Re: Entire LAN goes boo with 2.5.64

Re: Entire LAN goes boo with 2.5.64

[Jon Burgess]

Some things you might want to look at:
   Is the Linux box sending any traffic (look at the stats in ifconfig)?
   Does a packet sniffer like http://www.ethereal.com/ give any clues as to the
type of traffic on the network?
   Does the same thing occur if you run less processes, e.g. boot into run level
1 or 3?
   Are there any processes consuming an unreasonable amount of CPU time on the
Linux box?
   Is there a process which is being restarted many times a second, so top or ps
shows a radiply increasing PID?

It could be some network-aware process which has got stuck in a tight loop
sending requests to your windows box, e.g. a DHCP client.
I mention the DHCP client specifically because they sometimes get upset if you
don't enable some specific kernel networking options like CONFIG_PACKET or
CONFIG_FILTER & WinRoute might be acting as the DHCP server.

     Jon

Re: Entire LAN goes boo with 2.5.64

[Anders Widman]

> Some things you might want to look at:
>    Is the Linux box sending any traffic (look at the stats in ifconfig)?
>    Does a packet sniffer like http://www.ethereal.com/ give any clues as to the
> type of traffic on the network?
>    Does the same thing occur if you run less processes, e.g. boot into run level
> 1 or 3?
>    Are there any processes consuming an unreasonable amount of CPU time on the
> Linux box?
>    Is there a process which is being restarted many times a second, so top or ps
> shows a radiply increasing PID?

nope, nothing like that. Running just a bare minimum...

> It could be some network-aware process which has got stuck in a tight loop
> sending requests to your windows box, e.g. a DHCP client.
> I mention the DHCP client specifically because they sometimes get upset if you
> don't enable some specific kernel networking options like CONFIG_PACKET or
> CONFIG_FILTER & WinRoute might be acting as the DHCP server.

Yes,  you might be right. WinRoute is running as DCHP for the network,
and  the  problems  do  start as soon as Linux tries to lease an IP...
hm..

>      Jon


   



--------
PGP public key: https://tnonline.net/secure/pgp_key.txt

Re: Entire LAN goes boo with 2.5.64

[Ricky Beam]

On Fri, 7 Mar 2003, Anders Widman wrote:
>Yes,  you might be right. WinRoute is running as DCHP for the network,
>and  the  problems  do  start as soon as Linux tries to lease an IP...
>hm..

If you are using DHCP, it is critical that every machine have a correctly
set clock (see also: xntpd)  Otherwise, the client may see the lease expire
before it requested it. (what happens in such a case varies with the software
and version.)

--Ricky

Re: Entire LAN goes boo with 2.5.64

[Anders Widman]

> On Fri, 7 Mar 2003, Anders Widman wrote:
>>Yes,  you might be right. WinRoute is running as DCHP for the network,
>>and  the  problems  do  start as soon as Linux tries to lease an IP...
>>hm..

> If you are using DHCP, it is critical that every machine have a correctly
> set clock (see also: xntpd)  Otherwise, the client may see the lease expire
> before it requested it. (what happens in such a case varies with the software
> and version.)

Yes,   but  everything  works very good with 2.4.x kernels (all I have
tested  so far). So the problems comes with 2.5.64.. And, there is not
much network traffic either when this happens...

> --Ricky


   



--------
PGP public key: https://tnonline.net/secure/pgp_key.txt

Re: Entire LAN goes boo with 2.5.64

[Anders Widman]

Tried  with  a  Realtek  8139B and the Intel Pro100+ adapter. The same
thing   happens.   The   LAN   goes  crazy  and all programs trying to
access or use the LAN on the Linuxbox goes super-slow or crashes.

I am rather lost when it comes to where I should begin to look.

Have not compiled in IPX, network filtering and most other things. The
only  network  card  I  have compiled in is the Rtl8139 and the Becker
Intel Pro100+ driver.

Here is my net config: http://tnonline.net/conf.png

I  have  not compiled in ACPI or APM or APIC. And they are disabled in
BIOS too.

//Anders


> I've had this happen once, but with a 2.4 kernel. I had compiled in IPX
> and configured it for autodiscovery of frame type. On boot, it would
> flip back and forth between two different types rather fast (as fast as
> the 100base NIC could do it), freaking out every piece of networking
> equipment and every computer. See if you have IPX compiled in. 
> Otherwise, run ethereal or another sniffer to see what exactly the 
> network traffic is; that might be helpful.

> Alexander

> Anders Widman wrote:
>>    Hello,
>> 
>>    Trying  out  the  2.5.64  kernel  to try to solve some IDE specific
>>    problems  with 2.4.x kernels. Now I have another problem. We have a
>>    Windows LAN and a Windows XP with WinRoute Pro as gateway.
>> 
>>    When  booting  the linux-machine with the 2.5.64 kernel the windows
>>    machine goes to 100% cpu and the switch (Dlink) goes crazy (loosing
>>    link, other machines get 100k/s instead of 10-12MiB/s etc).
>> 
>>    I  compiled  the  2.5.64  with  as  few  options  as  possible,  no
>>    netfilter, or IPSec or similar stuff.
>> 
>>    What can be the problem?
>> 
>> 
>> --------
>> PGP public key: https://tnonline.net/secure/pgp_key.txt
>> 
>> -
>> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>> Please read the FAQ at  http://www.tux.org/lkml/


   



--------
PGP public key: https://tnonline.net/secure/pgp_key.txt

Re: Entire LAN goes boo with 2.5.64

[Anders Widman]

> Tried  with  a  Realtek  8139B and the Intel Pro100+ adapter. The same
> thing   happens.   The   LAN   goes  crazy  and all programs trying to
> access or use the LAN on the Linuxbox goes super-slow or crashes.

> I am rather lost when it comes to where I should begin to look.

> Have not compiled in IPX, network filtering and most other things. The
> only  network  card  I  have compiled in is the Rtl8139 and the Becker
> Intel Pro100+ driver.

> Here is my net config: http://tnonline.net/conf.png

> I  have  not compiled in ACPI or APM or APIC. And they are disabled in
> BIOS too.

Tried  with  all  of  them  enabled  too.  No change in state. LAN and
gateway goes down, or becomes incredibly slow and unusable...

//Anders


>> I've had this happen once, but with a 2.4 kernel. I had compiled in IPX
>> and configured it for autodiscovery of frame type. On boot, it would
>> flip back and forth between two different types rather fast (as fast as
>> the 100base NIC could do it), freaking out every piece of networking
>> equipment and every computer. See if you have IPX compiled in. 
>> Otherwise, run ethereal or another sniffer to see what exactly the 
>> network traffic is; that might be helpful.

>> Alexander

>> Anders Widman wrote:
>>>    Hello,
>>> 
>>>    Trying  out  the  2.5.64  kernel  to try to solve some IDE specific
>>>    problems  with 2.4.x kernels. Now I have another problem. We have a
>>>    Windows LAN and a Windows XP with WinRoute Pro as gateway.
>>> 
>>>    When  booting  the linux-machine with the 2.5.64 kernel the windows
>>>    machine goes to 100% cpu and the switch (Dlink) goes crazy (loosing
>>>    link, other machines get 100k/s instead of 10-12MiB/s etc).
>>> 
>>>    I  compiled  the  2.5.64  with  as  few  options  as  possible,  no
>>>    netfilter, or IPSec or similar stuff.
>>> 
>>>    What can be the problem?
>>> 
>>> 
>>> --------
>>> PGP public key: https://tnonline.net/secure/pgp_key.txt
>>> 
>>> -
>>> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
>>> the body of a message to majordomo@vger.kernel.org
>>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>> Please read the FAQ at  http://www.tux.org/lkml/


   



> --------
> PGP public key: https://tnonline.net/secure/pgp_key.txt

> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/


   



--------
PGP public key: https://tnonline.net/secure/pgp_key.txt

Re: Entire LAN goes boo with 2.5.64

[Anders Widman]

> Sounds like a bpdu storm, are you somehow looping the network?

Not  that  I  am aware of.. Running only one 8 port switch. Works good
with the 2.4.x kernels, but not with 2.5.. very odd.

//Anders

> -----Original Message-----
> From: Anders Widman [mailto:andewid@tnonline.net] 
> Sent: Thursday, March 06, 2003 6:02 AM
> To: Erik Hensema
> Cc: linux-kernel@vger.kernel.org; erik@hensema.net
> Subject: Re: Entire LAN goes boo with 2.5.64


>> Anders Widman (andewid@tnonline.net) wrote:
>>> 
>>>>    Hello,
>>> 
>>>>    Trying  out  the  2.5.64  kernel  to try to solve some IDE
> specific
>>>>    problems  with 2.4.x kernels. Now I have another problem. We have
> a
>>>>    Windows LAN and a Windows XP with WinRoute Pro as gateway.
>>> 
>>>>    When  booting  the linux-machine with the 2.5.64 kernel the
> windows
>>>>    machine goes to 100% cpu and the switch (Dlink) goes crazy
> (loosing
>>>>    link, other machines get 100k/s instead of 10-12MiB/s etc).
>>> 
>>>>    I  compiled  the  2.5.64  with  as  few  options  as  possible,
> no
>>>>    netfilter, or IPSec or similar stuff.
>>> 
>>>>    What can be the problem?
>>> 
>>>    Forgot to say I am using a Intel Pro100+ NIC and I have tested
> with
>>>    both the Becker driver and the Intel driver.

>> I've seen something similar [1] happen to a LAN with one Windows XP
>> machine running vcool: http://vcool.occludo.net/ . This is also 
>> available for Linux (http://vcool.occludo.net/VC_Linux.html). Are you
>> running this patch or a similar one?

> Nope, no vcool or anything similar. But it is very odd that the switch
> would go crazy too!

>> [1] all machines were seeing frame errors on packets > 250 bytes; it
>> was a 10 mbit coax lan.

> Using 100mbit switched network.


   



> --------
> PGP public key: https://tnonline.net/secure/pgp_key.txt

> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel"
> in the body of a message to majordomo@vger.kernel.org More majordomo
> info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/


   



--------
PGP public key: https://tnonline.net/secure/pgp_key.txt

Entire LAN goes boo with 2.5.64

[Anders Widman]

   Hello,

   Trying  out  the  2.5.64  kernel  to try to solve some IDE specific
   problems  with 2.4.x kernels. Now I have another problem. We have a
   Windows LAN and a Windows XP with WinRoute Pro as gateway.

   When  booting  the linux-machine with the 2.5.64 kernel the windows
   machine goes to 100% cpu and the switch (Dlink) goes crazy (loosing
   link, other machines get 100k/s instead of 10-12MiB/s etc).

   I  compiled  the  2.5.64  with  as  few  options  as  possible,  no
   netfilter, or IPSec or similar stuff.

   What can be the problem?


--------
PGP public key: https://tnonline.net/secure/pgp_key.txt

Re: Entire LAN goes boo with 2.5.64

[Anders Widman]

>    Hello,

>    Trying  out  the  2.5.64  kernel  to try to solve some IDE specific
>    problems  with 2.4.x kernels. Now I have another problem. We have a
>    Windows LAN and a Windows XP with WinRoute Pro as gateway.

>    When  booting  the linux-machine with the 2.5.64 kernel the windows
>    machine goes to 100% cpu and the switch (Dlink) goes crazy (loosing
>    link, other machines get 100k/s instead of 10-12MiB/s etc).

>    I  compiled  the  2.5.64  with  as  few  options  as  possible,  no
>    netfilter, or IPSec or similar stuff.

>    What can be the problem?

   Forgot to say I am using a Intel Pro100+ NIC and I have tested with
   both the Becker driver and the Intel driver.



--------
PGP public key: https://tnonline.net/secure/pgp_key.txt

Re: Entire LAN goes boo with 2.5.64

[Erik Hensema]

Anders Widman (andewid@tnonline.net) wrote:
> 
>>    Hello,
> 
>>    Trying  out  the  2.5.64  kernel  to try to solve some IDE specific
>>    problems  with 2.4.x kernels. Now I have another problem. We have a
>>    Windows LAN and a Windows XP with WinRoute Pro as gateway.
> 
>>    When  booting  the linux-machine with the 2.5.64 kernel the windows
>>    machine goes to 100% cpu and the switch (Dlink) goes crazy (loosing
>>    link, other machines get 100k/s instead of 10-12MiB/s etc).
> 
>>    I  compiled  the  2.5.64  with  as  few  options  as  possible,  no
>>    netfilter, or IPSec or similar stuff.
> 
>>    What can be the problem?
> 
>    Forgot to say I am using a Intel Pro100+ NIC and I have tested with
>    both the Becker driver and the Intel driver.

I've seen something similar [1] happen to a LAN with one Windows XP machine
running vcool: http://vcool.occludo.net/ . This is also available for Linux
(http://vcool.occludo.net/VC_Linux.html). Are you running this patch or a
similar one?

[1] all machines were seeing frame errors on packets > 250 bytes; it was a
10 mbit coax lan.
-- 
Erik Hensema <erik@hensema.net>

Re: Entire LAN goes boo with 2.5.64

[Anders Widman]

> Anders Widman (andewid@tnonline.net) wrote:
>> 
>>>    Hello,
>> 
>>>    Trying  out  the  2.5.64  kernel  to try to solve some IDE specific
>>>    problems  with 2.4.x kernels. Now I have another problem. We have a
>>>    Windows LAN and a Windows XP with WinRoute Pro as gateway.
>> 
>>>    When  booting  the linux-machine with the 2.5.64 kernel the windows
>>>    machine goes to 100% cpu and the switch (Dlink) goes crazy (loosing
>>>    link, other machines get 100k/s instead of 10-12MiB/s etc).
>> 
>>>    I  compiled  the  2.5.64  with  as  few  options  as  possible,  no
>>>    netfilter, or IPSec or similar stuff.
>> 
>>>    What can be the problem?
>> 
>>    Forgot to say I am using a Intel Pro100+ NIC and I have tested with
>>    both the Becker driver and the Intel driver.

> I've seen something similar [1] happen to a LAN with one Windows XP machine
> running vcool: http://vcool.occludo.net/ . This is also available for Linux
> (http://vcool.occludo.net/VC_Linux.html). Are you running this patch or a
> similar one?

Nope, no vcool or anything similar. But it is very odd that the switch
would go crazy too!

> [1] all machines were seeing frame errors on packets > 250 bytes; it was a
> 10 mbit coax lan.

Using 100mbit switched network.


   



--------
PGP public key: https://tnonline.net/secure/pgp_key.txt