X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=stat.c;h=23657cee6a7702c3aacd81783acb1516b3960a1b;hp=a837ed90252eff2af69beb88e03f1c2c1ba6d72b;hb=87622bf5295880682bfad5ba14116cf5facbaf2c;hpb=3d2d14bcb844e72809192311369a642c5d415472 diff --git a/stat.c b/stat.c index a837ed90..23657cee 100644 --- a/stat.c +++ b/stat.c @@ -14,6 +14,8 @@ #include "lib/output_buffer.h" #include "helper_thread.h" #include "smalloc.h" +#include "zbd.h" +#include "oslib/asprintf.h" #define LOG_MSEC_SLACK 1 @@ -138,7 +140,6 @@ unsigned int calc_clat_percentiles(uint64_t *io_u_plat, unsigned long long nr, { unsigned long long sum = 0; unsigned int len, i, j = 0; - unsigned int oval_len = 0; unsigned long long *ovals = NULL; bool is_last; @@ -158,7 +159,11 @@ unsigned int calc_clat_percentiles(uint64_t *io_u_plat, unsigned long long nr, * isn't a worry. Also note that this does not work for NaN values. */ if (len > 1) - qsort((void *)plist, len, sizeof(plist[0]), double_cmp); + qsort(plist, len, sizeof(plist[0]), double_cmp); + + ovals = malloc(len * sizeof(*ovals)); + if (!ovals) + return 0; /* * Calculate bucket values, note down max and min values @@ -166,14 +171,9 @@ unsigned int calc_clat_percentiles(uint64_t *io_u_plat, unsigned long long nr, is_last = false; for (i = 0; i < FIO_IO_U_PLAT_NR && !is_last; i++) { sum += io_u_plat[i]; - while (sum >= (plist[j].u.f / 100.0 * nr)) { + while (sum >= ((long double) plist[j].u.f / 100.0 * nr)) { assert(plist[j].u.f <= 100.0); - if (j == oval_len) { - oval_len += 100; - ovals = realloc(ovals, oval_len * sizeof(*ovals)); - } - ovals[j] = plat_idx_to_val(i); if (ovals[j] < *minv) *minv = ovals[j]; @@ -188,6 +188,9 @@ unsigned int calc_clat_percentiles(uint64_t *io_u_plat, unsigned long long nr, } } + if (!is_last) + log_err("fio: error calculating latency percentiles\n"); + *output = ovals; return len; } @@ -256,8 +259,7 @@ static void show_clat_percentiles(uint64_t *io_u_plat, unsigned long long nr, } out: - if (ovals) - free(ovals); + free(ovals); } bool calc_lat(struct io_stat *is, unsigned long long *min, @@ -362,7 +364,7 @@ static void stat_calc_lat(struct thread_stat *ts, double *dst, * To keep the terse format unaltered, add all of the ns latency * buckets to the first us latency bucket */ -void stat_calc_lat_nu(struct thread_stat *ts, double *io_u_lat_u) +static void stat_calc_lat_nu(struct thread_stat *ts, double *io_u_lat_u) { unsigned long ntotal = 0, total = ddir_rw_sum(ts->total_io_u); int i; @@ -412,25 +414,36 @@ static void display_lat(const char *name, unsigned long long min, free(maxp); } +static double convert_agg_kbytes_percent(struct group_run_stats *rs, int ddir, int mean) +{ + double p_of_agg = 100.0; + if (rs && rs->agg[ddir] > 1024) { + p_of_agg = mean * 100 / (double) (rs->agg[ddir] / 1024.0); + + if (p_of_agg > 100.0) + p_of_agg = 100.0; + } + return p_of_agg; +} + static void show_ddir_status(struct group_run_stats *rs, struct thread_stat *ts, int ddir, struct buf_output *out) { - const char *str[] = { " read", "write", " trim", "sync" }; unsigned long runt; unsigned long long min, max, bw, iops; double mean, dev; - char *io_p, *bw_p, *bw_p_alt, *iops_p; + char *io_p, *bw_p, *bw_p_alt, *iops_p, *post_st = NULL; int i2p; if (ddir_sync(ddir)) { if (calc_lat(&ts->sync_stat, &min, &max, &mean, &dev)) { log_buf(out, " %s:\n", "fsync/fdatasync/sync_file_range"); - display_lat(str[ddir], min, max, mean, dev, out); + display_lat(io_ddir_name(ddir), min, max, mean, dev, out); show_clat_percentiles(ts->io_u_sync_plat, ts->sync_stat.samples, ts->percentile_list, ts->percentile_precision, - str[ddir], out); + io_ddir_name(ddir), out); } return; } @@ -450,12 +463,26 @@ static void show_ddir_status(struct group_run_stats *rs, struct thread_stat *ts, iops = (1000 * (uint64_t)ts->total_io_u[ddir]) / runt; iops_p = num2str(iops, ts->sig_figs, 1, 0, N2S_NONE); + if (ddir == DDIR_WRITE) + post_st = zbd_write_status(ts); + else if (ddir == DDIR_READ && ts->cachehit && ts->cachemiss) { + uint64_t total; + double hit; + + total = ts->cachehit + ts->cachemiss; + hit = (double) ts->cachehit / (double) total; + hit *= 100.0; + if (asprintf(&post_st, "; Cachehit=%0.2f%%", hit) < 0) + post_st = NULL; + } - log_buf(out, " %s: IOPS=%s, BW=%s (%s)(%s/%llumsec)\n", - rs->unified_rw_rep ? "mixed" : str[ddir], + log_buf(out, " %s: IOPS=%s, BW=%s (%s)(%s/%llumsec)%s\n", + rs->unified_rw_rep ? "mixed" : io_ddir_name(ddir), iops_p, bw_p, bw_p_alt, io_p, - (unsigned long long) ts->runtime[ddir]); + (unsigned long long) ts->runtime[ddir], + post_st ? : ""); + free(post_st); free(io_p); free(bw_p); free(bw_p_alt); @@ -467,21 +494,62 @@ static void show_ddir_status(struct group_run_stats *rs, struct thread_stat *ts, display_lat("clat", min, max, mean, dev, out); if (calc_lat(&ts->lat_stat[ddir], &min, &max, &mean, &dev)) display_lat(" lat", min, max, mean, dev, out); + if (calc_lat(&ts->clat_high_prio_stat[ddir], &min, &max, &mean, &dev)) { + display_lat(ts->lat_percentiles ? "high prio_lat" : "high prio_clat", + min, max, mean, dev, out); + if (calc_lat(&ts->clat_low_prio_stat[ddir], &min, &max, &mean, &dev)) + display_lat(ts->lat_percentiles ? "low prio_lat" : "low prio_clat", + min, max, mean, dev, out); + } + + if (ts->slat_percentiles && ts->slat_stat[ddir].samples > 0) + show_clat_percentiles(ts->io_u_plat[FIO_SLAT][ddir], + ts->slat_stat[ddir].samples, + ts->percentile_list, + ts->percentile_precision, "slat", out); + if (ts->clat_percentiles && ts->clat_stat[ddir].samples > 0) + show_clat_percentiles(ts->io_u_plat[FIO_CLAT][ddir], + ts->clat_stat[ddir].samples, + ts->percentile_list, + ts->percentile_precision, "clat", out); + if (ts->lat_percentiles && ts->lat_stat[ddir].samples > 0) + show_clat_percentiles(ts->io_u_plat[FIO_LAT][ddir], + ts->lat_stat[ddir].samples, + ts->percentile_list, + ts->percentile_precision, "lat", out); if (ts->clat_percentiles || ts->lat_percentiles) { - const char *name = ts->clat_percentiles ? "clat" : " lat"; + const char *name = ts->lat_percentiles ? "lat" : "clat"; + char prio_name[32]; uint64_t samples; - if (ts->clat_percentiles) - samples = ts->clat_stat[ddir].samples; - else + if (ts->lat_percentiles) samples = ts->lat_stat[ddir].samples; + else + samples = ts->clat_stat[ddir].samples; - show_clat_percentiles(ts->io_u_plat[ddir], - samples, - ts->percentile_list, - ts->percentile_precision, name, out); + /* Only print this if some high and low priority stats were collected */ + if (ts->clat_high_prio_stat[ddir].samples > 0 && + ts->clat_low_prio_stat[ddir].samples > 0) + { + sprintf(prio_name, "high prio (%.2f%%) %s", + 100. * (double) ts->clat_high_prio_stat[ddir].samples / (double) samples, + name); + show_clat_percentiles(ts->io_u_plat_high_prio[ddir], + ts->clat_high_prio_stat[ddir].samples, + ts->percentile_list, + ts->percentile_precision, prio_name, out); + + sprintf(prio_name, "low prio (%.2f%%) %s", + 100. * (double) ts->clat_low_prio_stat[ddir].samples / (double) samples, + name); + show_clat_percentiles(ts->io_u_plat_low_prio[ddir], + ts->clat_low_prio_stat[ddir].samples, + ts->percentile_list, + ts->percentile_precision, prio_name, out); + } } + if (calc_lat(&ts->bw_stat[ddir], &min, &max, &mean, &dev)) { double p_of_agg = 100.0, fkb_base = (double)rs->kb_base; const char *bw_str; @@ -495,11 +563,7 @@ static void show_ddir_status(struct group_run_stats *rs, struct thread_stat *ts, else bw_str = "kB"; - if (rs->agg[ddir]) { - p_of_agg = mean * 100 / (double) (rs->agg[ddir] / 1024); - if (p_of_agg > 100.0) - p_of_agg = 100.0; - } + p_of_agg = convert_agg_kbytes_percent(rs, ddir, mean); if (rs->unit_base == 1) { min *= 8.0; @@ -619,8 +683,8 @@ static int block_state_category(int block_state) static int compare_block_infos(const void *bs1, const void *bs2) { - uint32_t block1 = *(uint32_t *)bs1; - uint32_t block2 = *(uint32_t *)bs2; + uint64_t block1 = *(uint64_t *)bs1; + uint64_t block2 = *(uint64_t *)bs2; int state1 = BLOCK_INFO_STATE(block1); int state2 = BLOCK_INFO_STATE(block2); int bscat1 = block_state_category(state1); @@ -668,9 +732,8 @@ static int calc_block_percentiles(int nr_block_infos, uint32_t *block_infos, * isn't a worry. Also note that this does not work for NaN values. */ if (len > 1) - qsort((void *)plist, len, sizeof(plist[0]), double_cmp); + qsort(plist, len, sizeof(plist[0]), double_cmp); - nr_uninit = 0; /* Start only after the uninit entries end */ for (nr_uninit = 0; nr_uninit < nr_block_infos @@ -771,6 +834,218 @@ static void show_ss_normal(struct thread_stat *ts, struct buf_output *out) free(p2); } +static void show_agg_stats(struct disk_util_agg *agg, int terse, + struct buf_output *out) +{ + if (!agg->slavecount) + return; + + if (!terse) { + log_buf(out, ", aggrios=%llu/%llu, aggrmerge=%llu/%llu, " + "aggrticks=%llu/%llu, aggrin_queue=%llu, " + "aggrutil=%3.2f%%", + (unsigned long long) agg->ios[0] / agg->slavecount, + (unsigned long long) agg->ios[1] / agg->slavecount, + (unsigned long long) agg->merges[0] / agg->slavecount, + (unsigned long long) agg->merges[1] / agg->slavecount, + (unsigned long long) agg->ticks[0] / agg->slavecount, + (unsigned long long) agg->ticks[1] / agg->slavecount, + (unsigned long long) agg->time_in_queue / agg->slavecount, + agg->max_util.u.f); + } else { + log_buf(out, ";slaves;%llu;%llu;%llu;%llu;%llu;%llu;%llu;%3.2f%%", + (unsigned long long) agg->ios[0] / agg->slavecount, + (unsigned long long) agg->ios[1] / agg->slavecount, + (unsigned long long) agg->merges[0] / agg->slavecount, + (unsigned long long) agg->merges[1] / agg->slavecount, + (unsigned long long) agg->ticks[0] / agg->slavecount, + (unsigned long long) agg->ticks[1] / agg->slavecount, + (unsigned long long) agg->time_in_queue / agg->slavecount, + agg->max_util.u.f); + } +} + +static void aggregate_slaves_stats(struct disk_util *masterdu) +{ + struct disk_util_agg *agg = &masterdu->agg; + struct disk_util_stat *dus; + struct flist_head *entry; + struct disk_util *slavedu; + double util; + + flist_for_each(entry, &masterdu->slaves) { + slavedu = flist_entry(entry, struct disk_util, slavelist); + dus = &slavedu->dus; + agg->ios[0] += dus->s.ios[0]; + agg->ios[1] += dus->s.ios[1]; + agg->merges[0] += dus->s.merges[0]; + agg->merges[1] += dus->s.merges[1]; + agg->sectors[0] += dus->s.sectors[0]; + agg->sectors[1] += dus->s.sectors[1]; + agg->ticks[0] += dus->s.ticks[0]; + agg->ticks[1] += dus->s.ticks[1]; + agg->time_in_queue += dus->s.time_in_queue; + agg->slavecount++; + + util = (double) (100 * dus->s.io_ticks / (double) slavedu->dus.s.msec); + /* System utilization is the utilization of the + * component with the highest utilization. + */ + if (util > agg->max_util.u.f) + agg->max_util.u.f = util; + + } + + if (agg->max_util.u.f > 100.0) + agg->max_util.u.f = 100.0; +} + +void print_disk_util(struct disk_util_stat *dus, struct disk_util_agg *agg, + int terse, struct buf_output *out) +{ + double util = 0; + + if (dus->s.msec) + util = (double) 100 * dus->s.io_ticks / (double) dus->s.msec; + if (util > 100.0) + util = 100.0; + + if (!terse) { + if (agg->slavecount) + log_buf(out, " "); + + log_buf(out, " %s: ios=%llu/%llu, merge=%llu/%llu, " + "ticks=%llu/%llu, in_queue=%llu, util=%3.2f%%", + dus->name, + (unsigned long long) dus->s.ios[0], + (unsigned long long) dus->s.ios[1], + (unsigned long long) dus->s.merges[0], + (unsigned long long) dus->s.merges[1], + (unsigned long long) dus->s.ticks[0], + (unsigned long long) dus->s.ticks[1], + (unsigned long long) dus->s.time_in_queue, + util); + } else { + log_buf(out, ";%s;%llu;%llu;%llu;%llu;%llu;%llu;%llu;%3.2f%%", + dus->name, + (unsigned long long) dus->s.ios[0], + (unsigned long long) dus->s.ios[1], + (unsigned long long) dus->s.merges[0], + (unsigned long long) dus->s.merges[1], + (unsigned long long) dus->s.ticks[0], + (unsigned long long) dus->s.ticks[1], + (unsigned long long) dus->s.time_in_queue, + util); + } + + /* + * If the device has slaves, aggregate the stats for + * those slave devices also. + */ + show_agg_stats(agg, terse, out); + + if (!terse) + log_buf(out, "\n"); +} + +void json_array_add_disk_util(struct disk_util_stat *dus, + struct disk_util_agg *agg, struct json_array *array) +{ + struct json_object *obj; + double util = 0; + + if (dus->s.msec) + util = (double) 100 * dus->s.io_ticks / (double) dus->s.msec; + if (util > 100.0) + util = 100.0; + + obj = json_create_object(); + json_array_add_value_object(array, obj); + + json_object_add_value_string(obj, "name", (const char *)dus->name); + json_object_add_value_int(obj, "read_ios", dus->s.ios[0]); + json_object_add_value_int(obj, "write_ios", dus->s.ios[1]); + json_object_add_value_int(obj, "read_merges", dus->s.merges[0]); + json_object_add_value_int(obj, "write_merges", dus->s.merges[1]); + json_object_add_value_int(obj, "read_ticks", dus->s.ticks[0]); + json_object_add_value_int(obj, "write_ticks", dus->s.ticks[1]); + json_object_add_value_int(obj, "in_queue", dus->s.time_in_queue); + json_object_add_value_float(obj, "util", util); + + /* + * If the device has slaves, aggregate the stats for + * those slave devices also. + */ + if (!agg->slavecount) + return; + json_object_add_value_int(obj, "aggr_read_ios", + agg->ios[0] / agg->slavecount); + json_object_add_value_int(obj, "aggr_write_ios", + agg->ios[1] / agg->slavecount); + json_object_add_value_int(obj, "aggr_read_merges", + agg->merges[0] / agg->slavecount); + json_object_add_value_int(obj, "aggr_write_merge", + agg->merges[1] / agg->slavecount); + json_object_add_value_int(obj, "aggr_read_ticks", + agg->ticks[0] / agg->slavecount); + json_object_add_value_int(obj, "aggr_write_ticks", + agg->ticks[1] / agg->slavecount); + json_object_add_value_int(obj, "aggr_in_queue", + agg->time_in_queue / agg->slavecount); + json_object_add_value_float(obj, "aggr_util", agg->max_util.u.f); +} + +static void json_object_add_disk_utils(struct json_object *obj, + struct flist_head *head) +{ + struct json_array *array = json_create_array(); + struct flist_head *entry; + struct disk_util *du; + + json_object_add_value_array(obj, "disk_util", array); + + flist_for_each(entry, head) { + du = flist_entry(entry, struct disk_util, list); + + aggregate_slaves_stats(du); + json_array_add_disk_util(&du->dus, &du->agg, array); + } +} + +void show_disk_util(int terse, struct json_object *parent, + struct buf_output *out) +{ + struct flist_head *entry; + struct disk_util *du; + bool do_json; + + if (!is_running_backend()) + return; + + if (flist_empty(&disk_list)) { + return; + } + + if ((output_format & FIO_OUTPUT_JSON) && parent) + do_json = true; + else + do_json = false; + + if (!terse && !do_json) + log_buf(out, "\nDisk stats (read/write):\n"); + + if (do_json) + json_object_add_disk_utils(parent, &disk_list); + else if (output_format & ~(FIO_OUTPUT_JSON | FIO_OUTPUT_JSON_PLUS)) { + flist_for_each(entry, &disk_list) { + du = flist_entry(entry, struct disk_util, list); + + aggregate_slaves_stats(du); + print_disk_util(&du->dus, &du->agg, terse, out); + } + } +} + static void show_thread_status_normal(struct thread_stat *ts, struct group_run_stats *rs, struct buf_output *out) @@ -783,7 +1058,7 @@ static void show_thread_status_normal(struct thread_stat *ts, if (!ddir_rw_sum(ts->io_bytes) && !ddir_rw_sum(ts->total_io_u)) return; - + memset(time_buf, 0, sizeof(time_buf)); time(&time_p); @@ -920,12 +1195,17 @@ static void show_ddir_status_terse(struct thread_stat *ts, else log_buf(out, ";%llu;%llu;%f;%f", 0ULL, 0ULL, 0.0, 0.0); - if (ts->clat_percentiles || ts->lat_percentiles) { - len = calc_clat_percentiles(ts->io_u_plat[ddir], + if (ts->lat_percentiles) + len = calc_clat_percentiles(ts->io_u_plat[FIO_LAT][ddir], + ts->lat_stat[ddir].samples, + ts->percentile_list, &ovals, &maxv, + &minv); + else if (ts->clat_percentiles) + len = calc_clat_percentiles(ts->io_u_plat[FIO_CLAT][ddir], ts->clat_stat[ddir].samples, ts->percentile_list, &ovals, &maxv, &minv); - } else + else len = 0; for (i = 0; i < FIO_IO_U_LIST_MAX_LEN; i++) { @@ -941,8 +1221,7 @@ static void show_ddir_status_terse(struct thread_stat *ts, else log_buf(out, ";%llu;%llu;%f;%f", 0ULL, 0ULL, 0.0, 0.0); - if (ovals) - free(ovals); + free(ovals); bw_stat = calc_lat(&ts->bw_stat[ddir], &min, &max, &mean, &dev); if (bw_stat) { @@ -972,18 +1251,63 @@ static void show_ddir_status_terse(struct thread_stat *ts, } } +static struct json_object *add_ddir_lat_json(struct thread_stat *ts, uint32_t percentiles, + struct io_stat *lat_stat, uint64_t *io_u_plat) +{ + char buf[120]; + double mean, dev; + unsigned int i, len; + struct json_object *lat_object, *percentile_object, *clat_bins_object; + unsigned long long min, max, maxv, minv, *ovals = NULL; + + if (!calc_lat(lat_stat, &min, &max, &mean, &dev)) { + min = max = 0; + mean = dev = 0.0; + } + lat_object = json_create_object(); + json_object_add_value_int(lat_object, "min", min); + json_object_add_value_int(lat_object, "max", max); + json_object_add_value_float(lat_object, "mean", mean); + json_object_add_value_float(lat_object, "stddev", dev); + json_object_add_value_int(lat_object, "N", lat_stat->samples); + + if (percentiles && lat_stat->samples) { + len = calc_clat_percentiles(io_u_plat, lat_stat->samples, + ts->percentile_list, &ovals, &maxv, &minv); + + if (len > FIO_IO_U_LIST_MAX_LEN) + len = FIO_IO_U_LIST_MAX_LEN; + + percentile_object = json_create_object(); + json_object_add_value_object(lat_object, "percentile", percentile_object); + for (i = 0; i < len; i++) { + snprintf(buf, sizeof(buf), "%f", ts->percentile_list[i].u.f); + json_object_add_value_int(percentile_object, buf, ovals[i]); + } + free(ovals); + + if (output_format & FIO_OUTPUT_JSON_PLUS) { + clat_bins_object = json_create_object(); + json_object_add_value_object(lat_object, "bins", clat_bins_object); + + for(i = 0; i < FIO_IO_U_PLAT_NR; i++) + if (io_u_plat[i]) { + snprintf(buf, sizeof(buf), "%llu", plat_idx_to_val(i)); + json_object_add_value_int(clat_bins_object, buf, io_u_plat[i]); + } + } + } + + return lat_object; +} + static void add_ddir_status_json(struct thread_stat *ts, struct group_run_stats *rs, int ddir, struct json_object *parent) { - unsigned long long min, max, minv, maxv; + unsigned long long min, max; unsigned long long bw_bytes, bw; - unsigned long long *ovals = NULL; double mean, dev, iops; - unsigned int len; - int i; - const char *ddirname[] = { "read", "write", "trim", "sync" }; - struct json_object *dir_object, *tmp_object, *percentile_object, *clat_bins_object = NULL; - char buf[120]; + struct json_object *dir_object, *tmp_object; double p_of_agg = 100.0; assert(ddir_rw(ddir) || ddir_sync(ddir)); @@ -993,7 +1317,7 @@ static void add_ddir_status_json(struct thread_stat *ts, dir_object = json_create_object(); json_object_add_value_object(parent, - ts->unified_rw_rep ? "mixed" : ddirname[ddir], dir_object); + ts->unified_rw_rep ? "mixed" : io_ddir_name(ddir), dir_object); if (ddir_rw(ddir)) { bw_bytes = 0; @@ -1017,116 +1341,55 @@ static void add_ddir_status_json(struct thread_stat *ts, json_object_add_value_int(dir_object, "short_ios", ts->short_io_u[ddir]); json_object_add_value_int(dir_object, "drop_ios", ts->drop_io_u[ddir]); - if (!calc_lat(&ts->slat_stat[ddir], &min, &max, &mean, &dev)) { - min = max = 0; - mean = dev = 0.0; - } - tmp_object = json_create_object(); + tmp_object = add_ddir_lat_json(ts, ts->slat_percentiles, + &ts->slat_stat[ddir], ts->io_u_plat[FIO_SLAT][ddir]); json_object_add_value_object(dir_object, "slat_ns", tmp_object); - json_object_add_value_int(tmp_object, "min", min); - json_object_add_value_int(tmp_object, "max", max); - json_object_add_value_float(tmp_object, "mean", mean); - json_object_add_value_float(tmp_object, "stddev", dev); - - if (!calc_lat(&ts->clat_stat[ddir], &min, &max, &mean, &dev)) { - min = max = 0; - mean = dev = 0.0; - } - tmp_object = json_create_object(); + + tmp_object = add_ddir_lat_json(ts, ts->clat_percentiles, + &ts->clat_stat[ddir], ts->io_u_plat[FIO_CLAT][ddir]); json_object_add_value_object(dir_object, "clat_ns", tmp_object); - json_object_add_value_int(tmp_object, "min", min); - json_object_add_value_int(tmp_object, "max", max); - json_object_add_value_float(tmp_object, "mean", mean); - json_object_add_value_float(tmp_object, "stddev", dev); - } else { - if (!calc_lat(&ts->sync_stat, &min, &max, &mean, &dev)) { - min = max = 0; - mean = dev = 0.0; - } - tmp_object = json_create_object(); + tmp_object = add_ddir_lat_json(ts, ts->lat_percentiles, + &ts->lat_stat[ddir], ts->io_u_plat[FIO_LAT][ddir]); json_object_add_value_object(dir_object, "lat_ns", tmp_object); + } else { json_object_add_value_int(dir_object, "total_ios", ts->total_io_u[DDIR_SYNC]); - json_object_add_value_int(tmp_object, "min", min); - json_object_add_value_int(tmp_object, "max", max); - json_object_add_value_float(tmp_object, "mean", mean); - json_object_add_value_float(tmp_object, "stddev", dev); + tmp_object = add_ddir_lat_json(ts, ts->lat_percentiles | ts->clat_percentiles, + &ts->sync_stat, ts->io_u_sync_plat); + json_object_add_value_object(dir_object, "lat_ns", tmp_object); } - if (ts->clat_percentiles || ts->lat_percentiles) { - if (ddir_rw(ddir)) { - len = calc_clat_percentiles(ts->io_u_plat[ddir], - ts->clat_stat[ddir].samples, - ts->percentile_list, &ovals, &maxv, - &minv); - } else { - len = calc_clat_percentiles(ts->io_u_sync_plat, - ts->sync_stat.samples, - ts->percentile_list, &ovals, &maxv, - &minv); - } - - if (len > FIO_IO_U_LIST_MAX_LEN) - len = FIO_IO_U_LIST_MAX_LEN; - } else - len = 0; - - percentile_object = json_create_object(); - json_object_add_value_object(tmp_object, "percentile", percentile_object); - for (i = 0; i < len; i++) { - snprintf(buf, sizeof(buf), "%f", ts->percentile_list[i].u.f); - json_object_add_value_int(percentile_object, (const char *)buf, ovals[i]); - } + if (!ddir_rw(ddir)) + return; - if (output_format & FIO_OUTPUT_JSON_PLUS) { - clat_bins_object = json_create_object(); - if (ts->clat_percentiles) - json_object_add_value_object(tmp_object, "bins", clat_bins_object); + /* Only print PRIO latencies if some high priority samples were gathered */ + if (ts->clat_high_prio_stat[ddir].samples > 0) { + const char *high, *low; - for(i = 0; i < FIO_IO_U_PLAT_NR; i++) { - if (ddir_rw(ddir)) { - if (ts->io_u_plat[ddir][i]) { - snprintf(buf, sizeof(buf), "%llu", plat_idx_to_val(i)); - json_object_add_value_int(clat_bins_object, (const char *)buf, ts->io_u_plat[ddir][i]); - } - } else { - if (ts->io_u_sync_plat[i]) { - snprintf(buf, sizeof(buf), "%llu", plat_idx_to_val(i)); - json_object_add_value_int(clat_bins_object, (const char *)buf, ts->io_u_sync_plat[i]); - } - } + if (ts->lat_percentiles) { + high = "lat_high_prio"; + low = "lat_low_prio"; + } else { + high = "clat_high_prio"; + low = "clat_low_prio"; } - } - if (!ddir_rw(ddir)) - return; + tmp_object = add_ddir_lat_json(ts, ts->clat_percentiles | ts->lat_percentiles, + &ts->clat_high_prio_stat[ddir], ts->io_u_plat_high_prio[ddir]); + json_object_add_value_object(dir_object, high, tmp_object); - if (!calc_lat(&ts->lat_stat[ddir], &min, &max, &mean, &dev)) { - min = max = 0; - mean = dev = 0.0; + tmp_object = add_ddir_lat_json(ts, ts->clat_percentiles | ts->lat_percentiles, + &ts->clat_low_prio_stat[ddir], ts->io_u_plat_low_prio[ddir]); + json_object_add_value_object(dir_object, low, tmp_object); } - tmp_object = json_create_object(); - json_object_add_value_object(dir_object, "lat_ns", tmp_object); - json_object_add_value_int(tmp_object, "min", min); - json_object_add_value_int(tmp_object, "max", max); - json_object_add_value_float(tmp_object, "mean", mean); - json_object_add_value_float(tmp_object, "stddev", dev); - if (output_format & FIO_OUTPUT_JSON_PLUS && ts->lat_percentiles) - json_object_add_value_object(tmp_object, "bins", clat_bins_object); - - if (ovals) - free(ovals); if (calc_lat(&ts->bw_stat[ddir], &min, &max, &mean, &dev)) { - if (rs->agg[ddir]) { - p_of_agg = mean * 100 / (double) (rs->agg[ddir] / 1024); - if (p_of_agg > 100.0) - p_of_agg = 100.0; - } + p_of_agg = convert_agg_kbytes_percent(rs, ddir, mean); } else { min = max = 0; p_of_agg = mean = dev = 0.0; } + json_object_add_value_int(dir_object, "bw_min", min); json_object_add_value_int(dir_object, "bw_max", max); json_object_add_value_float(dir_object, "bw_agg", p_of_agg); @@ -1145,6 +1408,16 @@ static void add_ddir_status_json(struct thread_stat *ts, json_object_add_value_float(dir_object, "iops_stddev", dev); json_object_add_value_int(dir_object, "iops_samples", (&ts->iops_stat[ddir])->samples); + + if (ts->cachehit + ts->cachemiss) { + uint64_t total; + double hit; + + total = ts->cachehit + ts->cachemiss; + hit = (double) ts->cachehit / (double) total; + hit *= 100.0; + json_object_add_value_float(dir_object, "cachehit", hit); + } } static void show_thread_status_terse_all(struct thread_stat *ts, @@ -1206,18 +1479,19 @@ static void show_thread_status_terse_all(struct thread_stat *ts, log_buf(out, ";%3.2f%%", io_u_lat_m[i]); /* disk util stats, if any */ - if (ver >= 3) + if (ver >= 3 && is_running_backend()) show_disk_util(1, NULL, out); /* Additional output if continue_on_error set - default off*/ if (ts->continue_on_error) log_buf(out, ";%llu;%d", (unsigned long long) ts->total_err_count, ts->first_error); - if (ver == 2) - log_buf(out, "\n"); /* Additional output if description is set */ - if (strlen(ts->description)) + if (strlen(ts->description)) { + if (ver == 2) + log_buf(out, "\n"); log_buf(out, ";%s", ts->description); + } log_buf(out, "\n"); } @@ -1236,12 +1510,8 @@ static void json_add_job_opts(struct json_object *root, const char *name, json_object_add_value_object(root, name, dir_object); flist_for_each(entry, opt_list) { - const char *pos = ""; - p = flist_entry(entry, struct print_option, list); - if (p->value) - pos = p->value; - json_object_add_value_string(dir_object, p->name, pos); + json_object_add_value_string(dir_object, p->name, p->value); } } @@ -1289,19 +1559,15 @@ static struct json_object *show_thread_status_json(struct thread_stat *ts, usr_cpu = 0; sys_cpu = 0; } + json_object_add_value_int(root, "job_runtime", ts->total_run_time); json_object_add_value_float(root, "usr_cpu", usr_cpu); json_object_add_value_float(root, "sys_cpu", sys_cpu); json_object_add_value_int(root, "ctx", ts->ctx); json_object_add_value_int(root, "majf", ts->majf); json_object_add_value_int(root, "minf", ts->minf); - - /* Calc % distribution of IO depths, usecond, msecond latency */ + /* Calc % distribution of IO depths */ stat_calc_dist(ts->io_u_map, ddir_rw_sum(ts->total_io_u), io_u_dist); - stat_calc_lat_n(ts, io_u_lat_n); - stat_calc_lat_u(ts, io_u_lat_u); - stat_calc_lat_m(ts, io_u_lat_m); - tmp = json_create_object(); json_object_add_value_object(root, "iodepth_level", tmp); /* Only show fixed 7 I/O depth levels*/ @@ -1314,6 +1580,44 @@ static struct json_object *show_thread_status_json(struct thread_stat *ts, json_object_add_value_float(tmp, (const char *)name, io_u_dist[i]); } + /* Calc % distribution of submit IO depths */ + stat_calc_dist(ts->io_u_submit, ts->total_submit, io_u_dist); + tmp = json_create_object(); + json_object_add_value_object(root, "iodepth_submit", tmp); + /* Only show fixed 7 I/O depth levels*/ + for (i = 0; i < 7; i++) { + char name[20]; + if (i == 0) + snprintf(name, 20, "0"); + else if (i < 6) + snprintf(name, 20, "%d", 1 << (i+1)); + else + snprintf(name, 20, ">=%d", 1 << i); + json_object_add_value_float(tmp, (const char *)name, io_u_dist[i]); + } + + /* Calc % distribution of completion IO depths */ + stat_calc_dist(ts->io_u_complete, ts->total_complete, io_u_dist); + tmp = json_create_object(); + json_object_add_value_object(root, "iodepth_complete", tmp); + /* Only show fixed 7 I/O depth levels*/ + for (i = 0; i < 7; i++) { + char name[20]; + if (i == 0) + snprintf(name, 20, "0"); + else if (i < 6) + snprintf(name, 20, "%d", 1 << (i+1)); + else + snprintf(name, 20, ">=%d", 1 << i); + json_object_add_value_float(tmp, (const char *)name, io_u_dist[i]); + } + + /* Calc % distribution of nsecond, usecond, msecond latency */ + stat_calc_dist(ts->io_u_map, ddir_rw_sum(ts->total_io_u), io_u_dist); + stat_calc_lat_n(ts, io_u_lat_n); + stat_calc_lat_u(ts, io_u_lat_u); + stat_calc_lat_m(ts, io_u_lat_m); + /* Nanosecond latency */ tmp = json_create_object(); json_object_add_value_object(root, "latency_ns", tmp); @@ -1381,7 +1685,7 @@ static struct json_object *show_thread_status_json(struct thread_stat *ts, snprintf(buf, sizeof(buf), "%f", ts->percentile_list[i].u.f); json_object_add_value_int(percentile_object, - (const char *)buf, + buf, percentiles[i]); } @@ -1399,7 +1703,7 @@ static struct json_object *show_thread_status_json(struct thread_stat *ts, if (ts->ss_dur) { struct json_object *data; struct json_array *iops, *bw; - int i, j, k; + int j, k, l; char ss_buf[64]; snprintf(ss_buf, sizeof(ss_buf), "%s%s:%f%s", @@ -1435,8 +1739,8 @@ static struct json_object *show_thread_status_json(struct thread_stat *ts, j = ts->ss_head; else j = ts->ss_head == 0 ? ts->ss_dur - 1 : ts->ss_head - 1; - for (i = 0; i < ts->ss_dur; i++) { - k = (j + i) % ts->ss_dur; + for (l = 0; l < ts->ss_dur; l++) { + k = (j + l) % ts->ss_dur; json_array_add_value_int(bw, ts->ss_bw_data[k]); json_array_add_value_int(iops, ts->ss_iops_data[k]); } @@ -1476,13 +1780,10 @@ struct json_object *show_thread_status(struct thread_stat *ts, return ret; } -static void sum_stat(struct io_stat *dst, struct io_stat *src, bool first) +static void __sum_stat(struct io_stat *dst, struct io_stat *src, bool first) { double mean, S; - if (src->samples == 0) - return; - dst->min_val = min(dst->min_val, src->min_val); dst->max_val = max(dst->max_val, src->max_val); @@ -1509,6 +1810,39 @@ static void sum_stat(struct io_stat *dst, struct io_stat *src, bool first) dst->samples += src->samples; dst->mean.u.f = mean; dst->S.u.f = S; + +} + +/* + * We sum two kinds of stats - one that is time based, in which case we + * apply the proper summing technique, and then one that is iops/bw + * numbers. For group_reporting, we should just add those up, not make + * them the mean of everything. + */ +static void sum_stat(struct io_stat *dst, struct io_stat *src, bool first, + bool pure_sum) +{ + if (src->samples == 0) + return; + + if (!pure_sum) { + __sum_stat(dst, src, first); + return; + } + + if (first) { + dst->min_val = src->min_val; + dst->max_val = src->max_val; + dst->samples = src->samples; + dst->mean.u.f = src->mean.u.f; + dst->S.u.f = src->S.u.f; + } else { + dst->min_val += src->min_val; + dst->max_val += src->max_val; + dst->samples += src->samples; + dst->mean.u.f += src->mean.u.f; + dst->S.u.f += src->S.u.f; + } } void sum_group_stats(struct group_run_stats *dst, struct group_run_stats *src) @@ -1540,26 +1874,30 @@ void sum_group_stats(struct group_run_stats *dst, struct group_run_stats *src) void sum_thread_stats(struct thread_stat *dst, struct thread_stat *src, bool first) { - int l, k; + int k, l, m; for (l = 0; l < DDIR_RWDIR_CNT; l++) { if (!dst->unified_rw_rep) { - sum_stat(&dst->clat_stat[l], &src->clat_stat[l], first); - sum_stat(&dst->slat_stat[l], &src->slat_stat[l], first); - sum_stat(&dst->lat_stat[l], &src->lat_stat[l], first); - sum_stat(&dst->bw_stat[l], &src->bw_stat[l], first); - sum_stat(&dst->iops_stat[l], &src->iops_stat[l], first); + sum_stat(&dst->clat_stat[l], &src->clat_stat[l], first, false); + sum_stat(&dst->clat_high_prio_stat[l], &src->clat_high_prio_stat[l], first, false); + sum_stat(&dst->clat_low_prio_stat[l], &src->clat_low_prio_stat[l], first, false); + sum_stat(&dst->slat_stat[l], &src->slat_stat[l], first, false); + sum_stat(&dst->lat_stat[l], &src->lat_stat[l], first, false); + sum_stat(&dst->bw_stat[l], &src->bw_stat[l], first, true); + sum_stat(&dst->iops_stat[l], &src->iops_stat[l], first, true); dst->io_bytes[l] += src->io_bytes[l]; if (dst->runtime[l] < src->runtime[l]) dst->runtime[l] = src->runtime[l]; } else { - sum_stat(&dst->clat_stat[0], &src->clat_stat[l], first); - sum_stat(&dst->slat_stat[0], &src->slat_stat[l], first); - sum_stat(&dst->lat_stat[0], &src->lat_stat[l], first); - sum_stat(&dst->bw_stat[0], &src->bw_stat[l], first); - sum_stat(&dst->iops_stat[0], &src->iops_stat[l], first); + sum_stat(&dst->clat_stat[0], &src->clat_stat[l], first, false); + sum_stat(&dst->clat_high_prio_stat[0], &src->clat_high_prio_stat[l], first, false); + sum_stat(&dst->clat_low_prio_stat[0], &src->clat_low_prio_stat[l], first, false); + sum_stat(&dst->slat_stat[0], &src->slat_stat[l], first, false); + sum_stat(&dst->lat_stat[0], &src->lat_stat[l], first, false); + sum_stat(&dst->bw_stat[0], &src->bw_stat[l], first, true); + sum_stat(&dst->iops_stat[0], &src->iops_stat[l], first, true); dst->io_bytes[0] += src->io_bytes[l]; @@ -1574,7 +1912,7 @@ void sum_thread_stats(struct thread_stat *dst, struct thread_stat *src, } } - sum_stat(&dst->sync_stat, &src->sync_stat, first); + sum_stat(&dst->sync_stat, &src->sync_stat, first, false); dst->usr_time += src->usr_time; dst->sys_time += src->sys_time; dst->ctx += src->ctx; @@ -1586,13 +1924,13 @@ void sum_thread_stats(struct thread_stat *dst, struct thread_stat *src, dst->io_u_submit[k] += src->io_u_submit[k]; dst->io_u_complete[k] += src->io_u_complete[k]; } - for (k = 0; k < FIO_IO_U_LAT_N_NR; k++) { + + for (k = 0; k < FIO_IO_U_LAT_N_NR; k++) dst->io_u_lat_n[k] += src->io_u_lat_n[k]; + for (k = 0; k < FIO_IO_U_LAT_U_NR; k++) dst->io_u_lat_u[k] += src->io_u_lat_u[k]; + for (k = 0; k < FIO_IO_U_LAT_M_NR; k++) dst->io_u_lat_m[k] += src->io_u_lat_m[k]; - } - for (k = 0; k < FIO_IO_U_PLAT_NR; k++) - dst->io_u_sync_plat[k] += src->io_u_sync_plat[k]; for (k = 0; k < DDIR_RWDIR_CNT; k++) { if (!dst->unified_rw_rep) { @@ -1608,20 +1946,36 @@ void sum_thread_stats(struct thread_stat *dst, struct thread_stat *src, dst->total_io_u[DDIR_SYNC] += src->total_io_u[DDIR_SYNC]; - for (k = 0; k < DDIR_RWDIR_CNT; k++) { - int m; + for (k = 0; k < FIO_LAT_CNT; k++) + for (l = 0; l < DDIR_RWDIR_CNT; l++) + for (m = 0; m < FIO_IO_U_PLAT_NR; m++) + if (!dst->unified_rw_rep) + dst->io_u_plat[k][l][m] += src->io_u_plat[k][l][m]; + else + dst->io_u_plat[k][0][m] += src->io_u_plat[k][l][m]; + + for (k = 0; k < FIO_IO_U_PLAT_NR; k++) + dst->io_u_sync_plat[k] += src->io_u_sync_plat[k]; + for (k = 0; k < DDIR_RWDIR_CNT; k++) { for (m = 0; m < FIO_IO_U_PLAT_NR; m++) { - if (!dst->unified_rw_rep) - dst->io_u_plat[k][m] += src->io_u_plat[k][m]; - else - dst->io_u_plat[0][m] += src->io_u_plat[k][m]; + if (!dst->unified_rw_rep) { + dst->io_u_plat_high_prio[k][m] += src->io_u_plat_high_prio[k][m]; + dst->io_u_plat_low_prio[k][m] += src->io_u_plat_low_prio[k][m]; + } else { + dst->io_u_plat_high_prio[0][m] += src->io_u_plat_high_prio[k][m]; + dst->io_u_plat_low_prio[0][m] += src->io_u_plat_low_prio[k][m]; + } + } } dst->total_run_time += src->total_run_time; dst->total_submit += src->total_submit; dst->total_complete += src->total_complete; + dst->nr_zone_resets += src->nr_zone_resets; + dst->cachehit += src->cachehit; + dst->cachemiss += src->cachemiss; } void init_group_run_stat(struct group_run_stats *gs) @@ -1645,6 +1999,8 @@ void init_thread_stat(struct thread_stat *ts) ts->slat_stat[j].min_val = -1UL; ts->bw_stat[j].min_val = -1UL; ts->iops_stat[j].min_val = -1UL; + ts->clat_high_prio_stat[j].min_val = -1UL; + ts->clat_low_prio_stat[j].min_val = -1UL; } ts->sync_stat.min_val = -1UL; ts->groupid = -1; @@ -1714,6 +2070,7 @@ void __show_run_stats(void) ts->clat_percentiles = td->o.clat_percentiles; ts->lat_percentiles = td->o.lat_percentiles; + ts->slat_percentiles = td->o.slat_percentiles; ts->percentile_precision = td->o.percentile_precision; memcpy(ts->percentile_list, td->o.percentile_list, sizeof(td->o.percentile_list)); opt_lists[j] = &td->opt_list; @@ -1725,10 +2082,11 @@ void __show_run_stats(void) /* * These are per-group shared already */ - strncpy(ts->name, td->o.name, FIO_JOBNAME_SIZE - 1); + snprintf(ts->name, sizeof(ts->name), "%s", td->o.name); if (td->o.description) - strncpy(ts->description, td->o.description, - FIO_JOBDESC_SIZE - 1); + snprintf(ts->description, + sizeof(ts->description), "%s", + td->o.description); else memset(ts->description, 0, FIO_JOBDESC_SIZE); @@ -1765,12 +2123,12 @@ void __show_run_stats(void) if (!td->error && td->o.continue_on_error && td->first_error) { ts->error = td->first_error; - ts->verror[sizeof(ts->verror) - 1] = '\0'; - strncpy(ts->verror, td->verror, sizeof(ts->verror) - 1); + snprintf(ts->verror, sizeof(ts->verror), "%s", + td->verror); } else if (td->error) { ts->error = td->error; - ts->verror[sizeof(ts->verror) - 1] = '\0'; - strncpy(ts->verror, td->verror, sizeof(ts->verror) - 1); + snprintf(ts->verror, sizeof(ts->verror), "%s", + td->verror); } } @@ -1935,19 +2293,14 @@ void __show_run_stats(void) buf_output_free(out); } + fio_idle_prof_cleanup(); + log_info_flush(); free(runstats); free(threadstats); free(opt_lists); } -void show_run_stats(void) -{ - fio_sem_down(stat_sem); - __show_run_stats(); - fio_sem_up(stat_sem); -} - void __show_running_run_stats(void) { struct thread_data *td; @@ -2001,8 +2354,6 @@ void __show_running_run_stats(void) fio_sem_up(stat_sem); } -static bool status_interval_init; -static struct timespec status_time; static bool status_file_disabled; #define FIO_STATUS_FILE "fio-dump-status" @@ -2024,6 +2375,9 @@ static int check_status_file(void) } if (temp_dir == NULL) temp_dir = "/tmp"; +#ifdef __COVERITY__ + __coverity_tainted_data_sanitize__(temp_dir); +#endif snprintf(fio_status_file_path, sizeof(fio_status_file_path), "%s/%s", temp_dir, FIO_STATUS_FILE); @@ -2042,16 +2396,6 @@ static int check_status_file(void) void check_for_running_stats(void) { - if (status_interval) { - if (!status_interval_init) { - fio_gettime(&status_time, NULL); - status_interval_init = true; - } else if (mtime_since_now(&status_time) >= status_interval) { - show_running_run_stats(); - fio_gettime(&status_time, NULL); - return; - } - } if (check_status_file()) { show_running_run_stats(); return; @@ -2213,18 +2557,20 @@ static struct io_logs *get_cur_log(struct io_log *iolog) * submissions, flag 'td' as needing a log regrow and we'll take * care of it on the submission side. */ - if (iolog->td->o.io_submit_mode == IO_MODE_OFFLOAD || + if ((iolog->td && iolog->td->o.io_submit_mode == IO_MODE_OFFLOAD) || !per_unit_log(iolog)) return regrow_log(iolog); - iolog->td->flags |= TD_F_REGROW_LOGS; - assert(iolog->pending->nr_samples < iolog->pending->max_samples); + if (iolog->td) + iolog->td->flags |= TD_F_REGROW_LOGS; + if (iolog->pending) + assert(iolog->pending->nr_samples < iolog->pending->max_samples); return iolog->pending; } static void __add_log_sample(struct io_log *iolog, union io_sample_data data, - enum fio_ddir ddir, unsigned int bs, - unsigned long t, uint64_t offset) + enum fio_ddir ddir, unsigned long long bs, + unsigned long t, uint64_t offset, uint8_t priority_bit) { struct io_logs *cur_log; @@ -2243,6 +2589,7 @@ static void __add_log_sample(struct io_log *iolog, union io_sample_data data, s->time = t + (iolog->td ? iolog->td->unix_epoch : 0); io_sample_set_ddir(iolog, s, ddir); s->bs = bs; + s->priority_bit = priority_bit; if (iolog->log_offset) { struct io_sample_offset *so = (void *) s; @@ -2259,16 +2606,19 @@ static void __add_log_sample(struct io_log *iolog, union io_sample_data data, static inline void reset_io_stat(struct io_stat *ios) { - ios->max_val = ios->min_val = ios->samples = 0; + ios->min_val = -1ULL; + ios->max_val = ios->samples = 0; ios->mean.u.f = ios->S.u.f = 0; } void reset_io_stats(struct thread_data *td) { struct thread_stat *ts = &td->ts; - int i, j; + int i, j, k; for (i = 0; i < DDIR_RWDIR_CNT; i++) { + reset_io_stat(&ts->clat_high_prio_stat[i]); + reset_io_stat(&ts->clat_low_prio_stat[i]); reset_io_stat(&ts->clat_stat[i]); reset_io_stat(&ts->slat_stat[i]); reset_io_stat(&ts->lat_stat[i]); @@ -2282,12 +2632,18 @@ void reset_io_stats(struct thread_data *td) ts->drop_io_u[i] = 0; for (j = 0; j < FIO_IO_U_PLAT_NR; j++) { - ts->io_u_plat[i][j] = 0; + ts->io_u_plat_high_prio[i][j] = 0; + ts->io_u_plat_low_prio[i][j] = 0; if (!i) ts->io_u_sync_plat[j] = 0; } } + for (i = 0; i < FIO_LAT_CNT; i++) + for (j = 0; j < DDIR_RWDIR_CNT; j++) + for (k = 0; k < FIO_IO_U_PLAT_NR; k++) + ts->io_u_plat[i][j][k] = 0; + ts->total_io_u[DDIR_SYNC] = 0; for (i = 0; i < FIO_IO_U_MAP_NR; i++) { @@ -2305,10 +2661,12 @@ void reset_io_stats(struct thread_data *td) ts->total_submit = 0; ts->total_complete = 0; + ts->nr_zone_resets = 0; + ts->cachehit = ts->cachemiss = 0; } static void __add_stat_to_log(struct io_log *iolog, enum fio_ddir ddir, - unsigned long elapsed, bool log_max) + unsigned long elapsed, bool log_max, uint8_t priority_bit) { /* * Note an entry in the log. Use the mean from the logged samples, @@ -2323,26 +2681,26 @@ static void __add_stat_to_log(struct io_log *iolog, enum fio_ddir ddir, else data.val = iolog->avg_window[ddir].mean.u.f + 0.50; - __add_log_sample(iolog, data, ddir, 0, elapsed, 0); + __add_log_sample(iolog, data, ddir, 0, elapsed, 0, priority_bit); } reset_io_stat(&iolog->avg_window[ddir]); } static void _add_stat_to_log(struct io_log *iolog, unsigned long elapsed, - bool log_max) + bool log_max, uint8_t priority_bit) { int ddir; for (ddir = 0; ddir < DDIR_RWDIR_CNT; ddir++) - __add_stat_to_log(iolog, ddir, elapsed, log_max); + __add_stat_to_log(iolog, ddir, elapsed, log_max, priority_bit); } static unsigned long add_log_sample(struct thread_data *td, struct io_log *iolog, union io_sample_data data, - enum fio_ddir ddir, unsigned int bs, - uint64_t offset) + enum fio_ddir ddir, unsigned long long bs, + uint64_t offset, uint8_t priority_bit) { unsigned long elapsed, this_window; @@ -2355,7 +2713,7 @@ static unsigned long add_log_sample(struct thread_data *td, * If no time averaging, just add the log sample. */ if (!iolog->avg_msec) { - __add_log_sample(iolog, data, ddir, bs, elapsed, offset); + __add_log_sample(iolog, data, ddir, bs, elapsed, offset, priority_bit); return 0; } @@ -2379,7 +2737,7 @@ static unsigned long add_log_sample(struct thread_data *td, return diff; } - __add_stat_to_log(iolog, ddir, elapsed, td->o.log_max != 0); + __add_stat_to_log(iolog, ddir, elapsed, td->o.log_max != 0, priority_bit); iolog->avg_last[ddir] = elapsed - (this_window - iolog->avg_msec); return iolog->avg_msec; @@ -2392,18 +2750,19 @@ void finalize_logs(struct thread_data *td, bool unit_logs) elapsed = mtime_since_now(&td->epoch); if (td->clat_log && unit_logs) - _add_stat_to_log(td->clat_log, elapsed, td->o.log_max != 0); + _add_stat_to_log(td->clat_log, elapsed, td->o.log_max != 0, 0); if (td->slat_log && unit_logs) - _add_stat_to_log(td->slat_log, elapsed, td->o.log_max != 0); + _add_stat_to_log(td->slat_log, elapsed, td->o.log_max != 0, 0); if (td->lat_log && unit_logs) - _add_stat_to_log(td->lat_log, elapsed, td->o.log_max != 0); + _add_stat_to_log(td->lat_log, elapsed, td->o.log_max != 0, 0); if (td->bw_log && (unit_logs == per_unit_log(td->bw_log))) - _add_stat_to_log(td->bw_log, elapsed, td->o.log_max != 0); + _add_stat_to_log(td->bw_log, elapsed, td->o.log_max != 0, 0); if (td->iops_log && (unit_logs == per_unit_log(td->iops_log))) - _add_stat_to_log(td->iops_log, elapsed, td->o.log_max != 0); + _add_stat_to_log(td->iops_log, elapsed, td->o.log_max != 0, 0); } -void add_agg_sample(union io_sample_data data, enum fio_ddir ddir, unsigned int bs) +void add_agg_sample(union io_sample_data data, enum fio_ddir ddir, unsigned long long bs, + uint8_t priority_bit) { struct io_log *iolog; @@ -2411,7 +2770,7 @@ void add_agg_sample(union io_sample_data data, enum fio_ddir ddir, unsigned int return; iolog = agg_io_log[ddir]; - __add_log_sample(iolog, data, ddir, bs, mtime_since_genesis(), 0); + __add_log_sample(iolog, data, ddir, bs, mtime_since_genesis(), 0, priority_bit); } void add_sync_clat_sample(struct thread_stat *ts, unsigned long long nsec) @@ -2423,32 +2782,60 @@ void add_sync_clat_sample(struct thread_stat *ts, unsigned long long nsec) add_stat_sample(&ts->sync_stat, nsec); } -static void add_clat_percentile_sample(struct thread_stat *ts, - unsigned long long nsec, enum fio_ddir ddir) +static void add_lat_percentile_sample_noprio(struct thread_stat *ts, + unsigned long long nsec, enum fio_ddir ddir, enum fio_lat lat) { unsigned int idx = plat_val_to_idx(nsec); assert(idx < FIO_IO_U_PLAT_NR); - ts->io_u_plat[ddir][idx]++; + ts->io_u_plat[lat][ddir][idx]++; +} + +static void add_lat_percentile_sample(struct thread_stat *ts, + unsigned long long nsec, enum fio_ddir ddir, uint8_t priority_bit, + enum fio_lat lat) +{ + unsigned int idx = plat_val_to_idx(nsec); + + add_lat_percentile_sample_noprio(ts, nsec, ddir, lat); + + if (!priority_bit) + ts->io_u_plat_low_prio[ddir][idx]++; + else + ts->io_u_plat_high_prio[ddir][idx]++; } void add_clat_sample(struct thread_data *td, enum fio_ddir ddir, - unsigned long long nsec, unsigned int bs, uint64_t offset) + unsigned long long nsec, unsigned long long bs, + uint64_t offset, uint8_t priority_bit) { + const bool needs_lock = td_async_processing(td); unsigned long elapsed, this_window; struct thread_stat *ts = &td->ts; struct io_log *iolog = td->clat_hist_log; - td_io_u_lock(td); + if (needs_lock) + __td_io_u_lock(td); add_stat_sample(&ts->clat_stat[ddir], nsec); + if (!ts->lat_percentiles) { + if (priority_bit) + add_stat_sample(&ts->clat_high_prio_stat[ddir], nsec); + else + add_stat_sample(&ts->clat_low_prio_stat[ddir], nsec); + } + if (td->clat_log) add_log_sample(td, td->clat_log, sample_val(nsec), ddir, bs, - offset); + offset, priority_bit); - if (ts->clat_percentiles) - add_clat_percentile_sample(ts, nsec, ddir); + if (ts->clat_percentiles) { + if (ts->lat_percentiles) + add_lat_percentile_sample_noprio(ts, nsec, ddir, FIO_CLAT); + else + add_lat_percentile_sample(ts, nsec, ddir, priority_bit, FIO_CLAT); + } if (iolog && iolog->hist_msec) { struct io_hist *hw = &iolog->hist_window[ddir]; @@ -2458,7 +2845,7 @@ void add_clat_sample(struct thread_data *td, enum fio_ddir ddir, if (!hw->hist_last) hw->hist_last = elapsed; this_window = elapsed - hw->hist_last; - + if (this_window >= iolog->hist_msec) { uint64_t *io_u_plat; struct io_u_plat_entry *dst; @@ -2470,13 +2857,13 @@ void add_clat_sample(struct thread_data *td, enum fio_ddir ddir, * located in iolog.c after printing this sample to the * log file. */ - io_u_plat = (uint64_t *) td->ts.io_u_plat[ddir]; + io_u_plat = (uint64_t *) td->ts.io_u_plat[FIO_CLAT][ddir]; dst = malloc(sizeof(struct io_u_plat_entry)); memcpy(&(dst->io_u_plat), io_u_plat, - FIO_IO_U_PLAT_NR * sizeof(unsigned int)); + FIO_IO_U_PLAT_NR * sizeof(uint64_t)); flist_add(&dst->list, &hw->list); __add_log_sample(iolog, sample_plat(dst), ddir, bs, - elapsed, offset); + elapsed, offset, priority_bit); /* * Update the last time we recorded as being now, minus @@ -2488,52 +2875,71 @@ void add_clat_sample(struct thread_data *td, enum fio_ddir ddir, } } - td_io_u_unlock(td); + if (needs_lock) + __td_io_u_unlock(td); } void add_slat_sample(struct thread_data *td, enum fio_ddir ddir, - unsigned long usec, unsigned int bs, uint64_t offset) + unsigned long long nsec, unsigned long long bs, uint64_t offset, + uint8_t priority_bit) { + const bool needs_lock = td_async_processing(td); struct thread_stat *ts = &td->ts; if (!ddir_rw(ddir)) return; - td_io_u_lock(td); + if (needs_lock) + __td_io_u_lock(td); - add_stat_sample(&ts->slat_stat[ddir], usec); + add_stat_sample(&ts->slat_stat[ddir], nsec); if (td->slat_log) - add_log_sample(td, td->slat_log, sample_val(usec), ddir, bs, offset); + add_log_sample(td, td->slat_log, sample_val(nsec), ddir, bs, offset, + priority_bit); + + if (ts->slat_percentiles) + add_lat_percentile_sample_noprio(ts, nsec, ddir, FIO_SLAT); - td_io_u_unlock(td); + if (needs_lock) + __td_io_u_unlock(td); } void add_lat_sample(struct thread_data *td, enum fio_ddir ddir, - unsigned long long nsec, unsigned int bs, uint64_t offset) + unsigned long long nsec, unsigned long long bs, + uint64_t offset, uint8_t priority_bit) { + const bool needs_lock = td_async_processing(td); struct thread_stat *ts = &td->ts; if (!ddir_rw(ddir)) return; - td_io_u_lock(td); + if (needs_lock) + __td_io_u_lock(td); add_stat_sample(&ts->lat_stat[ddir], nsec); if (td->lat_log) add_log_sample(td, td->lat_log, sample_val(nsec), ddir, bs, - offset); + offset, priority_bit); - if (ts->lat_percentiles) - add_clat_percentile_sample(ts, nsec, ddir); + if (ts->lat_percentiles) { + add_lat_percentile_sample(ts, nsec, ddir, priority_bit, FIO_LAT); + if (priority_bit) + add_stat_sample(&ts->clat_high_prio_stat[ddir], nsec); + else + add_stat_sample(&ts->clat_low_prio_stat[ddir], nsec); - td_io_u_unlock(td); + } + if (needs_lock) + __td_io_u_unlock(td); } void add_bw_sample(struct thread_data *td, struct io_u *io_u, unsigned int bytes, unsigned long long spent) { + const bool needs_lock = td_async_processing(td); struct thread_stat *ts = &td->ts; unsigned long rate; @@ -2542,16 +2948,19 @@ void add_bw_sample(struct thread_data *td, struct io_u *io_u, else rate = 0; - td_io_u_lock(td); + if (needs_lock) + __td_io_u_lock(td); add_stat_sample(&ts->bw_stat[io_u->ddir], rate); if (td->bw_log) add_log_sample(td, td->bw_log, sample_val(rate), io_u->ddir, - bytes, io_u->offset); + bytes, io_u->offset, io_u_is_prio(io_u)); td->stat_io_bytes[io_u->ddir] = td->this_io_bytes[io_u->ddir]; - td_io_u_unlock(td); + + if (needs_lock) + __td_io_u_unlock(td); } static int __add_samples(struct thread_data *td, struct timespec *parent_tv, @@ -2560,6 +2969,7 @@ static int __add_samples(struct thread_data *td, struct timespec *parent_tv, struct io_stat *stat, struct io_log *log, bool is_kb) { + const bool needs_lock = td_async_processing(td); unsigned long spent, rate; enum fio_ddir ddir; unsigned long next, next_log; @@ -2570,7 +2980,8 @@ static int __add_samples(struct thread_data *td, struct timespec *parent_tv, if (spent < avg_time && avg_time - spent >= LOG_MSEC_SLACK) return avg_time - spent; - td_io_u_lock(td); + if (needs_lock) + __td_io_u_lock(td); /* * Compute both read and write rates for the interval. @@ -2593,21 +3004,22 @@ static int __add_samples(struct thread_data *td, struct timespec *parent_tv, add_stat_sample(&stat[ddir], rate); if (log) { - unsigned int bs = 0; + unsigned long long bs = 0; if (td->o.min_bs[ddir] == td->o.max_bs[ddir]) bs = td->o.min_bs[ddir]; - next = add_log_sample(td, log, sample_val(rate), ddir, bs, 0); + next = add_log_sample(td, log, sample_val(rate), ddir, bs, 0, 0); next_log = min(next_log, next); } stat_io_bytes[ddir] = this_io_bytes[ddir]; } - timespec_add_msec(parent_tv, avg_time); + *parent_tv = *t; - td_io_u_unlock(td); + if (needs_lock) + __td_io_u_unlock(td); if (spent <= avg_time) next = avg_time; @@ -2627,18 +3039,22 @@ static int add_bw_samples(struct thread_data *td, struct timespec *t) void add_iops_sample(struct thread_data *td, struct io_u *io_u, unsigned int bytes) { + const bool needs_lock = td_async_processing(td); struct thread_stat *ts = &td->ts; - td_io_u_lock(td); + if (needs_lock) + __td_io_u_lock(td); add_stat_sample(&ts->iops_stat[io_u->ddir], 1); if (td->iops_log) add_log_sample(td, td->iops_log, sample_val(1), io_u->ddir, - bytes, io_u->offset); + bytes, io_u->offset, io_u_is_prio(io_u)); td->stat_io_blocks[io_u->ddir] = td->this_io_blocks[io_u->ddir]; - td_io_u_unlock(td); + + if (needs_lock) + __td_io_u_unlock(td); } static int add_iops_samples(struct thread_data *td, struct timespec *t) @@ -2718,3 +3134,4 @@ uint32_t *io_u_block_info(struct thread_data *td, struct io_u *io_u) assert(idx < td->ts.nr_block_infos); return info; } +