[libvirt] [PATCH] Use loop-control to allocate loopback device.

[Ian Main]

On Tue, Aug 27, 2013 at 10:34:54AM -0700, Ian Main wrote:
> On Tue, Aug 27, 2013 at 03:28:07PM +0100, Daniel P. Berrange wrote:
> > On Mon, Aug 26, 2013 at 12:30:39PM -0700, Ian Main wrote:
> > > This patch changes virFileLoopDeviceOpen() to use the new loop-control
> > > device to allocate a new loop device.  If this behavior is unsupported
> > > or an error occurs while trying to do this it falls back to the previous
> > > method of searching /dev for a free device.
> > > 
> > > With this patch you can start as many image based LXC domains as you
> > > like (well almost).
> > > 
> > > Fixes bug https://bugzilla.redhat.com/show_bug.cgi?id=995543
> > > 
> > > Signed-off-by: Ian Main <imain at redhat.com>
> > > ---
> > >  configure.ac       | 12 +++++++++++
> > >  src/util/virfile.c | 61 +++++++++++++++++++++++++++++++++++++++++++++++++++++-
> > >  2 files changed, 72 insertions(+), 1 deletion(-)
> > > 
> > 
> > 
> > >  #if defined(__linux__) && HAVE_DECL_LO_FLAGS_AUTOCLEAR
> > > -static int virFileLoopDeviceOpen(char **dev_name)
> > > +
> > > +#if HAVE_DECL_LOOP_CTL_GET_FREE
> > > +static int virFileLoopDeviceOpenLoopCtl(char **dev_name)
> > > +{
> > > +    int fd = -1;
> > > +    int devnr;
> > > +    char *looppath = NULL;
> > > +
> > > +    VIR_DEBUG("Opening loop-control device");
> > > +    if ((fd = open("/dev/loop-control", O_RDWR)) < 0) {
> > > +        virReportSystemError(errno, "%s",
> > > +                             _("Unable to open /dev/loop-control"));
> > > +        return -1;
> > 
> > We need to distinguish ENOENT (which should be non-fatal) from
> > other (fatal) errors here.
> > 
> > I'd suggest that we add an 'int *fd' parameter to this method,
> > and use the return value to indicate success / failure only.
> > 
> > eg, 
> > 
> >   if (virFileLoopDeviceOpenLoopCtl(devname, &fd) < 0)
> >     return -1;
> >   if (fd == -1)
> >     fd = virFileLoopDeviceOpenSearch(devname);
> >   return fd;
> 
> You bet.  To be clear does this mean you don't want to fall back to the
> search method in the case of an error while using loop-control
> allocation?  My thinking was we'd just give it a go regardless of the
> issue..

nm, I see. heh

	Ian