Re: CVS commit: src/sys/arch

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi

>
> Module Name:	src
> Committed By:	bouyer
> Date:		Sun Oct 21 15:41:04 UTC 2007
>
> Modified Files:
> 	src/sys/arch/amd64/amd64 [bouyer-xenamd64]: machdep.c
pmap.c
> 	src/sys/arch/xen/conf [bouyer-xenamd64]: files.xen
> 	src/sys/arch/xen/i386 [bouyer-xenamd64]: pmap.c
xpmap.c
> 	src/sys/arch/xen/include [bouyer-xenamd64]: xenfunc.h
xenpmap.h
> 	src/sys/arch/xen/x86 [bouyer-xenamd64]: xenfunc.c
> 	src/sys/arch/xenamd64/amd64 [bouyer-xenamd64]:
xpmap.c
> 	src/sys/arch/xenamd64/conf [bouyer-xenamd64]:
files.xenamd64
> Added Files:
> 	src/sys/arch/xen/x86 [bouyer-xenamd64]: x86_xpmap.c
>
> Log Message:
> Factorise some Xen pmap code in x86_xpmap.c.
> More xpmap_{ptom,mtop} -> xpmap_{ptom,mtop}_masked
>
> The xenamd64 kernel is now good enough to complete a
sysinst  
> install from
> xennet to xbd.
>

Good work, Manuel I'm going to test this kernel.
Can you upload kernel image somewhere ?
>

Regards

Adam.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)

iD8DBQFHG4GRlIxPgX3Go0MRAoGhAJ41QUI9R4qRmdu92x/5zYBsYZ73dQCg
2odg
Kys+CialIHfGiAfu8+imp9A=
=4VUt
-----END PGP SIGNATURE-----

On Sun, Oct 21, 2007 at 06:42:56PM +0200, Adam Hamsik
wrote:
> [...]
> >The xenamd64 kernel is now good enough to complete
a sysinst  
> >install from
> >xennet to xbd.
> >
> 
> Good work, Manuel I'm going to test this kernel.

thanks ! 

> Can you upload kernel image somewhere ?

OK, I'll also take the opportunity to post a status update.

With the code currently in-tree, the system is stable enough
to
install from sysinst over the network to a xbd, but it
occasinally
panics on a make -j4 dependall kernel build: it gets a page
fault trying to
read a PTE entry (although with the last code, I got only
one panic for
a full kernel build) So there's still some bug/race
condition in the pmap code.
There's also a few things that don't work as it should:
- 'systat vm' complains about "The alternate system
clock has died". I don't
  remember what this means exactly, but I suspect it could
be because the
  symbol table is not loaded. This may be fixed with a newer
Xen, or
  using the xentools from the NetBSD package (I'm currently
using a
  fedora 7 as dom0).
- all CPU time is accounted to system or interrupt, there's
no user time
  accounted. Probably missing assembly stubs at some
places.
- It would be nice if someone could test it on a box with
more than
  4GB ram 
- I'm not sure there's no delayed/missed interrupts. Also
linux has code to
  detect interrupt reentrancy and clean up the stack if this
happens,
  to avoid stack overflow. We should probably try to do
something similar.


Short term, the branch should be updated to a HEAD, new that
the yamt-x86pmap
branch has been merged. after that, and once the domU bugs
are tested,
the branch can be merged to HEAD, and I'll start working on
dom0
support.

Kernels from current sources are available from
ftp://asim.lip6.fr/outgoing/bouyer/amd64/
Note that if you use netbsd-INSTALL_XEN3_DOMU.gz to install
a domU, you'll
have to create the xbd devices yourself as there's no xen
devices in
the amd64 MAKEDEV yet (exept the one in the ramdisk in this
kernel).
To build from sources, checkout a full current source tree,
and run
cvs update -dPA -r bouyer-xenamd64
in src/sys

-- 
Manuel Bouyer <bouyerantioche.eu.org>
     NetBSD: 26 ans d'experience feront toujours la
difference
--