On Thu, Sep 15, 2011 at 09:01:38AM +0200, Paolo Bonzini wrote:
> On 09/15/2011 05:14 AM, David Gibson wrote:
> >Under PAPR, there is generally only
> >supposed to be one SCSI target (disk / cd / whatever) per virtual scsi
> >bus. But the generic qdev code will, by default, keep assigning
> >devices to the existing bus until it's full. Any thoughts on how to
> >sanely change that behaviour on a per-machine basis?
>
> You could change the if_max_devs array in blockdev.c to something
> provided by the machines.
>
> However, I'm not sure about this, for two reasons:
>
> 1) do you mean, in Linux terms, one target per SCSI _host_ or one
> target per SCSI _channel_? i.e. if you looks at
> /sys/bus/scsi/devices, right now it looks like
>
> 0:0:0:0 0:0:1:0 (two targets on the same host and channel)
>
> Should it be?
>
> 0:0:0:0 0:1:0:0 (one target per channel)
>
> or
>
> 0:0:0:0 1:0:0:0 (one target per host)
>
> If it is the former, then you are simply hitting a limitation of the
> SCSI layer in QEMU and I do have patches to make assignment more
> flexible. Based on the Linux VSCSI driver, and based on what SLOF
> does, I'd guess that's what you mean.
Well, now I'm confused. I had a look at a pHyp machine, and Linux
seemed to see it as multiple targets on a single channel, but I'm sure
the PAPR spec says you shouldn't have that. So I'm going to have to
look closer now.
> 2) does this matter at all? First, when doing "real world"
> virtualization you won't use the legacy options (neither -hda/-cdrom
> nor "-drive ...,if=scsi"), you would use -device to manually assign
> the devices to their buses.
Well, perhaps, but I really prefer to have sane defaults, rather than
having to build the machine myself on the command line.
> Second, why should you care in the case
> of SCSI? It seems like a very hard limitation to me, and unlike the
> PCI case it doesn't buy you anything in terms of isolation.
Ah, there is a good reason on this side. I forget the exact details,
but due to the protocol it uses there's some blocksize limit that is
only advertised per vscsi adaptor, whereas it should really be a
per-target quantity (and is different in practice for cdroms and
disks). Of course that's arguably a bug in the vscsi protocol, but we
can't fix that.
--
David Gibson | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_
| _way_ _around_!
http://www.ozlabs.org/~dgibson
