[uml-devel] UML development, anyone?

[uml-devel] UML development, anyone?

[Sakari Ailus]

Hello all,

I'm a long-time UML user and lately I've become a bit worried on the 
state of the UML. This list, as well as the -user one, is very quiet 
these days. Is there someone still actively developing the UML? I have 
to say I really like it! Running Linux on Linux reliably and 
efficiently, just what I need! :-)

I'm currently running three UML virtual machines on an server using 
Athlon XP-M. The host is running 2.6.31 and the guests are 2.6.27.

I cannot upgrade neither host nor guests due to these problems:

- SKAS 3 patches available up to 2.6.31 only. Merging this to current 
kernel might not be a problem, just some work, but:

- Severe data corruption on guest on recent guest kernel versions. I'm 
using a few hard disk partitions directly for UML guests. Something 
after 2.6.27 guest breaks this quite badly. Even for small data sets 
like 100 MiB, copying a file to a different location on guest and then 
running md5sum on both often produces a different result.

Do I have any options than to switch to Qemu and suffer the performance 
penalty from emulation (kqemu is obsolete, too) or to switch to newer 
hardware that supports kvm? Qemu performance is somewhat less than 10 % 
of the native (or UML) performance.

I also found something called lguest but haven't given it a try yet. 
Looks the most interesting but that doesn't seem to have much list 
activity either. So it works perfectly for almost everyone or has few 
users... ;)

How are others coping in the current situation? Or is it just me who has 
these troubles...?

Best regards,

-- 
Sakari Ailus
sakari.ailus@iki.fi

------------------------------------------------------------------------------
This SF.net Dev2Dev email is sponsored by:

Show off your parallel programming skills.
Enter the Intel(R) Threading Challenge 2010.
http://p.sf.net/sfu/intel-thread-sfd
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] UML development, anyone?

[Boaz Harrosh]

On 08/31/2010 01:17 AM, Sakari Ailus wrote:
> Hello all,
> 
> I'm a long-time UML user and lately I've become a bit worried on the 
> state of the UML. This list, as well as the -user one, is very quiet 
> these days. Is there someone still actively developing the UML? I have 
> to say I really like it! Running Linux on Linux reliably and 
> efficiently, just what I need! :-)
> 
> I'm currently running three UML virtual machines on an server using 
> Athlon XP-M. The host is running 2.6.31 and the guests are 2.6.27.
> 
> I cannot upgrade neither host nor guests due to these problems:
> 
> - SKAS 3 patches available up to 2.6.31 only. Merging this to current 
> kernel might not be a problem, just some work, but:
> 
> - Severe data corruption on guest on recent guest kernel versions. I'm 
> using a few hard disk partitions directly for UML guests. Something 
> after 2.6.27 guest breaks this quite badly. Even for small data sets 
> like 100 MiB, copying a file to a different location on guest and then 
> running md5sum on both often produces a different result.
> 
> Do I have any options than to switch to Qemu and suffer the performance 
> penalty from emulation (kqemu is obsolete, too) or to switch to newer 
> hardware that supports kvm? Qemu performance is somewhat less than 10 % 
> of the native (or UML) performance.
> 
> I also found something called lguest but haven't given it a try yet. 
> Looks the most interesting but that doesn't seem to have much list 
> activity either. So it works perfectly for almost everyone or has few 
> users... ;)
> 
> How are others coping in the current situation? Or is it just me who has 
> these troubles...?
> 

I hear you ...

I'm using UML for years, as my developing platform for new Kernel code.
You can imagine how it shortens the development cycle. For guests I must
use the absolute newest Kernel all the time. So ...

- Please send a detailed bug report about that data corruption so someone
  can reproduce it. I'm not sure what block device you are using but it has
  a bug. If you have a good version 2.6.27 and a bad version 2.6.28. It should
  be easy to "git bisect" to the exact patch that broke the system, and should
  be easy to fix. (Tell me if you need help, it's easy really)

- I probably don't have your problem because in my UMLs /home is an NFS
  mount to the office /home server. And any other data intensive work
  is done over iscsi mounts. Look into iscsi Qemu mangament tools , they
  should work well for UML as well. (Note you'll need to kconfig scsi and
  iscsi in your guest compile, they are off by default)
  (The general rule is any UML specific drivers tend to rot faster then
   any generic driver like iscsi or nfs)

- SKAS 3: haaa SKAS3, that one is abandoned. It will need a maintainer to
  get it alive again. How much better is it? do you have some numbers.
  (I never used it, the all point was to keep the desktop (host) with a
  distro Kernel, nvidia and all that)

> Best regards,
> 

Cheers
Boaz

------------------------------------------------------------------------------
This SF.net Dev2Dev email is sponsored by:

Show off your parallel programming skills.
Enter the Intel(R) Threading Challenge 2010.
http://p.sf.net/sfu/intel-thread-sfd
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] UML development, anyone?

[Sakari Ailus]

Hi Boaz,

Boaz Harrosh wrote:
> On 08/31/2010 01:17 AM, Sakari Ailus wrote:
>> Hello all,
>>
>> I'm a long-time UML user and lately I've become a bit worried on the
>> state of the UML. This list, as well as the -user one, is very quiet
>> these days. Is there someone still actively developing the UML? I have
>> to say I really like it! Running Linux on Linux reliably and
>> efficiently, just what I need! :-)
>>
>> I'm currently running three UML virtual machines on an server using
>> Athlon XP-M. The host is running 2.6.31 and the guests are 2.6.27.
>>
>> I cannot upgrade neither host nor guests due to these problems:
>>
>> - SKAS 3 patches available up to 2.6.31 only. Merging this to current
>> kernel might not be a problem, just some work, but:
>>
>> - Severe data corruption on guest on recent guest kernel versions. I'm
>> using a few hard disk partitions directly for UML guests. Something
>> after 2.6.27 guest breaks this quite badly. Even for small data sets
>> like 100 MiB, copying a file to a different location on guest and then
>> running md5sum on both often produces a different result.
>>
>> Do I have any options than to switch to Qemu and suffer the performance
>> penalty from emulation (kqemu is obsolete, too) or to switch to newer
>> hardware that supports kvm? Qemu performance is somewhat less than 10 %
>> of the native (or UML) performance.
>>
>> I also found something called lguest but haven't given it a try yet.
>> Looks the most interesting but that doesn't seem to have much list
>> activity either. So it works perfectly for almost everyone or has few
>> users... ;)
>>
>> How are others coping in the current situation? Or is it just me who has
>> these troubles...?
>>
>
> I hear you ...
>
> I'm using UML for years, as my developing platform for new Kernel code.
> You can imagine how it shortens the development cycle. For guests I must
> use the absolute newest Kernel all the time. So ...
>
> - Please send a detailed bug report about that data corruption so someone
>    can reproduce it. I'm not sure what block device you are using but it has
>    a bug. If you have a good version 2.6.27 and a bad version 2.6.28. It should
>    be easy to "git bisect" to the exact patch that broke the system, and should
>    be easy to fix. (Tell me if you need help, it's easy really)

I did write about this back then but it seems no-one else had the same 
problem. It's one thing to find it and the other to fix it. My wild 
guess is that it's not a change to that actual driver that broke it.

> - I probably don't have your problem because in my UMLs /home is an NFS
>    mount to the office /home server. And any other data intensive work
>    is done over iscsi mounts. Look into iscsi Qemu mangament tools , they
>    should work well for UML as well. (Note you'll need to kconfig scsi and
>    iscsi in your guest compile, they are off by default)
>    (The general rule is any UML specific drivers tend to rot faster then
>     any generic driver like iscsi or nfs)

Using NFS root would indeed be another option.

> - SKAS 3: haaa SKAS3, that one is abandoned. It will need a maintainer to
>    get it alive again. How much better is it? do you have some numbers.
>    (I never used it, the all point was to keep the desktop (host) with a
>    distro Kernel, nvidia and all that)

SKAS4 was supposed to replace it and I think I even used it for a while 
but the current SKAS4 patch is even older than SKAS3. SKAS 4, as far as 
I understand, would be a long-term solution since the /dev/mm approach 
of SKAS 3 wasn't accepted by Linus.

<URL:http://user-mode-linux.sourceforge.net/old/skas.html>

You do need that to get decent performance with UML and at some point 
some years ago perhaps, I couldn't get it even working without.

Regards,

-- 
Sakari Ailus
sakari.ailus@iki.fi

------------------------------------------------------------------------------
Start uncovering the many advantages of virtual appliances
and start using them to simplify application deployment and
accelerate your shift to cloud computing
http://p.sf.net/sfu/novell-sfdev2dev
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] UML development, anyone?

[Boaz Harrosh]

On 09/12/2010 09:34 AM, Sakari Ailus wrote:
>> - Please send a detailed bug report about that data corruption so someone
>>    can reproduce it. I'm not sure what block device you are using but it has
>>    a bug. If you have a good version 2.6.27 and a bad version 2.6.28. It should
>>    be easy to "git bisect" to the exact patch that broke the system, and should
>>    be easy to fix. (Tell me if you need help, it's easy really)
> 
> I did write about this back then but it seems no-one else had the same 
> problem. It's one thing to find it and the other to fix it. My wild 
> guess is that it's not a change to that actual driver that broke it.
> 

A "git bisect" should be the best way to find what exactly broke it. If it's
not the driver then fine, that will be found as well.
Can you reliably reproduce the data corruption? If yes then a git bisect should
be easy. Please try to see what you can do?

Thanks
Boaz

------------------------------------------------------------------------------
Start uncovering the many advantages of virtual appliances
and start using them to simplify application deployment and
accelerate your shift to cloud computing
http://p.sf.net/sfu/novell-sfdev2dev
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel