xway issues / bugs

xway issues / bugs

[Muhammad Atif]

[-- Attachment #1.1: Type: text/plain, Size: 1562 bytes --]

Dear Developers,
I have tried using the interdomain communication patch for xen (Xway, http://sourceforge.net/projects/xway) a link of which is there at Xen wiki. I am having troubles getting it going. 
I am using AMD Opteron 248 (64-bit), with 2 GHz ram and IWILL motherboard. My distribution is ubuntu 7.04 and I am using Xen 3.0.3 for i686, as xway patch suggests that it is only for 32 bit systems. The GCC version is 4.1.2. I have also tried  xway on another machine which has 32 bit intel processor.
During the boot time, i see in the dmesg output giving me "net/ipv4/xway_proto.c:xsw_accept:143: inet_accept() err=-11" on Opteron. The intel 32 bit gives NO message of this type. Now to the heart of the issue
I can get the xway daemon and other related stuff loaded without any errors or warning. Upon simple ssh from one domU to another (i have compiled the ip addresses as instructed) i get a printf on screen on the receiving domain "client_sock = 5" and that is where both the domains stop processing i.e. ssh doe not seem to progress any further. strace is not helping me that much. I have noticed that xwayd.c (daemon) gives a write() call and this call is returning 0, instead of the message size.

Has anyone ever come across this error/bug? or simply has anyone ever tried xway? 


 
Best Regards,
Atif




      ____________________________________________________________________________________
Looking for last minute shopping deals?  
Find them fast with Yahoo! Search.  http://tools.search.yahoo.com/newsearch/category.php?category=shopping

[-- Attachment #1.2: Type: text/html, Size: 1791 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

migrateDevices( ) ?

[tgh]

hi
  I read the code of migration,and i am confused about the function of 
migrateDevices(), it is called for 3times,and what is the function for 
it ? could someone help me out

Thanks in advance

Re: migrateDevices( ) ?

[Stefan Berger]

[-- Attachment #1.1: Type: text/plain, Size: 808 bytes --]

xen-devel-bounces@lists.xensource.com wrote on 01/02/2008 01:41:56 AM:

> hi
>   I read the code of migration,and i am confused about the function of 
> migrateDevices(), it is called for 3times,and what is the function for 
> it ? could someone help me out

The purpose of this function is to enable the migration of 'external 
devices', i.e. this would be disk images or TPM state, from one physical 
machine to another while migrating a vm.

The prototype of the migrate function explains the different steps.

http://xenbits.xensource.com/xen-unstable.hg?file/2491691e3e69/tools/python/xen/xend/server/DevController.py


   Stefan

> 
> Thanks in advance
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel

[-- Attachment #1.2: Type: text/html, Size: 1125 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: migrateDevices( ) ?

[tgh]

hi
Thank you for your explanation
and when xen processes " live migration",xend will fork a process to do 
xc_linux_save,and also call the helphandler which will call 
migrateDevices() in three steps,that is ,migrateDevices() will copy the 
devices' state to the destination when the iterative pre-copying of the 
VM memory state is completed ,is it? or what is the tricks behind? i 
read DevController.py,but still not clear,
and i am confused about which code is called in the destination node to 
cooperate with for migration, that is ,on the destination node ,which 
code will invoke the receiving function there? is it httpserver? then 
,which function code will it call to receive ,or what to do ?i guess 
some functions will be called to deal with something ,then call the 
restore function or something, but which function is called before the 
restore function is called ? i am lost in it

Thanks in advance



Stefan Berger 写道:
>
> xen-devel-bounces@lists.xensource.com wrote on 01/02/2008 01:41:56 AM:
>
> > hi
> > I read the code of migration,and i am confused about the function of
> > migrateDevices(), it is called for 3times,and what is the function for
> > it ? could someone help me out
>
> The purpose of this function is to enable the migration of 'external 
> devices', i.e. this would be disk images or TPM state, from one 
> physical machine to another while migrating a vm.
>
> The prototype of the migrate function explains the different steps.
>
> http://xenbits.xensource.com/xen-unstable.hg?file/2491691e3e69/tools/python/xen/xend/server/DevController.py 
>
>
>
> Stefan
>
> >
> > Thanks in advance
> >
> > _______________________________________________
> > Xen-devel mailing list
> > Xen-devel@lists.xensource.com
> > http://lists.xensource.com/xen-devel

device models?

[tgh]

hi
  what does the device models mean in the xen? I read it in summit pdf 
and some code reference somewhere ,and i am not clear about it? what 
does it refer to?

Thanks in advance

migration and relocation server?

[tgh]

hi
I read the code of migration,and i have got some confusions, the source 
node sock connect the destination node with which port? 8000,is it ？and 
when migration is processing ,the relocation server is invoked ,is it ? 
and the relocation server is using port:8002? is it ?and the source node 
copy the memory state ,say dirty page, through port:8000 to the 
destination node, while the VM-migrating's networking data is send to 
relocation server, is it? or how does it work? could someone help me

Thanks in advance

Re: device models?

[Samuel Thibault]

tgh, le Fri 04 Jan 2008 13:59:19 +0800, a écrit :
>  what does the device models mean in the xen?

That's the program which emulates devices by answering to i/o requests.
Currently, that's basically a qemu process in dom0.

Samuel

Re: xway issues / bugs

[Mark Williamson]

Hi

Sorry you're having trouble with xway.  I've not tried it myself and I've also 
not known of anyone else actually running it...  It might be a good idea to 
try and contact the xway developers directly in case they don't read this 
list, or you could cc them on these e-mails..

Perhaps there's a contact e-mail address for the developers somewhere in the 
source code?

I have a Powerpoint presentation on Xway that I downloaded when it was first 
announced; the presentation was by "Kangho Kim, System Software Team, ETRI".  
ETRI = Electronics and Telecommunications Research Institute.  Maybe you can 
find some contact details for him through Google?

Cheers,
Mark

> Dear Developers,
> I have tried using the interdomain communication patch for xen (Xway,
> http://sourceforge.net/projects/xway) a link of which is there at Xen wiki.
> I am having troubles getting it going. I am using AMD Opteron 248 (64-bit),
> with 2 GHz ram and IWILL motherboard. My distribution is ubuntu 7.04 and I
> am using Xen 3.0.3 for i686, as xway patch suggests that it is only for 32
> bit systems. The GCC version is 4.1.2. I have also tried  xway on another
> machine which has 32 bit intel processor. During the boot time, i see in
> the dmesg output giving me "net/ipv4/xway_proto.c:xsw_accept:143:
> inet_accept() err=-11" on Opteron. The intel 32 bit gives NO message of
> this type. Now to the heart of the issue I can get the xway daemon and
> other related stuff loaded without any errors or warning. Upon simple ssh
> from one domU to another (i have compiled the ip addresses as instructed) i
> get a printf on screen on the receiving domain "client_sock = 5" and that
> is where both the domains stop processing i.e. ssh doe not seem to progress
> any further. strace is not helping me that much. I have noticed that
> xwayd.c (daemon) gives a write() call and this call is returning 0, instead
> of the message size.
>
> Has anyone ever come across this error/bug? or simply has anyone ever tried
> xway?
>
>
>
> Best Regards,
> Atif
>
>
>
>
>      
> ___________________________________________________________________________
>_________ Looking for last minute shopping deals?
> Find them fast with Yahoo! Search. 
> http://tools.search.yahoo.com/newsearch/category.php?category=shopping



-- 
Dave: Just a question. What use is a unicyle with no seat?  And no pedals!
Mark: To answer a question with a question: What use is a skateboard?
Dave: Skateboards have wheels.
Mark: My wheel has a wheel!

code read problems

[tgh]

hi
  I try to read the code of migration,and there are some python code and 
some c code( in guestOS application) and some c code( in guestOS kernel 
) and some c code (in xen) , and i want to trace the processing path in 
these code , and how could i achieve it,
  in what way should i compile these code?

could someone  help me

Thanks in advance

Re: code read problems

[tgh]

could you reply in english, i could not read your letter

Thanks


clp@eclinso.com 写道:
> Vielen Dank für Ihre Nachricht. 
>
> Ich bin vom 22-02-08 bis 07-03-08 nicht im Hause und werde Ihre E-Mail ab dem 10-03-07  bearbeiten.
>
> In dringenden Fällen wenden Sie sich bitte an Herrn Fabio Lüdi, er wird Ihnen gerne weiterhelfen.
>
> Sie erreichen Herrn Lüdi unter:
> Phone +41 61 6666 406
> E-Mail fl@eclinso.com
>
>
>
>
>
>   

Re: code read problems

[Daniel Stodden]

On Tue, 2008-02-26 at 19:22 +0800, tgh wrote:
> could you reply in english, i could not read your letter

Don't bother. Just an autoreply generated to tell you the guy is on
vacation.

Regarding your problem: There is not much you can do to get some sort
execution traces enabled automatically. You probably want to enable
debugging when building Xen and the libraries. Then maybe add a couple
of debug-print statements to the code, whereever you see fit.

I believe migration support in xend and libxc should be understandable
in isolation. The tricky parts are definitely done in C. Last time I
checked, xend mainly performed a single call to the tools and library.

Also note that random instrumentation of all code executed for
translating and mapping of the domU address space within the hypervisor
would probably soon get more verbose than you asked for, since some of
the functions involved can be called at a comparatively high frequency.

Rule 1 when digging your way through complex systems: Divide and
Conquer. Division comes first. Understand one thing at a time, starting
at a comparatively high-level, then selectively dig deeper.

regards,
daniel


> clp@eclinso.com 写道:
> > Vielen Dank für Ihre Nachricht. 
> >
> > Ich bin vom 22-02-08 bis 07-03-08 nicht im Hause und werde Ihre
> E-Mail ab dem 10-03-07  bearbeiten.
> >
> > In dringenden Fällen wenden Sie sich bitte an Herrn Fabio Lüdi, er
> wird Ihnen gerne weiterhelfen.
> >
> > Sie erreichen Herrn Lüdi unter:
> > Phone +41 61 6666 406
> > E-Mail fl@eclinso.com

-- 
Daniel Stodden
LRR     -      Lehrstuhl für Rechnertechnik und Rechnerorganisation
Institut für Informatik der TU München             D-85748 Garching
http://www.lrr.in.tum.de/~stodden         mailto:stodden@cs.tum.edu
PGP Fingerprint: F5A4 1575 4C56 E26A 0B33  3D80 457E 82AE B0D8 735B

Re: code read problems

[Christoph Egger]

That was an automatic generated out-of-office mail in german language.
It says, clp is not available until March 10th
and in urgent cases, Mr. Fabio Luedl can help you out.


On Tuesday 26 February 2008 12:22:40 tgh wrote:
> could you reply in english, i could not read your letter
>
> Thanks
>
> clp@eclinso.com 写道:
> > Vielen Dank für Ihre Nachricht.
> >
> > Ich bin vom 22-02-08 bis 07-03-08 nicht im Hause und werde Ihre E-Mail ab
> > dem 10-03-07  bearbeiten.
> >
> > In dringenden Fällen wenden Sie sich bitte an Herrn Fabio Lüdi, er wird
> > Ihnen gerne weiterhelfen.
> >
> > Sie erreichen Herrn Lüdi unter:
> > Phone +41 61 6666 406
> > E-Mail fl@eclinso.com
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel



-- 
AMD Saxony, Dresden, Germany
Operating System Research Center

Legal Information:
AMD Saxony Limited Liability Company & Co. KG
Sitz (Geschäftsanschrift):
   Wilschdorfer Landstr. 101, 01109 Dresden, Deutschland
Registergericht Dresden: HRA 4896
vertretungsberechtigter Komplementär:
   AMD Saxony LLC (Sitz Wilmington, Delaware, USA)
Geschäftsführer der AMD Saxony LLC:
   Dr. Hans-R. Deppe, Thomas McCoy

Re: code read problems

[pradeep singh rautela]

[-- Attachment #1: Type: text/plain, Size: 1840 bytes --]

On 26/02/2008, Christoph Egger <Christoph.Egger@amd.com> wrote:
>
>  That was an automatic generated out-of-office mail in german language.
>  It says, clp is not available until March 10th
>  and in urgent cases, Mr. Fabio Luedl can help you out.

Too bad, i thought that was not new kind of Spam :(.
Will have to use google transtlator next time before marking SPAMs.

Thanks for Transalation.

Cu,
      --Pradeep
>
>
>
>  On Tuesday 26 February 2008 12:22:40 tgh wrote:
>  > could you reply in english, i could not read your letter
>  >
>  > Thanks
>  >
>  > clp@eclinso.com 写道:
>  > > Vielen Dank für Ihre Nachricht.
>  > >
>  > > Ich bin vom 22-02-08 bis 07-03-08 nicht im Hause und werde Ihre E-Mail ab
>  > > dem 10-03-07  bearbeiten.
>  > >
>  > > In dringenden Fällen wenden Sie sich bitte an Herrn Fabio Lüdi, er wird
>  > > Ihnen gerne weiterhelfen.
>  > >
>  > > Sie erreichen Herrn Lüdi unter:
>  > > Phone +41 61 6666 406
>  > > E-Mail fl@eclinso.com
>  >
>  > _______________________________________________
>  > Xen-devel mailing list
>  > Xen-devel@lists.xensource.com
>  > http://lists.xensource.com/xen-devel
>
>
>
>
> --
>  AMD Saxony, Dresden, Germany
>  Operating System Research Center
>
>  Legal Information:
>  AMD Saxony Limited Liability Company & Co. KG
>  Sitz (Geschäftsanschrift):
>    Wilschdorfer Landstr. 101, 01109 Dresden, Deutschland
>  Registergericht Dresden: HRA 4896
>  vertretungsberechtigter Komplementär:
>    AMD Saxony LLC (Sitz Wilmington, Delaware, USA)
>  Geschäftsführer der AMD Saxony LLC:
>    Dr. Hans-R. Deppe, Thomas McCoy
>
>
>
>  _______________________________________________
>  Xen-devel mailing list
>  Xen-devel@lists.xensource.com
>  http://lists.xensource.com/xen-devel
>


-- 
Pradeep Singh Rautela
http://eagain.wordpress.com
http://emptydomain.googlepages.com

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: code read problems

[tgh]

hi
  thank you for your reply
  and i read the code of xc_save.c and xc_restore.c, which maybe do the 
function of VMsaving and VMrestoring, but i wander, if the code of 
xc_save.c and xc_restore.c is called by some python code or c code 
during migration or xm save and xm restore or not? the code of xc_save.c 
is a main function ,and is it called by other program to do migration or 
not ?

  and in the code of xc_save.c, there seem no notification to the 
xenstore or devbackend, and how are these VMdevbackends destroyed in the 
dom0, when migration or save?
  and in the code of xc_restore.c,there seems to be an existing 
VMdomain,and the whole data from savefile or from  migration,will load 
into the VMdomain, is it？  then ,what code call these code of xc_save 
and xc_restore?  i am confused

could you help me out

Thanks in advance


Daniel Stodden 写道:
> On Tue, 2008-02-26 at 19:22 +0800, tgh wrote:
>   
>> could you reply in english, i could not read your letter
>>     
>
> Don't bother. Just an autoreply generated to tell you the guy is on
> vacation.
>
> Regarding your problem: There is not much you can do to get some sort
> execution traces enabled automatically. You probably want to enable
> debugging when building Xen and the libraries. Then maybe add a couple
> of debug-print statements to the code, whereever you see fit.
>
> I believe migration support in xend and libxc should be understandable
> in isolation. The tricky parts are definitely done in C. Last time I
> checked, xend mainly performed a single call to the tools and library.
>
> Also note that random instrumentation of all code executed for
> translating and mapping of the domU address space within the hypervisor
> would probably soon get more verbose than you asked for, since some of
> the functions involved can be called at a comparatively high frequency.
>
> Rule 1 when digging your way through complex systems: Divide and
> Conquer. Division comes first. Understand one thing at a time, starting
> at a comparatively high-level, then selectively dig deeper.
>
> regards,
> daniel
>
>
>   
>> clp@eclinso.com 写道:
>>     
>>> Vielen Dank für Ihre Nachricht. 
>>>
>>> Ich bin vom 22-02-08 bis 07-03-08 nicht im Hause und werde Ihre
>>>       
>> E-Mail ab dem 10-03-07  bearbeiten.
>>     
>>> In dringenden Fällen wenden Sie sich bitte an Herrn Fabio Lüdi, er
>>>       
>> wird Ihnen gerne weiterhelfen.
>>     
>>> Sie erreichen Herrn Lüdi unter:
>>> Phone +41 61 6666 406
>>> E-Mail fl@eclinso.com
>>>       
>
>   

Re: code read problems

[Mike Sun]

[-- Attachment #1: Type: text/plain, Size: 3064 bytes --]

I've had to trace through the migration code and made the following
notes.  Hope it helps:

http://msun.bluespot.org/wiki/doku.php?id=xen_migration_notes

On Wed, Feb 27, 2008 at 4:37 AM, tgh <wwwwww4187@sina.com.cn> wrote:
> hi
>   thank you for your reply
>   and i read the code of xc_save.c and xc_restore.c, which maybe do the
>  function of VMsaving and VMrestoring, but i wander, if the code of
>  xc_save.c and xc_restore.c is called by some python code or c code
>  during migration or xm save and xm restore or not? the code of xc_save.c
>  is a main function ,and is it called by other program to do migration or
>  not ?
>
>   and in the code of xc_save.c, there seem no notification to the
>  xenstore or devbackend, and how are these VMdevbackends destroyed in the
>  dom0, when migration or save?
>   and in the code of xc_restore.c,there seems to be an existing
>  VMdomain,and the whole data from savefile or from  migration,will load
>  into the VMdomain, is it？  then ,what code call these code of xc_save
>  and xc_restore?  i am confused
>
>  could you help me out
>
>  Thanks in advance
>
>
>  Daniel Stodden 写道:
>
>
> > On Tue, 2008-02-26 at 19:22 +0800, tgh wrote:
>  >
>  >> could you reply in english, i could not read your letter
>  >>
>  >
>  > Don't bother. Just an autoreply generated to tell you the guy is on
>  > vacation.
>  >
>  > Regarding your problem: There is not much you can do to get some sort
>  > execution traces enabled automatically. You probably want to enable
>  > debugging when building Xen and the libraries. Then maybe add a couple
>  > of debug-print statements to the code, whereever you see fit.
>  >
>  > I believe migration support in xend and libxc should be understandable
>  > in isolation. The tricky parts are definitely done in C. Last time I
>  > checked, xend mainly performed a single call to the tools and library.
>  >
>  > Also note that random instrumentation of all code executed for
>  > translating and mapping of the domU address space within the hypervisor
>  > would probably soon get more verbose than you asked for, since some of
>  > the functions involved can be called at a comparatively high frequency.
>  >
>  > Rule 1 when digging your way through complex systems: Divide and
>  > Conquer. Division comes first. Understand one thing at a time, starting
>  > at a comparatively high-level, then selectively dig deeper.
>  >
>  > regards,
>  > daniel
>  >
>  >
>  >
>  >> clp@eclinso.com 写道:
>  >>
>  >>> Vielen Dank für Ihre Nachricht.
>  >>>
>  >>> Ich bin vom 22-02-08 bis 07-03-08 nicht im Hause und werde Ihre
>  >>>
>  >> E-Mail ab dem 10-03-07  bearbeiten.
>  >>
>  >>> In dringenden Fällen wenden Sie sich bitte an Herrn Fabio Lüdi, er
>  >>>
>  >> wird Ihnen gerne weiterhelfen.
>  >>
>  >>> Sie erreichen Herrn Lüdi unter:
>  >>> Phone +41 61 6666 406
>  >>> E-Mail fl@eclinso.com
>  >>>
>  >
>  >
>
>
>
>
> _______________________________________________
>  Xen-devel mailing list
>  Xen-devel@lists.xensource.com
>  http://lists.xensource.com/xen-devel
>

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: code read problems

[Brendan Cully]

Nice summary!

In part 3, you have:

  ? What does that checkpoint flag do ?

When that flag is set (through xm save -c), the domain resumes
operation after the save operation. Normally "save" destroys the
domain.

You might be interested in my presentation at xen summit 4 about
checkpointing, which goes into some detail about the save/migration
process:

http://xen.org/files/xensummit_4/talk_Cully.pdf

Since then we've increased the speed of checkpointing considerably,
but the code to do it isn't in Xen yet. We have a paper about it in
the next NSDI conference which has the details, if you're interested.

On Wednesday, 27 February 2008 at 14:23, Mike Sun wrote:
> I've had to trace through the migration code and made the following
> notes.  Hope it helps:
> 
> http://msun.bluespot.org/wiki/doku.php?id=xen_migration_notes
> 
> On Wed, Feb 27, 2008 at 4:37 AM, tgh <wwwwww4187@sina.com.cn> wrote:
> > hi
> >   thank you for your reply
> >   and i read the code of xc_save.c and xc_restore.c, which maybe do the
> >  function of VMsaving and VMrestoring, but i wander, if the code of
> >  xc_save.c and xc_restore.c is called by some python code or c code
> >  during migration or xm save and xm restore or not? the code of xc_save.c
> >  is a main function ,and is it called by other program to do migration or
> >  not ?
> >
> >   and in the code of xc_save.c, there seem no notification to the
> >  xenstore or devbackend, and how are these VMdevbackends destroyed in the
> >  dom0, when migration or save?
> >   and in the code of xc_restore.c,there seems to be an existing
> >  VMdomain,and the whole data from savefile or from  migration,will load
> >  into the VMdomain, is it？  then ,what code call these code of xc_save
> >  and xc_restore?  i am confused
> >
> >  could you help me out
> >
> >  Thanks in advance
> >
> >
> >  Daniel Stodden 写道:
> >
> >
> > > On Tue, 2008-02-26 at 19:22 +0800, tgh wrote:
> >  >
> >  >> could you reply in english, i could not read your letter
> >  >>
> >  >
> >  > Don't bother. Just an autoreply generated to tell you the guy is on
> >  > vacation.
> >  >
> >  > Regarding your problem: There is not much you can do to get some sort
> >  > execution traces enabled automatically. You probably want to enable
> >  > debugging when building Xen and the libraries. Then maybe add a couple
> >  > of debug-print statements to the code, whereever you see fit.
> >  >
> >  > I believe migration support in xend and libxc should be understandable
> >  > in isolation. The tricky parts are definitely done in C. Last time I
> >  > checked, xend mainly performed a single call to the tools and library.
> >  >
> >  > Also note that random instrumentation of all code executed for
> >  > translating and mapping of the domU address space within the hypervisor
> >  > would probably soon get more verbose than you asked for, since some of
> >  > the functions involved can be called at a comparatively high frequency.
> >  >
> >  > Rule 1 when digging your way through complex systems: Divide and
> >  > Conquer. Division comes first. Understand one thing at a time, starting
> >  > at a comparatively high-level, then selectively dig deeper.
> >  >
> >  > regards,
> >  > daniel
> >  >
> >  >
> >  >
> >  >> clp@eclinso.com 写道:
> >  >>
> >  >>> Vielen Dank für Ihre Nachricht.
> >  >>>
> >  >>> Ich bin vom 22-02-08 bis 07-03-08 nicht im Hause und werde Ihre
> >  >>>
> >  >> E-Mail ab dem 10-03-07  bearbeiten.
> >  >>
> >  >>> In dringenden Fällen wenden Sie sich bitte an Herrn Fabio Lüdi, er
> >  >>>
> >  >> wird Ihnen gerne weiterhelfen.
> >  >>
> >  >>> Sie erreichen Herrn Lüdi unter:
> >  >>> Phone +41 61 6666 406
> >  >>> E-Mail fl@eclinso.com
> >  >>>
> >  >
> >  >
> >
> >
> >
> >
> > _______________________________________________
> >  Xen-devel mailing list
> >  Xen-devel@lists.xensource.com
> >  http://lists.xensource.com/xen-devel
> >

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

Re: code read problems

[Mike Sun]

[-- Attachment #1: Type: text/plain, Size: 4682 bytes --]

Hi Brendan,

I'd be very interested to see your NSDI paper.  Any way I could get a
pre-print?  We're actually looking ways of using COW memory to improve
the speed of the checkpoints in our research.

Thanks!
Mike

On Thu, Feb 28, 2008 at 1:52 PM, Brendan Cully <brendan@cs.ubc.ca> wrote:
> Nice summary!
>
>  In part 3, you have:
>
>   ? What does that checkpoint flag do ?
>
>  When that flag is set (through xm save -c), the domain resumes
>  operation after the save operation. Normally "save" destroys the
>  domain.
>
>  You might be interested in my presentation at xen summit 4 about
>  checkpointing, which goes into some detail about the save/migration
>  process:
>
>  http://xen.org/files/xensummit_4/talk_Cully.pdf
>
>  Since then we've increased the speed of checkpointing considerably,
>  but the code to do it isn't in Xen yet. We have a paper about it in
>  the next NSDI conference which has the details, if you're interested.
>
>
>
>  On Wednesday, 27 February 2008 at 14:23, Mike Sun wrote:
>  > I've had to trace through the migration code and made the following
>  > notes.  Hope it helps:
>  >
>  > http://msun.bluespot.org/wiki/doku.php?id=xen_migration_notes
>  >
>  > On Wed, Feb 27, 2008 at 4:37 AM, tgh <wwwwww4187@sina.com.cn> wrote:
>  > > hi
>  > >   thank you for your reply
>  > >   and i read the code of xc_save.c and xc_restore.c, which maybe do the
>  > >  function of VMsaving and VMrestoring, but i wander, if the code of
>  > >  xc_save.c and xc_restore.c is called by some python code or c code
>  > >  during migration or xm save and xm restore or not? the code of xc_save.c
>  > >  is a main function ,and is it called by other program to do migration or
>  > >  not ?
>  > >
>  > >   and in the code of xc_save.c, there seem no notification to the
>  > >  xenstore or devbackend, and how are these VMdevbackends destroyed in the
>  > >  dom0, when migration or save?
>  > >   and in the code of xc_restore.c,there seems to be an existing
>  > >  VMdomain,and the whole data from savefile or from  migration,will load
>  > >  into the VMdomain, is it？  then ,what code call these code of xc_save
>  > >  and xc_restore?  i am confused
>  > >
>  > >  could you help me out
>  > >
>  > >  Thanks in advance
>  > >
>  > >
>  > >  Daniel Stodden 写道:
>  > >
>  > >
>  > > > On Tue, 2008-02-26 at 19:22 +0800, tgh wrote:
>  > >  >
>  > >  >> could you reply in english, i could not read your letter
>  > >  >>
>  > >  >
>  > >  > Don't bother. Just an autoreply generated to tell you the guy is on
>  > >  > vacation.
>  > >  >
>  > >  > Regarding your problem: There is not much you can do to get some sort
>  > >  > execution traces enabled automatically. You probably want to enable
>  > >  > debugging when building Xen and the libraries. Then maybe add a couple
>  > >  > of debug-print statements to the code, whereever you see fit.
>  > >  >
>  > >  > I believe migration support in xend and libxc should be understandable
>  > >  > in isolation. The tricky parts are definitely done in C. Last time I
>  > >  > checked, xend mainly performed a single call to the tools and library.
>  > >  >
>  > >  > Also note that random instrumentation of all code executed for
>  > >  > translating and mapping of the domU address space within the hypervisor
>  > >  > would probably soon get more verbose than you asked for, since some of
>  > >  > the functions involved can be called at a comparatively high frequency.
>  > >  >
>  > >  > Rule 1 when digging your way through complex systems: Divide and
>  > >  > Conquer. Division comes first. Understand one thing at a time, starting
>  > >  > at a comparatively high-level, then selectively dig deeper.
>  > >  >
>  > >  > regards,
>  > >  > daniel
>  > >  >
>  > >  >
>  > >  >
>  > >  >> clp@eclinso.com 写道:
>  > >  >>
>  > >  >>> Vielen Dank für Ihre Nachricht.
>  > >  >>>
>  > >  >>> Ich bin vom 22-02-08 bis 07-03-08 nicht im Hause und werde Ihre
>  > >  >>>
>  > >  >> E-Mail ab dem 10-03-07  bearbeiten.
>  > >  >>
>  > >  >>> In dringenden Fällen wenden Sie sich bitte an Herrn Fabio Lüdi, er
>  > >  >>>
>  > >  >> wird Ihnen gerne weiterhelfen.
>  > >  >>
>  > >  >>> Sie erreichen Herrn Lüdi unter:
>  > >  >>> Phone +41 61 6666 406
>  > >  >>> E-Mail fl@eclinso.com
>  > >  >>>
>  > >  >
>  > >  >
>  > >
>  > >
>  > >
>  > >
>  > > _______________________________________________
>  > >  Xen-devel mailing list
>  > >  Xen-devel@lists.xensource.com
>  > >  http://lists.xensource.com/xen-devel
>  > >
>
>  > _______________________________________________
>  > Xen-devel mailing list
>  > Xen-devel@lists.xensource.com
>  > http://lists.xensource.com/xen-devel
>
>

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: code read problems

[tgh]

hi
  I read checkpoint code
  when checkpoint or migration ends,guestos continues processing by 
returning back from hypercall of suspend, that is ,take_machine_down() 
call the post_suspend() to continue ,is it right?
  while in post_suspend(),pfn_to_mfn_frame_list_list[] is converted by 
virt_to_mfn(),why do this convert? i could not find where it has been 
converted,and why should it be converted back?
  and in the xc_domain_restore()in the dom0,pagetable has been 
uncanonicalized, which is coupled with canonicalization in 
xc_domain_save(),is it ? ,and what is the reasons for 
pfn_to_mfn_frame_list_list[] virt_to_mfn in post_suspend()?
  could you help me

thanks

Re: code read problems

[Daniel Stodden]

Hi.

On Fri, 2008-02-29 at 14:44 +0800, tgh wrote:
> hi
>   I read checkpoint code
>   when checkpoint or migration ends,guestos continues processing by 
> returning back from hypercall of suspend, that is ,take_machine_down() 
> call the post_suspend() to continue ,is it right?
>   while in post_suspend(),pfn_to_mfn_frame_list_list[] is converted by 
> virt_to_mfn(),why do this convert? i could not find where it has been 
> converted,and why should it be converted back?

No, not converted. Re-initialized. 

Please compare post_suspend() with the original initialization at system
boot in setup_arch() (arch/x/kernel/setup-xen.c). It should come clear
then.

Observe that this is not canonicalization, because it is not the p2m,
but a directory containing the *machine frames* *carrying* the p2m.
Since the memory underlying the p2m has changed after resume, that
directory needs a reset.

>   and in the xc_domain_restore()in the dom0,pagetable has been 
> uncanonicalized, which is coupled with canonicalization in 
> xc_domain_save(),is it ? ,and what is the reasons for 
> pfn_to_mfn_frame_list_list[] virt_to_mfn in post_suspend()?
>   could you help me

Canonicalization forth and back is done by the migration code. Keeping
the allocated frame list up to date is done by the domain itself. I do
agree that this can be confusing indeed.

hth,
Daniel
 
-- 
Daniel Stodden
LRR     -      Lehrstuhl für Rechnertechnik und Rechnerorganisation
Institut für Informatik der TU München             D-85748 Garching
http://www.lrr.in.tum.de/~stodden         mailto:stodden@cs.tum.edu
PGP Fingerprint: F5A4 1575 4C56 E26A 0B33  3D80 457E 82AE B0D8 735B

Re: code read problems

[tgh]

Thank you for your reply
I encounter other confusions,

 1)when I read the code of restore() in checkpoint.py for paravirt
 and restore() calls for the dominfo.completeRestore(),which call 
_registerWatches(), and then call refreshShutdown(),
 and in refreshShutdown(),if   elif xeninfo['shutdown']:  and  if reason 
== 'suspend' ,self._unwatchVM()will be called ,

                   if reason == 'suspend':
                       self._stateSet(DOM_STATE_SUSPENDED)
                       # Don't destroy the domain.  XendCheckpoint will do
                       # this once it has finished.  However, stop watching
                       # the VM path now, otherwise we will end up with one
                       # watch for the old domain, and one for the new.
                       self._unwatchVm()

and then ,why do we call _registerWatches() before refreshShutdown(),i 
am confused about it,
and what is the function of dominfo.completeRestore()? it 
introducedomain(),and storeDomDetails() and registerWatches()and 
,refreshShutdown(),and why should it refreshShutdown()?and is it that in 
refreshShutdown(),the old domain is cleaned and destoryed, or the new 
domain is dealt with? i am confused

2) and the code of save() in the checkpoint.py, how does the 
saveInputHandler synchronize with the XC_SAVE ?  how many times do they 
synchronize with each other？ only ,XC_SAVE sends "suspend" to 
saveInputHandler,  and then saveInputHandler  send "done" to XC_SAVE   
,is it right?                           
       def saveInputHandler(line, tochild):
           log.debug("In saveInputHandler %s", line)
           if line == "suspend":
               log.debug("Suspending %d ...", dominfo.getDomid())
               dominfo.shutdown('suspend')
               dominfo.waitForShutdown()
               dominfo.migrateDevices(network, dst, 
DEV_MIGRATE_STEP2,domain_name)
               log.info("Domain %d suspended.", dominfo.getDomid())
               dominfo.migrateDevices(network, dst, 
DEV_MIGRATE_STEP3,domain_name)
               #send signal to device model for save
               if hvm:
                   log.info("release_devices for hvm domain")
                   dominfo._releaseDevices(True)
               tochild.write("done\n")
               tochild.flush()
               log.debug('Written done')

and how many times does XC_SAVE synchronize with Handler, i read the 
code , and just find only one time ,is it right?
3) and what is the function of the dominfo.migrateDevices with 
DEV_MIGRATE_STEP2 and DEV_MIGRATE_STEP3, i read the code ,but still 
confused, does  it  put the  information of the devices into the 
checkpoint file , or what?
4) and how does xen deal with the requests of devices in the ring and 
gnttb and backend ,how does it store them? i have not find the code to 
deal with it?


could you help me

Thanks in advance



Daniel Stodden 写道:
> Hi.
>
> On Fri, 2008-02-29 at 14:44 +0800, tgh wrote:
>   
>> hi
>>   I read checkpoint code
>>   when checkpoint or migration ends,guestos continues processing by 
>> returning back from hypercall of suspend, that is ,take_machine_down() 
>> call the post_suspend() to continue ,is it right?
>>   while in post_suspend(),pfn_to_mfn_frame_list_list[] is converted by 
>> virt_to_mfn(),why do this convert? i could not find where it has been 
>> converted,and why should it be converted back?
>>     
>
> No, not converted. Re-initialized. 
>
> Please compare post_suspend() with the original initialization at system
> boot in setup_arch() (arch/x/kernel/setup-xen.c). It should come clear
> then.
>
> Observe that this is not canonicalization, because it is not the p2m,
> but a directory containing the *machine frames* *carrying* the p2m.
> Since the memory underlying the p2m has changed after resume, that
> directory needs a reset.
>
>   
>>   and in the xc_domain_restore()in the dom0,pagetable has been 
>> uncanonicalized, which is coupled with canonicalization in 
>> xc_domain_save(),is it ? ,and what is the reasons for 
>> pfn_to_mfn_frame_list_list[] virt_to_mfn in post_suspend()?
>>   could you help me
>>     
>
> Canonicalization forth and back is done by the migration code. Keeping
> the allocated frame list up to date is done by the domain itself. I do
> agree that this can be confusing indeed.
>
> hth,
> Daniel
>  
>   

Re: code read problems

[tgh]

hi
  I try to understand the migration code ,and I guess ,during 
migration,there is a step to make some change to the network bridge, 
but  which  code  does  it ? could  someone help me

Thanks