Re: [Qemu-devel] iSCSI options for IQN with colons

[Pino Toscano <ptoscano@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 2561 bytes --]

On Tuesday 01 December 2015 10:27:28 Markus Armbruster wrote:
> Beware, I know next to nothing about iSCSI.
> 
> Pino Toscano <ptoscano@redhat.com> writes:
> 
> > Hi,
> >
> > while testing the integration of QEMU with iSCSI, I was setting up an
> > environment with both target and initiator IQNs with colons. Then I
> > tried to connect to two different targets using two different initiator
> > IQN, like the following:
> >
> >   $ qemu ... \
> >       -iscsi id=iqn.2015-11.com.bla:suffix1,initiator-name=iqn.2015-11.com.bla:suffix1-initiator \
> >       -iscsi id=iqn.2015-11.com.bla:suffix2,initiator-name=iqn.2015-11.com.bla:suffix2-initiator \
> >       -drive file=iscsi://server/iqn.2015-11.com.bla%3Asuffix1/0,format=raw,id=hd1,if=none \
> >       -drive file=iscsi://server/iqn.2015-11.com.bla%3Asuffix2/0,format=raw,id=hd2,if=none \
> >       ...
> >
> > which didn't work at first:
> >
> >   qemu-system-x86_64: -iscsi id=iqn.2015-11.com.bla:suffix1,initiator-name=iqn.2015-11.com.bla:suffix1-initiator: Parameter 'id' expects an identifier
> >
> > which, according to id_wellformed in id.c, is true. Allowing colons in
> > id=... like in the following patch
> >
> > diff --git a/util/id.c b/util/id.c
> > index bcc64d8..25fca9d 100644
> > --- a/util/id.c
> > +++ b/util/id.c
> > @@ -20,7 +20,7 @@ bool id_wellformed(const char *id)
> >          return false;
> >      }
> >      for (i = 1; id[i]; i++) {
> > -        if (!qemu_isalnum(id[i]) && !strchr("-._", id[i])) {
> > +        if (!qemu_isalnum(id[i]) && !strchr("-._:", id[i])) {
> >              return false;
> >          }
> >      }
> >
> > allowed me to work run QEMU with the attached disks.
> >
> > The question basically boils down to whether it is right to reject
> > colons in id:
> > - if so, then there should be a way to allow them only in id of -iscsi
> >   (since colons can be part of IQNs)
> > - if not, whether allowing them could cause regressions in option
> >   parsing
> 
> Have you tried
> 
>     -iscsi id=iscsi.1,initiator-name=iqn.2015-11.com.bla:suffix1-initiator \
>     -iscsi id=iscsi.2,initiator-name=iqn.2015-11.com.bla:suffix2-initiator \
>     -drive file=iscsi://server/iqn.2015-11.com.bla%3Asuffix1/0,format=raw,id=hd1,if=none \
>     -drive file=iscsi://server/iqn.2015-11.com.bla%3Asuffix2/0,format=raw,id=hd2,if=none \

This won't work, as the various parse_* in iscsi.c (e.g.
parse_initiator_name for the above cases) use the target IQN as
identifier for the parameters.

-- 
Pino Toscano

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 819 bytes --]