Add dm-integrity support

Create a linear LV with integrity added to it:

lvcreate --type integrity -n Name -L Size VG

or

lvcreate --integrity y -n Name -L Size VG

When creating, --zero y will zero the entire LV.
This is suggested to initialize integrity checksums
and avoid causing read errors in unwritten portions
of the new LV.

Options:

  --integrity internal
  integrity checksum metadata is interleaved with data.

  --integrity external
  integrity checksum metadata on separate LV, allows
  adding integrity to existing LV, or removing integrity.

  --integritymetadata LV
  use the specified LV for external metadata.

  --integritysettings String
  set dm-integrity parameters.

TODO:

- for internal, zero start of origin LV

- increase allocated LV size by an extra extent, and then
  round provided_data_sectors down to be a multiple of extent size

- add command to remove intregrity from an LV (only for external),
  i.e. lvconvert --integrity none LV

- let integrity be used by raid images

device_mapper/all.h

@@ -392,6 +392,15 @@ struct dm_status_writecache {
 int dm_get_status_writecache(struct dm_pool *mem, const char *params,
                              struct dm_status_writecache **status);
 
+struct dm_status_integrity {
+	uint64_t number_of_mismatches;
+	uint64_t provided_data_sectors;
+	uint64_t recalc_sector;
+};
+
+int dm_get_status_integrity(struct dm_pool *mem, const char *params,
+                            struct dm_status_integrity **status);
+
 /*
  * Parse params from STATUS call for snapshot target
  *
@@ -970,6 +979,35 @@ int dm_tree_node_add_writecache_target(struct dm_tree_node *node,
 				uint32_t writecache_block_size,
 				struct writecache_settings *settings);
 
+struct integrity_settings {
+	char mode[8];
+	uint32_t tag_size;
+	const char *internal_hash;
+
+	uint32_t journal_sectors;
+	uint32_t interleave_sectors;
+	uint32_t buffer_sectors;
+	uint32_t journal_watermark;
+	uint32_t commit_time;
+	uint32_t block_size;
+	uint32_t bitmap_flush_interval;
+	uint64_t sectors_per_bit;
+
+	unsigned journal_sectors_set:1;
+	unsigned interleave_sectors_set:1;
+	unsigned buffer_sectors_set:1;
+	unsigned journal_watermark_set:1;
+	unsigned commit_time_set:1;
+	unsigned block_size_set:1;
+	unsigned bitmap_flush_interval_set:1;
+	unsigned sectors_per_bit_set:1;
+};
+
+int dm_tree_node_add_integrity_target(struct dm_tree_node *node,
+				uint64_t size,
+				const char *origin_uuid,
+				const char *meta_uuid,
+				struct integrity_settings *settings);
 
 /*
  * VDO target

device_mapper/libdm-deptree.c

@@ -38,6 +38,7 @@ enum {
 	SEG_STRIPED,
 	SEG_ZERO,
 	SEG_WRITECACHE,
+	SEG_INTEGRITY,
 	SEG_THIN_POOL,
 	SEG_THIN,
 	SEG_VDO,
@@ -78,6 +79,7 @@ static const struct {
 	{ SEG_STRIPED, "striped" },
 	{ SEG_ZERO, "zero"},
 	{ SEG_WRITECACHE, "writecache"},
+	{ SEG_INTEGRITY, "integrity"},
 	{ SEG_THIN_POOL, "thin-pool"},
 	{ SEG_THIN, "thin"},
 	{ SEG_VDO, "vdo" },
@@ -221,6 +223,10 @@ struct load_segment {
 	int writecache_pmem;				/* writecache, 1 if pmem, 0 if ssd */
 	uint32_t writecache_block_size;			/* writecache, in bytes */
 	struct writecache_settings writecache_settings;	/* writecache */
+
+	uint64_t integrity_data_sectors;		/* integrity (provided_data_sectors) */
+	struct dm_tree_node *integrity_meta_node;	/* integrity */
+	struct integrity_settings integrity_settings;	/* integrity */
 };
 
 /* Per-device properties */
@@ -2705,6 +2711,82 @@ static int _writecache_emit_segment_line(struct dm_task *dmt,
 	return 1;
 }
 
+static int _integrity_emit_segment_line(struct dm_task *dmt,
+				    struct load_segment *seg,
+				    char *params, size_t paramsize)
+{
+	struct integrity_settings *set = &seg->integrity_settings;
+	int pos = 0;
+	int count;
+	char origin_dev[DM_FORMAT_DEV_BUFSIZE];
+	char meta_dev[DM_FORMAT_DEV_BUFSIZE];
+
+	if (!_build_dev_string(origin_dev, sizeof(origin_dev), seg->origin))
+		return_0;
+
+	if (seg->integrity_meta_node &&
+	    !_build_dev_string(meta_dev, sizeof(meta_dev), seg->integrity_meta_node))
+		return_0;
+
+	count = 1; /* for internal_hash which we always pass in */
+
+	if (seg->integrity_meta_node)
+		count++;
+
+	if (set->journal_sectors_set)
+		count++;
+	if (set->interleave_sectors_set)
+		count++;
+	if (set->buffer_sectors_set)
+		count++;
+	if (set->journal_watermark_set)
+		count++;
+	if (set->commit_time_set)
+		count++;
+	if (set->block_size_set)
+		count++;
+	if (set->bitmap_flush_interval_set)
+		count++;
+	if (set->sectors_per_bit_set)
+		count++;
+
+	EMIT_PARAMS(pos, "%s 0 %u %s %d internal_hash:%s",
+		    origin_dev,
+		    set->tag_size,
+		    set->mode,
+		    count,
+		    set->internal_hash);
+
+	if (seg->integrity_meta_node)
+		EMIT_PARAMS(pos, " meta_device:%s", meta_dev);
+
+	if (set->journal_sectors_set)
+		EMIT_PARAMS(pos, " journal_sectors:%u", set->journal_sectors);
+
+	if (set->interleave_sectors_set)
+		EMIT_PARAMS(pos, " ineterleave_sectors:%u", set->interleave_sectors);
+
+	if (set->buffer_sectors_set)
+		EMIT_PARAMS(pos, " buffer_sectors:%u", set->buffer_sectors);
+
+	if (set->journal_watermark_set)
+		EMIT_PARAMS(pos, " journal_watermark:%u", set->journal_watermark);
+
+	if (set->commit_time_set)
+		EMIT_PARAMS(pos, " commit_time:%u", set->commit_time);
+
+	if (set->block_size_set)
+		EMIT_PARAMS(pos, " block_size:%u", set->block_size);
+
+	if (set->bitmap_flush_interval_set)
+		EMIT_PARAMS(pos, " bitmap_flush_interval:%u", set->bitmap_flush_interval);
+
+	if (set->sectors_per_bit_set)
+		EMIT_PARAMS(pos, " sectors_per_bit:%llu", (unsigned long long)set->sectors_per_bit);
+
+	return 1;
+}
+
 static int _thin_pool_emit_segment_line(struct dm_task *dmt,
 					struct load_segment *seg,
 					char *params, size_t paramsize)
@@ -2889,6 +2971,10 @@ static int _emit_segment_line(struct dm_task *dmt, uint32_t major,
 		if (!_writecache_emit_segment_line(dmt, seg, params, paramsize))
 			return_0;
 		break;
+	case SEG_INTEGRITY:
+		if (!_integrity_emit_segment_line(dmt, seg, params, paramsize))
+			return_0;
+		break;
 	}
 
 	switch(seg->type) {
@@ -2901,6 +2987,7 @@ static int _emit_segment_line(struct dm_task *dmt, uint32_t major,
 	case SEG_THIN:
 	case SEG_CACHE:
 	case SEG_WRITECACHE:
+	case SEG_INTEGRITY:
 		break;
 	case SEG_CRYPT:
 	case SEG_LINEAR:
@@ -3738,6 +3825,39 @@ int dm_tree_node_add_writecache_target(struct dm_tree_node *node,
 	return 1;
 }
 
+int dm_tree_node_add_integrity_target(struct dm_tree_node *node,
+				  uint64_t size,
+				  const char *origin_uuid,
+				  const char *meta_uuid,
+				  struct integrity_settings *settings)
+{
+	struct load_segment *seg;
+
+	if (!(seg = _add_segment(node, SEG_INTEGRITY, size)))
+		return_0;
+
+	if (meta_uuid) {
+		if (!(seg->integrity_meta_node = dm_tree_find_node_by_uuid(node->dtree, meta_uuid))) {
+			log_error("Missing integrity's meta uuid %s.", meta_uuid);
+			return 0;
+		}
+
+		if (!_link_tree_nodes(node, seg->integrity_meta_node))
+			return_0;
+	}
+
+	if (!(seg->origin = dm_tree_find_node_by_uuid(node->dtree, origin_uuid))) {
+		log_error("Missing integrity's origin uuid %s.", origin_uuid);
+		return 0;
+	}
+	if (!_link_tree_nodes(node, seg->origin))
+		return_0;
+
+	memcpy(&seg->integrity_settings, settings, sizeof(struct integrity_settings));
+
+	return 1;
+}
+
 int dm_tree_node_add_replicator_target(struct dm_tree_node *node,
 				       uint64_t size,
 				       const char *rlog_uuid,

device_mapper/libdm-targets.c

@@ -380,6 +380,35 @@ int dm_get_status_writecache(struct dm_pool *mem, const char *params,
 	return 1;
 }
 
+int dm_get_status_integrity(struct dm_pool *mem, const char *params,
+			     struct dm_status_integrity **status)
+{
+	struct dm_status_integrity *s;
+	char recalc_str[8];
+
+	if (!(s = dm_pool_zalloc(mem, sizeof(struct dm_status_integrity))))
+		return_0;
+
+	memset(recalc_str, 0, sizeof(recalc_str));
+
+	if (sscanf(params, "%llu %llu %s",
+		   (unsigned long long *)&s->number_of_mismatches,
+		   (unsigned long long *)&s->provided_data_sectors,
+		   recalc_str) != 3) {
+		log_error("Failed to parse integrity params: %s.", params);
+		dm_pool_free(mem, s);
+		return 0;
+	}
+
+	if (recalc_str[0] == '-')
+		s->recalc_sector = 0;
+	else
+		s->recalc_sector = strtoull(recalc_str, NULL, 0);
+
+	*status = s;
+	return 1;
+}
+
 int parse_thin_pool_status(const char *params, struct dm_status_thin_pool *s)
 {
 	int pos;

lib/Makefile.in

@@ -20,6 +20,7 @@ SOURCES =\
 	activate/activate.c \
 	cache/lvmcache.c \
 	writecache/writecache.c \
+	integrity/integrity.c \
 	cache_segtype/cache.c \
 	commands/toolcontext.c \
 	config/config.c \
@@ -67,6 +68,7 @@ SOURCES =\
 	log/log.c \
 	metadata/cache_manip.c \
 	metadata/writecache_manip.c \
+	metadata/integrity_manip.c \
 	metadata/lv.c \
 	metadata/lv_manip.c \
 	metadata/merge.c \

lib/activate/activate.c

@@ -2870,6 +2870,7 @@ int deactivate_lv_with_sub_lv(const struct logical_volume *lv)
 int activate_lv(struct cmd_context *cmd, const struct logical_volume *lv)
 {
 	const struct logical_volume *active_lv;
+	const struct logical_volume *lv_use;
 	int ret;
 
 	/*
@@ -2888,19 +2889,30 @@ int activate_lv(struct cmd_context *cmd, const struct logical_volume *lv)
 		goto out;
 	}
 
+	if (lv->status & LV_UNCOMMITTED)
+		lv_use = lv;
+	else
+		lv_use = lv_committed(lv);
+
 	ret = lv_activate_with_filter(cmd, NULL, 0,
 				      (lv->status & LV_NOSCAN) ? 1 : 0,
 				      (lv->status & LV_TEMPORARY) ? 1 : 0,
-				      lv_committed(lv));
+				      lv_use);
 out:
 	return ret;
 }
 
 int deactivate_lv(struct cmd_context *cmd, const struct logical_volume *lv)
 {
+	const struct logical_volume *lv_use;
 	int ret;
 
-	ret = lv_deactivate(cmd, NULL, lv_committed(lv));
+	if (lv->status & LV_UNCOMMITTED)
+		lv_use = lv;
+	else
+		lv_use = lv_committed(lv);
+
+	ret = lv_deactivate(cmd, NULL, lv_use);
 
 	return ret;
 }

lib/activate/activate.h

@@ -39,6 +39,7 @@ typedef enum {
 	SEG_STATUS_THIN_POOL,
 	SEG_STATUS_VDO_POOL,
 	SEG_STATUS_WRITECACHE,
+	SEG_STATUS_INTEGRITY,
 	SEG_STATUS_UNKNOWN
 } lv_seg_status_type_t;
 
@@ -53,6 +54,7 @@ struct lv_seg_status {
 		struct dm_status_thin *thin;
 		struct dm_status_thin_pool *thin_pool;
 		struct dm_status_writecache *writecache;
+		struct dm_status_integrity *integrity;
 		struct lv_status_vdo vdo_pool;
 	};
 };
@@ -260,6 +262,7 @@ void fs_unlock(void);
 
 #define TARGET_NAME_CACHE "cache"
 #define TARGET_NAME_WRITECACHE "writecache"
+#define TARGET_NAME_INTEGRITY "integrity"
 #define TARGET_NAME_ERROR "error"
 #define TARGET_NAME_ERROR_OLD "erro"	/* Truncated in older kernels */
 #define TARGET_NAME_LINEAR "linear"
@@ -277,6 +280,7 @@ void fs_unlock(void);
 #define MODULE_NAME_CLUSTERED_MIRROR "clog"
 #define MODULE_NAME_CACHE TARGET_NAME_CACHE
 #define MODULE_NAME_WRITECACHE TARGET_NAME_WRITECACHE
+#define MODULE_NAME_INTEGRITY TARGET_NAME_INTEGRITY
 #define MODULE_NAME_ERROR TARGET_NAME_ERROR
 #define MODULE_NAME_LOG_CLUSTERED "log-clustered"
 #define MODULE_NAME_LOG_USERSPACE "log-userspace"

lib/activate/dev_manager.c

@@ -222,6 +222,10 @@ static int _get_segment_status_from_target_params(const char *target_name,
 		if (!dm_get_status_writecache(seg_status->mem, params, &(seg_status->writecache)))
 			return_0;
 		seg_status->type = SEG_STATUS_WRITECACHE;
+	} else if (segtype_is_integrity(segtype)) {
+		if (!dm_get_status_integrity(seg_status->mem, params, &(seg_status->integrity)))
+			return_0;
+		seg_status->type = SEG_STATUS_INTEGRITY;
 	} else
 		/*
 		 * TODO: Add support for other segment types too!
@@ -299,6 +303,9 @@ static int _info_run(const char *dlid, struct dm_info *dminfo,
 		if (lv_is_vdo_pool(seg_status->seg->lv))
 			length = get_vdo_pool_virtual_size(seg_status->seg);
 
+		if (lv_is_integrity(seg_status->seg->lv))
+			length = seg_status->seg->integrity_data_sectors;
+
 		do {
 			target = dm_get_next_target(dmt, target, &target_start,
 						    &target_length, &target_name, &target_params);
@@ -2620,6 +2627,10 @@ static int _add_lv_to_dtree(struct dev_manager *dm, struct dm_tree *dtree,
 			if (!_add_lv_to_dtree(dm, dtree, seg->writecache, dm->activation ? origin_only : 1))
 				return_0;
 		}
+		if (seg->integrity_meta_dev && seg_is_integrity(seg)) {
+			if (!_add_lv_to_dtree(dm, dtree, seg->integrity_meta_dev, dm->activation ? origin_only : 1))
+				return_0;
+		}
 		if (seg->pool_lv &&
 		    (lv_is_cache_pool(seg->pool_lv) || lv_is_cache_vol(seg->pool_lv) || dm->track_external_lv_deps) &&
 		    /* When activating and not origin_only detect linear 'overlay' over pool */
@@ -3076,6 +3087,11 @@ static int _add_segment_to_dtree(struct dev_manager *dm,
 				  lv_layer(seg->writecache)))
 		return_0;
 
+	if (seg->integrity_meta_dev && !laopts->origin_only &&
+	    !_add_new_lv_to_dtree(dm, dtree, seg->integrity_meta_dev, laopts,
+				  lv_layer(seg->integrity_meta_dev)))
+		return_0;
+
 	/* Add any LVs used by this segment */
 	for (s = 0; s < seg->area_count; ++s) {
 		if ((seg_type(seg, s) == AREA_LV) &&

lib/commands/toolcontext.c

@@ -1362,6 +1362,9 @@ static int _init_segtypes(struct cmd_context *cmd)
 		return 0;
 #endif
 
+	if (!init_integrity_segtypes(cmd, &seglib))
+		return 0;
+
 	return 1;
 }
 

lib/format_text/flags.c

@@ -104,8 +104,10 @@ static const struct flag _lv_flags[] = {
 	{LV_VDO_POOL, NULL, 0},
 	{LV_VDO_POOL_DATA, NULL, 0},
 	{WRITECACHE, NULL, 0},
+	{INTEGRITY, NULL, 0},
 	{LV_PENDING_DELETE, NULL, 0}, /* FIXME Display like COMPATIBLE_FLAG */
 	{LV_REMOVED, NULL, 0},
+	{LV_UNCOMMITTED, NULL, 0},
 	{0, NULL, 0}
 };
 

lib/integrity/integrity.c

@@ -0,0 +1,325 @@
+/*
+ * Copyright (C) 2013-2016 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "base/memory/zalloc.h"
+#include "lib/misc/lib.h"
+#include "lib/commands/toolcontext.h"
+#include "lib/metadata/segtype.h"
+#include "lib/display/display.h"
+#include "lib/format_text/text_export.h"
+#include "lib/config/config.h"
+#include "lib/datastruct/str_list.h"
+#include "lib/misc/lvm-string.h"
+#include "lib/activate/activate.h"
+#include "lib/metadata/metadata.h"
+#include "lib/metadata/lv_alloc.h"
+#include "lib/config/defaults.h"
+
+#define SEG_LOG_ERROR(t, p...) \
+        log_error(t " segment %s of logical volume %s.", ## p,	\
+                  dm_config_parent_name(sn), seg->lv->name), 0;
+
+static void _integrity_display(const struct lv_segment *seg)
+{
+	/* TODO: lvdisplay segments */
+}
+
+static int _integrity_text_import(struct lv_segment *seg,
+				   const struct dm_config_node *sn,
+				   struct dm_hash_table *pv_hash __attribute__((unused)))
+{
+	struct integrity_settings *set;
+	struct logical_volume *origin_lv = NULL;
+	struct logical_volume *meta_lv = NULL;
+	const char *origin_name = NULL;
+	const char *meta_dev = NULL;
+	const char *mode = NULL;
+	const char *hash = NULL;
+
+	memset(&seg->integrity_settings, 0, sizeof(struct integrity_settings));
+	set = &seg->integrity_settings;
+
+	/* origin always set */
+
+	if (!dm_config_has_node(sn, "origin"))
+		return SEG_LOG_ERROR("origin not specified in");
+
+	if (!dm_config_get_str(sn, "origin", &origin_name))
+		return SEG_LOG_ERROR("origin must be a string in");
+
+	if (!(origin_lv = find_lv(seg->lv->vg, origin_name)))
+		return SEG_LOG_ERROR("Unknown LV specified for integrity origin %s in", origin_name);
+
+	if (!set_lv_segment_area_lv(seg, 0, origin_lv, 0, 0))
+		return_0;
+
+	/* data_sectors always set */
+
+	if (!dm_config_get_uint64(sn, "data_sectors", &seg->integrity_data_sectors))
+		return SEG_LOG_ERROR("integrity data_sectors must be set in");
+
+	/* mode always set */
+
+	if (!dm_config_get_str(sn, "mode", &mode))
+		return SEG_LOG_ERROR("integrity mode must be set in");
+
+	if (strlen(mode) > 7)
+		return SEG_LOG_ERROR("integrity mode invalid in");
+
+	strncpy(set->mode, mode, 7);
+
+	/* tag_size always set */
+
+	if (!dm_config_get_uint32(sn, "tag_size", &set->tag_size))
+		return SEG_LOG_ERROR("integrity tag_size must be set in");
+
+	/* internal_hash always set */
+
+	if (!dm_config_get_str(sn, "internal_hash", &hash))
+		return SEG_LOG_ERROR("integrity internal_hash must be set in");
+
+	if (!(set->internal_hash = strdup(hash)))
+		return SEG_LOG_ERROR("integrity internal_hash failed to be set in");
+
+	/* meta_dev optional */
+
+	if (dm_config_has_node(sn, "meta_dev")) {
+		if (!dm_config_get_str(sn, "meta_dev", &meta_dev))
+			return SEG_LOG_ERROR("meta_dev must be a string in");
+
+		if (!(meta_lv = find_lv(seg->lv->vg, meta_dev)))
+			return SEG_LOG_ERROR("Unknown logical volume %s specified for integrity in", meta_dev);
+	}
+
+	/* the rest are optional */
+
+	if (dm_config_has_node(sn, "journal_sectors")) {
+		if (!dm_config_get_uint32(sn, "journal_sectors", &set->journal_sectors))
+			return SEG_LOG_ERROR("Unknown integrity_setting in");
+		set->journal_sectors_set = 1;
+	}
+
+	if (dm_config_has_node(sn, "interleave_sectors")) {
+		if (!dm_config_get_uint32(sn, "interleave_sectors", &set->interleave_sectors))
+			return SEG_LOG_ERROR("Unknown integrity_setting in");
+		set->interleave_sectors_set = 1;
+	}
+
+	if (dm_config_has_node(sn, "buffer_sectors")) {
+		if (!dm_config_get_uint32(sn, "buffer_sectors", &set->buffer_sectors))
+			return SEG_LOG_ERROR("Unknown integrity_setting in");
+		set->buffer_sectors_set = 1;
+	}
+
+	if (dm_config_has_node(sn, "journal_watermark")) {
+		if (!dm_config_get_uint32(sn, "journal_watermark", &set->journal_watermark))
+			return SEG_LOG_ERROR("Unknown integrity_setting in");
+		set->journal_watermark_set = 1;
+	}
+
+	if (dm_config_has_node(sn, "commit_time")) {
+		if (!dm_config_get_uint32(sn, "commit_time", &set->commit_time))
+			return SEG_LOG_ERROR("Unknown integrity_setting in");
+		set->commit_time_set = 1;
+	}
+
+	if (dm_config_has_node(sn, "block_size")) {
+		if (!dm_config_get_uint32(sn, "block_size", &set->block_size))
+			return SEG_LOG_ERROR("Unknown integrity_setting in");
+		set->block_size_set = 1;
+	}
+
+	if (dm_config_has_node(sn, "bitmap_flush_interval")) {
+		if (!dm_config_get_uint32(sn, "bitmap_flush_interval", &set->bitmap_flush_interval))
+			return SEG_LOG_ERROR("Unknown integrity_setting in");
+		set->bitmap_flush_interval_set = 1;
+	}
+
+	if (dm_config_has_node(sn, "sectors_per_bit")) {
+		if (!dm_config_get_uint64(sn, "sectors_per_bit", &set->sectors_per_bit))
+			return SEG_LOG_ERROR("Unknown integrity_setting in");
+		set->sectors_per_bit_set = 1;
+	}
+
+	seg->origin = origin_lv;
+	seg->integrity_meta_dev = meta_lv;
+	seg->lv->status |= INTEGRITY;
+
+	if (meta_lv && !add_seg_to_segs_using_this_lv(meta_lv, seg))
+		return_0;
+
+	return 1;
+}
+
+static int _integrity_text_import_area_count(const struct dm_config_node *sn,
+					      uint32_t *area_count)
+{
+	*area_count = 1;
+
+	return 1;
+}
+
+static int _integrity_text_export(const struct lv_segment *seg,
+				   struct formatter *f)
+{
+	const struct integrity_settings *set = &seg->integrity_settings;
+
+	outf(f, "origin = \"%s\"", seg_lv(seg, 0)->name);
+	outf(f, "data_sectors = %llu", (unsigned long long)seg->integrity_data_sectors);
+
+	outf(f, "mode = \"%s\"", set->mode);
+	outf(f, "tag_size = %u", set->tag_size);
+	outf(f, "internal_hash = \"%s\"", set->internal_hash);
+
+	if (seg->integrity_meta_dev)
+		outf(f, "meta_dev = \"%s\"", seg->integrity_meta_dev->name);
+
+	if (set->journal_sectors_set)
+		outf(f, "journal_sectors = %u", set->journal_sectors);
+
+	if (set->interleave_sectors_set)
+		outf(f, "interleave_sectors = %u", set->interleave_sectors);
+
+	if (set->buffer_sectors_set)
+		outf(f, "buffer_sectors = %u", set->buffer_sectors);
+
+	if (set->journal_watermark_set)
+		outf(f, "journal_watermark = %u", set->journal_watermark);
+
+	if (set->commit_time_set)
+		outf(f, "commit_time = %u", set->commit_time);
+
+	if (set->block_size_set)
+		outf(f, "block_size = %u", set->block_size);
+
+	if (set->bitmap_flush_interval)
+		outf(f, "bitmap_flush_interval = %u", set->bitmap_flush_interval);
+
+	if (set->sectors_per_bit)
+		outf(f, "sectors_per_bit = %llu", (unsigned long long)set->sectors_per_bit);
+
+	return 1;
+}
+
+static void _destroy(struct segment_type *segtype)
+{
+	free((void *) segtype);
+}
+
+#ifdef DEVMAPPER_SUPPORT
+
+static int _target_present(struct cmd_context *cmd,
+			   const struct lv_segment *seg __attribute__((unused)),
+			   unsigned *attributes __attribute__((unused)))
+{
+	static int _integrity_checked = 0;
+	static int _integrity_present = 0;
+
+	if (!activation())
+		return 0;
+
+	if (!_integrity_checked) {
+		_integrity_checked = 1;
+		_integrity_present =  target_present(cmd, TARGET_NAME_INTEGRITY, 0);
+	}
+
+	return _integrity_present;
+}
+
+static int _modules_needed(struct dm_pool *mem,
+			   const struct lv_segment *seg __attribute__((unused)),
+			   struct dm_list *modules)
+{
+	if (!str_list_add(mem, modules, MODULE_NAME_INTEGRITY)) {
+		log_error("String list allocation failed for integrity module.");
+		return 0;
+	}
+
+	return 1;
+}
+#endif /* DEVMAPPER_SUPPORT */
+
+#ifdef DEVMAPPER_SUPPORT
+static int _integrity_add_target_line(struct dev_manager *dm,
+				 struct dm_pool *mem,
+				 struct cmd_context *cmd __attribute__((unused)),
+				 void **target_state __attribute__((unused)),
+				 struct lv_segment *seg,
+				 const struct lv_activate_opts *laopts __attribute__((unused)),
+				 struct dm_tree_node *node, uint64_t len,
+				 uint32_t *pvmove_mirror_count __attribute__((unused)))
+{
+	char *origin_uuid;
+	char *meta_uuid = NULL;
+
+	if (!seg_is_integrity(seg)) {
+		log_error(INTERNAL_ERROR "Passed segment is not integrity.");
+		return 0;
+	}
+
+	if (!(origin_uuid = build_dm_uuid(mem, seg_lv(seg, 0), NULL)))
+		return_0;
+
+	if (seg->integrity_meta_dev) {
+		if (!(meta_uuid = build_dm_uuid(mem, seg->integrity_meta_dev, NULL)))
+			return_0;
+	}
+
+	if (!seg->integrity_data_sectors) {
+		log_error("_integrity_add_target_line zero size");
+		return_0;
+	}
+	
+	if (!dm_tree_node_add_integrity_target(node, seg->integrity_data_sectors,
+					       origin_uuid, meta_uuid,
+					       &seg->integrity_settings))
+		return_0;
+
+	return 1;
+}
+#endif /* DEVMAPPER_SUPPORT */
+
+static struct segtype_handler _integrity_ops = {
+	.display = _integrity_display,
+	.text_import = _integrity_text_import,
+	.text_import_area_count = _integrity_text_import_area_count,
+	.text_export = _integrity_text_export,
+#ifdef DEVMAPPER_SUPPORT
+	.add_target_line = _integrity_add_target_line,
+	.target_present = _target_present,
+	.modules_needed = _modules_needed,
+#endif
+	.destroy = _destroy,
+};
+
+int init_integrity_segtypes(struct cmd_context *cmd,
+			struct segtype_library *seglib)
+{
+	struct segment_type *segtype = zalloc(sizeof(*segtype));
+
+	if (!segtype) {
+		log_error("Failed to allocate memory for integrity segtype");
+		return 0;
+	}
+
+	segtype->name = SEG_TYPE_NAME_INTEGRITY;
+	segtype->flags = SEG_INTEGRITY;
+	segtype->ops = &_integrity_ops;
+
+	if (!lvm_register_segtype(seglib, segtype))
+		return_0;
+	log_very_verbose("Initialised segtype: %s", segtype->name);
+
+	return 1;
+}

lib/metadata/integrity_manip.c

@@ -0,0 +1,264 @@
+/*
+ * Copyright (C) 2014-2015 Red Hat, Inc. All rights reserved.
+ *
+ * This file is part of LVM2.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU Lesser General Public License v.2.1.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "lib/misc/lib.h"
+#include "lib/metadata/metadata.h"
+#include "lib/locking/locking.h"
+#include "lib/misc/lvm-string.h"
+#include "lib/commands/toolcontext.h"
+#include "lib/display/display.h"
+#include "lib/metadata/segtype.h"
+#include "lib/activate/activate.h"
+#include "lib/config/defaults.h"
+#include "lib/activate/dev_manager.h"
+
+#define DEFAULT_TAG_SIZE 4 /* bytes */
+#define DEFAULT_MODE 'B'
+#define DEFAULT_INTERNAL_HASH "crc32c"
+
+#define ONE_MB_IN_BYTES 1048576
+
+/*
+ * The user wants external metadata, but did not specify an existing
+ * LV to hold metadata, so create an LV for metadata.  Save it in
+ * lp->integrity_meta_lv and it will be passed to lv_add_integrity().
+ */
+int lv_create_integrity_metadata(struct cmd_context *cmd,
+				struct volume_group *vg,
+				struct lvcreate_params *lp)
+{
+	char metaname[NAME_LEN];
+	uint32_t extent_bytes;
+	struct logical_volume *lv;
+	struct lvcreate_params lp_meta = {
+		.activate = CHANGE_AN,
+		.alloc = ALLOC_INHERIT,
+		.major = -1,
+		.minor = -1,
+		.permission = LVM_READ | LVM_WRITE,
+		.pvh = &vg->pvs,
+		.read_ahead = DM_READ_AHEAD_NONE,
+		.stripes = 1,
+		.vg_name = vg->name,
+		.zero = 0,
+		.wipe_signatures = 0,
+	};
+
+	if (dm_snprintf(metaname, NAME_LEN, "%s_imeta", lp->lv_name) < 0) {
+		log_error("Failed to create metadata LV name.");
+		return 0;
+	}
+
+	if (!(lp_meta.lv_name = strdup(metaname)))
+		return_0;
+
+	/* TODO: set pvh per command line */
+	/* TODO: scale size according to LV size. */
+	extent_bytes = vg->extent_size * SECTOR_SIZE;
+	lp_meta.extents = (4 * ONE_MB_IN_BYTES) / extent_bytes;
+
+	log_debug("Creating integrity metadata LV with %u extents", lp_meta.extents);
+
+	dm_list_init(&lp_meta.tags);
+
+	if (!(lp_meta.segtype = get_segtype_from_string(vg->cmd, SEG_TYPE_NAME_STRIPED)))
+		return_0;
+
+	if (!(lv = lv_create_single(vg, &lp_meta))) {
+		log_error("Failed to create integrity metadata LV");
+		return 0;
+	}
+
+	lp->integrity_meta_lv = lv;
+	return 1;
+}
+
+static int _get_provided_data_sectors(struct logical_volume *lv, uint64_t *provided_data_sectors)
+{
+	struct lv_with_info_and_seg_status status;
+
+	memset(&status, 0, sizeof(status));
+	status.seg_status.type = SEG_STATUS_NONE;
+
+	status.seg_status.seg = first_seg(lv);
+
+	/* FIXME: why reporter_pool? */
+	if (!(status.seg_status.mem = dm_pool_create("reporter_pool", 1024))) {
+		log_error("Failed to get mem for LV status.");
+		return 0;
+	}
+
+	if (!lv_info_with_seg_status(lv->vg->cmd, first_seg(lv), &status, 1, 1)) {
+		log_error("Failed to get device mapper status for %s", display_lvname(lv));
+		goto fail;
+	}
+
+	if (!status.info.exists) {
+		log_error("No device mapper info exists for %s", display_lvname(lv));
+		goto fail;
+	}
+
+	if (status.seg_status.type != SEG_STATUS_INTEGRITY) {
+		log_error("Invalid device mapper status type (%d) for %s",
+			  (uint32_t)status.seg_status.type, display_lvname(lv));
+		goto fail;
+	}
+
+	*provided_data_sectors = status.seg_status.integrity->provided_data_sectors;
+
+	dm_pool_destroy(status.seg_status.mem);
+	return 1;
+
+fail:
+	dm_pool_destroy(status.seg_status.mem);
+	return 0;
+}
+
+int lv_add_integrity(struct logical_volume *lv, const char *arg,
+		     struct logical_volume *meta_lv_created,
+		     const char *meta_name,
+		     struct integrity_settings *settings)
+{
+	struct cmd_context *cmd = lv->vg->cmd;
+	struct integrity_settings *set;
+	struct logical_volume *lv_orig;
+	struct logical_volume *meta_lv = NULL;
+	const struct segment_type *segtype;
+	struct lv_segment *seg;
+	int ret;
+
+	if (!(segtype = get_segtype_from_string(cmd, SEG_TYPE_NAME_INTEGRITY)))
+		return_0;
+
+	/*
+	 * "lv_orig" is a new LV with new id, but with the segments from "lv".
+	 * "lv" keeps the existing name and id, but gets a new integrity segment,
+	 * in place of the segments that were moved to lv_orig.
+	 */
+
+	if (!(lv_orig = insert_layer_for_lv(cmd, lv, INTEGRITY, "_iorig")))
+		return_0;
+
+	seg = first_seg(lv);
+	seg->segtype = segtype;
+
+	lv->status |= INTEGRITY;
+
+	memcpy(&seg->integrity_settings, settings, sizeof(struct integrity_settings));
+	set = &seg->integrity_settings;
+
+	if (!set->mode[0])
+		set->mode[0] = DEFAULT_MODE;
+
+	if (!set->tag_size)
+		set->tag_size = DEFAULT_TAG_SIZE;
+
+	if (!set->internal_hash)
+	       	set->internal_hash = DEFAULT_INTERNAL_HASH;
+
+	/*
+	 * --integrity <arg> is y|external|internal
+	 */
+
+	if (!arg)
+		arg = "external";
+
+	if (!strcmp(arg, "none")) {
+		log_error("Invalid --integrity arg for lvcreate.");
+		return 0;
+	}
+
+	if (!strcmp(arg, "internal") && meta_name) {
+		log_error("Internal integrity cannot be used with integritymetadata option.");
+		return 0;
+	}
+
+	if (meta_lv_created)
+		meta_lv = meta_lv_created;
+	else if (meta_name) {
+		if (!(meta_lv = find_lv(lv->vg, meta_name))) {
+			log_error("LV %s not found.", meta_name);
+			return_0;
+		}
+	}
+
+	/*
+	 * When not using a meta_dev, dm-integrity needs to tell us what the
+	 * usable size of the LV is, which is the dev size minus the integrity
+	 * overhead.  To find that, we need to do a special, temporary activation
+	 * of the new LV, specifying a dm dev size of 1, then check the dm dev
+	 * status field provided_data_sectors, which is the actual size of the
+	 * LV.  We need to include provided_data_sectors in the metadata for the
+	 * new LV, and use this as the dm dev size for normal LV activation.
+	 *
+	 * When using a meta_dev, the dm dev size is the size of the data
+	 * device.  The necessary size of the meta_dev for the given data
+	 * device needs to be estimated.
+	 */
+
+	if (meta_lv) {
+		if (!sync_local_dev_names(cmd)) {
+			log_error("Failed to sync local devices.");
+			return 0;
+		}
+
+		if (!activate_and_wipe_lv(meta_lv, 0)) {
+			log_error("LV %s could not be zeroed.", display_lvname(meta_lv));
+			return 0;
+		}
+		seg->integrity_data_sectors = seg->len;
+		seg->integrity_meta_dev = meta_lv;
+		lv_set_hidden(meta_lv);
+		/* TODO: give meta_lv a suffix? e.g. _imeta */
+		ret = 1;
+	} else {
+		/* dm-integrity wants temp/fake size of 1 to report usable size */
+		seg->integrity_data_sectors = 1;
+
+		lv->status |= LV_TEMPORARY;
+		lv->status |= LV_NOSCAN;
+		lv->status |= LV_UNCOMMITTED;
+
+		log_debug("Activating temporary integrity LV to get data sectors.");
+
+		if (!activate_lv(cmd, lv)) {
+			log_error("Failed to activate temporary integrity.");
+			ret = 0;
+			goto out;
+		}
+
+		if (!_get_provided_data_sectors(lv, &seg->integrity_data_sectors)) {
+			log_error("Failed to get data sectors from dm-integrity");
+			ret = 0;
+		} else {
+			log_debug("Found integrity provided_data_sectors %llu", (unsigned long long)seg->integrity_data_sectors);
+			ret = 1;
+		}
+
+		lv->status |= LV_UNCOMMITTED;
+
+		if (!deactivate_lv(cmd, lv)) {
+			log_error("Failed to deactivate temporary integrity.");
+			ret = 0;
+		}
+
+		lv->status &= ~LV_UNCOMMITTED;
+		lv->status &= ~LV_NOSCAN;
+		lv->status &= ~LV_TEMPORARY;
+	}
+ out:
+	return ret;
+}
+
+

lib/metadata/lv.c

@@ -585,6 +585,8 @@ struct logical_volume *lv_origin_lv(const struct logical_volume *lv)
 		origin = first_seg(lv)->external_lv;
 	else if (lv_is_writecache(lv) && first_seg(lv)->origin)
 		origin = first_seg(lv)->origin;
+	else if (lv_is_integrity(lv) && first_seg(lv)->origin)
+		origin = first_seg(lv)->origin;
 
 	return origin;
 }

lib/metadata/lv_manip.c

@@ -1457,6 +1457,15 @@ static int _lv_reduce(struct logical_volume *lv, uint32_t extents, int delete)
 					return_0;
 			}
 
+			if (delete && seg_is_integrity(seg)) {
+				/* Remove integrity origin in addition to integrity layer. */
+				if (!lv_remove(seg_lv(seg, 0)))
+					return_0;
+				/* Remove integrity metadata. */
+				if (seg->integrity_meta_dev && !lv_remove(seg->integrity_meta_dev))
+					return_0;
+			}
+
 			if ((pool_lv = seg->pool_lv)) {
 				if (!detach_pool_lv(seg))
 					return_0;
@@ -5654,6 +5663,11 @@ int lv_resize(struct logical_volume *lv,
 		return 0;
 	}
 
+	if (lv_is_integrity(lv)) {
+		log_error("Resize not yet allowed on LVs with integrity.");
+		return 0;
+	}
+
 	if (!_lvresize_check(lv, lp))
 		return_0;
 
@@ -7410,6 +7424,12 @@ int insert_layer_for_segments_on_pv(struct cmd_context *cmd,
 	return 1;
 }
 
+/* FIXME: copied from label.c */
+#define BCACHE_BLOCK_SIZE_IN_SECTORS 256 /* 256*512 = 128K */
+#define BCACHE_BLOCK_SIZE_IN_BYTES 131072
+#define ONE_MB_IN_BYTES 1048576
+#define ONE_MB_IN_SECTORS 2048   /* 2048 * 512 = 1048576 */
+
 /*
  * Initialize the LV with 'value'.
  */
@@ -7468,7 +7488,44 @@ int wipe_lv(struct logical_volume *lv, struct wipe_params wp)
 			stack;
 	}
 
-	if (wp.do_zero) {
+	if (wp.do_zero && !wp.zero_value && (wp.zero_sectors >= ONE_MB_IN_SECTORS)) {
+		uint64_t off = 0, i = 0, j = 0;
+		uint64_t zero_bytes;
+		uint32_t extra_bytes;
+
+		zero_bytes = wp.zero_sectors * 512;
+
+		if ((extra_bytes = (zero_bytes % ONE_MB_IN_BYTES)))
+			zero_bytes -= extra_bytes;
+
+		log_print("Zeroing %llu MiB...", (unsigned long long)(zero_bytes / ONE_MB_IN_BYTES));
+
+		/*
+		 * Write 1MiB at a time to avoid going over bcache size.
+		 * Then write 128KiB at a time to cover remaining dev size.
+		 */
+
+		for (i = 0; i < (zero_bytes / ONE_MB_IN_BYTES); i++) {
+			off = i * ONE_MB_IN_BYTES;
+			if (!dev_write_zeros(dev, off, (size_t)ONE_MB_IN_BYTES))
+				log_error("Failed to zero LV at offset %llu.", (unsigned long long)off);
+		}
+
+		if (extra_bytes) {
+			log_warn("Zeroing %u bytes at %llu...", extra_bytes, (unsigned long long)off);
+
+			for (j = 0; j < (extra_bytes / BCACHE_BLOCK_SIZE_IN_BYTES); j++) {
+				off = i * ONE_MB_IN_BYTES + j * BCACHE_BLOCK_SIZE_IN_BYTES;
+				if (!dev_write_zeros(dev, off, (size_t)BCACHE_BLOCK_SIZE_IN_BYTES))
+					log_error("Failed to zero LV at offset %llu.", (unsigned long long)off);
+			}
+		}
+
+		/* FIXME: bcache can't write partial block yet */
+		if ((extra_bytes = (wp.zero_sectors * 512) - (i * ONE_MB_IN_BYTES + j * BCACHE_BLOCK_SIZE_IN_BYTES)))
+			log_warn("WARNING: last %llu bytes not zeroed.", (unsigned long long)extra_bytes);
+
+	} else if (wp.do_zero) {
 		zero_sectors = wp.zero_sectors ? : UINT64_C(4096) >> SECTOR_SHIFT;
 
 		if (zero_sectors > lv->size)
@@ -7679,6 +7736,11 @@ static int _should_wipe_lv(struct lvcreate_params *lp,
 	     first_seg(first_seg(lv)->pool_lv)->zero_new_blocks))
 		return 0;
 
+	if (seg_is_integrity(lp) && (!lp->zero || !(lv->status & LVM_WRITE))) {
+		log_warn("WARNING: --zero not enabled, integrity will not be initialized and may cause read errors.");
+		return 0;
+	}
+
 	/* Cannot zero read-only volume */
 	if ((lv->status & LVM_WRITE) &&
 	    (lp->zero || lp->wipe_signatures))
@@ -7934,6 +7996,18 @@ static struct logical_volume *_lv_create_an_lv(struct volume_group *vg,
 		/* FIXME Eventually support raid/mirrors with -m */
 		if (!(create_segtype = get_segtype_from_string(vg->cmd, SEG_TYPE_NAME_STRIPED)))
 			return_0;
+
+	} else if (seg_is_integrity(lp)) {
+		/*
+		 * TODO: if using internal metadata, estimate the amount of metadata
+		 * that will be needed, and add this to the amount of PV space being
+		 * allocated so that the usable LV size is what the user requested.
+		 * Or, just request an extra extent_size bytes, then round the
+		 * provided_data_sectors down to be an extent_size multiple.
+		 */
+		if (!(create_segtype = get_segtype_from_string(vg->cmd, SEG_TYPE_NAME_STRIPED)))
+			return_0;
+
 	} else if (seg_is_mirrored(lp) || (seg_is_raid(lp) && !seg_is_any_raid0(lp))) {
 		if (!(lp->region_size = adjusted_mirror_region_size(vg->cmd,
 								    vg->extent_size,
@@ -8158,6 +8232,12 @@ static struct logical_volume *_lv_create_an_lv(struct volume_group *vg,
 		}
 	}
 
+	if (seg_is_integrity(lp)) {
+		if (!lv_add_integrity(lv, lp->integrity_arg, lp->integrity_meta_lv,
+				      lp->integrity_meta_name, &lp->integrity_settings))
+			return_NULL;
+	}
+
 	lv_set_activation_skip(lv, lp->activation_skip & ACTIVATION_SKIP_SET,
 			       lp->activation_skip & ACTIVATION_SKIP_SET_ENABLED);
 	/*
@@ -8282,7 +8362,22 @@ static struct logical_volume *_lv_create_an_lv(struct volume_group *vg,
 		goto deactivate_and_revert_new_lv;
 	}
 
-	if (_should_wipe_lv(lp, lv, !lp->suppress_zero_warn)) {
+	if (seg_is_integrity(lp)) {
+		struct wipe_params wipe;
+
+		memset(&wipe, 0, sizeof(wipe));
+		wipe.do_zero = 1;
+		wipe.zero_sectors = first_seg(lv)->integrity_data_sectors;
+
+		if (!_should_wipe_lv(lp, lv, 1))
+			goto_out;
+
+		if (!wipe_lv(lv, wipe))
+			log_error("Failed to zero LV.");
+
+		goto out;
+
+	} else if (_should_wipe_lv(lp, lv, !lp->suppress_zero_warn)) {
 		if (!wipe_lv(lv, (struct wipe_params)
 			     {
 				     .do_zero = lp->zero,

lib/metadata/merge.c

@@ -742,6 +742,8 @@ int check_lv_segments(struct logical_volume *lv, int complete_vg)
 			seg_found++;
 		if (seg->metadata_lv == lv || seg->pool_lv == lv || seg->writecache == lv)
 			seg_found++;
+		if (seg->integrity_meta_dev == lv)
+			seg_found++;
 		if (seg_is_thin_volume(seg) && (seg->origin == lv || seg->external_lv == lv))
 			seg_found++;
 

lib/metadata/metadata-exported.h

@@ -84,12 +84,14 @@
 #define CONVERTING		UINT64_C(0x0000000000400000)	/* LV */
 
 #define MISSING_PV		UINT64_C(0x0000000000800000)	/* PV */
+#define INTEGRITY		UINT64_C(0x0000000000800000)	/* LV - Internal use only */
 #define PV_MOVED_VG		UINT64_C(0x4000000000000000)	/* PV - Moved to a new VG */
 #define PARTIAL_LV		UINT64_C(0x0000000001000000)	/* LV - derived flag, not
 							   written out in metadata*/
 
 //#define POSTORDER_FLAG	UINT64_C(0x0000000002000000) /* Not real flags, reserved for
 //#define POSTORDER_OPEN_FLAG	UINT64_C(0x0000000004000000)    temporary use inside vg_read_internal. */
+#define LV_UNCOMMITTED		UINT64_C(0x0000000002000000)
 #define VIRTUAL_ORIGIN		UINT64_C(0x0000000008000000)	/* LV - internal use only */
 
 #define MERGING			UINT64_C(0x0000000010000000)	/* LV SEG */
@@ -261,6 +263,7 @@
 #define lv_is_pool_metadata_spare(lv)	(((lv)->status & POOL_METADATA_SPARE) ? 1 : 0)
 #define lv_is_lockd_sanlock_lv(lv)	(((lv)->status & LOCKD_SANLOCK_LV) ? 1 : 0)
 #define lv_is_writecache(lv)   (((lv)->status & WRITECACHE) ? 1 : 0)
+#define lv_is_integrity(lv)   (((lv)->status & INTEGRITY) ? 1 : 0)
 
 #define lv_is_vdo(lv)		(((lv)->status & LV_VDO) ? 1 : 0)
 #define lv_is_vdo_pool(lv)	(((lv)->status & LV_VDO_POOL) ? 1 : 0)
@@ -519,6 +522,10 @@ struct lv_segment {
 	uint32_t writecache_block_size;		/* For writecache */
 	struct writecache_settings writecache_settings; /* For writecache */
 
+	uint64_t integrity_data_sectors;
+	struct logical_volume *integrity_meta_dev;
+	struct integrity_settings integrity_settings;
+
 	struct dm_vdo_target_params vdo_params;	/* For VDO-pool */
 	uint32_t vdo_pool_header_size;		/* For VDO-pool */
 	uint32_t vdo_pool_virtual_extents;	/* For VDO-pool */
@@ -992,6 +999,11 @@ struct lvcreate_params {
 	alloc_policy_t alloc; /* all */
 	struct dm_vdo_target_params vdo_params; /* vdo */
 
+	const char *integrity_arg;
+	const char *integrity_meta_name; /* external LV is user-specified */
+	struct logical_volume *integrity_meta_lv; /* external LV we create */
+	struct integrity_settings integrity_settings;
+
 	struct dm_list tags;	/* all */
 
 	int yes;
@@ -1385,4 +1397,12 @@ int vg_is_foreign(struct volume_group *vg);
 
 void vg_write_commit_bad_mdas(struct cmd_context *cmd, struct volume_group *vg);
 
+int lv_add_integrity(struct logical_volume *lv, const char *arg,
+		     struct logical_volume *meta_lv_created,
+		     const char *meta_name, struct integrity_settings *settings);
+int lv_create_integrity_metadata(struct cmd_context *cmd,
+			struct volume_group *vg,
+			struct lvcreate_params *lp);
+
+
 #endif

lib/metadata/segtype.h

@@ -67,6 +67,7 @@ struct dev_manager;
 #define SEG_RAID6_N_6		(1ULL << 35)
 #define SEG_RAID6		SEG_RAID6_ZR
 #define SEG_WRITECACHE		(1ULL << 36)
+#define SEG_INTEGRITY		(1ULL << 37)
 
 #define SEG_STRIPED_TARGET	(1ULL << 39)
 #define SEG_LINEAR_TARGET	(1ULL << 40)
@@ -84,6 +85,7 @@ struct dev_manager;
 #define SEG_TYPE_NAME_CACHE		"cache"
 #define SEG_TYPE_NAME_CACHE_POOL	"cache-pool"
 #define SEG_TYPE_NAME_WRITECACHE	"writecache"
+#define SEG_TYPE_NAME_INTEGRITY		"integrity"
 #define SEG_TYPE_NAME_ERROR		"error"
 #define SEG_TYPE_NAME_FREE		"free"
 #define SEG_TYPE_NAME_ZERO		"zero"
@@ -117,6 +119,7 @@ struct dev_manager;
 #define segtype_is_cache(segtype)	((segtype)->flags & SEG_CACHE ? 1 : 0)
 #define segtype_is_cache_pool(segtype)	((segtype)->flags & SEG_CACHE_POOL ? 1 : 0)
 #define segtype_is_writecache(segtype)	((segtype)->flags & SEG_WRITECACHE ? 1 : 0)
+#define segtype_is_integrity(segtype)	((segtype)->flags & SEG_INTEGRITY ? 1 : 0)
 #define segtype_is_mirrored(segtype)	((segtype)->flags & SEG_AREAS_MIRRORED ? 1 : 0)
 #define segtype_is_mirror(segtype)	((segtype)->flags & SEG_MIRROR ? 1 : 0)
 #define segtype_is_pool(segtype)	((segtype)->flags & (SEG_CACHE_POOL | SEG_THIN_POOL)  ? 1 : 0)
@@ -179,6 +182,7 @@ struct dev_manager;
 #define seg_is_cache(seg)	segtype_is_cache((seg)->segtype)
 #define seg_is_cache_pool(seg)	segtype_is_cache_pool((seg)->segtype)
 #define seg_is_writecache(seg)	segtype_is_writecache((seg)->segtype)
+#define seg_is_integrity(seg)	segtype_is_integrity((seg)->segtype)
 #define seg_is_used_cache_pool(seg) (seg_is_cache_pool(seg) && (!dm_list_empty(&(seg->lv)->segs_using_this_lv)))
 #define seg_is_linear(seg)	(seg_is_striped(seg) && ((seg)->area_count == 1))
 #define seg_is_mirror(seg)	segtype_is_mirror((seg)->segtype)
@@ -347,6 +351,8 @@ int init_vdo_segtypes(struct cmd_context *cmd, struct segtype_library *seglib);
 
 int init_writecache_segtypes(struct cmd_context *cmd, struct segtype_library *seglib);
 
+int init_integrity_segtypes(struct cmd_context *cmd, struct segtype_library *seglib);
+
 #define CACHE_FEATURE_POLICY_MQ			(1U << 0)
 #define CACHE_FEATURE_POLICY_SMQ		(1U << 1)
 #define CACHE_FEATURE_METADATA2			(1U << 2)

tools/args.h

@@ -272,6 +272,15 @@ arg(ignoreunsupported_ARG, '\0', "ignoreunsupported", 0, 0, 0,
     "and \\fBdiff\\fP types include unsupported settings in their output by default,\n"
     "all the other types ignore unsupported settings.\n")
 
+arg(integrity_ARG, '\0', "integrity", string_VAL, 0, 0,
+    "Controls if integrity metadata should be stored and checked for an LV.\n")
+
+arg(integritymetadata_ARG, '\0', "integritymetadata", lv_VAL, 0, 0,
+    "The name of an LV to hold integrity metadata.\n")
+
+arg(integritysettings_ARG, '\0', "integritysettings", string_VAL, ARG_GROUPABLE, 0,
+    "Set dm-integrity parameters.\n")
+
 arg(labelsector_ARG, '\0', "labelsector", number_VAL, 0, 0,
     "By default the PV is labelled with an LVM2 identifier in its second\n"
     "sector (sector 1). This lets you use a different sector near the\n"

tools/command-lines.in

@@ -1269,6 +1269,20 @@ FLAGS: SECONDARY_SYNTAX
 
 ---
 
+lvcreate --type integrity --size SizeMB VG
+OO: --integrity String, --integritymetadata LV, --integritysettings String, OO_LVCREATE
+OP: PV ...
+ID: lvcreate_integrity
+DESC: Create a LV with integrity.
+
+lvcreate --integrity String --size SizeMB VG
+OO: --type integrity, --integritymetadata LV, --integritysettings String, OO_LVCREATE
+OP: PV ...
+ID: lvcreate_integrity
+DESC: Create a LV with integrity (infers --type integrity).
+
+---
+
 lvdisplay
 OO: --aligned, --all, --binary, --colon, --columns,
 --configreport ConfigReport, --foreign, --history, --ignorelockingfailure,

tools/lv_types.h

@@ -34,5 +34,6 @@ lvt(raid10_LVT, "raid10", NULL)
 lvt(error_LVT, "error", NULL)
 lvt(zero_LVT, "zero", NULL)
 lvt(writecache_LVT, "writecache", NULL)
+lvt(integrity_LVT, "integrity", NULL)
 lvt(LVT_COUNT, "", NULL)
 

tools/lvcreate.c

@@ -785,6 +785,8 @@ static int _lvcreate_params(struct cmd_context *cmd,
 		mirror_default_cfg = (arg_uint_value(cmd, stripes_ARG, 1) > 1)
 			? global_raid10_segtype_default_CFG : global_mirror_segtype_default_CFG;
 		segtype_str = find_config_tree_str(cmd, mirror_default_cfg, NULL);
+	} else if (arg_is_set(cmd, integrity_ARG)) {
+		segtype_str = SEG_TYPE_NAME_INTEGRITY;
 	} else
 		segtype_str = SEG_TYPE_NAME_STRIPED;
 
@@ -1218,6 +1220,11 @@ static int _lvcreate_params(struct cmd_context *cmd,
 		}
 	}
 
+	if (seg_is_integrity(lp)) {
+		if (!get_integrity_options(cmd, &lp->integrity_arg, &lp->integrity_meta_name, &lp->integrity_settings))
+			return 0;
+	}
+
 	lcp->pv_count = argc;
 	lcp->pvs = argv;
 
@@ -1702,6 +1709,24 @@ static int _lvcreate_single(struct cmd_context *cmd, const char *vg_name,
 			    lp->pool_name ? : "with generated name", lp->vg_name, lp->segtype->name);
 	}
 
+	/*
+	 * Create an LV to hold integrity metadata when:
+	 *
+	 * . The user wants external (not internal) metadata.  External
+	 *   is indicated by: no option set (external is default), or
+	 *   --integrity y|external.
+	 *
+	 * . The user did not specify an existing metadata LV with
+	 *   --integritymetadata, saved as lp->integrity_meta_name.
+	 */
+	if (seg_is_integrity(lp) && !lp->integrity_meta_name &&
+	    (!lp->integrity_arg ||
+	     !strcmp(lp->integrity_arg, "external") ||
+	     !strcmp(lp->integrity_arg, "y"))) {
+		if (!lv_create_integrity_metadata(cmd, vg, lp))
+			goto_out;
+	}
+
 	if (vg->lock_type && !strcmp(vg->lock_type, "sanlock")) {
 		if (!handle_sanlock_lv(cmd, vg)) {
 			log_error("No space for sanlock lock, extend the internal lvmlock LV.");

tools/toollib.c

@@ -1414,6 +1414,174 @@ out:
 	return ok;
 }
 
+
+static int _get_one_integrity_setting(struct cmd_context *cmd, struct integrity_settings *settings,
+				       char *key, char *val)
+{
+	if (!strncmp(key, "mode", strlen("mode"))) {
+		if (*val == 'D')
+			settings->mode[0] = 'D';
+		else if (*val == 'J')
+			settings->mode[0] = 'J';
+		else if (*val == 'B')
+			settings->mode[0] = 'B';
+		else if (*val == 'R')
+			settings->mode[0] = 'R';
+		else
+			goto_bad;
+		/* lvm assigns a default if the user doesn't. */
+		return 1;
+	}
+
+	if (!strncmp(key, "tag_size", strlen("tag_size"))) {
+		if (sscanf(val, "%u", &settings->tag_size) != 1)
+			goto_bad;
+		/* lvm assigns a default if the user doesn't. */
+		return 1;
+	}
+
+	if (!strncmp(key, "internal_hash", strlen("internal_hash"))) {
+		if (!(settings->internal_hash = strdup(val)))
+			goto_bad;
+		/* lvm assigns a default if the user doesn't. */
+		return 1;
+	}
+
+	/*
+	 * For the following settings, lvm does not set a default value if the
+	 * user does not specify their own value, and lets dm-integrity use its
+	 * own default.
+	 */
+
+	if (!strncmp(key, "journal_sectors", strlen("journal_sectors"))) {
+		if (sscanf(val, "%u", &settings->journal_sectors) != 1)
+			goto_bad;
+		settings->journal_sectors_set = 1;
+		return 1;
+	}
+
+	if (!strncmp(key, "interleave_sectors", strlen("interleave_sectors"))) {
+		if (sscanf(val, "%u", &settings->interleave_sectors) != 1)
+			goto_bad;
+		settings->interleave_sectors_set = 1;
+		return 1;
+	}
+
+	if (!strncmp(key, "buffer_sectors", strlen("buffer_sectors"))) {
+		if (sscanf(val, "%u", &settings->buffer_sectors) != 1)
+			goto_bad;
+		settings->buffer_sectors_set = 1;
+		return 1;
+	}
+
+	if (!strncmp(key, "journal_watermark", strlen("journal_watermark"))) {
+		if (sscanf(val, "%u", &settings->journal_watermark) != 1)
+			goto_bad;
+		settings->journal_watermark_set = 1;
+		return 1;
+	}
+
+	if (!strncmp(key, "commit_time", strlen("commit_time"))) {
+		if (sscanf(val, "%u", &settings->commit_time) != 1)
+			goto_bad;
+		settings->commit_time_set = 1;
+		return 1;
+	}
+
+	if (!strncmp(key, "block_size", strlen("block_size"))) {
+		if (sscanf(val, "%u", &settings->block_size) != 1)
+			goto_bad;
+		settings->block_size_set = 1;
+		return 1;
+	}
+
+	if (!strncmp(key, "bitmap_flush_interval", strlen("bitmap_flush_interval"))) {
+		if (sscanf(val, "%u", &settings->bitmap_flush_interval) != 1)
+			goto_bad;
+		settings->bitmap_flush_interval_set = 1;
+		return 1;
+	}
+
+	if (!strncmp(key, "sectors_per_bit", strlen("sectors_per_bit"))) {
+		if (sscanf(val, "%llu", (unsigned long long *)&settings->sectors_per_bit) != 1)
+			goto_bad;
+		settings->sectors_per_bit_set = 1;
+		return 1;
+	}
+
+	log_error("Unknown setting: %s", key);
+	return 0;
+
+ bad:
+	log_error("Invalid setting: %s", key);
+	return 0;
+}
+
+static int _get_integrity_settings(struct cmd_context *cmd, struct integrity_settings *settings)
+{
+	struct arg_value_group_list *group;
+	const char *str;
+	char key[64];
+	char val[64];
+	int num;
+	int pos;
+
+	/*
+	 * "grouped" means that multiple --integritysettings options can be used.
+	 * Each option is also allowed to contain multiple key = val pairs.
+	 */
+
+	dm_list_iterate_items(group, &cmd->arg_value_groups) {
+		if (!grouped_arg_is_set(group->arg_values, integritysettings_ARG))
+			continue;
+
+		if (!(str = grouped_arg_str_value(group->arg_values, integritysettings_ARG, NULL)))
+			break;
+
+		pos = 0;
+
+		while (pos < strlen(str)) {
+			/* scan for "key1=val1 key2 = val2  key3= val3" */
+
+			memset(key, 0, sizeof(key));
+			memset(val, 0, sizeof(val));
+
+			if (sscanf(str + pos, " %63[^=]=%63s %n", key, val, &num) != 2) {
+				log_error("Invalid setting at: %s", str+pos);
+				return 0;
+			}
+
+			pos += num;
+
+			if (!_get_one_integrity_setting(cmd, settings, key, val))
+				return_0;
+		}
+	}
+
+	return 1;
+}
+
+int get_integrity_options(struct cmd_context *cmd, const char **arg, const char **meta_name,
+			  struct integrity_settings *set)
+{
+	*arg = NULL;
+	*meta_name = NULL;
+	memset(set, 0, sizeof(struct integrity_settings));
+
+	if (arg_is_set(cmd, integrity_ARG))
+		*arg = arg_str_value(cmd, integrity_ARG, NULL);
+
+	if (arg_is_set(cmd, integritymetadata_ARG))
+		*meta_name = arg_str_value(cmd, integritymetadata_ARG, NULL);
+
+	if (arg_is_set(cmd, integritysettings_ARG)) {
+		if (!_get_integrity_settings(cmd, set))
+			return_0;
+	}
+
+	return 1;
+}
+
 /* FIXME move to lib */
 static int _pv_change_tag(struct physical_volume *pv, const char *tag, int addtag)
 {
@@ -2579,6 +2747,8 @@ static int _lv_is_type(struct cmd_context *cmd, struct logical_volume *lv, int l
 		return seg_is_raid10(seg);
 	case writecache_LVT:
 		return seg_is_writecache(seg);
+	case integrity_LVT:
+		return seg_is_integrity(seg);
 	case error_LVT:
 		return !strcmp(seg->segtype->name, SEG_TYPE_NAME_ERROR);
 	case zero_LVT:
@@ -2637,6 +2807,8 @@ int get_lvt_enum(struct logical_volume *lv)
 		return raid10_LVT;
 	if (seg_is_writecache(seg))
 		return writecache_LVT;
+	if (seg_is_integrity(seg))
+		return integrity_LVT;
 
 	if (!strcmp(seg->segtype->name, SEG_TYPE_NAME_ERROR))
 		return error_LVT;

tools/tools.h

@@ -233,6 +233,9 @@ struct lv_prop *get_lv_prop(int lvp_enum);
 struct lv_type *get_lv_type(int lvt_enum);
 struct command *get_command(int cmd_enum);
 
+int get_integrity_options(struct cmd_context *cmd, const char **arg, const char **meta_name,
+                          struct integrity_settings *set);
+
 int lvchange_properties_cmd(struct cmd_context *cmd, int argc, char **argv);
 int lvchange_activate_cmd(struct cmd_context *cmd, int argc, char **argv);
 int lvchange_refresh_cmd(struct cmd_context *cmd, int argc, char **argv);