diff options
Diffstat (limited to '')
-rw-r--r-- | plugins/ocp/ocp-nvme.c | 788 | ||||
-rw-r--r-- | plugins/ocp/ocp-nvme.h | 25 |
2 files changed, 813 insertions, 0 deletions
diff --git a/plugins/ocp/ocp-nvme.c b/plugins/ocp/ocp-nvme.c new file mode 100644 index 0000000..6b7fc9d --- /dev/null +++ b/plugins/ocp/ocp-nvme.c @@ -0,0 +1,788 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* Copyright (c) 2022 Meta Platforms, Inc. + * + * Authors: Arthur Shau <arthurshau@fb.com>, + * Wei Zhang <wzhang@fb.com>, + * Venkat Ramesh <venkatraghavan@fb.com> + */ +#include <stdio.h> +#include <string.h> +#include <stdlib.h> +#include <inttypes.h> +#include <errno.h> +#include <limits.h> +#include <fcntl.h> +#include <unistd.h> + +#include "common.h" +#include "nvme.h" +#include "libnvme.h" +#include "plugin.h" +#include "linux/types.h" +#include "nvme-print.h" + +#define CREATE_CMD +#include "ocp-nvme.h" + +/* C0 SCAO Log Page */ +#define C0_SMART_CLOUD_ATTR_LEN 0x200 +#define C0_SMART_CLOUD_ATTR_OPCODE 0xC0 +#define C0_GUID_LENGTH 16 +#define C0_ACTIVE_BUCKET_TIMER_INCREMENT 5 +#define C0_ACTIVE_THRESHOLD_INCREMENT 5 +#define C0_MINIMUM_WINDOW_INCREMENT 100 + +static __u8 scao_guid[C0_GUID_LENGTH] = { 0xC5, 0xAF, 0x10, 0x28, 0xEA, 0xBF, + 0xF2, 0xA4, 0x9C, 0x4F, 0x6F, 0x7C, 0xC9, 0x14, 0xD5, 0xAF }; + +/* C3 Latency Monitor Log Page */ +#define C3_LATENCY_MON_LOG_BUF_LEN 0x200 +#define C3_LATENCY_MON_OPCODE 0xC3 +#define C3_LATENCY_MON_VERSION 0x0001 +#define C3_GUID_LENGTH 16 +static __u8 lat_mon_guid[C3_GUID_LENGTH] = { 0x92, 0x7a, 0xc0, 0x8c, 0xd0, 0x84, + 0x6c, 0x9c, 0x70, 0x43, 0xe6, 0xd4, 0x58, 0x5e, 0xd4, 0x85 }; + +#define READ 0 +#define WRITE 1 +#define TRIM 2 +#define RESERVED 3 + +typedef enum { + SCAO_PMUW = 0, /* Physical media units written */ + SCAO_PMUR = 16, /* Physical media units read */ + SCAO_BUNBR = 32, /* Bad user nand blocks raw */ + SCAO_BUNBN = 38, /* Bad user nand blocks normalized */ + SCAO_BSNBR = 40, /* Bad system nand blocks raw */ + SCAO_BSNBN = 46, /* Bad system nand blocks normalized */ + SCAO_XRC = 48, /* XOR recovery count */ + SCAO_UREC = 56, /* Uncorrectable read error count */ + SCAO_SEEC = 64, /* Soft ecc error count */ + SCAO_EECE = 72, /* End to end corrected errors */ + SCAO_EEDC = 76, /* End to end detected errors */ + SCAO_SDPU = 80, /* System data percent used */ + SCAO_RFSC = 81, /* Refresh counts */ + SCAO_MXUDEC = 88, /* Max User data erase counts */ + SCAO_MNUDEC = 92, /* Min User data erase counts */ + SCAO_NTTE = 96, /* Number of Thermal throttling events */ + SCAO_CTS = 97, /* Current throttling status */ + SCAO_EVF = 98, /* Errata Version Field */ + SCAO_PVF = 99, /* Point Version Field */ + SCAO_MIVF = 101, /* Minor Version Field */ + SCAO_MAVF = 103, /* Major Version Field */ + SCAO_PCEC = 104, /* PCIe correctable error count */ + SCAO_ICS = 112, /* Incomplete shutdowns */ + SCAO_PFB = 120, /* Percent free blocks */ + SCAO_CPH = 128, /* Capacitor health */ + SCAO_NEV = 130, /* NVMe Errata Version */ + SCAO_UIO = 136, /* Unaligned I/O */ + SCAO_SVN = 144, /* Security Version Number */ + SCAO_NUSE = 152, /* NUSE - Namespace utilization */ + SCAO_PSC = 160, /* PLP start count */ + SCAO_EEST = 176, /* Endurance estimate */ + SCAO_PLRC = 192, /* PCIe Link Retraining Count */ + SCAO_LPV = 494, /* Log page version */ + SCAO_LPG = 496, /* Log page GUID */ +} SMART_CLOUD_ATTRIBUTE_OFFSETS; + +struct __attribute__((__packed__)) ssd_latency_monitor_log { + __u8 feature_status; /* 0x00 */ + __u8 rsvd1; /* 0x01 */ + __le16 active_bucket_timer; /* 0x02 */ + __le16 active_bucket_timer_threshold; /* 0x04 */ + __u8 active_threshold_a; /* 0x06 */ + __u8 active_threshold_b; /* 0x07 */ + __u8 active_threshold_c; /* 0x08 */ + __u8 active_threshold_d; /* 0x09 */ + __le16 active_latency_config; /* 0x0A */ + __u8 active_latency_min_window; /* 0x0C */ + __u8 rsvd2[0x13]; /* 0x0D */ + + __le32 active_bucket_counter[4][4] ; /* 0x20 - 0x5F */ + __le64 active_latency_timestamp[4][3]; /* 0x60 - 0xBF */ + __le16 active_measured_latency[4][3]; /* 0xC0 - 0xD7 */ + __le16 active_latency_stamp_units; /* 0xD8 */ + __u8 rsvd3[0x16]; /* 0xDA */ + + __le32 static_bucket_counter[4][4] ; /* 0xF0 - 0x12F */ + __le64 static_latency_timestamp[4][3]; /* 0x130 - 0x18F */ + __le16 static_measured_latency[4][3]; /* 0x190 - 0x1A7 */ + __le16 static_latency_stamp_units; /* 0x1A8 */ + __u8 rsvd4[0x16]; /* 0x1AA */ + + __le16 debug_log_trigger_enable; /* 0x1C0 */ + __le16 debug_log_measured_latency; /* 0x1C2 */ + __le64 debug_log_latency_stamp; /* 0x1C4 */ + __le16 debug_log_ptr; /* 0x1CC */ + __le16 debug_log_counter_trigger; /* 0x1CE */ + __u8 debug_log_stamp_units; /* 0x1D0 */ + __u8 rsvd5[0x1D]; /* 0x1D1 */ + + __le16 log_page_version; /* 0x1EE */ + __u8 log_page_guid[0x10]; /* 0x1F0 */ +}; + +static long double int128_to_double(__u8 *data) +{ + int i; + long double result = 0; + + for (i = 0; i < 16; i++) { + result *= 256; + result += data[15 - i]; + } + return result; +} + +static int convert_ts(time_t time, char *ts_buf) +{ + struct tm gmTimeInfo; + time_t time_Human, time_ms; + char buf[80]; + + time_Human = time/1000; + time_ms = time % 1000; + + gmtime_r((const time_t *)&time_Human, &gmTimeInfo); + + strftime(buf, sizeof(buf), "%Y-%m-%d %H:%M:%S", &gmTimeInfo); + sprintf(ts_buf, "%s.%03ld GMT", buf, time_ms); + + return 0; +} + +static void ocp_print_C0_log_normal(void *data) +{ + __u8 *log_data = (__u8*)data; + uint16_t smart_log_ver = 0; + + printf("SMART Cloud Attributes :- \n"); + + printf(" Physical media units written - %"PRIu64" %"PRIu64"\n", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_PMUW+8] & 0xFFFFFFFFFFFFFFFF), + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_PMUW] & 0xFFFFFFFFFFFFFFFF)); + printf(" Physical media units read - %"PRIu64" %"PRIu64"\n", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_PMUR+8] & 0xFFFFFFFFFFFFFFFF), + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_PMUR] & 0xFFFFFFFFFFFFFFFF)); + printf(" Bad user nand blocks - Raw %"PRIu64"\n", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_BUNBR] & 0x0000FFFFFFFFFFFF)); + printf(" Bad user nand blocks - Normalized %d\n", + (uint16_t)le16_to_cpu(*(uint16_t *)&log_data[SCAO_BUNBN])); + printf(" Bad system nand blocks - Raw %"PRIu64"\n", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_BSNBR] & 0x0000FFFFFFFFFFFF)); + printf(" Bad system nand blocks - Normalized %d\n", + (uint16_t)le16_to_cpu(*(uint16_t *)&log_data[SCAO_BSNBN])); + printf(" XOR recovery count %"PRIu64"\n", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_XRC])); + printf(" Uncorrectable read error count %"PRIu64"\n", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_UREC])); + printf(" Soft ecc error count %"PRIu64"\n", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_SEEC])); + printf(" End to end corrected errors %"PRIu32"\n", + (uint32_t)le32_to_cpu(*(uint32_t *)&log_data[SCAO_EECE])); + printf(" End to end detected errors %"PRIu32"\n", + (uint32_t)le32_to_cpu(*(uint32_t *)&log_data[SCAO_EEDC])); + printf(" System data percent used %d\n", + (__u8)log_data[SCAO_SDPU]); + printf(" Refresh counts %"PRIu64"\n", + (uint64_t)(le64_to_cpu(*(uint64_t *)&log_data[SCAO_RFSC])& 0x00FFFFFFFFFFFFFF)); + printf(" Max User data erase counts %"PRIu32"\n", + (uint32_t)le32_to_cpu(*(uint32_t *)&log_data[SCAO_MXUDEC])); + printf(" Min User data erase counts %"PRIu32"\n", + (uint32_t)le32_to_cpu(*(uint32_t *)&log_data[SCAO_MNUDEC])); + printf(" Number of Thermal throttling events %d\n", + (__u8)log_data[SCAO_NTTE]); + printf(" Current throttling status 0x%x\n", + (__u8)log_data[SCAO_CTS]); + printf(" PCIe correctable error count %"PRIu64"\n", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_PCEC])); + printf(" Incomplete shutdowns %"PRIu32"\n", + (uint32_t)le32_to_cpu(*(uint32_t *)&log_data[SCAO_ICS])); + printf(" Percent free blocks %d\n", + (__u8)log_data[SCAO_PFB]); + printf(" Capacitor health %"PRIu16"\n", + (uint16_t)le16_to_cpu(*(uint16_t *)&log_data[SCAO_CPH])); + printf(" Unaligned I/O %"PRIu64"\n", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_UIO])); + printf(" Security Version Number %"PRIu64"\n", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_SVN])); + printf(" NUSE - Namespace utilization %"PRIu64"\n", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_NUSE])); + printf(" PLP start count %.0Lf\n", + int128_to_double(&log_data[SCAO_PSC])); + printf(" Endurance estimate %.0Lf\n", + int128_to_double(&log_data[SCAO_EEST])); + smart_log_ver = (uint16_t)le16_to_cpu(*(uint16_t *)&log_data[SCAO_LPV]); + printf(" Log page version %"PRIu16"\n",smart_log_ver); + printf(" Log page GUID 0x"); + printf("%"PRIx64"%"PRIx64"\n",(uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_LPG + 8]), + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_LPG])); + if(smart_log_ver > 2) { + printf(" Errata Version Field %d\n", + (__u8)log_data[SCAO_EVF]); + printf(" Point Version Field %"PRIu16"\n", + (uint16_t)log_data[SCAO_PVF]); + printf(" Minor Version Field %"PRIu16"\n", + (uint16_t)log_data[SCAO_MIVF]); + printf(" Major Version Field %d\n", + (__u8)log_data[SCAO_MAVF]); + printf(" NVMe Errata Version %d\n", + (__u8)log_data[SCAO_NEV]); + printf(" PCIe Link Retraining Count %"PRIu64"\n", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_PLRC])); + } + printf("\n"); +} + +static void ocp_print_C0_log_json(void *data) +{ + __u8 *log_data = (__u8*)data; + struct json_object *root; + struct json_object *pmuw; + struct json_object *pmur; + uint16_t smart_log_ver = 0; + + root = json_create_object(); + pmuw = json_create_object(); + pmur = json_create_object(); + + json_object_add_value_uint64(pmuw, "hi", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_PMUW+8] & 0xFFFFFFFFFFFFFFFF)); + json_object_add_value_uint64(pmuw, "lo", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_PMUW] & 0xFFFFFFFFFFFFFFFF)); + json_object_add_value_object(root, "Physical media units written", pmuw); + json_object_add_value_uint64(pmur, "hi", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_PMUR+8] & 0xFFFFFFFFFFFFFFFF)); + json_object_add_value_uint64(pmur, "lo", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_PMUR] & 0xFFFFFFFFFFFFFFFF)); + json_object_add_value_object(root, "Physical media units read", pmur); + json_object_add_value_uint64(root, "Bad user nand blocks - Raw", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_BUNBR] & 0x0000FFFFFFFFFFFF)); + json_object_add_value_uint(root, "Bad user nand blocks - Normalized", + (uint16_t)le16_to_cpu(*(uint16_t *)&log_data[SCAO_BUNBN])); + json_object_add_value_uint64(root, "Bad system nand blocks - Raw", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_BSNBR] & 0x0000FFFFFFFFFFFF)); + json_object_add_value_uint(root, "Bad system nand blocks - Normalized", + (uint16_t)le16_to_cpu(*(uint16_t *)&log_data[SCAO_BSNBN])); + json_object_add_value_uint64(root, "XOR recovery count", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_XRC])); + json_object_add_value_uint64(root, "Uncorrectable read error count", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_UREC])); + json_object_add_value_uint64(root, "Soft ecc error count", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_SEEC])); + json_object_add_value_uint(root, "End to end corrected errors", + (uint32_t)le32_to_cpu(*(uint32_t *)&log_data[SCAO_EECE])); + json_object_add_value_uint(root, "End to end detected errors", + (uint32_t)le32_to_cpu(*(uint32_t *)&log_data[SCAO_EEDC])); + json_object_add_value_uint(root, "System data percent used", + (__u8)log_data[SCAO_SDPU]); + json_object_add_value_uint64(root, "Refresh counts", + (uint64_t)(le64_to_cpu(*(uint64_t *)&log_data[SCAO_RFSC])& 0x00FFFFFFFFFFFFFF)); + json_object_add_value_uint(root, "Max User data erase counts", + (uint32_t)le32_to_cpu(*(uint32_t *)&log_data[SCAO_MXUDEC])); + json_object_add_value_uint(root, "Min User data erase counts", + (uint32_t)le32_to_cpu(*(uint32_t *)&log_data[SCAO_MNUDEC])); + json_object_add_value_uint(root, "Number of Thermal throttling events", + (__u8)log_data[SCAO_NTTE]); + json_object_add_value_uint(root, "Current throttling status", + (__u8)log_data[SCAO_CTS]); + json_object_add_value_uint64(root, "PCIe correctable error count", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_PCEC])); + json_object_add_value_uint(root, "Incomplete shutdowns", + (uint32_t)le32_to_cpu(*(uint32_t *)&log_data[SCAO_ICS])); + json_object_add_value_uint(root, "Percent free blocks", + (__u8)log_data[SCAO_PFB]); + json_object_add_value_uint(root, "Capacitor health", + (uint16_t)le16_to_cpu(*(uint16_t *)&log_data[SCAO_CPH])); + json_object_add_value_uint64(root, "Unaligned I/O", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_UIO])); + json_object_add_value_uint64(root, "Security Version Number", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_SVN])); + json_object_add_value_uint64(root, "NUSE - Namespace utilization", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_NUSE])); + json_object_add_value_uint(root, "PLP start count", + int128_to_double(&log_data[SCAO_PSC])); + json_object_add_value_uint64(root, "Endurance estimate", + int128_to_double(&log_data[SCAO_EEST])); + smart_log_ver = (uint16_t)le16_to_cpu(*(uint16_t *)&log_data[SCAO_LPV]); + json_object_add_value_uint(root, "Log page version", smart_log_ver); + char guid[40]; + memset((void*)guid, 0, 40); + sprintf((char*)guid, "0x%"PRIx64"%"PRIx64"",(uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_LPG + 8]), + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_LPG])); + json_object_add_value_string(root, "Log page GUID", guid); + if(smart_log_ver > 2){ + json_object_add_value_uint(root, "Errata Version Field", + (__u8)log_data[SCAO_EVF]); + json_object_add_value_uint(root, "Point Version Field", + (uint16_t)log_data[SCAO_PVF]); + json_object_add_value_uint(root, "Minor Version Field", + (uint16_t)log_data[SCAO_MIVF]); + json_object_add_value_uint(root, "Major Version Field", + (__u8)log_data[SCAO_MAVF]); + json_object_add_value_uint(root, "NVMe Errata Version", + (__u8)log_data[SCAO_NEV]); + json_object_add_value_uint(root, "PCIe Link Retraining Count", + (uint64_t)le64_to_cpu(*(uint64_t *)&log_data[SCAO_PLRC])); + } + json_print_object(root, NULL); + printf("\n"); + json_free_object(root); +} + +static int get_c0_log_page(int fd, char *format) +{ + int ret = 0; + int fmt = -1; + __u8 *data; + int i; + + fmt = validate_output_format(format); + if (fmt < 0) { + fprintf(stderr, "ERROR : OCP : invalid output format\n"); + return fmt; + } + + if ((data = (__u8 *) malloc(sizeof(__u8) * C0_SMART_CLOUD_ATTR_LEN)) == NULL) { + fprintf(stderr, "ERROR : OCP : malloc : %s\n", strerror(errno)); + return -1; + } + memset(data, 0, sizeof (__u8) * C0_SMART_CLOUD_ATTR_LEN); + + ret = nvme_get_log_simple(fd, C0_SMART_CLOUD_ATTR_OPCODE, + C0_SMART_CLOUD_ATTR_LEN, data); + + if (strcmp(format, "json")) + fprintf(stderr, "NVMe Status:%s(%x)\n", + nvme_status_to_string(ret, false), ret); + + if (ret == 0) { + + /* check log page guid */ + /* Verify GUID matches */ + for (i=0; i<16; i++) { + if (scao_guid[i] != data[SCAO_LPG + i]) { + fprintf(stderr, "ERROR : OCP : Unknown GUID in C0 Log Page data\n"); + int j; + fprintf(stderr, "ERROR : OCP : Expected GUID: 0x"); + for (j = 0; j<16; j++) { + fprintf(stderr, "%x", scao_guid[j]); + } + fprintf(stderr, "\nERROR : OCP : Actual GUID: 0x"); + for (j = 0; j<16; j++) { + fprintf(stderr, "%x", data[SCAO_LPG + j]); + } + fprintf(stderr, "\n"); + + ret = -1; + goto out; + } + } + + /* print the data */ + if (!data) { + fprintf(stderr, "ERROR : OCP : Invalid buffer to read 0xC0 log\n"); + ret = -1; + goto out; + } + switch (fmt) { + case NORMAL: + ocp_print_C0_log_normal(data); + break; + case JSON: + ocp_print_C0_log_json(data); + break; + } + } else { + fprintf(stderr, "ERROR : OCP : Unable to read C0 data from buffer\n"); + } + +out: + free(data); + return ret; +} + +static int ocp_smart_add_log(int argc, char **argv, struct command *cmd, + struct plugin *plugin) +{ + const char *desc = "Retrieve latency monitor log data."; + int fd; + int ret = 0; + + struct config { + char *output_format; + }; + + struct config cfg = { + .output_format = "normal", + }; + + OPT_ARGS(opts) = { + OPT_FMT("output-format", 'o', &cfg.output_format, "output Format: normal|json"), + OPT_END() + }; + + fd = parse_and_open(argc, argv, desc, opts); + if (fd < 0) + return fd; + + ret = get_c0_log_page(fd, cfg.output_format); + if (ret) + fprintf(stderr, "ERROR : OCP : Failure reading the C0 Log Page, ret = %d\n", + ret); + + return ret; +} + +static int ocp_print_C3_log_normal(int fd, struct ssd_latency_monitor_log *log_data) +{ + printf("-Latency Monitor/C3 Log Page Data- \n"); + printf(" Controller : %s\n", devicename); + int i, j; + int pos = 0; + char ts_buf[128]; + + printf(" Feature Status 0x%x \n", + log_data->feature_status); + printf(" Active Bucket Timer %d min \n", + C0_ACTIVE_BUCKET_TIMER_INCREMENT * + le16_to_cpu(log_data->active_bucket_timer)); + printf(" Active Bucket Timer Threshold %d min \n", + C0_ACTIVE_BUCKET_TIMER_INCREMENT * + le16_to_cpu(log_data->active_bucket_timer_threshold)); + printf(" Active Threshold A %d ms \n", + C0_ACTIVE_THRESHOLD_INCREMENT * + le16_to_cpu(log_data->active_threshold_a+1)); + printf(" Active Threshold B %d ms \n", + C0_ACTIVE_THRESHOLD_INCREMENT * + le16_to_cpu(log_data->active_threshold_b+1)); + printf(" Active Threshold C %d ms \n", + C0_ACTIVE_THRESHOLD_INCREMENT * + le16_to_cpu(log_data->active_threshold_c+1)); + printf(" Active Threshold D %d ms \n", + C0_ACTIVE_THRESHOLD_INCREMENT * + le16_to_cpu(log_data->active_threshold_d+1)); + printf(" Active Latency Minimum Window %d ms \n", + C0_MINIMUM_WINDOW_INCREMENT * + le16_to_cpu(log_data->active_latency_min_window)); + printf(" Active Latency Stamp Units %d \n", + le16_to_cpu(log_data->active_latency_stamp_units)); + printf(" Static Latency Stamp Units %d \n", + le16_to_cpu(log_data->static_latency_stamp_units)); + printf(" Debug Log Trigger Enable %d \n", + le16_to_cpu(log_data->debug_log_trigger_enable)); + + printf(" Read Write Deallocate/Trim \n"); + for (i = 0; i <= 3; i++) { + printf(" Active Latency Mode: Bucket %d %27d %27d %27d \n", + i, + log_data->active_latency_config & (1 << pos), + log_data->active_latency_config & (1 << pos), + log_data->active_latency_config & (1 << pos)); + } + printf("\n"); + for (i = 0; i <= 3; i++) { + printf(" Active Bucket Counter: Bucket %d %27d %27d %27d \n", + i, + le32_to_cpu(log_data->active_bucket_counter[i][READ]), + le32_to_cpu(log_data->active_bucket_counter[i][WRITE]), + le32_to_cpu(log_data->active_bucket_counter[i][TRIM])); + } + + for (i = 0; i <= 3; i++) { + printf(" Active Measured Latency: Bucket %d %27d ms %27d ms %27d ms \n", + i, + le16_to_cpu(log_data->active_measured_latency[i][READ]), + le16_to_cpu(log_data->active_measured_latency[i][WRITE]), + le16_to_cpu(log_data->active_measured_latency[i][TRIM])); + } + + for (i = 0; i <= 3; i++) { + printf(" Active Latency Time Stamp: Bucket %d ", i); + for (j = 0; j <= 2; j++) { + if (le64_to_cpu(log_data->active_latency_timestamp[i][j]) == -1) + printf(" N/A "); + else { + convert_ts(le64_to_cpu(log_data->active_latency_timestamp[i][j]), ts_buf); + printf("%s ", ts_buf); + } + } + printf("\n"); + } + + for (i = 0; i <= 3; i++) { + printf(" Static Bucket Counter: Bucket %d %27d %27d %27d \n", + i, + le32_to_cpu(log_data->static_bucket_counter[i][READ]), + le32_to_cpu(log_data->static_bucket_counter[i][WRITE]), + le32_to_cpu(log_data->static_bucket_counter[i][TRIM])); + } + + for (i = 0; i <= 3; i++) { + printf(" Static Measured Latency: Bucket %d %27d ms %27d ms %27d ms \n", + i, + le16_to_cpu(log_data->static_measured_latency[i][READ]), + le16_to_cpu(log_data->static_measured_latency[i][WRITE]), + le16_to_cpu(log_data->static_measured_latency[i][TRIM])); + } + + for (i = 0; i <= 3; i++) { + printf(" Static Latency Time Stamp: Bucket %d ", i); + for (j = 0; j <= 2; j++) { + if (le64_to_cpu(log_data->static_latency_timestamp[i][j]) == -1) + printf(" N/A "); + else { + convert_ts(le64_to_cpu(log_data->static_latency_timestamp[i][j]), ts_buf); + printf("%s ", ts_buf); + } + } + printf("\n"); + } + + return 0; +} + +static void ocp_print_C3_log_json(struct ssd_latency_monitor_log *log_data) +{ + int i, j; + int pos = 0; + char buf[128]; + char ts_buf[128]; + char *operation[3] = {"Read", "Write", "Trim"}; + struct json_object *root; + root = json_create_object(); + + json_object_add_value_uint(root, "Feature Status", + log_data->feature_status); + json_object_add_value_uint(root, "Active Bucket Timer", + C0_ACTIVE_BUCKET_TIMER_INCREMENT * + le16_to_cpu(log_data->active_bucket_timer)); + json_object_add_value_uint(root, "Active Bucket Timer Threshold", + C0_ACTIVE_BUCKET_TIMER_INCREMENT * + le16_to_cpu(log_data->active_bucket_timer_threshold)); + json_object_add_value_uint(root, "Active Threshold A", + C0_ACTIVE_THRESHOLD_INCREMENT * + le16_to_cpu(log_data->active_threshold_a+1)); + json_object_add_value_uint(root, "Active Threshold B", + C0_ACTIVE_THRESHOLD_INCREMENT * + le16_to_cpu(log_data->active_threshold_b+1)); + json_object_add_value_uint(root, "Active Threshold C", + C0_ACTIVE_THRESHOLD_INCREMENT * + le16_to_cpu(log_data->active_threshold_c+1)); + json_object_add_value_uint(root, "Active Threshold D", + C0_ACTIVE_THRESHOLD_INCREMENT * + le16_to_cpu(log_data->active_threshold_d+1)); + json_object_add_value_uint(root, "Active Lantency Minimum Window", + C0_MINIMUM_WINDOW_INCREMENT * + le16_to_cpu(log_data->active_latency_min_window)); + json_object_add_value_uint(root, "Active Latency Stamp Units", + le16_to_cpu(log_data->active_latency_stamp_units)); + json_object_add_value_uint(root, "Static Latency Stamp Units", + le16_to_cpu(log_data->static_latency_stamp_units)); + json_object_add_value_uint(root, "Debug Log Trigger Enable", + le16_to_cpu(log_data->debug_log_trigger_enable)); + + for (i = 0; i <= 3; i++) { + struct json_object *bucket; + bucket = json_create_object(); + sprintf(buf, "Active Latency Mode: Bucket %d", i); + for (j = 0; j <= 2; j++) { + json_object_add_value_uint(bucket, operation[j], + log_data->active_latency_config & (1 << pos)); + } + json_object_add_value_object(root, buf, bucket); + } + for (i = 0; i <= 3; i++) { + struct json_object *bucket; + bucket = json_create_object(); + sprintf(buf, "Active Bucket Counter: Bucket %d", i); + for (j = 0; j <= 2; j++) { + json_object_add_value_uint(bucket, operation[j], + le32_to_cpu(log_data->active_bucket_counter[i][j])); + } + json_object_add_value_object(root, buf, bucket); + } + for (i = 0; i <= 3; i++) { + struct json_object *bucket; + bucket = json_create_object(); + sprintf(buf, "Active Measured Latency: Bucket %d", i); + for (j = 0; j <= 2; j++) { + json_object_add_value_uint(bucket, operation[j], + le16_to_cpu(log_data->active_measured_latency[i][j])); + } + json_object_add_value_object(root, buf, bucket); + } + for (i = 0; i <= 3; i++) { + struct json_object *bucket; + bucket = json_create_object(); + sprintf(buf, "Active Latency Time Stamp: Bucket %d", i); + for (j = 0; j <= 2; j++) { + if (le64_to_cpu(log_data->active_latency_timestamp[i][j]) == -1) + json_object_add_value_string(bucket, operation[j], "NA"); + else { + convert_ts(le64_to_cpu(log_data->active_latency_timestamp[i][j]), ts_buf); + json_object_add_value_string(bucket, operation[j], ts_buf); + } + } + json_object_add_value_object(root, buf, bucket); + } + for (i = 0; i <= 3; i++) { + struct json_object *bucket; + bucket = json_create_object(); + sprintf(buf, "Static Bucket Counter: Bucket %d", i); + for (j = 0; j <= 2; j++) { + json_object_add_value_uint(bucket, operation[j], + le32_to_cpu(log_data->static_bucket_counter[i][j])); + } + json_object_add_value_object(root, buf, bucket); + } + for (i = 0; i <= 3; i++) { + struct json_object *bucket; + bucket = json_create_object(); + sprintf(buf, "Static Measured Latency: Bucket %d", i); + for (j = 0; j <= 2; j++) { + json_object_add_value_uint(bucket, operation[j], + le16_to_cpu(log_data->static_measured_latency[i][j])); + } + json_object_add_value_object(root, buf, bucket); + } + for (i = 0; i <= 3; i++) { + struct json_object *bucket; + bucket = json_create_object(); + sprintf(buf, "Static Latency Time Stamp: Bucket %d", i); + for (j = 0; j <= 2; j++) { + if (le64_to_cpu(log_data->static_latency_timestamp[i][j]) == -1) + json_object_add_value_string(bucket, operation[j], "NA"); + else { + convert_ts(le64_to_cpu(log_data->static_latency_timestamp[i][j]), ts_buf); + json_object_add_value_string(bucket, operation[j], ts_buf); + } + } + json_object_add_value_object(root, buf, bucket); + } + + json_print_object(root, NULL); + printf("\n"); + + json_free_object(root); +} + +static int get_c3_log_page(int fd, char *format) +{ + int ret = 0; + int fmt = -1; + __u8 *data; + int i; + struct ssd_latency_monitor_log *log_data; + + fmt = validate_output_format(format); + if (fmt < 0) { + fprintf(stderr, "ERROR : OCP : invalid output format\n"); + return fmt; + } + + if ((data = (__u8 *) malloc(sizeof(__u8) * C3_LATENCY_MON_LOG_BUF_LEN)) == NULL) { + fprintf(stderr, "ERROR : OCP : malloc : %s\n", strerror(errno)); + return -1; + } + memset(data, 0, sizeof (__u8) * C3_LATENCY_MON_LOG_BUF_LEN); + + ret = nvme_get_log_simple(fd, C3_LATENCY_MON_OPCODE, + C3_LATENCY_MON_LOG_BUF_LEN, data); + + if (strcmp(format, "json")) + fprintf(stderr, + "NVMe Status:%s(%x)\n", + nvme_status_to_string(ret, false), + ret); + + if (ret == 0) { + log_data = (struct ssd_latency_monitor_log*)data; + + /* check log page version */ + if (log_data->log_page_version != C3_LATENCY_MON_VERSION) { + fprintf(stderr, + "ERROR : OCP : invalid latency monitor version\n"); + ret = -1; + goto out; + } + + /* check log page guid */ + /* Verify GUID matches */ + for (i=0; i<16; i++) { + if (lat_mon_guid[i] != log_data->log_page_guid[i]) { + fprintf(stderr,"ERROR : OCP : Unknown GUID in C3 Log Page data\n"); + int j; + fprintf(stderr, "ERROR : OCP : Expected GUID: 0x"); + for (j = 0; j<16; j++) { + fprintf(stderr, "%x", lat_mon_guid[j]); + } + fprintf(stderr, "\nERROR : OCP : Actual GUID: 0x"); + for (j = 0; j<16; j++) { + fprintf(stderr, "%x", log_data->log_page_guid[j]); + } + fprintf(stderr, "\n"); + + ret = -1; + goto out; + } + } + + /* print the data */ + if (!log_data) { + fprintf(stderr, + "ERROR : OCP : Invalid C3 log data buffer\n"); + ret = -1; + goto out; + } + switch (fmt) { + case NORMAL: + ocp_print_C3_log_normal(fd, log_data); + break; + case JSON: + ocp_print_C3_log_json(log_data); + break; + } + } else { + fprintf(stderr, + "ERROR : OCP : Unable to read C3 data from buffer\n"); + } + +out: + free(data); + return ret; +} + +static int ocp_latency_monitor_log(int argc, char **argv, struct command *command, + struct plugin *plugin) +{ + const char *desc = "Retrieve latency monitor log data."; + int fd; + int ret = 0; + + struct config { + char *output_format; + }; + + struct config cfg = { + .output_format = "normal", + }; + + OPT_ARGS(opts) = { + OPT_FMT("output-format", 'o', &cfg.output_format, + "output Format: normal|json"), + OPT_END() + }; + + fd = parse_and_open(argc, argv, desc, opts); + if (fd < 0) + return fd; + + ret = get_c3_log_page(fd, cfg.output_format); + if (ret) + fprintf(stderr, + "ERROR : OCP : Failure reading the C3 Log Page, ret = %d\n", + ret); + + return ret; +} diff --git a/plugins/ocp/ocp-nvme.h b/plugins/ocp/ocp-nvme.h new file mode 100644 index 0000000..19378cd --- /dev/null +++ b/plugins/ocp/ocp-nvme.h @@ -0,0 +1,25 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* Copyright (c) 2022 Meta Platforms, Inc. + * + * Authors: Arthur Shau <arthurshau@fb.com>, + * Wei Zhang <wzhang@fb.com>, + * Venkat Ramesh <venkatraghavan@fb.com> + */ +#undef CMD_INC_FILE +#define CMD_INC_FILE plugins/ocp/ocp-nvme + +#if !defined(OCP_NVME) || defined(CMD_HEADER_MULTI_READ) +#define OCP_NVME + +#include "cmd.h" + +PLUGIN(NAME("ocp", "OCP cloud SSD extensions", NVME_VERSION), + COMMAND_LIST( + ENTRY("smart-add-log", "Retrieve extended SMART Information", ocp_smart_add_log) + ENTRY("latency-monitor-log", "Get Latency Monitor Log Page", ocp_latency_monitor_log) + ) +); + +#endif + +#include "define_cmd.h" |