X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libbcachefs%2Fsuper-io.c;h=332d41e1c0a32082988f70ded38bd004b35444c0;hb=9799b119c34d7be1ee96d143209cfe5fc543d92a;hp=738b68b5d35cbfad6460d4bf78f914b0cb383ffd;hpb=42cf74fd1d0ef58927967e6236988e86cfc0d086;p=bcachefs-tools-debian diff --git a/libbcachefs/super-io.c b/libbcachefs/super-io.c index 738b68b..332d41e 100644 --- a/libbcachefs/super-io.c +++ b/libbcachefs/super-io.c @@ -1,29 +1,90 @@ // SPDX-License-Identifier: GPL-2.0 #include "bcachefs.h" -#include "btree_update_interior.h" -#include "buckets.h" #include "checksum.h" +#include "counters.h" #include "disk_groups.h" #include "ec.h" #include "error.h" -#include "io.h" #include "journal.h" -#include "journal_io.h" #include "journal_sb.h" #include "journal_seq_blacklist.h" +#include "recovery.h" #include "replicas.h" #include "quota.h" +#include "sb-clean.h" +#include "sb-members.h" #include "super-io.h" #include "super.h" +#include "trace.h" #include "vstructs.h" -#include "counters.h" #include -#include #include -#include +static const struct blk_holder_ops bch2_sb_handle_bdev_ops = { +}; + +struct bch2_metadata_version { + u16 version; + const char *name; + u64 recovery_passes; +}; + +static const struct bch2_metadata_version bch2_metadata_versions[] = { +#define x(n, v, _recovery_passes) { \ + .version = v, \ + .name = #n, \ + .recovery_passes = _recovery_passes, \ +}, + BCH_METADATA_VERSIONS() +#undef x +}; + +void bch2_version_to_text(struct printbuf *out, unsigned v) +{ + const char *str = "(unknown version)"; + + for (unsigned i = 0; i < ARRAY_SIZE(bch2_metadata_versions); i++) + if (bch2_metadata_versions[i].version == v) { + str = bch2_metadata_versions[i].name; + break; + } + + prt_printf(out, "%u.%u: %s", BCH_VERSION_MAJOR(v), BCH_VERSION_MINOR(v), str); +} + +unsigned bch2_latest_compatible_version(unsigned v) +{ + if (!BCH_VERSION_MAJOR(v)) + return v; + + for (unsigned i = 0; i < ARRAY_SIZE(bch2_metadata_versions); i++) + if (bch2_metadata_versions[i].version > v && + BCH_VERSION_MAJOR(bch2_metadata_versions[i].version) == + BCH_VERSION_MAJOR(v)) + v = bch2_metadata_versions[i].version; + + return v; +} + +u64 bch2_upgrade_recovery_passes(struct bch_fs *c, + unsigned old_version, + unsigned new_version) +{ + u64 ret = 0; + + for (const struct bch2_metadata_version *i = bch2_metadata_versions; + i < bch2_metadata_versions + ARRAY_SIZE(bch2_metadata_versions); + i++) + if (i->version > old_version && i->version <= new_version) { + if (i->recovery_passes & RECOVERY_PASS_ALL_FSCK) + ret |= bch2_fsck_recovery_passes(); + ret |= i->recovery_passes; + } + + return ret &= ~RECOVERY_PASS_ALL_FSCK; +} const char * const bch2_sb_fields[] = { #define x(name, nr) #name, @@ -35,7 +96,7 @@ const char * const bch2_sb_fields[] = { static int bch2_sb_field_validate(struct bch_sb *, struct bch_sb_field *, struct printbuf *); -struct bch_sb_field *bch2_sb_field_get(struct bch_sb *sb, +struct bch_sb_field *bch2_sb_field_get_id(struct bch_sb *sb, enum bch_sb_field_type type) { struct bch_sb_field *f; @@ -90,7 +151,7 @@ static struct bch_sb_field *__bch2_sb_field_resize(struct bch_sb_handle *sb, void bch2_sb_field_delete(struct bch_sb_handle *sb, enum bch_sb_field_type type) { - struct bch_sb_field *f = bch2_sb_field_get(sb->sb, type); + struct bch_sb_field *f = bch2_sb_field_get_id(sb->sb, type); if (f) __bch2_sb_field_resize(sb, f, 0); @@ -102,7 +163,8 @@ void bch2_free_super(struct bch_sb_handle *sb) { kfree(sb->bio); if (!IS_ERR_OR_NULL(sb->bdev)) - blkdev_put(sb->bdev, sb->mode); + blkdev_put(sb->bdev, sb->holder); + kfree(sb->holder); kfree(sb->sb); memset(sb, 0, sizeof(*sb)); @@ -123,7 +185,7 @@ int bch2_sb_realloc(struct bch_sb_handle *sb, unsigned u64s) if (sb->sb && sb->buffer_size >= new_buffer_size) return 0; - if (sb->have_layout) { + if (sb->sb && sb->have_layout) { u64 max_bytes = 512 << sb->sb->layout.sb_max_size_bits; if (new_bytes > max_bytes) { @@ -137,14 +199,20 @@ int bch2_sb_realloc(struct bch_sb_handle *sb, unsigned u64s) return 0; if (dynamic_fault("bcachefs:add:super_realloc")) - return -ENOMEM; + return -BCH_ERR_ENOMEM_sb_realloc_injected; + + new_sb = krealloc(sb->sb, new_buffer_size, GFP_NOFS|__GFP_ZERO); + if (!new_sb) + return -BCH_ERR_ENOMEM_sb_buf_realloc; + + sb->sb = new_sb; if (sb->have_bio) { - unsigned nr_bvecs = DIV_ROUND_UP(new_buffer_size, PAGE_SIZE); + unsigned nr_bvecs = buf_pages(sb->sb, new_buffer_size); bio = bio_kmalloc(nr_bvecs, GFP_KERNEL); if (!bio) - return -ENOMEM; + return -BCH_ERR_ENOMEM_sb_bio_realloc; bio_init(bio, NULL, bio->bi_inline_vecs, nr_bvecs, 0); @@ -152,21 +220,16 @@ int bch2_sb_realloc(struct bch_sb_handle *sb, unsigned u64s) sb->bio = bio; } - new_sb = krealloc(sb->sb, new_buffer_size, GFP_NOFS|__GFP_ZERO); - if (!new_sb) - return -ENOMEM; - - sb->sb = new_sb; sb->buffer_size = new_buffer_size; return 0; } -struct bch_sb_field *bch2_sb_field_resize(struct bch_sb_handle *sb, +struct bch_sb_field *bch2_sb_field_resize_id(struct bch_sb_handle *sb, enum bch_sb_field_type type, unsigned u64s) { - struct bch_sb_field *f = bch2_sb_field_get(sb->sb, type); + struct bch_sb_field *f = bch2_sb_field_get_id(sb->sb, type); ssize_t old_u64s = f ? le32_to_cpu(f->u64s) : 0; ssize_t d = -old_u64s + u64s; @@ -183,16 +246,16 @@ struct bch_sb_field *bch2_sb_field_resize(struct bch_sb_handle *sb, /* XXX: we're not checking that offline device have enough space */ for_each_online_member(ca, c, i) { - struct bch_sb_handle *sb = &ca->disk_sb; + struct bch_sb_handle *dev_sb = &ca->disk_sb; - if (bch2_sb_realloc(sb, le32_to_cpu(sb->sb->u64s) + d)) { + if (bch2_sb_realloc(dev_sb, le32_to_cpu(dev_sb->sb->u64s) + d)) { percpu_ref_put(&ca->ref); return NULL; } } } - f = bch2_sb_field_get(sb->sb, type); + f = bch2_sb_field_get_id(sb->sb, type); f = __bch2_sb_field_resize(sb, f, u64s); if (f) f->type = cpu_to_le32(type); @@ -201,18 +264,15 @@ struct bch_sb_field *bch2_sb_field_resize(struct bch_sb_handle *sb, /* Superblock validate: */ -static inline void __bch2_sb_layout_size_assert(void) -{ - BUILD_BUG_ON(sizeof(struct bch_sb_layout) != 512); -} - static int validate_sb_layout(struct bch_sb_layout *layout, struct printbuf *out) { u64 offset, prev_offset, max_sectors; unsigned i; - if (uuid_le_cmp(layout->magic, BCACHE_MAGIC) && - uuid_le_cmp(layout->magic, BCHFS_MAGIC)) { + BUILD_BUG_ON(sizeof(struct bch_sb_layout) != 512); + + if (!uuid_equal(&layout->magic, &BCACHE_MAGIC) && + !uuid_equal(&layout->magic, &BCHFS_MAGIC)) { prt_printf(out, "Not a bcachefs superblock layout"); return -BCH_ERR_invalid_sb_layout; } @@ -252,40 +312,58 @@ static int validate_sb_layout(struct bch_sb_layout *layout, struct printbuf *out return 0; } -static int bch2_sb_validate(struct bch_sb_handle *disk_sb, struct printbuf *out, - int rw) +static int bch2_sb_compatible(struct bch_sb *sb, struct printbuf *out) { - struct bch_sb *sb = disk_sb->sb; - struct bch_sb_field *f; - struct bch_sb_field_members *mi; - enum bch_opt_id opt_id; - u32 version, version_min; - u16 block_size; - int ret; - - version = le16_to_cpu(sb->version); - version_min = version >= bcachefs_metadata_version_bkey_renumber - ? le16_to_cpu(sb->version_min) - : version; - - if (version >= bcachefs_metadata_version_max) { - prt_printf(out, "Unsupported superblock version %u (min %u, max %u)", - version, bcachefs_metadata_version_min, bcachefs_metadata_version_max); + u16 version = le16_to_cpu(sb->version); + u16 version_min = le16_to_cpu(sb->version_min); + + if (!bch2_version_compatible(version)) { + prt_str(out, "Unsupported superblock version "); + bch2_version_to_text(out, version); + prt_str(out, " (min "); + bch2_version_to_text(out, bcachefs_metadata_version_min); + prt_str(out, ", max "); + bch2_version_to_text(out, bcachefs_metadata_version_current); + prt_str(out, ")"); return -BCH_ERR_invalid_sb_version; } - if (version_min < bcachefs_metadata_version_min) { - prt_printf(out, "Unsupported superblock version %u (min %u, max %u)", - version_min, bcachefs_metadata_version_min, bcachefs_metadata_version_max); + if (!bch2_version_compatible(version_min)) { + prt_str(out, "Unsupported superblock version_min "); + bch2_version_to_text(out, version_min); + prt_str(out, " (min "); + bch2_version_to_text(out, bcachefs_metadata_version_min); + prt_str(out, ", max "); + bch2_version_to_text(out, bcachefs_metadata_version_current); + prt_str(out, ")"); return -BCH_ERR_invalid_sb_version; } if (version_min > version) { - prt_printf(out, "Bad minimum version %u, greater than version field %u", - version_min, version); + prt_str(out, "Bad minimum version "); + bch2_version_to_text(out, version_min); + prt_str(out, ", greater than version field "); + bch2_version_to_text(out, version); return -BCH_ERR_invalid_sb_version; } + return 0; +} + +static int bch2_sb_validate(struct bch_sb_handle *disk_sb, struct printbuf *out, + int rw) +{ + struct bch_sb *sb = disk_sb->sb; + struct bch_sb_field *f; + struct bch_sb_field_members_v1 *mi; + enum bch_opt_id opt_id; + u16 block_size; + int ret; + + ret = bch2_sb_compatible(sb, out); + if (ret) + return ret; + if (sb->features[1] || (le64_to_cpu(sb->features[0]) & (~0ULL << BCH_FEATURE_NR))) { prt_printf(out, "Filesystem has incompatible features"); @@ -300,13 +378,13 @@ static int bch2_sb_validate(struct bch_sb_handle *disk_sb, struct printbuf *out, return -BCH_ERR_invalid_sb_block_size; } - if (bch2_is_zero(sb->user_uuid.b, sizeof(uuid_le))) { + if (bch2_is_zero(sb->user_uuid.b, sizeof(sb->user_uuid))) { prt_printf(out, "Bad user UUID (got zeroes)"); return -BCH_ERR_invalid_sb_uuid; } - if (bch2_is_zero(sb->uuid.b, sizeof(uuid_le))) { - prt_printf(out, "Bad intenal UUID (got zeroes)"); + if (bch2_is_zero(sb->uuid.b, sizeof(sb->uuid))) { + prt_printf(out, "Bad internal UUID (got zeroes)"); return -BCH_ERR_invalid_sb_uuid; } @@ -333,13 +411,16 @@ static int bch2_sb_validate(struct bch_sb_handle *disk_sb, struct printbuf *out, if (rw == READ) { /* * Been seeing a bug where these are getting inexplicably - * zeroed, so we'r now validating them, but we have to be + * zeroed, so we're now validating them, but we have to be * careful not to preven people's filesystems from mounting: */ if (!BCH_SB_JOURNAL_FLUSH_DELAY(sb)) SET_BCH_SB_JOURNAL_FLUSH_DELAY(sb, 1000); if (!BCH_SB_JOURNAL_RECLAIM_DELAY(sb)) SET_BCH_SB_JOURNAL_RECLAIM_DELAY(sb, 1000); + + if (!BCH_SB_VERSION_UPGRADE_COMPLETE(sb)) + SET_BCH_SB_VERSION_UPGRADE_COMPLETE(sb, le16_to_cpu(sb->version)); } for (opt_id = 0; opt_id < bch2_opts_nr; opt_id++) { @@ -377,7 +458,7 @@ static int bch2_sb_validate(struct bch_sb_handle *disk_sb, struct printbuf *out, } /* members must be validated first: */ - mi = bch2_sb_get_members(sb); + mi = bch2_sb_field_get(sb, members_v1); if (!mi) { prt_printf(out, "Invalid superblock: member info area missing"); return -BCH_ERR_invalid_sb_members_missing; @@ -388,7 +469,7 @@ static int bch2_sb_validate(struct bch_sb_handle *disk_sb, struct printbuf *out, return ret; vstruct_for_each(sb, f) { - if (le32_to_cpu(f->type) == BCH_SB_FIELD_members) + if (le32_to_cpu(f->type) == BCH_SB_FIELD_members_v1) continue; ret = bch2_sb_field_validate(sb, f, out); @@ -404,7 +485,6 @@ static int bch2_sb_validate(struct bch_sb_handle *disk_sb, struct printbuf *out, static void bch2_sb_update(struct bch_fs *c) { struct bch_sb *src = c->disk_sb.sb; - struct bch_sb_field_members *mi = bch2_sb_get_members(src); struct bch_dev *ca; unsigned i; @@ -414,6 +494,7 @@ static void bch2_sb_update(struct bch_fs *c) c->sb.user_uuid = src->user_uuid; c->sb.version = le16_to_cpu(src->version); c->sb.version_min = le16_to_cpu(src->version_min); + c->sb.version_upgrade_complete = BCH_SB_VERSION_UPGRADE_COMPLETE(src); c->sb.nr_devices = src->nr_devices; c->sb.clean = BCH_SB_CLEAN(src); c->sb.encryption_type = BCH_SB_ENCRYPTION_TYPE(src); @@ -429,11 +510,13 @@ static void bch2_sb_update(struct bch_fs *c) c->sb.features = le64_to_cpu(src->features[0]); c->sb.compat = le64_to_cpu(src->compat[0]); - for_each_member_device(ca, c, i) - ca->mi = bch2_mi_to_cpu(mi->members + i); + for_each_member_device(ca, c, i) { + struct bch_member m = bch2_sb_member_get(src, i); + ca->mi = bch2_mi_to_cpu(&m); + } } -static void __copy_super(struct bch_sb_handle *dst_handle, struct bch_sb *src) +static int __copy_super(struct bch_sb_handle *dst_handle, struct bch_sb *src) { struct bch_sb_field *src_f, *dst_f; struct bch_sb *dst = dst_handle->sb; @@ -458,42 +541,47 @@ static void __copy_super(struct bch_sb_handle *dst_handle, struct bch_sb *src) memcpy(dst->compat, src->compat, sizeof(dst->compat)); for (i = 0; i < BCH_SB_FIELD_NR; i++) { + int d; + if ((1U << i) & BCH_SINGLE_DEVICE_SB_FIELDS) continue; - src_f = bch2_sb_field_get(src, i); - dst_f = bch2_sb_field_get(dst, i); + src_f = bch2_sb_field_get_id(src, i); + dst_f = bch2_sb_field_get_id(dst, i); + + d = (src_f ? le32_to_cpu(src_f->u64s) : 0) - + (dst_f ? le32_to_cpu(dst_f->u64s) : 0); + if (d > 0) { + int ret = bch2_sb_realloc(dst_handle, + le32_to_cpu(dst_handle->sb->u64s) + d); + + if (ret) + return ret; + + dst = dst_handle->sb; + dst_f = bch2_sb_field_get_id(dst, i); + } + dst_f = __bch2_sb_field_resize(dst_handle, dst_f, src_f ? le32_to_cpu(src_f->u64s) : 0); if (src_f) memcpy(dst_f, src_f, vstruct_bytes(src_f)); } + + return 0; } int bch2_sb_to_fs(struct bch_fs *c, struct bch_sb *src) { - struct bch_sb_field_journal *journal_buckets = - bch2_sb_get_journal(src); - unsigned journal_u64s = journal_buckets - ? le32_to_cpu(journal_buckets->field.u64s) - : 0; int ret; lockdep_assert_held(&c->sb_lock); - ret = bch2_sb_realloc(&c->disk_sb, - le32_to_cpu(src->u64s) - journal_u64s); - if (ret) - return ret; - - __copy_super(&c->disk_sb, src); - - ret = bch2_sb_replicas_to_cpu_replicas(c); - if (ret) - return ret; - - ret = bch2_sb_disk_groups_to_cpu(c); + ret = bch2_sb_realloc(&c->disk_sb, 0) ?: + __copy_super(&c->disk_sb, src) ?: + bch2_sb_replicas_to_cpu_replicas(c) ?: + bch2_sb_disk_groups_to_cpu(c); if (ret) return ret; @@ -503,21 +591,7 @@ int bch2_sb_to_fs(struct bch_fs *c, struct bch_sb *src) int bch2_sb_from_fs(struct bch_fs *c, struct bch_dev *ca) { - struct bch_sb *src = c->disk_sb.sb, *dst = ca->disk_sb.sb; - struct bch_sb_field_journal *journal_buckets = - bch2_sb_get_journal(dst); - unsigned journal_u64s = journal_buckets - ? le32_to_cpu(journal_buckets->field.u64s) - : 0; - unsigned u64s = le32_to_cpu(src->u64s) + journal_u64s; - int ret; - - ret = bch2_sb_realloc(&ca->disk_sb, u64s); - if (ret) - return ret; - - __copy_super(&ca->disk_sb, src); - return 0; + return __copy_super(&ca->disk_sb, c->disk_sb.sb); } /* read superblock: */ @@ -525,7 +599,6 @@ int bch2_sb_from_fs(struct bch_fs *c, struct bch_dev *ca) static int read_one_super(struct bch_sb_handle *sb, u64 offset, struct printbuf *err) { struct bch_csum csum; - u32 version, version_min; size_t bytes; int ret; reread: @@ -539,28 +612,15 @@ reread: return ret; } - if (uuid_le_cmp(sb->sb->magic, BCACHE_MAGIC) && - uuid_le_cmp(sb->sb->magic, BCHFS_MAGIC)) { + if (!uuid_equal(&sb->sb->magic, &BCACHE_MAGIC) && + !uuid_equal(&sb->sb->magic, &BCHFS_MAGIC)) { prt_printf(err, "Not a bcachefs superblock"); return -BCH_ERR_invalid_sb_magic; } - version = le16_to_cpu(sb->sb->version); - version_min = version >= bcachefs_metadata_version_bkey_renumber - ? le16_to_cpu(sb->sb->version_min) - : version; - - if (version >= bcachefs_metadata_version_max) { - prt_printf(err, "Unsupported superblock version %u (min %u, max %u)", - version, bcachefs_metadata_version_min, bcachefs_metadata_version_max); - return -BCH_ERR_invalid_sb_version; - } - - if (version_min < bcachefs_metadata_version_min) { - prt_printf(err, "Unsupported superblock version %u (min %u, max %u)", - version_min, bcachefs_metadata_version_min, bcachefs_metadata_version_max); - return -BCH_ERR_invalid_sb_version; - } + ret = bch2_sb_compatible(sb->sb, err); + if (ret) + return ret; bytes = vstruct_bytes(sb->sb); @@ -571,8 +631,9 @@ reread: } if (bytes > sb->buffer_size) { - if (bch2_sb_realloc(sb, le32_to_cpu(sb->sb->u64s))) - return -ENOMEM; + ret = bch2_sb_realloc(sb, le32_to_cpu(sb->sb->u64s)); + if (ret) + return ret; goto reread; } @@ -603,26 +664,34 @@ int bch2_read_super(const char *path, struct bch_opts *opts, struct printbuf err = PRINTBUF; __le64 *i; int ret; - - pr_verbose_init(*opts, ""); - +#ifndef __KERNEL__ +retry: +#endif memset(sb, 0, sizeof(*sb)); - sb->mode = FMODE_READ; + sb->mode = BLK_OPEN_READ; sb->have_bio = true; + sb->holder = kmalloc(1, GFP_KERNEL); + if (!sb->holder) + return -ENOMEM; + +#ifndef __KERNEL__ + if (opt_get(*opts, direct_io) == false) + sb->mode |= BLK_OPEN_BUFFERED; +#endif if (!opt_get(*opts, noexcl)) - sb->mode |= FMODE_EXCL; + sb->mode |= BLK_OPEN_EXCL; if (!opt_get(*opts, nochanges)) - sb->mode |= FMODE_WRITE; + sb->mode |= BLK_OPEN_WRITE; - sb->bdev = blkdev_get_by_path(path, sb->mode, sb); + sb->bdev = blkdev_get_by_path(path, sb->mode, sb->holder, &bch2_sb_handle_bdev_ops); if (IS_ERR(sb->bdev) && PTR_ERR(sb->bdev) == -EACCES && opt_get(*opts, read_only)) { - sb->mode &= ~FMODE_WRITE; + sb->mode &= ~BLK_OPEN_WRITE; - sb->bdev = blkdev_get_by_path(path, sb->mode, sb); + sb->bdev = blkdev_get_by_path(path, sb->mode, sb->holder, &bch2_sb_handle_bdev_ops); if (!IS_ERR(sb->bdev)) opt_set(*opts, nochanges, true); } @@ -694,7 +763,13 @@ int bch2_read_super(const char *path, struct bch_opts *opts, got_super: if (le16_to_cpu(sb->sb->block_size) << 9 < - bdev_logical_block_size(sb->bdev)) { + bdev_logical_block_size(sb->bdev) && + opt_get(*opts, direct_io)) { +#ifndef __KERNEL__ + opt_set(*opts, direct_io, false); + bch2_free_super(sb); + goto retry; +#endif prt_printf(&err, "block size (%u) smaller than device block size (%u)", le16_to_cpu(sb->sb->block_size) << 9, bdev_logical_block_size(sb->bdev)); @@ -712,7 +787,6 @@ got_super: goto err_no_print; } out: - pr_verbose_init(*opts, "ret %i", ret); printbuf_exit(&err); return ret; err: @@ -804,10 +878,9 @@ int bch2_write_super(struct bch_fs *c) closure_init_stack(cl); memset(&sb_written, 0, sizeof(sb_written)); - if (c->opts.version_upgrade) { - c->disk_sb.sb->magic = BCHFS_MAGIC; - c->disk_sb.sb->layout.magic = BCHFS_MAGIC; - } + /* Make sure we're using the new magic numbers: */ + c->disk_sb.sb->magic = BCHFS_MAGIC; + c->disk_sb.sb->layout.magic = BCHFS_MAGIC; le64_add_cpu(&c->disk_sb.sb->seq, 1); @@ -819,6 +892,7 @@ int bch2_write_super(struct bch_fs *c) SET_BCH_SB_BIG_ENDIAN(c->disk_sb.sb, CPU_BIG_ENDIAN); bch2_sb_counters_from_cpu(c); + bch_members_cpy_v2_v1(&c->disk_sb); for_each_online_member(ca, c, i) bch2_sb_from_fs(c, ca); @@ -941,458 +1015,32 @@ void __bch2_check_set_feature(struct bch_fs *c, unsigned feat) mutex_unlock(&c->sb_lock); } -/* BCH_SB_FIELD_members: */ - -static int bch2_sb_members_validate(struct bch_sb *sb, - struct bch_sb_field *f, - struct printbuf *err) -{ - struct bch_sb_field_members *mi = field_to_type(f, members); - unsigned i; - - if ((void *) (mi->members + sb->nr_devices) > - vstruct_end(&mi->field)) { - prt_printf(err, "too many devices for section size"); - return -BCH_ERR_invalid_sb_members; - } - - for (i = 0; i < sb->nr_devices; i++) { - struct bch_member *m = mi->members + i; - - if (!bch2_member_exists(m)) - continue; - - if (le64_to_cpu(m->nbuckets) > LONG_MAX) { - prt_printf(err, "device %u: too many buckets (got %llu, max %lu)", - i, le64_to_cpu(m->nbuckets), LONG_MAX); - return -BCH_ERR_invalid_sb_members; - } - - if (le64_to_cpu(m->nbuckets) - - le16_to_cpu(m->first_bucket) < BCH_MIN_NR_NBUCKETS) { - prt_printf(err, "device %u: not enough buckets (got %llu, max %u)", - i, le64_to_cpu(m->nbuckets), BCH_MIN_NR_NBUCKETS); - return -BCH_ERR_invalid_sb_members; - } - - if (le16_to_cpu(m->bucket_size) < - le16_to_cpu(sb->block_size)) { - prt_printf(err, "device %u: bucket size %u smaller than block size %u", - i, le16_to_cpu(m->bucket_size), le16_to_cpu(sb->block_size)); - return -BCH_ERR_invalid_sb_members; - } - - if (le16_to_cpu(m->bucket_size) < - BCH_SB_BTREE_NODE_SIZE(sb)) { - prt_printf(err, "device %u: bucket size %u smaller than btree node size %llu", - i, le16_to_cpu(m->bucket_size), BCH_SB_BTREE_NODE_SIZE(sb)); - return -BCH_ERR_invalid_sb_members; - } - } - - return 0; -} - -static void bch2_sb_members_to_text(struct printbuf *out, struct bch_sb *sb, - struct bch_sb_field *f) -{ - struct bch_sb_field_members *mi = field_to_type(f, members); - struct bch_sb_field_disk_groups *gi = bch2_sb_get_disk_groups(sb); - unsigned i; - - for (i = 0; i < sb->nr_devices; i++) { - struct bch_member *m = mi->members + i; - unsigned data_have = bch2_sb_dev_has_data(sb, i); - u64 bucket_size = le16_to_cpu(m->bucket_size); - u64 device_size = le64_to_cpu(m->nbuckets) * bucket_size; - - if (!bch2_member_exists(m)) - continue; - - prt_printf(out, "Device:"); - prt_tab(out); - prt_printf(out, "%u", i); - prt_newline(out); - - printbuf_indent_add(out, 2); - - prt_printf(out, "UUID:"); - prt_tab(out); - pr_uuid(out, m->uuid.b); - prt_newline(out); - - prt_printf(out, "Size:"); - prt_tab(out); - prt_units_u64(out, device_size << 9); - prt_newline(out); - - prt_printf(out, "Bucket size:"); - prt_tab(out); - prt_units_u64(out, bucket_size << 9); - prt_newline(out); - - prt_printf(out, "First bucket:"); - prt_tab(out); - prt_printf(out, "%u", le16_to_cpu(m->first_bucket)); - prt_newline(out); - - prt_printf(out, "Buckets:"); - prt_tab(out); - prt_printf(out, "%llu", le64_to_cpu(m->nbuckets)); - prt_newline(out); - - prt_printf(out, "Last mount:"); - prt_tab(out); - if (m->last_mount) - pr_time(out, le64_to_cpu(m->last_mount)); - else - prt_printf(out, "(never)"); - prt_newline(out); - - prt_printf(out, "State:"); - prt_tab(out); - prt_printf(out, "%s", - BCH_MEMBER_STATE(m) < BCH_MEMBER_STATE_NR - ? bch2_member_states[BCH_MEMBER_STATE(m)] - : "unknown"); - prt_newline(out); - - prt_printf(out, "Label:"); - prt_tab(out); - if (BCH_MEMBER_GROUP(m)) { - unsigned idx = BCH_MEMBER_GROUP(m) - 1; - - if (idx < disk_groups_nr(gi)) - prt_printf(out, "%s (%u)", - gi->entries[idx].label, idx); - else - prt_printf(out, "(bad disk labels section)"); - } else { - prt_printf(out, "(none)"); - } - prt_newline(out); - - prt_printf(out, "Data allowed:"); - prt_tab(out); - if (BCH_MEMBER_DATA_ALLOWED(m)) - prt_bitflags(out, bch2_data_types, BCH_MEMBER_DATA_ALLOWED(m)); - else - prt_printf(out, "(none)"); - prt_newline(out); - - prt_printf(out, "Has data:"); - prt_tab(out); - if (data_have) - prt_bitflags(out, bch2_data_types, data_have); - else - prt_printf(out, "(none)"); - prt_newline(out); - - prt_printf(out, "Discard:"); - prt_tab(out); - prt_printf(out, "%llu", BCH_MEMBER_DISCARD(m)); - prt_newline(out); - - prt_printf(out, "Freespace initialized:"); - prt_tab(out); - prt_printf(out, "%llu", BCH_MEMBER_FREESPACE_INITIALIZED(m)); - prt_newline(out); - - printbuf_indent_sub(out, 2); - } -} - -static const struct bch_sb_field_ops bch_sb_field_ops_members = { - .validate = bch2_sb_members_validate, - .to_text = bch2_sb_members_to_text, -}; - -/* BCH_SB_FIELD_crypt: */ - -static int bch2_sb_crypt_validate(struct bch_sb *sb, - struct bch_sb_field *f, - struct printbuf *err) -{ - struct bch_sb_field_crypt *crypt = field_to_type(f, crypt); - - if (vstruct_bytes(&crypt->field) < sizeof(*crypt)) { - prt_printf(err, "wrong size (got %zu should be %zu)", - vstruct_bytes(&crypt->field), sizeof(*crypt)); - return -BCH_ERR_invalid_sb_crypt; - } - - if (BCH_CRYPT_KDF_TYPE(crypt)) { - prt_printf(err, "bad kdf type %llu", BCH_CRYPT_KDF_TYPE(crypt)); - return -BCH_ERR_invalid_sb_crypt; - } - - return 0; -} - -static void bch2_sb_crypt_to_text(struct printbuf *out, struct bch_sb *sb, - struct bch_sb_field *f) -{ - struct bch_sb_field_crypt *crypt = field_to_type(f, crypt); - - prt_printf(out, "KFD: %llu", BCH_CRYPT_KDF_TYPE(crypt)); - prt_newline(out); - prt_printf(out, "scrypt n: %llu", BCH_KDF_SCRYPT_N(crypt)); - prt_newline(out); - prt_printf(out, "scrypt r: %llu", BCH_KDF_SCRYPT_R(crypt)); - prt_newline(out); - prt_printf(out, "scrypt p: %llu", BCH_KDF_SCRYPT_P(crypt)); - prt_newline(out); -} - -static const struct bch_sb_field_ops bch_sb_field_ops_crypt = { - .validate = bch2_sb_crypt_validate, - .to_text = bch2_sb_crypt_to_text, -}; - -/* BCH_SB_FIELD_clean: */ - -int bch2_sb_clean_validate_late(struct bch_fs *c, struct bch_sb_field_clean *clean, int write) +/* Downgrade if superblock is at a higher version than currently supported: */ +void bch2_sb_maybe_downgrade(struct bch_fs *c) { - struct jset_entry *entry; - int ret; - - for (entry = clean->start; - entry < (struct jset_entry *) vstruct_end(&clean->field); - entry = vstruct_next(entry)) { - ret = bch2_journal_entry_validate(c, NULL, entry, - le16_to_cpu(c->disk_sb.sb->version), - BCH_SB_BIG_ENDIAN(c->disk_sb.sb), - write); - if (ret) - return ret; - } - - return 0; -} - -int bch2_fs_mark_dirty(struct bch_fs *c) -{ - int ret; + lockdep_assert_held(&c->sb_lock); /* - * Unconditionally write superblock, to verify it hasn't changed before - * we go rw: + * Downgrade, if superblock is at a higher version than currently + * supported: */ - - mutex_lock(&c->sb_lock); - SET_BCH_SB_CLEAN(c->disk_sb.sb, false); - c->disk_sb.sb->features[0] |= cpu_to_le64(BCH_SB_FEATURES_ALWAYS); + if (BCH_SB_VERSION_UPGRADE_COMPLETE(c->disk_sb.sb) > bcachefs_metadata_version_current) + SET_BCH_SB_VERSION_UPGRADE_COMPLETE(c->disk_sb.sb, bcachefs_metadata_version_current); + if (c->sb.version > bcachefs_metadata_version_current) + c->disk_sb.sb->version = cpu_to_le16(bcachefs_metadata_version_current); + if (c->sb.version_min > bcachefs_metadata_version_current) + c->disk_sb.sb->version_min = cpu_to_le16(bcachefs_metadata_version_current); c->disk_sb.sb->compat[0] &= cpu_to_le64((1ULL << BCH_COMPAT_NR) - 1); - ret = bch2_write_super(c); - mutex_unlock(&c->sb_lock); - - return ret; -} - -static struct jset_entry *jset_entry_init(struct jset_entry **end, size_t size) -{ - struct jset_entry *entry = *end; - unsigned u64s = DIV_ROUND_UP(size, sizeof(u64)); - - memset(entry, 0, u64s * sizeof(u64)); - /* - * The u64s field counts from the start of data, ignoring the shared - * fields. - */ - entry->u64s = cpu_to_le16(u64s - 1); - - *end = vstruct_next(*end); - return entry; } -void bch2_journal_super_entries_add_common(struct bch_fs *c, - struct jset_entry **end, - u64 journal_seq) +void bch2_sb_upgrade(struct bch_fs *c, unsigned new_version) { - struct bch_dev *ca; - unsigned i, dev; - - percpu_down_read(&c->mark_lock); - - if (!journal_seq) { - for (i = 0; i < ARRAY_SIZE(c->usage); i++) - bch2_fs_usage_acc_to_base(c, i); - } else { - bch2_fs_usage_acc_to_base(c, journal_seq & JOURNAL_BUF_MASK); - } - - { - struct jset_entry_usage *u = - container_of(jset_entry_init(end, sizeof(*u)), - struct jset_entry_usage, entry); - - u->entry.type = BCH_JSET_ENTRY_usage; - u->entry.btree_id = BCH_FS_USAGE_inodes; - u->v = cpu_to_le64(c->usage_base->nr_inodes); - } - - { - struct jset_entry_usage *u = - container_of(jset_entry_init(end, sizeof(*u)), - struct jset_entry_usage, entry); - - u->entry.type = BCH_JSET_ENTRY_usage; - u->entry.btree_id = BCH_FS_USAGE_key_version; - u->v = cpu_to_le64(atomic64_read(&c->key_version)); - } - - for (i = 0; i < BCH_REPLICAS_MAX; i++) { - struct jset_entry_usage *u = - container_of(jset_entry_init(end, sizeof(*u)), - struct jset_entry_usage, entry); - - u->entry.type = BCH_JSET_ENTRY_usage; - u->entry.btree_id = BCH_FS_USAGE_reserved; - u->entry.level = i; - u->v = cpu_to_le64(c->usage_base->persistent_reserved[i]); - } - - for (i = 0; i < c->replicas.nr; i++) { - struct bch_replicas_entry *e = - cpu_replicas_entry(&c->replicas, i); - struct jset_entry_data_usage *u = - container_of(jset_entry_init(end, sizeof(*u) + e->nr_devs), - struct jset_entry_data_usage, entry); - - u->entry.type = BCH_JSET_ENTRY_data_usage; - u->v = cpu_to_le64(c->usage_base->replicas[i]); - memcpy(&u->r, e, replicas_entry_bytes(e)); - } - - for_each_member_device(ca, c, dev) { - unsigned b = sizeof(struct jset_entry_dev_usage) + - sizeof(struct jset_entry_dev_usage_type) * BCH_DATA_NR; - struct jset_entry_dev_usage *u = - container_of(jset_entry_init(end, b), - struct jset_entry_dev_usage, entry); - - u->entry.type = BCH_JSET_ENTRY_dev_usage; - u->dev = cpu_to_le32(dev); - u->buckets_ec = cpu_to_le64(ca->usage_base->buckets_ec); - - for (i = 0; i < BCH_DATA_NR; i++) { - u->d[i].buckets = cpu_to_le64(ca->usage_base->d[i].buckets); - u->d[i].sectors = cpu_to_le64(ca->usage_base->d[i].sectors); - u->d[i].fragmented = cpu_to_le64(ca->usage_base->d[i].fragmented); - } - } - - percpu_up_read(&c->mark_lock); - - for (i = 0; i < 2; i++) { - struct jset_entry_clock *clock = - container_of(jset_entry_init(end, sizeof(*clock)), - struct jset_entry_clock, entry); - - clock->entry.type = BCH_JSET_ENTRY_clock; - clock->rw = i; - clock->time = cpu_to_le64(atomic64_read(&c->io_clock[i].now)); - } -} - -void bch2_fs_mark_clean(struct bch_fs *c) -{ - struct bch_sb_field_clean *sb_clean; - struct jset_entry *entry; - unsigned u64s; - int ret; - - mutex_lock(&c->sb_lock); - if (BCH_SB_CLEAN(c->disk_sb.sb)) - goto out; - - SET_BCH_SB_CLEAN(c->disk_sb.sb, true); - - c->disk_sb.sb->compat[0] |= cpu_to_le64(1ULL << BCH_COMPAT_alloc_info); - c->disk_sb.sb->compat[0] |= cpu_to_le64(1ULL << BCH_COMPAT_alloc_metadata); - c->disk_sb.sb->features[0] &= cpu_to_le64(~(1ULL << BCH_FEATURE_extents_above_btree_updates)); - c->disk_sb.sb->features[0] &= cpu_to_le64(~(1ULL << BCH_FEATURE_btree_updates_journalled)); - - u64s = sizeof(*sb_clean) / sizeof(u64) + c->journal.entry_u64s_reserved; - - sb_clean = bch2_sb_resize_clean(&c->disk_sb, u64s); - if (!sb_clean) { - bch_err(c, "error resizing superblock while setting filesystem clean"); - goto out; - } - - sb_clean->flags = 0; - sb_clean->journal_seq = cpu_to_le64(atomic64_read(&c->journal.seq)); - - /* Trying to catch outstanding bug: */ - BUG_ON(le64_to_cpu(sb_clean->journal_seq) > S64_MAX); - - entry = sb_clean->start; - bch2_journal_super_entries_add_common(c, &entry, 0); - entry = bch2_btree_roots_to_journal_entries(c, entry, entry); - BUG_ON((void *) entry > vstruct_end(&sb_clean->field)); - - memset(entry, 0, - vstruct_end(&sb_clean->field) - (void *) entry); - - /* - * this should be in the write path, and we should be validating every - * superblock section: - */ - ret = bch2_sb_clean_validate_late(c, sb_clean, WRITE); - if (ret) { - bch_err(c, "error writing marking filesystem clean: validate error"); - goto out; - } - - bch2_write_super(c); -out: - mutex_unlock(&c->sb_lock); -} - -static int bch2_sb_clean_validate(struct bch_sb *sb, - struct bch_sb_field *f, - struct printbuf *err) -{ - struct bch_sb_field_clean *clean = field_to_type(f, clean); - - if (vstruct_bytes(&clean->field) < sizeof(*clean)) { - prt_printf(err, "wrong size (got %zu should be %zu)", - vstruct_bytes(&clean->field), sizeof(*clean)); - return -BCH_ERR_invalid_sb_clean; - } - - return 0; -} - -static void bch2_sb_clean_to_text(struct printbuf *out, struct bch_sb *sb, - struct bch_sb_field *f) -{ - struct bch_sb_field_clean *clean = field_to_type(f, clean); - struct jset_entry *entry; - - prt_printf(out, "flags: %x", le32_to_cpu(clean->flags)); - prt_newline(out); - prt_printf(out, "journal_seq: %llu", le64_to_cpu(clean->journal_seq)); - prt_newline(out); - - for (entry = clean->start; - entry != vstruct_end(&clean->field); - entry = vstruct_next(entry)) { - if (entry->type == BCH_JSET_ENTRY_btree_keys && - !entry->u64s) - continue; + lockdep_assert_held(&c->sb_lock); - bch2_journal_entry_to_text(out, NULL, entry); - prt_newline(out); - } + c->disk_sb.sb->version = cpu_to_le16(new_version); + c->disk_sb.sb->features[0] |= cpu_to_le64(BCH_SB_FEATURES_ALL); } -static const struct bch_sb_field_ops bch_sb_field_ops_clean = { - .validate = bch2_sb_clean_validate, - .to_text = bch2_sb_clean_to_text, -}; - static const struct bch_sb_field_ops *bch2_sb_field_ops[] = { #define x(f, nr) \ [BCH_SB_FIELD_##f] = &bch_sb_field_ops_##f, @@ -1400,21 +1048,27 @@ static const struct bch_sb_field_ops *bch2_sb_field_ops[] = { #undef x }; +static const struct bch_sb_field_ops bch2_sb_field_null_ops; + +static const struct bch_sb_field_ops *bch2_sb_field_type_ops(unsigned type) +{ + return likely(type < ARRAY_SIZE(bch2_sb_field_ops)) + ? bch2_sb_field_ops[type] + : &bch2_sb_field_null_ops; +} + static int bch2_sb_field_validate(struct bch_sb *sb, struct bch_sb_field *f, struct printbuf *err) { unsigned type = le32_to_cpu(f->type); struct printbuf field_err = PRINTBUF; + const struct bch_sb_field_ops *ops = bch2_sb_field_type_ops(type); int ret; - if (type >= BCH_SB_FIELD_NR) - return 0; - - ret = bch2_sb_field_ops[type]->validate(sb, f, &field_err); + ret = ops->validate ? ops->validate(sb, f, &field_err) : 0; if (ret) { prt_printf(err, "Invalid superblock section %s: %s", - bch2_sb_fields[type], - field_err.buf); + bch2_sb_fields[type], field_err.buf); prt_newline(err); bch2_sb_field_to_text(err, sb, f); } @@ -1427,13 +1081,12 @@ void bch2_sb_field_to_text(struct printbuf *out, struct bch_sb *sb, struct bch_sb_field *f) { unsigned type = le32_to_cpu(f->type); - const struct bch_sb_field_ops *ops = type < BCH_SB_FIELD_NR - ? bch2_sb_field_ops[type] : NULL; + const struct bch_sb_field_ops *ops = bch2_sb_field_type_ops(type); if (!out->nr_tabstops) printbuf_tabstop_push(out, 32); - if (ops) + if (type < BCH_SB_FIELD_NR) prt_printf(out, "%s", bch2_sb_fields[type]); else prt_printf(out, "(unknown field %u)", type); @@ -1441,9 +1094,9 @@ void bch2_sb_field_to_text(struct printbuf *out, struct bch_sb *sb, prt_printf(out, " (size %zu):", vstruct_bytes(f)); prt_newline(out); - if (ops && ops->to_text) { + if (ops->to_text) { printbuf_indent_add(out, 2); - bch2_sb_field_ops[type]->to_text(out, sb, f); + ops->to_text(out, sb, f); printbuf_indent_sub(out, 2); } } @@ -1474,7 +1127,6 @@ void bch2_sb_layout_to_text(struct printbuf *out, struct bch_sb_layout *l) void bch2_sb_to_text(struct printbuf *out, struct bch_sb *sb, bool print_layout, unsigned fields) { - struct bch_sb_field_members *mi; struct bch_sb_field *f; u64 fields_have = 0; unsigned nr_devices = 0; @@ -1482,15 +1134,8 @@ void bch2_sb_to_text(struct printbuf *out, struct bch_sb *sb, if (!out->nr_tabstops) printbuf_tabstop_push(out, 44); - mi = bch2_sb_get_members(sb); - if (mi) { - struct bch_member *m; - - for (m = mi->members; - m < mi->members + sb->nr_devices; - m++) - nr_devices += bch2_member_exists(m); - } + for (int i = 0; i < sb->nr_devices; i++) + nr_devices += bch2_dev_exists(sb, i); prt_printf(out, "External UUID:"); prt_tab(out); @@ -1514,12 +1159,17 @@ void bch2_sb_to_text(struct printbuf *out, struct bch_sb *sb, prt_str(out, "Version:"); prt_tab(out); - prt_printf(out, "%s", bch2_metadata_versions[le16_to_cpu(sb->version)]); + bch2_version_to_text(out, le16_to_cpu(sb->version)); + prt_newline(out); + + prt_str(out, "Version upgrade complete:"); + prt_tab(out); + bch2_version_to_text(out, BCH_SB_VERSION_UPGRADE_COMPLETE(sb)); prt_newline(out); prt_printf(out, "Oldest version on disk:"); prt_tab(out); - prt_printf(out, "%s", bch2_metadata_versions[le16_to_cpu(sb->version_min)]); + bch2_version_to_text(out, le16_to_cpu(sb->version_min)); prt_newline(out); prt_printf(out, "Created:");