Project Home
Project Home
Wiki
Wiki
Discussion Forums
Discussions
Project Information
Project Info
Forum Topic - Re: io-pkt leak data section - ppc8260 ethernet driver: (2 Items)
   
Re: io-pkt leak data section - ppc8260 ethernet driver  
Yes, but this is nor relevant to the problem that you originally reported
- the data size was increasing while running your driver. Running df in a
loop is not a normal situation. I¹ll agree that running df causes io-pkt
data size to increase, but that is not relevant to your driver.
 


On 2016-11-23, 9:29 AM, "Thomas Luetzel" <community-noreply@qnx.com> wrote:

>Hi, yes the VM is connected to the a network but if the VM is running
>without the shell script
>
>run.sh
>===============
>while [ 0 -lt 1 ]
>do
>   df >> /dev/null
>done
>
>the memory increases very fast!
>
>To the answer to Hugh - I am not running any BSP on a target hardware! I
>can reproduce this issue on a VM provided from the QNX homepage. My DUT
>(device under test) is delivered now from QNX, nothing implemented
>installed done from my side!
>
>Attched you will also find the timeline tracelog. I've marked a REPLY
>BLOCED from io-pkt by the df procress. This happens multiple times.
>
>Anything else what I can do to generate log which will help to isolate
>this behaviour.
>
>Thanks in advance
>
>
>
>_______________________________________________
>
>Networking Drivers
>http://community.qnx.com/sf/go/post117176
>To cancel your subscription to this discussion, please e-mail
>drivers-networking-unsubscribe@community.qnx.com

Re: io-pkt leak data section - ppc8260 ethernet driver  
Good Morning Hugh,

yes I agree, running "df" in a loop isn't a normal situation. this was just used to show and reproduce the leak in a 
fast way.
And yes I also agree - how this topic ended up, it has nothing to do with the original subject.

Somehow in our application the same thing happens but much more slowy with the result that after a couple of months 
systems performing a reset - critical for customers.

Still I don't get the relation to our network driver? As the vmware is now based on standard libs and binaries. I would 
assume. That if QNX Software Systems can provide us the io-pkt based on QNX 6.6.0 compiled for a PPC(BE) and all the 
related libs (libc.so.x ....) it should be possible to use it and test if this leak is still present.

From the currents test with 6.6.0 what I can confirm and Nick Reilly confirmed, this issue is not happening.
So from my understanding io-pkt must be replaced to solve my issue and not my network driver.

Or we do find a way what causes the leak when df send interacts with io-pkt and I can hunt for the same issue in our 
application and try to implement a workaround.

Kind Regards
Thomas