Re: splice(SPLICE_F_MOVE) problems

[Jens Axboe]

On Tue, May 02 2006, Oleg Nesterov wrote:
> On 05/01, Jens Axboe wrote:
> >
> > > If readahead doesn't work, SPLICE_F_MOVE is problematic too.
> > > add_to_page_cache_lru()->lru_cache_add() first increments
> > > page->count and adds this page to lru_add_pvecs. This means
> > > page_cache_pipe_buf_steal()->remove_mapping() will probably
> > > fail.
> > 
> > Because of the temporarily elevated page count?
> 
> Yes.
> 
> On the other hand, if readahead doesn't work we already have a
> bigger problem, and SPLICE_F_MOVE is not garanteed, so I think
> this is very minor.

Yes, clearly readahead has to work as expected. I haven't noticed any
problems, even on half cached workloads. With your handle_ra_miss()
addition and possibly killing the redundant !offset || nr_pages check,
it should be fine.

-- 
Jens Axboe

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/