xref: /freebsd/sys/dev/mlx5/mlx5_en/mlx5_en_ethtool.c (revision 59c8e88e72633afbc47a4ace0d2170d00d51f7dc)
1 /*-
2  * Copyright (c) 2015-2021 Mellanox Technologies. All rights reserved.
3  *
4  * Redistribution and use in source and binary forms, with or without
5  * modification, are permitted provided that the following conditions
6  * are met:
7  * 1. Redistributions of source code must retain the above copyright
8  *    notice, this list of conditions and the following disclaimer.
9  * 2. Redistributions in binary form must reproduce the above copyright
10  *    notice, this list of conditions and the following disclaimer in the
11  *    documentation and/or other materials provided with the distribution.
12  *
13  * THIS SOFTWARE IS PROVIDED BY AUTHOR AND CONTRIBUTORS `AS IS' AND
14  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
15  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
16  * ARE DISCLAIMED.  IN NO EVENT SHALL AUTHOR OR CONTRIBUTORS BE LIABLE
17  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
18  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
19  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
20  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
21  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
22  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
23  * SUCH DAMAGE.
24  */
25 
26 #include "opt_rss.h"
27 #include "opt_ratelimit.h"
28 
29 #include <dev/mlx5/mlx5_en/en.h>
30 #include <dev/mlx5/mlx5_en/port_buffer.h>
31 
32 void
33 mlx5e_create_stats(struct sysctl_ctx_list *ctx,
34     struct sysctl_oid_list *parent, const char *buffer,
35     const char **desc, unsigned num, u64 * arg)
36 {
37 	struct sysctl_oid *node;
38 	unsigned x;
39 
40 	sysctl_ctx_init(ctx);
41 
42 	node = SYSCTL_ADD_NODE(ctx, parent, OID_AUTO,
43 	    buffer, CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Statistics");
44 	if (node == NULL)
45 		return;
46 	for (x = 0; x != num; x++) {
47 		SYSCTL_ADD_UQUAD(ctx, SYSCTL_CHILDREN(node), OID_AUTO,
48 		    desc[2 * x], CTLFLAG_RD, arg + x, desc[2 * x + 1]);
49 	}
50 }
51 
52 void
53 mlx5e_create_counter_stats(struct sysctl_ctx_list *ctx,
54     struct sysctl_oid_list *parent, const char *buffer,
55     const char **desc, unsigned num, counter_u64_t *arg)
56 {
57 	struct sysctl_oid *node;
58 	unsigned x;
59 
60 	sysctl_ctx_init(ctx);
61 
62 	node = SYSCTL_ADD_NODE(ctx, parent, OID_AUTO,
63 	    buffer, CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Statistics");
64 	if (node == NULL)
65 		return;
66 	for (x = 0; x != num; x++) {
67 		SYSCTL_ADD_COUNTER_U64(ctx, SYSCTL_CHILDREN(node), OID_AUTO,
68 		    desc[2 * x], CTLFLAG_RD, arg + x, desc[2 * x + 1]);
69 	}
70 }
71 
72 static void
73 mlx5e_ethtool_sync_tx_completion_fact(struct mlx5e_priv *priv)
74 {
75 	/*
76 	 * Limit the maximum distance between completion events to
77 	 * half of the currently set TX queue size.
78 	 *
79 	 * The maximum number of queue entries a single IP packet can
80 	 * consume is given by MLX5_SEND_WQE_MAX_WQEBBS.
81 	 *
82 	 * The worst case max value is then given as below:
83 	 */
84 	uint64_t max = priv->params_ethtool.tx_queue_size /
85 	    (2 * MLX5_SEND_WQE_MAX_WQEBBS);
86 
87 	/*
88 	 * Update the maximum completion factor value in case the
89 	 * tx_queue_size field changed. Ensure we don't overflow
90 	 * 16-bits.
91 	 */
92 	if (max < 1)
93 		max = 1;
94 	else if (max > 65535)
95 		max = 65535;
96 	priv->params_ethtool.tx_completion_fact_max = max;
97 
98 	/*
99 	 * Verify that the current TX completion factor is within the
100 	 * given limits:
101 	 */
102 	if (priv->params_ethtool.tx_completion_fact < 1)
103 		priv->params_ethtool.tx_completion_fact = 1;
104 	else if (priv->params_ethtool.tx_completion_fact > max)
105 		priv->params_ethtool.tx_completion_fact = max;
106 }
107 
108 static int
109 mlx5e_getmaxrate(struct mlx5e_priv *priv)
110 {
111 	struct mlx5_core_dev *mdev = priv->mdev;
112 	u8 max_bw_unit[IEEE_8021QAZ_MAX_TCS];
113 	u8 max_bw_value[IEEE_8021QAZ_MAX_TCS];
114 	int err;
115 	int i;
116 
117 	PRIV_LOCK(priv);
118 	err = -mlx5_query_port_tc_rate_limit(mdev, max_bw_value, max_bw_unit);
119 	if (err)
120 		goto done;
121 
122 	for (i = 0; i <= mlx5_max_tc(mdev); i++) {
123 		switch (max_bw_unit[i]) {
124 		case MLX5_100_MBPS_UNIT:
125 			priv->params_ethtool.max_bw_value[i] = max_bw_value[i] * MLX5E_100MB;
126 			break;
127 		case MLX5_GBPS_UNIT:
128 			priv->params_ethtool.max_bw_value[i] = max_bw_value[i] * MLX5E_1GB;
129 			break;
130 		case MLX5_BW_NO_LIMIT:
131 			priv->params_ethtool.max_bw_value[i] = 0;
132 			break;
133 		default:
134 			priv->params_ethtool.max_bw_value[i] = -1;
135 			WARN_ONCE(true, "non-supported BW unit");
136 			break;
137 		}
138 	}
139 done:
140 	PRIV_UNLOCK(priv);
141 	return (err);
142 }
143 
144 static int
145 mlx5e_get_max_alloc(struct mlx5e_priv *priv)
146 {
147 	struct mlx5_core_dev *mdev = priv->mdev;
148 	int err;
149 	int x;
150 
151 	PRIV_LOCK(priv);
152 	err = -mlx5_query_port_tc_bw_alloc(mdev, priv->params_ethtool.max_bw_share);
153 	if (err == 0) {
154 		/* set default value */
155 		for (x = 0; x != IEEE_8021QAZ_MAX_TCS; x++) {
156 			priv->params_ethtool.max_bw_share[x] =
157 			    100 / IEEE_8021QAZ_MAX_TCS;
158 		}
159 		err = -mlx5_set_port_tc_bw_alloc(mdev,
160 		    priv->params_ethtool.max_bw_share);
161 	}
162 	PRIV_UNLOCK(priv);
163 
164 	return (err);
165 }
166 
167 static int
168 mlx5e_get_dscp(struct mlx5e_priv *priv)
169 {
170 	struct mlx5_core_dev *mdev = priv->mdev;
171 	int err;
172 
173 	if (MLX5_CAP_GEN(mdev, qcam_reg) == 0 ||
174 	    MLX5_CAP_QCAM_REG(mdev, qpts) == 0 ||
175 	    MLX5_CAP_QCAM_REG(mdev, qpdpm) == 0)
176 		return (EOPNOTSUPP);
177 
178 	PRIV_LOCK(priv);
179 	err = -mlx5_query_dscp2prio(mdev, priv->params_ethtool.dscp2prio);
180 	if (err)
181 		goto done;
182 
183 	err = -mlx5_query_trust_state(mdev, &priv->params_ethtool.trust_state);
184 	if (err)
185 		goto done;
186 done:
187 	PRIV_UNLOCK(priv);
188 	return (err);
189 }
190 
191 static void
192 mlx5e_tc_get_parameters(struct mlx5e_priv *priv,
193     u64 *new_bw_value, u8 *max_bw_value, u8 *max_bw_unit)
194 {
195 	const u64 upper_limit_mbps = 255 * MLX5E_100MB;
196 	const u64 upper_limit_gbps = 255 * MLX5E_1GB;
197 	u64 temp;
198 	int i;
199 
200 	memset(max_bw_value, 0, IEEE_8021QAZ_MAX_TCS);
201 	memset(max_bw_unit, 0, IEEE_8021QAZ_MAX_TCS);
202 
203 	for (i = 0; i <= mlx5_max_tc(priv->mdev); i++) {
204 		temp = (new_bw_value != NULL) ?
205 		    new_bw_value[i] : priv->params_ethtool.max_bw_value[i];
206 
207 		if (!temp) {
208 			max_bw_unit[i] = MLX5_BW_NO_LIMIT;
209 		} else if (temp > upper_limit_gbps) {
210 			max_bw_unit[i] = MLX5_BW_NO_LIMIT;
211 		} else if (temp <= upper_limit_mbps) {
212 			max_bw_value[i] = howmany(temp, MLX5E_100MB);
213 			max_bw_unit[i]  = MLX5_100_MBPS_UNIT;
214 		} else {
215 			max_bw_value[i] = howmany(temp, MLX5E_1GB);
216 			max_bw_unit[i]  = MLX5_GBPS_UNIT;
217 		}
218 	}
219 }
220 
221 static int
222 mlx5e_tc_maxrate_handler(SYSCTL_HANDLER_ARGS)
223 {
224 	struct mlx5e_priv *priv = arg1;
225 	struct mlx5_core_dev *mdev = priv->mdev;
226 	u8 max_bw_unit[IEEE_8021QAZ_MAX_TCS];
227 	u8 max_bw_value[IEEE_8021QAZ_MAX_TCS];
228 	u64 new_bw_value[IEEE_8021QAZ_MAX_TCS];
229 	u8 max_rates = mlx5_max_tc(mdev) + 1;
230 	u8 x;
231 	int err;
232 
233 	PRIV_LOCK(priv);
234 	err = SYSCTL_OUT(req, priv->params_ethtool.max_bw_value,
235 	    sizeof(priv->params_ethtool.max_bw_value[0]) * max_rates);
236 	if (err || !req->newptr)
237 		goto done;
238 	err = SYSCTL_IN(req, new_bw_value,
239 	    sizeof(new_bw_value[0]) * max_rates);
240 	if (err)
241 		goto done;
242 
243 	/* range check input value */
244 	for (x = 0; x != max_rates; x++) {
245 		if (new_bw_value[x] % MLX5E_100MB) {
246 			err = ERANGE;
247 			goto done;
248 		}
249 	}
250 
251 	mlx5e_tc_get_parameters(priv, new_bw_value, max_bw_value, max_bw_unit);
252 
253 	err = -mlx5_modify_port_tc_rate_limit(mdev, max_bw_value, max_bw_unit);
254 	if (err)
255 		goto done;
256 
257 	memcpy(priv->params_ethtool.max_bw_value, new_bw_value,
258 	    sizeof(priv->params_ethtool.max_bw_value));
259 done:
260 	PRIV_UNLOCK(priv);
261 	return (err);
262 }
263 
264 static int
265 mlx5e_tc_rate_share_handler(SYSCTL_HANDLER_ARGS)
266 {
267 	struct mlx5e_priv *priv = arg1;
268 	struct mlx5_core_dev *mdev = priv->mdev;
269 	u8 max_bw_share[IEEE_8021QAZ_MAX_TCS];
270 	u8 max_rates = mlx5_max_tc(mdev) + 1;
271 	int i;
272 	int err;
273 	int sum;
274 
275 	PRIV_LOCK(priv);
276 	err = SYSCTL_OUT(req, priv->params_ethtool.max_bw_share, max_rates);
277 	if (err || !req->newptr)
278 		goto done;
279 	err = SYSCTL_IN(req, max_bw_share, max_rates);
280 	if (err)
281 		goto done;
282 
283 	/* range check input value */
284 	for (sum = i = 0; i != max_rates; i++) {
285 		if (max_bw_share[i] < 1 || max_bw_share[i] > 100) {
286 			err = ERANGE;
287 			goto done;
288 		}
289 		sum += max_bw_share[i];
290 	}
291 
292 	/* sum of values should be as close to 100 as possible */
293 	if (sum < (100 - max_rates + 1) || sum > 100) {
294 		err = ERANGE;
295 		goto done;
296 	}
297 
298 	err = -mlx5_set_port_tc_bw_alloc(mdev, max_bw_share);
299 	if (err)
300 		goto done;
301 
302 	memcpy(priv->params_ethtool.max_bw_share, max_bw_share,
303 	    sizeof(priv->params_ethtool.max_bw_share));
304 done:
305 	PRIV_UNLOCK(priv);
306 	return (err);
307 }
308 
309 static int
310 mlx5e_get_prio_tc(struct mlx5e_priv *priv)
311 {
312 	struct mlx5_core_dev *mdev = priv->mdev;
313 	int err = 0;
314 	int i;
315 
316 	PRIV_LOCK(priv);
317 	if (!MLX5_CAP_GEN(priv->mdev, ets)) {
318 		PRIV_UNLOCK(priv);
319 		return (EOPNOTSUPP);
320 	}
321 
322 	for (i = 0; i != MLX5E_MAX_PRIORITY; i++) {
323 		err = -mlx5_query_port_prio_tc(mdev, i, priv->params_ethtool.prio_tc + i);
324 		if (err)
325 			break;
326 	}
327 	PRIV_UNLOCK(priv);
328 	return (err);
329 }
330 
331 static int
332 mlx5e_prio_to_tc_handler(SYSCTL_HANDLER_ARGS)
333 {
334 	struct mlx5e_priv *priv = arg1;
335 	struct mlx5_core_dev *mdev = priv->mdev;
336 	uint8_t temp[MLX5E_MAX_PRIORITY];
337 	int err;
338 	int i;
339 
340 	PRIV_LOCK(priv);
341 	err = SYSCTL_OUT(req, priv->params_ethtool.prio_tc, MLX5E_MAX_PRIORITY);
342 	if (err || !req->newptr)
343 		goto done;
344 	err = SYSCTL_IN(req, temp, MLX5E_MAX_PRIORITY);
345 	if (err)
346 		goto done;
347 
348 	for (i = 0; i != MLX5E_MAX_PRIORITY; i++) {
349 		if (temp[i] > mlx5_max_tc(mdev)) {
350 			err = ERANGE;
351 			goto done;
352 		}
353 	}
354 
355 	for (i = 0; i != MLX5E_MAX_PRIORITY; i++) {
356 		if (temp[i] == priv->params_ethtool.prio_tc[i])
357 			continue;
358 		err = -mlx5_set_port_prio_tc(mdev, i, temp[i]);
359 		if (err)
360 			goto done;
361 		/* update cached value */
362 		priv->params_ethtool.prio_tc[i] = temp[i];
363 	}
364 done:
365 	PRIV_UNLOCK(priv);
366 	return (err);
367 }
368 
369 int
370 mlx5e_fec_update(struct mlx5e_priv *priv)
371 {
372 	struct mlx5_core_dev *mdev = priv->mdev;
373 	u32 in[MLX5_ST_SZ_DW(pplm_reg)] = {};
374 	const int sz = MLX5_ST_SZ_BYTES(pplm_reg);
375 	int err;
376 
377 	if (!MLX5_CAP_GEN(mdev, pcam_reg))
378 		return (EOPNOTSUPP);
379 
380 	if (!MLX5_CAP_PCAM_REG(mdev, pplm))
381 		return (EOPNOTSUPP);
382 
383 	MLX5_SET(pplm_reg, in, local_port, 1);
384 
385 	err = -mlx5_core_access_reg(mdev, in, sz, in, sz, MLX5_REG_PPLM, 0, 0);
386 	if (err)
387 		return (err);
388 
389 	/* get 10x..25x mask */
390 	priv->params_ethtool.fec_mask_10x_25x[0] =
391 	    MLX5_GET(pplm_reg, in, fec_override_admin_10g_40g);
392 	priv->params_ethtool.fec_mask_10x_25x[1] =
393 	    MLX5_GET(pplm_reg, in, fec_override_admin_25g) &
394 	    MLX5_GET(pplm_reg, in, fec_override_admin_50g);
395 	priv->params_ethtool.fec_mask_10x_25x[2] =
396 	    MLX5_GET(pplm_reg, in, fec_override_admin_56g);
397 	priv->params_ethtool.fec_mask_10x_25x[3] =
398 	    MLX5_GET(pplm_reg, in, fec_override_admin_100g);
399 
400 	/* get 10x..25x available bits */
401 	priv->params_ethtool.fec_avail_10x_25x[0] =
402 	    MLX5_GET(pplm_reg, in, fec_override_cap_10g_40g);
403 	priv->params_ethtool.fec_avail_10x_25x[1] =
404 	    MLX5_GET(pplm_reg, in, fec_override_cap_25g) &
405 	    MLX5_GET(pplm_reg, in, fec_override_cap_50g);
406 	priv->params_ethtool.fec_avail_10x_25x[2] =
407 	    MLX5_GET(pplm_reg, in, fec_override_cap_56g);
408 	priv->params_ethtool.fec_avail_10x_25x[3] =
409 	    MLX5_GET(pplm_reg, in, fec_override_cap_100g);
410 
411 	/* get 50x mask */
412 	priv->params_ethtool.fec_mask_50x[0] =
413 	    MLX5_GET(pplm_reg, in, fec_override_admin_50g_1x);
414 	priv->params_ethtool.fec_mask_50x[1] =
415 	    MLX5_GET(pplm_reg, in, fec_override_admin_100g_2x);
416 	priv->params_ethtool.fec_mask_50x[2] =
417 	    MLX5_GET(pplm_reg, in, fec_override_admin_200g_4x);
418 	priv->params_ethtool.fec_mask_50x[3] =
419 	    MLX5_GET(pplm_reg, in, fec_override_admin_400g_8x);
420 
421 	/* get 50x available bits */
422 	priv->params_ethtool.fec_avail_50x[0] =
423 	    MLX5_GET(pplm_reg, in, fec_override_cap_50g_1x);
424 	priv->params_ethtool.fec_avail_50x[1] =
425 	    MLX5_GET(pplm_reg, in, fec_override_cap_100g_2x);
426 	priv->params_ethtool.fec_avail_50x[2] =
427 	    MLX5_GET(pplm_reg, in, fec_override_cap_200g_4x);
428 	priv->params_ethtool.fec_avail_50x[3] =
429 	    MLX5_GET(pplm_reg, in, fec_override_cap_400g_8x);
430 
431 	/* get current FEC mask */
432 	priv->params_ethtool.fec_mode_active =
433 	    MLX5_GET(pplm_reg, in, fec_mode_active);
434 
435 	return (0);
436 }
437 
438 static int
439 mlx5e_fec_mask_10x_25x_handler(SYSCTL_HANDLER_ARGS)
440 {
441 	struct mlx5e_priv *priv = arg1;
442 	struct mlx5_core_dev *mdev = priv->mdev;
443 	u32 out[MLX5_ST_SZ_DW(pplm_reg)] = {};
444 	u32 in[MLX5_ST_SZ_DW(pplm_reg)] = {};
445 	const int sz = MLX5_ST_SZ_BYTES(pplm_reg);
446 	u8 fec_mask_10x_25x[MLX5E_MAX_FEC_10X_25X];
447 	u8 fec_cap_changed = 0;
448 	u8 x;
449 	int err;
450 
451 	PRIV_LOCK(priv);
452 	err = SYSCTL_OUT(req, priv->params_ethtool.fec_mask_10x_25x,
453 	    sizeof(priv->params_ethtool.fec_mask_10x_25x));
454 	if (err || !req->newptr)
455 		goto done;
456 
457 	err = SYSCTL_IN(req, fec_mask_10x_25x,
458 	    sizeof(fec_mask_10x_25x));
459 	if (err)
460 		goto done;
461 
462 	if (!MLX5_CAP_GEN(mdev, pcam_reg)) {
463 		err = EOPNOTSUPP;
464 		goto done;
465 	}
466 
467 	if (!MLX5_CAP_PCAM_REG(mdev, pplm)) {
468 		err = EOPNOTSUPP;
469 		goto done;
470 	}
471 
472 	MLX5_SET(pplm_reg, in, local_port, 1);
473 
474 	err = -mlx5_core_access_reg(mdev, in, sz, in, sz, MLX5_REG_PPLM, 0, 0);
475 	if (err)
476 		goto done;
477 
478 	/* range check input value */
479 	for (x = 0; x != MLX5E_MAX_FEC_10X_25X; x++) {
480 		/* check only one bit is set, if any */
481 		if (fec_mask_10x_25x[x] & (fec_mask_10x_25x[x] - 1)) {
482 			err = ERANGE;
483 			goto done;
484 		}
485 		/* check a supported bit is set, if any */
486 		if (fec_mask_10x_25x[x] &
487 		    ~priv->params_ethtool.fec_avail_10x_25x[x]) {
488 			err = ERANGE;
489 			goto done;
490 		}
491 		fec_cap_changed |= (fec_mask_10x_25x[x] ^
492 		    priv->params_ethtool.fec_mask_10x_25x[x]);
493 	}
494 
495 	/* check for no changes */
496 	if (fec_cap_changed == 0)
497 		goto done;
498 
499 	memset(in, 0, sizeof(in));
500 
501 	MLX5_SET(pplm_reg, in, local_port, 1);
502 
503 	/* set new values */
504 	MLX5_SET(pplm_reg, in, fec_override_admin_10g_40g, fec_mask_10x_25x[0]);
505 	MLX5_SET(pplm_reg, in, fec_override_admin_25g, fec_mask_10x_25x[1]);
506 	MLX5_SET(pplm_reg, in, fec_override_admin_50g, fec_mask_10x_25x[1]);
507 	MLX5_SET(pplm_reg, in, fec_override_admin_56g, fec_mask_10x_25x[2]);
508 	MLX5_SET(pplm_reg, in, fec_override_admin_100g, fec_mask_10x_25x[3]);
509 
510 	/* preserve other values */
511 	MLX5_SET(pplm_reg, in, fec_override_admin_50g_1x, priv->params_ethtool.fec_mask_50x[0]);
512 	MLX5_SET(pplm_reg, in, fec_override_admin_100g_2x, priv->params_ethtool.fec_mask_50x[1]);
513 	MLX5_SET(pplm_reg, in, fec_override_admin_200g_4x, priv->params_ethtool.fec_mask_50x[2]);
514 	MLX5_SET(pplm_reg, in, fec_override_admin_400g_8x, priv->params_ethtool.fec_mask_50x[3]);
515 
516 	/* send new value to the firmware */
517 	err = -mlx5_core_access_reg(mdev, in, sz, out, sz, MLX5_REG_PPLM, 0, 1);
518 	if (err)
519 		goto done;
520 
521 	memcpy(priv->params_ethtool.fec_mask_10x_25x, fec_mask_10x_25x,
522 	    sizeof(priv->params_ethtool.fec_mask_10x_25x));
523 
524 	mlx5_toggle_port_link(priv->mdev);
525 done:
526 	PRIV_UNLOCK(priv);
527 	return (err);
528 }
529 
530 static int
531 mlx5e_fec_avail_10x_25x_handler(SYSCTL_HANDLER_ARGS)
532 {
533 	struct mlx5e_priv *priv = arg1;
534 	int err;
535 
536 	PRIV_LOCK(priv);
537 	err = SYSCTL_OUT(req, priv->params_ethtool.fec_avail_10x_25x,
538 	    sizeof(priv->params_ethtool.fec_avail_10x_25x));
539 	PRIV_UNLOCK(priv);
540 	return (err);
541 }
542 
543 static int
544 mlx5e_fec_mask_50x_handler(SYSCTL_HANDLER_ARGS)
545 {
546 	struct mlx5e_priv *priv = arg1;
547 	struct mlx5_core_dev *mdev = priv->mdev;
548 	u32 out[MLX5_ST_SZ_DW(pplm_reg)] = {};
549 	u32 in[MLX5_ST_SZ_DW(pplm_reg)] = {};
550 	const int sz = MLX5_ST_SZ_BYTES(pplm_reg);
551 	u16 fec_mask_50x[MLX5E_MAX_FEC_50X];
552 	u16 fec_cap_changed = 0;
553 	u8 x;
554 	int err;
555 
556 	PRIV_LOCK(priv);
557 	err = SYSCTL_OUT(req, priv->params_ethtool.fec_mask_50x,
558 	    sizeof(priv->params_ethtool.fec_mask_50x));
559 	if (err || !req->newptr)
560 		goto done;
561 
562 	err = SYSCTL_IN(req, fec_mask_50x,
563 	    sizeof(fec_mask_50x));
564 	if (err)
565 		goto done;
566 
567 	if (!MLX5_CAP_GEN(mdev, pcam_reg)) {
568 		err = EOPNOTSUPP;
569 		goto done;
570 	}
571 
572 	if (!MLX5_CAP_PCAM_REG(mdev, pplm)) {
573 		err = EOPNOTSUPP;
574 		goto done;
575 	}
576 
577 	MLX5_SET(pplm_reg, in, local_port, 1);
578 
579 	err = -mlx5_core_access_reg(mdev, in, sz, in, sz, MLX5_REG_PPLM, 0, 0);
580 	if (err)
581 		goto done;
582 
583 	/* range check input value */
584 	for (x = 0; x != MLX5E_MAX_FEC_50X; x++) {
585 		/* check only one bit is set, if any */
586 		if (fec_mask_50x[x] & (fec_mask_50x[x] - 1)) {
587 			err = ERANGE;
588 			goto done;
589 		}
590 		/* check a supported bit is set, if any */
591 		if (fec_mask_50x[x] &
592 		    ~priv->params_ethtool.fec_avail_50x[x]) {
593 			err = ERANGE;
594 			goto done;
595 		}
596 		fec_cap_changed |= (fec_mask_50x[x] ^
597 		    priv->params_ethtool.fec_mask_50x[x]);
598 	}
599 
600 	/* check for no changes */
601 	if (fec_cap_changed == 0)
602 		goto done;
603 
604 	memset(in, 0, sizeof(in));
605 
606 	MLX5_SET(pplm_reg, in, local_port, 1);
607 
608 	/* set new values */
609 	MLX5_SET(pplm_reg, in, fec_override_admin_50g_1x, fec_mask_50x[0]);
610 	MLX5_SET(pplm_reg, in, fec_override_admin_100g_2x, fec_mask_50x[1]);
611 	MLX5_SET(pplm_reg, in, fec_override_admin_200g_4x, fec_mask_50x[2]);
612 	MLX5_SET(pplm_reg, in, fec_override_admin_400g_8x, fec_mask_50x[3]);
613 
614 	/* preserve other values */
615 	MLX5_SET(pplm_reg, in, fec_override_admin_10g_40g, priv->params_ethtool.fec_mask_10x_25x[0]);
616 	MLX5_SET(pplm_reg, in, fec_override_admin_25g, priv->params_ethtool.fec_mask_10x_25x[1]);
617 	MLX5_SET(pplm_reg, in, fec_override_admin_50g, priv->params_ethtool.fec_mask_10x_25x[1]);
618 	MLX5_SET(pplm_reg, in, fec_override_admin_56g, priv->params_ethtool.fec_mask_10x_25x[2]);
619 	MLX5_SET(pplm_reg, in, fec_override_admin_100g, priv->params_ethtool.fec_mask_10x_25x[3]);
620 
621 	/* send new value to the firmware */
622 	err = -mlx5_core_access_reg(mdev, in, sz, out, sz, MLX5_REG_PPLM, 0, 1);
623 	if (err)
624 		goto done;
625 
626 	memcpy(priv->params_ethtool.fec_mask_50x, fec_mask_50x,
627 	    sizeof(priv->params_ethtool.fec_mask_50x));
628 
629 	mlx5_toggle_port_link(priv->mdev);
630 done:
631 	PRIV_UNLOCK(priv);
632 	return (err);
633 }
634 
635 static int
636 mlx5e_fec_avail_50x_handler(SYSCTL_HANDLER_ARGS)
637 {
638 	struct mlx5e_priv *priv = arg1;
639 	int err;
640 
641 	PRIV_LOCK(priv);
642 	err = SYSCTL_OUT(req, priv->params_ethtool.fec_avail_50x,
643 	    sizeof(priv->params_ethtool.fec_avail_50x));
644 	PRIV_UNLOCK(priv);
645 	return (err);
646 }
647 
648 static int
649 mlx5e_trust_state_handler(SYSCTL_HANDLER_ARGS)
650 {
651 	struct mlx5e_priv *priv = arg1;
652 	struct mlx5_core_dev *mdev = priv->mdev;
653 	int err;
654 	u8 result;
655 
656 	PRIV_LOCK(priv);
657 	result = priv->params_ethtool.trust_state;
658 	err = sysctl_handle_8(oidp, &result, 0, req);
659 	if (err || !req->newptr ||
660 	    result == priv->params_ethtool.trust_state)
661 		goto done;
662 
663 	switch (result) {
664 	case MLX5_QPTS_TRUST_PCP:
665 	case MLX5_QPTS_TRUST_DSCP:
666 		break;
667 	case MLX5_QPTS_TRUST_BOTH:
668 		if (!MLX5_CAP_QCAM_FEATURE(mdev, qpts_trust_both)) {
669 			err = EOPNOTSUPP;
670 			goto done;
671 		}
672 		break;
673 	default:
674 		err = ERANGE;
675 		goto done;
676 	}
677 
678 	err = -mlx5_set_trust_state(mdev, result);
679 	if (err)
680 		goto done;
681 
682 	priv->params_ethtool.trust_state = result;
683 
684 	/* update inline mode */
685 	mlx5e_refresh_sq_inline(priv);
686 #ifdef RATELIMIT
687 	mlx5e_rl_refresh_sq_inline(&priv->rl);
688 #endif
689 done:
690 	PRIV_UNLOCK(priv);
691 	return (err);
692 }
693 
694 static int
695 mlx5e_dscp_prio_handler(SYSCTL_HANDLER_ARGS)
696 {
697 	struct mlx5e_priv *priv = arg1;
698 	int prio_index = arg2;
699 	struct mlx5_core_dev *mdev = priv->mdev;
700 	uint8_t dscp2prio[MLX5_MAX_SUPPORTED_DSCP];
701 	uint8_t x;
702 	int err;
703 
704 	PRIV_LOCK(priv);
705 	err = SYSCTL_OUT(req, priv->params_ethtool.dscp2prio + prio_index,
706 	    sizeof(priv->params_ethtool.dscp2prio) / 8);
707 	if (err || !req->newptr)
708 		goto done;
709 
710 	memcpy(dscp2prio, priv->params_ethtool.dscp2prio, sizeof(dscp2prio));
711 	err = SYSCTL_IN(req, dscp2prio + prio_index, sizeof(dscp2prio) / 8);
712 	if (err)
713 		goto done;
714 	for (x = 0; x != MLX5_MAX_SUPPORTED_DSCP; x++) {
715 		if (dscp2prio[x] > 7) {
716 			err = ERANGE;
717 			goto done;
718 		}
719 	}
720 	err = -mlx5_set_dscp2prio(mdev, dscp2prio);
721 	if (err)
722 		goto done;
723 
724 	/* update local array */
725 	memcpy(priv->params_ethtool.dscp2prio, dscp2prio,
726 	    sizeof(priv->params_ethtool.dscp2prio));
727 done:
728 	PRIV_UNLOCK(priv);
729 	return (err);
730 }
731 
732 int
733 mlx5e_update_buf_lossy(struct mlx5e_priv *priv)
734 {
735 	struct ieee_pfc pfc;
736 
737 	PRIV_ASSERT_LOCKED(priv);
738 	bzero(&pfc, sizeof(pfc));
739 	pfc.pfc_en = priv->params.rx_priority_flow_control;
740 	return (-mlx5e_port_manual_buffer_config(priv, MLX5E_PORT_BUFFER_PFC,
741 	    priv->params_ethtool.hw_mtu, &pfc, NULL, NULL));
742 }
743 
744 static int
745 mlx5e_buf_size_handler(SYSCTL_HANDLER_ARGS)
746 {
747 	struct mlx5e_priv *priv;
748 	u32 buf_size[MLX5E_MAX_BUFFER];
749 	struct mlx5e_port_buffer port_buffer;
750 	int error, i;
751 
752 	priv = arg1;
753 	PRIV_LOCK(priv);
754 	error = -mlx5e_port_query_buffer(priv, &port_buffer);
755 	if (error != 0)
756 		goto done;
757 	for (i = 0; i < nitems(buf_size); i++)
758 		buf_size[i] = port_buffer.buffer[i].size;
759 	error = SYSCTL_OUT(req, buf_size, sizeof(buf_size));
760 	if (error != 0 || req->newptr == NULL)
761 		goto done;
762 	error = SYSCTL_IN(req, buf_size, sizeof(buf_size));
763 	if (error != 0)
764 		goto done;
765 	error = -mlx5e_port_manual_buffer_config(priv, MLX5E_PORT_BUFFER_SIZE,
766 	    priv->params_ethtool.hw_mtu, NULL, buf_size, NULL);
767 done:
768 	PRIV_UNLOCK(priv);
769 	return (error);
770 }
771 
772 static int
773 mlx5e_buf_prio_handler(SYSCTL_HANDLER_ARGS)
774 {
775 	struct mlx5e_priv *priv;
776 	struct mlx5_core_dev *mdev;
777 	u8 buffer[MLX5E_MAX_BUFFER];
778 	int error;
779 
780 	priv = arg1;
781 	mdev = priv->mdev;
782 	PRIV_LOCK(priv);
783 	error = -mlx5e_port_query_priority2buffer(mdev, buffer);
784 	if (error != 0)
785 		goto done;
786 	error = SYSCTL_OUT(req, buffer, MLX5E_MAX_BUFFER);
787 	if (error != 0 || req->newptr == NULL)
788 		goto done;
789 	error = SYSCTL_IN(req, buffer, MLX5E_MAX_BUFFER);
790 	if (error != 0)
791 		goto done;
792 	error = -mlx5e_port_manual_buffer_config(priv,
793 	    MLX5E_PORT_BUFFER_PRIO2BUFFER,
794 	    priv->params_ethtool.hw_mtu, NULL, NULL, buffer);
795 	if (error == 0)
796 		error = mlx5e_update_buf_lossy(priv);
797 done:
798 	PRIV_UNLOCK(priv);
799 	return (error);
800 }
801 
802 static int
803 mlx5e_cable_length_handler(SYSCTL_HANDLER_ARGS)
804 {
805 	struct mlx5e_priv *priv;
806 	u_int cable_len;
807 	int error;
808 
809 	priv = arg1;
810 	PRIV_LOCK(priv);
811 	cable_len = priv->dcbx.cable_len;
812 	error = sysctl_handle_int(oidp, &cable_len, 0, req);
813 	if (error == 0 && req->newptr != NULL &&
814 	    cable_len != priv->dcbx.cable_len) {
815 		error = -mlx5e_port_manual_buffer_config(priv,
816 		    MLX5E_PORT_BUFFER_CABLE_LEN, priv->params_ethtool.hw_mtu,
817 		    NULL, NULL, NULL);
818 		if (error == 0)
819 			priv->dcbx.cable_len = cable_len;
820 	}
821 	PRIV_UNLOCK(priv);
822 	return (error);
823 }
824 
825 static int
826 mlx5e_hw_temperature_handler(SYSCTL_HANDLER_ARGS)
827 {
828 	struct mlx5e_priv *priv = arg1;
829 	int err;
830 
831 	PRIV_LOCK(priv);
832 	err = SYSCTL_OUT(req, priv->params_ethtool.hw_val_temp,
833 	    sizeof(priv->params_ethtool.hw_val_temp[0]) *
834 	    priv->params_ethtool.hw_num_temp);
835 	if (err == 0 && req->newptr != NULL)
836 		err = EOPNOTSUPP;
837 	PRIV_UNLOCK(priv);
838 	return (err);
839 }
840 
841 int
842 mlx5e_hw_temperature_update(struct mlx5e_priv *priv)
843 {
844 	int err;
845 	u32 x;
846 
847 	if (priv->params_ethtool.hw_num_temp == 0) {
848 		u32 out_cap[MLX5_ST_SZ_DW(mtcap)] = {};
849 		const int sz_cap = MLX5_ST_SZ_BYTES(mtcap);
850 		u32 value;
851 
852 		err = -mlx5_core_access_reg(priv->mdev, NULL, 0, out_cap, sz_cap,
853 		    MLX5_ACCESS_REG_SUMMARY_CTRL_ID_MTCAP, 0, 0);
854 		if (err)
855 			goto done;
856 		value = MLX5_GET(mtcap, out_cap, sensor_count);
857 		if (value == 0)
858 			return (0);
859 		if (value > MLX5_MAX_TEMPERATURE)
860 			value = MLX5_MAX_TEMPERATURE;
861 		/* update number of temperature sensors */
862 		priv->params_ethtool.hw_num_temp = value;
863 	}
864 
865 	for (x = 0; x != priv->params_ethtool.hw_num_temp; x++) {
866 		u32 out_sensor[MLX5_ST_SZ_DW(mtmp_reg)] = {};
867 		const int sz_sensor = MLX5_ST_SZ_BYTES(mtmp_reg);
868 
869 		MLX5_SET(mtmp_reg, out_sensor, sensor_index, x);
870 
871 		err = -mlx5_core_access_reg(priv->mdev, out_sensor, sz_sensor,
872 		    out_sensor, sz_sensor,
873 		    MLX5_ACCESS_REG_SUMMARY_CTRL_ID_MTMP, 0, 0);
874 		if (err)
875 			goto done;
876 		/* convert from 0.125 celsius to millicelsius */
877 		priv->params_ethtool.hw_val_temp[x] =
878 		    (s16)MLX5_GET(mtmp_reg, out_sensor, temperature) * 125;
879 	}
880 done:
881 	return (err);
882 }
883 
884 #define	MLX5_PARAM_OFFSET(n)				\
885     __offsetof(struct mlx5e_priv, params_ethtool.n)
886 
887 static int
888 mlx5e_ethtool_handler(SYSCTL_HANDLER_ARGS)
889 {
890 	struct mlx5e_priv *priv = arg1;
891 	uint64_t value;
892 	int mode_modify;
893 	int was_opened;
894 	int error;
895 
896 	PRIV_LOCK(priv);
897 	value = priv->params_ethtool.arg[arg2];
898 	if (req != NULL) {
899 		error = sysctl_handle_64(oidp, &value, 0, req);
900 		if (error || req->newptr == NULL ||
901 		    value == priv->params_ethtool.arg[arg2])
902 			goto done;
903 
904 		/* assign new value */
905 		priv->params_ethtool.arg[arg2] = value;
906 	} else {
907 		error = 0;
908 	}
909 	/* check if device is gone */
910 	if (priv->gone) {
911 		error = ENXIO;
912 		goto done;
913 	}
914 	was_opened = test_bit(MLX5E_STATE_OPENED, &priv->state);
915 	mode_modify = MLX5_CAP_GEN(priv->mdev, cq_period_mode_modify);
916 
917 	switch (MLX5_PARAM_OFFSET(arg[arg2])) {
918 	case MLX5_PARAM_OFFSET(rx_coalesce_usecs):
919 		/* import RX coal time */
920 		if (priv->params_ethtool.rx_coalesce_usecs < 1)
921 			priv->params_ethtool.rx_coalesce_usecs = 0;
922 		else if (priv->params_ethtool.rx_coalesce_usecs >
923 		    MLX5E_FLD_MAX(cqc, cq_period)) {
924 			priv->params_ethtool.rx_coalesce_usecs =
925 			    MLX5E_FLD_MAX(cqc, cq_period);
926 		}
927 		priv->params.rx_cq_moderation_usec =
928 		    priv->params_ethtool.rx_coalesce_usecs;
929 
930 		/* check to avoid down and up the network interface */
931 		if (was_opened)
932 			error = mlx5e_refresh_channel_params(priv);
933 		break;
934 
935 	case MLX5_PARAM_OFFSET(rx_coalesce_pkts):
936 		/* import RX coal pkts */
937 		if (priv->params_ethtool.rx_coalesce_pkts < 1)
938 			priv->params_ethtool.rx_coalesce_pkts = 0;
939 		else if (priv->params_ethtool.rx_coalesce_pkts >
940 		    MLX5E_FLD_MAX(cqc, cq_max_count)) {
941 			priv->params_ethtool.rx_coalesce_pkts =
942 			    MLX5E_FLD_MAX(cqc, cq_max_count);
943 		}
944 		priv->params.rx_cq_moderation_pkts =
945 		    priv->params_ethtool.rx_coalesce_pkts;
946 
947 		/* check to avoid down and up the network interface */
948 		if (was_opened)
949 			error = mlx5e_refresh_channel_params(priv);
950 		break;
951 
952 	case MLX5_PARAM_OFFSET(tx_coalesce_usecs):
953 		/* import TX coal time */
954 		if (priv->params_ethtool.tx_coalesce_usecs < 1)
955 			priv->params_ethtool.tx_coalesce_usecs = 0;
956 		else if (priv->params_ethtool.tx_coalesce_usecs >
957 		    MLX5E_FLD_MAX(cqc, cq_period)) {
958 			priv->params_ethtool.tx_coalesce_usecs =
959 			    MLX5E_FLD_MAX(cqc, cq_period);
960 		}
961 		priv->params.tx_cq_moderation_usec =
962 		    priv->params_ethtool.tx_coalesce_usecs;
963 
964 		/* check to avoid down and up the network interface */
965 		if (was_opened)
966 			error = mlx5e_refresh_channel_params(priv);
967 		break;
968 
969 	case MLX5_PARAM_OFFSET(tx_coalesce_pkts):
970 		/* import TX coal pkts */
971 		if (priv->params_ethtool.tx_coalesce_pkts < 1)
972 			priv->params_ethtool.tx_coalesce_pkts = 0;
973 		else if (priv->params_ethtool.tx_coalesce_pkts >
974 		    MLX5E_FLD_MAX(cqc, cq_max_count)) {
975 			priv->params_ethtool.tx_coalesce_pkts =
976 			    MLX5E_FLD_MAX(cqc, cq_max_count);
977 		}
978 		priv->params.tx_cq_moderation_pkts =
979 		    priv->params_ethtool.tx_coalesce_pkts;
980 
981 		/* check to avoid down and up the network interface */
982 		if (was_opened)
983 			error = mlx5e_refresh_channel_params(priv);
984 		break;
985 
986 	case MLX5_PARAM_OFFSET(tx_queue_size):
987 		/* network interface must be down */
988 		if (was_opened)
989 			mlx5e_close_locked(priv->ifp);
990 
991 		/* import TX queue size */
992 		if (priv->params_ethtool.tx_queue_size <
993 		    (1 << MLX5E_PARAMS_MINIMUM_LOG_SQ_SIZE)) {
994 			priv->params_ethtool.tx_queue_size =
995 			    (1 << MLX5E_PARAMS_MINIMUM_LOG_SQ_SIZE);
996 		} else if (priv->params_ethtool.tx_queue_size >
997 		    priv->params_ethtool.tx_queue_size_max) {
998 			priv->params_ethtool.tx_queue_size =
999 			    priv->params_ethtool.tx_queue_size_max;
1000 		}
1001 		/* store actual TX queue size */
1002 		priv->params.log_sq_size =
1003 		    order_base_2(priv->params_ethtool.tx_queue_size);
1004 		priv->params_ethtool.tx_queue_size =
1005 		    1 << priv->params.log_sq_size;
1006 
1007 		/* verify TX completion factor */
1008 		mlx5e_ethtool_sync_tx_completion_fact(priv);
1009 
1010 		/* restart network interface, if any */
1011 		if (was_opened)
1012 			mlx5e_open_locked(priv->ifp);
1013 		break;
1014 
1015 	case MLX5_PARAM_OFFSET(rx_queue_size):
1016 		/* network interface must be down */
1017 		if (was_opened)
1018 			mlx5e_close_locked(priv->ifp);
1019 
1020 		/* import RX queue size */
1021 		if (priv->params_ethtool.rx_queue_size <
1022 		    (1 << MLX5E_PARAMS_MINIMUM_LOG_RQ_SIZE)) {
1023 			priv->params_ethtool.rx_queue_size =
1024 			    (1 << MLX5E_PARAMS_MINIMUM_LOG_RQ_SIZE);
1025 		} else if (priv->params_ethtool.rx_queue_size >
1026 		    priv->params_ethtool.rx_queue_size_max) {
1027 			priv->params_ethtool.rx_queue_size =
1028 			    priv->params_ethtool.rx_queue_size_max;
1029 		}
1030 		/* store actual RX queue size */
1031 		priv->params.log_rq_size =
1032 		    order_base_2(priv->params_ethtool.rx_queue_size);
1033 		priv->params_ethtool.rx_queue_size =
1034 		    1 << priv->params.log_rq_size;
1035 
1036 		/* restart network interface, if any */
1037 		if (was_opened)
1038 			mlx5e_open_locked(priv->ifp);
1039 		break;
1040 
1041 	case MLX5_PARAM_OFFSET(channels_rsss):
1042 		/* network interface must be down */
1043 		if (was_opened)
1044 			mlx5e_close_locked(priv->ifp);
1045 
1046 		/* import number of channels */
1047 		if (priv->params_ethtool.channels_rsss < 1)
1048 			priv->params_ethtool.channels_rsss = 1;
1049 		else if (priv->params_ethtool.channels_rsss > 128)
1050 			priv->params_ethtool.channels_rsss = 128;
1051 
1052 		priv->params.channels_rsss = priv->params_ethtool.channels_rsss;
1053 
1054 		/* restart network interface, if any */
1055 		if (was_opened)
1056 			mlx5e_open_locked(priv->ifp);
1057 		break;
1058 
1059 	case MLX5_PARAM_OFFSET(channels):
1060 		/* network interface must be down */
1061 		if (was_opened)
1062 			mlx5e_close_locked(priv->ifp);
1063 
1064 		/* import number of channels */
1065 		if (priv->params_ethtool.channels < 1)
1066 			priv->params_ethtool.channels = 1;
1067 		else if (priv->params_ethtool.channels >
1068 		    (u64) priv->mdev->priv.eq_table.num_comp_vectors) {
1069 			priv->params_ethtool.channels =
1070 			    (u64) priv->mdev->priv.eq_table.num_comp_vectors;
1071 		}
1072 		priv->params.num_channels = priv->params_ethtool.channels;
1073 
1074 		/* restart network interface, if any */
1075 		if (was_opened)
1076 			mlx5e_open_locked(priv->ifp);
1077 		break;
1078 
1079 	case MLX5_PARAM_OFFSET(rx_coalesce_mode):
1080 		/* network interface must be down */
1081 		if (was_opened != 0 && mode_modify == 0)
1082 			mlx5e_close_locked(priv->ifp);
1083 
1084 		/* import RX coalesce mode */
1085 		if (priv->params_ethtool.rx_coalesce_mode > 3)
1086 			priv->params_ethtool.rx_coalesce_mode = 3;
1087 		priv->params.rx_cq_moderation_mode =
1088 		    priv->params_ethtool.rx_coalesce_mode;
1089 
1090 		/* restart network interface, if any */
1091 		if (was_opened != 0) {
1092 			if (mode_modify == 0)
1093 				mlx5e_open_locked(priv->ifp);
1094 			else
1095 				error = mlx5e_refresh_channel_params(priv);
1096 		}
1097 		break;
1098 
1099 	case MLX5_PARAM_OFFSET(tx_coalesce_mode):
1100 		/* network interface must be down */
1101 		if (was_opened != 0 && mode_modify == 0)
1102 			mlx5e_close_locked(priv->ifp);
1103 
1104 		/* import TX coalesce mode */
1105 		if (priv->params_ethtool.tx_coalesce_mode != 0)
1106 			priv->params_ethtool.tx_coalesce_mode = 1;
1107 		priv->params.tx_cq_moderation_mode =
1108 		    priv->params_ethtool.tx_coalesce_mode;
1109 
1110 		/* restart network interface, if any */
1111 		if (was_opened != 0) {
1112 			if (mode_modify == 0)
1113 				mlx5e_open_locked(priv->ifp);
1114 			else
1115 				error = mlx5e_refresh_channel_params(priv);
1116 		}
1117 		break;
1118 
1119 	case MLX5_PARAM_OFFSET(hw_lro):
1120 		/* network interface must be down */
1121 		if (was_opened)
1122 			mlx5e_close_locked(priv->ifp);
1123 
1124 		/* import HW LRO mode */
1125 		if (priv->params_ethtool.hw_lro != 0 &&
1126 		    MLX5_CAP_ETH(priv->mdev, lro_cap)) {
1127 			priv->params_ethtool.hw_lro = 1;
1128 			/* check if feature should actually be enabled */
1129 			if (if_getcapenable(priv->ifp) & IFCAP_LRO) {
1130 				priv->params.hw_lro_en = true;
1131 			} else {
1132 				priv->params.hw_lro_en = false;
1133 
1134 				mlx5_en_warn(priv->ifp, "To enable HW LRO "
1135 				    "please also enable LRO via ifconfig(8).\n");
1136 			}
1137 		} else {
1138 			/* return an error if HW does not support this feature */
1139 			if (priv->params_ethtool.hw_lro != 0)
1140 				error = EINVAL;
1141 			priv->params.hw_lro_en = false;
1142 			priv->params_ethtool.hw_lro = 0;
1143 		}
1144 		/* restart network interface, if any */
1145 		if (was_opened)
1146 			mlx5e_open_locked(priv->ifp);
1147 		break;
1148 
1149 	case MLX5_PARAM_OFFSET(cqe_zipping):
1150 		/* network interface must be down */
1151 		if (was_opened)
1152 			mlx5e_close_locked(priv->ifp);
1153 
1154 		/* import CQE zipping mode */
1155 		if (priv->params_ethtool.cqe_zipping &&
1156 		    MLX5_CAP_GEN(priv->mdev, cqe_compression)) {
1157 			priv->params.cqe_zipping_en = true;
1158 			priv->params_ethtool.cqe_zipping = 1;
1159 		} else {
1160 			priv->params.cqe_zipping_en = false;
1161 			priv->params_ethtool.cqe_zipping = 0;
1162 		}
1163 		/* restart network interface, if any */
1164 		if (was_opened)
1165 			mlx5e_open_locked(priv->ifp);
1166 		break;
1167 
1168 	case MLX5_PARAM_OFFSET(tx_completion_fact):
1169 		/* network interface must be down */
1170 		if (was_opened)
1171 			mlx5e_close_locked(priv->ifp);
1172 
1173 		/* verify parameter */
1174 		mlx5e_ethtool_sync_tx_completion_fact(priv);
1175 
1176 		/* restart network interface, if any */
1177 		if (was_opened)
1178 			mlx5e_open_locked(priv->ifp);
1179 		break;
1180 
1181 	case MLX5_PARAM_OFFSET(modify_tx_dma):
1182 		/* check if network interface is opened */
1183 		if (was_opened) {
1184 			priv->params_ethtool.modify_tx_dma =
1185 			    priv->params_ethtool.modify_tx_dma ? 1 : 0;
1186 			/* modify tx according to value */
1187 			mlx5e_modify_tx_dma(priv, value != 0);
1188 		} else {
1189 			/* if closed force enable tx */
1190 			priv->params_ethtool.modify_tx_dma = 0;
1191 		}
1192 		break;
1193 
1194 	case MLX5_PARAM_OFFSET(modify_rx_dma):
1195 		/* check if network interface is opened */
1196 		if (was_opened) {
1197 			priv->params_ethtool.modify_rx_dma =
1198 			    priv->params_ethtool.modify_rx_dma ? 1 : 0;
1199 			/* modify rx according to value */
1200 			mlx5e_modify_rx_dma(priv, value != 0);
1201 		} else {
1202 			/* if closed force enable rx */
1203 			priv->params_ethtool.modify_rx_dma = 0;
1204 		}
1205 		break;
1206 
1207 	case MLX5_PARAM_OFFSET(diag_pci_enable):
1208 		priv->params_ethtool.diag_pci_enable =
1209 		    priv->params_ethtool.diag_pci_enable ? 1 : 0;
1210 
1211 		error = -mlx5_core_set_diagnostics_full(priv->mdev,
1212 		    priv->params_ethtool.diag_pci_enable,
1213 		    priv->params_ethtool.diag_general_enable);
1214 		break;
1215 
1216 	case MLX5_PARAM_OFFSET(diag_general_enable):
1217 		priv->params_ethtool.diag_general_enable =
1218 		    priv->params_ethtool.diag_general_enable ? 1 : 0;
1219 
1220 		error = -mlx5_core_set_diagnostics_full(priv->mdev,
1221 		    priv->params_ethtool.diag_pci_enable,
1222 		    priv->params_ethtool.diag_general_enable);
1223 		break;
1224 
1225 	case MLX5_PARAM_OFFSET(mc_local_lb):
1226 		/* check if mlx5ib is managing this feature */
1227 		if (MLX5_CAP_GEN(priv->mdev, port_type) != MLX5_CAP_PORT_TYPE_ETH) {
1228 			error = EOPNOTSUPP;
1229 			break;
1230 		}
1231 
1232 		priv->params_ethtool.mc_local_lb =
1233 		    priv->params_ethtool.mc_local_lb ? 1 : 0;
1234 
1235 		if (MLX5_CAP_GEN(priv->mdev, disable_local_lb_mc)) {
1236 			error = mlx5_nic_vport_modify_local_lb(priv->mdev,
1237 			    MLX5_LOCAL_MC_LB, priv->params_ethtool.mc_local_lb);
1238 		} else {
1239 			error = EOPNOTSUPP;
1240 		}
1241 		break;
1242 
1243 	case MLX5_PARAM_OFFSET(uc_local_lb):
1244 		/* check if mlx5ib is managing this feature */
1245 		if (MLX5_CAP_GEN(priv->mdev, port_type) != MLX5_CAP_PORT_TYPE_ETH) {
1246 			error = EOPNOTSUPP;
1247 			break;
1248 		}
1249 
1250 		priv->params_ethtool.uc_local_lb =
1251 		    priv->params_ethtool.uc_local_lb ? 1 : 0;
1252 
1253 		if (MLX5_CAP_GEN(priv->mdev, disable_local_lb_uc)) {
1254 			error = mlx5_nic_vport_modify_local_lb(priv->mdev,
1255 			    MLX5_LOCAL_UC_LB, priv->params_ethtool.uc_local_lb);
1256 		} else {
1257 			error = EOPNOTSUPP;
1258 		}
1259 		break;
1260 
1261 	case MLX5_PARAM_OFFSET(irq_cpu_base):
1262 	case MLX5_PARAM_OFFSET(irq_cpu_stride):
1263 		if (was_opened) {
1264 			/* network interface must toggled */
1265 			mlx5e_close_locked(priv->ifp);
1266 			mlx5e_open_locked(priv->ifp);
1267 		}
1268 		break;
1269 
1270 	default:
1271 		break;
1272 	}
1273 done:
1274 	PRIV_UNLOCK(priv);
1275 	return (error);
1276 }
1277 
1278 static const char *mlx5e_params_desc[] = {
1279 	MLX5E_PARAMS(MLX5E_STATS_DESC)
1280 };
1281 
1282 static const char *mlx5e_port_stats_debug_desc[] = {
1283 	MLX5E_PORT_STATS_DEBUG(MLX5E_STATS_DESC)
1284 };
1285 
1286 static int
1287 mlx5e_ethtool_debug_channel_info(SYSCTL_HANDLER_ARGS)
1288 {
1289 	struct mlx5e_priv *priv;
1290 	struct sbuf sb;
1291 	struct mlx5e_channel *c;
1292 	struct mlx5e_sq *sq;
1293 	struct mlx5e_rq *rq;
1294 	int error, i, tc;
1295 	bool opened;
1296 
1297 	priv = arg1;
1298 	error = sysctl_wire_old_buffer(req, 0);
1299 	if (error != 0)
1300 		return (error);
1301 	if (sbuf_new_for_sysctl(&sb, NULL, 1024, req) == NULL)
1302 		return (ENOMEM);
1303 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1304 
1305 	PRIV_LOCK(priv);
1306 	opened = test_bit(MLX5E_STATE_OPENED, &priv->state);
1307 
1308 	sbuf_printf(&sb, "pages irq %d\n",
1309 	    priv->mdev->priv.msix_arr[MLX5_EQ_VEC_PAGES].vector);
1310 	sbuf_printf(&sb, "command irq %d\n",
1311 	    priv->mdev->priv.msix_arr[MLX5_EQ_VEC_CMD].vector);
1312 	sbuf_printf(&sb, "async irq %d\n",
1313 	    priv->mdev->priv.msix_arr[MLX5_EQ_VEC_ASYNC].vector);
1314 
1315 	for (i = 0; i != priv->params.num_channels; i++) {
1316 		int eqn_not_used = -1;
1317 		int irqn = MLX5_EQ_VEC_COMP_BASE;
1318 
1319 		if (mlx5_vector2eqn(priv->mdev, i, &eqn_not_used, &irqn) != 0)
1320 			continue;
1321 
1322 		c = opened ? &priv->channel[i] : NULL;
1323 		rq = opened ? &c->rq : NULL;
1324 		sbuf_printf(&sb, "channel %d rq %d cq %d irq %d\n", i,
1325 		    opened ? rq->rqn : -1,
1326 		    opened ? rq->cq.mcq.cqn : -1,
1327 		    priv->mdev->priv.msix_arr[irqn].vector);
1328 
1329 		for (tc = 0; tc != priv->num_tc; tc++) {
1330 			sq = opened ? &c->sq[tc] : NULL;
1331 			sbuf_printf(&sb, "channel %d tc %d sq %d cq %d irq %d\n",
1332 			    i, tc,
1333 			    opened ? sq->sqn : -1,
1334 			    opened ? sq->cq.mcq.cqn : -1,
1335 			    priv->mdev->priv.msix_arr[irqn].vector);
1336 		}
1337 	}
1338 	PRIV_UNLOCK(priv);
1339 	error = sbuf_finish(&sb);
1340 	sbuf_delete(&sb);
1341 	return (error);
1342 }
1343 
1344 static int
1345 mlx5e_ethtool_debug_stats(SYSCTL_HANDLER_ARGS)
1346 {
1347 	struct mlx5e_priv *priv = arg1;
1348 	int sys_debug;
1349 	int error;
1350 
1351 	PRIV_LOCK(priv);
1352 	if (priv->gone != 0) {
1353 		error = ENODEV;
1354 		goto done;
1355 	}
1356 	sys_debug = priv->sysctl_debug;
1357 	error = sysctl_handle_int(oidp, &sys_debug, 0, req);
1358 	if (error != 0 || !req->newptr)
1359 		goto done;
1360 	sys_debug = sys_debug ? 1 : 0;
1361 	if (sys_debug == priv->sysctl_debug)
1362 		goto done;
1363 
1364 	if ((priv->sysctl_debug = sys_debug)) {
1365 		mlx5e_create_stats(&priv->stats.port_stats_debug.ctx,
1366 		    SYSCTL_CHILDREN(priv->sysctl_ifnet), "debug_stats",
1367 		    mlx5e_port_stats_debug_desc, MLX5E_PORT_STATS_DEBUG_NUM,
1368 		    priv->stats.port_stats_debug.arg);
1369 		SYSCTL_ADD_PROC(&priv->stats.port_stats_debug.ctx,
1370 		    SYSCTL_CHILDREN(priv->sysctl_ifnet), OID_AUTO,
1371 		    "hw_ctx_debug",
1372 		    CTLFLAG_RD | CTLFLAG_MPSAFE | CTLTYPE_STRING, priv, 0,
1373 		    mlx5e_ethtool_debug_channel_info, "S", "");
1374 	} else {
1375 		sysctl_ctx_free(&priv->stats.port_stats_debug.ctx);
1376 	}
1377 done:
1378 	PRIV_UNLOCK(priv);
1379 	return (error);
1380 }
1381 
1382 static void
1383 mlx5e_create_diagnostics(struct mlx5e_priv *priv)
1384 {
1385 	struct mlx5_core_diagnostics_entry entry;
1386 	struct sysctl_ctx_list *ctx;
1387 	struct sysctl_oid *node;
1388 	int x;
1389 
1390 	/* sysctl context we are using */
1391 	ctx = &priv->sysctl_ctx;
1392 
1393 	/* create root node */
1394 	node = SYSCTL_ADD_NODE(ctx,
1395 	    SYSCTL_CHILDREN(priv->sysctl_ifnet), OID_AUTO,
1396 	    "diagnostics", CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Diagnostics");
1397 	if (node == NULL)
1398 		return;
1399 
1400 	/* create PCI diagnostics */
1401 	for (x = 0; x != MLX5_CORE_PCI_DIAGNOSTICS_NUM; x++) {
1402 		entry = mlx5_core_pci_diagnostics_table[x];
1403 		if (mlx5_core_supports_diagnostics(priv->mdev, entry.counter_id) == 0)
1404 			continue;
1405 		SYSCTL_ADD_UQUAD(ctx, SYSCTL_CHILDREN(node), OID_AUTO,
1406 		    entry.desc, CTLFLAG_RD, priv->params_pci.array + x,
1407 		    "PCI diagnostics counter");
1408 	}
1409 
1410 	/* create general diagnostics */
1411 	for (x = 0; x != MLX5_CORE_GENERAL_DIAGNOSTICS_NUM; x++) {
1412 		entry = mlx5_core_general_diagnostics_table[x];
1413 		if (mlx5_core_supports_diagnostics(priv->mdev, entry.counter_id) == 0)
1414 			continue;
1415 		SYSCTL_ADD_UQUAD(ctx, SYSCTL_CHILDREN(node), OID_AUTO,
1416 		    entry.desc, CTLFLAG_RD, priv->params_general.array + x,
1417 		    "General diagnostics counter");
1418 	}
1419 }
1420 
1421 void
1422 mlx5e_create_ethtool(struct mlx5e_priv *priv)
1423 {
1424 	struct sysctl_oid *fec_node;
1425 	struct sysctl_oid *qos_node;
1426 	struct sysctl_oid *node;
1427 	const char *pnameunit;
1428 	struct mlx5e_port_buffer port_buffer;
1429 	unsigned x;
1430 	int i;
1431 
1432 	/* set some defaults */
1433 	priv->params_ethtool.irq_cpu_base = -1;	/* disabled */
1434 	priv->params_ethtool.irq_cpu_stride = 1;
1435 	priv->params_ethtool.tx_queue_size_max = 1 << MLX5E_PARAMS_MAXIMUM_LOG_SQ_SIZE;
1436 	priv->params_ethtool.rx_queue_size_max = 1 << MLX5E_PARAMS_MAXIMUM_LOG_RQ_SIZE;
1437 	priv->params_ethtool.tx_queue_size = 1 << priv->params.log_sq_size;
1438 	priv->params_ethtool.rx_queue_size = 1 << priv->params.log_rq_size;
1439 	priv->params_ethtool.channels = priv->params.num_channels;
1440 	priv->params_ethtool.channels_rsss = priv->params.channels_rsss;
1441 	priv->params_ethtool.coalesce_pkts_max = MLX5E_FLD_MAX(cqc, cq_max_count);
1442 	priv->params_ethtool.coalesce_usecs_max = MLX5E_FLD_MAX(cqc, cq_period);
1443 	priv->params_ethtool.rx_coalesce_mode = priv->params.rx_cq_moderation_mode;
1444 	priv->params_ethtool.rx_coalesce_usecs = priv->params.rx_cq_moderation_usec;
1445 	priv->params_ethtool.rx_coalesce_pkts = priv->params.rx_cq_moderation_pkts;
1446 	priv->params_ethtool.tx_coalesce_mode = priv->params.tx_cq_moderation_mode;
1447 	priv->params_ethtool.tx_coalesce_usecs = priv->params.tx_cq_moderation_usec;
1448 	priv->params_ethtool.tx_coalesce_pkts = priv->params.tx_cq_moderation_pkts;
1449 	priv->params_ethtool.hw_lro = priv->params.hw_lro_en;
1450 	priv->params_ethtool.cqe_zipping = priv->params.cqe_zipping_en;
1451 	mlx5e_ethtool_sync_tx_completion_fact(priv);
1452 
1453 	/* get default values for local loopback, if any */
1454 	if (MLX5_CAP_GEN(priv->mdev, disable_local_lb_mc) ||
1455 	    MLX5_CAP_GEN(priv->mdev, disable_local_lb_uc)) {
1456 		int err;
1457 		u8 val;
1458 
1459 		err = mlx5_nic_vport_query_local_lb(priv->mdev, MLX5_LOCAL_MC_LB, &val);
1460 		if (err == 0)
1461 			priv->params_ethtool.mc_local_lb = val;
1462 
1463 		err = mlx5_nic_vport_query_local_lb(priv->mdev, MLX5_LOCAL_UC_LB, &val);
1464 		if (err == 0)
1465 			priv->params_ethtool.uc_local_lb = val;
1466 	}
1467 
1468 	/* create root node */
1469 	node = SYSCTL_ADD_NODE(&priv->sysctl_ctx,
1470 	    SYSCTL_CHILDREN(priv->sysctl_ifnet), OID_AUTO,
1471 	    "conf", CTLFLAG_RW | CTLFLAG_MPSAFE, NULL, "Configuration");
1472 	if (node == NULL)
1473 		return;
1474 	for (x = 0; x != MLX5E_PARAMS_NUM; x++) {
1475 		/* check for read-only parameter */
1476 		if (strstr(mlx5e_params_desc[2 * x], "_max") != NULL ||
1477 		    strstr(mlx5e_params_desc[2 * x], "_mtu") != NULL) {
1478 			SYSCTL_ADD_PROC(&priv->sysctl_ctx, SYSCTL_CHILDREN(node), OID_AUTO,
1479 			    mlx5e_params_desc[2 * x], CTLTYPE_U64 | CTLFLAG_RD |
1480 			    CTLFLAG_MPSAFE, priv, x, &mlx5e_ethtool_handler, "QU",
1481 			    mlx5e_params_desc[2 * x + 1]);
1482 		} else if (strcmp(mlx5e_params_desc[2 * x], "hw_lro") == 0) {
1483 			/* read-only, but tunable parameters */
1484 			SYSCTL_ADD_PROC(&priv->sysctl_ctx, SYSCTL_CHILDREN(node), OID_AUTO,
1485 			    mlx5e_params_desc[2 * x], CTLTYPE_U64 | CTLFLAG_RDTUN |
1486 			    CTLFLAG_MPSAFE, priv, x, &mlx5e_ethtool_handler, "QU",
1487 			    mlx5e_params_desc[2 * x + 1]);
1488 		} else {
1489 			/*
1490 			 * NOTE: In FreeBSD-11 and newer the
1491 			 * CTLFLAG_RWTUN flag will take care of
1492 			 * loading default sysctl value from the
1493 			 * kernel environment, if any:
1494 			 */
1495 			SYSCTL_ADD_PROC(&priv->sysctl_ctx, SYSCTL_CHILDREN(node), OID_AUTO,
1496 			    mlx5e_params_desc[2 * x], CTLTYPE_U64 | CTLFLAG_RWTUN |
1497 			    CTLFLAG_MPSAFE, priv, x, &mlx5e_ethtool_handler, "QU",
1498 			    mlx5e_params_desc[2 * x + 1]);
1499 		}
1500 	}
1501 
1502 	/* create fec node */
1503 	fec_node = SYSCTL_ADD_NODE(&priv->sysctl_ctx,
1504 	    SYSCTL_CHILDREN(node), OID_AUTO,
1505 	    "fec", CTLFLAG_RW | CTLFLAG_MPSAFE, NULL,
1506 	    "Forward Error Correction");
1507 	if (fec_node == NULL)
1508 		return;
1509 
1510 	if (mlx5e_fec_update(priv) == 0) {
1511 		SYSCTL_ADD_U32(&priv->sysctl_ctx, SYSCTL_CHILDREN(fec_node), OID_AUTO,
1512 		    "mode_active", CTLFLAG_RD | CTLFLAG_MPSAFE,
1513 		    &priv->params_ethtool.fec_mode_active, 0,
1514 		    "Current FEC mode bit, if any.");
1515 
1516 		SYSCTL_ADD_PROC(&priv->sysctl_ctx, SYSCTL_CHILDREN(fec_node), OID_AUTO,
1517 		    "mask_10x_25x", CTLTYPE_U8 | CTLFLAG_RWTUN | CTLFLAG_MPSAFE,
1518 		    priv, 0, &mlx5e_fec_mask_10x_25x_handler, "CU",
1519 		    "Set FEC masks for 10G_40G, 25G_50G, 56G, 100G respectivly. "
1520 		    "0:Auto "
1521 		    "1:NOFEC "
1522 		    "2:FIRECODE "
1523 		    "4:RS");
1524 
1525 		SYSCTL_ADD_PROC(&priv->sysctl_ctx, SYSCTL_CHILDREN(fec_node), OID_AUTO,
1526 		    "avail_10x_25x", CTLTYPE_U8 | CTLFLAG_RD | CTLFLAG_MPSAFE,
1527 		    priv, 0, &mlx5e_fec_avail_10x_25x_handler, "CU",
1528 		    "Get available FEC bits for 10G_40G, 25G_50G, 56G, 100G respectivly. "
1529 		    "0:Auto "
1530 		    "1:NOFEC "
1531 		    "2:FIRECODE "
1532 		    "4:RS");
1533 
1534 		SYSCTL_ADD_PROC(&priv->sysctl_ctx, SYSCTL_CHILDREN(fec_node), OID_AUTO,
1535 		    "mask_50x", CTLTYPE_U16 | CTLFLAG_RWTUN | CTLFLAG_MPSAFE,
1536 		    priv, 0, &mlx5e_fec_mask_50x_handler, "SU",
1537 		    "Set FEC masks for 50G 1x, 100G 2x, 200G 4x, 400G 8x respectivly. "
1538 		    "0:Auto "
1539 		    "128:RS "
1540 		    "512:LL RS");
1541 
1542 		SYSCTL_ADD_PROC(&priv->sysctl_ctx, SYSCTL_CHILDREN(fec_node), OID_AUTO,
1543 		    "avail_50x", CTLTYPE_U16 | CTLFLAG_RD | CTLFLAG_MPSAFE,
1544 		    priv, 0, &mlx5e_fec_avail_50x_handler, "SU",
1545 		    "Get available FEC bits for 50G 1x, 100G 2x, 200G 4x, 400G 8x respectivly. "
1546 		    "0:Auto "
1547 		    "128:RS "
1548 		    "512:LL RS");
1549 	}
1550 
1551 	SYSCTL_ADD_PROC(&priv->sysctl_ctx, SYSCTL_CHILDREN(node), OID_AUTO,
1552 	    "debug_stats", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, priv,
1553 	    0, &mlx5e_ethtool_debug_stats, "I", "Extended debug statistics");
1554 
1555 	pnameunit = device_get_nameunit(priv->mdev->pdev->dev.bsddev);
1556 
1557 	SYSCTL_ADD_STRING(&priv->sysctl_ctx, SYSCTL_CHILDREN(node),
1558 	    OID_AUTO, "device_name", CTLFLAG_RD,
1559 	    __DECONST(void *, pnameunit), 0,
1560 	    "PCI device name");
1561 
1562 	/* Diagnostics support */
1563 	mlx5e_create_diagnostics(priv);
1564 
1565 	/* create qos node */
1566 	qos_node = SYSCTL_ADD_NODE(&priv->sysctl_ctx,
1567 	    SYSCTL_CHILDREN(node), OID_AUTO,
1568 	    "qos", CTLFLAG_RW | CTLFLAG_MPSAFE, NULL,
1569 	    "Quality Of Service configuration");
1570 	if (qos_node == NULL)
1571 		return;
1572 
1573 	/* Priority rate limit support */
1574 	if (mlx5e_getmaxrate(priv) == 0) {
1575 		SYSCTL_ADD_PROC(&priv->sysctl_ctx, SYSCTL_CHILDREN(qos_node),
1576 		    OID_AUTO, "tc_max_rate", CTLTYPE_U64 | CTLFLAG_RWTUN | CTLFLAG_MPSAFE,
1577 		    priv, 0, mlx5e_tc_maxrate_handler, "QU",
1578 		    "Max rate for priority, specified in kilobits, where kilo=1000, "
1579 		    "max_rate must be divisible by 100000");
1580 	}
1581 
1582 	/* Bandwidth limiting by ratio */
1583 	if (mlx5e_get_max_alloc(priv) == 0) {
1584 		SYSCTL_ADD_PROC(&priv->sysctl_ctx, SYSCTL_CHILDREN(qos_node),
1585 		    OID_AUTO, "tc_rate_share", CTLTYPE_U8 | CTLFLAG_RWTUN | CTLFLAG_MPSAFE,
1586 		    priv, 0, mlx5e_tc_rate_share_handler, "QU",
1587 		    "Specify bandwidth ratio from 1 to 100 "
1588 		    "for the available traffic classes");
1589 	}
1590 
1591 	/* Priority to traffic class mapping */
1592 	if (mlx5e_get_prio_tc(priv) == 0) {
1593 		SYSCTL_ADD_PROC(&priv->sysctl_ctx, SYSCTL_CHILDREN(qos_node),
1594 		    OID_AUTO, "prio_0_7_tc", CTLTYPE_U8 | CTLFLAG_RWTUN | CTLFLAG_MPSAFE,
1595 		    priv, 0, mlx5e_prio_to_tc_handler, "CU",
1596 		    "Set traffic class 0 to 7 for priority 0 to 7 inclusivly");
1597 	}
1598 
1599 	/* DSCP support */
1600 	if (mlx5e_get_dscp(priv) == 0) {
1601 		for (i = 0; i != MLX5_MAX_SUPPORTED_DSCP; i += 8) {
1602 			char name[32];
1603 			snprintf(name, sizeof(name), "dscp_%d_%d_prio", i, i + 7);
1604 			SYSCTL_ADD_PROC(&priv->sysctl_ctx, SYSCTL_CHILDREN(qos_node),
1605 				OID_AUTO, name, CTLTYPE_U8 | CTLFLAG_RWTUN | CTLFLAG_MPSAFE,
1606 				priv, i, mlx5e_dscp_prio_handler, "CU",
1607 				"Set DSCP to priority mapping, 0..7");
1608 		}
1609 #define	A	"Set trust state, 1:PCP 2:DSCP"
1610 #define	B	" 3:BOTH"
1611 		SYSCTL_ADD_PROC(&priv->sysctl_ctx, SYSCTL_CHILDREN(qos_node),
1612 		    OID_AUTO, "trust_state", CTLTYPE_U8 | CTLFLAG_RWTUN | CTLFLAG_MPSAFE,
1613 		    priv, 0, mlx5e_trust_state_handler, "CU",
1614 		    MLX5_CAP_QCAM_FEATURE(priv->mdev, qpts_trust_both) ?
1615 		    A B : A);
1616 #undef B
1617 #undef A
1618 	}
1619 
1620 	if (mlx5e_port_query_buffer(priv, &port_buffer) == 0) {
1621 		SYSCTL_ADD_PROC(&priv->sysctl_ctx, SYSCTL_CHILDREN(qos_node),
1622 		    OID_AUTO, "buffers_size",
1623 		    CTLTYPE_U32 | CTLFLAG_RWTUN | CTLFLAG_MPSAFE,
1624 		    priv, 0, mlx5e_buf_size_handler, "IU",
1625 		    "Set buffers sizes");
1626 		SYSCTL_ADD_PROC(&priv->sysctl_ctx, SYSCTL_CHILDREN(qos_node),
1627 		    OID_AUTO, "buffers_prio",
1628 		    CTLTYPE_U8 | CTLFLAG_RWTUN | CTLFLAG_MPSAFE,
1629 		    priv, 0, mlx5e_buf_prio_handler, "CU",
1630 		    "Set prio to buffers mapping");
1631 		SYSCTL_ADD_PROC(&priv->sysctl_ctx, SYSCTL_CHILDREN(qos_node),
1632 		    OID_AUTO, "cable_length",
1633 		    CTLTYPE_UINT | CTLFLAG_RWTUN | CTLFLAG_MPSAFE,
1634 		    priv, 0, mlx5e_cable_length_handler, "IU",
1635 		    "Set cable length in meters for xoff threshold calculation");
1636 	}
1637 
1638 	if (mlx5e_hw_temperature_update(priv) == 0) {
1639 		SYSCTL_ADD_PROC(&priv->sysctl_ctx, SYSCTL_CHILDREN(priv->sysctl_ifnet),
1640 		    OID_AUTO, "hw_temperature",
1641 		    CTLTYPE_S32 | CTLFLAG_RD | CTLFLAG_MPSAFE,
1642 		    priv, 0, mlx5e_hw_temperature_handler, "I",
1643 		    "HW temperature in millicelsius");
1644 	}
1645 }
1646