netdev - [PATCH net] sfc: check for zero length in EF10 RX prefix

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-ID: <20230831165811.18061-1-edward.cree@amd.com>
Date: Thu, 31 Aug 2023 17:58:11 +0100
From: <edward.cree@....com>
To: <linux-net-drivers@....com>, <davem@...emloft.net>, <kuba@...nel.org>,
	<edumazet@...gle.com>, <pabeni@...hat.com>
CC: Edward Cree <ecree.xilinx@...il.com>, <netdev@...r.kernel.org>,
	<habetsm.xilinx@...il.com>
Subject: [PATCH net] sfc: check for zero length in EF10 RX prefix

From: Edward Cree <ecree.xilinx@...il.com>

When EF10 RXDP firmware is operating in cut-through mode, packet length
 is not known at the time the RX prefix is generated, so it is left as
 zero and RX event merging is inhibited to ensure that the length is
 available in the RX event.  However, it has been found that in certain
 circumstances the RX events for these packets still get merged,
 meaning the driver cannot read the length from the RX event, and tries
 to use the length from the prefix.
The resulting zero-length SKBs cause crashes in GRO since commit
 1d11fa696733 ("net-gro: remove GRO_DROP"), so add a check to the driver
 to detect these zero-length RX events and discard the packet.

Signed-off-by: Edward Cree <ecree.xilinx@...il.com>
---
 drivers/net/ethernet/sfc/rx.c | 20 +++++++++++++++-----
 1 file changed, 15 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/sfc/rx.c b/drivers/net/ethernet/sfc/rx.c
index 2375cef577e4..f77a2d3ef37e 100644
--- a/drivers/net/ethernet/sfc/rx.c
+++ b/drivers/net/ethernet/sfc/rx.c
@@ -359,26 +359,36 @@ static bool efx_do_xdp(struct efx_nic *efx, struct efx_channel *channel,
 /* Handle a received packet.  Second half: Touches packet payload. */
 void __efx_rx_packet(struct efx_channel *channel)
 {
+	struct efx_rx_queue *rx_queue = efx_channel_get_rx_queue(channel);
 	struct efx_nic *efx = channel->efx;
 	struct efx_rx_buffer *rx_buf =
-		efx_rx_buffer(&channel->rx_queue, channel->rx_pkt_index);
+		efx_rx_buffer(rx_queue, channel->rx_pkt_index);
 	u8 *eh = efx_rx_buf_va(rx_buf);
 
 	/* Read length from the prefix if necessary.  This already
 	 * excludes the length of the prefix itself.
 	 */
-	if (rx_buf->flags & EFX_RX_PKT_PREFIX_LEN)
+	if (rx_buf->flags & EFX_RX_PKT_PREFIX_LEN) {
 		rx_buf->len = le16_to_cpup((__le16 *)
 					   (eh + efx->rx_packet_len_offset));
+		/* A known issue may prevent this being filled in;
+		 * if that happens, just drop the packet.
+		 * Must do that in the driver since passing a zero-length
+		 * packet up to the stack may cause a crash.
+		 */
+		if (unlikely(!rx_buf->len)) {
+			efx_free_rx_buffers(rx_queue, rx_buf,
+					    channel->rx_pkt_n_frags);
+			channel->n_rx_frm_trunc++;
+			goto out;
+		}
+	}
 
 	/* If we're in loopback test, then pass the packet directly to the
 	 * loopback layer, and free the rx_buf here
 	 */
 	if (unlikely(efx->loopback_selftest)) {
-		struct efx_rx_queue *rx_queue;
-
 		efx_loopback_rx_packet(efx, eh, rx_buf->len);
-		rx_queue = efx_channel_get_rx_queue(channel);
 		efx_free_rx_buffers(rx_queue, rx_buf,
 				    channel->rx_pkt_n_frags);
 		goto out;