X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libbcachefs%2Fsuper-io.c;h=326faf253255bea91a6b6628d1f9066ce3965b3d;hb=f76d246fff1c3c5b091fb2b910067d893620dc5e;hp=d87edaadaa27ab10e139271b6f0227282035c328;hpb=1ffc9b4554b29316a51f6bf3374540b3c878b783;p=bcachefs-tools-debian diff --git a/libbcachefs/super-io.c b/libbcachefs/super-io.c index d87edaa..326faf2 100644 --- a/libbcachefs/super-io.c +++ b/libbcachefs/super-io.c @@ -1,25 +1,92 @@ // 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-errors.h" +#include "sb-members.h" #include "super-io.h" #include "super.h" +#include "trace.h" #include "vstructs.h" #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, BCH_SB_FIELDS() @@ -30,11 +97,9 @@ 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; - /* XXX: need locking around superblock to access optional fields */ vstruct_for_each(sb, f) @@ -85,7 +150,7 @@ 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); @@ -95,10 +160,11 @@ void bch2_sb_field_delete(struct bch_sb_handle *sb, void bch2_free_super(struct bch_sb_handle *sb) { - if (sb->bio) - bio_put(sb->bio); + 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_name); kfree(sb->sb); memset(sb, 0, sizeof(*sb)); @@ -119,15 +185,13 @@ 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) { - char buf[BDEVNAME_SIZE]; - - pr_err("%s: superblock too big: want %zu but have %llu", - bdevname(sb->bdev, buf), new_bytes, max_bytes); - return -ENOSPC; + pr_err("%pg: superblock too big: want %zu but have %llu", + sb->bdev, new_bytes, max_bytes); + return -BCH_ERR_ENOSPC_sb; } } @@ -135,34 +199,37 @@ 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) { - bio = bio_kmalloc(GFP_KERNEL, - 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); - if (sb->bio) - bio_put(sb->bio); + kfree(sb->bio); 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; @@ -171,24 +238,22 @@ struct bch_sb_field *bch2_sb_field_resize(struct bch_sb_handle *sb, if (sb->fs_sb) { struct bch_fs *c = container_of(sb, struct bch_fs, disk_sb); - struct bch_dev *ca; - unsigned i; lockdep_assert_held(&c->sb_lock); /* 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; + for_each_online_member(c, ca) { + 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); @@ -197,35 +262,33 @@ 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)) { - pr_buf(out, "Not a bcachefs superblock layout"); - return -EINVAL; + 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; } if (layout->layout_type != 0) { - pr_buf(out, "Invalid superblock layout type %u", + prt_printf(out, "Invalid superblock layout type %u", layout->layout_type); - return -EINVAL; + return -BCH_ERR_invalid_sb_layout_type; } if (!layout->nr_superblocks) { - pr_buf(out, "Invalid superblock layout: no superblocks"); - return -EINVAL; + prt_printf(out, "Invalid superblock layout: no superblocks"); + return -BCH_ERR_invalid_sb_layout_nr_superblocks; } if (layout->nr_superblocks > ARRAY_SIZE(layout->sb_offset)) { - pr_buf(out, "Invalid superblock layout: too many superblocks"); - return -EINVAL; + prt_printf(out, "Invalid superblock layout: too many superblocks"); + return -BCH_ERR_invalid_sb_layout_nr_superblocks; } max_sectors = 1 << layout->sb_max_size_bits; @@ -236,10 +299,10 @@ static int validate_sb_layout(struct bch_sb_layout *layout, struct printbuf *out offset = le64_to_cpu(layout->sb_offset[i]); if (offset < prev_offset + max_sectors) { - pr_buf(out, "Invalid superblock layout: superblocks overlap\n" + prt_printf(out, "Invalid superblock layout: superblocks overlap\n" " (sb %u ends at %llu next starts at %llu", i - 1, prev_offset + max_sectors, offset); - return -EINVAL; + return -BCH_ERR_invalid_sb_layout_superblocks_overlap; } prev_offset = offset; } @@ -247,80 +310,129 @@ 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) +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; - u32 version, version_min; - u16 block_size; - int ret; - - version = le16_to_cpu(sb->version); - version_min = version >= bcachefs_metadata_version_new_versioning - ? le16_to_cpu(sb->version_min) - : version; - - if (version >= bcachefs_metadata_version_max) { - pr_buf(out, "Unsupported superblock version %u (min %u, max %u)", - version, bcachefs_metadata_version_min, bcachefs_metadata_version_max); - return -EINVAL; - } - - if (version_min < bcachefs_metadata_version_min) { - pr_buf(out, "Unsupported superblock version %u (min %u, max %u)", - version_min, bcachefs_metadata_version_min, bcachefs_metadata_version_max); - return -EINVAL; + 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 (!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) { - pr_buf(out, "Bad minimum version %u, greater than version field %u", - version_min, version); - return -EINVAL; + 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_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))) { - pr_buf(out, "Filesystem has incompatible features"); - return -EINVAL; + prt_printf(out, "Filesystem has incompatible features"); + return -BCH_ERR_invalid_sb_features; } block_size = le16_to_cpu(sb->block_size); if (block_size > PAGE_SECTORS) { - pr_buf(out, "Block size too big (got %u, max %u)", + prt_printf(out, "Block size too big (got %u, max %u)", block_size, PAGE_SECTORS); - return -EINVAL; + return -BCH_ERR_invalid_sb_block_size; } - if (bch2_is_zero(sb->user_uuid.b, sizeof(uuid_le))) { - pr_buf(out, "Bad user UUID (got zeroes)"); - return -EINVAL; + 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))) { - pr_buf(out, "Bad intenal UUID (got zeroes)"); - return -EINVAL; + if (bch2_is_zero(sb->uuid.b, sizeof(sb->uuid))) { + prt_printf(out, "Bad internal UUID (got zeroes)"); + return -BCH_ERR_invalid_sb_uuid; } if (!sb->nr_devices || sb->nr_devices > BCH_SB_MEMBERS_MAX) { - pr_buf(out, "Bad number of member devices %u (max %u)", + prt_printf(out, "Bad number of member devices %u (max %u)", sb->nr_devices, BCH_SB_MEMBERS_MAX); - return -EINVAL; + return -BCH_ERR_invalid_sb_too_many_members; } if (sb->dev_idx >= sb->nr_devices) { - pr_buf(out, "Bad dev_idx (got %u, nr_devices %u)", + prt_printf(out, "Bad dev_idx (got %u, nr_devices %u)", sb->dev_idx, sb->nr_devices); - return -EINVAL; + return -BCH_ERR_invalid_sb_dev_idx; } if (!sb->time_precision || le32_to_cpu(sb->time_precision) > NSEC_PER_SEC) { - pr_buf(out, "Invalid time precision: %u (min 1, max %lu)", + prt_printf(out, "Invalid time precision: %u (min 1, max %lu)", le32_to_cpu(sb->time_precision), NSEC_PER_SEC); - return -EINVAL; + return -BCH_ERR_invalid_sb_time_precision; + } + + if (rw == READ) { + /* + * Been seeing a bug where these are getting inexplicably + * 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++) { + const struct bch_option *opt = bch2_opt_table + opt_id; + + if (opt->get_sb != BCH2_NO_SB_OPT) { + u64 v = bch2_opt_from_sb(sb, opt_id); + + prt_printf(out, "Invalid option "); + ret = bch2_opt_validate(opt, v, out); + if (ret) + return ret; + + printbuf_reset(out); + } } /* validate layout */ @@ -330,23 +442,23 @@ static int bch2_sb_validate(struct bch_sb_handle *disk_sb, struct printbuf *out) vstruct_for_each(sb, f) { if (!f->u64s) { - pr_buf(out, "Invalid superblock: optional with size 0 (type %u)", + prt_printf(out, "Invalid superblock: optional field with size 0 (type %u)", le32_to_cpu(f->type)); - return -EINVAL; + return -BCH_ERR_invalid_sb_field_size; } if (vstruct_next(f) > vstruct_last(sb)) { - pr_buf(out, "Invalid superblock: optional field extends past end of superblock (type %u)", + prt_printf(out, "Invalid superblock: optional field extends past end of superblock (type %u)", le32_to_cpu(f->type)); - return -EINVAL; + return -BCH_ERR_invalid_sb_field_size; } } /* members must be validated first: */ - mi = bch2_sb_get_members(sb); + mi = bch2_sb_field_get(sb, members_v1); if (!mi) { - pr_buf(out, "Invalid superblock: member info area missing"); - return -EINVAL; + prt_printf(out, "Invalid superblock: member info area missing"); + return -BCH_ERR_invalid_sb_members_missing; } ret = bch2_sb_field_validate(sb, &mi->field, out); @@ -354,7 +466,7 @@ 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); @@ -370,9 +482,6 @@ 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; lockdep_assert_held(&c->sb_lock); @@ -380,6 +489,7 @@ 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); @@ -395,11 +505,13 @@ 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(c, ca) { + struct bch_member m = bch2_sb_member_get(src, ca->dev_idx); + 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; @@ -424,42 +536,47 @@ 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++) { - if (i == BCH_SB_FIELD_journal) + 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; @@ -469,21 +586,7 @@ 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: */ @@ -491,61 +594,47 @@ 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: - bio_reset(sb->bio); - bio_set_dev(sb->bio, sb->bdev); + bio_reset(sb->bio, sb->bdev, REQ_OP_READ|REQ_SYNC|REQ_META); sb->bio->bi_iter.bi_sector = offset; - bio_set_op_attrs(sb->bio, REQ_OP_READ, REQ_SYNC|REQ_META); bch2_bio_map(sb->bio, sb->sb, sb->buffer_size); ret = submit_bio_wait(sb->bio); if (ret) { - pr_buf(err, "IO error: %i", ret); + prt_printf(err, "IO error: %i", ret); return ret; } - if (uuid_le_cmp(sb->sb->magic, BCACHE_MAGIC)) { - pr_buf(err, "Not a bcachefs superblock"); - return -EINVAL; - } - - version = le16_to_cpu(sb->sb->version); - version_min = version >= bcachefs_metadata_version_new_versioning - ? le16_to_cpu(sb->sb->version_min) - : version; - - if (version >= bcachefs_metadata_version_max) { - pr_buf(err, "Unsupported superblock version %u (min %u, max %u)", - version, bcachefs_metadata_version_min, bcachefs_metadata_version_max); - return -EINVAL; + 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; } - if (version_min < bcachefs_metadata_version_min) { - pr_buf(err, "Unsupported superblock version %u (min %u, max %u)", - version_min, bcachefs_metadata_version_min, bcachefs_metadata_version_max); - return -EINVAL; - } + ret = bch2_sb_compatible(sb->sb, err); + if (ret) + return ret; bytes = vstruct_bytes(sb->sb); if (bytes > 512 << sb->sb->layout.sb_max_size_bits) { - pr_buf(err, "Invalid superblock: too big (got %zu bytes, layout max %lu)", + prt_printf(err, "Invalid superblock: too big (got %zu bytes, layout max %lu)", bytes, 512UL << sb->sb->layout.sb_max_size_bits); - return -EINVAL; + return -BCH_ERR_invalid_sb_too_big; } 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; } if (BCH_SB_CSUM_TYPE(sb->sb) >= BCH_CSUM_NR) { - pr_buf(err, "unknown checksum type %llu", BCH_SB_CSUM_TYPE(sb->sb)); - return -EINVAL; + prt_printf(err, "unknown checksum type %llu", BCH_SB_CSUM_TYPE(sb->sb)); + return -BCH_ERR_invalid_sb_csum_type; } /* XXX: verify MACs */ @@ -553,8 +642,8 @@ reread: null_nonce(), sb->sb); if (bch2_crc_cmp(csum, sb->sb->csum)) { - pr_buf(err, "bad checksum"); - return -EINVAL; + prt_printf(err, "bad checksum"); + return -BCH_ERR_invalid_sb_csum; } sb->seq = le64_to_cpu(sb->sb->seq); @@ -562,34 +651,47 @@ reread: return 0; } -int bch2_read_super(const char *path, struct bch_opts *opts, - struct bch_sb_handle *sb) +static int __bch2_read_super(const char *path, struct bch_opts *opts, + struct bch_sb_handle *sb, bool ignore_notbchfs_msg) { u64 offset = opt_get(*opts, sb); struct bch_sb_layout layout; struct printbuf err = PRINTBUF; + struct printbuf err2 = 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; + + sb->sb_name = kstrdup(path, GFP_KERNEL); + if (!sb->sb_name) + 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); } @@ -601,12 +703,12 @@ int bch2_read_super(const char *path, struct bch_opts *opts, ret = bch2_sb_realloc(sb, 0); if (ret) { - pr_buf(&err, "error allocating memory for superblock"); + prt_printf(&err, "error allocating memory for superblock"); goto err; } if (bch2_fs_init_fault("read_super")) { - pr_buf(&err, "dynamic fault"); + prt_printf(&err, "dynamic fault"); ret = -EFAULT; goto err; } @@ -618,18 +720,22 @@ int bch2_read_super(const char *path, struct bch_opts *opts, if (opt_defined(*opts, sb)) goto err; - printk(KERN_ERR "bcachefs (%s): error reading default superblock: %s", + prt_printf(&err2, "bcachefs (%s): error reading default superblock: %s\n", path, err.buf); + if (ret == -BCH_ERR_invalid_sb_magic && ignore_notbchfs_msg) + printk(KERN_INFO "%s", err2.buf); + else + printk(KERN_ERR "%s", err2.buf); + + printbuf_exit(&err2); printbuf_reset(&err); /* * Error reading primary superblock - read location of backup * superblocks: */ - bio_reset(sb->bio); - bio_set_dev(sb->bio, sb->bdev); + bio_reset(sb->bio, sb->bdev, REQ_OP_READ|REQ_SYNC|REQ_META); sb->bio->bi_iter.bi_sector = BCH_SB_LAYOUT_SECTOR; - bio_set_op_attrs(sb->bio, REQ_OP_READ, REQ_SYNC|REQ_META); /* * use sb buffer to read layout, since sb buffer is page aligned but * layout won't be: @@ -638,7 +744,7 @@ int bch2_read_super(const char *path, struct bch_opts *opts, ret = submit_bio_wait(sb->bio); if (ret) { - pr_buf(&err, "IO error: %i", ret); + prt_printf(&err, "IO error: %i", ret); goto err; } @@ -663,35 +769,54 @@ 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)) { - pr_buf(&err, "block size (%u) smaller than device block size (%u)", + 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)); - ret = -EINVAL; + ret = -BCH_ERR_block_size_too_small; goto err; } ret = 0; sb->have_layout = true; - ret = bch2_sb_validate(sb, &err); + ret = bch2_sb_validate(sb, &err, READ); if (ret) { - printk(KERN_ERR "bcachefs (%s): error validating superblock: %s", + printk(KERN_ERR "bcachefs (%s): error validating superblock: %s\n", path, err.buf); goto err_no_print; } out: - pr_verbose_init(*opts, "ret %i", ret); printbuf_exit(&err); return ret; err: - printk(KERN_ERR "bcachefs (%s): error reading superblock: %s", + printk(KERN_ERR "bcachefs (%s): error reading superblock: %s\n", path, err.buf); err_no_print: bch2_free_super(sb); goto out; } +int bch2_read_super(const char *path, struct bch_opts *opts, + struct bch_sb_handle *sb) +{ + return __bch2_read_super(path, opts, sb, false); +} + +/* provide a silenced version for mount.bcachefs */ + +int bch2_read_super_silent(const char *path, struct bch_opts *opts, + struct bch_sb_handle *sb) +{ + return __bch2_read_super(path, opts, sb, true); +} + /* write superblock: */ static void write_super_endio(struct bio *bio) @@ -700,7 +825,12 @@ static void write_super_endio(struct bio *bio) /* XXX: return errors directly */ - if (bch2_dev_io_err_on(bio->bi_status, ca, "superblock write error: %s", + if (bch2_dev_io_err_on(bio->bi_status, ca, + bio_data_dir(bio) + ? BCH_MEMBER_ERROR_write + : BCH_MEMBER_ERROR_read, + "superblock %s error: %s", + bio_data_dir(bio) ? "write" : "read", bch2_blk_status_to_str(bio->bi_status))) ca->sb_write_error = 1; @@ -713,12 +843,10 @@ static void read_back_super(struct bch_fs *c, struct bch_dev *ca) struct bch_sb *sb = ca->disk_sb.sb; struct bio *bio = ca->disk_sb.bio; - bio_reset(bio); - bio_set_dev(bio, ca->disk_sb.bdev); + bio_reset(bio, ca->disk_sb.bdev, REQ_OP_READ|REQ_SYNC|REQ_META); bio->bi_iter.bi_sector = le64_to_cpu(sb->layout.sb_offset[0]); bio->bi_end_io = write_super_endio; bio->bi_private = ca; - bio_set_op_attrs(bio, REQ_OP_READ, REQ_SYNC|REQ_META); bch2_bio_map(bio, ca->sb_read_scratch, PAGE_SIZE); this_cpu_add(ca->io_done->sectors[READ][BCH_DATA_sb], @@ -739,12 +867,10 @@ static void write_one_super(struct bch_fs *c, struct bch_dev *ca, unsigned idx) sb->csum = csum_vstruct(c, BCH_SB_CSUM_TYPE(sb), null_nonce(), sb); - bio_reset(bio); - bio_set_dev(bio, ca->disk_sb.bdev); + bio_reset(bio, ca->disk_sb.bdev, REQ_OP_WRITE|REQ_SYNC|REQ_META); bio->bi_iter.bi_sector = le64_to_cpu(sb->offset); bio->bi_end_io = write_super_endio; bio->bi_private = ca; - bio_set_op_attrs(bio, REQ_OP_WRITE, REQ_SYNC|REQ_META); bch2_bio_map(bio, sb, roundup((size_t) vstruct_bytes(sb), bdev_logical_block_size(ca->disk_sb.bdev))); @@ -759,14 +885,15 @@ static void write_one_super(struct bch_fs *c, struct bch_dev *ca, unsigned idx) int bch2_write_super(struct bch_fs *c) { struct closure *cl = &c->sb_write; - struct bch_dev *ca; struct printbuf err = PRINTBUF; - unsigned i, sb = 0, nr_wrote; + unsigned sb = 0, nr_wrote; struct bch_devs_mask sb_written; bool wrote, can_mount_without_written, can_mount_with_written; unsigned degraded_flags = BCH_FORCE_IF_DEGRADED; int ret = 0; + trace_and_count(c, write_super, c, _RET_IP_); + if (c->opts.very_degraded) degraded_flags |= BCH_FORCE_IF_LOST; @@ -775,22 +902,31 @@ int bch2_write_super(struct bch_fs *c) closure_init_stack(cl); memset(&sb_written, 0, sizeof(sb_written)); + /* 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); - if (test_bit(BCH_FS_ERROR, &c->flags)) + if (test_bit(BCH_FS_error, &c->flags)) SET_BCH_SB_HAS_ERRORS(c->disk_sb.sb, 1); - if (test_bit(BCH_FS_TOPOLOGY_ERROR, &c->flags)) + if (test_bit(BCH_FS_topology_error, &c->flags)) SET_BCH_SB_HAS_TOPOLOGY_ERRORS(c->disk_sb.sb, 1); SET_BCH_SB_BIG_ENDIAN(c->disk_sb.sb, CPU_BIG_ENDIAN); - for_each_online_member(ca, c, i) + bch2_sb_counters_from_cpu(c); + bch2_sb_members_from_cpu(c); + bch2_sb_members_cpy_v2_v1(&c->disk_sb); + bch2_sb_errors_from_cpu(c); + + for_each_online_member(c, ca) bch2_sb_from_fs(c, ca); - for_each_online_member(ca, c, i) { + for_each_online_member(c, ca) { printbuf_reset(&err); - ret = bch2_sb_validate(&ca->disk_sb, &err); + ret = bch2_sb_validate(&ca->disk_sb, &err, WRITE); if (ret) { bch2_fs_inconsistent(c, "sb invalid before write: %s", err.buf); percpu_ref_put(&ca->io_ref); @@ -801,16 +937,23 @@ int bch2_write_super(struct bch_fs *c) if (c->opts.nochanges) goto out; - for_each_online_member(ca, c, i) { + /* + * Defer writing the superblock until filesystem initialization is + * complete - don't write out a partly initialized superblock: + */ + if (!BCH_SB_INITIALIZED(c->disk_sb.sb)) + goto out; + + for_each_online_member(c, ca) { __set_bit(ca->dev_idx, sb_written.d); ca->sb_write_error = 0; } - for_each_online_member(ca, c, i) + for_each_online_member(c, ca) read_back_super(c, ca); closure_sync(cl); - for_each_online_member(ca, c, i) { + for_each_online_member(c, ca) { if (ca->sb_write_error) continue; @@ -820,7 +963,7 @@ int bch2_write_super(struct bch_fs *c) le64_to_cpu(ca->sb_read_scratch->seq), ca->disk_sb.seq); percpu_ref_put(&ca->io_ref); - ret = -EROFS; + ret = -BCH_ERR_erofs_sb_err; goto out; } @@ -830,14 +973,14 @@ int bch2_write_super(struct bch_fs *c) le64_to_cpu(ca->sb_read_scratch->seq), ca->disk_sb.seq); percpu_ref_put(&ca->io_ref); - ret = -EROFS; + ret = -BCH_ERR_erofs_sb_err; goto out; } } do { wrote = false; - for_each_online_member(ca, c, i) + for_each_online_member(c, ca) if (!ca->sb_write_error && sb < ca->disk_sb.sb->layout.nr_superblocks) { write_one_super(c, ca, sb); @@ -847,7 +990,7 @@ int bch2_write_super(struct bch_fs *c) sb++; } while (wrote); - for_each_online_member(ca, c, i) { + for_each_online_member(c, ca) { if (ca->sb_write_error) __clear_bit(ca->dev_idx, sb_written.d); else @@ -859,7 +1002,7 @@ int bch2_write_super(struct bch_fs *c) can_mount_with_written = bch2_have_enough_devs(c, sb_written, degraded_flags, false); - for (i = 0; i < ARRAY_SIZE(sb_written.d); i++) + for (unsigned i = 0; i < ARRAY_SIZE(sb_written.d); i++) sb_written.d[i] = ~sb_written.d[i]; can_mount_without_written = @@ -898,534 +1041,32 @@ void __bch2_check_set_feature(struct bch_fs *c, unsigned feat) mutex_unlock(&c->sb_lock); } -/* BCH_SB_FIELD_journal: */ - -static int u64_cmp(const void *_l, const void *_r) -{ - u64 l = *((const u64 *) _l), r = *((const u64 *) _r); - - return l < r ? -1 : l > r ? 1 : 0; -} - -static int bch2_sb_journal_validate(struct bch_sb *sb, - struct bch_sb_field *f, - struct printbuf *err) -{ - struct bch_sb_field_journal *journal = field_to_type(f, journal); - struct bch_member *m = bch2_sb_get_members(sb)->members + sb->dev_idx; - int ret = -EINVAL; - unsigned nr; - unsigned i; - u64 *b; - - nr = bch2_nr_journal_buckets(journal); - if (!nr) - return 0; - - b = kmalloc_array(sizeof(u64), nr, GFP_KERNEL); - if (!b) - return -ENOMEM; - - for (i = 0; i < nr; i++) - b[i] = le64_to_cpu(journal->buckets[i]); - - sort(b, nr, sizeof(u64), u64_cmp, NULL); - - if (!b[0]) { - pr_buf(err, "journal bucket at sector 0"); - goto err; - } - - if (b[0] < le16_to_cpu(m->first_bucket)) { - pr_buf(err, "journal bucket %llu before first bucket %u", - b[0], le16_to_cpu(m->first_bucket)); - goto err; - } - - if (b[nr - 1] >= le64_to_cpu(m->nbuckets)) { - pr_buf(err, "journal bucket %llu past end of device (nbuckets %llu)", - b[nr - 1], le64_to_cpu(m->nbuckets)); - goto err; - } - - for (i = 0; i + 1 < nr; i++) - if (b[i] == b[i + 1]) { - pr_buf(err, "duplicate journal buckets %llu", b[i]); - goto err; - } - - ret = 0; -err: - kfree(b); - return ret; -} - -static void bch2_sb_journal_to_text(struct printbuf *out, struct bch_sb *sb, - struct bch_sb_field *f) -{ - struct bch_sb_field_journal *journal = field_to_type(f, journal); - unsigned i, nr = bch2_nr_journal_buckets(journal); - - pr_buf(out, "Buckets: "); - for (i = 0; i < nr; i++) - pr_buf(out, " %llu", le64_to_cpu(journal->buckets[i])); - pr_newline(out); -} - -static const struct bch_sb_field_ops bch_sb_field_ops_journal = { - .validate = bch2_sb_journal_validate, - .to_text = bch2_sb_journal_to_text, -}; - -/* BCH_SB_FIELD_members: */ - -static int bch2_sb_members_validate(struct bch_sb *sb, - struct bch_sb_field *f, - struct printbuf *err) +/* Downgrade if superblock is at a higher version than currently supported: */ +void bch2_sb_maybe_downgrade(struct bch_fs *c) { - struct bch_sb_field_members *mi = field_to_type(f, members); - unsigned i; - - if ((void *) (mi->members + sb->nr_devices) > - vstruct_end(&mi->field)) { - pr_buf(err, "too many devices for section size"); - return -EINVAL; - } - - 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) { - pr_buf(err, "device %u: too many buckets (got %llu, max %lu)", - i, le64_to_cpu(m->nbuckets), LONG_MAX); - return -EINVAL; - } - - if (le64_to_cpu(m->nbuckets) - - le16_to_cpu(m->first_bucket) < BCH_MIN_NR_NBUCKETS) { - pr_buf(err, "device %u: not enough buckets (got %llu, max %u)", - i, le64_to_cpu(m->nbuckets), BCH_MIN_NR_NBUCKETS); - return -EINVAL; - } - - if (le16_to_cpu(m->bucket_size) < - le16_to_cpu(sb->block_size)) { - pr_buf(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 -EINVAL; - } - - if (le16_to_cpu(m->bucket_size) < - BCH_SB_BTREE_NODE_SIZE(sb)) { - pr_buf(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 -EINVAL; - } - } - - 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; - - pr_buf(out, "Device:"); - pr_tab(out); - pr_buf(out, "%u", i); - pr_newline(out); - - pr_indent_push(out, 2); - - pr_buf(out, "UUID:"); - pr_tab(out); - pr_uuid(out, m->uuid.b); - pr_newline(out); - - pr_buf(out, "Size:"); - pr_tab(out); - pr_units(out, device_size, device_size << 9); - pr_newline(out); - - pr_buf(out, "Bucket size:"); - pr_tab(out); - pr_units(out, bucket_size, bucket_size << 9); - pr_newline(out); - - pr_buf(out, "First bucket:"); - pr_tab(out); - pr_buf(out, "%u", le16_to_cpu(m->first_bucket)); - pr_newline(out); - - pr_buf(out, "Buckets:"); - pr_tab(out); - pr_buf(out, "%llu", le64_to_cpu(m->nbuckets)); - pr_newline(out); - - pr_buf(out, "Last mount:"); - pr_tab(out); - if (m->last_mount) - pr_time(out, le64_to_cpu(m->last_mount)); - else - pr_buf(out, "(never)"); - pr_newline(out); - - pr_buf(out, "State:"); - pr_tab(out); - pr_buf(out, "%s", - BCH_MEMBER_STATE(m) < BCH_MEMBER_STATE_NR - ? bch2_member_states[BCH_MEMBER_STATE(m)] - : "unknown"); - pr_newline(out); - - pr_buf(out, "Group:"); - pr_tab(out); - if (BCH_MEMBER_GROUP(m)) { - unsigned idx = BCH_MEMBER_GROUP(m) - 1; - - if (idx < disk_groups_nr(gi)) - pr_buf(out, "%s (%u)", - gi->entries[idx].label, idx); - else - pr_buf(out, "(bad disk labels section)"); - } else { - pr_buf(out, "(none)"); - } - pr_newline(out); - - pr_buf(out, "Data allowed:"); - pr_tab(out); - if (BCH_MEMBER_DATA_ALLOWED(m)) - bch2_flags_to_text(out, bch2_data_types, - BCH_MEMBER_DATA_ALLOWED(m)); - else - pr_buf(out, "(none)"); - pr_newline(out); - - pr_buf(out, "Has data:"); - pr_tab(out); - if (data_have) - bch2_flags_to_text(out, bch2_data_types, data_have); - else - pr_buf(out, "(none)"); - pr_newline(out); - - pr_buf(out, "Discard:"); - pr_tab(out); - pr_buf(out, "%llu", BCH_MEMBER_DISCARD(m)); - pr_newline(out); - - pr_indent_pop(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)) { - pr_buf(err, "wrong size (got %zu should be %zu)", - vstruct_bytes(&crypt->field), sizeof(*crypt)); - return -EINVAL; - } - - if (BCH_CRYPT_KDF_TYPE(crypt)) { - pr_buf(err, "bad kdf type %llu", BCH_CRYPT_KDF_TYPE(crypt)); - return -EINVAL; - } - - 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); - - pr_buf(out, "KFD: %llu", BCH_CRYPT_KDF_TYPE(crypt)); - pr_newline(out); - pr_buf(out, "scrypt n: %llu", BCH_KDF_SCRYPT_N(crypt)); - pr_newline(out); - pr_buf(out, "scrypt r: %llu", BCH_KDF_SCRYPT_R(crypt)); - pr_newline(out); - pr_buf(out, "scrypt p: %llu", BCH_KDF_SCRYPT_P(crypt)); - pr_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) -{ - 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, "superblock", 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) +void bch2_sb_upgrade(struct bch_fs *c, unsigned new_version) { - 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) -{ - 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); - u->buckets_unavailable = cpu_to_le64(ca->usage_base->buckets_unavailable); - - 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)) { - pr_buf(err, "wrong size (got %zu should be %zu)", - vstruct_bytes(&clean->field), sizeof(*clean)); - return -EINVAL; - } - - 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; - - pr_buf(out, "flags: %x", le32_to_cpu(clean->flags)); - pr_newline(out); - pr_buf(out, "journal_seq: %llu", le64_to_cpu(clean->journal_seq)); - pr_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); - pr_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, @@ -1433,22 +1074,28 @@ 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) { - pr_buf(err, "Invalid superblock section %s: %s", - bch2_sb_fields[type], - field_err.buf); - pr_newline(err); + prt_printf(err, "Invalid superblock section %s: %s", + bch2_sb_fields[type], field_err.buf); + prt_newline(err); bch2_sb_field_to_text(err, sb, f); } @@ -1460,24 +1107,23 @@ 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->tabstops[0]) - out->tabstops[0] = 32; + if (!out->nr_tabstops) + printbuf_tabstop_push(out, 32); - if (ops) - pr_buf(out, "%s", bch2_sb_fields[type]); + if (type < BCH_SB_FIELD_NR) + prt_printf(out, "%s", bch2_sb_fields[type]); else - pr_buf(out, "(unknown field %u)", type); + prt_printf(out, "(unknown field %u)", type); - pr_buf(out, " (size %zu):", vstruct_bytes(f)); - pr_newline(out); + prt_printf(out, " (size %zu):", vstruct_bytes(f)); + prt_newline(out); - if (ops && ops->to_text) { - pr_indent_push(out, 2); - bch2_sb_field_ops[type]->to_text(out, sb, f); - pr_indent_pop(out, 2); + if (ops->to_text) { + printbuf_indent_add(out, 2); + ops->to_text(out, sb, f); + printbuf_indent_sub(out, 2); } } @@ -1485,129 +1131,121 @@ void bch2_sb_layout_to_text(struct printbuf *out, struct bch_sb_layout *l) { unsigned i; - pr_buf(out, "Type: %u", l->layout_type); - pr_newline(out); + prt_printf(out, "Type: %u", l->layout_type); + prt_newline(out); - pr_buf(out, "Superblock max size: "); - pr_units(out, - 1 << l->sb_max_size_bits, - 512 << l->sb_max_size_bits); - pr_newline(out); + prt_str(out, "Superblock max size: "); + prt_units_u64(out, 512 << l->sb_max_size_bits); + prt_newline(out); - pr_buf(out, "Nr superblocks: %u", l->nr_superblocks); - pr_newline(out); + prt_printf(out, "Nr superblocks: %u", l->nr_superblocks); + prt_newline(out); - pr_buf(out, "Offsets: "); + prt_str(out, "Offsets: "); for (i = 0; i < l->nr_superblocks; i++) { if (i) - pr_buf(out, ", "); - pr_buf(out, "%llu", le64_to_cpu(l->sb_offset[i])); + prt_str(out, ", "); + prt_printf(out, "%llu", le64_to_cpu(l->sb_offset[i])); } - pr_newline(out); + prt_newline(out); } 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; - if (!out->tabstops[0]) - out->tabstops[0] = 32; + 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); - pr_buf(out, "External UUID:"); - pr_tab(out); + prt_printf(out, "External UUID:"); + prt_tab(out); pr_uuid(out, sb->user_uuid.b); - pr_newline(out); + prt_newline(out); - pr_buf(out, "Internal UUID:"); - pr_tab(out); + prt_printf(out, "Internal UUID:"); + prt_tab(out); pr_uuid(out, sb->uuid.b); - pr_newline(out); - - pr_buf(out, "Device index:"); - pr_tab(out); - pr_buf(out, "%u", sb->dev_idx); - pr_newline(out); - - pr_buf(out, "Label:"); - pr_tab(out); - pr_buf(out, "%.*s", (int) sizeof(sb->label), sb->label); - pr_newline(out); - - pr_buf(out, "Version:"); - pr_tab(out); - pr_buf(out, "%u", le16_to_cpu(sb->version)); - pr_newline(out); - - pr_buf(out, "Oldest version on disk:"); - pr_tab(out); - pr_buf(out, "%u", le16_to_cpu(sb->version_min)); - pr_newline(out); - - pr_buf(out, "Created:"); - pr_tab(out); + prt_newline(out); + + prt_str(out, "Device index:"); + prt_tab(out); + prt_printf(out, "%u", sb->dev_idx); + prt_newline(out); + + prt_str(out, "Label:"); + prt_tab(out); + prt_printf(out, "%.*s", (int) sizeof(sb->label), sb->label); + prt_newline(out); + + prt_str(out, "Version:"); + prt_tab(out); + 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); + bch2_version_to_text(out, le16_to_cpu(sb->version_min)); + prt_newline(out); + + prt_printf(out, "Created:"); + prt_tab(out); if (sb->time_base_lo) - pr_time(out, div_u64(le64_to_cpu(sb->time_base_lo), NSEC_PER_SEC)); + bch2_prt_datetime(out, div_u64(le64_to_cpu(sb->time_base_lo), NSEC_PER_SEC)); else - pr_buf(out, "(not set)"); - pr_newline(out); - - pr_buf(out, "Sequence number:"); - pr_tab(out); - pr_buf(out, "%llu", le64_to_cpu(sb->seq)); - pr_newline(out); - - pr_buf(out, "Superblock size:"); - pr_tab(out); - pr_buf(out, "%zu", vstruct_bytes(sb)); - pr_newline(out); - - pr_buf(out, "Clean:"); - pr_tab(out); - pr_buf(out, "%llu", BCH_SB_CLEAN(sb)); - pr_newline(out); - - pr_buf(out, "Devices:"); - pr_tab(out); - pr_buf(out, "%u", nr_devices); - pr_newline(out); - - pr_buf(out, "Sections:"); + prt_printf(out, "(not set)"); + prt_newline(out); + + prt_printf(out, "Sequence number:"); + prt_tab(out); + prt_printf(out, "%llu", le64_to_cpu(sb->seq)); + prt_newline(out); + + prt_printf(out, "Superblock size:"); + prt_tab(out); + prt_printf(out, "%zu", vstruct_bytes(sb)); + prt_newline(out); + + prt_printf(out, "Clean:"); + prt_tab(out); + prt_printf(out, "%llu", BCH_SB_CLEAN(sb)); + prt_newline(out); + + prt_printf(out, "Devices:"); + prt_tab(out); + prt_printf(out, "%u", nr_devices); + prt_newline(out); + + prt_printf(out, "Sections:"); vstruct_for_each(sb, f) fields_have |= 1 << le32_to_cpu(f->type); - pr_tab(out); - bch2_flags_to_text(out, bch2_sb_fields, fields_have); - pr_newline(out); - - pr_buf(out, "Features:"); - pr_tab(out); - bch2_flags_to_text(out, bch2_sb_features, - le64_to_cpu(sb->features[0])); - pr_newline(out); - - pr_buf(out, "Compat features:"); - pr_tab(out); - bch2_flags_to_text(out, bch2_sb_compat, - le64_to_cpu(sb->compat[0])); - pr_newline(out); - - pr_newline(out); - pr_buf(out, "Options:"); - pr_newline(out); - pr_indent_push(out, 2); + prt_tab(out); + prt_bitflags(out, bch2_sb_fields, fields_have); + prt_newline(out); + + prt_printf(out, "Features:"); + prt_tab(out); + prt_bitflags(out, bch2_sb_features, le64_to_cpu(sb->features[0])); + prt_newline(out); + + prt_printf(out, "Compat features:"); + prt_tab(out); + prt_bitflags(out, bch2_sb_compat, le64_to_cpu(sb->compat[0])); + prt_newline(out); + + prt_newline(out); + prt_printf(out, "Options:"); + prt_newline(out); + printbuf_indent_add(out, 2); { enum bch_opt_id id; @@ -1617,28 +1255,29 @@ void bch2_sb_to_text(struct printbuf *out, struct bch_sb *sb, if (opt->get_sb != BCH2_NO_SB_OPT) { u64 v = bch2_opt_from_sb(sb, id); - pr_buf(out, "%s:", opt->attr.name); - pr_tab(out); - bch2_opt_to_text(out, NULL, opt, v, OPT_HUMAN_READABLE|OPT_SHOW_FULL_LIST); - pr_newline(out); + prt_printf(out, "%s:", opt->attr.name); + prt_tab(out); + bch2_opt_to_text(out, NULL, sb, opt, v, + OPT_HUMAN_READABLE|OPT_SHOW_FULL_LIST); + prt_newline(out); } } } - pr_indent_pop(out, 2); + printbuf_indent_sub(out, 2); if (print_layout) { - pr_newline(out); - pr_buf(out, "layout:"); - pr_newline(out); - pr_indent_push(out, 2); + prt_newline(out); + prt_printf(out, "layout:"); + prt_newline(out); + printbuf_indent_add(out, 2); bch2_sb_layout_to_text(out, &sb->layout); - pr_indent_pop(out, 2); + printbuf_indent_sub(out, 2); } vstruct_for_each(sb, f) if (fields & (1 << le32_to_cpu(f->type))) { - pr_newline(out); + prt_newline(out); bch2_sb_field_to_text(out, sb, f); } }