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 "ice_devlink.h"
   9#include "ice_eswitch.h"
  10#include "ice_fw_update.h"
  11#include "ice_dcb_lib.h"
  12
  13static int ice_active_port_option = -1;
  14
  15/* context for devlink info version reporting */
  16struct 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
  34static 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
  44static 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
  56static 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
  64static 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
  72static 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
  79static 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
  87static void
  88ice_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
  98static 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
 105static void
 106ice_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
 116static 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
 123static void
 124ice_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
 132static 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
 139static void
 140ice_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
 148static void
 149ice_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
 154static 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
 165static 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
 172static void
 173ice_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
 186static void
 187ice_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
 196static 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
 207static 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
 232enum ice_version_type {
 233	ICE_VERSION_FIXED,
 234	ICE_VERSION_RUNNING,
 235	ICE_VERSION_STORED,
 236};
 237
 238static 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 */
 271static 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, ice_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, ice_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, ice_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, ice_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(req, key, ctx->buf);
 372			if (err) {
 373				NL_SET_ERR_MSG_MOD(extack, "Unable to set running version");
 374				goto out_free_ctx;
 375			}
 376			break;
 377		case ICE_VERSION_STORED:
 378			err = devlink_info_version_stored_put(req, key, ctx->buf);
 379			if (err) {
 380				NL_SET_ERR_MSG_MOD(extack, "Unable to set stored version");
 381				goto out_free_ctx;
 382			}
 383			break;
 384		}
 385	}
 386
 387out_free_ctx:
 388	kfree(ctx);
 389	return err;
 390}
 391
 392/**
 393 * ice_devlink_reload_empr_start - Start EMP reset to activate new firmware
 394 * @pf: pointer to the pf instance
 395 * @extack: netlink extended ACK structure
 396 *
 397 * Allow user to activate new Embedded Management Processor firmware by
 398 * issuing device specific EMP reset. Called in response to
 399 * a DEVLINK_CMD_RELOAD with the DEVLINK_RELOAD_ACTION_FW_ACTIVATE.
 400 *
 401 * Note that teardown and rebuild of the driver state happens automatically as
 402 * part of an interrupt and watchdog task. This is because all physical
 403 * functions on the device must be able to reset when an EMP reset occurs from
 404 * any source.
 405 */
 406static int
 407ice_devlink_reload_empr_start(struct ice_pf *pf,
 408			      struct netlink_ext_ack *extack)
 409{
 410	struct device *dev = ice_pf_to_dev(pf);
 411	struct ice_hw *hw = &pf->hw;
 412	u8 pending;
 413	int err;
 414
 415	err = ice_get_pending_updates(pf, &pending, extack);
 416	if (err)
 417		return err;
 418
 419	/* pending is a bitmask of which flash banks have a pending update,
 420	 * including the main NVM bank, the Option ROM bank, and the netlist
 421	 * bank. If any of these bits are set, then there is a pending update
 422	 * waiting to be activated.
 423	 */
 424	if (!pending) {
 425		NL_SET_ERR_MSG_MOD(extack, "No pending firmware update");
 426		return -ECANCELED;
 427	}
 428
 429	if (pf->fw_emp_reset_disabled) {
 430		NL_SET_ERR_MSG_MOD(extack, "EMP reset is not available. To activate firmware, a reboot or power cycle is needed");
 431		return -ECANCELED;
 432	}
 433
 434	dev_dbg(dev, "Issuing device EMP reset to activate firmware\n");
 435
 436	err = ice_aq_nvm_update_empr(hw);
 437	if (err) {
 438		dev_err(dev, "Failed to trigger EMP device reset to reload firmware, err %d aq_err %s\n",
 439			err, ice_aq_str(hw->adminq.sq_last_status));
 440		NL_SET_ERR_MSG_MOD(extack, "Failed to trigger EMP device reset to reload firmware");
 441		return err;
 442	}
 443
 444	return 0;
 445}
 446
 447/**
 448 * ice_devlink_reinit_down - unload given PF
 449 * @pf: pointer to the PF struct
 450 */
 451static void ice_devlink_reinit_down(struct ice_pf *pf)
 452{
 453	/* No need to take devl_lock, it's already taken by devlink API */
 454	ice_unload(pf);
 455	rtnl_lock();
 456	ice_vsi_decfg(ice_get_main_vsi(pf));
 457	rtnl_unlock();
 458	ice_deinit_dev(pf);
 459}
 460
 461/**
 462 * ice_devlink_reload_down - prepare for reload
 463 * @devlink: pointer to the devlink instance to reload
 464 * @netns_change: if true, the network namespace is changing
 465 * @action: the action to perform
 466 * @limit: limits on what reload should do, such as not resetting
 467 * @extack: netlink extended ACK structure
 468 */
 469static int
 470ice_devlink_reload_down(struct devlink *devlink, bool netns_change,
 471			enum devlink_reload_action action,
 472			enum devlink_reload_limit limit,
 473			struct netlink_ext_ack *extack)
 474{
 475	struct ice_pf *pf = devlink_priv(devlink);
 476
 477	switch (action) {
 478	case DEVLINK_RELOAD_ACTION_DRIVER_REINIT:
 479		if (ice_is_eswitch_mode_switchdev(pf)) {
 480			NL_SET_ERR_MSG_MOD(extack,
 481					   "Go to legacy mode before doing reinit\n");
 482			return -EOPNOTSUPP;
 483		}
 484		if (ice_is_adq_active(pf)) {
 485			NL_SET_ERR_MSG_MOD(extack,
 486					   "Turn off ADQ before doing reinit\n");
 487			return -EOPNOTSUPP;
 488		}
 489		if (ice_has_vfs(pf)) {
 490			NL_SET_ERR_MSG_MOD(extack,
 491					   "Remove all VFs before doing reinit\n");
 492			return -EOPNOTSUPP;
 493		}
 494		ice_devlink_reinit_down(pf);
 495		return 0;
 496	case DEVLINK_RELOAD_ACTION_FW_ACTIVATE:
 497		return ice_devlink_reload_empr_start(pf, extack);
 498	default:
 499		WARN_ON(1);
 500		return -EOPNOTSUPP;
 501	}
 502}
 503
 504/**
 505 * ice_devlink_reload_empr_finish - Wait for EMP reset to finish
 506 * @pf: pointer to the pf instance
 507 * @extack: netlink extended ACK structure
 508 *
 509 * Wait for driver to finish rebuilding after EMP reset is completed. This
 510 * includes time to wait for both the actual device reset as well as the time
 511 * for the driver's rebuild to complete.
 512 */
 513static int
 514ice_devlink_reload_empr_finish(struct ice_pf *pf,
 515			       struct netlink_ext_ack *extack)
 516{
 517	int err;
 518
 519	err = ice_wait_for_reset(pf, 60 * HZ);
 520	if (err) {
 521		NL_SET_ERR_MSG_MOD(extack, "Device still resetting after 1 minute");
 522		return err;
 523	}
 524
 525	return 0;
 526}
 527
 528/**
 529 * ice_devlink_port_opt_speed_str - convert speed to a string
 530 * @speed: speed value
 531 */
 532static const char *ice_devlink_port_opt_speed_str(u8 speed)
 533{
 534	switch (speed & ICE_AQC_PORT_OPT_MAX_LANE_M) {
 535	case ICE_AQC_PORT_OPT_MAX_LANE_100M:
 536		return "0.1";
 537	case ICE_AQC_PORT_OPT_MAX_LANE_1G:
 538		return "1";
 539	case ICE_AQC_PORT_OPT_MAX_LANE_2500M:
 540		return "2.5";
 541	case ICE_AQC_PORT_OPT_MAX_LANE_5G:
 542		return "5";
 543	case ICE_AQC_PORT_OPT_MAX_LANE_10G:
 544		return "10";
 545	case ICE_AQC_PORT_OPT_MAX_LANE_25G:
 546		return "25";
 547	case ICE_AQC_PORT_OPT_MAX_LANE_50G:
 548		return "50";
 549	case ICE_AQC_PORT_OPT_MAX_LANE_100G:
 550		return "100";
 551	}
 552
 553	return "-";
 554}
 555
 556#define ICE_PORT_OPT_DESC_LEN	50
 557/**
 558 * ice_devlink_port_options_print - Print available port split options
 559 * @pf: the PF to print split port options
 560 *
 561 * Prints a table with available port split options and max port speeds
 562 */
 563static void ice_devlink_port_options_print(struct ice_pf *pf)
 564{
 565	u8 i, j, options_count, cnt, speed, pending_idx, active_idx;
 566	struct ice_aqc_get_port_options_elem *options, *opt;
 567	struct device *dev = ice_pf_to_dev(pf);
 568	bool active_valid, pending_valid;
 569	char desc[ICE_PORT_OPT_DESC_LEN];
 570	const char *str;
 571	int status;
 572
 573	options = kcalloc(ICE_AQC_PORT_OPT_MAX * ICE_MAX_PORT_PER_PCI_DEV,
 574			  sizeof(*options), GFP_KERNEL);
 575	if (!options)
 576		return;
 577
 578	for (i = 0; i < ICE_MAX_PORT_PER_PCI_DEV; i++) {
 579		opt = options + i * ICE_AQC_PORT_OPT_MAX;
 580		options_count = ICE_AQC_PORT_OPT_MAX;
 581		active_valid = 0;
 582
 583		status = ice_aq_get_port_options(&pf->hw, opt, &options_count,
 584						 i, true, &active_idx,
 585						 &active_valid, &pending_idx,
 586						 &pending_valid);
 587		if (status) {
 588			dev_dbg(dev, "Couldn't read port option for port %d, err %d\n",
 589				i, status);
 590			goto err;
 591		}
 592	}
 593
 594	dev_dbg(dev, "Available port split options and max port speeds (Gbps):\n");
 595	dev_dbg(dev, "Status  Split      Quad 0          Quad 1\n");
 596	dev_dbg(dev, "        count  L0  L1  L2  L3  L4  L5  L6  L7\n");
 597
 598	for (i = 0; i < options_count; i++) {
 599		cnt = 0;
 600
 601		if (i == ice_active_port_option)
 602			str = "Active";
 603		else if ((i == pending_idx) && pending_valid)
 604			str = "Pending";
 605		else
 606			str = "";
 607
 608		cnt += snprintf(&desc[cnt], ICE_PORT_OPT_DESC_LEN - cnt,
 609				"%-8s", str);
 610
 611		cnt += snprintf(&desc[cnt], ICE_PORT_OPT_DESC_LEN - cnt,
 612				"%-6u", options[i].pmd);
 613
 614		for (j = 0; j < ICE_MAX_PORT_PER_PCI_DEV; ++j) {
 615			speed = options[i + j * ICE_AQC_PORT_OPT_MAX].max_lane_speed;
 616			str = ice_devlink_port_opt_speed_str(speed);
 617			cnt += snprintf(&desc[cnt], ICE_PORT_OPT_DESC_LEN - cnt,
 618					"%3s ", str);
 619		}
 620
 621		dev_dbg(dev, "%s\n", desc);
 622	}
 623
 624err:
 625	kfree(options);
 626}
 627
 628/**
 629 * ice_devlink_aq_set_port_option - Send set port option admin queue command
 630 * @pf: the PF to print split port options
 631 * @option_idx: selected port option
 632 * @extack: extended netdev ack structure
 633 *
 634 * Sends set port option admin queue command with selected port option and
 635 * calls NVM write activate.
 636 */
 637static int
 638ice_devlink_aq_set_port_option(struct ice_pf *pf, u8 option_idx,
 639			       struct netlink_ext_ack *extack)
 640{
 641	struct device *dev = ice_pf_to_dev(pf);
 642	int status;
 643
 644	status = ice_aq_set_port_option(&pf->hw, 0, true, option_idx);
 645	if (status) {
 646		dev_dbg(dev, "ice_aq_set_port_option, err %d aq_err %d\n",
 647			status, pf->hw.adminq.sq_last_status);
 648		NL_SET_ERR_MSG_MOD(extack, "Port split request failed");
 649		return -EIO;
 650	}
 651
 652	status = ice_acquire_nvm(&pf->hw, ICE_RES_WRITE);
 653	if (status) {
 654		dev_dbg(dev, "ice_acquire_nvm failed, err %d aq_err %d\n",
 655			status, pf->hw.adminq.sq_last_status);
 656		NL_SET_ERR_MSG_MOD(extack, "Failed to acquire NVM semaphore");
 657		return -EIO;
 658	}
 659
 660	status = ice_nvm_write_activate(&pf->hw, ICE_AQC_NVM_ACTIV_REQ_EMPR, NULL);
 661	if (status) {
 662		dev_dbg(dev, "ice_nvm_write_activate failed, err %d aq_err %d\n",
 663			status, pf->hw.adminq.sq_last_status);
 664		NL_SET_ERR_MSG_MOD(extack, "Port split request failed to save data");
 665		ice_release_nvm(&pf->hw);
 666		return -EIO;
 667	}
 668
 669	ice_release_nvm(&pf->hw);
 670
 671	NL_SET_ERR_MSG_MOD(extack, "Reboot required to finish port split");
 672	return 0;
 673}
 674
 675/**
 676 * ice_devlink_port_split - .port_split devlink handler
 677 * @devlink: devlink instance structure
 678 * @port: devlink port structure
 679 * @count: number of ports to split to
 680 * @extack: extended netdev ack structure
 681 *
 682 * Callback for the devlink .port_split operation.
 683 *
 684 * Unfortunately, the devlink expression of available options is limited
 685 * to just a number, so search for an FW port option which supports
 686 * the specified number. As there could be multiple FW port options with
 687 * the same port split count, allow switching between them. When the same
 688 * port split count request is issued again, switch to the next FW port
 689 * option with the same port split count.
 690 *
 691 * Return: zero on success or an error code on failure.
 692 */
 693static int
 694ice_devlink_port_split(struct devlink *devlink, struct devlink_port *port,
 695		       unsigned int count, struct netlink_ext_ack *extack)
 696{
 697	struct ice_aqc_get_port_options_elem options[ICE_AQC_PORT_OPT_MAX];
 698	u8 i, j, active_idx, pending_idx, new_option;
 699	struct ice_pf *pf = devlink_priv(devlink);
 700	u8 option_count = ICE_AQC_PORT_OPT_MAX;
 701	struct device *dev = ice_pf_to_dev(pf);
 702	bool active_valid, pending_valid;
 703	int status;
 704
 705	status = ice_aq_get_port_options(&pf->hw, options, &option_count,
 706					 0, true, &active_idx, &active_valid,
 707					 &pending_idx, &pending_valid);
 708	if (status) {
 709		dev_dbg(dev, "Couldn't read port split options, err = %d\n",
 710			status);
 711		NL_SET_ERR_MSG_MOD(extack, "Failed to get available port split options");
 712		return -EIO;
 713	}
 714
 715	new_option = ICE_AQC_PORT_OPT_MAX;
 716	active_idx = pending_valid ? pending_idx : active_idx;
 717	for (i = 1; i <= option_count; i++) {
 718		/* In order to allow switching between FW port options with
 719		 * the same port split count, search for a new option starting
 720		 * from the active/pending option (with array wrap around).
 721		 */
 722		j = (active_idx + i) % option_count;
 723
 724		if (count == options[j].pmd) {
 725			new_option = j;
 726			break;
 727		}
 728	}
 729
 730	if (new_option == active_idx) {
 731		dev_dbg(dev, "request to split: count: %u is already set and there are no other options\n",
 732			count);
 733		NL_SET_ERR_MSG_MOD(extack, "Requested split count is already set");
 734		ice_devlink_port_options_print(pf);
 735		return -EINVAL;
 736	}
 737
 738	if (new_option == ICE_AQC_PORT_OPT_MAX) {
 739		dev_dbg(dev, "request to split: count: %u not found\n", count);
 740		NL_SET_ERR_MSG_MOD(extack, "Port split requested unsupported port config");
 741		ice_devlink_port_options_print(pf);
 742		return -EINVAL;
 743	}
 744
 745	status = ice_devlink_aq_set_port_option(pf, new_option, extack);
 746	if (status)
 747		return status;
 748
 749	ice_devlink_port_options_print(pf);
 750
 751	return 0;
 752}
 753
 754/**
 755 * ice_devlink_port_unsplit - .port_unsplit devlink handler
 756 * @devlink: devlink instance structure
 757 * @port: devlink port structure
 758 * @extack: extended netdev ack structure
 759 *
 760 * Callback for the devlink .port_unsplit operation.
 761 * Calls ice_devlink_port_split with split count set to 1.
 762 * There could be no FW option available with split count 1.
 763 *
 764 * Return: zero on success or an error code on failure.
 765 */
 766static int
 767ice_devlink_port_unsplit(struct devlink *devlink, struct devlink_port *port,
 768			 struct netlink_ext_ack *extack)
 769{
 770	return ice_devlink_port_split(devlink, port, 1, extack);
 771}
 772
 773/**
 774 * ice_tear_down_devlink_rate_tree - removes devlink-rate exported tree
 775 * @pf: pf struct
 776 *
 777 * This function tears down tree exported during VF's creation.
 778 */
 779void ice_tear_down_devlink_rate_tree(struct ice_pf *pf)
 780{
 781	struct devlink *devlink;
 782	struct ice_vf *vf;
 783	unsigned int bkt;
 784
 785	devlink = priv_to_devlink(pf);
 786
 787	devl_lock(devlink);
 788	mutex_lock(&pf->vfs.table_lock);
 789	ice_for_each_vf(pf, bkt, vf) {
 790		if (vf->devlink_port.devlink_rate)
 791			devl_rate_leaf_destroy(&vf->devlink_port);
 792	}
 793	mutex_unlock(&pf->vfs.table_lock);
 794
 795	devl_rate_nodes_destroy(devlink);
 796	devl_unlock(devlink);
 797}
 798
 799/**
 800 * ice_enable_custom_tx - try to enable custom Tx feature
 801 * @pf: pf struct
 802 *
 803 * This function tries to enable custom Tx feature,
 804 * it's not possible to enable it, if DCB or ADQ is active.
 805 */
 806static bool ice_enable_custom_tx(struct ice_pf *pf)
 807{
 808	struct ice_port_info *pi = ice_get_main_vsi(pf)->port_info;
 809	struct device *dev = ice_pf_to_dev(pf);
 810
 811	if (pi->is_custom_tx_enabled)
 812		/* already enabled, return true */
 813		return true;
 814
 815	if (ice_is_adq_active(pf)) {
 816		dev_err(dev, "ADQ active, can't modify Tx scheduler tree\n");
 817		return false;
 818	}
 819
 820	if (ice_is_dcb_active(pf)) {
 821		dev_err(dev, "DCB active, can't modify Tx scheduler tree\n");
 822		return false;
 823	}
 824
 825	pi->is_custom_tx_enabled = true;
 826
 827	return true;
 828}
 829
 830/**
 831 * ice_traverse_tx_tree - traverse Tx scheduler tree
 832 * @devlink: devlink struct
 833 * @node: current node, used for recursion
 834 * @tc_node: tc_node struct, that is treated as a root
 835 * @pf: pf struct
 836 *
 837 * This function traverses Tx scheduler tree and exports
 838 * entire structure to the devlink-rate.
 839 */
 840static void ice_traverse_tx_tree(struct devlink *devlink, struct ice_sched_node *node,
 841				 struct ice_sched_node *tc_node, struct ice_pf *pf)
 842{
 843	struct devlink_rate *rate_node = NULL;
 844	struct ice_vf *vf;
 845	int i;
 846
 847	if (node->rate_node)
 848		/* already added, skip to the next */
 849		goto traverse_children;
 850
 851	if (node->parent == tc_node) {
 852		/* create root node */
 853		rate_node = devl_rate_node_create(devlink, node, node->name, NULL);
 854	} else if (node->vsi_handle &&
 855		   pf->vsi[node->vsi_handle]->vf) {
 856		vf = pf->vsi[node->vsi_handle]->vf;
 857		if (!vf->devlink_port.devlink_rate)
 858			/* leaf nodes doesn't have children
 859			 * so we don't set rate_node
 860			 */
 861			devl_rate_leaf_create(&vf->devlink_port, node,
 862					      node->parent->rate_node);
 863	} else if (node->info.data.elem_type != ICE_AQC_ELEM_TYPE_LEAF &&
 864		   node->parent->rate_node) {
 865		rate_node = devl_rate_node_create(devlink, node, node->name,
 866						  node->parent->rate_node);
 867	}
 868
 869	if (rate_node && !IS_ERR(rate_node))
 870		node->rate_node = rate_node;
 871
 872traverse_children:
 873	for (i = 0; i < node->num_children; i++)
 874		ice_traverse_tx_tree(devlink, node->children[i], tc_node, pf);
 875}
 876
 877/**
 878 * ice_devlink_rate_init_tx_topology - export Tx scheduler tree to devlink rate
 879 * @devlink: devlink struct
 880 * @vsi: main vsi struct
 881 *
 882 * This function finds a root node, then calls ice_traverse_tx tree, which
 883 * traverses the tree and exports it's contents to devlink rate.
 884 */
 885int ice_devlink_rate_init_tx_topology(struct devlink *devlink, struct ice_vsi *vsi)
 886{
 887	struct ice_port_info *pi = vsi->port_info;
 888	struct ice_sched_node *tc_node;
 889	struct ice_pf *pf = vsi->back;
 890	int i;
 891
 892	tc_node = pi->root->children[0];
 893	mutex_lock(&pi->sched_lock);
 894	devl_lock(devlink);
 895	for (i = 0; i < tc_node->num_children; i++)
 896		ice_traverse_tx_tree(devlink, tc_node->children[i], tc_node, pf);
 897	devl_unlock(devlink);
 898	mutex_unlock(&pi->sched_lock);
 899
 900	return 0;
 901}
 902
 903static void ice_clear_rate_nodes(struct ice_sched_node *node)
 904{
 905	node->rate_node = NULL;
 906
 907	for (int i = 0; i < node->num_children; i++)
 908		ice_clear_rate_nodes(node->children[i]);
 909}
 910
 911/**
 912 * ice_devlink_rate_clear_tx_topology - clear node->rate_node
 913 * @vsi: main vsi struct
 914 *
 915 * Clear rate_node to cleanup creation of Tx topology.
 916 *
 917 */
 918void ice_devlink_rate_clear_tx_topology(struct ice_vsi *vsi)
 919{
 920	struct ice_port_info *pi = vsi->port_info;
 921
 922	mutex_lock(&pi->sched_lock);
 923	ice_clear_rate_nodes(pi->root->children[0]);
 924	mutex_unlock(&pi->sched_lock);
 925}
 926
 927/**
 928 * ice_set_object_tx_share - sets node scheduling parameter
 929 * @pi: devlink struct instance
 930 * @node: node struct instance
 931 * @bw: bandwidth in bytes per second
 932 * @extack: extended netdev ack structure
 933 *
 934 * This function sets ICE_MIN_BW scheduling BW limit.
 935 */
 936static int ice_set_object_tx_share(struct ice_port_info *pi, struct ice_sched_node *node,
 937				   u64 bw, struct netlink_ext_ack *extack)
 938{
 939	int status;
 940
 941	mutex_lock(&pi->sched_lock);
 942	/* converts bytes per second to kilo bits per second */
 943	node->tx_share = div_u64(bw, 125);
 944	status = ice_sched_set_node_bw_lmt(pi, node, ICE_MIN_BW, node->tx_share);
 945	mutex_unlock(&pi->sched_lock);
 946
 947	if (status)
 948		NL_SET_ERR_MSG_MOD(extack, "Can't set scheduling node tx_share");
 949
 950	return status;
 951}
 952
 953/**
 954 * ice_set_object_tx_max - sets node scheduling parameter
 955 * @pi: devlink struct instance
 956 * @node: node struct instance
 957 * @bw: bandwidth in bytes per second
 958 * @extack: extended netdev ack structure
 959 *
 960 * This function sets ICE_MAX_BW scheduling BW limit.
 961 */
 962static int ice_set_object_tx_max(struct ice_port_info *pi, struct ice_sched_node *node,
 963				 u64 bw, struct netlink_ext_ack *extack)
 964{
 965	int status;
 966
 967	mutex_lock(&pi->sched_lock);
 968	/* converts bytes per second value to kilo bits per second */
 969	node->tx_max = div_u64(bw, 125);
 970	status = ice_sched_set_node_bw_lmt(pi, node, ICE_MAX_BW, node->tx_max);
 971	mutex_unlock(&pi->sched_lock);
 972
 973	if (status)
 974		NL_SET_ERR_MSG_MOD(extack, "Can't set scheduling node tx_max");
 975
 976	return status;
 977}
 978
 979/**
 980 * ice_set_object_tx_priority - sets node scheduling parameter
 981 * @pi: devlink struct instance
 982 * @node: node struct instance
 983 * @priority: value representing priority for strict priority arbitration
 984 * @extack: extended netdev ack structure
 985 *
 986 * This function sets priority of node among siblings.
 987 */
 988static int ice_set_object_tx_priority(struct ice_port_info *pi, struct ice_sched_node *node,
 989				      u32 priority, struct netlink_ext_ack *extack)
 990{
 991	int status;
 992
 993	if (priority >= 8) {
 994		NL_SET_ERR_MSG_MOD(extack, "Priority should be less than 8");
 995		return -EINVAL;
 996	}
 997
 998	mutex_lock(&pi->sched_lock);
 999	node->tx_priority = priority;
1000	status = ice_sched_set_node_priority(pi, node, node->tx_priority);
1001	mutex_unlock(&pi->sched_lock);
1002
1003	if (status)
1004		NL_SET_ERR_MSG_MOD(extack, "Can't set scheduling node tx_priority");
1005
1006	return status;
1007}
1008
1009/**
1010 * ice_set_object_tx_weight - sets node scheduling parameter
1011 * @pi: devlink struct instance
1012 * @node: node struct instance
1013 * @weight: value represeting relative weight for WFQ arbitration
1014 * @extack: extended netdev ack structure
1015 *
1016 * This function sets node weight for WFQ algorithm.
1017 */
1018static int ice_set_object_tx_weight(struct ice_port_info *pi, struct ice_sched_node *node,
1019				    u32 weight, struct netlink_ext_ack *extack)
1020{
1021	int status;
1022
1023	if (weight > 200 || weight < 1) {
1024		NL_SET_ERR_MSG_MOD(extack, "Weight must be between 1 and 200");
1025		return -EINVAL;
1026	}
1027
1028	mutex_lock(&pi->sched_lock);
1029	node->tx_weight = weight;
1030	status = ice_sched_set_node_weight(pi, node, node->tx_weight);
1031	mutex_unlock(&pi->sched_lock);
1032
1033	if (status)
1034		NL_SET_ERR_MSG_MOD(extack, "Can't set scheduling node tx_weight");
1035
1036	return status;
1037}
1038
1039/**
1040 * ice_get_pi_from_dev_rate - get port info from devlink_rate
1041 * @rate_node: devlink struct instance
1042 *
1043 * This function returns corresponding port_info struct of devlink_rate
1044 */
1045static struct ice_port_info *ice_get_pi_from_dev_rate(struct devlink_rate *rate_node)
1046{
1047	struct ice_pf *pf = devlink_priv(rate_node->devlink);
1048
1049	return ice_get_main_vsi(pf)->port_info;
1050}
1051
1052static int ice_devlink_rate_node_new(struct devlink_rate *rate_node, void **priv,
1053				     struct netlink_ext_ack *extack)
1054{
1055	struct ice_sched_node *node;
1056	struct ice_port_info *pi;
1057
1058	pi = ice_get_pi_from_dev_rate(rate_node);
1059
1060	if (!ice_enable_custom_tx(devlink_priv(rate_node->devlink)))
1061		return -EBUSY;
1062
1063	/* preallocate memory for ice_sched_node */
1064	node = devm_kzalloc(ice_hw_to_dev(pi->hw), sizeof(*node), GFP_KERNEL);
1065	*priv = node;
1066
1067	return 0;
1068}
1069
1070static int ice_devlink_rate_node_del(struct devlink_rate *rate_node, void *priv,
1071				     struct netlink_ext_ack *extack)
1072{
1073	struct ice_sched_node *node, *tc_node;
1074	struct ice_port_info *pi;
1075
1076	pi = ice_get_pi_from_dev_rate(rate_node);
1077	tc_node = pi->root->children[0];
1078	node = priv;
1079
1080	if (!rate_node->parent || !node || tc_node == node || !extack)
1081		return 0;
1082
1083	if (!ice_enable_custom_tx(devlink_priv(rate_node->devlink)))
1084		return -EBUSY;
1085
1086	/* can't allow to delete a node with children */
1087	if (node->num_children)
1088		return -EINVAL;
1089
1090	mutex_lock(&pi->sched_lock);
1091	ice_free_sched_node(pi, node);
1092	mutex_unlock(&pi->sched_lock);
1093
1094	return 0;
1095}
1096
1097static int ice_devlink_rate_leaf_tx_max_set(struct devlink_rate *rate_leaf, void *priv,
1098					    u64 tx_max, struct netlink_ext_ack *extack)
1099{
1100	struct ice_sched_node *node = priv;
1101
1102	if (!ice_enable_custom_tx(devlink_priv(rate_leaf->devlink)))
1103		return -EBUSY;
1104
1105	if (!node)
1106		return 0;
1107
1108	return ice_set_object_tx_max(ice_get_pi_from_dev_rate(rate_leaf),
1109				     node, tx_max, extack);
1110}
1111
1112static int ice_devlink_rate_leaf_tx_share_set(struct devlink_rate *rate_leaf, void *priv,
1113					      u64 tx_share, struct netlink_ext_ack *extack)
1114{
1115	struct ice_sched_node *node = priv;
1116
1117	if (!ice_enable_custom_tx(devlink_priv(rate_leaf->devlink)))
1118		return -EBUSY;
1119
1120	if (!node)
1121		return 0;
1122
1123	return ice_set_object_tx_share(ice_get_pi_from_dev_rate(rate_leaf), node,
1124				       tx_share, extack);
1125}
1126
1127static int ice_devlink_rate_leaf_tx_priority_set(struct devlink_rate *rate_leaf, void *priv,
1128						 u32 tx_priority, struct netlink_ext_ack *extack)
1129{
1130	struct ice_sched_node *node = priv;
1131
1132	if (!ice_enable_custom_tx(devlink_priv(rate_leaf->devlink)))
1133		return -EBUSY;
1134
1135	if (!node)
1136		return 0;
1137
1138	return ice_set_object_tx_priority(ice_get_pi_from_dev_rate(rate_leaf), node,
1139					  tx_priority, extack);
1140}
1141
1142static int ice_devlink_rate_leaf_tx_weight_set(struct devlink_rate *rate_leaf, void *priv,
1143					       u32 tx_weight, struct netlink_ext_ack *extack)
1144{
1145	struct ice_sched_node *node = priv;
1146
1147	if (!ice_enable_custom_tx(devlink_priv(rate_leaf->devlink)))
1148		return -EBUSY;
1149
1150	if (!node)
1151		return 0;
1152
1153	return ice_set_object_tx_weight(ice_get_pi_from_dev_rate(rate_leaf), node,
1154					tx_weight, extack);
1155}
1156
1157static int ice_devlink_rate_node_tx_max_set(struct devlink_rate *rate_node, void *priv,
1158					    u64 tx_max, struct netlink_ext_ack *extack)
1159{
1160	struct ice_sched_node *node = priv;
1161
1162	if (!ice_enable_custom_tx(devlink_priv(rate_node->devlink)))
1163		return -EBUSY;
1164
1165	if (!node)
1166		return 0;
1167
1168	return ice_set_object_tx_max(ice_get_pi_from_dev_rate(rate_node),
1169				     node, tx_max, extack);
1170}
1171
1172static int ice_devlink_rate_node_tx_share_set(struct devlink_rate *rate_node, void *priv,
1173					      u64 tx_share, struct netlink_ext_ack *extack)
1174{
1175	struct ice_sched_node *node = priv;
1176
1177	if (!ice_enable_custom_tx(devlink_priv(rate_node->devlink)))
1178		return -EBUSY;
1179
1180	if (!node)
1181		return 0;
1182
1183	return ice_set_object_tx_share(ice_get_pi_from_dev_rate(rate_node),
1184				       node, tx_share, extack);
1185}
1186
1187static int ice_devlink_rate_node_tx_priority_set(struct devlink_rate *rate_node, void *priv,
1188						 u32 tx_priority, struct netlink_ext_ack *extack)
1189{
1190	struct ice_sched_node *node = priv;
1191
1192	if (!ice_enable_custom_tx(devlink_priv(rate_node->devlink)))
1193		return -EBUSY;
1194
1195	if (!node)
1196		return 0;
1197
1198	return ice_set_object_tx_priority(ice_get_pi_from_dev_rate(rate_node),
1199					  node, tx_priority, extack);
1200}
1201
1202static int ice_devlink_rate_node_tx_weight_set(struct devlink_rate *rate_node, void *priv,
1203					       u32 tx_weight, struct netlink_ext_ack *extack)
1204{
1205	struct ice_sched_node *node = priv;
1206
1207	if (!ice_enable_custom_tx(devlink_priv(rate_node->devlink)))
1208		return -EBUSY;
1209
1210	if (!node)
1211		return 0;
1212
1213	return ice_set_object_tx_weight(ice_get_pi_from_dev_rate(rate_node),
1214					node, tx_weight, extack);
1215}
1216
1217static int ice_devlink_set_parent(struct devlink_rate *devlink_rate,
1218				  struct devlink_rate *parent,
1219				  void *priv, void *parent_priv,
1220				  struct netlink_ext_ack *extack)
1221{
1222	struct ice_port_info *pi = ice_get_pi_from_dev_rate(devlink_rate);
1223	struct ice_sched_node *tc_node, *node, *parent_node;
1224	u16 num_nodes_added;
1225	u32 first_node_teid;
1226	u32 node_teid;
1227	int status;
1228
1229	tc_node = pi->root->children[0];
1230	node = priv;
1231
1232	if (!extack)
1233		return 0;
1234
1235	if (!ice_enable_custom_tx(devlink_priv(devlink_rate->devlink)))
1236		return -EBUSY;
1237
1238	if (!parent) {
1239		if (!node || tc_node == node || node->num_children)
1240			return -EINVAL;
1241
1242		mutex_lock(&pi->sched_lock);
1243		ice_free_sched_node(pi, node);
1244		mutex_unlock(&pi->sched_lock);
1245
1246		return 0;
1247	}
1248
1249	parent_node = parent_priv;
1250
1251	/* if the node doesn't exist, create it */
1252	if (!node->parent) {
1253		mutex_lock(&pi->sched_lock);
1254		status = ice_sched_add_elems(pi, tc_node, parent_node,
1255					     parent_node->tx_sched_layer + 1,
1256					     1, &num_nodes_added, &first_node_teid,
1257					     &node);
1258		mutex_unlock(&pi->sched_lock);
1259
1260		if (status) {
1261			NL_SET_ERR_MSG_MOD(extack, "Can't add a new node");
1262			return status;
1263		}
1264
1265		if (devlink_rate->tx_share)
1266			ice_set_object_tx_share(pi, node, devlink_rate->tx_share, extack);
1267		if (devlink_rate->tx_max)
1268			ice_set_object_tx_max(pi, node, devlink_rate->tx_max, extack);
1269		if (devlink_rate->tx_priority)
1270			ice_set_object_tx_priority(pi, node, devlink_rate->tx_priority, extack);
1271		if (devlink_rate->tx_weight)
1272			ice_set_object_tx_weight(pi, node, devlink_rate->tx_weight, extack);
1273	} else {
1274		node_teid = le32_to_cpu(node->info.node_teid);
1275		mutex_lock(&pi->sched_lock);
1276		status = ice_sched_move_nodes(pi, parent_node, 1, &node_teid);
1277		mutex_unlock(&pi->sched_lock);
1278
1279		if (status)
1280			NL_SET_ERR_MSG_MOD(extack, "Can't move existing node to a new parent");
1281	}
1282
1283	return status;
1284}
1285
1286/**
1287 * ice_devlink_reinit_up - do reinit of the given PF
1288 * @pf: pointer to the PF struct
1289 */
1290static int ice_devlink_reinit_up(struct ice_pf *pf)
1291{
1292	struct ice_vsi *vsi = ice_get_main_vsi(pf);
1293	struct ice_vsi_cfg_params params;
1294	int err;
1295
1296	err = ice_init_dev(pf);
1297	if (err)
1298		return err;
1299
1300	params = ice_vsi_to_params(vsi);
1301	params.flags = ICE_VSI_FLAG_INIT;
1302
1303	rtnl_lock();
1304	err = ice_vsi_cfg(vsi, &params);
1305	rtnl_unlock();
1306	if (err)
1307		goto err_vsi_cfg;
1308
1309	/* No need to take devl_lock, it's already taken by devlink API */
1310	err = ice_load(pf);
1311	if (err)
1312		goto err_load;
1313
1314	return 0;
1315
1316err_load:
1317	rtnl_lock();
1318	ice_vsi_decfg(vsi);
1319	rtnl_unlock();
1320err_vsi_cfg:
1321	ice_deinit_dev(pf);
1322	return err;
1323}
1324
1325/**
1326 * ice_devlink_reload_up - do reload up after reinit
1327 * @devlink: pointer to the devlink instance reloading
1328 * @action: the action requested
1329 * @limit: limits imposed by userspace, such as not resetting
1330 * @actions_performed: on return, indicate what actions actually performed
1331 * @extack: netlink extended ACK structure
1332 */
1333static int
1334ice_devlink_reload_up(struct devlink *devlink,
1335		      enum devlink_reload_action action,
1336		      enum devlink_reload_limit limit,
1337		      u32 *actions_performed,
1338		      struct netlink_ext_ack *extack)
1339{
1340	struct ice_pf *pf = devlink_priv(devlink);
1341
1342	switch (action) {
1343	case DEVLINK_RELOAD_ACTION_DRIVER_REINIT:
1344		*actions_performed = BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT);
1345		return ice_devlink_reinit_up(pf);
1346	case DEVLINK_RELOAD_ACTION_FW_ACTIVATE:
1347		*actions_performed = BIT(DEVLINK_RELOAD_ACTION_FW_ACTIVATE);
1348		return ice_devlink_reload_empr_finish(pf, extack);
1349	default:
1350		WARN_ON(1);
1351		return -EOPNOTSUPP;
1352	}
1353}
1354
1355static const struct devlink_ops ice_devlink_ops = {
1356	.supported_flash_update_params = DEVLINK_SUPPORT_FLASH_UPDATE_OVERWRITE_MASK,
1357	.reload_actions = BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT) |
1358			  BIT(DEVLINK_RELOAD_ACTION_FW_ACTIVATE),
1359	.reload_down = ice_devlink_reload_down,
1360	.reload_up = ice_devlink_reload_up,
1361	.eswitch_mode_get = ice_eswitch_mode_get,
1362	.eswitch_mode_set = ice_eswitch_mode_set,
1363	.info_get = ice_devlink_info_get,
1364	.flash_update = ice_devlink_flash_update,
1365
1366	.rate_node_new = ice_devlink_rate_node_new,
1367	.rate_node_del = ice_devlink_rate_node_del,
1368
1369	.rate_leaf_tx_max_set = ice_devlink_rate_leaf_tx_max_set,
1370	.rate_leaf_tx_share_set = ice_devlink_rate_leaf_tx_share_set,
1371	.rate_leaf_tx_priority_set = ice_devlink_rate_leaf_tx_priority_set,
1372	.rate_leaf_tx_weight_set = ice_devlink_rate_leaf_tx_weight_set,
1373
1374	.rate_node_tx_max_set = ice_devlink_rate_node_tx_max_set,
1375	.rate_node_tx_share_set = ice_devlink_rate_node_tx_share_set,
1376	.rate_node_tx_priority_set = ice_devlink_rate_node_tx_priority_set,
1377	.rate_node_tx_weight_set = ice_devlink_rate_node_tx_weight_set,
1378
1379	.rate_leaf_parent_set = ice_devlink_set_parent,
1380	.rate_node_parent_set = ice_devlink_set_parent,
1381};
1382
1383static int
1384ice_devlink_enable_roce_get(struct devlink *devlink, u32 id,
1385			    struct devlink_param_gset_ctx *ctx)
1386{
1387	struct ice_pf *pf = devlink_priv(devlink);
1388
1389	ctx->val.vbool = pf->rdma_mode & IIDC_RDMA_PROTOCOL_ROCEV2 ? true : false;
1390
1391	return 0;
1392}
1393
1394static int
1395ice_devlink_enable_roce_set(struct devlink *devlink, u32 id,
1396			    struct devlink_param_gset_ctx *ctx)
1397{
1398	struct ice_pf *pf = devlink_priv(devlink);
1399	bool roce_ena = ctx->val.vbool;
1400	int ret;
1401
1402	if (!roce_ena) {
1403		ice_unplug_aux_dev(pf);
1404		pf->rdma_mode &= ~IIDC_RDMA_PROTOCOL_ROCEV2;
1405		return 0;
1406	}
1407
1408	pf->rdma_mode |= IIDC_RDMA_PROTOCOL_ROCEV2;
1409	ret = ice_plug_aux_dev(pf);
1410	if (ret)
1411		pf->rdma_mode &= ~IIDC_RDMA_PROTOCOL_ROCEV2;
1412
1413	return ret;
1414}
1415
1416static int
1417ice_devlink_enable_roce_validate(struct devlink *devlink, u32 id,
1418				 union devlink_param_value val,
1419				 struct netlink_ext_ack *extack)
1420{
1421	struct ice_pf *pf = devlink_priv(devlink);
1422
1423	if (!test_bit(ICE_FLAG_RDMA_ENA, pf->flags))
1424		return -EOPNOTSUPP;
1425
1426	if (pf->rdma_mode & IIDC_RDMA_PROTOCOL_IWARP) {
1427		NL_SET_ERR_MSG_MOD(extack, "iWARP is currently enabled. This device cannot enable iWARP and RoCEv2 simultaneously");
1428		return -EOPNOTSUPP;
1429	}
1430
1431	return 0;
1432}
1433
1434static int
1435ice_devlink_enable_iw_get(struct devlink *devlink, u32 id,
1436			  struct devlink_param_gset_ctx *ctx)
1437{
1438	struct ice_pf *pf = devlink_priv(devlink);
1439
1440	ctx->val.vbool = pf->rdma_mode & IIDC_RDMA_PROTOCOL_IWARP;
1441
1442	return 0;
1443}
1444
1445static int
1446ice_devlink_enable_iw_set(struct devlink *devlink, u32 id,
1447			  struct devlink_param_gset_ctx *ctx)
1448{
1449	struct ice_pf *pf = devlink_priv(devlink);
1450	bool iw_ena = ctx->val.vbool;
1451	int ret;
1452
1453	if (!iw_ena) {
1454		ice_unplug_aux_dev(pf);
1455		pf->rdma_mode &= ~IIDC_RDMA_PROTOCOL_IWARP;
1456		return 0;
1457	}
1458
1459	pf->rdma_mode |= IIDC_RDMA_PROTOCOL_IWARP;
1460	ret = ice_plug_aux_dev(pf);
1461	if (ret)
1462		pf->rdma_mode &= ~IIDC_RDMA_PROTOCOL_IWARP;
1463
1464	return ret;
1465}
1466
1467static int
1468ice_devlink_enable_iw_validate(struct devlink *devlink, u32 id,
1469			       union devlink_param_value val,
1470			       struct netlink_ext_ack *extack)
1471{
1472	struct ice_pf *pf = devlink_priv(devlink);
1473
1474	if (!test_bit(ICE_FLAG_RDMA_ENA, pf->flags))
1475		return -EOPNOTSUPP;
1476
1477	if (pf->rdma_mode & IIDC_RDMA_PROTOCOL_ROCEV2) {
1478		NL_SET_ERR_MSG_MOD(extack, "RoCEv2 is currently enabled. This device cannot enable iWARP and RoCEv2 simultaneously");
1479		return -EOPNOTSUPP;
1480	}
1481
1482	return 0;
1483}
1484
1485static const struct devlink_param ice_devlink_params[] = {
1486	DEVLINK_PARAM_GENERIC(ENABLE_ROCE, BIT(DEVLINK_PARAM_CMODE_RUNTIME),
1487			      ice_devlink_enable_roce_get,
1488			      ice_devlink_enable_roce_set,
1489			      ice_devlink_enable_roce_validate),
1490	DEVLINK_PARAM_GENERIC(ENABLE_IWARP, BIT(DEVLINK_PARAM_CMODE_RUNTIME),
1491			      ice_devlink_enable_iw_get,
1492			      ice_devlink_enable_iw_set,
1493			      ice_devlink_enable_iw_validate),
1494
1495};
1496
1497static void ice_devlink_free(void *devlink_ptr)
1498{
1499	devlink_free((struct devlink *)devlink_ptr);
1500}
1501
1502/**
1503 * ice_allocate_pf - Allocate devlink and return PF structure pointer
1504 * @dev: the device to allocate for
1505 *
1506 * Allocate a devlink instance for this device and return the private area as
1507 * the PF structure. The devlink memory is kept track of through devres by
1508 * adding an action to remove it when unwinding.
1509 */
1510struct ice_pf *ice_allocate_pf(struct device *dev)
1511{
1512	struct devlink *devlink;
1513
1514	devlink = devlink_alloc(&ice_devlink_ops, sizeof(struct ice_pf), dev);
1515	if (!devlink)
1516		return NULL;
1517
1518	/* Add an action to teardown the devlink when unwinding the driver */
1519	if (devm_add_action_or_reset(dev, ice_devlink_free, devlink))
1520		return NULL;
1521
1522	return devlink_priv(devlink);
1523}
1524
1525/**
1526 * ice_devlink_register - Register devlink interface for this PF
1527 * @pf: the PF to register the devlink for.
1528 *
1529 * Register the devlink instance associated with this physical function.
1530 *
1531 * Return: zero on success or an error code on failure.
1532 */
1533void ice_devlink_register(struct ice_pf *pf)
1534{
1535	struct devlink *devlink = priv_to_devlink(pf);
1536
1537	devlink_register(devlink);
1538}
1539
1540/**
1541 * ice_devlink_unregister - Unregister devlink resources for this PF.
1542 * @pf: the PF structure to cleanup
1543 *
1544 * Releases resources used by devlink and cleans up associated memory.
1545 */
1546void ice_devlink_unregister(struct ice_pf *pf)
1547{
1548	devlink_unregister(priv_to_devlink(pf));
1549}
1550
1551/**
1552 * ice_devlink_set_switch_id - Set unique switch id based on pci dsn
1553 * @pf: the PF to create a devlink port for
1554 * @ppid: struct with switch id information
1555 */
1556static void
1557ice_devlink_set_switch_id(struct ice_pf *pf, struct netdev_phys_item_id *ppid)
1558{
1559	struct pci_dev *pdev = pf->pdev;
1560	u64 id;
1561
1562	id = pci_get_dsn(pdev);
1563
1564	ppid->id_len = sizeof(id);
1565	put_unaligned_be64(id, &ppid->id);
1566}
1567
1568int ice_devlink_register_params(struct ice_pf *pf)
1569{
1570	struct devlink *devlink = priv_to_devlink(pf);
1571
1572	return devlink_params_register(devlink, ice_devlink_params,
1573				       ARRAY_SIZE(ice_devlink_params));
1574}
1575
1576void ice_devlink_unregister_params(struct ice_pf *pf)
1577{
1578	devlink_params_unregister(priv_to_devlink(pf), ice_devlink_params,
1579				  ARRAY_SIZE(ice_devlink_params));
1580}
1581
1582/**
1583 * ice_devlink_set_port_split_options - Set port split options
1584 * @pf: the PF to set port split options
1585 * @attrs: devlink attributes
1586 *
1587 * Sets devlink port split options based on available FW port options
1588 */
1589static void
1590ice_devlink_set_port_split_options(struct ice_pf *pf,
1591				   struct devlink_port_attrs *attrs)
1592{
1593	struct ice_aqc_get_port_options_elem options[ICE_AQC_PORT_OPT_MAX];
1594	u8 i, active_idx, pending_idx, option_count = ICE_AQC_PORT_OPT_MAX;
1595	bool active_valid, pending_valid;
1596	int status;
1597
1598	status = ice_aq_get_port_options(&pf->hw, options, &option_count,
1599					 0, true, &active_idx, &active_valid,
1600					 &pending_idx, &pending_valid);
1601	if (status) {
1602		dev_dbg(ice_pf_to_dev(pf), "Couldn't read port split options, err = %d\n",
1603			status);
1604		return;
1605	}
1606
1607	/* find the biggest available port split count */
1608	for (i = 0; i < option_count; i++)
1609		attrs->lanes = max_t(int, attrs->lanes, options[i].pmd);
1610
1611	attrs->splittable = attrs->lanes ? 1 : 0;
1612	ice_active_port_option = active_idx;
1613}
1614
1615static const struct devlink_port_ops ice_devlink_port_ops = {
1616	.port_split = ice_devlink_port_split,
1617	.port_unsplit = ice_devlink_port_unsplit,
1618};
1619
1620/**
1621 * ice_devlink_create_pf_port - Create a devlink port for this PF
1622 * @pf: the PF to create a devlink port for
1623 *
1624 * Create and register a devlink_port for this PF.
1625 * This function has to be called under devl_lock.
1626 *
1627 * Return: zero on success or an error code on failure.
1628 */
1629int ice_devlink_create_pf_port(struct ice_pf *pf)
1630{
1631	struct devlink_port_attrs attrs = {};
1632	struct devlink_port *devlink_port;
1633	struct devlink *devlink;
1634	struct ice_vsi *vsi;
1635	struct device *dev;
1636	int err;
1637
1638	devlink = priv_to_devlink(pf);
1639
1640	dev = ice_pf_to_dev(pf);
1641
1642	devlink_port = &pf->devlink_port;
1643
1644	vsi = ice_get_main_vsi(pf);
1645	if (!vsi)
1646		return -EIO;
1647
1648	attrs.flavour = DEVLINK_PORT_FLAVOUR_PHYSICAL;
1649	attrs.phys.port_number = pf->hw.bus.func;
1650
1651	/* As FW supports only port split options for whole device,
1652	 * set port split options only for first PF.
1653	 */
1654	if (pf->hw.pf_id == 0)
1655		ice_devlink_set_port_split_options(pf, &attrs);
1656
1657	ice_devlink_set_switch_id(pf, &attrs.switch_id);
1658
1659	devlink_port_attrs_set(devlink_port, &attrs);
 
1660
1661	err = devl_port_register_with_ops(devlink, devlink_port, vsi->idx,
1662					  &ice_devlink_port_ops);
1663	if (err) {
1664		dev_err(dev, "Failed to create devlink port for PF %d, error %d\n",
1665			pf->hw.pf_id, err);
1666		return err;
1667	}
1668
1669	return 0;
1670}
1671
1672/**
1673 * ice_devlink_destroy_pf_port - Destroy the devlink_port for this PF
1674 * @pf: the PF to cleanup
1675 *
1676 * Unregisters the devlink_port structure associated with this PF.
1677 * This function has to be called under devl_lock.
1678 */
1679void ice_devlink_destroy_pf_port(struct ice_pf *pf)
1680{
1681	devl_port_unregister(&pf->devlink_port);
1682}
1683
1684/**
1685 * ice_devlink_create_vf_port - Create a devlink port for this VF
1686 * @vf: the VF to create a port for
1687 *
1688 * Create and register a devlink_port for this VF.
1689 *
1690 * Return: zero on success or an error code on failure.
1691 */
1692int ice_devlink_create_vf_port(struct ice_vf *vf)
1693{
1694	struct devlink_port_attrs attrs = {};
1695	struct devlink_port *devlink_port;
1696	struct devlink *devlink;
1697	struct ice_vsi *vsi;
1698	struct device *dev;
1699	struct ice_pf *pf;
1700	int err;
1701
1702	pf = vf->pf;
1703	dev = ice_pf_to_dev(pf);
1704	devlink_port = &vf->devlink_port;
1705
1706	vsi = ice_get_vf_vsi(vf);
1707	if (!vsi)
1708		return -EINVAL;
1709
1710	attrs.flavour = DEVLINK_PORT_FLAVOUR_PCI_VF;
1711	attrs.pci_vf.pf = pf->hw.bus.func;
1712	attrs.pci_vf.vf = vf->vf_id;
1713
1714	ice_devlink_set_switch_id(pf, &attrs.switch_id);
1715
1716	devlink_port_attrs_set(devlink_port, &attrs);
1717	devlink = priv_to_devlink(pf);
1718
1719	err = devlink_port_register(devlink, devlink_port, vsi->idx);
1720	if (err) {
1721		dev_err(dev, "Failed to create devlink port for VF %d, error %d\n",
1722			vf->vf_id, err);
1723		return err;
1724	}
1725
1726	return 0;
1727}
1728
1729/**
1730 * ice_devlink_destroy_vf_port - Destroy the devlink_port for this VF
1731 * @vf: the VF to cleanup
1732 *
1733 * Unregisters the devlink_port structure associated with this VF.
1734 */
1735void ice_devlink_destroy_vf_port(struct ice_vf *vf)
1736{
1737	devl_rate_leaf_destroy(&vf->devlink_port);
1738	devlink_port_unregister(&vf->devlink_port);
1739}
1740
1741#define ICE_DEVLINK_READ_BLK_SIZE (1024 * 1024)
1742
1743static const struct devlink_region_ops ice_nvm_region_ops;
1744static const struct devlink_region_ops ice_sram_region_ops;
1745
1746/**
1747 * ice_devlink_nvm_snapshot - Capture a snapshot of the NVM flash contents
1748 * @devlink: the devlink instance
1749 * @ops: the devlink region to snapshot
1750 * @extack: extended ACK response structure
1751 * @data: on exit points to snapshot data buffer
1752 *
1753 * This function is called in response to a DEVLINK_CMD_REGION_NEW for either
1754 * the nvm-flash or shadow-ram region.
1755 *
1756 * It captures a snapshot of the NVM or Shadow RAM flash contents. This
1757 * snapshot can then later be viewed via the DEVLINK_CMD_REGION_READ netlink
1758 * interface.
1759 *
1760 * @returns zero on success, and updates the data pointer. Returns a non-zero
1761 * error code on failure.
1762 */
1763static int ice_devlink_nvm_snapshot(struct devlink *devlink,
1764				    const struct devlink_region_ops *ops,
1765				    struct netlink_ext_ack *extack, u8 **data)
1766{
1767	struct ice_pf *pf = devlink_priv(devlink);
1768	struct device *dev = ice_pf_to_dev(pf);
1769	struct ice_hw *hw = &pf->hw;
1770	bool read_shadow_ram;
1771	u8 *nvm_data, *tmp, i;
1772	u32 nvm_size, left;
1773	s8 num_blks;
1774	int status;
1775
1776	if (ops == &ice_nvm_region_ops) {
1777		read_shadow_ram = false;
1778		nvm_size = hw->flash.flash_size;
1779	} else if (ops == &ice_sram_region_ops) {
1780		read_shadow_ram = true;
1781		nvm_size = hw->flash.sr_words * 2u;
1782	} else {
1783		NL_SET_ERR_MSG_MOD(extack, "Unexpected region in snapshot function");
1784		return -EOPNOTSUPP;
1785	}
1786
1787	nvm_data = vzalloc(nvm_size);
1788	if (!nvm_data)
1789		return -ENOMEM;
1790
1791	num_blks = DIV_ROUND_UP(nvm_size, ICE_DEVLINK_READ_BLK_SIZE);
1792	tmp = nvm_data;
1793	left = nvm_size;
1794
1795	/* Some systems take longer to read the NVM than others which causes the
1796	 * FW to reclaim the NVM lock before the entire NVM has been read. Fix
1797	 * this by breaking the reads of the NVM into smaller chunks that will
1798	 * probably not take as long. This has some overhead since we are
1799	 * increasing the number of AQ commands, but it should always work
1800	 */
1801	for (i = 0; i < num_blks; i++) {
1802		u32 read_sz = min_t(u32, ICE_DEVLINK_READ_BLK_SIZE, left);
1803
1804		status = ice_acquire_nvm(hw, ICE_RES_READ);
1805		if (status) {
1806			dev_dbg(dev, "ice_acquire_nvm failed, err %d aq_err %d\n",
1807				status, hw->adminq.sq_last_status);
1808			NL_SET_ERR_MSG_MOD(extack, "Failed to acquire NVM semaphore");
1809			vfree(nvm_data);
1810			return -EIO;
1811		}
1812
1813		status = ice_read_flat_nvm(hw, i * ICE_DEVLINK_READ_BLK_SIZE,
1814					   &read_sz, tmp, read_shadow_ram);
1815		if (status) {
1816			dev_dbg(dev, "ice_read_flat_nvm failed after reading %u bytes, err %d aq_err %d\n",
1817				read_sz, status, hw->adminq.sq_last_status);
1818			NL_SET_ERR_MSG_MOD(extack, "Failed to read NVM contents");
1819			ice_release_nvm(hw);
1820			vfree(nvm_data);
1821			return -EIO;
1822		}
1823		ice_release_nvm(hw);
1824
1825		tmp += read_sz;
1826		left -= read_sz;
1827	}
1828
1829	*data = nvm_data;
1830
1831	return 0;
1832}
1833
1834/**
1835 * ice_devlink_nvm_read - Read a portion of NVM flash contents
1836 * @devlink: the devlink instance
1837 * @ops: the devlink region to snapshot
1838 * @extack: extended ACK response structure
1839 * @offset: the offset to start at
1840 * @size: the amount to read
1841 * @data: the data buffer to read into
1842 *
1843 * This function is called in response to DEVLINK_CMD_REGION_READ to directly
1844 * read a section of the NVM contents.
1845 *
1846 * It reads from either the nvm-flash or shadow-ram region contents.
1847 *
1848 * @returns zero on success, and updates the data pointer. Returns a non-zero
1849 * error code on failure.
1850 */
1851static int ice_devlink_nvm_read(struct devlink *devlink,
1852				const struct devlink_region_ops *ops,
1853				struct netlink_ext_ack *extack,
1854				u64 offset, u32 size, u8 *data)
1855{
1856	struct ice_pf *pf = devlink_priv(devlink);
1857	struct device *dev = ice_pf_to_dev(pf);
1858	struct ice_hw *hw = &pf->hw;
1859	bool read_shadow_ram;
1860	u64 nvm_size;
1861	int status;
1862
1863	if (ops == &ice_nvm_region_ops) {
1864		read_shadow_ram = false;
1865		nvm_size = hw->flash.flash_size;
1866	} else if (ops == &ice_sram_region_ops) {
1867		read_shadow_ram = true;
1868		nvm_size = hw->flash.sr_words * 2u;
1869	} else {
1870		NL_SET_ERR_MSG_MOD(extack, "Unexpected region in snapshot function");
1871		return -EOPNOTSUPP;
1872	}
1873
1874	if (offset + size >= nvm_size) {
1875		NL_SET_ERR_MSG_MOD(extack, "Cannot read beyond the region size");
1876		return -ERANGE;
1877	}
1878
1879	status = ice_acquire_nvm(hw, ICE_RES_READ);
1880	if (status) {
1881		dev_dbg(dev, "ice_acquire_nvm failed, err %d aq_err %d\n",
1882			status, hw->adminq.sq_last_status);
1883		NL_SET_ERR_MSG_MOD(extack, "Failed to acquire NVM semaphore");
1884		return -EIO;
1885	}
1886
1887	status = ice_read_flat_nvm(hw, (u32)offset, &size, data,
1888				   read_shadow_ram);
1889	if (status) {
1890		dev_dbg(dev, "ice_read_flat_nvm failed after reading %u bytes, err %d aq_err %d\n",
1891			size, status, hw->adminq.sq_last_status);
1892		NL_SET_ERR_MSG_MOD(extack, "Failed to read NVM contents");
1893		ice_release_nvm(hw);
1894		return -EIO;
1895	}
1896	ice_release_nvm(hw);
1897
1898	return 0;
1899}
1900
1901/**
1902 * ice_devlink_devcaps_snapshot - Capture snapshot of device capabilities
1903 * @devlink: the devlink instance
1904 * @ops: the devlink region being snapshotted
1905 * @extack: extended ACK response structure
1906 * @data: on exit points to snapshot data buffer
1907 *
1908 * This function is called in response to the DEVLINK_CMD_REGION_TRIGGER for
1909 * the device-caps devlink region. It captures a snapshot of the device
1910 * capabilities reported by firmware.
1911 *
1912 * @returns zero on success, and updates the data pointer. Returns a non-zero
1913 * error code on failure.
1914 */
1915static int
1916ice_devlink_devcaps_snapshot(struct devlink *devlink,
1917			     const struct devlink_region_ops *ops,
1918			     struct netlink_ext_ack *extack, u8 **data)
1919{
1920	struct ice_pf *pf = devlink_priv(devlink);
1921	struct device *dev = ice_pf_to_dev(pf);
1922	struct ice_hw *hw = &pf->hw;
1923	void *devcaps;
1924	int status;
1925
1926	devcaps = vzalloc(ICE_AQ_MAX_BUF_LEN);
1927	if (!devcaps)
1928		return -ENOMEM;
1929
1930	status = ice_aq_list_caps(hw, devcaps, ICE_AQ_MAX_BUF_LEN, NULL,
1931				  ice_aqc_opc_list_dev_caps, NULL);
1932	if (status) {
1933		dev_dbg(dev, "ice_aq_list_caps: failed to read device capabilities, err %d aq_err %d\n",
1934			status, hw->adminq.sq_last_status);
1935		NL_SET_ERR_MSG_MOD(extack, "Failed to read device capabilities");
1936		vfree(devcaps);
1937		return status;
1938	}
1939
1940	*data = (u8 *)devcaps;
1941
1942	return 0;
1943}
1944
1945static const struct devlink_region_ops ice_nvm_region_ops = {
1946	.name = "nvm-flash",
1947	.destructor = vfree,
1948	.snapshot = ice_devlink_nvm_snapshot,
1949	.read = ice_devlink_nvm_read,
1950};
1951
1952static const struct devlink_region_ops ice_sram_region_ops = {
1953	.name = "shadow-ram",
1954	.destructor = vfree,
1955	.snapshot = ice_devlink_nvm_snapshot,
1956	.read = ice_devlink_nvm_read,
1957};
1958
1959static const struct devlink_region_ops ice_devcaps_region_ops = {
1960	.name = "device-caps",
1961	.destructor = vfree,
1962	.snapshot = ice_devlink_devcaps_snapshot,
1963};
1964
1965/**
1966 * ice_devlink_init_regions - Initialize devlink regions
1967 * @pf: the PF device structure
1968 *
1969 * Create devlink regions used to enable access to dump the contents of the
1970 * flash memory on the device.
1971 */
1972void ice_devlink_init_regions(struct ice_pf *pf)
1973{
1974	struct devlink *devlink = priv_to_devlink(pf);
1975	struct device *dev = ice_pf_to_dev(pf);
1976	u64 nvm_size, sram_size;
1977
1978	nvm_size = pf->hw.flash.flash_size;
1979	pf->nvm_region = devlink_region_create(devlink, &ice_nvm_region_ops, 1,
1980					       nvm_size);
1981	if (IS_ERR(pf->nvm_region)) {
1982		dev_err(dev, "failed to create NVM devlink region, err %ld\n",
1983			PTR_ERR(pf->nvm_region));
1984		pf->nvm_region = NULL;
1985	}
1986
1987	sram_size = pf->hw.flash.sr_words * 2u;
1988	pf->sram_region = devlink_region_create(devlink, &ice_sram_region_ops,
1989						1, sram_size);
1990	if (IS_ERR(pf->sram_region)) {
1991		dev_err(dev, "failed to create shadow-ram devlink region, err %ld\n",
1992			PTR_ERR(pf->sram_region));
1993		pf->sram_region = NULL;
1994	}
1995
1996	pf->devcaps_region = devlink_region_create(devlink,
1997						   &ice_devcaps_region_ops, 10,
1998						   ICE_AQ_MAX_BUF_LEN);
1999	if (IS_ERR(pf->devcaps_region)) {
2000		dev_err(dev, "failed to create device-caps devlink region, err %ld\n",
2001			PTR_ERR(pf->devcaps_region));
2002		pf->devcaps_region = NULL;
2003	}
2004}
2005
2006/**
2007 * ice_devlink_destroy_regions - Destroy devlink regions
2008 * @pf: the PF device structure
2009 *
2010 * Remove previously created regions for this PF.
2011 */
2012void ice_devlink_destroy_regions(struct ice_pf *pf)
2013{
2014	if (pf->nvm_region)
2015		devlink_region_destroy(pf->nvm_region);
2016
2017	if (pf->sram_region)
2018		devlink_region_destroy(pf->sram_region);
2019
2020	if (pf->devcaps_region)
2021		devlink_region_destroy(pf->devcaps_region);
2022}