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)) && (td_min_bs(td) == td_max_bs(td)) &&
!td->o.time_based)
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;
}
/*
* 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];
}
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;
}
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;
}
/*
* Load a blktrace file by reading all the blk_io_trace entries, and storing
* them as io_pieces like the fio text version would do.
*/
int load_blktrace(struct thread_data *td, const char *filename, int need_swap)
{
struct blk_io_trace t;
unsigned long ios[DDIR_RWDIR_CNT], skipped_writes;
unsigned int rw_bs[DDIR_RWDIR_CNT];
struct fifo *fifo;
int fd, i, old_state;
struct fio_file *f;
int this_depth[DDIR_RWDIR_CNT], depth[DDIR_RWDIR_CNT], max_depth;
fd = open(filename, O_RDONLY);
if (fd < 0) {
td_verror(td, errno, "open blktrace file");
return 1;
}
fifo = fifo_alloc(TRACE_FIFO_SIZE);
old_state = td_bump_runstate(td, TD_SETTING_UP);
td->o.size = 0;
for (i = 0; i < DDIR_RWDIR_CNT; i++) {
ios[i] = 0;
rw_bs[i] = 0;
this_depth[i] = 0;
depth[i] = 0;
}
skipped_writes = 0;
do {
int ret = trace_fifo_get(td, fifo, fd, &t, sizeof(t));
if (ret < 0)
goto err;
else if (!ret)
break;
else if (ret < (int) sizeof(t)) {
log_err("fio: short fifo get\n");
break;
}
if (need_swap)
byteswap_trace(&t);
if ((t.magic & 0xffffff00) != BLK_IO_TRACE_MAGIC) {
log_err("fio: bad magic in blktrace data: %x\n",
t.magic);
goto err;
}
if ((t.magic & 0xff) != BLK_IO_TRACE_VERSION) {
log_err("fio: bad blktrace version %d\n",
t.magic & 0xff);
goto err;
}
ret = discard_pdu(td, fifo, fd, &t);
if (ret < 0) {
td_verror(td, ret, "blktrace lseek");
goto err;
} else if (t.pdu_len != ret) {
log_err("fio: discarded %d of %d\n", ret, t.pdu_len);
goto err;
}
if ((t.action & BLK_TC_ACT(BLK_TC_NOTIFY)) == 0) {
if ((t.action & 0xffff) == __BLK_TA_QUEUE)
depth_inc(&t, this_depth);
else if (((t.action & 0xffff) == __BLK_TA_BACKMERGE) ||
((t.action & 0xffff) == __BLK_TA_FRONTMERGE))
depth_dec(&t, this_depth);
else if ((t.action & 0xffff) == __BLK_TA_COMPLETE)
depth_end(&t, this_depth, depth);
if (t_is_write(&t) && read_only) {
skipped_writes++;
continue;
}
}
handle_trace(td, &t, ios, rw_bs);
} while (1);
for (i = 0; i < td->files_index; i++) {
f = td->files[i];
trace_add_open_close_event(td, f->fileno, FIO_LOG_CLOSE_FILE);
}
fifo_free(fifo);
close(fd);
td_restore_runstate(td, old_state);
if (!td->files_index) {
log_err("fio: did not find replay device(s)\n");
return 1;
}
/*
* For stacked devices, we don't always get a COMPLETE event so
* the depth grows to insane values. Limit it to something sane(r).
*/
max_depth = 0;
for (i = 0; i < DDIR_RWDIR_CNT; i++) {
if (depth[i] > 1024)
depth[i] = 1024;
else if (!depth[i] && ios[i])
depth[i] = 1;
max_depth = max(depth[i], max_depth);
}
if (skipped_writes)
log_err("fio: %s skips replay of %lu writes due to read-only\n",
td->o.name, skipped_writes);
if (!ios[DDIR_READ] && !ios[DDIR_WRITE]) {
log_err("fio: found no ios in blktrace data\n");
return 1;
} else if (ios[DDIR_READ] && !ios[DDIR_WRITE]) {
td->o.td_ddir = TD_DDIR_READ;
td->o.max_bs[DDIR_READ] = rw_bs[DDIR_READ];
} else if (!ios[DDIR_READ] && ios[DDIR_WRITE]) {
td->o.td_ddir = TD_DDIR_WRITE;
td->o.max_bs[DDIR_WRITE] = rw_bs[DDIR_WRITE];
} else {
td->o.td_ddir = TD_DDIR_RW;
td->o.max_bs[DDIR_READ] = rw_bs[DDIR_READ];
td->o.max_bs[DDIR_WRITE] = rw_bs[DDIR_WRITE];
td->o.max_bs[DDIR_TRIM] = rw_bs[DDIR_TRIM];
}
/*
* We need to do direct/raw ios to the device, to avoid getting
* read-ahead in our way. But only do so if the minimum block size
* is a multiple of 4k, otherwise we don't know if it's safe to do so.
*/
if (!fio_option_is_set(&td->o, odirect) && !(td_min_bs(td) & 4095))
td->o.odirect = 1;
/*
* If depth wasn't manually set, use probed depth
*/
if (!fio_option_is_set(&td->o, iodepth))
td->o.iodepth = td->o.iodepth_low = max_depth;
return 0;
err:
close(fd);
fifo_free(fifo);
return 1;
}
