Re: RFC: removing hardcoded "modprobe blktap" in xencommons

[George Dunlap <george.dunlap@eu.citrix.com>]

On 15/07/13 10:38, Wei Liu wrote:
> On Mon, Jul 15, 2013 at 10:34:02AM +0100, Jan Beulich wrote:
>>>>> On 15.07.13 at 10:26, Wei Liu <wei.liu2@citrix.com> wrote:
>>> So my point is, if every driver follows the kernel convention
>>> (open(/dev/XYZ) would cause the module to load automatically) there then
>>> would be minimal changes to libxl. It is unmaintained kernel modules
>>> like blktap makes it not feasible to make things simplier.
>>>
>>> (Wild guess below, correct me if I'm wrong)
>>>
>>> Say, if we could make it just like open(blkback_char_dev), the loading
>>> is just a one-liner and doesn't need to be using libxl's AO interface.
>> Once again - most if not all backends don't fit that model - blktap
>> as said doesn't and neither blkback nor netback don't as they don't
>> create any /dev nodes.
>>
> OK, I get what you mean.
>
>> Backend module loading might be triggerable through xenstore
>> node watches...
>>
> Sure.

If we did something like this, we'd still have to have the modprobes in 
xencommons for older kernels; we'd just have to have a way to disable it 
for newer kernels.

  -George