rxrpc: Fix congestion control algorithm
authorDavid Howells <dhowells@redhat.com>
Fri, 3 May 2024 15:07:39 +0000 (16:07 +0100)
committerJakub Kicinski <kuba@kernel.org>
Wed, 8 May 2024 15:05:03 +0000 (08:05 -0700)
commitba4e103848d3a2a28a0445e39f4a9564187efe54
tree116c13952ec07ff8e10b022d78c6d714fa8b5b86
parent9a169c267e946b0f47f67e8ccc70134708ccf3d4
rxrpc: Fix congestion control algorithm

Make the following fixes to the congestion control algorithm:

 (1) Don't vary the cwnd starting value by the size of RXRPC_TX_SMSS since
     that's currently held constant - set to the size of a jumbo subpacket
     payload so that we can create jumbo packets on the fly.  The current
     code invariably picks 3 as the starting value.

     Further, the starting cwnd needs to be an even number because we ack
     every other packet, so set it to 4.

 (2) Don't cut ssthresh when we see an ACK come from the peer with a
     receive window (rwind) less than ssthresh.  ssthresh keeps track of
     characteristics of the connection whereas rwind may be reduced by the
     peer for any reason - and may be reduced to 0.

Fixes: 1fc4fa2ac93d ("rxrpc: Fix congestion management")
Fixes: 0851115090a3 ("rxrpc: Reduce ssthresh to peer's receive window")
Signed-off-by: David Howells <dhowells@redhat.com>
Suggested-by: Simon Wilkinson <sxw@auristor.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
Reviewed-by: Jeffrey Altman <jaltman@auristor.com <mailto:jaltman@auristor.com>>
Link: https://lore.kernel.org/r/20240503150749.1001323-2-dhowells@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
net/rxrpc/ar-internal.h
net/rxrpc/call_object.c
net/rxrpc/input.c