static inline void
run_gen_bench_impl(const char *impl)
{
int fn, ncols;
uint64_t ds, iter_cnt, iter, disksize;
hrtime_t start;
double elapsed, d_bw;
/* Benchmark generate functions */
for (fn = 0; fn < RAIDZ_GEN_NUM; fn++) {
for (ds = MIN_CS_SHIFT; ds <= MAX_CS_SHIFT; ds++) {
/* create suitable raidz_map */
ncols = rto_opts.rto_dcols + fn + 1;
zio_bench.io_size = 1ULL << ds;
rm_bench = vdev_raidz_map_alloc(&zio_bench,
BENCH_ASHIFT, ncols, fn+1);
/* estimate iteration count */
iter_cnt = GEN_BENCH_MEMORY;
iter_cnt /= zio_bench.io_size;
start = gethrtime();
for (iter = 0; iter < iter_cnt; iter++)
vdev_raidz_generate_parity(rm_bench);
elapsed = NSEC2SEC((double) (gethrtime() - start));
disksize = (1ULL << ds) / rto_opts.rto_dcols;
d_bw = (double)iter_cnt * (double)disksize;
d_bw /= (1024.0 * 1024.0 * elapsed);
LOG(D_ALL, "%10s, %8s, %zu, %10llu, %lf, %lf, %u\n",
impl,
raidz_gen_name[fn],
rto_opts.rto_dcols,
(1ULL<<ds),
d_bw,
d_bw * (double)(ncols),
(unsigned) iter_cnt);
vdev_raidz_map_free(rm_bench);
}
}
}
static void
run_rec_bench_impl(const char *impl)
{
int fn, ncols, nbad;
uint64_t ds, iter_cnt, iter, disksize;
hrtime_t start;
double elapsed, d_bw;
static const int tgt[7][3] = {
{1, 2, 3}, /* rec_p: bad QR & D[0] */
{0, 2, 3}, /* rec_q: bad PR & D[0] */
{0, 1, 3}, /* rec_r: bad PQ & D[0] */
{2, 3, 4}, /* rec_pq: bad R & D[0][1] */
{1, 3, 4}, /* rec_pr: bad Q & D[0][1] */
{0, 3, 4}, /* rec_qr: bad P & D[0][1] */
{3, 4, 5} /* rec_pqr: bad & D[0][1][2] */
};
for (fn = 0; fn < RAIDZ_REC_NUM; fn++) {
for (ds = MIN_CS_SHIFT; ds <= MAX_CS_SHIFT; ds++) {
/* create suitable raidz_map */
ncols = rto_opts.rto_dcols + PARITY_PQR;
zio_bench.io_size = 1ULL << ds;
/*
* raidz block is too short to test
* the requested method
*/
if (zio_bench.io_size / rto_opts.rto_dcols <
(1ULL << BENCH_ASHIFT))
continue;
rm_bench = vdev_raidz_map_alloc(&zio_bench,
BENCH_ASHIFT, ncols, PARITY_PQR);
/* estimate iteration count */
iter_cnt = (REC_BENCH_MEMORY);
iter_cnt /= zio_bench.io_size;
/* calculate how many bad columns there are */
nbad = MIN(3, raidz_ncols(rm_bench) -
raidz_parity(rm_bench));
start = gethrtime();
for (iter = 0; iter < iter_cnt; iter++)
vdev_raidz_reconstruct(rm_bench, tgt[fn], nbad);
elapsed = NSEC2SEC((double) (gethrtime() - start));
disksize = (1ULL << ds) / rto_opts.rto_dcols;
d_bw = (double)iter_cnt * (double)(disksize);
d_bw /= (1024.0 * 1024.0 * elapsed);
LOG(D_ALL, "%10s, %8s, %zu, %10llu, %lf, %lf, %u\n",
impl,
raidz_rec_name[fn],
rto_opts.rto_dcols,
(1ULL<<ds),
d_bw,
d_bw * (double)ncols,
(unsigned) iter_cnt);
vdev_raidz_map_free(rm_bench);
}
}
}
static void
mmp_thread(void *arg)
{
spa_t *spa = (spa_t *)arg;
mmp_thread_t *mmp = &spa->spa_mmp;
boolean_t last_spa_suspended = spa_suspended(spa);
boolean_t last_spa_multihost = spa_multihost(spa);
callb_cpr_t cpr;
hrtime_t max_fail_ns = zfs_multihost_fail_intervals *
MSEC2NSEC(MAX(zfs_multihost_interval, MMP_MIN_INTERVAL));
mmp_thread_enter(mmp, &cpr);
/*
* The mmp_write_done() function calculates mmp_delay based on the
* prior value of mmp_delay and the elapsed time since the last write.
* For the first mmp write, there is no "last write", so we start
* with fake, but reasonable, default non-zero values.
*/
mmp->mmp_delay = MSEC2NSEC(MAX(zfs_multihost_interval,
MMP_MIN_INTERVAL)) / MAX(vdev_count_leaves(spa), 1);
mmp->mmp_last_write = gethrtime() - mmp->mmp_delay;
while (!mmp->mmp_thread_exiting) {
uint64_t mmp_fail_intervals = zfs_multihost_fail_intervals;
uint64_t mmp_interval = MSEC2NSEC(
MAX(zfs_multihost_interval, MMP_MIN_INTERVAL));
boolean_t suspended = spa_suspended(spa);
boolean_t multihost = spa_multihost(spa);
hrtime_t start, next_time;
start = gethrtime();
if (multihost) {
next_time = start + mmp_interval /
MAX(vdev_count_leaves(spa), 1);
} else {
next_time = start + MSEC2NSEC(MMP_DEFAULT_INTERVAL);
}
/*
* When MMP goes off => on, or spa goes suspended =>
* !suspended, we know no writes occurred recently. We
* update mmp_last_write to give us some time to try.
*/
if ((!last_spa_multihost && multihost) ||
(last_spa_suspended && !suspended)) {
mutex_enter(&mmp->mmp_io_lock);
mmp->mmp_last_write = gethrtime();
mutex_exit(&mmp->mmp_io_lock);
} else if (last_spa_multihost && !multihost) {
mutex_enter(&mmp->mmp_io_lock);
mmp->mmp_delay = 0;
mutex_exit(&mmp->mmp_io_lock);
}
last_spa_multihost = multihost;
last_spa_suspended = suspended;
/*
* Smooth max_fail_ns when its factors are decreased, because
* making (max_fail_ns < mmp_interval) results in the pool being
* immediately suspended before writes can occur at the new
* higher frequency.
*/
if ((mmp_interval * mmp_fail_intervals) < max_fail_ns) {
max_fail_ns = ((31 * max_fail_ns) + (mmp_interval *
mmp_fail_intervals)) / 32;
} else {
max_fail_ns = mmp_interval * mmp_fail_intervals;
}
/*
* Suspend the pool if no MMP write has succeeded in over
* mmp_interval * mmp_fail_intervals nanoseconds.
*/
if (!suspended && mmp_fail_intervals && multihost &&
(start - mmp->mmp_last_write) > max_fail_ns) {
cmn_err(CE_WARN, "MMP writes to pool '%s' have not "
"succeeded in over %llus; suspending pool",
spa_name(spa),
NSEC2SEC(start - mmp->mmp_last_write));
zio_suspend(spa, NULL);
}
if (multihost)
mmp_write_uberblock(spa);
CALLB_CPR_SAFE_BEGIN(&cpr);
(void) cv_timedwait_sig(&mmp->mmp_thread_cv,
&mmp->mmp_thread_lock, ddi_get_lbolt() +
((next_time - gethrtime()) / (NANOSEC / hz)));
CALLB_CPR_SAFE_END(&cpr, &mmp->mmp_thread_lock);
}
/* Outstanding writes are allowed to complete. */
if (mmp->mmp_zio_root)
zio_wait(mmp->mmp_zio_root);
mmp->mmp_zio_root = NULL;
mmp_thread_exit(mmp, &mmp->mmp_thread, &cpr);
}
