[ENBD] fixup_slot failed to find slot....

[scunacc]

Dear Peter,

> The segfault is the problem. No code may segfault. If it does,
> something is badly wrong.
> 
> > But I do indeed get a segfault and those remaining connections are not
> 
> Run a simple test - just one process. 

One process doesn't segfault.

It always segfaults on the 5th process.

> What "remaining connections"? 

I have 14 nbd-clients running to 14 different servers from the same
client machine.

The 1st 4 make connections, the remaining 10 segfault.

> If those are different clients running
> on the same driver, then the kernel has oopsed, since there is no
> other connection between them.

I see no other Kernel messages, except the one I originally included
about the fixups. I do not get that for the 1st 4.

> 
> > A couple of the processes end up as zombies, the rest
> > would hang if I tried to work with the devices they are connecting.
> 
> A device that is not connected is not working and can tell anyone who
> tries to talk to it that it is ill.  Are you maybe running in a sync
> mode?  Maybe you want "-e"

Cute :-)

> 
>      -e This option specifies that the ENBD device will transmit network
>      errors almost immediately to any higher level device, such as RAID,
>      that may want to be notified of them.  The flag is normally absent
>      and the ENBD device will hide network errors inter nally, blocking
>      new requests until they can be me


This is when I am simply running nbd-client to make the connections. No
higher process is running yet.

I haven't had time to do more analysis - what I've reported above is
still in the light of what I wrote before.

Kind regards

Derek.