X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=options.c;h=fd91eede7a2ae8a90fe78d59284627235fbecb7e;hp=06ac6596292851ed0f95a4df8d096a513076026b;hb=e90a0adf85f75a65c02b22ba3766dff0d9d82ae6;hpb=f20485ad77872d6c5084dead547420386a6ecd5e

diff --git a/options.c b/options.c
index 06ac6596..fd91eede 100644
--- a/options.c
+++ b/options.c
@@ -102,7 +102,7 @@ static int bssplit_ddir(struct thread_data *td, int ddir, char *str)
 		} else
 			perc = -1;
 
-		if (str_to_decimal(fname, &val, 1, 1, td)) {
+		if (str_to_decimal(fname, &val, 1, td)) {
 			log_err("fio: bssplit conversion failed\n");
 			free(td->o.bssplit);
 			return 1;
@@ -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;
@@ -213,7 +325,7 @@ static int str_rw_cb(void *data, const char *str)
 	else {
 		long long val;
 
-		if (str_to_decimal(nr, &val, 1, 0, td)) {
+		if (str_to_decimal(nr, &val, 1, td)) {
 			log_err("fio: rw postfix parsing failed\n");
 			free(nr);
 			return 1;
@@ -226,32 +338,12 @@ static int str_rw_cb(void *data, const char *str)
 	return 0;
 }
 
-#ifdef FIO_HAVE_LIBAIO
-static int str_libaio_cb(void *data, const char *str)
-{
-	struct thread_data *td = data;
-
-	if (!strcmp(str, "userspace_reap")) {
-		td->o.userspace_libaio_reap = 1;
-		return 0;
-	}
-
-	log_err("fio: bad libaio sub-option: %s\n", str);
-	return 1;
-}
-#endif
-
 static int str_mem_cb(void *data, const char *mem)
 {
 	struct thread_data *td = data;
 
-	if (td->o.mem_type == MEM_MMAPHUGE || td->o.mem_type == MEM_MMAP) {
+	if (td->o.mem_type == MEM_MMAPHUGE || td->o.mem_type == MEM_MMAP)
 		td->mmapfile = get_opt_postfix(mem);
-		if (td->o.mem_type == MEM_MMAPHUGE && !td->mmapfile) {
-			log_err("fio: mmaphuge:/path/to/file\n");
-			return 1;
-		}
-	}
 
 	return 0;
 }
@@ -260,12 +352,9 @@ 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;
@@ -276,13 +365,8 @@ 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();
-	return 0;
-}
-
-static int str_lockmem_cb(void fio_unused *data, unsigned long long *val)
-{
-	mlock_size = *val;
+	fio_clock_source_set = 1;
+	fio_clock_init();
 	return 0;
 }
 
@@ -470,13 +554,127 @@ 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)
+#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.trim_percentage = *val;
+	td->o.numa_memmask_set = 1;
 	return 0;
+
+out:
+	return 1;
 }
 #endif
 
@@ -494,7 +692,7 @@ 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;
@@ -510,6 +708,45 @@ 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
@@ -653,19 +890,6 @@ 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 long long *off)
-{
-	struct thread_data *td = data;
-
-	if (*off && *off < sizeof(struct verify_header)) {
-		log_err("fio: verify_offset too small\n");
-		return 1;
-	}
-
-	td->o.verify_offset = *off;
-	return 0;
-}
-
 static int str_verify_pattern_cb(void *data, const char *input)
 {
 	struct thread_data *td = data;
@@ -704,25 +928,31 @@ 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;
-}
-
-static int str_lockfile_cb(void *data, const char *str)
-{
-	struct thread_data *td = data;
-	char *nr = get_opt_postfix(str);
-
-	td->o.lockfile_batch = 1;
-	if (nr) {
-		td->o.lockfile_batch = atoi(nr);
-		free(nr);
-	}
 
 	return 0;
 }
@@ -749,6 +979,17 @@ 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;
@@ -758,6 +999,7 @@ static int str_gtod_reduce_cb(void *data, int *il)
 	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;
 
@@ -816,70 +1058,225 @@ static int gtod_cpu_verify(struct fio_option *o, void *data)
 	return 0;
 }
 
-static int kb_base_verify(struct fio_option *o, void *data)
+/*
+ * Option grouping
+ */
+static struct opt_group fio_opt_groups[] = {
+	{
+		.name	= "General",
+		.mask	= FIO_OPT_C_GENERAL,
+	},
+	{
+		.name	= "I/O",
+		.mask	= FIO_OPT_C_IO,
+	},
+	{
+		.name	= "File",
+		.mask	= FIO_OPT_C_FILE,
+	},
+	{
+		.name	= "Statistics",
+		.mask	= FIO_OPT_C_STAT,
+	},
+	{
+		.name	= "Logging",
+		.mask	= FIO_OPT_C_LOG,
+	},
+	{
+		.name	= "Profiles",
+		.mask	= FIO_OPT_C_PROFILE,
+	},
+	{
+		.name	= NULL,
+	},
+};
+
+static struct opt_group *__opt_group_from_mask(struct opt_group *ogs, unsigned int *mask,
+					       unsigned int inv_mask)
 {
-	struct thread_data *td = data;
+	struct opt_group *og;
+	int i;
 
-	if (td->o.kb_base != 1024 && td->o.kb_base != 1000) {
-		log_err("fio: kb_base set to nonsensical value: %u\n",
-				td->o.kb_base);
-		return 1;
+	if (*mask == inv_mask || !*mask)
+		return NULL;
+
+	for (i = 0; ogs[i].name; i++) {
+		og = &ogs[i];
+
+		if (*mask & og->mask) {
+			*mask &= ~(og->mask);
+			return og;
+		}
 	}
 
-	return 0;
+	return NULL;
 }
 
-#define __stringify_1(x)	#x
-#define __stringify(x)		__stringify_1(x)
+struct opt_group *opt_group_from_mask(unsigned int *mask)
+{
+	return __opt_group_from_mask(fio_opt_groups, mask, FIO_OPT_C_INVALID);
+}
+
+static struct opt_group fio_opt_cat_groups[] = {
+	{
+		.name	= "Rate",
+		.mask	= FIO_OPT_G_RATE,
+	},
+	{
+		.name	= "Zone",
+		.mask	= FIO_OPT_G_ZONE,
+	},
+	{
+		.name	= "Read/write mix",
+		.mask	= FIO_OPT_G_RWMIX,
+	},
+	{
+		.name	= "Verify",
+		.mask	= FIO_OPT_G_VERIFY,
+	},
+	{
+		.name	= "Trim",
+		.mask	= FIO_OPT_G_TRIM,
+	},
+	{
+		.name	= "I/O Logging",
+		.mask	= FIO_OPT_G_IOLOG,
+	},
+	{
+		.name	= "I/O Depth",
+		.mask	= FIO_OPT_G_IO_DEPTH,
+	},
+	{
+		.name	= "I/O Flow",
+		.mask	= FIO_OPT_G_IO_FLOW,
+	},
+	{
+		.name	= "Description",
+		.mask	= FIO_OPT_G_DESC,
+	},
+	{
+		.name	= "Filename",
+		.mask	= FIO_OPT_G_FILENAME,
+	},
+	{
+		.name	= "General I/O",
+		.mask	= FIO_OPT_G_IO_BASIC,
+	},
+	{
+		.name	= "Cgroups",
+		.mask	= FIO_OPT_G_CGROUP,
+	},
+	{
+		.name	= "Runtime",
+		.mask	= FIO_OPT_G_RUNTIME,
+	},
+	{
+		.name	= "Process",
+		.mask	= FIO_OPT_G_PROCESS,
+	},
+	{
+		.name	= "Job credentials / priority",
+		.mask	= FIO_OPT_G_CRED,
+	},
+	{
+		.name	= "Clock settings",
+		.mask	= FIO_OPT_G_CLOCK,
+	},
+	{
+		.name	= "I/O Type",
+		.mask	= FIO_OPT_G_IO_TYPE,
+	},
+	{
+		.name	= "I/O Thinktime",
+		.mask	= FIO_OPT_G_THINKTIME,
+	},
+	{
+		.name	= "Randomizations",
+		.mask	= FIO_OPT_G_RANDOM,
+	},
+	{
+		.name	= "I/O buffers",
+		.mask	= FIO_OPT_G_IO_BUF,
+	},
+	{
+		.name	= "Tiobench profile",
+		.mask	= FIO_OPT_G_TIOBENCH,
+	},
+
+	{
+		.name	= NULL,
+	}
+};
+
+struct opt_group *opt_group_cat_from_mask(unsigned int *mask)
+{
+	return __opt_group_from_mask(fio_opt_cat_groups, mask, FIO_OPT_G_INVALID);
+}
 
 /*
  * Map of job/command line options
  */
-static struct fio_option options[FIO_MAX_OPTS] = {
+static struct fio_option fio_options[FIO_MAX_OPTS] = {
 	{
 		.name	= "description",
+		.lname	= "Description of job",
 		.type	= FIO_OPT_STR_STORE,
 		.off1	= td_var_offset(description),
 		.help	= "Text job description",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_DESC,
 	},
 	{
 		.name	= "name",
+		.lname	= "Job name",
 		.type	= FIO_OPT_STR_STORE,
 		.off1	= td_var_offset(name),
 		.help	= "Name of this job",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_DESC,
+	},
+	{
+		.name	= "filename",
+		.lname	= "Filename(s)",
+		.type	= FIO_OPT_STR_STORE,
+		.off1	= td_var_offset(filename),
+		.cb	= str_filename_cb,
+		.prio	= -1, /* must come after "directory" */
+		.help	= "File(s) to use for the workload",
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_FILENAME,
 	},
 	{
 		.name	= "directory",
+		.lname	= "Directory",
 		.type	= FIO_OPT_STR_STORE,
 		.off1	= td_var_offset(directory),
 		.cb	= str_directory_cb,
 		.help	= "Directory to store files in",
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_FILENAME,
 	},
 	{
-		.name	= "filename",
+		.name	= "filename_format",
 		.type	= FIO_OPT_STR_STORE,
-		.off1	= td_var_offset(filename),
-		.cb	= str_filename_cb,
+		.off1	= td_var_offset(filename_format),
 		.prio	= -1, /* must come after "directory" */
-		.help	= "File(s) to use for the workload",
-	},
-	{
-		.name	= "kb_base",
-		.type	= FIO_OPT_INT,
-		.off1	= td_var_offset(kb_base),
-		.verify	= kb_base_verify,
-		.prio	= 1,
-		.def	= "1024",
-		.help	= "How many bytes per KB for reporting (1000 or 1024)",
+		.help	= "Override default $jobname.$jobnum.$filenum naming",
+		.def	= "$jobname.$jobnum.$filenum",
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_FILENAME,
 	},
 	{
 		.name	= "lockfile",
+		.lname	= "Lockfile",
 		.type	= FIO_OPT_STR,
-		.cb	= str_lockfile_cb,
 		.off1	= td_var_offset(file_lock_mode),
 		.help	= "Lock file when doing IO to it",
 		.parent	= "filename",
+		.hide	= 0,
 		.def	= "none",
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_FILENAME,
 		.posval = {
 			  { .ival = "none",
 			    .oval = FILE_LOCK_NONE,
@@ -898,13 +1295,17 @@ static struct fio_option options[FIO_MAX_OPTS] = {
 	},
 	{
 		.name	= "opendir",
+		.lname	= "Open directory",
 		.type	= FIO_OPT_STR_STORE,
 		.off1	= td_var_offset(opendir),
 		.cb	= str_opendir_cb,
 		.help	= "Recursively add files from this directory and down",
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_FILENAME,
 	},
 	{
 		.name	= "rw",
+		.lname	= "Read/write",
 		.alias	= "readwrite",
 		.type	= FIO_OPT_STR,
 		.cb	= str_rw_cb,
@@ -912,6 +1313,8 @@ static struct fio_option options[FIO_MAX_OPTS] = {
 		.help	= "IO direction",
 		.def	= "read",
 		.verify	= rw_verify,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IO_BASIC,
 		.posval = {
 			  { .ival = "read",
 			    .oval = TD_DDIR_READ,
@@ -921,6 +1324,10 @@ 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",
@@ -929,10 +1336,18 @@ 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"
@@ -941,10 +1356,13 @@ static struct fio_option options[FIO_MAX_OPTS] = {
 	},
 	{
 		.name	= "rw_sequencer",
+		.lname	= "RW Sequencer",
 		.type	= FIO_OPT_STR,
 		.off1	= td_var_offset(rw_seq),
 		.help	= "IO offset generator modifier",
 		.def	= "sequential",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IO_BASIC,
 		.posval = {
 			  { .ival = "sequential",
 			    .oval = RW_SEQ_SEQ,
@@ -959,10 +1377,13 @@ static struct fio_option options[FIO_MAX_OPTS] = {
 
 	{
 		.name	= "ioengine",
+		.lname	= "IO Engine",
 		.type	= FIO_OPT_STR_STORE,
 		.off1	= td_var_offset(ioengine),
 		.help	= "IO engine to use",
 		.def	= FIO_PREFERRED_ENGINE,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IO_BASIC,
 		.posval	= {
 			  { .ival = "sync",
 			    .help = "Use read/write",
@@ -973,31 +1394,30 @@ 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",
-			    .cb   = str_libaio_cb,
 			  },
 #endif
-#ifdef FIO_HAVE_POSIXAIO
+#ifdef CONFIG_POSIXAIO
 			  { .ival = "posixaio",
 			    .help = "POSIX asynchronous IO",
 			  },
 #endif
-#ifdef FIO_HAVE_SOLARISAIO
+#ifdef CONFIG_SOLARISAIO
 			  { .ival = "solarisaio",
 			    .help = "Solaris native asynchronous IO",
 			  },
 #endif
-#ifdef FIO_HAVE_WINDOWSAIO
+#ifdef CONFIG_WINDOWSAIO
 			  { .ival = "windowsaio",
 			    .help = "Windows native asynchronous IO"
-		  	  },
+			  },
 #endif
 			  { .ival = "mmap",
 			    .help = "Memory mapped IO"
 			  },
-#ifdef FIO_HAVE_SPLICE
+#ifdef CONFIG_LINUX_SPLICE
 			  { .ival = "splice",
 			    .help = "splice/vmsplice based IO",
 			  },
@@ -1016,15 +1436,10 @@ 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",
 			  },
@@ -1034,10 +1449,25 @@ 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 CONFIG_LINUX_EXT4_MOVE_EXTENT
+			  { .ival = "e4defrag",
+			    .help = "ext4 defrag engine",
+			  },
+#endif
+#ifdef CONFIG_LINUX_FALLOCATE
+			  { .ival = "falloc",
+			    .help = "fallocate() file based engine",
+			  },
 #endif
 			  { .ival = "external",
 			    .help = "Load external engine (append name)",
@@ -1046,168 +1476,312 @@ static struct fio_option options[FIO_MAX_OPTS] = {
 	},
 	{
 		.name	= "iodepth",
+		.lname	= "IO Depth",
 		.type	= FIO_OPT_INT,
 		.off1	= td_var_offset(iodepth),
 		.help	= "Number of IO buffers to keep in flight",
 		.minval = 1,
+		.interval = 1,
 		.def	= "1",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IO_BASIC,
 	},
 	{
 		.name	= "iodepth_batch",
+		.lname	= "IO Depth batch",
 		.alias	= "iodepth_batch_submit",
 		.type	= FIO_OPT_INT,
 		.off1	= td_var_offset(iodepth_batch),
 		.help	= "Number of IO buffers to submit in one go",
 		.parent	= "iodepth",
+		.hide	= 1,
 		.minval	= 1,
+		.interval = 1,
 		.def	= "1",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IO_BASIC,
 	},
 	{
 		.name	= "iodepth_batch_complete",
+		.lname	= "IO Depth batch complete",
 		.type	= FIO_OPT_INT,
 		.off1	= td_var_offset(iodepth_batch_complete),
 		.help	= "Number of IO buffers to retrieve in one go",
 		.parent	= "iodepth",
+		.hide	= 1,
 		.minval	= 0,
+		.interval = 1,
 		.def	= "1",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IO_BASIC,
 	},
 	{
 		.name	= "iodepth_low",
+		.lname	= "IO Depth batch low",
 		.type	= FIO_OPT_INT,
 		.off1	= td_var_offset(iodepth_low),
 		.help	= "Low water mark for queuing depth",
 		.parent	= "iodepth",
+		.hide	= 1,
+		.interval = 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IO_BASIC,
 	},
 	{
 		.name	= "size",
+		.lname	= "Size",
 		.type	= FIO_OPT_STR_VAL,
 		.cb	= str_size_cb,
 		.help	= "Total size of device or files",
+		.interval = 1024 * 1024,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "fill_device",
+		.lname	= "Fill device",
 		.alias	= "fill_fs",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(fill_device),
 		.help	= "Write until an ENOSPC error occurs",
 		.def	= "0",
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "filesize",
+		.lname	= "File size",
 		.type	= FIO_OPT_STR_VAL,
 		.off1	= td_var_offset(file_size_low),
 		.off2	= td_var_offset(file_size_high),
 		.minval = 1,
 		.help	= "Size of individual files",
+		.interval = 1024 * 1024,
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "offset",
+		.lname	= "IO offset",
 		.alias	= "fileoffset",
 		.type	= FIO_OPT_STR_VAL,
 		.off1	= td_var_offset(start_offset),
 		.help	= "Start IO from this offset",
 		.def	= "0",
+		.interval = 1024 * 1024,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_INVALID,
+	},
+	{
+		.name	= "offset_increment",
+		.lname	= "IO 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",
+		.hide	= 1,
+		.def	= "0",
+		.interval = 1024 * 1024,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "bs",
+		.lname	= "Block size",
 		.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",
 		.parent = "rw",
+		.hide	= 1,
+		.interval = 512,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "ba",
+		.lname	= "Block size align",
 		.alias	= "blockalign",
 		.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",
+		.hide	= 1,
+		.interval = 512,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "bsrange",
+		.lname	= "Block size range",
 		.alias	= "blocksize_range",
 		.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]),
+		.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",
+		.hide	= 1,
+		.interval = 4096,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "bssplit",
+		.lname	= "Block size split",
 		.type	= FIO_OPT_STR,
 		.cb	= str_bssplit_cb,
 		.help	= "Set a specific mix of block sizes",
 		.parent	= "rw",
+		.hide	= 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "bs_unaligned",
+		.lname	= "Block size unaligned",
 		.alias	= "blocksize_unaligned",
 		.type	= FIO_OPT_STR_SET,
 		.off1	= td_var_offset(bs_unaligned),
 		.help	= "Don't sector align IO buffer sizes",
 		.parent = "rw",
+		.hide	= 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "randrepeat",
+		.lname	= "Random repeatable",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(rand_repeatable),
 		.help	= "Use repeatable random IO pattern",
 		.def	= "1",
 		.parent = "rw",
+		.hide	= 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_RANDOM,
 	},
 	{
 		.name	= "use_os_rand",
+		.lname	= "Use OS random",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(use_os_rand),
 		.help	= "Set to use OS random generator",
 		.def	= "0",
 		.parent = "rw",
+		.hide	= 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_RANDOM,
 	},
 	{
 		.name	= "norandommap",
+		.lname	= "No randommap",
 		.type	= FIO_OPT_STR_SET,
 		.off1	= td_var_offset(norandommap),
 		.help	= "Accept potential duplicate random blocks",
 		.parent = "rw",
+		.hide	= 1,
+		.hide_on_set = 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_RANDOM,
 	},
 	{
 		.name	= "softrandommap",
+		.lname	= "Soft randommap",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(softrandommap),
 		.help	= "Set norandommap if randommap allocation fails",
 		.parent	= "norandommap",
+		.hide	= 1,
 		.def	= "0",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_RANDOM,
+	},
+	{
+		.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",
+			  },
+		},
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_RANDOM,
+	},
+	{
+		.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",
+			  },
+		},
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_RANDOM,
 	},
 	{
 		.name	= "nrfiles",
+		.lname	= "Number of files",
 		.alias	= "nr_files",
 		.type	= FIO_OPT_INT,
 		.off1	= td_var_offset(nr_files),
 		.help	= "Split job workload between this number of files",
 		.def	= "1",
+		.interval = 1,
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "openfiles",
+		.lname	= "Number of open files",
 		.type	= FIO_OPT_INT,
 		.off1	= td_var_offset(open_files),
 		.help	= "Number of files to keep open at the same time",
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "file_service_type",
+		.lname	= "File service type",
 		.type	= FIO_OPT_STR,
 		.cb	= str_fst_cb,
 		.off1	= td_var_offset(file_service_type),
 		.help	= "How to select which file to service next",
 		.def	= "roundrobin",
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_INVALID,
 		.posval	= {
 			  { .ival = "random",
 			    .oval = FIO_FSERVICE_RANDOM,
@@ -1223,14 +1797,18 @@ static struct fio_option options[FIO_MAX_OPTS] = {
 			  },
 		},
 		.parent = "nrfiles",
+		.hide	= 1,
 	},
-#ifdef FIO_HAVE_FALLOCATE
+#ifdef CONFIG_POSIX_FALLOCATE
 	{
 		.name	= "fallocate",
+		.lname	= "Fallocate",
 		.type	= FIO_OPT_STR,
 		.off1	= td_var_offset(fallocate_mode),
 		.help	= "Whether pre-allocation is performed when laying out files",
 		.def	= "posix",
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_INVALID,
 		.posval	= {
 			  { .ival = "none",
 			    .oval = FIO_FALLOCATE_NONE,
@@ -1240,7 +1818,7 @@ static struct fio_option options[FIO_MAX_OPTS] = {
 			    .oval = FIO_FALLOCATE_POSIX,
 			    .help = "Use posix_fallocate()",
 			  },
-#ifdef FIO_HAVE_LINUX_FALLOCATE
+#ifdef CONFIG_LINUX_FALLOCATE
 			  { .ival = "keep",
 			    .oval = FIO_FALLOCATE_KEEP_SIZE,
 			    .help = "Use fallocate(..., FALLOC_FL_KEEP_SIZE, ...)",
@@ -1257,38 +1835,54 @@ static struct fio_option options[FIO_MAX_OPTS] = {
 			  },
 		},
 	},
-#endif	/* FIO_HAVE_FALLOCATE */
+#endif	/* CONFIG_POSIX_FALLOCATE */
 	{
 		.name	= "fadvise_hint",
+		.lname	= "Fadvise hint",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(fadvise_hint),
 		.help	= "Use fadvise() to advise the kernel on IO pattern",
 		.def	= "1",
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "fsync",
+		.lname	= "Fsync",
 		.type	= FIO_OPT_INT,
 		.off1	= td_var_offset(fsync_blocks),
 		.help	= "Issue fsync for writes every given number of blocks",
 		.def	= "0",
+		.interval = 1,
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "fdatasync",
+		.lname	= "Fdatasync",
 		.type	= FIO_OPT_INT,
 		.off1	= td_var_offset(fdatasync_blocks),
 		.help	= "Issue fdatasync for writes every given number of blocks",
 		.def	= "0",
+		.interval = 1,
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "write_barrier",
+		.lname	= "Write barrier",
 		.type	= FIO_OPT_INT,
 		.off1	= td_var_offset(barrier_blocks),
 		.help	= "Make every Nth write a barrier write",
 		.def	= "0",
+		.interval = 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_INVALID,
 	},
-#ifdef FIO_HAVE_SYNC_FILE_RANGE
+#ifdef CONFIG_SYNC_FILE_RANGE
 	{
 		.name	= "sync_file_range",
+		.lname	= "Sync file range",
 		.posval	= {
 			  { .ival = "wait_before",
 			    .oval = SYNC_FILE_RANGE_WAIT_BEFORE,
@@ -1311,86 +1905,126 @@ static struct fio_option options[FIO_MAX_OPTS] = {
 		.cb	= str_sfr_cb,
 		.off1	= td_var_offset(sync_file_range),
 		.help	= "Use sync_file_range()",
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_INVALID,
 	},
 #endif
 	{
 		.name	= "direct",
+		.lname	= "Direct I/O",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(odirect),
 		.help	= "Use O_DIRECT IO (negates buffered)",
 		.def	= "0",
+		.inverse = "buffered",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IO_TYPE,
 	},
 	{
 		.name	= "buffered",
+		.lname	= "Buffered I/O",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(odirect),
 		.neg	= 1,
 		.help	= "Use buffered IO (negates direct)",
 		.def	= "1",
+		.inverse = "direct",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IO_TYPE,
 	},
 	{
 		.name	= "overwrite",
+		.lname	= "Overwrite",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(overwrite),
 		.help	= "When writing, set whether to overwrite current data",
 		.def	= "0",
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "loops",
+		.lname	= "Loops",
 		.type	= FIO_OPT_INT,
 		.off1	= td_var_offset(loops),
 		.help	= "Number of times to run the job",
 		.def	= "1",
+		.interval = 1,
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_RUNTIME,
 	},
 	{
 		.name	= "numjobs",
+		.lname	= "Number of jobs",
 		.type	= FIO_OPT_INT,
 		.off1	= td_var_offset(numjobs),
 		.help	= "Duplicate this job this many times",
 		.def	= "1",
+		.interval = 1,
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_RUNTIME,
 	},
 	{
 		.name	= "startdelay",
+		.lname	= "Start delay",
 		.type	= FIO_OPT_STR_VAL_TIME,
 		.off1	= td_var_offset(start_delay),
 		.help	= "Only start job when this period has passed",
 		.def	= "0",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_RUNTIME,
 	},
 	{
 		.name	= "runtime",
+		.lname	= "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",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_RUNTIME,
 	},
 	{
 		.name	= "time_based",
+		.lname	= "Time based",
 		.type	= FIO_OPT_STR_SET,
 		.off1	= td_var_offset(time_based),
 		.help	= "Keep running until runtime/timeout is met",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_RUNTIME,
 	},
 	{
 		.name	= "ramp_time",
+		.lname	= "Ramp time",
 		.type	= FIO_OPT_STR_VAL_TIME,
 		.off1	= td_var_offset(ramp_time),
 		.help	= "Ramp up time before measuring performance",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_RUNTIME,
 	},
 	{
 		.name	= "clocksource",
+		.lname	= "Clock source",
 		.type	= FIO_OPT_STR,
 		.cb	= fio_clock_source_cb,
 		.off1	= td_var_offset(clocksource),
 		.help	= "What type of timing source to use",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_CLOCK,
 		.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,
@@ -1402,11 +2036,14 @@ static struct fio_option options[FIO_MAX_OPTS] = {
 	{
 		.name	= "mem",
 		.alias	= "iomem",
+		.lname	= "I/O Memory",
 		.type	= FIO_OPT_STR,
 		.cb	= str_mem_cb,
 		.off1	= td_var_offset(mem_type),
 		.help	= "Backing type for IO buffers",
 		.def	= "malloc",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_INVALID,
 		.posval	= {
 			  { .ival = "malloc",
 			    .oval = MEM_MALLOC,
@@ -1437,20 +2074,27 @@ static struct fio_option options[FIO_MAX_OPTS] = {
 	{
 		.name	= "iomem_align",
 		.alias	= "mem_align",
+		.lname	= "I/O memory alignment",
 		.type	= FIO_OPT_INT,
 		.off1	= td_var_offset(mem_align),
 		.minval	= 0,
 		.help	= "IO memory buffer offset alignment",
 		.def	= "0",
 		.parent	= "iomem",
+		.hide	= 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "verify",
+		.lname	= "Verify",
 		.type	= FIO_OPT_STR,
 		.off1	= td_var_offset(verify),
 		.help	= "Verify data written",
 		.cb	= str_verify_cb,
 		.def	= "0",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_VERIFY,
 		.posval = {
 			  { .ival = "0",
 			    .oval = VERIFY_NONE,
@@ -1469,12 +2113,12 @@ 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,
@@ -1509,581 +2153,1141 @@ static struct fio_option options[FIO_MAX_OPTS] = {
 	},
 	{
 		.name	= "do_verify",
+		.lname	= "Perform verify step",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(do_verify),
 		.help	= "Run verification stage after write",
 		.def	= "1",
 		.parent = "verify",
+		.hide	= 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_VERIFY,
 	},
 	{
 		.name	= "verifysort",
+		.lname	= "Verify sort",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(verifysort),
 		.help	= "Sort written verify blocks for read back",
 		.def	= "1",
 		.parent = "verify",
+		.hide	= 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_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",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_VERIFY,
 	},
 	{
 		.name   = "verify_interval",
+		.lname	= "Verify interval",
 		.type   = FIO_OPT_INT,
 		.off1   = td_var_offset(verify_interval),
 		.minval	= 2 * sizeof(struct verify_header),
 		.help   = "Store verify buffer header every N bytes",
 		.parent	= "verify",
+		.hide	= 1,
+		.interval = 2 * sizeof(struct verify_header),
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_VERIFY,
 	},
 	{
 		.name	= "verify_offset",
+		.lname	= "Verify offset",
 		.type	= FIO_OPT_INT,
 		.help	= "Offset verify header location by N bytes",
-		.def	= "0",
-		.cb	= str_verify_offset_cb,
+		.off1	= td_var_offset(verify_offset),
+		.minval	= sizeof(struct verify_header),
 		.parent	= "verify",
+		.hide	= 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_VERIFY,
 	},
 	{
 		.name	= "verify_pattern",
+		.lname	= "Verify pattern",
 		.type	= FIO_OPT_STR,
 		.cb	= str_verify_pattern_cb,
 		.help	= "Fill pattern for IO buffers",
 		.parent	= "verify",
+		.hide	= 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_VERIFY,
 	},
 	{
 		.name	= "verify_fatal",
+		.lname	= "Verify fatal",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(verify_fatal),
 		.def	= "0",
 		.help	= "Exit on a single verify failure, don't continue",
 		.parent = "verify",
+		.hide	= 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_VERIFY,
 	},
 	{
 		.name	= "verify_dump",
+		.lname	= "Verify dump",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(verify_dump),
-		.def	= "1",
+		.def	= "0",
 		.help	= "Dump contents of good and bad blocks on failure",
 		.parent = "verify",
+		.hide	= 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_VERIFY,
 	},
 	{
 		.name	= "verify_async",
+		.lname	= "Verify asynchronously",
 		.type	= FIO_OPT_INT,
 		.off1	= td_var_offset(verify_async),
 		.def	= "0",
 		.help	= "Number of async verifier threads to use",
 		.parent	= "verify",
+		.hide	= 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_VERIFY,
 	},
 	{
 		.name	= "verify_backlog",
+		.lname	= "Verify backlog",
 		.type	= FIO_OPT_STR_VAL,
 		.off1	= td_var_offset(verify_backlog),
 		.help	= "Verify after this number of blocks are written",
 		.parent	= "verify",
+		.hide	= 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_VERIFY,
 	},
 	{
 		.name	= "verify_backlog_batch",
+		.lname	= "Verify backlog batch",
 		.type	= FIO_OPT_INT,
 		.off1	= td_var_offset(verify_batch),
 		.help	= "Verify this number of IO blocks",
 		.parent	= "verify",
+		.hide	= 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_VERIFY,
 	},
 #ifdef FIO_HAVE_CPU_AFFINITY
 	{
 		.name	= "verify_async_cpus",
+		.lname	= "Async verify CPUs",
 		.type	= FIO_OPT_STR,
 		.cb	= str_verify_cpus_allowed_cb,
 		.help	= "Set CPUs allowed for async verify threads",
 		.parent	= "verify_async",
+		.hide	= 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_VERIFY,
 	},
 #endif
+	{
+		.name	= "experimental_verify",
+		.off1	= td_var_offset(experimental_verify),
+		.type	= FIO_OPT_BOOL,
+		.help	= "Enable experimental verification",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_VERIFY,
+	},
 #ifdef FIO_HAVE_TRIM
 	{
 		.name	= "trim_percentage",
+		.lname	= "Trim percentage",
 		.type	= FIO_OPT_INT,
-		.cb	= str_verify_trim_cb,
+		.off1	= td_var_offset(trim_percentage),
+		.minval = 0,
 		.maxval = 100,
 		.help	= "Number of verify blocks to discard/trim",
 		.parent	= "verify",
 		.def	= "0",
+		.interval = 1,
+		.hide	= 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_TRIM,
 	},
 	{
 		.name	= "trim_verify_zero",
-		.type	= FIO_OPT_INT,
+		.lname	= "Verify trim zero",
+		.type	= FIO_OPT_BOOL,
 		.help	= "Verify that trim/discarded blocks are returned as zeroes",
 		.off1	= td_var_offset(trim_zero),
 		.parent	= "trim_percentage",
+		.hide	= 1,
 		.def	= "1",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_TRIM,
 	},
 	{
 		.name	= "trim_backlog",
+		.lname	= "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",
+		.hide	= 1,
+		.interval = 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_TRIM,
 	},
 	{
 		.name	= "trim_backlog_batch",
+		.lname	= "Trim backlog batch",
 		.type	= FIO_OPT_INT,
 		.off1	= td_var_offset(trim_batch),
 		.help	= "Trim this number of IO blocks",
 		.parent	= "trim_percentage",
+		.hide	= 1,
+		.interval = 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_TRIM,
 	},
 #endif
 	{
 		.name	= "write_iolog",
+		.lname	= "Write I/O log",
 		.type	= FIO_OPT_STR_STORE,
 		.off1	= td_var_offset(write_iolog_file),
 		.help	= "Store IO pattern to file",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IOLOG,
 	},
 	{
 		.name	= "read_iolog",
+		.lname	= "Read I/O log",
 		.type	= FIO_OPT_STR_STORE,
 		.off1	= td_var_offset(read_iolog_file),
 		.help	= "Playback IO pattern from file",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IOLOG,
 	},
 	{
 		.name	= "replay_no_stall",
-		.type	= FIO_OPT_INT,
+		.lname	= "Don't stall on replay",
+		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(no_stall),
 		.def	= "0",
 		.parent	= "read_iolog",
+		.hide	= 1,
 		.help	= "Playback IO pattern file as fast as possible without stalls",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IOLOG,
 	},
 	{
 		.name	= "replay_redirect",
+		.lname	= "Redirect device for replay",
 		.type	= FIO_OPT_STR_STORE,
 		.off1	= td_var_offset(replay_redirect),
 		.parent	= "read_iolog",
+		.hide	= 1,
 		.help	= "Replay all I/O onto this device, regardless of trace device",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IOLOG,
 	},
 	{
 		.name	= "exec_prerun",
+		.lname	= "Pre-execute runnable",
 		.type	= FIO_OPT_STR_STORE,
 		.off1	= td_var_offset(exec_prerun),
 		.help	= "Execute this file prior to running job",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "exec_postrun",
+		.lname	= "Post-execute runnable",
 		.type	= FIO_OPT_STR_STORE,
 		.off1	= td_var_offset(exec_postrun),
 		.help	= "Execute this file after running job",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_INVALID,
 	},
 #ifdef FIO_HAVE_IOSCHED_SWITCH
 	{
 		.name	= "ioscheduler",
+		.lname	= "I/O scheduler",
 		.type	= FIO_OPT_STR_STORE,
 		.off1	= td_var_offset(ioscheduler),
 		.help	= "Use this IO scheduler on the backing device",
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_INVALID,
 	},
 #endif
 	{
 		.name	= "zonesize",
+		.lname	= "Zone size",
 		.type	= FIO_OPT_STR_VAL,
 		.off1	= td_var_offset(zone_size),
+		.help	= "Amount of data to read per zone",
+		.def	= "0",
+		.interval = 1024 * 1024,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_ZONE,
+	},
+	{
+		.name	= "zonerange",
+		.lname	= "Zone range",
+		.type	= FIO_OPT_STR_VAL,
+		.off1	= td_var_offset(zone_range),
 		.help	= "Give size of an IO zone",
 		.def	= "0",
+		.interval = 1024 * 1024,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_ZONE,
 	},
 	{
 		.name	= "zoneskip",
+		.lname	= "Zone skip",
 		.type	= FIO_OPT_STR_VAL,
 		.off1	= td_var_offset(zone_skip),
 		.help	= "Space between IO zones",
 		.def	= "0",
+		.interval = 1024 * 1024,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_ZONE,
 	},
 	{
 		.name	= "lockmem",
+		.lname	= "Lock memory",
 		.type	= FIO_OPT_STR_VAL,
-		.cb	= str_lockmem_cb,
-		.help	= "Lock down this amount of memory",
+		.off1	= td_var_offset(lockmem),
+		.help	= "Lock down this amount of memory (per worker)",
 		.def	= "0",
+		.interval = 1024 * 1024,
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "rwmixread",
+		.lname	= "Read/write mix read",
 		.type	= FIO_OPT_INT,
 		.cb	= str_rwmix_read_cb,
 		.maxval	= 100,
 		.help	= "Percentage of mixed workload that is reads",
 		.def	= "50",
+		.interval = 5,
+		.inverse = "rwmixwrite",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_RWMIX,
 	},
 	{
 		.name	= "rwmixwrite",
+		.lname	= "Read/write mix write",
 		.type	= FIO_OPT_INT,
 		.cb	= str_rwmix_write_cb,
 		.maxval	= 100,
 		.help	= "Percentage of mixed workload that is writes",
 		.def	= "50",
+		.interval = 5,
+		.inverse = "rwmixread",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_RWMIX,
 	},
 	{
 		.name	= "rwmixcycle",
+		.lname	= "Read/write mix cycle",
 		.type	= FIO_OPT_DEPRECATED,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_RWMIX,
 	},
 	{
 		.name	= "nice",
+		.lname	= "Nice",
 		.type	= FIO_OPT_INT,
 		.off1	= td_var_offset(nice),
 		.help	= "Set job CPU nice value",
 		.minval	= -19,
 		.maxval	= 20,
 		.def	= "0",
+		.interval = 1,
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_CRED,
 	},
 #ifdef FIO_HAVE_IOPRIO
 	{
 		.name	= "prio",
+		.lname	= "I/O nice priority",
 		.type	= FIO_OPT_INT,
 		.cb	= str_prio_cb,
 		.help	= "Set job IO priority value",
 		.minval	= 0,
 		.maxval	= 7,
+		.interval = 1,
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_CRED,
 	},
 	{
 		.name	= "prioclass",
+		.lname	= "I/O nice priority class",
 		.type	= FIO_OPT_INT,
 		.cb	= str_prioclass_cb,
 		.help	= "Set job IO priority class",
 		.minval	= 0,
 		.maxval	= 3,
+		.interval = 1,
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_CRED,
 	},
 #endif
 	{
 		.name	= "thinktime",
+		.lname	= "Thinktime",
 		.type	= FIO_OPT_INT,
 		.off1	= td_var_offset(thinktime),
 		.help	= "Idle time between IO buffers (usec)",
 		.def	= "0",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_THINKTIME,
 	},
 	{
 		.name	= "thinktime_spin",
+		.lname	= "Thinktime spin",
 		.type	= FIO_OPT_INT,
 		.off1	= td_var_offset(thinktime_spin),
 		.help	= "Start think time by spinning this amount (usec)",
 		.def	= "0",
 		.parent	= "thinktime",
+		.hide	= 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_THINKTIME,
 	},
 	{
 		.name	= "thinktime_blocks",
+		.lname	= "Thinktime blocks",
 		.type	= FIO_OPT_INT,
 		.off1	= td_var_offset(thinktime_blocks),
 		.help	= "IO buffer period between 'thinktime'",
 		.def	= "1",
 		.parent	= "thinktime",
+		.hide	= 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_THINKTIME,
 	},
 	{
 		.name	= "rate",
+		.lname	= "I/O 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",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_RATE,
 	},
 	{
 		.name	= "ratemin",
+		.lname	= "I/O min rate",
 		.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",
+		.hide	= 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_RATE,
 	},
 	{
 		.name	= "rate_iops",
+		.lname	= "I/O 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",
+		.hide	= 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_RATE,
 	},
 	{
 		.name	= "rate_iops_min",
+		.lname	= "I/O min rate IOPS",
 		.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",
+		.hide	= 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_RATE,
 	},
 	{
 		.name	= "ratecycle",
+		.lname	= "I/O rate cycle",
 		.type	= FIO_OPT_INT,
 		.off1	= td_var_offset(ratecycle),
 		.help	= "Window average for rate limits (msec)",
 		.def	= "1000",
 		.parent = "rate",
+		.hide	= 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_RATE,
+	},
+	{
+		.name	= "max_latency",
+		.type	= FIO_OPT_INT,
+		.off1	= td_var_offset(max_latency),
+		.help	= "Maximum tolerated IO latency (usec)",
+		.category = FIO_OPT_C_IO,
+		.group = FIO_OPT_G_RATE,
 	},
 	{
 		.name	= "invalidate",
+		.lname	= "Cache invalidate",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(invalidate_cache),
 		.help	= "Invalidate buffer/page cache prior to running job",
 		.def	= "1",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IO_TYPE,
 	},
 	{
 		.name	= "sync",
+		.lname	= "Synchronous I/O",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(sync_io),
 		.help	= "Use O_SYNC for buffered writes",
 		.def	= "0",
 		.parent = "buffered",
-	},
-	{
-		.name	= "bwavgtime",
-		.type	= FIO_OPT_INT,
-		.off1	= td_var_offset(bw_avg_time),
-		.help	= "Time window over which to calculate bandwidth"
-			  " (msec)",
-		.def	= "500",
+		.hide	= 1,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IO_TYPE,
 	},
 	{
 		.name	= "create_serialize",
+		.lname	= "Create serialize",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(create_serialize),
 		.help	= "Serialize creating of job files",
 		.def	= "1",
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "create_fsync",
+		.lname	= "Create fsync",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(create_fsync),
 		.help	= "fsync file after creation",
 		.def	= "1",
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "create_on_open",
+		.lname	= "Create on open",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(create_on_open),
 		.help	= "Create files when they are opened for IO",
 		.def	= "0",
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_INVALID,
+	},
+	{
+		.name	= "create_only",
+		.type	= FIO_OPT_BOOL,
+		.off1	= td_var_offset(create_only),
+		.help	= "Only perform file creation phase",
+		.category = FIO_OPT_C_FILE,
+		.def	= "0",
 	},
 	{
 		.name	= "pre_read",
+		.lname	= "Pre-read files",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(pre_read),
 		.help	= "Pre-read files before starting official testing",
 		.def	= "0",
-	},
-	{
-		.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 (usecs)",
-		.def	= "50000",
-		.parent = "cpuload",
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_INVALID,
 	},
 #ifdef FIO_HAVE_CPU_AFFINITY
 	{
 		.name	= "cpumask",
+		.lname	= "CPU mask",
 		.type	= FIO_OPT_INT,
 		.cb	= str_cpumask_cb,
 		.help	= "CPU affinity mask",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_CRED,
 	},
 	{
 		.name	= "cpus_allowed",
+		.lname	= "CPUs allowed",
 		.type	= FIO_OPT_STR,
 		.cb	= str_cpus_allowed_cb,
 		.help	= "Set CPUs allowed",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_CRED,
+	},
+#endif
+#ifdef CONFIG_LIBNUMA
+	{
+		.name	= "numa_cpu_nodes",
+		.type	= FIO_OPT_STR,
+		.cb	= str_numa_cpunodes_cb,
+		.help	= "NUMA CPU nodes bind",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_INVALID,
+	},
+	{
+		.name	= "numa_mem_policy",
+		.type	= FIO_OPT_STR,
+		.cb	= str_numa_mpol_cb,
+		.help	= "NUMA memory policy setup",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_INVALID,
 	},
 #endif
 	{
 		.name	= "end_fsync",
+		.lname	= "End fsync",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(end_fsync),
 		.help	= "Include fsync at the end of job",
 		.def	= "0",
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "fsync_on_close",
+		.lname	= "Fsync on close",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(fsync_on_close),
 		.help	= "fsync files on close",
 		.def	= "0",
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "unlink",
+		.lname	= "Unlink file",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(unlink),
 		.help	= "Unlink created files after job has completed",
 		.def	= "0",
+		.category = FIO_OPT_C_FILE,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "exitall",
+		.lname	= "Exit-all on terminate",
 		.type	= FIO_OPT_STR_SET,
 		.cb	= str_exitall_cb,
 		.help	= "Terminate all jobs when one exits",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_PROCESS,
 	},
 	{
 		.name	= "stonewall",
+		.lname	= "Wait for previous",
 		.alias	= "wait_for_previous",
 		.type	= FIO_OPT_STR_SET,
 		.off1	= td_var_offset(stonewall),
 		.help	= "Insert a hard barrier between this job and previous",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_PROCESS,
 	},
 	{
 		.name	= "new_group",
+		.lname	= "New group",
 		.type	= FIO_OPT_STR_SET,
 		.off1	= td_var_offset(new_group),
 		.help	= "Mark the start of a new group (for reporting)",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_PROCESS,
 	},
 	{
 		.name	= "thread",
+		.lname	= "Thread",
 		.type	= FIO_OPT_STR_SET,
 		.off1	= td_var_offset(use_thread),
-		.help	= "Use threads instead of forks",
+		.help	= "Use threads instead of processes",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_PROCESS,
 	},
 	{
 		.name	= "write_bw_log",
+		.lname	= "Write bandwidth log",
 		.type	= FIO_OPT_STR,
-		.off1	= td_var_offset(write_bw_log),
+		.off1	= td_var_offset(bw_log_file),
 		.cb	= str_write_bw_log_cb,
 		.help	= "Write log of bandwidth during run",
+		.category = FIO_OPT_C_LOG,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "write_lat_log",
+		.lname	= "Write latency log",
 		.type	= FIO_OPT_STR,
-		.off1	= td_var_offset(write_lat_log),
+		.off1	= td_var_offset(lat_log_file),
 		.cb	= str_write_lat_log_cb,
 		.help	= "Write log of latency during run",
+		.category = FIO_OPT_C_LOG,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
-		.name	= "hugepage-size",
+		.name	= "write_iops_log",
+		.lname	= "Write IOPS log",
+		.type	= FIO_OPT_STR,
+		.off1	= td_var_offset(iops_log_file),
+		.cb	= str_write_iops_log_cb,
+		.help	= "Write log of IOPS during run",
+		.category = FIO_OPT_C_LOG,
+		.group	= FIO_OPT_G_INVALID,
+	},
+	{
+		.name	= "log_avg_msec",
+		.lname	= "Log averaging (msec)",
 		.type	= FIO_OPT_INT,
-		.off1	= td_var_offset(hugepage_size),
-		.help	= "When using hugepages, specify size of each page",
-		.def	= __stringify(FIO_HUGE_PAGE),
+		.off1	= td_var_offset(log_avg_msec),
+		.help	= "Average bw/iops/lat logs over this period of time",
+		.def	= "0",
+		.category = FIO_OPT_C_LOG,
+		.group	= FIO_OPT_G_INVALID,
+	},
+	{
+		.name	= "bwavgtime",
+		.lname	= "Bandwidth average time",
+		.type	= FIO_OPT_INT,
+		.off1	= td_var_offset(bw_avg_time),
+		.help	= "Time window over which to calculate bandwidth"
+			  " (msec)",
+		.def	= "500",
+		.parent	= "write_bw_log",
+		.hide	= 1,
+		.interval = 100,
+		.category = FIO_OPT_C_LOG,
+		.group	= FIO_OPT_G_INVALID,
+	},
+	{
+		.name	= "iopsavgtime",
+		.lname	= "IOPS average time",
+		.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",
+		.hide	= 1,
+		.interval = 100,
+		.category = FIO_OPT_C_LOG,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "group_reporting",
-		.type	= FIO_OPT_STR_SET,
+		.lname	= "Group reporting",
+		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(group_reporting),
 		.help	= "Do reporting on a per-group basis",
+		.def	= "1",
+		.category = FIO_OPT_C_STAT,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "zero_buffers",
+		.lname	= "Zero I/O buffers",
 		.type	= FIO_OPT_STR_SET,
 		.off1	= td_var_offset(zero_buffers),
 		.help	= "Init IO buffers to all zeroes",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IO_BUF,
 	},
 	{
 		.name	= "refill_buffers",
+		.lname	= "Refill I/O buffers",
 		.type	= FIO_OPT_STR_SET,
 		.off1	= td_var_offset(refill_buffers),
 		.help	= "Refill IO buffers on every IO submit",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IO_BUF,
+	},
+	{
+		.name	= "scramble_buffers",
+		.lname	= "Scramble I/O buffers",
+		.type	= FIO_OPT_BOOL,
+		.off1	= td_var_offset(scramble_buffers),
+		.help	= "Slightly scramble buffers on every IO submit",
+		.def	= "1",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IO_BUF,
+	},
+	{
+		.name	= "buffer_compress_percentage",
+		.lname	= "Buffer compression percentage",
+		.type	= FIO_OPT_INT,
+		.off1	= td_var_offset(compress_percentage),
+		.maxval	= 100,
+		.minval	= 1,
+		.help	= "How compressible the buffer is (approximately)",
+		.interval = 5,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IO_BUF,
+	},
+	{
+		.name	= "buffer_compress_chunk",
+		.lname	= "Buffer compression chunk size",
+		.type	= FIO_OPT_INT,
+		.off1	= td_var_offset(compress_chunk),
+		.parent	= "buffer_compress_percentage",
+		.hide	= 1,
+		.help	= "Size of compressible region in buffer",
+		.interval = 256,
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IO_BUF,
 	},
 	{
 		.name	= "clat_percentiles",
+		.lname	= "Completion latency percentiles",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(clat_percentiles),
 		.help	= "Enable the reporting of completion latency percentiles",
-		.def	= "0",
+		.def	= "1",
+		.category = FIO_OPT_C_STAT,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "percentile_list",
+		.lname	= "Completion latency percentile list",
 		.type	= FIO_OPT_FLOAT_LIST,
 		.off1	= td_var_offset(percentile_list),
-		.off2   = td_var_offset(overwrite_plist),
+		.off2	= td_var_offset(percentile_precision),
 		.help	= "Specify a custom list of percentiles to report",
+		.def    = "1:5:10:20:30:40:50:60:70:80:90:95:99:99.5:99.9:99.95:99.99",
 		.maxlen	= FIO_IO_U_LIST_MAX_LEN,
 		.minfp	= 0.0,
 		.maxfp	= 100.0,
+		.category = FIO_OPT_C_STAT,
+		.group	= FIO_OPT_G_INVALID,
 	},
 
 #ifdef FIO_HAVE_DISK_UTIL
 	{
 		.name	= "disk_util",
+		.lname	= "Disk utilization",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(do_disk_util),
 		.help	= "Log disk utilization statistics",
 		.def	= "1",
+		.category = FIO_OPT_C_STAT,
+		.group	= FIO_OPT_G_INVALID,
 	},
 #endif
 	{
 		.name	= "gtod_reduce",
+		.lname	= "Reduce gettimeofday() calls",
 		.type	= FIO_OPT_BOOL,
 		.help	= "Greatly reduce number of gettimeofday() calls",
 		.cb	= str_gtod_reduce_cb,
 		.def	= "0",
+		.hide_on_set = 1,
+		.category = FIO_OPT_C_STAT,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "disable_lat",
+		.lname	= "Disable all latency stats",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(disable_lat),
 		.help	= "Disable latency numbers",
 		.parent	= "gtod_reduce",
+		.hide	= 1,
 		.def	= "0",
+		.category = FIO_OPT_C_STAT,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "disable_clat",
+		.lname	= "Disable completion latency stats",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(disable_clat),
 		.help	= "Disable completion latency numbers",
 		.parent	= "gtod_reduce",
+		.hide	= 1,
 		.def	= "0",
+		.category = FIO_OPT_C_STAT,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "disable_slat",
+		.lname	= "Disable submission latency stats",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(disable_slat),
 		.help	= "Disable submission latency numbers",
 		.parent	= "gtod_reduce",
+		.hide	= 1,
 		.def	= "0",
+		.category = FIO_OPT_C_STAT,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "disable_bw_measurement",
+		.lname	= "Disable bandwidth stats",
 		.type	= FIO_OPT_BOOL,
 		.off1	= td_var_offset(disable_bw),
 		.help	= "Disable bandwidth logging",
 		.parent	= "gtod_reduce",
+		.hide	= 1,
 		.def	= "0",
+		.category = FIO_OPT_C_STAT,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "gtod_cpu",
+		.lname	= "Dedicated gettimeofday() CPU",
 		.type	= FIO_OPT_INT,
 		.cb	= str_gtod_cpu_cb,
 		.help	= "Set up dedicated gettimeofday() thread on this CPU",
 		.verify	= gtod_cpu_verify,
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_CLOCK,
 	},
 	{
-		.name	= "continue_on_error",
+		.name	= "unified_rw_reporting",
 		.type	= FIO_OPT_BOOL,
+		.off1	= td_var_offset(unified_rw_rep),
+		.help	= "Unify reporting across data direction",
+		.def	= "0",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_INVALID,
+	},
+	{
+		.name	= "continue_on_error",
+		.lname	= "Continue on error",
+		.type	= FIO_OPT_STR,
 		.off1	= td_var_offset(continue_on_error),
 		.help	= "Continue on non-fatal errors during IO",
+		.def	= "none",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_ERR,
+		.posval = {
+			  { .ival = "none",
+			    .oval = ERROR_TYPE_NONE,
+			    .help = "Exit when an error is encountered",
+			  },
+			  { .ival = "read",
+			    .oval = ERROR_TYPE_READ,
+			    .help = "Continue on read errors only",
+			  },
+			  { .ival = "write",
+			    .oval = ERROR_TYPE_WRITE,
+			    .help = "Continue on write errors only",
+			  },
+			  { .ival = "io",
+			    .oval = ERROR_TYPE_READ | ERROR_TYPE_WRITE,
+			    .help = "Continue on any IO errors",
+			  },
+			  { .ival = "verify",
+			    .oval = ERROR_TYPE_VERIFY,
+			    .help = "Continue on verify errors only",
+			  },
+			  { .ival = "all",
+			    .oval = ERROR_TYPE_ANY,
+			    .help = "Continue on all io and verify errors",
+			  },
+			  { .ival = "0",
+			    .oval = ERROR_TYPE_NONE,
+			    .help = "Alias for 'none'",
+			  },
+			  { .ival = "1",
+			    .oval = ERROR_TYPE_ANY,
+			    .help = "Alias for 'all'",
+			  },
+		},
+	},
+	{
+		.name	= "ignore_error",
+		.type	= FIO_OPT_STR,
+		.cb	= str_ignore_error_cb,
+		.help	= "Set a specific list of errors to ignore",
+		.parent	= "rw",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_ERR,
+	},
+	{
+		.name	= "error_dump",
+		.type	= FIO_OPT_BOOL,
+		.off1	= td_var_offset(error_dump),
 		.def	= "0",
+		.help	= "Dump info on each error",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_ERR,
 	},
 	{
 		.name	= "profile",
+		.lname	= "Profile",
 		.type	= FIO_OPT_STR_STORE,
 		.off1	= td_var_offset(profile),
 		.help	= "Select a specific builtin performance test",
+		.category = FIO_OPT_C_PROFILE,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "cgroup",
+		.lname	= "Cgroup",
 		.type	= FIO_OPT_STR_STORE,
 		.off1	= td_var_offset(cgroup),
 		.help	= "Add job to cgroup of this name",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_CGROUP,
+	},
+	{
+		.name	= "cgroup_nodelete",
+		.lname	= "Cgroup no-delete",
+		.type	= FIO_OPT_BOOL,
+		.off1	= td_var_offset(cgroup_nodelete),
+		.help	= "Do not delete cgroups after job completion",
+		.def	= "0",
+		.parent	= "cgroup",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_CGROUP,
 	},
 	{
 		.name	= "cgroup_weight",
+		.lname	= "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",
+		.parent	= "cgroup",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_CGROUP,
 	},
 	{
 		.name	= "uid",
+		.lname	= "User ID",
 		.type	= FIO_OPT_INT,
 		.off1	= td_var_offset(uid),
 		.help	= "Run job with this user ID",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_CRED,
 	},
 	{
 		.name	= "gid",
+		.lname	= "Group ID",
 		.type	= FIO_OPT_INT,
 		.off1	= td_var_offset(gid),
 		.help	= "Run job with this group ID",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_CRED,
+	},
+	{
+		.name	= "kb_base",
+		.lname	= "KB Base",
+		.type	= FIO_OPT_INT,
+		.off1	= td_var_offset(kb_base),
+		.prio	= 1,
+		.def	= "1024",
+		.posval = {
+			  { .ival = "1024",
+			    .oval = 1024,
+			    .help = "Use 1024 as the K base",
+			  },
+			  { .ival = "1000",
+			    .oval = 1000,
+			    .help = "Use 1000 as the K base",
+			  },
+		},
+		.help	= "How many bytes per KB for reporting (1000 or 1024)",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_INVALID,
+	},
+	{
+		.name	= "unit_base",
+		.lname	= "Base unit for reporting (Bits or Bytes)",
+		.type	= FIO_OPT_INT,
+		.off1	= td_var_offset(unit_base),
+		.prio	= 1,
+		.posval = {
+			  { .ival = "0",
+			    .oval = 0,
+			    .help = "Auto-detect",
+			  },
+			  { .ival = "8",
+			    .oval = 8,
+			    .help = "Normal (byte based)",
+			  },
+			  { .ival = "1",
+			    .oval = 1,
+			    .help = "Bit based",
+			  },
+		},
+		.help	= "Bit multiple of result summary data (8 for byte, 1 for bit)",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_INVALID,
+	},
+	{
+		.name	= "hugepage-size",
+		.lname	= "Hugepage size",
+		.type	= FIO_OPT_INT,
+		.off1	= td_var_offset(hugepage_size),
+		.help	= "When using hugepages, specify size of each page",
+		.def	= __fio_stringify(FIO_HUGE_PAGE),
+		.interval = 1024 * 1024,
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_INVALID,
+	},
+	{
+		.name	= "flow_id",
+		.lname	= "I/O flow ID",
+		.type	= FIO_OPT_INT,
+		.off1	= td_var_offset(flow_id),
+		.help	= "The flow index ID to use",
+		.def	= "0",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IO_FLOW,
+	},
+	{
+		.name	= "flow",
+		.lname	= "I/O flow weight",
+		.type	= FIO_OPT_INT,
+		.off1	= td_var_offset(flow),
+		.help	= "Weight for flow control of this job",
+		.parent	= "flow_id",
+		.hide	= 1,
+		.def	= "0",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IO_FLOW,
+	},
+	{
+		.name	= "flow_watermark",
+		.lname	= "I/O 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",
+		.hide	= 1,
+		.def	= "1024",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IO_FLOW,
+	},
+	{
+		.name	= "flow_sleep",
+		.lname	= "I/O 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",
+		.hide	= 1,
+		.def	= "0",
+		.category = FIO_OPT_C_IO,
+		.group	= FIO_OPT_G_IO_FLOW,
 	},
 	{
 		.name = NULL,
@@ -2091,33 +3295,26 @@ static struct fio_option options[FIO_MAX_OPTS] = {
 };
 
 static void add_to_lopt(struct option *lopt, struct fio_option *o,
-			const char *name)
+			const char *name, int val)
 {
 	lopt->name = (char *) name;
-	lopt->val = FIO_GETOPT_JOB;
+	lopt->val = val;
 	if (o->type == FIO_OPT_STR_SET)
 		lopt->has_arg = no_argument;
 	else
 		lopt->has_arg = required_argument;
 }
 
-void fio_options_dup_and_init(struct option *long_options)
+static void options_to_lopts(struct fio_option *opts,
+			      struct option *long_options,
+			      int i, int option_type)
 {
-	struct fio_option *o;
-	unsigned int i;
-
-	options_init(options);
-
-	i = 0;
-	while (long_options[i].name)
-		i++;
-
-	o = &options[0];
+	struct fio_option *o = &opts[0];
 	while (o->name) {
-		add_to_lopt(&long_options[i], o, 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);
+			add_to_lopt(&long_options[i], o, o->alias, option_type);
 		}
 
 		i++;
@@ -2126,6 +3323,43 @@ void fio_options_dup_and_init(struct option *long_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)
+{
+	unsigned int i;
+
+	options_init(fio_options);
+
+	i = 0;
+	while (long_options[i].name)
+		i++;
+
+	options_to_lopts(fio_options, long_options, i, FIO_GETOPT_JOB);
+}
+
 struct fio_keyword {
 	const char *word;
 	const char *desc;
@@ -2156,7 +3390,7 @@ 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);
@@ -2172,15 +3406,15 @@ void fio_keywords_init(void)
 
 static char *bc_calc(char *str)
 {
-	char *buf, *tmp, opt[80];
+	char buf[128], *tmp;
 	FILE *f;
 	int ret;
 
 	/*
 	 * No math, just return string
 	 */
-	if (!strchr(str, '+') && !strchr(str, '-') && !strchr(str, '*') &&
-	    !strchr(str, '/'))
+	if ((!strchr(str, '+') && !strchr(str, '-') && !strchr(str, '*') &&
+	     !strchr(str, '/')) || strchr(str, '\''))
 		return str;
 
 	/*
@@ -2191,37 +3425,89 @@ static char *bc_calc(char *str)
 		return str;
 
 	tmp++;
-	memset(opt, 0, sizeof(opt));
-	strncpy(opt, str, tmp - str);
 
-	buf = malloc(128);
+	/*
+	 * 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");
-		free(buf);
 		return NULL;
 	}
 
-	sprintf(buf, "echo %s | %s", tmp, BC_APP);
+	sprintf(buf, "echo '%s' | %s", tmp, BC_APP);
 	f = popen(buf, "r");
 	if (!f) {
-		free(buf);
 		return NULL;
 	}
 
-	ret = fread(buf, 1, 128, f);
+	ret = fread(&buf[tmp - str], 1, 128 - (tmp - str), f);
 	if (ret <= 0) {
-		free(buf);
 		return NULL;
 	}
 
-	buf[ret - 1] = '\0';
-	strcat(opt, buf);
-	strcpy(buf, opt);
 	pclose(f);
+	buf[(tmp - str) + ret - 1] = '\0';
+	memcpy(buf, str, tmp - str);
 	free(str);
-	return buf;
+	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);
 }
 
 /*
@@ -2231,6 +3517,7 @@ 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];
@@ -2260,83 +3547,139 @@ static char *fio_keyword_replace(char *opt)
 			 * replace opt and free the old opt
 			 */
 			opt = new;
-			//free(o_org);
+			free(o_org);
 
-			/*
-			 * Check for potential math and invoke bc, if possible
-			 */
-			opt = bc_calc(opt);
+			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, fio_options, num_opts);
+	opts_copy = dup_and_sub_options(opts, num_opts);
 
-	sort_options(opts, options, 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], fio_options,
+						&o, td);
 
-	for (ret = 0, i = 0; i < num_opts; i++) {
-		opts[i] = fio_keyword_replace(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;
 }
 
 int fio_cmd_option_parse(struct thread_data *td, const char *opt, char *val)
 {
-	return parse_cmd_option(opt, val, options, td);
+	return parse_cmd_option(opt, val, fio_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);
+	fill_default_options(td, fio_options);
 }
 
 int fio_show_option_help(const char *opt)
 {
-	return show_cmd_help(options, opt);
+	return show_cmd_help(fio_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, fio_options);
 
-void options_mem_free(struct thread_data fio_unused *td)
-{
-#if 0
-	__options_mem(td, 0);
-#endif
+	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);
+	}
 }
 
 unsigned int fio_get_kb_base(void *data)
@@ -2357,13 +3700,13 @@ int add_option(struct fio_option *o)
 	struct fio_option *__o;
 	int opt_index = 0;
 
-	__o = options;
+	__o = fio_options;
 	while (__o->name) {
 		opt_index++;
 		__o++;
 	}
 
-	memcpy(&options[opt_index], o, sizeof(*o));
+	memcpy(&fio_options[opt_index], o, sizeof(*o));
 	return 0;
 }
 
@@ -2371,7 +3714,7 @@ void invalidate_profile_options(const char *prof_name)
 {
 	struct fio_option *o;
 
-	o = options;
+	o = fio_options;
 	while (o->name) {
 		if (o->prof_name && !strcmp(o->prof_name, prof_name)) {
 			o->type = FIO_OPT_INVALID;
@@ -2386,7 +3729,7 @@ 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);
+	o = find_option(fio_options, optname);
 	if (!o)
 		return;
 
@@ -2405,7 +3748,7 @@ void del_opt_posval(const char *optname, const char *ival)
 	struct fio_option *o;
 	unsigned int i;
 
-	o = find_option(options, optname);
+	o = find_option(fio_options, optname);
 	if (!o)
 		return;
 
@@ -2419,3 +3762,13 @@ void del_opt_posval(const char *optname, const char *ival)
 		o->posval[i].help = NULL;
 	}
 }
+
+void fio_options_free(struct thread_data *td)
+{
+	options_free(fio_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;
+	}
+}