Hi Avi,
On Tue, May 24, 2011 at 5:37 PM, Avi Kivity <avi@xxxxxxxxxx> wrote:
> Another thing, I believe reuse of actual kernel code has limited utility.
OK, I don't agree with that.
On Tue, May 24, 2011 at 5:37 PM, Avi Kivity <avi@xxxxxxxxxx> wrote:
> The kernel has vast infrastructure and porting all of it to userspace would
> be a huge undertaking. It's pretty common for code to use printk(),
> kmalloc(), slab caches, rcu, various locks, per-cpu variables. Are you
> going to port all of it? and maintain it when it changes?
We will port whatever is useful and makes sense. Nothing new here -
we're following perf's lead, that's all.
As for the specific issues you mention, only RCU and locking
mechanisms seem like something that are not trivial to port (although
userspace RCU already exists).
[ Porting lockdep to userspace would be one cool feature, though. ]
On Tue, May 24, 2011 at 5:37 PM, Avi Kivity <avi@xxxxxxxxxx> wrote:
> It's a lot easier to use the native userspace stacks.
We will be using userspace libs where appropriate (like with libvnc).
Pekka
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
