xen,bproc and grub

xen,bproc and grub

[Adam Sulmicki]

Hello,
	I have recently ported bproc [1] which part of clustermatic [2]
suite to xen-2.0 [3] to run on domain0. It was pretty straightforward,
that it required few trival hacks to the kernel, few sym-links and
remembering to use "ARCH=xen" whenever appropriate.

However, clustermatic does use linuxbios on its nodes for booting and does
not use traditional bios at all. This means that the grub (with its
multiboot support) with its heavy dependencies on PC-BIOS Is out of
window.

Thus I'm trying to figure out how I could boot xen/linux-2.6.8.1-xen0 duo
on nodes. Are there any other options to boot it besides using grub[4][5]?
and preferably does not rquire multiboot [6].

I figure out I that maybe could load xen.gz [7] using FILO, Kexec, or
maybe etherboot but it still would not load linux kernel in turn (because
of the dependency on multiboot).

Anyone knows what are my options here? Ideas?

[1] http://bproc.sourceforge.net/
[2] http://www.clustermatic.org/
[3] http://www.cl.cam.ac.uk/Research/SRG/netos/xen/
[4] grub configuration file entry for xen

	title Xen 2.0 / XenLinux 2.6.8.1
	    kernel (hd0,0)/boot/xen.gz dom0_mem=65536
	    module (hd0,0)/boot/vmlinuz-2.6.8.1-xen0 root=/dev/hda1

[5] grub-multiboot, from info page

 - Command: module file ...
     Load a boot module FILE for a Multiboot format boot image (no
     interpretation of the file contents are made, so the user of this
     command must know what the kernel in question expects). The rest
     of the line is passed as the "module command-line", like the
     `kernel' command. You must load a Multiboot kernel image before
     loading any module. See also *Note modulenounzip::.

[6] http://www.gnu.org/software/grub/manual/multiboot/multiboot.html
    does anyone knows better references to multiboot specs?

[7] /boot/xen.gz

redbull:/etc/clustermatic # file /boot/xen.gz
/boot/xen.gz: gzip compressed data, from Unix, max compression

redbull:/etc/clustermatic # file /tmp/xen
/tmp/xen: ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV), statically linked, stripped

redbull:/etc/clustermatic # readelf -a /tmp/xen
ELF Header:
  Magic:   7f 45 4c 46 01 01 01 00 00 00 00 00 00 00 00 00
  Class:                             ELF32
  Data:                              2's complement, little endian
  Version:                           1 (current)
  OS/ABI:                            UNIX - System V
  ABI Version:                       0
  Type:                              EXEC (Executable file)
  Machine:                           Intel 80386
  Version:                           0x1
  Entry point address:               0x100000
  Start of program headers:          52 (bytes into file)
  Start of section headers:          266608 (bytes into file)
  Flags:                             0x0
  Size of this header:               52 (bytes)
  Size of program headers:           32 (bytes)
  Number of program headers:         1
  Size of section headers:           40 (bytes)
  Number of section headers:         3
  Section header string table index: 2

Section Headers:
  [Nr] Name              Type            Addr     Off    Size   ES Flg Lk Inf Al
  [ 0]                   NULL            00000000 000000 000000 00      0 0  0
  [ 1] .text             PROGBITS        00100000 000080 0410f0 00 WAX  0 0 64
  [ 2] .shstrtab         STRTAB          00000000 0411e8 000011 00      0 0  1
Key to Flags:
  W (write), A (alloc), X (execute), M (merge), S (strings)
  I (info), L (link order), G (group), x (unknown)
  O (extra OS processing required) o (OS specific), p (processor specific)

Program Headers:
  Type           Offset   VirtAddr   PhysAddr   FileSiz MemSiz  Flg Align
  LOAD           0x000080 0x00100000 0x00100000 0x410f0 0x764c0 RWE 0x40

 Section to Segment mapping:
  Segment Sections...
   00     .text

There is no dynamic segment in this file.

There are no relocations in this file.

There are no unwind sections in this file.

No version information found in this file.
redbull:/etc/clustermatic #




-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: xen,bproc and grub

[Ian Pratt]

> I figure out I that maybe could load xen.gz [7] using FILO, Kexec, or
> maybe etherboot but it still would not load linux kernel in turn (because
> of the dependency on multiboot).
> 
> Anyone knows what are my options here? Ideas?

We wrote a patch to add multiboot support to kexec (that's how
Xenoboot works).  It was submitted to the kexec maintainer, but I
don't know whether it was accepted or not.

One tool that might be useful is something to take an image (and
one or more modules) and roll it into something that looks like a
static multiboot layout, then wrap it to look line a linux
kernel.

It's a shame that multiboot hasn't been more widely adopted. 

Ron, what environment does linuxbios start kernels in?

Thanks,
Ian





-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: xen,bproc and grub

[Adam Sulmicki]

> > I figure out I that maybe could load xen.gz [7] using FILO, Kexec, or
> > maybe etherboot but it still would not load linux kernel in turn (because
> > of the dependency on multiboot).
> >
> > Anyone knows what are my options here? Ideas?
>
> We wrote a patch to add multiboot support to kexec (that's how
> Xenoboot works).  It was submitted to the kexec maintainer, but I
> don't know whether it was accepted or not.

heh, a small world. I assume this is the patch you refer to (and that it
is the latest one)

	http://www.tjd.phlegethon.org/software/kexec-multiboot.patch

given all options the kexec is probably the best pick since in a way it
would be direct replacement to "Two Kernel Monte" that I'm using right
now.

> One tool that might be useful is something to take an image (and
> one or more modules) and roll it into something that looks like a
> static multiboot layout, then wrap it to look line a linux
> kernel.

yeah, it sounds like a neat idea. for this matter, does it have to look
like kernel? making it an simple elf image would be somewhat simpler, and
still work with many boot loaders. If nothing else, an intermediate step.



-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: xen,bproc and grub

[Ian Pratt]

> > We wrote a patch to add multiboot support to kexec (that's how
> > Xenoboot works).  It was submitted to the kexec maintainer, but I
> > don't know whether it was accepted or not.
> 
> heh, a small world. I assume this is the patch you refer to (and that it
> is the latest one)
> 
>
> http://www.tjd.phlegethon.org/software/kexec-multiboot.patch

ACK.

> given all options the kexec is probably the best pick since in a way it
> would be direct replacement to "Two Kernel Monte" that I'm using right
> now.

kexec is definitely an improvement over TKM. We've still seen
problems with some drivers refusing to work without a reset
e.g. aacraid.
 
> > One tool that might be useful is something to take an image (and
> > one or more modules) and roll it into something that looks like a
> > static multiboot layout, then wrap it to look line a linux
> > kernel.
> 
> yeah, it sounds like a neat idea. for this matter, does it have to look
> like kernel? making it an simple elf image would be somewhat simpler, and
> still work with many boot loaders. If nothing else, an intermediate step.

I guess so. When I originally thought of the tool the idea was to
be able to use pxeboot. It might also be handy for isolinux too,
though the iso patch to grub works pretty well, unless you have
ancient hardware. 

Ian








-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: xen,bproc and grub

[Tim Deegan]

On Tue, Nov 02, 2004 at 12:05:18AM +0000, Ian Pratt wrote:
> > > One tool that might be useful is something to take an image (and
> > > one or more modules) and roll it into something that looks like a
> > > static multiboot layout, then wrap it to look line a linux
> > > kernel.
> > 
> > yeah, it sounds like a neat idea. for this matter, does it have to look
> > like kernel? making it an simple elf image would be somewhat simpler, and
> > still work with many boot loaders. If nothing else, an intermediate step.
> 
> I guess so. When I originally thought of the tool the idea was to
> be able to use pxeboot. It might also be handy for isolinux too,
> though the iso patch to grub works pretty well, unless you have
> ancient hardware. 

It should be possible to build that kind of tool (at least so it would
work for Xen) starting from what I did in that kexec patch.  I'll have a
go at that and see if it's as straightforward as it looks. :)

Tim.

-- 
Tim Deegan                           (My opinions, not the University's)
Systems Research Group
University of Cambridge Computer Laboratory


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: xen,bproc and grub

[Jacob Gorm Hansen]

Ian Pratt wrote:

> I guess so. When I originally thought of the tool the idea was to
> be able to use pxeboot. It might also be handy for isolinux too,
> though the iso patch to grub works pretty well, unless you have
> ancient hardware. 

If one can use kexec instead of shutdown in unpriv domains, isn't that a 
better solution to the problem of recreating rebooted domains, that is 
currently solved by xend?

Jacob


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: xen,bproc and grub

[Ian Pratt]

> Ian Pratt wrote:
> 
> > I guess so. When I originally thought of the tool the idea was to
> > be able to use pxeboot. It might also be handy for isolinux too,
> > though the iso patch to grub works pretty well, unless you have
> > ancient hardware. 
> 
> If one can use kexec instead of shutdown in unpriv domains, isn't that a 
> better solution to the problem of recreating rebooted domains, that is 
> currently solved by xend?

The current xend approach also works if the domain has crashed,
whereas kexec could only be used for planned reboots. There's
nothing to stop a domain doing kexec itself...

Ian


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: xen,bproc and grub

[Adam Sulmicki]

> > > We wrote a patch to add multiboot support to kexec (that's how
> > > Xenoboot works).  It was submitted to the kexec maintainer, but I
> > > don't know whether it was accepted or not.
> >
> > heh, a small world. I assume this is the patch you refer to (and that it
> > is the latest one)
> >
> >
> > http://www.tjd.phlegethon.org/software/kexec-multiboot.patch

argh, it seems to be using real-mode for booting, and it seems multiboot
specs seems to assume existence of pcbios.

i guess the right way will be just to strop xen.gz of all 16 bit code and
have it loaded and started in 32 bit mode. In that way kexec could treat
xen like elf32-x86 type file.





-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: xen,bproc and grub

[Keir Fraser]

> > > > We wrote a patch to add multiboot support to kexec (that's how
> > > > Xenoboot works).  It was submitted to the kexec maintainer, but I
> > > > don't know whether it was accepted or not.
> > >
> > > heh, a small world. I assume this is the patch you refer to (and that it
> > > is the latest one)
> > >
> > >
> > > http://www.tjd.phlegethon.org/software/kexec-multiboot.patch
> 
> argh, it seems to be using real-mode for booting, and it seems multiboot
> specs seems to assume existence of pcbios.
> 
> i guess the right way will be just to strop xen.gz of all 16 bit code and
> have it loaded and started in 32 bit mode. In that way kexec could treat
> xen like elf32-x86 type file.

Xen is already an elf32-x86 binary. There is no real-mode, nor any
16-bit protected-mode, code in Xen.

 -- Keir


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: xen,bproc and grub

[Ronald G. Minnich]

On Mon, 1 Nov 2004, Ian Pratt wrote:

> Ron, what environment does linuxbios start kernels in?


several. Linux, etherboot, FILO, and so on. Most of these do not support 
multiboot. 

ron


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

RE: xen,bproc and grub

[Ian Pratt]

 
> > > > We wrote a patch to add multiboot support to kexec (that's how 
> > > > Xenoboot works).  It was submitted to the kexec 
> maintainer, but I 
> > > > don't know whether it was accepted or not.
> > > http://www.tjd.phlegethon.org/software/kexec-multiboot.patch
> 
> argh, it seems to be using real-mode for booting, and it 
> seems multiboot specs seems to assume existence of pcbios.
> 
> i guess the right way will be just to strop xen.gz of all 16 
> bit code and have it loaded and started in 32 bit mode. In 
> that way kexec could treat xen like elf32-x86 type file.

I forget the context of what you're trying to do.

With Tim's multiboot patch, kexec is able to 'boot' into xen just fine.

I think Tim has recently been looking at how we could roll a xen and
dom0 image into a static multiboot file that is then wrapped to appear
like a standard linux bzImage. 

Ian


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_idU88&alloc_id\x12065&op=click

RE: xen,bproc and grub

[Adam Sulmicki]

> > > > > We wrote a patch to add multiboot support to kexec (that's how
> > > > > Xenoboot works).  It was submitted to the kexec
> > maintainer, but I
> > > > > don't know whether it was accepted or not.
> > > > http://www.tjd.phlegethon.org/software/kexec-multiboot.patch
> >
> > argh, it seems to be using real-mode for booting, and it
> > seems multiboot specs seems to assume existence of pcbios.
> >
> > i guess the right way will be just to strop xen.gz of all 16
> > bit code and have it loaded and started in 32 bit mode. In
> > that way kexec could treat xen like elf32-x86 type file.
>
> I forget the context of what you're trying to do.
>
> With Tim's multiboot patch, kexec is able to 'boot' into xen just fine.
>
> I think Tim has recently been looking at how we could roll a xen and
> dom0 image into a static multiboot file that is then wrapped to appear
> like a standard linux bzImage.

yeah, kexec+multiboot works just fine for me on the laptop with PCBIOS.

However, when I try kexec+multiboot on a node running LinuxBIOS the end
result is something that looks as if node has rebooted. Pooking around I
traced it down to the fact that it uses real mode. (If I try to load linux
kernel using real-mode and kexec I get the exactly the same reboot, if on
the other hand I try to load the same kernel using 32 bit mode it seems to
work fine).

Either way I'm leaving for SuperComputing tomorrow so it probably will be
few days before I get chance to look more as to what exactly is going
wrong.




-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: xen,bproc and grub

[Tim Deegan]

On Mon, Nov 08, 2004 at 01:45:59AM -0500, Adam Sulmicki wrote:
> yeah, kexec+multiboot works just fine for me on the laptop with PCBIOS.
> 
> However, when I try kexec+multiboot on a node running LinuxBIOS the end
> result is something that looks as if node has rebooted. Pooking around I
> traced it down to the fact that it uses real mode. (If I try to load linux
> kernel using real-mode and kexec I get the exactly the same reboot, if on
> the other hand I try to load the same kernel using 32 bit mode it seems to
> work fine).

The kexec start-a-new-kernel code is the same for multiboot as for
(32-bit) linux (unless you use the --reset-vga flag with kexec --load,
and even then it doesn't leave 32-bit protected mode).  It's quite
surprising that one should work and the other not.

Tim.

-- 
Tim Deegan                           (My opinions, not the University's)
Systems Research Group
University of Cambridge Computer Laboratory


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: xen,bproc and grub

[Adam Sulmicki]

On Mon, 8 Nov 2004, Tim Deegan wrote:

> > However, when I try kexec+multiboot on a node running LinuxBIOS the end
> > result is something that looks as if node has rebooted. Pooking around I
> > traced it down to the fact that it uses real mode. (If I try to load linux
> > kernel using real-mode and kexec I get the exactly the same reboot, if on
> > the other hand I try to load the same kernel using 32 bit mode it seems to
> > work fine).
>
> The kexec start-a-new-kernel code is the same for multiboot as for
> (32-bit) linux (unless you use the --reset-vga flag with kexec --load,
> and even then it doesn't leave 32-bit protected mode).  It's quite
> surprising that one should work and the other not.

Keir Fraser wrote:

> Xen is already an elf32-x86 binary. There is no real-mode, nor any
> 16-bit protected-mode, code in Xen.

bummer, then something else must be going wrong. hmmm.

by the way. I had been looking a bit more into the xircom_cb (cardbus)
driver. It basically does not seem to work under xen0. It seems that it
calls the pci_register_driver() and nothing happens afterwards. Ideas?




-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: xen,bproc and grub

[Tim Deegan]

On Mon, Nov 08, 2004 at 06:28:32AM -0000, Ian Pratt wrote:
> I think Tim has recently been looking at how we could roll a xen and
> dom0 image into a static multiboot file that is then wrapped to appear
> like a standard linux bzImage. 

An alpha release of this tool is now available from 
http://www.cl.cam.ac.uk/~tjd21/mbootpack-0.1a.tar.gz

It should let you wrap up xen, xenolinux-0, and the command lines into a
bzImage file that can be loaded using a standard bootloader.  I've
tested it succesfully with xen+xenolinux images from the 2.0 release, on
a 128MB PII-400, loading the bzImage using GRUB and LILO.  Bug reports
from other configurations are very welcome. :)

Tim.

-- 
Tim Deegan                           (My opinions, not the University's)
Systems Research Group
University of Cambridge Computer Laboratory


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

RE: xen,bproc and grub

[Ian Pratt]

> > I think Tim has recently been looking at how we could roll 
> a xen and 
> > dom0 image into a static multiboot file that is then 
> wrapped to appear 
> > like a standard linux bzImage.
> 
> An alpha release of this tool is now available from 
> http://www.cl.cam.ac.uk/~tjd21/mbootpack-0.1a.tar.gz
> 
> It should let you wrap up xen, xenolinux-0, and the command 
> lines into a bzImage file that can be loaded using a standard 
> bootloader.  I've tested it succesfully with xen+xenolinux 
> images from the 2.0 release, on a 128MB PII-400, loading the 
> bzImage using GRUB and LILO.  Bug reports from other 
> configurations are very welcome. :)

It would be cool if it was possible to override the command line passed
to both Xen and the dom0 kernel. Since lilo/syslinux/pxeboot etc only
allow a single command line to specified, I guess we could distinguish
the xen arguments by prefixing them with xen_ or such like.

What do you think?


Ian


-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8

RE: xen,bproc and grub

[Adam Sulmicki]

any idea if xen works with this cpu?

	CPU: Centaur VIA Samuel 2 stepping 03

linux boots on it just fine, but .. looking at comments in x86_32.S I see
it wants "P6+" cpu. I seem have some difficulty figuring out exactly what
this CPU supoprts but the url below states that if nothing else it should
be compiled as in "march=i586"... but for example if it supports 4MB pages
is not clear to me.

http://www3.sympatico.ca/howlettfamily/epia/epia.html




-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8

RE: xen,bproc and grub

[Ronald G. Minnich]

I think that cpu is an i386 basically ...

ron



-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8

RE: xen,bproc and grub

[Adam Sulmicki]

> any idea if xen works with this cpu?
>
> 	CPU: Centaur VIA Samuel 2 stepping 03
>
> linux boots on it just fine, but .. looking at comments in x86_32.S I see
> it wants "P6+" cpu. I seem have some difficulty figuring out exactly what
> this CPU supoprts but the url below states that if nothing else it should
> be compiled as in "march=i586"... but for example if it supports 4MB pages
> is not clear to me.
>
> http://www3.sympatico.ca/howlettfamily/epia/epia.html

ah, I think I found the datascheet, go to

	http://www.viatech.com/en/products/processors/c3/specs.jsp

then click on "Datasheet (Samuel 2)" what gives you:

	http://www.viatech.com/en/downloads/datasheets/processors/c3_samuel2.zip

so far I see:

	3 - Page Size Extensions (4MB) 0

whereas the x86_32.S seems assume 4MB pages support.




-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8

RE: xen,bproc and grub

[Ronald G. Minnich]

On Mon, 15 Nov 2004, Adam Sulmicki wrote:

> > any idea if xen works with this cpu?
> >
> > 	CPU: Centaur VIA Samuel 2 stepping 03
> >
> > linux boots on it just fine, but .. looking at comments in x86_32.S I see
> > it wants "P6+" cpu. I seem have some difficulty figuring out exactly what
> > this CPU supoprts but the url below states that if nothing else it should
> > be compiled as in "march=i586"... but for example if it supports 4MB pages
> > is not clear to me.
> >
> > http://www3.sympatico.ca/howlettfamily/epia/epia.html
> 
> ah, I think I found the datascheet, go to
> 
> 	http://www.viatech.com/en/products/processors/c3/specs.jsp
> 
> then click on "Datasheet (Samuel 2)" what gives you:
> 
> 	http://www.viatech.com/en/downloads/datasheets/processors/c3_samuel2.zip
> 
> so far I see:
> 
> 	3 - Page Size Extensions (4MB) 0
> 
> whereas the x86_32.S seems assume 4MB pages support.
> 
> 

oh oh!

seems like that assumption needs repair :-)

ron


-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8

Re: xen,bproc and grub

[Keir Fraser]

> > > any idea if xen works with this cpu?
> > >
> > > 	CPU: Centaur VIA Samuel 2 stepping 03
> > >
> > > linux boots on it just fine, but .. looking at comments in x86_32.S I see
> > > it wants "P6+" cpu. I seem have some difficulty figuring out exactly what
> > > this CPU supoprts but the url below states that if nothing else it should
> > > be compiled as in "march=i586"...
> > so far I see:
> > 	3 - Page Size Extensions (4MB) 0
> > whereas the x86_32.S seems assume 4MB pages support.
> oh oh!
> seems like that assumption needs repair :-)

It could certainly be repaired, but all new C3s (based on Nehemia
core) have PSE, and more recent ones (stepping 8+) even have a local
APIC. The Samuel-2 CPUS are rather weedy -- there is not much point in
virtualising them.

 -- Keir


-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8

Re: xen,bproc and grub

[Ronald G. Minnich]

On Tue, 16 Nov 2004, Keir Fraser wrote:

> It could certainly be repaired, but all new C3s (based on Nehemia
> core) have PSE, and more recent ones (stepping 8+) even have a local
> APIC. The Samuel-2 CPUS are rather weedy -- there is not much point in
> virtualising them.
> 

sadly, we hvae 60 of them.

ron


-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8

Re: xen,bproc and grub

[Keir Fraser]

> 
> 
> On Tue, 16 Nov 2004, Keir Fraser wrote:
> 
> > It could certainly be repaired, but all new C3s (based on Nehemia
> > core) have PSE, and more recent ones (stepping 8+) even have a local
> > APIC. The Samuel-2 CPUS are rather weedy -- there is not much point in
> > virtualising them.
> > 
> 
> sadly, we hvae 60 of them.
> 
> ron

Modify boot.S so that if the P6 check fails, rather than halting, you:
 1. create non-PSE pagetables
 2. remove PGE and PSE from mmu_cr4_features

The first of these is the trickiest -- you need to allocate some
memory from somewhere to populate the initial pagetable, but the
start-of-day memory layout is pretty weird.

Or... buy some remotely up-to-date CPUs. :-)

 -- Keir


-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8

Re: xen,bproc and grub

[Adam Sulmicki]

On Tue, 16 Nov 2004, Keir Fraser wrote:

> Modify boot.S so that if the P6 check fails, rather than halting, you:
>  1. create non-PSE pagetables
>  2. remove PGE and PSE from mmu_cr4_features
>
> The first of these is the trickiest -- you need to allocate some
> memory from somewhere to populate the initial pagetable, but the
> start-of-day memory layout is pretty weird.

Ah, at last, I'm through the crash course on paging.

(it still does not fully boot but I think the hardest part -- the
assembler code and no printk() is behind me.)

=============================================================================
VFS: Mounted root (nfs filesystem) readonly.
Freeing unused kernel memory: 100k freed

sh-2.05b# mount /proc

sh-2.05b# cat /proc/cpuinfo

processor       : 0
vendor_id       : CentaurHauls
cpu family      : 6
model           : 7
model name      : VIA Samuel 2
stepping        : 3
cpu MHz         : 533.502
cache size      : 64 KB
fdiv_bug        : no
hlt_bug         : no
f00f_bug        : no
coma_bug        : no
fpu             : yes
fpu_exception   : yes
cpuid level     : 1
wp              : yes
flags           : fpu de tsc msr cx8 mtrr pge mmx 3dnow
bogomips        : 1048.57

sh-2.05b# source /usr/src/cm/k

kexec_load succeeded (multiboot-x86)

sh-2.05b# /sbin/kexec -e

Starting new kernel
 __  __            ____    ___     _          _
 \ \/ /___ _ __   |___ \  / _ \   | |__   ___| |_ __ _
  \  // _ \ '_ \    __) || | | |__| '_ \ / _ \ __/ _` |
  /  \  __/ | | |  / __/ | |_| |__| |_) |  __/ || (_| |
 /_/\_\___|_| |_| |_____(_)___/   |_.__/ \___|\__\__,_|

 http://www.cl.cam.ac.uk/netos/xen
 University of Cambridge Computer Laboratory


 CTRL-A Z for help | 19200 8N1 | NOR | Minicom 2.00.0 | VT102 | Offline
=============================================================================




-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/

Re: xen,bproc and grub

[Adam Sulmicki]

> > > > any idea if xen works with this cpu?
> > > >
> > > > 	CPU: Centaur VIA Samuel 2 stepping 03
> > > >
> > > > linux boots on it just fine, but .. looking at comments in x86_32.S I see
> > > > it wants "P6+" cpu. I seem have some difficulty figuring out exactly what
> > > > this CPU supoprts but the url below states that if nothing else it should
> > > > be compiled as in "march=i586"...
> > > so far I see:
> > > 	3 - Page Size Extensions (4MB) 0
> > > whereas the x86_32.S seems assume 4MB pages support.
> > oh oh!
> > seems like that assumption needs repair :-)
>
> It could certainly be repaired, but all new C3s (based on Nehemia
> core) have PSE, and more recent ones (stepping 8+) even have a local
> APIC. The Samuel-2 CPUS are rather weedy -- there is not much point in
> virtualising them.

I'm sort of wondering if missing PSE is the only issue I would face here.

comparing c3-2 datascheet and xen sources/papers it seems that xen uses
following stuff that are misseed in c3-2

        4mb pages
        fast calls
        page attribute table
        virtual mode extension

maybe some others.. although i'm not sure if any of those are critical for
correct operation of xen.... in particular that PAT seems worriesome.

we already covered PSE, any comments on other features?



-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8

Re: xen,bproc and grub

[Keir Fraser]

> I'm sort of wondering if missing PSE is the only issue I would face here.
> 
> comparing c3-2 datascheet and xen sources/papers it seems that xen uses
> following stuff that are misseed in c3-2
> 
>         4mb pages
>         fast calls
>         page attribute table
>         virtual mode extension
> 
> maybe some others.. although i'm not sure if any of those are critical for
> correct operation of xen.... in particular that PAT seems worriesome.
> 
> we already covered PSE, any comments on other features?
> 

We don't use SYSCALL/SYSENTER, PAT or VME.

 -- Keir


-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8

Re: xen,bproc and grub

[Adam Sulmicki]

> > I'm sort of wondering if missing PSE is the only issue I would face here.
> >
> > comparing c3-2 datascheet and xen sources/papers it seems that xen uses
> > following stuff that are misseed in c3-2
> >
> >         4mb pages
> >         fast calls
> >         page attribute table
> >         virtual mode extension
> >
> > maybe some others.. although i'm not sure if any of those are critical for
> > correct operation of xen.... in particular that PAT seems worriesome.
> >
> > we already covered PSE, any comments on other features?
> >
>
> We don't use SYSCALL/SYSENTER, PAT or VME.

ah, good.




-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8

Re: xen,bproc and grub

[Kip Macy]

My understanding was that PAT would likely be used to allow VMs to use
PSE, is that not correct?

		-Kip

> > We don't use SYSCALL/SYSENTER, PAT or VME.
>
> ah, good.
>
>
>
>
> -------------------------------------------------------
> This SF.Net email is sponsored by: InterSystems CACHE
> FREE OODBMS DOWNLOAD - A multidimensional database that combines
> robust object and relational technologies, making it a perfect match
> for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/xen-devel
>


-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8

Re: xen,bproc and grub

[Keir Fraser]

The two features are completely unrelated.

 -- Keir

> My understanding was that PAT would likely be used to allow VMs to use
> PSE, is that not correct?
> 
> 		-Kip
> 
> > > We don't use SYSCALL/SYSENTER, PAT or VME.
> >
> > ah, good.
> >
> >
> >
> >
> > -------------------------------------------------------
> > This SF.Net email is sponsored by: InterSystems CACHE
> > FREE OODBMS DOWNLOAD - A multidimensional database that combines
> > robust object and relational technologies, making it a perfect match
> > for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8
> > _______________________________________________
> > Xen-devel mailing list
> > Xen-devel@lists.sourceforge.net
> > https://lists.sourceforge.net/lists/listinfo/xen-devel
> >



-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8

Re: xen,bproc and grub

[Kip Macy]

I was told there were security issues with allowing PSE in VMs outright.

				-Kip

>
> The two features are completely unrelated.
>
>  -- Keir
>
> > My understanding was that PAT would likely be used to allow VMs to use
> > PSE, is that not correct?
> >
> > 		-Kip
> >
> > > > We don't use SYSCALL/SYSENTER, PAT or VME.
> > >
> > > ah, good.
> > >
> > >
> > >
> > >
> > > -------------------------------------------------------
> > > This SF.Net email is sponsored by: InterSystems CACHE
> > > FREE OODBMS DOWNLOAD - A multidimensional database that combines
> > > robust object and relational technologies, making it a perfect match
> > > for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8
> > > _______________________________________________
> > > Xen-devel mailing list
> > > Xen-devel@lists.sourceforge.net
> > > https://lists.sourceforge.net/lists/listinfo/xen-devel
> > >
>
>


-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8

Re: xen,bproc and grub

[Keir Fraser]

Ah yes, if VMs use PSE then we were plannign to abuse PAT to provide
some protection for us. However, VMs can *not* currently use PSE at
the moment and, even if they could, we have since discovered that our
abuse of PAT would not work anyway. :-)

 -- Keir

> 
> I was told there were security issues with allowing PSE in VMs outright.
> 
> 				-Kip
> 
> >
> > The two features are completely unrelated.
> >
> >  -- Keir
> >
> > > My understanding was that PAT would likely be used to allow VMs to use
> > > PSE, is that not correct?
> > >
> > > 		-Kip
> > >
> > > > > We don't use SYSCALL/SYSENTER, PAT or VME.
> > > >
> > > > ah, good.
> > > >
> > > >
> > > >
> > > >
> > > > -------------------------------------------------------
> > > > This SF.Net email is sponsored by: InterSystems CACHE
> > > > FREE OODBMS DOWNLOAD - A multidimensional database that combines
> > > > robust object and relational technologies, making it a perfect match
> > > > for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8
> > > > _______________________________________________
> > > > Xen-devel mailing list
> > > > Xen-devel@lists.sourceforge.net
> > > > https://lists.sourceforge.net/lists/listinfo/xen-devel
> > > >
> >
> >



-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8

Re: xen,bproc and grub

[Kip Macy]

Any other possible solutions?


				-Kip


On Tue, 16 Nov 2004, Keir Fraser wrote:

>
> Ah yes, if VMs use PSE then we were plannign to abuse PAT to provide
> some protection for us. However, VMs can *not* currently use PSE at
> the moment and, even if they could, we have since discovered that our
> abuse of PAT would not work anyway. :-)
>
>  -- Keir
>
> >
> > I was told there were security issues with allowing PSE in VMs outright.
> >
> > 				-Kip
> >
> > >
> > > The two features are completely unrelated.
> > >
> > >  -- Keir
> > >
> > > > My understanding was that PAT would likely be used to allow VMs to use
> > > > PSE, is that not correct?
> > > >
> > > > 		-Kip
> > > >
> > > > > > We don't use SYSCALL/SYSENTER, PAT or VME.
> > > > >
> > > > > ah, good.
> > > > >
> > > > >
> > > > >
> > > > >
> > > > > -------------------------------------------------------
> > > > > This SF.Net email is sponsored by: InterSystems CACHE
> > > > > FREE OODBMS DOWNLOAD - A multidimensional database that combines
> > > > > robust object and relational technologies, making it a perfect match
> > > > > for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8
> > > > > _______________________________________________
> > > > > Xen-devel mailing list
> > > > > Xen-devel@lists.sourceforge.net
> > > > > https://lists.sourceforge.net/lists/listinfo/xen-devel
> > > > >
> > >
> > >
>
>
>
> -------------------------------------------------------
> This SF.Net email is sponsored by: InterSystems CACHE
> FREE OODBMS DOWNLOAD - A multidimensional database that combines
> robust object and relational technologies, making it a perfect match
> for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/xen-devel
>


-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8

Re: xen,bproc and grub

[Keir Fraser]

Oh yes. It's only a problem for SMP guests anyway (which are making
excellent progress but not ready for check-in yet). Since we currently
just enclose any suspect code in Xen with a "big fat domain lock"
right now, we couldn't have any problem as only one "virtual cpu" can
update pagetables at any point in time. The PAT trick was looking
forward to the possibility that we could allow vcpus to go at their
pagetables with no locking in Xen required. That now won't be
possible, but I expect that isn't a problem. I just like lock-free
code, is all. :-)

 -- Keir

> Any other possible solutions?
> 
> 
> 				-Kip
> 
> 
> On Tue, 16 Nov 2004, Keir Fraser wrote:
> 
> >
> > Ah yes, if VMs use PSE then we were plannign to abuse PAT to provide
> > some protection for us. However, VMs can *not* currently use PSE at
> > the moment and, even if they could, we have since discovered that our
> > abuse of PAT would not work anyway. :-)
> >
> >  -- Keir
> >
> > >
> > > I was told there were security issues with allowing PSE in VMs outright.
> > >
> > > 				-Kip
> > >
> > > >
> > > > The two features are completely unrelated.
> > > >
> > > >  -- Keir
> > > >
> > > > > My understanding was that PAT would likely be used to allow VMs to use
> > > > > PSE, is that not correct?
> > > > >
> > > > > 		-Kip
> > > > >
> > > > > > > We don't use SYSCALL/SYSENTER, PAT or VME.
> > > > > >
> > > > > > ah, good.
> > > > > >
> > > > > >
> > > > > >
> > > > > >
> > > > > > -------------------------------------------------------
> > > > > > This SF.Net email is sponsored by: InterSystems CACHE
> > > > > > FREE OODBMS DOWNLOAD - A multidimensional database that combines
> > > > > > robust object and relational technologies, making it a perfect match
> > > > > > for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8
> > > > > > _______________________________________________
> > > > > > Xen-devel mailing list
> > > > > > Xen-devel@lists.sourceforge.net
> > > > > > https://lists.sourceforge.net/lists/listinfo/xen-devel
> > > > > >
> > > >
> > > >
> >
> >
> >
> > -------------------------------------------------------
> > This SF.Net email is sponsored by: InterSystems CACHE
> > FREE OODBMS DOWNLOAD - A multidimensional database that combines
> > robust object and relational technologies, making it a perfect match
> > for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8
> > _______________________________________________
> > Xen-devel mailing list
> > Xen-devel@lists.sourceforge.net
> > https://lists.sourceforge.net/lists/listinfo/xen-devel
> >



-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8