[comp.unix.ultrix] panic: out of mbuf map

zemon@felix.UUCP (10/31/87)

  Hi

  We have here a uVax II running Ultrix 1.2m .

  It was installed six month ago. We used to start it up in the morning
and shut it down at night. This week we start letting it up and running
overnight, but the day after that we got the panic message 'out of mbuf map'.

  This has happened twice, so now we are rebooting at mornings in order
to get a 'fresh' kernel every day...

  What does this error mean ? Is there some bug ? 

  Also, the message 'qerestart: restarted qe0 n' with 0 < n < 10 is logged
every now and then to /usr/adm/messages. Has this anything to do with that ?

  Any help is greatly apreciated!

-Carlos Mendioroz <tron@mrecvax.UUCP>
 {uunet|pyramid|utai}!atina!mrecvax!tron

mason@pescadero.stanford.edu (Tony Mason) (11/03/87)

In article <11437@felix.UUCP> tron@mrecvax.UUCP (Carlos Mendioroz) writes:
>
>and shut it down at night. This week we start letting it up and running
>overnight, but the day after that we got the panic message 'out of mbuf map'.

We've seen this same thing on non-ultrix machines (4.3 BSD.)  The person who
tracked it down noted that some mbufs were being allocated but not freed.
The most noticeable symptom was the workstation that crashes so runs X - and
X uses a local loopback socket for client/server communications.  Fix?
Somewere in the kernel there is a bug...

>  Also, the message 'qerestart: restarted qe0 n' with 0 < n < 10 is logged
>every now and then to /usr/adm/messages. Has this anything to do with that ?

We had six microvaxes that all did the same thing - just varying how often
they did that.  We have been carrying out FCO upgrades on our deqnas and the
problem has gone away on all but one machine (and that one had a different
revision # than the successful ones.)  The deqna board should be at wrap 
revision E1.  We were replacing with an E5.  The last board, an E6, didn't
solve the problem.

Also, our hardware service rep told me she remembered a software patch for
the problem

I hope this helps.

Tony Mason
Distributed Systems Group
Stanford University
mason@pescadero.stanford.edu

mason@Pescadero.stanford.edu (Tony Mason) (11/13/87)

Reply-Path:



In article <11437@felix.UUCP> tron@mrecvax.UUCP (Carlos Mendioroz) writes:
>
>and shut it down at night. This week we start letting it up and running
>overnight, but the day after that we got the panic message 'out of mbuf map'.

We've seen this same thing on non-ultrix machines (4.3 BSD.)  The person who
tracked it down noted that some mbufs were being allocated but not freed.
The most noticeable symptom was the workstation that crashes so runs X - and
X uses a local loopback socket for client/server communications.  Fix?
Somewere in the kernel there is a bug...

>  Also, the message 'qerestart: restarted qe0 n' with 0 < n < 10 is logged
>every now and then to /usr/adm/messages. Has this anything to do with that ?

We had six microvaxes that all did the same thing - just varying how often
they did that.  We have been carrying out FCO upgrades on our deqnas and the
problem has gone away on all but one machine (and that one had a different
revision # than the successful ones.)  The deqna board should be at wrap 
revision E1.  We were replacing with an E5.  The last board, an E6, didn't
solve the problem.

Also, our hardware service rep told me she remembered a software patch for
the problem

I hope this helps.

Tony Mason
Distributed Systems Group
Stanford University
mason@pescadero.stanford.edu

pludwig@ccng.UUCP (Paul Ludwig) (11/13/87)

Reply-Path:


We have a number of MVII running ultrix and your mbuf map error
is due to the qeo error. I would assume that the deqna (ethernet-
controller ) is going bad. Usually a mbuf map message followed by 
a qe0 retart shows that the ethernet is sending or recieving other
peoples data and not clearing its buffers which is the mbuf map
problem. This problem is over come as a problem in the new 2.0
version of ultrix that is available now from digital. I would
also having the ethernet board changed or reseat the address
chip on the board.

chuck@felix.UUCP (11/18/87)

Reply-Path:


In article <11437@felix.UUCP> tron@mrecvax.UUCP (Carlos Mendioroz) writes:
>
>and shut it down at night. This week we start letting it up and running
>overnight, but the day after that we got the panic message 'out of mbuf map'.

We've seen this same thing on non-ultrix machines (4.3 BSD.)  The person who
tracked it down noted that some mbufs were being allocated but not freed.
The most noticeable symptom was the workstation that crashes so runs X - and
X uses a local loopback socket for client/server communications.  Fix?
Somewere in the kernel there is a bug...

>  Also, the message 'qerestart: restarted qe0 n' with 0 < n < 10 is logged
>every now and then to /usr/adm/messages. Has this anything to do with that ?

We had six microvaxes that all did the same thing - just varying how often
they did that.  We have been carrying out FCO upgrades on our deqnas and the
problem has gone away on all but one machine (and that one had a different
revision # than the successful ones.)  The deqna board should be at wrap 
revision E1.  We were replacing with an E5.  The last board, an E6, didn't
solve the problem.

Also, our hardware service rep told me she remembered a software patch for
the problem

I hope this helps.

Tony Mason
Distributed Systems Group
Stanford University
mason@pescadero.stanford.edu