From: YiPeng Chai <[email protected]> Add interface to obtain FW eeprom records at runtime.
Signed-off-by: YiPeng Chai <[email protected]> Reviewed-by: Hawking Zhang <[email protected]> Signed-off-by: Alex Deucher <[email protected]> --- drivers/gpu/drm/amd/ras/core/aca.c | 4 ++- drivers/gpu/drm/amd/ras/core/ras_process.c | 3 ++ drivers/gpu/drm/amd/ras/core/ras_umc.c | 40 ++++++++++++++++++++++ drivers/gpu/drm/amd/ras/core/ras_umc.h | 1 + 4 files changed, 47 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/ras/core/aca.c b/drivers/gpu/drm/amd/ras/core/aca.c index 07a9cc030ca14..a2529645af098 100644 --- a/drivers/gpu/drm/amd/ras/core/aca.c +++ b/drivers/gpu/drm/amd/ras/core/aca.c @@ -247,7 +247,9 @@ static int aca_log_bad_bank(struct ras_core_context *ras_core, ecc_err->total_de_count += bank_ecc->de_count; mutex_unlock(&ras_core->ras_aca.aca_lock); - if ((aca_blk->blk_info->ras_block_id == RAS_BLOCK_ID__UMC) && + /* Log ACA bad bank when FW management EEPROM is disabled. */ + if (!ras_eeprom_mgr_fw_record_enabled(ras_core) && + (aca_blk->blk_info->ras_block_id == RAS_BLOCK_ID__UMC) && bank_ecc->de_count) { struct ras_bank_ecc ras_ecc = {0}; diff --git a/drivers/gpu/drm/amd/ras/core/ras_process.c b/drivers/gpu/drm/amd/ras/core/ras_process.c index c001074c8c568..dd4a4c5673bdd 100644 --- a/drivers/gpu/drm/amd/ras/core/ras_process.c +++ b/drivers/gpu/drm/amd/ras/core/ras_process.c @@ -170,6 +170,9 @@ int ras_process_handle_ras_event(struct ras_core_context *ras_core) ras_aca_clear_fatal_flag(ras_core); ras_umc_log_pending_bad_bank(ras_core); + if (ras_eeprom_mgr_fw_record_enabled(ras_core)) + ras_umc_dump_fw_records(ras_core); + do { umc_event_count = atomic_read(&ras_proc->umc_interrupt_count); ret = ras_process_umc_event(ras_core, umc_event_count); diff --git a/drivers/gpu/drm/amd/ras/core/ras_umc.c b/drivers/gpu/drm/amd/ras/core/ras_umc.c index 9cf1515dfd02e..b9961796a5395 100644 --- a/drivers/gpu/drm/amd/ras/core/ras_umc.c +++ b/drivers/gpu/drm/amd/ras/core/ras_umc.c @@ -675,6 +675,10 @@ static int ras_umc_save_bad_pages(struct ras_core_context *ras_core) int save_count; int ret = -ENODATA; + /* Not need to save bad pages when FW management EEPROM is enabled. */ + if (ras_eeprom_mgr_fw_record_enabled(ras_core)) + return 0; + if (!data->bps) return -EINVAL; @@ -918,3 +922,39 @@ int ras_umc_record_to_nps_record(struct ras_core_context *ras_core, return ras_umc_eeprom_rec2nps_rec(ras_core, record, nps); } + +int ras_umc_dump_fw_records(struct ras_core_context *ras_core) +{ + struct eeprom_umc_record rec; + int eeprom_count, umc_count, new_count = 0; + uint32_t c; + int i, ret; + + eeprom_count = ras_eeprom_mgr_get_record_count(ras_core); + /* no bad page record, skip eeprom access */ + if (!eeprom_count) + return 0; + + umc_count = ras_umc_get_saved_eeprom_count(ras_core); + if (umc_count == eeprom_count) { + return 0; + } else if (umc_count > eeprom_count) { + RAS_DEV_ERR(ras_core->dev, "Invalid error count: eeprom:%d, umc:%d\n", + eeprom_count, umc_count); + return 0; + } + + for (i = umc_count; i < eeprom_count; i++) { + memset(&rec, 0, sizeof(rec)); + ret = ras_eeprom_mgr_get_records(ras_core, i, &rec, 1); + if (ret) + return 0; + + c = 0; + ret = ras_umc_add_bad_pages(ras_core, &rec, 1, &c); + if (!ret) + new_count += c; + } + + return new_count; +} diff --git a/drivers/gpu/drm/amd/ras/core/ras_umc.h b/drivers/gpu/drm/amd/ras/core/ras_umc.h index 221cd008e55a0..5a3e019af8ad9 100644 --- a/drivers/gpu/drm/amd/ras/core/ras_umc.h +++ b/drivers/gpu/drm/amd/ras/core/ras_umc.h @@ -224,4 +224,5 @@ int ras_umc_bank_to_umc_record(struct ras_core_context *ras_core, struct ras_bank_ecc *bank, struct eeprom_umc_record *record); int ras_umc_record_to_nps_record(struct ras_core_context *ras_core, struct eeprom_umc_record *record, uint32_t nps); +int ras_umc_dump_fw_records(struct ras_core_context *ras_core); #endif -- 2.55.0
