Jason Thorpe <thorpej@nas.nasa.gov> writes:
> Not.  What about the case where IDE disks, which CANNOT be issued a
> "flush your cache" call, are still writing their data?

If it's for that -- and thus far, nobody but you has claimed that it
is -- then it should be:

* by that code (i.e. the sync bits, and probably right after them),

* documented as such, and

* for instance, not invoked if not syncing.


you've argued for something else, not this.


cgd
-- 
Chris Demetriou - cgd@netbsd.org - http://www.netbsd.org/People/Pages/cgd.html
Disclaimer: Not speaking for NetBSD, just expressing my own opinion.