__rmnet_map_ingress_handler() casts skb->data to a struct rmnet_map_header and trusts the on-wire pkt_len field without verifying that the skb actually contains the bytes it is about to dereference. When ingress deaggregation is disabled (RMNET_FLAGS_INGRESS_DEAGGREGATION cleared), the skb is handed to __rmnet_map_ingress_handler() directly, bypassing the length check that rmnet_map_deaggregate() performs. Three distinct reads past skb->len are reachable, so three guards are needed: - The MAP header (flags, mux_id, pkt_len) is read on entry, before either csum branch and even on the command and non-csum paths, so a runt frame faults here regardless of data_format. Guard the header first. - On the MAPv5 next-header path, rmnet_map_process_next_hdr_packet() reads the v5 csum header at (skb->data + sizeof(map_header)); the entry guard only covers the header, so this needs its own check. - On the MAPv4 path, rmnet_map_checksum_downlink_packet() reads the csum trailer at (skb->data + len); guard it under the same CKSUMV4 condition the read uses, or valid non-csum frames would be dropped. Each guard uses pskb_may_pull() before its read. The MAPv4 check uses ntohs(pkt_len) (== payload + pad) directly rather than the derived len, so it is unaffected by the u16 underflow in len when pad > pkt_len. Well-formed frames always carry the header/trailer they declare, so only malformed packets are dropped; this mirrors the length check rmnet_map_deaggregate() already performs on the deaggregation path. BUG: KASAN: slab-out-of-bounds in rmnet_map_checksum_downlink_packet Read of size 1 at addr ffff88801118ed00 by task exploit/147 Call Trace: ... rmnet_map_checksum_downlink_packet (drivers/net/ethernet/qualcomm/rmnet/rmnet_map_data.c:413) __rmnet_map_ingress_handler (drivers/net/ethernet/qualcomm/rmnet/rmnet_handlers.c:96) rmnet_rx_handler (drivers/net/ethernet/qualcomm/rmnet/rmnet_handlers.c:129) __netif_receive_skb_core.constprop.0 (net/core/dev.c:6089) netif_receive_skb (net/core/dev.c:6460) tun_get_user (drivers/net/tun.c:1955) tun_chr_write_iter (drivers/net/tun.c:2001) vfs_write (fs/read_write.c:688) ksys_write (fs/read_write.c:740) do_syscall_64 (arch/x86/entry/syscall_64.c:94) ... Fixes: ceed73a2cf4a ("drivers: net: ethernet: qualcomm: rmnet: Initial implementation") Reported-by: Weiming Shi Assisted-by: Claude:claude-opus-4-8 Signed-off-by: Xiang Mei --- drivers/net/ethernet/qualcomm/rmnet/rmnet_handlers.c | 11 +++++++++++ 1 file changed, 11 insertions(+) diff --git a/drivers/net/ethernet/qualcomm/rmnet/rmnet_handlers.c b/drivers/net/ethernet/qualcomm/rmnet/rmnet_handlers.c index 9f3479500f85..83d011ed5942 100644 --- a/drivers/net/ethernet/qualcomm/rmnet/rmnet_handlers.c +++ b/drivers/net/ethernet/qualcomm/rmnet/rmnet_handlers.c @@ -61,6 +61,9 @@ __rmnet_map_ingress_handler(struct sk_buff *skb, u16 len, pad; u8 mux_id; + if (!pskb_may_pull(skb, sizeof(*map_header))) + goto free_skb; + if (map_header->flags & MAP_CMD_FLAG) { /* Packet contains a MAP command (not data) */ if (port->data_format & RMNET_FLAGS_INGRESS_MAP_COMMANDS) @@ -84,11 +87,19 @@ __rmnet_map_ingress_handler(struct sk_buff *skb, if ((port->data_format & RMNET_FLAGS_INGRESS_MAP_CKSUMV5) && (map_header->flags & MAP_NEXT_HEADER_FLAG)) { + if (!pskb_may_pull(skb, sizeof(*map_header) + + sizeof(struct rmnet_map_v5_csum_header))) + goto free_skb; if (rmnet_map_process_next_hdr_packet(skb, len)) goto free_skb; skb_pull(skb, sizeof(*map_header)); rmnet_set_skb_proto(skb); } else { + if (port->data_format & RMNET_FLAGS_INGRESS_MAP_CKSUMV4 && + !pskb_may_pull(skb, sizeof(*map_header) + + ntohs(map_header->pkt_len) + + sizeof(struct rmnet_map_dl_csum_trailer))) + goto free_skb; /* Subtract MAP header */ skb_pull(skb, sizeof(*map_header)); rmnet_set_skb_proto(skb); -- 2.43.0