xen discuss - Jul 2006 - my experiences with Xen until now and a lot of questions

Hi,

now that I got Xen working in 32 Bit Mode on my Ferrari I have some other
questions (see below). Please keep in mind that I''m Solaris Guy and do
not know much about the currentLinux distributions .

regards

Bernd


Xen Web Server
--------------

- I activated the Xen Web server in the Xen configuration file with 
(xend-http-server yes)
  The Xen Webserver (http://localhost:8000) only shows two links to 
localhost:8000  ...

Q: Is this working as designed?


General
-------

- in the /var/adm/messages of the Solaris Dom0 Domain are  lot of messages like

Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0: 74726573
7466615f 6c007265 5f747369
Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0: 32 bytes of
hardware descriptor at address c88becb0:-
Jul 26 22:40:26 unknown bge: [ID 801725 kern.warning] WARNING: bge0:
driver-detected hardware error: bad ring index!
Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0: 74736967
69746172 67006e6f 635f7465
Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0: 7474615f
75626972 00736574 65737973
Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0: 32 bytes of
hardware descriptor at address c88becb0:-
Jul 26 22:40:26 unknown last message repeated 1 time
Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0: PCI address
5f676f6c0074756f packet len 0x5f74 token 0x70756b flags 0x7665 index 0x7075
Jul 26 22:40:26 unknown bge: [ID 801725 kern.warning] WARNING: bge0:
driver-detected hardware error: bad ring index!
Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0: PCI address
887c600 buffer len 0x600 token 0xbcd57855
Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0: c188068a
840fc084 00000201 0fc87539
Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0: 0000bf8d
fb830000 56850f0a e9fffffe
Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0: 598d317e
51be0fff 18758bff 833c468b
Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0: 5fe850c0
89fffb76 c483c845 c04d8b10
Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0: 9090c3c9
89559090 0000b8e5 c0850000

Q: Do I have to care about the WARNING messages like
bge0: driver-detected hardware error: bad ring index! ?

Q: How can I get rid of these messages?


- while copying a big file (700 MB) over the network via ftp 
  the machine is  not usable anymore (the response of the keyboard 
  and the mouse is very, very, slow.) [Tested while the 32 Bit Xen 
  Kernel was booted and one small Linux Domain with 128 MB was running]


- after booting different Solaris Kernel (with and without Xen,
  32 Bit/64 Bit Kernel) the keyboard does not work in the Gnome/CDE 
  Logon Dialog of Solaris (it does work in the GRUB Menu).
  To get around this I need to telnet to the machine and issue a reconfigure
  boot by creating /reconfigure and reboot. After the reconfigure boot the 
  keyboard works again

  The workaround from the bug

  http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=6447269

  seems to fix this issue.


Q: Is this a known bug? Is there a better workaround for this bug?


- in all Linux domains are continuous messages telling that

	Received packet is 16 bytes before head.

Note that I already have "set ip:dohwcksum = 0" in /etc/system of
Solaris
running in the Dom0 domain

"ethtool -K eth0 tx off" in the Linux Domain does not change this
behaviour

The network adapter is a Broadcom LAN adapter; the driver used in the Solaris
Dom0 Domain is the bge driver from snv_43.


driver from
Q: How can I avoid these messages?



Centos Linux Image from the Xen Demo Disk
-----------------------------------------

- the Centos image from the Xen Demo Disk boots fine and get''s an IP
address
after manual issuing "ethtool -k eth0 tx off" the VNC connection to
the
VNC Server in the Linux Domain works

Q: In which (Linux) start script should I add the ethtool command to
   automatically set this while booting the Linux Domain?
   
   
Debian Linux Image from the Xen Demo Disk
-----------------------------------------

- the Debian Image from the Xen Demo Disk boots fine and get''s an IP
address
  I can ping it from the Solaris domain and I can ping the machines on the
  network from within the Linux Domain. All other IP connections (ssh, telnet,
  vnc, etc) do not work
  There is no ethtool to disable the checksum in this Linux Image
  
Q: How can I disable the network checksum in the Debian Linux Image?

 

OpenSUSE Linux Image from the Xen Demo Disk
-------------------------------------------

- the OpenSUSE Image from the Xen Demo Disk boots fine but does not get
  an IP address automatically. I can configure an IP address manually.
  And after issuing "ethtool -K eth2 tx off" the vnc and the other
network
  services work
  
Q: How to configure the IP interface in OpenSUSE automatically (including
   the ethtool command)?
 
 
This message posted from opensolaris.org

Bernd, this is a Ferrari 4000, yes?  Could you give some more details  
about the specific model?

On 26 Jul 2006, at 10:08pm, Bernd Schemmer wrote:
> Xen Web Server
> --------------
>
> - I activated the Xen Web server in the Xen configuration file  
> with  (xend-http-server yes)
>   The Xen Webserver (http://localhost:8000) only shows two links  
> to  localhost:8000  ...
>
> Q: Is this working as designed?
I''ve no experience with this, sorry.
> General
> -------
>
> - in the /var/adm/messages of the Solaris Dom0 Domain are  lot of  
> messages like
>
> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0:  
> 74726573 7466615f 6c007265 5f747369
> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0: 32  
> bytes of hardware descriptor at address c88becb0:-
> Jul 26 22:40:26 unknown bge: [ID 801725 kern.warning] WARNING:  
> bge0: driver-detected hardware error: bad ring index!
> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0:  
> 74736967 69746172 67006e6f 635f7465
> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0:  
> 7474615f 75626972 00736574 65737973
> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0: 32  
> bytes of hardware descriptor at address c88becb0:-
> Jul 26 22:40:26 unknown last message repeated 1 time
> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0:  
> PCI address 5f676f6c0074756f packet len 0x5f74 token 0x70756b flags  
> 0x7665 index 0x7075
> Jul 26 22:40:26 unknown bge: [ID 801725 kern.warning] WARNING:  
> bge0: driver-detected hardware error: bad ring index!
> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0:  
> PCI address 887c600 buffer len 0x600 token 0xbcd57855
> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0:  
> c188068a 840fc084 00000201 0fc87539
> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0:  
> 0000bf8d fb830000 56850f0a e9fffffe
> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0:  
> 598d317e 51be0fff 18758bff 833c468b
> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0:  
> 5fe850c0 89fffb76 c483c845 c04d8b10
> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0:  
> 9090c3c9 89559090 0000b8e5 c0850000
>
> Q: Do I have to care about the WARNING messages like
> bge0: driver-detected hardware error: bad ring index! ?
Yes, they look bad.  Is there anything ''interesting'' about
your
network configuration?  (jumbo frames, perhaps?)
> Q: How can I get rid of these messages?
They suggest a bug, though I''m not sure where (could be the driver,  
could be somewhere else).  They''ll occur until we fix the bug or find  
a workaround.
> - while copying a big file (700 MB) over the network via ftp
>   the machine is  not usable anymore (the response of the keyboard
>   and the mouse is very, very, slow.) [Tested while the 32 Bit Xen
>   Kernel was booted and one small Linux Domain with 128 MB was  
> running]
Whilst network performance of the domU is not good today, I''ve not  
seen this kind of problem with dom0 networking.  It may well be  
related to the bge problem above.
> - after booting different Solaris Kernel (with and without Xen,
>   32 Bit/64 Bit Kernel) the keyboard does not work in the Gnome/CDE
>   Logon Dialog of Solaris (it does work in the GRUB Menu).
>   To get around this I need to telnet to the machine and issue a  
> reconfigure
>   boot by creating /reconfigure and reboot. After the reconfigure  
> boot the
>   keyboard works again
>
>   The workaround from the bug
>
>   http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=6447269
>
>   seems to fix this issue.
>
>
> Q: Is this a known bug? Is there a better workaround for this bug?
Given that the bug is in the database, I''d say that this is  
known :-)  Which workaround did you use?  (USB keyboard or ddi- 
forceattach.)
> - in all Linux domains are continuous messages telling that
>
> 	Received packet is 16 bytes before head.
Interesting - the first time I''ve heard this.  Essentially, the  
frontend network driver in the domU is complaining that the buffers  
it receives from the dom0 don''t have 16 bytes of padding at the  
start.  The lack of this padding (which is to support some  
optimisations) causes the driver to copy the data in the packet to a  
new buffer (which does have the padding).

It''s relatively easy to change the Solaris dom0 driver to provide the  
padding and I''ll look into doing that (bugid 6453402).

Not sure why we''ve not seen the warning before.
> [...]
> Q: How can I avoid these messages?
You can''t.  They should be harmless (though irritating).

The remainder of the questions are Linux distribution specific, and  
unfortunately not something that I''m familiar with, sorry :-(

dme.

David,
> Bernd, this is a Ferrari 4000, yes?  Could you give some more details  
> about the specific model?
Acer Ferrari 4005, what othe information do you neeed?
> > Q: Do I have to care about the WARNING messages like
> > bge0: driver-detected hardware error: bad ring index! ?
> 
> Yes, they look bad.  Is there anything ''interesting''
about your
> network configuration?  (jumbo frames, perhaps?)
No, nothing I think. I use the bge driver from Solaris with the default values;
the only things I configure are the IP address, netmask and broadcast address.
The network is a simple home network with a 10/100 MB Switch, two routers (one
with DSL modem) and some other workstations.
> Given that the bug is in the database, I''d say that this is  
> known :-)  Which workaround did you use?  (USB keyboard or ddi- 
> forceattach.)
ddi-forceattach

 
regards

Bernd

-- 
Bernd Schemmer
Stalburgstr. 14
60318 Frankfurt


Echte DSL-Flatrate dauerhaft für 0,- Euro*. Nur noch kurze Zeit!
"Feel free" mit GMX DSL: http://www.gmx.net/de/go/dsl

On 27 Jul 2006, at 9:20am, Bernd Schemmer wrote:
>> Bernd, this is a Ferrari 4000, yes?  Could you give some more details
>> about the specific model?
>
> Acer Ferrari 4005, what othe information do you neeed?
That''s probably enough, thanks.  I was planning to ask if anyone else  
has one of those at Sun (seems likely), in case I need to get hold of  
one.
>>> Q: Do I have to care about the WARNING messages like
>>> bge0: driver-detected hardware error: bad ring index! ?
>>
>> Yes, they look bad.  Is there anything ''interesting''
about your
>> network configuration?  (jumbo frames, perhaps?)
>
> No, nothing I think. I use the bge driver from Solaris with the  
> default values; the only things I configure are the IP address,  
> netmask and broadcast address. The network is a simple home network  
> with a 10/100 MB Switch, two routers (one with DSL modem) and some  
> other workstations.
Nothing strange there then.

Do the same problems occur if you run our updated kernel without the  
hypervisor?

dme.

David Edmondson wrote:
>
> On 27 Jul 2006, at 9:20am, Bernd Schemmer wrote:
>>> Bernd, this is a Ferrari 4000, yes?  Could you give some more
details
>>> about the specific model?
>>
>> Acer Ferrari 4005, what othe information do you neeed?
>
> That''s probably enough, thanks.  I was planning to ask if anyone
else
> has one of those at Sun (seems likely), in case I need to get hold of 
> one.
I have one - I haven''t put Xen on it yet as I needed it for some 
presentations over the last couple of weeks, but was going to do it for 
the next round of talks and demos I was doing
>
>>>> Q: Do I have to care about the WARNING messages like
>>>> bge0: driver-detected hardware error: bad ring index! ?
What version of onnv/ Opensolaris/ Solaris Express are you running? As 
far as I''m aware prior to (something like) build 43 the bge driver 
didn''t work with the 4005, you needed to download the one from
Broadcoms
website (bcme).

Chris

On 27 Jul 2006, at 10:22am, Chris Beal - Solaris Sustaining  
Engineering wrote:
>>>>> Q: Do I have to care about the WARNING messages like
>>>>> bge0: driver-detected hardware error: bad ring index! ?
> What version of onnv/ Opensolaris/ Solaris Express are you running?  
> As far as I''m aware prior to (something like) build 43 the bge  
> driver didn''t work with the 4005, you needed to download the one  
> from Broadcoms website (bcme).
Note that the bcme driver will probably not work with the ethernet  
bridge software included in the Solaris on Xen code, which means that  
your guest domains won''t be able to access the network.

dme.

-------- Original-Nachricht --------
Datum: Thu, 27 Jul 2006 10:53:58 +0100
Von: David Edmondson <dme@Sun.COM>
An: Chris Beal - Solaris Sustaining Engineering <chris.beal@Sun.COM>
Betreff: Re: [xen-discuss] my experiences with Xen until now and a lot of
questions
> 
> On 27 Jul 2006, at 10:22am, Chris Beal - Solaris Sustaining  
> Engineering wrote:
> >>>>> Q: Do I have to care about the WARNING messages like
> >>>>> bge0: driver-detected hardware error: bad ring index!
?
> > What version of onnv/ Opensolaris/ Solaris Express are you running?  
> > As far as I''m aware prior to (something like) build 43 the
bge
> > driver didn''t work with the 4005, you needed to download the
one
> > from Broadcoms website (bcme).
> 
> Note that the bcme driver will probably not work with the ethernet  
> bridge software included in the Solaris on Xen code, which means that  
> your guest domains won''t be able to access the network.
> 
> dme.
Mmmh, my guest domains are able to access the network (tested telnet, ssh, and
DNS resolution).

I''m using snv_43 and indeed, the bge driver from previous snv version
did not work with the network adapter in the 4005.


regards

Bernd

-- 
Bernd Schemmer
Stalburgstr. 14
60318 Frankfurt


Echte DSL-Flatrate dauerhaft für 0,- Euro*. Nur noch kurze Zeit!
"Feel free" mit GMX DSL: http://www.gmx.net/de/go/dsl

On 27 Jul 2006, at 12:46pm, Bernd Schemmer wrote:
>> Note that the bcme driver will probably not work with the ethernet
>> bridge software included in the Solaris on Xen code, which means that
>> your guest domains won''t be able to access the network.
>
> Mmmh, my guest domains are able to access the network (tested  
> telnet, ssh, and DNS resolution).
You''re using bge though, not bcme - right?
> I''m using snv_43 and indeed, the bge driver from previous snv  
> version did not work with the network adapter in the 4005.
The OS you installed may be nv43, but the bfu images we supplied are  
nv41 based.  Whilst running our code you''re using a bge driver from  
nv41.

Does anyone know the symptoms of the pre-nv43 bge driver failing on  
the 4005?  Do they match what Bernd saw?

dme.

Hi,
> You''re using bge though, not bcme - right?
yes
 
> > I''m using snv_43 and indeed, the bge driver from previous snv
> > version did not work with the network adapter in the 4005.
> 
> The OS you installed may be nv43, but the bfu images we supplied are  
> nv41 based.  Whilst running our code you''re using a bge driver
from
> nv41.
Ahh,ok, I did not realize that the bfu may replace the bge driver. Should I try
to reinstall the bfe driver from snv_43 after installing the bfu image?

Bernd

-- 
Bernd Schemmer
Stalburgstr. 14
60318 Frankfurt


Echte DSL-Flatrate dauerhaft für 0,- Euro*. Nur noch kurze Zeit!
"Feel free" mit GMX DSL: http://www.gmx.net/de/go/dsl

On 27 Jul 2006, at 1:39pm, Bernd Schemmer wrote:
> Should I try to reinstall the bfe driver from snv_43 after  
> installing the bfu image?
That''s worth a try.

dme.

Bernd Schemmer wrote:
> I''m using snv_43 and indeed, the bge driver from previous snv
version did not work with the network adapter in the 4005.
>   
I just thought I''d mention it as the OpenSolaris drop of the xen stuff 
was on build 41

Chris

On 26 Jul 2006, at 10:08pm, Bernd Schemmer wrote:
> - in all Linux domains are continuous messages telling that
>
> 	Received packet is 16 bytes before head.
It turns out that I''d not seen this because my Linux domU is 3.0.0  
derived, and the message was added later.  I''ll fix it for the next  
drop.

dme.

> - while copying a big file (700 MB) over the network
> via ftp 
> the machine is  not usable anymore (the response of
>  the keyboard 
> and the mouse is very, very, slow.) [Tested while
>  the 32 Bit Xen 
> Kernel was booted and one small Linux Domain with
>  128 MB was running]
This ?could? be related to a VM bug which will be fixed shortly.
> 
> - after booting different Solaris Kernel (with and
> without Xen,
> 32 Bit/64 Bit Kernel) the keyboard does not work in
>  the Gnome/CDE 
> Logon Dialog of Solaris (it does work in the GRUB
>  Menu).
> To get around this I need to telnet to the machine
>  and issue a reconfigure
> boot by creating /reconfigure and reboot. After the
>  reconfigure boot the 
>  keyboard works again
> 
>   The workaround from the bug
> 
> ttp://bugs.opensolaris.org/bugdatabase/view_bug.do?bug
> _id=6447269
> 
>   seems to fix this issue.
> 
> Q: Is this a known bug? Is there a better workaround
> for this bug?
Yes, known bug. I updated the bug yesterday with a better workaround.
Not sure if you say that one..

# echo "ddi-forceattach=1;" > /kernel/drv/kb8042.conf


MRJ
 
 
This message posted from opensolaris.org

On 27 Jul 2006, at 6:09pm, Bernd Schemmer wrote:
>> Do the same problems occur if you run our updated kernel without  
>> the  hypervisor?
>>
> No, they don''t
In that case, I''m puzzled.  Perhaps Chris will be able to shed some  
more light if he is able to try on his 4005, or I could try to get  
hold of one.

dme.

Hi,
>>Yes, known bug. I updated the bug yesterday with a better workaround.
>>Not sure if you say that one..
>>
>># echo "ddi-forceattach=1;" > /kernel/drv/kb8042.conf
That''s what I did and it worked

Thanks

Bernd
 
 
This message posted from opensolaris.org

David Edmondson wrote:
>
> On 27 Jul 2006, at 6:09pm, Bernd Schemmer wrote:
>>> Do the same problems occur if you run our updated kernel without 
>>> the  hypervisor?
>>>
>> No, they don''t
>
> In that case, I''m puzzled.  Perhaps Chris will be able to shed
some
> more light if he is able to try on his 4005, or I could try to get 
> hold of one.
>
> dme.
I''ll try and take a look, I may not get to it till Monday now. (got 
putback to 10 to do where I''ve got a load of conflicts to resolve and 
retest :()

Chris

David Edmondson wrote:
>>>> Q: Do I have to care about the WARNING messages like
>>>> bge0: driver-detected hardware error: bad ring index! ?
>>>
>>>
>>> Yes, they look bad.  Is there anything
''interesting'' about your
>>> network configuration?  (jumbo frames, perhaps?)
>>
>>
>> No, nothing I think. I use the bge driver from Solaris with the  
>> default values; the only things I configure are the IP address,  
>> netmask and broadcast address. The network is a simple home network  
>> with a 10/100 MB Switch, two routers (one with DSL modem) and some  
>> other workstations.
>
>
> Nothing strange there then.
>
> Do the same problems occur if you run our updated kernel without the  
> hypervisor?
>
No, they don''t

Bernd


-- 
Bernd Schemmer, Frankfurt am Main, Germany
http://home.arcor.de/bnsmb/index.html

Más temprano que tarde el mundio cambiará
                        Fidel Castro

On 27 Jul 2006, at 10:30pm, Bernd Schemmer wrote:
> After reinstalling the original bge driver from snv_43 the messages
>
> >>bge0: driver-detected hardware error: bad ring index! ?
>
> went away.
>
> Now the networks works
Yay!

dme.

David Edmondson wrote:

Hi,

After reinstalling the original bge driver from snv_43 the messages

 >>bge0: driver-detected hardware error: bad ring index! ?

went away.

Now the networks works

regards

Bernd


-- 
Bernd Schemmer, Frankfurt am Main, Germany
http://home.arcor.de/bnsmb/index.html

Más temprano que tarde el mundio cambiará
                        Fidel Castro

David Edmondson wrote:

fwiw:

Hi,

 >>The remainder of the questions are Linux distribution specific, and  
unfortunately not >>something that I''m familiar with, sorry :-(

I got all three Linux images from the Xen Demo Disk (Centos, Debian, and 
OpenSUSE) working now - including the VNC Server so that I can connect 
via vncviewer ;-)

I also found some places in the init scripts to add the necessary 
ethtool command -  I don''t know if I did that the right way but it
works
so I don''t care.

I had to copy the ethtool from the centos image to the debian image to 
get the debian image working.

Xen on Solaris is real cool! Congratulation to all that made this work!

And I''m really looking forward to the time when the full virtual 
virtualization works in Solaris also so that I can boot any OS in a DomU!

regards and have a nice weekend


Bernd

> Bernd, this is a Ferrari 4000, yes?  Could you give some more details  
> about the specific model?
>
> On 26 Jul 2006, at 10:08pm, Bernd Schemmer wrote:
>
>> Xen Web Server
>> --------------
>>
>> - I activated the Xen Web server in the Xen configuration file  with  
>> (xend-http-server yes)
>>   The Xen Webserver (http://localhost:8000) only shows two links  to  
>> localhost:8000  ...
>>
>> Q: Is this working as designed?
>
>
> I''ve no experience with this, sorry.
>
>> General
>> -------
>>
>> - in the /var/adm/messages of the Solaris Dom0 Domain are  lot of  
>> messages like
>>
>> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0:  
>> 74726573 7466615f 6c007265 5f747369
>> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0: 32  
>> bytes of hardware descriptor at address c88becb0:-
>> Jul 26 22:40:26 unknown bge: [ID 801725 kern.warning] WARNING:  bge0: 
>> driver-detected hardware error: bad ring index!
>> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0:  
>> 74736967 69746172 67006e6f 635f7465
>> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0:  
>> 7474615f 75626972 00736574 65737973
>> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0: 32  
>> bytes of hardware descriptor at address c88becb0:-
>> Jul 26 22:40:26 unknown last message repeated 1 time
>> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0:  PCI 
>> address 5f676f6c0074756f packet len 0x5f74 token 0x70756b flags  
>> 0x7665 index 0x7075
>> Jul 26 22:40:26 unknown bge: [ID 801725 kern.warning] WARNING:  bge0: 
>> driver-detected hardware error: bad ring index!
>> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0:  PCI 
>> address 887c600 buffer len 0x600 token 0xbcd57855
>> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0:  
>> c188068a 840fc084 00000201 0fc87539
>> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0:  
>> 0000bf8d fb830000 56850f0a e9fffffe
>> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0:  
>> 598d317e 51be0fff 18758bff 833c468b
>> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0:  
>> 5fe850c0 89fffb76 c483c845 c04d8b10
>> Jul 26 22:40:26 unknown bge: [ID 801725 kern.info] NOTICE: bge0:  
>> 9090c3c9 89559090 0000b8e5 c0850000
>>
>> Q: Do I have to care about the WARNING messages like
>> bge0: driver-detected hardware error: bad ring index! ?
>
>
> Yes, they look bad.  Is there anything ''interesting''
about your
> network configuration?  (jumbo frames, perhaps?)
>
>> Q: How can I get rid of these messages?
>
>
> They suggest a bug, though I''m not sure where (could be the
driver,
> could be somewhere else).  They''ll occur until we fix the bug or
find
> a workaround.
>
>> - while copying a big file (700 MB) over the network via ftp
>>   the machine is  not usable anymore (the response of the keyboard
>>   and the mouse is very, very, slow.) [Tested while the 32 Bit Xen
>>   Kernel was booted and one small Linux Domain with 128 MB was 
running]
>
>
> Whilst network performance of the domU is not good today, I''ve not
> seen this kind of problem with dom0 networking.  It may well be  
> related to the bge problem above.
>
>> - after booting different Solaris Kernel (with and without Xen,
>>   32 Bit/64 Bit Kernel) the keyboard does not work in the Gnome/CDE
>>   Logon Dialog of Solaris (it does work in the GRUB Menu).
>>   To get around this I need to telnet to the machine and issue a  
>> reconfigure
>>   boot by creating /reconfigure and reboot. After the reconfigure  
>> boot the
>>   keyboard works again
>>
>>   The workaround from the bug
>>
>>   http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=6447269
>>
>>   seems to fix this issue.
>>
>>
>> Q: Is this a known bug? Is there a better workaround for this bug?
>
>
> Given that the bug is in the database, I''d say that this is  known
> :-)  Which workaround did you use?  (USB keyboard or ddi- forceattach.)
>
>> - in all Linux domains are continuous messages telling that
>>
>>     Received packet is 16 bytes before head.
>
>
> Interesting - the first time I''ve heard this.  Essentially, the  
> frontend network driver in the domU is complaining that the buffers  
> it receives from the dom0 don''t have 16 bytes of padding at the  
> start.  The lack of this padding (which is to support some  
> optimisations) causes the driver to copy the data in the packet to a  
> new buffer (which does have the padding).
>
> It''s relatively easy to change the Solaris dom0 driver to provide
the
> padding and I''ll look into doing that (bugid 6453402).
>
> Not sure why we''ve not seen the warning before.
>
>> [...]
>
>
>> Q: How can I avoid these messages?
>
>
> You can''t.  They should be harmless (though irritating).
>
> The remainder of the questions are Linux distribution specific, and  
> unfortunately not something that I''m familiar with, sorry :-(
>
> dme.
>
>

-- 
Bernd Schemmer, Frankfurt am Main, Germany
http://home.arcor.de/bnsmb/index.html

Más temprano que tarde el mundio cambiará
                        Fidel Castro