summaryrefslogtreecommitdiffstats
path: root/plugins/intel/intel-nvme.c
diff options
context:
space:
mode:
Diffstat (limited to 'plugins/intel/intel-nvme.c')
-rw-r--r--plugins/intel/intel-nvme.c1738
1 files changed, 1738 insertions, 0 deletions
diff --git a/plugins/intel/intel-nvme.c b/plugins/intel/intel-nvme.c
new file mode 100644
index 0000000..378ecc0
--- /dev/null
+++ b/plugins/intel/intel-nvme.c
@@ -0,0 +1,1738 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+#include <fcntl.h>
+#include <errno.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <inttypes.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 "intel-nvme.h"
+
+struct __packed nvme_additional_smart_log_item {
+ __u8 key;
+ __u8 _kp[2];
+ __u8 norm;
+ __u8 _np;
+ union __packed {
+ __u8 raw[6];
+ struct __packed wear_level {
+ __le16 min;
+ __le16 max;
+ __le16 avg;
+ } wear_level;
+ struct __packed thermal_throttle {
+ __u8 pct;
+ __u32 count;
+ } thermal_throttle;
+ };
+ __u8 _rp;
+};
+
+struct nvme_additional_smart_log {
+ struct nvme_additional_smart_log_item program_fail_cnt;
+ struct nvme_additional_smart_log_item erase_fail_cnt;
+ struct nvme_additional_smart_log_item wear_leveling_cnt;
+ struct nvme_additional_smart_log_item e2e_err_cnt;
+ struct nvme_additional_smart_log_item crc_err_cnt;
+ struct nvme_additional_smart_log_item timed_workload_media_wear;
+ struct nvme_additional_smart_log_item timed_workload_host_reads;
+ struct nvme_additional_smart_log_item timed_workload_timer;
+ struct nvme_additional_smart_log_item thermal_throttle_status;
+ struct nvme_additional_smart_log_item retry_buffer_overflow_cnt;
+ struct nvme_additional_smart_log_item pll_lock_loss_cnt;
+ struct nvme_additional_smart_log_item nand_bytes_written;
+ struct nvme_additional_smart_log_item host_bytes_written;
+ struct nvme_additional_smart_log_item host_ctx_wear_used;
+ struct nvme_additional_smart_log_item perf_stat_indicator;
+ struct nvme_additional_smart_log_item re_alloc_sectr_cnt;
+ struct nvme_additional_smart_log_item soft_ecc_err_rate;
+ struct nvme_additional_smart_log_item unexp_power_loss;
+ struct nvme_additional_smart_log_item media_bytes_read;
+ struct nvme_additional_smart_log_item avail_fw_downgrades;
+};
+
+struct nvme_vu_id_ctrl_field { /* CDR MR5 */
+ __u8 rsvd1[3];
+ __u8 ss;
+ __u8 health[20];
+ __u8 cls;
+ __u8 nlw;
+ __u8 scap;
+ __u8 sstat;
+ __u8 bl[8];
+ __u8 rsvd2[38];
+ __u8 ww[8]; /* little endian */
+ __u8 mic_bl[4];
+ __u8 mic_fw[4];
+};
+
+static void json_intel_id_ctrl(struct nvme_vu_id_ctrl_field *id,
+ char *health, char *bl, char *ww, char *mic_bl, char *mic_fw,
+ struct json_object *root)
+{
+ json_object_add_value_int(root, "ss", id->ss);
+ json_object_add_value_string(root, "health", health);
+ json_object_add_value_int(root, "cls", id->cls);
+ json_object_add_value_int(root, "nlw", id->nlw);
+ json_object_add_value_int(root, "scap", id->scap);
+ json_object_add_value_int(root, "sstat", id->sstat);
+ json_object_add_value_string(root, "bl", bl);
+ json_object_add_value_string(root, "ww", ww);
+ json_object_add_value_string(root, "mic_bl", mic_bl);
+ json_object_add_value_string(root, "mic_fw", mic_fw);
+}
+
+static void intel_id_ctrl(__u8 *vs, struct json_object *root)
+{
+ struct nvme_vu_id_ctrl_field *id = (struct nvme_vu_id_ctrl_field *)vs;
+
+ char health[21] = { 0 };
+ char bl[9] = { 0 };
+ char ww[19] = { 0 };
+ char mic_bl[5] = { 0 };
+ char mic_fw[5] = { 0 };
+
+ if (id->health[0] == 0)
+ snprintf(health, 21, "%s", "healthy");
+ else
+ snprintf(health, 21, "%s", id->health);
+
+ snprintf(bl, 9, "%s", id->bl);
+ snprintf(ww, 19, "%02X%02X%02X%02X%02X%02X%02X%02X", id->ww[7],
+ id->ww[6], id->ww[5], id->ww[4], id->ww[3], id->ww[2],
+ id->ww[1], id->ww[0]);
+ snprintf(mic_bl, 5, "%s", id->mic_bl);
+ snprintf(mic_fw, 5, "%s", id->mic_fw);
+
+ if (root) {
+ json_intel_id_ctrl(id, health, bl, ww, mic_bl, mic_fw, root);
+ return;
+ }
+
+ printf("ss : %d\n", id->ss);
+ printf("health : %s\n", health);
+ printf("cls : %d\n", id->cls);
+ printf("nlw : %d\n", id->nlw);
+ printf("scap : %d\n", id->scap);
+ printf("sstat : %d\n", id->sstat);
+ printf("bl : %s\n", bl);
+ printf("ww : %s\n", ww);
+ printf("mic_bl : %s\n", mic_bl);
+ printf("mic_fw : %s\n", mic_fw);
+}
+
+static int id_ctrl(int argc, char **argv, struct command *cmd, struct plugin *plugin)
+{
+ return __id_ctrl(argc, argv, cmd, plugin, intel_id_ctrl);
+}
+
+static void show_intel_smart_log_jsn(struct nvme_additional_smart_log *smart,
+ unsigned int nsid, const char *devname)
+{
+ struct json_object *root, *entry_stats, *dev_stats, *multi;
+
+ root = json_create_object();
+ json_object_add_value_string(root, "Intel Smart log", devname);
+
+ dev_stats = json_create_object();
+
+ entry_stats = json_create_object();
+ json_object_add_value_int(entry_stats, "normalized", smart->program_fail_cnt.norm);
+ json_object_add_value_int(entry_stats, "raw", int48_to_long(smart->program_fail_cnt.raw));
+ json_object_add_value_object(dev_stats, "program_fail_count", entry_stats);
+
+ entry_stats = json_create_object();
+ json_object_add_value_int(entry_stats, "normalized", smart->erase_fail_cnt.norm);
+ json_object_add_value_int(entry_stats, "raw", int48_to_long(smart->erase_fail_cnt.raw));
+ json_object_add_value_object(dev_stats, "erase_fail_count", entry_stats);
+
+ entry_stats = json_create_object();
+ json_object_add_value_int(entry_stats, "normalized", smart->wear_leveling_cnt.norm);
+ multi = json_create_object();
+ json_object_add_value_int(multi, "min", le16_to_cpu(smart->wear_leveling_cnt.wear_level.min));
+ json_object_add_value_int(multi, "max", le16_to_cpu(smart->wear_leveling_cnt.wear_level.max));
+ json_object_add_value_int(multi, "avg", le16_to_cpu(smart->wear_leveling_cnt.wear_level.avg));
+ json_object_add_value_object(entry_stats, "raw", multi);
+ json_object_add_value_object(dev_stats, "wear_leveling", entry_stats);
+
+ entry_stats = json_create_object();
+ json_object_add_value_int(entry_stats, "normalized", smart->e2e_err_cnt.norm);
+ json_object_add_value_int(entry_stats, "raw", int48_to_long(smart->e2e_err_cnt.raw));
+ json_object_add_value_object(dev_stats, "end_to_end_error_detection_count", entry_stats);
+
+ entry_stats = json_create_object();
+ json_object_add_value_int(entry_stats, "normalized", smart->crc_err_cnt.norm);
+ json_object_add_value_int(entry_stats, "raw", int48_to_long(smart->crc_err_cnt.raw));
+ json_object_add_value_object(dev_stats, "crc_error_count", entry_stats);
+
+ entry_stats = json_create_object();
+ json_object_add_value_int(entry_stats, "normalized", smart->timed_workload_media_wear.norm);
+ json_object_add_value_double(entry_stats, "raw", ((long double)int48_to_long(smart->timed_workload_media_wear.raw)) / 1024);
+ json_object_add_value_object(dev_stats, "timed_workload_media_wear", entry_stats);
+
+ entry_stats = json_create_object();
+ json_object_add_value_int(entry_stats, "normalized", smart->timed_workload_host_reads.norm);
+ json_object_add_value_int(entry_stats, "raw", int48_to_long(smart->timed_workload_host_reads.raw));
+ json_object_add_value_object(dev_stats, "timed_workload_host_reads", entry_stats);
+
+ entry_stats = json_create_object();
+ json_object_add_value_int(entry_stats, "normalized", smart->timed_workload_timer.norm);
+ json_object_add_value_int(entry_stats, "raw", int48_to_long(smart->timed_workload_timer.raw));
+ json_object_add_value_object(dev_stats, "timed_workload_timer", entry_stats);
+
+ entry_stats = json_create_object();
+ json_object_add_value_int(entry_stats, "normalized", smart->thermal_throttle_status.norm);
+ multi = json_create_object();
+ json_object_add_value_int(multi, "pct", smart->thermal_throttle_status.thermal_throttle.pct);
+ json_object_add_value_int(multi, "cnt", smart->thermal_throttle_status.thermal_throttle.count);
+ json_object_add_value_object(entry_stats, "raw", multi);
+ json_object_add_value_object(dev_stats, "thermal_throttle_status", entry_stats);
+
+ entry_stats = json_create_object();
+ json_object_add_value_int(entry_stats, "normalized", smart->retry_buffer_overflow_cnt.norm);
+ json_object_add_value_int(entry_stats, "raw", int48_to_long(smart->retry_buffer_overflow_cnt.raw));
+ json_object_add_value_object(dev_stats, "retry_buffer_overflow_count", entry_stats);
+
+ entry_stats = json_create_object();
+ json_object_add_value_int(entry_stats, "normalized", smart->pll_lock_loss_cnt.norm);
+ json_object_add_value_int(entry_stats, "raw", int48_to_long(smart->pll_lock_loss_cnt.raw));
+ json_object_add_value_object(dev_stats, "pll_lock_loss_count", entry_stats);
+
+ entry_stats = json_create_object();
+ json_object_add_value_int(entry_stats, "normalized", smart->nand_bytes_written.norm);
+ json_object_add_value_int(entry_stats, "raw", int48_to_long(smart->nand_bytes_written.raw));
+ json_object_add_value_object(dev_stats, "nand_bytes_written", entry_stats);
+
+ entry_stats = json_create_object();
+ json_object_add_value_int(entry_stats, "normalized", smart->host_bytes_written.norm);
+ json_object_add_value_int(entry_stats, "raw", int48_to_long(smart->host_bytes_written.raw));
+ json_object_add_value_object(dev_stats, "host_bytes_written", entry_stats);
+
+ entry_stats = json_create_object();
+ json_object_add_value_int(entry_stats, "normalized", smart->host_ctx_wear_used.norm);
+ json_object_add_value_int(entry_stats, "raw", int48_to_long(smart->host_ctx_wear_used.raw));
+ json_object_add_value_object(dev_stats, "host_ctx_wear_used", entry_stats);
+
+ entry_stats = json_create_object();
+ json_object_add_value_int(entry_stats, "normalized", smart->perf_stat_indicator.norm);
+ json_object_add_value_int(entry_stats, "raw", int48_to_long(smart->perf_stat_indicator.raw));
+ json_object_add_value_object(dev_stats, "perf_stat_indicator", entry_stats);
+
+ entry_stats = json_create_object();
+ json_object_add_value_int(entry_stats, "normalized", smart->re_alloc_sectr_cnt.norm);
+ json_object_add_value_int(entry_stats, "raw", int48_to_long(smart->re_alloc_sectr_cnt.raw));
+ json_object_add_value_object(dev_stats, "re_alloc_sectr_cnt", entry_stats);
+
+ entry_stats = json_create_object();
+ json_object_add_value_int(entry_stats, "normalized", smart->soft_ecc_err_rate.norm);
+ json_object_add_value_int(entry_stats, "raw", int48_to_long(smart->soft_ecc_err_rate.raw));
+ json_object_add_value_object(dev_stats, "soft_ecc_err_rate", entry_stats);
+
+ entry_stats = json_create_object();
+ json_object_add_value_int(entry_stats, "normalized", smart->unexp_power_loss.norm);
+ json_object_add_value_int(entry_stats, "raw", int48_to_long(smart->unexp_power_loss.raw));
+ json_object_add_value_object(dev_stats, "unexp_power_loss", entry_stats);
+
+ entry_stats = json_create_object();
+ json_object_add_value_int(entry_stats, "normalized", smart->media_bytes_read.norm);
+ json_object_add_value_int(entry_stats, "raw", int48_to_long(smart->media_bytes_read.raw));
+ json_object_add_value_object(dev_stats, "media_bytes_read", entry_stats);
+
+ entry_stats = json_create_object();
+ json_object_add_value_int(entry_stats, "normalized", smart->avail_fw_downgrades.norm);
+ json_object_add_value_int(entry_stats, "raw", int48_to_long(smart->avail_fw_downgrades.raw));
+ json_object_add_value_object(dev_stats, "avail_fw_downgrades", entry_stats);
+
+ json_object_add_value_object(root, "Device stats", dev_stats);
+
+ json_print_object(root, NULL);
+ json_free_object(root);
+}
+
+static char *id_to_key(__u8 id)
+{
+ switch (id) {
+ case 0xAB:
+ return "program_fail_count";
+ case 0xAC:
+ return "erase_fail_count";
+ case 0xAD:
+ return "wear_leveling_count";
+ case 0xB8:
+ return "e2e_error_detect_count";
+ case 0xC7:
+ return "crc_error_count";
+ case 0xE2:
+ return "media_wear_percentage";
+ case 0xE3:
+ return "host_reads";
+ case 0xE4:
+ return "timed_work_load";
+ case 0xEA:
+ return "thermal_throttle_status";
+ case 0xF0:
+ return "retry_buff_overflow_count";
+ case 0xF3:
+ return "pll_lock_loss_counter";
+ case 0xF4:
+ return "nand_bytes_written";
+ case 0xF5:
+ return "host_bytes_written";
+ case 0xF6:
+ return "host_context_wear_used";
+ case 0xF7:
+ return "performance_status_indicator";
+ case 0xF8:
+ return "media_bytes_read";
+ case 0xF9:
+ return "available_fw_downgrades";
+ case 0x05:
+ return "re-allocated_sector_count";
+ case 0x0D:
+ return "soft_ecc_error_rate";
+ case 0xAE:
+ return "unexpected_power_loss";
+ default:
+ return "Invalid ID";
+ }
+}
+
+static void print_intel_smart_log_items(struct nvme_additional_smart_log_item *item)
+{
+ if (!item->key)
+ return;
+
+ printf("%#x %-45s %3d %"PRIu64"\n",
+ item->key, id_to_key(item->key),
+ item->norm, int48_to_long(item->raw));
+}
+
+static void show_intel_smart_log(struct nvme_additional_smart_log *smart,
+ unsigned int nsid, const char *devname)
+{
+ struct nvme_additional_smart_log_item *iter = &smart->program_fail_cnt;
+ int num_items = sizeof(struct nvme_additional_smart_log) /
+ sizeof(struct nvme_additional_smart_log_item);
+
+ printf("Additional Smart Log for NVME device:%s namespace-id:%x\n",
+ devname, nsid);
+ printf("ID KEY Normalized Raw\n");
+
+ for (int i = 0; i < num_items; i++, iter++)
+ print_intel_smart_log_items(iter);
+}
+
+static int get_additional_smart_log(int argc, char **argv, struct command *cmd, struct plugin *plugin)
+{
+ const char *desc =
+ "Get Intel vendor specific additional smart log (optionally, for the specified namespace), and show it.";
+ const char *namespace = "(optional) desired namespace";
+ const char *raw = "Dump output in binary format";
+ const char *json = "Dump output in json format";
+
+ struct nvme_additional_smart_log smart_log;
+ struct nvme_dev *dev;
+ int err;
+
+ struct config {
+ __u32 namespace_id;
+ bool raw_binary;
+ bool json;
+ };
+
+ struct config cfg = {
+ .namespace_id = NVME_NSID_ALL,
+ };
+
+ OPT_ARGS(opts) = {
+ OPT_UINT("namespace-id", 'n', &cfg.namespace_id, namespace),
+ OPT_FLAG("raw-binary", 'b', &cfg.raw_binary, raw),
+ OPT_FLAG("json", 'j', &cfg.json, json),
+ OPT_END()
+ };
+
+ err = parse_and_open(&dev, argc, argv, desc, opts);
+ if (err)
+ return err;
+
+ err = nvme_get_log_simple(dev_fd(dev), 0xca, sizeof(smart_log),
+ &smart_log);
+ if (!err) {
+ if (cfg.json)
+ show_intel_smart_log_jsn(&smart_log, cfg.namespace_id,
+ dev->name);
+ else if (!cfg.raw_binary)
+ show_intel_smart_log(&smart_log, cfg.namespace_id,
+ dev->name);
+ else
+ d_raw((unsigned char *)&smart_log, sizeof(smart_log));
+ } else if (err > 0) {
+ nvme_show_status(err);
+ }
+ dev_close(dev);
+ return err;
+}
+
+static int get_market_log(int argc, char **argv, struct command *cmd, struct plugin *plugin)
+{
+ const char *desc = "Get Intel Marketing Name log and show it.";
+ const char *raw = "dump output in binary format";
+ struct nvme_dev *dev;
+ char log[512];
+ int err;
+
+ struct config {
+ bool raw_binary;
+ };
+
+ struct config cfg = {
+ };
+
+ OPT_ARGS(opts) = {
+ OPT_FLAG("raw-binary", 'b', &cfg.raw_binary, raw),
+ OPT_END()
+ };
+
+ err = parse_and_open(&dev, argc, argv, desc, opts);
+ if (err)
+ return err;
+
+ err = nvme_get_log_simple(dev_fd(dev), 0xdd, sizeof(log), log);
+ if (!err) {
+ if (!cfg.raw_binary)
+ printf("Intel Marketing Name Log:\n%s\n", log);
+ else
+ d_raw((unsigned char *)&log, sizeof(log));
+ } else if (err > 0)
+ nvme_show_status(err);
+ dev_close(dev);
+ return err;
+}
+
+struct intel_temp_stats {
+ __le64 curr;
+ __le64 last_overtemp;
+ __le64 life_overtemp;
+ __le64 highest_temp;
+ __le64 lowest_temp;
+ __u8 rsvd[40];
+ __le64 max_operating_temp;
+ __le64 min_operating_temp;
+ __le64 est_offset;
+};
+
+static void show_temp_stats(struct intel_temp_stats *stats)
+{
+ printf(" Intel Temperature Statistics\n");
+ printf("--------------------------------\n");
+ printf("Current temperature : %"PRIu64"\n", le64_to_cpu(stats->curr));
+ printf("Last critical overtemp flag : %"PRIu64"\n", le64_to_cpu(stats->last_overtemp));
+ printf("Life critical overtemp flag : %"PRIu64"\n", le64_to_cpu(stats->life_overtemp));
+ printf("Highest temperature : %"PRIu64"\n", le64_to_cpu(stats->highest_temp));
+ printf("Lowest temperature : %"PRIu64"\n", le64_to_cpu(stats->lowest_temp));
+ printf("Max operating temperature : %"PRIu64"\n", le64_to_cpu(stats->max_operating_temp));
+ printf("Min operating temperature : %"PRIu64"\n", le64_to_cpu(stats->min_operating_temp));
+ printf("Estimated offset : %"PRIu64"\n", le64_to_cpu(stats->est_offset));
+}
+
+static int get_temp_stats_log(int argc, char **argv, struct command *cmd, struct plugin *plugin)
+{
+ struct intel_temp_stats stats;
+ struct nvme_dev *dev;
+ int err;
+
+ const char *desc = "Get Temperature Statistics log and show it.";
+ const char *raw = "dump output in binary format";
+ struct config {
+ bool raw_binary;
+ };
+
+ struct config cfg = {
+ };
+
+ OPT_ARGS(opts) = {
+ OPT_FLAG("raw-binary", 'b', &cfg.raw_binary, raw),
+ OPT_END()
+ };
+
+ err = parse_and_open(&dev, argc, argv, desc, opts);
+ if (err)
+ return err;
+
+ err = nvme_get_log_simple(dev_fd(dev), 0xc5, sizeof(stats), &stats);
+ if (!err) {
+ if (!cfg.raw_binary)
+ show_temp_stats(&stats);
+ else
+ d_raw((unsigned char *)&stats, sizeof(stats));
+ } else if (err > 0)
+ nvme_show_status(err);
+ dev_close(dev);
+ return err;
+}
+
+struct intel_lat_stats {
+ __u16 maj;
+ __u16 min;
+ __u32 data[1216];
+};
+
+struct __packed optane_lat_stats {
+ __u16 maj;
+ __u16 min;
+ __u64 data[9];
+};
+
+#define MEDIA_MAJOR_IDX 0
+#define MEDIA_MINOR_IDX 1
+#define MEDIA_MAX_LEN 2
+#define OPTANE_V1000_BUCKET_LEN 8
+#define OPTANE_V1000_BUCKET_LEN 8
+#define NAND_LAT_STATS_LEN 4868
+
+
+static struct intel_lat_stats stats;
+static struct optane_lat_stats v1000_stats;
+
+struct v1000_thresholds {
+ __u32 read[OPTANE_V1000_BUCKET_LEN];
+ __u32 write[OPTANE_V1000_BUCKET_LEN];
+};
+
+static struct v1000_thresholds v1000_bucket;
+static __u16 media_version[MEDIA_MAX_LEN] = {0};
+
+enum FormatUnit {
+ US,
+ MS,
+ S
+};
+
+/*
+ * COL_WIDTH controls width of columns in human-readable output.
+ * BUFSIZE is for local temp char[]
+ * US_IN_S and US_IN_MS are for unit conversions when printing.
+ */
+#define COL_WIDTH 12
+#define BUFSIZE 10
+#define US_IN_S 1000000
+#define US_IN_MS 1000
+
+static enum FormatUnit get_seconds_magnitude(__u32 microseconds)
+{
+ if (microseconds > US_IN_S)
+ return S;
+ else if (microseconds > US_IN_MS)
+ return MS;
+ else
+ return US;
+}
+
+static float convert_seconds(__u32 microseconds)
+{
+ float divisor = 1.0;
+
+ if (microseconds > US_IN_S)
+ divisor = US_IN_S;
+ else if (microseconds > US_IN_MS)
+ divisor = US_IN_MS;
+ return microseconds / divisor;
+}
+
+/*
+ * For control over whether a string will format to +/-INF or
+ * print out ####.##US normally.
+ */
+enum inf_bound_type {
+ NEGINF,
+ POSINF,
+ NOINF
+};
+
+/*
+ * Edge buckets may have range [#s, inf) or (-inf, #US] in some
+ * latency statistics formats.
+ * Passing in NEGINF to POSINF to bound_type overrides the string to
+ * either of "-INF" or "+INF", respectively.
+ */
+static void set_unit_string(char *buffer, __u32 microseconds,
+ enum FormatUnit unit,
+ enum inf_bound_type bound_type)
+{
+ char *string;
+
+ if (bound_type != NOINF) {
+ snprintf(buffer, BUFSIZE, "%s", bound_type ? "+INF" : "-INF");
+ return;
+ }
+
+ switch (unit) {
+ case US:
+ string = "us";
+ break;
+ case MS:
+ string = "ms";
+ break;
+ case S:
+ string = "s";
+ break;
+ default:
+ string = "_s";
+ break;
+ }
+
+ snprintf(buffer, BUFSIZE, "%4.2f%s", convert_seconds(microseconds),
+ string);
+}
+
+static void init_buffer(char *buffer, size_t size)
+{
+ size_t i;
+
+ for (i = 0; i < size; i++)
+ buffer[i] = i + '0';
+}
+
+static void show_lat_stats_bucket(struct intel_lat_stats *stats,
+ __u32 lower_us, enum inf_bound_type start_type,
+ __u32 upper_us, enum inf_bound_type end_type, int i)
+{
+ enum FormatUnit fu = S;
+ char buffer[BUFSIZE];
+
+ init_buffer(buffer, BUFSIZE);
+ printf("%-*d", COL_WIDTH, i);
+
+ fu = get_seconds_magnitude(lower_us);
+ set_unit_string(buffer, lower_us, fu, start_type);
+ printf("%-*s", COL_WIDTH, buffer);
+
+ fu = get_seconds_magnitude(upper_us);
+ set_unit_string(buffer, upper_us, fu, end_type);
+ printf("%-*s", COL_WIDTH, buffer);
+
+ printf("%-*d\n", COL_WIDTH, stats->data[i]);
+}
+
+static void show_optane_lat_stats_bucket(struct optane_lat_stats *stats,
+ __u32 lower_us, enum inf_bound_type start_type,
+ __u32 upper_us, enum inf_bound_type end_type, int i)
+{
+ enum FormatUnit fu = S;
+ char buffer[BUFSIZE];
+
+ init_buffer(buffer, BUFSIZE);
+ printf("%-*d", COL_WIDTH, i);
+
+ fu = get_seconds_magnitude(lower_us);
+ set_unit_string(buffer, lower_us, fu, start_type);
+ printf("%-*s", COL_WIDTH, buffer);
+
+ fu = get_seconds_magnitude(upper_us);
+ set_unit_string(buffer, upper_us, fu, end_type);
+ printf("%-*s", COL_WIDTH, buffer);
+
+ printf("%-*lu\n", COL_WIDTH, (unsigned long)stats->data[i]);
+}
+
+
+static void show_lat_stats_linear(struct intel_lat_stats *stats,
+ __u32 start_offset, __u32 end_offset, __u32 bytes_per,
+ __u32 us_step, bool nonzero_print)
+{
+ for (int i = (start_offset / bytes_per) - 1;
+ i < end_offset / bytes_per; i++) {
+ if (nonzero_print && stats->data[i] == 0)
+ continue;
+ show_lat_stats_bucket(stats, us_step * i, NOINF,
+ us_step * (i + 1), NOINF, i);
+ }
+}
+
+/*
+ * For 4.0-4.5 revision.
+ */
+#define LATENCY_STATS_V4_BASE_BITS 6
+#define LATENCY_STATS_V4_BASE_VAL (1 << LATENCY_STATS_V4_BASE_BITS)
+
+static int lat_stats_log_scale(int i)
+{
+ // if (i < 128)
+ if (i < (LATENCY_STATS_V4_BASE_VAL << 1))
+ return i;
+
+ int error_bits = (i >> LATENCY_STATS_V4_BASE_BITS) - 1;
+ int base = 1 << (error_bits + LATENCY_STATS_V4_BASE_BITS);
+ int k = i % LATENCY_STATS_V4_BASE_VAL;
+
+ return base + ((k + 0.5) * (1 << error_bits));
+}
+
+/*
+ * Creates a subroot in the following manner:
+ * {
+ * "latstats" : {
+ * "type" : "write" or "read",
+ * "values" : {
+ */
+static void lat_stats_make_json_root(
+ struct json_object *root, struct json_object *bucket_list,
+ int write)
+{
+ struct json_object *subroot = json_create_object();
+
+ json_object_add_value_object(root, "latstats", subroot);
+ json_object_add_value_string(subroot, "type", write ? "write" : "read");
+ json_object_add_value_object(subroot, "values", bucket_list);
+}
+
+/*
+ * Creates a bucket under the "values" json_object. Format is:
+ * "values" : {
+ * "bucket" : {
+ * "id" : #,
+ * "start" : string,
+ * "end" : string,
+ * "value" : 0,
+ * },
+ */
+static void json_add_bucket(struct intel_lat_stats *stats,
+ struct json_object *bucket_list, __u32 id,
+ __u32 lower_us, enum inf_bound_type start_type,
+ __u32 upper_us, enum inf_bound_type end_type, __u32 val)
+{
+ char buffer[BUFSIZE];
+ struct json_object *bucket = json_create_object();
+
+ init_buffer(buffer, BUFSIZE);
+
+ json_object_array_add(bucket_list, bucket);
+ json_object_add_value_int(bucket, "id", id);
+
+ set_unit_string(buffer, lower_us,
+ get_seconds_magnitude(lower_us), start_type);
+ json_object_add_value_string(bucket, "start", buffer);
+
+ set_unit_string(buffer, upper_us,
+ get_seconds_magnitude(upper_us), end_type);
+ json_object_add_value_string(bucket, "end", buffer);
+
+ json_object_add_value_int(bucket, "value", val);
+}
+
+static void json_add_bucket_optane(struct json_object *bucket_list, __u32 id,
+ __u32 lower_us, enum inf_bound_type start_type,
+ __u32 upper_us, enum inf_bound_type end_type, __u32 val)
+{
+ char buffer[BUFSIZE];
+ struct json_object *bucket = json_create_object();
+
+ init_buffer(buffer, BUFSIZE);
+
+ json_object_array_add(bucket_list, bucket);
+ json_object_add_value_int(bucket, "id", id);
+
+ set_unit_string(buffer, lower_us,
+ get_seconds_magnitude(lower_us), start_type);
+ json_object_add_value_string(bucket, "start", buffer);
+
+ set_unit_string(buffer, upper_us,
+ get_seconds_magnitude(upper_us), end_type);
+ json_object_add_value_string(bucket, "end", buffer);
+
+ json_object_add_value_uint(bucket, "value", val);
+
+}
+
+static void json_lat_stats_linear(struct intel_lat_stats *stats,
+ struct json_object *bucket_list, __u32 start_offset,
+ __u32 end_offset, __u32 bytes_per,
+ __u32 us_step, bool nonzero_print)
+{
+ for (int i = (start_offset / bytes_per) - 1;
+ i < end_offset / bytes_per; i++) {
+ if (nonzero_print && stats->data[i] == 0)
+ continue;
+
+ json_add_bucket(stats, bucket_list,
+ i, us_step * i, NOINF, us_step * (i + 1),
+ NOINF, stats->data[i]);
+ }
+}
+
+static void json_lat_stats_3_0(struct intel_lat_stats *stats,
+ int write)
+{
+ struct json_object *root = json_create_object();
+ struct json_object *bucket_list = json_object_new_array();
+
+ lat_stats_make_json_root(root, bucket_list, write);
+
+ json_lat_stats_linear(stats, bucket_list, 4, 131, 4, 32, false);
+ json_lat_stats_linear(stats, bucket_list, 132, 255, 4, 1024, false);
+ json_lat_stats_linear(stats, bucket_list, 256, 379, 4, 32768, false);
+ json_lat_stats_linear(stats, bucket_list, 380, 383, 4, 32, true);
+ json_lat_stats_linear(stats, bucket_list, 384, 387, 4, 32, true);
+ json_lat_stats_linear(stats, bucket_list, 388, 391, 4, 32, true);
+
+ json_print_object(root, NULL);
+ json_free_object(root);
+}
+
+static void json_lat_stats_4_0(struct intel_lat_stats *stats,
+ int write)
+{
+ struct json_object *root = json_create_object();
+ struct json_object *bucket_list = json_object_new_array();
+
+ lat_stats_make_json_root(root, bucket_list, write);
+
+ __u32 lower_us = 0;
+ __u32 upper_us = 1;
+ bool end = false;
+ int max = 1216;
+
+ for (int i = 0; i < max; i++) {
+ lower_us = lat_stats_log_scale(i);
+ if (i >= max - 1)
+ end = true;
+ else
+ upper_us = lat_stats_log_scale(i + 1);
+
+ json_add_bucket(stats, bucket_list, i,
+ lower_us, NOINF, upper_us,
+ end ? POSINF : NOINF, stats->data[i]);
+ }
+ json_print_object(root, NULL);
+ json_free_object(root);
+}
+
+static void show_lat_stats_3_0(struct intel_lat_stats *stats)
+{
+ show_lat_stats_linear(stats, 4, 131, 4, 32, false);
+ show_lat_stats_linear(stats, 132, 255, 4, 1024, false);
+ show_lat_stats_linear(stats, 256, 379, 4, 32768, false);
+ show_lat_stats_linear(stats, 380, 383, 4, 32, true);
+ show_lat_stats_linear(stats, 384, 387, 4, 32, true);
+ show_lat_stats_linear(stats, 388, 391, 4, 32, true);
+}
+
+static void show_lat_stats_4_0(struct intel_lat_stats *stats)
+{
+ int lower_us = 0;
+ int upper_us = 1;
+ bool end = false;
+ int max = 1216;
+
+ for (int i = 0; i < max; i++) {
+ lower_us = lat_stats_log_scale(i);
+ if (i >= max - 1)
+ end = true;
+ else
+ upper_us = lat_stats_log_scale(i + 1);
+
+ show_lat_stats_bucket(stats, lower_us, NOINF,
+ upper_us, end ? POSINF : NOINF, i);
+ }
+}
+
+static void json_lat_stats_v1000_0(struct optane_lat_stats *stats, int write)
+{
+ int i;
+ struct json_object *root = json_create_object();
+ struct json_object *bucket_list = json_object_new_array();
+
+ lat_stats_make_json_root(root, bucket_list, write);
+
+ if (write) {
+ for (i = 0; i < OPTANE_V1000_BUCKET_LEN - 1; i++)
+ json_add_bucket_optane(bucket_list, i,
+ v1000_bucket.write[i], NOINF,
+ v1000_bucket.write[i + 1] - 1,
+ NOINF,
+ stats->data[i]);
+
+ json_add_bucket_optane(bucket_list,
+ OPTANE_V1000_BUCKET_LEN - 1,
+ v1000_bucket.write[i],
+ NOINF,
+ v1000_bucket.write[i],
+ POSINF, stats->data[i]);
+ } else {
+ for (i = 0; i < OPTANE_V1000_BUCKET_LEN - 1; i++)
+ json_add_bucket_optane(bucket_list, i,
+ v1000_bucket.read[i], NOINF,
+ v1000_bucket.read[i + 1] - 1,
+ NOINF,
+ stats->data[i]);
+
+ json_add_bucket_optane(bucket_list,
+ OPTANE_V1000_BUCKET_LEN - 1,
+ v1000_bucket.read[i],
+ NOINF,
+ v1000_bucket.read[i],
+ POSINF, stats->data[i]);
+ }
+
+ struct json_object *subroot = json_create_object();
+
+ json_object_add_value_object(root, "Average latency since last reset", subroot);
+
+ json_object_add_value_uint(subroot, "value in us", stats->data[8]);
+
+ json_print_object(root, NULL);
+ json_free_object(root);
+
+}
+
+static void show_lat_stats_v1000_0(struct optane_lat_stats *stats, int write)
+{
+ int i;
+
+ if (write) {
+ for (i = 0; i < OPTANE_V1000_BUCKET_LEN - 1; i++)
+ show_optane_lat_stats_bucket(stats,
+ v1000_bucket.write[i],
+ NOINF,
+ v1000_bucket.write[i + 1] - 1,
+ NOINF, i);
+
+ show_optane_lat_stats_bucket(stats, v1000_bucket.write[i],
+ NOINF, v1000_bucket.write[i],
+ POSINF, i);
+ } else {
+ for (i = 0; i < OPTANE_V1000_BUCKET_LEN - 1; i++)
+ show_optane_lat_stats_bucket(stats,
+ v1000_bucket.read[i],
+ NOINF,
+ v1000_bucket.read[i + 1] - 1,
+ NOINF, i);
+
+ show_optane_lat_stats_bucket(stats, v1000_bucket.read[i],
+ NOINF, v1000_bucket.read[i],
+ POSINF, i);
+ }
+
+ printf("Average latency since last reset: %lu us\n", (unsigned long)stats->data[8]);
+
+}
+
+static void json_lat_stats(int write)
+{
+ switch (media_version[MEDIA_MAJOR_IDX]) {
+ case 3:
+ json_lat_stats_3_0(&stats, write);
+ break;
+ case 4:
+ switch (media_version[MEDIA_MINOR_IDX]) {
+ case 0:
+ case 1:
+ case 2:
+ case 3:
+ case 4:
+ case 5:
+ json_lat_stats_4_0(&stats, write);
+ break;
+ default:
+ printf("Unsupported minor revision (%u.%u)\n",
+ stats.maj, stats.min);
+ break;
+ }
+ break;
+ case 1000:
+ switch (media_version[MEDIA_MINOR_IDX]) {
+ case 0:
+ json_lat_stats_v1000_0(&v1000_stats, write);
+ break;
+ default:
+ printf("Unsupported minor revision (%u.%u)\n",
+ stats.maj, stats.min);
+ break;
+ }
+ break;
+ default:
+ printf("Unsupported revision (%u.%u)\n",
+ stats.maj, stats.min);
+ break;
+ }
+ printf("\n");
+}
+
+static void print_dash_separator(int count)
+{
+ for (int i = 0; i < count; i++)
+ putchar('-');
+ putchar('\n');
+}
+
+static void show_lat_stats(int write)
+{
+ static const int separator_length = 50;
+
+ printf("Intel IO %s Command Latency Statistics\n",
+ write ? "Write" : "Read");
+ printf("Major Revision : %u\nMinor Revision : %u\n",
+ media_version[MEDIA_MAJOR_IDX], media_version[MEDIA_MINOR_IDX]);
+ print_dash_separator(separator_length);
+ printf("%-12s%-12s%-12s%-20s\n", "Bucket", "Start", "End", "Value");
+ print_dash_separator(separator_length);
+
+ switch (media_version[MEDIA_MAJOR_IDX]) {
+ case 3:
+ show_lat_stats_3_0(&stats);
+ break;
+ case 4:
+ switch (media_version[MEDIA_MINOR_IDX]) {
+ case 0:
+ case 1:
+ case 2:
+ case 3:
+ case 4:
+ case 5:
+ case 6:
+ show_lat_stats_4_0(&stats);
+ break;
+ default:
+ printf("Unsupported minor revision (%u.%u)\n",
+ stats.maj, stats.min);
+ break;
+ }
+ break;
+ case 1000:
+ switch (media_version[MEDIA_MINOR_IDX]) {
+ case 0:
+ show_lat_stats_v1000_0(&v1000_stats, write);
+ break;
+ default:
+ printf("Unsupported minor revision (%u.%u)\n",
+ stats.maj, stats.min);
+ break;
+ }
+ break;
+ default:
+ printf("Unsupported revision (%u.%u)\n",
+ stats.maj, stats.min);
+ break;
+ }
+}
+
+static int get_lat_stats_log(int argc, char **argv, struct command *cmd, struct plugin *plugin)
+{
+ __u8 data[NAND_LAT_STATS_LEN];
+ struct nvme_dev *dev;
+ int err;
+
+ const char *desc = "Get Intel Latency Statistics log and show it.";
+ const char *raw = "Dump output in binary format";
+ const char *json = "Dump output in json format";
+ const char *write = "Get write statistics (read default)";
+
+ struct config {
+ bool raw_binary;
+ bool json;
+ bool write;
+ };
+
+ struct config cfg = {
+ };
+
+ OPT_ARGS(opts) = {
+ OPT_FLAG("write", 'w', &cfg.write, write),
+ OPT_FLAG("raw-binary", 'b', &cfg.raw_binary, raw),
+ OPT_FLAG("json", 'j', &cfg.json, json),
+ OPT_END()
+ };
+
+ err = parse_and_open(&dev, argc, argv, desc, opts);
+ if (err)
+ return err;
+
+ /* For optate, latency stats are deleted every time their LID is pulled.
+ * Therefore, we query the longest lat_stats log page first.
+ */
+ err = nvme_get_log_simple(dev_fd(dev), cfg.write ? 0xc2 : 0xc1,
+ sizeof(data), &data);
+
+ media_version[0] = (data[1] << 8) | data[0];
+ media_version[1] = (data[3] << 8) | data[2];
+
+ if (err)
+ goto close_dev;
+
+ if (media_version[0] == 1000) {
+ __u32 thresholds[OPTANE_V1000_BUCKET_LEN] = {0};
+ __u32 result;
+
+ struct nvme_get_features_args args = {
+ .args_size = sizeof(args),
+ .fd = dev_fd(dev),
+ .fid = 0xf7,
+ .nsid = 0,
+ .sel = 0,
+ .cdw11 = cfg.write ? 0x1 : 0x0,
+ .uuidx = 0,
+ .data_len = sizeof(thresholds),
+ .data = thresholds,
+ .timeout = NVME_DEFAULT_IOCTL_TIMEOUT,
+ .result = &result,
+ };
+ err = nvme_get_features(&args);
+ if (err) {
+ fprintf(stderr, "Querying thresholds failed. ");
+ nvme_show_status(err);
+ goto close_dev;
+ }
+
+ /* Update bucket thresholds to be printed */
+ if (cfg.write) {
+ for (int i = 0; i < OPTANE_V1000_BUCKET_LEN; i++)
+ v1000_bucket.write[i] = thresholds[i];
+ } else {
+ for (int i = 0; i < OPTANE_V1000_BUCKET_LEN; i++)
+ v1000_bucket.read[i] = thresholds[i];
+
+ }
+
+ /* Move counter values to optate stats struct which has a
+ * smaller size
+ */
+ memcpy(&v1000_stats, (struct optane_lat_stats *)data,
+ sizeof(struct optane_lat_stats));
+ } else {
+ memcpy(&stats, (struct intel_lat_stats *)data,
+ sizeof(struct intel_lat_stats));
+ }
+
+ if (cfg.json)
+ json_lat_stats(cfg.write);
+ else if (!cfg.raw_binary)
+ show_lat_stats(cfg.write);
+ else {
+ if (media_version[0] == 1000)
+ d_raw((unsigned char *)&v1000_stats,
+ sizeof(v1000_stats));
+ else
+ d_raw((unsigned char *)&stats,
+ sizeof(stats));
+ }
+
+close_dev:
+ dev_close(dev);
+ return err;
+}
+
+struct intel_assert_dump {
+ __u32 coreoffset;
+ __u32 assertsize;
+ __u8 assertdumptype;
+ __u8 assertvalid;
+ __u8 reserved[2];
+};
+
+struct intel_event_dump {
+ __u32 numeventdumps;
+ __u32 coresize;
+ __u32 coreoffset;
+ __u32 eventidoffset[16];
+ __u8 eventIdValidity[16];
+};
+
+struct intel_vu_version {
+ __u16 major;
+ __u16 minor;
+};
+
+struct intel_event_header {
+ __u32 eventidsize;
+ struct intel_event_dump edumps[0];
+};
+
+struct intel_vu_log {
+ struct intel_vu_version ver;
+ __u32 header;
+ __u32 size;
+ __u32 numcores;
+ __u8 reserved[4080];
+};
+
+struct intel_vu_nlog {
+ struct intel_vu_version ver;
+ __u32 logselect;
+ __u32 totalnlogs;
+ __u32 nlognum;
+ __u32 nlogname;
+ __u32 nlogbytesize;
+ __u32 nlogprimarybuffsize;
+ __u32 tickspersecond;
+ __u32 corecount;
+ __u32 nlogpausestatus;
+ __u32 selectoffsetref;
+ __u32 selectnlogpause;
+ __u32 selectaddedoffset;
+ __u32 nlogbufnum;
+ __u32 nlogbufnummax;
+ __u32 coreselected;
+ __u32 reserved[3];
+};
+
+struct intel_cd_log {
+ union {
+ struct {
+ __u32 selectLog : 3;
+ __u32 selectCore : 2;
+ __u32 selectNlog : 8;
+ __u8 selectOffsetRef : 1;
+ __u32 selectNlogPause : 2;
+ __u32 reserved2 : 16;
+ } fields;
+ __u32 entireDword;
+ } u;
+};
+
+static void print_intel_nlog(struct intel_vu_nlog *intel_nlog)
+{
+ printf("Version Major %u\n"
+ "Version Minor %u\n"
+ "Log_select %u\n"
+ "totalnlogs %u\n"
+ "nlognum %u\n"
+ "nlogname %u\n"
+ "nlogbytesze %u\n"
+ "nlogprimarybuffsize %u\n"
+ "tickspersecond %u\n"
+ "corecount %u\n"
+ "nlogpausestatus %u\n"
+ "selectoffsetref %u\n"
+ "selectnlogpause %u\n"
+ "selectaddedoffset %u\n"
+ "nlogbufnum %u\n"
+ "nlogbufnummax %u\n"
+ "coreselected %u\n",
+ intel_nlog->ver.major, intel_nlog->ver.minor,
+ intel_nlog->logselect, intel_nlog->totalnlogs, intel_nlog->nlognum,
+ intel_nlog->nlogname, intel_nlog->nlogbytesize,
+ intel_nlog->nlogprimarybuffsize, intel_nlog->tickspersecond,
+ intel_nlog->corecount, intel_nlog->nlogpausestatus,
+ intel_nlog->selectoffsetref, intel_nlog->selectnlogpause,
+ intel_nlog->selectaddedoffset, intel_nlog->nlogbufnum,
+ intel_nlog->nlogbufnummax, intel_nlog->coreselected);
+}
+
+static int read_entire_cmd(struct nvme_passthru_cmd *cmd, int total_size,
+ const size_t max_tfer, int out_fd, int ioctl_fd,
+ __u8 *buf)
+{
+ int err = 0;
+ size_t dword_tfer = 0;
+
+ dword_tfer = min(max_tfer, total_size);
+ while (total_size > 0) {
+ err = nvme_submit_admin_passthru(ioctl_fd, cmd, NULL);
+ if (err) {
+ fprintf(stderr,
+ "failed on cmd.data_len %u cmd.cdw13 %u cmd.cdw12 %x cmd.cdw10 %u err %x remaining size %d\n",
+ cmd->data_len, cmd->cdw13, cmd->cdw12,
+ cmd->cdw10, err, total_size);
+ goto out;
+ }
+
+ if (out_fd > 0) {
+ err = write(out_fd, buf, cmd->data_len);
+ if (err < 0) {
+ perror("write failure");
+ goto out;
+ }
+ err = 0;
+ }
+ total_size -= dword_tfer;
+ cmd->cdw13 += dword_tfer;
+ cmd->cdw10 = dword_tfer = min(max_tfer, total_size);
+ cmd->data_len = (min(max_tfer, total_size)) * 4;
+ }
+
+ out:
+ return err;
+}
+
+static int write_header(__u8 *buf, int fd, size_t amnt)
+{
+ if (write(fd, buf, amnt) < 0)
+ return 1;
+ return 0;
+}
+
+static int read_header(struct nvme_passthru_cmd *cmd, __u8 *buf, int ioctl_fd,
+ __u32 dw12, int nsid)
+{
+ memset(cmd, 0, sizeof(*cmd));
+ memset(buf, 0, 4096);
+ cmd->opcode = 0xd2;
+ cmd->nsid = nsid;
+ cmd->cdw10 = 0x400;
+ cmd->cdw12 = dw12;
+ cmd->data_len = 0x1000;
+ cmd->addr = (unsigned long)(void *)buf;
+ return read_entire_cmd(cmd, 0x400, 0x400, -1, ioctl_fd, buf);
+}
+
+static int setup_file(char *f, char *file, int fd, int type)
+{
+ struct nvme_id_ctrl ctrl;
+ int err = 0, i = sizeof(ctrl.sn) - 1;
+
+ err = nvme_identify_ctrl(fd, &ctrl);
+ if (err)
+ return err;
+
+ /* Remove trailing spaces from the name */
+ while (i && ctrl.sn[i] == ' ') {
+ ctrl.sn[i] = '\0';
+ i--;
+ }
+
+ sprintf(f, "%s_%-.*s.bin", type == 0 ? "Nlog" :
+ type == 1 ? "EventLog" : "AssertLog",
+ (int)sizeof(ctrl.sn), ctrl.sn);
+ return err;
+}
+
+static int get_internal_log_old(__u8 *buf, int output, int fd,
+ struct nvme_passthru_cmd *cmd)
+{
+ struct intel_vu_log *intel;
+ int err = 0;
+ const int dwmax = 0x400;
+ const int dmamax = 0x1000;
+
+ intel = (struct intel_vu_log *)buf;
+
+ printf("Log major:%d minor:%d header:%d size:%d\n",
+ intel->ver.major, intel->ver.minor, intel->header, intel->size);
+
+ err = write(output, buf, 0x1000);
+ if (err < 0) {
+ perror("write failure");
+ goto out;
+ }
+ intel->size -= 0x400;
+ cmd->opcode = 0xd2;
+ cmd->cdw10 = min(dwmax, intel->size);
+ cmd->data_len = min(dmamax, intel->size);
+ err = read_entire_cmd(cmd, intel->size, dwmax, output, fd, buf);
+ if (err)
+ goto out;
+
+ err = 0;
+ out:
+ return err;
+}
+
+static int get_internal_log(int argc, char **argv, struct command *command,
+ struct plugin *plugin)
+{
+ __u8 buf[0x2000];
+ char f[0x100];
+ int err, output, i, j, count = 0, core_num = 1;
+ struct nvme_passthru_cmd cmd;
+ struct intel_cd_log cdlog;
+ struct intel_vu_log *intel = malloc(sizeof(struct intel_vu_log));
+ struct intel_vu_nlog *intel_nlog = (struct intel_vu_nlog *)buf;
+ struct intel_assert_dump *ad = (struct intel_assert_dump *) intel->reserved;
+ struct intel_event_header *ehdr = (struct intel_event_header *)intel->reserved;
+ struct nvme_dev *dev;
+
+ const char *desc = "Get Intel Firmware Log and save it.";
+ const char *log = "Log type: 0, 1, or 2 for nlog, event log, and assert log, respectively.";
+ const char *core = "Select which region log should come from. -1 for all";
+ const char *nlognum = "Select which nlog to read. -1 for all nlogs";
+ const char *file = "Output file; defaults to device name provided";
+ const char *verbose = "To print out verbose nlog info";
+ const char *namespace_id = "Namespace to get logs from";
+
+ struct config {
+ __u32 namespace_id;
+ __u32 log;
+ int core;
+ int lnum;
+ char *file;
+ bool verbose;
+ };
+
+ struct config cfg = {
+ .namespace_id = -1,
+ .file = NULL,
+ .lnum = -1,
+ .core = -1
+ };
+
+ OPT_ARGS(opts) = {
+ OPT_UINT("log", 'l', &cfg.log, log),
+ OPT_INT("region", 'r', &cfg.core, core),
+ OPT_INT("nlognum", 'm', &cfg.lnum, nlognum),
+ OPT_UINT("namespace-id", 'n', &cfg.namespace_id, namespace_id),
+ OPT_FILE("output-file", 'o', &cfg.file, file),
+ OPT_FLAG("verbose-nlog", 'v', &cfg.verbose, verbose),
+ OPT_END()
+ };
+
+ err = parse_and_open(&dev, argc, argv, desc, opts);
+ if (err) {
+ free(intel);
+ return err;
+ }
+
+ if (cfg.log > 2 || cfg.core > 4 || cfg.lnum > 255) {
+ err = -EINVAL;
+ goto out_free;
+ }
+
+ if (!cfg.file) {
+ err = setup_file(f, cfg.file, dev_fd(dev), cfg.log);
+ if (err)
+ goto out_free;
+ cfg.file = f;
+ }
+
+ cdlog.u.entireDword = 0;
+
+ cdlog.u.fields.selectLog = cfg.log;
+ cdlog.u.fields.selectCore = cfg.core < 0 ? 0 : cfg.core;
+ cdlog.u.fields.selectNlog = cfg.lnum < 0 ? 0 : cfg.lnum;
+
+ output = open(cfg.file, O_WRONLY | O_CREAT | O_TRUNC, 0666);
+ if (output < 0) {
+ err = output;
+ goto out_free;
+ }
+
+ err = read_header(&cmd, buf, dev_fd(dev), cdlog.u.entireDword,
+ cfg.namespace_id);
+ if (err)
+ goto out;
+ memcpy(intel, buf, sizeof(*intel));
+
+ /* for 1.1 Fultondales will use old nlog, but current assert/event */
+ if ((intel->ver.major < 1 && intel->ver.minor < 1) ||
+ (intel->ver.major <= 1 && intel->ver.minor <= 1 && cfg.log == 0)) {
+ cmd.addr = (unsigned long)(void *)buf;
+ err = get_internal_log_old(buf, output, dev_fd(dev), &cmd);
+ goto out;
+ }
+
+ if (cfg.log == 2) {
+ if (cfg.verbose)
+ printf("Log major:%d minor:%d header:%d size:%d numcores:%d\n",
+ intel->ver.major, intel->ver.minor,
+ intel->header, intel->size, intel->numcores);
+
+ err = write_header(buf, output, 0x1000);
+ if (err) {
+ perror("write failure");
+ goto out;
+ }
+
+ count = intel->numcores;
+ } else if (cfg.log == 0) {
+ if (cfg.lnum < 0)
+ count = intel_nlog->totalnlogs;
+ else
+ count = 1;
+ if (cfg.core < 0)
+ core_num = intel_nlog->corecount;
+ } else if (cfg.log == 1) {
+ core_num = intel->numcores;
+ count = 1;
+ err = write_header(buf, output, sizeof(*intel));
+ if (err)
+ goto out;
+ }
+
+ for (j = (cfg.core < 0 ? 0 : cfg.core);
+ j < (cfg.core < 0 ? core_num : cfg.core + 1);
+ j++) {
+ cdlog.u.fields.selectCore = j;
+ for (i = 0; i < count; i++) {
+ if (cfg.log == 2) {
+ if (!ad[i].assertvalid)
+ continue;
+ cmd.cdw13 = ad[i].coreoffset;
+ cmd.cdw10 = 0x400;
+ cmd.data_len = min(0x400, ad[i].assertsize) * 4;
+ err = read_entire_cmd(&cmd, ad[i].assertsize,
+ 0x400, output,
+ dev_fd(dev),
+ buf);
+ if (err)
+ goto out;
+
+ } else if (cfg.log == 0) {
+ /* If the user selected to read the entire nlog */
+ if (count > 1)
+ cdlog.u.fields.selectNlog = i;
+
+ err = read_header(&cmd, buf, dev_fd(dev),
+ cdlog.u.entireDword,
+ cfg.namespace_id);
+ if (err)
+ goto out;
+ err = write_header(buf, output, sizeof(*intel_nlog));
+ if (err)
+ goto out;
+ if (cfg.verbose)
+ print_intel_nlog(intel_nlog);
+ cmd.cdw13 = 0x400;
+ cmd.cdw10 = 0x400;
+ cmd.data_len = min(0x1000, intel_nlog->nlogbytesize);
+ err = read_entire_cmd(&cmd, intel_nlog->nlogbytesize / 4,
+ 0x400, output,
+ dev_fd(dev),
+ buf);
+ if (err)
+ goto out;
+ } else if (cfg.log == 1) {
+ cmd.cdw13 = ehdr->edumps[j].coreoffset;
+ cmd.cdw10 = 0x400;
+ cmd.data_len = 0x400;
+ err = read_entire_cmd(&cmd, ehdr->edumps[j].coresize,
+ 0x400, output,
+ dev_fd(dev),
+ buf);
+ if (err)
+ goto out;
+ }
+ }
+ }
+ err = 0;
+out:
+ if (err > 0) {
+ nvme_show_status(err);
+ } else if (err < 0) {
+ perror("intel log");
+ err = EIO;
+ } else
+ printf("Successfully wrote log to %s\n", cfg.file);
+ close(output);
+out_free:
+ free(intel);
+ dev_close(dev);
+ return err;
+}
+
+static int enable_lat_stats_tracking(int argc, char **argv,
+ struct command *command, struct plugin *plugin)
+{
+ const char *desc = (
+ "Enable/Disable Intel Latency Statistics Tracking.\n"
+ "No argument prints current status.");
+ const char *enable_desc = "Enable LST";
+ const char *disable_desc = "Disable LST";
+ const __u32 nsid = 0;
+ const __u8 fid = 0xe2;
+ const __u8 sel = 0;
+ const __u32 cdw11 = 0x0;
+ const __u32 cdw12 = 0x0;
+ const __u32 data_len = 32;
+ const __u32 save = 0;
+ struct nvme_dev *dev;
+ void *buf = NULL;
+ __u32 result;
+ int err;
+
+ struct config {
+ bool enable, disable;
+ };
+
+ struct config cfg = {
+ .enable = false,
+ .disable = false,
+ };
+
+ struct argconfig_commandline_options command_line_options[] = {
+ {"enable", 'e', "", CFG_FLAG, &cfg.enable, no_argument, enable_desc},
+ {"disable", 'd', "", CFG_FLAG, &cfg.disable, no_argument, disable_desc},
+ {NULL}
+ };
+
+ err = parse_and_open(&dev, argc, argv, desc, command_line_options);
+
+ enum Option {
+ None = -1,
+ True = 1,
+ False = 0,
+ };
+
+ enum Option option = None;
+
+ if (cfg.enable && cfg.disable)
+ printf("Cannot enable and disable simultaneously.");
+ else if (cfg.enable || cfg.disable)
+ option = cfg.enable;
+
+ if (err)
+ return err;
+
+ struct nvme_get_features_args args_get = {
+ .args_size = sizeof(args_get),
+ .fd = dev_fd(dev),
+ .fid = fid,
+ .nsid = nsid,
+ .sel = sel,
+ .cdw11 = cdw11,
+ .uuidx = 0,
+ .data_len = data_len,
+ .data = buf,
+ .timeout = NVME_DEFAULT_IOCTL_TIMEOUT,
+ .result = &result,
+ };
+
+ struct nvme_set_features_args args_set = {
+ .args_size = sizeof(args_set),
+ .fd = dev_fd(dev),
+ .fid = fid,
+ .nsid = nsid,
+ .cdw11 = option,
+ .cdw12 = cdw12,
+ .save = save,
+ .uuidx = 0,
+ .cdw15 = 0,
+ .data_len = data_len,
+ .data = buf,
+ .timeout = NVME_DEFAULT_IOCTL_TIMEOUT,
+ .result = &result,
+ };
+
+ switch (option) {
+ case None:
+ err = nvme_get_features(&args_get);
+ if (!err) {
+ printf(
+ "Latency Statistics Tracking (FID 0x%X) is currently (%i).\n",
+ fid, result);
+ } else {
+ printf("Could not read feature id 0xE2.\n");
+ dev_close(dev);
+ return err;
+ }
+ break;
+ case True:
+ case False:
+ err = nvme_set_features(&args_set);
+ if (err > 0) {
+ nvme_show_status(err);
+ } else if (err < 0) {
+ perror("Enable latency tracking");
+ fprintf(stderr, "Command failed while parsing.\n");
+ } else {
+ printf("Successfully set enable bit for FID (0x%X) to %i.\n",
+ fid, option);
+ }
+ break;
+ default:
+ printf("%d not supported.\n", option);
+ return -EINVAL;
+ }
+ dev_close(dev);
+ return err;
+}
+
+static int set_lat_stats_thresholds(int argc, char **argv,
+ struct command *command, struct plugin *plugin)
+{
+ const char *desc = "Write Intel Bucket Thresholds for Latency Statistics Tracking";
+ const char *bucket_thresholds = "Bucket Threshold List, comma separated list: 0, 10, 20 ...";
+ const char *write = "Set write bucket Thresholds for latency tracking (read default)";
+
+ const __u32 nsid = 0;
+ const __u8 fid = 0xf7;
+ const __u32 cdw12 = 0x0;
+ const __u32 save = 0;
+ struct nvme_dev *dev;
+ __u32 result;
+ int err, num;
+
+ struct config {
+ bool write;
+ char *bucket_thresholds;
+ };
+
+ struct config cfg = {
+ .write = false,
+ .bucket_thresholds = "",
+ };
+
+ OPT_ARGS(opts) = {
+ OPT_FLAG("write", 'w', &cfg.write, write),
+ OPT_LIST("bucket-thresholds", 't', &cfg.bucket_thresholds,
+ bucket_thresholds),
+ OPT_END()
+ };
+
+ err = parse_and_open(&dev, argc, argv, desc, opts);
+
+ if (err)
+ return err;
+
+ /* Query maj and minor version first to figure out the amount of
+ * valid buckets a user is allowed to modify. Read or write doesn't
+ * matter
+ */
+ err = nvme_get_log_simple(dev_fd(dev), 0xc2,
+ sizeof(media_version), media_version);
+ if (err) {
+ fprintf(stderr, "Querying media version failed. ");
+ nvme_show_status(err);
+ goto close_dev;
+ }
+
+ if (media_version[0] == 1000) {
+ int thresholds[OPTANE_V1000_BUCKET_LEN] = {0};
+
+ num = argconfig_parse_comma_sep_array(cfg.bucket_thresholds,
+ thresholds,
+ sizeof(thresholds));
+ if (num == -1) {
+ fprintf(stderr, "ERROR: Bucket list is malformed\n");
+ goto close_dev;
+
+ }
+
+ struct nvme_set_features_args args = {
+ .args_size = sizeof(args),
+ .fd = dev_fd(dev),
+ .fid = fid,
+ .nsid = nsid,
+ .cdw11 = cfg.write ? 0x1 : 0x0,
+ .cdw12 = cdw12,
+ .save = save,
+ .uuidx = 0,
+ .cdw15 = 0,
+ .data_len = sizeof(thresholds),
+ .data = thresholds,
+ .timeout = NVME_DEFAULT_IOCTL_TIMEOUT,
+ .result = &result,
+ };
+ err = nvme_set_features(&args);
+
+ if (err > 0) {
+ nvme_show_status(err);
+ } else if (err < 0) {
+ perror("Enable latency tracking");
+ fprintf(stderr, "Command failed while parsing.\n");
+ }
+ } else {
+ fprintf(stderr, "Unsupported command\n");
+ }
+
+close_dev:
+ dev_close(dev);
+ return err;
+}
+