virtio implementation?

virtio implementation?

[Gregory Haskins]

Hi Rusty,
  I was curious if there are any implementations of your virtio
substrate for KVM yet?  And if not, I can offer help to get this done as
I have an immediate need for a ring-buffer like PV channel.

Regards,
-Greg


-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/

Re: virtio implementation?

[Dor Laor]

>Hi Rusty,
>  I was curious if there are any implementations of your virtio
>substrate for KVM yet?  And if not, I can offer help to get this done
as
>I have an immediate need for a ring-buffer like PV channel.
>
>Regards,
>-Greg

Actually I was planning of doing it. I'm maintaining our not-merged PV
network drivers branch
[started by Ingo Molnar].
Since they already contain a ring + network logic I though of dropping
the front end 
part and hooking with virtio.	
I also want to add/use virtio layer for the host side - KVM, Xen and
lguest all have Linux host
backend it makes logic to use a generic host drivers for
networking/block device too.
Actually this is my KVM forum presentation.

You're welcome to rip my ring code ;)

btw: Rusty - what do you think of virtio for the host?



-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/

Re: virtio implementation?

[Rusty Russell]

On Tue, 2007-07-17 at 06:31 -0700, Dor Laor wrote:
> >Hi Rusty,
> >  I was curious if there are any implementations of your virtio
> >substrate for KVM yet?  And if not, I can offer help to get this done
> as
> >I have an immediate need for a ring-buffer like PV channel.
> >
> >Regards,
> >-Greg
> 
> Actually I was planning of doing it. I'm maintaining our not-merged PV
> network drivers branch
> [started by Ingo Molnar].
> Since they already contain a ring + network logic I though of dropping
> the front end 
> part and hooking with virtio.	
> I also want to add/use virtio layer for the host side - KVM, Xen and
> lguest all have Linux host
> backend it makes logic to use a generic host drivers for
> networking/block device too.
> Actually this is my KVM forum presentation.

OK, I've been working on GSO for virtio net.  I'll post once I have it
tested.

> You're welcome to rip my ring code ;)
> 
> btw: Rusty - what do you think of virtio for the host?

You mean backend?  For networking it makes a great deal of sense.  For
block it makes far less sense (COW, weird formats, etc).

Cheers,
Rusty.


-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/

Re: virtio implementation?

[Dor Laor]

>On Tue, 2007-07-17 at 06:31 -0700, Dor Laor wrote:
>> >Hi Rusty,
>> >  I was curious if there are any implementations of your virtio
>> >substrate for KVM yet?  And if not, I can offer help to get this
done
>> as
>> >I have an immediate need for a ring-buffer like PV channel.
>> >
>> >Regards,
>> >-Greg
>>
>> Actually I was planning of doing it. I'm maintaining our not-merged
PV
>> network drivers branch
>> [started by Ingo Molnar].
>> Since they already contain a ring + network logic I though of
dropping
>> the front end
>> part and hooking with virtio.
>> I also want to add/use virtio layer for the host side - KVM, Xen and
>> lguest all have Linux host
>> backend it makes logic to use a generic host drivers for
>> networking/block device too.
>> Actually this is my KVM forum presentation.
>
>OK, I've been working on GSO for virtio net.  I'll post once I have it
>tested.
>
>> You're welcome to rip my ring code ;)
>>
>> btw: Rusty - what do you think of virtio for the host?
>
>You mean backend?  For networking it makes a great deal of sense.  For
>block it makes far less sense (COW, weird formats, etc).

I meant networking, but there are more components we can add to the
host:
	- Block device can be directed into blkTap thus making the
result both flexible and zero-copy.
	- backend for virtual bus that drivers virqs (or even event
channels) into the guest
	- There's probably room for more


-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/

Re: virtio implementation?

[Rusty Russell]

On Tue, 2007-07-17 at 23:49 -0700, Dor Laor wrote:
> >On Tue, 2007-07-17 at 06:31 -0700, Dor Laor wrote:
> >> btw: Rusty - what do you think of virtio for the host?
> >
> >You mean backend?  For networking it makes a great deal of sense.  For
> >block it makes far less sense (COW, weird formats, etc).
> 
> I meant networking, but there are more components we can add to the
> host:
> 	- Block device can be directed into blkTap thus making the
> result both flexible and zero-copy.

	Let's avoid the term "zero-copy" since it usually indicates that the
people in the discussion can't count.  Let's further assume that kvm
changes to the lguest model where guest pages are user pages.

	AFAICT the only difference between a userspace host block driver and a
kernel host block driver is now the system call overhead, and the
difficulty of doing efficient write barriers from userspace.

> 	- backend for virtual bus that drivers virqs (or even event
> channels) into the guest

Well, kvm is going to need some method to inject interrupts, yes.
lguest already has one, but it's going to change (Carsten wants lguest
to adopt the same userspace interface as kvm).

I'm a little confused as to what we're actually trying to discuss, to be
honest.

Cheers,
Rusty.


-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/

Re: virtio implementation?

[Avi Kivity]

Rusty Russell wrote:
> On Tue, 2007-07-17 at 23:49 -0700, Dor Laor wrote:
>   
>>> On Tue, 2007-07-17 at 06:31 -0700, Dor Laor wrote:
>>>       
>>>> btw: Rusty - what do you think of virtio for the host?
>>>>         
>>> You mean backend?  For networking it makes a great deal of sense.  For
>>> block it makes far less sense (COW, weird formats, etc).
>>>       
>> I meant networking, but there are more components we can add to the
>> host:
>> 	- Block device can be directed into blkTap thus making the
>> result both flexible and zero-copy.
>>     
>
> 	Let's avoid the term "zero-copy" since it usually indicates that the
> people in the discussion can't count.  

Well, for block you can do zero copy, using O_DIRECT.

> Let's further assume that kvm
> changes to the lguest model where guest pages are user pages.
>   

That's a valid assumption.  The rest of the discussion doesn't depend on
it, does it?

> 	AFAICT the only difference between a userspace host block driver and a
> kernel host block driver is now the system call overhead, and the
> difficulty of doing efficient write barriers from userspace.
>
>   

True.  For mortal people setups where you have at most a few thousand
ops/sec, the overhead is negligible.  In-kernel host drivers make sense
only for 100+ disks.  These people are better off using blockdevs rather
than file, so an in-kernel block device need not support file-based storage.


-- 
Do not meddle in the internals of kernels, for they are subtle and quick to panic.


-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/

Re: virtio implementation?

[Gerd Hoffmann]

Rusty Russell wrote:
> You mean backend?  For networking it makes a great deal of sense.  For
> block it makes far less sense (COW, weird formats, etc).

For block you probably want both:  userspace driver which can handle all
sorts of funny image files, and a kernel driver doing a 1:1 mapping to a
block device (lvm volume).

cheers,
  Gerd

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/

Re: virtio implementation?

[Arnd Bergmann]

On Wednesday 18 July 2007, Gerd Hoffmann wrote:
> Rusty Russell wrote:
> > You mean backend?  For networking it makes a great deal of sense.  For
> > block it makes far less sense (COW, weird formats, etc).
> 
> For block you probably want both:  userspace driver which can handle all
> sorts of funny image files, and a kernel driver doing a 1:1 mapping to a
> block device (lvm volume).
> 

Interestingly, once you have the kernel driver that maps a block device,
you can do most of the useful user scenarios by means of /dev/loop
and/or device mapper.

	Arnd <><

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/

Re: virtio implementation?

[Anthony Liguori]

Arnd Bergmann wrote:
> On Wednesday 18 July 2007, Gerd Hoffmann wrote:
>   
>> Rusty Russell wrote:
>>     
>>> You mean backend?  For networking it makes a great deal of sense.  For
>>> block it makes far less sense (COW, weird formats, etc).
>>>       
>> For block you probably want both:  userspace driver which can handle all
>> sorts of funny image files, and a kernel driver doing a 1:1 mapping to a
>> block device (lvm volume).
>>
>>     
>
> Interestingly, once you have the kernel driver that maps a block device,
> you can do most of the useful user scenarios by means of /dev/loop
> and/or device mapper.
>   

Not quite.  Using device mapper to implement something like qcow turns 
out to be pretty painful.  qcow, for instance, doesn't guarantee that 
blocks will be aligned to a sector boundary.  It also doesn't guarantee 
that blocks are >= PAGE_SIZE.  The other problem is that device mapper 
doesn't very much like to have very large tables so you end up having to 
use the table as a cache.

See dm-userspace for an actual implementation of all of this.  I'm 
beginning to think that a more proper userspace block device would be 
interesting.

http://wiki.xensource.com/xenwiki/DmUserspace

Regards,

Anthony Liguori

> 	Arnd <><
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by DB2 Express
> Download DB2 Express C - the FREE version of DB2 express and take
> control of your XML. No limits. Just data. Click to get it now.
> http://sourceforge.net/powerbar/db2/
> _______________________________________________
> kvm-devel mailing list
> kvm-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
> https://lists.sourceforge.net/lists/listinfo/kvm-devel
>
>   


-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/

Re: virtio implementation?

[Arnd Bergmann]

On Thursday 19 July 2007, Anthony Liguori wrote:
> > Interestingly, once you have the kernel driver that maps a block device,
> > you can do most of the useful user scenarios by means of /dev/loop
> > and/or device mapper.
> 
> Not quite.  Using device mapper to implement something like qcow turns 
> out to be pretty painful. 

Right. Note that I said 'most scenarios', not all ;-)

I don't understand enough about qcow to see why you can't do the same
with simple dm snapshots, but I'm sure there is a reason to have it.

	Arnd <><

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/

Re: virtio implementation?

[Anthony Liguori]

Arnd Bergmann wrote:
> On Thursday 19 July 2007, Anthony Liguori wrote:
>   
>>> Interestingly, once you have the kernel driver that maps a block device,
>>> you can do most of the useful user scenarios by means of /dev/loop
>>> and/or device mapper.
>>>       
>> Not quite.  Using device mapper to implement something like qcow turns 
>> out to be pretty painful. 
>>     
>
> Right. Note that I said 'most scenarios', not all ;-)
>
> I don't understand enough about qcow to see why you can't do the same
> with simple dm snapshots, but I'm sure there is a reason to have it.
>   

The other thing that you would want is "dm-file".  The loopback device 
has some problems with data corruption under certain scenarios.  There 
were patches for dm-file floating around but I haven't seen much 
activity recently.

Regards,

Anthony Liguori

> 	Arnd <><
>
>   


-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/

Re: virtio implementation?

[Avi Kivity]

Arnd Bergmann wrote:
> On Thursday 19 July 2007, Anthony Liguori wrote:
>   
>>> Interestingly, once you have the kernel driver that maps a block device,
>>> you can do most of the useful user scenarios by means of /dev/loop
>>> and/or device mapper.
>>>       
>> Not quite.  Using device mapper to implement something like qcow turns 
>> out to be pretty painful. 
>>     
>
> Right. Note that I said 'most scenarios', not all ;-)
>
> I don't understand enough about qcow to see why you can't do the same
> with simple dm snapshots, but I'm sure there is a reason to have it.
>
>   

The biggest advantages of qcow over dm are that it's a file, not a block
device, so it's administrable as a user, can be transported as a file,
etc.  For non-root setups, it's unbeatable.


-- 
Do not meddle in the internals of kernels, for they are subtle and quick to panic.


-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/

Re: virtio implementation?

[Daniel P. Berrange]

On Thu, Jul 19, 2007 at 01:48:21AM +0200, Arnd Bergmann wrote:
> On Wednesday 18 July 2007, Gerd Hoffmann wrote:
> > Rusty Russell wrote:
> > > You mean backend?  For networking it makes a great deal of sense.  For
> > > block it makes far less sense (COW, weird formats, etc).
> > 
> > For block you probably want both:  userspace driver which can handle all
> > sorts of funny image files, and a kernel driver doing a 1:1 mapping to a
> > block device (lvm volume).
> > 
> 
> Interestingly, once you have the kernel driver that maps a block device,
> you can do most of the useful user scenarios by means of /dev/loop
> and/or device mapper.

If/when the 'DMLoop'[1] target for device mapper gets finished/merged that 
could be a viable alternative to traditional loop devices

Dan.

[1]  http://sources.redhat.com/lvm2/wiki/DMLoop
-- 
|=- Red Hat, Engineering, Emerging Technologies, Boston.  +1 978 392 2496 -=|
|=-           Perl modules: http://search.cpan.org/~danberr/              -=|
|=-               Projects: http://freshmeat.net/~danielpb/               -=|
|=-  GnuPG: 7D3B9505   F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505  -=| 

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/

Re: virtio implementation?

[Gregory Haskins]

On Tue, 2007-07-17 at 06:31 -0700, Dor Laor wrote:

> You're welcome to rip my ring code ;)

Point me at the code, if you would.  I would be happy to use it as a
base. :)




-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/

Re: virtio implementation?

[Dor Laor]

>> You're welcome to rip my ring code ;)
>
>Point me at the code, if you would.  I would be happy to use it as a
>base. :)
>

It's under git://kvm.qumranet.com/home/dor/export/[kvm.git &
kvm-userspace.git]
Enojy, I hope that in a week or two (busy) I'll start sending patches
with cleanup of the 
current code and starting from there integ with virtio.

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/