Date: 16 Oct 2001 20:44:57 +0200 From: Dag-Erling Smorgrav <des@ofug.org> To: freebsd-hackers@FreeBSD.org Cc: John Baldwin <jhb@FreeBSD.org>, cvs-all@FreeBSD.org, cvs-committers@FreeBSD.org, Bruce Evans <bde@zeta.org.au> Subject: Re: cvs commit: src/sys/vm vnode_pager.c Message-ID: <xzpadyrxw46.fsf@flood.ping.uio.no> In-Reply-To: <200110161800.f9GI0QP33797@apollo.backplane.com> References: <XFMail.011016103432.jhb@FreeBSD.org> <200110161800.f9GI0QP33797@apollo.backplane.com>
next in thread | previous in thread | raw e-mail | index | archive | help
--=-=-=
Matthew Dillon <dillon@apollo.backplane.com> writes:
> Just calling a mutex function hundreds of thousands of times is
> going to be enough to cause the problem. I see a way we can fix the
> problem... the mount point lock is locking the list, so if the
> information we need to determine whether we have work to do is stored in
> the vnode itself we can check and loop without having to mess
> around with any more mutexes. If the flag says that there
> may be something to do, *then* we do the hard work.
That sounds reasonable. One stray thought, by the way - could it be
that vnodes aren't being reclaimed as fast as they should? What's the
policy - do vnodes only get reclaimed when we start running out?
Should we re-evaluate the cost of having them slow down ffs_sync()
vs. what we save by keeping them around so we don't need to reallocate
them?
Attached are a patch that adds counters to ffs_sync() (to see how many
vnodes are traversed each time, and how many of those actually needed
syncing), and a script I'm using to roughly measure the performance of
ffs_sync().
DES
--
Dag-Erling Smorgrav - des@ofug.org
--=-=-=
Content-Type: text/x-patch
Content-Disposition: attachment; filename=ffs_sync.diff
Index: sys/ufs/ffs/ffs_vfsops.c
===================================================================
RCS file: /home/ncvs/src/sys/ufs/ffs/ffs_vfsops.c,v
retrieving revision 1.161
diff -u -r1.161 ffs_vfsops.c
--- sys/ufs/ffs/ffs_vfsops.c 2 Oct 2001 14:34:22 -0000 1.161
+++ sys/ufs/ffs/ffs_vfsops.c 16 Oct 2001 17:21:34 -0000
@@ -972,6 +972,9 @@
return (0);
}
+static int count_synced_vnodes = 0;
+SYSCTL_INT(_debug, OID_AUTO, count_synced_vnodes, CTLFLAG_RW, &count_synced_vnodes, 0, "");
+
/*
* Go through the disk queues to initiate sandbagged IO;
* go through the inodes to write those that have been modified;
@@ -991,6 +994,7 @@
struct ufsmount *ump = VFSTOUFS(mp);
struct fs *fs;
int error, count, wait, lockreq, allerror = 0;
+ int looped = 0, vn_traversed = 0, vn_synced = 0;
fs = ump->um_fs;
if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */
@@ -1008,6 +1012,7 @@
}
mtx_lock(&mntvnode_mtx);
loop:
+ ++looped;
for (vp = LIST_FIRST(&mp->mnt_vnodelist); vp != NULL; vp = nvp) {
/*
* If the vnode that we are about to sync is no longer
@@ -1017,6 +1022,7 @@
goto loop;
nvp = LIST_NEXT(vp, v_mntvnodes);
+ ++vn_traversed;
mtx_unlock(&mntvnode_mtx);
mtx_lock(&vp->v_interlock);
ip = VTOI(vp);
@@ -1027,6 +1033,7 @@
mtx_lock(&mntvnode_mtx);
continue;
}
+ ++vn_synced;
if (vp->v_type != VCHR) {
if ((error = vget(vp, lockreq, td)) != 0) {
mtx_lock(&mntvnode_mtx);
@@ -1045,6 +1052,12 @@
mtx_lock(&mntvnode_mtx);
}
mtx_unlock(&mntvnode_mtx);
+
+ if (count_synced_vnodes)
+ printf(__FUNCTION__
+ "(): %d loops, %d vnodes traversed, %d vnodes synced\n",
+ looped, vn_traversed, vn_synced);
+
/*
* Force stale file system control information to be flushed.
*/
--=-=-=
Content-Type: application/x-sh
Content-Disposition: attachment; filename=measure.sh
#!/bin/sh
vmstat -m | grep FFS
sysctl debug.count_synced_vnodes=1
time sync
sysctl debug.count_synced_vnodes=0
dmesg | grep ffs_sync | tail -5
--=-=-=--
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?xzpadyrxw46.fsf>