static int verify_io_u_meta(struct verify_header *hdr, struct vcont *vc)
{
struct thread_data *td = vc->td;
struct vhdr_meta *vh = hdr_priv(hdr);
struct io_u *io_u = vc->io_u;
int ret = EILSEQ;
dprint(FD_VERIFY, "meta verify io_u %p, len %u\n", io_u, hdr->len);
if (vh->offset == io_u->offset + vc->hdr_num * td->o.verify_interval)
ret = 0;
if (td->o.verify_pattern_bytes)
ret |= verify_io_u_pattern(hdr, vc);
/*
* 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))
if (!td->o.verify_only || td->o.loops == 0)
if (vh->numberio != io_u->numberio)
ret = EILSEQ;
if (!ret)
return 0;
vc->name = "meta";
log_verify_failure(hdr, vc);
return ret;
}
/**
* \ingroup btree_private
*
* Puts a new node (with a new key) into the btree.
*
* ---- Description ----
*
* \param td Non-NULL pointer to an initialized btree structre.
* \param key Non-NULL pointer to an initialized tdt_t structure.
* \param valueptr Offset to the written value node.
* \param keyptr On successful return, the offset of the written key node.
* \param previous Offset to the previous node.
* \param next Offset to the next node.
* \param flags Bits from the TDFLG_* values.
*
* \returns Zero on success, or any error of the underlying functions.
*/
int td_put_new
(
td_t* td,
const tdt_t* key,
unsigned valueptr,
unsigned* keyptr,
unsigned previous,
unsigned next,
unsigned flags
)
{
struct keyhead keyhead = TD_KEYHEAD_NULL;
keyhead.value = valueptr;
keyhead.size = sizeof(struct keyhead) + key->size;
keyhead.previous = previous;
keyhead.next = next;
if (flags & TDFLG_CHECKSUM) {
td_checksum_create(key->data, key->size, &(keyhead.checksum));
}
CHECK(td_claim(td, 1, keyptr, &(keyhead.size)));
CHECK(td_write_keyhead(td, (*keyptr), &keyhead));
CHECK(td_write(td, (*keyptr) + sizeof(struct keyhead), key->data, key->size));
++(td->header.nentries);
++(td->header.scn);
return 0;
}
/*
* Return the data direction for the next io_u. If the job is a
* mixed read/write workload, check the rwmix cycle and switch if
* necessary.
*/
static enum fio_ddir get_rw_ddir(struct thread_data *td)
{
enum fio_ddir ddir;
/*
* see if it's time to fsync
*/
if (td->o.fsync_blocks &&
!(td->io_issues[DDIR_WRITE] % td->o.fsync_blocks) &&
td->io_issues[DDIR_WRITE] && should_fsync(td))
return DDIR_SYNC;
/*
* see if it's time to fdatasync
*/
if (td->o.fdatasync_blocks &&
!(td->io_issues[DDIR_WRITE] % td->o.fdatasync_blocks) &&
td->io_issues[DDIR_WRITE] && should_fsync(td))
return DDIR_DATASYNC;
/*
* see if it's time to sync_file_range
*/
if (td->sync_file_range_nr &&
!(td->io_issues[DDIR_WRITE] % td->sync_file_range_nr) &&
td->io_issues[DDIR_WRITE] && should_fsync(td))
return DDIR_SYNC_FILE_RANGE;
if (td_rw(td)) {
/*
* Check if it's time to seed a new data direction.
*/
if (td->io_issues[td->rwmix_ddir] >= td->rwmix_issues) {
/*
* Put a top limit on how many bytes we do for
* one data direction, to avoid overflowing the
* ranges too much
*/
ddir = get_rand_ddir(td);
if (ddir != td->rwmix_ddir)
set_rwmix_bytes(td);
td->rwmix_ddir = ddir;
}
ddir = td->rwmix_ddir;
} else if (td_read(td))
ddir = DDIR_READ;
else if (td_write(td))
ddir = DDIR_WRITE;
else
ddir = DDIR_TRIM;
td->rwmix_ddir = rate_ddir(td, ddir);
return td->rwmix_ddir;
}
static void update_accounting(struct submit_worker *sw)
{
struct thread_data *src = &sw->td;
struct thread_data *dst = sw->wq->td;
if (td_read(src))
sum_ddir(dst, src, DDIR_READ);
if (td_write(src))
sum_ddir(dst, src, DDIR_WRITE);
if (td_trim(src))
sum_ddir(dst, src, DDIR_TRIM);
}
static int fio_mmapio_open(struct thread_data *td, struct fio_file *f)
{
int ret, flags;
ret = generic_open_file(td, f);
if (ret)
return ret;
/*
* for size checkup, don't mmap anything.
*/
if (!f->io_size)
return 0;
if (td_rw(td))
flags = PROT_READ | PROT_WRITE;
else if (td_write(td)) {
flags = PROT_WRITE;
if (td->o.verify != VERIFY_NONE)
flags |= PROT_READ;
} else
flags = PROT_READ;
f->mmap = mmap(NULL, f->io_size, flags, MAP_SHARED, f->fd, f->file_offset);
if (f->mmap == MAP_FAILED) {
f->mmap = NULL;
td_verror(td, errno, "mmap");
goto err;
}
if (file_invalidate_cache(td, f))
goto err;
if (!td_random(td)) {
if (madvise(f->mmap, f->io_size, MADV_SEQUENTIAL) < 0) {
td_verror(td, errno, "madvise");
goto err;
}
} else {
if (madvise(f->mmap, f->io_size, MADV_RANDOM) < 0) {
td_verror(td, errno, "madvise");
goto err;
}
}
return 0;
err:
td->io_ops->close_file(td, f);
return 1;
}
static void io_workqueue_update_acct_fn(struct submit_worker *sw)
{
struct thread_data *src = sw->priv;
struct thread_data *dst = sw->wq->td;
if (td_read(src))
sum_ddir(dst, src, DDIR_READ);
if (td_write(src))
sum_ddir(dst, src, DDIR_WRITE);
if (td_trim(src))
sum_ddir(dst, src, DDIR_TRIM);
}
static int fio_mmap_file(struct thread_data *td, struct fio_file *f,
size_t length, off_t off)
{
struct fio_mmap_data *fmd = FILE_ENG_DATA(f);
int flags = 0;
if (td_rw(td) && !td->o.verify_only)
flags = PROT_READ | PROT_WRITE;
else if (td_write(td) && !td->o.verify_only) {
flags = PROT_WRITE;
if (td->o.verify != VERIFY_NONE)
flags |= PROT_READ;
} else
flags = PROT_READ;
fmd->mmap_ptr = mmap(NULL, length, flags, MAP_SHARED, f->fd, off);
if (fmd->mmap_ptr == MAP_FAILED) {
fmd->mmap_ptr = NULL;
td_verror(td, errno, "mmap");
goto err;
}
if (!fio_madvise_file(td, f, length))
goto err;
if (posix_madvise(fmd->mmap_ptr, length, POSIX_MADV_DONTNEED) < 0) {
td_verror(td, errno, "madvise");
goto err;
}
#ifdef FIO_MADV_FREE
if (f->filetype == FIO_TYPE_BLOCK)
(void) posix_madvise(fmd->mmap_ptr, fmd->mmap_sz, FIO_MADV_FREE);
#endif
err:
if (td->error && fmd->mmap_ptr)
munmap(fmd->mmap_ptr, length);
return td->error;
}
/**
* \ingroup btree_private
*
* Set administrated empty space to zero.
*
* \param td Non-NULL pointer to initialized td_t structure.
* \param off Offset to the chunk to be zeroised.
*
* \returns Zero on success, or non-zero on failure.
*/
int td_yield_set_zero
(td_t* td, unsigned off)
{
uchar zerobuf[ 256 ];
unsigned size;
memset(zerobuf, 0, sizeof(zerobuf));
CHECK(td_read_uint(td, off + sizeof(unsigned), &size));
if (size < (2 * sizeof(unsigned))) {
return TDERR_BOUNDS;
}
off += (2 * sizeof(unsigned));
size -= (2 * sizeof(unsigned));
while (size) {
unsigned l = ((size > sizeof(zerobuf)) ? sizeof(zerobuf) : size);
CHECK(td_write(td, off, zerobuf, l));
off += l;
size -= l;
}
return 0;
}
/*
* This is the mmap execution function
*/
static int fio_libpmem_file(struct thread_data *td, struct fio_file *f,
size_t length, off_t off)
{
struct fio_libpmem_data *fdd = FILE_ENG_DATA(f);
int flags = 0;
void *addr = NULL;
dprint(FD_IO, "DEBUG fio_libpmem_file\n");
if (td_rw(td))
flags = PROT_READ | PROT_WRITE;
else if (td_write(td)) {
flags = PROT_WRITE;
if (td->o.verify != VERIFY_NONE)
flags |= PROT_READ;
} else
flags = PROT_READ;
dprint(FD_IO, "f->file_name = %s td->o.verify = %d \n", f->file_name,
td->o.verify);
dprint(FD_IO, "length = %ld flags = %d f->fd = %d off = %ld \n",
length, flags, f->fd,off);
addr = util_map_hint(length, 0);
fdd->libpmem_ptr = mmap(addr, length, flags, MAP_SHARED, f->fd, off);
if (fdd->libpmem_ptr == MAP_FAILED) {
fdd->libpmem_ptr = NULL;
td_verror(td, errno, "mmap");
}
if (td->error && fdd->libpmem_ptr)
munmap(fdd->libpmem_ptr, length);
return td->error;
}
/*
* Best effort calculation of the estimated pending runtime of a job.
*/
static int thread_eta(struct thread_data *td)
{
unsigned long long bytes_total, bytes_done;
unsigned long eta_sec = 0;
unsigned long elapsed;
elapsed = (mtime_since_now(&td->epoch) + 999) / 1000;
bytes_total = td->total_io_size;
/*
* if writing, bytes_total will be twice the size. If mixing,
* assume a 50/50 split and thus bytes_total will be 50% larger.
*/
if (td->o.do_verify && td->o.verify && td_write(td)) {
if (td_rw(td))
bytes_total = bytes_total * 3 / 2;
else
bytes_total <<= 1;
}
if (td->o.zone_size && td->o.zone_skip)
bytes_total /= (td->o.zone_skip / td->o.zone_size);
if (td->o.fill_device && td->o.size == -1ULL)
return 0;
if (td->runstate == TD_RUNNING || td->runstate == TD_VERIFYING) {
double perc, perc_t;
bytes_done = td->io_bytes[DDIR_READ] + td->io_bytes[DDIR_WRITE];
perc = (double) bytes_done / (double) bytes_total;
if (perc > 1.0)
perc = 1.0;
if (td->o.time_based) {
perc_t = (double) elapsed / (double) td->o.timeout;
if (perc_t < perc)
perc = perc_t;
}
eta_sec = (unsigned long) (elapsed * (1.0 / perc)) - elapsed;
if (td->o.timeout &&
eta_sec > (td->o.timeout + done_secs - elapsed))
eta_sec = td->o.timeout + done_secs - elapsed;
} else if (td->runstate == TD_NOT_CREATED || td->runstate == TD_CREATED
|| td->runstate == TD_INITIALIZED
|| td->runstate == TD_RAMP
|| td->runstate == TD_PRE_READING) {
int t_eta = 0, r_eta = 0;
/*
* We can only guess - assume it'll run the full timeout
* if given, otherwise assume it'll run at the specified rate.
*/
if (td->o.timeout) {
t_eta = td->o.timeout + td->o.start_delay;
if (in_ramp_time(td)) {
unsigned long ramp_left;
ramp_left = mtime_since_now(&td->start);
ramp_left = (ramp_left + 999) / 1000;
if (ramp_left <= t_eta)
t_eta -= ramp_left;
}
}
if (td->o.rate[0] || td->o.rate[1]) {
r_eta = (bytes_total / 1024) / (td->o.rate[0] + td->o.rate[1]);
r_eta += td->o.start_delay;
}
if (r_eta && t_eta)
eta_sec = min(r_eta, t_eta);
else if (r_eta)
eta_sec = r_eta;
else if (t_eta)
eta_sec = t_eta;
else
eta_sec = 0;
} else {
/*
* thread is already done or waiting for fsync
*/
eta_sec = 0;
}
return eta_sec;
}
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);
if (td->o.verify_dump)
dump_buf(p, hdr_len, io_u->offset + hdr_num * hdr_len,
"hdr_fail", io_u->file);
return EILSEQ;
}
/*
* Sets the status of the 'td' in the printed status map.
*/
static void check_str_update(struct thread_data *td)
{
char c = __run_str[td->thread_number - 1];
switch (td->runstate) {
case TD_REAPED:
if (td->error)
c = 'X';
else if (td->sig)
c = 'K';
else
c = '_';
break;
case TD_EXITED:
c = 'E';
break;
case TD_RAMP:
c = '/';
break;
case TD_RUNNING:
if (td_rw(td)) {
if (td_random(td)) {
if (td->o.rwmix[DDIR_READ] == 100)
c = 'r';
else if (td->o.rwmix[DDIR_WRITE] == 100)
c = 'w';
else
c = 'm';
} else {
if (td->o.rwmix[DDIR_READ] == 100)
c = 'R';
else if (td->o.rwmix[DDIR_WRITE] == 100)
c = 'W';
else
c = 'M';
}
} else if (td_read(td)) {
if (td_random(td))
c = 'r';
else
c = 'R';
} else if (td_write(td)) {
if (td_random(td))
c = 'w';
else
c = 'W';
} else {
if (td_random(td))
c = 'd';
else
c = 'D';
}
break;
case TD_PRE_READING:
c = 'p';
break;
case TD_VERIFYING:
c = 'V';
break;
case TD_FSYNCING:
c = 'F';
break;
case TD_FINISHING:
c = 'f';
break;
case TD_CREATED:
c = 'C';
break;
case TD_INITIALIZED:
case TD_SETTING_UP:
c = 'I';
break;
case TD_NOT_CREATED:
c = 'P';
break;
default:
log_err("state %d\n", td->runstate);
}
__run_str[td->thread_number - 1] = c;
update_condensed_str(__run_str, run_str);
}
/*
* Best effort calculation of the estimated pending runtime of a job.
*/
static int thread_eta(struct thread_data *td)
{
unsigned long long bytes_total, bytes_done;
unsigned long eta_sec = 0;
unsigned long elapsed;
uint64_t timeout;
elapsed = (mtime_since_now(&td->epoch) + 999) / 1000;
timeout = td->o.timeout / 1000000UL;
bytes_total = td->total_io_size;
if (td->o.fill_device && td->o.size == -1ULL) {
if (!td->fill_device_size || td->fill_device_size == -1ULL)
return 0;
bytes_total = td->fill_device_size;
}
if (td->o.zone_size && td->o.zone_skip && bytes_total) {
unsigned int nr_zones;
uint64_t zone_bytes;
zone_bytes = bytes_total + td->o.zone_size + td->o.zone_skip;
nr_zones = (zone_bytes - 1) / (td->o.zone_size + td->o.zone_skip);
bytes_total -= nr_zones * td->o.zone_skip;
}
/*
* if writing and verifying afterwards, bytes_total will be twice the
* size. In a mixed workload, verify phase will be the size of the
* first stage writes.
*/
if (td->o.do_verify && td->o.verify && td_write(td)) {
if (td_rw(td)) {
unsigned int perc = 50;
if (td->o.rwmix[DDIR_WRITE])
perc = td->o.rwmix[DDIR_WRITE];
bytes_total += (bytes_total * perc) / 100;
} else
bytes_total <<= 1;
}
if (td->runstate == TD_RUNNING || td->runstate == TD_VERIFYING) {
double perc, perc_t;
bytes_done = ddir_rw_sum(td->io_bytes);
if (bytes_total) {
perc = (double) bytes_done / (double) bytes_total;
if (perc > 1.0)
perc = 1.0;
} else
perc = 0.0;
if (td->o.time_based) {
if (timeout) {
perc_t = (double) elapsed / (double) timeout;
if (perc_t < perc)
perc = perc_t;
} else {
/*
* Will never hit, we can't have time_based
* without a timeout set.
*/
perc = 0.0;
}
}
eta_sec = (unsigned long) (elapsed * (1.0 / perc)) - elapsed;
if (td->o.timeout &&
eta_sec > (timeout + done_secs - elapsed))
eta_sec = timeout + done_secs - elapsed;
} else if (td->runstate == TD_NOT_CREATED || td->runstate == TD_CREATED
|| td->runstate == TD_INITIALIZED
|| td->runstate == TD_SETTING_UP
|| td->runstate == TD_RAMP
|| td->runstate == TD_PRE_READING) {
int t_eta = 0, r_eta = 0;
unsigned long long rate_bytes;
/*
* We can only guess - assume it'll run the full timeout
* if given, otherwise assume it'll run at the specified rate.
*/
if (td->o.timeout) {
uint64_t timeout = td->o.timeout;
uint64_t start_delay = td->o.start_delay;
uint64_t ramp_time = td->o.ramp_time;
t_eta = timeout + start_delay + ramp_time;
t_eta /= 1000000ULL;
if (in_ramp_time(td)) {
unsigned long ramp_left;
ramp_left = mtime_since_now(&td->epoch);
ramp_left = (ramp_left + 999) / 1000;
if (ramp_left <= t_eta)
t_eta -= ramp_left;
}
}
rate_bytes = ddir_rw_sum(td->o.rate);
if (rate_bytes) {
r_eta = (bytes_total / 1024) / rate_bytes;
r_eta += (td->o.start_delay / 1000000ULL);
}
if (r_eta && t_eta)
eta_sec = min(r_eta, t_eta);
else if (r_eta)
eta_sec = r_eta;
else if (t_eta)
eta_sec = t_eta;
else
eta_sec = 0;
} else {
/*
* thread is already done or waiting for fsync
*/
eta_sec = 0;
}
return eta_sec;
}
static void io_completed(struct thread_data *td, struct io_u *io_u,
struct io_completion_data *icd)
{
struct fio_file *f;
dprint_io_u(io_u, "io complete");
td_io_u_lock(td);
assert(io_u->flags & IO_U_F_FLIGHT);
io_u->flags &= ~(IO_U_F_FLIGHT | IO_U_F_BUSY_OK);
td_io_u_unlock(td);
if (ddir_sync(io_u->ddir)) {
td->last_was_sync = 1;
f = io_u->file;
if (f) {
f->first_write = -1ULL;
f->last_write = -1ULL;
}
return;
}
td->last_was_sync = 0;
td->last_ddir = io_u->ddir;
if (!io_u->error && ddir_rw(io_u->ddir)) {
unsigned int bytes = io_u->buflen - io_u->resid;
const enum fio_ddir idx = io_u->ddir;
const enum fio_ddir odx = io_u->ddir ^ 1;
int ret;
td->io_blocks[idx]++;
td->this_io_blocks[idx]++;
td->io_bytes[idx] += bytes;
if (!(io_u->flags & IO_U_F_VER_LIST))
td->this_io_bytes[idx] += bytes;
if (idx == DDIR_WRITE) {
f = io_u->file;
if (f) {
if (f->first_write == -1ULL ||
io_u->offset < f->first_write)
f->first_write = io_u->offset;
if (f->last_write == -1ULL ||
((io_u->offset + bytes) > f->last_write))
f->last_write = io_u->offset + bytes;
}
}
if (ramp_time_over(td) && (td->runstate == TD_RUNNING ||
td->runstate == TD_VERIFYING)) {
account_io_completion(td, io_u, icd, idx, bytes);
if (__should_check_rate(td, idx)) {
td->rate_pending_usleep[idx] =
(usec_for_io(td, idx) -
utime_since_now(&td->start));
}
if (idx != DDIR_TRIM && __should_check_rate(td, odx))
td->rate_pending_usleep[odx] =
(usec_for_io(td, odx) -
utime_since_now(&td->start));
}
if (td_write(td) && idx == DDIR_WRITE &&
td->o.do_verify &&
td->o.verify != VERIFY_NONE &&
!td->o.experimental_verify)
log_io_piece(td, io_u);
icd->bytes_done[idx] += bytes;
if (io_u->end_io) {
ret = io_u->end_io(td, io_u);
if (ret && !icd->error)
icd->error = ret;
}
} else if (io_u->error) {
icd->error = io_u->error;
io_u_log_error(td, io_u);
}
if (icd->error) {
enum error_type_bit eb = td_error_type(io_u->ddir, icd->error);
if (!td_non_fatal_error(td, eb, icd->error))
return;
/*
* If there is a non_fatal error, then add to the error count
* and clear all the errors.
*/
update_error_count(td, icd->error);
td_clear_error(td);
icd->error = 0;
io_u->error = 0;
}
}
/* This functions mimics the generic_file_open function, but issues
IME native calls instead of POSIX calls. */
static int fio_ime_open_file(struct thread_data *td, struct fio_file *f)
{
int flags = 0;
int ret;
uint64_t desired_fs;
char *ime_filename;
dprint(FD_FILE, "fd open %s\n", f->file_name);
if (td_trim(td)) {
td_verror(td, EINVAL, "IME does not support TRIM operation");
return 1;
}
if (td->o.oatomic) {
td_verror(td, EINVAL, "IME does not support atomic IO");
return 1;
}
if (td->o.odirect)
flags |= O_DIRECT;
if (td->o.sync_io)
flags |= O_SYNC;
if (td->o.create_on_open && td->o.allow_create)
flags |= O_CREAT;
if (td_write(td)) {
if (!read_only)
flags |= O_RDWR;
if (td->o.allow_create)
flags |= O_CREAT;
} else if (td_read(td)) {
flags |= O_RDONLY;
} else {
/* We should never go here. */
td_verror(td, EINVAL, "Unsopported open mode");
return 1;
}
ime_filename = fio_set_ime_filename(f->file_name);
if (ime_filename == NULL)
return 1;
f->fd = ime_native_open(ime_filename, flags, 0600);
if (f->fd == -1) {
char buf[FIO_VERROR_SIZE];
int __e = errno;
snprintf(buf, sizeof(buf), "open(%s)", f->file_name);
td_verror(td, __e, buf);
return 1;
}
/* Now we need to make sure the real file size is sufficient for FIO
to do its things. This is normally done before the file open function
is called, but because FIO would use POSIX calls, we need to do it
ourselves */
ret = fio_ime_get_file_size(td, f);
if (ret < 0) {
ime_native_close(f->fd);
td_verror(td, errno, "ime_get_file_size");
return 1;
}
desired_fs = f->io_size + f->file_offset;
if (td_write(td)) {
dprint(FD_FILE, "Laying out file %s%s\n",
DEFAULT_IME_FILE_PREFIX, f->file_name);
if (!td->o.create_on_open &&
f->real_file_size < desired_fs &&
ime_native_ftruncate(f->fd, desired_fs) < 0) {
ime_native_close(f->fd);
td_verror(td, errno, "ime_native_ftruncate");
return 1;
}
if (f->real_file_size < desired_fs)
f->real_file_size = desired_fs;
} else if (td_read(td) && f->real_file_size < desired_fs) {
ime_native_close(f->fd);
log_err("error: can't read %lu bytes from file with "
"%lu bytes\n", desired_fs, f->real_file_size);
return 1;
}
return 0;
}
/*
* Print status of the jobs we know about. This includes rate estimates,
* ETA, thread state, etc.
*/
int calc_thread_status(struct jobs_eta *je, int force)
{
struct thread_data *td;
int i;
unsigned long rate_time, disp_time, bw_avg_time, *eta_secs;
unsigned long long io_bytes[DDIR_RWDIR_CNT];
unsigned long long io_iops[DDIR_RWDIR_CNT];
struct timeval now;
static unsigned long long rate_io_bytes[DDIR_RWDIR_CNT];
static unsigned long long disp_io_bytes[DDIR_RWDIR_CNT];
static unsigned long long disp_io_iops[DDIR_RWDIR_CNT];
static struct timeval rate_prev_time, disp_prev_time;
if (!force) {
if (output_format != FIO_OUTPUT_NORMAL)
return 0;
if (temp_stall_ts || eta_print == FIO_ETA_NEVER)
return 0;
if (!isatty(STDOUT_FILENO) && (eta_print != FIO_ETA_ALWAYS))
return 0;
}
if (!ddir_rw_sum(rate_io_bytes))
fill_start_time(&rate_prev_time);
if (!ddir_rw_sum(disp_io_bytes))
fill_start_time(&disp_prev_time);
eta_secs = malloc(thread_number * sizeof(unsigned long));
memset(eta_secs, 0, thread_number * sizeof(unsigned long));
je->elapsed_sec = (mtime_since_genesis() + 999) / 1000;
io_bytes[DDIR_READ] = io_bytes[DDIR_WRITE] = io_bytes[DDIR_TRIM] = 0;
io_iops[DDIR_READ] = io_iops[DDIR_WRITE] = io_iops[DDIR_TRIM] = 0;
bw_avg_time = ULONG_MAX;
for_each_td(td, i) {
if (is_power_of_2(td->o.kb_base))
je->is_pow2 = 1;
if (td->o.bw_avg_time < bw_avg_time)
bw_avg_time = td->o.bw_avg_time;
if (td->runstate == TD_RUNNING || td->runstate == TD_VERIFYING
|| td->runstate == TD_FSYNCING
|| td->runstate == TD_PRE_READING) {
je->nr_running++;
if (td_read(td)) {
je->t_rate += td->o.rate[DDIR_READ];
je->t_iops += td->o.rate_iops[DDIR_READ];
je->m_rate += td->o.ratemin[DDIR_READ];
je->m_iops += td->o.rate_iops_min[DDIR_READ];
}
if (td_write(td)) {
je->t_rate += td->o.rate[DDIR_WRITE];
je->t_iops += td->o.rate_iops[DDIR_WRITE];
je->m_rate += td->o.ratemin[DDIR_WRITE];
je->m_iops += td->o.rate_iops_min[DDIR_WRITE];
}
if (td_trim(td)) {
je->t_rate += td->o.rate[DDIR_TRIM];
je->t_iops += td->o.rate_iops[DDIR_TRIM];
je->m_rate += td->o.ratemin[DDIR_TRIM];
je->m_iops += td->o.rate_iops_min[DDIR_TRIM];
}
je->files_open += td->nr_open_files;
} else if (td->runstate == TD_RAMP) {
je->nr_running++;
je->nr_ramp++;
} else if (td->runstate == TD_SETTING_UP)
je->nr_running++;
else if (td->runstate < TD_RUNNING)
je->nr_pending++;
if (je->elapsed_sec >= 3)
eta_secs[i] = thread_eta(td);
else
eta_secs[i] = INT_MAX;
check_str_update(td);
if (td->runstate > TD_RAMP) {
int ddir;
for (ddir = DDIR_READ; ddir < DDIR_RWDIR_CNT; ddir++) {
io_bytes[ddir] += td->io_bytes[ddir];
io_iops[ddir] += td->io_blocks[ddir];
}
}
}
if (exitall_on_terminate)
je->eta_sec = INT_MAX;
else
je->eta_sec = 0;
for_each_td(td, i) {
if (exitall_on_terminate) {
if (eta_secs[i] < je->eta_sec)
je->eta_sec = eta_secs[i];
} else {
if (eta_secs[i] > je->eta_sec)
je->eta_sec = eta_secs[i];
}
}
free(eta_secs);
fio_gettime(&now, NULL);
rate_time = mtime_since(&rate_prev_time, &now);
if (write_bw_log && rate_time > bw_avg_time && !in_ramp_time(td)) {
calc_rate(rate_time, io_bytes, rate_io_bytes, je->rate);
memcpy(&rate_prev_time, &now, sizeof(now));
add_agg_sample(je->rate[DDIR_READ], DDIR_READ, 0);
add_agg_sample(je->rate[DDIR_WRITE], DDIR_WRITE, 0);
add_agg_sample(je->rate[DDIR_TRIM], DDIR_TRIM, 0);
}
disp_time = mtime_since(&disp_prev_time, &now);
/*
* Allow a little slack, the target is to print it every 1000 msecs
*/
if (!force && disp_time < 900)
return 0;
calc_rate(disp_time, io_bytes, disp_io_bytes, je->rate);
calc_iops(disp_time, io_iops, disp_io_iops, je->iops);
memcpy(&disp_prev_time, &now, sizeof(now));
if (!force && !je->nr_running && !je->nr_pending)
return 0;
je->nr_threads = thread_number;
memcpy(je->run_str, run_str, thread_number * sizeof(char));
return 1;
}
int generic_open_file(struct thread_data *td, struct fio_file *f)
{
int is_std = 0;
int flags = 0;
int from_hash = 0;
dprint(FD_FILE, "fd open %s\n", f->file_name);
if (!strcmp(f->file_name, "-")) {
if (td_rw(td)) {
log_err("fio: can't read/write to stdin/out\n");
return 1;
}
is_std = 1;
/*
* move output logging to stderr, if we are writing to stdout
*/
if (td_write(td))
f_out = stderr;
}
if (td->o.odirect)
flags |= OS_O_DIRECT;
if (td->o.sync_io)
flags |= O_SYNC;
if (f->filetype != FIO_TYPE_FILE)
flags |= FIO_O_NOATIME;
if (td->o.create_on_open)
flags |= O_CREAT;
open_again:
if (td_write(td)) {
if (!read_only)
flags |= O_RDWR;
if (f->filetype == FIO_TYPE_FILE)
flags |= O_CREAT;
if (is_std)
f->fd = dup(STDOUT_FILENO);
else
from_hash = file_lookup_open(f, flags);
} else {
if (f->filetype == FIO_TYPE_CHAR && !read_only)
flags |= O_RDWR;
else
flags |= O_RDONLY;
if (is_std)
f->fd = dup(STDIN_FILENO);
else
from_hash = file_lookup_open(f, flags);
}
if (f->fd == -1) {
char buf[FIO_VERROR_SIZE];
int __e = errno;
if (__e == EPERM && (flags & FIO_O_NOATIME)) {
flags &= ~FIO_O_NOATIME;
goto open_again;
}
snprintf(buf, sizeof(buf) - 1, "open(%s)", f->file_name);
td_verror(td, __e, buf);
}
if (!from_hash && f->fd != -1) {
if (add_file_hash(f)) {
int fio_unused ret;
/*
* OK to ignore, we haven't done anything with it
*/
ret = generic_close_file(td, f);
goto open_again;
}
}
return 0;
}
/*
* Leaves f->fd open on success, caller must close
*/
static int extend_file(struct thread_data *td, struct fio_file *f)
{
int r, new_layout = 0, unlink_file = 0, flags;
unsigned long long left;
unsigned int bs;
char *b;
if (read_only) {
log_err("fio: refusing extend of file due to read-only\n");
return 0;
}
/*
* check if we need to lay the file out complete again. fio
* does that for operations involving reads, or for writes
* where overwrite is set
*/
if (td_read(td) || (td_write(td) && td->o.overwrite) ||
(td_write(td) && td->io_ops->flags & FIO_NOEXTEND))
new_layout = 1;
if (td_write(td) && !td->o.overwrite)
unlink_file = 1;
if (unlink_file || new_layout) {
dprint(FD_FILE, "layout unlink %s\n", f->file_name);
if ((unlink(f->file_name) < 0) && (errno != ENOENT)) {
td_verror(td, errno, "unlink");
return 1;
}
}
flags = O_WRONLY | O_CREAT;
if (new_layout)
flags |= O_TRUNC;
dprint(FD_FILE, "open file %s, flags %x\n", f->file_name, flags);
f->fd = open(f->file_name, flags, 0644);
if (f->fd < 0) {
td_verror(td, errno, "open");
return 1;
}
#ifdef CONFIG_POSIX_FALLOCATE
if (!td->o.fill_device) {
switch (td->o.fallocate_mode) {
case FIO_FALLOCATE_NONE:
break;
case FIO_FALLOCATE_POSIX:
dprint(FD_FILE, "posix_fallocate file %s size %llu\n",
f->file_name,
(unsigned long long) f->real_file_size);
r = posix_fallocate(f->fd, 0, f->real_file_size);
if (r > 0) {
log_err("fio: posix_fallocate fails: %s\n",
strerror(r));
}
break;
#ifdef CONFIG_LINUX_FALLOCATE
case FIO_FALLOCATE_KEEP_SIZE:
dprint(FD_FILE,
"fallocate(FALLOC_FL_KEEP_SIZE) "
"file %s size %llu\n", f->file_name,
(unsigned long long) f->real_file_size);
r = fallocate(f->fd, FALLOC_FL_KEEP_SIZE, 0,
f->real_file_size);
if (r != 0)
td_verror(td, errno, "fallocate");
break;
#endif /* CONFIG_LINUX_FALLOCATE */
default:
log_err("fio: unknown fallocate mode: %d\n",
td->o.fallocate_mode);
assert(0);
}
}
#endif /* CONFIG_POSIX_FALLOCATE */
if (!new_layout)
goto done;
/*
* The size will be -1ULL when fill_device is used, so don't truncate
* or fallocate this file, just write it
*/
if (!td->o.fill_device) {
dprint(FD_FILE, "truncate file %s, size %llu\n", f->file_name,
(unsigned long long) f->real_file_size);
if (ftruncate(f->fd, f->real_file_size) == -1) {
td_verror(td, errno, "ftruncate");
goto err;
}
}
b = malloc(td->o.max_bs[DDIR_WRITE]);
left = f->real_file_size;
while (left && !td->terminate) {
bs = td->o.max_bs[DDIR_WRITE];
if (bs > left)
bs = left;
fill_io_buffer(td, b, bs, bs);
r = write(f->fd, b, bs);
if (r > 0) {
left -= r;
continue;
} else {
if (r < 0) {
int __e = errno;
if (__e == ENOSPC) {
if (td->o.fill_device)
break;
log_info("fio: ENOSPC on laying out "
"file, stopping\n");
break;
}
td_verror(td, errno, "write");
} else
td_verror(td, EIO, "write");
break;
}
}
if (td->terminate) {
dprint(FD_FILE, "terminate unlink %s\n", f->file_name);
unlink(f->file_name);
} else if (td->o.create_fsync) {
if (fsync(f->fd) < 0) {
td_verror(td, errno, "fsync");
goto err;
}
}
if (td->o.fill_device && !td_write(td)) {
fio_file_clear_size_known(f);
if (td_io_get_file_size(td, f))
goto err;
if (f->io_size > f->real_file_size)
f->io_size = f->real_file_size;
}
free(b);
done:
return 0;
err:
close(f->fd);
f->fd = -1;
return 1;
}
/*
* Print status of the jobs we know about. This includes rate estimates,
* ETA, thread state, etc.
*/
int calc_thread_status(struct jobs_eta *je, int force)
{
struct thread_data *td;
int i, unified_rw_rep;
unsigned long rate_time, disp_time, bw_avg_time, *eta_secs;
unsigned long long io_bytes[DDIR_RWDIR_CNT];
unsigned long long io_iops[DDIR_RWDIR_CNT];
struct timeval now;
static unsigned long long rate_io_bytes[DDIR_RWDIR_CNT];
static unsigned long long disp_io_bytes[DDIR_RWDIR_CNT];
static unsigned long long disp_io_iops[DDIR_RWDIR_CNT];
static struct timeval rate_prev_time, disp_prev_time;
if (!force) {
if (output_format != FIO_OUTPUT_NORMAL &&
f_out == stdout)
return 0;
if (temp_stall_ts || eta_print == FIO_ETA_NEVER)
return 0;
if (!isatty(STDOUT_FILENO) && (eta_print != FIO_ETA_ALWAYS))
return 0;
}
if (!ddir_rw_sum(rate_io_bytes))
fill_start_time(&rate_prev_time);
if (!ddir_rw_sum(disp_io_bytes))
fill_start_time(&disp_prev_time);
eta_secs = malloc(thread_number * sizeof(unsigned long));
memset(eta_secs, 0, thread_number * sizeof(unsigned long));
je->elapsed_sec = (mtime_since_genesis() + 999) / 1000;
io_bytes[DDIR_READ] = io_bytes[DDIR_WRITE] = io_bytes[DDIR_TRIM] = 0;
io_iops[DDIR_READ] = io_iops[DDIR_WRITE] = io_iops[DDIR_TRIM] = 0;
bw_avg_time = ULONG_MAX;
unified_rw_rep = 0;
for_each_td(td, i) {
unified_rw_rep += td->o.unified_rw_rep;
if (is_power_of_2(td->o.kb_base))
je->is_pow2 = 1;
je->unit_base = td->o.unit_base;
if (td->o.bw_avg_time < bw_avg_time)
bw_avg_time = td->o.bw_avg_time;
if (td->runstate == TD_RUNNING || td->runstate == TD_VERIFYING
|| td->runstate == TD_FSYNCING
|| td->runstate == TD_PRE_READING
|| td->runstate == TD_FINISHING) {
je->nr_running++;
if (td_read(td)) {
je->t_rate[0] += td->o.rate[DDIR_READ];
je->t_iops[0] += td->o.rate_iops[DDIR_READ];
je->m_rate[0] += td->o.ratemin[DDIR_READ];
je->m_iops[0] += td->o.rate_iops_min[DDIR_READ];
}
if (td_write(td)) {
je->t_rate[1] += td->o.rate[DDIR_WRITE];
je->t_iops[1] += td->o.rate_iops[DDIR_WRITE];
je->m_rate[1] += td->o.ratemin[DDIR_WRITE];
je->m_iops[1] += td->o.rate_iops_min[DDIR_WRITE];
}
if (td_trim(td)) {
je->t_rate[2] += td->o.rate[DDIR_TRIM];
je->t_iops[2] += td->o.rate_iops[DDIR_TRIM];
je->m_rate[2] += td->o.ratemin[DDIR_TRIM];
je->m_iops[2] += td->o.rate_iops_min[DDIR_TRIM];
}
je->files_open += td->nr_open_files;
} else if (td->runstate == TD_RAMP) {
je->nr_running++;
je->nr_ramp++;
} else if (td->runstate == TD_SETTING_UP) {
je->nr_running++;
je->nr_setting_up++;
} else if (td->runstate < TD_RUNNING)
je->nr_pending++;
if (je->elapsed_sec >= 3)
eta_secs[i] = thread_eta(td);
else
eta_secs[i] = INT_MAX;
check_str_update(td);
if (td->runstate > TD_SETTING_UP) {
int ddir;
for (ddir = DDIR_READ; ddir < DDIR_RWDIR_CNT; ddir++) {
if (unified_rw_rep) {
io_bytes[0] += td->io_bytes[ddir];
io_iops[0] += td->io_blocks[ddir];
} else {
io_bytes[ddir] += td->io_bytes[ddir];
io_iops[ddir] += td->io_blocks[ddir];
}
}
}
}
int generic_open_file(struct thread_data *td, struct fio_file *f)
{
int is_std = 0;
int flags = 0;
int from_hash = 0;
dprint(FD_FILE, "fd open %s\n", f->file_name);
if (td_trim(td) && f->filetype != FIO_TYPE_BD) {
log_err("fio: trim only applies to block device\n");
return 1;
}
if (!strcmp(f->file_name, "-")) {
if (td_rw(td)) {
log_err("fio: can't read/write to stdin/out\n");
return 1;
}
is_std = 1;
/*
* move output logging to stderr, if we are writing to stdout
*/
if (td_write(td))
f_out = stderr;
}
if (td_trim(td))
goto skip_flags;
if (td->o.odirect)
flags |= OS_O_DIRECT;
if (td->o.sync_io)
flags |= O_SYNC;
if (td->o.create_on_open)
flags |= O_CREAT;
skip_flags:
if (f->filetype != FIO_TYPE_FILE)
flags |= FIO_O_NOATIME;
open_again:
if (td_write(td)) {
if (!read_only)
flags |= O_RDWR;
if (f->filetype == FIO_TYPE_FILE)
flags |= O_CREAT;
if (is_std)
f->fd = dup(STDOUT_FILENO);
else
from_hash = file_lookup_open(f, flags);
} else if (td_read(td)) {
if (f->filetype == FIO_TYPE_CHAR && !read_only)
flags |= O_RDWR;
else
flags |= O_RDONLY;
if (is_std)
f->fd = dup(STDIN_FILENO);
else
from_hash = file_lookup_open(f, flags);
} else { //td trim
flags |= O_RDWR;
from_hash = file_lookup_open(f, flags);
}
if (f->fd == -1) {
char buf[FIO_VERROR_SIZE];
int __e = errno;
if (__e == EPERM && (flags & FIO_O_NOATIME)) {
flags &= ~FIO_O_NOATIME;
goto open_again;
}
if (__e == EMFILE && file_close_shadow_fds(td))
goto open_again;
snprintf(buf, sizeof(buf), "open(%s)", f->file_name);
if (__e == EINVAL && (flags & OS_O_DIRECT)) {
log_err("fio: looks like your file system does not " \
"support direct=1/buffered=0\n");
}
td_verror(td, __e, buf);
}
if (!from_hash && f->fd != -1) {
if (add_file_hash(f)) {
int fio_unused ret;
/*
* Stash away descriptor for later close. This is to
* work-around a "feature" on Linux, where a close of
* an fd that has been opened for write will trigger
* udev to call blkid to check partitions, fs id, etc.
* That polutes the device cache, which can slow down
* unbuffered accesses.
*/
if (f->shadow_fd == -1)
f->shadow_fd = f->fd;
else {
/*
* OK to ignore, we haven't done anything
* with it
*/
ret = generic_close_file(td, f);
}
goto open_again;
}
}
return 0;
}
/*
* Lazy way of fixing up options that depend on each other. We could also
* define option callback handlers, but this is easier.
*/
static int fixup_options(struct thread_data *td)
{
struct thread_options *o = &td->o;
int ret = 0;
#ifndef FIO_HAVE_PSHARED_MUTEX
if (!o->use_thread) {
log_info("fio: this platform does not support process shared"
" mutexes, forcing use of threads. Use the 'thread'"
" option to get rid of this warning.\n");
o->use_thread = 1;
ret = warnings_fatal;
}
#endif
if (o->write_iolog_file && o->read_iolog_file) {
log_err("fio: read iolog overrides write_iolog\n");
free(o->write_iolog_file);
o->write_iolog_file = NULL;
ret = warnings_fatal;
}
/*
* only really works with 1 file
*/
if (o->zone_size && o->open_files == 1)
o->zone_size = 0;
/*
* If zone_range isn't specified, backward compatibility dictates it
* should be made equal to zone_size.
*/
if (o->zone_size && !o->zone_range)
o->zone_range = o->zone_size;
/*
* Reads can do overwrites, we always need to pre-create the file
*/
if (td_read(td) || td_rw(td))
o->overwrite = 1;
if (!o->min_bs[DDIR_READ])
o->min_bs[DDIR_READ] = o->bs[DDIR_READ];
if (!o->max_bs[DDIR_READ])
o->max_bs[DDIR_READ] = o->bs[DDIR_READ];
if (!o->min_bs[DDIR_WRITE])
o->min_bs[DDIR_WRITE] = o->bs[DDIR_WRITE];
if (!o->max_bs[DDIR_WRITE])
o->max_bs[DDIR_WRITE] = o->bs[DDIR_WRITE];
o->rw_min_bs = min(o->min_bs[DDIR_READ], o->min_bs[DDIR_WRITE]);
/*
* For random IO, allow blockalign offset other than min_bs.
*/
if (!o->ba[DDIR_READ] || !td_random(td))
o->ba[DDIR_READ] = o->min_bs[DDIR_READ];
if (!o->ba[DDIR_WRITE] || !td_random(td))
o->ba[DDIR_WRITE] = o->min_bs[DDIR_WRITE];
if ((o->ba[DDIR_READ] != o->min_bs[DDIR_READ] ||
o->ba[DDIR_WRITE] != o->min_bs[DDIR_WRITE]) &&
!o->norandommap) {
log_err("fio: Any use of blockalign= turns off randommap\n");
o->norandommap = 1;
ret = warnings_fatal;
}
if (!o->file_size_high)
o->file_size_high = o->file_size_low;
if (o->norandommap && o->verify != VERIFY_NONE
&& !fixed_block_size(o)) {
log_err("fio: norandommap given for variable block sizes, "
"verify disabled\n");
o->verify = VERIFY_NONE;
ret = warnings_fatal;
}
if (o->bs_unaligned && (o->odirect || td->io_ops->flags & FIO_RAWIO))
log_err("fio: bs_unaligned may not work with raw io\n");
/*
* thinktime_spin must be less than thinktime
*/
if (o->thinktime_spin > o->thinktime)
o->thinktime_spin = o->thinktime;
/*
* The low water mark cannot be bigger than the iodepth
*/
if (o->iodepth_low > o->iodepth || !o->iodepth_low) {
/*
* syslet work around - if the workload is sequential,
* we want to let the queue drain all the way down to
* avoid seeking between async threads
*/
if (!strcmp(td->io_ops->name, "syslet-rw") && !td_random(td))
o->iodepth_low = 1;
else
o->iodepth_low = o->iodepth;
}
/*
* If batch number isn't set, default to the same as iodepth
*/
if (o->iodepth_batch > o->iodepth || !o->iodepth_batch)
o->iodepth_batch = o->iodepth;
if (o->nr_files > td->files_index)
o->nr_files = td->files_index;
if (o->open_files > o->nr_files || !o->open_files)
o->open_files = o->nr_files;
if (((o->rate[0] + o->rate[1]) && (o->rate_iops[0] + o->rate_iops[1]))||
((o->ratemin[0] + o->ratemin[1]) && (o->rate_iops_min[0] +
o->rate_iops_min[1]))) {
log_err("fio: rate and rate_iops are mutually exclusive\n");
ret = 1;
}
if ((o->rate[0] < o->ratemin[0]) || (o->rate[1] < o->ratemin[1]) ||
(o->rate_iops[0] < o->rate_iops_min[0]) ||
(o->rate_iops[1] < o->rate_iops_min[1])) {
log_err("fio: minimum rate exceeds rate\n");
ret = 1;
}
if (!o->timeout && o->time_based) {
log_err("fio: time_based requires a runtime/timeout setting\n");
o->time_based = 0;
ret = warnings_fatal;
}
if (o->fill_device && !o->size)
o->size = -1ULL;
if (o->verify != VERIFY_NONE) {
if (td_write(td) && o->do_verify && o->numjobs > 1) {
log_info("Multiple writers may overwrite blocks that "
"belong to other jobs. This can cause "
"verification failures.\n");
ret = warnings_fatal;
}
o->refill_buffers = 1;
if (o->max_bs[DDIR_WRITE] != o->min_bs[DDIR_WRITE] &&
!o->verify_interval)
o->verify_interval = o->min_bs[DDIR_WRITE];
}
if (o->pre_read) {
o->invalidate_cache = 0;
if (td->io_ops->flags & FIO_PIPEIO) {
log_info("fio: cannot pre-read files with an IO engine"
" that isn't seekable. Pre-read disabled.\n");
ret = warnings_fatal;
}
}
#ifndef FIO_HAVE_FDATASYNC
if (o->fdatasync_blocks) {
log_info("fio: this platform does not support fdatasync()"
" falling back to using fsync(). Use the 'fsync'"
" option instead of 'fdatasync' to get rid of"
" this warning\n");
o->fsync_blocks = o->fdatasync_blocks;
o->fdatasync_blocks = 0;
ret = warnings_fatal;
}
#endif
#ifdef WIN32
/*
* Windows doesn't support O_DIRECT or O_SYNC with the _open interface,
* so fail if we're passed those flags
*/
if ((td->io_ops->flags & FIO_SYNCIO) && (td->o.odirect || td->o.sync_io)) {
log_err("fio: Windows does not support direct or non-buffered io with"
" the synchronous ioengines. Use the 'windowsaio' ioengine"
" with 'direct=1' and 'iodepth=1' instead.\n");
ret = 1;
}
#endif
/*
* For fully compressible data, just zero them at init time.
* It's faster than repeatedly filling it.
*/
if (td->o.compress_percentage == 100) {
td->o.zero_buffers = 1;
td->o.compress_percentage = 0;
}
return ret;
}
/*
* Lazy way of fixing up options that depend on each other. We could also
* define option callback handlers, but this is easier.
*/
static int fixup_options(struct thread_data *td)
{
struct thread_options *o = &td->o;
int ret = 0;
#ifndef FIO_HAVE_PSHARED_MUTEX
if (!o->use_thread) {
log_info("fio: this platform does not support process shared"
" mutexes, forcing use of threads. Use the 'thread'"
" option to get rid of this warning.\n");
o->use_thread = 1;
ret = warnings_fatal;
}
#endif
if (o->write_iolog_file && o->read_iolog_file) {
log_err("fio: read iolog overrides write_iolog\n");
free(o->write_iolog_file);
o->write_iolog_file = NULL;
ret = warnings_fatal;
}
/*
* only really works with 1 file
*/
if (o->zone_size && o->open_files > 1)
o->zone_size = 0;
/*
* If zone_range isn't specified, backward compatibility dictates it
* should be made equal to zone_size.
*/
if (o->zone_size && !o->zone_range)
o->zone_range = o->zone_size;
/*
* Reads can do overwrites, we always need to pre-create the file
*/
if (td_read(td) || td_rw(td))
o->overwrite = 1;
if (!o->min_bs[DDIR_READ])
o->min_bs[DDIR_READ] = o->bs[DDIR_READ];
if (!o->max_bs[DDIR_READ])
o->max_bs[DDIR_READ] = o->bs[DDIR_READ];
if (!o->min_bs[DDIR_WRITE])
o->min_bs[DDIR_WRITE] = o->bs[DDIR_WRITE];
if (!o->max_bs[DDIR_WRITE])
o->max_bs[DDIR_WRITE] = o->bs[DDIR_WRITE];
if (!o->min_bs[DDIR_TRIM])
o->min_bs[DDIR_TRIM] = o->bs[DDIR_TRIM];
if (!o->max_bs[DDIR_TRIM])
o->max_bs[DDIR_TRIM] = o->bs[DDIR_TRIM];
o->rw_min_bs = min(o->min_bs[DDIR_READ], o->min_bs[DDIR_WRITE]);
o->rw_min_bs = min(o->min_bs[DDIR_TRIM], o->rw_min_bs);
/*
* For random IO, allow blockalign offset other than min_bs.
*/
if (!o->ba[DDIR_READ] || !td_random(td))
o->ba[DDIR_READ] = o->min_bs[DDIR_READ];
if (!o->ba[DDIR_WRITE] || !td_random(td))
o->ba[DDIR_WRITE] = o->min_bs[DDIR_WRITE];
if (!o->ba[DDIR_TRIM] || !td_random(td))
o->ba[DDIR_TRIM] = o->min_bs[DDIR_TRIM];
if ((o->ba[DDIR_READ] != o->min_bs[DDIR_READ] ||
o->ba[DDIR_WRITE] != o->min_bs[DDIR_WRITE] ||
o->ba[DDIR_TRIM] != o->min_bs[DDIR_TRIM]) &&
!o->norandommap) {
log_err("fio: Any use of blockalign= turns off randommap\n");
o->norandommap = 1;
ret = warnings_fatal;
}
if (!o->file_size_high)
o->file_size_high = o->file_size_low;
if (o->start_delay_high)
o->start_delay = get_rand_start_delay(td);
if (o->norandommap && o->verify != VERIFY_NONE
&& !fixed_block_size(o)) {
log_err("fio: norandommap given for variable block sizes, "
"verify disabled\n");
o->verify = VERIFY_NONE;
ret = warnings_fatal;
}
if (o->bs_unaligned && (o->odirect || td->io_ops->flags & FIO_RAWIO))
log_err("fio: bs_unaligned may not work with raw io\n");
/*
* thinktime_spin must be less than thinktime
*/
if (o->thinktime_spin > o->thinktime)
o->thinktime_spin = o->thinktime;
/*
* The low water mark cannot be bigger than the iodepth
*/
if (o->iodepth_low > o->iodepth || !o->iodepth_low)
o->iodepth_low = o->iodepth;
/*
* If batch number isn't set, default to the same as iodepth
*/
if (o->iodepth_batch > o->iodepth || !o->iodepth_batch)
o->iodepth_batch = o->iodepth;
if (o->nr_files > td->files_index)
o->nr_files = td->files_index;
if (o->open_files > o->nr_files || !o->open_files)
o->open_files = o->nr_files;
if (((o->rate[DDIR_READ] + o->rate[DDIR_WRITE] + o->rate[DDIR_TRIM]) &&
(o->rate_iops[DDIR_READ] + o->rate_iops[DDIR_WRITE] + o->rate_iops[DDIR_TRIM])) ||
((o->ratemin[DDIR_READ] + o->ratemin[DDIR_WRITE] + o->ratemin[DDIR_TRIM]) &&
(o->rate_iops_min[DDIR_READ] + o->rate_iops_min[DDIR_WRITE] + o->rate_iops_min[DDIR_TRIM]))) {
log_err("fio: rate and rate_iops are mutually exclusive\n");
ret = 1;
}
if ((o->rate[DDIR_READ] < o->ratemin[DDIR_READ]) ||
(o->rate[DDIR_WRITE] < o->ratemin[DDIR_WRITE]) ||
(o->rate[DDIR_TRIM] < o->ratemin[DDIR_TRIM]) ||
(o->rate_iops[DDIR_READ] < o->rate_iops_min[DDIR_READ]) ||
(o->rate_iops[DDIR_WRITE] < o->rate_iops_min[DDIR_WRITE]) ||
(o->rate_iops[DDIR_TRIM] < o->rate_iops_min[DDIR_TRIM])) {
log_err("fio: minimum rate exceeds rate\n");
ret = 1;
}
if (!o->timeout && o->time_based) {
log_err("fio: time_based requires a runtime/timeout setting\n");
o->time_based = 0;
ret = warnings_fatal;
}
if (o->fill_device && !o->size)
o->size = -1ULL;
if (o->verify != VERIFY_NONE) {
if (td_write(td) && o->do_verify && o->numjobs > 1) {
log_info("Multiple writers may overwrite blocks that "
"belong to other jobs. This can cause "
"verification failures.\n");
ret = warnings_fatal;
}
o->refill_buffers = 1;
if (o->max_bs[DDIR_WRITE] != o->min_bs[DDIR_WRITE] &&
!o->verify_interval)
o->verify_interval = o->min_bs[DDIR_WRITE];
/*
* Verify interval must be smaller or equal to the
* write size.
*/
if (o->verify_interval > o->min_bs[DDIR_WRITE])
o->verify_interval = o->min_bs[DDIR_WRITE];
else if (td_read(td) && o->verify_interval > o->min_bs[DDIR_READ])
o->verify_interval = o->min_bs[DDIR_READ];
}
if (o->pre_read) {
o->invalidate_cache = 0;
if (td->io_ops->flags & FIO_PIPEIO) {
log_info("fio: cannot pre-read files with an IO engine"
" that isn't seekable. Pre-read disabled.\n");
ret = warnings_fatal;
}
}
if (!o->unit_base) {
if (td->io_ops->flags & FIO_BIT_BASED)
o->unit_base = 1;
else
o->unit_base = 8;
}
#ifndef CONFIG_FDATASYNC
if (o->fdatasync_blocks) {
log_info("fio: this platform does not support fdatasync()"
" falling back to using fsync(). Use the 'fsync'"
" option instead of 'fdatasync' to get rid of"
" this warning\n");
o->fsync_blocks = o->fdatasync_blocks;
o->fdatasync_blocks = 0;
ret = warnings_fatal;
}
#endif
#ifdef WIN32
/*
* Windows doesn't support O_DIRECT or O_SYNC with the _open interface,
* so fail if we're passed those flags
*/
if ((td->io_ops->flags & FIO_SYNCIO) && (td->o.odirect || td->o.sync_io)) {
log_err("fio: Windows does not support direct or non-buffered io with"
" the synchronous ioengines. Use the 'windowsaio' ioengine"
" with 'direct=1' and 'iodepth=1' instead.\n");
ret = 1;
}
#endif
/*
* For fully compressible data, just zero them at init time.
* It's faster than repeatedly filling it.
*/
if (td->o.compress_percentage == 100) {
td->o.zero_buffers = 1;
td->o.compress_percentage = 0;
}
/*
* Using a non-uniform random distribution excludes usage of
* a random map
*/
if (td->o.random_distribution != FIO_RAND_DIST_RANDOM)
td->o.norandommap = 1;
/*
* If size is set but less than the min block size, complain
*/
if (o->size && o->size < td_min_bs(td)) {
log_err("fio: size too small, must be larger than the IO size: %llu\n", (unsigned long long) o->size);
ret = 1;
}
/*
* O_ATOMIC implies O_DIRECT
*/
if (td->o.oatomic)
td->o.odirect = 1;
/*
* If randseed is set, that overrides randrepeat
*/
if (td->o.rand_seed)
td->o.rand_repeatable = 0;
if ((td->io_ops->flags & FIO_NOEXTEND) && td->o.file_append) {
log_err("fio: can't append/extent with IO engine %s\n", td->io_ops->name);
ret = 1;
}
return ret;
}
