Re: [lwip-users] Question about socket send

[megatron]

I think it's may not be the fault of WINXP,nor lwip,though lwip has the
problem that you describled above.It's probably  the bug  of the TCP test
softerware that I use,it doesn't  read the data that TCP has received,so the
window gets smaller and smaller.Is that possible?
Today I do the  test again on another PC,the same WINXP,the same TCP test
softeware,but I get a different result.It's interesting.The lwip doesn't get
deadlocked,but the speed of data transmission gets slow down.At the
beginning of transmission,I can get a high transmission speed , it varies 
from 120KBps to 400KBps,both in sending or receiving.After several
minutes,there are more than 100M bytes are transmitted ,when the speed gets
greatly slow down to only 5 or 6 KBps.I get two wireshark capture files.The
"32K_fast.pcap" is captured when the speed is high,the data size is 32K
bytes,the "32K_slow.pcap" is captured when slow,the same data size.
IP192.168.2.46 is the PC, IP 192.168.2.12 is lwip .The packet captured when
slow shows ,there are many duplicated ack and fast retransmission.How does
this happen? Due to occasionally run out of memery?The lwip stats shows
there are still occasionally run out of  memery.

Thanks a lot!


_______________________________________________
lwip-users mailing list
address@hidden
http://lists.nongnu.org/mailman/listinfo/lwip-users


http://www.nabble.com/file/p24581089/32K_fast.pcap 32K_fast.pcap 
http://www.nabble.com/file/p24581089/32K_slow.pcap 32K_slow.pcap 
-- 
View this message in context: 
http://www.nabble.com/Question-about-socket-send-tp24473527p24581089.html
Sent from the lwip-users mailing list archive at Nabble.com.