Re: finding the perpetrator of regular disk I/O

[Taylor R Campbell <campbell+netbsd%mumble.net@localhost>]

   Date: Thu, 24 Mar 2011 18:33:35 +0000
   From: David Laight <david%l8s.co.uk@localhost>

   The atime/mtime for /dev entries ??

No -- I use noatime and nodevmtime, and the 16KB writes I observe
don't seem to change the atime, ctime, or mtime of any file on the
disk.

/dev/cgd1a on / type ffs (synchronous, noatime, nodevmtime, local)
/dev/cgd1e on /var type ffs (synchronous, nosuid, nodev, noatime, local)
/dev/cgd1f on /var/chroot type ffs (synchronous, noatime, nodevmtime, local)
/dev/cgd1g on /pkg type ffs (synchronous, nodev, noatime, local)
/dev/cgd1h on /home type ffs (synchronous, nosuid, nodev, noatime, local)
/pkg/local on /usr/local type null (read-only, local)
kernfs on /kern type kernfs (local)
procfs on /proc type procfs (local)
tmpfs on /tmp type tmpfs (nosuid, nodev, local)

   You might be see inode flushes.

Can you elaborate?  Is this different from what the kernel's syncer
does?  I tried running sync(8) a dozen times before putting the disk
into standby, and it still spun right back up again.