X-Git-Url: http://git.ipfire.org/?a=blobdiff_plain;f=super1.c;h=9b991e6c3e9ba9567d47dfb2d6108765c05d5fde;hb=4a3a795a8b287c3817fec2f93b3be6329d146730;hp=1ce26683ac856edeff85ba442b3153b031865fe2;hpb=cc3130a7863c21d773e1afad703b3f252debe075;p=thirdparty%2Fmdadm.git diff --git a/super1.c b/super1.c index 1ce26683..9b991e6c 100644 --- a/super1.c +++ b/super1.c @@ -22,6 +22,7 @@ * Email: */ +#include #include "mdadm.h" /* * The version-1 superblock : @@ -57,7 +58,7 @@ struct mdp_superblock_1 { __u64 reshape_position; /* next address in array-space for reshape */ __u32 delta_disks; /* change in number of raid_disks */ __u32 new_layout; /* new layout */ - __u32 new_chunk; /* new chunk size (bytes) */ + __u32 new_chunk; /* new chunk size (sectors) */ __u32 new_offset; /* signed number to add to data_offset in new * layout. 0 == no-change. This can be * different on each device in the array. @@ -133,9 +134,20 @@ struct misc_dev_info { |MD_FEATURE_NEW_OFFSET \ ) -#ifndef offsetof -#define offsetof(t,f) ((size_t)&(((t*)0)->f)) -#endif +/* return how many bytes are needed for bitmap, for cluster-md each node + * should have it's own bitmap */ +static unsigned int calc_bitmap_size(bitmap_super_t *bms, unsigned int boundary) +{ + unsigned long long bits, bytes; + + bits = __le64_to_cpu(bms->sync_size) / (__le32_to_cpu(bms->chunksize)>>9); + bytes = (bits+7) >> 3; + bytes += sizeof(bitmap_super_t); + bytes = ROUND_UP(bytes, boundary); + + return bytes; +} + static unsigned int calc_sb_1_csum(struct mdp_superblock_1 * sb) { unsigned int disk_csum, csum; @@ -199,8 +211,7 @@ static int aread(struct align_fd *afd, void *buf, int len) if (!bsize || bsize > 4096 || len > 4096) { if (!bsize) - fprintf(stderr, "WARNING - aread() called with " - "invalid block size\n"); + fprintf(stderr, "WARNING - aread() called with invalid block size\n"); return -1; } b = ROUND_UP_PTR((char *)abuf, 4096); @@ -232,8 +243,7 @@ static int awrite(struct align_fd *afd, void *buf, int len) bsize = afd->blk_sz; if (!bsize || bsize > 4096 || len > 4096) { if (!bsize) - fprintf(stderr, "WARNING - awrite() called with " - "invalid block size\n"); + fprintf(stderr, "WARNING - awrite() called with invalid block size\n"); return -1; } b = ROUND_UP_PTR((char *)abuf, 4096); @@ -260,6 +270,7 @@ static int awrite(struct align_fd *afd, void *buf, int len) static void examine_super1(struct supertype *st, char *homehost) { struct mdp_superblock_1 *sb = st->sb; + bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb)+MAX_SB_SIZE); time_t atime; unsigned int d; int role; @@ -269,6 +280,7 @@ static void examine_super1(struct supertype *st, char *homehost) int l = homehost ? strlen(homehost) : 0; int layout; unsigned long long sb_offset; + struct mdinfo info; printf(" Magic : %08x\n", __le32_to_cpu(sb->magic)); printf(" Version : 1"); @@ -292,6 +304,8 @@ static void examine_super1(struct supertype *st, char *homehost) strncmp(sb->set_name, homehost, l) == 0) printf(" (local to host %s)", homehost); printf("\n"); + if (bms->nodes > 0) + printf(" Cluster Name : %-64s\n", bms->cluster_name); atime = __le64_to_cpu(sb->ctime) & 0xFFFFFFFFFFULL; printf(" Creation Time : %.24s\n", ctime(&atime)); c=map_num(pers, __le32_to_cpu(sb->level)); @@ -337,6 +351,13 @@ static void examine_super1(struct supertype *st, char *homehost) (unsigned long long)__le64_to_cpu(sb->super_offset)); if (__le32_to_cpu(sb->feature_map) & MD_FEATURE_RECOVERY_OFFSET) printf("Recovery Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb->recovery_offset)); + + st->ss->getinfo_super(st, &info, NULL); + if (info.space_after != 1 && + !(__le32_to_cpu(sb->feature_map) & MD_FEATURE_NEW_OFFSET)) + printf(" Unused Space : before=%llu sectors, after=%llu sectors\n", + info.space_before, info.space_after); + printf(" State : %s\n", (__le64_to_cpu(sb->resync_offset)+1)? "active":"clean"); printf(" Device UUID : "); for (i=0; i<16; i++) { @@ -396,14 +417,13 @@ static void examine_super1(struct supertype *st, char *homehost) if (sb->bblog_size && sb->bblog_offset) { printf(" Bad Block Log : %d entries available at offset %ld sectors", __le16_to_cpu(sb->bblog_size)*512/8, - (long)__le32_to_cpu(sb->bblog_offset)); + (long)(int32_t)__le32_to_cpu(sb->bblog_offset)); if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BAD_BLOCKS)) printf(" - bad blocks present."); printf("\n"); } - if (calc_sb_1_csum(sb) == sb->sb_csum) printf(" Checksum : %x - correct\n", __le32_to_cpu(sb->sb_csum)); else @@ -499,7 +519,6 @@ static void examine_super1(struct supertype *st, char *homehost) printf("\n"); } - static void brief_examine_super1(struct supertype *st, int verbose) { struct mdp_superblock_1 *sb = st->sb; @@ -679,12 +698,8 @@ static int copy_metadata1(struct supertype *st, int from, int to) /* have the header, can calculate * correct bitmap bytes */ bitmap_super_t *bms; - int bits; bms = (void*)buf; - bits = __le64_to_cpu(bms->sync_size) / (__le32_to_cpu(bms->chunksize)>>9); - bytes = (bits+7) >> 3; - bytes += sizeof(bitmap_super_t); - bytes = ROUND_UP(bytes, 512); + bytes = calc_bitmap_size(bms, 512); if (n > bytes) n = bytes; } @@ -738,6 +753,7 @@ err: static void detail_super1(struct supertype *st, char *homehost) { struct mdp_superblock_1 *sb = st->sb; + bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MAX_SB_SIZE); int i; int l = homehost ? strlen(homehost) : 0; @@ -746,6 +762,8 @@ static void detail_super1(struct supertype *st, char *homehost) sb->set_name[l] == ':' && strncmp(sb->set_name, homehost, l) == 0) printf(" (local to host %s)", homehost); + if (bms->nodes > 0) + printf("\n Cluster Name : %-64s", bms->cluster_name); printf("\n UUID : "); for (i=0; i<16; i++) { if ((i&3)==0 && i != 0) printf(":"); @@ -807,7 +825,7 @@ static int examine_badblocks_super1(struct supertype *st, int fd, char *devname) size = __le32_to_cpu(sb->bblog_size)* 512; if (posix_memalign((void**)&bbl, 4096, size) != 0) { - pr_err("%s could not allocate badblocks list\n", __func__); + pr_err("could not allocate badblocks list\n"); return 0; } offset = __le64_to_cpu(sb->super_offset) + @@ -889,6 +907,8 @@ static void getinfo_super1(struct supertype *st, struct mdinfo *info, char *map) info->array.state = (__le64_to_cpu(sb->resync_offset) == MaxSector) ? 1 : 0; + if (__le32_to_cpu(bsb->nodes) > 1) + info->array.state |= (1 << MD_SB_CLUSTERED); info->data_offset = __le64_to_cpu(sb->data_offset); info->component_size = __le64_to_cpu(sb->size); @@ -905,7 +925,10 @@ static void getinfo_super1(struct supertype *st, struct mdinfo *info, char *map) role = __le16_to_cpu(sb->dev_roles[__le32_to_cpu(sb->dev_number)]); super_offset = __le64_to_cpu(sb->super_offset); - data_size = __le64_to_cpu(sb->size); + if (info->array.level <= 0) + data_size = __le64_to_cpu(sb->data_size); + else + data_size = __le64_to_cpu(sb->size); if (info->data_offset < super_offset) { unsigned long long end; info->space_before = info->data_offset; @@ -953,6 +976,14 @@ static void getinfo_super1(struct supertype *st, struct mdinfo *info, char *map) info->space_before = 0; info->space_after = misc->device_size - data_size - info->data_offset; } + if (info->space_before == 0 && info->space_after == 0) { + /* It will look like we don't support data_offset changes, + * be we do - it's just that there is no room. + * A change that reduced the number of devices should + * still be allowed, so set the otherwise useless value of '1' + */ + info->space_after = 1; + } info->disk.raid_disk = -1; switch(role) { @@ -982,7 +1013,6 @@ static void getinfo_super1(struct supertype *st, struct mdinfo *info, char *map) info->disk.state |= 1 << MD_DISK_REPLACEMENT; } - if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RECOVERY_OFFSET)) info->recovery_start = __le32_to_cpu(sb->recovery_offset); else @@ -990,7 +1020,8 @@ static void getinfo_super1(struct supertype *st, struct mdinfo *info, char *map) if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE)) { info->reshape_active = 1; - if (info->array.level == 10) + if ((sb->feature_map & __le32_to_cpu(MD_FEATURE_NEW_OFFSET)) && + sb->new_offset != 0) info->reshape_active |= RESHAPE_NO_BACKUP; info->reshape_progress = __le64_to_cpu(sb->reshape_position); info->new_level = __le32_to_cpu(sb->new_level); @@ -1166,7 +1197,7 @@ static int update_super1(struct supertype *st, struct mdinfo *info, */ unsigned long long sb_offset = __le64_to_cpu(sb->super_offset); unsigned long long data_offset = __le64_to_cpu(sb->data_offset); - long bitmap_offset = (long)__le64_to_cpu(sb->bitmap_offset); + long bitmap_offset = (long)(int32_t)__le32_to_cpu(sb->bitmap_offset); long bm_sectors = 0; long space; @@ -1178,16 +1209,17 @@ static int update_super1(struct supertype *st, struct mdinfo *info, } #endif if (sb_offset < data_offset) { - /* 1.1 or 1.2. Put bbl just before data + /* 1.1 or 1.2. Put bbl after bitmap leaving at least 32K */ long bb_offset; - space = data_offset - sb_offset; - bb_offset = space - 8; + bb_offset = sb_offset + 8; if (bm_sectors && bitmap_offset > 0) - space -= (bitmap_offset + bm_sectors); - else - space -= 8; /* The superblock */ - if (space >= 8) { + bb_offset = bitmap_offset + bm_sectors; + while (bb_offset < (long)sb_offset + 8 + 32*2 + && bb_offset + 8+8 <= (long)data_offset) + /* too close to bitmap, and room to grow */ + bb_offset += 8; + if (bb_offset + 8 <= (long)data_offset) { sb->bblog_size = __cpu_to_le16(8); sb->bblog_offset = __cpu_to_le32(bb_offset); } @@ -1229,12 +1261,65 @@ static int update_super1(struct supertype *st, struct mdinfo *info, /* set data_size to device size less data_offset */ struct misc_dev_info *misc = (struct misc_dev_info*) (st->sb + MAX_SB_SIZE + BM_SUPER_SIZE); - printf("Size was %llu\n", (unsigned long long) - __le64_to_cpu(sb->data_size)); sb->data_size = __cpu_to_le64( misc->device_size - __le64_to_cpu(sb->data_offset)); - printf("Size is %llu\n", (unsigned long long) - __le64_to_cpu(sb->data_size)); + } else if (strcmp(update, "revert-reshape") == 0) { + rv = -2; + if (!(sb->feature_map & __cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE))) + pr_err("No active reshape to revert on %s\n", + devname); + else { + __u32 temp; + unsigned long long reshape_sectors; + long reshape_chunk; + rv = 0; + /* reshape_position is a little messy. + * Its value must be a multiple of the larger + * chunk size, and of the "after" data disks. + * So when reverting we need to change it to + * be a multiple of the new "after" data disks, + * which is the old "before". + * If it isn't already a multiple of 'before', + * the only thing we could do would be + * copy some block around on the disks, which + * is easy to get wrong. + * So we reject a revert-reshape unless the + * alignment is good. + */ + if (__le32_to_cpu(sb->level) >= 4 && + __le32_to_cpu(sb->level) <= 6) { + reshape_sectors = __le64_to_cpu(sb->reshape_position); + reshape_chunk = __le32_to_cpu(sb->new_chunk); + reshape_chunk *= __le32_to_cpu(sb->raid_disks) - __le32_to_cpu(sb->delta_disks) - + (__le32_to_cpu(sb->level)==6 ? 2 : 1); + if (reshape_sectors % reshape_chunk) { + pr_err("Reshape position is not suitably aligned.\n"); + pr_err("Try normal assembly and stop again\n"); + return -2; + } + } + sb->raid_disks = __cpu_to_le32(__le32_to_cpu(sb->raid_disks) - + __le32_to_cpu(sb->delta_disks)); + if (sb->delta_disks == 0) + sb->feature_map ^= __cpu_to_le32(MD_FEATURE_RESHAPE_BACKWARDS); + else + sb->delta_disks = __cpu_to_le32(-__le32_to_cpu(sb->delta_disks)); + + temp = sb->new_layout; + sb->new_layout = sb->layout; + sb->layout = temp; + + temp = sb->new_chunk; + sb->new_chunk = sb->chunksize; + sb->chunksize = temp; + + if (sb->feature_map & __cpu_to_le32(MD_FEATURE_NEW_OFFSET)) { + long offset_delta = (int32_t)__le32_to_cpu(sb->new_offset); + sb->data_offset = __cpu_to_le64(__le64_to_cpu(sb->data_offset) + offset_delta); + sb->new_offset = __cpu_to_le32(-offset_delta); + sb->data_size = __cpu_to_le64(__le64_to_cpu(sb->data_size) - offset_delta); + } + } } else if (strcmp(update, "_reshape_progress")==0) sb->reshape_position = __cpu_to_le64(info->reshape_progress); else if (strcmp(update, "writemostly")==0) @@ -1259,7 +1344,7 @@ static int init_super1(struct supertype *st, mdu_array_info_t *info, int sbsize; if (posix_memalign((void**)&sb, 4096, SUPER1_SIZE) != 0) { - pr_err("%s could not allocate superblock\n", __func__); + pr_err("could not allocate superblock\n"); return 0; } memset(sb, 0, SUPER1_SIZE); @@ -1423,8 +1508,6 @@ static int store_super1(struct supertype *st, int fd) return -EINVAL; } - - if (sb_offset != __le64_to_cpu(sb->super_offset) && 0 != __le64_to_cpu(sb->super_offset) ) { @@ -1480,14 +1563,13 @@ static int write_init_super1(struct supertype *st) int rfd; int rv = 0; unsigned long long bm_space; - unsigned long long reserved; struct devinfo *di; unsigned long long dsize, array_size; - unsigned long long sb_offset, headroom; + unsigned long long sb_offset; unsigned long long data_offset; for (di = st->info; di; di = di->next) { - if (di->disk.state == 1) + if (di->disk.state & (1 << MD_DISK_FAULTY)) continue; if (di->fd < 0) continue; @@ -1512,7 +1594,7 @@ static int write_init_super1(struct supertype *st) sb->events = 0; refst = dup_super(st); - if (load_super1(refst, di->fd, NULL)==0) { + if (load_super1(refst, di->fd, NULL)==0) { struct mdp_superblock_1 *refsb = refst->sb; memcpy(sb->device_uuid, refsb->device_uuid, 16); @@ -1542,7 +1624,6 @@ static int write_init_super1(struct supertype *st) goto error_out; } - /* * Calculate the position of the superblock. * It is always aligned to a 4K boundary and @@ -1550,35 +1631,25 @@ static int write_init_super1(struct supertype *st) * 0: At least 8K, but less than 12K, from end of device * 1: At start of device * 2: 4K from start of device. - * Depending on the array size, we might leave extra space - * for a bitmap. - * Also leave 4K for bad-block log. + * data_offset has already been set. */ array_size = __le64_to_cpu(sb->size); /* work out how much space we left for a bitmap, * Add 8 sectors for bad block log */ bm_space = choose_bm_space(array_size) + 8; - /* We try to leave 0.1% at the start for reshape - * operations, but limit this to 128Meg (0.1% of 10Gig) - * which is plenty for efficient reshapes - * However we make it at least 2 chunks as one chunk - * is minimum needed for reshape. - */ - headroom = 128 * 1024 * 2; - while (headroom << 10 > array_size && - headroom/2 >= __le32_to_cpu(sb->chunksize) * 2) - headroom >>= 1; - data_offset = di->data_offset; + if (data_offset == INVALID_SECTORS) + data_offset = st->data_offset; switch(st->minor_version) { case 0: + if (data_offset == INVALID_SECTORS) + data_offset = 0; sb_offset = dsize; sb_offset -= 8*2; sb_offset &= ~(4*2-1); + sb->data_offset = __cpu_to_le64(data_offset); sb->super_offset = __cpu_to_le64(sb_offset); - if (data_offset == INVALID_SECTORS) - sb->data_offset = 0; if (sb_offset < array_size + bm_space) bm_space = sb_offset - array_size; sb->data_size = __cpu_to_le64(sb_offset - bm_space); @@ -1589,78 +1660,54 @@ static int write_init_super1(struct supertype *st) break; case 1: sb->super_offset = __cpu_to_le64(0); - if (data_offset == INVALID_SECTORS) { - reserved = bm_space + 4*2; - if (reserved < headroom) - reserved = headroom; - if (reserved + array_size > dsize) - reserved = dsize - array_size; - /* Try for multiple of 1Meg so it is nicely aligned */ - #define ONE_MEG (2*1024) - if (reserved > ONE_MEG) - reserved = (reserved/ONE_MEG) * ONE_MEG; - - /* force 4K alignment */ - reserved &= ~7ULL; - - } else - reserved = data_offset; - - sb->data_offset = __cpu_to_le64(reserved); - sb->data_size = __cpu_to_le64(dsize - reserved); - if (reserved >= 16) { + if (data_offset == INVALID_SECTORS) + data_offset = 16; + + sb->data_offset = __cpu_to_le64(data_offset); + sb->data_size = __cpu_to_le64(dsize - data_offset); + if (data_offset >= 8 + 32*2 + 8) { sb->bblog_size = __cpu_to_le16(8); - sb->bblog_offset = __cpu_to_le32(reserved-8); + sb->bblog_offset = __cpu_to_le32(8 + 32*2); + } else if (data_offset >= 16) { + sb->bblog_size = __cpu_to_le16(8); + sb->bblog_offset = __cpu_to_le32(data_offset-8); } break; case 2: sb_offset = 4*2; - sb->super_offset = __cpu_to_le64(4*2); - if (data_offset == INVALID_SECTORS) { - if (4*2 + 4*2 + bm_space + array_size - > dsize) - bm_space = dsize - array_size - - 4*2 - 4*2; - - reserved = bm_space + 4*2 + 4*2; - if (reserved < headroom) - reserved = headroom; - if (reserved + array_size > dsize) - reserved = dsize - array_size; - /* Try for multiple of 1Meg so it is nicely aligned */ - #define ONE_MEG (2*1024) - if (reserved > ONE_MEG) - reserved = (reserved/ONE_MEG) * ONE_MEG; - - /* force 4K alignment */ - reserved &= ~7ULL; - - } else - reserved = data_offset; - - sb->data_offset = __cpu_to_le64(reserved); - sb->data_size = __cpu_to_le64(dsize - reserved); - if (reserved >= 16+16) { + sb->super_offset = __cpu_to_le64(sb_offset); + if (data_offset == INVALID_SECTORS) + data_offset = 24; + + sb->data_offset = __cpu_to_le64(data_offset); + sb->data_size = __cpu_to_le64(dsize - data_offset); + if (data_offset >= 16 + 32*2 + 8) { + sb->bblog_size = __cpu_to_le16(8); + sb->bblog_offset = __cpu_to_le32(8 + 32*2); + } else if (data_offset >= 16+16) { sb->bblog_size = __cpu_to_le16(8); /* '8' sectors for the bblog, and another '8' * because we want offset from superblock, not * start of device. */ - sb->bblog_offset = __cpu_to_le32(reserved-8-8); + sb->bblog_offset = __cpu_to_le32(data_offset-8-8); } break; default: - pr_err("Failed to write invalid " - "metadata format 1.%i to %s\n", + pr_err("Failed to write invalid metadata format 1.%i to %s\n", st->minor_version, di->devname); rv = -EINVAL; goto out; } + if (conf_get_create_info()->bblist == 0) { + sb->bblog_size = 0; + sb->bblog_offset = 0; + } sb->sb_csum = calc_sb_1_csum(sb); rv = store_super1(st, di->fd); if (rv == 0 && (__le32_to_cpu(sb->feature_map) & 1)) - rv = st->ss->write_bitmap(st, di->fd); + rv = st->ss->write_bitmap(st, di->fd, NoUpdate); close(di->fd); di->fd = -1; if (rv) @@ -1694,7 +1741,7 @@ static int compare_super1(struct supertype *st, struct supertype *tst) if (!first) { if (posix_memalign((void**)&first, 4096, SUPER1_SIZE) != 0) { - pr_err("%s could not allocate superblock\n", __func__); + pr_err("could not allocate superblock\n"); return 1; } memcpy(first, second, SUPER1_SIZE); @@ -1805,8 +1852,7 @@ static int load_super1(struct supertype *st, int fd, char *devname) } if (posix_memalign((void**)&super, 4096, SUPER1_SIZE) != 0) { - pr_err("%s could not allocate superblock\n", - __func__); + pr_err("could not allocate superblock\n"); return 1; } @@ -1846,6 +1892,8 @@ static int load_super1(struct supertype *st, int fd, char *devname) misc = (struct misc_dev_info*) (((char*)super)+MAX_SB_SIZE+BM_SUPER_SIZE); misc->device_size = dsize; + if (st->data_offset == INVALID_SECTORS) + st->data_offset = __le64_to_cpu(super->data_offset); /* Now check on the bitmap superblock */ if ((__le32_to_cpu(super->feature_map)&MD_FEATURE_BITMAP_OFFSET) == 0) @@ -1870,7 +1918,6 @@ static int load_super1(struct supertype *st, int fd, char *devname) return 0; } - static struct supertype *match_metadata_desc1(char *arg) { struct supertype *st = xcalloc(1, sizeof(*st)); @@ -1879,6 +1926,7 @@ static struct supertype *match_metadata_desc1(char *arg) st->ss = &super1; st->max_devs = MAX_DEVS; st->sb = NULL; + st->data_offset = INVALID_SECTORS; /* leading zeros can be safely ignored. --detail generates them. */ while (*arg == '0') arg++; @@ -1915,19 +1963,17 @@ static struct supertype *match_metadata_desc1(char *arg) * superblock type st, and reserving 'reserve' sectors for * a possible bitmap */ -static __u64 _avail_size1(struct supertype *st, __u64 devsize, - unsigned long long data_offset, int chunksize) +static __u64 avail_size1(struct supertype *st, __u64 devsize, + unsigned long long data_offset) { struct mdp_superblock_1 *super = st->sb; int bmspace = 0; + int bbspace = 0; if (devsize < 24) return 0; - if (super == NULL) - /* creating: allow suitable space for bitmap */ - bmspace = choose_bm_space(devsize); #ifndef MDASSEMBLE - else if (__le32_to_cpu(super->feature_map)&MD_FEATURE_BITMAP_OFFSET) { + if (__le32_to_cpu(super->feature_map)&MD_FEATURE_BITMAP_OFFSET) { /* hot-add. allow for actual size of bitmap */ struct bitmap_super_s *bsb; bsb = (struct bitmap_super_s *)(((char*)super)+MAX_SB_SIZE); @@ -1935,20 +1981,20 @@ static __u64 _avail_size1(struct supertype *st, __u64 devsize, } #endif /* Allow space for bad block log */ - if (super && super->bblog_size) - devsize -= __le16_to_cpu(super->bblog_size); - else - devsize -= 8; - + if (super->bblog_size) + bbspace = __le16_to_cpu(super->bblog_size); if (st->minor_version < 0) /* not specified, so time to set default */ st->minor_version = 2; + if (data_offset == INVALID_SECTORS) + data_offset = st->data_offset; + if (data_offset != INVALID_SECTORS) switch(st->minor_version) { case 0: - return devsize - data_offset - 8*2; + return devsize - data_offset - 8*2 - bbspace; case 1: case 2: return devsize - data_offset; @@ -1958,36 +2004,19 @@ static __u64 _avail_size1(struct supertype *st, __u64 devsize, devsize -= bmspace; - if (super == NULL && st->minor_version > 0) { - /* haven't committed to a size yet, so allow some - * slack for space for reshape. - * Limit slack to 128M, but aim for about 0.1% - */ - unsigned long long headroom = 128*1024*2; - while ((headroom << 10) > devsize && - (chunksize == 0 || - headroom / 2 >= ((unsigned)chunksize*2)*2)) - headroom >>= 1; - devsize -= headroom; - } switch(st->minor_version) { case 0: /* at end */ - return ((devsize - 8*2 ) & ~(4*2-1)); + return ((devsize - 8*2 - bbspace ) & ~(4*2-1)); case 1: /* at start, 4K for superblock and possible bitmap */ - return devsize - 4*2; + return devsize - 4*2 - bbspace; case 2: /* 4k from start, 4K for superblock and possible bitmap */ - return devsize - (4+4)*2; + return devsize - (4+4)*2 - bbspace; } return 0; } -static __u64 avail_size1(struct supertype *st, __u64 devsize, - unsigned long long data_offset) -{ - return _avail_size1(st, devsize, data_offset, 0); -} static int add_internal_bitmap1(struct supertype *st, @@ -2016,7 +2045,6 @@ add_internal_bitmap1(struct supertype *st, bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MAX_SB_SIZE); int uuid[4]; - if (__le64_to_cpu(sb->data_size) == 0) /* Must be creating the array, else data_size would be non-zero */ creating = 1; @@ -2030,8 +2058,8 @@ add_internal_bitmap1(struct supertype *st, * been left. */ offset = 0; - room = choose_bm_space(__le64_to_cpu(sb->size)); bbl_size = 8; + room = choose_bm_space(__le64_to_cpu(sb->size)) + bbl_size; } else { room = __le64_to_cpu(sb->super_offset) - __le64_to_cpu(sb->data_offset) @@ -2044,7 +2072,7 @@ add_internal_bitmap1(struct supertype *st, bbl_size = -bbl_offset; if (!may_change || (room < 3*2 && - __le32_to_cpu(sb->max_dev) <= 384)) { + __le32_to_cpu(sb->max_dev) <= 384)) { room = 3*2; offset = 1*2; bbl_size = 0; @@ -2057,8 +2085,8 @@ add_internal_bitmap1(struct supertype *st, case 2: /* between superblock and data */ if (creating) { offset = 4*2; - room = choose_bm_space(__le64_to_cpu(sb->size)); bbl_size = 8; + room = choose_bm_space(__le64_to_cpu(sb->size)) + bbl_size; } else { room = __le64_to_cpu(sb->data_offset) - __le64_to_cpu(sb->super_offset); @@ -2087,6 +2115,10 @@ add_internal_bitmap1(struct supertype *st, /* Limit to 128K of bitmap when chunk size not requested */ room = 128*2; + if (room <= 1) + /* No room for a bitmap */ + return 0; + max_bits = (room * 512 - sizeof(bitmap_super_t)) * 8; min_chunk = 4096; /* sub-page chunks don't work yet.. */ @@ -2130,6 +2162,10 @@ add_internal_bitmap1(struct supertype *st, bms->daemon_sleep = __cpu_to_le32(delay); bms->sync_size = __cpu_to_le64(size); bms->write_behind = __cpu_to_le32(write_behind); + bms->nodes = __cpu_to_le32(st->nodes); + if (st->cluster_name) + strncpy((char *)bms->cluster_name, + st->cluster_name, strlen(st->cluster_name)); *chunkp = chunk; return 1; @@ -2155,7 +2191,7 @@ static void locate_bitmap1(struct supertype *st, int fd) lseek64(fd, offset<<9, 0); } -static int write_bitmap1(struct supertype *st, int fd) +static int write_bitmap1(struct supertype *st, int fd, enum bitmap_update update) { struct mdp_superblock_1 *sb = st->sb; bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb)+MAX_SB_SIZE); @@ -2163,6 +2199,43 @@ static int write_bitmap1(struct supertype *st, int fd) void *buf; int towrite, n; struct align_fd afd; + unsigned int i = 0; + unsigned long long total_bm_space, bm_space_per_node; + + switch (update) { + case NameUpdate: + /* update cluster name */ + if (st->cluster_name) { + memset((char *)bms->cluster_name, 0, sizeof(bms->cluster_name)); + strncpy((char *)bms->cluster_name, st->cluster_name, 64); + } + break; + case NodeNumUpdate: + /* cluster md only supports superblock 1.2 now */ + if (st->minor_version != 2) { + pr_err("Warning: cluster md only works with superblock 1.2\n"); + return -EINVAL; + } + + /* Each node has an independent bitmap, it is necessary to calculate the + * space is enough or not, first get how many bytes for the total bitmap */ + bm_space_per_node = calc_bitmap_size(bms, 4096); + + total_bm_space = 512 * (__le64_to_cpu(sb->data_offset) - __le64_to_cpu(sb->super_offset)); + total_bm_space = total_bm_space - 4096; /* leave another 4k for superblock */ + + if (bm_space_per_node * st->nodes > total_bm_space) { + pr_err("Warning: The max num of nodes can't exceed %llu\n", + total_bm_space / bm_space_per_node); + return -ENOMEM; + } + + bms->nodes = __cpu_to_le32(st->nodes); + break; + case NoUpdate: + default: + break; + } init_afd(&afd, fd); @@ -2171,27 +2244,37 @@ static int write_bitmap1(struct supertype *st, int fd) if (posix_memalign(&buf, 4096, 4096)) return -ENOMEM; - memset(buf, 0xff, 4096); - memcpy(buf, (char *)bms, sizeof(bitmap_super_t)); - - towrite = __le64_to_cpu(bms->sync_size) / (__le32_to_cpu(bms->chunksize)>>9); - towrite = (towrite+7) >> 3; /* bits to bytes */ - towrite += sizeof(bitmap_super_t); - towrite = ROUND_UP(towrite, 512); - while (towrite > 0) { - n = towrite; - if (n > 4096) - n = 4096; - n = awrite(&afd, buf, n); - if (n > 0) - towrite -= n; + do { + /* Only the bitmap[0] should resync + * whole device on initial assembly + */ + if (i) + memset(buf, 0x00, 4096); else + memset(buf, 0xff, 4096); + memcpy(buf, (char *)bms, sizeof(bitmap_super_t)); + + towrite = calc_bitmap_size(bms, 4096); + while (towrite > 0) { + n = towrite; + if (n > 4096) + n = 4096; + n = awrite(&afd, buf, n); + if (n > 0) + towrite -= n; + else + break; + if (i) + memset(buf, 0x00, 4096); + else + memset(buf, 0xff, 4096); + } + fsync(fd); + if (towrite) { + rv = -2; break; - memset(buf, 0xff, 4096); - } - fsync(fd); - if (towrite) - rv = -2; + } + } while (++i < __le32_to_cpu(bms->nodes)); free(buf); return rv; @@ -2219,7 +2302,9 @@ static int validate_geometry1(struct supertype *st, int level, char *subdev, unsigned long long *freesize, int verbose) { - unsigned long long ldsize; + unsigned long long ldsize, devsize; + int bmspace; + unsigned long long headroom; int fd; if (level == LEVEL_CONTAINER) { @@ -2227,12 +2312,16 @@ static int validate_geometry1(struct supertype *st, int level, pr_err("1.x metadata does not support containers\n"); return 0; } - if (chunk && *chunk == UnSet) + if (*chunk == UnSet) *chunk = DEFAULT_CHUNK; if (!subdev) return 1; + if (st->minor_version < 0) + /* not specified, so time to set default */ + st->minor_version = 2; + fd = open(subdev, O_RDONLY|O_EXCL, 0); if (fd < 0) { if (verbose) @@ -2247,11 +2336,118 @@ static int validate_geometry1(struct supertype *st, int level, } close(fd); - *freesize = _avail_size1(st, ldsize >> 9, data_offset, *chunk); + devsize = ldsize >> 9; + if (devsize < 24) { + *freesize = 0; + return 0; + } + + /* creating: allow suitable space for bitmap */ + bmspace = choose_bm_space(devsize); + + if (data_offset == INVALID_SECTORS) + data_offset = st->data_offset; + if (data_offset == INVALID_SECTORS) + switch (st->minor_version) { + case 0: + data_offset = 0; + break; + case 1: + case 2: + /* Choose data offset appropriate for this device + * and use as default for whole array. + * The data_offset must allow for bitmap space + * and base metadata, should allow for some headroom + * for reshape, and should be rounded to multiple + * of 1M. + * Headroom is limited to 128M, but aim for about 0.1% + */ + headroom = 128*1024*2; + while ((headroom << 10) > devsize && + (*chunk == 0 || + headroom / 2 >= ((unsigned)(*chunk)*2)*2)) + headroom >>= 1; + data_offset = 12*2 + bmspace + headroom; + #define ONE_MEG (2*1024) + if (data_offset > ONE_MEG) + data_offset = (data_offset / ONE_MEG) * ONE_MEG; + break; + } + if (st->data_offset == INVALID_SECTORS) + st->data_offset = data_offset; + switch(st->minor_version) { + case 0: /* metadata at end. Round down and subtract space to reserve */ + devsize = (devsize & ~(4ULL*2-1)); + /* space for metadata, bblog, bitmap */ + devsize -= 8*2 + 8 + bmspace; + break; + case 1: + case 2: + devsize -= data_offset; + break; + } + *freesize = devsize; return 1; } #endif /* MDASSEMBLE */ +void *super1_make_v0(struct supertype *st, struct mdinfo *info, mdp_super_t *sb0) +{ + /* Create a v1.0 superblock based on 'info'*/ + void *ret; + struct mdp_superblock_1 *sb; + int i; + int rfd; + unsigned long long offset; + + if (posix_memalign(&ret, 4096, 1024) != 0) + return NULL; + sb = ret; + memset(ret, 0, 1024); + sb->magic = __cpu_to_le32(MD_SB_MAGIC); + sb->major_version = __cpu_to_le32(1); + + copy_uuid(sb->set_uuid, info->uuid, super1.swapuuid); + sprintf(sb->set_name, "%d", sb0->md_minor); + sb->ctime = __cpu_to_le32(info->array.ctime+1); + sb->level = __cpu_to_le32(info->array.level); + sb->layout = __cpu_to_le32(info->array.layout); + sb->size = __cpu_to_le64(info->component_size); + sb->chunksize = __cpu_to_le32(info->array.chunk_size/512); + sb->raid_disks = __cpu_to_le32(info->array.raid_disks); + if (info->array.level > 0) + sb->data_size = sb->size; + else + sb->data_size = st->ss->avail_size(st, st->devsize/512, 0); + sb->resync_offset = MaxSector; + sb->max_dev = __cpu_to_le32(MD_SB_DISKS); + sb->dev_number = __cpu_to_le32(info->disk.number); + sb->utime = __cpu_to_le64(info->array.utime); + + offset = st->devsize/512 - 8*2; + offset &= ~(4*2-1); + sb->super_offset = __cpu_to_le64(offset); + //*(__u64*)(st->other + 128 + 8 + 8) = __cpu_to_le64(offset); + + if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 || + read(rfd, sb->device_uuid, 16) != 16) { + __u32 r[4] = {random(), random(), random(), random()}; + memcpy(sb->device_uuid, r, 16); + } + if (rfd >= 0) + close(rfd); + + for (i = 0; i < MD_SB_DISKS; i++) { + int state = sb0->disks[i].state; + sb->dev_roles[i] = 0xFFFF; + if ((state & (1<dev_roles[i] = __cpu_to_le16(sb0->disks[i].raid_disk); + } + sb->sb_csum = calc_sb_1_csum(sb); + return ret; +} + struct superswitch super1 = { #ifndef MDASSEMBLE .examine_super = examine_super1,