svn commit: r247844 - vendor-sys/illumos/dist/uts/common/fs/zfs vendor-sys/illumos/dist/uts/common/fs/zfs/sys vendor/illumos/dist/lib/libzpool/common vendor/illumos/dist/lib/libzpool/common/sys
Martin Matuska
mm at FreeBSD.org
Tue Mar 5 15:33:30 UTC 2013
Author: mm
Date: Tue Mar 5 15:33:28 2013
New Revision: 247844
URL: http://svnweb.freebsd.org/changeset/base/247844
Log:
Update vendor/illumos/dist and vendor-sys/illumos/dist
to illumos-gate 13975:ef6409bc370f
Illumos ZFS issues:
3582 zfs_delay() should support a variable resolution
3584 DTrace sdt probes for ZFS txg states
Modified:
vendor-sys/illumos/dist/uts/common/fs/zfs/dsl_dir.c
vendor-sys/illumos/dist/uts/common/fs/zfs/dsl_pool.c
vendor-sys/illumos/dist/uts/common/fs/zfs/dsl_scan.c
vendor-sys/illumos/dist/uts/common/fs/zfs/spa_misc.c
vendor-sys/illumos/dist/uts/common/fs/zfs/sys/txg.h
vendor-sys/illumos/dist/uts/common/fs/zfs/sys/txg_impl.h
vendor-sys/illumos/dist/uts/common/fs/zfs/txg.c
Changes in other areas also in this revision:
Modified:
vendor/illumos/dist/lib/libzpool/common/kernel.c
vendor/illumos/dist/lib/libzpool/common/sys/zfs_context.h
Modified: vendor-sys/illumos/dist/uts/common/fs/zfs/dsl_dir.c
==============================================================================
--- vendor-sys/illumos/dist/uts/common/fs/zfs/dsl_dir.c Tue Mar 5 14:08:36 2013 (r247843)
+++ vendor-sys/illumos/dist/uts/common/fs/zfs/dsl_dir.c Tue Mar 5 15:33:28 2013 (r247844)
@@ -738,7 +738,8 @@ dsl_dir_tempreserve_space(dsl_dir_t *dd,
err = dsl_pool_tempreserve_space(dd->dd_pool, asize, tx);
} else {
if (err == EAGAIN) {
- txg_delay(dd->dd_pool, tx->tx_txg, 1);
+ txg_delay(dd->dd_pool, tx->tx_txg,
+ MSEC2NSEC(10), MSEC2NSEC(10));
err = ERESTART;
}
dsl_pool_memory_pressure(dd->dd_pool);
Modified: vendor-sys/illumos/dist/uts/common/fs/zfs/dsl_pool.c
==============================================================================
--- vendor-sys/illumos/dist/uts/common/fs/zfs/dsl_pool.c Tue Mar 5 14:08:36 2013 (r247843)
+++ vendor-sys/illumos/dist/uts/common/fs/zfs/dsl_pool.c Tue Mar 5 15:33:28 2013 (r247844)
@@ -58,6 +58,9 @@ kmutex_t zfs_write_limit_lock;
static pgcnt_t old_physmem = 0;
+hrtime_t zfs_throttle_delay = MSEC2NSEC(10);
+hrtime_t zfs_throttle_resolution = MSEC2NSEC(10);
+
int
dsl_pool_open_special_dir(dsl_pool_t *dp, const char *name, dsl_dir_t **ddp)
{
@@ -511,12 +514,13 @@ dsl_pool_sync(dsl_pool_t *dp, uint64_t t
* Weight the throughput calculation towards the current value:
* thru = 3/4 old_thru + 1/4 new_thru
*
- * Note: write_time is in nanosecs, so write_time/MICROSEC
- * yields millisecs
+ * Note: write_time is in nanosecs while dp_throughput is expressed in
+ * bytes per millisecond.
*/
ASSERT(zfs_write_limit_min > 0);
- if (data_written > zfs_write_limit_min / 8 && write_time > MICROSEC) {
- uint64_t throughput = data_written / (write_time / MICROSEC);
+ if (data_written > zfs_write_limit_min / 8 &&
+ write_time > MSEC2NSEC(1)) {
+ uint64_t throughput = data_written / NSEC2MSEC(write_time);
if (dp->dp_throughput)
dp->dp_throughput = throughput / 4 +
@@ -614,8 +618,10 @@ dsl_pool_tempreserve_space(dsl_pool_t *d
* the caller 1 clock tick. This will slow down the "fill"
* rate until the sync process can catch up with us.
*/
- if (reserved && reserved > (write_limit - (write_limit >> 3)))
- txg_delay(dp, tx->tx_txg, 1);
+ if (reserved && reserved > (write_limit - (write_limit >> 3))) {
+ txg_delay(dp, tx->tx_txg, zfs_throttle_delay,
+ zfs_throttle_resolution);
+ }
return (0);
}
Modified: vendor-sys/illumos/dist/uts/common/fs/zfs/dsl_scan.c
==============================================================================
--- vendor-sys/illumos/dist/uts/common/fs/zfs/dsl_scan.c Tue Mar 5 14:08:36 2013 (r247843)
+++ vendor-sys/illumos/dist/uts/common/fs/zfs/dsl_scan.c Tue Mar 5 15:33:28 2013 (r247844)
@@ -403,7 +403,7 @@ dsl_scan_check_pause(dsl_scan_t *scn, co
zfs_resilver_min_time_ms : zfs_scan_min_time_ms;
elapsed_nanosecs = gethrtime() - scn->scn_sync_start_time;
if (elapsed_nanosecs / NANOSEC > zfs_txg_timeout ||
- (elapsed_nanosecs / MICROSEC > mintime &&
+ (NSEC2MSEC(elapsed_nanosecs) > mintime &&
txg_sync_waiting(scn->scn_dp)) ||
spa_shutting_down(scn->scn_dp->dp_spa)) {
if (zb) {
@@ -1308,7 +1308,7 @@ dsl_scan_free_should_pause(dsl_scan_t *s
elapsed_nanosecs = gethrtime() - scn->scn_sync_start_time;
return (elapsed_nanosecs / NANOSEC > zfs_txg_timeout ||
- (elapsed_nanosecs / MICROSEC > zfs_free_min_time_ms &&
+ (NSEC2MSEC(elapsed_nanosecs) > zfs_free_min_time_ms &&
txg_sync_waiting(scn->scn_dp)) ||
spa_shutting_down(scn->scn_dp->dp_spa));
}
@@ -1433,7 +1433,7 @@ dsl_scan_sync(dsl_pool_t *dp, dmu_tx_t *
"free_bpobj/bptree txg %llu",
(longlong_t)scn->scn_visited_this_txg,
(longlong_t)
- (gethrtime() - scn->scn_sync_start_time) / MICROSEC,
+ NSEC2MSEC(gethrtime() - scn->scn_sync_start_time),
(longlong_t)tx->tx_txg);
scn->scn_visited_this_txg = 0;
/*
@@ -1481,7 +1481,7 @@ dsl_scan_sync(dsl_pool_t *dp, dmu_tx_t *
zfs_dbgmsg("visited %llu blocks in %llums",
(longlong_t)scn->scn_visited_this_txg,
- (longlong_t)(gethrtime() - scn->scn_sync_start_time) / MICROSEC);
+ (longlong_t)NSEC2MSEC(gethrtime() - scn->scn_sync_start_time));
if (!scn->scn_pausing) {
/* finished with scan. */
Modified: vendor-sys/illumos/dist/uts/common/fs/zfs/spa_misc.c
==============================================================================
--- vendor-sys/illumos/dist/uts/common/fs/zfs/spa_misc.c Tue Mar 5 14:08:36 2013 (r247843)
+++ vendor-sys/illumos/dist/uts/common/fs/zfs/spa_misc.c Tue Mar 5 15:33:28 2013 (r247844)
@@ -499,8 +499,8 @@ spa_add(const char *name, nvlist_t *conf
hdlr.cyh_arg = spa;
hdlr.cyh_level = CY_LOW_LEVEL;
- spa->spa_deadman_synctime = zfs_deadman_synctime *
- zfs_txg_synctime_ms * MICROSEC;
+ spa->spa_deadman_synctime = MSEC2NSEC(zfs_deadman_synctime *
+ zfs_txg_synctime_ms);
/*
* This determines how often we need to check for hung I/Os after
@@ -508,7 +508,7 @@ spa_add(const char *name, nvlist_t *conf
* an expensive operation we don't want to check too frequently.
* Instead wait for 5 synctimes before checking again.
*/
- when.cyt_interval = 5ULL * zfs_txg_synctime_ms * MICROSEC;
+ when.cyt_interval = MSEC2NSEC(5 * zfs_txg_synctime_ms);
when.cyt_when = CY_INFINITY;
mutex_enter(&cpu_lock);
spa->spa_deadman_cycid = cyclic_add(&hdlr, &when);
Modified: vendor-sys/illumos/dist/uts/common/fs/zfs/sys/txg.h
==============================================================================
--- vendor-sys/illumos/dist/uts/common/fs/zfs/sys/txg.h Tue Mar 5 14:08:36 2013 (r247843)
+++ vendor-sys/illumos/dist/uts/common/fs/zfs/sys/txg.h Tue Mar 5 15:33:28 2013 (r247844)
@@ -74,13 +74,8 @@ extern void txg_rele_to_quiesce(txg_hand
extern void txg_rele_to_sync(txg_handle_t *txghp);
extern void txg_register_callbacks(txg_handle_t *txghp, list_t *tx_callbacks);
-/*
- * Delay the caller by the specified number of ticks or until
- * the txg closes (whichever comes first). This is intended
- * to be used to throttle writers when the system nears its
- * capacity.
- */
-extern void txg_delay(struct dsl_pool *dp, uint64_t txg, int ticks);
+extern void txg_delay(struct dsl_pool *dp, uint64_t txg, hrtime_t delta,
+ hrtime_t resolution);
/*
* Wait until the given transaction group has finished syncing.
Modified: vendor-sys/illumos/dist/uts/common/fs/zfs/sys/txg_impl.h
==============================================================================
--- vendor-sys/illumos/dist/uts/common/fs/zfs/sys/txg_impl.h Tue Mar 5 14:08:36 2013 (r247843)
+++ vendor-sys/illumos/dist/uts/common/fs/zfs/sys/txg_impl.h Tue Mar 5 15:33:28 2013 (r247844)
@@ -23,6 +23,10 @@
* Use is subject to license terms.
*/
+/*
+ * Copyright (c) 2012 by Delphix. All rights reserved.
+ */
+
#ifndef _SYS_TXG_IMPL_H
#define _SYS_TXG_IMPL_H
@@ -36,14 +40,14 @@ extern "C" {
struct tx_cpu {
kmutex_t tc_lock;
kcondvar_t tc_cv[TXG_SIZE];
- uint64_t tc_count[TXG_SIZE];
+ uint64_t tc_count[TXG_SIZE]; /* tx hold count on each txg */
list_t tc_callbacks[TXG_SIZE]; /* commit cb list */
- char tc_pad[16];
+ char tc_pad[16]; /* pad to fill 3 cache lines */
};
typedef struct tx_state {
- tx_cpu_t *tx_cpu; /* protects right to enter txg */
- kmutex_t tx_sync_lock; /* protects tx_state_t */
+ tx_cpu_t *tx_cpu; /* protects access to tx_open_txg */
+ kmutex_t tx_sync_lock; /* protects the rest of this struct */
uint64_t tx_open_txg; /* currently open txg id */
uint64_t tx_quiesced_txg; /* quiesced txg waiting for sync */
uint64_t tx_syncing_txg; /* currently syncing txg id */
Modified: vendor-sys/illumos/dist/uts/common/fs/zfs/txg.c
==============================================================================
--- vendor-sys/illumos/dist/uts/common/fs/zfs/txg.c Tue Mar 5 14:08:36 2013 (r247843)
+++ vendor-sys/illumos/dist/uts/common/fs/zfs/txg.c Tue Mar 5 15:33:28 2013 (r247844)
@@ -232,7 +232,7 @@ txg_thread_exit(tx_state_t *tx, callb_cp
}
static void
-txg_thread_wait(tx_state_t *tx, callb_cpr_t *cpr, kcondvar_t *cv, uint64_t time)
+txg_thread_wait(tx_state_t *tx, callb_cpr_t *cpr, kcondvar_t *cv, clock_t time)
{
CALLB_CPR_SAFE_BEGIN(cpr);
@@ -353,6 +353,9 @@ txg_quiesce(dsl_pool_t *dp, uint64_t txg
ASSERT(txg == tx->tx_open_txg);
tx->tx_open_txg++;
+ DTRACE_PROBE2(txg__quiescing, dsl_pool_t *, dp, uint64_t, txg);
+ DTRACE_PROBE2(txg__opened, dsl_pool_t *, dp, uint64_t, tx->tx_open_txg);
+
/*
* Now that we've incremented tx_open_txg, we can let threads
* enter the next transaction group.
@@ -475,6 +478,7 @@ txg_sync_thread(dsl_pool_t *dp)
txg = tx->tx_quiesced_txg;
tx->tx_quiesced_txg = 0;
tx->tx_syncing_txg = txg;
+ DTRACE_PROBE2(txg__syncing, dsl_pool_t *, dp, uint64_t, txg);
cv_broadcast(&tx->tx_quiesce_more_cv);
dprintf("txg=%llu quiesce_txg=%llu sync_txg=%llu\n",
@@ -488,6 +492,7 @@ txg_sync_thread(dsl_pool_t *dp)
mutex_enter(&tx->tx_sync_lock);
tx->tx_synced_txg = txg;
tx->tx_syncing_txg = 0;
+ DTRACE_PROBE2(txg__synced, dsl_pool_t *, dp, uint64_t, txg);
cv_broadcast(&tx->tx_sync_done_cv);
/*
@@ -536,21 +541,22 @@ txg_quiesce_thread(dsl_pool_t *dp)
*/
dprintf("quiesce done, handing off txg %llu\n", txg);
tx->tx_quiesced_txg = txg;
+ DTRACE_PROBE2(txg__quiesced, dsl_pool_t *, dp, uint64_t, txg);
cv_broadcast(&tx->tx_sync_more_cv);
cv_broadcast(&tx->tx_quiesce_done_cv);
}
}
/*
- * Delay this thread by 'ticks' if we are still in the open transaction
- * group and there is already a waiting txg quiesing or quiesced. Abort
- * the delay if this txg stalls or enters the quiesing state.
+ * Delay this thread by delay nanoseconds if we are still in the open
+ * transaction group and there is already a waiting txg quiesing or quiesced.
+ * Abort the delay if this txg stalls or enters the quiesing state.
*/
void
-txg_delay(dsl_pool_t *dp, uint64_t txg, int ticks)
+txg_delay(dsl_pool_t *dp, uint64_t txg, hrtime_t delay, hrtime_t resolution)
{
tx_state_t *tx = &dp->dp_tx;
- clock_t timeout = ddi_get_lbolt() + ticks;
+ hrtime_t start = gethrtime();
/* don't delay if this txg could transition to quiesing immediately */
if (tx->tx_open_txg > txg ||
@@ -563,10 +569,11 @@ txg_delay(dsl_pool_t *dp, uint64_t txg,
return;
}
- while (ddi_get_lbolt() < timeout &&
- tx->tx_syncing_txg < txg-1 && !txg_stalled(dp))
- (void) cv_timedwait(&tx->tx_quiesce_more_cv, &tx->tx_sync_lock,
- timeout);
+ while (gethrtime() - start < delay &&
+ tx->tx_syncing_txg < txg-1 && !txg_stalled(dp)) {
+ (void) cv_timedwait_hires(&tx->tx_quiesce_more_cv,
+ &tx->tx_sync_lock, delay, resolution, 0);
+ }
mutex_exit(&tx->tx_sync_lock);
}
More information about the svn-src-all
mailing list