On Mon, Nov 26, 2012 at 06:21:37PM +0800, Osier Yang wrote: > Since "rawio" and "unpriv_sgio" are only valid for "lun", this I think that's a ugly attribute name, just use 'sgio'. > groups them together. And since both of them intend to allow > the unprivledged user to use the SCSI commands, they are must be > exclusive. Actually "unpriv_sgio" supersedes "rawio", as it > confines the capability per-device, unlike "rawio", which gives > the domain process broad capablity. I'd tend to say that the rawio behaviour you describe is really just a QEMU implementation detail. The XML itself is written on the basis that the 'rawio' flag is specific to one device. eg it would be possible to make the 'rawio' flag work on just a single device if we really wanted to do the kernel work. > --- > docs/formatdomain.html.in | 10 +++- > docs/schemas/domaincommon.rng | 52 ++++++++++---- > src/conf/domain_conf.c | 56 ++++++++++++---- > src/conf/domain_conf.h | 11 +++ > src/libvirt_private.syms | 4 + > src/qemu/qemu_process.c | 30 ++++++++ > src/util/util.c | 153 +++++++++++++++++++++++++++++++++++++++++ > src/util/util.h | 7 ++ > 8 files changed, 293 insertions(+), 30 deletions(-) > > diff --git a/docs/formatdomain.html.in b/docs/formatdomain.html.in > index 6a3b976..f3f6a9e 100644 > --- a/docs/formatdomain.html.in > +++ b/docs/formatdomain.html.in > @@ -1395,7 +1395,15 @@ > rawio='yes', rawio capability will be enabled for all disks in > the domain (because, in the case of QEMU, this capability can > only be set on a per-process basis). This attribute is only > - valid when device is "lun". > + valid when device is "lun". NB, <code>rawio</code> gives > + the domain process broad capability, to confined the capability > + as much as possible, one should use <code>unpriv_sgio</code> > + instead, which controls the capability per-device. > + The optional <code>unpriv_sgio</code> attribute > + (<span class="since">since 1.0.1</span>) indicates whether the > + disk will allow unprivileged SG_IO, valid settings are "yes" > + or "no" (defaults to "no"). Note that it's exclusive with > + attribute <code>rawio</code>; As above, I think this needs re-wording to clearly differentiate between the intended meaning of the 'rawio' attribute in general, vs the current QEMU specific impl of that attribute. Daniel -- |: http://berrange.com -o- http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :| -- libvir-list mailing list libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list