|Subject:||[lwip-devel] Help with TCP stream errors.|
|Date:||Wed, 29 Apr 2009 17:03:56 -0400|
Can someone please help to explain these errors? I think it’s occurring because I overrun my packet driver output code since this occurs only when using a hardware generated checksum. I don’t quite understand the errors and why the PC keeps sending ACKs. Also, I can’t get out of this state without closing and reopening the socket on the PC. After this error, I can send a TCP packet (32 bytes) to ping the lwip system, and the response to the PC is the ping plus a couple of packets from the errors that have previously occurred.
The attached pcap capture is from just before the error occurs. Before this point, somewhere between 400k and 1500k can be transferred OK, but this error happens every time. With a software checksum, which is slower, I can run all day. This consists of sending large (400k) blocks to the PC with small (my) ACKs or control packets back. The socket on the PC is opened and is never expected to be closed.
This is not the same way I got the “unrecoverable errors” from bug report https://savannah.nongnu.org/bugs/?24493 but the result is the same – errors continue in the data stream until the PC closes and reopens the socket. Like the socket connection doesn’t recover because I can’t get back to an error-free state.
This is a NIOS-II GCC compiled platform. lwIP is the 1.3.0 HEAD from a couple of months back. I have been waiting to update to the current HEAD when I can quantify any change in performance.
Description: Binary data
|[Prev in Thread]||Current Thread||[Next in Thread]|