X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=verify.c;h=91a9077ac272d2ffcc04f5b4e7e00a2eb879944e;hp=dc107087cefbabc5630a0f2cdcfec8e199743b4f;hb=323f58c5ee9eb695e4c8f257ee955625480b3729;hpb=03e20d687566753b90383571e5e152c5142bdffd

diff --git a/verify.c b/verify.c
index dc107087..19bec75a 100644
--- a/verify.c
+++ b/verify.c
@@ -6,12 +6,14 @@
 #include <string.h>
 #include <assert.h>
 #include <pthread.h>
+#include <libgen.h>
 
 #include "fio.h"
 #include "verify.h"
-#include "smalloc.h"
 #include "trim.h"
 #include "lib/rand.h"
+#include "lib/hweight.h"
+#include "lib/pattern.h"
 
 #include "crc/md5.h"
 #include "crc/crc64.h"
@@ -22,45 +24,92 @@
 #include "crc/sha256.h"
 #include "crc/sha512.h"
 #include "crc/sha1.h"
+#include "crc/xxhash.h"
+
+static void populate_hdr(struct thread_data *td, struct io_u *io_u,
+			 struct verify_header *hdr, unsigned int header_num,
+			 unsigned int header_len);
+static void fill_hdr(struct thread_data *td, struct io_u *io_u,
+		     struct verify_header *hdr, unsigned int header_num,
+		     unsigned int header_len, uint64_t rand_seed);
+static void __fill_hdr(struct thread_data *td, struct io_u *io_u,
+		       struct verify_header *hdr, unsigned int header_num,
+		       unsigned int header_len, uint64_t rand_seed);
+
+void fill_buffer_pattern(struct thread_data *td, void *p, unsigned int len)
+{
+	(void)cpy_pattern(td->o.buffer_pattern, td->o.buffer_pattern_bytes, p, len);
+}
 
-void fill_pattern(struct thread_data *td, void *p, unsigned int len, struct io_u *io_u)
+void __fill_buffer(struct thread_options *o, unsigned long seed, void *p,
+		   unsigned int len)
 {
-	switch (td->o.verify_pattern_bytes) {
-	case 0:
+	__fill_random_buf_percentage(seed, p, o->compress_percentage, len, len, o->buffer_pattern, o->buffer_pattern_bytes);
+}
+
+unsigned long fill_buffer(struct thread_data *td, void *p, unsigned int len)
+{
+	struct frand_state *fs = &td->verify_state;
+	struct thread_options *o = &td->o;
+
+	return fill_random_buf_percentage(fs, p, o->compress_percentage, len, len, o->buffer_pattern, o->buffer_pattern_bytes);
+}
+
+void fill_verify_pattern(struct thread_data *td, void *p, unsigned int len,
+			 struct io_u *io_u, unsigned long seed, int use_seed)
+{
+	struct thread_options *o = &td->o;
+
+	if (!o->verify_pattern_bytes) {
 		dprint(FD_VERIFY, "fill random bytes len=%u\n", len);
-		fill_random_buf(p, len);
-		break;
-	case 1:
-		if (io_u->buf_filled_len >= len) {
-			dprint(FD_VERIFY, "using already filled verify pattern b=0 len=%u\n", len);
-			return;
-		}
-		dprint(FD_VERIFY, "fill verify pattern b=0 len=%u\n", len);
-		memset(p, td->o.verify_pattern[0], len);
-		io_u->buf_filled_len = len;
-		break;
-	default: {
-		unsigned int i = 0, size = 0;
-		unsigned char *b = p;
-
-		if (io_u->buf_filled_len >= len) {
-			dprint(FD_VERIFY, "using already filled verify pattern b=%d len=%u\n",
-					td->o.verify_pattern_bytes, len);
-			return;
-		}
-		dprint(FD_VERIFY, "fill verify pattern b=%d len=%u\n",
-					td->o.verify_pattern_bytes, len);
-
-		while (i < len) {
-			size = td->o.verify_pattern_bytes;
-			if (size > (len - i))
-				size = len - i;
-			memcpy(b+i, td->o.verify_pattern, size);
-			i += size;
-		}
-		io_u->buf_filled_len = len;
-		break;
-		}
+
+		if (use_seed)
+			__fill_buffer(o, seed, p, len);
+		else
+			io_u->rand_seed = fill_buffer(td, p, len);
+		return;
+	}
+
+	/* Skip if we were here and we do not need to patch pattern
+	 * with format */
+	if (!td->o.verify_fmt_sz && io_u->buf_filled_len >= len) {
+		dprint(FD_VERIFY, "using already filled verify pattern b=%d len=%u\n",
+			o->verify_pattern_bytes, len);
+		return;
+	}
+
+	(void)paste_format(td->o.verify_pattern, td->o.verify_pattern_bytes,
+			   td->o.verify_fmt, td->o.verify_fmt_sz,
+			   p, len, io_u);
+	io_u->buf_filled_len = len;
+}
+
+static unsigned int get_hdr_inc(struct thread_data *td, struct io_u *io_u)
+{
+	unsigned int hdr_inc;
+
+	hdr_inc = io_u->buflen;
+	if (td->o.verify_interval && td->o.verify_interval <= io_u->buflen)
+		hdr_inc = td->o.verify_interval;
+
+	return hdr_inc;
+}
+
+static void fill_pattern_headers(struct thread_data *td, struct io_u *io_u,
+				 unsigned long seed, int use_seed)
+{
+	unsigned int hdr_inc, header_num;
+	struct verify_header *hdr;
+	void *p = io_u->buf;
+
+	fill_verify_pattern(td, p, io_u->buflen, io_u, seed, use_seed);
+
+	hdr_inc = get_hdr_inc(td, io_u);
+	header_num = 0;
+	for (; p < io_u->buf + io_u->buflen; p += hdr_inc) {
+		hdr = p;
+		populate_hdr(td, io_u, hdr, header_num, hdr_inc);
+		header_num++;
 	}
 }
 
@@ -86,7 +135,7 @@ static void hexdump(void *buffer, int len)
 }
 
 /*
- * Prepare for seperation of verify_header and checksum header
+ * Prepare for separation of verify_header and checksum header
  */
 static inline unsigned int __hdr_size(int verify_type)
 {
@@ -94,7 +143,9 @@ static inline unsigned int __hdr_size(int verify_type)
 
 	switch (verify_type) {
 	case VERIFY_NONE:
+	case VERIFY_HDR_ONLY:
 	case VERIFY_NULL:
+	case VERIFY_PATTERN:
 		len = 0;
 		break;
 	case VERIFY_MD5:
@@ -120,12 +171,14 @@ static inline unsigned int __hdr_size(int verify_type)
 	case VERIFY_SHA512:
 		len = sizeof(struct vhdr_sha512);
 		break;
-	case VERIFY_META:
-		len = sizeof(struct vhdr_meta);
+	case VERIFY_XXHASH:
+		len = sizeof(struct vhdr_xxhash);
 		break;
 	case VERIFY_SHA1:
 		len = sizeof(struct vhdr_sha1);
 		break;
+	case VERIFY_PATTERN_NO_HDR:
+		return 0;
 	default:
 		log_err("fio: unknown verify header!\n");
 		assert(0);
@@ -134,8 +187,12 @@ static inline unsigned int __hdr_size(int verify_type)
 	return len + sizeof(struct verify_header);
 }
 
-static inline unsigned int hdr_size(struct verify_header *hdr)
+static inline unsigned int hdr_size(struct thread_data *td,
+				    struct verify_header *hdr)
 {
+	if (td->o.verify == VERIFY_PATTERN_NO_HDR)
+		return 0;
+
 	return __hdr_size(hdr->verify_type);
 }
 
@@ -156,6 +213,7 @@ struct vcont {
 	 */
 	struct io_u *io_u;
 	unsigned int hdr_num;
+	struct thread_data *td;
 
 	/*
 	 * Output, only valid in case of error
@@ -166,6 +224,112 @@ struct vcont {
 	unsigned int crc_len;
 };
 
+#define DUMP_BUF_SZ	255
+static int dump_buf_warned;
+
+static void dump_buf(char *buf, unsigned int len, unsigned long long offset,
+		     const char *type, struct fio_file *f)
+{
+	char *ptr, fname[DUMP_BUF_SZ];
+	size_t buf_left = DUMP_BUF_SZ;
+	int ret, fd;
+
+	ptr = strdup(f->file_name);
+
+	memset(fname, 0, sizeof(fname));
+	if (aux_path)
+		sprintf(fname, "%s%s", aux_path, FIO_OS_PATH_SEPARATOR);
+
+	strncpy(fname + strlen(fname), basename(ptr), buf_left - 1);
+
+	buf_left -= strlen(fname);
+	if (buf_left <= 0) {
+		if (!dump_buf_warned) {
+			log_err("fio: verify failure dump buffer too small\n");
+			dump_buf_warned = 1;
+		}
+		free(ptr);
+		return;
+	}
+
+	snprintf(fname + strlen(fname), buf_left, ".%llu.%s", offset, type);
+
+	fd = open(fname, O_CREAT | O_TRUNC | O_WRONLY, 0644);
+	if (fd < 0) {
+		perror("open verify buf file");
+		return;
+	}
+
+	while (len) {
+		ret = write(fd, buf, len);
+		if (!ret)
+			break;
+		else if (ret < 0) {
+			perror("write verify buf file");
+			break;
+		}
+		len -= ret;
+		buf += ret;
+	}
+
+	close(fd);
+	log_err("       %s data dumped as %s\n", type, fname);
+	free(ptr);
+}
+
+/*
+ * Dump the contents of the read block and re-generate the correct data
+ * and dump that too.
+ */
+static void __dump_verify_buffers(struct verify_header *hdr, struct vcont *vc)
+{
+	struct thread_data *td = vc->td;
+	struct io_u *io_u = vc->io_u;
+	unsigned long hdr_offset;
+	struct io_u dummy;
+	void *buf;
+
+	if (!td->o.verify_dump)
+		return;
+
+	/*
+	 * Dump the contents we just read off disk
+	 */
+	hdr_offset = vc->hdr_num * hdr->len;
+
+	dump_buf(io_u->buf + hdr_offset, hdr->len, io_u->offset + hdr_offset,
+			"received", vc->io_u->file);
+
+	/*
+	 * Allocate a new buf and re-generate the original data
+	 */
+	buf = malloc(io_u->buflen);
+	dummy = *io_u;
+	dummy.buf = buf;
+	dummy.rand_seed = hdr->rand_seed;
+	dummy.buf_filled_len = 0;
+	dummy.buflen = io_u->buflen;
+
+	fill_pattern_headers(td, &dummy, hdr->rand_seed, 1);
+
+	dump_buf(buf + hdr_offset, hdr->len, io_u->offset + hdr_offset,
+			"expected", vc->io_u->file);
+	free(buf);
+}
+
+static void dump_verify_buffers(struct verify_header *hdr, struct vcont *vc)
+{
+	struct thread_data *td = vc->td;
+	struct verify_header shdr;
+
+	if (td->o.verify == VERIFY_PATTERN_NO_HDR) {
+		__fill_hdr(td, vc->io_u, &shdr, 0, vc->io_u->buflen, 0);
+		hdr = &shdr;
+	}
+
+	__dump_verify_buffers(hdr, vc);
+}
+
 static void log_verify_failure(struct verify_header *hdr, struct vcont *vc)
 {
 	unsigned long long offset;
@@ -181,6 +345,8 @@ static void log_verify_failure(struct verify_header *hdr, struct vcont *vc)
 		log_err("       Received CRC: ");
 		hexdump(vc->bad_crc, vc->crc_len);
 	}
+
+	dump_verify_buffers(hdr, vc);
 }
 
 /*
@@ -188,21 +354,77 @@ static void log_verify_failure(struct verify_header *hdr, struct vcont *vc)
  */
 static inline void *io_u_verify_off(struct verify_header *hdr, struct vcont *vc)
 {
-	return vc->io_u->buf + vc->hdr_num * hdr->len + hdr_size(hdr);
+	return vc->io_u->buf + vc->hdr_num * hdr->len + hdr_size(vc->td, hdr);
 }
 
-static int verify_io_u_meta(struct verify_header *hdr, struct thread_data *td,
-			    struct vcont *vc)
+static int verify_io_u_pattern(struct verify_header *hdr, struct vcont *vc)
 {
-	struct vhdr_meta *vh = hdr_priv(hdr);
+	struct thread_data *td = vc->td;
 	struct io_u *io_u = vc->io_u;
+	char *buf, *pattern;
+	unsigned int header_size = __hdr_size(td->o.verify);
+	unsigned int len, mod, i, pattern_size;
+	int rc;
+
+	pattern = td->o.verify_pattern;
+	pattern_size = td->o.verify_pattern_bytes;
+	assert(pattern_size != 0);
+
+	(void)paste_format_inplace(pattern, pattern_size,
+				   td->o.verify_fmt, td->o.verify_fmt_sz, io_u);
+
+	buf = (void *) hdr + header_size;
+	len = get_hdr_inc(td, io_u) - header_size;
+	mod = (get_hdr_inc(td, io_u) * vc->hdr_num + header_size) % pattern_size;
+
+	rc = cmp_pattern(pattern, pattern_size, mod, buf, len);
+	if (!rc)
+		return 0;
+
+	/* Slow path, compare each byte */
+	for (i = 0; i < len; i++) {
+		if (buf[i] != pattern[mod]) {
+			unsigned int bits;
+
+			bits = hweight8(buf[i] ^ pattern[mod]);
+			log_err("fio: got pattern '%02x', wanted '%02x'. Bad bits %d\n",
+				(unsigned char)buf[i],
+				(unsigned char)pattern[mod],
+				bits);
+			log_err("fio: bad pattern block offset %u\n", i);
+			dump_verify_buffers(hdr, vc);
+			return EILSEQ;
+		}
+		mod++;
+		if (mod == td->o.verify_pattern_bytes)
+			mod = 0;
+	}
+
+	/* Unreachable line */
+	assert(0);
+	return EILSEQ;
+}
+
+static int verify_io_u_xxhash(struct verify_header *hdr, struct vcont *vc)
+{
+	void *p = io_u_verify_off(hdr, vc);
+	struct vhdr_xxhash *vh = hdr_priv(hdr);
+	uint32_t hash;
+	void *state;
 
-	dprint(FD_VERIFY, "meta verify io_u %p, len %u\n", io_u, hdr->len);
+	dprint(FD_VERIFY, "xxhash verify io_u %p, len %u\n", vc->io_u, hdr->len);
 
-	if (vh->offset == io_u->offset + vc->hdr_num * td->o.verify_interval)
+	state = XXH32_init(1);
+	XXH32_update(state, p, hdr->len - hdr_size(vc->td, hdr));
+	hash = XXH32_digest(state);
+
+	if (vh->hash == hash)
 		return 0;
 
-	vc->name = "meta";
+	vc->name = "xxhash";
+	vc->good_crc = &vh->hash;
+	vc->bad_crc = &hash;
+	vc->crc_len = sizeof(hash);
 	log_verify_failure(hdr, vc);
 	return EILSEQ;
 }
@@ -212,14 +434,14 @@ static int verify_io_u_sha512(struct verify_header *hdr, struct vcont *vc)
 	void *p = io_u_verify_off(hdr, vc);
 	struct vhdr_sha512 *vh = hdr_priv(hdr);
 	uint8_t sha512[128];
-	struct sha512_ctx sha512_ctx = {
+	struct fio_sha512_ctx sha512_ctx = {
 		.buf = sha512,
 	};
 
 	dprint(FD_VERIFY, "sha512 verify io_u %p, len %u\n", vc->io_u, hdr->len);
 
-	sha512_init(&sha512_ctx);
-	sha512_update(&sha512_ctx, p, hdr->len - hdr_size(hdr));
+	fio_sha512_init(&sha512_ctx);
+	fio_sha512_update(&sha512_ctx, p, hdr->len - hdr_size(vc->td, hdr));
 
 	if (!memcmp(vh->sha512, sha512_ctx.buf, sizeof(sha512)))
 		return 0;
@@ -237,14 +459,15 @@ static int verify_io_u_sha256(struct verify_header *hdr, struct vcont *vc)
 	void *p = io_u_verify_off(hdr, vc);
 	struct vhdr_sha256 *vh = hdr_priv(hdr);
 	uint8_t sha256[64];
-	struct sha256_ctx sha256_ctx = {
+	struct fio_sha256_ctx sha256_ctx = {
 		.buf = sha256,
 	};
 
 	dprint(FD_VERIFY, "sha256 verify io_u %p, len %u\n", vc->io_u, hdr->len);
 
-	sha256_init(&sha256_ctx);
-	sha256_update(&sha256_ctx, p, hdr->len - hdr_size(hdr));
+	fio_sha256_init(&sha256_ctx);
+	fio_sha256_update(&sha256_ctx, p, hdr->len - hdr_size(vc->td, hdr));
+	fio_sha256_final(&sha256_ctx);
 
 	if (!memcmp(vh->sha256, sha256_ctx.buf, sizeof(sha256)))
 		return 0;
@@ -262,14 +485,15 @@ static int verify_io_u_sha1(struct verify_header *hdr, struct vcont *vc)
 	void *p = io_u_verify_off(hdr, vc);
 	struct vhdr_sha1 *vh = hdr_priv(hdr);
 	uint32_t sha1[5];
-	struct sha1_ctx sha1_ctx = {
+	struct fio_sha1_ctx sha1_ctx = {
 		.H = sha1,
 	};
 
 	dprint(FD_VERIFY, "sha1 verify io_u %p, len %u\n", vc->io_u, hdr->len);
 
-	sha1_init(&sha1_ctx);
-	sha1_update(&sha1_ctx, p, hdr->len - hdr_size(hdr));
+	fio_sha1_init(&sha1_ctx);
+	fio_sha1_update(&sha1_ctx, p, hdr->len - hdr_size(vc->td, hdr));
+	fio_sha1_final(&sha1_ctx);
 
 	if (!memcmp(vh->sha1, sha1_ctx.H, sizeof(sha1)))
 		return 0;
@@ -290,7 +514,7 @@ static int verify_io_u_crc7(struct verify_header *hdr, struct vcont *vc)
 
 	dprint(FD_VERIFY, "crc7 verify io_u %p, len %u\n", vc->io_u, hdr->len);
 
-	c = crc7(p, hdr->len - hdr_size(hdr));
+	c = fio_crc7(p, hdr->len - hdr_size(vc->td, hdr));
 
 	if (c == vh->crc7)
 		return 0;
@@ -311,7 +535,7 @@ static int verify_io_u_crc16(struct verify_header *hdr, struct vcont *vc)
 
 	dprint(FD_VERIFY, "crc16 verify io_u %p, len %u\n", vc->io_u, hdr->len);
 
-	c = crc16(p, hdr->len - hdr_size(hdr));
+	c = fio_crc16(p, hdr->len - hdr_size(vc->td, hdr));
 
 	if (c == vh->crc16)
 		return 0;
@@ -332,7 +556,7 @@ static int verify_io_u_crc64(struct verify_header *hdr, struct vcont *vc)
 
 	dprint(FD_VERIFY, "crc64 verify io_u %p, len %u\n", vc->io_u, hdr->len);
 
-	c = crc64(p, hdr->len - hdr_size(hdr));
+	c = fio_crc64(p, hdr->len - hdr_size(vc->td, hdr));
 
 	if (c == vh->crc64)
 		return 0;
@@ -353,7 +577,7 @@ static int verify_io_u_crc32(struct verify_header *hdr, struct vcont *vc)
 
 	dprint(FD_VERIFY, "crc32 verify io_u %p, len %u\n", vc->io_u, hdr->len);
 
-	c = crc32(p, hdr->len - hdr_size(hdr));
+	c = fio_crc32(p, hdr->len - hdr_size(vc->td, hdr));
 
 	if (c == vh->crc32)
 		return 0;
@@ -374,10 +598,7 @@ static int verify_io_u_crc32c(struct verify_header *hdr, struct vcont *vc)
 
 	dprint(FD_VERIFY, "crc32c verify io_u %p, len %u\n", vc->io_u, hdr->len);
 
-	if (hdr->verify_type == VERIFY_CRC32C_INTEL)
-		c = crc32c_intel(p, hdr->len - hdr_size(hdr));
-	else
-		c = crc32c(p, hdr->len - hdr_size(hdr));
+	c = fio_crc32c(p, hdr->len - hdr_size(vc->td, hdr));
 
 	if (c == vh->crc32)
 		return 0;
@@ -395,14 +616,15 @@ static int verify_io_u_md5(struct verify_header *hdr, struct vcont *vc)
 	void *p = io_u_verify_off(hdr, vc);
 	struct vhdr_md5 *vh = hdr_priv(hdr);
 	uint32_t hash[MD5_HASH_WORDS];
-	struct md5_ctx md5_ctx = {
+	struct fio_md5_ctx md5_ctx = {
 		.hash = hash,
 	};
 
 	dprint(FD_VERIFY, "md5 verify io_u %p, len %u\n", vc->io_u, hdr->len);
 
-	md5_init(&md5_ctx);
-	md5_update(&md5_ctx, p, hdr->len - hdr_size(hdr));
+	fio_md5_init(&md5_ctx);
+	fio_md5_update(&md5_ctx, p, hdr->len - hdr_size(vc->td, hdr));
+	fio_md5_final(&md5_ctx);
 
 	if (!memcmp(vh->md5_digest, md5_ctx.hash, sizeof(hash)))
 		return 0;
@@ -415,113 +637,175 @@ static int verify_io_u_md5(struct verify_header *hdr, struct vcont *vc)
 	return EILSEQ;
 }
 
-static unsigned int hweight8(unsigned int w)
+/*
+ * Push IO verification to a separate thread
+ */
+int verify_io_u_async(struct thread_data *td, struct io_u **io_u_ptr)
 {
-	unsigned int res = w - ((w >> 1) & 0x55);
+	struct io_u *io_u = *io_u_ptr;
 
-	res = (res & 0x33) + ((res >> 2) & 0x33);
-	return (res + (res >> 4)) & 0x0F;
-}
-
-int verify_io_u_pattern(char *pattern, unsigned long pattern_size,
-			char *buf, unsigned int len, unsigned int mod)
-{
-	unsigned int i;
+	pthread_mutex_lock(&td->io_u_lock);
 
-	for (i = 0; i < len; i++) {
-		if (buf[i] != pattern[mod]) {
-			unsigned int bits;
+	if (io_u->file)
+		put_file_log(td, io_u->file);
 
-			bits = hweight8(buf[i] ^ pattern[mod]);
-			log_err("fio: got pattern %x, wanted %x. Bad bits %d\n",
-				buf[i], pattern[mod], bits);
-			log_err("fio: bad pattern block offset %u\n", i);
-			return EILSEQ;
-		}
-		mod++;
-		if (mod == pattern_size)
-			mod = 0;
+	if (io_u->flags & IO_U_F_IN_CUR_DEPTH) {
+		td->cur_depth--;
+		io_u_clear(io_u, IO_U_F_IN_CUR_DEPTH);
 	}
+	flist_add_tail(&io_u->verify_list, &td->verify_list);
+	*io_u_ptr = NULL;
+	pthread_mutex_unlock(&td->io_u_lock);
 
+	pthread_cond_signal(&td->verify_cond);
 	return 0;
 }
 
 /*
- * Push IO verification to a separate thread
+ * Thanks Rusty, for spending the time so I don't have to.
+ *
+ * http://rusty.ozlabs.org/?p=560
  */
-int verify_io_u_async(struct thread_data *td, struct io_u *io_u)
+static int mem_is_zero(const void *data, size_t length)
 {
-	if (io_u->file)
-		put_file_log(td, io_u->file);
+	const unsigned char *p = data;
+	size_t len;
+
+	/* Check first 16 bytes manually */
+	for (len = 0; len < 16; len++) {
+		if (!length)
+			return 1;
+		if (*p)
+			return 0;
+		p++;
+		length--;
+	}
 
-	io_u->file = NULL;
+	/* Now we know that's zero, memcmp with self. */
+	return memcmp(data, p, length) == 0;
+}
 
-	pthread_mutex_lock(&td->io_u_lock);
-	
-	if (io_u->flags & IO_U_F_IN_CUR_DEPTH) {
-		td->cur_depth--;
-		io_u->flags &= ~IO_U_F_IN_CUR_DEPTH;
+static int mem_is_zero_slow(const void *data, size_t length, size_t *offset)
+{
+	const unsigned char *p = data;
+
+	*offset = 0;
+	while (length) {
+		if (*p)
+			break;
+		(*offset)++;
+		length--;
+		p++;
 	}
-	flist_del(&io_u->list);
-	flist_add_tail(&io_u->list, &td->verify_list);
-	io_u->flags |= IO_U_F_FREE_DEF;
-	pthread_mutex_unlock(&td->io_u_lock);
 
-	pthread_cond_signal(&td->verify_cond);
-	return 0;
+	return !length;
 }
 
 static int verify_trimmed_io_u(struct thread_data *td, struct io_u *io_u)
 {
-	static char zero_buf[1024];
-	unsigned int this_len, len;
-	int ret = 0;
-	void *p;
+	size_t offset;
 
 	if (!td->o.trim_zero)
 		return 0;
 
-	len = io_u->buflen;
-	p = io_u->buf;
-	do {
-		this_len = sizeof(zero_buf);
-		if (this_len > len)
-			this_len = len;
-		if (memcmp(p, zero_buf, this_len)) {
-			ret = EILSEQ;
-			break;
-		}
-		len -= this_len;
-		p += this_len;
-	} while (len);
-
-	if (!ret)
+	if (mem_is_zero(io_u->buf, io_u->buflen))
 		return 0;
 
+	mem_is_zero_slow(io_u->buf, io_u->buflen, &offset);
+
 	log_err("trim: verify failed at file %s offset %llu, length %lu"
 		", block offset %lu\n",
 			io_u->file->file_name, io_u->offset, io_u->buflen,
-			(unsigned long) (p - io_u->buf));
-	return ret;
+			(unsigned long) offset);
+	return EILSEQ;
+}
+
+static int verify_header(struct io_u *io_u, struct thread_data *td,
+			 struct verify_header *hdr, unsigned int hdr_num,
+			 unsigned int hdr_len)
+{
+	void *p = hdr;
+	uint32_t crc;
+
+	if (hdr->magic != FIO_HDR_MAGIC) {
+		log_err("verify: bad magic header %x, wanted %x",
+			hdr->magic, FIO_HDR_MAGIC);
+		goto err;
+	}
+	if (hdr->len != hdr_len) {
+		log_err("verify: bad header length %u, wanted %u",
+			hdr->len, hdr_len);
+		goto err;
+	}
+	if (hdr->rand_seed != io_u->rand_seed) {
+		log_err("verify: bad header rand_seed %"PRIu64
+			", wanted %"PRIu64,
+			hdr->rand_seed, io_u->rand_seed);
+		goto err;
+	}
+	if (hdr->offset != io_u->offset + hdr_num * td->o.verify_interval) {
+		log_err("verify: bad header offset %"PRIu64
+			", wanted %llu",
+			hdr->offset, io_u->offset);
+		goto err;
+	}
+
+	/*
+	 * For read-only workloads, the program cannot be certain of the
+	 * last numberio written to a block. Checking of numberio will be
+	 * done only for workloads that write data.  For verify_only,
+	 * numberio will be checked in the last iteration when the correct
+	 * state of numberio, that would have been written to each block
+	 * in a previous run of fio, has been reached.
+	 */
+	if ((td_write(td) || td_rw(td)) && (td_min_bs(td) == td_max_bs(td)) &&
+	    !td->o.time_based)
+		if (!td->o.verify_only || td->o.loops == 0)
+			if (hdr->numberio != io_u->numberio) {
+				log_err("verify: bad header numberio %"PRIu16
+					", wanted %"PRIu16,
+					hdr->numberio, io_u->numberio);
+				goto err;
+			}
+
+	crc = fio_crc32c(p, offsetof(struct verify_header, crc32));
+	if (crc != hdr->crc32) {
+		log_err("verify: bad header crc %x, calculated %x",
+			hdr->crc32, crc);
+		goto err;
+	}
+	return 0;
+
+err:
+	log_err(" at file %s offset %llu, length %u\n",
+		io_u->file->file_name,
+		io_u->offset + hdr_num * hdr_len, hdr_len);
+	return EILSEQ;
 }
 
-int verify_io_u(struct thread_data *td, struct io_u *io_u)
+int verify_io_u(struct thread_data *td, struct io_u **io_u_ptr)
 {
 	struct verify_header *hdr;
-	unsigned int hdr_size, hdr_inc, hdr_num = 0;
+	struct io_u *io_u = *io_u_ptr;
+	unsigned int header_size, hdr_inc, hdr_num = 0;
 	void *p;
 	int ret;
 
 	if (td->o.verify == VERIFY_NULL || io_u->ddir != DDIR_READ)
 		return 0;
+	/*
+	 * If the IO engine is faking IO (like null), then just pretend
+	 * we verified everything.
+	 */
+	if (td->io_ops->flags & FIO_FAKEIO)
+		return 0;
+
 	if (io_u->flags & IO_U_F_TRIMMED) {
 		ret = verify_trimmed_io_u(td, io_u);
 		goto done;
 	}
 
-	hdr_inc = io_u->buflen;
-	if (td->o.verify_interval)
-		hdr_inc = td->o.verify_interval;
+	hdr_inc = get_hdr_inc(td, io_u);
 
 	ret = 0;
 	for (p = io_u->buf; p < io_u->buf + io_u->buflen;
@@ -529,49 +813,43 @@ int verify_io_u(struct thread_data *td, struct io_u *io_u)
 		struct vcont vc = {
 			.io_u		= io_u,
 			.hdr_num	= hdr_num,
+			.td		= td,
 		};
+		unsigned int verify_type;
 
 		if (ret && td->o.verify_fatal)
 			break;
 
-		hdr_size = __hdr_size(td->o.verify);
+		header_size = __hdr_size(td->o.verify);
 		if (td->o.verify_offset)
-			memswp(p, p + td->o.verify_offset, hdr_size);
+			memswp(p, p + td->o.verify_offset, header_size);
 		hdr = p;
 
-		if (hdr->fio_magic != FIO_HDR_MAGIC) {
-			log_err("verify: bad magic header %x, wanted %x at file %s offset %llu, length %u\n",
-				hdr->fio_magic, FIO_HDR_MAGIC,
-				io_u->file->file_name,
-				io_u->offset + hdr_num * hdr->len, hdr->len);
-			return EILSEQ;
+		/*
+		 * Make rand_seed check pass when have verifysort or
+		 * verify_backlog.
+		 */
+		if (td->o.verifysort || (td->flags & TD_F_VER_BACKLOG))
+			io_u->rand_seed = hdr->rand_seed;
+
+		if (td->o.verify != VERIFY_PATTERN_NO_HDR) {
+			ret = verify_header(io_u, td, hdr, hdr_num, hdr_inc);
+			if (ret)
+				return ret;
 		}
 
-		if (td->o.verify_pattern_bytes) {
-			dprint(FD_VERIFY, "pattern verify io_u %p, len %u\n",
-								io_u, hdr->len);
-			ret = verify_io_u_pattern(td->o.verify_pattern,
-				  td->o.verify_pattern_bytes,
-				  p + hdr_size,
-				  hdr_inc - hdr_size,
-				  hdr_size % td->o.verify_pattern_bytes);
-
-			if (ret) {
-				log_err("pattern: verify failed at file %s offset %llu, length %u\n",
-					io_u->file->file_name,
-					io_u->offset + hdr_num * hdr->len,
-					hdr->len);
-			}
-
-			/*
-			 * Also verify the meta data, if applicable
-			 */
-			if (hdr->verify_type == VERIFY_META)
-				ret |= verify_io_u_meta(hdr, td, &vc);
-			continue;
-		}
-
-		switch (hdr->verify_type) {
+		if (td->o.verify != VERIFY_NONE)
+			verify_type = td->o.verify;
+		else
+			verify_type = hdr->verify_type;
+
+		switch (verify_type) {
+		case VERIFY_HDR_ONLY:
+			/* Header is always verified, check if pattern is left
+			 * for verification. */
+			if (td->o.verify_pattern_bytes)
+				ret = verify_io_u_pattern(hdr, &vc);
+			break;
 		case VERIFY_MD5:
 			ret = verify_io_u_md5(hdr, &vc);
 			break;
@@ -597,210 +875,245 @@ int verify_io_u(struct thread_data *td, struct io_u *io_u)
 		case VERIFY_SHA512:
 			ret = verify_io_u_sha512(hdr, &vc);
 			break;
-		case VERIFY_META:
-			ret = verify_io_u_meta(hdr, td, &vc);
+		case VERIFY_XXHASH:
+			ret = verify_io_u_xxhash(hdr, &vc);
 			break;
 		case VERIFY_SHA1:
 			ret = verify_io_u_sha1(hdr, &vc);
 			break;
+		case VERIFY_PATTERN:
+		case VERIFY_PATTERN_NO_HDR:
+			ret = verify_io_u_pattern(hdr, &vc);
+			break;
 		default:
 			log_err("Bad verify type %u\n", hdr->verify_type);
 			ret = EINVAL;
 		}
+
+		if (ret && verify_type != hdr->verify_type)
+			log_err("fio: verify type mismatch (%u media, %u given)\n",
+					hdr->verify_type, verify_type);
 	}
 
 done:
 	if (ret && td->o.verify_fatal)
-		td->terminate = 1;
+		fio_mark_td_terminate(td);
 
 	return ret;
 }
 
-static void fill_meta(struct verify_header *hdr, struct thread_data *td,
-		      struct io_u *io_u, unsigned int header_num)
+static void fill_xxhash(struct verify_header *hdr, void *p, unsigned int len)
 {
-	struct vhdr_meta *vh = hdr_priv(hdr);
+	struct vhdr_xxhash *vh = hdr_priv(hdr);
+	void *state;
 
-	vh->thread = td->thread_number;
-
-	vh->time_sec = io_u->start_time.tv_sec;
-	vh->time_usec = io_u->start_time.tv_usec;
-
-	vh->numberio = td->io_issues[DDIR_WRITE];
-
-	vh->offset = io_u->offset + header_num * td->o.verify_interval;
+	state = XXH32_init(1);
+	XXH32_update(state, p, len);
+	vh->hash = XXH32_digest(state);
 }
 
 static void fill_sha512(struct verify_header *hdr, void *p, unsigned int len)
 {
 	struct vhdr_sha512 *vh = hdr_priv(hdr);
-	struct sha512_ctx sha512_ctx = {
+	struct fio_sha512_ctx sha512_ctx = {
 		.buf = vh->sha512,
 	};
 
-	sha512_init(&sha512_ctx);
-	sha512_update(&sha512_ctx, p, len);
+	fio_sha512_init(&sha512_ctx);
+	fio_sha512_update(&sha512_ctx, p, len);
 }
 
 static void fill_sha256(struct verify_header *hdr, void *p, unsigned int len)
 {
 	struct vhdr_sha256 *vh = hdr_priv(hdr);
-	struct sha256_ctx sha256_ctx = {
+	struct fio_sha256_ctx sha256_ctx = {
 		.buf = vh->sha256,
 	};
 
-	sha256_init(&sha256_ctx);
-	sha256_update(&sha256_ctx, p, len);
+	fio_sha256_init(&sha256_ctx);
+	fio_sha256_update(&sha256_ctx, p, len);
+	fio_sha256_final(&sha256_ctx);
 }
 
 static void fill_sha1(struct verify_header *hdr, void *p, unsigned int len)
 {
 	struct vhdr_sha1 *vh = hdr_priv(hdr);
-	struct sha1_ctx sha1_ctx = {
+	struct fio_sha1_ctx sha1_ctx = {
 		.H = vh->sha1,
 	};
 
-	sha1_init(&sha1_ctx);
-	sha1_update(&sha1_ctx, p, len);
+	fio_sha1_init(&sha1_ctx);
+	fio_sha1_update(&sha1_ctx, p, len);
+	fio_sha1_final(&sha1_ctx);
 }
 
 static void fill_crc7(struct verify_header *hdr, void *p, unsigned int len)
 {
 	struct vhdr_crc7 *vh = hdr_priv(hdr);
 
-	vh->crc7 = crc7(p, len);
+	vh->crc7 = fio_crc7(p, len);
 }
 
 static void fill_crc16(struct verify_header *hdr, void *p, unsigned int len)
 {
 	struct vhdr_crc16 *vh = hdr_priv(hdr);
 
-	vh->crc16 = crc16(p, len);
+	vh->crc16 = fio_crc16(p, len);
 }
 
 static void fill_crc32(struct verify_header *hdr, void *p, unsigned int len)
 {
 	struct vhdr_crc32 *vh = hdr_priv(hdr);
 
-	vh->crc32 = crc32(p, len);
+	vh->crc32 = fio_crc32(p, len);
 }
 
 static void fill_crc32c(struct verify_header *hdr, void *p, unsigned int len)
 {
 	struct vhdr_crc32 *vh = hdr_priv(hdr);
 
-	if (hdr->verify_type == VERIFY_CRC32C_INTEL)
-		vh->crc32 = crc32c_intel(p, len);
-	else
-		vh->crc32 = crc32c(p, len);
+	vh->crc32 = fio_crc32c(p, len);
 }
 
 static void fill_crc64(struct verify_header *hdr, void *p, unsigned int len)
 {
 	struct vhdr_crc64 *vh = hdr_priv(hdr);
 
-	vh->crc64 = crc64(p, len);
+	vh->crc64 = fio_crc64(p, len);
 }
 
 static void fill_md5(struct verify_header *hdr, void *p, unsigned int len)
 {
 	struct vhdr_md5 *vh = hdr_priv(hdr);
-	struct md5_ctx md5_ctx = {
+	struct fio_md5_ctx md5_ctx = {
 		.hash = (uint32_t *) vh->md5_digest,
 	};
 
-	md5_init(&md5_ctx);
-	md5_update(&md5_ctx, p, len);
+	fio_md5_init(&md5_ctx);
+	fio_md5_update(&md5_ctx, p, len);
+	fio_md5_final(&md5_ctx);
+}
+
+static void __fill_hdr(struct thread_data *td, struct io_u *io_u,
+		       struct verify_header *hdr, unsigned int header_num,
+		       unsigned int header_len, uint64_t rand_seed)
+{
+	void *p = hdr;
+
+	hdr->magic = FIO_HDR_MAGIC;
+	hdr->verify_type = td->o.verify;
+	hdr->len = header_len;
+	hdr->rand_seed = rand_seed;
+	hdr->offset = io_u->offset + header_num * td->o.verify_interval;
+	hdr->time_sec = io_u->start_time.tv_sec;
+	hdr->time_usec = io_u->start_time.tv_usec;
+	hdr->thread = td->thread_number;
+	hdr->numberio = io_u->numberio;
+	hdr->crc32 = fio_crc32c(p, offsetof(struct verify_header, crc32));
+}
+
+
+static void fill_hdr(struct thread_data *td, struct io_u *io_u,
+		     struct verify_header *hdr, unsigned int header_num,
+		     unsigned int header_len, uint64_t rand_seed)
+{
+
+	if (td->o.verify != VERIFY_PATTERN_NO_HDR)
+		__fill_hdr(td, io_u, hdr, header_num, header_len, rand_seed);
+}
+
+static void populate_hdr(struct thread_data *td, struct io_u *io_u,
+			 struct verify_header *hdr, unsigned int header_num,
+			 unsigned int header_len)
+{
+	unsigned int data_len;
+	void *data, *p;
+
+	p = (void *) hdr;
+
+	fill_hdr(td, io_u, hdr, header_num, header_len, io_u->rand_seed);
+
+	data_len = header_len - hdr_size(td, hdr);
+
+	data = p + hdr_size(td, hdr);
+	switch (td->o.verify) {
+	case VERIFY_MD5:
+		dprint(FD_VERIFY, "fill md5 io_u %p, len %u\n",
+						io_u, hdr->len);
+		fill_md5(hdr, data, data_len);
+		break;
+	case VERIFY_CRC64:
+		dprint(FD_VERIFY, "fill crc64 io_u %p, len %u\n",
+						io_u, hdr->len);
+		fill_crc64(hdr, data, data_len);
+		break;
+	case VERIFY_CRC32C:
+	case VERIFY_CRC32C_INTEL:
+		dprint(FD_VERIFY, "fill crc32c io_u %p, len %u\n",
+						io_u, hdr->len);
+		fill_crc32c(hdr, data, data_len);
+		break;
+	case VERIFY_CRC32:
+		dprint(FD_VERIFY, "fill crc32 io_u %p, len %u\n",
+						io_u, hdr->len);
+		fill_crc32(hdr, data, data_len);
+		break;
+	case VERIFY_CRC16:
+		dprint(FD_VERIFY, "fill crc16 io_u %p, len %u\n",
+						io_u, hdr->len);
+		fill_crc16(hdr, data, data_len);
+		break;
+	case VERIFY_CRC7:
+		dprint(FD_VERIFY, "fill crc7 io_u %p, len %u\n",
+						io_u, hdr->len);
+		fill_crc7(hdr, data, data_len);
+		break;
+	case VERIFY_SHA256:
+		dprint(FD_VERIFY, "fill sha256 io_u %p, len %u\n",
+						io_u, hdr->len);
+		fill_sha256(hdr, data, data_len);
+		break;
+	case VERIFY_SHA512:
+		dprint(FD_VERIFY, "fill sha512 io_u %p, len %u\n",
+						io_u, hdr->len);
+		fill_sha512(hdr, data, data_len);
+		break;
+	case VERIFY_XXHASH:
+		dprint(FD_VERIFY, "fill xxhash io_u %p, len %u\n",
+						io_u, hdr->len);
+		fill_xxhash(hdr, data, data_len);
+		break;
+	case VERIFY_SHA1:
+		dprint(FD_VERIFY, "fill sha1 io_u %p, len %u\n",
+						io_u, hdr->len);
+		fill_sha1(hdr, data, data_len);
+		break;
+	case VERIFY_HDR_ONLY:
+	case VERIFY_PATTERN:
+	case VERIFY_PATTERN_NO_HDR:
+		/* nothing to do here */
+		break;
+	default:
+		log_err("fio: bad verify type: %d\n", td->o.verify);
+		assert(0);
+	}
+
+	if (td->o.verify_offset && hdr_size(td, hdr))
+		memswp(p, p + td->o.verify_offset, hdr_size(td, hdr));
 }
 
 /*
  * fill body of io_u->buf with random data and add a header with the
- * crc32 or md5 sum of that data.
+ * checksum of choice
  */
 void populate_verify_io_u(struct thread_data *td, struct io_u *io_u)
 {
-	struct verify_header *hdr;
-	void *p = io_u->buf, *data;
-	unsigned int hdr_inc, data_len, header_num = 0;
-
 	if (td->o.verify == VERIFY_NULL)
 		return;
 
-	fill_pattern(td, p, io_u->buflen, io_u);
-
-	hdr_inc = io_u->buflen;
-	if (td->o.verify_interval)
-		hdr_inc = td->o.verify_interval;
+	io_u->numberio = td->io_issues[io_u->ddir];
 
-	for (; p < io_u->buf + io_u->buflen; p += hdr_inc) {
-		hdr = p;
-
-		hdr->fio_magic = FIO_HDR_MAGIC;
-		hdr->verify_type = td->o.verify;
-		hdr->len = hdr_inc;
-		data_len = hdr_inc - hdr_size(hdr);
-
-		data = p + hdr_size(hdr);
-		switch (td->o.verify) {
-		case VERIFY_MD5:
-			dprint(FD_VERIFY, "fill md5 io_u %p, len %u\n",
-							io_u, hdr->len);
-			fill_md5(hdr, data, data_len);
-			break;
-		case VERIFY_CRC64:
-			dprint(FD_VERIFY, "fill crc64 io_u %p, len %u\n",
-							io_u, hdr->len);
-			fill_crc64(hdr, data, data_len);
-			break;
-		case VERIFY_CRC32C:
-		case VERIFY_CRC32C_INTEL:
-			dprint(FD_VERIFY, "fill crc32c io_u %p, len %u\n",
-							io_u, hdr->len);
-			fill_crc32c(hdr, data, data_len);
-			break;
-		case VERIFY_CRC32:
-			dprint(FD_VERIFY, "fill crc32 io_u %p, len %u\n",
-							io_u, hdr->len);
-			fill_crc32(hdr, data, data_len);
-			break;
-		case VERIFY_CRC16:
-			dprint(FD_VERIFY, "fill crc16 io_u %p, len %u\n",
-							io_u, hdr->len);
-			fill_crc16(hdr, data, data_len);
-			break;
-		case VERIFY_CRC7:
-			dprint(FD_VERIFY, "fill crc7 io_u %p, len %u\n",
-							io_u, hdr->len);
-			fill_crc7(hdr, data, data_len);
-			break;
-		case VERIFY_SHA256:
-			dprint(FD_VERIFY, "fill sha256 io_u %p, len %u\n",
-							io_u, hdr->len);
-			fill_sha256(hdr, data, data_len);
-			break;
-		case VERIFY_SHA512:
-			dprint(FD_VERIFY, "fill sha512 io_u %p, len %u\n",
-							io_u, hdr->len);
-			fill_sha512(hdr, data, data_len);
-			break;
-		case VERIFY_META:
-			dprint(FD_VERIFY, "fill meta io_u %p, len %u\n",
-							io_u, hdr->len);
-			fill_meta(hdr, td, io_u, header_num);
-			break;
-		case VERIFY_SHA1:
-			dprint(FD_VERIFY, "fill sha1 io_u %p, len %u\n",
-							io_u, hdr->len);
-			fill_sha1(hdr, data, data_len);
-			break;
-		default:
-			log_err("fio: bad verify type: %d\n", td->o.verify);
-			assert(0);
-		}
-		if (td->o.verify_offset)
-			memswp(p, p + td->o.verify_offset, hdr_size(hdr));
-		header_num++;
-	}
+	fill_pattern_headers(td, io_u, 0, 0);
 }
 
 int get_next_verify(struct thread_data *td, struct io_u *io_u)
@@ -817,11 +1130,27 @@ int get_next_verify(struct thread_data *td, struct io_u *io_u)
 		struct rb_node *n = rb_first(&td->io_hist_tree);
 
 		ipo = rb_entry(n, struct io_piece, rb_node);
+
+		/*
+		 * Ensure that the associated IO has completed
+		 */
+		read_barrier();
+		if (ipo->flags & IP_F_IN_FLIGHT)
+			goto nothing;
+
 		rb_erase(n, &td->io_hist_tree);
 		assert(ipo->flags & IP_F_ONRB);
 		ipo->flags &= ~IP_F_ONRB;
 	} else if (!flist_empty(&td->io_hist_list)) {
-		ipo = flist_entry(td->io_hist_list.next, struct io_piece, list);
+		ipo = flist_first_entry(&td->io_hist_list, struct io_piece, list);
+
+		/*
+		 * Ensure that the associated IO has completed
+		 */
+		read_barrier();
+		if (ipo->flags & IP_F_IN_FLIGHT)
+			goto nothing;
+
 		flist_del(&ipo->list);
 		assert(ipo->flags & IP_F_ONLIST);
 		ipo->flags &= ~IP_F_ONLIST;
@@ -832,10 +1161,12 @@ int get_next_verify(struct thread_data *td, struct io_u *io_u)
 
 		io_u->offset = ipo->offset;
 		io_u->buflen = ipo->len;
+		io_u->numberio = ipo->numberio;
 		io_u->file = ipo->file;
+		io_u_set(io_u, IO_U_F_VER_LIST);
 
 		if (ipo->flags & IP_F_TRIMMED)
-			io_u->flags |= IO_U_F_TRIMMED;
+			io_u_set(io_u, IO_U_F_TRIMMED);
 
 		if (!fio_file_open(io_u->file)) {
 			int r = td_io_open_file(td, io_u->file);
@@ -856,20 +1187,35 @@ int get_next_verify(struct thread_data *td, struct io_u *io_u)
 		remove_trim_entry(td, ipo);
 		free(ipo);
 		dprint(FD_VERIFY, "get_next_verify: ret io_u %p\n", io_u);
+
+		if (!td->o.verify_pattern_bytes) {
+			io_u->rand_seed = __rand(&td->verify_state);
+			if (sizeof(int) != sizeof(long *))
+				io_u->rand_seed *= __rand(&td->verify_state);
+		}
 		return 0;
 	}
 
+nothing:
 	dprint(FD_VERIFY, "get_next_verify: empty\n");
 	return 1;
 }
 
+void fio_verify_init(struct thread_data *td)
+{
+	if (td->o.verify == VERIFY_CRC32C_INTEL ||
+	    td->o.verify == VERIFY_CRC32C) {
+		crc32c_intel_probe();
+	}
+}
+
 static void *verify_async_thread(void *data)
 {
 	struct thread_data *td = data;
 	struct io_u *io_u;
 	int ret = 0;
 
-	if (td->o.verify_cpumask_set &&
+	if (fio_option_is_set(&td->o, verify_cpumask) &&
 	    fio_setaffinity(td->pid, td->o.verify_cpumask)) {
 		log_err("fio: failed setting verify thread affinity\n");
 		goto done;
@@ -901,15 +1247,16 @@ static void *verify_async_thread(void *data)
 			continue;
 
 		while (!flist_empty(&list)) {
-			io_u = flist_entry(list.next, struct io_u, list);
-			flist_del_init(&io_u->list);
+			io_u = flist_first_entry(&list, struct io_u, verify_list);
+			flist_del_init(&io_u->verify_list);
+
+			io_u_set(io_u, IO_U_F_NO_FILE_PUT);
+			ret = verify_io_u(td, &io_u);
 
-			ret = verify_io_u(td, io_u);
 			put_io_u(td, io_u);
 			if (!ret)
 				continue;
-			if (td->o.continue_on_error &&
-			    td_non_fatal_error(ret)) {
+			if (td_non_fatal_error(td, ERROR_TYPE_VERIFY_BIT, ret)) {
 				update_error_count(td, ret);
 				td_clear_error(td);
 				ret = 0;
@@ -920,7 +1267,7 @@ static void *verify_async_thread(void *data)
 	if (ret) {
 		td_verror(td, ret, "async_verify");
 		if (td->o.verify_fatal)
-			td->terminate = 1;
+			fio_mark_td_terminate(td);
 	}
 
 done:
@@ -988,3 +1335,376 @@ void verify_async_exit(struct thread_data *td)
 	free(td->verify_threads);
 	td->verify_threads = NULL;
 }
+
+int paste_blockoff(char *buf, unsigned int len, void *priv)
+{
+	struct io_u *io = priv;
+	unsigned long long off;
+
+	typecheck(typeof(off), io->offset);
+	off = cpu_to_le64((uint64_t)io->offset);
+	len = min(len, (unsigned int)sizeof(off));
+	memcpy(buf, &off, len);
+	return 0;
+}
+
+struct all_io_list *get_all_io_list(int save_mask, size_t *sz)
+{
+	struct all_io_list *rep;
+	struct thread_data *td;
+	size_t depth;
+	void *next;
+	int i, nr;
+
+	compiletime_assert(sizeof(struct all_io_list) == 8, "all_io_list");
+
+	/*
+	 * Calculate reply space needed. We need one 'io_state' per thread,
+	 * and the size will vary depending on depth.
+	 */
+	depth = 0;
+	nr = 0;
+	for_each_td(td, i) {
+		if (save_mask != IO_LIST_ALL && (i + 1) != save_mask)
+			continue;
+		td->stop_io = 1;
+		td->flags |= TD_F_VSTATE_SAVED;
+		depth += td->o.iodepth;
+		nr++;
+	}
+
+	if (!nr)
+		return NULL;
+
+	*sz = sizeof(*rep);
+	*sz += nr * sizeof(struct thread_io_list);
+	*sz += depth * sizeof(uint64_t);
+	rep = malloc(*sz);
+
+	rep->threads = cpu_to_le64((uint64_t) nr);
+
+	next = &rep->state[0];
+	for_each_td(td, i) {
+		struct thread_io_list *s = next;
+		unsigned int comps;
+
+		if (save_mask != IO_LIST_ALL && (i + 1) != save_mask)
+			continue;
+
+		if (td->last_write_comp) {
+			int j, k;
+
+			if (td->io_blocks[DDIR_WRITE] < td->o.iodepth)
+				comps = td->io_blocks[DDIR_WRITE];
+			else
+				comps = td->o.iodepth;
+
+			k = td->last_write_idx - 1;
+			for (j = 0; j < comps; j++) {
+				if (k == -1)
+					k = td->o.iodepth - 1;
+				s->offsets[j] = cpu_to_le64(td->last_write_comp[k]);
+				k--;
+			}
+		} else
+			comps = 0;
+
+		s->no_comps = cpu_to_le64((uint64_t) comps);
+		s->depth = cpu_to_le64((uint64_t) td->o.iodepth);
+		s->numberio = cpu_to_le64((uint64_t) td->io_issues[DDIR_WRITE]);
+		s->index = cpu_to_le64((uint64_t) i);
+		if (td->random_state.use64) {
+			s->rand.state64.s[0] = cpu_to_le64(td->random_state.state64.s1);
+			s->rand.state64.s[1] = cpu_to_le64(td->random_state.state64.s2);
+			s->rand.state64.s[2] = cpu_to_le64(td->random_state.state64.s3);
+			s->rand.state64.s[3] = cpu_to_le64(td->random_state.state64.s4);
+			s->rand.state64.s[4] = cpu_to_le64(td->random_state.state64.s5);
+			s->rand.state64.s[5] = 0;
+			s->rand.use64 = cpu_to_le64((uint64_t)1);
+		} else {
+			s->rand.state32.s[0] = cpu_to_le32(td->random_state.state32.s1);
+			s->rand.state32.s[1] = cpu_to_le32(td->random_state.state32.s2);
+			s->rand.state32.s[2] = cpu_to_le32(td->random_state.state32.s3);
+			s->rand.state32.s[3] = 0;
+			s->rand.use64 = 0;
+		}
+		s->name[sizeof(s->name) - 1] = '\0';
+		strncpy((char *) s->name, td->o.name, sizeof(s->name) - 1);
+		next = io_list_next(s);
+	}
+
+	return rep;
+}
+
+static int open_state_file(const char *name, const char *prefix, int num,
+			   int for_write)
+{
+	char out[64];
+	int flags;
+	int fd;
+
+	if (for_write)
+		flags = O_CREAT | O_TRUNC | O_WRONLY | O_SYNC;
+	else
+		flags = O_RDONLY;
+
+	verify_state_gen_name(out, sizeof(out), name, prefix, num);
+
+	fd = open(out, flags, 0644);
+	if (fd == -1) {
+		perror("fio: open state file");
+		return -1;
+	}
+
+	return fd;
+}
+
+static int write_thread_list_state(struct thread_io_list *s,
+				   const char *prefix)
+{
+	struct verify_state_hdr hdr;
+	uint64_t crc;
+	ssize_t ret;
+	int fd;
+
+	fd = open_state_file((const char *) s->name, prefix, s->index, 1);
+	if (fd == -1)
+		return 1;
+
+	crc = fio_crc32c((void *)s, thread_io_list_sz(s));
+
+	hdr.version = cpu_to_le64((uint64_t) VSTATE_HDR_VERSION);
+	hdr.size = cpu_to_le64((uint64_t) thread_io_list_sz(s));
+	hdr.crc = cpu_to_le64(crc);
+	ret = write(fd, &hdr, sizeof(hdr));
+	if (ret != sizeof(hdr))
+		goto write_fail;
+
+	ret = write(fd, s, thread_io_list_sz(s));
+	if (ret != thread_io_list_sz(s)) {
+write_fail:
+		if (ret < 0)
+			perror("fio: write state file");
+		log_err("fio: failed to write state file\n");
+		ret = 1;
+	} else
+		ret = 0;
+
+	close(fd);
+	return ret;
+}
+
+void __verify_save_state(struct all_io_list *state, const char *prefix)
+{
+	struct thread_io_list *s = &state->state[0];
+	unsigned int i;
+
+	for (i = 0; i < le64_to_cpu(state->threads); i++) {
+		write_thread_list_state(s,  prefix);
+		s = io_list_next(s);
+	}
+}
+
+void verify_save_state(int mask)
+{
+	struct all_io_list *state;
+	size_t sz;
+
+	state = get_all_io_list(mask, &sz);
+	if (state) {
+		char prefix[PATH_MAX];
+
+		if (aux_path)
+			sprintf(prefix, "%s%slocal", aux_path, FIO_OS_PATH_SEPARATOR);
+		else
+			strcpy(prefix, "local");
+
+		__verify_save_state(state, prefix);
+		free(state);
+	}
+}
+
+void verify_free_state(struct thread_data *td)
+{
+	if (td->vstate)
+		free(td->vstate);
+}
+
+static struct thread_io_list *convert_v1_list(struct thread_io_list_v1 *s)
+{
+	struct thread_io_list *til;
+	int i;
+
+	til = malloc(__thread_io_list_sz(s->no_comps));
+	til->no_comps = s->no_comps;
+	til->depth = s->depth;
+	til->numberio = s->numberio;
+	til->index = s->index;
+	memcpy(til->name, s->name, sizeof(til->name));
+
+	til->rand.use64 = 0;
+	for (i = 0; i < 4; i++)
+		til->rand.state32.s[i] = s->rand.s[i];
+
+	for (i = 0; i < s->no_comps; i++)
+		til->offsets[i] = s->offsets[i];
+
+	return til;
+}
+
+void verify_convert_assign_state(struct thread_data *td, void *p, int version)
+{
+	struct thread_io_list *til;
+	int i;
+
+	if (version == 1) {
+		struct thread_io_list_v1 *s = p;
+
+		s->no_comps = le64_to_cpu(s->no_comps);
+		s->depth = le64_to_cpu(s->depth);
+		s->numberio = le64_to_cpu(s->numberio);
+		for (i = 0; i < 4; i++)
+			s->rand.s[i] = le32_to_cpu(s->rand.s[i]);
+		for (i = 0; i < s->no_comps; i++)
+			s->offsets[i] = le64_to_cpu(s->offsets[i]);
+
+		til = convert_v1_list(s);
+		free(s);
+	} else {
+		struct thread_io_list *s = p;
+
+		s->no_comps = le64_to_cpu(s->no_comps);
+		s->depth = le64_to_cpu(s->depth);
+		s->numberio = le64_to_cpu(s->numberio);
+		s->rand.use64 = le64_to_cpu(s->rand.use64);
+
+		if (s->rand.use64) {
+			for (i = 0; i < 6; i++)
+				s->rand.state64.s[i] = le64_to_cpu(s->rand.state64.s[i]);
+		} else {
+			for (i = 0; i < 4; i++)
+				s->rand.state32.s[i] = le32_to_cpu(s->rand.state32.s[i]);
+		}
+		for (i = 0; i < s->no_comps; i++)
+			s->offsets[i] = le64_to_cpu(s->offsets[i]);
+
+		til = p;
+	}
+
+	td->vstate = til;
+}
+
+int verify_state_hdr(struct verify_state_hdr *hdr, struct thread_io_list *s,
+		     int *version)
+{
+	uint64_t crc;
+
+	hdr->version = le64_to_cpu(hdr->version);
+	hdr->size = le64_to_cpu(hdr->size);
+	hdr->crc = le64_to_cpu(hdr->crc);
+
+	if (hdr->version != VSTATE_HDR_VERSION ||
+	    hdr->version != VSTATE_HDR_VERSION_V1)
+		return 1;
+
+	crc = fio_crc32c((void *)s, hdr->size);
+	if (crc != hdr->crc)
+		return 1;
+
+	*version = hdr->version;
+	return 0;
+}
+
+int verify_load_state(struct thread_data *td, const char *prefix)
+{
+	struct verify_state_hdr hdr;
+	void *s = NULL;
+	uint64_t crc;
+	ssize_t ret;
+	int fd;
+
+	if (!td->o.verify_state)
+		return 0;
+
+	fd = open_state_file(td->o.name, prefix, td->thread_number - 1, 0);
+	if (fd == -1)
+		return 1;
+
+	ret = read(fd, &hdr, sizeof(hdr));
+	if (ret != sizeof(hdr)) {
+		if (ret < 0)
+			td_verror(td, errno, "read verify state hdr");
+		log_err("fio: failed reading verify state header\n");
+		goto err;
+	}
+
+	hdr.version = le64_to_cpu(hdr.version);
+	hdr.size = le64_to_cpu(hdr.size);
+	hdr.crc = le64_to_cpu(hdr.crc);
+
+	if (hdr.version != VSTATE_HDR_VERSION &&
+	    hdr.version != VSTATE_HDR_VERSION_V1) {
+		log_err("fio: bad version in verify state header\n");
+		goto err;
+	}
+
+	s = malloc(hdr.size);
+	ret = read(fd, s, hdr.size);
+	if (ret != hdr.size) {
+		if (ret < 0)
+			td_verror(td, errno, "read verify state");
+		log_err("fio: failed reading verity state\n");
+		goto err;
+	}
+
+	crc = fio_crc32c(s, hdr.size);
+	if (crc != hdr.crc) {
+		log_err("fio: verify state is corrupt\n");
+		goto err;
+	}
+
+	close(fd);
+
+	verify_convert_assign_state(td, s, hdr.version);
+	return 0;
+err:
+	if (s)
+		free(s);
+	close(fd);
+	return 1;
+}
+
+/*
+ * Use the loaded verify state to know when to stop doing verification
+ */
+int verify_state_should_stop(struct thread_data *td, struct io_u *io_u)
+{
+	struct thread_io_list *s = td->vstate;
+	int i;
+
+	if (!s)
+		return 0;
+
+	/*
+	 * If we're not into the window of issues - depth yet, continue. If
+	 * issue is shorter than depth, do check.
+	 */
+	if ((td->io_blocks[DDIR_READ] < s->depth ||
+	    s->numberio - td->io_blocks[DDIR_READ] > s->depth) &&
+	    s->numberio > s->depth)
+		return 0;
+
+	/*
+	 * We're in the window of having to check if this io was
+	 * completed or not. If the IO was seen as completed, then
+	 * lets verify it.
+	 */
+	for (i = 0; i < s->no_comps; i++)
+		if (io_u->offset == s->offsets[i])
+			return 0;
+
+	/*
+	 * Not found, we have to stop
+	 */
+	return 1;
+}