--- /dev/null
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2025 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "ras.h"
+#include "ras_umc.h"
+#include "ras_umc_v12_0.h"
+
+#define MAX_ECC_NUM_PER_RETIREMENT 16
+
+/* bad page timestamp format
+ * yy[31:27] mm[26:23] day[22:17] hh[16:12] mm[11:6] ss[5:0]
+ */
+#define EEPROM_TIMESTAMP_MINUTE 6
+#define EEPROM_TIMESTAMP_HOUR 12
+#define EEPROM_TIMESTAMP_DAY 17
+#define EEPROM_TIMESTAMP_MONTH 23
+#define EEPROM_TIMESTAMP_YEAR 27
+
+static uint64_t ras_umc_get_eeprom_timestamp(struct ras_core_context *ras_core)
+{
+ struct ras_time tm = {0};
+ uint64_t utc_timestamp = 0;
+ uint64_t eeprom_timestamp = 0;
+
+ utc_timestamp = ras_core_get_utc_second_timestamp(ras_core);
+ if (!utc_timestamp)
+ return utc_timestamp;
+
+ ras_core_convert_timestamp_to_time(ras_core, utc_timestamp, &tm);
+
+ /* the year range is 2000 ~ 2031, set the year if not in the range */
+ if (tm.tm_year < 2000)
+ tm.tm_year = 2000;
+ if (tm.tm_year > 2031)
+ tm.tm_year = 2031;
+
+ tm.tm_year -= 2000;
+
+ eeprom_timestamp = tm.tm_sec + (tm.tm_min << EEPROM_TIMESTAMP_MINUTE)
+ + (tm.tm_hour << EEPROM_TIMESTAMP_HOUR)
+ + (tm.tm_mday << EEPROM_TIMESTAMP_DAY)
+ + (tm.tm_mon << EEPROM_TIMESTAMP_MONTH)
+ + (tm.tm_year << EEPROM_TIMESTAMP_YEAR);
+ eeprom_timestamp &= 0xffffffff;
+
+ return eeprom_timestamp;
+}
+
+static const struct ras_umc_ip_func *ras_umc_get_ip_func(
+ struct ras_core_context *ras_core, uint32_t ip_version)
+{
+ switch (ip_version) {
+ case IP_VERSION(12, 0, 0):
+ return &ras_umc_func_v12_0;
+ default:
+ RAS_DEV_ERR(ras_core->dev,
+ "UMC ip version(0x%x) is not supported!\n", ip_version);
+ break;
+ }
+
+ return NULL;
+}
+
+int ras_umc_psp_convert_ma_to_pa(struct ras_core_context *ras_core,
+ struct umc_mca_addr *in, struct umc_phy_addr *out,
+ uint32_t nps)
+{
+ struct ras_ta_query_address_input addr_in;
+ struct ras_ta_query_address_output addr_out;
+ int ret;
+
+ if (!in)
+ return -EINVAL;
+
+ memset(&addr_in, 0, sizeof(addr_in));
+ memset(&addr_out, 0, sizeof(addr_out));
+
+ addr_in.ma.err_addr = in->err_addr;
+ addr_in.ma.ch_inst = in->ch_inst;
+ addr_in.ma.umc_inst = in->umc_inst;
+ addr_in.ma.node_inst = in->node_inst;
+ addr_in.ma.socket_id = in->socket_id;
+
+ addr_in.addr_type = RAS_TA_MCA_TO_PA;
+
+ ret = ras_psp_query_address(ras_core, &addr_in, &addr_out);
+ if (ret) {
+ RAS_DEV_WARN(ras_core->dev,
+ "Failed to query RAS physical address for 0x%llx, ret:%d",
+ in->err_addr, ret);
+ return -EREMOTEIO;
+ }
+
+ if (out) {
+ out->pa = addr_out.pa.pa;
+ out->bank = addr_out.pa.bank;
+ out->channel_idx = addr_out.pa.channel_idx;
+ }
+
+ return 0;
+}
+
+static int ras_umc_log_ecc(struct ras_core_context *ras_core,
+ unsigned long idx, void *data)
+{
+ struct ras_umc *ras_umc = &ras_core->ras_umc;
+ int ret;
+
+ mutex_lock(&ras_umc->tree_lock);
+ ret = radix_tree_insert(&ras_umc->root, idx, data);
+ if (!ret)
+ radix_tree_tag_set(&ras_umc->root, idx, UMC_ECC_NEW_DETECTED_TAG);
+ mutex_unlock(&ras_umc->tree_lock);
+
+ return ret;
+}
+
+int ras_umc_clear_logged_ecc(struct ras_core_context *ras_core)
+{
+ struct ras_umc *ras_umc = &ras_core->ras_umc;
+ uint64_t buf[8] = {0};
+ void **slot;
+ void *data;
+ void *iter = buf;
+
+ mutex_lock(&ras_umc->tree_lock);
+ radix_tree_for_each_slot(slot, &ras_umc->root, iter, 0) {
+ data = ras_radix_tree_delete_iter(&ras_umc->root, iter);
+ kfree(data);
+ }
+ mutex_unlock(&ras_umc->tree_lock);
+
+ return 0;
+}
+
+static void ras_umc_reserve_eeprom_record(struct ras_core_context *ras_core,
+ struct eeprom_umc_record *record)
+{
+ struct ras_umc *ras_umc = &ras_core->ras_umc;
+ uint64_t page_pfn[16];
+ int count = 0, i;
+
+ memset(page_pfn, 0, sizeof(page_pfn));
+ if (ras_umc->ip_func && ras_umc->ip_func->eeprom_record_to_nps_pages) {
+ count = ras_umc->ip_func->eeprom_record_to_nps_pages(ras_core,
+ record, record->cur_nps, page_pfn, ARRAY_SIZE(page_pfn));
+ if (count <= 0) {
+ RAS_DEV_ERR(ras_core->dev,
+ "Fail to convert error address! count:%d\n", count);
+ return;
+ }
+ }
+
+ /* Reserve memory */
+ for (i = 0; i < count; i++)
+ ras_core_event_notify(ras_core,
+ RAS_EVENT_ID__RESERVE_BAD_PAGE, &page_pfn[i]);
+}
+
+/* When gpu reset is ongoing, ecc logging operations will be pended.
+ */
+int ras_umc_log_bad_bank_pending(struct ras_core_context *ras_core, struct ras_bank_ecc *bank)
+{
+ struct ras_umc *ras_umc = &ras_core->ras_umc;
+ struct ras_bank_ecc_node *ecc_node;
+
+ ecc_node = kzalloc(sizeof(*ecc_node), GFP_KERNEL);
+ if (!ecc_node)
+ return -ENOMEM;
+
+ memcpy(&ecc_node->ecc, bank, sizeof(ecc_node->ecc));
+
+ mutex_lock(&ras_umc->pending_ecc_lock);
+ list_add_tail(&ecc_node->node, &ras_umc->pending_ecc_list);
+ mutex_unlock(&ras_umc->pending_ecc_lock);
+
+ return 0;
+}
+
+/* After gpu reset is complete, re-log the pending error banks.
+ */
+int ras_umc_log_pending_bad_bank(struct ras_core_context *ras_core)
+{
+ struct ras_umc *ras_umc = &ras_core->ras_umc;
+ struct ras_bank_ecc_node *ecc_node, *tmp;
+
+ mutex_lock(&ras_umc->pending_ecc_lock);
+ list_for_each_entry_safe(ecc_node,
+ tmp, &ras_umc->pending_ecc_list, node){
+ if (ecc_node && !ras_umc_log_bad_bank(ras_core, &ecc_node->ecc)) {
+ list_del(&ecc_node->node);
+ kfree(ecc_node);
+ }
+ }
+ mutex_unlock(&ras_umc->pending_ecc_lock);
+
+ return 0;
+}
+
+int ras_umc_log_bad_bank(struct ras_core_context *ras_core, struct ras_bank_ecc *bank)
+{
+ struct ras_umc *ras_umc = &ras_core->ras_umc;
+ struct eeprom_umc_record umc_rec;
+ struct eeprom_umc_record *err_rec;
+ int ret;
+
+ memset(&umc_rec, 0, sizeof(umc_rec));
+
+ mutex_lock(&ras_umc->bank_log_lock);
+ ret = ras_umc->ip_func->bank_to_eeprom_record(ras_core, bank, &umc_rec);
+ if (ret)
+ goto out;
+
+ err_rec = kzalloc(sizeof(*err_rec), GFP_KERNEL);
+ if (!err_rec) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ memcpy(err_rec, &umc_rec, sizeof(umc_rec));
+ ret = ras_umc_log_ecc(ras_core, err_rec->cur_nps_retired_row_pfn, err_rec);
+ if (ret) {
+ if (ret == -EEXIST) {
+ RAS_DEV_INFO(ras_core->dev, "The bad pages have been logged before.\n");
+ ret = 0;
+ }
+
+ kfree(err_rec);
+ goto out;
+ }
+
+ ras_umc_reserve_eeprom_record(ras_core, err_rec);
+
+ ret = ras_core_event_notify(ras_core,
+ RAS_EVENT_ID__BAD_PAGE_DETECTED, NULL);
+
+out:
+ mutex_unlock(&ras_umc->bank_log_lock);
+ return ret;
+}
+
+static int ras_umc_get_new_records(struct ras_core_context *ras_core,
+ struct eeprom_umc_record *records, u32 num)
+{
+ struct ras_umc *ras_umc = &ras_core->ras_umc;
+ struct eeprom_umc_record *entries[MAX_ECC_NUM_PER_RETIREMENT];
+ u32 entry_num = num < MAX_ECC_NUM_PER_RETIREMENT ? num : MAX_ECC_NUM_PER_RETIREMENT;
+ int count = 0;
+ int new_detected, i;
+
+ mutex_lock(&ras_umc->tree_lock);
+ new_detected = radix_tree_gang_lookup_tag(&ras_umc->root, (void **)entries,
+ 0, entry_num, UMC_ECC_NEW_DETECTED_TAG);
+ for (i = 0; i < new_detected; i++) {
+ if (!entries[i])
+ continue;
+
+ memcpy(&records[i], entries[i], sizeof(struct eeprom_umc_record));
+ count++;
+ radix_tree_tag_clear(&ras_umc->root,
+ entries[i]->cur_nps_retired_row_pfn, UMC_ECC_NEW_DETECTED_TAG);
+ }
+ mutex_unlock(&ras_umc->tree_lock);
+
+ return count;
+}
+
+static bool ras_umc_check_retired_record(struct ras_core_context *ras_core,
+ struct eeprom_umc_record *record, bool from_eeprom)
+{
+ struct ras_umc *ras_umc = &ras_core->ras_umc;
+ struct eeprom_store_record *data = &ras_umc->umc_err_data.rom_data;
+ uint32_t nps = 0;
+ int i, ret;
+
+ if (from_eeprom) {
+ nps = ras_umc->umc_err_data.umc_nps_mode;
+ if (ras_umc->ip_func && ras_umc->ip_func->eeprom_record_to_nps_record) {
+ ret = ras_umc->ip_func->eeprom_record_to_nps_record(ras_core, record, nps);
+ if (ret)
+ RAS_DEV_WARN(ras_core->dev,
+ "Failed to adjust eeprom record, ret:%d", ret);
+ }
+ return false;
+ }
+
+ for (i = 0; i < data->count; i++) {
+ if ((data->bps[i].retired_row_pfn == record->retired_row_pfn) &&
+ (data->bps[i].cur_nps_retired_row_pfn == record->cur_nps_retired_row_pfn))
+ return true;
+ }
+
+ return false;
+}
+
+/* alloc/realloc bps array */
+static int ras_umc_realloc_err_data_space(struct ras_core_context *ras_core,
+ struct eeprom_store_record *data, int pages)
+{
+ unsigned int old_space = data->count + data->space_left;
+ unsigned int new_space = old_space + pages;
+ unsigned int align_space = ALIGN(new_space, 512);
+ void *bps = kzalloc(align_space * sizeof(*data->bps), GFP_KERNEL);
+
+ if (!bps)
+ return -ENOMEM;
+
+ if (data->bps) {
+ memcpy(bps, data->bps,
+ data->count * sizeof(*data->bps));
+ kfree(data->bps);
+ }
+
+ data->bps = bps;
+ data->space_left += align_space - old_space;
+ return 0;
+}
+
+static int ras_umc_update_eeprom_rom_data(struct ras_core_context *ras_core,
+ struct eeprom_umc_record *bps)
+{
+ struct eeprom_store_record *data = &ras_core->ras_umc.umc_err_data.rom_data;
+
+ if (!data->space_left &&
+ ras_umc_realloc_err_data_space(ras_core, data, 256)) {
+ return -ENOMEM;
+ }
+
+ memcpy(&data->bps[data->count], bps, sizeof(*data->bps));
+ data->count++;
+ data->space_left--;
+ return 0;
+}
+
+static int ras_umc_update_eeprom_ram_data(struct ras_core_context *ras_core,
+ struct eeprom_umc_record *bps)
+{
+ struct ras_umc *ras_umc = &ras_core->ras_umc;
+ struct eeprom_store_record *data = &ras_umc->umc_err_data.ram_data;
+ uint64_t page_pfn[16];
+ int count = 0, j;
+
+ if (!data->space_left &&
+ ras_umc_realloc_err_data_space(ras_core, data, 256)) {
+ return -ENOMEM;
+ }
+
+ memset(page_pfn, 0, sizeof(page_pfn));
+ if (ras_umc->ip_func && ras_umc->ip_func->eeprom_record_to_nps_pages)
+ count = ras_umc->ip_func->eeprom_record_to_nps_pages(ras_core,
+ bps, bps->cur_nps, page_pfn, ARRAY_SIZE(page_pfn));
+
+ if (count > 0) {
+ for (j = 0; j < count; j++) {
+ bps->cur_nps_retired_row_pfn = page_pfn[j];
+ memcpy(&data->bps[data->count], bps, sizeof(*data->bps));
+ data->count++;
+ data->space_left--;
+ }
+ } else {
+ memcpy(&data->bps[data->count], bps, sizeof(*data->bps));
+ data->count++;
+ data->space_left--;
+ }
+
+ return 0;
+}
+
+/* it deal with vram only. */
+static int ras_umc_add_bad_pages(struct ras_core_context *ras_core,
+ struct eeprom_umc_record *bps,
+ int pages, bool from_eeprom)
+{
+ struct ras_umc *ras_umc = &ras_core->ras_umc;
+ struct ras_umc_err_data *data = &ras_umc->umc_err_data;
+ int i, ret = 0;
+
+ if (!bps || pages <= 0)
+ return 0;
+
+ mutex_lock(&ras_umc->umc_lock);
+ for (i = 0; i < pages; i++) {
+ if (ras_umc_check_retired_record(ras_core, &bps[i], from_eeprom))
+ continue;
+
+ ret = ras_umc_update_eeprom_rom_data(ras_core, &bps[i]);
+ if (ret)
+ goto out;
+
+ if (data->last_retired_pfn == bps[i].cur_nps_retired_row_pfn)
+ continue;
+
+ data->last_retired_pfn = bps[i].cur_nps_retired_row_pfn;
+
+ if (from_eeprom)
+ ras_umc_reserve_eeprom_record(ras_core, &bps[i]);
+
+ ret = ras_umc_update_eeprom_ram_data(ras_core, &bps[i]);
+ if (ret)
+ goto out;
+ }
+out:
+ mutex_unlock(&ras_umc->umc_lock);
+
+ return ret;
+}
+
+/*
+ * read error record array in eeprom and reserve enough space for
+ * storing new bad pages
+ */
+int ras_umc_load_bad_pages(struct ras_core_context *ras_core)
+{
+ struct eeprom_umc_record *bps;
+ uint32_t ras_num_recs;
+ int ret;
+
+ ras_num_recs = ras_eeprom_get_record_count(ras_core);
+ /* no bad page record, skip eeprom access */
+ if (!ras_num_recs ||
+ ras_core->ras_eeprom.record_threshold_config == DISABLE_RETIRE_PAGE)
+ return 0;
+
+ bps = kcalloc(ras_num_recs, sizeof(*bps), GFP_KERNEL);
+ if (!bps)
+ return -ENOMEM;
+
+ ret = ras_eeprom_read(ras_core, bps, ras_num_recs);
+ if (ret) {
+ RAS_DEV_ERR(ras_core->dev, "Failed to load EEPROM table records!");
+ } else {
+ ras_core->ras_umc.umc_err_data.last_retired_pfn = UMC_INV_MEM_PFN;
+ ret = ras_umc_add_bad_pages(ras_core, bps, ras_num_recs, true);
+ }
+
+ kfree(bps);
+ return ret;
+}
+
+/*
+ * write error record array to eeprom, the function should be
+ * protected by recovery_lock
+ * new_cnt: new added UE count, excluding reserved bad pages, can be NULL
+ */
+static int ras_umc_save_bad_pages(struct ras_core_context *ras_core)
+{
+ struct ras_umc *ras_umc = &ras_core->ras_umc;
+ struct eeprom_store_record *data = &ras_umc->umc_err_data.rom_data;
+ uint32_t eeprom_record_num;
+ int save_count;
+ int ret = 0;
+
+ if (!data->bps)
+ return 0;
+
+ eeprom_record_num = ras_eeprom_get_record_count(ras_core);
+ mutex_lock(&ras_umc->umc_lock);
+ save_count = data->count - eeprom_record_num;
+ /* only new entries are saved */
+ if (save_count > 0) {
+ if (ras_eeprom_append(ras_core,
+ &data->bps[eeprom_record_num],
+ save_count)) {
+ RAS_DEV_ERR(ras_core->dev, "Failed to save EEPROM table data!");
+ ret = -EIO;
+ goto exit;
+ }
+
+ RAS_DEV_INFO(ras_core->dev, "Saved %d pages to EEPROM table.\n", save_count);
+ }
+
+exit:
+ mutex_unlock(&ras_umc->umc_lock);
+ return ret;
+}
+
+int ras_umc_handle_bad_pages(struct ras_core_context *ras_core, void *data)
+{
+ struct eeprom_umc_record records[MAX_ECC_NUM_PER_RETIREMENT];
+ int count, ret;
+
+ memset(records, 0, sizeof(records));
+ count = ras_umc_get_new_records(ras_core, records, ARRAY_SIZE(records));
+ if (count <= 0)
+ return -ENODATA;
+
+ ret = ras_umc_add_bad_pages(ras_core, records, count, false);
+ if (ret) {
+ RAS_DEV_ERR(ras_core->dev, "Failed to add ras bad page!\n");
+ return -EINVAL;
+ }
+
+ ret = ras_umc_save_bad_pages(ras_core);
+ if (ret) {
+ RAS_DEV_ERR(ras_core->dev, "Failed to save ras bad page\n");
+ return -EINVAL;
+ }
+
+ return 0;
+}
+
+int ras_umc_sw_init(struct ras_core_context *ras_core)
+{
+ struct ras_umc *ras_umc = &ras_core->ras_umc;
+
+ memset(ras_umc, 0, sizeof(*ras_umc));
+
+ INIT_LIST_HEAD(&ras_umc->pending_ecc_list);
+
+ INIT_RADIX_TREE(&ras_umc->root, GFP_KERNEL);
+
+ mutex_init(&ras_umc->tree_lock);
+ mutex_init(&ras_umc->pending_ecc_lock);
+ mutex_init(&ras_umc->umc_lock);
+ mutex_init(&ras_umc->bank_log_lock);
+
+ return 0;
+}
+
+int ras_umc_sw_fini(struct ras_core_context *ras_core)
+{
+ struct ras_umc *ras_umc = &ras_core->ras_umc;
+ struct ras_umc_err_data *umc_err_data = &ras_umc->umc_err_data;
+ struct ras_bank_ecc_node *ecc_node, *tmp;
+
+ mutex_destroy(&ras_umc->umc_lock);
+ mutex_destroy(&ras_umc->bank_log_lock);
+
+ if (umc_err_data->rom_data.bps) {
+ umc_err_data->rom_data.count = 0;
+ kfree(umc_err_data->rom_data.bps);
+ umc_err_data->rom_data.bps = NULL;
+ umc_err_data->rom_data.space_left = 0;
+ }
+
+ if (umc_err_data->ram_data.bps) {
+ umc_err_data->ram_data.count = 0;
+ kfree(umc_err_data->ram_data.bps);
+ umc_err_data->ram_data.bps = NULL;
+ umc_err_data->ram_data.space_left = 0;
+ }
+
+ ras_umc_clear_logged_ecc(ras_core);
+
+ mutex_lock(&ras_umc->pending_ecc_lock);
+ list_for_each_entry_safe(ecc_node,
+ tmp, &ras_umc->pending_ecc_list, node){
+ list_del(&ecc_node->node);
+ kfree(ecc_node);
+ }
+ mutex_unlock(&ras_umc->pending_ecc_lock);
+
+ mutex_destroy(&ras_umc->tree_lock);
+ mutex_destroy(&ras_umc->pending_ecc_lock);
+
+ return 0;
+}
+
+int ras_umc_hw_init(struct ras_core_context *ras_core)
+{
+ struct ras_umc *ras_umc = &ras_core->ras_umc;
+ uint32_t nps;
+
+ nps = ras_core_get_curr_nps_mode(ras_core);
+
+ if (!nps || (nps >= UMC_MEMORY_PARTITION_MODE_UNKNOWN)) {
+ RAS_DEV_ERR(ras_core->dev, "Invalid memory NPS mode: %u!\n", nps);
+ return -ENODATA;
+ }
+
+ ras_umc->umc_err_data.umc_nps_mode = nps;
+
+ ras_umc->umc_vram_type = ras_core->config->umc_cfg.umc_vram_type;
+ if (!ras_umc->umc_vram_type) {
+ RAS_DEV_ERR(ras_core->dev, "Invalid UMC VRAM Type: %u!\n",
+ ras_umc->umc_vram_type);
+ return -ENODATA;
+ }
+
+ ras_umc->umc_ip_version = ras_core->config->umc_ip_version;
+ ras_umc->ip_func = ras_umc_get_ip_func(ras_core, ras_umc->umc_ip_version);
+ if (!ras_umc->ip_func)
+ return -EINVAL;
+
+ return 0;
+}
+
+int ras_umc_hw_fini(struct ras_core_context *ras_core)
+{
+ return 0;
+}
+
+int ras_umc_clean_badpage_data(struct ras_core_context *ras_core)
+{
+ struct ras_umc_err_data *data = &ras_core->ras_umc.umc_err_data;
+
+ mutex_lock(&ras_core->ras_umc.umc_lock);
+
+ kfree(data->rom_data.bps);
+ kfree(data->ram_data.bps);
+
+ memset(data, 0, sizeof(*data));
+ mutex_unlock(&ras_core->ras_umc.umc_lock);
+
+ return 0;
+}
+
+int ras_umc_fill_eeprom_record(struct ras_core_context *ras_core,
+ uint64_t err_addr, uint32_t umc_inst, struct umc_phy_addr *cur_nps_addr,
+ enum umc_memory_partition_mode cur_nps, struct eeprom_umc_record *record)
+{
+ struct eeprom_umc_record *err_rec = record;
+
+ /* Set bad page pfn and nps mode */
+ EEPROM_RECORD_SETUP_UMC_ADDR_AND_NPS(err_rec,
+ RAS_ADDR_TO_PFN(cur_nps_addr->pa), cur_nps);
+
+ err_rec->address = err_addr;
+ err_rec->ts = ras_umc_get_eeprom_timestamp(ras_core);
+ err_rec->err_type = RAS_EEPROM_ERR_NON_RECOVERABLE;
+ err_rec->cu = 0;
+ err_rec->mem_channel = cur_nps_addr->channel_idx;
+ err_rec->mcumc_id = umc_inst;
+ err_rec->cur_nps_retired_row_pfn = RAS_ADDR_TO_PFN(cur_nps_addr->pa);
+ err_rec->cur_nps_bank = cur_nps_addr->bank;
+ err_rec->cur_nps = cur_nps;
+ return 0;
+}
+
+int ras_umc_get_saved_eeprom_count(struct ras_core_context *ras_core)
+{
+ struct ras_umc_err_data *err_data = &ras_core->ras_umc.umc_err_data;
+
+ return err_data->rom_data.count;
+}
+
+int ras_umc_get_badpage_count(struct ras_core_context *ras_core)
+{
+ struct eeprom_store_record *data = &ras_core->ras_umc.umc_err_data.ram_data;
+
+ return data->count;
+}
+
+int ras_umc_get_badpage_record(struct ras_core_context *ras_core, uint32_t index, void *record)
+{
+ struct eeprom_store_record *data = &ras_core->ras_umc.umc_err_data.ram_data;
+
+ if (index >= data->count)
+ return -EINVAL;
+
+ memcpy(record, &data->bps[index], sizeof(struct eeprom_umc_record));
+ return 0;
+}
+
+bool ras_umc_check_retired_addr(struct ras_core_context *ras_core, uint64_t addr)
+{
+ struct ras_umc *ras_umc = &ras_core->ras_umc;
+ struct eeprom_store_record *data = &ras_umc->umc_err_data.ram_data;
+ uint64_t page_pfn = RAS_ADDR_TO_PFN(addr);
+ int i, ret = false;
+
+ mutex_lock(&ras_umc->umc_lock);
+ for (i = 0; i < data->count; i++) {
+ if (data->bps[i].cur_nps_retired_row_pfn == page_pfn) {
+ ret = true;
+ break;
+ }
+ }
+ mutex_unlock(&ras_umc->umc_lock);
+
+ return ret;
+}
+
+int ras_umc_translate_soc_pa_and_bank(struct ras_core_context *ras_core,
+ uint64_t *soc_pa, struct umc_bank_addr *bank_addr, bool bank_to_pa)
+{
+ struct ras_umc *ras_umc = &ras_core->ras_umc;
+ int ret = 0;
+
+ if (bank_to_pa)
+ ret = ras_umc->ip_func->bank_to_soc_pa(ras_core, *bank_addr, soc_pa);
+ else
+ ret = ras_umc->ip_func->soc_pa_to_bank(ras_core, *soc_pa, bank_addr);
+
+ return ret;
+}
--- /dev/null
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2025 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __RAS_UMC_H__
+#define __RAS_UMC_H__
+#include "ras.h"
+#include "ras_eeprom.h"
+#include "ras_cmd.h"
+
+#define UMC_VRAM_TYPE_UNKNOWN 0
+#define UMC_VRAM_TYPE_GDDR1 1
+#define UMC_VRAM_TYPE_DDR2 2
+#define UMC_VRAM_TYPE_GDDR3 3
+#define UMC_VRAM_TYPE_GDDR4 4
+#define UMC_VRAM_TYPE_GDDR5 5
+#define UMC_VRAM_TYPE_HBM 6
+#define UMC_VRAM_TYPE_DDR3 7
+#define UMC_VRAM_TYPE_DDR4 8
+#define UMC_VRAM_TYPE_GDDR6 9
+#define UMC_VRAM_TYPE_DDR5 10
+#define UMC_VRAM_TYPE_LPDDR4 11
+#define UMC_VRAM_TYPE_LPDDR5 12
+#define UMC_VRAM_TYPE_HBM3E 13
+
+#define UMC_ECC_NEW_DETECTED_TAG 0x1
+#define UMC_INV_MEM_PFN (0xFFFFFFFFFFFFFFFF)
+
+/* three column bits and one row bit in MCA address flip
+ * in bad page retirement
+ */
+#define UMC_PA_FLIP_BITS_NUM 4
+
+enum umc_memory_partition_mode {
+ UMC_MEMORY_PARTITION_MODE_NONE = 0,
+ UMC_MEMORY_PARTITION_MODE_NPS1 = 1,
+ UMC_MEMORY_PARTITION_MODE_NPS2 = 2,
+ UMC_MEMORY_PARTITION_MODE_NPS3 = 3,
+ UMC_MEMORY_PARTITION_MODE_NPS4 = 4,
+ UMC_MEMORY_PARTITION_MODE_NPS6 = 6,
+ UMC_MEMORY_PARTITION_MODE_NPS8 = 8,
+ UMC_MEMORY_PARTITION_MODE_UNKNOWN
+};
+
+struct ras_core_context;
+struct ras_bank_ecc;
+
+struct umc_flip_bits {
+ uint32_t flip_bits_in_pa[UMC_PA_FLIP_BITS_NUM];
+ uint32_t flip_row_bit;
+ uint32_t r13_in_pa;
+ uint32_t bit_num;
+};
+
+struct umc_mca_addr {
+ uint64_t err_addr;
+ uint32_t ch_inst;
+ uint32_t umc_inst;
+ uint32_t node_inst;
+ uint32_t socket_id;
+};
+
+struct umc_phy_addr {
+ uint64_t pa;
+ uint32_t bank;
+ uint32_t channel_idx;
+};
+
+struct umc_bank_addr {
+ uint32_t stack_id; /* SID */
+ uint32_t bank_group;
+ uint32_t bank;
+ uint32_t row;
+ uint32_t column;
+ uint32_t channel;
+ uint32_t subchannel; /* Also called Pseudochannel (PC) */
+};
+
+struct ras_umc_ip_func {
+ int (*bank_to_eeprom_record)(struct ras_core_context *ras_core,
+ struct ras_bank_ecc *bank, struct eeprom_umc_record *record);
+ int (*eeprom_record_to_nps_record)(struct ras_core_context *ras_core,
+ struct eeprom_umc_record *record, uint32_t nps);
+ int (*eeprom_record_to_nps_pages)(struct ras_core_context *ras_core,
+ struct eeprom_umc_record *record, uint32_t nps,
+ uint64_t *pfns, uint32_t num);
+ int (*bank_to_soc_pa)(struct ras_core_context *ras_core,
+ struct umc_bank_addr bank_addr, uint64_t *soc_pa);
+ int (*soc_pa_to_bank)(struct ras_core_context *ras_core,
+ uint64_t soc_pa, struct umc_bank_addr *bank_addr);
+};
+
+struct eeprom_store_record {
+ /* point to data records array */
+ struct eeprom_umc_record *bps;
+ /* the count of entries */
+ int count;
+ /* the space can place new entries */
+ int space_left;
+};
+
+struct ras_umc_err_data {
+ struct eeprom_store_record rom_data;
+ struct eeprom_store_record ram_data;
+ enum umc_memory_partition_mode umc_nps_mode;
+ uint64_t last_retired_pfn;
+};
+
+struct ras_umc {
+ u32 umc_ip_version;
+ u32 umc_vram_type;
+ const struct ras_umc_ip_func *ip_func;
+ struct radix_tree_root root;
+ struct mutex tree_lock;
+ struct mutex umc_lock;
+ struct mutex bank_log_lock;
+ struct mutex pending_ecc_lock;
+ struct ras_umc_err_data umc_err_data;
+ struct list_head pending_ecc_list;
+};
+
+int ras_umc_sw_init(struct ras_core_context *ras);
+int ras_umc_sw_fini(struct ras_core_context *ras);
+int ras_umc_hw_init(struct ras_core_context *ras);
+int ras_umc_hw_fini(struct ras_core_context *ras);
+int ras_umc_psp_convert_ma_to_pa(struct ras_core_context *ras_core,
+ struct umc_mca_addr *in, struct umc_phy_addr *out,
+ uint32_t nps);
+int ras_umc_handle_bad_pages(struct ras_core_context *ras_core, void *data);
+int ras_umc_log_bad_bank(struct ras_core_context *ras, struct ras_bank_ecc *bank);
+int ras_umc_log_bad_bank_pending(struct ras_core_context *ras_core, struct ras_bank_ecc *bank);
+int ras_umc_log_pending_bad_bank(struct ras_core_context *ras_core);
+int ras_umc_clear_logged_ecc(struct ras_core_context *ras_core);
+int ras_umc_load_bad_pages(struct ras_core_context *ras_core);
+int ras_umc_get_saved_eeprom_count(struct ras_core_context *ras_core);
+int ras_umc_clean_badpage_data(struct ras_core_context *ras_core);
+int ras_umc_fill_eeprom_record(struct ras_core_context *ras_core,
+ uint64_t err_addr, uint32_t umc_inst, struct umc_phy_addr *cur_nps_addr,
+ enum umc_memory_partition_mode cur_nps, struct eeprom_umc_record *record);
+
+int ras_umc_get_badpage_count(struct ras_core_context *ras_core);
+int ras_umc_get_badpage_record(struct ras_core_context *ras_core, uint32_t index, void *record);
+bool ras_umc_check_retired_addr(struct ras_core_context *ras_core, uint64_t addr);
+int ras_umc_translate_soc_pa_and_bank(struct ras_core_context *ras_core,
+ uint64_t *soc_pa, struct umc_bank_addr *bank_addr, bool bank_to_pa);
+#endif
projects / thirdparty / linux.git / commitdiff
