Re: [Xen-devel] [PATCH] tools: specifically enable VirtFS in Linux QEMU builds

[Stefano Stabellini <sstabellini@xxxxxxxxxx>]

On Fri, 7 Sep 2018, Paul Durrant wrote:
> > -----Original Message-----
> > From: George Dunlap [mailto:george.dunlap@xxxxxxxxxx]
> > Sent: 07 September 2018 16:26
> > To: Roger Pau Monne <roger.pau@xxxxxxxxxx>; Paul Durrant
> > <Paul.Durrant@xxxxxxxxxx>
> > Cc: 'Stefano Stabellini' <sstabellini@xxxxxxxxxx>; Wei Liu
> > <wei.liu2@xxxxxxxxxx>; Konrad Rzeszutek Wilk <konrad.wilk@xxxxxxxxxx>;
> > Andrew Cooper <Andrew.Cooper3@xxxxxxxxxx>; Tim (Xen.org)
> > <tim@xxxxxxx>; Julien Grall <julien.grall@xxxxxxx>; Jan Beulich
> > <jbeulich@xxxxxxxx>; Ian Jackson <Ian.Jackson@xxxxxxxxxx>; xen-
> > devel@xxxxxxxxxxxxxxxxxxxx
> > Subject: Re: [Xen-devel] [PATCH] tools: specifically enable VirtFS in Linux
> > QEMU builds
> > 
> > On 09/07/2018 03:57 PM, Roger Pau Monné wrote:
> > > On Fri, Sep 07, 2018 at 08:35:11AM +0000, Paul Durrant wrote:
> > >>> -----Original Message-----
> > >>> From: Stefano Stabellini [mailto:sstabellini@xxxxxxxxxx]
> > >>> Sent: 06 September 2018 19:12
> > >>> To: Paul Durrant <Paul.Durrant@xxxxxxxxxx>
> > >>> Cc: xen-devel@xxxxxxxxxxxxxxxxxxxx; Andrew Cooper
> > >>> <Andrew.Cooper3@xxxxxxxxxx>; George Dunlap
> > >>> <George.Dunlap@xxxxxxxxxx>; Ian Jackson <Ian.Jackson@xxxxxxxxxx>;
> > Jan
> > >>> Beulich <jbeulich@xxxxxxxx>; Julien Grall <julien.grall@xxxxxxx>;
> > Konrad
> > >>> Rzeszutek Wilk <konrad.wilk@xxxxxxxxxx>; Stefano Stabellini
> > >>> <sstabellini@xxxxxxxxxx>; Tim (Xen.org) <tim@xxxxxxx>; Wei Liu
> > >>> <wei.liu2@xxxxxxxxxx>
> > >>> Subject: Re: [PATCH] tools: specifically enable VirtFS in Linux QEMU
> > builds
> > >>>
> > >>> On Thu, 6 Sep 2018, Paul Durrant wrote:
> > >>>> 9pfs support has been a documented feature since Xen 4.9, but QEMU
> > will
> > >>>> not be built with backend support unless libcap and libattr dev
> > packages
> > >>>> are installed.
> > >>>>
> > >>>> This patch modifies the README to call out those packages as pre-
> > >>> requisites
> > >>>> for Linux builds and specifically enables VirtFS in the configure line
> > >>>> for QEMU so that an error message is displayed if they are missing.
> > >>>>
> > >>>> Signed-off-by: Paul Durrant <paul.durrant@xxxxxxxxxx>
> > >>>
> > >>> Thank you, Paul!
> > >>
> > >> NP. I'm at the point of seriously considering a Windows frontend, which 
> > >> is
> > why I've been playing.
> > >>
> > >> BTW, as a heads-up... Before I realised QEMU was not building the
> > backend, I tried a modprobe of the Xen 9pfs transport module in a Linux
> > guest and immediately hit a kernel BUG (NULL ptr IIRC) which was apparently
> > down to the frontend expecting the backend to be present when its probe
> > routine is run, rather than deferring things until the backend goes into
> > InitWait (which is what a frontend generally should do). Also, it's 
> > non-obvious
> > why I have to explicitly modprobe anyway... Shouldn't the module get
> > demand-loaded when I run mount -t 9pfs blah?
> > >>
> > >>> Do we need to do anything for the configure stuff
> > >>> (AC_CHECK_LIB in tools/configure.ac)?
> > >>
> > >> That's certainly a reasonable belt'n'braces approach so that folks don't
> > have to wait until their tools build fails to find out what they need. I'll 
> > have a
> > look at that.
> > >
> > > IMO I don't think we should be duplicating the checks that QEMU
> > > configure already does in our configure script. The best option would
> > > be to run whatever external configure scripts there are together with
> > > our own configure script, but this is quite complicated because the
> > > qemu-dir might have to be fetched at the point where configure runs.
> > 
> > Not the least because the dependencies may change.  I think adding an
> > "--enable-9pfs" option which will pass on the requisite "--enable" to
> > qemu (such that qemu build will fail if the prereqs are not present)
> > makes the most sense.
> 
> I'd be happy with that if others are.

If QEMU changes the virtfs default to "enable", then this option will
become ineffective maybe without us even noticing. However, I guess we
prefer to deal with duplicate enable/disable options in QEMU and Xen,
rather than duplicate dependency checks which might be harder to keep
up-to-date in the long run?

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel