| |
|
 |
|
 |
|
|
|
| |
|
|
xen-users
RE: [Xen-users] blktap/blktap2
>
>
>> What's the issue with blktap that precludes it from being added to the
>> kernel?
>>
>
>Just to make sure there's no confusion, we're talking about blktap,
>which is different from blkback/blkfront. Both blkback and blkfront are
>included in the Linux kernel since 3.0.0. These are the drivers that
>make paravirtualized IO possible, and are thus necessary for Xen PV domains.
>
>blktap is used to provide a high performance disk IO interface to
>virtual block devices (see http://wiki.xensource.com/xenwiki/blktap). It
>could, in fact, be implemented completely in userspace, and not as a
>kernel driver. For this reason, blktap is unlikely to ever be accepted
>into the mainline Linux kernel. I'm not a kernel dev myself, so that's
>the best I'll be able to do to explain why it's not going into mainline.
>
>There is talk of refactoring blktap so that it actually does exist as
>pure userspace code, and doesn't require a kernel module. I don't know
>of any concrete plans to actually do this in the near future, since it
>works just fine as an out-of-tree module, but in theory it would be
>possible to move it to userspace.
>
>Mike
>
Is there some performance hit changing it from kernel space to complete
userspace?
Regards
Matej Zary
_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users
|
|
|
| |
|
Home