The soft and hard busy mechanism rely on the vm object lock to work.

Unify the 2 concept into a real, minimal, sxlock where the shared
acquisition represent the soft busy and the exclusive acquisition
represent the hard busy.
The old VPO_WANTED mechanism becames the hard-path for this new lock
and it becomes per-page rather than per-object.
The vm_object lock becames an interlock for this functionality:
it can be held in both read or write mode.
However, if the vm_object lock is held in read mode while acquiring
or releasing the busy state, the thread owner cannot make any
assumption on the busy state unless it is also busying it.

Also:
- Add a new flag to directly shared busy pages while vm_page_alloc
  and vm_page_grab are being executed.  This will be very helpful
  once these functions happen under a read object lock.
- Move the swapping sleep into its own per-object flag

The KPI is heavilly changed this is why the version is bumped.
It is very likely that some VM ports users will need to change
their own code.

Sponsored by:	EMC / Isilon storage division
Discussed with:	alc
Reviewed by:	jeff, kib
Tested by:	gavin, bapt (older version)
Tested by:	pho, scottl

2 files changed, 7 insertions, 14 deletions

diff --git a/sys/fs/tmpfs/tmpfs_subr.c b/sys/fs/tmpfs/tmpfs_subr.c
index 4e94476..94498ff 100644
--- a/sys/fs/tmpfs/tmpfs_subr.c
+++ b/sys/fs/tmpfs/tmpfs_subr.c
@@ -1355,11 +1355,8 @@ tmpfs_reg_resize(struct vnode *vp, off_t newsize, boolean_t ignerr)
 retry:
 			m = vm_page_lookup(uobj, idx);
 			if (m != NULL) {
-				if ((m->oflags & VPO_BUSY) != 0 ||
-				    m->busy != 0) {
-					vm_page_sleep(m, "tmfssz");
+				if (vm_page_sleep_if_busy(m, "tmfssz"))
 					goto retry;
-				}
 				MPASS(m->valid == VM_PAGE_BITS_ALL);
 			} else if (vm_pager_has_page(uobj, idx, NULL, NULL)) {
 				m = vm_page_alloc(uobj, idx, VM_ALLOC_NORMAL);
@@ -1379,7 +1376,7 @@ retry:
 				if (rv == VM_PAGER_OK) {
 					vm_page_deactivate(m);
 					vm_page_unlock(m);
-					vm_page_wakeup(m);
+					vm_page_xunbusy(m);
 				} else {
 					vm_page_free(m);
 					vm_page_unlock(m);
diff --git a/sys/fs/tmpfs/tmpfs_vnops.c b/sys/fs/tmpfs/tmpfs_vnops.c
index d867612..900ea79 100644
--- a/sys/fs/tmpfs/tmpfs_vnops.c
+++ b/sys/fs/tmpfs/tmpfs_vnops.c
@@ -449,7 +449,7 @@ tmpfs_nocacheread(vm_object_t tobj, vm_pindex_t idx,
 
 	/*
 	 * Parallel reads of the page content from disk are prevented
-	 * by VPO_BUSY.
+	 * by exclusive busy.
 	 *
 	 * Although the tmpfs vnode lock is held here, it is
 	 * nonetheless safe to sleep waiting for a free page.  The
@@ -457,10 +457,8 @@ tmpfs_nocacheread(vm_object_t tobj, vm_pindex_t idx,
 	 * lock to page out tobj's pages because tobj is a OBJT_SWAP
 	 * type object.
 	 */
-	m = vm_page_grab(tobj, idx, VM_ALLOC_NORMAL | VM_ALLOC_RETRY |
-	    VM_ALLOC_NOBUSY);
+	m = vm_page_grab(tobj, idx, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
 	if (m->valid != VM_PAGE_BITS_ALL) {
-		vm_page_busy(m);
 		if (vm_pager_has_page(tobj, idx, NULL, NULL)) {
 			rv = vm_pager_get_pages(tobj, &m, 1, 0);
 			m = vm_page_lookup(tobj, idx);
@@ -483,8 +481,8 @@ tmpfs_nocacheread(vm_object_t tobj, vm_pindex_t idx,
 			}
 		} else
 			vm_page_zero_invalid(m, TRUE);
-		vm_page_wakeup(m);
 	}
+	vm_page_xunbusy(m);
 	vm_page_lock(m);
 	vm_page_hold(m);
 	vm_page_unlock(m);
@@ -574,10 +572,8 @@ tmpfs_mappedwrite(vm_object_t tobj, size_t len, struct uio *uio)
 	tlen = MIN(PAGE_SIZE - offset, len);
 
 	VM_OBJECT_WLOCK(tobj);
-	tpg = vm_page_grab(tobj, idx, VM_ALLOC_NORMAL | VM_ALLOC_NOBUSY |
-	    VM_ALLOC_RETRY);
+	tpg = vm_page_grab(tobj, idx, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
 	if (tpg->valid != VM_PAGE_BITS_ALL) {
-		vm_page_busy(tpg);
 		if (vm_pager_has_page(tobj, idx, NULL, NULL)) {
 			rv = vm_pager_get_pages(tobj, &tpg, 1, 0);
 			tpg = vm_page_lookup(tobj, idx);
@@ -600,8 +596,8 @@ tmpfs_mappedwrite(vm_object_t tobj, size_t len, struct uio *uio)
 			}
 		} else
 			vm_page_zero_invalid(tpg, TRUE);
-		vm_page_wakeup(tpg);
 	}
+	vm_page_xunbusy(tpg);
 	vm_page_lock(tpg);
 	vm_page_hold(tpg);
 	vm_page_unlock(tpg);