X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=io_u.c;h=f020cac38af38e5d2ebb150c059564f023b14b18;hp=d6816068063e9a4be50998fd21c6fcd64d1ae49e;hb=9dbc7bfe8aab6def3e2b51edd57942371e4e5366;hpb=51ede0b1e9c9b570b942b50b44d0455183a0d5ec

diff --git a/io_u.c b/io_u.c
index d6816068..f020cac3 100644
--- a/io_u.c
+++ b/io_u.c
@@ -10,7 +10,7 @@
 #include "verify.h"
 #include "trim.h"
 #include "lib/rand.h"
-#include "lib/bitmap.h"
+#include "lib/axmap.h"
 
 struct io_completion_data {
 	int nr;				/* input */
@@ -21,12 +21,12 @@ struct io_completion_data {
 };
 
 /*
- * The ->io_bitmap contains a map of blocks we have or have not done io
+ * The ->io_axmap contains a map of blocks we have or have not done io
  * to yet. Used to make sure we cover the entire range in a fair fashion.
  */
-static int random_map_free(struct fio_file *f, const unsigned long long block)
+static int random_map_free(struct fio_file *f, const uint64_t block)
 {
-	return !bitmap_isset(f->io_bitmap, block);
+	return !axmap_isset(f->io_axmap, block);
 }
 
 /*
@@ -36,23 +36,24 @@ static void mark_random_map(struct thread_data *td, struct io_u *io_u)
 {
 	unsigned int min_bs = td->o.rw_min_bs;
 	struct fio_file *f = io_u->file;
-	unsigned long long block;
 	unsigned int nr_blocks;
+	uint64_t block;
 
-	block = (io_u->offset - f->file_offset) / (unsigned long long) min_bs;
+	block = (io_u->offset - f->file_offset) / (uint64_t) min_bs;
 	nr_blocks = (io_u->buflen + min_bs - 1) / min_bs;
 
-	nr_blocks = bitmap_set_nr(f->io_bitmap, block, nr_blocks);
+	if (!(io_u->flags & IO_U_F_BUSY_OK))
+		nr_blocks = axmap_set_nr(f->io_axmap, block, nr_blocks);
 
 	if ((nr_blocks * min_bs) < io_u->buflen)
 		io_u->buflen = nr_blocks * min_bs;
 }
 
-static unsigned long long last_block(struct thread_data *td, struct fio_file *f,
-				     enum fio_ddir ddir)
+static uint64_t last_block(struct thread_data *td, struct fio_file *f,
+			   enum fio_ddir ddir)
 {
-	unsigned long long max_blocks;
-	unsigned long long max_size;
+	uint64_t max_blocks;
+	uint64_t max_size;
 
 	assert(ddir_rw(ddir));
 
@@ -66,35 +67,51 @@ static unsigned long long last_block(struct thread_data *td, struct fio_file *f,
 	if (td->o.zone_range)
 		max_size = td->o.zone_range;
 
-	max_blocks = max_size / (unsigned long long) td->o.ba[ddir];
+	max_blocks = max_size / (uint64_t) td->o.ba[ddir];
 	if (!max_blocks)
 		return 0;
 
 	return max_blocks;
 }
 
+struct rand_off {
+	struct flist_head list;
+	uint64_t off;
+};
+
 static int __get_next_rand_offset(struct thread_data *td, struct fio_file *f,
-				  enum fio_ddir ddir, unsigned long long *b)
+				  enum fio_ddir ddir, uint64_t *b)
 {
-	unsigned long long rmax, r, lastb;
+	uint64_t r, lastb;
 
 	lastb = last_block(td, f, ddir);
 	if (!lastb)
 		return 1;
 
-	rmax = td->o.use_os_rand ? OS_RAND_MAX : FRAND_MAX;
+	if (td->o.random_generator == FIO_RAND_GEN_TAUSWORTHE) {
+		uint64_t rmax;
 
-	if (td->o.use_os_rand) {
-		rmax = OS_RAND_MAX;
-		r = os_random_long(&td->random_state);
+		rmax = td->o.use_os_rand ? OS_RAND_MAX : FRAND_MAX;
+
+		if (td->o.use_os_rand) {
+			rmax = OS_RAND_MAX;
+			r = os_random_long(&td->random_state);
+		} else {
+			rmax = FRAND_MAX;
+			r = __rand(&td->__random_state);
+		}
+
+		dprint(FD_RANDOM, "off rand %llu\n", r);
+
+		*b = (lastb - 1) * (r / ((uint64_t) rmax + 1.0));
 	} else {
-		rmax = FRAND_MAX;
-		r = __rand(&td->__random_state);
-	}
+		uint64_t off = 0;
 
-	*b = (lastb - 1) * (r / ((unsigned long long) rmax + 1.0));
+		if (lfsr_next(&f->lfsr, &off, lastb))
+			return 1;
 
-	dprint(FD_RANDOM, "off rand %llu\n", r);
+		*b = off;
+	}
 
 	/*
 	 * if we are not maintaining a random map, we are done.
@@ -110,7 +127,7 @@ static int __get_next_rand_offset(struct thread_data *td, struct fio_file *f,
 
 	dprint(FD_RANDOM, "get_next_rand_offset: offset %llu busy\n", *b);
 
-	*b = bitmap_next_free(f->io_bitmap, *b);
+	*b = axmap_next_free(f->io_axmap, *b);
 	if (*b == (uint64_t) -1ULL)
 		return 1;
 ret:
@@ -119,7 +136,7 @@ ret:
 
 static int __get_next_rand_offset_zipf(struct thread_data *td,
 				       struct fio_file *f, enum fio_ddir ddir,
-				       unsigned long long *b)
+				       uint64_t *b)
 {
 	*b = zipf_next(&f->zipf);
 	return 0;
@@ -127,14 +144,22 @@ static int __get_next_rand_offset_zipf(struct thread_data *td,
 
 static int __get_next_rand_offset_pareto(struct thread_data *td,
 					 struct fio_file *f, enum fio_ddir ddir,
-					 unsigned long long *b)
+					 uint64_t *b)
 {
 	*b = pareto_next(&f->zipf);
 	return 0;
 }
 
-static int get_next_rand_offset(struct thread_data *td, struct fio_file *f,
-				enum fio_ddir ddir, unsigned long long *b)
+static int flist_cmp(void *data, struct flist_head *a, struct flist_head *b)
+{
+	struct rand_off *r1 = flist_entry(a, struct rand_off, list);
+	struct rand_off *r2 = flist_entry(b, struct rand_off, list);
+
+	return r1->off - r2->off;
+}
+
+static int get_off_from_method(struct thread_data *td, struct fio_file *f,
+			       enum fio_ddir ddir, uint64_t *b)
 {
 	if (td->o.random_distribution == FIO_RAND_DIST_RANDOM)
 		return __get_next_rand_offset(td, f, ddir, b);
@@ -147,14 +172,59 @@ static int get_next_rand_offset(struct thread_data *td, struct fio_file *f,
 	return 1;
 }
 
+static int get_next_rand_offset(struct thread_data *td, struct fio_file *f,
+				enum fio_ddir ddir, uint64_t *b)
+{
+	struct rand_off *r;
+	int i, ret = 1;
+
+	/*
+	 * If sort not enabled, or not a pure random read workload without
+	 * any stored write metadata, just return a random offset
+	 */
+	if (!td->o.verifysort_nr || !(ddir == DDIR_READ && td->o.do_verify &&
+	    td->o.verify != VERIFY_NONE && td_random(td)) ||
+	    td->o.random_generator == FIO_RAND_GEN_TAUSWORTHE)
+		return get_off_from_method(td, f, ddir, b);
+
+	if (!flist_empty(&td->next_rand_list)) {
+		struct rand_off *r;
+fetch:
+		r = flist_entry(td->next_rand_list.next, struct rand_off, list);
+		flist_del(&r->list);
+		*b = r->off;
+		free(r);
+		return 0;
+	}
+
+	for (i = 0; i < td->o.verifysort_nr; i++) {
+		r = malloc(sizeof(*r));
+
+		ret = get_off_from_method(td, f, ddir, &r->off);
+		if (ret) {
+			free(r);
+			break;
+		}
+
+		flist_add(&r->list, &td->next_rand_list);
+	}
+
+	if (ret && !i)
+		return ret;
+
+	assert(!flist_empty(&td->next_rand_list));
+	flist_sort(NULL, &td->next_rand_list, flist_cmp);
+	goto fetch;
+}
+
 static int get_next_rand_block(struct thread_data *td, struct fio_file *f,
-			       enum fio_ddir ddir, unsigned long long *b)
+			       enum fio_ddir ddir, uint64_t *b)
 {
 	if (!get_next_rand_offset(td, f, ddir, b))
 		return 0;
 
 	if (td->o.time_based) {
-		fio_file_reset(f);
+		fio_file_reset(td, f);
 		if (!get_next_rand_offset(td, f, ddir, b))
 			return 0;
 	}
@@ -165,7 +235,7 @@ static int get_next_rand_block(struct thread_data *td, struct fio_file *f,
 }
 
 static int get_next_seq_offset(struct thread_data *td, struct fio_file *f,
-			       enum fio_ddir ddir, unsigned long long *offset)
+			       enum fio_ddir ddir, uint64_t *offset)
 {
 	assert(ddir_rw(ddir));
 
@@ -173,7 +243,7 @@ static int get_next_seq_offset(struct thread_data *td, struct fio_file *f,
 		f->last_pos = f->last_pos - f->io_size;
 
 	if (f->last_pos < f->real_file_size) {
-		unsigned long long pos;
+		uint64_t pos;
 
 		if (f->last_pos == f->file_offset && td->o.ddir_seq_add < 0)
 			f->last_pos = f->real_file_size;
@@ -193,7 +263,7 @@ static int get_next_block(struct thread_data *td, struct io_u *io_u,
 			  enum fio_ddir ddir, int rw_seq)
 {
 	struct fio_file *f = io_u->file;
-	unsigned long long b, offset;
+	uint64_t b, offset;
 	int ret;
 
 	assert(ddir_rw(ddir));
@@ -223,7 +293,7 @@ static int get_next_block(struct thread_data *td, struct io_u *io_u,
 			ret = 1;
 		}
 	}
-	
+
 	if (!ret) {
 		if (offset != -1ULL)
 			io_u->offset = offset;
@@ -278,10 +348,12 @@ static int __get_next_offset(struct thread_data *td, struct io_u *io_u)
 
 static int get_next_offset(struct thread_data *td, struct io_u *io_u)
 {
-	struct prof_io_ops *ops = &td->prof_io_ops;
+	if (td->flags & TD_F_PROFILE_OPS) {
+		struct prof_io_ops *ops = &td->prof_io_ops;
 
-	if (ops->fill_io_u_off)
-		return ops->fill_io_u_off(td, io_u);
+		if (ops->fill_io_u_off)
+			return ops->fill_io_u_off(td, io_u);
+	}
 
 	return __get_next_offset(td, io_u);
 }
@@ -356,10 +428,12 @@ static unsigned int __get_next_buflen(struct thread_data *td, struct io_u *io_u)
 
 static unsigned int get_next_buflen(struct thread_data *td, struct io_u *io_u)
 {
-	struct prof_io_ops *ops = &td->prof_io_ops;
+	if (td->flags & TD_F_PROFILE_OPS) {
+		struct prof_io_ops *ops = &td->prof_io_ops;
 
-	if (ops->fill_io_u_size)
-		return ops->fill_io_u_size(td, io_u);
+		if (ops->fill_io_u_size)
+			return ops->fill_io_u_size(td, io_u);
+	}
 
 	return __get_next_buflen(td, io_u);
 }
@@ -471,6 +545,12 @@ static enum fio_ddir get_rw_ddir(struct thread_data *td)
 {
 	enum fio_ddir ddir;
 
+	/*
+	 * If verify phase started, it's always a READ
+	 */
+	if (td->runstate == TD_VERIFYING)
+		return DDIR_READ;
+
 	/*
 	 * see if it's time to fsync
 	 */
@@ -948,10 +1028,12 @@ out:
 
 static struct fio_file *get_next_file(struct thread_data *td)
 {
-	struct prof_io_ops *ops = &td->prof_io_ops;
+	if (!(td->flags & TD_F_PROFILE_OPS)) {
+		struct prof_io_ops *ops = &td->prof_io_ops;
 
-	if (ops->get_next_file)
-		return ops->get_next_file(td);
+		if (ops->get_next_file)
+			return ops->get_next_file(td);
+	}
 
 	return __get_next_file(td);
 }
@@ -1028,7 +1110,10 @@ again:
 
 static int check_get_trim(struct thread_data *td, struct io_u *io_u)
 {
-	if (td->o.trim_backlog && td->trim_entries) {
+	if (!(td->flags & TD_F_TRIM_BACKLOG))
+		return 0;
+
+	if (td->trim_entries) {
 		int get_trim = 0;
 
 		if (td->trim_batch) {
@@ -1051,7 +1136,10 @@ static int check_get_trim(struct thread_data *td, struct io_u *io_u)
 
 static int check_get_verify(struct thread_data *td, struct io_u *io_u)
 {
-	if (td->o.verify_backlog && td->io_hist_len) {
+	if (!(td->flags & TD_F_VER_BACKLOG))
+		return 0;
+
+	if (td->io_hist_len) {
 		int get_verify = 0;
 
 		if (td->verify_batch)
@@ -1082,7 +1170,7 @@ static int check_get_verify(struct thread_data *td, struct io_u *io_u)
 static void small_content_scramble(struct io_u *io_u)
 {
 	unsigned int i, nr_blocks = io_u->buflen / 512;
-	unsigned long long boffset;
+	uint64_t boffset;
 	unsigned int offset;
 	void *p, *end;
 
@@ -1100,9 +1188,9 @@ static void small_content_scramble(struct io_u *io_u)
 		 * and the actual offset.
 		 */
 		offset = (io_u->start_time.tv_usec ^ boffset) & 511;
-		offset &= ~(sizeof(unsigned long long) - 1);
-		if (offset >= 512 - sizeof(unsigned long long))
-			offset -= sizeof(unsigned long long);
+		offset &= ~(sizeof(uint64_t) - 1);
+		if (offset >= 512 - sizeof(uint64_t))
+			offset -= sizeof(uint64_t);
 		memcpy(p + offset, &boffset, sizeof(boffset));
 
 		end = p + 512 - sizeof(io_u->start_time);
@@ -1142,7 +1230,7 @@ struct io_u *get_io_u(struct thread_data *td)
 	/*
 	 * If using an iolog, grab next piece if any available.
 	 */
-	if (td->o.read_iolog_file) {
+	if (td->flags & TD_F_READ_IOLOG) {
 		if (read_iolog_get(td, io_u))
 			goto err_put;
 	} else if (set_io_u_file(td, io_u)) {
@@ -1163,12 +1251,12 @@ struct io_u *get_io_u(struct thread_data *td)
 		f->last_pos = io_u->offset + io_u->buflen;
 
 		if (io_u->ddir == DDIR_WRITE) {
-			if (td->o.refill_buffers) {
+			if (td->flags & TD_F_REFILL_BUFFERS) {
 				io_u_fill_buffer(td, io_u,
 					io_u->xfer_buflen, io_u->xfer_buflen);
-			} else if (td->o.scramble_buffers)
+			} else if (td->flags & TD_F_SCRAMBLE_BUFFERS)
 				do_scramble = 1;
-			if (td->o.verify != VERIFY_NONE) {
+			if (td->flags & TD_F_VER_NONE) {
 				populate_verify_io_u(td, io_u);
 				do_scramble = 0;
 			}
@@ -1261,7 +1349,8 @@ static void account_io_completion(struct thread_data *td, struct io_u *io_u,
 
 static long long usec_for_io(struct thread_data *td, enum fio_ddir ddir)
 {
-	unsigned long long secs, remainder, bps, bytes;
+	uint64_t secs, remainder, bps, bytes;
+
 	bytes = td->this_io_bytes[ddir];
 	bps = td->rate_bps[ddir];
 	secs = bytes / bps;
@@ -1336,7 +1425,8 @@ static void io_completed(struct thread_data *td, struct io_u *io_u,
 
 		if (td_write(td) && idx == DDIR_WRITE &&
 		    td->o.do_verify &&
-		    td->o.verify != VERIFY_NONE)
+		    td->o.verify != VERIFY_NONE &&
+		    !td->o.experimental_verify)
 			log_io_piece(td, io_u);
 
 		icd->bytes_done[idx] += bytes;