ip_vs_ftp_out() can link an expectation to an unconfirmed conntrack. A later extension add on the same packet, such as seqadj in ip_vs_update_conntrack() or NAT, then reallocates ct->ext and leaves the expectation pointing into the freed area: BUG: KASAN: slab-use-after-free in nf_ct_unlink_expect_report+0x490/0x4c0 Write of size 8 at addr ffff888004dd4e48 by task syz.2.17/649 Check the conntrack at the start of ip_vs_ftp_out() and drop the packet unless the conntrack is confirmed, is not a template and has the seqadj extension that ip_vs_update_conntrack() adds on the first packet. Fixes: b25adce16064 ("ipvs: correct usage/allocation of seqadj ext in ipvs") Suggested-by: Julian Anastasov Suggested-by: Pablo Neira Ayuso Cc: stable@vger.kernel.org Signed-off-by: Theodor Arsenij Larionov Trichkine --- v4: check the conntrack at the start of ip_vs_ftp_out(), as suggested by Julian. A server reply that fails the check is now dropped even when it is not a PASV/EPSV response. Without a conntrack, e.g. with a notrack rule, only PASV/EPSV replies were dropped before. v3: https://lore.kernel.org/netfilter-devel/20261116101506.98853-1-theodorlarionov@gmail.com/ net/netfilter/ipvs/ip_vs_ftp.c | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/net/netfilter/ipvs/ip_vs_ftp.c b/net/netfilter/ipvs/ip_vs_ftp.c index 44de99f4d187..216be082162f 100644 --- a/net/netfilter/ipvs/ip_vs_ftp.c +++ b/net/netfilter/ipvs/ip_vs_ftp.c @@ -29,6 +29,7 @@ #include #include #include +#include #include #include #include @@ -283,6 +284,12 @@ static int ip_vs_ftp_out(struct ip_vs_app *app, struct ip_vs_conn *cp, if (cp->state != IP_VS_TCP_S_ESTABLISHED) return 1; + ct = nf_ct_get(skb, &ctinfo); + if (!ct || !nf_ct_is_confirmed(ct) || nf_ct_is_template(ct)) + return 0; + if (!nfct_seqadj(ct)) + return 0; + /* Linear packets are much easier to deal with. */ if (skb_ensure_writable(skb, skb->len)) return 0; -- 2.34.1