Data throughput issues. Increasing throughput.

[Pau Espin Pedrol]

Hi,

On 04/04/18 16:59, Dave Ellis wrote:
> After that the instructions mention using openggsn.  I'm not sure if 
> there is further configuration required for the kernel gtp module - do I 
> need to create a tunnel? Or should this be done by osmo-ggsn as it did 
> for the gtpu-mode tun setting?

I'm not sure here since I didn't find the opportunity to use the gtp 
kernel module myself yet. Looking at code in osmo-ggsn gtp-kernel.c, 
gtp_kernel_init() calls gtp_dev_create() which seems to create a GTP 
tunnel device, so osmo-ggsn seems to be creating the device.

> 
> [Thread debugging using libthread_db enabled]
> Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
> <0002> ggsn.c:203 APN(S1U): Starting
> <0002> ggsn.c:281 APN(S1U): Opening Kernel GTP device gtp0
> <0002> gtp-kernel.c:75 Initialized GTP kernel mode (genl ID is 30)
> 
> Program received signal SIGSEGV, Segmentation fault.
> gtp_kernel_init (gsn=0x0, devname=0x6a6fe0 "gtp0",
>      prefix=prefix at entry=0x6a76e0, ipup=0x0) at gtp-kernel.c:94
> 94if (gtp_dev_create(-1, devname, gsn->fd0, gsn->fd1u) < 0) {
> 

This looks like a bug in the code (gsn=NULL). Please open a ticket in 
redmine in OsmoGGSN project to remember to have a look at it. Also 
compiling everything with CFLAGS="-g -O0", then running under gdb and 
providing a full backtrace may help pinpoint the issue quicker.

Feel free to open a new task also about the bad performance when running 
without gtp kernel support and providing some performance figures, and 
if possible a linux perf output of the process.


-- 
- Pau Espin Pedrol <pespin at sysmocom.de>         http://www.sysmocom.de/
=======================================================================
* sysmocom - systems for mobile communications GmbH
* Alt-Moabit 93
* 10559 Berlin, Germany
* Sitz / Registered office: Berlin, HRB 134158 B
* Geschaeftsfuehrer / Managing Director: Harald Welte