dm-integrity with raid1 support

Create a raid1 LV where the raid images use dm-integrity.
Only external metadata is currently enabled.

lvcreate --type raid1 --mirrors Num --integrity y ...

daemons/dmeventd/dmeventd.c

@@ -1745,7 +1745,8 @@ static void _init_thread_signals(void)
 	sigdelset(&my_sigset, SIGHUP);
 	sigdelset(&my_sigset, SIGQUIT);
 
-	pthread_sigmask(SIG_BLOCK, &my_sigset, NULL);
+	if (pthread_sigmask(SIG_BLOCK, &my_sigset, NULL))
+		log_sys_error("pthread_sigmask", "SIG_BLOCK");
 }
 
 /*

daemons/dmeventd/plugins/vdo/dmeventd_vdo.c

@@ -16,7 +16,12 @@
 #include "daemons/dmeventd/plugins/lvm2/dmeventd_lvm.h"
 #include "daemons/dmeventd/libdevmapper-event.h"
 
-/* Use parser from new device_mapper library */
+/*
+ * Use parser from new device_mapper library.
+ * Although during compilation we can see dm_vdo_status_parse()
+ * in runtime we are linked agains systems libdm 'older' library
+ * which does not provide this symbol and plugin fails to load
+ */
 #include "device_mapper/vdo/status.c"
 
 #include <sys/wait.h>

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 +979,37 @@ 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 recalculate; /* not persistent */
+};
+
+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-common.c

@@ -2012,7 +2012,8 @@ static int _sysfs_get_kernel_name(uint32_t major, uint32_t minor, char *buf, siz
 			log_sys_error("readlink", sysfs_path);
 		else {
 			log_sys_debug("readlink", sysfs_path);
-			return _sysfs_find_kernel_name(major, minor, buf, buf_size);
+			r = _sysfs_find_kernel_name(major, minor, buf, buf_size);
+			goto out;
 		}
 		goto bad;
 	}
@@ -2033,6 +2034,7 @@ static int _sysfs_get_kernel_name(uint32_t major, uint32_t minor, char *buf, siz
 	strcpy(buf, name);
 	r = 1;
 bad:
+out:
 	free(temp_buf);
 	free(sysfs_path);
 

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 +2711,88 @@ 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++;
+
+	if (set->recalculate)
+		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);
+
+	if (set->recalculate)
+		EMIT_PARAMS(pos, " recalculate");
+
+	return 1;
+}
+
 static int _thin_pool_emit_segment_line(struct dm_task *dmt,
 					struct load_segment *seg,
 					char *params, size_t paramsize)
@@ -2889,6 +2977,10 @@ 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 +2993,7 @@ 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 +3831,39 @@ 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/cache/lvmcache.c

@@ -21,7 +21,6 @@
 #include "lib/locking/locking.h"
 #include "lib/metadata/metadata.h"
 #include "lib/mm/memlock.h"
-#include "lib/datastruct/str_list.h"
 #include "lib/format_text/format-text.h"
 #include "lib/config/config.h"
 
@@ -1952,24 +1951,25 @@ struct lvmcache_info *lvmcache_add(struct labeller *labeller,
 
 			strncpy(dev->pvid, pvid_s, sizeof(dev->pvid));
 
-			/*
-			 * Keep the existing PV/dev in lvmcache, and save the
-			 * new duplicate in the list of duplicates.  After
-			 * scanning is complete, compare the duplicate devs
-			 * with those in lvmcache to check if one of the
-			 * duplicates is preferred and if so switch lvmcache to
-			 * use it.
-			 */
-
-			if (!(devl = zalloc(sizeof(*devl))))
-				return_NULL;
-			devl->dev = dev;
-
 			/* shouldn't happen */
 			if (dev_in_device_list(dev, &_initial_duplicates))
 				log_debug_cache("Initial duplicate already in list %s", dev_name(dev));
-			else
+			else {
+				/*
+				 * Keep the existing PV/dev in lvmcache, and save the
+				 * new duplicate in the list of duplicates.  After
+				 * scanning is complete, compare the duplicate devs
+				 * with those in lvmcache to check if one of the
+				 * duplicates is preferred and if so switch lvmcache to
+				 * use it.
+				 */
+
+				if (!(devl = zalloc(sizeof(*devl))))
+					return_NULL;
+				devl->dev = dev;
+
 				dm_list_add(&_initial_duplicates, &devl->list);
+			}
 
 			if (is_duplicate)
 				*is_duplicate = 1;

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/device/dev-type.c

@@ -42,7 +42,6 @@ int dev_is_pmem(struct device *dev)
 {
 	FILE *fp;
 	char path[PATH_MAX];
-	char buffer[64];
 	int is_pmem = 0;
 
 	if (dm_snprintf(path, sizeof(path), "%sdev/block/%d:%d/queue/dax",
@@ -56,27 +55,16 @@ int dev_is_pmem(struct device *dev)
 	if (!(fp = fopen(path, "r")))
 		return 0;
 
-	if (!fgets(buffer, sizeof(buffer), fp)) {
-		log_warn("Failed to read %s.", path);
-		if (fclose(fp))
-			log_sys_debug("fclose", path);
-		return 0;
-	} else if (sscanf(buffer, "%d", &is_pmem) != 1) {
-		log_warn("Failed to parse %s '%s'.", path, buffer);
-		if (fclose(fp))
-			log_sys_debug("fclose", path);
-		return 0;
-	}
+	if (fscanf(fp, "%d", &is_pmem) != 1)
+		log_warn("Failed to parse DAX %s.", path);
+
+	if (is_pmem)
+		log_debug("%s is pmem", dev_name(dev));
 
 	if (fclose(fp))
 		log_sys_debug("fclose", path);
 
-	if (is_pmem) {
-		log_debug("%s is pmem", dev_name(dev));
-		return 1;
-	}
-
-	return 0;
+	return is_pmem ? 1 : 0;
 }
 
 int dev_is_lv(struct device *dev)
@@ -84,6 +72,7 @@ int dev_is_lv(struct device *dev)
 	FILE *fp;
 	char path[PATH_MAX];
 	char buffer[64];
+	int ret = 0;
 
 	if (dm_snprintf(path, sizeof(path), "%sdev/block/%d:%d/dm/uuid",
 			dm_sysfs_dir(),
@@ -96,17 +85,15 @@ int dev_is_lv(struct device *dev)
 	if (!(fp = fopen(path, "r")))
 		return 0;
 
-	if (!fgets(buffer, sizeof(buffer), fp)) {
+	if (!fgets(buffer, sizeof(buffer), fp))
 		log_warn("Failed to read %s.", path);
-		fclose(fp);
-		return 0;
-	}
+	else if (!strncmp(buffer, "LVM-", 4))
+		ret = 1;
 
-	fclose(fp);
+	if (fclose(fp))
+		log_sys_debug("fclose", path);
 
-	if (!strncmp(buffer, "LVM-", 4))
-		return 1;
-	return 0;
+	return ret;
 }
 
 struct dev_types *create_dev_types(const char *proc_dir,

lib/format_text/flags.c

@@ -104,8 +104,11 @@ static const struct flag _lv_flags[] = {
 	{LV_VDO_POOL, NULL, 0},
 	{LV_VDO_POOL_DATA, NULL, 0},
 	{WRITECACHE, NULL, 0},
+	{INTEGRITY, NULL, 0},
+	{INTEGRITY_METADATA, NULL, 0},
 	{LV_PENDING_DELETE, NULL, 0}, /* FIXME Display like COMPATIBLE_FLAG */
 	{LV_REMOVED, NULL, 0},
+	{LV_UNCOMMITTED, NULL, 0},
 	{0, NULL, 0}
 };
 

lib/format_text/format-text.c

@@ -277,7 +277,6 @@ static int _raw_write_mda_header(const struct format_type *fmt,
 	dev_set_last_byte(dev, start_byte + MDA_HEADER_SIZE);
 
 	if (!dev_write_bytes(dev, start_byte, MDA_HEADER_SIZE, mdah)) {
-		dev_unset_last_byte(dev);
 		log_error("Failed to write mda header to %s fd %d", dev_name(dev), dev->bcache_fd);
 		return 0;
 	}
@@ -342,7 +341,7 @@ static struct raw_locn *_read_metadata_location_vg(struct device_area *dev_area,
 	 * Don't try to check existing metadata
 	 * if given vgname is an empty string.
 	 */
-	if (!*vgname)
+	if (!vgname || !*vgname)
 		return rlocn;
 
 	/*
@@ -532,29 +531,6 @@ static struct volume_group *_vg_read_precommit_raw(struct format_instance *fid,
 	return vg;
 }
 
-#define MAX_DESC_LEN 2048
-
-static char *_build_desc_write(struct cmd_context *cmd, struct volume_group *vg)
-{
-	size_t len = strlen(cmd->cmd_line) + 32;
-	char *desc;
-
-	if (len > MAX_DESC_LEN)
-		len = MAX_DESC_LEN;
-
-	if (!(desc = zalloc(len)))
-		return_NULL;
-
-	vg->write_count++;
-
-	if (vg->write_count == 1)
-		dm_snprintf(desc, len, "Write from %s.", cmd->cmd_line);
-	else
-		dm_snprintf(desc, len, "Write[%u] from %s.", vg->write_count, cmd->cmd_line);
-
-	return desc;
-}
-
 /*
  * VG metadata updates:
  *
@@ -599,6 +575,7 @@ static char *_build_desc_write(struct cmd_context *cmd, struct volume_group *vg)
 static int _vg_write_raw(struct format_instance *fid, struct volume_group *vg,
 			 struct metadata_area *mda)
 {
+	char desc[2048];
 	struct mda_context *mdac = (struct mda_context *) mda->metadata_locn;
 	struct text_fid_context *fidtc = (struct text_fid_context *) fid->private;
 	struct raw_locn *rlocn_old;
@@ -673,12 +650,15 @@ static int _vg_write_raw(struct format_instance *fid, struct volume_group *vg,
 		write_buf_size = fidtc->write_buf_size;
 		new_size = fidtc->new_metadata_size;
 	} else {
-		char *desc = _build_desc_write(fid->fmt->cmd, vg);
+		if (!vg->write_count++)
+			(void) dm_snprintf(desc, sizeof(desc), "Write from %s.", vg->cmd->cmd_line);
+		else
+			(void) dm_snprintf(desc, sizeof(desc), "Write[%u] from %s.", vg->write_count, vg->cmd->cmd_line);
+
 		new_size = text_vg_export_raw(vg, desc, &write_buf, &write_buf_size);
 		fidtc->write_buf = write_buf;
 		fidtc->write_buf_size = write_buf_size;
 		fidtc->new_metadata_size = new_size;
-		free(desc);
 	}
 
 	if (!new_size || !write_buf) {
@@ -989,7 +969,6 @@ static int _vg_write_raw(struct format_instance *fid, struct volume_group *vg,
 
 	if (!dev_write_bytes(mdac->area.dev, write1_start, (size_t)write1_size, write_buf)) {
 		log_error("Failed to write metadata to %s fd %d", devname, mdac->area.dev->bcache_fd);
-		dev_unset_last_byte(mdac->area.dev);
 		goto out;
 	}
 
@@ -1002,7 +981,6 @@ static int _vg_write_raw(struct format_instance *fid, struct volume_group *vg,
 		if (!dev_write_bytes(mdac->area.dev, write2_start, write2_size,
 				     write_buf + new_size - new_wrap)) {
 			log_error("Failed to write metadata wrap to %s fd %d", devname, mdac->area.dev->bcache_fd);
-			dev_unset_last_byte(mdac->area.dev);
 			goto out;
 		}
 	}

lib/format_text/text_label.c

@@ -332,14 +332,12 @@ static int _read_mda_header_and_metadata(const struct format_type *fmt,
 		log_warn("WARNING: bad metadata header on %s at %llu.",
 			 dev_name(mdac->area.dev),
 			 (unsigned long long)mdac->area.start);
-		if (mda)
-			mda->header_start = mdac->area.start;
+		mda->header_start = mdac->area.start;
 		*bad_fields |= BAD_MDA_HEADER;
 		return 0;
 	}
 
-	if (mda)
-		mda->header_start = mdah->start;
+	mda->header_start = mdah->start;
 
 	mda_set_ignored(mda, rlocn_is_ignored(mdah->raw_locns));
 

lib/integrity/integrity.c

@@ -0,0 +1,328 @@
+/*
+ * 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)
+		meta_lv->status |= INTEGRITY_METADATA;
+
+	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/label/hints.c

@@ -638,7 +638,8 @@ static int _read_hint_file(struct cmd_context *cmd, struct dm_list *hints, int *
 	char devpath[PATH_MAX];
 	FILE *fp;
 	struct dev_iter *iter;
-	struct hint *hint;
+	struct hint hint;
+	struct hint *alloc_hint;
 	struct device *dev;
 	char *split[HINT_LINE_WORDS];
 	char *name, *pvid, *devn, *vgname, *p, *filter_str = NULL;
@@ -662,11 +663,7 @@ static int _read_hint_file(struct cmd_context *cmd, struct dm_list *hints, int *
 		split[i] = NULL;
 
 	while (fgets(_hint_line, sizeof(_hint_line), fp)) {
-		if (!(hint = zalloc(sizeof(struct hint)))) {
-			ret = 0;
-			break;
-		}
-
+		memset(&hint, 0, sizeof(hint));
 		if (_hint_line[0] == '#')
 			continue;
 
@@ -704,13 +701,11 @@ static int _read_hint_file(struct cmd_context *cmd, struct dm_list *hints, int *
 			_filter_to_str(cmd, devices_global_filter_CFG, &filter_str);
 			if (!filter_str || strcmp(filter_str, _hint_line + keylen)) {
 				log_debug("ignore hints with different global_filter");
-				if (filter_str)
-					free(filter_str);
+				free(filter_str);
 				*needs_refresh = 1;
 				break;
 			}
-			if (filter_str)
-				free(filter_str);
+			free(filter_str);
 			continue;
 		}
 
@@ -719,23 +714,20 @@ static int _read_hint_file(struct cmd_context *cmd, struct dm_list *hints, int *
 			_filter_to_str(cmd, devices_filter_CFG, &filter_str);
 			if (!filter_str || strcmp(filter_str, _hint_line + keylen)) {
 				log_debug("ignore hints with different filter");
-				if (filter_str)
-					free(filter_str);
+				free(filter_str);
 				*needs_refresh = 1;
 				break;
 			}
-			if (filter_str)
-				free(filter_str);
+			free(filter_str);
 			continue;
 		}
 
 		keylen = strlen("scan_lvs:");
 		if (!strncmp(_hint_line, "scan_lvs:", keylen)) {
 			int scan_lvs = 0;
-			sscanf(_hint_line + keylen, "%u", &scan_lvs);
-
-			if (scan_lvs != cmd->scan_lvs) {
-				log_debug("ignore hints with different scan_lvs");
+			if ((sscanf(_hint_line + keylen, "%u", &scan_lvs) != 1) ||
+			    scan_lvs != cmd->scan_lvs) {
+				log_debug("ignore hints with different or unreadable scan_lvs");
 				*needs_refresh = 1;
 				break;
 			}
@@ -744,7 +736,11 @@ static int _read_hint_file(struct cmd_context *cmd, struct dm_list *hints, int *
 
 		keylen = strlen("devs_hash:");
 		if (!strncmp(_hint_line, "devs_hash:", keylen)) {
-			sscanf(_hint_line + keylen, "%u %u", &read_hash, &read_count);
+			if (sscanf(_hint_line + keylen, "%u %u", &read_hash, &read_count) != 2) {
+				log_debug("ignore hints with invalid devs_hash");
+				*needs_refresh = 1;
+				break;
+			}
 			continue;
 		}
 
@@ -764,19 +760,28 @@ static int _read_hint_file(struct cmd_context *cmd, struct dm_list *hints, int *
 		vgname = split[3];
 
 		if (name && !strncmp(name, "scan:", 5))
-			strncpy(hint->name, name+5, PATH_MAX);
+			if (!dm_strncpy(hint.name, name + 5, sizeof(hint.name)))
+				continue;
 
 		if (pvid && !strncmp(pvid, "pvid:", 5))
-			strncpy(hint->pvid, pvid+5, ID_LEN);
+			if (!dm_strncpy(hint.pvid, pvid + 5, sizeof(hint.pvid)))
+				continue;
 
 		if (devn && sscanf(devn, "devn:%d:%d", &major, &minor) == 2)
-			hint->devt = makedev(major, minor);
+			hint.devt = makedev(major, minor);
 
 		if (vgname && (strlen(vgname) > 3) && (vgname[4] != '-'))
-			strncpy(hint->vgname, vgname+3, NAME_LEN);
+			if (!dm_strncpy(hint.vgname, vgname + 3, sizeof(hint.vgname)))
+				continue;
 
-		log_debug("add hint %s %s %d:%d %s", hint->name, hint->pvid, major, minor, vgname);
-		dm_list_add(hints, &hint->list);
+		if (!(alloc_hint = malloc(sizeof(struct hint)))) {
+			ret = 0;
+			break;
+		}
+		memcpy(alloc_hint, &hint, sizeof(hint));
+
+		log_debug("add hint %s %s %d:%d %s", hint.name, hint.pvid, major, minor, vgname);
+		dm_list_add(hints, &alloc_hint->list);
 		found++;
 	}
 
@@ -911,13 +916,11 @@ int write_hint_file(struct cmd_context *cmd, int newhints)
 
 	_filter_to_str(cmd, devices_global_filter_CFG, &filter_str);
 	fprintf(fp, "global_filter:%s\n", filter_str ?: "-");
-	if (filter_str)
-		free(filter_str);
+	free(filter_str);
 
 	_filter_to_str(cmd, devices_filter_CFG, &filter_str);
 	fprintf(fp, "filter:%s\n", filter_str ?: "-");
-	if (filter_str)
-		free(filter_str);
+	free(filter_str);
 
 	fprintf(fp, "scan_lvs:%d\n", cmd->scan_lvs);
 
@@ -951,7 +954,6 @@ int write_hint_file(struct cmd_context *cmd, int newhints)
 		 * detect when the devices on the system change, which
 		 * invalidates the existing hints.
 		 */
-		memset(devpath, 0, sizeof(devpath));
 		strncpy(devpath, dev_name(dev), PATH_MAX);
 		hash = calc_crc(hash, (const uint8_t *)devpath, strlen(devpath));
 		count++;
@@ -1179,7 +1181,8 @@ static void _get_single_vgname_cmd_arg(struct cmd_context *cmd,
 
 	if (!(st = strchr(arg, '/'))) {
 		/* simple vgname */
-		name = strdup(arg);
+		if (!(name = strdup(arg)))
+			return;
 		goto check;
 	}
 
@@ -1187,7 +1190,8 @@ static void _get_single_vgname_cmd_arg(struct cmd_context *cmd,
 	for (p = arg; p < st; p++)
 		namebuf[i++] = *p;
 
-	name = strdup(namebuf);
+	if (!(name = strdup(namebuf)))
+		return;
 
 check:
 	/*
@@ -1202,6 +1206,8 @@ check:
 			return;
 		}
 	}
+
+	free(name);
 }
 
 /*
@@ -1261,8 +1267,9 @@ int get_hints(struct cmd_context *cmd, struct dm_list *hints_out, int *newhints,
 	 */
 	if (_newhints_exists()) {
 		log_debug("get_hints: newhints file");
-		if (!_hints_exists())
-			_touch_hints();
+		if (!_hints_exists() && !_touch_hints())
+			return 0;
+
 		if (!_lock_hints(cmd, LOCK_EX, NONBLOCK))
 			return 0;
 		/* create new hints after scan */
@@ -1355,6 +1362,9 @@ int get_hints(struct cmd_context *cmd, struct dm_list *hints_out, int *newhints,
 		  dm_list_size(devs_out), dm_list_size(devs_in));
 
 	dm_list_splice(hints_out, &hints_list);
+
+	free(vgname);
+
 	return 1;
 }
 

lib/label/label.c

@@ -218,7 +218,7 @@ int label_write(struct device *dev, struct label *label)
 
 	if (!dev_write_bytes(dev, offset, LABEL_SIZE, buf)) {
 		log_debug_devs("Failed to write label to %s", dev_name(dev));
-		r = 0;
+		return 0;
 	}
 
 	dev_unset_last_byte(dev);
@@ -252,8 +252,6 @@ struct label *label_create(struct labeller *labeller)
 /* global variable for accessing the bcache populated by label scan */
 struct bcache *scan_bcache;
 
-#define BCACHE_BLOCK_SIZE_IN_SECTORS 256 /* 256*512 = 128K */
-
 static bool _in_bcache(struct device *dev)
 {
 	if (!dev)
@@ -655,7 +653,6 @@ static int _scan_list(struct cmd_context *cmd, struct dev_filter *f,
 	int submit_count;
 	int scan_failed;
 	int is_lvm_device;
-	int error;
 	int ret;
 
 	dm_list_init(&wait_devs);
@@ -702,12 +699,11 @@ static int _scan_list(struct cmd_context *cmd, struct dev_filter *f,
 
 	dm_list_iterate_items_safe(devl, devl2, &wait_devs) {
 		bb = NULL;
-		error = 0;
 		scan_failed = 0;
 		is_lvm_device = 0;
 
 		if (!bcache_get(scan_bcache, devl->dev->bcache_fd, 0, 0, &bb)) {
-			log_debug_devs("Scan failed to read %s error %d.", dev_name(devl->dev), error);
+			log_debug_devs("Scan failed to read %s.", dev_name(devl->dev));
 			scan_failed = 1;
 			scan_read_errors++;
 			scan_failed_count++;
@@ -1451,6 +1447,7 @@ bool dev_write_bytes(struct device *dev, uint64_t start, size_t len, void *data)
 	if (!bcache_write_bytes(scan_bcache, dev->bcache_fd, start, len, data)) {
 		log_error("Error writing device %s at %llu length %u.",
 			  dev_name(dev), (unsigned long long)start, (uint32_t)len);
+		dev_unset_last_byte(dev);
 		label_scan_invalidate(dev);
 		return false;
 	}
@@ -1458,6 +1455,7 @@ bool dev_write_bytes(struct device *dev, uint64_t start, size_t len, void *data)
 	if (!bcache_flush(scan_bcache)) {
 		log_error("Error writing device %s at %llu length %u.",
 			  dev_name(dev), (unsigned long long)start, (uint32_t)len);
+		dev_unset_last_byte(dev);
 		label_scan_invalidate(dev);
 		return false;
 	}

lib/label/label.h

@@ -118,6 +118,14 @@ int label_scan_open(struct device *dev);
 int label_scan_open_excl(struct device *dev);
 int label_scan_open_rw(struct device *dev);
 
+/*
+ * These are the sizes the label.c uses to set up
+ * and use bcache (they are not bcache restrictions
+ * or defs.)
+ */
+#define BCACHE_BLOCK_SIZE_IN_SECTORS 256 /* 256*512 = 128K */
+#define BCACHE_BLOCK_SIZE_IN_BYTES 131072
+
 /*
  * Wrappers around bcache equivalents.
  * (these make it easier to disable bcache and revert to direct rw if needed)

lib/metadata/cache_manip.c

@@ -23,7 +23,6 @@
 #include "lib/config/defaults.h"
 #include "lib/metadata/lv_alloc.h"
 #include "lib/misc/lvm-signal.h"
-#include "lib/activate/dev_manager.h"
 
 /* https://github.com/jthornber/thin-provisioning-tools/blob/master/caching/cache_metadata_size.cc */
 #define DM_TRANSACTION_OVERHEAD		4096  /* KiB */

lib/metadata/integrity_manip.c

@@ -0,0 +1,662 @@
+/*
+ * 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
+
+int lv_is_integrity_origin(const struct logical_volume *lv)
+{
+	struct seg_list *sl;
+
+	dm_list_iterate_items(sl, &lv->segs_using_this_lv) {
+		if (!sl->seg || !sl->seg->lv || !sl->seg->origin)
+			continue;
+		if (lv_is_integrity(sl->seg->lv) && (sl->seg->origin == lv))
+			return 1;
+	}
+	return 0;
+}
+
+/*
+ * Every 500M of data needs 4M of metadata.
+ * (From trial and error testing.)
+ */
+static uint64_t _lv_size_bytes_to_integrity_meta_bytes(uint64_t lv_size_bytes)
+{
+	return ((lv_size_bytes / (500 * ONE_MB_IN_BYTES)) + 1) * (4 * ONE_MB_IN_BYTES);
+}
+
+/*
+ * 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];
+	uint64_t lv_size_bytes, meta_bytes, meta_sectors;
+	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,
+		.suppress_zero_warn = 1,
+	};
+
+	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;
+
+	lp_meta.pvh = lp->pvh;
+
+	lv_size_bytes = lp->extents * vg->extent_size * 512;
+	meta_bytes = _lv_size_bytes_to_integrity_meta_bytes(lv_size_bytes);
+	meta_sectors = meta_bytes / 512;
+	lp_meta.extents = meta_sectors / vg->extent_size;
+
+	log_print("Creating integrity metadata LV %s with size %s.",
+		  metaname, display_size(cmd, meta_sectors));
+
+	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;
+	uint64_t data_sectors, extra_sectors;
+
+	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;
+	}
+
+	data_sectors = status.seg_status.integrity->provided_data_sectors;
+
+	if ((extra_sectors = (data_sectors % lv->vg->extent_size))) {
+		data_sectors -= extra_sectors;
+		log_debug("Reduce provided_data_sectors by %llu to %llu for extent alignment",
+			  (unsigned long long)extra_sectors, (unsigned long long)data_sectors);
+	}
+
+	*provided_data_sectors = data_sectors;
+
+	dm_pool_destroy(status.seg_status.mem);
+	return 1;
+
+fail:
+	dm_pool_destroy(status.seg_status.mem);
+	return 0;
+}
+
+int lv_remove_integrity(struct logical_volume *lv)
+{
+	struct lv_segment *seg = first_seg(lv);
+	struct logical_volume *origin;
+	struct logical_volume *meta_lv;
+
+	if (!seg_is_integrity(seg)) {
+		log_error("LV %s segment is not integrity.", display_lvname(lv));
+		return 0;
+	}
+
+	if (!(meta_lv = seg->integrity_meta_dev)) {
+		log_error("LV %s segment has no integrity metadata device.", display_lvname(lv));
+		return 0;
+	}
+
+	if (!(origin = seg_lv(seg, 0))) {
+		log_error("LV %s integrity segment has no origin", display_lvname(lv));
+		return 0;
+	}
+
+	if (!remove_seg_from_segs_using_this_lv(seg->integrity_meta_dev, seg))
+		return_0;
+
+	lv_set_visible(seg->integrity_meta_dev);
+
+	lv->status &= ~INTEGRITY;
+	meta_lv->status &= ~INTEGRITY_METADATA;
+
+	seg->integrity_meta_dev = NULL;
+
+	if (!remove_layer_from_lv(lv, origin))
+		return_0;
+
+	if (!lv_remove(origin))
+		return_0;
+
+	if (!lv_remove(meta_lv))
+		log_warn("WARNING: failed to remove integrity metadata LV.");
+
+	return 1;
+}
+
+/*
+ * Add integrity to each raid image.
+ *
+ * for each rimage_N:
+ * . create and allocate a new linear LV rimage_N_imeta
+ * . move the segments from rimage_N to a new rimage_N_iorig
+ * . add an integrity segment to rimage_N with
+ *   origin=rimage_N_iorig, meta_dev=rimage_N_imeta
+ *
+ * Before:
+ * rimage_0
+ *   segment1: striped: pv0:A
+ * rimage_1
+ *   segment1: striped: pv1:B
+ *
+ * After:
+ * rimage_0
+ *   segment1: integrity: rimage_0_iorig, rimage_0_imeta
+ * rimage_1
+ *   segment1: integrity: rimage_1_iorig, rimage_1_imeta
+ * rimage_0_iorig
+ *   segment1: striped: pv0:A
+ * rimage_1_iorig
+ *   segment1: striped: pv1:B
+ * rimage_0_imeta
+ *   segment1: striped: pv2:A
+ * rimage_1_imeta
+ *   segment1: striped: pv2:B
+ *    
+ */
+
+static int _lv_add_integrity_to_raid(struct logical_volume *lv, const char *arg,
+				     struct integrity_settings *settings,
+				     struct dm_list *pvh,
+				     uint64_t *zero_data_sectors)
+{
+	struct lvcreate_params lp;
+	struct logical_volume *imeta_lvs[DEFAULT_RAID_MAX_IMAGES];
+	struct cmd_context *cmd = lv->vg->cmd;
+	struct volume_group *vg = lv->vg;
+	struct logical_volume *lv_image, *lv_imeta, *lv_iorig;
+	struct lv_segment *seg_top, *seg_image;
+	const struct segment_type *segtype;
+	struct integrity_settings *set;
+	uint64_t lv_image_size; /* sectors */
+	uint64_t status_data_sectors;
+	uint32_t area_count, s;
+	int internal_metadata;
+	int ret = 1;
+
+	memset(imeta_lvs, 0, sizeof(imeta_lvs));
+
+	if (dm_list_size(&lv->segments) != 1)
+		return_0;
+
+	seg_top = first_seg(lv);
+	area_count = seg_top->area_count;
+
+	if ((internal_metadata = !strcmp(arg, "internal"))) {
+		/*
+		 * FIXME: raid on internal integrity might not be used widely
+		 * enough to enable, given the additional complexity/support.
+		 * i.e. nearly everyone may just use external metadata.
+		 *
+		 * FIXME: _info_run() needs code to adjust the length, like
+		 * is done for if (lv_is_integrity()) length = ...
+		 */
+		/* goto skip_imeta; */
+		log_error("Internal integrity metadata is not yet supported with raid.");
+		return 0;
+	}
+
+	/*
+	 * For each rimage, create an _imeta LV for integrity metadata.
+	 * Each needs to be zeroed.
+	 */
+	for (s = 0; s < area_count; s++) {
+		struct logical_volume *meta_lv;
+		struct wipe_params wipe;
+
+		if (s >= DEFAULT_RAID_MAX_IMAGES)
+			return_0;
+
+		lv_image = seg_lv(seg_top, s);
+
+		if (!seg_is_striped(first_seg(lv_image))) {
+			log_error("raid1 image must be linear to add integrity");
+			return_0;
+		}
+
+		/*
+		 * allocate a new linear LV NAME_rimage_N_imeta
+		 * lv_create_integrity_metadata() returns its result in lp
+		 */
+		memset(&lp, 0, sizeof(lp));
+		lp.lv_name = lv_image->name;
+		lp.pvh = pvh;
+		lp.extents = lv_image->size / vg->extent_size;
+
+		if (!lv_create_integrity_metadata(cmd, vg, &lp)) {
+			return_0;
+		}
+		meta_lv = lp.integrity_meta_lv;
+
+		/*
+		 * dm-integrity requires the metadata LV header to be zeroed.
+		 */
+
+		if (!activate_lv(cmd, meta_lv)) {
+			log_error("Failed to activate LV %s to zero", display_lvname(meta_lv));
+			return 0;
+		}
+
+		memset(&wipe, 0, sizeof(wipe));
+		wipe.do_zero = 1;
+		wipe.zero_sectors = 8;
+
+		if (!wipe_lv(meta_lv, wipe)) {
+			log_error("Failed to zero LV for integrity metadata %s", display_lvname(meta_lv));
+			return 0;
+		}
+
+		if (!deactivate_lv(cmd, meta_lv)) {
+			log_error("Failed to deactivate LV %s after zero", display_lvname(meta_lv));
+			return 0;
+		}
+
+		/* Used below to set up the new integrity segment. */
+		imeta_lvs[s] = meta_lv;
+	}
+
+/* skip_imeta: */
+
+	/*
+	 * For each rimage, move its segments to a new rimage_iorig and give
+	 * the rimage a new integrity segment.
+	 */
+	for (s = 0; s < area_count; s++) {
+		lv_image = seg_lv(seg_top, s);
+
+		lv_image_size = lv_image->size; /* sectors */
+
+		if (!(segtype = get_segtype_from_string(cmd, SEG_TYPE_NAME_INTEGRITY)))
+			return_0;
+
+		log_debug("Adding integrity to raid image %s", lv_image->name);
+
+		/*
+		 * "lv_iorig" is a new LV with new id, but with the segments
+		 * from "lv_image". "lv_image" keeps the existing name and id,
+		 * but gets a new integrity segment, in place of the segments
+		 * that were moved to lv_iorig.
+		 */
+		if (!(lv_iorig = insert_layer_for_lv(cmd, lv_image, INTEGRITY, "_iorig")))
+			return_0;
+
+		lv_image->status |= INTEGRITY;
+
+		/*
+		 * Set up the new first segment of lv_image as integrity.
+		 */
+		seg_image = first_seg(lv_image);
+		seg_image->segtype = segtype;
+
+		if (internal_metadata) {
+			/* dm-integrity wants temp/fake size of 1 to report usable size */
+			lv_image->size = 1;
+			seg_image->integrity_data_sectors = 1;
+		} else {
+			lv_imeta = imeta_lvs[s]; /* external metadata lv created above */
+			lv_imeta->status |= INTEGRITY_METADATA;
+			lv_set_hidden(lv_imeta);
+			seg_image->integrity_data_sectors = lv_image_size;
+			seg_image->integrity_meta_dev = lv_imeta;
+		}
+
+		memcpy(&seg_image->integrity_settings, settings, sizeof(struct integrity_settings));
+		set = &seg_image->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;
+
+		set->recalculate = 1; /* enable integrity initialization in the kernel */
+	}
+
+	if (internal_metadata) {
+		/* Get the size from the first image, the others will be the same. */
+		lv_image = seg_lv(seg_top, 0);
+
+		lv_image->status |= LV_TEMPORARY;
+		lv_image->status |= LV_NOSCAN;
+		lv_image->status |= LV_UNCOMMITTED;
+
+		log_debug("Activating temporary integrity LV to get data sectors.");
+
+		if (!activate_lv(cmd, lv_image)) {
+			log_error("Failed to activate temporary integrity.");
+			ret = 0;
+			goto out;
+		}
+
+		if (!_get_provided_data_sectors(lv_image, &status_data_sectors)) {
+			log_error("Failed to get data sectors from dm-integrity");
+			ret = 0;
+		} else {
+			log_print("Found integrity provided_data_sectors %llu", (unsigned long long)status_data_sectors);
+			ret = 1;
+		}
+
+		if (!status_data_sectors) {
+			log_error("LV size too small to include metadata.");
+			ret = 0;
+		}
+
+		lv_image->status |= LV_UNCOMMITTED;
+
+		if (!deactivate_lv(cmd, lv_image)) {
+			log_error("Failed to deactivate temporary integrity.");
+			ret = 0;
+		}
+
+		if (!ret)
+			goto_out;
+
+		lv_image->status &= ~LV_UNCOMMITTED;
+		lv_image->status &= ~LV_NOSCAN;
+		lv_image->status &= ~LV_TEMPORARY;
+
+		for (s = 0; s < area_count; s++) {
+			lv_image = seg_lv(seg_top, s);
+			lv_image->size = lv_image_size; /* was 1 temporarily */
+			seg_image = first_seg(lv_image);
+			seg_image->integrity_data_sectors = status_data_sectors;
+		}
+	}
+
+	/*
+	 * recalculate above enables automatic integrity metadata intialization
+	 * in the kernel when first activated, instead of activating the raid
+	 * LV and writing zeroes to the entire thing from the lvcreate command
+	 * to initialize the metadata in the integrity images.
+	 *
+	 * Zeroing the raid LV to intialize integrity metadata on the images
+	 * does not work because dm-raid immediately reads the integrity
+	 * images when activated, and gets errors back because of no init.
+	 *
+	 * When activated with recalculate, dm-integrity will not return errors
+	 * to dm-raid when it reads uninitialized blocks.
+	 */
+	*zero_data_sectors = 0; /* disables intialization by lvcreate */
+
+	log_debug("Write VG with integrity added to LV");
+
+	if (!vg_write(vg) || !vg_commit(vg))
+		ret = 0;
+out:
+	return ret;
+}
+
+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 dm_list *pvh,
+		     uint64_t *zero_data_sectors)
+{
+	struct cmd_context *cmd = lv->vg->cmd;
+	struct volume_group *vg = lv->vg;
+	struct integrity_settings *set;
+	struct logical_volume *lv_orig;
+	struct logical_volume *meta_lv = NULL;
+	const struct segment_type *segtype;
+	struct lv_segment *seg;
+	uint64_t meta_bytes, meta_sectors;
+	uint64_t lv_size_sectors;
+	int ret = 1;
+
+	if (lv_is_raid(lv))
+		return _lv_add_integrity_to_raid(lv, arg, settings, pvh, zero_data_sectors);
+
+	lv_size_sectors = lv->size;
+
+	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(vg, meta_name))) {
+			log_error("LV %s not found.", meta_name);
+			return_0;
+		}
+
+		meta_bytes = _lv_size_bytes_to_integrity_meta_bytes(lv->size * 512);
+		meta_sectors = meta_bytes / 512;
+
+		if (meta_lv->size < meta_sectors) {
+			log_error("Integrity metadata needs %s, metadata LV is only %s.",
+				 display_size(cmd, meta_sectors), display_size(cmd, meta_lv->size));
+			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) {
+		struct wipe_params wipe;
+
+		if (!sync_local_dev_names(cmd)) {
+			log_error("Failed to sync local devices.");
+			return 0;
+		}
+
+		log_verbose("Zeroing LV for integrity metadata");
+
+		if (!lv_is_active(meta_lv)) {
+			if (!activate_lv(cmd, meta_lv)) {
+				log_error("Failed to activate LV %s to zero", display_lvname(meta_lv));
+				return 0;
+			}
+		}
+
+		memset(&wipe, 0, sizeof(wipe));
+		wipe.do_zero = 1;
+		wipe.zero_sectors = 8;
+
+		if (!wipe_lv(meta_lv, wipe)) {
+			log_error("Failed to zero LV for integrity metadata %s", display_lvname(meta_lv));
+			return 0;
+		}
+
+		if (!deactivate_lv(cmd, meta_lv)) {
+			log_error("Failed to deactivate LV %s after zero", display_lvname(meta_lv));
+			return 0;
+		}
+
+		meta_lv->status |= INTEGRITY_METADATA;
+		seg->integrity_data_sectors = lv_size_sectors;
+		seg->integrity_meta_dev = meta_lv;
+		lv_set_hidden(meta_lv);
+		/* TODO: give meta_lv a suffix? e.g. _imeta */
+	} 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;
+		}
+
+		if (!seg->integrity_data_sectors) {
+			log_error("LV size too small to include metadata.");
+			ret = 0;
+		}
+
+		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;
+	}
+
+	/*
+	 * Tells lvcreate to zero the final LV at the end of commands.
+	 */
+	*zero_data_sectors = seg->integrity_data_sectors;
+
+	log_debug("Write VG with integrity added to LV");
+
+	if (!vg_write(vg) || !vg_commit(vg))
+		ret = 0;
+ 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;
 }
@@ -1200,10 +1202,13 @@ char *lv_attr_dup_with_info_and_seg_status(struct dm_pool *mem, const struct lv_
 		repstr[0] = (lv_is_merging_origin(lv)) ? 'O' : 'o';
 	else if (lv_is_pool_metadata(lv) ||
 		 lv_is_pool_metadata_spare(lv) ||
-		 lv_is_raid_metadata(lv))
+		 lv_is_raid_metadata(lv) ||
+		 lv_is_integrity_metadata(lv))
 		repstr[0] = 'e';
 	else if (lv_is_cache_type(lv) || lv_is_writecache(lv))
 		repstr[0] = 'C';
+	else if (lv_is_integrity(lv))
+		repstr[0] = 'g';
 	else if (lv_is_raid(lv))
 		repstr[0] = (lv_is_not_synced(lv)) ? 'R' : 'r';
 	else if (lv_is_mirror(lv))

lib/metadata/lv_manip.c

@@ -28,6 +28,7 @@
 #include "lib/datastruct/str_list.h"
 #include "lib/config/defaults.h"
 #include "lib/misc/lvm-exec.h"
+#include "lib/misc/lvm-signal.h"
 #include "lib/mm/memlock.h"
 #include "lib/locking/lvmlockd.h"
 #include "lib/label/label.h"
@@ -55,6 +56,8 @@ typedef enum {
 #define A_POSITIONAL_FILL	0x40	/* Slots are positional and filled using PREFERRED */
 #define A_PARTITION_BY_TAGS	0x80	/* No allocated area may share any tag with any other */
 
+#define ONE_MB_IN_BYTES 1048576
+
 /*
  * Constant parameters during a single allocation attempt.
  */
@@ -134,7 +137,9 @@ enum {
 	LV_TYPE_SANLOCK,
 	LV_TYPE_CACHEVOL,
 	LV_TYPE_WRITECACHE,
-	LV_TYPE_WRITECACHEORIGIN
+	LV_TYPE_WRITECACHEORIGIN,
+	LV_TYPE_INTEGRITY,
+	LV_TYPE_INTEGRITYORIGIN
 };
 
 static const char *_lv_type_names[] = {
@@ -190,6 +195,8 @@ static const char *_lv_type_names[] = {
 	[LV_TYPE_CACHEVOL] =				"cachevol",
 	[LV_TYPE_WRITECACHE] =				"writecache",
 	[LV_TYPE_WRITECACHEORIGIN] =			"writecacheorigin",
+	[LV_TYPE_INTEGRITY] =				"integrity",
+	[LV_TYPE_INTEGRITYORIGIN] =			"integrityorigin",
 };
 
 static int _lv_layout_and_role_mirror(struct dm_pool *mem,
@@ -461,6 +468,43 @@ bad:
 	return 0;
 }
 
+static int _lv_layout_and_role_integrity(struct dm_pool *mem,
+				     const struct logical_volume *lv,
+				     struct dm_list *layout,
+				     struct dm_list *role,
+				     int *public_lv)
+{
+	int top_level = 0;
+
+	/* non-top-level LVs */
+	if (lv_is_integrity_metadata(lv)) {
+		if (!str_list_add_no_dup_check(mem, role, _lv_type_names[LV_TYPE_INTEGRITY]) ||
+		    !str_list_add_no_dup_check(mem, role, _lv_type_names[LV_TYPE_METADATA]))
+			goto_bad;
+	} else if (lv_is_integrity_origin(lv)) {
+		if (!str_list_add_no_dup_check(mem, role, _lv_type_names[LV_TYPE_INTEGRITY]) ||
+		    !str_list_add_no_dup_check(mem, role, _lv_type_names[LV_TYPE_ORIGIN]) ||
+		    !str_list_add_no_dup_check(mem, role, _lv_type_names[LV_TYPE_INTEGRITYORIGIN]))
+			goto_bad;
+	} else
+		top_level = 1;
+
+	if (!top_level) {
+		*public_lv = 0;
+		return 1;
+	}
+
+	/* top-level LVs */
+	if (lv_is_integrity(lv)) {
+		if (!str_list_add_no_dup_check(mem, layout, _lv_type_names[LV_TYPE_INTEGRITY]))
+			goto_bad;
+	}
+
+	return 1;
+bad:
+	return 0;
+}
+
 static int _lv_layout_and_role_thick_origin_snapshot(struct dm_pool *mem,
 						     const struct logical_volume *lv,
 						     struct dm_list *layout,
@@ -577,6 +621,11 @@ int lv_layout_and_role(struct dm_pool *mem, const struct logical_volume *lv,
 	    !_lv_layout_and_role_cache(mem, lv, *layout, *role, &public_lv))
 		goto_bad;
 
+	/* Integrity related */
+	if ((lv_is_integrity(lv) || lv_is_integrity_origin(lv) || lv_is_integrity_metadata(lv)) &&
+	    !_lv_layout_and_role_integrity(mem, lv, *layout, *role, &public_lv))
+		goto_bad;
+
 	/* VDO and related */
 	if (lv_is_vdo_type(lv) &&
 	    !_lv_layout_and_role_vdo(mem, lv, *layout, *role, &public_lv))
@@ -1457,6 +1506,15 @@ 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 +5712,11 @@ 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 +7473,90 @@ int insert_layer_for_segments_on_pv(struct cmd_context *cmd,
 	return 1;
 }
 
+int zero_lv_name(struct cmd_context *cmd, const char *vg_name, const char *lv_name, uint64_t lv_size_sectors)
+{
+	char name[PATH_MAX];
+	struct device *dev;
+	uint64_t off = 0, i = 0, j = 0;
+	uint64_t lv_size_bytes = lv_size_sectors * 512;
+	uint64_t zero_bytes;
+	uint32_t extra_bytes;
+
+	if (dm_snprintf(name, sizeof(name), "%s%s/%s", cmd->dev_dir, vg_name, lv_name) < 0) {
+		log_error("Device path name too long, device not zeroed (%s).", lv_name);
+		return 0;
+	}
+
+	if (!(dev = dev_cache_get(cmd, name, NULL))) {
+		log_error("Failed to find device %s: device not zeroed.", name);
+		return 0;
+	}
+
+	if (!label_scan_open_rw(dev)) {
+		log_error("Failed to open %s: device not zeroed.", name);
+		return 0;
+	}
+
+	zero_bytes = lv_size_bytes;
+
+	log_print("Zeroing %s %s... (cancel command to zero manually)",
+		  name, display_size(cmd, zero_bytes/512));
+
+	if ((extra_bytes = (zero_bytes % ONE_MB_IN_BYTES)))
+		zero_bytes -= extra_bytes;
+
+	/*
+	 * Write 1MiB at a time to avoid going over bcache size.
+	 * Then write 128KiB (bcache block sizes) at a time to
+	 * cover remaining dev size.
+	 */
+
+	sigint_allow();
+
+	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 (off && !(off % (1024 * ONE_MB_IN_BYTES)))
+			log_print("Zeroed %s...", display_size(cmd, off/512));
+
+		if (sigint_caught()) {
+			log_print("Zeroing canceled.");
+			goto out;
+		}
+
+	}
+
+	if (extra_bytes) {
+		log_debug("Zeroing final %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);
+
+			if (sigint_caught()) {
+				log_print("Zeroing canceled.");
+				goto out;
+			}
+		}
+	}
+
+	/*
+	 * FIXME: bcache can't write partial blocks yet.
+	 * This shouldn't actually happen given current
+	 * usage where LV size is a multiple of extents.
+	 */
+	if ((extra_bytes = lv_size_bytes - (i * ONE_MB_IN_BYTES + j * BCACHE_BLOCK_SIZE_IN_BYTES)))
+		log_warn("WARNING: last %llu bytes not zeroed.", (unsigned long long)extra_bytes);
+
+out:
+	sigint_restore();
+	label_scan_invalidate(dev);
+	return 1;
+}
+
 /*
  * Initialize the LV with 'value'.
  */
@@ -7679,6 +7826,15 @@ static int _should_wipe_lv(struct lvcreate_params *lp,
 	     first_seg(first_seg(lv)->pool_lv)->zero_new_blocks))
 		return 0;
 
+	/*
+	 * dm-integrity requires the first sector (integrity superblock) to be
+	 * zero when creating a new integrity device.
+	 * TODO: print a warning or error if the user specifically
+	 * asks for no wiping or zeroing?
+	 */
+	if (seg_is_integrity(lp) || (seg_is_raid1(lp) && lp->integrity_arg))
+		return 1;
+
 	/* Cannot zero read-only volume */
 	if ((lv->status & LVM_WRITE) &&
 	    (lp->zero || lp->wipe_signatures))
@@ -7934,6 +8090,11 @@ 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)) {
+		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,
@@ -8277,6 +8438,19 @@ static struct logical_volume *_lv_create_an_lv(struct volume_group *vg,
 			goto revert_new_lv;
 		}
 		lv->status &= ~LV_TEMPORARY;
+
+	} else if (seg_is_integrity(lp) || (seg_is_raid1(lp) && lp->integrity_arg)) {
+		/*
+		 * Activate the new origin LV so it can be zeroed/wiped
+		 * below before adding integrity.
+		 */
+		lv->status |= LV_TEMPORARY;
+		if (!activate_lv(cmd, lv)) {
+			log_error("Failed to activate new LV for wiping.");
+			goto revert_new_lv;
+		}
+		lv->status &= ~LV_TEMPORARY;
+
 	} else if (!lv_active_change(cmd, lv, lp->activate)) {
 		log_error("Failed to activate new LV %s.", display_lvname(lv));
 		goto deactivate_and_revert_new_lv;
@@ -8296,6 +8470,53 @@ static struct logical_volume *_lv_create_an_lv(struct volume_group *vg,
 		}
 	}
 
+	if (seg_is_integrity(lp) || (seg_is_raid1(lp) && lp->integrity_arg)) {
+		log_verbose("Adding integrity to new LV");
+
+		/* Origin is active from zeroing, deactivate to add integrity. */
+
+		if (!deactivate_lv(cmd, lv)) {
+			log_error("Failed to deactivate LV to add integrity");
+			goto revert_new_lv;
+		}
+
+		if (!lv_add_integrity(lv, lp->integrity_arg, lp->integrity_meta_lv,
+				      lp->integrity_meta_name, &lp->integrity_settings, lp->pvh,
+				      &lp->zero_data_sectors))
+			goto revert_new_lv;
+
+		backup(vg);
+
+		/*
+		 * The standard option combination should be -Zy -ay, in which
+		 * case we activate here, and zero at the end of the command.
+		 * The invalid combination -Zy -an has been prevented earlier.
+		 * The combination -Zn -an involves no zeroing or activation.
+		 * For combination -Zn -ay we activate here.
+		 */
+		if (lp->zero) {
+			/* Activate for zeroing at the end of lvcreate. */
+		       	if (!activate_lv(cmd, lv)) {
+				log_error("Failed to activate LV %s.", display_lvname(lv));
+				goto out;
+			}
+		} else if (!lp->zero && is_change_activating(lp->activate)) {
+		       	if (!activate_lv(cmd, lv)) {
+				log_error("Failed to activate LV %s.", display_lvname(lv));
+				goto out;
+			}
+		}
+
+		/*
+		 * The entire LV is zeroed, which can take a long time,
+		 * so defer this to the end of the command when no locks
+		 * are held, and the command can be canceled without
+		 * problems (if the user doesn't want to wait, or wants
+		 * to do the zeroing themselves.)
+		 */
+		goto out;
+	}
+
 	if (seg_is_vdo_pool(lp)) {
 		if (!convert_vdo_pool_lv(lv, &lp->vdo_params, &lp->virtual_extents)) {
 			stack;

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 +84,15 @@
 #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 INTEGRITY_METADATA	UINT64_C(0x0000000004000000)    /* LV - Internal use only */
+#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 +264,8 @@
 #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_integrity_metadata(lv)   (((lv)->status & INTEGRITY_METADATA) ? 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)
@@ -272,9 +277,11 @@
 /* Recognize component LV (matching lib/misc/lvm-string.c _lvname_has_reserved_component_string()) */
 #define lv_is_component(lv) (lv_is_cache_origin(lv) || \
 			     lv_is_writecache_origin(lv) || \
+			     lv_is_integrity_origin(lv) || \
 			     ((lv)->status & (\
 					      CACHE_POOL_DATA |\
 					      CACHE_POOL_METADATA |\
+					      INTEGRITY_METADATA |\
 					      LV_CACHE_VOL |\
 					      LV_VDO_POOL_DATA |\
 					      MIRROR_IMAGE |\
@@ -519,6 +526,10 @@ 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 +1003,12 @@ 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;
+	uint64_t zero_data_sectors; /* the resulting size that should be zeroed at the end */
+
 	struct dm_list tags;	/* all */
 
 	int yes;
@@ -1086,6 +1103,8 @@ int lv_is_cache_origin(const struct logical_volume *lv);
 int lv_is_writecache_origin(const struct logical_volume *lv);
 int lv_is_writecache_cachevol(const struct logical_volume *lv);
 
+int lv_is_integrity_origin(const struct logical_volume *lv);
+
 int lv_is_merging_cow(const struct logical_volume *cow);
 uint32_t cow_max_extents(const struct logical_volume *origin, uint32_t chunk_size);
 int cow_has_min_chunks(const struct volume_group *vg, uint32_t cow_extents, uint32_t chunk_size);
@@ -1385,4 +1404,16 @@ 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,
+		     struct dm_list *pvh,
+		     uint64_t *zero_data_sectors);
+int lv_create_integrity_metadata(struct cmd_context *cmd,
+			struct volume_group *vg,
+			struct lvcreate_params *lp);
+int lv_remove_integrity(struct logical_volume *lv);
+
+int zero_lv_name(struct cmd_context *cmd, const char *vg_name, const char *lv_name, uint64_t lv_size_sectors);
+
 #endif

lib/metadata/mirror.c

@@ -266,7 +266,6 @@ static int _write_log_header(struct cmd_context *cmd, struct logical_volume *lv)
 	dev_set_last_byte(dev, sizeof(log_header));
 
 	if (!dev_write_bytes(dev, UINT64_C(0), sizeof(log_header), &log_header)) {
-		dev_unset_last_byte(dev);
 		log_error("Failed to write log header to %s.", name);
 		return 0;
 	}

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)

lib/metadata/thin_manip.c

@@ -503,11 +503,11 @@ int update_pool_lv(struct logical_volume *lv, int activate)
 			 *   which Node has pool active.
 			 */
 			if (!activate_lv(lv->vg->cmd, lv)) {
-				init_dmeventd_monitor(monitored);
+				(void) init_dmeventd_monitor(monitored);
 				return_0;
 			}
 			if (!lv_is_active(lv)) {
-				init_dmeventd_monitor(monitored);
+				(void) init_dmeventd_monitor(monitored);
 				log_error("Cannot activate thin pool %s, perhaps skipped in lvm.conf volume_list?",
 					  display_lvname(lv));
 				return 0;

lib/metadata/writecache_manip.c

@@ -21,8 +21,6 @@
 #include "lib/metadata/segtype.h"
 #include "lib/activate/activate.h"
 #include "lib/config/defaults.h"
-#include "lib/metadata/lv_alloc.h"
-#include "lib/misc/lvm-signal.h"
 #include "lib/activate/dev_manager.h"
 
 int lv_is_writecache_origin(const struct logical_volume *lv)

lib/raid/raid.c

@@ -240,6 +240,26 @@ static int _raid_text_export(const struct lv_segment *seg, struct formatter *f)
 	return _raid_text_export_raid(seg, f);
 }
 
+static int _image_integrity_data_sectors(struct lv_segment *seg, uint64_t *data_sectors)
+{
+	struct logical_volume *lv_image;
+	struct lv_segment *seg_image;
+	int s;
+
+	*data_sectors = 0;
+
+	for (s = 0; s < seg->area_count; s++) {
+		 lv_image = seg_lv(seg, s);
+
+		 if (lv_is_integrity(lv_image)) {
+			 seg_image = first_seg(lv_image);
+			 *data_sectors = seg_image->integrity_data_sectors;
+			 return 1;
+		 }
+	}
+	return 1;
+}
+
 static int _raid_add_target_line(struct dev_manager *dm __attribute__((unused)),
 				 struct dm_pool *mem __attribute__((unused)),
 				 struct cmd_context *cmd __attribute__((unused)),
@@ -256,6 +276,7 @@ static int _raid_add_target_line(struct dev_manager *dm __attribute__((unused)),
 	uint64_t writemostly[RAID_BITMAP_SIZE] = { 0 };
 	struct dm_tree_node_raid_params_v2 params = { 0 };
 	unsigned attrs;
+	uint64_t integrity_data_sectors = 0;
 
 	if (seg_is_raid4(seg)) {
 		if (!_raid_target_present(cmd, NULL, &attrs) ||
@@ -351,6 +372,15 @@ static int _raid_add_target_line(struct dev_manager *dm __attribute__((unused)),
 	params.stripe_size = seg->stripe_size;
 	params.flags = flags;
 
+	if (!_image_integrity_data_sectors(seg, &integrity_data_sectors))
+		return_0;
+
+	if (integrity_data_sectors) {
+		log_debug("Reducing raid size from %llu to integrity_data_sectors %llu",
+			  (unsigned long long)len, (unsigned long long)integrity_data_sectors);
+		len = integrity_data_sectors;
+	}
+
 	if (!dm_tree_node_add_raid_target_with_params_v2(node, len, &params))
 		return_0;
 

libdaemon/server/daemon-server.c

@@ -656,7 +656,8 @@ void daemon_start(daemon_state s)
 		failed = 1; /* FD out of available selectable set */
 
 	sigfillset(&new_set);
-	sigprocmask(SIG_SETMASK, NULL, &old_set);
+	if (sigprocmask(SIG_SETMASK, NULL, &old_set))
+		perror("sigprocmask error");
 
 	while (!failed) {
 		_reset_timeout(s);

test/Makefile.in

@@ -315,7 +315,7 @@ lib/dm-version-expected: $(top_srcdir)/VERSION_DM .lib-dir-stamp
 CMDS = lvm $(shell cat $(top_builddir)/tools/.commands 2>/dev/null)
 LIB = $(addprefix lib/, $(LIB_SECURETEST) $(LIB_DMSECURETEST) $(LIB_SHARED) $(LIB_LOCAL) $(LIB_NOT) $(LIB_LINK_NOT) $(LIB_FLAVOURS))
 
-.tests-stamp: $(ALL) $(LIB) $(SUBDIRS) lib/version-expected lib/dm-version-expected unit-test
+.tests-stamp: .lib-dir-stamp $(ALL) $(LIB) $(SUBDIRS) lib/version-expected lib/dm-version-expected unit-test
 	@echo "    [TEST-STAMP]"
 	@if test "$(srcdir)" != . ; then \
 		echo "Linking tests to builddir."; \

test/shell/lvconvert-raid-reshape-stripes-load-reload.sh

@@ -56,7 +56,7 @@ mount "$DM_DEV_DIR/$vg/$lv1" "$mount_dir"
 
 echo 3 >/proc/sys/vm/drop_caches
 # FIXME: This is filling up ram disk. Use sane amount of data please! Rate limit the data written!
-dd if=/dev/urandom of="$mount_dir/random" bs=1M count=50 conv=fdatasync
+dd if=/dev/urandom of="$mount_dir/random" bs=1M count=4 conv=fdatasync
 checksum_ "$mount_dir/random" >MD5
 
 # FIXME: wait_for_sync - is this really testing anything under load?

test/shell/nomda-missing.sh

@@ -74,6 +74,8 @@ check active $vg mirror12
 vgchange -a n $vg
 aux enable_dev "$dev3"
 aux disable_dev "$dev4"
+dmsetup table
+dmsetup info -c
 vgchange -aey $vg
 not vgck $vg
 

tools/args.h

@@ -230,9 +230,6 @@ arg(errorwhenfull_ARG, '\0', "errorwhenfull", bool_VAL, 0, 0,
     "(Also see dm-thin-pool kernel module option no_space_timeout.)\n"
     "See \\fBlvmthin\\fP(7) for more information.\n")
 
-arg(file_long_ARG, '\0', "file", string_VAL, 0, 0,
-    "File name.\n")
-
 arg(force_long_ARG, '\0', "force", 0, ARG_COUNTABLE, 0,
     "Force metadata restore even with thin pool LVs.\n"
     "Use with extreme caution. Most changes to thin metadata\n"
@@ -275,6 +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

@@ -719,12 +719,12 @@ FLAGS: SECONDARY_SYNTAX
 # and the LV type is known.
 
 lvconvert --repair LV_cache_cachepool_mirror_raid_thinpool
-OO: --usepolicies, --interval Number, --poolmetadataspare Bool, --file_long String, OO_LVCONVERT
+OO: --usepolicies, --interval Number, --poolmetadataspare Bool, OO_LVCONVERT
 OP: PV ...
 ID: lvconvert_repair
 DESC: Replace failed PVs in a raid or mirror LV.
 DESC: Repair a thin pool.
-DESC: Repair a cache pool or cache.
+DESC: Repair a cache pool.
 RULE: all not lv_is_locked lv_is_pvmove
 RULE: --poolmetadataspare and LV_cache LV_cachepool LV_thinpool
 
@@ -757,6 +757,14 @@ FLAGS: SECONDARY_SYNTAX
 
 ---
 
+lvconvert --integrity String LV
+OO: OO_LVCONVERT
+OP: PV ...
+ID: lvconvert_integrity
+DESC: Remove integrity from an LV.
+
+---
+
 # --extents is not specified; it's an automatic alternative for --size
 
 OO_LVCREATE: --addtag Tag, --alloc Alloc, --autobackup Bool, --activate Active,
@@ -870,7 +878,8 @@ DESC: Create a raid1 or mirror LV (infers --type raid1|mirror).
 # R9,R10,R11,R12 (--type raid with any use of --stripes/--mirrors)
 lvcreate --type raid --size SizeMB VG
 OO: --mirrors PNumber, --stripes Number, --stripesize SizeKB,
---regionsize RegionSize, --minrecoveryrate SizeKB, --maxrecoveryrate SizeKB, OO_LVCREATE
+--regionsize RegionSize, --minrecoveryrate SizeKB, --maxrecoveryrate SizeKB,
+--integrity String, --integritysettings String, OO_LVCREATE
 OP: PV ...
 ID: lvcreate_raid_any
 DESC: Create a raid LV (a specific raid level must be used, e.g. raid1).
@@ -1269,6 +1278,20 @@ 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/lvconvert.c

@@ -2480,339 +2480,10 @@ deactivate_pmslv:
 	return 1;
 }
 
-/*
- * Use cache_repair to read metadata from cachevol_lv and
- * write repaired metadata to new file.
+/* TODO: lots of similar code with  thinpool repair
+ *       investigate possible better code sharing...
  */
-
-static int _run_cache_repair_from_cachevol(struct cmd_context *cmd,
-			     struct logical_volume *cachevol_lv,
-			     const char *meta_file)
-{
-	char cachevol_lv_path[PATH_MAX];
-	char *cachevol_lv_name;
-	const char *dmdir = dm_dir();
-	const char *cache_repair = find_config_tree_str_allow_empty(cmd, global_cache_repair_executable_CFG, NULL);
-	const struct dm_config_node *cn;
-	const struct dm_config_value *cv;
-	const char *argv[MAX_PDATA_ARGS + 7]; /* Max supported args */
-	int args = 0;
-	int status;
-
-	if (!cache_repair || !cache_repair[0]) {
-		log_error("No cache_repair command set in lvm.conf. Repair is disabled.");
-		return 0;
-	}
-
-	if (!(cachevol_lv_name = dm_build_dm_name(cmd->mem, cachevol_lv->vg->name, cachevol_lv->name, NULL))) {
-		log_error("Failed to create dm name for %s.", display_lvname(cachevol_lv));
-		return 0;
-	}
-
-	if (dm_snprintf(cachevol_lv_path, sizeof(cachevol_lv_path), "%s/%s", dmdir, cachevol_lv_name) < 0) {
-		log_error("Failed to create path for %s.", display_lvname(cachevol_lv));
-		return 0;
-	}
-
-	if (!(cn = find_config_tree_array(cmd, global_cache_repair_options_CFG, NULL))) {
-		log_error(INTERNAL_ERROR "Unable to find configuration for global/cache_repair_options");
-		return 0;
-	}
-
-	for (cv = cn->v; cv && args < MAX_PDATA_ARGS; cv = cv->next) {
-		if (cv->type != DM_CFG_STRING) {
-			log_error("Invalid string in config file: "
-				  "global/cache_repair_options");
-			return 0;
-		}
-		argv[++args] = cv->v.str;
-	}
-
-	if (args >= MAX_PDATA_ARGS) {
-		log_error("Too many cache_repair_options set in lvm.conf.");
-		return 0;
-	}
-
-	argv[0] = cache_repair;
-	argv[++args] = "-i";
-	argv[++args] = cachevol_lv_path;
-	argv[++args] = "-o";
-	argv[++args] = meta_file;
-	argv[++args] = NULL;
-
-	if (!exec_cmd(cmd, (const char * const *)argv, &status, 1)) {
-		log_error("The cache_repair command failed (status %d).", status);
-		return 0;
-	}
-
-	log_print("cache_repair of metadata completed from %s to %s", display_lvname(cachevol_lv), meta_file);
-
-	return 1;
-}
-
-static int _run_cache_writeback_from_cachevol(struct cmd_context *cmd,
-			        struct logical_volume *cache_lv,
-			        struct logical_volume *cachevol_lv,
-			        const char *meta_file,
-				uint64_t data_offset_bytes)
-{
-	char orig_path[PATH_MAX];
-	char data_path[PATH_MAX];
-	char *orig_name;
-	char *data_name;
-	const char *dmdir = dm_dir();
-	const char *cache_writeback = "cache_writeback";
-	const char *argv[MAX_PDATA_ARGS + 7]; /* Max supported args */
-	char offset_str[16];
-	int args = 0;
-	int status;
-
-	if (!(orig_name = dm_build_dm_name(cmd->mem, cache_lv->vg->name, cache_lv->name, NULL))) {
-		log_error("Failed to create dm name for %s.", display_lvname(cache_lv));
-	}
-
-	if (!(data_name = dm_build_dm_name(cmd->mem, cachevol_lv->vg->name, cachevol_lv->name, NULL))) {
-		log_error("Failed to create dm name for %s.", display_lvname(cachevol_lv));
-	}
-
-	if (dm_snprintf(orig_path, sizeof(orig_path), "%s/%s", dmdir, orig_name) < 0) {
-		log_error("Failed to create path for %s.", display_lvname(cache_lv));
-		return 0;
-	}
-
-	/* _off suffix is temp hack to allow manual offset */
-	if (dm_snprintf(data_path, sizeof(data_path), "%s/%s_off", dmdir, data_name) < 0) {
-		log_error("Failed to create path for %s.", display_lvname(cachevol_lv));
-		return 0;
-	}
-
-	if (dm_snprintf(offset_str, sizeof(offset_str), "%llu",
-			(unsigned long long)data_offset_bytes) < 0)
-		return_0;
-
-	/* let the user set up an _off dm wrapper */
-	if (yes_no_prompt("Done creating wrapper %s with offset %u sectors? [y/n]: ",
-			  data_path, (uint32_t)data_offset_bytes/512) == 'n')
-		return_0;
-
-	argv[0] = cache_writeback;
-	argv[++args] = "--metadata-device";
-	argv[++args] = meta_file;
-	argv[++args] = "--origin-device";
-	argv[++args] = orig_path;
-	argv[++args] = "--fast-device";
-	argv[++args] = data_path;
-	/*
-	argv[++args] = "--fast-device-offset";
-	argv[++args] = offset_str;
-	*/
-	argv[++args] = "--no-metadata-update";
-	argv[++args] = NULL;
-
-	if (!exec_cmd(cmd, (const char * const *)argv, &status, 1)) {
-		log_error("The cache_writeback command failed (status %d).", status);
-		return 0;
-	}
-
-	log_print("cache_writeback of data completed from %s to %s",
-		  display_lvname(cachevol_lv), display_lvname(cache_lv));
-
-	return 1;
-}
-
-static int _lvconvert_cache_repair_cachevol(struct cmd_context *cmd,
-				   struct logical_volume *cache_lv,
-				   struct dm_list *pvh)
-{
-	char filepath[PATH_MAX];
-	char zero_sector_buf[512];
-	struct volume_group *vg = cache_lv->vg;
-	struct logical_volume *cachevol_lv;
-	struct lv_segment *cache_seg, *cachevol_seg;
-	uint64_t metadata_len_sectors, data_offset_bytes;
-	uint64_t orig_cachevol_seglen;
-	const char *meta_file = NULL;
-	FILE *fp;
-	uint32_t prealloc_sectors_written = 0;
-	int cache_mode;
-	int ret = 0;
-
-	memset(zero_sector_buf, 0, 512);
-
-	if (!(cache_seg = first_seg(cache_lv)))
-		return_0;
-
-	if (!(cachevol_lv = cache_seg->pool_lv))
-		return_0;
-
-	if (!lv_is_cache_vol(cachevol_lv))
-		return_0;
-
-	if (cache_seg->metadata_start) {
-		/* metadata is always at 0, but check in case it changes in future */
-		log_error("Cannot handle non-zero metadata start locations.");
-		return 0;
-	}
-
-	if (lv_is_active(cache_lv)) {
-		log_error("LV %s must not be active.", display_lvname(cache_lv));
-		return 0;
-	}
-
-	if (lv_is_active(cachevol_lv)) {
-		log_error("LV %s must not be active.", display_lvname(cachevol_lv));
-		return 0;
-	}
-
-	if (vg_is_shared(vg)) {
-		/* cache LV using the cachevol cannot be active elsewhere */
-		if (!lockd_lv(cmd, cache_lv, "ex", 0))
-			return_0;
-
-		/* cachevol cannot be active elsewhere. */
-		if (!lockd_lv(cmd, cachevol_lv, "ex", 0))
-			return_0;
-	}
-
-	metadata_len_sectors = cache_seg->metadata_len;
-	cache_mode = cache_seg->cache_mode;
-
-	/* Disable syncing the cache during split */
-	if (cache_mode != CACHE_MODE_WRITETHROUGH)
-		cache_seg->cache_mode = CACHE_MODE_WRITETHROUGH;
-
-	if (!_lvconvert_split_and_keep_cachevol(cmd, cache_lv, cachevol_lv)) {
-		log_error("Failed to detach cachevol, not repairing.");
-		return_0;
-	}
-
-	/*
-	 * Note: following the spilt, cache_lv is no longer technically a cache
-	 * and cachevol_lv is no longer technically a cachevol.  The variable
-	 * names represent their previous types.
-	 */
-
-	if (cache_mode == CACHE_MODE_WRITETHROUGH) {
-		log_print("No repair or writeback required for writethrough cache mode.");
-		return 1;
-	}
-
-	/*
-	 * Prevent the active LVs from being exposed and used outside of lvm.
-	 */
-	cachevol_lv->status |= LV_TEMPORARY;
-	cache_lv->status |= LV_TEMPORARY;
-
-	/*
-	 * Temporarily change the cachevol LV size to be only the
-	 * metadata size, so cache_repair will only look at the
-	 * metadata portion.
-	 * seg->len and seg->metadata_len units are sectors
-	 */
-	cachevol_seg = first_seg(cachevol_lv);
-	orig_cachevol_seglen = cachevol_seg->len;
-	cachevol_seg->len = metadata_len_sectors / vg->extent_size;
-
-	if (!activate_lv(cmd, cachevol_lv)) {
-		log_error("Failed to activate LV %s for cache repair and writeback.", display_lvname(cachevol_lv));
-		goto out;
-	}
-
-	if (!sync_local_dev_names(cmd)) {
-		log_error("Failed to sync local devices before cache_repair.");
-		goto out;
-	}
-
-	/*
-	 * cache_repair wants the destination file to be preallocated.
-	 * FIXME: what's the relationship between the metadata size and
-	 * the file size?
-	 */
-
-	if (arg_is_set(cmd, file_long_ARG))
-		meta_file = arg_str_value(cmd, file_long_ARG, NULL);
-	else {
-		snprintf(filepath, PATH_MAX, "/tmp/lvconvert-cache-repair-metadata-%d", getpid());
-		meta_file = filepath;
-	}
-
-	log_print("Allocating %u MB for metadata in %s",
-		  (uint32_t)(metadata_len_sectors * 512 / 1048576), meta_file);
-
-	if (!(fp = fopen(meta_file, "w+"))) {
-		log_error("Failed to open file for metadata %s", meta_file);
-		meta_file = NULL;
-		goto out;
-	}
-	while (prealloc_sectors_written < metadata_len_sectors) {
-		if (!fwrite(zero_sector_buf, 512, 1, fp)) {
-			fclose(fp);
-			log_error("Failed to allocate space for metadata in %s", meta_file);
-			goto out;
-		}
-		prealloc_sectors_written++;
-	}
-	fflush(fp);
-	fclose(fp);
-
-	if (!_run_cache_repair_from_cachevol(cmd, cachevol_lv, meta_file)) {
-		log_error("cache_repair failed, cache data not written back to original volume.");
-		goto out;
-	}
-
-	/*
-	 * Deactivate cachevol LV, and reactivate with correct size.
-	 * Activate cache LV.
-	 */
-
-	if (!deactivate_lv(cmd, cachevol_lv)) {
-		log_error("Failed to deactivate LV %s", cachevol_lv->name);
-		goto out;
-	}
-
-	cachevol_seg->len = orig_cachevol_seglen;
-
-	if (!activate_lv(cmd, cachevol_lv)) {
-		log_error("Failed to activate LV %s for cache repair and writeback.", display_lvname(cachevol_lv));
-		goto out;
-	}
-
-	if (!activate_lv(cmd, cache_lv)) {
-		log_error("Failed to activate LV %s for cache repair and writeback.", display_lvname(cache_lv));
-		goto out;
-	}
-
-	if (!sync_local_dev_names(cmd)) {
-		log_error("Failed to sync local devices before cache_writeback.");
-		goto out;
-	}
-
-	data_offset_bytes = metadata_len_sectors * 512;
-
-	if (!_run_cache_writeback_from_cachevol(cmd, cache_lv, cachevol_lv, meta_file, data_offset_bytes)) {
-		log_error("cache_writeback failed");
-		goto_out;
-	}
-
-	log_print("cachevol writeback complete, %s now unused.", cachevol_lv->name);
-	ret = 1;
-out:
-	if (!deactivate_lv(cmd, cache_lv))
-		log_error("Failed to deactivate LV %s", cache_lv->name);
-
-	if (!deactivate_lv(cmd, cachevol_lv))
-		log_error("Failed to deactivate LV %s", cachevol_lv->name);
-
-	if (meta_file)
-		unlink(meta_file);
-
-	if (!ret)
-		log_error("Cache data on %s not written back to original LV %s.",
-			  display_lvname(cachevol_lv), display_lvname(cache_lv));
-	return ret;
-}
-
-static int _lvconvert_cache_repair_cachepool(struct cmd_context *cmd,
+static int _lvconvert_cache_repair(struct cmd_context *cmd,
 				   struct logical_volume *cache_lv,
 				   struct dm_list *pvh, int poolmetadataspare)
 {
@@ -2831,6 +2502,11 @@ static int _lvconvert_cache_repair_cachepool(struct cmd_context *cmd,
 	struct logical_volume *pmslv;
 	struct logical_volume *mlv;
 
+	if (lv_is_cache(cache_lv) && lv_is_cache_vol(first_seg(cache_lv)->pool_lv)) {
+		log_error("Manual repair required.");
+		return 0;
+	}
+
 	pool_lv = lv_is_cache_pool(cache_lv) ? cache_lv : first_seg(cache_lv)->pool_lv;
 	mlv = first_seg(pool_lv)->metadata_lv;
 
@@ -4093,7 +3769,7 @@ static int _lvconvert_repair_pvs(struct cmd_context *cmd, struct logical_volume
 	return ret ? ECMD_PROCESSED : ECMD_FAILED;
 }
 
-static int _lvconvert_repair_cache_or_thin(struct cmd_context *cmd, struct logical_volume *lv,
+static int _lvconvert_repair_cachepool_thinpool(struct cmd_context *cmd, struct logical_volume *lv,
 			struct processing_handle *handle)
 {
 	int poolmetadataspare = arg_int_value(cmd, poolmetadataspare_ARG, DEFAULT_POOL_METADATA_SPARE);
@@ -4114,13 +3790,8 @@ static int _lvconvert_repair_cache_or_thin(struct cmd_context *cmd, struct logic
 		if (!_lvconvert_thin_pool_repair(cmd, lv, use_pvh, poolmetadataspare))
 			return_ECMD_FAILED;
 	} else /* cache */ {
-		if (lv_is_cache(lv) && lv_is_cache_vol(first_seg(lv)->pool_lv)) {
-			if (!_lvconvert_cache_repair_cachevol(cmd, lv, use_pvh))
-				return_ECMD_FAILED;
-		} else {
-			if (!_lvconvert_cache_repair_cachepool(cmd, lv, use_pvh, poolmetadataspare))
-				return_ECMD_FAILED;
-		}
+		if (!_lvconvert_cache_repair(cmd, lv, use_pvh, poolmetadataspare))
+			return_ECMD_FAILED;
 	}
 
 	return ECMD_PROCESSED;
@@ -4132,7 +3803,7 @@ static int _lvconvert_repair_single(struct cmd_context *cmd, struct logical_volu
 	if (lv_is_thin_pool(lv) ||
 	    lv_is_cache(lv) ||
 	    lv_is_cache_pool(lv))
-		return _lvconvert_repair_cache_or_thin(cmd, lv, handle);
+		return _lvconvert_repair_cachepool_thinpool(cmd, lv, handle);
 
 	if (lv_is_raid(lv) || lv_is_mirror(lv))
 		return _lvconvert_repair_pvs(cmd, lv, handle);
@@ -4143,6 +3814,12 @@ static int _lvconvert_repair_single(struct cmd_context *cmd, struct logical_volu
 	return ECMD_FAILED;
 }
 
+/*
+ * FIXME: add option --repair-pvs to call _lvconvert_repair_pvs() directly,
+ * and option --repair-thinpool to call _lvconvert_repair_thinpool().
+ * and option --repair-cache to call _lvconvert_repair_cache().
+ * and option --repair-cachepool to call _lvconvert_repair_cachepool().
+ */
 int lvconvert_repair_cmd(struct cmd_context *cmd, int argc, char **argv)
 {
 	struct processing_handle *handle;
@@ -6053,6 +5730,75 @@ int lvconvert_to_cache_with_cachevol_cmd(struct cmd_context *cmd, int argc, char
 	return ret;
 }
 
+static int _lvconvert_integrity_single(struct cmd_context *cmd,
+					struct logical_volume *lv,
+					struct processing_handle *handle)
+{
+	struct volume_group *vg = lv->vg;
+	struct lv_segment *seg;
+	const char *arg = arg_str_value(cmd, integrity_ARG, NULL);
+
+	if (strcmp(arg, "none") && strcmp(arg, "n")) {
+		log_error("Integrity can only be removed from an existing LV (see --integrity none).");
+		return ECMD_FAILED;
+	}
+
+	if (!lv_is_integrity(lv)) {
+		log_error("LV does not have integrity.");
+		return ECMD_FAILED;
+	}
+
+	seg = first_seg(lv);
+
+	if (!seg->integrity_meta_dev) {
+		log_error("Internal integrity cannot be removed.");
+		return ECMD_FAILED;
+	}
+
+	/* TODO: lift this restriction */
+	if (lv_info(cmd, lv, 1, NULL, 0, 0)) {
+		log_error("LV must be inactive to remove integrity.");
+		return ECMD_FAILED;
+	}
+
+	if (!archive(vg))
+		return ECMD_FAILED;
+
+	if (!lv_remove_integrity(lv))
+		return ECMD_FAILED;
+
+	if (!vg_write(vg) || !vg_commit(vg))
+		return ECMD_FAILED;
+
+	backup(vg);
+
+	log_print_unless_silent("Logical volume %s has removed integrity.", display_lvname(lv));
+	return ECMD_PROCESSED;
+}
+
+int lvconvert_integrity_cmd(struct cmd_context *cmd, int argc, char **argv)
+{
+	struct processing_handle *handle;
+	struct lvconvert_result lr = { 0 };
+	int ret;
+
+	if (!(handle = init_processing_handle(cmd, NULL))) {
+		log_error("Failed to initialize processing handle.");
+		return ECMD_FAILED;
+	}
+
+	handle->custom_handle = &lr;
+
+	cmd->cname->flags &= ~GET_VGNAME_FROM_OPTIONS;
+
+	ret = process_each_lv(cmd, cmd->position_argc, cmd->position_argv, NULL, NULL, READ_FOR_UPDATE, handle, NULL,
+			      &_lvconvert_integrity_single);
+
+	destroy_processing_handle(cmd, handle);
+
+	return ret;
+}
+
 /*
  * All lvconvert command defs have their own function,
  * so the generic function name is unused.

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;
 
@@ -846,6 +848,8 @@ static int _lvcreate_params(struct cmd_context *cmd,
 	poolmetadataspare_ARG
 
 #define RAID_ARGS \
+	integrity_ARG,\
+	integritysettings_ARG,\
 	maxrecoveryrate_ARG,\
 	minrecoveryrate_ARG,\
 	raidmaxrecoveryrate_ARG,\
@@ -1218,6 +1222,11 @@ static int _lvcreate_params(struct cmd_context *cmd,
 		}
 	}
 
+	if (seg_is_integrity(lp) || seg_is_raid(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;
 
@@ -1575,6 +1584,13 @@ static int _check_zero_parameters(struct cmd_context *cmd, struct lvcreate_param
 	if (seg_is_thin(lp))
 		return 1;
 
+	if (seg_is_integrity(lp)) {
+		if (lp->zero && !is_change_activating(lp->activate)) {
+			log_error("Zeroing integrity is not compatible with inactive creation (-an).");
+			return 0;
+		}
+	}
+
 	/* If there is some problem, buffer will not be empty */
 	if (dm_snprintf(buf, sizeof(buf), "%s%s%s%s%s%s%s",
 			lp->origin_name ? "origin " : "",
@@ -1702,6 +1718,24 @@ 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.");
@@ -1775,5 +1809,17 @@ int lvcreate(struct cmd_context *cmd, int argc, char **argv)
 
 	_destroy_lvcreate_params(&lp);
 	destroy_processing_handle(cmd, handle);
+
+	/*
+	 * When set, the final LV should be zeroed (set by add_integrity
+	 * to intialize integrity metadata/checksums).
+	 */
+	if (lp.zero_data_sectors) {
+		if (!lp.zero)
+			log_warn("WARNING: not zeroing integrity LV, read errors are possible.");
+		else
+			zero_lv_name(cmd, lp.vg_name, lp.lv_name, lp.zero_data_sectors);
+	}
+
 	return ret;
 }

tools/lvmcmdline.c

@@ -149,6 +149,9 @@ static const struct command_function _command_functions[CMD_COUNT] = {
 	{ lvconvert_to_vdopool_CMD, lvconvert_to_vdopool_cmd },
 	{ lvconvert_to_vdopool_param_CMD, lvconvert_to_vdopool_param_cmd },
 
+	/* lvconvert for integrity */
+	{ lvconvert_integrity_CMD, lvconvert_integrity_cmd },
+
 	{ pvscan_display_CMD, pvscan_display_cmd },
 	{ pvscan_cache_CMD, pvscan_cache_cmd },
 };

tools/pvck.c

@@ -359,7 +359,10 @@ static int _dump_all_text(struct cmd_context *cmd, const char *tofile, struct de
 			continue;
 		}
 
-		sscanf(line, "seqno = %u", &seqno);
+		if (sscanf(line, "seqno = %u", &seqno) != 1) {
+			log_warn("WARNING: Failed to parse seqno.");
+			seqno = 0; /* Skip? */
+		}
 
 		/*
 		 * The first three lines look like metadata with
@@ -417,11 +420,11 @@ static int _dump_all_text(struct cmd_context *cmd, const char *tofile, struct de
 				memset(line, 0, sizeof(line));
 				_copy_line(str1, line, &len);
 				log_print("%s", line);
-			}
-			if ((str2 = strstr(str1, "creation_time = "))) {
-				memset(line, 0, sizeof(line));
-				_copy_line(str2, line, &len);
-				log_print("%s\n", line);
+				if ((str2 = strstr(str1, "creation_time = "))) {
+					memset(line, 0, sizeof(line));
+					_copy_line(str2, line, &len);
+					log_print("%s\n", line);
+				}
 			}
 		}
 
@@ -652,27 +655,31 @@ static int _dump_meta_area(struct device *dev, const char *tofile,
 	if (!tofile)
 		return_0;
 
-	if (!(meta_buf = malloc(mda_size)))
+	if (!(meta_buf = zalloc(mda_size)))
 		return_0;
-	memset(meta_buf, 0, mda_size);
 
 	if (!dev_read_bytes(dev, mda_offset, mda_size, meta_buf)) {
 		log_print("CHECK: failed to read metadata area at offset %llu size %llu",
 			  (unsigned long long)mda_offset, (unsigned long long)mda_size);
+		free(meta_buf);
 		return 0;
 	}
 
 	if (!(fp = fopen(tofile, "wx"))) {
-		log_error("Failed to create file %s", tofile);
+		log_error("Failed to create file %s.", tofile);
+		free(meta_buf);
 		return 0;
 	}
 
-	fwrite(meta_buf, mda_size - 512, 1, fp);
+	if (fwrite(meta_buf, mda_size - 512, 1, fp) < (mda_size - 512))
+		log_warn("WARNING: Failed to write " FMTu64 " bytes to file %s.", mda_size - 512, tofile);
 
+	free(meta_buf);
 	if (fflush(fp))
 		stack;
 	if (fclose(fp))
 		stack;
+
 	return 1;
 }
 
@@ -705,12 +712,11 @@ static int _dump_current_text(struct device *dev,
 	int ri = rlocn_index; /* 0 or 1 */
 	int bad = 0;
 
-	if (!(meta_buf = malloc(meta_size))) {
+	if (!(meta_buf = zalloc(meta_size))) {
 		log_print("CHECK: mda_header_%d.raw_locn[%d] no mem for metadata text size %llu", mn, ri,
 			  (unsigned long long)meta_size);
 		return 0;
 	}
-	memset(meta_buf, 0, meta_size);
 
 	/*
 	 * Read the metadata text specified by the raw_locn so we can
@@ -721,7 +727,7 @@ static int _dump_current_text(struct device *dev,
 	 * mda_offset + meta_offset.
 	 */
 	if (meta_offset + meta_size > mda_size) {
-	 	/* text metadata wraps to start of text metadata area */
+		/* text metadata wraps to start of text metadata area */
 		uint32_t wrap = (uint32_t) ((meta_offset + meta_size) - mda_size);
 		off_t offset_a = mda_offset + meta_offset;
 		uint32_t size_a = meta_size - wrap;
@@ -732,6 +738,7 @@ static int _dump_current_text(struct device *dev,
 			log_print("CHECK: failed to read metadata text at mda_header_%d.raw_locn[%d].offset %llu size %llu part_a %llu %llu", mn, ri,
 				  (unsigned long long)meta_offset, (unsigned long long)meta_size,
 				  (unsigned long long)offset_a, (unsigned long long)size_a);
+			free(meta_buf);
 			return 0;
 		}
 
@@ -739,12 +746,14 @@ static int _dump_current_text(struct device *dev,
 			log_print("CHECK: failed to read metadata text at mda_header_%d.raw_locn[%d].offset %llu size %llu part_b %llu %llu", mn, ri,
 				  (unsigned long long)meta_offset, (unsigned long long)meta_size,
 				  (unsigned long long)offset_b, (unsigned long long)size_b);
+			free(meta_buf);
 			return 0;
 		}
 	} else {
 		if (!dev_read_bytes(dev, mda_offset + meta_offset, meta_size, meta_buf)) {
 			log_print("CHECK: failed to read metadata text at mda_header_%d.raw_locn[%d].offset %llu size %llu", mn, ri,
 				  (unsigned long long)meta_offset, (unsigned long long)meta_size);
+			free(meta_buf);
 			return 0;
 		}
 	}
@@ -802,9 +811,9 @@ static int _dump_current_text(struct device *dev,
 	}
 
  out:
-	if (bad)
-		return 0;
-	return 1;
+	free(meta_buf);
+
+	return (!bad) ? 1 : 0;
 }
 
 static int _dump_label_and_pv_header(struct cmd_context *cmd, int print_fields,
@@ -848,6 +857,7 @@ static int _dump_label_and_pv_header(struct cmd_context *cmd, int print_fields,
 	if (!dev_read_bytes(dev, lh_offset, 512, buf)) {
 		log_print("CHECK: failed to read label_header at %llu",
 			  (unsigned long long)lh_offset);
+		free(buf);
 		return 0;
 	}
 
@@ -1042,9 +1052,9 @@ static int _dump_label_and_pv_header(struct cmd_context *cmd, int print_fields,
 			  (unsigned long long)xlate64(dlocn->size));
 	}
 
-	if (bad)
-		return 0;
-	return 1;
+	free(buf);
+
+	return (!bad) ? 1 : 0;
 }
 
 /*
@@ -1097,6 +1107,7 @@ static int _dump_mda_header(struct cmd_context *cmd,
 
 	if (!dev_read_bytes(dev, mda_offset, 512, buf)) {
 		log_print("CHECK: failed to read mda_header at %llu", (unsigned long long)mda_offset);
+		free(buf);
 		return 0;
 	}
 
@@ -1187,11 +1198,9 @@ static int _dump_mda_header(struct cmd_context *cmd,
 
 	/* Should we also check text metadata if it exists in rlocn1? */
  out:
-	if (buf)
-		free(buf);
-	if (bad)
-		return 0;
-	return 1;
+	free(buf);
+
+	return (!bad) ? 1 : 0;
 }
 
 static int _dump_headers(struct cmd_context *cmd,
@@ -1457,7 +1466,8 @@ static int _dump_search(struct cmd_context *cmd,
 		uint64_t dev_bytes;
 		uint64_t extra_bytes;
 
-		dev_get_size(dev, &dev_sectors);
+		if (!dev_get_size(dev, &dev_sectors))
+			return_ECMD_FAILED;
 
 		dev_bytes = dev_sectors * 512;
 		extra_bytes = dev_bytes % ONE_MB_IN_BYTES;
@@ -1484,9 +1494,8 @@ static int _dump_search(struct cmd_context *cmd,
 	log_print("Searching for metadata in mda%d at offset %llu size %llu", mda_num,
 		  (unsigned long long)mda_offset, (unsigned long long)mda_size);
 
-	if (!(buf = malloc(mda_size)))
+	if (!(buf = zalloc(mda_size)))
 		return ECMD_FAILED;
-	memset(buf, 0, mda_size);
 
 	if (!dev_read_bytes(dev, mda_offset, mda_size, buf)) {
 		log_print("CHECK: failed to read metadata area at offset %llu size %llu",

tools/pvscan.c

@@ -420,7 +420,7 @@ check_duplicate:
 		log_error("pvscan[%d] PV %s is duplicate for PVID %s on %d:%d and %d:%d.",
 			  getpid(), dev_name(dev), dev->pvid, major, minor, file_major, file_minor);
 
-	if (file_vgname[0] && strcmp(file_vgname, vgname))
+	if (file_vgname[0] && vgname && strcmp(file_vgname, vgname))
 		log_error("pvscan[%d] PV %s has unexpected VG %s vs %s.",
 			  getpid(), dev_name(dev), vgname, file_vgname);
 

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;
@@ -4284,7 +4456,8 @@ static int _process_pvs_in_vg(struct cmd_context *cmd,
 		/*
 		 * When processing only specific PVs, we can quit once they've all been found.
 	 	 */
-		if (!process_all_pvs && dm_list_empty(arg_tags) && dm_list_empty(arg_devices))
+		if (!process_all_pvs && dm_list_empty(arg_tags) &&
+		    (!arg_devices || dm_list_empty(arg_devices)))
 			break;
 		log_set_report_object_name_and_id(NULL, NULL);
 	}

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);
@@ -272,6 +275,8 @@ int lvconvert_merge_cmd(struct cmd_context *cmd, int argc, char **argv);
 int lvconvert_to_vdopool_cmd(struct cmd_context *cmd, int argc, char **argv);
 int lvconvert_to_vdopool_param_cmd(struct cmd_context *cmd, int argc, char **argv);
 
+int lvconvert_integrity_cmd(struct cmd_context *cmd, int argc, char **argv);
+
 int pvscan_display_cmd(struct cmd_context *cmd, int argc, char **argv);
 int pvscan_cache_cmd(struct cmd_context *cmd, int argc, char **argv);