xref: /linux/drivers/net/ethernet/intel/ice/devlink/devlink.c (revision 6ae67f115986734bc24dfed5cee1daafb5351623)
1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright (c) 2020, Intel Corporation. */
3 
4 #include <linux/vmalloc.h>
5 
6 #include "ice.h"
7 #include "ice_lib.h"
8 #include "devlink.h"
9 #include "port.h"
10 #include "ice_eswitch.h"
11 #include "ice_fw_update.h"
12 #include "ice_dcb_lib.h"
13 #include "ice_sf_eth.h"
14 
15 /* context for devlink info version reporting */
16 struct ice_info_ctx {
17 	char buf[128];
18 	struct ice_orom_info pending_orom;
19 	struct ice_nvm_info pending_nvm;
20 	struct ice_netlist_info pending_netlist;
21 	struct ice_hw_dev_caps dev_caps;
22 };
23 
24 /* The following functions are used to format specific strings for various
25  * devlink info versions. The ctx parameter is used to provide the storage
26  * buffer, as well as any ancillary information calculated when the info
27  * request was made.
28  *
29  * If a version does not exist, for example when attempting to get the
30  * inactive version of flash when there is no pending update, the function
31  * should leave the buffer in the ctx structure empty.
32  */
33 
34 static void ice_info_get_dsn(struct ice_pf *pf, struct ice_info_ctx *ctx)
35 {
36 	u8 dsn[8];
37 
38 	/* Copy the DSN into an array in Big Endian format */
39 	put_unaligned_be64(pci_get_dsn(pf->pdev), dsn);
40 
41 	snprintf(ctx->buf, sizeof(ctx->buf), "%8phD", dsn);
42 }
43 
44 static void ice_info_pba(struct ice_pf *pf, struct ice_info_ctx *ctx)
45 {
46 	struct ice_hw *hw = &pf->hw;
47 	int status;
48 
49 	status = ice_read_pba_string(hw, (u8 *)ctx->buf, sizeof(ctx->buf));
50 	if (status)
51 		/* We failed to locate the PBA, so just skip this entry */
52 		dev_dbg(ice_pf_to_dev(pf), "Failed to read Product Board Assembly string, status %d\n",
53 			status);
54 }
55 
56 static void ice_info_fw_mgmt(struct ice_pf *pf, struct ice_info_ctx *ctx)
57 {
58 	struct ice_hw *hw = &pf->hw;
59 
60 	snprintf(ctx->buf, sizeof(ctx->buf), "%u.%u.%u",
61 		 hw->fw_maj_ver, hw->fw_min_ver, hw->fw_patch);
62 }
63 
64 static void ice_info_fw_api(struct ice_pf *pf, struct ice_info_ctx *ctx)
65 {
66 	struct ice_hw *hw = &pf->hw;
67 
68 	snprintf(ctx->buf, sizeof(ctx->buf), "%u.%u.%u", hw->api_maj_ver,
69 		 hw->api_min_ver, hw->api_patch);
70 }
71 
72 static void ice_info_fw_build(struct ice_pf *pf, struct ice_info_ctx *ctx)
73 {
74 	struct ice_hw *hw = &pf->hw;
75 
76 	snprintf(ctx->buf, sizeof(ctx->buf), "0x%08x", hw->fw_build);
77 }
78 
79 static void ice_info_orom_ver(struct ice_pf *pf, struct ice_info_ctx *ctx)
80 {
81 	struct ice_orom_info *orom = &pf->hw.flash.orom;
82 
83 	snprintf(ctx->buf, sizeof(ctx->buf), "%u.%u.%u",
84 		 orom->major, orom->build, orom->patch);
85 }
86 
87 static void
88 ice_info_pending_orom_ver(struct ice_pf __always_unused *pf,
89 			  struct ice_info_ctx *ctx)
90 {
91 	struct ice_orom_info *orom = &ctx->pending_orom;
92 
93 	if (ctx->dev_caps.common_cap.nvm_update_pending_orom)
94 		snprintf(ctx->buf, sizeof(ctx->buf), "%u.%u.%u",
95 			 orom->major, orom->build, orom->patch);
96 }
97 
98 static void ice_info_nvm_ver(struct ice_pf *pf, struct ice_info_ctx *ctx)
99 {
100 	struct ice_nvm_info *nvm = &pf->hw.flash.nvm;
101 
102 	snprintf(ctx->buf, sizeof(ctx->buf), "%x.%02x", nvm->major, nvm->minor);
103 }
104 
105 static void
106 ice_info_pending_nvm_ver(struct ice_pf __always_unused *pf,
107 			 struct ice_info_ctx *ctx)
108 {
109 	struct ice_nvm_info *nvm = &ctx->pending_nvm;
110 
111 	if (ctx->dev_caps.common_cap.nvm_update_pending_nvm)
112 		snprintf(ctx->buf, sizeof(ctx->buf), "%x.%02x",
113 			 nvm->major, nvm->minor);
114 }
115 
116 static void ice_info_eetrack(struct ice_pf *pf, struct ice_info_ctx *ctx)
117 {
118 	struct ice_nvm_info *nvm = &pf->hw.flash.nvm;
119 
120 	snprintf(ctx->buf, sizeof(ctx->buf), "0x%08x", nvm->eetrack);
121 }
122 
123 static void
124 ice_info_pending_eetrack(struct ice_pf *pf, struct ice_info_ctx *ctx)
125 {
126 	struct ice_nvm_info *nvm = &ctx->pending_nvm;
127 
128 	if (ctx->dev_caps.common_cap.nvm_update_pending_nvm)
129 		snprintf(ctx->buf, sizeof(ctx->buf), "0x%08x", nvm->eetrack);
130 }
131 
132 static void ice_info_ddp_pkg_name(struct ice_pf *pf, struct ice_info_ctx *ctx)
133 {
134 	struct ice_hw *hw = &pf->hw;
135 
136 	snprintf(ctx->buf, sizeof(ctx->buf), "%s", hw->active_pkg_name);
137 }
138 
139 static void
140 ice_info_ddp_pkg_version(struct ice_pf *pf, struct ice_info_ctx *ctx)
141 {
142 	struct ice_pkg_ver *pkg = &pf->hw.active_pkg_ver;
143 
144 	snprintf(ctx->buf, sizeof(ctx->buf), "%u.%u.%u.%u",
145 		 pkg->major, pkg->minor, pkg->update, pkg->draft);
146 }
147 
148 static void
149 ice_info_ddp_pkg_bundle_id(struct ice_pf *pf, struct ice_info_ctx *ctx)
150 {
151 	snprintf(ctx->buf, sizeof(ctx->buf), "0x%08x", pf->hw.active_track_id);
152 }
153 
154 static void ice_info_netlist_ver(struct ice_pf *pf, struct ice_info_ctx *ctx)
155 {
156 	struct ice_netlist_info *netlist = &pf->hw.flash.netlist;
157 
158 	/* The netlist version fields are BCD formatted */
159 	snprintf(ctx->buf, sizeof(ctx->buf), "%x.%x.%x-%x.%x.%x",
160 		 netlist->major, netlist->minor,
161 		 netlist->type >> 16, netlist->type & 0xFFFF,
162 		 netlist->rev, netlist->cust_ver);
163 }
164 
165 static void ice_info_netlist_build(struct ice_pf *pf, struct ice_info_ctx *ctx)
166 {
167 	struct ice_netlist_info *netlist = &pf->hw.flash.netlist;
168 
169 	snprintf(ctx->buf, sizeof(ctx->buf), "0x%08x", netlist->hash);
170 }
171 
172 static void
173 ice_info_pending_netlist_ver(struct ice_pf __always_unused *pf,
174 			     struct ice_info_ctx *ctx)
175 {
176 	struct ice_netlist_info *netlist = &ctx->pending_netlist;
177 
178 	/* The netlist version fields are BCD formatted */
179 	if (ctx->dev_caps.common_cap.nvm_update_pending_netlist)
180 		snprintf(ctx->buf, sizeof(ctx->buf), "%x.%x.%x-%x.%x.%x",
181 			 netlist->major, netlist->minor,
182 			 netlist->type >> 16, netlist->type & 0xFFFF,
183 			 netlist->rev, netlist->cust_ver);
184 }
185 
186 static void
187 ice_info_pending_netlist_build(struct ice_pf __always_unused *pf,
188 			       struct ice_info_ctx *ctx)
189 {
190 	struct ice_netlist_info *netlist = &ctx->pending_netlist;
191 
192 	if (ctx->dev_caps.common_cap.nvm_update_pending_netlist)
193 		snprintf(ctx->buf, sizeof(ctx->buf), "0x%08x", netlist->hash);
194 }
195 
196 static void ice_info_cgu_fw_build(struct ice_pf *pf, struct ice_info_ctx *ctx)
197 {
198 	u32 id, cfg_ver, fw_ver;
199 
200 	if (!ice_is_feature_supported(pf, ICE_F_CGU))
201 		return;
202 	if (ice_aq_get_cgu_info(&pf->hw, &id, &cfg_ver, &fw_ver))
203 		return;
204 	snprintf(ctx->buf, sizeof(ctx->buf), "%u.%u.%u", id, cfg_ver, fw_ver);
205 }
206 
207 static void ice_info_cgu_id(struct ice_pf *pf, struct ice_info_ctx *ctx)
208 {
209 	if (!ice_is_feature_supported(pf, ICE_F_CGU))
210 		return;
211 	snprintf(ctx->buf, sizeof(ctx->buf), "%u", pf->hw.cgu_part_number);
212 }
213 
214 #define fixed(key, getter) { ICE_VERSION_FIXED, key, getter, NULL }
215 #define running(key, getter) { ICE_VERSION_RUNNING, key, getter, NULL }
216 #define stored(key, getter, fallback) { ICE_VERSION_STORED, key, getter, fallback }
217 
218 /* The combined() macro inserts both the running entry as well as a stored
219  * entry. The running entry will always report the version from the active
220  * handler. The stored entry will first try the pending handler, and fallback
221  * to the active handler if the pending function does not report a version.
222  * The pending handler should check the status of a pending update for the
223  * relevant flash component. It should only fill in the buffer in the case
224  * where a valid pending version is available. This ensures that the related
225  * stored and running versions remain in sync, and that stored versions are
226  * correctly reported as expected.
227  */
228 #define combined(key, active, pending) \
229 	running(key, active), \
230 	stored(key, pending, active)
231 
232 enum ice_version_type {
233 	ICE_VERSION_FIXED,
234 	ICE_VERSION_RUNNING,
235 	ICE_VERSION_STORED,
236 };
237 
238 static const struct ice_devlink_version {
239 	enum ice_version_type type;
240 	const char *key;
241 	void (*getter)(struct ice_pf *pf, struct ice_info_ctx *ctx);
242 	void (*fallback)(struct ice_pf *pf, struct ice_info_ctx *ctx);
243 } ice_devlink_versions[] = {
244 	fixed(DEVLINK_INFO_VERSION_GENERIC_BOARD_ID, ice_info_pba),
245 	running(DEVLINK_INFO_VERSION_GENERIC_FW_MGMT, ice_info_fw_mgmt),
246 	running("fw.mgmt.api", ice_info_fw_api),
247 	running("fw.mgmt.build", ice_info_fw_build),
248 	combined(DEVLINK_INFO_VERSION_GENERIC_FW_UNDI, ice_info_orom_ver, ice_info_pending_orom_ver),
249 	combined("fw.psid.api", ice_info_nvm_ver, ice_info_pending_nvm_ver),
250 	combined(DEVLINK_INFO_VERSION_GENERIC_FW_BUNDLE_ID, ice_info_eetrack, ice_info_pending_eetrack),
251 	running("fw.app.name", ice_info_ddp_pkg_name),
252 	running(DEVLINK_INFO_VERSION_GENERIC_FW_APP, ice_info_ddp_pkg_version),
253 	running("fw.app.bundle_id", ice_info_ddp_pkg_bundle_id),
254 	combined("fw.netlist", ice_info_netlist_ver, ice_info_pending_netlist_ver),
255 	combined("fw.netlist.build", ice_info_netlist_build, ice_info_pending_netlist_build),
256 	fixed("cgu.id", ice_info_cgu_id),
257 	running("fw.cgu", ice_info_cgu_fw_build),
258 };
259 
260 /**
261  * ice_devlink_info_get - .info_get devlink handler
262  * @devlink: devlink instance structure
263  * @req: the devlink info request
264  * @extack: extended netdev ack structure
265  *
266  * Callback for the devlink .info_get operation. Reports information about the
267  * device.
268  *
269  * Return: zero on success or an error code on failure.
270  */
271 static int ice_devlink_info_get(struct devlink *devlink,
272 				struct devlink_info_req *req,
273 				struct netlink_ext_ack *extack)
274 {
275 	struct ice_pf *pf = devlink_priv(devlink);
276 	struct device *dev = ice_pf_to_dev(pf);
277 	struct ice_hw *hw = &pf->hw;
278 	struct ice_info_ctx *ctx;
279 	size_t i;
280 	int err;
281 
282 	err = ice_wait_for_reset(pf, 10 * HZ);
283 	if (err) {
284 		NL_SET_ERR_MSG_MOD(extack, "Device is busy resetting");
285 		return err;
286 	}
287 
288 	ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
289 	if (!ctx)
290 		return -ENOMEM;
291 
292 	/* discover capabilities first */
293 	err = ice_discover_dev_caps(hw, &ctx->dev_caps);
294 	if (err) {
295 		dev_dbg(dev, "Failed to discover device capabilities, status %d aq_err %s\n",
296 			err, libie_aq_str(hw->adminq.sq_last_status));
297 		NL_SET_ERR_MSG_MOD(extack, "Unable to discover device capabilities");
298 		goto out_free_ctx;
299 	}
300 
301 	if (ctx->dev_caps.common_cap.nvm_update_pending_orom) {
302 		err = ice_get_inactive_orom_ver(hw, &ctx->pending_orom);
303 		if (err) {
304 			dev_dbg(dev, "Unable to read inactive Option ROM version data, status %d aq_err %s\n",
305 				err, libie_aq_str(hw->adminq.sq_last_status));
306 
307 			/* disable display of pending Option ROM */
308 			ctx->dev_caps.common_cap.nvm_update_pending_orom = false;
309 		}
310 	}
311 
312 	if (ctx->dev_caps.common_cap.nvm_update_pending_nvm) {
313 		err = ice_get_inactive_nvm_ver(hw, &ctx->pending_nvm);
314 		if (err) {
315 			dev_dbg(dev, "Unable to read inactive NVM version data, status %d aq_err %s\n",
316 				err, libie_aq_str(hw->adminq.sq_last_status));
317 
318 			/* disable display of pending Option ROM */
319 			ctx->dev_caps.common_cap.nvm_update_pending_nvm = false;
320 		}
321 	}
322 
323 	if (ctx->dev_caps.common_cap.nvm_update_pending_netlist) {
324 		err = ice_get_inactive_netlist_ver(hw, &ctx->pending_netlist);
325 		if (err) {
326 			dev_dbg(dev, "Unable to read inactive Netlist version data, status %d aq_err %s\n",
327 				err, libie_aq_str(hw->adminq.sq_last_status));
328 
329 			/* disable display of pending Option ROM */
330 			ctx->dev_caps.common_cap.nvm_update_pending_netlist = false;
331 		}
332 	}
333 
334 	ice_info_get_dsn(pf, ctx);
335 
336 	err = devlink_info_serial_number_put(req, ctx->buf);
337 	if (err) {
338 		NL_SET_ERR_MSG_MOD(extack, "Unable to set serial number");
339 		goto out_free_ctx;
340 	}
341 
342 	for (i = 0; i < ARRAY_SIZE(ice_devlink_versions); i++) {
343 		enum ice_version_type type = ice_devlink_versions[i].type;
344 		const char *key = ice_devlink_versions[i].key;
345 
346 		memset(ctx->buf, 0, sizeof(ctx->buf));
347 
348 		ice_devlink_versions[i].getter(pf, ctx);
349 
350 		/* If the default getter doesn't report a version, use the
351 		 * fallback function. This is primarily useful in the case of
352 		 * "stored" versions that want to report the same value as the
353 		 * running version in the normal case of no pending update.
354 		 */
355 		if (ctx->buf[0] == '\0' && ice_devlink_versions[i].fallback)
356 			ice_devlink_versions[i].fallback(pf, ctx);
357 
358 		/* Do not report missing versions */
359 		if (ctx->buf[0] == '\0')
360 			continue;
361 
362 		switch (type) {
363 		case ICE_VERSION_FIXED:
364 			err = devlink_info_version_fixed_put(req, key, ctx->buf);
365 			if (err) {
366 				NL_SET_ERR_MSG_MOD(extack, "Unable to set fixed version");
367 				goto out_free_ctx;
368 			}
369 			break;
370 		case ICE_VERSION_RUNNING:
371 			err = devlink_info_version_running_put_ext(req, key,
372 								   ctx->buf,
373 								   DEVLINK_INFO_VERSION_TYPE_COMPONENT);
374 			if (err) {
375 				NL_SET_ERR_MSG_MOD(extack, "Unable to set running version");
376 				goto out_free_ctx;
377 			}
378 			break;
379 		case ICE_VERSION_STORED:
380 			err = devlink_info_version_stored_put_ext(req, key,
381 								  ctx->buf,
382 								  DEVLINK_INFO_VERSION_TYPE_COMPONENT);
383 			if (err) {
384 				NL_SET_ERR_MSG_MOD(extack, "Unable to set stored version");
385 				goto out_free_ctx;
386 			}
387 			break;
388 		}
389 	}
390 
391 out_free_ctx:
392 	kfree(ctx);
393 	return err;
394 }
395 
396 /**
397  * ice_devlink_reload_empr_start - Start EMP reset to activate new firmware
398  * @pf: pointer to the pf instance
399  * @extack: netlink extended ACK structure
400  *
401  * Allow user to activate new Embedded Management Processor firmware by
402  * issuing device specific EMP reset. Called in response to
403  * a DEVLINK_CMD_RELOAD with the DEVLINK_RELOAD_ACTION_FW_ACTIVATE.
404  *
405  * Note that teardown and rebuild of the driver state happens automatically as
406  * part of an interrupt and watchdog task. This is because all physical
407  * functions on the device must be able to reset when an EMP reset occurs from
408  * any source.
409  */
410 static int
411 ice_devlink_reload_empr_start(struct ice_pf *pf,
412 			      struct netlink_ext_ack *extack)
413 {
414 	struct device *dev = ice_pf_to_dev(pf);
415 	struct ice_hw *hw = &pf->hw;
416 	u8 pending;
417 	int err;
418 
419 	err = ice_get_pending_updates(pf, &pending, extack);
420 	if (err)
421 		return err;
422 
423 	/* pending is a bitmask of which flash banks have a pending update,
424 	 * including the main NVM bank, the Option ROM bank, and the netlist
425 	 * bank. If any of these bits are set, then there is a pending update
426 	 * waiting to be activated.
427 	 */
428 	if (!pending) {
429 		NL_SET_ERR_MSG_MOD(extack, "No pending firmware update");
430 		return -ECANCELED;
431 	}
432 
433 	if (pf->fw_emp_reset_disabled) {
434 		NL_SET_ERR_MSG_MOD(extack, "EMP reset is not available. To activate firmware, a reboot or power cycle is needed");
435 		return -ECANCELED;
436 	}
437 
438 	dev_dbg(dev, "Issuing device EMP reset to activate firmware\n");
439 
440 	err = ice_aq_nvm_update_empr(hw);
441 	if (err) {
442 		dev_err(dev, "Failed to trigger EMP device reset to reload firmware, err %d aq_err %s\n",
443 			err, libie_aq_str(hw->adminq.sq_last_status));
444 		NL_SET_ERR_MSG_MOD(extack, "Failed to trigger EMP device reset to reload firmware");
445 		return err;
446 	}
447 
448 	return 0;
449 }
450 
451 /**
452  * ice_devlink_reinit_down - unload given PF
453  * @pf: pointer to the PF struct
454  */
455 static void ice_devlink_reinit_down(struct ice_pf *pf)
456 {
457 	/* No need to take devl_lock, it's already taken by devlink API */
458 	ice_unload(pf);
459 	rtnl_lock();
460 	ice_vsi_decfg(ice_get_main_vsi(pf));
461 	rtnl_unlock();
462 	ice_deinit_pf(pf);
463 	ice_deinit_dev(pf);
464 }
465 
466 /**
467  * ice_devlink_reload_down - prepare for reload
468  * @devlink: pointer to the devlink instance to reload
469  * @netns_change: if true, the network namespace is changing
470  * @action: the action to perform
471  * @limit: limits on what reload should do, such as not resetting
472  * @extack: netlink extended ACK structure
473  */
474 static int
475 ice_devlink_reload_down(struct devlink *devlink, bool netns_change,
476 			enum devlink_reload_action action,
477 			enum devlink_reload_limit limit,
478 			struct netlink_ext_ack *extack)
479 {
480 	struct ice_pf *pf = devlink_priv(devlink);
481 
482 	switch (action) {
483 	case DEVLINK_RELOAD_ACTION_DRIVER_REINIT:
484 		if (ice_is_eswitch_mode_switchdev(pf)) {
485 			NL_SET_ERR_MSG_MOD(extack,
486 					   "Go to legacy mode before doing reinit");
487 			return -EOPNOTSUPP;
488 		}
489 		if (ice_is_adq_active(pf)) {
490 			NL_SET_ERR_MSG_MOD(extack,
491 					   "Turn off ADQ before doing reinit");
492 			return -EOPNOTSUPP;
493 		}
494 		if (ice_has_vfs(pf)) {
495 			NL_SET_ERR_MSG_MOD(extack,
496 					   "Remove all VFs before doing reinit");
497 			return -EOPNOTSUPP;
498 		}
499 		ice_devlink_reinit_down(pf);
500 		return 0;
501 	case DEVLINK_RELOAD_ACTION_FW_ACTIVATE:
502 		return ice_devlink_reload_empr_start(pf, extack);
503 	default:
504 		WARN_ON(1);
505 		return -EOPNOTSUPP;
506 	}
507 }
508 
509 /**
510  * ice_devlink_reload_empr_finish - Wait for EMP reset to finish
511  * @pf: pointer to the pf instance
512  * @extack: netlink extended ACK structure
513  *
514  * Wait for driver to finish rebuilding after EMP reset is completed. This
515  * includes time to wait for both the actual device reset as well as the time
516  * for the driver's rebuild to complete.
517  */
518 static int
519 ice_devlink_reload_empr_finish(struct ice_pf *pf,
520 			       struct netlink_ext_ack *extack)
521 {
522 	int err;
523 
524 	err = ice_wait_for_reset(pf, 60 * HZ);
525 	if (err) {
526 		NL_SET_ERR_MSG_MOD(extack, "Device still resetting after 1 minute");
527 		return err;
528 	}
529 
530 	return 0;
531 }
532 
533 /**
534  * ice_get_tx_topo_user_sel - Read user's choice from flash
535  * @pf: pointer to pf structure
536  * @layers: value read from flash will be saved here
537  *
538  * Reads user's preference for Tx Scheduler Topology Tree from PFA TLV.
539  *
540  * Return: zero when read was successful, negative values otherwise.
541  */
542 static int ice_get_tx_topo_user_sel(struct ice_pf *pf, uint8_t *layers)
543 {
544 	struct ice_aqc_nvm_tx_topo_user_sel usr_sel = {};
545 	struct ice_hw *hw = &pf->hw;
546 	int err;
547 
548 	err = ice_acquire_nvm(hw, ICE_RES_READ);
549 	if (err)
550 		return err;
551 
552 	err = ice_aq_read_nvm(hw, ICE_AQC_NVM_TX_TOPO_MOD_ID, 0,
553 			      sizeof(usr_sel), &usr_sel, true, true, NULL);
554 	if (err)
555 		goto exit_release_res;
556 
557 	if (usr_sel.data & ICE_AQC_NVM_TX_TOPO_USER_SEL)
558 		*layers = ICE_SCHED_5_LAYERS;
559 	else
560 		*layers = ICE_SCHED_9_LAYERS;
561 
562 exit_release_res:
563 	ice_release_nvm(hw);
564 
565 	return err;
566 }
567 
568 /**
569  * ice_update_tx_topo_user_sel - Save user's preference in flash
570  * @pf: pointer to pf structure
571  * @layers: value to be saved in flash
572  *
573  * Variable "layers" defines user's preference about number of layers in Tx
574  * Scheduler Topology Tree. This choice should be stored in PFA TLV field
575  * and be picked up by driver, next time during init.
576  *
577  * Return: zero when save was successful, negative values otherwise.
578  */
579 static int ice_update_tx_topo_user_sel(struct ice_pf *pf, int layers)
580 {
581 	struct ice_aqc_nvm_tx_topo_user_sel usr_sel = {};
582 	struct ice_hw *hw = &pf->hw;
583 	int err;
584 
585 	err = ice_acquire_nvm(hw, ICE_RES_WRITE);
586 	if (err)
587 		return err;
588 
589 	err = ice_aq_read_nvm(hw, ICE_AQC_NVM_TX_TOPO_MOD_ID, 0,
590 			      sizeof(usr_sel), &usr_sel, true, true, NULL);
591 	if (err)
592 		goto exit_release_res;
593 
594 	if (layers == ICE_SCHED_5_LAYERS)
595 		usr_sel.data |= ICE_AQC_NVM_TX_TOPO_USER_SEL;
596 	else
597 		usr_sel.data &= ~ICE_AQC_NVM_TX_TOPO_USER_SEL;
598 
599 	err = ice_write_one_nvm_block(pf, ICE_AQC_NVM_TX_TOPO_MOD_ID, 2,
600 				      sizeof(usr_sel.data), &usr_sel.data,
601 				      true, NULL, NULL);
602 exit_release_res:
603 	ice_release_nvm(hw);
604 
605 	return err;
606 }
607 
608 /**
609  * ice_devlink_tx_sched_layers_get - Get tx_scheduling_layers parameter
610  * @devlink: pointer to the devlink instance
611  * @id: the parameter ID to set
612  * @ctx: context to store the parameter value
613  *
614  * Return: zero on success and negative value on failure.
615  */
616 static int ice_devlink_tx_sched_layers_get(struct devlink *devlink, u32 id,
617 					   struct devlink_param_gset_ctx *ctx)
618 {
619 	struct ice_pf *pf = devlink_priv(devlink);
620 	int err;
621 
622 	err = ice_get_tx_topo_user_sel(pf, &ctx->val.vu8);
623 	if (err)
624 		return err;
625 
626 	return 0;
627 }
628 
629 /**
630  * ice_devlink_tx_sched_layers_set - Set tx_scheduling_layers parameter
631  * @devlink: pointer to the devlink instance
632  * @id: the parameter ID to set
633  * @ctx: context to get the parameter value
634  * @extack: netlink extended ACK structure
635  *
636  * Return: zero on success and negative value on failure.
637  */
638 static int ice_devlink_tx_sched_layers_set(struct devlink *devlink, u32 id,
639 					   struct devlink_param_gset_ctx *ctx,
640 					   struct netlink_ext_ack *extack)
641 {
642 	struct ice_pf *pf = devlink_priv(devlink);
643 	int err;
644 
645 	err = ice_update_tx_topo_user_sel(pf, ctx->val.vu8);
646 	if (err)
647 		return err;
648 
649 	NL_SET_ERR_MSG_MOD(extack,
650 			   "Tx scheduling layers have been changed on this device. You must do the PCI slot powercycle for the change to take effect.");
651 
652 	return 0;
653 }
654 
655 /**
656  * ice_devlink_tx_sched_layers_validate - Validate passed tx_scheduling_layers
657  *                                        parameter value
658  * @devlink: unused pointer to devlink instance
659  * @id: the parameter ID to validate
660  * @val: value to validate
661  * @extack: netlink extended ACK structure
662  *
663  * Supported values are:
664  * - 5 - five layers Tx Scheduler Topology Tree
665  * - 9 - nine layers Tx Scheduler Topology Tree
666  *
667  * Return: zero when passed parameter value is supported. Negative value on
668  * error.
669  */
670 static int ice_devlink_tx_sched_layers_validate(struct devlink *devlink, u32 id,
671 						union devlink_param_value val,
672 						struct netlink_ext_ack *extack)
673 {
674 	if (val.vu8 != ICE_SCHED_5_LAYERS && val.vu8 != ICE_SCHED_9_LAYERS) {
675 		NL_SET_ERR_MSG_MOD(extack,
676 				   "Wrong number of tx scheduler layers provided.");
677 		return -EINVAL;
678 	}
679 
680 	return 0;
681 }
682 
683 /**
684  * ice_tear_down_devlink_rate_tree - removes devlink-rate exported tree
685  * @pf: pf struct
686  *
687  * This function tears down tree exported during VF's creation.
688  */
689 void ice_tear_down_devlink_rate_tree(struct ice_pf *pf)
690 {
691 	struct devlink *devlink;
692 	struct ice_vf *vf;
693 	unsigned int bkt;
694 
695 	devlink = priv_to_devlink(pf);
696 
697 	devl_lock(devlink);
698 	mutex_lock(&pf->vfs.table_lock);
699 	ice_for_each_vf(pf, bkt, vf) {
700 		if (vf->devlink_port.devlink_rate)
701 			devl_rate_leaf_destroy(&vf->devlink_port);
702 	}
703 	mutex_unlock(&pf->vfs.table_lock);
704 
705 	devl_rate_nodes_destroy(devlink);
706 	devl_unlock(devlink);
707 }
708 
709 /**
710  * ice_enable_custom_tx - try to enable custom Tx feature
711  * @pf: pf struct
712  *
713  * This function tries to enable custom Tx feature,
714  * it's not possible to enable it, if DCB or ADQ is active.
715  */
716 static bool ice_enable_custom_tx(struct ice_pf *pf)
717 {
718 	struct ice_port_info *pi = ice_get_main_vsi(pf)->port_info;
719 	struct device *dev = ice_pf_to_dev(pf);
720 
721 	if (pi->is_custom_tx_enabled)
722 		/* already enabled, return true */
723 		return true;
724 
725 	if (ice_is_adq_active(pf)) {
726 		dev_err(dev, "ADQ active, can't modify Tx scheduler tree\n");
727 		return false;
728 	}
729 
730 	if (ice_is_dcb_active(pf)) {
731 		dev_err(dev, "DCB active, can't modify Tx scheduler tree\n");
732 		return false;
733 	}
734 
735 	pi->is_custom_tx_enabled = true;
736 
737 	return true;
738 }
739 
740 /**
741  * ice_traverse_tx_tree - traverse Tx scheduler tree
742  * @devlink: devlink struct
743  * @node: current node, used for recursion
744  * @tc_node: tc_node struct, that is treated as a root
745  * @pf: pf struct
746  *
747  * This function traverses Tx scheduler tree and exports
748  * entire structure to the devlink-rate.
749  */
750 static void ice_traverse_tx_tree(struct devlink *devlink, struct ice_sched_node *node,
751 				 struct ice_sched_node *tc_node, struct ice_pf *pf)
752 {
753 	struct devlink_rate *rate_node = NULL;
754 	struct ice_dynamic_port *sf;
755 	struct ice_vf *vf;
756 	int i;
757 
758 	if (node->rate_node)
759 		/* already added, skip to the next */
760 		goto traverse_children;
761 
762 	if (node->parent == tc_node) {
763 		/* create root node */
764 		rate_node = devl_rate_node_create(devlink, node, node->name, NULL);
765 	} else if (node->vsi_handle &&
766 		   pf->vsi[node->vsi_handle]->type == ICE_VSI_VF &&
767 		   pf->vsi[node->vsi_handle]->vf) {
768 		vf = pf->vsi[node->vsi_handle]->vf;
769 		if (!vf->devlink_port.devlink_rate)
770 			/* leaf nodes doesn't have children
771 			 * so we don't set rate_node
772 			 */
773 			devl_rate_leaf_create(&vf->devlink_port, node,
774 					      node->parent->rate_node);
775 	} else if (node->vsi_handle &&
776 		   pf->vsi[node->vsi_handle]->type == ICE_VSI_SF &&
777 		   pf->vsi[node->vsi_handle]->sf) {
778 		sf = pf->vsi[node->vsi_handle]->sf;
779 		if (!sf->devlink_port.devlink_rate)
780 			/* leaf nodes doesn't have children
781 			 * so we don't set rate_node
782 			 */
783 			devl_rate_leaf_create(&sf->devlink_port, node,
784 					      node->parent->rate_node);
785 	} else if (node->info.data.elem_type != ICE_AQC_ELEM_TYPE_LEAF &&
786 		   node->parent->rate_node) {
787 		rate_node = devl_rate_node_create(devlink, node, node->name,
788 						  node->parent->rate_node);
789 	}
790 
791 	if (rate_node && !IS_ERR(rate_node))
792 		node->rate_node = rate_node;
793 
794 traverse_children:
795 	for (i = 0; i < node->num_children; i++)
796 		ice_traverse_tx_tree(devlink, node->children[i], tc_node, pf);
797 }
798 
799 /**
800  * ice_devlink_rate_init_tx_topology - export Tx scheduler tree to devlink rate
801  * @devlink: devlink struct
802  * @vsi: main vsi struct
803  *
804  * This function finds a root node, then calls ice_traverse_tx tree, which
805  * traverses the tree and exports it's contents to devlink rate.
806  */
807 int ice_devlink_rate_init_tx_topology(struct devlink *devlink, struct ice_vsi *vsi)
808 {
809 	struct ice_port_info *pi = vsi->port_info;
810 	struct ice_sched_node *tc_node;
811 	struct ice_pf *pf = vsi->back;
812 	int i;
813 
814 	tc_node = pi->root->children[0];
815 	mutex_lock(&pi->sched_lock);
816 	for (i = 0; i < tc_node->num_children; i++)
817 		ice_traverse_tx_tree(devlink, tc_node->children[i], tc_node, pf);
818 	mutex_unlock(&pi->sched_lock);
819 
820 	return 0;
821 }
822 
823 static void ice_clear_rate_nodes(struct ice_sched_node *node)
824 {
825 	node->rate_node = NULL;
826 
827 	for (int i = 0; i < node->num_children; i++)
828 		ice_clear_rate_nodes(node->children[i]);
829 }
830 
831 /**
832  * ice_devlink_rate_clear_tx_topology - clear node->rate_node
833  * @vsi: main vsi struct
834  *
835  * Clear rate_node to cleanup creation of Tx topology.
836  *
837  */
838 void ice_devlink_rate_clear_tx_topology(struct ice_vsi *vsi)
839 {
840 	struct ice_port_info *pi = vsi->port_info;
841 
842 	mutex_lock(&pi->sched_lock);
843 	ice_clear_rate_nodes(pi->root->children[0]);
844 	mutex_unlock(&pi->sched_lock);
845 }
846 
847 /**
848  * ice_set_object_tx_share - sets node scheduling parameter
849  * @pi: devlink struct instance
850  * @node: node struct instance
851  * @bw: bandwidth in bytes per second
852  * @extack: extended netdev ack structure
853  *
854  * This function sets ICE_MIN_BW scheduling BW limit.
855  */
856 static int ice_set_object_tx_share(struct ice_port_info *pi, struct ice_sched_node *node,
857 				   u64 bw, struct netlink_ext_ack *extack)
858 {
859 	int status;
860 
861 	mutex_lock(&pi->sched_lock);
862 	/* converts bytes per second to kilo bits per second */
863 	node->tx_share = div_u64(bw, 125);
864 	status = ice_sched_set_node_bw_lmt(pi, node, ICE_MIN_BW, node->tx_share);
865 	mutex_unlock(&pi->sched_lock);
866 
867 	if (status)
868 		NL_SET_ERR_MSG_MOD(extack, "Can't set scheduling node tx_share");
869 
870 	return status;
871 }
872 
873 /**
874  * ice_set_object_tx_max - sets node scheduling parameter
875  * @pi: devlink struct instance
876  * @node: node struct instance
877  * @bw: bandwidth in bytes per second
878  * @extack: extended netdev ack structure
879  *
880  * This function sets ICE_MAX_BW scheduling BW limit.
881  */
882 static int ice_set_object_tx_max(struct ice_port_info *pi, struct ice_sched_node *node,
883 				 u64 bw, struct netlink_ext_ack *extack)
884 {
885 	int status;
886 
887 	mutex_lock(&pi->sched_lock);
888 	/* converts bytes per second value to kilo bits per second */
889 	node->tx_max = div_u64(bw, 125);
890 	status = ice_sched_set_node_bw_lmt(pi, node, ICE_MAX_BW, node->tx_max);
891 	mutex_unlock(&pi->sched_lock);
892 
893 	if (status)
894 		NL_SET_ERR_MSG_MOD(extack, "Can't set scheduling node tx_max");
895 
896 	return status;
897 }
898 
899 /**
900  * ice_set_object_tx_priority - sets node scheduling parameter
901  * @pi: devlink struct instance
902  * @node: node struct instance
903  * @priority: value representing priority for strict priority arbitration
904  * @extack: extended netdev ack structure
905  *
906  * This function sets priority of node among siblings.
907  */
908 static int ice_set_object_tx_priority(struct ice_port_info *pi, struct ice_sched_node *node,
909 				      u32 priority, struct netlink_ext_ack *extack)
910 {
911 	int status;
912 
913 	if (priority >= 8) {
914 		NL_SET_ERR_MSG_MOD(extack, "Priority should be less than 8");
915 		return -EINVAL;
916 	}
917 
918 	mutex_lock(&pi->sched_lock);
919 	node->tx_priority = priority;
920 	status = ice_sched_set_node_priority(pi, node, node->tx_priority);
921 	mutex_unlock(&pi->sched_lock);
922 
923 	if (status)
924 		NL_SET_ERR_MSG_MOD(extack, "Can't set scheduling node tx_priority");
925 
926 	return status;
927 }
928 
929 /**
930  * ice_set_object_tx_weight - sets node scheduling parameter
931  * @pi: devlink struct instance
932  * @node: node struct instance
933  * @weight: value represeting relative weight for WFQ arbitration
934  * @extack: extended netdev ack structure
935  *
936  * This function sets node weight for WFQ algorithm.
937  */
938 static int ice_set_object_tx_weight(struct ice_port_info *pi, struct ice_sched_node *node,
939 				    u32 weight, struct netlink_ext_ack *extack)
940 {
941 	int status;
942 
943 	if (weight > 200 || weight < 1) {
944 		NL_SET_ERR_MSG_MOD(extack, "Weight must be between 1 and 200");
945 		return -EINVAL;
946 	}
947 
948 	mutex_lock(&pi->sched_lock);
949 	node->tx_weight = weight;
950 	status = ice_sched_set_node_weight(pi, node, node->tx_weight);
951 	mutex_unlock(&pi->sched_lock);
952 
953 	if (status)
954 		NL_SET_ERR_MSG_MOD(extack, "Can't set scheduling node tx_weight");
955 
956 	return status;
957 }
958 
959 /**
960  * ice_get_pi_from_dev_rate - get port info from devlink_rate
961  * @rate_node: devlink struct instance
962  *
963  * This function returns corresponding port_info struct of devlink_rate
964  */
965 static struct ice_port_info *ice_get_pi_from_dev_rate(struct devlink_rate *rate_node)
966 {
967 	struct ice_pf *pf = devlink_priv(rate_node->devlink);
968 
969 	return ice_get_main_vsi(pf)->port_info;
970 }
971 
972 static int ice_devlink_rate_node_new(struct devlink_rate *rate_node, void **priv,
973 				     struct netlink_ext_ack *extack)
974 {
975 	struct ice_sched_node *node;
976 	struct ice_port_info *pi;
977 
978 	pi = ice_get_pi_from_dev_rate(rate_node);
979 
980 	if (!ice_enable_custom_tx(devlink_priv(rate_node->devlink)))
981 		return -EBUSY;
982 
983 	/* preallocate memory for ice_sched_node */
984 	node = devm_kzalloc(ice_hw_to_dev(pi->hw), sizeof(*node), GFP_KERNEL);
985 	if (!node)
986 		return -ENOMEM;
987 
988 	*priv = node;
989 
990 	return 0;
991 }
992 
993 static int ice_devlink_rate_node_del(struct devlink_rate *rate_node, void *priv,
994 				     struct netlink_ext_ack *extack)
995 {
996 	struct ice_sched_node *node, *tc_node;
997 	struct ice_port_info *pi;
998 
999 	pi = ice_get_pi_from_dev_rate(rate_node);
1000 	tc_node = pi->root->children[0];
1001 	node = priv;
1002 
1003 	if (!rate_node->parent || !node || tc_node == node || !extack)
1004 		return 0;
1005 
1006 	if (!ice_enable_custom_tx(devlink_priv(rate_node->devlink)))
1007 		return -EBUSY;
1008 
1009 	/* can't allow to delete a node with children */
1010 	if (node->num_children)
1011 		return -EINVAL;
1012 
1013 	mutex_lock(&pi->sched_lock);
1014 	ice_free_sched_node(pi, node);
1015 	mutex_unlock(&pi->sched_lock);
1016 
1017 	return 0;
1018 }
1019 
1020 static int ice_devlink_rate_leaf_tx_max_set(struct devlink_rate *rate_leaf, void *priv,
1021 					    u64 tx_max, struct netlink_ext_ack *extack)
1022 {
1023 	struct ice_sched_node *node = priv;
1024 
1025 	if (!ice_enable_custom_tx(devlink_priv(rate_leaf->devlink)))
1026 		return -EBUSY;
1027 
1028 	if (!node)
1029 		return 0;
1030 
1031 	return ice_set_object_tx_max(ice_get_pi_from_dev_rate(rate_leaf),
1032 				     node, tx_max, extack);
1033 }
1034 
1035 static int ice_devlink_rate_leaf_tx_share_set(struct devlink_rate *rate_leaf, void *priv,
1036 					      u64 tx_share, struct netlink_ext_ack *extack)
1037 {
1038 	struct ice_sched_node *node = priv;
1039 
1040 	if (!ice_enable_custom_tx(devlink_priv(rate_leaf->devlink)))
1041 		return -EBUSY;
1042 
1043 	if (!node)
1044 		return 0;
1045 
1046 	return ice_set_object_tx_share(ice_get_pi_from_dev_rate(rate_leaf), node,
1047 				       tx_share, extack);
1048 }
1049 
1050 static int ice_devlink_rate_leaf_tx_priority_set(struct devlink_rate *rate_leaf, void *priv,
1051 						 u32 tx_priority, struct netlink_ext_ack *extack)
1052 {
1053 	struct ice_sched_node *node = priv;
1054 
1055 	if (!ice_enable_custom_tx(devlink_priv(rate_leaf->devlink)))
1056 		return -EBUSY;
1057 
1058 	if (!node)
1059 		return 0;
1060 
1061 	return ice_set_object_tx_priority(ice_get_pi_from_dev_rate(rate_leaf), node,
1062 					  tx_priority, extack);
1063 }
1064 
1065 static int ice_devlink_rate_leaf_tx_weight_set(struct devlink_rate *rate_leaf, void *priv,
1066 					       u32 tx_weight, struct netlink_ext_ack *extack)
1067 {
1068 	struct ice_sched_node *node = priv;
1069 
1070 	if (!ice_enable_custom_tx(devlink_priv(rate_leaf->devlink)))
1071 		return -EBUSY;
1072 
1073 	if (!node)
1074 		return 0;
1075 
1076 	return ice_set_object_tx_weight(ice_get_pi_from_dev_rate(rate_leaf), node,
1077 					tx_weight, extack);
1078 }
1079 
1080 static int ice_devlink_rate_node_tx_max_set(struct devlink_rate *rate_node, void *priv,
1081 					    u64 tx_max, struct netlink_ext_ack *extack)
1082 {
1083 	struct ice_sched_node *node = priv;
1084 
1085 	if (!ice_enable_custom_tx(devlink_priv(rate_node->devlink)))
1086 		return -EBUSY;
1087 
1088 	if (!node)
1089 		return 0;
1090 
1091 	return ice_set_object_tx_max(ice_get_pi_from_dev_rate(rate_node),
1092 				     node, tx_max, extack);
1093 }
1094 
1095 static int ice_devlink_rate_node_tx_share_set(struct devlink_rate *rate_node, void *priv,
1096 					      u64 tx_share, struct netlink_ext_ack *extack)
1097 {
1098 	struct ice_sched_node *node = priv;
1099 
1100 	if (!ice_enable_custom_tx(devlink_priv(rate_node->devlink)))
1101 		return -EBUSY;
1102 
1103 	if (!node)
1104 		return 0;
1105 
1106 	return ice_set_object_tx_share(ice_get_pi_from_dev_rate(rate_node),
1107 				       node, tx_share, extack);
1108 }
1109 
1110 static int ice_devlink_rate_node_tx_priority_set(struct devlink_rate *rate_node, void *priv,
1111 						 u32 tx_priority, struct netlink_ext_ack *extack)
1112 {
1113 	struct ice_sched_node *node = priv;
1114 
1115 	if (!ice_enable_custom_tx(devlink_priv(rate_node->devlink)))
1116 		return -EBUSY;
1117 
1118 	if (!node)
1119 		return 0;
1120 
1121 	return ice_set_object_tx_priority(ice_get_pi_from_dev_rate(rate_node),
1122 					  node, tx_priority, extack);
1123 }
1124 
1125 static int ice_devlink_rate_node_tx_weight_set(struct devlink_rate *rate_node, void *priv,
1126 					       u32 tx_weight, struct netlink_ext_ack *extack)
1127 {
1128 	struct ice_sched_node *node = priv;
1129 
1130 	if (!ice_enable_custom_tx(devlink_priv(rate_node->devlink)))
1131 		return -EBUSY;
1132 
1133 	if (!node)
1134 		return 0;
1135 
1136 	return ice_set_object_tx_weight(ice_get_pi_from_dev_rate(rate_node),
1137 					node, tx_weight, extack);
1138 }
1139 
1140 static int ice_devlink_set_parent(struct devlink_rate *devlink_rate,
1141 				  struct devlink_rate *parent,
1142 				  void *priv, void *parent_priv,
1143 				  struct netlink_ext_ack *extack)
1144 {
1145 	struct ice_port_info *pi = ice_get_pi_from_dev_rate(devlink_rate);
1146 	struct ice_sched_node *tc_node, *node, *parent_node;
1147 	u16 num_nodes_added;
1148 	u32 first_node_teid;
1149 	u32 node_teid;
1150 	int status;
1151 
1152 	tc_node = pi->root->children[0];
1153 	node = priv;
1154 
1155 	if (!extack)
1156 		return 0;
1157 
1158 	if (!ice_enable_custom_tx(devlink_priv(devlink_rate->devlink)))
1159 		return -EBUSY;
1160 
1161 	if (!parent) {
1162 		if (!node || tc_node == node || node->num_children)
1163 			return -EINVAL;
1164 
1165 		mutex_lock(&pi->sched_lock);
1166 		ice_free_sched_node(pi, node);
1167 		mutex_unlock(&pi->sched_lock);
1168 
1169 		return 0;
1170 	}
1171 
1172 	parent_node = parent_priv;
1173 
1174 	/* if the node doesn't exist, create it */
1175 	if (!node->parent) {
1176 		mutex_lock(&pi->sched_lock);
1177 		status = ice_sched_add_elems(pi, tc_node, parent_node,
1178 					     parent_node->tx_sched_layer + 1,
1179 					     1, &num_nodes_added, &first_node_teid,
1180 					     &node);
1181 		mutex_unlock(&pi->sched_lock);
1182 
1183 		if (status) {
1184 			NL_SET_ERR_MSG_MOD(extack, "Can't add a new node");
1185 			return status;
1186 		}
1187 
1188 		if (devlink_rate->tx_share)
1189 			ice_set_object_tx_share(pi, node, devlink_rate->tx_share, extack);
1190 		if (devlink_rate->tx_max)
1191 			ice_set_object_tx_max(pi, node, devlink_rate->tx_max, extack);
1192 		if (devlink_rate->tx_priority)
1193 			ice_set_object_tx_priority(pi, node, devlink_rate->tx_priority, extack);
1194 		if (devlink_rate->tx_weight)
1195 			ice_set_object_tx_weight(pi, node, devlink_rate->tx_weight, extack);
1196 	} else {
1197 		node_teid = le32_to_cpu(node->info.node_teid);
1198 		mutex_lock(&pi->sched_lock);
1199 		status = ice_sched_move_nodes(pi, parent_node, 1, &node_teid);
1200 		mutex_unlock(&pi->sched_lock);
1201 
1202 		if (status)
1203 			NL_SET_ERR_MSG_MOD(extack, "Can't move existing node to a new parent");
1204 	}
1205 
1206 	return status;
1207 }
1208 
1209 static void ice_set_min_max_msix(struct ice_pf *pf)
1210 {
1211 	struct devlink *devlink = priv_to_devlink(pf);
1212 	union devlink_param_value val;
1213 	int err;
1214 
1215 	err = devl_param_driverinit_value_get(devlink,
1216 					      DEVLINK_PARAM_GENERIC_ID_MSIX_VEC_PER_PF_MIN,
1217 					      &val);
1218 	if (!err)
1219 		pf->msix.min = val.vu32;
1220 
1221 	err = devl_param_driverinit_value_get(devlink,
1222 					      DEVLINK_PARAM_GENERIC_ID_MSIX_VEC_PER_PF_MAX,
1223 					      &val);
1224 	if (!err)
1225 		pf->msix.max = val.vu32;
1226 }
1227 
1228 /**
1229  * ice_devlink_reinit_up - do reinit of the given PF
1230  * @pf: pointer to the PF struct
1231  */
1232 static int ice_devlink_reinit_up(struct ice_pf *pf)
1233 {
1234 	struct ice_vsi *vsi = ice_get_main_vsi(pf);
1235 	struct device *dev = ice_pf_to_dev(pf);
1236 	bool need_dev_deinit = false;
1237 	int err;
1238 
1239 	err = ice_init_hw(&pf->hw);
1240 	if (err) {
1241 		dev_err(dev, "ice_init_hw failed: %d\n", err);
1242 		return err;
1243 	}
1244 
1245 	/* load MSI-X values */
1246 	ice_set_min_max_msix(pf);
1247 
1248 	err = ice_init_dev(pf);
1249 	if (err)
1250 		goto unroll_hw_init;
1251 
1252 	err = ice_init_pf(pf);
1253 	if (err) {
1254 		dev_err(dev, "ice_init_pf failed: %d\n", err);
1255 		goto unroll_dev_init;
1256 	}
1257 
1258 	vsi->flags = ICE_VSI_FLAG_INIT;
1259 
1260 	rtnl_lock();
1261 	err = ice_vsi_cfg(vsi);
1262 	rtnl_unlock();
1263 	if (err)
1264 		goto unroll_pf_init;
1265 
1266 	/* No need to take devl_lock, it's already taken by devlink API */
1267 	err = ice_load(pf);
1268 	if (err)
1269 		goto err_load;
1270 
1271 	return 0;
1272 
1273 err_load:
1274 	rtnl_lock();
1275 	ice_vsi_decfg(vsi);
1276 	rtnl_unlock();
1277 unroll_pf_init:
1278 	ice_deinit_pf(pf);
1279 unroll_dev_init:
1280 	need_dev_deinit = true;
1281 unroll_hw_init:
1282 	ice_deinit_hw(&pf->hw);
1283 	if (need_dev_deinit)
1284 		ice_deinit_dev(pf);
1285 	return err;
1286 }
1287 
1288 /**
1289  * ice_devlink_reload_up - do reload up after reinit
1290  * @devlink: pointer to the devlink instance reloading
1291  * @action: the action requested
1292  * @limit: limits imposed by userspace, such as not resetting
1293  * @actions_performed: on return, indicate what actions actually performed
1294  * @extack: netlink extended ACK structure
1295  */
1296 static int
1297 ice_devlink_reload_up(struct devlink *devlink,
1298 		      enum devlink_reload_action action,
1299 		      enum devlink_reload_limit limit,
1300 		      u32 *actions_performed,
1301 		      struct netlink_ext_ack *extack)
1302 {
1303 	struct ice_pf *pf = devlink_priv(devlink);
1304 
1305 	switch (action) {
1306 	case DEVLINK_RELOAD_ACTION_DRIVER_REINIT:
1307 		*actions_performed = BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT);
1308 		return ice_devlink_reinit_up(pf);
1309 	case DEVLINK_RELOAD_ACTION_FW_ACTIVATE:
1310 		*actions_performed = BIT(DEVLINK_RELOAD_ACTION_FW_ACTIVATE);
1311 		return ice_devlink_reload_empr_finish(pf, extack);
1312 	default:
1313 		WARN_ON(1);
1314 		return -EOPNOTSUPP;
1315 	}
1316 }
1317 
1318 static const struct devlink_ops ice_devlink_ops = {
1319 	.supported_flash_update_params = DEVLINK_SUPPORT_FLASH_UPDATE_OVERWRITE_MASK,
1320 	.reload_actions = BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT) |
1321 			  BIT(DEVLINK_RELOAD_ACTION_FW_ACTIVATE),
1322 	.reload_down = ice_devlink_reload_down,
1323 	.reload_up = ice_devlink_reload_up,
1324 	.eswitch_mode_get = ice_eswitch_mode_get,
1325 	.eswitch_mode_set = ice_eswitch_mode_set,
1326 	.info_get = ice_devlink_info_get,
1327 	.flash_update = ice_devlink_flash_update,
1328 
1329 	.rate_node_new = ice_devlink_rate_node_new,
1330 	.rate_node_del = ice_devlink_rate_node_del,
1331 
1332 	.rate_leaf_tx_max_set = ice_devlink_rate_leaf_tx_max_set,
1333 	.rate_leaf_tx_share_set = ice_devlink_rate_leaf_tx_share_set,
1334 	.rate_leaf_tx_priority_set = ice_devlink_rate_leaf_tx_priority_set,
1335 	.rate_leaf_tx_weight_set = ice_devlink_rate_leaf_tx_weight_set,
1336 
1337 	.rate_node_tx_max_set = ice_devlink_rate_node_tx_max_set,
1338 	.rate_node_tx_share_set = ice_devlink_rate_node_tx_share_set,
1339 	.rate_node_tx_priority_set = ice_devlink_rate_node_tx_priority_set,
1340 	.rate_node_tx_weight_set = ice_devlink_rate_node_tx_weight_set,
1341 
1342 	.rate_leaf_parent_set = ice_devlink_set_parent,
1343 	.rate_node_parent_set = ice_devlink_set_parent,
1344 
1345 	.port_new = ice_devlink_port_new,
1346 };
1347 
1348 static const struct devlink_ops ice_sf_devlink_ops;
1349 
1350 static int
1351 ice_devlink_enable_roce_get(struct devlink *devlink, u32 id,
1352 			    struct devlink_param_gset_ctx *ctx)
1353 {
1354 	struct ice_pf *pf = devlink_priv(devlink);
1355 	struct iidc_rdma_core_dev_info *cdev;
1356 
1357 	cdev = pf->cdev_info;
1358 	if (!cdev)
1359 		return -ENODEV;
1360 
1361 	ctx->val.vbool = !!(cdev->rdma_protocol & IIDC_RDMA_PROTOCOL_ROCEV2);
1362 
1363 	return 0;
1364 }
1365 
1366 static int ice_devlink_enable_roce_set(struct devlink *devlink, u32 id,
1367 				       struct devlink_param_gset_ctx *ctx,
1368 				       struct netlink_ext_ack *extack)
1369 {
1370 	struct ice_pf *pf = devlink_priv(devlink);
1371 	struct iidc_rdma_core_dev_info *cdev;
1372 	bool roce_ena = ctx->val.vbool;
1373 	int ret;
1374 
1375 	cdev = pf->cdev_info;
1376 	if (!cdev)
1377 		return -ENODEV;
1378 
1379 	if (!roce_ena) {
1380 		ice_unplug_aux_dev(pf);
1381 		cdev->rdma_protocol &= ~IIDC_RDMA_PROTOCOL_ROCEV2;
1382 		return 0;
1383 	}
1384 
1385 	cdev->rdma_protocol |= IIDC_RDMA_PROTOCOL_ROCEV2;
1386 	ret = ice_plug_aux_dev(pf);
1387 	if (ret)
1388 		cdev->rdma_protocol &= ~IIDC_RDMA_PROTOCOL_ROCEV2;
1389 
1390 	return ret;
1391 }
1392 
1393 static int
1394 ice_devlink_enable_roce_validate(struct devlink *devlink, u32 id,
1395 				 union devlink_param_value val,
1396 				 struct netlink_ext_ack *extack)
1397 {
1398 	struct ice_pf *pf = devlink_priv(devlink);
1399 	struct iidc_rdma_core_dev_info *cdev;
1400 
1401 	cdev = pf->cdev_info;
1402 	if (!cdev)
1403 		return -ENODEV;
1404 
1405 	if (!test_bit(ICE_FLAG_RDMA_ENA, pf->flags))
1406 		return -EOPNOTSUPP;
1407 
1408 	if (cdev->rdma_protocol & IIDC_RDMA_PROTOCOL_IWARP) {
1409 		NL_SET_ERR_MSG_MOD(extack, "iWARP is currently enabled. This device cannot enable iWARP and RoCEv2 simultaneously");
1410 		return -EOPNOTSUPP;
1411 	}
1412 
1413 	return 0;
1414 }
1415 
1416 static int
1417 ice_devlink_enable_iw_get(struct devlink *devlink, u32 id,
1418 			  struct devlink_param_gset_ctx *ctx)
1419 {
1420 	struct ice_pf *pf = devlink_priv(devlink);
1421 	struct iidc_rdma_core_dev_info *cdev;
1422 
1423 	cdev = pf->cdev_info;
1424 	if (!cdev)
1425 		return -ENODEV;
1426 
1427 	ctx->val.vbool = !!(cdev->rdma_protocol & IIDC_RDMA_PROTOCOL_IWARP);
1428 
1429 	return 0;
1430 }
1431 
1432 static int ice_devlink_enable_iw_set(struct devlink *devlink, u32 id,
1433 				     struct devlink_param_gset_ctx *ctx,
1434 				     struct netlink_ext_ack *extack)
1435 {
1436 	struct ice_pf *pf = devlink_priv(devlink);
1437 	struct iidc_rdma_core_dev_info *cdev;
1438 	bool iw_ena = ctx->val.vbool;
1439 	int ret;
1440 
1441 	cdev = pf->cdev_info;
1442 	if (!cdev)
1443 		return -ENODEV;
1444 
1445 	if (!iw_ena) {
1446 		ice_unplug_aux_dev(pf);
1447 		cdev->rdma_protocol &= ~IIDC_RDMA_PROTOCOL_IWARP;
1448 		return 0;
1449 	}
1450 
1451 	cdev->rdma_protocol |= IIDC_RDMA_PROTOCOL_IWARP;
1452 	ret = ice_plug_aux_dev(pf);
1453 	if (ret)
1454 		cdev->rdma_protocol &= ~IIDC_RDMA_PROTOCOL_IWARP;
1455 
1456 	return ret;
1457 }
1458 
1459 static int
1460 ice_devlink_enable_iw_validate(struct devlink *devlink, u32 id,
1461 			       union devlink_param_value val,
1462 			       struct netlink_ext_ack *extack)
1463 {
1464 	struct ice_pf *pf = devlink_priv(devlink);
1465 
1466 	if (!test_bit(ICE_FLAG_RDMA_ENA, pf->flags))
1467 		return -EOPNOTSUPP;
1468 
1469 	if (pf->cdev_info->rdma_protocol & IIDC_RDMA_PROTOCOL_ROCEV2) {
1470 		NL_SET_ERR_MSG_MOD(extack, "RoCEv2 is currently enabled. This device cannot enable iWARP and RoCEv2 simultaneously");
1471 		return -EOPNOTSUPP;
1472 	}
1473 
1474 	return 0;
1475 }
1476 
1477 #define DEVLINK_LOCAL_FWD_DISABLED_STR "disabled"
1478 #define DEVLINK_LOCAL_FWD_ENABLED_STR "enabled"
1479 #define DEVLINK_LOCAL_FWD_PRIORITIZED_STR "prioritized"
1480 
1481 /**
1482  * ice_devlink_local_fwd_mode_to_str - Get string for local_fwd mode.
1483  * @mode: local forwarding for mode used in port_info struct.
1484  *
1485  * Return: Mode respective string or "Invalid".
1486  */
1487 static const char *
1488 ice_devlink_local_fwd_mode_to_str(enum ice_local_fwd_mode mode)
1489 {
1490 	switch (mode) {
1491 	case ICE_LOCAL_FWD_MODE_ENABLED:
1492 		return DEVLINK_LOCAL_FWD_ENABLED_STR;
1493 	case ICE_LOCAL_FWD_MODE_PRIORITIZED:
1494 		return DEVLINK_LOCAL_FWD_PRIORITIZED_STR;
1495 	case ICE_LOCAL_FWD_MODE_DISABLED:
1496 		return DEVLINK_LOCAL_FWD_DISABLED_STR;
1497 	}
1498 
1499 	return "Invalid";
1500 }
1501 
1502 /**
1503  * ice_devlink_local_fwd_str_to_mode - Get local_fwd mode from string name.
1504  * @mode_str: local forwarding mode string.
1505  *
1506  * Return: Mode value or negative number if invalid.
1507  */
1508 static int ice_devlink_local_fwd_str_to_mode(const char *mode_str)
1509 {
1510 	if (!strcmp(mode_str, DEVLINK_LOCAL_FWD_ENABLED_STR))
1511 		return ICE_LOCAL_FWD_MODE_ENABLED;
1512 	else if (!strcmp(mode_str, DEVLINK_LOCAL_FWD_PRIORITIZED_STR))
1513 		return ICE_LOCAL_FWD_MODE_PRIORITIZED;
1514 	else if (!strcmp(mode_str, DEVLINK_LOCAL_FWD_DISABLED_STR))
1515 		return ICE_LOCAL_FWD_MODE_DISABLED;
1516 
1517 	return -EINVAL;
1518 }
1519 
1520 /**
1521  * ice_devlink_local_fwd_get - Get local_fwd parameter.
1522  * @devlink: Pointer to the devlink instance.
1523  * @id: The parameter ID to set.
1524  * @ctx: Context to store the parameter value.
1525  *
1526  * Return: Zero.
1527  */
1528 static int ice_devlink_local_fwd_get(struct devlink *devlink, u32 id,
1529 				     struct devlink_param_gset_ctx *ctx)
1530 {
1531 	struct ice_pf *pf = devlink_priv(devlink);
1532 	struct ice_port_info *pi;
1533 	const char *mode_str;
1534 
1535 	pi = pf->hw.port_info;
1536 	mode_str = ice_devlink_local_fwd_mode_to_str(pi->local_fwd_mode);
1537 	snprintf(ctx->val.vstr, sizeof(ctx->val.vstr), "%s", mode_str);
1538 
1539 	return 0;
1540 }
1541 
1542 /**
1543  * ice_devlink_local_fwd_set - Set local_fwd parameter.
1544  * @devlink: Pointer to the devlink instance.
1545  * @id: The parameter ID to set.
1546  * @ctx: Context to get the parameter value.
1547  * @extack: Netlink extended ACK structure.
1548  *
1549  * Return: Zero.
1550  */
1551 static int ice_devlink_local_fwd_set(struct devlink *devlink, u32 id,
1552 				     struct devlink_param_gset_ctx *ctx,
1553 				     struct netlink_ext_ack *extack)
1554 {
1555 	int new_local_fwd_mode = ice_devlink_local_fwd_str_to_mode(ctx->val.vstr);
1556 	struct ice_pf *pf = devlink_priv(devlink);
1557 	struct device *dev = ice_pf_to_dev(pf);
1558 	struct ice_port_info *pi;
1559 
1560 	pi = pf->hw.port_info;
1561 	if (pi->local_fwd_mode != new_local_fwd_mode) {
1562 		pi->local_fwd_mode = new_local_fwd_mode;
1563 		dev_info(dev, "Setting local_fwd to %s\n", ctx->val.vstr);
1564 		ice_schedule_reset(pf, ICE_RESET_CORER);
1565 	}
1566 
1567 	return 0;
1568 }
1569 
1570 /**
1571  * ice_devlink_local_fwd_validate - Validate passed local_fwd parameter value.
1572  * @devlink: Unused pointer to devlink instance.
1573  * @id: The parameter ID to validate.
1574  * @val: Value to validate.
1575  * @extack: Netlink extended ACK structure.
1576  *
1577  * Supported values are:
1578  * "enabled" - local_fwd is enabled, "disabled" - local_fwd is disabled
1579  * "prioritized" - local_fwd traffic is prioritized in scheduling.
1580  *
1581  * Return: Zero when passed parameter value is supported. Negative value on
1582  * error.
1583  */
1584 static int ice_devlink_local_fwd_validate(struct devlink *devlink, u32 id,
1585 					  union devlink_param_value val,
1586 					  struct netlink_ext_ack *extack)
1587 {
1588 	if (ice_devlink_local_fwd_str_to_mode(val.vstr) < 0) {
1589 		NL_SET_ERR_MSG_MOD(extack, "Error: Requested value is not supported.");
1590 		return -EINVAL;
1591 	}
1592 
1593 	return 0;
1594 }
1595 
1596 static int
1597 ice_devlink_msix_max_pf_validate(struct devlink *devlink, u32 id,
1598 				 union devlink_param_value val,
1599 				 struct netlink_ext_ack *extack)
1600 {
1601 	struct ice_pf *pf = devlink_priv(devlink);
1602 
1603 	if (val.vu32 > pf->hw.func_caps.common_cap.num_msix_vectors)
1604 		return -EINVAL;
1605 
1606 	return 0;
1607 }
1608 
1609 static int
1610 ice_devlink_msix_min_pf_validate(struct devlink *devlink, u32 id,
1611 				 union devlink_param_value val,
1612 				 struct netlink_ext_ack *extack)
1613 {
1614 	if (val.vu32 < ICE_MIN_MSIX)
1615 		return -EINVAL;
1616 
1617 	return 0;
1618 }
1619 
1620 static int ice_devlink_enable_rdma_validate(struct devlink *devlink, u32 id,
1621 					    union devlink_param_value val,
1622 					    struct netlink_ext_ack *extack)
1623 {
1624 	struct ice_pf *pf = devlink_priv(devlink);
1625 	bool new_state = val.vbool;
1626 
1627 	if (new_state && !test_bit(ICE_FLAG_RDMA_ENA, pf->flags))
1628 		return -EOPNOTSUPP;
1629 
1630 	return 0;
1631 }
1632 
1633 enum ice_param_id {
1634 	ICE_DEVLINK_PARAM_ID_BASE = DEVLINK_PARAM_GENERIC_ID_MAX,
1635 	ICE_DEVLINK_PARAM_ID_TX_SCHED_LAYERS,
1636 	ICE_DEVLINK_PARAM_ID_LOCAL_FWD,
1637 };
1638 
1639 static const struct devlink_param ice_dvl_rdma_params[] = {
1640 	DEVLINK_PARAM_GENERIC(ENABLE_ROCE, BIT(DEVLINK_PARAM_CMODE_RUNTIME),
1641 			      ice_devlink_enable_roce_get,
1642 			      ice_devlink_enable_roce_set,
1643 			      ice_devlink_enable_roce_validate),
1644 	DEVLINK_PARAM_GENERIC(ENABLE_IWARP, BIT(DEVLINK_PARAM_CMODE_RUNTIME),
1645 			      ice_devlink_enable_iw_get,
1646 			      ice_devlink_enable_iw_set,
1647 			      ice_devlink_enable_iw_validate),
1648 	DEVLINK_PARAM_GENERIC(ENABLE_RDMA, BIT(DEVLINK_PARAM_CMODE_DRIVERINIT),
1649 			      NULL, NULL, ice_devlink_enable_rdma_validate),
1650 };
1651 
1652 static const struct devlink_param ice_dvl_msix_params[] = {
1653 	DEVLINK_PARAM_GENERIC(MSIX_VEC_PER_PF_MAX,
1654 			      BIT(DEVLINK_PARAM_CMODE_DRIVERINIT),
1655 			      NULL, NULL, ice_devlink_msix_max_pf_validate),
1656 	DEVLINK_PARAM_GENERIC(MSIX_VEC_PER_PF_MIN,
1657 			      BIT(DEVLINK_PARAM_CMODE_DRIVERINIT),
1658 			      NULL, NULL, ice_devlink_msix_min_pf_validate),
1659 };
1660 
1661 static const struct devlink_param ice_dvl_sched_params[] = {
1662 	DEVLINK_PARAM_DRIVER(ICE_DEVLINK_PARAM_ID_TX_SCHED_LAYERS,
1663 			     "tx_scheduling_layers",
1664 			     DEVLINK_PARAM_TYPE_U8,
1665 			     BIT(DEVLINK_PARAM_CMODE_PERMANENT),
1666 			     ice_devlink_tx_sched_layers_get,
1667 			     ice_devlink_tx_sched_layers_set,
1668 			     ice_devlink_tx_sched_layers_validate),
1669 	DEVLINK_PARAM_DRIVER(ICE_DEVLINK_PARAM_ID_LOCAL_FWD,
1670 			     "local_forwarding", DEVLINK_PARAM_TYPE_STRING,
1671 			     BIT(DEVLINK_PARAM_CMODE_RUNTIME),
1672 			     ice_devlink_local_fwd_get,
1673 			     ice_devlink_local_fwd_set,
1674 			     ice_devlink_local_fwd_validate),
1675 };
1676 
1677 static void ice_devlink_free(void *devlink_ptr)
1678 {
1679 	devlink_free((struct devlink *)devlink_ptr);
1680 }
1681 
1682 /**
1683  * ice_allocate_pf - Allocate devlink and return PF structure pointer
1684  * @dev: the device to allocate for
1685  *
1686  * Allocate a devlink instance for this device and return the private area as
1687  * the PF structure. The devlink memory is kept track of through devres by
1688  * adding an action to remove it when unwinding.
1689  */
1690 struct ice_pf *ice_allocate_pf(struct device *dev)
1691 {
1692 	struct devlink *devlink;
1693 
1694 	devlink = devlink_alloc(&ice_devlink_ops, sizeof(struct ice_pf), dev);
1695 	if (!devlink)
1696 		return NULL;
1697 
1698 	/* Add an action to teardown the devlink when unwinding the driver */
1699 	if (devm_add_action_or_reset(dev, ice_devlink_free, devlink))
1700 		return NULL;
1701 
1702 	return devlink_priv(devlink);
1703 }
1704 
1705 /**
1706  * ice_allocate_sf - Allocate devlink and return SF structure pointer
1707  * @dev: the device to allocate for
1708  * @pf: pointer to the PF structure
1709  *
1710  * Allocate a devlink instance for SF.
1711  *
1712  * Return: ice_sf_priv pointer to allocated memory or ERR_PTR in case of error
1713  */
1714 struct ice_sf_priv *ice_allocate_sf(struct device *dev, struct ice_pf *pf)
1715 {
1716 	struct devlink *devlink;
1717 	int err;
1718 
1719 	devlink = devlink_alloc(&ice_sf_devlink_ops, sizeof(struct ice_sf_priv),
1720 				dev);
1721 	if (!devlink)
1722 		return ERR_PTR(-ENOMEM);
1723 
1724 	err = devl_nested_devlink_set(priv_to_devlink(pf), devlink);
1725 	if (err) {
1726 		devlink_free(devlink);
1727 		return ERR_PTR(err);
1728 	}
1729 
1730 	return devlink_priv(devlink);
1731 }
1732 
1733 /**
1734  * ice_devlink_register - Register devlink interface for this PF
1735  * @pf: the PF to register the devlink for.
1736  *
1737  * Register the devlink instance associated with this physical function.
1738  *
1739  * Return: zero on success or an error code on failure.
1740  */
1741 void ice_devlink_register(struct ice_pf *pf)
1742 {
1743 	struct devlink *devlink = priv_to_devlink(pf);
1744 
1745 	devl_register(devlink);
1746 }
1747 
1748 /**
1749  * ice_devlink_unregister - Unregister devlink resources for this PF.
1750  * @pf: the PF structure to cleanup
1751  *
1752  * Releases resources used by devlink and cleans up associated memory.
1753  */
1754 void ice_devlink_unregister(struct ice_pf *pf)
1755 {
1756 	devl_unregister(priv_to_devlink(pf));
1757 }
1758 
1759 int ice_devlink_register_params(struct ice_pf *pf)
1760 {
1761 	struct devlink *devlink = priv_to_devlink(pf);
1762 	union devlink_param_value value;
1763 	struct ice_hw *hw = &pf->hw;
1764 	int status;
1765 
1766 	status = devl_params_register(devlink, ice_dvl_rdma_params,
1767 				      ARRAY_SIZE(ice_dvl_rdma_params));
1768 	if (status)
1769 		return status;
1770 
1771 	status = devl_params_register(devlink, ice_dvl_msix_params,
1772 				      ARRAY_SIZE(ice_dvl_msix_params));
1773 	if (status)
1774 		goto unregister_rdma_params;
1775 
1776 	if (hw->func_caps.common_cap.tx_sched_topo_comp_mode_en)
1777 		status = devl_params_register(devlink, ice_dvl_sched_params,
1778 					      ARRAY_SIZE(ice_dvl_sched_params));
1779 	if (status)
1780 		goto unregister_msix_params;
1781 
1782 	value.vu32 = pf->msix.max;
1783 	devl_param_driverinit_value_set(devlink,
1784 					DEVLINK_PARAM_GENERIC_ID_MSIX_VEC_PER_PF_MAX,
1785 					value);
1786 	value.vu32 = pf->msix.min;
1787 	devl_param_driverinit_value_set(devlink,
1788 					DEVLINK_PARAM_GENERIC_ID_MSIX_VEC_PER_PF_MIN,
1789 					value);
1790 
1791 	value.vbool = test_bit(ICE_FLAG_RDMA_ENA, pf->flags);
1792 	devl_param_driverinit_value_set(devlink,
1793 					DEVLINK_PARAM_GENERIC_ID_ENABLE_RDMA,
1794 					value);
1795 
1796 	return 0;
1797 
1798 unregister_msix_params:
1799 	devl_params_unregister(devlink, ice_dvl_msix_params,
1800 			       ARRAY_SIZE(ice_dvl_msix_params));
1801 unregister_rdma_params:
1802 	devl_params_unregister(devlink, ice_dvl_rdma_params,
1803 			       ARRAY_SIZE(ice_dvl_rdma_params));
1804 	return status;
1805 }
1806 
1807 void ice_devlink_unregister_params(struct ice_pf *pf)
1808 {
1809 	struct devlink *devlink = priv_to_devlink(pf);
1810 	struct ice_hw *hw = &pf->hw;
1811 
1812 	devl_params_unregister(devlink, ice_dvl_rdma_params,
1813 			       ARRAY_SIZE(ice_dvl_rdma_params));
1814 	devl_params_unregister(devlink, ice_dvl_msix_params,
1815 			       ARRAY_SIZE(ice_dvl_msix_params));
1816 
1817 	if (hw->func_caps.common_cap.tx_sched_topo_comp_mode_en)
1818 		devl_params_unregister(devlink, ice_dvl_sched_params,
1819 				       ARRAY_SIZE(ice_dvl_sched_params));
1820 }
1821 
1822 #define ICE_DEVLINK_READ_BLK_SIZE (1024 * 1024)
1823 
1824 static const struct devlink_region_ops ice_nvm_region_ops;
1825 static const struct devlink_region_ops ice_sram_region_ops;
1826 
1827 /**
1828  * ice_devlink_nvm_snapshot - Capture a snapshot of the NVM flash contents
1829  * @devlink: the devlink instance
1830  * @ops: the devlink region to snapshot
1831  * @extack: extended ACK response structure
1832  * @data: on exit points to snapshot data buffer
1833  *
1834  * This function is called in response to a DEVLINK_CMD_REGION_NEW for either
1835  * the nvm-flash or shadow-ram region.
1836  *
1837  * It captures a snapshot of the NVM or Shadow RAM flash contents. This
1838  * snapshot can then later be viewed via the DEVLINK_CMD_REGION_READ netlink
1839  * interface.
1840  *
1841  * @returns zero on success, and updates the data pointer. Returns a non-zero
1842  * error code on failure.
1843  */
1844 static int ice_devlink_nvm_snapshot(struct devlink *devlink,
1845 				    const struct devlink_region_ops *ops,
1846 				    struct netlink_ext_ack *extack, u8 **data)
1847 {
1848 	struct ice_pf *pf = devlink_priv(devlink);
1849 	struct device *dev = ice_pf_to_dev(pf);
1850 	struct ice_hw *hw = &pf->hw;
1851 	bool read_shadow_ram;
1852 	u8 *nvm_data, *tmp, i;
1853 	u32 nvm_size, left;
1854 	s8 num_blks;
1855 	int status;
1856 
1857 	if (ops == &ice_nvm_region_ops) {
1858 		read_shadow_ram = false;
1859 		nvm_size = hw->flash.flash_size;
1860 	} else if (ops == &ice_sram_region_ops) {
1861 		read_shadow_ram = true;
1862 		nvm_size = hw->flash.sr_words * 2u;
1863 	} else {
1864 		NL_SET_ERR_MSG_MOD(extack, "Unexpected region in snapshot function");
1865 		return -EOPNOTSUPP;
1866 	}
1867 
1868 	nvm_data = vzalloc(nvm_size);
1869 	if (!nvm_data)
1870 		return -ENOMEM;
1871 
1872 	num_blks = DIV_ROUND_UP(nvm_size, ICE_DEVLINK_READ_BLK_SIZE);
1873 	tmp = nvm_data;
1874 	left = nvm_size;
1875 
1876 	/* Some systems take longer to read the NVM than others which causes the
1877 	 * FW to reclaim the NVM lock before the entire NVM has been read. Fix
1878 	 * this by breaking the reads of the NVM into smaller chunks that will
1879 	 * probably not take as long. This has some overhead since we are
1880 	 * increasing the number of AQ commands, but it should always work
1881 	 */
1882 	for (i = 0; i < num_blks; i++) {
1883 		u32 read_sz = min_t(u32, ICE_DEVLINK_READ_BLK_SIZE, left);
1884 
1885 		status = ice_acquire_nvm(hw, ICE_RES_READ);
1886 		if (status) {
1887 			dev_dbg(dev, "ice_acquire_nvm failed, err %d aq_err %d\n",
1888 				status, hw->adminq.sq_last_status);
1889 			NL_SET_ERR_MSG_MOD(extack, "Failed to acquire NVM semaphore");
1890 			vfree(nvm_data);
1891 			return -EIO;
1892 		}
1893 
1894 		status = ice_read_flat_nvm(hw, i * ICE_DEVLINK_READ_BLK_SIZE,
1895 					   &read_sz, tmp, read_shadow_ram);
1896 		if (status) {
1897 			dev_dbg(dev, "ice_read_flat_nvm failed after reading %u bytes, err %d aq_err %d\n",
1898 				read_sz, status, hw->adminq.sq_last_status);
1899 			NL_SET_ERR_MSG_MOD(extack, "Failed to read NVM contents");
1900 			ice_release_nvm(hw);
1901 			vfree(nvm_data);
1902 			return -EIO;
1903 		}
1904 		ice_release_nvm(hw);
1905 
1906 		tmp += read_sz;
1907 		left -= read_sz;
1908 	}
1909 
1910 	*data = nvm_data;
1911 
1912 	return 0;
1913 }
1914 
1915 /**
1916  * ice_devlink_nvm_read - Read a portion of NVM flash contents
1917  * @devlink: the devlink instance
1918  * @ops: the devlink region to snapshot
1919  * @extack: extended ACK response structure
1920  * @offset: the offset to start at
1921  * @size: the amount to read
1922  * @data: the data buffer to read into
1923  *
1924  * This function is called in response to DEVLINK_CMD_REGION_READ to directly
1925  * read a section of the NVM contents.
1926  *
1927  * It reads from either the nvm-flash or shadow-ram region contents.
1928  *
1929  * @returns zero on success, and updates the data pointer. Returns a non-zero
1930  * error code on failure.
1931  */
1932 static int ice_devlink_nvm_read(struct devlink *devlink,
1933 				const struct devlink_region_ops *ops,
1934 				struct netlink_ext_ack *extack,
1935 				u64 offset, u32 size, u8 *data)
1936 {
1937 	struct ice_pf *pf = devlink_priv(devlink);
1938 	struct device *dev = ice_pf_to_dev(pf);
1939 	struct ice_hw *hw = &pf->hw;
1940 	bool read_shadow_ram;
1941 	u64 nvm_size;
1942 	int status;
1943 
1944 	if (ops == &ice_nvm_region_ops) {
1945 		read_shadow_ram = false;
1946 		nvm_size = hw->flash.flash_size;
1947 	} else if (ops == &ice_sram_region_ops) {
1948 		read_shadow_ram = true;
1949 		nvm_size = hw->flash.sr_words * 2u;
1950 	} else {
1951 		NL_SET_ERR_MSG_MOD(extack, "Unexpected region in snapshot function");
1952 		return -EOPNOTSUPP;
1953 	}
1954 
1955 	if (offset + size >= nvm_size) {
1956 		NL_SET_ERR_MSG_MOD(extack, "Cannot read beyond the region size");
1957 		return -ERANGE;
1958 	}
1959 
1960 	status = ice_acquire_nvm(hw, ICE_RES_READ);
1961 	if (status) {
1962 		dev_dbg(dev, "ice_acquire_nvm failed, err %d aq_err %d\n",
1963 			status, hw->adminq.sq_last_status);
1964 		NL_SET_ERR_MSG_MOD(extack, "Failed to acquire NVM semaphore");
1965 		return -EIO;
1966 	}
1967 
1968 	status = ice_read_flat_nvm(hw, (u32)offset, &size, data,
1969 				   read_shadow_ram);
1970 	if (status) {
1971 		dev_dbg(dev, "ice_read_flat_nvm failed after reading %u bytes, err %d aq_err %d\n",
1972 			size, status, hw->adminq.sq_last_status);
1973 		NL_SET_ERR_MSG_MOD(extack, "Failed to read NVM contents");
1974 		ice_release_nvm(hw);
1975 		return -EIO;
1976 	}
1977 	ice_release_nvm(hw);
1978 
1979 	return 0;
1980 }
1981 
1982 /**
1983  * ice_devlink_devcaps_snapshot - Capture snapshot of device capabilities
1984  * @devlink: the devlink instance
1985  * @ops: the devlink region being snapshotted
1986  * @extack: extended ACK response structure
1987  * @data: on exit points to snapshot data buffer
1988  *
1989  * This function is called in response to the DEVLINK_CMD_REGION_TRIGGER for
1990  * the device-caps devlink region. It captures a snapshot of the device
1991  * capabilities reported by firmware.
1992  *
1993  * @returns zero on success, and updates the data pointer. Returns a non-zero
1994  * error code on failure.
1995  */
1996 static int
1997 ice_devlink_devcaps_snapshot(struct devlink *devlink,
1998 			     const struct devlink_region_ops *ops,
1999 			     struct netlink_ext_ack *extack, u8 **data)
2000 {
2001 	struct ice_pf *pf = devlink_priv(devlink);
2002 	struct device *dev = ice_pf_to_dev(pf);
2003 	struct ice_hw *hw = &pf->hw;
2004 	void *devcaps;
2005 	int status;
2006 
2007 	devcaps = vzalloc(ICE_AQ_MAX_BUF_LEN);
2008 	if (!devcaps)
2009 		return -ENOMEM;
2010 
2011 	status = ice_aq_list_caps(hw, devcaps, ICE_AQ_MAX_BUF_LEN, NULL,
2012 				  ice_aqc_opc_list_dev_caps, NULL);
2013 	if (status) {
2014 		dev_dbg(dev, "ice_aq_list_caps: failed to read device capabilities, err %d aq_err %d\n",
2015 			status, hw->adminq.sq_last_status);
2016 		NL_SET_ERR_MSG_MOD(extack, "Failed to read device capabilities");
2017 		vfree(devcaps);
2018 		return status;
2019 	}
2020 
2021 	*data = (u8 *)devcaps;
2022 
2023 	return 0;
2024 }
2025 
2026 static const struct devlink_region_ops ice_nvm_region_ops = {
2027 	.name = "nvm-flash",
2028 	.destructor = vfree,
2029 	.snapshot = ice_devlink_nvm_snapshot,
2030 	.read = ice_devlink_nvm_read,
2031 };
2032 
2033 static const struct devlink_region_ops ice_sram_region_ops = {
2034 	.name = "shadow-ram",
2035 	.destructor = vfree,
2036 	.snapshot = ice_devlink_nvm_snapshot,
2037 	.read = ice_devlink_nvm_read,
2038 };
2039 
2040 static const struct devlink_region_ops ice_devcaps_region_ops = {
2041 	.name = "device-caps",
2042 	.destructor = vfree,
2043 	.snapshot = ice_devlink_devcaps_snapshot,
2044 };
2045 
2046 /**
2047  * ice_devlink_init_regions - Initialize devlink regions
2048  * @pf: the PF device structure
2049  *
2050  * Create devlink regions used to enable access to dump the contents of the
2051  * flash memory on the device.
2052  */
2053 void ice_devlink_init_regions(struct ice_pf *pf)
2054 {
2055 	struct devlink *devlink = priv_to_devlink(pf);
2056 	struct device *dev = ice_pf_to_dev(pf);
2057 	u64 nvm_size, sram_size;
2058 
2059 	nvm_size = pf->hw.flash.flash_size;
2060 	pf->nvm_region = devl_region_create(devlink, &ice_nvm_region_ops, 1,
2061 					    nvm_size);
2062 	if (IS_ERR(pf->nvm_region)) {
2063 		dev_err(dev, "failed to create NVM devlink region, err %ld\n",
2064 			PTR_ERR(pf->nvm_region));
2065 		pf->nvm_region = NULL;
2066 	}
2067 
2068 	sram_size = pf->hw.flash.sr_words * 2u;
2069 	pf->sram_region = devl_region_create(devlink, &ice_sram_region_ops,
2070 					     1, sram_size);
2071 	if (IS_ERR(pf->sram_region)) {
2072 		dev_err(dev, "failed to create shadow-ram devlink region, err %ld\n",
2073 			PTR_ERR(pf->sram_region));
2074 		pf->sram_region = NULL;
2075 	}
2076 
2077 	pf->devcaps_region = devl_region_create(devlink,
2078 						&ice_devcaps_region_ops, 10,
2079 						ICE_AQ_MAX_BUF_LEN);
2080 	if (IS_ERR(pf->devcaps_region)) {
2081 		dev_err(dev, "failed to create device-caps devlink region, err %ld\n",
2082 			PTR_ERR(pf->devcaps_region));
2083 		pf->devcaps_region = NULL;
2084 	}
2085 }
2086 
2087 /**
2088  * ice_devlink_destroy_regions - Destroy devlink regions
2089  * @pf: the PF device structure
2090  *
2091  * Remove previously created regions for this PF.
2092  */
2093 void ice_devlink_destroy_regions(struct ice_pf *pf)
2094 {
2095 	if (pf->nvm_region)
2096 		devl_region_destroy(pf->nvm_region);
2097 
2098 	if (pf->sram_region)
2099 		devl_region_destroy(pf->sram_region);
2100 
2101 	if (pf->devcaps_region)
2102 		devl_region_destroy(pf->devcaps_region);
2103 }
2104