Are you _really_ sure?
We may not have any outstanding delayed ack's, but if the reader has now
opened the window significantly we should still send an _extra_ ack just
to tell the other end that we've opened the window. Otherwise the other
end ends up waiting with the zero-window-probe..
Note that lmbench tends to not show this, because with lmbench you have a
reader that is so quick that the reader always reads the data before the
delayed ack has had a chance to be sent out, so with lmbench the DACK
timer will always be set. But if you have a reader that is slower than the
DACK timer...
I'll make a pre-90-2 with this thing included, and we can remove it again
if it turns out that David is correct (which is likely, but I'd like to
hear more about this and I already applied the patch..)
Linus
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu