Add large block support to zpios(1) benchmark

As part of the large block support effort, it makes sense to add
support for large blocks to **zpios(1)**. The specifying of a zfs
block size for zpios is optional and will default to 128K if the
block size is not specified.

  `zpios ... -S size | --blocksize size ...`

This will use *size* ZFS blocks for each test, specified as a comma
delimited list with an optional unit suffix. The supported range is
powers of two from 128K through 16M. A range of block sizes can be
tested as follows: `-S 128K,256K,512K,1M`

Example run below
(non realistic results from a VM and output abbreviated for space)

```
 --regioncount=750 --regionsize=8M --chunksize=1M --offset=4K
 --threaddelay=0 --cleanup --human-readable --verbose --cleanup
 --blocksize=128K,256K,512K,1M

 th-cnt  rg-cnt  rg-sz  ch-sz  blksz  wr-data wr-bw   rd-data rd-bw
---------------------------------------------------------------------
 4       750     8m     1m     128k   5g      90.06m  5g      93.37m
 4       750     8m     1m     256k   5g      79.71m  5g      99.81m
 4       750     8m     1m     512k   5g      42.20m  5g      93.14m
 4       750     8m     1m     1m     5g      35.51m  5g      89.36m
 8       750     8m     1m     128k   5g      85.49m  5g      90.81m
 8       750     8m     1m     256k   5g      61.42m  5g      99.24m
 8       750     8m     1m     512k   5g      49.09m  5g     108.78m
 16      750     8m     1m     128k   5g      86.28m  5g      88.73m
 16      750     8m     1m     256k   5g      64.34m  5g      93.47m
 16      750     8m     1m     512k   5g      68.84m  5g     124.47m
 16      750     8m     1m     1m     5g      53.97m  5g      97.20m
---------------------------------------------------------------------
```

Signed-off-by: Don Brady <[email protected]>
Signed-off-by: Brian Behlendorf <[email protected]>
Closes #3795
Closes #2071

7 files changed, 127 insertions, 17 deletions

diff --git a/cmd/zpios/zpios.h b/cmd/zpios/zpios.h
index 92d96fcbe..4a69b9e54 100644
--- a/cmd/zpios/zpios.h
+++ b/cmd/zpios/zpios.h
@@ -29,6 +29,8 @@
  *
  *  You should have received a copy of the GNU General Public License along
  *  with ZPIOS.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ *  Copyright (c) 2015, Intel Corporation.
  */
 
 #ifndef _ZPIOS_H
@@ -39,10 +41,10 @@
 #define	VERSION_SIZE		64
 
 /* Regular expressions */
-#define	REGEX_NUMBERS		"^[0-9]*[0-9]$"
+#define	REGEX_NUMBERS		"^[0-9]+$"
 #define	REGEX_NUMBERS_COMMA	"^([0-9]+,)*[0-9]+$"
-#define	REGEX_SIZE		"^[0-9][0-9]*[kmgt]$"
-#define	REGEX_SIZE_COMMA	"^([0-9][0-9]*[kmgt]+,)*[0-9][0-9]*[kmgt]$"
+#define	REGEX_SIZE		"^[0-9]+[kKmMgGtT]?$"
+#define	REGEX_SIZE_COMMA	"^([0-9]+[kKmMgGtT]?,)*[0-9]+[kKmMgGtT]?$"
 
 /* Flags for low, high, incr */
 #define	FLAG_SET		0x01
@@ -82,10 +84,12 @@ typedef struct cmd_args {
 	range_repeat_t O;		/* Offset count */
 	range_repeat_t C;		/* Chunksize */
 	range_repeat_t S;		/* Regionsize */
+	range_repeat_t B;		/* Blocksize */
 
 	const char *pool;		/* Pool */
 	const char *name;		/* Name */
 	uint32_t flags;			/* Flags */
+	uint32_t block_size;		/* ZFS block size */
 	uint32_t io_type;		/* DMUIO only */
 	uint32_t verbose;		/* Verbose */
 	uint32_t human_readable;	/* Human readable output */
@@ -105,6 +109,7 @@ typedef struct cmd_args {
 	uint64_t current_C;
 	uint64_t current_S;
 	uint64_t current_O;
+	uint64_t current_B;
 
 	uint32_t rc;
 } cmd_args_t;
diff --git a/cmd/zpios/zpios_main.c b/cmd/zpios/zpios_main.c
index 971a886a3..e6e88f60d 100644
--- a/cmd/zpios/zpios_main.c
+++ b/cmd/zpios/zpios_main.c
@@ -29,6 +29,8 @@
  *
  *  You should have received a copy of the GNU General Public License along
  *  with ZPIOS.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ *  Copyright (c) 2015, Intel Corporation.
  */
 
 #include <stdlib.h>
@@ -44,7 +46,7 @@
 
 static const char short_opt[] =
 	"t:l:h:e:n:i:j:k:o:m:q:r:c:a:b:g:s:A:B:C:"
-	"L:p:M:xP:R:G:I:N:T:VzOfHv?";
+	"S:L:p:M:xP:R:G:I:N:T:VzOfHv?";
 static const struct option long_opt[] = {
 	{"threadcount",		required_argument,	0,	't' },
 	{"threadcount_low",	required_argument,	0,	'l' },
@@ -66,6 +68,7 @@ static const struct option long_opt[] = {
 	{"regionsize_low",	required_argument,	0,	'A' },
 	{"regionsize_high",	required_argument,	0,	'B' },
 	{"regionsize_incr",	required_argument,	0,	'C' },
+	{"blocksize",		required_argument,	0,	'S' },
 	{"load",		required_argument,	0,	'L' },
 	{"pool",		required_argument,	0,	'p' },
 	{"name",		required_argument,	0,	'M' },
@@ -116,6 +119,7 @@ usage(void)
 		"	--regionsize_low    -A    =value\n"
 		"	--regionsize_high   -B    =value\n"
 		"	--regionsize_incr   -C    =value\n"
+		"	--blocksize         -S    =values\n"
 		"	--load              -L    =dmuio|ssf|fpp\n"
 		"	--pool              -p    =pool name\n"
 		"	--name              -M    =test name\n"
@@ -143,6 +147,11 @@ static void args_fini(cmd_args_t *args)
 	free(args);
 }
 
+/* block size is 128K to 16M, power of 2 */
+#define	MIN_BLKSIZE	(128ULL << 10)
+#define	MAX_BLKSIZE	(16ULL << 20)
+#define	POW_OF_TWO(x)	(((x) & ((x) - 1)) == 0)
+
 static cmd_args_t *
 args_init(int argc, char **argv)
 {
@@ -152,7 +161,8 @@ args_init(int argc, char **argv)
 	uint32_t fl_of = 0;
 	uint32_t fl_rs = 0;
 	uint32_t fl_cs = 0;
-	int c, rc;
+	uint32_t fl_bs = 0;
+	int c, rc, i;
 
 	if (argc == 1) {
 		usage();
@@ -166,6 +176,11 @@ args_init(int argc, char **argv)
 
 	memset(args, 0, sizeof (*args));
 
+	/* provide a default block size of 128K */
+	args->B.next_val = 0;
+	args->B.val[0] = MIN_BLKSIZE;
+	args->B.val_count = 1;
+
 	while ((c = getopt_long(argc, argv, short_opt, long_opt, NULL)) != -1) {
 		rc = 0;
 
@@ -250,6 +265,10 @@ args_init(int argc, char **argv)
 			rc = set_lhi(REGEX_NUMBERS, &args->S, optarg,
 			    FLAG_INCR, &fl_rs, "regionsize_incr");
 			break;
+		case 'S': /* --blocksize */
+			rc = set_count(REGEX_SIZE, REGEX_SIZE_COMMA,
+			    &args->B, optarg, &fl_bs, "blocksize");
+			break;
 		case 'L': /* --load */
 			rc = set_load_params(args, optarg);
 			break;
@@ -339,6 +358,17 @@ args_init(int argc, char **argv)
 		return (NULL);
 	}
 
+	/* validate block size(s) */
+	for (i = 0; i < args->B.val_count; i++) {
+		int bs = args->B.val[i];
+
+		if (bs < MIN_BLKSIZE || bs > MAX_BLKSIZE || !POW_OF_TWO(bs)) {
+			fprintf(stderr, "Error: invalid block size %d\n", bs);
+			args_fini(args);
+			return (NULL);
+		}
+	}
+
 	return (args);
 }
 
@@ -480,7 +510,7 @@ get_next(uint64_t *val, range_repeat_t *range)
 
 static int
 run_one(cmd_args_t *args, uint32_t id, uint32_t T, uint32_t N,
-    uint64_t C, uint64_t S, uint64_t O)
+    uint64_t C, uint64_t S, uint64_t O, uint64_t B)
 {
 	zpios_cmd_t *cmd;
 	int rc, rc2, cmd_size;
@@ -506,6 +536,7 @@ run_one(cmd_args_t *args, uint32_t id, uint32_t T, uint32_t N,
 	cmd->cmd_region_count = N;
 	cmd->cmd_region_size = S;
 	cmd->cmd_offset = O;
+	cmd->cmd_block_size = B;
 	cmd->cmd_region_noise = args->regionnoise;
 	cmd->cmd_chunk_noise = args->chunknoise;
 	cmd->cmd_thread_delay = args->thread_delay;
@@ -541,7 +572,7 @@ run_offsets(cmd_args_t *args)
 	while (rc == 0 && get_next(&args->current_O, &args->O)) {
 		rc = run_one(args, args->current_id,
 		    args->current_T, args->current_N, args->current_C,
-		    args->current_S, args->current_O);
+		    args->current_S, args->current_O, args->current_B);
 		args->current_id++;
 	}
 
@@ -594,12 +625,26 @@ run_chunk_sizes(cmd_args_t *args)
 }
 
 static int
+run_block_sizes(cmd_args_t *args)
+{
+	int rc = 0;
+
+	while (rc == 0 && get_next(&args->current_B, &args->B)) {
+		rc = run_chunk_sizes(args);
+	}
+
+	args->B.next_val = 0;
+	return (rc);
+}
+
+
+static int
 run_thread_counts(cmd_args_t *args)
 {
 	int rc = 0;
 
 	while (rc == 0 && get_next((uint64_t *)&args->current_T, &args->T))
-		rc = run_chunk_sizes(args);
+		rc = run_block_sizes(args);
 
 	return (rc);
 }
diff --git a/cmd/zpios/zpios_util.c b/cmd/zpios/zpios_util.c
index b226322b0..2d248ed9a 100644
--- a/cmd/zpios/zpios_util.c
+++ b/cmd/zpios/zpios_util.c
@@ -29,6 +29,8 @@
  *
  *  You should have received a copy of the GNU General Public License along
  *  with ZPIOS.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ *  Copyright (c) 2015, Intel Corporation.
  */
 
 #include <stdlib.h>
@@ -185,6 +187,8 @@ int
 set_count(char *pattern1, char *pattern2, range_repeat_t *range,
     char *optarg, uint32_t *flags, char *arg)
 {
+	uint64_t count = range->val_count;
+
 	if (flags)
 		*flags |= FLAG_SET;
 
@@ -197,6 +201,9 @@ set_count(char *pattern1, char *pattern2, range_repeat_t *range,
 		fprintf(stderr, "Error: Incorrect pattern for %s, '%s'\n",
 		    arg, optarg);
 		return (EINVAL);
+	} else if (count == range->val_count) {
+		fprintf(stderr, "Error: input ignored for %s, '%s'\n",
+		    arg, optarg);
 	}
 
 	return (0);
@@ -314,14 +321,14 @@ print_stats_header(cmd_args_t *args)
 	if (args->verbose) {
 		printf(
 		    "status    name        id\tth-cnt\trg-cnt\trg-sz\t"
-		    "ch-sz\toffset\trg-no\tch-no\tth-dly\tflags\ttime\t"
+		    "ch-sz\toffset\trg-no\tch-no\tth-dly\tflags\tblksz\ttime\t"
 		    "cr-time\trm-time\twr-time\trd-time\twr-data\twr-ch\t"
 		    "wr-bw\trd-data\trd-ch\trd-bw\n");
 		printf(
-		    "------------------------------------------------"
-		    "------------------------------------------------"
-		    "------------------------------------------------"
-		    "----------------------------------------------\n");
+		    "-------------------------------------------------"
+		    "-------------------------------------------------"
+		    "-------------------------------------------------"
+		    "--------------------------------------------------\n");
 	} else {
 		printf(
 		    "status    name        id\t"
@@ -358,6 +365,7 @@ print_stats_human_readable(cmd_args_t *args, zpios_cmd_t *cmd)
 		printf("%s\t", uint64_to_kmgt(str, cmd->cmd_chunk_noise));
 		printf("%s\t", uint64_to_kmgt(str, cmd->cmd_thread_delay));
 		printf("%s\t", print_flags(str, cmd->cmd_flags));
+		printf("%s\t", uint64_to_kmgt(str, cmd->cmd_block_size));
 	}
 
 	if (args->rc) {
@@ -414,6 +422,7 @@ print_stats_table(cmd_args_t *args, zpios_cmd_t *cmd)
 		printf("%u\t", cmd->cmd_chunk_noise);
 		printf("%u\t", cmd->cmd_thread_delay);
 		printf("0x%x\t", cmd->cmd_flags);
+		printf("%u\t", cmd->cmd_block_size);
 	}
 
 	if (args->rc) {
diff --git a/include/zpios-ctl.h b/include/zpios-ctl.h
index 9a47ff91d..aee4f0a50 100644
--- a/include/zpios-ctl.h
+++ b/include/zpios-ctl.h
@@ -29,6 +29,8 @@
  *
  *  You should have received a copy of the GNU General Public License along
  *  with ZPIOS.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ *  Copyright (c) 2015, Intel Corporation.
  */
 
 #ifndef _ZPIOS_CTL_H
@@ -116,6 +118,7 @@ typedef struct zpios_cmd {
 	uint32_t cmd_chunk_noise;	/* Chunk noise */
 	uint32_t cmd_thread_delay;	/* Thread delay */
 	uint32_t cmd_flags;		/* Test flags */
+	uint32_t cmd_block_size;	/* ZFS block size */
 	char cmd_pre[ZPIOS_PATH_SIZE];	/* Pre-exec hook */
 	char cmd_post[ZPIOS_PATH_SIZE];	/* Post-exec hook */
 	char cmd_log[ZPIOS_PATH_SIZE];  /* Requested log dir */
diff --git a/include/zpios-internal.h b/include/zpios-internal.h
index 4b99b4ce3..dd2bd2343 100644
--- a/include/zpios-internal.h
+++ b/include/zpios-internal.h
@@ -29,6 +29,8 @@
  *
  *  You should have received a copy of the GNU General Public License along
  *  with ZPIOS.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ *  Copyright (c) 2015, Intel Corporation.
  */
 
 #ifndef _ZPIOS_INTERNAL_H
@@ -79,6 +81,7 @@ typedef struct run_args {
 	__u32 chunk_noise;
 	__u32 thread_delay;
 	__u32 flags;
+	__u32 block_size;
 	char pre[ZPIOS_PATH_SIZE];
 	char post[ZPIOS_PATH_SIZE];
 	char log[ZPIOS_PATH_SIZE];
diff --git a/man/man1/zpios.1 b/man/man1/zpios.1
index 4b362b09d..4334c03c0 100644
--- a/man/man1/zpios.1
+++ b/man/man1/zpios.1
@@ -22,6 +22,8 @@
 .\"
 .\" Copyright 2013 Darik Horn <[email protected]>. All rights reserved.
 .\"
+.\" Copyright (c) 2015, Intel Corporation.
+.\"
 .TH zpios 1 "2013 FEB 28" "ZFS on Linux" "User Commands"
 
 .SH NAME
@@ -36,10 +38,10 @@ not depend on the ZFS Posix Layer ("ZPL").
 
 .SH OPTIONS
 .HP
-.BI "\-s" " regex" ", \-\-threadcount" " regex"
+.BI "\-t" " regex" ", \-\-threadcount" " regex"
 .IP
 Start this many threads for each test series, specified as a comma
-delimited regular expression. (eg: "-s 1,2,3")
+delimited regular expression. (eg: "-t 1,2,3")
 .IP
 This option is mutually exclusive with the \fBthreadcount_*\fR
 options.
@@ -120,6 +122,35 @@ chunk size for the last test.
 These three options must be specified together and are mutually
 exclusive with the \fBchunksize\fR option.
 .HP
+.BI "\-s" " size" ", \-\-regionsize" " size"
+.IP
+Use \fIsize\fR regions for each test, specified as a comma delimited
+regular expression with an optional unit suffix. (eg: "-s 1M" means
+one megabyte.)
+.IP
+This option is mutually exclusive with the \fBregionsize_*\fB options.
+.HP
+.BI "\-A" " size_low" ", \-\-regionsize_low" " size_low"
+.HP
+.BI "\-B" " size_high" ", \-\-regionsize_high" " size_high"
+.HP
+.BI "\-C" " size_incr" ", \-\-regionsize_incr" " size_incr"
+.IP
+Use a \fIsize_low\fR region size for the first test, add \fIsize_incr\fR
+to the region size for each subsequent test, and use a \fIsize_high\fR
+region size for the last test.
+.IP
+These three options must be specified together and are mutually
+exclusive with the \fBregionsize\fR option.
+.HP
+.BI "\-S" " size | sizes" ", \-\-blocksize" " size | sizes"
+.IP
+Use \fIsize\fR ZFS blocks for each test, specified as a comma delimited
+regular expression with an optional unit suffix. (eg: "-S 1M" means
+one megabyte.) The supported range is powers of two from 128K through 16M.
+A range of blocks can be tested as follows: "-S 128K,256K,512K,1M".
+.IP
+.HP
 .BI "\-L" " dmu_flags" ", \-\-load" " dmu_flags"
 .IP
 Specify \fIdmuio\fR for regular DMU_IO, \fIssf\fR for single shared
diff --git a/module/zpios/pios.c b/module/zpios/pios.c
index e3a85c168..8f4a8fd69 100644
--- a/module/zpios/pios.c
+++ b/module/zpios/pios.c
@@ -29,10 +29,13 @@
  *
  *  You should have received a copy of the GNU General Public License along
  *  with ZPIOS.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ *  Copyright (c) 2015, Intel Corporation.
  */
 
 #include <sys/zfs_context.h>
 #include <sys/dmu.h>
+#include <sys/spa.h>
 #include <sys/txg.h>
 #include <sys/dsl_destroy.h>
 #include <linux/miscdevice.h>
@@ -129,8 +132,17 @@ zpios_dmu_object_create(run_args_t *run_args, objset_t *os)
 {
 	struct dmu_tx *tx;
 	uint64_t obj = 0ULL;
+	uint64_t blksize = run_args->block_size;
 	int rc;
 
+	if (blksize < SPA_MINBLOCKSIZE ||
+	    blksize > spa_maxblocksize(dmu_objset_spa(os)) ||
+	    !ISP2(blksize)) {
+		zpios_print(run_args->file,
+		    "invalid block size for pool: %d\n", (int)blksize);
+		return (obj);
+	}
+
 	tx = dmu_tx_create(os);
 	dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, OBJ_SIZE);
 	rc = dmu_tx_assign(tx, TXG_WAIT);
@@ -142,10 +154,11 @@ zpios_dmu_object_create(run_args_t *run_args, objset_t *os)
 	}
 
 	obj = dmu_object_alloc(os, DMU_OT_UINT64_OTHER, 0, DMU_OT_NONE, 0, tx);
-	rc = dmu_object_set_blocksize(os, obj, 128ULL << 10, 0, tx);
+	rc = dmu_object_set_blocksize(os, obj, blksize, 0, tx);
 	if (rc) {
 		zpios_print(run_args->file,
-			    "dmu_object_set_blocksize() failed: %d\n", rc);
+		    "dmu_object_set_blocksize to %d failed: %d\n",
+		    (int)blksize, rc);
 		dmu_tx_abort(tx);
 		return (obj);
 	}
@@ -295,6 +308,7 @@ zpios_setup_run(run_args_t **run_args, zpios_cmd_t *kcmd, struct file *file)
 	ra->chunk_noise		= kcmd->cmd_chunk_noise;
 	ra->thread_delay	= kcmd->cmd_thread_delay;
 	ra->flags		= kcmd->cmd_flags;
+	ra->block_size		= kcmd->cmd_block_size;
 	ra->stats.wr_data	= 0;
 	ra->stats.wr_chunks	= 0;
 	ra->stats.rd_data	= 0;