xref: /linux/drivers/net/ethernet/intel/ice/ice_eswitch.c (revision 860a9bed265146b10311bcadbbcef59c3af4454d)
1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright (C) 2019-2021, Intel Corporation. */
3 
4 #include "ice.h"
5 #include "ice_lib.h"
6 #include "ice_eswitch.h"
7 #include "ice_eswitch_br.h"
8 #include "ice_fltr.h"
9 #include "ice_repr.h"
10 #include "devlink/devlink.h"
11 #include "ice_tc_lib.h"
12 
13 /**
14  * ice_eswitch_setup_env - configure eswitch HW filters
15  * @pf: pointer to PF struct
16  *
17  * This function adds HW filters configuration specific for switchdev
18  * mode.
19  */
20 static int ice_eswitch_setup_env(struct ice_pf *pf)
21 {
22 	struct ice_vsi *uplink_vsi = pf->eswitch.uplink_vsi;
23 	struct net_device *netdev = uplink_vsi->netdev;
24 	bool if_running = netif_running(netdev);
25 	struct ice_vsi_vlan_ops *vlan_ops;
26 
27 	if (if_running && !test_and_set_bit(ICE_VSI_DOWN, uplink_vsi->state))
28 		if (ice_down(uplink_vsi))
29 			return -ENODEV;
30 
31 	ice_remove_vsi_fltr(&pf->hw, uplink_vsi->idx);
32 
33 	netif_addr_lock_bh(netdev);
34 	__dev_uc_unsync(netdev, NULL);
35 	__dev_mc_unsync(netdev, NULL);
36 	netif_addr_unlock_bh(netdev);
37 
38 	if (ice_vsi_add_vlan_zero(uplink_vsi))
39 		goto err_vlan_zero;
40 
41 	if (ice_cfg_dflt_vsi(uplink_vsi->port_info, uplink_vsi->idx, true,
42 			     ICE_FLTR_RX))
43 		goto err_def_rx;
44 
45 	if (ice_cfg_dflt_vsi(uplink_vsi->port_info, uplink_vsi->idx, true,
46 			     ICE_FLTR_TX))
47 		goto err_def_tx;
48 
49 	vlan_ops = ice_get_compat_vsi_vlan_ops(uplink_vsi);
50 	if (vlan_ops->dis_rx_filtering(uplink_vsi))
51 		goto err_vlan_filtering;
52 
53 	if (ice_vsi_update_security(uplink_vsi, ice_vsi_ctx_set_allow_override))
54 		goto err_override_uplink;
55 
56 	if (ice_vsi_update_local_lb(uplink_vsi, true))
57 		goto err_override_local_lb;
58 
59 	if (if_running && ice_up(uplink_vsi))
60 		goto err_up;
61 
62 	return 0;
63 
64 err_up:
65 	ice_vsi_update_local_lb(uplink_vsi, false);
66 err_override_local_lb:
67 	ice_vsi_update_security(uplink_vsi, ice_vsi_ctx_clear_allow_override);
68 err_override_uplink:
69 	vlan_ops->ena_rx_filtering(uplink_vsi);
70 err_vlan_filtering:
71 	ice_cfg_dflt_vsi(uplink_vsi->port_info, uplink_vsi->idx, false,
72 			 ICE_FLTR_TX);
73 err_def_tx:
74 	ice_cfg_dflt_vsi(uplink_vsi->port_info, uplink_vsi->idx, false,
75 			 ICE_FLTR_RX);
76 err_def_rx:
77 	ice_vsi_del_vlan_zero(uplink_vsi);
78 err_vlan_zero:
79 	ice_fltr_add_mac_and_broadcast(uplink_vsi,
80 				       uplink_vsi->port_info->mac.perm_addr,
81 				       ICE_FWD_TO_VSI);
82 	if (if_running)
83 		ice_up(uplink_vsi);
84 
85 	return -ENODEV;
86 }
87 
88 /**
89  * ice_eswitch_release_repr - clear PR VSI configuration
90  * @pf: poiner to PF struct
91  * @repr: pointer to PR
92  */
93 static void
94 ice_eswitch_release_repr(struct ice_pf *pf, struct ice_repr *repr)
95 {
96 	struct ice_vsi *vsi = repr->src_vsi;
97 
98 	/* Skip representors that aren't configured */
99 	if (!repr->dst)
100 		return;
101 
102 	ice_vsi_update_security(vsi, ice_vsi_ctx_set_antispoof);
103 	metadata_dst_free(repr->dst);
104 	repr->dst = NULL;
105 	ice_fltr_add_mac_and_broadcast(vsi, repr->parent_mac,
106 				       ICE_FWD_TO_VSI);
107 }
108 
109 /**
110  * ice_eswitch_setup_repr - configure PR to run in switchdev mode
111  * @pf: pointer to PF struct
112  * @repr: pointer to PR struct
113  */
114 static int ice_eswitch_setup_repr(struct ice_pf *pf, struct ice_repr *repr)
115 {
116 	struct ice_vsi *uplink_vsi = pf->eswitch.uplink_vsi;
117 	struct ice_vsi *vsi = repr->src_vsi;
118 	struct metadata_dst *dst;
119 
120 	ice_remove_vsi_fltr(&pf->hw, vsi->idx);
121 	repr->dst = metadata_dst_alloc(0, METADATA_HW_PORT_MUX,
122 				       GFP_KERNEL);
123 	if (!repr->dst)
124 		goto err_add_mac_fltr;
125 
126 	if (ice_vsi_update_security(vsi, ice_vsi_ctx_clear_antispoof))
127 		goto err_dst_free;
128 
129 	if (ice_vsi_add_vlan_zero(vsi))
130 		goto err_update_security;
131 
132 	netif_keep_dst(uplink_vsi->netdev);
133 
134 	dst = repr->dst;
135 	dst->u.port_info.port_id = vsi->vsi_num;
136 	dst->u.port_info.lower_dev = uplink_vsi->netdev;
137 
138 	return 0;
139 
140 err_update_security:
141 	ice_vsi_update_security(vsi, ice_vsi_ctx_set_antispoof);
142 err_dst_free:
143 	metadata_dst_free(repr->dst);
144 	repr->dst = NULL;
145 err_add_mac_fltr:
146 	ice_fltr_add_mac_and_broadcast(vsi, repr->parent_mac, ICE_FWD_TO_VSI);
147 
148 	return -ENODEV;
149 }
150 
151 /**
152  * ice_eswitch_update_repr - reconfigure port representor
153  * @repr_id: representor ID
154  * @vsi: VSI for which port representor is configured
155  */
156 void ice_eswitch_update_repr(unsigned long repr_id, struct ice_vsi *vsi)
157 {
158 	struct ice_pf *pf = vsi->back;
159 	struct ice_repr *repr;
160 	int ret;
161 
162 	if (!ice_is_switchdev_running(pf))
163 		return;
164 
165 	repr = xa_load(&pf->eswitch.reprs, repr_id);
166 	if (!repr)
167 		return;
168 
169 	repr->src_vsi = vsi;
170 	repr->dst->u.port_info.port_id = vsi->vsi_num;
171 
172 	if (repr->br_port)
173 		repr->br_port->vsi = vsi;
174 
175 	ret = ice_vsi_update_security(vsi, ice_vsi_ctx_clear_antispoof);
176 	if (ret) {
177 		ice_fltr_add_mac_and_broadcast(vsi, repr->parent_mac,
178 					       ICE_FWD_TO_VSI);
179 		dev_err(ice_pf_to_dev(pf), "Failed to update VSI of port representor %d",
180 			repr->id);
181 	}
182 }
183 
184 /**
185  * ice_eswitch_port_start_xmit - callback for packets transmit
186  * @skb: send buffer
187  * @netdev: network interface device structure
188  *
189  * Returns NETDEV_TX_OK if sent, else an error code
190  */
191 netdev_tx_t
192 ice_eswitch_port_start_xmit(struct sk_buff *skb, struct net_device *netdev)
193 {
194 	struct ice_repr *repr = ice_netdev_to_repr(netdev);
195 	unsigned int len = skb->len;
196 	int ret;
197 
198 	skb_dst_drop(skb);
199 	dst_hold((struct dst_entry *)repr->dst);
200 	skb_dst_set(skb, (struct dst_entry *)repr->dst);
201 	skb->dev = repr->dst->u.port_info.lower_dev;
202 
203 	ret = dev_queue_xmit(skb);
204 	ice_repr_inc_tx_stats(repr, len, ret);
205 
206 	return ret;
207 }
208 
209 /**
210  * ice_eswitch_set_target_vsi - set eswitch context in Tx context descriptor
211  * @skb: pointer to send buffer
212  * @off: pointer to offload struct
213  */
214 void
215 ice_eswitch_set_target_vsi(struct sk_buff *skb,
216 			   struct ice_tx_offload_params *off)
217 {
218 	struct metadata_dst *dst = skb_metadata_dst(skb);
219 	u64 cd_cmd, dst_vsi;
220 
221 	if (!dst) {
222 		cd_cmd = ICE_TX_CTX_DESC_SWTCH_UPLINK << ICE_TXD_CTX_QW1_CMD_S;
223 		off->cd_qw1 |= (cd_cmd | ICE_TX_DESC_DTYPE_CTX);
224 	} else {
225 		cd_cmd = ICE_TX_CTX_DESC_SWTCH_VSI << ICE_TXD_CTX_QW1_CMD_S;
226 		dst_vsi = FIELD_PREP(ICE_TXD_CTX_QW1_VSI_M,
227 				     dst->u.port_info.port_id);
228 		off->cd_qw1 = cd_cmd | dst_vsi | ICE_TX_DESC_DTYPE_CTX;
229 	}
230 }
231 
232 /**
233  * ice_eswitch_release_env - clear eswitch HW filters
234  * @pf: pointer to PF struct
235  *
236  * This function removes HW filters configuration specific for switchdev
237  * mode and restores default legacy mode settings.
238  */
239 static void ice_eswitch_release_env(struct ice_pf *pf)
240 {
241 	struct ice_vsi *uplink_vsi = pf->eswitch.uplink_vsi;
242 	struct ice_vsi_vlan_ops *vlan_ops;
243 
244 	vlan_ops = ice_get_compat_vsi_vlan_ops(uplink_vsi);
245 
246 	ice_vsi_update_local_lb(uplink_vsi, false);
247 	ice_vsi_update_security(uplink_vsi, ice_vsi_ctx_clear_allow_override);
248 	vlan_ops->ena_rx_filtering(uplink_vsi);
249 	ice_clear_dflt_vsi(uplink_vsi);
250 	ice_fltr_add_mac_and_broadcast(uplink_vsi,
251 				       uplink_vsi->port_info->mac.perm_addr,
252 				       ICE_FWD_TO_VSI);
253 }
254 
255 /**
256  * ice_eswitch_enable_switchdev - configure eswitch in switchdev mode
257  * @pf: pointer to PF structure
258  */
259 static int ice_eswitch_enable_switchdev(struct ice_pf *pf)
260 {
261 	struct ice_vsi *uplink_vsi;
262 
263 	uplink_vsi = ice_get_main_vsi(pf);
264 	if (!uplink_vsi)
265 		return -ENODEV;
266 
267 	if (netif_is_any_bridge_port(uplink_vsi->netdev)) {
268 		dev_err(ice_pf_to_dev(pf),
269 			"Uplink port cannot be a bridge port\n");
270 		return -EINVAL;
271 	}
272 
273 	pf->eswitch.uplink_vsi = uplink_vsi;
274 
275 	if (ice_eswitch_setup_env(pf))
276 		return -ENODEV;
277 
278 	if (ice_eswitch_br_offloads_init(pf))
279 		goto err_br_offloads;
280 
281 	pf->eswitch.is_running = true;
282 
283 	return 0;
284 
285 err_br_offloads:
286 	ice_eswitch_release_env(pf);
287 	return -ENODEV;
288 }
289 
290 /**
291  * ice_eswitch_disable_switchdev - disable eswitch resources
292  * @pf: pointer to PF structure
293  */
294 static void ice_eswitch_disable_switchdev(struct ice_pf *pf)
295 {
296 	ice_eswitch_br_offloads_deinit(pf);
297 	ice_eswitch_release_env(pf);
298 
299 	pf->eswitch.is_running = false;
300 }
301 
302 /**
303  * ice_eswitch_mode_set - set new eswitch mode
304  * @devlink: pointer to devlink structure
305  * @mode: eswitch mode to switch to
306  * @extack: pointer to extack structure
307  */
308 int
309 ice_eswitch_mode_set(struct devlink *devlink, u16 mode,
310 		     struct netlink_ext_ack *extack)
311 {
312 	struct ice_pf *pf = devlink_priv(devlink);
313 
314 	if (pf->eswitch_mode == mode)
315 		return 0;
316 
317 	if (ice_has_vfs(pf)) {
318 		dev_info(ice_pf_to_dev(pf), "Changing eswitch mode is allowed only if there is no VFs created");
319 		NL_SET_ERR_MSG_MOD(extack, "Changing eswitch mode is allowed only if there is no VFs created");
320 		return -EOPNOTSUPP;
321 	}
322 
323 	switch (mode) {
324 	case DEVLINK_ESWITCH_MODE_LEGACY:
325 		dev_info(ice_pf_to_dev(pf), "PF %d changed eswitch mode to legacy",
326 			 pf->hw.pf_id);
327 		xa_destroy(&pf->eswitch.reprs);
328 		NL_SET_ERR_MSG_MOD(extack, "Changed eswitch mode to legacy");
329 		break;
330 	case DEVLINK_ESWITCH_MODE_SWITCHDEV:
331 	{
332 		if (ice_is_adq_active(pf)) {
333 			dev_err(ice_pf_to_dev(pf), "Couldn't change eswitch mode to switchdev - ADQ is active. Delete ADQ configs and try again, e.g. tc qdisc del dev $PF root");
334 			NL_SET_ERR_MSG_MOD(extack, "Couldn't change eswitch mode to switchdev - ADQ is active. Delete ADQ configs and try again, e.g. tc qdisc del dev $PF root");
335 			return -EOPNOTSUPP;
336 		}
337 
338 		dev_info(ice_pf_to_dev(pf), "PF %d changed eswitch mode to switchdev",
339 			 pf->hw.pf_id);
340 		xa_init(&pf->eswitch.reprs);
341 		NL_SET_ERR_MSG_MOD(extack, "Changed eswitch mode to switchdev");
342 		break;
343 	}
344 	default:
345 		NL_SET_ERR_MSG_MOD(extack, "Unknown eswitch mode");
346 		return -EINVAL;
347 	}
348 
349 	pf->eswitch_mode = mode;
350 	return 0;
351 }
352 
353 /**
354  * ice_eswitch_mode_get - get current eswitch mode
355  * @devlink: pointer to devlink structure
356  * @mode: output parameter for current eswitch mode
357  */
358 int ice_eswitch_mode_get(struct devlink *devlink, u16 *mode)
359 {
360 	struct ice_pf *pf = devlink_priv(devlink);
361 
362 	*mode = pf->eswitch_mode;
363 	return 0;
364 }
365 
366 /**
367  * ice_is_eswitch_mode_switchdev - check if eswitch mode is set to switchdev
368  * @pf: pointer to PF structure
369  *
370  * Returns true if eswitch mode is set to DEVLINK_ESWITCH_MODE_SWITCHDEV,
371  * false otherwise.
372  */
373 bool ice_is_eswitch_mode_switchdev(struct ice_pf *pf)
374 {
375 	return pf->eswitch_mode == DEVLINK_ESWITCH_MODE_SWITCHDEV;
376 }
377 
378 /**
379  * ice_eswitch_start_all_tx_queues - start Tx queues of all port representors
380  * @pf: pointer to PF structure
381  */
382 static void ice_eswitch_start_all_tx_queues(struct ice_pf *pf)
383 {
384 	struct ice_repr *repr;
385 	unsigned long id;
386 
387 	if (test_bit(ICE_DOWN, pf->state))
388 		return;
389 
390 	xa_for_each(&pf->eswitch.reprs, id, repr)
391 		ice_repr_start_tx_queues(repr);
392 }
393 
394 /**
395  * ice_eswitch_stop_all_tx_queues - stop Tx queues of all port representors
396  * @pf: pointer to PF structure
397  */
398 void ice_eswitch_stop_all_tx_queues(struct ice_pf *pf)
399 {
400 	struct ice_repr *repr;
401 	unsigned long id;
402 
403 	if (test_bit(ICE_DOWN, pf->state))
404 		return;
405 
406 	xa_for_each(&pf->eswitch.reprs, id, repr)
407 		ice_repr_stop_tx_queues(repr);
408 }
409 
410 static void ice_eswitch_stop_reprs(struct ice_pf *pf)
411 {
412 	ice_eswitch_stop_all_tx_queues(pf);
413 }
414 
415 static void ice_eswitch_start_reprs(struct ice_pf *pf)
416 {
417 	ice_eswitch_start_all_tx_queues(pf);
418 }
419 
420 int
421 ice_eswitch_attach(struct ice_pf *pf, struct ice_vf *vf)
422 {
423 	struct ice_repr *repr;
424 	int err;
425 
426 	if (pf->eswitch_mode == DEVLINK_ESWITCH_MODE_LEGACY)
427 		return 0;
428 
429 	if (xa_empty(&pf->eswitch.reprs)) {
430 		err = ice_eswitch_enable_switchdev(pf);
431 		if (err)
432 			return err;
433 	}
434 
435 	ice_eswitch_stop_reprs(pf);
436 
437 	repr = ice_repr_add_vf(vf);
438 	if (IS_ERR(repr)) {
439 		err = PTR_ERR(repr);
440 		goto err_create_repr;
441 	}
442 
443 	err = ice_eswitch_setup_repr(pf, repr);
444 	if (err)
445 		goto err_setup_repr;
446 
447 	err = xa_insert(&pf->eswitch.reprs, repr->id, repr, GFP_KERNEL);
448 	if (err)
449 		goto err_xa_alloc;
450 
451 	vf->repr_id = repr->id;
452 
453 	ice_eswitch_start_reprs(pf);
454 
455 	return 0;
456 
457 err_xa_alloc:
458 	ice_eswitch_release_repr(pf, repr);
459 err_setup_repr:
460 	ice_repr_rem_vf(repr);
461 err_create_repr:
462 	if (xa_empty(&pf->eswitch.reprs))
463 		ice_eswitch_disable_switchdev(pf);
464 	ice_eswitch_start_reprs(pf);
465 
466 	return err;
467 }
468 
469 void ice_eswitch_detach(struct ice_pf *pf, struct ice_vf *vf)
470 {
471 	struct ice_repr *repr = xa_load(&pf->eswitch.reprs, vf->repr_id);
472 	struct devlink *devlink = priv_to_devlink(pf);
473 
474 	if (!repr)
475 		return;
476 
477 	ice_eswitch_stop_reprs(pf);
478 	xa_erase(&pf->eswitch.reprs, repr->id);
479 
480 	if (xa_empty(&pf->eswitch.reprs))
481 		ice_eswitch_disable_switchdev(pf);
482 
483 	ice_eswitch_release_repr(pf, repr);
484 	ice_repr_rem_vf(repr);
485 
486 	if (xa_empty(&pf->eswitch.reprs)) {
487 		/* since all port representors are destroyed, there is
488 		 * no point in keeping the nodes
489 		 */
490 		ice_devlink_rate_clear_tx_topology(ice_get_main_vsi(pf));
491 		devl_lock(devlink);
492 		devl_rate_nodes_destroy(devlink);
493 		devl_unlock(devlink);
494 	} else {
495 		ice_eswitch_start_reprs(pf);
496 	}
497 }
498 
499 /**
500  * ice_eswitch_rebuild - rebuild eswitch
501  * @pf: pointer to PF structure
502  */
503 void ice_eswitch_rebuild(struct ice_pf *pf)
504 {
505 	struct ice_repr *repr;
506 	unsigned long id;
507 
508 	if (!ice_is_switchdev_running(pf))
509 		return;
510 
511 	xa_for_each(&pf->eswitch.reprs, id, repr)
512 		ice_eswitch_detach(pf, repr->vf);
513 }
514 
515 /**
516  * ice_eswitch_get_target - get netdev based on src_vsi from descriptor
517  * @rx_ring: ring used to receive the packet
518  * @rx_desc: descriptor used to get src_vsi value
519  *
520  * Get src_vsi value from descriptor and load correct representor. If it isn't
521  * found return rx_ring->netdev.
522  */
523 struct net_device *ice_eswitch_get_target(struct ice_rx_ring *rx_ring,
524 					  union ice_32b_rx_flex_desc *rx_desc)
525 {
526 	struct ice_eswitch *eswitch = &rx_ring->vsi->back->eswitch;
527 	struct ice_32b_rx_flex_desc_nic_2 *desc;
528 	struct ice_repr *repr;
529 
530 	desc = (struct ice_32b_rx_flex_desc_nic_2 *)rx_desc;
531 	repr = xa_load(&eswitch->reprs, le16_to_cpu(desc->src_vsi));
532 	if (!repr)
533 		return rx_ring->netdev;
534 
535 	return repr->netdev;
536 }
537