1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright (c) 2020, Intel Corporation. */
4 #include <linux/vmalloc.h>
8 #include "ice_devlink.h"
9 #include "ice_eswitch.h"
10 #include "ice_fw_update.h"
11 #include "ice_dcb_lib.h"
13 static int ice_active_port_option = -1;
15 /* context for devlink info version reporting */
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;
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
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.
34 static void ice_info_get_dsn(struct ice_pf *pf, struct ice_info_ctx *ctx)
38 /* Copy the DSN into an array in Big Endian format */
39 put_unaligned_be64(pci_get_dsn(pf->pdev), dsn);
41 snprintf(ctx->buf, sizeof(ctx->buf), "%8phD", dsn);
44 static void ice_info_pba(struct ice_pf *pf, struct ice_info_ctx *ctx)
46 struct ice_hw *hw = &pf->hw;
49 status = ice_read_pba_string(hw, (u8 *)ctx->buf, sizeof(ctx->buf));
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",
56 static void ice_info_fw_mgmt(struct ice_pf *pf, struct ice_info_ctx *ctx)
58 struct ice_hw *hw = &pf->hw;
60 snprintf(ctx->buf, sizeof(ctx->buf), "%u.%u.%u",
61 hw->fw_maj_ver, hw->fw_min_ver, hw->fw_patch);
64 static void ice_info_fw_api(struct ice_pf *pf, struct ice_info_ctx *ctx)
66 struct ice_hw *hw = &pf->hw;
68 snprintf(ctx->buf, sizeof(ctx->buf), "%u.%u.%u", hw->api_maj_ver,
69 hw->api_min_ver, hw->api_patch);
72 static void ice_info_fw_build(struct ice_pf *pf, struct ice_info_ctx *ctx)
74 struct ice_hw *hw = &pf->hw;
76 snprintf(ctx->buf, sizeof(ctx->buf), "0x%08x", hw->fw_build);
79 static void ice_info_orom_ver(struct ice_pf *pf, struct ice_info_ctx *ctx)
81 struct ice_orom_info *orom = &pf->hw.flash.orom;
83 snprintf(ctx->buf, sizeof(ctx->buf), "%u.%u.%u",
84 orom->major, orom->build, orom->patch);
88 ice_info_pending_orom_ver(struct ice_pf __always_unused *pf,
89 struct ice_info_ctx *ctx)
91 struct ice_orom_info *orom = &ctx->pending_orom;
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);
98 static void ice_info_nvm_ver(struct ice_pf *pf, struct ice_info_ctx *ctx)
100 struct ice_nvm_info *nvm = &pf->hw.flash.nvm;
102 snprintf(ctx->buf, sizeof(ctx->buf), "%x.%02x", nvm->major, nvm->minor);
106 ice_info_pending_nvm_ver(struct ice_pf __always_unused *pf,
107 struct ice_info_ctx *ctx)
109 struct ice_nvm_info *nvm = &ctx->pending_nvm;
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);
116 static void ice_info_eetrack(struct ice_pf *pf, struct ice_info_ctx *ctx)
118 struct ice_nvm_info *nvm = &pf->hw.flash.nvm;
120 snprintf(ctx->buf, sizeof(ctx->buf), "0x%08x", nvm->eetrack);
124 ice_info_pending_eetrack(struct ice_pf *pf, struct ice_info_ctx *ctx)
126 struct ice_nvm_info *nvm = &ctx->pending_nvm;
128 if (ctx->dev_caps.common_cap.nvm_update_pending_nvm)
129 snprintf(ctx->buf, sizeof(ctx->buf), "0x%08x", nvm->eetrack);
132 static void ice_info_ddp_pkg_name(struct ice_pf *pf, struct ice_info_ctx *ctx)
134 struct ice_hw *hw = &pf->hw;
136 snprintf(ctx->buf, sizeof(ctx->buf), "%s", hw->active_pkg_name);
140 ice_info_ddp_pkg_version(struct ice_pf *pf, struct ice_info_ctx *ctx)
142 struct ice_pkg_ver *pkg = &pf->hw.active_pkg_ver;
144 snprintf(ctx->buf, sizeof(ctx->buf), "%u.%u.%u.%u",
145 pkg->major, pkg->minor, pkg->update, pkg->draft);
149 ice_info_ddp_pkg_bundle_id(struct ice_pf *pf, struct ice_info_ctx *ctx)
151 snprintf(ctx->buf, sizeof(ctx->buf), "0x%08x", pf->hw.active_track_id);
154 static void ice_info_netlist_ver(struct ice_pf *pf, struct ice_info_ctx *ctx)
156 struct ice_netlist_info *netlist = &pf->hw.flash.netlist;
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);
165 static void ice_info_netlist_build(struct ice_pf *pf, struct ice_info_ctx *ctx)
167 struct ice_netlist_info *netlist = &pf->hw.flash.netlist;
169 snprintf(ctx->buf, sizeof(ctx->buf), "0x%08x", netlist->hash);
173 ice_info_pending_netlist_ver(struct ice_pf __always_unused *pf,
174 struct ice_info_ctx *ctx)
176 struct ice_netlist_info *netlist = &ctx->pending_netlist;
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);
187 ice_info_pending_netlist_build(struct ice_pf __always_unused *pf,
188 struct ice_info_ctx *ctx)
190 struct ice_netlist_info *netlist = &ctx->pending_netlist;
192 if (ctx->dev_caps.common_cap.nvm_update_pending_netlist)
193 snprintf(ctx->buf, sizeof(ctx->buf), "0x%08x", netlist->hash);
196 static void ice_info_cgu_fw_build(struct ice_pf *pf, struct ice_info_ctx *ctx)
198 u32 id, cfg_ver, fw_ver;
200 if (!ice_is_feature_supported(pf, ICE_F_CGU))
202 if (ice_aq_get_cgu_info(&pf->hw, &id, &cfg_ver, &fw_ver))
204 snprintf(ctx->buf, sizeof(ctx->buf), "%u.%u.%u", id, cfg_ver, fw_ver);
207 static void ice_info_cgu_id(struct ice_pf *pf, struct ice_info_ctx *ctx)
209 if (!ice_is_feature_supported(pf, ICE_F_CGU))
211 snprintf(ctx->buf, sizeof(ctx->buf), "%u", pf->hw.cgu_part_number);
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 }
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.
228 #define combined(key, active, pending) \
229 running(key, active), \
230 stored(key, pending, active)
232 enum ice_version_type {
238 static const struct ice_devlink_version {
239 enum ice_version_type type;
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),
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
266 * Callback for the devlink .info_get operation. Reports information about the
269 * Return: zero on success or an error code on failure.
271 static int ice_devlink_info_get(struct devlink *devlink,
272 struct devlink_info_req *req,
273 struct netlink_ext_ack *extack)
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;
282 err = ice_wait_for_reset(pf, 10 * HZ);
284 NL_SET_ERR_MSG_MOD(extack, "Device is busy resetting");
288 ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
292 /* discover capabilities first */
293 err = ice_discover_dev_caps(hw, &ctx->dev_caps);
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");
301 if (ctx->dev_caps.common_cap.nvm_update_pending_orom) {
302 err = ice_get_inactive_orom_ver(hw, &ctx->pending_orom);
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));
307 /* disable display of pending Option ROM */
308 ctx->dev_caps.common_cap.nvm_update_pending_orom = false;
312 if (ctx->dev_caps.common_cap.nvm_update_pending_nvm) {
313 err = ice_get_inactive_nvm_ver(hw, &ctx->pending_nvm);
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));
318 /* disable display of pending Option ROM */
319 ctx->dev_caps.common_cap.nvm_update_pending_nvm = false;
323 if (ctx->dev_caps.common_cap.nvm_update_pending_netlist) {
324 err = ice_get_inactive_netlist_ver(hw, &ctx->pending_netlist);
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));
329 /* disable display of pending Option ROM */
330 ctx->dev_caps.common_cap.nvm_update_pending_netlist = false;
334 ice_info_get_dsn(pf, ctx);
336 err = devlink_info_serial_number_put(req, ctx->buf);
338 NL_SET_ERR_MSG_MOD(extack, "Unable to set serial number");
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;
346 memset(ctx->buf, 0, sizeof(ctx->buf));
348 ice_devlink_versions[i].getter(pf, ctx);
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.
355 if (ctx->buf[0] == '\0' && ice_devlink_versions[i].fallback)
356 ice_devlink_versions[i].fallback(pf, ctx);
358 /* Do not report missing versions */
359 if (ctx->buf[0] == '\0')
363 case ICE_VERSION_FIXED:
364 err = devlink_info_version_fixed_put(req, key, ctx->buf);
366 NL_SET_ERR_MSG_MOD(extack, "Unable to set fixed version");
370 case ICE_VERSION_RUNNING:
371 err = devlink_info_version_running_put(req, key, ctx->buf);
373 NL_SET_ERR_MSG_MOD(extack, "Unable to set running version");
377 case ICE_VERSION_STORED:
378 err = devlink_info_version_stored_put(req, key, ctx->buf);
380 NL_SET_ERR_MSG_MOD(extack, "Unable to set stored version");
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
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.
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
407 ice_devlink_reload_empr_start(struct ice_pf *pf,
408 struct netlink_ext_ack *extack)
410 struct device *dev = ice_pf_to_dev(pf);
411 struct ice_hw *hw = &pf->hw;
415 err = ice_get_pending_updates(pf, &pending, extack);
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.
425 NL_SET_ERR_MSG_MOD(extack, "No pending firmware update");
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");
434 dev_dbg(dev, "Issuing device EMP reset to activate firmware\n");
436 err = ice_aq_nvm_update_empr(hw);
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");
448 * ice_devlink_reload_down - prepare for reload
449 * @devlink: pointer to the devlink instance to reload
450 * @netns_change: if true, the network namespace is changing
451 * @action: the action to perform
452 * @limit: limits on what reload should do, such as not resetting
453 * @extack: netlink extended ACK structure
456 ice_devlink_reload_down(struct devlink *devlink, bool netns_change,
457 enum devlink_reload_action action,
458 enum devlink_reload_limit limit,
459 struct netlink_ext_ack *extack)
461 struct ice_pf *pf = devlink_priv(devlink);
464 case DEVLINK_RELOAD_ACTION_DRIVER_REINIT:
465 if (ice_is_eswitch_mode_switchdev(pf)) {
466 NL_SET_ERR_MSG_MOD(extack,
467 "Go to legacy mode before doing reinit\n");
470 if (ice_is_adq_active(pf)) {
471 NL_SET_ERR_MSG_MOD(extack,
472 "Turn off ADQ before doing reinit\n");
475 if (ice_has_vfs(pf)) {
476 NL_SET_ERR_MSG_MOD(extack,
477 "Remove all VFs before doing reinit\n");
482 case DEVLINK_RELOAD_ACTION_FW_ACTIVATE:
483 return ice_devlink_reload_empr_start(pf, extack);
491 * ice_devlink_reload_empr_finish - Wait for EMP reset to finish
492 * @pf: pointer to the pf instance
493 * @extack: netlink extended ACK structure
495 * Wait for driver to finish rebuilding after EMP reset is completed. This
496 * includes time to wait for both the actual device reset as well as the time
497 * for the driver's rebuild to complete.
500 ice_devlink_reload_empr_finish(struct ice_pf *pf,
501 struct netlink_ext_ack *extack)
505 err = ice_wait_for_reset(pf, 60 * HZ);
507 NL_SET_ERR_MSG_MOD(extack, "Device still resetting after 1 minute");
515 * ice_devlink_port_opt_speed_str - convert speed to a string
516 * @speed: speed value
518 static const char *ice_devlink_port_opt_speed_str(u8 speed)
520 switch (speed & ICE_AQC_PORT_OPT_MAX_LANE_M) {
521 case ICE_AQC_PORT_OPT_MAX_LANE_100M:
523 case ICE_AQC_PORT_OPT_MAX_LANE_1G:
525 case ICE_AQC_PORT_OPT_MAX_LANE_2500M:
527 case ICE_AQC_PORT_OPT_MAX_LANE_5G:
529 case ICE_AQC_PORT_OPT_MAX_LANE_10G:
531 case ICE_AQC_PORT_OPT_MAX_LANE_25G:
533 case ICE_AQC_PORT_OPT_MAX_LANE_50G:
535 case ICE_AQC_PORT_OPT_MAX_LANE_100G:
542 #define ICE_PORT_OPT_DESC_LEN 50
544 * ice_devlink_port_options_print - Print available port split options
545 * @pf: the PF to print split port options
547 * Prints a table with available port split options and max port speeds
549 static void ice_devlink_port_options_print(struct ice_pf *pf)
551 u8 i, j, options_count, cnt, speed, pending_idx, active_idx;
552 struct ice_aqc_get_port_options_elem *options, *opt;
553 struct device *dev = ice_pf_to_dev(pf);
554 bool active_valid, pending_valid;
555 char desc[ICE_PORT_OPT_DESC_LEN];
559 options = kcalloc(ICE_AQC_PORT_OPT_MAX * ICE_MAX_PORT_PER_PCI_DEV,
560 sizeof(*options), GFP_KERNEL);
564 for (i = 0; i < ICE_MAX_PORT_PER_PCI_DEV; i++) {
565 opt = options + i * ICE_AQC_PORT_OPT_MAX;
566 options_count = ICE_AQC_PORT_OPT_MAX;
569 status = ice_aq_get_port_options(&pf->hw, opt, &options_count,
570 i, true, &active_idx,
571 &active_valid, &pending_idx,
574 dev_dbg(dev, "Couldn't read port option for port %d, err %d\n",
580 dev_dbg(dev, "Available port split options and max port speeds (Gbps):\n");
581 dev_dbg(dev, "Status Split Quad 0 Quad 1\n");
582 dev_dbg(dev, " count L0 L1 L2 L3 L4 L5 L6 L7\n");
584 for (i = 0; i < options_count; i++) {
587 if (i == ice_active_port_option)
589 else if ((i == pending_idx) && pending_valid)
594 cnt += snprintf(&desc[cnt], ICE_PORT_OPT_DESC_LEN - cnt,
597 cnt += snprintf(&desc[cnt], ICE_PORT_OPT_DESC_LEN - cnt,
598 "%-6u", options[i].pmd);
600 for (j = 0; j < ICE_MAX_PORT_PER_PCI_DEV; ++j) {
601 speed = options[i + j * ICE_AQC_PORT_OPT_MAX].max_lane_speed;
602 str = ice_devlink_port_opt_speed_str(speed);
603 cnt += snprintf(&desc[cnt], ICE_PORT_OPT_DESC_LEN - cnt,
607 dev_dbg(dev, "%s\n", desc);
615 * ice_devlink_aq_set_port_option - Send set port option admin queue command
616 * @pf: the PF to print split port options
617 * @option_idx: selected port option
618 * @extack: extended netdev ack structure
620 * Sends set port option admin queue command with selected port option and
621 * calls NVM write activate.
624 ice_devlink_aq_set_port_option(struct ice_pf *pf, u8 option_idx,
625 struct netlink_ext_ack *extack)
627 struct device *dev = ice_pf_to_dev(pf);
630 status = ice_aq_set_port_option(&pf->hw, 0, true, option_idx);
632 dev_dbg(dev, "ice_aq_set_port_option, err %d aq_err %d\n",
633 status, pf->hw.adminq.sq_last_status);
634 NL_SET_ERR_MSG_MOD(extack, "Port split request failed");
638 status = ice_acquire_nvm(&pf->hw, ICE_RES_WRITE);
640 dev_dbg(dev, "ice_acquire_nvm failed, err %d aq_err %d\n",
641 status, pf->hw.adminq.sq_last_status);
642 NL_SET_ERR_MSG_MOD(extack, "Failed to acquire NVM semaphore");
646 status = ice_nvm_write_activate(&pf->hw, ICE_AQC_NVM_ACTIV_REQ_EMPR, NULL);
648 dev_dbg(dev, "ice_nvm_write_activate failed, err %d aq_err %d\n",
649 status, pf->hw.adminq.sq_last_status);
650 NL_SET_ERR_MSG_MOD(extack, "Port split request failed to save data");
651 ice_release_nvm(&pf->hw);
655 ice_release_nvm(&pf->hw);
657 NL_SET_ERR_MSG_MOD(extack, "Reboot required to finish port split");
662 * ice_devlink_port_split - .port_split devlink handler
663 * @devlink: devlink instance structure
664 * @port: devlink port structure
665 * @count: number of ports to split to
666 * @extack: extended netdev ack structure
668 * Callback for the devlink .port_split operation.
670 * Unfortunately, the devlink expression of available options is limited
671 * to just a number, so search for an FW port option which supports
672 * the specified number. As there could be multiple FW port options with
673 * the same port split count, allow switching between them. When the same
674 * port split count request is issued again, switch to the next FW port
675 * option with the same port split count.
677 * Return: zero on success or an error code on failure.
680 ice_devlink_port_split(struct devlink *devlink, struct devlink_port *port,
681 unsigned int count, struct netlink_ext_ack *extack)
683 struct ice_aqc_get_port_options_elem options[ICE_AQC_PORT_OPT_MAX];
684 u8 i, j, active_idx, pending_idx, new_option;
685 struct ice_pf *pf = devlink_priv(devlink);
686 u8 option_count = ICE_AQC_PORT_OPT_MAX;
687 struct device *dev = ice_pf_to_dev(pf);
688 bool active_valid, pending_valid;
691 status = ice_aq_get_port_options(&pf->hw, options, &option_count,
692 0, true, &active_idx, &active_valid,
693 &pending_idx, &pending_valid);
695 dev_dbg(dev, "Couldn't read port split options, err = %d\n",
697 NL_SET_ERR_MSG_MOD(extack, "Failed to get available port split options");
701 new_option = ICE_AQC_PORT_OPT_MAX;
702 active_idx = pending_valid ? pending_idx : active_idx;
703 for (i = 1; i <= option_count; i++) {
704 /* In order to allow switching between FW port options with
705 * the same port split count, search for a new option starting
706 * from the active/pending option (with array wrap around).
708 j = (active_idx + i) % option_count;
710 if (count == options[j].pmd) {
716 if (new_option == active_idx) {
717 dev_dbg(dev, "request to split: count: %u is already set and there are no other options\n",
719 NL_SET_ERR_MSG_MOD(extack, "Requested split count is already set");
720 ice_devlink_port_options_print(pf);
724 if (new_option == ICE_AQC_PORT_OPT_MAX) {
725 dev_dbg(dev, "request to split: count: %u not found\n", count);
726 NL_SET_ERR_MSG_MOD(extack, "Port split requested unsupported port config");
727 ice_devlink_port_options_print(pf);
731 status = ice_devlink_aq_set_port_option(pf, new_option, extack);
735 ice_devlink_port_options_print(pf);
741 * ice_devlink_port_unsplit - .port_unsplit devlink handler
742 * @devlink: devlink instance structure
743 * @port: devlink port structure
744 * @extack: extended netdev ack structure
746 * Callback for the devlink .port_unsplit operation.
747 * Calls ice_devlink_port_split with split count set to 1.
748 * There could be no FW option available with split count 1.
750 * Return: zero on success or an error code on failure.
753 ice_devlink_port_unsplit(struct devlink *devlink, struct devlink_port *port,
754 struct netlink_ext_ack *extack)
756 return ice_devlink_port_split(devlink, port, 1, extack);
760 * ice_tear_down_devlink_rate_tree - removes devlink-rate exported tree
763 * This function tears down tree exported during VF's creation.
765 void ice_tear_down_devlink_rate_tree(struct ice_pf *pf)
767 struct devlink *devlink;
771 devlink = priv_to_devlink(pf);
774 mutex_lock(&pf->vfs.table_lock);
775 ice_for_each_vf(pf, bkt, vf) {
776 if (vf->devlink_port.devlink_rate)
777 devl_rate_leaf_destroy(&vf->devlink_port);
779 mutex_unlock(&pf->vfs.table_lock);
781 devl_rate_nodes_destroy(devlink);
782 devl_unlock(devlink);
786 * ice_enable_custom_tx - try to enable custom Tx feature
789 * This function tries to enable custom Tx feature,
790 * it's not possible to enable it, if DCB or ADQ is active.
792 static bool ice_enable_custom_tx(struct ice_pf *pf)
794 struct ice_port_info *pi = ice_get_main_vsi(pf)->port_info;
795 struct device *dev = ice_pf_to_dev(pf);
797 if (pi->is_custom_tx_enabled)
798 /* already enabled, return true */
801 if (ice_is_adq_active(pf)) {
802 dev_err(dev, "ADQ active, can't modify Tx scheduler tree\n");
806 if (ice_is_dcb_active(pf)) {
807 dev_err(dev, "DCB active, can't modify Tx scheduler tree\n");
811 pi->is_custom_tx_enabled = true;
817 * ice_traverse_tx_tree - traverse Tx scheduler tree
818 * @devlink: devlink struct
819 * @node: current node, used for recursion
820 * @tc_node: tc_node struct, that is treated as a root
823 * This function traverses Tx scheduler tree and exports
824 * entire structure to the devlink-rate.
826 static void ice_traverse_tx_tree(struct devlink *devlink, struct ice_sched_node *node,
827 struct ice_sched_node *tc_node, struct ice_pf *pf)
829 struct devlink_rate *rate_node = NULL;
834 /* already added, skip to the next */
835 goto traverse_children;
837 if (node->parent == tc_node) {
838 /* create root node */
839 rate_node = devl_rate_node_create(devlink, node, node->name, NULL);
840 } else if (node->vsi_handle &&
841 pf->vsi[node->vsi_handle]->vf) {
842 vf = pf->vsi[node->vsi_handle]->vf;
843 if (!vf->devlink_port.devlink_rate)
844 /* leaf nodes doesn't have children
845 * so we don't set rate_node
847 devl_rate_leaf_create(&vf->devlink_port, node,
848 node->parent->rate_node);
849 } else if (node->info.data.elem_type != ICE_AQC_ELEM_TYPE_LEAF &&
850 node->parent->rate_node) {
851 rate_node = devl_rate_node_create(devlink, node, node->name,
852 node->parent->rate_node);
855 if (rate_node && !IS_ERR(rate_node))
856 node->rate_node = rate_node;
859 for (i = 0; i < node->num_children; i++)
860 ice_traverse_tx_tree(devlink, node->children[i], tc_node, pf);
864 * ice_devlink_rate_init_tx_topology - export Tx scheduler tree to devlink rate
865 * @devlink: devlink struct
866 * @vsi: main vsi struct
868 * This function finds a root node, then calls ice_traverse_tx tree, which
869 * traverses the tree and exports it's contents to devlink rate.
871 int ice_devlink_rate_init_tx_topology(struct devlink *devlink, struct ice_vsi *vsi)
873 struct ice_port_info *pi = vsi->port_info;
874 struct ice_sched_node *tc_node;
875 struct ice_pf *pf = vsi->back;
878 tc_node = pi->root->children[0];
879 mutex_lock(&pi->sched_lock);
881 for (i = 0; i < tc_node->num_children; i++)
882 ice_traverse_tx_tree(devlink, tc_node->children[i], tc_node, pf);
883 devl_unlock(devlink);
884 mutex_unlock(&pi->sched_lock);
889 static void ice_clear_rate_nodes(struct ice_sched_node *node)
891 node->rate_node = NULL;
893 for (int i = 0; i < node->num_children; i++)
894 ice_clear_rate_nodes(node->children[i]);
898 * ice_devlink_rate_clear_tx_topology - clear node->rate_node
899 * @vsi: main vsi struct
901 * Clear rate_node to cleanup creation of Tx topology.
904 void ice_devlink_rate_clear_tx_topology(struct ice_vsi *vsi)
906 struct ice_port_info *pi = vsi->port_info;
908 mutex_lock(&pi->sched_lock);
909 ice_clear_rate_nodes(pi->root->children[0]);
910 mutex_unlock(&pi->sched_lock);
914 * ice_set_object_tx_share - sets node scheduling parameter
915 * @pi: devlink struct instance
916 * @node: node struct instance
917 * @bw: bandwidth in bytes per second
918 * @extack: extended netdev ack structure
920 * This function sets ICE_MIN_BW scheduling BW limit.
922 static int ice_set_object_tx_share(struct ice_port_info *pi, struct ice_sched_node *node,
923 u64 bw, struct netlink_ext_ack *extack)
927 mutex_lock(&pi->sched_lock);
928 /* converts bytes per second to kilo bits per second */
929 node->tx_share = div_u64(bw, 125);
930 status = ice_sched_set_node_bw_lmt(pi, node, ICE_MIN_BW, node->tx_share);
931 mutex_unlock(&pi->sched_lock);
934 NL_SET_ERR_MSG_MOD(extack, "Can't set scheduling node tx_share");
940 * ice_set_object_tx_max - sets node scheduling parameter
941 * @pi: devlink struct instance
942 * @node: node struct instance
943 * @bw: bandwidth in bytes per second
944 * @extack: extended netdev ack structure
946 * This function sets ICE_MAX_BW scheduling BW limit.
948 static int ice_set_object_tx_max(struct ice_port_info *pi, struct ice_sched_node *node,
949 u64 bw, struct netlink_ext_ack *extack)
953 mutex_lock(&pi->sched_lock);
954 /* converts bytes per second value to kilo bits per second */
955 node->tx_max = div_u64(bw, 125);
956 status = ice_sched_set_node_bw_lmt(pi, node, ICE_MAX_BW, node->tx_max);
957 mutex_unlock(&pi->sched_lock);
960 NL_SET_ERR_MSG_MOD(extack, "Can't set scheduling node tx_max");
966 * ice_set_object_tx_priority - sets node scheduling parameter
967 * @pi: devlink struct instance
968 * @node: node struct instance
969 * @priority: value representing priority for strict priority arbitration
970 * @extack: extended netdev ack structure
972 * This function sets priority of node among siblings.
974 static int ice_set_object_tx_priority(struct ice_port_info *pi, struct ice_sched_node *node,
975 u32 priority, struct netlink_ext_ack *extack)
980 NL_SET_ERR_MSG_MOD(extack, "Priority should be less than 8");
984 mutex_lock(&pi->sched_lock);
985 node->tx_priority = priority;
986 status = ice_sched_set_node_priority(pi, node, node->tx_priority);
987 mutex_unlock(&pi->sched_lock);
990 NL_SET_ERR_MSG_MOD(extack, "Can't set scheduling node tx_priority");
996 * ice_set_object_tx_weight - sets node scheduling parameter
997 * @pi: devlink struct instance
998 * @node: node struct instance
999 * @weight: value represeting relative weight for WFQ arbitration
1000 * @extack: extended netdev ack structure
1002 * This function sets node weight for WFQ algorithm.
1004 static int ice_set_object_tx_weight(struct ice_port_info *pi, struct ice_sched_node *node,
1005 u32 weight, struct netlink_ext_ack *extack)
1009 if (weight > 200 || weight < 1) {
1010 NL_SET_ERR_MSG_MOD(extack, "Weight must be between 1 and 200");
1014 mutex_lock(&pi->sched_lock);
1015 node->tx_weight = weight;
1016 status = ice_sched_set_node_weight(pi, node, node->tx_weight);
1017 mutex_unlock(&pi->sched_lock);
1020 NL_SET_ERR_MSG_MOD(extack, "Can't set scheduling node tx_weight");
1026 * ice_get_pi_from_dev_rate - get port info from devlink_rate
1027 * @rate_node: devlink struct instance
1029 * This function returns corresponding port_info struct of devlink_rate
1031 static struct ice_port_info *ice_get_pi_from_dev_rate(struct devlink_rate *rate_node)
1033 struct ice_pf *pf = devlink_priv(rate_node->devlink);
1035 return ice_get_main_vsi(pf)->port_info;
1038 static int ice_devlink_rate_node_new(struct devlink_rate *rate_node, void **priv,
1039 struct netlink_ext_ack *extack)
1041 struct ice_sched_node *node;
1042 struct ice_port_info *pi;
1044 pi = ice_get_pi_from_dev_rate(rate_node);
1046 if (!ice_enable_custom_tx(devlink_priv(rate_node->devlink)))
1049 /* preallocate memory for ice_sched_node */
1050 node = devm_kzalloc(ice_hw_to_dev(pi->hw), sizeof(*node), GFP_KERNEL);
1056 static int ice_devlink_rate_node_del(struct devlink_rate *rate_node, void *priv,
1057 struct netlink_ext_ack *extack)
1059 struct ice_sched_node *node, *tc_node;
1060 struct ice_port_info *pi;
1062 pi = ice_get_pi_from_dev_rate(rate_node);
1063 tc_node = pi->root->children[0];
1066 if (!rate_node->parent || !node || tc_node == node || !extack)
1069 if (!ice_enable_custom_tx(devlink_priv(rate_node->devlink)))
1072 /* can't allow to delete a node with children */
1073 if (node->num_children)
1076 mutex_lock(&pi->sched_lock);
1077 ice_free_sched_node(pi, node);
1078 mutex_unlock(&pi->sched_lock);
1083 static int ice_devlink_rate_leaf_tx_max_set(struct devlink_rate *rate_leaf, void *priv,
1084 u64 tx_max, struct netlink_ext_ack *extack)
1086 struct ice_sched_node *node = priv;
1088 if (!ice_enable_custom_tx(devlink_priv(rate_leaf->devlink)))
1094 return ice_set_object_tx_max(ice_get_pi_from_dev_rate(rate_leaf),
1095 node, tx_max, extack);
1098 static int ice_devlink_rate_leaf_tx_share_set(struct devlink_rate *rate_leaf, void *priv,
1099 u64 tx_share, struct netlink_ext_ack *extack)
1101 struct ice_sched_node *node = priv;
1103 if (!ice_enable_custom_tx(devlink_priv(rate_leaf->devlink)))
1109 return ice_set_object_tx_share(ice_get_pi_from_dev_rate(rate_leaf), node,
1113 static int ice_devlink_rate_leaf_tx_priority_set(struct devlink_rate *rate_leaf, void *priv,
1114 u32 tx_priority, struct netlink_ext_ack *extack)
1116 struct ice_sched_node *node = priv;
1118 if (!ice_enable_custom_tx(devlink_priv(rate_leaf->devlink)))
1124 return ice_set_object_tx_priority(ice_get_pi_from_dev_rate(rate_leaf), node,
1125 tx_priority, extack);
1128 static int ice_devlink_rate_leaf_tx_weight_set(struct devlink_rate *rate_leaf, void *priv,
1129 u32 tx_weight, struct netlink_ext_ack *extack)
1131 struct ice_sched_node *node = priv;
1133 if (!ice_enable_custom_tx(devlink_priv(rate_leaf->devlink)))
1139 return ice_set_object_tx_weight(ice_get_pi_from_dev_rate(rate_leaf), node,
1143 static int ice_devlink_rate_node_tx_max_set(struct devlink_rate *rate_node, void *priv,
1144 u64 tx_max, struct netlink_ext_ack *extack)
1146 struct ice_sched_node *node = priv;
1148 if (!ice_enable_custom_tx(devlink_priv(rate_node->devlink)))
1154 return ice_set_object_tx_max(ice_get_pi_from_dev_rate(rate_node),
1155 node, tx_max, extack);
1158 static int ice_devlink_rate_node_tx_share_set(struct devlink_rate *rate_node, void *priv,
1159 u64 tx_share, struct netlink_ext_ack *extack)
1161 struct ice_sched_node *node = priv;
1163 if (!ice_enable_custom_tx(devlink_priv(rate_node->devlink)))
1169 return ice_set_object_tx_share(ice_get_pi_from_dev_rate(rate_node),
1170 node, tx_share, extack);
1173 static int ice_devlink_rate_node_tx_priority_set(struct devlink_rate *rate_node, void *priv,
1174 u32 tx_priority, struct netlink_ext_ack *extack)
1176 struct ice_sched_node *node = priv;
1178 if (!ice_enable_custom_tx(devlink_priv(rate_node->devlink)))
1184 return ice_set_object_tx_priority(ice_get_pi_from_dev_rate(rate_node),
1185 node, tx_priority, extack);
1188 static int ice_devlink_rate_node_tx_weight_set(struct devlink_rate *rate_node, void *priv,
1189 u32 tx_weight, struct netlink_ext_ack *extack)
1191 struct ice_sched_node *node = priv;
1193 if (!ice_enable_custom_tx(devlink_priv(rate_node->devlink)))
1199 return ice_set_object_tx_weight(ice_get_pi_from_dev_rate(rate_node),
1200 node, tx_weight, extack);
1203 static int ice_devlink_set_parent(struct devlink_rate *devlink_rate,
1204 struct devlink_rate *parent,
1205 void *priv, void *parent_priv,
1206 struct netlink_ext_ack *extack)
1208 struct ice_port_info *pi = ice_get_pi_from_dev_rate(devlink_rate);
1209 struct ice_sched_node *tc_node, *node, *parent_node;
1210 u16 num_nodes_added;
1211 u32 first_node_teid;
1215 tc_node = pi->root->children[0];
1221 if (!ice_enable_custom_tx(devlink_priv(devlink_rate->devlink)))
1225 if (!node || tc_node == node || node->num_children)
1228 mutex_lock(&pi->sched_lock);
1229 ice_free_sched_node(pi, node);
1230 mutex_unlock(&pi->sched_lock);
1235 parent_node = parent_priv;
1237 /* if the node doesn't exist, create it */
1238 if (!node->parent) {
1239 mutex_lock(&pi->sched_lock);
1240 status = ice_sched_add_elems(pi, tc_node, parent_node,
1241 parent_node->tx_sched_layer + 1,
1242 1, &num_nodes_added, &first_node_teid,
1244 mutex_unlock(&pi->sched_lock);
1247 NL_SET_ERR_MSG_MOD(extack, "Can't add a new node");
1251 if (devlink_rate->tx_share)
1252 ice_set_object_tx_share(pi, node, devlink_rate->tx_share, extack);
1253 if (devlink_rate->tx_max)
1254 ice_set_object_tx_max(pi, node, devlink_rate->tx_max, extack);
1255 if (devlink_rate->tx_priority)
1256 ice_set_object_tx_priority(pi, node, devlink_rate->tx_priority, extack);
1257 if (devlink_rate->tx_weight)
1258 ice_set_object_tx_weight(pi, node, devlink_rate->tx_weight, extack);
1260 node_teid = le32_to_cpu(node->info.node_teid);
1261 mutex_lock(&pi->sched_lock);
1262 status = ice_sched_move_nodes(pi, parent_node, 1, &node_teid);
1263 mutex_unlock(&pi->sched_lock);
1266 NL_SET_ERR_MSG_MOD(extack, "Can't move existing node to a new parent");
1273 * ice_devlink_reload_up - do reload up after reinit
1274 * @devlink: pointer to the devlink instance reloading
1275 * @action: the action requested
1276 * @limit: limits imposed by userspace, such as not resetting
1277 * @actions_performed: on return, indicate what actions actually performed
1278 * @extack: netlink extended ACK structure
1281 ice_devlink_reload_up(struct devlink *devlink,
1282 enum devlink_reload_action action,
1283 enum devlink_reload_limit limit,
1284 u32 *actions_performed,
1285 struct netlink_ext_ack *extack)
1287 struct ice_pf *pf = devlink_priv(devlink);
1290 case DEVLINK_RELOAD_ACTION_DRIVER_REINIT:
1291 *actions_performed = BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT);
1292 return ice_load(pf);
1293 case DEVLINK_RELOAD_ACTION_FW_ACTIVATE:
1294 *actions_performed = BIT(DEVLINK_RELOAD_ACTION_FW_ACTIVATE);
1295 return ice_devlink_reload_empr_finish(pf, extack);
1302 static const struct devlink_ops ice_devlink_ops = {
1303 .supported_flash_update_params = DEVLINK_SUPPORT_FLASH_UPDATE_OVERWRITE_MASK,
1304 .reload_actions = BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT) |
1305 BIT(DEVLINK_RELOAD_ACTION_FW_ACTIVATE),
1306 .reload_down = ice_devlink_reload_down,
1307 .reload_up = ice_devlink_reload_up,
1308 .eswitch_mode_get = ice_eswitch_mode_get,
1309 .eswitch_mode_set = ice_eswitch_mode_set,
1310 .info_get = ice_devlink_info_get,
1311 .flash_update = ice_devlink_flash_update,
1313 .rate_node_new = ice_devlink_rate_node_new,
1314 .rate_node_del = ice_devlink_rate_node_del,
1316 .rate_leaf_tx_max_set = ice_devlink_rate_leaf_tx_max_set,
1317 .rate_leaf_tx_share_set = ice_devlink_rate_leaf_tx_share_set,
1318 .rate_leaf_tx_priority_set = ice_devlink_rate_leaf_tx_priority_set,
1319 .rate_leaf_tx_weight_set = ice_devlink_rate_leaf_tx_weight_set,
1321 .rate_node_tx_max_set = ice_devlink_rate_node_tx_max_set,
1322 .rate_node_tx_share_set = ice_devlink_rate_node_tx_share_set,
1323 .rate_node_tx_priority_set = ice_devlink_rate_node_tx_priority_set,
1324 .rate_node_tx_weight_set = ice_devlink_rate_node_tx_weight_set,
1326 .rate_leaf_parent_set = ice_devlink_set_parent,
1327 .rate_node_parent_set = ice_devlink_set_parent,
1331 ice_devlink_enable_roce_get(struct devlink *devlink, u32 id,
1332 struct devlink_param_gset_ctx *ctx)
1334 struct ice_pf *pf = devlink_priv(devlink);
1336 ctx->val.vbool = pf->rdma_mode & IIDC_RDMA_PROTOCOL_ROCEV2 ? true : false;
1342 ice_devlink_enable_roce_set(struct devlink *devlink, u32 id,
1343 struct devlink_param_gset_ctx *ctx)
1345 struct ice_pf *pf = devlink_priv(devlink);
1346 bool roce_ena = ctx->val.vbool;
1350 ice_unplug_aux_dev(pf);
1351 pf->rdma_mode &= ~IIDC_RDMA_PROTOCOL_ROCEV2;
1355 pf->rdma_mode |= IIDC_RDMA_PROTOCOL_ROCEV2;
1356 ret = ice_plug_aux_dev(pf);
1358 pf->rdma_mode &= ~IIDC_RDMA_PROTOCOL_ROCEV2;
1364 ice_devlink_enable_roce_validate(struct devlink *devlink, u32 id,
1365 union devlink_param_value val,
1366 struct netlink_ext_ack *extack)
1368 struct ice_pf *pf = devlink_priv(devlink);
1370 if (!test_bit(ICE_FLAG_RDMA_ENA, pf->flags))
1373 if (pf->rdma_mode & IIDC_RDMA_PROTOCOL_IWARP) {
1374 NL_SET_ERR_MSG_MOD(extack, "iWARP is currently enabled. This device cannot enable iWARP and RoCEv2 simultaneously");
1382 ice_devlink_enable_iw_get(struct devlink *devlink, u32 id,
1383 struct devlink_param_gset_ctx *ctx)
1385 struct ice_pf *pf = devlink_priv(devlink);
1387 ctx->val.vbool = pf->rdma_mode & IIDC_RDMA_PROTOCOL_IWARP;
1393 ice_devlink_enable_iw_set(struct devlink *devlink, u32 id,
1394 struct devlink_param_gset_ctx *ctx)
1396 struct ice_pf *pf = devlink_priv(devlink);
1397 bool iw_ena = ctx->val.vbool;
1401 ice_unplug_aux_dev(pf);
1402 pf->rdma_mode &= ~IIDC_RDMA_PROTOCOL_IWARP;
1406 pf->rdma_mode |= IIDC_RDMA_PROTOCOL_IWARP;
1407 ret = ice_plug_aux_dev(pf);
1409 pf->rdma_mode &= ~IIDC_RDMA_PROTOCOL_IWARP;
1415 ice_devlink_enable_iw_validate(struct devlink *devlink, u32 id,
1416 union devlink_param_value val,
1417 struct netlink_ext_ack *extack)
1419 struct ice_pf *pf = devlink_priv(devlink);
1421 if (!test_bit(ICE_FLAG_RDMA_ENA, pf->flags))
1424 if (pf->rdma_mode & IIDC_RDMA_PROTOCOL_ROCEV2) {
1425 NL_SET_ERR_MSG_MOD(extack, "RoCEv2 is currently enabled. This device cannot enable iWARP and RoCEv2 simultaneously");
1432 static const struct devlink_param ice_devlink_params[] = {
1433 DEVLINK_PARAM_GENERIC(ENABLE_ROCE, BIT(DEVLINK_PARAM_CMODE_RUNTIME),
1434 ice_devlink_enable_roce_get,
1435 ice_devlink_enable_roce_set,
1436 ice_devlink_enable_roce_validate),
1437 DEVLINK_PARAM_GENERIC(ENABLE_IWARP, BIT(DEVLINK_PARAM_CMODE_RUNTIME),
1438 ice_devlink_enable_iw_get,
1439 ice_devlink_enable_iw_set,
1440 ice_devlink_enable_iw_validate),
1444 static void ice_devlink_free(void *devlink_ptr)
1446 devlink_free((struct devlink *)devlink_ptr);
1450 * ice_allocate_pf - Allocate devlink and return PF structure pointer
1451 * @dev: the device to allocate for
1453 * Allocate a devlink instance for this device and return the private area as
1454 * the PF structure. The devlink memory is kept track of through devres by
1455 * adding an action to remove it when unwinding.
1457 struct ice_pf *ice_allocate_pf(struct device *dev)
1459 struct devlink *devlink;
1461 devlink = devlink_alloc(&ice_devlink_ops, sizeof(struct ice_pf), dev);
1465 /* Add an action to teardown the devlink when unwinding the driver */
1466 if (devm_add_action_or_reset(dev, ice_devlink_free, devlink))
1469 return devlink_priv(devlink);
1473 * ice_devlink_register - Register devlink interface for this PF
1474 * @pf: the PF to register the devlink for.
1476 * Register the devlink instance associated with this physical function.
1478 * Return: zero on success or an error code on failure.
1480 void ice_devlink_register(struct ice_pf *pf)
1482 struct devlink *devlink = priv_to_devlink(pf);
1484 devlink_register(devlink);
1488 * ice_devlink_unregister - Unregister devlink resources for this PF.
1489 * @pf: the PF structure to cleanup
1491 * Releases resources used by devlink and cleans up associated memory.
1493 void ice_devlink_unregister(struct ice_pf *pf)
1495 devlink_unregister(priv_to_devlink(pf));
1499 * ice_devlink_set_switch_id - Set unique switch id based on pci dsn
1500 * @pf: the PF to create a devlink port for
1501 * @ppid: struct with switch id information
1504 ice_devlink_set_switch_id(struct ice_pf *pf, struct netdev_phys_item_id *ppid)
1506 struct pci_dev *pdev = pf->pdev;
1509 id = pci_get_dsn(pdev);
1511 ppid->id_len = sizeof(id);
1512 put_unaligned_be64(id, &ppid->id);
1515 int ice_devlink_register_params(struct ice_pf *pf)
1517 struct devlink *devlink = priv_to_devlink(pf);
1519 return devlink_params_register(devlink, ice_devlink_params,
1520 ARRAY_SIZE(ice_devlink_params));
1523 void ice_devlink_unregister_params(struct ice_pf *pf)
1525 devlink_params_unregister(priv_to_devlink(pf), ice_devlink_params,
1526 ARRAY_SIZE(ice_devlink_params));
1530 * ice_devlink_set_port_split_options - Set port split options
1531 * @pf: the PF to set port split options
1532 * @attrs: devlink attributes
1534 * Sets devlink port split options based on available FW port options
1537 ice_devlink_set_port_split_options(struct ice_pf *pf,
1538 struct devlink_port_attrs *attrs)
1540 struct ice_aqc_get_port_options_elem options[ICE_AQC_PORT_OPT_MAX];
1541 u8 i, active_idx, pending_idx, option_count = ICE_AQC_PORT_OPT_MAX;
1542 bool active_valid, pending_valid;
1545 status = ice_aq_get_port_options(&pf->hw, options, &option_count,
1546 0, true, &active_idx, &active_valid,
1547 &pending_idx, &pending_valid);
1549 dev_dbg(ice_pf_to_dev(pf), "Couldn't read port split options, err = %d\n",
1554 /* find the biggest available port split count */
1555 for (i = 0; i < option_count; i++)
1556 attrs->lanes = max_t(int, attrs->lanes, options[i].pmd);
1558 attrs->splittable = attrs->lanes ? 1 : 0;
1559 ice_active_port_option = active_idx;
1562 static const struct devlink_port_ops ice_devlink_port_ops = {
1563 .port_split = ice_devlink_port_split,
1564 .port_unsplit = ice_devlink_port_unsplit,
1568 * ice_devlink_create_pf_port - Create a devlink port for this PF
1569 * @pf: the PF to create a devlink port for
1571 * Create and register a devlink_port for this PF.
1573 * Return: zero on success or an error code on failure.
1575 int ice_devlink_create_pf_port(struct ice_pf *pf)
1577 struct devlink_port_attrs attrs = {};
1578 struct devlink_port *devlink_port;
1579 struct devlink *devlink;
1580 struct ice_vsi *vsi;
1584 dev = ice_pf_to_dev(pf);
1586 devlink_port = &pf->devlink_port;
1588 vsi = ice_get_main_vsi(pf);
1592 attrs.flavour = DEVLINK_PORT_FLAVOUR_PHYSICAL;
1593 attrs.phys.port_number = pf->hw.bus.func;
1595 /* As FW supports only port split options for whole device,
1596 * set port split options only for first PF.
1598 if (pf->hw.pf_id == 0)
1599 ice_devlink_set_port_split_options(pf, &attrs);
1601 ice_devlink_set_switch_id(pf, &attrs.switch_id);
1603 devlink_port_attrs_set(devlink_port, &attrs);
1604 devlink = priv_to_devlink(pf);
1606 err = devlink_port_register_with_ops(devlink, devlink_port, vsi->idx,
1607 &ice_devlink_port_ops);
1609 dev_err(dev, "Failed to create devlink port for PF %d, error %d\n",
1618 * ice_devlink_destroy_pf_port - Destroy the devlink_port for this PF
1619 * @pf: the PF to cleanup
1621 * Unregisters the devlink_port structure associated with this PF.
1623 void ice_devlink_destroy_pf_port(struct ice_pf *pf)
1625 devlink_port_unregister(&pf->devlink_port);
1629 * ice_devlink_create_vf_port - Create a devlink port for this VF
1630 * @vf: the VF to create a port for
1632 * Create and register a devlink_port for this VF.
1634 * Return: zero on success or an error code on failure.
1636 int ice_devlink_create_vf_port(struct ice_vf *vf)
1638 struct devlink_port_attrs attrs = {};
1639 struct devlink_port *devlink_port;
1640 struct devlink *devlink;
1641 struct ice_vsi *vsi;
1647 dev = ice_pf_to_dev(pf);
1648 devlink_port = &vf->devlink_port;
1650 vsi = ice_get_vf_vsi(vf);
1654 attrs.flavour = DEVLINK_PORT_FLAVOUR_PCI_VF;
1655 attrs.pci_vf.pf = pf->hw.bus.func;
1656 attrs.pci_vf.vf = vf->vf_id;
1658 ice_devlink_set_switch_id(pf, &attrs.switch_id);
1660 devlink_port_attrs_set(devlink_port, &attrs);
1661 devlink = priv_to_devlink(pf);
1663 err = devlink_port_register(devlink, devlink_port, vsi->idx);
1665 dev_err(dev, "Failed to create devlink port for VF %d, error %d\n",
1674 * ice_devlink_destroy_vf_port - Destroy the devlink_port for this VF
1675 * @vf: the VF to cleanup
1677 * Unregisters the devlink_port structure associated with this VF.
1679 void ice_devlink_destroy_vf_port(struct ice_vf *vf)
1681 devl_rate_leaf_destroy(&vf->devlink_port);
1682 devlink_port_unregister(&vf->devlink_port);
1685 #define ICE_DEVLINK_READ_BLK_SIZE (1024 * 1024)
1687 static const struct devlink_region_ops ice_nvm_region_ops;
1688 static const struct devlink_region_ops ice_sram_region_ops;
1691 * ice_devlink_nvm_snapshot - Capture a snapshot of the NVM flash contents
1692 * @devlink: the devlink instance
1693 * @ops: the devlink region to snapshot
1694 * @extack: extended ACK response structure
1695 * @data: on exit points to snapshot data buffer
1697 * This function is called in response to a DEVLINK_CMD_REGION_NEW for either
1698 * the nvm-flash or shadow-ram region.
1700 * It captures a snapshot of the NVM or Shadow RAM flash contents. This
1701 * snapshot can then later be viewed via the DEVLINK_CMD_REGION_READ netlink
1704 * @returns zero on success, and updates the data pointer. Returns a non-zero
1705 * error code on failure.
1707 static int ice_devlink_nvm_snapshot(struct devlink *devlink,
1708 const struct devlink_region_ops *ops,
1709 struct netlink_ext_ack *extack, u8 **data)
1711 struct ice_pf *pf = devlink_priv(devlink);
1712 struct device *dev = ice_pf_to_dev(pf);
1713 struct ice_hw *hw = &pf->hw;
1714 bool read_shadow_ram;
1715 u8 *nvm_data, *tmp, i;
1720 if (ops == &ice_nvm_region_ops) {
1721 read_shadow_ram = false;
1722 nvm_size = hw->flash.flash_size;
1723 } else if (ops == &ice_sram_region_ops) {
1724 read_shadow_ram = true;
1725 nvm_size = hw->flash.sr_words * 2u;
1727 NL_SET_ERR_MSG_MOD(extack, "Unexpected region in snapshot function");
1731 nvm_data = vzalloc(nvm_size);
1735 num_blks = DIV_ROUND_UP(nvm_size, ICE_DEVLINK_READ_BLK_SIZE);
1739 /* Some systems take longer to read the NVM than others which causes the
1740 * FW to reclaim the NVM lock before the entire NVM has been read. Fix
1741 * this by breaking the reads of the NVM into smaller chunks that will
1742 * probably not take as long. This has some overhead since we are
1743 * increasing the number of AQ commands, but it should always work
1745 for (i = 0; i < num_blks; i++) {
1746 u32 read_sz = min_t(u32, ICE_DEVLINK_READ_BLK_SIZE, left);
1748 status = ice_acquire_nvm(hw, ICE_RES_READ);
1750 dev_dbg(dev, "ice_acquire_nvm failed, err %d aq_err %d\n",
1751 status, hw->adminq.sq_last_status);
1752 NL_SET_ERR_MSG_MOD(extack, "Failed to acquire NVM semaphore");
1757 status = ice_read_flat_nvm(hw, i * ICE_DEVLINK_READ_BLK_SIZE,
1758 &read_sz, tmp, read_shadow_ram);
1760 dev_dbg(dev, "ice_read_flat_nvm failed after reading %u bytes, err %d aq_err %d\n",
1761 read_sz, status, hw->adminq.sq_last_status);
1762 NL_SET_ERR_MSG_MOD(extack, "Failed to read NVM contents");
1763 ice_release_nvm(hw);
1767 ice_release_nvm(hw);
1779 * ice_devlink_nvm_read - Read a portion of NVM flash contents
1780 * @devlink: the devlink instance
1781 * @ops: the devlink region to snapshot
1782 * @extack: extended ACK response structure
1783 * @offset: the offset to start at
1784 * @size: the amount to read
1785 * @data: the data buffer to read into
1787 * This function is called in response to DEVLINK_CMD_REGION_READ to directly
1788 * read a section of the NVM contents.
1790 * It reads from either the nvm-flash or shadow-ram region contents.
1792 * @returns zero on success, and updates the data pointer. Returns a non-zero
1793 * error code on failure.
1795 static int ice_devlink_nvm_read(struct devlink *devlink,
1796 const struct devlink_region_ops *ops,
1797 struct netlink_ext_ack *extack,
1798 u64 offset, u32 size, u8 *data)
1800 struct ice_pf *pf = devlink_priv(devlink);
1801 struct device *dev = ice_pf_to_dev(pf);
1802 struct ice_hw *hw = &pf->hw;
1803 bool read_shadow_ram;
1807 if (ops == &ice_nvm_region_ops) {
1808 read_shadow_ram = false;
1809 nvm_size = hw->flash.flash_size;
1810 } else if (ops == &ice_sram_region_ops) {
1811 read_shadow_ram = true;
1812 nvm_size = hw->flash.sr_words * 2u;
1814 NL_SET_ERR_MSG_MOD(extack, "Unexpected region in snapshot function");
1818 if (offset + size >= nvm_size) {
1819 NL_SET_ERR_MSG_MOD(extack, "Cannot read beyond the region size");
1823 status = ice_acquire_nvm(hw, ICE_RES_READ);
1825 dev_dbg(dev, "ice_acquire_nvm failed, err %d aq_err %d\n",
1826 status, hw->adminq.sq_last_status);
1827 NL_SET_ERR_MSG_MOD(extack, "Failed to acquire NVM semaphore");
1831 status = ice_read_flat_nvm(hw, (u32)offset, &size, data,
1834 dev_dbg(dev, "ice_read_flat_nvm failed after reading %u bytes, err %d aq_err %d\n",
1835 size, status, hw->adminq.sq_last_status);
1836 NL_SET_ERR_MSG_MOD(extack, "Failed to read NVM contents");
1837 ice_release_nvm(hw);
1840 ice_release_nvm(hw);
1846 * ice_devlink_devcaps_snapshot - Capture snapshot of device capabilities
1847 * @devlink: the devlink instance
1848 * @ops: the devlink region being snapshotted
1849 * @extack: extended ACK response structure
1850 * @data: on exit points to snapshot data buffer
1852 * This function is called in response to the DEVLINK_CMD_REGION_TRIGGER for
1853 * the device-caps devlink region. It captures a snapshot of the device
1854 * capabilities reported by firmware.
1856 * @returns zero on success, and updates the data pointer. Returns a non-zero
1857 * error code on failure.
1860 ice_devlink_devcaps_snapshot(struct devlink *devlink,
1861 const struct devlink_region_ops *ops,
1862 struct netlink_ext_ack *extack, u8 **data)
1864 struct ice_pf *pf = devlink_priv(devlink);
1865 struct device *dev = ice_pf_to_dev(pf);
1866 struct ice_hw *hw = &pf->hw;
1870 devcaps = vzalloc(ICE_AQ_MAX_BUF_LEN);
1874 status = ice_aq_list_caps(hw, devcaps, ICE_AQ_MAX_BUF_LEN, NULL,
1875 ice_aqc_opc_list_dev_caps, NULL);
1877 dev_dbg(dev, "ice_aq_list_caps: failed to read device capabilities, err %d aq_err %d\n",
1878 status, hw->adminq.sq_last_status);
1879 NL_SET_ERR_MSG_MOD(extack, "Failed to read device capabilities");
1884 *data = (u8 *)devcaps;
1889 static const struct devlink_region_ops ice_nvm_region_ops = {
1890 .name = "nvm-flash",
1891 .destructor = vfree,
1892 .snapshot = ice_devlink_nvm_snapshot,
1893 .read = ice_devlink_nvm_read,
1896 static const struct devlink_region_ops ice_sram_region_ops = {
1897 .name = "shadow-ram",
1898 .destructor = vfree,
1899 .snapshot = ice_devlink_nvm_snapshot,
1900 .read = ice_devlink_nvm_read,
1903 static const struct devlink_region_ops ice_devcaps_region_ops = {
1904 .name = "device-caps",
1905 .destructor = vfree,
1906 .snapshot = ice_devlink_devcaps_snapshot,
1910 * ice_devlink_init_regions - Initialize devlink regions
1911 * @pf: the PF device structure
1913 * Create devlink regions used to enable access to dump the contents of the
1914 * flash memory on the device.
1916 void ice_devlink_init_regions(struct ice_pf *pf)
1918 struct devlink *devlink = priv_to_devlink(pf);
1919 struct device *dev = ice_pf_to_dev(pf);
1920 u64 nvm_size, sram_size;
1922 nvm_size = pf->hw.flash.flash_size;
1923 pf->nvm_region = devlink_region_create(devlink, &ice_nvm_region_ops, 1,
1925 if (IS_ERR(pf->nvm_region)) {
1926 dev_err(dev, "failed to create NVM devlink region, err %ld\n",
1927 PTR_ERR(pf->nvm_region));
1928 pf->nvm_region = NULL;
1931 sram_size = pf->hw.flash.sr_words * 2u;
1932 pf->sram_region = devlink_region_create(devlink, &ice_sram_region_ops,
1934 if (IS_ERR(pf->sram_region)) {
1935 dev_err(dev, "failed to create shadow-ram devlink region, err %ld\n",
1936 PTR_ERR(pf->sram_region));
1937 pf->sram_region = NULL;
1940 pf->devcaps_region = devlink_region_create(devlink,
1941 &ice_devcaps_region_ops, 10,
1942 ICE_AQ_MAX_BUF_LEN);
1943 if (IS_ERR(pf->devcaps_region)) {
1944 dev_err(dev, "failed to create device-caps devlink region, err %ld\n",
1945 PTR_ERR(pf->devcaps_region));
1946 pf->devcaps_region = NULL;
1951 * ice_devlink_destroy_regions - Destroy devlink regions
1952 * @pf: the PF device structure
1954 * Remove previously created regions for this PF.
1956 void ice_devlink_destroy_regions(struct ice_pf *pf)
1959 devlink_region_destroy(pf->nvm_region);
1961 if (pf->sram_region)
1962 devlink_region_destroy(pf->sram_region);
1964 if (pf->devcaps_region)
1965 devlink_region_destroy(pf->devcaps_region);