Xen devel - Nov 2010 - any plans to sopport virtio-serial ?

Hello. Does Xen developers have any plans to support qemu virtio-serial
or may be some ideas to implement some analog?

If no, can You provide my some tech info, what i need to write this?
(xenstore not good for two way communication and have scalability
problems, for example like tdb suport transactions..) Xway not
supporting, XenSocket use another protocol and need to rewrite programs
to support it..

-- 
Vasiliy G Tolstov <v.tolstov@selfip.ru>
Selfip.Ru


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Tue, 2010-11-23 at 08:00 +0000, Vasiliy G Tolstov
wrote:
> Hello. Does Xen developers have any plans to support qemu virtio-serial
> or may be some ideas to implement some analog?
> 
> If no, can You provide my some tech info, what i need to write this?
> (xenstore not good for two way communication and have scalability
> problems, for example like tdb suport transactions..) Xway not
> supporting, XenSocket use another protocol and need to rewrite programs
> to support it..
We are certainly interested in virtio generally and hopefully once the
glue layer for Xen is written support for the specific devices ought to
fall out quite easily.

As I understand it (which is very vague) virtio allows each architecture
to supply device discovery, shared ring and event notification backends
which it then builds the specific device services on top of. I think the
most obvious solution is to do device discovery using XenStore but to do
the shared ring and event notification bits with page grants and event
channels respectively. I was under the impression that virtio was
designed with the possibility of this sort of Xen backend in mind.

Ian.


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Tue, 23 Nov 2010, Vasiliy G Tolstov wrote:
> Hello. Does Xen developers have any plans to support qemu virtio-serial
> or may be some ideas to implement some analog?
> 
> If no, can You provide my some tech info, what i need to write this?
> (xenstore not good for two way communication and have scalability
> problems, for example like tdb suport transactions..) Xway not
> supporting, XenSocket use another protocol and need to rewrite programs
> to support it..
 
You can try to make the current virtio-serial backend in Qemu work on
Xen, that shouldn''t be difficult, but it would only work with HVM
guests.

Otherwise you can try to use a Xen PV serial for your purposes: multiple
PV serials don''t work out of the box, but I have tested them a little
while ago and they do work fine.
You need to hack xl to create multiple PV serials frontend/backend pairs
(see init_console_info in tools/libxl/xl_cmdimpl.c), then you need these
two patches in the guest kernel (unfortunately they are not yet
upstream):

- hvc_xen: support PV on HVM consoles
e76b17aae62d13cebd4cc33415b4f0cc72316345

- hvc_xen: implement multiconsole support
d6edaed101e089b80257bcbb3141308a97ea489e

look at this branch:

git://xenbits.xen.org/people/sstabellini/linux-pvhvm.git 2.6.36-rc6-pvhvm-v8

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Tue, 2010-11-23 at 12:19 +0000, Stefano Stabellini
wrote:
> On Tue, 23 Nov 2010, Vasiliy G Tolstov wrote:
> > Hello. Does Xen developers have any plans to support qemu
virtio-serial
> > or may be some ideas to implement some analog?
> > 
> > If no, can You provide my some tech info, what i need to write this?
> > (xenstore not good for two way communication and have scalability
> > problems, for example like tdb suport transactions..) Xway not
> > supporting, XenSocket use another protocol and need to rewrite
programs
> > to support it..
>  
> You can try to make the current virtio-serial backend in Qemu work on
> Xen, that shouldn''t be difficult, but it would only work with HVM
> guests.
> 
> Otherwise you can try to use a Xen PV serial for your purposes: multiple
> PV serials don''t work out of the box, but I have tested them a
little
> while ago and they do work fine.
> You need to hack xl to create multiple PV serials frontend/backend pairs
> (see init_console_info in tools/libxl/xl_cmdimpl.c), then you need these
> two patches in the guest kernel (unfortunately they are not yet
> upstream):
> 
> - hvc_xen: support PV on HVM consoles
> e76b17aae62d13cebd4cc33415b4f0cc72316345
> 
> - hvc_xen: implement multiconsole support
> d6edaed101e089b80257bcbb3141308a97ea489e
> 
> look at this branch:
> 
> git://xenbits.xen.org/people/sstabellini/linux-pvhvm.git
2.6.36-rc6-pvhvm-v8
Many thank''s. Does this patches come to 2.6.32 stable?
If i patch stable-2.6.32.x with this patches, can i use them? 
And some more question - if i have already running domain, can i create
second serial console via xenapi ? (not xapi)


-- 
Vasiliy G Tolstov <v.tolstov@selfip.ru>
Selfip.Ru


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Tue, 23 Nov 2010, Vasiliy G Tolstov wrote:
> On Tue, 2010-11-23 at 12:19 +0000, Stefano Stabellini wrote:
> > On Tue, 23 Nov 2010, Vasiliy G Tolstov wrote:
> > > Hello. Does Xen developers have any plans to support qemu
virtio-serial
> > > or may be some ideas to implement some analog?
> > > 
> > > If no, can You provide my some tech info, what i need to write
this?
> > > (xenstore not good for two way communication and have scalability
> > > problems, for example like tdb suport transactions..) Xway not
> > > supporting, XenSocket use another protocol and need to rewrite
programs
> > > to support it..
> >  
> > You can try to make the current virtio-serial backend in Qemu work on
> > Xen, that shouldn''t be difficult, but it would only work with
HVM
> > guests.
> > 
> > Otherwise you can try to use a Xen PV serial for your purposes:
multiple
> > PV serials don''t work out of the box, but I have tested them
a little
> > while ago and they do work fine.
> > You need to hack xl to create multiple PV serials frontend/backend
pairs
> > (see init_console_info in tools/libxl/xl_cmdimpl.c), then you need
these
> > two patches in the guest kernel (unfortunately they are not yet
> > upstream):
> > 
> > - hvc_xen: support PV on HVM consoles
> > e76b17aae62d13cebd4cc33415b4f0cc72316345
> > 
> > - hvc_xen: implement multiconsole support
> > d6edaed101e089b80257bcbb3141308a97ea489e
> > 
> > look at this branch:
> > 
> > git://xenbits.xen.org/people/sstabellini/linux-pvhvm.git
2.6.36-rc6-pvhvm-v8
> 
> Many thank''s. Does this patches come to 2.6.32 stable?
Nope.
> If i patch stable-2.6.32.x with this patches, can i use them? 
Yes. I was planning to port those patches to stable-2.6.32.x myself but
I haven''t had the time yet.
> And some more question - if i have already running domain, can i create
> second serial console via xenapi ? (not xapi)
I doubt it but my knowledge of xenapi is very limited...


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Tue, 2010-11-23 at 12:19 +0000, Stefano Stabellini
wrote:
> On Tue, 23 Nov 2010, Vasiliy G Tolstov wrote:
> > Hello. Does Xen developers have any plans to support qemu
virtio-serial
> > or may be some ideas to implement some analog?
> > 
> > If no, can You provide my some tech info, what i need to write this?
> > (xenstore not good for two way communication and have scalability
> > problems, for example like tdb suport transactions..) Xway not
> > supporting, XenSocket use another protocol and need to rewrite
programs
> > to support it..
>  
> You can try to make the current virtio-serial backend in Qemu work on
> Xen, that shouldn''t be difficult, but it would only work with HVM
> guests.
> 
> Otherwise you can try to use a Xen PV serial for your purposes: multiple
> PV serials don''t work out of the box, but I have tested them a
little
> while ago and they do work fine.
> You need to hack xl to create multiple PV serials frontend/backend pairs
> (see init_console_info in tools/libxl/xl_cmdimpl.c), then you need these
> two patches in the guest kernel (unfortunately they are not yet
> upstream):
> 
> - hvc_xen: support PV on HVM consoles
> e76b17aae62d13cebd4cc33415b4f0cc72316345
> 
> - hvc_xen: implement multiconsole support
> d6edaed101e089b80257bcbb3141308a97ea489e
> 
> look at this branch:
> 
> git://xenbits.xen.org/people/sstabellini/linux-pvhvm.git
2.6.36-rc6-pvhvm-v8
I''m clone this repo, but can''t find sha commits like
e76b17aae62d13cebd4cc33415b4f0cc72316345 or
d6edaed101e089b80257bcbb3141308a97ea489e


-- 
Vasiliy G Tolstov <v.tolstov@selfip.ru>
Selfip.Ru


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Tue, 23 Nov 2010, Vasiliy G Tolstov wrote:
> On Tue, 2010-11-23 at 12:19 +0000, Stefano Stabellini wrote:
> > On Tue, 23 Nov 2010, Vasiliy G Tolstov wrote:
> > > Hello. Does Xen developers have any plans to support qemu
virtio-serial
> > > or may be some ideas to implement some analog?
> > > 
> > > If no, can You provide my some tech info, what i need to write
this?
> > > (xenstore not good for two way communication and have scalability
> > > problems, for example like tdb suport transactions..) Xway not
> > > supporting, XenSocket use another protocol and need to rewrite
programs
> > > to support it..
> >  
> > You can try to make the current virtio-serial backend in Qemu work on
> > Xen, that shouldn''t be difficult, but it would only work with
HVM
> > guests.
> > 
> > Otherwise you can try to use a Xen PV serial for your purposes:
multiple
> > PV serials don''t work out of the box, but I have tested them
a little
> > while ago and they do work fine.
> > You need to hack xl to create multiple PV serials frontend/backend
pairs
> > (see init_console_info in tools/libxl/xl_cmdimpl.c), then you need
these
> > two patches in the guest kernel (unfortunately they are not yet
> > upstream):
> > 
> > - hvc_xen: support PV on HVM consoles
> > e76b17aae62d13cebd4cc33415b4f0cc72316345
> > 
> > - hvc_xen: implement multiconsole support
> > d6edaed101e089b80257bcbb3141308a97ea489e
> > 
> > look at this branch:
> > 
> > git://xenbits.xen.org/people/sstabellini/linux-pvhvm.git
2.6.36-rc6-pvhvm-v8
> 
> I''m clone this repo, but can''t find sha commits like
> e76b17aae62d13cebd4cc33415b4f0cc72316345 or
> d6edaed101e089b80257bcbb3141308a97ea489e
 
If you checkout the branch named 2.6.36-rc6-pvhvm-v8, the two commits in
question are the last two...

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Tue, 2010-11-23 at 15:07 +0000, Stefano Stabellini
wrote:
> On Tue, 23 Nov 2010, Vasiliy G Tolstov wrote:
> > On Tue, 2010-11-23 at 12:19 +0000, Stefano Stabellini wrote:
> > > On Tue, 23 Nov 2010, Vasiliy G Tolstov wrote:
> > > > Hello. Does Xen developers have any plans to support qemu
virtio-serial
> > > > or may be some ideas to implement some analog?
> > > > 
> > > > If no, can You provide my some tech info, what i need to
write this?
> > > > (xenstore not good for two way communication and have
scalability
> > > > problems, for example like tdb suport transactions..) Xway
not
> > > > supporting, XenSocket use another protocol and need to
rewrite programs
> > > > to support it..
> > >  
> > > You can try to make the current virtio-serial backend in Qemu
work on
> > > Xen, that shouldn''t be difficult, but it would only work
with HVM
> > > guests.
> > > 
> > > Otherwise you can try to use a Xen PV serial for your purposes:
multiple
> > > PV serials don''t work out of the box, but I have tested
them a little
> > > while ago and they do work fine.
> > > You need to hack xl to create multiple PV serials
frontend/backend pairs
> > > (see init_console_info in tools/libxl/xl_cmdimpl.c), then you
need these
> > > two patches in the guest kernel (unfortunately they are not yet
> > > upstream):
> > > 
> > > - hvc_xen: support PV on HVM consoles
> > > e76b17aae62d13cebd4cc33415b4f0cc72316345
> > > 
> > > - hvc_xen: implement multiconsole support
> > > d6edaed101e089b80257bcbb3141308a97ea489e
> > > 
> > > look at this branch:
> > > 
> > > git://xenbits.xen.org/people/sstabellini/linux-pvhvm.git
2.6.36-rc6-pvhvm-v8
> > 
> > I''m clone this repo, but can''t find sha commits like
> > e76b17aae62d13cebd4cc33415b4f0cc72316345 or
> > d6edaed101e089b80257bcbb3141308a97ea489e
>  
> If you checkout the branch named 2.6.36-rc6-pvhvm-v8, the two commits in
> question are the last two...
Yes, sorry, i''m try to find it into master. Ok. I''m try to use
them.

-- 
Vasiliy G Tolstov <v.tolstov@selfip.ru>
Selfip.Ru


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Tue, 2010-11-23 at 15:07 +0000, Stefano Stabellini
wrote:
> On Tue, 23 Nov 2010, Vasiliy G Tolstov wrote:
> > On Tue, 2010-11-23 at 12:19 +0000, Stefano Stabellini wrote:
> > > On Tue, 23 Nov 2010, Vasiliy G Tolstov wrote:
> > > > Hello. Does Xen developers have any plans to support qemu
virtio-serial
> > > > or may be some ideas to implement some analog?
> > > > 
> > > > If no, can You provide my some tech info, what i need to
write this?
> > > > (xenstore not good for two way communication and have
scalability
> > > > problems, for example like tdb suport transactions..) Xway
not
> > > > supporting, XenSocket use another protocol and need to
rewrite programs
> > > > to support it..
> > >  
> > > You can try to make the current virtio-serial backend in Qemu
work on
> > > Xen, that shouldn''t be difficult, but it would only work
with HVM
> > > guests.
> > > 
> > > Otherwise you can try to use a Xen PV serial for your purposes:
multiple
> > > PV serials don''t work out of the box, but I have tested
them a little
> > > while ago and they do work fine.
> > > You need to hack xl to create multiple PV serials
frontend/backend pairs
> > > (see init_console_info in tools/libxl/xl_cmdimpl.c), then you
need these
> > > two patches in the guest kernel (unfortunately they are not yet
> > > upstream):
> > > 
> > > - hvc_xen: support PV on HVM consoles
> > > e76b17aae62d13cebd4cc33415b4f0cc72316345
> > > 
> > > - hvc_xen: implement multiconsole support
> > > d6edaed101e089b80257bcbb3141308a97ea489e
> > > 
> > > look at this branch:
> > > 
> > > git://xenbits.xen.org/people/sstabellini/linux-pvhvm.git
2.6.36-rc6-pvhvm-v8
> > 
> > I''m clone this repo, but can''t find sha commits like
> > e76b17aae62d13cebd4cc33415b4f0cc72316345 or
> > d6edaed101e089b80257bcbb3141308a97ea489e
>  
> If you checkout the branch named 2.6.36-rc6-pvhvm-v8, the two commits in
> question are the last two...
> 
Stefano, is there this branch tested for live migration issues? For
example, if i compile kernel with config, that working on
stable-2.6.32.x does it work on pvhvm branch?

-- 
Vasiliy G Tolstov <v.tolstov@selfip.ru>
Selfip.Ru


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Tue, 23 Nov 2010, Vasiliy G Tolstov wrote:
> Stefano, is there this branch tested for live migration issues? For
> example, if i compile kernel with config, that working on
> stable-2.6.32.x does it work on pvhvm branch?
 
PV on HVM should work on stable-2.6.32.x but I received a bug
report on live migration issues precisely on that kernel, so if it
doesn''t work for you I wouldn''t be surprised.

Alternatively I suggest you keep using 2.6.36-rc6-pvhvm-v8.


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Wed, 2010-11-24 at 12:55 +0000, Stefano Stabellini
wrote:
> On Tue, 23 Nov 2010, Vasiliy G Tolstov wrote:
> > Stefano, is there this branch tested for live migration issues? For
> > example, if i compile kernel with config, that working on
> > stable-2.6.32.x does it work on pvhvm branch?
>  
> PV on HVM should work on stable-2.6.32.x but I received a bug
> report on live migration issues precisely on that kernel, so if it
> doesn''t work for you I wouldn''t be surprised.
> 
> Alternatively I suggest you keep using 2.6.36-rc6-pvhvm-v8.
> 
Ok. I''m try to use pvhvm and deeper testing it. Does pvhvm branch pull
changes from stable-2.6.32.x ?


-- 
Vasiliy G Tolstov <v.tolstov@selfip.ru>
Selfip.Ru


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Wed, 24 Nov 2010, Vasiliy G Tolstov wrote:
> On Wed, 2010-11-24 at 12:55 +0000, Stefano Stabellini wrote:
> > On Tue, 23 Nov 2010, Vasiliy G Tolstov wrote:
> > > Stefano, is there this branch tested for live migration issues?
For
> > > example, if i compile kernel with config, that working on
> > > stable-2.6.32.x does it work on pvhvm branch?
> >  
> > PV on HVM should work on stable-2.6.32.x but I received a bug
> > report on live migration issues precisely on that kernel, so if it
> > doesn''t work for you I wouldn''t be surprised.
> > 
> > Alternatively I suggest you keep using 2.6.36-rc6-pvhvm-v8.
> > 
> 
> Ok. I''m try to use pvhvm and deeper testing it. Does pvhvm branch
pull
> changes from stable-2.6.32.x ?
it is the other way around: stable-2.6.32.x pulls changes from the pvhvm
branch, so you should just use stable-2.6.32.x

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Wed, 2010-11-24 at 12:59 +0000, Stefano Stabellini
wrote:
> On Wed, 24 Nov 2010, Vasiliy G Tolstov wrote:
> > On Wed, 2010-11-24 at 12:55 +0000, Stefano Stabellini wrote:
> > > On Tue, 23 Nov 2010, Vasiliy G Tolstov wrote:
> > > > Stefano, is there this branch tested for live migration
issues? For
> > > > example, if i compile kernel with config, that working on
> > > > stable-2.6.32.x does it work on pvhvm branch?
> > >  
> > > PV on HVM should work on stable-2.6.32.x but I received a bug
> > > report on live migration issues precisely on that kernel, so if
it
> > > doesn''t work for you I wouldn''t be surprised.
> > > 
> > > Alternatively I suggest you keep using 2.6.36-rc6-pvhvm-v8.
> > > 
> > 
> > Ok. I''m try to use pvhvm and deeper testing it. Does pvhvm
branch pull
> > changes from stable-2.6.32.x ?
> 
> it is the other way around: stable-2.6.32.x pulls changes from the pvhvm
> branch, so you should just use stable-2.6.32.x
Hmm. Ok this means, that a can use stable-2.6.32.x and get second serial
console support in domU ?

-- 
Vasiliy G Tolstov <v.tolstov@selfip.ru>
Selfip.Ru


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Hi,
> As I understand it (which is very vague) virtio allows each architecture
> to supply device discovery, shared ring and event notification backends
> which it then builds the specific device services on top of.
Yes. I think there are three backends now: pci, lguest, s390.
> I think the
> most obvious solution is to do device discovery using XenStore but to do
> the shared ring and event notification bits with page grants and event
> channels respectively. I was under the impression that virtio was
> designed with the possibility of this sort of Xen backend in mind.
Yes, that should be possible.

Another option would be to just use the pci backend.  As I understand it 
nowdays pci is available even in pv guests for pci device passthrough, 
so in theory this should work fine for both pv and hvm guests ...

cheers,
   Gerd


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Wed, 24 Nov 2010, Vasiliy G Tolstov wrote:
> On Wed, 2010-11-24 at 12:59 +0000, Stefano Stabellini wrote:
> > On Wed, 24 Nov 2010, Vasiliy G Tolstov wrote:
> > > On Wed, 2010-11-24 at 12:55 +0000, Stefano Stabellini wrote:
> > > > On Tue, 23 Nov 2010, Vasiliy G Tolstov wrote:
> > > > > Stefano, is there this branch tested for live migration
issues? For
> > > > > example, if i compile kernel with config, that working
on
> > > > > stable-2.6.32.x does it work on pvhvm branch?
> > > >  
> > > > PV on HVM should work on stable-2.6.32.x but I received a
bug
> > > > report on live migration issues precisely on that kernel, so
if it
> > > > doesn''t work for you I wouldn''t be
surprised.
> > > > 
> > > > Alternatively I suggest you keep using 2.6.36-rc6-pvhvm-v8.
> > > > 
> > > 
> > > Ok. I''m try to use pvhvm and deeper testing it. Does
pvhvm branch pull
> > > changes from stable-2.6.32.x ?
> > 
> > it is the other way around: stable-2.6.32.x pulls changes from the
pvhvm
> > branch, so you should just use stable-2.6.32.x
> 
> Hmm. Ok this means, that a can use stable-2.6.32.x and get second serial
> console support in domU ?
I see what you mean now.
stable-2.6.32.x does not support multiple PV serials but the pvhvm
branch does.

So you have to use the pvhvm branch on Jeremy''s tree or
2.6.36-rc6-pvhvm-v8 on my tree.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Wed, 2010-11-24 at 13:12 +0000, Stefano Stabellini
wrote:
> On Wed, 24 Nov 2010, Vasiliy G Tolstov wrote:
> > On Wed, 2010-11-24 at 12:59 +0000, Stefano Stabellini wrote:
> > > On Wed, 24 Nov 2010, Vasiliy G Tolstov wrote:
> > > > On Wed, 2010-11-24 at 12:55 +0000, Stefano Stabellini wrote:
> > > > > On Tue, 23 Nov 2010, Vasiliy G Tolstov wrote:
> > > > > > Stefano, is there this branch tested for live
migration issues? For
> > > > > > example, if i compile kernel with config, that
working on
> > > > > > stable-2.6.32.x does it work on pvhvm branch?
> > > > >  
> > > > > PV on HVM should work on stable-2.6.32.x but I received
a bug
> > > > > report on live migration issues precisely on that
kernel, so if it
> > > > > doesn''t work for you I wouldn''t be
surprised.
> > > > > 
> > > > > Alternatively I suggest you keep using
2.6.36-rc6-pvhvm-v8.
> > > > > 
> > > > 
> > > > Ok. I''m try to use pvhvm and deeper testing it.
Does pvhvm branch pull
> > > > changes from stable-2.6.32.x ?
> > > 
> > > it is the other way around: stable-2.6.32.x pulls changes from
the pvhvm
> > > branch, so you should just use stable-2.6.32.x
> > 
> > Hmm. Ok this means, that a can use stable-2.6.32.x and get second
serial
> > console support in domU ?
> 
> I see what you mean now.
> stable-2.6.32.x does not support multiple PV serials but the pvhvm
> branch does.
> 
> So you have to use the pvhvm branch on Jeremy''s tree or
> 2.6.36-rc6-pvhvm-v8 on my tree.
Thank You for answers =).

-- 
Vasiliy G Tolstov <v.tolstov@selfip.ru>
Selfip.Ru


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Wed, 24 Nov 2010, Gerd Hoffmann wrote:
>    Hi,
> 
> > As I understand it (which is very vague) virtio allows each
architecture
> > to supply device discovery, shared ring and event notification
backends
> > which it then builds the specific device services on top of.
> 
> Yes. I think there are three backends now: pci, lguest, s390.
> 
> > I think the
> > most obvious solution is to do device discovery using XenStore but to
do
> > the shared ring and event notification bits with page grants and event
> > channels respectively. I was under the impression that virtio was
> > designed with the possibility of this sort of Xen backend in mind.
> 
> Yes, that should be possible.
> 
> Another option would be to just use the pci backend.  As I understand it 
> nowdays pci is available even in pv guests for pci device passthrough, 
> so in theory this should work fine for both pv and hvm guests ...
For PV guests the pci bus is only available through the pcifront driver,
it wouldn''t interact with qemu at all.
But it should work fine for PV on HVM guests.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Wed, 2010-11-24 at 12:55 +0000, Stefano Stabellini
wrote:
> On Tue, 23 Nov 2010, Vasiliy G Tolstov wrote:
> > Stefano, is there this branch tested for live migration issues? For
> > example, if i compile kernel with config, that working on
> > stable-2.6.32.x does it work on pvhvm branch?
>  
> PV on HVM should work on stable-2.6.32.x but I received a bug
> report on live migration issues precisely on that kernel, so if it
> doesn''t work for you I wouldn''t be surprised.
> 
> Alternatively I suggest you keep using 2.6.36-rc6-pvhvm-v8.
> 
domU compiled with this config (in attach) silently destroyed after
start. Does i need to modify python xm create to start this domain?

domU config:

kernel= "/mnt/storage/repos/test/vmlinuz-pvhvm"
name = test21-388"
vif = [ ''mac=00:16:3e:00:05:34,ip=188.127.245.186'' ]
disk = [''file:/mnt/storage/sas/21-718,xvda,w'']

memory = 8192
maxmem = 8192

vcpus = 2
maxvcpus = 2
cpu_cap = 200
cpu_weight = 8192

extra = '' root=/dev/xvda1 tmem=1 console=hvc0 earlyprintk=xen debug
loglevel=10 ''



xm dmesg:
(XEN) cpupool_add_domain(dom=13,pool=0) n_dom 5
(XEN) tmem: all pools frozen for all domains
(XEN) tmem: all pools thawed for all domains
(XEN) cpupool_rm_domain(dom=13,pool=0) n_dom 4
(XEN) mm.c:2562:d0 Unknown domain ''13''
(XEN) mm.c:2562:d0 Unknown domain ''13''
(XEN) mm.c:2562:d0 Unknown domain ''13''
(XEN) mm.c:2562:d0 Unknown domain ''13''


qemu-dm:
domid: 13
Warning: vlan 0 is not connected to host network
Using xvda for guest''s hda
Using xvdb for guest''s hdb
Watching /local/domain/0/device-model/13/logdirty/cmd
Watching /local/domain/0/device-model/13/command
char device redirected to /dev/pts/2
/usr/src/packages/BUILD/xen-4.0.0-testing/tools/ioemu-dir/hw/xen_blktap.c:715:
Init blktap pipes
xen be: console-0: xen be: console-0: connect() failed
connect() failed
Log-dirty: no command yet.
xen be: console-0: xen be: console-0: connect() failed
connect() failed
xen be: console-0: xen be: console-0: connect() failed
connect() failed
xen be: console-0: xen be: console-0: connect() failed
connect() failed


-- 
Vasiliy G Tolstov <v.tolstov@selfip.ru>
Selfip.Ru


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Wed, 2010-11-24 at 19:54 +0300, Vasiliy G Tolstov
wrote:
> On Wed, 2010-11-24 at 12:55 +0000, Stefano Stabellini wrote:
> > On Tue, 23 Nov 2010, Vasiliy G Tolstov wrote:
> > > Stefano, is there this branch tested for live migration issues?
For
> > > example, if i compile kernel with config, that working on
> > > stable-2.6.32.x does it work on pvhvm branch?
> >  
> > PV on HVM should work on stable-2.6.32.x but I received a bug
> > report on live migration issues precisely on that kernel, so if it
> > doesn''t work for you I wouldn''t be surprised.
> > 
> > Alternatively I suggest you keep using 2.6.36-rc6-pvhvm-v8.
> > 
Stefano, i''m complete investigate, that last two commits, breaks domain
creating (use 2.6.36-rc6-pvhvm-v8). If i roll back this patches, xm
create construct domains and boot. If apply, domain after creating
destroyed.

>From e76b17aae62d13cebd4cc33415b4f0cc72316345 Mon Sep 17 00:00:00 2001
From: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Date: Thu, 5 Aug 2010 16:59:46 +0100
Subject: [PATCH 1/2] hvc_xen: support PV on HVM consoles

>From d6edaed101e089b80257bcbb3141308a97ea489e Mon Sep 17 00:00:00 2001
From: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Date: Mon, 9 Aug 2010 14:20:55 +0100
Subject: [PATCH 2/2] hvc_xen: implement multiconsole support

This patch implements support for multiple consoles:
consoles other than the first one are setup using the traditional xenbus
and grant-table based mechanism.
We use a list to keep track of the allocated consoles, we don''t
expect too many of them anyway.



-- 
Vasiliy G Tolstov <v.tolstov@selfip.ru>
Selfip.Ru


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Thu, 25 Nov 2010, Vasiliy G Tolstov wrote:
> On Wed, 2010-11-24 at 19:54 +0300, Vasiliy G Tolstov wrote:
> > On Wed, 2010-11-24 at 12:55 +0000, Stefano Stabellini wrote:
> > > On Tue, 23 Nov 2010, Vasiliy G Tolstov wrote:
> > > > Stefano, is there this branch tested for live migration
issues? For
> > > > example, if i compile kernel with config, that working on
> > > > stable-2.6.32.x does it work on pvhvm branch?
> > >  
> > > PV on HVM should work on stable-2.6.32.x but I received a bug
> > > report on live migration issues precisely on that kernel, so if
it
> > > doesn''t work for you I wouldn''t be surprised.
> > > 
> > > Alternatively I suggest you keep using 2.6.36-rc6-pvhvm-v8.
> > > 
> 
> Stefano, i''m complete investigate, that last two commits, breaks
domain
> creating (use 2.6.36-rc6-pvhvm-v8). If i roll back this patches, xm
> create construct domains and boot. If apply, domain after creating
> destroyed.
> 
I think I found and fixed the issue, try to clone again
2.6.37-rc4-pvhvm-pvconsole and this time the guest should boot.
However I believe that pv consoles for hvm guests won''t still work
properly with xend.
If you want to try two pv consoles with both pv and hvm guests apply
the appended patch to xen-unstable and use xl to start a guest:

---

diff -r 8420b82c22c2 tools/libxl/xl_cmdimpl.c
--- a/tools/libxl/xl_cmdimpl.c	Tue Nov 30 11:34:08 2010 +0000
+++ b/tools/libxl/xl_cmdimpl.c	Tue Dec 14 17:25:34 2010 +0000
@@ -1666,6 +1666,12 @@ start:
 
         init_console_info(&console, 0, &state);
         console.domid = domid;
+        console.consback = LIBXL_CONSBACK_IOEMU;
+        libxl_device_console_add(&ctx, domid, &console);
+        libxl_device_console_destroy(&console);
+        init_console_info(&console, 1, NULL);
+        console.domid = domid;
+        console.consback = LIBXL_CONSBACK_IOEMU;
         libxl_device_console_add(&ctx, domid, &console);
         libxl_device_console_destroy(&console);
 
@@ -1686,13 +1692,17 @@ start:
 
         init_console_info(&console, 0, &state);
         console.domid = domid;
-        if (d_config.num_vfbs)
-             console.consback = LIBXL_CONSBACK_IOEMU;
+        console.consback = LIBXL_CONSBACK_IOEMU;
         libxl_device_console_add(&ctx, domid, &console);
         libxl_device_console_destroy(&console);
 
-        if (d_config.num_vfbs)
-            libxl_create_xenpv_qemu(&ctx, domid, d_config.vfbs,
&dm_starting);
+        init_console_info(&console, 1, NULL);
+        console.domid = domid;
+        console.consback = LIBXL_CONSBACK_IOEMU;
+        libxl_device_console_add(&ctx, domid, &console);
+        libxl_device_console_destroy(&console);
+
+        libxl_create_xenpv_qemu(&ctx, domid, d_config.vfbs,
&dm_starting);
     }
 
     if (dm_starting)

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel