Re: [Xen-devel] [PATCH] xl: Support backend domain ID for disks

[Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx>]

Daniel De Graaf writes ("[Xen-devel] [PATCH] xl: Support backend domain ID for 
disks"):
> Allow specification of backend domains for disks, either in the config
> file or via block-attach

This is probably going in the right direction but I have some
questions and observations.

Firstly, much of the rest of the code in libxl assumes that the
pdev_path string can (depending on the backend type) be dereferenced
by libxl.  That is, if libxl is running in dom0, it assumes that the
block device can be dereferenced in dom0.

So for this to work properly I think at least you need to investigate
the backend type selection machinery and the pdev_path validation and
make sure they are somehow disabled.

Having never done driver domains: how does the backend domain know
what it is supposed to be doing ?  Does it just get the pdev_path via
xenstore and do the rest itself ?  Does it get told the backend type ?
What is the resulting xenstore protocol ?

Is this a reason to preserve the arrangement whereby the target of
blkback is set up by a hotplug script, rather than by a script
executed directly by libxl ?


Finally, one other relatively minor thing.  I don't think "backend" is
the appropriate name for "backend domid".  How about "backenddomain" ?
(This may not be compatible with xm I guess...)

Ian.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel