[PATCH 0/6] Add raid0<->raid10 takeover support for imsm metadata

[PATCH 0/6] Add raid0<->raid10 takeover support for imsm metadata

[Krzysztof Wojcik]

The following series implements raid0<->raid10 takeover support
for imsm metadata format.
It contains also some fixes for generic takeover flow.

---

Krzysztof Wojcik (6):
      Define imsm_analyze_change function
      Set reshape.after.data_disks for raid0<->raid10 takeover
      Unfreeze for non re-striping transitions
      Add raid10 -> raid0 takeover support
      raid0->raid10 takeover- create metadata update
      raid0->raid10 takeover- allocate memory for added disks


 Grow.c        |    5 +
 super-intel.c |  289 +++++++++++++++++++++++++++++++++++++++++++++++++++++++--
 2 files changed, 282 insertions(+), 12 deletions(-)

-- 
Krzyszto Wojcik

[PATCH 1/6] Define imsm_analyze_change function

[Krzysztof Wojcik]

Function intended to use for single volume migration.
Function analyze transition and validate if it is supported.

Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com>
---
 super-intel.c |  135 ++++++++++++++++++++++++++++++++++++++++++++++++++++-----
 1 files changed, 124 insertions(+), 11 deletions(-)

diff --git a/super-intel.c b/super-intel.c
index 4cbb070..c12ba74 100644
--- a/super-intel.c
+++ b/super-intel.c
@@ -6576,16 +6576,110 @@ static void imsm_update_metadata_locally(struct supertype *st,
 	}
 }
 
-/*******************************************************************************
+/***************************************************************************
 * Function:	imsm_analyze_change
-* Description:	Function analyze and validate change for single volume migration
+* Description:	Function analyze change for single volume
+* 		and validate if transition is supported
 * Parameters:	Geometry parameters, supertype structure
 * Returns:	Operation type code on success, -1 if fail
-********************************************************************************/
-enum imsm_reshape_type imsm_analyze_change(
-	struct supertype *st, struct geo_params *geo)
+****************************************************************************/
+enum imsm_reshape_type imsm_analyze_change(struct supertype *st,
+						struct geo_params *geo)
 {
-	return -1;
+	struct mdinfo info;
+	int change = -1;
+	int check_devs = 0;
+
+	getinfo_super_imsm_volume(st, &info, NULL);
+
+	if ((geo->level != info.array.level) &&
+		(geo->level >= 0) &&
+		(geo->level != UnSet)) {
+		switch (info.array.level) {
+		case 0:
+			if (geo->level == 5) {
+				change = CH_LEVEL_MIGRATION;
+				check_devs = 1;
+			}
+			if (geo->level == 10) {
+				change = CH_TAKEOVER;
+				check_devs = 1;
+			}
+			break;
+		case 5:
+			if (geo->level != 0)
+				change = CH_LEVEL_MIGRATION;
+			break;
+		case 10:
+			if (geo->level == 0) {
+				change = CH_TAKEOVER;
+				check_devs = 1;
+			}
+			break;
+		}
+		if (change == -1) {
+			fprintf(stderr, Name " Error. Level Migration from %d to %d "\
+				"not supported!\n", info.array.level, geo->level);
+			goto analyse_change_exit;
+		}
+	} else {
+		geo->level = info.array.level;
+	}
+
+	if ((geo->layout != info.array.layout)
+		&& ((geo->layout != UnSet) && (geo->layout != -1))) {
+		change = CH_LEVEL_MIGRATION;
+		if ((info.array.layout == 0) && (info.array.level == 5)
+			&& (geo->layout == 5)) {
+			/* reshape 5 -> 4 */
+		} else if ((info.array.layout == 5)
+				&& (info.array.level == 5)
+				&& (geo->layout == 0)) {
+			/* reshape 4 -> 5 */
+			geo->layout = 0;
+			geo->level = 5;
+		} else {
+			fprintf(stderr, Name " Error. Layout Migration from %d to %d "\
+				"not supported!\n", info.array.layout, geo->layout);
+			change = -1;
+			goto analyse_change_exit;
+		}
+	} else {
+		geo->layout = info.array.layout;
+	}
+
+	if ((geo->chunksize > 0) && (geo->chunksize != UnSet)
+		&& (geo->chunksize != info.array.chunk_size)) {
+		change = CH_CHUNK_MIGR;
+	} else {
+		geo->chunksize = info.array.chunk_size;
+	}
+
+	if (!validate_geometry_imsm(st,
+					geo->level,
+					geo->layout,
+					geo->raid_disks,
+					(geo->chunksize / 1024),
+					geo->size,
+					0, 0, 1))
+		change = -1;
+
+	if (check_devs) {
+		struct intel_super *super = st->sb;
+		struct imsm_super *mpb = super->anchor;
+
+		if (mpb->num_raid_devs > 1) {
+			fprintf(stderr, Name " Error. Cannot perform operation on %s"\
+				"- for this operation it MUST be single "\
+			       "array in container\n",
+				geo->dev_name);
+			change = -1;
+		}
+	}
+
+analyse_change_exit:
+
+	return change;
 }
 
 static int imsm_reshape_super(struct supertype *st, long long size, int level,
@@ -6649,18 +6743,37 @@ static int imsm_reshape_super(struct supertype *st, long long size, int level,
 		* - chunk size migration
 		* - migration: raid5 -> raid0; raid0 -> raid5
 		*/
-		int change;
+		struct intel_super *super = st->sb;
+		struct intel_dev *dev = super->devlist;
+		int change, devnum;
 		dprintf("imsm: info: Volume operation\n");
+		/* find requested device */
+		while (dev) {
+			imsm_find_array_minor_by_subdev(dev->index, st->container_dev, &devnum);
+			if (devnum == geo.dev_id)
+				break;
+			dev = dev->next;
+		}
+		if (dev == NULL) {
+			fprintf(stderr, Name " Cannot find %s (%i) subarray\n",
+					geo.dev_name, geo.dev_id);
+				goto exit_imsm_reshape_super;
+		}
+		super->current_vol = dev->index;
 		change = imsm_analyze_change(st, &geo);
 		switch (change) {
 			case CH_TAKEOVER:
-			break;
+				ret_val = 0;
+				break;
 			case CH_CHUNK_MIGR:
-			break;
+				ret_val = 0;
+				break;
 			case CH_LEVEL_MIGRATION:
-			break;
+				ret_val = 0;
+				break;
+			default:
+				ret_val = 1;
 		}
-		ret_val = 0;
 	}
 
 exit_imsm_reshape_super:

Re: [PATCH 1/6] Define imsm_analyze_change function

[NeilBrown]

On Thu, 13 Jan 2011 16:32:27 +0100 Krzysztof Wojcik
<krzysztof.wojcik@intel.com> wrote:

> Function intended to use for single volume migration.
> Function analyze transition and validate if it is supported.
> 
> Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com>

Applied, thanks.
Though I reformatted it a bit.

NeilBrown


> ---
>  super-intel.c |  135 ++++++++++++++++++++++++++++++++++++++++++++++++++++-----
>  1 files changed, 124 insertions(+), 11 deletions(-)
> 
> diff --git a/super-intel.c b/super-intel.c
> index 4cbb070..c12ba74 100644
> --- a/super-intel.c
> +++ b/super-intel.c
> @@ -6576,16 +6576,110 @@ static void imsm_update_metadata_locally(struct supertype *st,
>  	}
>  }
>  
> -/*******************************************************************************
> +/***************************************************************************
>  * Function:	imsm_analyze_change
> -* Description:	Function analyze and validate change for single volume migration
> +* Description:	Function analyze change for single volume
> +* 		and validate if transition is supported
>  * Parameters:	Geometry parameters, supertype structure
>  * Returns:	Operation type code on success, -1 if fail
> -********************************************************************************/
> -enum imsm_reshape_type imsm_analyze_change(
> -	struct supertype *st, struct geo_params *geo)
> +****************************************************************************/
> +enum imsm_reshape_type imsm_analyze_change(struct supertype *st,
> +						struct geo_params *geo)
>  {
> -	return -1;
> +	struct mdinfo info;
> +	int change = -1;
> +	int check_devs = 0;
> +
> +	getinfo_super_imsm_volume(st, &info, NULL);
> +
> +	if ((geo->level != info.array.level) &&
> +		(geo->level >= 0) &&
> +		(geo->level != UnSet)) {
> +		switch (info.array.level) {
> +		case 0:
> +			if (geo->level == 5) {
> +				change = CH_LEVEL_MIGRATION;
> +				check_devs = 1;
> +			}
> +			if (geo->level == 10) {
> +				change = CH_TAKEOVER;
> +				check_devs = 1;
> +			}
> +			break;
> +		case 5:
> +			if (geo->level != 0)
> +				change = CH_LEVEL_MIGRATION;
> +			break;
> +		case 10:
> +			if (geo->level == 0) {
> +				change = CH_TAKEOVER;
> +				check_devs = 1;
> +			}
> +			break;
> +		}
> +		if (change == -1) {
> +			fprintf(stderr, Name " Error. Level Migration from %d to %d "\
> +				"not supported!\n", info.array.level, geo->level);
> +			goto analyse_change_exit;
> +		}
> +	} else {
> +		geo->level = info.array.level;
> +	}
> +
> +	if ((geo->layout != info.array.layout)
> +		&& ((geo->layout != UnSet) && (geo->layout != -1))) {
> +		change = CH_LEVEL_MIGRATION;
> +		if ((info.array.layout == 0) && (info.array.level == 5)
> +			&& (geo->layout == 5)) {
> +			/* reshape 5 -> 4 */
> +		} else if ((info.array.layout == 5)
> +				&& (info.array.level == 5)
> +				&& (geo->layout == 0)) {
> +			/* reshape 4 -> 5 */
> +			geo->layout = 0;
> +			geo->level = 5;
> +		} else {
> +			fprintf(stderr, Name " Error. Layout Migration from %d to %d "\
> +				"not supported!\n", info.array.layout, geo->layout);
> +			change = -1;
> +			goto analyse_change_exit;
> +		}
> +	} else {
> +		geo->layout = info.array.layout;
> +	}
> +
> +	if ((geo->chunksize > 0) && (geo->chunksize != UnSet)
> +		&& (geo->chunksize != info.array.chunk_size)) {
> +		change = CH_CHUNK_MIGR;
> +	} else {
> +		geo->chunksize = info.array.chunk_size;
> +	}
> +
> +	if (!validate_geometry_imsm(st,
> +					geo->level,
> +					geo->layout,
> +					geo->raid_disks,
> +					(geo->chunksize / 1024),
> +					geo->size,
> +					0, 0, 1))
> +		change = -1;
> +
> +	if (check_devs) {
> +		struct intel_super *super = st->sb;
> +		struct imsm_super *mpb = super->anchor;
> +
> +		if (mpb->num_raid_devs > 1) {
> +			fprintf(stderr, Name " Error. Cannot perform operation on %s"\
> +				"- for this operation it MUST be single "\
> +			       "array in container\n",
> +				geo->dev_name);
> +			change = -1;
> +		}
> +	}
> +
> +analyse_change_exit:
> +
> +	return change;
>  }
>  
>  static int imsm_reshape_super(struct supertype *st, long long size, int level,
> @@ -6649,18 +6743,37 @@ static int imsm_reshape_super(struct supertype *st, long long size, int level,
>  		* - chunk size migration
>  		* - migration: raid5 -> raid0; raid0 -> raid5
>  		*/
> -		int change;
> +		struct intel_super *super = st->sb;
> +		struct intel_dev *dev = super->devlist;
> +		int change, devnum;
>  		dprintf("imsm: info: Volume operation\n");
> +		/* find requested device */
> +		while (dev) {
> +			imsm_find_array_minor_by_subdev(dev->index, st->container_dev, &devnum);
> +			if (devnum == geo.dev_id)
> +				break;
> +			dev = dev->next;
> +		}
> +		if (dev == NULL) {
> +			fprintf(stderr, Name " Cannot find %s (%i) subarray\n",
> +					geo.dev_name, geo.dev_id);
> +				goto exit_imsm_reshape_super;
> +		}
> +		super->current_vol = dev->index;
>  		change = imsm_analyze_change(st, &geo);
>  		switch (change) {
>  			case CH_TAKEOVER:
> -			break;
> +				ret_val = 0;
> +				break;
>  			case CH_CHUNK_MIGR:
> -			break;
> +				ret_val = 0;
> +				break;
>  			case CH_LEVEL_MIGRATION:
> -			break;
> +				ret_val = 0;
> +				break;
> +			default:
> +				ret_val = 1;
>  		}
> -		ret_val = 0;
>  	}
>  
>  exit_imsm_reshape_super:

[PATCH 2/6] Set reshape.after.data_disks for raid0<->raid10 takeover

[Krzysztof Wojcik]

reshape.after.data_disks field must be initiated
for raid0<->rai10 transition.
Instead calculated spares_needed variable in reshape_array
function has random value.

Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com>
---
 Grow.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/Grow.c b/Grow.c
index 763287b..47a8fb8 100644
--- a/Grow.c
+++ b/Grow.c
@@ -966,6 +966,7 @@ char *analyse_change(struct mdinfo *info, struct reshape *re)
 		re->level = 0;
 		re->parity = 0;
 		re->before.data_disks = new_disks;
+		re->after.data_disks = re->before.data_disks;
 		re->before.layout = 0;
 		re->backup_blocks = 0;
 		return NULL;
@@ -1003,6 +1004,7 @@ char *analyse_change(struct mdinfo *info, struct reshape *re)
 			re->parity = 0;
 			re->before.data_disks = (info->array.raid_disks +
 						 info->delta_disks);
+			re->after.data_disks = re->before.data_disks;
 			re->before.layout = info->new_layout;
 			re->backup_blocks = 0;
 			return NULL;

Re: [PATCH 2/6] Set reshape.after.data_disks for raid0<->raid10 takeover

[NeilBrown]

On Thu, 13 Jan 2011 16:32:35 +0100 Krzysztof Wojcik
<krzysztof.wojcik@intel.com> wrote:

> reshape.after.data_disks field must be initiated
> for raid0<->rai10 transition.
> Instead calculated spares_needed variable in reshape_array
> function has random value.
> 
> Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com>
> ---
>  Grow.c |    2 ++
>  1 files changed, 2 insertions(+), 0 deletions(-)
> 
> diff --git a/Grow.c b/Grow.c
> index 763287b..47a8fb8 100644
> --- a/Grow.c
> +++ b/Grow.c
> @@ -966,6 +966,7 @@ char *analyse_change(struct mdinfo *info, struct reshape *re)
>  		re->level = 0;
>  		re->parity = 0;
>  		re->before.data_disks = new_disks;
> +		re->after.data_disks = re->before.data_disks;
>  		re->before.layout = 0;
>  		re->backup_blocks = 0;
>  		return NULL;
> @@ -1003,6 +1004,7 @@ char *analyse_change(struct mdinfo *info, struct reshape *re)
>  			re->parity = 0;
>  			re->before.data_disks = (info->array.raid_disks +
>  						 info->delta_disks);
> +			re->after.data_disks = re->before.data_disks;
>  			re->before.layout = info->new_layout;
>  			re->backup_blocks = 0;
>  			return NULL;


apply, thanks.

NeilBrown

[PATCH 3/6] Unfreeze for non re-striping transitions

[Krzysztof Wojcik]

For non re-striping transitions array must be unfrozen
before end of processing.

Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com>
---
 Grow.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/Grow.c b/Grow.c
index 47a8fb8..c0f9419 100644
--- a/Grow.c
+++ b/Grow.c
@@ -1751,7 +1751,7 @@ static int reshape_array(char *container, int cfd, int fd, char *devname,
 				printf("chunk size for %s set to %d\n",
 				       devname, info->array.chunk_size);
 		}
-
+		unfreeze(st);
 		return rv;
 	}
 

Re: [PATCH 3/6] Unfreeze for non re-striping transitions

[NeilBrown]

On Thu, 13 Jan 2011 16:32:44 +0100 Krzysztof Wojcik
<krzysztof.wojcik@intel.com> wrote:

> For non re-striping transitions array must be unfrozen
> before end of processing.
> 
> Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com>
> ---
>  Grow.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/Grow.c b/Grow.c
> index 47a8fb8..c0f9419 100644
> --- a/Grow.c
> +++ b/Grow.c
> @@ -1751,7 +1751,7 @@ static int reshape_array(char *container, int cfd, int fd, char *devname,
>  				printf("chunk size for %s set to %d\n",
>  				       devname, info->array.chunk_size);
>  		}
> -
> +		unfreeze(st);
>  		return rv;
>  	}
>  
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html


I already had an unfreeze in there, condition on 'forked', how I think yours
is more correct - I don't want the condition there.

Applied thanks,

NeilBrown

[PATCH 4/6] Add raid10 -> raid0 takeover support

[Krzysztof Wojcik]

The patch introduces takeover form level 10 to level 0 for imsm
metadata. This patch contains procedures connected with preparing
and applying metadata update during 10 -> 0 takeover.
When performing takeover 10->0 mdmon should update the external
metadata (due to disk slot and level changes).
To achieve that mdadm calls reshape_super() and prepare
the "update_takeover" metadata update type.
Prepared update is processed by mdmon in process_update().

Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com>
---
 Grow.c        |    1 
 super-intel.c |  126 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 126 insertions(+), 1 deletions(-)

diff --git a/Grow.c b/Grow.c
index c0f9419..e1ec01c 100644
--- a/Grow.c
+++ b/Grow.c
@@ -1463,6 +1463,7 @@ int Grow_reshape(char *devname, int fd, int quiet, char *backup_file,
 			rv = 1;
 			goto release;
 		}
+		ping_monitor(container);
 	}
 
 	info.array = array;
diff --git a/super-intel.c b/super-intel.c
index c12ba74..7d2d777 100644
--- a/super-intel.c
+++ b/super-intel.c
@@ -299,6 +299,7 @@ enum imsm_update_type {
 	update_rename_array,
 	update_add_remove_disk,
 	update_reshape_container_disks,
+	update_takeover
 };
 
 struct imsm_update_activate_spare {
@@ -319,6 +320,17 @@ struct geo_params {
 	int raid_disks;
 };
 
+#define TAKEOVER_DISKS 2
+enum takeover_direction {
+	R10_TO_R0,
+	R0_TO_R10
+};
+struct imsm_update_takeover {
+	enum imsm_update_type type;
+	int subarray;
+	enum takeover_direction direction;
+	int disks[TAKEOVER_DISKS];
+};
 
 struct imsm_update_reshape {
 	enum imsm_update_type type;
@@ -5793,6 +5805,56 @@ update_reshape_exit:
 	return ret_val;
 }
 
+static int apply_takeover_update(struct imsm_update_takeover *u,
+				struct intel_super *super)
+{
+	struct imsm_dev *dev = NULL;
+	struct imsm_map *map;
+	struct dl *dm, *du;
+	int *tab;
+	int i;
+
+	dev = get_imsm_dev(super, u->subarray);
+
+	if (dev == NULL)
+		return 0;
+
+	map = get_imsm_map(dev, 0);
+	tab = (int *)&map->disk_ord_tbl;
+
+	if (u->direction == R10_TO_R0) {
+		/* iterate through devices to mark removed disks as spare */
+		for (i = 0; i < TAKEOVER_DISKS; i++) {
+			for (dm = super->disks; dm; dm = dm->next) {
+				if (((unsigned int)dm->major != major(u->disks[i])) ||
+				    ((unsigned int)dm->minor != minor(u->disks[i])))
+					continue;
+				for (du = super->disks; du; du = du->next)
+				    if ((du->index > dm->index) && (du->index > 0))
+					du->index--;
+				dm->disk.status = SPARE_DISK;
+				dm->index = -1;
+			}
+		}
+		/* update disk order table */
+		i = 0;
+		for (du = super->disks; du; du = du->next) {
+			if (du->index >= 0) {
+				tab[du->index] = i;
+				i++;
+			}
+		}
+		/* update map */
+		map->num_members = 2;
+		map->map_state = IMSM_T_STATE_NORMAL;
+		map->num_domains = 1;
+		map->raid_level = 0;
+		map->failed_disk_num = -1;
+	}
+
+	return 1;
+}
+
 static void imsm_process_update(struct supertype *st,
 			        struct metadata_update *update)
 {
@@ -5835,6 +5897,13 @@ static void imsm_process_update(struct supertype *st,
 	mpb = super->anchor;
 
 	switch (type) {
+	case update_takeover: {
+		struct imsm_update_takeover *u = (void *)update->buf;
+		if (apply_takeover_update(u, super))
+			super->updates_pending++;
+		break;
+	}
+
 	case update_reshape_container_disks: {
 		struct imsm_update_reshape *u = (void *)update->buf;
 		if (apply_reshape_container_disks_update(
@@ -6682,6 +6751,61 @@ analyse_change_exit:
 	return change;
 }
 
+int imsm_takeover(struct supertype *st, struct geo_params *geo)
+{
+	struct intel_super *super = st->sb;
+	struct imsm_update_takeover *u;
+	struct mdinfo *info;
+	struct mdinfo *newdi;
+	struct dl *dl;
+	int i, fd;
+	int found = 0;
+	char buf[PATH_MAX];
+
+	sprintf(buf, "/dev/md%i", geo->dev_id);
+	fd = open(buf, O_RDONLY);
+	if (!fd) {
+		fprintf(stderr, Name "Cannot open %s", buf);
+		return 1;
+	}
+	info = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
+	if (!info) {
+		fprintf(stderr, Name "Cannot load sysfs information for %s (%i)",
+			geo->dev_name, geo->dev_id);
+		return 1;
+	}
+
+	u = malloc(sizeof(struct imsm_update_takeover));
+	if (u == NULL)
+		return 1;
+
+	u->type = update_takeover;
+	u->subarray = super->current_vol;
+
+	/* 10->0 transition - mark disks to remove */
+	if (geo->level == 0) {
+		u->direction = R10_TO_R0;
+		i = 0;
+		for (dl = super->disks; dl; dl = dl->next) {
+			found = 0;
+			for (newdi = info->devs; newdi; newdi = newdi->next) {
+				if ((dl->major != newdi->disk.major) ||
+					    (dl->minor != newdi->disk.minor) ||
+					    (newdi->disk.raid_disk < 0))
+					continue;
+				found = 1;
+				break;
+			}
+			/* if disk not found, mark it for remove */
+			if ((found == 0) && (!(dl->disk.status & SPARE_DISK)))
+				u->disks[i++] = makedev(dl->major, dl->minor);
+		}
+	}
+
+	append_metadata_update(st, u, sizeof(struct imsm_update_takeover));
+	return 0;
+}
+
 static int imsm_reshape_super(struct supertype *st, long long size, int level,
 			      int layout, int chunksize, int raid_disks,
 			      char *backup, char *dev, int verbose)
@@ -6763,7 +6887,7 @@ static int imsm_reshape_super(struct supertype *st, long long size, int level,
 		change = imsm_analyze_change(st, &geo);
 		switch (change) {
 			case CH_TAKEOVER:
-				ret_val = 0;
+				ret_val = imsm_takeover(st, &geo);
 				break;
 			case CH_CHUNK_MIGR:
 				ret_val = 0;

Re: [PATCH 4/6] Add raid10 -> raid0 takeover support

[NeilBrown]

On Thu, 13 Jan 2011 16:32:52 +0100 Krzysztof Wojcik
<krzysztof.wojcik@intel.com> wrote:

> The patch introduces takeover form level 10 to level 0 for imsm
> metadata. This patch contains procedures connected with preparing
> and applying metadata update during 10 -> 0 takeover.
> When performing takeover 10->0 mdmon should update the external
> metadata (due to disk slot and level changes).
> To achieve that mdadm calls reshape_super() and prepare
> the "update_takeover" metadata update type.
> Prepared update is processed by mdmon in process_update().

You are using sysfs_read inside super-intel again.  That is the wrong thing
to do for the same reasons as last time.

And I'm not very comfortable about sending a list of devices to delete from
mdadm to mdmon.
I would think that mdadm would simply fail/remove the devices from the array,
then tell mdmon to update the metadata.  It would see which ones are still
working and construct the RAID0 out of those.

NeilBrown


> 
> Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com>
> ---
>  Grow.c        |    1 
>  super-intel.c |  126 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>  2 files changed, 126 insertions(+), 1 deletions(-)
> 
> diff --git a/Grow.c b/Grow.c
> index c0f9419..e1ec01c 100644
> --- a/Grow.c
> +++ b/Grow.c
> @@ -1463,6 +1463,7 @@ int Grow_reshape(char *devname, int fd, int quiet, char *backup_file,
>  			rv = 1;
>  			goto release;
>  		}
> +		ping_monitor(container);
>  	}
>  
>  	info.array = array;
> diff --git a/super-intel.c b/super-intel.c
> index c12ba74..7d2d777 100644
> --- a/super-intel.c
> +++ b/super-intel.c
> @@ -299,6 +299,7 @@ enum imsm_update_type {
>  	update_rename_array,
>  	update_add_remove_disk,
>  	update_reshape_container_disks,
> +	update_takeover
>  };
>  
>  struct imsm_update_activate_spare {
> @@ -319,6 +320,17 @@ struct geo_params {
>  	int raid_disks;
>  };
>  
> +#define TAKEOVER_DISKS 2
> +enum takeover_direction {
> +	R10_TO_R0,
> +	R0_TO_R10
> +};
> +struct imsm_update_takeover {
> +	enum imsm_update_type type;
> +	int subarray;
> +	enum takeover_direction direction;
> +	int disks[TAKEOVER_DISKS];
> +};
>  
>  struct imsm_update_reshape {
>  	enum imsm_update_type type;
> @@ -5793,6 +5805,56 @@ update_reshape_exit:
>  	return ret_val;
>  }
>  
> +static int apply_takeover_update(struct imsm_update_takeover *u,
> +				struct intel_super *super)
> +{
> +	struct imsm_dev *dev = NULL;
> +	struct imsm_map *map;
> +	struct dl *dm, *du;
> +	int *tab;
> +	int i;
> +
> +	dev = get_imsm_dev(super, u->subarray);
> +
> +	if (dev == NULL)
> +		return 0;
> +
> +	map = get_imsm_map(dev, 0);
> +	tab = (int *)&map->disk_ord_tbl;
> +
> +	if (u->direction == R10_TO_R0) {
> +		/* iterate through devices to mark removed disks as spare */
> +		for (i = 0; i < TAKEOVER_DISKS; i++) {
> +			for (dm = super->disks; dm; dm = dm->next) {
> +				if (((unsigned int)dm->major != major(u->disks[i])) ||
> +				    ((unsigned int)dm->minor != minor(u->disks[i])))
> +					continue;
> +				for (du = super->disks; du; du = du->next)
> +				    if ((du->index > dm->index) && (du->index > 0))
> +					du->index--;
> +				dm->disk.status = SPARE_DISK;
> +				dm->index = -1;
> +			}
> +		}
> +		/* update disk order table */
> +		i = 0;
> +		for (du = super->disks; du; du = du->next) {
> +			if (du->index >= 0) {
> +				tab[du->index] = i;
> +				i++;
> +			}
> +		}
> +		/* update map */
> +		map->num_members = 2;
> +		map->map_state = IMSM_T_STATE_NORMAL;
> +		map->num_domains = 1;
> +		map->raid_level = 0;
> +		map->failed_disk_num = -1;
> +	}
> +
> +	return 1;
> +}
> +
>  static void imsm_process_update(struct supertype *st,
>  			        struct metadata_update *update)
>  {
> @@ -5835,6 +5897,13 @@ static void imsm_process_update(struct supertype *st,
>  	mpb = super->anchor;
>  
>  	switch (type) {
> +	case update_takeover: {
> +		struct imsm_update_takeover *u = (void *)update->buf;
> +		if (apply_takeover_update(u, super))
> +			super->updates_pending++;
> +		break;
> +	}
> +
>  	case update_reshape_container_disks: {
>  		struct imsm_update_reshape *u = (void *)update->buf;
>  		if (apply_reshape_container_disks_update(
> @@ -6682,6 +6751,61 @@ analyse_change_exit:
>  	return change;
>  }
>  
> +int imsm_takeover(struct supertype *st, struct geo_params *geo)
> +{
> +	struct intel_super *super = st->sb;
> +	struct imsm_update_takeover *u;
> +	struct mdinfo *info;
> +	struct mdinfo *newdi;
> +	struct dl *dl;
> +	int i, fd;
> +	int found = 0;
> +	char buf[PATH_MAX];
> +
> +	sprintf(buf, "/dev/md%i", geo->dev_id);
> +	fd = open(buf, O_RDONLY);
> +	if (!fd) {
> +		fprintf(stderr, Name "Cannot open %s", buf);
> +		return 1;
> +	}
> +	info = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
> +	if (!info) {
> +		fprintf(stderr, Name "Cannot load sysfs information for %s (%i)",
> +			geo->dev_name, geo->dev_id);
> +		return 1;
> +	}
> +
> +	u = malloc(sizeof(struct imsm_update_takeover));
> +	if (u == NULL)
> +		return 1;
> +
> +	u->type = update_takeover;
> +	u->subarray = super->current_vol;
> +
> +	/* 10->0 transition - mark disks to remove */
> +	if (geo->level == 0) {
> +		u->direction = R10_TO_R0;
> +		i = 0;
> +		for (dl = super->disks; dl; dl = dl->next) {
> +			found = 0;
> +			for (newdi = info->devs; newdi; newdi = newdi->next) {
> +				if ((dl->major != newdi->disk.major) ||
> +					    (dl->minor != newdi->disk.minor) ||
> +					    (newdi->disk.raid_disk < 0))
> +					continue;
> +				found = 1;
> +				break;
> +			}
> +			/* if disk not found, mark it for remove */
> +			if ((found == 0) && (!(dl->disk.status & SPARE_DISK)))
> +				u->disks[i++] = makedev(dl->major, dl->minor);
> +		}
> +	}
> +
> +	append_metadata_update(st, u, sizeof(struct imsm_update_takeover));
> +	return 0;
> +}
> +
>  static int imsm_reshape_super(struct supertype *st, long long size, int level,
>  			      int layout, int chunksize, int raid_disks,
>  			      char *backup, char *dev, int verbose)
> @@ -6763,7 +6887,7 @@ static int imsm_reshape_super(struct supertype *st, long long size, int level,
>  		change = imsm_analyze_change(st, &geo);
>  		switch (change) {
>  			case CH_TAKEOVER:
> -				ret_val = 0;
> +				ret_val = imsm_takeover(st, &geo);
>  				break;
>  			case CH_CHUNK_MIGR:
>  				ret_val = 0;

[PATCH 5/6] raid0->raid10 takeover- create metadata update

[Krzysztof Wojcik]

Create metadata update for raid0 -> raid10 takeover.
Because we have no mdmon running for raid0 we have to
update metadata using local update mechanism

Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com>
---
 super-intel.c |   14 +++++++++++++-
 1 files changed, 13 insertions(+), 1 deletions(-)

diff --git a/super-intel.c b/super-intel.c
index 7d2d777..63bd2d2 100644
--- a/super-intel.c
+++ b/super-intel.c
@@ -6802,7 +6802,19 @@ int imsm_takeover(struct supertype *st, struct geo_params *geo)
 		}
 	}
 
-	append_metadata_update(st, u, sizeof(struct imsm_update_takeover));
+	/* 0->10 transition- nothing to do here
+	 * all work will be done in prepare_update, process_update */
+	if (geo->level == 10)
+		u->direction = R0_TO_R10;
+
+	/* update metadata locally */
+	imsm_update_metadata_locally(st, u, sizeof(struct imsm_update_takeover));
+	/* and possibly remotely */
+	if (st->update_tail)
+		append_metadata_update(st, u, sizeof(struct imsm_update_takeover));
+	else
+		free(u);
+
 	return 0;
 }
 

[PATCH 6/6] raid0->raid10 takeover- allocate memory for added disks

[Krzysztof Wojcik]

Allocate memory necessary to add new dummy disks
in process_update.

Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com>
---
 super-intel.c |   18 ++++++++++++++++++
 1 files changed, 18 insertions(+), 0 deletions(-)

diff --git a/super-intel.c b/super-intel.c
index 63bd2d2..a7d7d08 100644
--- a/super-intel.c
+++ b/super-intel.c
@@ -6224,6 +6224,24 @@ static void imsm_prepare_update(struct supertype *st,
 	size_t len = 0;
 
 	switch (type) {
+	case update_takeover: {
+		struct imsm_update_takeover *u = (void *)update->buf;
+		if (u->direction == R0_TO_R10) {
+			void **tail = (void**)&update->space_list;
+			void *space;
+			int size = sizeof_imsm_dev(super->devlist->dev, 0) +
+				(TAKEOVER_DISKS * sizeof(__u32));
+			space = malloc(size);
+			if (!space)
+				break;
+			*tail = space;
+			tail = space;
+			*tail = NULL;
+
+			len = disks_to_mpb_size(TAKEOVER_DISKS * 2);
+		}
+		break;
+	}
 	case update_reshape_container_disks: {
 		/* Every raid device in the container is about to
 		 * gain some more devices, and we will enter a