Re: [PATCH RFC 00/10] libxl: new hotplug calling convention

["Roger Pau Monné" <roger.pau@citrix.com>]

On 17/01/13 16:57, Ian Campbell wrote:
> On Thu, 2013-01-17 at 15:47 +0000, Roger Pau Monne wrote:
>> On 17/01/13 16:40, Ian Campbell wrote:
>>> On Thu, 2013-01-17 at 15:30 +0000, Roger Pau Monne wrote:
>>>> We still need pdev_path, I would say only internally because once the
>>>> "target" has been attached to the Dom0, pdev_path has the path to the
>>>> block device that will be used in blkback, but the caller doesn't need
>>>> to know about this.
>>>
>>> So in the new scheme both pdev_path are valid and contain different
>>> information?
>>
>> Yes, target would be what's passed by the user, like:
>>
>> "iqn=iqn.1994-04.org.netbsd.iscsi-target:target0,portal=192.168.1.128"
>>
>> And pdev_path would be filled after the device has been attached, and it
>> will be something like:
>>
>> /dev/sdb
> 
> I see. When we login and create /dev/sdb we can't overwrite target? Or
> move pdev_path into some purely internal datastructure?

Yes that's another option, what I'm doing in this series is overwriting
pdev_path, we could just change the name of pdev_path to target (with a
union as you said) and keep doing it (overwriting target with the
physical path of the attached device).

I'm afraid there's no internal disk related structure, only
libxl_device_disk (public) or libxl__device, but libxl__device is too
general, and I don't want to start overloading it with disk specific values.

> 
>> The main reason we need the block device is to be able to call pygrub or
>> to pass it to Qemu.
> 
> I see.
> 
> Ian.
>