doc: Fix up note about missing tcpi_snd_wnd in man page

The behaviour without tcpi_snd_wnd changed: the only difference now
is the advertised window, which corresponds to the queried sending
buffer size.

Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
This commit is contained in:
Stefano Brivio 2021-10-12 22:56:36 +02:00
parent c9d57fee7c
commit 4869d309e1

10
passt.1
View file

@ -695,13 +695,9 @@ option, see \fBtcp\fR(7). Before Linux 5.3, i.e. before Linux kernel commit
8f7baad7f035 ("tcp: Add snd_wnd to TCP_INFO"), the sending window
(\fIsnd_wnd\fR field) is not available.
If the sending window cannot be queried, it will always be announced as a fixed
value to guest or target namespace (14 600 bytes, suggested by RFC 6928), and
segments received by guest or target namespace will be acknowledged as soon as
the corresponding payload is enqueued to the corresponding socket. The normal
behaviour is to acknowledge segments only as the remote peer acknowledges the
corresponding payload, in order to reflect the congestion control dynamic back
to the sender. This might affect throughput of TCP connections.
If the sending window cannot be queried, it will always be announced as the
current sending buffer size to guest or target namespace. This might affect
throughput of TCP connections.
.SH LIMITATIONS