|From:||Enrico Murador - Research & Development - CET|
|Subject:||[lwip-users] Pbuf pool size and network interface|
|Date:||Thu, 23 Jun 2011 17:08:40 +0200|
|User-agent:||Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:18.104.22.168) Gecko/20110616 Thunderbird/3.1.11|
I'm using the callback API's of LWIP, with a network interface based on the Coldfire MCF5223x driver.
The Coldfire interface is DMA based, and the driver pre-allocates a number of pbufs from the pool that will be "filled" by DMA.
the maximum number of buffers that my interface can allocate is 4 (of PBUF_POOL_BUFSIZE maximum size), as the original driver.
By watching the Lwip code (only of callback ApI's), my network interface is the only place where PBUF_POOL type buffer allocation is used
(excluding SLIP interface, that I don't use).
Therefore, might be correct to define PBUF_POOL_SIZE not more than the maximum number of netif's available buffers (4 in my example)?
Supposing I have a 500 bytes TCP_WND size, and my server application allows 10 simultaneous TCP connections;
do I need at least (10 * 500) bytes of total pbuf pool space (available to the network interface) to be sure to avoid "TCP deadlocks"?
Thank you for your patience
|[Prev in Thread]||Current Thread||[Next in Thread]|