[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20120920135257.GD1998@localhost.localdomain>
Date: Thu, 20 Sep 2012 09:52:59 -0400
From: Konrad Rzeszutek Wilk <konrad.wilk@...cle.com>
To: Oliver Chick <oliver.chick@...rix.com>
Cc: "xen-devel@...ts.xen.org" <xen-devel@...ts.xen.org>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] Persistent grant maps for xen blk drivers
On Thu, Sep 20, 2012 at 02:12:25PM +0100, Oliver Chick wrote:
> Thank you for all your useful feedback, Konrad.
>
> I have made most of the changes you suggest. I will test them and
> repost. I have made a few points below.
Great. Looking forward to your next patch. Also pls include it
as an attachment as I could not apply it last time.. Not sure if
that is something with your mailer - but attachments usually survive
any MUA mangling.
> > > + page_to_pfn(pers_gnt->page));
> >
> > Would it make sense to cache this in the 'pers_gnt' structure?
>
> As far as I can tell, we only need this value when mapping, and
> unmapping. So if we cache it, we will use it a maximum of one time. I
> think it's cheap to calculate. Am I right?
Then lets not cache it.
> > > @@ -688,7 +804,7 @@ static int dispatch_rw_block_io(struct xen_blkif *blkif,
> > > for (i = 0; i < nseg; i++) {
> > > while ((bio == NULL) ||
> > > (bio_add_page(bio,
> > > - blkbk->pending_page(pending_req, i),
> >
> > Can we get rid of pending_page macro?
>
> Unfortunately not, it is still used in the non-persistent mode to
> populate the pages[].
Fair enough
> > > + memcpy(shared_data + sg->offset,
> > > + bvec_data + sg->offset,
> > > + sg->length);
> >
> > Do we need to worry about it spilling over a page? Should we check that
> > sg>offset + sg->length < PAGE_SIZE ?
>
> I agree, this is probably a worthwhile thing to check.
>
> >
> > Also this would imply that based on the offset (so say it is 3999) that the old data (0->3998)
> > might be still there - don't know how important that is?
>
> This is true. I spoke with IanC about this, and we *think* that this is
> ok. Any old data that is there will have already been given to blkback,
> so we're not really leaking data that we shouldn't be.
Put a comment in saying that. In case in the future we want to share
the persistent grants with other domains, we would need to address this.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists