Spdk/test/unit/lib/ftl/ftl_rwb.c/ftl_rwb_ut.c
Claire J. In 658d118c06 lib/ftl: consider 3D TLC NAND read unit size
For the latest TLC NAND, one write buffer unit (rwb batch)
needs to be spread over three PUs instead of being allocated
to a single PU for better sequential read performance
since the optimal write size(ws_opt) of 3D TLC NAND is
3 times bigger than the optimal read size(rs_opt).

I added num_interleave_units in 'struct spdk_ftl_conf'
to configure the number of interleaving units per ws_opt.
If num_interleave_units is set as 1, the whole of the ws_opt
blocks are placed sequentially around each PU.
If num_interleave_units is set as N, the 1/N of the ws_opt
blocks are staggered. So consecutively numbered blocks
are separated by ws_opt / num_interleave_units.

The sequential read performance is improved from 1.9GiB/s
up to 2.97GiB/S with this patch on our system. No performance
degradation is observed on sequential writes or
4KB random reads/writes.

Please refer to the Trello card for more details.
https://trello.com/c/Osol93ZU

Change-Id: I371e72067b278ef43c3ac87a3d9ce9010d3fcb15
Signed-off-by: Claire J. In <claire.in@circuitblvd.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450976
Reviewed-by: Young Tack Jin <youngtack.jin@circuitblvd.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
2019-05-24 23:29:22 +00:00

590 lines
15 KiB
C

/*-
* BSD LICENSE
*
* Copyright (c) Intel Corporation.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
*
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in
* the documentation and/or other materials provided with the
* distribution.
* * Neither the name of Intel Corporation nor the names of its
* contributors may be used to endorse or promote products derived
* from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "spdk/stdinc.h"
#include "spdk_cunit.h"
#include "common/lib/test_env.c"
#include "ftl/ftl_rwb.c"
struct ftl_rwb_ut {
/* configurations */
struct spdk_ftl_conf conf;
size_t metadata_size;
size_t num_punits;
size_t xfer_size;
/* the fields below are calculated by the configurations */
size_t max_batches;
size_t max_active_batches;
size_t max_entries;
size_t max_allocable_entries;
size_t interleave_offset;
size_t num_entries_per_worker;
};
static struct ftl_rwb *g_rwb;
static struct ftl_rwb_ut g_ut;
static int _init_suite(void);
static int
init_suite1(void)
{
g_ut.conf.rwb_size = 1024 * 1024;
g_ut.conf.num_interleave_units = 1;
g_ut.metadata_size = 64;
g_ut.num_punits = 4;
g_ut.xfer_size = 16;
return _init_suite();
}
static int
init_suite2(void)
{
g_ut.conf.rwb_size = 2 * 1024 * 1024;
g_ut.conf.num_interleave_units = 4;
g_ut.metadata_size = 64;
g_ut.num_punits = 8;
g_ut.xfer_size = 16;
return _init_suite();
}
static int
_init_suite(void)
{
struct spdk_ftl_conf *conf = &g_ut.conf;
if (conf->num_interleave_units == 0 ||
g_ut.xfer_size % conf->num_interleave_units ||
g_ut.num_punits == 0) {
return -1;
}
g_ut.max_batches = conf->rwb_size / (FTL_BLOCK_SIZE * g_ut.xfer_size);
if (conf->num_interleave_units > 1) {
g_ut.max_batches += g_ut.num_punits;
g_ut.max_active_batches = g_ut.num_punits;
} else {
g_ut.max_batches++;
g_ut.max_active_batches = 1;
}
g_ut.max_entries = g_ut.max_batches * g_ut.xfer_size;
g_ut.max_allocable_entries = (g_ut.max_batches / g_ut.max_active_batches) *
g_ut.max_active_batches * g_ut.xfer_size;
g_ut.interleave_offset = g_ut.xfer_size / conf->num_interleave_units;
/* if max_batches is less than max_active_batches * 2, */
/* test_rwb_limits_applied will be failed. */
if (g_ut.max_batches < g_ut.max_active_batches * 2) {
return -1;
}
g_ut.num_entries_per_worker = 16 * g_ut.max_allocable_entries;
return 0;
}
static void
setup_rwb(void)
{
g_rwb = ftl_rwb_init(&g_ut.conf, g_ut.xfer_size,
g_ut.metadata_size, g_ut.num_punits);
SPDK_CU_ASSERT_FATAL(g_rwb != NULL);
}
static void
cleanup_rwb(void)
{
ftl_rwb_free(g_rwb);
g_rwb = NULL;
}
static void
test_rwb_acquire(void)
{
struct ftl_rwb_entry *entry;
size_t i;
setup_rwb();
/* Verify that it's possible to acquire all of the entries */
for (i = 0; i < g_ut.max_allocable_entries; ++i) {
entry = ftl_rwb_acquire(g_rwb, FTL_RWB_TYPE_USER);
SPDK_CU_ASSERT_FATAL(entry);
ftl_rwb_push(entry);
}
entry = ftl_rwb_acquire(g_rwb, FTL_RWB_TYPE_USER);
CU_ASSERT_PTR_NULL(entry);
cleanup_rwb();
}
static void
test_rwb_pop(void)
{
struct ftl_rwb_entry *entry;
struct ftl_rwb_batch *batch;
size_t entry_count, i, i_reset = 0, i_offset = 0;
uint64_t expected_lba;
setup_rwb();
/* Acquire all entries */
for (i = 0; i < g_ut.max_allocable_entries; ++i) {
entry = ftl_rwb_acquire(g_rwb, FTL_RWB_TYPE_USER);
SPDK_CU_ASSERT_FATAL(entry);
entry->lba = i;
ftl_rwb_push(entry);
}
/* Pop all batches and free them */
for (i = 0; i < g_ut.max_allocable_entries / g_ut.xfer_size; ++i) {
batch = ftl_rwb_pop(g_rwb);
SPDK_CU_ASSERT_FATAL(batch);
entry_count = 0;
ftl_rwb_foreach(entry, batch) {
if (i % g_ut.max_active_batches == 0) {
i_offset = i * g_ut.xfer_size;
}
if (entry_count % g_ut.interleave_offset == 0) {
i_reset = i % g_ut.max_active_batches +
(entry_count / g_ut.interleave_offset) *
g_ut.max_active_batches;
}
expected_lba = i_offset +
i_reset * g_ut.interleave_offset +
entry_count % g_ut.interleave_offset;
CU_ASSERT_EQUAL(entry->lba, expected_lba);
entry_count++;
}
CU_ASSERT_EQUAL(entry_count, g_ut.xfer_size);
ftl_rwb_batch_release(batch);
}
/* Acquire all entries once more */
for (i = 0; i < g_ut.max_allocable_entries; ++i) {
entry = ftl_rwb_acquire(g_rwb, FTL_RWB_TYPE_USER);
SPDK_CU_ASSERT_FATAL(entry);
ftl_rwb_push(entry);
}
/* Pop one batch and check we can acquire xfer_size entries */
for (i = 0; i < g_ut.max_active_batches; i++) {
batch = ftl_rwb_pop(g_rwb);
SPDK_CU_ASSERT_FATAL(batch);
ftl_rwb_batch_release(batch);
}
for (i = 0; i < g_ut.xfer_size * g_ut.max_active_batches; ++i) {
entry = ftl_rwb_acquire(g_rwb, FTL_RWB_TYPE_USER);
SPDK_CU_ASSERT_FATAL(entry);
ftl_rwb_push(entry);
}
entry = ftl_rwb_acquire(g_rwb, FTL_RWB_TYPE_USER);
CU_ASSERT_PTR_NULL(entry);
/* Pop and Release all batches */
for (i = 0; i < g_ut.max_allocable_entries / g_ut.xfer_size; ++i) {
batch = ftl_rwb_pop(g_rwb);
SPDK_CU_ASSERT_FATAL(batch);
ftl_rwb_batch_release(batch);
}
cleanup_rwb();
}
static void
test_rwb_disable_interleaving(void)
{
struct ftl_rwb_entry *entry;
struct ftl_rwb_batch *batch;
size_t entry_count, i;
setup_rwb();
ftl_rwb_disable_interleaving(g_rwb);
/* Acquire all entries and assign sequential lbas */
for (i = 0; i < g_ut.max_allocable_entries; ++i) {
entry = ftl_rwb_acquire(g_rwb, FTL_RWB_TYPE_USER);
SPDK_CU_ASSERT_FATAL(entry);
entry->lba = i;
ftl_rwb_push(entry);
}
/* Check for expected lbas */
for (i = 0; i < g_ut.max_allocable_entries / g_ut.xfer_size; ++i) {
batch = ftl_rwb_pop(g_rwb);
SPDK_CU_ASSERT_FATAL(batch);
entry_count = 0;
ftl_rwb_foreach(entry, batch) {
CU_ASSERT_EQUAL(entry->lba, i * g_ut.xfer_size + entry_count);
entry_count++;
}
CU_ASSERT_EQUAL(entry_count, g_ut.xfer_size);
ftl_rwb_batch_release(batch);
}
cleanup_rwb();
}
static void
test_rwb_batch_revert(void)
{
struct ftl_rwb_batch *batch;
struct ftl_rwb_entry *entry;
size_t i;
setup_rwb();
for (i = 0; i < g_ut.max_allocable_entries; ++i) {
entry = ftl_rwb_acquire(g_rwb, FTL_RWB_TYPE_USER);
SPDK_CU_ASSERT_FATAL(entry);
ftl_rwb_push(entry);
}
/* Pop one batch and revert it */
batch = ftl_rwb_pop(g_rwb);
SPDK_CU_ASSERT_FATAL(batch);
ftl_rwb_batch_revert(batch);
/* Verify all of the batches */
for (i = 0; i < g_ut.max_allocable_entries / g_ut.xfer_size; ++i) {
batch = ftl_rwb_pop(g_rwb);
CU_ASSERT_PTR_NOT_NULL_FATAL(batch);
}
cleanup_rwb();
}
static void
test_rwb_entry_from_offset(void)
{
struct ftl_rwb_entry *entry;
struct ftl_ppa ppa = { .cached = 1 };
size_t i;
setup_rwb();
for (i = 0; i < g_ut.max_allocable_entries; ++i) {
ppa.offset = i;
entry = ftl_rwb_entry_from_offset(g_rwb, i);
CU_ASSERT_EQUAL(ppa.offset, entry->pos);
}
cleanup_rwb();
}
static void *
test_rwb_worker(void *ctx)
{
struct ftl_rwb_entry *entry;
unsigned int *num_done = ctx;
size_t i;
for (i = 0; i < g_ut.num_entries_per_worker; ++i) {
while (1) {
entry = ftl_rwb_acquire(g_rwb, FTL_RWB_TYPE_USER);
if (entry) {
entry->flags = 0;
ftl_rwb_push(entry);
break;
} else {
/* Allow other threads to run under valgrind */
pthread_yield();
}
}
}
__atomic_fetch_add(num_done, 1, __ATOMIC_SEQ_CST);
return NULL;
}
static void
test_rwb_parallel(void)
{
struct ftl_rwb_batch *batch;
struct ftl_rwb_entry *entry;
#define NUM_PARALLEL_WORKERS 4
pthread_t workers[NUM_PARALLEL_WORKERS];
unsigned int num_done = 0;
size_t i, num_entries = 0;
bool all_done = false;
int rc;
setup_rwb();
for (i = 0; i < NUM_PARALLEL_WORKERS; ++i) {
rc = pthread_create(&workers[i], NULL, test_rwb_worker, (void *)&num_done);
CU_ASSERT_TRUE(rc == 0);
}
while (1) {
batch = ftl_rwb_pop(g_rwb);
if (batch) {
ftl_rwb_foreach(entry, batch) {
num_entries++;
}
ftl_rwb_batch_release(batch);
} else {
if (NUM_PARALLEL_WORKERS == __atomic_load_n(&num_done, __ATOMIC_SEQ_CST)) {
if (!all_done) {
/* Pop all left entries from rwb */
all_done = true;
continue;
}
for (i = 0; i < NUM_PARALLEL_WORKERS; ++i) {
pthread_join(workers[i], NULL);
}
break;
}
/* Allow other threads to run under valgrind */
pthread_yield();
}
}
CU_ASSERT_TRUE(num_entries == NUM_PARALLEL_WORKERS * g_ut.num_entries_per_worker);
cleanup_rwb();
}
static void
test_rwb_limits_base(void)
{
struct ftl_rwb_entry *entry;
size_t limits[FTL_RWB_TYPE_MAX];
setup_rwb();
ftl_rwb_get_limits(g_rwb, limits);
CU_ASSERT_TRUE(limits[FTL_RWB_TYPE_INTERNAL] == ftl_rwb_entry_cnt(g_rwb));
CU_ASSERT_TRUE(limits[FTL_RWB_TYPE_USER] == ftl_rwb_entry_cnt(g_rwb));
/* Verify it's possible to acquire both type of entries */
entry = ftl_rwb_acquire(g_rwb, FTL_RWB_TYPE_INTERNAL);
CU_ASSERT_PTR_NOT_NULL_FATAL(entry);
entry = ftl_rwb_acquire(g_rwb, FTL_RWB_TYPE_USER);
CU_ASSERT_PTR_NOT_NULL_FATAL(entry);
cleanup_rwb();
}
static void
test_rwb_limits_set(void)
{
size_t limits[FTL_RWB_TYPE_MAX], check[FTL_RWB_TYPE_MAX];
size_t i;
setup_rwb();
/* Check valid limits */
ftl_rwb_get_limits(g_rwb, limits);
memcpy(check, limits, sizeof(limits));
ftl_rwb_set_limits(g_rwb, limits);
ftl_rwb_get_limits(g_rwb, limits);
CU_ASSERT(memcmp(check, limits, sizeof(limits)) == 0);
for (i = 0; i < FTL_RWB_TYPE_MAX; ++i) {
ftl_rwb_get_limits(g_rwb, limits);
limits[i] = 0;
}
memcpy(check, limits, sizeof(limits));
ftl_rwb_set_limits(g_rwb, limits);
ftl_rwb_get_limits(g_rwb, limits);
CU_ASSERT(memcmp(check, limits, sizeof(limits)) == 0);
cleanup_rwb();
}
static void
test_rwb_limits_applied(void)
{
struct ftl_rwb_entry *entry;
struct ftl_rwb_batch *batch;
size_t limits[FTL_RWB_TYPE_MAX];
const size_t test_limit = g_ut.xfer_size * g_ut.max_active_batches;
size_t i;
setup_rwb();
/* Check that it's impossible to acquire any entries when the limits are */
/* set to 0 */
ftl_rwb_get_limits(g_rwb, limits);
limits[FTL_RWB_TYPE_USER] = 0;
ftl_rwb_set_limits(g_rwb, limits);
entry = ftl_rwb_acquire(g_rwb, FTL_RWB_TYPE_USER);
CU_ASSERT_PTR_NULL(entry);
limits[FTL_RWB_TYPE_USER] = ftl_rwb_entry_cnt(g_rwb);
limits[FTL_RWB_TYPE_INTERNAL] = 0;
ftl_rwb_set_limits(g_rwb, limits);
entry = ftl_rwb_acquire(g_rwb, FTL_RWB_TYPE_INTERNAL);
CU_ASSERT_PTR_NULL(entry);
/* Check positive limits */
limits[FTL_RWB_TYPE_USER] = ftl_rwb_entry_cnt(g_rwb);
limits[FTL_RWB_TYPE_INTERNAL] = test_limit;
ftl_rwb_set_limits(g_rwb, limits);
for (i = 0; i < test_limit; ++i) {
entry = ftl_rwb_acquire(g_rwb, FTL_RWB_TYPE_INTERNAL);
SPDK_CU_ASSERT_FATAL(entry);
entry->flags = FTL_IO_INTERNAL;
ftl_rwb_push(entry);
}
/* Now we expect null, since we've reached threshold */
entry = ftl_rwb_acquire(g_rwb, FTL_RWB_TYPE_INTERNAL);
CU_ASSERT_PTR_NULL(entry);
for (i = 0; i < test_limit / g_ut.xfer_size; ++i) {
/* Complete the entries and check we can retrieve the entries once again */
batch = ftl_rwb_pop(g_rwb);
SPDK_CU_ASSERT_FATAL(batch);
ftl_rwb_batch_release(batch);
}
entry = ftl_rwb_acquire(g_rwb, FTL_RWB_TYPE_INTERNAL);
SPDK_CU_ASSERT_FATAL(entry);
entry->flags = FTL_IO_INTERNAL;
/* Set the same limit but this time for user entries */
limits[FTL_RWB_TYPE_USER] = test_limit;
limits[FTL_RWB_TYPE_INTERNAL] = ftl_rwb_entry_cnt(g_rwb);
ftl_rwb_set_limits(g_rwb, limits);
for (i = 0; i < test_limit; ++i) {
entry = ftl_rwb_acquire(g_rwb, FTL_RWB_TYPE_USER);
SPDK_CU_ASSERT_FATAL(entry);
ftl_rwb_push(entry);
}
/* Now we expect null, since we've reached threshold */
entry = ftl_rwb_acquire(g_rwb, FTL_RWB_TYPE_USER);
CU_ASSERT_PTR_NULL(entry);
/* Check that we're still able to acquire a number of internal entries */
/* while the user entires are being throttled */
for (i = 0; i < g_ut.xfer_size; ++i) {
entry = ftl_rwb_acquire(g_rwb, FTL_RWB_TYPE_INTERNAL);
SPDK_CU_ASSERT_FATAL(entry);
}
cleanup_rwb();
}
int
main(int argc, char **argv)
{
CU_pSuite suite1, suite2;
unsigned int num_failures;
if (CU_initialize_registry() != CUE_SUCCESS) {
return CU_get_error();
}
suite1 = CU_add_suite("suite1", init_suite1, NULL);
if (!suite1) {
CU_cleanup_registry();
return CU_get_error();
}
suite2 = CU_add_suite("suite2", init_suite2, NULL);
if (!suite2) {
CU_cleanup_registry();
return CU_get_error();
}
if (
CU_add_test(suite1, "test_rwb_acquire",
test_rwb_acquire) == NULL
|| CU_add_test(suite1, "test_rwb_pop",
test_rwb_pop) == NULL
|| CU_add_test(suite1, "test_rwb_disable_interleaving",
test_rwb_disable_interleaving) == NULL
|| CU_add_test(suite1, "test_rwb_batch_revert",
test_rwb_batch_revert) == NULL
|| CU_add_test(suite1, "test_rwb_entry_from_offset",
test_rwb_entry_from_offset) == NULL
|| CU_add_test(suite1, "test_rwb_parallel",
test_rwb_parallel) == NULL
|| CU_add_test(suite1, "test_rwb_limits_base",
test_rwb_limits_base) == NULL
|| CU_add_test(suite1, "test_rwb_limits_set",
test_rwb_limits_set) == NULL
|| CU_add_test(suite1, "test_rwb_limits_applied",
test_rwb_limits_applied) == NULL
|| CU_add_test(suite2, "test_rwb_acquire",
test_rwb_acquire) == NULL
|| CU_add_test(suite2, "test_rwb_pop",
test_rwb_pop) == NULL
|| CU_add_test(suite2, "test_rwb_disable_interleaving",
test_rwb_disable_interleaving) == NULL
|| CU_add_test(suite2, "test_rwb_batch_revert",
test_rwb_batch_revert) == NULL
|| CU_add_test(suite2, "test_rwb_entry_from_offset",
test_rwb_entry_from_offset) == NULL
|| CU_add_test(suite2, "test_rwb_parallel",
test_rwb_parallel) == NULL
|| CU_add_test(suite2, "test_rwb_limits_base",
test_rwb_limits_base) == NULL
|| CU_add_test(suite2, "test_rwb_limits_set",
test_rwb_limits_set) == NULL
|| CU_add_test(suite2, "test_rwb_limits_applied",
test_rwb_limits_applied) == NULL
) {
CU_cleanup_registry();
return CU_get_error();
}
CU_basic_set_mode(CU_BRM_VERBOSE);
CU_basic_run_tests();
num_failures = CU_get_number_of_failures();
CU_cleanup_registry();
return num_failures;
}