Re: [LARTC] Detecting stale TCP/IP

Linux Advanced Routing and Traffic Control

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Jasper,

Thanks. This works absolutelly perfectly!

I found that simply waiting for the TIOCOUTQ to fall to zero gives a perfect indication of data having been sent correctly, and if not, that the connection is stale.

I have put this into a:

while (test) sleep(1);

But this waists time. With a ping of 300ms being common, a sleep of 1000ms is way to much to check every send, so I can only check agregates of sending. Would anybody know of a better method?

Ben

Jasper Spaans wrote:

On Tue, Jul 29, 2003 at 08:56:24AM +0000, Ben Clewett wrote:

Dear lartc,

I have a TCP/IP server connecting to a GPRS PDA.

Unfortuntatelly GPRS seems unstable, and for this reason or another, a interupt to the TCP/IP connection (like switching the clinet off suddenly) does not terminate the TCP/IP connection. It can still be seen in 'netstat', and a test by the application shows it present. It may time out after about 40 minutes.

My real problem is a call to 'write' in non-blocking mode returns success when the other end of the TCP/IP connection is not there.

Does any member know a reference to where I can test for the reply TCP/IP 'ACK' packet for this write call, and therefore timeout and terminate the connection?


You can try using the ioctl TIOCOUTQ, see socket(7). Although this manpage
states otherwise, at least on 2.4.21 and newer it returns the number of
unacked bytes.

Implement a select-loop, and see if the value of this ioctl changes. If not,
close the connection, and the kernel should handle the rest.


VrGr,




[Index of Archives]     [LARTC Home Page]     [Netfilter]     [Netfilter Development]     [Network Development]     [Bugtraq]     [GCC Help]     [Yosemite News]     [Linux Kernel]     [Fedora Users]
  Powered by Linux