[lwip-users] RE: PBUF Leak Problems...

[Ramanathan Ramadass]

Tom,

 

My situation is different. As mentioned, I don’t use the whole lwIP framework but only the TCP layer and implement my application using the event callbacks from that layer. The limits that you run into at the various interfaces should be easily configurable enabling you to get around them.

 

What I am interested in is the correctness of the OOSEQ handling code. Reassembly is of course, pretty tricky and there could be a problem there causing us to lose pbufs.

 

Btw – I use only pbuf pool (no pbuf ram/rom/ref etc.) configured such that sizeof(each node in pool) == sizeof(struct pbuf) + PBUF_POOL_BUFSIZE == 2K.

 

Regards

Ram

 

---

From: Thomas Catalino [mailto:address@hidden
Sent: Wednesday, September 19, 2007 4:57 PM
To: address@hidden; address@hidden
Subject: Re: PBUF Leak Problems...

 

Ram -

We are seeing a lot of OOSEQ packets. It it looking to us like when this happens our processing load / cpu utilization is increasing significantly. Have you observed anything like this?

If true, this seems to be having a cascading effect where the application does not get enough cpu time to drain data from the API quickly enough -- which seems to be resulting in us overflowing the mailbox between the stack and the API (we are using the sockets API). This looks like where we are losing pbufs. At least it is the first place we have found.

Thanks,
Tom