[2.2] BRT and other fixes into 2.2.3-staging

cmd/zdb/zdb_il.c

@@ -168,7 +168,7 @@ zil_prt_rec_write(zilog_t *zilog, int txtype, const void *arg)
 	    (u_longlong_t)lr->lr_foid, (u_longlong_t)lr->lr_offset,
 	    (u_longlong_t)lr->lr_length);
 
-	if (txtype == TX_WRITE2 || verbose < 5)
+	if (txtype == TX_WRITE2 || verbose < 4)
 		return;
 
 	if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) {
@@ -178,6 +178,8 @@ zil_prt_rec_write(zilog_t *zilog, int txtype, const void *arg)
 		    "will claim" : "won't claim");
 		print_log_bp(bp, tab_prefix);
 
+		if (verbose < 5)
+			return;
 		if (BP_IS_HOLE(bp)) {
 			(void) printf("\t\t\tLSIZE 0x%llx\n",
 			    (u_longlong_t)BP_GET_LSIZE(bp));
@@ -202,6 +204,9 @@ zil_prt_rec_write(zilog_t *zilog, int txtype, const void *arg)
 		if (error)
 			goto out;
 	} else {
+		if (verbose < 5)
+			return;
+
 		/* data is stored after the end of the lr_write record */
 		data = abd_alloc(lr->lr_length, B_FALSE);
 		abd_copy_from_buf(data, lr + 1, lr->lr_length);
@@ -217,6 +222,28 @@ zil_prt_rec_write(zilog_t *zilog, int txtype, const void *arg)
 	abd_free(data);
 }
 
+static void
+zil_prt_rec_write_enc(zilog_t *zilog, int txtype, const void *arg)
+{
+	(void) txtype;
+	const lr_write_t *lr = arg;
+	const blkptr_t *bp = &lr->lr_blkptr;
+	int verbose = MAX(dump_opt['d'], dump_opt['i']);
+
+	(void) printf("%s(encrypted)\n", tab_prefix);
+
+	if (verbose < 4)
+		return;
+
+	if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) {
+		(void) printf("%shas blkptr, %s\n", tab_prefix,
+		    !BP_IS_HOLE(bp) &&
+		    bp->blk_birth >= spa_min_claim_txg(zilog->zl_spa) ?
+		    "will claim" : "won't claim");
+		print_log_bp(bp, tab_prefix);
+	}
+}
+
 static void
 zil_prt_rec_truncate(zilog_t *zilog, int txtype, const void *arg)
 {
@@ -312,11 +339,34 @@ zil_prt_rec_clone_range(zilog_t *zilog, int txtype, const void *arg)
 {
 	(void) zilog, (void) txtype;
 	const lr_clone_range_t *lr = arg;
+	int verbose = MAX(dump_opt['d'], dump_opt['i']);
 
 	(void) printf("%sfoid %llu, offset %llx, length %llx, blksize %llx\n",
 	    tab_prefix, (u_longlong_t)lr->lr_foid, (u_longlong_t)lr->lr_offset,
 	    (u_longlong_t)lr->lr_length, (u_longlong_t)lr->lr_blksz);
 
+	if (verbose < 4)
+		return;
+
+	for (unsigned int i = 0; i < lr->lr_nbps; i++) {
+		(void) printf("%s[%u/%llu] ", tab_prefix, i + 1,
+		    (u_longlong_t)lr->lr_nbps);
+		print_log_bp(&lr->lr_bps[i], "");
+	}
+}
+
+static void
+zil_prt_rec_clone_range_enc(zilog_t *zilog, int txtype, const void *arg)
+{
+	(void) zilog, (void) txtype;
+	const lr_clone_range_t *lr = arg;
+	int verbose = MAX(dump_opt['d'], dump_opt['i']);
+
+	(void) printf("%s(encrypted)\n", tab_prefix);
+
+	if (verbose < 4)
+		return;
+
 	for (unsigned int i = 0; i < lr->lr_nbps; i++) {
 		(void) printf("%s[%u/%llu] ", tab_prefix, i + 1,
 		    (u_longlong_t)lr->lr_nbps);
@@ -327,6 +377,7 @@ zil_prt_rec_clone_range(zilog_t *zilog, int txtype, const void *arg)
 typedef void (*zil_prt_rec_func_t)(zilog_t *, int, const void *);
 typedef struct zil_rec_info {
 	zil_prt_rec_func_t	zri_print;
+	zil_prt_rec_func_t	zri_print_enc;
 	const char		*zri_name;
 	uint64_t		zri_count;
 } zil_rec_info_t;
@@ -341,7 +392,9 @@ static zil_rec_info_t zil_rec_info[TX_MAX_TYPE] = {
 	{.zri_print = zil_prt_rec_remove,   .zri_name = "TX_RMDIR           "},
 	{.zri_print = zil_prt_rec_link,	    .zri_name = "TX_LINK            "},
 	{.zri_print = zil_prt_rec_rename,   .zri_name = "TX_RENAME          "},
-	{.zri_print = zil_prt_rec_write,    .zri_name = "TX_WRITE           "},
+	{.zri_print = zil_prt_rec_write,
+	    .zri_print_enc = zil_prt_rec_write_enc,
+	    .zri_name = "TX_WRITE           "},
 	{.zri_print = zil_prt_rec_truncate, .zri_name = "TX_TRUNCATE        "},
 	{.zri_print = zil_prt_rec_setattr,  .zri_name = "TX_SETATTR         "},
 	{.zri_print = zil_prt_rec_acl,	    .zri_name = "TX_ACL_V0          "},
@@ -358,6 +411,7 @@ static zil_rec_info_t zil_rec_info[TX_MAX_TYPE] = {
 	{.zri_print = zil_prt_rec_rename,   .zri_name = "TX_RENAME_EXCHANGE "},
 	{.zri_print = zil_prt_rec_rename,   .zri_name = "TX_RENAME_WHITEOUT "},
 	{.zri_print = zil_prt_rec_clone_range,
+	    .zri_print_enc = zil_prt_rec_clone_range_enc,
 	    .zri_name = "TX_CLONE_RANGE     "},
 };
 
@@ -384,6 +438,8 @@ print_log_record(zilog_t *zilog, const lr_t *lr, void *arg, uint64_t claim_txg)
 	if (txtype && verbose >= 3) {
 		if (!zilog->zl_os->os_encrypted) {
 			zil_rec_info[txtype].zri_print(zilog, txtype, lr);
+		} else if (zil_rec_info[txtype].zri_print_enc) {
+			zil_rec_info[txtype].zri_print_enc(zilog, txtype, lr);
 		} else {
 			(void) printf("%s(encrypted)\n", tab_prefix);
 		}

include/os/linux/spl/sys/kmem_cache.h

@@ -70,8 +70,6 @@ typedef enum kmem_cbrc {
 #define	KMC_REAP_CHUNK		INT_MAX
 #define	KMC_DEFAULT_SEEKS	1
 
-#define	KMC_RECLAIM_ONCE	0x1	/* Force a single shrinker pass */
-
 extern struct list_head spl_kmem_cache_list;
 extern struct rw_semaphore spl_kmem_cache_sem;
 

include/sys/dataset_kstats.h

@@ -71,6 +71,7 @@ typedef struct dataset_kstats {
 
 int dataset_kstats_create(dataset_kstats_t *, objset_t *);
 void dataset_kstats_destroy(dataset_kstats_t *);
+void dataset_kstats_rename(dataset_kstats_t *dk, const char *);
 
 void dataset_kstats_update_write_kstats(dataset_kstats_t *, int64_t);
 void dataset_kstats_update_read_kstats(dataset_kstats_t *, int64_t);

include/sys/dsl_crypt.h

@@ -206,6 +206,7 @@ void dsl_dataset_promote_crypt_sync(dsl_dir_t *target, dsl_dir_t *origin,
     dmu_tx_t *tx);
 int dmu_objset_create_crypt_check(dsl_dir_t *parentdd,
     dsl_crypto_params_t *dcp, boolean_t *will_encrypt);
+boolean_t dmu_objset_crypto_key_equal(objset_t *osa, objset_t *osb);
 void dsl_dataset_create_crypt_sync(uint64_t dsobj, dsl_dir_t *dd,
     struct dsl_dataset *origin, dsl_crypto_params_t *dcp, dmu_tx_t *tx);
 uint64_t dsl_crypto_key_create_sync(uint64_t crypt, dsl_wrapping_key_t *wkey,

lib/libspl/include/assert.h

@@ -64,6 +64,9 @@ libspl_assert(const char *buf, const char *file, const char *func, int line)
 #undef verify
 #endif
 
+#define	PANIC(fmt, a...)						\
+	libspl_assertf(__FILE__, __FUNCTION__, __LINE__, fmt, ## a)
+
 #define	VERIFY(cond)							\
 	(void) ((!(cond)) &&						\
 	    libspl_assert(#cond, __FILE__, __FUNCTION__, __LINE__))

man/man4/spl.4

@@ -31,14 +31,6 @@ for use by the kmem caches.
 For the majority of systems and workloads only a small number of threads are
 required.
 .
-.It Sy spl_kmem_cache_reclaim Ns = Ns Sy 0 Pq uint
-When this is set it prevents Linux from being able to rapidly reclaim all the
-memory held by the kmem caches.
-This may be useful in circumstances where it's preferable that Linux
-reclaim memory from some other subsystem first.
-Setting this will increase the likelihood out of memory events on a memory
-constrained system.
-.
 .It Sy spl_kmem_cache_obj_per_slab Ns = Ns Sy 8 Pq uint
 The preferred number of objects per slab in the cache.
 In general, a larger value will increase the caches memory footprint

man/man7/zpool-features.7

@@ -364,9 +364,12 @@ When this feature is enabled ZFS will use block cloning for operations like
 Block cloning allows to create multiple references to a single block.
 It is much faster than copying the data (as the actual data is neither read nor
 written) and takes no additional space.
-Blocks can be cloned across datasets under some conditions (like disabled
-encryption and equal
-.Nm recordsize ) .
+Blocks can be cloned across datasets under some conditions (like equal
+.Nm recordsize ,
+the same master encryption key, etc.).
+ZFS tries its best to clone across datasets including encrypted ones.
+This is limited for various (nontrivial) reasons depending on the OS
+and/or ZFS internals.
 .Pp
 This feature becomes
 .Sy active

module/os/freebsd/spl/spl_kstat.c

@@ -187,45 +187,40 @@ kstat_sysctl_dataset_string(SYSCTL_HANDLER_ARGS)
 static int
 kstat_sysctl_io(SYSCTL_HANDLER_ARGS)
 {
-	struct sbuf *sb;
+	struct sbuf sb;
 	kstat_t *ksp = arg1;
 	kstat_io_t *kip = ksp->ks_data;
 	int rc;
 
-	sb = sbuf_new_auto();
-	if (sb == NULL)
-		return (ENOMEM);
+	sbuf_new_for_sysctl(&sb, NULL, 0, req);
+
 	/* Update the aggsums before reading */
 	(void) ksp->ks_update(ksp, KSTAT_READ);
 
 	/* though wlentime & friends are signed, they will never be negative */
-	sbuf_printf(sb,
+	sbuf_printf(&sb,
 	    "%-8llu %-8llu %-8u %-8u %-8llu %-8llu "
 	    "%-8llu %-8llu %-8llu %-8llu %-8u %-8u\n",
 	    kip->nread, kip->nwritten,
 	    kip->reads, kip->writes,
 	    kip->wtime, kip->wlentime, kip->wlastupdate,
 	    kip->rtime, kip->rlentime, kip->rlastupdate,
 	    kip->wcnt,  kip->rcnt);
-	rc = sbuf_finish(sb);
-	if (rc == 0)
-		rc = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb));
-	sbuf_delete(sb);
+	rc = sbuf_finish(&sb);
+	sbuf_delete(&sb);
 	return (rc);
 }
 
 static int
 kstat_sysctl_raw(SYSCTL_HANDLER_ARGS)
 {
-	struct sbuf *sb;
+	struct sbuf sb;
 	void *data;
 	kstat_t *ksp = arg1;
 	void *(*addr_op)(kstat_t *ksp, loff_t index);
 	int n, has_header, rc = 0;
 
-	sb = sbuf_new_auto();
-	if (sb == NULL)
-		return (ENOMEM);
+	sbuf_new_for_sysctl(&sb, NULL, PAGE_SIZE, req);
 
 	if (ksp->ks_raw_ops.addr)
 		addr_op = ksp->ks_raw_ops.addr;
@@ -258,8 +253,10 @@ kstat_sysctl_raw(SYSCTL_HANDLER_ARGS)
 	if (has_header) {
 		if (rc == ENOMEM && !kstat_resize_raw(ksp))
 			goto restart_headers;
-		if (rc == 0)
-			sbuf_printf(sb, "\n%s", ksp->ks_raw_buf);
+		if (rc == 0) {
+			sbuf_cat(&sb, "\n");
+			sbuf_cat(&sb, ksp->ks_raw_buf);
+		}
 	}
 
 	while ((data = addr_op(ksp, n)) != NULL) {
@@ -270,22 +267,19 @@ kstat_sysctl_raw(SYSCTL_HANDLER_ARGS)
 			if (rc == ENOMEM && !kstat_resize_raw(ksp))
 				goto restart;
 			if (rc == 0)
-				sbuf_printf(sb, "%s", ksp->ks_raw_buf);
+				sbuf_cat(&sb, ksp->ks_raw_buf);
 
 		} else {
 			ASSERT3U(ksp->ks_ndata, ==, 1);
-			sbuf_hexdump(sb, ksp->ks_data,
+			sbuf_hexdump(&sb, ksp->ks_data,
 			    ksp->ks_data_size, NULL, 0);
 		}
 		n++;
 	}
 	free(ksp->ks_raw_buf, M_TEMP);
 	mutex_exit(ksp->ks_lock);
-	sbuf_trim(sb);
-	rc = sbuf_finish(sb);
-	if (rc == 0)
-		rc = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb));
-	sbuf_delete(sb);
+	rc = sbuf_finish(&sb);
+	sbuf_delete(&sb);
 	return (rc);
 }
 

module/os/freebsd/zfs/zio_crypt.c

@@ -1251,7 +1251,7 @@ zio_crypt_init_uios_zil(boolean_t encrypt, uint8_t *plainbuf,
 	iovec_t *dst_iovecs;
 	zil_chain_t *zilc;
 	lr_t *lr;
-	uint64_t txtype, lr_len;
+	uint64_t txtype, lr_len, nused;
 	uint_t crypt_len, nr_iovecs, vec;
 	uint_t aad_len = 0, total_len = 0;
 
@@ -1268,7 +1268,10 @@ zio_crypt_init_uios_zil(boolean_t encrypt, uint8_t *plainbuf,
 	zilc = (zil_chain_t *)src;
 	slrp = src + sizeof (zil_chain_t);
 	aadp = aadbuf;
-	blkend = src + ((byteswap) ? BSWAP_64(zilc->zc_nused) : zilc->zc_nused);
+	nused = ((byteswap) ? BSWAP_64(zilc->zc_nused) : zilc->zc_nused);
+	ASSERT3U(nused, >=, sizeof (zil_chain_t));
+	ASSERT3U(nused, <=, datalen);
+	blkend = src + nused;
 
 	/*
 	 * Calculate the number of encrypted iovecs we will need.
@@ -1287,6 +1290,8 @@ zio_crypt_init_uios_zil(boolean_t encrypt, uint8_t *plainbuf,
 			txtype = lr->lrc_txtype;
 			lr_len = lr->lrc_reclen;
 		}
+		ASSERT3U(lr_len, >=, sizeof (lr_t));
+		ASSERT3U(lr_len, <=, blkend - slrp);
 
 		nr_iovecs++;
 		if (txtype == TX_WRITE && lr_len != sizeof (lr_write_t))

module/os/freebsd/zfs/zvol_os.c

@@ -1333,6 +1333,7 @@ zvol_os_rename_minor(zvol_state_t *zv, const char *newname)
 		}
 	}
 	strlcpy(zv->zv_name, newname, sizeof (zv->zv_name));
+	dataset_kstats_rename(&zv->zv_kstat, newname);
 }
 
 /*

module/os/linux/spl/spl-kmem-cache.c

@@ -76,17 +76,6 @@ module_param(spl_kmem_cache_magazine_size, uint, 0444);
 MODULE_PARM_DESC(spl_kmem_cache_magazine_size,
 	"Default magazine size (2-256), set automatically (0)");
 
-/*
- * The default behavior is to report the number of objects remaining in the
- * cache.  This allows the Linux VM to repeatedly reclaim objects from the
- * cache when memory is low satisfy other memory allocations.  Alternately,
- * setting this value to KMC_RECLAIM_ONCE limits how aggressively the cache
- * is reclaimed.  This may increase the likelihood of out of memory events.
- */
-static unsigned int spl_kmem_cache_reclaim = 0 /* KMC_RECLAIM_ONCE */;
-module_param(spl_kmem_cache_reclaim, uint, 0644);
-MODULE_PARM_DESC(spl_kmem_cache_reclaim, "Single reclaim pass (0x1)");
-
 static unsigned int spl_kmem_cache_obj_per_slab = SPL_KMEM_CACHE_OBJ_PER_SLAB;
 module_param(spl_kmem_cache_obj_per_slab, uint, 0644);
 MODULE_PARM_DESC(spl_kmem_cache_obj_per_slab, "Number of objects per slab");

module/os/linux/zfs/zio_crypt.c

@@ -1405,7 +1405,7 @@ zio_crypt_init_uios_zil(boolean_t encrypt, uint8_t *plainbuf,
     boolean_t *no_crypt)
 {
 	int ret;
-	uint64_t txtype, lr_len;
+	uint64_t txtype, lr_len, nused;
 	uint_t nr_src, nr_dst, crypt_len;
 	uint_t aad_len = 0, nr_iovecs = 0, total_len = 0;
 	iovec_t *src_iovecs = NULL, *dst_iovecs = NULL;
@@ -1432,7 +1432,10 @@ zio_crypt_init_uios_zil(boolean_t encrypt, uint8_t *plainbuf,
 	zilc = (zil_chain_t *)src;
 	slrp = src + sizeof (zil_chain_t);
 	aadp = aadbuf;
-	blkend = src + ((byteswap) ? BSWAP_64(zilc->zc_nused) : zilc->zc_nused);
+	nused = ((byteswap) ? BSWAP_64(zilc->zc_nused) : zilc->zc_nused);
+	ASSERT3U(nused, >=, sizeof (zil_chain_t));
+	ASSERT3U(nused, <=, datalen);
+	blkend = src + nused;
 
 	/* calculate the number of encrypted iovecs we will need */
 	for (; slrp < blkend; slrp += lr_len) {
@@ -1445,6 +1448,8 @@ zio_crypt_init_uios_zil(boolean_t encrypt, uint8_t *plainbuf,
 			txtype = BSWAP_64(lr->lrc_txtype);
 			lr_len = BSWAP_64(lr->lrc_reclen);
 		}
+		ASSERT3U(lr_len, >=, sizeof (lr_t));
+		ASSERT3U(lr_len, <=, blkend - slrp);
 
 		nr_iovecs++;
 		if (txtype == TX_WRITE && lr_len != sizeof (lr_write_t))

module/os/linux/zfs/zvol_os.c

@@ -1521,6 +1521,8 @@ zvol_os_rename_minor(zvol_state_t *zv, const char *newname)
 	 */
 	set_disk_ro(zv->zv_zso->zvo_disk, !readonly);
 	set_disk_ro(zv->zv_zso->zvo_disk, readonly);
+
+	dataset_kstats_rename(&zv->zv_kstat, newname);
 }
 
 void