X-Git-Url: https://git.kernel.dk/?a=blobdiff_plain;f=init.c;h=89c66afa0e1b8b439ce5a534f76e5d83183431ce;hb=f2e714707b27f5918a4020e3090d2b346094379c;hp=194242439dc93b61b4126ce8a1a9fa3808b08980;hpb=320beefe2d0c5f05f24b4e0c1d13fcb2f1286a78;p=fio.git diff --git a/init.c b/init.c index 19424243..63826662 100644 --- a/init.c +++ b/init.c @@ -8,8 +8,6 @@ #include #include #include -#include -#include #include #include #include @@ -17,533 +15,54 @@ #include "fio.h" #include "parse.h" +#include "smalloc.h" +#include "filehash.h" +#include "verify.h" +#include "profile.h" + +#include "lib/getopt.h" + +static char fio_version_string[] = "fio 1.53"; #define FIO_RANDSEED (0xb1899bedUL) -#define td_var_offset(var) ((size_t) &((struct thread_data *)0)->var) +static char **ini_file; +static int max_jobs = MAX_JOBS; +static int dump_cmdline; -static int str_ioengine_cb(void *, const char *); -static int str_mem_cb(void *, const char *); -static int str_lockmem_cb(void *, unsigned long *); -#ifdef FIO_HAVE_IOPRIO -static int str_prio_cb(void *, unsigned int *); -static int str_prioclass_cb(void *, unsigned int *); -#endif -static int str_exitall_cb(void); -static int str_cpumask_cb(void *, unsigned int *); +static struct thread_data def_thread; +struct thread_data *threads = NULL; -#define __stringify_1(x) #x -#define __stringify(x) __stringify_1(x) +int exitall_on_terminate = 0; +int terse_output = 0; +int eta_print; +unsigned long long mlock_size = 0; +FILE *f_out = NULL; +FILE *f_err = NULL; +char **job_sections = NULL; +int nr_job_sections = 0; +char *exec_profile = NULL; +int warnings_fatal = 0; -/* - * Map of job/command line options - */ -static struct fio_option options[] = { - { - .name = "description", - .type = FIO_OPT_STR_STORE, - .off1 = td_var_offset(description), - .help = "Text job description", - }, - { - .name = "name", - .type = FIO_OPT_STR_STORE, - .off1 = td_var_offset(name), - .help = "Name of this job", - }, - { - .name = "directory", - .type = FIO_OPT_STR_STORE, - .off1 = td_var_offset(directory), - .help = "Directory to store files in", - }, - { - .name = "filename", - .type = FIO_OPT_STR_STORE, - .off1 = td_var_offset(filename), - .help = "Force the use of a specific file", - }, - { - .name = "rw", - .type = FIO_OPT_STR, - .off1 = td_var_offset(td_ddir), - .help = "IO direction", - .def = "read", - .posval = { - { .ival = "read", .oval = TD_DDIR_READ }, - { .ival = "write", .oval = TD_DDIR_WRITE }, - { .ival = "randread", .oval = TD_DDIR_RANDREAD }, - { .ival = "randwrite", .oval = TD_DDIR_RANDWRITE }, - { .ival = "rw", .oval = TD_DDIR_RW }, - { .ival = "randrw", .oval = TD_DDIR_RANDRW }, - }, - }, - { - .name = "ioengine", - .type = FIO_OPT_STR, - .cb = str_ioengine_cb, - .help = "IO engine to use", - .def = "sync", - .posval = { - { .ival = "sync", }, -#ifdef FIO_HAVE_LIBAIO - { .ival = "libaio", }, -#endif -#ifdef FIO_HAVE_POSIXAIO - { .ival = "posixaio", }, -#endif - { .ival = "mmap", }, -#ifdef FIO_HAVE_SPLICE - { .ival = "splice", }, -#endif -#ifdef FIO_HAVE_SGIO - { .ival = "sg", }, -#endif - { .ival = "null", }, { .ival = "net", }, -#ifdef FIO_HAVE_SYSLET - { .ival = "syslet-rw", }, -#endif - { .ival = "external", }, - }, - }, - { - .name = "iodepth", - .type = FIO_OPT_INT, - .off1 = td_var_offset(iodepth), - .help = "Amount of IO buffers to keep in flight", - .def = "1", - }, - { - .name = "iodepth_batch", - .type = FIO_OPT_INT, - .off1 = td_var_offset(iodepth_batch), - .help = "Number of IO to submit in one go", - }, - { - .name = "iodepth_low", - .type = FIO_OPT_INT, - .off1 = td_var_offset(iodepth_low), - .help = "Low water mark for queuing depth", - }, - { - .name = "size", - .type = FIO_OPT_STR_VAL, - .off1 = td_var_offset(total_file_size), - .help = "Size of device or file", - }, - { - .name = "bs", - .type = FIO_OPT_STR_VAL_INT, - .off1 = td_var_offset(bs[DDIR_READ]), - .off2 = td_var_offset(bs[DDIR_WRITE]), - .help = "Block size unit", - .def = "4k", - }, - { - .name = "bsrange", - .type = FIO_OPT_RANGE, - .off1 = td_var_offset(min_bs[DDIR_READ]), - .off2 = td_var_offset(max_bs[DDIR_READ]), - .off3 = td_var_offset(min_bs[DDIR_WRITE]), - .off4 = td_var_offset(max_bs[DDIR_WRITE]), - .help = "Set block size range (in more detail than bs)", - }, - { - .name = "bs_unaligned", - .type = FIO_OPT_STR_SET, - .off1 = td_var_offset(bs_unaligned), - .help = "Don't sector align IO buffer sizes", - }, - { - .name = "offset", - .type = FIO_OPT_STR_VAL, - .off1 = td_var_offset(start_offset), - .help = "Start IO from this offset", - .def = "0", - }, - { - .name = "randrepeat", - .type = FIO_OPT_BOOL, - .off1 = td_var_offset(rand_repeatable), - .help = "Use repeatable random IO pattern", - .def = "1", - }, - { - .name = "norandommap", - .type = FIO_OPT_STR_SET, - .off1 = td_var_offset(norandommap), - .help = "Accept potential duplicate random blocks", - }, - { - .name = "nrfiles", - .type = FIO_OPT_INT, - .off1 = td_var_offset(nr_files), - .help = "Split job workload between this number of files", - .def = "1", - }, - { - .name = "file_service_type", - .type = FIO_OPT_STR, - .off1 = td_var_offset(file_service_type), - .help = "How to select which file to service next", - .def = "roundrobin", - .posval = { - { .ival = "random", .oval = FIO_FSERVICE_RANDOM }, - { .ival = "roundrobin", .oval = FIO_FSERVICE_RR }, - }, - }, - { - .name = "fsync", - .type = FIO_OPT_INT, - .off1 = td_var_offset(fsync_blocks), - .help = "Issue fsync for writes every given number of blocks", - .def = "0", - }, - { - .name = "direct", - .type = FIO_OPT_BOOL, - .off1 = td_var_offset(odirect), - .help = "Use O_DIRECT IO (negates buffered)", - .def = "0", - }, - { - .name = "buffered", - .type = FIO_OPT_BOOL, - .off1 = td_var_offset(odirect), - .neg = 1, - .help = "Use buffered IO (negates direct)", - .def = "1", - }, - { - .name = "overwrite", - .type = FIO_OPT_BOOL, - .off1 = td_var_offset(overwrite), - .help = "When writing, set whether to overwrite current data", - .def = "0", - }, - { - .name = "loops", - .type = FIO_OPT_INT, - .off1 = td_var_offset(loops), - .help = "Number of times to run the job", - .def = "1", - }, - { - .name = "numjobs", - .type = FIO_OPT_INT, - .off1 = td_var_offset(numjobs), - .help = "Duplicate this job this many times", - .def = "1", - }, - { - .name = "startdelay", - .type = FIO_OPT_INT, - .off1 = td_var_offset(start_delay), - .help = "Only start job when this period has passed", - .def = "0", - }, - { - .name = "runtime", - .alias = "timeout", - .type = FIO_OPT_STR_VAL_TIME, - .off1 = td_var_offset(timeout), - .help = "Stop workload when this amount of time has passed", - .def = "0", - }, - { - .name = "mem", - .type = FIO_OPT_STR, - .cb = str_mem_cb, - .off1 = td_var_offset(mem_type), - .help = "Backing type for IO buffers", - .def = "malloc", - .posval = { - { .ival = "malloc", .oval = MEM_MALLOC }, - { .ival = "shm", .oval = MEM_SHM }, -#ifdef FIO_HAVE_HUGETLB - { .ival = "shmhuge", .oval = MEM_SHMHUGE }, -#endif - { .ival = "mmap", .oval = MEM_MMAP }, -#ifdef FIO_HAVE_HUGETLB - { .ival = "mmaphuge", .oval = MEM_MMAPHUGE }, -#endif - }, - }, - { - .name = "verify", - .type = FIO_OPT_STR, - .off1 = td_var_offset(verify), - .help = "Verify sum function", - .def = "0", - .posval = { - { .ival = "0", .oval = VERIFY_NONE }, - { .ival = "crc32", .oval = VERIFY_CRC32 }, - { .ival = "md5", .oval = VERIFY_MD5 }, - }, - }, - { - .name = "write_iolog", - .type = FIO_OPT_STR_STORE, - .off1 = td_var_offset(write_iolog_file), - .help = "Store IO pattern to file", - }, - { - .name = "read_iolog", - .type = FIO_OPT_STR_STORE, - .off1 = td_var_offset(read_iolog_file), - .help = "Playback IO pattern from file", - }, - { - .name = "exec_prerun", - .type = FIO_OPT_STR_STORE, - .off1 = td_var_offset(exec_prerun), - .help = "Execute this file prior to running job", - }, - { - .name = "exec_postrun", - .type = FIO_OPT_STR_STORE, - .off1 = td_var_offset(exec_postrun), - .help = "Execute this file after running job", - }, -#ifdef FIO_HAVE_IOSCHED_SWITCH - { - .name = "ioscheduler", - .type = FIO_OPT_STR_STORE, - .off1 = td_var_offset(ioscheduler), - .help = "Use this IO scheduler on the backing device", - }, -#endif - { - .name = "zonesize", - .type = FIO_OPT_STR_VAL, - .off1 = td_var_offset(zone_size), - .help = "Give size of an IO zone", - .def = "0", - }, - { - .name = "zoneskip", - .type = FIO_OPT_STR_VAL, - .off1 = td_var_offset(zone_skip), - .help = "Space between IO zones", - .def = "0", - }, - { - .name = "lockmem", - .type = FIO_OPT_STR_VAL, - .cb = str_lockmem_cb, - .help = "Lock down this amount of memory", - .def = "0", - }, - { - .name = "rwmixcycle", - .type = FIO_OPT_INT, - .off1 = td_var_offset(rwmixcycle), - .help = "Cycle period for mixed read/write workloads (msec)", - .def = "500", - }, - { - .name = "rwmixread", - .type = FIO_OPT_INT, - .off1 = td_var_offset(rwmixread), - .maxval = 100, - .help = "Percentage of mixed workload that is reads", - .def = "50", - }, - { - .name = "rwmixwrite", - .type = FIO_OPT_INT, - .off1 = td_var_offset(rwmixwrite), - .maxval = 100, - .help = "Percentage of mixed workload that is writes", - .def = "50", - }, - { - .name = "nice", - .type = FIO_OPT_INT, - .off1 = td_var_offset(nice), - .help = "Set job CPU nice value", - .minval = -19, - .maxval = 20, - .def = "0", - }, -#ifdef FIO_HAVE_IOPRIO - { - .name = "prio", - .type = FIO_OPT_INT, - .cb = str_prio_cb, - .help = "Set job IO priority value", - .minval = 0, - .maxval = 7, - }, - { - .name = "prioclass", - .type = FIO_OPT_INT, - .cb = str_prioclass_cb, - .help = "Set job IO priority class", - .minval = 0, - .maxval = 3, - }, -#endif - { - .name = "thinktime", - .type = FIO_OPT_INT, - .off1 = td_var_offset(thinktime), - .help = "Idle time between IO buffers (usec)", - .def = "0", - }, - { - .name = "thinktime_spin", - .type = FIO_OPT_INT, - .off1 = td_var_offset(thinktime_spin), - .help = "Start thinktime by spinning this amount (usec)", - .def = "0", - }, - { - .name = "thinktime_blocks", - .type = FIO_OPT_INT, - .off1 = td_var_offset(thinktime_blocks), - .help = "IO buffer period between 'thinktime'", - .def = "1", - }, - { - .name = "rate", - .type = FIO_OPT_INT, - .off1 = td_var_offset(rate), - .help = "Set bandwidth rate", - }, - { - .name = "ratemin", - .type = FIO_OPT_INT, - .off1 = td_var_offset(ratemin), - .help = "The bottom limit accepted", - }, - { - .name = "ratecycle", - .type = FIO_OPT_INT, - .off1 = td_var_offset(ratecycle), - .help = "Window average for rate limits (msec)", - .def = "1000", - }, - { - .name = "invalidate", - .type = FIO_OPT_BOOL, - .off1 = td_var_offset(invalidate_cache), - .help = "Invalidate buffer/page cache prior to running job", - .def = "1", - }, - { - .name = "sync", - .type = FIO_OPT_BOOL, - .off1 = td_var_offset(sync_io), - .help = "Use O_SYNC for buffered writes", - .def = "0", - }, - { - .name = "bwavgtime", - .type = FIO_OPT_INT, - .off1 = td_var_offset(bw_avg_time), - .help = "Time window over which to calculate bandwidth (msec)", - .def = "500", - }, - { - .name = "create_serialize", - .type = FIO_OPT_BOOL, - .off1 = td_var_offset(create_serialize), - .help = "Serialize creating of job files", - .def = "1", - }, - { - .name = "create_fsync", - .type = FIO_OPT_BOOL, - .off1 = td_var_offset(create_fsync), - .help = "Fsync file after creation", - .def = "1", - }, - { - .name = "cpuload", - .type = FIO_OPT_INT, - .off1 = td_var_offset(cpuload), - .help = "Use this percentage of CPU", - }, - { - .name = "cpuchunks", - .type = FIO_OPT_INT, - .off1 = td_var_offset(cpucycle), - .help = "Length of the CPU burn cycles", - }, -#ifdef FIO_HAVE_CPU_AFFINITY - { - .name = "cpumask", - .type = FIO_OPT_INT, - .cb = str_cpumask_cb, - .help = "CPU affinity mask", - }, -#endif - { - .name = "end_fsync", - .type = FIO_OPT_BOOL, - .off1 = td_var_offset(end_fsync), - .help = "Include fsync at the end of job", - .def = "0", - }, - { - .name = "unlink", - .type = FIO_OPT_BOOL, - .off1 = td_var_offset(unlink), - .help = "Unlink created files after job has completed", - .def = "0", - }, - { - .name = "exitall", - .type = FIO_OPT_STR_SET, - .cb = str_exitall_cb, - .help = "Terminate all jobs when one exits", - }, - { - .name = "stonewall", - .type = FIO_OPT_STR_SET, - .off1 = td_var_offset(stonewall), - .help = "Insert a hard barrier between this job and previous", - }, - { - .name = "thread", - .type = FIO_OPT_STR_SET, - .off1 = td_var_offset(use_thread), - .help = "Use threads instead of forks", - }, - { - .name = "write_bw_log", - .type = FIO_OPT_STR_SET, - .off1 = td_var_offset(write_bw_log), - .help = "Write log of bandwidth during run", - }, - { - .name = "write_lat_log", - .type = FIO_OPT_STR_SET, - .off1 = td_var_offset(write_lat_log), - .help = "Write log of latency during run", - }, - { - .name = "hugepage-size", - .type = FIO_OPT_STR_VAL, - .off1 = td_var_offset(hugepage_size), - .help = "When using hugepages, specify size of each page", - .def = __stringify(FIO_HUGE_PAGE), - }, - { - .name = NULL, - }, -}; +int write_bw_log = 0; +int read_only = 0; + +static int def_timeout; +static int write_lat_log; + +static int prev_group_jobs; -#define FIO_JOB_OPTS (sizeof(options) / sizeof(struct fio_option)) -#define FIO_CMD_OPTS (16) -#define FIO_GETOPT_JOB (0x89988998) +unsigned long fio_debug = 0; +unsigned int fio_debug_jobno = -1; +unsigned int *fio_debug_jobp = NULL; + +static char cmd_optstr[256]; /* * Command line options. These will contain the above, plus a few * extra that only pertain to fio itself and not jobs. */ -static struct option long_options[FIO_JOB_OPTS + FIO_CMD_OPTS] = { +static struct option l_opts[FIO_NR_OPTIONS] = { { .name = "output", .has_arg = required_argument, @@ -584,30 +103,51 @@ static struct option long_options[FIO_JOB_OPTS + FIO_CMD_OPTS] = { .has_arg = optional_argument, .val = 'c', }, + { + .name = "showcmd", + .has_arg = no_argument, + .val = 's', + }, + { + .name = "readonly", + .has_arg = no_argument, + .val = 'r', + }, + { + .name = "eta", + .has_arg = required_argument, + .val = 'e', + }, + { + .name = "debug", + .has_arg = required_argument, + .val = 'd', + }, + { + .name = "section", + .has_arg = required_argument, + .val = 'x', + }, + { + .name = "alloc-size", + .has_arg = required_argument, + .val = 'a', + }, + { + .name = "profile", + .has_arg = required_argument, + .val = 'p', + }, + { + .name = "warnings-fatal", + .has_arg = no_argument, + .val = 'w', + }, { .name = NULL, }, }; -static int def_timeout = 0; - -static char fio_version_string[] = "fio 1.12"; - -static char **ini_file; -static int max_jobs = MAX_JOBS; - -struct thread_data def_thread; -struct thread_data *threads = NULL; - -int exitall_on_terminate = 0; -int terse_output = 0; -unsigned long long mlock_size = 0; -FILE *f_out = NULL; -FILE *f_err = NULL; - -static int write_lat_log = 0; -int write_bw_log = 0; - FILE *get_f_out() { return f_out; @@ -627,12 +167,22 @@ static struct thread_data *get_new_job(int global, struct thread_data *parent) if (global) return &def_thread; - if (thread_number >= max_jobs) + if (thread_number >= max_jobs) { + log_err("error: maximum number of jobs (%d) reached.\n", + max_jobs); return NULL; + } td = &threads[thread_number++]; *td = *parent; + td->o.uid = td->o.gid = -1U; + + dup_files(td, parent); + options_mem_dupe(td); + + profile_add_hooks(td); + td->thread_number = thread_number; return td; } @@ -642,100 +192,225 @@ static void put_job(struct thread_data *td) if (td == &def_thread) return; + profile_td_exit(td); + if (td->error) - fprintf(f_out, "fio: %s\n", td->verror); + log_info("fio: %s\n", td->verror); memset(&threads[td->thread_number - 1], 0, sizeof(*td)); thread_number--; } +static int __setup_rate(struct thread_data *td, enum fio_ddir ddir) +{ + unsigned int bs = td->o.min_bs[ddir]; + unsigned long long bytes_per_sec; + + assert(ddir_rw(ddir)); + + if (td->o.rate[ddir]) + bytes_per_sec = td->o.rate[ddir]; + else + bytes_per_sec = td->o.rate_iops[ddir] * bs; + + if (!bytes_per_sec) { + log_err("rate lower than supported\n"); + return -1; + } + + td->rate_nsec_cycle[ddir] = 1000000000ULL / bytes_per_sec; + td->rate_pending_usleep[ddir] = 0; + return 0; +} + +static int setup_rate(struct thread_data *td) +{ + int ret = 0; + + if (td->o.rate[DDIR_READ] || td->o.rate_iops[DDIR_READ]) + ret = __setup_rate(td, DDIR_READ); + if (td->o.rate[DDIR_WRITE] || td->o.rate_iops[DDIR_WRITE]) + ret |= __setup_rate(td, DDIR_WRITE); + + return ret; +} + +static int fixed_block_size(struct thread_options *o) +{ + return o->min_bs[DDIR_READ] == o->max_bs[DDIR_READ] && + o->min_bs[DDIR_WRITE] == o->max_bs[DDIR_WRITE] && + o->min_bs[DDIR_READ] == o->min_bs[DDIR_WRITE]; +} + /* * Lazy way of fixing up options that depend on each other. We could also * define option callback handlers, but this is easier. */ -static void fixup_options(struct thread_data *td) +static int fixup_options(struct thread_data *td) { - if (!td->rwmixread && td->rwmixwrite) - td->rwmixread = 100 - td->rwmixwrite; - - if (td->write_iolog_file && td->read_iolog_file) { - log_err("fio: read iolog overrides write_iolog\n"); - free(td->write_iolog_file); - td->write_iolog_file = NULL; + struct thread_options *o = &td->o; + int ret = 0; + +#ifndef FIO_HAVE_PSHARED_MUTEX + if (!o->use_thread) { + log_info("fio: this platform does not support process shared" + " mutexes, forcing use of threads. Use the 'thread'" + " option to get rid of this warning.\n"); + o->use_thread = 1; + ret = warnings_fatal; } +#endif - if (td->io_ops->flags & FIO_SYNCIO) - td->iodepth = 1; - else { - if (!td->iodepth) - td->iodepth = td->nr_files; + if (o->write_iolog_file && o->read_iolog_file) { + log_err("fio: read iolog overrides write_iolog\n"); + free(o->write_iolog_file); + o->write_iolog_file = NULL; + ret = warnings_fatal; } /* * only really works for sequential io for now, and with 1 file */ - if (td->zone_size && td_random(td) && td->nr_files == 1) - td->zone_size = 0; + if (o->zone_size && td_random(td) && o->open_files == 1) + o->zone_size = 0; /* * Reads can do overwrites, we always need to pre-create the file */ if (td_read(td) || td_rw(td)) - td->overwrite = 1; + o->overwrite = 1; - if (!td->min_bs[DDIR_READ]) - td->min_bs[DDIR_READ]= td->bs[DDIR_READ]; - if (!td->max_bs[DDIR_READ]) - td->max_bs[DDIR_READ] = td->bs[DDIR_READ]; - if (!td->min_bs[DDIR_WRITE]) - td->min_bs[DDIR_WRITE]= td->bs[DDIR_WRITE]; - if (!td->max_bs[DDIR_WRITE]) - td->max_bs[DDIR_WRITE] = td->bs[DDIR_WRITE]; + if (!o->min_bs[DDIR_READ]) + o->min_bs[DDIR_READ] = o->bs[DDIR_READ]; + if (!o->max_bs[DDIR_READ]) + o->max_bs[DDIR_READ] = o->bs[DDIR_READ]; + if (!o->min_bs[DDIR_WRITE]) + o->min_bs[DDIR_WRITE] = o->bs[DDIR_WRITE]; + if (!o->max_bs[DDIR_WRITE]) + o->max_bs[DDIR_WRITE] = o->bs[DDIR_WRITE]; - td->rw_min_bs = min(td->min_bs[DDIR_READ], td->min_bs[DDIR_WRITE]); + o->rw_min_bs = min(o->min_bs[DDIR_READ], o->min_bs[DDIR_WRITE]); + + /* + * For random IO, allow blockalign offset other than min_bs. + */ + if (!o->ba[DDIR_READ] || !td_random(td)) + o->ba[DDIR_READ] = o->min_bs[DDIR_READ]; + if (!o->ba[DDIR_WRITE] || !td_random(td)) + o->ba[DDIR_WRITE] = o->min_bs[DDIR_WRITE]; + + if ((o->ba[DDIR_READ] != o->min_bs[DDIR_READ] || + o->ba[DDIR_WRITE] != o->min_bs[DDIR_WRITE]) && + !o->norandommap) { + log_err("fio: Any use of blockalign= turns off randommap\n"); + o->norandommap = 1; + ret = warnings_fatal; + } - if (td_read(td) && !td_rw(td)) - td->verify = 0; + if (!o->file_size_high) + o->file_size_high = o->file_size_low; - if (td->norandommap && td->verify != VERIFY_NONE) { - log_err("fio: norandommap given, verify disabled\n"); - td->verify = VERIFY_NONE; + if (o->norandommap && o->verify != VERIFY_NONE + && !fixed_block_size(o)) { + log_err("fio: norandommap given for variable block sizes, " + "verify disabled\n"); + o->verify = VERIFY_NONE; + ret = warnings_fatal; } - if (td->bs_unaligned && (td->odirect || td->io_ops->flags & FIO_RAWIO)) + if (o->bs_unaligned && (o->odirect || td->io_ops->flags & FIO_RAWIO)) log_err("fio: bs_unaligned may not work with raw io\n"); - /* - * O_DIRECT and char doesn't mix, clear that flag if necessary. - */ - if (td->filetype == FIO_TYPE_CHAR && td->odirect) - td->odirect = 0; - /* * thinktime_spin must be less than thinktime */ - if (td->thinktime_spin > td->thinktime) - td->thinktime_spin = td->thinktime; + if (o->thinktime_spin > o->thinktime) + o->thinktime_spin = o->thinktime; /* * The low water mark cannot be bigger than the iodepth */ - if (td->iodepth_low > td->iodepth || !td->iodepth_low) { + if (o->iodepth_low > o->iodepth || !o->iodepth_low) { /* * syslet work around - if the workload is sequential, * we want to let the queue drain all the way down to * avoid seeking between async threads */ if (!strcmp(td->io_ops->name, "syslet-rw") && !td_random(td)) - td->iodepth_low = 1; + o->iodepth_low = 1; else - td->iodepth_low = td->iodepth; + o->iodepth_low = o->iodepth; } /* * If batch number isn't set, default to the same as iodepth */ - if (td->iodepth_batch > td->iodepth || !td->iodepth_batch) - td->iodepth_batch = td->iodepth; + if (o->iodepth_batch > o->iodepth || !o->iodepth_batch) + o->iodepth_batch = o->iodepth; + + if (o->nr_files > td->files_index) + o->nr_files = td->files_index; + + if (o->open_files > o->nr_files || !o->open_files) + o->open_files = o->nr_files; + + if (((o->rate[0] + o->rate[1]) && (o->rate_iops[0] + o->rate_iops[1]))|| + ((o->ratemin[0] + o->ratemin[1]) && (o->rate_iops_min[0] + + o->rate_iops_min[1]))) { + log_err("fio: rate and rate_iops are mutually exclusive\n"); + ret = 1; + } + if ((o->rate[0] < o->ratemin[0]) || (o->rate[1] < o->ratemin[1]) || + (o->rate_iops[0] < o->rate_iops_min[0]) || + (o->rate_iops[1] < o->rate_iops_min[1])) { + log_err("fio: minimum rate exceeds rate\n"); + ret = 1; + } + + if (!o->timeout && o->time_based) { + log_err("fio: time_based requires a runtime/timeout setting\n"); + o->time_based = 0; + ret = warnings_fatal; + } + + if (o->fill_device && !o->size) + o->size = -1ULL; + + if (o->verify != VERIFY_NONE) { + if (td_write(td) && o->do_verify && o->numjobs > 1) { + log_info("Multiple writers may overwrite blocks that " + "belong to other jobs. This can cause " + "verification failures.\n"); + ret = warnings_fatal; + } + + o->refill_buffers = 1; + if (o->max_bs[DDIR_WRITE] != o->min_bs[DDIR_WRITE] && + !o->verify_interval) + o->verify_interval = o->min_bs[DDIR_WRITE]; + } + + if (o->pre_read) { + o->invalidate_cache = 0; + if (td->io_ops->flags & FIO_PIPEIO) { + log_info("fio: cannot pre-read files with an IO engine" + " that isn't seekable. Pre-read disabled.\n"); + ret = warnings_fatal; + } + } + +#ifndef FIO_HAVE_FDATASYNC + if (o->fdatasync_blocks) { + log_info("fio: this platform does not support fdatasync()" + " falling back to using fsync(). Use the 'fsync'" + " option instead of 'fdatasync' to get rid of" + " this warning\n"); + o->fsync_blocks = o->fdatasync_blocks; + o->fdatasync_blocks = 0; + ret = warnings_fatal; + } +#endif + + return ret; } /* @@ -759,6 +434,111 @@ static char *to_kmg(unsigned int val) return buf; } +/* External engines are specified by "external:name.o") */ +static const char *get_engine_name(const char *str) +{ + char *p = strstr(str, ":"); + + if (!p) + return str; + + p++; + strip_blank_front(&p); + strip_blank_end(p); + return p; +} + +static int exists_and_not_file(const char *filename) +{ + struct stat sb; + + if (lstat(filename, &sb) == -1) + return 0; + + /* \\.\ is the device namespace in Windows, where every file + * is a device node */ + if (S_ISREG(sb.st_mode) && strncmp(filename, "\\\\.\\", 4) != 0) + return 0; + + return 1; +} + +static void td_fill_rand_seeds_os(struct thread_data *td) +{ + os_random_seed(td->rand_seeds[0], &td->bsrange_state); + os_random_seed(td->rand_seeds[1], &td->verify_state); + os_random_seed(td->rand_seeds[2], &td->rwmix_state); + + if (td->o.file_service_type == FIO_FSERVICE_RANDOM) + os_random_seed(td->rand_seeds[3], &td->next_file_state); + + os_random_seed(td->rand_seeds[5], &td->file_size_state); + os_random_seed(td->rand_seeds[6], &td->trim_state); + + if (!td_random(td)) + return; + + if (td->o.rand_repeatable) + td->rand_seeds[4] = FIO_RANDSEED * td->thread_number; + + os_random_seed(td->rand_seeds[4], &td->random_state); +} + +static void td_fill_rand_seeds_internal(struct thread_data *td) +{ + init_rand_seed(&td->__bsrange_state, td->rand_seeds[0]); + init_rand_seed(&td->__verify_state, td->rand_seeds[1]); + init_rand_seed(&td->__rwmix_state, td->rand_seeds[2]); + + if (td->o.file_service_type == FIO_FSERVICE_RANDOM) + init_rand_seed(&td->__next_file_state, td->rand_seeds[3]); + + init_rand_seed(&td->__file_size_state, td->rand_seeds[5]); + init_rand_seed(&td->__trim_state, td->rand_seeds[6]); + + if (!td_random(td)) + return; + + if (td->o.rand_repeatable) + td->rand_seeds[4] = FIO_RANDSEED * td->thread_number; + + init_rand_seed(&td->__random_state, td->rand_seeds[4]); +} + +void td_fill_rand_seeds(struct thread_data *td) +{ + if (td->o.use_os_rand) + td_fill_rand_seeds_os(td); + else + td_fill_rand_seeds_internal(td); +} + +/* + * Initialize the various random states we need (random io, block size ranges, + * read/write mix, etc). + */ +static int init_random_state(struct thread_data *td) +{ + int fd; + + fd = open("/dev/urandom", O_RDONLY); + if (fd == -1) { + td_verror(td, errno, "open"); + return 1; + } + + if (read(fd, td->rand_seeds, sizeof(td->rand_seeds)) < + (int) sizeof(td->rand_seeds)) { + td_verror(td, EIO, "read"); + close(fd); + return 1; + } + + close(fd); + td_fill_rand_seeds(td); + return 0; +} + /* * Adds a job to the list of things todo. Sanitizes the various options * to make sure we don't have conflicts, and initializes various @@ -768,9 +548,10 @@ static int add_job(struct thread_data *td, const char *jobname, int job_add_num) { const char *ddir_str[] = { NULL, "read", "write", "rw", NULL, "randread", "randwrite", "randrw" }; - struct stat sb; - int numjobs, i; - struct fio_file *f; + unsigned int i; + const char *engine; + char fname[PATH_MAX]; + int numjobs, file_alloced; /* * the def_thread is just for options, it's not a real job @@ -778,107 +559,112 @@ static int add_job(struct thread_data *td, const char *jobname, int job_add_num) if (td == &def_thread) return 0; - assert(td->io_ops); + /* + * if we are just dumping the output command line, don't add the job + */ + if (dump_cmdline) { + put_job(td); + return 0; + } - if (td->odirect) - td->io_ops->flags |= FIO_RAWIO; + if (profile_td_init(td)) + return 1; - td->filetype = FIO_TYPE_FILE; - if (td->filename && !lstat(td->filename, &sb)) { - if (S_ISBLK(sb.st_mode)) - td->filetype = FIO_TYPE_BD; - else if (S_ISCHR(sb.st_mode)) - td->filetype = FIO_TYPE_CHAR; + engine = get_engine_name(td->o.ioengine); + td->io_ops = load_ioengine(td, engine); + if (!td->io_ops) { + log_err("fio: failed to load engine %s\n", engine); + goto err; } - fixup_options(td); - - if (td->filename) - td->nr_uniq_files = 1; + if (td->o.use_thread) + nr_thread++; else - td->nr_uniq_files = td->nr_files; + nr_process++; - if (td->filetype == FIO_TYPE_FILE || td->filename) { - char tmp[PATH_MAX]; - int len = 0; + if (td->o.odirect) + td->io_ops->flags |= FIO_RAWIO; - if (td->directory && td->directory[0] != '\0') { - if (lstat(td->directory, &sb) < 0) { - log_err("fio: %s is not a directory\n", td->directory); - td_verror(td, errno, "lstat"); - return 1; - } - if (!S_ISDIR(sb.st_mode)) { - log_err("fio: %s is not a directory\n", td->directory); - return 1; + file_alloced = 0; + if (!td->o.filename && !td->files_index && !td->o.read_iolog_file) { + file_alloced = 1; + + if (td->o.nr_files == 1 && exists_and_not_file(jobname)) + add_file(td, jobname); + else { + for (i = 0; i < td->o.nr_files; i++) { + sprintf(fname, "%s.%d.%d", jobname, + td->thread_number, i); + add_file(td, fname); } - len = sprintf(tmp, "%s/", td->directory); } + } - td->files = malloc(sizeof(struct fio_file) * td->nr_files); + if (fixup_options(td)) + goto err; - for_each_file(td, f, i) { - memset(f, 0, sizeof(*f)); - f->fd = -1; + if (td->io_ops->flags & FIO_DISKLESSIO) { + struct fio_file *f; - if (td->filename) - sprintf(tmp + len, "%s", td->filename); - else - sprintf(tmp + len, "%s.%d.%d", jobname, td->thread_number, i); - f->file_name = strdup(tmp); - } - } else { - td->nr_files = 1; - td->files = malloc(sizeof(struct fio_file)); - f = &td->files[0]; - - memset(f, 0, sizeof(*f)); - f->fd = -1; - f->file_name = strdup(jobname); + for_each_file(td, f, i) + f->real_file_size = -1ULL; } - for_each_file(td, f, i) { - f->file_size = td->total_file_size / td->nr_files; - f->file_offset = td->start_offset; - } - - fio_sem_init(&td->mutex, 0); + td->mutex = fio_mutex_init(0); td->ts.clat_stat[0].min_val = td->ts.clat_stat[1].min_val = ULONG_MAX; td->ts.slat_stat[0].min_val = td->ts.slat_stat[1].min_val = ULONG_MAX; + td->ts.lat_stat[0].min_val = td->ts.lat_stat[1].min_val = ULONG_MAX; td->ts.bw_stat[0].min_val = td->ts.bw_stat[1].min_val = ULONG_MAX; + td->ddir_seq_nr = td->o.ddir_seq_nr; - if (td->stonewall && td->thread_number > 1) + if ((td->o.stonewall || td->o.new_group) && prev_group_jobs) { + prev_group_jobs = 0; groupid++; + } td->groupid = groupid; + prev_group_jobs++; + + if (init_random_state(td)) + goto err; if (setup_rate(td)) goto err; - if (td->write_lat_log) { + if (td->o.write_lat_log) { + setup_log(&td->ts.lat_log); setup_log(&td->ts.slat_log); setup_log(&td->ts.clat_log); } - if (td->write_bw_log) + if (td->o.write_bw_log) setup_log(&td->ts.bw_log); - if (!td->name) - td->name = strdup(jobname); + if (!td->o.name) + td->o.name = strdup(jobname); if (!terse_output) { if (!job_add_num) { - if (td->io_ops->flags & FIO_CPUIO) - fprintf(f_out, "%s: ioengine=cpu, cpuload=%u, cpucycle=%u\n", td->name, td->cpuload, td->cpucycle); - else { + if (!strcmp(td->io_ops->name, "cpuio")) { + log_info("%s: ioengine=cpu, cpuload=%u," + " cpucycle=%u\n", td->o.name, + td->o.cpuload, + td->o.cpucycle); + } else { char *c1, *c2, *c3, *c4; - c1 = to_kmg(td->min_bs[DDIR_READ]); - c2 = to_kmg(td->max_bs[DDIR_READ]); - c3 = to_kmg(td->min_bs[DDIR_WRITE]); - c4 = to_kmg(td->max_bs[DDIR_WRITE]); + c1 = to_kmg(td->o.min_bs[DDIR_READ]); + c2 = to_kmg(td->o.max_bs[DDIR_READ]); + c3 = to_kmg(td->o.min_bs[DDIR_WRITE]); + c4 = to_kmg(td->o.max_bs[DDIR_WRITE]); - fprintf(f_out, "%s: (g=%d): rw=%s, bs=%s-%s/%s-%s, ioengine=%s, iodepth=%u\n", td->name, td->groupid, ddir_str[td->td_ddir], c1, c2, c3, c4, td->io_ops->name, td->iodepth); + log_info("%s: (g=%d): rw=%s, bs=%s-%s/%s-%s," + " ioengine=%s, iodepth=%u\n", + td->o.name, td->groupid, + ddir_str[td->o.td_ddir], + c1, c2, c3, c4, + td->io_ops->name, + td->o.iodepth); free(c1); free(c2); @@ -886,210 +672,110 @@ static int add_job(struct thread_data *td, const char *jobname, int job_add_num) free(c4); } } else if (job_add_num == 1) - fprintf(f_out, "...\n"); + log_info("...\n"); } /* * recurse add identical jobs, clear numjobs and stonewall options * as they don't apply to sub-jobs */ - numjobs = td->numjobs; + numjobs = td->o.numjobs; while (--numjobs) { - struct thread_data *td_new = get_new_job(0, td); - - if (!td_new) - goto err; - - td_new->numjobs = 1; - td_new->stonewall = 0; - job_add_num = numjobs - 1; - - if (add_job(td_new, jobname, job_add_num)) - goto err; - } - return 0; -err: - put_job(td); - return -1; -} - -/* - * Initialize the various random states we need (random io, block size ranges, - * read/write mix, etc). - */ -int init_random_state(struct thread_data *td) -{ - unsigned long seeds[5]; - int fd, num_maps, blocks, i; - struct fio_file *f; - - if (td->io_ops->flags & FIO_CPUIO) - return 0; - - fd = open("/dev/urandom", O_RDONLY); - if (fd == -1) { - td_verror(td, errno, "open"); - return 1; - } - - if (read(fd, seeds, sizeof(seeds)) < (int) sizeof(seeds)) { - td_verror(td, EIO, "read"); - close(fd); - return 1; - } - - close(fd); - - os_random_seed(seeds[0], &td->bsrange_state); - os_random_seed(seeds[1], &td->verify_state); - os_random_seed(seeds[2], &td->rwmix_state); - - if (td->file_service_type == FIO_FSERVICE_RANDOM) - os_random_seed(seeds[3], &td->next_file_state); - - if (!td_random(td)) - return 0; - - if (td->rand_repeatable) - seeds[4] = FIO_RANDSEED * td->thread_number; - - if (!td->norandommap) { - for_each_file(td, f, i) { - blocks = (f->real_file_size + td->rw_min_bs - 1) / td->rw_min_bs; - num_maps = (blocks + BLOCKS_PER_MAP-1)/ BLOCKS_PER_MAP; - f->file_map = malloc(num_maps * sizeof(long)); - f->num_maps = num_maps; - memset(f->file_map, 0, num_maps * sizeof(long)); - } - } - - os_random_seed(seeds[4], &td->random_state); - return 0; -} - -static void fill_cpu_mask(os_cpu_mask_t cpumask, int cpu) -{ -#ifdef FIO_HAVE_CPU_AFFINITY - unsigned int i; - - CPU_ZERO(&cpumask); - - for (i = 0; i < sizeof(int) * 8; i++) { - if ((1 << i) & cpu) - CPU_SET(i, &cpumask); - } -#endif -} - -static int is_empty_or_comment(char *line) -{ - unsigned int i; - - for (i = 0; i < strlen(line); i++) { - if (line[i] == ';') - return 1; - if (line[i] == '#') - return 1; - if (!isspace(line[i]) && !iscntrl(line[i])) - return 0; - } - - return 1; -} - -/* - * Check if mmap/mmaphuge has a :/foo/bar/file at the end. If so, return that. - */ -static char *get_mmap_file(const char *str) -{ - char *p = strstr(str, ":"); - - if (!p) - return NULL; - - p++; - strip_blank_front(&p); - strip_blank_end(p); - return strdup(p); -} - -static int str_mem_cb(void *data, const char *mem) -{ - struct thread_data *td = data; - - if (td->mem_type == MEM_MMAPHUGE || td->mem_type == MEM_MMAP) { - td->mmapfile = get_mmap_file(mem); - if (td->mem_type == MEM_MMAPHUGE && !td->mmapfile) { - log_err("fio: mmaphuge:/path/to/file\n"); - return 1; - } - } - - return 0; -} - -/* External engines are specified by "external:name.o") */ -static const char *get_engine_name(const char *str) -{ - char *p = strstr(str, ":"); + struct thread_data *td_new = get_new_job(0, td); - if (!p) - return str; + if (!td_new) + goto err; - p++; - strip_blank_front(&p); - strip_blank_end(p); - return p; -} + td_new->o.numjobs = 1; + td_new->o.stonewall = 0; + td_new->o.new_group = 0; -static int str_ioengine_cb(void *data, const char *str) -{ - struct thread_data *td = data; - const char *name = get_engine_name(str); + if (file_alloced) { + td_new->o.filename = NULL; + td_new->files_index = 0; + td_new->files_size = 0; + td_new->files = NULL; + } - td->io_ops = load_ioengine(td, name); - if (td->io_ops) - return 0; + job_add_num = numjobs - 1; - return 1; -} + if (add_job(td_new, jobname, job_add_num)) + goto err; + } -static int str_lockmem_cb(void fio_unused *data, unsigned long *val) -{ - mlock_size = *val; return 0; +err: + put_job(td); + return -1; } -#ifdef FIO_HAVE_IOPRIO -static int str_prioclass_cb(void *data, unsigned int *val) +/* + * Parse as if 'o' was a command line + */ +void add_job_opts(const char **o) { - struct thread_data *td = data; + struct thread_data *td, *td_parent; + int i, in_global = 1; + char jobname[32]; - td->ioprio |= *val << IOPRIO_CLASS_SHIFT; - return 0; + i = 0; + td_parent = td = NULL; + while (o[i]) { + if (!strncmp(o[i], "name", 4)) { + in_global = 0; + if (td) + add_job(td, jobname, 0); + td = NULL; + sprintf(jobname, "%s", o[i] + 5); + } + if (in_global && !td_parent) + td_parent = get_new_job(1, &def_thread); + else if (!in_global && !td) { + if (!td_parent) + td_parent = &def_thread; + td = get_new_job(0, td_parent); + } + if (in_global) + fio_options_parse(td_parent, (char **) &o[i], 1); + else + fio_options_parse(td, (char **) &o[i], 1); + i++; + } + + if (td) + add_job(td, jobname, 0); } -static int str_prio_cb(void *data, unsigned int *val) +static int skip_this_section(const char *name) { - struct thread_data *td = data; + int i; - td->ioprio |= *val; - return 0; -} -#endif + if (!nr_job_sections) + return 0; + if (!strncmp(name, "global", 6)) + return 0; -static int str_exitall_cb(void) -{ - exitall_on_terminate = 1; - return 0; + for (i = 0; i < nr_job_sections; i++) + if (!strcmp(job_sections[i], name)) + return 0; + + return 1; } -static int str_cpumask_cb(void *data, unsigned int *val) +static int is_empty_or_comment(char *line) { - struct thread_data *td = data; + unsigned int i; - fill_cpu_mask(td->cpumask, *val); - return 0; + for (i = 0; i < strlen(line); i++) { + if (line[i] == ';') + return 1; + if (line[i] == '#') + return 1; + if (!isspace(line[i]) && !iscntrl(line[i])) + return 0; + } + + return 1; } /* @@ -1100,34 +786,80 @@ static int parse_jobs_ini(char *file, int stonewall_flag) unsigned int global; struct thread_data *td; char *string, *name; - fpos_t off; FILE *f; char *p; int ret = 0, stonewall; + int first_sect = 1; + int skip_fgets = 0; + int inside_skip = 0; + char **opts; + int i, alloc_opts, num_opts; + + if (!strcmp(file, "-")) + f = stdin; + else + f = fopen(file, "r"); - f = fopen(file, "r"); if (!f) { perror("fopen job file"); return 1; } string = malloc(4096); - name = malloc(256); - memset(name, 0, 256); + + /* + * it's really 256 + small bit, 280 should suffice + */ + name = malloc(280); + memset(name, 0, 280); + + alloc_opts = 8; + opts = malloc(sizeof(char *) * alloc_opts); + num_opts = 0; stonewall = stonewall_flag; do { - p = fgets(string, 4095, f); - if (!p) - break; + /* + * if skip_fgets is set, we already have loaded a line we + * haven't handled. + */ + if (!skip_fgets) { + p = fgets(string, 4095, f); + if (!p) + break; + } + + skip_fgets = 0; + strip_blank_front(&p); + strip_blank_end(p); + if (is_empty_or_comment(p)) continue; - if (sscanf(p, "[%255s]", name) != 1) + if (sscanf(p, "[%255[^\n]]", name) != 1) { + if (inside_skip) + continue; + log_err("fio: option <%s> outside of [] job section\n", + p); + break; + } + + name[strlen(name) - 1] = '\0'; + + if (skip_this_section(name)) { + inside_skip = 1; continue; + } else + inside_skip = 0; global = !strncmp(name, "global", 6); - name[strlen(name) - 1] = '\0'; + if (dump_cmdline) { + if (first_sect) + log_info("fio "); + if (!global) + log_info("--name=%s ", name); + first_sect = 0; + } td = get_new_job(global, &def_thread); if (!td) { @@ -1139,44 +871,68 @@ static int parse_jobs_ini(char *file, int stonewall_flag) * Seperate multiple job files by a stonewall */ if (!global && stonewall) { - td->stonewall = stonewall; + td->o.stonewall = stonewall; stonewall = 0; } - fgetpos(f, &off); + num_opts = 0; + memset(opts, 0, alloc_opts * sizeof(char *)); + while ((p = fgets(string, 4096, f)) != NULL) { if (is_empty_or_comment(p)) continue; strip_blank_front(&p); - if (p[0] == '[') + /* + * new section, break out and make sure we don't + * fgets() a new line at the top. + */ + if (p[0] == '[') { + skip_fgets = 1; break; + } strip_blank_end(p); - fgetpos(f, &off); + if (num_opts == alloc_opts) { + alloc_opts <<= 1; + opts = realloc(opts, + alloc_opts * sizeof(char *)); + } - /* - * Don't break here, continue parsing options so we - * dump all the bad ones. Makes trial/error fixups - * easier on the user. - */ - ret |= parse_option(p, options, td); + opts[num_opts] = strdup(p); + num_opts++; } + ret = fio_options_parse(td, opts, num_opts); if (!ret) { - fsetpos(f, &off); + if (dump_cmdline) + for (i = 0; i < num_opts; i++) + log_info("--%s ", opts[i]); + ret = add_job(td, name, 0); } else { log_err("fio: job %s dropped\n", name); put_job(td); } + + for (i = 0; i < num_opts; i++) + free(opts[i]); + num_opts = 0; } while (!ret); + if (dump_cmdline) + log_info("\n"); + + for (i = 0; i < num_opts; i++) + free(opts[i]); + free(string); free(name); - fclose(f); + free(opts); + if (f != stdin) + fclose(f); return ret; } @@ -1184,30 +940,87 @@ static int fill_def_thread(void) { memset(&def_thread, 0, sizeof(def_thread)); - if (fio_getaffinity(getpid(), &def_thread.cpumask) == -1) { - perror("sched_getaffinity"); - return 1; - } + fio_getaffinity(getpid(), &def_thread.o.cpumask); /* * fill default options */ - fill_default_options(&def_thread, options); + fio_fill_default_options(&def_thread); - def_thread.timeout = def_timeout; - def_thread.write_bw_log = write_bw_log; - def_thread.write_lat_log = write_lat_log; + def_thread.o.timeout = def_timeout; + return 0; +} -#ifdef FIO_HAVE_DISK_UTIL - def_thread.do_disk_util = 1; -#endif +static void free_shm(void) +{ + struct shmid_ds sbuf; + + if (threads) { + void *tp = threads; + + threads = NULL; + file_hash_exit(); + fio_debug_jobp = NULL; + shmdt(tp); + shmctl(shm_id, IPC_RMID, &sbuf); + } + + scleanup(); +} + +/* + * The thread area is shared between the main process and the job + * threads/processes. So setup a shared memory segment that will hold + * all the job info. We use the end of the region for keeping track of + * open files across jobs, for file sharing. + */ +static int setup_thread_area(void) +{ + void *hash; + + /* + * 1024 is too much on some machines, scale max_jobs if + * we get a failure that looks like too large a shm segment + */ + do { + size_t size = max_jobs * sizeof(struct thread_data); + + size += file_hash_size; + size += sizeof(unsigned int); + + shm_id = shmget(0, size, IPC_CREAT | 0600); + if (shm_id != -1) + break; + if (errno != EINVAL) { + perror("shmget"); + break; + } + + max_jobs >>= 1; + } while (max_jobs); + + if (shm_id == -1) + return 1; + + threads = shmat(shm_id, NULL, 0); + if (threads == (void *) -1) { + perror("shmat"); + return 1; + } + memset(threads, 0, max_jobs * sizeof(struct thread_data)); + hash = (void *) threads + max_jobs * sizeof(struct thread_data); + fio_debug_jobp = (void *) hash + file_hash_size; + *fio_debug_jobp = -1; + file_hash_init(hash); + atexit(free_shm); return 0; } -static void usage(void) +static void usage(const char *name) { - printf("%s\n", fio_version_string); + printf("%s [options] [job options] \n", name); + printf("\t--debug=options\tEnable debug logging\n"); printf("\t--output\tWrite output to file\n"); printf("\t--timeout\tRuntime in seconds\n"); printf("\t--latency-log\tGenerate per-job latency logs\n"); @@ -1215,23 +1028,138 @@ static void usage(void) printf("\t--minimal\tMinimal (terse) output\n"); printf("\t--version\tPrint version info and exit\n"); printf("\t--help\t\tPrint this page\n"); - printf("\t--cmdhelp=cmd\tPrint command help, \"all\" for all of them\n"); + printf("\t--cmdhelp=cmd\tPrint command help, \"all\" for all of" + " them\n"); + printf("\t--showcmd\tTurn a job file into command line options\n"); + printf("\t--eta=when\tWhen ETA estimate should be printed\n"); + printf("\t \tMay be \"always\", \"never\" or \"auto\"\n"); + printf("\t--readonly\tTurn on safety read-only checks, preventing" + " writes\n"); + printf("\t--section=name\tOnly run specified section in job file\n"); + printf("\t--alloc-size=kb\tSet smalloc pool to this size in kb" + " (def 1024)\n"); + printf("\t--warnings-fatal Fio parser warnings are fatal\n"); + printf("\nFio was written by Jens Axboe "); + printf("\n Jens Axboe \n"); +} + +#ifdef FIO_INC_DEBUG +struct debug_level debug_levels[] = { + { .name = "process", .shift = FD_PROCESS, }, + { .name = "file", .shift = FD_FILE, }, + { .name = "io", .shift = FD_IO, }, + { .name = "mem", .shift = FD_MEM, }, + { .name = "blktrace", .shift = FD_BLKTRACE }, + { .name = "verify", .shift = FD_VERIFY }, + { .name = "random", .shift = FD_RANDOM }, + { .name = "parse", .shift = FD_PARSE }, + { .name = "diskutil", .shift = FD_DISKUTIL }, + { .name = "job", .shift = FD_JOB }, + { .name = "mutex", .shift = FD_MUTEX }, + { .name = "profile", .shift = FD_PROFILE }, + { .name = "time", .shift = FD_TIME }, + { .name = NULL, }, +}; + +static int set_debug(const char *string) +{ + struct debug_level *dl; + char *p = (char *) string; + char *opt; + int i; + + if (!strcmp(string, "?") || !strcmp(string, "help")) { + log_info("fio: dumping debug options:"); + for (i = 0; debug_levels[i].name; i++) { + dl = &debug_levels[i]; + log_info("%s,", dl->name); + } + log_info("all\n"); + return 1; + } + + while ((opt = strsep(&p, ",")) != NULL) { + int found = 0; + + if (!strncmp(opt, "all", 3)) { + log_info("fio: set all debug options\n"); + fio_debug = ~0UL; + continue; + } + + for (i = 0; debug_levels[i].name; i++) { + dl = &debug_levels[i]; + found = !strncmp(opt, dl->name, strlen(dl->name)); + if (!found) + continue; + + if (dl->shift == FD_JOB) { + opt = strchr(opt, ':'); + if (!opt) { + log_err("fio: missing job number\n"); + break; + } + opt++; + fio_debug_jobno = atoi(opt); + log_info("fio: set debug jobno %d\n", + fio_debug_jobno); + } else { + log_info("fio: set debug option %s\n", opt); + fio_debug |= (1UL << dl->shift); + } + break; + } + + if (!found) + log_err("fio: debug mask %s not found\n", opt); + } + return 0; +} +#else +static int set_debug(const char *string) +{ + log_err("fio: debug tracing not included in build\n"); + return 1; +} +#endif + +static void fio_options_fill_optstring(void) +{ + char *ostr = cmd_optstr; + int i, c; + + c = i = 0; + while (l_opts[i].name) { + ostr[c++] = l_opts[i].val; + if (l_opts[i].has_arg == required_argument) + ostr[c++] = ':'; + else if (l_opts[i].has_arg == optional_argument) { + ostr[c++] = ':'; + ostr[c++] = ':'; + } + i++; + } + ostr[c] = '\0'; } static int parse_cmd_line(int argc, char *argv[]) { struct thread_data *td = NULL; - int c, ini_idx = 0, lidx, ret, dont_add_job = 0; + int c, ini_idx = 0, lidx, ret = 0, do_exit = 0, exit_val = 0; + char *ostr = cmd_optstr; - while ((c = getopt_long(argc, argv, "", long_options, &lidx)) != -1) { + while ((c = getopt_long_only(argc, argv, ostr, l_opts, &lidx)) != -1) { switch (c) { + case 'a': + smalloc_pool_size = atoi(optarg); + break; case 't': def_timeout = atoi(optarg); break; case 'l': write_lat_log = 1; break; - case 'w': + case 'b': write_bw_log = 1; break; case 'o': @@ -1246,20 +1174,54 @@ static int parse_cmd_line(int argc, char *argv[]) terse_output = 1; break; case 'h': - usage(); + usage(argv[0]); exit(0); case 'c': - ret = show_cmd_help(options, optarg); - exit(ret); + exit(fio_show_option_help(optarg)); + case 's': + dump_cmdline = 1; + break; + case 'r': + read_only = 1; + break; case 'v': - printf("%s\n", fio_version_string); + log_info("%s\n", fio_version_string); exit(0); + case 'e': + if (!strcmp("always", optarg)) + eta_print = FIO_ETA_ALWAYS; + else if (!strcmp("never", optarg)) + eta_print = FIO_ETA_NEVER; + break; + case 'd': + if (set_debug(optarg)) + do_exit++; + break; + case 'x': { + size_t new_size; + + if (!strcmp(optarg, "global")) { + log_err("fio: can't use global as only " + "section\n"); + do_exit++; + exit_val = 1; + break; + } + new_size = (nr_job_sections + 1) * sizeof(char *); + job_sections = realloc(job_sections, new_size); + job_sections[nr_job_sections] = strdup(optarg); + nr_job_sections++; + break; + } + case 'p': + exec_profile = strdup(optarg); + break; case FIO_GETOPT_JOB: { - const char *opt = long_options[lidx].name; + const char *opt = l_opts[lidx].name; char *val = optarg; if (!strncmp(opt, "name", 4) && td) { - ret = add_job(td, td->name ?: "fio", 0); + ret = add_job(td, td->o.name ?: "fio", 0); if (ret) { put_job(td); return 0; @@ -1267,31 +1229,41 @@ static int parse_cmd_line(int argc, char *argv[]) td = NULL; } if (!td) { - int global = !strncmp(val, "global", 6); + int is_section = !strncmp(opt, "name", 4); + int global = 0; + + if (!is_section || !strncmp(val, "global", 6)) + global = 1; + + if (is_section && skip_this_section(val)) + continue; td = get_new_job(global, &def_thread); if (!td) return 0; } - ret = parse_cmd_option(opt, val, options, td); - if (ret) - dont_add_job = 1; + ret = fio_cmd_option_parse(td, opt, val); break; } + case 'w': + warnings_fatal = 1; + break; default: + do_exit++; + exit_val = 1; break; } } + if (do_exit) + exit(exit_val); + if (td) { - if (dont_add_job) + if (!ret) + ret = add_job(td, td->o.name ?: "fio", 0); + if (ret) put_job(td); - else { - ret = add_job(td, td->name ?: "fio", 0); - if (ret) - put_job(td); - } } while (optind < argc) { @@ -1304,83 +1276,6 @@ static int parse_cmd_line(int argc, char *argv[]) return ini_idx; } -static void free_shm(void) -{ - struct shmid_ds sbuf; - - if (threads) { - shmdt((void *) threads); - threads = NULL; - shmctl(shm_id, IPC_RMID, &sbuf); - } -} - -/* - * The thread area is shared between the main process and the job - * threads/processes. So setup a shared memory segment that will hold - * all the job info. - */ -static int setup_thread_area(void) -{ - /* - * 1024 is too much on some machines, scale max_jobs if - * we get a failure that looks like too large a shm segment - */ - do { - size_t size = max_jobs * sizeof(struct thread_data); - - shm_id = shmget(0, size, IPC_CREAT | 0600); - if (shm_id != -1) - break; - if (errno != EINVAL) { - perror("shmget"); - break; - } - - max_jobs >>= 1; - } while (max_jobs); - - if (shm_id == -1) - return 1; - - threads = shmat(shm_id, NULL, 0); - if (threads == (void *) -1) { - perror("shmat"); - return 1; - } - - atexit(free_shm); - return 0; -} - -/* - * Copy the fio options into the long options map, so we mirror - * job and cmd line options. - */ -static void dupe_job_options(void) -{ - struct fio_option *o; - unsigned int i; - - i = 0; - while (long_options[i].name) - i++; - - o = &options[0]; - while (o->name) { - long_options[i].name = 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; - - i++; - o++; - assert(i < FIO_JOB_OPTS + FIO_CMD_OPTS); - } -} - int parse_options(int argc, char *argv[]) { int job_files, i; @@ -1388,9 +1283,8 @@ int parse_options(int argc, char *argv[]) f_out = stdout; f_err = stderr; - options_init(options); - - dupe_job_options(); + fio_options_fill_optstring(); + fio_options_dup_and_init(l_opts); if (setup_thread_area()) return 1; @@ -1408,11 +1302,25 @@ int parse_options(int argc, char *argv[]) } free(ini_file); + options_mem_free(&def_thread); if (!thread_number) { - log_err("No jobs defined(s)\n"); + if (dump_cmdline) + return 0; + if (exec_profile) + return 0; + + log_err("No jobs(s) defined\n\n"); + usage(argv[0]); return 1; } + if (def_thread.o.gtod_offload) { + fio_gtod_init(); + fio_gtod_offload = 1; + fio_gtod_cpu = def_thread.o.gtod_cpu; + } + + log_info("%s\n", fio_version_string); return 0; }