From owner-freebsd-bugs  Wed Apr 15 11:22:29 1998
Return-Path: <owner-freebsd-bugs@FreeBSD.ORG>
Received: (from majordom@localhost)
          by hub.freebsd.org (8.8.8/8.8.8) id LAA03732
          for freebsd-bugs-outgoing; Wed, 15 Apr 1998 11:22:29 -0700 (PDT)
          (envelope-from owner-freebsd-bugs@FreeBSD.ORG)
Received: from flea.best.net (root@flea.best.net [206.184.139.131])
          by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id SAA03694;
          Wed, 15 Apr 1998 18:22:17 GMT
          (envelope-from dillon@flea.best.net)
Received: (from dillon@localhost) by flea.best.net (8.8.8/8.7.3) id LAA28517; Wed, 15 Apr 1998 11:21:32 -0700 (PDT)
Date: Wed, 15 Apr 1998 11:21:32 -0700 (PDT)
From: Matt Dillon <dillon@best.net>
Message-Id: <199804151821.LAA28517@flea.best.net>
To: Poul-Henning Kamp <phk@FreeBSD.ORG>
Cc: freebsd-bugs@FreeBSD.ORG
Subject: Re: kern/6258
Sender: owner-freebsd-bugs@FreeBSD.ORG
Precedence: bulk
X-Loop: FreeBSD.org

:
:Old Synopsis: A fix required to prevent kernel lockups in brelse causes the dump program to lockup in 'newbuf'
:New Synopsis: A fix required to prevent kernel lockups in brelse causes the dump program to lockup in 'newbuf' [2.2 ISSUE]
:
:State-Changed-From-To: open-analyzed
:State-Changed-By: phk
:State-Changed-When: Wed Apr 15 10:08:26 PDT 1998
:State-Changed-Why: 
:merge to 2.2 issue

    Oh, I should add something to the bug report... in email I think the 
    FreeBSD folks decided to fold in the -current changes to fix this problem.  
    The patch I included in the bug report would therefore not be used.

    However, since I did submit the patch I should probably submit an update
    to it as the patch I submitted has a bug in.  The bp scan I added in the
    patch overwrites the 'bp' variable which is used later on.

					-Matt

    Matthew Dillon   Engineering, BEST Internet Communications, Inc.
		    <dillon@best.net>
    [always include a portion of the original email in any response!]


--- LINK/vfs_bio.c	Fri Mar 13 13:13:57 1998
+++ vfs_bio.c	Mon Apr 13 17:21:49 1998
@@ -597,10 +597,12 @@
 		LIST_REMOVE(bp, b_hash);
 		LIST_INSERT_HEAD(&invalhash, bp, b_hash);
 		bp->b_dev = NODEV;
+#ifdef notdef
 		/*
 		 * Get rid of the kva allocation *now*
 		 */
 		bfreekva(bp);
+#endif
 		if (needsbuffer) {
 			wakeup(&needsbuffer);
 			needsbuffer=0;
@@ -986,9 +988,34 @@
 		 */
 		if (vm_map_findspace(buffer_map,
 			vm_map_min(buffer_map), maxsize, &addr)) {
-			bp->b_flags |= B_INVAL;
-			brelse(bp);
-			goto trytofreespace;
+				struct buf *bp2;
+
+			/*
+			 * Matt hack.  Since we can't call bfreekva() in
+			 * brelse(), the bp's on the EMPTY list may all
+			 * still have allocated KVM.  If we can't find
+			 * unused space in the buffer_map, we should try
+			 * to defragment the map by freeing as much from
+			 * the empty list as possible.
+			 */
+			printf("vm_map_findspace() failed, defragmenting freelist\n");
+			for (bp2 = TAILQ_FIRST(&bufqueues[QUEUE_EMPTY]);
+				bp2;
+				bp2 = TAILQ_NEXT(bp2, b_freelist)
+			) {
+			    if (bp2->b_kvasize)
+				bfreekva(bp2);
+			    if (bp2->b_qindex != QUEUE_EMPTY)
+				break;
+			}
+			addr = 0;
+			if (vm_map_findspace(buffer_map,
+				vm_map_min(buffer_map), maxsize, &addr)) {
+
+				bp->b_flags |= B_INVAL;
+				brelse(bp);
+				goto trytofreespace;
+			}
 		}
 	}
 

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-bugs" in the body of the message