diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c')
-rw-r--r-- | drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c | 226 |
1 files changed, 129 insertions, 97 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c index 8a32b5c93778..2a8e04895595 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c @@ -27,7 +27,8 @@ #include <linux/bits.h> #include "smu_v11_0_i2c.h" -#define EEPROM_I2C_TARGET_ADDR 0xA0 +#define EEPROM_I2C_TARGET_ADDR_ARCTURUS 0xA8 +#define EEPROM_I2C_TARGET_ADDR_VEGA20 0xA0 /* * The 2 macros bellow represent the actual size in bytes that @@ -83,7 +84,7 @@ static int __update_table_header(struct amdgpu_ras_eeprom_control *control, { int ret = 0; struct i2c_msg msg = { - .addr = EEPROM_I2C_TARGET_ADDR, + .addr = 0, .flags = 0, .len = EEPROM_ADDRESS_SIZE + EEPROM_TABLE_HEADER_SIZE, .buf = buff, @@ -93,6 +94,8 @@ static int __update_table_header(struct amdgpu_ras_eeprom_control *control, *(uint16_t *)buff = EEPROM_HDR_START; __encode_table_header_to_buff(&control->tbl_hdr, buff + EEPROM_ADDRESS_SIZE); + msg.addr = control->i2c_address; + ret = i2c_transfer(&control->eeprom_accessor, &msg, 1); if (ret < 1) DRM_ERROR("Failed to write EEPROM table header, ret:%d", ret); @@ -100,7 +103,101 @@ static int __update_table_header(struct amdgpu_ras_eeprom_control *control, return ret; } -static uint32_t __calc_hdr_byte_sum(struct amdgpu_ras_eeprom_control *control); + + +static uint32_t __calc_hdr_byte_sum(struct amdgpu_ras_eeprom_control *control) +{ + int i; + uint32_t tbl_sum = 0; + + /* Header checksum, skip checksum field in the calculation */ + for (i = 0; i < sizeof(control->tbl_hdr) - sizeof(control->tbl_hdr.checksum); i++) + tbl_sum += *(((unsigned char *)&control->tbl_hdr) + i); + + return tbl_sum; +} + +static uint32_t __calc_recs_byte_sum(struct eeprom_table_record *records, + int num) +{ + int i, j; + uint32_t tbl_sum = 0; + + /* Records checksum */ + for (i = 0; i < num; i++) { + struct eeprom_table_record *record = &records[i]; + + for (j = 0; j < sizeof(*record); j++) { + tbl_sum += *(((unsigned char *)record) + j); + } + } + + return tbl_sum; +} + +static inline uint32_t __calc_tbl_byte_sum(struct amdgpu_ras_eeprom_control *control, + struct eeprom_table_record *records, int num) +{ + return __calc_hdr_byte_sum(control) + __calc_recs_byte_sum(records, num); +} + +/* Checksum = 256 -((sum of all table entries) mod 256) */ +static void __update_tbl_checksum(struct amdgpu_ras_eeprom_control *control, + struct eeprom_table_record *records, int num, + uint32_t old_hdr_byte_sum) +{ + /* + * This will update the table sum with new records. + * + * TODO: What happens when the EEPROM table is to be wrapped around + * and old records from start will get overridden. + */ + + /* need to recalculate updated header byte sum */ + control->tbl_byte_sum -= old_hdr_byte_sum; + control->tbl_byte_sum += __calc_tbl_byte_sum(control, records, num); + + control->tbl_hdr.checksum = 256 - (control->tbl_byte_sum % 256); +} + +/* table sum mod 256 + checksum must equals 256 */ +static bool __validate_tbl_checksum(struct amdgpu_ras_eeprom_control *control, + struct eeprom_table_record *records, int num) +{ + control->tbl_byte_sum = __calc_tbl_byte_sum(control, records, num); + + if (control->tbl_hdr.checksum + (control->tbl_byte_sum % 256) != 256) { + DRM_WARN("Checksum mismatch, checksum: %u ", control->tbl_hdr.checksum); + return false; + } + + return true; +} + +int amdgpu_ras_eeprom_reset_table(struct amdgpu_ras_eeprom_control *control) +{ + unsigned char buff[EEPROM_ADDRESS_SIZE + EEPROM_TABLE_HEADER_SIZE] = { 0 }; + struct amdgpu_ras_eeprom_table_header *hdr = &control->tbl_hdr; + int ret = 0; + + mutex_lock(&control->tbl_mutex); + + hdr->header = EEPROM_TABLE_HDR_VAL; + hdr->version = EEPROM_TABLE_VER; + hdr->first_rec_offset = EEPROM_RECORD_START; + hdr->tbl_size = EEPROM_TABLE_HEADER_SIZE; + + control->tbl_byte_sum = 0; + __update_tbl_checksum(control, NULL, 0, 0); + control->next_addr = EEPROM_RECORD_START; + + ret = __update_table_header(control, buff); + + mutex_unlock(&control->tbl_mutex); + + return ret; + +} int amdgpu_ras_eeprom_init(struct amdgpu_ras_eeprom_control *control) { @@ -109,7 +206,7 @@ int amdgpu_ras_eeprom_init(struct amdgpu_ras_eeprom_control *control) unsigned char buff[EEPROM_ADDRESS_SIZE + EEPROM_TABLE_HEADER_SIZE] = { 0 }; struct amdgpu_ras_eeprom_table_header *hdr = &control->tbl_hdr; struct i2c_msg msg = { - .addr = EEPROM_I2C_TARGET_ADDR, + .addr = 0, .flags = I2C_M_RD, .len = EEPROM_ADDRESS_SIZE + EEPROM_TABLE_HEADER_SIZE, .buf = buff, @@ -119,9 +216,15 @@ int amdgpu_ras_eeprom_init(struct amdgpu_ras_eeprom_control *control) switch (adev->asic_type) { case CHIP_VEGA20: + control->i2c_address = EEPROM_I2C_TARGET_ADDR_VEGA20; ret = smu_v11_0_i2c_eeprom_control_init(&control->eeprom_accessor); break; + case CHIP_ARCTURUS: + control->i2c_address = EEPROM_I2C_TARGET_ADDR_ARCTURUS; + ret = smu_i2c_eeprom_init(&adev->smu, &control->eeprom_accessor); + break; + default: return 0; } @@ -131,6 +234,8 @@ int amdgpu_ras_eeprom_init(struct amdgpu_ras_eeprom_control *control) return ret; } + msg.addr = control->i2c_address; + /* Read/Create table header from EEPROM address 0 */ ret = i2c_transfer(&control->eeprom_accessor, &msg, 1); if (ret < 1) { @@ -143,25 +248,18 @@ int amdgpu_ras_eeprom_init(struct amdgpu_ras_eeprom_control *control) if (hdr->header == EEPROM_TABLE_HDR_VAL) { control->num_recs = (hdr->tbl_size - EEPROM_TABLE_HEADER_SIZE) / EEPROM_TABLE_RECORD_SIZE; + control->tbl_byte_sum = __calc_hdr_byte_sum(control); + control->next_addr = EEPROM_RECORD_START; + DRM_DEBUG_DRIVER("Found existing EEPROM table with %d records", control->num_recs); } else { DRM_INFO("Creating new EEPROM table"); - hdr->header = EEPROM_TABLE_HDR_VAL; - hdr->version = EEPROM_TABLE_VER; - hdr->first_rec_offset = EEPROM_RECORD_START; - hdr->tbl_size = EEPROM_TABLE_HEADER_SIZE; - - adev->psp.ras.ras->eeprom_control.tbl_byte_sum = - __calc_hdr_byte_sum(&adev->psp.ras.ras->eeprom_control); - ret = __update_table_header(control, buff); + ret = amdgpu_ras_eeprom_reset_table(control); } - /* Start inserting records from here */ - adev->psp.ras.ras->eeprom_control.next_addr = EEPROM_RECORD_START; - return ret == 1 ? 0 : -EIO; } @@ -173,6 +271,9 @@ void amdgpu_ras_eeprom_fini(struct amdgpu_ras_eeprom_control *control) case CHIP_VEGA20: smu_v11_0_i2c_eeprom_control_fini(&control->eeprom_accessor); break; + case CHIP_ARCTURUS: + smu_i2c_eeprom_fini(&adev->smu, &control->eeprom_accessor); + break; default: return; @@ -226,8 +327,8 @@ static void __decode_table_record_from_buff(struct amdgpu_ras_eeprom_control *co record->offset = (le64_to_cpu(tmp) & 0xffffffffffff); i += 6; - buff[i++] = record->mem_channel; - buff[i++] = record->mcumc_id; + record->mem_channel = buff[i++]; + record->mcumc_id = buff[i++]; memcpy(&tmp, buff + i, 6); record->retired_page = (le64_to_cpu(tmp) & 0xffffffffffff); @@ -266,87 +367,18 @@ static uint32_t __correct_eeprom_dest_address(uint32_t curr_address) return curr_address; } - -static uint32_t __calc_hdr_byte_sum(struct amdgpu_ras_eeprom_control *control) -{ - int i; - uint32_t tbl_sum = 0; - - /* Header checksum, skip checksum field in the calculation */ - for (i = 0; i < sizeof(control->tbl_hdr) - sizeof(control->tbl_hdr.checksum); i++) - tbl_sum += *(((unsigned char *)&control->tbl_hdr) + i); - - return tbl_sum; -} - -static uint32_t __calc_recs_byte_sum(struct eeprom_table_record *records, - int num) -{ - int i, j; - uint32_t tbl_sum = 0; - - /* Records checksum */ - for (i = 0; i < num; i++) { - struct eeprom_table_record *record = &records[i]; - - for (j = 0; j < sizeof(*record); j++) { - tbl_sum += *(((unsigned char *)record) + j); - } - } - - return tbl_sum; -} - -static inline uint32_t __calc_tbl_byte_sum(struct amdgpu_ras_eeprom_control *control, - struct eeprom_table_record *records, int num) -{ - return __calc_hdr_byte_sum(control) + __calc_recs_byte_sum(records, num); -} - -/* Checksum = 256 -((sum of all table entries) mod 256) */ -static void __update_tbl_checksum(struct amdgpu_ras_eeprom_control *control, - struct eeprom_table_record *records, int num, - uint32_t old_hdr_byte_sum) -{ - /* - * This will update the table sum with new records. - * - * TODO: What happens when the EEPROM table is to be wrapped around - * and old records from start will get overridden. - */ - - /* need to recalculate updated header byte sum */ - control->tbl_byte_sum -= old_hdr_byte_sum; - control->tbl_byte_sum += __calc_tbl_byte_sum(control, records, num); - - control->tbl_hdr.checksum = 256 - (control->tbl_byte_sum % 256); -} - -/* table sum mod 256 + checksum must equals 256 */ -static bool __validate_tbl_checksum(struct amdgpu_ras_eeprom_control *control, - struct eeprom_table_record *records, int num) -{ - control->tbl_byte_sum = __calc_tbl_byte_sum(control, records, num); - - if (control->tbl_hdr.checksum + (control->tbl_byte_sum % 256) != 256) { - DRM_WARN("Checksum mismatch, checksum: %u ", control->tbl_hdr.checksum); - return false; - } - - return true; -} - int amdgpu_ras_eeprom_process_recods(struct amdgpu_ras_eeprom_control *control, struct eeprom_table_record *records, bool write, int num) { int i, ret = 0; - struct i2c_msg *msgs; - unsigned char *buffs; + struct i2c_msg *msgs, *msg; + unsigned char *buffs, *buff; + struct eeprom_table_record *record; struct amdgpu_device *adev = to_amdgpu_device(control); - if (adev->asic_type != CHIP_VEGA20) + if (adev->asic_type != CHIP_VEGA20 && adev->asic_type != CHIP_ARCTURUS) return 0; buffs = kcalloc(num, EEPROM_ADDRESS_SIZE + EEPROM_TABLE_RECORD_SIZE, @@ -373,9 +405,9 @@ int amdgpu_ras_eeprom_process_recods(struct amdgpu_ras_eeprom_control *control, * 256b */ for (i = 0; i < num; i++) { - unsigned char *buff = &buffs[i * (EEPROM_ADDRESS_SIZE + EEPROM_TABLE_RECORD_SIZE)]; - struct eeprom_table_record *record = &records[i]; - struct i2c_msg *msg = &msgs[i]; + buff = &buffs[i * (EEPROM_ADDRESS_SIZE + EEPROM_TABLE_RECORD_SIZE)]; + record = &records[i]; + msg = &msgs[i]; control->next_addr = __correct_eeprom_dest_address(control->next_addr); @@ -383,8 +415,8 @@ int amdgpu_ras_eeprom_process_recods(struct amdgpu_ras_eeprom_control *control, * Update bits 16,17 of EEPROM address in I2C address by setting them * to bits 1,2 of Device address byte */ - msg->addr = EEPROM_I2C_TARGET_ADDR | - ((control->next_addr & EEPROM_ADDR_MSB_MASK) >> 15); + msg->addr = control->i2c_address | + ((control->next_addr & EEPROM_ADDR_MSB_MASK) >> 15); msg->flags = write ? 0 : I2C_M_RD; msg->len = EEPROM_ADDRESS_SIZE + EEPROM_TABLE_RECORD_SIZE; msg->buf = buff; @@ -415,8 +447,8 @@ int amdgpu_ras_eeprom_process_recods(struct amdgpu_ras_eeprom_control *control, if (!write) { for (i = 0; i < num; i++) { - unsigned char *buff = &buffs[i*(EEPROM_ADDRESS_SIZE + EEPROM_TABLE_RECORD_SIZE)]; - struct eeprom_table_record *record = &records[i]; + buff = &buffs[i*(EEPROM_ADDRESS_SIZE + EEPROM_TABLE_RECORD_SIZE)]; + record = &records[i]; __decode_table_record_from_buff(control, record, buff + EEPROM_ADDRESS_SIZE); } |