Re: [Xen-devel] [PATCH v2 01/17] configure: make the libaio test conditional on blktap1

[Ian Campbell <Ian.Campbell@xxxxxxxxxx>]

On Thu, 2014-06-05 at 17:49 +0200, Roger Pau Monnà wrote:
> On 05/06/14 17:06, Ian Campbell wrote:
> > On Mon, 2014-06-02 at 17:08 +0200, Roger Pau Monne wrote:
> >> libaio is only required for blktap1, so make the check conditional on
> >> whether the user has requested blktap1 or not.
> > 
> > I think it is used for blktap2 as well.
> 
> OK, I have such a mess with blktap versions and which ones are inside of
> the tree. Right now configure only allows enabling/disabling blktap1,
> and blktap2 is compiled depending on the host OS version. Should I
> change blktap2 to be selected on configure and disabled by default?
> (just like blktap1)

That might be an interesting change but I wouldn't want to intertwine it
with the FreeBSD support for fear that someone objects.

Less controversial would be to move the host OS version check for
blktap2 into configure and use it as the default, and then base the
libaio check on that?

> > I thought blktap2 was Linux specific (and hence that the libaio could be
> > too) but, from tools/Makefile:
> > SUBDIRS-$(CONFIG_Linux) += blktap2
> > SUBDIRS-$(CONFIG_NetBSD) += blktap2
> > 
> > I'm totally surprised by that! And TBH I find it a little hard to
> > believe...
> 
> Yes, this is certainly a mistake, NetBSD never had anything similar to
> blktap2 AFAIK.

Looking at tools/blktap2/Makefile it looks like the lvm and vhd bits
would be build on BSD but not drivers or control. I've no idea how
useful the lvm and vhd stuff is on NetBSD...

However: the aio stuff is strictly in the drivers subtree. So I think
you can just make the libaio check conditional on Linux only. If you
also want to improve the situation around blktap2 configuration then I
think that would be wonderful, but not required for this series.

Ian.


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel