[openib-general] [SDP] having moving data on ttcp.aio connection

[Libor Michalek]

On Tue, Apr 26, 2005 at 03:36:33PM -0700, Tom Duffy wrote:
> Has anybody seen this type of error when doing SDP?
> 
>  ERR: : VMA lock <508000:65536> error <-12> <16:0:8>
> WARN: <1> <0404:2480> Error <-12> IOCB lock <65536:0>
> 
> ttcp reports:
> 
> [root at sins-stinger-10 ~]# ./ttcp -r -l 65536 -a 20
> ttcp-r: buflen = 65536 nbuf = 0 align = 16384/0 port = 5001
> ttcp-r: socket
> ttcp-r: accept from 192.168.0.26
> ttcp-r: Event error <-12> <5275648>
> ttcp-r: 0 bytes in 2.50 real seconds = 0.00 Mbit/sec +++
> ttcp-r: 2 I/O calls, usec/call = 1248114.00, calls/sec = 0.80
> ttcp-r: user: 0 sys: 41994 total: 41994 real: 2496228 (microseconds)
> 
> BTW, this with both ends on 2.6.11 with stock openib revision 2214, not
> my modified 2.6.12-rc3 version.  Also, I am using opensm for my SM (back
> to back configuration).

  The error -12 is errno ENOMEM, and the most common caused for ENOMEM
with AIO, especially on an unloaded system, is that the mlock failed,
and the most common reason is that you cannot lock as many pages as
you are attempting to lock. You should increase the amount of memory 
that the user is allowed to lock. The following command in each shell
from which you are running ttcp:

  limit memorylocked unlimited

I've already told Hal I'd add this to the README, I'll do it right now.

-Libor