> What is the mechanism that holds off the sender of data until there is room in the
> transmit buffer for another packet?
No explicit mechanism exists. Only congestion avoidance implied by tcp algo.
> Do you have any suggestions on how I can debug this?
Certainly, with tcpdump.
> I've been frustrated with this
> problem for quite awhile now; the "bug" has appeared since somewhere around 2.1.9x
> or so.
Then it looks as another problem, because this particular hole was always
present.
Practical solutions are tuning tx queue length and/or
reducing window clamp (on receiving side) to lower value (sort of ~>mss*3):
32K for slow links is actual infinity and means "no flow control at all".
Alexey
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/