Re: [PATCH] Fixing soft NFS umount -f, round 1

[Chuck Silvers <chuq%chuq.com@localhost>]

On Fri, Jun 26, 2015 at 02:58:55PM +0000, Emmanuel Dreyfus wrote:
> On Fri, Jun 26, 2015 at 07:51:27AM -0700, Chuck Silvers wrote:
> > I'm not sure what you mean about a partial write, could you explain?
> 
> This is the problem of pages not being written by the filesystem: I 
> have not yet understood how to discard them. Setting bp->b_error
> is enough?

as long as all write I/Os eventually complete (with either success or failure),
the pages being written will eventually become clean and can be discarded.
in the failure case, bp->b_error should be non-zero, yes.
you shouldn't need to add anything new to explicitly discard the pages,
you just need to make sure that every write I/O eventually completes
(where timing out and failing counts as "complete").

-Chuck