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 */
 20static 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_set_dflt_vsi(uplink_vsi))
 42		goto err_def_rx;
 43
 44	if (ice_cfg_dflt_vsi(uplink_vsi->port_info, uplink_vsi->idx, true,
 45			     ICE_FLTR_TX))
 46		goto err_def_tx;
 47
 48	vlan_ops = ice_get_compat_vsi_vlan_ops(uplink_vsi);
 49	if (vlan_ops->dis_rx_filtering(uplink_vsi))
 50		goto err_vlan_filtering;
 51
 52	if (ice_vsi_update_security(uplink_vsi, ice_vsi_ctx_set_allow_override))
 53		goto err_override_uplink;
 54
 55	if (ice_vsi_update_local_lb(uplink_vsi, true))
 56		goto err_override_local_lb;
 57
 58	if (if_running && ice_up(uplink_vsi))
 59		goto err_up;
 60
 61	return 0;
 62
 63err_up:
 64	ice_vsi_update_local_lb(uplink_vsi, false);
 65err_override_local_lb:
 66	ice_vsi_update_security(uplink_vsi, ice_vsi_ctx_clear_allow_override);
 67err_override_uplink:
 68	vlan_ops->ena_rx_filtering(uplink_vsi);
 69err_vlan_filtering:
 70	ice_cfg_dflt_vsi(uplink_vsi->port_info, uplink_vsi->idx, false,
 71			 ICE_FLTR_TX);
 72err_def_tx:
 73	ice_cfg_dflt_vsi(uplink_vsi->port_info, uplink_vsi->idx, false,
 74			 ICE_FLTR_RX);
 75err_def_rx:
 76	ice_vsi_del_vlan_zero(uplink_vsi);
 77err_vlan_zero:
 78	ice_fltr_add_mac_and_broadcast(uplink_vsi,
 79				       uplink_vsi->port_info->mac.perm_addr,
 80				       ICE_FWD_TO_VSI);
 81	if (if_running)
 82		ice_up(uplink_vsi);
 83
 84	return -ENODEV;
 85}
 86
 87/**
 88 * ice_eswitch_release_repr - clear PR VSI configuration
 89 * @pf: poiner to PF struct
 90 * @repr: pointer to PR
 91 */
 92static void
 93ice_eswitch_release_repr(struct ice_pf *pf, struct ice_repr *repr)
 94{
 95	struct ice_vsi *vsi = repr->src_vsi;
 96
 97	/* Skip representors that aren't configured */
 98	if (!repr->dst)
 99		return;
100
101	ice_vsi_update_security(vsi, ice_vsi_ctx_set_antispoof);
102	metadata_dst_free(repr->dst);
103	repr->dst = NULL;
104	ice_fltr_add_mac_and_broadcast(vsi, repr->parent_mac,
105				       ICE_FWD_TO_VSI);
106}
107
108/**
109 * ice_eswitch_setup_repr - configure PR to run in switchdev mode
110 * @pf: pointer to PF struct
111 * @repr: pointer to PR struct
112 */
113static int ice_eswitch_setup_repr(struct ice_pf *pf, struct ice_repr *repr)
114{
115	struct ice_vsi *uplink_vsi = pf->eswitch.uplink_vsi;
116	struct ice_vsi *vsi = repr->src_vsi;
117	struct metadata_dst *dst;
118
119	repr->dst = metadata_dst_alloc(0, METADATA_HW_PORT_MUX,
120				       GFP_KERNEL);
121	if (!repr->dst)
122		return -ENOMEM;
123
124	netif_keep_dst(uplink_vsi->netdev);
125
126	dst = repr->dst;
127	dst->u.port_info.port_id = vsi->vsi_num;
128	dst->u.port_info.lower_dev = uplink_vsi->netdev;
129
130	return 0;
131}
132
133/**
134 * ice_eswitch_cfg_vsi - configure VSI to work in slow-path
135 * @vsi: VSI structure of representee
136 * @mac: representee MAC
137 *
138 * Return: 0 on success, non-zero on error.
139 */
140int ice_eswitch_cfg_vsi(struct ice_vsi *vsi, const u8 *mac)
141{
142	int err;
143
144	ice_remove_vsi_fltr(&vsi->back->hw, vsi->idx);
145
146	err = ice_vsi_update_security(vsi, ice_vsi_ctx_clear_antispoof);
147	if (err)
148		goto err_update_security;
149
150	err = ice_vsi_add_vlan_zero(vsi);
151	if (err)
152		goto err_vlan_zero;
153
154	return 0;
155
156err_vlan_zero:
157	ice_vsi_update_security(vsi, ice_vsi_ctx_set_antispoof);
158err_update_security:
159	ice_fltr_add_mac_and_broadcast(vsi, mac, ICE_FWD_TO_VSI);
160
161	return err;
162}
163
164/**
165 * ice_eswitch_decfg_vsi - unroll changes done to VSI for switchdev
166 * @vsi: VSI structure of representee
167 * @mac: representee MAC
168 */
169void ice_eswitch_decfg_vsi(struct ice_vsi *vsi, const u8 *mac)
170{
171	ice_vsi_update_security(vsi, ice_vsi_ctx_set_antispoof);
172	ice_fltr_add_mac_and_broadcast(vsi, mac, ICE_FWD_TO_VSI);
173}
174
175/**
176 * ice_eswitch_update_repr - reconfigure port representor
177 * @repr_id: representor ID
178 * @vsi: VSI for which port representor is configured
179 */
180void ice_eswitch_update_repr(unsigned long *repr_id, struct ice_vsi *vsi)
181{
182	struct ice_pf *pf = vsi->back;
183	struct ice_repr *repr;
184	int err;
185
186	if (!ice_is_switchdev_running(pf))
187		return;
188
189	repr = xa_load(&pf->eswitch.reprs, *repr_id);
190	if (!repr)
191		return;
192
193	repr->src_vsi = vsi;
194	repr->dst->u.port_info.port_id = vsi->vsi_num;
195
196	if (repr->br_port)
197		repr->br_port->vsi = vsi;
198
199	err = ice_eswitch_cfg_vsi(vsi, repr->parent_mac);
200	if (err)
201		dev_err(ice_pf_to_dev(pf), "Failed to update VSI of port representor %d",
202			repr->id);
203
204	/* The VSI number is different, reload the PR with new id */
205	if (repr->id != vsi->vsi_num) {
206		xa_erase(&pf->eswitch.reprs, repr->id);
207		repr->id = vsi->vsi_num;
208		if (xa_insert(&pf->eswitch.reprs, repr->id, repr, GFP_KERNEL))
209			dev_err(ice_pf_to_dev(pf), "Failed to reload port representor %d",
210				repr->id);
211		*repr_id = repr->id;
212	}
213}
214
215/**
216 * ice_eswitch_port_start_xmit - callback for packets transmit
217 * @skb: send buffer
218 * @netdev: network interface device structure
219 *
220 * Returns NETDEV_TX_OK if sent, else an error code
221 */
222netdev_tx_t
223ice_eswitch_port_start_xmit(struct sk_buff *skb, struct net_device *netdev)
224{
225	struct ice_repr *repr = ice_netdev_to_repr(netdev);
226	unsigned int len = skb->len;
227	int ret;
228
229	skb_dst_drop(skb);
230	dst_hold((struct dst_entry *)repr->dst);
231	skb_dst_set(skb, (struct dst_entry *)repr->dst);
232	skb->dev = repr->dst->u.port_info.lower_dev;
233
234	ret = dev_queue_xmit(skb);
235	ice_repr_inc_tx_stats(repr, len, ret);
236
237	return ret;
238}
239
240/**
241 * ice_eswitch_set_target_vsi - set eswitch context in Tx context descriptor
242 * @skb: pointer to send buffer
243 * @off: pointer to offload struct
244 */
245void
246ice_eswitch_set_target_vsi(struct sk_buff *skb,
247			   struct ice_tx_offload_params *off)
248{
249	struct metadata_dst *dst = skb_metadata_dst(skb);
250	u64 cd_cmd, dst_vsi;
251
252	if (!dst) {
253		cd_cmd = ICE_TX_CTX_DESC_SWTCH_UPLINK << ICE_TXD_CTX_QW1_CMD_S;
254		off->cd_qw1 |= (cd_cmd | ICE_TX_DESC_DTYPE_CTX);
255	} else {
256		cd_cmd = ICE_TX_CTX_DESC_SWTCH_VSI << ICE_TXD_CTX_QW1_CMD_S;
257		dst_vsi = FIELD_PREP(ICE_TXD_CTX_QW1_VSI_M,
258				     dst->u.port_info.port_id);
259		off->cd_qw1 = cd_cmd | dst_vsi | ICE_TX_DESC_DTYPE_CTX;
260	}
261}
262
263/**
264 * ice_eswitch_release_env - clear eswitch HW filters
265 * @pf: pointer to PF struct
266 *
267 * This function removes HW filters configuration specific for switchdev
268 * mode and restores default legacy mode settings.
269 */
270static void ice_eswitch_release_env(struct ice_pf *pf)
271{
272	struct ice_vsi *uplink_vsi = pf->eswitch.uplink_vsi;
273	struct ice_vsi_vlan_ops *vlan_ops;
274
275	vlan_ops = ice_get_compat_vsi_vlan_ops(uplink_vsi);
276
277	ice_vsi_update_local_lb(uplink_vsi, false);
278	ice_vsi_update_security(uplink_vsi, ice_vsi_ctx_clear_allow_override);
279	vlan_ops->ena_rx_filtering(uplink_vsi);
280	ice_cfg_dflt_vsi(uplink_vsi->port_info, uplink_vsi->idx, false,
281			 ICE_FLTR_TX);
282	ice_cfg_dflt_vsi(uplink_vsi->port_info, uplink_vsi->idx, false,
283			 ICE_FLTR_RX);
284	ice_fltr_add_mac_and_broadcast(uplink_vsi,
285				       uplink_vsi->port_info->mac.perm_addr,
286				       ICE_FWD_TO_VSI);
287}
288
289/**
290 * ice_eswitch_enable_switchdev - configure eswitch in switchdev mode
291 * @pf: pointer to PF structure
292 */
293static int ice_eswitch_enable_switchdev(struct ice_pf *pf)
294{
295	struct ice_vsi *uplink_vsi;
296
297	uplink_vsi = ice_get_main_vsi(pf);
298	if (!uplink_vsi)
299		return -ENODEV;
300
301	if (netif_is_any_bridge_port(uplink_vsi->netdev)) {
302		dev_err(ice_pf_to_dev(pf),
303			"Uplink port cannot be a bridge port\n");
304		return -EINVAL;
305	}
306
307	pf->eswitch.uplink_vsi = uplink_vsi;
308
309	if (ice_eswitch_setup_env(pf))
310		return -ENODEV;
311
312	if (ice_eswitch_br_offloads_init(pf))
313		goto err_br_offloads;
314
315	pf->eswitch.is_running = true;
316
317	return 0;
318
319err_br_offloads:
320	ice_eswitch_release_env(pf);
321	return -ENODEV;
322}
323
324/**
325 * ice_eswitch_disable_switchdev - disable eswitch resources
326 * @pf: pointer to PF structure
327 */
328static void ice_eswitch_disable_switchdev(struct ice_pf *pf)
329{
330	ice_eswitch_br_offloads_deinit(pf);
331	ice_eswitch_release_env(pf);
332
333	pf->eswitch.is_running = false;
334}
335
336/**
337 * ice_eswitch_mode_set - set new eswitch mode
338 * @devlink: pointer to devlink structure
339 * @mode: eswitch mode to switch to
340 * @extack: pointer to extack structure
341 */
342int
343ice_eswitch_mode_set(struct devlink *devlink, u16 mode,
344		     struct netlink_ext_ack *extack)
345{
346	struct ice_pf *pf = devlink_priv(devlink);
347
348	if (pf->eswitch_mode == mode)
349		return 0;
350
351	if (ice_has_vfs(pf)) {
352		dev_info(ice_pf_to_dev(pf), "Changing eswitch mode is allowed only if there is no VFs created");
353		NL_SET_ERR_MSG_MOD(extack, "Changing eswitch mode is allowed only if there is no VFs created");
354		return -EOPNOTSUPP;
355	}
356
357	switch (mode) {
358	case DEVLINK_ESWITCH_MODE_LEGACY:
359		dev_info(ice_pf_to_dev(pf), "PF %d changed eswitch mode to legacy",
360			 pf->hw.pf_id);
361		xa_destroy(&pf->eswitch.reprs);
362		NL_SET_ERR_MSG_MOD(extack, "Changed eswitch mode to legacy");
363		break;
364	case DEVLINK_ESWITCH_MODE_SWITCHDEV:
365	{
366		if (ice_is_adq_active(pf)) {
367			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");
368			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");
369			return -EOPNOTSUPP;
370		}
371
372		dev_info(ice_pf_to_dev(pf), "PF %d changed eswitch mode to switchdev",
373			 pf->hw.pf_id);
374		xa_init(&pf->eswitch.reprs);
375		NL_SET_ERR_MSG_MOD(extack, "Changed eswitch mode to switchdev");
376		break;
377	}
378	default:
379		NL_SET_ERR_MSG_MOD(extack, "Unknown eswitch mode");
380		return -EINVAL;
381	}
382
383	pf->eswitch_mode = mode;
384	return 0;
385}
386
387/**
388 * ice_eswitch_mode_get - get current eswitch mode
389 * @devlink: pointer to devlink structure
390 * @mode: output parameter for current eswitch mode
391 */
392int ice_eswitch_mode_get(struct devlink *devlink, u16 *mode)
393{
394	struct ice_pf *pf = devlink_priv(devlink);
395
396	*mode = pf->eswitch_mode;
397	return 0;
398}
399
400/**
401 * ice_is_eswitch_mode_switchdev - check if eswitch mode is set to switchdev
402 * @pf: pointer to PF structure
403 *
404 * Returns true if eswitch mode is set to DEVLINK_ESWITCH_MODE_SWITCHDEV,
405 * false otherwise.
406 */
407bool ice_is_eswitch_mode_switchdev(struct ice_pf *pf)
408{
409	return pf->eswitch_mode == DEVLINK_ESWITCH_MODE_SWITCHDEV;
410}
411
412/**
413 * ice_eswitch_start_all_tx_queues - start Tx queues of all port representors
414 * @pf: pointer to PF structure
415 */
416static void ice_eswitch_start_all_tx_queues(struct ice_pf *pf)
417{
418	struct ice_repr *repr;
419	unsigned long id;
420
421	if (test_bit(ICE_DOWN, pf->state))
422		return;
423
424	xa_for_each(&pf->eswitch.reprs, id, repr)
425		ice_repr_start_tx_queues(repr);
426}
427
428/**
429 * ice_eswitch_stop_all_tx_queues - stop Tx queues of all port representors
430 * @pf: pointer to PF structure
431 */
432void ice_eswitch_stop_all_tx_queues(struct ice_pf *pf)
433{
434	struct ice_repr *repr;
435	unsigned long id;
436
437	if (test_bit(ICE_DOWN, pf->state))
438		return;
439
440	xa_for_each(&pf->eswitch.reprs, id, repr)
441		ice_repr_stop_tx_queues(repr);
442}
443
444static void ice_eswitch_stop_reprs(struct ice_pf *pf)
445{
446	ice_eswitch_stop_all_tx_queues(pf);
447}
448
449static void ice_eswitch_start_reprs(struct ice_pf *pf)
450{
451	ice_eswitch_start_all_tx_queues(pf);
452}
453
454static int
455ice_eswitch_attach(struct ice_pf *pf, struct ice_repr *repr, unsigned long *id)
456{
457	int err;
458
459	if (pf->eswitch_mode == DEVLINK_ESWITCH_MODE_LEGACY)
460		return 0;
461
462	if (xa_empty(&pf->eswitch.reprs)) {
463		err = ice_eswitch_enable_switchdev(pf);
464		if (err)
465			return err;
466	}
467
468	ice_eswitch_stop_reprs(pf);
469
470	err = repr->ops.add(repr);
471	if (err)
472		goto err_create_repr;
473
474	err = ice_eswitch_setup_repr(pf, repr);
475	if (err)
476		goto err_setup_repr;
477
478	err = xa_insert(&pf->eswitch.reprs, repr->id, repr, GFP_KERNEL);
479	if (err)
480		goto err_xa_alloc;
481
482	*id = repr->id;
483
484	ice_eswitch_start_reprs(pf);
485
486	return 0;
487
488err_xa_alloc:
489	ice_eswitch_release_repr(pf, repr);
490err_setup_repr:
491	repr->ops.rem(repr);
492err_create_repr:
493	if (xa_empty(&pf->eswitch.reprs))
494		ice_eswitch_disable_switchdev(pf);
495	ice_eswitch_start_reprs(pf);
496
497	return err;
498}
499
500/**
501 * ice_eswitch_attach_vf - attach VF to a eswitch
502 * @pf: pointer to PF structure
503 * @vf: pointer to VF structure to be attached
504 *
505 * During attaching port representor for VF is created.
506 *
507 * Return: zero on success or an error code on failure.
508 */
509int ice_eswitch_attach_vf(struct ice_pf *pf, struct ice_vf *vf)
510{
511	struct ice_repr *repr = ice_repr_create_vf(vf);
512	struct devlink *devlink = priv_to_devlink(pf);
513	int err;
514
515	if (IS_ERR(repr))
516		return PTR_ERR(repr);
517
518	devl_lock(devlink);
519	err = ice_eswitch_attach(pf, repr, &vf->repr_id);
520	if (err)
521		ice_repr_destroy(repr);
522	devl_unlock(devlink);
523
524	return err;
525}
526
527/**
528 * ice_eswitch_attach_sf - attach SF to a eswitch
529 * @pf: pointer to PF structure
530 * @sf: pointer to SF structure to be attached
531 *
532 * During attaching port representor for SF is created.
533 *
534 * Return: zero on success or an error code on failure.
535 */
536int ice_eswitch_attach_sf(struct ice_pf *pf, struct ice_dynamic_port *sf)
537{
538	struct ice_repr *repr = ice_repr_create_sf(sf);
539	int err;
540
541	if (IS_ERR(repr))
542		return PTR_ERR(repr);
543
544	err = ice_eswitch_attach(pf, repr, &sf->repr_id);
545	if (err)
546		ice_repr_destroy(repr);
547
548	return err;
549}
550
551static void ice_eswitch_detach(struct ice_pf *pf, struct ice_repr *repr)
552{
553	ice_eswitch_stop_reprs(pf);
554	repr->ops.rem(repr);
555
556	xa_erase(&pf->eswitch.reprs, repr->id);
557
558	if (xa_empty(&pf->eswitch.reprs))
559		ice_eswitch_disable_switchdev(pf);
560
561	ice_eswitch_release_repr(pf, repr);
562	ice_repr_destroy(repr);
563
564	if (xa_empty(&pf->eswitch.reprs)) {
565		struct devlink *devlink = priv_to_devlink(pf);
566
567		/* since all port representors are destroyed, there is
568		 * no point in keeping the nodes
569		 */
570		ice_devlink_rate_clear_tx_topology(ice_get_main_vsi(pf));
571		devl_rate_nodes_destroy(devlink);
572	} else {
573		ice_eswitch_start_reprs(pf);
574	}
575}
576
577/**
578 * ice_eswitch_detach_vf - detach VF from a eswitch
579 * @pf: pointer to PF structure
580 * @vf: pointer to VF structure to be detached
581 */
582void ice_eswitch_detach_vf(struct ice_pf *pf, struct ice_vf *vf)
583{
584	struct ice_repr *repr = xa_load(&pf->eswitch.reprs, vf->repr_id);
585	struct devlink *devlink = priv_to_devlink(pf);
586
587	if (!repr)
588		return;
589
590	devl_lock(devlink);
591	ice_eswitch_detach(pf, repr);
592	devl_unlock(devlink);
593}
594
595/**
596 * ice_eswitch_detach_sf - detach SF from a eswitch
597 * @pf: pointer to PF structure
598 * @sf: pointer to SF structure to be detached
599 */
600void ice_eswitch_detach_sf(struct ice_pf *pf, struct ice_dynamic_port *sf)
601{
602	struct ice_repr *repr = xa_load(&pf->eswitch.reprs, sf->repr_id);
603
604	if (!repr)
605		return;
606
607	ice_eswitch_detach(pf, repr);
608}
609
610/**
611 * ice_eswitch_get_target - get netdev based on src_vsi from descriptor
612 * @rx_ring: ring used to receive the packet
613 * @rx_desc: descriptor used to get src_vsi value
614 *
615 * Get src_vsi value from descriptor and load correct representor. If it isn't
616 * found return rx_ring->netdev.
617 */
618struct net_device *ice_eswitch_get_target(struct ice_rx_ring *rx_ring,
619					  union ice_32b_rx_flex_desc *rx_desc)
620{
621	struct ice_eswitch *eswitch = &rx_ring->vsi->back->eswitch;
622	struct ice_32b_rx_flex_desc_nic_2 *desc;
623	struct ice_repr *repr;
624
625	desc = (struct ice_32b_rx_flex_desc_nic_2 *)rx_desc;
626	repr = xa_load(&eswitch->reprs, le16_to_cpu(desc->src_vsi));
627	if (!repr)
628		return rx_ring->netdev;
629
630	return repr->netdev;
631}