In mptcp_pm_create_subflow_or_signal_addr(), rcu_read_(un)lock() are
used as expected to iterate over the list of local addresses, but
list_for_each_entry() was used instead of list_for_each_entry_rcu() in
__lookup_addr(). It is important to use this variant which adds the
required READ_ONCE() (and diagnostic checks if enabled).
Because __lookup_addr() is also used in mptcp_pm_nl_set_flags() where it
is called under the pernet->lock and not rcu_read_lock(), an extra
condition is then passed to help the diagnostic checks making sure
either the associated spin lock or the RCU lock is held.
Fixes: 86e39e04482b ("mptcp: keep track of local endpoint still available for each msk")
Cc: stable@xxxxxxxxxxxxxxx
Reviewed-by: Geliang Tang <geliang@xxxxxxxxxx>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@xxxxxxxxxx>
Link: https://patch.msgid.link/20241112-net-mptcp-misc-6-12-pm-v1-3-b835580cefa8@xxxxxxxxxx
Signed-off-by: Jakub Kicinski <kuba@xxxxxxxxxx>
diff --git a/net/mptcp/pm_netlink.c b/net/mptcp/pm_netlink.c
index db586a5b3866..45a2b5f05d38 100644
--- a/net/mptcp/pm_netlink.c
+++ b/net/mptcp/pm_netlink.c
@@ -524,7 +524,8 @@ __lookup_addr(struct pm_nl_pernet *pernet, const struct mptcp_addr_info *info)
{
struct mptcp_pm_addr_entry *entry;
- list_for_each_entry(entry, &pernet->local_addr_list, list) {
+ list_for_each_entry_rcu(entry, &pernet->local_addr_list, list,
+ lockdep_is_held(&pernet->lock)) {
if (mptcp_addresses_equal(&entry->addr, info, entry->addr.port))
return entry;
}
