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);
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);
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;
{
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;
}
#endif
+#ifdef FIO_HAVE_LIBNUMA
+static int str_numa_cpunodes_cb(void *data, char *input)
+{
+ struct thread_data *td = data;
+
+ /* numa_parse_nodestring() parses a character string list
+ * of nodes into a bit mask. The bit mask is allocated by
+ * numa_allocate_nodemask(), so it should be freed by
+ * numa_free_nodemask().
+ */
+ td->o.numa_cpunodesmask = numa_parse_nodestring(input);
+ if (td->o.numa_cpunodesmask == NULL) {
+ log_err("fio: numa_parse_nodestring failed\n");
+ td_verror(td, 1, "str_numa_cpunodes_cb");
+ return 1;
+ }
+
+ td->o.numa_cpumask_set = 1;
+ return 0;
+}
+
+static int str_numa_mpol_cb(void *data, char *input)
+{
+ struct thread_data *td = data;
+ const char * const policy_types[] =
+ { "default", "prefer", "bind", "interleave", "local" };
+ int i;
+
+ char *nodelist = strchr(input, ':');
+ if (nodelist) {
+ /* NUL-terminate mode */
+ *nodelist++ = '\0';
+ }
+
+ for (i = 0; i <= MPOL_LOCAL; i++) {
+ if (!strcmp(input, policy_types[i])) {
+ td->o.numa_mem_mode = i;
+ break;
+ }
+ }
+ if (i > MPOL_LOCAL) {
+ log_err("fio: memory policy should be: default, prefer, bind, interleave, local\n");
+ goto out;
+ }
+
+ switch (td->o.numa_mem_mode) {
+ case MPOL_PREFERRED:
+ /*
+ * Insist on a nodelist of one node only
+ */
+ if (nodelist) {
+ char *rest = nodelist;
+ while (isdigit(*rest))
+ rest++;
+ if (*rest) {
+ log_err("fio: one node only for \'prefer\'\n");
+ goto out;
+ }
+ } else {
+ log_err("fio: one node is needed for \'prefer\'\n");
+ goto out;
+ }
+ break;
+ case MPOL_INTERLEAVE:
+ /*
+ * Default to online nodes with memory if no nodelist
+ */
+ if (!nodelist)
+ nodelist = strdup("all");
+ break;
+ case MPOL_LOCAL:
+ case MPOL_DEFAULT:
+ /*
+ * Don't allow a nodelist
+ */
+ if (nodelist) {
+ log_err("fio: NO nodelist for \'local\'\n");
+ goto out;
+ }
+ break;
+ case MPOL_BIND:
+ /*
+ * Insist on a nodelist
+ */
+ if (!nodelist) {
+ log_err("fio: a nodelist is needed for \'bind\'\n");
+ goto out;
+ }
+ break;
+ }
+
+
+ /* numa_parse_nodestring() parses a character string list
+ * of nodes into a bit mask. The bit mask is allocated by
+ * numa_allocate_nodemask(), so it should be freed by
+ * numa_free_nodemask().
+ */
+ switch (td->o.numa_mem_mode) {
+ case MPOL_PREFERRED:
+ td->o.numa_mem_prefer_node = atoi(nodelist);
+ break;
+ case MPOL_INTERLEAVE:
+ case MPOL_BIND:
+ td->o.numa_memnodesmask = numa_parse_nodestring(nodelist);
+ if (td->o.numa_memnodesmask == NULL) {
+ log_err("fio: numa_parse_nodestring failed\n");
+ td_verror(td, 1, "str_numa_memnodes_cb");
+ return 1;
+ }
+ break;
+ case MPOL_LOCAL:
+ case MPOL_DEFAULT:
+ default:
+ break;
+ }
+
+ td->o.numa_memmask_set = 1;
+ return 0;
+
+out:
+ return 1;
+}
+#endif
+
#ifdef FIO_HAVE_TRIM
static int str_verify_trim_cb(void *data, unsigned long long *val)
{
}
#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
}
}
}
+
+ /*
+ * 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;
}
.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",
.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"
{ .ival = "rdma",
.help = "RDMA IO engine",
},
+#endif
+#ifdef FIO_HAVE_FUSION_AW
+ { .ival = "fusion-aw-sync",
+ .help = "Fusion-io atomic write engine",
+ },
+#endif
+#ifdef FIO_HAVE_E4_ENG
+ { .ival = "e4defrag",
+ .help = "ext4 defrag engine",
+ },
+#endif
+#ifdef FIO_HAVE_FALLOC_ENG
+ { .ival = "falloc",
+ .help = "fallocate() file based engine",
+ },
#endif
{ .ival = "external",
.help = "Load external engine (append name)",
.help = "Start IO from this offset",
.def = "0",
},
+ {
+ .name = "offset_increment",
+ .type = FIO_OPT_STR_VAL,
+ .off1 = td_var_offset(offset_increment),
+ .help = "What is the increment from one offset to the next",
+ .parent = "offset",
+ .def = "0",
+ },
{
.name = "bs",
.alias = "blocksize",
.type = FIO_OPT_INT,
.off1 = td_var_offset(bs[DDIR_READ]),
.off2 = td_var_offset(bs[DDIR_WRITE]),
+ .off3 = td_var_offset(bs[DDIR_TRIM]),
.minval = 1,
.help = "Block size unit",
.def = "4k",
.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",
.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",
.parent = "norandommap",
.def = "0",
},
+ {
+ .name = "random_distribution",
+ .type = FIO_OPT_STR,
+ .off1 = td_var_offset(random_distribution),
+ .cb = str_random_distribution_cb,
+ .help = "Random offset distribution generator",
+ .def = "random",
+ .posval = {
+ { .ival = "random",
+ .oval = FIO_RAND_DIST_RANDOM,
+ .help = "Completely random",
+ },
+ { .ival = "zipf",
+ .oval = FIO_RAND_DIST_ZIPF,
+ .help = "Zipf distribution",
+ },
+ { .ival = "pareto",
+ .oval = FIO_RAND_DIST_PARETO,
+ .help = "Pareto distribution",
+ },
+ },
+ },
{
.name = "nrfiles",
.alias = "nr_files",
.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,
.name = "zonesize",
.type = FIO_OPT_STR_VAL,
.off1 = td_var_offset(zone_size),
+ .help = "Amount of data to read per zone",
+ .def = "0",
+ },
+ {
+ .name = "zonerange",
+ .type = FIO_OPT_STR_VAL,
+ .off1 = td_var_offset(zone_range),
.help = "Give size of an IO zone",
.def = "0",
},
{
.name = "rate",
.type = FIO_OPT_INT,
- .off1 = td_var_offset(rate[0]),
- .off2 = td_var_offset(rate[1]),
+ .off1 = td_var_offset(rate[DDIR_READ]),
+ .off2 = td_var_offset(rate[DDIR_WRITE]),
+ .off3 = td_var_offset(rate[DDIR_TRIM]),
.help = "Set bandwidth rate",
},
{
.name = "ratemin",
.type = FIO_OPT_INT,
- .off1 = td_var_offset(ratemin[0]),
- .off2 = td_var_offset(ratemin[1]),
+ .off1 = td_var_offset(ratemin[DDIR_READ]),
+ .off2 = td_var_offset(ratemin[DDIR_WRITE]),
+ .off3 = td_var_offset(ratemin[DDIR_TRIM]),
.help = "Job must meet this rate or it will be shutdown",
.parent = "rate",
},
{
.name = "rate_iops",
.type = FIO_OPT_INT,
- .off1 = td_var_offset(rate_iops[0]),
- .off2 = td_var_offset(rate_iops[1]),
+ .off1 = td_var_offset(rate_iops[DDIR_READ]),
+ .off2 = td_var_offset(rate_iops[DDIR_WRITE]),
+ .off3 = td_var_offset(rate_iops[DDIR_TRIM]),
.help = "Limit IO used to this number of IO operations/sec",
},
{
.name = "rate_iops_min",
.type = FIO_OPT_INT,
- .off1 = td_var_offset(rate_iops_min[0]),
- .off2 = td_var_offset(rate_iops_min[1]),
+ .off1 = td_var_offset(rate_iops_min[DDIR_READ]),
+ .off2 = td_var_offset(rate_iops_min[DDIR_WRITE]),
+ .off3 = td_var_offset(rate_iops_min[DDIR_TRIM]),
.help = "Job must meet this rate or it will be shut down",
.parent = "rate_iops",
},
.def = "1000",
.parent = "rate",
},
+ {
+ .name = "max_latency",
+ .type = FIO_OPT_INT,
+ .off1 = td_var_offset(max_latency),
+ .help = "Maximum tolerated IO latency (usec)",
+ },
{
.name = "invalidate",
.type = FIO_OPT_BOOL,
.help = "Create files when they are opened for IO",
.def = "0",
},
+ {
+ .name = "create_only",
+ .type = FIO_OPT_BOOL,
+ .off1 = td_var_offset(create_only),
+ .help = "Only perform file creation phase",
+ .def = "0",
+ },
{
.name = "pre_read",
.type = FIO_OPT_BOOL,
.cb = str_cpus_allowed_cb,
.help = "Set CPUs allowed",
},
+#endif
+#ifdef FIO_HAVE_LIBNUMA
+ {
+ .name = "numa_cpu_nodes",
+ .type = FIO_OPT_STR,
+ .cb = str_numa_cpunodes_cb,
+ .help = "NUMA CPU nodes bind",
+ },
+ {
+ .name = "numa_mem_policy",
+ .type = FIO_OPT_STR,
+ .cb = str_numa_mpol_cb,
+ .help = "NUMA memory policy setup",
+ },
#endif
{
.name = "end_fsync",
.cb = str_write_iops_log_cb,
.help = "Write log of IOPS during run",
},
+ {
+ .name = "log_avg_msec",
+ .type = FIO_OPT_INT,
+ .off1 = td_var_offset(log_avg_msec),
+ .help = "Average bw/iops/lat logs over this period of time",
+ .def = "0",
+ },
{
.name = "hugepage-size",
.type = FIO_OPT_INT,
.help = "Slightly scramble buffers on every IO submit",
.def = "1",
},
+ {
+ .name = "buffer_compress_percentage",
+ .type = FIO_OPT_INT,
+ .off1 = td_var_offset(compress_percentage),
+ .maxval = 100,
+ .minval = 1,
+ .help = "How compressible the buffer is (approximately)",
+ },
+ {
+ .name = "buffer_compress_chunk",
+ .type = FIO_OPT_INT,
+ .off1 = td_var_offset(compress_chunk),
+ .parent = "buffer_compress_percentage",
+ .help = "Size of compressible region in buffer",
+ },
{
.name = "clat_percentiles",
.type = FIO_OPT_BOOL,
},
{
.name = "continue_on_error",
- .type = FIO_OPT_BOOL,
+ .type = FIO_OPT_STR,
.off1 = td_var_offset(continue_on_error),
.help = "Continue on non-fatal errors during IO",
+ .def = "none",
+ .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",
+ },
+ {
+ .name = "error_dump",
+ .type = FIO_OPT_BOOL,
+ .off1 = td_var_offset(error_dump),
.def = "0",
+ .help = "Dump info on each error",
},
+
{
.name = "profile",
.type = FIO_OPT_STR_STORE,
.off1 = td_var_offset(gid),
.help = "Run job with this group ID",
},
+ {
+ .name = "flow_id",
+ .type = FIO_OPT_INT,
+ .off1 = td_var_offset(flow_id),
+ .help = "The flow index ID to use",
+ .def = "0",
+ },
+ {
+ .name = "flow",
+ .type = FIO_OPT_INT,
+ .off1 = td_var_offset(flow),
+ .help = "Weight for flow control of this job",
+ .parent = "flow_id",
+ .def = "0",
+ },
+ {
+ .name = "flow_watermark",
+ .type = FIO_OPT_INT,
+ .off1 = td_var_offset(flow_watermark),
+ .help = "High watermark for flow control. This option"
+ " should be set to the same value for all threads"
+ " with non-zero flow.",
+ .parent = "flow_id",
+ .def = "1024",
+ },
+ {
+ .name = "flow_sleep",
+ .type = FIO_OPT_INT,
+ .off1 = td_var_offset(flow_sleep),
+ .help = "How many microseconds to sleep after being held"
+ " back by the flow control mechanism",
+ .parent = "flow_id",
+ .def = "0",
+ },
{
.name = NULL,
},
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);