int r, new_layout = 0, unlink_file = 0, flags;
unsigned long long left;
unsigned int bs;
- char *b;
+ char *b = NULL;
if (read_only) {
log_err("fio: refusing extend of file due to read-only\n");
* does that for operations involving reads, or for writes
* where overwrite is set
*/
- if (td_read(td) || (td_write(td) && td->o.overwrite) ||
+ if (td_read(td) ||
+ (td_write(td) && td->o.overwrite && !td->o.file_append) ||
(td_write(td) && td->io_ops->flags & FIO_NOEXTEND))
new_layout = 1;
- if (td_write(td) && !td->o.overwrite)
+ if (td_write(td) && !td->o.overwrite && !td->o.file_append)
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)) {
+ if ((td_io_unlink_file(td, f) < 0) && (errno != ENOENT)) {
td_verror(td, errno, "unlink");
return 1;
}
}
- flags = O_WRONLY | O_CREAT;
+ flags = O_WRONLY;
+ if (td->o.allow_create)
+ flags |= O_CREAT;
if (new_layout)
flags |= O_TRUNC;
+#ifdef WIN32
+ flags |= _O_BINARY;
+#endif
+
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");
+ int err = errno;
+
+ if (err == ENOENT && !td->o.allow_create)
+ log_err("fio: file creation disallowed by "
+ "allow_file_create=0\n");
+ else
+ td_verror(td, err, "open");
return 1;
}
if (td->terminate) {
dprint(FD_FILE, "terminate unlink %s\n", f->file_name);
- unlink(f->file_name);
+ td_io_unlink_file(td, f);
} else if (td->o.create_fsync) {
if (fsync(f->fd) < 0) {
td_verror(td, errno, "fsync");
err:
close(f->fd);
f->fd = -1;
+ if (b)
+ free(b);
return 1;
}
static int pre_read_file(struct thread_data *td, struct fio_file *f)
{
- int r, did_open = 0, old_runstate;
+ int ret = 0, r, did_open = 0, old_runstate;
unsigned long long left;
unsigned int bs;
char *b;
b = malloc(bs);
memset(b, 0, bs);
- lseek(f->fd, f->file_offset, SEEK_SET);
+ if (lseek(f->fd, f->file_offset, SEEK_SET) < 0) {
+ td_verror(td, errno, "lseek");
+ log_err("fio: failed to lseek pre-read file\n");
+ ret = 1;
+ goto error;
+ }
+
left = f->io_size;
while (left && !td->terminate) {
}
}
+error:
td_restore_runstate(td, old_runstate);
if (did_open)
td->io_ops->close_file(td, f);
+
free(b);
- return 0;
+ return ret;
}
-static unsigned long long get_rand_file_size(struct thread_data *td)
+unsigned long long get_rand_file_size(struct thread_data *td)
{
unsigned long long ret, sized;
+ uint64_t frand_max;
unsigned long r;
- if (td->o.use_os_rand) {
- r = os_random_long(&td->file_size_state);
- sized = td->o.file_size_high - td->o.file_size_low;
- ret = (unsigned long long) ((double) sized * (r / (OS_RAND_MAX + 1.0)));
- } else {
- r = __rand(&td->__file_size_state);
- sized = td->o.file_size_high - td->o.file_size_low;
- ret = (unsigned long long) ((double) sized * (r / (FRAND_MAX + 1.0)));
- }
-
+ frand_max = rand_max(&td->file_size_state);
+ r = __rand(&td->file_size_state);
+ sized = td->o.file_size_high - td->o.file_size_low;
+ ret = (unsigned long long) ((double) sized * (r / (frand_max + 1.0)));
ret += td->o.file_size_low;
ret -= (ret % td->o.rw_min_bs);
return ret;
unsigned long long off,
unsigned long long len)
{
- int ret = 0;
+ int errval = 0, ret = 0;
+
+#ifdef CONFIG_ESX
+ return 0;
+#endif
if (len == -1ULL)
len = f->io_size;
dprint(FD_IO, "invalidate cache %s: %llu/%llu\n", f->file_name, off,
len);
- /*
- * FIXME: add blockdev flushing too
- */
- if (f->mmap_ptr) {
- ret = posix_madvise(f->mmap_ptr, f->mmap_sz, POSIX_MADV_DONTNEED);
-#ifdef FIO_MADV_FREE
- if (f->filetype == FIO_TYPE_BD)
- (void) posix_madvise(f->mmap_ptr, f->mmap_sz, FIO_MADV_FREE);
-#endif
+ if (td->io_ops->invalidate) {
+ ret = td->io_ops->invalidate(td, f);
+ if (ret < 0)
+ errval = ret;
} else if (f->filetype == FIO_TYPE_FILE) {
ret = posix_fadvise(f->fd, off, len, POSIX_FADV_DONTNEED);
+ if (ret)
+ errval = ret;
} else if (f->filetype == FIO_TYPE_BD) {
+ int retry_count = 0;
+
ret = blockdev_invalidate_cache(f);
+ while (ret < 0 && errno == EAGAIN && retry_count++ < 25) {
+ /*
+ * Linux multipath devices reject ioctl while
+ * the maps are being updated. That window can
+ * last tens of milliseconds; we'll try up to
+ * a quarter of a second.
+ */
+ usleep(10000);
+ ret = blockdev_invalidate_cache(f);
+ }
if (ret < 0 && errno == EACCES && geteuid()) {
if (!root_warn) {
log_err("fio: only root may flush block "
}
ret = 0;
}
+ if (ret < 0)
+ errval = errno;
} else if (f->filetype == FIO_TYPE_CHAR || f->filetype == FIO_TYPE_PIPE)
ret = 0;
- if (ret < 0) {
- td_verror(td, errno, "invalidate_cache");
- return 1;
- } else if (ret > 0) {
- td_verror(td, ret, "invalidate_cache");
- return 1;
- }
+ /*
+ * Cache flushing isn't a fatal condition, and we know it will
+ * happen on some platforms where we don't have the proper
+ * function to flush eg block device caches. So just warn and
+ * continue on our way.
+ */
+ if (errval)
+ log_info("fio: cache invalidation of %s failed: %s\n", f->file_name, strerror(errval));
- return ret;
+ return 0;
}
from_hash = 0;
}
+#ifdef WIN32
+ flags |= _O_BINARY;
+#endif
+
f->fd = open(f->file_name, flags, 0600);
return from_hash;
}
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");
}
if (td->o.sync_io)
flags |= O_SYNC;
- if (td->o.create_on_open)
+ if (td->o.create_on_open && td->o.allow_create)
flags |= O_CREAT;
skip_flags:
if (f->filetype != FIO_TYPE_FILE)
if (!read_only)
flags |= O_RDWR;
- if (f->filetype == FIO_TYPE_FILE)
+ if (f->filetype == FIO_TYPE_FILE && td->o.allow_create)
flags |= O_CREAT;
if (is_std)
}
td_verror(td, __e, buf);
+ return 1;
}
if (!from_hash && f->fd != -1) {
if (td->error != ENOENT) {
log_err("%s\n", td->verror);
err = 1;
+ break;
}
clear_error(td);
}
} else if (f->filetype != FIO_TYPE_FILE)
continue;
- strcpy(buf, f->file_name);
+ buf[255] = '\0';
+ strncpy(buf, f->file_name, 255);
if (stat(buf, &sb) < 0) {
if (errno != ENOENT)
if (fm)
continue;
- fm = malloc(sizeof(*fm));
- strcpy(fm->__base, buf);
+ fm = calloc(1, sizeof(*fm));
+ strncpy(fm->__base, buf, sizeof(fm->__base) - 1);
fm->base = basename(fm->__base);
fm->key = sb.st_dev;
flist_add(&fm->list, &list);
fm = flist_entry(n, struct fio_mount, list);
flist_del(&fm->list);
- sz = get_fs_size(fm->base);
+ sz = get_fs_free_size(fm->base);
if (sz && sz != -1ULL)
ret += sz;
return f->real_file_size;
return td->o.start_offset +
- (td->thread_number - 1) * td->o.offset_increment;
+ td->subjob_number * td->o.offset_increment;
}
/*
* Calculate per-file size and potential extra size for the
* first files, if needed.
*/
- if (!o->file_size_low) {
+ if (!o->file_size_low && o->nr_files) {
uint64_t all_fs;
fs = o->size / o->nr_files;
}
}
- if (!o->size || o->size > total_size)
+ if (td->o.block_error_hist) {
+ int len;
+
+ assert(td->o.nr_files == 1); /* checked in fixup_options */
+ f = td->files[0];
+ len = f->io_size / td->o.bs[DDIR_TRIM];
+ if (len > MAX_NR_BLOCK_INFOS || len <= 0) {
+ log_err("fio: cannot calculate block histogram with "
+ "%d trim blocks, maximum %d\n",
+ len, MAX_NR_BLOCK_INFOS);
+ td_verror(td, EINVAL, "block_error_hist");
+ goto err_out;
+ }
+
+ td->ts.nr_block_infos = len;
+ for (i = 0; i < len; i++)
+ td->ts.block_infos[i] =
+ BLOCK_INFO(0, BLOCK_STATE_UNINIT);
+ } else
+ td->ts.nr_block_infos = 0;
+
+ if (!o->size || (total_size && o->size > total_size))
o->size = total_size;
+ if (o->size < td_min_bs(td)) {
+ log_err("fio: blocksize too large for data set\n");
+ goto err_out;
+ }
+
/*
* See if we need to extend some files
*/
if (need_extend) {
temp_stall_ts = 1;
- if (output_format == FIO_OUTPUT_NORMAL)
+ if (output_format & FIO_OUTPUT_NORMAL)
log_info("%s: Laying out IO file(s) (%u file(s) /"
" %lluMB)\n", o->name, need_extend,
extend_size >> 20);
err = __file_invalidate_cache(td, f, old_len,
extend_len);
- close(f->fd);
+
+ /*
+ * Shut up static checker
+ */
+ if (f->fd != -1)
+ close(f->fd);
+
f->fd = -1;
if (err)
break;
* iolog already set the total io size, if we read back
* stored entries.
*/
- if (!o->read_iolog_file)
- td->total_io_size = o->size * o->loops;
+ if (!o->read_iolog_file) {
+ if (o->io_limit)
+ td->total_io_size = o->io_limit * o->loops;
+ else
+ td->total_io_size = o->size * o->loops;
+ }
done:
if (o->create_only)
{
unsigned int range_size, seed;
unsigned long nranges;
- uint64_t file_size;
+ uint64_t fsize;
range_size = min(td->o.min_bs[DDIR_READ], td->o.min_bs[DDIR_WRITE]);
- file_size = min(f->real_file_size, f->io_size);
+ fsize = min(f->real_file_size, f->io_size);
- nranges = (file_size + range_size - 1) / range_size;
+ nranges = (fsize + range_size - 1) / range_size;
seed = jhash(f->file_name, strlen(f->file_name), 0) * td->thread_number;
if (!td->o.rand_repeatable)
if (td->o.random_distribution == FIO_RAND_DIST_ZIPF)
zipf_init(&f->zipf, nranges, td->o.zipf_theta.u.f, seed);
- else
+ else if (td->o.random_distribution == FIO_RAND_DIST_PARETO)
pareto_init(&f->zipf, nranges, td->o.pareto_h.u.f, seed);
+ else if (td->o.random_distribution == FIO_RAND_DIST_GAUSS)
+ gauss_init(&f->gauss, nranges, td->o.gauss_dev.u.f, seed);
return 1;
}
return 1;
}
+/*
+ * Check if the number of blocks exceeds the randomness capability of
+ * the selected generator. Tausworthe is 32-bit, the others are fullly
+ * 64-bit capable.
+ */
+static int check_rand_gen_limits(struct thread_data *td, struct fio_file *f,
+ uint64_t blocks)
+{
+ if (blocks <= FRAND32_MAX)
+ return 0;
+ if (td->o.random_generator != FIO_RAND_GEN_TAUSWORTHE)
+ return 0;
+
+ /*
+ * If the user hasn't specified a random generator, switch
+ * to tausworthe64 with informational warning. If the user did
+ * specify one, just warn.
+ */
+ log_info("fio: file %s exceeds 32-bit tausworthe random generator.\n",
+ f->file_name);
+
+ if (!fio_option_is_set(&td->o, random_generator)) {
+ log_info("fio: Switching to tausworthe64. Use the "
+ "random_generator= option to get rid of this "
+ " warning.\n");
+ td->o.random_generator = FIO_RAND_GEN_TAUSWORTHE64;
+ return 0;
+ }
+
+ /*
+ * Just make this information to avoid breaking scripts.
+ */
+ log_info("fio: Use the random_generator= option to switch to lfsr or "
+ "tausworthe64.\n");
+ return 0;
+}
+
int init_random_map(struct thread_data *td)
{
unsigned long long blocks;
return 0;
for_each_file(td, f, i) {
- uint64_t file_size = min(f->real_file_size, f->io_size);
+ uint64_t fsize = min(f->real_file_size, f->io_size);
+
+ blocks = fsize / (unsigned long long) td->o.rw_min_bs;
- blocks = file_size / (unsigned long long) td->o.rw_min_bs;
+ if (check_rand_gen_limits(td, f, blocks))
+ return 1;
if (td->o.random_generator == FIO_RAND_GEN_LFSR) {
unsigned long seed;
seed = td->rand_seeds[FIO_RAND_BLOCK_OFF];
-
- if (!lfsr_init(&f->lfsr, blocks, seed, 0))
+
+ if (!lfsr_init(&f->lfsr, blocks, seed, 0)) {
+ fio_file_set_lfsr(f);
continue;
+ }
} else if (!td->o.norandommap) {
f->io_axmap = axmap_new(blocks);
- if (f->io_axmap)
+ if (f->io_axmap) {
+ fio_file_set_axmap(f);
continue;
+ }
} else if (td->o.norandommap)
continue;
dprint(FD_FILE, "close files\n");
for_each_file(td, f, i) {
+ if (td->o.unlink && f->filetype == FIO_TYPE_FILE) {
+ dprint(FD_FILE, "free unlink %s\n", f->file_name);
+ td_io_unlink_file(td, f);
+ }
+
if (fio_file_open(f))
td_io_close_file(td, f);
if (td->o.unlink && f->filetype == FIO_TYPE_FILE) {
dprint(FD_FILE, "free unlink %s\n", f->file_name);
- unlink(f->file_name);
+ td_io_unlink_file(td, f);
}
sfree(f->file_name);
f->file_name = NULL;
- axmap_free(f->io_axmap);
- f->io_axmap = NULL;
+ if (fio_file_axmap(f)) {
+ axmap_free(f->io_axmap);
+ f->io_axmap = NULL;
+ }
sfree(f);
}
}
}
-static void set_already_allocated(const char *fname) {
- struct file_name *fn;
+static int __is_already_allocated(const char *fname)
+{
+ struct flist_head *entry;
+ char *filename;
- fn = malloc(sizeof(struct file_name));
- fn->filename = strdup(fname);
- flist_add_tail(&fn->list, &filename_list);
+ if (flist_empty(&filename_list))
+ return 0;
+
+ flist_for_each(entry, &filename_list) {
+ filename = flist_entry(entry, struct file_name, list)->filename;
+
+ if (strcmp(filename, fname) == 0)
+ return 1;
+ }
+
+ return 0;
}
static int is_already_allocated(const char *fname)
{
- struct flist_head *entry;
- char *filename;
+ int ret;
- if (!flist_empty(&filename_list))
- {
- flist_for_each(entry, &filename_list) {
- filename = flist_entry(entry, struct file_name, list)->filename;
+ fio_file_hash_lock();
+ ret = __is_already_allocated(fname);
+ fio_file_hash_unlock();
+ return ret;
+}
- if (strcmp(filename, fname) == 0)
- return 1;
- }
+static void set_already_allocated(const char *fname)
+{
+ struct file_name *fn;
+
+ fn = malloc(sizeof(struct file_name));
+ fn->filename = strdup(fname);
+
+ fio_file_hash_lock();
+ if (!__is_already_allocated(fname)) {
+ flist_add_tail(&fn->list, &filename_list);
+ fn = NULL;
}
+ fio_file_hash_unlock();
- return 0;
+ if (fn) {
+ free(fn->filename);
+ free(fn);
+ }
}
-static void free_already_allocated() {
+
+static void free_already_allocated(void)
+{
struct flist_head *entry, *tmp;
struct file_name *fn;
- if (!flist_empty(&filename_list))
- {
- flist_for_each_safe(entry, tmp, &filename_list) {
- fn = flist_entry(entry, struct file_name, list);
- free(fn->filename);
- flist_del(&fn->list);
- free(fn);
- }
+ if (flist_empty(&filename_list))
+ return;
+
+ fio_file_hash_lock();
+ flist_for_each_safe(entry, tmp, &filename_list) {
+ fn = flist_entry(entry, struct file_name, list);
+ free(fn->filename);
+ flist_del(&fn->list);
+ free(fn);
}
+
+ fio_file_hash_unlock();
+}
+
+static struct fio_file *alloc_new_file(struct thread_data *td)
+{
+ struct fio_file *f;
+
+ f = smalloc(sizeof(*f));
+ if (!f) {
+ log_err("fio: smalloc OOM\n");
+ assert(0);
+ return NULL;
+ }
+
+ f->fd = -1;
+ f->shadow_fd = -1;
+ fio_file_reset(td, f);
+ return f;
}
int add_file(struct thread_data *td, const char *fname, int numjob, int inc)
dprint(FD_FILE, "add file %s\n", fname);
if (td->o.directory)
- len = set_name_idx(file_name, td->o.directory, numjob);
+ len = set_name_idx(file_name, PATH_MAX, td->o.directory, numjob);
sprintf(file_name + len, "%s", fname);
if (numjob && is_already_allocated(file_name))
return 0;
- f = smalloc(sizeof(*f));
- if (!f) {
- log_err("fio: smalloc OOM\n");
- assert(0);
- }
-
- f->fd = -1;
- f->shadow_fd = -1;
- fio_file_reset(td, f);
+ f = alloc_new_file(td);
if (td->files_size <= td->files_index) {
unsigned int new_size = td->o.nr_files + 1;
set_already_allocated(file_name);
- /*
- * For adding files after the fact - if openfiles= isn't
- * given as an option, ensure we allow at least one file open
- */
- if (!td->o.open_files)
- td->o.open_files = 1;
-
if (inc)
td->o.nr_files++;
if (--f->references)
return 0;
- if (should_fsync(td) && td->o.fsync_on_close)
+ if (should_fsync(td) && td->o.fsync_on_close) {
f_ret = fsync(f->fd);
+ if (f_ret < 0)
+ f_ret = errno;
+ }
if (td->io_ops->close_file)
ret = td->io_ops->close_file(td, f);
if (lstat(full_path, &sb) == -1) {
if (errno != ENOENT) {
td_verror(td, errno, "stat");
- return 1;
+ ret = 1;
+ break;
}
}
for_each_file(org, f, i) {
struct fio_file *__f;
- __f = smalloc(sizeof(*__f));
- if (!__f) {
- log_err("fio: smalloc OOM\n");
- assert(0);
- }
- __f->fd = -1;
- fio_file_reset(td, __f);
+ __f = alloc_new_file(td);
if (f->file_name) {
__f->file_name = smalloc_strdup(f->file_name);
void free_release_files(struct thread_data *td)
{
close_files(td);
+ td->o.nr_files = 0;
+ td->o.open_files = 0;
td->files_index = 0;
td->nr_normal_files = 0;
}
void fio_file_reset(struct thread_data *td, struct fio_file *f)
{
- f->last_pos = f->file_offset;
- f->last_start = -1ULL;
- if (f->io_axmap)
+ int i;
+
+ for (i = 0; i < DDIR_RWDIR_CNT; i++) {
+ f->last_pos[i] = f->file_offset;
+ f->last_start[i] = -1ULL;
+ }
+
+ if (fio_file_axmap(f))
axmap_reset(f->io_axmap);
- if (td->o.random_generator == FIO_RAND_GEN_LFSR)
+ else if (fio_file_lfsr(f))
lfsr_reset(&f->lfsr, td->rand_seeds[FIO_RAND_BLOCK_OFF]);
}
}
/* free memory used in initialization phase only */
-void filesetup_mem_free() {
+void filesetup_mem_free(void)
+{
free_already_allocated();
}