Hi Sasha, On 03/03/2025 17:29, Sasha Levin wrote: > From: "Matthieu Baerts (NGI0)" <matttbe@xxxxxxxxxx> > > [ Upstream commit db75a16813aabae3b78c06b1b99f5e314c1f55d3 ] > > Recently, some fallback have been initiated, while the connection was > not supposed to fallback. > > Add a safety check with a warning to detect when an wrong attempt to > fallback is being done. This should help detecting any future issues > quicker. > > Acked-by: Paolo Abeni <pabeni@xxxxxxxxxx> > Signed-off-by: Matthieu Baerts (NGI0) <matttbe@xxxxxxxxxx> > Link: https://patch.msgid.link/20250224-net-mptcp-misc-fixes-v1-3-f550f636b435@xxxxxxxxxx > Signed-off-by: Jakub Kicinski <kuba@xxxxxxxxxx> > Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx> Thank you for backporting this patch, but is it OK to delay it a bit on v6.13 and older please? This patch depends on its parent commit, commit 8668860b0ad3 ("mptcp: reset when MPTCP opts are dropped after join"), on kernels >=v5.19, to avoid a WARN(). > diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h > index b70a303e08287..7e2f70f22b05b 100644 > --- a/net/mptcp/protocol.h > +++ b/net/mptcp/protocol.h > @@ -1194,6 +1194,8 @@ static inline void __mptcp_do_fallback(struct mptcp_sock *msk) > pr_debug("TCP fallback already done (msk=%p)\n", msk); > return; > } > + if (WARN_ON_ONCE(!READ_ONCE(msk->allow_infinite_fallback))) An alternative is to drop the WARN_ON_ONCE() when backporting this to stable, and only keep the condition here: if (!READ_ONCE(msk->allow_infinite_fallback)) > + return; > set_bit(MPTCP_FALLBACK_DONE, &msk->flags); > } > Cheers, Matt -- Sponsored by the NGI0 Core fund.
