From: Maxim Mikityanskiy <maximmi@nvidia.com>
Date: Thu, 29 Sep 2022 00:21:43 -0700
Subject: net/mlx5e: xsk: Use XSK frame size as striding RQ page size
Patch-mainline: v6.1-rc1
Git-commit: fa5573359aae63d36518e2821f968ff2f5eae02e
References: jsc#PED-1549

XSK RQs support striding RQ linear mode, but the stride size is always
set to PAGE_SIZE. It may be larger than the XSK frame size,
unnecessarily reducing the useful space in a WQE, but more importantly
causing UMEM data corruption in certain cases.

Normally, stride size bigger than XSK frame size is not a problem if the
hardware enforces the MTU. However, traffic between vports skips the
hardware MTU check, and oversized packets may be received.

If an oversized packet is bigger than the XSK frame but not bigger than
the stride, it will cause overwriting of the adjacent UMEM region. If
the packet takes more than one stride, they can be recycled for reuse
so it's not a problem when the XSK frame size matches the stride size.

To reduce the impact of the above issue, attempt to use the MTT page
size for striding RQ that matches the XSK frame size, allowing to safely
use 2048-byte frames on an up-to-date firmware.

Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Acked-by: Thomas Bogendoerfer <tbogendoerfer@suse.de>
---
 drivers/net/ethernet/mellanox/mlx5/core/en/params.c |   37 ++++++++++++++++----
 1 file changed, 31 insertions(+), 6 deletions(-)

--- a/drivers/net/ethernet/mellanox/mlx5/core/en/params.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/params.c
@@ -6,6 +6,7 @@
 #include "en/port.h"
 #include "en_accel/en_accel.h"
 #include "en_accel/ipsec.h"
+#include <net/xdp_sock_drv.h>
 
 static u8 mlx5e_mpwrq_min_page_shift(struct mlx5_core_dev *mdev)
 {
@@ -16,8 +17,8 @@ static u8 mlx5e_mpwrq_min_page_shift(str
 
 u8 mlx5e_mpwrq_page_shift(struct mlx5_core_dev *mdev, struct mlx5e_xsk_param *xsk)
 {
+	u8 req_page_shift = xsk ? order_base_2(xsk->chunk_size) : PAGE_SHIFT;
 	u8 min_page_shift = mlx5e_mpwrq_min_page_shift(mdev);
-	u8 req_page_shift = PAGE_SHIFT;
 
 	/* Regular RQ uses order-0 pages, the NIC must be able to map them. */
 	if (WARN_ON_ONCE(!xsk && req_page_shift < min_page_shift))
@@ -933,12 +934,36 @@ static u8 mlx5e_build_icosq_log_wq_sz(st
 	/* If XDP program is attached, XSK may be turned on at any time without
 	 * restarting the channel. ICOSQ must be big enough to fit UMR WQEs of
 	 * both regular RQ and XSK RQ.
-	 * Although mlx5e_mpwqe_get_log_rq_size accepts mlx5e_xsk_param, it
-	 * doesn't affect its return value, as long as params->xdp_prog != NULL,
-	 * so we can just multiply by 2.
+	 *
+	 * XSK uses different values of page_shift, and the total number of UMR
+	 * WQEBBs depends on it. This dependency is complex and not monotonic,
+	 * especially taking into consideration that some of the parameters come
+	 * from capabilities. Hence, we have to try all valid values of XSK
+	 * frame size (and page_shift) to find the maximum.
 	 */
-	if (params->xdp_prog)
-		wqebbs *= 2;
+	if (params->xdp_prog) {
+		u32 max_xsk_wqebbs = 0;
+		u8 frame_shift;
+
+		for (frame_shift = XDP_UMEM_MIN_CHUNK_SHIFT;
+		     frame_shift <= PAGE_SHIFT; frame_shift++) {
+			/* The headroom doesn't affect the calculation. */
+			struct mlx5e_xsk_param xsk = {
+				.chunk_size = 1 << frame_shift,
+			};
+			u32 xsk_wqebbs;
+			u8 page_shift;
+
+			page_shift = mlx5e_mpwrq_page_shift(mdev, &xsk);
+			xsk_wqebbs = mlx5e_mpwrq_umr_wqebbs(mdev, page_shift) *
+				(1 << mlx5e_mpwqe_get_log_rq_size(mdev, params, &xsk));
+
+			if (xsk_wqebbs > max_xsk_wqebbs)
+				max_xsk_wqebbs = xsk_wqebbs;
+		}
+
+		wqebbs += max_xsk_wqebbs;
+	}
 
 	if (params->packet_merge.type == MLX5E_PACKET_MERGE_SHAMPO)
 		wqebbs += mlx5e_shampo_icosq_sz(mdev, params, rqp);