Jason,
thanks.
Will integrate and test.
Can you send me some metadata samples (in personal email)
created with "dmraid -rD" for referenece, please ?
Heinz
On Thu, Jun 28, 2007 at 04:53:39PM -0700, Jason Gaston wrote:
> This patch adds raid10(0+1) nested raid level support to the Intel isw
> metadata format handler in dmraid 1.0.0.rc14.
>
> Signed-off-by: Jason Gaston <jason.d.gaston@xxxxxxxxx>
>
> --- 1.0.0.rc14/lib/format/ataraid/isw.c.orig 2006-09-22 11:11:37.000000000
> -0700
> +++ 1.0.0.rc14/lib/format/ataraid/isw.c 2007-06-28 14:06:31.000000000 -0700
> @@ -25,30 +25,59 @@
> #endif
>
> static const char *handler = HANDLER;
> +static struct isw_disk *_get_disk(struct isw *isw, struct dev_info *di);
> +int is_raid10(struct isw *isw);
>
> -/*
> - * Make up RAID set name from family_num and volume name.
> - */
> -static size_t _name(struct isw *isw, struct isw_dev *dev,
> - char *str, size_t len)
> +static size_t _name(struct isw *isw, char *str, size_t len, int subset,
> + int num, struct isw_dev *dev)
> {
> - return snprintf(str, len, dev ? "isw_%u_%s" : "isw_%u",
> + switch(subset) {
> + case 2:
> + return snprintf(str, len, "isw_%u", isw->family_num);
> + case 1:
> + if(!is_raid10(isw))
> + return snprintf(str, len, "isw_%u_%s",
> isw->family_num, (char*) dev->volume);
> -}
> + else
> + return snprintf(str, len, "isw_%u_%s-%u",
> + isw->family_num, (char*) dev->volume, num);
> + case 0:
> + return snprintf(str, len, "isw_%u_%s", isw->family_num,
> + (char*) dev->volume);
> + }
> + return 0;
> +}
> +
> +static char *name(struct lib_context *lc, struct raid_dev *rd,
> + unsigned int subset)
> +{
> + size_t len;
> + char *ret = NULL;
> + int id = 0;
> + struct isw *isw = META(rd, isw);
> + struct isw_disk *disk = isw->disk;
> + struct isw_dev *dev = (struct isw_dev*) (isw->disk + isw->num_disks);
>
> -static char *name(struct lib_context *lc, struct isw *isw, struct isw_dev
> *dev)
> -{
> - size_t len;
> - char *ret;
> + if((subset == 1) && (is_raid10(isw))) {
> + if ((disk = _get_disk(isw, rd->di))) {
> + if(disk == isw->disk) id = 0;
> + else if(disk == isw->disk + 1) id = 1;
> + else if(disk == isw->disk + 2) id = 2;
> + else if(disk == isw->disk + 3) id = 3;
> + else return ret;
> + }
> + id = id % 2;
> + }
>
> - if ((ret = dbg_malloc((len = _name(isw, dev, NULL, 0) + 1)))) {
> - _name(isw, dev, ret, len);
> - mk_alpha(lc, ret + HANDLER_LEN, len - HANDLER_LEN -
> - (dev ? strlen((char*) dev->volume) - 2 : 1));
> - } else
> + if ((ret = dbg_malloc((len = _name(isw, ret, 0, subset, id,
> + dev) + 1)))) {
> + _name(isw, ret, len, subset, id, dev);
> + mk_alpha(lc, ret + HANDLER_LEN, snprintf(ret, 0, "%u",
> + isw->family_num));
> + } else
> log_alloc_err(lc, handler);
>
> - return ret;
> + return ret;
> }
>
> /* Find a disk table slot by serial number. */
> @@ -59,7 +88,7 @@
>
> do {
> if (!strncmp(di->serial, (const char*) disk->serial,
> - MAX_RAID_SERIAL_LEN))
> + MAX_RAID_SERIAL_LEN))
> return disk;
> } while (++disk < isw->disk + isw->num_disks);
> }
> @@ -102,21 +131,38 @@
> }
>
> /* Neutralize disk type. */
> -static enum type type(struct raid_dev *rd)
> +static enum type type(struct isw *isw)
> {
> /* Mapping of Intel types to generic types. */
> static struct types types[] = {
> { ISW_T_RAID0, t_raid0},
> { ISW_T_RAID1, t_raid1},
> { ISW_T_RAID5, t_raid5_la},
> + { ISW_T_RAID10, t_raid0},
> { 0, t_undef},
> };
> - struct isw_dev *dev = rd->private.ptr;
> +
> + struct isw_dev *dev = (struct isw_dev*) (isw->disk + isw->num_disks);
> +
> + if(is_raid10(isw))
> + dev->vol.map.raid_level = ISW_T_RAID10;
>
> return dev ? rd_type(types, (unsigned int) dev->vol.map.raid_level) :
> t_group;
> }
>
> +int is_raid10(struct isw *isw)
> +{
> + int ret;
> + struct isw_dev *dev = (struct isw_dev*) (isw->disk + isw->num_disks);
> +
> + ret = ((dev->vol.map.raid_level == ISW_T_RAID10) ||
> + (dev->vol.map.raid_level == ISW_T_RAID1 &&
> + isw->num_disks > 3));
> +
> + return ret;
> +}
> +
> /*
> * Generate checksum of Raid metadata for mpb_size/sizeof(u32) words
> * (checksum field itself ignored for this calculation).
> @@ -298,7 +344,8 @@
> }
>
> /* Check for RAID disk ok. */
> -static int disk_ok(struct lib_context *lc, struct dev_info *di, struct isw
> *isw)
> +static int disk_ok(struct lib_context *lc, struct dev_info *di,
> + struct isw *isw)
> {
> struct isw_disk *disk = get_disk(lc, di, isw);
>
> @@ -306,15 +353,15 @@
> }
>
> static void *isw_read_metadata(struct lib_context *lc, struct dev_info *di,
> - size_t *sz, uint64_t *offset,
> - union read_info *info)
> + size_t *sz, uint64_t *offset,
> + union read_info *info)
> {
> size_t size = ISW_DISK_BLOCK_SIZE;
> uint64_t isw_sboffset = ISW_CONFIGOFFSET;
> struct isw *isw;
>
> if (!(isw = alloc_private_and_read(lc, handler, size,
> - di->path, isw_sboffset)))
> + di->path, isw_sboffset)))
> goto out;
>
> /*
> @@ -391,7 +438,7 @@
> */
> /* Check state if isw device map. */
> static int _check_map_state(struct lib_context *lc, struct raid_dev *rd,
> - struct isw_dev *dev)
> + struct isw_dev *dev)
> {
> /* FIXME: FAILED_MAP etc. */
> switch (dev->vol.map.map_state) {
> @@ -415,20 +462,27 @@
> struct raid_dev *r;
>
> if (!_check_map_state(lc, rd, dev) ||
> - !(r = alloc_raid_dev(lc, handler)))
> + !(r = alloc_raid_dev(lc, handler)))
> return NULL;
>
> if (!(r->private.ptr = alloc_private(lc, handler, sizeof(*dev))))
> goto free;
> -
> memcpy(r->private.ptr, dev, sizeof(*dev));
> - if ((r->type = type(r)) == t_undef) {
> +
> + if (!(r->meta_areas = alloc_meta_areas(lc, rd, handler, 1)))
> + return 0;
> +
> + r->meta_areas->offset = rd->meta_areas->offset;
> + r->meta_areas->size = rd->meta_areas->size;
> + r->meta_areas->area = rd->meta_areas->area;
> +
> + if ((r->type = type(isw)) == t_undef) {
> log_err(lc, "%s: RAID type %u not supported",
> handler, (unsigned int) dev->vol.map.raid_level);
> goto free;
> }
>
> - if (!(r->name = name(lc, isw, dev)))
> + if (!(r->name = name(lc, rd, 1)))
> goto free;
>
> r->di = rd->di;
> @@ -460,43 +514,73 @@
> return _get_disk(isw, RD(new)->di) < _get_disk(isw, RD(pos)->di);
> }
>
> +static void super_created(struct raid_set *super, void *private)
> +{
> + super->type = t_raid1;
> + super->stride = ((struct isw_dev*) private)->vol.map.blocks_per_strip;
> +}
> +
> +static int set_sort(struct list_head *pos, struct list_head *new)
> +{
> + return 0;
> +}
> +
> /*
> * rs_group contains the top-level group RAID set (type: t_group) on entry
> * and shall be returned on success (or NULL on error).
> */
> static struct raid_set *group_rd(struct lib_context *lc,
> - struct raid_set *rs_group,
> - struct raid_dev *rd_meta)
> + struct raid_set *rs_group,
> + struct raid_dev *rd_meta)
> {
> unsigned int d;
> void *private;
> struct isw *isw = META(rd_meta, isw);
> struct isw_dev *dev;
> struct raid_dev *rd;
> - struct raid_set *rs;
> -
> + struct raid_set *rs, *ss;
> + char *nm = NULL;
> +
> /* Loop the device/volume table. */
> for (d = 0; d < isw->num_raid_devs; d++) {
> dev = raiddev(isw, d);
> -
> +
> if (!(rd = _create_rd(lc, rd_meta, isw, dev)))
> return NULL;
>
> - if (!(rs = find_or_alloc_raid_set(lc, rd->name, FIND_ALL,
> - rd, &rs_group->sets,
> - create_rs, dev))) {
> - free_raid_dev(lc, &rd);
> - return NULL;
> + if(is_raid10(isw)) {
> + nm = name(lc, rd, 0);
> + ss = find_or_alloc_raid_set(lc, nm, FIND_TOP, rd,
> + LC_RS(lc), super_created, dev);
> + if (!(rs = find_or_alloc_raid_set(lc, rd->name,
> + FIND_ALL, rd, &ss->sets, create_rs,
> + dev))) {
> + free_raid_dev(lc, &rd);
> + return NULL;
> + }
> + } else {
> + if (!(rs = find_or_alloc_raid_set(lc, rd->name,
> + FIND_ALL, rd, &rs_group->sets,
> + create_rs, dev))) {
> + free_raid_dev(lc, &rd);
> + return NULL;
> + }
> }
>
> + rs->status = s_ok;
> +
> /* Save and set to enable dev_sort(). */
> private = rd->private.ptr;
> rd->private.ptr = isw;
> -
> list_add_sorted(lc, &rs->devs, &rd->devs, dev_sort);
> -
> /* Restore. */
> rd->private.ptr = private;
> +
> + if(is_raid10(isw)) {
> + ss = join_superset(lc, name, super_created, set_sort,
> + rs, rd);
> + return ss;
> + }
> }
>
> return rs_group;
> @@ -504,31 +588,35 @@
>
> /* Add an Intel SW RAID device to a set */
> static struct raid_set *isw_group(struct lib_context *lc,
> - struct raid_dev *rd_meta)
> + struct raid_dev *rd_meta)
> {
> - struct raid_set *rs_group;
> + struct raid_set *rs_group = NULL;
> + struct isw *isw = META(rd_meta, isw);
>
> if (T_SPARE(rd_meta))
> return NULL;
>
> - /*
> - * Once we get here, an Intel SW RAID disk containing a metadata area
> - * with a volume table has been discovered by isw_read.
> - */
> - /* Check if a top level group RAID set already exists. */
> - if (!(rs_group = find_or_alloc_raid_set(lc, rd_meta->name, FIND_TOP,
> - rd_meta, LC_RS(lc),
> - NO_CREATE, NO_CREATE_ARG)))
> - return NULL;
> + if(!is_raid10(isw)) {
> + /*
> + * Once we get here, an Intel SW RAID disk containing a metadata
> + * area with a volume table has been discovered by isw_read.
> + */
> + /* Check if a top level group RAID set already exists. */
> + if (!(rs_group = find_or_alloc_raid_set(lc, rd_meta->name,
> + FIND_TOP, rd_meta, LC_RS(lc),
> + NO_CREATE, NO_CREATE_ARG)))
> + return NULL;
>
> - /*
> - * Add the whole underlying (meta) RAID device to the group set.
> - * Sorting is no problem here, because RAID sets and devices will
> - * be created for all the Volumes of an ISW set and those need sorting.
> - */
> - rd_meta->private.ptr = rd_meta->meta_areas->area;
> - list_add_sorted(lc, &rs_group->devs, &rd_meta->devs, dev_sort);
> - rd_meta->private.ptr = NULL;
> + /*
> + * Add the whole underlying (meta) RAID device to the group set.
> + * Sorting is no problem here, because RAID sets and devices
> + * will be created for all the Volumes of an ISW set and those
> + * need sorting.
> + */
> + rd_meta->private.ptr = rd_meta->meta_areas->area;
> + list_add_sorted(lc, &rs_group->devs, &rd_meta->devs, dev_sort);
> + rd_meta->private.ptr = NULL;
> + }
>
> /*
> * We need to run through the volume table and create a RAID set and
> @@ -552,17 +640,16 @@
> }
>
> static int check_rd(struct lib_context *lc, struct raid_set *rs,
> - struct raid_dev *rd, void *context)
> + struct raid_dev *rd, void *context)
> {
> struct isw_dev *dev = rd->private.ptr;
>
> /* FIXME: more status checks ? */
> if (dev->status)
> LOG_ERR(lc, 0, "%s device for volume \"%s\" broken on %s "
> - "in RAID set \"%s\"",
> - handler, dev->volume, rd->di->path, rs->name);
> -
> - return 1;
> + "in RAID set \"%s\"", handler, dev->volume,
> + rd->di->path, rs->name);
> + return 1;
> }
>
> static int _isw_check(struct lib_context *lc, struct raid_set *rs)
> @@ -572,9 +659,14 @@
>
> static int isw_check(struct lib_context *lc, struct raid_set *rs)
> {
> - return T_GROUP(rs) ? _isw_check(lc, rs) : 0;
> + /* If it is a stacked set like RAID10 */
> + if((!T_GROUP(rs)) && SETS(rs)) {
> + return check_raid_set(lc, rs, devices, NULL,
> + NO_CHECK_RD, NULL, handler);
> + } else {
> + return T_GROUP(rs) ? _isw_check(lc, rs) : 0;
> + }
> }
> -
> /*
> * IO error event handler.
> */
> @@ -585,7 +677,7 @@
> struct isw_disk *disk;
>
> if (!(disk = get_disk(lc, rd->di, isw)))
> - LOG_ERR(lc, 0, "%s: disk", handler);
> + LOG_ERR(lc, 0, "%s: disk", handler);
>
> /* Avoid write trashing. */
> if (S_BROKEN(status(lc, rd)))
> @@ -629,7 +721,7 @@
> for (i = 0; i < ISW_FILLERS; i++) {
> if (isw->filler[i])
> P("filler[%i]: %u", isw,
> - isw->filler[i], i, isw->filler[i]);
> + isw->filler[i], i, isw->filler[i]);
> }
>
> /* Disk table. */
> @@ -744,7 +836,7 @@
> static struct dmraid_format isw_format = {
> .name = HANDLER,
> .descr = "Intel Software RAID",
> - .caps = "0,1",
> + .caps = "0,1,10",
> .format = FMT_RAID,
> .read = isw_read,
> .write = isw_write,
> @@ -797,5 +889,7 @@
> rd->status = status(lc, rd);
> rd->type = t_group;
>
> - return (rd->name = name(lc, isw, NULL)) ? 1 : 0;
> + return (rd->name = name(lc, rd, 2)) ? 1 : 0;
> }
> +
> +
> --- 1.0.0.rc14/lib/format/ataraid/isw.h.orig 2006-01-24 07:02:54.000000000
> -0800
> +++ 1.0.0.rc14/lib/format/ataraid/isw.h 2007-06-27 15:24:19.000000000 -0700
> @@ -68,6 +68,7 @@
> uint8_t raid_level;
> #define ISW_T_RAID0 0
> #define ISW_T_RAID1 1
> +#define ISW_T_RAID10 2
> #define ISW_T_RAID5 5 // since metadata version 1.2.02 ?
> uint8_t num_members; // number of member disks
> uint8_t reserved[3];
--
Regards,
Heinz -- The LVM Guy --
*** Software bugs are stupid.
Nevertheless it needs not so stupid people to solve them ***
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Heinz Mauelshagen Red Hat GmbH
Consulting Development Engineer Am Sonnenhang 11
Storage Development 56242 Marienrachdorf
Germany
Mauelshagen@xxxxxxxxxx PHONE +49 171 7803392
FAX +49 2626 924446
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
_______________________________________________
Ataraid-list mailing list
Ataraid-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/ataraid-list
