X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=stat.c;h=ec75de241be623756c88c1b88edb69473a0dfb97;hp=91c74abef420f8fb0e8453e5a9be1fecff1e6f56;hb=96563db9995c5cebdce526f2bf97923153d4456b;hpb=b427f2e12beba7b0c9a655f8ecbd19187c0b6029 diff --git a/stat.c b/stat.c index 91c74abe..ec75de24 100644 --- a/stat.c +++ b/stat.c @@ -1,10 +1,7 @@ #include #include #include -#include #include -#include -#include #include #include "fio.h" @@ -17,10 +14,11 @@ #include "lib/output_buffer.h" #include "helper_thread.h" #include "smalloc.h" +#include "zbd.h" -#define LOG_MSEC_SLACK 10 +#define LOG_MSEC_SLACK 1 -struct fio_mutex *stat_mutex; +struct fio_sem *stat_sem; void clear_rusage_stat(struct thread_data *td) { @@ -135,15 +133,15 @@ static int double_cmp(const void *a, const void *b) return cmp; } -unsigned int calc_clat_percentiles(unsigned int *io_u_plat, unsigned long nr, +unsigned int calc_clat_percentiles(uint64_t *io_u_plat, unsigned long long nr, fio_fp64_t *plist, unsigned long long **output, unsigned long long *maxv, unsigned long long *minv) { - unsigned long sum = 0; + unsigned long long sum = 0; unsigned int len, i, j = 0; unsigned int oval_len = 0; unsigned long long *ovals = NULL; - int is_last; + bool is_last; *minv = -1ULL; *maxv = 0; @@ -166,7 +164,7 @@ unsigned int calc_clat_percentiles(unsigned int *io_u_plat, unsigned long nr, /* * Calculate bucket values, note down max and min values */ - is_last = 0; + 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)) { @@ -183,7 +181,7 @@ unsigned int calc_clat_percentiles(unsigned int *io_u_plat, unsigned long nr, if (ovals[j] > *maxv) *maxv = ovals[j]; - is_last = (j == len - 1); + is_last = (j == len - 1) != 0; if (is_last) break; @@ -198,18 +196,19 @@ unsigned int calc_clat_percentiles(unsigned int *io_u_plat, unsigned long nr, /* * Find and display the p-th percentile of clat */ -static void show_clat_percentiles(unsigned int *io_u_plat, unsigned long nr, +static void show_clat_percentiles(uint64_t *io_u_plat, unsigned long long nr, fio_fp64_t *plist, unsigned int precision, - struct buf_output *out) + const char *pre, struct buf_output *out) { unsigned int divisor, len, i, j = 0; unsigned long long minv, maxv; unsigned long long *ovals; - int is_last, per_line, scale_down, time_width; + int per_line, scale_down, time_width; + bool is_last; char fmt[32]; len = calc_clat_percentiles(io_u_plat, nr, plist, &ovals, &maxv, &minv); - if (!len) + if (!len || !ovals) goto out; /* @@ -219,15 +218,15 @@ static void show_clat_percentiles(unsigned int *io_u_plat, unsigned long nr, if (minv > 2000000 && maxv > 99999999ULL) { scale_down = 2; divisor = 1000000; - log_buf(out, " clat percentiles (msec):\n |"); + log_buf(out, " %s percentiles (msec):\n |", pre); } else if (minv > 2000 && maxv > 99999) { scale_down = 1; divisor = 1000; - log_buf(out, " clat percentiles (usec):\n |"); + log_buf(out, " %s percentiles (usec):\n |", pre); } else { scale_down = 0; divisor = 1; - log_buf(out, " clat percentiles (nsec):\n |"); + log_buf(out, " %s percentiles (nsec):\n |", pre); } @@ -243,7 +242,7 @@ static void show_clat_percentiles(unsigned int *io_u_plat, unsigned long nr, log_buf(out, " |"); /* end of the list */ - is_last = (j == len - 1); + is_last = (j == len - 1) != 0; for (i = 0; i < scale_down; i++) ovals[j] = (ovals[j] + 999) / 1000; @@ -297,14 +296,14 @@ void show_group_stats(struct group_run_stats *rs, struct buf_output *out) if (!rs->max_run[i]) continue; - io = num2str(rs->iobytes[i], 4, 1, i2p, N2S_BYTE); - ioalt = num2str(rs->iobytes[i], 4, 1, !i2p, N2S_BYTE); - agg = num2str(rs->agg[i], 4, 1, i2p, rs->unit_base); - aggalt = num2str(rs->agg[i], 4, 1, !i2p, rs->unit_base); - min = num2str(rs->min_bw[i], 4, 1, i2p, rs->unit_base); - minalt = num2str(rs->min_bw[i], 4, 1, !i2p, rs->unit_base); - max = num2str(rs->max_bw[i], 4, 1, i2p, rs->unit_base); - maxalt = num2str(rs->max_bw[i], 4, 1, !i2p, rs->unit_base); + io = num2str(rs->iobytes[i], rs->sig_figs, 1, i2p, N2S_BYTE); + ioalt = num2str(rs->iobytes[i], rs->sig_figs, 1, !i2p, N2S_BYTE); + agg = num2str(rs->agg[i], rs->sig_figs, 1, i2p, rs->unit_base); + aggalt = num2str(rs->agg[i], rs->sig_figs, 1, !i2p, rs->unit_base); + min = num2str(rs->min_bw[i], rs->sig_figs, 1, i2p, rs->unit_base); + minalt = num2str(rs->min_bw[i], rs->sig_figs, 1, !i2p, rs->unit_base); + max = num2str(rs->max_bw[i], rs->sig_figs, 1, i2p, rs->unit_base); + maxalt = num2str(rs->max_bw[i], rs->sig_figs, 1, !i2p, rs->unit_base); log_buf(out, "%s: bw=%s (%s), %s-%s (%s-%s), io=%s (%s), run=%llu-%llumsec\n", rs->unified_rw_rep ? " MIXED" : str[i], agg, aggalt, min, max, minalt, maxalt, io, ioalt, @@ -322,7 +321,7 @@ void show_group_stats(struct group_run_stats *rs, struct buf_output *out) } } -void stat_calc_dist(unsigned int *map, unsigned long total, double *io_u_dist) +void stat_calc_dist(uint64_t *map, unsigned long total, double *io_u_dist) { int i; @@ -341,7 +340,7 @@ void stat_calc_dist(unsigned int *map, unsigned long total, double *io_u_dist) } static void stat_calc_lat(struct thread_stat *ts, double *dst, - unsigned int *src, int nr) + uint64_t *src, int nr) { unsigned long total = ddir_rw_sum(ts->total_io_u); int i; @@ -364,7 +363,7 @@ 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; @@ -417,13 +416,25 @@ static void display_lat(const char *name, unsigned long long min, 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" }; 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(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, + io_ddir_name(ddir), out); + } + return; + } + assert(ddir_rw(ddir)); if (!ts->runtime[ddir]) @@ -433,18 +444,32 @@ static void show_ddir_status(struct group_run_stats *rs, struct thread_stat *ts, runt = ts->runtime[ddir]; bw = (1000 * ts->io_bytes[ddir]) / runt; - io_p = num2str(ts->io_bytes[ddir], 4, 1, i2p, N2S_BYTE); - bw_p = num2str(bw, 4, 1, i2p, ts->unit_base); - bw_p_alt = num2str(bw, 4, 1, !i2p, ts->unit_base); + io_p = num2str(ts->io_bytes[ddir], ts->sig_figs, 1, i2p, N2S_BYTE); + bw_p = num2str(bw, ts->sig_figs, 1, i2p, ts->unit_base); + bw_p_alt = num2str(bw, ts->sig_figs, 1, !i2p, ts->unit_base); iops = (1000 * (uint64_t)ts->total_io_u[ddir]) / runt; - iops_p = num2str(iops, 4, 1, 0, N2S_NONE); - - log_buf(out, " %s: IOPS=%s, BW=%s (%s)(%s/%llumsec)\n", - rs->unified_rw_rep ? "mixed" : str[ddir], + 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)%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); @@ -457,11 +482,19 @@ static void show_ddir_status(struct group_run_stats *rs, struct thread_stat *ts, if (calc_lat(&ts->lat_stat[ddir], &min, &max, &mean, &dev)) display_lat(" lat", min, max, mean, dev, out); - if (ts->clat_percentiles) { + if (ts->clat_percentiles || ts->lat_percentiles) { + const char *name = ts->clat_percentiles ? "clat" : " lat"; + uint64_t samples; + + if (ts->clat_percentiles) + samples = ts->clat_stat[ddir].samples; + else + samples = ts->lat_stat[ddir].samples; + show_clat_percentiles(ts->io_u_plat[ddir], - ts->clat_stat[ddir].samples, + samples, ts->percentile_list, - ts->percentile_precision, out); + ts->percentile_precision, 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; @@ -509,20 +542,21 @@ static void show_ddir_status(struct group_run_stats *rs, struct thread_stat *ts, } } -static int show_lat(double *io_u_lat, int nr, const char **ranges, - const char *msg, struct buf_output *out) +static bool show_lat(double *io_u_lat, int nr, const char **ranges, + const char *msg, struct buf_output *out) { - int new_line = 1, i, line = 0, shown = 0; + bool new_line = true, shown = false; + int i, line = 0; for (i = 0; i < nr; i++) { if (io_u_lat[i] <= 0.0) continue; - shown = 1; + shown = true; if (new_line) { if (line) log_buf(out, "\n"); log_buf(out, " lat (%s) : ", msg); - new_line = 0; + new_line = false; line = 0; } if (line) @@ -530,13 +564,13 @@ static int show_lat(double *io_u_lat, int nr, const char **ranges, log_buf(out, "%s%3.2f%%", ranges[i], io_u_lat[i]); line++; if (line == 5) - new_line = 1; + new_line = true; } if (shown) log_buf(out, "\n"); - return shown; + return true; } static void show_lat_n(double *io_u_lat_n, struct buf_output *out) @@ -599,8 +633,8 @@ 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); @@ -650,7 +684,6 @@ static int calc_block_percentiles(int nr_block_infos, uint32_t *block_infos, if (len > 1) qsort((void *)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 @@ -734,17 +767,17 @@ static void show_ss_normal(struct thread_stat *ts, struct buf_output *out) bw_mean = steadystate_bw_mean(ts); iops_mean = steadystate_iops_mean(ts); - p1 = num2str(bw_mean / ts->kb_base, 4, ts->kb_base, i2p, ts->unit_base); - p1alt = num2str(bw_mean / ts->kb_base, 4, ts->kb_base, !i2p, ts->unit_base); - p2 = num2str(iops_mean, 4, 1, 0, N2S_NONE); + p1 = num2str(bw_mean / ts->kb_base, ts->sig_figs, ts->kb_base, i2p, ts->unit_base); + p1alt = num2str(bw_mean / ts->kb_base, ts->sig_figs, ts->kb_base, !i2p, ts->unit_base); + p2 = num2str(iops_mean, ts->sig_figs, 1, 0, N2S_NONE); log_buf(out, " steadystate : attained=%s, bw=%s (%s), iops=%s, %s%s=%.3f%s\n", - ts->ss_state & __FIO_SS_ATTAINED ? "yes" : "no", + ts->ss_state & FIO_SS_ATTAINED ? "yes" : "no", p1, p1alt, p2, - ts->ss_state & __FIO_SS_IOPS ? "iops" : "bw", - ts->ss_state & __FIO_SS_SLOPE ? " slope": " mean dev", + ts->ss_state & FIO_SS_IOPS ? "iops" : "bw", + ts->ss_state & FIO_SS_SLOPE ? " slope": " mean dev", ts->ss_criterion.u.f, - ts->ss_state & __FIO_SS_PCT ? "%" : ""); + ts->ss_state & FIO_SS_PCT ? "%" : ""); free(p1); free(p1alt); @@ -792,6 +825,9 @@ static void show_thread_status_normal(struct thread_stat *ts, show_latencies(ts, out); + if (ts->sync_stat.samples) + show_ddir_status(rs, ts, DDIR_SYNC, out); + runtime = ts->total_run_time; if (runtime) { double runt = (double) runtime; @@ -828,12 +864,13 @@ static void show_thread_status_normal(struct thread_stat *ts, io_u_dist[1], io_u_dist[2], io_u_dist[3], io_u_dist[4], io_u_dist[5], io_u_dist[6]); - log_buf(out, " issued rwt: total=%llu,%llu,%llu," - " short=%llu,%llu,%llu," - " dropped=%llu,%llu,%llu\n", + log_buf(out, " issued rwts: total=%llu,%llu,%llu,%llu" + " short=%llu,%llu,%llu,0" + " dropped=%llu,%llu,%llu,0\n", (unsigned long long) ts->total_io_u[0], (unsigned long long) ts->total_io_u[1], (unsigned long long) ts->total_io_u[2], + (unsigned long long) ts->total_io_u[3], (unsigned long long) ts->short_io_u[0], (unsigned long long) ts->short_io_u[1], (unsigned long long) ts->short_io_u[2], @@ -896,7 +933,7 @@ 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 (ts->clat_percentiles) { + if (ts->clat_percentiles || ts->lat_percentiles) { len = calc_clat_percentiles(ts->io_u_plat[ddir], ts->clat_stat[ddir].samples, ts->percentile_list, &ovals, &maxv, @@ -952,95 +989,136 @@ 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 bw; + 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"}; - struct json_object *dir_object, *tmp_object, *percentile_object, *clat_bins_object; + struct json_object *dir_object, *tmp_object, *percentile_object, *clat_bins_object = NULL; char buf[120]; double p_of_agg = 100.0; - assert(ddir_rw(ddir)); + assert(ddir_rw(ddir) || ddir_sync(ddir)); if (ts->unified_rw_rep && ddir != DDIR_READ) return; 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; + bw = 0; + iops = 0.0; + if (ts->runtime[ddir]) { + uint64_t runt = ts->runtime[ddir]; + + bw_bytes = ((1000 * ts->io_bytes[ddir]) / runt); /* Bytes/s */ + bw = bw_bytes / 1024; /* KiB/s */ + iops = (1000.0 * (uint64_t) ts->total_io_u[ddir]) / runt; + } - bw = 0; - iops = 0.0; - if (ts->runtime[ddir]) { - uint64_t runt = ts->runtime[ddir]; + json_object_add_value_int(dir_object, "io_bytes", ts->io_bytes[ddir]); + json_object_add_value_int(dir_object, "io_kbytes", ts->io_bytes[ddir] >> 10); + json_object_add_value_int(dir_object, "bw_bytes", bw_bytes); + json_object_add_value_int(dir_object, "bw", bw); + json_object_add_value_float(dir_object, "iops", iops); + json_object_add_value_int(dir_object, "runtime", ts->runtime[ddir]); + json_object_add_value_int(dir_object, "total_ios", ts->total_io_u[ddir]); + 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(); + 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(); + 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; + } - bw = ((1000 * ts->io_bytes[ddir]) / runt) / 1024; /* KiB/s */ - iops = (1000.0 * (uint64_t) ts->total_io_u[ddir]) / runt; + tmp_object = json_create_object(); + json_object_add_value_object(dir_object, "lat_ns", tmp_object); + 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); } - json_object_add_value_int(dir_object, "io_bytes", ts->io_bytes[ddir]); - json_object_add_value_int(dir_object, "io_kbytes", ts->io_bytes[ddir] >> 10); - json_object_add_value_int(dir_object, "bw", bw); - json_object_add_value_float(dir_object, "iops", iops); - json_object_add_value_int(dir_object, "runtime", ts->runtime[ddir]); - json_object_add_value_int(dir_object, "total_ios", ts->total_io_u[ddir]); - 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(); - 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 (ts->clat_percentiles || ts->lat_percentiles) { + if (ddir_rw(ddir)) { + uint64_t samples; - if (!calc_lat(&ts->clat_stat[ddir], &min, &max, &mean, &dev)) { - min = max = 0; - mean = dev = 0.0; - } - tmp_object = json_create_object(); - 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); + if (ts->clat_percentiles) + samples = ts->clat_stat[ddir].samples; + else + samples = ts->lat_stat[ddir].samples; - if (ts->clat_percentiles) { - len = calc_clat_percentiles(ts->io_u_plat[ddir], - ts->clat_stat[ddir].samples, + len = calc_clat_percentiles(ts->io_u_plat[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 < FIO_IO_U_LIST_MAX_LEN; i++) { - if (i >= len) { - json_object_add_value_int(percentile_object, "0.00", 0); - continue; - } + 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 (output_format & FIO_OUTPUT_JSON_PLUS) { clat_bins_object = json_create_object(); - json_object_add_value_object(tmp_object, "bins", clat_bins_object); + if (ts->clat_percentiles) + json_object_add_value_object(tmp_object, "bins", clat_bins_object); + for(i = 0; i < FIO_IO_U_PLAT_NR; i++) { - 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]); + 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 (!ddir_rw(ddir)) + return; + if (!calc_lat(&ts->lat_stat[ddir], &min, &max, &mean, &dev)) { min = max = 0; mean = dev = 0.0; @@ -1051,6 +1129,9 @@ static void add_ddir_status_json(struct thread_stat *ts, 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); @@ -1082,6 +1163,16 @@ 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, @@ -1143,7 +1234,7 @@ 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*/ @@ -1160,7 +1251,7 @@ static void show_thread_status_terse_all(struct thread_stat *ts, } static void json_add_job_opts(struct json_object *root, const char *name, - struct flist_head *opt_list, bool num_jobs) + struct flist_head *opt_list) { struct json_object *dir_object; struct flist_head *entry; @@ -1176,8 +1267,6 @@ static void json_add_job_opts(struct json_object *root, const char *name, const char *pos = ""; p = flist_entry(entry, struct print_option, list); - if (!num_jobs && !strcmp(p->name, "numjobs")) - continue; if (p->value) pos = p->value; json_object_add_value_string(dir_object, p->name, pos); @@ -1211,11 +1300,12 @@ static struct json_object *show_thread_status_json(struct thread_stat *ts, } if (opt_list) - json_add_job_opts(root, "job options", opt_list, true); + json_add_job_opts(root, "job options", opt_list); add_ddir_status_json(ts, rs, DDIR_READ, root); add_ddir_status_json(ts, rs, DDIR_WRITE, root); add_ddir_status_json(ts, rs, DDIR_TRIM, root); + add_ddir_status_json(ts, rs, DDIR_SYNC, root); /* CPU Usage */ if (ts->total_run_time) { @@ -1227,19 +1317,15 @@ 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*/ @@ -1252,6 +1338,44 @@ 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); @@ -1337,23 +1461,23 @@ 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", - ts->ss_state & __FIO_SS_IOPS ? "iops" : "bw", - ts->ss_state & __FIO_SS_SLOPE ? "_slope" : "", + ts->ss_state & FIO_SS_IOPS ? "iops" : "bw", + ts->ss_state & FIO_SS_SLOPE ? "_slope" : "", (float) ts->ss_limit.u.f, - ts->ss_state & __FIO_SS_PCT ? "%" : ""); + ts->ss_state & FIO_SS_PCT ? "%" : ""); tmp = json_create_object(); json_object_add_value_object(root, "steadystate", tmp); json_object_add_value_string(tmp, "ss", ss_buf); json_object_add_value_int(tmp, "duration", (int)ts->ss_dur); - json_object_add_value_int(tmp, "attained", (ts->ss_state & __FIO_SS_ATTAINED) > 0); + json_object_add_value_int(tmp, "attained", (ts->ss_state & FIO_SS_ATTAINED) > 0); snprintf(ss_buf, sizeof(ss_buf), "%f%s", (float) ts->ss_criterion.u.f, - ts->ss_state & __FIO_SS_PCT ? "%" : ""); + ts->ss_state & FIO_SS_PCT ? "%" : ""); json_object_add_value_string(tmp, "criterion", ss_buf); json_object_add_value_float(tmp, "max_deviation", ts->ss_deviation.u.f); json_object_add_value_float(tmp, "slope", ts->ss_slope.u.f); @@ -1369,12 +1493,12 @@ static struct json_object *show_thread_status_json(struct thread_stat *ts, ** otherwise it actually points to the second element ** in the list */ - if ((ts->ss_state & __FIO_SS_ATTAINED) || !(ts->ss_state & __FIO_SS_BUFFER_FULL)) + if ((ts->ss_state & FIO_SS_ATTAINED) || !(ts->ss_state & FIO_SS_BUFFER_FULL)) 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]); } @@ -1414,13 +1538,10 @@ 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); @@ -1447,6 +1568,39 @@ 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) @@ -1471,6 +1625,8 @@ void sum_group_stats(struct group_run_stats *dst, struct group_run_stats *src) dst->kb_base = src->kb_base; if (!dst->unit_base) dst->unit_base = src->unit_base; + if (!dst->sig_figs) + dst->sig_figs = src->sig_figs; } void sum_thread_stats(struct thread_stat *dst, struct thread_stat *src, @@ -1480,22 +1636,22 @@ void sum_thread_stats(struct thread_stat *dst, struct thread_stat *src, 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->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->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]; @@ -1510,24 +1666,25 @@ void sum_thread_stats(struct thread_stat *dst, struct thread_stat *src, } } + 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; dst->majf += src->majf; dst->minf += src->minf; - for (k = 0; k < FIO_IO_U_MAP_NR; k++) + for (k = 0; k < FIO_IO_U_MAP_NR; k++) { dst->io_u_map[k] += src->io_u_map[k]; - for (k = 0; k < FIO_IO_U_MAP_NR; k++) dst->io_u_submit[k] += src->io_u_submit[k]; - for (k = 0; k < FIO_IO_U_MAP_NR; 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) { @@ -1541,6 +1698,8 @@ 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; @@ -1555,6 +1714,9 @@ void sum_thread_stats(struct thread_stat *dst, struct thread_stat *src, 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) @@ -1579,6 +1741,7 @@ void init_thread_stat(struct thread_stat *ts) ts->bw_stat[j].min_val = -1UL; ts->iops_stat[j].min_val = -1UL; } + ts->sync_stat.min_val = -1UL; ts->groupid = -1; } @@ -1588,8 +1751,8 @@ void __show_run_stats(void) struct thread_data *td; struct thread_stat *threadstats, *ts; int i, j, k, nr_ts, last_ts, idx; - int kb_base_warned = 0; - int unit_base_warned = 0; + bool kb_base_warned = false; + bool unit_base_warned = false; struct json_object *root = NULL; struct json_array *array = NULL; struct buf_output output[FIO_OUTPUT_NR]; @@ -1645,6 +1808,7 @@ void __show_run_stats(void) ts = &threadstats[j]; ts->clat_percentiles = td->o.clat_percentiles; + ts->lat_percentiles = td->o.lat_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; @@ -1677,15 +1841,16 @@ void __show_run_stats(void) ts->kb_base = td->o.kb_base; ts->unit_base = td->o.unit_base; + ts->sig_figs = td->o.sig_figs; ts->unified_rw_rep = td->o.unified_rw_rep; } else if (ts->kb_base != td->o.kb_base && !kb_base_warned) { log_info("fio: kb_base differs for jobs in group, using" " %u as the base\n", ts->kb_base); - kb_base_warned = 1; + kb_base_warned = true; } else if (ts->unit_base != td->o.unit_base && !unit_base_warned) { log_info("fio: unit_base differs for jobs in group, using" " %u as the base\n", ts->unit_base); - unit_base_warned = 1; + unit_base_warned = true; } ts->continue_on_error = td->o.continue_on_error; @@ -1739,6 +1904,7 @@ void __show_run_stats(void) rs = &runstats[ts->groupid]; rs->kb_base = ts->kb_base; rs->unit_base = ts->unit_base; + rs->sig_figs = ts->sig_figs; rs->unified_rw_rep += ts->unified_rw_rep; for (j = 0; j < DDIR_RWDIR_CNT; j++) { @@ -1786,13 +1952,14 @@ void __show_run_stats(void) char time_buf[32]; struct timeval now; unsigned long long ms_since_epoch; + time_t tv_sec; gettimeofday(&now, NULL); ms_since_epoch = (unsigned long long)(now.tv_sec) * 1000 + (unsigned long long)(now.tv_usec) / 1000; - os_ctime_r((const time_t *) &now.tv_sec, time_buf, - sizeof(time_buf)); + tv_sec = now.tv_sec; + os_ctime_r(&tv_sec, time_buf, sizeof(time_buf)); if (time_buf[strlen(time_buf) - 1] == '\n') time_buf[strlen(time_buf) - 1] = '\0'; @@ -1802,7 +1969,7 @@ void __show_run_stats(void) json_object_add_value_int(root, "timestamp_ms", ms_since_epoch); json_object_add_value_string(root, "time", time_buf); global = get_global_options(); - json_add_job_opts(root, "global options", &global->opt_list, false); + json_add_job_opts(root, "global options", &global->opt_list); array = json_create_array(); json_object_add_value_array(root, "jobs", array); } @@ -1863,19 +2030,14 @@ 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_mutex_down(stat_mutex); - __show_run_stats(); - fio_mutex_up(stat_mutex); -} - void __show_running_run_stats(void) { struct thread_data *td; @@ -1883,7 +2045,7 @@ void __show_running_run_stats(void) struct timespec ts; int i; - fio_mutex_down(stat_mutex); + fio_sem_down(stat_sem); rt = malloc(thread_number * sizeof(unsigned long long)); fio_gettime(&ts, NULL); @@ -1909,7 +2071,7 @@ void __show_running_run_stats(void) continue; if (td->rusage_sem) { td->update_rusage = 1; - fio_mutex_down(td->rusage_sem); + fio_sem_down(td->rusage_sem); } td->update_rusage = 0; } @@ -1926,12 +2088,12 @@ void __show_running_run_stats(void) } free(rt); - fio_mutex_up(stat_mutex); + fio_sem_up(stat_sem); } -static int status_interval_init; +static bool status_interval_init; static struct timespec status_time; -static int status_file_disabled; +static bool status_file_disabled; #define FIO_STATUS_FILE "fio-dump-status" @@ -1962,7 +2124,7 @@ static int check_status_file(void) log_err("fio: failed to unlink %s: %s\n", fio_status_file_path, strerror(errno)); log_err("fio: disabling status file updates\n"); - status_file_disabled = 1; + status_file_disabled = true; } return 1; @@ -1973,7 +2135,7 @@ void check_for_running_stats(void) if (status_interval) { if (!status_interval_init) { fio_gettime(&status_time, NULL); - status_interval_init = 1; + status_interval_init = true; } else if (mtime_since_now(&status_time) >= status_interval) { show_running_run_stats(); fio_gettime(&status_time, NULL); @@ -2141,17 +2303,19 @@ 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, + enum fio_ddir ddir, unsigned long long bs, unsigned long t, uint64_t offset) { struct io_logs *cur_log; @@ -2159,7 +2323,7 @@ static void __add_log_sample(struct io_log *iolog, union io_sample_data data, if (iolog->disabled) return; if (flist_empty(&iolog->io_logs)) - iolog->avg_last = t; + iolog->avg_last[ddir] = t; cur_log = get_cur_log(iolog); if (cur_log) { @@ -2209,10 +2373,15 @@ void reset_io_stats(struct thread_data *td) ts->short_io_u[i] = 0; ts->drop_io_u[i] = 0; - for (j = 0; j < FIO_IO_U_PLAT_NR; j++) + for (j = 0; j < FIO_IO_U_PLAT_NR; j++) { ts->io_u_plat[i][j] = 0; + if (!i) + ts->io_u_sync_plat[j] = 0; + } } + ts->total_io_u[DDIR_SYNC] = 0; + for (i = 0; i < FIO_IO_U_MAP_NR; i++) { ts->io_u_map[i] = 0; ts->io_u_submit[i] = 0; @@ -2228,6 +2397,8 @@ 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, @@ -2261,9 +2432,11 @@ static void _add_stat_to_log(struct io_log *iolog, unsigned long elapsed, __add_stat_to_log(iolog, ddir, elapsed, log_max); } -static 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) +static unsigned long add_log_sample(struct thread_data *td, + struct io_log *iolog, + union io_sample_data data, + enum fio_ddir ddir, unsigned long long bs, + uint64_t offset) { unsigned long elapsed, this_window; @@ -2290,19 +2463,19 @@ static long add_log_sample(struct thread_data *td, struct io_log *iolog, * If period hasn't passed, adding the above sample is all we * need to do. */ - this_window = elapsed - iolog->avg_last; - if (elapsed < iolog->avg_last) - return iolog->avg_last - elapsed; + this_window = elapsed - iolog->avg_last[ddir]; + if (elapsed < iolog->avg_last[ddir]) + return iolog->avg_last[ddir] - elapsed; else if (this_window < iolog->avg_msec) { - int diff = iolog->avg_msec - this_window; + unsigned long diff = iolog->avg_msec - this_window; if (inline_log(iolog) || diff > LOG_MSEC_SLACK) return diff; } - _add_stat_to_log(iolog, elapsed, td->o.log_max != 0); + __add_stat_to_log(iolog, ddir, elapsed, td->o.log_max != 0); - iolog->avg_last = elapsed - (this_window - iolog->avg_msec); + iolog->avg_last[ddir] = elapsed - (this_window - iolog->avg_msec); return iolog->avg_msec; } @@ -2324,7 +2497,7 @@ void finalize_logs(struct thread_data *td, bool unit_logs) _add_stat_to_log(td->iops_log, elapsed, td->o.log_max != 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) { struct io_log *iolog; @@ -2335,6 +2508,15 @@ void add_agg_sample(union io_sample_data data, enum fio_ddir ddir, unsigned int __add_log_sample(iolog, data, ddir, bs, mtime_since_genesis(), 0); } +void add_sync_clat_sample(struct thread_stat *ts, unsigned long long nsec) +{ + unsigned int idx = plat_val_to_idx(nsec); + assert(idx < FIO_IO_U_PLAT_NR); + + ts->io_u_sync_plat[idx]++; + 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) { @@ -2345,13 +2527,16 @@ static void add_clat_percentile_sample(struct thread_stat *ts, } 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) { + 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); @@ -2372,7 +2557,7 @@ void add_clat_sample(struct thread_data *td, enum fio_ddir ddir, this_window = elapsed - hw->hist_last; if (this_window >= iolog->hist_msec) { - unsigned int *io_u_plat; + uint64_t *io_u_plat; struct io_u_plat_entry *dst; /* @@ -2382,7 +2567,7 @@ 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 = (unsigned int *) td->ts.io_u_plat[ddir]; + io_u_plat = (uint64_t *) td->ts.io_u_plat[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)); @@ -2400,36 +2585,43 @@ 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 usec, unsigned long long bs, uint64_t offset) { + 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); if (td->slat_log) add_log_sample(td, td->slat_log, sample_val(usec), ddir, bs, offset); - 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) { + 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); @@ -2437,12 +2629,17 @@ void add_lat_sample(struct thread_data *td, enum fio_ddir ddir, add_log_sample(td, td->lat_log, sample_val(nsec), ddir, bs, offset); - td_io_u_unlock(td); + if (ts->lat_percentiles) + add_clat_percentile_sample(ts, nsec, ddir); + + 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; @@ -2451,7 +2648,8 @@ 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); @@ -2460,7 +2658,9 @@ void add_bw_sample(struct thread_data *td, struct io_u *io_u, bytes, io_u->offset); 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, @@ -2469,9 +2669,10 @@ 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 int next, next_log; + unsigned long next, next_log; next_log = avg_time; @@ -2479,7 +2680,8 @@ 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. @@ -2502,7 +2704,7 @@ 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]; @@ -2516,7 +2718,8 @@ static int __add_samples(struct thread_data *td, struct timespec *parent_tv, timespec_add_msec(parent_tv, avg_time); - td_io_u_unlock(td); + if (needs_lock) + __td_io_u_unlock(td); if (spent <= avg_time) next = avg_time; @@ -2536,9 +2739,11 @@ 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); @@ -2547,7 +2752,9 @@ void add_iops_sample(struct thread_data *td, struct io_u *io_u, bytes, io_u->offset); 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) @@ -2596,7 +2803,7 @@ int calc_log_samples(void) void stat_init(void) { - stat_mutex = fio_mutex_init(FIO_MUTEX_UNLOCKED); + stat_sem = fio_sem_init(FIO_SEM_UNLOCKED); } void stat_exit(void) @@ -2605,8 +2812,8 @@ void stat_exit(void) * When we have the mutex, we know out-of-band access to it * have ended. */ - fio_mutex_down(stat_mutex); - fio_mutex_remove(stat_mutex); + fio_sem_down(stat_sem); + fio_sem_remove(stat_sem); } /*