X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=options.c;h=53c3a82674f1821961c41ec8fb15cd44f29ced12;hp=c273da761457435fef000668e2167cc0e50418d0;hb=1647f592cc93364e018ed149cb1a5b6a7dc598b9;hpb=86375e3446fdeb6c4dd1406cc0e4c456af954a2c diff --git a/options.c b/options.c index c273da76..53c3a826 100644 --- a/options.c +++ b/options.c @@ -3,7 +3,6 @@ #include #include #include -#include #include #include #include @@ -14,8 +13,9 @@ #include "verify.h" #include "parse.h" #include "lib/fls.h" +#include "options.h" -#define td_var_offset(var) ((size_t) &((struct thread_options *)0)->var) +#include "crc/crc32c.h" /* * Check if mmap/mmaphuge has a :/foo/bar/file at the end. If so, return that. @@ -33,6 +33,26 @@ static char *get_opt_postfix(const char *str) return strdup(p); } +static int converthexchartoint(char a) +{ + int base; + + switch(a) { + case '0'...'9': + base = '0'; + break; + case 'A'...'F': + base = 'A' - 10; + break; + case 'a'...'f': + base = 'a' - 10; + break; + default: + base = 0; + } + return (a - base); +} + static int bs_cmp(const void *p1, const void *p2) { const struct bssplit *bsp1 = p1; @@ -82,7 +102,7 @@ static int bssplit_ddir(struct thread_data *td, int ddir, char *str) } else perc = -1; - if (str_to_decimal(fname, &val, 1)) { + if (str_to_decimal(fname, &val, 1, td)) { log_err("fio: bssplit conversion failed\n"); free(td->o.bssplit); return 1; @@ -182,12 +202,27 @@ static int str_rw_cb(void *data, const char *str) struct thread_data *td = data; char *nr = get_opt_postfix(str); - td->o.ddir_nr = 1; - if (nr) { - td->o.ddir_nr = atoi(nr); - free(nr); + td->o.ddir_seq_nr = 1; + td->o.ddir_seq_add = 0; + + if (!nr) + return 0; + + if (td_random(td)) + td->o.ddir_seq_nr = atoi(nr); + else { + long long val; + + if (str_to_decimal(nr, &val, 1, td)) { + log_err("fio: rw postfix parsing failed\n"); + free(nr); + return 1; + } + + td->o.ddir_seq_add = val; } + free(nr); return 0; } @@ -206,13 +241,37 @@ static int str_mem_cb(void *data, const char *mem) return 0; } -static int str_lockmem_cb(void fio_unused *data, unsigned long *val) +static int str_verify_cb(void *data, const char *mem) +{ + struct thread_data *td = data; + + if (td->o.verify != VERIFY_CRC32C_INTEL) + return 0; + + if (!crc32c_intel_works()) { + log_info("fio: System does not support hw accelerated crc32c. Falling back to sw crc32c.\n"); + td->o.verify = VERIFY_CRC32C; + } + + return 0; +} + +static int fio_clock_source_cb(void *data, const char *str) +{ + struct thread_data *td = data; + + fio_clock_source = td->o.clocksource; + fio_time_init(); + return 0; +} + +static int str_lockmem_cb(void fio_unused *data, unsigned long long *val) { mlock_size = *val; return 0; } -static int str_rwmix_read_cb(void *data, unsigned int *val) +static int str_rwmix_read_cb(void *data, unsigned long long *val) { struct thread_data *td = data; @@ -221,7 +280,7 @@ static int str_rwmix_read_cb(void *data, unsigned int *val) return 0; } -static int str_rwmix_write_cb(void *data, unsigned int *val) +static int str_rwmix_write_cb(void *data, unsigned long long *val) { struct thread_data *td = data; @@ -231,7 +290,7 @@ static int str_rwmix_write_cb(void *data, unsigned int *val) } #ifdef FIO_HAVE_IOPRIO -static int str_prioclass_cb(void *data, unsigned int *val) +static int str_prioclass_cb(void *data, unsigned long long *val) { struct thread_data *td = data; unsigned short mask; @@ -247,7 +306,7 @@ static int str_prioclass_cb(void *data, unsigned int *val) return 0; } -static int str_prio_cb(void *data, unsigned int *val) +static int str_prio_cb(void *data, unsigned long long *val) { struct thread_data *td = data; @@ -271,7 +330,7 @@ static int str_exitall_cb(void) } #ifdef FIO_HAVE_CPU_AFFINITY -static int str_cpumask_cb(void *data, unsigned int *val) +static int str_cpumask_cb(void *data, unsigned long long *val) { struct thread_data *td = data; unsigned int i; @@ -285,7 +344,7 @@ static int str_cpumask_cb(void *data, unsigned int *val) return 1; } - max_cpu = sysconf(_SC_NPROCESSORS_ONLN); + max_cpu = cpus_online(); for (i = 0; i < sizeof(int) * 8; i++) { if ((1 << i) & *val) { @@ -322,7 +381,7 @@ static int set_cpus_allowed(struct thread_data *td, os_cpu_mask_t *mask, strip_blank_front(&str); strip_blank_end(str); - max_cpu = sysconf(_SC_NPROCESSORS_ONLN); + max_cpu = cpus_online(); while ((cpu = strsep(&str, ",")) != NULL) { char *str2, *cpu2; @@ -356,7 +415,7 @@ static int set_cpus_allowed(struct thread_data *td, os_cpu_mask_t *mask, ret = 1; break; } - + dprint(FD_PARSE, "set cpu allowed %d\n", icpu); fio_cpu_set(mask, icpu); icpu++; @@ -396,6 +455,16 @@ static int str_verify_cpus_allowed_cb(void *data, const char *input) } #endif +#ifdef FIO_HAVE_TRIM +static int str_verify_trim_cb(void *data, unsigned long long *val) +{ + struct thread_data *td = data; + + td->o.trim_percentage = *val; + return 0; +} +#endif + static int str_fst_cb(void *data, const char *str) { struct thread_data *td = data; @@ -410,8 +479,25 @@ static int str_fst_cb(void *data, const char *str) return 0; } +#ifdef FIO_HAVE_SYNC_FILE_RANGE +static int str_sfr_cb(void *data, const char *str) +{ + struct thread_data *td = data; + char *nr = get_opt_postfix(str); + + td->sync_file_range_nr = 1; + if (nr) { + td->sync_file_range_nr = atoi(nr); + free(nr); + } + + return 0; +} +#endif + static int check_dir(struct thread_data *td, char *fname) { +#if 0 char file[PATH_MAX], *dir; int elen = 0; @@ -424,7 +510,6 @@ static int check_dir(struct thread_data *td, char *fname) sprintf(file + elen, "%s", fname); dir = dirname(file); -#if 0 { struct stat sb; /* @@ -449,6 +534,52 @@ static int check_dir(struct thread_data *td, char *fname) return 0; } +/* + * Return next file in the string. Files are separated with ':'. If the ':' + * is escaped with a '\', then that ':' is part of the filename and does not + * indicate a new file. + */ +static char *get_next_file_name(char **ptr) +{ + char *str = *ptr; + char *p, *start; + + if (!str || !strlen(str)) + return NULL; + + start = str; + do { + /* + * No colon, we are done + */ + p = strchr(str, ':'); + if (!p) { + *ptr = NULL; + break; + } + + /* + * We got a colon, but it's the first character. Skip and + * continue + */ + if (p == start) { + str = ++start; + continue; + } + + if (*(p - 1) != '\\') { + *p = '\0'; + *ptr = p + 1; + break; + } + + memmove(p - 1, p, strlen(p) + 1); + str = p; + } while (1); + + return start; +} + static int str_filename_cb(void *data, const char *input) { struct thread_data *td = data; @@ -462,7 +593,7 @@ static int str_filename_cb(void *data, const char *input) if (!td->files_index) td->o.nr_files = 0; - while ((fname = strsep(&str, ":")) != NULL) { + while ((fname = get_next_file_name(&str)) != NULL) { if (!strlen(fname)) break; if (check_dir(td, fname)) { @@ -507,7 +638,7 @@ static int str_opendir_cb(void *data, const char fio_unused *str) return add_dir_files(td, td->o.opendir); } -static int str_verify_offset_cb(void *data, unsigned int *off) +static int str_verify_offset_cb(void *data, unsigned long long *off) { struct thread_data *td = data; @@ -520,22 +651,50 @@ static int str_verify_offset_cb(void *data, unsigned int *off) return 0; } -static int str_verify_pattern_cb(void *data, unsigned int *off) +static int str_verify_pattern_cb(void *data, const char *input) { struct thread_data *td = data; - unsigned int msb; - - msb = __fls(*off); - if (msb <= 8) - td->o.verify_pattern_bytes = 1; - else if (msb <= 16) - td->o.verify_pattern_bytes = 2; - else if (msb <= 24) - td->o.verify_pattern_bytes = 3; - else - td->o.verify_pattern_bytes = 4; - - td->o.verify_pattern = *off; + long off; + int i = 0, j = 0, len, k, base = 10; + char* loc1, * loc2; + + loc1 = strstr(input, "0x"); + loc2 = strstr(input, "0X"); + if (loc1 || loc2) + base = 16; + off = strtol(input, NULL, base); + if (off != LONG_MAX || errno != ERANGE) { + while (off) { + td->o.verify_pattern[i] = off & 0xff; + off >>= 8; + i++; + } + } else { + len = strlen(input); + k = len - 1; + if (base == 16) { + if (loc1) + j = loc1 - input + 2; + else + j = loc2 - input + 2; + } else + return 1; + if (len - j < MAX_PATTERN_SIZE * 2) { + while (k >= j) { + off = converthexchartoint(input[k--]); + if (k >= j) + off += (converthexchartoint(input[k--]) + * 16); + td->o.verify_pattern[i++] = (char) off; + } + } + } + td->o.verify_pattern_bytes = i; + /* + * VERIFY_META could already be set + */ + if (td->o.verify == VERIFY_NONE) + td->o.verify = VERIFY_PATTERN; return 0; } @@ -575,21 +734,34 @@ static int str_write_lat_log_cb(void *data, const char *str) return 0; } +static int str_write_iops_log_cb(void *data, const char *str) +{ + struct thread_data *td = data; + + if (str) + td->o.iops_log_file = strdup(str); + + td->o.write_iops_log = 1; + return 0; +} + static int str_gtod_reduce_cb(void *data, int *il) { struct thread_data *td = data; int val = *il; + td->o.disable_lat = !!val; td->o.disable_clat = !!val; td->o.disable_slat = !!val; td->o.disable_bw = !!val; + td->o.clat_percentiles = !val; if (val) td->tv_cache_mask = 63; return 0; } -static int str_gtod_cpu_cb(void *data, int *il) +static int str_gtod_cpu_cb(void *data, long long *il) { struct thread_data *td = data; int val = *il; @@ -599,6 +771,20 @@ static int str_gtod_cpu_cb(void *data, int *il) return 0; } +static int str_size_cb(void *data, unsigned long long *__val) +{ + struct thread_data *td = data; + unsigned long long v = *__val; + + if (parse_is_percent(v)) { + td->o.size = 0; + td->o.size_percent = -1ULL - v; + } else + td->o.size = v; + + return 0; +} + static int rw_verify(struct fio_option *o, void *data) { struct thread_data *td = data; @@ -640,13 +826,10 @@ static int kb_base_verify(struct fio_option *o, void *data) return 0; } -#define __stringify_1(x) #x -#define __stringify(x) __stringify_1(x) - /* * Map of job/command line options */ -static struct fio_option options[] = { +static struct fio_option options[FIO_MAX_OPTS] = { { .name = "description", .type = FIO_OPT_STR_STORE, @@ -678,9 +861,10 @@ static struct fio_option options[] = { .name = "kb_base", .type = FIO_OPT_INT, .off1 = td_var_offset(kb_base), - .help = "How many bytes per KB for reporting (1000 or 1024)", .verify = kb_base_verify, + .prio = 1, .def = "1024", + .help = "How many bytes per KB for reporting (1000 or 1024)", }, { .name = "lockfile", @@ -749,12 +933,30 @@ static struct fio_option options[] = { }, }, }, + { + .name = "rw_sequencer", + .type = FIO_OPT_STR, + .off1 = td_var_offset(rw_seq), + .help = "IO offset generator modifier", + .def = "sequential", + .posval = { + { .ival = "sequential", + .oval = RW_SEQ_SEQ, + .help = "Generate sequential offsets", + }, + { .ival = "identical", + .oval = RW_SEQ_IDENT, + .help = "Generate identical offsets", + }, + }, + }, + { .name = "ioengine", .type = FIO_OPT_STR_STORE, .off1 = td_var_offset(ioengine), .help = "IO engine to use", - .def = "sync", + .def = FIO_PREFERRED_ENGINE, .posval = { { .ival = "sync", .help = "Use read/write", @@ -763,7 +965,7 @@ static struct fio_option options[] = { .help = "Use pread/pwrite", }, { .ival = "vsync", - .help = "Use readv/writev", + .help = "Use readv/writev", }, #ifdef FIO_HAVE_LIBAIO { .ival = "libaio", @@ -779,9 +981,14 @@ static struct fio_option options[] = { { .ival = "solarisaio", .help = "Solaris native asynchronous IO", }, +#endif +#ifdef FIO_HAVE_WINDOWSAIO + { .ival = "windowsaio", + .help = "Windows native asynchronous IO" + }, #endif { .ival = "mmap", - .help = "Memory mapped IO", + .help = "Memory mapped IO" }, #ifdef FIO_HAVE_SPLICE { .ival = "splice", @@ -808,12 +1015,22 @@ static struct fio_option options[] = { }, #endif { .ival = "cpuio", - .help = "CPU cycler burner engine", + .help = "CPU cycle burner engine", }, #ifdef FIO_HAVE_GUASI { .ival = "guasi", .help = "GUASI IO engine", }, +#endif +#ifdef FIO_HAVE_BINJECT + { .ival = "binject", + .help = "binject direct inject block engine", + }, +#endif +#ifdef FIO_HAVE_RDMA + { .ival = "rdma", + .help = "RDMA IO engine", + }, #endif { .ival = "external", .help = "Load external engine (append name)", @@ -824,7 +1041,7 @@ static struct fio_option options[] = { .name = "iodepth", .type = FIO_OPT_INT, .off1 = td_var_offset(iodepth), - .help = "Amount of IO buffers to keep in flight", + .help = "Number of IO buffers to keep in flight", .minval = 1, .def = "1", }, @@ -857,12 +1074,12 @@ static struct fio_option options[] = { { .name = "size", .type = FIO_OPT_STR_VAL, - .off1 = td_var_offset(size), - .minval = 1, + .cb = str_size_cb, .help = "Total size of device or files", }, { .name = "fill_device", + .alias = "fill_fs", .type = FIO_OPT_BOOL, .off1 = td_var_offset(fill_device), .help = "Write until an ENOSPC error occurs", @@ -940,6 +1157,14 @@ static struct fio_option options[] = { .def = "1", .parent = "rw", }, + { + .name = "use_os_rand", + .type = FIO_OPT_BOOL, + .off1 = td_var_offset(use_os_rand), + .help = "Set to use OS random generator", + .def = "0", + .parent = "rw", + }, { .name = "norandommap", .type = FIO_OPT_STR_SET, @@ -957,6 +1182,7 @@ static struct fio_option options[] = { }, { .name = "nrfiles", + .alias = "nr_files", .type = FIO_OPT_INT, .off1 = td_var_offset(nr_files), .help = "Split job workload between this number of files", @@ -991,6 +1217,40 @@ static struct fio_option options[] = { }, .parent = "nrfiles", }, +#ifdef FIO_HAVE_FALLOCATE + { + .name = "fallocate", + .type = FIO_OPT_STR, + .off1 = td_var_offset(fallocate_mode), + .help = "Whether pre-allocation is performed when laying out files", + .def = "posix", + .posval = { + { .ival = "none", + .oval = FIO_FALLOCATE_NONE, + .help = "Do not pre-allocate space", + }, + { .ival = "posix", + .oval = FIO_FALLOCATE_POSIX, + .help = "Use posix_fallocate()", + }, +#ifdef FIO_HAVE_LINUX_FALLOCATE + { .ival = "keep", + .oval = FIO_FALLOCATE_KEEP_SIZE, + .help = "Use fallocate(..., FALLOC_FL_KEEP_SIZE, ...)", + }, +#endif + /* Compatibility with former boolean values */ + { .ival = "0", + .oval = FIO_FALLOCATE_NONE, + .help = "Alias for 'none'", + }, + { .ival = "1", + .oval = FIO_FALLOCATE_POSIX, + .help = "Alias for 'posix'", + }, + }, + }, +#endif /* FIO_HAVE_FALLOCATE */ { .name = "fadvise_hint", .type = FIO_OPT_BOOL, @@ -1012,6 +1272,40 @@ static struct fio_option options[] = { .help = "Issue fdatasync for writes every given number of blocks", .def = "0", }, + { + .name = "write_barrier", + .type = FIO_OPT_INT, + .off1 = td_var_offset(barrier_blocks), + .help = "Make every Nth write a barrier write", + .def = "0", + }, +#ifdef FIO_HAVE_SYNC_FILE_RANGE + { + .name = "sync_file_range", + .posval = { + { .ival = "wait_before", + .oval = SYNC_FILE_RANGE_WAIT_BEFORE, + .help = "SYNC_FILE_RANGE_WAIT_BEFORE", + .or = 1, + }, + { .ival = "write", + .oval = SYNC_FILE_RANGE_WRITE, + .help = "SYNC_FILE_RANGE_WRITE", + .or = 1, + }, + { + .ival = "wait_after", + .oval = SYNC_FILE_RANGE_WAIT_AFTER, + .help = "SYNC_FILE_RANGE_WAIT_AFTER", + .or = 1, + }, + }, + .type = FIO_OPT_STR_MULTI, + .cb = str_sfr_cb, + .off1 = td_var_offset(sync_file_range), + .help = "Use sync_file_range()", + }, +#endif { .name = "direct", .type = FIO_OPT_BOOL, @@ -1050,7 +1344,7 @@ static struct fio_option options[] = { }, { .name = "startdelay", - .type = FIO_OPT_INT, + .type = FIO_OPT_STR_VAL_TIME, .off1 = td_var_offset(start_delay), .help = "Only start job when this period has passed", .def = "0", @@ -1075,6 +1369,29 @@ static struct fio_option options[] = { .off1 = td_var_offset(ramp_time), .help = "Ramp up time before measuring performance", }, + { + .name = "clocksource", + .type = FIO_OPT_STR, + .cb = fio_clock_source_cb, + .off1 = td_var_offset(clocksource), + .help = "What type of timing source to use", + .posval = { + { .ival = "gettimeofday", + .oval = CS_GTOD, + .help = "Use gettimeofday(2) for timing", + }, + { .ival = "clock_gettime", + .oval = CS_CGETTIME, + .help = "Use clock_gettime(2) for timing", + }, +#ifdef ARCH_HAVE_CPU_CLOCK + { .ival = "cpu", + .oval = CS_CPUCLOCK, + .help = "Use CPU private clock", + }, +#endif + }, + }, { .name = "mem", .alias = "iomem", @@ -1125,6 +1442,7 @@ static struct fio_option options[] = { .type = FIO_OPT_STR, .off1 = td_var_offset(verify), .help = "Verify data written", + .cb = str_verify_cb, .def = "0", .posval = { { .ival = "0", @@ -1159,6 +1477,10 @@ static struct fio_option options[] = { .oval = VERIFY_CRC7, .help = "Use crc7 checksums for verification", }, + { .ival = "sha1", + .oval = VERIFY_SHA1, + .help = "Use sha1 checksums for verification", + }, { .ival = "sha256", .oval = VERIFY_SHA256, .help = "Use sha256 checksums for verification", @@ -1212,7 +1534,7 @@ static struct fio_option options[] = { }, { .name = "verify_pattern", - .type = FIO_OPT_INT, + .type = FIO_OPT_STR, .cb = str_verify_pattern_cb, .help = "Fill pattern for IO buffers", .parent = "verify", @@ -1225,6 +1547,14 @@ static struct fio_option options[] = { .help = "Exit on a single verify failure, don't continue", .parent = "verify", }, + { + .name = "verify_dump", + .type = FIO_OPT_BOOL, + .off1 = td_var_offset(verify_dump), + .def = "0", + .help = "Dump contents of good and bad blocks on failure", + .parent = "verify", + }, { .name = "verify_async", .type = FIO_OPT_INT, @@ -1233,6 +1563,20 @@ static struct fio_option options[] = { .help = "Number of async verifier threads to use", .parent = "verify", }, + { + .name = "verify_backlog", + .type = FIO_OPT_STR_VAL, + .off1 = td_var_offset(verify_backlog), + .help = "Verify after this number of blocks are written", + .parent = "verify", + }, + { + .name = "verify_backlog_batch", + .type = FIO_OPT_INT, + .off1 = td_var_offset(verify_batch), + .help = "Verify this number of IO blocks", + .parent = "verify", + }, #ifdef FIO_HAVE_CPU_AFFINITY { .name = "verify_async_cpus", @@ -1241,6 +1585,39 @@ static struct fio_option options[] = { .help = "Set CPUs allowed for async verify threads", .parent = "verify_async", }, +#endif +#ifdef FIO_HAVE_TRIM + { + .name = "trim_percentage", + .type = FIO_OPT_INT, + .cb = str_verify_trim_cb, + .maxval = 100, + .help = "Number of verify blocks to discard/trim", + .parent = "verify", + .def = "0", + }, + { + .name = "trim_verify_zero", + .type = FIO_OPT_INT, + .help = "Verify that trim/discarded blocks are returned as zeroes", + .off1 = td_var_offset(trim_zero), + .parent = "trim_percentage", + .def = "1", + }, + { + .name = "trim_backlog", + .type = FIO_OPT_STR_VAL, + .off1 = td_var_offset(trim_backlog), + .help = "Trim after this number of blocks are written", + .parent = "trim_percentage", + }, + { + .name = "trim_backlog_batch", + .type = FIO_OPT_INT, + .off1 = td_var_offset(trim_batch), + .help = "Trim this number of IO blocks", + .parent = "trim_percentage", + }, #endif { .name = "write_iolog", @@ -1254,6 +1631,21 @@ static struct fio_option options[] = { .off1 = td_var_offset(read_iolog_file), .help = "Playback IO pattern from file", }, + { + .name = "replay_no_stall", + .type = FIO_OPT_INT, + .off1 = td_var_offset(no_stall), + .def = "0", + .parent = "read_iolog", + .help = "Playback IO pattern file as fast as possible without stalls", + }, + { + .name = "replay_redirect", + .type = FIO_OPT_STR_STORE, + .off1 = td_var_offset(replay_redirect), + .parent = "read_iolog", + .help = "Replay all I/O onto this device, regardless of trace device", + }, { .name = "exec_prerun", .type = FIO_OPT_STR_STORE, @@ -1392,7 +1784,7 @@ static struct fio_option options[] = { .type = FIO_OPT_INT, .off1 = td_var_offset(rate_iops_min[0]), .off2 = td_var_offset(rate_iops_min[1]), - .help = "Job must meet this rate or it will be shutdown", + .help = "Job must meet this rate or it will be shut down", .parent = "rate_iops", }, { @@ -1425,6 +1817,15 @@ static struct fio_option options[] = { .help = "Time window over which to calculate bandwidth" " (msec)", .def = "500", + .parent = "write_bw_log", + }, + { + .name = "iopsavgtime", + .type = FIO_OPT_INT, + .off1 = td_var_offset(iops_avg_time), + .help = "Time window over which to calculate IOPS (msec)", + .def = "500", + .parent = "write_iops_log", }, { .name = "create_serialize", @@ -1437,7 +1838,7 @@ static struct fio_option options[] = { .name = "create_fsync", .type = FIO_OPT_BOOL, .off1 = td_var_offset(create_fsync), - .help = "Fsync file after creation", + .help = "fsync file after creation", .def = "1", }, { @@ -1447,11 +1848,11 @@ static struct fio_option options[] = { .help = "Create files when they are opened for IO", .def = "0", }, - { + { .name = "pre_read", .type = FIO_OPT_BOOL, .off1 = td_var_offset(pre_read), - .help = "Preread files before starting official testing", + .help = "Pre-read files before starting official testing", .def = "0", }, { @@ -1511,6 +1912,7 @@ static struct fio_option options[] = { }, { .name = "stonewall", + .alias = "wait_for_previous", .type = FIO_OPT_STR_SET, .off1 = td_var_offset(stonewall), .help = "Insert a hard barrier between this job and previous", @@ -1541,12 +1943,19 @@ static struct fio_option options[] = { .cb = str_write_lat_log_cb, .help = "Write log of latency during run", }, + { + .name = "write_iops_log", + .type = FIO_OPT_STR, + .off1 = td_var_offset(write_iops_log), + .cb = str_write_iops_log_cb, + .help = "Write log of IOPS during run", + }, { .name = "hugepage-size", .type = FIO_OPT_INT, .off1 = td_var_offset(hugepage_size), .help = "When using hugepages, specify size of each page", - .def = __stringify(FIO_HUGE_PAGE), + .def = __fio_stringify(FIO_HUGE_PAGE), }, { .name = "group_reporting", @@ -1566,6 +1975,31 @@ static struct fio_option options[] = { .off1 = td_var_offset(refill_buffers), .help = "Refill IO buffers on every IO submit", }, + { + .name = "scramble_buffers", + .type = FIO_OPT_BOOL, + .off1 = td_var_offset(scramble_buffers), + .help = "Slightly scramble buffers on every IO submit", + .def = "1", + }, + { + .name = "clat_percentiles", + .type = FIO_OPT_BOOL, + .off1 = td_var_offset(clat_percentiles), + .help = "Enable the reporting of completion latency percentiles", + .def = "1", + }, + { + .name = "percentile_list", + .type = FIO_OPT_FLOAT_LIST, + .off1 = td_var_offset(percentile_list), + .off2 = td_var_offset(overwrite_plist), + .help = "Specify a custom list of percentiles to report", + .maxlen = FIO_IO_U_LIST_MAX_LEN, + .minfp = 0.0, + .maxfp = 100.0, + }, + #ifdef FIO_HAVE_DISK_UTIL { .name = "disk_util", @@ -1582,6 +2016,14 @@ static struct fio_option options[] = { .cb = str_gtod_reduce_cb, .def = "0", }, + { + .name = "disable_lat", + .type = FIO_OPT_BOOL, + .off1 = td_var_offset(disable_lat), + .help = "Disable latency numbers", + .parent = "gtod_reduce", + .def = "0", + }, { .name = "disable_clat", .type = FIO_OPT_BOOL, @@ -1594,7 +2036,7 @@ static struct fio_option options[] = { .name = "disable_slat", .type = FIO_OPT_BOOL, .off1 = td_var_offset(disable_slat), - .help = "Disable submissionn latency numbers", + .help = "Disable submission latency numbers", .parent = "gtod_reduce", .def = "0", }, @@ -1610,24 +2052,115 @@ static struct fio_option options[] = { .name = "gtod_cpu", .type = FIO_OPT_INT, .cb = str_gtod_cpu_cb, - .help = "Setup dedicated gettimeofday() thread on this CPU", + .help = "Set up dedicated gettimeofday() thread on this CPU", .verify = gtod_cpu_verify, }, { .name = "continue_on_error", .type = FIO_OPT_BOOL, .off1 = td_var_offset(continue_on_error), - .help = "Continue on non-fatal errors during I/O", + .help = "Continue on non-fatal errors during IO", + .def = "0", + }, + { + .name = "profile", + .type = FIO_OPT_STR_STORE, + .off1 = td_var_offset(profile), + .help = "Select a specific builtin performance test", + }, + { + .name = "cgroup", + .type = FIO_OPT_STR_STORE, + .off1 = td_var_offset(cgroup), + .help = "Add job to cgroup of this name", + }, + { + .name = "cgroup_weight", + .type = FIO_OPT_INT, + .off1 = td_var_offset(cgroup_weight), + .help = "Use given weight for cgroup", + .minval = 100, + .maxval = 1000, + }, + { + .name = "cgroup_nodelete", + .type = FIO_OPT_BOOL, + .off1 = td_var_offset(cgroup_nodelete), + .help = "Do not delete cgroups after job completion", .def = "0", }, + { + .name = "uid", + .type = FIO_OPT_INT, + .off1 = td_var_offset(uid), + .help = "Run job with this user ID", + }, + { + .name = "gid", + .type = FIO_OPT_INT, + .off1 = td_var_offset(gid), + .help = "Run job with this group ID", + }, { .name = NULL, }, }; +static void add_to_lopt(struct option *lopt, struct fio_option *o, + const char *name, int val) +{ + lopt->name = (char *) name; + lopt->val = val; + if (o->type == FIO_OPT_STR_SET) + lopt->has_arg = no_argument; + else + lopt->has_arg = required_argument; +} + +static void options_to_lopts(struct fio_option *opts, + struct option *long_options, + int i, int option_type) +{ + struct fio_option *o = &opts[0]; + while (o->name) { + add_to_lopt(&long_options[i], o, o->name, option_type); + if (o->alias) { + i++; + add_to_lopt(&long_options[i], o, o->alias, option_type); + } + + i++; + o++; + assert(i < FIO_NR_OPTIONS); + } +} + +void fio_options_set_ioengine_opts(struct option *long_options, + struct thread_data *td) +{ + unsigned int i; + + i = 0; + while (long_options[i].name) { + if (long_options[i].val == FIO_GETOPT_IOENGINE) { + memset(&long_options[i], 0, sizeof(*long_options)); + break; + } + i++; + } + + /* + * Just clear out the prior ioengine options. + */ + if (!td || !td->eo) + return; + + options_to_lopts(td->io_ops->options, long_options, i, + FIO_GETOPT_IOENGINE); +} + void fio_options_dup_and_init(struct option *long_options) { - struct fio_option *o; unsigned int i; options_init(options); @@ -1636,30 +2169,276 @@ void fio_options_dup_and_init(struct option *long_options) while (long_options[i].name) i++; - o = &options[0]; - while (o->name) { - long_options[i].name = (char *) o->name; - long_options[i].val = FIO_GETOPT_JOB; - if (o->type == FIO_OPT_STR_SET) - long_options[i].has_arg = no_argument; - else - long_options[i].has_arg = required_argument; + options_to_lopts(options, long_options, i, FIO_GETOPT_JOB); +} - i++; - o++; - assert(i < FIO_NR_OPTIONS); +struct fio_keyword { + const char *word; + const char *desc; + char *replace; +}; + +static struct fio_keyword fio_keywords[] = { + { + .word = "$pagesize", + .desc = "Page size in the system", + }, + { + .word = "$mb_memory", + .desc = "Megabytes of memory online", + }, + { + .word = "$ncpus", + .desc = "Number of CPUs online in the system", + }, + { + .word = NULL, + }, +}; + +void fio_keywords_init(void) +{ + unsigned long long mb_memory; + char buf[128]; + long l; + + sprintf(buf, "%lu", page_size); + fio_keywords[0].replace = strdup(buf); + + mb_memory = os_phys_mem() / (1024 * 1024); + sprintf(buf, "%llu", mb_memory); + fio_keywords[1].replace = strdup(buf); + + l = cpus_online(); + sprintf(buf, "%lu", l); + fio_keywords[2].replace = strdup(buf); +} + +#define BC_APP "bc" + +static char *bc_calc(char *str) +{ + char buf[128], *tmp; + FILE *f; + int ret; + + /* + * No math, just return string + */ + if ((!strchr(str, '+') && !strchr(str, '-') && !strchr(str, '*') && + !strchr(str, '/')) || strchr(str, '\'')) + return str; + + /* + * Split option from value, we only need to calculate the value + */ + tmp = strchr(str, '='); + if (!tmp) + return str; + + tmp++; + + /* + * Prevent buffer overflows; such a case isn't reasonable anyway + */ + if (strlen(str) >= 128 || strlen(tmp) > 100) + return str; + + sprintf(buf, "which %s > /dev/null", BC_APP); + if (system(buf)) { + log_err("fio: bc is needed for performing math\n"); + return NULL; } + + sprintf(buf, "echo '%s' | %s", tmp, BC_APP); + f = popen(buf, "r"); + if (!f) { + return NULL; + } + + ret = fread(&buf[tmp - str], 1, 128 - (tmp - str), f); + if (ret <= 0) { + return NULL; + } + + pclose(f); + buf[(tmp - str) + ret - 1] = '\0'; + memcpy(buf, str, tmp - str); + free(str); + return strdup(buf); +} + +/* + * Return a copy of the input string with substrings of the form ${VARNAME} + * substituted with the value of the environment variable VARNAME. The + * substitution always occurs, even if VARNAME is empty or the corresponding + * environment variable undefined. + */ +static char *option_dup_subs(const char *opt) +{ + char out[OPT_LEN_MAX+1]; + char in[OPT_LEN_MAX+1]; + char *outptr = out; + char *inptr = in; + char *ch1, *ch2, *env; + ssize_t nchr = OPT_LEN_MAX; + size_t envlen; + + if (strlen(opt) + 1 > OPT_LEN_MAX) { + log_err("OPT_LEN_MAX (%d) is too small\n", OPT_LEN_MAX); + return NULL; + } + + in[OPT_LEN_MAX] = '\0'; + strncpy(in, opt, OPT_LEN_MAX); + + while (*inptr && nchr > 0) { + if (inptr[0] == '$' && inptr[1] == '{') { + ch2 = strchr(inptr, '}'); + if (ch2 && inptr+1 < ch2) { + ch1 = inptr+2; + inptr = ch2+1; + *ch2 = '\0'; + + env = getenv(ch1); + if (env) { + envlen = strlen(env); + if (envlen <= nchr) { + memcpy(outptr, env, envlen); + outptr += envlen; + nchr -= envlen; + } + } + + continue; + } + } + + *outptr++ = *inptr++; + --nchr; + } + + *outptr = '\0'; + return strdup(out); +} + +/* + * Look for reserved variable names and replace them with real values + */ +static char *fio_keyword_replace(char *opt) +{ + char *s; + int i; + int docalc = 0; + + for (i = 0; fio_keywords[i].word != NULL; i++) { + struct fio_keyword *kw = &fio_keywords[i]; + + while ((s = strstr(opt, kw->word)) != NULL) { + char *new = malloc(strlen(opt) + 1); + char *o_org = opt; + int olen = s - opt; + int len; + + /* + * Copy part of the string before the keyword and + * sprintf() the replacement after it. + */ + memcpy(new, opt, olen); + len = sprintf(new + olen, "%s", kw->replace); + + /* + * If there's more in the original string, copy that + * in too + */ + opt += strlen(kw->word) + olen; + if (strlen(opt)) + memcpy(new + olen + len, opt, opt - o_org - 1); + + /* + * replace opt and free the old opt + */ + opt = new; + free(o_org); + + docalc = 1; + } + } + + /* + * Check for potential math and invoke bc, if possible + */ + if (docalc) + opt = bc_calc(opt); + + return opt; +} + +static char **dup_and_sub_options(char **opts, int num_opts) +{ + int i; + char **opts_copy = malloc(num_opts * sizeof(*opts)); + for (i = 0; i < num_opts; i++) { + opts_copy[i] = option_dup_subs(opts[i]); + if (!opts_copy[i]) + continue; + opts_copy[i] = fio_keyword_replace(opts_copy[i]); + } + return opts_copy; } int fio_options_parse(struct thread_data *td, char **opts, int num_opts) { - int i, ret; + int i, ret, unknown; + char **opts_copy; sort_options(opts, options, num_opts); + opts_copy = dup_and_sub_options(opts, num_opts); + + for (ret = 0, i = 0, unknown = 0; i < num_opts; i++) { + struct fio_option *o; + int newret = parse_option(opts_copy[i], opts[i], options, &o, + td); - for (ret = 0, i = 0; i < num_opts; i++) - ret |= parse_option(opts[i], options, td); + if (opts_copy[i]) { + if (newret && !o) { + unknown++; + continue; + } + free(opts_copy[i]); + opts_copy[i] = NULL; + } + + ret |= newret; + } + if (unknown) { + ret |= ioengine_load(td); + if (td->eo) { + sort_options(opts_copy, td->io_ops->options, num_opts); + opts = opts_copy; + } + for (i = 0; i < num_opts; i++) { + struct fio_option *o = NULL; + int newret = 1; + if (!opts_copy[i]) + continue; + + if (td->eo) + newret = parse_option(opts_copy[i], opts[i], + td->io_ops->options, &o, + td->eo); + + ret |= newret; + if (!o) + log_err("Bad option <%s>\n", opts[i]); + + free(opts_copy[i]); + opts_copy[i] = NULL; + } + } + + free(opts_copy); return ret; } @@ -1668,6 +2447,12 @@ int fio_cmd_option_parse(struct thread_data *td, const char *opt, char *val) return parse_cmd_option(opt, val, options, td); } +int fio_cmd_ioengine_option_parse(struct thread_data *td, const char *opt, + char *val) +{ + return parse_cmd_option(opt, val, td->io_ops->options, td); +} + void fio_fill_default_options(struct thread_data *td) { fill_default_options(td, options); @@ -1678,40 +2463,124 @@ int fio_show_option_help(const char *opt) return show_cmd_help(options, opt); } -static void __options_mem(struct thread_data *td, int alloc) +void options_mem_dupe(void *data, struct fio_option *options) { - struct thread_options *o = &td->o; - struct fio_option *opt; + struct fio_option *o; char **ptr; - int i; - for (i = 0, opt = &options[0]; opt->name; i++, opt = &options[i]) { - if (opt->type != FIO_OPT_STR_STORE) + for (o = &options[0]; o->name; o++) { + if (o->type != FIO_OPT_STR_STORE) continue; - ptr = (void *) o + opt->off1; - if (*ptr) { - if (alloc) - *ptr = strdup(*ptr); - else { - free(*ptr); - *ptr = NULL; - } - } + ptr = td_var(data, o->off1); + if (*ptr) + *ptr = strdup(*ptr); } } /* * dupe FIO_OPT_STR_STORE options */ -void options_mem_dupe(struct thread_data *td) +void fio_options_mem_dupe(struct thread_data *td) { - __options_mem(td, 1); + options_mem_dupe(&td->o, options); + + if (td->eo && td->io_ops) { + void *oldeo = td->eo; + + td->eo = malloc(td->io_ops->option_struct_size); + memcpy(td->eo, oldeo, td->io_ops->option_struct_size); + options_mem_dupe(td->eo, td->io_ops->options); + } } -void options_mem_free(struct thread_data fio_unused *td) +unsigned int fio_get_kb_base(void *data) { -#if 0 - __options_mem(td, 0); -#endif + struct thread_data *td = data; + unsigned int kb_base = 0; + + if (td) + kb_base = td->o.kb_base; + if (!kb_base) + kb_base = 1024; + + return kb_base; +} + +int add_option(struct fio_option *o) +{ + struct fio_option *__o; + int opt_index = 0; + + __o = options; + while (__o->name) { + opt_index++; + __o++; + } + + memcpy(&options[opt_index], o, sizeof(*o)); + return 0; +} + +void invalidate_profile_options(const char *prof_name) +{ + struct fio_option *o; + + o = options; + while (o->name) { + if (o->prof_name && !strcmp(o->prof_name, prof_name)) { + o->type = FIO_OPT_INVALID; + o->prof_name = NULL; + } + o++; + } +} + +void add_opt_posval(const char *optname, const char *ival, const char *help) +{ + struct fio_option *o; + unsigned int i; + + o = find_option(options, optname); + if (!o) + return; + + for (i = 0; i < PARSE_MAX_VP; i++) { + if (o->posval[i].ival) + continue; + + o->posval[i].ival = ival; + o->posval[i].help = help; + break; + } +} + +void del_opt_posval(const char *optname, const char *ival) +{ + struct fio_option *o; + unsigned int i; + + o = find_option(options, optname); + if (!o) + return; + + for (i = 0; i < PARSE_MAX_VP; i++) { + if (!o->posval[i].ival) + continue; + if (strcmp(o->posval[i].ival, ival)) + continue; + + o->posval[i].ival = NULL; + o->posval[i].help = NULL; + } +} + +void fio_options_free(struct thread_data *td) +{ + options_free(options, td); + if (td->eo && td->io_ops && td->io_ops->options) { + options_free(td->io_ops->options, td->eo); + free(td->eo); + td->eo = NULL; + } }