq = &vm_page_queues[queue];
 	TAILQ_REMOVE(pq->pl, m, pageq);
-	--(*pq->cnt);
-	--(*pq->lcnt);
+	(*pq->cnt)--;
+	(*pq->lcnt)--;
 	oldobject = NULL;
 	if (qtype == PQ_ZERO) {
 		m->flags = PG_ZERO|PG_BUSY;
 	} else if (qtype == PQ_CACHE) {
 		oldobject = m->object;
+		m->flags |= PG_BUSY;
 		vm_page_remove(m);
 		m->flags = PG_BUSY;
 	} else {
@@ -953,24 +980,34 @@ vm_page_activate(m)
 	register vm_page_t m;
 {
 	int s;
+	vm_page_t np;
+	vm_object_t object;
 
 	s = splvm();
-	if (m->queue == PQ_ACTIVE)
-		panic("vm_page_activate: already active");
-
-	if ((m->queue - m->pc) == PQ_CACHE)
-		cnt.v_reactivated++;
+	if (m->queue != PQ_ACTIVE) {
+		if ((m->queue - m->pc) == PQ_CACHE)
+			cnt.v_reactivated++;
 
-	vm_page_unqueue(m);
+		vm_page_unqueue(m);
 
-	if (m->wire_count == 0) {
-		m->queue = PQ_ACTIVE;
-		++(*vm_page_queues[PQ_ACTIVE].lcnt);
-		TAILQ_INSERT_TAIL(&vm_page_queue_active, m, pageq);
+		if (m->wire_count == 0) {
+			m->queue = PQ_ACTIVE;
+			++(*vm_page_queues[PQ_ACTIVE].lcnt);
+			TAILQ_INSERT_TAIL(&vm_page_queue_active, m, pageq);
+			if (m->act_count < ACT_INIT)
+				m->act_count = ACT_INIT;
+			cnt.v_active_count++;
+		}
+	} else {
 		if (m->act_count < ACT_INIT)
 			m->act_count = ACT_INIT;
-		cnt.v_active_count++;
 	}
+
+	object = m->object;
+	TAILQ_REMOVE(&object->memq, m, listq);
+	TAILQ_INSERT_TAIL(&object->memq, m, listq);
+	object->generation++;
+
 	splx(s);
 }
 
@@ -986,9 +1023,7 @@ vm_page_freechk_and_unqueue(m)
 	oldobject = m->object;
 
 #if !defined(MAX_PERF)
-	if (m->busy ||
-		(m->flags & PG_BUSY) ||
-		((m->queue - m->pc) == PQ_FREE) ||
+	if (m->busy || ((m->queue - m->pc) == PQ_FREE) ||
 		(m->hold_count != 0)) {
 		printf("vm_page_free: pindex(%ld), busy(%d), PG_BUSY(%d), hold(%d)\n",
 			m->pindex, m->busy,
@@ -1000,17 +1035,20 @@ vm_page_freechk_and_unqueue(m)
 	}
 #endif
 
-	vm_page_remove(m);
 	vm_page_unqueue_nowakeup(m);
+	vm_page_remove(m);
+
 	if ((m->flags & PG_FICTITIOUS) != 0) {
 		return 0;
 	}
+
 	if (m->wire_count != 0) {
 		if (m->wire_count > 1) {
 			panic("vm_page_free: invalid wire count (%d), pindex: 0x%x",
 				m->wire_count, m->pindex);
 		}
 		m->wire_count = 0;
+		m->object->wire_count--;
 		cnt.v_wire_count--;
 	}
 
@@ -1147,8 +1185,9 @@ vm_page_wire(m)
 		vm_page_unqueue(m);
 		splx(s);
 		cnt.v_wire_count++;
+		m->object->wire_count++;
 	}
-	++(*vm_page_queues[PQ_NONE].lcnt);
+	(*vm_page_queues[PQ_NONE].lcnt)++;
 	m->wire_count++;
 	m->flags |= PG_MAPPED;
 }
@@ -1169,15 +1208,18 @@ vm_page_unwire(m)
 
 	s = splvm();
 
-	if (m->wire_count > 0)
+	if (m->wire_count > 0) {
 		m->wire_count--;
-		
-	if (m->wire_count == 0) {
-		cnt.v_wire_count--;
-		TAILQ_INSERT_TAIL(&vm_page_queue_active, m, pageq);
-		m->queue = PQ_ACTIVE;
-		++(*vm_page_queues[PQ_ACTIVE].lcnt);
-		cnt.v_active_count++;
+		if (m->wire_count == 0) {
+			m->object->wire_count--;
+			cnt.v_wire_count--;
+			TAILQ_INSERT_TAIL(&vm_page_queue_active, m, pageq);
+			m->queue = PQ_ACTIVE;
+			(*vm_page_queues[PQ_ACTIVE].lcnt)++;
+			cnt.v_active_count++;
+		}
+	} else {
+		panic("vm_page_unwire: invalid wire count: %d\n", m->wire_count);
 	}
 	splx(s);
 }
@@ -1246,9 +1288,10 @@ vm_page_cache(m)
 	s = splvm();
 	vm_page_unqueue_nowakeup(m);
 	m->queue = PQ_CACHE + m->pc;
-	++(*vm_page_queues[m->queue].lcnt);
+	(*vm_page_queues[m->queue].lcnt)++;
 	TAILQ_INSERT_TAIL(vm_page_queues[m->queue].pl, m, pageq);
 	cnt.v_cache_count++;
+	m->object->cache_count++;
 	vm_page_free_wakeup();
 	splx(s);
 }
@@ -1415,9 +1458,7 @@ again1:
 						goto again1;
 					}
 				}
-				if ((m->dirty == 0) &&
-					(m->busy == 0) &&
-					(m->hold_count == 0))
+				if ((m->dirty == 0) && (m->busy == 0) && (m->hold_count == 0))
 					vm_page_cache(m);
 			}
 
@@ -1449,9 +1490,7 @@ again1:
 						goto again1;
 					}
 				}
-				if ((m->dirty == 0) &&
-					(m->busy == 0) &&
-					(m->hold_count == 0))
+				if ((m->dirty == 0) && (m->busy == 0) && (m->hold_count == 0))
 					vm_page_cache(m);
 			}
 
@@ -1479,11 +1518,13 @@ again1:
 			vm_page_t m = &pga[i];
 
 			pqtype = m->queue - m->pc;
-			if (pqtype == PQ_CACHE)
+			if (pqtype == PQ_CACHE) {
+				m->flags |= PG_BUSY;
 				vm_page_free(m);
+			}
 
 			TAILQ_REMOVE(vm_page_queues[m->queue].pl, m, pageq);
-			--(*vm_page_queues[m->queue].lcnt);
+			(*vm_page_queues[m->queue].lcnt)--;
 			cnt.v_free_count--;
 			m->valid = VM_PAGE_BITS_ALL;
 			m->flags = 0;
diff --git a/sys/vm/vm_pageout.c b/sys/vm/vm_pageout.c
index 8a6f97f..0ae68ea 100644
--- a/sys/vm/vm_pageout.c
+++ b/sys/vm/vm_pageout.c
@@ -65,7 +65,7 @@
  * any improvements or extensions that they make and grant Carnegie the
  * rights to redistribute these changes.
  *
- * $Id: vm_pageout.c,v 1.108 1998/01/17 09:17:01 dyson Exp $
+ * $Id: vm_pageout.c,v 1.109 1998/01/22 17:30:42 dyson Exp $
  */
 
 /*
@@ -403,8 +403,7 @@ vm_pageout_flush(mc, count, sync)
 			 * page so it doesn't clog the inactive list.  (We
 			 * will try paging out it again later).
 			 */
-			if (mt->queue == PQ_INACTIVE)
-				vm_page_activate(mt);
+			vm_page_activate(mt);
 			break;
 		case VM_PAGER_AGAIN:
 			break;
@@ -505,6 +504,7 @@ vm_pageout_object_deactivate_pages(map, object, desired, map_remove_only)
 						splx(s);
 					}
 				} else {
+					vm_page_activate(p);
 					p->flags &= ~PG_REFERENCED;
 					if (p->act_count < (ACT_MAX - ACT_ADVANCE))
 						p->act_count += ACT_ADVANCE;
@@ -595,18 +595,17 @@ vm_pageout_page_free(vm_page_t m) {
 	vm_object_t object;
 
 	object = m->object;
-	vp = NULL;
-
 	object->ref_count++;
+
 	if (object->type == OBJT_VNODE) {
 		vp = object->handle;
 		vp->v_usecount++;
 		if (VSHOULDBUSY(vp))
 			vbusy(vp);
 	}
+
 	m->flags |= PG_BUSY;
 	vm_page_protect(m, VM_PROT_NONE);
-	PAGE_WAKEUP(m);
 	vm_page_free(m);
 	vm_object_deallocate(object);
 }
@@ -778,7 +777,8 @@ rescan0:
 				continue;
 			}
 
-			if (object->type == OBJT_VNODE && (object->flags & OBJ_DEAD) == 0) {
+			if ((object->type == OBJT_VNODE) &&
+				(object->flags & OBJ_DEAD) == 0) {
 				vp = object->handle;
 				if (VOP_ISLOCKED(vp) ||
 				    vget(vp, LK_EXCLUSIVE|LK_NOOBJ, curproc)) {
diff --git a/sys/vm/vnode_pager.c b/sys/vm/vnode_pager.c
index aec5c55..2c2ec01 100644
--- a/sys/vm/vnode_pager.c
+++ b/sys/vm/vnode_pager.c
@@ -38,7 +38,7 @@
  * SUCH DAMAGE.
  *
  *	from: @(#)vnode_pager.c	7.5 (Berkeley) 4/20/91
- *	$Id: vnode_pager.c,v 1.79 1998/01/06 05:26:17 dyson Exp $
+ *	$Id: vnode_pager.c,v 1.80 1998/01/17 09:17:02 dyson Exp $
  */
 
 /*
@@ -315,7 +315,6 @@ void
 vnode_pager_freepage(m)
 	vm_page_t m;
 {
-	PAGE_WAKEUP(m);
 	vm_page_free(m);
 }