]> Gentwo Git Trees - linux/.git/commit
tcp: take care of zero tp->window_clamp in tcp_set_rcvlowat()
authorEric Dumazet <edumazet@google.com>
Fri, 3 Oct 2025 18:41:19 +0000 (18:41 +0000)
committerJakub Kicinski <kuba@kernel.org>
Mon, 6 Oct 2025 20:08:48 +0000 (13:08 -0700)
commit21b29e74ffe5a6c851c235bb80bf5ee26292c67b
treeacaf3c0e18ef261259cdc2f3893c5dc798ceab91
parent95920c2ed02bde551ab654e9749c2ca7bc3100e0
tcp: take care of zero tp->window_clamp in tcp_set_rcvlowat()

Some applications (like selftests/net/tcp_mmap.c) call SO_RCVLOWAT
on their listener, before accept().

This has an unfortunate effect on wscale selection in
tcp_select_initial_window() during 3WHS.

For instance, tcp_mmap was negotiating wscale 4, regardless
of tcp_rmem[2] and sysctl_rmem_max.

Do not change tp->window_clamp if it is zero
or bigger than our computed value.

Zero value is special, it allows tcp_select_initial_window()
to enable autotuning.

Note that SO_RCVLOWAT use on listener is probably not wise,
because tp->scaling_ratio has a default value, possibly wrong.

Fixes: d1361840f8c5 ("tcp: fix SO_RCVLOWAT and RCVBUF autotuning")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com>
Reviewed-by: Neal Cardwell <ncardwell@google.com>
Link: https://patch.msgid.link/20251003184119.2526655-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
net/ipv4/tcp.c