[plug] NBD issues

[Craig Ringer]

Hi folks

I'm running into some odd issues with NBD that I haven't seen before, 
and was hoping someone might be able to shed some light.

If the NBD client I'm using dies without informing the server (ie a 
hardware reset), the nbd server seems to fail to accept any more 
connections from that IP. This is a bit of an issue, but not the only one.

I'm also seeing kernel oops-es when disconnecting an NBD device:

NBD_DISCONNECT
NBD: receive - sock=-1041996592 at buf=-1041375512, size=16 returned 0.
NBD: Recv control failed.(result 0)
req should never be null
nbd: shutting down socket
Unable to handle kernel NULL pointer dereference at virtual address 00000008
  printing eip:
c0193b4b
*pde = 00000000
Oops: 0000
CPU:    0
EIP:    0010:[<c0193b4b>]    Not tainted
EFLAGS: 00010282
eax: 00000000   ebx: c0311fb0   ecx: 00000001   edx: c1ef2000
esi: c0311f90   edi: 00000001   ebp: 00000004   esp: c1eddf18
ds: 0018   es: 0018   ss: 0018
Process nbd-client (pid: 115, stackpage=c1edd000)
Stack: c0262249 00000000 c191c4ac 00000001 00000000 00000000 00030002 
00000000
        00000000 00000000 00000000 00000000 00000000 00000000 00000000 
00000000
        00000000 00000000 00000000 00000000 00000000 00000000 00000000 
00000000
Call Trace:    [<c01379df>] [<c013e251>] [<c0108967>]

Code: 8b 50 08 6a 03 50 ff 52 28 c7 46 0c 00 00 00 00 83 c4 0c 89


It strikes me that this isn't good. If nobody's seen this before, I'll 
have to find out how to make a bug report that's useful to the kernel 
folks, to be sent via the nbd maintainer.

It's been working like a dream before, so I'm a little puzzled as to why 
it's playing up now. I may have to look into alternate methods of 
fetching the data I need - however NBD is very handy for a variety of 
reasons.

Craig Ringer

_______________________________________________
plug mailing list
plug at plug.linux.org.au
http://mail.plug.linux.org.au/cgi-bin/mailman/listinfo/plug