xref: /linux/drivers/net/ethernet/intel/ice/ice_sched.c (revision 4d66c56f7efe122d09d06cd3ebfa52a43d51a9cb)
1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright (c) 2018, Intel Corporation. */
3 
4 #include "ice_sched.h"
5 
6 /**
7  * ice_sched_add_root_node - Insert the Tx scheduler root node in SW DB
8  * @pi: port information structure
9  * @info: Scheduler element information from firmware
10  *
11  * This function inserts the root node of the scheduling tree topology
12  * to the SW DB.
13  */
14 static enum ice_status
15 ice_sched_add_root_node(struct ice_port_info *pi,
16 			struct ice_aqc_txsched_elem_data *info)
17 {
18 	struct ice_sched_node *root;
19 	struct ice_hw *hw;
20 
21 	if (!pi)
22 		return ICE_ERR_PARAM;
23 
24 	hw = pi->hw;
25 
26 	root = devm_kzalloc(ice_hw_to_dev(hw), sizeof(*root), GFP_KERNEL);
27 	if (!root)
28 		return ICE_ERR_NO_MEMORY;
29 
30 	/* coverity[suspicious_sizeof] */
31 	root->children = devm_kcalloc(ice_hw_to_dev(hw), hw->max_children[0],
32 				      sizeof(*root), GFP_KERNEL);
33 	if (!root->children) {
34 		devm_kfree(ice_hw_to_dev(hw), root);
35 		return ICE_ERR_NO_MEMORY;
36 	}
37 
38 	memcpy(&root->info, info, sizeof(*info));
39 	pi->root = root;
40 	return 0;
41 }
42 
43 /**
44  * ice_sched_find_node_by_teid - Find the Tx scheduler node in SW DB
45  * @start_node: pointer to the starting ice_sched_node struct in a sub-tree
46  * @teid: node TEID to search
47  *
48  * This function searches for a node matching the TEID in the scheduling tree
49  * from the SW DB. The search is recursive and is restricted by the number of
50  * layers it has searched through; stopping at the max supported layer.
51  *
52  * This function needs to be called when holding the port_info->sched_lock
53  */
54 struct ice_sched_node *
55 ice_sched_find_node_by_teid(struct ice_sched_node *start_node, u32 teid)
56 {
57 	u16 i;
58 
59 	/* The TEID is same as that of the start_node */
60 	if (ICE_TXSCHED_GET_NODE_TEID(start_node) == teid)
61 		return start_node;
62 
63 	/* The node has no children or is at the max layer */
64 	if (!start_node->num_children ||
65 	    start_node->tx_sched_layer >= ICE_AQC_TOPO_MAX_LEVEL_NUM ||
66 	    start_node->info.data.elem_type == ICE_AQC_ELEM_TYPE_LEAF)
67 		return NULL;
68 
69 	/* Check if TEID matches to any of the children nodes */
70 	for (i = 0; i < start_node->num_children; i++)
71 		if (ICE_TXSCHED_GET_NODE_TEID(start_node->children[i]) == teid)
72 			return start_node->children[i];
73 
74 	/* Search within each child's sub-tree */
75 	for (i = 0; i < start_node->num_children; i++) {
76 		struct ice_sched_node *tmp;
77 
78 		tmp = ice_sched_find_node_by_teid(start_node->children[i],
79 						  teid);
80 		if (tmp)
81 			return tmp;
82 	}
83 
84 	return NULL;
85 }
86 
87 /**
88  * ice_aqc_send_sched_elem_cmd - send scheduling elements cmd
89  * @hw: pointer to the HW struct
90  * @cmd_opc: cmd opcode
91  * @elems_req: number of elements to request
92  * @buf: pointer to buffer
93  * @buf_size: buffer size in bytes
94  * @elems_resp: returns total number of elements response
95  * @cd: pointer to command details structure or NULL
96  *
97  * This function sends a scheduling elements cmd (cmd_opc)
98  */
99 static enum ice_status
100 ice_aqc_send_sched_elem_cmd(struct ice_hw *hw, enum ice_adminq_opc cmd_opc,
101 			    u16 elems_req, void *buf, u16 buf_size,
102 			    u16 *elems_resp, struct ice_sq_cd *cd)
103 {
104 	struct ice_aqc_sched_elem_cmd *cmd;
105 	struct ice_aq_desc desc;
106 	enum ice_status status;
107 
108 	cmd = &desc.params.sched_elem_cmd;
109 	ice_fill_dflt_direct_cmd_desc(&desc, cmd_opc);
110 	cmd->num_elem_req = cpu_to_le16(elems_req);
111 	desc.flags |= cpu_to_le16(ICE_AQ_FLAG_RD);
112 	status = ice_aq_send_cmd(hw, &desc, buf, buf_size, cd);
113 	if (!status && elems_resp)
114 		*elems_resp = le16_to_cpu(cmd->num_elem_resp);
115 
116 	return status;
117 }
118 
119 /**
120  * ice_aq_query_sched_elems - query scheduler elements
121  * @hw: pointer to the HW struct
122  * @elems_req: number of elements to query
123  * @buf: pointer to buffer
124  * @buf_size: buffer size in bytes
125  * @elems_ret: returns total number of elements returned
126  * @cd: pointer to command details structure or NULL
127  *
128  * Query scheduling elements (0x0404)
129  */
130 enum ice_status
131 ice_aq_query_sched_elems(struct ice_hw *hw, u16 elems_req,
132 			 struct ice_aqc_get_elem *buf, u16 buf_size,
133 			 u16 *elems_ret, struct ice_sq_cd *cd)
134 {
135 	return ice_aqc_send_sched_elem_cmd(hw, ice_aqc_opc_get_sched_elems,
136 					   elems_req, (void *)buf, buf_size,
137 					   elems_ret, cd);
138 }
139 
140 /**
141  * ice_sched_add_node - Insert the Tx scheduler node in SW DB
142  * @pi: port information structure
143  * @layer: Scheduler layer of the node
144  * @info: Scheduler element information from firmware
145  *
146  * This function inserts a scheduler node to the SW DB.
147  */
148 enum ice_status
149 ice_sched_add_node(struct ice_port_info *pi, u8 layer,
150 		   struct ice_aqc_txsched_elem_data *info)
151 {
152 	struct ice_sched_node *parent;
153 	struct ice_aqc_get_elem elem;
154 	struct ice_sched_node *node;
155 	enum ice_status status;
156 	struct ice_hw *hw;
157 
158 	if (!pi)
159 		return ICE_ERR_PARAM;
160 
161 	hw = pi->hw;
162 
163 	/* A valid parent node should be there */
164 	parent = ice_sched_find_node_by_teid(pi->root,
165 					     le32_to_cpu(info->parent_teid));
166 	if (!parent) {
167 		ice_debug(hw, ICE_DBG_SCHED,
168 			  "Parent Node not found for parent_teid=0x%x\n",
169 			  le32_to_cpu(info->parent_teid));
170 		return ICE_ERR_PARAM;
171 	}
172 
173 	/* query the current node information from FW  before additing it
174 	 * to the SW DB
175 	 */
176 	status = ice_sched_query_elem(hw, le32_to_cpu(info->node_teid), &elem);
177 	if (status)
178 		return status;
179 
180 	node = devm_kzalloc(ice_hw_to_dev(hw), sizeof(*node), GFP_KERNEL);
181 	if (!node)
182 		return ICE_ERR_NO_MEMORY;
183 	if (hw->max_children[layer]) {
184 		/* coverity[suspicious_sizeof] */
185 		node->children = devm_kcalloc(ice_hw_to_dev(hw),
186 					      hw->max_children[layer],
187 					      sizeof(*node), GFP_KERNEL);
188 		if (!node->children) {
189 			devm_kfree(ice_hw_to_dev(hw), node);
190 			return ICE_ERR_NO_MEMORY;
191 		}
192 	}
193 
194 	node->in_use = true;
195 	node->parent = parent;
196 	node->tx_sched_layer = layer;
197 	parent->children[parent->num_children++] = node;
198 	memcpy(&node->info, &elem.generic[0], sizeof(node->info));
199 	return 0;
200 }
201 
202 /**
203  * ice_aq_delete_sched_elems - delete scheduler elements
204  * @hw: pointer to the HW struct
205  * @grps_req: number of groups to delete
206  * @buf: pointer to buffer
207  * @buf_size: buffer size in bytes
208  * @grps_del: returns total number of elements deleted
209  * @cd: pointer to command details structure or NULL
210  *
211  * Delete scheduling elements (0x040F)
212  */
213 static enum ice_status
214 ice_aq_delete_sched_elems(struct ice_hw *hw, u16 grps_req,
215 			  struct ice_aqc_delete_elem *buf, u16 buf_size,
216 			  u16 *grps_del, struct ice_sq_cd *cd)
217 {
218 	return ice_aqc_send_sched_elem_cmd(hw, ice_aqc_opc_delete_sched_elems,
219 					   grps_req, (void *)buf, buf_size,
220 					   grps_del, cd);
221 }
222 
223 /**
224  * ice_sched_remove_elems - remove nodes from HW
225  * @hw: pointer to the HW struct
226  * @parent: pointer to the parent node
227  * @num_nodes: number of nodes
228  * @node_teids: array of node teids to be deleted
229  *
230  * This function remove nodes from HW
231  */
232 static enum ice_status
233 ice_sched_remove_elems(struct ice_hw *hw, struct ice_sched_node *parent,
234 		       u16 num_nodes, u32 *node_teids)
235 {
236 	struct ice_aqc_delete_elem *buf;
237 	u16 i, num_groups_removed = 0;
238 	enum ice_status status;
239 	u16 buf_size;
240 
241 	buf_size = sizeof(*buf) + sizeof(u32) * (num_nodes - 1);
242 	buf = devm_kzalloc(ice_hw_to_dev(hw), buf_size, GFP_KERNEL);
243 	if (!buf)
244 		return ICE_ERR_NO_MEMORY;
245 
246 	buf->hdr.parent_teid = parent->info.node_teid;
247 	buf->hdr.num_elems = cpu_to_le16(num_nodes);
248 	for (i = 0; i < num_nodes; i++)
249 		buf->teid[i] = cpu_to_le32(node_teids[i]);
250 
251 	status = ice_aq_delete_sched_elems(hw, 1, buf, buf_size,
252 					   &num_groups_removed, NULL);
253 	if (status || num_groups_removed != 1)
254 		ice_debug(hw, ICE_DBG_SCHED, "remove node failed FW error %d\n",
255 			  hw->adminq.sq_last_status);
256 
257 	devm_kfree(ice_hw_to_dev(hw), buf);
258 	return status;
259 }
260 
261 /**
262  * ice_sched_get_first_node - get the first node of the given layer
263  * @pi: port information structure
264  * @parent: pointer the base node of the subtree
265  * @layer: layer number
266  *
267  * This function retrieves the first node of the given layer from the subtree
268  */
269 static struct ice_sched_node *
270 ice_sched_get_first_node(struct ice_port_info *pi,
271 			 struct ice_sched_node *parent, u8 layer)
272 {
273 	return pi->sib_head[parent->tc_num][layer];
274 }
275 
276 /**
277  * ice_sched_get_tc_node - get pointer to TC node
278  * @pi: port information structure
279  * @tc: TC number
280  *
281  * This function returns the TC node pointer
282  */
283 struct ice_sched_node *ice_sched_get_tc_node(struct ice_port_info *pi, u8 tc)
284 {
285 	u8 i;
286 
287 	if (!pi || !pi->root)
288 		return NULL;
289 	for (i = 0; i < pi->root->num_children; i++)
290 		if (pi->root->children[i]->tc_num == tc)
291 			return pi->root->children[i];
292 	return NULL;
293 }
294 
295 /**
296  * ice_free_sched_node - Free a Tx scheduler node from SW DB
297  * @pi: port information structure
298  * @node: pointer to the ice_sched_node struct
299  *
300  * This function frees up a node from SW DB as well as from HW
301  *
302  * This function needs to be called with the port_info->sched_lock held
303  */
304 void ice_free_sched_node(struct ice_port_info *pi, struct ice_sched_node *node)
305 {
306 	struct ice_sched_node *parent;
307 	struct ice_hw *hw = pi->hw;
308 	u8 i, j;
309 
310 	/* Free the children before freeing up the parent node
311 	 * The parent array is updated below and that shifts the nodes
312 	 * in the array. So always pick the first child if num children > 0
313 	 */
314 	while (node->num_children)
315 		ice_free_sched_node(pi, node->children[0]);
316 
317 	/* Leaf, TC and root nodes can't be deleted by SW */
318 	if (node->tx_sched_layer >= hw->sw_entry_point_layer &&
319 	    node->info.data.elem_type != ICE_AQC_ELEM_TYPE_TC &&
320 	    node->info.data.elem_type != ICE_AQC_ELEM_TYPE_ROOT_PORT &&
321 	    node->info.data.elem_type != ICE_AQC_ELEM_TYPE_LEAF) {
322 		u32 teid = le32_to_cpu(node->info.node_teid);
323 
324 		ice_sched_remove_elems(hw, node->parent, 1, &teid);
325 	}
326 	parent = node->parent;
327 	/* root has no parent */
328 	if (parent) {
329 		struct ice_sched_node *p;
330 
331 		/* update the parent */
332 		for (i = 0; i < parent->num_children; i++)
333 			if (parent->children[i] == node) {
334 				for (j = i + 1; j < parent->num_children; j++)
335 					parent->children[j - 1] =
336 						parent->children[j];
337 				parent->num_children--;
338 				break;
339 			}
340 
341 		p = ice_sched_get_first_node(pi, node, node->tx_sched_layer);
342 		while (p) {
343 			if (p->sibling == node) {
344 				p->sibling = node->sibling;
345 				break;
346 			}
347 			p = p->sibling;
348 		}
349 
350 		/* update the sibling head if head is getting removed */
351 		if (pi->sib_head[node->tc_num][node->tx_sched_layer] == node)
352 			pi->sib_head[node->tc_num][node->tx_sched_layer] =
353 				node->sibling;
354 	}
355 
356 	/* leaf nodes have no children */
357 	if (node->children)
358 		devm_kfree(ice_hw_to_dev(hw), node->children);
359 	devm_kfree(ice_hw_to_dev(hw), node);
360 }
361 
362 /**
363  * ice_aq_get_dflt_topo - gets default scheduler topology
364  * @hw: pointer to the HW struct
365  * @lport: logical port number
366  * @buf: pointer to buffer
367  * @buf_size: buffer size in bytes
368  * @num_branches: returns total number of queue to port branches
369  * @cd: pointer to command details structure or NULL
370  *
371  * Get default scheduler topology (0x400)
372  */
373 static enum ice_status
374 ice_aq_get_dflt_topo(struct ice_hw *hw, u8 lport,
375 		     struct ice_aqc_get_topo_elem *buf, u16 buf_size,
376 		     u8 *num_branches, struct ice_sq_cd *cd)
377 {
378 	struct ice_aqc_get_topo *cmd;
379 	struct ice_aq_desc desc;
380 	enum ice_status status;
381 
382 	cmd = &desc.params.get_topo;
383 	ice_fill_dflt_direct_cmd_desc(&desc, ice_aqc_opc_get_dflt_topo);
384 	cmd->port_num = lport;
385 	status = ice_aq_send_cmd(hw, &desc, buf, buf_size, cd);
386 	if (!status && num_branches)
387 		*num_branches = cmd->num_branches;
388 
389 	return status;
390 }
391 
392 /**
393  * ice_aq_add_sched_elems - adds scheduling element
394  * @hw: pointer to the HW struct
395  * @grps_req: the number of groups that are requested to be added
396  * @buf: pointer to buffer
397  * @buf_size: buffer size in bytes
398  * @grps_added: returns total number of groups added
399  * @cd: pointer to command details structure or NULL
400  *
401  * Add scheduling elements (0x0401)
402  */
403 static enum ice_status
404 ice_aq_add_sched_elems(struct ice_hw *hw, u16 grps_req,
405 		       struct ice_aqc_add_elem *buf, u16 buf_size,
406 		       u16 *grps_added, struct ice_sq_cd *cd)
407 {
408 	return ice_aqc_send_sched_elem_cmd(hw, ice_aqc_opc_add_sched_elems,
409 					   grps_req, (void *)buf, buf_size,
410 					   grps_added, cd);
411 }
412 
413 /**
414  * ice_aq_cfg_sched_elems - configures scheduler elements
415  * @hw: pointer to the HW struct
416  * @elems_req: number of elements to configure
417  * @buf: pointer to buffer
418  * @buf_size: buffer size in bytes
419  * @elems_cfgd: returns total number of elements configured
420  * @cd: pointer to command details structure or NULL
421  *
422  * Configure scheduling elements (0x0403)
423  */
424 static enum ice_status
425 ice_aq_cfg_sched_elems(struct ice_hw *hw, u16 elems_req,
426 		       struct ice_aqc_conf_elem *buf, u16 buf_size,
427 		       u16 *elems_cfgd, struct ice_sq_cd *cd)
428 {
429 	return ice_aqc_send_sched_elem_cmd(hw, ice_aqc_opc_cfg_sched_elems,
430 					   elems_req, (void *)buf, buf_size,
431 					   elems_cfgd, cd);
432 }
433 
434 /**
435  * ice_aq_suspend_sched_elems - suspend scheduler elements
436  * @hw: pointer to the HW struct
437  * @elems_req: number of elements to suspend
438  * @buf: pointer to buffer
439  * @buf_size: buffer size in bytes
440  * @elems_ret: returns total number of elements suspended
441  * @cd: pointer to command details structure or NULL
442  *
443  * Suspend scheduling elements (0x0409)
444  */
445 static enum ice_status
446 ice_aq_suspend_sched_elems(struct ice_hw *hw, u16 elems_req,
447 			   struct ice_aqc_suspend_resume_elem *buf,
448 			   u16 buf_size, u16 *elems_ret, struct ice_sq_cd *cd)
449 {
450 	return ice_aqc_send_sched_elem_cmd(hw, ice_aqc_opc_suspend_sched_elems,
451 					   elems_req, (void *)buf, buf_size,
452 					   elems_ret, cd);
453 }
454 
455 /**
456  * ice_aq_resume_sched_elems - resume scheduler elements
457  * @hw: pointer to the HW struct
458  * @elems_req: number of elements to resume
459  * @buf: pointer to buffer
460  * @buf_size: buffer size in bytes
461  * @elems_ret: returns total number of elements resumed
462  * @cd: pointer to command details structure or NULL
463  *
464  * resume scheduling elements (0x040A)
465  */
466 static enum ice_status
467 ice_aq_resume_sched_elems(struct ice_hw *hw, u16 elems_req,
468 			  struct ice_aqc_suspend_resume_elem *buf,
469 			  u16 buf_size, u16 *elems_ret, struct ice_sq_cd *cd)
470 {
471 	return ice_aqc_send_sched_elem_cmd(hw, ice_aqc_opc_resume_sched_elems,
472 					   elems_req, (void *)buf, buf_size,
473 					   elems_ret, cd);
474 }
475 
476 /**
477  * ice_aq_query_sched_res - query scheduler resource
478  * @hw: pointer to the HW struct
479  * @buf_size: buffer size in bytes
480  * @buf: pointer to buffer
481  * @cd: pointer to command details structure or NULL
482  *
483  * Query scheduler resource allocation (0x0412)
484  */
485 static enum ice_status
486 ice_aq_query_sched_res(struct ice_hw *hw, u16 buf_size,
487 		       struct ice_aqc_query_txsched_res_resp *buf,
488 		       struct ice_sq_cd *cd)
489 {
490 	struct ice_aq_desc desc;
491 
492 	ice_fill_dflt_direct_cmd_desc(&desc, ice_aqc_opc_query_sched_res);
493 	return ice_aq_send_cmd(hw, &desc, buf, buf_size, cd);
494 }
495 
496 /**
497  * ice_sched_suspend_resume_elems - suspend or resume HW nodes
498  * @hw: pointer to the HW struct
499  * @num_nodes: number of nodes
500  * @node_teids: array of node teids to be suspended or resumed
501  * @suspend: true means suspend / false means resume
502  *
503  * This function suspends or resumes HW nodes
504  */
505 static enum ice_status
506 ice_sched_suspend_resume_elems(struct ice_hw *hw, u8 num_nodes, u32 *node_teids,
507 			       bool suspend)
508 {
509 	struct ice_aqc_suspend_resume_elem *buf;
510 	u16 i, buf_size, num_elem_ret = 0;
511 	enum ice_status status;
512 
513 	buf_size = sizeof(*buf) * num_nodes;
514 	buf = devm_kzalloc(ice_hw_to_dev(hw), buf_size, GFP_KERNEL);
515 	if (!buf)
516 		return ICE_ERR_NO_MEMORY;
517 
518 	for (i = 0; i < num_nodes; i++)
519 		buf->teid[i] = cpu_to_le32(node_teids[i]);
520 
521 	if (suspend)
522 		status = ice_aq_suspend_sched_elems(hw, num_nodes, buf,
523 						    buf_size, &num_elem_ret,
524 						    NULL);
525 	else
526 		status = ice_aq_resume_sched_elems(hw, num_nodes, buf,
527 						   buf_size, &num_elem_ret,
528 						   NULL);
529 	if (status || num_elem_ret != num_nodes)
530 		ice_debug(hw, ICE_DBG_SCHED, "suspend/resume failed\n");
531 
532 	devm_kfree(ice_hw_to_dev(hw), buf);
533 	return status;
534 }
535 
536 /**
537  * ice_alloc_lan_q_ctx - allocate LAN queue contexts for the given VSI and TC
538  * @hw: pointer to the HW struct
539  * @vsi_handle: VSI handle
540  * @tc: TC number
541  * @new_numqs: number of queues
542  */
543 static enum ice_status
544 ice_alloc_lan_q_ctx(struct ice_hw *hw, u16 vsi_handle, u8 tc, u16 new_numqs)
545 {
546 	struct ice_vsi_ctx *vsi_ctx;
547 	struct ice_q_ctx *q_ctx;
548 
549 	vsi_ctx = ice_get_vsi_ctx(hw, vsi_handle);
550 	if (!vsi_ctx)
551 		return ICE_ERR_PARAM;
552 	/* allocate LAN queue contexts */
553 	if (!vsi_ctx->lan_q_ctx[tc]) {
554 		vsi_ctx->lan_q_ctx[tc] = devm_kcalloc(ice_hw_to_dev(hw),
555 						      new_numqs,
556 						      sizeof(*q_ctx),
557 						      GFP_KERNEL);
558 		if (!vsi_ctx->lan_q_ctx[tc])
559 			return ICE_ERR_NO_MEMORY;
560 		vsi_ctx->num_lan_q_entries[tc] = new_numqs;
561 		return 0;
562 	}
563 	/* num queues are increased, update the queue contexts */
564 	if (new_numqs > vsi_ctx->num_lan_q_entries[tc]) {
565 		u16 prev_num = vsi_ctx->num_lan_q_entries[tc];
566 
567 		q_ctx = devm_kcalloc(ice_hw_to_dev(hw), new_numqs,
568 				     sizeof(*q_ctx), GFP_KERNEL);
569 		if (!q_ctx)
570 			return ICE_ERR_NO_MEMORY;
571 		memcpy(q_ctx, vsi_ctx->lan_q_ctx[tc],
572 		       prev_num * sizeof(*q_ctx));
573 		devm_kfree(ice_hw_to_dev(hw), vsi_ctx->lan_q_ctx[tc]);
574 		vsi_ctx->lan_q_ctx[tc] = q_ctx;
575 		vsi_ctx->num_lan_q_entries[tc] = new_numqs;
576 	}
577 	return 0;
578 }
579 
580 /**
581  * ice_aq_rl_profile - performs a rate limiting task
582  * @hw: pointer to the HW struct
583  * @opcode:opcode for add, query, or remove profile(s)
584  * @num_profiles: the number of profiles
585  * @buf: pointer to buffer
586  * @buf_size: buffer size in bytes
587  * @num_processed: number of processed add or remove profile(s) to return
588  * @cd: pointer to command details structure
589  *
590  * RL profile function to add, query, or remove profile(s)
591  */
592 static enum ice_status
593 ice_aq_rl_profile(struct ice_hw *hw, enum ice_adminq_opc opcode,
594 		  u16 num_profiles, struct ice_aqc_rl_profile_generic_elem *buf,
595 		  u16 buf_size, u16 *num_processed, struct ice_sq_cd *cd)
596 {
597 	struct ice_aqc_rl_profile *cmd;
598 	struct ice_aq_desc desc;
599 	enum ice_status status;
600 
601 	cmd = &desc.params.rl_profile;
602 
603 	ice_fill_dflt_direct_cmd_desc(&desc, opcode);
604 	desc.flags |= cpu_to_le16(ICE_AQ_FLAG_RD);
605 	cmd->num_profiles = cpu_to_le16(num_profiles);
606 	status = ice_aq_send_cmd(hw, &desc, buf, buf_size, cd);
607 	if (!status && num_processed)
608 		*num_processed = le16_to_cpu(cmd->num_processed);
609 	return status;
610 }
611 
612 /**
613  * ice_aq_add_rl_profile - adds rate limiting profile(s)
614  * @hw: pointer to the HW struct
615  * @num_profiles: the number of profile(s) to be add
616  * @buf: pointer to buffer
617  * @buf_size: buffer size in bytes
618  * @num_profiles_added: total number of profiles added to return
619  * @cd: pointer to command details structure
620  *
621  * Add RL profile (0x0410)
622  */
623 static enum ice_status
624 ice_aq_add_rl_profile(struct ice_hw *hw, u16 num_profiles,
625 		      struct ice_aqc_rl_profile_generic_elem *buf,
626 		      u16 buf_size, u16 *num_profiles_added,
627 		      struct ice_sq_cd *cd)
628 {
629 	return ice_aq_rl_profile(hw, ice_aqc_opc_add_rl_profiles,
630 				 num_profiles, buf,
631 				 buf_size, num_profiles_added, cd);
632 }
633 
634 /**
635  * ice_aq_remove_rl_profile - removes RL profile(s)
636  * @hw: pointer to the HW struct
637  * @num_profiles: the number of profile(s) to remove
638  * @buf: pointer to buffer
639  * @buf_size: buffer size in bytes
640  * @num_profiles_removed: total number of profiles removed to return
641  * @cd: pointer to command details structure or NULL
642  *
643  * Remove RL profile (0x0415)
644  */
645 static enum ice_status
646 ice_aq_remove_rl_profile(struct ice_hw *hw, u16 num_profiles,
647 			 struct ice_aqc_rl_profile_generic_elem *buf,
648 			 u16 buf_size, u16 *num_profiles_removed,
649 			 struct ice_sq_cd *cd)
650 {
651 	return ice_aq_rl_profile(hw, ice_aqc_opc_remove_rl_profiles,
652 				 num_profiles, buf,
653 				 buf_size, num_profiles_removed, cd);
654 }
655 
656 /**
657  * ice_sched_del_rl_profile - remove RL profile
658  * @hw: pointer to the HW struct
659  * @rl_info: rate limit profile information
660  *
661  * If the profile ID is not referenced anymore, it removes profile ID with
662  * its associated parameters from HW DB,and locally. The caller needs to
663  * hold scheduler lock.
664  */
665 static enum ice_status
666 ice_sched_del_rl_profile(struct ice_hw *hw,
667 			 struct ice_aqc_rl_profile_info *rl_info)
668 {
669 	struct ice_aqc_rl_profile_generic_elem *buf;
670 	u16 num_profiles_removed;
671 	enum ice_status status;
672 	u16 num_profiles = 1;
673 
674 	if (rl_info->prof_id_ref != 0)
675 		return ICE_ERR_IN_USE;
676 
677 	/* Safe to remove profile ID */
678 	buf = (struct ice_aqc_rl_profile_generic_elem *)
679 		&rl_info->profile;
680 	status = ice_aq_remove_rl_profile(hw, num_profiles, buf, sizeof(*buf),
681 					  &num_profiles_removed, NULL);
682 	if (status || num_profiles_removed != num_profiles)
683 		return ICE_ERR_CFG;
684 
685 	/* Delete stale entry now */
686 	list_del(&rl_info->list_entry);
687 	devm_kfree(ice_hw_to_dev(hw), rl_info);
688 	return status;
689 }
690 
691 /**
692  * ice_sched_clear_rl_prof - clears RL prof entries
693  * @pi: port information structure
694  *
695  * This function removes all RL profile from HW as well as from SW DB.
696  */
697 static void ice_sched_clear_rl_prof(struct ice_port_info *pi)
698 {
699 	u16 ln;
700 
701 	for (ln = 0; ln < pi->hw->num_tx_sched_layers; ln++) {
702 		struct ice_aqc_rl_profile_info *rl_prof_elem;
703 		struct ice_aqc_rl_profile_info *rl_prof_tmp;
704 
705 		list_for_each_entry_safe(rl_prof_elem, rl_prof_tmp,
706 					 &pi->rl_prof_list[ln], list_entry) {
707 			struct ice_hw *hw = pi->hw;
708 			enum ice_status status;
709 
710 			rl_prof_elem->prof_id_ref = 0;
711 			status = ice_sched_del_rl_profile(hw, rl_prof_elem);
712 			if (status) {
713 				ice_debug(hw, ICE_DBG_SCHED,
714 					  "Remove rl profile failed\n");
715 				/* On error, free mem required */
716 				list_del(&rl_prof_elem->list_entry);
717 				devm_kfree(ice_hw_to_dev(hw), rl_prof_elem);
718 			}
719 		}
720 	}
721 }
722 
723 /**
724  * ice_sched_clear_agg - clears the aggregator related information
725  * @hw: pointer to the hardware structure
726  *
727  * This function removes aggregator list and free up aggregator related memory
728  * previously allocated.
729  */
730 void ice_sched_clear_agg(struct ice_hw *hw)
731 {
732 	struct ice_sched_agg_info *agg_info;
733 	struct ice_sched_agg_info *atmp;
734 
735 	list_for_each_entry_safe(agg_info, atmp, &hw->agg_list, list_entry) {
736 		struct ice_sched_agg_vsi_info *agg_vsi_info;
737 		struct ice_sched_agg_vsi_info *vtmp;
738 
739 		list_for_each_entry_safe(agg_vsi_info, vtmp,
740 					 &agg_info->agg_vsi_list, list_entry) {
741 			list_del(&agg_vsi_info->list_entry);
742 			devm_kfree(ice_hw_to_dev(hw), agg_vsi_info);
743 		}
744 		list_del(&agg_info->list_entry);
745 		devm_kfree(ice_hw_to_dev(hw), agg_info);
746 	}
747 }
748 
749 /**
750  * ice_sched_clear_tx_topo - clears the scheduler tree nodes
751  * @pi: port information structure
752  *
753  * This function removes all the nodes from HW as well as from SW DB.
754  */
755 static void ice_sched_clear_tx_topo(struct ice_port_info *pi)
756 {
757 	if (!pi)
758 		return;
759 	/* remove RL profiles related lists */
760 	ice_sched_clear_rl_prof(pi);
761 	if (pi->root) {
762 		ice_free_sched_node(pi, pi->root);
763 		pi->root = NULL;
764 	}
765 }
766 
767 /**
768  * ice_sched_clear_port - clear the scheduler elements from SW DB for a port
769  * @pi: port information structure
770  *
771  * Cleanup scheduling elements from SW DB
772  */
773 void ice_sched_clear_port(struct ice_port_info *pi)
774 {
775 	if (!pi || pi->port_state != ICE_SCHED_PORT_STATE_READY)
776 		return;
777 
778 	pi->port_state = ICE_SCHED_PORT_STATE_INIT;
779 	mutex_lock(&pi->sched_lock);
780 	ice_sched_clear_tx_topo(pi);
781 	mutex_unlock(&pi->sched_lock);
782 	mutex_destroy(&pi->sched_lock);
783 }
784 
785 /**
786  * ice_sched_cleanup_all - cleanup scheduler elements from SW DB for all ports
787  * @hw: pointer to the HW struct
788  *
789  * Cleanup scheduling elements from SW DB for all the ports
790  */
791 void ice_sched_cleanup_all(struct ice_hw *hw)
792 {
793 	if (!hw)
794 		return;
795 
796 	if (hw->layer_info) {
797 		devm_kfree(ice_hw_to_dev(hw), hw->layer_info);
798 		hw->layer_info = NULL;
799 	}
800 
801 	if (hw->port_info)
802 		ice_sched_clear_port(hw->port_info);
803 
804 	hw->num_tx_sched_layers = 0;
805 	hw->num_tx_sched_phys_layers = 0;
806 	hw->flattened_layers = 0;
807 	hw->max_cgds = 0;
808 }
809 
810 /**
811  * ice_sched_add_elems - add nodes to HW and SW DB
812  * @pi: port information structure
813  * @tc_node: pointer to the branch node
814  * @parent: pointer to the parent node
815  * @layer: layer number to add nodes
816  * @num_nodes: number of nodes
817  * @num_nodes_added: pointer to num nodes added
818  * @first_node_teid: if new nodes are added then return the TEID of first node
819  *
820  * This function add nodes to HW as well as to SW DB for a given layer
821  */
822 static enum ice_status
823 ice_sched_add_elems(struct ice_port_info *pi, struct ice_sched_node *tc_node,
824 		    struct ice_sched_node *parent, u8 layer, u16 num_nodes,
825 		    u16 *num_nodes_added, u32 *first_node_teid)
826 {
827 	struct ice_sched_node *prev, *new_node;
828 	struct ice_aqc_add_elem *buf;
829 	u16 i, num_groups_added = 0;
830 	enum ice_status status = 0;
831 	struct ice_hw *hw = pi->hw;
832 	size_t buf_size;
833 	u32 teid;
834 
835 	buf_size = struct_size(buf, generic, num_nodes - 1);
836 	buf = devm_kzalloc(ice_hw_to_dev(hw), buf_size, GFP_KERNEL);
837 	if (!buf)
838 		return ICE_ERR_NO_MEMORY;
839 
840 	buf->hdr.parent_teid = parent->info.node_teid;
841 	buf->hdr.num_elems = cpu_to_le16(num_nodes);
842 	for (i = 0; i < num_nodes; i++) {
843 		buf->generic[i].parent_teid = parent->info.node_teid;
844 		buf->generic[i].data.elem_type = ICE_AQC_ELEM_TYPE_SE_GENERIC;
845 		buf->generic[i].data.valid_sections =
846 			ICE_AQC_ELEM_VALID_GENERIC | ICE_AQC_ELEM_VALID_CIR |
847 			ICE_AQC_ELEM_VALID_EIR;
848 		buf->generic[i].data.generic = 0;
849 		buf->generic[i].data.cir_bw.bw_profile_idx =
850 			cpu_to_le16(ICE_SCHED_DFLT_RL_PROF_ID);
851 		buf->generic[i].data.cir_bw.bw_alloc =
852 			cpu_to_le16(ICE_SCHED_DFLT_BW_WT);
853 		buf->generic[i].data.eir_bw.bw_profile_idx =
854 			cpu_to_le16(ICE_SCHED_DFLT_RL_PROF_ID);
855 		buf->generic[i].data.eir_bw.bw_alloc =
856 			cpu_to_le16(ICE_SCHED_DFLT_BW_WT);
857 	}
858 
859 	status = ice_aq_add_sched_elems(hw, 1, buf, buf_size,
860 					&num_groups_added, NULL);
861 	if (status || num_groups_added != 1) {
862 		ice_debug(hw, ICE_DBG_SCHED, "add node failed FW Error %d\n",
863 			  hw->adminq.sq_last_status);
864 		devm_kfree(ice_hw_to_dev(hw), buf);
865 		return ICE_ERR_CFG;
866 	}
867 
868 	*num_nodes_added = num_nodes;
869 	/* add nodes to the SW DB */
870 	for (i = 0; i < num_nodes; i++) {
871 		status = ice_sched_add_node(pi, layer, &buf->generic[i]);
872 		if (status) {
873 			ice_debug(hw, ICE_DBG_SCHED,
874 				  "add nodes in SW DB failed status =%d\n",
875 				  status);
876 			break;
877 		}
878 
879 		teid = le32_to_cpu(buf->generic[i].node_teid);
880 		new_node = ice_sched_find_node_by_teid(parent, teid);
881 		if (!new_node) {
882 			ice_debug(hw, ICE_DBG_SCHED,
883 				  "Node is missing for teid =%d\n", teid);
884 			break;
885 		}
886 
887 		new_node->sibling = NULL;
888 		new_node->tc_num = tc_node->tc_num;
889 
890 		/* add it to previous node sibling pointer */
891 		/* Note: siblings are not linked across branches */
892 		prev = ice_sched_get_first_node(pi, tc_node, layer);
893 		if (prev && prev != new_node) {
894 			while (prev->sibling)
895 				prev = prev->sibling;
896 			prev->sibling = new_node;
897 		}
898 
899 		/* initialize the sibling head */
900 		if (!pi->sib_head[tc_node->tc_num][layer])
901 			pi->sib_head[tc_node->tc_num][layer] = new_node;
902 
903 		if (i == 0)
904 			*first_node_teid = teid;
905 	}
906 
907 	devm_kfree(ice_hw_to_dev(hw), buf);
908 	return status;
909 }
910 
911 /**
912  * ice_sched_add_nodes_to_layer - Add nodes to a given layer
913  * @pi: port information structure
914  * @tc_node: pointer to TC node
915  * @parent: pointer to parent node
916  * @layer: layer number to add nodes
917  * @num_nodes: number of nodes to be added
918  * @first_node_teid: pointer to the first node TEID
919  * @num_nodes_added: pointer to number of nodes added
920  *
921  * This function add nodes to a given layer.
922  */
923 static enum ice_status
924 ice_sched_add_nodes_to_layer(struct ice_port_info *pi,
925 			     struct ice_sched_node *tc_node,
926 			     struct ice_sched_node *parent, u8 layer,
927 			     u16 num_nodes, u32 *first_node_teid,
928 			     u16 *num_nodes_added)
929 {
930 	u32 *first_teid_ptr = first_node_teid;
931 	u16 new_num_nodes, max_child_nodes;
932 	enum ice_status status = 0;
933 	struct ice_hw *hw = pi->hw;
934 	u16 num_added = 0;
935 	u32 temp;
936 
937 	*num_nodes_added = 0;
938 
939 	if (!num_nodes)
940 		return status;
941 
942 	if (!parent || layer < hw->sw_entry_point_layer)
943 		return ICE_ERR_PARAM;
944 
945 	/* max children per node per layer */
946 	max_child_nodes = hw->max_children[parent->tx_sched_layer];
947 
948 	/* current number of children + required nodes exceed max children ? */
949 	if ((parent->num_children + num_nodes) > max_child_nodes) {
950 		/* Fail if the parent is a TC node */
951 		if (parent == tc_node)
952 			return ICE_ERR_CFG;
953 
954 		/* utilize all the spaces if the parent is not full */
955 		if (parent->num_children < max_child_nodes) {
956 			new_num_nodes = max_child_nodes - parent->num_children;
957 			/* this recursion is intentional, and wouldn't
958 			 * go more than 2 calls
959 			 */
960 			status = ice_sched_add_nodes_to_layer(pi, tc_node,
961 							      parent, layer,
962 							      new_num_nodes,
963 							      first_node_teid,
964 							      &num_added);
965 			if (status)
966 				return status;
967 
968 			*num_nodes_added += num_added;
969 		}
970 		/* Don't modify the first node TEID memory if the first node was
971 		 * added already in the above call. Instead send some temp
972 		 * memory for all other recursive calls.
973 		 */
974 		if (num_added)
975 			first_teid_ptr = &temp;
976 
977 		new_num_nodes = num_nodes - num_added;
978 
979 		/* This parent is full, try the next sibling */
980 		parent = parent->sibling;
981 
982 		/* this recursion is intentional, for 1024 queues
983 		 * per VSI, it goes max of 16 iterations.
984 		 * 1024 / 8 = 128 layer 8 nodes
985 		 * 128 /8 = 16 (add 8 nodes per iteration)
986 		 */
987 		status = ice_sched_add_nodes_to_layer(pi, tc_node, parent,
988 						      layer, new_num_nodes,
989 						      first_teid_ptr,
990 						      &num_added);
991 		*num_nodes_added += num_added;
992 		return status;
993 	}
994 
995 	status = ice_sched_add_elems(pi, tc_node, parent, layer, num_nodes,
996 				     num_nodes_added, first_node_teid);
997 	return status;
998 }
999 
1000 /**
1001  * ice_sched_get_qgrp_layer - get the current queue group layer number
1002  * @hw: pointer to the HW struct
1003  *
1004  * This function returns the current queue group layer number
1005  */
1006 static u8 ice_sched_get_qgrp_layer(struct ice_hw *hw)
1007 {
1008 	/* It's always total layers - 1, the array is 0 relative so -2 */
1009 	return hw->num_tx_sched_layers - ICE_QGRP_LAYER_OFFSET;
1010 }
1011 
1012 /**
1013  * ice_sched_get_vsi_layer - get the current VSI layer number
1014  * @hw: pointer to the HW struct
1015  *
1016  * This function returns the current VSI layer number
1017  */
1018 static u8 ice_sched_get_vsi_layer(struct ice_hw *hw)
1019 {
1020 	/* Num Layers       VSI layer
1021 	 *     9               6
1022 	 *     7               4
1023 	 *     5 or less       sw_entry_point_layer
1024 	 */
1025 	/* calculate the VSI layer based on number of layers. */
1026 	if (hw->num_tx_sched_layers > ICE_VSI_LAYER_OFFSET + 1) {
1027 		u8 layer = hw->num_tx_sched_layers - ICE_VSI_LAYER_OFFSET;
1028 
1029 		if (layer > hw->sw_entry_point_layer)
1030 			return layer;
1031 	}
1032 	return hw->sw_entry_point_layer;
1033 }
1034 
1035 /**
1036  * ice_rm_dflt_leaf_node - remove the default leaf node in the tree
1037  * @pi: port information structure
1038  *
1039  * This function removes the leaf node that was created by the FW
1040  * during initialization
1041  */
1042 static void ice_rm_dflt_leaf_node(struct ice_port_info *pi)
1043 {
1044 	struct ice_sched_node *node;
1045 
1046 	node = pi->root;
1047 	while (node) {
1048 		if (!node->num_children)
1049 			break;
1050 		node = node->children[0];
1051 	}
1052 	if (node && node->info.data.elem_type == ICE_AQC_ELEM_TYPE_LEAF) {
1053 		u32 teid = le32_to_cpu(node->info.node_teid);
1054 		enum ice_status status;
1055 
1056 		/* remove the default leaf node */
1057 		status = ice_sched_remove_elems(pi->hw, node->parent, 1, &teid);
1058 		if (!status)
1059 			ice_free_sched_node(pi, node);
1060 	}
1061 }
1062 
1063 /**
1064  * ice_sched_rm_dflt_nodes - free the default nodes in the tree
1065  * @pi: port information structure
1066  *
1067  * This function frees all the nodes except root and TC that were created by
1068  * the FW during initialization
1069  */
1070 static void ice_sched_rm_dflt_nodes(struct ice_port_info *pi)
1071 {
1072 	struct ice_sched_node *node;
1073 
1074 	ice_rm_dflt_leaf_node(pi);
1075 
1076 	/* remove the default nodes except TC and root nodes */
1077 	node = pi->root;
1078 	while (node) {
1079 		if (node->tx_sched_layer >= pi->hw->sw_entry_point_layer &&
1080 		    node->info.data.elem_type != ICE_AQC_ELEM_TYPE_TC &&
1081 		    node->info.data.elem_type != ICE_AQC_ELEM_TYPE_ROOT_PORT) {
1082 			ice_free_sched_node(pi, node);
1083 			break;
1084 		}
1085 
1086 		if (!node->num_children)
1087 			break;
1088 		node = node->children[0];
1089 	}
1090 }
1091 
1092 /**
1093  * ice_sched_init_port - Initialize scheduler by querying information from FW
1094  * @pi: port info structure for the tree to cleanup
1095  *
1096  * This function is the initial call to find the total number of Tx scheduler
1097  * resources, default topology created by firmware and storing the information
1098  * in SW DB.
1099  */
1100 enum ice_status ice_sched_init_port(struct ice_port_info *pi)
1101 {
1102 	struct ice_aqc_get_topo_elem *buf;
1103 	enum ice_status status;
1104 	struct ice_hw *hw;
1105 	u8 num_branches;
1106 	u16 num_elems;
1107 	u8 i, j;
1108 
1109 	if (!pi)
1110 		return ICE_ERR_PARAM;
1111 	hw = pi->hw;
1112 
1113 	/* Query the Default Topology from FW */
1114 	buf = devm_kzalloc(ice_hw_to_dev(hw), ICE_AQ_MAX_BUF_LEN, GFP_KERNEL);
1115 	if (!buf)
1116 		return ICE_ERR_NO_MEMORY;
1117 
1118 	/* Query default scheduling tree topology */
1119 	status = ice_aq_get_dflt_topo(hw, pi->lport, buf, ICE_AQ_MAX_BUF_LEN,
1120 				      &num_branches, NULL);
1121 	if (status)
1122 		goto err_init_port;
1123 
1124 	/* num_branches should be between 1-8 */
1125 	if (num_branches < 1 || num_branches > ICE_TXSCHED_MAX_BRANCHES) {
1126 		ice_debug(hw, ICE_DBG_SCHED, "num_branches unexpected %d\n",
1127 			  num_branches);
1128 		status = ICE_ERR_PARAM;
1129 		goto err_init_port;
1130 	}
1131 
1132 	/* get the number of elements on the default/first branch */
1133 	num_elems = le16_to_cpu(buf[0].hdr.num_elems);
1134 
1135 	/* num_elems should always be between 1-9 */
1136 	if (num_elems < 1 || num_elems > ICE_AQC_TOPO_MAX_LEVEL_NUM) {
1137 		ice_debug(hw, ICE_DBG_SCHED, "num_elems unexpected %d\n",
1138 			  num_elems);
1139 		status = ICE_ERR_PARAM;
1140 		goto err_init_port;
1141 	}
1142 
1143 	/* If the last node is a leaf node then the index of the queue group
1144 	 * layer is two less than the number of elements.
1145 	 */
1146 	if (num_elems > 2 && buf[0].generic[num_elems - 1].data.elem_type ==
1147 	    ICE_AQC_ELEM_TYPE_LEAF)
1148 		pi->last_node_teid =
1149 			le32_to_cpu(buf[0].generic[num_elems - 2].node_teid);
1150 	else
1151 		pi->last_node_teid =
1152 			le32_to_cpu(buf[0].generic[num_elems - 1].node_teid);
1153 
1154 	/* Insert the Tx Sched root node */
1155 	status = ice_sched_add_root_node(pi, &buf[0].generic[0]);
1156 	if (status)
1157 		goto err_init_port;
1158 
1159 	/* Parse the default tree and cache the information */
1160 	for (i = 0; i < num_branches; i++) {
1161 		num_elems = le16_to_cpu(buf[i].hdr.num_elems);
1162 
1163 		/* Skip root element as already inserted */
1164 		for (j = 1; j < num_elems; j++) {
1165 			/* update the sw entry point */
1166 			if (buf[0].generic[j].data.elem_type ==
1167 			    ICE_AQC_ELEM_TYPE_ENTRY_POINT)
1168 				hw->sw_entry_point_layer = j;
1169 
1170 			status = ice_sched_add_node(pi, j, &buf[i].generic[j]);
1171 			if (status)
1172 				goto err_init_port;
1173 		}
1174 	}
1175 
1176 	/* Remove the default nodes. */
1177 	if (pi->root)
1178 		ice_sched_rm_dflt_nodes(pi);
1179 
1180 	/* initialize the port for handling the scheduler tree */
1181 	pi->port_state = ICE_SCHED_PORT_STATE_READY;
1182 	mutex_init(&pi->sched_lock);
1183 	for (i = 0; i < ICE_AQC_TOPO_MAX_LEVEL_NUM; i++)
1184 		INIT_LIST_HEAD(&pi->rl_prof_list[i]);
1185 
1186 err_init_port:
1187 	if (status && pi->root) {
1188 		ice_free_sched_node(pi, pi->root);
1189 		pi->root = NULL;
1190 	}
1191 
1192 	devm_kfree(ice_hw_to_dev(hw), buf);
1193 	return status;
1194 }
1195 
1196 /**
1197  * ice_sched_query_res_alloc - query the FW for num of logical sched layers
1198  * @hw: pointer to the HW struct
1199  *
1200  * query FW for allocated scheduler resources and store in HW struct
1201  */
1202 enum ice_status ice_sched_query_res_alloc(struct ice_hw *hw)
1203 {
1204 	struct ice_aqc_query_txsched_res_resp *buf;
1205 	enum ice_status status = 0;
1206 	__le16 max_sibl;
1207 	u16 i;
1208 
1209 	if (hw->layer_info)
1210 		return status;
1211 
1212 	buf = devm_kzalloc(ice_hw_to_dev(hw), sizeof(*buf), GFP_KERNEL);
1213 	if (!buf)
1214 		return ICE_ERR_NO_MEMORY;
1215 
1216 	status = ice_aq_query_sched_res(hw, sizeof(*buf), buf, NULL);
1217 	if (status)
1218 		goto sched_query_out;
1219 
1220 	hw->num_tx_sched_layers = le16_to_cpu(buf->sched_props.logical_levels);
1221 	hw->num_tx_sched_phys_layers =
1222 		le16_to_cpu(buf->sched_props.phys_levels);
1223 	hw->flattened_layers = buf->sched_props.flattening_bitmap;
1224 	hw->max_cgds = buf->sched_props.max_pf_cgds;
1225 
1226 	/* max sibling group size of current layer refers to the max children
1227 	 * of the below layer node.
1228 	 * layer 1 node max children will be layer 2 max sibling group size
1229 	 * layer 2 node max children will be layer 3 max sibling group size
1230 	 * and so on. This array will be populated from root (index 0) to
1231 	 * qgroup layer 7. Leaf node has no children.
1232 	 */
1233 	for (i = 0; i < hw->num_tx_sched_layers - 1; i++) {
1234 		max_sibl = buf->layer_props[i + 1].max_sibl_grp_sz;
1235 		hw->max_children[i] = le16_to_cpu(max_sibl);
1236 	}
1237 
1238 	hw->layer_info = devm_kmemdup(ice_hw_to_dev(hw), buf->layer_props,
1239 				      (hw->num_tx_sched_layers *
1240 				       sizeof(*hw->layer_info)),
1241 				      GFP_KERNEL);
1242 	if (!hw->layer_info) {
1243 		status = ICE_ERR_NO_MEMORY;
1244 		goto sched_query_out;
1245 	}
1246 
1247 sched_query_out:
1248 	devm_kfree(ice_hw_to_dev(hw), buf);
1249 	return status;
1250 }
1251 
1252 /**
1253  * ice_sched_find_node_in_subtree - Find node in part of base node subtree
1254  * @hw: pointer to the HW struct
1255  * @base: pointer to the base node
1256  * @node: pointer to the node to search
1257  *
1258  * This function checks whether a given node is part of the base node
1259  * subtree or not
1260  */
1261 static bool
1262 ice_sched_find_node_in_subtree(struct ice_hw *hw, struct ice_sched_node *base,
1263 			       struct ice_sched_node *node)
1264 {
1265 	u8 i;
1266 
1267 	for (i = 0; i < base->num_children; i++) {
1268 		struct ice_sched_node *child = base->children[i];
1269 
1270 		if (node == child)
1271 			return true;
1272 
1273 		if (child->tx_sched_layer > node->tx_sched_layer)
1274 			return false;
1275 
1276 		/* this recursion is intentional, and wouldn't
1277 		 * go more than 8 calls
1278 		 */
1279 		if (ice_sched_find_node_in_subtree(hw, child, node))
1280 			return true;
1281 	}
1282 	return false;
1283 }
1284 
1285 /**
1286  * ice_sched_get_free_qparent - Get a free LAN or RDMA queue group node
1287  * @pi: port information structure
1288  * @vsi_handle: software VSI handle
1289  * @tc: branch number
1290  * @owner: LAN or RDMA
1291  *
1292  * This function retrieves a free LAN or RDMA queue group node
1293  */
1294 struct ice_sched_node *
1295 ice_sched_get_free_qparent(struct ice_port_info *pi, u16 vsi_handle, u8 tc,
1296 			   u8 owner)
1297 {
1298 	struct ice_sched_node *vsi_node, *qgrp_node = NULL;
1299 	struct ice_vsi_ctx *vsi_ctx;
1300 	u16 max_children;
1301 	u8 qgrp_layer;
1302 
1303 	qgrp_layer = ice_sched_get_qgrp_layer(pi->hw);
1304 	max_children = pi->hw->max_children[qgrp_layer];
1305 
1306 	vsi_ctx = ice_get_vsi_ctx(pi->hw, vsi_handle);
1307 	if (!vsi_ctx)
1308 		return NULL;
1309 	vsi_node = vsi_ctx->sched.vsi_node[tc];
1310 	/* validate invalid VSI ID */
1311 	if (!vsi_node)
1312 		goto lan_q_exit;
1313 
1314 	/* get the first queue group node from VSI sub-tree */
1315 	qgrp_node = ice_sched_get_first_node(pi, vsi_node, qgrp_layer);
1316 	while (qgrp_node) {
1317 		/* make sure the qgroup node is part of the VSI subtree */
1318 		if (ice_sched_find_node_in_subtree(pi->hw, vsi_node, qgrp_node))
1319 			if (qgrp_node->num_children < max_children &&
1320 			    qgrp_node->owner == owner)
1321 				break;
1322 		qgrp_node = qgrp_node->sibling;
1323 	}
1324 
1325 lan_q_exit:
1326 	return qgrp_node;
1327 }
1328 
1329 /**
1330  * ice_sched_get_vsi_node - Get a VSI node based on VSI ID
1331  * @hw: pointer to the HW struct
1332  * @tc_node: pointer to the TC node
1333  * @vsi_handle: software VSI handle
1334  *
1335  * This function retrieves a VSI node for a given VSI ID from a given
1336  * TC branch
1337  */
1338 static struct ice_sched_node *
1339 ice_sched_get_vsi_node(struct ice_hw *hw, struct ice_sched_node *tc_node,
1340 		       u16 vsi_handle)
1341 {
1342 	struct ice_sched_node *node;
1343 	u8 vsi_layer;
1344 
1345 	vsi_layer = ice_sched_get_vsi_layer(hw);
1346 	node = ice_sched_get_first_node(hw->port_info, tc_node, vsi_layer);
1347 
1348 	/* Check whether it already exists */
1349 	while (node) {
1350 		if (node->vsi_handle == vsi_handle)
1351 			return node;
1352 		node = node->sibling;
1353 	}
1354 
1355 	return node;
1356 }
1357 
1358 /**
1359  * ice_sched_calc_vsi_child_nodes - calculate number of VSI child nodes
1360  * @hw: pointer to the HW struct
1361  * @num_qs: number of queues
1362  * @num_nodes: num nodes array
1363  *
1364  * This function calculates the number of VSI child nodes based on the
1365  * number of queues.
1366  */
1367 static void
1368 ice_sched_calc_vsi_child_nodes(struct ice_hw *hw, u16 num_qs, u16 *num_nodes)
1369 {
1370 	u16 num = num_qs;
1371 	u8 i, qgl, vsil;
1372 
1373 	qgl = ice_sched_get_qgrp_layer(hw);
1374 	vsil = ice_sched_get_vsi_layer(hw);
1375 
1376 	/* calculate num nodes from queue group to VSI layer */
1377 	for (i = qgl; i > vsil; i--) {
1378 		/* round to the next integer if there is a remainder */
1379 		num = DIV_ROUND_UP(num, hw->max_children[i]);
1380 
1381 		/* need at least one node */
1382 		num_nodes[i] = num ? num : 1;
1383 	}
1384 }
1385 
1386 /**
1387  * ice_sched_add_vsi_child_nodes - add VSI child nodes to tree
1388  * @pi: port information structure
1389  * @vsi_handle: software VSI handle
1390  * @tc_node: pointer to the TC node
1391  * @num_nodes: pointer to the num nodes that needs to be added per layer
1392  * @owner: node owner (LAN or RDMA)
1393  *
1394  * This function adds the VSI child nodes to tree. It gets called for
1395  * LAN and RDMA separately.
1396  */
1397 static enum ice_status
1398 ice_sched_add_vsi_child_nodes(struct ice_port_info *pi, u16 vsi_handle,
1399 			      struct ice_sched_node *tc_node, u16 *num_nodes,
1400 			      u8 owner)
1401 {
1402 	struct ice_sched_node *parent, *node;
1403 	struct ice_hw *hw = pi->hw;
1404 	enum ice_status status;
1405 	u32 first_node_teid;
1406 	u16 num_added = 0;
1407 	u8 i, qgl, vsil;
1408 
1409 	qgl = ice_sched_get_qgrp_layer(hw);
1410 	vsil = ice_sched_get_vsi_layer(hw);
1411 	parent = ice_sched_get_vsi_node(hw, tc_node, vsi_handle);
1412 	for (i = vsil + 1; i <= qgl; i++) {
1413 		if (!parent)
1414 			return ICE_ERR_CFG;
1415 
1416 		status = ice_sched_add_nodes_to_layer(pi, tc_node, parent, i,
1417 						      num_nodes[i],
1418 						      &first_node_teid,
1419 						      &num_added);
1420 		if (status || num_nodes[i] != num_added)
1421 			return ICE_ERR_CFG;
1422 
1423 		/* The newly added node can be a new parent for the next
1424 		 * layer nodes
1425 		 */
1426 		if (num_added) {
1427 			parent = ice_sched_find_node_by_teid(tc_node,
1428 							     first_node_teid);
1429 			node = parent;
1430 			while (node) {
1431 				node->owner = owner;
1432 				node = node->sibling;
1433 			}
1434 		} else {
1435 			parent = parent->children[0];
1436 		}
1437 	}
1438 
1439 	return 0;
1440 }
1441 
1442 /**
1443  * ice_sched_calc_vsi_support_nodes - calculate number of VSI support nodes
1444  * @hw: pointer to the HW struct
1445  * @tc_node: pointer to TC node
1446  * @num_nodes: pointer to num nodes array
1447  *
1448  * This function calculates the number of supported nodes needed to add this
1449  * VSI into Tx tree including the VSI, parent and intermediate nodes in below
1450  * layers
1451  */
1452 static void
1453 ice_sched_calc_vsi_support_nodes(struct ice_hw *hw,
1454 				 struct ice_sched_node *tc_node, u16 *num_nodes)
1455 {
1456 	struct ice_sched_node *node;
1457 	u8 vsil;
1458 	int i;
1459 
1460 	vsil = ice_sched_get_vsi_layer(hw);
1461 	for (i = vsil; i >= hw->sw_entry_point_layer; i--)
1462 		/* Add intermediate nodes if TC has no children and
1463 		 * need at least one node for VSI
1464 		 */
1465 		if (!tc_node->num_children || i == vsil) {
1466 			num_nodes[i]++;
1467 		} else {
1468 			/* If intermediate nodes are reached max children
1469 			 * then add a new one.
1470 			 */
1471 			node = ice_sched_get_first_node(hw->port_info, tc_node,
1472 							(u8)i);
1473 			/* scan all the siblings */
1474 			while (node) {
1475 				if (node->num_children < hw->max_children[i])
1476 					break;
1477 				node = node->sibling;
1478 			}
1479 
1480 			/* tree has one intermediate node to add this new VSI.
1481 			 * So no need to calculate supported nodes for below
1482 			 * layers.
1483 			 */
1484 			if (node)
1485 				break;
1486 			/* all the nodes are full, allocate a new one */
1487 			num_nodes[i]++;
1488 		}
1489 }
1490 
1491 /**
1492  * ice_sched_add_vsi_support_nodes - add VSI supported nodes into Tx tree
1493  * @pi: port information structure
1494  * @vsi_handle: software VSI handle
1495  * @tc_node: pointer to TC node
1496  * @num_nodes: pointer to num nodes array
1497  *
1498  * This function adds the VSI supported nodes into Tx tree including the
1499  * VSI, its parent and intermediate nodes in below layers
1500  */
1501 static enum ice_status
1502 ice_sched_add_vsi_support_nodes(struct ice_port_info *pi, u16 vsi_handle,
1503 				struct ice_sched_node *tc_node, u16 *num_nodes)
1504 {
1505 	struct ice_sched_node *parent = tc_node;
1506 	enum ice_status status;
1507 	u32 first_node_teid;
1508 	u16 num_added = 0;
1509 	u8 i, vsil;
1510 
1511 	if (!pi)
1512 		return ICE_ERR_PARAM;
1513 
1514 	vsil = ice_sched_get_vsi_layer(pi->hw);
1515 	for (i = pi->hw->sw_entry_point_layer; i <= vsil; i++) {
1516 		status = ice_sched_add_nodes_to_layer(pi, tc_node, parent,
1517 						      i, num_nodes[i],
1518 						      &first_node_teid,
1519 						      &num_added);
1520 		if (status || num_nodes[i] != num_added)
1521 			return ICE_ERR_CFG;
1522 
1523 		/* The newly added node can be a new parent for the next
1524 		 * layer nodes
1525 		 */
1526 		if (num_added)
1527 			parent = ice_sched_find_node_by_teid(tc_node,
1528 							     first_node_teid);
1529 		else
1530 			parent = parent->children[0];
1531 
1532 		if (!parent)
1533 			return ICE_ERR_CFG;
1534 
1535 		if (i == vsil)
1536 			parent->vsi_handle = vsi_handle;
1537 	}
1538 
1539 	return 0;
1540 }
1541 
1542 /**
1543  * ice_sched_add_vsi_to_topo - add a new VSI into tree
1544  * @pi: port information structure
1545  * @vsi_handle: software VSI handle
1546  * @tc: TC number
1547  *
1548  * This function adds a new VSI into scheduler tree
1549  */
1550 static enum ice_status
1551 ice_sched_add_vsi_to_topo(struct ice_port_info *pi, u16 vsi_handle, u8 tc)
1552 {
1553 	u16 num_nodes[ICE_AQC_TOPO_MAX_LEVEL_NUM] = { 0 };
1554 	struct ice_sched_node *tc_node;
1555 	struct ice_hw *hw = pi->hw;
1556 
1557 	tc_node = ice_sched_get_tc_node(pi, tc);
1558 	if (!tc_node)
1559 		return ICE_ERR_PARAM;
1560 
1561 	/* calculate number of supported nodes needed for this VSI */
1562 	ice_sched_calc_vsi_support_nodes(hw, tc_node, num_nodes);
1563 
1564 	/* add VSI supported nodes to TC subtree */
1565 	return ice_sched_add_vsi_support_nodes(pi, vsi_handle, tc_node,
1566 					       num_nodes);
1567 }
1568 
1569 /**
1570  * ice_sched_update_vsi_child_nodes - update VSI child nodes
1571  * @pi: port information structure
1572  * @vsi_handle: software VSI handle
1573  * @tc: TC number
1574  * @new_numqs: new number of max queues
1575  * @owner: owner of this subtree
1576  *
1577  * This function updates the VSI child nodes based on the number of queues
1578  */
1579 static enum ice_status
1580 ice_sched_update_vsi_child_nodes(struct ice_port_info *pi, u16 vsi_handle,
1581 				 u8 tc, u16 new_numqs, u8 owner)
1582 {
1583 	u16 new_num_nodes[ICE_AQC_TOPO_MAX_LEVEL_NUM] = { 0 };
1584 	struct ice_sched_node *vsi_node;
1585 	struct ice_sched_node *tc_node;
1586 	struct ice_vsi_ctx *vsi_ctx;
1587 	enum ice_status status = 0;
1588 	struct ice_hw *hw = pi->hw;
1589 	u16 prev_numqs;
1590 
1591 	tc_node = ice_sched_get_tc_node(pi, tc);
1592 	if (!tc_node)
1593 		return ICE_ERR_CFG;
1594 
1595 	vsi_node = ice_sched_get_vsi_node(hw, tc_node, vsi_handle);
1596 	if (!vsi_node)
1597 		return ICE_ERR_CFG;
1598 
1599 	vsi_ctx = ice_get_vsi_ctx(hw, vsi_handle);
1600 	if (!vsi_ctx)
1601 		return ICE_ERR_PARAM;
1602 
1603 	prev_numqs = vsi_ctx->sched.max_lanq[tc];
1604 	/* num queues are not changed or less than the previous number */
1605 	if (new_numqs <= prev_numqs)
1606 		return status;
1607 	status = ice_alloc_lan_q_ctx(hw, vsi_handle, tc, new_numqs);
1608 	if (status)
1609 		return status;
1610 
1611 	if (new_numqs)
1612 		ice_sched_calc_vsi_child_nodes(hw, new_numqs, new_num_nodes);
1613 	/* Keep the max number of queue configuration all the time. Update the
1614 	 * tree only if number of queues > previous number of queues. This may
1615 	 * leave some extra nodes in the tree if number of queues < previous
1616 	 * number but that wouldn't harm anything. Removing those extra nodes
1617 	 * may complicate the code if those nodes are part of SRL or
1618 	 * individually rate limited.
1619 	 */
1620 	status = ice_sched_add_vsi_child_nodes(pi, vsi_handle, tc_node,
1621 					       new_num_nodes, owner);
1622 	if (status)
1623 		return status;
1624 	vsi_ctx->sched.max_lanq[tc] = new_numqs;
1625 
1626 	return 0;
1627 }
1628 
1629 /**
1630  * ice_sched_cfg_vsi - configure the new/existing VSI
1631  * @pi: port information structure
1632  * @vsi_handle: software VSI handle
1633  * @tc: TC number
1634  * @maxqs: max number of queues
1635  * @owner: LAN or RDMA
1636  * @enable: TC enabled or disabled
1637  *
1638  * This function adds/updates VSI nodes based on the number of queues. If TC is
1639  * enabled and VSI is in suspended state then resume the VSI back. If TC is
1640  * disabled then suspend the VSI if it is not already.
1641  */
1642 enum ice_status
1643 ice_sched_cfg_vsi(struct ice_port_info *pi, u16 vsi_handle, u8 tc, u16 maxqs,
1644 		  u8 owner, bool enable)
1645 {
1646 	struct ice_sched_node *vsi_node, *tc_node;
1647 	struct ice_vsi_ctx *vsi_ctx;
1648 	enum ice_status status = 0;
1649 	struct ice_hw *hw = pi->hw;
1650 
1651 	ice_debug(pi->hw, ICE_DBG_SCHED, "add/config VSI %d\n", vsi_handle);
1652 	tc_node = ice_sched_get_tc_node(pi, tc);
1653 	if (!tc_node)
1654 		return ICE_ERR_PARAM;
1655 	vsi_ctx = ice_get_vsi_ctx(hw, vsi_handle);
1656 	if (!vsi_ctx)
1657 		return ICE_ERR_PARAM;
1658 	vsi_node = ice_sched_get_vsi_node(hw, tc_node, vsi_handle);
1659 
1660 	/* suspend the VSI if TC is not enabled */
1661 	if (!enable) {
1662 		if (vsi_node && vsi_node->in_use) {
1663 			u32 teid = le32_to_cpu(vsi_node->info.node_teid);
1664 
1665 			status = ice_sched_suspend_resume_elems(hw, 1, &teid,
1666 								true);
1667 			if (!status)
1668 				vsi_node->in_use = false;
1669 		}
1670 		return status;
1671 	}
1672 
1673 	/* TC is enabled, if it is a new VSI then add it to the tree */
1674 	if (!vsi_node) {
1675 		status = ice_sched_add_vsi_to_topo(pi, vsi_handle, tc);
1676 		if (status)
1677 			return status;
1678 
1679 		vsi_node = ice_sched_get_vsi_node(hw, tc_node, vsi_handle);
1680 		if (!vsi_node)
1681 			return ICE_ERR_CFG;
1682 
1683 		vsi_ctx->sched.vsi_node[tc] = vsi_node;
1684 		vsi_node->in_use = true;
1685 		/* invalidate the max queues whenever VSI gets added first time
1686 		 * into the scheduler tree (boot or after reset). We need to
1687 		 * recreate the child nodes all the time in these cases.
1688 		 */
1689 		vsi_ctx->sched.max_lanq[tc] = 0;
1690 	}
1691 
1692 	/* update the VSI child nodes */
1693 	status = ice_sched_update_vsi_child_nodes(pi, vsi_handle, tc, maxqs,
1694 						  owner);
1695 	if (status)
1696 		return status;
1697 
1698 	/* TC is enabled, resume the VSI if it is in the suspend state */
1699 	if (!vsi_node->in_use) {
1700 		u32 teid = le32_to_cpu(vsi_node->info.node_teid);
1701 
1702 		status = ice_sched_suspend_resume_elems(hw, 1, &teid, false);
1703 		if (!status)
1704 			vsi_node->in_use = true;
1705 	}
1706 
1707 	return status;
1708 }
1709 
1710 /**
1711  * ice_sched_rm_agg_vsi_entry - remove aggregator related VSI info entry
1712  * @pi: port information structure
1713  * @vsi_handle: software VSI handle
1714  *
1715  * This function removes single aggregator VSI info entry from
1716  * aggregator list.
1717  */
1718 static void
1719 ice_sched_rm_agg_vsi_info(struct ice_port_info *pi, u16 vsi_handle)
1720 {
1721 	struct ice_sched_agg_info *agg_info;
1722 	struct ice_sched_agg_info *atmp;
1723 
1724 	list_for_each_entry_safe(agg_info, atmp, &pi->hw->agg_list,
1725 				 list_entry) {
1726 		struct ice_sched_agg_vsi_info *agg_vsi_info;
1727 		struct ice_sched_agg_vsi_info *vtmp;
1728 
1729 		list_for_each_entry_safe(agg_vsi_info, vtmp,
1730 					 &agg_info->agg_vsi_list, list_entry)
1731 			if (agg_vsi_info->vsi_handle == vsi_handle) {
1732 				list_del(&agg_vsi_info->list_entry);
1733 				devm_kfree(ice_hw_to_dev(pi->hw),
1734 					   agg_vsi_info);
1735 				return;
1736 			}
1737 	}
1738 }
1739 
1740 /**
1741  * ice_sched_is_leaf_node_present - check for a leaf node in the sub-tree
1742  * @node: pointer to the sub-tree node
1743  *
1744  * This function checks for a leaf node presence in a given sub-tree node.
1745  */
1746 static bool ice_sched_is_leaf_node_present(struct ice_sched_node *node)
1747 {
1748 	u8 i;
1749 
1750 	for (i = 0; i < node->num_children; i++)
1751 		if (ice_sched_is_leaf_node_present(node->children[i]))
1752 			return true;
1753 	/* check for a leaf node */
1754 	return (node->info.data.elem_type == ICE_AQC_ELEM_TYPE_LEAF);
1755 }
1756 
1757 /**
1758  * ice_sched_rm_vsi_cfg - remove the VSI and its children nodes
1759  * @pi: port information structure
1760  * @vsi_handle: software VSI handle
1761  * @owner: LAN or RDMA
1762  *
1763  * This function removes the VSI and its LAN or RDMA children nodes from the
1764  * scheduler tree.
1765  */
1766 static enum ice_status
1767 ice_sched_rm_vsi_cfg(struct ice_port_info *pi, u16 vsi_handle, u8 owner)
1768 {
1769 	enum ice_status status = ICE_ERR_PARAM;
1770 	struct ice_vsi_ctx *vsi_ctx;
1771 	u8 i;
1772 
1773 	ice_debug(pi->hw, ICE_DBG_SCHED, "removing VSI %d\n", vsi_handle);
1774 	if (!ice_is_vsi_valid(pi->hw, vsi_handle))
1775 		return status;
1776 	mutex_lock(&pi->sched_lock);
1777 	vsi_ctx = ice_get_vsi_ctx(pi->hw, vsi_handle);
1778 	if (!vsi_ctx)
1779 		goto exit_sched_rm_vsi_cfg;
1780 
1781 	ice_for_each_traffic_class(i) {
1782 		struct ice_sched_node *vsi_node, *tc_node;
1783 		u8 j = 0;
1784 
1785 		tc_node = ice_sched_get_tc_node(pi, i);
1786 		if (!tc_node)
1787 			continue;
1788 
1789 		vsi_node = ice_sched_get_vsi_node(pi->hw, tc_node, vsi_handle);
1790 		if (!vsi_node)
1791 			continue;
1792 
1793 		if (ice_sched_is_leaf_node_present(vsi_node)) {
1794 			ice_debug(pi->hw, ICE_DBG_SCHED,
1795 				  "VSI has leaf nodes in TC %d\n", i);
1796 			status = ICE_ERR_IN_USE;
1797 			goto exit_sched_rm_vsi_cfg;
1798 		}
1799 		while (j < vsi_node->num_children) {
1800 			if (vsi_node->children[j]->owner == owner) {
1801 				ice_free_sched_node(pi, vsi_node->children[j]);
1802 
1803 				/* reset the counter again since the num
1804 				 * children will be updated after node removal
1805 				 */
1806 				j = 0;
1807 			} else {
1808 				j++;
1809 			}
1810 		}
1811 		/* remove the VSI if it has no children */
1812 		if (!vsi_node->num_children) {
1813 			ice_free_sched_node(pi, vsi_node);
1814 			vsi_ctx->sched.vsi_node[i] = NULL;
1815 
1816 			/* clean up aggregator related VSI info if any */
1817 			ice_sched_rm_agg_vsi_info(pi, vsi_handle);
1818 		}
1819 		if (owner == ICE_SCHED_NODE_OWNER_LAN)
1820 			vsi_ctx->sched.max_lanq[i] = 0;
1821 	}
1822 	status = 0;
1823 
1824 exit_sched_rm_vsi_cfg:
1825 	mutex_unlock(&pi->sched_lock);
1826 	return status;
1827 }
1828 
1829 /**
1830  * ice_rm_vsi_lan_cfg - remove VSI and its LAN children nodes
1831  * @pi: port information structure
1832  * @vsi_handle: software VSI handle
1833  *
1834  * This function clears the VSI and its LAN children nodes from scheduler tree
1835  * for all TCs.
1836  */
1837 enum ice_status ice_rm_vsi_lan_cfg(struct ice_port_info *pi, u16 vsi_handle)
1838 {
1839 	return ice_sched_rm_vsi_cfg(pi, vsi_handle, ICE_SCHED_NODE_OWNER_LAN);
1840 }
1841 
1842 /**
1843  * ice_sched_rm_unused_rl_prof - remove unused RL profile
1844  * @pi: port information structure
1845  *
1846  * This function removes unused rate limit profiles from the HW and
1847  * SW DB. The caller needs to hold scheduler lock.
1848  */
1849 static void ice_sched_rm_unused_rl_prof(struct ice_port_info *pi)
1850 {
1851 	u16 ln;
1852 
1853 	for (ln = 0; ln < pi->hw->num_tx_sched_layers; ln++) {
1854 		struct ice_aqc_rl_profile_info *rl_prof_elem;
1855 		struct ice_aqc_rl_profile_info *rl_prof_tmp;
1856 
1857 		list_for_each_entry_safe(rl_prof_elem, rl_prof_tmp,
1858 					 &pi->rl_prof_list[ln], list_entry) {
1859 			if (!ice_sched_del_rl_profile(pi->hw, rl_prof_elem))
1860 				ice_debug(pi->hw, ICE_DBG_SCHED,
1861 					  "Removed rl profile\n");
1862 		}
1863 	}
1864 }
1865 
1866 /**
1867  * ice_sched_update_elem - update element
1868  * @hw: pointer to the HW struct
1869  * @node: pointer to node
1870  * @info: node info to update
1871  *
1872  * It updates the HW DB, and local SW DB of node. It updates the scheduling
1873  * parameters of node from argument info data buffer (Info->data buf) and
1874  * returns success or error on config sched element failure. The caller
1875  * needs to hold scheduler lock.
1876  */
1877 static enum ice_status
1878 ice_sched_update_elem(struct ice_hw *hw, struct ice_sched_node *node,
1879 		      struct ice_aqc_txsched_elem_data *info)
1880 {
1881 	struct ice_aqc_conf_elem buf;
1882 	enum ice_status status;
1883 	u16 elem_cfgd = 0;
1884 	u16 num_elems = 1;
1885 
1886 	buf.generic[0] = *info;
1887 	/* Parent TEID is reserved field in this aq call */
1888 	buf.generic[0].parent_teid = 0;
1889 	/* Element type is reserved field in this aq call */
1890 	buf.generic[0].data.elem_type = 0;
1891 	/* Flags is reserved field in this aq call */
1892 	buf.generic[0].data.flags = 0;
1893 
1894 	/* Update HW DB */
1895 	/* Configure element node */
1896 	status = ice_aq_cfg_sched_elems(hw, num_elems, &buf, sizeof(buf),
1897 					&elem_cfgd, NULL);
1898 	if (status || elem_cfgd != num_elems) {
1899 		ice_debug(hw, ICE_DBG_SCHED, "Config sched elem error\n");
1900 		return ICE_ERR_CFG;
1901 	}
1902 
1903 	/* Config success case */
1904 	/* Now update local SW DB */
1905 	/* Only copy the data portion of info buffer */
1906 	node->info.data = info->data;
1907 	return status;
1908 }
1909 
1910 /**
1911  * ice_sched_cfg_node_bw_alloc - configure node BW weight/alloc params
1912  * @hw: pointer to the HW struct
1913  * @node: sched node to configure
1914  * @rl_type: rate limit type CIR, EIR, or shared
1915  * @bw_alloc: BW weight/allocation
1916  *
1917  * This function configures node element's BW allocation.
1918  */
1919 static enum ice_status
1920 ice_sched_cfg_node_bw_alloc(struct ice_hw *hw, struct ice_sched_node *node,
1921 			    enum ice_rl_type rl_type, u8 bw_alloc)
1922 {
1923 	struct ice_aqc_txsched_elem_data buf;
1924 	struct ice_aqc_txsched_elem *data;
1925 	enum ice_status status;
1926 
1927 	buf = node->info;
1928 	data = &buf.data;
1929 	if (rl_type == ICE_MIN_BW) {
1930 		data->valid_sections |= ICE_AQC_ELEM_VALID_CIR;
1931 		data->cir_bw.bw_alloc = cpu_to_le16(bw_alloc);
1932 	} else if (rl_type == ICE_MAX_BW) {
1933 		data->valid_sections |= ICE_AQC_ELEM_VALID_EIR;
1934 		data->eir_bw.bw_alloc = cpu_to_le16(bw_alloc);
1935 	} else {
1936 		return ICE_ERR_PARAM;
1937 	}
1938 
1939 	/* Configure element */
1940 	status = ice_sched_update_elem(hw, node, &buf);
1941 	return status;
1942 }
1943 
1944 /**
1945  * ice_set_clear_cir_bw - set or clear CIR BW
1946  * @bw_t_info: bandwidth type information structure
1947  * @bw: bandwidth in Kbps - Kilo bits per sec
1948  *
1949  * Save or clear CIR bandwidth (BW) in the passed param bw_t_info.
1950  */
1951 static void
1952 ice_set_clear_cir_bw(struct ice_bw_type_info *bw_t_info, u32 bw)
1953 {
1954 	if (bw == ICE_SCHED_DFLT_BW) {
1955 		clear_bit(ICE_BW_TYPE_CIR, bw_t_info->bw_t_bitmap);
1956 		bw_t_info->cir_bw.bw = 0;
1957 	} else {
1958 		/* Save type of BW information */
1959 		set_bit(ICE_BW_TYPE_CIR, bw_t_info->bw_t_bitmap);
1960 		bw_t_info->cir_bw.bw = bw;
1961 	}
1962 }
1963 
1964 /**
1965  * ice_set_clear_eir_bw - set or clear EIR BW
1966  * @bw_t_info: bandwidth type information structure
1967  * @bw: bandwidth in Kbps - Kilo bits per sec
1968  *
1969  * Save or clear EIR bandwidth (BW) in the passed param bw_t_info.
1970  */
1971 static void
1972 ice_set_clear_eir_bw(struct ice_bw_type_info *bw_t_info, u32 bw)
1973 {
1974 	if (bw == ICE_SCHED_DFLT_BW) {
1975 		clear_bit(ICE_BW_TYPE_EIR, bw_t_info->bw_t_bitmap);
1976 		bw_t_info->eir_bw.bw = 0;
1977 	} else {
1978 		/* EIR BW and Shared BW profiles are mutually exclusive and
1979 		 * hence only one of them may be set for any given element.
1980 		 * First clear earlier saved shared BW information.
1981 		 */
1982 		clear_bit(ICE_BW_TYPE_SHARED, bw_t_info->bw_t_bitmap);
1983 		bw_t_info->shared_bw = 0;
1984 		/* save EIR BW information */
1985 		set_bit(ICE_BW_TYPE_EIR, bw_t_info->bw_t_bitmap);
1986 		bw_t_info->eir_bw.bw = bw;
1987 	}
1988 }
1989 
1990 /**
1991  * ice_set_clear_shared_bw - set or clear shared BW
1992  * @bw_t_info: bandwidth type information structure
1993  * @bw: bandwidth in Kbps - Kilo bits per sec
1994  *
1995  * Save or clear shared bandwidth (BW) in the passed param bw_t_info.
1996  */
1997 static void
1998 ice_set_clear_shared_bw(struct ice_bw_type_info *bw_t_info, u32 bw)
1999 {
2000 	if (bw == ICE_SCHED_DFLT_BW) {
2001 		clear_bit(ICE_BW_TYPE_SHARED, bw_t_info->bw_t_bitmap);
2002 		bw_t_info->shared_bw = 0;
2003 	} else {
2004 		/* EIR BW and Shared BW profiles are mutually exclusive and
2005 		 * hence only one of them may be set for any given element.
2006 		 * First clear earlier saved EIR BW information.
2007 		 */
2008 		clear_bit(ICE_BW_TYPE_EIR, bw_t_info->bw_t_bitmap);
2009 		bw_t_info->eir_bw.bw = 0;
2010 		/* save shared BW information */
2011 		set_bit(ICE_BW_TYPE_SHARED, bw_t_info->bw_t_bitmap);
2012 		bw_t_info->shared_bw = bw;
2013 	}
2014 }
2015 
2016 /**
2017  * ice_sched_calc_wakeup - calculate RL profile wakeup parameter
2018  * @bw: bandwidth in Kbps
2019  *
2020  * This function calculates the wakeup parameter of RL profile.
2021  */
2022 static u16 ice_sched_calc_wakeup(s32 bw)
2023 {
2024 	s64 bytes_per_sec, wakeup_int, wakeup_a, wakeup_b, wakeup_f;
2025 	s32 wakeup_f_int;
2026 	u16 wakeup = 0;
2027 
2028 	/* Get the wakeup integer value */
2029 	bytes_per_sec = div64_long(((s64)bw * 1000), BITS_PER_BYTE);
2030 	wakeup_int = div64_long(ICE_RL_PROF_FREQUENCY, bytes_per_sec);
2031 	if (wakeup_int > 63) {
2032 		wakeup = (u16)((1 << 15) | wakeup_int);
2033 	} else {
2034 		/* Calculate fraction value up to 4 decimals
2035 		 * Convert Integer value to a constant multiplier
2036 		 */
2037 		wakeup_b = (s64)ICE_RL_PROF_MULTIPLIER * wakeup_int;
2038 		wakeup_a = div64_long((s64)ICE_RL_PROF_MULTIPLIER *
2039 					   ICE_RL_PROF_FREQUENCY,
2040 				      bytes_per_sec);
2041 
2042 		/* Get Fraction value */
2043 		wakeup_f = wakeup_a - wakeup_b;
2044 
2045 		/* Round up the Fractional value via Ceil(Fractional value) */
2046 		if (wakeup_f > div64_long(ICE_RL_PROF_MULTIPLIER, 2))
2047 			wakeup_f += 1;
2048 
2049 		wakeup_f_int = (s32)div64_long(wakeup_f * ICE_RL_PROF_FRACTION,
2050 					       ICE_RL_PROF_MULTIPLIER);
2051 		wakeup |= (u16)(wakeup_int << 9);
2052 		wakeup |= (u16)(0x1ff & wakeup_f_int);
2053 	}
2054 
2055 	return wakeup;
2056 }
2057 
2058 /**
2059  * ice_sched_bw_to_rl_profile - convert BW to profile parameters
2060  * @bw: bandwidth in Kbps
2061  * @profile: profile parameters to return
2062  *
2063  * This function converts the BW to profile structure format.
2064  */
2065 static enum ice_status
2066 ice_sched_bw_to_rl_profile(u32 bw, struct ice_aqc_rl_profile_elem *profile)
2067 {
2068 	enum ice_status status = ICE_ERR_PARAM;
2069 	s64 bytes_per_sec, ts_rate, mv_tmp;
2070 	bool found = false;
2071 	s32 encode = 0;
2072 	s64 mv = 0;
2073 	s32 i;
2074 
2075 	/* Bw settings range is from 0.5Mb/sec to 100Gb/sec */
2076 	if (bw < ICE_SCHED_MIN_BW || bw > ICE_SCHED_MAX_BW)
2077 		return status;
2078 
2079 	/* Bytes per second from Kbps */
2080 	bytes_per_sec = div64_long(((s64)bw * 1000), BITS_PER_BYTE);
2081 
2082 	/* encode is 6 bits but really useful are 5 bits */
2083 	for (i = 0; i < 64; i++) {
2084 		u64 pow_result = BIT_ULL(i);
2085 
2086 		ts_rate = div64_long((s64)ICE_RL_PROF_FREQUENCY,
2087 				     pow_result * ICE_RL_PROF_TS_MULTIPLIER);
2088 		if (ts_rate <= 0)
2089 			continue;
2090 
2091 		/* Multiplier value */
2092 		mv_tmp = div64_long(bytes_per_sec * ICE_RL_PROF_MULTIPLIER,
2093 				    ts_rate);
2094 
2095 		/* Round to the nearest ICE_RL_PROF_MULTIPLIER */
2096 		mv = round_up_64bit(mv_tmp, ICE_RL_PROF_MULTIPLIER);
2097 
2098 		/* First multiplier value greater than the given
2099 		 * accuracy bytes
2100 		 */
2101 		if (mv > ICE_RL_PROF_ACCURACY_BYTES) {
2102 			encode = i;
2103 			found = true;
2104 			break;
2105 		}
2106 	}
2107 	if (found) {
2108 		u16 wm;
2109 
2110 		wm = ice_sched_calc_wakeup(bw);
2111 		profile->rl_multiply = cpu_to_le16(mv);
2112 		profile->wake_up_calc = cpu_to_le16(wm);
2113 		profile->rl_encode = cpu_to_le16(encode);
2114 		status = 0;
2115 	} else {
2116 		status = ICE_ERR_DOES_NOT_EXIST;
2117 	}
2118 
2119 	return status;
2120 }
2121 
2122 /**
2123  * ice_sched_add_rl_profile - add RL profile
2124  * @pi: port information structure
2125  * @rl_type: type of rate limit BW - min, max, or shared
2126  * @bw: bandwidth in Kbps - Kilo bits per sec
2127  * @layer_num: specifies in which layer to create profile
2128  *
2129  * This function first checks the existing list for corresponding BW
2130  * parameter. If it exists, it returns the associated profile otherwise
2131  * it creates a new rate limit profile for requested BW, and adds it to
2132  * the HW DB and local list. It returns the new profile or null on error.
2133  * The caller needs to hold the scheduler lock.
2134  */
2135 static struct ice_aqc_rl_profile_info *
2136 ice_sched_add_rl_profile(struct ice_port_info *pi,
2137 			 enum ice_rl_type rl_type, u32 bw, u8 layer_num)
2138 {
2139 	struct ice_aqc_rl_profile_generic_elem *buf;
2140 	struct ice_aqc_rl_profile_info *rl_prof_elem;
2141 	u16 profiles_added = 0, num_profiles = 1;
2142 	enum ice_status status;
2143 	struct ice_hw *hw;
2144 	u8 profile_type;
2145 
2146 	if (layer_num >= ICE_AQC_TOPO_MAX_LEVEL_NUM)
2147 		return NULL;
2148 	switch (rl_type) {
2149 	case ICE_MIN_BW:
2150 		profile_type = ICE_AQC_RL_PROFILE_TYPE_CIR;
2151 		break;
2152 	case ICE_MAX_BW:
2153 		profile_type = ICE_AQC_RL_PROFILE_TYPE_EIR;
2154 		break;
2155 	case ICE_SHARED_BW:
2156 		profile_type = ICE_AQC_RL_PROFILE_TYPE_SRL;
2157 		break;
2158 	default:
2159 		return NULL;
2160 	}
2161 
2162 	if (!pi)
2163 		return NULL;
2164 	hw = pi->hw;
2165 	list_for_each_entry(rl_prof_elem, &pi->rl_prof_list[layer_num],
2166 			    list_entry)
2167 		if (rl_prof_elem->profile.flags == profile_type &&
2168 		    rl_prof_elem->bw == bw)
2169 			/* Return existing profile ID info */
2170 			return rl_prof_elem;
2171 
2172 	/* Create new profile ID */
2173 	rl_prof_elem = devm_kzalloc(ice_hw_to_dev(hw), sizeof(*rl_prof_elem),
2174 				    GFP_KERNEL);
2175 
2176 	if (!rl_prof_elem)
2177 		return NULL;
2178 
2179 	status = ice_sched_bw_to_rl_profile(bw, &rl_prof_elem->profile);
2180 	if (status)
2181 		goto exit_add_rl_prof;
2182 
2183 	rl_prof_elem->bw = bw;
2184 	/* layer_num is zero relative, and fw expects level from 1 to 9 */
2185 	rl_prof_elem->profile.level = layer_num + 1;
2186 	rl_prof_elem->profile.flags = profile_type;
2187 	rl_prof_elem->profile.max_burst_size = cpu_to_le16(hw->max_burst_size);
2188 
2189 	/* Create new entry in HW DB */
2190 	buf = (struct ice_aqc_rl_profile_generic_elem *)
2191 		&rl_prof_elem->profile;
2192 	status = ice_aq_add_rl_profile(hw, num_profiles, buf, sizeof(*buf),
2193 				       &profiles_added, NULL);
2194 	if (status || profiles_added != num_profiles)
2195 		goto exit_add_rl_prof;
2196 
2197 	/* Good entry - add in the list */
2198 	rl_prof_elem->prof_id_ref = 0;
2199 	list_add(&rl_prof_elem->list_entry, &pi->rl_prof_list[layer_num]);
2200 	return rl_prof_elem;
2201 
2202 exit_add_rl_prof:
2203 	devm_kfree(ice_hw_to_dev(hw), rl_prof_elem);
2204 	return NULL;
2205 }
2206 
2207 /**
2208  * ice_sched_cfg_node_bw_lmt - configure node sched params
2209  * @hw: pointer to the HW struct
2210  * @node: sched node to configure
2211  * @rl_type: rate limit type CIR, EIR, or shared
2212  * @rl_prof_id: rate limit profile ID
2213  *
2214  * This function configures node element's BW limit.
2215  */
2216 static enum ice_status
2217 ice_sched_cfg_node_bw_lmt(struct ice_hw *hw, struct ice_sched_node *node,
2218 			  enum ice_rl_type rl_type, u16 rl_prof_id)
2219 {
2220 	struct ice_aqc_txsched_elem_data buf;
2221 	struct ice_aqc_txsched_elem *data;
2222 
2223 	buf = node->info;
2224 	data = &buf.data;
2225 	switch (rl_type) {
2226 	case ICE_MIN_BW:
2227 		data->valid_sections |= ICE_AQC_ELEM_VALID_CIR;
2228 		data->cir_bw.bw_profile_idx = cpu_to_le16(rl_prof_id);
2229 		break;
2230 	case ICE_MAX_BW:
2231 		/* EIR BW and Shared BW profiles are mutually exclusive and
2232 		 * hence only one of them may be set for any given element
2233 		 */
2234 		if (data->valid_sections & ICE_AQC_ELEM_VALID_SHARED)
2235 			return ICE_ERR_CFG;
2236 		data->valid_sections |= ICE_AQC_ELEM_VALID_EIR;
2237 		data->eir_bw.bw_profile_idx = cpu_to_le16(rl_prof_id);
2238 		break;
2239 	case ICE_SHARED_BW:
2240 		/* Check for removing shared BW */
2241 		if (rl_prof_id == ICE_SCHED_NO_SHARED_RL_PROF_ID) {
2242 			/* remove shared profile */
2243 			data->valid_sections &= ~ICE_AQC_ELEM_VALID_SHARED;
2244 			data->srl_id = 0; /* clear SRL field */
2245 
2246 			/* enable back EIR to default profile */
2247 			data->valid_sections |= ICE_AQC_ELEM_VALID_EIR;
2248 			data->eir_bw.bw_profile_idx =
2249 				cpu_to_le16(ICE_SCHED_DFLT_RL_PROF_ID);
2250 			break;
2251 		}
2252 		/* EIR BW and Shared BW profiles are mutually exclusive and
2253 		 * hence only one of them may be set for any given element
2254 		 */
2255 		if ((data->valid_sections & ICE_AQC_ELEM_VALID_EIR) &&
2256 		    (le16_to_cpu(data->eir_bw.bw_profile_idx) !=
2257 			    ICE_SCHED_DFLT_RL_PROF_ID))
2258 			return ICE_ERR_CFG;
2259 		/* EIR BW is set to default, disable it */
2260 		data->valid_sections &= ~ICE_AQC_ELEM_VALID_EIR;
2261 		/* Okay to enable shared BW now */
2262 		data->valid_sections |= ICE_AQC_ELEM_VALID_SHARED;
2263 		data->srl_id = cpu_to_le16(rl_prof_id);
2264 		break;
2265 	default:
2266 		/* Unknown rate limit type */
2267 		return ICE_ERR_PARAM;
2268 	}
2269 
2270 	/* Configure element */
2271 	return ice_sched_update_elem(hw, node, &buf);
2272 }
2273 
2274 /**
2275  * ice_sched_get_node_rl_prof_id - get node's rate limit profile ID
2276  * @node: sched node
2277  * @rl_type: rate limit type
2278  *
2279  * If existing profile matches, it returns the corresponding rate
2280  * limit profile ID, otherwise it returns an invalid ID as error.
2281  */
2282 static u16
2283 ice_sched_get_node_rl_prof_id(struct ice_sched_node *node,
2284 			      enum ice_rl_type rl_type)
2285 {
2286 	u16 rl_prof_id = ICE_SCHED_INVAL_PROF_ID;
2287 	struct ice_aqc_txsched_elem *data;
2288 
2289 	data = &node->info.data;
2290 	switch (rl_type) {
2291 	case ICE_MIN_BW:
2292 		if (data->valid_sections & ICE_AQC_ELEM_VALID_CIR)
2293 			rl_prof_id = le16_to_cpu(data->cir_bw.bw_profile_idx);
2294 		break;
2295 	case ICE_MAX_BW:
2296 		if (data->valid_sections & ICE_AQC_ELEM_VALID_EIR)
2297 			rl_prof_id = le16_to_cpu(data->eir_bw.bw_profile_idx);
2298 		break;
2299 	case ICE_SHARED_BW:
2300 		if (data->valid_sections & ICE_AQC_ELEM_VALID_SHARED)
2301 			rl_prof_id = le16_to_cpu(data->srl_id);
2302 		break;
2303 	default:
2304 		break;
2305 	}
2306 
2307 	return rl_prof_id;
2308 }
2309 
2310 /**
2311  * ice_sched_get_rl_prof_layer - selects rate limit profile creation layer
2312  * @pi: port information structure
2313  * @rl_type: type of rate limit BW - min, max, or shared
2314  * @layer_index: layer index
2315  *
2316  * This function returns requested profile creation layer.
2317  */
2318 static u8
2319 ice_sched_get_rl_prof_layer(struct ice_port_info *pi, enum ice_rl_type rl_type,
2320 			    u8 layer_index)
2321 {
2322 	struct ice_hw *hw = pi->hw;
2323 
2324 	if (layer_index >= hw->num_tx_sched_layers)
2325 		return ICE_SCHED_INVAL_LAYER_NUM;
2326 	switch (rl_type) {
2327 	case ICE_MIN_BW:
2328 		if (hw->layer_info[layer_index].max_cir_rl_profiles)
2329 			return layer_index;
2330 		break;
2331 	case ICE_MAX_BW:
2332 		if (hw->layer_info[layer_index].max_eir_rl_profiles)
2333 			return layer_index;
2334 		break;
2335 	case ICE_SHARED_BW:
2336 		/* if current layer doesn't support SRL profile creation
2337 		 * then try a layer up or down.
2338 		 */
2339 		if (hw->layer_info[layer_index].max_srl_profiles)
2340 			return layer_index;
2341 		else if (layer_index < hw->num_tx_sched_layers - 1 &&
2342 			 hw->layer_info[layer_index + 1].max_srl_profiles)
2343 			return layer_index + 1;
2344 		else if (layer_index > 0 &&
2345 			 hw->layer_info[layer_index - 1].max_srl_profiles)
2346 			return layer_index - 1;
2347 		break;
2348 	default:
2349 		break;
2350 	}
2351 	return ICE_SCHED_INVAL_LAYER_NUM;
2352 }
2353 
2354 /**
2355  * ice_sched_get_srl_node - get shared rate limit node
2356  * @node: tree node
2357  * @srl_layer: shared rate limit layer
2358  *
2359  * This function returns SRL node to be used for shared rate limit purpose.
2360  * The caller needs to hold scheduler lock.
2361  */
2362 static struct ice_sched_node *
2363 ice_sched_get_srl_node(struct ice_sched_node *node, u8 srl_layer)
2364 {
2365 	if (srl_layer > node->tx_sched_layer)
2366 		return node->children[0];
2367 	else if (srl_layer < node->tx_sched_layer)
2368 		/* Node can't be created without a parent. It will always
2369 		 * have a valid parent except root node.
2370 		 */
2371 		return node->parent;
2372 	else
2373 		return node;
2374 }
2375 
2376 /**
2377  * ice_sched_rm_rl_profile - remove RL profile ID
2378  * @pi: port information structure
2379  * @layer_num: layer number where profiles are saved
2380  * @profile_type: profile type like EIR, CIR, or SRL
2381  * @profile_id: profile ID to remove
2382  *
2383  * This function removes rate limit profile from layer 'layer_num' of type
2384  * 'profile_type' and profile ID as 'profile_id'. The caller needs to hold
2385  * scheduler lock.
2386  */
2387 static enum ice_status
2388 ice_sched_rm_rl_profile(struct ice_port_info *pi, u8 layer_num, u8 profile_type,
2389 			u16 profile_id)
2390 {
2391 	struct ice_aqc_rl_profile_info *rl_prof_elem;
2392 	enum ice_status status = 0;
2393 
2394 	if (layer_num >= ICE_AQC_TOPO_MAX_LEVEL_NUM)
2395 		return ICE_ERR_PARAM;
2396 	/* Check the existing list for RL profile */
2397 	list_for_each_entry(rl_prof_elem, &pi->rl_prof_list[layer_num],
2398 			    list_entry)
2399 		if (rl_prof_elem->profile.flags == profile_type &&
2400 		    le16_to_cpu(rl_prof_elem->profile.profile_id) ==
2401 		    profile_id) {
2402 			if (rl_prof_elem->prof_id_ref)
2403 				rl_prof_elem->prof_id_ref--;
2404 
2405 			/* Remove old profile ID from database */
2406 			status = ice_sched_del_rl_profile(pi->hw, rl_prof_elem);
2407 			if (status && status != ICE_ERR_IN_USE)
2408 				ice_debug(pi->hw, ICE_DBG_SCHED,
2409 					  "Remove rl profile failed\n");
2410 			break;
2411 		}
2412 	if (status == ICE_ERR_IN_USE)
2413 		status = 0;
2414 	return status;
2415 }
2416 
2417 /**
2418  * ice_sched_set_node_bw_dflt - set node's bandwidth limit to default
2419  * @pi: port information structure
2420  * @node: pointer to node structure
2421  * @rl_type: rate limit type min, max, or shared
2422  * @layer_num: layer number where RL profiles are saved
2423  *
2424  * This function configures node element's BW rate limit profile ID of
2425  * type CIR, EIR, or SRL to default. This function needs to be called
2426  * with the scheduler lock held.
2427  */
2428 static enum ice_status
2429 ice_sched_set_node_bw_dflt(struct ice_port_info *pi,
2430 			   struct ice_sched_node *node,
2431 			   enum ice_rl_type rl_type, u8 layer_num)
2432 {
2433 	enum ice_status status;
2434 	struct ice_hw *hw;
2435 	u8 profile_type;
2436 	u16 rl_prof_id;
2437 	u16 old_id;
2438 
2439 	hw = pi->hw;
2440 	switch (rl_type) {
2441 	case ICE_MIN_BW:
2442 		profile_type = ICE_AQC_RL_PROFILE_TYPE_CIR;
2443 		rl_prof_id = ICE_SCHED_DFLT_RL_PROF_ID;
2444 		break;
2445 	case ICE_MAX_BW:
2446 		profile_type = ICE_AQC_RL_PROFILE_TYPE_EIR;
2447 		rl_prof_id = ICE_SCHED_DFLT_RL_PROF_ID;
2448 		break;
2449 	case ICE_SHARED_BW:
2450 		profile_type = ICE_AQC_RL_PROFILE_TYPE_SRL;
2451 		/* No SRL is configured for default case */
2452 		rl_prof_id = ICE_SCHED_NO_SHARED_RL_PROF_ID;
2453 		break;
2454 	default:
2455 		return ICE_ERR_PARAM;
2456 	}
2457 	/* Save existing RL prof ID for later clean up */
2458 	old_id = ice_sched_get_node_rl_prof_id(node, rl_type);
2459 	/* Configure BW scheduling parameters */
2460 	status = ice_sched_cfg_node_bw_lmt(hw, node, rl_type, rl_prof_id);
2461 	if (status)
2462 		return status;
2463 
2464 	/* Remove stale RL profile ID */
2465 	if (old_id == ICE_SCHED_DFLT_RL_PROF_ID ||
2466 	    old_id == ICE_SCHED_INVAL_PROF_ID)
2467 		return 0;
2468 
2469 	return ice_sched_rm_rl_profile(pi, layer_num, profile_type, old_id);
2470 }
2471 
2472 /**
2473  * ice_sched_set_eir_srl_excl - set EIR/SRL exclusiveness
2474  * @pi: port information structure
2475  * @node: pointer to node structure
2476  * @layer_num: layer number where rate limit profiles are saved
2477  * @rl_type: rate limit type min, max, or shared
2478  * @bw: bandwidth value
2479  *
2480  * This function prepares node element's bandwidth to SRL or EIR exclusively.
2481  * EIR BW and Shared BW profiles are mutually exclusive and hence only one of
2482  * them may be set for any given element. This function needs to be called
2483  * with the scheduler lock held.
2484  */
2485 static enum ice_status
2486 ice_sched_set_eir_srl_excl(struct ice_port_info *pi,
2487 			   struct ice_sched_node *node,
2488 			   u8 layer_num, enum ice_rl_type rl_type, u32 bw)
2489 {
2490 	if (rl_type == ICE_SHARED_BW) {
2491 		/* SRL node passed in this case, it may be different node */
2492 		if (bw == ICE_SCHED_DFLT_BW)
2493 			/* SRL being removed, ice_sched_cfg_node_bw_lmt()
2494 			 * enables EIR to default. EIR is not set in this
2495 			 * case, so no additional action is required.
2496 			 */
2497 			return 0;
2498 
2499 		/* SRL being configured, set EIR to default here.
2500 		 * ice_sched_cfg_node_bw_lmt() disables EIR when it
2501 		 * configures SRL
2502 		 */
2503 		return ice_sched_set_node_bw_dflt(pi, node, ICE_MAX_BW,
2504 						  layer_num);
2505 	} else if (rl_type == ICE_MAX_BW &&
2506 		   node->info.data.valid_sections & ICE_AQC_ELEM_VALID_SHARED) {
2507 		/* Remove Shared profile. Set default shared BW call
2508 		 * removes shared profile for a node.
2509 		 */
2510 		return ice_sched_set_node_bw_dflt(pi, node,
2511 						  ICE_SHARED_BW,
2512 						  layer_num);
2513 	}
2514 	return 0;
2515 }
2516 
2517 /**
2518  * ice_sched_set_node_bw - set node's bandwidth
2519  * @pi: port information structure
2520  * @node: tree node
2521  * @rl_type: rate limit type min, max, or shared
2522  * @bw: bandwidth in Kbps - Kilo bits per sec
2523  * @layer_num: layer number
2524  *
2525  * This function adds new profile corresponding to requested BW, configures
2526  * node's RL profile ID of type CIR, EIR, or SRL, and removes old profile
2527  * ID from local database. The caller needs to hold scheduler lock.
2528  */
2529 static enum ice_status
2530 ice_sched_set_node_bw(struct ice_port_info *pi, struct ice_sched_node *node,
2531 		      enum ice_rl_type rl_type, u32 bw, u8 layer_num)
2532 {
2533 	struct ice_aqc_rl_profile_info *rl_prof_info;
2534 	enum ice_status status = ICE_ERR_PARAM;
2535 	struct ice_hw *hw = pi->hw;
2536 	u16 old_id, rl_prof_id;
2537 
2538 	rl_prof_info = ice_sched_add_rl_profile(pi, rl_type, bw, layer_num);
2539 	if (!rl_prof_info)
2540 		return status;
2541 
2542 	rl_prof_id = le16_to_cpu(rl_prof_info->profile.profile_id);
2543 
2544 	/* Save existing RL prof ID for later clean up */
2545 	old_id = ice_sched_get_node_rl_prof_id(node, rl_type);
2546 	/* Configure BW scheduling parameters */
2547 	status = ice_sched_cfg_node_bw_lmt(hw, node, rl_type, rl_prof_id);
2548 	if (status)
2549 		return status;
2550 
2551 	/* New changes has been applied */
2552 	/* Increment the profile ID reference count */
2553 	rl_prof_info->prof_id_ref++;
2554 
2555 	/* Check for old ID removal */
2556 	if ((old_id == ICE_SCHED_DFLT_RL_PROF_ID && rl_type != ICE_SHARED_BW) ||
2557 	    old_id == ICE_SCHED_INVAL_PROF_ID || old_id == rl_prof_id)
2558 		return 0;
2559 
2560 	return ice_sched_rm_rl_profile(pi, layer_num,
2561 				       rl_prof_info->profile.flags,
2562 				       old_id);
2563 }
2564 
2565 /**
2566  * ice_sched_set_node_bw_lmt - set node's BW limit
2567  * @pi: port information structure
2568  * @node: tree node
2569  * @rl_type: rate limit type min, max, or shared
2570  * @bw: bandwidth in Kbps - Kilo bits per sec
2571  *
2572  * It updates node's BW limit parameters like BW RL profile ID of type CIR,
2573  * EIR, or SRL. The caller needs to hold scheduler lock.
2574  */
2575 static enum ice_status
2576 ice_sched_set_node_bw_lmt(struct ice_port_info *pi, struct ice_sched_node *node,
2577 			  enum ice_rl_type rl_type, u32 bw)
2578 {
2579 	struct ice_sched_node *cfg_node = node;
2580 	enum ice_status status;
2581 
2582 	struct ice_hw *hw;
2583 	u8 layer_num;
2584 
2585 	if (!pi)
2586 		return ICE_ERR_PARAM;
2587 	hw = pi->hw;
2588 	/* Remove unused RL profile IDs from HW and SW DB */
2589 	ice_sched_rm_unused_rl_prof(pi);
2590 	layer_num = ice_sched_get_rl_prof_layer(pi, rl_type,
2591 						node->tx_sched_layer);
2592 	if (layer_num >= hw->num_tx_sched_layers)
2593 		return ICE_ERR_PARAM;
2594 
2595 	if (rl_type == ICE_SHARED_BW) {
2596 		/* SRL node may be different */
2597 		cfg_node = ice_sched_get_srl_node(node, layer_num);
2598 		if (!cfg_node)
2599 			return ICE_ERR_CFG;
2600 	}
2601 	/* EIR BW and Shared BW profiles are mutually exclusive and
2602 	 * hence only one of them may be set for any given element
2603 	 */
2604 	status = ice_sched_set_eir_srl_excl(pi, cfg_node, layer_num, rl_type,
2605 					    bw);
2606 	if (status)
2607 		return status;
2608 	if (bw == ICE_SCHED_DFLT_BW)
2609 		return ice_sched_set_node_bw_dflt(pi, cfg_node, rl_type,
2610 						  layer_num);
2611 	return ice_sched_set_node_bw(pi, cfg_node, rl_type, bw, layer_num);
2612 }
2613 
2614 /**
2615  * ice_sched_set_node_bw_dflt_lmt - set node's BW limit to default
2616  * @pi: port information structure
2617  * @node: pointer to node structure
2618  * @rl_type: rate limit type min, max, or shared
2619  *
2620  * This function configures node element's BW rate limit profile ID of
2621  * type CIR, EIR, or SRL to default. This function needs to be called
2622  * with the scheduler lock held.
2623  */
2624 static enum ice_status
2625 ice_sched_set_node_bw_dflt_lmt(struct ice_port_info *pi,
2626 			       struct ice_sched_node *node,
2627 			       enum ice_rl_type rl_type)
2628 {
2629 	return ice_sched_set_node_bw_lmt(pi, node, rl_type,
2630 					 ICE_SCHED_DFLT_BW);
2631 }
2632 
2633 /**
2634  * ice_sched_validate_srl_node - Check node for SRL applicability
2635  * @node: sched node to configure
2636  * @sel_layer: selected SRL layer
2637  *
2638  * This function checks if the SRL can be applied to a selected layer node on
2639  * behalf of the requested node (first argument). This function needs to be
2640  * called with scheduler lock held.
2641  */
2642 static enum ice_status
2643 ice_sched_validate_srl_node(struct ice_sched_node *node, u8 sel_layer)
2644 {
2645 	/* SRL profiles are not available on all layers. Check if the
2646 	 * SRL profile can be applied to a node above or below the
2647 	 * requested node. SRL configuration is possible only if the
2648 	 * selected layer's node has single child.
2649 	 */
2650 	if (sel_layer == node->tx_sched_layer ||
2651 	    ((sel_layer == node->tx_sched_layer + 1) &&
2652 	    node->num_children == 1) ||
2653 	    ((sel_layer == node->tx_sched_layer - 1) &&
2654 	    (node->parent && node->parent->num_children == 1)))
2655 		return 0;
2656 
2657 	return ICE_ERR_CFG;
2658 }
2659 
2660 /**
2661  * ice_sched_save_q_bw - save queue node's BW information
2662  * @q_ctx: queue context structure
2663  * @rl_type: rate limit type min, max, or shared
2664  * @bw: bandwidth in Kbps - Kilo bits per sec
2665  *
2666  * Save BW information of queue type node for post replay use.
2667  */
2668 static enum ice_status
2669 ice_sched_save_q_bw(struct ice_q_ctx *q_ctx, enum ice_rl_type rl_type, u32 bw)
2670 {
2671 	switch (rl_type) {
2672 	case ICE_MIN_BW:
2673 		ice_set_clear_cir_bw(&q_ctx->bw_t_info, bw);
2674 		break;
2675 	case ICE_MAX_BW:
2676 		ice_set_clear_eir_bw(&q_ctx->bw_t_info, bw);
2677 		break;
2678 	case ICE_SHARED_BW:
2679 		ice_set_clear_shared_bw(&q_ctx->bw_t_info, bw);
2680 		break;
2681 	default:
2682 		return ICE_ERR_PARAM;
2683 	}
2684 	return 0;
2685 }
2686 
2687 /**
2688  * ice_sched_set_q_bw_lmt - sets queue BW limit
2689  * @pi: port information structure
2690  * @vsi_handle: sw VSI handle
2691  * @tc: traffic class
2692  * @q_handle: software queue handle
2693  * @rl_type: min, max, or shared
2694  * @bw: bandwidth in Kbps
2695  *
2696  * This function sets BW limit of queue scheduling node.
2697  */
2698 static enum ice_status
2699 ice_sched_set_q_bw_lmt(struct ice_port_info *pi, u16 vsi_handle, u8 tc,
2700 		       u16 q_handle, enum ice_rl_type rl_type, u32 bw)
2701 {
2702 	enum ice_status status = ICE_ERR_PARAM;
2703 	struct ice_sched_node *node;
2704 	struct ice_q_ctx *q_ctx;
2705 
2706 	if (!ice_is_vsi_valid(pi->hw, vsi_handle))
2707 		return ICE_ERR_PARAM;
2708 	mutex_lock(&pi->sched_lock);
2709 	q_ctx = ice_get_lan_q_ctx(pi->hw, vsi_handle, tc, q_handle);
2710 	if (!q_ctx)
2711 		goto exit_q_bw_lmt;
2712 	node = ice_sched_find_node_by_teid(pi->root, q_ctx->q_teid);
2713 	if (!node) {
2714 		ice_debug(pi->hw, ICE_DBG_SCHED, "Wrong q_teid\n");
2715 		goto exit_q_bw_lmt;
2716 	}
2717 
2718 	/* Return error if it is not a leaf node */
2719 	if (node->info.data.elem_type != ICE_AQC_ELEM_TYPE_LEAF)
2720 		goto exit_q_bw_lmt;
2721 
2722 	/* SRL bandwidth layer selection */
2723 	if (rl_type == ICE_SHARED_BW) {
2724 		u8 sel_layer; /* selected layer */
2725 
2726 		sel_layer = ice_sched_get_rl_prof_layer(pi, rl_type,
2727 							node->tx_sched_layer);
2728 		if (sel_layer >= pi->hw->num_tx_sched_layers) {
2729 			status = ICE_ERR_PARAM;
2730 			goto exit_q_bw_lmt;
2731 		}
2732 		status = ice_sched_validate_srl_node(node, sel_layer);
2733 		if (status)
2734 			goto exit_q_bw_lmt;
2735 	}
2736 
2737 	if (bw == ICE_SCHED_DFLT_BW)
2738 		status = ice_sched_set_node_bw_dflt_lmt(pi, node, rl_type);
2739 	else
2740 		status = ice_sched_set_node_bw_lmt(pi, node, rl_type, bw);
2741 
2742 	if (!status)
2743 		status = ice_sched_save_q_bw(q_ctx, rl_type, bw);
2744 
2745 exit_q_bw_lmt:
2746 	mutex_unlock(&pi->sched_lock);
2747 	return status;
2748 }
2749 
2750 /**
2751  * ice_cfg_q_bw_lmt - configure queue BW limit
2752  * @pi: port information structure
2753  * @vsi_handle: sw VSI handle
2754  * @tc: traffic class
2755  * @q_handle: software queue handle
2756  * @rl_type: min, max, or shared
2757  * @bw: bandwidth in Kbps
2758  *
2759  * This function configures BW limit of queue scheduling node.
2760  */
2761 enum ice_status
2762 ice_cfg_q_bw_lmt(struct ice_port_info *pi, u16 vsi_handle, u8 tc,
2763 		 u16 q_handle, enum ice_rl_type rl_type, u32 bw)
2764 {
2765 	return ice_sched_set_q_bw_lmt(pi, vsi_handle, tc, q_handle, rl_type,
2766 				      bw);
2767 }
2768 
2769 /**
2770  * ice_cfg_q_bw_dflt_lmt - configure queue BW default limit
2771  * @pi: port information structure
2772  * @vsi_handle: sw VSI handle
2773  * @tc: traffic class
2774  * @q_handle: software queue handle
2775  * @rl_type: min, max, or shared
2776  *
2777  * This function configures BW default limit of queue scheduling node.
2778  */
2779 enum ice_status
2780 ice_cfg_q_bw_dflt_lmt(struct ice_port_info *pi, u16 vsi_handle, u8 tc,
2781 		      u16 q_handle, enum ice_rl_type rl_type)
2782 {
2783 	return ice_sched_set_q_bw_lmt(pi, vsi_handle, tc, q_handle, rl_type,
2784 				      ICE_SCHED_DFLT_BW);
2785 }
2786 
2787 /**
2788  * ice_cfg_rl_burst_size - Set burst size value
2789  * @hw: pointer to the HW struct
2790  * @bytes: burst size in bytes
2791  *
2792  * This function configures/set the burst size to requested new value. The new
2793  * burst size value is used for future rate limit calls. It doesn't change the
2794  * existing or previously created RL profiles.
2795  */
2796 enum ice_status ice_cfg_rl_burst_size(struct ice_hw *hw, u32 bytes)
2797 {
2798 	u16 burst_size_to_prog;
2799 
2800 	if (bytes < ICE_MIN_BURST_SIZE_ALLOWED ||
2801 	    bytes > ICE_MAX_BURST_SIZE_ALLOWED)
2802 		return ICE_ERR_PARAM;
2803 	if (ice_round_to_num(bytes, 64) <=
2804 	    ICE_MAX_BURST_SIZE_64_BYTE_GRANULARITY) {
2805 		/* 64 byte granularity case */
2806 		/* Disable MSB granularity bit */
2807 		burst_size_to_prog = ICE_64_BYTE_GRANULARITY;
2808 		/* round number to nearest 64 byte granularity */
2809 		bytes = ice_round_to_num(bytes, 64);
2810 		/* The value is in 64 byte chunks */
2811 		burst_size_to_prog |= (u16)(bytes / 64);
2812 	} else {
2813 		/* k bytes granularity case */
2814 		/* Enable MSB granularity bit */
2815 		burst_size_to_prog = ICE_KBYTE_GRANULARITY;
2816 		/* round number to nearest 1024 granularity */
2817 		bytes = ice_round_to_num(bytes, 1024);
2818 		/* check rounding doesn't go beyond allowed */
2819 		if (bytes > ICE_MAX_BURST_SIZE_KBYTE_GRANULARITY)
2820 			bytes = ICE_MAX_BURST_SIZE_KBYTE_GRANULARITY;
2821 		/* The value is in k bytes */
2822 		burst_size_to_prog |= (u16)(bytes / 1024);
2823 	}
2824 	hw->max_burst_size = burst_size_to_prog;
2825 	return 0;
2826 }
2827 
2828 /**
2829  * ice_sched_replay_node_prio - re-configure node priority
2830  * @hw: pointer to the HW struct
2831  * @node: sched node to configure
2832  * @priority: priority value
2833  *
2834  * This function configures node element's priority value. It
2835  * needs to be called with scheduler lock held.
2836  */
2837 static enum ice_status
2838 ice_sched_replay_node_prio(struct ice_hw *hw, struct ice_sched_node *node,
2839 			   u8 priority)
2840 {
2841 	struct ice_aqc_txsched_elem_data buf;
2842 	struct ice_aqc_txsched_elem *data;
2843 	enum ice_status status;
2844 
2845 	buf = node->info;
2846 	data = &buf.data;
2847 	data->valid_sections |= ICE_AQC_ELEM_VALID_GENERIC;
2848 	data->generic = priority;
2849 
2850 	/* Configure element */
2851 	status = ice_sched_update_elem(hw, node, &buf);
2852 	return status;
2853 }
2854 
2855 /**
2856  * ice_sched_replay_node_bw - replay node(s) BW
2857  * @hw: pointer to the HW struct
2858  * @node: sched node to configure
2859  * @bw_t_info: BW type information
2860  *
2861  * This function restores node's BW from bw_t_info. The caller needs
2862  * to hold the scheduler lock.
2863  */
2864 static enum ice_status
2865 ice_sched_replay_node_bw(struct ice_hw *hw, struct ice_sched_node *node,
2866 			 struct ice_bw_type_info *bw_t_info)
2867 {
2868 	struct ice_port_info *pi = hw->port_info;
2869 	enum ice_status status = ICE_ERR_PARAM;
2870 	u16 bw_alloc;
2871 
2872 	if (!node)
2873 		return status;
2874 	if (bitmap_empty(bw_t_info->bw_t_bitmap, ICE_BW_TYPE_CNT))
2875 		return 0;
2876 	if (test_bit(ICE_BW_TYPE_PRIO, bw_t_info->bw_t_bitmap)) {
2877 		status = ice_sched_replay_node_prio(hw, node,
2878 						    bw_t_info->generic);
2879 		if (status)
2880 			return status;
2881 	}
2882 	if (test_bit(ICE_BW_TYPE_CIR, bw_t_info->bw_t_bitmap)) {
2883 		status = ice_sched_set_node_bw_lmt(pi, node, ICE_MIN_BW,
2884 						   bw_t_info->cir_bw.bw);
2885 		if (status)
2886 			return status;
2887 	}
2888 	if (test_bit(ICE_BW_TYPE_CIR_WT, bw_t_info->bw_t_bitmap)) {
2889 		bw_alloc = bw_t_info->cir_bw.bw_alloc;
2890 		status = ice_sched_cfg_node_bw_alloc(hw, node, ICE_MIN_BW,
2891 						     bw_alloc);
2892 		if (status)
2893 			return status;
2894 	}
2895 	if (test_bit(ICE_BW_TYPE_EIR, bw_t_info->bw_t_bitmap)) {
2896 		status = ice_sched_set_node_bw_lmt(pi, node, ICE_MAX_BW,
2897 						   bw_t_info->eir_bw.bw);
2898 		if (status)
2899 			return status;
2900 	}
2901 	if (test_bit(ICE_BW_TYPE_EIR_WT, bw_t_info->bw_t_bitmap)) {
2902 		bw_alloc = bw_t_info->eir_bw.bw_alloc;
2903 		status = ice_sched_cfg_node_bw_alloc(hw, node, ICE_MAX_BW,
2904 						     bw_alloc);
2905 		if (status)
2906 			return status;
2907 	}
2908 	if (test_bit(ICE_BW_TYPE_SHARED, bw_t_info->bw_t_bitmap))
2909 		status = ice_sched_set_node_bw_lmt(pi, node, ICE_SHARED_BW,
2910 						   bw_t_info->shared_bw);
2911 	return status;
2912 }
2913 
2914 /**
2915  * ice_sched_replay_q_bw - replay queue type node BW
2916  * @pi: port information structure
2917  * @q_ctx: queue context structure
2918  *
2919  * This function replays queue type node bandwidth. This function needs to be
2920  * called with scheduler lock held.
2921  */
2922 enum ice_status
2923 ice_sched_replay_q_bw(struct ice_port_info *pi, struct ice_q_ctx *q_ctx)
2924 {
2925 	struct ice_sched_node *q_node;
2926 
2927 	/* Following also checks the presence of node in tree */
2928 	q_node = ice_sched_find_node_by_teid(pi->root, q_ctx->q_teid);
2929 	if (!q_node)
2930 		return ICE_ERR_PARAM;
2931 	return ice_sched_replay_node_bw(pi->hw, q_node, &q_ctx->bw_t_info);
2932 }
2933