lwip-users
[Top][All Lists]
Advanced

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

Re: [lwip-users] Decreasing window size problem


From: Per-Henrik Lundblom
Subject: Re: [lwip-users] Decreasing window size problem
Date: Fri, 8 Jun 2007 10:58:03 +0200
User-agent: Mutt/1.5.13 (2006-08-11)

* Per-Henrik Lundblom <address@hidden> [070608 10:21]:
 
> > > In my case, A's window update is lost by B so that B doesn't know that A
> > > now accepts data packets with a size equal or larger than the packet in
> > > B's unsent queue. If A's window whould have been 0 bytes, then
> > > implementation of TCP persist timer into lwIP whould have solved the
> > > problem by letting B probe A for a window update. Now when A's window
> > > update is lost, I'm in a deadlock situation.
> > 
> > Hmm, yes, I can see what you're getting at.  The only solution here is
> > to get B to do zero-window style probes to get a window update from A.
> > Normally it would only have to do this when it actually has a zero
> > window available to send into, but we need it to probe whenever it has
> > seen a window that prevents it sending.  Splitting the data might also
> > work, but would rather complicate matters, and probably add more code.  
> 
> I will look into implementing the persistent timer and using the same
> mechanism for sending probes without the remote peer having a zero
> window.

Mmm, the probes are basically splitted packets from the unsent queue
with a data length of 1 byte. So, the splitting must still be performed
=/

I didn't mention it but everything I already have described assumes that
the unacked queue is empty.

/PH

--
Per-Henrik Lundblom           epost: address@hidden
telefon: 0733-20 71 26        hemsida: www.whatever.nu





reply via email to

[Prev in Thread] Current Thread [Next in Thread]