PAE on xen-2.0 testing?

PAE on xen-2.0 testing?

[stoeni]

hi,

is it possible to use PAE on 2.0-testing sources? grepping around shows
some PAE stuff, but how to enable it? make ARCH=xen menuconfig shows
only 4GB ...


-- 
Tschaw/2,
Stoeni!

++ Please remember: rm -rf means "read mail -really fast" ++

Re: PAE on xen-2.0 testing?

[David F Barrera]

On Mon, 2005-10-17 at 20:16 +0200, stoeni wrote:
> hi,
> 
> is it possible to use PAE on 2.0-testing sources? grepping around shows
> some PAE stuff, but how to enable it? make ARCH=xen menuconfig shows
> only 4GB ...

I haven't tried it on 2.0-testing, but on xen-unstable, you build Xen
this way: make XEN_TARGET_X86_PAE=y world
> 
> 
-- 
Regards,

David F Barrera
Linux Technology Center
Systems and Technology Group, IBM

"The wisest men follow their own direction. "
                                                        Euripides

Re: PAE on xen-2.0 testing?

[Mark Williamson]

I'm afraid if you want PAE you'll want unstable/3.0 - it'll probably never be 
backported to 2.0.  Xen-unstable is relatively stable now though, so you 
might want to try it out for evaluation purposes since it will become 3.0 
eventually.

Cheers,
Mark

On Monday 17 October 2005 19:16, stoeni wrote:
> hi,
>
> is it possible to use PAE on 2.0-testing sources? grepping around shows
> some PAE stuff, but how to enable it? make ARCH=xen menuconfig shows
> only 4GB ...

Re: PAE on xen-2.0 testing?

[stoeni]

Mark Williamson schrieb:
> I'm afraid if you want PAE you'll want unstable/3.0 - it'll probably never be 
> backported to 2.0.  Xen-unstable is relatively stable now though, so you 
> might want to try it out for evaluation purposes since it will become 3.0 
> eventually.

ok, good to know. the 3.0 thing doesn't really work, after mailing
hundreds :) of mails with keir yesterday the source from today broke the
system again. gdth causes a panic .. if it compiles some vif are coming
up, some not. same show as on saturday/sunday :(

i'll wait the next days for an upcoming source that will work on this
maschine ;-)

-- 
Tschaw/2,
Stoeni!

++ Please remember: rm -rf means "read mail -really fast" ++

Re: PAE on xen-2.0 testing?

[Ted Kaczmarek]

On Mon, 2005-10-17 at 20:26 +0200, stoeni wrote:
> Mark Williamson schrieb:
> > I'm afraid if you want PAE you'll want unstable/3.0 - it'll probably never be 
> > backported to 2.0.  Xen-unstable is relatively stable now though, so you 
> > might want to try it out for evaluation purposes since it will become 3.0 
> > eventually.
> 
> ok, good to know. the 3.0 thing doesn't really work, after mailing
> hundreds :) of mails with keir yesterday the source from today broke the
> system again. gdth causes a panic .. if it compiles some vif are coming
> up, some not. same show as on saturday/sunday :(
> 
> i'll wait the next days for an upcoming source that will work on this
> maschine ;-)
> 
How long ago did you start seeing these issues and on what hardware.
With two or three nics I see this on an SMP Athlon and a UP P4.
By this I mean vif not mapping to a bridge or mapping ro the wrong
bridge.

 If I go back to a single nic setup using a secondary ip on the dom0 for
the domU gateway things are playing nice.

 With two and three nics and bridge between dom0's I just increased the
rate at which of my hair follicle count declines :-)


Regards,
Ted

Re: PAE on xen-2.0 testing?

[stoeni]

Ted Kaczmarek schrieb:
 > How long ago did you start seeing these issues and on what hardware.

box is: 2*2.8xeon (fsb 800) with intel chipset 7230, 6x1gb ram, single
e100 and dual e1000 (only the first e1000 active) and 150gb raid5 on
icp/gdth with debian 3.1 (sarge) as base system.

first try with xen 3.0 because auf 6gb ram. xen 2.0.7 sees only 3g ram
in this machine, don't know why, other boxes (ten at production) using
full 4gb ram on xen 2.0.7.

> With two or three nics I see this on an SMP Athlon and a UP P4.
> By this I mean vif not mapping to a bridge or mapping ro the wrong
> bridge.

same behavior with the sources from today. vif assigned to first domU,
second domU crashes, third domU got a vif but doesnt work. very strange.

actual i did a make uninstall an used release  "Sun Oct 16 17:25:15 2005
+0100 7396:9b51e7637676", starting three domains gives the following
results:

one network vif3.1 (why 3.1 and not 2.1?)

domU 1 up, no networking
domU 2 up, no networking
domU 3 did not start

also nice, all domains configured with 1024 mb ram, but look here:

Name              ID  Mem(MiB)  CPU  VCPUs  State   Time(s)
Domain-0           0       123    -      4  r-----     19.7
vm-tempo           1      1023    0      1  -b----     12.5
vm-tempo1          2      1024    0      1  -b----      0.6

why 1023 and 1024? ;-)

very strange. yesterday with some help from keir i got it up and running
with the sources from sunday, but can't get it up and running after
using the sources from today (doing make uninstall an so on). restored
some backups etc. but it's not working at the moment.

after shuting down this domains vm-tempo doesn't find a root fs ...
arghl .. all domains are equal except the logical volumes and networking
stuff.

>  If I go back to a single nic setup using a secondary ip on the dom0 for
> the domU gateway things are playing nice.

thanks for the tip, it's worth a try. maybe i'll try the e100 instead of
the e1000.

>  With two and three nics and bridge between dom0's I just increased the
> rate at which of my hair follicle count declines :-)

hrhr ... until xen-unstable i also had much and long hairs :))

-- 
Tschaw/2,
Stoeni!

++ Please remember: rm -rf means "read mail -really fast" ++

RE: PAE on xen-2.0 testing?

[Ian Pratt]

 
> ok, good to know. the 3.0 thing doesn't really work, after 
> mailing hundreds :) of mails with keir yesterday the source 
> from today broke the system again. gdth causes a panic .. if 
> it compiles some vif are coming up, some not. same show as on 
> saturday/sunday :(

There's a good chance Christian has fixed the vif/block dev issue in the
staging tree. The race seemed to bite some people hard and other not at
all.

Ian

Re: PAE on xen-2.0 testing?

[stoeni]

Ian Pratt wrote:
>>ok, good to know. the 3.0 thing doesn't really work, after 
>>mailing hundreds :) of mails with keir yesterday the source 
>>from today broke the system again. gdth causes a panic .. if 
>>it compiles some vif are coming up, some not. same show as on 
>>saturday/sunday :(
> 
> 
> There's a good chance Christian has fixed the vif/block dev issue in the
> staging tree. The race seemed to bite some people hard and other not at
> all.

great! todays unstable-src snapshot shows same behaviour.

where can i get the staging tree so i can test it?

-- 
Tschaw/2,
Stoeni!

++ Please remember: rm -rf means "read mail -really fast" ++

Re: PAE on xen-2.0 testing?

[Steven Hand]

>Ian Pratt wrote:
>>>ok, good to know. the 3.0 thing doesn't really work, after 
>>>mailing hundreds :) of mails with keir yesterday the source 
>>>from today broke the system again. gdth causes a panic .. if 
>>>it compiles some vif are coming up, some not. same show as on 
>>>saturday/sunday :(
>> 
>> 
>> There's a good chance Christian has fixed the vif/block dev issue in the
>> staging tree. The race seemed to bite some people hard and other not at
>> all.
>
>great! todays unstable-src snapshot shows same behaviour.
>
>where can i get the staging tree so i can test it?

It'll be pushed to the public tree once it's passed the (short) 
xen regression test. Should be later today. 

cheers,

S.

Re: PAE on xen-2.0 testing?

[stoeni]

>>where can i get the staging tree so i can test it?
> It'll be pushed to the public tree once it's passed the (short) 
> xen regression test. Should be later today. 

ok. i'll wait for the src build tomorrow :)

-- 
Tschaw/2,
Stoeni!

++ Please remember: rm -rf means "read mail -really fast" ++