[uml-devel] Strange behaviour in eth assignments

[uml-devel] Strange behaviour in eth assignments

[Peter Hovorka]

Hi,

Jeff told me to drop a note here about the following occurence:

Having compiled a 2.6.16.27 guest kernel with a bb1 patchset, I was 
unable to bring up an eth0 via tun/tap. The kernel runs well, but a 
command line of

./um2.6.16.27-bb1 ubd0=root_fs ubd1=swap_fs eth0=tuntap,tap112 mem=100M

starts the instance well, but a command inside the instance of 'ifconfig 
-a' results in an 'eth2' device being ready to be brought up. I haven't 
found a way to bring up an 'eth0' or an 'eth1' device inside the 
instance.

Has anyone had this kind of problem before? I've got no clue about its 
cause, this is what I found out until now:

- The problem seems to be independent from the Host Kernel. Using the 
above Guest Kernel, the system will always prepare eth2.

- The problem seems to be independent from most of the other options in 
the Kernel config as there are statically linked, network options etc.

- The problem is not (!) present with older Kernels. I did run some 
tests with other Kernels, the following ones failed to bring up eth0:

2.6.16.27 with bb1
2.6.16.9 with bs2

The following versions were ok in bringing up eth0 as usual:

2.6.14.3 with bs3
2.6.13.4 with bs5

- I haven't tried non-SKAS3 usage yet

As Jeff told me yesterday that there was possibly no change in that 
region of the UML code, I suspect that any kind of change in the recent 
Kernel sources has caused this artefact to show, sadly I haven't got 
the faintest idea of Kernel development myself, so I'm writing this to 
the devel list.

Kind regards,
Peter












-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
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] Strange behaviour in eth assignments

[Blaisorblade]

On Thursday 17 August 2006 12:56, Peter Hovorka wrote:
> Hi,
>
> Jeff told me to drop a note here about the following occurence:
>
> Having compiled a 2.6.16.27 guest kernel with a bb1 patchset, I was
> unable to bring up an eth0 via tun/tap. The kernel runs well, but a
> command line of
>
> ./um2.6.16.27-bb1 ubd0=root_fs ubd1=swap_fs eth0=tuntap,tap112

Do you really mean "tap112"? Nobody I think tried this. I've checked for bugs 
in parsing, but there is none. Please give more details and try a saner 
setting, or elaborate on the reason of this strange setting. At least let's 
determine if changing it causes the bug not to show up.

The other possibility is that something in the guest image (I'm not talking 
about the kernel, but about the rootfs) is calling ifrename. But since some 
kernels don't do that, this is probably not the case.

> mem=100M 
You'd better use a round amount (128 or 96M), I don't know if that's a problem 
but that's strange.

> starts the instance well, but a command inside the instance of 'ifconfig
> -a' results in an 'eth2' device being ready to be brought up. I haven't
> found a way to bring up an 'eth0' or an 'eth1' device inside the
> instance.

> Has anyone had this kind of problem before? I've got no clue about its
> cause, this is what I found out until now:

> - The problem is not (!) present with older Kernels. I did run some
> tests with other Kernels, the following ones failed to bring up eth0:

> 2.6.16.27 with bb1
> 2.6.16.9 with bs2

> The following versions were ok in bringing up eth0 as usual:

> 2.6.14.3 with bs3
> 2.6.13.4 with bs5

> - I haven't tried non-SKAS3 usage yet

> As Jeff told me yesterday that there was possibly no change in that
> region of the UML code, I suspect that any kind of change in the recent
> Kernel sources has caused this artefact to show, sadly I haven't got
> the faintest idea of Kernel development myself, so I'm writing this to
> the devel list.

If this were, say, a buffer overflow, or a bug because tap112 does not exist, 
the source causes "unspecified behaviour", so totally unrelated changes can 
change the actual behaviour to change.

> Kind regards,
> Peter


-- 
Inform me of my mistakes, so I can keep imitating Homer Simpson's "Doh!".
Paolo Giarrusso, aka Blaisorblade
http://www.user-mode-linux.org/~blaisorblade
Chiacchiera con i tuoi amici in tempo reale! 
 http://it.yahoo.com/mail_it/foot/*http://it.messenger.yahoo.com 


-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
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] Strange behaviour in eth assignments

[Peter Hovorka]

Hi,

> > ./um2.6.16.27-bb1 ubd0=root_fs ubd1=swap_fs eth0=tuntap,tap112
>
> Do you really mean "tap112"? 

Yes, Sir.

> Nobody I think tried this. > I've checked  
> for bugs in parsing, but there is none. Please give more details and
> try a saner setting, or elaborate on the reason of this strange
> setting. At least let's determine if changing it causes the bug not
> to show up.

I've used taps according to the used IP in a range of 1-254 since 2002 - 
and I've never experienced any problems in doing so. So tap112 is quite 
a sane setting - for me at least it is.

I have tried other tap settings in a range from tap1-254 (and even 
simple tunctls without assigning a special number) - with the same 
results.

> The other possibility is that something in the guest image (I'm not
> talking about the kernel, but about the rootfs) is calling ifrename.
> But since some kernels don't do that, this is probably not the case.

I have tried this with three more or less different file systems after 
your mail to the list (and I should have done earlier - only thing I 
hadn't thought of), as there are:

- Ubuntu Server 6.0.6 LTS
- Fedora Core 5
- Debian Sarge

The problem seems to be Ubuntu - FC5 and Debian Sarge do well. No 
problems with them. I would like to use Ubuntu, though - so it would be 
nice to have a workaround or at least the cause for this to happen.

> > mem=100M
>
> You'd better use a round amount (128 or 96M), I don't know if that's
> a problem but that's strange.

There has never been a single problem in assigning memory outside the 
'round amount' row. I've assigned thousands of instances with 34, 35, 
36 MB or 67, 68, 69 MB or 130, 131, 132 MB and so on. 

I have tried other memory options after your comment from above, but it 
doesn't change the strange behaviour.

> If this were, say, a buffer overflow, or a bug because tap112 does
> not exist, the source causes "unspecified behaviour", so totally
> unrelated changes can change the actual behaviour to change.

Tap112 did exist, I set them up manually not by the uml_helper. As I 
said, older Kernel versions did not show any problems of this kind. I 
double and triple checked everything I could from a user's perspective 
and it took me several days until I packed my bag to look out for Jeff 
in the #uml and to finally post to uml-devel - the only thing I didn't 
take a closer look at was the file system respectively the distro.

Kind regards,
Peter

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
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] Strange behaviour in eth assignments

[Blaisorblade]

On Saturday 19 August 2006 18:38, Peter Hovorka wrote:
> Hi,
>
> > > ./um2.6.16.27-bb1 ubd0=root_fs ubd1=swap_fs eth0=tuntap,tap112
> >
> > Do you really mean "tap112"?
>
> Yes, Sir.
I didn't want to be rude, sorry.

> > Nobody I think tried this. > I've checked
> > for bugs in parsing, but there is none. Please give more details and
> > try a saner setting, or elaborate on the reason of this strange
> > setting. At least let's determine if changing it causes the bug not
> > to show up.
>
> I've used taps according to the used IP in a range of 1-254 since 2002 -
> and I've never experienced any problems in doing so. So tap112 is quite
> a sane setting - for me at least it is.

I was just surprised, and one of my thoughts was "it is barfing on 112 because 
maybe the code is stupid and expects 1 digit". So I asked just to make sure.

The other was "why is he doing this? Does he need? Probably yes, I won't take 
that for granted and instead I'll ask him".

> > The other possibility is that something in the guest image (I'm not
> > talking about the kernel, but about the rootfs) is calling ifrename.
> > But since some kernels don't do that, this is probably not the case.

> I have tried this with three more or less different file systems after
> your mail to the list (and I should have done earlier - only thing I
> hadn't thought of), as there are:

> - Ubuntu Server 6.0.6 LTS
> - Fedora Core 5
> - Debian Sarge

> The problem seems to be Ubuntu - FC5 and Debian Sarge do well.
> No 
> problems with them. I would like to use Ubuntu, though - so it would be
> nice to have a workaround or at least the cause for this to happen.

About Ubuntu, make sure ifrename is not installed and /etc/network/if-* 
and /etc/rc2.d do not mention ifrename, and that ifrename is not installed.
I can only think that some hotplug script is receiving an event through 
sysfs / hotplug / some such hell and is renaming the interface.

Bye
-- 
Inform me of my mistakes, so I can keep imitating Homer Simpson's "Doh!".
Paolo Giarrusso, aka Blaisorblade
http://www.user-mode-linux.org/~blaisorblade
Chiacchiera con i tuoi amici in tempo reale! 
 http://it.yahoo.com/mail_it/foot/*http://it.messenger.yahoo.com 


-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
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] Strange behaviour in eth assignments

[Peter Hovorka]

Hi,

> > Yes, Sir.
>
> I didn't want to be rude, sorry.

Oh no - I guess I should have attached some kind of smiley there, I 
didn't think of you as rude. It's quite contrary: I'm very glad that 
you do all this work! So don't mind!

> I was just surprised, and one of my thoughts was "it is barfing on
> 112 because maybe the code is stupid and expects 1 digit". So I asked
> just to make sure.

And that was totally right, be assured of that.

> About Ubuntu, make sure ifrename is not installed and
> /etc/network/if-* and /etc/rc2.d do not mention ifrename, and that
> ifrename is not installed. I can only think that some hotplug script
> is receiving an event through sysfs / hotplug / some such hell and is
> renaming the interface.

Thanks for your thoughts - I'll try to gather information about that in 
the ubuntu forum/s and/or their list/s about that. If I get a solution 
from somewhere, I'll post it to the uml user's list.

Once again many thanks,

Peter

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
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] Strange behaviour in eth assignments

[alessandro salvatori]

On 8/20/06, Peter Hovorka <peter@fusselt.net> wrote:
> Hi,
>
> > > Yes, Sir.
> >
> > I didn't want to be rude, sorry.
>
> Oh no - I guess I should have attached some kind of smiley there, I
> didn't think of you as rude. It's quite contrary: I'm very glad that
> you do all this work! So don't mind!

it's not the smile, it's rather that something's lost in translation :)

at first "yes, sir" always sounds a bit grumpy for us Italians, but
that's because we very seldomly use that expression and when we do it
is as if we wanted to underline we are compelled to answer yes (the
reason is that it is the typical answer a soldier would give to any
order a superior gives him)

cheers
-- 
Alessandro

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
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] Strange behaviour in eth assignments

[Peter Hovorka]

Hi there,

my sincere apologies to all of you for bothering about the Ubuntu eth 
problem. After talking to the ubuntu team we found out that the ubuntu 
installer makes some entries in /etc/iftab to help udev whilst booting 
the system. There were the two entries from the host system with eth0 
and eth1 - entries deleted the whole system runs well.
 
Kind regards,
Peter

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
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] Strange behaviour in eth assignments

[Blaisorblade]

On Saturday 26 August 2006 14:35, Peter Hovorka wrote:
> Hi there,
>
> my sincere apologies to all of you for bothering about the Ubuntu eth
> problem.

Thanks for your apologies but thanks anyway for your competent cooperation.

> After talking to the ubuntu team we found out that the ubuntu 
> installer makes some entries in /etc/iftab to help udev whilst booting
> the system. There were the two entries from the host system with eth0
> and eth1 - entries deleted the whole system runs well.

I'm starting to get sick of udev - on Ubuntu currently I'm unable to compile a 
vanilla kernel that works (I must still do more complete tests but I'm 
already beyond the "I've misconfigured something" moment).
-- 
Inform me of my mistakes, so I can keep imitating Homer Simpson's "Doh!".
Paolo Giarrusso, aka Blaisorblade
http://www.user-mode-linux.org/~blaisorblade
Chiacchiera con i tuoi amici in tempo reale! 
 http://it.yahoo.com/mail_it/foot/*http://it.messenger.yahoo.com 


-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
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] Strange behaviour in eth assignments

[Nix]

On Sat, 26 Aug 2006, blaisorblade@yahoo.it stipulated:
> I'm starting to get sick of udev - on Ubuntu currently I'm unable to compile a 
> vanilla kernel that works (I must still do more complete tests but I'm 
> already beyond the "I've misconfigured something" moment).

All udev needs is CONFIG_HOTPLUG and sysfs, as far as I know.

-- 
`In typical emacs fashion, it is both absurdly ornate and
 still not really what one wanted.' --- jdev

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
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] Strange behaviour in eth assignments

[Jeff Dike]

On Sat, Aug 19, 2006 at 04:53:43PM +0200, Blaisorblade wrote:
> Do you really mean "tap112"? Nobody I think tried this. 

I have :-)  Read The Book (and BTW, I need to figure out a way to get you 
a copy).

> If this were, say, a buffer overflow, or a bug because tap112 does not exist,
> the source causes "unspecified behaviour", so totally unrelated changes can 
> change the actual behaviour to change.

I don't see anything on the host affecting this.  Something funky is happening
when the UML network subsystem chooses names for new interfaces.

				Jeff

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
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] Strange behaviour in eth assignments

[Blaisorblade]

On Sunday 20 August 2006 02:42, Jeff Dike wrote:
> On Sat, Aug 19, 2006 at 04:53:43PM +0200, Blaisorblade wrote:
> > Do you really mean "tap112"? Nobody I think tried this.
>
> I have :-)
I forgot!
> Read The Book (and BTW, I need to figure out a way to get you 
> a copy).

I reviewed it at the time, I've not seen later changes (and not had the time - 
I'm concentrating on avoiding housekeeping and doing instead big UML jobs 
like remap_file_pages for this).

And I still need to send the document to AWL to get paid (I think it's too 
late probably... but I should!).
-- 
Inform me of my mistakes, so I can keep imitating Homer Simpson's "Doh!".
Paolo Giarrusso, aka Blaisorblade
http://www.user-mode-linux.org/~blaisorblade
Chiacchiera con i tuoi amici in tempo reale! 
 http://it.yahoo.com/mail_it/foot/*http://it.messenger.yahoo.com 


-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel