Merge pull request openzfs#816 from ahrens/merge

Merge with `Allow MMP to bypass waiting for other threads`
pcd1193182 · Apr 21, 2023 · bbc2095 · bbc2095
2 parents 1f28ff8 + 8b8b49f
commit bbc2095
Show file tree

Hide file tree

Showing 4 changed files with 36 additions and 6 deletions.
diff --git a/include/sys/spa.h b/include/sys/spa.h
@@ -980,6 +980,8 @@ extern int spa_import_progress_set_state(uint64_t pool_guid,
 extern int spa_config_tryenter(spa_t *spa, int locks, const void *tag,
     krw_t rw);
 extern void spa_config_enter(spa_t *spa, int locks, const void *tag, krw_t rw);
+extern void spa_config_enter_mmp(spa_t *spa, int locks, const void *tag,
+    krw_t rw);
 extern void spa_config_exit(spa_t *spa, int locks, const void *tag);
 extern int spa_config_held(spa_t *spa, int locks, krw_t rw);
 extern void spa_config_enter_read_priority(spa_t *, int);

diff --git a/module/zfs/mmp.c b/module/zfs/mmp.c
@@ -445,7 +445,7 @@ mmp_write_uberblock(spa_t *spa)
 	uint64_t offset;
 
 	hrtime_t lock_acquire_time = gethrtime();
-	spa_config_enter(spa, SCL_STATE, mmp_tag, RW_READER);
+	spa_config_enter_mmp(spa, SCL_STATE, mmp_tag, RW_READER);
 	lock_acquire_time = gethrtime() - lock_acquire_time;
 	if (lock_acquire_time > (MSEC2NSEC(MMP_MIN_INTERVAL) / 10))
 		zfs_dbgmsg("MMP SCL_STATE acquisition pool '%s' took %llu ns "

diff --git a/module/zfs/spa_misc.c b/module/zfs/spa_misc.c
@@ -551,8 +551,9 @@ spa_config_enter_read_priority(spa_t *spa, int locks)
 	}
 }
 
-void
-spa_config_enter(spa_t *spa, int locks, const void *tag, krw_t rw)
+static void
+spa_config_enter_impl(spa_t *spa, int locks, const void *tag, krw_t rw,
+    int mmp_flag)
 {
 	(void) tag;
 	int wlocks_held = 0;
@@ -570,7 +571,8 @@ spa_config_enter(spa_t *spa, int locks, const void *tag, krw_t rw)
 			continue;
 		mutex_enter(&scl->scl_lock);
 		if (rw == RW_READER) {
-			while (scl->scl_writer || scl->scl_write_wanted) {
+			while (scl->scl_writer ||
+			    (!mmp_flag && scl->scl_write_wanted)) {
 				cv_wait(&scl->scl_cv, &scl->scl_lock);
 			}
 		} else {
@@ -598,6 +600,27 @@ spa_config_enter(spa_t *spa, int locks, const void *tag, krw_t rw)
 	ASSERT3U(wlocks_held, <=, locks);
 }
 
+void
+spa_config_enter(spa_t *spa, int locks, const void *tag, krw_t rw)
+{
+	spa_config_enter_impl(spa, locks, tag, rw, 0);
+}
+
+/*
+ * The spa_config_enter_mmp() allows the mmp thread to cut in front of
+ * outstanding write lock requests. This is needed since the mmp updates are
+ * time sensitive and failure to service them promptly will result in a
+ * suspended pool. This pool suspension has been seen in practice when there is
+ * a single disk in a pool that is responding slowly and presumably about to
+ * fail.
+ */
+
+void
+spa_config_enter_mmp(spa_t *spa, int locks, const void *tag, krw_t rw)
+{
+	spa_config_enter_impl(spa, locks, tag, rw, 1);
+}
+
 void
 spa_config_exit(spa_t *spa, int locks, const void *tag)
 {

diff --git a/tests/zfs-tests/tests/functional/rsend/send-c_volume.ksh b/tests/zfs-tests/tests/functional/rsend/send-c_volume.ksh
@@ -29,6 +29,7 @@
 
 function cleanup
 {
+	rm $BACKDIR/copy
 	log_must_busy zfs destroy -r $vol
 	cleanup_pool $POOL2
 }
@@ -60,7 +61,9 @@ log_must eval "zfs recv -d $POOL2 <$BACKDIR/full"
 
 verify_stream_size $BACKDIR/full $vol
 verify_stream_size $BACKDIR/full $vol2
-md5=$(dd if=$voldev2 bs=1024k count=$megs 2>/dev/null | md5digest)
+block_device_wait $voldev2
+log_must dd if=$voldev2 of=$BACKDIR/copy bs=1024k count=$megs
+md5=$(md5digest $BACKDIR/copy)
 [[ $md5 = $md5_1 ]] || log_fail "md5 mismatch: $md5 != $md5_1"
 
 # Repeat, for an incremental send
@@ -72,7 +75,9 @@ log_must eval "zfs recv -d $POOL2 <$BACKDIR/inc"
 
 verify_stream_size $BACKDIR/inc $vol 90 $vol@snap
 verify_stream_size $BACKDIR/inc $vol2 90 $vol2@snap
-md5=$(dd skip=$megs if=$voldev2 bs=1024k count=$megs 2>/dev/null | md5digest)
+block_device_wait $voldev2
+log_must dd skip=$megs if=$voldev2 of=$BACKDIR/copy bs=1024k count=$megs
+md5=$(md5digest $BACKDIR/copy)
 [[ $md5 = $md5_2 ]] || log_fail "md5 mismatch: $md5 != $md5_2"
 
 log_pass "Verify compressed send works with volumes"