Date: Tue, 17 May 2016 11:42:41 +0300 From: Konstantin Belousov <kostikbel@gmail.com> To: Bruce Evans <brde@optusnet.com.au> Cc: fs@freebsd.org Subject: Re: fix for per-mount i/o counting in ffs Message-ID: <20160517084241.GY89104@kib.kiev.ua> In-Reply-To: <20160517072104.I2137@besplex.bde.org> References: <20160517072104.I2137@besplex.bde.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, May 17, 2016 at 07:26:08AM +1000, Bruce Evans wrote:
> Counting of i/o's in g_vfs_strategy() requires the fs to initialize
> devvp->v_rdev->si_mountpt to non-null. This seems to be done correctly
> in ext2fs and msdosfs, but in ffs it is not done for ro mounts, or for
> rw mounts that started as ro. The bug is most obvious for the root
> file system since it always starts as ro.
I committed the comments updates.
For the accounting patch, don't we want to account for all io, including
the mount-time metadata reads and initial superblock update ?
diff --git a/sys/ufs/ffs/ffs_vfsops.c b/sys/ufs/ffs/ffs_vfsops.c
index 9776554..712fc21 100644
--- a/sys/ufs/ffs/ffs_vfsops.c
+++ b/sys/ufs/ffs/ffs_vfsops.c
@@ -780,6 +780,8 @@ ffs_mountfs(devvp, mp, td)
mp->mnt_iosize_max = MAXPHYS;
devvp->v_bufobj.bo_ops = &ffs_ops;
+ if (devvp->v_type == VCHR)
+ devvp->v_rdev->si_mountpt = mp;
fs = NULL;
sblockloc = 0;
@@ -1049,8 +1051,6 @@ ffs_mountfs(devvp, mp, td)
ffs_flushfiles(mp, FORCECLOSE, td);
goto out;
}
- if (devvp->v_type == VCHR && devvp->v_rdev != NULL)
- devvp->v_rdev->si_mountpt = mp;
if (fs->fs_snapinum[0] != 0)
ffs_snapshot_mount(mp);
fs->fs_fmod = 1;
@@ -1083,6 +1083,8 @@ ffs_mountfs(devvp, mp, td)
out:
if (bp)
brelse(bp);
+ if (devvp->v_type == VCHR && devvp->v_rdev != NULL)
+ devvp->v_rdev->si_mountpt = NULL;
if (cp != NULL) {
DROP_GIANT();
g_topology_lock();
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20160517084241.GY89104>