X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=stat.c;h=77e389cb1f7f85baa0c600a56adfc5e32c0274f7;hp=28acc233870fea2fa9e29abd8d5664b2101e7b89;hb=9966af7bc3b881e814d337c2417c8033a8848a66;hpb=eef02441621aa969f01a1a331e0215dd587d25af diff --git a/stat.c b/stat.c index 28acc233..77e389cb 100644 --- a/stat.c +++ b/stat.c @@ -14,6 +14,8 @@ #include "lib/getrusage.h" #include "idletime.h" +static struct fio_mutex *stat_mutex; + void update_rusage_stat(struct thread_data *td) { struct thread_stat *ts = &td->ts; @@ -65,12 +67,12 @@ static unsigned int plat_val_to_idx(unsigned int val) /* * Discard the error bits and apply the mask to find the - * index for the buckets in the group + * index for the buckets in the group */ offset = (FIO_IO_U_PLAT_VAL - 1) & (val >> error_bits); /* Make sure the index does not exceed (array size - 1) */ - idx = (base + offset) < (FIO_IO_U_PLAT_NR - 1)? + idx = (base + offset) < (FIO_IO_U_PLAT_NR - 1) ? (base + offset) : (FIO_IO_U_PLAT_NR - 1); return idx; @@ -88,11 +90,11 @@ static unsigned int plat_idx_to_val(unsigned int idx) /* MSB <= (FIO_IO_U_PLAT_BITS-1), cannot be rounded off. Use * all bits of the sample as index */ - if (idx < (FIO_IO_U_PLAT_VAL << 1) ) + if (idx < (FIO_IO_U_PLAT_VAL << 1)) return idx; /* Find the group and compute the minimum value of that group */ - error_bits = (idx >> FIO_IO_U_PLAT_BITS) -1; + error_bits = (idx >> FIO_IO_U_PLAT_BITS) - 1; base = 1 << (error_bits + FIO_IO_U_PLAT_BITS); /* Find its bucket number of the group */ @@ -116,11 +118,9 @@ static int double_cmp(const void *a, const void *b) return cmp; } -static unsigned int calc_clat_percentiles(unsigned int *io_u_plat, - unsigned long nr, fio_fp64_t *plist, - unsigned int **output, - unsigned int *maxv, - unsigned int *minv) +unsigned int calc_clat_percentiles(unsigned int *io_u_plat, unsigned long nr, + fio_fp64_t *plist, unsigned int **output, + unsigned int *maxv, unsigned int *minv) { unsigned long sum = 0; unsigned int len, i, j = 0; @@ -144,7 +144,7 @@ static unsigned int calc_clat_percentiles(unsigned int *io_u_plat, * 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((void *)plist, len, sizeof(plist[0]), double_cmp); /* * Calculate bucket values, note down max and min values @@ -182,7 +182,7 @@ static unsigned int calc_clat_percentiles(unsigned int *io_u_plat, * Find and display the p-th percentile of clat */ static void show_clat_percentiles(unsigned int *io_u_plat, unsigned long nr, - fio_fp64_t *plist, uint64_t precision) + fio_fp64_t *plist, unsigned int precision) { unsigned int len, j = 0, minv, maxv; unsigned int *ovals; @@ -205,7 +205,7 @@ static void show_clat_percentiles(unsigned int *io_u_plat, unsigned long nr, log_info(" clat percentiles (usec):\n |"); } - snprintf(fmt, sizeof(fmt), "%%1.%luf", precision); + snprintf(fmt, sizeof(fmt), "%%1.%uf", precision); per_line = (80 - 7) / (precision + 14); for (j = 0; j < len; j++) { @@ -240,18 +240,16 @@ out: free(ovals); } -static int calc_lat(struct io_stat *is, unsigned long *min, unsigned long *max, - double *mean, double *dev) +int calc_lat(struct io_stat *is, unsigned long *min, unsigned long *max, + double *mean, double *dev) { - double n = is->samples; + double n = (double) is->samples; - if (is->samples == 0) + if (n == 0) return 0; *min = is->min_val; *max = is->max_val; - - n = (double) is->samples; *mean = is->mean.u.f; if (n > 1.0) @@ -276,15 +274,17 @@ void show_group_stats(struct group_run_stats *rs) if (!rs->max_run[i]) continue; - p1 = num2str(rs->io_kb[i], 6, rs->kb_base, i2p); - p2 = num2str(rs->agg[i], 6, rs->kb_base, i2p); - p3 = num2str(rs->min_bw[i], 6, rs->kb_base, i2p); - p4 = num2str(rs->max_bw[i], 6, rs->kb_base, i2p); + p1 = num2str(rs->io_kb[i], 6, rs->kb_base, i2p, 8); + p2 = num2str(rs->agg[i], 6, rs->kb_base, i2p, rs->unit_base); + p3 = num2str(rs->min_bw[i], 6, rs->kb_base, i2p, rs->unit_base); + p4 = num2str(rs->max_bw[i], 6, rs->kb_base, i2p, rs->unit_base); - log_info("%s: io=%sB, aggrb=%sB/s, minb=%sB/s, maxb=%sB/s," + log_info("%s: io=%s, aggrb=%s/s, minb=%s/s, maxb=%s/s," " mint=%llumsec, maxt=%llumsec\n", rs->unified_rw_rep ? " MIXED" : ddir_str[i], - p1, p2, p3, p4, rs->min_run[i], rs->max_run[i]); + p1, p2, p3, p4, + (unsigned long long) rs->min_run[i], + (unsigned long long) rs->max_run[i]); free(p1); free(p2); @@ -293,12 +293,7 @@ void show_group_stats(struct group_run_stats *rs) } } -#define ts_total_io_u(ts) \ - ((ts)->total_io_u[DDIR_READ] + (ts)->total_io_u[DDIR_WRITE] +\ - (ts)->total_io_u[DDIR_TRIM]) - -static void stat_calc_dist(unsigned int *map, unsigned long total, - double *io_u_dist) +void stat_calc_dist(unsigned int *map, unsigned long total, double *io_u_dist) { int i; @@ -319,7 +314,7 @@ static void stat_calc_dist(unsigned int *map, unsigned long total, static void stat_calc_lat(struct thread_stat *ts, double *dst, unsigned int *src, int nr) { - unsigned long total = ts_total_io_u(ts); + unsigned long total = ddir_rw_sum(ts->total_io_u); int i; /* @@ -336,28 +331,33 @@ static void stat_calc_lat(struct thread_stat *ts, double *dst, } } -static void stat_calc_lat_u(struct thread_stat *ts, double *io_u_lat) +void stat_calc_lat_u(struct thread_stat *ts, double *io_u_lat) { stat_calc_lat(ts, io_u_lat, ts->io_u_lat_u, FIO_IO_U_LAT_U_NR); } -static void stat_calc_lat_m(struct thread_stat *ts, double *io_u_lat) +void stat_calc_lat_m(struct thread_stat *ts, double *io_u_lat) { stat_calc_lat(ts, io_u_lat, ts->io_u_lat_m, FIO_IO_U_LAT_M_NR); } -static int usec_to_msec(unsigned long *min, unsigned long *max, double *mean, - double *dev) +static void display_lat(const char *name, unsigned long min, unsigned long max, + double mean, double dev) { - if (*min > 1000 && *max > 1000 && *mean > 1000.0 && *dev > 1000.0) { - *min /= 1000; - *max /= 1000; - *mean /= 1000.0; - *dev /= 1000.0; - return 0; - } + const char *base = "(usec)"; + char *minp, *maxp; - return 1; + if (!usec_to_msec(&min, &max, &mean, &dev)) + base = "(msec)"; + + minp = num2str(min, 6, 1, 0, 0); + maxp = num2str(max, 6, 1, 0, 0); + + log_info(" %s %s: min=%s, max=%s, avg=%5.02f," + " stdev=%5.02f\n", name, base, minp, maxp, mean, dev); + + free(minp); + free(maxp); } static void show_ddir_status(struct group_run_stats *rs, struct thread_stat *ts, @@ -379,68 +379,28 @@ 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], 6, 1, i2p); - bw_p = num2str(bw, 6, 1, i2p); + io_p = num2str(ts->io_bytes[ddir], 6, 1, i2p, 8); + bw_p = num2str(bw, 6, 1, i2p, ts->unit_base); iops = (1000 * (uint64_t)ts->total_io_u[ddir]) / runt; - iops_p = num2str(iops, 6, 1, 0); + iops_p = num2str(iops, 6, 1, 0, 0); - log_info(" %s: io=%sB, bw=%sB/s, iops=%s, runt=%6llumsec\n", + log_info(" %s: io=%s, bw=%s/s, iops=%s, runt=%6llumsec\n", rs->unified_rw_rep ? "mixed" : ddir_str[ddir], - io_p, bw_p, iops_p, ts->runtime[ddir]); + io_p, bw_p, iops_p, + (unsigned long long) ts->runtime[ddir]); free(io_p); free(bw_p); free(iops_p); - if (calc_lat(&ts->slat_stat[ddir], &min, &max, &mean, &dev)) { - const char *base = "(usec)"; - char *minp, *maxp; - - if (!usec_to_msec(&min, &max, &mean, &dev)) - base = "(msec)"; - - minp = num2str(min, 6, 1, 0); - maxp = num2str(max, 6, 1, 0); - - log_info(" slat %s: min=%s, max=%s, avg=%5.02f," - " stdev=%5.02f\n", base, minp, maxp, mean, dev); - - free(minp); - free(maxp); - } - if (calc_lat(&ts->clat_stat[ddir], &min, &max, &mean, &dev)) { - const char *base = "(usec)"; - char *minp, *maxp; - - if (!usec_to_msec(&min, &max, &mean, &dev)) - base = "(msec)"; - - minp = num2str(min, 6, 1, 0); - maxp = num2str(max, 6, 1, 0); - - log_info(" clat %s: min=%s, max=%s, avg=%5.02f," - " stdev=%5.02f\n", base, minp, maxp, mean, dev); - - free(minp); - free(maxp); - } - if (calc_lat(&ts->lat_stat[ddir], &min, &max, &mean, &dev)) { - const char *base = "(usec)"; - char *minp, *maxp; - - if (!usec_to_msec(&min, &max, &mean, &dev)) - base = "(msec)"; - - minp = num2str(min, 6, 1, 0); - maxp = num2str(max, 6, 1, 0); - - log_info(" lat %s: min=%s, max=%s, avg=%5.02f," - " stdev=%5.02f\n", base, minp, maxp, mean, dev); + if (calc_lat(&ts->slat_stat[ddir], &min, &max, &mean, &dev)) + display_lat("slat", min, max, mean, dev); + if (calc_lat(&ts->clat_stat[ddir], &min, &max, &mean, &dev)) + display_lat("clat", min, max, mean, dev); + if (calc_lat(&ts->lat_stat[ddir], &min, &max, &mean, &dev)) + display_lat(" lat", min, max, mean, dev); - free(minp); - free(maxp); - } if (ts->clat_percentiles) { show_clat_percentiles(ts->io_u_plat[ddir], ts->clat_stat[ddir].samples, @@ -448,8 +408,15 @@ static void show_ddir_status(struct group_run_stats *rs, struct thread_stat *ts, ts->percentile_precision); } if (calc_lat(&ts->bw_stat[ddir], &min, &max, &mean, &dev)) { - double p_of_agg = 100.0; - const char *bw_str = "KB"; + double p_of_agg = 100.0, fkb_base = (double)rs->kb_base; + const char *bw_str = (rs->unit_base == 1 ? "Kbit" : "KB"); + + if (rs->unit_base == 1) { + min *= 8.0; + max *= 8.0; + mean *= 8.0; + dev *= 8.0; + } if (rs->agg[ddir]) { p_of_agg = mean * 100 / (double) rs->agg[ddir]; @@ -457,15 +424,15 @@ static void show_ddir_status(struct group_run_stats *rs, struct thread_stat *ts, p_of_agg = 100.0; } - if (mean > 999999.9) { - min /= 1000.0; - max /= 1000.0; - mean /= 1000.0; - dev /= 1000.0; - bw_str = "MB"; + if (mean > fkb_base * fkb_base) { + min /= fkb_base; + max /= fkb_base; + mean /= fkb_base; + dev /= fkb_base; + bw_str = (rs->unit_base == 1 ? "Mbit" : "MB"); } - log_info(" bw (%s/s) : min=%5lu, max=%5lu, per=%3.2f%%," + log_info(" bw (%-4s/s): min=%5lu, max=%5lu, per=%3.2f%%," " avg=%5.02f, stdev=%5.02f\n", bw_str, min, max, p_of_agg, mean, dev); } @@ -518,25 +485,28 @@ static void show_lat_m(double *io_u_lat_m) show_lat(io_u_lat_m, FIO_IO_U_LAT_M_NR, ranges, "msec"); } -static void show_latencies(double *io_u_lat_u, double *io_u_lat_m) +static void show_latencies(struct thread_stat *ts) { + double io_u_lat_u[FIO_IO_U_LAT_U_NR]; + double io_u_lat_m[FIO_IO_U_LAT_M_NR]; + + stat_calc_lat_u(ts, io_u_lat_u); + stat_calc_lat_m(ts, io_u_lat_m); + show_lat_u(io_u_lat_u); show_lat_m(io_u_lat_m); } -void show_thread_status(struct thread_stat *ts, struct group_run_stats *rs) +static void show_thread_status_normal(struct thread_stat *ts, + struct group_run_stats *rs) { double usr_cpu, sys_cpu; unsigned long runtime; double io_u_dist[FIO_IO_U_MAP_NR]; - double io_u_lat_u[FIO_IO_U_LAT_U_NR]; - double io_u_lat_m[FIO_IO_U_LAT_M_NR]; time_t time_p; char time_buf[64]; - if (!(ts->io_bytes[DDIR_READ] + ts->io_bytes[DDIR_WRITE] + - ts->io_bytes[DDIR_TRIM]) && !(ts->total_io_u[DDIR_READ] + - ts->total_io_u[DDIR_WRITE] + ts->total_io_u[DDIR_TRIM])) + if (!ddir_rw_sum(ts->io_bytes) && !ddir_rw_sum(ts->total_io_u)) return; time(&time_p); @@ -563,9 +533,7 @@ void show_thread_status(struct thread_stat *ts, struct group_run_stats *rs) if (ts->io_bytes[DDIR_TRIM]) show_ddir_status(rs, ts, DDIR_TRIM); - stat_calc_lat_u(ts, io_u_lat_u); - stat_calc_lat_m(ts, io_u_lat_m); - show_latencies(io_u_lat_u, io_u_lat_m); + show_latencies(ts); runtime = ts->total_run_time; if (runtime) { @@ -578,10 +546,13 @@ void show_thread_status(struct thread_stat *ts, struct group_run_stats *rs) sys_cpu = 0; } - log_info(" cpu : usr=%3.2f%%, sys=%3.2f%%, ctx=%lu, majf=%lu," - " minf=%lu\n", usr_cpu, sys_cpu, ts->ctx, ts->majf, ts->minf); + log_info(" cpu : usr=%3.2f%%, sys=%3.2f%%, ctx=%llu," + " majf=%llu, minf=%llu\n", usr_cpu, sys_cpu, + (unsigned long long) ts->ctx, + (unsigned long long) ts->majf, + (unsigned long long) ts->minf); - stat_calc_dist(ts->io_u_map, ts_total_io_u(ts), io_u_dist); + stat_calc_dist(ts->io_u_map, ddir_rw_sum(ts->total_io_u), io_u_dist); log_info(" IO depths : 1=%3.1f%%, 2=%3.1f%%, 4=%3.1f%%, 8=%3.1f%%," " 16=%3.1f%%, 32=%3.1f%%, >=64=%3.1f%%\n", io_u_dist[0], io_u_dist[1], io_u_dist[2], @@ -600,18 +571,31 @@ void show_thread_status(struct thread_stat *ts, struct group_run_stats *rs) 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_info(" issued : total=r=%lu/w=%lu/d=%lu," - " short=r=%lu/w=%lu/d=%lu\n", - ts->total_io_u[0], ts->total_io_u[1], - ts->total_io_u[2], - ts->short_io_u[0], ts->short_io_u[1], - ts->short_io_u[2]); + log_info(" issued : total=r=%llu/w=%llu/d=%llu," + " short=r=%llu/w=%llu/d=%llu," + " drop=r=%llu/w=%llu/d=%llu\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->short_io_u[0], + (unsigned long long) ts->short_io_u[1], + (unsigned long long) ts->short_io_u[2], + (unsigned long long) ts->drop_io_u[0], + (unsigned long long) ts->drop_io_u[1], + (unsigned long long) ts->drop_io_u[2]); if (ts->continue_on_error) { - log_info(" errors : total=%lu, first_error=%d/<%s>\n", - ts->total_err_count, + log_info(" errors : total=%llu, first_error=%d/<%s>\n", + (unsigned long long)ts->total_err_count, ts->first_error, strerror(ts->first_error)); } + if (ts->latency_depth) { + log_info(" latency : target=%llu, window=%llu, percentile=%.2f%%, depth=%u\n", + (unsigned long long)ts->latency_target, + (unsigned long long)ts->latency_window, + ts->latency_percentile.u.f, + ts->latency_depth); + } } static void show_ddir_status_terse(struct thread_stat *ts, @@ -634,8 +618,9 @@ static void show_ddir_status_terse(struct thread_stat *ts, iops = (1000 * (uint64_t) ts->total_io_u[ddir]) / runt; } - log_info(";%llu;%llu;%llu;%llu", ts->io_bytes[ddir] >> 10, bw, iops, - ts->runtime[ddir]); + log_info(";%llu;%llu;%llu;%llu", + (unsigned long long) ts->io_bytes[ddir] >> 10, bw, iops, + (unsigned long long) ts->runtime[ddir]); if (calc_lat(&ts->slat_stat[ddir], &min, &max, &mean, &dev)) log_info(";%lu;%lu;%f;%f", min, max, mean, dev); @@ -720,6 +705,9 @@ static void add_ddir_status_json(struct thread_stat *ts, json_object_add_value_int(dir_object, "bw", bw); json_object_add_value_int(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; @@ -775,7 +763,7 @@ static void add_ddir_status_json(struct thread_stat *ts, if (ovals) free(ovals); - if (!calc_lat(&ts->bw_stat[ddir], &min, &max, &mean, &dev)) { + if (calc_lat(&ts->bw_stat[ddir], &min, &max, &mean, &dev)) { if (rs->agg[ddir]) { p_of_agg = mean * 100 / (double) rs->agg[ddir]; if (p_of_agg > 100.0) @@ -787,13 +775,13 @@ static void add_ddir_status_json(struct thread_stat *ts, } 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", mean); + json_object_add_value_float(dir_object, "bw_agg", p_of_agg); json_object_add_value_float(dir_object, "bw_mean", mean); json_object_add_value_float(dir_object, "bw_dev", dev); } static void show_thread_status_terse_v2(struct thread_stat *ts, - struct group_run_stats *rs) + struct group_run_stats *rs) { double io_u_dist[FIO_IO_U_MAP_NR]; double io_u_lat_u[FIO_IO_U_LAT_U_NR]; @@ -821,11 +809,13 @@ static void show_thread_status_terse_v2(struct thread_stat *ts, sys_cpu = 0; } - log_info(";%f%%;%f%%;%lu;%lu;%lu", usr_cpu, sys_cpu, ts->ctx, ts->majf, - ts->minf); + log_info(";%f%%;%f%%;%llu;%llu;%llu", usr_cpu, sys_cpu, + (unsigned long long) ts->ctx, + (unsigned long long) ts->majf, + (unsigned long long) ts->minf); /* Calc % distribution of IO depths, usecond, msecond latency */ - stat_calc_dist(ts->io_u_map, ts_total_io_u(ts), io_u_dist); + stat_calc_dist(ts->io_u_map, ddir_rw_sum(ts->total_io_u), io_u_dist); stat_calc_lat_u(ts, io_u_lat_u); stat_calc_lat_m(ts, io_u_lat_m); @@ -842,11 +832,11 @@ static void show_thread_status_terse_v2(struct thread_stat *ts, log_info(";%3.2f%%", io_u_lat_m[i]); /* Additional output if continue_on_error set - default off*/ if (ts->continue_on_error) - log_info(";%lu;%d", ts->total_err_count, ts->first_error); + log_info(";%llu;%d", (unsigned long long) ts->total_err_count, ts->first_error); log_info("\n"); /* Additional output if description is set */ - if (ts->description) + if (strlen(ts->description)) log_info(";%s", ts->description); log_info("\n"); @@ -883,11 +873,13 @@ static void show_thread_status_terse_v3_v4(struct thread_stat *ts, sys_cpu = 0; } - log_info(";%f%%;%f%%;%lu;%lu;%lu", usr_cpu, sys_cpu, ts->ctx, ts->majf, - ts->minf); + log_info(";%f%%;%f%%;%llu;%llu;%llu", usr_cpu, sys_cpu, + (unsigned long long) ts->ctx, + (unsigned long long) ts->majf, + (unsigned long long) ts->minf); /* Calc % distribution of IO depths, usecond, msecond latency */ - stat_calc_dist(ts->io_u_map, ts_total_io_u(ts), io_u_dist); + stat_calc_dist(ts->io_u_map, ddir_rw_sum(ts->total_io_u), io_u_dist); stat_calc_lat_u(ts, io_u_lat_u); stat_calc_lat_m(ts, io_u_lat_m); @@ -904,11 +896,12 @@ static void show_thread_status_terse_v3_v4(struct thread_stat *ts, log_info(";%3.2f%%", io_u_lat_m[i]); /* disk util stats, if any */ - show_disk_util(1, NULL); + if (is_backend) + show_disk_util(1, NULL); /* Additional output if continue_on_error set - default off*/ if (ts->continue_on_error) - log_info(";%lu;%d", ts->total_err_count, ts->first_error); + log_info(";%llu;%d", (unsigned long long) ts->total_err_count, ts->first_error); /* Additional output if description is set */ if (strlen(ts->description)) @@ -954,7 +947,7 @@ static struct json_object *show_thread_status_json(struct thread_stat *ts, /* Calc % distribution of IO depths, usecond, msecond latency */ - stat_calc_dist(ts->io_u_map, ts_total_io_u(ts), io_u_dist); + stat_calc_dist(ts->io_u_map, ddir_rw_sum(ts->total_io_u), io_u_dist); stat_calc_lat_u(ts, io_u_lat_u); stat_calc_lat_m(ts, io_u_lat_m); @@ -991,7 +984,14 @@ static struct json_object *show_thread_status_json(struct thread_stat *ts, /* Additional output if continue_on_error set - default off*/ if (ts->continue_on_error) { json_object_add_value_int(root, "total_err", ts->total_err_count); - json_object_add_value_int(root, "total_err", ts->first_error); + json_object_add_value_int(root, "first_error", ts->first_error); + } + + if (ts->latency_depth) { + json_object_add_value_int(root, "latency_depth", ts->latency_depth); + json_object_add_value_int(root, "latency_target", ts->latency_target); + json_object_add_value_float(root, "latency_percentile", ts->latency_percentile.u.f); + json_object_add_value_int(root, "latency_window", ts->latency_window); } /* Additional output if description is set */ @@ -1012,6 +1012,18 @@ static void show_thread_status_terse(struct thread_stat *ts, log_err("fio: bad terse version!? %d\n", terse_version); } +struct json_object *show_thread_status(struct thread_stat *ts, + struct group_run_stats *rs) +{ + if (output_format == FIO_OUTPUT_TERSE) + show_thread_status_terse(ts, rs); + else if (output_format == FIO_OUTPUT_JSON) + return show_thread_status_json(ts, rs); + else + show_thread_status_normal(ts, rs); + return NULL; +} + static void sum_stat(struct io_stat *dst, struct io_stat *src, int nr) { double mean, S; @@ -1116,9 +1128,11 @@ void sum_thread_stats(struct thread_stat *dst, struct thread_stat *src, int nr) if (!dst->unified_rw_rep) { dst->total_io_u[k] += src->total_io_u[k]; dst->short_io_u[k] += src->short_io_u[k]; + dst->drop_io_u[k] += src->drop_io_u[k]; } else { dst->total_io_u[0] += src->total_io_u[k]; dst->short_io_u[0] += src->short_io_u[k]; + dst->drop_io_u[0] += src->drop_io_u[k]; } } @@ -1162,13 +1176,14 @@ void init_thread_stat(struct thread_stat *ts) ts->groupid = -1; } -void show_run_stats(void) +void __show_run_stats(void) { struct group_run_stats *runstats, *rs; struct thread_data *td; struct thread_stat *threadstats, *ts; int i, j, nr_ts, last_ts, idx; int kb_base_warned = 0; + int unit_base_warned = 0; struct json_object *root = NULL; struct json_array *array = NULL; @@ -1225,13 +1240,18 @@ void show_run_stats(void) /* * These are per-group shared already */ - strncpy(ts->name, td->o.name, FIO_JOBNAME_SIZE); + strncpy(ts->name, td->o.name, FIO_JOBNAME_SIZE - 1); if (td->o.description) strncpy(ts->description, td->o.description, - FIO_JOBNAME_SIZE); + FIO_JOBDESC_SIZE - 1); else - memset(ts->description, 0, FIO_JOBNAME_SIZE); + memset(ts->description, 0, FIO_JOBDESC_SIZE); + /* + * If multiple entries in this group, this is + * the first member. + */ + ts->thread_number = td->thread_number; ts->groupid = td->groupid; /* @@ -1240,11 +1260,16 @@ void show_run_stats(void) ts->pid = td->pid; ts->kb_base = td->o.kb_base; + ts->unit_base = td->o.unit_base; 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; + } 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; } ts->continue_on_error = td->o.continue_on_error; @@ -1254,13 +1279,20 @@ void show_run_stats(void) if (!td->error && td->o.continue_on_error && td->first_error) { ts->error = td->first_error; - strcpy(ts->verror, td->verror); + ts->verror[sizeof(ts->verror) - 1] = '\0'; + strncpy(ts->verror, td->verror, sizeof(ts->verror) - 1); } else if (td->error) { ts->error = td->error; - strcpy(ts->verror, td->verror); + ts->verror[sizeof(ts->verror) - 1] = '\0'; + strncpy(ts->verror, td->verror, sizeof(ts->verror) - 1); } } + ts->latency_depth = td->latency_qd; + ts->latency_target = td->o.latency_target; + ts->latency_percentile = td->o.latency_percentile; + ts->latency_window = td->o.latency_window; + sum_thread_stats(ts, &td->ts, idx); } @@ -1270,6 +1302,7 @@ void show_run_stats(void) ts = &threadstats[i]; rs = &runstats[ts->groupid]; rs->kb_base = ts->kb_base; + rs->unit_base = ts->unit_base; rs->unified_rw_rep += ts->unified_rw_rep; for (j = 0; j < DDIR_RWDIR_CNT; j++) { @@ -1333,7 +1366,7 @@ void show_run_stats(void) struct json_object *tmp = show_thread_status_json(ts, rs); json_array_add_value_object(array, tmp); } else - show_thread_status(ts, rs); + show_thread_status_normal(ts, rs); } if (output_format == FIO_OUTPUT_JSON) { /* disk util stats, if any */ @@ -1358,15 +1391,33 @@ void show_run_stats(void) if (is_backend) fio_server_send_du(); - else if (output_format == FIO_OUTPUT_NORMAL) + else if (output_format == FIO_OUTPUT_NORMAL) { show_disk_util(0, NULL); + show_idle_prof_stats(FIO_OUTPUT_NORMAL, NULL); + } - show_idle_prof_stats(FIO_OUTPUT_NORMAL, NULL); + if ( !(output_format == FIO_OUTPUT_TERSE) && append_terse_output) { + log_info("\nAdditional Terse Output:\n"); + for (i = 0; i < nr_ts; i++) { + ts = &threadstats[i]; + rs = &runstats[ts->groupid]; + show_thread_status_terse(ts, rs); + } + } + + log_info_flush(); free(runstats); free(threadstats); } +void show_run_stats(void) +{ + fio_mutex_down(stat_mutex); + __show_run_stats(); + fio_mutex_up(stat_mutex); +} + static void *__show_running_run_stats(void *arg) { struct thread_data *td; @@ -1374,6 +1425,8 @@ static void *__show_running_run_stats(void *arg) struct timeval tv; int i; + fio_mutex_down(stat_mutex); + rt = malloc(thread_number * sizeof(unsigned long long)); fio_gettime(&tv, NULL); @@ -1386,14 +1439,22 @@ static void *__show_running_run_stats(void *arg) if (td_trim(td) && td->io_bytes[DDIR_TRIM]) td->ts.runtime[DDIR_TRIM] += rt[i]; - update_rusage_stat(td); + td->update_rusage = 1; td->ts.io_bytes[DDIR_READ] = td->io_bytes[DDIR_READ]; td->ts.io_bytes[DDIR_WRITE] = td->io_bytes[DDIR_WRITE]; td->ts.io_bytes[DDIR_TRIM] = td->io_bytes[DDIR_TRIM]; td->ts.total_run_time = mtime_since(&td->epoch, &tv); } - show_run_stats(); + for_each_td(td, i) { + if (td->rusage_sem) { + td->update_rusage = 1; + fio_mutex_down(td->rusage_sem); + } + td->update_rusage = 0; + } + + __show_run_stats(); for_each_td(td, i) { if (td_read(td) && td->io_bytes[DDIR_READ]) @@ -1405,6 +1466,8 @@ static void *__show_running_run_stats(void *arg) } free(rt); + fio_mutex_up(stat_mutex); + free(arg); return NULL; } @@ -1415,10 +1478,80 @@ static void *__show_running_run_stats(void *arg) */ void show_running_run_stats(void) { - pthread_t thread; + pthread_t *thread; + + thread = calloc(1, sizeof(*thread)); + if (!thread) + return; + + if (!pthread_create(thread, NULL, __show_running_run_stats, thread)) { + int err; + + err = pthread_detach(*thread); + if (err) + log_err("fio: DU thread detach failed: %s\n", strerror(err)); + + return; + } + + free(thread); +} + +static int status_interval_init; +static struct timeval status_time; +static int status_file_disabled; + +#define FIO_STATUS_FILE "fio-dump-status" + +static int check_status_file(void) +{ + struct stat sb; + const char *temp_dir; + char fio_status_file_path[PATH_MAX]; + + if (status_file_disabled) + return 0; + + temp_dir = getenv("TMPDIR"); + if (temp_dir == NULL) { + temp_dir = getenv("TEMP"); + if (temp_dir && strlen(temp_dir) >= PATH_MAX) + temp_dir = NULL; + } + if (temp_dir == NULL) + temp_dir = "/tmp"; - pthread_create(&thread, NULL, __show_running_run_stats, NULL); - pthread_detach(thread); + snprintf(fio_status_file_path, sizeof(fio_status_file_path), "%s/%s", temp_dir, FIO_STATUS_FILE); + + if (stat(fio_status_file_path, &sb)) + return 0; + + if (unlink(fio_status_file_path) < 0) { + 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; + } + + return 1; +} + +void check_for_running_stats(void) +{ + if (status_interval) { + if (!status_interval_init) { + fio_gettime(&status_time, NULL); + status_interval_init = 1; + } 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; + } } static inline void add_stat_sample(struct io_stat *is, unsigned long data) @@ -1442,24 +1575,55 @@ static inline void add_stat_sample(struct io_stat *is, unsigned long data) static void __add_log_sample(struct io_log *iolog, unsigned long val, enum fio_ddir ddir, unsigned int bs, - unsigned long t) + unsigned long t, uint64_t offset) { - const int nr_samples = iolog->nr_samples; + uint64_t nr_samples = iolog->nr_samples; + struct io_sample *s; + + if (iolog->disabled) + return; if (!iolog->nr_samples) iolog->avg_last = t; if (iolog->nr_samples == iolog->max_samples) { - int new_size = sizeof(struct io_sample) * iolog->max_samples*2; + size_t new_size; + void *new_log; + + new_size = 2 * iolog->max_samples * log_entry_sz(iolog); + + if (iolog->log_gz && (new_size > iolog->log_gz)) { + if (iolog_flush(iolog, 0)) { + log_err("fio: failed flushing iolog! Will stop logging.\n"); + iolog->disabled = 1; + return; + } + nr_samples = iolog->nr_samples; + } else { + new_log = realloc(iolog->log, new_size); + if (!new_log) { + log_err("fio: failed extending iolog! Will stop logging.\n"); + iolog->disabled = 1; + return; + } + iolog->log = new_log; + iolog->max_samples <<= 1; + } + } + + s = get_sample(iolog, nr_samples); + + s->val = val; + s->time = t; + io_sample_set_ddir(iolog, s, ddir); + s->bs = bs; - iolog->log = realloc(iolog->log, new_size); - iolog->max_samples <<= 1; + if (iolog->log_offset) { + struct io_sample_offset *so = (void *) s; + + so->offset = offset; } - iolog->log[nr_samples].val = val; - iolog->log[nr_samples].time = t; - iolog->log[nr_samples].ddir = ddir; - iolog->log[nr_samples].bs = bs; iolog->nr_samples++; } @@ -1469,9 +1633,76 @@ static inline void reset_io_stat(struct io_stat *ios) 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; + + for (i = 0; i < DDIR_RWDIR_CNT; i++) { + reset_io_stat(&ts->clat_stat[i]); + reset_io_stat(&ts->slat_stat[i]); + reset_io_stat(&ts->lat_stat[i]); + reset_io_stat(&ts->bw_stat[i]); + reset_io_stat(&ts->iops_stat[i]); + + ts->io_bytes[i] = 0; + ts->runtime[i] = 0; + + for (j = 0; j < FIO_IO_U_PLAT_NR; j++) + ts->io_u_plat[i][j] = 0; + } + + for (i = 0; i < FIO_IO_U_MAP_NR; i++) { + ts->io_u_map[i] = 0; + ts->io_u_submit[i] = 0; + ts->io_u_complete[i] = 0; + ts->io_u_lat_u[i] = 0; + ts->io_u_lat_m[i] = 0; + ts->total_submit = 0; + ts->total_complete = 0; + } + + for (i = 0; i < 3; i++) { + ts->total_io_u[i] = 0; + ts->short_io_u[i] = 0; + ts->drop_io_u[i] = 0; + } +} + +static void _add_stat_to_log(struct io_log *iolog, unsigned long elapsed) +{ + /* + * Note an entry in the log. Use the mean from the logged samples, + * making sure to properly round up. Only write a log entry if we + * had actual samples done. + */ + if (iolog->avg_window[DDIR_READ].samples) { + unsigned long mr; + + mr = iolog->avg_window[DDIR_READ].mean.u.f + 0.50; + __add_log_sample(iolog, mr, DDIR_READ, 0, elapsed, 0); + } + if (iolog->avg_window[DDIR_WRITE].samples) { + unsigned long mw; + + mw = iolog->avg_window[DDIR_WRITE].mean.u.f + 0.50; + __add_log_sample(iolog, mw, DDIR_WRITE, 0, elapsed, 0); + } + if (iolog->avg_window[DDIR_TRIM].samples) { + unsigned long mw; + + mw = iolog->avg_window[DDIR_TRIM].mean.u.f + 0.50; + __add_log_sample(iolog, mw, DDIR_TRIM, 0, elapsed, 0); + } + + reset_io_stat(&iolog->avg_window[DDIR_READ]); + reset_io_stat(&iolog->avg_window[DDIR_WRITE]); + reset_io_stat(&iolog->avg_window[DDIR_TRIM]); +} + static void add_log_sample(struct thread_data *td, struct io_log *iolog, unsigned long val, enum fio_ddir ddir, - unsigned int bs) + unsigned int bs, uint64_t offset) { unsigned long elapsed, this_window; @@ -1484,7 +1715,7 @@ static void add_log_sample(struct thread_data *td, struct io_log *iolog, * If no time averaging, just add the log sample. */ if (!iolog->avg_msec) { - __add_log_sample(iolog, val, ddir, bs, elapsed); + __add_log_sample(iolog, val, ddir, bs, elapsed, offset); return; } @@ -1502,35 +1733,27 @@ static void add_log_sample(struct thread_data *td, struct io_log *iolog, if (this_window < iolog->avg_msec) return; - /* - * Note an entry in the log. Use the mean from the logged samples, - * making sure to properly round up. Only write a log entry if we - * had actual samples done. - */ - if (iolog->avg_window[DDIR_READ].samples) { - unsigned long mr; - - mr = iolog->avg_window[DDIR_READ].mean.u.f + 0.50; - __add_log_sample(iolog, mr, DDIR_READ, 0, elapsed); - } - if (iolog->avg_window[DDIR_WRITE].samples) { - unsigned long mw; + _add_stat_to_log(iolog, elapsed); - mw = iolog->avg_window[DDIR_WRITE].mean.u.f + 0.50; - __add_log_sample(iolog, mw, DDIR_WRITE, 0, elapsed); - } - if (iolog->avg_window[DDIR_TRIM].samples) { - unsigned long mw; + iolog->avg_last = elapsed; +} - mw = iolog->avg_window[DDIR_TRIM].mean.u.f + 0.50; - __add_log_sample(iolog, mw, DDIR_TRIM, 0, elapsed); - } +void finalize_logs(struct thread_data *td) +{ + unsigned long elapsed; + elapsed = mtime_since_now(&td->epoch); - reset_io_stat(&iolog->avg_window[DDIR_READ]); - reset_io_stat(&iolog->avg_window[DDIR_WRITE]); - reset_io_stat(&iolog->avg_window[DDIR_TRIM]); - iolog->avg_last = elapsed; + if (td->clat_log) + _add_stat_to_log(td->clat_log, elapsed); + if (td->slat_log) + _add_stat_to_log(td->slat_log, elapsed); + if (td->lat_log) + _add_stat_to_log(td->lat_log, elapsed); + if (td->bw_log) + _add_stat_to_log(td->bw_log, elapsed); + if (td->iops_log) + _add_stat_to_log(td->iops_log, elapsed); } void add_agg_sample(unsigned long val, enum fio_ddir ddir, unsigned int bs) @@ -1541,7 +1764,7 @@ void add_agg_sample(unsigned long val, enum fio_ddir ddir, unsigned int bs) return; iolog = agg_io_log[ddir]; - __add_log_sample(iolog, val, ddir, bs, mtime_since_genesis()); + __add_log_sample(iolog, val, ddir, bs, mtime_since_genesis(), 0); } static void add_clat_percentile_sample(struct thread_stat *ts, @@ -1554,7 +1777,7 @@ static void add_clat_percentile_sample(struct thread_stat *ts, } void add_clat_sample(struct thread_data *td, enum fio_ddir ddir, - unsigned long usec, unsigned int bs) + unsigned long usec, unsigned int bs, uint64_t offset) { struct thread_stat *ts = &td->ts; @@ -1564,14 +1787,14 @@ void add_clat_sample(struct thread_data *td, enum fio_ddir ddir, add_stat_sample(&ts->clat_stat[ddir], usec); if (td->clat_log) - add_log_sample(td, td->clat_log, usec, ddir, bs); + add_log_sample(td, td->clat_log, usec, ddir, bs, offset); if (ts->clat_percentiles) add_clat_percentile_sample(ts, usec, ddir); } void add_slat_sample(struct thread_data *td, enum fio_ddir ddir, - unsigned long usec, unsigned int bs) + unsigned long usec, unsigned int bs, uint64_t offset) { struct thread_stat *ts = &td->ts; @@ -1581,11 +1804,11 @@ void add_slat_sample(struct thread_data *td, enum fio_ddir ddir, add_stat_sample(&ts->slat_stat[ddir], usec); if (td->slat_log) - add_log_sample(td, td->slat_log, usec, ddir, bs); + add_log_sample(td, td->slat_log, usec, ddir, bs, offset); } void add_lat_sample(struct thread_data *td, enum fio_ddir ddir, - unsigned long usec, unsigned int bs) + unsigned long usec, unsigned int bs, uint64_t offset) { struct thread_stat *ts = &td->ts; @@ -1595,7 +1818,7 @@ void add_lat_sample(struct thread_data *td, enum fio_ddir ddir, add_stat_sample(&ts->lat_stat[ddir], usec); if (td->lat_log) - add_log_sample(td, td->lat_log, usec, ddir, bs); + add_log_sample(td, td->lat_log, usec, ddir, bs, offset); } void add_bw_sample(struct thread_data *td, enum fio_ddir ddir, unsigned int bs, @@ -1621,11 +1844,15 @@ void add_bw_sample(struct thread_data *td, enum fio_ddir ddir, unsigned int bs, if (!delta) continue; /* No entries for interval */ - rate = delta * 1000 / spent / 1024; + if (spent) + rate = delta * 1000 / spent / 1024; + else + rate = 0; + add_stat_sample(&ts->bw_stat[ddir], rate); if (td->bw_log) - add_log_sample(td, td->bw_log, rate, ddir, bs); + add_log_sample(td, td->bw_log, rate, ddir, bs, 0); td->stat_io_bytes[ddir] = td->this_io_bytes[ddir]; } @@ -1633,7 +1860,7 @@ void add_bw_sample(struct thread_data *td, enum fio_ddir ddir, unsigned int bs, fio_gettime(&td->bw_sample_time, NULL); } -void add_iops_sample(struct thread_data *td, enum fio_ddir ddir, +void add_iops_sample(struct thread_data *td, enum fio_ddir ddir, unsigned int bs, struct timeval *t) { struct thread_stat *ts = &td->ts; @@ -1656,14 +1883,33 @@ void add_iops_sample(struct thread_data *td, enum fio_ddir ddir, if (!delta) continue; /* No entries for interval */ - iops = (delta * 1000) / spent; + if (spent) + iops = (delta * 1000) / spent; + else + iops = 0; + add_stat_sample(&ts->iops_stat[ddir], iops); if (td->iops_log) - add_log_sample(td, td->iops_log, iops, ddir, 0); + add_log_sample(td, td->iops_log, iops, ddir, bs, 0); td->stat_io_blocks[ddir] = td->this_io_blocks[ddir]; } fio_gettime(&td->iops_sample_time, NULL); } + +void stat_init(void) +{ + stat_mutex = fio_mutex_init(FIO_MUTEX_UNLOCKED); +} + +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); +}