Peter Teoh wrote:
2009/10/11 andi <andi.platschek@xxxxxxxxx>:
Ozan Türkyılmaz wrote:
hey all,
it's not really kernel reletad but, kernel peaple should be trasted
with hardware questions.
my question is this: today a lot of CPUs (x36 based ones) come with
visualization extensions and
a lot of visualization software supports this extensions. my
question is that this visualization extensions
can really make difference of performance of guest OS.
ps: i got AMD 64 cpu with AMD-V extensions.
I think, what you are talking about is virtualization, not visualization?
And, yes these extensions really speed up virtualization, the reason for
this is, that a guest OS that
is running in Ring3 (unprivileged mode) is not allowed to do the priviledged
well....i am not sure of all scenario...but i can describe what i know....
You are right, there are different approaches, and I just started to
look into
virtualization... I just found a blog on that topic:
http://mrpointy.wordpress.com/2009/05/12/is-kvm-a-type-1-or-a-type-2/
This article uses the terms type1 and type2 hypervisor, where in the
type1 the guest
runs in ring1 and in type2 it runs in ring3. So both variants are in use...
Say VMWare. In normal OS only ring 0 and 3 are used. So VMware
used ring 1 to emulate the guest's higher privilege mode, and ring 3
for the application. Same design with Xen, Lguest too. But if u
I thought lguest runs in ring3 too, but a look at the documentation
shows, that you are right, it runs
the guest in ring0:
__init void lguest_init(void)
{
/* We're under lguest. */
pv_info.name = "lguest";
/* Paravirt is enabled. */
pv_info.paravirt_enabled = 1;
/* We're running at privilege level 1, not 0 as normal. */
pv_info.kernel_rpl = 1;
/* Everyone except Xen runs with this set. */
pv_info.shared_kernel_pmd = 1;
are running in HVM mode, a higher privilege than ring 0 (called ring
-1) is used for the hypervisor (or host) OS, and ring 0 for the guest
OS. but QEMU may be different....not sure....possibly the same
design with UML too....which is emulating everything in ring 3???? not
sure of details....
I think that UML might be a little different. afaik they introduced a
own architecture (arch/um)
in order to replace all the hardware dependent stuff by code that can be
run in userspace.
(But I might be totally wrong here!!)
Anyway, if you want to know how big the difference in speed is, why not
run the VMM of your choice
once without and once with use of the hardware support? should not be
too hard to see a difference.
greetings,
andi
operations. If you do not have
a hardware support for virtualization, these priviledged operations have to
be emulated in software, which
is really slow.
regards,
andi
--
To unsubscribe from this list: send an email with
"unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx
Please read the FAQ at http://kernelnewbies.org/FAQ
--
To unsubscribe from this list: send an email with
"unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx
Please read the FAQ at http://kernelnewbies.org/FAQ
