#include "mdmon.h"
#include "sha1.h"
#include <values.h>
+#include <stddef.h>
/* a non-official T10 name for creation GUIDs */
static char T10[] = "Linux-MD";
* for concat I hope) */
be64 array_blocks; /* blocks in array */
__u8 pad1[8];
- be32 spare_refs[8];
+ be32 spare_refs[8]; /* This is used to detect missing spares.
+ * As we don't have an interface for that
+ * the values are ignored.
+ */
__u8 cache_pol[8];
__u8 bg_rate;
__u8 pad2[3];
char *name, char *homehost,
int *uuid, unsigned long long data_offset);
-#ifndef offsetof
-#define offsetof(t,f) ((size_t)&(((t*)0)->f))
-#endif
-
#if DEBUG
static void pr_state(struct ddf_super *ddf, const char *msg)
{
static void pr_state(const struct ddf_super *ddf, const char *msg) {}
#endif
-static void _ddf_set_updates_pending(struct ddf_super *ddf, const char *func)
+static void _ddf_set_updates_pending(struct ddf_super *ddf, struct vd_config *vc,
+ const char *func)
{
+ if (vc) {
+ vc->timestamp = cpu_to_be32(time(0)-DECADE);
+ vc->seqnum = cpu_to_be32(be32_to_cpu(vc->seqnum) + 1);
+ }
if (ddf->updates_pending)
return;
ddf->updates_pending = 1;
pr_state(ddf, func);
}
-#define ddf_set_updates_pending(x) _ddf_set_updates_pending((x), __func__)
+#define ddf_set_updates_pending(x,v) _ddf_set_updates_pending((x), (v), __func__)
static be32 calc_crc(void *buf, int len)
{
rlq = DDF_RAID1_SIMPLE;
prim_elmnt_count = cpu_to_be16(2);
sec_elmnt_count = array->raid_disks / 2;
+ srl = DDF_2SPANNED;
+ prl = DDF_RAID1;
} else if (array->raid_disks % 3 == 0
&& array->layout == 0x103) {
rlq = DDF_RAID1_MULTI;
prim_elmnt_count = cpu_to_be16(3);
sec_elmnt_count = array->raid_disks / 3;
+ srl = DDF_2SPANNED;
+ prl = DDF_RAID1;
+ } else if (array->layout == 0x201) {
+ prl = DDF_RAID1E;
+ rlq = DDF_RAID1E_OFFSET;
+ } else if (array->layout == 0x102) {
+ prl = DDF_RAID1E;
+ rlq = DDF_RAID1E_ADJACENT;
} else
return err_bad_md_layout(array);
- srl = DDF_2SPANNED;
- prl = DDF_RAID1;
break;
default:
return err_bad_md_layout(array);
return err_bad_ddf_layout(conf);
level = 1;
break;
+ case DDF_RAID1E:
+ if (conf->rlq == DDF_RAID1E_ADJACENT)
+ layout = 0x102;
+ else if (conf->rlq == DDF_RAID1E_OFFSET)
+ layout = 0x201;
+ else
+ return err_bad_ddf_layout(conf);
+ level = 10;
+ break;
case DDF_RAID4:
if (conf->rlq != DDF_RAID4_N)
return err_bad_ddf_layout(conf);
printf("%d", j);
else
printf("--");
+ printf("@%lluK", (unsigned long long) be64_to_cpu(LBA_OFFSET(sb, vc)[i])/2);
}
printf(")\n");
if (vc->chunk_shift != 255)
static void detail_super_ddf(struct supertype *st, char *homehost)
{
- /* FIXME later
- * Could print DDF GUID
- * Need to find which array
- * If whole, briefly list all arrays
- * If one, give name
- */
+ struct ddf_super *sb = st->sb;
+ int cnt = be16_to_cpu(sb->virt->populated_vdes);
+
+ printf(" Container GUID : "); print_guid(sb->anchor.guid, 1);
+ printf("\n");
+ printf(" Seq : %08x\n", be32_to_cpu(sb->active->seq));
+ printf(" Virtual Disks : %d\n", cnt);
+ printf("\n");
}
+#endif
static const char *vendors_with_variable_volume_UUID[] = {
"LSI ",
memcpy(uuid, sha, 4*4);
}
+#ifndef MDASSEMBLE
static void brief_detail_super_ddf(struct supertype *st)
{
struct mdinfo info;
cptr = (__u32 *)(ddf->anchor.guid + 16);
info->array.ctime = DECADE + __be32_to_cpu(*cptr);
- info->array.utime = 0;
info->array.chunk_size = 0;
info->container_enough = 1;
if (info->disk.raid_disk >= 0)
pde = ddf->phys->entries + info->disk.raid_disk;
if (pde &&
- !(be16_to_cpu(pde->state) & DDF_Failed))
+ !(be16_to_cpu(pde->state) & DDF_Failed) &&
+ !(be16_to_cpu(pde->state) & DDF_Missing))
info->disk.state = (1 << MD_DISK_SYNC) | (1 << MD_DISK_ACTIVE);
else
info->disk.state = 1 << MD_DISK_FAULTY;
- info->events = be32_to_cpu(ddf->active->seq);
} else {
+ /* There should always be a dlist, but just in case...*/
info->disk.number = -1;
info->disk.raid_disk = -1;
-// info->disk.raid_disk = find refnum in the table and use index;
info->disk.state = (1 << MD_DISK_SYNC) | (1 << MD_DISK_ACTIVE);
}
+ info->events = be32_to_cpu(ddf->active->seq);
+ info->array.utime = DECADE + be32_to_cpu(ddf->active->timestamp);
info->recovery_start = MaxSector;
info->reshape_active = 0;
uuid_from_super_ddf(st, info->uuid);
if (map) {
- int i;
- for (i = 0 ; i < map_disks; i++) {
- if (i < info->array.raid_disks &&
- !(be16_to_cpu(ddf->phys->entries[i].state)
+ int i, e = 0;
+ int max = be16_to_cpu(ddf->phys->max_pdes);
+ for (i = e = 0 ; i < map_disks ; i++, e++) {
+ while (e < max &&
+ be32_to_cpu(ddf->phys->entries[e].refnum) == 0xffffffff)
+ e++;
+ if (i < info->array.raid_disks && e < max &&
+ !(be16_to_cpu(ddf->phys->entries[e].state)
& DDF_Failed))
map[i] = 1;
else
int cd = ddf->currentdev;
int n_prim;
int j;
- struct dl *dl;
+ struct dl *dl = NULL;
int map_disks = info->array.raid_disks;
__u32 *cptr;
struct vd_config *conf;
info->array.ctime = DECADE + __be32_to_cpu(*cptr);
info->array.utime = DECADE + be32_to_cpu(vc->conf.timestamp);
info->array.chunk_size = 512 << vc->conf.chunk_shift;
- info->custom_array_size = 0;
+ info->custom_array_size = be64_to_cpu(vc->conf.array_blocks);
conf = &vc->conf;
n_prim = be16_to_cpu(conf->prim_elmnt_count);
struct phys_disk *pd;
struct virtual_disk *vd;
- if (data_offset != INVALID_SECTORS) {
- pr_err("data-offset not supported by DDF\n");
- return 0;
- }
-
if (st->sb)
return init_super_ddf_bvd(st, info, size, name, homehost, uuid,
data_offset);
memset(&vd->entries[i], 0xff, sizeof(struct virtual_entry));
st->sb = ddf;
- ddf_set_updates_pending(ddf);
+ ddf_set_updates_pending(ddf, NULL);
return 1;
}
static struct extent *get_extents(struct ddf_super *ddf, struct dl *dl)
{
- /* Find a list of used extents on the give physical device
+ /* Find a list of used extents on the given physical device
* (dnum) of the given ddf.
* Return a malloced array of 'struct extent'
*/
rv[n].size = 0;
return rv;
}
+
+static unsigned long long find_space(
+ struct ddf_super *ddf, struct dl *dl,
+ unsigned long long data_offset,
+ unsigned long long *size)
+{
+ /* Find if the requested amount of space is available.
+ * If it is, return start.
+ * If not, set *size to largest space.
+ * If data_offset != INVALID_SECTORS, then the space must start
+ * at this location.
+ */
+ struct extent *e = get_extents(ddf, dl);
+ int i = 0;
+ unsigned long long pos = 0;
+ unsigned long long max_size = 0;
+
+ if (!e) {
+ *size = 0;
+ return INVALID_SECTORS;
+ }
+ do {
+ unsigned long long esize = e[i].start - pos;
+ if (data_offset != INVALID_SECTORS &&
+ pos <= data_offset &&
+ e[i].start > data_offset) {
+ pos = data_offset;
+ esize = e[i].start - pos;
+ }
+ if (data_offset != INVALID_SECTORS &&
+ pos != data_offset) {
+ i++;
+ continue;
+ }
+ if (esize >= *size) {
+ /* Found! */
+ free(e);
+ return pos;
+ }
+ if (esize > max_size)
+ max_size = esize;
+ pos = e[i].start + e[i].size;
+ i++;
+ } while (e[i-1].size);
+ *size = max_size;
+ free(e);
+ return INVALID_SECTORS;
+}
#endif
static int init_super_ddf_bvd(struct supertype *st,
vcl->next = ddf->conflist;
ddf->conflist = vcl;
ddf->currentconf = vcl;
- ddf_set_updates_pending(ddf);
+ ddf_set_updates_pending(ddf, NULL);
return 1;
}
#ifndef MDASSEMBLE
static void add_to_super_ddf_bvd(struct supertype *st,
- mdu_disk_info_t *dk, int fd, char *devname)
+ mdu_disk_info_t *dk, int fd, char *devname,
+ unsigned long long data_offset)
{
/* fd and devname identify a device within the ddf container (st).
* dk identifies a location in the new BVD.
struct ddf_super *ddf = st->sb;
struct vd_config *vc;
unsigned int i;
- unsigned long long blocks, pos, esize;
- struct extent *ex;
+ unsigned long long blocks, pos;
unsigned int raid_disk = dk->raid_disk;
if (fd == -1) {
raid_disk %= n;
}
- ex = get_extents(ddf, dl);
- if (!ex)
- return;
-
- i = 0; pos = 0;
blocks = be64_to_cpu(vc->blocks);
if (ddf->currentconf->block_sizes)
blocks = ddf->currentconf->block_sizes[dk->raid_disk];
- /* First-fit */
- do {
- esize = ex[i].start - pos;
- if (esize >= blocks)
- break;
- pos = ex[i].start + ex[i].size;
- i++;
- } while (ex[i-1].size);
-
- free(ex);
- if (esize < blocks)
+ pos = find_space(ddf, dl, data_offset, &blocks);
+ if (pos == INVALID_SECTORS)
return;
ddf->currentdev = dk->raid_disk;
__func__, dl->pdnum, be32_to_cpu(dl->disk.refnum),
ddf->currentconf->vcnum, guid_str(vc->guid),
dk->raid_disk);
- ddf_set_updates_pending(ddf);
+ ddf_set_updates_pending(ddf, vc);
}
static unsigned int find_unused_pde(const struct ddf_super *ddf)
__u64 wsp = cfs - t;
if (wsp > 1024*1024*2ULL && wsp > dl->size / 16) {
pr_err("%s: %x:%x: workspace size 0x%llx too big, ignoring\n",
- __func__, dl->major, dl->minor, wsp);
+ __func__, dl->major, dl->minor,
+ (unsigned long long)wsp);
} else
cfs = t;
}
pde->config_size = cpu_to_be64(cfs);
dprintf("%s: %x:%x config_size %llx, DDF structure is %llx blocks\n",
- __func__, dl->major, dl->minor, cfs, dl->size-cfs);
+ __func__, dl->major, dl->minor,
+ (unsigned long long)cfs, (unsigned long long)(dl->size-cfs));
}
/* Add a device to a container, either while creating it or while
__u32 *tptr;
if (ddf->currentconf) {
- add_to_super_ddf_bvd(st, dk, fd, devname);
+ add_to_super_ddf_bvd(st, dk, fd, devname, data_offset);
return 0;
}
} else {
dd->next = ddf->dlist;
ddf->dlist = dd;
- ddf_set_updates_pending(ddf);
+ ddf_set_updates_pending(ddf, NULL);
}
return 0;
i, be32_to_cpu(d->disk.refnum),
guid_str(vdc->guid),
vdc->sec_elmnt_seq);
- vdc->seqnum = header->seq;
vdc->crc = calc_crc(vdc, conf_size);
memcpy(conf + i*conf_size, vdc, conf_size);
} else
*/
get_dev_size(fd, NULL, &size);
size /= 512;
+ memcpy(&ddf->anchor, ddf->active, 512);
if (be64_to_cpu(d->workspace_lba) != 0ULL)
ddf->anchor.workspace_lba = d->workspace_lba;
else
else
ddf->anchor.secondary_lba =
cpu_to_be64(size - 32*1024*2);
- ddf->anchor.seq = ddf->active->seq;
ddf->anchor.timestamp = cpu_to_be32(time(0) - DECADE);
memcpy(&ddf->primary, &ddf->anchor, 512);
memcpy(&ddf->secondary, &ddf->anchor, 512);
len);
append_metadata_update(st, vc, tlen);
- /* FIXME I need to close the fds! */
return 0;
} else {
struct dl *d;
if (!currentconf)
for (d = ddf->dlist; d; d=d->next)
while (Kill(d->devname, NULL, 0, -1, 1) == 0);
+ /* Note: we don't close the fd's now, but a subsequent
+ * ->free_super() will
+ */
return __write_init_super_ddf(st);
}
}
static int reserve_space(struct supertype *st, int raiddisks,
unsigned long long size, int chunk,
+ unsigned long long data_offset,
unsigned long long *freesize)
{
/* Find 'raiddisks' spare extents at least 'size' big (but
}
/* Now find largest extent on each device */
for (dl = ddf->dlist ; dl ; dl=dl->next) {
- struct extent *e = get_extents(ddf, dl);
- unsigned long long pos = 0;
- int i = 0;
- int found = 0;
- unsigned long long minsize = size;
-
- if (size == 0)
- minsize = chunk;
+ unsigned long long minsize = ULLONG_MAX;
- if (!e)
- continue;
- do {
- unsigned long long esize;
- esize = e[i].start - pos;
- if (esize >= minsize) {
- found = 1;
- minsize = esize;
- }
- pos = e[i].start + e[i].size;
- i++;
- } while (e[i-1].size);
- if (found) {
+ find_space(ddf, dl, data_offset, &minsize);
+ if (minsize >= size && minsize >= (unsigned)chunk) {
cnt++;
dl->esize = minsize;
}
- free(e);
}
if (cnt < raiddisks) {
pr_err("not enough devices with space to create array.\n");
* chosen so that add_to_super/getinfo_super
* can return them.
*/
- return reserve_space(st, raiddisks, size, *chunk, freesize);
+ return reserve_space(st, raiddisks, size, *chunk,
+ data_offset, freesize);
}
return 1;
}
*/
fd = open(dev, O_RDONLY|O_EXCL, 0);
if (fd >= 0) {
- sra = sysfs_read(fd, NULL, GET_VERSION);
close(fd);
- if (sra && sra->array.major_version == -1 &&
- strcmp(sra->text_version, "ddf") == 0) {
- /* load super */
- /* find space for 'n' devices. */
- /* remember the devices */
- /* Somehow return the fact that we have enough */
- }
-
+ /* Just a bare device, no good to us */
if (verbose)
pr_err("ddf: Cannot create this array "
"on device %s - a container is required.\n",
struct stat stb;
struct ddf_super *ddf = st->sb;
struct dl *dl;
- unsigned long long pos = 0;
unsigned long long maxsize;
- struct extent *e;
- int i;
/* ddf/bvd supports lots of things, but not containers */
if (level == LEVEL_CONTAINER) {
if (verbose)
if (minsize == 0)
minsize = 8;
for (dl = ddf->dlist; dl ; dl = dl->next) {
- int found = 0;
- pos = 0;
-
- i = 0;
- e = get_extents(ddf, dl);
- if (!e) continue;
- do {
- unsigned long long esize;
- esize = e[i].start - pos;
- if (esize >= minsize)
- found = 1;
- pos = e[i].start + e[i].size;
- i++;
- } while (e[i-1].size);
- if (found)
+ if (find_space(ddf, dl, data_offset, &minsize)
+ != INVALID_SECTORS)
dcnt++;
- free(e);
}
if (dcnt < raiddisks) {
if (verbose)
dev);
return 0;
}
- e = get_extents(ddf, dl);
- maxsize = 0;
- i = 0;
- if (e)
- do {
- unsigned long long esize;
- esize = e[i].start - pos;
- if (esize >= maxsize)
- maxsize = esize;
- pos = e[i].start + e[i].size;
- i++;
- } while (e[i-1].size);
+ maxsize = ULLONG_MAX;
+ find_space(ddf, dl, data_offset, &maxsize);
*freesize = maxsize;
- // FIXME here I am
return 1;
}
}
for (i = 0; i < conf->sec_elmnt_count; i++) {
if (!__was_sec_seen(i)) {
- pr_err("BVD %d is missing\n", i);
+ /* pr_err("BVD %d is missing\n", i); */
return -1;
}
}
be16_set(ddf->phys->entries[pd].state,
cpu_to_be16(DDF_Failed|DDF_Missing));
vc->phys_refnum[n_bvd] = cpu_to_be32(0);
- ddf_set_updates_pending(ddf);
+ ddf_set_updates_pending(ddf, vc);
}
/* Mark the array as Degraded */
(ddf->virt->entries[inst].state & ~DDF_state_mask)
| state;
a->check_degraded = 1;
- ddf_set_updates_pending(ddf);
+ ddf_set_updates_pending(ddf, vc);
}
}
}
int old = ddf->virt->entries[inst].state;
if (consistent == 2) {
handle_missing(ddf, a, inst);
- /* Should check if a recovery should be started FIXME */
consistent = 1;
if (!is_resync_complete(&a->info))
consistent = 0;
else
ddf->virt->entries[inst].state |= DDF_state_inconsistent;
if (old != ddf->virt->entries[inst].state)
- ddf_set_updates_pending(ddf);
+ ddf_set_updates_pending(ddf, NULL);
old = ddf->virt->entries[inst].init_state;
ddf->virt->entries[inst].init_state &= ~DDF_initstate_mask;
else
ddf->virt->entries[inst].init_state |= DDF_init_quick;
if (old != ddf->virt->entries[inst].init_state)
- ddf_set_updates_pending(ddf);
+ ddf_set_updates_pending(ddf, NULL);
dprintf("ddf mark %d/%s (%d) %s %llu\n", inst,
guid_str(ddf->virt->entries[inst].guid), a->curr_state,
unsigned int i, n_bvd, working = 0;
unsigned int n_prim = be16_to_cpu(vc->prim_elmnt_count);
int pd, st, state;
+ char *avail = xcalloc(1, n_prim);
+ mdu_array_info_t array;
+
+ layout_ddf2md(vc, &array);
+
for (i = 0; i < n_prim; i++) {
if (!find_index_in_bvd(ddf, vc, i, &n_bvd))
continue;
continue;
st = be16_to_cpu(ddf->phys->entries[pd].state);
if ((st & (DDF_Online|DDF_Failed|DDF_Rebuilding))
- == DDF_Online)
+ == DDF_Online) {
working++;
+ avail[i] = 1;
+ }
}
state = DDF_state_degraded;
else if (working >= 2)
state = DDF_state_part_optimal;
break;
+ case DDF_RAID1E:
+ if (!enough(10, n_prim, array.layout, 1, avail))
+ state = DDF_state_failed;
+ break;
case DDF_RAID4:
case DDF_RAID5:
if (working < n_prim - 1)
int pd;
struct mdinfo *mdi;
struct dl *dl;
+ int update = 0;
dprintf("%s: %d to %x\n", __func__, n, state);
if (vc == NULL) {
dprintf("%s: array %u disk %u ref %08x pd %d\n",
__func__, inst, n_bvd,
be32_to_cpu(vc->phys_refnum[n_bvd]), pd);
- if ((state & DS_INSYNC) && ! (state & DS_FAULTY)) {
- pd = dl->pdnum; /* FIXME: is this really correct ? */
+ if ((state & DS_INSYNC) && ! (state & DS_FAULTY) &&
+ dl->pdnum >= 0) {
+ pd = dl->pdnum;
vc->phys_refnum[n_bvd] = dl->disk.refnum;
LBA_OFFSET(ddf, vc)[n_bvd] =
cpu_to_be64(mdi->data_offset);
cpu_to_be16(DDF_Global_Spare));
be16_set(ddf->phys->entries[pd].type,
cpu_to_be16(DDF_Active_in_VD));
- ddf_set_updates_pending(ddf);
+ update = 1;
}
} else {
be16 old = ddf->phys->entries[pd].state;
cpu_to_be16(DDF_Rebuilding));
}
if (!be16_eq(old, ddf->phys->entries[pd].state))
- ddf_set_updates_pending(ddf);
+ update = 1;
}
dprintf("ddf: set_disk %d (%08x) to %x->%02x\n", n,
ddf->virt->entries[inst].state =
(ddf->virt->entries[inst].state & ~DDF_state_mask)
| state;
- ddf_set_updates_pending(ddf);
+ update = 1;
}
-
+ if (update)
+ ddf_set_updates_pending(ddf, vc);
}
static void ddf_sync_metadata(struct supertype *st)
append_metadata_update(st, vd, len);
} else {
_kill_subarray_ddf(ddf, conf->guid);
- ddf_set_updates_pending(ddf);
+ ddf_set_updates_pending(ddf, NULL);
ddf_sync_metadata(st);
}
return 0;
conf->sec_elmnt_seq, guid_str(conf->guid));
}
+static void ddf_process_phys_update(struct supertype *st,
+ struct metadata_update *update)
+{
+ struct ddf_super *ddf = st->sb;
+ struct phys_disk *pd;
+ unsigned int ent;
+
+ pd = (struct phys_disk*)update->buf;
+ ent = be16_to_cpu(pd->used_pdes);
+ if (ent >= be16_to_cpu(ddf->phys->max_pdes))
+ return;
+ if (be16_and(pd->entries[0].state, cpu_to_be16(DDF_Missing))) {
+ struct dl **dlp;
+ /* removing this disk. */
+ be16_set(ddf->phys->entries[ent].state,
+ cpu_to_be16(DDF_Missing));
+ for (dlp = &ddf->dlist; *dlp; dlp = &(*dlp)->next) {
+ struct dl *dl = *dlp;
+ if (dl->pdnum == (signed)ent) {
+ close(dl->fd);
+ dl->fd = -1;
+ *dlp = dl->next;
+ update->space = dl->devname;
+ *(void**)dl = update->space_list;
+ update->space_list = (void**)dl;
+ break;
+ }
+ }
+ ddf_set_updates_pending(ddf, NULL);
+ return;
+ }
+ if (!all_ff(ddf->phys->entries[ent].guid))
+ return;
+ ddf->phys->entries[ent] = pd->entries[0];
+ ddf->phys->used_pdes = cpu_to_be16
+ (1 + be16_to_cpu(ddf->phys->used_pdes));
+ ddf_set_updates_pending(ddf, NULL);
+ if (ddf->add_list) {
+ struct active_array *a;
+ struct dl *al = ddf->add_list;
+ ddf->add_list = al->next;
+
+ al->next = ddf->dlist;
+ ddf->dlist = al;
+
+ /* As a device has been added, we should check
+ * for any degraded devices that might make
+ * use of this spare */
+ for (a = st->arrays ; a; a=a->next)
+ a->check_degraded = 1;
+ }
+}
+
+static void ddf_process_virt_update(struct supertype *st,
+ struct metadata_update *update)
+{
+ struct ddf_super *ddf = st->sb;
+ struct virtual_disk *vd;
+ unsigned int ent;
+
+ vd = (struct virtual_disk*)update->buf;
+
+ if (vd->entries[0].state == DDF_state_deleted) {
+ if (_kill_subarray_ddf(ddf, vd->entries[0].guid))
+ return;
+ } else {
+ ent = find_vde_by_guid(ddf, vd->entries[0].guid);
+ if (ent != DDF_NOTFOUND) {
+ dprintf("%s: VD %s exists already in slot %d\n",
+ __func__, guid_str(vd->entries[0].guid),
+ ent);
+ return;
+ }
+ ent = find_unused_vde(ddf);
+ if (ent == DDF_NOTFOUND)
+ return;
+ ddf->virt->entries[ent] = vd->entries[0];
+ ddf->virt->populated_vdes =
+ cpu_to_be16(
+ 1 + be16_to_cpu(
+ ddf->virt->populated_vdes));
+ dprintf("%s: added VD %s in slot %d(s=%02x i=%02x)\n",
+ __func__, guid_str(vd->entries[0].guid), ent,
+ ddf->virt->entries[ent].state,
+ ddf->virt->entries[ent].init_state);
+ }
+ ddf_set_updates_pending(ddf, NULL);
+}
+
+static void ddf_remove_failed(struct ddf_super *ddf)
+{
+ /* Now remove any 'Failed' devices that are not part
+ * of any VD. They will have the Transition flag set.
+ * Once done, we need to update all dl->pdnum numbers.
+ */
+ unsigned int pdnum;
+ unsigned int pd2 = 0;
+ struct dl *dl;
+
+ for (pdnum = 0; pdnum < be16_to_cpu(ddf->phys->max_pdes);
+ pdnum++) {
+ if (be32_to_cpu(ddf->phys->entries[pdnum].refnum) ==
+ 0xFFFFFFFF)
+ continue;
+ if (be16_and(ddf->phys->entries[pdnum].state,
+ cpu_to_be16(DDF_Failed))
+ && be16_and(ddf->phys->entries[pdnum].state,
+ cpu_to_be16(DDF_Transition))) {
+ /* skip this one unless in dlist*/
+ for (dl = ddf->dlist; dl; dl = dl->next)
+ if (dl->pdnum == (int)pdnum)
+ break;
+ if (!dl)
+ continue;
+ }
+ if (pdnum == pd2)
+ pd2++;
+ else {
+ ddf->phys->entries[pd2] =
+ ddf->phys->entries[pdnum];
+ for (dl = ddf->dlist; dl; dl = dl->next)
+ if (dl->pdnum == (int)pdnum)
+ dl->pdnum = pd2;
+ pd2++;
+ }
+ }
+ ddf->phys->used_pdes = cpu_to_be16(pd2);
+ while (pd2 < pdnum) {
+ memset(ddf->phys->entries[pd2].guid, 0xff,
+ DDF_GUID_LEN);
+ pd2++;
+ }
+}
+
+static void ddf_update_vlist(struct ddf_super *ddf, struct dl *dl)
+{
+ struct vcl *vcl;
+ unsigned int vn = 0;
+ int in_degraded = 0;
+
+ if (dl->pdnum < 0)
+ return;
+ for (vcl = ddf->conflist; vcl ; vcl = vcl->next) {
+ unsigned int dn, ibvd;
+ const struct vd_config *conf;
+ int vstate;
+ dn = get_pd_index_from_refnum(vcl,
+ dl->disk.refnum,
+ ddf->mppe,
+ &conf, &ibvd);
+ if (dn == DDF_NOTFOUND)
+ continue;
+ dprintf("dev %d/%08x has %s (sec=%u) at %d\n",
+ dl->pdnum,
+ be32_to_cpu(dl->disk.refnum),
+ guid_str(conf->guid),
+ conf->sec_elmnt_seq, vn);
+ /* Clear the Transition flag */
+ if (be16_and
+ (ddf->phys->entries[dl->pdnum].state,
+ cpu_to_be16(DDF_Failed)))
+ be16_clear(ddf->phys
+ ->entries[dl->pdnum].state,
+ cpu_to_be16(DDF_Transition));
+ dl->vlist[vn++] = vcl;
+ vstate = ddf->virt->entries[vcl->vcnum].state
+ & DDF_state_mask;
+ if (vstate == DDF_state_degraded ||
+ vstate == DDF_state_part_optimal)
+ in_degraded = 1;
+ }
+ while (vn < ddf->max_part)
+ dl->vlist[vn++] = NULL;
+ if (dl->vlist[0]) {
+ be16_clear(ddf->phys->entries[dl->pdnum].type,
+ cpu_to_be16(DDF_Global_Spare));
+ if (!be16_and(ddf->phys
+ ->entries[dl->pdnum].type,
+ cpu_to_be16(DDF_Active_in_VD))) {
+ be16_set(ddf->phys
+ ->entries[dl->pdnum].type,
+ cpu_to_be16(DDF_Active_in_VD));
+ if (in_degraded)
+ be16_set(ddf->phys
+ ->entries[dl->pdnum]
+ .state,
+ cpu_to_be16
+ (DDF_Rebuilding));
+ }
+ }
+ if (dl->spare) {
+ be16_clear(ddf->phys->entries[dl->pdnum].type,
+ cpu_to_be16(DDF_Global_Spare));
+ be16_set(ddf->phys->entries[dl->pdnum].type,
+ cpu_to_be16(DDF_Spare));
+ }
+ if (!dl->vlist[0] && !dl->spare) {
+ be16_set(ddf->phys->entries[dl->pdnum].type,
+ cpu_to_be16(DDF_Global_Spare));
+ be16_clear(ddf->phys->entries[dl->pdnum].type,
+ cpu_to_be16(DDF_Spare));
+ be16_clear(ddf->phys->entries[dl->pdnum].type,
+ cpu_to_be16(DDF_Active_in_VD));
+ }
+}
+
+static void ddf_process_conf_update(struct supertype *st,
+ struct metadata_update *update)
+{
+ struct ddf_super *ddf = st->sb;
+ struct vd_config *vc;
+ struct vcl *vcl;
+ struct dl *dl;
+ unsigned int ent;
+ unsigned int pdnum, len;
+
+ vc = (struct vd_config*)update->buf;
+ len = ddf->conf_rec_len * 512;
+ if ((unsigned int)update->len != len * vc->sec_elmnt_count) {
+ pr_err("%s: %s: insufficient data (%d) for %u BVDs\n",
+ __func__, guid_str(vc->guid), update->len,
+ vc->sec_elmnt_count);
+ return;
+ }
+ for (vcl = ddf->conflist; vcl ; vcl = vcl->next)
+ if (memcmp(vcl->conf.guid, vc->guid, DDF_GUID_LEN) == 0)
+ break;
+ dprintf("%s: conf update for %s (%s)\n", __func__,
+ guid_str(vc->guid), (vcl ? "old" : "new"));
+ if (vcl) {
+ /* An update, just copy the phys_refnum and lba_offset
+ * fields
+ */
+ unsigned int i;
+ unsigned int k;
+ copy_matching_bvd(ddf, &vcl->conf, update);
+ for (k = 0; k < be16_to_cpu(vc->prim_elmnt_count); k++)
+ dprintf("BVD %u has %08x at %llu\n", 0,
+ be32_to_cpu(vcl->conf.phys_refnum[k]),
+ be64_to_cpu(LBA_OFFSET(ddf,
+ &vcl->conf)[k]));
+ for (i = 1; i < vc->sec_elmnt_count; i++) {
+ copy_matching_bvd(ddf, vcl->other_bvds[i-1],
+ update);
+ for (k = 0; k < be16_to_cpu(
+ vc->prim_elmnt_count); k++)
+ dprintf("BVD %u has %08x at %llu\n", i,
+ be32_to_cpu
+ (vcl->other_bvds[i-1]->
+ phys_refnum[k]),
+ be64_to_cpu
+ (LBA_OFFSET
+ (ddf,
+ vcl->other_bvds[i-1])[k]));
+ }
+ } else {
+ /* A new VD_CONF */
+ unsigned int i;
+ if (!update->space)
+ return;
+ vcl = update->space;
+ update->space = NULL;
+ vcl->next = ddf->conflist;
+ memcpy(&vcl->conf, vc, len);
+ ent = find_vde_by_guid(ddf, vc->guid);
+ if (ent == DDF_NOTFOUND)
+ return;
+ vcl->vcnum = ent;
+ ddf->conflist = vcl;
+ for (i = 1; i < vc->sec_elmnt_count; i++)
+ memcpy(vcl->other_bvds[i-1],
+ update->buf + len * i, len);
+ }
+ /* Set DDF_Transition on all Failed devices - to help
+ * us detect those that are no longer in use
+ */
+ for (pdnum = 0; pdnum < be16_to_cpu(ddf->phys->max_pdes);
+ pdnum++)
+ if (be16_and(ddf->phys->entries[pdnum].state,
+ cpu_to_be16(DDF_Failed)))
+ be16_set(ddf->phys->entries[pdnum].state,
+ cpu_to_be16(DDF_Transition));
+
+ /* Now make sure vlist is correct for each dl. */
+ for (dl = ddf->dlist; dl; dl = dl->next)
+ ddf_update_vlist(ddf, dl);
+ ddf_remove_failed(ddf);
+
+ ddf_set_updates_pending(ddf, vc);
+}
+
static void ddf_process_update(struct supertype *st,
struct metadata_update *update)
{
* and offset. This will also mark the spare as active with
* a spare-assignment record.
*/
- struct ddf_super *ddf = st->sb;
be32 *magic = (be32 *)update->buf;
- struct phys_disk *pd;
- struct virtual_disk *vd;
- struct vd_config *vc;
- struct vcl *vcl;
- struct dl *dl;
- unsigned int ent;
- unsigned int pdnum, pd2, len;
dprintf("Process update %x\n", be32_to_cpu(*magic));
if (be32_eq(*magic, DDF_PHYS_RECORDS_MAGIC)) {
- if (update->len != (sizeof(struct phys_disk) +
+ if (update->len == (sizeof(struct phys_disk) +
sizeof(struct phys_disk_entry)))
- return;
- pd = (struct phys_disk*)update->buf;
-
- ent = be16_to_cpu(pd->used_pdes);
- if (ent >= be16_to_cpu(ddf->phys->max_pdes))
- return;
- if (be16_and(pd->entries[0].state, cpu_to_be16(DDF_Missing))) {
- struct dl **dlp;
- /* removing this disk. */
- be16_set(ddf->phys->entries[ent].state,
- cpu_to_be16(DDF_Missing));
- for (dlp = &ddf->dlist; *dlp; dlp = &(*dlp)->next) {
- struct dl *dl = *dlp;
- if (dl->pdnum == (signed)ent) {
- close(dl->fd);
- dl->fd = -1;
- /* FIXME this doesn't free
- * dl->devname */
- update->space = dl;
- *dlp = dl->next;
- break;
- }
- }
- ddf_set_updates_pending(ddf);
- return;
- }
- if (!all_ff(ddf->phys->entries[ent].guid))
- return;
- ddf->phys->entries[ent] = pd->entries[0];
- ddf->phys->used_pdes = cpu_to_be16
- (1 + be16_to_cpu(ddf->phys->used_pdes));
- ddf_set_updates_pending(ddf);
- if (ddf->add_list) {
- struct active_array *a;
- struct dl *al = ddf->add_list;
- ddf->add_list = al->next;
-
- al->next = ddf->dlist;
- ddf->dlist = al;
-
- /* As a device has been added, we should check
- * for any degraded devices that might make
- * use of this spare */
- for (a = st->arrays ; a; a=a->next)
- a->check_degraded = 1;
- }
+ ddf_process_phys_update(st, update);
} else if (be32_eq(*magic, DDF_VIRT_RECORDS_MAGIC)) {
- if (update->len != (sizeof(struct virtual_disk) +
+ if (update->len == (sizeof(struct virtual_disk) +
sizeof(struct virtual_entry)))
- return;
- vd = (struct virtual_disk*)update->buf;
-
- if (vd->entries[0].state == DDF_state_deleted) {
- if (_kill_subarray_ddf(ddf, vd->entries[0].guid))
- return;
- } else {
- ent = find_vde_by_guid(ddf, vd->entries[0].guid);
- if (ent != DDF_NOTFOUND) {
- dprintf("%s: VD %s exists already in slot %d\n",
- __func__, guid_str(vd->entries[0].guid),
- ent);
- return;
- }
- ent = find_unused_vde(ddf);
- if (ent == DDF_NOTFOUND)
- return;
- ddf->virt->entries[ent] = vd->entries[0];
- ddf->virt->populated_vdes =
- cpu_to_be16(
- 1 + be16_to_cpu(
- ddf->virt->populated_vdes));
- dprintf("%s: added VD %s in slot %d(s=%02x i=%02x)\n",
- __func__, guid_str(vd->entries[0].guid), ent,
- ddf->virt->entries[ent].state,
- ddf->virt->entries[ent].init_state);
- }
- ddf_set_updates_pending(ddf);
- }
-
- else if (be32_eq(*magic, DDF_VD_CONF_MAGIC)) {
- vc = (struct vd_config*)update->buf;
- len = ddf->conf_rec_len * 512;
- if ((unsigned int)update->len != len * vc->sec_elmnt_count) {
- pr_err("%s: %s: insufficient data (%d) for %u BVDs\n",
- __func__, guid_str(vc->guid), update->len,
- vc->sec_elmnt_count);
- return;
- }
- for (vcl = ddf->conflist; vcl ; vcl = vcl->next)
- if (memcmp(vcl->conf.guid, vc->guid, DDF_GUID_LEN) == 0)
- break;
- dprintf("%s: conf update for %s (%s)\n", __func__,
- guid_str(vc->guid), (vcl ? "old" : "new"));
- if (vcl) {
- /* An update, just copy the phys_refnum and lba_offset
- * fields
- */
- unsigned int i;
- unsigned int k;
- copy_matching_bvd(ddf, &vcl->conf, update);
- for (k = 0; k < be16_to_cpu(vc->prim_elmnt_count); k++)
- dprintf("BVD %u has %08x at %llu\n", 0,
- be32_to_cpu(vcl->conf.phys_refnum[k]),
- be64_to_cpu(LBA_OFFSET(ddf,
- &vcl->conf)[k]));
- for (i = 1; i < vc->sec_elmnt_count; i++) {
- copy_matching_bvd(ddf, vcl->other_bvds[i-1],
- update);
- for (k = 0; k < be16_to_cpu(
- vc->prim_elmnt_count); k++)
- dprintf("BVD %u has %08x at %llu\n", i,
- be32_to_cpu
- (vcl->other_bvds[i-1]->
- phys_refnum[k]),
- be64_to_cpu
- (LBA_OFFSET
- (ddf,
- vcl->other_bvds[i-1])[k]));
- }
- } else {
- /* A new VD_CONF */
- unsigned int i;
- if (!update->space)
- return;
- vcl = update->space;
- update->space = NULL;
- vcl->next = ddf->conflist;
- memcpy(&vcl->conf, vc, len);
- ent = find_vde_by_guid(ddf, vc->guid);
- if (ent == DDF_NOTFOUND)
- return;
- vcl->vcnum = ent;
- ddf->conflist = vcl;
- for (i = 1; i < vc->sec_elmnt_count; i++)
- memcpy(vcl->other_bvds[i-1],
- update->buf + len * i, len);
- }
- /* Set DDF_Transition on all Failed devices - to help
- * us detect those that are no longer in use
- */
- for (pdnum = 0; pdnum < be16_to_cpu(ddf->phys->max_pdes);
- pdnum++)
- if (be16_and(ddf->phys->entries[pdnum].state,
- cpu_to_be16(DDF_Failed)))
- be16_set(ddf->phys->entries[pdnum].state,
- cpu_to_be16(DDF_Transition));
- /* Now make sure vlist is correct for each dl. */
- for (dl = ddf->dlist; dl; dl = dl->next) {
- unsigned int vn = 0;
- int in_degraded = 0;
-
- if (dl->pdnum < 0)
- continue;
- for (vcl = ddf->conflist; vcl ; vcl = vcl->next) {
- unsigned int dn, ibvd;
- const struct vd_config *conf;
- int vstate;
- dn = get_pd_index_from_refnum(vcl,
- dl->disk.refnum,
- ddf->mppe,
- &conf, &ibvd);
- if (dn == DDF_NOTFOUND)
- continue;
- dprintf("dev %d/%08x has %s (sec=%u) at %d\n",
- dl->pdnum,
- be32_to_cpu(dl->disk.refnum),
- guid_str(conf->guid),
- conf->sec_elmnt_seq, vn);
- /* Clear the Transition flag */
- if (be16_and
- (ddf->phys->entries[dl->pdnum].state,
- cpu_to_be16(DDF_Failed)))
- be16_clear(ddf->phys
- ->entries[dl->pdnum].state,
- cpu_to_be16(DDF_Transition));
- dl->vlist[vn++] = vcl;
- vstate = ddf->virt->entries[vcl->vcnum].state
- & DDF_state_mask;
- if (vstate == DDF_state_degraded ||
- vstate == DDF_state_part_optimal)
- in_degraded = 1;
- }
- while (vn < ddf->max_part)
- dl->vlist[vn++] = NULL;
- if (dl->vlist[0]) {
- be16_clear(ddf->phys->entries[dl->pdnum].type,
- cpu_to_be16(DDF_Global_Spare));
- if (!be16_and(ddf->phys
- ->entries[dl->pdnum].type,
- cpu_to_be16(DDF_Active_in_VD))) {
- be16_set(ddf->phys
- ->entries[dl->pdnum].type,
- cpu_to_be16(DDF_Active_in_VD));
- if (in_degraded)
- be16_set(ddf->phys
- ->entries[dl->pdnum]
- .state,
- cpu_to_be16
- (DDF_Rebuilding));
- }
- }
- if (dl->spare) {
- be16_clear(ddf->phys->entries[dl->pdnum].type,
- cpu_to_be16(DDF_Global_Spare));
- be16_set(ddf->phys->entries[dl->pdnum].type,
- cpu_to_be16(DDF_Spare));
- }
- if (!dl->vlist[0] && !dl->spare) {
- be16_set(ddf->phys->entries[dl->pdnum].type,
- cpu_to_be16(DDF_Global_Spare));
- be16_clear(ddf->phys->entries[dl->pdnum].type,
- cpu_to_be16(DDF_Spare));
- be16_clear(ddf->phys->entries[dl->pdnum].type,
- cpu_to_be16(DDF_Active_in_VD));
- }
- }
-
- /* Now remove any 'Failed' devices that are not part
- * of any VD. They will have the Transition flag set.
- * Once done, we need to update all dl->pdnum numbers.
- */
- pd2 = 0;
- for (pdnum = 0; pdnum < be16_to_cpu(ddf->phys->max_pdes);
- pdnum++) {
- if (be32_to_cpu(ddf->phys->entries[pdnum].refnum) ==
- 0xFFFFFFFF)
- continue;
- if (be16_and(ddf->phys->entries[pdnum].state,
- cpu_to_be16(DDF_Failed))
- && be16_and(ddf->phys->entries[pdnum].state,
- cpu_to_be16(DDF_Transition))) {
- /* skip this one unless in dlist*/
- for (dl = ddf->dlist; dl; dl = dl->next)
- if (dl->pdnum == (int)pdnum)
- break;
- if (!dl)
- continue;
- }
- if (pdnum == pd2)
- pd2++;
- else {
- ddf->phys->entries[pd2] =
- ddf->phys->entries[pdnum];
- for (dl = ddf->dlist; dl; dl = dl->next)
- if (dl->pdnum == (int)pdnum)
- dl->pdnum = pd2;
- pd2++;
- }
- }
- ddf->phys->used_pdes = cpu_to_be16(pd2);
- while (pd2 < pdnum) {
- memset(ddf->phys->entries[pd2].guid, 0xff,
- DDF_GUID_LEN);
- pd2++;
- }
-
- ddf_set_updates_pending(ddf);
+ ddf_process_virt_update(st, update);
+ } else if (be32_eq(*magic, DDF_VD_CONF_MAGIC)) {
+ ddf_process_conf_update(st, update);
}
/* case DDF_SPARE_ASSIGN_MAGIC */
}
-static void ddf_prepare_update(struct supertype *st,
- struct metadata_update *update)
+static int ddf_prepare_update(struct supertype *st,
+ struct metadata_update *update)
{
/* This update arrived at managemon.
* We are about to pass it to monitor.
* If a malloc is needed, do it here.
*/
struct ddf_super *ddf = st->sb;
- be32 *magic = (be32 *)update->buf;
+ be32 *magic;
+ if (update->len < 4)
+ return 0;
+ magic = (be32 *)update->buf;
if (be32_eq(*magic, DDF_VD_CONF_MAGIC)) {
struct vcl *vcl;
- struct vd_config *conf = (struct vd_config *) update->buf;
+ struct vd_config *conf;
+ if (update->len < (int)sizeof(*conf))
+ return 0;
+ conf = (struct vd_config *) update->buf;
if (posix_memalign(&update->space, 512,
offsetof(struct vcl, conf)
+ ddf->conf_rec_len * 512) != 0) {
update->space = NULL;
- return;
+ return 0;
}
vcl = update->space;
vcl->conf.sec_elmnt_count = conf->sec_elmnt_count;
if (alloc_other_bvds(ddf, vcl) != 0) {
free(update->space);
update->space = NULL;
+ return 0;
}
}
+ return 1;
}
/*
struct mdinfo *d2;
int is_global = 0;
int is_dedicated = 0;
- struct extent *ex;
- unsigned int j;
be16 state;
if (dl->pdnum < 0)
if (dl->spare) {
if (dl->spare->type & DDF_spare_dedicated) {
/* check spare_ents for guid */
+ unsigned int j;
for (j = 0 ;
j < be16_to_cpu
(dl->spare
/* We are allowed to use this device - is there space?
* We need a->info.component_size sectors */
- ex = get_extents(ddf, dl);
- if (!ex) {
- dprintf("cannot get extents\n");
- continue;
- }
- j = 0; pos = 0;
- esize = 0;
-
- do {
- esize = ex[j].start - pos;
- if (esize >= a->info.component_size)
- break;
- pos = ex[j].start + ex[j].size;
- j++;
- } while (ex[j-1].size);
+ esize = a->info.component_size;
+ pos = find_space(ddf, dl, INVALID_SECTORS, &esize);
- free(ex);
if (esize < a->info.component_size) {
dprintf("%x:%x has no room: %llu %llu\n",
dl->major, dl->minor,
projects / thirdparty / mdadm.git / blobdiff