[fio.git] / iolog.c

/*
 * Code related to writing an iolog of what a thread is doing, and to
 * later read that back and replay
 */
#include <stdio.h>
#include <stdlib.h>
#include <libgen.h>
#include <assert.h>
#include "flist.h"
#include "fio.h"
#include "verify.h"
#include "trim.h"
#include "filelock.h"

static const char iolog_ver2[] = "fio version 2 iolog";

void queue_io_piece(struct thread_data *td, struct io_piece *ipo)
{
	flist_add_tail(&ipo->list, &td->io_log_list);
	td->total_io_size += ipo->len;
}

void log_io_u(struct thread_data *td, struct io_u *io_u)
{
	const char *act[] = { "read", "write", "sync", "datasync",
				"sync_file_range", "wait", "trim" };

	assert(io_u->ddir <= 6);

	if (!td->o.write_iolog_file)
		return;

	fprintf(td->iolog_f, "%s %s %llu %lu\n", io_u->file->file_name,
						act[io_u->ddir], io_u->offset,
						io_u->buflen);
}

void log_file(struct thread_data *td, struct fio_file *f,
	      enum file_log_act what)
{
	const char *act[] = { "add", "open", "close" };

	assert(what < 3);

	if (!td->o.write_iolog_file)
		return;


	/*
	 * this happens on the pre-open/close done before the job starts
	 */
	if (!td->iolog_f)
		return;

	fprintf(td->iolog_f, "%s %s\n", f->file_name, act[what]);
}

static void iolog_delay(struct thread_data *td, unsigned long delay)
{
	unsigned long usec = utime_since_now(&td->last_issue);

	if (delay < usec)
		return;

	delay -= usec;

	/*
	 * less than 100 usec delay, just regard it as noise
	 */
	if (delay < 100)
		return;

	usec_sleep(td, delay);
}

static int ipo_special(struct thread_data *td, struct io_piece *ipo)
{
	struct fio_file *f;
	int ret;

	/*
	 * Not a special ipo
	 */
	if (ipo->ddir != DDIR_INVAL)
		return 0;

	f = td->files[ipo->fileno];

	switch (ipo->file_action) {
	case FIO_LOG_OPEN_FILE:
		ret = td_io_open_file(td, f);
		if (!ret)
			break;
		td_verror(td, ret, "iolog open file");
		return -1;
	case FIO_LOG_CLOSE_FILE:
		td_io_close_file(td, f);
		break;
	case FIO_LOG_UNLINK_FILE:
		unlink(f->file_name);
		break;
	default:
		log_err("fio: bad file action %d\n", ipo->file_action);
		break;
	}

	return 1;
}

int read_iolog_get(struct thread_data *td, struct io_u *io_u)
{
	struct io_piece *ipo;
	unsigned long elapsed;

	while (!flist_empty(&td->io_log_list)) {
		int ret;

		ipo = flist_entry(td->io_log_list.next, struct io_piece, list);
		flist_del(&ipo->list);
		remove_trim_entry(td, ipo);

		ret = ipo_special(td, ipo);
		if (ret < 0) {
			free(ipo);
			break;
		} else if (ret > 0) {
			free(ipo);
			continue;
		}

		io_u->ddir = ipo->ddir;
		if (ipo->ddir != DDIR_WAIT) {
			io_u->offset = ipo->offset;
			io_u->buflen = ipo->len;
			io_u->file = td->files[ipo->fileno];
			get_file(io_u->file);
			dprint(FD_IO, "iolog: get %llu/%lu/%s\n", io_u->offset,
						io_u->buflen, io_u->file->file_name);
			if (ipo->delay)
				iolog_delay(td, ipo->delay);
		} else {
			elapsed = mtime_since_genesis();
			if (ipo->delay > elapsed)
				usec_sleep(td, (ipo->delay - elapsed) * 1000);
		}

		free(ipo);

		if (io_u->ddir != DDIR_WAIT)
			return 0;
	}

	td->done = 1;
	return 1;
}

void prune_io_piece_log(struct thread_data *td)
{
	struct io_piece *ipo;
	struct rb_node *n;

	while ((n = rb_first(&td->io_hist_tree)) != NULL) {
		ipo = rb_entry(n, struct io_piece, rb_node);
		rb_erase(n, &td->io_hist_tree);
		remove_trim_entry(td, ipo);
		td->io_hist_len--;
		free(ipo);
	}

	while (!flist_empty(&td->io_hist_list)) {
		ipo = flist_entry(td->io_hist_list.next, struct io_piece, list);
		flist_del(&ipo->list);
		remove_trim_entry(td, ipo);
		td->io_hist_len--;
		free(ipo);
	}
}

/*
 * log a successful write, so we can unwind the log for verify
 */
void log_io_piece(struct thread_data *td, struct io_u *io_u)
{
	struct rb_node **p, *parent;
	struct io_piece *ipo, *__ipo;

	ipo = malloc(sizeof(struct io_piece));
	init_ipo(ipo);
	ipo->file = io_u->file;
	ipo->offset = io_u->offset;
	ipo->len = io_u->buflen;
	ipo->numberio = io_u->numberio;
	ipo->flags = IP_F_IN_FLIGHT;

	io_u->ipo = ipo;

	if (io_u_should_trim(td, io_u)) {
		flist_add_tail(&ipo->trim_list, &td->trim_list);
		td->trim_entries++;
	}

	/*
	 * We don't need to sort the entries, if:
	 *
	 *	Sequential writes, or
	 *	Random writes that lay out the file as it goes along
	 *
	 * For both these cases, just reading back data in the order we
	 * wrote it out is the fastest.
	 *
	 * One exception is if we don't have a random map AND we are doing
	 * verifies, in that case we need to check for duplicate blocks and
	 * drop the old one, which we rely on the rb insert/lookup for
	 * handling.
	 */
	if (((!td->o.verifysort) || !td_random(td) || !td->o.overwrite) &&
	      (file_randommap(td, ipo->file) || td->o.verify == VERIFY_NONE)) {
		INIT_FLIST_HEAD(&ipo->list);
		flist_add_tail(&ipo->list, &td->io_hist_list);
		ipo->flags |= IP_F_ONLIST;
		td->io_hist_len++;
		return;
	}

	RB_CLEAR_NODE(&ipo->rb_node);

	/*
	 * Sort the entry into the verification list
	 */
restart:
	p = &td->io_hist_tree.rb_node;
	parent = NULL;
	while (*p) {
		parent = *p;

		__ipo = rb_entry(parent, struct io_piece, rb_node);
		if (ipo->file < __ipo->file)
			p = &(*p)->rb_left;
		else if (ipo->file > __ipo->file)
			p = &(*p)->rb_right;
		else if (ipo->offset < __ipo->offset)
			p = &(*p)->rb_left;
		else if (ipo->offset > __ipo->offset)
			p = &(*p)->rb_right;
		else {
			dprint(FD_IO, "iolog: overlap %llu/%lu, %llu/%lu",
				__ipo->offset, __ipo->len,
				ipo->offset, ipo->len);
			td->io_hist_len--;
			rb_erase(parent, &td->io_hist_tree);
			remove_trim_entry(td, __ipo);
			free(__ipo);
			goto restart;
		}
	}

	rb_link_node(&ipo->rb_node, parent, p);
	rb_insert_color(&ipo->rb_node, &td->io_hist_tree);
	ipo->flags |= IP_F_ONRB;
	td->io_hist_len++;
}

void write_iolog_close(struct thread_data *td)
{
	fflush(td->iolog_f);
	fclose(td->iolog_f);
	free(td->iolog_buf);
	td->iolog_f = NULL;
	td->iolog_buf = NULL;
}

/*
 * Read version 2 iolog data. It is enhanced to include per-file logging,
 * syncs, etc.
 */
static int read_iolog2(struct thread_data *td, FILE *f)
{
	unsigned long long offset;
	unsigned int bytes;
	int reads, writes, waits, fileno = 0, file_action = 0; /* stupid gcc */
	char *fname, *act;
	char *str, *p;
	enum fio_ddir rw;

	free_release_files(td);

	/*
	 * Read in the read iolog and store it, reuse the infrastructure
	 * for doing verifications.
	 */
	str = malloc(4096);
	fname = malloc(256+16);
	act = malloc(256+16);

	reads = writes = waits = 0;
	while ((p = fgets(str, 4096, f)) != NULL) {
		struct io_piece *ipo;
		int r;

		r = sscanf(p, "%256s %256s %llu %u", fname, act, &offset,
									&bytes);
		if (r == 4) {
			/*
			 * Check action first
			 */
			if (!strcmp(act, "wait"))
				rw = DDIR_WAIT;
			else if (!strcmp(act, "read"))
				rw = DDIR_READ;
			else if (!strcmp(act, "write"))
				rw = DDIR_WRITE;
			else if (!strcmp(act, "sync"))
				rw = DDIR_SYNC;
			else if (!strcmp(act, "datasync"))
				rw = DDIR_DATASYNC;
			else if (!strcmp(act, "trim"))
				rw = DDIR_TRIM;
			else {
				log_err("fio: bad iolog file action: %s\n",
									act);
				continue;
			}
			fileno = get_fileno(td, fname);
		} else if (r == 2) {
			rw = DDIR_INVAL;
			if (!strcmp(act, "add")) {
				fileno = add_file(td, fname, 0, 1);
				file_action = FIO_LOG_ADD_FILE;
				continue;
			} else if (!strcmp(act, "open")) {
				fileno = get_fileno(td, fname);
				file_action = FIO_LOG_OPEN_FILE;
			} else if (!strcmp(act, "close")) {
				fileno = get_fileno(td, fname);
				file_action = FIO_LOG_CLOSE_FILE;
			} else {
				log_err("fio: bad iolog file action: %s\n",
									act);
				continue;
			}
		} else {
			log_err("bad iolog2: %s", p);
			continue;
		}

		if (rw == DDIR_READ)
			reads++;
		else if (rw == DDIR_WRITE) {
			/*
			 * Don't add a write for ro mode
			 */
			if (read_only)
				continue;
			writes++;
		} else if (rw == DDIR_WAIT) {
			waits++;
		} else if (rw == DDIR_INVAL) {
		} else if (!ddir_sync(rw)) {
			log_err("bad ddir: %d\n", rw);
			continue;
		}

		/*
		 * Make note of file
		 */
		ipo = malloc(sizeof(*ipo));
		init_ipo(ipo);
		ipo->ddir = rw;
		if (rw == DDIR_WAIT) {
			ipo->delay = offset;
		} else {
			ipo->offset = offset;
			ipo->len = bytes;
			if (rw != DDIR_INVAL && bytes > td->o.max_bs[rw])
				td->o.max_bs[rw] = bytes;
			ipo->fileno = fileno;
			ipo->file_action = file_action;
		}

		queue_io_piece(td, ipo);
	}

	free(str);
	free(act);
	free(fname);

	if (writes && read_only) {
		log_err("fio: <%s> skips replay of %d writes due to"
			" read-only\n", td->o.name, writes);
		writes = 0;
	}

	if (!reads && !writes && !waits)
		return 1;
	else if (reads && !writes)
		td->o.td_ddir = TD_DDIR_READ;
	else if (!reads && writes)
		td->o.td_ddir = TD_DDIR_WRITE;
	else
		td->o.td_ddir = TD_DDIR_RW;

	return 0;
}

/*
 * open iolog, check version, and call appropriate parser
 */
static int init_iolog_read(struct thread_data *td)
{
	char buffer[256], *p;
	FILE *f;
	int ret;

	f = fopen(td->o.read_iolog_file, "r");
	if (!f) {
		perror("fopen read iolog");
		return 1;
	}

	p = fgets(buffer, sizeof(buffer), f);
	if (!p) {
		td_verror(td, errno, "iolog read");
		log_err("fio: unable to read iolog\n");
		fclose(f);
		return 1;
	}

	/*
	 * version 2 of the iolog stores a specific string as the
	 * first line, check for that
	 */
	if (!strncmp(iolog_ver2, buffer, strlen(iolog_ver2)))
		ret = read_iolog2(td, f);
	else {
		log_err("fio: iolog version 1 is no longer supported\n");
		ret = 1;
	}

	fclose(f);
	return ret;
}

/*
 * Set up a log for storing io patterns.
 */
static int init_iolog_write(struct thread_data *td)
{
	struct fio_file *ff;
	FILE *f;
	unsigned int i;

	f = fopen(td->o.write_iolog_file, "a");
	if (!f) {
		perror("fopen write iolog");
		return 1;
	}

	/*
	 * That's it for writing, setup a log buffer and we're done.
	  */
	td->iolog_f = f;
	td->iolog_buf = malloc(8192);
	setvbuf(f, td->iolog_buf, _IOFBF, 8192);

	/*
	 * write our version line
	 */
	if (fprintf(f, "%s\n", iolog_ver2) < 0) {
		perror("iolog init\n");
		return 1;
	}

	/*
	 * add all known files
	 */
	for_each_file(td, ff, i)
		log_file(td, ff, FIO_LOG_ADD_FILE);

	return 0;
}

int init_iolog(struct thread_data *td)
{
	int ret = 0;

	if (td->o.read_iolog_file) {
		int need_swap;

		/*
		 * Check if it's a blktrace file and load that if possible.
		 * Otherwise assume it's a normal log file and load that.
		 */
		if (is_blktrace(td->o.read_iolog_file, &need_swap))
			ret = load_blktrace(td, td->o.read_iolog_file, need_swap);
		else
			ret = init_iolog_read(td);
	} else if (td->o.write_iolog_file)
		ret = init_iolog_write(td);

	if (ret)
		td_verror(td, EINVAL, "failed initializing iolog");

	return ret;
}

void setup_log(struct io_log **log, unsigned long avg_msec, int log_type)
{
	struct io_log *l = malloc(sizeof(*l));

	memset(l, 0, sizeof(*l));
	l->nr_samples = 0;
	l->max_samples = 1024;
	l->log_type = log_type;
	l->log = malloc(l->max_samples * sizeof(struct io_sample));
	l->avg_msec = avg_msec;
	*log = l;
}

#ifdef CONFIG_SETVBUF
static void *set_file_buffer(FILE *f)
{
	size_t size = 1048576;
	void *buf;

	buf = malloc(size);
	setvbuf(f, buf, _IOFBF, size);
	return buf;
}

static void clear_file_buffer(void *buf)
{
	free(buf);
}
#else
static void *set_file_buffer(FILE *f)
{
	return NULL;
}

static void clear_file_buffer(void *buf)
{
}
#endif

void __finish_log(struct io_log *log, const char *name)
{
	unsigned int i;
	void *buf;
	FILE *f;

	f = fopen(name, "a");
	if (!f) {
		perror("fopen log");
		return;
	}

	buf = set_file_buffer(f);

	for (i = 0; i < log->nr_samples; i++) {
		fprintf(f, "%lu, %lu, %u, %u\n",
				(unsigned long) log->log[i].time,
				(unsigned long) log->log[i].val,
				log->log[i].ddir, log->log[i].bs);
	}

	fclose(f);
	clear_file_buffer(buf);
	free(log->log);
	free(log);
}

static int finish_log_named(struct thread_data *td, struct io_log *log,
			    const char *prefix, const char *postfix,
			    int trylock)
{
	char file_name[256];

	snprintf(file_name, sizeof(file_name), "%s_%s.log", prefix, postfix);

	if (trylock) {
		if (fio_trylock_file(file_name))
			return 1;
	} else
		fio_lock_file(file_name);

	if (td->client_type == FIO_CLIENT_TYPE_GUI) {
		fio_send_iolog(td, log, file_name);
		free(log->log);
		free(log);
	} else
		__finish_log(log, file_name);

	fio_unlock_file(file_name);
	return 0;
}

static int finish_log(struct thread_data *td, struct io_log *log,
		      const char *name, int trylock)
{
	return finish_log_named(td, log, td->o.name, name, trylock);
}

static int write_this_log(struct thread_data *td, struct io_log *log,
			  const char *log_file, const char *name, int try)
{
	int ret;

	if (!log)
		return 0;

	if (log_file)
		ret = finish_log_named(td, log, log_file, name, try);
	else
		ret = finish_log(td, log, name, try);

	return ret;
}

static int write_iops_log(struct thread_data *td, int try)
{
	struct thread_options *o = &td->o;

	return write_this_log(td, td->iops_log, o->iops_log_file, "iops", try);
}

static int write_slat_log(struct thread_data *td, int try)
{
	struct thread_options *o = &td->o;

	return write_this_log(td, td->slat_log, o->lat_log_file, "slat", try);
}

static int write_clat_log(struct thread_data *td, int try)
{
	struct thread_options *o = &td->o;

	return write_this_log(td, td->clat_log, o->lat_log_file, "clat" , try);
}

static int write_lat_log(struct thread_data *td, int try)
{
	struct thread_options *o = &td->o;

	return write_this_log(td, td->lat_log, o->lat_log_file, "lat", try);
}

static int write_bandw_log(struct thread_data *td, int try)
{
	struct thread_options *o = &td->o;

	return write_this_log(td, td->bw_log, o->bw_log_file, "bw", try);
}

enum {
	BW_LOG_MASK	= 1,
	LAT_LOG_MASK	= 2,
	SLAT_LOG_MASK	= 4,
	CLAT_LOG_MASK	= 8,
	IOPS_LOG_MASK	= 16,

	ALL_LOG_NR	= 5,
};

struct log_type {
	unsigned int mask;
	int (*fn)(struct thread_data *, int);
};

static struct log_type log_types[] = {
	{
		.mask	= BW_LOG_MASK,
		.fn	= write_bandw_log,
	},
	{
		.mask	= LAT_LOG_MASK,
		.fn	= write_lat_log,
	},
	{
		.mask	= SLAT_LOG_MASK,
		.fn	= write_slat_log,
	},
	{
		.mask	= CLAT_LOG_MASK,
		.fn	= write_clat_log,
	},
	{
		.mask	= IOPS_LOG_MASK,
		.fn	= write_iops_log,
	},
};

void fio_writeout_logs(struct thread_data *td)
{
	unsigned int log_mask = 0;
	unsigned int log_left = ALL_LOG_NR;
	int old_state, i;

	old_state = td_bump_runstate(td, TD_FINISHING);

	finalize_logs(td);

	while (log_left) {
		int prev_log_left = log_left;

		for (i = 0; i < ALL_LOG_NR && log_left; i++) {
			struct log_type *lt = &log_types[i];
			int ret;

			if (!(log_mask & lt->mask)) {
				ret = lt->fn(td, log_left != 1);
				if (!ret) {
					log_left--;
					log_mask |= lt->mask;
				}
			}
		}

		if (prev_log_left == log_left)
			usleep(5000);
	}

	td_restore_runstate(td, old_state);
}
Commit	Line	Data
ac9b9101 JA	1	/*
	2	* Code related to writing an iolog of what a thread is doing, and to
	3	* later read that back and replay
	4	*/
	5	#include <stdio.h>
	6	#include <stdlib.h>
	7	#include <libgen.h>
	8	#include <assert.h>
	9	#include "flist.h"
	10	#include "fio.h"
	11	#include "verify.h"
	12	#include "trim.h"
243bfe19	13	#include "filelock.h"
ac9b9101 JA	14
	15	static const char iolog_ver2[] = "fio version 2 iolog";
	16
	17	void queue_io_piece(struct thread_data td, struct io_piece ipo)
	18	{
	19	flist_add_tail(&ipo->list, &td->io_log_list);
	20	td->total_io_size += ipo->len;
	21	}
	22
	23	void log_io_u(struct thread_data td, struct io_u io_u)
	24	{
	25	const char *act[] = { "read", "write", "sync", "datasync",
	26	"sync_file_range", "wait", "trim" };
	27
	28	assert(io_u->ddir <= 6);
	29
	30	if (!td->o.write_iolog_file)
	31	return;
	32
	33	fprintf(td->iolog_f, "%s %s %llu %lu\n", io_u->file->file_name,
	34	act[io_u->ddir], io_u->offset,
	35	io_u->buflen);
	36	}
	37
	38	void log_file(struct thread_data td, struct fio_file f,
	39	enum file_log_act what)
	40	{
	41	const char *act[] = { "add", "open", "close" };
	42
	43	assert(what < 3);
	44
	45	if (!td->o.write_iolog_file)
	46	return;
	47
	48
	49	/*
	50	* this happens on the pre-open/close done before the job starts
	51	*/
	52	if (!td->iolog_f)
	53	return;
	54
	55	fprintf(td->iolog_f, "%s %s\n", f->file_name, act[what]);
	56	}
	57
	58	static void iolog_delay(struct thread_data *td, unsigned long delay)
	59	{
	60	unsigned long usec = utime_since_now(&td->last_issue);
	61
	62	if (delay < usec)
	63	return;
	64
	65	delay -= usec;
	66
	67	/*
	68	* less than 100 usec delay, just regard it as noise
	69	*/
	70	if (delay < 100)
	71	return;
	72
	73	usec_sleep(td, delay);
	74	}
	75
	76	static int ipo_special(struct thread_data td, struct io_piece ipo)
	77	{
78	struct fio_file *f;
79	int ret;
80
81	/*
82	* Not a special ipo
83	*/
84	if (ipo->ddir != DDIR_INVAL)
85	return 0;
86
87	f = td->files[ipo->fileno];
88
89	switch (ipo->file_action) {
90	case FIO_LOG_OPEN_FILE:
91	ret = td_io_open_file(td, f);
92	if (!ret)
93	break;
94	td_verror(td, ret, "iolog open file");
95	return -1;
96	case FIO_LOG_CLOSE_FILE:
97	td_io_close_file(td, f);
98	break;
99	case FIO_LOG_UNLINK_FILE:
100	unlink(f->file_name);
101	break;
102	default:
103	log_err("fio: bad file action %d\n", ipo->file_action);
104	break;
105	}
106
107	return 1;
108	}
109
110	int read_iolog_get(struct thread_data td, struct io_u io_u)
111	{
112	struct io_piece *ipo;
113	unsigned long elapsed;
3c3ed070	114
ac9b9101 JA	115	while (!flist_empty(&td->io_log_list)) {
	116	int ret;
	117
	118	ipo = flist_entry(td->io_log_list.next, struct io_piece, list);
	119	flist_del(&ipo->list);
	120	remove_trim_entry(td, ipo);
	121
	122	ret = ipo_special(td, ipo);
	123	if (ret < 0) {
	124	free(ipo);
	125	break;
	126	} else if (ret > 0) {
	127	free(ipo);
	128	continue;
	129	}
	130
	131	io_u->ddir = ipo->ddir;
	132	if (ipo->ddir != DDIR_WAIT) {
	133	io_u->offset = ipo->offset;
	134	io_u->buflen = ipo->len;
	135	io_u->file = td->files[ipo->fileno];
	136	get_file(io_u->file);
	137	dprint(FD_IO, "iolog: get %llu/%lu/%s\n", io_u->offset,
	138	io_u->buflen, io_u->file->file_name);
	139	if (ipo->delay)
	140	iolog_delay(td, ipo->delay);
	141	} else {
	142	elapsed = mtime_since_genesis();
	143	if (ipo->delay > elapsed)
	144	usec_sleep(td, (ipo->delay - elapsed) * 1000);
ac9b9101 JA	145	}
	146
	147	free(ipo);
3c3ed070	148
ac9b9101 JA	149	if (io_u->ddir != DDIR_WAIT)
	150	return 0;
	151	}
	152
	153	td->done = 1;
	154	return 1;
	155	}
	156
	157	void prune_io_piece_log(struct thread_data *td)
	158	{
	159	struct io_piece *ipo;
	160	struct rb_node *n;
	161
	162	while ((n = rb_first(&td->io_hist_tree)) != NULL) {
	163	ipo = rb_entry(n, struct io_piece, rb_node);
	164	rb_erase(n, &td->io_hist_tree);
	165	remove_trim_entry(td, ipo);
	166	td->io_hist_len--;
	167	free(ipo);
	168	}
	169
	170	while (!flist_empty(&td->io_hist_list)) {
	171	ipo = flist_entry(td->io_hist_list.next, struct io_piece, list);
	172	flist_del(&ipo->list);
	173	remove_trim_entry(td, ipo);
	174	td->io_hist_len--;
	175	free(ipo);
	176	}
	177	}
	178
	179	/*
	180	* log a successful write, so we can unwind the log for verify
	181	*/
	182	void log_io_piece(struct thread_data td, struct io_u io_u)
	183	{
	184	struct rb_node *p, parent;
	185	struct io_piece ipo, __ipo;
	186
	187	ipo = malloc(sizeof(struct io_piece));
	188	init_ipo(ipo);
	189	ipo->file = io_u->file;
	190	ipo->offset = io_u->offset;
	191	ipo->len = io_u->buflen;
da0a7bd2	192	ipo->numberio = io_u->numberio;
f9401285 JA	193	ipo->flags = IP_F_IN_FLIGHT;
	194
	195	io_u->ipo = ipo;
ac9b9101 JA	196
	197	if (io_u_should_trim(td, io_u)) {
	198	flist_add_tail(&ipo->trim_list, &td->trim_list);
	199	td->trim_entries++;
	200	}
	201
	202	/*
	203	* We don't need to sort the entries, if:
	204	*
	205	* Sequential writes, or
	206	* Random writes that lay out the file as it goes along
	207	*
	208	* For both these cases, just reading back data in the order we
	209	* wrote it out is the fastest.
	210	*
	211	* One exception is if we don't have a random map AND we are doing
	212	* verifies, in that case we need to check for duplicate blocks and
	213	* drop the old one, which we rely on the rb insert/lookup for
	214	* handling.
	215	*/
c4b6117b	216	if (((!td->o.verifysort) \|\| !td_random(td) \|\| !td->o.overwrite) &&
ac9b9101 JA	217	(file_randommap(td, ipo->file) \|\| td->o.verify == VERIFY_NONE)) {
	218	INIT_FLIST_HEAD(&ipo->list);
	219	flist_add_tail(&ipo->list, &td->io_hist_list);
	220	ipo->flags \|= IP_F_ONLIST;
	221	td->io_hist_len++;
	222	return;
	223	}
	224
	225	RB_CLEAR_NODE(&ipo->rb_node);
	226
	227	/*
	228	* Sort the entry into the verification list
	229	*/
	230	restart:
	231	p = &td->io_hist_tree.rb_node;
	232	parent = NULL;
	233	while (*p) {
	234	parent = *p;
	235
	236	__ipo = rb_entry(parent, struct io_piece, rb_node);
	237	if (ipo->file < __ipo->file)
	238	p = &(*p)->rb_left;
	239	else if (ipo->file > __ipo->file)
	240	p = &(*p)->rb_right;
	241	else if (ipo->offset < __ipo->offset)
	242	p = &(*p)->rb_left;
	243	else if (ipo->offset > __ipo->offset)
	244	p = &(*p)->rb_right;
	245	else {
885ac623 JA	246	dprint(FD_IO, "iolog: overlap %llu/%lu, %llu/%lu",
	247	__ipo->offset, __ipo->len,
	248	ipo->offset, ipo->len);
ac9b9101 JA	249	td->io_hist_len--;
	250	rb_erase(parent, &td->io_hist_tree);
	251	remove_trim_entry(td, __ipo);
	252	free(__ipo);
	253	goto restart;
	254	}
	255	}
	256
	257	rb_link_node(&ipo->rb_node, parent, p);
	258	rb_insert_color(&ipo->rb_node, &td->io_hist_tree);
	259	ipo->flags \|= IP_F_ONRB;
	260	td->io_hist_len++;
	261	}
	262
	263	void write_iolog_close(struct thread_data *td)
	264	{
	265	fflush(td->iolog_f);
	266	fclose(td->iolog_f);
	267	free(td->iolog_buf);
	268	td->iolog_f = NULL;
	269	td->iolog_buf = NULL;
	270	}
	271
	272	/*
	273	* Read version 2 iolog data. It is enhanced to include per-file logging,
	274	* syncs, etc.
	275	*/
	276	static int read_iolog2(struct thread_data td, FILE f)
	277	{
	278	unsigned long long offset;
	279	unsigned int bytes;
	280	int reads, writes, waits, fileno = 0, file_action = 0; /* stupid gcc */
	281	char fname, act;
	282	char str, p;
	283	enum fio_ddir rw;
	284
	285	free_release_files(td);
	286
	287	/*
	288	* Read in the read iolog and store it, reuse the infrastructure
	289	* for doing verifications.
	290	*/
	291	str = malloc(4096);
	292	fname = malloc(256+16);
	293	act = malloc(256+16);
	294
	295	reads = writes = waits = 0;
	296	while ((p = fgets(str, 4096, f)) != NULL) {
	297	struct io_piece *ipo;
	298	int r;
	299
	300	r = sscanf(p, "%256s %256s %llu %u", fname, act, &offset,
	301	&bytes);
	302	if (r == 4) {
	303	/*
	304	* Check action first
	305	*/
	306	if (!strcmp(act, "wait"))
	307	rw = DDIR_WAIT;
	308	else if (!strcmp(act, "read"))
	309	rw = DDIR_READ;
	310	else if (!strcmp(act, "write"))
	311	rw = DDIR_WRITE;
	312	else if (!strcmp(act, "sync"))
313	rw = DDIR_SYNC;
314	else if (!strcmp(act, "datasync"))
315	rw = DDIR_DATASYNC;
316	else if (!strcmp(act, "trim"))
317	rw = DDIR_TRIM;
318	else {
319	log_err("fio: bad iolog file action: %s\n",
320	act);
321	continue;
322	}
033ace1e	323	fileno = get_fileno(td, fname);
ac9b9101 JA	324	} else if (r == 2) {
	325	rw = DDIR_INVAL;
	326	if (!strcmp(act, "add")) {
5903e7b7	327	fileno = add_file(td, fname, 0, 1);
ac9b9101 JA	328	file_action = FIO_LOG_ADD_FILE;
	329	continue;
	330	} else if (!strcmp(act, "open")) {
	331	fileno = get_fileno(td, fname);
	332	file_action = FIO_LOG_OPEN_FILE;
	333	} else if (!strcmp(act, "close")) {
	334	fileno = get_fileno(td, fname);
	335	file_action = FIO_LOG_CLOSE_FILE;
	336	} else {
	337	log_err("fio: bad iolog file action: %s\n",
	338	act);
	339	continue;
	340	}
	341	} else {
	342	log_err("bad iolog2: %s", p);
	343	continue;
	344	}
	345
	346	if (rw == DDIR_READ)
	347	reads++;
	348	else if (rw == DDIR_WRITE) {
	349	/*
	350	* Don't add a write for ro mode
	351	*/
	352	if (read_only)
	353	continue;
	354	writes++;
	355	} else if (rw == DDIR_WAIT) {
	356	waits++;
	357	} else if (rw == DDIR_INVAL) {
	358	} else if (!ddir_sync(rw)) {
	359	log_err("bad ddir: %d\n", rw);
	360	continue;
	361	}
	362
	363	/*
	364	* Make note of file
	365	*/
	366	ipo = malloc(sizeof(*ipo));
	367	init_ipo(ipo);
	368	ipo->ddir = rw;
	369	if (rw == DDIR_WAIT) {
	370	ipo->delay = offset;
	371	} else {
	372	ipo->offset = offset;
	373	ipo->len = bytes;
42793d94	374	if (rw != DDIR_INVAL && bytes > td->o.max_bs[rw])
ac9b9101 JA	375	td->o.max_bs[rw] = bytes;
	376	ipo->fileno = fileno;
	377	ipo->file_action = file_action;
	378	}
3c3ed070	379
ac9b9101 JA	380	queue_io_piece(td, ipo);
	381	}
	382
	383	free(str);
	384	free(act);
	385	free(fname);
	386
	387	if (writes && read_only) {
	388	log_err("fio: <%s> skips replay of %d writes due to"
	389	" read-only\n", td->o.name, writes);
	390	writes = 0;
	391	}
	392
	393	if (!reads && !writes && !waits)
	394	return 1;
	395	else if (reads && !writes)
	396	td->o.td_ddir = TD_DDIR_READ;
	397	else if (!reads && writes)
	398	td->o.td_ddir = TD_DDIR_WRITE;
	399	else
	400	td->o.td_ddir = TD_DDIR_RW;
	401
	402	return 0;
	403	}
	404
	405	/*
	406	* open iolog, check version, and call appropriate parser
	407	*/
	408	static int init_iolog_read(struct thread_data *td)
	409	{
	410	char buffer[256], *p;
	411	FILE *f;
	412	int ret;
	413
	414	f = fopen(td->o.read_iolog_file, "r");
	415	if (!f) {
	416	perror("fopen read iolog");
	417	return 1;
	418	}
	419
	420	p = fgets(buffer, sizeof(buffer), f);
	421	if (!p) {
	422	td_verror(td, errno, "iolog read");
	423	log_err("fio: unable to read iolog\n");
	424	fclose(f);
	425	return 1;
	426	}
	427
	428	/*
	429	* version 2 of the iolog stores a specific string as the
	430	* first line, check for that
	431	*/
	432	if (!strncmp(iolog_ver2, buffer, strlen(iolog_ver2)))
	433	ret = read_iolog2(td, f);
	434	else {
	435	log_err("fio: iolog version 1 is no longer supported\n");
	436	ret = 1;
	437	}
	438
	439	fclose(f);
	440	return ret;
	441	}
	442
	443	/*
444	* Set up a log for storing io patterns.
445	*/
446	static int init_iolog_write(struct thread_data *td)
447	{
448	struct fio_file *ff;
449	FILE *f;
450	unsigned int i;
451
452	f = fopen(td->o.write_iolog_file, "a");
453	if (!f) {
454	perror("fopen write iolog");
455	return 1;
456	}
457
458	/*
459	* That's it for writing, setup a log buffer and we're done.
460	*/
461	td->iolog_f = f;
462	td->iolog_buf = malloc(8192);
463	setvbuf(f, td->iolog_buf, _IOFBF, 8192);
464
465	/*
466	* write our version line
467	*/
468	if (fprintf(f, "%s\n", iolog_ver2) < 0) {
469	perror("iolog init\n");
470	return 1;
471	}
472
473	/*
474	* add all known files
475	*/
476	for_each_file(td, ff, i)
477	log_file(td, ff, FIO_LOG_ADD_FILE);
478
479	return 0;
480	}
481
482	int init_iolog(struct thread_data *td)
483	{
484	int ret = 0;
485
486	if (td->o.read_iolog_file) {
d95b34a6 JA	487	int need_swap;
d95b34a6 JA	488
ac9b9101 JA	489	/*
	490	* Check if it's a blktrace file and load that if possible.
	491	* Otherwise assume it's a normal log file and load that.
	492	*/
d95b34a6 JA	493	if (is_blktrace(td->o.read_iolog_file, &need_swap))
d95b34a6 JA	494	ret = load_blktrace(td, td->o.read_iolog_file, need_swap);
ac9b9101 JA	495	else
	496	ret = init_iolog_read(td);
	497	} else if (td->o.write_iolog_file)
	498	ret = init_iolog_write(td);
	499
f01b34ae JA	500	if (ret)
	501	td_verror(td, EINVAL, "failed initializing iolog");
	502
ac9b9101 JA	503	return ret;
	504	}
	505
ea51b956	506	void setup_log(struct io_log **log, unsigned long avg_msec, int log_type)
ac9b9101 JA	507	{
	508	struct io_log l = malloc(sizeof(l));
	509
b8bc8cba	510	memset(l, 0, sizeof(*l));
ac9b9101 JA	511	l->nr_samples = 0;
ac9b9101 JA	512	l->max_samples = 1024;
ea51b956	513	l->log_type = log_type;
ac9b9101	514	l->log = malloc(l->max_samples * sizeof(struct io_sample));
b8bc8cba	515	l->avg_msec = avg_msec;
ac9b9101 JA	516	*log = l;
	517	}
	518
2e802282 JA	519	#ifdef CONFIG_SETVBUF
	520	static void set_file_buffer(FILE f)
	521	{
	522	size_t size = 1048576;
	523	void *buf;
	524
	525	buf = malloc(size);
	526	setvbuf(f, buf, _IOFBF, size);
	527	return buf;
	528	}
	529
	530	static void clear_file_buffer(void *buf)
	531	{
	532	free(buf);
	533	}
	534	#else
	535	static void set_file_buffer(FILE f)
	536	{
	537	return NULL;
	538	}
	539
	540	static void clear_file_buffer(void *buf)
	541	{
	542	}
	543	#endif
	544
ac9b9101 JA	545	void __finish_log(struct io_log log, const char name)
	546	{
	547	unsigned int i;
2e802282	548	void *buf;
ac9b9101 JA	549	FILE *f;
	550
	551	f = fopen(name, "a");
	552	if (!f) {
	553	perror("fopen log");
	554	return;
	555	}
	556
2e802282 JA	557	buf = set_file_buffer(f);
2e802282 JA	558
ac9b9101	559	for (i = 0; i < log->nr_samples; i++) {
1fae4855 JA	560	fprintf(f, "%lu, %lu, %u, %u\n",
	561	(unsigned long) log->log[i].time,
	562	(unsigned long) log->log[i].val,
	563	log->log[i].ddir, log->log[i].bs);
ac9b9101 JA	564	}
	565
	566	fclose(f);
2e802282	567	clear_file_buffer(buf);
ac9b9101 JA	568	free(log->log);
	569	free(log);
	570	}
	571
905e3d4f JA	572	static int finish_log_named(struct thread_data td, struct io_log log,
	573	const char prefix, const char postfix,
	574	int trylock)
ac9b9101	575	{
e53cf97b	576	char file_name[256];
ac9b9101	577
98ffb8f3	578	snprintf(file_name, sizeof(file_name), "%s_%s.log", prefix, postfix);
1b42725f	579
243bfe19 JA	580	if (trylock) {
	581	if (fio_trylock_file(file_name))
	582	return 1;
	583	} else
	584	fio_lock_file(file_name);
	585
f5ed765a	586	if (td->client_type == FIO_CLIENT_TYPE_GUI) {
e53cf97b	587	fio_send_iolog(td, log, file_name);
f5ed765a JA	588	free(log->log);
	589	free(log);
	590	} else
e53cf97b	591	__finish_log(log, file_name);
243bfe19 JA	592
	593	fio_unlock_file(file_name);
	594	return 0;
ac9b9101 JA	595	}
ac9b9101 JA	596
905e3d4f JA	597	static int finish_log(struct thread_data td, struct io_log log,
905e3d4f JA	598	const char *name, int trylock)
ac9b9101	599	{
243bfe19	600	return finish_log_named(td, log, td->o.name, name, trylock);
ac9b9101	601	}
905e3d4f JA	602
	603	static int write_this_log(struct thread_data td, struct io_log log,
	604	const char log_file, const char name, int try)
	605	{
	606	int ret;
	607
	608	if (!log)
	609	return 0;
	610
	611	if (log_file)
	612	ret = finish_log_named(td, log, log_file, name, try);
	613	else
	614	ret = finish_log(td, log, name, try);
	615
	616	return ret;
	617	}
	618
	619	static int write_iops_log(struct thread_data *td, int try)
	620	{
	621	struct thread_options *o = &td->o;
	622
	623	return write_this_log(td, td->iops_log, o->iops_log_file, "iops", try);
	624	}
	625
	626	static int write_slat_log(struct thread_data *td, int try)
	627	{
	628	struct thread_options *o = &td->o;
	629
	630	return write_this_log(td, td->slat_log, o->lat_log_file, "slat", try);
	631	}
	632
	633	static int write_clat_log(struct thread_data *td, int try)
	634	{
	635	struct thread_options *o = &td->o;
	636
	637	return write_this_log(td, td->clat_log, o->lat_log_file, "clat" , try);
	638	}
	639
	640	static int write_lat_log(struct thread_data *td, int try)
	641	{
	642	struct thread_options *o = &td->o;
	643
	644	return write_this_log(td, td->lat_log, o->lat_log_file, "lat", try);
	645	}
	646
	647	static int write_bandw_log(struct thread_data *td, int try)
	648	{
	649	struct thread_options *o = &td->o;
	650
	651	return write_this_log(td, td->bw_log, o->bw_log_file, "bw", try);
	652	}
	653
	654	enum {
	655	BW_LOG_MASK = 1,
	656	LAT_LOG_MASK = 2,
	657	SLAT_LOG_MASK = 4,
	658	CLAT_LOG_MASK = 8,
	659	IOPS_LOG_MASK = 16,
	660
905e3d4f JA	661	ALL_LOG_NR = 5,
	662	};
	663
	664	struct log_type {
	665	unsigned int mask;
	666	int (fn)(struct thread_data , int);
	667	};
	668
	669	static struct log_type log_types[] = {
	670	{
	671	.mask = BW_LOG_MASK,
	672	.fn = write_bandw_log,
	673	},
	674	{
	675	.mask = LAT_LOG_MASK,
	676	.fn = write_lat_log,
	677	},
	678	{
	679	.mask = SLAT_LOG_MASK,
	680	.fn = write_slat_log,
	681	},
	682	{
	683	.mask = CLAT_LOG_MASK,
	684	.fn = write_clat_log,
	685	},
	686	{
	687	.mask = IOPS_LOG_MASK,
	688	.fn = write_iops_log,
	689	},
	690	};
	691
	692	void fio_writeout_logs(struct thread_data *td)
	693	{
ea5409f9	694	unsigned int log_mask = 0;
905e3d4f JA	695	unsigned int log_left = ALL_LOG_NR;
	696	int old_state, i;
	697
	698	old_state = td_bump_runstate(td, TD_FINISHING);
	699
	700	finalize_logs(td);
	701
	702	while (log_left) {
	703	int prev_log_left = log_left;
	704
	705	for (i = 0; i < ALL_LOG_NR && log_left; i++) {
	706	struct log_type *lt = &log_types[i];
	707	int ret;
	708
ea5409f9	709	if (!(log_mask & lt->mask)) {
905e3d4f JA	710	ret = lt->fn(td, log_left != 1);
	711	if (!ret) {
	712	log_left--;
ea5409f9	713	log_mask \|= lt->mask;
905e3d4f JA	714	}
	715	}
	716	}
	717
	718	if (prev_log_left == log_left)
	719	usleep(5000);
	720	}
	721
	722	td_restore_runstate(td, old_state);
	723	}