1/*
2 * Copyright (c) 2018, Mellanox Technologies. All rights reserved.
3 *
4 * This software is available to you under a choice of one of two
5 * licenses.  You may choose to be licensed under the terms of the GNU
6 * General Public License (GPL) Version 2, available from the file
7 * COPYING in the main directory of this source tree, or the
8 * OpenIB.org BSD license below:
9 *
10 *     Redistribution and use in source and binary forms, with or
11 *     without modification, are permitted provided that the following
12 *     conditions are met:
13 *
14 *      - Redistributions of source code must retain the above
15 *        copyright notice, this list of conditions and the following
16 *        disclaimer.
17 *
18 *      - Redistributions in binary form must reproduce the above
19 *        copyright notice, this list of conditions and the following
20 *        disclaimer in the documentation and/or other materials
21 *        provided with the distribution.
22 *
23 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
24 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
25 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
26 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
27 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
28 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
29 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
30 * SOFTWARE.
31 */
32#ifndef __MLX5_EN_XDP_H__
33#define __MLX5_EN_XDP_H__
34
35#include <linux/indirect_call_wrapper.h>
36#include <net/xdp_sock.h>
37
38#include "en.h"
39#include "en/txrx.h"
40
41#define MLX5E_XDP_MIN_INLINE (ETH_HLEN + VLAN_HLEN)
42
43#define MLX5E_XDP_INLINE_WQE_MAX_DS_CNT 16
44#define MLX5E_XDP_INLINE_WQE_SZ_THRSD \
45	(MLX5E_XDP_INLINE_WQE_MAX_DS_CNT * MLX5_SEND_WQE_DS - \
46	 sizeof(struct mlx5_wqe_inline_seg))
47
48struct mlx5e_xdp_buff {
49	struct xdp_buff xdp;
50	struct mlx5_cqe64 *cqe;
51	struct mlx5e_rq *rq;
52};
53
54/* XDP packets can be transmitted in different ways. On completion, we need to
55 * distinguish between them to clean up things in a proper way.
56 */
57enum mlx5e_xdp_xmit_mode {
58	/* An xdp_frame was transmitted due to either XDP_REDIRECT from another
59	 * device or XDP_TX from an XSK RQ. The frame has to be unmapped and
60	 * returned.
61	 */
62	MLX5E_XDP_XMIT_MODE_FRAME,
63
64	/* The xdp_frame was created in place as a result of XDP_TX from a
65	 * regular RQ. No DMA remapping happened, and the page belongs to us.
66	 */
67	MLX5E_XDP_XMIT_MODE_PAGE,
68
69	/* No xdp_frame was created at all, the transmit happened from a UMEM
70	 * page. The UMEM Completion Ring producer pointer has to be increased.
71	 */
72	MLX5E_XDP_XMIT_MODE_XSK,
73};
74
75/* xmit_mode entry is pushed to the fifo per packet, followed by multiple
76 * entries, as follows:
77 *
78 * MLX5E_XDP_XMIT_MODE_FRAME:
79 *    xdpf, dma_addr_1, dma_addr_2, ... , dma_addr_num.
80 *    'num' is derived from xdpf.
81 *
82 * MLX5E_XDP_XMIT_MODE_PAGE:
83 *    num, page_1, page_2, ... , page_num.
84 *
85 * MLX5E_XDP_XMIT_MODE_XSK:
86 *    frame.xsk_meta.
87 */
88#define MLX5E_XDP_FIFO_ENTRIES2DS_MAX_RATIO 4
89
90union mlx5e_xdp_info {
91	enum mlx5e_xdp_xmit_mode mode;
92	union {
93		struct xdp_frame *xdpf;
94		dma_addr_t dma_addr;
95	} frame;
96	union {
97		struct mlx5e_rq *rq;
98		u8 num;
99		struct page *page;
100	} page;
101	struct xsk_tx_metadata_compl xsk_meta;
102};
103
104struct mlx5e_xsk_param;
105int mlx5e_xdp_max_mtu(struct mlx5e_params *params, struct mlx5e_xsk_param *xsk);
106bool mlx5e_xdp_handle(struct mlx5e_rq *rq,
107		      struct bpf_prog *prog, struct mlx5e_xdp_buff *mlctx);
108void mlx5e_xdp_mpwqe_complete(struct mlx5e_xdpsq *sq);
109bool mlx5e_poll_xdpsq_cq(struct mlx5e_cq *cq);
110void mlx5e_free_xdpsq_descs(struct mlx5e_xdpsq *sq);
111void mlx5e_set_xmit_fp(struct mlx5e_xdpsq *sq, bool is_mpw);
112void mlx5e_xdp_rx_poll_complete(struct mlx5e_rq *rq);
113int mlx5e_xdp_xmit(struct net_device *dev, int n, struct xdp_frame **frames,
114		   u32 flags);
115
116extern const struct xdp_metadata_ops mlx5e_xdp_metadata_ops;
117extern const struct xsk_tx_metadata_ops mlx5e_xsk_tx_metadata_ops;
118
119INDIRECT_CALLABLE_DECLARE(bool mlx5e_xmit_xdp_frame_mpwqe(struct mlx5e_xdpsq *sq,
120							  struct mlx5e_xmit_data *xdptxd,
121							  int check_result,
122							  struct xsk_tx_metadata *meta));
123INDIRECT_CALLABLE_DECLARE(bool mlx5e_xmit_xdp_frame(struct mlx5e_xdpsq *sq,
124						    struct mlx5e_xmit_data *xdptxd,
125						    int check_result,
126						    struct xsk_tx_metadata *meta));
127INDIRECT_CALLABLE_DECLARE(int mlx5e_xmit_xdp_frame_check_mpwqe(struct mlx5e_xdpsq *sq));
128INDIRECT_CALLABLE_DECLARE(int mlx5e_xmit_xdp_frame_check(struct mlx5e_xdpsq *sq));
129
130static inline void mlx5e_xdp_tx_enable(struct mlx5e_priv *priv)
131{
132	set_bit(MLX5E_STATE_XDP_TX_ENABLED, &priv->state);
133
134	if (priv->channels.params.xdp_prog)
135		set_bit(MLX5E_STATE_XDP_ACTIVE, &priv->state);
136}
137
138static inline void mlx5e_xdp_tx_disable(struct mlx5e_priv *priv)
139{
140	if (priv->channels.params.xdp_prog)
141		clear_bit(MLX5E_STATE_XDP_ACTIVE, &priv->state);
142
143	clear_bit(MLX5E_STATE_XDP_TX_ENABLED, &priv->state);
144	/* Let other device's napi(s) and XSK wakeups see our new state. */
145	synchronize_net();
146}
147
148static inline bool mlx5e_xdp_tx_is_enabled(struct mlx5e_priv *priv)
149{
150	return test_bit(MLX5E_STATE_XDP_TX_ENABLED, &priv->state);
151}
152
153static inline bool mlx5e_xdp_is_active(struct mlx5e_priv *priv)
154{
155	return test_bit(MLX5E_STATE_XDP_ACTIVE, &priv->state);
156}
157
158static inline void mlx5e_xmit_xdp_doorbell(struct mlx5e_xdpsq *sq)
159{
160	if (sq->doorbell_cseg) {
161		mlx5e_notify_hw(&sq->wq, sq->pc, sq->uar_map, sq->doorbell_cseg);
162		sq->doorbell_cseg = NULL;
163	}
164}
165
166/* Enable inline WQEs to shift some load from a congested HCA (HW) to
167 * a less congested cpu (SW).
168 */
169static inline bool mlx5e_xdp_get_inline_state(struct mlx5e_xdpsq *sq, bool cur)
170{
171	u16 outstanding = sq->xdpi_fifo_pc - sq->xdpi_fifo_cc;
172
173#define MLX5E_XDP_INLINE_WATERMARK_LOW	10
174#define MLX5E_XDP_INLINE_WATERMARK_HIGH 128
175
176	if (cur && outstanding <= MLX5E_XDP_INLINE_WATERMARK_LOW)
177		return false;
178
179	if (!cur && outstanding >= MLX5E_XDP_INLINE_WATERMARK_HIGH)
180		return true;
181
182	return cur;
183}
184
185static inline bool mlx5e_xdp_mpwqe_is_full(struct mlx5e_tx_mpwqe *session, u8 max_sq_mpw_wqebbs)
186{
187	if (session->inline_on)
188		return session->ds_count + MLX5E_XDP_INLINE_WQE_MAX_DS_CNT >
189		       max_sq_mpw_wqebbs * MLX5_SEND_WQEBB_NUM_DS;
190
191	return mlx5e_tx_mpwqe_is_full(session, max_sq_mpw_wqebbs);
192}
193
194struct mlx5e_xdp_wqe_info {
195	u8 num_wqebbs;
196	u8 num_pkts;
197};
198
199static inline void
200mlx5e_xdp_mpwqe_add_dseg(struct mlx5e_xdpsq *sq,
201			 struct mlx5e_xmit_data *xdptxd,
202			 struct mlx5e_xdpsq_stats *stats)
203{
204	struct mlx5e_tx_mpwqe *session = &sq->mpwqe;
205	struct mlx5_wqe_data_seg *dseg =
206		(struct mlx5_wqe_data_seg *)session->wqe + session->ds_count;
207	u32 dma_len = xdptxd->len;
208
209	session->pkt_count++;
210	session->bytes_count += dma_len;
211
212	if (session->inline_on && dma_len <= MLX5E_XDP_INLINE_WQE_SZ_THRSD) {
213		struct mlx5_wqe_inline_seg *inline_dseg =
214			(struct mlx5_wqe_inline_seg *)dseg;
215		u16 ds_len = sizeof(*inline_dseg) + dma_len;
216		u16 ds_cnt = DIV_ROUND_UP(ds_len, MLX5_SEND_WQE_DS);
217
218		inline_dseg->byte_count = cpu_to_be32(dma_len | MLX5_INLINE_SEG);
219		memcpy(inline_dseg->data, xdptxd->data, dma_len);
220
221		session->ds_count += ds_cnt;
222		stats->inlnw++;
223		return;
224	}
225
226	dseg->addr       = cpu_to_be64(xdptxd->dma_addr);
227	dseg->byte_count = cpu_to_be32(dma_len);
228	dseg->lkey       = sq->mkey_be;
229	session->ds_count++;
230}
231
232static inline void
233mlx5e_xdpi_fifo_push(struct mlx5e_xdp_info_fifo *fifo,
234		     union mlx5e_xdp_info xi)
235{
236	u32 i = (*fifo->pc)++ & fifo->mask;
237
238	fifo->xi[i] = xi;
239}
240
241static inline union mlx5e_xdp_info
242mlx5e_xdpi_fifo_pop(struct mlx5e_xdp_info_fifo *fifo)
243{
244	return fifo->xi[(*fifo->cc)++ & fifo->mask];
245}
246#endif
247