On Tue, 9 Jun 1998 12:10:37 -0700 
 "Erik E. Fair" <fair@clock.org> wrote:

 > The other issue here might simply be a failure of the kernel to wait long
 > enough for I/O to complete.

If you issue the SYNCHRONIZE CACHE command _without_ the Immed bit set,
and wait for it to complete, then you are guaranteed that the data is on
stable storage.

As far as the kernel knew, the I/O was done; it got an interrupt saying
so... I don't think inserting a delay is going to do any good... you just
need to make sure the cache is flushed by forcing it out.

 > This is not to say that a "SYNCHRONIZE CACHE" command sent to *all* disks
 > at any halt or powerdown event (and maybe in sync(2)?) is not a good idea...

I dunno about sync(2) ... there's not really a mechanism for the VFS code
to do this...

Jason R. Thorpe                                       thorpej@nas.nasa.gov
NASA Ames Research Center                            Home: +1 408 866 1912
NAS: M/S 258-5                                       Work: +1 650 604 0935
Moffett Field, CA 94035                             Pager: +1 650 428 6939