OpenZFS: MFV 2.0-rc3-gfc5966

- Annotate FreeBSD sysctls with CTLFLAG_MPSAFE
- Reduce stack usage of Lua
- Don't save user FPU context in kernel threads
- Add support for procfs_list
- Code cleanup in zio_crypt
- Add DB_RF_NOPREFETCH to dbuf_read()s in dnode.c
- Drop references when skipping dmu_send due to EXDEV
- Eliminate gratuitous bzeroing in dbuf_stats_hash_table_data
- Fix legacy compat for platform IOCs

8 files changed, 47 insertions, 43 deletions

diff --git a/sys/contrib/openzfs/module/zfs/arc.c b/sys/contrib/openzfs/module/zfs/arc.c
index 7a499298f75c..3ba198380733 100644
--- a/sys/contrib/openzfs/module/zfs/arc.c
+++ b/sys/contrib/openzfs/module/zfs/arc.c
@@ -308,6 +308,7 @@
 #include <sys/aggsum.h>
 #include <cityhash.h>
 #include <sys/vdev_trim.h>
+#include <sys/zstd/zstd.h>
 
 #ifndef _KERNEL
 /* set with ZFS_DEBUG=watch, to enable watchpoints on frozen buffers */
@@ -4859,6 +4860,7 @@ static boolean_t
 arc_reap_cb_check(void *arg, zthr_t *zthr)
 {
 	int64_t free_memory = arc_available_memory();
+	static int reap_cb_check_counter = 0;
 
 	/*
 	 * If a kmem reap is already active, don't schedule more.  We must
@@ -4883,6 +4885,14 @@ arc_reap_cb_check(void *arg, zthr_t *zthr)
 		arc_no_grow = B_FALSE;
 	}
 
+	/*
+	 * Called unconditionally every 60 seconds to reclaim unused
+	 * zstd compression and decompression context. This is done
+	 * here to avoid the need for an independent thread.
+	 */
+	if (!((reap_cb_check_counter++) % 60))
+		zfs_zstd_cache_reap_now();
+
 	return (B_FALSE);
 }
 
diff --git a/sys/contrib/openzfs/module/zfs/dbuf_stats.c b/sys/contrib/openzfs/module/zfs/dbuf_stats.c
index a2f3c580ee6c..12bb568a08cc 100644
--- a/sys/contrib/openzfs/module/zfs/dbuf_stats.c
+++ b/sys/contrib/openzfs/module/zfs/dbuf_stats.c
@@ -134,7 +134,8 @@ dbuf_stats_hash_table_data(char *buf, size_t size, void *data)
 
 	ASSERT3S(dsh->idx, >=, 0);
 	ASSERT3S(dsh->idx, <=, h->hash_table_mask);
-	memset(buf, 0, size);
+	if (size)
+		buf[0] = 0;
 
 	mutex_enter(DBUF_HASH_MUTEX(h, dsh->idx));
 	for (db = h->hash_table[dsh->idx]; db != NULL; db = db->db_hash_next) {
diff --git a/sys/contrib/openzfs/module/zfs/dmu_send.c b/sys/contrib/openzfs/module/zfs/dmu_send.c
index 33e99c2e02ab..9480c8b75497 100644
--- a/sys/contrib/openzfs/module/zfs/dmu_send.c
+++ b/sys/contrib/openzfs/module/zfs/dmu_send.c
@@ -643,7 +643,7 @@ dump_freeobjects(dmu_send_cookie_t *dscp, uint64_t firstobj, uint64_t numobjs)
 	 * receiving side.
 	 */
 	if (maxobj > 0) {
-		if (maxobj < firstobj)
+		if (maxobj <= firstobj)
 			return (0);
 
 		if (maxobj < firstobj + numobjs)
@@ -663,8 +663,6 @@ dump_freeobjects(dmu_send_cookie_t *dscp, uint64_t firstobj, uint64_t numobjs)
 			return (SET_ERROR(EINTR));
 		dscp->dsc_pending_op = PENDING_NONE;
 	}
-	if (numobjs == 0)
-		numobjs = UINT64_MAX - firstobj;
 
 	if (dscp->dsc_pending_op == PENDING_FREEOBJECTS) {
 		/*
@@ -2686,12 +2684,15 @@ dmu_send_obj(const char *pool, uint64_t tosnap, uint64_t fromsnap,
 			bcopy(fromredact, dspp.fromredactsnaps, size);
 		}
 
-		if (!dsl_dataset_is_before(dspp.to_ds, fromds, 0)) {
+		boolean_t is_before =
+		    dsl_dataset_is_before(dspp.to_ds, fromds, 0);
+		dspp.is_clone = (dspp.to_ds->ds_dir !=
+		    fromds->ds_dir);
+		dsl_dataset_rele(fromds, FTAG);
+		if (!is_before) {
+			dsl_pool_rele(dspp.dp, FTAG);
 			err = SET_ERROR(EXDEV);
 		} else {
-			dspp.is_clone = (dspp.to_ds->ds_dir !=
-			    fromds->ds_dir);
-			dsl_dataset_rele(fromds, FTAG);
 			err = dmu_send_impl(&dspp);
 		}
 	} else {
diff --git a/sys/contrib/openzfs/module/zfs/dnode.c b/sys/contrib/openzfs/module/zfs/dnode.c
index 30d20bfefa12..23364dbae897 100644
--- a/sys/contrib/openzfs/module/zfs/dnode.c
+++ b/sys/contrib/openzfs/module/zfs/dnode.c
@@ -1355,7 +1355,8 @@ dnode_hold_impl(objset_t *os, uint64_t object, int flag, int slots,
 	 * We do not need to decrypt to read the dnode so it doesn't matter
 	 * if we get the encrypted or decrypted version.
 	 */
-	err = dbuf_read(db, NULL, DB_RF_CANFAIL | DB_RF_NO_DECRYPT);
+	err = dbuf_read(db, NULL, DB_RF_CANFAIL |
+	    DB_RF_NO_DECRYPT | DB_RF_NOPREFETCH);
 	if (err) {
 		DNODE_STAT_BUMP(dnode_hold_dbuf_read);
 		dbuf_rele(db, FTAG);
@@ -2396,7 +2397,8 @@ dnode_next_offset_level(dnode_t *dn, int flags, uint64_t *offset,
 			return (SET_ERROR(ESRCH));
 		}
 		error = dbuf_read(db, NULL,
-		    DB_RF_CANFAIL | DB_RF_HAVESTRUCT | DB_RF_NO_DECRYPT);
+		    DB_RF_CANFAIL | DB_RF_HAVESTRUCT |
+		    DB_RF_NO_DECRYPT | DB_RF_NOPREFETCH);
 		if (error) {
 			dbuf_rele(db, FTAG);
 			return (error);
diff --git a/sys/contrib/openzfs/module/zfs/dsl_crypt.c b/sys/contrib/openzfs/module/zfs/dsl_crypt.c
index 33c21e86c4d7..26d4c2fe7e33 100644
--- a/sys/contrib/openzfs/module/zfs/dsl_crypt.c
+++ b/sys/contrib/openzfs/module/zfs/dsl_crypt.c
@@ -235,11 +235,7 @@ dsl_crypto_params_create_nvlist(dcp_cmd_t cmd, nvlist_t *props,
 	return (0);
 
 error:
-	if (wkey != NULL)
-		dsl_wrapping_key_free(wkey);
-	if (dcp != NULL)
-		kmem_free(dcp, sizeof (dsl_crypto_params_t));
-
+	kmem_free(dcp, sizeof (dsl_crypto_params_t));
 	*dcp_out = NULL;
 	return (ret);
 }
diff --git a/sys/contrib/openzfs/module/zfs/spa_misc.c b/sys/contrib/openzfs/module/zfs/spa_misc.c
index 41f0ddbde288..04210472886c 100644
--- a/sys/contrib/openzfs/module/zfs/spa_misc.c
+++ b/sys/contrib/openzfs/module/zfs/spa_misc.c
@@ -2169,6 +2169,7 @@ spa_import_progress_init(void)
 	    spa_import_progress_list;
 
 	procfs_list_install("zfs",
+	    NULL,
 	    "import_progress",
 	    0644,
 	    &spa_import_progress_list->procfs_list,
diff --git a/sys/contrib/openzfs/module/zfs/spa_stats.c b/sys/contrib/openzfs/module/zfs/spa_stats.c
index cf0be3c45dc8..c3eacc14239e 100644
--- a/sys/contrib/openzfs/module/zfs/spa_stats.c
+++ b/sys/contrib/openzfs/module/zfs/spa_stats.c
@@ -122,14 +122,11 @@ static void
 spa_read_history_init(spa_t *spa)
 {
 	spa_history_list_t *shl = &spa->spa_stats.read_history;
-	char *module;
 
 	shl->size = 0;
-
-	module = kmem_asprintf("zfs/%s", spa_name(spa));
-
 	shl->procfs_list.pl_private = shl;
-	procfs_list_install(module,
+	procfs_list_install("zfs",
+	    spa_name(spa),
 	    "reads",
 	    0600,
 	    &shl->procfs_list,
@@ -137,8 +134,6 @@ spa_read_history_init(spa_t *spa)
 	    spa_read_history_show_header,
 	    spa_read_history_clear,
 	    offsetof(spa_read_history_t, srh_node));
-
-	kmem_strfree(module);
 }
 
 static void
@@ -293,14 +288,11 @@ static void
 spa_txg_history_init(spa_t *spa)
 {
 	spa_history_list_t *shl = &spa->spa_stats.txg_history;
-	char *module;
 
 	shl->size = 0;
-
-	module = kmem_asprintf("zfs/%s", spa_name(spa));
-
 	shl->procfs_list.pl_private = shl;
-	procfs_list_install(module,
+	procfs_list_install("zfs",
+	    spa_name(spa),
 	    "txgs",
 	    0644,
 	    &shl->procfs_list,
@@ -308,8 +300,6 @@ spa_txg_history_init(spa_t *spa)
 	    spa_txg_history_show_header,
 	    spa_txg_history_clear,
 	    offsetof(spa_txg_history_t, sth_node));
-
-	kmem_strfree(module);
 }
 
 static void
@@ -699,14 +689,12 @@ static void
 spa_mmp_history_init(spa_t *spa)
 {
 	spa_history_list_t *shl = &spa->spa_stats.mmp_history;
-	char *module;
 
 	shl->size = 0;
 
-	module = kmem_asprintf("zfs/%s", spa_name(spa));
-
 	shl->procfs_list.pl_private = shl;
-	procfs_list_install(module,
+	procfs_list_install("zfs",
+	    spa_name(spa),
 	    "multihost",
 	    0644,
 	    &shl->procfs_list,
@@ -714,8 +702,6 @@ spa_mmp_history_init(spa_t *spa)
 	    spa_mmp_history_show_header,
 	    spa_mmp_history_clear,
 	    offsetof(spa_mmp_history_t, smh_node));
-
-	kmem_strfree(module);
 }
 
 static void
diff --git a/sys/contrib/openzfs/module/zfs/zfs_log.c b/sys/contrib/openzfs/module/zfs/zfs_log.c
index 4eae855f4274..fb44007fefc3 100644
--- a/sys/contrib/openzfs/module/zfs/zfs_log.c
+++ b/sys/contrib/openzfs/module/zfs/zfs_log.c
@@ -584,15 +584,22 @@ zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype,
 		    (wr_state == WR_COPIED ? len : 0));
 		lr = (lr_write_t *)&itx->itx_lr;
 
-		DB_DNODE_ENTER(db);
-		if (wr_state == WR_COPIED && dmu_read_by_dnode(DB_DNODE(db),
-		    off, len, lr + 1, DMU_READ_NO_PREFETCH) != 0) {
-			zil_itx_destroy(itx);
-			itx = zil_itx_create(txtype, sizeof (*lr));
-			lr = (lr_write_t *)&itx->itx_lr;
-			wr_state = WR_NEED_COPY;
+		/*
+		 * For WR_COPIED records, copy the data into the lr_write_t.
+		 */
+		if (wr_state == WR_COPIED) {
+			int err;
+			DB_DNODE_ENTER(db);
+			err = dmu_read_by_dnode(DB_DNODE(db), off, len, lr + 1,
+			    DMU_READ_NO_PREFETCH);
+			if (err != 0) {
+				zil_itx_destroy(itx);
+				itx = zil_itx_create(txtype, sizeof (*lr));
+				lr = (lr_write_t *)&itx->itx_lr;
+				wr_state = WR_NEED_COPY;
+			}
+			DB_DNODE_EXIT(db);
 		}
-		DB_DNODE_EXIT(db);
 
 		itx->itx_wr_state = wr_state;
 		lr->lr_foid = zp->z_id;