On Sun, Feb 16, 2020 at 10:56:54PM +0530, Amol Grover wrote: > ip_set_type_list is traversed using list_for_each_entry_rcu > outside an RCU read-side critical section but under the protection > of ip_set_type_mutex. > > Hence, add corresponding lockdep expression to silence false-positive > warnings, and harden RCU lists. > > Signed-off-by: Amol Grover <frextrite@xxxxxxxxx> > --- Hi David Could you please go through this patch aswell? This patch was directed to preemptively fix the _suspicious RCU usage_ warning which is now also being reported by Kernel Test Robot. [ 11.654186] ============================= [ 11.654619] WARNING: suspicious RCU usage [ 11.655022] 5.6.0-rc1-00179-gdb4ead2cd5253 #1 Not tainted [ 11.655583] ----------------------------- [ 11.656001] net/netfilter/ipset/ip_set_core.c:89 RCU-list traversed in non-reader section!! Thanks Amol > net/netfilter/ipset/ip_set_core.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/net/netfilter/ipset/ip_set_core.c b/net/netfilter/ipset/ip_set_core.c > index cf895bc80871..97c851589160 100644 > --- a/net/netfilter/ipset/ip_set_core.c > +++ b/net/netfilter/ipset/ip_set_core.c > @@ -86,7 +86,8 @@ find_set_type(const char *name, u8 family, u8 revision) > { > struct ip_set_type *type; > > - list_for_each_entry_rcu(type, &ip_set_type_list, list) > + list_for_each_entry_rcu(type, &ip_set_type_list, list, > + lockdep_is_held(&ip_set_type_mutex)) > if (STRNCMP(type->name, name) && > (type->family == family || > type->family == NFPROTO_UNSPEC) && > -- > 2.24.1 >