Can somebody tell me what skb_checksum does? How is it different from
computing a TCP checksum and why is it necessary?
What does it return?
In tcp_input.c I see the following lines
if (tcp_checksum_complete_user(sk,, skb))
goto csum_error;
tcp_checksum_complete goes through a series of function calls that finally
result in a call to skb_checksum?
Why does it have to return a 0?
The reason for these questions ..........
I am facing a peculiar problem with a kernel version 2.4.25.
My TCP sessions occasionally hang due to this error. If I have two
machines talking via a TCP session and attempt to send a largish amount of
data quickly results in this error. Any subsequent packets sent on this
connection (where a checksum error has been detected) are discarded by the
receiving machine. The connection is still maintained.
I tried using an ftp with large files to simulate the problem and I
noticed that it happens invariably when a packet has been dropped and the
sending machine tries to resend it. It always seems to have an
skb_checksum error?
I can open new connections and do telnet etc.... and as long as the data
rate is moderate things are fine.
I have tried bypassing the checksum error and I've noticed that the
received packet with the skb checksum error has some discrepancy in the
data. The size remains the same but a character or two is corrupted in the
data buffer.
I would have suspected some hardware issues but if I use UDP (tftp etc) I
have no problems transmitting or receiving data (with a larger volume and
a faster rate).
I posted an abbreiviated query related to this on the linux-net list but I
got no answers. Maybe somebody here could help.
Thanks
S
--
Kernelnewbies: Help each other learn about the Linux kernel.
Archive: http://mail.nl.linux.org/kernelnewbies/
FAQ: http://kernelnewbies.org/faq/
