Buffer cache revamp

1) remove multiple size queues, introduced as a stopgap.
2) decouple pages containing data from their mappings
3) only keep buffers mapped when they actually have to be mapped
  (right now, this is when buffers are B_BUSY)
4) New functions to make a buffer busy, and release the busy flag
   (buf_acquire and buf_release)
5) Move high/low water marks and statistics counters into a structure
6) Add a sysctl to retrieve buffer cache statistics

Tested in several variants and beat upon by bob and art for a year. run
accidentally on henning's nfs server for a few months...

ok deraadt@, krw@, art@ - who promises to be around to deal with any fallout

1 files changed, 10 insertions, 6 deletions

diff --git a/sys/kern/vfs_subr.c b/sys/kern/vfs_subr.c
index 3ac320e5bc1..65153b0cdfc 100644
--- a/sys/kern/vfs_subr.c
+++ b/sys/kern/vfs_subr.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: vfs_subr.c,v 1.167 2008/06/09 23:38:37 millert Exp $	*/
+/*	$OpenBSD: vfs_subr.c,v 1.168 2008/06/10 20:14:36 beck Exp $	*/
 /*	$NetBSD: vfs_subr.c,v 1.53 1996/04/22 01:39:13 christos Exp $	*/
 
 /*
@@ -120,7 +120,7 @@ void
 vntblinit(void)
 {
 	/* buffer cache may need a vnode for each buffer */
-	maxvnodes = desiredvnodes;
+	maxvnodes = bufpages;
 	pool_init(&vnode_pool, sizeof(struct vnode), 0, 0, 0, "vnodes",
 	    &pool_allocator_nointr);
 	TAILQ_INIT(&vnode_hold_list);
@@ -1256,8 +1256,12 @@ vfs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
 
 		free(tmpvfsp, M_TEMP);
 		return (ret);
-	}
+	case VFS_BCACHESTAT:	/* buffer cache statistics */
+		ret = sysctl_rdstruct(oldp, oldlenp, newp, &bcstats,
+		    sizeof(struct bcachestats));
+		return(ret);
 
+	}
 	return (EOPNOTSUPP);
 }
 
@@ -1664,7 +1668,7 @@ vfs_syncwait(int verbose)
 			if (bp->b_flags & B_DELWRI) {
 				s = splbio();
 				bremfree(bp);
-				bp->b_flags |= B_BUSY;
+				buf_acquire(bp);
 				splx(s);
 				nbusy++;
 				bawrite(bp);
@@ -1835,7 +1839,7 @@ loop:
 				break;
 			}
 			bremfree(bp);
-			bp->b_flags |= B_BUSY;
+			buf_acquire(bp);
 			/*
 			 * XXX Since there are no node locks for NFS, I believe
 			 * there is a slight chance that a delayed write will
@@ -1873,7 +1877,7 @@ loop:
 		if ((bp->b_flags & B_DELWRI) == 0)
 			panic("vflushbuf: not dirty");
 		bremfree(bp);
-		bp->b_flags |= B_BUSY;
+		buf_acquire(bp);
 		splx(s);
 		/*
 		 * Wait for I/O associated with indirect blocks to complete,