Re: aborting IO requests

On Tue, 2002-12-03 at 18:03, Henrik Nordstrom wrote:
> On 3 Dec 2002, Robert Collins wrote:
>
> > This appears to be inconsistent across fs types. Sigh. I'll just let
> > everything complete, at a worst case of what, 4Kb, it's not a great
> > overhead, and it leaves everything simpler while I get rid of the huge
> > wads of duplicate code.
>
> Actually it is quite consistent at the moment.. initiated reads are
> allowed to complete and is then cancelled before copying the data into the
> supplied buffer
>
> The aufs store has lots of code to deal with aborted reads but most is
> actually disabled at the moment.

More specifically, what I meant was that the store object close routines
appear to take different approaches in aufs and diskd. The end behaviour
may be consistent.

Rob