X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=options.c;h=1760762040c65fb83d0eb10a0466738c514551ec;hp=f62ab6d573cd3369cbeced32edec4a4fe4c666d4;hb=1ae83d45ed853cd73b95b89ae14cacac5b97187e;hpb=0329ecda39a8ed42e173d25e16d39c968568d783 diff --git a/options.c b/options.c index f62ab6d5..17607620 100644 --- a/options.c +++ b/options.c @@ -166,7 +166,7 @@ static int bssplit_ddir(struct thread_data *td, int ddir, char *str) static int str_bssplit_cb(void *data, const char *input) { struct thread_data *td = data; - char *str, *p, *odir; + char *str, *p, *odir, *ddir; int ret = 0; p = str = strdup(input); @@ -176,7 +176,21 @@ static int str_bssplit_cb(void *data, const char *input) odir = strchr(str, ','); if (odir) { - ret = bssplit_ddir(td, DDIR_WRITE, odir + 1); + ddir = strchr(odir + 1, ','); + if (ddir) { + ret = bssplit_ddir(td, DDIR_TRIM, ddir + 1); + if (!ret) + *ddir = '\0'; + } else { + char *op; + + op = strdup(odir + 1); + ret = bssplit_ddir(td, DDIR_TRIM, op); + + free(op); + } + if (!ret) + ret = bssplit_ddir(td, DDIR_WRITE, odir + 1); if (!ret) { *odir = '\0'; ret = bssplit_ddir(td, DDIR_READ, str); @@ -185,18 +199,116 @@ static int str_bssplit_cb(void *data, const char *input) char *op; op = strdup(str); + ret = bssplit_ddir(td, DDIR_WRITE, op); + free(op); + if (!ret) { + op = strdup(str); + ret = bssplit_ddir(td, DDIR_TRIM, op); + free(op); + } ret = bssplit_ddir(td, DDIR_READ, str); - if (!ret) - ret = bssplit_ddir(td, DDIR_WRITE, op); - - free(op); } free(p); return ret; } +static int str2error(char *str) +{ + const char * err[] = {"EPERM", "ENOENT", "ESRCH", "EINTR", "EIO", + "ENXIO", "E2BIG", "ENOEXEC", "EBADF", + "ECHILD", "EAGAIN", "ENOMEM", "EACCES", + "EFAULT", "ENOTBLK", "EBUSY", "EEXIST", + "EXDEV", "ENODEV", "ENOTDIR", "EISDIR", + "EINVAL", "ENFILE", "EMFILE", "ENOTTY", + "ETXTBSY","EFBIG", "ENOSPC", "ESPIPE", + "EROFS","EMLINK", "EPIPE", "EDOM", "ERANGE"}; + int i = 0, num = sizeof(err) / sizeof(void *); + + while( i < num) { + if (!strcmp(err[i], str)) + return i + 1; + i++; + } + return 0; +} + +static int ignore_error_type(struct thread_data *td, int etype, char *str) +{ + unsigned int i; + int *error; + char *fname; + + if (etype >= ERROR_TYPE_CNT) { + log_err("Illegal error type\n"); + return 1; + } + + td->o.ignore_error_nr[etype] = 4; + error = malloc(4 * sizeof(struct bssplit)); + + i = 0; + while ((fname = strsep(&str, ":")) != NULL) { + + if (!strlen(fname)) + break; + + /* + * grow struct buffer, if needed + */ + if (i == td->o.ignore_error_nr[etype]) { + td->o.ignore_error_nr[etype] <<= 1; + error = realloc(error, td->o.ignore_error_nr[etype] + * sizeof(int)); + } + if (fname[0] == 'E') { + error[i] = str2error(fname); + } else { + error[i] = atoi(fname); + if (error[i] < 0) + error[i] = error[i]; + } + if (!error[i]) { + log_err("Unknown error %s, please use number value \n", + fname); + return 1; + } + i++; + } + if (i) { + td->o.continue_on_error |= 1 << etype; + td->o.ignore_error_nr[etype] = i; + td->o.ignore_error[etype] = error; + } + return 0; + +} + +static int str_ignore_error_cb(void *data, const char *input) +{ + struct thread_data *td = data; + char *str, *p, *n; + int type = 0, ret = 1; + p = str = strdup(input); + + strip_blank_front(&str); + strip_blank_end(str); + + while (p) { + n = strchr(p, ','); + if (n) + *n++ = '\0'; + ret = ignore_error_type(td, type, p); + if (ret) + break; + p = n; + type++; + } + free(str); + return ret; +} + static int str_rw_cb(void *data, const char *str) { struct thread_data *td = data; @@ -245,12 +357,9 @@ 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; + if (td->o.verify == VERIFY_CRC32C_INTEL || + td->o.verify == VERIFY_CRC32C) { + crc32c_intel_probe(); } return 0; @@ -261,7 +370,8 @@ 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(); + fio_clock_source_set = 1; + fio_clock_init(); return 0; } @@ -455,6 +565,130 @@ static int str_verify_cpus_allowed_cb(void *data, const char *input) } #endif +#ifdef CONFIG_LIBNUMA +static int str_numa_cpunodes_cb(void *data, char *input) +{ + struct thread_data *td = data; + + /* numa_parse_nodestring() parses a character string list + * of nodes into a bit mask. The bit mask is allocated by + * numa_allocate_nodemask(), so it should be freed by + * numa_free_nodemask(). + */ + td->o.numa_cpunodesmask = numa_parse_nodestring(input); + if (td->o.numa_cpunodesmask == NULL) { + log_err("fio: numa_parse_nodestring failed\n"); + td_verror(td, 1, "str_numa_cpunodes_cb"); + return 1; + } + + td->o.numa_cpumask_set = 1; + return 0; +} + +static int str_numa_mpol_cb(void *data, char *input) +{ + struct thread_data *td = data; + const char * const policy_types[] = + { "default", "prefer", "bind", "interleave", "local" }; + int i; + + char *nodelist = strchr(input, ':'); + if (nodelist) { + /* NUL-terminate mode */ + *nodelist++ = '\0'; + } + + for (i = 0; i <= MPOL_LOCAL; i++) { + if (!strcmp(input, policy_types[i])) { + td->o.numa_mem_mode = i; + break; + } + } + if (i > MPOL_LOCAL) { + log_err("fio: memory policy should be: default, prefer, bind, interleave, local\n"); + goto out; + } + + switch (td->o.numa_mem_mode) { + case MPOL_PREFERRED: + /* + * Insist on a nodelist of one node only + */ + if (nodelist) { + char *rest = nodelist; + while (isdigit(*rest)) + rest++; + if (*rest) { + log_err("fio: one node only for \'prefer\'\n"); + goto out; + } + } else { + log_err("fio: one node is needed for \'prefer\'\n"); + goto out; + } + break; + case MPOL_INTERLEAVE: + /* + * Default to online nodes with memory if no nodelist + */ + if (!nodelist) + nodelist = strdup("all"); + break; + case MPOL_LOCAL: + case MPOL_DEFAULT: + /* + * Don't allow a nodelist + */ + if (nodelist) { + log_err("fio: NO nodelist for \'local\'\n"); + goto out; + } + break; + case MPOL_BIND: + /* + * Insist on a nodelist + */ + if (!nodelist) { + log_err("fio: a nodelist is needed for \'bind\'\n"); + goto out; + } + break; + } + + + /* numa_parse_nodestring() parses a character string list + * of nodes into a bit mask. The bit mask is allocated by + * numa_allocate_nodemask(), so it should be freed by + * numa_free_nodemask(). + */ + switch (td->o.numa_mem_mode) { + case MPOL_PREFERRED: + td->o.numa_mem_prefer_node = atoi(nodelist); + break; + case MPOL_INTERLEAVE: + case MPOL_BIND: + td->o.numa_memnodesmask = numa_parse_nodestring(nodelist); + if (td->o.numa_memnodesmask == NULL) { + log_err("fio: numa_parse_nodestring failed\n"); + td_verror(td, 1, "str_numa_memnodes_cb"); + return 1; + } + break; + case MPOL_LOCAL: + case MPOL_DEFAULT: + default: + break; + } + + td->o.numa_memmask_set = 1; + return 0; + +out: + return 1; +} +#endif + #ifdef FIO_HAVE_TRIM static int str_verify_trim_cb(void *data, unsigned long long *val) { @@ -479,7 +713,7 @@ static int str_fst_cb(void *data, const char *str) return 0; } -#ifdef FIO_HAVE_SYNC_FILE_RANGE +#ifdef CONFIG_SYNC_FILE_RANGE static int str_sfr_cb(void *data, const char *str) { struct thread_data *td = data; @@ -495,6 +729,45 @@ static int str_sfr_cb(void *data, const char *str) } #endif +static int str_random_distribution_cb(void *data, const char *str) +{ + struct thread_data *td = data; + double val; + char *nr; + + if (td->o.random_distribution == FIO_RAND_DIST_ZIPF) + val = 1.1; + else if (td->o.random_distribution == FIO_RAND_DIST_PARETO) + val = 0.2; + else + return 0; + + nr = get_opt_postfix(str); + if (nr && !str_to_float(nr, &val)) { + log_err("fio: random postfix parsing failed\n"); + free(nr); + return 1; + } + + free(nr); + + if (td->o.random_distribution == FIO_RAND_DIST_ZIPF) { + if (val == 1.00) { + log_err("fio: zipf theta must different than 1.0\n"); + return 1; + } + td->o.zipf_theta = val; + } else { + if (val <= 0.00 || val >= 1.00) { + log_err("fio: pareto input out of range (0 < input < 1.0)\n"); + return 1; + } + td->o.pareto_h = val; + } + + return 0; +} + static int check_dir(struct thread_data *td, char *fname) { #if 0 @@ -689,12 +962,32 @@ static int str_verify_pattern_cb(void *data, const char *input) } } } + + /* + * Fill the pattern all the way to the end. This greatly reduces + * the number of memcpy's we have to do when verifying the IO. + */ + while (i > 1 && i * 2 <= MAX_PATTERN_SIZE) { + memcpy(&td->o.verify_pattern[i], &td->o.verify_pattern[0], i); + i *= 2; + } + if (i == 1) { + /* + * The code in verify_io_u_pattern assumes a single byte pattern + * fills the whole verify pattern buffer. + */ + memset(td->o.verify_pattern, td->o.verify_pattern[0], + MAX_PATTERN_SIZE); + } + 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; } @@ -915,6 +1208,10 @@ static struct fio_option options[FIO_MAX_OPTS] = { .oval = TD_DDIR_WRITE, .help = "Sequential write", }, + { .ival = "trim", + .oval = TD_DDIR_TRIM, + .help = "Sequential trim", + }, { .ival = "randread", .oval = TD_DDIR_RANDREAD, .help = "Random read", @@ -923,10 +1220,18 @@ static struct fio_option options[FIO_MAX_OPTS] = { .oval = TD_DDIR_RANDWRITE, .help = "Random write", }, + { .ival = "randtrim", + .oval = TD_DDIR_RANDTRIM, + .help = "Random trim", + }, { .ival = "rw", .oval = TD_DDIR_RW, .help = "Sequential read and write mix", }, + { .ival = "readwrite", + .oval = TD_DDIR_RW, + .help = "Sequential read and write mix", + }, { .ival = "randrw", .oval = TD_DDIR_RANDRW, .help = "Random read and write mix" @@ -967,12 +1272,12 @@ static struct fio_option options[FIO_MAX_OPTS] = { { .ival = "vsync", .help = "Use readv/writev", }, -#ifdef FIO_HAVE_LIBAIO +#ifdef CONFIG_LIBAIO { .ival = "libaio", .help = "Linux native asynchronous IO", }, #endif -#ifdef FIO_HAVE_POSIXAIO +#ifdef CONFIG_POSIXAIO { .ival = "posixaio", .help = "POSIX asynchronous IO", }, @@ -990,7 +1295,7 @@ static struct fio_option options[FIO_MAX_OPTS] = { { .ival = "mmap", .help = "Memory mapped IO" }, -#ifdef FIO_HAVE_SPLICE +#ifdef CONFIG_LINUX_SPLICE { .ival = "splice", .help = "splice/vmsplice based IO", }, @@ -1009,15 +1314,10 @@ static struct fio_option options[FIO_MAX_OPTS] = { { .ival = "net", .help = "Network IO", }, -#ifdef FIO_HAVE_SYSLET - { .ival = "syslet-rw", - .help = "syslet enabled async pread/pwrite IO", - }, -#endif { .ival = "cpuio", .help = "CPU cycle burner engine", }, -#ifdef FIO_HAVE_GUASI +#ifdef CONFIG_GUASI { .ival = "guasi", .help = "GUASI IO engine", }, @@ -1027,10 +1327,25 @@ static struct fio_option options[FIO_MAX_OPTS] = { .help = "binject direct inject block engine", }, #endif -#ifdef FIO_HAVE_RDMA +#ifdef CONFIG_RDMA { .ival = "rdma", .help = "RDMA IO engine", }, +#endif +#ifdef CONFIG_FUSION_AW + { .ival = "fusion-aw-sync", + .help = "Fusion-io atomic write engine", + }, +#endif +#ifdef FIO_HAVE_E4_ENG + { .ival = "e4defrag", + .help = "ext4 defrag engine", + }, +#endif +#ifdef FIO_HAVE_FALLOC_ENG + { .ival = "falloc", + .help = "fallocate() file based engine", + }, #endif { .ival = "external", .help = "Load external engine (append name)", @@ -1101,12 +1416,21 @@ static struct fio_option options[FIO_MAX_OPTS] = { .help = "Start IO from this offset", .def = "0", }, + { + .name = "offset_increment", + .type = FIO_OPT_STR_VAL, + .off1 = td_var_offset(offset_increment), + .help = "What is the increment from one offset to the next", + .parent = "offset", + .def = "0", + }, { .name = "bs", .alias = "blocksize", .type = FIO_OPT_INT, .off1 = td_var_offset(bs[DDIR_READ]), .off2 = td_var_offset(bs[DDIR_WRITE]), + .off3 = td_var_offset(bs[DDIR_TRIM]), .minval = 1, .help = "Block size unit", .def = "4k", @@ -1118,6 +1442,7 @@ static struct fio_option options[FIO_MAX_OPTS] = { .type = FIO_OPT_INT, .off1 = td_var_offset(ba[DDIR_READ]), .off2 = td_var_offset(ba[DDIR_WRITE]), + .off3 = td_var_offset(ba[DDIR_TRIM]), .minval = 1, .help = "IO block offset alignment", .parent = "rw", @@ -1130,6 +1455,8 @@ static struct fio_option options[FIO_MAX_OPTS] = { .off2 = td_var_offset(max_bs[DDIR_READ]), .off3 = td_var_offset(min_bs[DDIR_WRITE]), .off4 = td_var_offset(max_bs[DDIR_WRITE]), + .off5 = td_var_offset(min_bs[DDIR_TRIM]), + .off6 = td_var_offset(max_bs[DDIR_TRIM]), .minval = 1, .help = "Set block size range (in more detail than bs)", .parent = "rw", @@ -1180,6 +1507,45 @@ static struct fio_option options[FIO_MAX_OPTS] = { .parent = "norandommap", .def = "0", }, + { + .name = "random_generator", + .type = FIO_OPT_STR, + .off1 = td_var_offset(random_generator), + .help = "Type of random number generator to use", + .def = "tausworthe", + .posval = { + { .ival = "tausworthe", + .oval = FIO_RAND_GEN_TAUSWORTHE, + .help = "Strong Tausworthe generator", + }, + { .ival = "lfsr", + .oval = FIO_RAND_GEN_LFSR, + .help = "Variable length LFSR", + }, + }, + }, + { + .name = "random_distribution", + .type = FIO_OPT_STR, + .off1 = td_var_offset(random_distribution), + .cb = str_random_distribution_cb, + .help = "Random offset distribution generator", + .def = "random", + .posval = { + { .ival = "random", + .oval = FIO_RAND_DIST_RANDOM, + .help = "Completely random", + }, + { .ival = "zipf", + .oval = FIO_RAND_DIST_ZIPF, + .help = "Zipf distribution", + }, + { .ival = "pareto", + .oval = FIO_RAND_DIST_PARETO, + .help = "Pareto distribution", + }, + }, + }, { .name = "nrfiles", .alias = "nr_files", @@ -1279,7 +1645,7 @@ static struct fio_option options[FIO_MAX_OPTS] = { .help = "Make every Nth write a barrier write", .def = "0", }, -#ifdef FIO_HAVE_SYNC_FILE_RANGE +#ifdef CONFIG_SYNC_FILE_RANGE { .name = "sync_file_range", .posval = { @@ -1376,14 +1742,18 @@ static struct fio_option options[FIO_MAX_OPTS] = { .off1 = td_var_offset(clocksource), .help = "What type of timing source to use", .posval = { +#ifdef CONFIG_GETTIMEOFDAY { .ival = "gettimeofday", .oval = CS_GTOD, .help = "Use gettimeofday(2) for timing", }, +#endif +#ifdef CONFIG_CLOCK_GETTIME { .ival = "clock_gettime", .oval = CS_CGETTIME, .help = "Use clock_gettime(2) for timing", }, +#endif #ifdef ARCH_HAVE_CPU_CLOCK { .ival = "cpu", .oval = CS_CPUCLOCK, @@ -1462,12 +1832,12 @@ static struct fio_option options[FIO_MAX_OPTS] = { .help = "Use crc32 checksums for verification", }, { .ival = "crc32c-intel", - .oval = VERIFY_CRC32C_INTEL, - .help = "Use hw crc32c checksums for verification", + .oval = VERIFY_CRC32C, + .help = "Use crc32c checksums for verification (hw assisted, if available)", }, { .ival = "crc32c", .oval = VERIFY_CRC32C, - .help = "Use crc32c checksums for verification", + .help = "Use crc32c checksums for verification (hw assisted, if available)", }, { .ival = "crc16", .oval = VERIFY_CRC16, @@ -1516,6 +1886,16 @@ static struct fio_option options[FIO_MAX_OPTS] = { .def = "1", .parent = "verify", }, + { + .name = "verifysort_nr", + .type = FIO_OPT_INT, + .off1 = td_var_offset(verifysort_nr), + .help = "Pre-load and sort verify blocks for a read workload", + .minval = 0, + .maxval = 131072, + .def = "1024", + .parent = "verify", + }, { .name = "verify_interval", .type = FIO_OPT_INT, @@ -1767,30 +2147,34 @@ static struct fio_option options[FIO_MAX_OPTS] = { { .name = "rate", .type = FIO_OPT_INT, - .off1 = td_var_offset(rate[0]), - .off2 = td_var_offset(rate[1]), + .off1 = td_var_offset(rate[DDIR_READ]), + .off2 = td_var_offset(rate[DDIR_WRITE]), + .off3 = td_var_offset(rate[DDIR_TRIM]), .help = "Set bandwidth rate", }, { .name = "ratemin", .type = FIO_OPT_INT, - .off1 = td_var_offset(ratemin[0]), - .off2 = td_var_offset(ratemin[1]), + .off1 = td_var_offset(ratemin[DDIR_READ]), + .off2 = td_var_offset(ratemin[DDIR_WRITE]), + .off3 = td_var_offset(ratemin[DDIR_TRIM]), .help = "Job must meet this rate or it will be shutdown", .parent = "rate", }, { .name = "rate_iops", .type = FIO_OPT_INT, - .off1 = td_var_offset(rate_iops[0]), - .off2 = td_var_offset(rate_iops[1]), + .off1 = td_var_offset(rate_iops[DDIR_READ]), + .off2 = td_var_offset(rate_iops[DDIR_WRITE]), + .off3 = td_var_offset(rate_iops[DDIR_TRIM]), .help = "Limit IO used to this number of IO operations/sec", }, { .name = "rate_iops_min", .type = FIO_OPT_INT, - .off1 = td_var_offset(rate_iops_min[0]), - .off2 = td_var_offset(rate_iops_min[1]), + .off1 = td_var_offset(rate_iops_min[DDIR_READ]), + .off2 = td_var_offset(rate_iops_min[DDIR_WRITE]), + .off3 = td_var_offset(rate_iops_min[DDIR_TRIM]), .help = "Job must meet this rate or it will be shut down", .parent = "rate_iops", }, @@ -1802,6 +2186,12 @@ static struct fio_option options[FIO_MAX_OPTS] = { .def = "1000", .parent = "rate", }, + { + .name = "max_latency", + .type = FIO_OPT_INT, + .off1 = td_var_offset(max_latency), + .help = "Maximum tolerated IO latency (usec)", + }, { .name = "invalidate", .type = FIO_OPT_BOOL, @@ -1855,6 +2245,13 @@ static struct fio_option options[FIO_MAX_OPTS] = { .help = "Create files when they are opened for IO", .def = "0", }, + { + .name = "create_only", + .type = FIO_OPT_BOOL, + .off1 = td_var_offset(create_only), + .help = "Only perform file creation phase", + .def = "0", + }, { .name = "pre_read", .type = FIO_OPT_BOOL, @@ -1889,6 +2286,20 @@ static struct fio_option options[FIO_MAX_OPTS] = { .cb = str_cpus_allowed_cb, .help = "Set CPUs allowed", }, +#endif +#ifdef CONFIG_LIBNUMA + { + .name = "numa_cpu_nodes", + .type = FIO_OPT_STR, + .cb = str_numa_cpunodes_cb, + .help = "NUMA CPU nodes bind", + }, + { + .name = "numa_mem_policy", + .type = FIO_OPT_STR, + .cb = str_numa_mpol_cb, + .help = "NUMA memory policy setup", + }, #endif { .name = "end_fsync", @@ -1996,6 +2407,21 @@ static struct fio_option options[FIO_MAX_OPTS] = { .help = "Slightly scramble buffers on every IO submit", .def = "1", }, + { + .name = "buffer_compress_percentage", + .type = FIO_OPT_INT, + .off1 = td_var_offset(compress_percentage), + .maxval = 100, + .minval = 1, + .help = "How compressible the buffer is (approximately)", + }, + { + .name = "buffer_compress_chunk", + .type = FIO_OPT_INT, + .off1 = td_var_offset(compress_chunk), + .parent = "buffer_compress_percentage", + .help = "Size of compressible region in buffer", + }, { .name = "clat_percentiles", .type = FIO_OPT_BOOL, @@ -2110,6 +2536,21 @@ static struct fio_option options[FIO_MAX_OPTS] = { }, }, }, + { + .name = "ignore_error", + .type = FIO_OPT_STR, + .cb = str_ignore_error_cb, + .help = "Set a specific list of errors to ignore", + .parent = "rw", + }, + { + .name = "error_dump", + .type = FIO_OPT_BOOL, + .off1 = td_var_offset(error_dump), + .def = "0", + .help = "Dump info on each error", + }, + { .name = "profile", .type = FIO_OPT_STR_STORE, @@ -2149,6 +2590,40 @@ static struct fio_option options[FIO_MAX_OPTS] = { .off1 = td_var_offset(gid), .help = "Run job with this group ID", }, + { + .name = "flow_id", + .type = FIO_OPT_INT, + .off1 = td_var_offset(flow_id), + .help = "The flow index ID to use", + .def = "0", + }, + { + .name = "flow", + .type = FIO_OPT_INT, + .off1 = td_var_offset(flow), + .help = "Weight for flow control of this job", + .parent = "flow_id", + .def = "0", + }, + { + .name = "flow_watermark", + .type = FIO_OPT_INT, + .off1 = td_var_offset(flow_watermark), + .help = "High watermark for flow control. This option" + " should be set to the same value for all threads" + " with non-zero flow.", + .parent = "flow_id", + .def = "1024", + }, + { + .name = "flow_sleep", + .type = FIO_OPT_INT, + .off1 = td_var_offset(flow_sleep), + .help = "How many microseconds to sleep after being held" + " back by the flow control mechanism", + .parent = "flow_id", + .def = "0", + }, { .name = NULL, }, @@ -2250,7 +2725,7 @@ void fio_keywords_init(void) char buf[128]; long l; - sprintf(buf, "%lu", page_size); + sprintf(buf, "%lu", (unsigned long) page_size); fio_keywords[0].replace = strdup(buf); mb_memory = os_phys_mem() / (1024 * 1024);