Implement physical rewrites

amotin · amotin · commit 6f4269254b18 · 2025-07-24T21:22:22.000-04:00
Based on previous commit this implements `zfs rewrite -P` flag,
making ZFS to keep block logical birth times while rewriting
files.  It should exclude the rewritten blocks from incremental
sends, snapshot diffs, etc.  Snapshots space usage same time will
reflect the additional space usage from newly allocated blocks.

Since this begins to use new "rewrite" flag in the block pointers,
this commit introduces a new read-compatible per-dataset feature
physical_rewrite.  It must be enabled for the command to not fail,
it is activated on first use and deactivated on deletion of the
last affected dataset.

Signed-off-by:  Alexander Motin &lt;alexander.motin@TrueNAS.com&gt;
diff --git a/cmd/zfs/zfs_main.c b/cmd/zfs/zfs_main.c
@@ -440,7 +440,7 @@ get_usage(zfs_help_t idx)
 		return (gettext("\tredact <snapshot> <bookmark> "
 		    "<redaction_snapshot> ...\n"));
 	case HELP_REWRITE:
-		return (gettext("\trewrite [-rvx] [-o <offset>] [-l <length>] "
+		return (gettext("\trewrite [-Prvx] [-o <offset>] [-l <length>] "
 		    "<directory|file ...>\n"));
 	case HELP_JAIL:
 		return (gettext("\tjail <jailid|jailname> <filesystem>\n"));
@@ -9177,8 +9177,11 @@ zfs_do_rewrite(int argc, char **argv)
 	zfs_rewrite_args_t args;
 	memset(&args, 0, sizeof (args));
 
-	while ((c = getopt(argc, argv, "l:o:rvx")) != -1) {
+	while ((c = getopt(argc, argv, "Pl:o:rvx")) != -1) {
 		switch (c) {
+		case 'P':
+			args.flags |= ZFS_REWRITE_PHYSICAL;
+			break;
 		case 'l':
 			args.len = strtoll(optarg, NULL, 0);
 			break;
diff --git a/include/sys/dbuf.h b/include/sys/dbuf.h
@@ -164,6 +164,7 @@ typedef struct dbuf_dirty_record {
 			boolean_t dr_nopwrite;
 			boolean_t dr_brtwrite;
 			boolean_t dr_diowrite;
+			boolean_t dr_rewrite;
 			boolean_t dr_has_raw_params;
 
 			/* Override and raw params are mutually exclusive. */
diff --git a/include/sys/dmu.h b/include/sys/dmu.h
@@ -822,6 +822,7 @@ struct blkptr *dmu_buf_get_blkptr(dmu_buf_t *db);
  */
 void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx);
 void dmu_buf_will_dirty_flags(dmu_buf_t *db, dmu_tx_t *tx, dmu_flags_t flags);
+void dmu_buf_will_rewrite(dmu_buf_t *db, dmu_tx_t *tx);
 boolean_t dmu_buf_is_dirty(dmu_buf_t *db, dmu_tx_t *tx);
 void dmu_buf_set_crypt_params(dmu_buf_t *db_fake, boolean_t byteorder,
     const uint8_t *salt, const uint8_t *iv, const uint8_t *mac, dmu_tx_t *tx);
diff --git a/include/sys/fs/zfs.h b/include/sys/fs/zfs.h
@@ -1627,6 +1627,9 @@ typedef struct zfs_rewrite_args {
 	uint64_t	arg;
 } zfs_rewrite_args_t;
 
+/* zfs_rewrite_args flags */
+#define	ZFS_REWRITE_PHYSICAL	0x1	/* Preserve logical birth time. */
+
 #define	ZFS_IOC_REWRITE		_IOW(0x83, 3, zfs_rewrite_args_t)
 
 /*
diff --git a/include/sys/zio.h b/include/sys/zio.h
@@ -374,6 +374,7 @@ typedef struct zio_prop {
 	boolean_t		zp_encrypt;
 	boolean_t		zp_byteorder;
 	boolean_t		zp_direct_write;
+	boolean_t		zp_rewrite;
 	uint8_t			zp_salt[ZIO_DATA_SALT_LEN];
 	uint8_t			zp_iv[ZIO_DATA_IV_LEN];
 	uint8_t			zp_mac[ZIO_DATA_MAC_LEN];
diff --git a/include/zfeature_common.h b/include/zfeature_common.h
@@ -88,6 +88,7 @@ typedef enum spa_feature {
 	SPA_FEATURE_LONGNAME,
 	SPA_FEATURE_LARGE_MICROZAP,
 	SPA_FEATURE_DYNAMIC_GANG_HEADER,
+	SPA_FEATURE_PHYSICAL_REWRITE,
 	SPA_FEATURES
 } spa_feature_t;
 
diff --git a/lib/libuutil/libuutil.abi b/lib/libuutil/libuutil.abi
@@ -1475,6 +1475,11 @@
       <parameter type-id='80f4b756' name='name'/>
       <return type-id='a27af98c'/>
     </function-decl>
+    <function-decl name='zfs_tunable_iter' mangled-name='zfs_tunable_iter' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_tunable_iter'>
+      <parameter type-id='d8d5f4ab' name='cb'/>
+      <parameter type-id='eaa32e2f' name='arg'/>
+      <return type-id='48b5725f'/>
+    </function-decl>
     <function-decl name='zfs_tunable_set' mangled-name='zfs_tunable_set' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_tunable_set'>
       <parameter type-id='a27af98c' name='zt'/>
       <parameter type-id='80f4b756' name='val'/>
@@ -1486,11 +1491,6 @@
       <parameter type-id='b59d7dce' name='valsz'/>
       <return type-id='95e97e5e'/>
     </function-decl>
-    <function-decl name='zfs_tunable_iter' mangled-name='zfs_tunable_iter' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_tunable_iter'>
-      <parameter type-id='d8d5f4ab' name='cb'/>
-      <parameter type-id='eaa32e2f' name='arg'/>
-      <return type-id='48b5725f'/>
-    </function-decl>
     <function-type size-in-bits='64' id='92f86508'>
       <parameter type-id='a27af98c'/>
       <parameter type-id='eaa32e2f'/>
diff --git a/lib/libzfs/libzfs.abi b/lib/libzfs/libzfs.abi
@@ -635,7 +635,7 @@
     <elf-symbol name='fletcher_4_superscalar_ops' size='128' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
     <elf-symbol name='libzfs_config_ops' size='16' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
     <elf-symbol name='sa_protocol_names' size='16' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
-    <elf-symbol name='spa_feature_table' size='2520' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
+    <elf-symbol name='spa_feature_table' size='2576' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
     <elf-symbol name='zfeature_checks_disable' size='4' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
     <elf-symbol name='zfs_deleg_perm_tab' size='528' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
     <elf-symbol name='zfs_history_event_names' size='328' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
@@ -1655,6 +1655,11 @@
       <parameter type-id='80f4b756' name='name'/>
       <return type-id='a27af98c'/>
     </function-decl>
+    <function-decl name='zfs_tunable_iter' mangled-name='zfs_tunable_iter' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_tunable_iter'>
+      <parameter type-id='d8d5f4ab' name='cb'/>
+      <parameter type-id='eaa32e2f' name='arg'/>
+      <return type-id='48b5725f'/>
+    </function-decl>
     <function-decl name='zfs_tunable_set' mangled-name='zfs_tunable_set' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_tunable_set'>
       <parameter type-id='a27af98c' name='zt'/>
       <parameter type-id='80f4b756' name='val'/>
@@ -1666,11 +1671,6 @@
       <parameter type-id='b59d7dce' name='valsz'/>
       <return type-id='95e97e5e'/>
     </function-decl>
-    <function-decl name='zfs_tunable_iter' mangled-name='zfs_tunable_iter' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_tunable_iter'>
-      <parameter type-id='d8d5f4ab' name='cb'/>
-      <parameter type-id='eaa32e2f' name='arg'/>
-      <return type-id='48b5725f'/>
-    </function-decl>
     <function-type size-in-bits='64' id='92f86508'>
       <parameter type-id='a27af98c'/>
       <parameter type-id='eaa32e2f'/>
@@ -6394,7 +6394,8 @@
       <enumerator name='SPA_FEATURE_LONGNAME' value='42'/>
       <enumerator name='SPA_FEATURE_LARGE_MICROZAP' value='43'/>
       <enumerator name='SPA_FEATURE_DYNAMIC_GANG_HEADER' value='44'/>
-      <enumerator name='SPA_FEATURES' value='45'/>
+      <enumerator name='SPA_FEATURE_PHYSICAL_REWRITE' value='45'/>
+      <enumerator name='SPA_FEATURES' value='46'/>
     </enum-decl>
     <typedef-decl name='spa_feature_t' type-id='33ecb627' id='d6618c78'/>
     <qualified-type-def type-id='80f4b756' const='yes' id='b99c00c9'/>
@@ -9585,8 +9586,8 @@
     </function-decl>
   </abi-instr>
   <abi-instr address-size='64' path='module/zcommon/zfeature_common.c' language='LANG_C99'>
-    <array-type-def dimensions='1' type-id='83f29ca2' size-in-bits='20160' id='b948da70'>
-      <subrange length='45' type-id='7359adad' id='cb8ddca0'/>
+    <array-type-def dimensions='1' type-id='83f29ca2' size-in-bits='20608' id='b9408bab'>
+      <subrange length='46' type-id='7359adad' id='8b86bc1b'/>
     </array-type-def>
     <enum-decl name='zfeature_flags' id='6db816a4'>
       <underlying-type type-id='9cac1fee'/>
@@ -9664,7 +9665,7 @@
     <pointer-type-def type-id='611586a1' size-in-bits='64' id='2e243169'/>
     <qualified-type-def type-id='eaa32e2f' const='yes' id='83be723c'/>
     <pointer-type-def type-id='83be723c' size-in-bits='64' id='7acd98a2'/>
-    <var-decl name='spa_feature_table' type-id='b948da70' mangled-name='spa_feature_table' visibility='default' elf-symbol-id='spa_feature_table'/>
+    <var-decl name='spa_feature_table' type-id='b9408bab' mangled-name='spa_feature_table' visibility='default' elf-symbol-id='spa_feature_table'/>
     <var-decl name='zfeature_checks_disable' type-id='c19b74c3' mangled-name='zfeature_checks_disable' visibility='default' elf-symbol-id='zfeature_checks_disable'/>
     <function-decl name='opendir' visibility='default' binding='global' size-in-bits='64'>
       <parameter type-id='80f4b756'/>
diff --git a/lib/libzfs_core/libzfs_core.abi b/lib/libzfs_core/libzfs_core.abi
@@ -1426,6 +1426,11 @@
       <parameter type-id='80f4b756' name='name'/>
       <return type-id='a27af98c'/>
     </function-decl>
+    <function-decl name='zfs_tunable_iter' mangled-name='zfs_tunable_iter' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_tunable_iter'>
+      <parameter type-id='d8d5f4ab' name='cb'/>
+      <parameter type-id='eaa32e2f' name='arg'/>
+      <return type-id='48b5725f'/>
+    </function-decl>
     <function-decl name='zfs_tunable_set' mangled-name='zfs_tunable_set' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_tunable_set'>
       <parameter type-id='a27af98c' name='zt'/>
       <parameter type-id='80f4b756' name='val'/>
@@ -1437,11 +1442,6 @@
       <parameter type-id='b59d7dce' name='valsz'/>
       <return type-id='95e97e5e'/>
     </function-decl>
-    <function-decl name='zfs_tunable_iter' mangled-name='zfs_tunable_iter' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_tunable_iter'>
-      <parameter type-id='d8d5f4ab' name='cb'/>
-      <parameter type-id='eaa32e2f' name='arg'/>
-      <return type-id='48b5725f'/>
-    </function-decl>
     <function-type size-in-bits='64' id='92f86508'>
       <parameter type-id='a27af98c'/>
       <parameter type-id='eaa32e2f'/>
diff --git a/man/man7/zpool-features.7 b/man/man7/zpool-features.7
@@ -842,6 +842,23 @@ when the
 command is used on a top-level vdev, and will never return to being
 .Sy enabled .
 .
+.feature com.truenas physical_rewrite yes extensible_dataset
+This feature enables physical block rewriting that preserves logical birth
+times, avoiding unnecessary inclusion of rewritten blocks in incremental
+.Nm zfs Cm send
+streams.
+When enabled, the
+.Nm zfs Cm rewrite Fl P
+command can be used.
+.Pp
+This feature becomes
+.Sy active
+the first time
+.Nm zfs Cm rewrite Fl P
+is used on any dataset, and will return to being
+.Sy enabled
+once all datasets that have ever used physical rewrite are destroyed.
+.
 .feature org.zfsonlinux project_quota yes extensible_dataset
 This feature allows administrators to account the spaces and objects usage
 information against the project identifier
diff --git a/man/man8/zfs-rewrite.8 b/man/man8/zfs-rewrite.8
@@ -31,7 +31,7 @@
 .Sh SYNOPSIS
 .Nm zfs
 .Cm rewrite
-.Oo Fl rvx Ns Oc
+.Oo Fl Prvx Ns Oc
 .Op Fl l Ar length
 .Op Fl o Ar offset
 .Ar file Ns | Ns Ar directory Ns …
@@ -43,6 +43,15 @@ as is without modification at a new location and possibly with new
 properties, such as checksum, compression, dedup, copies, etc,
 as if they were atomically read and written back.
 .Bl -tag -width "-r"
+.It Fl P
+Perform physical rewrite, preserving logical birth time of blocks.
+By default, rewrite updates logical birth times, making blocks appear
+as modified in snapshots and incremental send streams.
+Physical rewrite preserves logical birth times, avoiding unnecessary
+inclusion in incremental streams.
+Physical rewrite requires the
+.Sy physical_rewrite
+feature to be enabled on the pool.
 .It Fl l Ar length
 Rewrite at most this number of bytes.
 .It Fl o Ar offset
@@ -60,17 +69,22 @@ same as some property changes may increase pool space usage.
 Holes that were never written or were previously zero-compressed are
 not rewritten and will remain holes even if compression is disabled.
 .Pp
-Rewritten blocks will be seen as modified in next snapshot and as such
-included into the incremental
-.Nm zfs Cm send
-stream.
-.Pp
 If a
 .Fl l
 or
 .Fl o
 value request a rewrite to regions past the end of the file, then those
 regions are silently ignored, and no error is reported.
+.Pp
+By default, rewritten blocks update their logical birth time,
+meaning they will be included in incremental
+.Nm zfs Cm send
+streams as modified data.
+When the
+.Fl P
+flag is used, rewritten blocks preserve their logical birth time, since
+there are no user data changes.
 .
 .Sh SEE ALSO
-.Xr zfsprops 7
+.Xr zfsprops 7 ,
+.Xr zpool-features 7
diff --git a/module/zcommon/zfeature_common.c b/module/zcommon/zfeature_common.c
@@ -792,6 +792,18 @@ zpool_feature_init(void)
 	    ZFEATURE_FLAG_MOS | ZFEATURE_FLAG_NO_UPGRADE,
 	    ZFEATURE_TYPE_BOOLEAN, NULL, sfeatures);
 
+	{
+		static const spa_feature_t physical_rewrite_deps[] = {
+			SPA_FEATURE_EXTENSIBLE_DATASET,
+			SPA_FEATURE_NONE
+		};
+		zfeature_register(SPA_FEATURE_PHYSICAL_REWRITE,
+		    "com.truenas:physical_rewrite", "physical_rewrite",
+		    "Support for preserving logical birth time during rewrite.",
+		    ZFEATURE_FLAG_READONLY_COMPAT | ZFEATURE_FLAG_PER_DATASET,
+		    ZFEATURE_TYPE_BOOLEAN, physical_rewrite_deps, sfeatures);
+	}
+
 	zfs_mod_list_supported_free(sfeatures);
 }
 
diff --git a/module/zfs/dbuf.c b/module/zfs/dbuf.c
@@ -2152,6 +2152,12 @@ dbuf_redirty(dbuf_dirty_record_t *dr)
 			ASSERT(arc_released(db->db_buf));
 			arc_buf_thaw(db->db_buf);
 		}
+
+		/*
+		 * Clear the rewrite flag since this is now a logical
+		 * modification.
+		 */
+		dr->dt.dl.dr_rewrite = B_FALSE;
 	}
 }
 
@@ -2699,6 +2705,38 @@ dmu_buf_will_dirty(dmu_buf_t *db_fake, dmu_tx_t *tx)
 	dmu_buf_will_dirty_flags(db_fake, tx, DMU_READ_NO_PREFETCH);
 }
 
+void
+dmu_buf_will_rewrite(dmu_buf_t *db_fake, dmu_tx_t *tx)
+{
+	dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
+
+	ASSERT(tx->tx_txg != 0);
+	ASSERT(!zfs_refcount_is_zero(&db->db_holds));
+
+	/*
+	 * If the dbuf is already dirty in this txg, it will be written
+	 * anyway, so there's nothing to do.
+	 */
+	mutex_enter(&db->db_mtx);
+	if (dbuf_find_dirty_eq(db, tx->tx_txg) != NULL) {
+		mutex_exit(&db->db_mtx);
+		return;
+	}
+	mutex_exit(&db->db_mtx);
+
+	/*
+	 * The dbuf is not dirty, so we need to make it dirty and
+	 * mark it for rewrite (preserve logical birth time).
+	 */
+	dmu_buf_will_dirty_flags(db_fake, tx, DMU_READ_NO_PREFETCH);
+
+	mutex_enter(&db->db_mtx);
+	dbuf_dirty_record_t *dr = dbuf_find_dirty_eq(db, tx->tx_txg);
+	if (dr != NULL && db->db_level == 0)
+		dr->dt.dl.dr_rewrite = B_TRUE;
+	mutex_exit(&db->db_mtx);
+}
+
 boolean_t
 dmu_buf_is_dirty(dmu_buf_t *db_fake, dmu_tx_t *tx)
 {
@@ -5330,6 +5368,24 @@ dbuf_write(dbuf_dirty_record_t *dr, arc_buf_t *data, dmu_tx_t *tx)
 
 	dmu_write_policy(os, dn, db->db_level, wp_flag, &zp);
 
+	/*
+	 * Set rewrite properties for zfs_rewrite() operations.
+	 */
+	if (db->db_level == 0 && dr->dt.dl.dr_rewrite) {
+		zp.zp_rewrite = B_TRUE;
+
+		/*
+		 * Mark physical rewrite feature for activation.
+		 * This will be activated automatically during dataset sync.
+		 */
+		dsl_dataset_t *ds = os->os_dsl_dataset;
+		if (!dsl_dataset_feature_is_active(ds,
+		    SPA_FEATURE_PHYSICAL_REWRITE)) {
+			ds->ds_feature_activation[
+			    SPA_FEATURE_PHYSICAL_REWRITE] = (void *)B_TRUE;
+		}
+	}
+
 	/*
 	 * We copy the blkptr now (rather than when we instantiate the dirty
 	 * record), because its value can change between open context and
@@ -5400,6 +5456,7 @@ EXPORT_SYMBOL(dbuf_release_bp);
 EXPORT_SYMBOL(dbuf_dirty);
 EXPORT_SYMBOL(dmu_buf_set_crypt_params);
 EXPORT_SYMBOL(dmu_buf_will_dirty);
+EXPORT_SYMBOL(dmu_buf_will_rewrite);
 EXPORT_SYMBOL(dmu_buf_is_dirty);
 EXPORT_SYMBOL(dmu_buf_will_clone_or_dio);
 EXPORT_SYMBOL(dmu_buf_will_not_fill);
diff --git a/module/zfs/dmu.c b/module/zfs/dmu.c
@@ -2508,6 +2508,7 @@ dmu_write_policy(objset_t *os, dnode_t *dn, int level, int wp, zio_prop_t *zp)
 	zp->zp_encrypt = encrypt;
 	zp->zp_byteorder = ZFS_HOST_BYTEORDER;
 	zp->zp_direct_write = (wp & WP_DIRECT_WR) ? B_TRUE : B_FALSE;
+	zp->zp_rewrite = B_FALSE;
 	memset(zp->zp_salt, 0, ZIO_DATA_SALT_LEN);
 	memset(zp->zp_iv, 0, ZIO_DATA_IV_LEN);
 	memset(zp->zp_mac, 0, ZIO_DATA_MAC_LEN);
diff --git a/module/zfs/zfs_vnops.c b/module/zfs/zfs_vnops.c
diff --git a/module/zfs/zio.c b/module/zfs/zio.c
diff --git a/tests/runfiles/common.run b/tests/runfiles/common.run
diff --git a/tests/runfiles/sanity.run b/tests/runfiles/sanity.run
diff --git a/tests/zfs-tests/tests/Makefile.am b/tests/zfs-tests/tests/Makefile.am
diff --git a/tests/zfs-tests/tests/functional/cli_root/zfs_rewrite/zfs_rewrite_physical.ksh b/tests/zfs-tests/tests/functional/cli_root/zfs_rewrite/zfs_rewrite_physical.ksh
diff --git a/tests/zfs-tests/tests/functional/cli_root/zpool_get/zpool_get.cfg b/tests/zfs-tests/tests/functional/cli_root/zpool_get/zpool_get.cfg
