>On Mon, 8 Mar 1999, David Miller wrote:
>
>> People seem to have their understanding of these roles reversed.
Note: I never thought that the probe0 would assure ack reliability.
>You are right, Andrea and me where wrong. Anyway I think I got closer
As said above my last patch (or DaveM latest patch) was not intended to
assure reliability of the receiver ack, but just to don't care if the ack
from the receiver would be dropped. If it would be dropped the sender now
using probe0 is just able to request the ack again or to send out a
fragmented packet (if the snd_window wasn't really 0 yet).
And the patch from DaveM gone in 2.2.3 is sure obviously right and
_necessary_ to allow the receiver to advertise the new window ASAP, but as
I just pointed out the last _late_ night is _not_ the fix for the
_deadlock_ because I still think there _was_n't a mechanism that would
allow the connection to go ahead if the ack sent by the receiver during
the read(2) would be dropped by the netowrk.
Andrea Arcangeli
-
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/