Re: [gentoo-user] gentoo-sources and xen blktap driver?

[Michael Mol <mikemol@gmail.com>]

On Sat, Jan 7, 2012 at 7:36 PM, Pandu Poluan <pandu@poluan.info> wrote:
> On Jan 8, 2012 12:43 AM, "Michael Mol" <mikemol@gmail.com> wrote:
>> On Sat, Jan 7, 2012 at 11:35 AM, Pandu Poluan <pandu@poluan.info> wrote:
>> > On Jan 7, 2012 8:44 PM, "victor romanchuk" <rom@persimplex.net> wrote:
>> >>
>> >> Konstantinos Agouros wrote, at 01/07/2012 03:51 PM:
>> >> > since xen got into the mainstream kernel the way to go is to use
>> >> > gentoo-sources for dom0 and the domUs. However the blktap modules are
>> >> > not
>> >> > there. Is there any way to get this to work?
>> >>
>> >> blktap drivers were excluded from kernel mainline since 3.x, these two
>> >> threads
>> >> from xen-users mailing list might put some light in that context:
>> >>
>> >>
>> >>
>> >> http://old-list-archives.xen.org/archives/html/xen-users/2011-07/msg00637.html
>> >>
>> >>
>> >> http://old-list-archives.xen.org/archives/html/xen-users/2011-10/msg00065.html
>> >>
>> >> the latest sys-kernel/xen-sources containing working blktap (not
>> >> blktap2)
>> >> is
>> >> 2.6.38 (this is buggy from my point of view; i'm still sitting on
>> >> 2.6.34-r5 for
>> >> production installations)
>> >>
>> >
>> > Can someone shed a light on the importance of blktap, i.e., why one
>> > would
>> > want to use it when -- as someone explained in the first email thread
>> > you
>> > gave -- blkfront+blkend is enough for paravirtualization?
>>
>> Reading through the linked threads, it sounds like the benefit stems
>> from being able to shim things in between the front and back ends.
>>
>> You might want that for any number of reasons:
>> * a block encryption layer
>> * a metering layer
>> * a read/write masking layer
>> * an intercept to have the block device exist on (or be mirrored to)
>> on another system.
>>
>> etc.
>>
>
> Ah yes, of course.
>
> One of the threads also mentioned that blktap might be better implemented in
> userspace.

Well, there's that, too. For a while, there's been a long push in
Linux to get anything that could remotely reasonably be done in
userspace out of kernelspace.

-- 
:wq