/*
* mdadm - manage Linux "md" devices aka RAID arrays.
*
- * Copyright (C) 2006-2007 Neil Brown <neilb@suse.de>
+ * Copyright (C) 2006-2009 Neil Brown <neilb@suse.de>
*
*
* This program is free software; you can redistribute it and/or modify
struct phys_disk *phys;
struct virtual_disk *virt;
int pdsize, vdsize;
- int max_part, mppe, conf_rec_len;
+ unsigned int max_part, mppe, conf_rec_len;
int currentdev;
int updates_pending;
struct vcl {
struct vcl *next;
__u64 *lba_offset; /* location in 'conf' of
* the lba table */
- int vcnum; /* index into ->virt */
+ unsigned int vcnum; /* index into ->virt */
__u64 *block_sizes; /* NULL if all the same */
};
};
unsigned long long size; /* sectors */
int pdnum; /* index in ->phys */
struct spare_assign *spare;
+ void *mdupdate; /* hold metadata update */
+
+ /* These fields used by auto-layout */
+ int raiddisk; /* slot to fill in autolayout */
+ __u64 esize;
};
};
struct disk_data disk;
- void *mdupdate; /* hold metadata update */
struct vcl *vlist[0]; /* max_part in size */
} *dlist, *add_list;
};
#endif
-static int calc_crc(void *buf, int len)
+static unsigned int calc_crc(void *buf, int len)
{
/* crcs are always at the same place as in the ddf_header */
struct ddf_header *ddf = buf;
newcrc = crc32(0, buf, len);
ddf->crc = oldcrc;
- return newcrc;
+ /* The crc is store (like everything) bigendian, so convert
+ * here for simplicity
+ */
+ return __cpu_to_be32(newcrc);
}
static int load_ddf_header(int fd, unsigned long long lba,
else
offset += __be64_to_cpu(super->active->secondary_lba);
- if (lseek64(fd, offset<<9, 0) != (offset<<9)) {
+ if ((unsigned long long)lseek64(fd, offset<<9, 0) != (offset<<9)) {
if (dofree)
free(buf);
return NULL;
}
- if (read(fd, buf, len<<9) != (len<<9)) {
+ if ((unsigned long long)read(fd, buf, len<<9) != (len<<9)) {
if (dofree)
free(buf);
return NULL;
struct dl *dl;
struct stat stb;
char *conf;
- int i;
+ unsigned int i;
+ unsigned int confsec;
int vnum;
- int max_virt_disks = __be16_to_cpu(super->active->max_vd_entries);
+ unsigned int max_virt_disks = __be16_to_cpu(super->active->max_vd_entries);
unsigned long long dsize;
/* First the local disk info */
if (get_dev_size(fd, devname, &dsize))
dl->size = dsize >> 9;
dl->spare = NULL;
- for (i=0 ; i < super->max_part ; i++)
+ for (i = 0 ; i < super->max_part ; i++)
dl->vlist[i] = NULL;
super->dlist = dl;
dl->pdnum = -1;
- for (i=0; i < __be16_to_cpu(super->active->max_pd_entries); i++)
+ for (i = 0; i < __be16_to_cpu(super->active->max_pd_entries); i++)
if (memcmp(super->phys->entries[i].guid,
dl->disk.guid, DDF_GUID_LEN) == 0)
dl->pdnum = i;
0);
vnum = 0;
- for (i = 0;
- i < __be32_to_cpu(super->active->config_section_length);
- i += super->conf_rec_len) {
+ for (confsec = 0;
+ confsec < __be32_to_cpu(super->active->config_section_length);
+ confsec += super->conf_rec_len) {
struct vd_config *vd =
- (struct vd_config *)((char*)conf + i*512);
+ (struct vd_config *)((char*)conf + confsec*512);
struct vcl *vcl;
if (vd->magic == DDF_SPARE_ASSIGN_MAGIC) {
static int load_super_ddf_all(struct supertype *st, int fd,
void **sbp, char *devname, int keep_fd);
#endif
+
+static void free_super_ddf(struct supertype *st);
+
static int load_super_ddf(struct supertype *st, int fd,
char *devname)
{
if (get_dev_size(fd, devname, &dsize) == 0)
return 1;
+ if (test_partition(fd))
+ /* DDF is not allowed on partitions */
+ return 1;
+
/* 32M is a lower bound */
if (dsize <= 32*1024*1024) {
- if (devname) {
+ if (devname)
fprintf(stderr,
Name ": %s is too small for ddf: "
"size is %llu sectors.\n",
devname, dsize>>9);
- return 1;
- }
+ return 1;
}
if (dsize & 511) {
- if (devname) {
+ if (devname)
fprintf(stderr,
Name ": %s is an odd size for ddf: "
"size is %llu bytes.\n",
devname, dsize);
- return 1;
- }
+ return 1;
}
+ free_super_ddf(st);
+
if (posix_memalign((void**)&super, 512, sizeof(*super))!= 0) {
fprintf(stderr, Name ": malloc of %zu failed.\n",
sizeof(*super));
return rv;
}
+ if (st->subarray[0]) {
+ unsigned long val;
+ struct vcl *v;
+ char *ep;
+
+ val = strtoul(st->subarray, &ep, 10);
+ if (*ep != '\0') {
+ free(super);
+ return 1;
+ }
+
+ for (v = super->conflist; v; v = v->next)
+ if (v->vcnum == val)
+ super->currentconf = v;
+ if (!super->currentconf) {
+ free(super);
+ return 1;
+ }
+ }
+
/* Should possibly check the sections .... */
st->sb = super;
free(d->spare);
free(d);
}
+ while (ddf->add_list) {
+ struct dl *d = ddf->add_list;
+ ddf->add_list = d->next;
+ if (d->fd >= 0)
+ close(d->fd);
+ if (d->spare)
+ free(d->spare);
+ free(d);
+ }
free(ddf);
st->sb = NULL;
}
return map[i].num2;
}
+static int all_ff(char *guid)
+{
+ int i;
+ for (i = 0; i < DDF_GUID_LEN; i++)
+ if (guid[i] != (char)0xff)
+ return 0;
+ return 1;
+}
+
#ifndef MDASSEMBLE
static void print_guid(char *guid, int tstamp)
{
printf("%02X", guid[i]&255);
}
- printf(" (");
+ printf("\n (");
while (l && guid[l-1] == ' ')
l--;
for (i=0 ; i<l ; i++) {
struct vcl *vcl;
for (vcl = sb->conflist ; vcl ; vcl = vcl->next) {
+ unsigned int i;
struct vd_config *vc = &vcl->conf;
if (calc_crc(vc, crl*512) != vc->crc)
continue;
/* Ok, we know about this VD, let's give more details */
- printf(" Raid Devices[%d] : %d\n", n,
+ printf(" Raid Devices[%d] : %d (", n,
__be16_to_cpu(vc->prim_elmnt_count));
+ for (i = 0; i < __be16_to_cpu(vc->prim_elmnt_count); i++) {
+ int j;
+ int cnt = __be16_to_cpu(sb->phys->used_pdes);
+ for (j=0; j<cnt; j++)
+ if (vc->phys_refnum[i] == sb->phys->entries[j].refnum)
+ break;
+ if (i) printf(" ");
+ if (j < cnt)
+ printf("%d", j);
+ else
+ printf("--");
+ }
+ printf(")\n");
+ if (vc->chunk_shift != 255)
printf(" Chunk Size[%d] : %d sectors\n", n,
1 << vc->chunk_shift);
printf(" Raid Level[%d] : %s\n", n,
map_num(ddf_sec_level, vc->srl) ?: "-unknown-");
}
printf(" Device Size[%d] : %llu\n", n,
- __be64_to_cpu(vc->blocks)/2);
+ (unsigned long long)__be64_to_cpu(vc->blocks)/2);
printf(" Array Size[%d] : %llu\n", n,
- __be64_to_cpu(vc->array_blocks)/2);
+ (unsigned long long)__be64_to_cpu(vc->array_blocks)/2);
}
}
for (i=0; i<cnt; i++) {
struct virtual_entry *ve = &sb->virt->entries[i];
+ printf("\n");
printf(" VD GUID[%d] : ", i); print_guid(ve->guid, 1);
printf("\n");
printf(" unit[%d] : %d\n", i, __be16_to_cpu(ve->unit));
int i;
struct dl *dl;
printf(" Physical Disks : %d\n", cnt);
+ printf(" Number RefNo Size Device Type/State\n");
for (i=0 ; i<cnt ; i++) {
struct phys_disk_entry *pd = &sb->phys->entries[i];
int type = __be16_to_cpu(pd->type);
int state = __be16_to_cpu(pd->state);
- printf(" PD GUID[%d] : ", i); print_guid(pd->guid, 0);
- printf("\n");
- printf(" ref[%d] : %08x\n", i,
+ //printf(" PD GUID[%d] : ", i); print_guid(pd->guid, 0);
+ //printf("\n");
+ printf(" %3d %08x ", i,
__be32_to_cpu(pd->refnum));
- printf(" mode[%d] : %s%s%s%s%s\n", i,
+ printf("%8lluK ",
+ (unsigned long long)__be64_to_cpu(pd->config_size)>>1);
+ for (dl = sb->dlist; dl ; dl = dl->next) {
+ if (dl->disk.refnum == pd->refnum) {
+ char *dv = map_dev(dl->major, dl->minor, 0);
+ if (dv) {
+ printf("%-15s", dv);
+ break;
+ }
+ }
+ }
+ if (!dl)
+ printf("%15s","");
+ printf(" %s%s%s%s%s",
(type&2) ? "active":"",
- (type&4) ? "Global Spare":"",
+ (type&4) ? "Global-Spare":"",
(type&8) ? "spare" : "",
(type&16)? ", foreign" : "",
(type&32)? "pass-through" : "");
- printf(" state[%d] : %s%s%s%s%s%s%s\n", i,
+ if (state & DDF_Failed)
+ /* This over-rides these three */
+ state &= ~(DDF_Online|DDF_Rebuilding|DDF_Transition);
+ printf("/%s%s%s%s%s%s%s",
(state&1)? "Online": "Offline",
(state&2)? ", Failed": "",
(state&4)? ", Rebuilding": "",
(state&8)? ", in-transition": "",
- (state&16)? ", SMART errors": "",
- (state&32)? ", Unrecovered Read Errors": "",
+ (state&16)? ", SMART-errors": "",
+ (state&32)? ", Unrecovered-Read-Errors": "",
(state&64)? ", Missing" : "");
- printf(" Avail Size[%d] : %llu K\n", i,
- __be64_to_cpu(pd->config_size)>>1);
- for (dl = sb->dlist; dl ; dl = dl->next) {
- if (dl->disk.refnum == pd->refnum) {
- char *dv = map_dev(dl->major, dl->minor, 0);
- if (dv)
- printf(" Device[%d] : %s\n",
- i, dv);
- }
- }
printf("\n");
}
}
static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info);
+static void uuid_from_super_ddf(struct supertype *st, int uuid[4]);
-static void brief_examine_super_ddf(struct supertype *st)
+static void brief_examine_super_ddf(struct supertype *st, int verbose)
{
/* We just write a generic DDF ARRAY entry
*/
char nbuf[64];
getinfo_super_ddf(st, &info);
fname_from_uuid(st, &info, nbuf, ':');
- printf("ARRAY /dev/ddf metadata=ddf UUID=%s\n", nbuf + 5);
+
+ printf("ARRAY metadata=ddf UUID=%s\n", nbuf + 5);
+}
+
+static void brief_examine_subarrays_ddf(struct supertype *st, int verbose)
+{
+ /* We just write a generic DDF ARRAY entry
+ */
+ struct ddf_super *ddf = st->sb;
+ struct mdinfo info;
+ unsigned int i;
+ char nbuf[64];
+ getinfo_super_ddf(st, &info);
+ fname_from_uuid(st, &info, nbuf, ':');
+
+ for (i = 0; i < __be16_to_cpu(ddf->virt->max_vdes); i++) {
+ struct virtual_entry *ve = &ddf->virt->entries[i];
+ struct vcl vcl;
+ char nbuf1[64];
+ if (all_ff(ve->guid))
+ continue;
+ memcpy(vcl.conf.guid, ve->guid, DDF_GUID_LEN);
+ ddf->currentconf =&vcl;
+ uuid_from_super_ddf(st, info.uuid);
+ fname_from_uuid(st, &info, nbuf1, ':');
+ printf("ARRAY container=%s member=%d UUID=%s\n",
+ nbuf+5, i, nbuf1+5);
+ }
}
+static void export_examine_super_ddf(struct supertype *st)
+{
+ struct mdinfo info;
+ char nbuf[64];
+ getinfo_super_ddf(st, &info);
+ fname_from_uuid(st, &info, nbuf, ':');
+ printf("MD_METADATA=ddf\n");
+ printf("MD_LEVEL=container\n");
+ printf("MD_UUID=%s\n", nbuf+5);
+}
+
+
static void detail_super_ddf(struct supertype *st, char *homehost)
{
/* FIXME later
* the hostname
*/
struct ddf_super *ddf = st->sb;
- int len = strlen(homehost);
+ unsigned int len;
+
+ if (!homehost)
+ return 0;
+ len = strlen(homehost);
return (memcmp(ddf->controller.guid, T10, 8) == 0 &&
len < sizeof(ddf->controller.vendor_data) &&
}
#ifndef MDASSEMBLE
-static struct vd_config *find_vdcr(struct ddf_super *ddf, int inst)
+static struct vd_config *find_vdcr(struct ddf_super *ddf, unsigned int inst)
{
struct vcl *v;
/* Find the entry in phys_disk which has the given refnum
* and return it's index
*/
- int i;
- for (i=0; i < __be16_to_cpu(ddf->phys->max_pdes); i++)
+ unsigned int i;
+ for (i = 0; i < __be16_to_cpu(ddf->phys->max_pdes); i++)
if (ddf->phys->entries[i].refnum == phys_refnum)
return i;
return -1;
sha1_init_ctx(&ctx);
sha1_process_bytes(guid, DDF_GUID_LEN, &ctx);
- if (vcl && vcl->conf.sec_elmnt_count > 1)
- sha1_process_bytes(&vcl->conf.sec_elmnt_seq, 1, &ctx);
sha1_finish_ctx(&ctx, buf);
memcpy(uuid, buf, 4*4);
}
(ddf->anchor.guid+16));
info->array.utime = 0;
info->array.chunk_size = 0;
+ info->container_enough = 1;
info->disk.major = 0;
info->component_size = ddf->dlist->size - info->data_offset;
} else {
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);
+ info->disk.state = (1 << MD_DISK_SYNC) | (1 << MD_DISK_ACTIVE);
+ info->recovery_start = MaxSector;
info->reshape_active = 0;
info->name[0] = 0;
struct ddf_super *ddf = st->sb;
struct vcl *vc = ddf->currentconf;
int cd = ddf->currentdev;
+ int j;
+ struct dl *dl;
/* FIXME this returns BVD info - what if we want SVD ?? */
__be32_to_cpu(*(__u32*)(vc->conf.guid+16));
info->array.utime = DECADE + __be32_to_cpu(vc->conf.timestamp);
info->array.chunk_size = 512 << vc->conf.chunk_shift;
+ info->custom_array_size = 0;
- if (cd >= 0 && cd < ddf->mppe) {
+ if (cd >= 0 && (unsigned)cd < ddf->mppe) {
info->data_offset = __be64_to_cpu(vc->lba_offset[cd]);
if (vc->block_sizes)
info->component_size = vc->block_sizes[cd];
info->component_size = __be64_to_cpu(vc->conf.blocks);
}
+ for (dl = ddf->dlist; dl ; dl = dl->next)
+ if (dl->raiddisk == info->disk.raid_disk)
+ break;
info->disk.major = 0;
info->disk.minor = 0;
+ if (dl) {
+ info->disk.major = dl->major;
+ info->disk.minor = dl->minor;
+ }
// info->disk.number = __be32_to_cpu(ddf->disk.refnum);
// info->disk.raid_disk = find refnum in the table and use index;
// info->disk.state = ???;
info->container_member = ddf->currentconf->vcnum;
+ info->recovery_start = MaxSector;
info->resync_start = 0;
+ info->reshape_active = 0;
if (!(ddf->virt->entries[info->container_member].state
& DDF_state_inconsistent) &&
(ddf->virt->entries[info->container_member].init_state
& DDF_initstate_mask)
== DDF_init_full)
- info->resync_start = ~0ULL;
+ info->resync_start = MaxSector;
uuid_from_super_ddf(st, info->uuid);
- info->container_member = atoi(st->subarray);
info->array.major_version = -1;
info->array.minor_version = -2;
sprintf(info->text_version, "/%s/%s",
st->subarray);
info->safe_mode_delay = 200;
- info->name[0] = 0;
+ memcpy(info->name, ddf->virt->entries[info->container_member].name, 16);
+ info->name[16]=0;
+ for(j=0; j<16; j++)
+ if (info->name[j] == ' ')
+ info->name[j] = 0;
}
static void make_header_guid(char *guid)
{
__u32 stamp;
- int rfd;
/* Create a DDF Header of Virtual Disk GUID */
/* 24 bytes of fiction required.
memcpy(guid+12, &stamp, 4);
stamp = __cpu_to_be32(time(0) - DECADE);
memcpy(guid+16, &stamp, 4);
- rfd = open("/dev/urandom", O_RDONLY);
- if (rfd < 0 || read(rfd, &stamp, 4) != 4)
- stamp = random();
+ stamp = random32();
memcpy(guid+20, &stamp, 4);
- if (rfd >= 0) close(rfd);
}
static int init_super_ddf_bvd(struct supertype *st,
struct phys_disk *pd;
struct virtual_disk *vd;
- if (!info) {
- st->sb = NULL;
- return 0;
- }
if (st->sb)
- return init_super_ddf_bvd(st, info, size, name, homehost,
- uuid);
+ return init_super_ddf_bvd(st, info, size, name, homehost, uuid);
if (posix_memalign((void**)&ddf, 512, sizeof(*ddf)) != 0) {
fprintf(stderr, Name ": %s could not allocate superblock\n", __func__);
memset(ddf, 0, sizeof(*ddf));
ddf->dlist = NULL; /* no physical disks yet */
ddf->conflist = NULL; /* No virtual disks yet */
+ st->sb = ddf;
+
+ if (info == NULL) {
+ /* zeroing superblock */
+ return 0;
+ }
/* At least 32MB *must* be reserved for the ddf. So let's just
* start 32MB from the end, and put the primary header there.
memcpy(ddf->controller.product_id, "What Is My PID??", 16);
memset(ddf->controller.pad, 0xff, 8);
memset(ddf->controller.vendor_data, 0xff, 448);
+ if (homehost && strlen(homehost) < 440)
+ strcpy((char*)ddf->controller.vendor_data, homehost);
if (posix_memalign((void**)&pd, 512, pdsize) != 0) {
fprintf(stderr, Name ": %s could not allocate pd\n", __func__);
memset(pd, 0xff, pdsize);
memset(pd, 0, sizeof(*pd));
- pd->magic = DDF_PHYS_DATA_MAGIC;
+ pd->magic = DDF_PHYS_RECORDS_MAGIC;
pd->used_pdes = __cpu_to_be16(0);
pd->max_pdes = __cpu_to_be16(max_phys_disks);
memset(pd->pad, 0xff, 52);
return 1;
}
-static int all_ff(char *guid)
-{
- int i;
- for (i = 0; i < DDF_GUID_LEN; i++)
- if (guid[i] != (char)0xff)
- return 0;
- return 1;
-}
static int chunk_to_shift(int chunksize)
{
return ffs(chunksize/512)-1;
}
break;
case 5:
- case 6:
switch(layout) {
case ALGORITHM_LEFT_ASYMMETRIC:
return DDF_RAID5_N_RESTART;
case ALGORITHM_RIGHT_ASYMMETRIC:
- if (level == 5)
- return DDF_RAID5_0_RESTART;
- else
- return DDF_RAID6_0_RESTART;
+ return DDF_RAID5_0_RESTART;
case ALGORITHM_LEFT_SYMMETRIC:
return DDF_RAID5_N_CONTINUE;
case ALGORITHM_RIGHT_SYMMETRIC:
return -1; /* not mentioned in standard */
}
+ case 6:
+ switch(layout) {
+ case ALGORITHM_ROTATING_N_RESTART:
+ return DDF_RAID5_N_RESTART;
+ case ALGORITHM_ROTATING_ZERO_RESTART:
+ return DDF_RAID6_0_RESTART;
+ case ALGORITHM_ROTATING_N_CONTINUE:
+ return DDF_RAID5_N_CONTINUE;
+ }
}
return -1;
}
case DDF_RAID6:
switch(rlq) {
case DDF_RAID5_N_RESTART:
- return ALGORITHM_LEFT_ASYMMETRIC;
+ return ALGORITHM_ROTATING_N_RESTART;
case DDF_RAID6_0_RESTART:
- return ALGORITHM_RIGHT_ASYMMETRIC;
+ return ALGORITHM_ROTATING_ZERO_RESTART;
case DDF_RAID5_N_CONTINUE:
- return ALGORITHM_LEFT_SYMMETRIC;
+ return ALGORITHM_ROTATING_N_CONTINUE;
default:
return -1;
}
*/
struct extent *rv;
int n = 0;
- int i, j;
+ unsigned int i, j;
rv = malloc(sizeof(struct extent) * (ddf->max_part + 2));
if (!rv)
struct vcl *v = dl->vlist[i];
if (v == NULL)
continue;
- for (j=0; j < v->conf.prim_elmnt_count; j++)
+ for (j = 0; j < v->conf.prim_elmnt_count; j++)
if (v->conf.phys_refnum[j] == dl->disk.refnum) {
/* This device plays role 'j' in 'v'. */
rv[n].start = __be64_to_cpu(v->lba_offset[j]);
* We need to create a new vd_config and a new virtual_entry
*/
struct ddf_super *ddf = st->sb;
- int venum;
+ unsigned int venum;
struct virtual_entry *ve;
struct vcl *vcl;
struct vd_config *vc;
return 0;
}
+ if (name)
+ for (venum = 0; venum < __be16_to_cpu(ddf->virt->max_vdes); venum++)
+ if (!all_ff(ddf->virt->entries[venum].guid)) {
+ char *n = ddf->virt->entries[venum].name;
+
+ if (strncmp(name, n, 16) == 0) {
+ fprintf(stderr, Name ": This ddf already"
+ " has an array called %s\n",
+ name);
+ return 0;
+ }
+ }
+
for (venum = 0; venum < __be16_to_cpu(ddf->virt->max_vdes); venum++)
if (all_ff(ddf->virt->entries[venum].guid))
break;
memset(vc->vendor, 0xff, 32);
memset(vc->phys_refnum, 0xff, 4*ddf->mppe);
- memset(vc->phys_refnum+(ddf->mppe * 4), 0x00, 8*ddf->mppe);
+ memset(vc->phys_refnum+ddf->mppe, 0x00, 8*ddf->mppe);
vcl->next = ddf->conflist;
ddf->conflist = vcl;
* the phys_refnum and lba_offset for the newly created vd_config.
* We might also want to update the type in the phys_disk
* section.
+ *
+ * Alternately: fd == -1 and we have already chosen which device to
+ * use and recorded in dlist->raid_disk;
*/
struct dl *dl;
struct ddf_super *ddf = st->sb;
struct vd_config *vc;
__u64 *lba_offset;
- int working;
- int i;
+ unsigned int working;
+ unsigned int i;
unsigned long long blocks, pos, esize;
struct extent *ex;
- for (dl = ddf->dlist; dl ; dl = dl->next)
- if (dl->major == dk->major &&
- dl->minor == dk->minor)
- break;
+ if (fd == -1) {
+ for (dl = ddf->dlist; dl ; dl = dl->next)
+ if (dl->raiddisk == dk->raid_disk)
+ break;
+ } else {
+ for (dl = ddf->dlist; dl ; dl = dl->next)
+ if (dl->major == dk->major &&
+ dl->minor == dk->minor)
+ break;
+ }
if (!dl || ! (dk->state & (1<<MD_DISK_SYNC)))
return;
vc->phys_refnum[dk->raid_disk] = dl->disk.refnum;
lba_offset[dk->raid_disk] = __cpu_to_be64(pos);
- for (i=0; i < ddf->max_part ; i++)
+ for (i = 0; i < ddf->max_part ; i++)
if (dl->vlist[i] == NULL)
break;
if (i == ddf->max_part)
return;
dl->vlist[i] = ddf->currentconf;
- dl->fd = fd;
- dl->devname = devname;
+ if (fd >= 0)
+ dl->fd = fd;
+ if (devname)
+ dl->devname = devname;
/* Check how many working raid_disks, and if we can mark
* array as optimal yet
*/
working = 0;
- for (i=0; i < __be16_to_cpu(vc->prim_elmnt_count); i++)
+ for (i = 0; i < __be16_to_cpu(vc->prim_elmnt_count); i++)
if (vc->phys_refnum[i] != 0xffffffff)
working++;
/* add a device to a container, either while creating it or while
* expanding a pre-existing container
*/
-static void add_to_super_ddf(struct supertype *st,
+static int add_to_super_ddf(struct supertype *st,
mdu_disk_info_t *dk, int fd, char *devname)
{
struct ddf_super *ddf = st->sb;
struct tm *tm;
unsigned long long size;
struct phys_disk_entry *pde;
- int n, i;
+ unsigned int n, i;
struct stat stb;
if (ddf->currentconf) {
add_to_super_ddf_bvd(st, dk, fd, devname);
- return;
+ return 0;
}
/* This is device numbered dk->number. We need to create
fprintf(stderr, Name
": %s could allocate buffer for new disk, aborting\n",
__func__);
- abort();
+ return 1;
}
dd->major = major(stb.st_rdev);
dd->minor = minor(stb.st_rdev);
tm = localtime(&now);
sprintf(dd->disk.guid, "%8s%04d%02d%02d",
T10, tm->tm_year+1900, tm->tm_mon+1, tm->tm_mday);
- *(__u32*)(dd->disk.guid + 16) = random();
- *(__u32*)(dd->disk.guid + 20) = random();
+ *(__u32*)(dd->disk.guid + 16) = random32();
+ *(__u32*)(dd->disk.guid + 20) = random32();
do {
/* Cannot be bothered finding a CRC of some irrelevant details*/
- dd->disk.refnum = random();
- for (i = __be16_to_cpu(ddf->active->max_pd_entries) - 1;
- i >= 0; i--)
- if (ddf->phys->entries[i].refnum == dd->disk.refnum)
+ dd->disk.refnum = random32();
+ for (i = __be16_to_cpu(ddf->active->max_pd_entries);
+ i > 0; i--)
+ if (ddf->phys->entries[i-1].refnum == dd->disk.refnum)
break;
- } while (i >= 0);
+ } while (i > 0);
dd->disk.forced_ref = 1;
dd->disk.forced_guid = 1;
ddf->dlist = dd;
ddf->updates_pending = 1;
}
+
+ return 0;
}
/*
char *null_aligned = (char*)((((unsigned long)null_conf)+511)&~511UL);
if (null_conf[0] != 0xff)
memset(null_conf, 0xff, sizeof(null_conf));
- int togo = conf_size;
+ unsigned int togo = conf_size;
while (togo > sizeof(null_conf)-512) {
if (write(fd, null_aligned, sizeof(null_conf)-512) < 0)
break;
static int write_init_super_ddf(struct supertype *st)
{
+ struct ddf_super *ddf = st->sb;
+ struct vcl *currentconf = ddf->currentconf;
+
+ /* we are done with currentconf reset it to point st at the container */
+ ddf->currentconf = NULL;
if (st->update_tail) {
/* queue the virtual_disk and vd_config as metadata updates */
struct virtual_disk *vd;
struct vd_config *vc;
- struct ddf_super *ddf = st->sb;
int len;
- if (!ddf->currentconf) {
+ if (!currentconf) {
int len = (sizeof(struct phys_disk) +
sizeof(struct phys_disk_entry));
len = sizeof(struct virtual_disk) + sizeof(struct virtual_entry);
vd = malloc(len);
*vd = *ddf->virt;
- vd->entries[0] = ddf->virt->entries[ddf->currentconf->vcnum];
- vd->populated_vdes = __cpu_to_be16(ddf->currentconf->vcnum);
+ vd->entries[0] = ddf->virt->entries[currentconf->vcnum];
+ vd->populated_vdes = __cpu_to_be16(currentconf->vcnum);
append_metadata_update(st, vd, len);
/* Then the vd_config */
len = ddf->conf_rec_len * 512;
vc = malloc(len);
- memcpy(vc, &ddf->currentconf->conf, len);
+ memcpy(vc, ¤tconf->conf, len);
append_metadata_update(st, vc, len);
/* FIXME I need to close the fds! */
return 0;
- } else
+ } else {
+ struct dl *d;
+ for (d = ddf->dlist; d; d=d->next)
+ while (Kill(d->devname, NULL, 0, 1, 1) == 0);
return __write_init_super_ddf(st, 1);
+ }
}
#endif
}
#ifndef MDASSEMBLE
+
+static int reserve_space(struct supertype *st, int raiddisks,
+ unsigned long long size, int chunk,
+ unsigned long long *freesize)
+{
+ /* Find 'raiddisks' spare extents at least 'size' big (but
+ * only caring about multiples of 'chunk') and remember
+ * them.
+ * If the cannot be found, fail.
+ */
+ struct dl *dl;
+ struct ddf_super *ddf = st->sb;
+ int cnt = 0;
+
+ for (dl = ddf->dlist; dl ; dl=dl->next) {
+ dl->raiddisk = -1;
+ dl->esize = 0;
+ }
+ /* 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;
+
+ 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) {
+ cnt++;
+ dl->esize = minsize;
+ }
+ free(e);
+ }
+ if (cnt < raiddisks) {
+ fprintf(stderr, Name ": not enough devices with space to create array.\n");
+ return 0; /* No enough free spaces large enough */
+ }
+ if (size == 0) {
+ /* choose the largest size of which there are at least 'raiddisk' */
+ for (dl = ddf->dlist ; dl ; dl=dl->next) {
+ struct dl *dl2;
+ if (dl->esize <= size)
+ continue;
+ /* This is bigger than 'size', see if there are enough */
+ cnt = 0;
+ for (dl2 = dl; dl2 ; dl2=dl2->next)
+ if (dl2->esize >= dl->esize)
+ cnt++;
+ if (cnt >= raiddisks)
+ size = dl->esize;
+ }
+ if (chunk) {
+ size = size / chunk;
+ size *= chunk;
+ }
+ *freesize = size;
+ if (size < 32) {
+ fprintf(stderr, Name ": not enough spare devices to create array.\n");
+ return 0;
+ }
+ }
+ /* We have a 'size' of which there are enough spaces.
+ * We simply do a first-fit */
+ cnt = 0;
+ for (dl = ddf->dlist ; dl && cnt < raiddisks ; dl=dl->next) {
+ if (dl->esize < size)
+ continue;
+
+ dl->raiddisk = cnt;
+ cnt++;
+ }
+ return 1;
+}
+
+
+
static int
validate_geometry_ddf_container(struct supertype *st,
int level, int layout, int raiddisks,
verbose);
}
- if (st->sb) {
- /* A container has already been opened, so we are
- * creating in there. Maybe a BVD, maybe an SVD.
- * Should make a distinction one day.
- */
- return validate_geometry_ddf_bvd(st, level, layout, raiddisks,
- chunk, size, dev, freesize,
- verbose);
- }
if (!dev) {
/* Initial sanity check. Exclude illegal levels. */
int i;
for (i=0; ddf_level_num[i].num1 != MAXINT; i++)
if (ddf_level_num[i].num2 == level)
break;
- if (ddf_level_num[i].num1 == MAXINT)
+ if (ddf_level_num[i].num1 == MAXINT) {
+ if (verbose)
+ fprintf(stderr, Name ": DDF does not support level %d arrays\n",
+ level);
return 0;
+ }
/* Should check layout? etc */
+
+ if (st->sb && freesize) {
+ /* --create was given a container to create in.
+ * So we need to check that there are enough
+ * free spaces and return the amount of space.
+ * We may as well remember which drives were
+ * chosen so that add_to_super/getinfo_super
+ * can return them.
+ */
+ return reserve_space(st, raiddisks, size, chunk, freesize);
+ }
return 1;
}
+ if (st->sb) {
+ /* A container has already been opened, so we are
+ * creating in there. Maybe a BVD, maybe an SVD.
+ * Should make a distinction one day.
+ */
+ return validate_geometry_ddf_bvd(st, level, layout, raiddisks,
+ chunk, size, dev, freesize,
+ verbose);
+ }
/* This is the first device for the array.
* If it is a container, we read it in and do automagic allocations,
* no other devices should be given.
if (verbose)
fprintf(stderr,
Name ": ddf: Cannot create this array "
- "on device %s\n",
+ "on device %s - a container is required.\n",
dev);
return 0;
}
close(fd);
*freesize = avail_size_ddf(st, ldsize >> 9);
+ if (*freesize == 0)
+ return 0;
return 1;
}
struct extent *e;
int i;
/* ddf/bvd supports lots of things, but not containers */
- if (level == LEVEL_CONTAINER)
+ if (level == LEVEL_CONTAINER) {
+ if (verbose)
+ fprintf(stderr, Name ": DDF cannot create a container within an container\n");
return 0;
+ }
/* We must have the container info already read in. */
if (!ddf)
return 0;
if ((S_IFMT & stb.st_mode) != S_IFBLK)
return 0;
for (dl = ddf->dlist ; dl ; dl = dl->next) {
- if (dl->major == major(stb.st_rdev) &&
- dl->minor == minor(stb.st_rdev))
+ if (dl->major == (int)major(stb.st_rdev) &&
+ dl->minor == (int)minor(stb.st_rdev))
break;
}
if (!dl) {
return 1;
}
if (st->subarray[0]) {
+ unsigned long val;
struct vcl *v;
+ char *ep;
+
+ val = strtoul(st->subarray, &ep, 10);
+ if (*ep != '\0') {
+ free(super);
+ return 1;
+ }
for (v = super->conflist; v; v = v->next)
- if (v->vcnum == atoi(st->subarray))
+ if (v->vcnum == val)
super->currentconf = v;
- if (!super->currentconf)
+ if (!super->currentconf) {
+ free(super);
return 1;
+ }
}
+
*sbp = super;
if (st->ss == NULL) {
st->ss = &super_ddf;
for (vc = ddf->conflist ; vc ; vc=vc->next)
{
- int i;
+ unsigned int i;
+ unsigned int j;
struct mdinfo *this;
this = malloc(sizeof(*this));
memset(this, 0, sizeof(*this));
this->resync_start = 0;
} else {
this->array.state = 1;
- this->resync_start = ~0ULL;
+ this->resync_start = MaxSector;
}
- memcpy(this->name, ddf->virt->entries[i].name, 32);
- this->name[32]=0;
+ memcpy(this->name, ddf->virt->entries[i].name, 16);
+ this->name[16]=0;
+ for(j=0; j<16; j++)
+ if (this->name[j] == ' ')
+ this->name[j] = 0;
memset(this->uuid, 0, sizeof(this->uuid));
this->component_size = __be64_to_cpu(vc->conf.blocks);
devnum2devname(st->container_dev),
this->container_member);
- for (i=0 ; i < ddf->mppe ; i++) {
+ for (i = 0 ; i < ddf->mppe ; i++) {
struct mdinfo *dev;
struct dl *d;
+ int stt;
if (vc->conf.phys_refnum[i] == 0xFFFFFFFF)
continue;
- this->array.working_disks++;
-
for (d = ddf->dlist; d ; d=d->next)
if (d->disk.refnum == vc->conf.phys_refnum[i])
break;
if (d == NULL)
- break;
+ /* Haven't found that one yet, maybe there are others */
+ continue;
+ stt = __be16_to_cpu(ddf->phys->entries[d->pdnum].state);
+ if ((stt & (DDF_Online|DDF_Failed|DDF_Rebuilding))
+ != DDF_Online)
+ continue;
+
+ this->array.working_disks++;
dev = malloc(sizeof(*dev));
memset(dev, 0, sizeof(*dev));
dev->disk.minor = d->minor;
dev->disk.raid_disk = i;
dev->disk.state = (1<<MD_DISK_SYNC)|(1<<MD_DISK_ACTIVE);
+ dev->recovery_start = MaxSector;
dev->events = __be32_to_cpu(ddf->primary.seq);
dev->data_offset = __be64_to_cpu(vc->lba_offset[i]);
return rest;
}
-static int store_zero_ddf(struct supertype *st, int fd)
+static int store_super_ddf(struct supertype *st, int fd)
{
+ struct ddf_super *ddf = st->sb;
unsigned long long dsize;
void *buf;
int rc;
+ if (!ddf)
+ return 1;
+
+ /* ->dlist and ->conflist will be set for updates, currently not
+ * supported
+ */
+ if (ddf->dlist || ddf->conflist)
+ return 1;
+
if (!get_dev_size(fd, NULL, &dsize))
return 1;
if (consistent == 2) {
/* Should check if a recovery should be started FIXME */
consistent = 1;
- if (!is_resync_complete(a))
+ if (!is_resync_complete(&a->info))
consistent = 0;
}
if (consistent)
old = ddf->virt->entries[inst].init_state;
ddf->virt->entries[inst].init_state &= ~DDF_initstate_mask;
- if (is_resync_complete(a))
+ if (is_resync_complete(&a->info))
ddf->virt->entries[inst].init_state |= DDF_init_full;
- else if (a->resync_start == 0)
+ else if (a->info.resync_start == 0)
ddf->virt->entries[inst].init_state |= DDF_init_not;
else
ddf->virt->entries[inst].init_state |= DDF_init_quick;
ddf->updates_pending = 1;
dprintf("ddf mark %d %s %llu\n", inst, consistent?"clean":"dirty",
- a->resync_start);
+ a->info.resync_start);
return consistent;
}
+#define container_of(ptr, type, member) ({ \
+ const typeof( ((type *)0)->member ) *__mptr = (ptr); \
+ (type *)( (char *)__mptr - offsetof(type,member) );})
/*
* The state of each disk is stored in the global phys_disk structure
* in phys_disk.entries[n].state.
static void ddf_set_disk(struct active_array *a, int n, int state)
{
struct ddf_super *ddf = a->container->sb;
- int inst = a->info.container_member;
+ unsigned int inst = a->info.container_member;
struct vd_config *vc = find_vdcr(ddf, inst);
int pd = find_phys(ddf, vc->phys_refnum[n]);
int i, st, working;
+ struct mdinfo *mdi;
+ struct dl *dl;
if (vc == NULL) {
dprintf("ddf: cannot find instance %d!!\n", inst);
return;
}
- if (pd < 0) {
- /* disk doesn't currently exist. If it is now in_sync,
- * insert it. */
+ /* Find the matching slot in 'info'. */
+ for (mdi = a->info.devs; mdi; mdi = mdi->next)
+ if (mdi->disk.raid_disk == n)
+ break;
+ if (!mdi)
+ return;
+
+ /* and find the 'dl' entry corresponding to that. */
+ for (dl = ddf->dlist; dl; dl = dl->next)
+ if (mdi->state_fd >= 0 &&
+ mdi->disk.major == dl->major &&
+ mdi->disk.minor == dl->minor)
+ break;
+ if (!dl)
+ return;
+
+ if (pd < 0 || pd != dl->pdnum) {
+ /* disk doesn't currently exist or has changed.
+ * If it is now in_sync, insert it. */
if ((state & DS_INSYNC) && ! (state & DS_FAULTY)) {
- /* Find dev 'n' in a->info->devs, determine the
- * ddf refnum, and set vc->phys_refnum and update
- * phys->entries[]
- */
- /* FIXME */
+ struct vcl *vcl;
+ pd = dl->pdnum;
+ vc->phys_refnum[n] = dl->disk.refnum;
+ vcl = container_of(vc, struct vcl, conf);
+ vcl->lba_offset[n] = mdi->data_offset;
+ ddf->phys->entries[pd].type &=
+ ~__cpu_to_be16(DDF_Global_Spare);
+ ddf->phys->entries[pd].type |=
+ __cpu_to_be16(DDF_Active_in_VD);
+ ddf->updates_pending = 1;
}
} else {
int old = ddf->phys->entries[pd].state;
struct vd_config *vc;
struct vcl *vcl;
struct dl *dl;
- int mppe;
- int ent;
+ unsigned int mppe;
+ unsigned int ent;
dprintf("Process update %x\n", *magic);
dprintf("len %d %d\n", update->len, ddf->conf_rec_len);
mppe = __be16_to_cpu(ddf->anchor.max_primary_element_entries);
- if (update->len != ddf->conf_rec_len * 512)
+ if ((unsigned)update->len != ddf->conf_rec_len * 512)
return;
vc = (struct vd_config*)update->buf;
for (vcl = ddf->conflist; vcl ; vcl = vcl->next)
mppe * (sizeof(__u32) + sizeof(__u64)));
} else {
/* A new VD_CONF */
+ if (!update->space)
+ return;
vcl = update->space;
update->space = NULL;
vcl->next = ddf->conflist;
memcpy(&vcl->conf, vc, update->len);
vcl->lba_offset = (__u64*)
&vcl->conf.phys_refnum[mppe];
+ for (ent = 0;
+ ent < __be16_to_cpu(ddf->virt->populated_vdes);
+ ent++)
+ if (memcmp(vc->guid, ddf->virt->entries[ent].guid,
+ DDF_GUID_LEN) == 0) {
+ vcl->vcnum = ent;
+ break;
+ }
ddf->conflist = vcl;
}
/* Now make sure vlist is correct for each dl. */
for (dl = ddf->dlist; dl; dl = dl->next) {
- int dn;
- int vn = 0;
+ unsigned int dn;
+ unsigned int vn = 0;
+ int in_degraded = 0;
for (vcl = ddf->conflist; vcl ; vcl = vcl->next)
for (dn=0; dn < ddf->mppe ; dn++)
if (vcl->conf.phys_refnum[dn] ==
dl->disk.refnum) {
+ int vstate;
dprintf("dev %d has %p at %d\n",
dl->pdnum, vcl, vn);
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;
break;
}
while (vn < ddf->max_part)
if (dl->vlist[0]) {
ddf->phys->entries[dl->pdnum].type &=
~__cpu_to_be16(DDF_Global_Spare);
- ddf->phys->entries[dl->pdnum].type |=
- __cpu_to_be16(DDF_Active_in_VD);
+ if (!(ddf->phys->entries[dl->pdnum].type &
+ __cpu_to_be16(DDF_Active_in_VD))) {
+ ddf->phys->entries[dl->pdnum].type |=
+ __cpu_to_be16(DDF_Active_in_VD);
+ if (in_degraded)
+ ddf->phys->entries[dl->pdnum].state |=
+ __cpu_to_be16(DDF_Rebuilding);
+ }
}
if (dl->spare) {
ddf->phys->entries[dl->pdnum].type &=
struct ddf_super *ddf = st->sb;
__u32 *magic = (__u32*)update->buf;
if (*magic == DDF_VD_CONF_MAGIC)
- posix_memalign(&update->space, 512,
+ if (posix_memalign(&update->space, 512,
offsetof(struct vcl, conf)
- + ddf->conf_rec_len * 512);
+ + ddf->conf_rec_len * 512) != 0)
+ update->space = NULL;
}
/*
int is_global = 0;
int is_dedicated = 0;
struct extent *ex;
- int j;
+ unsigned int j;
/* If in this array, skip */
for (d2 = a->info.devs ; d2 ; d2 = d2->next)
- if (d2->disk.major == dl->major &&
+ if (d2->state_fd >= 0 &&
+ d2->disk.major == dl->major &&
d2->disk.minor == dl->minor) {
dprintf("%x:%x already in array\n", dl->major, dl->minor);
break;
esize = ex[j].start - pos;
if (esize >= a->info.component_size)
break;
- pos = ex[i].start + ex[i].size;
- i++;
- } while (ex[i-1].size);
+ pos = ex[j].start + ex[j].size;
+ j++;
+ } while (ex[j-1].size);
free(ex);
if (esize < a->info.component_size) {
- dprintf("%x:%x has no room: %llu %llu\n", dl->major, dl->minor,
+ dprintf("%x:%x has no room: %llu %llu\n",
+ dl->major, dl->minor,
esize, a->info.component_size);
/* No room */
continue;
/* Cool, we have a device with some space at pos */
di = malloc(sizeof(*di));
+ if (!di)
+ continue;
memset(di, 0, sizeof(*di));
di->disk.number = i;
di->disk.raid_disk = i;
di->disk.major = dl->major;
di->disk.minor = dl->minor;
di->disk.state = 0;
+ di->recovery_start = 0;
di->data_offset = pos;
di->component_size = a->info.component_size;
di->container_member = dl->pdnum;
* phys_refnum and lba_offset values
*/
mu = malloc(sizeof(*mu));
+ if (mu && posix_memalign(&mu->space, 512, sizeof(struct vcl)) != 0) {
+ free(mu);
+ mu = NULL;
+ }
+ if (!mu) {
+ while (rv) {
+ struct mdinfo *n = rv->next;
+
+ free(rv);
+ rv = n;
+ }
+ return NULL;
+ }
+
mu->buf = malloc(ddf->conf_rec_len * 512);
- posix_memalign(&mu->space, 512, sizeof(struct vcl));
- mu->len = ddf->conf_rec_len;
+ mu->len = ddf->conf_rec_len * 512;
+ mu->space = NULL;
mu->next = *updates;
vc = find_vdcr(ddf, a->info.container_member);
memcpy(mu->buf, vc, ddf->conf_rec_len * 512);
}
#endif /* MDASSEMBLE */
+static int ddf_level_to_layout(int level)
+{
+ switch(level) {
+ case 0:
+ case 1:
+ return 0;
+ case 5:
+ return ALGORITHM_LEFT_SYMMETRIC;
+ case 6:
+ return ALGORITHM_ROTATING_N_CONTINUE;
+ case 10:
+ return 0x102;
+ default:
+ return UnSet;
+ }
+}
+
struct superswitch super_ddf = {
#ifndef MDASSEMBLE
.examine_super = examine_super_ddf,
.brief_examine_super = brief_examine_super_ddf,
+ .brief_examine_subarrays = brief_examine_subarrays_ddf,
+ .export_examine_super = export_examine_super_ddf,
.detail_super = detail_super_ddf,
.brief_detail_super = brief_detail_super_ddf,
.validate_geometry = validate_geometry_ddf,
.load_super = load_super_ddf,
.init_super = init_super_ddf,
- .store_super = store_zero_ddf,
+ .store_super = store_super_ddf,
.free_super = free_super_ddf,
.match_metadata_desc = match_metadata_desc_ddf,
.container_content = container_content_ddf,
+ .default_layout = ddf_level_to_layout,
.external = 1,
.prepare_update = ddf_prepare_update,
.activate_spare = ddf_activate_spare,
#endif
+ .name = "ddf",
};
projects / thirdparty / mdadm.git / blobdiff